Dyna Edition 188 - December of 2014

DYNA Journal of the Facultad de Minas, Universidad Nacional de Colombia - Medellin Campus

DYNA 81 (188), December, 2014 - ISSN 0012-7353 Tarifa Postal Reducida No. 2014-287 4-72 La Red Postal de Colombia, Vence 31 de Dic. 2014.

FACULTAD DE MINAS

DYNA

http://dyna.medellin.unal.edu.co/

DYNA is an international journal published by the Facultad de Minas, Universidad Nacional de Colombia, Medellín Campus since 1933. DYNA publishes peer-reviewed scientific articles covering all aspects of engineering. Our objective is the dissemination of original, useful and relevant research presenting new knowledge about theoretical or practical aspects of methodologies and methods used in engineering or leading to improvements in professional practices. All conclusions presented in the articles must be based on the current state-of-the-art and supported by a rigorous analysis and a balanced appraisal. The journal publishes scientific and technological research articles, review articles and case studies.

DYNA publishes articles in the following areas: Organizational Engineering Civil Engineering Materials and Mines Engineering

Geosciences and the Environment Systems and Informatics Chemistry and Petroleum

Mechatronics Bio-engineering Other areas related to engineering

Publication Information

Indexing and Databases

DYNA (ISSN 0012-73533, printed; 2346-2183, online) is published by the Facultad de Minas, Universidad Nacional de Colombia, with a bimonthly periodicity (February, April, June, August, October, and December). Circulation License Resolution 000584 de 1976 from the Ministry of the Government.

DYNA is admitted in:

Contact information Web page: E-mail: Mail address: Facultad de Minas

http://dyna.unalmed.edu.co dyna@unal.edu.co Revista DYNA Universidad Nacional de Colombia Medellín Campus Carrera 80 No. 65-223 Bloque M9 - Of.:107 Telephone: (574) 4255068 Fax: (574) 4255343 Medellín - Colombia © Copyright 2014. Universidad Nacional de Colombia The complete or partial reproduction of texts with educational ends is permitted, granted that the source is duly cited. Unless indicated otherwise.

Notice All statements, methods, instructions and ideas are only responsibility of the authors and not necessarily represent the view of the Universidad Nacional de Colombia. The publisher does not accept responsibility for any injury and/or damage for the use of the content of this journal.

The concepts and opinions expressed in the articles are the exclusive responsibility of the authors.

The National System of Indexation and Homologation of Specialized Journals CT+I-PUBLINDEX, Category A1 Science Citation Index Expanded Journal Citation Reports - JCR Science Direct SCOPUS Chemical Abstract - CAS Scientific Electronic Library on Line - SciELO GEOREF PERIÓDICA Data Base Latindex Actualidad Iberoaméricana RedALyC - Scientific Information System Directory of Open Acces Journals - DOAJ PASCAL CAPES UN Digital Library - SINAB CAPES

Publisher’s Office Juan David Velásquez Henao, Director Mónica del Pilar Rada T., Editorial Coordinator Catalina Cardona A., Editorial Assistant Amilkar Álvarez C., Diagrammer Byron Llano V., Editorial Assistant Institutional Exchange Request DYNA may be requested as an institutional exchange through the Andrew Mark Bailey, English Style Editor Alberto Gutiérrez, IT Assistant e-mail canjebib_med@unal.edu.co or to the postal address: Biblioteca Central “Efe Gómez” Universidad Nacional de Colombia, Sede Medellín Calle 59A No 63-20 Teléfono: (57+4) 430 97 86 Medellín - Colombia

Reduced Postal Fee Tarifa Postal Reducida # 2014-287 4-72. La Red Postal de Colombia, expires Dec. 31st, 2014

SEDE MEDELLÍN

SEDE MEDELLÍN

COUNCIL OF THE FACULTAD DE MINAS

JOURNAL EDITORIAL BOARD

Dean John Willian Branch Bedoya, PhD

Editor-in-Chief Juan David Velásquez Henao, PhD Universidad Nacional de Colombia, Colombia

Vice-Dean Pedro Nel Benjumea Hernández, PhD Vice-Dean of Research and Extension Verónica Botero Fernández, PhD Director of University Services Carlos Alberto Graciano, PhD Academic Secretary Carlos Alberto Zarate Yepes, PhD Representative of the Curricular Area Directors Gaspar Monsalve Mejía, PhD Elkin Rodríguez Velásquez, PhD Representative of the Basic Units of AcademicAdministrative Management Germán Alberto Sierra Gallego, PhD

Editors George Barbastathis, PhD Massachusetts Institute of Technology, USA Tim A. Osswald, PhD University of Wisconsin, USA Juan De Pablo, PhD University of Wisconsin, USA Hans Christian Öttinger, PhD Swiss Federal Institute of Technology (ETH), Switzerland Patrick D. Anderson, PhD Eindhoven University of Technology, the Netherlands Igor Emri, PhD Associate Professor, University of Ljubljana, Slovenia

Representative of the Basic Units of AcademicAdministrative Management Juan David Velásquez Henao, PhD

Dietmar Drummer, PhD Institute of Polymer Technology University ErlangenNürnberg, Germany

Professor Representative Jaime Ignacio Vélez Upegui, PhD

Ting-Chung Poon, PhD Virginia Polytechnic Institute and State University, USA

Delegate of the University Council Pedro Ignacio Torres Trujillo, PhD

Pierre Boulanger, PhD University of Alberta, Canadá

Undergraduate Student Representative Rubén David Montoya Pérez

Jordi Payá Bernabeu, Ph.D. Instituto de Ciencia y Tecnología del Hormigón (ICITECH) Universitat Politècnica de València, España

FACULTY EDITORIAL BOARD

Javier Belzunce Varela, Ph.D. Universidad de Oviedo, España

Dean John Willian Branch Bedoya, PhD Vice-Dean of Research and Extension Santiago Arango Aramburo, PhD Members Hernán Darío Álvarez Zapata, PhD Oscar Jaime Restrepo Baena, PhD Juan David Velásquez Henao, PhD Jaime Aguirre Cardona, PhD Mónica del Pilar Rada Tobón MSc

Luis Gonzaga Santos Sobral, PhD Centro de Tecnología Mineral - CETEM, Brasil Agustín Bueno, PhD Universidad de Alicante, España Henrique Lorenzo Cimadevila, PhD Universidad de Vigo, España Mauricio Trujillo, PhD Universidad Nacional Autónoma de México, México

Carlos Palacio, PhD Universidad de Antioquia, Colombia Jorge Garcia-Sucerquia, PhD Universidad Nacional de Colombia, Colombia Juan Pablo Hernández, PhD Universidad Nacional de Colombia, Colombia John Willian Branch Bedoya, PhD Universidad Nacional de Colombia, Colombia Enrique Posada, Msc INDISA S.A, Colombia Oscar Jaime Restrepo Baena, PhD Universidad Nacional de Colombia, Colombia Moisés Oswaldo Bustamante Rúa, PhD Universidad Nacional de Colombia, Colombia Hernán Darío Álvarez, PhD Universidad Nacional de Colombia, Colombia Jaime Aguirre Cardona, PhD Universidad Nacional de Colombia, Colombia

DYNA 81 (188), December, 2014. Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online

CONTENTS Editorial Juan D. Velásquez

Corrosion resistance of hybrid films applied on tin plate: Precursor solution acidified with nitric acid (pH=3) Sandra R. Kunst, Gustavo A. Ludwig, Maria R. Vega, Cláudia T. Oliveira & Célia F. Malfatti

Plane geometry drawing tutorial Eduardo Gutiérrez de Ravé, Francisco J. Jiménez-Hornero & Ana B. Ariza-Villaverde

Analytical model of signal generation for radio over fiber systems Gustavo Adolfo Puerto-Leguizamón & Carlos Arturo Suárez-Fajardo

Recycling of agroindustrial solid wastes as additives in brick manufacturing for development of sustainable construction materials Lisset Maritza Luna-Cañas, Carlos Alberto Ríos-Reyes & Luz Amparo Quintero-Ortíz

The use of gypsum mining by-product and lime on the engineering properties of compressed earth blocks Eliana Rocío Jaramillo-Pérez, Josue Mauricio Plata-Chaves & Carlos Alberto Ríos-Reyes

Practical lessons learnt from the application of X-ray computed tomography to evaluate the internal structure of asphalt mixtures Allex Eduardo Alvarez-Lugo & Juan Sebastián Carvajal-Muñoz

Rail vehicle passing through a turnout: Influence of the track elasticity Rodrigo F. Lagos-Cereceda, Kenny L. Alvarez-C., Jordi Vinolas-Prat & Asier Alonso-Pazos

Evolution of the passive harmonic filters optimization problem in industrial power systems Jandecy Cabral-Leite, Ignacio Pérez-Abril, Manoel Socorro Santos-Azevedo, Maria Emilia de Lima-Tostes & Ubiratan Holanda-Bezerra

Restricting the use of cars by license plate numbers: A misguided urban transport policy

9 11 20 26 34 42 52 60 67

Víctor Cantillo & Juan de Dios Ortúzar

75

Active vibration control in building-like structures submitted to earthquakes using multiple positive position feedback and sliding modes

83

Josué Enríquez-Zárate & Gerardo Silva-Navarro

Analysis of customer satisfaction using surveys with open questions José Amelio Medina-Merodio, Carmen de Pablos-Heredero, María Lourdes Jiménez-Rodríguez, Luis de Marcos-Ortega, Roberto Barchino-Plata, Daniel Rodríguez-García & Daniel Gómez-Aguado

Analysis of the economic impact of environmental biosafety works projects in healthcare centres in Extremadura (Spain) Justo García Sanz-Calcedo & Pedro Monzón-González

Assessing the performance of a differential evolution algorithm in structural damage detection by varying the objective function Jesús Daniel Villalba-Morales & José Elias Laier

Efficient reconstruction of Raman spectroscopy imaging based on compressive sensing Diana Fernanda Galvis-Carreño, Yuri Hercilia Mejía-Melgarejo & Henry Arguello-Fuentes

Apply multicriteria methods for critical alternative selection Rosario Garza-Ríos & Caridad González-Sánchez

Flatness-based fault tolerant control César Martínez-Torres, Loïc Lavigne, Franck Cazaurang, Efraín Alcorta-García & David A. Díaz-Romero

Dynamic wired-wireless architecture for WDM stacking access networks Gustavo Adolfo Puerto-Leguizamón, Laura Camila Realpe-Mancipe & Carlos Arturo Suárez-Fajardo

Influence of osmotic pre-treatment on convective drying of yellow pitahaya Alfredo Ayala-Aponte, Liliana Serna-Cock, Jimena Libreros-Triana, Claudia Prieto & Karina Di Scala

Evaluation of thermal behavior for an asymmetric greenhouse by means of dynamic simulations Ervin Alvarez-Sánchez, Gustavo Leyva-Retureta, Edgar Portilla-Flores & Andrés López-Velázquez

A comparative study TiC/TiN and TiN/CN multilayers Miguel J. Espitia-Rico, Gladys Casiano-Jiménez, César Ortega-López, Nicolás De la Espriella-Vélez & Luis Sánchez-Pacheco

Developing a fast cordless soldering iron via induction heating Ernesto Edgar Mazón-Valadez, Alfonso Hernández-Sámano, Juan Carlos Estrada-Gutiérrez, José Ávila-Paz & Mario Eduardo Cano-González

92 100 106 116 125 131 139 145 152 160 166

Selecting working fluids in an organic Rankine cycle for power generation from low temperature heat sources Fredy Vélez

Approach to biomimetic design. Learning and application Ignacio López-Forniés & Luis Berges-Muro

Structural control using magnetorheological dampers governed by predictive and dynamic inverse models Luis Augusto Lara-Valencia, José Luis Vital-de Brito & Yamile Valencia-González

Design of boundary combined footings of rectangular shape using a new model Arnulfo Luévanos-Rojas

Effect of additives on diffusion coefficient for cupric ions and kinematics viscosity in CuSO4H2SO4 solution at 60°C Eugenia Araneda-Hernández, Froilán Vergara-Gutiérrez & Antonio Pagliero-Neira

Influence of silicon on wear behaviour of “Silal” cast irons Ana I. García-Diez, Carolina Camba-Fabal, Ángel Varela-Lafuente, Víctor Blázquez-Martínez, José Luís Mier-Buenhombre & Benito Del Río-López

Effect of cationic polyelectrolytes addition in cement cohesion Edison Albert Zuluaga-Hernández & Bibian A. Hoyos

A new dynamic visualization technique for system dynamics simulations Ricardo Sotaquirá-Gutiérrez

Scenarios of photovoltaic grid parity in Colombia Maritza Jiménez, Lorena Cadavid & Carlos J. Franco

Our cover Image alluding to Article: Efficient reconstruction of Raman spectroscopy imaging based on compressive sensing Authors: Diana Fernanda Galvis-Carreño, Yuri Hercilia Mejía-Melgarejo & Henry Arguello-Fuentes

173 181 191 199 209 216 222 229 237

DYNA 81 (188), December, 2014. Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online

CONTENIDO Editorial Juan D. Velásquez

Resistencia a la corrosión de películas híbridas sobre láminas de estaño: Solución precursora acidificada con ácido nítrico (pH=3) Sandra R. Kunst, Gustavo A. Ludwig, Maria R. Vega, Cláudia T. Oliveira & Célia F. Malfatti

Tutorial de dibujo geométrico Eduardo Gutiérrez de Ravé, Francisco J. Jiménez-Hornero & Ana B. Ariza-Villaverde

Modelo analítico de generación de señales para sistemas radio sobre fibra

11 20

Gustavo Adolfo Puerto-Leguizamón & Carlos Arturo Suárez-Fajardo

26

Reciclaje de residuos sólidos agroindustriales como aditivos en la fabricación de ladrillos para el desarrollo sostenible de materiales de construcción

34

Lisset Maritza Luna-Cañas, Carlos Alberto Ríos-Reyes & Luz Amparo Quintero-Ortíz

El uso de residuos de minería de yeso y cal sobre las propiedades de ingeniería de los bloques de tierra comprimida Eliana Rocío Jaramillo-Pérez, Josue Mauricio Plata-Chaves & Carlos Alberto Ríos-Reyes

42

Lecciones prácticas aprendidas a partir de la aplicación de la tomografía computarizada de rayos-x para evaluar la estructura interna de mezclas asfálticas

52

Allex Eduardo Alvarez-Lugo & Juan Sebastián Carvajal-Muñoz

Influencia de la elasticidad de vía al circular por un desvío ferroviario Rodrigo F. Lagos-Cereceda, Kenny L. Alvarez-C., Jordi Vinolas-Prat & Asier Alonso-Pazos

Evolución del problema de optimización de ls filtros pasivos de armónicos en sistemas eléctricos industriales Jandecy Cabral-Leite, Ignacio Pérez-Abril, Manoel Socorro Santos-Azevedo, Maria Emilia de Lima-Tostes & Ubiratan Holanda-Bezerra

Restricción vehicular según número de patente: Una política de transporte errónea

60 67

Víctor Cantillo & Juan de Dios Ortúzar

75

Control activo de vibraciones en estructuras tipo edificio sometidas a sismos utilizando múltiple retroalimentación positiva de la posición y modos deslizantes

83

Josué Enríquez-Zárate & Gerardo Silva-Navarro

Análisis de la satisfacción de cliente mediante el uso de cuestionarios con preguntas abiertas José Amelio Medina-Merodio, Carmen de Pablos-Heredero, María Lourdes Jiménez-Rodríguez , Luis de Marcos-Ortega, Roberto Barchino-Plata, Daniel Rodríguez-García & Daniel Gómez-Aguado

Análisis del impacto económico de la bioseguridad ambiental en proyectos de obras en centros sanitarios de Extremadura (España)

92

Justo García Sanz-Calcedo & Pedro Monzón-González

100

Valoración del desempeño de un algoritmo de evolución diferencial en detección de daño estructural considerando diversas funciones objetivo

106

Jesús Daniel Villalba-Morales & José Elias Laier

Reconstrucción eficiente de imágenes a partir de espectroscopia Raman basada en la técnica de sensado compresivo Diana Fernanda Galvis-Carreño, Yuri Hercilia Mejía-Melgarejo & Henry Arguello-Fuentes

Selección de alternativas críticas aplicando un enfoque multicriterio Rosario Garza-Ríos & Caridad González-Sánchez

Control tolerante a fallas basado en planitud César Martínez-Torres, Loïc Lavigne, Franck Cazaurang, Efraín Alcorta-García & David A. Díaz-Romero

Arquitectura dinámica fija-móvil para redes de acceso WDM apiladas Gustavo Adolfo Puerto-Leguizamón, Laura Camila Realpe-Mancipe & Carlos Arturo Suárez-Fajardo

Influencia de un pre-tratamiento osmótico sobre el secado convectivo de pitahaya amarilla Alfredo Ayala-Aponte, Liliana Serna-Cock, Jimena Libreros-Triana, Claudia Prieto & Karina Di Scala

Evaluación del comportamiento térmico de un invernadero asimétrico mediante simulaciones dinámicas Ervin Alvarez-Sánchez, Gustavo Leyva-Retureta, Edgar Portilla-Flores & Andres López-Velázquez

9

116 125 131 139 145 152

Un estudio comparativo de las multicapas TIC/TiN y TiN/CN Miguel J. Espitia-Rico, Gladys Casiano-Jiménez, César Ortega-López, Nicolás De la Espriella-Vélez & Luis Sánchez-Pacheco

Desarrollo de un cautín inalámbrico rápido a través de calentamiento por inducción Ernesto Edgar Mazón-Valadez, Alfonso Hernández-Sámano, Juan Carlos Estrada-Gutiérrez, José Ávila-Paz & Mario Eduardo Cano-González

Seleccionando fluidos de trabajo en ciclos Rankine para generación de energía a partir de fuentes de calor de baja temperatura Fredy Vélez

Aproximación al diseño biomimético. Aprendizaje y aplicación

166 173

Ignacio López-Forniés & Luis Berges-Muro

181

Control estructural utilizando amortiguadores magnetoreológicos gobernados por un modelo predictivo y por un modelo inverso dinámico

191

Luis Augusto Lara-Valencia, José Luis Vital-de Brito & Yamile Valencia-González

Diseño de zapatas combinadas de lindero de forma rectangular utilizando un nuevo modelo Arnulfo Luévanos-Rojas

Efecto de aditivos en el coeficiente de difusión de iones cúpricos y la viscosidad cinemática en solución CuSO4 H2SO4 a 60° Eugenia Araneda-Hernández, Froilán Vergara-Gutierrez & Antonio Pagliero-Neira

Influencia del silicio en el comportamiento al desgaste de las fundiciones tipo “Silal” Ana I. García-Diez, Carolina Camba-Fabal, Ángel Varela-Lafuente, Víctor Blázquez-Martínez, José Luís Mier-Buenhombre & Benito Del Río-López

Efecto de la adición de polielectrólitos catiónicos en la cohesión del cemento Edison Albert Zuluaga-Hernández & Bibian A. Hoyos

Una nueva técnica de visualización dinámica para simulaciones en dinámica de sistemas Ricardo Sotaquirá-Gutiérrez

Escenarios de paridad de red fotovoltaica en Colombia Maritza Jiménez, Lorena Cadavid & Carlos J. Franco

Nuestra carátula Imágenes alusivas al artículo: Reconstrucción eficiente de imágenes a partir de espectroscopia Raman basada en la técnica de sensado compresivo Autores: Diana Fernanda Galvis-Carreño, Yuri Hercilia Mejía-Melgarejo & Henry Arguello-Fuentes

160

199 209 216 222 229 237

Editorial

Una guía corta para escribir Revisiones Sistemáticas de Literatura Parte 2 En esta editorial se discuten los orígenes de la revisión sistemática de literatura (SLR) y sus principales aplicaciones en gestión e ingeniería.

1

Orígenes de la Revisión Sistemática de Literatura

La metodología de revisión sistemática de literatura (SLR) surge originalmente a partir del concepto de evidence-based medicine (EBM), que se refiere al hecho de que el individuo en su práctica profesional debe tomar decisiones soportadas en su experiencia, juicio profesional y en la evidencia objetiva más rigurosa que este disponible [1]; de ahí que el énfasis de la actividad investigativa este orientada a demostrar objetiva y transparentemente qué es lo que realmente funciona y que el énfasis de la práctica profesional este orientado a usar dicha información para tomar mejores decisiones. La EBM nace como respuesta a que la mayoría de estudios primarios en medicina y ciencias de la salud carecían de un rigor apropiado, o presentaban resultados contradictorios; y a la dificultad de poder sintetizar adecuadamente grandes volúmenes de evidencia cuestionable [2]; en consecuencia, muchas revisiones de literatura presentaban conclusiones deficientes, inapropiadas o sesgadas [3]. Estas situaciones causaron que la evidencia tomara un rol central en la investigación y el ejercicio profesional [3]. El concepto de EBM fue posteriormente extendido en UK (y otros países), desde la década de los 80s, a la política pública y la práctica profesional (evidence-based policy and practice —EBPP—) pero particularmente se difundió en las ciencias sociales, la educación y la justicia criminal [2]; como consecuencia, se desarrollaron muchas guías y manuales de buenas prácticas [3]. Tanto el concepto y práctica de la EBM como de la EBPP implican la realización de estudios primarios que provean evidencias con altos estándares de rigurosidad, transparencia, calidad y objetividad; recursos para almacenar y hacer disponible la evidencia recolectada a la comunidad científica y profesional; y mecanismos para su sintetización y análisis.

En este contexto, la revisión sistemática de literatura (SLR) entra a jugar un papel fundamental como un mecanismo para recolectar, organizar, evaluar y sintetizar toda la evidencia disponible respecto a un fenómeno de interés, ya sea para mejorar la práctica actual (mostrar que es lo que realmente

funciona) o para sugerir nuevas direcciones de investigación.

Pero para ello, la revisión de literatura debe cumplir con los mismos estándares de calidad con que se realizan los estudios primarios de la más alta calidad. Es así como emerge la metodología de SLR en respuesta a dicha necesidad. Ya que la EBM se sustenta fundamentalmente en estudios cuantitativos y métodos estadísticos de análisis, el desarrollo de guías para realizar SLRs ha estado fundamentalmente orientado hacia estos fines, y particularmente a la utilización del meta-análisis, que es un procedimiento estadístico para la agregación de los resultados cuantitativos provenientes de varios estudios empíricos, con el fin de inferir estadísticamente resultados más confiables de los que se pueden obtener por la realización de estudios individuales [1][3].

2

Aplicaciones de SLR en Gestión e Ingeniería

Claramente el concepto de la EBPP puede ser aplicado en todas la disciplinas profesionales, pero particularmente la ingeniería puede obtener grandes beneficios; esto es especialmente importante en aquellas áreas de rápido desarrollo, tales la computación, la energía y la electrónica, en las cuales los desarrollos conceptuales pueden provenir de forma independiente desde diferentes áreas; esto puede dificultar la búsqueda y recopilación de evidencias. Así mismo, las revisiones de literatura en la ingeniería son tradicionalmente narrativas —excepto en la ingeniería de software y la política energética— y adolecen de todas las limitantes que ya se han discutido.

Dados los beneficios de la EBPP, no resulta extraño que dichas prácticas se hayan extendido a otras disciplinas.

Tranfield et al [3] propone el uso de la metodología de SLR en el área de la gestión, discute sus beneficios, y como las diferencias entre dicha área y la medicina pueden afectar el proceso para realizar SLRs. Kitchenham y Charters [1] prepararon unos lineamientos con base en las guías existentes para el desarrollo de SLR en medicina y ciencias sociales, y particularmente en los preparados por el Centre for Reviews and Dissemination (CRD) [4], para que fueran usados por investigadores, profesionales y estudiantes de postgrado en el área de la ingeniería de software en la preparación de revisiones de literatura rigurosas.

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (187), pp. 9-10. October, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI:

Velásquez / DYNA 81 (187), pp. 9-10. October, 2014.

Mientras que en las ciencias de la vida y la salud existen abundantes estudios que usan la metodología de SLR, existen muy pocos ejemplos en ingeniería –excepto en el campo de la ingeniería de software–. La metodología de SLR ha sido usada para: analizar las herramientas para medir desempeño de construcciones en Nigeria [5]; identificar temáticas en la enseñanza de estructuras de datos y matemáticas discretas en currículos de ciencias de la computación [6]; analizar los problemas de adopción y difusión en sistemas de información, tecnologías de la información y tecnologías de la comunicación [7]; realizar un análisis de las técnicas para dar accesibilidad a la Web [8]; discutir las técnicas no lineales usadas en el pronóstico de la demanda de electricidad [9]; analizar el pronóstico de índices de mercado usando lógica difusa [10]; analizar y discutir los factores críticos de éxito en la presentación oral científica [11]; discutir la aplicación de técnicas de decisión para la construcción de modelos para la selección de proveedores [12]; analizar como el concepto de ‘leagility’ ha evolucionado en la literatura científica actual sobre administración [13]; integrar el extenso cuerpo de literatura sobre ventas y planeamiento de la operación [14]; determinar las tendencias actuales, autores más importantes y problemas de investigación, así como sintetizar el conocimiento existente sobre la identificación de radiofrecuencias [15]; para analizar los métodos de ensamble de redes neuronales artificiales en el pronóstico de series de tiempo económicas o financieras [16].

En general, las ingenierías modernas son disciplinas jóvenes en comparación con la medicina, y al igual a como ocurre en la gestión [3], los estudios en estas áreas difícilmente comparten los mismos objetivos o investigan los mismos interrogantes. Es así como para cada tópico particular existe un número relativamente bajo de estudios, posiblemente realizados desde diferentes ópticas; pero más aún, en el caso de estudios cuantitativos, difícilmente se usan los mismos datos experimentales, de tal forma que se hace imposible la agregación de estudios para aumentar la confiabilidad de los resultados. Existen contadas excepciones, en las que se ha recopilado y puesto a disposición de la comunidad científica bases de datos de problemas con el fin de que los resultados de diferentes investigaciones sean comparables; un ejemplo es el UCI Machine Learning Repository en el que se pone a disposición de la comunidad más de 280 conjuntos de datos para la experimentación con técnicas de aprendizaje de máquinas; sin embargo, los investigadores no tiene la obligación de usar estos conjuntos de datos. Sin embargo, y a diferencia de muchas de las guías existentes, los lineamientos de Kitchenham y Charters [1] y de Tranfield et al [3] no enfatizan el meta-análisis como una herramienta fundamental debido a que existe poca evidencia empírica cuantitativa en comparación con otras áreas de investigación [1].

Referencias [1] [2] [3]

[4]

[5]

[6]

[7]

[8]

[9] [10]

[11] [12]

[13] [14] [15] [16]

Juan D. Velásquez, MSc, PhD Profesor Titular Universidad Nacional de Colombia E-mail: jdvelasq@unal.edu.co http://orcid.org/0000-0003-3043-3037

10

Kitchenham, B.A.; Charters, S. (2007). Guidelines for Performing Systematic Literature Reviews in Software Engineering. Technical Report EBSE-2007-01.

S. Sorrell. “Improving the evidence base for energy policy: The role of systematic reviews”, Energy Policy, vol. 35, no. 3, pp. 18581871, 2007.

D. Tranfield, D. Denyer and P. Smart. “Towards a Methodology for Developing Evidence-Informed Management Knowledge by Means of Systematic Review,” British Journal of Management, vol. 14, pp.

207-222, 2003.

K. S. Khan, G. ter Riet, Gerben. J. Glanville, A. Sowden, and J. Kleijnen (eds). “Undertaking Systematic Review of Research on Effectiveness. CRD’s Guidance for those Carrying Out or Commissioning Reviews,” CRD Report Number 4 (2nd Edition), NHS Centre for Reviews and Dissemination, University of York, March 2001.

H. Koleoso, M. Omirin, Y. Adewunmi and G. Babawale.

“Application of existing performance evaluation tools and concepts to Nigerian facilities management practice,” International Journal of Strategic Property Management, vol. 17, no. 4, pp. 361-376, 2013.

T. Whelan, S. bergin and J. F. Power. “Teaching Discrete Structures: A systematic review of the literature,” SIGCSE '11 Proceedings of the 42nd ACM technical symposium on Computer science education, pp. 275-280, 2011.

Y. K. Dwivedi, M. D. Williams, B. Lai and N. Mustafee. “An Analysis of Literature on Consumer Adoption and Diffusion of Information System/Information Technology/Information and Communication Technology,” International Journal of Electronic Government Research, vol. 6, no. 4, pp. 58-73, 2010.

A. P. Freire, R. Goularte and R. P. de M. Fortes. “Techniques for developing more accessible web applications: a survey towards a process classification”. SIGDOC '07 Proceedings of the 25th annual ACM international conference on Design of communication, pp.

162-169, 2007.

V. M. Rueda, J. D. Velásquez and C. J. Franco. “Recent advances in load forecasting using nonlinear models”. DYNA, vol. 78, no. 167, pp. 7-16, 2011.

A. Arango, J. D. Velásquez, C. J. Franco. “Técnicas de Lógica Difusa en la Predicción de Índices de Mercados de Valores: Una Revisión de Literatura”. Revista Ingenierías, vol. 11, no. 22, pp.

115-123. 2013.

J. D. Velasquez. “Factores de Éxito en la Comunicación Oral Científica,” Facultad de Minas, Universidad Nacional de Colombia, 2011. No publicado.

J. Chai, J. N. K. Liu, E. W. T. Ngai. “Application of decisionmaking techniques in supplier selection: A systematic review of literatura,” Expert Systems with Applications, vol. 40, no. 10, pp.

3872-3885, 2013.

M. M. Naim and J. Grosling. “On leanness, agility and leagile supply chains,” International Journal of Production Economics, vol.

131, no. 1, pp. 342-354, 2011.

A. M. T. Thomé, L. F. Scavarda, N. S. Fernandez, A. J. Scavarda.

“Sales and operations planning: A research synthesis,” International Journal of Production Economics, vol. 138, no. 1, pp. 1-13, 2012.

Z. Irani, A. Gunasekaran and Y. K. Dwivedi. “Radio frequency identification (RFID): research trends and framework,” International Journal of Production Research, vol. 48, no. 9, pp. 2485-2511, 2010.

L. F. Rodríguez, J. D. Velásquez and C. J. Franco. “Una investigación científica acerca del progreso de métodos de ensamble basados en inteligencia computacional para predicción de series de tiempo económicas y financieras” in Modelos no Lineales en Series Económicas y/o Financieras. Universidad de Guadalajar, 2012.

Corrosion resistance of hybrid films applied on tin plate: Precursor solution acidified with nitric acid (pH=3) Sandra R. Kunst a, Gustavo A. Ludwig b, Maria R. Vega c, Cláudia T. Oliveira d & Célia F. Malfatti e a b

Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, tessaro.sandra@gmail.com Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, gustavolludwig@gmail.com c Universidade Federal do Rio Grande do Sul, Porto Alegre, ortega.vega@ufrgs.br d Univerrsity Feevale, Novo Hamburgo, Brazil, ctofeevale@gmail.com e Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, celia.malfatti@ufrgs.br

Received: April 29th, 2013. Received in revised form: August 13th, 2014. Accepted: October 25th, 2014.

Abstract Siloxane – poly (methylmethacrylate)-based materials are systems formed by a silicon network, to which chains of poly (methylmethacrylate) are linked by covalent bonds or by physical interactions. Their stability and adherence allow their application on substrates like tin plate in order to increase the corrosion resistance. The aim of this work is to coat tin plate with a hybrid film obtained from a sol consisting of alkoxide precursors: 3 - (trimethoxysilylpropyl) methacrylate (TMSM) and poly (methyl methacrylate) PMMA. Effect of tetraethoxysilane (TEOS) addition was evaluated. Morphology was evaluated by SEM and contact angle. Electrochemical behavior was evaluated by open circuit potential (OCP), potentiodynamic polarization and electrochemical impedance spectroscopy (EIS).

Results showed that siloxane-PMMA film obtained with a higher addition of TEOS had higher thickness. However, intense densification caused by TEOS addition promoted crack formation, thereby compromising the corrosion resistance.

Keywords: Hybrid films; nitric acid; tin plate; pH=3; TEOS; corrosion.

Resistencia a la corrosión de películas híbridas sobre láminas de estaño: Solución precursora acidificada con ácido nítrico (pH=3) Resume Los materiales híbridos a base de siloxano y PMMA se constituyen por una red de silicona, a la que están enlazadas cadenas de polimetilmetacrilato. Su estabilidad y adherencia permiten su aplicación en láminas de estaño para aumentar la resistencia a la corrosión.

En este trabajo se recubrió una lámina de estaño con una película híbrida a partir de un sol que consta de los precursores alcóxidos: 3trimetoxisililpropilmetacrilato (TMSM) y polimetilmetacrilado (PMMA). Se evaluó la influencia de la adición de tetraetoxisilano (TEOS).

Se evaluó la morfología de los filmes por SEM y ángulo de contacto. Se evaluó el comportamiento electroquímico con medidas de potencial a circuito abierto, polarización potenciodinámica e impedancia electroquímica. Los resultados muestran que la película con mayor adición de TEOS tuvo mayor espesor; sin embargo, el TEOS intensificó la densificación. Así, se formaron grietas que comprometieron la resistencia a la corrosión.

Palabra clave: Películas híbridas, ácido nítrico, láminas de estaño, pH = 3; TEOS corrosión.

1. Introduction During the last years, polymeric-based surface treatments have arisen as non-toxic alternatives to traditional chromating process. Among those alternatives, pretreatments based on siloxane-PMMA have showed promising results, attracting attention of industries. These hybrids films improve corrosion resistance [1] and adhesion properties of

organic layers. Besides, these hybrids films reduce the environmental impacts compared to the chromatization process [2]. Moreover, siloxane-PMMA hybrid films promote an excellent anchorage on the metallic surface of tin plate [3,4] for posterior paint application, which can be done by the releasing of silicone from functional siloxanes using an ultraviolet cure [5].

PMMA-based materials, such as polymeric blend [6,7]

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 11-19. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.37998

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

and organic-inorganic hybrid materials [3,8,9,10,11] have been investigated intending to get better properties. These materials are classified according to their chemical bond characteristics. Class I hybrids have their organic and inorganic components bonded by Van der Waals forces, such as PMMA on ZrO2 [9]. Class II hybrids are created by merging the inorganic and organic phases by covalent bonds, like those obtained through sol-gel method using TMSM and TEOS as precursors [11]. These formulations combine the hardness, scratch resistance and thermal stability of the ceramic component with flexibility, transparency and tunable adhesion of the organic substances. Class III hybrids, basically consist of merging class I with class II hybrids, getting the adhesion properties of one with the inorganicorganic network provided by the other. This way, the strategy used for researchers is to employ a coupling agent such as trialkoxides silane functionalized with vinylic ligands, which allow the connexion between organic and inorganic phases [12,13,14]. A trialkoxide silane widely used a coupling agent is 3-methacriloxi-propyl-trimethoxy-silane (TMSM, also known as MPTS). TMSM, besides being a great coupling agent between organic and inorganic phases, it is photo sensitive to UV radiation, which makes it possible to control its refractive index, as in optical devices and in materials with low dielectric constant [15,16,17]. TMSM can promote the formation of uncracked, micrometric films. Also the film refractive index can be controlled by the methacrylate introduction in the hybrid structure and the curing method, such as UV or heat treatment [18] or by heat treatments, allow the increase of the refractive index.

The aim of this work is to study the behavior of hybrid films and the effect of the addition of tetraethoxysilane (TEOS) on corrosion on a tin plate substrate. The tin plate was coated with a hybrid film obtained from a sol constituted by the alkoxide precursors: 3 - (trimetoxisililpropil) methacrylate (TMSM) and poly (methyl methacrylate) PMMA, and it was evaluated the effect of the addition of tetraethoxysilane (TEOS). The films were obtained by a dipcoating process and cured for 3 hours at 160 °C. The hydrolysis was carried out at pH=3 using nitric acid.

Tetraethyl orthosilicate (TEOS)

3-(trimethoxysilylpropyl) methacrylate (TMSM)

Methyl methacrylate (MMA)

CH3 H2C COOCH Figure 1. Structural formula of the organic and inorganic precursors. Source: Adapted from Kunst, S. R .et al 2014.

from ethyl alcohol. The structural formula of the organic and inorganic precursors are shown in Figure 1.

The sol-gel method was employed in the preparation of hybrid organic-inorganic materials. In the synthesis of the inorganic phase, precursors TMSM and TEOS were mixed at 60 °C for 1 hour. Hydrolysis was performed at medium pH=3 using nitric acid as a catalyst and ethanol and water as solvents. The organic phase consisted of homogenization of MMA at room temperature, where the thermal initiator was benzoyl peroxide (BPO). Finally, the two solutions (inorganic and organic) were mixed. The films were obtained by a dip-coating process, with a removal rate of 14 cm.min-1.

Subsequently, the coated substrates were heat treated (cured) at a temperature of 160 °C for 3 hours at a heating rate of 5 °C.min - 1 . This treatment increases the degree of polymerization because it promotes the formation of free radicals from the C=C bonds existing in TMSM and MMA. Table 1 shows the description of samples used.

2. Experimental Method 2.1. Substrate treatment The tin plates were rinsed with acetone and dried. After that, they were degreased with neutral detergent at 70 °C by immersion for 10 minutes. Then, they were properly washed and dried.

2.2. Elaboration of Siloxane-PMMA hybrid films

Table 1. Description of the samples.

Samples Description Tin plate Tin plate without coating Tin plate coated with hybrid film with addition of T1N3 TEOS (level 1), nitric acid and pH=3 de TEOS. Tin plate coated with hybrid film with addition of T2N3 TEOS (level 2, increased concentration of TEOS in the level 1), and nitric acid pH=3 Tin plate coated with hybrid film with addition of T3N3 TEOS (level 3, increased concentration of TEOS in the level 1 and level 2), and nitric acid pH=3 Source: The authors.

The hybrid films were obtained on a tin plate substrate from a sol constituted by the silane precursors: γmethacryloxypropyl-trimethoxysilane (TMSPMA – C10H20O5Si) and tetraethoxysilane (TEOS - C8H20O4Si) with addition of methyl methacrylate (MMA), which was distilled to remove the polymerization inhibitor (hydroquinone) and impurities, and stored in a freezer prior to use. Benzoyl peroxide, BPO (reagent), was recrystallized 12

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

Samples containing a small amount of TEOS in the hybrid film were named level 1, while the samples containing high amount to TEOS were named level 2 and the sample that has the greatest amount of TEOS was named level 3.

Small discontinuities

a)

2.3. Characterization of Siloxane-PMMA hybrid films In order to evaluate the films morphology, scanning electron microscopy (SEM) and the sessile drop method were performed.

Scanning electron microscopy (SEM) in a JEOL-JSM 5800 with an acceleration voltage of 20 keV was used to evaluate the thickness, by cross section, and the morphology of hybrid films.

The film hydrophobicity was determined by the contact angle measurement from the sessile drop method in equipment developed by the research laboratory in corrosion (LAPEC) of UFRGS. The contact angle was determined by using image analyses software (also developed by this laboratory).

The corrosion performance of the films was evaluated by the following electrochemical techniques: open circuit potential (OCP) monitoring, potentiodynamic polarization and electrochemical impedance spectroscopy (EIS) measurements in a 0.05M NaCl solution. For the electrochemical characterization, a three-electrode cell was used, with platinum as the counter electrode and SCE (Saturated Calomel Electrode) as the reference electrode. The work electrode area was 0.626 cm². The OCP monitoring and potenciodynamic polarization curves were performed in a potentiostat PAR 273, following the ASTM G5 [19] standard. The OCP was monitored for the first hour of immersion in the electrolyte before the polarization curves and the EIS measurements. The polarization curves were measured with a scan rate of 1 mV.s-1 in a range from 200 mV (under OCP) up to 400 mV (above OCP).

Potentiodynamic polarization was carried out in the same PAR 273 as the OCP measurements. The equipment used for the EIS measurements was a potentiostat (Omnimetra Mod. PG-05) coupled to a frequency response analyzer model Solartron 1255. The amplitude of the EIS perturbation signal was 10 mV, and the frequency range studied was from 1 x 105 to 1 x 10-2 Hz, according to the ASTM G106 [20] standard.

b)

Discontinuities

Delamination

Cracks

c)

Figure 2. Micrographs obtained from SEM of hybrid films: (a) T1N3, (b) T2N3 and (c) T3N3.

Source: The authors.

3. Results and discussion The thickness of the films was determined by the analysis of images obtained by SEM of the cross section (Figure 3) and the results are shown in Table 2. It is observed that the sample with highest thickness was the sample with greater addition of TEOS (T2N3 and T3N3), (Figure 3-b and 3-c). This is associated with higher content of silanol groups (the inorganic phase) due to the greater amount of TEOS in the film. The real density of hybrid siloxane-PMMA increases with siloxane content in the film, because the siloxane phase density (2.2 g.cm3) is higher than the PMMA density (1.2 g.cm3).

However, it was verified that the sample T3N3 (Fig. 2-c) showed cracks due to intense densification of the film promoted by the excessive addition of TEOS.

3.1. Morphological characterization Figure 2 shows the SEM micrographs for the films studied: T1N3 (with TEOS addition of level 1), T2N3 (with TEOS addition of level 2) and T3N3 (with TEOS addition of level 3). From the SEM micrographs one can observe that the sample T1N3 (Figure 2-a) showed small discontinuities in the coating while for the sample T2N3 (Figure 2-b), the formation of the discontinuities were more accentuated. The sample T3N3 (Figure 2-c) showed cracks and delamination of the hybrid film. This is due to the formation of a complete porous ceramic structure, which is brittle [21], after hydrolysis and cross-linking of TEOS. Therefore, the TEOS addition contributed to the irregular coverage observed.

13

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

a)

Resin

a)

b)

c)

d)

Hybrid film

Tin Plate

b)

Hybrid film

Figure 4. Images obtained for the contact angle determination by the sessile drop method: (a) tin plate, (b) T1N3, (c) T2N3 and (d) T3N3.

Source: The authors.

Resin Tin Plate

Table 3. Contact angle values obtained by the sessile drop method.

Samples Contact Angle Standard deviation

c)

Tin plate T1N3 T2N3 T3N3 Source: The authors.

Hybrid film

73° 69° 65° 58°

1.9 0.5 0.6 0.4

Cracks highest wettability (lowest contact angle values), among the samples studied. This behavior can be associated to the hydrophilic behavior resulting from excessive TEOS addition [22] and also to the film.

Tin Plate Resin Figure 3. Cross section micrographs obtained by SEM: (a) T1N3, (b) T2N3 and (c) T3N3.

Source: The authors.

3.2. Electrochemical characterization Measurements of open circuit potential (OCP) were made in order to monitor the variation of potential with soaking time in 0.05 M NaCl solution as illustrated in Figure 5. The samples T1N3 and T2N3 showed that these films presented open circuit potential values displaced towards passive potentials related to the tin plate substrate, i.e., the hybrid films obtained promoted the formation of a barrier between the substrate and the medium.

This is due to the formation of PMMA chains cross-linked with polysiloxane nanodomains, hindering the passage of the electrolyte through the film towards the substrate (tin plate).

Sample T3N3 suffered electrolyte permeation due to the presence of cracks on the film surface, as well as delamination problems; henceforth, the OCP values for this sample were more active than the ones for the tin plate substrate and the other studied systems.

OCP shows stabilization after 1200 seconds of soaking time for all the hybrid films due to surface homogeneity, shown in T1N3 and T2N3 by the regular morphology of the obtained film, as observed in the SEM micrographs (Figure 1-a, Figure 1-b). Meanwhile, for the system T3N3, OCP stabilization can be explained because there was the formation of tin oxides due to the electrolyte permeation provoked by cracks and delamination.

Table 2. Thickness of siloxane-PMMA hybrid films.

Samples

Thickness (µm)

T1N3

1.36

Standard deviation (µm) 0.39

T2N3

5.71

0.22

T3N3

4.07

0.27

Source: The authors.

The results of the contact angle area showed in Figure 4 and Table 3. The value of the contact angle obtained for tin plate (Figure 4-a and Table 3) is associated with the fact that this behavior is probably due to the presence of tin oxides (SnOx), as SnO, SnO2 and its hydrated forms, which confer a complete or partial coverage of the surface, increasing substrate hydrophobicity. The hybrid film T1N3 showed the highest contact angle value and, consequently, lower wettability. This behavior is related to the formation of more siloxane groups in the sample with TEOS, forming a more compact network preventing the absorption of water, making the film more hydrophobic. Regular morphology can be observed (Figure 2-b). The sample T3N3 showed the 14

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

Table 4. Obtained data from Tafel extrapolation.

Ecorr Samples icorr (A/cm²)

Rp (â„Ś.cm2)

( V) -439

5.54 x 104

T1N3

1.87 x 10

-8

-534

1.40 x 106

T2N3

3.25 x 10-8

-547

8.01 x 106

T3N3

-7

-472

1.01 x 105

Tin plate

4.71 x 10-7

2.62 x 10

Source: The authors.

The anodic branch of the polarization curve is very similar for all the hybrid films meanwhile the tin plate shows a small passivation region caused by tin oxides. For the hybrid films, the anodic reaction is not inhibited by high potential, associated to two factors: lower O2 diffusion or cathodic area decreasing.

Figure 7 and Figure 8 show electrochemical impedance Bode diagrams for hybrid siloxane-PMMA films (T1N3, T2N3 and T3N3) and for tin plate substrates obtained for 24 and 96 hours, respectively, of immersion in a 0.05 M NaCl solution.

After 24 h of immersion (Figure 7), tin plate presented only one time constant in the medium to low frequency range, which was attributed to a passivating top layer of tin oxides (SnOx), such as SnO, SnO2 [24]. The hybrid siloxane-PMMA films (T1N3, T2N3 and T3N3) films showed a time constant at a higher frequencies range, less intense than that observed in 24 hours of immersion, and a time constant in the medium to low frequency range, indicating overlapped time constants (Figure 7). The high frequency phenomenon is associated with the barrier properties of the siloxane-PMMA film. On the other hand, the phenomenon observed at the medium to low frequency, indicates an acceleration of the interfacial process associated with the tin oxides on the substrate surface [25].

After 96 h of immersion (Figure 8), tin plate presented two overlapped time constants, indicating the degradation of the tin oxides. At the end of the experiment, red corrosion products were observed on the electrode surface, indicating that during the experiment, iron was dissolved and diffused to the surface forming iron oxides (Figure 9-a).

For the T3N3 film, a time constant observed in high frequency for 24h disappears, possibly related to poor and/or very cracked film formation. Besides, a time constant is observed in the medium to low frequency range, indicating an acceleration of the interfacial process associated with the tin oxides on the substrate surface.

In Figure 8, the samples T1N3 and T2N3 film presents one time constant at high frequency, but less significant than 24 hours of immersion in NaCl 0.05M. This could be explained by the T1N3 and T2N3 film structure where the radical of MMA covalently bonds to the TMSM moieties through polymerization reactions that can only interact by weak Van der Waals forces. Taking into account these structural considerations, a lower thickness of the film is expected, due to the weak bonds (Figure 3 and Table 2), and consequently not resistant to long periods of immersion in a NaCl 0.05 M solution. In this film, a time constant in the medium to low frequency range was also observed, indicating an acceleration of the interfacial process associated with the tin oxides on the substrate surface.

Figure 5. Open circuit potential curves for the films (T1N3, T2N3 and T3N3) and for tin plate without coating.

Source: The authors.

The results (Fig. 6-b and Table 4) showed that the hybrid films siloxane-PMMA with nitric acid and pH=3 (T1N3, T2N3 and T3N3) promoted the increase in polarization resistance (Rp) and the decrease in corrosion current density (icorr) values related to tin plate, evidencing the protective behavior of these films.

All the studied hybrid films displaced the corrosion potential towards the cathodic side, meanwhile they diminished the corrosion current density and increased the polarization resistance, as can be observed in Fig. 6 and Table 4. Compared to tin plate, T1N3 and T2N3 reduced the corrosion current density by one order of magnitude and the polarization resistance for these samples augmented by two orders of magnitude. T3N3 kept the corrosion current density in the same order of magnitude as tin plate but increased the polarization resistance just in one order of magnitude. It is in agreement with that observed in the OCP monitoring results.

These phenomena can be explained regarding the amount of TEOS present in the precursor solution of sample T3N3, since the excess of TEOS bring about the formation of a porous ceramic structure, which is brittle. Hence, excess TEOS could promote the formation of cracks that allow the electrolyte to permeate, making the film T3N3 less polarization resistant [23].

Figure 6. Polarization curves in a 0.05M NaCl solution.

Source: The authors.

15

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

hybrid films. A value of n=1 corresponds to smooth surface, therefore CPE should be substituted by an ideal capacitor C. n=0.5 suggests a response of diffusion or porous material, and 0.5<n<1 was associated to heterogeneous, rough or non-homogeneous current distribution [26,27]. On the other hand, Re is the electrolyte resistance, RMF and CPEMF represent a medium frequency range which was attributed to the tin oxides (SnOx), passivating top layer [24]. RHF and CPEHF represent a higher frequency range, associated to the siloxane-PMMA film. RLF and CPELF represent a low frequency range phenomenon, associated to the degradation of the tin oxides.

Figure 7. Bode diagrams obtained for the uncoated tin plate and post-treated with siloxane-PMMA hybrid films in a 0.05 M NaCl solution: 24 hours immersion.

Source: The authors.

Table 5.

Electrical elements fitted values for tin plate for 24 and 96 h of immersion in a 0.05M NaCl solution.

Tinplate

24h

96h

1586

157.9

1.86 x 10-6 0.83 2.85 x 105

2.86 x 10-5 0.78 3.91 x 102 3.02 x 10-4 0.49 7.28 x 103

Fitted circuit Re (Ω.cm2)

Figure 8. Bode diagrams obtained for the uncoated tin plate and post-treated with siloxane-PMMA hybrid films in a 0.05 M NaCl solution: 96 hours immersion.

Source: author

CPEHF (Fcm-2) n RHF (Ω.cm2) CPEMF (Fcm-2) n RMF (Ω.cm2) CPELF (Fcm-2) n RLF(Ω.cm2) Source: The authors.

Table 6.

Electrical elements fitted values for T1N3 for 24 and 96 h of immersion in a 0.05M NaCl solution.

T1N3 24h 96h Fitted circuit Re (Ω.cm2) 158.2 161.2 CPEHF (Fcm-2) 4.12 x 10-6 5.20 x 10-6 n 0.77 0.78 RHF (Ω.cm2) 4392 1280 CPEMF (Fcm-2) 2.16 x 10-5 3.36 x 10-5 n 0. 59 0.59 RMF (Ω.cm2) 6.47 x 109 5.73 x 104 Source: The authors.

Table 7. Electrical elements fitted values for T2N3 for 24 and 96 h of immersion in a 0.05M NaCl solution.

Electrical circuit models were used to fit impedance curves. Table 5, Table 6, Table 7 and Table 8 show electrical elements values obtained by fitting for coated and uncoated tin plate at 24 and 96 h of immersion in 0.05 M NaCl solution. In the circuits, capacitances were substituted by constant phase elements (CPE) in order to take into account the non-ideality of the tin plate and 16

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

T2N3

24h

96h

141.0 5.47 x 10-6 0.71 1445 2.66 x 10-5 0.62 1.98 x 104

139.2 4.59 x 10-5 0.68 531.7 4.59 x 10-5 0.57 3.91 x 102

a)

Fitted circuit Re (Ω.cm2) CPEHF (Fcm-2) n RHF (Ω.cm2) CPEMF(Fcm-2) n RMF (Ω.cm2) Source: The authors.

b)

Table 8.

Electrical elements fitted values for T3N3 for 24 and 96 h of immersion in a 0.05M NaCl solution.

T3N3 24h 96h Fitted circuit Re (Ω.cm2) CPEHF (Fcm-2) n RHF (Ω.cm2) CPEMF (Fcm-2) n RMF (Ω.cm2) CPELF (Fcm-2) n RLF (Ω.cm2) Source: The authors.

c) 149.4 3.38 x 10-5 0.73 590.2 3.22 x 10-5 0.64 1.18 x 105

152.0

6.95 x 10-5 0.61 1.21 x 103 5.81 x 10-4 0.44 7.45 x 103

d)

Figure 9 shows the images obtained after 96 hours of EIS in 0.05 M NaCl. At the end of the experiment, red corrosion products were observed on the electrode surface, indicating the formation of iron oxides. The hybrid films obtained with nitric acid (pH=3) and excess TEOS addition (T3N3) showed more red corrosion products on the electrode surface, indicating the formation of iron oxides (Figure 9-d), and hybrid films obtained with nitric acid (pH=3) and with addition of TEOS (level 2, increased concentration of TEOS in the level 1) T2N3 showed corrosion products from darkening indicating degradation of the tin oxides (Figure 9c) when compared to the hybrid films obtained with nitric acid (pH=3) with the lowest amount of TEOS (T1N3) (Figure 9-b), as was expected by the results of the EIS bode diagrams (Figure 7 and Figure 8).

Figure 9. Micrographs obtained after 96 hours of EIS in 0.05 M NaCl: (a) tin plate, (b) T1N3, (c) T2N3 and (d) T3N3.

Source: The authors.

Moreover, the systems showed the best performance in the electrochemical tests, evidencing greater resistance to corrosion even after 96 hours of immersion in saline solution.

Furthermore, this film presented a hydrophobic behavior, evidenced by higher contact angle values and, consequently, lower wettability compared to the T3N3 film.

It also showed a more regular coverage and lower roughness.

The hybrid film siloxane-PMMA obtained with higher addition of TEOS (T3N3) showed greater thickness, however, due to intense densification of the film promoted by the addition of TEOS, there was the formation of cracks, thereby compromising the corrosion resistance. This is linked to the fact that TEOS addition leads to a permeable structure, which is brittle and further contributes to irregular coverage of the surface and to the poor corrosion protection results as shown in the electrochemical characterization.

The present work was carried out with support of CAPES, a Brazilian Government entity focused in human resources formation. The authors also thank the financial support of the Brazilian agencies: CNPq and FAPERGS.

4. Conclusions From the obtained results, it was verified that hybrid siloxane-PMMA films (T1N3 and T2N3) promoted the polarization resistance increase and the corrosion current decrease compared to the uncoated tin plate substrate and T3N3, evidencing the protective effect of these coatings in saline solutions. A shift in open circuit potential to less active potentials was observed after hybrid films application.

17

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

References

[1]

[2]

[3]

[4]

[5] [6]

[7]

[8]

[9]

[10]

[11]

[12] [13]

[14] [15]

[16]

Hu, H., Li, N., Cheng, J. and Chen, L., Corrosion behavior of chromium-free dacromet coating in seawater, Journal of Alloys and Compounds, 472, pp.

219-224, 2009. http://dx.doi.org/10.1016/j.jallcom.2008.04.029 Trabelsi, W., Triki, E., Dhouibi, L., Ferreira, M.G.S. and Montemor, M.F., An electrochemical and analytical assessment on the early corrosion behavior of galvanized steel pretreated with aminosilanes.

Surface & Coatings Technology, 2004.

Sarmento, V.H.V., Dahmouche, K., Pulcinelli, S H., Santilli, C.V. and Craievichi, A.F., Small- angle X-ray and nuclear magnetic resonance study of siloxane-PMMA hybrids prepared by the sol gel process, Journal of Applied Crystallography, 36, pp. 473-477, 2003. http://dx.doi.org/10.1107/S0021889803000384 Bernardo, P.E.M., Camargo, C.D., Costa, N.G., Avaliação do processo de corrosão em folhas de flandres com e sem revestimento orgânico interno, utilizadas em conservas de pêssego, 6° COTEQ – Conferência sobre Tecnologia de Equipamentos, Brasil, Bahia, 2002.

Dóhler, H., Ferenz, M., Herweth, S., Utilização de silanos epóxifuncionais como aditivos de aderência para revestimentos de liberação de silicone – Patente PI0603786-0A, 2007.

Cangialosi, D., Mcgrail, P.T., Emmerson, G., Valenza, A., Calderaro, E. and Spadaro, D., Properties and morphology of PMMA/ABN blends obtained via MMA in situ polymerization through γ-rays, Nuclear Instruments & Methods in Physics Research B, 185, pp. 262266, 2001. http://dx.doi.org/10.1016/S0168-583X(01)00805-9 Shieh, Y.T., Liu, K.H. and Lin, T.L., Effetc of supercritical CO2 on morphology of compatible crystalline/amorphous PEO/PMMA blends, The Journal Supercritical Fluids, 28, pp. 101-112, 2004. http://dx.doi.org/10.1016/S0896-8446(03)00009-3 Langroudi, A.E., Gharazi, S., Rahimi, A. and Ghasemi, D., Synthesis and morphological study on the nanocomposite hydrophilic coatings, Applied Surface Science, 255, pp. 5746-5754, 2009. http://dx.doi.org/10.1016/j.apsusc.2008.12.078 Kozhukharov, S., Kozhukharov, V., Schem, M., Aslan, M., Wittmar, M., Wittmar, A. and Veith, M., Protective ability of hybrid nanocomposite coatings with cerium sulphate as inhibitor against corrosion of AA 2024 aluminium alloy, Progress in Organic Coatings, 73, pp.

95-103, 2012. http://dx.doi.org/10.1016/j.porgcoat.2011.09.005 Fedel, M., Druart, M.E., Oliver, M., Poelman, M., Deflorian, F. and Rossi, S., Compatibility between cataphoretic electro-coanting and silane surface layer for the corrosion protection of galvanized steel, Progress in Organic Coatings, 69, pp. 118-125, 2010. http://dx.doi.org/10.1016/j.porgcoat.2010.04.003 Montemor, M.F. and Ferreira, M.G.S., Electrochemical study of modified bis-[triethoxysilylpropyl] tetrasulfide silane films applied on the AZ31 Mg alloy. Electrochimica Acta, 52, pp. 7486-7495, 2007. http://dx.doi.org/10.1016/j.electacta.2006.12.086 Gu, S., Kondo, T. and Konno, M., Preparation of silica–polystyrene core–shell particles up to micron sizes. J. Colloid Interface Sci., 272, pp. 314-320, 2004. http://dx.doi.org/10.1016/j.jcis.2004.01.056 Luna-Xavier, J., Guyot, A. and Bourgeat-Lami, E., Synthesis and characterization of silica/poly (methyl methacrylate) nanocomposite latex particles through emulsion polymerization using a cationic azo initiator. J. Colloid Interface Sci., 250, pp. 82-92, 2002. http://dx.doi.org/10.1006/jcis.2002.8310 Sertchook, H. and Avnir, D., Submicron silica/polystyrene composite particles prepared by a one-step sol–gel process. Chem. Mater., 15, pp. 1690-1694, 2003. http://dx.doi.org/10.1021/cm020980h Park, J.U., Kim, W.S. and Bae, B.S., Photoinduced low refractive index in a photosensitive organic-inorganic hybrid material, Journal of Materials Chemistry, 13, pp.

738-741, 2003. http://dx.doi.org/10.1039/b211094f Zhu, A., Shi, Z., Cai, A., Zhao, F. and Liao, T. Synthesis of core-shell PMMA-SiO2 nanoparticles with suspension-dispersionpolymerization in an aqueous system and its effect on mechanical properties of PVC composites, Polymer Testing., 27, pp. 540,547, 2008.

[17] Katsikis, N., Zahradnik, F., Helmschrott, A., Munstedt, H. and Vital, A., Thermal stability of poly(methyl methacrylate)/silica nano- and microcomposites as investigated by dynamic-mechanical experiments, Polym. Degrad. Stab. 92, pp. 1966-1976, 2007. http://dx.doi.org/10.1016/j.polymdegradstab.2007.08.009 [18] Delattre, L., Dupuy, C. and Babonneau, F., Characterization on the hydrolysis and polymerization processes of methacryloxypropyltrimethoxylane, Journal of Sol-gel Science and Thecnology, 2, pp.

185-188, 1994. http://dx.doi.org/10.1007/BF00486238 [19] ATMS - G01 Committee, ASTM G5 - Reference Test Method for Making Potentiodynamic Anodic Polarization Measurements, ASTM International, 4, 2013.

[20] G01 Committee, ASTM G106 - Practice for Verification of Algorithm and Equipment for Electrochemical Impedance Measurements, ASTM International, 2010.

[21] Zhang, X., Wu, Y., He, S. and Yang, D., Structural characterization of sol–gel composites using TEOS/MEMO as precursors, Surf. Coat.

Technol., 201, pp.

6051-6058, 2007. http://dx.doi.org/10.1016/j.surfcoat.2006.11.012 [22] Cairú, A. and Mittal, K.L., UV-Resistant and Superhydrophobic SelfCleaning Surfaces using Sol-gel Process. Ed. Netherlands / Brill; 2009.

[23] Kunst, S.R., Cardoso, H.R.P, Oliveira, C.T., Santana, J.A., Sarmento, V.H.V., Muller, I.L. and Malfatti, C.F., Corrosion resistance of siloxane–poly(methyl methacrylate) hybrid films modified with acetic acid on tin plate substrates: Influence of tetraethoxysilane addition. Applied Surface Science, 298, pp. 1-11, 2014. http://dx.doi.org/10.1016/j.apsusc.2013.09.182 [24] Sakai, R.T., Cruz, F.M.D.L., Melo, H.G., Benedetti, A.V., Santilli, C.V. and Suegama, P.H., Electrochemical study of TEOS, TEOS/MPTS, MPTS/MMA and TEOS/MPTS/MMA films on tin coated steel in 3.5% NaCl solution, Prog. Org. Coatings, 74, pp. 288301, 2012. http://dx.doi.org/10.1016/j.porgcoat.2012.01.001 [25] Suegama, P.H., Sarmento, V.H.V., Montemor, M.F., Benedetti, A.V., Melo, H.G., Aoki, I.V. and et al., Effect of cerium (IV) ions on the anticorrosion properties of siloxane-poly(methyl methacrylate) based film applied on tin coated steel, Electrochimica Acta, 55, pp. 51005109, 2010. http://dx.doi.org/10.1016/j.electacta.2010.04.002 [26] Conde, A. and Damborenea, J., Electrochemical impedance spectroscopy for studying the degradation of enamel coatings, Corros.

Sci., 44, pp. 1555-1567, 2002. http://dx.doi.org/10.1016/S0010938X(01)00149-4 [27] Morales, U.P., Camargo, A.M. and Flóres, J.J.O., Electrochemical Impedance – Interpretation of Typical Diagrams with Equivalent Circuits. DYNA, 77, pp. 69-75, 2010.

S.R. Kunst, possui curso técnico profissionalizante em Química pela Fundação Liberato Salzano Vieira da Cunha. Possui bacharelado em Engenharia Industrial Química. Mestre em engenharia com dedicação exclusiva na Universidade Federal do Rio Grande do Sul, Brasil (PPGEM - Capes 7). Doutoranda acadêmica com dedicação exclusiva na Universidade Federal do Rio Grande do Sul, Brasil (PPGEM - Capes 7), sendo as principais linhas de pesquisas: Elaboração e Caracterização de precursores silanos na proteção do aço galvanizado, flandres e alumínio e elaboração e caracterização de camada de difusão gasosa para células a combustíveis de hidrogênio.

G.A. Ludwig, possui graduação em Engenharia Industrial Mecânica pela Universidade Feevale em 2011. Mestre em Engenharia em 2013, pela Universidade Federal do Rio Grande do Sul, Brasil (PPGEM - Capes 7).

Doudorando acadêmico com dedicação exclusiva pela Universidade Federal do Rio Grande do Sul, Brasil (PPGEM - Capes 7) Principais linhas de pesquisa: Elaboração e caracterização de revestimentos metálicos protetores para substratos metálicos utilizados em alta temperatura (SOFC).

M.R.O. Veja, possui graduação em Chemical Engineering pela Universidad del Atlantico em 2011. Mestranda em Engenharia pela Universidade Federal do Rio Grande do Sul, Brasil (PPGEM - Capes 7).

18

Kunst et al / DYNA 81 (188), pp. 11-19. December, 2014.

C.T. Oliveira, professora e pesquisadora, doutora em engenharia na área de ciênica dos materiais e engenheira metalúrgica, possui experiência na área de tratamento de superfície, principalmente em revestimentos protetores, porosos e não-porosos, para finalidade de proteção contra corrosão, aderência de tintas, uso como dielétricos em capacitores eletrolíticos, e obtenção de nano-óxidos para aplicação diversificada.

C.F. Malfatti, possui graduação em Engenharia Metalúrgica pela Universidade Federal do Rio Grande do Sul, Brasil com Mestrado e Doutorado em Engenharia - área de concentração Ciência e Tecnologia dos Materiais, pela Universidade Federal do Rio Grande do Sul e pela Université Paul Sabatier. Atualmente é professora e pesquisadora na Universidade Federal do Rio Grande do Sul, Brasil. Tem experiência na área de Engenharia de Materiais e Metalúrgica, com ênfase em eletroquímica, revestimentos inorgânicos e revestimentos compósitos, revestimentos metálicos e corrosão. Atua principalmente no desenvolvimento relacionado aos seguintes temas: nanotecnologia aplicada ao tratamento de superfícies metálicas, tratamento de superfície metálicas para aplicações na área de biomateriais, tecnologias e materiais para conversão e estocagem de energia, revestimentos protetores e caracterização eletroquímica.

Área Curricular de Ingeniería Geológica e Ingeniería de Minas y Metalurgia Oferta de Posgrados    

Especialización en Materiales y Procesos Maestría en Ingeniería - Materiales y Procesos Maestría en Ingeniería - Recursos Minerales Doctorado en Ingeniería - Ciencia y Tecnología de Materiales

Mayor información: Néstor Ricardo Rojas Reyes Director de Área curricular acgeomin_med@unal.edu.co (57-4) 425 53 68

19

Plane geometry drawing tutorial Eduardo Gutiérrez de Ravé, Francisco J. Jiménez-Hornero & Ana B. Ariza-Villaverde Deptartamet of Graphic Engineering, University of Córdoba, Córdoba, Spain, eduardo@uco.es Received: May 18th, 2013.Received in revised form: June 19th, 2013.Accepted: September 25th, 2013.

Abstract A tutorial has been developed with the aim of making geometry drawing teaching easier. It arises as a helpful tool for lecturers and engineering undergraduate students, with the objective of giving support to theoretical and practical lessons in plane geometry drawing.

This easy-to-use tutorial provides users with interactivity, practical videos, self-evaluation, fundamental lessons, and “step by step” teaching due to the different levels of conceptual complexity included in its contents.

Keywords: Plane Geometry, Geometric Design, Geometric constructions, CAGD.

Tutorial de dibujo geométrico Resumen Se ha desarrollado un tutorial para facilitar la docencia del dibujo geométrico. Con la idea de servir de apoyo a las explicaciones teóricas y prácticas de los conceptos correspondientes a los trazados geométricos planos necesarios en la ingeniería. Este tutorial es de fácil manejo y permite interactividad con el usuario, animaciones prácticas, autoevaluaciones, explicaciones amplias del temario y la enseñanza "paso a paso" de los conceptos gracias a los diferentes niveles de complejidad conceptual que incluye en su contenido.

Palabras clave: Geometría métrica, Diseño Geométrico, Construcciones Geométrica, CAGD.

1. Introduction In technical drawing, geometric construction constitutes one of the fundamentals for engineering students. Generally, the main goal of geometry education is to improve spatial skills [1]. Most of this ability is acquired during elementary and high school courses, but some acquisition is left for undergraduate studies [2]. In general, geometry is a discipline which interrelates with other subjects such as computer science, mathematics, computational geometry, computer-aided design and geometric solid modeling [3].

Geometry is necessary for work in various fields such as computer graphics, engineering, architecture, and cartography.

Geometry belongs to school curricula. On the other hand, geometric construction teaching is being abandoned in the training of future engineers due to the incorporation of Computational Geometry (CG) and Dynamic Geometry (DG) teaching, that brings in dynamicity to the traditional ruler-and-compass geometry learning process. During the 1980s several programs were set up for the purpose of simulating geometric constructions carried out with traditional methods [4]. Botana & Valcarce (2002) [5]

introduced “Discover”, a program for learning and teaching geometry that permits the replacement of the traditional ruler-and-compass by electronic substitutes. During the 1990s, DG has been increasingly used for teaching, mainly in high schools, although the traditional Euclidean tools are still being replaced by virtual tools in computers. Current software DG systems, such as Cabri-Géomètre [6], SketchPad [7], Cinderella [8,9] iGeom [10], and Geometry Expert [11-13] present web-based versions, enabling students to use them worldwide through a web browser. Liu et al. (2007) [14] propose a pen-based intelligent dynamic lecture system for geometry teachers. Plane Geometry plays an important role in learning, research and engineering.

Pythagoras is a software simulator for dynamic geometry [15]. The use of Internet and computers can bring great benefits to the teaching of geometry, choosing or creating appropriate programs and methodologies that take advantage of the computer’s positive characteristics [16].

DG programs have proven to be an excellent resource for teachers and students [17]. C.a.R. is a dynamic geometry program simulating compass and ruler constructions on a computer [18]. However, on a computer, much more is possible, and as Hoyles & Noss (2003) [19] have shown,

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 20-25. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.38147

Gutiérrez de Ravé et al / DYNA 81 (188), pp. 20-25. December, 2014.

program structure allows the user to access at each moment to its contents. The subject matter is clearly ordered in chapters, sections, subsections and methods.

Each part is explained on the blackboard, where the user is able to visualize and control the method's execution. c) Practices and exercises: These permit the improvement of theoretical and practical knowledge interactively using a dynamic data input.

Dynamic Geometry systems are ‘‘pedagogic tools finely tuned for the exploration of a mathematical domain’’. DG can be understood as being an alternative to traditional ruler-and-compass geometry, which produces static constructions. However, for engineering students and, specifically, mechanical engineering students (designing manufacturing products) [3] it is necessary to learn the fundamentals and step-by-step processes in geometrical construction. Also, they must acquire a knowledge of Geometry and Drawing, which has not been sufficiently promoted in pre-university or university education during recent years [2]. The use of tools that explain underlying processes in a comprehensive way [20] introduces immersive collaborative learning into geometry education and applies 3D dynamic geometry to make it easier. Some authors have understood that it is necessary to gain prior knowledge of plane geometry drawing in order to learn 3D construction, and to improve spatial skills. Computer-Aided Teaching has been of great benefit to teachers and students [14].

In this work, a computer application called PGDT, acronym of Plane Geometry Drawing Tutorial, has been developed based on computational and dynamic geometry with the aim of facilitating the teaching and learning of plane geometry concepts. Thus, the topics included in PGDT are the following ones: 1. Drawing tools, 2.

Elementary Geometric Constructions, 3. Working with line segments, 4. Angles, 5. Triangles, 6. Regular Polygons, 7.

Tangencies, 8. Quadrilaterals, 9. Circles and connection, 10.

Power, 11. Inversive geometry, 12. Scales, 13. Geometric movements, 14. Homothetic transformation and Similarity (geometry), 15. Affinity.

From a total of 225 methods, 26 of them are explicative and do not require any data introduction. All the topics considered here are useful for engineering and they are mainly focused on geometric constructions.

2.1. Application structure PGDT has two levels: interface and processing, with three and two independents units, respectively. Those levels are described as follows: a) Interface level including the following units i) Application management: used for controlling user operations such as application closing, resizing and restoring. ii) Lessons index: this shows all the contents and methods of PGDT. iii) Videos: Tutorial videos.

At this level, three windows are available to the user: main, indexes and video windows. These are described below. b) Processing level consisting of the next units i) Operations: for videos visualization and lesson-method explanation. ii) Help: On line assessment 2.1.1. Interface level (PGDT interface has three windows. User can introduce or select parameters and receive some messages by means of dialog boxes.

Main window.

In Fig. 1, the main window components can be observed. This window is used as a system communication and is divided into two views: Theory-Property Manager and Blackboard, the first one showing the theoretical explanation of the selected method as well as input data properties. In the second view, the graphic output of the selected method can be observed.

2. Methods PGDT has been developed under the Windows operating system (WOS) for the purpose of providing the user with a visual, practical and easy-to-use tool for the execution of different Graphical methods. Although Windows is not the only operating system offering those features, it has been chosen because it is one of those most used by students in their personal computers. Delphi is the programming language selected for implementing the application. It is a powerful compiler that easily manages windows and icons of work environments running under WOS. In PGDT, the following relevant aspects have been taken into account: i) to visualize the parameters of resulting geometric elements (i.e. coordinates, perimeter, area); ii) to set an error control during the step by step learning process of geometric constructions iii) to describe basic mathematic functions and drawing algorithms. These aspects have been developed in the following contents included in PGDT: a) Learning of drawing tool handling by means of videos that explain step by step the use of the ruler-andcompass. b) Theoretical description of each topic: the on-page

Figure 1. Interface.

Source: The authors.

21

Gutiérrez de Ravé et al / DYNA 81 (188), pp. 20-25. December, 2014.

The program enables the creation of geometric objects, such as points, lines, arcs, circumferences, rectangles, ellipses, by using dynamic measures and the modification of their properties.

In addition, options for saving and printing the theory and the graphical methodology are provided, as well as graphic tools like zoom, gridding, rulers or squared patterns. Three colors are used for geometric object representation on the blackboard, red for input, blue for auxiliary construction elements and green for solutions Finally, PGDT includes on-line help.

PGDT main windows contain the following items: i) Window frame, ii) Application menu, iii) Toolbar I, iv) Blackboard, v) Theory and Property manager, vi) Toolbar II, vii) Console message, viii) Status Bar. • Window frame. The functions of this application facilitate window handling. It is composed of: Title Bar, Menu access (maximization, minimization and restoration of window, movement and resizing and application closing), and fast access to main functions of WOS menu.

• Menu. Includes the following submenus: File (Print and Save in RTF type for text and BMP for blackboard graphics), Draw (this allows graphic object selection), Edition (to copy, cut, paste or delete the graphic objects on the blackboard), Display (with an index showing all the lessons and methods), Videos, Zoom, Rules, Grid, Help on-line and References.

• Toolbar I. Direct access to: forward and backward, index, video, saving, printing, copying, pasting, cutting, deleting, showing grids and rules on the blackboard, graphic object display and zoom tools.

• Blackboard. This is rectangular with the origin of the coordinates in the top left hand corner. Blackboard is used for drawing and showing results.

• Theory and Property manager. Shows theory contents related to the selected method and properties of graphic elements on the blackboard. It is possible to modify properties and geometric parameters of input graphic elements. Theory shows the text, including explicative steps. Property shows characteristics of selected graphic objects, coordinates, color and nomenclature used.

• Toolbar II. This is divided into two sections. In the section on the left the following actions are found: Act/Des, Line, Arc, Rectangle, Ellipse, Circle, Point. In the right hand section the following actions are found: blackboard cleaning, starting method, “backward” and “forward”, and ending method and index.

• Console message. This shows messages when user needs to put data on the blackboard.

• Status bar. This shows the actions in progress.

Index window This window allows the selection of lessons included in PGDT, Fig. 2.

Video window This window shows videos contained in the tutorial (Fig. 3). These videos are focused on geometric constructions by means of rulers (square and bevel).

Figure 2. PGDT Window Index and methods.

Source: The authors.

Figure 3. Video Window.

Source: The authors.

Selecting properties of graphic elements, v) Drawing on the blackboard i) To show a video, the icon access is found in Toolbar I. Moreover, a voice explaining the method shown can be heard. ii) Selecting topic-method, this is activated by pressing icon “Muestra el Índice”. A window emerges containing 15 tutorial topics and their corresponding methods. iii) Explanation of the selected method. The icons “backward” and “forward” are in Toolbar II and permit the user to run through the steps of the method.

It is possible to apply the method by introducing new data through the icon “Limpiar dibujo”. If the selected method is only a theoretical one, these icons will not be shown in Toolbar II. iv) Selecting properties of graphic objects. This option permits users to choose and change properties of input graphic objects that will be shown in red on the blackboard. Only the configurable properties can be modified (Fig. 4)

2.1.2. Processing level In this section, PGDT execution procedure is explained.

Options described are: i) Video showing ii) Selecting topicmethod, iii) Explanation of method selected, iv) 22

GutiĂŠrrez de RavĂŠ et al / DYNA 81 (188), pp. 20-25. December, 2014.

v)

Drawing on the blackboard, by using the mouse.

3. Example of application The background knowledge related to tangency is a relevant geometry subject in technical design because it is involved in many frequent systems such as rod-crank and gears. For this reason, the drawings of tangent lines to circles have been chosen as examples of application.

The first example corresponds to drawing tangent lines from point P exterior to a circle. The proposed method is composed of 6 steps. The required graphic object inputs are completed in the first two steps and they are placed on the blackboard. Thus, the circle is introduced by drawing it on the blackboard.

Coordinates of its center and radius can be modified through the window properties by the keyboard (Fig. 5).

Figure 6. First step of drawing tangent lines from point P exterior to a circle of center O and radius R. Source: The authors.

Figure 4. Properties window.

Source: The authors.

Figure 7. Graphic solution to drawing tangent lines from point P exterior to a circle of center O and radius R. Source: The authors.

The following steps begin by drawing the line linking point P and center O (Fig. 6) and finish by calculating the geometric solutions shown in Fig. 7.

The second application example is shown in Fig. 8. It deals with the calculation of common tangents to two independent circles in 9 steps. Fig. 8a shows the two circles drawn on the board by the user, and Fig. 8b depicts internal and external tangents drawn by the program and the corresponding tangency points

Figure 5. Drawing tangent lines from point P exterior to a circle: variables set up shown in property window.

Source: The authors.

23

Gutiérrez de Ravé et al / DYNA 81 (188), pp. 20-25. December, 2014.

References [1]

[2]

[3]

[4]

[5] [6] [7] [8] [9] [10] [11] [12]

[13] [14] Figure 8. Common tangents to two independent circles.

Source: The authors.

[15]

4. Conclusions [16]

This work introduces an application to be used in graphics engineering teaching and learning. One of the main difficulties in encouraging a wider use of DG programs is creating content and evaluating and guiding students during learning activities. This problem is overcome by using the proposed computer application due to the dynamic data entry corresponding to each exercise, and showing the solution step by step. Thereby, teachers can choose the necessary parameters in an easy way. One way to motivate students to think about their own mental models is to allow them to manipulate variables, by changing their values and trying to observe how they behave. It is essential to fulfil this requirement in this kind of computer tool. The application is able to run on computers with low-processing capabilities thus facilitating its required diffusion through elearning platforms or web.

[17]

[18] [19]

[20]

24

Kaufmann, H. and Schmalstieg, D. Mathematics and geometry education with collaborative augmented reality, Computers & Graphics, 27, pp.

339-345, 2003. http://dx.doi.org/10.1016/S00978493(03)00028-1 Moura, J.G., Branddo, L. O. and Brandao, A. A. F. A Web-based learning management system with automatic assessment resources.

Proceedings of the 37th Annual Frontiers in Education Conference, pp.

766-771, 2007.

Lowther, J. L. and Shene, C. K. Computing with geometry as an undergraduate course: A three-year experience, Proceedings of the 32nd Technical Symposium on Computer Science Education. SIGCSE Bulletin, 33 (1), pp.

119-123, 2001. http://dx.doi.org/10.1145/366413.364558 Ruthven, K., Hennessy, S. and Deaney, R. Constructions of dynamic geometry: A study of the interpretative flexibility of educational software in classroom practice, Computers & Education, 51 (1), pp.

297-317, 2008. http://dx.doi.org/10.1016/j.compedu.2007.05.013 Botanaa, F. and Valcarce, J. L. A dynamic–symbolic interface for geometric theorem discovery. Computers & Education, 38 (1–3), pp.

21-35, 2002. http://dx.doi.org/10.1016/S0360-1315(01)00089-6 Bellemain, F. Conception, realisation and experimentation of software for teaching geometry: Cabri-géométre, PhD. dissertation. LSD2-IMAG Laboratory, Grenoble, France, 1992.

Jackiw, R. N. and Finzer, W. F. The geometer's sketchpad: programming by geometry, in Watch what I do: programming by demonstration, Cambridge, MIT Press, 1993. pp. 293-307.

Kortenkamp, U. Foundations of Dynamic Geometry, PhD. dissertation, ETH, Institut für Theoretische Informatik, Zurich, Switzerland, 1999.

Kortenkamp, U. Foundations of Dynamic Geometry, Journal für Mathematik-Didaktik, 21, pp.

161-162, 2000. http://dx.doi.org/10.1007/BF03338916 Brandão, L. O. iGeom: a free software for dynamic geometry into the web. International Conference on Sciences and Mathematics Education, Rio de Janeiro, Brazil, 2002.

Gao, X. S. Building Dynamic Mathematical Models with Geometry Expert, III. A Geometry Deductive Database, Proceedings of the fourth Asian Technology Conference in Mathematics, pp. 153-162, 1999.

Gao, X. S., Zhu, C. C. and Huang, Y. Building a Dynamic Mathematical Models with Geometry Expert, I. Geometric Transformations, Functions and Plane Curves, Proceedings of the third Asian Technology Conference in Mathematics, pp. 216-224, 1998.

Gao, X. S. Zhu, C. C. and Huang, Y. Building Dynamic Mathematical Models with Geometry Expert, II. Linkages, Proceedings of the third Asian Technology Conference in Mathematics, pp. 15-22, 1998.

Liu, Y., Lin, Q. and Dai, G. PIGP: A Pen-Based Intelligent Dynamic Lecture System for Geometry Teaching in Hui K.-c. et al. Edutainment, LNCS 4469, pp. 381-390, 2007.

Ciobanu, G. and Rusu, D. Pythagoras: An interactive environment for plane geometry. Proceedings of the Intelligent Computer Communication and Processing Conference, pp. 283-286, 2007.

Isotani, S. and Brandão, L. O. An algorithm for automatic checking of exercises in a dynamic geometry system: iGeom. Computers & Education., 51, pp.

1283-1303, 2008. http://dx.doi.org/10.1016/j.compedu.2007.12.004 Sinclair, M. P. Peer interactions in a computer lab: Reflections on results of a case study involving web-based dynamic geometry sketches. The Journal of Mathematical Behavior, 24, pp. 89-107, 2005. http://dx.doi.org/10.1016/j.jmathb.2004.12.003 Grothman, R. C.a.R.: Compass and Rules, 1999. Available at http://www.z-u-l.de/doc_en/index.html Hoyles, C. and Noss, R. What can digital technologies take from and bring to research in mathematics education? in Bishop, A. J. et al.

Second International Handbook of Mathematics Education, 2nd edition, Dordrecht, Kluwer Academic, 2003. pp.

323-349. http://dx.doi.org/10.1007/978-94-010-0273-8_11 Kaufmann, H. and Schmalstieg, D. Designing Immersive Virtual Reality for Geometry Education. Proceedings of the Virtual Reality Conference, pp. 51-58, 2006.

Gutiérrez de Ravé et al / DYNA 81 (188), pp. 20-25. December, 2014.

E. Gutiérrez de Ravé obtained his agronomic engineering PhD (1987) from the University of Córdoba (Spain). He is currently full professor in the Department of Graphic Engineering and Geomatics at the University of Córdoba. His research is focused in computer aided design, curves and surfaces and computer graphics.

F.J. Jiménez-Hornero obtained his agronomic engineering PhD (2003) from the University of Córdoba (Spain). He is currently full professor in the Department of Graphic Engineering and Geomatics at the University of Córdoba. His research is focused in computational fluids mechanics, multifractals and computer graphics.

A.B. Ariza-Villaverde obtained her agronomic engineering PhD (2013) from the University of Córdoba (Spain). She is currently developing her research activity as grant holder in the Department of Graphic Engineering and Geomatics at the University of Córdoba. Her research lines are GIS, axial maps, multifractals and computer graphics.

Área Curricular de Ingeniería Civil Oferta de Posgrados     

Especialización en Vías y Transportes Especialización en Estructuras Maestría en Ingeniería - Infraestructura y Sistemas de Transporte Maestría en Ingeniería – Geotecnia Doctorado en Ingeniería - Ingeniería Civil

Mayor información: John Jairo Blandón Valencia Director de Área curricular asisacic_med@unal.edu.co (57-4) 425 5172

25

Analytical model of signal generation for radio over fiber systems Gustavo Adolfo Puerto-Leguizamón a & Carlos Arturo Suárez-Fajardo b a b

Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. gapuerto@udistrital.edu.co Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. csuarezf@udistrital.edu.co Received: August 31th, 2013. Received in revised form: February 18th, 2014. Accepted: November 12th, 2014.

Abstract This paper presents an analytical model that describes the elements involved in the signal generation feasible to be used in radio over fiber systems for the transport of information. In these systems, the radio frequency carriers are conveyed as optical subcarriers over an optical fiber link in a point-to-point or point-to-multipoint connectivity paradigm. The model is based on the definition of both the electrical field and power at the output of the optical modulator for baseband and radiofrequency modulated signals. By modeling the electrical field, it was found a trade-off between the modulation depths of both signals; as well as the their optimum value in order to assure a good signal quality in reception.

Keywords: microwave photonics; modulation depth; optical modulation; radio over fiber.

Modelo analítico de generación de señales para sistemas radio sobre fibra Resumen Este artículo presenta un modelo analítico que describe los elementos involucrados en la generación de señales proclives de ser usadas en sistemas de radio sobre fibra para el transporte de información. En estos sistemas las portadoras de radiofrecuencia se transportan como subportadoras de canal óptico sobre un enlace de fibra óptica en conexiones punto-punto o punto-multipunto. El presente modelo tiene como base fundamental la definición del campo eléctrico y potencia a la salida del modulador óptico que conjuga una señal modulada en banda base y una señal de radiofrecuencia en cualquier banda. Mediante modelado de la función de campo se encuentra que los índices de modulación presentan una fuerte dependencia entre si y se obtiene su valor óptimo a fin de garantizar una buena calidad para ambas señales en el receptor.

Palabras clave: fotónica de microondas; índice de modulación; modulación óptica; radio sobre fibra.

1. Introducción La demanda de altas tasas de transmisión y gran ancho de banda en redes fijas e inalámbricas se ha incrementado en los últimos años y se pronostica un comportamiento similar en los años venideros. El fabricante de equipos de telecomunicaciones Cisco Systems publicó un reporte basado en mediciones del tráfico actual de Internet en donde se pronostica el tráfico de datos a nivel mundial hasta el año 2017 [1]. El reporte especifica que el tráfico IP global anual superará el umbral del zettabyte alcanzando los 1,4 zettabytes a finales de 2017. En general, el tráfico IP crecerá a una tasa de crecimiento anual compuesta del 23% desde 2012 hasta 2017. El tráfico en el segmento metropolitano superará el tráfico de la red de transporte en 2014 y representará el 58% del tráfico IP total en 2017. Entre 2012 y 2017 en este

segmento de red, el tráfico crecerá casi dos veces más que el tráfico de la red troncal. Del mismo modo se pronostica que para 2017 casi la mitad de todo el tráfico IP se originará en dispositivos que no son computadores personales. En 2012, sólo el 26% del tráfico IP de consumo se originó en dispositivos no-PC, pero para 2017 la proporción de tráfico IP no-PC crecerá hasta un 49%. El tráfico originado en computadores personales crecerá a una tasa compuesta anual de un 14%, mientras que el tráfico originado en tablets, teléfonos móviles, televisores y dispositivos de comunicación máquina-máquina (M2M) tendrán tasas de crecimiento del tráfico generado del 104%, 79%, 24% y 82% respectivamente.

El tráfico de los dispositivos inalámbricos y móviles superará el tráfico de dispositivos cableados en 2016. Así, a corto plazo se prevé un gran aumento del ancho de banda originado desde dispositivos móviles.

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 26-33. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.39715

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.

calidad de un sistema de transmisión RoF considerando requerimientos generados por efectos de propagación de ondas milimétricas además de la evaluación de diferentes formatos de modulación. Asimismo, en [13] se presenta un estudio sobre los efectos de ruido e intermodulación en enlaces RoF que permiten identificar las limitaciones en el desempeño del sistema, también un análisis de la relación portadora-banda lateral [14], así como estrategias para reducir la figura de ruido en enlaces de fibra óptica modulados con señales analógicas [15] y un análisis sobre la optimización del punto de cuadratura de un modulador óptico tipo MZ para aplicaciones RoF [16]. Recientemente se publicó un estudio sobre el análisis de moduladores ópticos tipo MZ y de electro-absorción para aplicaciones de RoF y comunicaciones ópticas de espacio libre [17].

En este artículo se presenta un modelo analítico que a partir de la función de campo obtenida a la salida de un modulador tipo MZ, se modela el comportamiento y prestaciones de un sistema de transmisión RoF en función de la variación de los índices de modulación de las señales que alimentan el modulador óptico. Para ello, en el modelo se considera que la señal de entrada al modulador MZ es la combinación directa de la señal en RF y cualquier señal en banda base transportada en la portadora óptica.

Hoy en día las redes de acceso ópticas pueden proporcionar gran ancho de banda a usuarios fijos. Por otro lado, las redes inalámbricas ofrecen una movilidad deseable a los usuarios, pero no cumplen con los requerimientos de ancho de banda. Además, el tener redes de acceso separadas genera un alto costo en términos de operación y mantenimiento. Todo esto sugiere la integración de dichas redes en una sola infraestructura compartida para la futura distribución de contenidos a usuarios fijos y móviles. En este contexto, los sistemas de Radio sobre Fibra (RoF) y fibra hasta el hogar (FTTH) son dos candidatos serios para consolidarse en el segmento de redes de acceso inalámbrico y fijo respectivamente debido al gran ancho de banda que soportan. Aunque los sistemas RoF se pueden implementar en redes troncales cubriendo largas distancias [2], el mayor punto de atracción radica en el despliegue de sistemas RoF en redes de acceso inalámbrico de banda ancha que permita el transporte y la distribución de las portadoras de RF de cualquier red inalámbrica utilizando una infraestructura de red FTTH. Este hecho sumado al gran ancho de banda ofrecido por la fibra óptica, constituyen las principales razones por las cuales son atractivas las tecnologías de transmisión RoF. En este escenario la principal preocupación y a la vez el desafío se traduce en cómo transmitir las señales de banda base (BB) de las redes FTTH y de radiofrecuencia (RF) de las redes inalámbricas en una sola longitud de onda sobre una sola fibra de una manera costo-efectiva y sobre todo con una calidad aceptable debido a los efectos de degradación que pueden experimentar las señales desde el propio proceso de generación debido al limitado rango dinámico y ancho de banda del modulador óptico.

A la fecha se han realizado diferentes trabajos que demuestran las ventajas de implementar sistemas de transporte de señales de radiofrecuencia sobre enlaces de fibra óptica en diferentes escenarios y aplicaciones. En [3] se discuten las tecnologías habilitantes que permiten el desarrollo de sistemas RoF que incluye esquemas para generación óptica de ondas milimétricas y elevadores de frecuencia, en [4] se presentan varias técnicas para la implementación de redes de acceso ópticas-inalámbricas basadas en procesos de modulación externa realizada en un modulador tipo Mach-Zehnder (MZ). Por otro lado, la modulación simultánea de señales en BB y RF en un modulador óptico se demostró experimentalmente en [5].

En el contexto de estudios sobre las limitaciones en las prestaciones y desempeño en la transmisión de señales analógicas en enlaces de fibra, en [6] se presentó el efecto de intermodulaciones en banda y fuera de banda de canales multiplexados en longitud de onda (WDM) en una red de transporte RoF. Asimismo se han reportado estudios sobre los efectos de las características no-lineales de diodos láser configurados en modulación directa [7] y efectos de distorsión causados por moduladores tipo MZ en enlaces RoF [8,9]. Posteriormente en [10,11] se definieron parámetros para identificar el máximo rango dinámico y las mínimas pérdidas de inserción en un enlace óptico modulado externamente con subportadoras de RF y en [12] se ampliaron los estudios mencionados anteriormente con el reporte y demostración de un análisis de prestaciones de

2. Materiales y métodos Los mecanismos de modulación óptica se basan en alterar alguno de los parámetros de una señal óptica de forma proporcional a una segunda señal eléctrica. El efecto de la señal moduladora sobre la señal modulada se evalúa de forma cuantitativa mediante el índice de modulación, cuanto mayor es dicho índice, mayor es la variación del parámetro modificado en la señal portadora para la misma señal moduladora. El presente estudio está basado en el paradigma de modulación externa en moduladores tipo Mach-Zehnder (MZ), los cuales se basan en el efecto electroóptico lineal o efecto Pockels [18]. Un modulador tipo MZ utiliza una estructura interferométrica implementada sobre un sustrato de Niobato de Litio (LiNbO3), como se muestra en la Fig. 1. En este dispositivo, la luz se divide en partes iguales entre dos guías de ondas paralelas en la superficie del sustrato y se recombina en la salida. La forma en que la variación en el índice de refracción se traduce en un cambio de alguna de las propiedades de la señal óptica: amplitud, frecuencia, fase o polarización, depende de la configuración del dispositivo, en particular, de las corrientes de polarización que controlan el elemento modulador y que tienen incidencia directa en los índices de modulación, relación de extinción y potencia de la señal modulada. Cuando no hay tensión, el desplazamiento de fase relativa es cero y la señal recombinada sale del dispositivo sin atenuación (a excepción de las pérdidas en las guías de ondas). Cuando se aplica una tensión que produce un desplazamiento de fase de π (Vπ es un parámetro propio de cada modulador) entre los dos brazos, la señal se extingue, por lo tanto, el dispositivo actúa como un interruptor controlado por tensión. Estos dispositivos funcionan a velocidades de hasta

27

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.

φ1(RF)+ φ1(dc) A1 Entrada A2

∆φ1 Guía de ∆φ2 onda φ2(RF)+ φ2(dc)

Salida

Figura 1. Modulador MZ con control doble.

Fuente: Los autores

50 Gb/s [19]. Un modulador tipo MZ también permite la aplicación de voltajes de control en ambas guías de onda, lo que resulta en un dispositivo modulador externo de doble control. Esta característica permite la generación de señales en cuadratura arbitrarias [20] que encuentran aplicación en procedimientos de transmisión, tales como modulación de portadora óptica suprimida. Actualmente se cuenta con tecnologías de fabricación de los moduladores ópticos integrados a un nivel actual de madurez muy alto [21,22].

La Fig. 2 muestra la curva de transferencia de un modulador tipo MZ. Esta curva representa la transferencia de potencia óptica del dispositivo, en función del desfase electroóptico inducido sobre la señal óptica.

Este desfase depende a su vez de la tensión de polarización aplicada sobre los electrodos. Como se puede observar, existe una región donde la función de transferencia tiene carácter lineal, y que por consiguiente, resulta óptima para la modulación de la señal eléctrica sobre la portadora óptica. El dispositivo trabaja en régimen lineal cuando se aplica una tensión de polarización tal que se induce un desfase sobre la señal óptica igual a π/2, y además los niveles de tensión aplicados son lo suficientemente pequeños para no distorsionar la señal de información. Las ecuaciones de modelado del modulador MZ tienen como finalidad obtener la expresión del campo eléctrico de la señal óptica a la salida del dispositivo, en función de las diferentes señales de entrada y de algunos parámetros del mismo. Como punto de partida, se formula una primera aproximación que proporciona el campo eléctrico a la salida del dispositivo en función del campo eléctrico a la entrada y de los desfases inducidos por las señales eléctricas aplicadas sobre los electrodos del modulador MZ como consecuencia del efecto electroóptico. La ecuación de campo se establece a partir de la geometría del modulador. En el dispositivo mostrado en la Fig. 1, se aplica una señal eléctrica sobre uno de los dos brazos del interferómetro. Esta señal provoca mediante el efecto electroóptico un cambio de fase sobre la señal óptica que se propaga por dicho brazo. La función de transferencia se puede expresar en términos del coeficiente de acoplo de la propagación del campo y el desfase producido en la señal en ambas ramas del interferómetro de la siguiente forma [13]:



A e  j1   A2 e  j 2   1  E out (t )     E in (t ) 1   



Figura 2. Curva de transferencia de potencia de un modulador MZ. Fuente: Los autores

Donde  es la atenuación de la señal a su paso por el dispositivo y A1 y A2 representan los coeficientes de acoplo de las ramas superior e inferior respectivamente en el modelo del modulador de control doble mostrado en la Fig. 1, φ1, φ2 representa los desfases en cada una de las ramas debido al efecto electroóptico. Estas variables se pueden expresar de la siguiente forma:

1 

 2 

 V ( RF )

 V ( RF )

A1  a  0.5  

(2)

A2  1  a 2

(3)

V1( RF ) (t ) 

V2 ( RF ) (t ) 

 V ( dc )

 V ( dc )

V1( dc )

V2 ( dc )

(4)

(5)

El término ε en (2) representa la diferencia entre los coeficientes de acoplo de propagación de energía de la rama superior e inferior. Para un modulador MZ ideal, ε=0, lo cual indica que la potencia es dividida en partes iguales en las dos ramas. A su vez, los términos de variación de fase Δφ dependen de la tensión de polarización V(dc), este valor se define como aquella tensión que aplicada sobre los electrodos del dispositivo provoca un cambio de fase de 180º sobre la señal óptica que se propaga por la guía de onda como consecuencia del efecto electroóptico. Del mismo modo depende VRF, el cual representa el voltaje necesario en las entradas de RF (superior e inferior) para provocar un cambio de fase de 180º entre los dos brazos del interferómetro. Si se asume un coeficiente de acoplo A1=A2=0.5, y α = 2 (para unas pérdidas de inserción típicas de un modulador MZ de 6 dB), la ecuación (1) se puede expresar como:

(1)

28

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.  1    1    exp j1      exp j2    1   2  2   Eout (t )     Ein (t ) 2  2

En el presente trabajo se modela el sistema descrito en la Fig. 3 utilizando el software de simulación VPI Transmission Maker, en donde los parámetros establecidos de entrada definen una función de campo a la salida del modulador MZ como la descrita en la expresión (11).

(6)

Simplificando se obtiene la expresión general del campo eléctrico a la salida del modulador MZ: 1 E out (t )    E in (t )exp  j1   exp  j 2  4

3. Resultados y discusión

(7)

Con base en la función de campo obtenida para el sistema propuesto, la potencia óptica a la salida del transmisor está dada por:

Finalmente, la expresión de una señal en un sistema radio sobre fibra se puede definir mediante:

V ( RF ) t   c (t )  e (t )  cos(  e t   e )

Pout (t ) 

(8)

Donde c(t) es cualquier señal en banda base con cierta velocidad binaria transportada en la portadora óptica y e(t) es la señal de datos que modula la portadora de RF definida por  e  2f e más una constante de fase αe. El índice de

    3   1  cos c(t )  em(t )   cos    5 Pin     2     8  3  sin  c(t )  em(t )   sin      2   5 

modulación se define a través de la relación entre la amplitud de la señal moduladora y la tensión de desplazamiento de fase Vπ.

IM Señal 

VSeñal V

(9)

Pout (t ) 

   1  sin  c(t )  em(t )    5      1  sin  c(t )  e(t )  cos( e t   e )   5   

(12) En donde representando esta función en series de funciones de Bessel se puede identificar la magnitud de las componentes espectrales. Para tal fin se ha asumido  e = 0 a fin de simplificar la expresión.

Pout t  

Pin 8

          k 1  sin  c(t )    J 0  e(t )   2 (1) J 2 k  e(t )  cos(2kct ) 5   5  5  k 1  

        cos c(t )  2 (1) k J 2 k 1  e(t )  cos((2k  1)ct )   5   k 0 5  

(10)



Pin           1  sin  c(t )    J 0  e(t )   2 J 2  e(t )  cos(2ct )  2 J 4  e(t )  cos(4c t ) 8  5   5  5  5 

    2 J 6  e(t )  cos(6c t )  ...  5  

Que representa la función de campo eléctrico a la salida del modulador. Reemplazando (4) y (5) en (7) y usando (8), con V2(RF)=0 dado que se está asumiendo un modulador MZ de control sencillo y con valores estándar de: V(RF)=5, V(dc)=5 y V1(dc)=V2(dc)=3.75, se obtiene:

        cos c(t )   2 J1  e(t ) cos(ct )   2 J 3  e(t )  cos(3ct )  5   5  5 

       1 Eout t     Ein (t ) exp j  c(t )  em(t )  3.75  exp  j  3.75   4 5 5        Eout t  

Pin 8

Pin 8

La combinación directa de la señal de RF multiplexada en subportadora de RF es el método más directo para la generación de señales en un sistema de radio sobre fibra. De forma directa las señales en BB y RF se mezclan en el dominio eléctrico, posteriormente la señal combinada se convierte al dominio óptico mediante un modulador MZ de control sencillo. El esquema del generador se muestra en la Fig. 3. Como se puede apreciar, las características inherentes de este tipo de generación de señales requiere el uso de dispositivos de microondas necesarios para combinar la señal de RF y la señal en BB. A partir de la expresión de campo eléctrico a la salida del modulador de la ecuación (7) se obtiene:  1   2   1   2  1 E out (t )   2 E in (t ) exp j  cos  2 2 4    

1 * E out t   E out t   2 Pin 1  cos1   2   8 Pin     1  cos c(t )  em(t )  3.75  3.75    8  5 5 

      2 J 5  e(t )  cos(5c t ) 2 J 7  e(t )  cos(7ct )  ...  5  5  

(13)

A partir de la expresión anterior se deduce que la potencia óptica a la salida del modulador MZ tiene componentes frecuenciales en todos los armónicos, pares e impares, de la subportadora de RF, cos(  e t   e ) tal y como se muestra en la Fig. 4, la cual representa el espectro óptico obtenido de la combinación directa de la señal en BB

 j  1  Ein (t )  exp  c(t )  e(t )  cos(et   e )    2  2 5  1  3   cos  c(t )  e(t )  cos(et   e )     2  2 5  

(11) 29

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.

Oscilador RF

Datos a subportadora RF

Señal

Láser

en banda base

Mezclador RF 1

Mezclador RF 2

Señal combinada: banda base y RF

Modulador MZ

Figura 3. Esquema de modulación óptica para generación de señales RoF mediante combinación directa de subportadora RF y banda base.

Fuente: Los autores

y una subportadora de RF en 18 GHz y el efecto que impone la utilización de un valor bajo de índice de modulación de BB (IMBB=0.1) Fig. 4(a) y un valor alto del mismo (IMBB=0.9) Fig. 4(b) en donde se puede apreciar como al aumentar el índice modulación de BB, los niveles de potencia de dicha señal aumentan mientras que la señal de RF disminuye. La subportadora de 18 GHz se seleccionó de forma arbitraria y a nivel de simulación no genera ningún cambio en los resultados, no obstante, en la práctica este valor debe escogerse cuidadosamente debido a las restricciones que impone sobre el filtro óptico en el proceso de detección de las señales. Con este método de generación, el espectro óptico del paquete óptico es de doble banda lateral con frecuencia central en la longitud de onda del láser y con una separación de las bandas laterales equivalente a la frecuencia RF. Esto puede provocar serias limitaciones debido al efecto de desvanecimiento por dispersión si la detección de la señal de RF se realiza directamente junto con la portadora óptica.

Por otro lado, el modulador MZ de control sencillo no permite controlar el chirp sobre la modulación de BB, ocasionando posibles problemas en redes de larga distancia con dispersión acumulada y velocidades binarias de la señal de banda base superiores a 10 Gb/s si no se realiza regeneración en los nodos de la red. En la Fig. 4 se observa que los índices de modulación tienen una fuerte dependencia entre sí, esta dependencia afecta a la señal en BB ya que la corriente fotodetectada es proporcional a la potencia óptica que llega al receptor y como se puede ver en (13), la potencia óptica de salida incluye un término de señal de RF J0(e(t)π/5).

Este componente se puede modelar como ruido a la señal en BB, por lo tanto, cuanto mayor sea el valor de la señal de RF, más interferencia pasa a la señal en BB y por eso sufrirá una mayor distorsión.

(a)

(b)

Figura 4. Espectro óptico de una señal multiplexada en subportadora para diferentes valores de índice de modulación de carga. (a) IMc=0.1. (b) IMc=1. Fuente: Los autores 30

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.

En este contexto, la calidad de la señal de RF decrece al aumentar el índice de modulación de la señal en BB ya que la modulación de dicha señal se hace en los extremos de la curva de transferencia del modulador y al aumentar el índice de modulación de la señal en BB la subportadora de RF se acerca cada vez más a los extremos de dicha curva de transferencia donde la respuesta es menos lineal y por lo tanto su amplitud disminuye. Para evaluar el comportamiento del sistema y comprobar la calidad de las señales generadas, la función de campo eléctrico a la salida del modulador definida en (11) se evaluó para diferentes niveles de amplitud de la señal en BB y la señal de RF para un mismo valor de Vπ en un proceso de detección directa del campo eléctrico Otros mecanismos de detección pueden implementar técnicas de filtrado óptica con técnicas de sintonización como las descritas en [23]. Con este procedimiento se busca establecer diferentes índices de modulación para ambas señales y modelar el comportamiento general del sistema en función de dicho parámetro. La evaluación de la calidad de la señal generada se realiza mediante el factor de calidad Q, el cual se define como:

Q

m1  m0 1   0

Figura 5. Modelo del comportamiento del factor de calidad de la señal en banda base en función de su índice de modulación para varios niveles de señal de RF. Fuente: Los autores

Así, la fotocorriente detectada es proporcional a la potencia óptica que llega al receptor y como se discutió anteriormente, la señal de RF actuaría como ruido sobre la señal en BB. De la misma manera, la Fig. 6 muestra el modelado del factor de calidad de la señal de RF en función tanto de su índice de modulación como el del establecido para la señal en BB. Dicho modelado representa una función monótona decreciente con respecto a IMBB y aumenta al incrementarse IMRF. En la Fig. 6 se puede apreciar que para un IMRF constante, cuando IMBB aumenta, el factor de calidad de la señal de RF disminuye.

Este comportamiento se debe a que la modulación de la etiqueta se hace en los extremos de la curva de transferencia del modulador, ver Fig. 2, de tal forma que al aumentar IMBB, la señal de RF se acerca cada vez más a los extremos de dicha curva donde es menos lineal. Como consecuencia del achatamiento de los extremos de la curva del modulador, la amplitud de la señal de RF modulada disminuye.

Así, la fotocorriente detectada es proporcional a la potencia óptica que llega al receptor y como se discutió anteriormente, la señal de RF actuaría como ruido sobre la señal en BB. De la misma manera, la Fig. 6 muestra el modelado del factor de calidad de la señal de RF en función tanto de su índice de modulación como el del establecido para la señal en BB. Dicho modelado representa una función monótona decreciente con respecto a IMBB y aumenta al incrementarse IMRF. En la Fig. 6 se puede apreciar que para un IMRF constante, cuando IMBB aumenta, el factor de calidad de la señal de RF disminuye. Este comportamiento se debe a que la modulación de la etiqueta se hace en los extremos de la curva de transferencia del modulador, ver Fig. 2, de tal forma que al aumentar IMBB, la señal de RF se acerca cada vez más a los extremos de dicha curva donde es menos lineal. Como consecuencia del achatamiento de los extremos de la curva del modulador, la amplitud de la señal de RF modulada disminuye.

(14)

Donde mi y σi representan la potencia media y desviación típica de potencia respectivamente del bit i en el instante de decisión en el receptor. La Fig. 5 muestra el comportamiento general que representa el factor de calidad Q de la señal en BB en función de su índice de modulación para diferentes valores del índice de modulación de la señal de RF. Se observa que el factor Q de la señal en BB aumenta cuando IMBB crece. Esto se debe a que cuando IMBB es mayor, el nivel de señal de BB crece y por lo tanto la señal detectada será mayor. También se observa que para un IMBB constante, al incrementar el valor de IMRF se reduce el factor Q de la señal en BB.

Figura 5. Modelo del comportamiento del factor de calidad de la señal en banda base en función de su índice de modulación para varios niveles de señal de RF. Fuente: Los autores 31

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.

4. Conclusiones Este artículo presentó un modelo analítico que representa el proceso de generación de señales para sistemas Radio sobre Fibra empleando un modulador óptico tipo MZ de control sencillo. El modelo describe el proceso de modulación óptica de señales multiplexadas eléctricamente en banda base y en radiofrecuencia. Las señales en banda base se transmiten en la portadora óptica y las señales de radio frecuencia se transportan como subportadoras de la portadora óptica en una frecuencia de la banda de microondas. Este paradigma de transmisión de señales es la base fundamental para las redes de nueva generación ya que permite establecer estrategias de convergencia entre redes fijas y redes inalámbricas i.e. redes móviles celulares que distribuyen sus portadoras entre diferentes estaciones base usando recursos físicos de una red FTTH. Con base en la función de transferencia de la configuración interferométrica de modulador tipo MZ, se derivaron las expresiones de campo eléctrico y potencia óptica a la salida del sistema cuando por uno de los electrodos del modulador se aplica una señal combinada de banda base y radiofrecuencia. Asimismo, al variar los índices de modulación de dichas señales en la función de campo y bajo un mecanismo de detección directa, se modeló el comportamiento del sistema y se encontraron los valores óptimos para los índices de modulación de la señal en banda base y en radiofrecuencia. En este contexto, se encontró que los índices de modulación de ambas señales tienen una fuerte dependencia entre ellos reflejándose dicho comportamiento en el factor de calidad de la señal y que presenta un mínimo de penalización entre ellas cuando el índice de modulación de la señal en banda base (IMBB) está entre 0.25 y 0.65 y el índice de modulación de la señal de RF (IMRF) se encuentra entre 0.2 y 0.4.

Figura 6. Modelo del comportamiento del factor de calidad de la señal de RF en función de su índice de modulación para varios niveles de señal en banda base.

Fuente: Los autores

Referencias [1] Figura 7. Factor de mérito normalizado de los valores del factor de calidad Q de la señal en BB y RF. Fuente: Los autores

[2]

Finalmente, con el propósito de encontrar el punto óptimo del valor de los índices de modulación, se define un factor de mérito normalizado del factor de calidad Q como el producto de IMBB e IMRF. La Fig. 7 representa el factor de mérito normalizado de los índices de modulación.

En este contexto se tomará como referencia un valor de factor de mérito normalizado entre el 80% y 100% para determinar los rangos de índices de modulación óptimos en el proceso de generación de señales RoF. En la Fig, 7 se observa que el rango de valores óptimos para IMBB comprende valores entre 0.25 y 0.65, con valores de IMRF entre 0.2 y 0.4 tal y como se muestra en la zona definida por el recuadro de la Fig. 7. Nótese que el punto óptimo de modulación para la señal en BB y RF se obtiene cuando IMBB=IMRF=0.4.

[3]

[4]

[5]

[6]

32

Cisco Systems, Cisco visual networking index: Forecast and methodology, 2012-2017., [Online], [Date of reference: August 31th of 2013], Available at: http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537 /ns705/ns827/white_paper_c11-481360.pdf Marciniak, M., Towards broadband global optical and wireless networking, Proceedings of 11th Management Committee Meeting of COST Action 273, Towards Mobile Broadband Multimedia Networks, pp. 13-16, 2004.

Zhensheng, J., Jianjun, Y., Georgios, E. and Gee-Kung, C., Key enabling technologies for optical–wireless networks: Optical millimeter-wave generation, wavelength reuse, and architecture, J. Lightwave Tech., 25 (11), pp.

3452-3471, 2007. http://dx.doi.org/10.1109/JLT.2007.909201 Gee-Kung, C., Jianjun, Y. and Zhensheng, J., Architectures and enabling technologies for super-broadband radio-over-fiber opticalwireless access networks, Proceedings of IEEE International Topical Meeting on Microwave Photonics, pp. 24-28, 2007.

Chun-Ting, L., Jason, C., Peng-Chun, P., Cheng-Feng, P., Wei-Ren, P., Bi-Shiou, C. and Sien, C., Hybrid optical access network integrating fiber-to-the-home and radio-over-fiber systems. IEEE Photon Technol. Lett., 19 (8), pp. 610-612, 2007. http://dx.doi.org/10.1109/LPT.2007.894326 Castleford, D., Nirmalathas, A., Novak, D. and Tucker, R., Optical crosstalk in fiber-radio WDM networks. IEEE Trans. Microw.

Theory Tech., 49 (10), pp.

2030-2035, 2001. http://dx.doi.org/10.1109/22.954826

Puerto-Leguizamón & Suárez-Fajardo / DYNA 81 (188), pp. 26-33. December, 2014.

[7]

[8] [9] [10]

[11] [12] [13]

[14]

[15] [16]

[17]

[18]

[19]

[20] [21]

[22]

[23]

Mizuguti, H., Okuno,T., Komaki, S. and Morinaga, N., Performance analysis of optical fiber link for microcellular mobile communication systems. IEICE Trans. Electron., E76-C, pp. 271278, 1993.

Way, W., Optical fiber-based microcellular systems: An overview.

IEICE Trans. Commun., E76-B (9), pp. 1091-1102, 1993.

Cox, C., High-performance fiber-optic links for microwave applications, Proceeding of IEEE MTT-S Int. Microw. Symp., pp.719-722, 1993.

Ackerman, E., Wanuga, S., Kasemset, D., Daryoush, A. and Samant, N., Maximum dynamic range operation of a microwave external modulation fiber-optic link. IEEE Trans. Microw. Theory Tech., 41 (8), pp. 1299-1306, 1993. http://dx.doi.org/10.1109/22.241670 Sabido, D. and Kazovsky, L., Dynamic range of optically amplified RF optical links. IEEE Trans. Microw. Theory Tech., 49 (10), pp.

1950-1955, 2001. http://dx.doi.org/10.1109/22.954813 Sabella, R., Performance analysis of wireless broadband systems employing optical fiber links, IEEE Trans. Commun., 47(5), pp.

715-721, 1999. http://dx.doi.org/10.1109/26.768765 Smith, G.H., Novak, D. and Ahmed, Z., Overcoming chromaticdispersion effects in fiber-wireless systems incorporating external modulators. IEEE Trans. Microw. Theory Tech., 45 (8), pp. 14101415, 1997. http://dx.doi.org/10.1109/22.618444 Lim, C., Attygalle, M., Nirmalathas, A., Novak, D. and Waterhouse, R., Analysis of optical carrier-to-sideband ratio for improving transmission performance in fiber-radio links. IEEE Trans. Microw.

Theory Tech., 54 (5), pp.

2181-2187, 2006. http://dx.doi.org/10.1109/TMTT.2006.872809 Karim, A. and Devenport, J., Noise figure reduction in externally modulated analog fiber-optic links. IEEE Photon. Technol. Lett., 19 (5), pp. 312-314, 2007. http://dx.doi.org/10.1109/LPT.2007.891591 Zongjie, H Zhang, X., Shilie Z., Xiaofeng, J. and Hao, C., Any bias point control of mach-zehnder electrooptic modulator and its applications in optimization of radio-over-fiber links, Proceedings of 2011 International Topical Meeting on Microwave Photonics, & Microwave Photonics Conference, Asia-Pacific, MWP/APMP, pp.

218-221, 2011.

Prabu, K., Bose, S. and Kumar, D.S., Analysis of optical modulators for radio over free space optical communication systems and radio over fiber systems, Proceedings of 2012 Annual IEEE India Conference (INDICON), pp.

1176-1179, 2012. http://dx.doi.org/10.1109/INDCON.2012.6420795 Cho, H.R., Shin, M.J., Han, S.H. and Wu, J.W., Mach–Zehnder interferometer measurement of the Pockels effect in a poled polymer film with a coplanar electrode structure. Applied Physics Lett., 69 (25), pp. 3788-3790, 1996. http://dx.doi.org/10.1063/1.116999 Yamada, E., Shibata, Y., Watanabe, K., Yasui, T., Ohki, A., Mawatari, H., Kanazawa, S., Iga, R. and Ishii, H., Demonstration of 50 Gbit/s 16QAM signal generation by novel 16QAM generation method using a dual-drive InP Mach-Zehnder modulator, Proceedings of Optical Fiber Communication Conference and Exposition (OFC/NFOEC), pp. 1-3, 2011.

Ho, K., Generation of arbitrary quadrature signals using one dual drive modulator. IEEE/OSA J. Lightwave Technol., 23 (2), pp. 764770, 2005. http://dx.doi.org/10.1109/JLT.2004.838855 Kikuchi, N., Yamada, E., Shibata, Y. and Ishii, H., High-Speed InPbased Mach-Zehnder modulator for advanced modulation formats, Proceedings of IEEE Compound Semiconductor Integrated Circuit Symposium (CSICS), pp. 1-4, 2012.

Kaiser, R., Velthaus, K. O., Brast, T., Gruner, M., Hamacher, M., Hoffmann, D. and Schell, M., Medium and large scale MachZehnder modulator ICs on InP for fabrication of advanced transmitters, Proceedings of 14th International Conference on Transparent Optical Networks (ICTON), pp. 1-4, 2012.

Aguiar, M., Gómez, J. y Torres, P., Modelamiento térmico y vibratorio de una cápsula para sensores de fibra óptica adaptables a mediciones en sistemas eléctricos de potencia. DYNA, 76 (157), pp.

243-250, 2009.

en 2008 e investigador posdoctoral en el Instituto de Telecomunicaciones y Aplicaciones Multimedia de la misma universidad hasta 2011. Desde 2012 es Profesor Asistente de la Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. A la fecha ha publicado más de 40 artículos en revistas y congresos internacionales en el campo de redes ópticas, es par evaluador de Colciencias y de las revistas IEEE Journal on Lightwave Technologies, IEEE Photonic Technology Letters y Optics Express. Sus intereses de investigación incluyen sistemas de radio sobre fibra, networking óptico y redes de acceso ópticas.

C.A. Suárez-Fajardo, es Ing. Electrónico por la Universidad Distrital y Licenciado en matemáticas por la Universidad Pedagógica Nacional, Bogotá, Colombia. Inicia sus estudios doctorales en 2002, para lo cual se integra como investigador adscrito al grupo de radiación electromagnética (GRE) de la Universidad Politécnica de Valencia, España. En el 2003 obtiene el título de especialista en Telecomunicaciones, en el 2004 obtiene el título de MSc.en Telecomunicaciones y el de Dr. en Telecomunicaciones en 2006 por la Universidad Politécnica de Valencia, España. Actualmente ocupa el cargo de Profesor Titular en la Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. Es autor de más de 40 artículos en revistas indexadas y en congresos internacionales. Es par evaluador de proyectos de Colciencias y de revistas tales como: Revista Ingeniería de la Universidad de Antioquia, Revista Ingeniería de la Universidad Javeriana, Revista Ingeniería y Desarrollo de la Universidad del Norte, revista Chilena de Ingeniería (Ingeniare) y Journal of Antennas and Propagation (IJAP).

Área Curricular de Ingeniería Eléctrica e Ingeniería de Control Oferta de Posgrados



Maestría en Ingeniería - Ingeniería Eléctrica

Mayor información: Javier Gustavo Herrera Murcia Director de Área curricular ingelcontro_med@unal.edu.co (57-4) 425 52 64

G.A. Puerto-Leguizamón, es Ing. de Telecomunicaciones. En 2003 se vinculó al Grupo de Comunicaciones Ópticas y Cuánticas de la Universidad Politécnica de Valencia, España. Dr. en Telecomunicaciones 33

Recycling of agroindustrial solid wastes as additives in brick manufacturing for development of sustainable construction materials Lisset Maritza Luna-Cañas a, Carlos Alberto Ríos-Reyes b & Luz Amparo Quintero-Ortíz c a

Escuela de Geología, Universidad Industrial de Santander, Bucaramanga, Colombia, lissetluna1829@hotmail.com b Escuela de Geología, Universidad Industrial de Santander, Bucaramanga, Colombia, carios@uis.edu.co c Escuela de Ingeniería Metalúrgica y Ciencia de Materiales, Universidad Industrial de Santander, Bucaramanga, Colombia, luzquint@uis.edu.co Received: August 31th, 2013. Received in revised form: March 13th, 2014. Accepted: October 25th, 2014.

Abstract Accumulation of unmanaged agroindustrial solid wastes especially in developing countries has resulted in an increased environmental concern. Recycling of such wastes as a sustainable construction material appears to be a viable solution not only to the pollution problem but also an economical option to design green buildings. This paper studies the application of several agroindustrial wastes in brick manufacturing, which include cocoa shell, sawdust, rice husk and sugarcane. First, the mineralogical and chemical composition of the wastes and clayey soil were determined. Next, bricks were fabricated with different quantities of waste (5%, 10% and 20%). The effect of adding these wastes on the technological behavior of the brick was assessed by compressive strength, flexural strength and durability tests. Based on the results obtained, the optimum amounts of agroindustrial waste to obtain bricks were mixing 10% of cocoa shell and 90% of clayey soil. These percentages produced bricks whose mechanical properties were suitable for use as secondary raw materials in the brick production.

Keywords: agroindustrial wastes; bricks; recycling; construction material; engineering properties.

Reciclaje de residuos sólidos agroindustriales como aditivos en la fabricación de ladrillos para el desarrollo sostenible de materiales de construcción Resumen La acumulación de residuos sólidos agroindustriales no administrados especialmente en los países en vías de desarrollo ha dado lugar a una creciente preocupación ambiental. El reciclaje de tales residuos como un material de construcción sostenible parece ser una solución viable no sólo al problema de la contaminación, sino también una opción económica para diseñar edificios verdes. El presente trabajo estudia la aplicación de varios residuos agroindustriales en la fabricación de ladrillos, que incluyen cáscara de cacao, aserrín, cáscara de arroz y caña de azúcar. En primer lugar, se determinó la composición mineralógica y química de los residuos y del suelo arcilloso. A continuación, los ladrillos se fabricaron con diferentes cantidades de residuos (5%, 10% y 20%). El efecto de la adición de estos residuos en el comportamiento tecnológico del ladrillo se evaluó mediante ensayos de resistencia a la compresión, resistencia a la flexión y durabilidad. Con base en los resultados obtenidos, las cantidades óptimas de residuos agroindustriales para obtener ladrillos fueron mezclando 10% de cáscara de cacao y 90% de suelo arcilloso. Estos porcentajes producen ladrillos cuyas propiedades mecánicas eran adecuadas para su uso como materias primas secundarias en la producción de ladrillos.

Palabras clave: residuos agroindustriales; ladrillos; reciclaje; materiales de construcción; propiedades de ingeniería.

1. Introduction Agroindustry generates considerable quantities of solid wastes which are rich in organic matter and could constitute new materials for value added products. Because of their biodegradable nature, several agroindustrial residues can be

safely disposed of; however, the amount of discharged residues is expected to increase dramatically in the future.

In Colombia, they are mostly underutilized, untreated and thus in most cases disposed off by unplanned landfilling.

Due to increasing landfill costs, stricter environmental regulation and current interest in sustainable development,

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 34-41. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.39717

Luna-Cañas et al / DYNA 81 (188), pp. 34-41. December, 2014.

10 10000

20 20000

40 40000

50 50000

Osb Osb

Osb Osb

Osb Osb

Osb Osb

30 30000

Osb Osb Osb Osb

Hay Hay

Hay

Dnp Dnp

Hay

Qtz Qtz Hay Hay Mnt Mnt

Mnt Mnt

00

Hay Hay

Intensity (arbitrary units)

Qtz Qtz

the effective recycling of agroindustrial residues for the manufacture of bricks of greater value to mitigate the depletion of resources and environmental impact has become an increasing concern in recent years. Traditional construction materials, including bricks, are being produced from existing natural resources, which is destroying the environment due to their continuous exploration and depletion. On the other hand, large concentrations of toxic substances are emitted into the atmosphere during the manufacturing process of construction materials, which has a strong negative environmental impact. Consequently, major changes regarding the conservation of resources and recycling of wastes by proper management are taking place in our ways of living and working [1]. Many authorities and investigators are lately working to have the privilege of reusing the wastes in environmentally and economically sustainable ways [2]. Therefore to satisfy the continuously increasing demand, researchers are incorporating solid wastes for the manufacturing of novel construction materials to develop sustainable alternative solutions. From the standpoint of energy saving and conservation of natural resources, the use of alternative constituents in construction materials is now a global concern [1]. It well known that almost all the buildings comprise a structure of reinforced concrete and facade made of brick walls [3]. Attempts have been made to incorporate several industrial wastes in the manufacturing of bricks, including paper-making pulp [4-5], cigarette butts [6], steel slag [7-8], fly ash [9-10], water treatment sludge [11-12], thin film transistor liquid crystal display optical glass [13], processed tea [14], sawdust [1516], cotton waste [17], polystyrene fabric [18], rubber [19], granite sludge [20], limestone powder waste [15,17] and waste foundry sands [21]. These studies demonstrated that the use of waste materials can save energy and enhance brick quality. The purpose of our research was to develop a comparative study on the use of several agroindustrial wastes (cocoa shell, sawdust, rice husk and sugarcane) in the manufacturing of bricks. The experimental study includes a laboratory simulation of the industrial brickmaking process to assess technological feasibility, and technological trial to validate prior results.

60 60000

70 7000

o2θ(CuKα)

Figure 1. XRD pattern of the clay-rich material. Mnt, montmorillonite; Hay, halloysite; Qtz, quartz; Dnp, donpeacorite; Osb, osbornite.

Source: The authors.

clayey soil and mixed in a mortar to obtain good homogenization. To enable comparative results, three samples per series were prepared for the tests. The necessary amount of water was added to the samples to obtain adequate plasticity and absence of defects, mainly cracks, during the semi-dry molding stage, using a mold of 50 x 60 x 90 mm. Agroindustrial solid waste-free mixtures were also prepared as a control. Therefore, ASWBs with a cross section of 50x 60 mm and a length of 90 mm were obtained. Samples were fired in a laboratory furnace at 800 o C for 4 h. Samples were then cooled to room temperature by natural convection inside the furnace. The shaped samples were designated as C (control) for the bricks without agroindustrial solid waste and ASWxB for the mixtures, where ASW and x denotes the type of residue incorporated (CS - cocoa shell; SD - sawdust; RH - rice husk, and SC - sugarcane) and its content (%) in the clay matrix, respectively.

2.2. Properties of materials Qualitative determination of major crystalline phases present in the clay-rich material was achieved by using a powder X-ray diffractometer (PhilipsPW1710), operating in Bragg–Brentano geometry with Cu Kα radiation (k = 1.5406 Å), 40 kV and 40 mA, and secondary monochromation. Data was collectedin the 2–70º 2θ range (0.02º step size). The crystalline patterns were compared with the standard line patterns from the Powder Diffraction File database supplied by the International Centre for Diffraction Data (ICDD), with the help of the Joint Committee on Powder Diffraction Standards (JCPDS) files for inorganic compounds. The major crystalline phases found in the clay-rich material are quartz, montmorillonite, halloysite, donpeacorite and osbornite (Fig. 1).

The morphology of the agroindustrial wastes (Fig. 2) was examined by environmental scanning electron microscopy (ESEM) (FEI Quanta 650), under the following analytical conditions: magnification = 183x, WD = 17.217.5, HV = 20.0 kV, spot = 3.0, mode SE, detector LFD.

2. Experimental procedure 2.1. Preparation of the samples The materials used for the manufacture of agroindustrial solid waste-based bricks (ASWBs) consisted of raw clayrich material, cocoa shell, sawdust, rice husk and sugarcane.

The raw clay-rich material used in this study was supplied

by ERGO Durán & García Brick Company Ltda., from the brick plant in Girón, Santander (Colombia).The clayey soil is currently used by this company to make fired bricks of different shapes and sizes with dimensional tolerances that conform to ASTM Standards. Agroindustrial solid wastes were obtained from the supply and storage center (Centroabastos), Santander (Colombia). Their use should be promoted as an appropriate and alternative low cost but high quality building technology. Calculated amounts of cocoa shell, sawdust, rice husk and sugarcane were added to the 35

Luna-Cañas et al / DYNA 81 (188), pp. 34-41. December, 2014.

Figure 2. SEM images of the agroindustrial wastes.

Source: The authors.

The particle size distribution of the clay-rich material was obtained by Niño et al. [22], combining sieve and hydrometer analyses according to the standards ASTM C136-06 [23] and ASTM D1140-00[24], revealing that it is mainly composed of sand particles (87.80%), with 13.63% of fine particles and 1.57% of gravel particles, corresponding to a sandy clay soil. Niño et al. [22] also reported the Atterberg’s limits of the clay-rich material according to the standard ASTM D4318-10 [25], with the following results: liquid limit of 35%, plastic limit of 17% and plasticity index of 18%.

Figure 3. Experimental scheme followed for manufacturing ASWBs. Source: The authors.

The units of ASWBs were manufactured with cuboidal shape and standard size (60 x 50 x 90 mm). The specimens were dried at 100 oC for 24 hours, removed from the molds and were fired in a (TERRIGENO) furnace at 800 oC. The fired samples were tested for compressive strength, flexural strength, and Mg2SO4 and H2SO4 attack. All tests were carried out according to ASTM standards and the results reported are the mean of three values.

Fig. 4 illustrates the preparation of the ASWBs. In order to obtain comparable results, a total of 12 ASWBs (3 for each mixture) were prepared for testing four different series.

The shape and size tolerances have been respected. ASWBs were cured for 28 days under the following environmental conditions: average temperature of 25 oC and relative humidity of 80%. Too much clay will cause cracks in the blocks while too much sand will cause the blocks to crumble. The suitable soil must contain the right proportions of sand, silt, clay and water.

2.3. Sample preparation, mix compositions and testing Fig. 3 illustrates a block diagram showing the methodology followed in the manufacturing of the ASWBs during their study. The raw clay-rich material was naturally dried during 3 weeks under the following environmental conditions: average temperature of 24 o C and relative humidity of 83.5%. Then, it was subjected to the following steps: rough crushing with a Retsch Jaw Crusher BB200 to ~2 mm, milling with a Retsch RM100 mortar grinder mill to clay particle size and sieving with a 200 mesh Ro-Tap sieve shaker (using 4, 10, 20, 40, 60, 100 and 200 mesh series).

The agroindustrial residues were dried for 24 hours under the direct sunlight to remove the excess moisture. Then, they were cut in fragments of different average dimensions.

In order to determine the effect of the addition of agroindustrial residues on the engineering properties of ASWBs. Different amounts of ASWBs (5%, 10% and 20%) were chosen for the mix design of the ASWBs. The mix proportions were prepared based on the dry weights of the ingredients. The quantities of the dry materials obtained from the mix design were measured in each case with the aid of a weighing balance. First, the dry materials were mixed by hand with a spade on a hard surface until they reached a uniform color. Then, water was added and mixing continued until a homogeneous mixture was obtained. The resultant mixtures were compacted manually in appropriate molds using predetermined masses corresponding to the maximum density (found from standard compaction tests).

Figure 4. Stages during preparation of ASWBs. (a) Agroindustrial solid waste (cocoa shell). (b) Adding cocoa shell to the clayey soil. (c) Mix of materials. (d) Molding process. (e) 28-days cured ASWBs. (f) Sintered ASWBs. Source: The authors.

36

Luna-Ca単as et al / DYNA 81 (188), pp. 34-41. December, 2014.

and the ASWBs. A loading steel roller identical to the two described above was set on top of the ASWBs. The load was applied via a steel roller, identical to those described above, directly onto the ASWBs. The maximum load until the occurrence of the first crack was recorded as flexural strength. Upon crack occurrence, the strength decreased.

This test provides values for the modulus of rupture (MR) of the ASWBs. MR can be calculated according to Varela et al.

[31] using Eq. (1).

The swelling/shrinkage behavior of the 28-days cured ASWBs was determined as follows. Immediately after the fabrication of the ASWBs, their dimensions were recorded and at the end of the 28-day curing period, a record of their dimensions was also taken. There was no significant dimensional or volume increase in any of the ASWBs. No defects such as cracks and bloating were observed after firing. However, a texture characterized by black cores are developed after firing, which can be attributed to organic matter that is not completely burned during firing [26-27].

In general, the color of the fired samples was reddish, which is similar to that observed in the formulas without wastes, although somewhat lighter as the proportion of waste increases. Engineering tests were conducted in a computerized device for mechanical assays according to the ASTM C67-11 standard [28]. A Universal Testing Machine (MTS 810) with a maximum load of 500000 N was used in the testing procedure, taking into account its accuracy (0.01), flexibility, high performance, and innovative standard features; large test space to accommodate standard, medium and large size specimens, grips, fixtures and environmental subsystem, and environmental chamber dimension: 500 x 255 x 350 mm. Data were recorded automatically to the computer system which the user can manipulate the collected results. The compressive strength test was conducted with a crosshead speed of 0.5 mm/min.

This test was performed according to ASTM D 2166-00e1 [29]. The test was carried out as follows: ASWBs were placed between two steel bearing plates (on the top and on the bottom), which were identical (length, width and thickness were respectively 100 x 40 x 5 mm). The load strain reading at failure was recorded; it was the maximum load the specimen could carry in compression. The threepoint bending flexural strength test was conducted with a cross head speed of 0.2 mm/s and a distance between the supports of 90 mm. This test was performed according to ASTM D 1635-00 [30]. The procedure performed on the ASWBs was as follows: two cylindrical steel rollers (length of 100 mm and diameter of 5 mm) were set at a distance of 99-129 mm apart on the bottom steel support plate (length, width and thickness were respectively100 x 40 x 5 mm).

The ASWBs were placed over the bottom steel support plate, which reduced the frictional forces between the rollers

MR =

3Pa

(1)

2bd2

Where MR is the flexural modulus of rupture (MPa), P is the maximum applied load (N), a is the distance between line of fracture and the nearest support (mm), b and d are the width and thickness of the specimen (mm), respectively.

The total water absorption capacity of the ASWBs established by the water absorption (WA) test. After 28 days of curing time, the dry specimens were weighed. Then, they were subjected to 24 h submersion. The water of absorption can be determined from the moist weight of specimens after submersion according to the standard ASTM C67-11 [28].

3. Results and discussion Some of the physical and chemical properties of the ASWBs are presented in Tables 1-2. The ASWBs containing residues expanded slightly when fired at 800oC, resulting in a typical behavior of porous bodies. This may be due to the high content in quartz of the clay that is inert at the studied temperature which reduces the contraction of the piece, as well as to the increase in porosity due to the high content in organic matter in the organic residues. All ASWBs showed a contraction at this temperature. The weight loss experienced by the samples upon temperature increased with respect to the residue content at 800 oC for all types of wastes. This weight loss could be due to the elimination of the organic matter from the clay and residue by means of combustion and to the elimination of water content from clay mineral due to dehydroxylation reactions in the clay as suggested by Eliche-Quesada et al. [32].

Table 1.

Average results for compressive and flexural strength tests of the ASWBs. Compressive strength Flexural strength Trial (T) Mix Code Av. Stress b a (mm) b (mm) P (N) L (mm) c (mm) d (mm) (MPa) (mm) T1 SDB5 60 93 17800 3.20 6.4 4.3 5.8 8.9 T2 SDB10 60 92 36500 6.60 6.4 4.7 6.3 8.9 T3 SDB20 59 91 5900 1.10 6.5 4.5 6.0 9.0 T4 CSB5 56 88 49200 10.00 6.4 4.4 6.0 8.9 T5 CSB10 60 89 28700 5.40 6.4 4.4 5.5 8.9 T6 CSB20 ----------------T7 RHB5 56 90 2200 4.40 6.6 4.7 5.9 9.1 T8 RHB10 58 90 7500 1.40 6.6 4.5 6.0 9.1 T9 RHB20 58 92 4900 0.90 7.1 4.3 6.0 9.6 T10 SCB5 59 92 11500 2.10 6.6 4.4 5.5 9.1 T11 SCB10 60 91 9400 1.70 6.8 4.5 6.0 9.3 T12 SCB20 62 95 3300 0.60 6.5 4.3 5.7 9.0 SD, sawdust; CS, cocoa shell; RH, rice husk; SC, sugarcane; P, point load; MR, Modulus of rupture Source: The authors.

37

P (N) 4200 2900 1400 4900 2800 --4800 1700 1000 3800 1300 1100

Av. MR (MPa) 2.79 1.49 0.84 2.97 2.02 --2.90 1.04 0.69 2.83 0.82 0.77

Luna-Ca単as et al / DYNA 81 (188), pp. 34-41. December, 2014.

Table 2.

Results for compressive and flexural strength tests of the ASWBs under Mg2SO4 and H2SO4 attack, respectively Mg2SO4 Compressive Flexural attack strength H2SO4 strength Trial Mix (T) Code Stress P MR t (days) P (N) [M] (MPa) (N) (MPa) T1-1 SDB5 7 15000 2.7 0.25 4200 2.70 T2-1 SDB10 7 10300 2.9 0.25 1400 0.82 T3-1 SDB20 7 3000 0.5 0.25 2800 1.62 T1-2 SDB5 15 14100 2.6 0.50 2900 1.89 T2-2 SDB10 15 11100 1.9 0.50 4900 3.00 T3-2 SDB20 15 3000 0.6 0.50 4800 2.97 T1-3 SDB5 30 11500 2.2 ------T2-3 SDB10 30 6600 1.3 ------T3-3 SDB20 30 4100 0.8 ------T4-1 CSB5 7 9500 1.8 0.25 1700 1.14 T5-1 CSB10 7 12900 2.5 0.25 3800 2.40 T6-1 CSB20 7 ----------T4-2 CSB5 15 22200 4.4 0.50 1000 0.70 T5-2 CSB10 15 11900 2.4 0.50 1300 0.85 T6-2 CSB20 15 ----------T4-3 CSB5 30 18500 3.7 ------T5-3 CSB10 30 15000 3.0 ------T6-3 CSB20 30 ----------T7-1 RHB5 7 10100 2.1 0.25 1100 0.62 T8-1 RHB10 7 3200 0.6 0.25 2900 1.74 T9-1 RHB20 7 1200 0.2 0.25 4900 2.30 T7-2 RHB5 15 11900 2.0 0.50 4200 2.19 T8-2 RHB10 15 3700 0.6 0.50 1400 0.83 T9-2 RHB20 15 1500 0.2 0.50 2800 1.40 T7-3 RHB5 30 12500 2.5 ------T8-3 RHB10 30 3500 0.6 ------T9-3 RHB20 30 2200 0.4 ------T10SCB5 7 0.25 1 16600 3.4 4800 3.18 T11SCB10 7 0.25 1 1100 0.2 1000 0.72 T12SCB20 7 0.25 1 1400 0.3 1300 0.91 T10SCB5 15 0.50 2 14300 2.7 1700 1.08 T11SCB10 15 0.50 2 2000 0.4 3800 2.54 T12SCB20 15 0.50 2 900 0.2 1100 0.75 T10SCB5 30 3 12800 2.5 ------T11SCB10 30 3 1600 0.3 ------T12SCB20 30 3 1300 0.3 ------SD, sawdust; CS, cocoa shell; RH, rice husk; SC, sugarcane; P, point load; MR, Modulus of rupture Source: The authors.

According to Romero et al. [33], it is apparent that open porosity in the ASWBs decreases when the amount of liquid phase tends to approach the particles. The temperature decreased the porosity of the ASWBs. The changes in this property were notable with the addition of sawdust, while the addition of cocoa shell, rice husk and sugarcane produced minor differences in apparent porosity similar to results obtained by Eliche-Quesada et al. [32]. The addition of agroindustrial wastes increased the porosity of the ASWBs, however this effect is expected, since the organic matter of the wastes were eliminated during the thermal process, leading to an increase in the open porosity of the

ceramic bodies. Water absorption, firing temperature and type and content of the residue affects the quality of the final material and its durability significantly [32].

According to Ni単o et al. [22], the clayey soil showed an average water absorption of 30.21%. The addition of waste should produce a significant increase in water absorption.

However, the combustion of organic matter acted differently in the formation of interconnected surface porosity [32].

Fig. 5 illustrates the compressive strength and flexion strength test and set up.

Figure 5. Above, compression strength test; (a) experimental set up, and (b) view of the resultant ASWB after testing. Below, flexural strength test; (c) experimental set up, and (d) view of the resultant ASWB after testing.

Source: The authors.

Figure 6. Average compressive strength for all ASWBs. Source: The authors.

The typical load and compressive strength test is shown in Fig. 5a-5b. The average compressive strength of the ASWBs as a function of waste content is presented in Fig. 6 and results are depicted in Table 1.

38

Luna-Ca単as et al / DYNA 81 (188), pp. 34-41. December, 2014.

Figure 7. Average MR for all ASWBs. Source: The authors.

Figure 9. Average compressive strength for all ASWBs after several days of Mg2SO4 attack.

Source: The authors.

is shown in Fig. 5c-5d and results are depicted in Table 1.

The average MR of the ASWBs as a function of waste content is presented in Fig. 7 and results are depicted in Table 1.

Fig. 8 illustrates the experimental set up for ASWBs under Mg2SO4 and H2SO4 attack.

Figs. 9-10 illustrate the average compressive strength and MR for all ASWBs attacked by Mg2SO4 and H2SO4, respectively, and results are depicted in Table 2.

Fig. 9 shows the average compressive strength of ASWBs with different percentages and type of agroindustrial wastes before and after Mg2SO4 attack. As can be seen, ASWBs shows a strong decrease in the compressive strength after Mg2SO4 attack, with residual values that decreased with days of attack, although with cocoa shell-based bricks kept slightly higher compressive strength values compared with the rest of the ASWBs, and the highest compressive strength values after 15 days of Mg2SO4 attack. However, data from 20 wt.% of cocoa shell addition were not reported.

Fig. 10 shows the average MR of the ASWBs with different percentages and type of agroindustrial wastes before and after H2SO4 attack. As can be seen, ASWBs (5 wt.% addition) show lower MRs than bricks made solely with the clayey soil, except for sugarcane-based bricks, which showed higher MR values after H2SO4 0.25M attack.

On the other hand, the MR decreased with increasing H2SO4 concentration, except for rice husk-based bricks. With 10 wt.% of waste addition, at low H2SO4 concentration, the MR decreased for sawdust- and sugarcane-based bricks, and increased for cocoa shell- and rice husk-based bricks. These results were also obtained at high H2SO4 concentration.

However, with increasing H2SO4 concentration, the MR decreased for cocoa shell- and rice husk-based bricks and

Figure 8. Above, sulfuric acid attack test; (a) experimental set up, (b) and (c) views of the resultant slight efflorescent and non-efflorescent sawdustbased bricks, respectively, after 0.25M sulfuric acid attack. Below, magnesium sulfate attack test; (d) experimental set up, (e) and (f) views of the resultant 10% sawdust-based bricks after 15 and 30 days of magnesium sulfate attack testing, respectively.

Source: The authors.

Fig. 6 shows that the compressive strength tends to decrease with the waste addition except for sawdust-based bricks, which can be related to the higher apparent porosity than clay bricks without residues. The results were better for cocoa shell, with a higher compressive strength using 5 wt.% of cocoa shell, which can be explained by the presence of oil in the waste as suggested by Eliche-Quesada et al. [32], oily films form between particles, acting as lubricants during formation of the clay body and permitting more efficient packing. This phenomenon would promote an increase in mechanical properties of ASWBs. However, a greater percentage of cocoa shell (10 and 20 wt.%) may generate oil pockets that result in pores after firing and contribute to a decrease in compressive strength as reported

by Monteiro and Vieira [34]. Nevertheless, in all ASWBs, including those with higher percentages of waste addition, compressive strengths are always less than the minimum amount (10 MPa) required by existing regulations, except for the 5 wt.% of cocoa shell-based bricks, which produced compressive strength values around 10 MPa. The typical load and deflection in the beam-flexural test 39

Luna-Cañas et al / DYNA 81 (188), pp. 34-41. December, 2014.

compressive strength of ceramic materials is the most relevant engineering quality index for building, the cocoa shell-based bricks obtained at 800 oC had the best quality.

The results indicated that is possible to obtain ASWBs mixing 10% of cocoa shell and 90% of clayey soil, which fulfill the technological standards for traditional bricks and possess mechanical properties similar to those of clay bricks without this waste. Use of these residues could have practical implications as a means of recycling and for achieving cost savings in brick production, since fewer raw materials would be required.

Acknowledgments This research forms part of the L.M. Luna’s undergraduate thesis from the Universidad Industrial de Santander. We gratefully acknowledge the Universidad Industrial de Santander for the use of research facilities (Xray diffraction and Microscopy laboratories) and human resources; with special thanks to Luis Garrido from the Laboratory of Crushing, Milling and Grinding of the School of Geology (Universidad Industrial de Santander). The authors also thank the ERGO Durán & García Brick Company Ltda. for providing facilities to study the mining area and to Ing. Luz Torrado and Vicente Paez from the Concrete and Strength of Materials Laboratory of the Civil Engineering Program (Universidad Pontificia Bolivariana) for their assistance with experimental tests. The authors also acknowledge the anonymous referees for their critical and insightful reading of the manuscript and are most grateful to the above-named people and institutions for support.

Figure 10. Average MR for all ASWBs after H2SO4 attack.

Source: The authors.

increased for sawdust- and sugarcane-based bricks. With 20 wt.% of waste addition, at low H2SO4 concentration, the MR increased for all ASWBs, except for cocoa shell-based bricks (data not reported). These results were also obtained at high H2SO4 concentration. However, with increasing H2SO4 concentration, the MR decreased for all ASWBs, except for sawdust-based bricks. For cocoa shell-based bricks, data were not reported.

After performing durability and strength tests on the ASWBs, results show that most of them perform below the acceptable level in all tests, except for the cocoa shell-based bricks.

References [1]

4. Conclusions [2]

During different agroindustrial activities, huge quantity of solid wastes can be generated as by-products, which pose major environmental problems as well as occupy a large area of land for their storage/disposal. There is a tremendous scope for setting up secondary industries for recycling and using such huge quantities of solid wastes such as minerals or resources in the production of construction materials. Environment-friendly, energyefficient, and cost-effective alternative materials produced from solid wastes will show a good market potential to fulfill people’s needs in rural and urban areas. This study shows that viable bricks can be manufactured with the addition of different percentages of agroindustrial wastes to the traditional clay mix. At the temperature investigated, changes occurred in the values of the bulk density, water absorption and apparent porosity with waste addition, which in turn caused changes in the porosity. Apparent porosity and water absorption values increased with the addition of residues. During the sintering process, the development of a liquid phase caused a decrease in open porosity and water absorption. This increased the compressive strength by reducing the porosity content. Because water absorption is related to the durability of bricks and because the

[3] [4]

[5]

[6]

[7] [8]

40

Safiuddin, Md., Jumaat. M.Z., Salam, M.A., Islam, M.S. and Hashim, R., Utilization of solid wastes in construction materials. International Journal of Physics Sciences, 5 (13), pp. 1952-1963, 2010.

Aubert, J.E., Husson, B. and Sarramone, N., Utilization of municipal solid waste incineration (MSWI) fly ash in blended cement: Part 1: Processing and characterization of MSWI fly ash. Journal of Hazardous Materials, 136, pp.

624-631, 2006. http://dx.doi.org/10.1016/j.jhazmat.2005.12.041 Martínez-Lage, I., Vázquez-Herrero, C., González-Fonteboa, B. and Martínez-Abella, F., Generation of recycled aggregates and technical requiremetns for some applications. DYNA, 77 (161), pp. 89-97, 2010.

Maschio, S., Furlani, E., Tonello, G., Faraone, N., Aneggi, E., Minichelli, D., Fedrizzi, L., Bacchiorrini, A. and Bruckner, S., Fast firing of tiles containing paper mill sludge, glass cullet and clay. Waste Management, 29, pp.

2880-2885, 2009. http://dx.doi.org/10.1016/j.wasman.2009.06.016 Mucahit, S. and Sedat, A., The use of recycled paper processing residue in making porous brick with reduced thermal conductivity. Ceramics International, 35, pp.

2625-2631, 2009. http://dx.doi.org/10.1016/j.ceramint.2009.02.027 Aeslina, A.K., Abbas, M., Felicity, R. and John, B., Density, strength, thermal conductivity and leachate characteristics of light weight fired clay bricks incorporating cigarette butts. International Journal of Environmental Science and Engineering, 2 (4), pp. 179-184, 2010.

Shih, P., Wu, Z. and Chiang, H., Characteristics of bricks made from waste steel slag. Waste Management, 24, pp. 1043-1047, 2004. http://dx.doi.org/10.1016/j.wasman.2004.08.006 El-Mahllawy, M.S., Characteristics of acid resisting bricks made from quarry residues and waste steel slag. Construction and Building Materials, 22, pp.

1887-1896, 2008. http://dx.doi.org/10.1016/j.conbuildmat.2007.04.007

Luna-Cañas et al / DYNA 81 (188), pp. 34-41. December, 2014.

[9]

[10]

[11]

[12]

[13]

[14] [15] [16] [17] [18] [19]

[20]

[21]

[22]

[23] [24] [25] [26]

[27] [28]

Sarkar, R., Singh, N. and Das, S.K., Effect of addition of pond ash and fly ash on properties of ash-clay burnt bricks. Waste Management Research, 25, pp.

566-571, 2007. http://dx.doi.org/10.1177/0734242X07080114 Cultrone, G. and Sebastián, E., Fly ash addition in clayey materials to improve the quality of solid bricks. Construction and Building Materials, 23, pp.

1178-1184, 2009. http://dx.doi.org/10.1016/j.conbuildmat.2008.07.001 Monteiro, S.N., Alexandre, J., Margem, J.I., Sánchez, R. and Vieira, C.M.F., Incorporation of sludge waste from water treatment plant into red ceramic. Construction and Building Materials, 22, pp. 1281-1287, 2008. http://dx.doi.org/10.1016/j.conbuildmat.2007.01.013 Chiang, K-Y., Chou, P-H., Hua, C-R., Chien, K-L. and Chesseman, C., Lightweight-bricks manufactured from water treatment sludge and rice husks. Journal of Hazardous Materials, 171, pp. 76-82, 2009. http://dx.doi.org/10.1016/j.jhazmat.2009.05.144 Dondi, M., Guarini, G., Raimondo, M. and Zanelli, C., Recycling PC and TV waste glass in clay bricks and roof tiles. Waste Management, 29, pp.

1945-1951, 2009. http://dx.doi.org/10.1016/j.wasman.2008.12.003 Demir, I., An investigation on the production of construction brick with processed waste tea. Building and Environment, 41, pp. 1274-1278, 2006. http://dx.doi.org/10.1016/j.buildenv.2005.05.004 Turgut, P. and Halil, M.A., Limestone dust and wood sawdust as brick material, Building and Environment, 42, pp. 3399-3403, 2007. http://dx.doi.org/10.1016/j.buildenv.2006.08.012 Demir, I., Effect of organic residues addition on the technological properties of clay bricks, Waste Management, 28, pp. 622-627, 2008. http://dx.doi.org/10.1016/j.wasman.2007.03.019 Halil, M.A. and Turgut, P., Cotton and limestone powder waste as brick material, Construction and Building Materials, 22, pp. 1074-1080, 2008. http://dx.doi.org/10.1016/j.conbuildmat.2007.03.006 Sohrab, V. and Ali, A.Y., The use of polystyrene in lightweight brick production. Iranian Polymer Journal, 12 (4), pp. 323-329, 2003.

Turgut, P. and Bulent, Y., Physico-mechanical and thermal performances of newly develop rubber-added bricks, Energy and Buildings, 40, pp.

679-688, 2008. http://dx.doi.org/10.1016/j.enbuild.2007.05.002 Torres, P., Fernandes, H.R., Olhero, S. and Ferreira, J.M.F., Incorporation of wastes from granite rock cutting and polishing industries to produce roof tiles. Journal of European Ceramic Society, 29, pp.

23-30, 2009. http://dx.doi.org/10.1016/j.jeurceramsoc.2008.05.045 Alonso-Santurde, R., Andrés, A., Viguri, J.R., Raimondo, M., Guarini, G., Zanelli, C. and Dondi, M., Technological behaviour and recycling potential of spent foundry sands in clay bricks. Journal of Environmental Management, 92 (3), pp. 994-1002, 2011. http://dx.doi.org/10.1016/j.jenvman.2010.11.004 Niño, M.C., Spinosi, V. and Ríos, C.A. and Sandoval, R., Effect of the addition of coal-ash and cassava peels on the engineering properties of compressed earth blocks. Construction and Building Materials, 36, pp.

276-286, 2012. http://dx.doi.org/10.1016/j.conbuildmat.2012.04.056 ASTM C136-06, Standard test method for sieve analysis of fine and coarse aggregates. American Society for Testing and Materials, West Conshohocken, Pennsylvania, PA 19428, USA; 2006.

ASTM D1140-00, Standard test methods for amount of material in soils finer than No. 200 (75-μm) sieve. American Society for Testing and Materials, West Conshohocken, Pennsylvania, PA 19428, USA; 2006.

ASTM D4318-10, Standard test methods for liquid limit, plastic limit, and plasticity index of soils. American Society for Testing and Materials, West Conshohocken, Pennsylvania, PA 19428, USA; 2010.

De la Casa, J.A., Lorite, M., Jiménez, J. and Castro, E., Valorisation of wastewater from two-phase olive oil extraction in fired clay, Journal of Hazardous Materials, 169, pp.

271-278, 2009. http://dx.doi.org/10.1016/j.jhazmat.2009.03.095 Kadir, A.A. and Sarani. N.A., An Overview of wastes recycling in fired clay bricks. International Journal of Integrated Engineering, 4 (2), pp.

53-69, 2012.

ASTM C67-11, Standard test methods for sampling and testing brick and structural clay tile. American Society for Testing and Materials, West Conshohocken, Pennsylvania, PA 19428, USA; 2011.

[29]

[30]

[31]

[32]

[33]

[34] [35]

ASTM D 2166-00e1, Standard Test method for unconfined compressive strength of cohesive soils. American Society for Testing and Materials, West Conshohocken, Pennsylvania, PA 19428, USA; 2004.

ASTM D 1635-00, Standard test method for flexural strength of soilcement using simple beam with 3rd point loading. American Society for Testing and Materials, West Conshohocken, Pennsylvania, PA 19428, USA; 2006.

Varela, P.G., Cotella, N.G., Oviedo, O.E., Radevich, O.A. y Kohl, R.G., Influencia de la velocidad de ensayo sobre el módulo de ruptura en moldes para fundición de precisión. Jornadas SAM 2000 - IV Coloquio Latinoamericano de Fractura y Fatiga, Neuquén, Argentina, pp. 101107, 2000.

Eliche-Quesada, D., Corpas-Iglesias, F.A., Pérez-Villarejo, L. and Iglesias-Godino, F.J., Recycling of sawdust, spent earth from oil filtration, compost and marble residues for brick manufacturing.

Construction and Building Materials, 34, pp. 275-284, 2012. http://dx.doi.org/10.1016/j.conbuildmat.2012.02.079 Romero, M., Andrés, A., Alonso, R., Viguri, J. and Rincón, J.M., Sintering behavior of ceramic bodies from contaminated marine sediments. Ceramics International, 34, pp. 1917-1924, 2008. http://dx.doi.org/10.1016/j.ceramint.2007.07.002 Monteiro, S.N. and Vieira, C.M.F., Effect of oily waste addition to clay ceramic. Ceramics International, 31, pp. 353-358, 2005. http://dx.doi.org/10.1016/j.ceramint.2004.05.002 Masonry Standards Joint Committee (MSJC). Building Code Requirements for Masonry Structures. ACI 530-08/ASCE 5-08/TMS 402-08, ACI, Farmington Hills, Mich.; ASCE, Reston, Va.; TMS, Boulder, Colorado, 2008.

L. M. Luna-Cañas, received a BSc in Geology in 2012 from the Universidad Industrial de Santander, Bucaramanga, Colombia. She has worked as Assistant Officer in Ingeniería de Corrosión ICL in 2013, consulting for diagnosis, design and geotechnical monitoring, as a Geologist in ConstruSuelos Colombia S.A.S. in 2013, conducting geological exploration of surface and geomorphic units, and is a Junior Geologist since 2011 in AUX Colombia, applying geotechnical logging in the core description in situ and logging area.

C.A. Ríos-Reyes, received a BSc in Geology in 1989 and the Postgraduate Diploma in University Teaching in 1995 from the Universidad Industrial de Santander, Bucaramanga, Colombia. The Shimane University, Matsue, Japón, conferred him the degree of MSc in Geology in 1999. The University of Wolverhampton, Wolverhampton, England, conferred him the degree of PhD in Applied Sciences in 2008. He has been working as a full-time Lecturer of the School of Geology since 1992, where he developed his professional experience at a University teaching level during the last 22 years in the field of Mineralogy and fieldwork on crystalline basement complexes in different areas of Colombia. During the last 6 years he has focused his research interests on the use of low-cost raw materials with potential application in the mitigation of environmental problems.

Actually, he is the director of the Research Group in Basic and Applied Geology at the School of Geology of the Universidad Industrial de Santander, Colombia and the director of the Microscopy Laboratory of the Guatiguará Technological Park.

L.A. Quintero-Ortíz, received a Bs Eng in Metallurgic Engineering in 1983, a Post-graduate Diploma in University Teaching in 2002, and an MSc degree in Metallurgic and Material Science Engineering in 2002, from the Universidad Industrial de Santander, Bucaramanga, Colombia. She has been working as a full-time Lecturer at the School of Metallurgic and Material Science Engineering since 1983, when she received her professional degree. Actually, she is the director of the Research Group in the Development and Technology of New Materials at the School of Metallurgic and Material Science Engineering of the Universidad Industrial de Santander, Colombia. Her research interests include of the improvement of the manufacturing, characterization and evaluation processes of materials, training human resources to boost the development of new materials, and developing materials for technological applications. Her main interests are developing materials for technological applications, improving and innovating the process of obtaining materials via casting and performing basic research on materials.

41

The use of gypsum mining by-product and lime on the engineering properties of compressed earth blocks Eliana Rocío Jaramillo-Pérez a, Josue Mauricio Plata-Chaves b & Carlos Alberto Ríos-Reyes c b

a Escuela de Geología, Universidad Industrial de Santander, Bucaramanga, Colombia, elijarap@gmail.com Escuela de Geología, Universidad Industrial de Santander, Bucaramanga, Colombia, mauriciowhl@gmail.com c Escuela de Geología, Universidad Industrial de Santander, Bucaramanga, Colombia, carios@uis.edu.co

Received: September 1th, 2013.Received in revised form: March 16th, 2014.Accepted: September 25th, 2014.

Abstract Disadvantages of compressed earth blocks are their poor mechanical properties and low resistance to water damage. Therefore, their use is vulnerable to deterioration and require care and maintenance, which depends on the degree of stabilization and compaction of the clay soil. Gypsum mining wastes and lime used as stabilization materials to improve the properties of these construction materials. The compressive and flexural strength, softening in water, drying shrinkage and unit weight determined. Strength values increased with both mining waste additions. Highest resistance against softening in water obtained with a 25% of mining waste. Drying shrinkage reduced with increasing mining waste content. Dry unit weight was not in the recommended standards. Results showed that gypsum mining wastes can be used as alternative materials to stabilize compressed earth blocks.

Keywords: compressed earth blocks; construction materials; gypsum mining by-product; stabilization; environmental.

El uso de residuos de minería de yeso y cal sobre las propiedades de ingeniería de los bloques de tierra comprimida Resumen Las desventajas de los bloques de tierra comprimida son sus bajas propiedades mecánicas y resistencia al daño al agua. Por lo tanto, su uso es vulnerable al deterioro y requiere cuidado y mantenimiento, dependiendo del grado de estabilización y compactación del suelo arcilloso. Residuos de minería del yeso y cal se utilizaron como estabilizantes para mejorar las propiedades de estos materiales de construcción. Resistencia a compresión y flexión, ablandamiento en agua, retracción por secado y peso unitario se determinaron. La resistencia aumento con la adición de residuo de minería. La resistencia al ablandamiento en agua fue mayor con 25% de residuo de minería. La contracción por secado disminuyo con el aumento del contenido de residuo de minería. El peso unitario seco no estaba en los estándares recomendados. Los resultados mostraron que los residuos de minería del yeso pueden utilizarse como materiales alternativos en la estabilización de bloques de tierra comprimida.

Palabras claves: bloques de tierra comprimida; materiales de construcción; residuos de minería del yeso; estabilización; medio ambiente.

1. Introduction Compressed earth blocks (CEBs) play a major role in improving the environmental efficiency and sustainability of buildings and contributes to worldwide economic prosperity and infrastructural development. On the other hand, the production processes of construction materials have a considerable impact on the environment. The utilization of earth in housing construction is one of the oldest and most common methods used. CEBs are one of the oldest identifiable man-made building materials which

are becoming more popular due to their simplicity and low cost, relative abundance of materials, good performance (good thermal and acoustic properties), and at the end of a building's life the clay material can easily be reused by grinding, wetting or returned to the ground without any damage to the environment [1]. However, despite their advances, further studies are needed in order to improve their durability and mechanical properties, both important quality control measures for manufacturers and builders.

Many additives such as cement, lime, asphalt emulsions, bituminous materials, and natural and industrial by-products

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 42-51. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.39725

Jaramillo-PĂŠrez et al / DYNA 81 (188), pp. 42-51. December, 2014.

alternative to traditional building practices that is relatively inexpensive, uses local resources, and in some cases, has been found to last several millennia [10]. A number of standards have also developed for CEB test procedures [1012]. However, unlike other masonry units, there is little general consensus on test procedure for CEBs. The main objective of this study is to investigate the effects of the aforementioned types of industrial residues on the properties of CEBs. Results of experimental studies are also presented.

The compressive strength of blocks measured by different tests is also compared with other parameters, such as threepoint bending strength.

have been tested to improve the mechanical properties and to enhance the durability of the compacted blocks [1-6].

Portland cement has been by far the most used material for soil stabilization [2,5,6]. However, due to the high energy consumption necessary for its manufacture and the consequent environmental damage caused by the release of high quantities of greenhouse gases during its production, the cement industry has been highlighted as one of the major contributors of anthropogenic CO2 emissions emitting about 5% globally [7-8]. In view of the above mentioned, several research activities are directed towards partial or total substitution of Portland cement by pozzolanic binders, e.g. lime, fly ash, and natural pozzolans among others. The worldwide development of mining produces large volumes of mining wastes and their disposal cause major challenges and serious economic and environmental problems. Mining of industrial minerals is a special case as far as mining waste generation is concerned, since they are mostly inert used directly in restoration work. The problem is the need for integrated management, including the removal and restoration, rather than the generation of hazardous materials. Gypsum is mined from Cretaceous sedimentary rocks at various locations in Colombia. However, the mining wastes produced after the extraction of gypsum is not used for restoration. Lately, researchers are making efforts to reduce the amount of waste by finding alternative uses for it. The need to conserve the traditional building materials that are facing depletion has necessitated the search for alternative materials [9]. In the 1970s and 1980s a new generation of manual, mechanical and motor-driven presses appeared, leading to the emergence today of a genuine market for the production and application of CEBs [4]. They have excellent insulating properties - reducing heating and cooling costs. The compressive strengths of the blocks depend on their densities. The compressive strength of a soil can be increased by chemical stabilization. This project was designed to prepare locally available soils, make building blocks with a block press and test them to determine the engineering properties of CEBs. The objective was to test local soils to see if they could be used for low housing construction. CEB technology offers an

2. Materials and methods 2.1. Materials The materials used for the industrial trial consisted of raw clay-rich material and gypsum mining waste (Fig. 1), and lime.

2.1.1. Raw clay-rich material The raw clay-rich material used in this study is extracted

by Polypus of Colombia for the development of the housing project â&#x20AC;&#x153;Prados de Laurentiaâ&#x20AC;? at Floridablanca (Santander), which offers an innovative construction system. The clay soil forms part of the Quaternary Fine Member of the Bucaramanga Formation and presents characteristics suitable for the production of CEBs [13] with dimensional tolerances conform to ASTM Standards.

2.1.2. Gypsum mining by-product Gypsum mining wastes, which are disposed after extraction of gypsum from Cretaceous sedimentary rocks of the Rosablanca Formation in several mines located around Los Santos (Santander), was used as a chemical additive to protect CEBs against moisture decomposition and stabilize them.

2.1.3 Lime

An industrial lime was also used as a stabilizer.

Figure 1. Raw materials Source: The authors.

43

Jaramillo-Pérez et al / DYNA 81 (188), pp. 42-51. December, 2014.

2.2. Properties of materials Qualitative determination of major crystalline phases of the raw clay-rich material and the gypsum mining by-product was carried out by using a Siemens D500 X-Ray Diffractometer, operating in the Bragg–Brentano geometry with CuK1 radiation (=1.5406 Å), at 40 kV and 30 mA, and a graphite monochromator. Data was collected in the 2-70° 2θ range (0.02° step size). The crystalline patterns were compared with the standard line patterns from the Powder Diffraction File database supplied by the International Centre for Diffraction Data (ICDD), with the help of Joint Committee on Powder Diffraction Standards (JCPDS) files for inorganic compounds.

The major crystalline phases found in the clay-rich material are quartz, microcline, muscovite, anatase and kaolinite (Fig. 2a).

As shown in Fig. 2b, the gypsum mining by-product is characterized by the occurrence of quartz, clinochlore, gypsum, dolomite, Mg-calcite and calcite. The chemical composition of this was investigated by X-ray fluorescence using a Shimazu EDX 800 HS XRF spectrometer to quantify the elements in the gypsum mining waste using the method of fundamental parameters (FP) with the software DXP-700E Version 1.00 Rel. 014. The chemical composition of the gypsum mining waste used in this study was 48.64% CaO, 27.31% SiO2, 9.16% MgO, 6.13% SO3, 4.81% Al2O3, 2.41% Fe2O3, 1.53% K2O, 0.47 SrO%, 0.20 MnO, 0.11% BaO and 0.02% CuO. The particle size distribution (the relative content of clay, sand and gravel) of the clay-rich material (Fig. 3) obtained by combined sieve and hydrometer analyses according to the standards ASTM C136-06 [14] and ASTM D1140-00 [15].

Figure 3. Particle size distribution of the clay-rich material Source: The authors.

Fig. 3a reveals that the clay-rich material is within the recommended limits for the manufacture of CEBs, which according to Houben et al. [16], are: gravel (0-40%), sand (25-80%), silt (10-25%) and clay (8-30%). The fine grained portion (amount of soil to pass a No. 200 mesh) was 21.1%, which was used in determining the percentages of clay (13.5%) and silt (7.5%) by the hydrometry test (Fig. 3b).

According to Cuéllar et al. [17], clay properties depend on the structural characteristics and particle size (< 2µm).

Therefore, this test only measures the clay/silt ratio for the fine grained portion of the soil and not the entire soil itself.

The Atterberg's limits of the clay-rich material determined according to the standard ASTM D4318-10 [18], using the plasticity chart with the following results: liquid limit (LL) of 27.3%, plastic limit (PL) of 21.2% and plasticity index (PI) of 6.1%, with an acceptable correlation (R2 = 0.739). The clay-rich material can be classified as SM-SC (silty-clayey sand with low plasticity) using the ASTM D2487-11 [19]. It corresponds to a coarse-grained (> 50% retained on No. 200 mesh) sandy (> 50% of coarse fraction is < 4.75 mm (No. 4 mesh)). Similarly, it contains > 12 % of material passing the No. 200 mesh, LL < 50%, 4 ≤ IP ≤ 7 and Atterberg's limits on or above the “A” Line.

For the purpose of sample preparation, dry density and moisture content values were established. Therefore, Proctor Compaction tests were carried out in accordance to the standard ASTM D1557-12 [20] in order to establish values of the maximum dry density and optimum moisture (Fig. 4) for the non-stabilized and stabilized CEBs. This was to guide the research on the possible range of moisture contents at which the dry unit weight of the clay-rich soil

Figure 2. XRD pattern of raw clay-rich material (upper part) and and gypsum mining by-product (lower part). Ms, muscovite; Kao, kaolinite; Qtz, quartz; Mic, microcline; Ana, anatase; Ccs, clinocrysotile; Gyp, gypsum; Cal, calcite; Py, pyrite; Or, orthoclase; Dol, dolomite Source: The authors.

44

Jaramillo-PĂŠrez et al / DYNA 81 (188), pp. 42-51. December, 2014.

Figure 4. Results from Proctor Compaction tests Source: The authors.

will be a maximum and to achieve the best compaction effort. The clay-rich material used in this study has a natural moisture content (optimum moisture) of 11.5%.

The moisture content of the sample at the point of testing includes this natural moisture in the clay soil and the water added at the point of mixing. The clay was used at its natural moisture content because in practice, an oven drying operation will not be feasible.

Figure 5. Methodology for manufacturing CABs Source: The authors.

2.3. Sample preparation, mix compositions and testing Fig. 5 illustrates a block diagram showing the methodology followed in the manufacturing of the CEBs during their study. The raw clay-rich material and the gypsum mining by-product were naturally dried for three weeks under the following environmental conditions: average temperature of 24oC and relative humidity of 83.5%. The mining waste subjected to rough crushing with a Retsch Jaw Crusher BB200 to ~ 2 mm and milling with a Retsch RM100 mortar grinder mill to clay particle size. Both raw clay-rich material and gypsum mining byproduct sieved with a Ro-Tap sieve shaker (using 4, 10, 20, 40, 60, 100 and 200 mesh series). The mining waste sieved and the particle size below 200 mesh used. In order to evaluate the engineering properties of CEBs, with gypsum mining by-product as stabilizing agent, several mixtures were prepared for mix design of preparation of CEBs. The mix proportions were prepared based on the dry weights of the ingredients. The quantities of materials obtained from the mix design was measured with the aid of a weighing balance. CEBs were produced with a Cinva-Ram block making machine, a technology that offers an alternative kind of building construction which is more accessible and of high quality. For testing, 91 CEBs (13 for each mixture) were prepared. The cuboidal shape and size (290 x 100 x 140 mm) tolerances of the masonry units respected.

Fig. 6 illustrates the preparation of the CEBs. Several mixtures were loaded into the block making machine. Table 1 reports the details of the mixture compositions and the assessment of the process of manufacture of CEBs produced during the tests.

Figure 6. Stages of characterization of raw materials and preparation of CABs. (a)-(b) Combined sieve and hydrometer analyses of the clay-rich material. (c) Sieve analyses of the gypsum mining waste. (d)-(e) Test of the Atterberg's limits (limits of consistency). (f) Proctor Compaction test. (g) Mix of materials. (h) Mix in the Cinva-Ram block-making machine before pressing process. (i) Resultant set of CABs Source: The authors.

Table 1.

Details of mix composition used during industrial trial.

Trial Mix CS MR WA Mix proportions (%) (T) code CRM GMW CaO H2O (MPa) (MPa) (%) T1 CEB1 88.5 0 0 11.5 0.25 0.291 --T2 CEB2 86 2.5 0 11.5 0.62 0.333 --T3 CEB3 81.5 2.5 3 13 1.57 0.446 25.5 T4 CEB4 83.5 5 0 11.5 0.47 0.262 --T5 CEB5 79 5 3 13 1.21 0.580 22.4 T6 CEB6 78.5 10 0 11.5 0.90 0.357 --T7 CEB7 74 10 3 13 1.32 0.334 23.4 CEB, compressed earth block; CRM, clay-rich material; GMW, gypsum mining waste; CS, Compression strength; MR, Modulus of rupture; WA, Water absorption Source: The authors.

45

Jaramillo-P茅rez et al / DYNA 81 (188), pp. 42-51. December, 2014.

To CEBs not stabilized with lime, the water content was increased by 1.5% compared to 3% lime, as suggested in Perez & Pach贸n [21] and given the reaction ratio of water/lime which corresponds to 0.5:1. These authors suggest minimal use of lime, as it can generate a low performance. An automated hydraulic pump, connected to a mold frame by a hydraulic hose and cylinder, was used to gradually pressurize the cylinder which in turn applied pressure to the mixture in the mold. After the pressure was applied for a few seconds, it was released and a CEB was extruded by the machine onto a conveyor belt for transfer to storage. In order to obtain comparable results, five different series of samples were prepared for the tests, a separate series for each percent material addition. CEBs were kept undisturbed under controlled environmental conditions (average temperature of 25 oC and relative humidity of 80%) during the curing phase (28 days). No detrimental effects due to shrinking/swelling, such as cracking, were observed. Engineering tests conducted in a computerized device for mechanical assays according to the standard ASTM C67-11 [22]. A Universal Testing Machine (PINZUAR, model PC-160/116) with a maximum load of 1000kN was used in the testing procedure, taking into account its accuracy, flexibility, high performance, and innovative standard features; large test space to accommodate standard, medium and large size specimens, grips, fixtures and environmental subsystem, and environmental chamber dimension: 500 x 255 x 350 mm.

Data was recorded automatically to the computer system.

All CEBs were subjected to a compressive load at a crosshead speed of 0.5 mm/min. A test of compressive strength was conducted to determine the level of deformation of the material. The three-point bending flexural strength test was conducted with a crosshead speed of 0.2 mm/s and a distance between the supports of 90 mm.

The test provides values for the modulus of rupture (MR) of the CEBs. MR was calculated using the following equation:

Figure 7. Compression strength test, showing experimental set up and resultant CABs after testing Source: The authors.

(2) Where Wd is the mass of the dry specimens before submersion (g) and Ww is the wet mass of the specimen after being removed from the water tank (g).

3. Test results and discussion Table 1 shows the average values of results for the compression, flexural and water absorption tests. Each value represents the average of 5 specimens. The number and series of specimens was according to ASTM standards and depending on the number of different mixtures tested, with a minimum of five specimens per batch.

3.1. Compressive strength of the CEBs The uniaxial compressive stress is reached when the material fails completely. The compressive strength test determines the relationship stress vs. strain of the CEBs. Fig. 7 shows a representative set of the experimental test to determine the compressive strength of the CEBs and results are depicted in Table 1.

Fig. 8 illustrates the average compressive strength of the CEBs. It also shows the influence of the gypsum mining byproduct on the compressive strength of specimens obtained after 28 days of curing time under dry conditions. From trial T1, the simple compression test reached an average value for 5 units of 0.251 MPa. This value must be taken into account to compare the results with other trials. CEB1 tends to separate at the ends while the center remains consistent.

From trial T2, the simple compression test reached an average value for 5 units of 0.624 MPa, higher than that obtained in the trial T1. CEBs tend to separate at the ends while the center remains consistent. From trial T3, the simple compression test reached an average value for 5 units of 1.574 MPa, higher than that obtained in trials T1 and T2. CEBs tend to separate at the ends, with some of them displaying broken side surfaces, while the front and center remain consistent. From trial T4, the simple compression test reached an average value for 5 units of 0.466 MPa, slightly higher than that obtained in the trial T1.

Deep cracks were observed on the CEBs although they did not disintegrate completely. From trial T5, the simple compression test reached an average value for 5 units of 1.206 MPa, slightly lower than that obtained in the trial T3.

Some CEBs displayed cracks and others tended to separate in their external surfaces while the center remains

(1) Where MR is the flexural modulus of rupture (MPa), P is the maximum applied load (N), a is the distance between line of fracture and the nearest support (mm), b and d are the width and thickness of the specimen (mm), respectively.

The durability of the CEBs was assessed as follows: after 28 days of curing time, the CEBs were weighed; then, they were submerged in water for 24 h and then tested in compression after repeated wetting and drying on their unconfined compressive strength values. Repeated wetting and drying of the blocks can alter the soil structure and create concentrated weaknesses through cracking and the infiltration of water. The total water absorption capacity of the CEBs was established by the water absorption (WA) test.

The water of absorption can be determined from the moist weight of specimens after submersion according to the standard ASTM C67-11 [22]. The water absorption during immersion was calculated using the following equation:

46

Jaramillo-Pérez et al / DYNA 81 (188), pp. 42-51. December, 2014.

3.2. Flexural strength characteristic of the CEBs It is the ability of a masonry brick, beam or slab to resist failure in bending. The typical load and deflection from beam-flexural test is shown in Fig. 9 and results are depicted in Table 1.

Fig. 10 illustrates the average MR of the CEBs. The nonstabilized CEBs from the trial T1 have a load carrying capacity of 1133 N. This mixture achieved a MR in the range of 0.202–0.369 MPa (with an average of 0.291 MPa).

The non-stabilized CEBs from the trial T2 containing clayrich material (86%) and gypsum mining waste (2.5%) have a load carrying capacity of 1300 N. This mixture achieved a higher MR, in the range of 0.278–0.395 MPa (with an average of 0.334 MPa). The addition of gypsum mining waste helps to stabilize the clay-rich material, improving the engineering properties of CEBs. The stabilized CEBs from the trial T3 containing clay-rich material (81.5%), gypsum mining waste (2.5%) and lime (3%) have a load carrying capacity of 1767 N, which is higher than the CEBs obtained in the trials T1 and T2, increasing the stress resistance. This mixture achieved an MR in the range of 0.405–0.505 MPa (average of 0.446 MPa). The non-stabilized CEBs from the trial T4, containing clay-rich material (83.5%) and gypsum mining waste (5%), have the lower load carrying capacity of 1000 N. This mixture achieved an MR in the range of 0.233–0.316 MPa (with an average of 0.262 MPa). The stabilized CEBs from the trial T5, containing clay-rich material (79%), gypsum mining waste (5%) and lime (3%), have the higher load carrying capacity of 2233 N. The higher MR values (0.482–0.758 MPa; an average of 0.580 MPa) obtained in the mixtures with a 5% of gypsum mining waste. The non-stabilized CEBs from the trial T6, containing clay-rich material (78.5%) and gypsum mining waste (10%) have a load carrying capacity of 1433 N. This mixture achieved an MR in the range of 0.328–0.404 MPa (an average of 0.357 MPa). The stabilized CEBs from the trial T7, containing clay-rich material (74%), gypsum mining waste (10%) and lime (3%) have a load carrying capacity of 1400 N. This mixture achieved an MR in the range of 0.301–0.363 MPa (with an average of 0.334 MPa).

Figure 8. Average compressive strength for all CEBs Source: The authors.

Figure 9. Flexural strength test set up, showing experimental set up and resultant CABs after testing Source: The authors.

consistent. From trial T6, the simple compression test reached an average value for 5 units of 0.901 MPa, slightly lower than that obtained in the trial T5. Deep cracks were observed on the CEBs and they tended to separate in their external surfaces although they did not disintegrate completely. From trial T7, the simple compression test reached an average value for 5 units of 1.319 MPa, which is in the range of values obtained between the trials T3 and T5. CEBs tend to separate on the sides and front, and some of them were crossed by cracks along their front.

From Fig. 8, we observe that the behavior of unstabilized CEBs (trials T1, T2, T4 and T6), lies well below the values recommended by Colombian technical standards, which suggest a minimum value of compressive strength of 1.2 MPa. Therefore, it would not be advisable to use them for the development of individual blocks. The best performances were obtained in the mixtures stabilized with lime, obtaining the best value for the mixtures from the trial T3 containing 2.5% gypsum mining waste and 3% lime, showing an improvement of 52.7% (6 times better) with respect to CEBs obtained from the trial T1.

Increasing the gypsum mining waste content above 2.5% (trials T5 and T7), promote a increase in the compressive strength but remains below the value obtained from the trial T3. Using the gypsum mining waste without the presence of lime, the results showed an improvement of between 150 and 250% (up to 3 times better). For all cases, mixtures stabilized without using lime showed compressive strength values lower than those obtained from lime stabilized CEBS.

Figure 10. Average MR for all CEBs Source: The authors.

47

Jaramillo-Pérez et al / DYNA 81 (188), pp. 42-51. December, 2014.

These results show that although the gypsum mining waste meets the expectations and proposals for the compaction point load strength of the CEBs, doses greater than 5% strongly affect the mixture. These results confirmed the results obtained from the compressive strength test.

According to the Masonry Standards Joint Committee (MSJC) [23], the allowable flexural tensile stress, or modulus of rupture, for clay and concrete masonry is 0.21 MPa. Using this as the quality standard, the allowable rupture load could be determined. The CEBs showed flexural strengths between 0.262 and 0.580 MPa, which is below the range of 0.5-2 MPa reported in previous studies [24-25], except for the MR obtained from the trial T5 (0.580 MPa). From trial T1, the population of data has a mean of 0.175 MPa and their standard deviation is 0.152. From trial T2, the population of data has a mean of 0.200 MPa and their standard deviation is 0.168. From trial T3, the population of data has a mean of 0.268 MPa and their standard deviation is 0.221. From trial T4, the population of data has a mean of 0.157 MPa and their standard deviation is 0.132. From trial T5, the population of data has a mean of 0.348 MPa and their standard deviation is 0.300. From trial T6, the population of data has a mean of 0.214 MPa and their standard deviation is 0.177. From trial T7, the population of data has a mean of 0.201 MPa and their standard deviation is 0.165. The fourth population has the smaller standard deviation than the other populations because its values are mostly close to 0.132. After performing durability and strength tests on the CEBs, results show that most of them perform at an acceptable level in all tests. However, gypsum mining waste doses of 10% or more will reduce workability of the CEBs.

Figure 11. Durability test of the CEBs Source: The authors.

consistency, disintegrating completely, and developing a silty sand mixture, in which the gypsum mining waste separated and was easily differentiated from the mixture.

The stabilized CEBs from the trial T3 containing clayrich material (81.5%), gypsum mining waste (2.5%) and lime (3%), retained their shape but their size increased from 5 to 10 mm, and deformation was observed at their edges and corners. The average water absorption was 25.502%, being the highest from the CEBs that retained their shape. The compressive strength after water absorption showed an average value of 0.580 MPa. The non-stabilized CEBs from the trial T4, containing clayrich material (83.5%) and gypsum mining waste (5%), showed a similar behavior to that observed in the trial T2, with CEBs losing consistency, disintegrating completely, and developing a silty sand mixture, in which the gypsum mining waste separated and was easily differentiated from the mixture. The stabilized CEBs from the trial T5, containing clay-rich material (79%), gypsum mining waste (5%) and lime (3%), retained their shape but their size increased by up to 5 mm, and deformation was observed at their edges and corners. The average water absorption was 22.378%, being the lowest from CEBs that retained their shape. The compressive strength after water absorption showed an average value of 0.860 Mpa, which is higher when compared with that observed in the trial T3. The non-stabilized CEBs from the trial T6, containing clay-rich material (78.5%) and gypsum mining waste (10%), showed a similar behavior to that observed in trials T2 and T4, with CEBs losing consistency, disintegrating completely, and developing a silty sand mixture, in which the gypsum mining waste separated and was easily differentiated from the mixture.

The stabilized CEBs from the trial T7, containing clayrich material (74%), gypsum mining waste (10%) and lime (3%), showed a great loss of material to the edges and corners, which took a rounded shape. The average water absorption was 23.360%, although this percentage is not representative because possibly could correspond to the loss of material and not to the degree of water absorption. Therefore, the compressive strength test after water absorption was not performed for this trial. The water absorption tests reveal that stabilizing the mixtures with lime ensures better structural consistency.

According to Fig. 11, the results obtained were not

3.3. Durability testing of the CEBs The durability of the CEBs assessed by determining the effect of wetting and drying on their compressive strength values, although without the number of saturation cycles suggested by Krosnowski [25], which can alter the soil structure and create concentrated weaknesses through cracking and the infiltration of water. Results obtained from the water absorption test give a general idea to assess the behavior of the CEBs under extreme conditions. In the case of Bucaramanga and its metropolitan area, one of these extreme conditions is the possibility of a flood, particularly affecting the CEBs that form the base of a wall, which are more likely to be submerged completely and should bear the burden of the entire wall. As the density of soil is increased, its porosity reduced and less water can penetrate it [26].

Water absorption is used as an indicator for the specimen’s resistance to immersion. Table 1 and Fig. 11 present results from the durability test. During the saturation, each CEB was carefully examined for any observable cracking or degradation effects.

Fig. 12 shows CEBs soaked in water and the detrimental effects of saturation. The non-stabilized CEBs from the trial T1 showed a loss of consistency, disintegrating completely, and developing a silty sand mixture. The non-stabilized CEBs from the trial T2 containing clay-rich material (86%) and gypsum mining waste (2.5%), showed a loss of 48

Jaramillo-Pérez et al / DYNA 81 (188), pp. 42-51. December, 2014.

satisfactory regarding CEBS from trials T1, T2, T4 and T6 prepared in the absence of lime, which showed results losing consistency, disintegrating completely, and developing a silty sand mixture, in which the gypsum mining waste separated and easily differentiated from the mixture. Therefore, these mixtures are described as inefficient and their behavior can be explained due to the presence of sulphates in the gypsum mining waste that are sensitive to water so that when wetted they may become easily detached. CEBs from trials T3, T5 and T7 containing 3% lime, kept in shape, showing deviation in their lengths (~ 5-10 mm) and disintegration at their edges and corners. Although CEBs from trial T3 holds its shape, CEBs from trial T5 showed the best consistency, keeping their shape better. CEBs from trial T7 suffered further disintegration, showing a great loss of material to the edges and corners, which took a rounded shape.

Regarding the percentage of moisture in weight, CEBs from trials T3 and T5 showed values of 25.502% and 22.378%, respectively; CEBs from the trial T7 showed a percentage of moisture in weight of 23.036%, although due to the loss of material at the edges, cannot be considered as representative. These results are considered good considering the values obtained in previous studies [1,13,27-29]. After water absorption, the non-stabilized and stabilized CEBs were subjected to simple compression test to assess the degree of consistency while being subjected to excessive moisture. Results show that the mixtures have a cohesive behavior when having excess moisture, with the clay-rich material taking a plastic behavior. Compressive strength values of 0.593 MPa and 0.885 MPa for CEBs from trials T3 and T5, respectively, are still sufficient to maintain a standing wall in a building. According to Krosnowski [25], a decrease in the compressive strength values with the number of saturation cycles is expected to occur; however, the rate at which the compressive strength decreases also appears to decrease as the number of saturation cycles increases.

4. Conclusions A laboratory test program conducted to evaluate the potential use of gypsum mining waste to produce CEBs. The hardened properties such as compressive strength, flexural strength, and water absorption was investigated.

Subordinately, test results may provide a means to reduce a waste disposal problem while providing the construction industry with a new, useful, low cost raw material. Based on the experimental tests conducted on the CEBs, the following conclusions can be drawn: The liquid and plastic limits of the clay-rich material are appropriated for the production of CEBs, although it is advisable to test a number of natural fibers to increase compressive and flexural strength and to avoid excessive cracking.

Clay-rich material correspond to a granular soil, with > 50% of sand and gravel size, but the soil used is a sandy soil because > 50% of the coarse fraction is < 4.75 mm (No. 4 mesh ASTM). According to this and the behavior of the fine fraction of the soil, it classified as a clayed silty sand soil, settling near the boundary line suitable for the preparation of CEBs. The chemical composition of the gypsum mining waste reveals that the elemental content would be suitable in principle for chemical stabilization, avoiding a waste with high levels of visible gypsum as this could create adverse conditions for the development of CEBs. Non-stabilized CEBs showed values of compressive strength up to 0.251 MPa, which are below recommended limits. However, CEBs from the trial T3 (2.5% of gypsum mining waste and 3% of lime), the compressive strength was improved by up to 500% (5 times) reaching values of 1.574 MPa, that is within the minimum range required by Colombian construction standards.

Stabilized CEBs showed much better values of modulus of rupture compared with those obtained from nonstabilized CEBs. CEBs from the trial T5(5% of gypsum mining waste and 3% of lime), showed the highest values of MR, achieving high levels of rigidity, although in the compressive strength test they are lower than those obtained for CEBs from the trial T3.

CEBs containing 10% gypsum mining waste showed compressive strength values lower than those obtained for CEBs containing 5 or 2.5% gypsum mining waste.

Non-stabilized CEBs from trials T2, T4 and T6, showed a slight improvement in the engineering properties with respect to Non-stabilized CEBs from trials T1, although not as pronounced as observed in lime stabilized CEBs. A significant improvement was displayed by lime stabilized CEBs in extremely humid conditions, retaining their shape after being submerged in water 24 hours that confirms the activating ability of lime to generate reactions cementing between the clay-rich material and gypsum mining waste.

Non-stabilized CEBs, containing gypsum mining waste in several percentages, after water absorption, showed a completely unacceptable behavior; they completely disintegrated, making them unsuitable in extreme conditions.

Figure 12. (a) CEB before soaking in water. (b) CEB during soaking in water. (c) CEB after soaking in water. (d) CEB during compressive strength test Source: The authors.

49

Jaramillo-Pérez et al / DYNA 81 (188), pp. 42-51. December, 2014.

The results of this study reveal that the engineering properties of the CEBs were not satisfactory in the criterion of authors, suggesting additional experimental work to improve the engineering properties of CEBs.

[18]

ASTM D4318-10. Standard test methods for liquid limit, plastic limit, and plasticity index of soils. American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2010.

[19] ASTM D2487-11. Standard practice for classification of soils for engineering purposes (Unified Soil Classification System). American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2011.

[20] ASTM D1557-12. Standard test methods for laboratory compaction characteristics of soil using modified effort (56,000 ft-lbf/ft3 (2,700 kNm/m3)). American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2012.

[21] Pérez, P. y Pachón C., Determinación de los módulos elástico, plástico y de rotura en material para tapia pisada. Tesis de Grado, Universidad Industrial de Santander, Bucaramanga, Colombia, 98p, 2003.

[22] ASTM C67-11. Standard test methods for sampling and testing brick and structural clay tile. American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2011.

[23] Masonry standards joint committee. [Online],[date of reference November 25th of 2012]. Available at: http://www.acronymfinder.com/MasonryStandards-Joint-Committee-(ACI)-(MSJC).html [24] Galán-Marín, C., Rivera-Gómez, C. and Petric, J., Clay-based composite stabilized with natural polymer and fibre. Construction and Building Materials, 24 (8), pp.

1462-1468, 1991. http://dx.doi.org/10.1016/j.conbuildmat.2010.01.008 [25] Krosnowski, A.D., A proposed best practice method of defining a standard of care for stabilized compressed earthen block production. MSc. Thesis, University of Colorado, Boulder, USA, 2011.

[26] Morel, J.-C., Pkla, A. and Walker, P., Compressive strength testing of compressed earth blocks. Construction and Building Materials, 21, pp.

303-309, 2007. http://dx.doi.org/10.1016/j.conbuildmat.2005.08.021 [27] Demir, I., Effect of organic residues addition on the technological properties of clay bricks. Waste Management, 28, pp. 622-627, 2008 http://dx.doi.org/10.1016/j.wasman.2007.03.019 [28] Niño, M.C. y Spinosi, V.C., Caracterización fisicoquímica y mecánica de suelos residuales utilizados en la construcción de ecoviviendas con tecnologías sostenibles en el Munipicio de Girón, Santander. Tesis de Grado, Universidad Industrial de Santander, Bucaramanga, Colombia, 2011.

[29] Niño, M.C., Spinosi, V.C., Ríos, C.A. and Sandoval, R., Effect of the addition of coal-ash and cassava peels on the engineering properties of compressed earth blocks. Construction and Building Materials, 36, pp.

276-286, 2012. http://dx.doi.org/10.1016/j.conbuildmat.2012.04.056

Acknowledgments This research forms part of the undergraduate thesis of E. Jaramillo and J. Plata. The authors acknowledge Andina Ingeniería Ltda. for their laboratory facilities and Polypus of Colombia for field work support. We are indebted to Universidad Industrial de Santander for providing research facilities. The authors also acknowledge to the anonymous referees for their critical and insightful reading of the manuscript and are most grateful to the above-named people and institutions for support.

References [1]

[2]

[3] [4] [5] [6] [7]

[8] [9] [10] [11] [12] [13]

[14] [15] [16] [17]

Oti, J.E., Kinuthia, J.M. and Bai, J., Compressive strength and microstructural analysis of unfired clay masonry bricks. Engineering Geology, 109, pp.

230-240, 2009. http://dx.doi.org/10.1016/j.enggeo.2009.08.010 Yamín-Lacouture, L.E., Phillips-Bernal, C., Reyes-Ortíz, J.C. y RuizValencia, D., Estudio de vulnerabilidad sísmica, rehabilitación y refuerzo de casas de adobe y tapia pisada a nivel nacional. Centro de Estudios de Desastres y Riesgos (CEDERI), Universidad de los Andes, Bogotá, Colombia, 2007.

Oti, J.E., Kinuthia, J.M. and Bai, J., Engineering properties of unfired clay masonry bricks. Engineering Geology, 107, pp. 130-139, 2009. http://dx.doi.org/10.1016/j.enggeo.2009.05.002 Guillaud, H., Joffroy, T. and Odul, P., Compressed earth blocks, in: Manual of design and construction Vol. 2. Eshborn: Vieweg, 1995.

Deboucha, S. and Hashim, R., A review on bricks and stabilized compressed earth blocks. Scientific Research and Essays, 6 (3), pp. 499506, 2011.

Guettala, A., Houari, H., Mezghiche, B. and Chebili, R., Durability of lime stabilized earth blocks. Courrier du Savoir, 2, pp. 61-66, 2002.

Mckinley, J.D., Thomas, H.R., Williams, K.P. and Reid, J.M. Chemical analysis of contaminated soil strengthened by the addition of lime.

Engineering Geology, 60 (1-4), pp.

181-92, 2011. http://dx.doi.org/10.1016/S0013-7952(00)00100-9 Rao, S.M. and Shivananda, P., Role of curing temperature in progress of lime-soil reactions. Geotechnical and Geological Engineering, 23 (1), pp.

79-85, 2005. http://dx.doi.org/10.1007/s10706-003-3157-5 Chandra, S., Waste materials used in concrete manufacturing. New Jersey: Noyes Publications Westwood, 1977.

ASTM E2392/M-10. Standard guide for design of earthen wall building systems. American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2010.

Walker, P., Specifications for stabilised pressed earth blocks. Masonry International, 10 (1), pp. 1-6, 1996.

Centre for the Development of Enterprise. Compressed earth blocks testing procedures, CDE, Brussels, Belgium, 2000.

Jaramillo, E.R. y Plata, J.M., Caracterización de geomateriales y evaluación de su uso en la preparación de materiales para la construcción de vivienda sustentable en Bucaramanga. Tesis de grado, Universidad Industrial de Santander, Bucaramanga, Colombia, 2012.

ASTM C136-06. Standard test method for sieve analysis of fine and coarse aggregates. American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2006.

ASTM D1140-00. Standard test methods for amount of material in soils finer than No. 200 (75-μm) Sieve. American Society for Testing Materials, West Conshohocken, Pennsylvania, PA 19428, USA, 2006.

Houben, H., Rigassi, V. and Garnier, P., Compressed earth blocks production equipment, CDI and CRATerre-EAG, Brussels, 1994.

Cuéllar, A., Mesa, F.A., Vargas, C. y Perilla, J.E., Arcillas modificadas y caracterizadas por micro-raman y difracción de rayos X. DYNA, 77 (164), pp. 39-44, 2010.

E.R. Jaramillo-Pérez, received the BSc in Geology in 2012 from the Universidad Industrial de Santander, Bucaramanga, Colombia. She has worked since 2012 as a consultant for many companies within the petroleum industry, particularly with the Instituto Colombiano del Petróleo (ICP), basically oriented to the development and project management, integration and management of geological information and sedimentological studies.

J.M. Plata-Chaves, received the BSc in Geology in 2012 from the Universidad Industrial de Santander, Bucaramanga, Colombia. He has been working as Junior Geologist in Digital Rock Physics Laboratory since 2012, specifically in the field of TAC and core description applied to sedimentology and oil industry.

C.A. Ríos-Reyes, received the BSc in Geology in 1989 and the Postgraduate Diploma in University Teaching in 1995 from the Universidad Industrial de Santander, Bucaramanga, Colombia. The Shimane University, Matsue, Japan, conferred on him the degree of MSc in Geology in 1999.

The University of Wolverhampton, Wolverhampton, Inglaterra, conferred him the degree of PhD in Applied Sciences in 2008. He has been working as a full-time Lecturer of the School of Geology, in the Universidad Industrial de Santander, Colombia since 1992, where he developed his professional experience at University teaching level during the last 22 years on the field of Mineralogy, Metamorphic Petrology and fieldworks on crystalline basement complexes in different areas of Colombia.

Actually, he is the director of the Research Group in Basic and Applied Geology at the School of Geology of the Universidad Industrial de Santander and the director of the Microscopy Laboratory of the Guatiguará 50

Jaramillo-Pérez et al / DYNA 81 (188), pp. 42-51. December, 2014.

Technological Park. He is specialist in mineralogy, experimental geology, petrology and geochemistry of metamorphic rocks and has extensive research experience in geological mapping, experimental and environmental mineralogy and metamorphic petrology.

Área Curricular de Ingeniería Geológica e Ingeniería de Minas y Metalurgia Oferta de Posgrados    

Especialización en Materiales y Procesos Maestría en Ingeniería - Materiales y Procesos Maestría en Ingeniería - Recursos Minerales Doctorado en Ingeniería - Ciencia y Tecnología de Materiales

Mayor información: Néstor Ricardo Rojas Reyes Director de Área curricular acgeomin_med@unal.edu.co (57-4) 425 53 68

51

Practical lessons learnt from the application of X-ray computed tomography to evaluate the internal structure of asphalt mixtures Allex Eduardo Alvarez-Lugo a & Juan Sebastián Carvajal-Muñoz b a

University of Magdalena, Santa Marta, Colombia, allexalvarez@yahoo.com University of Magdalena, Santa Marta, Colombia, juancarvajal@tamu.edu

b

Received: September 27th, 2013. Received in revised form: June 24th, 2014. Accepted:, September 16th 2014.

Abstract X-ray Computed Tomography (X-ray CT) has allowed for the efficient non-destructive characterization of the internal structure of paving asphalt mixtures (AM), and has led to multiple practical lessons learnt based on the analysis of laboratory- and field-produced AM. This paper aims at summarizing these practical lessons, to facilitate their future application and further developments, in terms of: (i) fabrication of laboratory specimens, (ii) comparison of laboratory- and field-compacted mixtures, (iii) comparison of hot-mix asphalt and warm-mix asphalt mixtures, (iv) effects of additives, temperature, and compaction, (v) stone-on-stone contact, (vi) relationship between internal structure and performance, and (vii) modeling applications. These practical lessons are primarily gathered from the analysis of the air void distribution of laboratory-and field-produced AM, evaluated through X-ray CT, which has led to relevant inputs for the assessment of the response and performance of AM. X-ray CT enables computation of the AM internal structure with multiple practical applications and future opportunities to enhance the microstructure of AM and, consequently, optimize their performance.

Key words: X-ray computed tomography; internal structure; air voids; asphalt mixture; pavements engineering.

Lecciones prácticas aprendidas a partir de la aplicación de la tomografía computarizada de rayos-x para evaluar la estructura interna de mezclas asfálticas Resumen La Tomografía Computarizada de rayos-X (TC-rX) ha permitido una eficiente caracterización no destructiva de mezclas asfálticas (MA) de pavimentación y ha generado múltiples lecciones prácticas aprendidas a partir del análisis de MA producidas en campo y laboratorio. El presente artículo tiene como objetivo resumir las lecciones prácticas aprendidas, con el fin de facilitar su aplicación futura y desarrollos posteriores, en términos de: (i) fabricación de especímenes de laboratorio, (ii) comparación de mezclas compactadas en campo y laboratorio, (iii) comparación entre mezclas asfálticas en caliente y mezclas tibias, (iv) efectos de aditivos, temperatura y compactación, (v) contacto agregado-agregado, (vi) relación entre la estructura interna y el desempeño, y (vii) aplicaciones de modelación. Estas lecciones prácticas se recopilaron principalmente a partir del análisis de la distribución de vacíos en MA producidas en campo y laboratorio, evaluadas a través de TC-rX, el cual generó información relevante para evaluar la respuesta y el desempeño de MA. La TC-rX hizo expedito el cálculo de la estructura interna de MA con múltiples aplicaciones prácticas y posibilidades futuras para mejorar la microestructura de MA y, consecuentemente, optimizar su desempeño.

Palabras clave: tomografía computarizada de rayos-X; estructura interna; vacíos con aire; mezclas asfálticas; ingeniería de pavimentos.

1. Introduction The paving industry has taken advantage of the research conducted in material characterization to produce high-quality asphalt mixes (AM) that allow for good performance, comfort, and safety for road users over time. One of those research topics concerns the use of X-ray Computed Tomography (Xray CT) to determine the internal structure (or microstructure)

of paving materials. Considering the relevance of X-ray CT in the research on AM [1,2], this paper aims at synthesizing some practical lessons learnt from the application of X-ray CT, mostly based on the computation of air void (AV) related characteristics, to facilitate their future application and future developments of AM. The first section of the paper includes a brief description of X-ray CT and image analysis techniques. Next, seven

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 52-59. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.40042

Alvarez-Lugo & Carvajal-Muñoz / DYNA 81 (188), pp. 52-59. December, 2014.

AM specimen and the detector receives the remaining energy that got through the specimen and sends the data to a processing unit.

2. The processing unit transforms the data into 2D-grayscale images (raw images) with a specific resolution N_mm/pixel that depends on the equipment configuration and characteristics.

3. The raw 2D images are analyzed by means of specialized software (e.g., Image J [4], Image-Pro® Plus [5], and iPas [6]) based on mathematical algorithms to compute the internal structure’s characteristics.

Further analyses can be performed based on the internal structure’s characteristics; for instance, subsequent studies can be developed to assess the internal structure’s influence on the AM performance by means of predictive mathematical models (Section 9).

By means of image analysis, the internal structure of AM can be quantified in terms of both the AV characteristics and aggregate characteristics. The AV evaluation include, for example, computation of the total air void (TAV) content, air void size (AVR), connected air void (CAV) content, and ratio of total to connected AV content (TAV/CAV). The assessment of aggregate characteristics can include computation of its size distribution, contacts (stone-on-stone contact), and orientation.

Based on the aforementioned indexes, the following sections summarize practical lessons learnt from the application of X-ray CT and image analysis to evaluate the internal structure of AM.

Figure 1. Evaluation of the internal structure of AM by means of X-ray CT Source: The Authors..

sections introduce the main practical lessons documented with corresponding suggestions for future research. The paper is completed with a section on conclusions and recommendations.

2. X-ray CT and Image Analysis Essentially, X-ray CT is a technology used to generate images of density distribution for the cross sectional area of a material specimen. The device is composed by an X-ray source with collimator, a rotating-specimen holder and a detector with a black filter (Fig. 1). The technology uses the ability of short wavelength radiation to pass through objects and measures the difference between X-ray intensities before and after passing through.

More specifically, when the initial radiation (I0) passes through the object—material specimen—, some of it is absorbed and scattered while the other portion (I) gets through and, it is consequently, detected. The amount of penetration and absorption are a function of the: a) linear attenuation coefficient, which depends on the material density, b) X-ray’s energy, c) atomic number, d) density, and e) object thickness [3]. The generation of X-ray CT images involves mathematical processes for converting the different intensities’ measurements into two dimensional slice images (2D) in grayscale (i.e., consisting of 256 ranges of gray intensities, from 0-black to 255-white).

Consequently, the resulting image is a depiction of the varying spatial densities within the object planes, which subsequently allows for identifying its internal structure [3].

Although this technology was originally developed for medical purposes, paving engineering discovered its potential applications for materials research and adapted it.

Therefore, in the last decade and a half, X-ray CT has been frequently used to assess the internal structure of AM produced under varying conditions in both the laboratory and field.

The image acquisition and analysis for AM can be performed in three basic steps (Fig. 1): 1. A source of X-ray energy emits pulses to irradiate the

3. Fabrication of Laboratory Specimens Recent studies [7-12] have systematically concluded that fabrication of laboratory-compacted specimens using the Superpave gyratory compactor (SGC) should be modified in order to reduce the high vertical- and horizontalheterogeneity in the internal structure [7-11,13-16] (e.g., vertical and horizontal distributions of the: a) TAV content, b) AVR, and c) CAV content).

Fig. 2 summarizes the typical vertical distribution of TAV content values previously reported for SGC specimens compacted at different heights for three types of AM. These heterogeneous distributions show that the top- and bottomportions of the specimens have the highest TAV content values as compared to the more homogeneous central portions. In addition, the “shape” of the vertical distribution of TAV content in the SGC specimens is related to the specimen height as supported by Fig. 2c, d, e, and f. The mixture characteristics can also be related to the vertical distribution of AV content [17].

Corresponding specific recommendations for improvement of the vertical homogeneity of SGC specimens include cutting 10 to 20 mm from the top- and bottom-portions of the specimens—with proper consideration of the nominal-maximum aggregate size coverage requirements and the aspect ratio of the specimen—. These recommendations were suggested for improvement of SGC specimens of: a) permeable friction course (PFC) (or open-graded friction course, OGFC) mixtures [9,10], b) dense-graded hot mix asphalt (HMA) 53

Alvarez-Lugo & Carvajal-Mu単oz / DYNA 81 (188), pp. 52-59. December, 2014.

mixtures [8,10,18,19], and c) dense-graded warm mix asphalt (WMA) mixtures [11,12]. Similarly, Muraya [10] proposed the reduction of the height of SGC specimens

from 150 to 121 mm to enhance the homogeneity of stone matrix asphalt- (SMA), dense graded-, and porous-mixtures.

PFC MIXTURES

Position (mm)

Position (mm)

Air Voids Content (%)

Air Voids Content (%)

a) Specimen height : 115+5 mm. Source: [13]

b) Specimen height : 115+5 mm. Source: [9]

DENSE-GRADED HMA MIXTURES

Position (mm)

Position (mm)

Air Voids Content (%)

Air Voids Content (%)

d) Specimen height: 115 mm. Source: [14]

Position (mm)

Position (mm)

c) Specimen height: 64 mm. Source: [11]

Air Voids Content (%)

Air Voids Content (%)

e) Specimen height: 150-160 mm. Source: [8,15]

f) Specimen height: 165-182 mm. Source: [10]

SMA MIXTURES

Position (mm)

Position (mm)

Air Voids Content (%)

Air Voids Content (%)

g) Specimen height: 115 mm. Source: [7]

h) Specimen height: 165-182 mm. Source: [10] Figure 2. Typical vertical distributions of TAV content in different types of AM produced using the SGC. 54

Alvarez-Lugo & Carvajal-Muñoz / DYNA 81 (188), pp. 52-59. December, 2014.

determined using either the SGC or the linear kneading compactor (LKC). However, additional research is needed to comprehensively relate the laboratory- and fieldcompaction methods and, hence, ensure more homogeneity in the AM microstructure.

As a result of the previous conclusions, it is crucial that further research focus on developing compaction methodologies that allow for better comparisons between the internal structures of laboratory- and field-compacted AM to: a) improve their compatibility, b) allow for a more homogeneous internal structure, and c) eventually, ensure similar or superior performance.

In terms of reducing the horizontal heterogeneity of SGC compacted specimens, Muraya [10] and Dubois et al.

[17] suggested coring the outside part of the specimens (i.e., dense-graded HMA and PFC mixtures). This suggestion is consistent with the recommendations reported for PFC mixtures [9] of coring from 152 to 102 mm, which was supported in data similar to that shown in Fig. 2b. This figure compares the typical heterogeneous distribution of TAV content for three concentric cores in a PFC specimen, where the e-ring 1 corresponds to the innermost core of the specimen. Similarly, differences between the AV content in the core and the outer portions of Marshall specimens were previously reported [16].

Nevertheless, the coring and sawing processes may induce variations due to the enlargement or filling of AV with fine-aggregate particles and asphalt mastic. As a result, future research was suggested [9] to fully validate the effects of coring PFC mixtures given the possibility of rearranging the internal structure during this process.

The aforementioned modifications in the specimen fabrication process—leading to a more homogeneous internal structure—would allow for more representative results in the laboratory tests performed for mixture design and evaluation. In fact, some authors [20-23] emphasized the need to produce specimens with uniform AV distributions to enhance testing repeatability based on avoiding high variations of stress and strain in a particular specimen.

5. Microstructure of HMA- and WMA-Mixtures The limited information available at this time [7,11,12,30] on the comparison of the microstructure of laboratory-compacted (i.e., SGC specimens) HMA- and WMA-mixtures leads to conclude that they are comparable in terms of the vertical distribution of: TAV content, CAV content, and AVR. More specifically, some of the WMA additives evaluated (i.e., Asphamin®, Evotherm®, Rediset®, and Sasobit®) [7,11,12,30], and corresponding fabrication processes of WMA mixtures, induced limited differences in the mixture internal structure as compared to those obtained with the conventional—control—HMA mixtures.

In addition, Masad et al. [31] concluded that the use of Evotherm® allowed for a more uniform distribution of TAV and AVR as compared to that obtained in HMA mixtures fabricated using unmodified binders. Estakhri et al. [30] concluded that the use Evotherm® and Aspha-min® induced changes in the compactability and thus, influenced a particular internal structure that could be equivalent to that of HMA mixtures depending on the materials used for its production. However, this conclusion is not warranted due to the limited research conducted on the WMA’s internal structure.

At this point, more research is required to further assess the influence of different aggregate gradations, specimen height and diameter, and the use of additional WMA additives and compaction methods on the microstructure of WMA-mixtures.

4. Laboratory- and Field-Compacted Mixtures The studies available [7,9,14,19,24-28] on the comparison of laboratory-compacted and field-compacted AM—assessed through analysis of road cores— systematically concluded that minimum or sometimes even no similarities can be found between their microstructure.

This conclusion was substantiated based on the analysis of road cores and SGC compacted specimens of HMA mixtures (i.e., dense-graded [7,14,24-26,28], PFC [9], and SMA [27]) and WMA mixtures (i.e., dense-graded [7]). The data in Fig. 2g exemplify this aspect including the comparison of both SGC compacted specimens and road cores, identified as “Lab” and “RC”, respectively. In fact, the vertical AV distributions showed dissimilar tendencies that do not allow for further macroscopic comparisons.

Additional information on the distribution of TAV in road cores was reported by Masad et al. [29].

Recently, Dubois et al. [17] reported homogeneous internal structures (i.e., TAV content and AVR in dense-, open- and gap-graded HMA mixtures) and improved comparability with field-compacted specimens using a laboratory roller compactor. Similarly, the use of double static compression permitted homogeneity in longitudinal— vertical—AV contents, although they exhibited larger transverse scattering as compared to gyratory compaction [17].

On the other hand, Masad and Button [14] concluded that a relationship between field- and laboratory-compaction is available for determining the required field compaction effort based on the slope of a laboratory compaction curve

6. Effects of Additives, Temperature, and Compaction 6.1. Additive effects Currently, a wide range of additives are available in the global market for the production of modified asphalts, and more recently of WMA mixtures. In addition, the effects of using additives on the internal structure of AM need to be accounted for, since they may differ from one mix to another as a result of factors including the: a) type of mix, b) type of additive, c) additive-asphalt-aggregate compatibility, and d) mixture fabrication protocol.

As a result, in terms of AM microstructure, the main practical lesson concerns the careful selection of the additive for the production of AM in accordance with the materials involved in its production and the specific mixture design. For instance, according to Wegan and Nielsen [32] 55

Alvarez-Lugo & Carvajal-Muñoz / DYNA 81 (188), pp. 52-59. December, 2014.

produce AM with proper performance [33].

Previous research [14,17,19,34] concluded that the compactor type and its mode of energy application, as well as the compaction energy influence to a great extent the internal structure of AM. Regarding the type of compactor, Harvey and Monismith [35] reported dissimilar mechanical responses when using different compaction methods (i.e., rolling wheel, kneading compaction, and gyratory compaction), since they induced different internal structures. In general, the strongest conclusion of their study regards the disability to interchange different compaction methods in the laboratory due to their strong influence on AM performance.

More recently, Jonsson et al. [36] compared the rolling wheel, Marshall, and gyratory compaction using X-ray CT and concluded that none of these methods produced homogeneous distributions of TAV content. The Marshall compaction led to the most homogenous horizontal and vertical distributions of TAV [36]. In terms of the gyratory compaction, numerous studies [8-10,13,14,17] have reported the production of heterogeneous distributions of TAV content —as shown in Fig. 2 and previously discussed—and CAV content [9].

According to Verdi et al. [37], the AM final density depends principally on the number of gyrations and less on the vertical pressure applied by the gyratory compactor. In addition, recent research [10,17] suggested that the number of gyrations cannot be used as a control parameter for compaction due to the scattered AV distributions.

In this regard, although “theoretically” higher compaction efforts are used to get homogeneous distributions of AV, this case does not often occur, because of external factors affecting the aggregates packing in the AM [31]. Fig. 2f and 2h and Fig. 3 show data supporting the effect of the compaction energy—number of gyrations—for specimens fabricated with different mixture types and at two heights. Similar results were reported for PFC mixtures [9].

Future studies should be able to determine which types of compaction—equipment and methods—are more efficient for obtaining homogeneous AM specimens in both the laboratory and field.

the effect of using polymer modified asphalt on the microstructure of AM depends on the type of neat binder, the mixture gradation, as well as the temperature and mode of compaction.

In addition, the comparison of PFC mixtures produced with polymer modified asphalt and asphalt rubber [13], led to conclude that these mixtures are different materials with particular internal structures that require the use of different fabrication protocols (i.e., specifications for mix design), as well as specific handling and control techniques.

Consequently, additives play a specific and relevant role on the AM production and its internal structure, which should be further studied in order to better define the most appropriate additives doses, types, and conditions for a homogeneous internal structure, and high-quality performance in the field.

6.2. Temperature effects The mixture temperature during compaction in both the field and laboratory has to be carefully controlled to minimize the temperature gradients, which can lead to diverse heterogeneous microstructures in AM. However, additional information is still required to clearly assess the effect of the temperature gradients, especially in field compaction conditions.

In the laboratory, based on the compaction of SGC specimens, Masad et al. [31] concluded that more homogeneous temperature profiles can lead to more homogeneity in the internal structure (i.e., vertical distribution of TAV content). However, the relationship was reported as weak and does not warrant improved homogeneity. In addition, the authors concluded that the heterogeneity is not only the result of the temperature profile, but it is also due to the friction against the metallicmold boundaries.

Research on WMA mixtures [11] concluded that different compaction temperatures generated internal structures—TAV content and AVR—with insignificant statistical differences (see Fig. 2c; specimens fabricated using Evotherm®). On the contrary, Estakhri et al. [30] stated that different laboratory compaction temperatures induced changes in the TAV content of both HMA- and WMA-mixtures (i.e., fabricated using Evotherm DAT®, Sasobit®, and Advera®), which is due to the differences among the physico-chemical properties of the additives used.

Future research should look for improved comparisons between different types of AM to determine the most suitable conditions of temperature control to promote homogeneity in the internal structure of both field- and laboratory-compacted AM. 6.3. Compaction effects Compaction is one of the most critical variables to be considered in the design and construction of AM, since both the density and microstructure acquired by AM is related to their performance. Therefore, adequate compaction control in both the field and laboratory has to be conducted to

Figure 3. Vertical distribution of TAV content under varying compaction energies in the SGC. Source: [14].

56

Alvarez-Lugo & Carvajal-Muñoz / DYNA 81 (188), pp. 52-59. December, 2014.

field- and laboratory-specimens, and the Hamburg PD test results were more affected by the AV content rather than by the microstructure. However, specimens with a uniform distribution of TAV content had less variation in the Overlay test (i.e., cracking resistance) than those with a heterogeneous microstructure.

Studies on PFC mixtures [9,46] concluded that their particular responses to mechanical tests, used to assess the mixture performance as a result of their specific gradations, type of binders, and mix design, are highly dependent on the internal structure and exhibited different responses to those typically identified for dense-graded HMA. In years to come, research on assessing performance of AM should further account for the microstructure to determine their relationship. This will lead to a better understanding on the best scenarios for obtaining the performance pursued when designing different types of AM.

7. Stone-on-Stone Contact The arrangement—microstructure—of the structural skeleton formed by the aggregates in the compacted AM determines to a vast extent its mechanical response. This aspect is particularly critical to ensure the proper performance of PFC- and SMA-mixtures based on a fully developed condition of stone-on-stone contact in the coarse aggregate. This contact condition is mostly evaluated conducting macroscopic evaluations according to the VCA—voids in the coarse aggregate—method [38,39].

Additional research [40], based on the application of X-ray CT and image analysis, further validated the VCA method.

However, the authors [40] highlighted the advantage of image analysis over the VCA method, since it can allow the quantification of the number of contact points as well as optimizing the aggregate gradation.

As in the case of the TAV- and CAV-content, a homogeneous vertical distribution of the number of contact points is desirable. However, a recent study based on X-ray CT and image analysis showed the existence of heterogeneous distributions of contact points (as an index of stone-on-stone contact) in SGC compacted specimens of PFC mixtures [41]. Thus, the need for a homogeneous microstructure still includes both the AV- and aggregatesarrangement.

Recent research [42-45] also determined the influence of stone-on-stone contact and aggregates characteristics on the internal structure and corresponding mechanical response of the dense-graded and SMA mixtures. In particular, the main practical lesson derived from these studies [42-45] concerns maximizing the number of contact points for a stiff structural skeleton capable of providing a good mixture performance. The ways of obtaining these may include use of adequate compaction: a) vertical pressures and b) angles—gyratory compactor—to get an appropriate orientation of the aggregates and enhance the mechanical response.

Future research should focus on determining the relationship between the aggregate geometric characteristics (e.g., evaluated using nondestructive technology) and the internal structure of AM to obtain conclusive results that can be used as inputs for mix design, modeling, and field applications.

9. Modeling Applications The X-ray CT outputs constitute a vast amount of relevant information for the development of mathematical models used to determine the possible scenarios at which the pavement materials could fail, be susceptible to damage or develop inappropriate performance in the field.

Therefore, the integration of X-ray CT and mathematical models turn out to be critical and essential to have a deeper understanding on the AM characteristics and account for new approaches in their design, production, and placement in the field.

In response to this need of understanding the relation between the mixture’s internal structure and the distresses affecting the AM, numerous mathematical-modeling applications have been accomplished by means of the information available from X-ray CT. Ultimately, the main objective of these studies has to do with finding accurate simulations of field response, based on inputs from the laboratory, and get valid information for improving the design, construction, and forensic evaluation of AM [14].

In particular, based on X-ray CT images, Caro [47] proposed a coupled micromechanical model to evaluate moisture damage induced in AM. This research, along with similar studies [48-50] led to a more comprehensive analysis of the microstructure effects on the performance and damage phenomena of AM as compared to the conventional phenomenological approaches applied for their study. One of the strongest recommendations offered by the authors concerns the advance in innovative probabilistic models and enhancement of the existing ones to have more accurate predictions of moisture induced damage and have improved predictions of the AM field performance.

Mathematical models have also been proposed to predict the fatigue life, aging process, influence of chemical substances [51], and permeability of dense-graded HMA [29] and PFC mixtures [52]. Additional information on similar studies related to modeling and simulation on AM can be found in work by Liu et al. [51].

All the efforts and innovation conducted in the modeling of AM damage by means of non-destructive techniques

8. Relationship between Internal Structure and Performance The AM field performance largely depends on the mixture microstructure [8,14]. Previous research [31] stated that gyratory-compacted specimens exhibited the poorest mechanical response, evaluated by means of the shear permanent deformation (PD) test, while linear kneadingcompacted specimens had the highest resistance to PD, and the rolling wheel-compacted specimens reported intermediate results. The differences in the internal structure of the specimens—exemplified in Fig. 2d—can partially explain their diverse mechanical response.

In addition, Masad et al. [31] concluded that similar TAV contents can lead to comparable rut depth between 57

Alvarez-Lugo & Carvajal-Muñoz / DYNA 81 (188), pp. 52-59. December, 2014.

constitute a highly valuable element of research that allows for future pavement structures with enhanced durability, better functionality, and performance. Therefore, there is a long way to go in this regard for future research.

[3]

10. Conclusions and Recommendations

[6]

This paper summarized the main practical lessons learnt from the use of X-ray CT and image analysis for the evaluation of the internal structure of AM (mainly based on the analysis of the distributions of AV content). Based on the information presented, the following conclusions and recommendations are offered:  X-ray CT and image analysis enable the computation of the internal structure for AM, which have allowed a better understanding of AM characteristics and response including, for example, the nondestructive analysis of the AV characteristics and the computation of aggregates distribution, orientation, and contact.  This feasible quantification of the AM microstructure, based at present primarily on the analysis of AV characteristics, supported extensive analysis of both laboratory- and field-compacted AM leading to multiple practical applications as previously documented.  The control of the variables affecting the AM internal structure (e.g., asphalt-, aggregate-, and additiveproperties, mixture type, production methods, temperature, and compaction) is to be further studied to define the conditions to ensure an optimum microstructure and, consequently, an optimum AM performance.  In the coming years, the advancement of non-destructive material characterization techniques should allow for a more comprehensive evaluation of the AM microstructure that lead to better mixture designs and higher durability over time in the field. For this purpose, more detailed studies of internal structure of different AM need to be performed as well as development of new devices and equipment with higher accuracy and representativeness.

[7]

[4] [5]

[8]

[9]

[10] [11] [12] [13]

[14] [15] [16]

[17]

11. Disclaimer

[18]

This paper does not constitute a standard, specification, nor is it intended for design, construction, bidding, contracting, or permit purposes. Trade names were used solely for information and not for product endorsement.

[19]

Acknowledgements [20]

The authors thank the support provided by the Viceresearch Office at the University of Magdalena for the successful completion of this study.

[21]

References [1] [2]

Shashidhar, N. X-ray tomography of asphalt concrete.

Transportation Research Record, 1681, pp. 186-192, 1999. http://dx.doi.org/10.3141/1681-22 Masad, E. X-ray computed tomography of aggregates and asphalt mixes. Materials Evaluation Journal, 62 (7), pp. 775-783, 2004.

[22]

58

Hassan, A. Microstructural characterisation of rubber modified asphalt mixtures, Doctoral Thesis, University of Nottingham, 2012.

Rasband, W. Imagej 1.41o. National Institutes of Health, Available at: http://imagej.nih.gov/ij/, Retrieved on June 2011.

Media Cybernetics, L. P. Image-pro® plus, version 5.1.2. Georgia, M.D., 1999.

RILEM TG-2; The University of Wisconsin-Madison; Michigan State University. Ipas - image processing & analysis system. 2011.

Alvarez, A. E., Carvajal, J. S., Reyes, O. J., Estakhri, C. and Walubita, L. F., Image analysis of the internal structure of warm mix asphalt (wma) mixtures, Proceedings of TRB 91st Annual Meeting, pp. 1-17, 2012.

Walubita, L. F., Jamison, B., Alvarez, A. E., Hu, X. and Mushota, C. Air void characterisation of hma gyratory laboratory-moulded samples and field cores using x-ray computed tomography (x-ray ct). Journal of the South African Institution of Civil Engineering, 54 (1), pp. 22-31, 2012.

Alvarez, A. E., Epps Martin, A. and Estakhri, C. Internal structure of compacted permeable friction course mixtures. Construction and Building Materials, 24 (6), pp.

1027-1035, 2010. http://dx.doi.org/10.1016/j.conbuildmat.2009.11.015 Muraya, P. M. Homogeneous test specimens from gyratory compaction. International Journal of Pavement Engineering, 8 (3), pp. 225-235, 2007. http://dx.doi.org/10.1080/10298430701289323 Alvarez, A. E., Carvajal, J. S. and Reyes, O. J. Internal structure of laboratory compacted warm mix asphalt. DYNA, 79 (172), pp. 3845, 2012.

Alvarez, A. E., Macias, N. and Fuentes, L. G. Analysis of connected air voids in warm mix asphalt. DYNA, 79 (172), pp. 29-37, 2012.

Alvarez, A. E., Fernandez, E. M., Epps Martin, A., Reyes, O. J., Walubita, L. F. and Simate, G. S. Comparison of permeable friction course mixtures fabricated using asphalt rubber and performancegrade asphalt binders. Construction and Building Materials, 28 (1), pp.

427-436, 2012. http://dx.doi.org/10.1016/j.conbuildmat.2011.08.085 Masad, E. and Button, J. Implications of experimental measurements and analyses of the internal structure of hot-mix asphalt.

Transportation Research Record, (1891), pp. 212-220, 2004.

Alvarez, A. E., Carvajal, J. S. and Walubita, L. Comparison of internal structure of different hot mix asphalt types. Ingeniare.

Revista chilena de ingeniería, 22 (1), pp. 74-87, 2014.

Partl, M., Flisch, A. and Jonsson, M. Comparison of laboratory compaction methods using x-ray computer tomography.

International Journal of Road Materials and Pavement Design, 8 (2), pp.

139-164, 2007. http://dx.doi.org/10.1080/14680629.2007.9690071 Dubois, V., De La Roche, C. and Burban, O. Influence of the compaction process on the air void homogeneity of asphalt mixtures samples. Construction and Building Materials, (24), pp. 885-897, 2010. http://dx.doi.org/10.1016/j.conbuildmat.2009.12.004.

Tashman, L., Masad, E., D'Angelo, J., Bukowski, J. and Harman, T. X-ray tomography to characterize air void distribution in superpave gyratory compacted specimens. The International Journal of Pavement Engineering, 3 (1), pp.

19-28, 2002. http://dx.doi.org/10.1080/10298430290029902a Partl, A., Flisch, A. and Jönsson, M. Gyratory compaction analysis with computer tomography. International Journal of Road Materials and Pavement Design, 4 (4), pp. 401-422, 2003. http://dx.doi.org/10.1080/14680629.2003.9689956 Masad, E., Muhunthan, B., Shashidhar, N. and Harman, T. Internal structure characterization of asphalt concrete using image analysis.

Journal of Computing in Civil Engineering, 13 (2), pp. 88-95, 1999. http://dx.doi.org/10.1061/(ASCE)0887-3801(1999)13:2(88) Chehab, G. R., O'Quinn, E. and Kim, R. Y. Specimen geometry study for direct tension test based on mechanical tests and air void variation in asphalt concrete specimens compacted by superpave gyratory compactor. Transportation Research Record, (1723), pp.

125-132, 2000.

Romero, P. and Masad, E. Relationship between the representative volume element and mechanical properties of asphalt concrete.

Journal of Materials in Civil Engineering, 13 (1), pp. 77-84, 2001. http://dx.doi.org/10.1061/(ASCE)0899-1561(2001)13:1(77)

Alvarez-Lugo & Carvajal-Muñoz / DYNA 81 (188), pp. 52-59. December, 2014.

[23] Hashin, Z. Analysis of composite materials-a survey. Journal of Applied Mechanics, 50 (3), pp., 1983.

[24] Consuegra, A., Little, D. N., Von Quintus, H. and Burati, J. J. Comparative evaluation of laboratory compaction devices based on their ability to produce mixtures with engineering properties similar to those produced in the field. Transportation Research Record, 1228, pp. 80-87, 1989.

[25] Button, J. W., Little, D. N., Jagadam, V. and Pendleton, O. J. Correlation of selected laboratory compaction methods with field compaction. Transportation Research Record, (1454), pp. 193-201, 1994.

[26] D'Angelo, J., Paugh, C. and Hannan, T. Comparison of the superpave gyratory compactor to the marshall for field quality control. Journal of the Association of Asphalt Paving Technologists, (64), pp., 1995.

[27] Carvajal, J. S., Reyes, O. J., Alvarez, A. E. and Fuentes, L. G., Evaluation of the internal structure of warm mix asphalt (wma) using x-ray computed tomography images, Proceedings of International Road Federation 17th International Meeting, pp., 2013.

[28] Peterson, R., Mahboub, K., Anderson, M., Masad, E. and Tashman, L. Superpave laboratory compaction versus field compaction.

Transportation Research Record, (1832), pp. 201-208, 2003.

[29] Masad, E., Birgisson, B., Al-Omari, A. and Cooley, A. Analytical derivation of permeability and numerical simulation of fluid flow in hot-mix asphalt. Journal of Materials in Civil Engineering, 16 (5), pp. 487-496, 2004. http://dx.doi.org/10.1061/(ASCE)08991561(2004)16:5(487) [30] Estakhri, C., Button, J. and Alvarez, A. E. Field and laboratory investigation of warm mix asphalt in texas, Report No FHWA/TX-10/05597-2, Texas Transportation Institute-Texas A&M University, 2010.

[31] Masad, E., Kassem, E. and Chowdhury, A. Application of imaging technology to improve the laboratory and field compaction of hma, Report No FHWA/TX-09/0-5261-1, Texas Transportation InstituteTexas A&M University, 2009, P. [32] Wegan, V. and Nielsen, C. B. Microstructure of polymer modified binders in bituminous mixtures, Report 109, Danish Road Institute (DRI), 2001, P. [33] Epps, J., Gallaway, B., Harper, W. and Scott, W. Compaction of asphalt concrete pavements, Research Report 90-2F, The Texas Highway Department and US Department of Transportation, 1969.

[34] Alvarez, A. E., Epps Martin, A. and Estakhri, C. Effects of densification on permeable friction course mixtures. Journal of Testing and Evaluation, 37 (1), pp. 11-20, 2009.

[35] Harvey, J. and Monismith, C. L. Effects of laboratory asphalt concrete specimen preparation variables on fatigue and permanent deformation test results using strategic highway research program a003a proposed testing equipment. Transportation Research Record, 1417, pp. 38-57, 1993.

[36] Jonsson, M., Partl, M. N. and Flisch, A. Comparison of different compaction methods using x-ray tomography, EMPA-No. FE840544, Road Engineering/Sealing Components Highway Engineering; Swedish Royal Institute of Technology, 2002, P. [37] Verdi, A., Shah, S. and Arena, P. Compaction of asphaltic concrete pavement with high intensity pneumatic roller. Part i, Research Report No. 10, Louisiana Department of Highways in Cooperation with the Bureau of Public Roads, 1963, P. [38] Brown, E. R. and Mallick, R. B. Evaluation of stone-on-stone contact in stone-matrix asphalt. Transportation Research Record, (1492), pp. 208-219, 1995.

[39] Kandhal, P. S. Design, construction, and maintenance of opengraded asphalt friction courses, Information series 115, National Asphalt Pavement Association, 2002, P. [40] Watson, D. E., Masad, E., Moore, K. A., Williams, K. and Cooley Jr., L. A. Verification of voids in coarse aggregate testing, determining stone-on-stone contact of hot-mix asphalt mixtures.

Transportation Research Record, 1891, pp. 182-190, 2004. http://dx.doi.org/10.3141/1891-21 [41] Alvarez, A. E., Mora, J. C. and Caballero, M. M., Análisis del contacto agregado-agregado en mezclas drenantes empleando tomografía computarizada con rayos-x y análisis de imágenes, Proceedings of XIX Simposio Colombiano Sobre Ingeniería de Pavimentos, pp., 2013.

[42] Hunter, A. E., Airey, G. D. and Collop, A. C. Aggregate orientation and segregation in laboratory-compacted asphalt samples.

Transportation Research Record, 1891, pp. 8-15, 2004. http://dx.doi.org/10.3141/1891-02 [43] Yue, Z. Q. and Morin, I. Digital image processing for aggregate orientation in asphalt concrete mixtures. Canadian Journal of Civil Engineering, 23, pp. 480-489, 1996. http://dx.doi.org/10.1139/l96052 [44] Saadeh, S., Tashman, L., Masad, E. and Mogawer, W. Spatial and directional distributions of aggregates in asphalt mixes. Journal of Testing and Evaluation, 30 (6), pp. 483-491, 2002.

[45] Wang, L. B., Paul, H. S., Harman, T. and Angelo, J. D. Characterization of aggregates and asphalt concrete using x-ray computerized tomography: A state-of-the-art report. Journal of the Association of Asphalt Paving Technologists, 73, pp. 467-500, 2004.

[46] Lu, Q. and Harvey, J. T., Laboratory evaluation of open-graded asphalt mixes with small aggregates and various binders and additives, Proceedings of Transportation Research Board 90th Annual Meeting pp. 1-18, 2011.

[47] Caro, S. A coupled micromechanical model of moisture-induced damage in asphalt mixtures: Formulation and applications, Ph.D. Thesis, Texas A&M University, College Station, TX., 2009.

[48] Al-Omari, A., Tashman, L., Masad, E., Cooley, A. and Harman, T. Proposed methodology for predicting hma permeability. Journal of the Association of Asphalt Paving Technologists, 71, pp. 30-58, 2002.

[49] Birgisson, B., Roque, R. and Page, G. Evaluation of water damage using hot mix asphalt fracture mechanics. Journal of the Association of the Asphalt Paving Technologists, 72, pp. 424-462, 2003.

[50] Birgisson, B., Roque, R. and Page, G. The use of a performancebased fracture criterion for the evaluation of moisture susceptibility in hot mix asphalt. Transportation Research Record, 3431, pp. 5561, 2004. http://dx.doi.org/10.3141/1891-08 [51] Liu, Y., You, Z., Li, L. and Wang, W. Review on advances in modeling and simulation of stone-based paving materials.

Construction and Building Materials, (43), pp. 408-417, 2013. http://dx.doi.org/10.1016/j.conbuildmat.2013.02.043 [52] Alvarez, A. E., Epps Martin, A. and Estakhri, C. Drainability of permeable friction course mixtures. Journal of Materials in Civil Engineering, ASCE, 22 (6), pp.

556-564, 2010. http://dx.doi.org/10.1061/(ASCE)MT.1943-5533.0000053 A.E. Alvarez-Lugo, received a Bs. Eng. in Civil Engineering in 1998 from the Universidad Nacional de Colombia. He got his MSc. degree in Civil Engineering in 2001 from the Universidad de los Andes, and his Ph.D. (Materials-Pavements) in 2009 from Texas A&M University. From 1998 to 2000 and 2001 to 2002, he worked in consulting engineering in the pavements engineering sector, and since 2002 he has been working as a professor in civil engineering and researcher. Currently, he is an Associate Professor and Head of the Civil Engineering Department at the University of Magdalena, Santa Marta, Colombia. His research interests include: bituminous materials, nanotechnologies, development of new materials for the pavement industry, non-destructive analysis of pavements, and design and implementation of permeable friction courses (PFC) mixtures.

ORCID: 0000-0001-9010-7642 J.S. Carvajal-Muñoz, received a Bs. Eng. in Environmental and Sanitary Engineering in 2012 and a Bs. Eng. in Civil Engineering in 2013 from the University of Magdalena. He has research experience since 2008 working as research assistant for the GIRPSU- and GIIC-research groups at the same institution. In 2013 he worked as Occasional Professor for the Civil Engineering program at the University of Magdalena. He is currently a M.Sc. student (Materials) at Texas A&M University. His research interests include both environmental sciences and civil engineering. In the first he is interested in biological fertilization of soils, recycling of materials, photocatalysis and advanced treatment of waste waters. In the civil area he is interested in Materials science and characterization, use of recycled materials in pavements and soils, bituminous materials, and nondestructive characterization of asphalt mixes.

ORCID: 0000-0003-3977-5076

59

Rail vehicle passing through a turnout: Influence of the track elasticity Rodrigo F. Lagos-Cereceda a, Kenny L. Alvarez-C. a, Jordi Vinolas-Prat b & Asier Alonso-Pazos c a

Escuela de Ingeniería Mecánica, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile. rodrigo.lagos@ucv.cl, kenny.alvarez.c@mail.ucv.cl a Escuela de Ingeniería Mecánica, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile. kenny.alvarez.c@mail.ucv.cl b Escuela Politécnica Superior y Escuela de Arquitectura, Universidad Antonio de Nebrija, Madrid, España. jvinolas@nebrija.es c Centro de Estudiose Investigaciones Técnicas - CEIT, San Sebastián, España. aalonso@ceit.es Received: September 27th, 2013. Received in revised form: June 24th, 2014. Accepted: September 16th 2014

Abstract In recent years, the different transport systems have been largely improved. However, some problems need to be still solved if railways want to be the transport of the future. Among others: the high cost of infrastructure and its maintenance, the requirements to assure low noise levels and low transmission of vibration.

The turnouts, one of the most critical elements on the track, are studied in this paper. Mathematical models have been developed in order to analyse in detail a vehicle passing through a turnout. These models are used to implement multibody models, which reproduce the phenomena of the vehicle/turnout interaction.

The research analyses the validity and accuracy of multibody tools and how they could allow improving the design process of turnouts.

In particular the influence of the turnout geometry and track elasticity has been investigated and also how these parameters could be optimized.

Keywords: Turnouts; switches and crossings; track geometry; track elasticity; dynamic vehicle–turnout interaction; railway; PAD; multibody models.

Influencia de la elasticidad de vía al circular por un desvío ferroviario Resumen En la actualidad, el transporte ferroviario en sus diversas formas se encuentra en pleno auge. Sin embargo, algunos problemas aún deben ser resueltos sí se desea que sea el transporte del futuro. Algunos de estos problemas son: el alto costo de la infraestructura y su mantenimiento, los requisitos para asegurar bajos niveles de ruido y baja transmisión de las vibraciones en las zonas.

Los desvíos ferroviarios es uno de los elementos más críticos en la vía, y se estudian en este trabajo. Modelos matemáticos se han desarrollado con el fin de analizar en detalle un vehículo que por un desvío. Estos modelos se utilizan para implementar modelos multicuerpo, que reproducen los fenómenos de la interacción vehículo/desvío.

La investigación analiza la validez y exactitud de herramientas multicuerpo y cómo podrían permitir mejorar el proceso de diseño de los desvíos.

En particular, la influencia de la geometría de la participación y la elasticidad de la vía ha sido investigada y también cómo estos parámetros podrían ser optimizados.

Palabras clave: desvíos ferroviarios; elasticidad de vía; dinámica ferroviaria; PAD; geometría de vía; interacción vehículo-vía; modelos multicuerpo; ferrocarril.

1. Introducción Actualmente el trasporte ferroviario está en constante crecimiento debido a sus cualidades únicas como por ejemplo: economía, rapidez, bajas emisiones contaminantes, etc. Un ejemplo de estas ventajas se puede observar en el trayecto Barcelona – Madrid, este corredor pasó de tener una cuota de mercado del 13,7% el año 2007 a un 45,6% el

año 2010, dos años después de la inauguración del corredor de Alta Velocidad (Fuente: Renfe y Comisión Europea).

Otros ejemplos se pueden ver en [1-3].

Es por ello que en las últimas décadas se ha potenciado exponencialmente la investigación con el objetivo de mejorar las condiciones de funcionamiento y así lograr un mejor servicio a la sociedad.

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 60-66. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.40047

Lagos-Cereceda et al / DYNA 81 (188), pp. 60-66. December, 2014.

recta, y la aguja recta separada, el vehículo circulará por vía desviada. Es por esto que la zona de cambio es la que define si el tren circula por vía desviada, o vía directa (véase Fig. 2).

Por otra parte, la zona de carriles de unión tiene por objetivo permitir la continuidad entre la zona del cambio y de cruzamiento. En general, el perfil de vía en esta zona es el mismo que el carril convencional.

La zona de cruzamiento es considerada una de las zonas más críticas, debido a su compleja forma geométrica y los fenómenos que aquí se producen. Las principales partes que componen esta zona son: corazón, pata de liebre y contracarril (véase Fig.3). Se pueden distinguir distintos tipos de cruzamiento, pueden ser con corazón fijo, móvil, o con pata de liebre móvil.

Variados investigadores han demostrado extensamente la importancia que presenta la elasticidad de la vía durante la circulación de un vehículo [3,5-7]. No obstante, la relevancia de este parámetro en los desvíos aún no ha sido estudiado en profundidad.

Figura 1: Partes de un desvío típico Fuente: Elaboración propia.

Figura 2: Partes de un desvío típico, zona del cambio Fuente: Elaboración propia.

2. Elasticidad de Vía Uno de los elementos propios que componen un sistema ferroviario son los carriles o vías, que permiten la circulación del vehículo. Dentro de las vías, diversos estudios, han identificado los desvíos como un elemento crítico durante la circulación Los desvíos son dispositivos que se encuentran directamente en la vía y permiten que el vehículo circule de una línea a otra, ya sea convergente o divergente. Estos aparatos son responsables directos de diversos fenómenos dinámicos como por ejemplo: contaminación acústica en las zonas colindantes al desvío, vibraciones que se trasmiten tanto al vehículo como a la infraestructura y restricciones en las velocidades de circulación. En general, estos fenómenos afectan principalmente tanto al confort, como a la vida útil del desvío y del vehículo. Diversos estudios han demostrado lo crítico de estos aparatos, por ejemplo García Díaz de Villegas en su trabajo [4] demostró la importancia que presentan el guiado de las agujas en el comportamiento de los ejes y el desgaste producido en las agujas producto de un guiado deficiente. Por otra parte, Kassa y Andersson fueron de los primeros investigadores desarrollar modelos que intentaban predecir la problemática de los desvíos, no obstante sus estudios evidencian la necesidad desarrollar estudios de mayor complejidad que considerasen frecuencias más amplias.

Los desvíos están compuestos principalmente por tres zonas: cambio, carriles de unión y cruzamiento (véase Fig. 1).

La zona de cambio se compone de dos conjuntos principales: agujas y contra-agujas. Las agujas son las partes móviles que inducen la circulación de una vía a otra, siendo una recta y la otra curva. Cuando la aguja recta se encuentra en contacto con la contra-aguja curva, la aguja curva se encuentra separada de la contra-aguja recta, permitiendo que el tren circule por vía directa. En cambio, cuando la aguja curva se encuentra en contacto con la contra-aguja

La elasticidad global de vía está compuesta, en general, por la elasticidad del conjunto de elementos que componen la vía [8,9]. Por lo tanto, la elasticidad de vía depende del tipo de carril, soportes elásticos (PADS), balasto/placa, subbalasto, traviesas, etc. (véase Fig. 4).

Figura 3: Partes de un desvío típico, zona del cruzamiento Fuente: Elaboración propia.

Figura 4. Esquema de vía sobre balasto Fuente: Elaboración propia.

61

Lagos-Cereceda et al / DYNA 81 (188), pp. 60-66. December, 2014.

Cuando un vehículo circula por vía convencional, la elasticidad está definida por los elementos mencionados anteriormente, pero en un desvío, la elasticidad se ve afectada por los cambios geométricos que se producen producto de las agujas, el corazón y demás elementos que componen un desvío.

Para lograr un adecuado análisis de la influencia de la elasticidad, esta investigación se dividió en dos partes: • En la primera se buscó determinar el efecto que presenta el cambio de la elasticidad en los desvíos y para ello se desarrollaron dos modelos de desvío en un software multicuerpo (Simpack), desvío A y B, simulando la circulación de un vehículo tipo metro. En cada modelo de desvío se realizaron variaciones de la elasticidad nominal de vía y con ello se obtuvieron las fuerzas verticales máximas en cada zona de interés, producidas por la circulación del vehículo. Es importante destacar que el software multicuerpo no considera (por defecto) las variaciones geométricas de la vía, por lo que la elasticidad de la vía no cambia, cuando la geometría varía.

• En la segunda parte del estudio, una vez analizada la influencia del valor de la elasticidad, se procedió a modelar por el método de elementos finitos el desvío A, para así determinar la elasticidad teórica en cada punto de la vía, considerando así los cambios geométricos producto del desvío. Una vez conocida la variación de la elasticidad a lo largo de la vía, esta se utilizó para complementar el modelo multicuerpo de Simpack permitiendo así considerar los cambios de elasticidad producto de los cambios de geometría inducidos por la aparición del desvío. Con dichas consideraciones en el modelo se volvió a analizar la influencia de la misma en la circulación.

Para facilitar el estudio se analizaron dos tipos de desvíos denominados A y B. El desvío A ha sido modelado como un desvío optimizado con el sistema FAKOP y corazón de punta móvil, por otra parte el desvío B es un desvío convencional con corazón de punta fija. Ambos desvíos presentan un radio de 500 metros, con vía desviada hacia la derecha de la vía principal.

El sistema FAKOP consiste en desplazar lateralmente la contra-aguja recta hacia el exterior de la vía [4], dejando una concavidad hacia el interior de la vía tal y como se indica en la Fig. 5. Esta geometría permite una variación en el radio de rodadura, lo que provoca que el eje se desplace lateralmente hacia la concavidad, compensando así el cambio del radio, lo cual permite evitar el ataque de la rueda hacia la aguja recta. Además, la zona de cruzamiento está compuesta por un corazón móvil y pata de liebre fija.

Figura 5. Sistema FAKOP Fuente: Elaboración propia.

Figura 6. Comparación fuerzas verticales en desvío por vía directa Fuente: Elaboración propia.

modelos en total). Luego, en cada modelo se definió un valor nominal de rigidez y amortiguamiento de vía mediante muelles y amortiguadores respectivamente. Y a partir de estos modelos “base”, se obtuvieron nuevos modelos a partir de la variación porcentual del valor nominal de amortiguamiento y rigidez de vía, con el objetivo de determinar así la influencia de la elasticidad en cada desvío (A y B). Las variaciones realizadas fueron de: -50%, -25%, 25% y 50% respecto del valor nominal de elasticidad.

Para poder obtener la influencia de la rigidez vertical, se calcularon las fuerzas que se producen en el punto de contacto rueda-carril en cada modelo desarrollado. Así se obtuvieron los picos de fuerzas en las zonas de cambio y cruzamiento. Para cada desvío, estos cálculos fueron graficados y comparados con los valores de una vía rígida, tanto como por circulación directa como por desviada para las zonas de cambio y de cruzamiento (véase Fig. 7 y 8).

En el caso de circulación por vía directa (véase Fig. 7), se distingue claramente que para ambos desvíos, A y B, la fuerza vertical se incrementa al aumentar la rigidez de la vía. Si se compara los resultados obtenidos de los dos modelos, se distingue que las magnitudes de las fuerzas verticales son mayores en el desvío B que en el A, y a su vez el modelo B presenta una gran sensibilidad a los cambios de rigidez.

El análisis de los resultados obtenidos al circular por vía desviada (véase Fig. 8), son semejantes a los anteriores, ya que se observa que la fuerza vertical aumenta cuando la rigidez es mayor. En ambos casos la fuerza vertical posee grandes variaciones en el desvío B, producto de la variación

3. Modelos multicuerpo con elasticidad de vía constante Para simular la circulación de un vehículo por cada desvío se desarrollaron los respectivos modelos de vía/desvío y vehículo en el software multicuerpo Simpack [10-12] (véase Fig. 6), definiendo la rigidez y la elasticidad de vía mediante muelles y amortiguadores respectivamente.

Primero se modelaron los desvíos A y B en Simpack, tanto para la circulación por vía directa como desviada (4 62

Lagos-Cereceda et al / DYNA 81 (188), pp. 60-66. December, 2014.

en la rigidez vertical, principalmente en la zona de cambio.

Es importante destacar que las condiciones de carga para un desvío son recomendadas principalmente por cada fabricante de desvíos, y por los administradores de la vía, por lo que las condiciones de carga modeladas en el presente trabajo no se ajustan necesariamente a los requerimientos de cada administración o fabricante.

Como conclusión de esta parte del estudio, se puede observar que la magnitud de la fuerza vertical en la vía aumenta si la rigidez de la vía es mayor, además se puede observar que si se implementa el sistema FAKOP, es decir, sí se realizan cambios geométricos en la zona de cambio y se implementa un sistema de corazón móvil, las fuerzas disminuyen considerablemente.

Figura 8. Comparación fuerzas verticales en desvío por vía desviada Fuente: Elaboración propia.

Con este objetivo se elaboró un modelo 3D para el carril derecho e izquierdo, y se cuantificó la rigidez vertical de vía por medio del método de elementos finitos [17], para luego utilizar dicha elasticidad en los modelos dinámicos en Simpack.

En estos modelos de elementos finitos se consideró la forma geométrica correspondiente a cada carril en el desvío, considerando la elasticidad de los soportes elásticos (PADS) y la distancia entre traviesas. Por otra parte, para simplificar el problema se consideró que las traviesas y placas poseían una rigidez infinita.

5. Cálculo de la elasticidad de vía Figura 7. Comparación fuerzas verticales en desvío por vía directa Fuente: Elaboración propia.

Una vez separada la vía en derecha e izquierda, cada carril fue dividido en secciones según la cantidad de zonas de interés. De esta forma el carril izquierdo se modeló en tres zonas, la zona de cambio, de carriles de unión y de cruzamiento. En cambio el carril derecho que requería una mayor partición debido a la complejidad que éste presenta, se dividió en seis zonas, dos divisiones para la zona de cambio, una para la zona de carriles de unión y tres para la zona de cruzamiento.

Una vez desarrollados los modelos, se asignaron las siguientes propiedades del material (acero): E=210.000 N/mm2, υ=0,3 y ρ=7.850 kg/m3.

Para poder completar el modelo, fue necesario asignar la rigidez de vía, esta rigidez se definió por los soportes elásticos dispuestos en el patín del carril, es decir en la parte inferior del carril, espaciados según la distribución de las traviesas (proporcionada por el fabricante del desvío). La

4. Modelos multicuerpo con elasticidad de vía variable Se ha demostrado al estudiar el comportamiento dinámico de un sistema convencional vehículo-vía que los cambios de rigidez son de segundo orden, y por lo tanto su influencia es menor en comparación de otras variables. Sin embargo, al estudiar la circulación por un desvío, las condiciones de elasticidad se ven alteradas drásticamente producto de los abruptos cambios geométricos existentes en un desvío [13].

Para poder estudiar el comportamiento de vía bajo condiciones de elasticidad variable, cosa que por defecto no permite Simpack, se desarrolló un modelo en ABAQUS [14–16] del desvío tipo A, para circulación por vía directa.

63

Lagos-Cereceda et al / DYNA 81 (188), pp. 60-66. December, 2014.

Figura 9. Parte del modelo en Abaqus de la zona del corazón de punta móvil Fuente: Elaboración propia.

Figura 10. Rigidez vertical calculada en ABAQUS, carril derecho Fuente: Elaboración propia.

rigidez de estos soportes está dada por cuatro casos, tres casos con rigidez constante (200, 100 y 20 kN/mm) y uno con rigidez variable que cambia en función de la posición en la vía. Los valores utilizados en el caso de rigidez variables fueron considerados en función de los estudios iniciales y definidos arbitrariamente por el investigador, considerando que ocurren cambios geométricos en las vías.

Lo anterior fue desarrollado para ambos carriles, derecho e izquierdo, obteniendo así ocho modelos en total.

El siguiente paso consistió en confeccionar el mallado de los carriles, siendo uno de los procedimientos más laboriosos del modelo, debido que para obtener resultados más precisos, fue necesario confeccionar un mallado muy fino, aumentando considerablemente los requerimientos computacionales. Es por este motivo que se decidió dividir los carriles aislando aquellas zonas de geometría más compleja (véase Fig. 9), de esta forma se logró definir un mallado fino en las zonas complejas, como por ejemplo en el corazón, y en las aquellas zonas de menor complejidad geométrica se optó por un mallado más basto. Con este procedimiento cada modelo no sobrepasó los 550.000 elementos.

Una vez definidos los modelos, se procedió a definir todas aquellas posiciones a lo largo del carril en los que se deseaba conocer la rigidez de la vía. Luego, para cada una de estas posiciones se procedió a calcular la rigidez vertical de la vía, para luego post-procesar los resultados e implementarlos en los modelos dinámicos Simpack.

El procedimiento para calcular la rigidez vertical del carril en cada posición consistió en aplicar una carga vertical estática sobre la cabeza del carril, y luego calcular la deformación de la vía en ese punto. Con la deformación obtenida producto de la aplicación de la carga se obtiene la rigidez vertical de vía. Este procedimiento se repite en todas aquellas posiciones en los que se deseaba conocer la elasticidad.

Para poder realizar los cálculos de las deformaciones verticales de vía, se creó una batería de cálculo en MATLAB, debido a que se necesitaban tantos cálculos como posiciones de aplicación de la carga. Es así como se creó un fichero con la información de cada punto de aplicación de la carga, para luego calcular la deformación del carril en cada punto de aplicación, de esta forma se pudo calcular la rigidez vertical de vía para ambos carriles en función de la posición.

Figura 11. Rigidez vertical calculada en ABAQUS, carril izquierdo Fuente: Elaboración propia.

De acuerdo a los resultados obtenidos, mostrados en las Fig. 10 y 11, se puede observar que en la zona de cambio y de cruzamiento, la rigidez vertical varía en función de la posición para cierta carga aplicada. Además, la magnitud de la rigidez de vía depende también de la rigidez de los soportes, si la rigidez de los soportes es mayor, la rigidez de vía también lo será, y por último si se compara los dos carriles para una misma rigidez, se observa claramente que las variaciones son mayores en el carril derecho que en el izquierdo producto que en el primero se observan los mayores cambios geométricos.

6. Introducción de la elasticidad en los modelos multicuerpo de Simpack Una vez obtenida la rigidez en Abaqus, se incorporó dicha rigidez en los modelos de Simpack. Para considerar la elasticidad calculada previamente en Simpack se desarrollaron ocho ficheros de texto, cada uno con los valores de rigidez en cada posición del carril obtenidos en Abaqus (cuatro ficheros para el carril derecho y cuatro para 64

Lagos-Cereceda et al / DYNA 81 (188), pp. 60-66. December, 2014.

Por último, en la segunda zona de cruzamiento también se tienen variaciones de las fuerzas verticales (Fig. 14), se aprecia que las variaciones de las fuerza son semejantes entre una magnitud de rigidez y otra, por lo tanto la influencia de la rigidez en esta zona pasa a ser despreciable, y solamente se puede considerar que la variación de las fuerzas se debe en su totalidad al cambio geométrico de los carriles. Por lo tanto la geometría del carril en este punto es la que determina el comportamientos de la fuerzas.

Figura 12. Fuerza vertical, zona de las agujas Fuente: Elaboración propia.

el izquierdo). Una vez listos los ficheros, se crearon y definieron las respectivas funciones en Simpack dependientes de estos ficheros. De esta manera en cada modelo la rigidez vertical quedó definida por una determinada función exclusiva para cada caso de estudio.

La rigidez vertical de vía que se implementó en Simpack corresponde a las cuatro rigideces calculadas en Abaqus para los distintos tipos de rigidez de definidos en los soportes, y una quinta con la rigidez nominal constante proporcionada por Simpack.

Al analizar los resultados, se distinguen tres zonas irregulares a lo largo de los carriles, la primera corresponde a la zona de cambio y las otras dos a la zona de cruzamiento. En la zona de cambio representada en la Fig. 12, se observa que la fuerza varía considerablemente. Esta variación se debe principalmente a los cambios geométricos que posee la vía en esa zona, debido a que el vehículo se ve afectado repentinamente por la incorporación de las agujas.

Además se observa que la magnitud de la fuerza disminuye cuando la rigidez es menor.

Por otra parte, en la primera zona de cruzamiento (Fig. 13), la dinámica del vehículo se ve afectada exclusivamente por la elasticidad de vía y no por sus cambios geométricos producto de la aparición del corazón como era de esperar.

De esta figura se concluye que la las fuerzas en el contacto al internarse en la zona del cruzamiento se ven influenciadas por la rigidez equivalente de la vía.

Figura 14. Fuerza vertical, zona del corazón, influencia geométrica Fuente:Elaboración propia.

7. Comentarios finales y conclusiones Al realizar un estudio del comportamiento dinámico del sistema, principalmente de la influencia de la elasticidad de vía en los desvíos ferroviarios, se pudo demostrar que el sistema FAKOP logra minimizar los efectos dinámicos ocurridos en las zonas de cambio y cruzamiento. Los efectos dinámicos se ven reflejados principalmente a través de las fuerzas que se inducen en la vía. En concreto, si se considera una vía rígida, o una rigidez constante, las fuerzas inducidas en la vía son mayores, y a la vez estas aumentan al aumentar la rigidez vertical.

Por otra parte, se desarrollaron modelos para calcular la rigidez variable de la vía por medio del método de elementos finitos, lográndose así obtener la rigidez de la vía para distintos valores de rigidez vertical, en función de la posición del carril. Esto se desarrolló con el fin de implementar estos resultados en un modelo de simulación para luego poder evaluar la influencia de ésta rigidez en la vía. Como resultado se obtuvo que al aumentar la rigidez de vía, considerando los cambios de la elasticidad producidos por el desvío, las fuerzas verticales aumentan, principalmente en la zona de cambio, y en la primera zona de cruzamiento. Otro fenómeno importante que ocurre en la zona de cruzamiento, es que los cambios abruptos de rigidez provocan aceleraciones verticales en el vehículo, lo que se traduce en un aumento en las fuerzas verticales.

Referencias [1]

Figura 13. Fuerza vertical, zona de la punta del corazón, influencia de la elasticidad Fuente: Elaboración propia.

65

Button, K., Transportation economics: Some developments over the past 30 years, Transportation Research Forum, 45 (2), pp. 7-30, 2006.

Lagos-Cereceda et al / DYNA 81 (188), pp. 60-66. December, 2014.

[2] [3]

[4] [5]

[6]

[7] [8]

[9]

[10] [11] [12]

[13] [14] [15]

[16]

[17]

Givoni, M. and Banister, D., Role of the railways in the future of air transport, Transportation Planning and Technology, 30 (1), pp. 95112, 2007. http://dx.doi.org/10.1080/03081060701208100 Lagos, R.F., Alonso, A., Vinolas, J., and Perez, X., Rail vehicle passing through a turnout: Analysis of different turnout designs and wheel profiles, Proceedings of the Institution of Mechanical Engineers, Journal of Rail and Rapid Transit, 226 (F6), pp. 587-602 2012. http://dx.doi.org/10.1177/0954409712445114 García-Díaz De Villegas, J.M. and Bugarín, M.R., Improvements in railway switches, J. Rail Rapid Transit, 216 (4), pp. 275-286, 2002. http://dx.doi.org/10.1243/095440902321029226 Kassa, E. and Nielsen, J.C.O., Dynamic interaction between train and railway turnout: Full-scale field test and validation of simulation models, Vehicle System Dynamics: International Journal of Vehicle Mechanics and Mobility, 46 (Suppl.1), pp. 521-534, 2008.

Johansson, A., Pålsson, B., Ekh, M., Nielsen, J.C.O., Ander, M.K.A., Brouzoulis, J. and Kassa, E., Simulation of wheel–rail contact and damage in switches & crossings, Wear, 271 (1-2), pp.

472-481, 2011. http://dx.doi.org/10.1016/j.wear.2010.10.014 Gurule, S. and Wilson, N., Simulation of wheel/rail interaction in turnouts and special track work, Vehicle System Dynamics, 33 (Suppl.), pp. 143-154, 2000.

Lei, X.-Y. and Zhang, B., Analysis on dynamic behavior of ballastless track based on vehicle and track elements with finite element method, Tiedao Xuebao, Journal China Railway Society, 33 (7), pp. 78-85, 2011.

Remennikov, A.M. and Kaewunruen, S., Experimental investigation on dynamic railway sleeper/ballast interaction, Experimental Mechanics, 46 (1), pp.

57-66, 2006. http://dx.doi.org/10.1007/s11340-006-5868-z Schupp, G., Netter, H., Mauer, L. and Gretzschel, M., Multibody system simulation of railway vehicles with SIMPACK, Vehicle System Dynamics, 31 (Suppl.), pp. 101-118, 1999.

Iwnicki, S., The Manchester benchmarks for rail vehicle simulation, Vehicle System Dynamics, 30 (S), pp. 295-313, 1998 Lagos, R.F., Alonso, A., Vinolas, J. y Sanchez, J.C., Simulación del comportamiento dinámico de vehículos ferroviarios a su paso por desvíos. Influencia de parámetros, XVIII Congreso Nacional de Ingeniería Mecánica, Asociación Española de Ingeniería Mecánica, Ciudad Real, 10 P., 2010.

Iwnicki, S., Handbook of railway vehicle dynamics, CRC/Taylor & Francis, 2006.

Galeano, C., Mantilla, J., Duque, C. and Mejía, M., Herramientas de software con licencia pública general para el modelado por elementos finitos, DYNA, 74 (153), pp. 313-324, 2007.

Wiest, M., Kassa, E., Daves, W., Nielsen, J.C.O. and Ossberger, H., Assessment of methods for calculating contact pressure in wheelrail/switch contact, Wear, 265 (9-10), pp. 1439-1445, 2008. http://dx.doi.org/10.1016/j.wear.2008.02.039 Wiest, M., Daves, W., Fischer, F.D. and Ossberger, H., Deformation and damage of a crossing nose due to wheel passages, Wear, 265 (910), pp.

1431-1438, 2008. http://dx.doi.org/10.1016/j.wear.2008.01.033 Menssen, R. and Kik, W., Running through a switch - simulation and test, Vehicle System Dynamics, 23 (Suppl), pp. 378–389, 1994. http://dx.doi.org/10.1080/00423119308969528

en la investigación Elasticidad de Vías Ferroviarias, modelando componentes estructurales en Abaqus. En la actualidad participa en el desarrollo de proyectos de viviendas energéticamente sustentables y estudios en dinámica ferroviaria. Además se encuentra preparando la defensa de su tesis de grado “Desarrollo y validación de una metodología para medición de ruido generado al paso de un vehículo ferroviario” para optar al Título de Ingeniero Mecánico, en la Pontificia Universidad Católica de Valparaíso, Chile.

A. Alonso-Pazos, se recibió como Ing. Industrial en 2002 y en 2005 se recibe como Dr. en Ingeniería Mecánica, por la Universidad de Navarra, España. Desde el año 2002 es miembro del departamento de Mecánica Aplicada del CEIT, desarrollando su trabajo principalmente en el grupo de Dinámica Ferroviaria. Actualmente es jefe del laboratorio de ferrocarriles CAF-CEIT. Sus investigaciones se centran en el contacto rueda-carril, estabilidad dinámica de vehículos ferroviarios, transmisión de vibraciones, ruido ferroviario y caracterización de propiedades dinámicas en materiales viscoelásticos. Ha participado en la supervisión de 5 tesis doctorales, es autor o co-autor de más de 16 publicaciones indexadas ISI y ha participado y/o dirigido más de 14 proyectos de investigación nacionales y europeos.

J. Vinolas-Prat, es Ing. Mecánico por la Universidad de Navarra, España, obtuvo el Dr. en esta misma institución en 1991. Sus intereses científicos se han centrado en los campos de la dinámica de máquinas, ruido y vibraciones, dinámica ferroviaria e interacción con la infraestructura.

Obtuvo la acreditación como profesor catedrático de la ANECA en 2010.

En su haber consta la publicación de alrededor de 50 artículos científicos en revistas indexadas (ISI) en áreas tales como análisis estructural, dinámica ferroviaria, suspensiones activas, trenes basculantes, reducción de ruido, interacción rueda/carril y otros aspectos relacionados con la optimización de los componentes de vehículos y máquinas. Es miembro del comité científico de la Conferencia Internacional sobre ruido y vibraciones ISMA, miembro del Consejo Editorial de la Revista IMech Journal of Rail and Rapid Transit y del Consejo Editorial de la Revista International Journal of Rail Transportation (IJRT).

Área Curricular de Ingeniería Civil Oferta de Posgrados   

R.F. Lagos-Cereceda, se recibió como Ing. Mecánico en 2004 en la Universidad de Santiago de Chile, Chile, en 2009 se recibe como Ing. Civil Industrial por la Universidad Técnica Federico Santa María, Chile. En 2012 obtiene el grado de Dr. en Ingeniería Industrial por la Universidad de Navarra, España. Desde 2003 a 2005 trabajó como ingeniero de proyectos, para luego desarrollarse como ingeniero calculista en el ámbito estructural y gerente de producción. A partir del año 2008 se especializa en el área ferroviaria. Y a partir del año 2012 forma parte del cuerpo docente de la Escuela de Ingeniería Mecánica de la Pontificia Universidad Católica de Valparaíso, Chile. Su investigación se centra principalmente en la interacción vehículo/desvío, desvíos ferroviarios, modelación multicuerpo, MEF.

 

Especialización en Vías y Transportes Especialización en Estructuras Maestría en Ingeniería - Infraestructura y Sistemas de Transporte Maestría en Ingeniería – Geotecnia Doctorado en Ingeniería - Ingeniería Civil Mayor información:

John Jairo Blandón Valencia Director de Área curricular asisacic_med@unal.edu.co (57-4) 425 5172

K.L. Alvarez-C., desde el año 2012 participa activamente en diversos proyectos de investigación de la Escuela de Ingeniería Mecánica de la Pontificia Universidad Católica de Valparaíso, Chile. En el 2013 participó 66

Evolution of the passive harmonic filters optimization problem in industrial power systems Jandecy Cabral-Leite a, Ignacio Pérez-Abril b, Manoel Socorro Santos-Azevedo c, Maria Emilia de Lima-Tostes d & Ubiratan Holanda-Bezerra e a Instituto de Tecnología y Educación Galileo de la Amazonía, Manaus, Brasil. jandecycabral@hotmail.com Centro de Estudios Electroenergéticos, Universidad Central de Las Villas, Santa Clara, Cuba. iperez@uclv.edu.cu c Universidad del Estados de Amazonas, Manaus, Brasil. manoelazevedo@yahoo.com.br d Departamento de Ingeniería Eléctrica y Computación, Universidade Federal do Pará, Belém, Brasil. tostes@ufpa.br e Departamento de Ingeniería Eléctrica y Computación, Universidade Federal do Pará, Belém, Brasil. bira@ufpa.br b

Received: September 30th, 2013. Received in revised form: June 12th, 2014. Accepted: October 30th, 2014.

Abstract Several authors have treated the optimization of passive filters in electric distribution systems. Optimization methods like: sequential quadratic programming (SQP), simulated annealing (SA), differential evolution (DE), artificial neural networks (ANN), particle swarm optimization (PSO), genetic algorithm (GA), etc., have been employed for optimizing certain configurations of passive filters. These optimization methods have been employed to solve several formulations of the problem of the project of filters. These formulations can be classified in: formulations of one or several objectives. The objective of the present work is to show the evolution of the formulation of this problem in the lasts years respect to the objective functions and constraints used. This analysis shows how the formulations employed have been upgraded from single-objective to multi-objective formulations to achieve a better representation of this complex problem.

Keywords: harmonics; passive filters; optimization.

Evolución del problema de optimización de los filtros pasivos de armónicos en sistemas eléctricos industriales Resumen Varios autores han tratado la optimización de filtros pasivos en sistemas eléctricos de distribución. Métodos de optimización como: programación cuadrática secuencial (SQP), simulación del recocido (SA), evolución diferencial (DE), redes neuronales (ANN), calentamiento de partículas (PSO), algoritmo genético (GA), etc., han sido empleados para la optimización de ciertas configuraciones de filtros pasivos. Estos métodos de optimización se han empleado para resolver múltiples formulaciones del problema de diseño de estos filtros. Estas formulaciones pueden clasificarse en: formulaciones de uno o de varios objetivos. El objetivo del presente trabajo es mostrar la evolución que ha tenido la formulación de este problema en los últimos años en cuanto a funciones objetivo y restricciones se refiere. Este análisis muestra como se ha transitado de formulaciones mono-objetivo a formulaciones multi-objetivo para lograr una mejor representación de este complejo problema.

Palabras clave: armónicos; filtros pasivos; optimización.

1. Introducción Dugan [1] señala que los filtros pasivos de armónicos tipo paralelo son dispositivos cuya función es “cortocircuitar” las corrientes de armónico tan cerca cómo se pueda de la fuente de distorsión, es decir, de las cargas no lineales.

En la práctica es imposible “cortocircuitar” la corriente de

armónico Ih de la carga no-lineal y lo que se produce es una división de esta corriente en: la corriente por el filtro Ifh y la corriente por el sistema Ish (Fig. 1).

Evidentemente, la acción del filtro será mejor en tanto reduzca más la corriente Ish por el sistema, lo cual redundará en una reducción de la tensión de armónico Vh en el punto de conexión del filtro.

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 67-74. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.40063

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

seguridad en caso de que cambios en los parámetros de los filtros eleven la frecuencia del pico de resonancia hasta la frecuencia h.

Para evitar problemas con este pico de resonancia, Dugan [1] recomienda añadir los filtros comenzando por el armónico significativo de más bajo orden del sistema. Es decir, para instalar un filtro de séptimo armónico, debe existir también un filtro de quinto armónico.

Varias contribuciones se han dirigido a determinar la ubicación más apropiada de los filtros en el sistema eléctrico [5-6]. En general emplean indicadores de sensibilidad de la distorsión armónica a la conexión de filtros para desarrollar una estrategia para la ubicación de los mismos.

Como se sabe, el filtro limita la circulación de corrientes armónicas solo desde su localización hasta la fuente, aunque siempre va a mejorar la calidad de la tensión en todos los nodos. Además, se destaca [1] la importancia de escoger localizaciones en que la impedancia del sistema de suministro sea bastante estable.

Varios autores han tratado la optimización de filtros pasivos en sistemas eléctricos de distribución. Métodos de optimización como: programación cuadrática secuencial (SQP), simulación del recocido (SA), evolución diferencial (DE), redes neuronales (ANN), calentamiento de partículas (PSO), algoritmos genéticos (GA), etc., han sido empleados para la optimización de ciertas configuraciones de filtros pasivos.

Estos métodos de optimización se han empleado para resolver múltiples tipos de representaciones del problema de diseño de filtros, las que pueden clasificarse en: formulaciones de uno o de varios objetivos.

El objetivo del presente trabajo es mostrar la evolución que ha tenido este problema en los últimos años en cuanto a funciones objetivo y restricciones se refiere.

Este análisis muestra como se ha transitado de formulaciones mono-objetivo a formulaciones multiobjetivo para lograr una mejor representación de este complejo problema.

Figura 1. Acción del filtro.

Fuente: J. Arrillaga and N.R. Watson

Una metodología para evaluar la eficiencia de los filtros ha sido propuesta por Czarnecki y Ginn [2,3]. En la misma se “mide” dicha eficiencia como la reducción de la distorsión armónica total THD de la corriente y la tensión en el punto donde se conecta el filtro.

La eficiencia del filtro respecto a su función de reducción de la distorsión de corriente (εi) se calcula como:

i 

THD( Is 0 )  THD( Is) THD( Is 0 )

Donde Is0 es la corriente por el sistema antes de conexión del filtro e Is es la corriente resultante después su conexión.

Por otra parte, la eficiencia del filtro respecto a función de reducción de la distorsión de tensión (εv) calcula como:

v 

THD(V 0 )  THD(V ) THD(V 0 )

(1) la de su se

(2)

Donde V0 es la tensión en el punto de conexión del filtro antes de su instalación y V es la tensión en dicho punto resultante por la instalación del filtro.

Como los filtros sintonizados solo “eliminan” una frecuencia determinada, normalmente pueden formarse arreglos de filtros sintonizados que pueden incluir incluso filtros amortiguados.

En este caso se presenta el problema de cómo distribuir la potencia reactiva total del filtro entre las diferentes ramas del mismo. La solución tradicional para este problema [4] ha sido emplear capacitores iguales en las diferentes ramas, no obstante, como apuntan Czarnecki y Ginn [2-3], la distribución de la potencia reactiva entre las ramas influye decisivamente en la eficiencia del filtro, ya que las magnitudes de los diferentes armónicos a filtrar son diferentes.

Una particularidad importante de los filtros sintonizados [1] es que los mismos crean un pico de resonancia paralelo con la impedancia equivalente del sistema a una frecuencia inferior a su frecuencia de sintonía.

Este pico de resonancia debe quedar lejos de cualquier armónico significativo, por lo que los filtros comúnmente se sintonizan a una frecuencia ligeramente inferior a la frecuencia h que se quiere “eliminar” como margen de

2. Formulaciones mono-objetivo Las formulaciones mono-objetivo fundamentan la determinación de los parámetros del filtro o conjunto de filtros en la optimización de una función objetivo dada, sujeta a un conjunto de restricciones. Algunas de estas formulaciones se discuten seguidamente: Haozhong Cheng et. al. [7] presentan en 1995 la determinación de un conjunto de filtros sintonizados bajo dos criterios diferentes: mínima potencia reactiva a la frecuencia fundamental Qs1 sujeto a un nivel máximo de distorsión total de la tensión THDV, o mínima distorsión total de la tensión sujeto a cumplir un nivel deseado de potencia reactiva a la frecuencia fundamental.

En 1998 Tien-Ting Chang y Hong-Chan Chang [8] formulan el problema de planificación de filtros pasivos sintonizados en sistemas eléctricos de distribución a partir de la minimización de una función de costo dada por f:

min f  C P  C F  68

(3)

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

Donde CP representa el costo anual de las pérdidas de energía considerando varios estados de carga del sistema y CF el costo de los filtros sintonizados a instalar.

Evidentemente, para reunir en esta función el costo anual de pérdidas con el costo de inversión de los filtros, este tiene que llevarse a unas bases anuales.

Como restricciones al problema, se considera limitar la tensión V en cada nodo a un intervalo permisible, así como la distorsión armónica total de tensión THDV en cada nodo por debajo de un valor máximo preestablecido.

Vi (min)  Vi  Vi (max)

(4)

THDVi  THDV(max)

(5)

Vi ,h / Vi ,1  IHDV (max)

En esta contribución se emplea el método de SQP con buenos resultados.

En 2005, Ahmed Faheem Zobaa [11] utiliza una función de costo semejante a la (6), pero trata la determinación de un solo filtro sintonizado para compensar una carga nolineal y al mismo tiempo garantizar un factor de potencia superior al 90% mediante la restricción:

PF1  0.9

Xeqh  0

(6)

(7)

QF1 (min)  QF1  QF1 (max)

(8)

Vc  1.1Vcnom

(13)

Vc pico  1.1 2 Vcnom

(14)

El método de optimización empleado en este caso es el de Búsqueda de la Sección Dorada.

Este mismo autor publica en 2006 [12] una nueva versión de su programa que completa ahora las restricciones necesarias para evitar la sobrecarga del capacitor por corriente Ic y por potencia reactiva Qc de acuerdo a los límites de la IEEE Std. 18-1992 [13].

En esta contribución no se consideran estados de carga u operación diferentes para el sistema eléctrico.

En 2003, I. Pérez y J. A. González [10] formulan el problema de maximizar el Valor Presente Neto VPN del proyecto de compensación que comprende el costo de las pérdidas de energía y el costo de inversión e instalación de los filtros.

Considerando un período de evaluación de N años con una razón de interés i, el VPN del proyecto de compensación de potencia reactiva se calcula como: N   0 max  f  VPN  C F   (C P  C P ) /(1  i ) k  k 1  

(12)

Además, este autor introduce restricciones adicionales para evitar la sobrecarga del capacitor del filtro por sobretensión en valor eficaz Vc y en valor pico Vcpico.

El costo de cada filtro se considera proporcional Kch, Klh a la potencia reactiva del capacitor Qch e inductor Qlh. Como restricciones al problema, se vuelven a incluir los límites (4) y (5) para la tensión y el THDV en cada nodo, no obstante, se añaden otras dos condiciones límites adicionales: la primera a la distorsión total de la corriente THDI en el punto de conexión común PCC, y la segunda a la potencia reactiva total de frecuencia fundamental de los filtros QF1.

THDI  THDI (max)

(11)

Un detalle interesante de este acercamiento es que emplea un conjunto de restricciones para evitar la resonancia de la impedancia equivalente del sistema Zeqh para cualquier frecuencia armónica presente en el espectro de la carga no-lineal. Debido a la sencillez del circuito, este tipo de restricción puede expresarse evitando la aparición de una reactancia equivalente Xeqh cero (condición de resonancia), es decir:

Se emplea el método DE para solucionar el problema que se plantea en variables discretas.

En 2001 Ying-Tung Hsiao [9] emplea el método de SA para optimizar la suma de los costos de inversión y el costo de instalación Ki de un conjunto de filtros sintonizados.

  min  f  Ki   Kch Qch  Klh Qlh  h  

(10)

Ic  1.8Icnom

(15)

Qc  1.35Qcnom

(16)

Su nueva función de costo a minimizar corresponde con el costo de pérdidas anuales más el costo de inversión del filtro actualizado a un año considerando N años de evaluación con una razón de interés i.

  i (1  i ) N   C F  min  f  C P   N  (1  i )  1   

(9)

(17)

En 2006, Gary Chang, et. al. [14] presentan una formulación que utiliza hasta cuatro funciones objetivo alternativas que se optimizan mediante un algoritmo genético.

Es decir, se escoge una de entre las siguientes cuatro funciones:

Como restricciones al problema, se utilizan los límites (4) y (5) para la tensión y el THDV en cada nodo y en este caso se añade una condición adicional a la distorsión individual de la tensión IHDV en cada nodo.

69

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

En 2008, Alex-Sander, et. al. [17] tratan el problema de diseñar un filtro de mínima potencia reactiva compuesto por varias ramas sintonizadas.

En 2009, Gary W. Chang et al. [18] determinan un conjunto de filtros sintonizados por la minimización del costo de los mismos sujeto a las restricciones de calidad de la energía y estrés de los filtros. La parte distintiva de este trabajo consiste en el análisis probabilístico que se emplea para considerar las variaciones en las fuentes de armónicos y en la impedancia del sistema.

Ying-Yi Hong y Ching-Sheng Chiu [19] presentan en 2010 la minimización del costo de un conjunto de filtros sintonizados sujeto a las restricciones de distorsión máxima, estrés de los filtros y factor de potencia deseado. Esta contribución utiliza el método de “Simultaneous Perturbation Stochastic Approximation”.

En 2010, Ahmed Faheem Zobaa et. al. [20] tratan el problema de determinar las variaciones de los parámetros de un filtro sintonizado para adaptarse a variaciones dinámicas de la carga. La optimización se realiza para uno de dos criterios posibles: mínimas pérdidas en el sistema o máximo factor de potencia.

L. I. Kovernikova1 y Nguyen Chi Thanh [21] presentan en 2012 el cálculo de los parámetros de diferentes tipos de filtros para sistemas de alta tensión que cumplen una potencia reactiva especificada a la frecuencia fundamental.

Como función objetivo a minimizar seleccionan las pérdidas en los filtros, donde IFh y RFh son la corriente y resistencia del filtro al armónico h.

1) La suma cuadrática de la distorsión armónica total en todos los m nodos THDVm. M  2 min  f   THDV m   m 1  

(18)

2) La suma cuadrática de los factores de influencia telefónica ITFm en los nodos.

M  2 min  f   ITFm   m 1  

(19)

3) La suma de las pérdidas de potencia en la red para todos los armónicos superiores HPLh. H   min  f   HPLh  h2  

(20)

1)

El costo de inversión en los filtros seleccionados mediante una función semejante a (6).

Esta contribución utiliza cuatro tipos de restricciones para el problema: los límites de THDV (5), de distorsión individual (10), de tensión límite en el capacitor de cada filtro (13) y una restricción adicional para evitar la desintonización de los filtros.

Como los valores los componentes de los filtros pueden variar debido a tolerancias de fabricación o condiciones ambientales (-7% a +12% para las capacitancias y ±3% para las inductancias), se plantea que la reactancia XL del filtro a frecuencia fundamental debe cumplir la desigualdad.

0.907

XC X  X L  1.156 2C h2 h

  min f  PF   IFh2 RFh  h  

(21)

Como restricciones se plantea un rango válido para las tensiones armónicas con respecto a la tensión nominal en cada nodo.

Donde XC es la reactancia del filtro a frecuencia fundamental y h su frecuencia de sintonía.

M. Ghiasi, V. Rashtchi y H. Hoseini [15] formulan el problema en 2005 como la minimización del costo total de los filtros sintonizados que se seleccionan y ubican en un circuito de distribución. En este sentido usa una función objetivo semejante a la (6).

Utiliza un Algoritmo Genético Simple, donde las restricciones de THDV máximo (7) e IHDV máximo (10) se incluyen mediante una función de penalidad.

En 2008, Zhang Ruihua, Liu Yuhong y Li Yaohua [16] tratan el proyecto de dos filtros sintonizados al quinto y séptimo armónicos y un filtro de segundo orden que generan una potencia reactiva especificada al sistema. Nuevamente la función objetivo es el costo de los filtros.

Como restricciones de utilizan las de máximo THDV (7), IHDV (10) y de máxima distorsión individual de la corriente IHDI.

I h / I1  IHDI(max)

(23)

Vh (min)/ Vnom  Vi ,h / Vnom  Vh (max)/ Vnom (24) La optimización se realiza con la técnica de PSO. En 2012, Shady et. al. [22] publican la optimización del un filtro tipo C para mínima distorsión de tensión THDV, sujeto a las restricciones de calidad, estrés del filtro y factor de potencia deseado. El objetivo es minimizar la distorsión de tensión en el nodo en que se conecta el filtro.

min f  THDV 

(25)

V. Ravikumar Pandi et. al. [23] presentan en 2012 la optimización de filtros sintonizados y de segundo orden bajo el concepto de minimizar el máximo THDV de todos los nodos del sistema.

min f  maxk THDVk 

(22)

(26)

Este trabajo emplea el método de PSO para resolver el problema.

En 2012, empleando el método de SA, M. Mustafa Ertay

Para la solución del problema se utiliza un Algoritmo Genético Simple.

70

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

técnicas “fuzzy” y el método SA. Ying-Pin Chang et. al. [29] regresan al problema presentado en [25, 27] empleando Redes Neuronales Secuenciales y Arreglos Ortogonales para su solución.

Por su parte, Lina Huang et. al. [30], presentan una formulación semejante a la (30), con solo la sustitución de la función de máxima distorsión de tensión por la de potencia reactiva total a la fundamental que en este caso se maximiza, es decir:

et. al. [24] utilizan como función objetivo a minimizar la suma de las corrientes de armónicos en el PCC, es decir.

  min f   Ish  h  

(27)

3. Formulaciones multi-objetivo Ying-Pin Chang y Chi-Jui Wu [25] presentan en 2004 una formulación de optimización multi-objetivo para filtros de diverso tipo, que se basa en tres objetivos fundamentales a minimizar: la distorsión total de la demanda de corriente TDD en el punto común de conexión PCC, el THDV de las tensiones de los nodos del sistema y las pérdidas de los filtros PF. Estos tres objetivos se unen en una sola función a minimizar compuesta por:

min f1  THDI  min f 2  C F 

max f 3  QF1  Esta formulación se resuelve con ayuda del método de PSO. Una formulación idéntica a (31) es considerada en 2008 y 2009 por Na He et. al. [31,32]. En la misma se optimizan filtros sintonizados y de segundo orden teniendo en cuenta casos pesimistas de desvalorización de las componentes de los filtros. Se utilizan restricciones especiales del tipo (12) para evitar resonancias en el sistema.

Chia-Nan Ko et. al. [33] regresan en 2009 a las formulaciones [25,27,29] con la modificación de añadir las pérdidas de los filtros al problema de la siguiente manera:

  min  f  a  TDD   b  THDV i  c  PF  (28) i   Donde las constantes a, b y c, que son suministradas por el usuario, determinan la importancia relativa de cada uno de los objetivos a minimizar.

Las restricciones consideradas incluyen las establecidas por la IEEE Std. 519-92 [26], es decir: las restricciones de máximo THDV (5) y de distorsión individual (10) de las tensiones, así como las restricciones de máxima TDD y máxima distorsión individual de la demanda de corriente en el PCC.

TDD  TDD (max)

(29)

I h / I L  IDDh (max)

(30)

n   min  f   wi  THDVi  wn1  TDD  wn 2  PF  wn3  PF  i 1   (33)

La solución se realiza por el método PSO con “Nonlinear Time-varying Evolution” (PSO-NTVE).

Adel M. Sharaf y Adel A. A. El-Gammal [34] plantean un método en 2009 que utilizando el método de “Discrete Particle Swarm Optimization (DPSO)”, minimiza una función compuesta por varios objetivos: mínima corriente de distorsión Ish en el sistema, máxima corriente de armónicos Ifh en el filtro, mínima tensión armónica Vh en el punto de conexión del filtro, todos con respecto a la corriente de armónico de la carga Ih, así como mínima relación cuadrática de la corriente del sistema con respecto a la del filtro.

Donde IL es la corriente máxima promedio de la instalación industrial.

Es interesante, que los límites de distorsión para las frecuencias inferiores a la de sintonía de los filtros se reducen a la tercera parte, para evitar el fenómeno de resonancia a estas frecuencias inferiores.

En 2005, Yin-Pin Chang et. al. [27] continúan la formulación anterior considerando un enfoque probabilístico para los escenarios de operación de la red y la desvalorización de los componentes de los filtros.

Yuan-Lin Chen [28] presenta en este mismo año un enfoque multiobjetivo basado en tres funciones independientes a minimizar: distorsión total de la corriente THDI en el PCC, máxima distorsión total de tensión THDV en el sistema y costo de los filtros sintonizados a emplear.

min  f   1 ( Is h / I h )  ...

 2 /( IF h / I h )   3 (V h/ I h )   4 ( Is h / IFh ) 2 

(34)

En otra contribución [35], estos autores utilizan el método “Discrete Multi-Objective Particle Swarm Optimization (MOPSO)” y reformulan el problema mediante las tres funciones objetivos siguientes:

min f1  THDI 

min f 2  max k THDVk 

(32)

min  f1  Is h / I h 

(31)

max f1  IFh / I h 

min f 3  C F 

(35)

min  f 3  Vh / I h 

El problema se resuelve con un procedimiento basado en

En 2010 y 2011, José M. Maza-Ortega et. al. [36,37] 71

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

max f1  QF1 

publican un programa para diseño de filtros que puede emplear diferentes tipos de funciones objetivos, dentro de las que se encuentran: mínima corriente RMS en el sistema, mínimo THDI de dicha corriente, mínimo THDV, combinación lineal de THDI y THDV, costo de los filtros, etc. Se expresan rigurosamente un conjunto de restricciones para evitar resonancias del sistema con los filtros.

Vishal Verma y Bhim Singh [38] presentan una función objetivo compuesta para optimizar con SGA. Esta función incluye la minimización de: la corriente de armónicos Ish en el sistema, la corriente de frecuencia fundamental en el filtro (equivale a minimizar la admitancia del filtro a dicha frecuencia 1/ZF1) y la diferencia entre la potencia reactiva deseada a frecuencia fundamental Qs1 y la suma de potencia reactiva generada por cada filtro k, QFk,1.

  min  f   

2          Is  Qs QF   h  h ZF  1 k k ,1   1  

2

min f 2  THDI 

min f 3  THDV  min f 4  C F 

Finalmente, Junpeng Ji et. al. [44], presentan otra aplicación del método de PSO para resolver una formulación multiobjetivo compuesta por tres funciones a minimizar: el costo de los filtros, el negativo de la potencia reactiva a la fundamental (para maximizar esta) y una suma pesada de las distorsiones de tensión y corriente.

min f1  C F 

min f 2  C  CF1 

Donde C es una constante de alto valor para que f2 siempre sea positiva.

4. Conclusiones A partir del análisis realizado, pueden formularse las siguientes conclusiones: 1) La diferencia fundamental entre las formulaciones mono-objetivo y las de múltiples objetivos es que las primeras se dirigen generalmente a determinar los filtros de mínimo costo que garantizan el cumplimiento de las restricciones de distorsión pertinentes, mientras que las segundas agregan como objetivos la minimización de los índices de distorsión fundamentales.

2) Las formulaciones multi-objetivo son superiores a las primeras, ya que: - El cumplimiento de las normas de calidad de la energía no implica la ausencia de problemas por este concepto.

La situación ideal es la reducción a cero de la distorsión de tensión y corriente.

- La máxima eficiencia en las funciones de filtrado solo se consigue expresando dichas funciones dentro de los objetivos de la optimización.

- Una solución de igual o ligeramente inferior efectividad económica estimada puede tener un desempeño muy superior desde el punto de vista de la reducción de los índices de distorsión.

3) Varias formulaciones multi-objetivo utilizan un método de optimización no adaptado a este tipo de problemas. Las mismas emplean una única función objetivo compuesta por la suma ponderada de varios sub-objetivos, lo cual trae aparejado varios problemas: - La solución que se obtiene es altamente dependiente de los pesos relativos de cada objetivo, los que se dejan a la selección del usuario.

- Los diferentes tipos de objetivos pueden tener distinto significado físico y magnitud.

- Se obtiene una única solución al problema.

4) Las formulaciones multi-objetivo que utilizan métodos desarrollados para este tipo de problemas, permiten la obtención de un conjunto de soluciones óptimas que constituyen la frontera de Pareto del problema. Una vez

(37)

Sanjeev Singh y Bhim Singh [40] utilizan en 2011 un algoritmo basado en PSO para maximizar una función objetivo que comprende el promedio del factor de potencia FPk y la reducción del THDIk para un conjunto de M estados de carga diferentes:

 max  f  

M

1 M

 w P F k 1

1

k

  w2 (1  THDI k )  

(38)

La optimización se realiza para filtros sintonizados y de segundo orden para convertidores de doce pulsos.

Hao Yue et. al. [41] presentan una aplicación del Algoritmo Genético por Ordenamiento No-dominado NSGA-II [42] al problema de los filtros de armónicos. La formulación utilizada incluye la minimización de: el costo de los filtros CF y las pérdidas en la red ∆P.

min f1  C F 

min f 2  P

(41)

max f 3  1  THDV   2  THDI 

  (36)   

Otra función objetivo compuesta se presenta por Rachid Dehini y Slimane Sefiane en 2011 [39] para ser resuelta con “Ant Colony Optimization (ACO)”. La misma incluye el costo de los filtros sintonizados y el THDI de la corriente en el PCC.

min f  C F  THDI 

(40)

(39)

En 2012, Shengqing Li et. al. [43] aplican el método de “Multi-island PSO” para optimizar una formulación con cuatro objetivos diferentes: maximizar la potencia reactiva a la frecuencia fundamental, así como minimizar el THDI, el THDV y el costo de los filtros.

72

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

obtenido este conjunto de soluciones, el usuario debe ejercer su criterio para seleccionar cual o cuales de estas soluciones son las mejores para el problema considerado.

[18] Chang, G.W., Wang, H.-L., Chuang, G.S. and Chu, S.-Y., Passive harmonic filter planning in a power system with considering probabilistic constraints, IEEE Transactions on Power Delivery, 24 (1), pp. 208-218, 2009.

[19] Hong, Y.-Y. and Chiu, Ch.-S.. Passive filter planning using simultaneous perturbation stochastic approximation, IEEE Transactions on Power Delivery, 25 (2), pp. 939-946, 2010.

[20] Zobaa, A.F., Vaccaro, A., Zeineldin, H.H., Lecci, A. and AbdelMonem, A.M., Sizing of passive filters in time-varying nonsinusoidal environments, pp. 1-8, IEEE 2010. DOI: 10.1109/ICHQP.2010.5625346 [21] Kovernikova, L.I. and Thanh, N.Ch., An optimization approach to calculation of passive filter parameters based on particle swarm optimization, International Conference on Renewable Energies and Power Quality (ICREPQ’12), Santiago de Compostela, Spain, 2012.

[22] Aleem, S.H.E.A, Zobaa, A.F. and Aziz, M.M.A., Optimal C-type passive filter based on minimization of the voltage harmonic distortion for nonlinear loads, IEEE Transactions on Industrial Electronics, 59 (1), pp. 281-289, 2012.

[23] Pandi, V.R., Zeineldin, H.H. and Xiao, W., Passive harmonic filter planning to overcome power quality issues in radial distribution systems, IEEE Conference Power and Energy Society General Meeting, pp. 1-6, IEEE 2012. DOI: 10.1109/PESGM.2012.6345247 [24] Ertay, M.M., Tosun, S. and Zengin, A., Simulated annealing based passive power filter design for a medium voltage power system, International Symposium on Innovations in Intelligent Systems and Applications (INISTA) 2012, pp.

1-5, 2012.

DOI: 10.1109/INISTA.2012.6247042 [25] Chang, Y.-P. and Wu, Ch.-J., Design of harmonic filters using combined feasible direction method and differential evolution, International Conference on Power System Technology, POWERCON 2004, Singapore, Vol. 1, pp. 812-817, 2004. DOI: 10.1109/ICPST.2004.1460105 [26] IEEE Std. 519-1992, IEEE Recommended Practices and Requirements for Harmonic Control in Electrical Power Systems, 1993.

[27] Chang, Y.-P., Tseng, W.-K. and Tsao, T.-F., Application of combined feasible-direction method and genetic algorithm to optimal planning of harmonic filters considering uncertainty conditions, IEE Proc. Generation Transmission and Distribution, 152 (5), pp. 729-736, 2005.

[28] Chen, Y.-L., Optimal multi-objective single-tuned harmonic filter planning, IEEE Transactions on Power Delivery, 20 (2), pp. 11911197, 2005.

[29] Chang, Y.-P., Low, Ch. and Wu, Ch.-J., Optimal design of discretevalue passive harmonic filters using sequential neural-network approximation and orthogonal array, IEEE Transactions on Power Delivery, 22 (3), pp. 1813-1821, 2007.

[30] Huang, L., He, N. and Xu, D., Optimal design for passive power filters in hybrid power filter based on particle swarm optimization, Proceedings of the IEEE International Conference on Automation and Logistics, Jinan, China, pp. 1468-1472, 2007. DOI: 10.1109/ICAL.2007.4338802 [31] He, N., Huang, L., Wu, J., Xu, D., Study on optimal design method for Passive Power Filters set at high voltage bus considering many practical aspects, Applied Power Electronics Conference and Exposition APEC 2008, pp.

396-401, 2008.

DOI: 10.1109/APEC.2008.4522752 [32] He, N., Xu, D. and Huang, L., The application of particle swarm optimization to passive and hybrid active power filter design, IEEE Transactions on Industrial Electronics, 56 (8), pp. 2841-2851, 2009.

[33] Ko, Ch.-N., Chang, Y.-P. and Wu, Ch.-J., A PSO method with nonlinear time-varying evolution for optimal design of harmonic filters, IEEE Transactions on Power Systems, 24 (1), pp. 437-444, 2009.

[34] Sharaf, A.M. and El-Gammal, A.A.A., A discrete particle swarm optimization technique (DPSO) for power filter design, 4th International Design and Test Workshop (IDT), pp. 1-6. 2009. DOI: 10.1109/IDT.2009.5404376 [35] Sharaf, A.M. and El-Gammal, A.A.A., A novel discrete multiobjective particle swarm optimization (MOPSO) of optimal shunt

Agradecimientos A el Instituto de Tecnología Galileo de la Amazonía ITEGAM, a la Fundación para la Investigación del Estado de Amazonas - FAPEAM por apoyar financieramente esta investigación.

Referencias [1] [2] [3] [4] [5]

[6]

[7]

[8]

[9] [10] [11] [12] [13] [14]

[15]

[16]

[17]

Dugan, R.C., McGranaghan, M.F., Santoso, S. and Beaty, H.W., Electrical Power Systems Quality, Second Edition 2004.

Czarnecki, L.S. and Ginn III, H.L., The effect of the design method on efficiency of resonant harmonic filters, IEEE Transactions on Power Delivery, 20 (1), pp.286-291, 2005.

Ginn III, H.L., Improvement of resonant harmonic filter effectiveness in the presence of distribution voltage distortion, PhD Dissertation Louisiana State University, USA, 2002.

Arrillaga, J. and Watson, N.R., Power systems harmonic, 2nd ed, New York Wiley, 2003.

Au, M.T. and Milanovic, J.V., Planning Approaches for the Strategic Placement of Passive Harmonic Filters in Radial Distribution Networks, IEEE Transactions on Power Delivery, 22 (1), pp.347-353, 2007.

Stone, P.E.C., et al., Efficient harmonic filter allocation in an industrial distribution system, IEEE Transactions on Industrial Electronics, 59 (2), pp.

740-751 2012.

DOI: 10.1109/TIE.2011.2157279 Cheng H., et al., New method for both harmonic voltage and harmonic current suppression and power factor correction in industrial power systems, Proceedings of the 1995 Industrial and Commercial Power Systems Technical Conference, DOI: 10.1109/ICPS.1995.526986, IEEE 1995.

Chang, T.-T. and Chang, H-Ch, Application of differential evolution to passive shunt harmonic filter planning, international conference on harmonics and quality of power ICHQP '98, Athens, Greece, Oct. 1998. pp; 149-153, DOI: 10.1109/ICHQP.1998.759866 Hsiao, Y.-T., Design of filters for reducing harmonic distortion and correcting power factor in industrial distribution systems, Tamkang Journal of Science and Engineering, 4 (3), pp.193-199, 2001.

Pérez-Abril, I. and González-Quintero, J.A., VAR Compensation by sequential quadratic programming, IEEE Transactions on Power Systems, 18 (1), pp.36-41, 2003.

Zobaa, A.F., Cost-effective applications of power factor correction for nonlinear loads, IEEE Transactions on Power Delivery, 20 (1), pp.359-365, 2005. DOI: 10.1109/TPWRD.2003.817532 Zobaa, A.F., Maintaining a good power factor and saving money for industrial loads, IEEE Transactions on Industrial Electronics, 53 (2), pp. 710-711, 2006.

IEEE Std. 18-1992, IEEE Standard for shunt power capacitors, 1992.

Chang, G.W., Chu, S.-Y. and Wang, H.-L.. A new method of passive harmonic filter planning for controlling voltage distortion in a power system, IEEE Transactions on Power Delivery, 21 (1), pp.305-312, 2006.

Ghiasi, M., Rashtchi, V. and Hoseini, H., Optimum location and sizing of passive filters in distribution networks using genetic algorithm, International Conference on Emerging Technologies IEEE-ICET 2008, Rawalpindi, Pakistan, pp.18-19, 2008 Ruihua, Z., Yuhong, L. and Yaohua, L., Optimal parameters for filter using improved genetic algorithms. International Conference on Sustainable Power Generation and Supply, SUPERGEN 2009. pp. 1-5, 2009. DOI: 10.1109/SUPERGEN.2009.5348353 Luiz, A.-S.A. and Cardoso-Filho, B.J., Minimum reactive power filter design for high power converters, 13th International Power Electronics and Motion Control Conference. EPE-PEMC 2008, pp.

1345-1352, 2008. DOI: 10.1109/EPEPEMC.2008.4635455.

73

Cabral-Leite et al / DYNA 81 (188), pp. 67-74. December, 2014.

[36]

[37]

[38] [39]

[40]

[41]

[42] [43]

[44]

power filter, IEEE/PES Power Systems Conference and Exposition, PSCE '09. pp. 1-7, 2009. DOI: 10.1109/PSCE.2009.4839957 Maza-Ortega, J.M. and Burgos-Payán, M., A Software-based tool for optimal design of passive tuned filters, IEEE International Symposiun on Industrial Electronic, Bari, Italy, pp. 3273-3278, 2010.

Churio-Barboza, J.C., Maza-Ortega, J.M. and Burgos-Payán, M., Optimal design of passive tuned filters for time varying non-linear loads, IEEE Proceedings of the International Conference on Power Engineering, Energy and Electrical Drives, Torremolinos, Málaga, Spain. pp. 1-6, 2011. DOI: 10.1109/PowerEng.2011.6036491 Verma, V. and Singh, B., Genetic-Algorithm-based design of passive filters for offshore applications, IEEE Transactions on Industry Applications, 46 (4), pp. 1295-1303, 2010.

Dehini R. and Sefiane, S., Power quality and cost improvement by passive power filters synthesis using ant colony algorithm, Journal of Theoretical and Applied Information Technology 23 (2), pp. 7079, 2011.

Singh S. and Singh, B., Passive filter design for a 12-pulse converter fed LCI-synchronous motor drive, Joint International Conference on Power Electronics, Drives and Energy Systems (PEDES) Power, India, pp. 1-8, 2010. DOI: 10.1109/PEDES.2010.5712371 Yue, H., Li, G., Zhou, M., Wang, K. and Wang, J., Multi-objective optimal power filter planning in distribution network based on fast nondominated sorting genetic Algorithms, 4th International Conference on Electric Utility Deregulation and Restructuring and Power Technologies (DRPT), pp. 234-240, 2011. DOI: 10.1109/DRPT.2011.5993895 Frutos, M. and Tohmé, F., Evolutionary multi-objective scheduling procedures in non-standardized production processes, DYNA, 79 (172), pp. 101-107, 2003.

Li, S., Li, Y., Luo, X., Zeng, L. and He, Z., Multi-objective optimal design for passive power filters in hybrid power filter system based on multi-island particle swarm optimization, IEEE 7th International Power Electronics and Motion Control Conference - ECCE Asia, Harbin, China, Vol. 4, pp. 2859-2863, 2012. DOI: 10.1109/IPEMC.2012.6259320 Ji, J., Liu, H., Zeng, G. and Zhang, J., The Multi-objective optimization design of passive power filter based on PSO, AsiaPacific Power and Energy Engineering Conference (APPEEC), pp.

1-4, 2012. DOI: 10.1109/APPEEC.2012.6307477

Computacionales con énfasis en Modelos Analíticos y simulación de sistemas. Es coordinador de proyectos en ITEGAM. M. E. de Lima-Tostes, graduada de Ing. Eléctrica en 1997 de la Universidad Federal de Pará (UFPA), Brasil. MSc. en Ing.Eléctrica en 1992 y Dra. en Ing. Eléctrica en 2003 por la Universidad Federal de Pará (UFPA), Brasil. Su área de interés es la calidad de la energía eléctrica en los sistemas eléctricos de potencia.

U. Holanda-Bezerra, graduado en Ing. Eléctrica en 1976 de la Universidad Federal de Pará (UFPA), Brasil. MSc. en Ing. Eléctrica en 1980 de la Universidad Federal Escuela de Itajubá (EFEI)-MG y Dr. en Ing. Eléctrica en 1988, de la Universidad Federal de Rio de Janeiro (COPPE/UFRJ), Brasil. Es profesor de la UFPA desde 1977. Actualmente es profesor en el área de Conversión de Energía. Su área de interés es Seguridad, Calidad de la Energía y Energía Renovable.

Área Curricular de Ingeniería Eléctrica e Ingeniería de Control

J. Cabral-Leite, graduado en Matemática en la Universidad Federal de Rondónia (UNIR) en 1987 y en Ingeniería en Producciones Eléctricas del Centro de Educación Superior FUCAPI en 2006, MSc en Ing. Industrial y en Sistemas en la Universidad Federal de Santa Catarina (UFSC) en 2001 y Dr. en Ing. Eléctrica de la Universidad Federal de Pará (UFPA) en 2013. Es Director Presidente e investigador del Instituto de Tecnología y Educación Galileo de la Amazonía (ITEGAM), Brasil. Su área de interés incluye calidad de la energía, análisis, diseño y optimización de sistemas eléctricos de potencia.

Oferta de Posgrados 

I. Pérez-Abril, graduado en Ing. Eléectrica en 1984, en la Universidad Central de Las Villas, Cuba. Dr. en Ing. Eléctrica en 1995, de la Universidad Central de Las Villas, Cuba. Es director del Centro de Estudios Electroenergéticos y profesor de Sistemas Eléctricos Industriales de la Universidad Central de Las Villas, Cuba. Su área de interés incluyes calidad de la energía y análisis, diseño y optimización de sistemas eléctricos

Maestría en Ingeniería - Ingeniería Eléctrica

Mayor información: Javier Gustavo Herrera Murcia Director de Área curricular ingelcontro_med@unal.edu.co (57-4) 425 52 64

de potencia.

M.S. Santos-Azevedo, graduado en Procesamiento de Datos en 1999 y en Matemática en 1985 de la Universidad Federal de Amazonas, Brasil. MSc. en Ingeniería Electrica en 2006 de la Universidad Federal de Campina Grande, Brasil. Es profesor asistente de la Universidad del Estado de Amazonas (UEA), Brasil.

Tiene experiencia en el área de Sistemas

74

Restricting the use of cars by license plate numbers: A misguided urban transport policy Víctor Cantillo a & Juan de Dios Ortúzar b a

Department of Civil and Environmental Engineering, Universidad del Norte, Barranquilla, Colombia. vcantill@uninorte.edu.co Department of Transport Engineering and Logistics, Pontificia Universidad Católica de Chile, Santiago, Chile. jos@ing.puc.cl

b

Received: October 1th, 2013. Received in revised form: July 30th, 2014. Accepted: October 25th, 2014.

Abstract Several conurbations in Latin America have implemented policies to restrict the use of private cars by prohibiting those with a license plate number ending on some digit, which vary each day, from circulating on that weekday. These policies have been formulated as a “second best” option to congestion charging as the latter policy has been deemed politically unfeasible so far. The idea underlying license plate restrictions is very simple: by not allowing a proportion of the total number of vehicles to circulate on a given day, congestion levels and other externalities are expected to decrease, improving the quality of life for people.

Based on a simple microeconomic analysis, supported by a review of the literature and by evidence collected in several cities where these restrictions have been implemented, it is concluded that the policy is apparently effective only in the very short-term, but ultimately it fails to achieve its objectives and usually leads to a worse outcome.

Keywords: Car restrictions; externalities; urban policies.

Restricción vehicular según número de patente: Una política de transporte errónea Resumen Varias ciudades latinoamericanas han implementado políticas restrictivas al uso de vehículos particulares consistentes en prohibir su circulación según el número de su patente o placa. La política ha sido planteada como una alternativa de segunda mejor, al cargo por congestión; este último consiste en el cobro directo de la externalidad a los usuarios de vías congestionadas, pero es considerado políticamente inviable en la actualidad en muchos casos. La idea detrás de la restricción según patente es muy simple: al restringir una proporción del parque automotor se disminuye los niveles de congestión y otras externalidades negativas, especialmente las ambientales, mejorando la calidad de vida de las personas.

A través de un sencillo análisis microeconómico, apoyado en evidencia recogida en algunas ciudades donde se ha implementado esta política, se demuestra que sólo puede funcionar en el muy corto plazo y, finalmente, no logra el objetivo deseado en el largo plazo.

Palabras clave: Restricción vehicular; externalidades; políticas urbanas.

1. Introduction Cities in the developing world have been experiencing a rapid increase in motorization. To deal with the growing congestion, smog, and other externalities associated with urban transport, it has become common for some conurbations in Latin America to restrict the use of private cars, by prohibiting vehicles with license plate numbers ending with certain pre-specified digits from circulating on certain days of the week. In Santiago de Chile, since 1986, vehicles lacking catalytic converters have been subject to

this restriction during the autumn and winter periods. More recently, this restriction has also applied to vehicles with catalytic converters on emergency days [1]. In 1989, the government of Mexico City introduced the Hoy No Circula (which roughly translates as "Do Not Circulate Today") program, which bans drivers from using their cars on one day of the week also based on the last digit of the cars' license plate number; nowadays, a restriction is in force in the central zone of Mexico´s metropolitan area, where 20% of all vehicles concentrate daily. In Colombia, Bogotá instituted the vehicular restriction known as Pico y Placa

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 75-82. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.40081

Cantillo & Ortúzar / DYNA 81 (188), pp. 75-82. December, 2014.

greenhouse gases resulting in a rise in global temperature and climate change. A troubling aspect of externalities from the transport sector is that they are increasing rapidly in Latin America and other developing countries where, with the ongoing rapid increase in population, expansion of middle class, and availability of cheaper vehicles, the always present desire to own private vehicles is more within reach now than ever for millions of people.

Fig. 1 illustrates the economic analysis of externalities related to transport. For the sake of simplicity, we shall consider just one origin and a destination connected by a road on which individuals travel in identical vehicles in static equilibrium [7]. Therefore, we are ignoring other effects such as a redistribution of trips to other modes of transport (in the case of compulsory trips) or to other days, in the case of non-compulsory trips. Neither are we including induced demand by users with a right to circulate taking advantage of the improved flow conditions in the short term. Moreover, the traffic flow, speed, and density are the same all along the road. The horizontal axis indicates the volume of traffic flow (v), while the vertical axis represents the generalized cost of the trip (C), which includes operational costs, travel time, and other charges. When traffic volumes are low, vehicles can travel at their desired speeds (i.e. free flow). When flows (v) increase, congestion arises, the speed of traffic decreases, and costs also increase. In this model we are not considering other effects such as mode changes and redistribution of trips.

The model proposed is a static analysis of congestion. In Fig. 1, the AvC(v) curve represents the average (or private) costs perceived by drivers and the MgC(v) curve represents the marginal (or social) costs; both costs are related by equation (1), where the second term on the right represents the externalities; i.e., the (marginal) social costs are the (average) private costs plus the externalities.

(which roughly translates as "Peak and Plate") in 1998. But, unlike the previous cities' restrictions, this measure was a response to traffic congestion and not to poor air quality [2].

The measure has been extended to other Colombian cities, including Medellín in 2005.

Outside Latin America, Manila has had a license plate number restriction in place for some time prohibiting certain vehicles from circulating on high traffic arteries during the rush hour [3]. An interesting case is the special odd–even license plate restriction that was used in Beijing during the 2008 Olympic Games where an evident reduction of congestion and mobile source pollution was confirmed during this period [4]. The benefits of such restriction seemed so attractive that authorities decided to resume it with a similar, but less restrictive, one-day-a-week licensing scheme, which stops 20% of private vehicles from using roads on weekdays.

According to local planners and politicians, the policy constitutes a second best alternative to congestion charging (which entails directly charging drivers using congested roads for a specific externality) [5], as any class of road pricing has been considered politically unfeasible so far.

The virtues of this simple policy include its acceptance by large sectors of society, especially those who do not own cars (i.e. the majority of Latin Americans). Furthermore, as a result of the restriction it is expected that some drivers may experiment with using public transport and this is clearly a desirable effect.

Nevertheless, several disadvantages associated with this measure have also been identified, including the increase in automobile stock due to the fact that drivers with sufficient purchasing power tend to buy a second vehicle (often more polluting), the increase in the number of trips by taxi and motorbike, the risk of fraud inherent to the policy (such as counterfeit license plates), and the difficulty of dealing with exceptions provided for certain types of vehicles, people, or organizations [2]. The literature has examined consumer behavior in relation to these measures; for example, Wang et al. [6] used revealed preference data in Beijing to examine compliance with the measure, finding that almost 48% of drivers did not comply.

In this paper we conduct a detailed analysis of the vehicular restriction policy based on license plate numbers.

We perform a complete review of the literature and also evaluate the policy's short- and long-term effects. Our evaluation is supported by empirical evidence from Latin American cities where the restriction has been applied for many years.

MgC  v   AvC  v   v

AvC  v  v

(1)

If we interpret the flow volume as the number of trips demanded per unit of time, then we obtain the demand curve and the supply-demand diagram shown in Fig. 1, where D1 is defined as the initial demand. If there is no intervention or any additional charge to the operational costs and travel times, then we obtain the initial market equilibrium at flow Vm; the generalized private costs are expressed as GCpM, and the social costs as GCsM. The figure also shows that the social optimum, which corresponds to the intersection of the demand curve and the social costs, should take place when the equilibrium flow is VO and social costs are GCsO. To reach the social optimum it is necessary to charge an amount E (equal to the congestion externality), as the market equilibrium subtracts a social loss (black area) from the social benefit (lightly shaded area). The philosophy underlying the congestion charge policy is to reach an equilibrium corresponding to the social optimum. The policy has been successfully implemented in cities such as London, Singapore, and Stockholm [8]. However, as it is a

2. Economic analysis of license plate restriction policies Being a market with strong externalities is one of transport's most relevant characteristics, especially in urban areas. In simple terms, an externality is an action carried out by an economic agent (company or individual) that has a direct impact upon the productive processes of another company and/or the well-being of another individual.

Transport supply is associated with concomitant effects that introduce distortions, such as congestion, accidents, noise and air pollution, including the emission of global 76

Cantillo & OrtĂşzar / DYNA 81 (188), pp. 75-82. December, 2014.

However, a more detailed analysis of the measure shows that the effect is not the same. We know that the demand curve reflects consumers' willingness to pay, such that, their inability to use their vehicles generates a social loss; this is represented by the darkly shaded area between curves D1 and D2. We must remember that the equilibrium described relates to the short-term.

In the long-term car users take decisions that change the aforementioned equilibrium as illustrated in Fig. 3. There is a willingness to pay among drivers for using their cars, reflected in the area corresponding to the social loss. A large number of car users express this willingness by purchasing a second or third car, with license plate numbers that are not restricted on the same days as the other car(s) they own.

Accordingly, the demand curve stubbornly moves to the right until say position D3, which can, even in a short period of time, coincide with or even be more to the right of D1, thus returning to a situation similar to or worse than the initial situation (remember that the vehicles that enter the system in this manner are neither new nor in optimal condition, as they are only sporadically used.).

Clearly, in time the desired effects are lost both with respect to congestion and pollution. This latter issue is exacerbated by the fact that, as pointed out earlier, not only do new vehicles enter the system but also older and more polluting vehicles which are cheaper to acquire as â&#x20AC;&#x153;replacement carsâ&#x20AC;?. In fact, such vehicles are generally purchased by people who lack the disposable income needed to buy a new second car. Another aggravating effect is that all these families, once they buy their second car, tend to use both of them on days where neither of the vehicles is restricted. This is confirmed by the noticeable increase in car flows on Saturdays and Sundays, and at the hours during which the restriction does not apply.

It is worth mentioning that prior to their implementation these restrictive measures received, in all cases, strong opposition from automobile users and certain business guilds (especially merchants). However, they were ultimately accepted thanks to their apparent initial success.

Nevertheless, with time these initial benefits have become less clear largely because of an increase in automobile stock. In response to this, the authorities have tended to make the measures even harsher in some places by increasing the license plate numbers subject to the restriction on a given day, extending the hours during which the restriction applies, or eliminating exceptions. On the other hand, cities commonly rotate the days on which certain license plate numbers are restricted.

Figure 1. Market equilibrium and social optimum in a congested road.

Source: Adapted from Lindsey, R. and Verhoef, E. 2001

Figure 2: Equilibrium after applying the restriction policy.

Source: Own drawing

supply oriented measure which entails charging drivers, it has not yet proven possible to implement it in Latin America (and in developing countries in general), as it is perceived as politically unfeasible [9].

In some cities decision makers have preferred restrictive circulation measures applied to demand. The logic of the implementation is very simple: move the demand curve in order to achieve a new equilibrium with lower costs. As shown in Fig. 2, the license plate restriction measure moves the demand curve to position D2 toward the left of the original curve D1. For our analysis, we shall assume the most favorable of circumstances, in which the movement is such that the new market equilibrium, which corresponds to a private cost GCpM, leads to a flow that is identical to the social optimum, i.e., VO. In other words, we shall assume that there is no need to implement the apparently unfeasible congestion charge in order to achieve flows similar to those that would have been achieved with this optimal policy.

3. Ex-post evaluation of the policy in Latin American cities Various research projects support the thesis that congestion charging (i.e., road pricing) is more efficient than draconian measures, such as prohibiting circulation based on license plate numbers [10]; this is because car users can keep driving by paying the externality and the income generated from road pricing can be used for a variety of purposes, such as, for example, substantially improving public transport services (as has occurred in 77

Cantillo & Ortúzar / DYNA 81 (188), pp. 75-82. December, 2014.

4, or 6 digits were subject to restriction), Fresard [11] concluded that there was no noticeable effect between the restrictions for two or four digits although a 10% decrease was expected. Also, when the restriction was extended from two to six digits, the average reduction observed only slightly surpassed 5% (although a 20% reduction was expected). Another interesting fact is that the measure had a lesser effect upon higher income automobile users. His conclusions were that the measure has been ineffective (i.e., it did not achieve the desired effects) as people found ways to avoid it, such as buying a second car. Finally, people's travel habits underwent changes reflected by the fact that some choose to travel at different hours; but the increased flows during hours when the restriction did not apply were also due to the fact that people started to move spontaneously away from the more congested times.

A paper by De Grange and Troncoso [12] looked again at the effects of the two restrictions that have been applied in Santiago (i.e. the traditional measure applied from April through August to vehicles without catalytic converters and the additional measure that bans the use of vehicles with catalytic converters between 7:30 am and 9:00 pm on days declared as environmental ‘‘pre-emergencies’’); they conclude that the traditional restriction has had no impact on the use of private cars while the additional restriction curtailed their use only by 5.5% when the restriction applied to 20% of cars.

Figure 3: Long-term equilibrium after applying the vehicular restriction.

Source: Own drawing

London). We show below that the evidence from four cities where cars are restricted based on license plate numbers: Santiago (Chile), México City, Medellín and Bogotá, supports this view.

This analysis is also supported by recent research on vehicle restriction and data from various sources in the cities considered. The main variables analysed are pollutant emissions, changes in car registrations and car flow measurements. Unfortunately, the available information is not the same in all cases making it difficult to build homogeneous indicators. Nevertheless, it is possible to derive general conclusions on the implementation of the policy.

3.2. Mexico City Similar to the case of Santiago, car restrictions in Mexico City were motivated by environmental reasons.

Davis [13] analyzed the effect of the restriction implemented in 1989 (Hoy No Circula) on air quality, by making high frequency measurements at several monitoring stations. His work shows no evidence that the restriction improved air quality; furthermore, it reveals that the restriction undoubtedly led to an increase in the number of vehicles in circulation and to a shift in the composition of the automobile stock towards more polluting vehicles from other cities. Also, his analysis shows that there was a relative increase in air pollution and noise on weekends and during the hours of the day when the restriction did not apply. Moreover, while it was expected that the program would encourage drivers to use low-emission means of transport, contrary evidence was found suggesting that drivers tended to use taxis when they could not use their cars.

To analyze the effects of the Hoy no Circula (HNC) policy on car use, Gallego et al. [14] developed a (linearcity) model that captures in a simple way two essential elements of the problem: the allocation of existing vehicle capacity to competing uses (peak vs. off-peak hours) and how that capacity is adjusted in response to a policy shock, because their short and long run effects can vary widely (i.e., whether and how fast households adjust their vehicle stock). They argue that, potentially, the driving restriction can also have a positive effect on car travel in the long-run giving that for some households owning a car is not that attractive any more (although using it, is). However, other

3.1. Santiago, Chile Since 1986, in Santiago, which has suffered intense pollution problems due to its geographical setting, during autumn and winter (where a thermal inversion problem sets in) vehicles lacking catalytic converters have been subject to a restriction based on the last digit of their license plate numbers. In the last decade, this restriction has also applied to vehicles with catalytic converters on emergency days.

Originally the latter measure applied to 20% of these cars, but in 2008 it was increased to 40% of the fleet.

Fresard [11] investigated the real effects of the vehicular restriction in Santiago. He found that an important disadvantage was that it did not distinguish between drivers with different income levels who obviously ascribed different subjective values to time; thus, higher income drivers simply used less polluting cars that were unrestricted. Another, more undesirable consequence, was that a large proportion of families purchased a second car in order to avoid the restriction. Moreover, these second cars were in most cases older and more polluting than the families' first cars.

Based on the traffic capacities at periods corresponding to different vehicular restriction conditions (i.e. whether 2, 78

Cantillo & Ortúzar / DYNA 81 (188), pp. 75-82. December, 2014.

households will find attractive to increase the size of their car-bundle and buy a second car; not only by-passing the driving restriction altogether but, what is worse, increasing car travel during both peak and off-peak. Obviously, the increase in stock is much larger as only few households consider the possibility of selling their cars because of the policy.

Clearly, then, in a short amount of time the desired effects of the policy are lost both with respect to congestion and pollution. This latter issue is exacerbated by the fact that, as pointed out earlier, not only do new vehicles enter the system as usual but also, as an effect of the policy, older and more polluting vehicles enter the system too, usually from other cities. Such vehicles are generally purchased by people who lack the disposable income needed to buy a new second car. Another aggravating effect is that all these families, once they buy their second car, tend to use both vehicles on days where neither of them is restricted.

Likewise, there is a noticeable increase in car flows on Saturdays and Sundays, and at the hours during which the restriction does not apply.

Other empirical results for HNC [14] show statistically significant reductions of CO in the short run (i.e., during the first month of implementation), of 11% and 8% for peak and off-peak hours, respectively. This short-run result is entirely consistent with the perception that compliance with the program has always been high. In the longer run though (i.e., after the first year of implementation), estimates show an increase of 13% during peak hours and of 8% during of off-peak hours. On the other hand, estimates for weekends show no reduction in the short-run, as expected, and a significant increase in the long run of 14%. These results are consistent with the fact that the policy effectively constrained week days only; hence, a strong immediate effect should be seen in peak hours and no effect during weekends. Additionally, the results are consistent with the hypothesis that in the margin households will tend to buy cars to try and return to their pre-policy use in the peak, but this behaviour will leave them with additional cars to use during weekends and, eventually, off-peak hours. This, in turn, should increase pollution and congestion during such hours.

Similar evidence was found by Eskeland and Feyzioglu [15], who demonstrated that the ban restricting cars from circulating on a specified weekday had increased total driving in Mexico City. Because of the ban, cars effectively represented “driving permits,” and some households just bought an additional one and increased their total driving.

Greater use of old cars, congestion effects, and increased weekend driving may also have contributed to the disappointing results. The ban has high welfare costs and does not deliver the intended benefits of reduced driving— quite the contrary. Conclusions were founded in the estimation of a demand function based on time series, aggregate data, to analyze the aggregate reductions in driving resulting from the driving ban. The time series analysis indicated, strongly, that total car use in Mexico City shifted upwards as a consequence of the regulation, suggesting that positive net car purchases should play a major role (since one would expect the many households

Table 1.

Effect of car restriction on variation of car ownership in Mexico City (annual average).

Before Under Indicator Regulation Regulation 1983-89 1990-93 154 Sales of new vehicles Mexico City 80 (thousands) Rest of Mexico 127 237 Increase in vehicles Mexico City 7 239 registered (thousands) Rest of Mexico 174 250 Net import of vehicles Mexico City -74 85 (thousands) Rest of Mexico 47 13 GDP (% growth) Mexico City 1.1 3.8 Whole country 1.1 4.2 Metro ridership in Mexico City(percentage growth) 5.7 -2.4 Source: Asociación Mexicana de la Industria Automotriz and Instituto Nacional de Estadística y Geografía, cited by [15]

Table 2.

Evolution of mode share in Mexico City.

Mode

1983

Bus/Minibus 48.2 Taxi 5.0 Car 23.0 Light Rail - Trolley 1.7 Metro 22.1 Source: Cometravi, 1999. [16]

1986 47.8 5.0 24.9 3.2 19.1

1989 53.6 5.9 16.4 3.1 21.0

1992 59.7 8.2 17.8 1.1 13.2

1995 55.0 8.5 22.2 1.2 13.1

that would not increase car ownership to reduce, or keep unchanged, their car use).

Aggregate annual data presented in Table 1 support these conclusions; it shows that car restrictions turned Mexico City from net used car exporter (exporting an average of 74,000 cars per year) to importer (importing an average of 84,000 cars per year). Table 1 also indicates that after regulation ridership of the metro system declined. It is worth indicating that although the average increase in gross domestic product of Mexico City was higher than in the rest of the country during the years following the implementation of the car restriction, this fact does not fully explain the changes listed. On the other hand, Table 2 shows the evolution of mode share in Mexico City. It is evident with when HNC was imposed, in 1989, car´s market share decreased; however, few years after, car users increase to original levels.

HNC results are consistent in showing how rapidly households adjust their vehicle stocks leaving little room for policy makers to make ex-post adjustments to policies It is clear that this policy was ineffective in moving people away from their cars and reducing pollution, but it is less clear how much of that result can be exported to other driving restriction programs including elements that this one did not include, at least in their early years of implementation (i.e., incentives towards a faster and cleaner fleet turnover).

3.3. Bogotá Innovative policies have been implemented in Bogotá to transform a car-centred transportation system into a peopleoriented one. In addition to innovative strategic urban planning, a successful BRT system of three trunk lines 79

Cantillo & Ortúzar / DYNA 81 (188), pp. 75-82. December, 2014.

(Transmilenio) and strong support for the use of bicycles, the city implemented a vehicular restriction system called “pico y placa” in 1998. The Office of the Mayor decided to restrict automobile use at rush hours depending on the cars' license plate numbers. So, initially, 20% of cars could not circulate between 7:00 and 9:00 am and between 5:30 and 7:30 pm on week days. In 2002 the ban was increased to 40% of cars and set from 6:30 to 9:00 and from 17:00 to 19:00. In 2005 it was increased to 6:00 to 9:00 am and 4:00 to 7:30 pm and since 2009 the restriction spans almost the whole day, from 6:00 to 20:00 hrs. Although the measure was proposed as a provisional one, it ended up being adopted permanently by subsequent Mayors. Unlike Mexico City and Santiago, the main motivation for its implementation was congestion, but it was also supported for environmental reasons.

Unfortunately, the Office of the Mayor did not implement a monitoring system for determining the measure's impact on traffic. So, there are no systematic measures of average traffic speeds before and after the introduction of the policy. This lack of information makes it difficult to reach conclusions about certain essential topics, such as: (i) what exactly happened to the 40% of trips by car after the restriction took effect; (ii) which were the changes in the chosen modes of transport; (iii) which trips were no longer made, and (iv) did the hours during which e people travel changed.

It does not seem correct to assume a reduction in the number of vehicles per km (veh/km) equal to the proportion of cars that ceased to circulate (40%). In fact, considering the transport cost flexibility vs. veh/km, it is very likely that some car drivers may have decided to drive more during times when their cars were not subject to restriction, given that the restriction may have temporarily reduced their generalized costs of trips by car. Based on an understanding of certain parameters reflecting the current situation (i.e., with the circulation restriction), especially those relating to speeds and road use, Bocarejo [17] examined the policy's economic impact, calculating the average generalized cost perceived by car users along with the social cost in situations with and without the restriction. He also evaluated the change in consumer surplus. The results of this analysis indicate that the circulation restriction during rush hours in Bogotá (before 2009) could have generated a loss in consumer utility (users) of some 118 million Euros/year.

While the values of time and elasticities considered in the analysis may be improved in a more precise study, this result shows at least the potential harm that a policy of this nature can inflict to society.

Fig. 4 shows (a) the evolution of new registered automobiles and (b) the evolution of GDP in Bogotá and in the rest of Colombia from 1998 (when Pico y Placa was implemented in Bogotá) to 2005 (when Pico y Placa was also implemented in other important cities like Medellín, Cali and Bucaramanga, involving about 75% of the cars in Colombia). The reference point is 1998, to which we assigned a value of 100 in both cases.

Growth in car ownership is linked to an increase in income [19], and although Bogotá has a higher GDP per capita than the rest of Colombia, in the period analysed the capital experienced less growth in its GDP. Between 1998

Figure 4. Evolution of a) New cars registered and b) GDP in Bogotá and in the rest of Colombia.

Source: Ministerio de Transporte and DANE [18]

and 2001, a severe economic crisis reduced the number of automobile registrations, affecting Bogotá more; nevertheless, we can see that in the long term far more new vehicles entered the automobile stock in Bogotá than in the rest of the country during the years in which Pico y Placa has been in force (although tendencies in economic growth were comparable). It is important to mention that in these years no relevant changes in operational or acquisition car costs were observed.

The short term impact of Pico y Placa on the mode share is evident in Table 3. Trips whose purpose was work and education decreased significantly when the policy was implemented. However, in the medium and long term as the table shows, trips in cars return to their original levels.

Table 5 also shows the impact of the introduction of Transmilenio (by December 2000) on modal split.

According to records of the Bogota air quality network, the main pollutant is PM10, which often exceeds the legal norm [20]. Fig. 5a) shows the PM10 annual percentage exceeding set limits (70μg/m3), calculated from the averages reported by the stations, while Fig. 5b) shows the overall averages for the city. It can be seen that with the introduction of the restriction in 1998 there was a decrease in the indicators for that pollutant, but a couple of years later it increased again. Similarly, extensions of the policy had marginal impacts on the PM10 levels. In 2002, 2005 and 2009, the schedule of the restriction was increased, and the Fig. 5 shows that effect.

Observations of classified ads for the sale of used vehicles in Bogotá let see that the last digit of the license plate number is an important decision-making factor. This provides a clear market signal. Individuals, particularly those who already own a car, consider the license plate number to be a relevant factor when deciding to buy a second-hand car in order to guarantee that the new vehicle and the car already owned will not be subject to restriction on the same days.

Table 3.

Evolution of mode share (%) in Bogotá. Trips with purpose work and education Mode 1998 2000 2002 Bus 64 65 50 Transmilenio NA NA 12 Walking 7 12 12 Car 17 12 14 Bicycle 1 3 4 Others 11 8 8 Source: Bogotá Cómo vamos.2013 [21] 80

Cantillo & Ortúzar / DYNA 81 (188), pp. 75-82. December, 2014.

Table 4.

Variation in concentration of airborne particulate matter (μg/m3) in Medellin Station 2004 2005 2006 2007 Aguinaga 105 102 101 126 Guayabal 105 98 101 105 Politécnico 108 111 100 110 UdeA 94 91 93 113 Udn 80 75 79 85 Unalmed 129 118 133 141 Unibol 80 71 80 84 Source: Metropolitan area of Medellín air quality monitoring network [19] Figure 5. Evolution of a) PM10 annual percentage exceeding legal limits; b) PM10 average in Bogotá.

Source: Bogota air quality network 2009. [22].

Table 5.

Variations in concentration of PM10 (μg/m3) in Medellin.

Station 2004 2005 2006 2007 Aguinaga 62 60 58 73 Corantioquia 61 59 58 67 Guayabal 68 63 63 65 Source: Metropolitan area of Medellín air quality monitoring network [19].

positive short-term effect. Emissions decreased in the year following the policy's implementation, but its reduction was inferior to the percentage decrease in automobile stock affected (20%). However, only two years later all the indicators increased and even surpassed the levels prior to the introduction of the measure. This reveals that in the medium- and long-term, the policy's benefits were lost, as the real automobile stock actually increased.

By analysing the traffic flows, Posada et al. [24] concluded that some car drivers initially re-scheduled their trips away from the peak hours as a consequence of the implementation of Pico y Placa in Medellín. However, the policy turned obsolete in only three years as a consequence on the growth in car ownership. They remark that in this case “the cure could be worse than the illness”. In the same line, Gonzalez-Calderon et al [27] show that “píco y placa” is only an inefficient palliative. They suggest the implementation of congestion pricing in Medellin as a strategy to improve the efficiency of the city´s road infrastructure

Figure 6. Evolution of cars registered in Medellin.

Source: Ministerio de Transporte [25]

3.4. Medellin In a recent analysis of the evolution of the Pico y Placa policy, which began in 2005 in Medellin, Sarmiento and Zuleta [23] and Posada et al. [24] show that, as a result of the measure, the city no longer has clearly defined peak hours as before. This is because car users have moved many of their activities to hours during which the restriction does not apply. Furthermore, a before and after analysis of nine of the city's key intersections revealed that traffic flows during rush hours decreased during the first year that the measure was implemented (recall that it affected 20% of vehicles), but only two years later traffic flows were back to pre-measure levels. As a result, the traffic authorities extended the restriction to 40% of the vehicles and 20% of two-stroke engine motorcycles in 2008.

Fig. 6 shows the evolution of registered vehicles in Medellín from 2003 (two years before Pico y Placa was implemented) to 2010; the reference point is 2004, to which we assigned a value of 100 It may be noted that after the implementation of the policy there was an increase in the rate of growth of automobile stock.

Emissions measurements were also taken from 2004 to 2007 at different stations in Medellin [26]. Table 4 shows the annual variation in airborne particulate matter (TSP), while Table 4 shows the same measurements for PM10 particulate matter. In all cases, we can see that the implementation of the vehicular restriction in 2005 had a

4. Discussion and conclusions We have analysed a policy, common to several Latin American cities, that involves prohibiting the circulation of a proportion of private cars based on their license plate numbers. This type of measure has been conceived as a way to reduce two of the most relevant urban transport externalities: congestion and air pollution.

Diverse empirical evidence reveals that the measure involves a reduction in social benefits and that it only has apparently positive short-term effects (reduced travel costs and improved air quality). However, these disappear quite rapidly and the situation tends to get worse. The reason for this is, primarily, that a greater willingness-to-pay for driving among car users who are subject to the circulation prohibition, leads to their purchasing vehicles, both new and used (the latter are usually brought from other localities), that increase the city's automobile stock.

The justification for restricting license plate numbers has 81

Cantillo & Ortúzar / DYNA 81 (188), pp. 75-82. December, 2014.

been centred upon the proposition that there is a general benefit (to society) that takes priority over individual benefits. However, our analysis reveals that, in reality, the approach produces a net loss to society. Worse still, upon discovering that the measure does not work as expected, local politicians and transport planners have tended to apply the prohibition to a larger share of the vehicle fleet and increase the hours during which the circulation restriction applies. We show that, fairly rapidly, these measures do not only cease to be effective, but also tend to exacerbate the negative results that had not been initially considered.

One wonders why this measure is still in operation or being considered. An explanation is the short-term vision of local politicians. Transport authorities become hostage of the policy, fearing that if they disassemble it might lead to immediate demand growth.

The transportation-related externalities in Latin American cities have become increasingly critical.

Therefore, these cities must become bolder in terms of the transport policies they adopt. In Latin America's main cities, significant progress has been made in terms of public transport, zoning laws, and control over public space. So, the conditions are there for moving towards the necessary implementation of road pricing policies, which have had such good results in the developed world.

[11] [12] [13] [14] [15] [16] [17] [18]

[19] [20]

Acknowledgement

[21]

Authors gratefully acknowledge the financial support of the Colombian Science and Technology National Agency and the company Terpel, under the contract number 12155022-7963 CT 755-2011. Support by Colciencias does not constitute endorsement of the views expressed in this paper.

[22] [23]

References [1] [2]

[3] [4] [5] [6] [7] [8] [9] [10]

[24]

Bull, A., Congestión de tránsito: El problema y cómo enfrentarlo.

United Nations, Santiago, Chile, 2003.

Breithaupt, M. and Fjellstrom, K., Transport demand management: towards an integrated approach. Proceedings Regional Workshop on Transport Planning, Demand Management and Air Quality, Manila, Philippines, 2002.

Thomson, J.M., Reflections on the economics of traffic congestion.

Journal of Transport Economics and Policy 32, pp. 93-112, 1988.

Wang, R., Shaping urban transport policies in China: Will copying foreign policies work? Transport Policy 17, pp. 147-152, 2010. http://dx.doi.org/10.1016/j.tranpol.2010.01.001 Lindsey, R. and Verhoef, E., Traffic congestion and congestion pricing, in: Hensher, D.A. and Button ,K.J. eds., Handbook of Transport Systems and Traffic Control, Pergamon, Oxford, pp. 77-105, 2001.

Wang, L., Xu, J., Zheng X. and Qin, P., Will a driving restriction policy reduce car trips? A case study of Beijing, China. Enviroment for Development. Discussion Paper DP 13-11, 2013.

Small, K.A. and Verhoef, E.T., The economics of urban transportation.

London: Routledge, 2007.

Rouwendal, J. and Verhoef, E.T., Basic economic principles of road pricing: From theory to applications. Transport Policy 13, pp. 106-114, 2006. http://dx.doi.org/10.1016/j.tranpol.2005.11.007 Mahendra, A., Vehicle restrictions in four Latin American cities: Is congestion pricing possible? Transport Reviews 28, pp. 105-110, 2008. http://dx.doi.org/10.1080/01441640701458265 Hau, T.D., Congestion charging mechanisms for roads: An evaluation of current practice. Working paper, Transport Division, Infrastructure and Urban Development Department, The World Bank, Washington, D.C., 12-14, 1993.

[25] [26] [27]

Fresard, F., Efecto real de la restricción vehicular en Santiago de Chile.

Proceedings X Congreso Panamericano de Ingeniería de Tránsito y Transporte, Santander, Spain, 1998.

De Grange, L. and Troncoso, R., Impacts of vehicle restrictions on urban transport flows: the case of Santiago, Chile. Transport Policy 18, pp. 862-869, 201.

Davis, L.W., The effect of driving restrictions on air quality in Mexico City. Journal of Political Economy 16, pp. 38-81, 2008. http://dx.doi.org/10.1086/529398 Gallego, F., Montero, J.P. and Salas, C., The effect of transport policies on car use: A bundling model with applications. Energy Economics 40, pp. S85 – S97, 2013. http://dx.doi.org/10.1016/j.eneco.2013.09.018 Eskeland, G., and Feyzioglu, T., Rationing can backfire: The "day without a car" in Mexico City. The World Bank Economic Review 11, pp. 383-408, 1997. http://dx.doi.org/10.1093/wber/11.3.383 Comisión Metropolitana de Transporte y Vialidad, Diagnóstico de las condiciones del transporte y sus implicaciones sobre la calidad del aire en la ZMVM, Cometravi, 1999.

Bocarejo, J.P., Evaluation de l´impact economique des politiques liées à la mobilité: Les cas de Paris, Londres, Bogotá et Santiago. Dr. Thesis, Université Paris-Est, Paris, France, 2008.

Departamento Administrativo Nacional de Estadística DANE, Indicador de seguimiento a la economía, [Online], [date of reference: June 5th of 2013].

Available at https://www.dane.gov.co/index.php/cuentas-economicas/indicador-deseguimiento-a-la-economia-ise.

Dargay, J., Gately, D. and Sommer, M., Vehicle ownership and income growth, worldwide: 1960-2030. Energy Journal 28, pp. 143-170, 2007. http://dx.doi.org/10.5547/ISSN0195-6574-EJ-Vol28-No4-7 Gaitan, M. and Behrentz, E., Evaluación del estado de la calidad del aire en Bogotá, MSc. Tesis, Universidad de los Andes, Bogotá, Colombia, 2009.

Bogotá cómo vamos. Informe de movilidad. [Online], [date of reference: June 8th of 2013].

Available at https://www.bogotacomovamos.org Alcaldía Mayor de Bogotá, Secretaría Distrital de Ambiente. Informe anual de calidad del aire en Bogotá, Red de monitoreo de calidad del aire en Bogotá, Colombia, 2009.

Sarmiento, I. and Zuleta, N., Análisis de la evolución del Pico y Placa en Medellín. Memorias IX Simposio de Ingeniería de Tránsito y Transporte, Villa de Leyva, Colombia, 2009.

Posada, J.J., Farbiarz, V. and González, C., Análisis del Pico y Placa como restricción a la circulación vehicular en Medellín – basado en volúmenes vehiculares. DYNA 78 (165), pp. 112-121, 2011.

Ministerio de Transporte. Estadísticas, [Online], [date of reference: September 6th of 2011].

Available at https://www.mintrasporte.gov.co/documentos.php?id=15.

Area Metropolitana del Valle de Aburra, Monitoreo de la calidad del aire. [Online], [date of reference June 5th of 2011] Available at http://www.metropol.gov.co/aire/contenidos.php?seccion=1.

Gonzalez-Calderon C., Posada, J.J. and Sanchez-Díaz, I., The need for congestion pricing in Medellin: An economic perspective. DYNA 79 (171), pp. 123-131. 2012.

V. Cantillo, received a BSc. Eng in Civil Engineering in 1987 from Universidad del Norte, Barranquilla, Colombia, a MSc in Transport and Traffic Engineering from Universidad del Cauca, Colombia, and a PhD in Engineering Sciences in 2004 from Pontificia Universidad Católica de Chile, Chile. He is Associate Professor of the Department of Civil and Environmental Engineering at Universidad del Norte, Barranquilla, Colombia. His research interests include transport modelling, planning and economics.

J. de D. Ortúzar, received his Civil Eng. Title from Pontificia Universidad Católica de Chile (PUC), Chile in 1971, and then obtained a MSc in Transport Planning and Engineering in 1975 and a PhD in 1980 from Leeds University, United Kingdom. He is Professor of Transport Engineering at PUC, and member of the Millennium Institute in Complex Engineering Systems and the Centre for Urban Sustainable Development (CEDEUS). His research interests include discrete choice modeling, data collection techniques and social evaluation of projects.

82

Active vibration control in building-like structures submitted to earthquakes using multiple positive position feedback and sliding modes Josué Enríquez-Zárate a & Gerardo Silva-Navarro b a

Sección de Mecatrónica, Centro de Investigación y de Estudios Avanzados del I.P.N., México, D.F., México. enriquezz@cinvestav.mx b Sección de Mecatrónica, Centro de Investigación y de Estudios Avanzados del I.P.N., México, D.F., México. gsilva@cinvestav.mx Received: October 2th, 2013. Received in revised form: May 9th, 2014. Accepted: September 28th, 2014.

Abstract This work deals with the structural and dynamic analysis of a building-like structure consisting of a three-story building with one passive/active vibration absorber. The base of the structure is perturbed using a shaker, providing excitation forces and noisy excitations emulating ground transportation, underground railways and earthquakes, quite common in Mexico City. It is considered a realistic seismic record of 8.1Mw occurred at Mexico City, containing some resonant frequencies of the structure. The mechanical structure is modeled using Euler-Lagrange methodology and validated using experimental modal analysis techniques. The active control scheme is synthesized to actively attenuate the noise and vibration system response, caused by noisy excitation forces acting on the base, by employing Multiple Positive Position Feedback and Sliding Mode Control to improve the closed-loop system response and, simultaneously, attenuate three vibration modes. Simulation and experimental results describe the overall system performance.

Keywords: Modal analysis; Active vibration control; Earthquakes; Positive Position Feedback; Sliding mode control.

Control activo de vibraciones en estructuras tipo edificio sometidas a sismos utilizando múltiple retroalimentación positiva de la posición y modos deslizantes Resumen Este trabajo considera el análisis estructural y dinámico de una estructura tipo edificio de tres pisos con un absorbedor de vibraciones pasivo/activo. La base de la estructura se perturba con un generador de vibraciones, proporcionando fuerzas de excitación y ruido emulando transporte terrestre, trenes subterráneos y sismos, comunes en la Ciudad de México. Se considera un registro sísmico real de magnitud 8.1Mw, conteniendo algunas frecuencias resonantes de la estructura. La estructura mecánica se modela con la metodología de Euler-Lagrange y se valida con técnicas de análisis modal experimental. El esquema de control activo se diseña para atenuar la respuesta vibratoria del sistema, ocasionada por las fuerzas armónicas que actúan en la base, empleando múltiple retroalimentación positiva de la posición y control por modos deslizantes para mejorar la robustez del sistema en lazo cerrado y atenuar tres modos de vibración.

Resultados en simulación y experimentales describen el desempeño del sistema completo.

Palabra Clave: Análisis modal; Control activo de vibraciones; Sismo; Retroalimentación Positiva de la Posición; Modos deslizantes.

1. Introducción El estudio sobre el control de las vibraciones mecánicas en el diseño de edificios es un tema de investigación que actualmente es de gran interés sobre todo en las grandes ciudades. En particular, este trabajo de investigación se centra en el análisis de la respuesta de estructuras tipo edificio, que se

encuentran perturbadas por fenómenos sísmicos.

Nuestra investigación se centra en las edificaciones de la Ciudad de México, debido a que por su tipo de suelo, la mayor parte está clasificada como zona lacustre y de transición, teniendo baja resistencia sísmica, y debido a esto las estructuras son más vulnerables y se encuentran afectadas por fenómenos como el ruido, transporte, trenes

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 83-91. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.40097

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

subterráneos y especialmente por sismos [1].

Las investigaciones recientes sobre el control de vibraciones en edificios, se refieren generalmente al uso de los esquemas pasivo, semiactivo y activo [2]. Los esquemas pasivos, conocidos como Tuned Mass Damper (TMD, por sus siglas en inglés), generalmente se encuentran limitados en su respuesta, debido a que se diseñan para la atenuación o minimización de una frecuencia o modo de vibración específico de la estructura. Los esquemas activos, denominados Active Mass Damper (AMD, por sus siglas en inglés), ofrecen un controlador en lazo cerrado para un amplio rango de frecuencias o modos de vibración. Generalmente este diseño agrega un grado de libertad y un actuador al modelo original de la estructura tipo edificio, añadiendo complejidad a la dinámica del sistema. La supresión o atenuación de las frecuencias vibratorias en el sistema se realiza agregando una fuerza de control proporcionada por un actuador electrohidráulico o electromecánico.

En este trabajo se considera una estructura tipo edificio de tres niveles, con excitación armónica y sísmica en la base, la cual se controla con un absorbedor de vibraciones pasivo/activo o híbrido y leyes de control basadas en múltiple retroalimentación positiva de la posición (MPPF, por sus siglas en inglés), en combinación con la técnica de control por modos deslizantes, para mejorar la robustez y minimización la respuesta dinámica del sistema.

Figura 1. Integración del sistema tipo edificio completo.

Fuente: Elaboración propia.

2. Sistema completo Se presenta el diseño y construcción a escala de una estructura tipo edificio de tres pisos sometida a perturbaciones del tipo sísmico. El objetivo consiste en analizar su respuesta dinámica, en simulación numérica y experimental, utilizando esquemas de control pasivo/activo para minimizar su respuesta frecuencial en lazo cerrado.

Específicamente, el sistema es una estructura tipo edificio a escala de tres pisos, diseñada y construida de aleación de aluminio. La altura máxima de la estructura, sin considerar el sistema de absorción de vibraciones es de 950 mm, con una base rectangular de 300 200 mm. La estructura completa se encuentra montada sobre un riel con rodamiento de bolas sin fricción y la base móvil se encuentra conectada directamente a un generador de vibraciones electromecánico modelo ET-139, el cual se utiliza para perturbar al sistema con componentes armónicos de baja frecuencia, obtenidos directamente del registro sísmico del temblor ocurrido en la Ciudad de México en 1985, con epicentro en la costa de Zihuatanejo, Michoacán [3]. El generador de vibraciones se controla por un amplificador de potencia Labworks®, modelo PA-138. Para propósitos de control se utiliza como variable de control en lazo cerrado la posición del sistema. Se utiliza un sistema de adquisición de datos NI-CompactDAQ de National Instruments®, chasis modelo NI-DAQ-9172 y módulo de acelerómetros modelo 9133 conectado vía USB a una computadora, para obtener la señal de un acelerómetro colocado en un piso de la estructura, que mide directamente la aceleración y procesa a través de las plataformas de Labview® y Matlab/Simulink® para calcular los desplazamientos (ver Fig. 1).

Figura. 2. Estructura experimental.

Fuente: Elaboración propia.

Por último, se considera que el actuador en el absorbedor pasivo/activo de vibraciones es un motor de cd, acoplado directamente a un sistema masa-resorte con una base móvil, soportada sobre una guía de bolas montada sobre el tercer piso de la estructura. El desplazamiento de la masa del absorbedor se mide con un decodificador óptico en la flecha del motor, señal que se emplea para determinar el valor de la fuerza requerida por la acción de control, por medio de la variación en su valor en corriente. Para el esquema de control retroalimentado se utiliza una tarjeta de adquisición de datos Sensoray® modelo 626. En la Fig. 1 se describe un diagrama esquemático de la integración del sistema completo y en la Fig. 2 se muestra el prototipo de la estructura experimental.

La estructura tipo edificio consta de tres masas ଵ ଶ ଷ interconectadas cada una por cuatro columnas flexibles en cada masa, expresado por medio de rigideces equivalentes ଵ ଶ ଷ y se considera amortiguamiento viscoso (lineal) equivalente ଵ ଶ ଷ sobre cada grado de libertad (ver Fig. 3) [4].

84

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

Canal − 1 (V) Canal − 2 (N90E) Canal − 3 (N00E)

2

Aceleración [m/s ]

0.25 0.2 0.15 0.1 0.05 0 0

5

10

15

20 25 30 Frecuencia [Hz]

35

40

45

50

Figura 5. Espectro en frecuencia del registro sísmico de 1985 en la Ciudad de México.

Fuente: Elaboración propia.

mediciones de la aceleración se obtienen usando un acelerómetro conectado sobre la masa ଷ .

El sistema primario y el TMD están formados por elementos mecánicos . Las ௜ ௜ ௜ , vibraciones no deseadas que causan el desplazamiento de la estructura se generan por el desplazamiento en términos de la aceleración .

2

Aceleración [m/s ]

Figura 3. Modelo esquemático de la estructura tipo edificio.

Fuente: Elaboración propia.

3. Registro sísmico de 1985

Canal − 1 (V)

0.2 0.1

La señal perturbadora se representa por un registro sísmico real ocurrido en la Ciudad de México el 19 de septiembre de 1985 en hora epicentro (GMT): 13:17:42,6.

[3,5]. Esta señal sísmica se representa por tres componentes (ver Figs. 4a,b,c), es decir, se tienen dos en el plano horizontal (llamadas longitudinal y transversal) y otra en sentido vertical [6,7]. La respuesta espectral de las tres componentes se observa en la Fig. 5.

La respuesta en frecuencia corresponde a señales de baja frecuencia, las cuales tienen un fuerte impacto sobre estructuras (e.g., edificios, casas) de gran altura o de varios pisos y de gran masa, debido a que tienen períodos naturales de vibración largos [1]. Los sismos de mayor magnitud tienen mayor duración (cerca o superior a un minuto) y liberan más energía a bajas frecuencias. Las frecuencias altas muestran una mayor tendencia a atenuarse por encima de los 5 o 10 Hz, dependiendo de la distancia a la fuente sísmica y las condiciones geológicas del sitio. Las frecuencias más altas se registran en sitios de roca y en las cercanías de la fuente sísmica [8]. La estructura experimental es de baja altura y poca masa, y sus modos de vibración difícilmente son excitados por la señal sísmica real (Fig. 4). Por lo tanto, es conveniente escalar la componente de amplitud del registro sísmico real por una constante de veinte veces su valor, esto con la finalidad de excitar los primeros modos de vibración de la estructura, en el rango de frecuencias del registro sísmico real.

0 −0.1 −0.2 0

20

40

60

80 100 Tiempo [s]

120

140

160

180

Figure 4a. Componente vertical del registro sísmico.

Fuente: Elaboración propia.

Canal − 2 (N90E)

2

Aceleración [m/s ]

0.4 0.2 0 −0.2 −0.4 0

20

40

60

80 100 Tiempo [s]

120

140

160

180

Figure 4b. Primera componente horizontal del registro sísmico. Fuente: propia.

Canal − 3 (N00E)

2

Aceleración [m/s ]

0.4 0.2 0 −0.2 −0.4 0

20

40

60

80 100 Tiempo [s]

120

140

160

180

Figure 4c. Segunda componente horizontal del registro sísmico.

Fuente: Elaboración propia.

4. Estructura tipo edificio de tres niveles La fuerza de control se describe como , que afecta directamente a un TMD, a través de ସ , para lograr los objetivos de control deseados de atenuación y/o cancelación de vibraciones. La base de la estructura está perturbada por un desplazamiento en la base producida por una fuerza lateral de un generador de vibraciones electromagnético conectado a la estructura. Las

Un modelo matemático simplificado para un edificio de tres pisos sometido a movimiento por aceleración en la base se obtiene como ଷ

85

ଷ

ଷ

ଷ ଷ

(1)

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014. ் ଷ donde es el vector de coordenadas ଵ ଶ ଷ generalizadas de desplazamientos relativos con respecto al marco de referencia principal, es la aceleración de la señal sísmica en la base de la estructura y ଷ ଷ ଷ son matrices de masa, amortiguamiento y rigidez de ் ଷ respectivamente. El vector ଷ es el vector de influencia, que representa el desplazamiento de cada masa debido al desplazamiento del suelo. Las matrices de masa, rigidez y amortiguamiento para el edificio de tres pisos son

Tabla 1.

Parámetros del sistema. ଵ

ଵ

ଷ ଵ

ଶ ଶ

ଶ ଶ

ଷ ଷ

ଵ

ଶ ଶ

ଷ

ଷ ଷ

ଶ ଶ

ଷ ଷ

ଷ ଷ

ଷ ଶ

଴

௝

ఠ೔ ାఠೕ

଴

଴ ଷ,

ଷ

, con

௜

y

௝

donde

ଶఠ೔ ൈఠೕ ଴

௜

ఠ೔ ାఠೕ

En la estructura tipo edificio mostrada en la Fig. 3 el TMD se representa en términos de los parámetros ସ ସ ସ y se diseña para atenuar una de las frecuencias de resonancia más críticas en el sistema. En este caso el efecto del desplazamiento en la base se logra atenuar en la respuesta del sistema primario, para un rango pequeño de frecuencias de excitación y bajo condiciones de operación estables [2].

La principal desventaja de estos esquemas de control en lazo abierto es la ausencia o poca robustez con respecto a incertidumbres o parámetros desconocidos o variaciones en las frecuencias de excitación.

El edificio de tres pisos con un TMD sobre la masa del tercer piso de la estructura, sometida a un movimiento sísmico , de manera que el TMD se diseña para compensar el segundo modo de vibración. Las ecuaciones de movimiento del sistema de cuatro grados de libertad son

la estructura y ௜ y ௝ la proporción de amortiguamiento en la estructura para los modos y respectivamente.

La estructura tipo edifico de tres pisos se perturba con una señal sísmica en su componente vertical (ver Figs. 4 y 5), reproducida con un generador de vibraciones electromagnético para amplitud y frecuencia variable en un tiempo de 179 s [3].

La función de respuesta frecuencial (FRF) experimental en la estructura se obtiene mediante la transformada rápida de Fourier (FFT, por sus siglas en inglés) y la aplicación de técnicas de análisis modal experimental como Peak Peaking (ver Fig. 6).

Los parámetros del sistema sin TMD ni control activo se presentan en la Tabla 1.

FRF experimental Peak Peaking

1

2

Aceleración [m/s ]

ସ

5

3 2

2 3

1 2

4

6

8 10 12 Frecuencia (Hz)

14

16

ସ

ସ

ସ ସ

(3)

் ସ donde es el vector de coordenadas ଵ ଶ ଷ ସ generalizadas de desplazamientos relativos con respecto al marco de referencia principal, es la aceleración de la señal sísmica en la base de la estructura y ସ ସ ସ son las matrices de masa, rigidez y amortiguamiento de , respectivamente. ் ସ En este caso ସ es el vector de influencia, que representa el acoplamiento inercial entre los pisos de la estructura y el desplazamiento del suelo. Las matrices de masa, amortiguamiento y rigidez se expresan como

4

0

ଷ

5. Estructura tipo edificio con TMD

y

las frecuencias resonantes de

6

ଶ

Tabla 2.

Parámetros modales de la estructura tipo edificio de tres pisos.

Modo Varia- Amortiguamiento Frecuencia natural ௜ [Hz] ción modal experimental (%) Numérico Experimental ௜ 1 1.7020 1.4343 0.0044 15.72 2 5.1607 4.9703 -3.69 0.0011 3 8.0274 8.9493 11.48 0.000446 Fuente: Elaboración propia.

La matriz de masas ଷ se determina a partir de las masas ( ଵ ଶ ଷ ) por cada piso del edificio. Los valores de la matriz de rigidez ଷ se obtienen considerando las dimensiones y propiedades mecánicas del material en cada piso del edificio.

Para propósitos de análisis modal se considera amortiguamiento proporcional o de Rayleigh [9,10], es decir, que

ଷ

Los parámetros de amortiguamiento se obtuvieron indirectamente, a partir de la FRF, considerando el amortiguamiento modal ௜ y la frecuencia modal de la estructura obtenidos experimentalmente.

El ௜ amortiguamiento modal es aproximado por medio de los métodos de Peak Peaking y Curve Fitting.

Una comparación de las frecuencias de resonancia obtenidos en forma experimental y numéricamente, se presentan en la Tabla 2, donde los resultados son muy cercanos para validar el modelo simplificado de la estructura tipo edificio de tres pisos en los primeros tres modos de vibración.

ଶ

ଷ

ଷ

ଶ

Fuente: Elaboración propia.

(2)

ଵ ଷ

ଶ

ଵ

18

Figura 6. FRF experimental de la estructura de tres pisos.

Fuente: Elaboración propia.

86

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

Tabla 3.

Parámetros del sistema.

7

FRF experimental Peak Peaking

1

ଶ

ଵ

ଷ

ଶ

ଵ

ସ

ଷ

ଶ

2

ଵ

Aceleración [m/s ]

6

ସ

ଷ

ସ

3 2

2

0

4 3

2

4

6

8 10 12 Frecuencia [Hz]

14

16

18

Figura 7. FRF experimental del edificio de tres pisos con TMD. Fuente: Elaboración propia.

ଵ ଶ

ସ

ଷ

Tabla 4.

Parámetros modales con TMD. Frecuencia natural Modo ௜ [Hz] Numérica Experimental 1 1.4745 1.4153 2 4.5737 4.7341 3 6.1259 7.3509 4 8.0802 9.2163 Fuente: Elaboración propia.

ସ ଵ

ଶ ଶ

ଶ ଶ

ଷ ଷ

ଷ ଷ

ସ

ଶ ଶ

(4)

ସ

ସ ଵ ସ

4

1

Fuente: Elaboración propia.

ସ

5

ସ

ଶ ଶ

ଷ ଷ

ଷ ଷ

ସ ସ

ସ

Variación Amortigua-miento (%) modal experimental ௜ -4.01 0.0046 3.5 0.0016 19.99 0.000949 14.06 0.000620

ସ

Para propósitos de análisis modal se propone que el sistema posee amortiguamiento de tipo proporcional, de ଶఠ೔ ൈఠೕ manera que ସ , donde ଴ y ଴ ସ ଴ ସ ௜ ఠ ାఠ ଶ con ௜ y ௝ , las frecuencias modales೔ deೕ la ଴ ௝ ఠ ାఠ estructura y ೔ ௜ yೕ ௝ la proporción de amortiguamiento en la estructura para los modos y respectivamente.

El TMD se diseña para atenuar pasivamente la -ésima ௞ర frecuencia a partir de la expresión ௝ donde ௝ es la ௠ర -th frecuencia de resonancia de la estructura que será atenuada y ସ y ସ son la masa y la rigidez equivalente del TMD. En particular, el TMD se diseña para atenuar el segundo modo de vibración experimental, localizado en Hz. ଶ Los parámetros del edificio de tres pisos con TMD se proporcionan en la Tabla 3.

Los parámetros de amortiguamiento se obtuvieron indirectamente, a partir de la FRF, considerando el amortiguamiento modal ௜ y la frecuencia modal de la estructura obtenidos experimentalmente.

El ௜ amortiguamiento modal se estima con los métodos de Peak Peaking y Curve Fitting. La estructura tipo edificio de tres pisos con TMD se caracteriza aplicando un movimiento armónico en el suelo, utilizando la componente vertical del registro sísmico, que tiene una amplitud variable aproximadamente entre 0 y 0.20 m/s2 y frecuencias de excitación de 0 a 50 Hz (ver Fig. 5).

La FRF experimental se describe en la Fig. 7. Los resultados del análisis modal experimental se resumen en la Tabla 4, donde se puede apreciar un aproximamiento razonable con respecto a los resultados numéricos.

La respuesta dinámica experimental (desplazamiento y aceleración) en el tercer piso de la estructura tipo edificio con TMD, cuando se somete a movimiento sísmico en el suelo se muestra en las Figs. 8a,b,c.

Aceleracion en el tercer piso [m/s2]

Figure 8a. Desplazamiento experimental de la estructura con TMD. Fuente: Elaboración propia.

6

Lazo abierto TMD Experimental

4 2 0 -2 -4

0

20

40

60

80 100 Tiempo [s]

120

140

160

180

Figure 8b. Aceleración experimental de la estructura con TMD. Fuente: Elaboración propia.

La respuesta del sistema cerca del segundo modo de vibración se atenúa en 39.78%. La FRF en simulación del tercer piso considerando el TMD se presenta en la Fig. 9. En este caso los últimos modos se reducen.

Dado que el TMD está diseñado para atenuar una frecuencia de resonancia específica se considera que esta solución no es lo suficientemente robusta para atenuar en 87

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

Aceleración [m/s2]

8

Considere el edifico de tres pisos con TMD colocado en el tercer piso, la aceleración del suelo o señal sísmica y una fuerza de control activa que actúa sobre el TMD, que se representa como

TMD Experimental Lazo abierto

6

4

ସ

2

4

6

8 10 12 Frecuencia (Hz)

14

16

18

20

Figure 8c. FRF experimental con TMD en el 3er piso.

Fuente: Elaboración propia.

Aceleración [m/s2]

8

Lazo abierto TMD

6

4

2

0 0

1

2

3

4 5 6 Frecuencia (Hz)

7

8

9

ସ

ସ ସ

(5)

௙

் ସ donde es el vector de ଵ ଶ ଷ ସ coordenadas generalizadas, es la aceleración del suelo generado por la señal sísmica en la base del sistema y ସ ସ ସ son las matrices de masas, amortiguamiento y ் rigidez de 4 4, respectivamente. Aquí ସ ସ es el vector de influencia que representa el acoplamiento inercial entre los pisos y el movimiento del suelo. El control de fuerza afecta directamente a la masa ସ del TMD ் y ௙ representa una matriz de entrada para el sistema completo.

Note que el sistema (5) es completamente controlable a partir de la fuerza de control y observable a partir del desplazamiento del TMD ସ .

En nuestro caso, la estructura mecánica tiene únicamente un sensor de movimiento (aceleración) en el TMD ( ସ ) y una fuerza de control que actúa directamente sobre su masa ( ସ ). El esquema de control PPF para la estructura tipo edificio de cuatro grados de libertad (5) resulta en el sistema en lazo cerrado

2

0 0

ସ

10

Figura 9. FRF en simulación en el 3er piso con TMD. Fuente: Elaboración propia.

ସ

forma simultánea varios modos de vibración en presencia de vibraciones con variación en las frecuencias de excitación y amplitud, como las frecuencias de una señal sísmica.

Involucrando otros efectos no deseables como roll-off. Es por tal motivo, que aplicamos esquemas de control pasivo/activo de vibraciones con técnicas de retroalimentación positiva de posición.

ସ

ସ

௙

ସ ସ ଶ ௙

௙

(6)

௙

ଶ ் ௙ ௙

(7)

ଶ ௙

(8)

ସ donde y es el movimiento del suelo. El absorbedor pasivo virtual , con relación de amortiguamiento y frecuencia natural se ௙ ௙, ଶ ் retroalimenta con el sistema primario mediante ௙ ௙ . La ley de control PPF se representa por la fuerza de control , donde es una ganancia de control.

En forma compacta, el sistema en lazo cerrado se describe como

6. Estructura tipo edificio con control pasivo/activo PPF El esquema de control por retroalimentación positiva de posición (PPF, por sus siglas en inglés) se utiliza para compensar la respuesta del sistema en presencia de la perturbación generada por la señal sísmica . Este controlador es ampliamente conocido en la literatura como un método de control modal para atenuación de vibraciones [9,11,12,14,15].

Este esquema de control agrega un grado de libertad adicional al sistema mecánico original, considerado como un absorbedor pasivo virtual o como un filtro de segundo orden. Los parámetros de este controlador se pueden obtener utilizando datos experimentales, lo que hace que la técnica PPF sea muy utilizada entre otro tipo de aplicaciones estructurales y de control [11]. El término de retroalimentación positiva de la posición significa que la coordenada de posición (desplazamiento) del sistema primario se retroalimenta positivamente al filtro y la coordenada de posición (desplazamiento) del compensador (sistema secundario) se retroalimenta positivamente al sistema primario [11].

ସ

ସ ௙ ସ ଶ ் ௙ ௙

௙ ଶ ௙

௙ ଶ ௙

(9) ସ ସ

Observe que la matriz de masa ସ es simétrica y definida positiva, por lo tanto, la matriz de masa completa en (9) es también simétrica y definida positiva. La matriz de amortiguamiento proporcional ସ es simétrica y definida positiva, y de igual forma, la matriz de amortiguamiento completa tiene las mismas propiedades. Sin embargo, la matriz de rigidez ସ es simétrica y definida positiva, y para garantizar la estabilidad asintótica en lazo cerrado es suficiente el seleccionar constantes apropiadas y ௙ en el control PPF, de tal forma que la matriz de rigidez en lazo cerrado sea simétrica y definida positiva [9,12]. Para 88

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

mostrar esto considere que ் ்

ଶ

ଶ ௙

௙

் ଶ

் ଵ

ଶ ் ௙ ௙

ଶ ் ௙ ௙ ௙

Para mejorar el desempeño de la respuesta vibratoria del esquema de control activo/pasivo usando TMD y control PPF, proponemos agregar múltiples controladores PPF para obtener mayor robustez para un amplio rango de frecuencias.

ଵ

ଶ ௙

ଶ

ଵ ଶ ௙

்

் ௙ ଵ

் ௙ ଵ

ଶ

7. Estructura tipo edificio con un control MPPF y TMD

ଶ

El control basado en múltiple absorbedores pasivos virtuales (MPPF) es una extensión del control PPF. El propósito de este esquema es agregar ௣ absorbedores pasivos virtuales conectados en paralelo hacia el sistema primario, para atenuar las vibraciones [9,16].

Considere el edificio de tres pisos con TMD sobre el tercer piso, la aceleración del suelo y una fuerza de control activa operando sobre el TMD. El diseño de un control MPPF para una estructura tipo edificio de cuatro grados de libertad (5) resulta en un sistema en lazo cerrado de la forma

siendo el segundo término siempre no negativo y para asegurar la estabilidad asintótica en lazo cerrado, la matriz de rigidez debe ser definida positiva, esto se obtiene seleccionando y ௙ de tal forma que la matriz ் ଶ ଶ sea definida positiva [9]. ௙ ௙ ௙ Las Figs. 10a,b,c describen los resultados numéricos aplicando el esquema de control activo/pasivo de vibraciones (6)-(8). Los parámetros del control PPF son y . La respuesta completa del ௙ ଵ, ௙ sistema se atenúa en alrededor del 49% con respecto a la dinámica de la estructura en lazo abierto, con pequeños esfuerzos de control.

ସ

ସ

Lazo abierto Control PPF + TMD

5

்

(11) (12)

4

ସ donde y es el movimiento en la base, ௡೛ son coordenadas de ௣ absorbedores pasivos virtuales, con matrices diagonales de relaciones de amortiguamiento y frecuencias naturales . En este caso ௙ଵ ௙௡೛ . La ley de control MPPF está representada ௡೛ por la fuerza de control , donde es una matriz diagonal de ganancias de control.

En forma compacta el sistema en lazo cerrado es expresado de la siguiente forma

3 2

0 0

1

2

3

4 5 6 Frecuencia (Hz)

7

8

9

10

Figure 10a. FRF con TMD y control PPF. Fuente: Elaboración propia.

ସ 6

ସ

Lazo abierto Control PPF + TMD

2

Aceleracion en el tercer piso [m/s ]

ଶ ଶ

1

4

ସ ଶ

(13)

ଶ ்

ସ ସ

ଶ

2

Note que la matriz ସ es simétrica y definida positiva, además es la matriz identidad y, entonces, la matriz de masa completa en (13) es también simétrica y definida positiva. La matriz ସ es simétrica y definida positiva, luego la matriz de amortiguamiento completo es también simétrica y definida positiva. La matriz de rigidez ସ es simétrica y definida positiva, y para garantizar la estabilidad asintótica en lazo cerrado resulta suficiente seleccionar constantes adecuadas ௜ y ௙ ௜ , de tal forma que la matriz de rigidez en lazo cerrado sea simétrica y definida positiva [9]. Para ் , con la forma cuadrática mostrarlo considere

0 −2 −4 −6 0

20

40

60

80 100 Tiempo [s]

120

140

160

180

Figure 10b. Respuesta dinámica de la estructura.

Fuente: Elaboración propia.

Esfuerzo de control [N]

(10)

ସ ସ ଶ

6

Aceleración [m/s2]

ସ

4 2 0

்

்

ଶ

்

ଶ

−2

் −4 0

20

40

60

80 100 Tiempo [s]

120

140

160

ଶ ଶ

180

Figure 10c. Esfuerzo de control usando PPF. Fuente: Elaboración propia.

்

ଶ

்

் ்

ଶ

(14)

்

donde el segundo término es siempre no negativo y con 89

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

esto, para asegurar la estabilidad asintótica en lazo cerrado, la matriz de rigidez en (14) debe ser definida positiva, que se obtiene seleccionando las matrices y de tal ଶ ଶ ் forma que sea definida positiva [3].

Las Figs. 11a,b,c, muestran los resultados numéricos del esquema de control pasivo/activo (9)-(11). Los parámetros del control MPPF son ௙ଵ , ଵ , ଵ, ௙ଵ , ଶ y ௙ଷ , ଷ ௙ଶ ଷ , ௙ଶ ସ , ௙ଷ La respuesta del sistema se atenúa aproximadamente en 54% con respecto a la respuesta en lazo abierto, empleando esfuerzos de control muy pequeños.

Para optimizar el desempeño del control MPPF, proponemos agregar un esquema de control PPF combinado con una componente de control basada en modos deslizantes para obtener una dinámica más robusta.

importante de los controladores por modos deslizantes es su alta robustez, ante la presencia de incertidumbres no modeladas, como son las perturbaciones externas [13].

Para compensar en forma simultánea varios modos de vibración, un control PPF con modos deslizantes se expresa como ௙

ଶ ௙

௙

் ௙

் ௙

ଵ ଶ ௙

(15)

ଶ ் ௙ ௙

(16)

் ௙

ଶ

(17)

௦௠

donde es la superficie de conmutación, establecida en ் términos del error virtual ௙ , con los parámetros positivos ଵ y ଶ y es la función signum, que para propósitos prácticos se aproxima por la función continua

8. Estructura tipo edificio con un control TMD y PPF combinado con modos deslizantes La técnica de control por modos deslizantes es un esquema de control robusto que puede utilizarse tanto para sistemas lineales como no lineales. Una característica 6

Lazo abierto Control MPPF + TMD

5 Aceleración [m/s2]

suficientemente pequeño, con un parámetro positivo dependiendo de las limitaciones físicas y ancho de banda del motor de cd.

En las Figs. 12a,b,c se ilustran los resultados numéricos, aplicando el esquema de control pasivo/activo de vibraciones (15)-(17). Los parámetros del control PPF con modos deslizantes son ௙ଵ , ଵ , ଵ , ௙ଵ , ଵ y ଶ y ௦௠ . Se observa la buena atenuación en la respuesta frecuencial, aunque en este caso el esfuerzo de control es mayor, debido al término del control por modos deslizantes.

4 3 2 1 0 0

1

2

3

4 5 6 Frecuencia (Hz)

7

8

9

10

Figure 11a. FRF con TMD y control MPPF. Fuente: Elaboración propia.

6

Lazo abierto Control Modos Deslizantes + PPF + TMD

2

Aceleracción [m/s ]

5

Aceleración del tercer piso [m/s2]

6

Lazo abierto Control MPPF + TMD

4 2

3 2 1

0

0 0

−2

20

40

60

80 100 Tiempo [s]

120

140

160

2

3

4 5 6 Frecuencia (Hz)

7

8

9

10

180

Aceleracion en el tercer piso [m/s2]

Figure 11b. Respuesta dinámica de la estructura.

Fuente: Elaboración propia.

20 10 0 −10 −20 0

1

Figure 12a. FRF con TMD y control PPF con Modos deslizantes.

Fuente: Elaboración propia.

−4 −6 0

Esfuerzo de control [N]

4

20

40

60

80 100 Tiempo [s]

120

140

160

6

2 0 -2 -4 -6

180

Figure 11c. Esfuerzo de control usando MPPF. Fuente: Elaboración propia.

Lazo abierto Control Modos Deslizantes + PPF + TMD

4

0

20

40

60

80 100 Tiempo [s]

120

140

Figure 12b. Respuesta dinámica de la estructura.

Fuente: Elaboración propia.

90

160

180

Esfuerzo de control [N]

Enríquez-Zárate & Silva-Navarro / DYNA 81 (188), pp. 83-91. December, 2014.

[10] Gawronski, W.K., Advanced structural dynamics and active control of structures. NY: Springer-Verlag, 2002.

[11] Inman, D.J., Tarazaga, P.A. and Salehian, A., Active and passive damping of structures, Proc. International Congress on Sound and Vibration ICSV13, pp. 1-8, 2006.

[12] Friswell, M.I. and Inman, D.J. The relationship between positive position feedback and output feedback controllers. Smart Materials and Structures, 8, pp.

285-291, 1999. http://dx.doi.org/10.1088/0964-1726/8/3/301 [13] Utkin, V.I., Sliding modes in control and optimization. Berlin: Springer-Verlag, 1992. http://dx.doi.org/10.1007/978-3-642-843792 [14] Baz, A. and Poh, S., Optimal vibration control with modal positive position feedback. Optim. Control Appl. Meth., 17 (2), pp. 141-149, 1996. http://dx.doi.org/10.1002/(SICI)10991514(199604/06)17:2<141::AID-OCA566>3.0.CO;2-D [15] Baz, A. and Hong, J.T., Adaptive control of flexible structures using modal positive position feedback. International Journal of Adaptive Control and Signal Processing, 11 (3), pp. 231-253, 1997. http://dx.doi.org/10.1002/(SICI)1099-1115(199705)11:3<231::AIDACS435>3.0.CO;2-8 [16] Moon K.K. and Seok H., Active vibration control of smart grid structure by multiinput and multioutput positive position feedback controller. Journal of Sound and Vibration, 304 (1-2), pp. 230-245, 2007. http://dx.doi.org/10.1016/j.jsv.2007.02.021

2 1 0 −1 −2 0

20

40

60

80 100 Tiempo [s]

120

140

160

180

Figure 12c. Esfuerzo de control PPF con Modos deslizantes.

Fuente: Elaboración propia.

En general, el controlador PPF combinado con modos deslizantes resulta más sencillo de implementar que el controlador MPPF, produciendo una mejor respuesta y empleando esfuerzos de control ligeramente mayores, pero añadiendo robustez ante las perturbaciones exógenas desconocidas, como realmente ocurre en los sismos.

9. Conclusiones Se utiliza un absorbedor pasivo/activo con esquemas de control por retroalimentación positiva de la posición y modos deslizantes para una estructura tipo edificio de tres pisos, perturbada en su base por un registro sísmico de magnitud 8.1 Mw, que afectó a la Ciudad de México en 1985. El desempeño del sistema completo, en lazo abierto y lazo cerrado, se valida a través de resultados en simulación y experimentales. Los resultados son satisfactorios, debido a que los esquemas de control atenúan altamente la respuesta del sistema en presencia de la señal excitadora, utilizando pequeños esfuerzos de control y recurriendo únicamente a la retroalimentación de la posición del TMD.

J. Enríquez-Zárate, es Ing. en Cibernética de la Universidad del Sol, Cuernavaca, México, con grado de MSc en Ing. Eléctrica - Mecatrónica, y de Dr en Ing. en Diseño Mecánico, ambos obtenidos en la UNAM, Mexico D.F, México. Actualmente realiza una estancia Posdoctoral en la Sección de Mecatrónica en el CINVESTAV-IPN, México, con intereses de investigación en absorbedores de vibraciones activo/pasivo y semi-activo en estructuras mecánicas (edificios, plataformas marinas, eólicos), diseño mecánico y control de sistemas no lineales.

G. Silva-Navarro, es Ing. Mecánico (Diseño Mecánico e Ingeniería Térmica) con grado de MSc. en Inge. Eléctrica (Control) ambos obtenidos en el Instituto Tecnológico de la Laguna, Torreón, México, y con grado de Dr. en Ing. Eléctrica (Control Automático) obtenido en el CINVESTAVIPN, México. Actualmente es investigador titular en la Sección de Mecatrónica en el CINVESTAV-IPN. México, con intereses de investigación en mecatrónica, absorbedores de vibraciones activo/pasivo para sistemas mecánicos, estructuras inteligentes, máquinas rotativas, diseño mecánico y control de sistemas no lineales.

Referencias [1] [2] [3] [4] [5]

[6]

[7]

[8] [9]

Meli, R. y Bazán, E., Diseño sísmico de edificios. México: Editorial LIMUSA, 2013.

Gómez, G., Marulanda, J. and Thomson, P., Control systems for dynamic loading protection of civil structures. DYNA, 75 (155), pp.

77-89, 2008.

Centro de Instrumentación y Registro Sísmico, A.C. Datos del registro sísmico del temblor de la Ciudad de México en 1985.

[Online] Disponible en: www.cires.org.mx Rios-Gutiérrez, M. and Silva-Navarro, G., Active vibration control in building-like structures using piezoelectric actuators and positive acceleration feedback. DYNA, 80 (179), pp. 116-125, 2013.

Auvinet G. y Mendoza M.J., Comportamiento de diversos tipos de cimentación en la zona lacustre de la Ciudad de México durante el sismo del 19 de septiembre de 1985. Proc. Symposium: Los Sismos de 1985; Casos de Mecánica de Suelos, 1986.

Schmidt-Díaz, V. y Quirós-Serrano, C. Caracterización de los registros acelerográficos obtenidos en el laboratorio de ingeniería sísmica de la Universidad de Costa Rica. Ingeniería, 17 (1), pp. 2741, 2007.

Kappos, A.J. and Anastasios, G.S., Protection of buildings from earthquake-induced vibration, en Crocker M.J. In: Handbook of Noise and Vibration Control, Wiley, 2007.pp. 1393-1403. http://dx.doi.org/10.1002/9780470209707.ch117 Santana, G., Sismo de Cóbano 25 de marzo de 1990 efectos sobre suelos y edificaciones, Costa Rica: Univ. de Costa Rica, Inst. Inv. en Ingeniería, Lab. de Ing. Sísmica, 1990.

Cabrera-Amado, A. and Silva-Navarro, G., Semiactive vibration absorption in a rotor-bearing system using a PPF control scheme, Proc. International Conference on Noise and Vibration Engineering ISMA2012+USD2012, pp. 209–221, 2012.

91

Analysis of customer satisfaction using surveys with open questions José Amelio Medina-Merodio a, Carmen de Pablos-Heredero b, María Lourdes Jiménez-Rodríguez c, Luis de Marcos-Ortega d, Roberto Barchino-Plata e, Daniel Rodríguez-García f & Daniel Gómez-Aguado g b

a Escuela Politécnica Superior. Universidad de Alcalá. Alcalá, España, josea.medina@uah.es Facultad de Ciencias Jurídicas y Sociales. Universidad Rey Juan Carlos. Madrid, España, carmen.depablos@urjc.es c Escuela Politécnica Superior. Universidad de Alcalá. Alcalá, España, lou.jimenez@uah.es d Escuela Politécnica Superior. Universidad de Alcalá. Alcalá, España, luis.demarcos@uah.es e Escuela Politécnica Superior. Universidad de Alcalá. Alcalá, España, roberto.barchino@uah.es f Escuela Politécnica Superior. Universidad de Alcalá. Alcalá, España, daniel.rodriguezg@uah.es g Escuela Politécnica. Universidad de Alcalá. Alcalá, España, daniel.gomeza@edu.uah.es

Received: October 7th, 2013. Received in revised form: March 3th, 2014. Accepted: October 27th, 2014.

Abstract In this paper the use of open-ended questionnaires to improve the evaluation of customer satisfaction according to ISO 9001 in small and medium-sized enterprises is analyzed. By obtaining more information in comparison to the closed questions questionnaire some limitations coming from the second one are removed. The open-ended questionnaire is analyzed by applying a semantic study to obtain the root of each word and remove the word that is not relevant for the information needs of the organization. This way the positive or negative trend for each response is identified. This study proofs that the use of open-ended questionnaires facilitates the fulfilment of the ISO 9001 standard. It allows the comparison between the data coming from the Customer Relationship Management System (CRM) and the data obtained through the questionnaire. Furthermore it opens new areas of research based in the use of semantic analysis in quality systems and marketing.

Keywords: Customer’s satisfaction, ISO 9001, semantic analysis, Lucene.

Análisis de la satisfacción de cliente mediante el uso de cuestionarios con preguntas abiertas Resumen En este trabajo se analiza, cómo el uso de cuestionarios de preguntas abiertas permite a las pequeñas y medianas empresas, mejorar la evaluación del grado de satisfacción de clientes según la norma ISO 9001. Al obtener mayor información que con los cuestionarios de preguntas cerradas, se eliminan las limitaciones de estos últimos. Para conseguir este objetivo, se han analizado las preguntas abiertas mediante su estudio semántico, obteniendo previamente la raíz de cada palabra y eliminando las que no aportan información, detectando la tendencia positiva y negativa de cada una de las respuestas. Este estudio prueba que el uso de cuestionarios de preguntas abiertas, facilita cumplir con la norma ISO 9001 y permite su comparación con los datos del sistema de gestión de las relaciones con los clientes (CRM – del ingles Customer Relationship Management). Además abre nuevas líneas de investigación de la semántica en los sistemas de calidad y de marketing.

Palabras clave: Satisfacción de cliente; ISO 9001; análisis semántico; Lucene.

1. Introducción En un mundo cada vez más globalizado [1] conocer los gustos y las necesidades de los clientes genera una importante ventaja competitiva [2] al permitir descubrir nuevos clientes y fidelizarlos.

Con el fin de obtener la ventaja competitiva muchas

empresas han optado por certificar sus organizaciones con la norma ISO 9001 [8] como se desprende de los últimos estudios presentados por ISO sobre la evolución de certificados de calidad [9].

El objetivo, de este trabajo se centra en analizar en qué medida el uso de cuestionarios de preguntas abiertas permite mejorar la evaluación del grado de satisfacción de

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 92-99. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.40144

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

creciente número de páginas que permiten realizar y publicar estos cuestionarios [17].

Es habitual que estos cuestionarios o encuestas de satisfacción se realicen mediante preguntas cerradas [10,13,16] permitiendo al usuario seleccionar la respuesta entre un número finito de posibles respuestas. La cantidad de soluciones de cada pregunta va a estar en función del nivel de escala Likert que seleccionemos y esta puede ser de 5, 7 y 10 niveles [15]. Lo que significa que si tomamos escala de nivel 5 la pregunta tendrá 5 posibles respuestas.

Además el tratamiento de estos cuestionarios es mucho mas sencillo, no provoca complicaciones y el coste de la relación información/precio frente a otras herramientas similares es mucho mejor.

Por otro lado, el uso de este tipo de evaluaciones conlleva distintos problemas, los más conocidos son la contestación automática al cuestionario marcado todos 1 o todos 5 sin examinar a que equivale el valor de uno y de cinco, motivo por el cual se añaden preguntas de control [14]. Otro problema es la falta de personalización de los cuestionarios en función del cliente.

Al convertir los cuestionarios cerrados en preguntas abiertas se permite a la empresa llevar a cabo un seguimiento de la satisfacción de cliente de forma más eficiente. De este modo, al disponer de la información necesaria, se puede conocer de la fuente original lo que puede mejorar en sus procesos de una forma más real.

Los cuestionarios de preguntas abiertas son difíciles de implementar y controlar porque nunca se sabe lo que un encuestado puede contestar. Además, no debemos olvidar que independientemente de que el cuestionario esté confeccionado con preguntas abiertas o cerradas, debe validarse.

Cabe destacar que para sondear y conseguir información del mercado, se suele utilizar también como herramienta el cuestionario que permite obtener la información con un coste relativamente bajo [12].

cliente según la normativa ISO 9001 al obtener mayor información de los clientes.

Este análisis, se ha centrado en las pequeñas y medianas empresas, debido a que el tejido empresarial español está constituido básicamente por micro, pequeñas y medianas empresas, conocidas como Pyme (entre 0 y 249 asalariados) [6]. Según el directorio central de empresas (DIRCE), a 1 de enero del 2013 hay en España 3.195.210 empresas de las cuales 3.191.416 (99,88%) son Pyme [7].

Aunque este estudio es aplicable a todas las organizaciones independientemente de su tamaño y actividad.

Autores como Zaidi [3] y Prieto [4] indican que el 96 % de los clientes insatisfechos no se quejan, el 91% de los clientes insatisfechos no repiten y que cada cliente insatisfecho se lo dice a 10 personas. Además estadísticamente se ha calculado que cada reclamación significa aproximadamente 23 clientes perdidos y 250 informes negativos [5].

Por todo ello, se considera que la evaluación de satisfacción del cliente propuesta por la norma ISO 9001 mediante herramientas como los cuestionarios, es fundamental para conocer el grado de satisfacción del cliente. Al mismo tiempo los cuestionarios con preguntas abiertas permiten conocer mejor las necesidades de los clientes al no quedar limitadas a opciones prefijadas.

La estructura de este artículo está compuesta por una breve introducción, un análisis del estado de la cuestión, la metodología y arquitectura empleadas en el estudio.

Por último se presentan los resultados, conclusiones y las futuras líneas de investigación.

2. Estado de la cuestión Las organizaciones buscan adaptarse a un mercado cada vez mas global, competitivo, y cambiante. En este ambiente cada empresa busca diferenciarse de su competencia de distintas formas: una de ellas, mediante la certificación de la empresa a través de uno de los estándares existentes de gestión de la calidad, p. e. la ISO 9001 [8] o a través de la autoevaluación prevista por el modelo propuesto por la Fundación Europea para la Gestión de la Calidad (EFQM – del inglés European Foundation for Quality Management) [11].

Para estos sistemas de gestión de la calidad los clientes son las figuras más importantes a la hora de definir los procesos y los requisitos organizativos. Uno de estos procesos pone especial hincapié en que las empresas deben evaluar la satisfacción del cliente, ya que conocer lo que piensan los clientes sobre los productos, los procesos de venta, gestión de incidencia, etc., facilita que la organización cumpla con los requisitos del cliente.

La propia norma ISO 9001 en su apartado 8.2.1 indica que la empresa tiene que establecer medidas necesarias para “realizar el seguimiento de la información relativa a la percepción del cliente respecto al cumplimiento de sus requisitos por parte de la organización” [2] Aunque son muchas las posibles opciones para medir la satisfacción de cliente, esta se ha centrado en los cuestionarios porque permiten una más amplia cobertura, ahorro de tiempo, ya que se pueden obtener respuestas a miles de cuestionarios en horas o días. Prueba de ello es el

3. Metodología Para llevar a cabo la investigación se ha diseñado un cuestionario de satisfacción de cliente según la norma ISO 9001 en el que las preguntas que se plantean son abiertas, es decir el encuestado puede contestar sin limitarse a las respuestas que vienen preestablecidas en los cuestionarios cerrados.

Posteriormente se ha justificado y evaluado la hipótesis planteada, concluyendo que los cuestionarios de preguntas abiertas permiten mejorar la evaluación del grado de satisfacción de cliente según la norma ISO 9001.

Con el desarrollo de un cuestionario y el tratamiento semántico de las cuestiones planteadas, se determinará el grado de información obtenido y se dará solución a los problemas derivados del uso de cuestionarios de preguntas cerradas mediante la realización de un caso práctico que presentamos a continuación.

4. Arquitectura del sistema El entorno de trabajo que se ha diseñado para analizar las respuestas a las preguntas abiertas y estudiar su 93

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

En este caso y debido al idioma castellano el análisis de la respuesta tiene que ser capaz de tener en cuenta en cada palabra el genero, es decir, si se trata de masculino o femenino, y el número, plural o singular, a diferencia de otros idiomas como el inglés que no es necesario al no existir esta discriminación.

No hay que olvidar que el análisis morfológico es el estudio de la estructura de las palabras de un idioma y que a través del mismo se identifican sus respectivos morfemas.

Por el contrario, el análisis semántico clasifica las palabras según su significado.

Si observamos una palabra de tendencia positiva como podría ser “bueno” y sacamos la raíz de dicha palabra, la cual seria “buen” vemos que seguiría siendo de tendencia positiva por lo que todas las palabras derivadas de la raíz “buen” también serán de tendencia positiva (bueno, buena, buenos, buenas, etc.).

De tal manera que si sabemos que la raíz “buen” y sus derivadas son de tendencia positiva, como se han sacado todas las raíces de las palabras, con buscar “buen” dentro de la contestación del cliente, estaríamos buscando implícitamente todas las palabras derivadas de buen, ahorrándonos así buscar el resto de palabras que tengan la raíz “buen”.

Es por ello, que cada respuesta del cuestionario debe ser analizada, obteniendo cada una de las palabras que componen la oración.

Una vez realizado el proceso anterior se eliminan las palabras vacías (en inglés stopwords) [20] o que no aportan información, como son los determinantes, preposiciones y conjunciones que la componen, seguidamente se obtienen la raíces de las palabras y por último se analizan la tendencia positiva, negativa o neutra de cada una de ellas.

Se puede dar varios casos: el caso primero, que la respuesta a la pregunta del cuestionario tenga una única palabra y esta sea un adjetivo, por ejemplo “buenos” o “malos”, en ese caso estos son reducidos a su raíz, “buen-” o “mal-”, y se analizaría si tienen tendencia positiva o negativa.

Por otro lado, las respuestas pueden estar compuestas de varias palabras por ejemplo “son buenas” o “son buenas frente a la competencia”. En estos casos, el sistema debe ser capaz de eliminar aquellas palabras que no tienen un valor sustancial con el fin de detectar si tiene una tendencia positiva o negativa En el primer ejemplo “son buenas” al analizar se obtiene dos palabras “son” puede ser sustantivo o verbo y “buenas” que es adjetivo. En esta situación “son” es neutra, y “buen-” tiene tendencia positiva. En el segundo ejemplo “buenas frente a la competencia” al analizar se obtienen cinco palabras “buenas” “frente” “a”, “la” “competencia”, “a”, “las”, deben ser eliminadas al no aportar información, y las palabras restantes son reducidas a su raíz, por ejemplo “buenas” es reducida a “buen-”, “frente” a “frent-” y “competencia” a “compet-” como apreciamos “buen-“ tiene una clara tendencia positiva mientras que la raíz compet y frent son neutras.

En el caso de los verbos, se realiza un análisis morfológico de verbo, de manera automática, Consiste en encontrar la flexión de un verbo en infinitivo en cualquier

tendencia positiva o negativa permitirá a la empresa evaluar el grado de satisfacción de cliente y al mismo tiempo realizar investigación de campo.

4.1. Justificación del cuestionario según la norma ISO El cuestionario esta compuesto por tres bloques: uno centrado en el producto, otro centrado en los servicios que proporciona la empresa y el último global. La justificación de la elección de estas preguntas y de estos grupos proviene de la norma ISO 9001:2008 Las primeras cuatro preguntas están relacionadas con el apartado 7.2 “Procesos relacionados con el cliente” y más concretamente con el punto 7.2.1. “Determinación de los requisitos relacionados con el producto”. En él se indica que la organización debe determinar (i) los requisitos especificados por el cliente, incluyendo los requisitos para las actividades de entrega y las posteriores a la misma, (ii) los requisitos no establecidos por el cliente pero necesarios para el uso especificado o para el uso previsto, (iii) cuando sea conocido, los requisitos legales y reglamentarios aplicables al producto, y (iv) cualquier requisito adicional que la organización considere necesario.

Además se incluye una nota para las actividades posteriores a la entrega, por ejemplo, acciones cubiertas por la garantía, obligaciones contractuales como servicios de mantenimiento, y servicios suplementarios como el reciclaje o la disposición final.

Las siguientes cinco preguntas están vinculadas al apartado 7.2 Procesos relacionados con el cliente y más concretamente con el punto 7.2.3 Comunicación con el cliente, aunque cada pregunta hace referencia a algún apartado o punto de la norma.

En este apartado 7.2.3. se indica que la organización debe determinar e implementar disposiciones eficaces para la comunicación con los clientes, relativas a la información sobre el producto, las consultas, contratos o atención de pedidos, incluyendo las modificaciones, y la retroalimentación del cliente y sus quejas.

En la última pregunta del cuestionario propuesto: “¿Qué nos diferencia frente a la competencia?”, se trata de obtener la percepción que los clientes tienen de la empresa con respecto a las demás empresas del sector y los matices que los diferencian de ellas, para lo cual se propone la posibilidad de responder mediante contestación múltiple a las características que más diferencien a la empresa.

4.2.

Proceso de evaluación de la tendencia de las cuestiones

Si se analiza una pregunta y su contestación en un cuestionario con preguntas cerradas desarrollado en escala Likert 5, por ejemplo la pregunta “¿Los plazos de entrega son?”, un cliente podría seleccionar entre cinco posibles respuestas “Muy bueno” “Bueno” “Regular” “Malo” y “Muy Malo”.

Si por el contrario, las preguntas del cuestionario se transforman en preguntas abiertas evitamos que el cliente tienda a contestar automáticamente y necesitará escribir el texto de la respuesta, por ejemplo “Buenos”.

94

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

Figura 1. Diagrama de bloques de sistema.

Fuente: Elaboración propia Figura 2. Fases del Sistema.

Fuente: Elaboración propia

tiempo y conseguir lematizarlo a su forma de infinitivo.

Existen diversos lematizadores que usan como idioma base el inglés [27-30], pero en el caso del español es más complicado [31,32]. Una de esas características o limitaciones es la gran cantidad de conjugaciones que puede poseer un verbo en castellano, lo cual dificulta enormemente la generación automática de dichas conjugaciones.

Si la contestación fuera del tipo “no es bueno” el sistema lo detectaría como negativa debido al uso de un adverbio de negación antes de la raíz “buen”, convirtiendo esta raíz positiva a negativa.

Es importante destacar, que no hay la misma complejidad cuando se analiza morfológicamente una palabra en ingles, que una palabra en castellano. El análisis morfológico de verbos en castellano no es una tarea fácil, aunque hay autores que lo han analizado [24,25] debido a sus peculiaridades características.

no está en la base de datos, será añadida previa consulta, en cuanto a si se trata de una palabra de tendencia positiva, negativa o neutra, una vez confirmada, se almacenará en la base de datos en su lugar correspondiente.

Para la representación gráfica [18] se ha optado por un sistema de indicadores mostrado por un cuadro de mando [22]. Esta herramienta nos permitirá facilitar la toma de decisiones en función de los resultados obtenidos.

Todo el desarrollo de la aplicación se ha basado en software libre. Por un lado, el desarrollo de la página web se ha realizado en PHP, ya que puede ser utilizado en los principales sistemas operativos, y se soporta en la mayoría de servidores Web. El núcleo del sistema se ha desarrollado utilizando Lucene [19] por su potencia y posibilidad de modificación, facilita la búsqueda por campos y rangos.

4.3. Integración y conexión entre sistemas

4.4. Ensayo del sistema

La búsqueda de la ventaja competitiva ha facilitado el desarrollo de aplicaciones CRM (Customer Relationship Management). Esta filosofía de negocio busca entender las necesidades de los clientes, por lo que la obtención de información a través de los cuestionarios nos permitirá cotejar que la información recibida por este sistema está en la línea correcta.

El sistema presentado en la Fig. 1 permite que el cuestionario se introduzca por un usuario de forma online mediante una Web o por el contrario se introduzca en el sistema después de recibirse en la empresa por fax o por correo electrónico proporcionando gran versatilidad para los usuarios.

Los resultados de los cuestionarios obtenidos se almacenan en una base de datos MySQL. Una vez introducidos los datos a través del formulario, mediante una aplicación en Java se capturarán todos los datos de las contestaciones de los clientes para saber si dichas contestaciones tienen una tendencia positiva o negativa, para esto se verá si las palabras que forman la contestación son como se ha comentado anteriormente positivas o negativas.

Cuando se realiza el procesamiento y análisis de las respuestas, si el programa detecta alguna palabra nueva que

Como se ha comentado antes, una de las funciones principales de la aplicación es el tratamiento de datos para que posteriormente sean utilizados por los demás programas. En la Fig. 2 se muestra las fases más importantes del proceso en la cual una de las funciones de esta aplicación es leer las respuestas de los cuestionarios que han sido contestados vía Web y tratadas de la manera adecuada, transformando esta información en información válida o necesaria para el uso en las aplicaciones .

Lo que se realiza es una lectura completa de todas y cada una de las contestaciones de los clientes que están almacenadas en una tabla de una base de datos. Pero de estas contestaciones, no todo aporta información ya que hay partes que lo único que hacen es difuminar la información existente, consumir espacio de manera inútil o ralentizar el propósito de nuestro sistema.

Por ello, como se ha comentado anteriormente, una vez almacenadas las cadenas de texto contestadas en el cuestionario, son extraídas por el programa y se eliminarán las palabras que no aportan información significativa.

Estas palabras llamadas comúnmente palabras vacías o en ingles, stopwords [20], son aquellas palabras sin significado como artículos, pronombres, preposiciones, etc.

Una vez eliminadas las palabras vacías, para que la 95

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

Seguidamente se procedió a leer todas las palabras restantes y se pasará a extraer las raíces de cada una.

Para sacar las raíces de las palabras, se utiliza un lematizador (stemmer o algoritmo de stemming en inglés) que permite reducir una palabra a su raíz o (en inglés) a un stem o lema.

Cabe destacar que hay muchos para la lengua inglesa, pero por el contrario es difícil encontrar alguno para el idioma castellano.

La lematización de palabras se ha usado en un proyecto llamado Snowball que posee una librería aplicable a la programación en Java y se integra de forma perfecta con la herramienta de indexación y búsqueda Lucene.

En este caso la utilizarnos junto al algoritmo de Snowball para eliminar las palabras vacías y obtener de manera menos compleja y eficiente las raíces de las palabras e indexar las raíces de estas palabras obtenidas.

Snowball es un pequeño lenguaje para el manejo de strings que permite implementar algoritmos de normalización del lenguaje (stemming algorithms) mediante sencillos scripts. Posteriormente mediante un compilador se genera una salida en C o en Java.

Una vez indexadas las palabras en un índice, Lucene permite buscar las palabras dentro de dicho índice, mostrando la cantidad de palabras que ha encontrado dentro y que son iguales a la que se le ha indicado buscar.

Esto es muy útil en este caso, que queremos saber la cantidad de palabras positivas y negativas para obtener la tendencia de la contestación.

Todos estos algoritmos de lematización están basados en un algoritmo que se usaba inicialmente para la lematización de las palabras inglesas, y este es el algoritmo de Porter [26].

Una vez obtenidas las raíces de todas las palabras, sólo tendríamos que buscar entre ellas cuales son positivas y cuales negativas. Dependiendo de la cantidad de palabras positivas y negativas que haya en una contestación de una pregunta, se podrá obtener si dicha respuesta es de tendencia positiva o negativa

eficiencia de la aplicación sea aun mayor, realizaremos un barrido de stemming. El stemming es el método que permitirá reducir una palabra a su raíz.

Esto facilitará el procesado de la aplicación, ya que se buscan palabras positivas y negativas, pero si de dichas palabras se obtiene solo la raíz se disminuirá considerablemente el número de palabras con las que trabajar, ya que si ponemos como ejemplo una palabra positiva que podría ser bueno, su raíz seria buen y buscando esta raíz “buen” nos evitaríamos buscar el resto de palabras con dicha raíz como podrían ser buen, bueno, buenos, buenas. En vez de buscar entre todas esas palabras buscaríamos sólo sobre una, la cual es la raíz “buen”.

En este caso se utilizará el algoritmo de steamming Snowball junto a Lucene, que es una API (del inglés Application Programming Interface) de código abierto para recuperación de información implementada en Java [21,23].

La eliminación de estas palabras hace que sea más ligero el algoritmo de procesamiento de palabras, ya que elimina palabras que para nuestro fin, no son relevantes en ningún caso y tendremos que tratar con menos palabras y esto provocará que el sistema sea bastante más rápido y eficiente.

4.4.1. Motor de búsqueda El núcleo principal de esta aplicación, que procesa la mayor parte de información, se ha desarrollado con el fin de interactuar con Lucene.

Lucene no es un motor de búsqueda, pero se puede utilizar para implementar un motor de búsqueda, ya que es una herramienta de indexado y búsqueda de texto completo que posee funciones muy variadas e increíblemente útiles para este fin.

Lucene es una herramienta que permite tanto la indexación cómo la búsqueda de documentos y está implementada completamente en Java. Como se comentó anteriormente, Lucene no es un programa, sino una API, a través de la cual se añaden capacidades de indexación y búsqueda.

Se va a utilizar Lucene para las siguientes funcionalidades: (i) Eliminación de las palabras vacías, (ii) transformación de palabras a sus raíces, (iii) inserción de dichas palabras dentro de un índice de Lucene y (iv) poder buscar dentro de ese índice.

Lucene posee funciones por defecto para eliminar las palabras vacías, pero esta función se realiza para filtrar sólo palabras en inglés, en nuestro caso, ha sido necesario modificarlo creando una lista con todas o las máximas palabras vacías que hay en el castellano (se puede encontrar listas de palabras vacías fácilmente por internet).

Esto se puede realizar con la herramienta Lucene creando un array de tipo string que contenga todas las palabras vacías y pasar ese array a una función predefinida de Lucene junto con el texto o frase a la que se le quieren eliminar las palabras vacías, devolviendo esta función a un texto en el que se han eliminado todas esas palabras

5. Resultados Todo el proceso para analizar cada contestación de cada cliente no tendría ningún valor si no pudiéramos convertir esos datos en información.

Con ese fin se ha generado un cuadro de mando como herramienta que permita la toma de decisiones y se han implementado gráficas que facilitan su compresión de una manera rápida. Para ello nos hemos basado en la norma UNE 66175 [33].

En la Fig. 3 se puede ver cómo a partir de la contestación de los cuestionarios recibidos y su análisis, se muestran las preguntas positivas y negativas así como su porcentaje.

Durante la exposición del cuestionario propuesto se han desarrollado dos grandes grupos de cuestiones (i) uno referente a producto y (ii) otro referente a la información.

En la Fig. 4 se observa el valor de tendencias positivas y negativas por pregunta referente al grupo de cuestiones de producto.

96

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

En la Fig. 5 se puede observar el valor de tendencias positivas y negativas por pregunta.

Finalmente, en la Fig. 6 se representa la información mediante un diagrama radial en el que se compara lo que se opina de la empresa frente a la competencia.

6. Conclusiones y futuros trabajos En este artículo se ha descrito una aplicación que ayuda a las organizaciones que la usan a cumplir con la norma ISO 9001, facilitando la comparación entre los datos de CRM y los obtenidos a través del cuestionario.

El uso de cuestionarios Web facilita a la empresa obtener mayor número de cuestionarios a un coste relativamente bajo. Ya que este tipo cuestionario siempre tiene unas respuestas predefinidas para que el cliente conteste con una de ellas, al mismo tiempo, se obtienen las ventajas de las entrevistas.

Los cuestionarios tal y como se han implementado en este proyecto no son comunes, por la complejidad de los algoritmos usados para poder sacar las tendencias de las palabras.

Estos algoritmos de lematización son más comunes en idiomas como el inglés, pero en el caso de buscarlos con el idioma español son casi inexistentes por el gran incremento de complejidad que sufre el algoritmo debido a las formas verbales del español.

Este sistema permitirá a las empresas conocer más y mejor la opinión de sus clientes, no solo la respuesta sino que también cualquier matización de ella.

Además elimina errores cometidos en los cuestionarios de preguntas cerradas como son la contestación automática al cuestionario marcado todos 1 o todos 5 sin examinar a que equivale el valor de uno y de cinco y la eliminación de preguntas de control.

Dentro de las futuras líneas de investigación merece la pena analizar el grado de tendencia de las contestaciones a las cuestiones planteadas, así como relacionar este sistema con los CRM con el fin de realizar un análisis semántico de los informes de cada cliente y su verificación con los datos obtenidos de los cuestionarios de evaluación.

Aunque los cuestionarios son específicos como hemos indicado anteriormente, uno de los campos donde su aplicación puede ser importante a partir de los datos obtenidos en nuestro estudio es la investigación de mercados, donde los cuestionarios son una herramienta indispensable para desarrollar estudios de mercado.

Por otro lado como futura línea de trabajo se analizará el grado de positivismo y negatividad de la respuesta, ya que no es lo mismo bueno que excelente, pasando de un valor bivaluado por palabra, a un valor multivaluado en cada una.

Figura 3. Contestación respuestas por pregunta.

Fuente: Elaboración propia

Figura 4. Información sobre los productos.

Fuente: Elaboración propia

Figura 5. Tendencia por pregunta.

Fuente: Elaboración propia

Referencias [1] [2] [3] Figura 6. Información sobre lo que nos diferencia frente a la competencia Fuente: Elaboración propia 97

Friedman, T. L. La tierra es plana: Breve Historia del Mundo Globalizado del Siglo XXI. Ciudad de edición, Ediciones Martínez Roca 2006.

Soriano, C. L. La ventaja competitiva. Madrid, Ediciones Díaz de Santos S.A., 1997.

Zaidi, A. QFD Despliegue de la función calidad, Madrid, Díaz Santos, 1993.

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

[4] [5] [6]

[7] [8] [9] [10] [11] [12] [13]

[14]

[15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

Prieto, J., El servicio en acción: la única forma de ganar a todos, Bogotá, ECOE Ediciones, 2005.

Plaza, A.S., Apuntes teóricos y ejercicios de aplicación de gestión del mantenimiento industrial- Integración con calidad y riesgos laborales, lulu.com, 2009.

Moreno, M.E., Pyme Española: Características y Sectores. [en línea] [consulta: abril 8 de 2013] Disponible en: http://www.microsoft.com/business/es-es/ content/paginas/article.aspx?cbcid=108.

DIRCE (Directorio Central de Empresas), Retrato de las Pyme 2013, [en línea], [consulta: abril 23 de 2013]. Disponible en: http://www.ipyme.org/Publicaciones/Retrato_PYME_2013.pdf ISO 9001:2008, Sistema de gestión de la calidad. Requisitos, Madrid, Aenor, 2008, 42 P. ISO Survey 2011, World distribution of certificates in 2011 [online], [date of reference may 12th of 2013]. Available on: http://www.iso.org/iso/home/standards/certification/iso-survey.htm.

UNE 66176, Guía para la medición, seguimiento y análisis de la satisfacción del cliente, Madrid, Aenor, 2005, 30 P. EFQM, EFQM MODEL [en línea] [consulta, mayo 22 de 2013].

Disponible en: http://www.efqm.org.

Díaz-de Rada, V., Flavián, C. y Guinaliu M., Encuesta en Internet; Algo más que una versión mejorada de la tradicional encuesta autoadministrada, Investigación y Marketing, 82, pp. 45-56, 2004.

García, T., El cuestionario como instrumento de investigación/evaluación [en línea] [consulta, mayo 15 de 2013] Disponible en: http://www.univsantana.com/sociologia/El_Cuestionario.pdf . del Castillo, A.M. Investigación de mercados, [en línea] [consulta, febrero 10 de 2013] Disponible en: http://openmultimedia.ie.edu/OpenProducts/usmc/usmc/pdf/USMC. pdf.

DeVellis, R.F., Scale development: Theory and applications, Thousand Oaks Publicacions Inc., 2012.

Dutka, A., Manual de la A.M.A. Para la satisfacción de cliente.

Buenos Aires, Ediciones Granica, S.A., 1998.

Google, Cómo crear un formulario de Google, [en línea] [consulta, marzo 13 de 2013] Disponible en: https://support.google.com/docs/answer/87809?hl=es Chart J., [en línea] [consulta, marzo 20 de 2013]. Disponible en: http://www.jfree.org/jfreechart/.

Lucene, [en línea] [consulta, marzo 1 de 2013]. Disponible en http://lucene.apache.org/.

Stop-words. [en línea] [consulta, marzo 12 de 2013] Disponible en: https://code.google.com/p/stop-words/ [citado 12 de marzo de 2013].

Configure and manage stopwords and stoplists for full-text search, [en línea] [consulta, mayo 28 de 2013].

Disponible en: http://msdn.microsoft.com/en-us/library/ms142551.aspx Kaplan, R. and Norton, D., Cuadro de mando integral. Gestión 2.000. Barcelona. 2000.

Cutting, D., Lucene, [en línea] [consulta, mayo 28 de 2013].

Disponible en: http://apachefoundation.wikispaces.com/Apache+Lucene.

Zapata, C.M, and Mesa, J.E., Una propuesta para el análisis morfológico de verbos del español, DYNA, 76 (157), pp 27-36, 2009 Zapata, C. M and Mesa, J.E., Los modelos de diálogos y sus aplicaciones en sistemas de diálogo hombre-maquina: Revisión de la literatura, DYNA, 76 (160), pp. 305-315, 2009 The Porter stemming Algorithm, [onlíne], [date of reference, june of 2013].

Available on: 13th http://snowball.tartarus.org/algorithms/porter/stemmer.html The English (Porter 2) stemming Algorithm, [onlíne], [date of of 2013].

Available on: reference june 10th http://snowball.tartarus.org/algorithms/english/stemmer.html Stemming early English, [onlíne], Available on: http://snowball.tartarus.org/texts/earlyenglish.html Muise, C., McIlraith, S., Baier, J.A. and Reimer, M., Exploiting Ngram analysis to predict operator sequences 2009. In 19th International Conference on Automated Planning and Scheduling (ICAPS09), Thessaloniki, Greece. A version of this paper also appeared in the ICAPS09 Workshop on Learning in Planning.

[30] Rojas-Galeano, S.A., Revealing non-alphabetical guises of spamtrigger vocables, DYNA, 80 (186), pp 50-57. 2013 [31] Molino de Ideas, Lematizador, [en línea] [consultado junio 12 de 2013].Disponible en: http://www.molinolabs.com/lematizador.html [32] Grampal, Lematizador para el español desarrollado en la Universidad Autónoma de Madrid, [en línea] [consultado junio 13 de 2013].

Disponible en: http://cartago.lllf.uam.es/grampal/grampal.cgi [33] UNE 66175:2003. Sistema de gestión de la calidad. Guía para la implantación de sistemas de indicadores, Aenor, 2003, 30 P. J. A. Medina-Merodio, Ingeniero Técnico en Telecomunicaciones en 2001 e Ingeniero Técnico Industrial en 2010 por la Universidad de Alcalá, Licenciado en Investigación y Técnicas de Mercado en 2007 y Doctor en Dirección de Empresas por la Universidad Rey Juan Carlos de Madrid en 2010. En la actualidad es Profesor Ayudante Doctor en el Departamento de Ciencias de la Computación en la Universidad de Alcalá. Anteriormente desempeñó funciones de Técnico de Gestión (Sistemas y Tecnologías de la Información) en el Servicio de Salud de Castilla la Mancha, Técnico Informático en ADAC- Asociación para el Desarrollo de la Alcarria y la Campiña y trabajó como Responsable de Calidad, Medioambiente y Servicios Informáticos en Tecnivial, S.A. ORCID ID: http://orcid.org/0000-0003-3359-4952 C. de Pablos-Heredero, Doctora en Ciencias Económicas y Empresariales es profesora en el área de Organización de Empresas en la Universidad Rey Juan Carlos de Madrid desde 1994. Responsable del programa de Doctorado en Organización de Empresas. Está especializada en el impacto de las tecnologías de información y comunicación en los sistemas organizativos donde desarrolla sus investigaciones principalmente. Ha dirigido tesis doctorales y proyectos en esta temática y publicado 78 artículos en revistas especializadas con índices de impacto y escrito 8 libros. Ha trabajado como consultora en Primma Consulting y es la Directora Académica del Master Oficial en Organización de Empresas y co-directora del Master en Emprendimiento y el Master en gestión de proyectos logísticos SAP ERP en la Universidad Rey Juan Carlos.

ORCID ID: http://orcid.org/0000-0003-0457-3730 M. L. Jiménez-Rodríguez, es Licenciada en Ciencias Matemáticas por la Universidad Complutense de Madrid, en 2002 y Doctora por la Universidad de Alcalá, en 2006. En la actualidad es Profesora Titular en el área de Lenguajes y Sistemas Informáticos adscrito al Departamento de Ciencias de la Computación de la Universidad de Alcalá y profesora-tutora de la Universidad Nacional de Educación a Distancia (UNED) en los estudios de Grado en Ingeniería Informática y Grado en Administración y Dirección de Empresas.

ORCID ID: http://orcid.org/0000-0003-4398-5404 L. de Marcos-Ortega, Ingeniero Informática en 2005 y Doctor en Informática en 2009 por la Universidad de Alcalá donde actualmente ocupa plaza de profesor ayudante doctor en el Departamento de CC. de la Computación. Cuenta con más de 90 publicaciones en revistas y congresos.

Entre sus áreas de interés destacan los sistemas educativos y la formación en el ámbito de la informática (objetos docentes, competencias, sistemas adaptativos, movilidad y estandarización en las TIC aplicadas al aprendizaje); junto con distintos aspectos del ámbito de la computación evolutiva (algoritmos genéticos, enjambres partículas y optimización combinatoria). Ha participado en numerosos proyectos de investigación y cuenta con amplia experiencia en tareas de gestión y administración de estos proyectos. Ha completado estancias en las universidades de Lund (Suecia) y Reading (Reino Unido), y en el Instituto Tecnológico de Monterrey (México) donde ha realizado diversas actividades relacionadas con la docencia y la investigación.

ORCID ID: http://orcid.org/0000-0003-0718-8774 R. Barchino-Plata, es Ingeniero en Informática por la Universidad Politécnica de Madrid en 2003 y Doctor por la Universidad de Alcalá en 2007. En la actualidad es Profesor Titular del Área de Lenguajes y Sistemas Informáticos adscrito al Departamento de Ciencias de la Computación de la Universidad de Alcalá además de ser Profesor Tutor de la Universidad Nacional de Educación a Distancia – UNED. Ocupa el puesto de Coordinador de Estudios Virtuales dentro del Instituto de 98

Medina-Merodio et al / DYNA 81 (188), pp. 92-99. December, 2014.

Ciencias de la Educación de la Universidad de Alcalá. Es miembro, en calidad de vocal, del comité técnico de normalización 71, Subcomité 36: Tecnologías de la Información para el Aprendizaje de AENOR – Asociación Española de Normalización y Certificación.

ORCID ID: http://orcid.org/0000-0002-5657-5191 D. Rodríguez-García, es licenciado en informática por la Universidad del País Vasco (EHU/UPV) en 1995 y doctorado en informática por la Universidad de Reading, Reino Unido, en 2003, donde ha sido profesor desde octubre del 2001 hasta septiembre 2006. Daniel obtuvo el certificado en educación universitaria por la Universidad de Reading en julio del 2005.

Desde octubre 2006, es profesor en el departamento de Ciencias de la Computación de la Universidad de Alcalá, Madrid, consultor on-line en la UOC colabora habitualmente con Oxford Brookes Reino Unido. Además, Daniel tiene 2 años de experiencia en la empresa privada como ingeniero de software y consultor.

Sus intereses en la investigación se centran en la ingeniería del software, y la aplicación de distintas técnicas de la minería de datos y computación a la ingeniería del software. Daniel es miembro de las asociaciones profesionales ACM e IEEE. ORCID ID: http://orcid.org/0000-0002-2887-0185

Área Curricular de Ingeniería Administrativa e Ingeniería Industrial Oferta de Posgrados     

Especialización en Gestión Empresarial Especialización en Ingeniería Financiera Maestría en Ingeniería Administrativa Maestría en Ingeniería Industrial Doctorado en Ingeniería - Industria y Organizaciones

Mayor información: Elkin Rodríguez Velásquez Director de Área curricular acia_med@unal.edu.co (57-4) 425 52 02

99

Analysis of the economic impact of environmental biosafety works projects in healthcare centres in Extremadura (Spain) Justo García Sanz-Calcedo a & Pedro Monzón-González b b

a Centro Universitario de Mérida, Universidad de Extremadura, Mérida, España. jgsanz@unex.es Escuela Politécnica de Cáceres, Universidad de Extremadura, Cáceres, España. pmonzon@sas.juntaex.es

Received: December 2th, 2013. Received in revised form: July 14th, 2013. Accepted: October 31th, 2014.

Abstract The aim of this paper is to analyze the results obtained in the methodological application of techniques aimed at the maintenance of environmental biosafety in works of reform and expansion of healthcare centres in Extremadura, Spain during 2004-2010, assessing the costs of its implementation and contrasting if the use of a BSA project in phase of works affects the probability of nosocomial infection and the conditions of health and safety. The average investment accounted for a cost of 5.5 €/m2 under construction in critical areas and 0.9 €/m2 in the rest works. The impact of biosafety on the work budget draft was a 1.07% at hospitals and a 0.57% at healthcare centers.

In critical areas, the sectorization was an average investment of the 26.89%, while in other areas it was the 29.48%. The largest investment in classification corresponds to small venues in critical areas involving the 40.64%.

Keywords: biosafety; public works; project management; health management; nosocomial infection.

Análisis del impacto económico de la bioseguridad ambiental en proyectos de obras en centros sanitarios de Extremadura (España) Resumen Se analizan los resultados obtenidos en la aplicación metodológica de técnicas destinadas al mantenimiento de la bioseguridad ambiental en obras de reforma y ampliación de centros sanitarios en Extremadura (España) durante 2004-2010, evaluando los costes derivados de su implantación y contrastando si la utilización de un proyecto de BSA en fase de obras afecta a la probabilidad de infección nosocomial y a las condiciones de seguridad y salud. La inversión media fue de 5,5 €/m2 en obras en zonas críticas y 0,9 €/m2 en el resto. La repercusión media del proyecto de bioseguridad sobre el presupuesto de obra, fue de un 1,07% en hospitales y de un 0,57% en Centros de Salud. En zonas críticas la sectorización supuso una inversión media del 26,89%, mientras que en el resto de zonas fue del 29,48%. La mayor inversión en sectorización corresponde a pequeñas actuaciones en zonas críticas (40,64%).

Palabras clave: bioseguridad; obras públicas; gestión de proyectos; gestión sanitaria; infección nosocomial.

1. Introducción Los principales agentes causantes de las infecciones asociados a obras en centros sanitarios son los hongos, siendo el Aspergillus (Fumigatus, Flavus, Niger, Terreus…) el más frecuente, así como las bacterias, entre las que destaca la Legionella (Pneumophila y Bozemanii) [1]. Los primeros se encuentran presentes en el suelo, el agua y la vegetación en descomposición, presentando esporas que pueden permanecer suspendidas en el aire durante periodos prolongados de tiempo, aumentando la probabilidad de inhalación o de contaminación de las superficies expuestas [2]. Estas esporas pueden dispersarse en las partículas de

polvo que se generan en los procesos de construcción o a través de la suciedad que se produce y se acumula en los suelos, paredes y techos [3].

La Legionella se encuentra presente en torres de refrigeración, redes hídricas y equipos de producción de agua caliente sanitaria, fundamentalmente en los depósitos de acumulación, donde las condiciones de estratificación del agua favorecen su proliferación [4]. Durante las obras de reforma o ampliación de un centro sanitario se producen cortes e interrupciones en el suministro que provocan que en algún tramo de las redes hídricas se den las condiciones de temperatura y estratificación requeridas por la bacteria, contaminando el agua al restablecerse el suministro [5].

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 100-105. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41030

Sanz-Calcedo & Monzón-González / DYNA 81 (188), pp. 100-105. December, 2014.

Además, la tierra y el polvo, que contienen formas inactivas de Legionella, pueden desplazarse por el aire durante las excavaciones y movimientos de tierra en obras próximas, depositándose y contaminando las bandejas de condensados de las torres de refrigeración [6].

Las resultados obtenidos en la aplicación de medidas de bioseguridad ambiental en obras de reforma de centros sanitarios, los costes invertidos en su aplicación y las pautas seguidas en su desarrollo no están suficientemente explicitados en la literatura científica [7]. Aunque la relación entre brotes de infecciones nosocomiales y la ejecución de obras de reforma en centros sanitarios es un hecho bien documentado desde hace más de 25 años [8-11], no está cuantificada la magnitud real del problema por motivos de infra-diagnóstico y sesgo de no publicación [12]. Sin embargo, las consecuencias de una infección fúngica o bacteriana asociadas a contaminación ambiental derivada del proceso de obras en los hospitales están perfectamente documentadas [13].

La concentración media admisible de Aspergillus dentro de un hospital varía entre 3–105 UFC/m3, según la bibliografía [14-17].

Curtis et al realizaron durante un año el control en un hospital terciario y observaron numerosos picos en la concentración aérea de hongos filamentosos, como consecuencia de intervenciones o incidentes en los sistemas de filtración del aire [18].

Ante la complejidad organizativa de los hospitales y centros sanitarios en general, es necesario elaborar una estrategia común, para conseguir un nivel de prevención que resulte suficiente para evitar las infecciones durante la realización de obras [19]. Dicha estrategia se inicia en la fase de elaboración del proyecto, donde los profesionales sanitarios deben colaborar estrechamente con el equipo técnico redactor, identificando los grupos de riesgo y/o la criticidad de las zonas de actuación [20]. El mantenimiento de un sistema de registro continuo de los niveles de bioseguridad ambiental es básico, ya que en la mayor parte de las ocasiones en las que se detectan recuentos fúngicos elevados, no se puede identificar la causa que los ha producido [21].

Se han editado distintas guías, manuales e instrucciones técnicas, encaminados a la gestión de la bioseguridad ambiental (BSA) en fase de ejecución de obras. En 2000, el Instituto Nacional de la Salud de España, en colaboración con la Sociedad Española de Medicina Preventiva, Salud Pública e Higiene, editó la guía “Recomendaciones para la vigilancia, prevención y control de infecciones en hospitales en obras” [22].

La Sociedad Española de Enfermedades Infecciosas y Microbiología Clínica editó en 2010 el documento “Recomendaciones sobre la prevención de la infección fúngica invasora por hongos filamentosos” [23], que presenta un conjunto de pautas de actuación y organización, basadas en evitar que se produzcan los factores de riesgo asociados a las infecciones nosocomiales. No obstante, en ninguna de las anteriores guías, ni en la literatura científica se han encontrado precedentes sobre los resultados prácticos de su aplicación en muestras homogéneas de edificios [24].

Algunos autores han manifestado la idoneidad de incluir dentro del Estudio de Seguridad y Salud, los documentos referidos a la Bioseguridad Ambiental [25] que, entre otras ventajas, puede agilizar la redacción, facilitar la comprensión y reducir la extensa documentación de un proyecto.

Otros autores consideran que el ambiente inanimado es un factor que contribuye en una pequeña proporción al desarrollo de infecciones nosocomiales, sobre todo cuando se compara con otros factores tales como la adhesión del personal sanitario a las medidas habituales de prevención de la infección [26].

El Servicio Extremeño de Salud ha realizado durante el periodo 2004-2010, la reforma y ampliación de gran parte de su parque inmobiliario, utilizando en el desarrollo de las obras de reforma y ampliación la metodología citada.

El objetivo de este trabajo es analizar los resultados obtenidos en la aplicación de una metodología de gestión del mantenimiento de la bioseguridad ambiental en obras de reforma y ampliación de hospitales y otros centros sociosanitarios, cuantificando el impacto económico de la gestión de la bioseguridad ambiental.

2. Materiales y métodos En el desarrollo de este trabajo, relacionado con otros previos de los autores [27-29], se analizaron 42 obras de reforma y ampliación de centros sanitarios, realizadas por el Servicio Extremeño de Salud, desarrolladas entre 2004 y 2010 en Extremadura (España), 30 realizadas en hospitales (71%) y 12 en centros de salud (29%), con un presupuesto total superior a 58.000.000 € y se evaluaron las incidencias que en materia de bioseguridad fueron detectadas Para determinar la repercusión de la bioseguridad en el desarrollo de cada una de las obras, se revisó el expediente técnico de la misma, el proyecto de ejecución, así como las incidencias que fueron documentadas en materia de bioseguridad ambiental en la Subdirección de Obras, Instalaciones y Equipamiento del Servicio Extremeño de Salud.

Para determinar la incidencia del Plan de Bioseguridad en las condiciones de Seguridad y Salud de la obra, se realizaron encuestas y entrevistas personales a los coordinadores de Seguridad y Salud de cada una de las obras analizadas.

Para realizar el análisis cuantitativo las obras se clasificaron según la producción de polvo, en cuatro grupos, en base al documento “Construction-related Nosocomial Infections in Patients in Health Care Facilities[30]. Los grupos fueron los siguientes: 1º Trabajos de inspección y actividades no invasivas.

2º Trabajos a pequeña escala o con un nivel mínimo de generación de polvo.

3º Trabajos que generan un nivel considerable de polvo o que conllevan la demolición de algún elemento.

4º Trabajos de demolición, construcción o reforma.

Las áreas afectadas por las obras se clasificaron en cinco zonas según su ubicación, el área afectada y su nivel de criticidad, utilizando como base lo establecido en el documento “Recomendaciones para la vigilancia, prevención y control de Infecciones en Hospitales en Obras” [22] e incluyendo modificaciones para extenderlo a otros edificios asistenciales: Z1: zonas interiores críticas Z1a: grandes actuaciones en Z1 Z1b: actuaciones menores en Z1

101

Sanz-Calcedo & Monzón-González / DYNA 81 (188), pp. 100-105. December, 2014.

Tabla 1.

Medidas de prevención en función del grupo de riesgo y el tipo de obra.

INSPECCION Y ACTIVIDADES NO INVASIVAS

BAJA EMISIÓN DE POLVO

NIVEL MODERADO O ALTO DE POLVO

DEMOLICIÓN, CONTRUCCIÓN Y REFORMA

zonas comunes, oficinas, salas de baja ocupación,…

I

II

II

II-III

Z3 atención al paciente, zonas pacientes ambulantes, admisión,…

I

II

III

III

zonas contiguas a zonas críticas

II

IV

IV

IV

V

V

V

V

I

II

III

III

CLASIFICACIÓN DE ÁREAS HOSPITALARIAS

Z2

quirófanos alto riesgo, pacientes neutropénicos, UCI, REA,…

Figura 1. Diagrama de bloques de la metodología utilizada.

Fuente: Elaboración propia.

Z2: zonas interiores contiguas a las críticas Z3: zonas interiores no incluidas en Z1 y Z2 Z4: zonas exteriores CS: centros de salud En cada una de las obras analizadas, se examinaron los siguientes documentos:  Plan de Bioseguridad.  Documento de aprobación del Plan de Bioseguridad.  Anexos al Plan de Bioseguridad.  Modificaciones del Plan de BSA y sus aprobaciones.  Informe de las inspecciones periódicas realizadas por técnicos del Servicio de Obras.  Incidencias de Bioseguridad Ambiental producidas en el desarrollo de la obra.  Informe de inspección final de las obras.  En la Fig. 1 se puede observar el proceso que se siguió durante el desarrollo y ejecución de las obras.

Para el mantenimiento de la bioseguridad en fase de obras se exigió un Proyecto de Bioseguridad Ambiental redactado por el equipo redactor del proyecto de obras, que a modo de documento independiente del proyecto pero inseparable de este, contenía toda la información necesaria para su comprensión, desarrollo y puesta en obra y que estaba compuesto por los siguientes documentos:  Memoria  Pliego de Condiciones  Planos  Mediciones y Presupuesto El primer documento, la Memoria, describía de forma detallada la obra justificando pormenorizadamente las soluciones adoptadas, clasificando las áreas hospitalarias y pacientes de riesgo afectados, para determinar las áreas interiores y el nivel de criticidad de las mismas y definiendo el factor de riesgo de los pacientes. Para determinar las medidas de prevención a adoptar, se utilizó la Tabla 1, en la

Z1 resto quirófanos, REA, quemados, aislados,… todas las zonas CS asistenciales del centro de salud

Fuente: Elaboración propia a partir de Public Health Agency of Canada, [30]

Tabla 2.

Medidas de prevención a adoptar en función del nivel de riesgo.

Nivel

Riesgo

I Muy bajo II Bajo III Moderado IV Alto V Muy alto Fuente: Elaboración propia

Medidas Mínimas Simples Medias Elevadas Extremas

que se representa mediante doble entrada, la elección de las medidas de prevención.

Según la zona de obras, en función del riesgo de los pacientes y la tipología de las mismas, se determinaron un paquete de medidas que detallaban todas las disposiciones de prevención, especialmente aquellas medidas destinadas a minimizar la generación de polvo, los procedimientos, la tipología y las características principales de la sectorización previa al comienzo de cada fase de obra [31]. En la Tabla 2 se exponen el tipo de medidas adoptadas en función del nivel de riesgo intrínseco de cada obra.

Se realizó una división por fases y sectorización de las obras definiendo las fases en las que se divide la obra, consensuadas previamente con los equipos asistenciales, evitando dejar al azar el orden en la realización de los trabajos de construcción y buscando el equilibrio entre la protección y la funcionalidad del hospital.

En el documento Pliego de Condiciones, se expresaban los procedimientos constructivos para sectorizar, el

102

Sanz-Calcedo & Monzón-González / DYNA 81 (188), pp. 100-105. December, 2014.

protocolo de realización de los trabajos, los medios, herramientas, maquinaria y mano de obra necesaria para conseguir que se realicen en condiciones que minimicen la generación de polvo, los criterios de medición de las partidas abonables, así como otras cuestiones.

Otro de los documentos que conforman el proyecto de Bioseguridad, es el denominado Mediciones y Presupuestos, que incluía las partidas necesarias para realizar las medidas preventivas, la sectorización y el sellado de la zona de obras, la mano de obra precisa para el mantenimiento de dichas medidas, así como el desmontaje de las mismas, la limpieza y la desinfección necesaria.

El último documento, los Planos, aportaba la documentación gráfica necesaria para definir las zonas afectadas y su nivel de riesgo, el estado actual, el estado reformado, la división por fases y la programación, la sectorización, los sellados y los detalles que definen además las actuaciones necesarias para garantizar la bioseguridad de las zonas expuestas, indicando los espacios destinados a acopios, escombros y accesos de obra.

La Comisión de Obras [32] estaba formada por el Hospital (Dirección, Servicios de Ingeniería y Mantenimiento, Medicina Preventiva y responsables asistenciales de las áreas afectadas), la Dirección Facultativa de las obras, la Empresa Constructora y los Técnicos del Servicio de Salud, en representación del Órgano de Contratación. Elaboraba el documento: “Condicionantes de Bioseguridad en la Programación de las Obras”, en el que se analizaban pormenorizadamente las obras a realizar, la prioridad en el orden de ejecución, los traslados de pacientes y el mantenimiento o paro de la actividad asistencial. Este documento se utilizó como base para la elaboración del proyecto de bioseguridad ambiental.

El Proyecto de BSA, una vez supervisado con el resto del proyecto por la Oficina de Supervisión, era revisado por la Comisión de Obras al objeto de determinar su validez. En caso de que surgiese algún cambio asistencial no contemplado en los documentos previos, se remitía al Director del proyecto para adaptar el mismo a las nuevas necesidades asistenciales. Una vez realizadas y comprobadas las correcciones, la Comisión de Obras informaba favorablemente el Plan para que la dirección del centro sanitario procediese a su aprobación.

3. Resultados Se ha analizado la distribución porcentual de las actuaciones realizadas en función de la actividad del área de intervención, observándose que las zonas críticas (Z1) representan el 17% de la muestra, las zonas colindantes con las críticas (Z2) el 33% y las zonas no incluidas en los anteriores apartados, denominadas Z3, el 50% de la muestra.

En la Tabla 3 se refleja las actuaciones realizadas durante el periodo 2004-2010, indicando el tipo de edificio, la criticidad del área afectada, la inversión en BSA realizada, el costo en medidas de BSA, el coste en medidas de sectorización y de ejecución, así como la repercusión de la BSA por unidad de superficie y la repercusión porcentual sobre el presupuesto de la obra.

Tabla 3.

Inversión en BSA en obras sanitarias del Servicio Extremeño de Salud (2004-2010).

PRESUPUESTO ÁREA

€

SECTORIZACIÓN

€

Z1a 28.633 24.868 Z1b 4.173 2.477 Z2 27.922 17.993 Z3 2.223 1.703 CS 1.816 1.568 Fuente: Elaboración propia

EJECUCIÓN TRABAJOS

%

€

%

86,85% 59,36% 64,44% 76,61% 86,34%

3.765 1.696 9.929 520 248

13,15% 40,64% 35,56% 23,39% 13,66%

Figura 2. Inversión en BSA por unidad de superficie construida y zona de actuación.

Fuente: Elaboración propia.

Se contrastó que en ninguna de las actuaciones analizadas se produjeron brotes de infección nosocomial por Aspergillus o Legionella en pacientes ingresados en el centro sanitario, por causa de la ejecución de obras de reforma y/o ampliación en el mismo.

En la Fig. 2 se ha representado la inversión por unidad de superficie construida en bioseguridad, en función de la zona donde se ha producido la intervención. Se observa que la inversión en BSA en obras de centros sanitarios osciló entre los 0,67 y los 6,15 €/ m2, teniendo una mayor repercusión porcentual en las obras de menor escala que se realizan en las áreas críticas (Z1) de centros hospitalarios.

Por otro lado se detectó que en actuaciones realizadas en unidades críticas de bajo volumen de obra, la repercusión por unidad de superficie es de 6,15 €/m2 mayor que en actuaciones de mayor envergadura, cuya media es de 4,95 €/m2. También se apreció que conforme es menos trascendente la zona de intervención, y está más alejada de las zonas críticas, zonas Z2 y Z3, la repercusión por unidad de superficie disminuye proporcionalmente, obteniéndose valores de 1,08€/m2 y 0,67 €/m2 respectivamente.

En la Fig. 3 se ha representado la relación entre la inversión en BSA y el presupuesto de la obra expresado porcentualmente, en función de la zona donde se ha producido la intervención. La repercusión porcentual media del proyecto de bioseguridad sobre el presupuesto de ejecución material de la obra, fue de un 1,07% en obras de reforma y ampliación de hospitales y de un 0,57% en obras de reforma de Centros de Salud.

103

Sanz-Calcedo & Monzón-González / DYNA 81 (188), pp. 100-105. December, 2014.

Figura 3. Relación entre inversión en BSA y presupuesto de obra por zonas de actuación.

Fuente: Elaboración propia.

La mayor parte de la inversión en bioseguridad se realiza en tareas de sectorización. En zonas críticas la sectorización supuso el 26,89% del presupuesto, mientras que en el resto de zonas el porcentaje fue del 29,48%. La inversión mayor en sectorización corresponde a pequeñas actuaciones en zonas críticas, cuya sectorización consumió el 40,64% del presupuesto de la obra.

Los técnicos redactores de los proyectos incrementaron su volumen de trabajo en la fase de redacción del proyecto, aunque durante la ejecución de la obra, en su faceta de directores de ejecución, vieron como disminuía su responsabilidad directa, según entrevistas realizadas tras la finalización de las obras.

La mejora de las condiciones de seguridad y salud en la obra, mejoraron en el 92,3% de los casos según una encuesta efectuada a cada Coordinador de Seguridad y Salud, debido a la mejora de los aspectos relativos al orden, la limpieza y las circulaciones en la zona afectada por la obra. De esta forma se disminuye proporcionalmente la probabilidad de accidentes durante la fase de ejecución.

4. Discusión Se ha demostrado que la gestión de la bioseguridad ambiental en obra supone una baja inversión, que implantada previamente a la ejecución de las obras, minimiza la generación y transmisión de polvo y evita la estratificación del aire, disminuyendo la probabilidad de infección nosocomial. La repercusión media del proyecto de bioseguridad sobre el presupuesto de obra en las zonas críticas, ha sido superior en pequeñas obras que en grandes reformas de centros sanitarios. La sectorización en actuaciones menores, requerirán mayor esfuerzo en sectorización que en otras mayores, en las que será necesario clausurar completamente el área, con lo que la sectorización tendrá un peso menor.

En la distribución del presupuesto, las medidas destinadas a sectorizar las zonas afectadas supusieron un peso elevado respecto a otras medias relacionadas con los métodos de trabajo.

El hecho de tener valoradas y presupuestadas previamente las medidas de contención necesarias para garantizar la BSA en la fase de obra permite garantizar la adecuada financiación de las mismas, al preverse con

antelación los recursos necesarios para su ejecución.

Además, la mejora de las condiciones de seguridad y salud en fase de obra y la mínima inversión realizada, producen una rápida amortización, eliminando riesgos y sus posibles consecuencias.

Es aconsejable utilizar el Plan de BSA como estudio singular con entidad propia, independiente del Estudio de Seguridad y Salud de la obra, para definir el tipo de medidas a emplear en función del tipo de obra. Aunque a priori puede parecer lógico incluir la BSA dentro del marco de la Seguridad y Salud de las obras, en el transcurso de este trabajo se ha puesto de manifiesto que es preferible separar ambos documentos, pues tienen objetivos distintos, distinta tramitación e intervienen distintos agentes.

Es conveniente dar traslado de la importancia de las medidas de bioseguridad y salud al Órgano de Contratación, para que incluya en los Pliegos de Condiciones cláusulas que obliguen contractualmente a los contratistas de las obras a respetar las medidas y los procedimientos que se adopten, indicando que algunas de estas medidas afectará a los procedimientos constructivos y al orden de ejecución y a la división por fases de los trabajos, disminuyendo posiblemente el rendimiento y por tanto alterando la productividad de los mismos, variables que deberán ser tenidos en cuenta a la hora de ofertar en el proceso de licitación de las obras.

5. Conclusiones El proyecto de BSA mejora la planificación y el desarrollo de las obras evitando dejar actuaciones al azar o al criterio de las empresas contratistas y su dotación económica permite exigir su ejecución durante las obras.

Con respecto a otros estudios similares, se aportan datos reales que permiten evaluar las medidas de bioseguridad para presupuestar adecuadamente, analizando una muestra muy amplia, que sirve de referencia para evaluar el coste de las medidas de BSA, pues la estrategia de prevención tiene dos fases, la previa o proyectual y la de obras, y las últimas no pueden desarrollarse plenamente si no han sido previamente previstas, planificadas y dotadas presupuestariamente. El estudio presenta las limitaciones propias de estar circunscrito a un territorio concreto, Extremadura, y la posibilidad de infra-notificación de incidencias por los responsables de control y evaluación de las medidas.

References [1] [2] [3]

[4]

104

Centers for Disease Control and Prevention, Prevention and control of nosocomial pulmonary aspergillosis. Guidelines for Prevention of Nosocomial Pneumonia. MMWR 46 pp. 58-62, 1997.

Astagneau, P. y Gambotti, L., Infecciones nosocomiales. EMCTratado de Medicina, 11 (2), pp.1-5, 2007. ttp://dx.doi.org/10.1016/S1636-5410(07)70640-2 Iwen, P.C., Davis, J.C., Reed, E.C. et al., Airborne fungal spore monitoring in a protective environment during hospital construction, and correlation with an outbreak of invasive aspergillosis. Infection Control and Hospital Epidemiology, 15 (5), pp. 303-306, 1994. http://dx.doi.org/10.2307/30146558 http://dx.doi.org/10.1086/646916 Martínez-Hernández, J., Manual de higiene y medicina preventiva hospitalaria. Madrid: Diaz de Santos, 2006.

Sanz-Calcedo & Monzón-González / DYNA 81 (188), pp. 100-105. December, 2014.

[5] [6] [7] [8]

[9]

[10]

[11]

[12] [13]

[14]

[15]

[16]

[17] [18] [19]

[20]

[21]

[22]

Albarca, S., Candau, A., Cisneros, J.M. et al., Manual para la prevención y control de la legionelosis, aspergilosis y tuberculosis en instalaciones sanitarias. Junta de Andalucía, España, 2002.

Generalitat de Catalunya, Prevenció de la infección nosocomial relacionada amb el desenvolupament d´obres als centres sanitaris.

Barcelona: Departamento de Salud de Pública, España, 2007.

Lago, J. and Falcón, D., El taller de hospitales en obras: Una experiencia de participación. Gestión y Evaluación de Costes Sanitarios, 3 (4), pp. 417-421, 2002.

Gaspar, C., Mariano, A., Cuesta, J. et al., Brote de micosis pulmonar invasiva en pacientes hematológicos neutropénicos en relación con obras de remodelación. Enfermedades Infecciosas y Microbiología Clínica, 17 (3), pp.113-118, 1999.

Goebes, M.D., Baron, E.J., Mathews, K.L., et al., Effects of buildings constructions in a hospital. Infection Control and Hospital Epidemiology, 29, pp.

462-464, 2008. http://dx.doi.org/10.1086/587189 Bouza, E., Peláez, T., Pérez-Molina, J. et al., Demolition of a hospital building by controlled explosión: The impact on filamentous fungal load in internal and external air. Journal of Hospital Infection, 52, pp.

234-242, 2002. http://dx.doi.org/10.1053/jhin.2002.1316 Dettenkofer, M., Seegers, S., Antes, G. et al., Does the architecture of hospital facilities influence nosocomial infection rates? A systematic review. Infection Control and Hospital Epidemiology, 25, pp. 21-25, 2004. http://dx.doi.org/10.1086/502286 Sánchez-Payá, J., Prevención de la aspergilosis nosocomial. Revista Iberoamericana de Micología, 17, pp. 100-102, 2000.

Gangneux, J.P., Adjide, C.C., Bernard, L. et al., Quantitative assessment of fungal risk in the case of construction works in healthcare establishments: Proposed indicators for the determination of the impact of management precautions on the risk of fungal infection. Journal of Mycologie Medicale, 22 (1), pp. 64-71, 2012. http://dx.doi.org/10.1016/j.mycmed.2012.01.003 Falvey, D.G. and Streifel, A.J., Ten-year air sample analysis of Aspergillus prevalence in a university hospital. Journal of Hospital Infection, 67, pp.

35-41, 2007. http://dx.doi.org/10.1016/j.jhin.2007.06.008 Lee, L., Berkheiser, M., Jiang, Y. et al., Risk of bioaerosol contamination with Aspergillus species before and after cleaning in rooms filtered with high-efficiency particulate air filters that house patients with hematologic malignancy. Infection Control and Hospital Epidemiology, 28, pp.

1066-1070, 2007. http://dx.doi.org/10.1086/519866 Panagopoulou, P., Filioti, J., Petrikkos, G. et al., Environmental surveillance of filamentous fungi in three tertiary care hospitals in Greece. Journal of Hospital Infection, 52, pp. 185-191, 2002. http://dx.doi.org/10.1053/jhin.2002.1298 Molina, J., Bolaños, M. y Santandreu, E., Baja incidencia de aspergilosis invasiva en un área hospitalaria en obras. Medicina Clínica, 127 (1), pp. 127-595, 2006.

Curtis, L., Cali, S., Conroy, L. et al,. Aspergillus surveillance project at a large tertiary-care hospital. Journal of Hospital Infection, 59, pp.

188-196, 2005. http://dx.doi.org/10.1016/j.jhin.2004.05.017 López, F., Cuadros, F., Segador, C., Ruiz, A., García Sanz-Calcedo, J. et al., Edificio PETER. Un ejemplo de construcción bioclimática y de integración de energías renovables. Dyna Ingeniería e Industria, 86 (2), pp. 212-221, 2011. http://dx.doi.org/10.6036/3911 Simmons, B.P., Guidelines for the prevention and control of nosocomial infections. Guidelines for hospital environmental control. Infection Control and Hospital Epidemiology, 11, pp. 183199, 1983.

Robles, M., Dierssen, T., Llorca, F.J. et al., Prevención de la infección nosocomial de origen fúngico: verificación de la bioseguridad ambiental en quirófanos. Revista Clínica Española, 205 (12), pp. 601-606, 2005. http://dx.doi.org/10.1016/S00142565(05)72653-9 Instituto Nacional de la Salud, Recomendaciones para la vigilancia, prevención y control de Infecciones en Hospitales en Obras. Grupo de trabajo de la Sociedad Española de Medicina Preventiva, Salud Pública e Higiene e Insalud. Madrid: Insalud, 2000.

[23] Ruiz-Camps, I., Aguado, J.M., Almirante, B. et al., Recomendaciones sobre la prevención de la infección fúngica invasora por hongos filamentosos de la Sociedad Española de Enfermedades Infecciosas y Microbiología Clínica, 28 (3), pp. 28172, 2010.

[24] Bradley, V., Watts-Douglas, V.C., Brian, S. and Peter-Mills, D. Developing unique engineering solutions to improve patient safety.

Journal of Healthcare Engineering, 3 (3), pp. 431-442, 2012. http://dx.doi.org/10.1260/2040-2295.3.3.431 [25] Mulá, F. and Valerí, J.M., El proyecto y el estudio de seguridad y salud. Control de infecciones por obras en centros sanitarios.

Asociación Española de Ingeniería Hospitalaria. Madrid, España, 2005.

[26] Morillo-García, A., Bioseguridad ambiental en la ejecución de obras hospitalarias. Anuario. Madrid: Asociación Española de Ingeniería Hospitalaria, 2013.

[27] García Sanz-Calcedo, J., López-Rodríguez, F. and Cuadros, F., Quantitative analysis on energy efficiency of health centers according to their size. Energy and Buildings, 73, pp. 7-12, 2014. http://dx.doi.org/10.1016/j.enbuild.2014.01.021 [28] García Sanz-Calcedo, J., Trejo, P., Rubio, B. et al., Proyecto y ejecución de obras hospitalarias: protocolos de bioseguridad y salud en zonas en obras. Medicina Preventiva, 13, pp. 159-160, 2007.

[29] Monzón-González, P., García Sanz-Calcedo, J., Fernández-Tardío, F.D. y Domínguez-Serrano, M.V. Gestión de la bioseguridad ambiental en fase de proyecto de obras en recintos hospitalarios.

Todo Hospital, 277, pp. 9-20, 2012.

[30] Public Health Agency of Canada. Construction-related nosocomial infections in patients in health care facilities. Decreasing the risk of aspergillus, legionella and other infections. Canada Communicable Disease Report, v. 27S2. Ottawa, Canada, 2001.

[31] Llevot, J., Una introducción a la evaluación de la ventilación en quirófanos. Todo Hospital, 262, pp. 759-763, 2009.

[32] Benítez, J.M., Bioseguridad ambiental en un hospital con grandes obras: medidas correctoras y organización de control. Todo Hospital, 229, pp. 439-445, 2006.

J. García Sanz-Calcedo, es Dr. Ing. Industrial e Ing. Mecánico por la Universidad de Extremadura, España, MSc. en Gestión de Instituciones Sanitarias por la Universidad Autónoma de Madrid, España y Máster en Prevención de Riesgos Laborales. Ha sido alumno del programa de Alta Dirección en Instituciones Sanitarias del IESE. Formado en la empresa privada, se incorpora al área sanitaria pública en el año 1990 como Jefe de Servicio de Obras y Mantenimiento, Subdirector de Gestión y Subdirector General de Obras, Instalaciones y Equipamiento. Actualmente es Profesor Contratado Doctor en la Universidad de Extremadura, España. Como investigador, se ha especializado en infraestructuras sanitarias, participando en proyectos de investigación internacionales y dirigiendo varias tesis relacionadas con la gestión sanitaria. Ha publicado multitud de artículos y libros relacionados con la ingeniería hospitalaria y ha realizado estancias en universidades de Alemania, Polonia, Eslovenia, Rumanía, Francia, Chile y Colombia.

P. Monzón-González, es Ing. de la Edificación y Arquitecto Técnico por la Universidad de Extremadura, España. Ha trabajado en la empresa privada y desde 2002 es Jefe de Sección de Obras e Instalaciones del Servicio Extremeño de Salud. Ha investigado en asuntos relacionados con la ingeniería hospitalaria y la bioseguridad.

105

Assessing the performance of a differential evolution algorithm in structural damage detection by varying the objective function Jesús Daniel Villalba-Moralesa & José Elias Laier b a

Facultad de Ingeniería, Pontificia Universidad Javeriana, Bogotá, Colombia. jesus.villalba@javeriana.edu.co b Escuela de Ingeniería de São Carlos, Universidad de São Paulo, São Paulo Brasil. jelaier@sc.usp.br Received: December 9th, 2013. Received in revised form: March 11th, 2014. Accepted: November 10 th, 2014.

Abstract Structural damage detection has become an important research topic in certain segments of the engineering community. These methodologies occasionally formulate an optimization problem by defining an objective function based on dynamic parameters, with metaheuristics used to find the solution. In this study, damage localization and quantification is performed by an Adaptive Differential Evolution algorithm, which solves the associated optimization problem. Furthermore, this paper looks at the proposed methodology’s performance when using different functions based on natural frequencies, mode shapes, modal flexibilities, modal strain energies and the residual force vector. Simple and multiple damage scenarios are numerically imposed on truss structures to assess the performance of the proposed methodology. Results show that damage scenarios can be reliably determined by using the analyzed objective functions.

However, the methodology does not perform well when the objective function based on natural frequencies and modal strain energies is employed.

Keywords: damage detection; metaheuristics; optimization; dynamic parameters; differential evolution.

Valoración del desempeño de un algoritmo de evolución diferencial en detección de daño estructural considerando diversas funciones objetivo Resumen Detección de daño estructural es actualmente un importante tema de investigación para diferentes comunidades en ingeniería. Algunas de las metodologías de detección de daño reportadas en la literatura formulan un problema de optimización mediante una función objetivo basada en la respuesta dinámica de la estructura y el uso de metaheurísticas para resolverlo. En este estudio, la localización y cuantificación del daño se realiza utilizando un algoritmo de evolución diferencial con parámetros adaptativos. El desempeño de la metodología propuesta es evaluado utilizando diversas funciones objetivo basadas en frecuencias naturales, formas modales, flexibilidad modal, energía de deformación modal y el vector de fuerza residual. Escenarios de daño simple y múltiple son simulados para estructuras de tipo armadura con el objetivo de determinar el desempeño de la metodología propuesta. Los resultados muestran que el algoritmo utilizado puede determinar confiablemente los escenarios buscados para varias de las funciones objetivo utilizadas. Sin embargo, no se obtuvo buenos resultados cuando se utilizó la función basada en frecuencias naturales y energía de deformación modal.

Palabras clave: detección de daño; metaheurísticas; optimización parámetros dinámicos; evolución diferencial.

1. Introduction Metaheuristics are a set of computational techniques that help find suitable solutions for optimization problems reasonably quickly. These computational techniques are characterized by the ease of their design and implementation [1]; they are primarily used to tackle complex optimization

problems that do not count on a specific algorithm for their solution, such as the damage detection problem. When it comes to finding a solution for this type of problem, metaheuristics often prove to be the most effective tool [2].

Using metaheuristics to solve this problem offers several advantages over other, more conventional optimization technique—the ability to find a global solution, lower

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 106-115. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41105

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

dependence on the initial solution, less sensitivity to noise in measurements and freedom from computing derivatives to guide the search [3]. On the flip side, metaheuristics´ drawbacks include the definition of their own parameters, e.g. crossover and mutation rates for genetic algorithms, a process generally carried out by trial and error.

Previous studies have demonstrated the potential of using metaheuristics for structural damage detection [3-8].

However, no study has defined the best metaheuristic to be used. Consequently, determining the performance of metaheuristics to detect damage proves to be a very interesting topic of research. One of the most promising metaheuristics is the Differential Evolution Algorithm (DE), which solves real-value optimization problems by evolving a population of possible solutions. Only a few damage detection methodologies, such as [7], rely on the DE approach. That proposal required analysis in sub-domains to improve the methodology performance. As damage representation implies one optimization variable for each structural element, the division into sub-domains reduced the number of optimization variables for this problem. In addition to damage identification, DE has been put to use in the field of structural engineering, such as structural identification [9] and structural optimization [10]. The successful application to the aforementioned type of problem turns DE into an appealing alternative for damage identification.

Using DE entails the definition of a few parameters: population size (NP), crossover rate (CR) and amplification factor (AF), all three of which control the evolutionary process. While papers that propose methodologies to control DE parameters abound in the literature, such as reference [11,12], there is not sufficient research that focuses on metaheuristics that do not depend on their parameters for damage detection.

This paper is aimed at addressing this lack by evaluating the performance of an adaptive differential algorithm to detect structural damage. A simple method of adaptation is used to avoid defining DE parameters by trials, leaving NP as the only uncontrolled parameter. A second main contribution of this paper relates to the definition of the objective function. That is to say, the possibility of selecting different dynamic parameters to form the objective function permits the determination of the best parameters for use. To that end, several objective functions are proposed, which are based on dynamic parameters such as natural frequencies, modal flexibilities, modal strain energies and residual force vectors. Methodologyâ&#x20AC;&#x2122;s performance is assessed in two parts. First, it is applied to the detection of simulated simple and multiple damage scenarios in truss structures. Then, the results for each function are compared in terms of the differences between computed and simulated damage scenarios to choose the parameters best suited for the damage detection problem in trusses.

Begin 1.

Generate initial vectors.

2.

Compute cost of each vector.

3.

Run from i=1 to NP - Define the i-th vector in the population as Target Vector.

-Create Trial Vector.

-Compute cost of Trial Vector.

-Compare Trial and Target Vectors; choose the best one.

4.

Verify convergence criteria. If unmet, repeat from Step 2. Otherwise, go to Step 5.

5.

Show best solution vector in population.

End Figure 1. Differential Evolution Algorithm.

Source: The authors

The algorithm iteratively modifies a set of solutions (population) by means of specific operators to find the best solution. Each solution in the population is called a solution vector. A value for NP must be defined in Step 1, with initial solutions either heuristically or randomly generated.

The quality (read: cost) of each initial solution (Step 2) is quantified by using the objective function. For a maximization problem, the higher an objective functionâ&#x20AC;&#x2122;s value, the higher the quality of the solution. Step 3 compares the cost of a Target Vector (selected from the current population) and a new vector, referred to as a Trial Vector. The vector with the highest cost will form part of the population in the next iteration; this process is carried out for each vector in the current population, one vector at a time. After all vectors in the population have been exhausted, DE moves on to Step 4, which consists of verifying the convergence criteria. If these criteria are not reached, Step 3 is carried out again. Otherwise, the algorithm advances to the next step. Finally, the solution to the problem is revealed.

The creation of trial vectors is laid out below: First, a mutation vector is created from the variation of the current best vector [14]: (1) where I is the position of the target vector in the population, xbest is the best solution vector in the current iteration t. AF is a DE parameter that allows for control of the variation amplitude introduced by Vamp, with Vamp being a quantity vector computed as [15] (2)

2 Adaptive Differential Evolutionary Algorithm De [13] is an algorithm used to solve optimization problems involving continuous domains. The process of DE for optimization problems is summarized in Fig. 1.

where xk are vectors in the current population, which are randomly chosen. Not only must these xk vectors be different from each other, but they must also be different from the current best solution vector.

107

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

Next, the i-th trial vector can be computed using the binary crossover [14]: ,

(3)

,

Each term in the trial vector is assigned as , ,

(4)

,

where tg is the target vector, nrj is a random number generated for each position j of the i-th vector, which ranges between 0 and 1. The latter parameter controls the role played by both mutation and target vectors in defining the trial vector.

Values for cr and af must be previously assigned before employing the DE. An iterative process that sets DE parameters is frequently relied on to define the most appropriate values. To avoid this process, the present study gives each solution vector its own cr and af parameters.

These values can evolve throughout the DE´s execution, and their initial values are randomly generated between for af [0.7-0.9] and cr [0.4-0.6]. Far from arbitrary, these ranges are chosen because they account for the recommended values of 0.8 and 0.5 for af [15] and cr [16], respectively. cr and af values for the iteration t+1 can be computed as (5)

of a reduction in the stiffness matrix of the damaged element by means of a stiffness reduction factor β. A value equal to 0 for this factor indicates no damage and 1 total damage.

Optimization aims to arrive at the correct set of β factors, i. e., those which allow the dynamic parameters of an updated FEM to match those measured experimentally for the current structure. Each possible solution vector corresponds to a different damage scenario, and each position in the vector represents the β factor of one structural element. For an undamped structure with a mass matrix M and a stiffness matrix K, the modal parameters are given by (7) where ωj is the natural frequency corresponding to the jth mode shape ϕj. Thus, any change in the stiffness or mass matrix will lead to changes in the dynamic parameters.

Step 2 experimentally determines the structure’s dynamic behavior. As this is a numerical study, the current dynamic parameters are obtained by introducing the simulated damage scenario into the undamaged structure’s FEM. Then, the eigen-value problem, given in Eq. 7, is solved. These parameters are numerically perturbed to simulate noise in a real measurement. Such perturbations equal 1% for the natural frequencies (noiseω) and 3% for the mode shapes (noiseΦ) [18]. The equations that introduce noise in the modal parameters are [3]: (8)

(6) where best corresponds to the best vector in the current population. Ri,1 and Ri,2 are random numbers between 0 and 1 generated for solution vector i.

Concerning the definition of NP, Storn and Price [15] proposed a size between 5 and 10 the number of design variables. However, NP definition utilized in this research will be explained in Section 4.

3. Proposed damage detection methodology The methodology employed throughout the present paper (Fig. 2) can be consulted in [17], with one notable exception: for our purposes, the genetic algorithm has been replaced by a DE. This methodology assumes the existence of a finite element model (FEM) that is able to accurately represent the undamaged condition (Step 1). Damage takes the form Begin 1.

Define the finite-element model of initial structure.

2.

Determine dynamic parameters of current structure.

3.

Define the objective function.

4.

Apply DE algorithm 5.

Show damage scenario found.

End Figure 2. Damage Detection Methodology.

Source: Adapted from Villalba, J. D. and Laier, J. E.2012

where Rand is a random number between -1 and 1.

Letter n means a parameter with noise.

Armed with the available modal data, an objective function must be defined (Step 3), which generally has the following form: (9) where curr_dyn_resp and upd_dyn_resp refer to the dynamic parameters for the current structure and updated model, respectively. A detailed version of the objective functions relied on to carry out this research appears in Section 3.1.

Step 4 applies the DE described in Section 2 in tandem with the heuristic shown in Section 3.2. In the same vein, convergence criteria are presented along with the numerical example because the maximum number of iterations depends on the structure size.

Finally, the computed damage scenario is reported in the form of stiffness reduction factor (β) value for each of the 61 elements in the structure.

3.1. Objective Functions In this study, the damage detection problem was treated as a maximization one that is equivalent to the minimization formulation in Eq 9. The objective functions below were those used for result comparison.

108

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.





3.2. Heuristic for the generation of the initial population vector

Natural Frequencies and Mode Shapes (10):

To accelerate DE convergence, the initial population was generated heuristically based on two characteristics assumed for the damage scenarios. First, damaged element quantity is considered low in light of its relation to the total number of elements in the structure [19]. Second, damage extent is not expected to be severe. The proposed heuristic generates a random number between 0 and 1 for each βi factor in each initial solution vector. If the random number is smaller than 0.5, then βi takes a random value between 0.1 and 0.5; otherwise, βi is set to zero.

Natural Frequencies and Strain Modal Energies (11)

4. Numerical Examples 

Residual Force Vector (12)



Modal Flexibility (13)

The symbols used in the above functions are: nm is the number of modes available, c1,2 are constants defined as 200 and 1, respectively, ωj is the j-th natural frequency, ϕij is a value for the i-th degree of freedom of the j-th mode shape, dea is a value from the DE solution and ex is experimental information, ngll is the number of degrees of freedom of the structure, W1,2=2.0 weight factor, und refers to the undamaged structure, msej is the modal strain energy for the j-th mode shape and MFij is a value corresponding to the position (i,j) of the modal flexibility.

The terms MSEj, RFV and MF are given by (14) (15) (16) All functions are based on the difference between the dynamic parameters computed for the FEM of a specific vector solution –possible damage scenario– and those obtained experimentally. Thus, each possible damage scenario in the population displays a given probability of being the correct one. The objective function is put into use one at a time, so the performance of the proposed methodology using this function can be measured.

The proposed methodology was applied to detect damage in truss structures with different configurations (Fig. 3). The FEM uses conventional 2D bar elements assuming perfectly pinned joints. Vertical and horizontal elements have a length equal to 1.0m. Elements have elastic modulus E=200 x 10E9 N/m2, density ρ=7800 kg/m3 and cross-sectional area A=0.001 m2. Information regarding the first eight natural frequencies and complete mode shapes of the damaged structure were considered available.

Applying the proposed methodology to a total of 21 simulated damage scenarios (see Tables 1-4, where DS is a damage scenario identifier) makes it possible to evaluate the methodology’s performance. Initially, results for the 61element truss will be used to compare how well the four proposed objective functions work.

Tables 1 and 2 include identified damaged elements and damage extents (βi) for simulated damage scenarios. Scenarios seen in Table 1 correspond to damage present in only one element, whether the damage be in a vertical (DS-1, DS-4 and DS-7), horizontal (DS-2, DS-5, DS-8) or diagonal (DS-3, DS6, and DS-9) element. Multiple damaged elements are considered in scenarios DS-10, DS-11 and DS-12 (Table 2).

The focus of the results will be on the correct identification of damaged elements and the presence of misidentified elements.

The objective functions producing the best damage identification are then used to detect damage in the other trusses. Simulated damage scenarios for these trusses are shown in Table 3-5, which include damage located in a specific zone of the truss or spread out along the structure. The reason for using only cases of multiple damage scenarios is brought up later. In this order of ideas, more detail-oriented analysis for these results will be undertaken in section 5.

Table 1.

Simulated simple damage scenarios for the 61-element truss structure.

Damage Scenario Damaged Element Damage Extent (β) DS-1 3 0.15 DS-2 24 0.20 DS-3 52 0.40 DS-4 6 0.20 DS-5 33 0.20 DS-6 55 0.20 DS-7 10 0.20 DS-8 27 0.20 DS-9 50 0.20 Source: The authors.

109

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

Figure 3. Analyzed structures. a) 15, b) 45, c) 47 and d) 61 element trusses.

Source: Adapted from Villalba, J. D. and Laier, J. E.2012.

Table 2.

Simulated multiple damage scenarios for the 61-element truss structure.

DS-10 DS-11 DS-12 Elem βi Elem βi Elem βi 3 0.150 4 0.250 3 0.170 28 0.150 15 0.200 10 0.210 52 0.150 18 0.230 14 0.140 24 0.220 17 0.180 34 0.260 37 0.200 52 0.200 Source: The authors.

Table 5.

Simulated multiple damage scenarios for the 47-element truss structure.

T19 T20 T21 Elem βi Elem βi Elem βi 3 0.26 4 0.24 2 0.25 8 0.22 10 0.28 10 0.25 13 0.31 14 0.25 Source: The authors.

optimization variables. Yet, the relationship between NP and structure size has not quite been elucidated for the damage detection problem. As a result, NP was set at 200 after a process of trial and error. The criterion for halting the algorithm’s execution is either reaching a maximum of 200 iterations or 50 iterations without significant changes in the best current vector’s cost. 30 executions were done for each example studied to ensure the proposed methodology reliably detects the real damage scenario. The computed damage scenario that corresponds to the run with the highest cost solution is then reported to the user.

Table 3.

Simulated multiple damage scenarios for the 15-element truss structure.

DS-13 DS-14 DS-15 Elem βi Elem βi Elem βi 3 0.26 4 0.24 2 0.25 8 0.22 10 0.28 13 0.25 13 0.31 15 0.25 Source: The authors.

Table 4.

Simulated multiple damage scenarios for the 45-element truss structure.

T16 T17 T18 Elem βi Elem βi Elem βi 5 0.32 1 0.23 7 0.30 26 0.29 13 0.31 23 0.30 38 0.24 32 0.26 35 0.30 41 0.18 41 0.30 Source: The authors.

5. Results and Discussion

In addition to the structure’s physical characteristics, terms related to the DE execution require definition. All DE parameters were defined for the 61-element structure before being brought to bear on the other structures. First, observation shows that NP depends on the number of

Results for cases with only one damaged element in the 61-element structure are shown in Figs. 4-6. The methodology detects the real damaged element and damage extent accurately. Functions F1 and F4 provided the best results across all cases analyzed. They properly gauged damage extent, with the difference between real and simulated damage less than 0.08. Moreover, few elements were misidentified.

The function based on the residual force vector (F3) misidentified elements on numerous occasions. Naturally, these errors translate into low methodology reliability when function F3 is used. For all cases, reliance upon an objective function based on natural frequencies and modal strain energy left the real damaged element undiscovered.

110

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

Figure 4. Results for Damage Scenario DS-1.

Source: The authors.

Figure 5. Results for Damage Scenario DS-2.

Source: The authors.

Figure 6. Results for Damage Scenario DS-3.

Source: The authors.

Table 6 presents methodology performance when an energy-based objective function was used to detect simple damage in the 61-element truss. Results evince the low reliability when using Function F2, for there is a significant gap between real and computed damage extents, as well as many misidentified damage elements. Table 7 confirms this low reliability by displaying the best five runs for a scenario

wherein Element 50 is damaged (only βi factors ≥ 0.05 are shown): not a single computed scenario included the real damaged element. The algorithm could not distinguish between completely different solutions - scenarios with varying damaged elements and damage extents - and converge on any of these solutions. Therefore, the results for the detection of multiple damage scenarios using function F2 will not be shown.

Table 6.

Performance of the proposed methodology to detect simple damage using an objective function based on strain modal energy.

DS-4 DS-5 DS-6 DS-7 DS-8 Element βreal βcomp Element βreal βcomp Element βreal βcomp Element βreal βcomp Element βreal 2 --0.54 3 --0.10 2 --0.06 10 0.20 0.05 2 --6 0.20 0.04 8 --0.05 11 --0.53 12 --0.60 4 --12 --0.51 11 --0.53 12 --0.59 34 -0.09 7 --36 --0.41 33 0.20 0.00 31 --0.1 12 --36 --0.49 39 --0.07 14 --44 --0.35 55 0.20 --27 0.20 56 --0.10 36 --44 --52 --60 --Source: The authors.

111

βcomp 0.07 0.05 0.15 0.57 0.09 --0.37 0.15 0.09 0.09

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

Table 7.

Comparison of different solutions found for damage scenario DS-9 with the objective function based on strain modal energy.

Element 1 2 7 8 11 13 25 34 35 36 47 50 βreal ----------------------0.20 βsol1 ------0.05 --0.06 ------0.40 ----βsol2 ----------------0.37 --0.35 --βsol3 --0.05 0.06 0.08 0.48 --0.39 0.03 --------βsol4 --------0.50 ----0.06 --------βsol5 0.09 0.19 0.22 0.17 0.35 --0.41 0.03 --------Source: The authors.

Results obtained for simulated multiple damage scenarios in the 61-element structure are shown in Figs.

7-9. The algorithm detected all real damaged elements when objective functions F1, F3 and F4 were used. With respect to the damage extent, the difference between computed and real extent is less than 0.1 for every element in all three scenarios. As was the case for simple damage scenarios, objective function F3 produced the most misidentified elements. Though, as previously

59 ------0.05 0.39 ---

60 --0.39 --0.12 --0.07

Cost --1350.4 1343.8 1329.3 1325.8 1319.3

mentioned, most functions included misidentification, Function F1 identified only the real damaged elements for Scenario DS-10 (Fig. 7). Function F3 misidentified 40% of the elements for scenario DS-11, and seven of these misidentified elements exhibited a damage extent higher than 0.1. Therefore, the logical conclusion is that objective function F3 is less reliable than the others. For scenario DS-12, Function F4 produced the best results, with an average difference of 0.047 between simulated and real damage extent for all real damage elements.

Figure7. Results for Damage Scenario DS-10.

Source: The authors.

In order to more thoroughly understand the methodology’s effectiveness in terms of computing real damage extent, readers are guided to Tables 8-10, which lay out the differences between real (βreal) and computed (βcomp) damage extent. The value denoted in parentheses indicates such difference in terms of percentage. Looking at these results, it can be inferred that as the number of damaged elements increases, error extent also increases.

For example, when using Function F4, the average error was 11, 24 and 25 % for 3, 5 and 6 damaged elements,

respectively. In spite of this fact, results for Functions F1 and F4 do not display a significant difference between the computed and real damage extent. Function F2 presents an average error of 19% for the three damage scenarios.

Then there is the case of Function F4, which underestimated the damage extent for element 14 in scenario DS-12. It is important to recall that element 14 had a low real damage value. In this sense, dynamic parameter changes caused by low damage values can be masked by noise in the measurements.

Figure 8. Results for Damage Scenario DS-11.

Source: The authors.

112

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

Figure 9. Results for Damage Scenario DS-12.

Source: The authors.

Table 8.

Comparing the real and computed damage extent for scenario DS-10 using functions F1, F3, F4. βcomp Element βreal F1 F3 F4 0.10 (33) 0.18 (20) 0.15 (0) 3 0.15 0.16 (7) 0.13(13) 0.13 (13) 28 0.15 52 0.15 0.15(0) 0.18(20) 0.18 (20) Source: The authors.

Table 9.

Comparing the real and computed damage extent for scenario DS-11 using functions F1, F3, F4. βcomp Element βreal F1 F3 F4 4 0.25 0.31 (24) 0.20 (20) 0.21 (16) 0.16 (20) 0.26 (30) 0.15(25) 15 0.2 18 0.23 0.18 (22) 0.25 (9) 0.22 (4) 24 0.22 0.17 (23) 0.20 (9) 0.12 (45) 34 0.23 0.28 (22) 0.28 (22) 0.16 (30) Source: The authors.

Table 10.

Comparing the real and computed damage extent for scenario DS-12 using functions F1, F3, F4. βcomp Element βreal F1 F3 F4 0.18 (6) 0.03 (82) 0.21(24) 3 0.17 0.11(48) 0.20 (5) 0.26 (24) 10 0.21 0.15 (7) 0.18 (29) 0.04 (71) 14 0.14 0.18 (0) 0.19 (6) 0.15 (17) 17 0.18 0.18 (10) 0.18 (10) 0.17 (15) 37 0.2 52 0.2 0.16 (20) 0.17 (15) 0.20 (0) Source: The authors.

Table 11 shows the number of executions for which the algorithm found all real damaged elements. Here, one should keep in mind that the proposed DE performs 30 runs to detect each damage scenario. Based on the results of this study, each objective function possesses a different reliability level; however, it is safe to say objective function F4 (modal flexibility) was the group’s weak link. Despite the high success rate of detecting real damaged elements with F3, results for this function also presented many misidentified elements. On the whole, simple damage scenarios were detected more reliably than multiple damage

scenarios, regardless of the objective function employed. In Table 11.

DE Performance when identifying all real damaged elements.

Damage Number of successful executions Scenario F1 F3 DS-1 DS-2 DS-3 DS-4 DS-5 DS-6 Source: The authors.

30 30 30 15 19 7

23 29 30 15 18 19

F4 26 25 18 8 8 8

response to this fact, the performance assessment for the other trusses entailed multiple damage scenarios.

As the function F3 reported many misidentified elements, results for the other trusses will only include Functions F1 and F4. Tables 12-14, 15-17 and 18-20 depict the application of the proposed methodology for 15, 45, 47-element trusses, respectively. Reported damage scenarios only include those elements presenting β 0.05. As previously mentioned, the percentage difference between the computed and real damage extent is placed in parentheses: results show that the methodology precisely computes damage extent for both functions. The exact scenario DS-13 was obtained with Function F1 for the smallest truss structure. It is worth mentioning that the search space correlates to the number of elements in the truss structure; thus, as the number of elements increases, the search space increases, adding another level of complexity to the optimization process.

Detecting element 2 in Scenario DS-15 was difficult for function F4, exposing the difficulty to detect damage in some elements. While the average error for all the scenarios is 4.2 and 14.3 for function F1 and F4, respectively. Function F4 means more misidentified elements compared with the results obtained using F1.

These elements presented low damage extent for both functions in most cases. However, the existence of misidentified elements with high damage values is possible, i. e., Element 5 in Scenario DS-17 as computed with Function F1 or Element 10 in Scenario DS-15 as computed with Function F4. Overall, F1 and F4 produce similar results, albeit with greater reliability in the case of F1.

113

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

Table 12.

Results for damage scenario DS-13 using functions F1 and F4. βcomp Element βreal F1 F4 0.26 (0) 0.23 (12) 3 0.26 8 0.22 0.22 (0) 0.20 (9) Source: The authors.

Table 17.

Results for damage scenario DS-18 using functions F1 and F4. βcomp Element βreal F1 1 --0.09 7 0.30 0.27 (10) 0.29 (3) 23 0.30 --28 --35 0.30 0.30 (0) 41 0.30 0.30 (0) Source: The authors.

Table 13.

Results for damage scenario DS-14 using functions F1 and F4. βcomp Element βreal F1 F4 1 ----0.07 0.24 (0) 0.21 (13) 4 0.24 10 0.28 0.28 (0) 0.17 (39) 13 0.31 0.31 (0) 0.25 (17) 14 ----0.28 Source: The authors.

F4 0.28 (7) 0.29 (3) 0.05 0.31 (3) 0.30 (0)

Table 18.

Results for damage scenario DS-19 using functions F1 and F4. βcomp Element βreal F1 F4 3 ----0.05 --0.06 4 --13 0.19 0.20 (5) 0.18 (5) 15 ----0.13 22 --0.06 --25 0.24 0.24 (0) 0.22 (8) 28 0.20 0.19 (5) 0.15 (25) 29 --0.06 --30 --0.06 --34 --0.06 --Source: The authors.

Table 14.

Results for damage scenario DS-15 using functions F1 and F4. βcomp Element βreal F1 F4 2 0.25 0.26 (4) 0.08 (68) --0.17 10 --13 0.25 0.25 (0) 0.39 (56) 15 0.25 0.26 (4) 0.28 (12) Source: The authors.

Table 19.

Results for damage scenario DS-20 using functions F1 and F4. βcomp Element βreal F1 F4 4 ----0.05 9 0.32 0.31(3) 0.33 (3) 11 ----0.05 0.27 (4) 0.29 (4) 14 0.28 --0.06 15 ----0.05 27 --30 --0.05 --34 0.22 0.25 (14) 0.23(5) 37 0.26 0.26 (0) 0.21(19) Source: The authors.

Table 15.

Results for damage scenario DS-16 using functions F1 and F4. βcomp Element βreal F1 F4 2 ----0.14 4 ----0.06 5 0.32 0.35(9) 0.18 (44) 11 ----0.07 0.29 (0) 0.29 (0) 26 0.29 --0.05 35 --38 0.24 0.24 (0) 0.25 (4) Source: The authors.

Table 20.

Results for damage scenario DS-21 using functions F1 and F4. βcomp Element βreal F1 F4 2 0.18 0.19 (0) 0.18 (0) 0.05 --8 ----0.05 () 26 --27 0.18 0.19 (6) 0.13 (28) 28 0.18 0.18 (0) 0.20 (11) 29 0.18 0.17 (6) 0.16 (11) 45 ----0.05 Source: The authors.

Table 16.

Results for damage scenario DS-17 using functions F1 and F4. βcomp Element βreal F1 F4 1 0.23 0.17 (26) 0.24 (4) 4 ----0.05 5 --0.20 0.05 7 ----0.05 11 ----0.09 12 ----0.07 0.32 (3) 0.31(0) 13 0.31 32 0.26 0.28 (8) 0.23(12) 41 0.18 0.20 (11) 0.17 (6) 43 ----0.06 Source: The authors.

6. Conclusions

The aforementioned result highlights the importance of decreasing computational error in terms of damage extent and misidentified elements quantity in order to enhance methodology performance.

This paper explores the application of the Differential Evolution (DE) to the damage detection problem. DE operators stem from the relevant scientific literature, with adaptive parameters based on the evolution of those parameters belonging to the best individual in the current

114

Villalba-Moralesa & Laier / DYNA 81 (188), pp. 106-115. December, 2014.

iteration. Such adaptation can be understood as one of the key contributions of this research as the final user need only define the population size. This methodology was tested on truss-type structures. Results show that the methodology detect the different damage scenarios (i.e. identifies both damaged elements and damage extent) with varying levels of reliability, depending on the objective function used. In this study, four objective functions were used, and the most reliable results were obtained by objective function F1, which is based on natural frequencies and mode shapes.

Results for the function based on modal flexibility proved to be similar to those obtained with F1; even though the former generated more of misidentified elements. A different objective function, one based on natural frequencies and modal strain energies (F2), failed to locate the damaged element in most of the simple damage scenarios. As regards the objective function based on the residual force vector, the results were unreliable given that it led to numerous misidentified elements. As far as the ability to identify simple and multiple damage scenarios is concerned, the logical conclusion that simple damage scenarios are more reliably identified was confirmed. In conclusion, more research should be geared toward developing damage detection methodologies that find scenarios with many damaged elements.

Acknowledgements The authors would like to acknowledge CNPq (Brazilian National Council for Technological and Scientific Development) for the financial support provided to this research.

[10]

[11]

[12]

[13]

[14]

[15] [16]

[17]

[18]

[19]

References [1] [2]

[3] [4]

[5]

[6]

[7]

[8]

[9]

Talbi, E., Metaheuristics: From design to implementation. New Jersey: Ed. J. Wiley & Sons, 2009. http://dx.doi.org/10.1002/9780470496916 Navarro, R., Puris, A. and Bello, R., The performance of some metaheuristics in continuous problems studied according to the location of the optima in the search space. DYNA, 80 (180), pp. 6066, 2013.

Begambre, O., Hybrid algorithm for damage detection: A heuristic approach, Ph. D. dissertation, University of Sao Paulo, Sao Carlos, SP, Brazil, 2007.

Mares, C. and Surace, C., An application of genetic algorithm to identify damage in elastic structures. Journal of Sound and Vibration, 195 (2), pp.

95-215, 1996. http://dx.doi.org/10.1006/jsvi.1996.0416 Moradi, S., Razi, P. and Fatahi, L., On the application of bees algorithm to the problem of crack detection of beam-type structures.

Computers and Structures, 89, pp. 2169-2175, 2011. http://dx.doi.org/10.1016/j.compstruc.2011.08.020 Fadel, L., Fadel, L.F. Kaminski, J. and Riera, J.D., Damage detection under ambient vibration by harmony search algorithm.

Expert Systems with Applicatioon, 39, pp. 9704-9714, 2012. http://dx.doi.org/10.1016/j.eswa.2012.02.147 Casciati, S., Stiffness identification and damage localization via differential evolution algorithms. Structural Control and Health Monitoring, 15, pp.

436-449, 2008. http://dx.doi.org/10.1002/stc.236 Sandesh, K., Shankar, K., Application of a hybrid of particle swarm and genetic algorithm for structural damage detection. Inverse Problems in Science and Engineering, 18 (7), pp. 997-1021, 2012. http://dx.doi.org/10.1080/17415977.2010.500381 Kyprianou, K. and Worden K., Identification of hysteretic systems sing the differential evolution algorithm. Journal of Sound and

Vibration, 248 (2), pp.

289-314, 2001. http://dx.doi.org/10.1006/jsvi.2001.3798 Patrick, H. Tinker, M. and Dozier, G., Evolutionary optimization of a geometrically refined truss. Structural Multidisciplinary.

Optimization, 31, pp.

311-319, 2006. http://dx.doi.org/10.1007/s00158-005-0564-7 Zhang, J. and Sanderson A.C., JADE: Adaptive differential evolution with optional external archive. IEEE Transactions on Evolutionary Computation, 13 (5), pp. 945-958, 2009. http://dx.doi.org/10.1109/TEVC.2009.2014613 Brest, J. and Maucec, M.S., Self-adaptive differential evolution algorithm using population size reduction and three strategies. Soft Computing, 15, pp.

2157-2174, 2011. http://dx.doi.org/10.1007/s00500-010-0644-5 Storn, R. and Price, K.V., Differential evolution- A simple and efficient adaptive scheme for global optimization over continuous spaces. Technical Report TR-95-012. Berkeley, USA: Intern.

Computer Science Institute, 1995.

Storn, R. and Price, K.V., Differential evolution- A simple and efficient heuristic for global optimization over continuous spaces, Journal of Global Optimization, 11 (4), pp. 341-359, 1997. http://dx.doi.org/10.1023/A:1008202821328 Tvridk, J., Differential evolution: competitive setting of control parameters. Proceedings of the International Multiconference on Computer Science and Information Technology, pp. 207-213, 2006.

Ali, M.M. and Torn, A., Population set-based global optimization algorithms: some modifications and numerical studies. Computers and Operation. Research, 31, pp. 1703-1725, 2004. http://dx.doi.org/10.1016/S0305-0548(03)00116-3 Villalba, J.D. and Laier, J.E., Localising and quantifying damage by means of a multi-chromosome genetic algorithm. Advances in Engineering Software, 50, pp.

150-157, 2012. http://dx.doi.org/10.1016/j.advengsoft.2012.02.002 Perera, R. and Torres, R., Structural damage detection via modal data with genetic algorithms. Journal of Structural Engineering, 132 (93), pp. 1491-1501, 2006. http://dx.doi.org/10.1061/(ASCE)07339445(2006)132:9(1491) Raich, A. and Liszkai, T., Improving the performance of structural damage detection methods using advanced genetic algorithms.

Journal of Structural Engineering, 133 (3), pp. 449-461, 2007. http://dx.doi.org/10.1061/(ASCE)0733-9445(2007)133:3(449)

J.D. Villalba-Morales, received his BSc. Eng in Civil Engineering in 2005 at the Industrial University of Santander, Bucaramanga, Colombia. He obtained his MSc and Dr.in 2007 and 2012 at the University of Sao Paulo, Brazil. After Sao Paulo, Dr. Villalba began to work as an assistant professor in the Pontificia Universidad Javeriana, Bogotรก, Colombia, from 2013 to the present, where he teaches courses on numerical methods for engineers as well as the finite element method. His research interests include structural health monitoring, earthquake engineering, steel structures, optimization and computational intelligence technique application and development.

J.E. Laier, received his BSc. Eng in Civil Engineering in 1971 at the University of Sao Paulo, Brazil, the university from which he also obtained his MSc. In 1975 and Dr. in Structural Engineering in 1978. He continued research as a PhD at The Computational Mechanics Center, Southampton, UK. Currently, he teaches as a full professor at the University of Sao Paulo, Brasil. His experience in the Civil Engineering field, focused on structural mechanics, includes work on numerical analysis, structural dynamics, high-rise building analysis, damage detection and structural identification.

115

Efficient reconstruction of Raman spectroscopy imaging based on compressive sensing Diana Fernanda Galvis-Carreño a, Yuri Hercilia Mejía-Melgarejo b & Henry Arguello-Fuentes c b

a , Escuela de Ingeniería Química, Universidad Industrial de Santander. Bucaramanga, Colombia. diana.galvis1@correo.uis.edu.co Escuela de Ingenierías Eléctrica, Electrónica y de Telecomunicaciones, Universidad Industrial de Santander. Bucaramanga, Colombia. yuri.mejia@correo.uis.edu.co c Escuela de Ingeniería de Sistemas e Informática, Universidad Industrial de Santander. Bucaramanga, Colombia. henarfu@uis.edu.co

Received: December 12th, 2013.Received in revised form: March 10th, 2014.Accepted: September 25th, 2014.

Abstract Raman Spectroscopy Imaging requires long periods of time for the data acquisition and subsequent treatment of the spectral chemical images. Recently, Compressed Sensing (CS) technique has been used satisfactorily in Raman Spectroscopy Imaging, reducing the acquisition time by simultaneously sensing and compressing the underlying Raman spectral signals. The Coded Aperture Snapshot Spectral Imager (CASSI) is an optical architecture that applied effectively the CS technique in Raman Spectroscopy Imaging. The main optical element of CASSI system is a coded aperture, which can transmit or block the information from the underlying scene. The principal design variable in the coded apertures is the percentage of transmissive elements or transmittance. This paper describes the technique of CS in Raman Spectroscopy imaging by using the CASSI system and realizes the selection of the optimal transmittance values of the coded apertures to ensure an efficient recovery of Raman Images. Diverse simulations are performed to determine the Peak Signal to Noise Ratio (PSNR) of the reconstructed Raman data cubes as a function of the transmittance of the coded apertures, the size of the underlying Raman data cubes and the number of projections expressed in terms of the compression ratio.

Keywords: Raman Spectroscopy; Spectral Imaging; Compressed Sensing; Coded Aperture.

Reconstrucción eficiente de imágenes a partir de espectroscopia Raman basada en la técnica de sensado compresivo Resumen La Espectroscopia Raman de Imágenes requiere largos periodos de tiempo en la adquisición como en el tratamiento de datos para la construcción de imágenes químicas. Para reducir el tiempo se ha empleado la técnica de Sensado Compresivo (SC) gracias a la detección y compresión simultánea de las señales. El sistema de adquisición de imágenes basado en una apertura codificada (CASSI) es una arquitectura óptica que aplica de manera eficiente los conceptos de SC. El principal elemento del sistema CASSI es una apertura codificada, la cual puede ser vista como un filtro que transmite o bloquea información de una escena. El porcentaje de elementos transmisores es conocido como la transmitancia y esta es una variable de diseño. Este trabajo describe la técnica de SC aplicada a la Espectroscopia Raman de Imágenes empleando el sistema CASSI y realiza la selección de los valores óptimos de transmitancia que garantizan una eficiente reconstrucción de imágenes. Se realizaron diversas simulaciones para determinar la relación señal a ruido (PSNR) de la reconstrucción de un cubo de datos Raman como función de la transmitancia, el tamaño del cubo y el número de capturas expresadas en términos de la relación de compresión.

Palabras clave: Espectroscopia Raman; Imágenes Espectrales; Sensado Compresivo; Aperturas Codificadas.

1. Introduction Spectral imaging is a technology that can obtain the spatial map of spectral variations of a scene; these spatial maps are useful in many applications including military target discrimination, biomedical, biochemical, agriculture,

mineralogy, biophysics, environmental remote sensing, among others [1-3]. Different spectroscopic techniques can be employed to obtain this kind of images for determining relevant chemical information.

Raman Spectroscopy is currently one of the most used analytical techniques in several areas of modern science and

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 116-124. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41162

Galvis-Carre単o et al / DYNA 81 (188), pp. 116-124. December, 2014.

it is used to analyze chemical composition and construction of spectral images of different compounds [4-6]. Raman analysis presents the relevant advantage of being a noninvasive technique, not requiring the addition of chemical agents or labels for the sample identification, and it has a relative low cost compared with other spectroscopy techniques. Furthermore, there is extensive information in the literature regarding this technique [7, 8].

Currently, Raman Spectroscopy Imaging emerges as a tool to create chemical images of the distribution of the components from simultaneous measurement of spectra and spatial information. Raman chemical images can be obtained through subsequent measurements of several sample points and a reconstruction process. These images are useful to chemical identification and classification [5,9].

Despite the broad advantages and applications of this technique, it requires long periods of time for the data acquisition and subsequent treatment for spectral images.

More specifically, typically spectral detection methods requires in the order of per spectrum which is impractical for the collection of large spectral images. For example, the collection of 1 megapixel image would require or days [8]. For this reason, various studies and modifications have been made on the optical architecture of Raman Spectroscopy Imaging.

There are two general approaches to obtain Raman Images: serial and direct imaging. The serial imaging approach as point (whiskbroom spectrometer) and line (pushbroom spectrometer) Raman mapping, requires numerous spectra for reconstructing the entire Raman image at a given wavenumber. In point mapping, a laser spot is raster scanned, in two spatial dimensions with a spectrum being recorded at each position; the entire Raman spectrum is obtained at each point. For line mapping, a laser line is raster scanned along either the or axis, using a twodimensional charge-coupled device detector to collect the spectral and spatial information; the entire Raman spectrum is obtained at each line. In contrast, in direct approach (snapshot imaging spectrometer) all spatial points of the Raman image at a specific wavenumber are determined simultaneously from a single measurement of a globally illuminated sample.

Between the named methodologies, the number of data obtained during the sensing varies considerably and even in the case of direct imaging, the data processing takes considerable time for the formation of the spectral images [9,10]. Fig. 1 shows the above mentioned approaches.

Figure 1. Approaches to obtain Raman Images. a) Serial approach (per point or per line), b) Direct image approach. and represents the spatial coordinates and the spectral coordinate.

Source: [10]

An innovative alternative to reduce the time of acquisition and processing of the signals obtained from Raman Spectroscopy is to employ the Compressed Sensing (CS) technique. Different studies have shown that this technique can be used satisfactorily in Raman Spectroscopy Imaging [4, 6]. CS efficiently reduces the acquisition time by simultaneously sensing and compressing the underlying spectral signals. Instead of sensing directly the spectral signal, CS senses random projections. These projections are then used to recover the underlying Raman spectral signal by solving a minimization problem [11,12].

Different architectures for obtaining spectral images based on Raman Spectroscopy have been developed under the concepts of Compressed Sensing technique [13]. The Coded Aperture Snapshot Spectral Imager (CASSI) system is a remarkable optical architecture that effectively exploits CS principles in Raman Spectroscopy Imaging [6]. In CASSI the coded measurements captured by the detector are mathematically equivalent to compressive random projections in CS. In CASSI system, the coding is applied to the (spatialspectral) Raman image source density by means of a coded aperture as realized by the CASSI system depicted in Fig. 2, where are the spatial coordinates and is the Raman shift [14]. The resulting coded field is subsequently modified by a

Figure 2. CASSI system representation. Black elements in coded aperture are blocking elements and white are transmissive. Source: [14] 117

Galvis-CarreĂąo et al / DYNA 81 (188), pp. 116-124. December, 2014.

dispersive element before it impinges onto the detector. The compressive measurements across the detector are realized by the integration of the field over the spectral range sensitivity of the detector. The recovery of the underlying hyperspectral signal in CASSI entails solving an undetermined linear system of equations. The quality of the CS Raman reconstructions depends on the correct selection of the coded aperture patterns . For this reason, the most important component in CASSI Raman system is a set of coded apertures, which need to be properly designed taking in account variables as the percentage of transmissive elements of the coded aperture or transmittance, the Raman data cube size and the number of projections expressed in terms of the compression ratio.

Using the compressive sensing by CASSI system in Raman Spectroscopy Imaging the time of acquisition and signal processing is reduced significantly, opening a broad range of applications of this technique in different scientific areas, including biomedical, chemical, biochemical, environmental, among others.

This paper describes the technique of CS in Raman Spectroscopy Imaging by using the Coded Aperture Snapshot Spectral Imager (CASSI) and realizes the selection of the optimal transmittance values in the coded apertures needed in this system. The main contribution of this work is to design the optimal transmittance values of the coded apertures, which allow an efficient Raman image reconstruction.

Diverse simulations in MatLab are performed to determine the Peak Signal to Noise Ratio (PSNR) of the reconstructed Raman data cubes. The spatial and spectral analyses of the reconstructions allow establishing optimal values of transmittance of the coded aperture as a function of the size of the underlying Raman data cubes and the compression ratio.

2. Raman spectroscopy Raman Spectroscopy analysis is based on the study of light scattered by a material when a beam of monochromatic light is incident on the underlying material [15]. The interaction of the electric field of an electromagnetic wave with the electrons interacting with the system leads to the scattering of incident light.

The scattering process is show in Fig. 3. Most of the light scatters of the same energy of the incident beam, .

This energy light is said to be elastically or Rayleigh scattered and it does not provide information about the sample. Valuable information can be obtained from the light that changes energy upon scattering, (where is related with the energy of the sample). This light is said to be inelastically or Raman scattered. If the scattered light loses energy , it emerges at a longer wavelength and this effect is known as Raman Stokes. The light can scatter with an increase of energy, , and concomitant shorter wavelength and this is called the Raman anti-Stokes effect.

The change in the wavelength of the light (to either longer or shorter wavelength) is known as the Raman shift.

Frequency variations seen in Raman scattering phenomenon are equivalents to energy variations (Fig. 3).

Figure 3. Scheme of the Rayleigh and linear spontaneous Raman scattering.

Source: The Authors

Because of light-material interaction, the molecules may become temporarily to a virtual energy state that must return to an allowed vibrational energy level, this is done by scattering a beam of light. The frequency at which the beam is scattered depends on the energetic jump performed for the molecule. Because the energy level of a molecule depends crucially on the composition of this, the spectrum of Raman shifts is a highly specific â&#x20AC;&#x153;fingerprintâ&#x20AC;? of the internal energy level structure. As such, it can be used for extremely precise chemical detection and identification.

Recently, Raman Spectroscopy Imaging has emerged from the Raman Spectroscopy technique as a new modality which enables real time, noninvasive, high-resolution imaging to probing the chemical composition of materials with no sample preparation. Thousands of Raman spectrums are acquired from all over the field of view to created chemical images. These chemical images, as functions of Raman intensity and spatial coordinates, allow an assessment of the chemical heterogeneity of a specimen in terms of the spatial distribution of the sample and its underlying molecular constituents [9, 16].

Applications of Raman Imaging techniques cover a wide range of scientific disciplines spanning biology, medicine, and material sciences, as for example, in the analysis of cells, tissues, pharmaceuticals, semiconductors, polymers, artwork, and minerals [17,18].

3. Compressive sensing Compressive sensing has emerged as a promising research area that can enable the acquisition of signals at sampling rates below the Nyquist- criterion or the equivalently scanning methods. In CS traditional sampling is replaced by measurements of inner products with random vectors. The signals are then reconstructed by solving an inverse problem such as a linear program or a greedy pursuit in a basis where these admit sparse representations [11,19].

One of the key concepts in compressed sensing is called sparsity. This concept establishes that most of the energy of a signal is concentrated in a small set of its components.

Most real signals are not sparse themselves, however, one can find a sparse representation in a given basis [11, 19]. In general, spectral signal can be expressed as being a basis representation matrix, with , where represents the size of the spectral signal or data cube and being the representations coefficients of in domain [14, 21].

118

Galvis-Carreño et al / DYNA 81 (188), pp. 116-124. December, 2014.

4. Coded Aperture Snapshot Spectral Raman Imaging System The Coded Aperture Snapshot Spectral Imager (CASSI) architecture implements CS in spectral Raman Imaging [6].

The CASSI first introduced in [2], is a remarkable imaging architecture that effectively senses the three dimensional (3D) spectral information of a scene, using a single 2D coded random projections measurements.

Projections in CASSI are attained using a coded aperture and a dispersive element. The principal components in CASSI are illustrated in Fig. 2. The mathematical model of the CASSI system has been extensively studied in [2,22]. Suppose that the power spectral density of the image of the scene formed by the objective lens at the plane of the coded aperture is denoted by where and index the spatial coordinates and indexes Raman spectrum or Raman Shift. Referring to Fig. 2 and denoting the coded aperture transmission function by , the power spectral density immediately after spatially modulated by the coded aperture is 1

0

(1)

Formally the transference function of the coded aperture is designed as an array of square features (pixels) with size equal to the Focal Plane Array (FPA) detector pixels . can be described as (2)

, ,

where represents the rectangular step ∆ ∆ function accounting for the features shape and , represents the binary value (blocking or transmissive) at the element with 1 representing a transmissive coded element and a 0 representing a blocking code element.

After propagation through relay optic lens and the dispersive element, the power spectral density in front of the detector is given by (3) where represents the relay lenses and the dispersive element operation, and the dispersion induced by the dispersive element. Finally, the detector measures the intensity of the incident light rather than the spectral density as in spectrometers. This is realized by the integration of the power spectral density along the wavelength axis over the FPA spectral range . Then, the measurements at the FPA are given by

(4) Replacing (3) in (4) conduces to

(5)

is shift Assuming, (i) the PSF invariant, (ii) the dispersion by the dispersive element is linear, and (iii) that there is one-to-one mapping between elements of the coded aperture and the detector pixels, the detector measurement can be succinctly expressed as

(6)

In discrete form, the measurement at the detector pixel is given by

,

,

, ,

(7)

where L represent the number of bands of the Raman data cube. Additionally, the measurement can be expressed in matrix form as (8) where is a matrix that accounts for the effects of the coded aperture and the dispersive element, on the data cube .

For spectrally rich scenes or very detailed spatial scenes, a single shot CASSI measurement may not provide a sufficient number of compressive measurements. Increasing the number of shots multiplies the number of measurements, thus rapidly overcoming such limitations [21]. The CASSI spectral imager architecture has been extended to admit multiple measurement shots in [20,23,24]. The multiple measurements are attained as separate FPA measurements, each with a distinct coded aperture that remains fixed during the integration time of the detector. In matrix form, the mathematical model for multi-shot CASSI system is similar to that shown in eq. (8) for the CASSI system (9) for , where is the number of shots. The coded aperture pattern ℓ used to sense ℓ is different for each projection.

A typical example of the measurement process is shown in Fig. 4. This figure shows the three steps of sensing a data cube: spatial encoding, spectral shift and the integration on the detector for three shots. The multi-shot approach allows obtaining different information from the same scene as different coded patterns are used.

Assuming that the Raman data cube size is as shown in Fig. 4, the dispersive element shifted each band one pixel horizontally, causing that the spectral modulated and dispersed image impinges on pixels in the detector then, the CASSI sensing matrix is of size . Notice that, the number of the detector pixels is smaller than the number of the voxels of the discretized 3D data cube. Thus the compressive measurements representing by eq. (8) is an under-determined system of equations.

119

Galvis-Carreño et al / DYNA 81 (188), pp. 116-124. December, 2014.

Coded Apertures

Coded Raman Data Cube

Dispersive Element Operation

Detector N+L-1

N

N

N

Raman Data Cube N

Spatial Sp at ia l

N

L

Spectral

Figure 4. The process of CASSI imaging is depicted for three shots.

Source: The Authors

Several numerical algorithms based in a regularization framework, exploring and exploiting additional properties or structures in the data cube to obtain the image estimation.

The Gradient Projection for Sparse Reconstruction (GPSR) method is an algorithm used for spectral image estimation with the assumption that the signal of interest is sparse or compressible in some basis , where the coefficients of the data cube in this basis are represented by . Specifically, the Raman Data cube is represented by and the corresponding CASSI measurement by . Then the reconstruction consists on recovering such that the cost function is minimized as

higher discrete gradients horizontally and vertically. Whit this regularizer, the TwIST estimate the Raman data cube, corresponds to finding a compromise between the lack of fitness of a candidate estimate to the measurement data and its degree or undesirability, given by the penalty term .

The TV norm measures how much an image varies across pixels, so that a highly textured or noised image will have a large TV norm, whereas a smooth or piecewise constant image would have a small TV norm. A tunning parameter in eq. (11) specifies the relative weight of the constraints versus the data fidelity term.

5. Coded aperture design

(10) where is an S-sparse representation of on the basis and is a regularization constant. The penalty term drives small components of to zero and helps promote sparse solutions [25].

The Two-Step Iterative Shrinkage/Thresholding algorithm or TwIST algorithm [26], is another algorithm framework used frequently in the CS literature. TwIST describes a Raman data cube as the solution to the minimization problem (11)

The quality of the reconstructed signal depends on the correct selection of the coded aperture used for sensing the signals. Coded apertures traditionally employed in CASSI system include, random codes, boolean codes, binary codes, the grayscale codes and Hadamard codes [27]. Boolean codes have proven to have the best results for the reconstruction of spectral images whose number bands is less than the spatial distribution of the image [28].

For the development of this work, random codes are employed and these entries satisfy ℓ, where indicates the number of the projection, with ℓ, representing a transmissive code element and ℓ, representing a blocking code element.

The transmittance of the coded aperture is given by

where the choices for the regularization function include but are not limited to the norm.

Traditionally, TwIST use the total variation (TV) regularizer given by

(13)

where represents the size of the coded aperture.

Fig. 5 shows the example of three coded apertures with

1, , , , , 1, , , .

(12) transmittances of 0.1, 0.5 and 0.8 respectively. For instance, , 0.1 of transmittance refers to 10% of the elements in coded The TV terms penalizes the solution candidates with aperture are transmissive and the remaining are blocking.

120

Galvis-Carreño et al / DYNA 81 (188), pp. 116-124. December, 2014.

a) Tr=0.1

a) Tr=0.5

a) Tr=0.8

Figure 5. Transmittance in Coded Apertures. a) 10% b) 50% and c) 80% of the elements in coded aperture are transmissive.

Source: The Authors

In addition to the transmittance, the number of the captured projections or shots affects the quality of the reconstructions. The number of shots can be expressed in terms of the compression ratio. The latter is defined as (14)

Figure 6. Synthetic Raman data cube.

Source: The Authors

Table 1. Spatial analysis of the reconstruction results.

Raman Data Cube size

where represents the number of shots, and and the spatial and spectral dimensions of the data cube respectively. Equation (14) can be seen as the ratio between the number of measurements and the number of pixels in the reconstructed data cube.

16x16x1024

32x32x1024

6. Simulations and results To test the CASSI system in Raman Spectroscopy, several simulations in MatLab were realized. Three different parts of a synthetic Raman data cube with spatial resolution of , and pixels and with L = 1024 spectral bands were used. This data cubes are part of a pharmaceutical tablet image; the spatial information contains three different compounds, Aspirin, Caffeine and Paracetamol as shown the Fig. 6. The spectral information of the synthetic data cube has the Raman spectrum of each of these three compounds between 642 and 1665 cm-1.

The transmittance given in eq. (13) is varying between 0.1 and 0.8 in order to determine the optimal value of this parameter. Further, the compression ratio established in eq.

(14) is analyzed. Compression ratios of 0.125, 0.24 and 0.5 are employed with different coded aperture in each projection. TwIST algorithm is used for reconstructions purposes established in eq. (11). The optimal value of the regularization parameter is found for each transmittance and each compression ratio. Simulation results are analyzed in terms of PSNR (Peak-Signal-to- Noise-Ratio) of the reconstructed images. Spatial and spectral analyzes of the reconstructions were performed separately, giving flexibility to the system user, who can use certain codes depending on the desired results, higher spatial resolution or higher spectral resolution. The simulations were conducted using an Intel Core i7 3960X 3.3 GHz processor, and 32 GB RAM memory. Each experiment is repeated five times and the respective results are average.

Numerical results for spatial and spectral analysis in the reconstruction are summarizes in Tables I and II respectively. Both, spatial and spectral analysis of the

64x64x1024

Compression Ratio

Optimal Transmittance

0.125 0.25 0.50 0.125 0.25 0.50 0.125 0.25 0.50

0.2 0.3 0.3 0.1 0.2 0.225 0.125 0.225 0.125

Spatial PSNR (dB) 9.2391 13.0343 16.7966 11.5391 14.6968 18.8215 9.6404 13.112 17.4113

Source: The Authors

reconstructions allows to establish optimal values of transmittance among 0.1 and 0.3. The PSNR always increases when the compression ratio increases.

An analysis of the spatial and spectral information of the reconstructions is realized using the parameters summarized in tables I , II and are shown in Fig, 7, 8 for three spatial regions of the data cube.

Fig. 7 shows a comparison between the spatial information of the reconstructions obtained with different compression ratio. Columns a, b and c show the spatial information of the bands number 162, 970 and 626 for the three different sizes of the data cube, , , and , respectively. The improvement in the spatial quality can be observed when take a greater compression ratio.

On the other hand, the spectral quality of the reconstructed data cube was also analyzed, for this, one spatial point from each scene was chosen randomly and its corresponding spectral signature (Raman shift) was plotted.

Fig. 8 shows the comparison between the spectral signatures of the original data cube with the corresponding reconstructions obtained. Rows a, b and c show the spectral reconstructions of the points (5, 2), (13, 15) and (38, 45) selected from the three different sizes of the data cube , and , respectively for different compression ratios. The Raman shifts of these points correspond to aspirin, paracetamol and

121

Galvis-Carre単o et al / DYNA 81 (188), pp. 116-124. December, 2014.

Table 2. Spectral analysis of the reconstruction results.

Raman Data Cube Compression Ratio size 0.125 16x16x1024 0.25 0.50 0.125 0.25 32x32x1024 0.50 0.125 0.25 64x64x1024 0.50 Source: The Authors

Optimal Transmittance 0.1 0.1 0.1 0.1 0.125 0.225 0.125 0.1 0.1

Spectral PSNR (dB) 19.9357 27.5535 38.0908 23.0679 28.5752 35.5224 20.5457 26.1435 37.926

Figure 7. Spectral information for compression ratio of 0.125, 0.25 and 0.5. Original spectrum and reconstructed spectrum. Points a) (5, 2), b) (13, 15) and , and respectively. c) (38, 45) for Raman data cubes of size Source: The Authors

caffeine respectively. Notice that the spectral signatures of the reconstruction data cube are very close to the original allowing the mapping and chemical classification of a synthetic data cube.

7. Conclusions The CASSI system has been successfully used to sense and reconstruct three differences parts of a synthetic data cube obtained by Raman Spectroscopy technique. The

spatial and spectral analyses of the reconstructions allow establishing optimal values of transmittance among 0.1 and 0.3. The PSNR for sustained reductions of 50% in the Nyquist criterion (using a compression ratio of 0.5) are up to 16.5 dB and 35.5 dB for spatial and spectral analyses respectively in the reconstructed Raman data cubes. Further, the spectral signatures of the reconstructions get closer to the original, showing that CASSI system in Raman Spectroscopy Imaging has promising results for an optimal chemical classification.

122

Galvis-Carreño et al / DYNA 81 (188), pp. 116-124. December, 2014.

[6] [7]

[8]

[9]

[10]

[11] [12]

Figure 8. Spatial information of the reconstructions for compression ratios of 0.125, 0.25 and 0.5. Bands number a) 162 b) 970 and c) 626 for Raman data , and cube of size respesctively.

Source: The Authors

[13] [14]

[15]

7. Acknowledgments The authors gratefully acknowledge to COLCIENCIAS, especially the “Jóvenes Investigadores e Innovadores” program, who supported the work of the engineer Diana Fernanda Galvis Carreño, one of the authors of this work and the Vicerrectoría de Investigación y Extensión of the Universidad Industrial de Santander for supporting this research registered under the project title: Optimal design of coded apertures for compressive spectral imaging, (VIE 1368 code).

[16] [17]

[18]

References [1] [2]

[3]

[4]

[5]

Yaohai, L., Guangming, S., Dahua, G. and Danhua, L. Highresolution spectral imaging based on coded dispersion. Applied Optics, vol. 52 (5), pp. 1041-1049, 2013.

Wagadarikar, A., John, R., Willett, R. and Brady, D. Single disperser design for coded aperture snapshot spectral imaging.

Applied Optics, vol.

47 (10), pp.

44-51, 2008. http://dx.doi.org/10.1364/AO.47.000B44 Lau, D., Villis, C., Furman, S. and Livett, M. Multispectral and hyperspectral image analysis of elemental and micro-Raman maps of cross-sections from a 16th century painting. Analytica Chimica Acta. vol.

610 (1), pp.

15-24, 2008. http://dx.doi.org/10.1016/j.aca.2007.12.043 McCain, S. T., Gehm, M. E., Wang, Y., Pitsianis, N. P., Brady, D. J. Coded Aperture Raman Spectroscopy for quantitative measurements of ethanol in a tissue phantom. Applied Spectroscopy, vol. 60 (6), pp. 663-671, 2006. http://dx.doi.org/10.1366/000370206777670693 Majzner, K., Kaczor, A., Kachamakova- Trojanowska, N., Fedorowicz, A., Chlopicki, S. and Baranska, M. 3D confocal Raman imaging of endothelial cells and vascular wall. Perspectives in analytical spectroscopy of biomedical research. Analyst, vol. 138 (2), pp. 603-610, 2013. http://dx.doi.org/10.1039/c2an36222h

[19]

[20] [21]

[22]

[23]

[24]

123

Hagen, N. and Brady, D. Coded Aperture DUV spectrometer for standoff Raman Spectroscopy. Proc. SPIE 7319, Next Generation Spectroscopic Technologies II, vol. 7319, 2009.

McCain, S. T., Gehm, M. E., Wang, Y., Pitsianis, N. P. and Brady, D. J. Multimodal multiplex Raman Spectroscopy optimized for in vivo chemometrics. Biomedical vibrational Spectroscopy III: Advances in Research and Industry, pp. 1-8, 2006.

Davis, B. M., Hemphill, A. J., Maltas, D. C., Zipper, M. A., Wang, P. and Ben-Amotz, D. Multivariate Hyperspectral Raman Imaging Using Compressive Detection. Analytical Chemistry, vol. 83 (12), pp. 5086-5092, 2011. http://dx.doi.org/10.1021/ac103259v Schlücker, S., Schaeberle, M. D., Huffman, S. W. and Levin, I. W. Raman Microspectroscopy: a comparison of point, line and widefield imaging methodologies. Analytical Chemistry, vol. 75 (16), pp.

4312-4318, 2003.

Hagen N., Kester R., Gao L. and Tkackyk T. Snapshot advantage: a review of the light collection improvement for parallel highdimensional measurement systems. Optical Engineering, vol. 51 (11), pp. 111702 1-7, 2012.

Donoho, D. Compressed Sensing. IEEE Transactions on Information Theory, vol.

52 (4), pp.

1289-1306, 2006. http://dx.doi.org/10.1109/TIT.2006.871582 Candes, E., Romberg, J. and Tao, T. Robust uncertainty principles: Exact signal reconstruction from highly incomplete frequency information. IEEE Transactions on Information Theory, vol. 52 (2), pp. 489-509, 2006. http://dx.doi.org/10.1109/TIT.2005.862083 Willet, R., Marcia, R. and Nichols, J. Compressed Sensing for Practical Optical Imaging Systems: a tutorial. Optical Engineering, vol. 50 (7), 2011.

Arguello, H. and Arce, G. R. Rank minimization Coded Aperture design for spectrally selective Compressive Imaging. IEEE Transactions on Image Processing, vol. 22 (3), pp. 941-954, 2013. http://dx.doi.org/10.1109/TIP.2012.2222899 Joya, M., Barba. J. and Pizani, P. Efectos estructurales en el semiconductor INSB, por la aplicación de diferentes métodos de presión. Dyna, vol. 79 (15), pp. 137-141, 2012.

Abramczyk, H. and Brozek-Pluska, B. Raman Imaging in Biochemical and Biomedical Applications. Diagnosis and Treatment of Breast Cancer. To appear in Chemical Reviews, 2104.

Mogilevsky, G., Borland, L., Brickhouse, M. and Fountain, A. W. Raman Spectroscopy for Homeland Security Applications.

International Journal of Spectroscopy [Online], 2012. . [Date of reference July 25th of 2013].

Available at: http://www.hindawi.com/journals/ijs/2012/808079/ Maltaşa, D. C., Kwokb, K., Wanga, P., Taylorb, L. S. and BenAmotz, D. Rapid classification of pharmaceutical ingredients with Raman spectroscopy using compressive detection strategy with PLSDA multivariate filters. Journal of Pharmaceutical and Biomedical Analysis, vol. 80, pp. 63-68, 2013.

Donoho, D., Tsaig, Y., Drori, I. and Starck, J. Sparse solution of underdetermined systems of linear equations by stagewise orthogonal matching pursuit. IEEE Transactions on Information Theory, vol.

58 (2) pp.

1094-1121, 2012. http://dx.doi.org/10.1109/TIT.2011.2173241 Duarte, M. and Baraniuk, R. Kronecker Compressive Sensing. IEEE Transactions on Image Processing, vol. 21 (2), pp. 494-504, 2012. http://dx.doi.org/10.1109/TIP.2011.2165289 Arguello, H. and Arce, G. R. Code Aperture Optimization for Spectrally Agile Compressive Imaging. Journal of the Optical Society of America A, vol. 23 (11), pp. 2400-2413, 2011. http://dx.doi.org/10.1364/JOSAA.28.002400 Arguello, H., Rueda, H., Wu, Y., Prather, D. Arce, G. R. Higherorder computational model for coded aperture spectral imaging.

Applied Optics, vol. 52 (10), pp. D12- D21, 2012. http://dx.doi.org/10.1364/AO.52.000D12 Arce, G. R., Brady, D. J., Carin, L. and Arguello, H. Compressive Coded Aperture Spectral Imaging: An Introduction. IEEE Signal Processing Magazine, vol. 31 (1), pp. 105-115, 2014. http://dx.doi.org/10.1109/MSP.2013.2278763 Rueda, H. and Arguello, H. Spatial super- resolution in coded aperturebased optical compressive hyperspectral imaging systems. Revista Facultad de Ingeniería Universidad de Antioquia, pp. 7-18, 2013.

Galvis-Carreño et al / DYNA 81 (188), pp. 116-124. December, 2014.

[25] Wagadarikar, A., Pitsianis, N. P., Sun, X. and Brady, D. J. Spectral image estimation for coded aperture snapshot spectral imagers.

Proceedings of SPIE, vol. 7076, pp. 707602-707615, 2008. http://dx.doi.org/10.1117/12.795545 [26] Bioucas-Dias, J. and Figueiredo, M. A new TwIST: Two-step iterative shrinking/thresholding algorithms for image restoration.

IEEE Transactions Image Processing, vol. 16, pp. 2992-3004, 2007. http://dx.doi.org/10.1109/TIP.2007.909319 [27] Arguello, H. and Arce, G. R. Restricted Isometry Property in Coded Aperture Compressive Spectral Imaging. IEEE Statistical Signal Processing Workshop, Ann Arbor, MI, USA, 2012.

[28] Arguello, H., Correa, C. V. and Arce, G. R Fast lapped block reconstructions in compressive spectral imaging. Applied Optics, vol.

52 (10), pp.

D32-D45, 2013. http://dx.doi.org/10.1364/AO.52.000D32 D.F. Galvis-Carreño, graduated as BSc. of Chemical Engineering in 2011.

She is currently doing his MSc studies in Chemical Engineering at the Industrial University of Santander, Colombia. Her main research areas include Raman Spectroscopy, Compresive Sensing, coded apertures design and image processing.

Y. Mejía-Melgarejo, received the MSc degree from the department of Electrical, Electronics, and Telecommunication, Universidad Industrial de Santander, Colombia, in 2014. Her main research areas are computational spectral imaging, digital signal processing, optical coded apertures design, and image processing.

H. Arguello-Fuentes, graduated as BSc Electrical Engineer in 2000 and as a MSc. degree in electrical power in 2003, both of them from de Universidad Industrial de Santander, Colombia, and the PhD degree in Electrical and Computer Engineering from the University of Delaware, United States. He has working as assistant professor in full-time dedication of the School of Engineering and Computer Systems of the Universidad Industrial de Santander.

His research interests include digital signal processing, compressive sensing, artificial intelligence and telecommunications.

Área Curricular de Ingeniería Química e Ingeniería de Petróleos Oferta de Posgrados   

Maestría en Ingeniería - Ingeniería Química Maestría en Ingeniería - Ingeniería de Petróleos Doctorado en Ingeniería - Sistemas Energéticos

Mayor información: Abel De Jesús Naranjo Agudelo Director de Área curricular qcaypet_med@unal.edu.co (57-4) 425 5317

124

Apply multicriteria methods for critical alternative selection Rosario Garza-Ríos a & Caridad González-Sánchez b b

a Departamento de Industrial, Instituto Superior Politécnico José A. Echeverría, La Habana, Cuba. rosariog@ind.cujae.edu.cu Centro de Enseñanza de Matemática para Ciencias Técnicas, Instituto Superior Politécnico José A. Echeverría, La Habana, Cuba. caryg@cemat.cujae.edu.cu

Received: December 16th, 2013. Received in revised form: March 3th, 2014. Accepted: September 25th, 2014.

Abstract The selection of critical alternatives has been an issue of great interest for both utilities and services, performing this process usually using monocriterial techniques. In practice there are problems it is necessary to apply more than one criterion for selection since this contributes to increasing the quality of decisions and increase the efficiency and effectiveness of the organization.

Multicriteria techniques allow ordering alternatives or determine what is the best, however it is sometimes necessary to determine within this set which will be called criticism for its impact on organizational performance.

In this paper we show two methods using multicriteria techniques, including ELECTRE II are possible to determine within an ordered set of alternatives which are considered critical. The application of these methods is illustrated on three examples.

Keywords: Alternative reviews; multicriterio; models of excellence.

Selección de alternativas críticas aplicando un enfoque multicriterio Resumen La selección de alternativas críticas ha sido un problema de gran interés para las empresas productoras y de servicios, realizándose este proceso usualmente con el uso de técnicas monocriteriales. La aplicación de más de un criterio para la selección contribuye al incremento de la calidad de las decisiones y a elevar la eficiencia y la eficacia de la organización.

Las técnicas multicriteriales permiten ordenar las alternativas o determinar cuál es la mejor, sin embargo en ocasiones es necesario determinar dentro de este conjunto las que serán denominadas críticas por su impacto en el desempeño de la organización.

En el presente trabajo se muestran dos métodos que haciendo uso de técnicas multicriteriales, entre las que se encuentra ELECTRE II, permiten determinar dentro de un conjunto de alternativas ordenadas cuales serán consideradas como críticas. Se ilustra la aplicación de estos métodos en tres ejemplos prácticos.

Palabras clave: alternativas críticas; multicriterio: modelos de excelencia.

1. Introducción Es muy común encontrarse en la gestión empresarial con la problemática de determinar las alternativas críticas entre un conjunto de alternativas, pudiendo definirse como alternativas, los procesos de una organización, los datos que afectan el desarrollo de esta, los requisitos críticos para el desarrollo de un software, los criterios y factores que no permiten que la misma obtenga un nivel de excelencia dado y sobre los cuales deberá trabajar para elevar el nivel alcanzado; estos son algunos de los problemas a los que trataremos de darle solución en el presente trabajo.

Es muy usual encontrar organizaciones que utilizan una vía muy cómoda y viable para la selección de las alternativas críticas, la cual está basada en un sólo criterio: el proceso que

más afecta la obtención del resultado de la empresa. Sin embargo, si se tratara de la obtención de resultados que puedan ajustarse mejor al cumplimiento de diferentes criterios los cuales van desde procesos que puedan garantizar elevar la satisfacción de los clientes, hasta aquellos en los que se generen costos mayores a los esperados o cuya ejecución resulte nociva al medio ambiente, esta práctica resulta ineficiente.

La tendencia de considerar varios criterios para escoger la mejor alternativa resulta novedosa en el sector empresarial, lo que se ha evidenciado en la realización de diferentes trabajos 1-5, en los cuales se muestra la utilización de herramientas multicriterio para la selección de la mejor alternativa.

Según la opinión de las autoras la aplicación de las técnicas multicriterio aporta la flexibilidad necesaria para la determinación de las alternativas críticas, sea cual sea el

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 125-130. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41220

Garza-Ríos a & González-Sánchez / DYNA 81 (188), pp. 125-130. December, 2014.

marco en que esto se presente, sin embargo es necesario establecer una escala que apoye esta decisión 6.

El problema que se pretende solucionar con el desarrollo de la presente investigación es: “La no utilización de técnicas matemáticas, específicamente multicriterio, en la determinación de alternativas críticas en una entidad, limita la flexibilidad y fundamentación en la toma de decisiones, no permitiendo obtener la eficiencia, eficacia y competitividad empresarial requerida.”

En el presente trabajo se exponen los resultados obtenidos de la investigación realizada: en el primer epígrafe aparece las técnicas propuestas por las autoras para la determinación de las alternativas críticas en una organización, en el segundo se muestran algunas aplicaciones del mismo y en el último aparecen las conclusiones a las que se arriban.

3.1. Selección de procesos críticos La selección de los procesos críticos en las organizaciones resulta de vital importancia, ya que estos serán los procesos a los que se les deberá asignar la mayor cantidad de recursos y cuya ejecución no puede perderse de vista porque pondría en peligro el cumplimiento de los objetivos de la empresa Una empresa tiene identificados sus procesos y clasificados en estratégicos, claves y de apoyo de acuerdo a lo que establece la Norma ISO 9000: 2008 , sin embargo necesita determinar cuáles de estos procesos son críticos (aquel/aquellos en que la reducción de la diferencia entre el rendimiento actual y el deseado tendrá un

Buscar la columna del número de criterios

2. Materiales y métodos Buscar el rango . del >

En trabajos anteriores se demuestra cómo el uso de las diferentes técnicas de consenso y herramientas multicriterios viabiliza y hace mucho más efectiva la selección de alternativas críticas 7,8, sin embargo en ninguno queda claro ni se fundamenta, el porqué de las razones o parámetros utilizados en la determinación de la decisión final.

Las autoras proponen dos formas de obtener el conjunto de alternativas críticas, una es el establecimiento de una escala que ayude al análisis del comportamiento de los resultados obtenidos y la otra es utilizando un enfoque de ELECTRE II para definir los mismos.

A continuación se explican ambas propuestas: 1ra En esta forma primeramente es necesario obtener la evaluación de cada una de las alternativas, utilizando para ello cualquiera de los métodos multicriterio desarrollados y estandarizar o normalizar los resultados obtenidos 9-12, para su utilización se sugiere hacer uso del procedimiento que aparece en la Fig. 1 que a su vez se auxilia de la Fig. 2, indicando el camino a seguir para encontrar el número de alternativas críticas.

2da Una alternativa es considerada crítica (eq.1), si su peso o importancia Wj es mayor que el peso medio eq. (2) de los criterios, filosofía del test de veto del método ELECTRE II 13,

Comprobar sí alternativas cuyos ese mismo rango

¿Э

S

Comprobar sí cantidad de alternativas en ese rango es igual al número máximo permisible de alternativas críticas

existen otras pertenezcan a

N

Analizar sí existen alternativas cuyos pertenezcan al rango inferior.

Serán críticas aquellas con

¿Э?

No Sí

pertenecient es al rango mayor

(1) Comprobar si

(2)

¿Se cumple

donde: Ccrit : Conjunto de alternativas consideradas como críticas Wj: Peso o importancia de la alternativa j Wmed: Peso o importancia promedio m: Total de alternativas consideradas

No

¿Se cumple?

Sí Alternativ a Crítica

3. Resultados

En este acápite se presentan los resultados de la aplicación de la metodología explicada anteriormente.

No

Sí

Figura 1. Procedimiento para la utilización de la escala.

Fuente. Elaboración propia

126

Garza-Ríos a & González-Sánchez / DYNA 81 (188), pp. 125-130. December, 2014.

Figura 2. Número de alternativas críticas.

Fuente: Elaboración propia

Figura 4. Salida del PRESS. Fuente. Software PRESS

impacto muy significativo en el éxito y los resultados de la organización, función o departamento), en este caso se utilizará el primero de los métodos propuestos, antes de aplicar este es necesario definir el conjunto de criterios que se tendrán en cuenta para la toma de decisiones.

Los criterios seleccionados por la empresa para determinar sus procesos críticos son:

 El comportamiento deficiente de los indicadores de evaluación influye en la eficiencia y eficacia del proceso.  Impacto en la satisfacción del cliente.  Incidencia en el cumplimiento de los objetivos estratégicos de la empresa.  Actualización y vigencia de las tecnologías.

Un análisis de las diferentes formas en que los expertos pueden expresar sus preferencias, en cuanto a los criterios para evaluar los procesos, permitió decidir que la escala que mejor se ajusta a las características del problema analizado, por el enfoque con que se definieron los criterios, es la propuesta por el método PRESS 9,14.

En la Fig. 3 se muestran los datos de entrada del caso de estudio y en la Fig. 4 se muestran los resultados obtenidos con la aplicación del software PRESS basado en el método del mismo nombre, donde se puede observar que se han incluido como alternativas todos los procesos de la entidad.

Tabla 1. Índices PRESS obtenidos del proceso de normalización Ïndice PRESS Alternativas normalizado Producción y comercialización de tangibles 38.55 Negocios Tecnológicos 24.73 Consultoría tecnológica y del cambio 17,96 organizacional Formación, producción y difusión de 16,05 conocimientos Dirección 1,52 Mercadotecnia 1,16 Fuente. Elaboración propia

Para aplicar el procedimiento propuesto es necesario estandarizar los resultados obtenidos del índice PRESS, para ello se siguen los siguientes pasos: 1) Eliminar los Índices PRESS inferiores a 1, en este caso: Recursos Humanos, Aseguramiento, Supervisión y Gestión Económica.

2) Ordenar los Índices PRESS atendiendo a su valor numérico de forma descendente.

3) Normalizar los Índices PRESS. La eq. (3) muestra como se calculan los índices PRESS normalizados: (3) : Índice PRESS normalizado : Índice PRESS calculado

Figura 3. Pantalla de captación de datos.

Fuente. Software PRESS

Los resultados obtenidos de la estandarización se muestran en la Tabla 1.

Pasos del procedimiento para la selección de los procesos críticos:  En la Fig. 2 seleccionar la columna correspondiente a la cantidad de alternativas, en este caso nos paramos en el número 6 que es el número de alternativas a considerar.  Utilizando el procedimiento descrito en la Fig. 1, seleccionar la alternativa que tenga mayor en este caso , por lo que el número máximo de 127

Garza-Ríos a & González-Sánchez / DYNA 81 (188), pp. 125-130. December, 2014.

procesos críticos será 3 (ver Fig. 2).  Comprobar si existen otras alternativas cuyos pertenecen a ese mismo rango, en este caso no existe ninguna alternativa que pertenezca a este rango.  Analizar si existen alternativas con perteneciente al rango inferior, en este caso existe una alternativa cuyo , y el rango inferior es 20, por lo que es afirmativa la respuesta, pasando al siguiente bloque.  Comprobar si , ; , entonces: No se cumple la condición, por lo que serán críticos los procesos perteneciente al rango mayor, en este caso tendremos solo un proceso crítico, el proceso Producción y Comercialización de Tangibles con 38,55.

En este sentido deben analizarse de forma detallada las causas que pueden provocar el comportamiento deficiente de los criterios medidos en este proceso, para emprender un plan de acciones que permitan mejorar el desempeño del mismo.

La escala de puntuación diseñada por el Comité de la Calidad para evaluar cada criterio se muestra en la Tabla 3 La construcción de un Diagrama Radar nos permite visualizar cual es la brecha, entre el estado actual y el nivel óptimo deseado.

En la Fig. 5 se evidencia la distancia a recorrer desde el estado actual hasta un nivel óptimo o superior, observándose que los criterios con más dificultades son: satisfacción del cliente; liderazgo; gestión y desarrollo del capital humano; calidad de los procesos, lo cual coincide con lo planteado en la forma 2, donde el conjunto de alternativas críticas está constituido por las alternativas cuyo peso es mayor que el peso medio.

3.2. Determinar los criterios críticos para la evaluación de la Excelencia Empresarial Una terminal de contenedores se prepara para su presentación al Premio de Excelencia, para ello ha seleccionado el Modelo de Excelencia EFQM 15-17, en la Tabla 2 se muestra la puntuación alcanzada una vez realizada una guía de autoevaluación.

Tabla 2.

Puntuación obtenida de la aplicación de la Guía de autoevaluación Criterios Ptos. Máx/Plan Ptos. Obt./Real % Liderazgo 172 135.708 78.9 Política y 105 79.695 75.9 Estrategia Satisfacción 218 153.69 70.5 Cliente Calidad de los 135 114.345 84.7 Procesos Gestión y Desarrollo del 187 123.046 65.8 Capital Humano Información y análisis de la 78 68.016 87.2 calidad Recursos y resultados 51 46.869 91.9 económicos Impacto en la 54 47.79 88.5 Sociedad Fuente. Elaboración propia

Tabla 3.

Escala de puntuación y nivel de excelencia Nivel Escala de puntuación 0 0 1 0-20 2 21-30 3 31-40 4 41-50 5 51-60 6 61-70 7 71-84 8 85-100 Fuente. Elaboración propia

Nivel de Excelencia Deficiente Deficiente Bajo Bajo Bajo Medio-Bajo Medio-Alto Alto Superior

Figura 5. Diagrama Radar con los Niveles de Excelencia obtenidos Fuente. Elaboración propia

En nuestro caso el número de criterios críticos encontrados fue de 4.

Ccrit = {Satisfacción del Cliente; Liderazgo; Gestión y Desarrollo del Capital Humano; Calidad de los Procesos} Sin embargo, ¿cuántos criterios podrán encontrarse en un nivel inferior al deseado y que de todas formas la entidad pueda presentarse a un proceso de evaluación de la excelencia?. Para esto se propone lo siguiente: Condición a cumplir:

donde: : Conjunto de criterios que pertenecen al conjunto y que tienen evaluación medio-bajo, bajo o deficiente Conjunto de criterios críticos : Nivel de evaluación del criterio sí Card L ≤ 1- c entonces, puede optar por el Premio. c: índice de concordancia, puede considerarse como 3/4 o 2/3 16, se propone utilizar el valor de c = 3/4, lo que significa que hasta el 25% de los criterios críticos 128

Garza-Ríos a & González-Sánchez / DYNA 81 (188), pp. 125-130. December, 2014.

pertenecientes al conjunto L, pueden tener NE Medio o por debajo de este, para poder presentarse al premio.

En caso de no cumplirse la condición, entonces la empresa aunque se encuentre en un Nivel de Excelencia (NE) Alto/Superior de forma general, no deberá presentarse al premio hasta haber cumplido al 100% el Plan de Acción de Mejoras a elaborar y posteriormente volverse a autoevaluar, para poder conocer la evolución de la organización y sí aún tienen brechas considerables o no para poderse presentar a un Premio.

Anteriormente se determinaron los siguientes criterios críticos =  Satisfacción del Cliente; Liderazgo; Gestión y Desarrollo del Capital Humano; Calidad de los Procesos  De los cuatro criterios críticos encontrados, ninguno corresponde a un Nivel de Excelencia Medio-Bajo, Bajo o Deficiente: Gestión y Desarrollo del Capital Humano y Satisfacción de los Clientes (Medio-Alto), Liderazgo (Alto) y Calidad de los Procesos (Superior) lo que representa un 100% del total de criterios críticos, es por ello que la empresa no se puede presentar al Premio pues los criterios críticos no cumple con la condición planteada. ¾

optar por el Premio.

entonces,

puede Figura 6. Escala propuesta para determinar las alternativas críticas.

Fuente: Elaboración propia

3.3. Calidad de los datos Una empresa de servicios telefónicos desea determinar cuáles son los datos críticos de sus clientes que hacen que brinden un servicio deficiente en la confección del directorio telefónico.

Se desea mejorar la calidad de los datos en la creación del directorio telefónico de una ciudad.

Criterios utilizados para la evaluación de la calidad de los datos son: exactitud, integridad y actualidad.

Las alternativas o datos son: nombre, primer apellido, dirección y número telefónico.

En este caso se propone aplicar la 1ra forma propuesta, utilizando para la obtención de los resultados el método ELECTRE II, por la experiencia y competitividad del grupo de expertos involucrados.

En la Tabla 4 se muestran los resultados obtenidos de la aplicación del método ELECTRE II, note que en el dato o alternativa dirección ha sido obtenido un índice de calidad de la alternativa negativo por lo que se eliminará del conjunto.

Tabla 4.

Resultados de la aplicación del método ELECTRE Fuerza

Debilidad

Índice de Calidad

Nombre

3

2

1

Primer apellido

3

2

1

Número Teléfono Dirección

3

2

1

0

3

-3

Datos

del

Fuente: Elaboración propia

Tabla 5.

Estandarización de los resultados obtenidos de la aplicación del ELECTRE II. Datos Estandarización Nombre 33.3 Primer apellido 33.3 Número del Teléfono 3.33 Fuente. Elaboración propia

La eq. (4) permite obtener la estandarización de los resultados

(4) donde: Wi: índice de estandarización del dato i ICi: índice de calidad obtenido de la aplicación del ELECTRE II para el dato i m: cantidad de datos o alternativas En la Tabla 5 se muestran los resultados obtenidos de la estandarización al aplicar la ecuación 4.

De la aplicación del procedimiento para la implementación de la 1ra forma, se obtiene que existen 3 alternativas críticas (datos críticos) en la Fig. 6 se muestra como se obtuvo esto.

Como todas las alternativas poseen el mismo índice estandarizado (33.3) los tres datos son críticos, nombre, primer apellido y número telefónico, por lo que debe tenerse especial cuidado al incluirlo en el directorio garantizando que no se cometan errores.

4. Discusión En el trabajo se presentan dos formas con enfoque multicriterio para determinar las alternativas críticas en tres casos diferentes.

La forma 1 es útil si solo se desean obtener las alternativas críticas, sin embargo la forma 2 es preferible si además de obtener estas se requiere determinar cuáles vetan una decisión posterior.

129

Garza-Ríos a & González-Sánchez / DYNA 81 (188), pp. 125-130. December, 2014.

4. Conclusiones La propuesta de dos métodos utilizando el enfoque multicriterio para la determinación de las alternativas críticas, constituye el aporte más significativo de la presente investigación.

El uso del enfoque del test de veto del ELECTRE II con las modificaciones realizadas para determinar las alternativas críticas con un nivel por debajo del deseado propone una solución factible y fácil de utilizar en un problema de decisión multicriterio.

Se diseñó una escala de apoyo para la toma de decisiones que permite la determinación de las alternativas críticas, tomando en cuenta rangos de valores flexibles, demostrándose su aplicabilidad con la utilización del método PRESS y ELECTRE II. El procedimiento propuesto para la selección de alternativas críticas presenta la flexibilidad necesaria para la toma de decisiones.

Referencias [1]

Alvarado, L., et al., Jerarquización multicriterio de la banca: Una herramienta de apoyo a la toma de decisiones en las cajas de ahorro del municipio Guanare, Venezuela, Revista Venezolana de Análisis de Coyuntura, XV (1), pp. 199-217, 2009.

[2] Fürst, E., Evaluación multicriterio social: ¿Una metodología de ayuda a la toma de decisiones o un aprendizaje social sujeto a una reinterpretación institucional-evolucionista?, Revista Iberoamericana de Economía Ecológica, 8, pp. 1-13, 2008.

[3] Soto-de la Vega, D., Vidal-Vieira, J. y Vitor-Toso, E.A. Metodología para localización de centros de distribución a través de análisis multicriterio y optimización. Revista DYNA 81 (184), pp.

28-35, 2014. http://dx.doi.org/10.15446/dyna.v81n184.39654 [4] Policani, A., Using a MCDA approach for the service quality sorting problem. Memorias electrónicas del XII CLAIO. La Habana, Cuba, T 006, 2004.

[5] Barker, T.J. and Zabinsky, Z.B., A multicriteria decision making model for reverse logistics using analytical hierarchy process, Revista Omega, 39 (5), pp.

558–573, 2011. http://dx.doi.org/10.1016/j.omega.2010.12.002 [6] González, C. y Garza, R., Una escala para la selección multicriterio de alternativas críticas. Memorias electrónicas del 5to Congreso Internacional de Ingeniería Eléctrica, Electromecánica y Sistemas, pp. 904 -908, 2008.

[7] García, C. y Muñoz, P., Localización empresarial en Aragón: Una aplicación empírica de la ayuda a la decisión multicriterio tipo ELECTRE I y III. Robustez de los resultados obtenidos, Revista de Métodos Cuantitativos para la Economía y la Empresa 7, pp.31-56, 2009.

[8] Vilalta J.A., Procedimiento para el diagnóstico de la calidad de los datos en organizaciones cubanas. Tesis Doctoral, Departamento de Ingeniería Industrial, Instituto Superior Politécnico ¨José A. Echeverría¨, La Habana, Cuba, 2008.

[9] Aragonés, P., Aproximación a la toma de decisiones en proyectos.

Implementación de una metodología multicriterio y multiexperto: PRESS II. Tesis Doctoral, Universidad Politécnica de Valencia, Valencia, España, 1997.

[10] Saaty, T., Toma de decisiones para líderes. El proceso analítico jerárquico. La toma de decisiones en un mundo complejo. RWS Publications, 1997.

[11] Romero, B. and Pomerol, J.Ch., Decisiones multicriterio: Fundamentos teóricos y utilización práctica. Colección de Economía, Universidad de Alcalá, España, 1997 [12] Doumpos M. and Zopounidis C. A multicriteria outranking modeling approach for credit rating. Decision Sciences, 42, pp. 721– 742, 2011. http://dx.doi.org/10.1111/j.1540-5915.2011.00328.x

[13] Roy, B., The outranking approach and the foundations of ELECTRE methods, in Bana, E. and Costa, C.A. (Eds.), Reading in multiple criteria decision aid, Springer-Verlag, Berlin, 1990. http://dx.doi.org/10.1007/978-3-642-75935-2_8 [14] Heras, N., Salinas, E., Martínez, L., González, C. y Garza, R., Sistema de ayuda a la toma de decisiones: Jerarquias y Press.

Memorias electrónicas del XII CLAIO 2004. La Habana, Cuba, T 278, 2004.

[15] Benítez, Y. y Vizcaíno, D., Utilización de herramientas matemáticas en la adaptación de modelos de Excelencia, Tesis de Grado Departamento de Ingeniería Industrial, Instituto Superior Politécnico ¨José Antonio Echeverría¨, La Habana, Cuba, 2007.

[16] Pino, G., Garza, R. and Pérez, I., Searching for entrepreneurial excellence: An approach to ELECTRE II philosophy. Proceeding of the MS´10 International Conference, published by World Scientific Proceeding Series on Computer Engineering and Information Science Vol. 3. Valencia, España, 2010.

[17] Modelo EFQM de excelencia. [Online], [Date de referencia: agosto 13 de 2013], Disponible en: http://www.efqm.org R. Garza-Ríos, es Ingeniera Industrial en 1980, MSc. en Optimización y técnicas de ayuda a la decisión, en 1996 y en el año 2001 Dra. en Ciencias Técnicas. Es profesora Titular en Matemática Aplicada del Dpto. de Ingeniería Industrial y profesora de Investigación de Operaciones y Simulación en pregrado. Imparte clases de postgrado en materias asociadas a problemas de decisión con incertidumbre, multicriterio y multiexperto.

C. González-Sánchez, es Ingeniera Industrial en 1970, Esp. en Sistemas Automatizados de Producción y Distribución, y en 1976 Dra. en Ciencias Económicas en el año 1981. Es profesora Titular y Consultante en Matemática en el Centro de Estudios de Matemática para Ciencias Técnicas, profesora de Matemática en pregrado. Imparte postgrado en materias asociadas a problemas de decisión utilizando técnicas matemáticas.

130

Área Curricular de Ingeniería Administrativa e Ingeniería Industrial Oferta de Posgrados     

Especialización en Gestión Empresarial Especialización en Ingeniería Financiera Maestría en Ingeniería Administrativa Maestría en Ingeniería Industrial Doctorado en Ingeniería - Industria y Organizaciones Mayor información:

Elkin Rodríguez Velásquez Director de Área curricular acia_med@unal.edu.co (57-4) 425 52 02

Garza-Ríos a & González-Sánchez / DYNA 81 (188), pp. 125-130. December, 2014.

131

Flatness-based fault tolerant control César Martínez-Torres a, Loïc Lavigne b, Franck Cazaurang b, Efraín Alcorta-García a & David A. Díaz-Romero a b

a Universidad Autónoma de Nuevo León, Nuevo León, México., efrain.alcortagr@uanl.edu.mx Université Bordeaux I, Bordeaux, Francia.,loic.lavigne@ims-bordeaux.fr, franck.cazaurang@ims-bordeaux.fr

Received: December 18th, 2013. Received in revised form: March 10th, 2014. Accepted: September 18th, 2014

Abstract This paper presents a Fault Tolerant control approach for nonlinear flat systems. Flatness property affords analytical redundancy and permit to compute the states and control inputs of the system. Residual signals are computed by comparing real measures and the computed signals obtained using the differentially flat equations. Multiplicative and additive faults can be handled indistinctly. The redundant signals obtained with the differentially flat equations are used to reconfigure the faulty system. Feasibility of this approach is verified for additive faults in a three tank system.

Keywords: Fault tolerance, differential flatness, three tank system.

Control tolerante a fallas basado en planitud Resumen Este artículo presenta un método de control tolerante a fallas para sistemas no lineales planos. Las propiedades intrínsecas de los sistemas planos generan redundancia analítica y permiten calcular todos los estados y las entradas de control del sistema. Los residuos son calculados comparando las medidas reales provenientes de los sensores y las señales obtenidas gracias al conjunto de ecuaciones del sistema plano. Fallas multiplicativas y aditivas se pueden manejar de manera indistinta. Las señales redundantes obtenidas con las ecuaciones del sistema plano son usadas para reconfigurar el sistema con falla. La factibilidad del método propuesto es verificada para fallas aditivas en un sistema de tres tanques.

Palabras clave: Tolerancia a fallas, planitud diferencial, sistema de tres tanques.

1. Introduction Since early in the 90's, Fault Tolerant Control (FTC) algorithms become an active research area due to the importance of operate reliable and/or profitable production systems, see for instance [1]. FTC system means that if a fault occurs in a system, the control system could be capable of overcome the fault effect and continue working (sometimes in a degraded mode). In the frame of analytical redundancy (in contrast with physically redundancy), FTC could be achieved basically in two known ways: The first one considers into the control design the possible effect of faults on the system. This approach is denominated passive and it has a close relationship to robust control design [2].

As it is well known, as far as more faults are considered in the design, the performance of the controller becomes conservative, i.e. the performance is degraded. The second one uses a different strategy: the FTC-algorithm responds to a system fault by modifying the control loop. This approach

is denominated active and the main characteristic is the high performance that could be reached [1]. Furthermore, active FTC could cover a larger span of faults.

There are not too much results reported in the literature considering flatness based approach. An early FTC flatness-based approach has been considered in [3], where the fault tolerance is carried out by using the fault estimation (using an algebraic estimation approach). Such estimates are obtained using the differentially flat equations to compute a fault-free version of the states and then compare them versus the faulty sensor. Such operation will provide an estimate of the fault. The approach is intended for actuator faults. According to the authors additive and multiplicative faults could be treated indistinctly. In [4] is presented an approach based also on fault estimation, which is obtained using the differentially flat equations to compute a fault-free version of the states and then compare them versus the faulty sensor. Such operation will provide an estimate of the fault. Then the

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 131-138. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41275

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014.

signal is conditioned using B-splines. The obtained trajectory is subtracted from the measure of the faulty sensor. Only sensor faults are taken into account. This approach is applied to linear systems with a focus on sensor faults. The main disadvantage of both techniques is the fact that estimation plus signal conditioning could take some time to be accomplished. Such time delay could lead the system to instability.

This work presents a FTC flatness-based approach which overcome the time delay of early approaches, the analytical redundancy needed to compute the residual signals is obtained from the inherent properties of the flat systems, in fact if a linear or nonlinear system is flat each state and control input could be expressed as function of a so-called flat output vector, this property will provide the redundancy needed to compute the residual signals.

Furthermore a fault-free version of the states and control inputs which are not part of the flat output vector is computed, such reference is the used to hide recover the faulty system. For sensors additive faults are considered in both stages FDI and control reconfiguration, however for faults affecting control inputs only FDI is considered, the fault effect is rejected by the controller.

This paper is organized as follows: section 2 presents the differential flatness property and the flatness motion planning. The FTC approach is presented in section 3.

Section 4 presents the results of the proposed approach applied in a classical three tank system, thanks to its versatility this system is widely used in the FTC community, see for instance [5-7]. Section 5 is devoted to present the conclusion.

2. Differential flatness The flatness theory search to determine if a system of differential equations could be parameterized by arbitrary functions. The first works have been carried out in [8], aiming aeronautical applications. The theory development continued in the PhD. dissertation of P. Martin [9], this work has led to the formal concept of flatness presented by M. Fliess et al. in [10].

The differential flatness of non-linear and linear systems could be described by using mathematical formalisms, and specifically differential algebra or differential geometry.

A non-linear or linear system is flat if there exists a set of variables differentially independent, called flat outputs, whose number is equal to the quantity of control inputs, such as, the vector state and the control inputs can be expressed as functions of the flat outputs and a finite number of its time derivatives. By consequence, state and control inputs trajectories can be obtained by planning only the flat output trajectories, this property can be particularly exploited on trajectory planning, see [11-14] and trajectory tracking [15,16]. Flatness could be used to design robust controllers, see for instance [17,18].

Definition 1: Let us consider the nonlinear system ୬ ୫ the state vector, the control vector ஶ and a function of and . The system is differentially ୫ flat if, and only if, it exists a flat output vector such as:



The flat output vector is expressed as function of the state and the control input and a finite number of its time derivatives. ሺஓሻ

୸



(1)

The state and the control input are expressed as functions of the vector and a finite number of its time derivatives. ஑ (2) ୶ ஑ାଵ

୳

Where

஑

denotes the

୲୦

(3)

time derivative of .

2.1. Flatness-based motion planning The goal of motion planning is to find control actions that move the concerned system from a start state to a goal condition, while respecting constraints and avoiding collision. Differential flatness is especially helpful to this, because if nominal trajectories for the flat outputs are available it is possible to find open-loop control inputs to drive the system to the final condition. If the nonlinear system is not flat create such trajectories requires an iterative solution by numerical methods. This iterative process can be solved by using optimal control techniques, however for nonlinear systems some problems still unsolved. Besides this solution needs to integrate the system equations in order to evaluate the solution proposed.

Motion planning by flatness, does not need to integrate the system equations and for a flat output trajectory, command inputs can be computed directly, the vector resultant always respect the system dynamics, see eq. (3).

By consequence the solutions of the set of differential equations are found. See [11,19].

Definition 1 implies that every system variable can be expressed in terms of the flat outputs and a finite number of its time derivatives. By consequence if we want to compute a trajectory whose initial and final conditions are specified, it suffices to construct a flat output trajectory to obtain the open loop control inputs satisfying the system output desired. In order to compute all the system variables, the flat output trajectory created needs to be at least times differentiable, where is the maximal time derivative of the flat output appearing in the differential flat equations. Additionally this trajectory is not required to satisfy any differential equation.

By consequence the flat outputs trajectories can be created by using a simple polynomial approach. If the trajectories needs to be optimal in some sense, a more advanced trajectory generation technique has to be used, some application examples can be found in [11,12,14,20].

3. Fault tolerant control approach The FTC proposed approach presented in Figure 1 keeps the nominal control in order to reduce the time response after a fault, the idea is to couple the FDI stage together with the fault recovery strategy, in fact some signals used to compute the residues are not affected by the fault, by consequence those signals are used to response to the fault effect.

131

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014.

3.1. Fault Detection and Isolation Let us consider a nonlinear flat model of dimension , and control inputs, with ஑ as first set of flat outputs, which corresponds to components of the state vector, also suppose that the full state is measured, it is always possible to compute residues:  State residues, because the full state is supposed to be measured.  Control inputs residues.

The residual signals are computed as follows: ୨୶

୫୩

୩

୨୳

୫୪

୪

(4)

where ୫୩ and ୫୪ are the ୲୦ and ୲୦ measured state and control input respectively and ୩ and ୪ are the ୲୦ and ୲୦ state and control input calculated using the differentially flat equations. Suppose that we have a nonlinear system composed by four states, ୘ ୬ and two control inputs ଵ ଶ ୘ ଵ ଶ ଷ ସ ୫ , as depicted in definition 1, the number of control inputs are equal to the number of flat outputs, by consequence ୫ , suppose too that the nonlinear system is flat, ଵ ଶ ୘ with the flat output vector equal to ஑ ஑ଵ ஑ଶ ୘ ୫ . ଵ ଶ This stage can be improved if a second set of flat outputs is found, see [21] for further details.

3.1.1. Residues computation and fault recovery The proposed approach has the next consequences:  The maximal number of residues is four.  Sensor faults not affecting flat outputs can be isolated depending on the system.  Flat output sensor faults can be detected but cannot be isolated.

The residual signals are computed as follows: ଵ୶ ଶ୶ ଵ୳

୫ଷ ୫ସ ୫ଵ

ଶ୳

୫ଶ

Where

ୟ

where

ଵ

ଵ

ଶ

ଶ

ଷ

୶

ଵ

ଵ

ଶ

ଶ

ସ

୳

ଵ

ଵ

ଶ

ଶ

ଵ

୳

ଵ

ଵ

ଶ

ଶ

ଶ

୬

and

where

୶

ୠ

ୟ ୫

.

୘

(5)

୘

ୟ ୠ

૛‫ܠ‬

1 1 0 1 0 0

૚‫ܝ‬

1 1 0 0 1 0

૛‫ܝ‬

1 1 0 0 0 1

detected but it cannot be isolated. The Table 1 presents each fault signature.

Fault isolation is assured for each state not included in the flat output vector, additionally the flat outputs are considered fault-free at any time, as consequence the right part of the eq. (5) will be fault-free at any time, such reference is then used to recover the system after the fault.

The controller reference is changed by a switch, which is triggered by a decision algorithm.

Fig. 2 presents the control reconfiguration proposed approach, in this, the nominal trajectory is calculated by creating trajectories for each flat output and then, using the eq. (3) to compute the nominal control inputs. Additive and multiplicative faults are consider for both, sensors and actuators. In order to create the residual signals (eq.(4)), each flat output needs to be measured, such measure is then introduced in eq. (2) and (3), by consequence control inputs and states estimations are available, such estimations are helpful to recover the system after fault by simply changing the faulty measure for the estimated one.

Let us retake the example presented previously. If a fault affects the measure signal ୫ଷ , the fault will be detected and isolated, see Table 1. Considering that the flat output vector is compound by ଵ ଶ and such states are consider fault-free at any time, an unfaulty reference of ଷ can be estimated using the right part of the eq. (5), such reference is then used to feed the controller and recover the system.

3.2. Derivatives estimation In order to compute the system states and the control inputs of the system, and consequently the residual signals, the time derivatives of the flat outputs of the system has to be estimated.

In this work a high-gain observer [22] is used to evaluate the time derivative of noisy signals. In order to improve the performance of the high-gain observer, a low-pass filter is synthesized. The delay introduced by the filter could affect the control reconfiguration; by consequence especially attention on its design has to be considered.

୘

୘

Table 1.

Source Faults signatures Fault ૚‫ܠ‬ 1 ୶ଵ 1 ୶ଶ 1 ୶ଷ 0 ୶ସ 0 ୳ଵ 0 ୳ଶ Source: The Authors.

ୠ

3.3. Detection robustness Analyzing the eq. (5) it is straightforward to see that if a fault affects the state measure of ୫ଷ , the residual ଵ୶ will be affected, the rest of residues are independent of this measure, so they will not be affected by the fault. A fault affecting ୫ସ or the control inputs can be analyzed in the same manner.

When a fault affects one of the flat outputs, all the residues will be affected, by consequence the fault can be

For this work the fault detection is achieved by simply comparing the residual amplitude versus a fixed detection threshold. The amplitude of the detection threshold is fixed by running series of fault-free simulations of the system.

Three different simulations are run, the first one by changing each parameter individually in the same percentage upwards and downwards. The two final simulations are run by varying all the parameters plus and minus the same percentage used in the previous simulation.

132

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014.

Figure 1 FDI and FTC proposed approach Source: The Authors.

Finally, the amplitude of the detection threshold is fixed by selecting the worst case among all the results of the simulations, plus a security marge. Such marge is added in order to avoid false alarms caused by the unknown perturbations or modeling errors.

4. Three tank system The proposed approach is applied to a classical three tank system; this classical system is used such system is composed by three tank, a central reservoir and two pumps in charge of introduce liquid to the system. Each tank is linked to the central reservoir by means of a pipe with transversal section equal to S. The tanks are linked between them with a pipe with the same section. See Fig. 2.

The nonlinear mathematical model is obtained as follows: ଵ

ଵ଴

ଵ

ଵଷ

ଵ

ଷ

ଶ

ଶ଴

ଶ

ଷଶ

ଶ

ଷ

ଷ

ଵଷ

ଵ

ଷ

ଷଶ

ଶ

ଵ

(6)

ଶ ଷ

ଷ଴

Table 2.

Coefficients values Coefficient az1 az2 az3 Sn S Source: The Authors.

Value 0.75 0.76 0.75 ିହ

0.0154

Where, xi, i=1,2,3, denotes the individual tank; Qi0, i=1,2,3 represents the outflow between each tank and the central reservoir, Q13 and Q32 are the outflow between tank one and tank three and the outflow between tanks three and two respectively, u1 and u2 are the incoming flows of each pump.

The valves connecting tanks one and three with the central reservoir are considered closed, so Q10 and Q30 are always equal to zero.

The flows Q13, Q32 and Q20 can be expressed as follows:

ଷ ଵଷ

ଵ

ଷ

୸ଵ ୬ ଶ଴

ଷଶ

ଶ

ଷ

ଶ ୸ଷ ୬

ଵ ୸ଶ

ଷ ୬

ଷ

ଶ

(7)

ଶ

Where Sn represents the transverse section of the pipes connecting the tanks and azr, r=1,2,3 represents the flow coefficients. Coefficients values are depicted in Table 2.

4.1. Flat model A system is flat if and only if each state and control input is expressed in function of the flat output, see definition 1 for more details. Let us define the flat output ୘ ୘ ଶ vector as: , the ஑ ଵ ଷ ஑ଵ ஑ଶ differentially flat equations can be written as follows: Figure 2 Three tank schema Source. Noura et al. 2009.

133

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014. ஑ ଵ

஑ଵ ஑ଶ

஑ ଷ

஑ଶ

஑ ଵ

஑ଵ

஑ ଶ

஑ ଶ

஑୶

ଶ ୸ଵ ୬

஑ ଶ

஑ଵ

஑ଵ

஑ଶ

஑ଶ

୸ଷ ୬

(8) ୸ଵ ୬

஑ଵ

୸ଷ ୬

஑୳

஑ ଶ

஑ଶ

஑ ଵ

஑ଶ

஑ଵ

஑ ଶ

஑ଶ

୸ଶ ୬

஑ ୘ ଷ

(9) ஑ ଵ

஑ଶ

஑ ଶ

Figure 3 Normalized residues for Source: The Authors.

஑ ୘ ଶ

It is straightforward to see that the three tank system is flat, eq. (8) prove that each single state, and the two control inputs can be written as a function of the flat output vector and a finite number of its time derivatives. Eq. (9) groups such functions in a vector, in order to simplify the notation.

ଵ

ିଷ

(10) ଶ

ିଷ

Flat outputs trajectories were generated by using a fifth order polynomial. White noise is added to the measured outputs. Derivatives are estimated by using a high-gain observer [22] coupled to a low-pass filter to reduce the amplitude of the noise and improve the derivative estimation.

According to the process described in section 3.1.1 the residuals are obtained by computing the difference between the value given by the measuring sensor and the estimated

fault.

value obtained with the differentially flat equations, eq. (8) and eq. (9). Equation (11) presents the three residuals. ଵ୶

4.2. Fault Detection and Isolation Additive faults affecting level sensor ୫ଶ and flow actuators are considered. For such faults, a +10cm measure error is considered in sensors and an extra flow of ିହ ଷ is added to input flows. Only a single fault may be present at a time, once the fault appears (at 100 s) it is recurrent until the end of the simulation.

Two PID controllers are connected to high measures of tanks one and two, see (10). Such controllers are synthetized by making a compromise between the fault rejection dynamic and the noise level presented in the measurements.

As described previously faults are detected by simply comparing the residual signal amplitude versus the threshold amplitude. The detection threshold was defined by changing the flow parameters in the range of 10%; afterwards the maximal value for each residue (positive and negative) plus an error margin is used as the final amplitude of the detection threshold. This margin adds robustness and avoids false alarms, if a residue exceeds the threshold, such residue is taken into account to construct the fault signature and by consequence isolate the fault.

ଶ

୶

୫ଶ

ଵ୳

୫ଵ

ଶ୳

୫ଶ

஑ଵ

௨ ௨

஑ଶ

஑ଵ ஑ଵ

஑ଶ

஑ଶ

஑ଵ ஑ଶ

ଶ

஑ଶ ஑ଶ

ଵ

୘ ୘ ଶ

(11) ୘

The results are consistent with the analysis presented in section 3.1.1. Let us analyze for instance the fault of the high measure of tank number two. Looking in detail the eq. (11) it is straightforward to see that the residue ଵ୶ is the only one depending of the measure ୫ଶ , by consequence, this residue is affected, however, since the feedback controller is connected to this measure , the pump number two, which is directly connected to this tank reacts to the fault as well, by consequence the residue ଶ୳ is affected too. See Fig. 3.

Faults affecting flow pumps could be analyzed in the same manner, for instance a fault affecting the measure of the pump number one affects directly the residue ଵ୳ . Due to the fact that we work in closed loop the residue ଵ୶ is affected too but in a smaller quantity, such amplitude variation is not enough to exceed the threshold and by consequence this is not taken into account to construct the fault signature, see Fig. 4.

For a fault in pump number two the residue affected directly is the residual signal ଶ୳ , for this case and since in the three tank system the high level of the two other tanks depends of the high measure of tank number one the pump one increase its flow and by consequence the measure of the first tank is affected, but the amplitude is not enough to exceed the threshold, see Fig. 5.

As expected for an individual fault affecting the measures of the flat outputs the three residues are triggered, by consequence each individual fault can be detected but it is impossible isolate them by simply comparing the fault signature, see Figs. 6 and 7. However none of the PID is connected to the measure of tank three, so this fault will not affect the final position, this results in a non-optimal isolation between faults in tank one and tank three. By this way every sensor and actuator fault can be detected and isolated. Table 3 presents a summary of the residues triggered by each fault.

All the residual signals are normalized between 1 and -1; the boundaries are the maximal and minimal value of the threshold.

134

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014.

differentially flat equations eq. (8).

Figure 4 Normalized residues for Source: The Authors.

Figure 5 Normalized residues for Source. Noura et al. 2009.

ଵ

fault

Figure 7 Normalized residues for Source: The Authors.

fault.

૛‫ܝ‬

1 1 1 0 1

Figure 8 . Control reconfiguration for ଶ sensor fault.

---, No reconfiguration; …, Nominal; __, Reconfiguration .

Source: The Authors.

Figure 6, Normalized residues for Source: The Authors.

ଵ

Fig. 8 shows the position error between nominal trajectory and trajectories with control reconfiguration and without reconfiguration. It is straightforward to see that in case of not reconfigure the controller the error is bigger and permanent, however for the reconfiguration case the error is close to zero at the end of the simulation.

Fig. 9 presents the errors for the final position of the tank number two in the two cases, with and without reconfiguration; it is clearly to see that if the control reconfiguration is not carried out the final trajectory is not followed.

Actuators faults are compensated by the controller, tank three sensor fault does not affect the final position and faults affecting water level measure of tank one can be isolated, but a non-faulty measure is not available. By consequence if such fault affects the system, the system cannot be reconfigured.

ଶ

Table 3.

Faults signatures for the three tank system Source.

Fault ૚‫ܠ‬ ૚‫ܝ‬ 1 1 ୶ଵ 1 0 ୶ଶ 1 1 ୶ଷ 0 1 ୳ଵ 0 0 ୳ଶ Source: The Authors.

ଷ

fault

4.3. Control reconfiguration Once the fault is detected the signal measure is changed by switching between this and the one computed with the 135

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014.

[5] [6]

[7]

[8]

Figure 9 Measure errors Source: The Authors.

[9]

5. Conclusions

[10]

This paper presents a flatness-based Fault Tolerant Control approach. The feasibility of the proposed approach is investigated in a classical three tank system. For this particular system faults affecting actuators are detected. The fault affecting the high measure sensor of tank number two is detected and isolated by simply comparing the amplitude of the residual signal versus a threshold, additionally thanks to the properties of the flat systems a fault-free version of the measure is estimated, such signal is then used to recover the system after the fault.

Actuators faults are detected and isolated in the same manner. Active reconfiguration is not considered since those faults are rejected by the controller.

Faults affecting the flat outputs can be detected but cannot be isolated using the threshold-based approach, however since the state feedback controller does not depend of the measure of tank three this fault will not impact the final position, contrary of the fault measure of tank number one, as consequence this result in a non-optimal identification between such faults. This can be avoided if as in [23] a second set of flat outputs is found, by this way not only one but two redundant signals are available for reconfiguration purposes, see [23] for further details.

[11]

[12]

[13]

[14]

[15] [16] [17]

Acknowledgment The fourth author thanks CONACYT Mexico for the financial support through the grant 178282-CB-2012-01.

[18] [19]

References [1] [2]

[3]

[4]

Isermann, R., Fault-diagnosis systems: An introduction from fault detection to fault tolerance, Springer, 2006. http://dx.doi.org/10.1007/3-540-30368-5 Benosman, M., Passive fault tolerant control, robust control, Theory and applications, [Online] Available from: http://www.intechopen.com/books/robust-control-theoryandapplications/passive-fault-tolerant-control, 2011.

Mai, P., Join, C., Reger, J. et al., Flatness-based fault tolerant control of a non-linear MIMO system using algebraic derivative estimation, in: 3rd IFAC Symposium on System, Structure and Control, SSC’07, 2007.

Suryawan, F., De Dona, J. and Seron, M. Fault detection, isolation, and recovery using spline tools and differential flatness with application to a magnetic levitation system, in Control and Fault-

[20] [21]

[22]

[23]

136

Tolerant Systems (SysTol), 2010 Conference on, IEEE, pp. 293-298, 2010 Noura, H., Theilliol, D., Ponsart, J.C. and Chamseddine, A., Faulttolerant control systems: Design and practical applications, Springer, 2009. http://dx.doi.org/10.1007/978-1-84882-653-3 Adam-Medina M., Theilliol, D., Astorga-Zaragoza, C.M., GuerreroRamírez, G. and Vela-Valdés, L.G., Diagnostico de fallas basado en un filtro desacoplado para sistemas no lineales representados por un enfoque multi-modelos, DYNA, 77 (162), pp. 313-323, 2010.

Köppen-Seliger B., Alcorta-García, E. and Frank, P.M., Fault detection: different strategies for modelling applied to the three tank benchmark-a case study, Proceedings of European Control Conference, 1999.

Charlet, B., Lévine J. and Marino, R., Sufficient conditions for dynamic state feedback linearization, SIAM Journal on Control and Optimization, 29 (1), pp.

38-57, 1991. http://dx.doi.org/10.1137/0329002 Martin, P., Contribution à l’étude des systèmes différentiellement plats, PhD. Thesis, École des Mines, Paris, France 1992.

Fliess, M., Lévine J., Martin, P. and Rouchon, P., Flatness and defect of non-linear systems: Introductory theory and examples, International Journal of Control, 61n(6), pp. 1327-1361, 1995. http://dx.doi.org/10.1080/00207179508921959 Louembet, C., Génération de trajectoires optimales pour systèmes différentiellement plats: Application aux manoeuves d’attitude sur orbite, PhD. Thesis, Université de Bordeaux I, Bordeaux, France, 2007.

Louembet, C., Cazaurang, F. and Zolghadri, A., Motion planning for flat systems using positive B-splines: An LMI approach, Automatica, 46 (8), pp.

1305-1309, 2010. http://dx.doi.org/10.1016/j.automatica.2010.05.001 Milam, M.B., Franz, R. Hauser J.E. and Murray, R.M., Receding horizon control of vectored thrust flight experiment, IEE Proceedings-Control Theory and Applications, 152 (3), pp. 340-348, 2005. http://dx.doi.org/10.1049/ip-cta:20059031 Van-Nieuwstadt, M.J. and Murray, R.M., Real-time trajectory generation for differentially flat systems, International Journal of Robust and Nonlinear Control, 8 (11), pp. 995-1020, 1998. http://dx.doi.org/10.1002/(SICI)1099-1239(199809)8:11<995::AIDRNC373>3.0.CO;2-W http://dx.doi.org/10.1002/(SICI)1099-1239(199809)8:11<995::AIDRNC373>3.3.CO;2-N Antritter, F., Müller B. and Deutscher J., Tracking control for nonlinear flat systems by linear dynamic output feedback, Proceedings NOLCOS 2004, Stuttgart, 2004.

Stumper, J., Svariceck F. and Kennel, R., Trajectory tracking control with flat inputs and a dynamic compensator. arXiv preprint arXiv: pp. 1211-5759, 2012.

Cazaurang, F., Commande robuste des systèmes plats application à la commande d’une machine synchrone, PhD. Thesis, Université Sciences et Technologies-Bordeaux I, Bordeaux, France,1997.

Lavigne, L., Outils d’analyse et de synthèse des lois de commande robuste des systèmes dynamiques plats, PhD. Thesis, Université Sciences et Technologies-Bordeaux I, Bordeaux, France, 2003.

Morio, V., Contribution au développement d’une loi de guidage autonome par platitude. Application à une mission de rentrée atmosphérique, PhD. Thesis, Université Sciences et TechnologiesBordeaux I, Bordeaux, France, 2009.

Cazaurang, F. and Lavigne, L., Satellite path planning by flatness approach, International Review of Aerospace Engineering, 2 (3), pp.

123-132, 2009.

Martínez-Torres, C., Lavigne, L.F., Cazaurang, F., Alcorta-García, E. and Diaz-Romero, D., Fault detection and isolation on a three tank system using differential flatness, in European Control Conference. Zurich, Switzerland, 2013.

Vasiljevic, L.K. and Khalil, H.K. Error bounds in differentiation of noisy signals by high-gain observers, Systems & Control Letters, 57 (10), pp.

856-862, 2008. http://dx.doi.org/10.1016/j.sysconle.2008.03.018 Martínez-Torres, C., Lavigne, L.F., Cazaurang, F., Alcorta-García, E. and Diaz-Romero, D., Fault tolerant control of a three tank

Martínez-Torres et al / DYNA 81 (188), pp. 131-138. December, 2014. system: A flatness based approach, in 2nd International Conference on Control and Fault-Tolerant Systems. Nice, France, 2013.

C. Martinez-Torres, obtained his B.Sc degree in Electronics and Communication Engineering from the Universidad Autónoma de Nuevo León (UANL), Mexico; the MSc degree in Aerospace Engineering in 2010, from Bordeaux University, France and the PhD degree in Automatic Control in 2014, from the UANL and the University of Bordeaux. His current research interest is fault tolerance guidance based on flatness approach dedicated to nonlinear systems.

L. Lavigne, obtained his PhD. degree in Automatic Control in 2003 from the University of Bordeaux I, France. Since September 2005 he has been Associate Professor at Bordeaux I University. His main research interests include fault detection and diagnosis, path planning, flat systems and robust control dedicated to aeronautics and space domains. Dr. Lavigne has been involved in two different European projects (GARTEUR, SIRASAS) which deals respectively on robust control and Fault Detection and Diagnosis in the Flight Control System F. Cazaurang, obtained the BSc and MSc degrees in Electrical Engineering from ENS Cachan in 1988 and 1990, respectively, the PhD. degree in Automatic Control in 1997from University Bordeaux I, France.

From September 1992 to August 1998 he worked as lecturer at Bordeaux University. From September 1998 to August 2010 he worked as Associate Professor of Control engineering in Bordeaux University. Since 2010 he has been Professor of Control Engineering in Bordeaux University. His main research interests include path planning, fault tolerant guidance dynamic inversion and robust control dedicated to aeronautics and space domains.

E. Alcorta-García was born in Monterrey, Nuevo León, Mexico in 1968.

He received the BSc. degree in Electronics and Communication Engineering in 1989 and the MSc. in Electrical Engineering (Automatic Control) in 1992 from the Universidad Autónoma de Nuevo León (UANL), Mexico and the Dr.-Ing. in Electrical Engineering (Automatic Control) in 1999 from the Univeristy Gerhard Mercator of Duisburg (actually Duisburg-Essen University), Germany. Since November 1999 he has held a teaching and research position at the UANL. His research interests include model-based fault diagnosis, fault tolerant control and observers.

D. Diaz-Romero, received his BSc Eng. and MSc. degrees from Universidad Autónoma de Nuevo León, Mexico and his PhD degree in automatic control theory from The University of Sheffield, U.K. He has worked in industry and academics. He currently holds the position of full time researcher at Universidad Autónoma de Nuevo León, Mexico.

137

Área Curricular de Ingeniería Eléctrica e Ingeniería de Control Oferta de Posgrados



Maestría en Ingeniería - Ingeniería Eléctrica

Mayor información: Javier Gustavo Herrera Murcia Director de Área curricular ingelcontro_med@unal.edu.co (57-4) 425 52 64

Dynamic wired-wireless architecture for WDM stacking access networks Gustavo Adolfo Puerto-Leguizamón a, Laura Camila Realpe-Mancipe b & Carlos Arturo Suárez-Fajardo c a

Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. gapuerto@udistrital.edu.co b Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. lcrealpem@correo.udistrital.edu.co c Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. csuarezf@udistrital.edu.co Received: December 22th, 2013. Received in revised form: March 3th, 2014. Accepted: September 25th, 2014.

Abstract This paper presents a dynamic architecture for convergent wired and wireless access networks in Time Division Multiplexing (TDM) based Passive Optical Network (PON) featuring wavelength stacking. Four wavelengths for wired services carrying 10 Gb/s traffic load, one shared extra reconfigurable wavelength and one wavelength common to all Optical Network Units (ONUs) for the transport of wireless services were launched into a 1:64 splitting ratio PON network. In the ONU, a tunable free spaced Fourier optics based filter selects one of the wavelengths conveying wired services and a tunable Fiber Bragg Grating (FBG) filters out the wavelength carrying the wireless services. In the uplink direction, subcarrier multiplexing (SCM) was used for the combined transport of the wired and wireless signals to the Central Office (CO).

Keywords: Access Networks, Dynamic Wavelength Allocation, Optical Filters, WDM Stacking, Wired-Wireless Convergence.

Arquitectura dinámica fija-móvil para redes de acceso WDM apiladas Resumen Este artículo presenta una arquitectura dinámica para redes de acceso convergentes fijas e inalámbricas en red óptica pasiva (PON) basada en multiplexación en división de tiempo (TDM) bajo el paradigma de apilamiento de longitudes de onda. Cuatro longitudes de onda para servicios fijos transportando una carga de tráfico de 10 Gb/s, una longitud de onda reconfigurable extra y una longitud de onda común a todas las unidades de red óptica (ONU) para el transporte de servicios inalámbricos se enviaron a una PON con un relación de división de 1:64. En la ONU un filtro sintonizable basado en óptica de Fourier de espacio libre selecciona una de las longitudes de onda que transporta servicios fijos y un filtro basado en redes de difracción de Bragg (FBG) extrae la longitud de onda que transporta servicios inalámbricos. En el enlace de subida se utilizó multiplexación por división de subportadora (SCM) para el transporte combinado de señales en banda base e inalámbricas a la oficina central (CO).

Palabras clave: Asignación Dinámica de Longitudes de Onda, Apilamiento WDM, Convergencia fija-inalámbrica, Filtros Ópticos, Redes de Acceso.

1. Introduction High capacity optical access networks providing high bandwidth and reliable services to fixed users can also be exploited to deal with the transport of wireless services.

Such hybrid architecture might emerge as a viable access solution where the optical network provides flexible high capacity backhaul to mobile end users [6-9]. To date, several approaches for converged transport of wired and wireless services in optical access networks have been

proposed. A study on the effects of simultaneous wired and wireless transmission on fiber is discussed in [10], an analysis on the optical mobile backhauling is presented in [11] and a discussion on the requirements posed on the optical network technologies from the cellular mobile network is presented in [12]. Recently, a study dealing with the optimized placement of base band units acting as hotels that group several remote radio heads in a converged WDM access network was discussed in [13] and [14] described a view on the evolution of radio access

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 139-144. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41312

Puerto-Leguizamón et al / DYNA 81 (188), pp. 139-144. December, 2014.

networks supported by WDM exploitation in an already deployed infrastructure.

On the other hand, different approaches regarding dynamic capacity allocation have also been proposed. In [15], a remote node based on the combination of an optical switch and an Arrayed Waveguide Grating (AWG) enables the dynamic wavelength assignation among different ONUs, also active routing using Semiconductor Optical Amplifiers (SOA) have been proposed as a solution to dynamically distribute wavelength channels in the access network [16-18]. Consequently, it is clear that fixed and mobile convergence has become a hot topic in the optical networking field, in this context Radio over Fiber (RoF) systems technologies will enable combined transport of fixed and mobile users in the future access networks [19], [20]. As seen, advanced features such as dynamic capacity allocation and capacity upgrade have been proposed based on improvements performed on the remote node or point of wavelength distribution. Such enhancement involves the use of active components that are highly sensitive to polarization and leads to precise control and maintenance of it. Therefore, the challenge is to enable convergence and dynamic wavelength allocation among some other characteristics in an already deployed infrastructure, where the high bandwidth provided by the optical fiber and the ubiquity and flexible connectivity of the wireless access can be merged in a unified optical access platform. To address such requirements, namely, the increment of capacity, dynamic convergent wired and wireless access networks, load balancing and resilience while enabling a seamless way to evolution over the next years by reusing the current fiber infrastructure, a dynamic optical access platform based on a novel (CO) architecture that performs the wavelength stacking of several TDM-PON systems has been proposed and experimentally demonstrated.

wavelengths for overlay wireless traffic. For capacity upgrade of the wired traffic, an extra wavelength common to all the ONUs is mapped to an input port of the AWGs resulting in a selective capability to upgrade capacity among different ONUs. Inset (a) in Fig. 1 shows the downstream wavelengths: four fixed, one extrawavelength and one wavelength for wireless transport.

After fiber transmission and distribution, the downstream signal reach the ONU where an optical coupler splits the signal in two in order to recover the wireless and wired wavelengths respectively. For the wireless traffic a tunable Fiber Bragg Grating (FBG) filters out from the downstream signals λx’RF, whereas for the wired data, a tunable free-spaced Fourier optics (FSFO) based optical filter selects one from four possible wavelengths (λx). In both cases, the wavelength allocation is dynamic as each ONU can be assigned with different wavelength channels depending of the traffic load on the network. Inset (b) depicts the architecture of the ONU. In the upstream, all the users share the same wavelength and both wired and wireless traffic is transported using subcarrier multiplexing (SCM) as seen in inset (c). Wireless and wired signals are electrically multiplexed. Direct modulation and Time Division Multiple Access (TDMA) are considered for the uplink connection. In the uplink, in order to avoid carrier suppression effects a FBG removes one of the sidebands of the upstream signal as seen in the spectrum shown in inset (d).

Subsequently down conversion of the wireless subcarrier and direct detection of both wired and wireless signals are performed. Overall, the presented architecture exploits the WDM stacking to enable the deployment of PON based access networks featuring converged transport and dynamic allocations of resources.

2. Materials and methods

For the experimental demonstration, continuous wave (CW) laser sources generating four wavelengths spaced 0.8 nm with an average modulated optical power of 15 dBm were used at the wavelength feeder. The four fixed wavelength channels were λ1 = 1546.64 nm, λ2 = 1547.44 nm, λ3 = 1548.24 nm and λ4 = 1549.04 nm. The extra wavelength λ-Extra is centered at 1549.86 nm and λ-RF is 1553.08 nm which correspond to the next upper FSR of the 1X8 AWG. The four multiplexed wavelength channels for wired services convey 10 Gb/s traffic Non Return to Zero (NRZ) encoded and the wavelength for the transport of wireless services conveys 10 MBauds, 16QAM modulated onto 5 GHz. For the uplink, all the ONUs share the same wavelength centered at 1532.7 nm that conveys SCMcombined 2.5 Gb/s and 5 MBauds 4QAM modulated onto 5 GHz. The passive matrix is intended to distribute a wavelength channel between different PONs, for demonstration purposes the wavelength channels were directly multiplexed through a gaussian band-pass profile 1X8 AWG with roughly 3 dB insertion losses. A circulator allows the use of a single strand of fiber for bidirectional transmission.

2.1. Architecture description The approach for a wavelength stacked access network featuring dynamic wavelength allocation and convergent wired-wireless traffic in both CO and ONU is depicted in Fig. 1. In the CO, four wavelengths, 0.8 nm spaced, are multiplexed and broadcasted by means of the combination of a passive matrix and an Arrayed Waveguide Grating (AWG). This arrangement allows the even and low loss distribution of the four wavelengths between several PONs. An optical switch placed between the wavelength feeder and the AWG provides dynamic wavelength assignment for wireless distribution and capacity upgrade purposes. In particular, for the wireless traffic, a set of wavelengths denoted as (λx’) use the next upper Free Spectral Range (FSR) of the AWG, these wavelengths are assigned individually depending on the demand to each one of their peers (λx) at the wavelength feeder. As a result, they are also broadcasted to each one of the ONUs enabling a converged platform for the distribution of wireless signals and with the capability of adding up to 4

2.2. Implementation details

140

Puerto-Leguizamón et al / DYNA 81 (188), pp. 139-144. December, 2014.

The optical circulator accounts for insertion losses of approximately 0.7 dB. After 20 km of optical transmission through Standard Monomode Fiber (SMF) and a 1:64 splitting ratio, the measured losses were close to 25 dB. FBG filters featuring a bandwidth of 20 GHz were used to separate the wavelength channel that transports the wireless data λx’RF. While a stretching mechanism enabled FBG tunability to λ-RF, thermal control as described in [21] was

used to assure stable operation on this wavelength. For demonstration purposes the fixed channels were recovered by using a bulk tunable (FSFO) filter with a band-pass of 25 GHz. It should be pointed out that other filtering technologies such as Fabry-Perot filters or FBG can be used; however the FSFO filter provided more flexibility to the experimental demonstration as it allowed changing the band-pass width.

Figure 1. Layout of the wavelength stacked approach featuring dynamic allocations of wavelengths for wired and wireless services Source: The Authors

3. Results and discussions Fig. 2 depicts two different scenarios demonstrating the operation and feasibility of the proposed architecture. Each scenario shows the wavelength channel selection in ONU’s belonging to different PONs. Scenario 1 shows the wavelength allocation in ONU 1 at four different PONs. PON 1 has been assigned with λ1, PON 2 with λ4, PON 3 and PON 4 with λ2. In scenario 2, λ4 is allocated to PON 1, λ1 to PON 2, PON 3 is assigned with λ-Extra due to the high load of the other wavelengths at that time and λ3 is allocated to PON 4. Finally, λ-RF is allocated to all ONUs in the four PONs. The signal degradation for the downlink and uplink was measured for wired and wireless services.

For the experimental evaluation, the quality of signals in ONU 1 and ONU 64 in four different PONs were measured under a dynamic wavelength allocation environment following the two scenarios described above.

Fig. 3(a) shows the Bit Error Rate (BER) performance of the examined wired services showing in all cases a penalty of approximately 2 dB for 1x10-12 BER compared to the

back-to-back curve. Fig. 3(b) shows the quality of the wireless services, degradation of the 16QAM signal was measured showing an Error Vector Magnitude (EVM) below 4 % for received optical powers above -24 dBm and with a degradation of roughly 0.5% compared to the backto-back value. The observed low penalties are caused mostly by the inherent insertion losses of signal transmission through the fiber, the optical coupler based splitting and crosstalk from the adjacent channels in the filtering process at the ONUs. The experimental results showing the quality of the upstream wired signal are shown in Fig. 4(a). We evaluated the upstream signals from four ONU’s at different PONs. Overall, the full penalty was measured to be 3.7 dB for 1x1012. Fig. 4(b) shows the results obtained from the upstream wireless service, the measured EVM is roughly 5% for received powers below -31 dBm and showing a degradation of 1.25% as compared to the back-to-back of the signal. The degradation of the uplink signals is caused by the crosstalk coming from the wired and wireless signal remains in the process of detection and down conversion respectively.

141

Puerto-Leguizam贸n et al / DYNA 81 (188), pp. 139-144. December, 2014.

Figure 3. Experimental results. (a) Downlink BER. (b) Downlink EVM Source: The Authors

Negligible quality differences were found in the uplink services coming from different PONs. 4. Conclusions

Figure 2. Experimental results. Scenarios of dynamic wavelength allocation in wavelength stacked access networks Source: The Authors

An optical architecture for converged transport of wired and wireless signals featuring dynamic allocation in multiwavelength WDM-TDM PON access networks was presented. Four wavelengths for wired services 0.8 nm spaced common to all ONUs and one dynamically routed wavelength were broadcasted in the downlink direction. For the wireless services, up to four dedicated and common wavelengths to all ONUs can be used. Exploitation of the wavelength stacking paradigm to implement dynamic wavelength allocation, load balancing and capacity upgrade for converged transport while allow a seamless way to evolution by reusing the current fiber infrastructure constitutes the novelty and fundamental basis of this approach. This proposal aims at upgrading both the CO and ONU architectures in order to provide WDM connectivity between them. The fact that a stack of wavelengths are broadcasted from the CO to different PON may generate a potential limiting factor in the power budget due to the insertion losses imposed by the optical components placed at the CO and the high splitting ratio proposed. However, the power requirements can be relaxed by using high

142

Puerto-Leguizamón et al / DYNA 81 (188), pp. 139-144. December, 2014.

5. Acknowledgement The authors wish to acknowledge the Optical and Quantum Communications Group of the Universidad Politécnica de Valencia, the E.U. funded project ALPHA 212 352 and the Universidad Distrital Francisco José de Caldas for supporting the realization of this paper References [1] [2]

[3]

[4]

[5]

[6]

[7] Figure 4. Experimental results. (a) Uplink BER. (b) Uplink EVM Source: The Authors

[8]

sensitivity optical receivers at the ONU, typical values of commercial optical detector for PON applications ranges from -27 to -32 dBm. Therefore, the approach is feasible in terms of optical power as long as an appropriate power budget is assured by selecting the right optical devices to implement the CO and OLT. As far as the dynamic WDM behavior, tunable filters are used to extract the wired and wireless wavelengths from the downstream channels; in particular a 20 GHz narrow FBG was used to filter the wavelength carrying the wireless services and a free spaced Fourier optics based filter featuring a bandwidth of 25 GHz was used to select one of the wavelengths carrying the baseband data. Regarding the tunable filters, so far no practical candidate technology can perform fast selection, while there are several technologies for slow selection, such as Fabry-Perot filters, thermally tuned semiconductor optical filters, FBGs or the used in the experiments based on free spaced Fourier optics. Therefore, further research is needed to implement wavelength-tunable optical filters featuring fast optical channel selection. As for the quality of the transported signals, the experimental measurements confirm the good performance of the system, 0.5% degradation for EVM and 2 dB penalties in average for 1x10-12 BER in the downlink whereas 1.25% degradation for EVM with 3.7 dB penalty in average for 1x10-12 BER in the uplink was measured.

[9]

[10]

[11]

[12]

[13]

[14]

143

Moeyaert, V. and Maier, G., Network technologies for broadband access, 13th International Conference on Transparent Optical Networks (ICTON), pp. 1-5, 2011.

Davey, R., Kani, J., Bourgart, F. and McCammon, K., Options for future optical access networks, IEEE Communications Magazine, 44 (10), pp.

50-56, 2006. http://dx.doi.org/10.1109/MCOM.2006.1710412 Kazovsky, L., Shaw, W., Gutierrez, D., Cheng, N. and Wong, S,.

Next-generation optical access networks, J. Lightwave Technol. 25 (11), pp.

3428-3442, 2007. http://dx.doi.org/10.1109/JLT.2007.907748 Hsueh, Y., Rogge, M. Yamamoto, S. and Kazovsky L., A highly flexible and efficient passive optical network employing dynamic wavelength allocation, IEEE J. Lightw. Technol., 23 (1), pp. 277– 286, 2005. http://dx.doi.org/10.1109/JLT.2004.838811 Kani, J., Enabling technologies for future scalable and flexible WDM-PON and WDM/TDM-PON systems, IEEE Journal of Selected Topics in Quantum Electronics, 16 (5), pp. 1290-1297, 2010. http://dx.doi.org/10.1109/JSTQE.2009.2035640 Venkatesan, G. and Kulkarni, K., Wireless backhaul for LTErequirements, challenges and options, 2nd International Symposium on Advanced Networks and Telecommunication Systems, pp. 1-3, 2008.

Popov, M., The convergence of wired and wireless services delivery in access and home networks, Optical Fiber Communication Conference and Exposition (OFC/NFOEC) and the National Fiber Optic Engineers Conference, pp. 1-3, 2010.

Chanclou, P., Belfqih, Z., Charbonnier, B., Duong, T., Frank, F., Genay, N., Huchard, M., Guignard, P., Guillo, L., Landousies, B., Pizzinat, A., Ramanitra, H. and Saliou, F., Optical access evolutions and their impact on the metropolitan and home networks, 34th European Conference on Optical Communication, pp.1-3, 2008.

Ali, M.A., Ellinas, G., Erkan, H., Hadjiantonis, A. and Dorsinville, R., On the vision of complete fixed-mobile convergence, J. Lightwave Technol., 28 (16), pp. 2343-2357, 2010. http://dx.doi.org/10.1109/JLT.2010.2050861 Yong-Yuk, W., Moon-Ki, H., Yong-Hwan, S. and Sang-Kook, H. Colorless two different gigabit data access transmissions using optical double sideband suppressed carrier and optical sideband slicing, IEEE/OSA Journal of Optical Communications and Networking, 5 (6), pp.

544-553, 2013. http://dx.doi.org/10.1364/JOCN.5.000544 Laraqui, K., Small cell optical mobile backhauling: Architectures, challenges, and solutions, 39th European Conference and Exhibition on Optical Communication, pp.

1-3, 2013. http://dx.doi.org/10.1049/cp.2013.1298 Kellerer, W., Kiess, W., Scalia, L., Biermann, T., Choi, C. and Kozu, K., Novel cellular optical access network and convergence with FTTH, Optical Fiber Communication Conference and Exposition (OFC/NFOEC) and the National Fiber Optic Engineers Conference, pp. 1-3, 2012.

Carapellese, N., Tornatore, M. and Pattavina, A., Placement of baseband units (BBUs) over fixed/mobile converged multi-stage WDMPONs, 17th International Conference on Optical Network Design and Modeling (ONDM), pp. 246-251, 2013.

Ponzini, F., Giorgi, L., Bianchi, A. and Sabella, R., Centralized radio access networks over wavelength-division multiplexing: a plug-andplay implementation, IEEE Communications Magazine, 51 (9), pp.

94-99, 2013. http://dx.doi.org/10.1109/MCOM.2013.6588656

Puerto-Leguizamón et al / DYNA 81 (188), pp. 139-144. December, 2014.

[15] Ortega, B., Mora, J., Puerto, G. and Capmany, J. Symmetric reconfigurable capacity assignment in a bidirectional DWDM access network, Optics Express, 15 (25), pp. 16781-16786, 2007. http://dx.doi.org/10.1364/OE.15.016781 [16] Yang, H., Shi, Y., Okonkwo, C.M., Tangdiongga, E. and Koonen, A.M.J., Dynamic capacity allocation in radio-over-fiber links, IEEE Topical Meeting on Microwave Photonics (MWP), pp. 181-184, 2010.

[17] Nguyen-Cac, T., Hyun-Do, J., Okonkwo, C., Tangdiongga, E., Koonen, T. Dynamically delivering radio signals by the active routing optical access network, IEEE Photonics Technology Letters, 24(3), pp.

182-184, 2012. http://dx.doi.org/10.1109/LPT.2011.2175910 [18] Zou, S., Okonkwo, C.M., Cao, Z., Nguyen-Cac, T., Tangdiongga, E. and Koonen, T,. Dynamic optical routing and simultaneous generation of millimeter-wave signals for in-building access network, Optical Fiber Communication Conference and Exposition (OFC/NFOEC) and the National Fiber Optic Engineers Conference, pp. 1-3, 2012.

[19] Dat, P.T., Kanno, A., Inagaki, K. and Kawanishi, T., High-Capacity wireless backhaul network using seamless convergence of radioover-fiber and 90-GHz millimeter-wave, J. of Lightwave Technol., 32 (20) pp.

3910-3923, 2014. http://dx.doi.org/10.1109/JLT.2014.2315800 [20] Kyung W.L., Jung H.P. and Hyun D.J., Comparison of digitized and analog radio-over-fiber systems over WDM-PON networks, International Conference on ICT Convergence (ICTC), pp. 705-706, 2013.

[21] Aguiar, M., Gómez, J. y Torres, P., Modelamiento térmico y vibratorio de una cápsula para sensores de fibra óptica adaptables a mediciones en sistemas eléctricos de potencia. Revista DYNA, 76 (157), pp. 243-250, 2009.

G.A. Puerto-Leguizamón, received the BSc. in Telecommunications Engineering in 2002. He joined the Institute of Telecommunications and Multimedia Applications at the Universitat Politècnica de València in Spain, where he received the Advanced Research Studies in 2005 and the PhD. degree in 2008. As postdoctoral researcher he performed as co-leader of the workpackage about new generation of physical technologies for optical networks in the framework of the European funded project ALPHA (Architectures for Flexible Photonics Home and Access Networks). Since 2012 he is an Assistant Professor at the Universidad Distrital Francisco José de Caldas in Bogotá where he is with the Laboratory of Microwave, Electromagnetism and Radiation (LIMER). He has published more than 40 papers in journals and international conferences and he is a reviewer of the IEEE Journal on Lightwave Technologies and IEEE Photonic Technology Letters. His research interests include optical networking and radio over fiber systems.

L.C. Realpe-Mancipe, is student of the Universidad Distrital Francisco José de Caldas in Bogotá, Colombia, where she is pursuing the BSc. in Electronic Engineering. She is actively involved in the activities of the IEEE student branch at the university. In 2013 she joined the Laboratory of Microwave, Electromagnetism and Radiation (LIMER) as an assistant researcher in the framework of the research project “Dynamic architectures for converged optical access networks” where she is developing her final project. His research interests include optical networking and optical access networks.

C.A. Suárez-Fajardo, received the MSc. and PhD. degrees in Telecommunications Engineering from the Universitat Politècnica de València, Valencia, Spain, in 2003 and 2006, respectively. In 2006, he founded the Laboratory of Microwave, Electromagnetism and Radiation (LIMER) research group at the University Distrital Francisco José de Caldas in Bogotá, Colombia, and in 2007 he became as an associate professor at the same University. He has published more than 40 papers in journals and international conferences and he is a reviewer of the Chilean Journal Engineering and Journal of Antennas and Propagation (IJAP). His research interests include wideband and multi-band planar antenna design and optimization, microwave engineering, applied electromagnetic and small satellite communication systems.

144

Área Curricular de Ingeniería de Sistemas e Informática Oferta de Posgrados    

Especialización en Sistemas Especialización en Mercados de Energía Maestría en Ingeniería - Ingeniería de Sistemas Doctorado en Ingeniería- Sistema e Informática

Mayor información: Carlos Jaime Franco Cardona Director de Área curricular acsei_med@unal.edu.co (57-4) 425 5365

Influence of osmotic pre-treatment on convective drying of yellow pitahaya Alfredo Ayala-Aponte a, Liliana Serna-Cock b, Jimena Libreros-Triana a, Claudia Prieto a & Karina Di Scala c,d a School of Food Engineering. Universidad del Valle. Cali. Colombia, alfredo.ayala@correounivalle.edu.co Facultad de Ingeniería y Administración. Universidad Nacional de Colombia, Palmira. Colombia, lserna@unal.edu.co c Food Engineering Research Group. Universidad Nacional de Mar del Plata, Buenos Aires, Argentina, kdiscala@gmail.com d CONICET - Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina. b

Received: December 23th, 2013. Received in revised form: September 19th. 2014. Accepted: November 18th, 2014.

Abstract Cylinders of pitahayas were osmotically dehydrated in sucrose 55 % (w/w) during 45 minutes as a pretreatment of convective drying.

Hot air drying was performed at 50, 60 and 70ºC. Drying kinetics of osmodehydrated samples was compared to untreated samples.

Effective moisture diffusion coefficients were determined. Values of activation energy were 29.56 and 16.93 kJ mol-1 for pretreated samples and untreated samples, respectively. Mathematical modeling was applied to simulate the experimental drying curves of pitahayas. Results indicated that the Weibull model could be used to simulate experimental drying data. Furthermore, shrinkage of samples due to changes in volume was more pronounced for fruits dehydrated without pretreatment.

Keywords: drying kinetics; effective diffusion coefficient; mathematical modelling; shrinkage; pitahayas.

Influencia de un pre-tratamiento osmótico sobre el secado convectivo de pitahaya amarilla Resumen Cilindros de pitahayas fueron osmóticamente deshidratados en sacarosa 55 % (p/p) durante 45 minutos como un pre-tratamiento al secado convectivo. El secado por aire caliente se realizó a 50, 60 y 70ºC. Se comparó la cinética de secado de las muestras osmodeshidratadas con las muestras no tratadas. Se determinaron los coeficientes de difusión efectiva. Los valores de la energía de activación fueron 29.56 y 16.93 kJ mol-1 para las muestras pre-tratadas y para las no tratadas, respectivamente. Se aplicó modelado matemático para simular las curvas experimentales de secado de las pitahayas. Los resultados indicaron que el modelo de Weibull podría ser usado para simular los datos de secado experimental. Además, el encogimiento de las muestras debido a cambios en el volumen fue más pronunciado en las muestras sin pretratar.

Palabras clave: cinética de secado; coeficiente efectivo de difusión; modelado matemático; encogimiento; pitahayas.

1. Introduction Yellow pitahaya (Selenicereus megalanthus) is a tropical plant native to Central and South America which belongs to the Cactaceae family. Its fruits have yellow skin and a sweet and aromatic white flesh with small black seeds [1].

Growing exports in Colombia are driven by high demand in Europe and the Middle East. Moreover, this pitahaya is a polyphenol-rich fruit and a good source of antioxidants components, like phenolic compounds and ascorbic acid.

Therefore, its consumption may be associated with nutraceutical properties due to its antioxidant capacity generated by their effect on free radicals, reducing the risk of chronic diseases [2]. In addition, pitahaya is a source of

glucose, fructose, dietary fiber, vitamins, and minerals [3,4].

Because of its high water content, pitahaya is susceptible to deterioration, thus stabilization of the fruits requires reducing the moisture content up to certain level, at which microbial spoilage and chemical reactions deterioration are greatly minimised is desirable for its preservation [5].

Conventional air-drying is a simultaneous heat and mass transfer process and it is a high cost process. Moreover, it brings about substantial reduction in weight and volume, minimising packing, storage and transportation costs and enabling storability of the product under ambient temperatures [6]. However, during drying foods undergo physical, structural, chemical and nutritional changes that cause quality degradation [7].

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 145-151. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41321

Ayala-Aponte et al / DYNA 81 (188), pp. 145-151. December, 2014.

In recent years, osmotic dehydration (OD) has received increasing attention in the field of fruit preservation in order to reduce energy consumption and improve quality of fruit product [8,9]. Moreover, the effects of osmotic pretreatment on drying rates have been investigated by several authors and vary according to the raw material used and the drying conditions [10,11]. Sucrose is considered one of the best osmotic substances, especially when the OD is employed before drying. Dehydrated fruits have a rapid growing market due to consumer’s demands for high quality dry products [9].

In order to control and optimize the drying process, it is necessary to use mathematical equations to simulate water transport phenomena. Therefore, the purpose of this work was to evaluate the drying kinetics of pitahaya subjected to osmotic dehydration in sucrose solution and to compare with the drying kinetics of untreated fruits. Simulation of drying kinetics by means of mathematical models was evaluated. Moreover, effective diffusion coefficients were determined and the effects of drying temperature on kinetics parameters as well as on samples collapse were also studied.

2. Materials and methods 2.1. Preparation of raw material Yellow pitahayas (Selenicereus megalanthus) from the Department of Valle del Cauca, Colombia were used. Fruits were harvested at maturity state level four according to classification of Norma Técnica Colombiana (NTC) 3554 [12]. The fruits were washed with chlorinated water (200 μl/l) and peeled with stainless steel knifes. Then, they were cut into cylinders of 3 mm of length and 30 mm of diameter using a cutter machine (SKYMSEN, Poli, Brusque, Brasil) using a cylindrical stainless steel a cork borer.

2.2. Physico-chemical analysis The moisture content was determined by direct heating in a drying oven at 105 °C for 48 h according to the AOAC method 931.04 [13]. Soluble solids (ºBrix) were measured using a refractometer (ABBE, 1T, Tokio, Japan). The water activity was measured with a aw-meter (Aqualab, Serie 4TE, Pullman,WA, US) with an accuracy of ± 0.003. The sample volume (V) was determined using the Eq. (1). A caliper was used for measuring heights (h) and diameters (D). All measurements were done in triplicate.

V 

 D2 h 4

(1)

2.3. Osmotic dehydration The cylinders of pitahaya were immersed in the osmotic solution of sucrose 55 % (w/w) at 27±0.2°C during 45 minutes.

These osmotic treatment conditions were selected according to a prior study on kinetics of osmotic dehydration of yellow pitahaya fruit [15]. Experiments were performed with a constant magnetic agitation of 500 rpm with an agitation equipment (Kika Labor Technik Pol Co, US). The solution was agitated continuously with a magnetic stirrer to maintain a uniform temperature throughout the experiment, thus, enhancing equilibrium conditions. The weight ratio of osmotic solution to pitahaya was 15:1 to maintain a constant concentration of the osmotic solution during OD. After 45 minutes, the samples were removed from the solution, drained, and blotted with absorbent paper to remove the excess solution.

2.4. Convective dehydration The non-treated as well as the pretreated samples (dehydrated by osmotic dehydration) were put in a convective lab dryer (Armfield, UOP8, US) at three different air drying temperatures (50, 60 and 70ºC) employing a constant air flow of 0.77 m/s (perpendicular direction to samples). This dryer consists in a tunnel with an air flower. Samples were measured during drying at 10, 20, 30, 40, 50, 60, 80, 100, 130, 160, 190, 230, 270, 320, 370 and 420 min by means of an electronic balance (Ohaus, Adventurer, NJ,US) with an accuracy of ±0.01 g. Initial and final moisture content of samples were determined according to AOAC (1990).

2.5. Estimation of water diffusion coefficient Fick’s second diffusion law has been widely used to describe drying kinetics and osmotic drying kinetics for biological materials [7, 16]. In this model, the dependent variable is the moisture ratio (MR) which relates the sample moisture content in real time to initial moisture contents Eq. (3). In this study, equilibrium moisture was assumed to be negligible. When internal mass transfer is the controlling mechanism and unidimensional transport in a slab or a cylinder with constant effective diffusivity can be assumed, the solution of the Fick’s second law is given by Eq. (4) and Eq. (6). For sufficiently long drying times, the first term in the series expansion gives a good estimate of the solutions, Eq. (5) and Eq. (7) [17]. Based on Eqs. (4) and (6) and applying Newman’s rule for a finite cylinder (Eq.(8)), a relationship between logarithm of dimensionless moisture content for plate and cylinder and time is obtained, which can be used to determine water diffusion coefficient (Dwe).

Shrinkage of pitahaya during dehydration can be correlated with the dimensionless moisture content (Xwt/Xw0), by means of empirical polynomial equations according to the work of Mayor & Sereno, 2004 [14]: 3

MR 

(2)

146

(3)

For an infinite slab:

2

X  X  X  V  a  wt   b wt   c  wt   d V0 X X  w0   X w0   w0 

X wt X wo

MR 

  2i  12 Dwe  2t  1 exp    4 L2  2 i 0 2i  12   8



(4)

Ayala-Aponte et al / DYNA 81 (188), pp. 145-151. December, 2014.

MR 

  Dwe 2 t  exp   2 2  4L  8

and T is the absolute temperature (K). Eq. (13) can be linearized by applying natural log at both sides and a plot of ln Y versus 1/T produce a straight line, from which the activation energy (Ea, kJ mol-1) can be determined [20].

(5)

For an infinite cylinder:

2.7. Statistical evaluation of the models

MR 

MR

cyl

   Dwe t  exp  2   r 

4

2

 MR

inf plate

 MR

inf cyl

(6)

Fit quality of the proposed models for simulating the drying kinetics data was evaluated by means of statistical tests including determination correlation coefficient (r2) [Eq. (14)] and sum squared errors (SSE) [Eq. (15)].

(7) (8)

Where Xwt is the moisture content (g water g-1 dry matter), Xwo is the initial moisture content (g water g-1 dry matter), Dwe is the diffusion coefficient (m2 s-1), t is the drying time (s), L is the half-thickness of the slab (m), r is de radius of the cylinder and =2.405 is the root of Bessel function for n=1.

2.6. Modelling of drying kinetics Numerous mathematical models have been proposed to describe the characteristics of agricultural products during drying [6]. In this research, four of the more known models were used to fit the drying experimental data, including Newton (Eq 9), Henderson-Pabis (Eq. 10), Peleg (Eq. 11) and standardized Weibull model (Eq. 12). Applications of mentioned models can be found in different previous works [18,19]:

MR  exp(k1t )

(9)

MR  n1  exp(  k 2 t )

(10)

MR  1  (

1  t  )  X wo  A  B t 

N

( MRcalc,i  MRexp,i )

i 1

( MRexp,i  MRexp,i ) 2

r2  

SSE 

1 N 2  MRexp,i  MRcalc ,i  N i 1

 t  MR  exp      

  

(15)

3. Results and discussion 3.1. Physicochemical properties and experimental drying curves

(11)

1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0

50

0

A 

(14)

Where N is the number of data values and z is the number of constants.

Coefficients of Eq.. (2) were estimated by means of mathematical routines applying the function “lsqcurvefit” of the program Matlab 7.7.

Moisture Ratio

  2 Dwe t  4  MR   n exp  n 1  r2   

1

2

3

4

5

60

6

70

7

8

Time (h)

(12)

- Ea ) (13) RT Where R is the universal gas constant (8.314 J K-1 mol-1) Deff  Deff 0 exp(

147

Moisure Ratio

1

Where ki is the kinetic parameters (min-1), ni (i = 1…3) are the empirical parameters (dimensionless), α is the shape parameter (dimensionless) and β is the scale parameter (min) of the Weibull model, t is the drying time (min) and i is the number of terms.

In order to determine the influence of the process temperature on the diffusion coefficient (Dwe) and kinetics parameters, an Arrhenius-type equation was applied (Eq. 13).

50

0,8

60

70

0,6 0,4 0,2 0

B

0

1

2

3

4

5

6

7

8

Time (h) Figure 1. Experimental drying curves for pitahayas. A) without osmotic dehydration, B) with osmotic dehydration. Values are mean ± s.d. (n=3).

Source: from own data.

Ayala-Aponte et al / DYNA 81 (188), pp. 145-151. December, 2014.

Table 1 Coefficients of Eq (2), adjusted to represent the volumetric shrinkage rate of pitahaya at different conditions of temperature for samples with and without osmotic dehydration (OD).

Coeff.

Without OD 50 ºC 60 º C 70 ºC a 0.1025 0.0846 0.1648 b -0.4151 -0.3117 -0.6087 c 0.5493 0.4290 0.6988 d 0.5317 0.4693 0.3264 0.9803 0.9614 0.9425 r2 With OD a 0.0738 0.0011 0.1170 b -0.2829 -0.0276 -0.3592 c 0.3841 0.2179 0.4084 d 0.6991 0.6134 0.5226 0.9609 0.9933 0.9892 r2 Source: from own data.

0,9 0,7

50°C 60°C

aw

0,5

70°C 50°C OD

0,3

60°C OD

0,1

70°C OD

0

1

2

3

4

5

6

7

8

Time (h)

temperatures, for pitahayas fruits with and without osmotic dehydration.

The measurement and prediction of water activity provide the best available tool for evaluating the stability of foods. Thus, the end point of drying is the residual moisture content of the final product which ensures economic viability and microbiological safety, i.e. a water activity value lower than 0.60 [24,25]. Therefore, according to Singh & Heldman [26], values of water activity among 0.20 and 0.40 ensures the stability of the product storage against browning and hydrolytically reactions, liquid oxidation and enzymatic activity. Regarding Fig. 2, treated and untreated samples had water activity values between 0.20 and 0.40, except 50 C.

Figure 2. Water activity of pitahaya fruits as function of drying temperatures and process time. Values are mean ± s.d. (n=3). OD: with osmotic pretreatment.

Source: from own data.

0,0 0,00 -0,2

5,00

10,00

V

-0,4 -0,6 -0,8 50°C 70°C OD 60°C

Time (h)

60°C OD 50°C OD 70°C

Fig. 3 present the fruits volume variation (V=

Figure 3. Volume change of pitahaya fruits as function of drying temperatures. OD = osmotic dehydrated samples.

Source: from own data.

Moisture content of fresh pitahayas was evaluated as 2.63±0.15 kg water /kg dry matter. Total soluble solids were determined as 21.32±0.25 and 25.92±0.42 °Brix for fresh and osmo- dried pitahayas, respectively. Small seeds of the fruits were also part of the insoluble solids fraction (8.18 %).

Fig. 1 presents the experimental drying curves of pitahayas with and without osmotic dehydration. Since osmodehydrated samples have lower initial moisture content than fresh samples, process times needed to reach final water content is lower than those correspondents to untreated samples. Moreover, samples subjected to osmotic dehydration prior to convective drying has modified their internal structure affecting the action of different mass transport mechanisms, and thus contributing to enhance the water transport out of the solid [21]. Comparable results were reported for different foods subjected to osmotic dehydration like pumpkins [22] and pineapples [23].

3.2. Changes in water activity and volume samples Fig. 2 and 3 present the variation of water activity and volume with process time during air dehydration at different

V  V0 ) V0

during processing. Both treated and untreated samples exhibited changes in volume. Shrinkage of foodstuff during drying is unavoidable because heating and removal of water from the food matrix may cause stresses in the cellular structure, hence leading to structural collapse, changes in volume, shape deformation and capillaries contraction [14].

Fruit with osmotic dehydration showed changes in volume lower that those only process by convective dehydration. Thus, osmodehydrated samples exhibited less shrinkage. Volume changes during OD are mainly due to compositional changes and mechanical stresses associated to mass fluxes [14]. Ideally, it can be considered that the shrinkage of the material is equal to the volume of the removed water. Therefore, a mathematical relationship can be obtained that relates the volume shrinkage to the moisture content of the material.

According to Eq. (2), Table 1 shows the parameters related to this equation which showed satisfactory results (r2>0.9425).

Simulation of reduction of volume based on Eq.(2) indicated that shrinkage increases with decreasing moisture content in general, but shrinkage characteristics vary among drying products and drying methods. This is probably due to the unique biopolymer structure of individual agricultural product and the combined effect of process conditions that determines the type and extent of shrinkage [27].

148

Ayala-Aponte et al / DYNA 81 (188), pp. 145-151. December, 2014.

Comparable values for activation energy were reported in previous works for watermelon [35]; mango [34] and pear [36].

Table 2 Effective moisture diffusion coeficientes (m2/s) for pithayas.

Temp. (°C) With OD Without OD 50 1.849  10-10 2.082  10-10 -10 60 2.282  10-10 2.847  10 70 2.669  10-10 3.957  10-10 Source::from own data.

3.4. Mathematical modeling of drying curves

3.3. Determination of water diffusion coefficient (Dwe) Based on Fick’s equation, effective diffusion coefficient (Dwe) can be calculated from Eqs. (4) and (6) for each work temperature. Estimated Dwe values are informed in Table 2.

It can be observed that moisture diffusion coefficients for pitahaya subjected to osmotic dehydration prior convective drying showed higher values than non-treated samples for all the temperatures under study. During osmotic dehydration, many aspects of cell structures are affected such as alteration of cell walls, splitting of the middle lamella, lysis of membranes, tissue shrinkage which could strongly influence the transport properties of the product during processing [28].

Comparable values were reported by other researchers related to osmodried fruits and vegetables: carambola [11], carrot cubes [29], West Indian cherry [30], apricots [16], apples [31], pomegranate arils [32] and Aloe Vera [33].

From the relationship of water diffusion coefficients and drying temperature, activation energy from the inverse slope of the line plot of ln Dwe versus T-1 can be obtained. In this research, activation energy values of 16.94 kJ/mol (r2=0.999) and 29.56 kJ/mol (r2=0.995) were obtained for untreated and treated samples, respectively. The Ea is a measure of dependence of the mass transfer process on temperature; lager magnitude of Ea is associated with higher temperature dependence. Higher activation energy therefore implies greater temperature sensitivity, and a smaller temperature change is needed for the mass transport to proceed more rapidly. The obtained values are within the range (0-63 kJ/mol) for diffusion-controlled processes [34].

3.5. Statistical analysis of models Table 4 shows the results of statistical tests (r2 and SSE) performed to the proposed models. These statistical tests evaluate the goodness of fit on the experimental data and they have been reported by other researchers during food drying analysis [19]. All the proposed models showed a good fit with high values of r2 (>0.90) and values close to zero for SSE. According to these results, the models that best fitted the experimental data, considering the statistical test applied, were the Peleg model (r2 = 0.9904; SSE= 0.0011) and Weibull (r2 =0.9957; SSE= 0.1040).

149

Estimated MR (Weibull)

Table 3.

Coefficients of Eqs (9-12) to represent the drying kinetics of pitahaya at different conditions of temperature for samples with and without osmotic dehydration (OD).

Without OD Eq. Coeff.

50ºC 60ºC 70ºC (9) k1 (1/h) 0.50±0.02 0.32 0.03 0.40 .02 (10) k2 (1/h) 0.340.01 0.420.01 0.51 0.01 n 1.040.06 1.040.03 1.020.02 (11) A -3.24±0.58 -2.44±0.27 -1.83±0.14 B -0.59±0.09 -0.65±0.05 -0.72±0.02 (12) 1.150.141 1.110.05 1.080.06   (1/h) 3.060.22 2.480.11 2.010.08 Without OD (9) k1 (1/h) 0.360.05 0.440.06 0.71 0.01 (10) k2 (1/h) 0.370.01 0.450.02 0.740.001 n 1.030.07 1.030.07 1.030.01 (11) A -2.84±0.72 -2.26±0.66 -1.17±0.02 B -0.63±0.11 -0.66±0.12 -0.78±0.01 (12) 1.090.17 1.120.20 1.090.02   (1/h) 2.800.31 2.270.29 1.390.01 Source: from own data.

Tables 3 shows the average values and standard deviations of the kinetic and empirical parameters ki (i = 1, 2), n, A, B, α and β, obtained for all the proposed models.

Except for de n and , a tendency with increasing temperature was observed for each of the rest of the parameters, since an increase in drying air temperature showed an increase in the parameters values.

The -parameter of Weibull is related to the velocity of the mass transfer at the beginning, e.g., the lower the  value, the faster the drying rate at the beginning [19]. In addition, parameter  decreases as temperature increases for both untreated and treated samples. Some authors suggested that parameter  represents the time needed to accomplish approximately 63% of the process [37]. Comparable results were reported in previous works for pepino fruit [6,19].

Regarding to parameter A for Peleg model, it shows a clear tendency to increase as temperature increases indicating that the higher the temperature the higher the water absorption rate. Similar tendencies for this parameter have been reported for chestnuts [38].

50 ºC

60 ºC

70 ºC

50ºC OD

60ºC OD

70 ºC OD

1,2 1 0,8 0,6 0,4 0,2 0 0

0,2

0,4

0,6

0,8

1

Experimental MR Figure 4. Experimental and Weibull-simulated moisture ratio (MR) for dehydrated pitahaya fruits.

Source: from own data.

Ayala-Aponte et al / DYNA 81 (188), pp. 145-151. December, 2014.

Table 4 Statistical tests performed to selected models to simulate the drying curves of pitahaya fruits.

50ºC 60ºC 70ºC 50ºC MODEL r2 SSE r2 SSE r2 SSE r2 SSE Newton 0.9830 0.0980 0.9933 0.1057 0.9945 0.1129 0.9866 0.0980 Henderson-Pabis 0.9921 0.0980 0.9967 0.1057 0.9963 0.1084 0.9947 0.0985 Peleg 0.9999 0.0007 0.9923 0.0008 0.9909 0.0010 0.9901 0.0801 Weibull 0.9920 0.0980 0.9978 0.1057 0.9971 0.1084 0.9945 0.0985 Source: From own data.

Therefore, Fig. 4 shows the experimental MR versus the corresponding estimate by the Weibull model to notice the goodness of fit of this mathematical model for all drying working temperatures, for both samples with and without osmotic dehydration. Similar observations were made by Cunha et al., (2001) modelling water losses during osmotic dehydration of apple [39], Corzo et al. (2008) [40] for air drying of coroba slices and Uribe et al., [19], for convective dehydration of pepino fruits.

4. Conclusion

range of drying conditions under studied.

Therefore, osmotic dehydration combined with

convective dehydration provides an opportunity to produce novel shelf stable high quality pitahaya fruits for the local as well as for export markets.

The authors thank the Ministry of Agriculture and Rural Development of Colombia for the support provided to this article.

[4]

[8]

[10]

[11]

[12] [13] [14]

[16]

[17]

References

[3]

[7]

[15]

Acknowledgements

[2]

[6]

[9]

Osmotic dehydration with sucrose 55 % (w/w) during 45 min was applied in order to evaluate the influence of pretreatment on convective drying rates of pitahayas at three different temperatures. Due to modifications in structure, osmodehydrated samples exhibited increasing drying rates with effective moisture coefficients greater that non-treated samples. Shrinkage due to variation in volume samples was more pronounced in untreated samples and was satisfactorily modeled by means of a polynomial equation.

Different mathematical models were used to simulate experimental drying characteristics. Based on statistical results, the Peleg and Weibull models can be appropriate used to simulate experimental drying curves in the

[1]

[5]

Ayala-Aponte, A., Serna Cock, L. and Rodriguez-de la Pava, G., Moisture adsorption in yellow pitahaya (Selenicereus megalanthus).

DYNA, 78 (170), pp.7-14, 2011.

Beltran, M.C., Oliva-Coba, T.G., Gallardo-Velasquez, T. and Osorio-Revilla, G., Ascorbic acid, phenolic content and antioxidant capacity red, cherry, yellow and white types of pitahaya cactus fruit (Stenocereus stellatus Riccobono). Agrociencia, 43, pp. 153-162, 2009.

Barbeu, G., The strawberry pear, a new tropical fruit. Fruits, 45, pp.

141-147, 1990.

Wu, M.C. and Chen, C.S., Variation of sugar content in various parts of pitahaya fruit. Proceedings of the Florida State Horticultural Society, 110, pp. 225-227, 1997.

[18] [19]

[20]

[21]

150

60ºC r2 SSE 0.9872 0.0941 0.9948 0.1097 0.9910 0.0011 0.9963 0.1098

70ºC r2 SSE 0.9945 0.1128 0.9955 0.1128 0.9857 0.0016 0.9964 0.1128

Sacilik, K. and Elicin, A.K., The thin layer drying characteristics of organic apple slices. Journal of Food Engineering, 73 (3), pp. 281289, 2006. http://dx.doi.org/10.1016/j.jfoodeng.2005.03.024 Doymaz, I., Convective drying kinetics of strawberry. Chemical Engineering and Processing, 47 (5), pp. 914-919, 2008. http://dx.doi.org/10.1016/j.cep.2007.02.003 Di Scala, K. and Crapiste, H., Drying kinetics and quality changes during drying of red pepper. LWT- Food Science and Technology, 41, pp. 789-795, 2008.

Gomes, A., Lucena-Barbosa Jr., J., Colato, A.G. and Xidieh-Murr, F.E., Osmotic dehydration of acerola fruit (Malpighia punicifolia L.). Journal of Food Engineering 68, pp. 99-103, 2005. http://dx.doi.org/10.1016/j.jfoodeng.2004.05.042 Pereira-da Silva, W., Silva-do Amaral, D., Duarte, M., Mata, M., Silva, C., Pinheiro, R. and Pessoa, T., Description of the osmotic dehydration and convective drying of coconut (Cocos nucifera L.) Pieces: A three-dimensional approach. Journal of Food Engineering, 115, pp.

121-131, 2013. http://dx.doi.org/10.1016/j.jfoodeng.2012.10.007 Ashwini, N.B., Sowbhagya, H.B. and Rastogi, N.K., Osmotic dehydration assisted impregnation of curcuminoids in coconut slices. Journal of Food Engineering, 105 (3), pp. 453-459, 2011. http://dx.doi.org/10.1016/j.jfoodeng.2011.03.002 Ruiz-López, I.I., Ruiz-Espinosa, H., Herman-Lara, E. and ZárateCastillo, G., Modeling of kinetics, equilibrium and distribution data of osmotically dehydrated carambola (Averrhoa carambola L.) in sugar solutions. Journal of Food Engineering 104, pp. 218-226, 2011. http://dx.doi.org/10.1016/j.jfoodeng.2010.12.013 ICONTEC. Norma Técnica Colombiana. NTC 3554. Frutas frescas.

Pitahaya amarilla. Bogotá: ICONTEC, pp. 1-14, 1996.

AOAC. Association of Official Analytical Chemists. Official method of analysis, Association of Official Analytical Chemists No. 934.06 (15th Ed), Arlington, MA, Washington, 1990.

Mayor, L. and Sereno, A.M., Modelling shrinkage during convective drying of food materials: A review. Journal of Food Engineering, 61(3), pp. 373-386, 2004. http://dx.doi.org/10.1016/S02608774(03)00144-4 Ayala-Aponte, A., Giraldo, C.J. and Serna-Cock, L., Kinetics of osmotic dehydration of yellow pitahaya fruit (Selenicereus megalanthus). Interciencia, 35, pp. 539-544, 2010.

Ispir, A. and Togrul, I.T., Osmotic dehydration of apricot: Kinetics and the effect of process parameters. Chemical Engineering Research and Design, 87, pp.

166-180, 2009. http://dx.doi.org/10.1016/j.cherd.2008.07.011 Crank, J., The mathematics of diffusion, second ed. Oxford University Press, London, UK, 1975.

Peleg, M., An empirical model for the description of moisture sorption curves. Journal of Food Science, 53 (4), 1216-1219, 1998. http://dx.doi.org/10.1111/j.1365-2621.1988.tb13565.x Uribe, E., Vega-Gálvez, A., Di Scala, K., Oyanadel, R., Saavedra, J. and Miranda, M., Characteristics of convective drying of pepino fruit (Solanum muricatum Ait.): Application of weibull distribution.

Food and Bioprocess Technology, 4, pp. 1349-1356, 2011. http://dx.doi.org/10.1007/s11947-009-0230-y Sharma, G.P. and Prasad, S., Effective moisture diffusivity of garlic cloves underging microwave-convective drying. Journal of Food Engineering, 65, pp.

609-617, 2004. http://dx.doi.org/10.1016/j.jfoodeng.2004.02.027 Chiralt, A. and Fito, P., Transport mechanisms in osmotic dehydration: The role of the structure. Food Science and Technology

Ayala-Aponte et al / DYNA 81 (188), pp. 145-151. December, 2014.

[22]

[23]

[24]

[25]

[26] [27]

[28]

[29]

[30] [31]

[32]

[33] [34]

[35]

[36] [37]

[38] [39]

International, 9 (3), pp.

179-186, 2003. http://dx.doi.org/10.1177/1082013203034757 Garcia, C.C., Mauro, M.A. and Kimura, M. Kinetics of osmotic dehydration and air-drying of pumpkins (Cucurbita moschata).

Journal of Food Engineering, 82 (3), pp. 284-291, 2007. http://dx.doi.org/10.1016/j.jfoodeng.2007.02.004 Lombard, G.E., Oliveira, J.C., Fito, P. and André, A., Osmotic dehydration of pineapple as a pre-treatment for further drying.

Journal of Food Engineering, 85 (2), pp. 277-284, 2008. http://dx.doi.org/10.1016/j.jfoodeng.2007.07.009 Manjarres-Pinzón, K., Cortes-Rodriguez, M. and RodríguezSandoval, E., Effect of drying conditions on the physical properties of impregnated orange peel. Brazilian Journal of Chemical Engineering, 30 (3), pp.

667-676, 2013. http://dx.doi.org/10.1590/S0104-66322013000300023 Proton, F. and Ahrné, L., Application of the Guggenheim, Anderson and De Boer model to correlate water activity and moisture content during osmotic dehydration of apples. Journal of Food Engineering, 61 (3), pp. 467-470, 2004. http://dx.doi.org/10.1016/S02608774(03)00119-5 Singh, R.P. and Heldman, D.R., Introduction to food engineering, 2nd ed. Academic Press, Inc., San Diego, pp.139-141, 1993.

Ong, S.P. and Law, C.L., Hygrothermal properties of various foods, vegetables and fruits, in drying of foods, vegetables and fruits – Vol. 1, en Ed. Jangam, S.V., Law, C.L. and Mujumdar, A.S., ISBN - 978981-08-6759-1, Published in Singapore, pp. 31-58, 2012.

Azarpazhooh, E. and Ramaswamy, H., modeling and optimization of microwave osmotic dehydration of apple cylinders under continuous-flow spray mode processing conditions. Food and Bioprocess Technology, 5 (5), pp. 1486-1501, 2012. http://dx.doi.org/10.1007/s11947-010-0471-9 Singh, B., Kumar, A. and Gupta, A.K., Study of mass transfer kinetics and effective diffusivity during osmotic dehydration of carrot cubes. Journal of Food Engineering, 79 (2), pp. 471-480, 2007. http://dx.doi.org/10.1016/j.jfoodeng.2006.01.074 http://dx.doi.org/10.1016/j.jfoodeng.2006.01.073 Silva, M., Da Silva, Z., Mariani, V. and Darche, S., Mass transfer during the osmotic dehydration of West Indian cherry. LWT - Food Science and Technology, 45(2), pp. 246-252, 2012.

Souraki, B.A., Ghavami, M. and Tondro, H., Correction of moisture and sucrose effective diffusivities for shrinkage during osmotic dehydration of apple in sucrose solution. Food and Bioproducts Processing, 92 (1), pp.

1-8, 2014. http://dx.doi.org/10.1016/j.fbp.2013.07.002 Mundada, M., Bahadur-Singh, H. and Swati, M., Convective dehydration kinetics of osmotically pretreated pomegranate arils.

Biosystems Engineering, 107, pp.

307-310, 2010. http://dx.doi.org/10.1016/j.biosystemseng.2010.09.002 Pisalkar, P.S., Jain, N.K. and Jain, S.K., Osmo-air drying of aloe vera gel cubes. Journal of Food Science and Technology, 48 (2), pp.

183-189, 2011. http://dx.doi.org/10.1007/s13197-010-0121-2 Alakali, J.S., Ariahu, C.C. and Nkpa, C.C., Kinetics of osmotic dehydration of mango. Journal of Food Processing and Preservation, 30, pp. 597-607, 2006. http://dx.doi.org/10.1111/j.17454549.2006.00080.x Falade, G. and Ayanwuyi, F.A., Kinetics of mass transfer, and colour changes during osmotic dehydration of watermelon. Journal of Food Engineering, 80, pp.

979-985, 2007. http://dx.doi.org/10.1016/j.jfoodeng.2006.06.033 Park, J., Bin, A. and Reis-Brod, F.P., Drying of pear d’Anjou with and without osmotic dehydration. Journal of Food Engineering, 56, pp. 97-103 2002. http://dx.doi.org/10.1016/S0260-8774(02)00152-8 Marabi, A., Livings, S., Jacobsons, M. and Saguy, I.S., Normalized weibull distribution for modeling rehydration of food particulates.

European Food Research Technology, 217, pp. 311-318, 2003. http://dx.doi.org/10.1007/s00217-003-0719-y Moreira, R., Chenlo, F., Torres, M. and Vázques, G., Effect of stirring in the osmotic dehydration of chestnut using glycerol solutions. Journal of Food Science and Technology, 5, pp. 1507-1514, 2007.

Cunha, L.M., Oliveira, F.A.R., Aboim, A.P. and Frías, J.M., Stochastic approach to the modelling of water losses during osmotic dehydration and improved parameter estimation. International

Journal of Food Science and Technology, 36, pp. 253-262, 2001. http://dx.doi.org/10.1046/j.1365-2621.2001.t01-1-00447.x [40] Corzo, O., Bracho, N., Pereira, A. and Vásquez, A., Weibull distribution for modeling air drying of coroba slices. LWT-Food Science and Technology, 41(10), pp. 2023-2028, 2008. http://dx.doi.org/10.1016/j.lwt.2008.01.002 A.A. Ayala-Aponte, received the BSc. Eng in Agricultural Engineering in 1993 from the Universidad del Valle, Cali, Colombia and the PhD degree in Science and Food Technology in 2011 from the Universidad Politécnica de Valencia, España. He is a professor in the area of Food Technology and Engineering, in the Universidad del Valle, Cali, Colombia. His research interests include: preservation and food processing.

L. Serna-Cock, received the BSc. in Bacteriology from the Universidad Católica de Manizales, Colombia in 1987, and the PhD degree in Food Engineering from the Universidad del Valle, Cali, Colombia. She is a professor in the area of Biotechnology in the Universidad Nacional de Colombia, Palmira, Colombia. Her research interests include: preservation, food processing and biotechnology.

J. Libreros-Triana, received the BSc. in Food Engineering from the Universidad del Valle, Cali, Colombia in 2010. She works in the food industry. Her research interests include: preservation and food processing.

C.M. Prieto, received the BSc. in Food Engineering from the Universidad del Valle, Cali, Colombia in 2010. She works in the food industry. Her research interests include: preservation and food processing.

K. Di Scala, received the BSc. Eng in Chemical Engineering in 1997 from the Universidad Nacional de Mar del Plata, Buenos Aires, Argentina, and the PhD degree in Chemical Engineering in 2006 from the Universidad Nacional del Sur, Bahía Blanca, Buenos Aires, Argentina. She is an assistant professor in the area of Food Engineering, in the Universidad Nacional de Mar del Plata Buenos Aires, Argentina. Her research interests include: food processing, food quality, and mathematical simulation and optimization of food processes.

151

Área Curricular de Ingeniería Química e Ingeniería de Petróleos Oferta de Posgrados   

Maestría en Ingeniería - Ingeniería Química Maestría en Ingeniería - Ingeniería de Petróleos Doctorado en Ingeniería - Sistemas Energéticos Mayor información:

Abel de Jesús Naranjo Agudelo Director de Área curricular qcaypet_med@unal.edu.co (57-4) 425 5317

Evaluation of thermal behavior for an asymmetric greenhouse by means of dynamic simulations Ervin Alvarez-Sánchez a, Gustavo Leyva-Retureta b, Edgar Portilla-Flores c & Andres López-Velázquez d a

Facultad de Ingeniería Mecánica y Eléctrica, Universidad Veracruzana, Veracruz, México, eralvarez@uv.m Facultad de Ingeniería Mecánica y Eléctrica, Universidad Veracruzana, Veracruz, México, guleyva@uv.mx c Instituto Politécnico Nacional, CIDETEC, México D.F, México, aportilla@ipn.mx d Facultad de Ingeniería Mecánica y Eléctrica, Universidad Veracruzana, Veracruz, México, andlopez@uv.mx b

Received: December 28th, 2013. Received in revised form: June 22th, 2014. Accepted: October 31th, 2014

Abstract In this paper is presented the design of an asymmetric greenhouse, with a simulation and evaluation of its microclimate by means of DesignBuilder®. Climatic and geographical conditions of the greenhouse location are processed with DesignBuilder®, which also allows to include the geometrical characteristics of the enclosure and its roofing, the properties of its building materials and a description of the climate control system. Three configurations of this asymmetric greenhouse are evaluated in order to analyze and compare their internal gains.

Keywords: asymmetric greenhouse, climate control, dynamic simulation, microclimate.

Evaluación del comportamiento térmico de un invernadero asimétrico mediante simulaciones dinámicas Resumen En este trabajo se lleva a cabo el diseño, simulación y evaluación del comportamiento del microclima de un invernadero asimétrico mediante el software DesignBuilder®, incorporando las condiciones climáticas y geográficas del lugar donde se pretende instalar el invernadero, sus características geométricas y las propiedades de los materiales del recinto y su techado, así como del sistema de control de climatización. Se evaluaron tres diferentes configuraciones para la envolvente del invernadero para comparar el comportamiento de sus ganancias internas.

Palabras clave: control de climatización, invernadero asimétrico, microclima, simulación dinámica.

1. Introduction An advantage of farming based on greenhouses is that it makes possible to supervise closely every factor related with crop behavior [1], even in zones with low quality soils if a hydroponic technique is used [2]. Additionally, inner temperature, ambient humidity and soil humidity can be sensed in an automatic way, to send all this data to a control system that modifies the internal conditions of the greenhouse. This represents an advantage in respect to open field cultivation, since the damage on crops due to plagues is minimized [3].

Accordingly to IDAE (Instituto para la Diversificación y Ahorro de Energía, Spain),the study on energy efficiency of

the greenhouses is vital for country development, as it could be a great support for natural resources conservation and the reduction on the emission of greenhouse gases, generating big savings on energy consumption in parallel.

The thermal performance of a greenhouse depends basically on climate conditions, structural shape, orientation and its building materials [4]. The analysis of thermal performance is a complex task, so a software for developing dynamic simulations is needed, such as DesignBuilder [5], [6], EnergyPlus [7] or Ecotect Analysis [8].

In recent years, computational fluid dynamics has been used to develop numerical models that improve the understanding of the variables interaction that make up climate inside greenhouses. In addition, the development of

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 152-159. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41338

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

more powerful software and hardware has been a key to the generation of improved results in the past five years [9].

An understanding of energy behavior is important in order to design efficient greenhouses that include climate system automation; for example, there is an Energy Audit protocol for greenhouses based on energy balance by a mathematical model for heat transfer phenomena, that serves to analyze the nature of the thermal and energetic behavior in terms of its interacting variables, as solar radiation, construction materials, and active cooling systems [10].

In this paper, the thermal behavior of three kinds of greenhouses is studied. Each enclosure is built using different covers with passive climate control systems, applying similar criteria to those used for buildings design in order to decrease the necessity of systems for heating or cooling, which require a huge amounts of energy. The organization of this work is as follows: Section II gives a theoretical framework about earth's movement and thermal phenomena in greenhouses.

Section III contains an explanation about the asymmetric greenhouses design, while simulation parameters are considered in Section IV. Finally, results and conclusions are included in Section V. 2. Theoretical framework Different techniques and technologies are applied to crop production in greenhouses, to modify the regular cycles of cultivation and increase the product quality and the number of harvesting periods, in order to improve the sale conditions [11]. This is possible by means of automated systems for irrigation and controls for temperature and relative humidity, to help farmers to optimize their productive resources, diminishing water consumption and electric energy demand in order to lower production costs [12].

The use of production systems with greenhouses in Mexico is extended to 24 of his 32 states. However, approximately 75% of this production is concentrated in just five entities: Sinaloa, North Baja California, South Baja California, Sonora and Jalisco [13], with tomato as main crop [14]. Applied technologies include irrigation, automatic ventilation, water heating, thermal screens, automatic control and hydroponics. Most Mexican regions have optimal conditions for greenhouse production, because of long duration of days and enough intensity of winter's solar energy [15].

The earth rotates around her polar axis and an angular change is produced between the equatorial plane and the lines passing through earth center and sun center, causing a solar declination (Fig. 1).

This phenomenon is magnified during the solstices because in summer there is a 23.5° solar declination that produces long days, while in winter the angular magnitude is the same but in the opposite direction, i.e. -23.5°, causing short days. However, during vernal and autumnal equinoxes the solar declination is zero because the sun is over the celestial equator, which implies that day and night have the same number of hours.

Figure 1. Sun’s apparent motion.

Source: Adapted from Strobel, N., 2013. [16]

The geographic location of the greenhouse also affects its design, since sunlight is perpendicular to equator but deviates accordingly to the location on North or South due to earth curvature. So, the first parameter for designing the solar collection surface is the latitude (L), which is the angle between the equatorial line and the meridian where the installation site is located.

2.1. Thermal phenomena in greenhouses The envelope and the internal elements of a greenhouse affect the difference between the inner and outer climates, producing an energetic interchange that defines its thermal and environmental behavior [17]. For this reason, the inner microclimate needs to be regulated in order to obtain optimal conditions for crops growth.

The relation between the inner and outer maximum temperatures is obtained by an adequate selection of materials for walls, floor and roof, especially considering their thickness. However, mechanical systems for heating or cooling are required when the external climate conditions do not allow optimal inner conditions. The passive design helps to minimize the use of these mechanical systems and to diminish the energy required for its operation [18].

The glazing system is the simplest and most economical passive design. The strategy is based on an adequate positioning of the glazing in order to have a maximum solar radiation in winter and a minimum solar radiation in summer, so greenhouse geometry is an important aspect.

The geometry parameters of the enclosure for an adequate operation include length, width and height [18].

2.2. Computational fluid dynamics and the measuring of greenhouses variables In the last decade, the use of numerical methods such as the Computational Fluid Dynamics (CFD) which is based on Navier-Stokes equations, has proved to be a good tool for developing models in order to understand relationships among the interacting variables in the behavior of climate for a greenhouse [9].

153

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

Simulations are used frequently to predict the behavior of the greenhouses; several studies used computational dynamics analysis to investigate the climatic conditions inside them.

These computational tools increase the degree of realism by simulating insect-proof environments (among other 3D models) and analyzing their effect on crops, considering it as a porous medium. The results have improved our understanding of the phenomenon in a greenhouse.

The computational dynamics produces a model for simulating night-time climate and condensation in the greenhouse. The model was applied to a enclosure with a four-span plastic cover. The condensation film was simulated by adding a user defined function (UDF) to the commercial CFD software. The results showed the importance of heat transfer losses by radiation, particularly for low values of soil heat flux (SHF); they also showed the roof was the coolest surface in the greenhouse, and therefore it is the condenser and sink for the water vapor produced by the crop. The CFD condensation model is intended to be used for the design of strategies for humidity control, particularly in unheated greenhouses [19].

There are numerical simulations for the distribution of climate parameters within a tomato greenhouse ventilated by tunnel, with variable outside conditions. The boundary climatic conditions were determined by experimental measurements, and the sun position was calculated for every time interval considered. Results are presented for spring equinox and summer solstice; these results highlight the combined influence of sun position, wind direction and the greenhouse microclimate. In this work, the possibility of using this methodology as conceptual tool for designers or in association with a control climate model for farmers is discussed [20].

As mentioned before, another important item is the study and measure of the variables interacting on a greenhouse. Advanced systems monitor these variables using wireless sensor networks that consider humidity, temperature, light, and volumetric water content in the soil.

These systems are flexible enough to be adapted to any greenhouse; since they are based on wireless technology, their nodes can establish links automatically, and have implemented functions for saving energy which extend the life of batteries up to a crop year without maintenance [21].

Figure 2. Cover’s angle for equinoxes.

Source: Own elaboration

Figure 3. Cover’s angle for winter.

Source: Own elaboration

In the north hemisphere exists a solar declination of approximately -23.5° during the solstices, as shown in Fig. 3, which produces a slope angle βz = 42.5°.

The ideal slope for the cover can be calculated from this adjustment, to receive the solar radiation perpendicularly at noon. Since in equinoxes the angle is 19° and in solstices it is approximately 43°, the selection of an angle between this two values is needed. For simulation purposes the slope angle selected is 26°, because an angle near to 43° could produce instability in the greenhouse structure under strong winds.

3. Asymmetric greenhouse design The construction virtual site is near Mexico City, due to the low solar-radiation produced by cloudiness, the high pollution rate and the predominant low temperatures [22]. In order to increase the solar radiation received, the greenhouse is designed with a larger area on the roof side facing south, with an orientation that is parallel to sun's apparent movement, i.e. east to west. The slope on this south face produces minimal variations of the solar energy collected during the year. The slope angle (βz) is defined by means of the difference between the latitude (L) of the location site and the solar declination (d), i.e. βz = L – d.

Taking into account that latitude of Mexico City is 19° and that in equinoxes the solar declination is zero, the resultant slope angle is 19°, as shown in Fig. 2.

Figure 4. Final design of asymmetric greenhouse.

Source: Own elaboration

154

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

Fig. 4 shows the designed asymmetric greenhouse. The enclosure has a bamboo structure with two modules of 9 m x 8 m, a lateral height of 3 m and 5 m in the ridge, with 144 m2 of total area and a total volume of 576 m3. The dimensions were selected taking as a base the maximum longitude of the bamboo beams, which is 12 m.

4. Simulation parameters In order to carry out dynamic simulations with DesignBuilder is necessary to define important parameters such as climatic data, characteristics of the materials of the cover, and HVAC systems.

Figure 61. Ground temperatures in DesignBuilder.

Source: Adapted from U.S. Department of Energy.

4.1. Climate data The climate data of Mexico City for this work consist of a TMY3 that was made from the 1991 to 2005 database; the climate file is under the designation of the World Meteorological Organization (WMO) and the source format is IWEC (International Weather for Energy Calculations). In Fig. 5, the weather can be seen in the template for DesignBuilder, which is necessary to describe the location of data such as latitude, longitude and altitude above sea level; this climate template is created with the data for every hour.

Fig. 6 shows the template field soil temperatures obtained with the EcotectAnalisys software; this is an important parameter since represents the heat transfer by conduction through soil terrain, and soil temperature varies depending on the year. So it would be a mistake to consider the soil temperature as constant throughout the year; it can be seen that it is coldest during the month of January with 15 ° C, while in July is the hottest, with 18.7 ° C.

templates, that were developed for polycarbonate which has a total solar transmission of 0.780, a light transmission of 0.790 and U-value (thermal conductivity) of 3.50 W / m 2K. In the case of concrete blocks, the template included in the package is used.

4.3. HVAC Systems HVAC systems are divided into active heating and cooling; the heating system runs on natural gas, with a COP of 0.83 and a flow temperature of 35 ° C, the system is programmed to operate whenever the inner temperature is 22 ° C; the cooling system operates with electric power, has a COP of 1.67 and air drive ranges between 7 ° C and 7.5 ° C, and is scheduled to run whenever the inner temperature reaches 26 ° C.

Figure 7. Template glazing.

Source: Adapted from U.S. Department of Energy.

Figure 5. Template climate of Mexico City.

Source: Adapted from U.S. Department of Energy.

4.4. Shading System

4.2. Characteristics of the building materials Templates and their glazing materials are created to simulate the characteristics of the walls and cover with DesignBuilder. Fig. 7 shows the properties of the glazing

Among the four types of shading devices provided by DesignBuilder, slatted blinds are chosen because they block the direct radiation and reduce the solar gain; however they allow to pass a portion of the diffuse radiation, which serves to decrease lighting property, as would do a diffusing screen or curtain.

155

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

For this case was used a template already defined in the software, called Blind slats with medium reflectivity. Its properties are established as follows: the shading screen distance is 0.05m, a horizontal orientation, the width of the blind slats is 0.0250m, the spacing between slats is 0.0188m, with a 0.0010m thickness, and the tilt angle with respect to glazing is 45 °.

Finally, an indoor air temperature control system (OnIfHighZoneAirTemp) was used; this is activated whenever the temperature of the air inside the area during the previous simulation step exceeds the control value (C or F).

4. 5 Inner-air set point for temperature control Simulations with DesignBuilder-EnergyPlus generate extensive data on environmental conditions within the enclosure. The set point temperature defines the ideal temperature (i.e. the setting of the heating thermostat) in the range when heating or cooling is required; its interpretation depends on the Temperature control calculation option. For this case the Internal operative temperature was considered; this is calculated by the mean of the internal air and radiant temperatures, where Internal air temperature is the calculated average temperature of the air inside and the Internal radiant temperature of the zone is calculated as if the sensor is in the center of the zone, with no weighting for any particular surface.

5. Results and discussion The simulations were developed using DesignBuilder® and the weather data archive for Mexico City, obtained from [23]. This archive is based on measurements obtained by weather stations and includes temperature, ambient humidity, solar radiation and wind velocity for each hour of year 2002, which is a representative year of typical weather conditions for the zone.

From the average temperatures for that year it can be noticed that the coldest months were January and December, while the hottest were April and May. The maximum temperature reached was 19.1°C in May and the minimum was 13.6°C in December.

In order to show the behavior of the designed greenhouse, three configurations were developed with polycarbonate as a cover due to its thermal conductance and its common use for building greenhouses [23]. The crop selected as reference was tomato, with an optimal growth temperature during daylight between 21°C and 27°C [24].

The basic configuration (A) for the asymmetric greenhouse, constructed with bamboo, has a polycarbonate envelope both in enclosures as in the roof. The second one (B) has the same structural materials, but also includes a climate control system programmed to maintain the inner temperature between 22°C and 26°C in every time.

Finally, the third configuration (C) has concrete walls 1.5m height on south, east, and west sides, while the north wall is completely of concrete and the remaining structure is constituted by glazing with polycarbonate. For inner weather control this configuration uses active and passive climate control systems.

Figure 8. Thermal behavior for the three greenhouse configurations.

Source: Own elaboration

The active climate control is a heating system programmed to work when the temperature is under 22°C, only in November and December. From March to October

156

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

Table 1.

Construction cost for configurations A and B. Cost Configuration Area (m2) (USD/m2) A 144 20.42 B 144 35.74 Source: Own elaboration

Table 2.

Construction cost for C configuration.

Area (m2) Cost (USD/m2) 144 37.74 No. Blocks Cost per unit($) 1535 0.503 Source: Own elaboration

Total (USD) 2,940.48 5,146.56

Total (USD) 5,434.56 Total ($) 772.10

Table 3.

Operating costs for B configuration Heating Cost Kwh year 34,333.72 Cooling Cost

MMBtu year 117.15

Kwh year 28,130.78 Source: Own elaboration

Figure 9. Inner gains for greenhouses A, B and C. Source: Own elaboration

the passive climate control, consisting of external shading curtains, is activated when the temperature is above 23°C. The thermal behavior for the three configurations is shown in Fig. 8. In configuration A the temperature oscillates from 20.15°C to 27.81°C, which is out of the optimal limits for crop growth.

The temperature in configuration B oscillates from 23.5°C to 24.11 °C which is between the optimal limits, because the active climate control system operates the whole year. The greenhouse of configuration C has temperature oscillations from 23.09°C to 25.58°C, utilizing active climate control only for two months.

Natural Gas Cost (USD/MMBtu) 3.693

Total USD

Cost per Kwh of energy 0.105

Total USD

432.635

2,953.732

Fig. 9 shows the inner gains for the three configurations, these are: the solar gains of exterior windows, the zone of sensitive heating and the zone of sensitive cooling.

The solar gain for the greenhouses A and B is the same, since they both have an envelope built with polycarbonate glazing, which produces gains near to 21500 KWh from March to August. Configuration B requires to extract or supply almost 5000 KWh for reaching the desired temperature. In order to do this, the active climate control systems are functioning in January, February, November and December, which implies a huge demand of electric energy during four months.

Configuration C shows the advantage of the passive climate control using shading curtains, since the solar gains diminish from 21500 KWh to 12500 KWh, implying that a cooling system is not necessary. On the other hand, a heating system is not necessary for January-February, while the use of this system in November-December decreases in approximately 1000 KWh, because of its structure of concrete walls.

Table 1 shows the construction cost for each configuration.

The costs are calculated based on the area and the cost per m2, depending on the type of technology [25,26].

Table 2 shows the number of blocks required for building the configuration C walls, with a total of 1,535 blocks. The cost of solid concrete block is $0.503 USD each piece, and the walls have a cost of $772.10 USD, resulting in total cost of $5,919.14 USD for a C configuration construction.

The operating cost is assessed with DesignBuilder using the fuel consumption per year for each greenhouse topology to determine the cost of an operating year. Fuel prices used correspond to $103.72 USD per kWh of electricity on

157

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

real measurements.

Table 4.

Operating costs for C configuration.

Heating Cost Kwh year 7381 Cooling Cost

MMBtu year 25.18

Kwh year 0 Source: Own elaboration

References

Natural Gas Cost (USD/MMBtu) 3.693

Total USD

Cost per Kwh of energy 0.105

Total ($)

[1] 92.989 [2]

0.00

[3] [4]

downtown area and $ 3,693 USD/ MMBtu. Table 3 and 4 are used for configurations B and C in order to determine the operating costs for each system.

It can be observed from Tables 1 and 2 that the most expensive greenhouse configuration is C, using polycarbonate with both active and passive cooling, for a total of $6,206.66 USD, followed by the type B with polycarbonate and active cooling with a total of $5,146.56 USD. Finally, the greenhouse with only polycarbonate is the cheapest, with a total of $2,940.48 USD. In terms of operating cost, configuration A is the most economical because does not have a HVAC system installed. Configuration B is the most expensive, with an operating cost of $3,386.36 USD per year, since cooling and heating remain active throughout the year. The C Configuration has an operating cost of $92.98 USD per year, reducing fuel consumption in a 100%, and up to 78% of natural gas consumption.

[5]

[6]

[7]

[8]

[9]

6. Conclusions [10]

The results obtained shown that an asymmetric greenhouse with concrete enclosure, a polycarbonate cover and shading curtains diminishes the electric energy consumption at the same time that maintains the inner temperature inside the desired limits, with an increment of the thermal efficiency of the greenhouse. This design can be implemented in crop zones similar to the one used in this work, taking into account the correct location of the structure, in order to have the larger solar reception area oriented southwards.

Because of its design, the operating costs for the greenhouses were reduced, with a total quantity of $772.50 USD in concrete blocks and the implementation of passive cooling systems. On the other hand, although the structure is very important the dynamic simulations showed that the solar gains are related mainly to the size of reception area.

These results demonstrate the useful of virtual construction of greenhouses using specialized software for analyze the interaction between internal and external weather conditions.

Future work shall consider an increment on the number of natural systems inside a virtual greenhouse, in order to have a much more real simulation. These systems could be: humidity, infiltrations and thermal loads of ground and crop, to mention a few. It is also necessary to build a greenhouse with the designed characteristics to verify the results from the dynamic simulations, comparing them with

[11]

[12]

[13]

[14]

[15]

[16] [17]

158

Straten, G., Willigenburg, G., Henten, E. and Ooteghem, R., Optimal Control of Greenhouse Cultivation. CRC Press, 2010. http://dx.doi.org/10.1201/b10321 Sethi, V.P., Pal, S.R. and Dubey, R.K., Self regulating wick type zero energy hydroponics system for greenhouse tomatoes. Jounal of Agricultural Engineering, 50 (3), pp. 66-69, 2013.

Albright, L.D., Controlling greenhouse environments. Acta Horticulturae, 578, pp. 47-54, 2002.

Fitz-Rodriguez, E., Kubota, C., Tignor, M.E., Wilson, S.B. and McMahon, M., Dynamic modeling and simulation of greenhouse enviroments under several scenarios: A web-based application.

Computers and Electronics in Agriculture, 70 (1), pp. 105-116, 2010. http://dx.doi.org/10.1016/j.compag.2009.09.010 Rahman, M.-M., Rasul, M.-G. and Khan, M.-K., Energy conservation measures in a institutional building by dynamic simulation using design builder, Proceedings of the 3rd IASME/WSEAS International Conference on Energy & environment, pp. 192-197, 2008.

Kim, Y.-K. and Altan, H., Using dynamic simulation for demonstrating the impact of energy consumption by retrofit and behavioural change. Proceedings of 13th Conference of International Building Performance Simulation Asociation (BS2013), pp. 24512457, 2013.

Reinhart, C.F. and Wienold, J., The dayligthing dashboard – A simulation – based design analysis for daylit spaces. Building and Environment, 46 (2), pp.

386-396, 2011. http://dx.doi.org/10.1016/j.buildenv.2010.08.001 Stromann-Andersen, J. and Sattrup, P.A., The urban canyon and building energy use: Urban density versus dayligth and passive solar gains. Energy and Buildings, 43 (8), pp. 2011-2022, 2011. http://dx.doi.org/10.1016/j.enbuild.2011.04.007 De la Torre, G., Soto, G., López, I., Torres, I. and Rico, E., Computational fluid dynamics in greenhouses: A review. African Journal of Biotechnology, 10 (77), pp. 17651-17662, 2011 Valera, D.L., Molina, F.D. and Alvarez, A.J., Energy audit protocol greenhouse. Energy audit of a greenhouse for growing cut flowers Mendigorría. Spain: IDEA, 2008 Gruda, N., Impact of environmental factors on product quality of greenhouse vegetables for fresh consumption. Critital Reviews in Plant Sciences, 24 (3), pp.

227-247, 2005. http://dx.doi.org/10.1080/07352680591008628 Park, D.I., Kang, B.J., Cho, K.R., Shin, C.S., Cho, S.E., Park, J.W. and Yang, W.M., A study in a greenhouse automatic control system based on wireless sensor network. Wireless Personal Communications, 56 (1), pp.

117-130, 2011. http://dx.doi.org/10.1007/s11277-009-9881-2 Flores, D. and Ford, M., Mexico greenhouse and shade house production to continue increasing. USDA Foreign Agricultural Service. Report Number: MX0024. 2010, [Online], [date of of 2014].

Available at: reference November 11th http://gain.fas.usda.gov/Recent%20GAIN%20Publications/Greenho use%20and%20Shade%20House%20Production%20to%20Continue %20Increasing_Mexico_Mexico_4-22-2010.pdf SAGARPA. Protected Agriculture 2012. [Online], [date of reference November 11th of 2014]. Available at: http://20062012.sagarpa.gob.mx/agricultura/Paginas/AgriculturaProtegida2012.aspx Kipp, J., Optimal climate regions in Mexico for greenhouse crop production.Ministry of Agriculture and Food Quality. Rapport GTB1024. 2010. [Online], [date of reference November 11th of 2014].

Available at: http://edepot.wur.nl/144644 Strobel, N., Astronomy Notes 2013. US: McGraw-Hill, 2013.

Pasgianos, G.D., Arvanitis, K.G., Polycarpou, P. and Sigrimis, N., A nonlinear feedback technique for greenhouse environmental control.

Computers and Electronics in Agriculture, 40 (1-3), pp. 153-177, 2003. http://dx.doi.org/10.1016/S0168-1699(03)00018-8

Alvarez-Sánchez et al / DYNA 81 (188), pp. 152-159. December, 2014.

[18] Oehler, M., The earth-sheltered solar greenhouse book. How to build an energy-free year-round greenhouse. USA: Mole Publishing Company, 2007.

[19] Piscia, D., Montero, J., Baeza, E. and Bailey B., A CFD greenhouse night-time condensation model. Biosystems Engineering, 111 (2), pp.

141-154, 2012. http://dx.doi.org/10.1016/j.biosystemseng.2011.11.006 [20] Nebbali, R., Roy, J. and Boulard, T., Dynamic simulation of the distributed radiative and convective climate within a cropped greenhouse. Renewable Energy, 43, pp. 111-129, 2012. http://dx.doi.org/10.1016/j.renene.2011.12.003 [21] Cama, A., Gil, F., Gómez, J., García, A. and Manzano, F., Wireless surveillance system for greenhouse crops. DYNA, 81 (184), pp. 164170, 2014. http://dx.doi.org/10.15446/dyna.v81n184.37034 [22] Vega, E., Eidels, S., Ruiz, H., López-Veneroni, D., Sosa, G., Gonzalez, E., Gasca, J., Mora, V., Reyes, E., Sánchez-Reyna, G., Villaseñor, R., Chow, J. C., Watson, J. G. and Edgerton, S., A. particulate air pollution in Mexico city: A detailed view. Aerosol and Air Quality Research,10, pp. 193-211, 2010.

[23] U.S. Department of Energy., Energy, Efficiency & Reneawable Energy. [Online], [date of reference November 11th of 2014].

Available at: http://apps1.eere.energy.gov/buildings/energyplus/cfm/weather_data 3.cfm/region=4_north_and_central_america_wmo_region_4/country =MEX/cname=Mexico [24] Janjai, S., Intawee, P., Kaewkiew, J., Sritus, C. and Khamvongsa, V., A large-scale solar greenhouse dryier using polycarbonate cover: Modeling and testing in a tropical enviroment of Lao People’s Democratic Republic. Renewable Energy, 36 (3), pp. 1053-1062, 2011. http://dx.doi.org/10.1016/j.renene.2010.09.008 [25] Amundson, S.K., Cultural techniques to improve yield and cost efficiency of greenhouse grown tomatoes, MSc. Thesis, University of Tenessee, Knoxville, USA, 2012.

[26] Chávez, J., and Sanchez, N., Suggested price ranges for 5 types of greenhouses in Mexico. AMCI-SAGARPA, 2010. [Online], [date of of 2014].

Available at: reference November 11th http://www.firco.gob.mx/Proyectos/Proap/Documents/Presentacion_ Rangos_Precios_Proap_2010.Pdf

A. López-Velázquez, received the BSc. Eng. in Mechanical Industrial Engineering 1990 from Instituto Tecnológico de Veracruz, Mexico, the MSc. degree in Mechanical Engineering in 1997 and PhD degree in Mechanical Engineering in 2007 from Universidad Central “Marta Abreu” de las Villas, Santa Clara, Cuba. Since 2009 he is Full Time Professor in the Mechanical and Electrical Engineering Faculty, Universidad Veracruzana, Mexico. His research interests include: tribology, materials science and engineering.

E. Alvarez-Sánchez, received the BSc. Eng. in Mechanical Electrical Engineering in 2000 from Universidad Veracruzana, Veracruz, México; the MSc. degree and the PhD. degree in Electrical Engineering in 2003 and 2007, respectively, both from the Centro de Investigación y de Estudios Avanzados del Instituto Politecnico Nacional, México D.F., Mexico. From 2008 to 2010, he courses a Postdoctoral in Centro de Investigación Científica y de Educación Superior de Ensenada, Ensenada, Mexico. Since 2011 he is Full Time Professor in the Mechanical and Electrical Engineering, Faculty campus Xalapa, Universidad Veracruzana, Mexico.

His research interests include: mechanical design, linear and nonlinear control, mechatronics, seismic vibrations and domotics.

G. Leyva-Retureta, received the BSc. Eng. in Mechanical Electrical Engineering in 2010 from Universidad Veracruzana, Mexico, the MSc. degree in Energetic Engineering in 2013 in the Universidad Veracruzana, Xalapa, Veracruz, Mexico. Since 2012 he is technical academic in the Mechanical and Electrical Engineering Faculty, Universidad Veracruzana, Mexico. His research interests include: bioclimatic architecture, greenhouses, energy efficiency and domotics E. Portilla-Flores, received the BSc. Eng. in Electronic Engineering in 1990 from Universidad Autónoma Metropolitana, México D.F., Mexico, the MSc. degree in Mechanical Engineering in 2000 from Instituto Tecnológico de Puebla, México and the PhD. degree in Electrical Engineering in 2006 from the Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional, Mexico D.F., Mexico. A Postdoctoral in Universidade Estadual do Campinas, Brazil, was made in 2012. Since 2008 is Full Time Professor in the Centro de Innovación y Desarrollo Tecnológico en Cómputo del Instituto Politécnico Nacional, Mexico D.F., Mexico. His research interests include: mechanical design, mechatronics, concurrent design and gripper design.

159

Área Curricular de Ingeniería Mecánica Oferta de Posgrados



Maestría en Ingeniería - Ingeniería Mecánica

Mayor información: Wilfredo Montealegre Rubio Director de Área curricular acmecanica_med@unal.edu.co (57-4) 4259262

A comparative study TiC/TiN and TiN/CN multilayers Miguel J. Espitia-Rico a, Gladys Casiano-Jiménez b, César Ortega-López b, Nicolás De la Espriella-Vélez b & Luis Sánchez-Pacheco b b

a Grupo GEFEM Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Bogotá, Colombia. mespitiar@udistrital.edu.co Departamento de Física, Universidad de Córdoba, Montería, Colombia. gcasianoj@correo.unicordoba.edu.co, cortegal@correo.unicordoba.edu.co, naespriellav@correo.unicordoba.edu.co, lcpacheco@correo.unicordoba.edu.co.

Received: January 19th, 2014. Received in revised form: September 10th, 2013. Accepted: September 30th, 2014.

Abstract We carried out a comparative study of TiC/TiN and TiN/CN multilayers using the linearized augmented plane wave (LAPW) scheme and density functional theory as implemented in the Wien2k code. Initially, we optimized the structural properties of the TiC/TiN and TiN/CN multilayers in the volume with NaCl structure, and the ground state energy, the bulk modulus, and the cohesive energy were determined. To determine the energy of formation, the total energy for TiN, TiC, and CN compounds was calculated. Finally, we determined the DOS (density of states) of the two multilayers. The analysis of the partial density of states reveals that multilayers has metallic behavior that can be explained by the strong p–d hybridization of N and Ti atoms.

Keywords: DFT, multilayers, electronic and structural properties.

Un estudio comparativo de las multicapas TIC/TiN y TiN/CN Resumen En este trabajo se hace un estudio comparativo de las multicapas TiN/CN y TiC/TiN usando el esquema ondas plana aumentadas y linealizadas y la teoría del funcional de la densidad tal como se halla implementado en el código WIEN2k. Inicialmente, se optimizaron las propiedades estructurales de las multicapas TiN/CN y TiC/TiN en volumen en la estructura NaCl y se obtuvo la energía de estado base, el módulo de volumen y la energía de cohesión. Para obtener la energía de formación, se calcula la energía total para los compuestos TiN, TiC, CN. Finalmente, se obtiene la densidad de estados los dos multicapas. El análisis de la densidad de estados revela que las multicapas poseen un comportamiento metálico que puede ser explicado por la fuerte hibridación entre los orbitales p-d de los átomos de N y Ti. Palabras clave: DFT, multicapas, Propiedades electrónicas y estructurales.

1. Introduction Because of its hardness, high chemical and thermal stability, and high resistance to wear, oxidation, and corrosion, both titanium nitride (TiN) and titanium carbide (TiC) have been widely used as hard coatings in high-speed cutting tools and in devices operating at high degrees of power and high temperatures [1-4]. Unfortunately, hard coatings based only on traditional TiN or TiC offer limited benefit in terms of prevention of rupture or adherence to the workpiece, which is one of the main problems associated with tribological applications [5-7]. For this reason, many studies have sought to improve the mechanical properties of thin films of these nitrides and carbides through the growth of multilayers or nanocomposites based on TiN or TiC [5, 8-11].

A TiN/CN multilayer grown via magnetron sputtering [5] has been found to possess a higher resistance to wear, oxidation, and corrosion than TiN or TiC [5]; therefore hard coatings based on a TiN/CN multilayer will have better mechanical and tribological properties than those based on traditional carbides or nitrides. Furthermore, a titanium carbide/titanium nitride (TiC/TiN) multilayer is a very interesting material, because it combines the high degree of hardness and the low friction coefficient of TiC with the high degree of wear resistance of TiN [12-16]. This unusual combination of properties makes the TiC/TiN multilayer a material with great chemical stability and excellent mechanical properties, such as extreme hardness, a low friction coefficient, a high melting point, high electrical conductivity and a high degree of wear resistance [17-23].

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 160-165. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41637

Espitia-Rico et al / DYNA 81 (188), pp. 160-165. December, 2014.

Due to this combination of properties, hard coatings based on a TiC/TiN multilayer improve the performance of highspeed cutting tools, since the high thermal conductivity of the new compound reduces thermal gradients and reduces stress cracks as a result of thermal shock in the tools [14,25], whereby the high-speed cutting tools have a longer lifespan [23,26]. Finally, we haven’t found any scientific publications that compare the properties of these multilayers. For this reason, the main objective of this paper is do a comparative study of the structural and electronic properties of TiN/CN and TiC/TiN multilayers using calculations based on density functional theory.

TiC/TiN and TiN/CN multilayers, with a concentration 5050, we calculated the energy of formation of the multilayer in the rock salt structure. The energy of formation of a ternary (LMN) compound is defined as the difference between the total energy of the ternary L1-xMxN phase and the total energy of the binary compounds in the ground states of LN and MN,

therefore, the energy of formation is given by Eq. (2) [30], ௙௔௦௘ ௅భషೣ ெೣ ே

௙

௙௔௦௘ ௅ெ

௙௔௦௘ ெே

(2)

3. Results

2. Computational method Calculations were performed within the framework of density functional theory (DFT) and using the full potential augmented plane wave (FP-LAPW) method implemented in the Wien2k package [27]. The correlation and exchange effects of electrons were dealt with using the generalized gradient approximation (GGA) of Perdew, Burke, and Ernzerhof (PBE) [28]. In the LAPW method, the cell is divided into two types of regions, atomic spheres centered at nuclear sites and the interstitial region between the nonoverlapping spheres. Within the atomic spheres, wave functions are linear combinations of products of radial and spherical harmonic functions, while in the interstitial region, the function expands as a linear combination of plane waves. The charge density and potentials are expanded in spherical harmonics up to lmax = 10 within the atomic spheres, and the wave function in the interstitial region is expanded in plane waves with a cutoff parameter of RmtKmax = 8, where Rmt is the smallest radius of the atomic sphere in ଶ ଶ the unit cell and Kmax limits the kinetic energy of the plane waves, where ௠௔௫ . To ensure convergence in the integration of the first Brillouin zone, 54 points were used at the irreducible first Brillouin zone. The integrals over the Brillouin zone were solved using special Monkhorst-Pack alignment points [29]. Self-consistency was achieved by requiring that the convergence of the total energy be less than 10-4 Ry. For expanding the potential in the interstitial region, Gmax =12 was considered. Muffin-tin radii were 1.60 for N, 1.95 for Ti, and 1.60 bohr for C. Calculations were performed taking into account the spin polarization, in order to check possible magnetic properties of the compounds.

For the lattice constant, the minimum volume, the bulk modulus, and the cohesive energy of each structure studied, the calculated data were fit to the Murnaghan equation of state, Eq. (1)

3.1. Structural properties We first optimized the structural parameters of the binary compounds TiN, TiC, and CN, because these parameters were used to model the TiN/CN and TiC/TiN multilayers. CN has two stable crystalline structures, one hexagonal-type graphite and the other cubical [31]. The lattice constant a, the bulk modulus (B0), and the minimum energy (E0) of the binary compounds are shown in Table 1, together with experimental and theoretical values reported by other authors. Note that the lattice constant and the bulk modulus B0 calculated in this study agree well with the experimental values, which shows the reliability of the DFT-based computations.

Furthermore, it was found that both the TiN/CN multilayer modeled in the NaCl phase by inserting a layer of TiN and a layer of CN along the z axis as the TiC/TiN multilayer crystallize in a tetragonal structure with space Table 1.

Lattice constant a and bulk modulus B0 for the binary compounds, along with experimental and theoretical values, included for comparison. Also shown is the cohesion energy E0.

This Theoretical E0 (eV) Parameter word Exp. other works 4.26 4.24a - 14.55 4.249b , 4.32c a (Å) 277.2 288d, 297c , 317f, g TiN B0 (GPa) 320a 4.249b , 4.32c TiC a (Å) 4.336 4.33a, e - 15.50 B0 (GPa) 250 252h, 267c 240a 3.620 - 8.43 CN a (Å) 293.2 cubic B0 (GPa) CN graphite

଴

଴ ଴ ᇱ ଴

ᇱ ଴

Reference [32] Reference [33] c Reference [34] d Reference [35] e Reference [36] f Reference [37] g Reference [38] h Reference [39] i Reference [40] j Reference [41] b

(1)

where B0 is the bulk modulus, its first derivative is B’0, V0 is equilibrium volume of the cell, and E0 is the cohesive energy.

In order to verify the thermodynamic stability of the

a (Å) B0 (GPa)

Source: The Authors a

଴ ᇱ ଴

fase fase ELN and EMN respectively;

161

6.542 412.3

6,44i -

427j

- 7.51

Espitia-Rico et al / DYNA 81 (188), pp. 160-165. December, 2014.

Table 3.

Energy of formation Multilayer TiC/TiN TiN/CN Source: The Authors

E0 (eV) - 11.37 - 10.51

Ef (eV) 3.69 5.22

Figure 1. Total energy as a function of volume and adjusted to the Murnaghan equation of state for the TiN/CN multilayer (black curve) and the TiC/TiN multilayer (gray curve). Energies and volumes are given by number of atoms in the cell.

Source: The Authors

Table 2.

Lattice constant a and bulk modulus B0 for the binary compounds, along with experimental and theoretical values, included for comparison. Also are shown the values c/a and the cohesion energy E0.

Parameter This Other c/a E0 (eV) Word works a (Å) 4.343 4.30k, 4.27l 1.412 - 11.37 TiC/TiN B0 (GPa) 269.7 283k, 325l TiN/CN

a (Å) B0 (GPa) Source: The Authors k l

2.923 236.8

-

1.413

- 10.51

Reference [42] Reference [43]

group 123 (P4=mmm). Fig. 1 shows the total energy curves as a function of volume and fitted to the Murnaghan equation of state (Eq. 1) for the TiN/CN multilayer (black curve) and the TiC/TiN multilayer (gray curve).

We have taken as zero the sum of the energies of neutral atoms of isolated Ti, C, and N. Therefore, the absolute value of the minimum energy of each curve is the cohesive energy of the TiN/CN and TiC/TiN multilayers in the NaCl phase.

Fig. 1 shows that each structure is considered metastable, since there is a minimum of energy in the corresponding curve. The lattice constant, the bulk modulus (B0), and the minimum energy (E0) of the TiN/CN and TiC/TiN multilayers calculated in the NaCl structure are shown in Table 2.

We note that the lattice constants calculated in this paper are in good agreement with the experimental and theoretical results reported by other authors. Additionally, we see that the values of the volume modules of the TiN/CN and TiC/TiN multilayers are quite high, confirming the hardness of these materials and making them ideal candidates for hard coatings on high-speed cutting tools and devices that must operate at a high degree of power and at high temperatures.

Figure 2. Total density of states (TDOS) and partial density of states (PDOS) of the TiN/CN multilayer in the NaCl phase at the equilibrium volume.

Source: The Authors

As a next step, we calculated the formation energy_Ef of the TiC/TiN and TiN/CN multilayers, with a 5050concentration, that is, x = 50% TiC molecules, x = 50% TiN molecules, and x = 50% NC molecules. For this, we calculated the cohesive energy of the binary compounds ே௔஼௟ TiC, TiN, and CN in their ground states, ்௜ே ே௔஼௟ ே௔஼௟ , ்௜஼ , and ஼ே .

Table 3 shows the values of the energy of formation that we calculated using Eq. (2) The energy of formation of the TiC/TiN and TiN/CN multilayers was calculated to be 3.69 and 5.22 eV, respectively. Because these values are positive as compared with the reference state for both multilayers, these TiC/TiN and TiN/CN multilayers are thermodynamically unstable, i.e. they cannot be formed under equilibrium conditions.

Therefore, only when kinetic constrains prevent the system from reaching the equilibrium state, as found e.g. in plasmainduced physical and chemical vapor deposition (PVD and PCVD) at relatively low temperatures, can the metastable phase of TiC/TiN and TiN/CN multilayers in the rock salt structure be obtained, similar to the result predicted by Zhang and Veprek [30] for the ternary Ti0,5Al0,5N phase.

3.2. Electronic Properties The total density of states (TDOS) and partial density of states (PDOS) for TiN/CN and TiC/TiN multilayers in the NaCl phase are shown in Figs. 2 and 3, respectively.

162

Espitia-Rico et al / DYNA 81 (188), pp. 160-165. December, 2014.

intraband gaps, one in the region between ~-13.0 eV and ~ 10.8 eV and the other in the region between ~ -7.7 eV and ~ 6.3 eV, with widths of ~ -2.2 eV and ~ -1.4 eV, respectively.

As before, the multilayer does not exhibit magnetic properties, because the total density of spin up is fully offset by the total density of spin down.

Conclusions

Figure 3. Total density of states (TDOS) and partial density of states (PDOS) of the TiC/TiN multilayer in the NaCl phase at the equilibrium volume.

Source: The Authors

It can be observed that both TiC/TiN and TiN/CN multilayers exhibit a metallic behavior, determined by the Ti-d, C-p, and N-p hybrid states that cross the Fermi level.

Moreover, according to the theory proposed by Jhi et al.

[44], the highly directional coupling between metal d and non-metal p electrons results in covalent bonding, and this bonding makes a positive contribution to the hardness, which is responsible for the high degree of hardness of the TiC/CN and TiC/TiN multilayers. A similar result was found by Restrepo, Parra et al. [45] in their study of TIN and TiC. Additionally, it can be seen that the DOS of the TiC/CN multilayer at the Fermi level, mainly dominated by Ti-d states, is smaller than total density of states when compared with TiC/TiN. This may be caused by the metallic d窶電 interactions, which make a negative contribution to the bulk modulus [44]. This might be the cause of the low degree of hardness of TiN/CN [43]. Similar behavior was reported for the binary compounds TiC and TiN [46], and NbC and NbN [47].

For the total density of states of the TiN/CN multilayer, it can be observed that the valence band is divided into two regions. The first one, between ~ - 16.0 eV and ~ -13.0 eV, is mainly governed by the N-s states, with a small contribution from C-s states. The region between ~ -11.6 eV and the Fermi level is in great part composed of the N-p states, with a small contribution of Ti-d and C-p electrons.

Between ~ -13.0 eV and ~ -11.6 eV, an intraband gap of ~ 1.4 eV is present. The multilayer does not exhibit magnetic properties, because the total density of spin up is fully offset by the total density of spin down.

In the total density of states of the multilayer it can be observed that valence band is divided into three regions. The first, between ~ -15.0 eV and -13.0 eV, is mainly governed by the N-s states, with a small contribution from C-s states. The region between ~ -10.8 eV and ~ -7.7 eV is mainly dominated

by C-s states, with a small contribution of N-p electrons. The region between ~ -6.3 eV and the Fermi level is governed by the Ti-d, C-p, and N-p states. The TiCN compound has two

We reported first-principles calculations to determine the structural and electronic properties of TiN/CN and TiC/TiN multilayers and carried out a comparative study of them. The calculated lattice constants are in good agreement with experimental data. The calculated values of bulk modulus are quite high, which means that these materials are very rigid. This property is due to the strong covalent bonds that exist between the metallic Ti-d and the nonmetallic C-p and N-p states. This confirms that these materials exhibit a high degree of hardness, which makes them attractive for potential applications at high temperature and for hard coatings. However, we found that the bulk modulus of the TiC/TiN multilayer was larger than that of the TiN/CN multilayer.

From the density of states it was found that the compounds exhibit metallic behavior, because of the Ti-d, C-p, and N-p hybrid orbitals traversing the Fermi level.

Acknowledgements The authors thank the Research Center of the University of Cordoba CUIC for its financial support.

References [1]

[2] [3] [4] [5]

[6]

[7]

[8]

[9]

163

Chen, G.S., Guo, J.J., Lin, C.K., Hsu, C.S. and Fang, J.S., Study on the characteristics of TiN thin film deposited by the atomic layer chemical vapor deposition. J. Vac. Sci. Technol. A, 20, pp. 479-484, 2002. http://dx.doi.org/10.1116/1.1450580 Hainsworth, S.V. and Soh, W.C., The effect on the mechanical properties of TiN coatings. Surf. Coat. Technol., 163, pp. 515-520, 2003. http://dx.doi.org/10.1016/S0257-8972(02)00652-7 Patsalas, P. and Logothetidis, S., Interface properties and structural evolution of TiN/Si and TiN/GaN heterostructures. J. Appl. Phys., 93, pp. 989-903, 2003. http://dx.doi.org/10.1063/1.1531812 Hu, S.B., Tu, J.P., Mei, Z., Li, Z.Z. and Zhang, X.B., Corrosion resistance of multi-layered PAPVD TiN and CrN coatings. Surf.

Coat. Technol., 141, pp. 164174-, 2001.

Liu, D.G, Tu, J.P., Gu, C.D., Chen, R. and Hong, C.F., Tribological and mechanical behaviors of TiN/CNx multilayer films deposited by magnetron sputtering. Thin Solid Films, 519, pp. 4842-4848, 2011. http://dx.doi.org/10.1016/j.tsf.2011.01.039 Takadoum, J. and Houmid-Bennani, H., Influence of substrate roughness and coating thickness on adhesion, friction and wear of TiN films Surf. Coat. Technol., 96, pp. 272-282, 1997. http://dx.doi.org/10.1016/S0257-8972(97)00182-5 Lahres, M., Muller-Hummel, P. and Doerfel, O., Applicability of different hard coatings in dry milling aluminium alloys. Surf. Coat.

Technol., 91 (2), pp.

116-121, 1997. http://dx.doi.org/10.1016/S0257-8972(96)03121-0 PalDey, S. and Deevi, S.C., Single layer and multilayer wear resistant coatings of (Ti,Al) N: A review. Mater. Sci. Eng. A, 342 (1-2), pp. 58-79, 2003. http://dx.doi.org/10.1016/S09215093(02)00259-9 Barshilia H.C., Prakash, M.S., Jain A. and Rajam, K.S., Structure, hardness and thermal stability of TiAlN and nanolayered TiAlN/CrN

Espitia-Rico et al / DYNA 81 (188), pp. 160-165. December, 2014.

[10]

[11]

[12]

[13] [14] [15]

[16] [17]

[18] [19] [20]

[21]

[22]

[23]

[24] [25]

[26] [27] [28]

multilayer films. Vacuum, 77(2), pp. 169-179, 2005. http://dx.doi.org/10.1016/j.vacuum.2004.08.020 Parlinska-Wojtan, M., Meier, S. and Patscheider, J., Transmission electron microscopy characterization of TiN/SiNx multilayered coatings plastically deformed by nanoindentation. Thin Solid Films, 518, pp.

4890-4897, 2010. http://dx.doi.org/10.1016/j.tsf.2010.02.064 Lin, Y. and Munroe, P.R., Deformation behavior of complex carbon nitride and metal nitride based bi-layer coatings. Thin Solid Films, 517, pp.

4862-4866, 2009. http://dx.doi.org/10.1016/j.tsf.2009.03.182 Knotek, O., Loffler, F. and Kramer, G., Deposition, properties and performance behaviour of carbide and carbonitride PVD coatings.

Surf. Coat. Technol., 61 (1-3), pp. 320-325, 1993. http://dx.doi.org/10.1016/0257-8972(93)90246-K Guu, Y.Y. and Lin, J.F., Analysis of wear behaviour of titanium carbonitride coatings, Wear, 210 (1-2), pp. 245-254, 1997. http://dx.doi.org/10.1016/S0043-1648(97)00056-2 Hsieh, J.H., Tan, A.L.K. and Zeng, X.T., Oxidation and wear behaviors of Ti-based thin films. Surf. Coat. Technol., 201(7), PP. 4094-4098, 2006.

Baravian, G., Sultan, G., Damond, E. and Detour, H., Optical emission spectroscopy of active species in a TiCN PVD arc discharge. Surf. Coat. Technol., 77, pp. 687-693, 1995. http://dx.doi.org/10.1016/02578-9729(68)00077http://dx.doi.org/10.1016/0257-8972(96)80007-7 Gergmann, E., Kaufmann, H., Schmid, R. and Vogel, J., Ion-plated titanium carbonitride films Surf. Coat. Technol., 42 (3), pp. 237-251, 1990. http://dx.doi.org/10.1016/0257-8972(90)90156-7 Narasimhan, K. Boppana, S.P. and Bhat, D.G., Development of agraded TiCN coating for cemented carbide cutting tools a design approach, Wear 188 (1–2), pp.123-129, 1995. http://dx.doi.org/10.1016/0043-1648(95)06635-7 Destefani J., Cutting tools 101, Manufacturing Engineering, 129 (3), pp.18-22, 2002.

Yang, Y.L., Zhang, D. Kou, H.S. and Liu, S.C., Laser cladded TiCN coatings on the surface of titanium. Acta Metallurgica Sinica (English Letters), 20 (3), pp, 210-216, 2007.

Yang, Y. Yao, W. and Zhang, H., Phase constituents and mechanical properties of laser in-situ synthesized TiCN/TiN composite coating on Ti–6Al–4V. Surface and Coatings Technology, 205 (2), pp. 620624, 2010. http://dx.doi.org/10.1016/j.surfcoat.2010.07.058 Aslan, E., Experimental investigation of cutting tool performance in high speed cutting of hardened X210 Cr12 cold-work tool steel (62 HRC). Materials and Design, 26 (1), pp. 21-27, 2005. http://dx.doi.org/10.1016/j.matdes.2004.04.004 Tsao, C.C. and Hong, H., Comparison of the tool life of tungsten carbides coated by multi-layer TiCN and TiAlCN for end mills using the Taguchi method. Journal of Materials Processing Technology, 123 (1), pp.

1-4, 2002. http://dx.doi.org/10.1016/S0924-0136(01)01152-9 Siow, P.C., Ghani, J.A. and Ghazali M.J.. Characterization of TiCN and TiCN/ZrN coatings for cutting tool application. Ceramics International, 39, pp.

1293-1298, 2013. http://dx.doi.org/10.1016/j.ceramint.2012.07.061 Zhang, S.Y., Titanium carbonitride based cermets: processes and properties. Mater Sci Eng A, 163, pp.141-147, 1993. http://dx.doi.org/10.1016/0921-5093(93)90588-6 Peng, Y., Miao, H. and Peng, Z., Development of TiCN-based cermets: Mechanical properties and wear mechanism. Int. Journal of Refractory Metals and Hard Materials, 39, pp. 78-89, 2013. http://dx.doi.org/10.1016/j.ijrmhm.2012.07.001 Van Den Berg, H., Hardmetals: Trends in development and application, Powder Metallurgy, 50 (1), pp. 7-10, 2007. http://dx.doi.org/10.1179/174329007X186822 Blaha, P., Schwarz, K. and Trickey, S.B., Electronic structure of solid with WIEN2k, Molecular Physics, 108 (21), pp. 3147-3166, 2010.

Perdew, J. Burke, K. and Ernzerhof, M., Generalized gradient approximation made simple. Physical Review Letter, 77 (18), pp. 38653868, 1996. http://dx.doi.org/10.1103/PhysRevLett.77.3865

[29] [30]

[31]

[32]

[33] [34]

[35]

[36]

[37]

[38]

[39] [40]

[41] [42]

[43]

[44] [45] [46]

164

Monkhorst, H.J. and Pack, A.T., Special points for brillouin-zone integrations. Physical Review B, 13 (12), pp. 5188-5192, 1976. http://dx.doi.org/10.1103/PhysRevB.13.5188 Zhang, R.F. and Veprek, S., Metastable phases and spinodal decomposition in Ti1−xAlxN system studied by ab initio and thermodynamic modeling, a comparison with the TiN–Si3N4 system.

Materials Science and Engineering A, 448, pp. 111-119, 2007. http://dx.doi.org/10.1016/j.msea.2006.10.012 Yu, D.L., Tiana, Y.J., He, J.L., Xiao, F.R., Wang, T.S., Li, D.C., Zheng, G. and Yanagisawa, O. Preparation of CNx/TiNy multilayers by ion beam sputtering. Journal of Crystal Growth, 233, pp. 303-311, 2001. http://dx.doi.org/10.1016/S0022-0248(01)01492-0 Zhukov, V.P., Gubanov, V.A., Jepsen, O., Christensen, N.E. and Anderson, O.K., Calculated energy band structures and chemical bonding in titanium and vanadium carbides, nitrides and oxides. J. Phys.

Chem.

Solids, 49 (7), pp.

841-849, 1988. http://dx.doi.org/10.1016/0022-3697(88)90037-6 González-Hernández, R., López-Pérez, W. and Rodríguez-Martínez, J., Ab initio study on the structural and electronic properties of TiN, Revista Colombiana de Física, 39 (2), pp. 519-522, 2011.

Grossman, J.C., Mizel, A., Cote, M., Cohen, M.L. and Louie, S.G., Transition metals and their carbides and nitrides: Trends in electronic and structural properties. Phys. Rev. B 60 (9), pp. 6343-6347, 1999. http://dx.doi.org/10.1103/PhysRevB.60.6343 Stampfl, C., Mannstadt, W., Asahi, R. and Freeman, J., Electronic structure and physical properties of early transition metal mononitrides: Density functional theory LDA, GGA, and screened-exchange LDA FLAPW calculations. Physical Review B, 63, pp. 55106(1)-55106(5), 2001.

Dunand, A., Flack. H.D. and Yvon, K., Bonding study of TiC and TiN. I. High-precision x-ray-diffraction determination of the valence-electron density distribution, Debye-Waller temperature factors, and atomic static displacements in TiC0.94 and TiN 0.99. Phys. Rev. B, 31, pp.

2299-2315, 1985. http://dx.doi.org/10.1103/PhysRevB.31.2299 Dridi, Z., Bouhafs, B., Ruterana, P. and Aourag, H., First-principles calculations of vacancy effects on structural and electronic properties of TiCx and TiNx, J. Phys.: Condens. Matter., 14, pp. 10237-10249, 2002. http://dx.doi.org/10.1088/0953-8984/14/43/320 Guemmaz, M., Mosser, A. Ahujab, R. and Johansson, B., Elastic properties of sub-stoichiometric titanium carbides: Comparison of FPLMTO calculations and experimental results. Solid State Communications, 110 (6), 299-303, 1999. http://dx.doi.org/10.1016/S0038-1098(99)00091-5 Liu L.M., Wang S.Q. and Ye H.Q., Adhesion and bonding of the Al/TiC interface. J. Surface Science, 550, pp. 46-56. 2004. http://dx.doi.org/10.1016/j.susc.2003.12.031 Wu, M.L., Qian, W.D., Chung, Y.W., Wang, Y.Y., Wong M.S. and Sproul, W.D., Superhard Coatings of CNx/ZrN Multilayers Prepared by DC Magnetron Sputtering, Thin Solid Films, 308-309, pp. 113-117, 1997. http://dx.doi.org/10.1016/S0040-6090(97)00429-X López-Fernández, V.. Nanomateriales basados en carbono. Doctoral Thesis, Departament of Chemical, Universidad Autonoma de Madrid, España, 2009.

Kuptsov, K.A., Kiryukhantsev, K.V., Sheveyko, A.N. and Shtansky, D.V., Comparative study of electrochemical and impact wear behavior of TiCN, TiSiCN, TiCrSiCN, and TiAlSiCN coatings. Surface and Coatings Technology, 216, pp.

273-281, 2013. http://dx.doi.org/10.1016/j.surfcoat.2012.11.058 Wenxia, F., Shouxin, C., Haiquan H., Guiqing Z. and Zengtao, L., Electronic structure and elastic constants of TiCxN1-x, ZrxNb1-xC and HfCxN1-x alloys: A first-principles study. Physica B, 406 (19), pp.

3631-3635, 2011. http://dx.doi.org/10.1016/j.physb.2011.06.058 Jhi, S.H., Ihm, J., Louie, S.G. and Cohen, M.L., Electronic mechanism of hardness enhancement in transition-metal carbonitrides. Nature, 399, pp. 132–134, 1999. http://dx.doi.org/10.1038/20148 Restrepo-Parra, E., Arango, P. and Casanova, S., Some concepts about Titanium nitride and titanium carbide. DYNA, 74 (157), pp. 213-224, 2009.

Yang, Y., Lu, H., Yu, C. and Chen, J.M., First-principles calculations of mechanical properties of TiC and TiN. Journal of Alloys and Compounds, 485 (1-2), pp.

542–547, 2009. http://dx.doi.org/10.1016/j.jallcom.2009.06.023

Espitia-Rico et al / DYNA 81 (188), pp. 160-165. December, 2014.

[47]

Amriou, T., Bouhafs, B., Aourag, H., Khelifa, B., Bresson, S. and Mathieu, C., FP-LAPW investigations of electronic structure and bonding mechanism of NbC and NbN compounds. Physica B, 325, pp.

46-56, 2003. http://dx.doi.org/10.1016/S0921-4526(02)01429-1

M.J. Espitia-Rico, received the BSc in Physicist degree in 1999 from the Universidad de Córdoba, Colombia; the MSc degree in Physical Science and is Dr. candidate in Physical Sciences, from the Universidad Nacional de Colombia. Bogotá, Colombia. From 1999 to 2004, he worked as a professor in the Physical Department, Universidad de Córdoba. Montería, Colombia. Since 2005 to day, he is a Full Professor in the Universidad Distrital Francisco José de Caldas, Colombia. His research interests include: simulation, computational calculation of new materials.

G.R. Casiano-Jiménez, received the BSc in Physicist degree in 1990 from the Universidad de Córdoba, Colombia, the MSc degree in Physical Science and is Dr. candidate in Physical Sciences, from the Universidad de Córdoba. Montería, Colombia. Since 2005 to day, she is a Full Professor in the Universidad de Córdoba, Colombia. His research interests include: simulation, modeling and computational calculation of new materials.

C. Ortega-López, received the BSc in Physicist degree in 1989 from the Universidad de Córdoba, Colombia, the MSc degree in Physical Science in 2002, and the PhD degree in Physical Science in 2009, from the Universidad Nacional de Colombia, Bogotá, Colombia. Since 1997 to day, he is a Full Professor in the Universidad Córdoba, Colombia. His research interests include: simulation, modeling and computational calculation of new materials.

N.A. De la Espriella-Vélez, received the BSc in Physicist degree in 1987 from the Universidad de Córdoba, Colombia, the MSc degree in Physical Science in 2002 from the Universidad Nacional de Colombia, Colombia, and the PhD degree in Physical Science in 2010, from the Universidad Central de Venezuela, Caracas, Venezuela. Since 1997 to day, he is a Full Professor in the Universidad Córdoba, Colombia. His research interests include: simulation, modeling and computational calculation of new materials.

L.C. Sánchez-Pacheco, received the BSc in Physicist degree in 2003, the MSc degree in Physical Science in 2006, and the PhD degree in Physical Science in 2012, all of them from Universidad de Antioquia, Medellín, Colombia. Since 2010 to day, he is a Full Professor in the Universidad Córdoba, Colombia. His research interests include: simulation, modeling and computational calculation of new materials.

165

Área Curricular de Ingeniería Geológica e Ingeniería de Minas y Metalurgia Oferta de Posgrados    

Especialización en Materiales y Procesos Maestría en Ingeniería - Materiales y Procesos Maestría en Ingeniería - Recursos Minerales Doctorado en Ingeniería - Ciencia y Tecnología de Materiales

Mayor información: Néstor Ricardo Rojas Reyes Director de Área curricular acgeomin_med@unal.edu.co (57-4) 425 53 68

Developing a fast cordless soldering iron via induction heating Ernesto Edgar Mazón-Valadez a, Alfonso Hernández-Sámano a, Juan Carlos Estrada-Gutiérrez a, José Ávila-Paz a & Mario Eduardo Cano-González a* a

Centro Universitario de la Ciénega, Universidad de Guadalajara, Guadalajara, México. mazon_valadez@hotmail.com, h.s.alfonso@gmail.com, jcarlosredes@gmail.com, jocmos@hotmail.com, meduardo2001@hotmail.com*.

Received: January 19th, de 2014. Received in revised form: June 10th, 2014. Accepted: October 17th, 2014

Abstract This study aims to present the design of a new soldering iron for welding electronic components that work via AC magnetic fields.

Moreover, the device has been designed to operate in cordless mode. The system comprises of a resonant inverter that is capable of generating an alternating magnetic field of 250 kHz in the center of a coil of 11 wires. The heating element is a cylindrical piece of ferromagnetic stainless steel with a concentric core of copper, which is maintained in contact with a standard and commercial tip.

Additionally, we determined the power factor and the efficiency of the energy transfer with a maximum power consumption of 134 watts.

The system represents a good tool suitable for the realization of development boards or electronic tasks.

Keywords: Electromagnetic induction, Heating, Resonant inverter, Soldering.

Desarrollo de un cautín inalámbrico rápido a través de calentamiento por inducción Resumen Se presenta el desarrollo de un dispositivo para soldar componentes electrónicos, el cual funciona a base de campos magnéticos alternos.

El dispositivo ha sido diseñado para trabajar sin cableado. El nuevo soldador se compone de un inversor resonante capaz de generar campos magnéticos alternos de 250 kHz en el centro de una bobina de 11 espiras. El elemento calefactor es una pequeña pieza cilíndrica de acero inoxidable magnético con un núcleo concéntrico de cobre, el cual se encuentra unido a una punta reemplazable para cautín comercial. Adicionalmente hemos determinado el factor de potencia y la eficiencia en la transferencia de energía con un máximo de consumo de potencia de 134 Watts. El dispositivo representa una buena herramienta adecuada para la realización de tarjetas impresas para circuitos o tareas de electrónica.

Palabras clave: Inducción Electromagnética, Calentamiento, Inversor Resonante, Soldador.

1. Introduction The induction heating process is a physical phenomenon that is widely used for the fast heating of ferromagnetic materials. A focalized distribution of eddy currents on the material is the origin of a quick heating [1]. Indeed, this kind of heating is cleaner than the conventional method of electric resistances as it does not involve any contact with the material [2]; nevertheless, the use of resistances implies a simpler technology and lower manufacturing cost.

At present, induction heaters are used in industrial applications for metallic pieces, such as dilatation, furnacing and welding [3]. Moreover, it is also used in daily life tasks, such as cooking [4], sealing plastic bags [5], and ironing

clothes [6]. Other new applications are focused on the heating of magnetic iron oxide nanoparticles for the treatment of tumors [7] and for the study of the specific absorption rate of magnetic materials [8].

On the other hand, there are two types of soldering irons (guns or pencil presentations) especially designed to perform the welding of electronic components on printed circuit boards (PCB), mostly in electronic labs. The more common ones are mainly composed of an electrical resistance wrapped around a heating element that transmits the heat to a soldering tip to melt the wire of lead/tin.

However, in the last few years, it has been possible to acquire high frequency soldering irons that work by using magnetic induction [9]. Due to a very different physical

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 166-173. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41635

Mazón-Valadez et al / DYNA 81 (188), pp. 166-172. December, 2014.

principle, these devices exhibit a higher performance to melt the wire in contrast to the devices of resistance. These presentations of soldering irons (with resistances or induction) possess a cable directly connected to a domestic tension line or a control station. This fact represents a source of possible accidents, imprecise welds or an obstacle in the soldering process. In order to tackle this issue, the present work aims to develop a cordless soldering iron via induction heating. This device is a powerful tool, which allows comfortable welding, and is a good alternative for the development of tasks in the electronic lab.

2. Theoretical Background 2.1. Induction heating The AC current flow in a cylindrical conductor rod with radius a have been studied for several years [10]. In case the conductor has ferromagnetic properties, this current flow could be induced by AC magnetic fields, producing a rapid heating in the material, formed by the distribution of eddy currents [1]. Fig. 1 shows the typical procedure to heat a ferromagnetic piece placed inside the region of a magnetic flux. The magnetic field is generated by the use of a coil with N loops and inductance L, which is powered with an AC power supply of high frequency. Moreover, the magnetic permeability of the metal depends on the temperature; indeed, the ferromagnetic ordering of the material could vanish in the Curie point (Curie temperature TC) where it becomes a paramagnetic material [11,12].

In agreement with Brown [13], the magnitude of the magnetic field intensity H in the rod satisfies eq. (1).

d 2 H 1 dH   H ( j 2f )  0, dr 2 r dr

(1)

r H  AJ 0 (  2 j ), s where, s

physically

1 dH  E.  dr

as

the

(3)

Then, it is possible to determine the power dissipated ( p  E 2 ) for the ferromagnetic rod in watts per unit length, as expressed in eq. (4) [15]. r a

P

 2rE dr.

2

(4)

r 0

With a further analysis using the Frobenius solution of H, assuming uniform magnetic field intensity H0 in all the metal and imposing boundary conditions on the surface of the cylinder, the value of A is determined. Thus, P is obtained in two ranges of (a/s) as shown in eqs. (5) and (6) [13].

4



P

2H 02  a    ;  s

a  1, s

(5)

P

8H 02 a ;  s

a  5. s

(6)

and

describes an oscillatory behavior of H(r) in r = 0.

This solution can be rewritten as shown in eq. (2) [13].

Figure 1. A ferromagnetic rod or cylinder heated with an AC magnetic field.

Source: Own

known

penetration of the radiofrequency (skin thickness) in the material [14], and the constant A depends on the initial conditions. With the new variables in eq. (2), it is possible to obtain the solution as a function of the radius of the cylinder and the skin thickness.

The relationship between the electric field due to the induced currents in the material and the magnetic field intensity is given in eq. (3). This expression can be obtained from the Maxwell-Ampere equation [15] neglecting the dielectric component of the material.

where, f is the frequency, μ is the magnetic permeability and σ is the conductivity of the material.

The solutions of eq. (1) are the typical Bessel functions of first class and order zero H  AJ 0 r  j 2f , which



 1 f is

(2)

Among the two equations, eq. (6) is more commonly used in industrial applications of induction heating; besides, the design of the heaters must take into account the values of μ and σ of the materials in order to establish the frequency range for heating. For example, if the goal is to heat a 1 cm diameter magnetic stainless steel rod, considering the definition of s, the regime a s  5 and the values of μ = 1.76 × 10−5 N A−2 and σ = 1.45 × 106 Siemens/m, respectively, the minimum frequency suitable to heat is 100 Hz. If we increase the frequency, the material will also be heated and the skin will diminish. This material is a good electrical conductor but a bad conductor of heat, whereas copper has a high σ, a low μ, and transmits heat very quickly and high frequencies are needed [13].

167

Mazón-Valadez et al / DYNA 81 (188), pp. 166-172. December, 2014.

2.2. Resonant inverter Various configurations of switching circuits can be designed to generate AC magnetic fields. The resonant inverters are a good alternative to obtain AC current. They could be composed of an H-bridge (full or half) or “push pull” configurations of transistors to feed a series or parallel LC circuit or “tank” [16]. The tank resonates when the difference between the capacitive and the inductive reactance is null. The commutation ON/OFF of the transistors is carried out using a driver circuit. In some applications, a phase follower stage (PLL) or frequency compensation is also needed [17,18].

Indeed, various reports have shown that the resonant inverters can be used as correctors of power factor under certain operating conditions [19,20].

The half H-bridge inverter (Fig. 2) is an attractive choice in low-power and high-frequency applications as it exhibits good stability and simple topology to feed L-C, C-L-C or LC-L resonant circuits with low total harmonic distortion.

Fig. 2 shows a well-known resonant inverter (half or full bridge) [21,22], which feeds a CR-LR parallel circuit with correct impedance coupling through CM-LM circuit. The capacitor CM limits the current to the resonant tank (CR-LR) and diminishes unwanted DC voltages. The inductor LM is a filter for higher frequencies and gives the initial impulse to the tank circuit in each cycle of commutation. This resonant inverter also contains identical zener diodes to maintain the rectangular shape of the pulse, opposite identical ultrafast diodes to drain the back-electromotive forces and RC snubber circuits to suppress voltage transients returning from the tank. It is important to note that the maximum current value and magnetic field of the tank, in ideal conditions, satisfy eqs. (7) and (8), respectively; whereas, the frequency is approximate by eq. (9).

IR 

VDC CR LR

BMax 

fR 

.

0 I RMax N

(7)

.

(8)

1 .

2 LRCR

(9)

h

Figure 3. (A) New soldering iron with all the pieces, (B) lateral view of the soldering iron and the support with the resonant coil (5) and sensor (6), and (C) the full device including the control box (7).

Source: Own

3. Materials and Methods 3.1. The industrial design of the soldering iron Initially, we used the handle of a commercial soldering iron with pencil design based on a resistance; thus, a new set of pieces was adapted to work via magnetic induction.

Fig. 3(A) shows the design of the new soldering iron, which contains a thermal insulating handle (1), a thin cylindrical piece of copper (2), a thick cylindrical piece of magnetic stainless steel (3) and a standard tip (4).

The piece (3) is heated by the resonant inverter and works like a small thermal reservoir due to its diameter and low thermal conductivity. A radial heating is received by the concentric cylinder (2) and is quickly transferred to the tip (4).

Fig. 3(B) shows a lateral view of the new soldering iron and its station. In this station, there is an AC magnetic field generator (5) to heat an extreme (3) and an infrared sensor (6). Fig. 3(C) exhibits the full system including the control box (7). In the coil (5), we have winding Litz wire on a rod of fiberglass to diminish the heating on the copper due to the skin thickness and support the temperature of the soldering iron, respectively [23].

3.2. Electronic features of the soldering iron

Figure 2. Diagram of a resonant inverter L-C-L with a half H-bridge.

Source: Own

In order to develop the cordless soldering iron, we used a resonant inverter, like the one shown in Fig. 2 but without snubbers, transistors STW14NK50Z (MOSFET technology) of very low cost and the passive components CM = 300 nF, LM = 10 μH, CR = 188 nF, and LR = 2.15μH. The device also contains a microcontroller (MCU) PIC18F4550 for driving the inverter, to digitizing and displaying the temperature readings. Initially, the MCU produces two digital control signals with a 60% duty cycle, 250 kHz frequency with 180° of phase shift and a delay of 0.5 μs between them.

168

Mazón-Valadez et al / DYNA 81 (188), pp. 166-172. December, 2014.

electronic stage of frequency compensation or PLL circuits can be neglected, but it is also needed to turn-off the system when the soldering iron has been removed.

The DC power supply of the inverter has been constructed using EMI/RFI filter to avoid any interference with other equipments [24], then the AC current is rectified by four diodes, and another LC filter (150 μH and 1 μF) is added to increase the power factor; besides, we observed a diminution of the total harmonic distortion in the output current. Fig. 6 shows the schematic circuit.

Figure 4. (A) Block diagram of the driver and (B) corresponding output signal of the driver.

Source: Own

Figure 6. Schematic diagram of the DC power supply.

Source: Own.

The pulses are pre-amplified with a pair of MOSFETs BS170 to feed an isolation transformer of 45 turns. Next, each signal is an input of the MOSFET driver UCC37321P and UCC37322P, respectively, to invert each one. Then, the stages of low and high powers are galvanically isolated. Fig. 4(A) shows the block diagram of the driver circuit, while Fig. 4(B) shows the signals to switch the current in the inverter obtained from the oscilloscope and measured in each gate of the transistor. These signals have rectangular shape with a 40% duty cycle and 0.5 μs dead time. The negative amplitude of the signal guarantees the fast cut-off of the transistors, whereas the positive one ensures its saturation status.

In order to have a continuous monitoring of the welding tip temperature, an infrared thermopile MLX90616 is calibrated.

This sensor is connected to an integrator and a non-inverting amplification using operational amplifiers. The total gain of the amplification is 7.3 and the circuit is shown in Fig. 5.

Protection of the device is essential for having control on the maximum temperature of the stainless steel, since a strong diminution of the resonant charge could damage the circuits.

This fact becomes very critical when the temperature in (3) reaches the value T = TC, or when the user heats different materials with the same resonant coil. Our observations in the range from room temperature to 500°C of the inductance allow a maximum increasing of 5%. For this reason, an

Figure 5. Amplification circuit of the thermopile infrared sensor.

Source: Own

Figure 7. Welding cycles with (A) de = 0.762 cm, (B) de = 0.889 cm, and (C) de = 1.016 cm.

Source: Own 169

Mazón-Valadez et al / DYNA 81 (188), pp. 166-172. December, 2014.

V/div), with maximum RMS value of 16.26 A and 56.57 V, respectively. Moreover, a phase shift of 82° was observed.

These measurements exhibit a good agreement with the ideal values simulated in the LT-Spice platform shown in Fig. 8(B). Furthermore, a very small distortion was observed in the current signal.

Fig. 9(A) shows the signal of the current (1 A/mV) in a larger interval, which is modulated in amplitude, the modulating signal is 60 Hz of the domestic line tension.

This observation is also in a good agreement with the theoretical predictions of the simulation shown in Fig. 9(B).

The main difference is in the zero crossing, but this is due to the capacitance of the LC circuit to correct the power factor.

Figure 8. (A) Curves of voltage (black line) and current (gray line) obtained experimentally and (B) ideal curves simulated with the LT-Spice.

Source: Own

4. Experimental Results and Discussions With the aim of establishing a “welding cycle”, we studied the heating of the tip with various diameters of the stainless steel piece (3) as shown in Fig. 3(A). The readings of the infrared sensor were obtained using the oscilloscope with scales of 50 s/div and 120°C/div.

Fig. 7 exhibits the welding cycles considering external diameters de of 0.762 cm, 0.889 cm and 1.016 cm. For example, with de = 0.762 cm and starting from room temperature (25°C), the value T = 500°C is reached in 45 s.

This value is maintained for a time lapse, before the inverter is turned off and the temperature diminishes to reach T = 200°C in 90 s. At this point, it is not possible to melt the welding wire, but by turning on the inverter, the tip again reaches T = 500°C in 20 s. Thus, ideally, we have 20 s to heat the tip and 90 s for soldering electronic components.

We chose de = 1.016 cm, with a cycle 32 s to heat, 155 s for soldering and initial heating of 80 s, the latter is similar to the initial heating time of a commercial soldering iron with resistance (pencil).

Following with the characterization of the device, using a current probe of Hall Effect A6303 with its corresponding amplifier AM 503 connected to an oscilloscope, the curves of voltage and current in the resonant tank were studied while the tip was heated. Fig. 8(A) shows a sinusoidal shape of 250 kHz in the current (0.5 A/mV) and voltage (50

Figure 9. (A) Curves of current obtained experimentally and (B) simulated with the LT-Spice.

Source: Own

Figure 10. Curves of the current (gray line) consumption and voltage (dark line) obtained experimentally.

Source: own

170

Mazón-Valadez et al / DYNA 81 (188), pp. 166-172. December, 2014.

Additionally, in order to study the power consumption of the device, the curves of current and voltage in the input were also analyzed. Fig. 10 shows those curves in the scale of (0.1 A/mV) and (100 V/div). In those measurements, a phase shift   6º is observed, which allows a power factor Cos  6º  0.995 , current consumption IRMS = 1.06 A with VRMS = 127 V. We analyzed the efficiency with a simple calculus of the maximum input and output power of the device; this value is given by eq. (10).



Poutput Pinput

,

[2]

[3]

[4]

[5]

(10)

[6]

where, Pinput and Poutput satisfy eqs. (11) and (12), respectively.

[7]

Pinput  I inputVinput cos  input .

(11)

Poutput  I outVout cos  out .

(12)

[8]

Regarding the RMS values shown in Figs. 8(A) and 10, Poutput = 128 W, Pinput = 134 W and η =95.5%.

[9]

5. Conclusions

[10]

In this study, we designed and built a new and efficient soldering iron with pencil presentation and replaceable tips, which works via high frequency induction heating. Moreover, the pencil does not contain any cord to the power supply in order to avoid accidents in the welding process. The device was developed using a resonant inverter with half H-bridge topology controlled using a microcontroller.

The microcontroller reads the temperature of the tip and turns off the device before it reaches the Curie temperature or if the pencil is removed. The device is characterized by determining its curvers of input and output current and voltage.

It exhibits an acceptable power factor and efficiency of 95.5% and a low total harmonic distortion. Furthermore, the device is built using low cost components. It represents a suitable alternative in the welding tasks in comparison to the current devices based on resistances. Indeed, the device is currently being used in our Biophysics lab and has proved to be a very useful tool for welding fragile electronic components on PCBs. Our observations indicate that with the “welding cycle” chosen, an experienced electronics technician may perform up to 45 welds per cycle.

[11]

[12] [13] [14]

[15] [16]

[17] [18]

Acknowledgments All the authors are grateful to the Mexican institution CONACYT for its valuable support.

[19]

References [1]

Field, A. B. Eddy Currents in large Slot-Wound conductors.

American Institute of Electrical Engineers, Transactions of the 26, pp. 761-788, 1905.

[20] [21]

171

Boadi, A., Tsuchida,Y., Todaka,T. and Enokizono, M. Designing of suitable construction of high-frequency induction heating coil by using finite-element method. Magnetics. IEEE Transactions, 41(10), pp. 4048–4050, 2005.

Bayındır, N.S., Kükrer, O. and Yakup, M. DSP-based PLLcontrolled 50–100 kHz 20 kW high frequency induction heating system for surface hardening and welding applications. IEE Proceedings - Electric Power Applications, 150(3), pp. 365-371, 2003. http://dx.doi.org/10.1049/ip-epa:20030096 Burdío, M., Monterde, F., Garcia, J. R., Barragan, L. A. and Martinez, A. A two-output series-resonant inverter for inductionheating cooking appliances. Power Electronics, IEEE Transactions on, 20(4), pp. 815-822, 2005.

Grooms, J. P., Mattson, L. J., Method for induction sealing an inner bag to an outer container, US 5416303A, 16 May 1995.

Lung W. Ch. Induction ironing apparatus and method, US 7681342 B2, 9 October 2006 Jordan, A., Scholz, R., Maier-Hauff, K., Johannsen, Wust, M., Nadobny, P. J., Schirra, H., Schmidt, H., Deger, S., Loening, S., Lanksch, W. and Felix, R. Presentation of a new magnetic field therapy system for the treatment of human solid tumors with magnetic fluid hyperthermia. Journal of Magnetism and. Magnetic.

Materials, 225 (1–2), pp. 118–126, 2001. http://dx.doi.org/10.1016/S0304-8853(00)01239-7 Cano, M. E., Barrera, A., Estrada, J. C., Hernandez, A. and Córdova, T. An induction heater device for studies of magnetic hyperthermia and specific absorption ratio measurements. Review of Scientific Instruments, 82 (11), pp.

114904-114904-6, 2011. http://dx.doi.org/10.1063/1.3658818 Mitsuhiko M.. System and Method for Induction Heating of a Soldering Iron, US/2010/0258554 A1, October 14 of 2010.

Snown, C., Alternating current distribution in cylindrical conductors, in: Scientific Papers of the Bureau of Standards, 20, Washington, USA, pp. 277-338, 1925.

Buschow, K. H. J. Encyclopedia of Materials: Science and Technology.Michigan: University of Michigan, vol 8, Elsevier, 2001. http://dx.doi.org/10.1016/B0-08-043152-6/00016-4, http://dx.doi.org/10.1016/B0-08-043152-6/01367-X, http://dx.doi.org/10.1016/B0-08-043152-6/00841-X Kittel, C. Introduction to Solid State Physics, New York: John Wiley & Sons, 6th ed, 1986.

Brown, G. H., Hoyler, C. N., Bierwirth, R. A., Theory and Applications of the Radiofrequency Heating, New York: D. Van Nostrand Company, 1947.

Dwight, H. B, A Precise Method of Calculation of Skin Effect in Isolated Tubes. Journal of the American Institute of Electrical Engineers, 42(8), pp. 830, 1923. http://dx.doi.org/10.1109/JoAIEE.1923.6593471 Jackson, J.D., Classical Electrodynamics, 3rd ed., Wiley, New York, 1998.

Llorente, S., Monterde, F., Burdio, J.M., Acero, J., A comparative study of resonant inverter topologies used in induction cookers, Applied Power Electronics Conference and Exposition, 7th. Annual IEEE, pp.1168-1174, 2002.

Calleja, H., Fast Response Control Circuit for Resonant Inverters, International Journal of Electronics, 89(3), pp. 233-244, 2002. http://dx.doi.org/10.1080/00207210210122550 Kamli, M., Yamamoto, S. and Abe, M., A 50-150 kHz Half-Bridge Inverter for Induction Heating Applications.IEEE Transactions on Industrial Electronics, 43(1), pp. 163–172, 1996. http://dx.doi.org/10.1109/41.481422 Kawamura,Y., TokiwaM., Kim Y.J., Nakaoka, M., New induction heated fluid energy conversion processing appliance incorporating auto-tuning PID control-based PWM resonant IGBT inverter with sensorless power factor correction, Power Electronics Specialists Conference, Record., 26th Annual IEEE, pp.1191-1197, 1995.

Calleja, H. and Ordonez, R., Induction heating inverter with active power factor correction. International Journal of Electronics, 86(9), pp. 1113-1121, 1999. http://dx.doi.org/10.1080/002072199132888 Ye, Z., Jain, P. K. and Sen, P. C., Full-Bridge Resonant Inverter With Modified Phase-Shift Modulation for High-Frequency AC

Mazón-Valadez et al / DYNA 81 (188), pp. 166-172. December, 2014.

Power Distribution Systems, IEEE Transactions on Industrial Electronics, 54 (1), pp. 2831-2845, 2007.

[22] Goya Gerardo Fabian, Cassinelli Nicolas, Ibarra García Manuel Ricardo, Magnetic Hyperthermia Application Device, PCT/ES2009/000235, November 12 of 2009.

[23] Lacroix, L.M., Carrey, J. and Respaud, M., A frequency-adjustable electromagnet for hyperthermia measurements on magnetic nanoparticles, Review of Scientific Instruments, 79(9), pp. 093909093909-5, 2008. http://dx.doi.org/10.1063/1.2972172 [24] Tai, C. C., Cheng, M. K., Anti-interference Design of Quasiresonant Tank for Magnetic Induction Heating System. PIERS Online, 4(4), pp. 417-420, 2008. http://dx.doi.org/10.2529/PIERS070907021437 E.E. Mazón-Valadez, received, the BIE degree in 2012 from the Universidad of Guadalajara, México, he is Student in the Masters in physics in the Centro Universitario de la Ciénega of the Universidad de Guadalajara, in México. His research interest include: Magnetic Hyperthermia, Resonant Inverters and SAR. A. Hernández-Sámano, received the MPhys degree in 2013 from the University of Guanajuato, México, he is Student in the PhD in physics in the Centro Universitario de la Ciénega of the Universidad de Guadalajara, in México. His research interest include: Magnetic Hyperthermia, Resonant Inverters and Magnetometry.

J.C. Estrada-Gutiérrez, received the MComp. in 2005 from the Universidad of Guadalajara, México, Full Professor en the Technologic Sciences department in the Centro Universitario de la Ciénega of the Universidad de Guadalajara, in México. His research interest include: Magnetic Hyperthermia, Automation and Control and Embedded Systems.

J. Ávila-Paz, received MSc.in 2003 from the Universidad Central Marta Abreu de las Villas, Cuba, Full Professor in the Technologic Sciences department in the Centro Universitario de la Ciénega of the Universidad de Guadalajara, in México. His research interest include: Magnetic Hyperthermia, Automation and Control, and Embedded Systems.

M.E. Cano-González, received the PhD degree in Physics in 2007 from the University of Guanajuato, México. Full professor in the Basic Sciences department in the Centro Universitario de la Ciénega of the Universidad de Guadalajara, in México. His research interest include: Magnetic Hyperthermia, Resonant Inverters and Monte Carlo Simulations.

172

Área Curricular de Ingeniería Eléctrica e Ingeniería de Control Oferta de Posgrados 

Maestría en Ingeniería - Ingeniería Eléctrica

Mayor información: Javier Gustavo Herrera Murcia Director de Área curricular ingelcontro_med@unal.edu.co (57-4) 425 52 64

Selecting working fluids in an organic Rankine cycle for power generation from low temperature heat sources Fredy Vélez Centro Tecnológico CARTIF, Valladolid, España. frevel@cartif.es Received: January 21th, de 2014. Received in revised form: April 29th, 2014. Accepted: May 14th, 2014

Abstract This paper presents a thermodynamic study carried out on the use of low-temperature heat sources for power generation through a subcritical Rankine cycle with organic working fluids. An analysis of the state of the art of this technology shows the selection of the working fluid as an open research line, since until now there is no fluid that can meet all environmental and technical aspects to be considered in these cycles. Hence, we have developed a series of simulations that allow us to study the behavior of the Rankine cycle with different configurations and fluids (wet, dry and isentropic) which has led us to observe the influence on the overall cycle efficiency when we change the type of fluids used (refrigerants, hydrocarbons and water) as well as the conditions of temperature, pressure, flow, etc. With the work realized, the viability of this type of processes is demonstrated for the recovery of heat in industry and/or the use of renewable sources of low and medium temperature for the production of electricity.

Keywords: energy efficiency; organic Rankine cycle; power generation; waste heat; renewable energy.

Seleccionando fluidos de trabajo en ciclos Rankine para generación de energía a partir de fuentes de calor de baja temperatura Resumen Este trabajo presenta un estudio termodinámico realizado sobre el uso de fuentes de calor de baja temperatura para la generación de energía a través de un ciclo Rankine subcrítico con fluidos de trabajo orgánicos. Un análisis del estado del arte de esta tecnología muestra como línea de investigación abierta, la selección del fluido de trabajo, pues hasta ahora, no existe un fluido que satisfaga todos los aspectos medioambientales y técnicos a tener en cuenta en estos ciclos. Por ello, se ha desarrollado una serie de simulaciones que permiten estudiar el comportamiento del ciclo Rankine con diferentes configuraciones y fluidos (húmedo, seco e isoentrópico), permitiendo con ello observar de qué manera influyen cambios tanto en esos tipos de fluidos utilizados (refrigerantes, hidrocarburos y agua), como de condiciones de temperatura, presión, flujo, etc., sobre la eficiencia total del ciclo. Con el trabajo realizado se demuestra la viabilidad de este tipo de proceso en la recuperación de calores en la industria y/o aprovechamiento de fuentes renovables de baja y media temperatura para la producción de energía eléctrica.

Palabras clave: eficiencia energética; ciclo Rankine orgánico; generación de energía; calor residual; energías renovables.

1. Introduction There is no doubt that the implementation of projects based on non-conventional technologies for environmentfriendly energy generation, and with occupational health and safety criteria are essential to ensure social equity and to facilitate economic activities related to the public service of energy off-grid areas, mainly in underdeveloped countries where majority of the territory corresponds to isolated areas with high levels of poverty and without any opportunities for socio-economic development, which is caused, among others,

by the lack of an adequate energy supply. The use of fossil fuels (e.g., oil and coal) as an energy source has many negative environmental impacts, such as the release of pollutants and resource depletion. A high consumption rate of fossil-fuels will result in an increase of environmental pollution during the next century, due to the emission of CO2 and other gases that cause global warming through what is known as the greenhouse effect [1]. These problems encourage efforts to develop new technologies that transform renewable energy sources such as solar, biomass, geothermal, as well as the use of waste and/or low-enthalpy heat now

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 173-180. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41666

VĂŠlez / DYNA 81 (188), pp. 173-180. December, 2014.

discharged in the industry (which represents a 50% or more of the heat generated therein) into electric and/or mechanical energy [2]. However, given that the traditional Rankine cycle does not present a good performance to recover or use low temperature heat, it is necessary to analyze other processes such as the Organic Rankine Cycle (ORC), which allows to leverage different sources of low temperature to power output as was proposed by [2-6] and many other authors. In this sense, the ORC is a promising process for the conversion of these low and medium temperature heat sources from renewable energy and waste heat from industry into electricity [7-10]. The ORC process works like a conventional steam Rankine cycle, with the difference that the former uses an organic compound of low boiling point as the working fluid (refrigerant and hydrocarbon) instead of water vapor, thereby reducing the evaporation temperature. In recent years, commercial applications of this technology, with power ranges from 200-2000 kWe, have increased sharply worldwide.

ORMAT, Barber-Nichols, UTC Power, Turboden, etc, are some of the companies that have developed plants for using geothermal energy, waste heat in industry, biomass, etc.

However, lower power modules are in a prototype stage due to lack of appropriate equipment (mainly the turbine) and the difficulty in selecting a suitable working fluid [3]. The latter characteristic has great influence on the design and behavior of the process [6], because the fluid must have optimal thermodynamic properties at low temperatures and pressures and also satisfy many criteria such as being economical, nontoxic, nonflammable, environmentally friendly, and allow high utilization of the available energy from the heat source, etc; all depending on the application, the source and the heat level to use. If all these aspects are considered, very few fluids can be used [3,11]. In [11], it is shown that these working fluid properties are keys in the performance cycle. In addition, the low temperatures that take advantage of the ORC cause that the efficiency of the cycle is highly sensitive to inefficiencies in heat transfer, which depends strongly on the thermodynamic properties of the fluid and the conditions to which it is operating. A study of 68 potential working fluids conducted in [12] in 1985, gave the best results only for three of them (R11, R113 and R114) which are nowadays not recommended by global environmental conservation policies [13]. In [2], ORC efficiency was analyzed using azeotropic mixtures of 85fluorinol, 2-methyl pyridine/water and substances such as benzene, ammonia, R134a, R113, R11 and R12, achieving greater efficiencies for the last two, however, they are substances of limited use. Other researchers who have analyzed the characteristics and outcomes of different fluids for use in ORC systems are among others [4, 6-12] and [14-16], of whose research can conclude that: R245fa and R134a as good candidates in subcritical cycles and CO2 in transcritical cycles for processes whose heat source is low temperature. Recent publications such as [17] have studied the behavior of R134a, R123, R227ea, R245fa, R290, and n-pentane from the point of view of energy production capacity and component size. They have shown the existence of an optimum pressure that minimizes the heat exchange area, which obviously depends on the source and the source temperature. In [11], a review of thermodynamic properties, physical stability, environmental impact, safety, cost and availability of 35 possible fluids to be

used in a Rankine cycle for the conversion of low temperature heat into electricity has been performed.

The present paper analyzes the behavior of working fluids, such as water and some hydrocarbons and refrigerants in a Rankine cycle in different configurations and conditions of temperature, pressure and flow, thus establishing the viability of this type of process in the recovery and/or utilization of low-temperature energy.

Nowadays, there is no a unique fluid that fulfills all aspects to be considered to use in ORC cycles. This study has selected R600, R600a, R113, R718, R425fa and Toluene, trying to encompass and take into account the optimal physical and thermodynamic properties, economic, toxicity, flammability, environmental friendly (with null Ozone Depletion Potential (ODP) and Global Warming Potential (GWP)). Of the fluids selected, R600a and R600 are hydrocarbons used as refrigerants with zero potential degradation of the ozone layer and very low global warming potential, however, they are highly flammable. Toluene is similar in its behavior, with the drawback of its toxicity. Refrigerant R113 is not flammable, nor toxic, but has high global warming potential and high destruction potential of the ozone layer. R245fa is a safe refrigerant, nonflammable, nontoxic, has zero ozone degradation potential, relatively high molecular weight, good thermophysical properties, but has a moderate potential for global warming. Finally, water (R718), as reference fluid to the being the working fluid in the conventional Rankine cycle.

Additionally, in this study the difference in the type of fluid has been considered and thus, the results obtained enable to determine whether the increase in the inlet temperature increases or decreases turbine cycle performance qualitatively but also, allows to know how this increases or decreases.

2. Description of the cycle The operating principle of the ORC is the same as the conventional Rankine cycle. A pump pressurizes the liquid fluid, which is injected into an evaporator to produce a vapor that is expanded in a turbine connected to a generator.

Finally, the steam is condensed and sucked by the pump, starting the cycle again. It may also include an internal heat exchanger (IHX) to further exploit the energy of the expanded steam, preheating the fluid exiting the pump and entering the evaporator, as is shown in Fig. 1.

Figure 2 represents the power cycle on a T-s diagram according to the state points displayed in the flow diagram of Fig. 1. As an example, an ideal cycle process is shown by the segments constructed from state points 1, 2is, 3 and 4is. Line segment 1-2is represents an isentropic expansion with work production. Heat is extracted from 2is to 3, through a constant subcritical pressure line to subsequently result to ideal compression of the subcooled liquid from the pressure at state point 3 to state point 4is. Finally, the segment 4is-1 corresponds to the heat added to a constant subcritical pressure up to the highest temperature point in the cycle in the state point 1.

A cycle, in which both the expansion and the compression are not ideal, is reproduced by segments in points 1, 2, 3 and 4 of the same Fig. 2.

174

VĂŠlez / DYNA 81 (188), pp. 173-180. December, 2014.

Figure 1. Schematic diagram of organic Rankine cycle with internal heat exchanger. Source: compiled by author.

Figure 2. Diagram Temperature - Entropy of the ideal and real Rankine cycle. Source: [18]

3. Selection of the working fluid As mentioned previously, the selection of the working fluid for use in ORC cycles is a crucial point, because depending on the application, the source and the heat level to use, the fluid must have optimal thermodynamic properties at lowest temperatures and pressures and also satisfy several criteria such as being economical, nontoxic, nonflammable, environmentally friendly, allow a high utilization of the available energy from the heat source, among others [3]. This limits the list to only a few fluids when considering all restrictions.

Some of these limitations are [3]: Environmental: Some fluids are restricted by International Agreements depending on their Ozone Depletion Potential (ODP) defined and limited by Montreal Protocol or Global Warming Potential (GWP) by Kyoto Protocol, which, intend to prevent the destruction of the ozone layer and emission of gases that cause the greenhouse

effect, respectively.

Security: The fluid must be non-toxic (because of the problems that can occur in the case of leaks in the atmosphere or in handling), non-corrosive (it obviously avoids major maintenance costs and/or installation damage) and non-flammable. For this, the standard security classification 34 of ASHRAE is often used as an indicator of the danger level of fluids.

Stability: The chemical stability of the used fluid limits the heat source temperature. When fluids are exposed to certain temperatures, they could decompose, producing substances which could cause a different cycle operation to that initially designed. Moreover, toxic and irritating compounds could induce health problems in case leaks occur.

Pressure: A fluid requiring high pressure to achieve an efficient process increases the cost of equipment due to the greater resistance required, increasing also the complexity of the plant.

Availability and low cost: A fluid of low availability and/or high cost limits its use in ORC plants for obvious reasons in the financial viability of the projects.

Latent heat and molecular weight: With higher molecular weight and the latent heat of the fluid, more energy from the heat source in the evaporator will be absorbed and, thus, reducing the size of the installation and use of the pump due to lower mass flow required.

Low freezing point: The freezing point of the fluid must be lower than the lowest temperature of the cycle.

Curve of saturation: The thermodynamic properties of the fluid imply that the slope of the saturation curve thereof is negative, vertical or positive, which markedly affecting the design and efficiency of the ORC. In Fig. 3.a, b and c, a schematic diagram is shown of Temperature - Entropy (T-s) for fluids with a negative saturation curve (a), vertical (b) and positive (c), called wet, isentropic and dry, respectively.

Since the objective of the ORC focuses on the use of heat of low and medium temperature, the overheating of steam as in a conventional Rankine cycle is not appropriate.

Furthermore, as is seen in Fig. 3.a, when a wet fluid is expanded without overheating (represented by the segment of points 1-2), it falls in the liquid/vapor zone causing damages to the expander and cycle inefficiencies. This happens, among other reasons, because of the phase shift.

Nevertheless, the contrary occurs with isentropic and dry fluids which, without any type of overheating, suffer expansion falling into the saturated vapor zone, Fig. 3.b, and/or superheated, Fig. 3.c, respectively, and thus, in the latter case, it could include an internal heat exchanger to the cycle which allows to use even more the energy from the expanded steam, preheating fluid from the pump to enter the evaporator, thereby increasing the efficiency of the cycle.

In summary, for an organic Rankine cycle, the ideal working fluid will be that whose saturated vapor line is parallel to the expansion of the turbine, ensuring maximum efficiency. In this case, turbine will be always working in the dry steam area. If both lines converge, the turbine would operate in the wet steam region. In order to avoid this, it should superheat the working fluid before the expansion. If, however, the lines mentioned above were divergent, the

175

Vélez / DYNA 81 (188), pp. 173-180. December, 2014.

output fluid from the turbine would exit overheated. This would increase significantly the size of the condenser surface, an aspect that is related to the operating pressure.

The major disadvantage of ORC is its relatively low efficiency (inherent to the thermodynamic constraints) and the relatively large sizes of heat exchangers equipment.

4. Modelling of the process In both cycles, the equations used for performance calculation assume steady state and constant efficiencies of 75% for both the pump and the turbine. In addition, there are no heat or pressure losses in the evaporator, heat exchanger or pipes, and an overall cycle efficiency (ɳ) is considered as:



Wt  W p Q e

(1)

Where, W is the power, “t” refers the turbine, “p” to the pump and "e” evaporator. Hence:

W t  m   t  h1  h 2 

W p  m  h3  h4  /  p

m is the mass flow, point “i”. as:

(2) (3)

hi is the enthalpy in the state

 e is the heat input in the evaporator defined Whereas Q Q e  m  h1  h4  or

Q e  m  h1  h4 IHX 

(4)

Through the theoretical basis presented in the preceding sections and using the software HYSYS®, a series of simulations has been developed that allows to study the behavior of the ORC with different configurations and fluids (as refrigerants, hydrocarbons and even water), thereby observing how the changes influence the overall cycle efficiency for these types of fluids used varying conditions of temperature, pressure, flow, etc.,. This simulator is useful for thermodynamic analysis, especially in steady-state conditions, and it has the advantage of including fluid properties and optimization tools. Its predictions have been compared with those of [19] and the results are very similar. The flow chart of the simulation is the same as it was presented in Fig. 1, and the method to solve all the components is fully described in [7,10] and [14].

5. Results and discussions The results obtained with the simulations for water, hydrocarbons R600, R600a, Toluene and refrigerants R113 and R245fa are shown in this section.

5.1.

Working Fluids: Water, R600, R600a, Toluene and R113

5.1.1. Influence of inlet temperature in the turbine on the overall cycle efficiency Figure 3. Diagram T-s for fluids wet (a), isentropic (b) and dry (c).

Source: [12]

The influence of inlet temperature to the turbine on the 176

VĂŠlez / DYNA 81 (188), pp. 173-180. December, 2014.

Figure 4. Influence of inlet temperature to the turbine on the total efficiency of the cycle at a constant condensation pressure.

Source: compiled by author.

overall cycle efficiency is presented in Fig. 4. Here, conditions of input and output pressure in the turbine and the fluid flow are kept constant for the five fluids analyzed and for the different simulations carried out with values of 20 bar, 2 bar and 2.8 kg/s, respectively. It is evident that the efficiency increases linearly with the inlet temperature to the turbine in the case of R718 (Water). Unlike this wet fluid, the dry fluids R600, R600a, R113 and Toluene showed a decrease of efficiency with the increase in inlet temperature to the turbine.

Thus, there is a big difference in the efficiencies obtained according to the type of fluid used, the highest efficiency obtained is for the case of R600a, followed by R600, R718, R113 and Toluene. It is also interesting to note that as Fig. 4 shows, the use of one fluid or another will depend on the heat source that is intended to be recovered, and the thermophysical properties of the fluid, since there are fluids which would obtain relatively high efficiencies at relatively low temperature ranges when compared to other fluids that are useful in other temperature ranges.

5.1.2. Influence of the working fluid flow on the overall efficiency of the cycle Fig. 5 shows the influence of the working fluid flow on the efficiency of the cycle with the condensation pressure constant and equal (2 bar) for all the studied fluids, maintaining constant in addition the inlet pressure to the turbine at 20 bar for the five fluids analyzed and for all simulations realized whereas the rate of energy supplied to the evaporator is also constant for all simulations carried out but different for each fluid. Values achieved were 1500 kW, 1650 kW, 7778 kW, 795 kW and 2270 kW for the R600a, R600, R718, R113 and Toluene, respectively.

Thermodynamic properties of each fluid are different, therefore, for the same state point, different values of enthalpy, entropy, etc., could be obtained for two or more types of fluids. In addition, the molecular weight and the vaporization latent heat of a specific fluid, determines the amount of fluid flow (greater or smaller) that is required to have to a given power, compared to other fluid. In this sense, to make the flow values similar, i.e. to keep them in

Figure 5. Influence of the working fluid flow on the total efficiency of the cycle at a constant condensation pressure.

Source: compiled by author.

Figure 6. Influence of the inlet pressure to the turbine on the total efficiency of the cycle at a constant condensation pressure. Source: compiled by author.

the same range, the input power to the evaporator was fixed, for the range of study. In Fig. 5, it is observed that a slight influence exists of the working fluid on the overall efficiency of the cycle for the case of R600a, R600, R113 and Toluene, whereas for the R718, the effect is clear but opposite to that shown previously, i.e, a decrease of the cycle efficiency when the flow of R718 increases.

5.1.3. Influence of the input pressure to the turbine on the overall efficiency of the cycle.

Based on the results shown in Fig. 4 and Fig. 6, the influence of the input pressure to the turbine on the overall efficiency of the cycle is analyzed, the fluid flow of 2.8 kg/s was held constant for the five fluids analyzed and for the different simulations, the rate of energy supplied to the evaporator is also constant for the different simulations but different for each fluid and whose values were 1500 kW, 1650 kW, 7778 kW, 795 kW and 2270 kW for R600a, R600, R718, R113 and Toluene, respectively. Fig. 6 shows that the system gains in efficiency with the increase of the

177

Vélez / DYNA 81 (188), pp. 173-180. December, 2014.

input pressure to the turbine. Results are consistent for all the fluids utilized. Higher input pressures to the turbine raise the net work that leads to an improvement of the efficiency.

Although the efficiency of the system also increases when the pressure of the system rises, this increase of the pressure in the system is not always feasible for economic reasons since the capital costs for the waste heat of a boiler and piping systems, as well as the complexity of the system and the component selection of the materials must also be taken into account.

5.2. Working fluid: R245fa As it was discussed in the introduction to this paper, the refrigerant R245fa is a very promising working fluid for using in low temperature ORC because it meets many of the aspects to be considered in these applications. The R245fa fluid is a safe refrigerant (nonflammable, nontoxic), has a potential ozone degradation of zero, relatively high molecular weight, good thermophysical properties (critical temperature and pressure, boiling point, etc.), has a positive saturation curve, which gives higher yields, allowing the preheating of the fluid exiting the pump using the steam exiting the turbine, etc. However, it has a moderate potential for global warming as its main disadvantage.

For those reasons, the approach to the thermodynamic study of this fluid has been done differently from that illustrated in the previous section. In this case, we have analyzed the effect of inlet pressure and temperature to the turbine on the other parameters (efficiency, flow, net work and heat supplied to the evaporator) to an ORC with a fixed output gross power of 125 kWe and an IHX to preheat the liquid fluid from the pump with the output steam from turbine (see Fig. 1). Simulations considered a P2=2.5 bar and a condensing temperature of the working fluid (T3) of 39.7°C. In addition, the heat required for the evaporator ( Q e ) does not take into account the efficiency for example

Figure 7. Mollier diagram of R245fa.

Source: [20].

Figure 8. Influence of the inlet pressure P1 to the turbine on the flow of R245fa in a cycle with IHX. Source: compiled by author.

from boiler and heat transfer in the heat exchanger.

5.2.1. Influence of the inlet pressure to the turbine on the fluid flow in the cycle As in this analysis, the work produced by the turbine (

W t ) was constant and equal to 125 kWe for all

simulations, it can be stated that: keeping the condition of turbine discharge constant P2=2.5 bar and considering the fact that the work produced by the turbine is given by the eq. (2), and, supported by the Mollier diagram of this fluid (Fig. 7), we see that, for any constant inlet temperature to the turbine (for example, T1=145°C), increasing the inlet pressure to the turbine (P1), the delta enthalpy, ∆h, increases from 30 kJ/kg to 42 kJ/kg (segment 1-2, red and blue colors of this Fig. 7, respectively) and therefore, the mass flow ( m  ) necessarily decreases (for any of the three temperatures studied 90°C, 120°C and 145°C), as depicted in Fig. 8 and thus to maintain constant value of the work produced by the turbine.

Figure 9. Influence of the inlet pressure to the turbine P1 on the net work of the cycle with IHX and R245fa.

Source: compiled by author.

178

Vélez / DYNA 81 (188), pp. 173-180. December, 2014.

5.2.2. Influence of the inlet pressure to the turbine on the net work of the cycle

o r d e r

t o

c a r r y

o u t

Likewise, as is shown in Fig. 9, as inlet pressure increases in the turbine, it causes a decrease in the net work produced by the system ( W net  W t  W p ), because for the turbine work to be constant with increasing pressure, the work consumed by the pump is increased for any of the three temperatures studied 90°C, 120°C and 145°C. 5.2.3. Influence of the inlet pressure to the turbine on the efficiency and the heat required in the cycle.

Similarly, and despite what was discussed above, the net work ( W net ) always decreases with increasing inlet pressure to the turbine P1, system efficiency increases, as is shown in Fig. 10. This occurs because increasing such pressure P1, as is seen in Fig. 10, the amount of heat required in the evaporator decreases with a decreasing mass flow (Fig. 8).

This is produced because the mass needed to be heated/evaporated to maintain the work produced by the turbine constant is less with respect to its own net work.

The previous situation occurs until that by the own characteristics of this fluid, there is a specific value for the pressure (the optimum in efficiency terms), in which an increase of the variable P1 induces a negligible increase of ∆h (see Fig. 7), and therefore, a lower value of the mass flow, causing the heat from the evaporator to drop a little but with sufficient value to counteract the decrease in the net work and therefore maintaining the efficiency nearly constant. In this case, and according to Figs. 8 to 10, this optimum inlet pressure to the turbine is close to 20 bar for the temperature 145°C, 15 bar for the temperature 120°C and 10 bar for 90°C.

Figure 10. Influence of the inlet pressure to the turbine P1 on the efficiency and the heat required by the evaporator in a cycle with IHX and R245fa.

Source: compiled by author.

his doctoral thesis, on which this paper is based. In addition, the author expresses his special gratitude to the Department of Renewable Energy Service of the Spanish Company Nicolás Correa, SA, for all the technical support provided in the preparation of this research.

References [1] [2]

[3]

4. Conclusions Based on the discussed in this paper, ORC technology seems to have an enormous potential with renewable energy sources such as biomass, solar and geothermal energy as well as waste heat from industrial processes or other cycle for transforming low temperature heat sources into electricity. However, the thermophysical properties, environmental impact, safety and stability as well as its availability and cost, play a critical role in the selection of the working fluid in these cycles.

Additionally, according to research carried out in this work, it can be concluded that using organic working fluids in Rankine cycles, relatively good efficiencies are obtained recovering and/or using heats of low/medium temperature to convert them into electricity.

[4]

[5]

[6]

[7]

[8]

Acknowledgements Fredy Vélez would like to thank the scholarship granted by the “Programa Iberoamericano de Ciencia y Tecnología para el Desarrollo”, CYTED, CARTIF Technological Center and the University of Valladolid in

[9]

179

Realpe, A. and Diazgranados, J. A. Electricity generation and wind potential assessment in regions of Colombia. Dyna, 79(171), pp. 116-122, 2012.

Hung, T. C., Shai, T. Y. and Wang, S. K. A review of organic Rankine cycles (ORC`s) for the recovery of low-grade waste heat.

Energy, vol.

22(7), pp.

661-667, 1997. http://dx.doi.org/10.1016/S0360-5442(96)00165-X Vélez, F., Segovia, J., Martín, M. C., Antolín, G., Chejne, F. and Quijano A. A Technical, economical and market review of organic rankine cycles for the conversion of low-grade heat for power generation. Renewable & Sustainable Energy Reviews, 16(6) pp.

4175– 4189, 2012. http://dx.doi.org/10.1016/j.rser.2012.03.022 Saleh, B., Koglbauer, G., Wendland, M. and Fischer, J. Working fluids for low temperature organic Rankine cycles. Energy, vol.

32(7), pp.

1210–1221, 2007. http://dx.doi.org/10.1016/j.energy.2006.07.001 Quoilin, S., Declaye, S., Tchange, B. F. and Lemort, V. Thermoeconomic optimization of waste heat recovery organic Rankine cycles. Applied Thermal Engineering, 31(14), pp. 2885-2893, 2011. http://dx.doi.org/10.1016/j.applthermaleng.2011.05.014 Tchange, B.F., Papadakis, G., Lambrinos, G. and Frangoudakis, A., Fluid selection for a low-temperature solar organic Rankine cycle. Applied Thermal Engineering, 29(1), pp. 2468–2476, 2009. http://dx.doi.org/10.1016/j.applthermaleng.2008.12.025 Vélez, F., Chejne, F., Antolín, G. and Quijano, A. Theoretical analysis of a transcritical power cycle for power generation from waste energy at low temperature heat source. Energy Conversion and Management, vol.

60, pp.188–195, 2012. http://dx.doi.org/10.1016/j.enconman.2012.01.023 Mago, P. J., Chamra, L. M., Srinivasan, K. and Somayaji, C. An examination of regenerative organic Rankine cycles using dry fluids. Applied Thermal Engineering, 28(8), pp. 998–1007, 2008. http://dx.doi.org/10.1016/j.applthermaleng.2007.06.025 Schuster, A., Karellas, S., Kakaras, E. and Spliethoff, H., Energetic and economic investigation of organic Rankine cycle applications.

Applied Thermal Engineering, 29(8-9), pp. 1809–1817, 2009. http://dx.doi.org/10.1016/j.applthermaleng.2008.08.016

Vélez / DYNA 81 (188), pp. 173-180. December, 2014.

[10] Vélez, F., Segovia, J., Chejne, F., Antolín, G., Quijano, A. and Martín, M.C. Low temperature heat source for power generation: exhaustive analysis of a carbon dioxide transcritical power cycle.

Energy, 36(9), pp. 5497-5507, 2011. http://dx.doi.org/10.1016/j.energy.2011.07.027 [11] Chen, H., Goswami, Y. and Stefanakos, E. A review of thermodynamic cycles and working fluids for the conversion of low-grade heat. Renewable and Sustainable Energy Reviews, 14(9), pp.

3059–3067, 2010. http://dx.doi.org/10.1016/j.rser.2010.07.006 [12] Badr, O., Probert, S. D. and O'callaghan, P.W., Selecting a working fluid for a Rankine-cycle engine. Applied Energy, 21(1), pp. 1-42, 1985. http://dx.doi.org/10.1016/0306-2619(85)90037-6 http://dx.doi.org/10.1016/0306-2619(85)90072-8 http://dx.doi.org/10.1016/0306-2619(85)90033-9 [13] U.S. Environmental Protection Agency. Class I Ozone Depleting Substances. [Online]. [date of reference March 11th of 2013] Available at: www.epa.gov/ozone/science/ods/classone.html.

[14] Velez, F., Segovia, J., Martín, M. C., Antolín, G., Chejne, F. and Quijano, A. Comparative study of working fluids for a Rankine cycle operating at low temperature. Fuel Processing Technology, vol.

103, pp.

71–77, 2012. http://dx.doi.org/10.1016/j.fuproc.2011.09.017 [15] Yamamoto, T., Furuhata, T., Arai, N. and Mori, K. Design and testing of the organic Rankine cycle. Energy, 26(3), pp. 239–251, 2001. http://dx.doi.org/10.1016/S0360-5442(00)00063-3 [16] Drescher, U. and Brüggemann, D. Fluid selection for the organic Rankine cycle (ORC) in biomass power and heat plants. Applied Thermal Engineering, 27(1), pp.

223–228, 2007. http://dx.doi.org/10.1016/j.applthermaleng.2006.04.024 [17] Lakew, A., Bolland, O. Working fluids for low temperature heat source. Applied Thermal Engineering, 30(10), pp. 1262–1268, 2010. http://dx.doi.org/10.1016/j.applthermaleng.2010.02.009 [18] Monsalve, E., Cecchi, P., Vidal, A., Zuñiga, A. Informe I: Ciclo de Rankine., Departamento de Ingeniería Mecánica, Facultad de Ciencias Físicas y Matemáticas, Universidad de Chile, Santiago de Chile, Chile, 2008.

[19] Lemmon, E. W., Huber, M. L. and Mclinden, M. O. Reference fluid thermodynamic and transport properties (REFPROP). NIST Standard Reference Database 23, Version 8.0; 2007.

[20] Organic-rankine-cycle.blogspot. [en línea] [date of reference January 15th of 2014]. Available at: http://goo.gl/Oj7sxf.

F. Vélez, PhD. in Energetic and Fluid-Mechanical Engineering by University of Valladolid (2011). MSci in Engineering -Emphasis in Chemical Engineering- (2007) and Chemical Engineering (2004), both

by Colombian National University, where he was also an Associate Professor at the Energy and Process School. He started to work as researcher in the field of renewable energies in 2004. Since 2007 he is actively working in the Energy Department of CARTIF in RTD projects about energy efficiency and integration of renewable energy (solar, geothermal, biomass) for the production of heating and cooling and/or electricity generation in buildings (to achieve zero emission buildings and near-zero energy balance) and industrial processes. He has experience in national and international (European and Latin American) projects, and has published many papers in peer review and technical journals and contributions in conferences about these themes.

ORCID 0000-0003-0764-1321

180

Área Curricular de Ingeniería Química e Ingeniería de Petróleos Oferta de Posgrados   

Maestría en Ingeniería - Ingeniería Química Maestría en Ingeniería - Ingeniería de Petróleos Doctorado en Ingeniería - Sistemas Energéticos

Mayor información: Abel de Jesús Naranjo Agudelo Director de Área curricular qcaypet_med@unal.edu.co (57-4) 425 5317

Approach to biomimetic design. Learning and application Ignacio López-Forniés a & Luis Berges-Muro b a

Departamento de Ingeniería de Diseño y Fabricación, Universidad de Zaragoza, Zaragoza, España. ignlopez@unizar.es Departamento de Ingeniería de Diseño y Fabricación, Universidad de Zaragoza, Zaragoza, España. bergesl@unizar.es

b

Received: January 22th, 2014. Received in revised form: July 23th, 2014. Accepted: July 28th, 2014.

Abstract Learning biomimetic design methods can be difficult and involve a long period of time, previous experiences have shown that the application of biomimetics in industrial design and product development projects requires an adaptation to traditional methods through a simpler model and a learning system. There are various biomimetic design processes, adaptation, integration and application corresponds to objectives of each project.

A learning method for the application of biomimetics in design projects, step by step or globally, is presented. The knowledge of biological principles and translating them into engineering principles underlie each method, learning is augmented by adding methods based on creativity and functional analysis to achieve innovative products thanks to a biomimetic approach.

Keywords: Design methods, learning biomimetics, biomimetic design methodology, bioinspired design, creative process, functional innovation, solutions from nature.

Aproximación al diseño biomimético. Aprendizaje y aplicación Resumen El aprendizaje de los métodos de diseño biomimético puede resultar difícil e implicar un largo periodo de tiempo. Experiencias previas han demostrado que la aplicación de la biomimética en proyectos de diseño industrial y desarrollo de producto necesita de una adaptación a los métodos tradicionales por medio de un modelo más sencillo y a través de un sistema de aprendizaje. Existen diversos procesos de diseño biomimético, su adaptación, integración y aplicación se corresponde con objetivos propios de cada proyecto.

Se presenta un modelo de aprendizaje de la biomimética para aplicación en proyectos de diseño, de sus partes y de manera global. El conocimiento de los principios biológicos y su traducción a los principios de ingeniería son la base de cada método, incrementando el aprendizaje con aquellos que se basan en técnicas creativas y de análisis funcional para conseguir innovación funcional de productos gracias al enfoque biomimético.

Palabras clave: Métodos de diseño, aprendizaje en biomimética, metodología de diseño biomimético, diseño bioinspirado, proceso creativo, innovación funcional, soluciones naturales.

1. Introducción La naturaleza es un modelo para la tecnología y por lo tanto fundamental para sentar las bases para el desarrollo de nuevos productos [1]. La biomimética está empezando a formar parte del aprendizaje de métodos de diseño y desarrollo de producto, se usa como fuente creativa y por el potencial demostrado en combinación con otras metodologías [2].

La biomimética, como observación de la naturaleza de la que extraer soluciones para la resolución de problemas técnicos, ha tomado un fuerte empuje en la investigación en ingeniería [3]. Áreas como la biomecánica [4,5], ciencia de los materiales [6,7] o robótica [8] entre otras cuentan con

numerosos casos de éxito, y es en el campo del diseño de producto donde comienzan a emerger metodologías de trabajo.

No ha sido desarrollado un método general y único para la biomimética, actualmente se están desarrollando métodos para buscar analogías funcionales en la literatura biológica para llevarlas a la práctica [3,9].

Con el modelo propuesto se consiguen mejoras funcionales de producto, aspecto poco explorado en biomimética. Para hacer la biomimética más accesible a los diseñadores es preciso un método generalizado, que identifique y utilice los fenómenos biológicos relevantes aplicables a problemas de ingeniería de una manera objetiva y repetible [10].

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 181-190. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41671

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

Existen esencialmente dos procesos de diseño biomimético. Los procesos dónde el diseño se inspira en la biología y ante un problema de diseño se emplean referentes de la naturaleza para dar una solución. Y los procesos de investigación biológica que producen conocimiento científico con valor añadido, que permiten desarrollarlo y traducirlo a soluciones aplicables en el ámbito artificial.

Según los investigadores se han denominado por pares y de diversas formas: directos/indirectos, bottom-up/topdown, diseño bioinspirado dirigido por una solución encontrada/ diseño bioinspirado dirigido por un problema dado o el caso particular del BioTRIZ [3,11-15]. El modelo presentado aúna ambos procesos de manera individual e integrada en un aprendizaje por experimentación.

La búsqueda de alternativas funcionales de producto tiene un carácter exploratorio, investigar las funciones resueltas por la naturaleza permite comprender como se resuelven para aplicarlas al diseño. La innovación de productos implica la integración y uso de nuevos métodos de diseño según modelos basados en analogías de estructura-función-comportamiento, donde la definición de las características estéticas y técnicas es determinante [1620].

Este modelo de aprendizaje biomimético se fundamenta en proyectos de aprendizaje basado en proyectos [21] y aprendizaje por experimentación [22]. Está validado teórica y empíricamente por el Cuadro de Validación [23], y cualitativa y cuantitativamente por una encuesta a quienes lo han puesto en práctica por su aplicabilidad, beneficios y utilidad [24].

2. Contexto Existe una gran variedad de métodos y procesos que han sido estudiados y empleados en diseño industrial [25-28] y entre ellos se encuentra la biomimética, el método aquí propuesto destaca por presentar a los diseñadores la observación de la naturaleza como ámbito de estudio, siendo un espacio en el que encontrar soluciones análogas aplicables a funciones innovadoras.

La biomimética no es una enseñanza generalizada en los grados de diseño industrial y son pocas las universidades que la integran en su currículo. Las metodologías biomiméticas son difíciles de aprender y de aplicar por los diseñadores noveles, además implican una excesiva carga de trabajo en ámbitos desconocidos como la biología [29].

Por esta razón, evidenciada en experiencias previas [30], es necesario realizar un nuevo modelo y un sistema de aprendizaje, como se ha realizado desde el Biomimicry Institute [31] en sus programas de inmersión.

El objetivo principal es familiarizar al diseñador con el diseño biomimético y con la comprensión de los principios observados en las soluciones funcionales de la naturaleza.

Otro objetivo es realizar un aprendizaje incremental por medio de métodos parciales de experimentación, por ello se proponen tres experiencias de aprendizaje, todas ellas efectivas en si mismas que afianzan una serie de conocimientos sobre biomimética que harán más sencillos los futuros proyectos.

Si bien la aplicación de la metodología completa supone

un trabajo muy amplio, cada uno de estos métodos se puede utilizar por separado, esto ha sido probado en ejercicios y proyectos realizados con dos grupos de alumnos de la asignatura de Biónica en el grado de Diseño Industrial y Desarrollo de Producto y un grupo de alumnos del curso “Biomimética, la naturaleza como fuente de soluciones” organizado por Universa y la propia Universidad de Zaragoza.

La utilización de dos grupos permite contrastar los resultados de los proyectos y la evaluación y valoración personal de todos los integrantes. Estos grupos están constituidos por sujetos con diferentes perfiles.

El primero es un grupo homogéneo y con formación en diseño, pertenecen a la titulación de Grado en Ingeniería en Diseño Industrial y Desarrollo de Producto y cursan Biónica como optativa, asignatura de reciente incorporación a los estudios de diseño para mejorar los contenidos formativos.

El grupo tiene 36 componentes, que trabajaron individualmente, en parejas y en 9 grupos de 4 componentes, según el ejercicio desarrollado. Todos los integrantes tienen experiencia en aplicación de métodos de diseño y están familiarizados con metodologías de diseño por fases. Previamente han realizado ejercicios y proyectos en los que realizan fases según el proceso tradicional. Para el primer grupo no se plantean dificultades en la aplicación de esta metodología, ni en el desarrollo de este proceso, ya que utilizan métodos conocidos y los relacionan con otros nuevos.

El segundo grupo es heterogéneo y está constituido por licenciados y diplomados de diferentes titulaciones, ingeniería mecánica, electrónica, arquitectura, estadística, química, veterinaria, geología y algunos diseñadores. El grupo está formado por 12 componentes, que formaron 4 grupos de 3 integrantes. Tan solo los diseñadores tenían experiencia previa en desarrollo de este tipo de proyectos por lo que se formaron tres grupos con un diseñador cada uno, los ingenieros y arquitectos tenían experiencia en proyectos pero no de este tipo. Para alguno de los componentes este modo de trabajo era completamente nuevo y no tenían referencia anterior.

Los alumnos del segundo grupo pertenecen a un curso de formación complementaria no reglada, por lo que la duración y dedicación al proyecto es menor. Hay que destacar que en este grupo todos los individuos tienen interés en la biomimética, por lo que su actitud era positiva y participativa. Su madurez les ha ayudado a superar la falta de hábito con la metodología, asimilar los conocimientos de manera rápida, y a tomar decisiones de una manera eficaz.

3. Descripción del modelo La principal contribución del modelo propuesto al proceso de enseñanza de diseño industrial radica en que la biomimética se utiliza como un método creativo de los clasificados como analógicos [32], que recurren a la aproximación de componentes, estructuras o funciones del ámbito natural al tecnológico, de este modo los métodos creativos aplicados al diseño se enriquecen con una gran diversidad de soluciones.

La analogía constituye la relación de similitud entre

182

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

elementos de dos hechos u objetos, que permite deducir mentalmente cierto grado de vínculo entre dichos hechos u objetos. El proceso creativo por analogía biológica se establece en la búsqueda de la relación de similitud, se genera una imagen mental del problema a solucionar y se vincula a seres vivos que ya han solucionado dicho problema. La analogía biológica permite hacer una interpretación técnica de algo que ya existe en la naturaleza sin necesidad de generar una invención, establece una transferencia de conocimiento.

Se define experiencia de aprendizaje (EA) como la actividad realizada para llegar a conocer un método parcial y su integración en una metodología global. Estas experiencias permiten evaluar el aprendizaje del método y su aplicabilidad, en estas experiencias se ha evidenciado la posibilidad de conexión entre métodos. Se entiende por analogía biomimética la acción de aislar el principio de la naturaleza y representarlo de una manera técnica, como un principio de física, química, ciencia de los materiales, de ingeniería, etc. De esta manera se relacionan el principio biológico y el ingenieril.

La Tabla 1 presenta un resumen del modelo, que integra diversos métodos de diseño, relacionados con la analogía biomimética y aplicables al diseño conceptual. Su ensayo consiste en realizar ejercicios como partes de un proyecto en el que dar solución a un problema técnico basándose en soluciones extraídas de la naturaleza. En la EA3 se integran los resultados del aprendizaje de la EA1 y EA2. En la columna derecha de la Tabla 1 se observan las actividades realizadas y los aspectos valorados en cada experiencia de aprendizaje.

Tabla 1.

Actividades y valoración del aprendizaje.

Experiencia de aprendizaje.

Actividades y valoración de la experiencia.

Selección del caso, su referente natural y el principio biológico utilizado.

Principio ingenieril del caso y su relación con el principio biológico.

EA1 Análisis de un caso Principio ingenieril de la propuesta y su relación con el principio biológico. real de biomimética.

Diferencias entre el principio ingenieril del caso y de la propuesta.

Fuentes de información utilizadas.

Individuo natural y característica propia.

Análisis del principio biológico.

EA2 Solución natural Traducción al principio ingenieril. Grado de aplicable al diseño de un relación entre el principio biológico y el objeto. ingenieril.

3.1. EA1. Estudio de un caso de biomimética.

La primera experiencia de aprendizaje se realiza de forma individual y es un ejercicio de investigación documental, se busca información para describir un caso de bomimética, y adquirir conocimiento suficiente que se utilice en el diseño de un objeto, distinto al objeto de referencia. Se aprende a hacer búsquedas en bases de datos específicas, en publicaciones científicas o en internet y se valora la calidad de la información encontrada y su utilidad.

Las búsquedas principalmente se realizan en www.asknature.com que tiene un repositorio de ejemplos, referencias y casos de estudio fácilmente accesible, además cuenta con links a los artículos científicos para examinar el principio biológico.

El objetivo de este ejercicio es comprender cómo en el caso el análisis del principio biológico es aplicado en una solución técnica o en un producto gracias a la descripción de un principio de ingeniería. Una vez que este principio se conoce se reutiliza aplicándolo en una nueva solución o producto, definiendo a nivel conceptual un producto. Se define el tipo de analogía biomimética entre la solución natural y la técnica, describiendo si se trata de una copia, imitación, emulación, inspiración, etc., y justificándolo adecuadamente.

En la Fig.1 se describe el proceso de diseño aplicado a esta EA1, que se inicia con la selección de un caso en las fuentes antes citadas en el que se describe la analogía biológica para comprender la traducción del principio biológico en ingenieril, posteriormente el alumno puede crear una nueva aplicación de este mismo principio que le servirá para definir una nueva solución de diseño a nivel conceptual.

Definimos principio biológico como la base o razón fundamental que da sentido y explica un fenómeno que se da en la naturaleza, ya sea de carácter funcional, estructural, material, formal o de comportamiento. Llamamos principio ingenieril a la justificación y demostración del principio técnico obtenido de la transformación del principio biológico, es decir, es una explicación técnica razonada y útil que toma como base de entendimiento un fundamento de la naturaleza.

A modo de ejemplo podemos citar un ejercicio en el que se seleccionó el caso del vendaje inteligente [33], cuya característica es que el textil de polipropileno tiene estructuras microscópicas en forma de sacos que imitan las células humanas, causando que las bacterias nocivas las ataquen. Cuando las bacterias dañinas liberan toxinas o

Propuesta de aplicación conceptual.

Definición del marco de trabajo.

Definición de las funciones clave.

Tabla biomimética, búsqueda de los referentes EA3 Problema técnico y naturales. resolución gracias a la Grado de relación. Analogía entre principio biomimética. biomimético y principio ingenieril.

Propuesta conceptual.

Aspectos destacables de los resultados respecto a la metodología.

Fuente: Elaboración propia.

1

SELECCIÓN DE UN CASO DE BIOMIMÉTICA

WEBS, ARTÍCULOS CIENTÍFICOS, BASES DE DATOS

2

COMPRENSIÓN DE LA ANALOGÍA BIOLÓGICA

TRADUCCIÓN DEL PRINCIPIO BIOLÓGICO EN INGENIERIL

3

DEFINICIÓN DEL PRINCIPIO INGENIERIL

NUEVA APLICACIÓN DEL PRINCIPIO INGENIERIL

4 DISEÑO CONCEPTUAL Figura 1. Proceso de diseño de la EA1.

Fuente: Elaboración propia.

183

DEFINICIÓN DE LA NUEVA SOLUCIÓN DE DISEÑO

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

enzimas para romper la pared de estas estructuras que imitan a las células, los agentes antibacterianos son liberados, matando inmediatamente a las bacterias ofensivas. Este principio ingenieril que se basa en la degradación del material se aplica de manera conceptual en una “jeringa inteligente” que con un textil similar puede detectar la presencia de patógenos en las extracciones de sangre.

La traducción en un fundamento ingenieril puede tener distintos niveles de analogía (ver Fig. 2). Cuando el principio ingenieril tiene un parecido bajo con la naturaleza se puede decir que son analogías basadas en la inspiración, la sugerencia o la interpretación, con un grado de resolución técnica bajo, mientras que cuando el parecido es alto se trata de réplicas o copias de la naturaleza y la resolución técnica es alta y muy sofisticada.

Vincent [34] plantea en los mapas biomiméticos que cuanto más abstracto es el principio biológico de origen más potencial tendrá el concepto, sin embargo la solución técnica requiere ser lo más específica posible y eso implica que se debe tender a una réplica del proceso natural como se plantea en el diseño de materiales.

La inspiración formal y ornamental [35] ha sido un elemento clave en el diseño de objetos aplicando un alto grado de abstracción del parecido con la naturaleza pero con una baja componente técnica, ejemplos claros se encuentran en el diseño de muebles o en iluminación.

Sin embargo cuando tenemos necesidad de un alto grado de resolución técnica el parecido con la naturaleza es muy alto y se llega a la copia literal de estructuras o materiales de los que existen numerosos ejemplos [36,37], como el ejemplo de los “Double Layer Anti Reflective (DLAR) coatings” empleados en los recubrimientos de las placas solares para aumentar la eficiencia energética y que se inspiran en los ojos de la polilla.

3.2. EA2. Aplicación de una solución de la naturaleza.

Se realiza en parejas y se trata de un ejercicio en el que una solución de la naturaleza se aplica al diseño de un objeto. En la Fig. 3 se resume el proceso de la EA2 que comienza por la búsqueda y estudio de algún individuo o grupo de individuos que tenga alguna característica, principio o facultad destacable, la elección de la característica es libre y se debe aplicar al diseño conceptual posteriormente. Una vez obtenida dicha característica se estudia y se define el principio biológico por medio de documentación específica y científica de la biología, cuanta más calidad y detalle tenga la información más

Figura 2. Grados de analogía en el principio ingenieril.

Fuente: Elaboración propia.

1

SELECCIÓN DE INDIVIDUO DE LA NATURALEZA

BÚSQUEDA DE UNA CARACTERÍSTICA DESTACABLE

2

ANÁLISIS DEL PRINCIPIO BIOLÓGICO

DEFINICIÓN DE LA CARACTERÍSTICA

3

TRADUCCIÓN AL PRINCIPIO INGENIERIL

DEFINICIÓN DEL GRADO DE RELACIÓN ENTRE PB Y PI

4 DISEÑO CONCEPTUAL

DEFINICIÓN DE LA NUEVA SOLUCIÓN DE DISEÑO

Figura 3. Proceso de diseño de la EA2.

Fuente: Elaboración propia.

posibilidades de obtener un grado de resolución técnica alto y menor será el grado de abstracción. A continuación se establece la analogía y traducción al principio ingenieril, definiendo a su vez el grado de relación entre ambos principios aplicando el aprendizaje adquirido en la definición de la analogía biomimética, experimentado en la EA1. Finalmente se aplica dicho principio ingenieril al diseño conceptual de un objeto tratando de superar los ya existentes.

El objetivo de este ejercicio es familiarizarse con la descripción de una característica natural y el modo en que se puede aplicar técnicamente, buscando la posibilidad de utilización en diferentes objetos por la diversidad de opciones de aplicación. La característica no necesariamente ha de ser la función principal del objeto, además no debe depender del objeto diseñado, por ejemplo si analizamos un mejillón por su característica filtrante no necesariamente debemos diseñar un filtro sino que debemos abstraer la característica de filtro de fluidos para aplicarla a objetos que necesiten de esta función.

En esta EA2 ponemos como ejemplo dos ejercicios, el primero con una documentación muy precisa que llevará a una resolución técnica alta y el segundo con una documentación de menor calidad. En el primer ejercicio se selecciona la “LABIA MINOR” de los Dermaptera, un pequeño artrópodo conocido comúnmente como Tijereta o Cortapicos, su característica destacable es el plegado y articulación de las alas por medio de su fluido sanguíneo, las articulaciones flexibles, la naturaleza flexible del ala, los patrones de plegado y la presencia de la resilina [38], el estudio pormenorizado detalla la importancia del cambio de presiones en el fluido en combinación con el material del ala para evitar la rotura. Este estudio ha llevado a la definición del principio ingenieril y a un pequeño prototipo realizado con láminas plásticas y globos inflables que dan como resultado el diseño conceptual de una bisagra que se puede aplicar en diversos dispositivos como por ejemplo en equipamiento médico para el despliegue de un globo esofágico traqueal por medio de fluidos en vez de gases.

En el segundo ejercicio el ser vivo es el ASTACUS conocida como cangrejo de río y la característica elegida es la articulación estanca con varios grados de libertad y restricción de movimiento, la calidad de la información es inferior y se basa en libros de biología, páginas web con poco carácter científico. El análisis del principio biológico resulta suficiente para definir el principio ingenieril que se basa en la geometría, el cambio de sección y el material. Se

184

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

define a nivel conceptual la aplicación en carcasas articuladas estancas para contener fluidos para uso en cosmética o alimentación. En este caso la falta de precisión en la información de partida impide el realizar una mejor relación entre el principio ingenieril y el biológico.

3.3. EA3. Solución de problema técnico.

Esta experiencia de aprendizaje se realiza por medio de un proyecto en grupos de 4 alumnos en el que se da solución a un problema técnico existente y común a un grupo de productos. Se pretende que la resolución del problema técnico sea extrapolable a productos distintos entre sí. Muchos objetos cotidianos con una característica común están condicionados por las soluciones técnicas existentes, siendo necesario encontrar soluciones más innovadoras.

En la figura 4 se observa la comparativa del modelo genérico de diseño frente al modelo propuesto con aplicación de biomimética. Inicialmente se establece un marco de trabajo que incluye el grupo de objetos que pueden presentar el problema técnico y se definen por medio de una búsqueda de referencias existentes en el mercado que describan el estado actual de la técnica. El tipo de problema técnico se asocia a una característica deseada en el producto que puede ser del tipo estanqueidad, flexibilidad, modularidad, adaptabilidad, etc. En un modelo genérico de diseño se correspondería en cierto modo a la fase del encargo de diseño donde se establecen los requisitos iniciales del proyecto. En esta fase se determina la característica que posteriormente analizaremos por su principio biológico.

El objetivo es establecer el espacio de diseño en el que el diseñador desarrolla su proyecto, a modo de ejemplo podemos determinar que la característica deseada es la ventilación de una carcasa, el marco de trabajo nos limita a objetos o productos que requieran ventilación como por ejemplo máquinas eléctricas, dispositivos electrónicos, contenedores de productos químicos, etc. Cada uno de ellos representa un espacio de diseño, pero estos espacios se pueden ampliar estableciendo un marco de trabajo más

genérico o especializar al definir concretamente un objeto que necesita una innovación vinculada a esta característica; hablaríamos genéricamente de una carcasa que permita el flujo de un fluido o concretamente de un objeto como una máquina herramienta eléctrica de mano que debe disipar calor.

A continuación, se identifican las funciones clave como oportunidades de innovación, se hace un listado de funciones clave, por medio de los mapas mentales. El objetivo es definir una potencial innovación funcional. Los mapas mentales o pensamiento radiante [39] se relacionan con el método de análisis funcional estableciendo niveles y jerarquías, desde funciones genéricas hasta funciones muy específicas y concretas.

Las funciones específicas pueden ser requisitos de diseño concretos, en ellas se ha perdido el carácter de abstracto y su definición técnica debe ser muy alta.

Una vez que las funciones clave están seleccionadas, se realiza un estudio del estado del arte, tanto biológico como industrial. El estado del arte en el ámbito industrial se plantea como búsqueda de objetos o productos que tengan la característica descrita y sean paradigmáticas, utilizando bases de datos de patentes. El objetivo es evitar que se planteen soluciones ya existentes, aportando otras nuevas o alternativas.

Para la investigación en la naturaleza es necesario utilizar la tabla de referentes naturales (ver tabla 2), las funciones se sitúan como objetivos a conseguir y por medio de la técnica creativa “el arte de preguntar” se van perfilando los posibles seres vivos que mejor desarrollen esa función, estas preguntas se orientan hacia el campo biológico [40]. Aprovechando el potencial de la diversidad biológica, y dado que en la naturaleza existen diferentes seres vivos con soluciones diferentes para un mismo problema técnico, se buscan los mejores candidatos para establecer la analogía biomimética. Del mismo modo se consiguen diferentes aplicaciones de las soluciones en objetos distintos por mera adaptación, ya que el propio Tabla 2.

Ejemplo de búsqueda de referentes naturales.

Función Clave

MODELO DE DISEÑO APLICANDO BIOMIMÉTICA

ENCARGO DE PROYECTO

ESTABLECER EL MARCO DE TRABAJO

PLIEGO DE ESPECIFICACIONES DOCUMENTACIÓN E INFORMACIÓN + ANÁLISIS DE PRODUCTO GENERACIÓN DE IDEAS DISEÑO CONCEPTUAL

PROCESO DE DISEÑO EN FASES

MODELO GENÉRICO DE PROCESO DE DISEÑO

TRADUCCIÓN DE SOLUCIONES NATURALES EN FUNCIONES CLAVE DISEÑO CONCEPTUAL

Figura 4. Cuadro comparativo entre modelos de proceso de diseño.

Fuente: Elaboración propia.

¿Para qué?

- Evitar oxidación. ¿Cómo?

- Sacando fluidos - Mediante válvula antirretorno.

- Diferencia de presiones.

- Atmósfera estéril.

- Arañas, mosquitos, Sanguijuelas, caracoles y babosas.

- Utricularia.

- Caracoles y lapas.

- Cráneo

Evitar agresiones

¿Dónde?

- Cualquier entorno. ¿Cómo?

- Contrataque - Mimetización con entorno - Defensa - Sonido - Color (advertencia) ¿Para qué?

- Proteger

- Escarabajo violín, mofeta.

- Camaleón, insecto palo, pulpo.

- Escarabajo tortuga, armadillo, pez globo.

- Babuinos, serpiente de cascabel.

Fuente: Elaboración propia.

185

Referente Natural

Hermetizar cíclico y al vacío.

DEFINICIÓN DE FUNCIONES CLAVE INVESTIGACIÓN BIOLÓGICA

Preguntas Biologizantes

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

marco de trabajo establece la relación entre objetos. El objetivo es conocer una serie de candidatos naturales capaces de aportar innovación funcional.

En la Tabla 2 se presentan dos funciones clave diferentes en la columna izquierda, en la columna central se plantean las “preguntas biologizantes” con las que identificar los posibles candidatos y en la columna derecha se listan una serie de candidatos naturales que de algún modo exhiben la característica perseguida.

Por último se realiza la traducción de las soluciones naturales en funciones clave, empleando la tabla en sentido inverso, aislamos los mejores candidatos para dar solución a las funciones clave, se utilizan los métodos aprendidos en las experiencias de aprendizaje previas para la traducción al principio ingenieril.

Los referentes naturales y la forma en que solucionan cada una de las funciones clave sirven para aplicar el método de definición de especificaciones de diseño utilizado generalizadamente en todos los procesos de conceptualización de producto. El objetivo es conocer la importancia de las especificaciones de diseño y su valor para la definición de conceptos.

En otros modelos se desarrolla el diseño de detalle, sin embargo en este no se contempla esta fase ya que en algunos casos su aplicación es directa pero en otros es necesaria una fase de investigación y desarrollo. Un estudio pormenorizado del individuo natural podría solventarse con una consulta a un especialista o una búsqueda en revistas especializadas, sin embargo el diseñador carece de ese conocimiento y el proyecto necesita una fase específica.

También puede ser necesario el desarrollo de prototipos, experimentos y pruebas que evidencien la correcta traducción del principio biológico en el ingenieril. Por medio de modelos funcionales y virtuales generados con herramientas CAD, módulos de cálculo, simulación o animación se integran los métodos de representación al diseño biomimético en la fase final del proceso.

Es en esta última experiencia de aprendizaje donde se observa claramente la integración de métodos, por la combinación de fases analíticas de investigación técnica y biológica, creativas, y de definición de producto o prototipado.

4. Resultados 4.1. EA 1. Estudio de un caso de biomimética.

En experiencias previas [30] se evidencio la dificultad de comprender el principio biológico y hacer su traducción a un principio ingenieril, se propuso hacer la EA1 para aprender y conocer esa fase y poder hacerla por sí mismos.

Se realizaron 41 ejercicios individuales, en la Tabla 3 se observa un resumen de los resultados de la EA1. Los resultados se evalúan Primero por la idoneidad del caso seleccionado y por la cantidad y calidad de la información para describir la analogía biomimética. Todos se consideran idóneos.

Segundo, por el análisis del caso y la definición del grado de relación entre principio biológico e ingenieril (GR1), si la relación entre el principio biológico y el

ingenieril se trata de una mera inspiración se considera muy bajo, si se aproxima a una copia entonces es muy alto, siendo estos los extremos de una escala de 5 niveles (ver Fig. 3). Ninguno se considera muy bajo, 3 se consideran bajos y 3 normales. En 15 ejercicios se considera alto y en 20 muy alto, lo que indica que los casos seleccionados explican detalladamente el principio biológico y su aplicación técnica.

Tercero, por el grado de relación de su propia propuesta (GR2), que utiliza la misma escala. El GR2 indica si el diseñador ha comprendido el GR1 y lo ha sabido aprovechar. Ninguno se considera muy bajo y 3 bajos, una parte muy pequeña de la muestra. Hay 11 ejercicios con un nivel normal, 13 alto y 14 muy alto, lo que indica que la información de los casos es válida y se han aplicado correctamente.

Además se evalúa de manera individual para los 41 ejercicios si el caso original se iguala o se supera. Se compara el GR2 respecto al GR1, se considera superado si el GR2 supera en la escala al GR1 (diferencia positiva, +2 o +1), se iguala cuando no cambia (diferencia igual a cero) y empeora si no lo supera (diferencia negativa. -1 o -2).

Ningún caso de los propuestos por los alumnos supera ampliamente al original, solo 3 lo superan, sin embargo se observa que la mayoría (26 ejercicios) mantiene el nivel, siendo en 12 ocasiones empeorado. Esto indica que en general han comprendido el caso y han aprendido a hacer su propia propuesta, no superando el caso cuando el nivel de la analogía y definición técnica es muy alto pero consiguiéndolo cuando es alto o normal.

4.2. EA 2. Aplicación de una solución de la naturaleza.

El aprendizaje en la EA1 permite al diseñador estar familiarizado con la relación entre el principio biológico e ingenieril y en la EA2 debe aplicarlo al saber identificar un ser vivo con una característica destacable, su comprensión y aplicación técnica. Se han realizado un total de 22 ejercicios en parejas, se encontraron 11 repeticiones en 5 seres vivos comunes, y los resultados se resumen en la Tabla 4.

Primero se evalúa la definición del principio biológico, si han sabido encontrar la información y analizarla. Se evalúa como válido (V) cuando la definición del principio biológico es útil para hacer una analogía biomimética, y no válido (X) cuando no se puede utilizar. Se han clasificado 18 ejercicios como válidos y tan solo 1 como no válido.

Existen casos imprecisos que resulta difícil calificarlos como válido o no válido y que evaluaremos como “indefinido” (I), de estos se han observado 3 ejercicios.

Segundo, se evalúa la definición del principio ingenieril, si utilizando la definición realizada en el principio biológico Tabla 3.

Resumen de resultados de la EA1.

2 BAJO

GR1

1 MUY BAJO 0

3

3

15

5 MUY ALTO 20

GR2

0

3

11

13

14

COMPARACIÓN GR2-GR1

-2 5

-1 7

0 26

+1 3

+2 0

ANALOGÍA BIOMIMÉTICA

Fuente: Elaboración propia.

186

3 NORMAL

4 ALTO

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

han sabido hacer una traducción técnica. Se utiliza el mismo criterio de válido o no válido anterior. En 15 ejercicios se considera como válido, 2 como no válido y se observan 5 calificables como indefinidos.

En general se observa que la definición del principio biológico condiciona la definición del principio ingenieril (17 ejercicios), una buena o mala definición lleva a una buena o mala traducción respectivamente, es decir, se mantiene (=). Y en 5 ejercicios se observa que la traducción es errónea (E), no se aprovecha la definición del principio biológico y lleva a un principio ingenieril no válido, es decir, empeora.

Individualmente se observa que normalmente se mantiene, los que tienen bien definido el principio biológico también lo hacen con el ingenieril, 14 casos (V-V), 2 casos (I-I) y 1 caso (X-X). Y que ninguno puede hacer una mejora es decir si el principio biológico está mal definido o indefinido no es posible tener bien definido el ingenieril (IX, V-I).

Y tercero, se evalúa el grado de relación entre el principio biológico y el ingenieril, con la misma escala de niveles utilizada en la EA1. Solo 2 ejercicios se consideran con un nivel bajo y ninguno con muy bajo. En 6 ejercicios se obtiene un nivel normal, en 11 alto y en 3 muy alto. Los resultados son satisfactorios, 14 de los 22 ejercicios obtienen un buen nivel de definición de los principios biológico e ingenieril y su analogía bomimética es correcta y útil para generar conceptos de diseño.

4.3. EA 3. Solución de problema técnico.

Se han desarrollado 13 proyectos en grupos de 3 o 4 diseñadores, la Tabla 5 recoge un resumen de los resultados.

La elección del marco es válida si la característica técnica se puede implementar en esos productos. Se proponen 3 marcos de trabajo, cada grupo elige el apropiado para su proyecto, 6 grupos han elegido el marco de alimentación, 3 el de electrónica y 3 de sumergible. Uno de los grupos no ha definido un marco sino que ha mezclado varios de ellos.

Todos ellos se consideran válidos.

Primero se evalúa el número de funciones clave generadas, los mapas mentales [39] son una herramienta eficaz y eficiente en la búsqueda de funciones potencialmente innovadoras, aunque no exclusiva. Se han encontrado 89 en los diferentes proyectos, 9 se clasifican con un alto grado de oportunidad de generar innovación, 47 grado medio y 33 grado bajo. Se clasifican con un grado Tabla 4.

Resumen de resultados de la EA2.

DEFINICIÓN

X

I

V

PB

1

3

18

PI

2

5

15

COMPARACIÓN

X-X

I-X

I-I

INDIVIDUAL

1

1

1 MUY BAJO

2 BAJO

0

2

6

ANALOGÍA BIOMIMÉTICA GR PB-PI

=

E

17

5

V-I

V-V

2

4

14

3 NORMAL

4 ALTO

5 MUY ALTO

11

3

(X) no válido / (I) indefinido / (V) válido / (=) se mantiene / (E) empeora

Fuente: Elaboración propia.

Tabla 5. Resumen de resultados de la EA3.

MARCO

NO VALIDO

GRUPOS

VALIDO

0

13

GRADO DE OPORTUNIDAD

TOTAL

BAJO

MEDIO

ALTO

FUNCION CLAVE

89

33

47

9

ANALOGÍA BIOMIMETICA

1 MUY BAJO

2 BAJO

3 NORMAL

4 ALTO

5 MUY ALTO

GR PB-PI

0

2

3

5

3

EDP CONCEPTO

INCOMPLETO

COMPLETO

GRUPOS

5

8

Fuente: Elaboración propia.

bajo aquellas funciones ya solucionadas, y para las cuales la naturaleza difícilmente aporta valor o diferenciación. Un grado medio se considera a funciones que existen pero se pueden diferenciar y aportar valor. Por último, un grado alto implica que no existen, o que existen en muy pocas aplicaciones, y sería interesante encontrar seres vivos que permitieran desarrollarlas.

Algunos grupos no utilizan los mapas mentales, se limitan a hacer listados organizados de funciones. Los listados de funciones no pueden establecer vínculos entre niveles, sin embargo demuestran ser válidos, se basan en la técnica de listado de atributos [41], propia de la generación de nuevos productos y en la mejora de servicios o productos existentes.

Las funciones clave agrupadas por temas genéricos, como por ejemplo barrera térmica, regulador de humedad y temperatura, resistencia al choque o evitar agresiones, tienen un grado de oportunidad bajo y parece difícil que superen el estado actual de la técnica.

Sin embargo las funciones específicas se clasifican con un grado de oportunidad medio o alto, por lo que dar solución a estas funciones podría tener aplicaciones de interés, por ejemplo una barrera selectiva unidireccional/bidireccional, integridad del hermético dependiente de las condiciones del entorno inspirada en bivalvos como la almeja o evitar contaminación al cocinar alimentos inspirado en la creación de un biofilm como hacen algunas algas con características antibacterianas.

En segundo lugar se evalúa la obtención de referentes naturales por medio de las tablas biomiméticas, método para encontrar referentes naturales a las funciones clave, partiendo de la función clave permite concretar soluciones o seres vivos que desarrollan esa función (ver Tabla 2).

Se ha observado que no todos los grupos las aplican estrictamente, omiten alguna parte, plantean una alternativa o bien obtienen el mismo resultado sin necesidad de ponerlo en un formato de tabla. Hay un total de 328 propuestas, algunas repetidas en varios grupos, el mayor número es 50 y el menor 9, salvando una excepción que no propone tabla biomimética.

El elevado número de seres vivos propuestos para desarrollar el principio biológico evidencia la utilidad de las tablas y de la EA2 en cuanto a la búsqueda de candidatos.

187

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

Además se tiene en cuenta la calidad de las propuestas, ya que son necesarias para definir el principio biológico.

Tercero, como en las experiencias de aprendizaje anteriores se evalúa el grado de relación entre principio biológico e ingenieril, con la misma clasificación. Es de destacar que 3 proyectos se clasifican con un grado muy alto, evidenciando que el principio natural está bien definido, con suficiente información y de calidad, y además la relación es clara y fundamentada de modo que el principio ingenieril está satisfactoriamente definido y se puede aplicar al diseño. Con un grado alto hay 5 proyectos, 3 con medio y 2 con bajo, es decir, 8 de 11 se consideran como alto o muy alto para grados de relación definidos por los alumnos, demostrando que el aprendizaje anterior es válido.

En estos proyectos además se valora la propuesta conceptual por la calidad de las especificaciones de diseño que se utilizan para describir el concepto en relación a la innovación pretendida. Se consideran solo dos opciones, que la especificación de diseño de producto está completa, define el concepto y utiliza la función clave (8 casos) o que está incompleta, el concepto no queda claro y bien definido o no utiliza la función clave (5 casos).

Los resultados de las propuestas conceptuales son óptimos, los 8 proyectos considerados con una descripción completa tienen una buena definición conceptual, con un listado de especificaciones vinculadas a los resultados del trabajo realizado con las tablas biomiméticas.

La calidad de las propuestas responde a las experiencias previas y ese es un indicador de los buenos resultados, todos los equipos en las experiencias de aprendizaje anteriores han definido un objeto o producto a nivel conceptual, planteando requisitos de diseño y necesidades a satisfacer por medio del principio ingenieril. Todos los equipos presentan varias opciones conceptuales y eligen una de ellas para desarrollar.

En la propuesta de proyecto no se define un objeto, se solicita resolver el problema técnico o característica de producto por lo que el objeto a diseñar puede ser con propuestas muy abiertas, por esta razón no se precisa coherencia entre los resultados de los diferentes proyectos y lo que realmente se analiza para evaluar es como se definen los requisitos funcionales partiendo del principio ingenieril.

4.4. Resultados globales.

La aplicación satisfactoria de este modelo ha servido para que en el último proyecto los diseñadores hayan sido capaces de ir aplicando los diferentes métodos aprendidos de manera parcial en las EA1 y EA2 y relacionándolos con otros métodos de diseño.

Para la validación del modelo se realizó una encuesta con cuatro bloques de preguntas, que se corresponden con las características de comprensión, aplicabilidad, beneficios y utilidad. En la Tabla 6 se presentan las preguntas con los valores medios de las respuestas de cada grupo, M-DI para el grupo de diseñadores y M-U para el grupo de Universa, además del total del grupo (M) junto con la desviación típica (σ). Las preguntas se formulan en positivo para obtener respuestas numéricas, donde 1 significa muy

negativo (muy en desacuerdo) y 5 muy positivo (totalmente de acuerdo).

Los resultados de la encuesta han permitido realizar un pequeño estudio estadístico (ver Tabla 6), y otro cualitativo gracias a preguntas abiertas para cada bloque, donde escribir observaciones o comentarios, además existe un bloque final en el que se solicita al alumno una evaluación personal que refleje el valor que ha tenido la utilización de la metodología y sus conclusiones.

El resultado general refleja que el modelo es aplicable por diseñadores y personas de otros campos de conocimiento, fácil de comprender y permite alcanzar los objetivos definidos. No se declaran dificultades de aplicabilidad y se reconoce que se emplearía en futuros proyectos teniendo la debida información y conocimiento sobre los requisitos del principio ingenieril. Se destaca como una metodología beneficiosa para el proceso creativo e innovador, y que permite conocer y aplicar el método por partes y de manera global. Los propios diseñadores evalúan positivamente la práctica con este modelo de diseño conceptual.

Además se reconoce como útil en la definición de objetivos y nuevos proyectos, en el establecimiento de ideas y conceptualización así como apoyo a otros métodos en las diferentes fases de diseño. Estos resultados se observan en la columna de valores medios, destacados en gris de la Tabla 6.

5. Conclusiones La innovación hoy en día exige nuevos procesos creativos y métodos alternativos en diseño industrial. Aquí se ha presentado un modelo de aprendizaje donde cada experiencia de aprendizaje afianza conocimientos útiles aplicables, el diseñador utiliza la naturaleza como fuente de ideas y soluciones técnicas. La resolución de ejercicios de dificultad incremental simplifica la resolución de un problema técnico, obteniendo resultados innovadores al integrar la biomimética.

Se han descrito relaciones entre análisis funcional y creatividad, el proceso creativo establece diferentes niveles en los que encontrar nuevas funciones. Se ha planteado un marco de trabajo en el que la relación entre naturaleza y diseño da buenos resultados.

La calidad de la información dada por el detalle del principio biológico e ingenieril permite al diseñador hacer nuevas y válidas propuestas de la utilización de un principio biológico.

El modelo presentado aporta una visión nueva respecto a la resolución funciones distintas de la principal, así el diseño de producto se beneficia de las soluciones a funciones complementarias. Muchos productos son eficaces por su función principal, pero la diferenciación e innovación se puede encontrar en funciones complementarias o en nuevas formas de conseguirlas.

Si bien existen trabajos de investigación y métodos basados en la explotación de soluciones de la naturaleza sus campos de actividad no se centran en la definición de nuevas funciones, por ello la exploración de soluciones naturales para la resolución de problemas técnicos de carácter funcional se convierte en un rasgo de innovación.

188

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

COMPRENSIÓN

Media Media Media DI UNIV

Sobre las metodologías expuestas en la parte teórica y aplicadas en la parte práctica:

En general se entienden con claridad Se entienden todas y cada una de las fases del proceso Marcan un proceso fácil de seguir

BENEFICIOS

APLICACIÓN

Tienen unos objetivos bien definidos y alcanzables Existen dificultades de aplicación en el proceso de diseño Los ejemplos facilitan la aplicación

Sobre la aplicación de las metodologías de biomimética:

Aplicar metodologías de biomimética…

Son necesarios conocimiento s previos Son necesarios requisitos iniciales de información Es necesario Conocer los requisitos del principio ingenieril Volvería a aplicarlas en otros proyectos Permite generar mayor número de ideas y conceptos Permite generar ideas novedosas e innovadoras Ayuda al diseñador a establecer objetivos de proyecto Ayuda al diseñador en el proceso creativo Ayuda al diseñador en la fase conceptual

4,07

4,13

4,09

4,31

4,00

3,55

Desv T 0,69

0,86

3,89

3,89

3,91

0,80

3,84

3,88

3,73

0,93

3,82

3,86

3,70

0,72

4,65

4,66

4,64

0,57

3,09

3,15

2,91

1,18

3,59

3,60

3,55

1,05

3,80

4,00

3,18

0,84

4,39

4,40

4,36

0,74

4,52

4,49

4,64

0,69

Ayuda al diseñador en la fase de diseño de detalle La definición de objetivos y nuevos proyectos La fase de definición de ideas y conceptualiza ción La fase de desarrollo, definición de materiales y procesos La fase de detalle y definición final del producto La metodología La estética y biomimética es caracterizaci útil para… ón formal Aplicar soluciones naturales (EA2) al diseño de producto La solución de problemas técnicos (EA3) Aplicarla junto con otros métodos y herramientas (p.ej. El análisis funcional y otros) Fuente: Elaboración propia.

VALIDACIÓN

Tabla 6.

Resumen estadístico encuesta.

4,80

4,55

0,49

[2] [3]

3,91

3,94

3,82

0,78 [4]

4,48

4,49

4,45

0,55

[5] [6]

4,33

4,46

3,91

3,77

3,55

0,98

3,89

3,86

4,00

0,80

4,24

4,29

4,09

0,79

3,70

3,69

3,73

0,92

3,48

3,54

3,27

1,07

3,67

3,71

3,55

0,97

4,46

4,54

4,18

0,55

4,33

4,31

4,36

0,79

4,41

4,43

4,36

0,80

References [1]

4,74

3,72

0,73 [7]

189

Drachsler, K., 2012. Bionik – Mit Einer Neuen Systematik Schneller Zu Innovationen, [Online], [date of reference February 20th of 2012], Available at: : http://w3.ipa.fhg.de/PresseMedien/interaktiv/interaktiv_2003_01.pdf Viñolas, I. y Marlet, J., Diseño ecológico: Hacia un diseño y una producción en armonía con la naturaleza. Barcelona: Blume, 2005.

Vincent, J.F.V., Bogatyreva, O.A., Bogatyrev, N.R., Bowyer, A. and Pahl, A.K., Biomimetics: Its practice and theory. Journal of the Royal Society Interface, 3 (9), pp. 471-482, 2006. http://dx.doi.org/10.1098/rsif.2006.0127 Vogel, S., Comparative biomechanics: Life's physical world.

Princeton: Princeton University Press, 2003.

Alcocer, W., Vela, L., Blanco, A., Gonzalez, J. and Oliver, M., Major trends in the development of ankle rehabilitation devices, DYNA, 79 (176), pp. 45-55, 2012.

Bhushan, B., Biomimetics: Lessons from Nature - an Overview.

Philosophical Transactions, Series A, Mathematical, Physical, and Engineering Sciences, 367 (1893), pp. 1445-86, 2009.

Asdrúbal, G., Del editor. El diseño y modelamiento de materiales.

DYNA, 75 (156), pp. 251-269, 2008

López-Forniés & Berges-Muro / DYNA 81 (188), pp. 181-190. December, 2014.

[8] [9]

[10] [11] [12] [13] [14] [15]

[16] [17] [18] [19]

[20]

[21]

[22] [23] [24]

[25] [26] [27] [28]

[29]

Bar-Cohen, Y., Biomimetics: Biologically inspired technologies.

Boca Raton: CRC/Taylor & Francis, 2006. http://dx.doi.org/10.1088/1748-3182/1/1/P01 Shu, L.H., A natural-language approach to biomimetic design.

Artificial intelligence for engineering design, Analysis and Manufacturing.

24 (4), pp.

507-519.

2010. http://dx.doi.org/10.1017/S0890060410000363 Mak, T.W. and Shu, L.H., Abstraction of biological analogies for design. Cirp Annals-Manufacturing Technology, 53 (1), pp.117-120.

2004. http://dx.doi.org/10.1016/S0007-8506(07)60658-1 Pedersen, Z., Biomimetic approaches to architectural design for increased sustainability. Sustainable Building Conference, 2007.

Roshko, T., The pedagogy of bio-design: Methodology development. WIT Transactions on Ecology and the Environment.

138, pp. 545-558, 2010. http://dx.doi.org/10.2495/DN100491 Helms, M., Vattam, S. and Goel, A., Biologically inspired design: process and products. Design Studies. 30 (5). pp. 606-622, 2009. http://dx.doi.org/10.1016/j.destud.2009.04.003 Speck, T. and Speck, O., Process sequences in biomimetic research.

WIT Transactions on Ecology and the Environment. 114, pp. 3-11.

2008. http://dx.doi.org/10.2495/DN080011 Bogatyreva, O.A., Pahl, A.K. and Vincent, J.F., Enriching TRIZ with biology: The biological effects database and implications for teleology and epistemology. ETRIA World Conference-2002, Strasbourg, pp. 3-1-307. 2002.

Manchado-Pérez, E. and Berges-Muro, L., Sistemas de retículas: Un método para diseñar nuevos conceptos de producto hacia el usuario.

DYNA, 80 (181), pp. 16-24, 2013.

Goel, A.K. and Bhatta ,S.R., Use of design patterns in analogy-based design. Advanced Engineering Informatics. 18 (2) pp. 85-94. 2004. http://dx.doi.org/10.1016/j.aei.2004.09.003 Gero, J., The situated function-behaviour-structure framework.

Design Studies.

25 (4) pp.

373-391, 2004. http://dx.doi.org/10.1016/j.destud.2003.10.010 Deng, Y.M., Britton, G.A. and Tor, S.B., Constraint-based functional design verification for conceptual design. ComputerAided Design.

32 (14) pp.

889-899.

2000. http://dx.doi.org/10.1016/S0010-4485(00)00077-4 Yasushi, U., Ishii, M., Yoshioka, M., Shimomura, Y. and Tomiyama, T., Supporting conceptual design based on the functionbehavior-state modeler. Artificial intelligence for engineering design, Analysis and Manufacturing: Ai Edam. 10 (4) pp. 275. 1996.

Mills, J., Engineering education – Is problem based or project-based learning the answer? Australasian J. of Engineering Education, [Online], [date of reference December 20th of 2003], Available: http://www.aaee.com.au/journal/2003/mills_treagust03.pdf Anzai, Y., The theory of learning by doing. Carnegie-Mellon University, Pittsburgh, 1978.

Pedersen, K., et al. Validating design methods and research: The validation square, DECTC'00, 2000 ASME Design Engineering Technical Conferences, Baltimore, MA, 2000.

López-Forniés, I., Modelo metodológico de diseño conceptual con enfoque biomimético, PhD. Thesis Dissertation, Design and Manufacturing Department, Universidad de Zaragoza, Zaragoza, España, 2012.

Tomiyama, T, Gu, P., Jin, Y., Lutters, D., Kind, Ch. and Kimura, F., Design methodologies: Industrial and educational applications. Cirp Annals - Manufacturing Technology, 58 (2) pp. 543-565. 2009.

Cross, N., Engineering design methods: Strategies for product design. Chichester: Wiley, 2000.

Pahl, G, and Wolfgang B., Engineering design: A systematic approach. London: Springer, 1996. http://dx.doi.org/10.1007/978-14471-3581-4 Design Council. Introducing design methods, [Online], [date of reference, July 25th of 2014], Available at: http://www.designcouncil.org.uk/news-opinion/introducing-designmethods Santulli, C. and Langella, C., Introducing students to bio-inspiration and biomimetic design: A workshop experience, International Journal of Technology and Design Education, 21 (4), pp 471-485, 2011. http://dx.doi.org/10.1007/s10798-010-9132-6

[30] López-Forniés, I., Berges-Muro, L., Relation between biomimetic and functional analysis in product design methodology. WIT Transactions on Ecology and the Environment, 138, pp 317-328, 2010. http://dx.doi.org/10.2495/DN100271 [31] Biomimicry Institute. Educating. [Online], [date of reference, July 25th of 2014]. Available at: http://www.biomimicry.net/educating/ [32] Marín, R. y De La Torre, S., Manual de creatividad. Barcelona: Vicens Vives, 2000.

[33] Ask Nature. Smart Bandage. [Online], [date of reference, July 25th of 2014]].

Available at: http://www.asknature.org/product/84b05783b105ac6f511cb22758a9 4d13 [34] Vincent, J., Stealing ideas from nature, RSA Journal London, pp.

36-43, 1997 [35] Reyes, F., Nature: Inspiration for Art & Design, Barcelona: Monsa., 2008.

[36] Takemura, S-Y., Stavenga, D.G. and Arikawa, K., Absence of eye shine and tapetum in the heterogeneous eye of Anthocharis butterflies (Pieridae), The Journal of Experimental Biolog,. 210, pp.3075-3081. 2007. http://dx.doi.org/10.1242/jeb.002725 [37] Jinkuk, K., Park, J.H., Hong, J., Choi, S.J., Kang, G.H., Yu, G.J., Kim, N.S. and Song, H., Double antireflection coating layer with silicon nitride and silicon oxide for crystalline silicon solar cell, Journal of Electroceramics, 30, pp.

41-45, 2013. http://dx.doi.org/10.1007/s10832-012-9710-y [38] Haas, F., Elastic joints in dermapteran hind wings: Materials and wing folding. Arthropod Structure & Development. 29 (2) pp. 137146. 2000. http://dx.doi.org/10.1016/S1467-8039(00)00025-6 [39] Buzan, T., The mind map book. London:BBC Books, 1993.

[40] Gruber, P., Skin in architecture: Towards bioinspired facades, WIT Transactions on Ecology and the Environment, 138, pp 503-513, 2010. http://dx.doi.org/10.2495/DN100451 [41] Crawford, R., The techniques of creative thinking: How to use your ideas to achieve success, Fraser Pub. Co, Wells, 1964.

I. López-Forniés es Dr por la Universidad de Zaragoza. Ingeniero en Organización Industrial, Bachelor Arts in Consumer Product Design por la Coventry University e Ingeniero Técnico Industrial. Actualmente Profesor Colaborador e investigador perteneciente al I3A de la Universidad de Zaragoza, España. Socio fundador del estudio de diseño industrial y gráfico Mil Asociados, Presidente y fundador de la Asociación de Profesionales de Diseño Industrial de Aragón (DIN-A). Coordinador, responsable y ponente de múltiples cursos de diseño industrial en diversos organismos.

L. Berges-Muro es Dr Ing. Industrial. Profesor Titular de Ingeniería de los Procesos de Fabricación de la Universidad de Zaragoza, España, ha desempeñado numerosos cargos como Vicerrector de Infraestructuras y Servicios Universitarios, Director de la Oficina de Transferencia de Resultados de Investigación. Director del Departamento de Ingeniería de Diseño y Fabricación de la Universidad de Zaragoza, España. Miembro de la Junta de Gobierno del Colegio Oficial de Ingenieros Industriales de Aragón y La Rioja.

190

Structural control using magnetorheological dampers governed by predictive and dynamic inverse models Luis Augusto Lara-Valencia a, José Luis Vital-de Brito b & Yamile Valencia-Gonzalez c a

Universidad Nacional de Colombia, Medellín, Colombia. lualarava@unal.edu.co b University of Brasilia, Brasilia, Brazil. jlbrito@unb.br c Universidad Nacional de Colombia, Medellín, Colombia. yvalenc0@unal.edu.co

Received: January 27th, 2014. Received in revised form: July 2th, 2014. Accepted: July 24th, 2014.

Abstract The present paper implements a novelty semi-active structural control design on a two-story building, with the aim of reducing vibrations caused by transient type loads. The analyzed structure corresponds to an experimental prototype that was fully characterized and modeled according to the diaphragm hypothesis. The controller used was based on the action of a pair of real magnetorheological (MR) dampers whose operation is emulated by the phenomenological model. These mechanisms are governed by a numerical system that is based on non-linear autoregressive model with exogenous inputs (NARX)-type artificial neural networks, which have the ability to determine the necessary optimal control forces and the voltages required for the development of these forces through a prediction model and an inverse model, which are pioneers in this kind of systems. The results obtained show that the control design based on neural networks that was developed in the present study is a reliable and efficient, achieving reductions of up to 69% for the peak response value.

Keywords: Dynamics of structures, semi-active control of structures, inverse models, predictive models, neural networks, magnetorheological dampers.

Control estructural utilizando amortiguadores magnetoreológicos gobernados por un modelo predictivo y por un modelo inverso dinámico Resumen En este artículo se implementa um novedoso proyecto de control estructural numérico en una edificación de dos pisos con el objetivo de reducir vibraciones debidas a cargas de tipo transiente. La estructura analizada corresponde a un prototipo experimental debidamente caracterizado y modelado de acuerdo con la hipótesis del diafragma. El controlador utilizado se basa en la acción de un par de amortiguadores magnetoreológicos (MR) reales cuyo funcionamiento es emulado a través del denominado modelo fenomenológico. Los disipadores son gobernados por un sistema numérico basado en redes neuronales artificiales del tipo NARX con la capacidad de determinar fuerzas óptimas de control y voltajes a través de un modelo de predicción y un modelo inverso, los cuales son de uso inédito en este tipo de sistemas. Los resultados obtenidos muestran que el proyecto de control basado en redes neuronales desarrollado en este trabajo es un controlador confiable y eficiente, consiguiendo reducciones de hasta 69% en los valores pico de respuesta.

Palabras Clave: Dinámica de estructuras, control semi-activo de estructuras, modelos inversos, modelos predictivos, redes neuronales, amortiguadores magnetoreológicos.

1. Introduction The proposed control algorithm calculates the optimal control force required by the MR dampers to reduce the movement of the protected structure. However, the algorithm must also determine the voltage required for the

controller because the increases or decreases in the forces produced by the damper are indirectly controlled by the voltages applied to the device.

In this work, the capacity and efficiency of the control design that has been proposed for a building was evaluated.

Thus, a numerical model was built for a 2-story gantry,

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 191-198. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41774

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014.

where 2 MR dampers were installed and controlled by the developed control algorithm. The structure was subjected to acceleration at the base, and the response values of the system both with and without control were calculated to evaluate the operation of the control strategy presented.

2. Studied Model The model studied in the present paper consisted of a 2story gantry, analyzed in 3 dimensions, with 3 degrees of freedom per floor (horizontal displacements in the X and Y axes and rotation around the Z axis). In addition, the model considered the use of a pair of MR dampers installed at the height of the first floor of the building that control the system. The model was a 2:3 scale experimental prototype built at the Laboratory of the Department of Structures, Geotechnics and Applied Geology of the University of Basilicata in Italy that was used in a joint research project between the Italian Seismic Engineering University Laboratories Network (Rede de Laboratórios Universitários Italianos de Engenharia Sísmica - ReLUIS) and the Italian Civil Protection Department (Departamento de Proteção Civil Italiano - DPC).

2.1. Parameters and properties of the building

Table 1.

Properties of the RD-1005-3 MR damper.

Damper properties Compressed length (mm) Extended length (mm) Body diameter (mm) Maximum extension force (N) Maximum operating temperature (°C) Maximum input current (A)

Values 155 208 41.4 4448 71 Continuous = 1 Intermittent = 2 Input voltage (V) 12 DC Electrical resistance at room temperature (Ohms) 5 Response time (ms) < 15 Source: Lord Corporation, Lord technical data RD-1005-3 Damper, 2006.

Table 2.

Constant parameters of the RD-1005-3 MR damper A β γ k1(x-x0) (N) (m) (m) (m) 20000 10000 10000 60

k0 (N/m) 2020

n 2

Source: Taken from Basili, M., 2006.

The mass, stiffness and damping matrix of the analyzed building can be consulted in [2] 2.2. Parameters and properties of the MR dampers

The 2-story gantry was 2 m tall between the floors, and the building was a rectangle with a distance of 3 m between the axes of the pillars in the Y direction and 4 m in the X direction. The gantry modeling was conducted in 3 dimensions, adopting the diaphragm hypothesis that assumes that each slab is rigid in its own plane and flexible in the perpendicular direction. It was also established that the horizontal displacements of all floor nodes were related to 3 rigid body displacements that were defined in the center of mass of each floor, i: uxi translations in the x direction, uyi in the y direction, and uθi torsion rotation around the z vertical axis. Fig. 1 shows a photograph of the actual model, located at the University of Basilicata in Italy, taken by [1].

The devices used to control the structure were a pair of compact RD-1005-3 MR dampers, manufactured by the Lord Corporation in Cary, NC, USA. To numerically simulate the behavior of these devices, the phenomenological model proposed in [3], was used.

Table 1 shows the primary properties of RD-1005-3 MR dampers, according to the technical specifications published by the manufacturer [4].

In [5] the parameters that characterize the behavior of the RD-1005-3 MR damper are identified. Was found that some of these parameters remained constant under varying operating conditions; thus, for example, fixed values of k0, n, and k1(x-x0) were defined based on tests seeking to determine the mechanical characteristics of the damper, while others, such as the A, β, and γ values, were constant values suggested in the literature [3]. The damper parameters that were assumed to be constant values are listed in Table 2.

However, was identified that the parameters α, c0, and c1 of the RD-1005-3 MR damper to be voltage-dependent parameters [5]. The equations that describe these relationships are the following: ଶ

ଶ

଴

ଵ

Figure 1 - Experimental frame located in the University of Basilicata Source: Carneiro, R., 2009.

192

ଶ

(1)

(2)

(3)

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014.

3. Controller based on a predictive model and an inverse dynamic model developed through NARX-Type artificial neural networks

Input S(n)

x’(n) x’’(n) v(n)

The primary purpose of the control algorithm based on ANNs herein presented is to define a model capable of calculating the optimal control force to be applied by the energy dissipation mechanism such that it reduces the movement of the protected structure as much as possible.

However, the control design must also focus on determining the voltage to be applied to the controller because the increase or decrease of forces produced by MR dampers is indirectly controlled through the voltage applied to the device. To determine these two fundamental parameters, the optimal force and voltage, two properly trained NARX networks were used. The first network simulates a prediction model tasked with determining the optimal control force required for the MR to minimize the vibrations of the structure in the most efficient manner possible when it is subjected to external forces on its base. The second network works as an inverse model, i.e., the network determines the input of the control design based on the delayed outputs of the system. Thus, the second network is occupied with defining the proper voltage to be applied to the control device such that it will apply a force to the structure close to the optimal force, which was calculated by the first neural network.

Fig. 2 presents a diagram of the control based on the ANN developed to reduce the vibrations of the analyzed structure.

3.1. Prediction model of the optimal control force The proposed prediction model of the optimal force is formed by a NARX-type neural network that is completely interconnected and configured with a layer of sensory units composed of fifteen input signals and one bias term, a computational processing layer consisting of sixteen hidden neurons, and a layer of results formed by a single output. A diagram of the network used in the prediction model of the force is shown in Fig. 3.

Figure 2. Controller design based on artificial neural networks.

Source: The Authors

x(n) z-1

x(n-1)

S(n-1)

x’(n-1) z-1

x’’(n-1) v(n-1) x(n-2)

S(n-2)

f(n+1) z-1

Output f(n)

x’(n-2) x’’(n-2)

Bias

v(n-2)

f(n-2) z-1

f(n-1) z-1

f(n)

Figure 3. Neural network used in the predictive model.

Source: The Authors

The selection of the number of layers and components per layer for each of the neural networks developed does not follow a specific procedure; it varies from application to application and is essentially a trial-and-error exercise. In general, the use of a hidden layer is adequate to model highly complex functional dependencies. This ability was demonstrated in our early experiences, where we attempted to reproduce the operation of the proposed NARX networks (predictive and inverse model). When the networks were modeled with two or more hidden layers, the results obtained from the processors were virtually identical to the results achieved by the networks whose computational units were distributed in a single hidden layer. Moreover, the computational efficiency of the networks that had a hidden layer was remarkably greater, which supports the choice of the number of network layers necessary for optimal performance.

The choice of the number of neurons in the hidden layers was decided through a survey process. NARX networks were analyzed with a hidden layer and various amounts of neurons (ranging from 8 to 30 neurons). Each of these networks was evaluated according to two specific parameters, processing time and performance (measured from the mean square error of the training process), ranging from 0 to 1. The value of 1 was given to the network with the best performance among all the networks, and the value 0 was represented the network with the poorest performance within the analyzed group of networks. The networks in between the extremes received a weighting between these two values based on the estimation of the equivalent percentage of the evaluated parameters compared with the best performance parameter values. The result of the process showed that the ideal number of neurons for a hidden layer in the NARX networks should be equal to 16.

Finally, the inputs of the neural networks were determined based on the work of He and Asada [6]. In this work, it was shown that a second-order input model was adequate to identify the characteristics of an MR damper, and based on this finding, it was decided to use delays in the inputs of the processors of one and two units of time, as shown in Fig. 3.

The hyperbolic tangent sigmoid function and piecewise linear function were used for activating the neurons in the

193

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014. x(n) x’(n) x’’(n)

z-1 f(n)

x(n-1) x’(n-1)

S(n-1) z-1

x’’(n-1) f(n-1) x(n-2)

S(n-2)

v(n+1) z-1

Output v(n)

x’(n-2) x’’(n-2)

Bias

f(n-2)

v(n-2) z-1 v(n-1) z-1 v(n)

Figure 4. Neural network used in the inverse model.

Source: The Authors

hidden and ouput layers respectively; the LevenbergMarquadt algorithm was used for training the synaptic conections of the artificial neural network.

3.2. Inverse model applied to determine the voltage of the control device The inverse model proposed to determine the voltage to be applied to the MR damper consists of a fully interconnected NARX network. Similarly to the prediction model, the network is configured with an input layer composed of fifteen input signals, a hidden layer with sixteen neurons, and a output layer with a single output.

This neural network uses the hyperbolic tangent sigmoid function and piecewise linear function for activating the neurons in the hidden and ouput layers respectively.

The input layer of the neural network that composes the inverse model is formed by the displacement, speed, and acceleration values of the structure, the optimal control force values calculated in the prediction model, and the feedback inherent to the recurrent network with the output value (voltage). The choice of order of the network delay lines was again based on the results obtained by [6]. Fig. 4 shows the neural network model used in the inverse model.

validation of the neural network designed for the predictive model was generated by 2 normally distributed series of random numbers for 2 specific parameters: acceleration and voltage. The acceleration values generated in the random series were applied at the base of the structure and were discretized in the numerical model. These acceleration values were produced according to the ordering in time and magnitude of the possible model responses. For such purposes, the sampling frequency of the acceleration parameter was 1 x 10-3 s, and the amplitude values ranged approximately within the interval [-6, 6] m/s2. The application of random acceleration on the base of the structure works as a type of filter, with the obtained responses (displacement, velocity, and acceleration values) in the state representation of the analyzed building creating consistent values to feed the network. Thus, the input dataset for the training processes and the validation of the prediction model were as follows: the responses obtained fro. m the structure (displacement, velocity, and acceleration); the voltage values generated from a series of normally distributed data, with a sampling frequency of 1 x 10-3 s and an amplitude of 2.5 v; and the optimal control force values generated in the neural network output, which enters the system through the use of a delay line to produce system feedback.

Figs. 5 and 6, respectively, present the voltage and acceleration values over time, generated based on the series of normally distributed random data.

With the excitation of the structure defined, the response values could then be obtained from the modeled gantry.

These values, along with the voltage values shown in Fig. 5 and the optimal force values determined by the model, 3

Voltage (v)

Input S(n)

2,5 2 1,5 1 0,5 0 0

2

4

6

8

10

Time (s)

4. Conditioning of both the predictive model and the inverse model used in the control design

Figure 5. Voltage values generated for the training and validation of the predictive model Source: The Authors

4.1. Conditioning the prediction model for the optimal control force The dataset used for the training and subsequent

Aceleration at the base (m/s2)

As previously mentioned, the control design was based on a predictive model and an inverse model acting together.

The predictive model determined the optimal control force values, while the inverse model defined the voltage values applied in the damper. Both models were run based on NARX-type neural networks.

Because the use of neural networks leads to a series of conditioning tasks (training and validation), the detailed procedure used by [7] to condition the prediction models used in the present study is presented below.

7 5 3 1 ‐1 ‐3 ‐5 ‐7 0

2

4

6

8

10

Time (s)

Figure 6. Acceleration values generated for use at the base of the structure in the predictive model.

Source: The Authors

194

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014.

comprised the set of sensory units that constituted the input layer of the prediction model. Fig. 7 shows the response values obtained by applying the excitation, shown in Fig. 6, to the structure. a)

Displacement (m)

0,004 0,002

mechanical model depended on the voltage values and the structure responses. Thus, working with the parameters shown in Figs. 5 and 7 in the model proposed in [3], the control force values were obtained according to the input parameters of the established neural network. Fig. 8 shows the desired control force values (target output) that were originated by the phenomenological model of the MR dampers as a result of the introduction of the responses and voltages specified in the input layer of the network.

0

4.2.

Conditioning of the applied inverse model to determine the voltage values for the control device

‐0,002 ‐0,004 0

2

4

6

8

10

Time (s)

b)

Velocity (m/s)

0,2 0,1 0 ‐0,1 ‐0,2 0

2

4

6

8

10

6

8

10

Time (s)

c)

Aceleration (m/s2)

15 10 5 0 ‐5 ‐10

The data used for the training and validation of the network that composed the inverse model were developed based on 2 random series of numbers generated from a normal distribution. In the specific case of the inverse model, the random parameters that generated the input values for the system were the output of the NARX network (voltage) and an acceleration value that was applied at the base of the studied gantry. The response values of the structure were the result of the application of the random acceleration to the gantry and were determined from the state representation of the system, and the control force values were the result of the insertion of the voltage values and structure responses into the mechanical model of the MR dampers. The sampling frequency values of the generated voltage and acceleration were both 1 x 10-3s, while the amplitudes of the generated parameters were approximately 2.5 V e ± 6 m/s2. Fig. 9 and 10 show the variation, over time, of the randomly generated voltage and acceleration values, respectively, that enabled the training and validation of the network proposed in the inverse model.

‐15 0

2

4

3

Time (s)

Voltage (v)

Figure 7. Response values used for the training and validation of the prediction model: a) Displacement; b) Velocity; c) Acceleration.

Source: The Authors

2,5 2 1,5 1 0,5

1500

0 0

Force (N)

1000

2

4

6

8

10

Time (s)

500

Figure 9. Voltage values generated for the training and validation of the inverse model.

Source: The Authors

0 ‐500

‐1000 ‐1500 0

2

4

6

8

10

Aceleration at the base (m/s2)

Tempo (s)

Figure 8. Control force values obtained from the phenomenological model for the MR dampers. The data were acquired for the training and validation of the predictive model.

Source: The Authors

The network training and validation dataset was complemented with the definition of the target output values for the system. For the specific case of the prediction model, the desired outputs were the control force values obtained from the phenomenological model of the MR dampers. The

8 6 4 2 0 ‐2 ‐4 ‐6 ‐8 0

2

4

6

8

10

Time (s)

Figure 10. Acceleration values generated for use at the base of the structure in the inverse model.

Source: The Authors

195

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014.

1,5

a)

1

Aceleration (m/s2)

Displacement (m)

0,004 0,002 0 ‐0,002

2

4

6

8

20

30 40 Time (s)

50

60

70

Figure 13 Acceleration record used.

Source: The Authors

b)

0,2

10

10

Time (s)

Velocity (m/s)

‐1 0

0

0,1

shown in Fig. 12. Although the control force values for the training and validation of the system were dependent on the response and voltage values of the analyzed model, these values were part of the system input, generating output values that correspond to the input values from the control plant.

0 ‐0,1 ‐0,2 0

2

4

6

8

10

Time (s)

c)

15

Aceleration (m/s2)

0 ‐0,5

‐1,5

‐0,004

10 5 0 ‐5 ‐10 ‐15 0

2

4

6

8

10

Time (s)

Figure 11. Response values used for the training and validation of the inverse model: a) Displacement; b) Velocity; c) Acceleration.

Source: The Authors

1500 1000

Force (N)

0,5

500

5. Performance of the proposed control model The proposed control model was tested on the studied gantry. The base of the structure was subjected to the excitation action shown in Fig. 13. The acceleration record that was used to excite the structure was scaled in time and magnitude to make it compatible with the structure dimensions.

For the specific case of a decrease in the displacement peaks of each floor of the structure, the values obtained when control was managed by neural networks were 66.67% for the first floor and 68.70% for the second floor when compared with the displacement peaks of the uncontrolled structure, which correlates to peak response values of 0.0017 m and 0.0036 m for the first and second floors, respectively. When a comparative exercise was performed using the RMS (RootMean-Square) values of the displacement, decreases in the displacement values of 78.69% and 79.40% were observed for the first and second floors of the structure.

Table 3.

Responses obtained from the system controlled by the semi-active algorithm based on NARX networks. ‐1000 Response values First floor Second floor ‐1500 Maximum peak (m) 0.0017 0.0036 0 2 4 6 8 10 Time (s) Decrease peak (%) 66.67 68.70 Figure 12. Control force values obtained from the phenomenological model RMS value (m) 0.0003 0.0006 for MR dampers. The data were acquired for the training and validation of RMS value decrease (%) 78.69 79.40 the inverse model.

Source: The Authors Maximum peak (m/s) 0.0535 0.0979 0

Acceleration

The displacement, velocity, and acceleration values of the structure, which were obtained from the application of the acceleration values shown in Fig. 10, are presented in Fig. 11. These variables were the response values of the input layer used in the training and validation of the neural network that comprises the inverse model.

Finally, the control force values obtained from the phenomenological model of the device, which also served as source nodes in the sensory unit of the inverse model, are

Velocity

Displacement

‐500

Peak decrease (%)

65.12

71.61

RMS value (m/s)

0.0070

0.0155

RMS value decrease (%)

82.76

83.21

Maximum peak (m/s2)

2.3828

3.4982

Peak decrease (%)

51.62

66.91

RMS value (m/s2)

0.2690

0.4568

RMS value decrease (%)

78.63

83.88

Source: The Authors

196

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014.

0,004 0,002 0 ‐0,002 ‐0,004 ‐0,006 10

20

30 40 Time (s)

50

60

0,004 2,9E‐17 ‐0,004 ‐0,008

70

0

10

20

b)

Velocity first floor

0,2

30 40 Time (s)

50

60

0,1 0 ‐0,1 ‐0,2

70

b)

Velocity second floor 0,4

Velocity (m/s)

Velocity (m/s)

0,008

‐0,012 0

0,2 0 ‐0,2 ‐0,4

0

10

20

30 40 Time (s)

50

60

70

0

10

20

c) c)

Aceleration first floor

6

30 40 Time (s)

50

60

Uncontrolled

4

Controlled

2 0 ‐2 ‐4 ‐6

70

c)

Aceleration second floor 12

Aceleration (m/s2)

Aceleration (m/s2)

a)

Displacement second floor 0,012

Displacement (m)

Displacement (m)

a)

Displacement first floor

0,006

Uncontrolled

8

Controlled

4 0 ‐4 ‐8 ‐12

10

20

30 40 Time (s)

50

60

70

Figure 14. Responses on the first floor of the structure for both the uncontrolled model and the model controlled by ANN: a) Displacement; b) Velocity; and c) Acceleration.

Source: The Authors

Table 3 shows the response values of the structure when managed by the control design proposed in the present study. The decreases in the values of these responses were calculated when compared with the values obtained in the uncontrolled structure.

Figs. 14 and 15 shows the variation in the responses of the structure over time. The cases shown correspond to the model controlled by the NARX neural networks and to the model where no control is exerted.

The behavior of the prediction model for the generated controller can be observed in Fig. 16. In this graph, it is possible to observe how the selected voltage values vary over time according to the system requirements. The initial voltage value for the MR dampers is 1.5 volts; therefore, the network started with this voltage value to determine in which direction the voltage value would generate control force values that approached the desired control force values obtained from the prediction model. As observed, the voltage in this case never reached 0; this result is primarily due to the nature of the excitation, which, during the time of the analysis, never ceased to influence the structure.

Considering that this influence was small in the last 10 s, it should be noted that the proposed neural networks were designed with 2 delay lines, which means that the neural

0

10

20

30 40 Time (s)

50

60

70

Figure 15. Responses on the second floor of the structure for both the uncontrolled model and the model controlled by ANN: a) Displacement; b) Velocity; and c) Acceleration.

Source: The Authors

3 2,5

Voltage (V)

0

2 1,5 1 0,5 0 0

10

20

30 40 Time (s)

50

60

70

Figure 16. Voltage value variations in the system controlled by ANN. Source: The Authors

networks made decisions based on up to 2 instants of past time; therefore, when the structure responses became stable, the system entered into a repetition of output values, resulting in a virtually fixed voltage value or, in this specific case, a voltage value with little variation at the end of the observation period.

In addition, Fig. 17 shows the variation in the control force values of the system caused by the voltage value variations generated in the prediction model of the controller. The control force values were examined with respect to time, displacement, and velocity.

197

Lara-Valencia et al / DYNA 81 (188), pp. 191-198. December, 2014.

Force (N)

demand for processing time, which hindered its execution in real time and would raise the cost of project implementation due to the need for elements with high computing power to solve the problem more quickly.

a)

Force vs. time 1000 750 500 250 0 ‐250 ‐500 ‐750 ‐1000

Acknowledgements

0

10

20

30 40 Time (s)

50

60

70

c) b)

Force vs. displacement 1000

Force (N)

500 0

References

‐500

[1]

‐1000 ‐0,0025

‐0,00125

‐1E‐18 Displacement (m)

0,00125

0,0025

c) c)

Force vs. velocity 1000

[2]

500

Force (N)

The authors acknowledge the support provided by the University of Brasília (Universidade de Brasília), the National University of Colombia, Medellin campus (Universidad Nacional de Colombia, sede Medellín), and the National Council for Scientific and Technological Development (Conselho Nacional de Desenvolvimento Científico e Tecnológico - CNPq) for the development of the present study.

[3]

0 ‐500 ‐1000 ‐0,06

[4] ‐0,04

‐0,02 0 Velocity (m/s)

0,02

0,04

0,06

[5]

Figure 17. Behavior of MR damper forces in the controlled system.

Variations are shown with respect to a) Time; b) Displacement; and c) Velocity.

Source: The Authors

[6]

6. Conclusions

[7]

A semi-active control design was developed in the present article using MR dampers that were managed by a control algorithm based on artificial neural networks. To measure the functionality and performance of the proposed system, a numerical application was developed using the control design on a 3-dimensional, 2-story structure that was subjected to the actions of a transient load.

The controller that was developed based on neural networks was able to reduce the peak and the RMS response values for the displacement of the structure by 67% and 79%, respectively, on average. For velocity, the peak and RMS response values were decreased by approximately 69% and 83%, respectively. Finally, for acceleration, an average reduction of 57% and 81% was achieved for the peak and response RMS values, respectively.

Based on the obtained numerical results, the control design based on neural networks that was developed in the present study can be considered an efficient, robust, reliable, and constant controller that was able to reduce the response values of the analyzed model. To accomplish this, the predictive and inverse models acted in a competent, appropriate, and synchronized manner, despite the complexity of the problem and solution. Perhaps the greatest weakness for this control alternative was the

Carneiro, R., Controle semi-ativo de vibrações em estruturas utilizando amortecedor magnetorreológico PhD. Thesis, Doutorado em Estruturas e Construção Civil, Universidade de Brasília, Distrito Federal, Brasilia, Brasil, 2009, 135 P. Lara, L., Brito, J. y Valencia, Y. Reducción de vibraciones en un edificio mediante la utilización de amortiguadores magnetoreológicos. DYNA, 79 (171), pp. 205-214, 2012.

Spencer Jr., B.F., Dyke, S.J., Sain, M.K. and Carlson, J.D., Phenomenological model of a magnetorheological damper. Journal of engineering mechanics, 123 (3), pp. 230-238, 1997. http://dx.doi.org/10.1061/(ASCE)0733-9399(1997)123:3(230) Lord Corporation, Lord technical data RD-1005-3 Damper, Technical data, Lord Corporation, Cary, North Carolina, 2006.

Basili, M., Controllo semi attivo di strutture adiacenti mediante dispositivi magnetoreologici: Teoria, sperimentazione e modellazione PhD. Thesis, in Structural Engineering, Università degli studi di Roma “La Sapienza”, Roma, Italia, 2006.

He, X. and Asada, H., A new method for identifying orders of inputoutput models for nonlinear dynamic systems, Proceedings of the American Control Conference, San Francisco, California, USA, pp.

2520-2523, 1993.

Lara, L.A., Estudo de algoritmos de controle semi-ativo aplicados a amortecedores magnetorréologicos, PhD. Thesis, Doutorado em Estruturas e Construção Civil,: Universidade de Brasília, Distrito Federal, Brasilia, Brasil, 2011, 223 P.

L.A. Lara-Valencia, received the BSc. in Civil Engineering in 2005 from Universidad Nacional de Colombia, campus Medellin, Colombia; the MSc. and Dr. degrees in Structures and Civil Construction in 2007 and 2011, respectively from the University of Brasilia, Brazil. Currently he is a full professor in the Civil Engineering department of the Universidad Nacional de Colombia, campus Medellin, Colombia. His research interest includes: vibration control of structures, dynamics of structures, linear and nonlinear finite elements modeling, foundations and tropical soils.

J.L. Vital-de Brito, received the BSc. in Civil Engineering in 1974 from University Estadual Paulista, Brasil the MSc. and Dr. degrees in Civil Engineering in 1979 and 1995, respectively from the University Federal do Rio Grande do Sul, Brasil. Currently he is a full professor in the Civil and Environmental Engineering department of the University of Brasilia and an active reviewer from the Journal of the Brazilian Society of Mechanical Sciences and Engineering. His research interest includes: dynamic of structures, structural stability, aerodynamics and vibration control.

Y. Valencia-Gonzalez, received the BSc. in Civil Engineering in 2001, the MSc. degree in Civil Engineering-Geotechnical in 2005, both from Universidad Nacional de Colombia, campus Medellin, Colombia. In 2009 received the Dr. degree in Geotechnical Follow by a year as postdoctoral fellow, all of them in the University of Brasilia, Brasil. Currently she is a full professor in the Civil Engineering department of the Universidad Nacional de Colombia campus Medellin, Colombia. Her research interest includes: tropical soils, biotechnology, foundations and vibration control.

198

Design of boundary combined footings of rectangular shape using a new model Arnulfo Luévanos-Rojas University of Durango State, Gómez Palacio, Durango, México. arnulfol_2007@hotmail.com Received: January 28th, 2014. Received in revised form: August 6th, 2014. Accepted: August 11th, 2014.

Abstract This paper presents the design of boundary combined footings of rectangular shape using a new model to consider real soil pressure acting on the contact surface of the footing; such pressure is presented in terms of an axial load, moment around the “X” axis and moment around the “Y” axis to each column. The classic model considers an axial load and moment around the transverse axis applied in each column, and when the moments in two directions are taken into account, the maximum pressure throughout the contact surface of the footing is considered the same. The main part of this research is that the proposed model considers real soil pressure and the classic model takes into account the maximum pressure and uniform is considered. It is concluded that the proposed model is more suited to the real conditions and is more economical.

Keywords: boundary combined footings; resultant force; center of gravity; bending moment; bending shear; punching shear.

Diseño de zapatas combinadas de lindero de forma rectangular utilizando un nuevo modelo Resumen Este documento presenta el diseño de zapatas combinadas de lindero de forma rectangular utilizando un nuevo modelo para considerar la presión real del suelo que actúan en la superficie de contacto de la zapata, dicha presión se presenta en función de una carga axial, momento alrededor del eje “X” y momento alrededor del eje “Y” de cada columna. El modelo clásico considera una carga axial y un momento alrededor del eje transversal aplicada en cada columna, y cuando los momentos en dos direcciones son tomados en cuenta, la presión máxima en toda la superficie de contacto de la zapata se considera la misma. La parte principal de esta investigación es que el modelo propuesto considera la presión real del suelo y el modelo clásico toma en cuenta la presión máxima y la considera uniforme. Se concluye que el nuevo modelo es el más apropiado, ya que se apega más a las condiciones reales y es más económico.

Palabras clave: zapatas combinadas de lindero; Fuerza resultante; Centro de gravedad; Momento flexionante; Fuerza cortante por flexión; Fuerza cortante por penetración.

1. Introduction The foundation is the part of the structure which transmits the loads to the soil. Each building demands the need to solve a problem of foundation. The foundations are classified into superficial and deep, which have important differences: in terms of geometry, the behavior of the soil, its structural functionality and its constructive systems [1,2].

Superficial foundations may be of various types according to their function; isolated footing, combined footing, strip footing, or mat foundation [1-4].

The distribution of soil pressure under a footing is a function of the type of soil, the relative rigidity of the soil

and the footing, and the depth of foundation at level of contact between footing and soil. A concrete footing on sand will have a pressure distribution similar to Fig. 1(a).

When a rigid footing is resting on sandy soil, the sand near the edges of the footing tends to displace laterally when the footing is loaded. This tends to decrease in soil pressure near the edges, whereas soil away from the edges of footing is relatively confined. On the other hand, the pressure distribution under a footing on clay is similar to Fig. 1(b).

As the footing is loaded, the soil under the footing deflects in a bowl-shaped depression, relieving the pressure under the middle of the footing. For design purposes, it is common to assume the soil pressures are linearly distributed.

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 199-208. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41800

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

Figure 1. Pressure distribution under footing: (a) footing on sand; (b) footing on clay; (c) equivalent uniform distribution.

Source: Bowles, 1996

a2’-a2’ with a width “b2” that are parallel to axis “Y-Y”, and moments around of an axis b’-b’, c’-c’, d’-d’ and e’-e’ that are parallel to axis “X-X”; 2) Bending shear; 3) Punching shear for footings which support a boundary column and other inner column subject to axial load and moment in two directions (bidirectional bending), where pressures are different in the four corners, these pressures are presented in terms of the mechanical elements (axial load, moment around the axis “X-X” and moment around the axis “Y-Y”).

2. Methodology

The pressure distribution will be uniform if the centroid of the footing coincides with the resultant of the applied loads, as shown in Fig. 1(c) [1].

In the design of superficial foundations, in the specific case of isolated footings, there are of three types in terms of the application of loads: 1) The footings subjected to concentric axial load, 2) The footings subjected to axial load and moment in one direction (unidirectional bending), 3) The footings subjected to axial load and moment in two directions (bidirectional bending) [1,2,5,6].

The hypothesis used in the classical model considers the axial load and moment around an axis transverse to the combined footing for the geometric proportions and shape are so fixed that the centroid of the footing area coincides with the resultant of the column loads. This results in uniform pressure below all the contact area of the footing.

Then the equation of the bidirectional bending is used to obtain the stresses acting on the contact surface of the combined footings, which must meet the following conditions: 1) The minimum stress should be equal to or greater than zero, because the soil is not capable of withstand tensile stresses, 2) The maximum stress must be equal or less than the allowable capacity that can withstand the soil [1,2,5,6].

A combined footing is a long footing supporting two or more columns in (typically two) one row. The combined footing may be rectangular, trapezoidal or Tee-shaped in plan. Rectangular footing is provided when one of the projections of the footing is restricted or the width of the footing is restricted. Trapezoidal footing is provided when one column load is much more than the other. As a result, both projections of the footing beyond the faces of the columns will be restricted [7-9].

Some papers present the use of load testing on foundations: Non-destructive load test in pilots [10]; Evaluation of the integrity of deep foundations: analysis and in situ verification [11]; Other, shows the use of static load tests in the geotechnical design of foundations [12]; Comparison between resonant-column and bender element test on three types of soils [13].

Mathematical models have been developed to obtain the dimensions of rectangular, square and circular isolated footings subjected to axial load and moments in two directions (bidirectional bending) [14-16]. Also, a mathematical model was presented for design of isolated footings of rectangular shape using a new model [17].

This paper presents a full mathematical model for the design of boundary combined footings to obtain: 1) Moments around of an axis a1’-a1’ with a width “b1” and

2.2. General conditions

According to Building Code Requirements for Structural Concrete (ACI 318-13) and Commentary the critical sections are: 1) the maximum moment is located in face of column, pedestal, or wall, for footings supporting a concrete column, pedestal, or wall; 2) bending shear is presented at a distance “d” (distance from extreme compression fiber to centroid of longitudinal tension reinforcement) shall be measured from face of column, pedestal, or wall, for footings supporting a column, pedestal, or wall; 3) punching shear is localized so that its perimeter “bo” is a minimum but need not approach closer than “d/2” to: (a) Edges or corners of columns, concentrated loads, or reaction areas; and (b) Changes in slab thickness such as edges of capitals, drop panels, or shear caps [18].

The general equation for any type of footings subjected to bidirectional bending [14-17, 19-21]: (1) where: σ is the stress exerted by the soil on the footing (soil pressure), A is the contact area of the footing, P is the axial load applied at the center of gravity of the footing, Mx is the moment around the axis “X”, My is the moment around the axis “Y”, Cx is the distance in the direction “X” measured from the axis “Y” up to the farthest end, Cy is the distance in direction “Y” measured from the axis “X” up to the farthest end, Iy is the moment of inertia around the axis “Y” and Ix is the moment of inertia around the axis “X”.

2.2. New model Fig. 2 shows a combined footing supporting two rectangular columns of different dimensions (a boundary column and other inner column) subject to axial load and moments in two directions in each column.

Fig. 3 presents a combined footing due to the equivalent loads. The mechanical elements of the components P1, Mx1, My1 are equivalent to a normal force “P1” acting on the point with coordinates (ex1, ey1), and for the components of P2, Mx2, My2 are equivalent to a normal force “P2” acting on the point with coordinates (ex2, ey2).

The general equation of the bidirectional bending is:

200

(2)

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

Figure 4. Pressure of the foundation on soil.

Source: Prepared by the author.

Figure 2. Boundary combined footing subjected to the real loads.

Source: Prepared by the author.

Figure 5. Boundary combined footing in plan.

Source: Prepared by the author.

Figure 3. Combined footing due to the equivalent loads.

Source: Prepared by the author.

where: σadm is the capacity of available allowable load of the soil, R is the resultant force of the forces, yc is the distance from the center of the contact area of the footing in the direction “Y” to the resultant, xc is the distance from the center of the contact area of the footing in the direction “X” to the resultant.

Now the sum of moments around the axis “X1” is obtained to find “yR” and the resultant force is made to coincide with the gravity center of the area of the footing with the position of the resultant force in the direction “Y”, therefore there is not moment around the axis “X” and the value of “yc” is zero, “xR = xc” is the sum of moments around the axis “Y” divided by the resultant, which is:

Fig. 5 presents a boundary combined footing to obtain the stresses anywhere of the contact surface of the structural member due to the pressure that is exerted by the soil.  In the longitudinal direction: (5) 

In the transverse direction:  To the boundary column is: (6)  To the intermediate column is: (7)

(3) Substituting equation (3) into equation (2) is transformed into a unidirectional bending system as follows:

where: b1 = c1+d/2 is the width of the failure surface, b2 = c3+d.

2.2.1. Model to obtain the bending moments

(4) Fig. 4 shows pressure diagram for combined footings subject to axial load and moment in one direction (unidirectional bending) in each column, where the pressures are presented at two different corners varying linearly along the contact surface, because there is not moment around the axis “X”.

Critical sections for bending moments are shown in Fig. 6, these are presented in sections a1’-a1’, a2’-a2’, b’-b’, c’c’, d’-d’ and e’-e’.

2.2.1.1. Moment around the axis a1’-a1’

201

The resultant force “FRa1’” is found through the volume

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014. /

/ /

/

/

/ /

/

(13)

The moment around the axis a2’-a2’ is: (14) Figure 6. Critical sections for bending moments.

Source: Prepared by the author.

Substituting the equation (12) and (13) into equation (14) is obtained:

of pressure the area formed by the axis a1’-a1’ with a width “b1 = c1+d/2” and the free end of the rectangular footing, where the higher pressure is presented: /

/

/

2.2.1.3. Moment around the axis b’-b’ The resultant force “FRb’” is the force “P1” acting on column 1 less the volume of pressure the area formed by the axis b'-b’ and the corners 1 and 2 to the left of the footing, this is presented of the follows:

(8)

/

(15)

The center of gravity “xca1’” is obtained by the equation: /

/

/

/ /

(16)

/

/ / /

/

/ /

(9)

The center of gravity “ycb’” with respect to axis b’-b’ is: (17) The moment around the axis b’-b’ is:

The moment around the axis a1’-a1’ is: (10)

(18)

Substituting the equation (8) and (9) into equation (10) is obtained:

Substituting the equation (16) and (17) into equation (18) is obtained:

(11)

(19)

2.2.1.2. Moment around the axis a2’-a2’

2.2.1.4. Moment around the axis c’-c’

The resultant force “FRa2’’” is obtained through the volume of pressure the area formed by the axis a2’-a2’ with a width “b2 = c3+d” and the free end of the rectangular footing, where the higher pressure is presented:

First, the position of the axis c’-c’ must be localized, which is where the maximum moment is located.

When the shear force is zero, the moment should be the maximum, then the shear force is presented at a distance “ym”, this is shown as follows:

/

/ /

/

/

/

(12) /

The center of gravity “xca2’” is obtained by the equation:

(20)

Now the equation (20) is equal to zero and we obtain: 202

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

(21) Then the maximum moment is obtained as follows: (22) Substituting the equation (21) into equation (22) is: (23) 2.2.1.5. Moment around the axis d’-d’ The resultant force “FRd’” is the force “P1” acting on column 1 less the volume of pressure the area formed by the axis d'-d’ and the corners 1 and 2, which is found to the left of the footing, this is as follows: /

Figure 7. Critical sections for bending shear.

Source: Prepared by the author.

free end of the rectangular footing, where the greatest pressure is presented:

/ ⁄

/

(24)

/

/

/

/

(28) The moment around the axis d’-d’ is: (25) 2.2.1.6. Moment around the axis e’-e’ The resultant force “FRe’” is the sum of the force “P1” acting on column 1 and the force “P2” acting on column 2 less the volume of pressure the area formed by the axis e'-e’ and the corners 1 and 2, which is found to the left of the footing, this is as follows:

2.2.2.2. Bending shear in axis f2’-f2’ Bending shear acting on the axis f2’-f2’ of the footing “Vff2’” is obtained through the volume of pressure the area formed by the axis f2’-f2’ with a width “b2 = c3+d” and the free end of the rectangular footing, where the greatest pressure is presented: / /

/

/ /

/ ⁄

/

(29) (26) 2.2.2.3. Bending shear in axis g’-g’

The moment around the axis e’-e’ is: (27) 2.2.2. Model to obtain the bending shear

Bending shear acting on the axis g’-g’ of the footing “Vfg’” is the force “P1” acting on column 1 less the volume of pressure the area formed by the axis g'-g’ and the corners 1 and 2 to the left of the footing, this is as follows: /

The critical sections for bending shear are obtained at a distance “d” starting the junction of the column with the footing as seen in Fig. 7, these are presented in sections f1’f1’, f2’-f2’, g’-g’, h’-h’ and i’-i’.

⁄

/ /

(30)

2.2.2.1. Bending shear in axis f1’-f1’

2.2.2.4. Bending shear in axis h’-h’

Bending shear acting on the axis f1’-f1’ of the footing “Vff1’” is obtained through the volume of pressure the area formed by the axis f1’-f1’ with a width “b1 = c1+d/2” and the

Bending shear acting on the axis h’-h’ of the footing “Vfh’” is the force “P1” acting in column 1 less the volume of pressure the area formed by the axis h'-h’ and the corners

203

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

1 and 2, which is found to the left of the footing, this is: /

/

’

⁄

(31)

/

2.2.2.5. Bending shear in axis i’-i’ Bending shear acting on the axis i’-i’ of the footing “Vfi’” is the sum of the force “P1” acting on column 1 and the force “P2” acting on column 2 less the volume of pressure the area formed by the axis i’-i’ and the corners 1 and 2, which is found to the left of the footing, this: /

/

’

⁄

/

Figure 8. Critical sections for punching shear.

Source: Prepared by the author.

(32) 

In the longitudinal direction:

2.2.3. Model to obtain the punching shear

(35)

The critical section for the punching shear appears at a distance “d/2” starting the junction of the column with the footing in the two directions.



In the transverse direction:  To the boundary column is:

2.2.3.1. Punching shear for boundary column

(36)

The critical section for the punching shear is presented in rectangular section formed by points 3, 4, 5 and 6, as shown in Fig. 8. Punching shear acting on the footing “Vp1” is the force “P1” which acting on column 1 less the volume of pressure the area formed by the points 3, 4, 5 and 6: ⁄

⁄ ⁄

 To the intermediate column is: (37) 2.3.1. Model to obtain the moments

(33)

⁄

Critical sections for bending moments are shown in Fig. 6, these are presented in sections a1’-a1’, a2’-a2’, b’-b’, c’-c’, d’-d’ and e’-e’. The bending moment in each section is:

2.2.3.2. Punching shear for inner column The critical section for the punching shear is presented in rectangular section formed by points 7, 8, 9 and 10, as shown in Fig. 8. Punching shear acting on the footing “Vp2” is the force “P2” which acting on column 2 less the volume of pressure the area formed by the points 7, 8, 9 and 10: ⁄ ⁄

(39) (40)

⁄ ⁄

(38)

(34)

2.3. Classic model

(41) (42)

This model takes into account only the maximum pressure of the soil for design of footings and it is considered uniform at all points on contact area of footings.

The maximum pressure is: 204

(43)

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

To xR ≥ b/6:

2.3.2. Model to obtain the bending shear The critical sections for bending shear (seen in Fig. 7), these are presented in sections f1’-f1’, f2’-f2’, g’-g’, h’-h’ and i’-i’. The bending shear in each section is: (44) (45) (46) (47)

(54) where: b is the dimension of the parallel footing the axis “X”, My=My1+My2.

Note: if in the combinations are included the wind and/or the earthquake, the load capacity the soil should be increased by 33% [18].

Step 4: The mechanical elements (P, Mx, My) acting on the footing are factored [18].

Step 5: The bending moments acting on the combined footing are obtained.

Step 6: The effective depth “d” for the maximum moment is found by the following expression [18]:

(48) (55)

2.3.3. Model to obtain the punching shear The critical sections for the punching shear are presented in Fig. 8.  The punching shear for boundary column (49)  The punching shear for inner column

where: Mu is the factored maximum moment at section acting on the footing, Øf is the strength reduction factor by bending and its value is 0.90, bw is width of analysis in structural member, ρ is ratio of “As” to “bwd”, fy is the specified yield strength of reinforcement of steel, f’c is the specified compressive strength of concrete at 28 days.

Step 7: Bending shear resisted by the concrete “Vcf” is [18]:

(50) (56) 2.4. Procedure of design Step 1: The mechanical elements (P, Mx, My) acting on the footing is obtained by the sum of: the dead loads, live loads and accidental loads (wind or earthquake) from each of these effects [20,21].

Step 2: The available load capacity the soil “σadm” is [20, 21]: (51) where: qa is the allowable load capacity the soil, γppz is the self-weight of the footing, γpps is the self-weight the soil fill.

Step 3: The value of “a” is selected according to the following equation:

To bending shear acting on the footing (Vf) is compared vs. bending shear resisting by concrete (Vcf) and is [18]: (57) where: Øv is the strength reduction factor by shear is 0.85.

Step 8: Punching shear (shear force bidirectional) resisted by the concrete “Vcp” is given [18]: (58a) where: βc is the ratio of long side to short side of the column and b0 is the perimeter of the critical section.

(52) (58b) where: a is the dimension of the parallel footing the axis “Y”, R=P1+P2, Mx=Mx1+Mx2.

The value of “b” is: To xR ≤ b/6:

where: αs is 40 for interior columns, 30 for edge columns, and 20 for corner columns.

(58c)

(53)

where: ØvVcp must be the value smallest of equations (58a), (58b) and (58c).

205

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

To punching shear acting on the footing (Vp) is compared vs. punching shear resisting by concrete (Vcp) and must comply with the following expression [18]: (59) Step 9: The main reinforcement steel “Asp” is [18]: (60) where: w is 0.85f’c /fy.

The minimum steel “Asmin” percentage “ρmin” by rule are [18]:

and

the

minimum (61) (62)

The reinforcement steel by temperature is found [18]: (63) where: t is the total thickness of the footing.

Step 10: The development length in tension of deformed bars “ld” is expressed [18]: Steel reinforcement in the top: (64) Steel reinforcement in the bottom: (65) where: ψt is the traditional reinforcement location factor to reflect the adverse effects of the top reinforcement casting position, ψe is a coating factor reflecting the effects of epoxy coating, db is the diameter of the bars, λ is modification factor reflecting the reduced mechanical properties of lightweight concrete, all relative to normalweight concrete of the same compressive strength.

The development length for deformed bars “ld” is compared vs. the available length of the footing “la” and must comply with the following expression [18]: (66) 3. Application The design of a boundary combined footing supporting two square columns is presented in Fig. 9, with the basic information following: c1 = 40x40 cm; c2 = 40x40 cm; L = 6.00 m; H = 1.5 m; MDx1 = 140 kN-m; MLx1 = 100 kN-m; MDy1 = 120 kN-m; MLy1 = 80 kN-m; PD1 = 700 kN; PL1 =

Figure 9. Combined footing supporting two square columns.

Source: Prepared by the author.

500 kN; MDx2 = 280 kN-m; MLx2 = 200 kN-m; MDy2 = 240 kN-m; MLy2 = 160 kN-m; PD2 = 1400 kN; PL2 = 1000 kN; f’c = 21 MPa; fy = 420 MPa; qa = 220 kN/m2; γppz = 24 kN/m3; γpps = 15 kN/m3.

Where: H is the depth of the footing, PD is the dead load, PL is the live load, MDx is the moment around the axis “X-X” of the dead load, MLx is the moment around the axis “X-X” of the live load, MDy is the moment around the axis “Y-Y” of the dead load, MLy is the moment around the axis “Y-Y” of the live load.

Step 1: The loads and moments acting on soil: P1 = 1200 kN; Mx1 = 240 kN-m; My1 = 200 kN-m; P2 = 2400 kN; Mx2 = 480 kN-m; My2 = 400 kN-m.

Step 2: The available load capacity the soil: The thickness “t” of the footing is proposed, the first proposal is the minimum thickness of 25 cm marking regulations, subsequently the thickness is revised to meet the following conditions: moments, bending shear and punching shear. If such conditions are not satisfied is proposed a greater thickness until it fulfills the three conditions mentioned. The thickness of the footing that fulfills the three conditions listed above is 95 cm for new model and for classic model is 120 cm. Using the equation (51) is obtained the available load capacity of the soil “σadm” is 188.95 kN/m2 (new model) and 186.70 kN/m2 (classic model).

Step 3: The value of “a” by equation (52) is obtained: a = 8.00 m. The value of “b” by equation (53) is found: b = 3.20 m. These values are for the two models. This value of “b” is verified to xR ≤ b/6 and meets.

Step 4: The mechanical elements (P, Mx, My) acting on the footing is factored: Pu1 = 1640 kN; Mux1 = 328 kN-m; Muy1 = 272 kN-m; Pu2 = 3280 kN; Mux2 = 656 kN-m; Muy2 = 544 kN-m.

Step 5: The bending moments acting on the footing of the two models are presented in Table 1.

Table 1.

Bending moments Moments

New model (kN-m) Parallel to axis “Y-Y” 612.88 1225.77 Parallel to axis “X-X” 278.80 1858.67 ‒1558.00 ‒1771.20 Source: Prepared by the author.

206

Classic model (kN-m) 658.44 1316.88 263.50 1338.33 ‒5000.32 ‒6343.80

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

Table 2.

Dimensions Concept Parallel to axis “Y-Y” Parallel to axis “X-X” Effective depth after performing different proposals Coating Total thickness Source: Prepared by the author.

New model (cm) 44.35 34.41 87

Classic model (cm) 42.02 63.57 112

8 95

8 120

Table 3.

Bending shear Bending shear

New model (kN) Parallel to axis “Y-Y” 481.04 342.10(O.K.) 731.65 684.21(O.K.) Parallel to axis “X-X” 1843.52 858.95(O.K.) ‒1514.95(O.K.) 448.95(O.K.) Source: Prepared by the author.

Table 4.

Punching shear Punching shear

New model (kN) Boundary column 5081.19 8995.07 3287.83 1436.19(O.K.) Intermediate column 8779.74 12645.97 5681.01 2970.02(O.K.) Source: Prepared by the author.

Table 5.

Reinforcement steel Reinforcement steel

Classic model New model cm2 cm2 Longitudinal reinforcement steel (direction of axis “Y”)

Steel at the top

Steel in the bottom Classic model (kN) 711.98 188.13(O.K.) 1127.30 376.25(O.K.)

Main steel

57.94

31.95

Minimum steel

92.71

119.35

Steel proposed

96.27(19Ø1”)

121.61(24Ø1”)

Main steel

55.14

158.04

Minimum steel

92.71

119.35

Steel proposed

96.27(19Ø1”)

162.15(32Ø1”)

Transverse reinforcement steel (direction of axis “X”) Steel at the top

2373.26 616.08(O.K.) ‒2294.45(O.K.) ‒1142.92(O.K.)

Steel in the bottom under the boundary column Steel in the bottom under the inner column

Classic model (kN) 7653.77 14657.68 4952.44 1272.35(O.K.)

Temperature steel Steel proposed

136.80

172.80 173.86(61Ø3/4”)

Main steel

136.81(48Ø3/ 4”) 19.24

Minimum steel

24.19

35.80

Steel proposed

37.05(13Ø3/4”)

Main steel

25.65(9Ø3/4” ) 38.88

Minimum steel

36.79

56.69

Steel proposed

39.90(14Ø3/4 ”) 100.80

57.00(20Ø3/4”)

102.61(36Ø3/ 4”)

119.71(42Ø3/4”)

Steel in Temperature bottom of steel the excess Steel proposed parts of the columns Source: Prepared by the author.

13527.59 20625.03 8753.14 2697.89(O.K.)

Table 6.

Development length Concept

Step 6: The effective depth for the bending moment is found by equation (55); these are shown in Table 2.

Step 7: Bending shear appear in Table 3.

Step 8: Punching shear is presented in Table 4.

Step 9: The reinforcement steel is shown in Table 5.

Step 10: The minimum development length for deformed bars appear in Table 6.

Fig. 10 shows the dimensions and the reinforcement steel of the boundary combined footing for the two models.

4. Conclusions The foundation is a part essential of a structure, because permits the transmission of loads from the structure to the soil. The mathematical approach suggested in this paper produces results that have a tangible accuracy for all problems, main part of this research for find the solution more economical.

The proposed model presented in this paper for the structural design of boundary combined footings subjected

Steel at the top

ψt 1.3 ψe = λ 1.0 178.02 ld (cm) 259.00(O.K.) la(cm) Source: Prepared by the author.

15.83

31.80

119.23

Steel in the bottom 1.0 1.0 83.36 132.00(O.K.)

to an axial load and moment in two directions, also it can be applied to others cases: 1) The footings subjected to a concentric axial load, 2) The footings subjected to a axial load and moment in one direction.

The model presented in this paper applies only for design of boundary combined footings, the structural member is assumed to be rigid and the supporting soil layers elastic, which meet expression of the bidirectional bending, i.e., the variation of pressure is linear. The suggestions for future research, when is presented another type of soil, by example in totally cohesive soils (clay soils) and totally granular soils (sandy soils), the pressure diagram is not linear and should be treated differently (see Fig. 1).

207

Luévanos-Rojas / DYNA 81 (188), pp. 199-208. December, 2014.

Figure 10. Final design of the boundary combined footing: (a) New model; (b) Classic model.

Source: Prepared by the author.

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17]

Bowles, J.E., Foundation analysis and design, McGraw-Hill, New York, 1996.

Das, B.M., Sordo-Zabay, E. and Arrioja-Juárez, R., Principios de ingeniería de cimentaciones, Cengage Learning Latin America, México, 2006.

Calabera-Ruiz, J., Calculo de estructuras de cimentación, Intemac Ediciones, México, 2000.

Tomlinson, M.J., Cimentaciones, diseño y construcción, Trillas, México, 2008.

Mosley, W.H., Bungey, J.H. and Hulse, R., Reinforced concrete design, Palgrave Macmillan, New York, 1999.

Gambhir, M.L., Fundamentals of reinforced concrete design, Prentice-Hall, of India Private Limited, 2008.

Kurian, N.P., Design of foundation systems, Alpha Science Int'l Ltd., India, 2005.

Punmia, B.C., Kumar-Jain, A., and Kumar-Jain, A., Limit state design of reinforced concrete, Laxmi Publications (P) Limited, New Delhi, India, 2007.

Varghese, P.C., Design of reinforced concrete foundations, PHI Learning Pvt. Ltd., New Delhi, India, 2009.

Ibañez-Mora, L., Pruebas de carga no destructivas en pilotes, DYNA, 75 (155), pp. 57-61, 2008.

Gaviria, C.A., Gómez, D. and Thomson, P., Evaluación de la integridad de cimentaciones profundas: análisis y verificación in situ, DYNA, 76 (159), pp. 23-33, 2009.

Valencia, Y., Camapum, J. and Lara, L., Aplicaciones adicionales de los resultados de pruebas de carga estáticas en el diseño geotécnico de cimentaciones, DYNA, 175, pp. 182-190, 2012.

Camacho-Tauta, J.F., Reyes-Ortiz, O.J. and Jimenez-Alvarez, J.D., Comparison between resonant-column and bender element test on three types of soils, DYNA, 80 (182), pp. 163-172, 2013.

Luévanos-Rojas, A., A mathematical model for dimensioning of footings rectangular, ICIC Express Letters Part B: Applications, 4, pp.269-274, 2013.

Luévanos-Rojas, A., A mathematical model for dimensioning of footings square, International Review Civil Engineering (IRECE), 3, pp.346-350, 2012.

Luévanos-Rojas, A., A mathematical model for the dimensioning of circular footings, Far East Journal of Mathematical Sciences, 71, pp.

357-367, 2012.

Luévanos-Rojas, A., Faudoa-Herrera, J.G., Andrade-Vallejo, R.A. and Cano-Alvarez, M.A., Design of isolated footings of rectangular form using a new model, International Journal of Innovative Computing, Information and Control, 9, pp. 4001-4022, 2013.

[18] ACI 318S-13 (American Concrete Institute), Building Code Requirements for Structural Concrete and Commentary, Committee 318, 2013.

[19] Gere, J.M. and Goodo, B.J., Mecánica de materiales, Cengage Learning, México, 2009.

[20] González-Cuevas, O.M. and Robles-Fernández-Villegas, F., Aspectos fundamentales del concreto reforzado, Limusa, México, 2005.

[21] McCormac, J.C. and Brown, R.H., Design of reinforced concrete, John Wiley & Sons, New York, 2013.

A. Luévanos-Rojas, received the BSc. Eng in Civil Engineering in 1981, the MSc degree in Planneation and Construction in 1996, and the Engineering Dr. degree in Planneation and Construction in 2009, all of them from the Facultad de Ingeniería, Ciencias y Arquitectura of the Universidad Juárez del Estado de Durango, Gómez Palacio, Durango, México. The MSc degree in Structures in 1983, from the Escuela Superior de Ingeniería y Arquitectura the Instituto Politécnico Nacional, Distrito Federal, México. The MSc degree in Administration in 2004, from the Facultad de Contaduría y Administración of the Universidad Autónoma de Coahuila, Torreón, Coahuila, México. From 1983 to 2009, he is a full time professor and from 2009 to 2014, he is professor and researcher for the Facultad de Ingeniería, Ciencias y Arquitectura of the Universidad Juárez del Estado de Durango. His research interests include: mathematical models applied to structures: methods of structural analysis, members design of concrete and steel, analysis of non-prismatic members. Also he is Associate Editor the journal “ICIC Express Letters Part B: Applications”.

ORCID: 0000-0002-0198-3614.

208

Effect of additives on diffusion coefficient for cupric ions and kinematics viscosity in CuSO4H2SO4 solution at 60°C Eugenia Araneda-Hernández a, Froilán Vergara-Gutierrez b & Antonio Pagliero-Neira.c a

Departamento de Metalurgia, Facultad de Ingeniería, Universidad de Concepción. Concepión, Chile. euaraned@udec.cl Departamento de Metalurgia, Facultad de Ingeniería, Universidad de Concepción. Concepión, Chile. fvergar@udec.cl c Departamento de Metalurgia, Facultad de Ingeniería, Universidad de Concepción. Concepión, Chile. apaglier@udec.cl b

Received: January 29th, de 2014. Received in revised form: April 10th, 2014. Accepted: April 25th, 2014

Abstract The effect of levelling and grain-refining agents on the diffusion coefficient of cupric ion was studied. Tests were performed using synthetic solutions with Cu2+ concentration and acidity similar to those of a copper electro-refining industrial electrolyte by means of rotating disc electrode (RDE) technique. The diffusion coefficient was calculated according to Levich’s expression from measurements of limiting current for different rotating speed of the RDE. Arrhenius dependence of the diffusion coefficient with temperature in the absence of additives was verified. In presence of additives, the variation of the limiting current is mainly attributed to changes in diffusion coefficient of cupric ion observed.

Keywords: diffusion coefficient cupric ions; electrorefining copper; glue; thiourea; chloride.

Efecto de aditivos en el coeficiente de difusión de iones cúpricos y la viscosidad cinemática en solución CuSO4H2SO4 a 60° Resumen En este estudio se determinó el efecto de la adición de agentes inhibidores y/o afinadores de grano comúnmente utilizados en la electrorefinación industrial de cobre sobre el coeficiente de difusión de ión cúprico a composición de ácido, cobre y temperatura en el rango industrial promedio. El coeficiente de difusión del ión cúprico DCu2+ se obtuvo de acuerdo a la expresión de Levich a través de mediciones de corriente límite sobre electrodo de disco rotatorio de acero inoxidable 316 L en soluciones sintéticas de composición 40 g/L de Cu2+ y 200 g/L de H2SO4 a 60°C con adición de cola, tiourea y cloruro a tres niveles de concentración y se compararon con el valor obtenido en soluciones en ausencia de estos aditivos. La incorporación de aditivos resulta en una leve disminución del coeficiente de difusión del Cu2+ siendo más significativa en presencia de iones cloruro.

Palabras clave: coeficiente de difusión iones cúpricos; electrorefinación cobre; cola; tiourea; cloruro.

1. Introducción El proceso de electrorefinación de cobre desde un sistema CuSO4−H2SO4 ocurre mediante las semi reacciones anódicas y catódicas: Cu2+ + 2e → Cu0 catódica E° = +0.34 VENH

(1)

Cu0 → Cu2+ + 2e

(2)

anódica E°= 0.34 VENH

La electrodepositación de cobre involucra además una semirreacción intermedia de transferencia de carga dada

por la reducción de Cu2+ a Cu+ a 0.15 mVENH. La electrodepositación de cobre de acuerdo a la semirreacción (1) comprende tres etapas, a) la transferencia de masa de iones a través del electrolito soporte hacia las vecindades del electrodo, b) la transferencia de carga para obtener adátomos o ad-iones y c) la incorporación de los ad-átomos o ad-iones a la superficie catódica. La primera de ellas ocurre principalmente por la difusión de iones cúpricos hacia las proximidades del cátodo a través de la capa difusiva y el flujo másico global está dado por la primera ley de Fick cuya constante de proporcionalidad corresponde al coeficiente de difusión de la especie electroactiva, en este caso, iones cúpricos.

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 209-215. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41815

Araneda-Hernández et al / DYNA 81 (188), pp. 209-215. December, 2014.

Por otro lado, en la producción industrial de cátodos electrorefinados es común el uso de agentes nivelantes e inhibidores adicionados al electrolito en pequeñas concentraciones para la obtención de depósitos más compactos y de menor rugosidad [1-4]. El mecanismo de acción de estos aditivos en la morfología de los depósitos se explica dependiendo de su naturaleza. En el caso de aditivos orgánicos: cola animal y tiourea, diversos estudios sostienen que estos actúan bloqueando sitios activos en la superficie del cátodo por fenómenos de adsorción superficial mientras que el ión cloruro actuaría como catalizador de la reacción de reducción global (1).

Se han realizado algunos estudios [5-7] para determinar el coeficiente de difusión del ión cúprico, referidos principalmente al efecto de la composición de cobre y acidez y en algunos casos se agrega el efecto de la temperatura, pero en rangos diferentes a los usados industrialmente [8,9] ni tampoco consideran el efecto de los agentes nivelantes y/o afinadores de grano.

El presente estudio pretende entregar una contribución a la data disponible del coeficiente de difusión del ión cúprico en el sistema CuSO4−H2SO4 a composición de cobre y ácido en el rango industrial y el efecto que tendrían sobre este parámetro los aditivos más utilizados en los procesos de electro refinación de cobre. Todas las experiencias fueron realizadas usando la técnica de electrodo de disco rotatorio (EDR) y el coeficiente de difusión se determinó a partir de mediciones de corriente límite para diferentes velocidades de rotación.

Adicionalmente se realizaron mediciones de viscosidad cinemática del electrolito soporte bajo las diferentes condiciones experimentales consideradas para ser utilizadas en el cálculo del coeficiente de difusión.

2.2. Mediciones de viscosidad Las mediciones de viscosidad de cada una de las soluciones se realizaron en triplicado en un ViscoSystem® AVS 370 de Schott Instruments con viscosímetros capilares Cannon−Fenske routine.

2.3. Mediciones electroquímicas Como electrodo de trabajo se utilizó un electrodo rotatorio EDI 101 de Radiometer, con un disco de platino de 0.0314 cm2. El que fue sometido a una limpieza electrolítica en solución ácida libre de cobre previo a cada ensayo. El electrodo de referencia correspondió a un electrodo de calomel saturado (ECS) y un alambre de platino de alta pureza se empleó como contra electrodo.

El registro de las variaciones de potencial y de corriente se realizó con un potenciostato Voltalab 40 PGZ301 de Radiometer. Las pruebas voltamétricas se llevaron a cabo en el rango 200 a 1200 mVECS con una velocidad de barrido de 5 mV/s.

3. Resultados y Discusiones 3.1. Viscosidad cinemática del electrolito 3.1.1. Efecto de la temperatura La viscosidad cinemática en sistemas líquidos se relaciona exponencialmente con la temperatura de acuerdo a [10]: ఔ

(3)

௢

2. Metodología experimental

donde  es la viscosidad cinemática en cm2/s, o una constante, T es la temperatura absoluta en Kelvin, R es la constante de los gases y E es la energía de activación de la viscosidad en kJ/mol.

Este comportamiento se verifica claramente en Fig. 1 en el caso de la solución base en el rango de temperatura analizado, encontrando la correlación empírica dada por la ec. (4) con un coeficiente de correlación de 0.99:

2.1. Electrolito soporte El electrolito soporte utilizado en las mediciones de viscosidad y en las pruebas electroquímicas consistió de una solución ácida sintética de composición 40 g/L de Cu2+ (a la forma de CuSO4·5H2O pro análisis) y 200 g/L de H2SO4, denominada solución base. Los aditivos considerados en este estudio corresponden a tiourea, cola animal y cloruro, que son los habituales en el proceso de electro refinación de cobre. Los dos primeros proporcionados por la refinería de Chuquicamata y el último adicionado a la forma de KCl pro análisis. Estos se adicionaron individualmente a la solución justo antes de cada prueba, previa dilución en agua destilada, cada uno de los test se realizó dentro de los 30 minutos de adicionado el aditivo.

Se consideraron concentraciones de 1, 10 y 100 mg/L para cada uno de ellos. Las soluciones se mantuvieron en contacto con la atmósfera con agitación magnética hasta alcanzar la temperatura requerida del ensayo. Durante cada prueba la solución fue calentada y mantenida a la temperatura previamente fijada, con variaciones de ±1°C, usando una celda electrolítica de doble camisa conectada con un baño termostático.

1359.2

ି1

8.7

(4)

De la misma figura, se observa una disminución de la viscosidad cinemática con un aumento en la temperatura de la solución lo que concuerda con lo reportado por Price y Davenport [11] en el mismo rango de temperatura (para concentraciones de Cu y H2SO4 en los rangos 5 a 55 g/L y 10 a 165 g/L, respectivamente) y Subbaiah y Das [12] a 20, 30, 40 y 60ºC (para concentraciones de Cu y H2SO4 en los rangos 1.04 a 43.5 g/L y 147.1 a 450 g/L, respectivamente.

A partir de la ec. (4) se determinó la energía de activación para la viscosidad cinemática a partir de la expresión para la solución base en el rango de temperatura 20 a 80°C la que resulta en 11,3 kJ/mol.

210

Araneda-Hernández et al / DYNA 81 (188), pp. 209-215. December, 2014.

0.0 -3.8

-0.2 -4.0

-0.4

ln 

2

i (A/cm )

-4.2 -4.4 -4.6

200 rpm 400 rpm 600 rpm 800 rpm 1000 rpm 2000 rpm

-0.8

-4.8

-1.0

2

R =0.99 -5.0 0.0026

-0.6

0.0028

0.0030

0.0032

0.0034

0.0036

-1.2 100

1/T (K-1)

0

-100

-200

-300

-400

-500

-600

E (mVECS)

Figura 1. Dependencia de la viscosidad cinemática con la temperatura en solución de composición 40 g/L Cu2+ y 200 g/L H2SO4.

Fuente: Los autores.

Figura 2. Detalle de curvas de polarización en el EDR de platino para soluciones de composición 40 g/L Cu2+ y 200 g/L H2SO4 en ausencia de aditivos a 60°C a distintas velocidades de rotación. 5 mV/s.

Fuente: Los autores.

Tabla 1.

Resultados de viscosidad cinemática en función de la concentración de aditivos en solución de composición 40 g/L Cu2+ y 200 g/L H2SO4 a 60ºC. Aditivo Concentración  aditivo (mm2/s) (mg/L) Sin aditivo 0 0.9813 Tiourea 1 0.9912 10 1.0440 100 1.0580 Cola 1 1.0031 10 0.9917 100 0.9976 Cloruro 1 1.0049 10 0.9889 100 1.0095 Fuente: Los autores.

Tabla 2.

Corriente límite en función de la velocidad de rotación del EDR en electrolito de composición 40 g/L Cu2+ y 200 g/L H2SO4 en ausencia de aditivos a 60°C. Velocidad de iL IL ∙102 rotación (rpm) (A/cm2) (A) 200 -0.358 1.12 400 -0.544 1.71 600 -0.646 2.03 800 -0.765 2.40 1000 -0.840 2.64 2000 -1.188 3.73 Fuente: Los autores.

voltagramas de Fig. 2. A bajas velocidades de rotación, inferior a 800 rpm, y a bajos potenciales catódicos, inferiores a −300 mVECS (considerados en valor absoluto), la corriente aumenta hasta alcanzar un valor límite. A velocidades de rotación sobre 800 rpm este valor límite prácticamente desaparece por la alta turbulencia originada en el electrolito.

Los valores de corriente límite a distintas rpm extraídos de los voltagramas se presentan en Tabla 2. Estos muestran un aumento en el valor de IL al incrementar la velocidad de rotación del EDR. Graficando el valor de corriente límite IL en función de la velocidad de rotación del EDR, se verifica una dependencia lineal entre ambos (Fig. 3), tal como lo predice Levich [10] para régimen laminar controlado por difusión de acuerdo a la expresión:

3.1.2. Efecto de los aditivos La viscosidad cinemática medida al electrolito en presencia de aditivos para tres concentraciones se muestran en Tabla 1. Se observa un leve aumento de la viscosidad en presencia de cola y cloruro con diferencias en el rango 0.77% a 2.9% respecto de la solución base a la misma temperatura. A 10 y 100 mg/L de tiourea las desviaciones observadas respecto de la solución de composición base son de 6.4% y 7.8%, respectivamente.

En general, la presencia de cola, tiourea y cloruro no modificaría de manera significativa la movilidad iónica en el electrolito soporte.

3.2. Determinación de corriente límite IL

௅

3.2.1. Efecto de la velocidad de rotación

0.62

଴

2/3 ି1/6

1/2

(5)

donde IL es la corriente límite catódica en amperes, n es el número de moles equivalentes, F es la constante de Faraday en coulomb, C0 es la concentración de iones cúpricos en el seno del electrolito en mol/cm3, A es el área efectiva catódica en cm2 (asumiendo área geométrica del

Un detalle de las curvas de la solución base para diferentes velocidades de rotación del EDR bajo las condiciones experimentales se muestran en los 211

Araneda-Hernández et al / DYNA 81 (188), pp. 209-215. December, 2014.

electrodo),  es la viscosidad cinemática en cm2/s,  es la velocidad de rotación del EDR en rpm y D es el coeficiente de difusión de los iones cúpricos en cm2/s.

3.2.3. Efecto de la concentración de aditivos Los resultados muestran en general una relación lineal entre la corriente límite experimental IL y la velocidad de rotación del EDR en soluciones en presencia de aditivos (Fig. 5), lo que indica que se mantiene el mecanismo difusional como controlante de la transferencia de masa en el electrolito. Sin embargo, en presencia de cola (Fig. 5a) y tiourea (Fig. 5b) esta presenta una leve disminución respecto del valor observado en soluciones en ausencia de aditivos aún a bajas concentraciones de ambos inhibidores (1 mg/L) y se mantiene constante al aumentar la concentración a 10 y 100 mg/L.

3.2.2. Efecto de la temperatura Los resultados de la variación de corriente límite en función de la velocidad de rotación del EDR para distintas temperaturas se muestran en Fig. 4. Se mantiene la dependencia lineal descrita por Levich de acuerdo a la ec.

(5) con un coeficiente de correlación de entre 0.97 a 0.99 en todo el rango de temperatura estudiado, observándose un aumento de la corriente límite con la temperatura asociado principalmente a un incremento en la movilidad iónica y a una disminución de la viscosidad del electrolito soporte de acuerdo al modelo de Arrhenius verificado en Fig. 1.

2.5

IL x 10-2 (A)

2.0

5

4

1 mg/L GL 10 mg/L GL 100 mg/L GL Sin aditivos

1.5 1.0 0.5

3

0.0 0.0

-2

IL x10 (A)

a)

3.0

2.0

4.0

2

8.0

10.0

b)

3.0

1

2.5

0

5

10

15

20

2.0 IL x10-2 (A)

0

6.0 ( / s-1)1/2

( / s-1)1/2

Figura 3. Dependencia de la corriente límite con la velocidad de rotación del EDR para un electrolito de composición 40 g/L Cu2+ y 200 g/L H2SO4 en ausencia de aditivos a 60°C. Fuente: Los autores.

1 mg/L Tu 10 mg/ L Tu 100 mg/L Tu Sin aditivos

1.5 1.0 0.5 0.0 0.0

2.0

4.0

4 20 °C 40 °C 60 °C 80 °C

3

8.0

10.0

c)

3.0 2.5

2

2.0 IL x10-2 (A)

2

IL x 10 (A)

6.0 (/ s-1)1/2

1

1 mg/L Cl10 mg/L Cl100 mg/L ClSin aditivos

1.5 1.0

0

0.5

0

2

4

6

8

10

12

0.0 0.0

( / s-1)1/2

2.0

4.0

6.0

8.0

10.0

( / s-1)1/2

Figura 4. Dependencia de la corriente límite de iones cúpricos con la temperatura del electrolito de composición 40 g/L Cu2+y 200 g/L H2SO4 en ausencia de aditivos. Velocidad de rotación del EDR: 200, 400, 600, 800 y 1000 rpm.

Fuente: Los autores.

Figura 5. Efecto de la concentración de aditivos en la corriente límite en función de la velocidad de rotación en soluciones de composición 40 g/L Cu2+ y 200 g/L H2SO4 a 60ºC a) Cola (GL), b) Tiourea (Tu), c) Cloruro (Cl−).

Fuente: Los autores.

212

Araneda-Hernández et al / DYNA 81 (188), pp. 209-215. December, 2014.

Tabla 3. Efecto de la temperatura en el coeficiente de difusión del ión cúprico en solución de composición 40 g/L Cu2+ y 200 g/L H2SO4 en ausencia de aditivos.

DCu2+ Temperatura (105 cm2/s) (ºC)

0.4647 0.7030 1.1560 1.7728

-11

ln DCu2+

20 40 60 80

-10

Fuente: Los autores.

-12

-13

En el caso de soluciones con adición de ión cloruro (Fig. 5c), a baja concentración no se observan variaciones significativas en los valores de corriente límite respecto de los obtenidos para la solución en ausencia de Cl. A mayor concentración de ión cloruro se observa una disminución significativa de la corriente límite respecto de la solución base en todo el rango de velocidad de rotación considerado.

Considerando que la temperatura de la solución se mantiene fija y que el gradiente de concentración de iones cúpricos entre el seno de la solución y la interfase catódica permanece prácticamente constante, esta disminución en la corriente límite medida experimentalmente estaría asociada más bien a una disminución en el área efectiva disponible por una adsorción de los aditivos en la interfase del electrodo.

-14

0.0028

0.0032

0.0036

0.0040

-1

(T / K) Figura 6. Dependencia de ln DCu2+ vs. 1/T en solución de composición 40 g/L Cu2+ y 200 g/L H2SO4.

Fuente: Los autores.

஽ ଴

(6)

donde D es el coeficiente de difusión en cm2/s, D0 es el factor de frecuencia, ED es la energía de activación para difusión en J/mol, R es la constante de los gases (8.314 J mol/K) y T es la temperatura absoluta. A partir de esta expresión y por lo tanto, de la pendiente de la recta de la misma figura se encontró que el coeficiente de difusión del ión cúprico se relaciona con la temperatura por la expresión ln D = 2328.2 T1  4.3703, con D en cm2/s. De esta correlación se obtiene que la energía de activación por difusión es de 19.36 kJ/mol, muy cercana a 19.2 kJ/mol obtenida por Moats et al.

[14]. Un valor mayor, 26.8 kJ/mol, es el que reportan Gladysz et al.[13] para un electrolito industrial que contiene ión cloruro y varias impurezas como Ni, As, Fe, Sb y Bi. Considerando la variación de la viscosidad cinemática del electrolito con la temperatura, la energía de activación efectiva para un flujo laminar Elam 10] en un disco rotatorio (Reynolds del orden de 104) expresada en términos de la energía de activación para el coeficiente de difusión ED y la viscosidad cinemática Eν está dada por:

3.3. Coeficiente de difusión del ión cúprico 3.3.1. Efecto de la temperatura A partir de la pendiente de la gráfica de Fig. 3 y de la expresión de Levich (ec. 5) se determinó el coeficiente de difusión del ión cúprico para cuatro valores de temperatura del electrolito considerando en este cálculo la variación de la viscosidad cinemática medida en función de la temperatura. Los resultados obtenidos para el coeficiente de difusión se muestran en Tabla 3 siguiente.

Estos resultados concuerdan con los reportados en estudios previos [1-3] para similar composición de cobre y rango de temperatura utilizando en algunos casos otras técnicas electroquímicas.

Para un electrolito de similar composición Gladysz et al. [13] utilizando una configuración de dos electrodos, un disco de oro (ultramicroelectrodo) como electrodo de trabajo y una placa de cobre como electrodo auxiliar obtienen un coeficiente de difusión de 0.70010−5 cm2/s para 36.9 g/L Cu2+ a 40ºC muy cercano a 0.70310-5 cm2/s que es el obtenido en este estudio a la misma temperatura.

Moats et al. [14] reportan un coeficiente de difusión a 65ºC para un electrolito de 40 g/L Cu, y 160 g/L H2SO4 de 1.2310-5 cm2/s, levemente superior a 1.16 10-5 cm2/s obtenido a 60ºC en este estudio. Esta diferencia estaría asociada principalmente a las diferentes temperaturas entre ambas experiencias, que como se sabe tiene un efecto significativo en la difusividad de especies iónicas.

El coeficiente de difusión del ión cúprico de acuerdo a la correlación de Levich en función de T1, se muestra en Fig. 6 el que exhibe un comportamiento lineal asimilable al de tipo Arrhenius con la temperatura de acuerdo a:

஽ ௟௔௠

ఔ

(6)

donde ED y Eν corresponden a la energía de activación para difusión y para la viscosidad cinemática, respectivamente.

El valor de Eν fue de 11.6 kJ/mol y se determinó a partir de la correlación obtenida de la gráfica de Fig. 1 (ec. 4) con un coeficiente de correlación de 0.99, resultando en un valor para la energía de activación Elam de 14.8 kJ/mol, levemente inferior a 15.3 kJ/mol, obtenido por Moats et al.

[14] quienes estimaron la viscosidad cinemática a partir de la correlación empírica obtenida por Price y Davenport [15] que considera la presencia de impurezas en solución provenientes de la disolución anódica, las que no fueron abordadas en este estudio.

213

Araneda-Hernández et al / DYNA 81 (188), pp. 209-215. December, 2014.

Tabla 4.

Efecto de la adición de aditivos en el coeficiente de difusión del ión cúprico a 60°C para un electrolito de composición base 40 g/L Cu2+ y 200 g/L H2SO4.

Concentración DCu2+ Aditivo aditivo 5 cm2/s) (10 (mg/L) Sin aditivo 0 1.1560 Cola 1 0.933 10 0.957 100 0.949 Tiourea 1 0.886 10 1.002 100 0.943 Cloruro 1 1.155 10 0.771 100 0.755 Fuente: Los autores.

4. Conclusiones Se verifica una disminución de la viscosidad cinemática del electrolito con un aumento en la temperatura, en el rango 20 a 80°C, en ausencia de aditivos. Esta dependencia de tipo Arrhenius está asociada principalmente a un incremento en la movilidad iónica en el electrolito soporte.

Se constató una disminución en la corriente límite en el electrolito de composición 40 g/L de Cu2+ y 200 g/L de H2SO4 a 60°C con la presencia de aditivos en solución, en el rango de concentración estudiado, sin embargo, en el caso de la cola y la tiourea, esta disminución es prácticamente constante para los tres niveles de concentración utilizados 1, 10 y 100 mg/L. Para el ión cloruro, esta disminución se observó a concentraciones de 10 y 100 mg/L. Utilizando el criterio de Levich, se obtuvo una correlación empírica para el coeficiente de difusión del ión cúprico en solución de composición base (40 g/L de Cu2+ y 200 g/L de H2SO4) en ausencia de aditivos, en función de la temperatura de la solución dada por la expresión ln D = 2328.2 (1/T)  4.3703, con D en cm2/s, en el rango de temperatura estudiado. De acuerdo a esta correlación, en una solución de composición base a 60°C y en ausencia de aditivos, el coeficiente de difusión del ión cúprico es de 1.15610−5 cm2/s.

La adición de cola y tiourea, en soluciones de composición base a 60°C resulta en una leve disminución del coeficiente de difusión del ión cúprico. El mismo efecto se observó en soluciones en presencia de cloruro pero a concentración de 10 y 100 mg/L. El efecto de tiourea y cola sobre la disminución del coeficiente de difusión de ión cúprico en la interface cátodo-solución se originaría en un mecanismo de bloqueo de sitios activos en presencia de moléculas absorbidas en la superficie del electrodo. Para el cloruro, en cambio, resultaría de la formación de un film de cloruro cuproso a una concentración crítica de cloruro en solución, la que en este estudio sería cercana a 10 mg/L.

3.1.2. Efecto de los aditivos Los resultados del coeficiente de difusión del ión cúprico en presencia de aditivos a diferentes concentraciones se muestran en Tabla 4 siguiente: A baja concentración de cola y tiourea (1 mg/L) el coeficiente de difusión del ión cúprico disminuye levemente respecto del observado en la solución de composición base en 18.7% y en 23.4% en presencia de cola y tiourea, respectivamente. Al aumentar la concentración a 10 y 100 mg/L para ambos aditivos, el coeficiente de difusión se mantiene prácticamente constante. Sin embargo, si se considera que estos dos aditivos actúan mediante un mecanismo de adsorción sobre la superficie catódica [16-19], la presencia de moléculas adsorbidas disminuiría el área efectiva disponible para la semi reacción de reducción en concordancia con la disminución observada en la corriente límite en presencia de estos aditivos sumando a esto el nulo efecto de la incorporación de aditivos en la movilidad iónica del electrolito.

En el caso del ión cloruro, a bajas concentraciones (1 mg/L) el coeficiente de difusión es muy cercano al de la solución base a la misma temperatura, sin embargo, al aumentar la concentración de cloruro a niveles de 10 mg/L, el valor del coeficiente de difusión disminuye significativamente en 34.7% respecto del obtenido en soluciones en ausencia de aditivos, manteniéndose constante a aumentos posteriores de 10 y 100 mg/L. Esta disminución en el coeficiente de difusión podría estar asociada a la formación de un film de cloruro cuproso en las vecindades del cátodo. El CuCl presenta una constante de equilibrio a 60°C de 1.14106 y considerando el equilibrio: Cu2+ + Cu = Cu+

K= 2.259×10-5

Referencias [1]

[2] [3]

(7)

[4]

2+

que para una concentración de Cu de 40 g/L resulta en una concentración de equilibrio de cloruro de aproximadamente 8.2 mg/L. Dado el alto valor de la constante de equilibrio del CuCl, un aumento en la concentración de Cl promueve la reducción de Cu2+ a Cu+ por sobre la reducción directa de Cu2+ a Cu metálico.

[5] [6]

214

Alodan, M.A. and Smyrl, W.H.. Confocal laser scanning microscopy, electrochemistry, and quartz crystal microbalance studies of leveling effects of thiourea on copper deposition. Journal of the Electrochemical Society, 145 (3), pp. 957-963, 1998. http://dx.doi.org/10.1149/1.1838372 Ke, B., Hoekstra, J.J., Sison, B.C. and Trivich, D, Role of thiourea in the electrodeposition of copper. Journal of the Electrochemical Society, 106 (12), pp. 1081-1082, 1959.

Sun, M. and O'Keefe, T.J., The effect of additives on the nucleation and growth of copper onto stainless steel cathodes.

Metallurgical and Materials Transactions B, 23B (5), pp. 591-599, 1992.

Arango, L., Carmona, J., Vallejo, E. y Hoyos, B., Efecto de aditivos en la dureza de los depósitos de níquel utilizando corriente pulsante. DYNA, 73 (150), pp. 59-66, 2006.

Quickenden, T.I. and Jiang, X., The diffusion coefficient of sulphate in aqueous solution. Electrochimica Acta, 29 (6), pp. 693700, 1984. http://dx.doi.org/10.1016/0013-4686(84)80002-X Hinatsu, J.T. and Foulkes, F.R., Diffusion coefficients for copper (II) in aqueous cupric sulfate-sulfuric acid solutions. J.

Araneda-Hernández et al / DYNA 81 (188), pp. 209-215. December, 2014.

[7]

[8]

[9] [10] [11] [12] [13]

[14]

[15]

[16]

[17]

[18]

[19]

Electrochem.

Soc., 136 (1), pp.

125-132, 1989. http://dx.doi.org/10.1149/1.2096571 Machardy, S.J. and Janssen, L.J.J., The diffusion coefficient of Cu(II) ions in sulfuric acid-aqueous and methanesulfonic acidmethanol solutions. Journal of Applied Electrochemistry, 34 (2), pp.

169-174, 2004. http://dx.doi.org/10.1023/B:JACH.0000009956.75577.ef Cifuentes, G., Vargas, C. and Simpson, J., Analysis of the main variables that influence in the cathodic rejection in copper electrorefining, Proceedings of The Seminar First Meeting on Minor Contaminants in Copper Metallurgy. Concepción, Chile, pp.

37-43. October 25-26, 2007.

Biswas, A.K. and Davenport, W.G., Extractive metallurgy of copper. Fourth Edition, Pergamon Press, 2002.

Levich, V.G., Physicochemical hydrodynamics, Prentice Hall, Inc. Englewood Cliffs, N.J., 1962.

Price, D.C. and Davenport, W.G., Physico-chemical properties of copper electrorefining and electrowinning electrolytes, Metallurgical Transactions B, 12B, pp. 639-643, 1981.

Subbaiah, T. and Das, S.C., Physico-chemical properties of copper electrolytes, Metallurgical Transactions B, 20B (3), pp. 375-380, 1989.

Gladysz. O., Los, P. and Krzyzak, E., Influence of concentrations of copper, levelling agents and temperature on the diffusion coefficient of cupric ions in industrial electro-refining electrolytes.

Journal Applied Electrochemistry, 37 (10), pp. 1093-1097, 2007. http://dx.doi.org/10.1007/s10800-007-9363-8 Moats, M.S., Hiskey, J.B. and Collins, D.W., The effect of copper, acid, and temperature on the diffusion coefficient of cupric ions simulated electrorefining electrolytes. Hydrometallurgy, 56 (3), pp.

255-268, 2000.

Price, D. and Davenport, W., Densities, electrical conductivities and viscosities of CuSO4/H2SO4 solutions in the range of modern electrorefining and electrowinning electrolytes. Metallurgical Transactions B, 11B (1), pp. 159-163, 1980.

Quinet, M., Lallemand, F., Ricq, L., Hihn, J. and Delobelle, P., Adsorption of thiourea on polycrystalline platinum. Surface & Coatings Technology, 204 (20), pp. 3108-3117, 2010. http://dx.doi.org/10.1016/j.surfcoat.2010.01.025 Tarallo, A. and Heerman, L., Influence of thiourea on the nucleation of copper on polycrystalline platinum. Journal of Applied Electrochemistry, 29 (5), pp. 585-591, 1999. http://dx.doi.org/10.1023/A:1003410720266 Tantavichet, N., Damronglerd, S. and Chailapakul, O., Influence of the interaction between chloride and thiourea on copper electrodeposition. Electrochimica Acta, 55 (1), pp. 240-249, 2009. http://dx.doi.org/10.1016/j.electacta.2009.08.045 Bletcha, V.K., Wang, Z.Z. and Krueger, D.W., Glue analysis and behavior in copper electrolyte, Metallurgical Transactions B, 24B (2), pp. 277-287, 1993.

Área Curricular de Ingeniería Geológica e Ingeniería de Minas y Metalurgia Oferta de Posgrados    

Especialización en Materiales y Procesos Maestría en Ingeniería - Materiales y Procesos Maestría en Ingeniería - Recursos Minerales Doctorado en Ingeniería - Ciencia y Tecnología de Materiales

Mayor información: Néstor Ricardo Rojas Reyes Director de Área curricular acgeomin_med@unal.edu.co (57-4) 425 53 68

E. Araneda, recibió su título en Ingeniería Metalúrgica en 1998 y es candidata al grado de Doctor en Ingeniería Metalúrgica en la Universidad de Concepción, Chile. Desde el año 2001 se desempeña como Investigador y Colaborador Académico en el Departamento de Ingeniería Metalúrgica de la Universidad de Concepción, participando en diversos proyectos de investigación y desarrollo en minería con énfasis en las áreas de electroquímica y caracterización avanzada de materiales.

F. Vergara, recibió su título de Ingeniero Metalúrgico en 1975 y en 1978, el grado de Doctor de la Université Scientifique et Medicale de Grenoble, Francia. Desde el año 1978 se desempeña como Profesor en el Departamento de Ingeniería Metalúrgica realizando docencia e investigación en el área electrometalurgia. Actualmente es Director del mismo departamento.

A. Pagliero, recibió su título de Ingeniero Químico en 1972 y en 1976, el grado de Doctor del Institut National Polytechnique de Grenoble, Francia.

Desde el año 1972 se desempeña como Profesor en el Departamento de Ingeniería Metalúrgica realizando docencia de pre y postgrado e investigación en el área electroquímica y electrometalurgia del cobre.

215

Influence of silicon on wear behaviour of “Silal” cast irons Ana I. García-Diez a, Carolina Camba-Fabal b, Ángel Varela-Lafuente c, Víctor Blázquez-Martínez d, José Luís Mier-Buenhombre e & Benito Del Río-López f a

Escuela Politécnica Superior, Universidade da Coruña, A Coruña, España. agd@cdf.udc.es Escuela Politécnica Superior, Universidade da Coruña, A Coruña, España. ccamba@.udc.es c Escuela Politécnica Superior, Universidade da Coruña, A Coruña, España. anvalaco@cdf.udc.es d Escuela Técnica Superior de Ingenieros Industriales, Universidad Politécnica de Madrid, Madrid, España. vblazquez@etsii.upm.es e Escuela Politécnica Superior, Universidade da Coruña, A Coruña, España. jlmier@.udc.es f Escuela Técnica Superior de Ingenieros Industriales, Universidad Politécnica de Madrid, Madrid, España. bdelrio@etsii.upm.es b

Received: January 29th, 2014. Received in revised form: July 29th, 2014. Accepted: August 5th, 2014.

Abstract The tribological and mechanical properties of various cast irons with different contents of carbon and silicon are studied. The Brinell hardness or every cast iron was measured and their wear resistance was calculated according to ASTM G99 standard, "Standard Test Method for Wear Testing with a Pin-on-Disk Apparatus. These results were related to microstructure, composition and casting processes of the cast irons. The best wear behaviour was experienced by the cast irons with spheroidal graphite and a ferrite matrix in their microstructure. At the same time, the cast irons with the lowest wear resistance are those that have flaky graphite and a purely ferrite matrix in their microstructure. In addition, various casting processes have been tested in order to determine their influence on the appearance of spheroidal graphite.

Keywords: High-alloy graphitic irons; silicon; wear.

Influencia del silicio en el comportamiento al desgaste de las fundiciones tipo “Silal” Resumen En este trabajo se ha evaluado la influencia del contenido de silicio en la resistencia frente al desgaste y en las propiedades mecánicas de diferentes fundiciones con alto contenido en silicio del tipo Silal. Para ello se ha variado el contenido en carbono y en silicio de las muestras, evaluando posteriormente la modificación en la dureza, determinada mediante la escala Brinell, y en el comportamiento tribológico, estudiado mediante ensayos de desgaste normalizados Pin-on-Disk. Todo ello se ha complementado con un estudio metalográfico para determinar la influencia de la microestructura en las propiedades analizadas. Se ha determinado que el mejor comportamiento al desgaste lo presentan aquellas aleaciones cuya composición y proceso de colada garantizan el grafito esferoidal frente a las que lo presentan con estructura laminar. Adicionalmente, se ha experimentado con diferentes procesos de colada para la fabricación de las fundiciones para determinar su influencia en la aparición del grafito esferoidal.

Palabras clave: Fundiciones grises de alta aleación; silicio; desgaste.

1. Introduction Among cast irons, there are ferritic cast irons with medium and high silicon content, known as Duriron and Silal, respectively. As well as being resistant to heat, the latter does not experience swelling when it is heated repeatedly or subjected to temperatures as high as 850ºC. A point to bear in mind is that it swells during casting because the graphite flakes are porous. These pores let oxidant gases

pass so that they themselves are oxidized internally.

Consequently, if it is heated beyond eutectic transformation temperature, the percentage of free graphite increases. This in turn may lead to a significant increase in its volume when ordinary grey cast irons are used.

Silal-type cast irons present a ferrite microstructure with graphite that is stable up to eutectic transformation temperature. Consequently, its capacity to resist oxidation at temperatures of 850ºC is very high and the problem of

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 216-221. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41809

García-Diez et al / DYNA 81 (188), pp. 216-221. December, 2014.

swelling is avoided. If these cast irons present spheroidal graphite, their behaviour can be improved even more [1].

On the whole, its composition has a silicon content between 5.5 and 7%, while the carbon content usually does not exceed 2.3%. The silicon dissolved in the ferrite weakens it so that its toughness and mechanical resistance are low. It has little resistance to thermal shock, although it performs better than grey cast irons.

Its mechanical properties and resistance to oxidation can be improved by adding magnesium to spheroidize the graphite. This step raises its tensile resistance by 20% while greatly improving its toughness. A further addition of molybdenum boosts its creep resistance. These cast irons tend to be used for oven racks or ashtrays, along with steam turbine components for continuous or intermittent work at temperatures ranging from 650 to 825ºC. By partially substituting silicon for aluminum in percentages of between 4 and 6%, it is possible to make it more resistant to oxidization at high temperatures and increase its mechanical strength. However, alloys with aluminum are difficult to produce because that metal has a strong oxygen affinity [2-8].

The silicon content for the cast iron Duriron is higher than that of Silal: between 12 and 16%. This high content means that it has a ferrite microstructure due to the alphagenic characteristics of the silicon. It also means its carbon content is lower than that of Silal. It is extremely resistant to dry corrosion from acid oxidants but it is much more fragile than Silal. Very hard and difficult to manipulate mechanically, Duriron is also inferior to Silal in terms of its tensile strength and thermal shock resistance.

Resistance to corrosion is improved with the addition of molybdenum, but this is not the case with its mechanical properties. Duriron is used to make components that have to withstand electrochemical corrosion [9-15].

The performance of this type of casts iron on corrosion and high temperature conditions is well known, but there is little information about their tribological behaviour.

Therefore, the purpose of this work is to evaluate their response to wear taking into account the composition and microstructure, in particular the presence of spheroidal graphite. Furthermore, different casting processes of these alloys have been tested to determine the best way to get this kind of graphite.

2. Experimental methods This study was conducted on ten distinct castings prepared in the Pilot Foundry Plant at the Escuela Técnica Superior de Ingenieros Industriales of the Universidad Politécnica in Madrid. In each of these, the carbon and silicon contents were varied. The carbon content was varied between 1.5 and 2.7%, where as the silicon content was between 3.3 and 9.3%. The influence of silicon content on wear was evaluated in a wider range than usual composition for these alloys while the content of carbon was maintained within the usual range. The casting process was also changed from one alloy to another. Nickel magnesium was added in all the cases except in Casting 4 in which magnesium chloride was added. The amount of nickel

magnesium in Casting 1 was five times less than in all other cases. The amount of Ni-Mg added was 2.7 % of the total weight of the castings in all cases except in the case of Casting 1, which was 0.5 %. On the other hand, the amount of MgCl2 added in the Casting 4 was 3.7 %. For all ten, the carbon and silicon contents were determined. The former was analyzed with a Leco CS-300 device, while the procedure from UNE 7-028-75 “Determining silicon gravimetric analysis in silicon and cast irons” [16] was followed to test the silicon content. Moreover, the hardness was tested for all ten, applying the Brinell scale and using a Hoytom tester with 2.5 mm diameter ballpoint indenter and an applied load of 187.5 kg [17].

Pin-on-disk tests were carried out to measure the wear resistance for each according to the mass loss caused by the casting process. These were informed by the procedure in ASTM G99 “Standard test method for wear testing with a pin-on-disk apparatus” [18]. Before doing these tests, it was necessary to prepare the surface of each sample so that the roughness was more uniform among all ten cast irons. By decreasing the roughness to minimum levels, the surface finish would not influence the results of the wear tests. With this surface preparation, the average roughness Ra in all ten samples was below 1.5μm. The samples were collected from the central part of the casting specimens, disposing the ends as they are the zones most susceptible to heterogeneities. A profilometry study of the samples surface was performed by making multiple passes in different directions. This permitted to discard the presence of porosity and surface defects that alter the results.

The following parameters were used. The tungsten carbide pin had a 4 mm diameter and a hardness of 75HRC. The samples were subjected to a load of 10N and each test included 1000 m of linear wear at a speed of 0.25m/s. The track radius on the surface was 8 mm. All tests were done at ambient temperature. The friction coefficient was also determined throughout the testing.

A minimum of three valid tests were performed for each simple. The Pearson coefficient was used to assess the accuracy of the results. This coefficient determines the relationship between the deviation of data and their mean.

Values of this coefficient higher than 0.20 were discarded.

Once the preparatory steps had been taken, the metalographic study of each cast iron was performed with an etching agent made up of 70% alcohol, 20% de nitric acid and 10% hydrofluoric acid, and an immersion time of five seconds per sample. Micrographies were taken of all ten samples.

3. Results Table 1 offers data on the carbon and silicon content for each sample.

In Table 1 it is possible to observe that the carbon content in the samples falls between 1.56 and 2.71%. The lowest figure corresponds with Cast Iron 5, while highest is for Sample 4. In terms of the silicon, the values have an interval between 3.35 and 9.11%. At the lowest end of the scale is Sample 6 and, at the other end, is Sample 9.

217

García-Diez et al / DYNA 81 (188), pp. 216-221. December, 2014.

boundary. It is the only cast iron that shows this graphite form. Its matrix is entirely ferrite like in Cast Iron 4. The microstructure of Cast Iron 6 is shown in Fig. 2-below. It consists of flaky graphite, although in a smaller amount than in Cast Iron 4. Once again, it has an entirely ferritic matrix.

Fig. 3-top left shows the microstructure of Cast Iron 7.

A spheroidal graphite with needle patterns- the more widespread characteristic- can be observed. Its matrix is ferrite.

As for Cast Iron 8, Fig. 3-top right presents a microstructure formed by spheroidal graphite; it is flaky over a ferrite matrix. In this sample, the presence of graphite is lower than in the previous case. At the same time, Fig. 3below is the microstructure of Cast Iron 9. Spheroidal graphite appears, with a martensite and austenite matrix.

The last one is Fig. 4, with the microstructure of Cast Iron 10, composed of spheroidal graphite and a ferrite matrix. This microstructure is similar to that of the Casting 1 and Casting 2 with the difference that the matrix is entirely ferritic, without the presence of pearlite.

Table 1.

Carbon and silicon content for each cast iron Sample Nº 1 Nº 2 Nº 3 Nº 4 Nº 5 Nº 6 Nº 7 Nº 8 Nº 9 Nº 10 Source: The authors

C [%]

Si [%]

2.05 2.03 2.02 2.71 1.56 2.33 2.09 2.31 1.82 2.39

6.22 4.66 6.62 5.29 6.76 3.35 6.92 5.94 9.11 5.33

Figure 1. Microstructure of Cast Iron 1(top left), Cast Iron 2 (top right) and Cast Iron 3 (below) X200.

Source: The authors

Fig. 1 shows the micrographies of Cast Iron 1, 2 and 3.

For the Cast Iron 1, it is possible to notice that the microstructure consists in spheroidal graphite (despite five times less nickel magnesium was used in casting) on a ferrite-pearlite matrix.

Cast Iron 2 shows spheroidal graphite that is less perfect in shape than the one in Cast Iron 1. In this case, graphite is distributed in a large amount of particles with widely varying sizes. That can be explained by the less silicon content (around 1.8% lower) when carbon content is constant. This is another case of the matrix becoming ferrite-pearlite, with the latter in lesser quantities The microstructure for Cast Iron 3, also has spheroidal graphite and a ferrite-pearlite matrix.

Fig. 2-top left corresponds to the microstructure of Cast Iron 4. In this case, the graphite is not spheroidal, but rather flaky. Its matrix is entirely ferrite and not ferrite-pearlite, as in the previous cases. It is the only of the ten analyzed cast irons that shows entirely laminar graphite. This means that the casting process with magnesium chloride addition does not achieve the objective sought: the graphite spheroidization. Fig. 2-top right shows the microstructure of Cast Iron 5, with interdendritic graphite along the grain

Figure 2. Microstructure for Cast Iron 4 (top left), Cast Iron 5 (top right) and Cast Iron 6 (below) X200.

Source: The authors

Figure 3. Microstructure for Cast Iron 7 (top left), Cast Iron 8 (top right) and Cast Iron 9 (below) X200.

Source: The authors

218

García-Diez et al / DYNA 81 (188), pp. 216-221. December, 2014.

Iron 5.

Fig. 5 displays the friction coefficient patterns obtained during the pin-on-disk wear tests carried out on the ten cast irons.

In the first, the patterns correspond with Cast Irons 1, 2, 8 and 10. It can be observed the coefficient of friction increases gradually during the test and stabilizes toward the end. The second portrays the friction coefficient throughout the wear test of Cast Irons 3, 5 and 9. The friction coefficient progressively increases during the initial third of the test and remains constant for the other two thirds. This pattern encompasses the highest friction coefficient values for the ten cast irons analyzed.

Figure 4. Microstructure for Cast Iron 10. X200.

Source: The authors

Table 2.

Hardness values for the cast irons Sample Nº 1 Nº 2 Nº 3 Nº 4 Nº 5 Nº 6 Nº 7 Nº 8 Nº 9 Nº 10 Source: The authors

Table 3.

Wear test results for cast irons Sample Mass loss [mg] Nº 1 5.39 Nº 2 3.00 Nº 3 3.06 Nº 4 6.13 Nº 5 1.87 Nº 6 54.25 Nº 7 66.22 Nº 8 13.26 Nº 9 10.64 Nº 10 21.63 Source: The authors

HB 255 252 285 136 237 229 205 234 329 269

Pearson coefficient 0.06 0.07 0.13 0.12 0.02 0.06 0.06 0.13 0.13 0.17

Friction coefficient 0.49 0.58 0.62 0.46 0.66 0.38 0.35 0.56 0.58 0.53

The last one is Figure 4, with the microstructure of Cast Iron 10, composed of spheroidal graphite and a ferrite matrix. This microstructure is similar to that of the Casting 1 and Casting 2 with the difference that the matrix is entirely ferritic, without the presence of pearlite.

Table 2 shows the hardness measurements for the different cast irons using the Brinell scale. The value shown is the average value of ten valid tests.

The hardness values range between 136 and 329 HB; the lowest results correspond with Cast Iron 4 and the highest, with Cast Iron 9.

The wear tests results of all samples can be seen in Table 3. The table represents wear resistance as mass loss expressed in milligrams, the variation coefficients for mass loss and the average friction coefficients in each test.

In Table 3 the mass loss values fall between 1.87 and 66.22 mg. The lowest value corresponds to Cast Iron 5, which also has the highest friction coefficient. The second lowest value relates to the mass loss Cast Iron 7; its friction coefficient was significantly lower than in the case of Cast

Figure 4. Microstructure for Cast Iron 10. X200.

Source: The authors

219

García-Diez et al / DYNA 81 (188), pp. 216-221. December, 2014.

The third has the friction coefficient pattern for Cast Irons 4 and 6. At the start of the testing, it grows rapidly, but remains steady thereafter. The friction coefficient values are the lowest for the ten samples. The behaviour pattern of the friction coefficient of Cast Iron 7 throughout the wear test can be seen in the last one, the fourth one. This coefficient rises sharply at the outset of the testing only to drop dramatically at its midpoint and then remain steady for the rest of the time. After all of the results were analyzed, the influence of the chemical composition on the microstructure and the tribological properties of the material can be evaluated. The measured friction coefficient is higher in Casting 2 than Casting 1, whereas the weight loss in the wear test is higher in Casting 1 and the hardness is similar in both alloys. This can be explained by the distribution of the graphite particles which came off more easily in Casting 1, leading to a friction coefficient decrease (The detached graphite acts as a lubricant), while wear weight loss increases. Casting 3 microstructure is similar to Casting 1, however it has considerably higher hardness and better wear behaviour. The Casting 4 hardness is clearly the lowest of all the alloys tested, and the weight loss during wear is one of the highest along with Casting 6 which also shows laminar graphite. Furthermore, the friction coefficient of both cast irons follows the same pattern (third curve of the Fig. 5). This is a consequence of the graphite detachment which decreases the friction coefficient due to its lubricant properties.

Casting 5 is the only one that shows interdendritic graphite which is more difficult to come off and negatively affects to the lubricant properties of the alloy. This explains the high value of the friction coefficient Casting 7 is the second in silicon content, whereas its carbon content is similar to those of Casting 1, Casting 2 and Casting 3. However, Casting 7 shows graphite mostly in needle form and in spheroidal form in much less extent. The friction coefficient pattern during the wear test is unique and that indicates a large mass detachment at the beginning of the test which acts as a lubricant and gives the lowest friction coefficient of all, but this does not prevent the wear of the ferritic matrix Casting 8 has one of the highest carbon content whereas its silicon content is medium. The microstructure is similar to those of Casting 1 and Casting 2 except that Casting 8 shows graphite flakes in small quantities and, therefore, its hardness is significantly lower than the other two. The same happens with its wear resistance.

Despite the low carbon content, Casting 9 has the highest hardness due to the presence of martensite in an austenitic matrix. However, the mass loss values are higher than expected. The only explanation would be that the martensite needles were detached acting as abrasive particles during testing.

The Casting 10 microstructure is similar to those of Casting 1 a Casting 2 with the difference that the first one shows a completely ferritic matrix and the other two have a ferritic-perlitic structure. This last matrix is softer and gives a higher weight loss during wear because resists less abrasion and favours the graphite detachment [19].

An increase in hardness leads to a reduction in the mass

loss during the wear test. There are some exceptions, all of these related to the cast irons that have a purely ferrite matrix in their microstructure.

4. Conclusions The influence of carbon and silicon content, microstructure and Brinell hardness on wear behaviour of ten Silal-type cast irons is studied. Ni-Mg was added in nine of these cast iron, while only one was made from magnesium chloride.

The casting process with magnesium chloride is not effective to obtain spheroidal graphite and this implies worse results in wear tests.

The best wear behaviour was obtained in the cast irons with spheroidal graphite and ferrite matrix in their microstructure, and with medium silicon and carbon contents. The exception is cast iron 5 (with the lowest carbon content of this group), which is the only one with interdendritic graphite and experiences a weight loss slightly lower than the others. At the same time, the cast irons with the lowest wear resistance are those that have flaky graphite and a purely ferrite matrix in their microstructure. One of them is made with MgCl2 and the other one has a lower silicon content.

It has been possible to prove that, in general, while the carbon content in these kinds of cast irons increases, so does the mass loss.

References [1]

Gonzaga-Cinco; R. and Fernández-Carrasquilla, J., Dependencia de las propiedades mecánicas y de la composición química en la fundición de grafito esferoidal, Rev. Metal. Madrid, 42 (2) pp. 91102, 2006.

[2] Pearce, J., Inoculation of cast irons: Practices and developments, Metal Casting Technologies, 53, pp. 16-22, 2007.

[3] Harvey, J.N. and Noble, G.A., Inoculation of cast irons-an overview.

55th Indian Foundry Congress, India, pp. 343-367, 2007.

[4] Loizaga, A., Sertucha, J. and Suárez, R., Influencia de los tratamientos realizados con diferentes ferroaleaciones de magnesio en la evolución de la calidad metalúrgica y los procesos de solidificación de las fundiciones esferoidales, Rev Metal Madrid, 44 (5) pp. 432-446, 2008.

[5] Skjegstad, N.T. and Skaland, T., Inoculation of grey and ductile iron. Bombay Foundry Congress. Bombay, India. pp. 1-23, 1996.

[6] Liu, S.L., Loper Jr., C.R. and Witter, T.H., The role of graphitic inoculants in ductile iron. AFS Transactions, 100, pp. 899-906, 1992.

[7] Subramanian, S.V., Kay, D.A.R. and Purdy, G.R., Compacted graphite morphology control. AFS Transactions, 90, pp. 589-603, 1982.

[8] Riposan, I., Chisamera, M., Stan, S. and Skaland, T., A new approach to graphite nucleation mechanism in gray iron.

Proceedings of the AFS Cast Iron Inoculation Conference. Illinois, EEUU, pp. 31-41, 2005.

[9] Lu, Z.L., Zhou, Y.X., Rao, Q.Ch. and Jin, Z.H. An investigation of the abrasive wear behaviour of ductile cast iron. J Mater Process Tech, 116 (2-3) pp.

176-181, 2001. http://dx.doi.org/10.1016/S0924-0136(01)01013-5 [10] Hirasata, K., Hayashi, K. and Matsunami, H., Friction and wear of spheroidal graphite cast iron under severe sliding conditions. 29th Leeds-Lyon Symposium on Tribology. Lyon, Francia, 41, pp. 643652, 2003.

220

García-Diez et al / DYNA 81 (188), pp. 216-221. December, 2014.

[11] Takeuchi, E., The mechanism of wear of spheroidal graphite cast iron in dry sliding, Wear, 19, pp. 267-276, 1972. http://dx.doi.org/10.1016/0043-1648(72)90119-6 [12] Lu, Z.L., Zhou, Y.X., Rao, Q. Ch. and Jin, Z.H., An investigation of the abrasive behavior of ductile cast iron. J Mater Process Tech, 116 (2-3) pp. 176-181, 2001. http://dx.doi.org/10.1016/S09240136(01)01013-5 [13] Shin, J.S., Lee, S.M., Moon, B.M., Lee, H.M., Lee, T.D. and Lee, Z.H., The effects of heat treatment condition and Si distribution on order-disorder transition in high Si steels, Met Mater Int, 10 (6) pp.

581-587, 2004. http://dx.doi.org/10.1007/BF03027422 [14] Blázquez-Martínez, V., Metalografía de las aleaciones férreas, E.T.S.I.I., Madrid, 1991.

[15] Castillo, R., Bermont, V. and Martínez, V., Relations between microstructure and mechanical properties in ductile cast irons: A rewiew. Rev Metal Madrid, 35, pp. 329-334, 1999. http://dx.doi.org/10.3989/revmetalm.1999.v35.i5.641 [16] Norma UNE 7-028, Determinación gravimétrica de silicio en aceros y fundiciones, 1975.

[17] Norma UNE-EN ISO 6506-1, Materiales metálicos. Ensayo de dureza Brinell. Parte 1: Método de ensayo, 2005.

[18] ASTM Standard G99, Standard test method for wear testing with a pinon-disk apparatus, ASTM International, West Conshohocken, PA, 2008.

[19] Sierra, H., Vélez, J. and Herrera, C., Resistencia a la abrasión de fundición nodular aleada con cobre, austemperada a 300 ºC. DYNA, 137, pp. 51-59, 2002.

received his PhD degree from the University of A Coruña, Spain in 2013. He is currently working as assistant lecturer in the field of Materials Science and Metallurgical Engineering at Engineering from Polytechnic University of Madrid. His research activity mainly concerns wear of ferrous alloys.

A.I. García-Diez, received her PhD in Industrial Engineering from Coruña University, Spain, in 2004 about abrasion-resistant ferrous alloys for use as carbon mills coatings. She is currently working as a lecturer at the Department of Industrial Engineering of Coruna University, Spain and is also member of the Materials Science and Engineering research group. Her research interests deal with the abrasive wear behavior of cast iron, manganese steel and copper. She has published about forty papers mainly on tribology of ferrous materials

Oferta de Posgrados

C. Camba-Fabal, studied at the University of A Coruña, Spain, graduating in Industrial Engineering within the Materials Science specialty in 2007.

She received her PhD degree from the same university in 2011. She is currently working as assistant lecturer in the field of Materials Science and Metallurgical Engineering at Higher Polytechnic School of Ferrol. Her research activity mainly concerns wear of ferrous alloys.

A.Varela-Lafuente, is MSc and PhD in Industrial Engineering from Polytechnic University of Madrid, Spain. He was Director of the Polytechnic High School at Coruna University, Spain from 2001 to 2005 Currently he is a lecturer at the same center He is member of the Materials Science and Engineering research group. He has many years of working experience in the field of metallurgy and his main scientific interests include the influence of the thermal and thermochemical treatments on the wear response of ferrous alloys on what he has published several papers during last years. He is member of the Spanish Mechanical Engineering Association V. Blázquez-Martínez, is MSc and PhD in Industrial Engineering from Polytechnic University of Madrid. Currently he is a lecturer at the same center. He has many years of working experience in the field of metallurgy and his main scientific interests include the influence of the thermal and thermochemical treatments on the wear response of ferrous alloys on what he has published several papers during last years.

J.L. Mier-Buenhombre, is a lecturer and Head of the Materials Science and Engineering research group (CIM) at the Department of Industrial Engineering of the University of Coruna, Spain. He received his PhD in Metallurgy from Complutense University (UCM), Spain, in 1993. He achieved an award from UCM for his PhD thesis in 1995. He has published 40 publications related to extractive metallurgy and wear behavior of metallic materials, many of them in international publications. He is a member of the Spanish Mechanical Engineering Association.

B. Del Río-López, studied at the Polytechnic University of Madrid, Spain, graduating in Industrial Engineering within the Metallurgical specialty. He 221

Área Curricular de Ingeniería Geológica e Ingeniería de Minas y Metalurgia

   

Especialización en Materiales y Procesos Maestría en Ingeniería - Materiales y Procesos Maestría en Ingeniería - Recursos Minerales Doctorado en Ingeniería - Ciencia y Tecnología de Materiales Mayor información: Néstor Ricardo Rojas Reyes Director de Área curricular acgeomin_med@unal.edu.co (57-4) 425 53 68

Effect of cationic polyelectrolytes addition in cement cohesion Edison Albert Zuluaga-Hernández a & Bibian A. Hoyos b a

Facultad de Minas, Universidad Nacional de Colombia, Medellín, Colombia. eazuluagh@unal.edu.co Facultad de Minas, Universidad Nacional de Colombia, Medellín, Colombia. bahoyos@unal.edu.co

b

Received: January 30th, 2014. Received in revised form: July 30th, 2014. Accepted: August 19th, 2014

Abstract Here is studied the variation in cohesion of cement main phase (C-S-H) as a result of cationic polyelectrolytes addition (quaternary amines spermine and norspermidine). Cohesion study was carried out by molecular simulation techniques (Monte Carlo) using a primitive model in a canonical ensemble (NVT). The proposed model takes into account the influence of ionic size of each particle and the addition of polyelectrolytes with different charge number and separation. The results obtained show that electrostatic interactions are responsible for the cohesion of the hardened cement. It was found that in absence of cationic polyelectrolytes, cohesion is lost when the C-S-H lamellae are at separations larger than 1 nm. Adding cationic polyelectrolytes generates a distribution of hydroxide ions around the polyelectrolyte charges, facilitates the distribution of calcium and sodium ions in the entire space between C-S-H surfaces; this allows the cohesive forces exist at greater distances of separation between the surfaces.

Keywords: cement cohesion, molecular simulation, cationic polyelectrolytes, osmotic pressure, primitive model.

Efecto de la adición de polielectrólitos catiónicos en la cohesión del cemento Resumen Se estudia la variación en la cohesión de la fase principal del cemento (C-S-H) por efecto de la adición de polielectrolitos catiónicos (aminas cuaternarias espermina y norespermidina). El estudio de la cohesión se realiza mediante técnicas de simulación molecular (Monte Carlo) usando un modelo primitivo en el ensamble Canónico (NVT). El modelo propuesto tiene en cuenta la influencia del tamaño iónico de cada partícula y la adición diferentes polielectrólitos. Los resultados muestran que las interacciones electroestáticas son las responsables de la cohesión del cemento. Sin polielectrolitos la cohesión se pierde cuando las láminas de C-S-H tienen separaciones mayores de 1 nm. Los polielectrolitos generan que los iones hidróxido se distribuyan alrededor de las cargas de estos, facilitando la distribución de los iones de calcio y de sodio en el espacio de separación entre las superficies del C-S-H, esto permite que las fuerzas de cohesión existan a mayores distancias de separación.

Palabras clave: cohesión del cemento, simulación molecular, polielectrolitos catiónicos, presión osmótica, modelo primitivo.

1. Introduction Cement is considered an artificial mineral, whose raw materials for elaboration are limestone, clay, sandstone and gypsum. These materials are put through a series of processes to complete the elaboration of the final product, which include grinding, homogenization (wet or dry), calcination and clinker trituration [1]. The lamellae of calcium silicate hydrate (C-S-H) constitute between 50 to 67% of the final volume of hydration products, becoming the main component of hardened cement. C-S-H is therefore responsible of the main macroscopic mechanic properties of

cement, specially the resistance to compression stress and durability.

To give an appropriate description of factors which influence the cement of cohesion, it must consider the interactions that can occur among C-S-H lamellae, ions in interstitial solution and the interaction between these two factors. A detailed description of cohesion, it should contemplate in its model the attractive forces which generate surfaces of C-S-H charged negatively and confined ions that are located among C-S-H lamellae.

Up to now it has postulated two interactions that give explanation to the cement of cohesion phenomenon: the

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 222-228. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41834

Zuluaga-HernĂĄndez & Hoyos / DYNA 81 (188), pp. 222-228. December, 2014.

attraction between the ions of interstitial solution with lamellae, due to electrostatic attractions generated by the high ionic correlation [2], this occurs with high pH values; and the interaction between ions that are embedded in the C-S-H lamellae which generate attractions closer to a covalent bond character [3]. Hence the system presents characteristics of ionic interactions and covalent bonds, which allow the lamellae remain linked together, these interactions are mainly responsible for the cohesion of hardened cement.

The debate that has taken place among researchers to elucidate the interactions is more representative, it has been continuous in recent years [4-5]. However, the character of covalent bond is not clearly defined. The results obtained in a previous study [4] show that about 60% are interactions of ionic character, which implies that the covalent character is around 40%, this could give clue that calcium ions are embedded in C-S-H structure which have a less interchangeable character and they have a limited movement. While interstitial ions (among lamellae) have a mobile more character.

In this article, we wish to present a model that describes the most representative interactions which generate cement cohesion in C-S-H lamellae with separations of up to 2 nm and the cations presence between the lamellae. For this model only considered electrostatic interactions (ionic character), with the understanding that such interactions are the most representative. This has been widely studied, both theoretically and experimentally [6-11].

The Poisson-Boltzmann and the DLVO theories were the first two theoretical explanations to the cohesion phenomenon. These theories do not consider the high ionic correlation generated between divalent ions of Ca2+ and the C-S-H lamellae, and inaccurately predict the cohesion of hardened cement paste [6,8,11-14].

The primitive models have allowed explaining the cohesion of cement in an adequate manner, and allowing calculating the attraction forces generated between the ions present in solution and the C-S-H lamellae, using Coulombâ&#x20AC;&#x2122;s potential. Results show that these kinds of models are able to represent the strong ionic correlation in cement at a molecular scale, due to the coupling between co-ions and counterions in the electrolytic solution and the lamellae of C-S-H (with high density of negative superficial charges). The strength of attraction forces is associated with the cementâ&#x20AC;&#x2122;s capability of resisting compression stress, but is also associated with a weak response to traction stress [45,11].

Recent studies propose a possible way to modify the molecular interactions in order to modulate the ductility of cement by implementing changes in the interactions between the electrolytic solution and the C-S-H lamellae [5]. This consists in adding polyelectrolytes (to the clinker) with the goal of creating cement with higher resistance to traction stress due to higher interaction range between C-SH lamellae and the electrolytic solution.

Literature shows important breakthroughs in this matter, both theoretical and experimental [5,15-18]. However, there is still work left in the way to improve the macroscopic properties of cement, modifying the interactions solution-

lamella by the inclusion of new compounds in the solution.

In this work, Monte Carlo molecular simulation is used to determine the changes in cohesion forces due to the addition of cationic polyelectrolytes to the clinker and to establish the variation of this cohesion force with the separation between C-S-H lamellae. Thus, an important step is made towards accurately modeling and understanding cement in a nanometric scale, which allow modulating and ultimately improving the macroscopic properties of this important material.

2. Model For the development of this model it was considered that the interactions between nanoparticles of C-S-H can be represented as the interaction of two flat walls negatively charged, with a charge density correspondent to the natural mineral tobermorite. All the ions in the interstitial solution were represented explicitly, and water was modeled as a continuum dielectric medium, characterized only by its relative permittivity. The confined solution was represented as a solution of electrolytes and polyelectrolytes with positively and negatively charged ions in order to guarantee neutral charge of the system as a whole, and whose mean concentration varies with the separation between the C-S-H lamellae.

A

B Figure 1. Schematic representation of the system: the C-S-H lamellae negatively charged and in solution ions of Ca+2, Na+, OH- , norspermidine (A) and spermine (B).

Source: The authors.

223

Zuluaga-Hernández & Hoyos / DYNA 81 (188), pp. 222-228. December, 2014.

For the simulations without polyelectrolytes, the interstitial solution was represented with calcium, sodium and hydroxide ions. In order to study the effect of the addition of cationic polyelectrolytes, this compounds were represented as quaternary amines, as they are the only stable polyelectrolytes at the pH conditions that characterize the cement paste (pH>12.5). The two polyelectrolytes chosen for this study were norspermidine (H3N+-(CH2)3-N+H2(CH2)3-N+H3) (Fig. 1 A) and spermine (H3N+-(CH2)3-N+H2(CH2)4-N+H2-(CH2)3-N+H3) (Fig. 1 B).

In order to represent the attractive and repulsive interactions generated at a molecular scale in cement cohesion, a primitive model was used, with Lennard-Jones and Coulomb’s potential.

The model here proposed presents two novel aspects respect to previous studies [5-6,8]. The first aspect is the consideration of different atomic sizes of the species involved in the electrolytic solution. The second aspect is the addition of a cationic polyelectrolyte with different separation distance between the positively charged amine groups.

The primitive model represents the ions in the electrolytic solutions as soft spheres with a positive or negative electric charge, depending on the nature of the ion.

These ions are separated a distance ( ܑ ‫) ܒ‬. The solvent in which ions are submerged is modeled as a continuum medium with a constant relative permittivity ( ‫) ܚ‬.

Coulomb’s potential is used to describe the electrostatic interactions between all ions in the solution, and between the ions and the negatively charged sites at the surfaces of C-S-H: ୯౟ ୯ౠ ୧

୨

ସ஠஫బ ஫౨ ሺ୰౟ ି୰ౠ ሻ

୧

୨

Where ଴ and d are the vacuum permittivity and the characteristic diameter of each ion present in the system, respectively.

Short range interactions are described using LennardJones potential: ଵଶ ௅௃

௜௝

଺

௜௝

௜௝

௜௝

௜௝

The values of parameters ε and σ were taken from the general force field CLAYFF [19], and are listed in Table 1.

Cationic polyelectrolytes were added in the center of the simulation box and were distributed equally and perpendicular to the surfaces of C-S-H. The potentials for bonds between carbons and hydrogens in the polyelectrolytes were not considered explicitly, as they are not relevant in electrostatic interactions. The polyelectrolyte norspermidine have three positive charges, and the spermine is a quaternary amine that has four positive charges. The distance between each amine group is illustrated in Fig. 2.

The interaction parameters between atoms of different nature are evaluated using the Lorentz-Berthelot mixing rules:

Table 1.

Lennard-Jones potential parameters used in the model.

Species (kcal/mol) OH0.1554 Ca2+ 0.1000 Na+ 0.1301 R 3N + 0.0650 Source: Adapted from [19]

(Å) 3.5532 3.2237 2.6378 3.3262

Figure 2. Distance of amine groups for Norspermidine (A) and spermine (B).

Source: The authors.

୧୨

୧

୧୨

୨

୧ ୨

The lamellae of C-S-H were modeled as uniform surfaces of 60 nm long (X axis) and 30 nm wide (Y axis). For the separation between lamellae of C-S-H, four different separations were considered in simulations (0.5, 1, 1.5 and 2 nm).

Assuming C-S-H has a structure similar to tobermorite [13], the surface charge density for the walls was set to 4.8 e-/nm2. Although size and charge density of C-S-H lamellae can vary with concentration of species and pH, these variations are not considered in this work.

For flat surfaces, cohesion force is directly related to the osmotic pressure of the system, and this osmotic pressure is also related to the interactions between particles. To obtain the net osmotic pressure it is necessary to evaluate the confinement osmotic pressure and the bulk osmotic pressure when the system is equilibrated. The confinement osmotic pressure of the interstitial solution can be calculated with the ionic distribution near the C-S-H surface: ଶ ୡ୭୬୤ ୭ୱ୫

୵ୟ୪୪ ୧

୆ ୧

୰ ୭

With the concentration of ions near the charged wall it is possible to evaluate the expression ୆ ୧୵ୟ୪୪ . The sum of the interactions of a wall with all the ions present in the simulation box and the other charged wall is obtained by the Maxwell term

224

஢మ ଶ஫౨ ஫౥

[6].

Zuluaga-Hernández & Hoyos / DYNA 81 (188), pp. 222-228. December, 2014.

The bulk osmotic pressure considers the effects of the ions fraction volume ( ), the ion bulk concentration for each ion species ( ୧ ), the average energy of the system and the correlation function between the ions ୲୭୲ ୧୨

Table 2.

Wall separation (z) and number of ions for simulations. Norspermidine (Nor) and spermine (Spe).

Na+ OHNor Spe z (nm) Simulation Ca2+ 1 2 3 4 5 1 6 7 8 1.5 9 10 11 2 12 Source: The authors

11 11 11 22 22 22 33 33 33 44 44 44

0.5

୦ୡ

୆୳୪୩ ୭ୱ୫

୆୳୪୩ ୧

୆ ୧

୧୨ ୧

୦ୡ

୲୭୲ ୆

୆୳୪୩ ୆୳୪୩ ୧ ୨

୨

For the calculation of ୧୵ୟ୪୪ and ୧୆୳୪୩ in eq. (5 and 6), it is considered that the "wall" region extends from each wall to a distance of 0.6 ୓ୌ (2.13 Å), and the “bulk” region the remaining space in between them.

The correlation function between ions indicates the direct influence of a given particle over another; located at a distance d (the diameter of the ion). This function is calculated as: ୧ ୨ ୧୨

୦ୡ

ୡ୭୬୤ ୭ୱ୫

୧

3. Simulation details Monte Carlo simulations were performed using a canonical ensemble (NVT) at 25°C, with periodic boundary conditions in two dimensions (X and Y axis). The cut radius for electrostatic interactions was set to 60 nm, which corresponds to the length of the simulation box.

For the analysis of the cohesion between C-S-H lamellae, three different systems were used: with no polyelectrolytes (only ions of Ca+2, Na+ and OH-); with norspermidine: H3N+(CH2)3-N+H2-(CH2)3-N+H3, and lastly with spermine: H3N+(CH2)3-N+H2-(CH2)4-N+H2-(CH2)3-N+H3. For each system four simulations were conducted, at four different separation distances, for a total of twelve simulations.

Table 2 summarizes the amount of ions used in each simulation. The polyelectrolytes were added in the Z axis, perpendicular to the surfaces of C-S-H. The number of ions was determined considering that the concentration of the solution was 20mM for Ca2+ ions, 100 mM for Na+ ions and 15 mM for polyelectrolytes (Norspermidine and spermine) [5,11]. OH- ions were added as necessary in order to guarantee the electric neutrality of the system. Thus, it is possible to obtain a similar pH to that observed in real cement paste (pH>13).

0 8 0 0 16 0 0 24 0 0 32 0

0 0 8 0 0 16 0 0 24 0 0 32

ଷ ୧

୧

୆୳୪୩ ୭ୱ୫

Simulation results are the equilibrium distributions of all ions in the interstitial solution, which allow calculating the osmotic pressure. This osmotic pressure is an indicator of the cohesion force of the composed system of the C-S-H lamellae and the electrolytic interstitial solution.

76 100 108 152 200 216 228 300 324 304 400 432

For all simulations, 3x106 equilibration cycles and 1x106 production cycles were conducted, with a random initial configuration for ions in the interstitial solution and the charges distributed in the surface of C-S-H. In order to determine density profiles (dimensionless) of ionic species, the number of particles contained in thin layers, of thickness ∆Z, parallel to the C-S-H lamellae was determined:

Therefore, the net osmotic pressure of the system is obtained by: ୭ୱ୫

54 54 54 108 108 108 162 162 162 216 216 216

4. Results The results of different simulations conducted in the canonical ensemble (NVT) were divided in three groups as follows: in absence of polyelectrolytes, with norspermidine and with spermine. For each case, four different simulations were conducted, at four different separation distances (0.5, 1, 1.5 y 2 nm). As some of the results show similar patterns, only the most representative are shown.

From density profile of the system in absence of polyelectrolytes (Fig. 3) it can be seen that the Ca2+ ions are distributed along the surface of the two lamellae of C-S-H. This favors the cohesion between the walls, given that the OH- ions (Fig. 4) are mostly located in the middle and attract the Ca2+ ions that are distributed evenly at the two walls.

Figure 3. Density profile for Ca2+ (black line) and Na+ (gray line) without polyelectrolytes for 0.5 nm C-S-H separation.

Source: The authors.

225

Zuluaga-Hernรกndez & Hoyos / DYNA 81 (188), pp. 222-228. December, 2014.

Figure 4. Density profile for OH- without polyelectrolytes for 0.5 nm C-SH separation.

Source: The authors.

Figure 6. Density profile for OH- without polyelectrolytes for 2nm C-S-H separation.

Source: The authors.

Figure 5. Density profile for Ca2+ (black line) and Na+ (gray line) without polyelectrolyte for 2nm C-S-H separation.

Source: The authors.

Figure 7. Density profile for Ca2+ (black line) and Na+ (gray line) with norspermidine for 1.5 nm C-S-H separation.

Source: The authors.

It can be seen from Fig. 3 that the sodium ions (gray line) are mostly located near the two surfaces of C-S-H, but unlike the calcium ions, some of these ions are located near the middle plane, aiding to stabilize the region with high concentration of OH- ions. Up until a separation distance of 1 nm a strong attraction between C-S-H lamellae was observed.

Unlike the separations of 0.5 and 1 nm, when the separation between the walls is 1.5 and 2 nm, the calcium and sodium ions are located near only one wall (Fig. 5) at a distance of approximately 0.5 nm, and the OH- ions (Fig. 6) are located in a layer at a further distance. This means that all the species are adhered at just one surface, which is the main reason why the cohesion in cement is lost.

The results obtained for the system at separation distances of 1.5 and 2 nm are similar. For greater separation distances, a rupture in the system is observed, generating the adhesion of the species at only one lamella causing loss of cohesion.

These results are associated with the macroscopic property of cement causing great resistance to compression stress but very little to traction stress.

The results obtained with the addition of cationic polyelectrolytes show that attraction forces that guarantee cohesion can be achieved between the lamellae of C-S-H at

Figure 8. Density profile for OH- with norspermidine for 1.5 nm C-S-H separation.

Source: The authors.

separations of 1.5 and 2 nm, which are not observed in the absence of these polyelectrolytes in the electrolytic solution.

This behavior can be corroborated with the density profiles of calcium, sodium and hydroxide ions (Fig. 7-10) obtained for simulations at separations of 1.5 and 2 nm with the presence of norspermidine and spermine, respectively.

226

Zuluaga-Hernรกndez & Hoyos / DYNA 81 (188), pp. 222-228. December, 2014.

Figure 9. Density profile for Ca2+ (black line) and Na+ (gray line) with spermine for 2 nm C-S-H separation.

Source: The authors.

Figure 11. Osmotic pressure between C-S-H surfaces. Without polyelectrolyte (solid line), with norspermidine (dashed line) and spermine (dotted line).

Source: The authors.

Figure 10. Density profile for OH- with spermine for 2 nm C-S-H separation.

Source: The authors.

Fig. 7 and Fig. 8 show that the addition of norspermidine generates a distribution of hydroxide ions around the positive charges. This in turn, by means of electrostatic attraction, enables the calcium and sodium distribution in the entire separation space between C-S-H lamellae.

The addition of norspermidine generates attraction forces between the two negatively charged surfaces of C-SH, at separations of 1.5 and 2 nm. These results indicate that the presence of norspermidine indirectly enables the distribution of the calcium and sodium ions throughout the separation space between the C-S-H lamellae, which allows cohesion forces to exist at greater separation distances between the surfaces. The same phenomenon is observed with the addition of spermine (Fig. 9 and Fig. 10). This polyelectrolyte has four different positive charges at different bond distances between charged amine groups.

From the analysis of the previous results, it can be established that the systems with the addition of cationic polyelectrolytes (norspermidine and spermine) in presence of sodium, calcium and hydroxide ions evidence attraction forces strong enough to accomplish cohesion at separations greater than 1 nm.

The attraction present in cement is possible due to electrostatic interactions between the highly charged C-S-H walls and the ions present in the interstitial solution. This interaction was evaluated by the calculation of the net

osmotic pressure (Fig. 11). For the system in absence of electrolytes it is observed that for the separations of 1.5 and 2 nm there is no evidence of cohesive forces, given that the ions distribute along a single surface of C-S-H. The opposite is observed when cationic polyelectrolytes are added to the system, where an attractive force between C-SH lamellae is evident at separations greater than 1nm.

From Fig. 11 it can be seen that the results with spermine are better than for norspermidine (greater net cohesion force and longer range effect). This behavior is due to the additional charge on the spermine structure, and a greater chain length, which allow the spermine to generate attractive forces at a greater distance between negative charges of the C-S-H lamellae. This is congruent with studies previously conducted [5,13], which explain the cohesion of the cement paste due to the strong Coulombic interactions between the charged surfaces of C-S-H and the confined ions between this surfaces, showing that the two main factors that govern cohesion are the superficial charge density of C-S-H lamellae and the valence of positively charged ions present in the interstitial solution.

The attraction forces between the C-S-H lamellae can be modified with the addition of quaternary amines, since these generate a distribution of the sodium and calcium ion throughout the entire separation space between C-S-H lamellae. This causes a net osmotic pressure that guarantee cohesion force at greater separation distances than in the absence of amines.

5. Conclusions Results obtained in this work allow concluding that the primitive model here developed is an adequate representation of the electrostatic interactions between the C-S-H lamellae and the electrolytic solution. This

227

Zuluaga-Hernández & Hoyos / DYNA 81 (188), pp. 222-228. December, 2014.

corroborates that the electrostatic interactions are responsible for the cohesion of hardened cement, and are crucial to widen the understanding of the phenomenon that take place at a nanometric scale. In this manner, it is possible to design strategies to modify and improve the macroscopic mechanic properties of cement.

In the absence of polyelectrolytes, cohesion is lost at separation distances greater than 1 nm. Adding polyelectrolytes, such as quaternary amines, causes the hydroxide ions to distribute around the positive amine charges, and these hydroxide ions, by means of electrostatic interactions, also enable the distribution of calcium and sodium ions throughout the entire space of separation between the C-S-H lamellae. This allows the cohesion forces to exist at greater separation distances between surfaces.

The results here obtained show that the spermine produces a greater net cohesion force, and an effect of longer range than norspermidine.

With the obtained results it is expected that systems with a higher charge density and polyelectrolytes with a higher number of positive charges would produce greater ranges of separation between the C-S-H lamellae.

References [1]

Neville, A.M. and Brooks, J.J., Concrete technology. London: Prentice Hall, 2010.

[2] Bellotto, M. and Bozzetto, G., On the origin of cement setting, and how to control the properties of the fresh and hardened material, Special Issue for International Congress on Materials & Structural Stability, pp. 74-79, 2013.

[3] Thomas, J., Scherer, W., Biernacki, G., Luttge, A., Bullard, J., Bishnoi, S., and Dolado, J., Modeling and simulation of cement hydration kinetics and microstructure development, Cement and Concrete Research, 41, pp.

1257-1278, 2011. http://dx.doi.org/10.1016/j.cemconres.2010.10.004 [4] Pellenq, R., Lequeux, N. and Damme, H., Engineering the bonding scheme in C–S–H: The iono-covalent framework, Cement and Concrete Research, 38, pp.

159-174, 2008. http://dx.doi.org/10.1016/j.cemconres.2007.09.026 [5] Pochard, I., Labbez, C., Nonat, A., Vija, H. and Jönsson, B., The effect of polycations on early cement paste, Cement and Concrete Research, 40, pp.

1488-1494, 2010. http://dx.doi.org/10.1016/j.cemconres.2010.06.002 [6] Kjellander, R., Marcelja, S. and Quirk, J., Attractive double-layer interactions between calcium clay particles. Journal of Colloid and Interface Science, 126, pp.

194-211, 1988. http://dx.doi.org/10.1016/0021-9797(88)90113-0 [7] Manzano, H., Dolado, J.S., and Ayuela, A., Elastic properties of the main species present in portland cement pastes, Acta Materialia, 57, pp.

1666-1674, 2009. http://dx.doi.org/10.1016/j.actamat.2008.12.007 [8] Jonsson, B., Wennerstrom, H., Nonat, A. and Cabane, B., Onset of cohesion in cement paste, Langmuir, 20, pp. 6702-6709, 2004. http://dx.doi.org/10.1021/la0498760 [9] Labbez, C., Nonat, A., Pochard, I. and Jönsson, B., Experimental and theoretical evidence of overcharging of calcium silicate hydrate, Journal of Colloid and Interface Science, 309, pp. 303-307, 2007. http://dx.doi.org/10.1016/j.jcis.2007.02.048 [10] Pellenq, R., Caillol, J. and Delville, A., Electrostatic attraction between two charged surfaces: A (N,V,T) Monte Carlo simulation, The Journal of Physical Chemestry, 101, pp. 8584-8594, 1997. http://dx.doi.org/10.1021/jp971273s [11] Jonsson, B., Nonat, C., Labbez, B., Cabane, B. and Wennerstrom, H., Controlling the cohesion of cement paste, Langmuir, 21, pp.

9211-9221, 2005. http://dx.doi.org/10.1021/la051048z

[12] Lesko, S., Lesniewskaa, E., Nonat, A., Mutinb, J. and Goudonneta, J., Investigation by atomic force microscopy of forces at the origin of cement cohesion, Ultramicroscopy, 86, pp. 11-21, 2001. http://dx.doi.org/10.1016/S0304-3991(00)00091-7 [13] Labbez, C., Jonsson, B., Pochard, I., Nonat, A. and Cabane, B., Surface charge density and electrokinetic potential of highly charged minerals: Experiments and Monte Carlo simulations on calcium silicate hydrate, The Journal of Physical Chemistry B, 110, pp.

9219-9230, 2006. http://dx.doi.org/10.1021/jp057096+ [14] Kjellander, R., Marcelja, S., Pashley, R. and Quirk, J., Double-layer ion correlation forces restrict calcium-clay swelling, The Journal of Physical Chemestry, 92, pp.

6489-6492, 1988. http://dx.doi.org/10.1021/j100334a005 [15] Delville, A., Influence of interionic correlations on the free energy of charged interfaces: a direct derivation from (N,V,T) Monte Carlo simulations, The Journal of Physical Chemistry, 108, pp. 9984-9988, 2004. http://dx.doi.org/10.1021/jp049189h [16] Labbez, C. and Jonsson, B., New Monte Carlo method for the titration of molecules and minerals, Computer Science, pp. 4699, 6672, 2007.

[17] Labbez, C., Jonsson, B., Skarba, M. and Borkovec, M., Ion-ion correlation and charge reversal at titrating solid interfaces, Langmuir, 25, pp.

7209-7213, 2009. http://dx.doi.org/10.1021/la900853e [18] Delville, A., Gasmi, N., Pellenq, R., Caillol, J. and Damme, H., Correlations between the stability of charged interfaces and ionic exchange capacity: A Monte Carlo study, Langmuir, 14, pp. 50775082, 1998. http://dx.doi.org/10.1021/la9802872 [19] Randall, T., Jian-jie, L. and Kalinichev, A., Molecular models of hydroxide, oxyhydroxide, and clay phases and the development of a general force field, The Journal of Physical Chemestry, 108, pp.

1255-1266, 2004. http://dx.doi.org/10.1021/jp0363287 E.A. Zuluaga-Hernández, received the BSc in 2010 and MSc in 2014 degrees both in Chemical Engineering; all of them from the Universidad Nacional de Colombia, Medellín, Colombia. Currently He is a student of the PhD in Universidad Nacional de Colombia, Medellín, Colombia.

B.A. Hoyos, received the BSc. in 1994 and MSc in 2003 degrees both in Chemical Engineering; and the PhD degree in Energy Systems in 2010, all of them from the Universidad Nacional de Colombia, Medellín, Colombia.

Currently he is a Full Professor in the Departamento de Procesos y Energía, Facultad de Minas, Universidad Nacional de Colombia, Medellín, Colombia. His research interests include molecular simulation, modeling asphaltene aggregation and viscosity, modelling wettability alteration by surfactants and specific adsorption.

228

A new dynamic visualization technique for system dynamics simulations Ricardo Sotaquirá-Gutiérrez a a

Facultad de Ingeniería, Universidad de La Sabana, Bogotá, Colombia. ricardosg@unisabana.edu.co

Received: January 30th, de 2014. Received in revised form: September 10th, 2014. Accepted: September 29th, 2014

Abstract This article addresses a main research question in System Dynamics, the correct understanding of the relationship between model structure and behavior, from a new perspective: the field of Information Visualization. We propose a new dynamic visualization technique that is different from the current static visualization techniques. The technique was applied to a classical case and tested through visual and interactive software built as proof of concept. The case illustrated the potential of the new technique to confront the problem by providing the user a visual experience in real time, so his/her understanding could be more direct and immediate. This visualization technique could be adapted and generalized to other simulation methodologies.

Keywords: information visualization, system dynamics, simulation, visual and interactive computing, human-computer interaction.

Una nueva técnica de visualización dinámica para simulaciones en dinámica de sistemas Resumen Se aborda un problema central de investigación en Dinámica de Sistemas, el de la correcta comprensión de la relación entre la estructura y el comportamiento de un modelo, a partir de un enfoque novedoso: el de Visualización de información. Se propone una nueva técnica de visualización que a diferencia de las existentes de naturaleza estática, ofrece un modo dinámico para representar resultados de simulación. Se aplica esta técnica para un modelo clásico mediante la creación de un software visual e interactivo como prueba de concepto. El caso ilustra el potencial de la técnica para atacar la problemática porque brinda al usuario una experiencia visual en tiempo real que facilita una comprensión más directa e inmediata. Esta técnica puede adaptarse y generalizarse para la visualización de datos de simulación obtenidos no solamente con Dinámica de Sistemas.

Palabras clave: visualización de información, dinámica de sistemas, simulación, computación visual e interactiva, interacción personacomputador.

1. Introducción En un artículo clásico en simulación de fenómenos complejos, John Sterman muestra un experimento en el que personas que toman decisiones no comprenden adecuadamente cómo una estructura de relaciones causaefecto genera un determinado comportamiento [1]. Esta es una de las dificultades más comunes a las que se enfrenta un usuario de la Dinámica de Sistemas (DS) y se hace evidente, por lo menos, en dos situaciones típicas: cuando nuevos usuarios de la metodología la están aprendiendo y cuando se comunican los resultados de un proyecto de simulación a beneficiarios e interesados que no son conocedores de la metodología. En ambos casos el usuario no experto se

enfrenta a un reto de comprensión. Por un lado tiene un modelo matemático con una red de relaciones causales cíclicas y, por el otro, un conjunto de resultados de simulación, pero no entiende de inmediato cuál es la unidad de estos dos aspectos del modelo. Tiene problemas para explicar crecimientos, decrecimientos y en general cambios que aprecia en las gráficas de simulación en términos de las relaciones causales y los ciclos de realimentación. Estos son obstáculos claves para el uso adecuado de la Dinámica de Sistemas y es por ello que han sido tema recurrente de investigación [1-3].

El presente artículo ofrece una perspectiva novedosa en el abordaje de esta limitación fundamental para el modelado y la simulación en Dinámica de Sistemas, al ubicar la

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 229-236. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.41843

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

problemática dentro de un campo nuevo que ha surgido del encuentro entre la computación y el diseño visual: la Visualización de información o la Visualización Científica [4-7]. Desde esta perspectiva la función primordial de visualizar los resultados de una simulación consistiría en facilitar al usuario comprender los patrones de comportamiento de las variables allí representadas como productos emergentes de las relaciones causales. Si se ubica esta función dentro del marco conceptual de la Visualización de información (VI), se puede decir que se trata de una tarea de visualización sinóptica [8]: lo que importa al usuario no son simplemente los datos individuales sino apreciar de manera comprensiva patrones de comportamiento en dichos datos.

Para lograr esta percepción sinóptica de los datos de una simulación se dispone de dos tipos generales de visualizaciones: estáticas y dinámicas [9]. Las primeras son más utilizadas, las más clásicas. En ellas tanto el tiempo como los datos de las variables son ubicados dentro de un mismo espacio de representación visual. Usualmente el tiempo es colocado sobre el eje X en una visualización cartesiana bidimensional. Se obtiene entonces la conocida gráfica de líneas para la evolución de cada variable en el tiempo. En cambio, en las visualizaciones dinámicas el tiempo no es convertido en una variable más dentro del espacio de visualización, no se “congela” sino que transcurre. Para ello se construye una serie de visualizaciones en un determinado espacio y se presentan sucesivamente una por una durante un intervalo de tiempo.

Cada visualización que hace parte de este conjunto se denomina cuadro. Se obtiene entonces una visualización en forma de animación por computador.

En Dinámica de Sistemas han dominado las visualizaciones estáticas. En el primer libro de DS de Jay Forrester, "Industrial Dynamics" [11], se visualizan los resultados en gráficos de líneas producidos simplemente con caracteres ASCII. Esto resulta explicable por la tecnología computacional disponible en la década de 1960. Sin embargo, cincuenta años después estas curvas siguen siendo la forma dominante, y casi exclusiva, para representar los resultados de una simulación en las herramientas de software disponibles (Vensim, Powersim, Ithink, Evolución, entre otros).

La intención de este trabajo fue explorar las potencialidades de una forma alternativa para representar las simulaciones: la visualización dinámica. Con referencia a este tipo de visualización Aigner y colegas afirman que: "En la propia percepción humana hay una comprensión intrínseca del tiempo, en especial del transcurrir del tiempo, y la visualización puede aprovechar esta cualidad si esa dimensión del tiempo es transformada en la dinámica de una representación visual” [9]. En consecuencia, sería de esperar que una visualización dinámica resulte más coherente con nuestra percepción humana que los gráficos estáticos tradicionales. Adicionalmente estos autores sugieren que “la mejor manera de representar numerosos datos de procesos fuertemente dinámicos es a través de animaciones, puesto que ellas comunican muy bien la dinámica subyacente en los datos” [9]. La idea de que la animación puede facilitar los procesos de aprendizaje con herramientas de simulación

en Ingeniería también ha sido planteada previamente por Valencia y otros [10].

De modo que una visualización dinámica podría servir mejor a la función primordial de la visualización en DS: el descubrimiento de patrones de comportamiento que emergen de una red causal. El artículo presenta entonces una aplicación de la VI a un conjunto de datos obtenidos por simulación con Dinámica de Sistemas.

Primero se examina el estado del arte en la visualización dinámica. Luego se reseñan los antecedentes de visualización de información en el campo de la DS y se ponderan sus limitaciones. Posteriormente se presenta el diseño de la nueva técnica de visualización dinámica, así como una prueba de concepto a nivel de software. La técnica propuesta fue aplicada y probada con un caso clásico de modelado con DS: el modelo Bass de difusión de tecnología [13]. Por último, se discuten los resultados obtenidos con este ejemplo de aplicación y se plantean las conclusiones del trabajo.

2. Técnicas de Visualización de Información En el campo de Visualización de Información se denominan “time-oriented data” (datos orientados en el tiempo) al tipo de datos que se producen en simulaciones en donde la variable independiente es el tiempo. En una cuidadosa revisión Aigner y colegas recopilaron 101 técnicas de visualización de datos orientados en el tiempo [9], de las cuales solamente 11 ofrecen formas de visualización dinámica. De acuerdo con este marco, los datos que se obtienen en simulaciones en Dinámica de Sistemas tienen las siguientes características: corresponden a múltiples variables dependientes del tiempo; no son georeferenciados sino abstractos; corresponden a instantes más que a intervalos; y por último son datos discretos.

Dentro de la revisión de Aigner se encuentran entonces tres técnicas de visualización afines a este tipo de datos: Trendalyzer [14], TimeRider [15] y Virtual Instruments [16].

Las visualizaciones creadas con Trendalyzer se han hecho bien conocidas a través una presentación de Rosling en una conferencia TED [14]. Esta herramienta permite analizar indicadores demográficos, socioeconómicos y de salud de múltiples países. Utiliza una representación cartesiana bidimensional, en cada eje se ubica una variable.

El cambio de estas variables en el tiempo se representa mediante animación. Otras características distintivas del software son: utiliza el tamaño del punto como variable visual para representar qué tan grande es la población de los países; el cambio en el tiempo de los valores de una variable determinada para un país deja un rastro visual; el usuario maneja a voluntad el paso del tiempo mediante una barra de deslizamiento. Es decir, a diferencia de las visualizaciones de simulación donde el tiempo simplemente avanza, en esta el usuario puede explorar los resultados también en dirección opuesta o repasar intervalos claves del comportamiento de las variables. Ninguna de estas características están presentes en las herramientas software actuales en Dinámica de Sistemas.

TimeRider [15] es un software orientado a VI sobre

230

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

pacientes con diabetes. Su espacio de representación es cartesiano bidimensional y como Trendalyzer el tiempo aparece en la animación y los datos dejan rastro visual.

Dado el interés que tienen los médicos en comparar conjuntos de datos de múltiples pacientes, el software ofrece interactividad para seleccionar y resaltar los datos de uno o varios pacientes en particular. Esta última característica podría ser útil para análisis de sensibilidad en DS, pero no es el caso que se aborda en este artículo.

Por último Matkovic y colegas aplican conceptos de VI al diseño de indicadores visuales en la industria automotriz, en el formato de tacómetros [16]. Se trata de variables continuas de desempeño de un automóvil en un laboratorio de pruebas. El tacómetro virtual no solamente indica la velocidad actual mediante una manecilla sino que muestra el rastro de las velocidades recientes. Lo mismo ocurre con los demás indicadores.

Ninguno de estos tres casos de visualización dinámica trabaja con conjuntos de datos obtenidos a partir de una simulación, su información es obtenida por observación o experimentación. Sin embargo, a partir de ellos se formularon los siguientes principios de diseño para ser aplicados en una visualización dinámica en DS:  Utilizar las variables visuales de color, tamaño y posición para representar cambios en los datos.  Aplicar una gama de tonos de un mismo color para dejar un rastro visual de los valores anteriores de cada variable.  Facilitar al usuario opciones de interacción que le permitan manejar a voluntad el tiempo de la animación: avanzar, retroceder y cambiar la velocidad.

Por otro lado, al interior del campo de la Dinámica de Sistemas no solamente se utilizan las formas de visualización indicadas en la sección anterior y hay antecedentes de innovación en la visualización de los resultados de simulación. Además de los gráficos de líneas utilizados desde los primeros trabajos en DS hasta la actualidad, ocasionalmente puede ser necesario comparar resultados de simulaciones sucesivas en un análisis de sensibilidad [13]. En tal caso se superponen bidimensionalmente o tridimensionalmente los gráficos de líneas en el tiempo. También para casos específicos, como por ejemplo el análisis de la estabilidad de algunos sistemas, se utilizan los diagramas de fase o diagramas de plano de fase propios del campo del cálculo diferencial no-lineal [17]. Ambos casos completan el repertorio de alternativas de visualización de tipo estático usadas en Dinámica de Sistemas.

En cuanto a innovaciones en la visualización en DS hay que resaltar el proyecto pionero de Howie y colegas para mejorar la interfaz de usuario de un micromundo gerencial y facilitar a los usuarios una mejor comprensión del sistema simulado [18]. Aunque esta experiencia no acude específicamente a principios de VI, que para la fecha estaban hasta ahora consolidándose, sí se pueden interpretar algunos cambios de la interfaz como mejoras en la visualización. Buena parte de los datos de las variables claves que le eran suministrados al jugador simplemente de manera numérica son reemplazados por gráficos de barras

que además muestran los posibles límites superior e inferior de cada variable.

Por otro lado las interfaces de los micromundos de DS se han visto mejoradas con las nuevas tecnologías visuales e interactivas [19-21]. En la actualidad lo usual es que cualquiera de estos simuladores incluya algunas visualizaciones estáticas de los datos. Sin embargo, este no es el tipo de simulaciones que interesan en el presente artículo. En los micromundos no se despliega el comportamiento completo de una variable de una sola vez sino que se van mostrando sus estados turno por turno del juego. En consecuencia aquí resultan pertinentes los conceptos de visualización estática más que los de visualización dinámica.

Por último, un tipo de visualización ciertamente dinámica usada en DS fue la planteada por Davidsen bajo la denominación “diagramas de comportamiento” [22]. Cada una de las variables de un modelo se acompaña de una miniatura que contiene una gráfica de líneas de la evolución de esta variable en el tiempo. En su momento esta forma de visualización se introdujo en el software Powersim, pero hoy está presente en otras herramientas de simulación con DS. Estas visualizaciones en miniatura no son dinámicas en sentido estricto. Se despliegan en pantalla en menos de dos segundos, para la percepción del usuario es tan breve la animación que solo termina apreciando el estado final, es decir una gráfica estática.

En resumen, en el campo de la DS solamente se encuentran técnicas de visualización estática y, por otro lado, en el área de Visualización de información hay casos pertinentes de visualización dinámica que brindan un punto de partida para el diseño de una nueva técnica de visualización para la DS. 3. Diseño de la técnica de Visualización Dinámica En síntesis, para el caso de la Dinámica de Sistemas se requiere una técnica que ofrezca una visualización de carácter dinámico que facilite a un usuario, especialmente al no-experto, “ver” cómo una estructura de relaciones causales cíclicas entre variables produce un determinado comportamiento apreciable en cada una de estas variables.

Para lograr este propósito se creó una técnica de visualización sustentada en tres principios de diseño, que incorporán los enunciados en la sección anterior, y que se explican con detalle:  Composición visual soportada en diagramas usados tradicionalmente.  Simplicidad en el cambio visual.  Alto grado de control del tiempo.

3.1. Composición visual soportada en diagramas Un primer aspecto que es necesario definir es la composición visual de la información. Para ello es necesario recordar dos características de los datos de simulación de DS según la tipología de Aigner [9]: son abstractos y multivariados. Aunque puede que haya casos específicos de modelos en DS con información georeferenciada o que corresponde a un fenómeno espacial real, este no es el caso

231

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

Figura 1. Ejemplo de un mapa de influencias.

Fuente: Autor.

general. Para datos georeferenciados o espaciales la composición la ofrecería el propio mapa al que pertenecen los datos. No es este el caso para los datos de simulación en DS, en este sentido son abstractos. En segundo lugar, dado que interesa la estructura de causas-efectos se requiere entonces que la visualización incluya información de múltiples variables que están relacionadas entre sí, por esto se afirma que se trata de datos multivariados.

Establecidas estas dos características entonces ¿Cómo ubicar en un espacio visual los datos de estas múltiples variables? Un diagrama utilizado tradicionalmente en DS ofrece un principio apropiado para la solución de este problema visual: el conocido diagrama causal o mapa de influencias [13] que muestra las variables del modelo, sus relaciones y los ciclos de realimentación (Fig. 1).

Para empezar se pueden disponer los elementos de la visualización dinámica de una manera similar. Con ello se resuelve el problema de la composición y se acude a una distribución que es común en DS. En síntesis la información que aparecerá en la composición visual consiste en un conjunto limitado de variables, para cada una de ellas debe aparecer un nombre o una identificación distintiva, así como sus valores a través del tiempo (fig. 2).

3.2. Simplicidad en el cambio visual Uno de los principios fundamentales del diseño visual e interactivo formulado por Maeda [23] es el de obtener simplicidad mediante la “selección cuidadosa” de los elementos visuales. En el caso de la visualización dinámica es necesario ser cuidadoso con aquello que cambia de un cuadro a otro en la animación. Una animación sobrecargada de cambios se hará menos comprensible para el usuario y esto estropearía el objetivo de la misma. En este sentido, la técnica propuesta limita este cambio entre cuadros únicamente a dos variables visuales: tamaño de punto y color. Para cada variable aparecerá un círculo cuyo diámetro será directamente proporcional al valor que esta variable tenga en cada instante de la simulación. En la animación sus valores son transformados entonces en un círculo de diámetro cambiante. La composición visual será entonces la de un conjunto de círculos que crecen o decrecen cada uno según la dinámica que emerge de las relaciones entre las variables. La segunda variable visual, color, se utiliza para dejar rastro del crecimiento o

Figura 2. Diseño visual de la técnica propuesta Fuente: Autor.

decrecimiento de un círculo, es decir, de los valores de una variable, mediante atenuaciones (Fig. 2).

3.3. Alto grado de control del tiempo Para mejorar la comprensión de la relación entre estructura y comportamiento en los modelos en DS resultan claves tanto la duración total de la animación como el grado de control que pueda tener el usuario de la variable tiempo. Si la duración de la animación es muy breve el usuario no-experto puede pasar por alto cambios claves en los valores de las variables. Si es muy prolongada el usuario puede perder concentración o enfoque. Por otro lado, si la animación aparece solamente una sola vez seguramente no bastará para que además de observar los cambios en valores el usuario también comprenda la relación de estos con la estructura causal, que es lo que interesa de fondo. Esto se resuelve brindando al usuario un control completo del tiempo de visualización. Para ello el usuario cuenta con dos posibilidades, similares a las que incluye la reseñada técnica de Trendalyzer [14]: ejecutar la visualización durante un tiempo fijo adecuado (20 segundos); o interactuar mediante un control sobre una barra deslizante del tiempo que le permita avanzar o retroceder a voluntad y a la velocidad deseada, de modo que pueda repetir la visualización cuantas veces necesite.

La Fig. 2 integra los tres principios de diseño de la técnica de visualización dinámica propuesta para DS. Se muestra un ejemplo genérico de un modelo con cuatro variables relacionadas a través de dos ciclos de realimentación. Se ilustran diferentes tamaños para cada variable según sean sus datos en un momento dado. Sobre la variable cuatro se dibuja su tamaño actual así como el rastro de un tamaño mayor que tenía previamente. Por último aparecen las dos opciones de control de la animación, automática (el botón triangular de reproducción) o manejada a voluntad (barra de deslizamiento).

4. Aplicación a un modelo en Dinámica de Sistemas 4.1. Modelo seleccionado

El modelo seleccionado cumple dos requisitos, el primero con respecto a su estructura causal, se trata de un

232

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

modelo sencillo en cuanto al número de variables y de ciclos de realimentación. Esto con el fin de que sea viable que un usuario no-experto de la DS pueda comprender la relación entre estructura y comportamiento. En segundo lugar, con respecto a su comportamiento éste no debe ser trivial sino que debe retar al usuario en cuanto a su comprensión.

El modelo escogido es denominado en la literatura “modelo Bass de difusión de tecnología” [24]. Se trata de un modelo clásico en Dinámica de Sistemas y citado frecuentemente [13]. Como se muestra a continuación cumple con los dos requisitos establecidos: una estructura causal sencilla unida a un comportamiento no trivial.

El modelo Bass describe un proceso de difusión de un producto tecnológicamente novedoso dentro de una población objetivo. Para ello se definen dos variables de nivel: el número de personas que potencialmente puede adquirir el producto pero que aún no lo tienen (p); y la cantidad de personas que adquirieron y ya están usando el producto (a). Estas variables se conectan a través de una variable de flujo: el número de nuevos usuarios en un período de tiempo definido (n). Estos adoptan la tecnología impulsados por dos dinámicas: por la difusión o el mercadeo del producto a través de propaganda, o por el proceso de voz a voz, es decir, la referencia que hace un usuario a un no usuario. Esta es la representación matemática discreta del modelo: a(t + Δt) = a(t) + n(t, t + Δt) * Δt p(t + Δt) = p(t) - n(t, t + Δt) * Δt n(t, t + Δt) = α*p(t) + β*p(t)*( a(t) / (a(t)+p(t) ) Siendo α y β los coeficientes de innovación (por propaganda) y de imitación (por voz a voz).

4.2. Conjunto de datos de simulación El escenario de simulación para el modelo Bass es el siguiente:    

t: [0-100]. Δt = 1. a(0) = 2 p(0) = 998 α = 0.0025 β = 0.007

El comportamiento generado se aprecia en la Fig. 3.

Las variables de nivel exhiben el típico comportamiento en S, creciente para los “usuarios actuales” y decreciente para los “potenciales”. Del mismo modo se muestra la curva de campana de la variable de flujo “nuevos usuarios”. La escala de la izquierda aplica para las variables de nivel y la de la derecha para la variable de flujo.

El máximo del flujo así como el punto de inflexión de las curvas de los niveles están ubicados en el centro de la gráfica (más precisamente en el periodo 46). Al finalizar los 100 periodos los niveles terminan con 19 usuarios potenciales y 981 usuarios actuales, es decir, una difusión prácticamente completa de la tecnología. Lo que se espera

Figura 3. Curvas de comportamiento del modelo Bass.

Fuente: Autor.

que un usuario interprete de estos resultados es que, en primer lugar, el crecimiento en “usuarios actuales” no es ilimitado sino que tiene una cota superior. En segundo lugar, que este crecimiento no es gradual sino que hay un cambio pronunciado e intempestivo ubicado en el intervalo 30-50. Estos cambios súbitos en los comportamientos usualmente no son advertidos por usuarios no-expertos.

Con lo anterior puede verse que el conjunto de datos de simulación del modelo Bass ofrece un reto a la intuición de un usuario no-experto. Se trata entonces de un conjunto de datos idóneo para aplicar la técnica de visualización dinámica propuesta.

4.3. Software para la técnica de visualización dinámica Para lograr una visión completa de la técnica de visualización animada aplicada al caso del modelo Bass se desarrolló un software. No se trata de una herramienta completa o de uso general para simular en DS. Es solamente una prueba de concepto aplicada al modelo específico con el fin de observar en acción la técnica de visualización diseñada. Para esto se utilizó Processing [12], un lenguaje de programación creado por el Medialab de MIT especialmente para desarrollar software de visualización e interacción. Processing combina la programación orientada a objetos con facilidades específicas para construir animaciones sobre las cuales pueda interactuar un usuario.

El modelo matemático es programado en el software mediante la declaración de cada una de sus variables y luego con un método denominado simulate() que en cada paso de simulación calcula los nuevos valores de las variables de acuerdo al método de Euler. Estos datos son entonces visualizados en el formato de una animación.

Processing ofrece un método, denominado draw(), que permite componer los distintos elementos visuales de cada cuadro de la animación. Se siguieron las pautas definidas en el diseño visual de la sección anterior, es decir, se utilizaron círculos para representar los valores de las variables en un momento dado. Adicionalmente se dibujan las flechas que

233

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

indican la relación de causalidad y los ciclos de realimentación propios del mapa de influencias del modelo.

Finalmente se implementaron los dos modos de interacción del usuario, el de reproducción automática y el de desplazamiento a voluntad de un indicador del tiempo.

Para ello se programaron métodos que se ejecutan de acuerdo con los movimientos y los eventos de selección del mouse. En la sección de resultados del artículo se incluye un conjunto de imágenes que muestran el software en funcionamiento.

Como se observa se trata de una herramienta de software independiente (stand-alone) y específica para el modelo Bass. Sobre la base de esta herramienta podrían hacerse adaptaciones para otros modelos. En tal caso habría que introducir en el código el modelo correspondiente, sus variables y modificar el método simulate(). Pero además habría que ajustar el diseño visual, aplicando la técnica presentada al conjunto de variables que se deseen mostrar.

Si se quisiera generalizar la aplicación de la técnica de visualización sería necesario que las herramientas de modelado existentes o algunas de las más usadas (como Ithink, Vensim y Powersim) tuviesen formatos de almacenamiento de los modelos que fuesen abiertos. Este es un asunto que ha sido objeto de discusión en la comunidad de DS por buen tiempo. Pero solo muy recientemente se observan avances en la posibilidad de un lenguaje común para intercambiar modelos entre herramientas de software, denominado XMILE [25]. Una primera y preliminar especificación se ha puesto en discusión en 2014.

Posiblemente en unos años algunas de las herramientas usuales de modelado incluyan esta especificación y esto haga posible “fomentar la innovación permitiendo a proveedores no comerciales ofrecer utilidades adicionales” [25].

5. Resultados y Discusión

Figura 4. Imagen de pantalla del software de visualización dinámica.

Fuente: Autor.

Figura 5. Primera serie. Visualización en tiempo real, segundos 1, 6, 11 y 16.

Fuente: Autor.

5.1. Resultados: Visualización dinámica del modelo Bass A continuación se presentan imágenes de la visualización dinámica que se obtuvo con el software implementado (Fig. 4). De izquierda a derecha aparecen tres círculos correspondientes a las tres variables principales del modelo: usuarios potenciales (variable de nivel), nuevos usuarios (variable de flujo) y usuarios actuales (variable de nivel). Dentro de cada círculo aparece el valor de la variable en cada momento. Se muestran además los dos ciclos de realimentación correspondientes. Por último en el área inferior del cuadro aparecen los elementos visuales e interactivos relacionados con el tiempo.

Como el resultado es una animación, se capturaron varios cuadros sucesivos de la misma y se dispusieron cronológicamente para registrarlos en la Fig. 5. La duración total de la animación es de 20 segundos y se muestran cuadros cada 4 segundos. Para evitar confusiones, se denominará tiempo real a este tiempo en el que ocurre la animación, mientras que la variable tiempo propia del modelo y la simulación se le denominará tiempo de simulación. En la parte inferior derecha de cada cuadro se muestra su correspondiente tiempo real.

Si se sigue cada imagen se pueden observar los cambios en los tamaños de los círculos de cada variable. El círculo del nivel de usuarios actuales (tercero) exhibe de manera animada un comportamiento que depende de la curva en S. Es creciente en todos los cuadros, pero es acotado, su tamaño al final de la animación se va estabilizando, como puede apreciarse en las últimas imágenes de la pantalla. La otra característica que se hace visible en esta técnica es que el cambio es más pronunciado en los cuadros intermedios (en los segundos 6 y 11). Este evento que ocurre en el tiempo real del usuario es el que precisamente produce la percepción de súbito y acentuado crecimiento y genera una experiencia visual significativa sobre un crecimiento en S. El círculo de usuarios potenciales exhibe un proceso animado inverso, decrecimiento lento, luego disminución acelerada y termina en su estabilización. Por último, el flujo describe una dinámica de lento aumento, súbito crecimiento hasta un máximo y luego gradual decrecimiento prácticamente hasta desaparecer.

Para examinar con mayor detalle las diferentes velocidades de cambio en la animación se comparan dos

234

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

Figura 6. Segunda serie. Comparación de dos intervalos de visualización.

Fuente: Autor.

intervalos de un segundo cada uno (Fig. 6). Entre el segundo 5 y el 6 el cambio es pronunciado, en cambio entre los segundos 10 y 11 el cambio es mínimo. Es así como el comportamiento en forma de “campana” que tiene el flujo (Fig. 3) se convierte a través de la animación en un momento de crecimiento acelerado seguido por un estancamiento y luego un decrecimiento acentuado. Esta es la experiencia que “vive” el usuario con el software. Para percibir este cambio repentino no requiere interpretar una visualización estática y transformarla mentalmente. Es una experiencia visual directa e inmediata. Aquí radica el poder de esta técnica de visualización dinámica para transmitir a cualquier usuario, incluso al no experto, el comportamiento del modelo.

Además, esta animación puede ser repetida por el usuario las veces que se necesite a través del botón y del deslizamiento del marcador del tiempo. Del mismo modo el usuario puede a voluntad repetir un intervalo cualquiera, por ejemplo el de cambios súbitos, arrastrando el marcador del tiempo entre el momento inicial y el final de este periodo.

Esta experimentación interactiva con la visualización dinámica permite reforzar su entendimiento de la dinámica del modelo.

5.2. Discusión de resultados El caso de aplicación de la técnica de visualización muestra con claridad la tarea a la que se enfrenta un usuario al interpretar una simulación hecha con DS. Como se indicó en la introducción se requiere una visualización sinóptica [8], es decir, que permita comprender un patrón de comportamiento global y no simplemente datos individuales. Para ello el usuario debe lograr un doble grado de integración de información. En primer lugar, necesita entender que los diferentes valores que toman las variables son resultado de la interrelación entre ellas. En el caso presentado se trata de entender que el nivel de usuarios potenciales se ve afectado por el flujo de nuevos usuarios, que este flujo incide sobre el nivel de usuarios actuales y que depende de ambos niveles. Esta estructura es la que produce los valores observados. Y en segundo lugar, el usuario debe integrar lo anterior con el paso del tiempo. Es

decir, estos valores corresponden a la evolución de la variables a través del tiempo.

En la visualización estática dominante (fig. 3) estos conjuntos de valores, incluido el tiempo, aparecen mezclados en una misma imagen. Para un usuario no experto esta visualización plantea un importante desafío interpretativo.

Por otro lado, la animación presentada en la sección anterior facilita la visualización sinóptica y supera el doble reto de integración de información. En cuanto a observar las relaciones entre las variables, la visualización dinámica en todo momento muestra solamente tres valores correspondientes a cada variable y hace visible las relaciones existente entre ellas, así promueve el primer grado de integración. Además, en la visualización dinámica el tiempo no es simplemente un dato numérico presentado en el espacio de visualización, sino que el tiempo pasa, el tiempo ocurre, es decir el usuario vive una experiencia temporal de visualización de datos. De esta manera se facilita el segundo grado de integración: entre estructura, comportamiento y tiempo. Se consigue en mayor medida una visualización sinóptica de la simulación. Esto es lo que se obtiene en el caso de aplicación cuando, por ejemplo, el usuario puede apreciar los cambios súbitos en los valores de las variables gracias a una visualización dinámica en tiempo real.

6. Conclusiones La técnica de visualización dinámica propuesta ofrece una sencilla alternativa al problema de comprensión estructura-comportamiento que tienen los usuarios no expertos en Dinámica de Sistemas. Se mostró que prácticamente la totalidad de visualizaciones disponibles en DS son estáticas. Ahora puede apreciarse por contraste con la visualización dinámica, que los gráficos estáticos imponen una barrera de comprensión al usuario. El ser humano tiene una capacidad intrínseca de percepción del tiempo que solo se utiliza si la simulación es una experiencia temporal, es decir dinámica.

La nueva técnica suministra un espacio visual en donde el usuario puede observar durante un período significativo los cambios en el comportamiento de las variables del modelo. Proporciona así una experiencia visual en tiempo real que le resulta inmediata y por tanto más directamente significativa que la interpretación secundaria de un gráfico estático. De modo que se introdujo al proceso de modelado y simulación una etapa o un momento específicamente diseñado para que el usuario realice la comprensión de los resultados de simulación con base en la estructura causal. Se brindó una técnica de visualización y un software para esta etapa, a nivel de prueba de concepto. Con ello se ataca de una nueva manera la problemática de falta de comprensión señalada.

Aun cuando el software se creó particularmente para un caso de aplicación, ofrece una visión completa de lo que podría llegar a ser una herramienta software general para visualización dinámica en DS. Para su uso generalizado habría que desarrollar una versión que incluyese algún tipo de editor del modelo o si llega a hacerse realidad el lenguaje

235

Sotaquirá-Gutiérrez et al / DYNA 81 (188), pp. 229-236. December, 2014.

de intercambio común de modelos (XMILE) entonces se podría tener como una utilidad que se enlace con modelos construidos con otras herramientas existentes.

Por último, cabe resaltar que esta técnica puede aplicarse prácticamente sin modificaciones a otras metodologías de simulación, tanto discreta como continua, que se utilizan en Ciencias, Ingeniería y Administración.

El autor expresa su agradecimiento a la Universidad de La Sabana por el sostenimiento del proyecto de investigación que hizo posible estos resultados.

[3]

[4] [5] [6] [7] [8]

[9] [10] [11] [12] [13] [14] [15]

[16] [17] [18]

[22] [23] [24]

Referencias

[2]

[20] [21]

Agradecimientos

[1]

[19]

Sterman, J.D., Modeling managerial behavior: Misperceptions of feedback in a dynamic decision making experiment. Management Science, 35 (3), pp.

321-339, 1989. http://dx.doi.org/10.1287/mnsc.35.3.321 Moxnes, E., Not only the tragedy of the commons: Misperceptions of feedback and policies for sustainable development. System Dynamics Review, 16 (4), pp.

325-348, 2000. http://dx.doi.org/10.1002/sdr.201 http://dx.doi.org/10.1002/sdr.201.abs Sterman, J.D., Does formal system dynamics training improve people’s understanding of accumulation ?. System Dynamics Review, 26 (4), pp.

316-334, 2010. http://dx.doi.org/10.1002/sdr.447 McCormick, B.H., DeFanti, T.A. and Brown, M.D., Visualization in scientific computing. Computer Graphics, 21 (6), 1987.

Card, S.K., Mackinlay, J.D. and Shneiderman, B., Readings in information visualization. San Francisco: Morgan Kaufmann, 1999.

Tufte, E.R., The visual display of quantitative information.

Connecticut: Graphics Press, 1983.

Spence, R., Information visualization: Design for interaction. New York: Adisson-Wesley, 2001.

Andrienko, G., Andrienko, N., Bak, P., Keim, D., Kisilevich, S. and Wrobel, S., A conceptual framework and taxonomy of techniques for analyzing movement. Journal of Visual Languages and Computing, 22 (3), pp.

213-232, 2011. http://dx.doi.org/10.1016/j.jvlc.2011.02.003 Aigner, W., Miksch, S., Schumann, H. and Tominski, C., Visualization of time-oriented data. London: Springer, 2011. http://dx.doi.org/10.1007/978-0-85729-079-3 Valencia, A.L., Ramírez J.M., Gómez, D. and Thomson, P., Aplicación interactiva para la educación en dinámica estructural.

DYNA, 78 (165), pp. 72-83, 2011.

Forrester, J.W., Industrial dynamics. Waltham: Pegasus Communications, 1961.

Reas, C. and Fry, B., Processing: A programming handbook for visual designers and artists. Cambridge: The MIT Press, 2007, 736 P. Sterman, J.D., Business dynamics: Systems thinking and modeling for a complex world. New York: McGraw-Hill, 2000.

Rosling, H., Making data dance. Economist, 397 (8712), pp. 1-5, 2010.

Rind, A., Aigner, W., Miksch, S., Wiltner, S., Pohl, M., Drexler, F., Neubauer, B. and Suchy, N., Visually exploring multivariate trends in patient cohorts using animated scatter plots. HCI International, pp. 139-148, 2011.

Matkovic, K., Hauser, H., Sainitzer, R. and Groller, M., Process visualization with levels of detail. IEEE Symposium on Information Visualization, pp. 67-70, 2002.

Kemal, S. and Barlas, Y., Verifying system dynamics simulation results by analytical phase plane tools. 2006 International System Dynamics Conference, 110 P, 2006.

Howie, E., Sy, S., Ford, L. and Vicente, K.J., Human – computer interface design can reduce misperceptions of feedback. System

[25]

Dynamics Review, 16 (3), pp.

151-171, 2000. http://dx.doi.org/10.1002/1099-1727(200023)16:3<151::AIDSDR191>3.0.CO;2-0 Sweeney, L.B. and Sterman, J.D., Bathtub dynamics: Initial results of a systems thinking inventory. System Dynamics Review, 16 (4), pp. 249-286, 2000. http://dx.doi.org/10.1002/sdr.198 Meadows, D., A brief and incomplete history of operational gaming in system dynamics. System Dynamics Review, 23 (2), pp. 199-203, 2007. http://dx.doi.org/10.1002/sdr.372 Forio Business Simulation, Forio Corporation. [en línea]. [Consulta: 25 de septiembre de 2014].

Available at: http://www.forio.com/services/training-simulations.

Davidsen, P., Perspectives on teaching system dynamics. 1994 International System Dynamics Conference, pp. 10-21, 1994.

Maeda, J., The laws of simplicity. Cambridge: MIT Press, 120 P, 2006.

Bass, F.M., A new product growth for model consumer durables.

Management Science, 15 (5), pp.

215-227, 1969. http://dx.doi.org/10.1287/mnsc.15.5.215 Chichakly, K., Baxter, G., Eberlein, R., Glass-Husain, W. Powers, R. and Schoenberg, W., XML Interchange language for system dynamics (XMILE) Version 1.0. OASIS Committee Specification Draft 01, 88 P, 2014 [en línea], [Consulta: 25 de septiembre de 2014].

Available at: http://docs.oasisopen.org/xmile/xmile/v1.0/csd01/xmile-v1.0-csd01.pdf

R. Sotaquirá-Gutiérrez, es Ing. de Sistemas y MaSc. en Informática en 1994 y 1999 respectivamente, de la Universidad Industrial de Santander, Colombia,; y Dr. en Ciencias Aplicadas mención Sistemología Interpretativa en 2008, de la Universidad de los Andes, Mérida, Venezuela.

Es miembro fundador del capítulo Colombiano de Dinámica de Sistemas y de la Escuela Latinoamericana de Pensamiento Sistémico y coautor de un libro en el área. Actualmente es profesor de tiempo completo de la Facultad de Ingeniería de la Universidad de La Sabana, director del programa de Ingeniería Informática y miembro del grupo de investigación en HumanCentered Design. Sus intereses de investigación incluyen: la dinámica de sistemas, el pensamiento sistémico aplicado a problemáticas organizacionales y sociales, la interacción persona-computador y el diseño de experiencias de usuario.

236

Scenarios of photovoltaic grid parity in Colombia Maritza Jiménez a, Lorena Cadavid b, Carlos J. Franco c a

Facultad de Minas, Universidad Nacional de Colombia, Medellín, Colombia. mjimenezz@unal.edu.co b Facultad de Minas, Universidad Nacional de Colombia, Medellín, Colombia. dlcadavi@unal.edu.co c Facultad de Minas, Universidad Nacional de Colombia, Medellín, Colombia. cjfranco@unal.edu.co

Received: September 6th, 2012. Received in revised form: November 1th, 2013. Accepted: November 25th, 2013.

Abstract In this article, we determined the residential Photovoltaic Grid Parity in 11 Colombian cities, by comparing grid energy prices offered by the local companies and the photovoltaic microgeneration cost for an average household. In order to do that, we developed a financial model, which considers the initial investment, cost of battery replacement, efficiency loss of photovoltaic technology and discount rate, among other variables, to determine the solar technology investment feasibility. It was found, in the base scenario, that on 2014 most of the considered cities have reached grid parity, while three of them will reach it between 2015 and 2021. Other scenarios, which consider higher discount rate, higher initial investment and slower learning curves, show similar results, where most cities will reach the grid parity before of 2028.

Keywords: Energy Microgeneration; learning curves; Grid Parity; Photovoltaic System

Escenarios de paridad de red fotovoltaica en Colombia Resumen En este artículo se determina la paridad de red fotovoltaica residencial para 11 ciudades colombianas, comparando los precios de la electricidad ofrecida por las compañías locales y el costo promedio de generación fotovoltaica para un hogar. Para ello, los autores desarrollan un modelo financiero que considera la inversión inicial, el costo del reemplazo de las baterías, la pérdida de eficiencia del sistema fotovoltaico y la tasa de descuento, entre otras variables, con el fin de determinar la viabilidad de la inversión en energía solar.

Los resultados indican, para el escenario base, que en el año 2014 la mayoría de ciudades han logrado la paridad de red, mientras tres de ellas la alcanzarán entre 2015 y 2021.Otros escenarios, que consideran mayores tasas de descuento, mayor inversión inicial y curvas de aprendizaje más lentas, muestran resultados similares, en los cuales la mayoría de ciudades alcanza la paridad de red antes del 2028.

Palabras clave: Microgeneración Eléctrica; curvas de aprendizaje, Paridad de Red, Sistemas Fotovoltaicos.

1. Introduction Colombian energy regulation, issued by the “Comisión de Regulación de Energía y Gas” (CREG), defines “autogenerator”, in the resolution 084 of 1996 [1], as the person who produces energy just for satisfy his own needs, can have the national interconnected system (SIN) support, but cannot sell the auto-generated energy to the national electricity grid.

Recently, Colombian government enacted the Renewable Energy Act [2], which aims to promote the development and use of non-conventional energy sources, especially those from renewable sources in the national energy system. It is expected this act to be formalized along the year 2014. Although this act covers microgeneration

activity, it is not defined the interval of energy generation in which the generation activity can be considered on the micro level.

In this way, a possible option for energy microgeneration is through photovoltaic solar system (from now on, PV) implementation. This alternative is known for giving significant advantages over other generation systems; it uses solar sunshine for generation (which is available all over the world) [3], has low maintenance requirements [4], allows easily to add extra generation capacity [3,4], has a high modularity [5,6] and produces zero noises [5].

However, so far it is not done a viability evaluation for the PV system technology in Colombian cities from a financial microgeneration point of view. Whereby, a comparison between the grid tariff and the PV

© The author; licensee Universidad Nacional de Colombia.

DYNA 81 (188), pp. 237-245. December, 2014 Medellín. ISSN 0012-7353 Printed, ISSN 2346-2183 Online DOI: http://dx.doi.org/10.15446/dyna.v81n188.42165

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

microgeneration cost is relevant in this territory to determine when the PV solar technology will be competitive with the grid support, which is known as Grid Parity [7].

A revision of the basic concepts about the PV technology, its components, operation, panel types and its efficiencies are presented below, followed by an explanation of the methodology, the main assumptions and the scenarios developed in the analysis, to finish with the results, discussion and conclusions about the Grid Parity State in 11 Colombian cities. Results show most of the considered cities have reached grid parity by the year 2014, while four of them will reach it at some point between 2015 and 2021.

The scope of this article is to discuss about feasibility of residential solar energy implementation, by making a financial evaluation to identify the PV grid parity time for several Colombian cities. At the end of the evaluation and from the results that we got, we highlight some policy issues that could act in benefit of the diffusion of the PV solar technology. Limitations were related with the availability of data about the solar irradiation for each city, and accurate information about the Grid Tariff offered by the local energy companies of the cities.

2. Photovoltaic solar system operation PV system development began in 1870, when William Grylls Adams and Richard Evens Day discovered photovoltaic effect by using Selenium [8]. This effect consists in the transformation of sunshine in electrical power through photovoltaic cells [8]. The development of the first solar cell was in 1954 [3,9], and its commercial use began in the 80s [4]. Besides the cell, PV system includes other elements that make possible to operate them.

Among them, the batteries are responsible for providing power in the off sunshine periods, and answer to the photovoltaics intermittency [10]. For PV systems, batteries must be replaced after a certain amount of using years.

2.1. Efficiency and panels types A critical factor in solar energy generation is the PV system efficiency, which, according to Lynn [11], refers to the percentage of solar radiation that the module can transform into electricity. In other words, with a specific power, a big efficiency implies a low panel superficial area requirements. It is important to note that PV technology can convert sunshine into direct electricity and diffuse solar irradiation [9].

Theoretical efficiency limit of silicon crystalline cells is 28% in average [8], and in practice, the reached efficiency is very close to this value. Looking for improving their efficiency [9] and decreasing its high production costs [11], new types of cells different from traditional single and poly crystalline cells have been developed during last years [8, 9,11]; this is the case of single crystal, multi-crystalline thin film, amorphous, CIGS, CdTe and organic cells. The main characteristics of these new cells are related with alternative materials for their manufacturing, which can improve its

efficiency, decrease its producing costs or make the cells lighter. However, so far the highest efficiency reached for a solar cell is still under 50% [12].

PV solar systems lose efficiency according to their use.

Most of the guarantees stipulate that in the last years of the system’s life, it will work with an 80% of efficiency [13].

According to Lynn [11], and Ramadhan and Naseeb [14], the life time of the PV solar systems is about 20 years.

2.2. Grid Parity analysis of PV solar systems The grid parity analysis consists in the comparison between the cost of electricity produced by a specific generation system (which is not connected to the national electricity grid), and the price of purchasing power from the national electricity grid [15].

In order to determining the cost of electricity produced

by PV systems and to perform a grid parity analysis, is it common to follow the Levelized Cost Of Energy (from this point on LCOE) methodology [3-7,9,14-20]. This methodology consists in a relation between the total expenditures of the solar technology (considering all its operation and maintenance requirements), and the electricity generated by the solar system, all during the PV system’s lifetime. Notice that although the evaluation is done for a specific point in time, the methodology collects information from the whole PV system’s lifetime in future.

Mathematical formulation of LCOE is presented in eq.

(1). ∑ ∑

(1)

Where LCOE: Average lifetime levelized cost of energy [$/kWh] = expenditures associated with the solar system in time = t (it includes investment, operation and maintenance expenditures and fuel expenditures) r = discount rate = electricity produced by the solar system in time = t n = lifetime of the system Note that this methodology considers the value of the money over time (which can be understood from the use of the discount rate concept in the numerator of eq. (1)), and attempts to distribute the costs among the total electricity produced. As a result, it is possible to find a kind of Net Present Value of the energy generated with the alternative system during the evaluation time. A formalization of the use of this methodology for the PV solar evaluation is developed by Hernández-Moro and Martínez-Duart [17].

Once the LCOE has been calculated, the final step of the grid parity analysis implies to compare this value with the current average price of electricity from the grid. It must be noted that this methodology does not allow considering expectations about changes on grid electricity prices over the financial evaluation horizon.

Different authors that implemented the LCOE method, have concluded that model results are specially sensible to specific variables, like local prices of electricity, PV system

238

JimĂŠnez et al / DYNA 81 (188), pp. 237-245. December, 2014.

price, solar irradiation availability, financing methods [15] and discount rate [9,14].

World market selected places include those with the best combinations of solar irradiation [9] and high electricity prices [3,6], and countries with a leadership in the PV system installation (like Germany) [4]. One of these studies include more than 150 countries, considering around the 97.7% of world population and 99.6% of global electricity consumption [3].

Main results of LCOE application conclude that countries with higher solar irradiation and higher grid tariffs are the first to reach the grid parity state; this is the case of Germany [3], Italy, Hawaii and some areas in the United States of America [6,18]. Also, countries with good enough solar irradiation availability and high grid tariff, like South Africa and Egypt, will be the ones next to reach it [3].Some analysis has been done for mature markets characterized by the absence of public policy incentives (like Italy [21], for which authors conclude that the payback time is about five to six years for residential consumption). In some not so lighting areas in USA, the technology would reach grid parity around 2020 [19]. However, other countries with not special conditions but an interest in fighting against the climate change have also been include in the analysis, like Malaysia [20] (for which authors conclude that it would take up to 16 years for the country to achieve PV grid parity).

*Values in constant 2013 dollars, excluding subsidies and contributions Figure 1. Average Price of electricity for householders in Colombia.

Source: The authors. Data from CREG [27] and SUI [28]

Fig. 1 presents the growth in the average price electricity for household users in Colombia, without considering subsidies or contributions. It shows how the residential electricity price increased by 29.8% between 2000 and 2012. This fact represents opportunities for the microgeneration in homes through alternative sources, such as PV systems.

3. Photovoltaic solar generation in the Colombian market

4. Methodology

Since 1994, the electricity sector in Colombia was liberalized, and generation, transmission, distribution and trading activities were open to competence. The final tariff, in general, consists in the sum of the prices of each activity [22-24]. Colombia has an annual demand of 62,882 GWh, with a peak demand of 9,380 MW in 2013 [25] The generation price is formed through an electricity market, in which generators sell electricity to traders and other generators, through the pool or by bilateral contracts between generators and traders. Installed generation capacity is 14.5 GW [25]. Generation depends mainly on hydro resources (74% of generation) and thermal plants (19% of generation). Despite the great potential of the country for generating electricity by using renewable sources of energy (mainly mini-hydro, wind, solar and biomass sources), alternative generation has not been adequately explored in Colombia, and large hydropower plants and thermal dominate current expansion plans.

Meanwhile, the transmission price is set by the government, and is the same for all the country. The distribution price is set by the government as well, but depends on the demand served by each distribution firm.

The demand is served by traders and it is possible to find differences in the prices of electricity among the diverse Colombian areas, not only because of the prices achieved by traders in the energy market, but because of the demand in each area.

Residential sector is the largest consumer of electricity (41% of the total energy consumed) followed by the industry sector (31% of the total energy consumed) [26].

This section explains the built model, the application case selected and the scenarios designed for the evaluation of residential PV grid parity in Colombia.

It was developed a financial model, which allows calculating the LCOE for a typical house in each city considered in this research, and comparing it with the price offered by the local utilities companies through the SIN. The model is aggregated at the monthly level, it is to mean, that it is calculated the monthly cost of generating electricity with the PV solar system, versus the cost of buying the same amount of electricity in the national electricity grid (variations intra-day for every month are not considered). This is made for every one of the months that compose the horizon. The final aggregated cost of both types of generation is brought to Net Present Value and translated in terms of equivalent tariff. As a result, model allows knowing what cities have reached the grid parity for PV solar technology, and what cities have not. The main variables of the model are explained below.

Horizon: the time for the financial evaluation was 20 years. This horizon was chosen considering the warranty period offered by equipmentâ&#x20AC;&#x2122;s suppliers in the Colombian market [13], which, as Branker et al [15] argue, it is usually a reference point for the financeable project lifetime; it is reflected in the Lynn [11] and Ramadhan and Naseeb [14] works, who use the same period for their analysis.

Note that, according to eq. (1), the shorter horizon for evaluation the higher LCOE, because there is less generated electricity for distributing the same initial investment. So, the 20-year period can be considered a conservative selection, if it is considered that other authors argue that it is

239

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

possible to include more years in the financial evaluation, such as 25 years [6,29], 25 to 30 [3], or even 25 to 40 [4].

Discount rate: given that LCOE is compared with the current electricity fee for the year 2014, without considering any possible changes on that tariff during the whole financial evaluation horizon (not even Colombian inflation projections), the discount rate should be free of inflationary effects and, at the same time, should reflect the opportunity cost regarding the closest investment option for the decision maker. For the Colombian case, we selected the risk-free rate reflected in the Fixed Term Certificates (CDT, because of its initials in Spanish), and subtracted from it the average inflation in last five years. The final discount rate can be understood as the “inflation premium” for the country.

Initial investment: it was considered a single-crystalline cell PV technology, given its high efficiency [11,12] and its market availability. The single-crystalline cells PV technology has a market price in Colombia of $USD 2430 in 2014 for a 0.6 kW module [30] (best commercial price found to date). Because the technology evaluated in this research, there are not operational expenditures to consider.

Besides, this system doesn’t need maintenance during the lifetime guarantee period [4,13].

Battery replacement: the analyzed PV system works with batteries, which must be replaced every five years. The 100 Ah batteries have a commercial price in 2014 of $USD 215 [13] (best commercial price found to date).

Learning curve: the learning curve describes the way that the technology’s price is reduced as a result of the learning obtained by doing and cumulative production [6,7,29]. For the specific case, the PV solar technology has shown that when the cumulative market production is increased by 10 times, it cost is halved; at the same time, the production is doubled each two years [31]. It means that by 2021, the price should decrease about 50% regarding the price in 2014, and an additional 25% by 2028. It is important to clarify that a global PV learning rate can be used as a local rate, because it doesn´t make a distinction between global and local results [4].

Consequently, the evaluation was made for the following time intervals: (1) [year 2014, year 2034], (2) [year 2021, year 2041] and (3) [year 2028, year 2048]. In each one of those intervals, the price of the equipment is different, according to the learning curve described.

It is important to clarify that, in our model, the decreasing ratio of the technology price (because of the experience curve), applies only for the equipment, because there is little information about experience curves for batteries. Therefore, it was considered that the price of batteries remains the same as the price in 2014 over the evaluation horizon.

The authors consider that the generated electricity depends only on the installed capacity of the PV system, the availability of sunshine in the specific place and the degradation system rate. This formulation is shown in eq.

(2).

(2) Where:

= electricity produced by the solar system in time = t Capacity: total initial electricity that can be produced for the PV system [15] cf : charge factor : degradation ratio in time = t Taking into account the current electricity market regulation in Colombia, it is assumed that the energy that a household can use from the PV system will be the minimum between the electricity demand of a typical household and the electricity generated for the PV system according to eq.

(3). It is to say, it is considered that in case of the electricity demand of a typical household in a specific city is lower than the electricity generated by the PV system, the excess electricity cannot be used or sold to the national electricity grid. Electricity demand of a typical household in each city considered in this analysis is assumed fixed over the whole evaluation horizon. Capacity, degradation rate and charge factor are explained below.

Capacity: it was considered a PV system with a capacity of 0.6 kW [30]. Notice that installed capacity remains invariable over analysis horizon, but the effective capacity decreases because of loss of efficiency of the equipment.

The consideration of larger capacities would increase the total of energy generated, but also the investment price and battery replacement cost.

Degradation ratio: it was assumed an efficiency loss for the PV system at a rate of 1% per annum (degradation ratio); i.e, each year the energy generated by the PV system is decreasing in this value in regard to the energy generated in the last year. The assumed ratio corresponds with a conservative scenario, as other authors consider a value about 0.2% to 0.5% as reasonable [15].

Charge factor: it was considered the sunshine duration in a day. The sunshine duration in a period is defined by the World Meteorological Organization (WMO) as “the sum of that sub-period for which the direct solar irradiance exceeds 120 Wm-2” [32], and is measured in hours [32].

Table 1 presents the annual average of the available sunshine for the analyzed Colombian cities, according to historical data. Therefore, the analysis considers that PV solar panel can generate electricity only during the fraction of the day in which there is sunshine, according its capacity and depending on the degradation rate.

Finally, the electricity grid tariff for comparing the LCOE calculated for each city corresponds to the average of monthly tariffs for the first half of 2014 for each specific city, without considering subsides or taxes to the electricity for final consumer (in Colombia, it corresponds to the socioeconomic level No. 4 – middle-high). Because of the high uncertainty of the path of those tariffs in the future, these values don’t increase over the time in our evaluation. The grid tariff considered for each city is presented in Table 1.

5. The Colombian case application The PV grid parity state in Colombia was calculated for 11 cities of the country. The cities were selected according to its population and sunshine factor, trying to consider both light and dark cities cases. Table 1 presents the attributes for those cities.

240

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

Table 1.

Attributes of the cities included in the analysis Av. Average 2014 Sunsh. grid tariff factor [$US/kWh] Bogotá 7,674,366 16.3% 18.2%* $ 0.18 Medellín 2,417,325 5.1% 21.2% $ 0.19 Cali 2,319,684 4.9% 21.1% $ 0.19 Barranquilla 1,206,946 2.6% 27.6% $ 0.16 Bucaramanga 526,827 1.1% 16.8% $ 0.18 Cartagena 978,600 2.1% 29.5%* $ 0.16 Cúcuta 637,302 1.4% 25.4%* $ 0.19 Ibagué 542,876 1.2% 24.1%* $ 0.20 Santa Marta 469,066 1.0% 32.0%* $ 0.16 Manizales 393,167 0.8% 17.5% $ 0.19 Riohacha 240,951 0.5% 31.3%* $ 0.17 *Measure taken from the brightest point of the city Source: population data from DANE [33], sunshine data from the Colombian Institute of Meteorology and Environmental Studies (IDEAM) [34], grid tariff data published by Sistema Único de Información de Servicios Públicos (SUI) [35] City

Population

%

Note that together, the cities group about 40% of Colombian population. Territories like Santa Marta have a high sunshine factor (32,0%), while other, like Note that, according to eq. (1), the shorter horizon for evaluation the higher LCOE, because there is less generated electricity for distributing the same initial investment. So, the 20-year period can be considered a conservative selection, if it is considered that other authors argue that it is possible to include more years in the financial evaluation, such as 25 years [6,29], 25 to 30 [3], or even 25 to 40 [4].

Discount rate: given that LCOE is compared with the current electricity fee for the year 2014, without considering any possible changes on that tariff during the whole financial evaluation horizon (not even Colombian inflation projections), the discount rate should be free of inflationary effects and, at the same time, should reflect the opportunity cost regarding the closest investment option for the decision maker. For the Colombian case, we selected the risk-free rate reflected in the Fixed Term Certificates (CDT, because of its initials in Spanish), and subtracted from it the average inflation in last five years. The final discount rate can be understood as the “inflation premium” for the country.

Initial investment: it was considered a single-crystalline cell PV technology, given its high efficiency [11,12] and its market availability. The single-crystalline cells PV technology has a market price in Colombia of $USD 2430 in 2014 for a 0.6 kW module [30] (best commercial price found to date). Because the technology evaluated in this research, there are not operational expenditures to consider.

Besides, this system doesn’t need maintenance during the lifetime guarantee period [4,13].

Battery replacement: the analyzed PV system works with batteries, which must be replaced every five years. The 100 Ah batteries have a commercial price in 2014 of $USD 215 [13] (best commercial price found to date).

Learning curve: the learning curve describes the way that the technology’s price is reduced as a result of the learning obtained by doing and cumulative production [6,7,29]. For the specific case, the PV solar technology has shown that when the cumulative market production is increased by 10 times, it cost is halved; at the same time, the production is doubled each two years [31]. It means that by

2021, the price should decrease about 50% regarding the price in 2014, and an additional 25% by 2028. It is important to clarify that a global PV learning rate can be used as a local rate, because it doesn´t make a distinction between global and local results [4].

Consequently, the evaluation was made for the following time intervals: (1) [year 2014, year 2034], (2) [year 2021, year 2041] and (3) [year 2028, year 2048]. In each one of those intervals, the price of the equipment is different, according to the learning curve described.

It is important to clarify that, in our model, the decreasing ratio of the technology price (because of the experience curve), applies only for the equipment, because there is little information about experience curves for batteries. Therefore, it was considered that the price of batteries remains the same as the price in 2014 over the evaluation horizon.

The authors consider that the generated electricity depends only on the installed capacity of the PV system, the availability of sunshine in the specific place and the degradation system rate. This formulation is shown in eq.

(2).

(2) Where: = electricity produced by the solar system in time = t Capacity: total initial electricity that can be produced for the PV system [15] cf : charge factor : degradation ratio in time = t Taking into account the current electricity market regulation in Colombia, it is assumed that the energy that a household can use from the PV system will be the minimum between the electricity demand of a typical household and the electricity generated for the PV system according to eq.

(3). It is to say, it is considered that in case of the electricity demand of a typical household in a specific city is lower than the electricity generated by the PV system, the excess electricity cannot be used or sold to the national electricity grid. Electricity demand of a typical household in each city considered in this analysis is assumed fixed over the whole evaluation horizon. Capacity, degradation rate and charge factor are explained below.

Capacity: it was considered a PV system with a capacity of 0.6 kW [30]. Notice that installed capacity remains invariable over analysis horizon, but the effective capacity decreases because of loss of efficiency of the equipment.

The consideration of larger capacities would increase the total of energy generated, but also the investment price and battery replacement cost.

Degradation ratio: it was assumed an efficiency loss for the PV system at a rate of 1% per annum (degradation ratio); i.e, each year the energy generated by the PV system is decreasing in this value in regard to the energy generated in the last year. The assumed ratio corresponds with a conservative scenario, as other authors consider a value about 0.2% to 0.5% as reasonable [15].

Charge factor: it was considered the sunshine duration in a day. The sunshine duration in a period is defined by the

241

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

World Meteorological Organization (WMO) as “the sum of that sub-period for which the direct solar irradiance exceeds 120 Wm-2” [32], and is measured in hours [32].

Table 1 presents the annual average of the available sunshine for the analyzed Colombian cities, according to historical data. Therefore, the analysis considers that PV solar panel can generate electricity only during the fraction of the day in which there is sunshine, according its capacity and depending on the degradation rate.

Finally, the electricity grid tariff for comparing the LCOE calculated for each city corresponds to the average of monthly tariffs for the first half of 2014 for each specific city, without considering subsides or taxes to the electricity for final consumer (in Colombia, it corresponds to the socioeconomic level No. 4 – middle-high). Because of the high uncertainty of the path of those tariffs in the future, these values don’t

increase over the time in our evaluation. The grid tariff considered for each city is presented in Table 1.

Bucaramanga and Manizales, have lower sunshine factors (around 16.8% and 17.5%, respectively), although have different grid tariff conditions ($USD/kWh 0.18 and $USD/kWh 0.19, respectively).

5.1. Scenarios In order to determine the project sensibility, we built a baseline scenario and stressed the model in three different ways to identify its answer to unlike system conditions and, consequently, if and when grid parity could occur in each case. Table 2 presents the value of the variables in each scenario.

Table 2 Scenarios Variable Horizon Discount rate Initial investment Funding method Battery replacement Learning curve Capacity Efficiecy Charge factor Electricity grid tariff Source: the authors

Baseline 20 years 1.39% EA $USD 2430 own equity $USD 215 (-50%, 2021), (-75% 2028) 0.6 kW -1% per year depends on each city depends on each city

Scenarios Risk averse Higher cost 20 years 20 years 20% EA 1.39% EA $USD 3645 $USD 2430 own equity own equity $USD 215 $USD 215 (-50%, 2021), (-75% (-50%, 2021), (-75% 2028) 2028) 0.6 kW 0.6 kW -1% per year -1% per year depends on each city depends on each city depends on each city depends on each city

In all scenarios, the evaluation horizon is 20 years, the price of batteries is $USD 215 and they must be replaced every five years, the equipment capacity is 0.6 kW (with a degradation ratio of 1% per year and a charge factor that depends on the specific city). For the baseline scenario, the discount rate is 1.39% EA, the learning curve occurs according to literature information [31], and the initial investment is $USD 2430 (made from the investor’s own equity).

The other four scenarios were designed by changing the discount rate, the value of the initial investment, the learning curve achieved and the funding method, thus: 1. Risk averse scenario: this scenario considers the high-risk perception associated with a new technology. In this case, investors can demand a higher discount rate in the project cash flow, in order to cover the higher perceived risk. In order to include this risk factor, a discount rate of 20% EA was considered in the risk averse scenario.

2. Higher price scenario: because the PV solar technology is not produced in Colombia, but it must be imported, there is no certainty about its final price for the consumers. Additional taxes and duties can take place at the moment of importation and legalization of the technology in the territory. Hence, this scenario considers a 50% higher technology acquisition price.

3. Slower learning curve scenario: a slower learning curve behavior was assumed in this scenario, given it is impossible to know with certainty which the global production of technology will be. This scenario considers

Slower learning curve 20 years 1.39% EA $USD 2430 own equity $USD 215 (-25%, 2021), (-50% 2028) 0.6 kW -1% per year depends on each city depends on each city

Bank funding 20 years 1.39% EA $USD 2430 bank funding $USD 215 (-50%, 2021), (-75% 2028) 0.6 kW -1% per year depends on each city depends on each city

that price decreases over time by a half of what is expected according to literature.

4. Bank funding scenario: in order to include the possible situation in which investor gets the funding from the banking system, this scenario evaluates the project considering that equipment can be funded through a bank loan. This option mitigates the initial effort that an investor must do for purchasing the equipment by using his own equity. For the model, the financial interest rate is the regular in the Colombian market for a consumer loan (about 25% EA), and the financial horizon is five years.

Reader must keep on mind that, in the four scenarios, the evaluation was made for the three-time intervals described before.

6. Results and discussion Results for each scenario are presented below by graphs, which show the time when, given the specific place conditions, Colombian cities reach the grid parity.

X axe (horizontal) represents the hours per month of sunshine, and the Y axe (vertical) represents the equivalent electricity tariff reached with the PV solar generation (in $USD per kWh). Cities are located in the 2D graph according to their sunshine and their 2014 local grid tariff.

The lines show the grid parity price for the three years considered. Note that the lines get down over time because of the learning curve considered in the model. Cities located above a specific line have reached the grid parity for the

242

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

Figure 2. Grid Parity: Baseline scenario.

Source: the authors

Figure 4. Grid Parity: higher prices scenario.

Source: The authors

Figure 5. Grid Parity: Slower learning curves scenario.

Source: The authors

Figure 3. Grid Parity: risk averse scenario.

Source: Author’s elaboration

year associated with the line, and cities below have not reached the grid parity yet. Fig. 2 shows the results for the baseline scenario.

In the baseline scenario, eight Colombian cities (Cali, Medellín, Ibagué, Cúcuta, Barranquilla, Cartagena, Riohacha and Santa Marta) have reached the Grid Parity in 2014, and other three cities (Bucaramanga, Manizales and Bogotá) will reach the grid parity before 2021.

Note that those cities with the higher number of sunshine hours reach the grid parity faster than those ones with lower sunshine. For some cities, like Medellín and Cali, the equivalent solar tariff is really close to the grid tariff (with differences about 7.5% and 7.1% regarding the grid tariff, respectively), but for cities like Santa Marta or Riohacha, the difference can be considered relevant (29.8% and 29.6% regarding the grid tariff, respectively).

For this scenario, the equivalent solar tariff is placed between $USD/kWh 0.11 and $USD/kWh 0.22 for the year 2014, USD/kWh 0.07 and $USD/kWh 0.13 for the year 2021 and USD/kWh 0.04 and $USD/kWh 0.08 for the year 2028.

Figs. 3 to 6, present the results with different system conditions. Fig. 3 shows the results for the risk averse scenario.

When it is considered a higher discount rate (20%), no city reaches the grid parity state by 2014; 10 out of 11 cities do it at some point between 2021 and 2028, and just one

city (Bucaramanga) reaches the grid parity only after 2028.

So, the more risk aversion, the more delay reaching the grid parity. Even so, with a discount rate which is more than 14 times the normal rate for a regular project, most of the Colombian cities included in this analysis achieves the grid parity before 2028.

Unlike the baseline scenario, the corresponding solar tariff for the risk averse scenario is placed between $USD/kWh 0.33 and $USD/kWh 0.63 for the year 2014, USD/kWh 0.17 and $USD/kWh 0.33 for the year 2021, and USD/kWh 0.10 and $USD/kWh 0.18 for the year 2028. It means that, in average, the equivalent tariffs are 2.56 times higher regarding the baseline scenario under the risk averse scenario’s conditions. Fig. 4 presents the results for the higher price scenario.

Fig. 4 shows the results when the technology acquisition price is 50% higher than the considered in baseline scenario.

With this assumption, the grid parity is reached by two cities in 2014, and the other nine cities reach it at some point between 2014 and 2021.

The equivalent solar tariff for the higher price scenario takes place in between $USD/kWh 0.16 and $USD/kWh 0.31 for the year 2014, USD/kWh 0.09 and $USD/kWh 0.17 for the year 2021, and USD/kWh 0.06 and $USD/kWh 0.31 for the year 2028. It means that, in average, the corresponding tariffs are 1.34 times higher than the tariff in the baseline scenario.

243

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

Figure 6. Grid Parity: Bank funding scenario.

Source: The authors

In the Fig. 5, the results for the slower learning curve scenario are shown.

The effect of reducing the speed of the learning curve by half of the expected speed doesn’t change the grid parity state for any of the analyzed cities. Like in the baseline scenario, Bogotá, Manizales and Bucaramanga reach the grid parity state between 2014 and 2021 in spite of the slower reduction on equipment’s prices.

However, the percentage difference between the equivalent solar tariff and the grid tariff is reduced for Bogotá, Manizales and Bucaramanga. In this baseline scenario, the corresponding solar tariff was 33.7%, 35.1% and 28.7% (respectively) lower than the grid tariff, while for the slower learning curves scenario, those differences are 11.0%, 13.0% and 4.4% respectively.

Fig. 6 presents the results for the Bank funding scenario.

Cali, Medellín, Ibagué, Cúcuta, Barranquilla, Cartagena, Riohacha and Santa Marta cities reach the grid parity at some point between 2014 and 2021 when equipment’s bank funding is considered. Other cities, like Bucaramanga, Manizales and Bogotá reach this state at some point between 2021 and 2028 (this point seems close to 2021).

Although the cash flow for individuals is relaxed in this scenario, the inclusion of the bank funding delays the grid parity achievement regarding the baseline scenario. This is because the financial rate is higher than the discount rate, it is mean, higher than the second most profitable alternative for individual’s investment. It supposes individual’s choices are not as profitable as the banking system is, which is truth for most of the people in developing countries, like Colombia.

7. Conclusions According to the results, grid parity is reached for eight cities (Cali, Medellín, Ibagué, Cúcuta, Barranquilla, Cartagena, Riohacha and Santa Marta) in the baseline scenario before the year 2014 (and, of course, in the slower learning curve scenario for the same year, because prices are still the same). For all the 11 cities in the baseline scenario, the slower learning curve scenario, and the higher price scenario, the grid parity is reached before the year 2021; and for all the cities in all considered scenarios before 2028 (except Bucaramanga in the risk averse scenario).

These results point out the viability for installing solar generation in Colombia in the consumer side, since in most of the scenarios and cities the grid parity will be achieved before 2021. Because the results in the baseline scenario are not affected by government mechanisms, it is possible to say that this parity grid is reached with the current state of the technology, the sunshine conditions of the territory and the electricity tariffs in the cities.

Considering a sensitive analysis on risk aversion, price of the technology, speed of learning curves and funding options, the whole spectrum of tariffs is covered for two scenarios: the baseline scenario, in the lower limit, and the risk aversion scenario, in the upper limit. For the year 2014, equivalent solar tariff for the territories with low sunshine goes from 0.33 to 0.63, which represents a difference of 2.92 times from the risk averse scenario tariff regarding the baseline scenario. This difference is getting shorter along time, and it is of 2.59 times for the year 2021, and 2.16 times for the year 2028. While for the baseline scenario some cities get the grid parity before the year 2014, for the risk aversion scenario the grid parity is reached only after the year 2028 for all cities. In this sense, the baseline scenario is an optimistic case.

It is important to consider that the project’s viability also depends on policy issues, such as subsidy mechanisms oriented to decrease the value of the initial investment, or to provide cheaper credit lines for green investors could bring forward the grid parity time of PV solar systems, which could even accelerate the learning curve and, therefore, rebounds in more proper conditions for the diffusion of this technology.

Sensitive analysis gives some clues about these policy issues. Baseline scenario can only be reached having a suitable vigilance of the technology’s prices, so that price remains at affordable levels for consumers; otherwise, with an increase in the prices, subsidies (directed to suppliers or consumers) and other government market mechanisms will be required to reduce the final price of the technology.

Related to this, it is also important to set a low interest rate in case of bank loans; it could be achieved with subsidies addressed to the funding institutions instead to suppliers.

Moreover, it is necessary to keep a low discount rate, which means to keep a low risk aversion; strategies oriented to familiarize the population with the technology (like demonstration projects or free trials, among others) can be helpful in this goal. The lower the prices of the technology, credit lines and risk aversion, the closer the baseline scenario and the faster the diffusion of the PV solar systems.

Whereby, it is important to make, as future work, different analysis of financial, tributary and fiscal incentives to improve the PV solar implementation in Colombia.

Also, analysis oriented to evaluate the acceptance of the technology by the population, which go beyond the costbenefit evaluations, could enrich the understanding of the diffusion of the PV solar systems in the Colombian market.

References [1] [2]

244

CREG, Resolución 084 de 1996. 1996, Congreso de Colombia. Ley 1715 de 2014. 2014 [Online]. Available at:

Jiménez et al / DYNA 81 (188), pp. 237-245. December, 2014.

[3] [4]

[5]

[6]

[7] [8] [9]

[10]

[11]

[12] [13] [14]

[15]

[16]

[17]

[18]

[19] [20]

[21]

http://www.secretariasenado.gov.co/senado/basedoc/ley/2006/ley_1 117_2006.html.

Breyer, C. and Gerlach, A., Global overview on grid-parity, Progress in Photovoltaics: Research and Applications, 21 (1), pp. 121-136, 2013. http://dx.doi.org/10.1002/pip.1254 Bhandari, R. and Stadler, I., Grid parity analysis of solar photovoltaic systems in Germany using experience curves, Solar Energy, 83 (9), pp.

1634-1644, 2009. http://dx.doi.org/10.1016/j.solener.2009.06.001 Blum, N.U., Wakeling, R.S. and Schmidt, T.S., Rural electrification through village grids—Assessing the cost competitiveness of isolated renewable energy technologies in Indonesia, Renewable and Sustainable Energy Reviews, 22, pp. 482-496, 2013. http://dx.doi.org/10.1016/j.rser.2013.01.049 Breyer, C., Gerlach, A., Mueller, J., Behacker, H. and Milner, A., Grid-parity analysis for EU and US regions and market segments– Dynamics of grid-parity and dependence on solar irradiance, local electricity prices and PV progress ratio, 24th European Photovoltaic Solar Energy Conference, Hamburg, Germany, pp. 21-25, 2009.

Pérez, D., Cervantes, V., Báez, M. J. and González, J., Pv Grid Parity Monitor, Eclaeron, 2012.

Chen, C.J., Physics of solar energy, Wiley, Hoboken, NJ, USA, pp.

177-208, 2011. http://dx.doi.org/10.1002/9781118172841.ch9 http://dx.doi.org/10.1002/9781118172841 Peters, M., Schmidt, T.S., Wiederkehr, D. and Schneider, M., Shedding light on solar technologies—A techno-economic assessment and its policy implications, Energy Policy, 39 (10), pp.

6422-6439, 2011. http://dx.doi.org/10.1016/j.enpol.2011.07.045 Joshi, A.S., Dincer, I. and Reddy, B.V., Performance analysis of photovoltaic systems: A review, Renewable and Sustainable Energy Reviews, 13 (8), pp.

1884-1897, 2009. http://dx.doi.org/10.1016/j.rser.2009.01.009 Lynn, P.A., Electricity from sunlight : An introduction to photovoltaics, Wiley, Hoboken, NJ, USA, pp. 25-72, 2010. http://dx.doi.org/10.1002/9780470710111 http://dx.doi.org/10.1002/9780470710111.ch2 NREL. Best research- cell efficiencies, National Renewable Energy Laboratory. 2013 [Online], [Accessed: May 13th of 2013] Available at: http://www.nrel.gov/ncpv/images/efficiency_chart.jpg..

Sersolar. Energía solar [Online], [Accessed: May 14th of 2013] Available at: http://www.sersolar.ca/..

Ramadhan, M. and Naseeb, A., The cost benefit analysis of implementing photovoltaic solar system in the state of Kuwait, Renewable Energy, 36 (4), pp.

1272-1276, 2011. http://dx.doi.org/10.1016/j.renene.2010.10.004 Branker, K., Pathak, M.J.M. and Pearce, J.M., A review of solar photovoltaic levelized cost of electricity, Renewable and Sustainable Energy Reviews, 15 (9), pp.

4470-4482, 2011. http://dx.doi.org/10.1016/j.rser.2011.07.104 Bazilian, M., Onyeji, I., Liebreich, M., MacGill, I., Chase, J., Shah, J., Gielen, D., Arent, D., Landfear, D. and Zhengrong, S., Reconsidering the economics of photovoltaic power, Renewable Energy, 53, pp.

329-338, 2013. http://dx.doi.org/10.1016/j.renene.2012.11.029 Hernández-Moro, J. and Martínez-Duart, J.M., Analytical model for solar PV and CSP electricity costs: Present LCOE values and their future evolution, Renewable and Sustainable Energy Reviews, 20, pp. 119-132, 2013. http://dx.doi.org/10.1016/j.rser.2012.11.082 Swift, K.D., A comparison of the cost and financial returns for solar photovoltaic systems installed by businesses in different locations across the United States, Renewable Energy, 57, pp. 137-143, 2013. http://dx.doi.org/10.1016/j.renene.2013.01.011 Reichelstein, S., and Yorston, M., The prospects for cost competitive solar PV power, Energy Policy, 55, pp. 117-127, 2013. http://dx.doi.org/10.1016/j.enpol.2012.11.003 Lau, C.Y., Gan, C.K. and Tan, P.H., Evaluation of solar photovoltaic Levelized Cost of Energy for PV grid parity analysis in Malaysia, International Journal of Renewable Energy Resources, 4 (1), pp.2834, 2014.

Chiaroni, D., Chiesa, V., Colasanti, L., Cucchiella, F., D’Adamo, I. and Frattini, F., Evaluating solar energy profitability: A focus on the role of self-consumption, Energy Conversion and Management, 88,

[22] [23] [24] [25] [26]

[27] [28] [29]

[30] [31] [32] [33]

[34]

[35]

pp.

317-331, 2014. http://dx.doi.org/10.1016/j.enconman.2014.08.044 Congreso de Colombia. Ley 142 de 1994 - Ley de Servicios Públicos Domiciliarios, 1994.

Congreso de Colombia. Ley 143 de 1994 - Ley Eléctrica, 1994.

Comisión de Regulación de Energía y Gas - CREG, 2008, Resolución No. 097 de 2008.

XM. NEÓN - Información Inteligente. 2013 [Online]. [Accessed: February 1 of 2013], Available at: http://sv04.xm.com.co/neonweb/.

Unidad de Planeación Minero Energética (UPME). Balance Minero Energético - 2010. 2010 [Online], [Accessed: February 1 of 2013], Available at: http://www.upme.gov.co/GeneradorConsultas/Consulta_Balance.asp x?IdModulo=3..

Comisión de Regulación de Energía y Gas - CREG. 2013 [Online].

[Accessed: November 19th of 2013], Available at: http://www.creg.gov.co/html/i_portals/index.php.

Sistema Único de Información de Servicios Públicos - SUI. 2013 [Online], [Accessed: November 19th of 2013], Available at: http://www.sui.gov.co/SUIAuth/logon.jsp.

Lund, P. D. Boosting new renewable technologies towards grid parity – Economic and policy aspects, Renewable Energy, 36 (11), pp.

2776-2784, 2011. http://dx.doi.org/10.1016/j.renene.2011.04.025 Alta Ingeniería. Energía Solar en Colombia. Energía Solar en Colombia y Renovables.

2014.

[Online], Available: http://www.altaingenieriaxxi.com/.

Partain, L.D. and Fraas, L.M., Wiley Series in Microwave and optical engineering : Solar cells and their applications, 2nd ed., Wiley, Hoboken, NJ, USA, pp. 3-153, 2010.

World Meteorological Organization, Guide to meteorological instruments and methods of observation, 7th ed. (8), Chairperson, Publications Board, Geneva 2, pp. I.8-1 to I.9-1, 2008.

Departamento Nacional de Estadística, Estimación y proyección de población nacional, departamental y municipal por área 1985-2020, DANE Proyecciones Poblac. [Online], [Accessed: November 21th of 2013].

Available at: http://www.dane.gov.co/index.php/poblacion-ydemografia/proyecciones-de-poblacion.

Instituto de Hidrología, Meteorología y Estudios Ambientales (IDEAM), Promedios Climatológicos 1981-2010., Características Clim. Colomb. [Online]. [Accessed: February 14th of 2014], Available at: http://institucional.ideam.gov.co/jsp/812..

Sistema Único de Información de Servicios Públicos - SUI. 2014 [Online], [Accessed: November 19th of 2013], Available at: http://reportes.sui.gov.co/fabricaReportes/frameSet.jsp?idreporte=el e_com_096.

M. Jiménez-Zapata, received the BSc. Eng in Management Engineering in 2014, and currently she is a student in the Energetic Systems Master program at Universidad Nacional de Colombia, Medellin, Colombia. She works in several projects in the electricity market field, with emphasis on photovoltaic renewable energy microgeneration systems since 2012, at Universidad Nacional de Colombia, Medellín, Colombia.

L. Cadavid, received the BSc. Eng in Management Engineering in 2006, and the MSc degree in Systems Engineering in 2010, both from Universidad Nacional de Colombia, Medellín, Colombia. Nowadays, she is a PhD candidate in Dr, Systems Engineering program at the same place.

Her research interests include Innovations Diffusion and Agent-Based Modeling and Simulation, as well as cleaner technologies.

Carlos J. Franco, He has a degree in Civil Engineering, a MSc. degree in Water Resources Management, and a PhD in Engineering, all from the Universidad Nacional de Colombia, Medellín, Colombia. Is a Full Professor in the Department of Computer and Decision Sciences at Universidad Nacional de Colombia, Medellín, Colombia. He is professor in subjects such as complex systems, system modeling and energy markets.

His research area is on energy systems analysis, including policy evaluation and strategy formulation. His recent work includes low carbon economies, demand response, electric market's integration and bio-fuels, among others.

245

DYNA

81 (188), December 2014 is an edition consisting of 250 printed issues which was finished printing in the month of December of 2014 in Todograficas Ltda. MedellĂ­n - Colombia The cover was printed on Propalcote C1S 250 g, the interior pages on Hanno Mate 90 g.

The fonts used are Times New Roman, Imprint MT Shadow

• Corrosion resistance of hybrid films applied on tin plate: Precursor solution acidified with nitric acid (pH=3) • Plane geometry drawing tutorial • Analytical model of signal generation for radio over fiber systems • Recycling of agroindustrial solid wastes as additives in brick manufacturing for development of sustainable construction materials • The use of gypsum mining by-product and lime on the engineering properties of compressed earth blocks • Practical lessons learnt from the application of X-ray computed tomography to evaluate the internal structure of asphalt mixtures • Rail vehicle passing through a turnout: Influence of the track elasticity • Evolution of the passive harmonic filters optimization problem in industrial power systems • Restricting the use of cars by license plate numbers: A misguided urban transport policy • Active vibration control in building-like structures submitted to earthquakes using multiple positive position feedback and sliding modes • Analysis of customer satisfaction using surveys with open questions • Analysis of the economic impact of environmental biosafety works projects in healthcare centres in Extremadura (Spain) • Assessing the performance of a differential evolution algorithm in structural damage detection by varying the objective function • Efficient reconstruction of Raman spectroscopy imaging based on compressive sensing • Apply multicriteria methods for critical alternative selection • Flatness-based fault tolerant control • Dynamic wired-wireless architecture for WDM stacking access networks • Influence of osmotic pre-treatment on convective drying of yellow pitahaya • Evaluation of thermal behavior for an asymmetric greenhouse by means of dynamic simulations • A comparative study TiC/TiN and TiN/CN multilayers • Developing a fast cordless soldering iron via induction heating • Selecting working fluids in an organic Rankine cycle for power generation from low temperature heat sources • Approach to biomimetic design. Learning and application • Structural control using magnetorheological dampers governed by predictive and dynamic inverse models • Design of boundary combined footings of rectangular shape using a new model • Effect of additives on diffusion coefficient for cupric ions and kinematics viscosity in CuSO4 H2SO4 solution at 60°C • Influence of silicon on wear behaviour of “Silal” cast irons • Effect of cationic polyelectrolytes addition in cement cohesion • A new dynamic visualization technique for system dynamics simulations • Scenarios of photovoltaic grid parity in Colombia

DYNA

Publication admitted to the Sistema Nacional de Indexación y Homologación de Revistas Especializadas CT+I - PUBLINDEX, Category A1

• Resistencia a la corrosión de películas híbridas sobre láminas de estaño: Solución precursora acidificada con ácido nítrico (pH=3) • Tutorial de dibujo geométrico • Modelo analítico de generación de señales para sistemas radio sobre fibra • Reciclaje de residuos sólidos agroindustriales como aditivos en la fabricación de ladrillos para el desarrollo sostenible de materiales de construcción • El uso de residuos de minería de yeso y cal sobre las propiedades de ingeniería de los bloques de tierra comprimida • Lecciones prácticas aprendidas a partir de la aplicación de la tomografía computarizada de rayos-x para evaluar la estructura interna de mezclas asfálticas • Influencia de la elasticidad de vía al circular por un desvío ferroviario • Evolución del problema de optimización de ls filtros pasivos de armónicos en sistemas eléctricos industriales • Restricción vehicular según número de patente: Una política de transporte errónea • Control activo de vibraciones en estructuras tipo edificio sometidas a sismos utilizando múltiple retroalimentación positiva de la posición y modos deslizantes • Análisis de la satisfacción de cliente mediante el uso de cuestionarios con preguntas abiertas • Análisis del impacto económico de la bioseguridad ambiental en proyectos de obras en centros sanitarios de Extremadura (España) • Valoración del desempeño de un algoritmo de evolución diferencial en detección de daño estructural considerando diversas funciones objetivo • Reconstrucción eficiente de imágenes a partir de espectroscopia Raman basada en la técnica de sensado compresivo • Selección de alternativas críticas aplicando un enfoque multicriterio • Control tolerante a fallas basado en planitud • Arquitectura dinámica fija-móvil para redes de acceso WDM apiladas • Influencia de un pre-tratamiento osmótico sobre el secado convectivo de pitahaya amarilla • Evaluación del comportamiento térmico de un invernadero asimétrico mediante simulaciones dinámicas • Un estudio comparativo de las multicapas TIC/TiN y TiN/CN • Desarrollo de un cautín inalámbrico rápido a través de calentamiento por inducción • Seleccionando fluidos de trabajo en ciclos Rankine para generación de energía a partir de fuentes de calor de baja temperatura • Aproximación al diseño biomimético. Aprendizaje y aplicación • Control estructural utilizando amortiguadores magnetoreológicos gobernados por un modelo predictivo y por un modelo inverso dinámico • Diseño de zapatas combinadas de lindero de forma rectangular utilizando un nuevo modelo • Efecto de aditivos en el coeficiente de difusión de iones cúpricos y la viscosidad cinemática en solución CuSO4 H2SO4 a 60° • Influencia del silicio en el comportamiento al desgaste de las fundiciones tipo “Silal” • Efecto de la adición de polielectrólitos catiónicos en la cohesión del cemento • Una nueva técnica de visualización dinámica para simulaciones en dinámica de sistemas • Escenarios de paridad de red fotovoltaica en Colombia

Red Colombiana de Revistas de Ingeniería