Vol. 6 N.4 - Journal of Aerospace Technology and Management by Journal of Aerospace Technology and Management

Journal of Aerospace Technology and Management

JOURNAL OF

AEROSPACE TECHNOLOGY AND MANAGEMENT Vol. 6 N. 4 Oct./Dec. 2014 ISSN 1984-9648 ISSN 2175-9146 (online)

www.jatm.com.br

V. 6, n. 4, Oct./Dec., 2014

Journal of Aerospace Technology and Management

GENERAL INFORMATION Journal of Aerospace Technology and Management (JATM) is a techno-scientific publication serialized, published by Departamento de Ciência e Tecnologia Aeroespacial (DCTA) and aims to serve the international aerospace community. It contains articles that have been selected by an Editorial Committee composed of researchers and technologists from the scientific community. The journal is quarterly published, and its main objective is to provide an archival form of presenting scientific and technological research results related to the aerospace field, as well as promote an additional source of diffusion and interaction, providing public access to all of its contents, following the principle of making free access to research and generate a greater global exchange of knowledge. JATM is added/indexed in the following databases: • CAS • CLASE/PERIÓDICA • DOAJ • EBSCO • EZB • GOOGLE SCHOLAR • J-GATE • LATINDEX

• LIVRE • PERIÓDICOS CAPES • PKP • REDALYC • SCOPUS • SOCOL@R • SUMÁRIOS.ORG • ULRICHSWEB

Correspondence All correspondence should be sent to: Dr Ana Cristina Avelar Journal of Aerospace Technology and Management Instituto de Aeronáutica e Espaço Praça Mal. Eduardo Gomes, 50 - Vila das Acácias CEP 12228-901 São José dos Campos/ São Paulo/Brazil Contact Phone: (55) 12-3947- 5115/5004 E-mail: editor@jatm.com.br Web: http://www.jatm.com.br Published by: Departamento de Ciência e Tecnologia Aeroespacial Distributed by: Instituto de Aeronáutica e Espaço Editing, proofreading and standardization: Zeppelini Editorial Printing: RR Donnelley Edition: 500 São José dos Campos, SP, Brazil ISSN 1984-9648

In WEB QUALIS System, JATM is classified as B3 and B4 in the Interdisciplinary and Engineering III areas respectively. The journal uses CROSSCHECK to prevent plagyarism and all published articles contain DOI numbers attributed by CROSSREF. JATM is an official publication of AAB - Associação Aeroespacial Brasileira and is affiliated to ABEC - Associação Brasileira de Editores Científicos.

JATM IS SUPPORTED BY:

Journal of Aerospace Technology and Management Vol. 6, n. 4 (Oct./Dec. 2014) – São José dos Campos: Zeppelini Editorial, 2014 Quartely issued Aerospace sciences Technologies Aerospace engineering CDU: 629.73

HISTORICAL NOTE: JATM was created in 2009 after the initiative of the Director of Instituto de Aeronáutica e Espaço (IAE), Brigadeiro Engenheiro Francisco Carlos Melo Pantoja. From September 2011, it has been edited by the Departamento de Ciência e Tecnologia Aeroespacial (DCTA), and it also started to be financially supported by Fundação Conrado Wessel (FCW). In order to reach the goal of becoming a journal that represents knowledge in science and aerospace technology, JATM searched for partnerships with others institutions in the same field from the beginning. In January 2014, an important step was achieved and JATM merged with Journal of Aerospace Engineering and Applications (JAESA) becoming an official publication of Associação Aeroespacial Brasileira (AAB).

ISSN 1984-9648 ISSN 2175-9146 (online)

JOURNAL OF AEROSPACE TECHNOLOGY AND MANAGEMENT Vol. 6 No. 4 - Out./Dec. 2014 EDITORS IN CHIEF

EXECUTIVE EDITOR

ASSISTANT EDITOR

Ana Cristina Avelar Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil editor@jatm.com.br

Ana Marlene F. Morais Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil secretary@jatm.com.br

Roberto Gil Annes da Silva Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil submission@jatm.com.br

Angelo Passaro Instituto de Estudos Avançados São José dos Campos/SP – Brazil

Eduardo Morgado Belo Escola de Engenharia de São Carlos São Carlos/SP – Brazil

Mischel Carmen N. Belderrain Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

Antonio Pascoal Del’Arco Jr Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Francisco Cristovão L. Melo Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Rita de Cássia L. Dutra Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Carlos Henrique Netto Lahoz Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

João Marcos T. Romano Universidade Estadual de Campinas Campinas/SP – Brazil

Waldemar Castro Leite Filho Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Antônio F. Bertachini Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil antonio.prado@inpe.br

SCIENTIFIC COUNCIL

José Nivaldo Hinckel Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

ASSOCIATE EDITORS ACOUSTICS

APPLIED COMPUTATION

CERAMIC MATERIALS

Marcello A. Faraco de Medeiros Escola de Engenharia de São Carlos São Carlos/SP – Brazil

Romis R. F. Attux Universidade Estadual de Campinas Campinas/SP – Brasil

CIRCUITRY

Bert Pluymers Katholieke Universiteit Leuven Leuven – Belgium

AERODYNAMICS

Acir Mércio Loredo Souza Universidade Federal do Rio Grande do Sul Porto Alegre/RS – Brazil João Luiz F. Azevedo Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

AEROSPACE METEOROLOGY Gilberto Fisch Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil Willian W. Vaughan University of Alabama Huntsville/AL – USA

Leandro Baroni Universidade Federal do ABC Santo André/SP – Brazil

ASTRODYNAMICS

José Maria Fonte Ferreira Universidade de Aveiro Aveiro – Portugal Altamiro Susin Universidade Federal do Rio Grande do Sul Porto Alegre/RS – Brazil

Othon Cabo Winter Faculdade de Engenharia de Guaratinguetá Guaratinguetá/SP – Brazil

Raimundo Freire Universidade Federal de Campina Grande Campina Grande/PB – Brazil

Anna Guerman Universidade da Beira Interior Covilhã – Portugal

COMPUTATIONAL FLUID DYNAMICS

Vivian Martins Gomes Faculdade de Engenharia de Guaratinguetá Guaratinguetá/SP – Brazil Josep J. Masdemont Universitat Politecnica de Catalunya Barcelona – Spain

Joern Sesterhenn Technische Universität Berlin Berlin – Germany John Cater University of Auckland Auckland – New Zealand

Paulo Celso Greco Escola de Engenharia de São Carlos São Carlos/SP – Brazil

COMPOSITES

METALLIC MATERIALS

ROBOTICS AND AUTOMATION

DEFENSE SYSTEMS

PHOTONICS

Sadek Crisostomo Absi Alfaro Universidade de Brasília Brasília/DF – Brazil

Edson Cocchieri Botelho Faculdade de Engenharia de Guaratinguetá Guaratinguetá/SP – Brazil Adam S. Cumming Defence Science and Technology Laboratory Salisbury/Wiltshire – England

ENERGETIC MATERIALS Elizabeth da Costa Mattos Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

José Leandro Andrade Campos Universidade de Coimbra Coimbra – Portugal

FLUID DYNAMICS AND TURBULENCE

Vassilis Theofilis Universidad Politécnica de Madrid Madrid – Spain

GUIDANCE, NAVIGATION AND CONTROL Arun Misra McGill University Montreal – Canada

Daniel Alazard Institut Supérieur de l’Aéronautique et de l’Espace Toulouse – France David Murray–Smith University of Glasgow Glasgow – Scotland

MANAGEMENT SYSTEMS Adiel Teixeira de Almeida Universidade Federal de Pernambuco Recife/PE – Brazil

Antonio Henriques de Araújo Jr Universidade Estadual do Rio de Janeiro Resende/RJ – Brazil

Kyriakos I. Kourousis University of Limerick Limerick - Ireland

Álvaro Damião Instituto de Estudos Avançados São José dos Campos/SP – Brazil

POLIMERIC MATERIALS Cristina Tristão de Andrade Instituto de Macromoléculas Rio de Janeiro/RJ – Brazil Mirabel Cerqueira Rezende Instituto de Ciência e Tecnologia São José dos Campos/SP – Brazil

PROCESSING OF AEROSPACE MATERIALS Alexandre Queiroz Bracarense Universidade Federal de Minas Gerais Belo Horizonte/MG – Brazil

PROPULSION AND COMBUSTION

Fernando de Souza Costa Instituto Nacional de Pesquisa Espacial São José dos Campos/SP – Brazil

RADARS AND TRACKING SYSTEMS Cynthia C.M. Junqueira Instituto de Aeronáutica e Espaço São José de Campos/SP – Brazil

Hugo H. Figueroa Universidade Estadual de Campinas Campinas/SP – Brazil Marc Lesturgie Office National d’Etudes et de Recherches Aérospatiales Palaiseau – France

André Fenili Universidade Federal do ABC Santo André/SP – Brazil

STRUCTURES

Sérgio Frascino M. Almeida Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

SYNTHESIS AND CHARACTERIZATION OF AEROSPACE MATERIALS

Gilson da Silva Instituto Nacional da Propriedade Industrial Rio de Janeiro/RJ – Brazil Roberto Costa Lima Instituto de Pesquisas da Marinha Rio de Janeiro/RJ – Brazil

THERMAL SCIENCES

Márcia B. H. Mantelli Universidade Federal de Santa Catarina Florianópolis/SC – Brazil Renato Machado Cotta Universidade Federal do Rio de Janeiro Rio de Janeiro/RJ – Brazil

VIBRATION AND STRUCTURAL DYNAMICS Carlos Cesnik University of Michigan Ann Arbor/MI – USA

Luiz Carlos S. Góes Instituto Tecnológico da Aeronáutica São José dos Campos/SP – Brazil Valder Steffen Junior Universidade Federal de Uberlândia Uberlândia/MG – Brazil

EDITORIAL PRODUCTION Glauco da Silva Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Mauricio Andrés Varela Morales Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

Rosely A. Montoro Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Lucia Helena de Oliveira Depart. Ciência e Tecnologia Aeroespacial São José dos Campos/SP – Brazil

Mônica E. Rocha de Oliveira Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Rosilene Maria M. Costa Instituto de Estudos Avançados São José dos Campos/SP – Brazil

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, 2014

ISSN 1984-9648 | ISSN 2175-9146 (online)

CONTENTS EDITORIAL 361 Mathematics of Nanosatellites Georgi Smirnov, Antonio Fernando Bertachini de Almeida Prado ORIGINAL PAPERS 363 Experimental Investigation of a Monopropellant Thruster Using Nitrous Oxide Luciano Hennemann, José Carlos de Andrade, Fernando de Souza Costa 373 Risk Assessment for Hexamine Nitration into RDX Erick Braga Ferrão Galante, Daniele Mesquita Bordalo da Costa, Assed Naked Haddad, Isaac José Antônio Luquetti dos Santos 389 Microstructural Analysis of Co-Free Maraging Steel Aged Aline Castilho Rodrigues, Heide Heloise Bernardi, Jorge Otubo 395 Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients Hamid Farrokhfal, Ahmad Reza Pishevar 407 On Using Quasi-Newton Algorithms of the Broyden Class for Model-to-Test Correlation Jan Klement 415 Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover Luca De Filippis, Enrico Gaia, Giorgio Guglieri, Marco Re, Claudia Ricco 431 A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception Valentino Lau, Fabiano Luis de Sousa, Roberto Luiz Galski, Evandro Marconi Rocco, José Carlos Becceneri, Walter Abrahão dos Santos, Sandra Aparecida Sandri 447 FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study Arineiza C. Pinheiro, Adenilso Simão, Ana Maria Ambrosio 462 The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry Fabiano Armellini, Paulo Carlos Kaminski, Catherine Beaudry COMMUNICATION 475 Development Status of L75: A Brazilian Liquid Propellant Rocket Engine Daniel Soares de Almeida, Cristiane Maria de Moraes Pagliuco ADHOC REFEREES 485 ADHOC Referees INSTRUCTIONS TO AUTHORS 487 Instructions to Authors

doi: 10.5028/jatm.v6i4.442

EDITORIAL Mathematics of Nanosatellites Georgi Smirnov1, Antonio Fernando Bertachini de Almeida Prado2

he role of small inexpensive satellites continuously grows in the modern space exploration. Their use can significantly reduce the cost of the mission. However, the control of such satellites is a challenge. One of the major issues here is that such satellites usually do not possess a complex attitude control system, and three-axis stabilization might be unavailable. As a consequence, the thrust vector of the orbit control system cannot be arbitrary oriented in space and rather involved mathematical methods are needed in order to compensate the control system’s simplicity. The most frequently used simple and lightweight passive systems of one-axis stabilization are spin, passive magnetic or aerodynamic stabilization. In this case, one or two orbit control thrusters can be installed along the stabilized axis, so the orientation of the thrust vector at any given moment of time is determined by the orientation of stabilized axis. A very similar situation we face if the satellite uses the radiation pressure from the Sun as a motive force. Recently, a serious progress was achieved in the analysis and solution of the respective control problems. For example, dampers which use magnetic hysteresis rods in order to dissipate the energy of undesired angular motions — occurred during deployment or caused by perturbations — are used in attitude control systems of small satellites since the 1960s (Fischell and Mobley, 1964). The mathematical modeling of such systems is quite a difficult task, since the majority of existent hysteresis models result in differential equations with discontinuous right-hand side. The analysis of dynamics for attitude control systems with magnetic hysteresis dampers and optimization of their parameters have been done in Sarychev et al. (1988) and Guerman et al. (1989), and the results of these studies have been implemented in real missions (Ovchinnikov et al., 2000; Santoni and Zelli, 2009). However, these studies lacked an accurate theoretical basis for application of

averaging methods to such problems. Recently, an adequate mathematical approach has been developed in Gama et al. (2011, 2013). It allows one to apply averaging technique to discontinuous systems and, therefore, rigorously justifies the main results from Sarychev et al. (1988) and Guerman et al. (1989). Another example concerns formation flying problems. A formation of nano- or picosatellites can perform a mission usually destined to big satellites. The main problem of a formation flying mission design is the maintenance of the required spatial configuration of satellites. Many works are focused on compensation of the satellites’ relative drift caused by the J2 harmonic of the Earth’s gravitational potential. In Guerman et al. (2012), a possibility to obtain a periodic relative motion of the chief and deputy satellites has been demonstrated for several types of single-input control, including the control oriented, along the geomagnetic field, and the control along an axis fixed in the absolute space. In each case, sufficient controllability conditions have been deduced. These conditions can be concisely formulated as follows: the vector of control direction should have non-zero components both in the orbital plane and along the normal to the orbit. To describe the relative motion of formation flying spacecraft, the Schweighart and Sedwick (2002) modification of the Hill-Clohessy-Wiltshire equations was used. In Guerman et al. (2014), the ideas from Guerman et al. (2012) were completed with a Newton-type method, which allows one to construct an exact trajectory and corresponding control law based on those obtained from the linearized model. Being implemented, such an iterative procedure significantly reduced the modeling error and confirmed the validity of results theoretically previously proved for the linearized system. Problems of orbital control become more involved if a gravitational field created by two bodies is considered, for example, orbital maneuver in the vicinity of Lagrangian

1.Universidade do Minho – Braga – Portugal 2.Instituto Nacional de Pesquisas Espaciais – São José dos Campos/SP – Brazil Author for correspondence: Gueorgui Smirnov – Deptartamento de Matemática e Aplicações – Universidade do Minho – Campus de Gualtar, 4710-057 – Braga – Portugal | Email: smirnov@math.uminho.pt

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.361-362, Oct.-Dec., 2014

362

Smirnov, G. and Prado, A.F.B.A.

points is a challenge. The MEC/MCTI/CAPES/CNPq/FAPs project “Orbital maneuvers of nano-satellites with constraint orientation of the trust”, under the leadership of Antonio Fernando Bertachini de Almeida Prado and Georgi Smirnov, deals with the problems of the described type and it is focused on the development of analytical and computational methods of control system design. The most important expected results concern orbital corrections and de-orbiting of nano-satellites. The main mathematical tools used by the team members are averaging techniques, the Lyapunov function method combined with local controllability conditions, and the ideas from the theory of differential games. These instruments allow the team members to analytically solve some of the problems and to develop effective numerical methods when the problem cannot be analytically solved. For example,

the team members analytically and numerically studied the de-orbiting problem for a satellite equipped with an inflatable balloon for a large range of orbital inclinations. They developed a new de-orbiting control algorithm and obtained an estimate for the de-orbiting time. Those topics are helping to define new lines of research, as well as new topics for Thesis and Dissertations in the Space Engineering and Technology graduate school at the Instituto Nacional de Pesquisas Espaciais (INPE). We expect an increase in the production of papers in well known journals in the field in the next years, thanks to the state of the art research performed by the group of researches and students involved in this project. It is also expected an increase in the number of students looking to perform research in actual and important topics under development in those areas at INPE.

REFERENCES Fischell, R. and Mobley, F.F., 1964, “A system for passive gravitygradient stabilization of earth satellites”, Guidance and Control, Vol. 2, pp. 37–71. Gama, R., Guerman, A. and Smirnov, G., 2011, “On the asymptotic stability of discontinuous systems analysed via the averaging method”, Nonlinear Analysis: Theory, Methods and Applications A, Vol. 74, No. 4, pp. 1513–1522. doi:10.1016/j.na.2010.10.024. Gama, R., Guerman, R., Seabra, A. and Smirnov., 2013, “Averaging Methods for Design of Spacecraft Hysteresis Damper”, Mathematical Problems in Engineering, Vol. 2013, Article ID 483457, 7 p. doi: 10.1155/2013/483457. Guerman, A.D., Ovchinnikov, M.Y., Pen’kov, V.I. and Sarychev, V.A., 1989, “Non-resonant motions of a satellite with hysteresis rods under conditions of gravity orientation”, Mechanics of Solids, Vol. 24, pp. 1–11. Guerman, A., Ovchinnikov, M., Smirnov, G. and Trofimov, S., 2012, “Closed relative trajectories for formation flying with single-input control”, Mathematical Problems in Engineering, Vol. 2012, 20 p. doi:10.1155/2012/967248.

Guerman, A., Ovchinnikov, M., Smirnov, G. and Trofimov, S., 2014, “High-precision single-input control of relative motion in spacecraft formation”, Acta Astronautica, Vol. 94, No. 1, pp. 375-382. doi:10.1016/j.actaastro.2013.02.014. Ovchinnikov, M., Pen’kov, V., Norberg, O. and Barabash, S., 2000, “Attitude control system for the first sweedish nanosatellite MUNIN”, Acta Astronautica, Vol. 46, No. 2–6, pp. 319–326. doi:10.1016/ S0094-5765(99)00226-X. Santoni, F. and Zelli, M., 2009, “Passive magnetic attitude stabilization of the UNISAT-4 microsatellite”, Acta Astronautica, Vol. 65, No. 5-6, pp. 792–803. doi:10.1016/j.actaastro.2009.03.012. Sarychev, V.A., Penkov, V.I., Ovchinnikov, M.Y. and Guerman, A.D., 1988, “Motions of a gravitationally stabilized satellite with hysteresis rods in a polar orbit”, Cosmic Research, Vol. 26, No. 5, pp. 561–574. Schweighart, S.A. and Sedwick, R.J., 2002, “High-fidelity linearized J2 model for satellite formation flight”, Journal of Guidance, Control, and Dynamics, Vol. 25, No. 6, pp. 1073-1080. doi:10.2514/2.4986.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.361-362, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.382

Experimental Investigation of a Monopropellant Thruster Using Nitrous Oxide Luciano Hennemann1, José Carlos de Andrade1, Fernando de Souza Costa1

ABSTRACT: Nitrous oxide is a monopropellant with potential for use in propulsion systems for orbit correction, positioning and attitude control of satellites. This paper presents experimental results with a 2N thruster prototype employing nitrous oxide gas decomposed by a rhodium oxide catalyst supported on alumina. Initially, the thruster design, the test bench and test procedures are described. Then the main experimental results, including thrust curves, mass flow rates, pressures and temperatures along the thruster are presented for pulsed operation. At the end specific impulses, characteristic velocities and thruster efficiencies are determined, thus indicating a catalytic decomposition efficiency of about 88%. KEYWORDS: Satellite propulsion, Nitrous oxide, Rhodium oxide catalyst, Propulsive efficiency.

INTRODUCTION Several factors, such as atmospheric drag, the gravitational attractions of the Moon and the Sun, the non-uniformity of the Earth gravitational field and solar radiation contribute to change the attitude and move satellites from their original orbits (Sutton, 2001). For example, in geo-stationary satellites, Luni-solar perturbations cause a north-south drift that requires corrections of 50 m/s/year and the variation of the Earth gravitational field can generate an east-west drift of 1 degree in 60 days (Stark and Swinerd, 2003). Therefore, the use of an orbit correction and attitude control system is necessary for continuity of the normal service conditions of a satellite. At first, the control system of the rocket launcher provides input for the upper propulsion stages, for initial stabilization in orbit, alignment and firing of small thrusters (actuators). These are responsible for positioning and attitude control of the satellite in its initial orbit, having sensors, algorithms and actuators to aid in these tasks. The sensors determine the satellite’s instantaneous position in orbit and provide reference to an algorithm that controls the thrusters to reposition the satellite (Plumlee and Steciak, 2004; Sanscrainte, 1961). Liquid propellant thrusters can generate high specific impulses and thrusts, and allow precise control of thrust and exhaustion kinetic energy. In general, these systems use pressurizing tanks with an inert gas (nitrogen or helium). In bipropellant systems, the fuel and oxidant are injected, mixed and burned in the combustion chamber to form gaseous products at elevated temperatures that are ejected through a nozzle to generate thrust. Monopropellant thrusters use a single propellant decomposed by a catalyst, or by heating, in order to form hot gases to be ejected at high speeds through the nozzle.

1.Instituto Nacional de Pesquisas Espaciais – Cachoeira Paulista/SP – Brazil. Author for correspondence: Luciano Hennemann | Laboratório Associado de Combustão e Propulsão – Instituto Nacional de Pesquisas Espaciais | Caixa Postal 1, Cachoeira Paulista/SP | CEP: 12.630-000 – Brazil | Email: hennemann@lcp.inpe.br Received: 06/10/2014 | Accepted: 10/02/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

364

Hennemann, L., Andrade, J.C. and Costa, F.S.

Both systems can operate in pulsed mode for attitude control and orbit correction and stay idle long periods of time. However, liquid monopropellant systems can provide smaller and more precise pulsed firings and have a lower number of components compared to bipropellant systems. These can be expensive and difficult to be applied in some cases, and require care in the design to operate with zero gravity (Brown, 1995; Humble et al., 1995). The use of monopropellant or bipropellant systems depends on the desired application. Comparatively, bipropellant systems have specific impulses between 300 s to 450 s, producing a thrust up to 1.2×107 N. Monopropellant systems operate with specific impulses of 130 s to 280 s providing 0.5 N to 500 N of thrust (Makled and Belal, 2009). Nevertheless, monopropellant systems are more commonly used than bipropellant systems for orbit correction and attitude control of satellites due to the low thrust levels required, lower mass and greater simplicity in construction. Hydrazine (N2H4) is the most used liquid monopropellant for space applications; however, in the last decade, there has been a significant interest for using green monopropellants, such as nitrous oxide (N2O) and hydrogen peroxide (H2O2). Nitrous oxide can be used as an energy source and for generation of oxygen on board spacecrafts or launch rockets, in cold gas propulsion systems for attitude control, in monopropellant or bipropellant thrusters, in gas generator systems, turbine operation and resistojets (Zakirov et al., 2000). A research team from the University of Surrey, in England, has tested several catalysts for decomposition of nitrous oxide. The tests were performed without choking the nozzle, at ambient pressure. The reported temperature for initiation of decomposition was 250°C and the final decomposition temperature reached 1500°C. It was concluded that the catalytic decomposition process depends on the mass flow rate and also that the start of the catalytic decomposition can be faster with higher preheat energy supply (Zakirov et al., 2001). At Tsinghua University, in China, hot tests were made with nozzle chocking. Based on the research from the University of Surrey, the catalytic bed dimensions were reduced and the transient preheat time decreased from 15 min to 52 s (Zakirov and Li, 2004). Also in Tsinghua, a computational model of the operation of a nitrous oxide monopropellant thruster was developed (Zakirov and Zhang, 2008). In Surrey and Tsinghua Universities, the focus of research was in the decomposition of nitrous oxide. In contrast, researchers at Stanford University evaluated the characteristic velocity efficiency and predicted the thrust of a nitrous oxide thruster,

although without providing information on the thruster design. A gas generator was developed to operate at high mass flow rates of nitrous oxide ranging from 0.9 g/s to 2.3 g/s, thus obtaining characteristic velocities with high efficiency in steady state. The mixing of CH4 and N2O was also studied for initiation of the reaction (Lohner et al., 2007; Scherson and Lohner, 2009). In the University of Beihang, in China, the effect of the catalytic decomposition of nitrous oxide on specific impulse was studied in a 140 mN monopropellant thruster. Based on preliminary studies with a numerical model considering effects of turbulence, the thruster was designed and tested in vacuum conditions. The catalyst used was alumina impregnated with iridium; yielding decomposition starting around 523 K. The experiments have shown that the bed length influences the nitrous oxide catalytic decomposition with longer beds, favoring complete decomposition. Therefore, it was found that to achieve high specific impulse, the optimization of the catalyst bed is required to assure that the products reach a high temperature at the nozzle inlet. On the other hand, effects such as heat losses to the environment cooperate negatively to reaction temperature. A model for the consumption of nitrous oxide and its self-pressurizing conditions in the tank was also developed. It follows from experimental and simulated data that the tank pressure decreases during the feeding process of self-pressurized nitrous oxide. Large pressure drops in the final stages of the feeding process can interfere with the required mass flow (Cai et al., 2011). Therefore, this work describes an experimental study of a 2 N monopropellant thruster using nitrous oxide as propellant. The propulsive parameters are determined experimentally and the decomposition performance is evaluated. A glow-plug heater system and a rhodium oxide catalyst supported on alumina were used to heat and decompose the gaseous propellant.

NITROUS OXIDE CHARACTERISTICS Nitrous oxide is a low cost non-toxic, non-flammable, nonexplosive, non-harmful , and easy to purchase or manufacture propellant (Merrill, 2008; Junior, 2009). It remains stable under normal conditions and is compatible with materials with simple structure; it is storable as a compressed gas or a liquid in wide temperature ranges, theoretically limited by its triple point (–90.8°C). However its storage is recommended in temperatures between –34°C to 60°C. N2O decomposes exothermically with aid of a catalyst or thermally, and can be stored in saturation

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

Experimental Investigation of a Monopropellant Thruster Using Nitrous Oxide

conditions in about 50.8 bar at 20°C. A comparison of nitrous oxide with other monopropellants is shown in Table 1. Nitrous oxide starts decomposition between 520°C to 850°C, forming mostly nitrogen (N2) and oxygen (O2), and releasing heat. Due to the high amount of energy needed to initiate the decomposition, catalysts are used to reduce the activation energy of the reaction, then yielding a lower decomposition temperature, around 250°C. However, the system must be preheated to this temperature to start the exothermic decomposition and attain a self-sustaining reaction. Reaching this condition, the adiabatic decomposition temperature is around 1640°C, producing a theoretical specific impulse of 206 s (Zakirov et al., 2001).

365

(4) or by a heterogeneous recombination: (5) Secondary reactions involved mainly in the formation of NO are: (6) (7) (8)

NITROUS OXIDE DECOMPOSITION

(9)

The chemical mechanism of the gas phase decomposition of nitrous oxide starts with:

The previous equations can be combined, yielding the following global reactions:

(1)

(10)

where M is a third body (Karabeyoglu et al., 2008). The O radical reacts with N2O according to the two parallel reactions:

(11)

(2) (3) The O radical can be eliminated by a termolecular homogeneous recombination:

The reaction involving N2/O2 is exothermic and the NO/N2 reaction is endothermic. Catalytic reactions offer advantages since they favor the N2/O2 self-sustaining reaction and improve the reaction efficiency (Wilson et al., 2012; Lohner et al., 2007; Zakirov and Zhang, 2008). The catalytic decomposition of nitrous oxide starts

Table 1. Comparison of nitrous oxide with others monopropellants (Zakirov et al., 2001). Propellant

Nitrous Oxide

Hydrogen Peroxide

Hydrazine

Chemical formula

N2O

H2O2

N2H4

Theoretical Isp (s)

206

179

245

Storable

yes

Density (kg/m )

745 at 20°C and 52 bar

1347

1004

Vapor pressure (bar)

50.8 at 20°C

0.00345 at 20°C

0.0214 at 26,7°C

Storage temperature (°C)

–34 to 60

–7 to 38

9 to 60

Toxic

burns skin

very toxic

Flammable

yes

Isp in vacuum conditions with nozzle expansion ratio 200; all propellants stored in liquid state; hydrogen peroxide 89 % m/m.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

366

Hennemann, L., Andrade, J.C. and Costa, F.S.

with the adsortion of a N2O molecule in an active site on the catalyst surface. After that, the N2 molecule is released while the oxygen atom remains adsorved on the active site. At last, the oxygen atom is released when it combines with another oxygen atom in a nearby active site, forming O2 or by direct reaction with another N2O molecule. These steps are described by: (12)

• Chemical kinetics, limited by the speed of N2O adsorption

on the catalyst, that is, the speed of combination of N2O with a rhodium molecule in an active site and release of oxygen and nitrogen; • Diffusion in the catalyst pores, where the reaction is limited by diffusion of reactants and products into and out of these microscopic pores; and • Convection in the diffusion boundary layer around each pellet catalyst, which cannot carry the products and reactants from the surface.

(13) (14) (15) In reactions (12-15) the O-M term refers to an oxygen radical and the N 2 O-M term refers to a N 2 O molecule adsorved on the catalyst surface. The []-M term corresponds to an empty site on the catalyst surface. During N 2 O decomposition, the adsortion of oxygen radicals coming from the catalytic surface is often considered a limiting factor. For this reason, the decomposition rate of N2O is intimately linked to the bind energy between the catalyst metal and the oxygen atom (Scherson and Lohner, 2009; Kapteijn et al., 1996).

CATALYST FOR N2O DECOMPOSITION A catalyst is used to provide a rapid propellant decomposition, without changing the global reaction. The reactants bind chemically to the surface of the catalyst active sites, which are atoms or groups of atoms available for reaction. When the catalyst is solid, the active material is composed of metal or metal oxides which are dispersed over a high surface area material, increasing the contact area between reactants and active sites (Batta et al., 1962; Vieira et al., 2003). In the case of gaseous nitrous oxide, this is obtained by making it pass through a material with transition metals, such as Ir, Rh, Ru, Ni, Co, Cu or Zr. There are several characteristics which affect the catalyst performance. The most important one is the decomposition limiting factor. The reaction can be controlled by different processes:

Knowing which of these modes is important will determine how the catalyst bed should be sized (Figueiredo and Ribeiro, 1989). Tests are needed also to determine the critical temperature range. It is necessary to know when the decomposition reaction changes from slow to fast and complete. The catalyst must be tested to determine the maximum temperature at which it can be used without deterioration of its properties. This should be checked since the active material (Rh, Ru, Ir) to be deposited on the support material surface also will be subjected to high temperatures. Placing these active elements at very high temperatures can cause their vaporization and removal of the support surface, destroying the catalyst (Figueiredo and Ribeiro, 1989). Being commercially available and having a relatively low cost, supports based on alumina (Al2O3) are widely used for catalytic purposes in propulsion. The aluminum oxide is formed at room temperature in γ-Al2O3 phase and has an open structure consisting of amorphous aggregates of Al12O18. When the temperature increases (around 1200°C) aluminum oxide goes through different phases irreversibly to form α-Al2O3 which is the final stage of its internal structure. Although the stages of transition are not well understood, it is well known that for each transition, the available surface area per gram of oxide decreases significantly, causing a decrease in the number of active sites available for decomposition (Arai and Machida, 1996; Vafai, 2005). Because of its ready availability and decomposition efficiency, a catalyst of rhodium (Rh2O3) oxide supported on alumina (Al2O3) was chosen to decompose nitrous oxide in a 2 N thruster prototype. Catalyst pellets (cylindrical extrudate) approximately 2 mm diameter and 3 mm length were used, containing 5% by volume of rhodium oxide. The pellets are shown in Fig. 1.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

Experimental Investigation of a Monopropellant Thruster Using Nitrous Oxide

367

Table 2. Thruster theoretical parameters.

Figure 1. Rh2O3/Al2O3 catalyst.

THRUSTER DESIGN The thruster design was based on a performance theoretical study carried out with aid of CEA2 NASA (2012) program. This program initially determines the chemical equilibrium composition and temperature in the chamber for constant pressure and enthalpy, through minimization of the Gibbs free energy. Then it solves the conservation equations for one-dimensional isentropic steady flow along the nozzle considering equilibrium or frozen conditions. As input parameters, the properties of pure nitrous oxide gas were considered at a temperature of 298.16 K, atmospheric pressure of 0.95 bar, 5 bar pressure in the chamber and assuming frozen flow in the nozzle. Data obtained with the NASA CEA2 code are shown in Tables 2 and 3. Table 3 confirms that the main decomposition products are O2 and N2. Table 4 shows some dimensions of the nozzle and the mass flow rate of propellant calculated to yield 2 N thrust. Estimates of the length and diameter of the catalytic chamber were made based on the literature of monopropellant catalytic thrusters (Lohner et al., 2007; Scherson and Lohner, 2009; Cai et al., 2011). Thus, the reference values adopted for initial design of the catalyst chamber were 15 mm in diameter and 70 mm in length (D15L70). All thruster parts (flange injection, chamber, catalyst bed and nozzle) were manufactured in 316 stainless steel. Computational views of thruster (3D drawings) are shown in Fig. 2, whereas Fig. 3a depicts the disassembled thruster and Fig. 3b, thruster mounted in the thrust balance.

Thruster parameters

Thruster chamber

Nozzle throat

Nozzle exhaustion

P (bar)

5.0000

2.7262

0.9505

T (K)

1906.92

1658.40

1293.97

ρ (kg/m3)

0.92529

0.58010

0.25924

cp (kJ/kg/K)

1.2403

1.2213

1.1842

1.2961

1.3021

1.3146

Mach

0.000

1.000

1.758

Ae/Ag

1.000

1.4343

c* (m/s)

1101.8

0.7099

1.1076

Ivac (m/s)

1383.0

1520.9

Isp (m/s)

782.2

1220.4

P = chamber pressure; T = chamber temperature; ρ = density of products; cp = specific heat of the products; γ = ratio of specific heats of products; Mach = Mach number; Ae = exhaustion area; Ag = throat area; c* = characteristic velocity; CF = coefficient of thrust; Ivac = specific impulse in vacuum; Isp = specific impulse with adapted nozzle.

Table 3. N2O decomposition products. Products

Molar fractions

0.06370

0.00008

NO2

0.00004

0.32988

0.66327

Table 4. Thruster parameters and dimensions. Parameters

Values

Ag (mm2)

3.61

Dg (mm)

2.14

Ae (mm2)

5.17

De (mm)

2.56

m. (/s)

1.63

TEST BENCH A test bench was used to determine thrust, mass flow rate, temperatures and pressures in the supply line and in the thruster. A schematic of the test bench is showed in Fig. 4. It contains

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

368

Hennemann, L., Andrade, J.C. and Costa, F.S.

Figure 2. 3D drawings of the thruster with instrumentation and cut views.

(a)

(b)

Figure 3. Thruster photos. Control valve Flow valve Thermocouples T4 T2 Tinj Bypass valve

Pressure transducer Ppc

Shut-off valve

Flow meter Check valve

Filter

Line pressure gauge Pressure regulator with heather

Thruster with thermal insultation Tpc T3 T1 Pressure transducer Pinj

Load cell

Power supply

Figure 4. Test bench scheme. J. Aerosp. Technol. Manag., SĂŁo JosĂŠ dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

Nitrous oxide bottle

Experimental Investigation of a Monopropellant Thruster Using Nitrous Oxide

a gas cylinder with nitrous oxide (99.99%), a heated pressure regulator, sintered filters in the gas inlet valves, manual lock valves, two-way solenoid valve with direct action, check valve, pressure transmitters, K-type thermocouples, HBM load cell with 5 N capacity (± 0.01 N), and a Sevenstar nitrous oxidizer mass flow meter. Preheating of the nitrous oxide to start decomposition was performed by a Duratherm Bosch glowplug placed at the entrance of the catalyst bed, as depicted in Fig. 2. The injection temperature, i.e., the gaseous nitrous oxide temperature at the entrance of the catalyst bed, was adjusted by a temperature controller (Novus 480I) which monitored the glow-plug heating. A data control system using a National Instruments SC-2345 rack, TCC02 modules (for signals in mV), FT01 modules (with channels 0-10 V), Compact Daq USB module (extra channels of analog inputs and communication module) and a PCI-6221 data acquisition board were used for observations of the experiments variables in real time. A Labview control software monitored variables of pressure, mass flow rate, thrust, temperature and operation of the glow-plug. For this purpose, analog and digital input/output were used, providing data on a computer in graphical form and saving in its database.

RESULTS AND DISCUSSION A pressure of 7.5 bar at the exit of the regulator valve was adopted to provide the total pressure drop on the various devices installed before the thruster. Once the temperature (thermocouple Tinj) measured in the glow-plug heater section reached 220°C the injection of nitrous oxide was performed. This is evidenced in Fig. 5 which shows that a pulse with a mass flow rate (Q) of 1.6 g/s increased rapidly the catalytic bed temperature, thus indicating the occurrence of catalytic decomposition. Locations of Tinj, T1, T2, T3 and T4 thermocouples along the catalytic chamber can be seen in Figs. 2 and 4. Tpc is the measured temperature after the catalytic bed (post-chamber). Prior to testing, pre-heating was kept for several minutes for thermal stabilization of the catalyst bed. After the preheating period, the glow plug temperature controller was turned off. The mass flow rate was gradually increased to the nominal flow rate and heating was maintained only by the decomposition of N2O flow. Under these conditions, it was possible to apply multiple sequences of pulses with different duty cycles, i.e., periods with mass flow on and off. Very long shots could compromise the thruster and thermocouples, due to

700

2.9

Tinj

500

2.4

1.9 400 1.4 300 0.9

200

0.9

100 0

Mass Flow Rate (g/s)

Temperature (ºC)

Tpc 600

369

100

150

200

-0.1 250

Time (s)

Figure 5. Initial heating using N2O. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

370

Hennemann, L., Andrade, J.C. and Costa, F.S.

the high temperatures that could be reached in the bed. During intervals, there occurred a temperature uniformization along the thruster, especially in the catalytic bed. Figure 6 shows curves of pressures, temperatures and mass flow rates measured in the thrust balance. A sequence of 5 pulses with 25 s on and 10 s off is shown, followed by sequences of 3

Thrust (N), Pressure (bar)

7.0

Tpc

Tinj

6.0 5.0 4.0 3.0 2.0 1.0 0.0 770

870

970

Time (s)

1070

1170

1070

1170

Pressure (bar), Mass Flow Rate (g/s)

(a)

7.0

Ppc

Pinj

6.0 5.0 4.0 3.0 2.0 1.0 0.0 770

870

970

Time (s)

Pressure (bar), Mass Flow Rate (g/s)

(b)

1400

Tinj

Tpc

1200 1000 800 600 400 200 0 770

870

970

Time (s)

1070

1170

(c)

Figure 6. D15L70 thruster duty cycle testing.

1270

pulses with 15 s on and 10 s off. It is observed that the injection pressures (Pinj) and post-chamber (Ppc) reached 5.29 bar and 5.15 bar, respectively. The temperature at the end of the catalytic bed (T4) reached 1400°C with an average mass flow rate (Q) of 1.65 g/s. The thrust (F) measured on the load cell reached 1.93 N, close to the design value of 2 N. Figure 7 shows the gas temperature evolution inside the catalytic chamber of the thruster. A MATLAB algorithm using a cubic interpolation method was developed to determine the temperature distributions during the thermal-catalytic decomposition process. Taking as reference the pulse in the interval of 860 s to 900 s in Fig. 6c, it was observed that the highest temperature (1400°C) was measured by the thermocouple located at 62 mm from the injection, as shown in Figs. 2 and 4. This corresponds to the last catalytic bed thermocouple (T4). The maximum temperature reached at the end of the catalytic chamber indicates a bed with optimum length. The hot expanding gases from decomposition pass through the post-chamber and follow to the nozzle for generating thrust. Pulses present similar profiles for several sequences and in different periods. However, the fatigue of the catalyst can be observed in the range of 1136 s to 1200 s in Fig. 6, where pulses are deliberately shown with a decrease in catalyst efficiency. This is due to the pellet contraction and loss of rhodium oxide. The contraction of the pellets is caused by the low calcination temperatures and phase change of alumina, as was previously described by Arai and Machida (1996). The rhodium oxide loss is due to the high temperatures reached in the catalytic bed. The decrease in pellet volume also creates preferential paths for N2O flow which pass through the catalytic bed without decomposition. Besides, it causes a decrease in the catalyst specific area, reducing the area of active sites available for catalytic reaction. If the surface available for decomposition is internal to the pellets, with the alumina retraction the active sites become inaccessible for N 2O, preventing its decomposition. However, there was no clustering of catalyst grains or weight loss. There is also a loss of efficiency concerning the reduction of the adsorption capacity of the molecules of N2O, as seen in the last three pulses in Figs. 6, 8 and 9. This may be explained by the adsortion of water and oxygen which reduces the number of active sites available for N2O decomposition (Esclapez, 2011). This factor, jointly with the retraction of alumina and heat losses, can decrease the reaction temperature in

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

Experimental Investigation of a Monopropellant Thruster Using Nitrous Oxide

895

1100 1000 900

880

800 700

875

600 870

Specific Impulse (m/s)

1200

885

Time (s)

1200

1300

890

1000 800 600 400 200

500 400

865

0 770

300 10

Position (mm)

371

870

970

Time (s)

1070

1170

1070

1170

(a)

1200

CONCLUSIONS

800 600 400 200 0 770

870

970

Time (s)

(b)

Figure 8. Specific impulse (a) and characteristic velocitie(b).

100 90 80 70

ηc*(%)

the bed. It should be noted that using shorter pulse lengths would provide lower temperatures and a slower degradation of the catalyst. Figure 8 shows the specific impulse, , and the characteristic velocity, , obtained experimentally. The pulses shown in Fig. 6 were taken as reference, since they were obtained with the thruster in full operation at the start of degradation of the catalyst. The average specific impulse obtained experimentally was about 1120 m/s, which is close to the theoretical value, 1220.4 m/s. Therefore, the average specific impulse efficiency, , was around 91.7 %. The characteristic velocity efficiency, , is shown in Fig. 9. The characteristic velocity value obtained in the experiments was about 970 m/s, whereas the average theoretical value is about 1102 m/s, giving an average efficiency of about 88%. Considering the characteristic velocity efficiency obtained experimentally, it is observed that the catalytic decomposition of nitrous oxide in the bed was not complete.

Characteristic Velocity (ms)

Figure 7. Gas temperature (°C) evolution along the catalytic bed and post-chamber section.

60 50 40 30 20

An experimental investigation of a monopropellant 2 N thermal-catalytic prototype thruster using gaseous nitrous oxide was presented. The propellant was decomposed by a catalyst of rhodium oxide supported on alumina, with preheating of both catalyst and propellant by a glow-plug system.

10 0 770

870

970

1070

1170

Time (s) Figure 9. Characteristic velocity efficiency.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

372

Hennemann, L., Andrade, J.C. and Costa, F.S.

Relatively high decomposition efficiency (~ 88 %) was observed prior to catalyst degradation. Changes in the catalytic support preparation are required to reduce the volume loss of the support under the test conditions and to improve catalyst performance for longer firing times. Other factors such as optimization of catalytic bed dimensions, improvements in the injector design and better thermal insulation may increase the thruster’s efficiency.

ACKNOWLEDGEMENTS The authors acknowledge the Brazilian Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for providing a doctoral scholarship to the first author and a research grant to the second one. The catalyst preparation was performed with the kind help of Dr. Ricardo Vieira from Brazilian Instituto Nacional de Pesquisas Espaciais (INPE).

REFERENCES Amaral, L., 1995, “Química”, Loyola, São Paulo, 23-24. Arai, H. and Machida, M., 1996, “Thermal stabilization of catalyst supports and their application to High Temperature catalytic combustion”. Applied Catalysis A: General, Vol. 138, No. 2, pp. 161-176. doi: 10.1016/0926-860X(95)00294-4.

NASA, “Chemical Equilibrium with Applications (CEA)”. Cleveland, OH: Glenn Research. Center NASA, Retrieved on March 5, 2012, from http://www.grc.nasa.gov/WWW/CEAWeb/ceaguiDownloadwin. htm. Neto, T.G.S., 2011, “Handout catalysis”. Instituto Nacional de Pesquisas Espaciais, Curso de Engenharia e Tecnologias Espaciais.

Batta, I., Solymost, F. and Szabo, Z., 1962, “Decomposition of nitrous oxide on some doped cupric oxide catalysts”. Journal of Catalysis, Vol. 1, No. 2, pp. 103-112. doi: 10.1016/0021-9517(62)90014-3.

Plumlee, D. and Steciak, J., 2004, “Development of a monopropellant micro-nozzle and Ion Mobility spectrometer in LTCC”, In: Ceramic Interconnect Technology Workshop, Denver, CO.

Brown, C.D., 1995, “Spacecraft Propulsion”. Ohio: Series Editor-in-Chief. AAIA Education Series. Air Force Institute Technology, pp. 224.

Scherson, Y. and Lohner, K., 2009, “A Monopropellant Gas Generator Based on N2O Decomposition for “Green” Propulsion and Power Applications”. American Institute of Aeronautics and Astronautics, 45th AIAA/ASME/ SAE/ASEE Joint Propulsion Conference & Exhibit, Denver, Colorado, EUA.

Cai, G., Sun, W., Fang, J., Li, M., Cong,Y. and Yang, Z., 2011, “Design and performance characterization of a sub-Newton N2O monopropellant Thruster”. Aerospace Science and Technology, Vol. 23, No. 1, pp. 439451. doi: 10.1016/j.ast.2011.10.003. Esclapez, S.P., 2011, “N2O Decomposition rhodium/ceria catalysts: From principles to pratical application”, PhD Thesis, Inorganic Chemistry Department, Alicante University, Spain, pp. 201. Figueiredo, J.L. and Ribeiro, F.R., 1989, “Catálise Heterogênea”. Calouste Gulbenkian Foundation, Porto, Portugal, pp. 353. Junior, J.A., 2009, “Desenvolvimento de um Propulsor Catalítico para Satélites usando Óxido Nitroso como Propelente”, Master Thesis, Instituto Nacional de Pesquisas Espaciais - INPE, Cachoeira Paulista, Brazil, pp. 168. Humble, R.W., Henry, G.N. and Larson, W.J., 1995, “Space Propulsion Analysis and Designer”. MacGraw-Hill Companies, Inc., Primis Custom Publishing. Kapteijn, F., Mirasol, J.R. and Moulijn, J.A., 1996, “Review: Heterogeneous catalytic decomposition of nitrous oxide”. Applied Catalysis B: Environmental 9, pp. 25-64. doi: 10.1016/0926-3373(96)90072-7. Karabeyoglu, A., Dyer, J., Stevens, J. and Cantwell, B., 2008, “Modeling of N2O Decomposition Events”. American Institute of Aeronautics and Astronautics, 44th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Hartford. Lohner, K., Dyer, J., Doran, E., Dunn, Z., Krieger, B., Decker, V., Wooley, E., Sadhwani, A., Cantwell, B. and Kenny, T., 2007, “Design and Development of a Sub-Scale Nitrous Oxide Monopropellant Gas Generator”, American Institute of Aeronautics and Astronautics, Stanford University, Stanford, USA. Makled, A.E. and Belal, H., 2009, “Modeling of Hydrazine Decomposition for Monopropellant Thrusters”. 13th International Conference on Aerospace Sciences & Aviation Technology, ASAT - 13, Military Technical College, Kobry Elkobbah, Cairo, Egypt. Merril, C., 2008, “Nitrous Oxide Explosive Hazards”, Air Force Research Laboratory, Edwards AFB, CA.

Sanscrainte, W., 1961, “Hydrogen Peroxide Attitude Control Systems”. Planetary and Space Science, Vol. 4, pp. 184-193. doi: 10.1016/00320633(61)90131-3. Stark, J.P.W.J. and Swinerd, G.G., 2003, “Mission Analysis”, In: Fortescue, P., Stark, J. and Swinerd, G., (Eds.), Spacecraft Systems Engineering, 3rd ed., John Wiley & Sons, pp 111-167. Sutton, G.P., 2001, “Rocket propulsion elements: an introduction to the engineering of rockets”. 7. ed. John Wiley & Sons, pp. 751. Vafai, K., 2005, “Handbook of Porous Media”. 2nd Edition. Taylor & Francis Group, LLC, London, UK, pp. 742. Vieira, R., Cuong, P., Keller, N. and Ledoux, M.J., “Novos Materiais à base de Nanofibras de Carbono como Suporte de Catalisador na Decomposição da Hidrazina”. Química Nova, Vol. 26, No. 5, pp. 665669. doi: 10.1590/S0100-40422003000500008. Wilson, M.D., Eilers, S.D. and Whitmore, S.A., 2012, “Catalytic Decomposition of Nitrous Oxide Monopropellant for Hybrid Motor Re-Ignition”. American Institute of Aeronautics and Astronautics, 48th AIAA/ASME/SAE/ ASEE Joint Propulsion Conference & Exhibit, Atlanta, Georgia, USA. Zakirov, A.V. and Zhang, H., 2008, “A model for the operation of nitrous oxide monopropellant”. Aerospace Science and Technology, Vol. 12, No. 4, pp.318-323. doi: 10.1016/j.ast.2007.08.003. Zakirov, V., Sweeting, M., Goeman, V. and Lawrence, T., 2000, “Surrey research on nitrous oxide catalytic decomposition for space applications”. 14th AIAAUSU Conference on Small Satellites, pp. 1–9. Zakirov, V. and Li, L., 2004, “Small Satellite Propulsion Challenges”. European Conference for Aerospace Sciences (EUCASS), Tsinghua University, P.R. China. Zakirov, V., Sweeting, M., Lawrence, T. and Sellers, J., 2001, “Nitrous oxide as rocket propellant”. Acta Astronautica, Vol. 48, No. 5-12, pp. 353-362. doi: 10.1016/S0094-5765(01)00047-9.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.363-372, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.380

Risk Assessment for Hexamine Nitration into RDX Erick Braga Ferrão Galante1, Daniele Mesquita Bordalo da Costa2, Assed Naked Haddad2, Isaac José Antônio Luquetti dos Santos3

ABSTRACT: Explosives are widely used in extraction of ores, melts and fuels and in the safe and economical demolition of structures. These applications demonstrate the value of explosives to society and the reason why they are manufactured worldwide despite the hazard of their preparation. Hence, this paper focuses on the manufactory of the military explosive Hexogen (also known as RDX). The characteristics of the process are described by the analysis of a specify manufactory plant. The chosen analysis tool is the Hazard and Operability Study – HazOp. Historically, this methodology has been applied at chemical industries and is used in industrial process operability analysis. This study analyses seven critical nodes in the RDX manufacturing process, identifies several deviations, and causes and consequences. From these results, improvements in the units are suggested and actual conditions discussed. It is important to remark that this work is an initial approach to analysis of the manufacturing process of RDX using the HazOp methodology. KEYWORDS: Risk Assessment, Explosives, Nitration, Hexogen, RDX, HazOp.

INTRODUCTION Explosives are energetic materials that have been used since the basic foundations of modern civilization in minerals, ores, metals and fuel exploitation. Due to their importance, explosives are manufactured worldwide. Every operation involving explosive usage offers risks. Hexogen (also known as RDX) is a high explosive, largely used in the oil industry (Galante et al., 2013). Hence, the risk of explosion is always present in production processes (Khan and Abbasi, 1999), which may lead to large accidents, such as the one in Bhopal (Bisarya and Puri, 2005; Chouhan, 2005; Eckerman, 2005; Gehlawat, 2005). An “explosive” is defined as solid or liquid substances that can combust quickly in an exothermic reaction (Meyer et al., 2007). Instability is a major characteristic of explosives, which undergo reaction triggered by flame, friction or heat. A quick release of gases at high-pressure gases, with a great energy emission, characterizes the explosion. Due to these characteristics, the manufacture, testing, sale, storage, and transportation of explosives require special consideration, which can be addressed by performing risk analyses, shuck as HazOp or Hazard Preliminary analyses. According to Taylor (2007), design issues can be addressed during any stage of a chemical unit, even one already operating. From that perspective, this work assesses the existing risks at a nitration process in the manufacture of RDX, emphasizing the perspective presented by Steen and Aven (2011), who discusse engineering risks. From the works of Aven (2012) and Held (2012), one realizes that a risk analysis of a RDX manufacturing unit is feasible through HazOp methodology. Furthermore, Held (2012) reported and discussed detonation of a small amount

1.Instituto Militar de Engenharia – Rio de Janeiro/RJ – Brazil 2.Universidade Federal do Rio de Janeiro – Rio de Janeiro/RJ – Brazil 3.Comissão Nacional de Energia Nuclear – Rio de Janeiro/RJ – Brazil. Author for correspondence: Erick Braga Ferrão Galante | Instituto Militar de Engenharia – Departamento de Engenharia Química | SE/5 - Praça Marechal Tibúrcio 80, Rio de Janeiro/RJ | CEP: 22290-270 – Brazil | Email: egalante@ime.eb.br Received: 06/09/2014 | Accepted: 09/26/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

374

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

of explosive in the piping just after batch nitration of HMX. In this accident, one operator was injured with shrapnel and then hospitalized. Damage to the plant was minor. Held (2012) reported that the investigation of this accident showed that the sensitivity of intermediate products was underestimated and that a HazOp (Boonthum et al., 2014; Dunjó et al., 2010; Rausand and Hoyland, 2011; Rausand and Utne, 2009) should be used to assess risk prior to the start-up of the unit. Hence, the main goal of this paper is to analyze and diagnose risks in a chemical unit used to manufacture RDX. Graf and Schmidt-Traub (2000), among others, proposed the use of HazOp for addressing safe operation of chemical plants, which in this particular case is a unit to nitrate Hexamine into RDX. Therefore, the HazOp was chosen as the most suitable method for this paper.

CASE OF STUDY Explosives are substances of great importance in human development. Besides their traditional use by the military, RDX is used in civilian applications, such as tunnels, construction, and exploitation of natural resources such as mining and oil exploration. Explosives are chemical substances or mixtures of substances, which react rapidly by heating or attrition and generate large volumes of gas and heat. Explosives are classified by several standards (for example STANAG (NATO), MILSTD (USA) and TM-9-1903 (USA)). Classifying explosives according to their detonation velocity is far common. RDX is classified as a high explosive (also known as “secondary explosive”) due to its output of energy during detonation, when compared to a low explosive (with a detonation velocity in the same magnitude of the speed of sound) (Akhavan, 2011; Cooper, 1996; Galante et al., 2014; Meyer et al. , 2007; Urbanski, 1984). Hexogen, RDX, T4 or cyclone are common names for 1,3,5-trinitroperhydro-1,3,5-triazine. Hexogen is an explosive used as main charge in military warheads, as well as in mining explosives (Galante et al., 2013; Galante and Haddad, 2009) or additive to propellants. It is soluble in acetone, insoluble in water, and partly soluble in ether and ethanol. According to Meyer et al. (2007), RDX may be the most important explosive in terms of brisance; besides that, its explosive power is high, as well as its detonation velocity (Mei et al., 2012).

Due to the inherent risk contact, it is necessary to observe RDX storage and transport conditions. This explosive is very sensitive to electrical sparks, shock, and heat and impact, among some other stimuli. Achuthan and Mullick (1983) studied and discussed risks including fire, explosion, and toxic hazards during manufacture of RDX. PRODUCTION PROCESS Bachmann and Sheehan (1949); Leach and Staples (1981); Lukasavage and Slagg (1993) and Meredith (1976) were the first to describe the synthesis of the high explosive hexogen (RDX). The Chemical production of RDX occurs mainly in nitration vessels. Hence, this is the core of any manufacturing facility. According to the analysis of manufacturing procedures (Akhavan, 2011; Cooper, 1996; IMBEL, 2006; Lukasavage and Slagg, 1993; Meredith, 1976; Urbanski, 1984), the typical capability of production in a semi-batch type unit (Fogler, 1999) is 50kg dry hexamine per hour/batch (IMBEL, 2006). The process transforms hexamine into RDX via nitration (using HNO3) and paraformaldehyde (the paraformaldehyde is added to the vessel manually and before the production starts). This case of study produces Hexogen (C3H6O6N6) by nitrating hexamethylaminetriamine (C6H12N4) using strong nitric acid. The later precipitation of the explosive occurs by adding cold water. The industrial manufacture of RDX used for this study parallels processes patented by Luksavage and Slagg (1993) and Meredith (1976). The flowchart under study (Fig. 1) basically consists of the mixture in nitration vessels, boilers, followed by filtration of the final product. Although this is not the most recent method for manufacturing RDX, it is still largely used worldwide, reason why it was chosen for this study. In this process, the reaction temperature is kept between 12°C and 15°C during the addition of nitric acid, the reaction is allowed to go up to 18°C, temperature in which it is quenched. After finishing the nitration, the mixture is heated to 60°C for 60 minutes. After this period, the mixture is cooled and the RDX precipitate is filtered, washed and dried. Each stage (nitration, heating and cooling) is held in a different vessel. The equipment specified in Table 1, which also shows the available quantities, comprises the manufacturing facility. All the vessels listed in Table 1 have a nominated volume of 300 liters and are made of stainless steel. Each and every vessel is equipped with a safety release valve, placed in the bottom end of the vessel. These safety valves operate under air pressure, which means that they will open if the pressure in an air pipeline

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

HNO3@99%, forward flow

Cooling fluid, forward flow

Hexamine feeder

Cold water, backward flow

Hexamine feeder

Coolant water, return flow Cold water, forward flow Heat exchanger 1

Vessel for HNO3@99% HNO3@99%, backward flow

Nitration vessel 1

Exhauster

Nitration Nitration vessel 2 vessel 3 Boiler 1

Cooling fluid, backward flow

375

Boiler 2

Heat exchanger Feed water

Hexogen Vacuum filter 1 Hexogen Vacuum filter 2

Emergency vessel for disposure

Vacuum pump Support tank Separation for the vessel Vessel vacuum pump Waste water for HNO 3 vessel @ 10%

Vessel for HNO3 @ 60%

Pump for HNO3 @ 60%

Residual acid (HNO3@60%) Vessel for HNO3 @ 60% 2 Feed water

Emergency vessel for neutralization 1

Emergency vessel for neutralization 2

Vessel for feed water

Figure 1. Production flowchart.

Table 1. Equipment list. Quantity

Type of equipment

Hexamine feeder

Pump for HNO3 @ 60%

Vacuum pump

Vessel for feed water

Hexamine rated feeder

Exhauster

Boiler

Vacuum filter

Nitration vessel

Emergency neutralization vessel

Waste water vessel

Support tank for the vacuum pump

Vessel for HNO3 @ 10%

Vessel for HNO3 @ 60%

Vessel for HNO3 @ 99%

Separation vessel

Heat exchanger

drops, may it be by design (operator acting upon a switch) or by a power loss in the unit. These vales will automatically open if the reaction temperature within the nitration vessel moves above 18°C. All safety valves are connected to an emergency neutralization vessel, diluting the nitric acid will be diluted and the reaction quenched. MATERIALS AND METHODS: HAZOP Seccatore et al. (2013) stated that the rock blasting, carried out on every day activities in thousands of locations around the world, is a primary activity in mining and civil excavation. They stated that in blasting and mining activities, risk management in explosive operations aims personal and health safety. However, there are other risks involved in the employing of explosives. They have suggested the use of HazOp as the main risk analysis technique for deviations. HazOp is a method for risk assessment that favors the decision making for corrective actions, where applicable. In general, it involves several professionals from different specialties, in order to evaluate different aspects of the studied object and favors process industries due to the easiness of applying control points (Khan and Abbasi, 1999). One of the first publications related to the methodology became available in 1974 and was called “Operability studies J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

376

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

and hazard analysis” (Dunjó et al., 2010; Kletz, 2009). This paper provided a guide for applying the methodology created by the Chemical Industries Association in the UK in 1977. Dunjó et al. (2010) analyzed the increase in publications in subsequent years, especially between 1996 and 2000, in which more than 40 papers on the theme were published. Initially, HazOp (Dunjó et al., 2010; Rossing et al., 2010) was developed to analyze chemical process systems (Pérez-Marin and Rodríguez-Toral, 2013; Schüller et al., 1997), but over time, its application was extended to other types of systems and complex operations of various kinds, including computer programs (Rausand and Hoyland, 2011). Galante and Haddad (2009) proposed a variation for use in the explosives industry. Furthermore, there are several approaches of this methodology, which could be used isolated or in parallel to other techniques (Boonthum et al., 2014; Van den Bosch and Weterings, 2005; CCPS, 1989; Defence, 2012; Haddad et al., 2012; Kennedy and Kirwan, 1998; Sahar et al., 2010; Schuller et al., 1997; Shimada et al., 2012). A HazOp report may review all possible deviations, as well as their causes and consequences and proposes mitigation and active or passive protection (Labovský et al., 2007). According to Schüller et al. (1997), the soundness of the methodology is a function of the extent of the known interactions evaluated and the detail level of the analysis, and the depth of study of the identified consequences. A limitation of HazOp is that this approach is inherently qualitative (a “diagnostic tool”) (Crawley et al., 2000). Moreover, there is the difficulty to estimate the time required for a complete HazOp study (Dunjó et al., 2010; Freeman and Mcnamara, 1992; Khan and Abbasi, 1997) as well as a lack of risk acceptability criteria and international standards, as discussed by Rouhiainen & Gunnerhed (2002) and Labovský et al. (2007). The application of HazOp is based on the formulation of questions in a structured and systematic approach, through the appropriate use of guide words applied to critical points within the process being studied. From the guidewords and process parameters, deviations can be identified and further analyzed. Table 2 shows a series of standard guide words for basic application of HazOp, as presented by Rausand and Hoyland (2011). As previously stated, each guideword is applied to a process parameter to determine a deviation. Every deviation is studied. Table 3 provides possible relationships between guidewords and the consequent deviations.

Table 2. List of guide words. Meaning

Example

None

None of the objectives is achieved

No flow

More, Bigger

Quantitative increase in a parameter

More temperature

Less, Minor

Quantitative decrease in a parameter

Less pressure

Part of

Just part of the objectives is achieved

Part of the yield

Reverse

Occurs the opposite of what one expects

Reverse flow

Other

Full replacement

Liquids in a gas pipe

Table 3. Deviations and parameters. Parameter

Guide word

Deviation

Flow

None, Less, More, Reverse, Other, Also

None, Less, More, Reverse, Other, Contamination

Pressure

More, Less

More pressure, Less pressure

Temperature

More, Less

Higher temperature, Lower temperature

Viscosity

More, Less

More viscosity, Less viscosity

Reaction

None, Less, More

No reaction, Reaction incomplete, Intense reaction

RESULTS This work is a reference for conducting further analysis in the unit by addressing a greater number of nodes. The nodes evaluated here were chosen because they are considered critical in the production sequence, following the criterion of a node before each “key” component. The component keys identified were two reaction tanks, the transfer of process fluids, product purification, cooling and output of the final product. Therefore, for this study of HazOp, the process flowchart was divided into seven nodes, where each node was study according to the parameters presented in Table 4. The selected nodes are listed as follows:

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

• Array of vessels for waste and clean water for dilute HNO3

(Table 5);

• Connection between nitration vessels (Table 6); • Connection between the nitration vessel and the boiler • • • •

(Table 7); Connection between boilers (Table 8); Connection between boiler and heat exchanger (Table 9); Connection between the heat exchanger and the vacuum filter (Table 10); and Connection between the vacuum filter and the RDX recovery pan (Table 11).

These seven nodes were chosen in order to allow the study to focus on the most hazardous equipment within processes. It is reported in the literature (Akhavan, 2011; Cooper, 1996; IMBEL, 2013; Lukasavage and Slagg, 1993; Meyer et al., 2007; Urbanski, 1984) that the most critical operations are the nitration, followed by those dealing with heating and gaseous extraction. The study also incorporated some preparation vessels and filtering unities, due to the acid present in those equipment. Hence, there are seven nodes selected for evaluation. The chosen parameters were: flow, pressure, temperature, and reaction. The viscosity parameter was not considered significant to the process, and therefore not used. Table 4 summarizes the parameters used for each node. The guidewords used were: none, more and less. It is important to highlight that “REVERSE” and “OTHER THAN” are not considered, since the production under study is not a flow process; instead it is organized in batches. NODE #1 The node #1 includes all tanks containing water for washing and HNO3, i.e., waste water vessel, vessel for HNO3@10% and vessel for HNO3@60%. These were analyzed for flow, pressure and temperature. Since the process at this node is the dilution of HNO3 and no chemical reaction occurs, the parameter “reaction” was not considered. Table 5 is the result of the HazOp for this node. Regarding the results for the parameter “flow”, there is the need to install a valve that allows the flow to increase (in cases where there is no flow or low flow) or to stop (when flow is increased). From the analysis of the parameter “pressure”, the final recommendation is to install a regulatory valve at the flow into the tanks, allowing consequently the increase or decrease in pressure. Accompanying this parameter is the suggestion to install a pressure gauge to monitor it. Furthermore, one can

377

Table 4. Parameter used in each node. Node

Parameter

Flow, Pressure, Temperature

Flow, Temperature, Reaction rate

Flow, Pressure, Temperature

Flow, Temperature

determine deviations in temperature via thermocouples installed in the tanks (monitored via a control panel). As a solution, it is possible to increase or decrease the flow, depending on the temperature. One solution to the problem is to correlate this assembly to the heat exchanger or the inflow of brine. NODE #2 The second node (node #2) connects nitration vessels 2 and 3. For this node, the parameters flow, temperature and reaction were analyzed. Table 6 is the result of the HazOp for this node. The pressure switch was not evaluated since its variation results from changing other parameters, making its analysis therefore redundant. Possible variations in the mass flow were analyzed using keywords “none” and “less”. All of them are related to errors in hexamine dosing and possible flaws in this operation. In order to minimize the occurrence of this event, implementing automatic control in this equipment is suggested. Regarding the parameter “flow”, there is also the possibility of potential physical damage to the pipes, occurring in the guidewords “none” and “less”. In these cases, the implementation of a program of preventative maintenance or replacement of the pipelines, when necessary, is recommended. In the case of increased flow (“more flow”), venting the equipment in order to prevent overpressure and the consequent risk of explosion is recommended. For this, it is necessary to empty the tank. Draining is possible through bottom discharge valves. Variation in flow is identifiable by checking discharge into the emergency vessel, which may interrupt the chemical reaction. This tank is connected to the output of all the reactors (nitration vessel and boilers). When the temperature increases above the set points (which are established from the safety limits of the reactions), the control system interrupts the supply of

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

378

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

Table 5. Node #1 – Array of vessels for waste and clean water to dilute HNO3. Guide word

Deviation

Cause

Vacuum filter malfunctioning Defect in the tank for emergency transfer

Flow

Consequences

Indication in the flow meter

Reduced yield factor for nitration Failure separation tank

Visual inspection

Flow Increase for the baffle for wastewater

Under functioning pump to HNO3@60%

Leakage or malfunction of the pump to HNO3

Flow decrease to the tank to vacuum pump Flow decreased to absorption of HNO3 Breaking the vacuum pump

Less

Flow Decreased

Lowering the temperature

Flow Increase

Increasing the temperature

Increased pressure within the vessel Flow increase for the baffle for wastewater

Reduced yield factor for nitration Flow decreased to baffle for wastewater

Flow increase for the baffle for wastewater Breaking the vacuum pump

Less

Precipitation of the product

(< 12°C)

Vacuum pump failure

Temperature change in pump HNO3@60%

Checking through instruments (thermometer) Inspection through the inspection location

Flow increase for the tank to HNO3, 60% through valves

Stopping the flow

Flow increase for the tank to vacuum pump

Pressure

Failed vacuum filter

Flow reversal of the tank to the vacuum pump and boiler 2 through valves

Increased temperature and fire hazard

Over functioning pump to HNO3@60%

Temperature

Mitigation

Breaking the vacuum pump

None

Less

Detection

Decrease Flow to tank for HNO3@60% Installation of measuring instrument Pressure Increasing the flow to the tank HNO3@60% Installation of measuring instrument Pressure Decrease flow to tank for HNO3@60% Perform preventive maintenance control systems and operation of feed valves of the heating fluid and cooling

Over-Pressure vessels Increased speed reaction Risk of explosion due to a sudden pressure relief in a closed vessel Leak of NOx

(> 15°C)

Interruption of production Loss of quality Yield loss

Increasing the flow to the tank HNO3 @60% Perform preventive maintenance on control systems and operation of temperature control valves

Risk of fire from overheating

compressed air to the bottom valve, opening it (the valve fails in the open position). In case of safety valve opening, the vessel contents flow into the dump tank emergency. The tank normally

is partly filled with water at room temperature. This water dilutes the acid and equalizes the temperature, thus stopping the reaction and preventing an explosion.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

379

Table 6. Node #2 – Connection between nitration vessels. Guide word

Deviation

Cause

Hexamine feeder failure

Consequences

Decreased production of the nitration vessel 3

Less

Flow

Detection

Failure dosage HNO3@99%

Overheating in nitration vessel 3 Increased production of the nitration vessel 3 Overheating in nitration vessel 3 Flow in emergency dump tank Reduction of refrigerant flow return with brine

Less (< 12°C) Hexamine feeder failure Failure dosage HNO3@99%

Temperature More (> 15°C)

Failure in the heat exchanger

Checking through existing instruments (thermometer)

Interruption of production downstream Overload nitration vessel 3 Pressure and consequent increased risk of explosion Interruption of operation of the discharge valve background Increased return chilled water brine Decreased feed flow of refrigerant brine

Mitigation

Manual duct repair Replacing hexamine feeder Replacing hexamine feeder Emptying the nitration vessel through bottom discharge valves

Flow Decrease feed coolant brine

Damage to the product Interruption of Reaction Over-Pressure vessels Increased reaction speed

Increasing the flow of refrigerant return with brine Checking through instruments (thermometer)

Risk of explosion due to a sudden pressure relief in a closed vessel Leak of NOx in the workshop

Increasing flow of refrigerant feed brine

Interruption of production Loss of quality Yield loss

None Less Reaction More

Failure feeder hexamine Failure dosage HNO3@99%

Checking through instruments (thermometer) Increased Pressure in the first nitration vessel Checking through existing instruments (thermometer)

Regarding the “temperature” parameter, its failures relate to issues within the feeder of hexamine, failure dosing HNO 3 @99% (failure in the tank for HNO 3 @99%), and failure in the heat exchanger. The precautionary measure is to decrease or increase the flow of coolant brine according to the temperature increase or decrease, respectively. A final consideration is that, due to the nature of the reaction, a thermal explosion is very unlikely to occur, since the increase

Risk of fire from overheating Decreased production of nitrous gases Overheating the first and second nitration vessels Increased rate of release of nitrous gases Overheating in nitration vessels 1 and 2

Flow decreased to the tank to discharge emergency

Flow increased to the tank to discharge emergency

of flow would increase the rate in which NOx is released, creating an overpressure before a thermal explosion. Analyzing the parameter “reaction”, is relevant because the nitration step is essential for RDX production. In relation to this parameter, one identifies the possibility of occurrence of the following deviations: none, less and more. For them, there is a possibility of failure in nitration vessel 3 and/or nitration vessel 2 and hexamine feeder HNO3 @99% dosage failure.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

380

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

The suggested remedial measure is to regulate the flow into the tank for discharge in emergency (depending on the nature of the deviation) via regulating valves already existing. NODE #3 The third node (node #3) is located between the hexamine feeder and boiler 1. An analysis was made for flow, temperature and reaction. Table 7 is the result of the HazOp for this node. Among the operations occurring in node #3, there is hexamine addition into the mixture and heating in a vessel to 70°C (after nitration, which is carried out at 18°C). This process completes the conversion of hexamine into RDX. Nitrogen oxide gases are formed as by-products. These are released to another vessel to prevent encapsulation within the crystals of RDX in the form of acids. Following nitration, which occurs at 18°C, the boiler heats the mixture to 70°C. Boiling safely completes the conversion of hexamine into RDX. Boiling also releases the nitrogen oxide gases. The flow in this node can increase due to failures in tank RDX or HNO3 @99% Feed. Decrease of flow is related to obstruction or crushing of the supply line. The reaction parameter may be increased or decreased depending on the increase or decrease of flow equipment. As a solution to the differences observed in this node, it is suggested to carry out regular maintenance to prevent the clogging of pipelines and of feeder for hexamine, which is replaced when necessary. Another measure is that it is necessary to interrupt the operation in case of overflow. Deviations in temperature are directly related to flow. Identification of these deviations can occur by inspection and by instruments (thermal sensors). The main consequence of temperature deviations is decrease of yield and interruption of the manufacturing process. Over-temperature increases pressure in vessels and reaction rate. Control valves and exhaust valves are required. The control valves operate integrated to sensors that measure temperature in the reactor and provide this information to a controller, which operates the valves to match set points. This safety system, if well stabilized, prevents any deviation in temperature. Furthermore, as part of redundancy due to the explosive nature of the mix, as the temperature rises in this node, the control system enters a “state of alert”. When the temperature rises to 12°C the Programmable Logic Controller (PLC) alerts the operator by a light in the panel, allowing the operator to monitor the situation closer. Continued increase in temperature

to 15° C triggers a siren, thereby increasing the alertness level. At 18°C, a bottom valve opens and discharges the contents. Dilution by water lowers the temperature, dilutes the acid thereby interrupting the reaction, and eliminates the risk of explosion. Reaction time may differ slightly due to variation in feeding hexamine and hence to temperature change in equipment (boilers). There is also a risk of losing the batch in an accidental fire. NODE #4 The forth node (node #4) extends between the two boilers. Analyzed parameters are “flow”, “pressure”, and “temperature”. Table 8 is the result of the HazOp for this node. The reaction parameter was not considered here because there are no chemical processes before or after this node. The parameter “flow” may increase by issues in the vessel for HNO3 @99% or due to failure in the first boiler. Regarding interruption or flow reduction, there is the possibility of obstruction in the piping. To solve this problem it is suggested to implement maintenance programs and piping repair and replacement, as necessary. Deviations for flow can be identified by the presence of material in the emergency dump vessel. In this situation, the risks involved and response mechanism are analogous to those in node #2. Pressure changes according to the flow from boiler 1 regarding the keywords “none,” “less”, and “more”. These keywords lead to variation in the flow of cold water (feed water). To address this situation, one can establish a maintenance program for the boilers and install a pressure gauge for the control parameter. The temperature (node #4) may increase or decrease, depending upon the decrease or increase in the pressure in boiler 1 or variations in the flow of wastewater, respectively. Temperature effects can be controlled with the implementation of instruments and a command and control system. NODE #5 The fifth node (node # 5) is positioned between the second boiler and the second heat exchanger. Analysis of this node considers “flow”, “pressure” and “temperature”. Table 9 is the result of the HazOp for this node. Like node #4, there is no need to address the reaction parameter. No reaction occurs at this node. The parameter “flow” may be increased by faults in the tank for HNO3 @60% or because of failure in boiler 1. Flow interruption or reduction is due to obstruction in the piping, as well as the causes listed

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

381

Table 7. Node #3 – Connection between the nitration vessel and the boiler. Guide word

Deviation

Cause

Detection

None

Failure feeder RDX

Overheating in boiler

Less

Failure tank HNO3@99% Clogging the duct Driving

Flow

Less (< 12°C)

(> 15°C)

Damage to the boiler

Replacing the feeder for hexamine

Product loss Risk of fire

Visual Inspection

Damage to the boiler

Failure tank HNO3@99%

flow Increase for boiled water baffle

Product loss

Decreased flow between equipment

Increased flow between equipment

Duct obstruction driving

Checking through instruments (thermometer)

Interruption of proceedings Over-pressure vessels Increased speed reaction Risk of explosion due to a sudden pressure relief in a closed vessel Leak of NOx Interruption of production

Cooling in boilers 1 and 2

Duct repair manual

Maintenance dosing for hexamine Interruption of proceedings

Loss of income from reaction Verification by local inspection

Maintenance dosing for hexamine Replacing the feeder for hexamine

Overheating in boiler

Failure feeder RDX Less

Reduced flow baffle for boiled water

Mitigation

Failure feeder RDX

Temperature More

Visual Inspection

Consequences

Loss of quality Yield loss System under heating

Use of temperature control valves

Use of drain valves for regularization Use of temperature control valves Interrupting the process to repair feeder in

Interruption of production

Reaction More

Failure feeder RDX

Heating in boilers 1 and 2

Loss of quality Yield loss Risk of fire from overheating

Interrupting the process to repair feeder RDX

Leak of NOx in the workshop

for flow increase. Dosing accurately the hexamine can mitigate both deviations. Pressure at node #5 varies due to gas leakage. This is a case of pressure reduction, which may be caused by an increase in the generation of nitrogen oxide gases, as well as a decrease in the capability of releasing pressure. Thus, there must be a pressure gauge at this point, and a pressure relief valve and enabling the discharge of nitrogen oxide gases. The increase or decrease in temperature at this point (node #5) is related to decreased pressure or reduced inflow of cold water in the heat exchanger, as well as an increase in the pressure and inflow of cold water in the heat exchanger, respectively. These

variations relates to the variation in the return flow of cooling water feed, as well as the flow of coolant water. The consequences of temperature increase at this point (node #5) do not generate events with potential major accidents or explosions, but can produce burns in the case of leakage. Inspection routines or changing the system for the return flow of chilled water and cold water feed will control temperature deviations. NODE #6 The sixth node (node #6) is located between the heat exchanger and vacuum filters. To study this node, the recommended parameters are “flow” and “temperature”, since there is no

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

382

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

Table 8. Node #4 – Connection between boilers. Guide word

Deviation

Cause

None

Failure at first boiler

Less

Failure tank HNO3@99% Duct obstruction in driving

Detection

Consequences

Interruption of the production process Under heating the boiler 2 Reduction of flow baffle for wastewater Interruption of the production process

Flow Failure boiler 1 More

Less

Failure tank HNO3@99%

Decreased flow at first boiler

Overheating the boiler 2 Flow presence of the boiler 1 to the emergency dump tank Flow Increase for the baffle wastewater Interruption of the production process Verification by instruments Breaking duct driving

Pressure More

Less (< 12°C) Temperature More (> 15°C)

Increased flow at first boiler Decreased Pressure on boiler 1 Flow Increase for the baffle for wastewater Increased Pressure in the kettle 1 Flow decreased to baffle for wastewater

Checking through the instruments consolidated control panel unit

Verification by local inspection Checking through instruments (thermometer) Verification flow to the baffles

chemical reaction near this node. It remains at atmospheric pressure. Table 10 is the result of the HazOp for this node. Flow (at node #6) can increase due to a failure in the heat exchanger, a failure in the tank to HNO3@60%. Due to the absence of reaction, any flow issue relates to a failure in the tank and/or an obstruction in the flow driving. The consequence of these deviations is the changing of the load on the vacuum filter (increase, decrease or lack of flow).

Interruption of production Loss of production quality due to the contribution of the advancement of chilled water

Insufficient progress chilled water to maintain operations Interruption of operation of the discharge valve background

Decreased Flow for the heat exchanger Increased feed of cold water Risk of explosion due to a sudden pressure relief in a closed vessel

Mitigation

Maintenance first boiler Maintaining the tank HNO3@99% Duct repair manual

Maintenance first boiler Maintaining the tank HNO3@99%

Maintenance first boiler Installing a meter or pressure gate Installation and calibration of Pressure relief valves are configured as recommended in regulations

Decrease in advance of cold water Increase in reaction time Reduced yield factor

Risk of explosion due to a sudden pressure relief in a closed vessel

Installation of control valves temperature Installation flush valves bottom

Risk of Explosive initiation

Correction of deviations to flow requires installation of a flow control valve between equipment, periodic maintenance and a flow meter in the piping. Maintenance is a critical aspect for this node as well as the previous nodes. Further analysis of deviation in temperature at node #6 indicates that it relates to changes in the flow of cold water. Installing equipment for monitoring as well as for cooling or heating the product to be filtered can control these deviations.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

383

Table 9. Node #5 – Connection between boiler and heat exchanger. Guide word

Deviation

Cause

None

Failure in the tank for HNO3@60%

Less

Failure in the first boiler

Detection

Under heating of the heat exchanger

Duct obstruction in driving

Flow

Failure in the tank for HNO3@60% Failure in the first boiler

None

Less

Gas Leak

Increased of NOx output

Pressure

Overheating of the heat exchanger Visual inspection Checking through instruments (Pressure gauge) or control panel operating parameters Decrease the temperature in the heat exchanger Visual inspection

Decreased output of nitrous gases

Decreased pressure Less (< 60°C)

Reduction of flow entry of cold water in the heat exchanger

Temperature Increased pressure More (> 80°C)

Flow of increased cold water inlet in the heat exchanger

Consequences

Mitigation

Pressure decrease the heat exchanger

Increasing the flow of feedstock feeder RDX

Heating in the heat exchanger

Increased Pressure in the heat exchanger Cooling Heat Exchanger

Reducing the flow of raw material in the feeder RDX

Interruption of production Ineffectiveness of heat exchange, compromising the yield and economic balance of reaction

Checking through instruments (Pressure gauge) or control panel operating parameters

Risk of explosion due to a sudden pressure relief in a closed vessel

Increased temperature in the heat exchanger

Risk of explosive initiation

Installing a meter at this point Pressure Installing a valve for controlling the output of nitrous gases at this point

Verification by local inspection Checking through instruments (thermometers) Reduction of flow chilled water return

Ineffectiveness of heat exchange, compromising the yield and economic balance of reaction Use valve for controlling the water outlet from the cold heat exchanger

flow increase in feed cold water Verification by local inspection Checking through instruments (thermometers) Increasing the flow of chilled water return

There is no risk of explosion, just leak fluid heated, which creates potential for burns

Decreased feed flow cold water

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

384

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

Table 10. Node #6 – Connection between the heat exchanger and the vacuum filter. Guide word

Deviation

Cause

Detection

Consequences

Mitigation

None Mechanical failure in the heat exchanger Less Flow

Less

Failure in the tank for HNO3@60% Duct obstruction driving

Increasing the amount of produced hexogen

Failure in the tank for HNO3@60%

Identification of overflow material

Flow decrease in cold water

Flow increase in cold water

Under load in the vacuum filter

Installing a control valve flow Duct repair

Mechanical failure in the heat exchanger

Temperature More

Decreasing the amount of produced hexogen

Regular benchmarking of verification of equipment (meter)

Verification by local inspection Checking through instruments (thermometer)

NODE #7 The seventh node (node # 7) is located between the vacuum filters and the output of RDX. Flow is the only relevant parameter here since there is no chemical reaction and the flux between filters connection and product exit occurs in atmospheric pressure. Table 11 is the result of the HazOp for this node. The parameter, “flow”, indicates the failure of production in previous equipment, which is critical in this node. In order to identify the causes of deviation in the flow, one uses the keywords “none,” “less”, and “more”. As a general rule, all deviation in flow relates to failure in vacuum filters (clogged or damaged) or obstruction of piping. Deviations for this node (node #7) are observable through loss of product quality and operational variation of the exhaust system. A possible mitigation is to ensure maintenance of this equipment.

Overhead in the vacuum filter

Interruption of operation of the filter Interruption of operation of the filter Damage to the filter

Regularization of flow through safety valves Performing batch filtering

Establish procedures and regular targets for maintenance of existing control systems Installing a hardware check (thermometer) Interrupting the natural cooling process for product

DISCUSSION Addressing the HazOp results from a management perspective, operation of the control system around the same variables and set points is critically important. This starting point provides the means to observe and address all possible (and sometimes expected) deviations. This integrated unit could be a PLC panel that monitors and controls all parameters relevant to the operation (temperature, pressure and flow rates). Deviations in temperature are controllable through valves and/or temperature control equipment. In the RDX manufacturing unit, these valves should be installed along the pipeline to ensure safety. A possible suggestion for improvement of in-service units is to modernize the existing control panel to ensure the most

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

385

Table 11. Node #7 – Connection between the vacuum filter and the RDX recovery pan. Guide word

Deviation

None Less Flow More

Cause

Failure in vacuum filters Clogging the duct Driving

Failure in vacuum filters

Detection

Visual Inspection Verification of the final product - not or poorly crystallized Flow Increase for the tank HNO3@60%

accurate levels of control. However, this intervention would require readjustment of the major part of the unit, making this option costly and sometimes financially unviable. During the study, some safety bypasses were plotted, but neglected for HazOp purposes. Usually a bypass is located in stretches of critical piping and valves between departure and arrival of fluids. These safety devices ensure operational continuity of flow since they provide an alternate path for process fluids. They are, therefore, vital to solving the identified deviations in the parameter, “flow”. This statement allows one to conclude that design engineers addressed the same issues discussed in this paper. Data loggers found in most RDX units offer redundant storage of measurements in addition to storage devices present in the control panel. In the storage vessels, one solution to the problem of deviations in the flow is an emergency discharge system. In the first node, for example, in case of failure there is the alternative of pumping fluid into the HNO 3 vessel. This will stop the process and reaction bulk would flow into the tank that contains HNO3@60%. Furthermore, from the analysis of the design of an RDX manufacturing unit, one should conclude that some equipment provides redundancy, most likely for predicting the occurrence of failures as discussed in this paper. Action taken as a result of this analysis ensures, in some cases, the continuity of the process and in others, safe termination. Since many of the control measures suggested in response to the possible consequences listed in this study are already present in most RDX plants, there is the logical assumption that a HazOp-type study has occurred prior to project implementation. However, absence of documentation of such studies in the open literature increases the relevance of this paper.

Consequences

Mitigation

Low yield factor

Repair of vacuum filters

Under functioning the hood

Maintenance or replacement duct

Low yield factor

Maintenance or replacement duct

Over functioning the hood

Driving

RDX manufacturing plants usually contain pneumatically controlled and powered equipment. This requires a source of compressed air. The compressed air system was not addressed as part of this work. Installation of electronic sensing and control devices in RDX plants should increase accuracy and precision of operational parameters. However, these devices will not guarantee greater reliability in the safety system. Bottom valves, due to pneumatic operation and control, must remain in the CLOSED position during operation and fail in the OPEN position. This increases safety, since any loss of energy (or another deviation) will cause the valve to open, thereby discharging the reacting material into the dump vessel. Most RDX units contain PLCs to monitor and control operating parameters. These communication systems allow the transmission of broadband data through the electrical grid. Thus would be allowed to monitor the unit for any computer, provided they are properly authorized, through Internet, favoring the monitoring of security conditions. Another advantage of the system is that its implementation requires no physical alterations on site, once the grid is already in place.

CONCLUSION This paper assessed the risks of operation of an RDX manufacturing unit using HazOp methodology. This paper focused on the core issues of such a manufacturing unit, since the nodes were applied only at seven points, considered critical to the overall operation. Upon conclusion of this particular HazOp, each node provided key information that should be

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

386

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

Table 12. Study contributions and final results. Node

Node description

Key equipment

Array of vessels for waste and clean water for dilute HNO3

Water vessels

Recommendations

Installation of control valves to assure flow management. Implementation of level and temperature instruments. All instrumentation should be integrated. Implementation of a system to dose hexamine automatically

Connection between nitration vessels

Connection between the nitration vessel and the boiler

Nitration vessel

Installation of a venting system in the equipment to prevent overpressure and the consequent risk of explosion is recommended. Use of an empty dump tank and a draining valve at bottom. Favor mechanical valves over electronic ones in this key position. This operation (boiling) will release the nitrogen oxide gases, requesting a venting/exhaustion system.

Post-nitration boiler

It is necessary a mechanism to interrupt the operation in case of overflow. Deviations in temperature are directly related to flow.

Connection between boilers

Boiler

Deviations for flow can be identified by the presence of material in the emergency dump vessel. In this situation the risks involved and response mechanism are analogous to those in node #2. Installation of a pressure gauge for the control parameter. Temperature effects should be controlled with the implementation of instruments and a command and control system.

Connection between boiler and heat exchanger

Connection between the heat exchanger and the vacuum filter

Vacuum filter

Connection between the vacuum filter and the RDX recovery pan

Filtering system

Heat exchanger

Dosing accurately the hexamine is key to avoid deviations in flow and temperature. This operation (boiling) will release the nitrogen oxide gases, requesting a venting/exhaustion system. Installation of control valves and periodic maintenance, as well as a flow meter should be considered Installing equipment for monitoring as well as for cooling or heating the product to be filtered should be considered

considered when designing or operating any RDX nitration unit. Those key aspects are consolidated in Table 12. Furthermore, the results show that the guidewords used for selected nodes (flow, pressure, temperature and reaction) and their deviations made possible a complete diagnostic of the RDX nitration unit intrinsic risks. This approach also made possible the analysis of types of detection, consequences of failure, and steps to be taken for each deviation identified.

Deviations for this node (node #7) are observable through loss of product quality and operational variation of the exhaust system. AÂ possible mitigation is to ensure maintenance of this equipment.

This study also indicates that a risk assessment had occurred on the unit examined in this study previously. This is evident mainly due to existence in the flowcharts of the required mitigation measurements (such as safety valves, bypasses and control instruments). Finally, it is important to mention that HazOp methodology does not quantify risk. For this reason, other tools for risk assessment should complement HazOp.

J. Aerosp. Technol. Manag., SĂŁo JosĂŠ dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

Risk Assessment for Hexamine Nitration into RDX

387

REFERENCES Achuthan, C.P. and Mullick, G., 1983, “Hazards in the manufacture of RDX and HMX”, Defence Science Journal, Vol. 33, No. 1, pp. 91-95.

Management”, Proceedings of the European Safety and Reliability Conference, ESREL 2009. [S.l: s.n.], p. 2009.

Akhavan, J., 2011, “The Chemistry of Explosives”, Third Ed., Norfolk: Biddles Ltd., Kings Lynn, Norfolk.

Galante, E., Haddad, A. and Marques, N., 2013, “Application of Explosives in the Oil Industry”, International Journal of Oil, Gas and Coal Engineering (OGCE), Vol. 1, No. 2, pp. 16-22. doi:10.11648/j. ogce.20130102.11.

Aven, T., 2012, “On the critique of Beck’s view on risk and risk analysis”, Safety Science, Vol. 50, No. 4, pp. 1043-1048. doi:10.1016/j.ssci.2011.12.004. Bachmann, W.E. and Sheehan, J.C., 1949, “A new method of preparing the high explosive RDX”, Journal of the American Chemical Society, Vol. 71, No. 5, pp. 1842-1845. doi: 10.1021/ ja01173a092. Bisarya, R.K. and Puri, S., 2005, “The Bhopal gas tragedy – A perspective”, Journal of Loss Prevention in the Process Industries, Vol. 18, No. 4-6, pp. 209-212. doi:10.1016/j.jlp.2005.07.006. Boonthum, N., Mulalee, U. and Srinophakun, T., 2014, “A systematic formulation for HAZOP analysis based on structural model”, Reliability Engineering & System Safety, Vol. 121, pp. 152-163. doi:10.1016/j. ress.2013.08.008. van den Bosch, C.J.H. and Weterings, R.A.P.M., 2005, “Yellow Book” - CPR 14E - Methods for the calculation of Physical Effects. Third ed. [S.l.]: Committee for the Prevention of Disasters. CCPS, 1989, “Guidelines for Chemical Process Quantitative Risk Analysis”, CCPS - CEN ed. New York: Center for Chemical Process Safety of the American Institute of Chemical Engineers. Chouhan, T.R., 2005, “The unfolding of Bhopal disaster”, Journal of Loss Prevention in the Process Industries, Vol. 18, No. 4-6, pp. 205208. doi:10.1016/j.jlp.2005.07.025. Cooper, P.,1996, “Explosives Engineering”, [S.l.]: Wiley-VCH, Inc. United States of America. Crawley, F., Preston, M. and Tyler, B., 2000, “Hazop: Guide to best practice. Guidelines to best practice for the process and chemical industries”, [S.l.]: Institution of Chemical Engineers (Great Britain), European Process Safety Centre. United Kingdom: The Cromwell Press. Defence, D.D., 2012, “MIL STD 882-E - Standard Practice for System Safety”. Dunjó, J., Fthenakis, V., Vílchez, J.A. and Arnaldos, J., 2010, “Hazard and operability (HAZOP) analysis. A literature review”, Journal of hazardous materials, Vol. 173, No. 1-3, pp. 19-32. doi:10.1016/j. jhazmat.2009.08.076. Eckerman, I., 2005, “The Bhopal gas leak: Analyses of causes and consequences by three different models”, Journal of Loss Prevention in the Process Industries”, Vol. 18, No. 4-6, pp. 213-217. doi:10.1016/j.jlp.2005.07.007. Fogler, H., 1999, “Elements of chemical reaction engineering”, [S.l: s.n.]. Freeman, K., Lee, R. and Mcnamara, T., 1992, “Plan HAZOP studies with an expert system”, Chemical Engineering Progress, Vol. 88, No. 8, pp. 28-32. Galante, E.B.F., Boer, A.H.D. and Bonifácio, D., 2014, “Life Cicle Inventory for Lead Azide Manufacture”, Journal of Aerospace Technology and Management, Vol. 6, No. 1, pp. 53-60. doi:10.5028/jatm.v6i1.289. Galante, E. and Haddad, A., 2009, “Risk analysis procedures for explosives manufact ... Advances in Safety, Reliability and Risk

Gehlawat, J.K., 2005, “Bhopal disaster – a personal experience”, Journal of Loss Prevention in the Process Industries, Vol. 18, No. 4-6, pp. 261-263. doi:10.1016/j.jlp.2005.07.002. Graf, H. and Schmidt-Traub, H., 2000, “Early hazard identification of chemical plants with statechart modelling techniques”, Safety science, Vol. 36, No, 1, pp.49-67. doi: 10.1016/S0925-7535(00)00034-5. Haddad, A., Galante, E., Caldas, R. and Morgado, C., 2012 “Hazard Matrix Application in Health, Safety and Environmental Management Risk Evaluation”. InTech. [S.l: s.n.]. doi: 10.5772/17159. Held, M., 2012, “Explosion in an HMX (octogen) plant”, In: AICHE ANNUAL MEETING. Anais... [S.l: s.n.]. IMBEL, 2006, “IMBEL”, Retrieved in July 17, 2013, from www.imbel. gov.br Júnior, J.B.C., 2002, “Metodologia de Análise de Risco em Sistemas Computacionais de Aplicação Crítica”, Escola Politécnica da Universidade de São Paulo - [S.l.]. Kennedy, R. and Kirwan, B., 1998, “Development of a Hazard and Operability-based method for identifying safety management vulnerabilities in high risk systems”, Safety Science, Vol. 30, No. 3, pp. 249-274. doi:10.1016/S0925-7535(98)00025-3. Khan, F.I. and Abbasi, S.A., 1999, “Major accidents in process industries and an analysis of causes and consequences”, Journal of Loss Prevention in the Process Industries, Vol. 12, No. 5, pp. 361378. doi:10.1016/S0950-4230(98)00062-X. Khan, F.I. and Abbasi, S.A., 1997, “Tophazop: a knowledge-based software tool for conducting HAZOP in a rapid, efficient yet inexpensive manner”, Journal of Loss Prevention in the Process Industries, Vol. 10, No. 5-6, pp. 333-343. Kletz, T., 2009, “What Went Wrong?, Fifth Edition: Case Histories of Process Plant Disasters and How They Could Have Been Avoided”, 5th. ed. Oxford, UK: Elsevier Inc. Labovský, J., Svandová, Z., Markos, J. and Jelemenský, L., 2007, “Model-based HAZOP study of a real MTBE plant”, Journal of Loss Prevention in the Process Industries, Vol. 20, No. 7, pp. 230-237. doi: 10.1016/j.jlp.2007.03.015. Leach, J.T. and Staples, J.M., 1981, “Hazards of inadvertent mixing of chemicals used in the Bachmann processes for manufacturing the military explosives RDX and HMX”, Journal of Hazardous Materials, Vol. 4, No. 3, pp. 271-281. doi:10.1016/03043894(81)80005-2. Lukasavage, W.J. and Slagg, N., 1993, “Process for preparation of RDX. n. Patent number US5250687”, International Classification: C07D25106. Mei, Z., Qian, H., Liu, D., Ye, Z. and Lü, C., 2012, “Preparation of emulsion explosive for military-civil”, Huozhayao Xuebao/Chinese Journal of Explosives and Propellants, Vol. 35, No. 1, pp. 32-34.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

388

Galante, E.B.F., Costa, D.M.B., Haddad, A.N. and Santos, I.J.A.L.

Meredith, J.A., 1976, “Preparation of RDX”. n. [s.n.]. Meyer, R., Köhler, J. and Homburg, A., 2007, “Explosives”. 6th, Compl. ed. Weinheim: Wiley-VCH Verlag GmbH, Weinheim. Pérez-Marín, M. and Rodríguez-Toral, M.A., 2013, “HAZOP – Local approach in the Mexican oil & gas industry”, Journal of Loss Prevention in the Process Industries, Vol. 26, No. 5, pp. 1-5. doi:10.1016/j. jlp.2013.03.008. Rausand, M. and Hoyland, A., 2011, “HAZOP – Hazard and Operability Study. System Reliability Theory”, Models, Statistical Methods and Applications. Second Edi ed. New York: John Wiley & Sons. Rausand, M. and Utne, I.B., 2009, “Product safety – Principles and practices in a life cycle perspective”, Safety Science, Vol. 47, No. 7, pp. 939-947. doi:10.1016/j.ssci.2008.10.004. Rossing, N.L., Lind, M., Jensen, N and Jørgensen, S.B., 2010, “A functional HAZOP methodology”, Computers & Chemical Engineering, Vol. 34, No. 2, pp. 244-253. doi:10.1016/j. compchemeng.2009.06.028. Rouhiainen, V. and Gunnerhed, M., 2002, “Development of international risk analysis standards”, Safety Science, Vol. 40, No. 1-4, pp. 57-67. doi:10.1016/S0925-7535(01)00034-0. Sahar, N.B.M., Ardi, S., Kazuhiko, S., Yoshiomi, M. and Hirotsugu, M., 2010, “HAZOP analysis management system with dynamic visual model aid”, American Journal of Applied Sciences, Vol. 7, No. 7, pp. 930-935.

Schüller, J.C.H., Brinkman, J.L., Van Gestel, P.J. and van Otterloo, R.W., 1997, “Red Book” - CPR 12E - Methods for determiming and processing probabilities, Netherlands: CPR 12 E - Committee for Prevention of Disasters. Seccatore, J., Origliasso, C. and Tomi, G., 2013, “Assessing a risk analysis methodology for rock blasting operations”, In: Paper Presented at the Blasting in Mines - New Trends: Workshop Hosted by Fragblast 10 - The 10th International Symposium on Rock Fragmentation By Blasting. Anais... [S.l: s.n.]. Shimada, Y., Kitajima, T. and Ling, Y., 2012, “Integrated risk management system based on IPL concept for chemical processes”, In: 11th International Probabilistic Safety Assessment and Management Conference and The Annual European Safety and Reliability Conference 2012, PSAM11 ESREL 2012,. Anais... [S.l: s.n.]. Steen, R. and Aven, T., 2011, “A risk perspective suitable for resilience engineering”, Safety Science, Vol. 49, No. 2, pp. 292-297. doi:10.1016/j.ssci.2010.09.003. Taylor, J.R., 2007, “Understanding and combating design error in process plant design”, Safety Science, Vol. 45, No. 1-2, pp. 75-105. doi:10.1016/j.ssci.2006.08.014. Urbanski, T., 1984, “Chemistry and tecnology of explosives”, Vol 4. New York, USA: Pergamon Press.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.373-388, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.400

Microstructural Analysis of Co-Free Maraging Steel Aged Aline Castilho Rodrigues1, Heide Heloise Bernardi1, Jorge Otubo2

ABSTRACT: Maraging steels have low carbon content and are highly alloyed, having as main feature the ability to increase the mechanical strength after thermal aging. Therefore, the objective of this work is to analyze the effect of aging in a Cofree maraging steel, Fe – 0.014% C – 0.3% Mn – 3.9% Mo – 2.1% Cu – 0.19% Si – 11.8% Cr – 9.1% Ni – 1.0% Ti (wt), at different heat treatment times (10 min - 960 min) at constant temperature (550°C), after cold rolling up to 66% and 77% area reduction. Microstructures were analyzed by scanning electron microscopy (SEM - EDS) and mechanical properties by Vickers hardness measurements. The results showed a significant increase in the hardness of the material after aging heat treatment on the solution treated and deformed samples. The aging heat treatment which was harder about 650 HV was 550°C/60 min. KEYWORDS: Maraging steel, Cold rolling, Hardness test Aging.

INTRODUCTION Maraging steel was developed in the 1950’s, to support the demand for a low density and temperature resistant steel (250°C – 300°C), for improving aircraft performance. This new material was created adding aluminum (Al) and titanium (Ti) in steel with nickel (Ni). In the 1960’s, cobalt (Co) and molybdenum (Mo) were added into the composition of maraging steels. Clarence George Bieber, in International Nickey Company, studied the addition of these elements in the alloy and concluded that there is a significant increasing in mechanical resistance (Lopes, 2007; Méndez, 2000). The classification of commercial maraging steels, according to the percentage of nickel, is: 18Ni (250), 18Ni (300) and 18Ni (350) (Lopes, 2007). Maraging steel received this name due to two words: martensite and aging, which mean martensite aged. Low carbon and alloyed, the maraging steels are capable of increasing their mechanical resistance and hardness after aging heat treatments. The treated solution state has a martensitic structure with high ductility and toughness, and it can be modified after aging heat treatments (Cardoso et al., 2013a). Due to the ductility of martensite, the maraging steel allows cold mechanical forming, not requiring intermediate heat treatment. A mechanical processing technique that can be used in maraging steel is cold rolling. During the cold rolling, plastic deformation and increase of hardness occur, therefore, there is an increasing of dislocation density and crystalline defects, hindering the movement between them. Due to the plastic deformation, there is an increase in tensile strength and yield point (Callister, 2012). On the other hand, these mechanical properties can also be improved by thermal aging, in which the formation of intermetallic precipitates occurs. In Table 1,the types of precipitates which may be formed

1.Faculdade de Tecnologia de São José dos Campos – São José dos Campos/SP – Brazil 2.Instituto Tecnológico de Aeronáutica – São José dos Campos/SP – Brazil Author for correspondence: Aline Castilho Rodrigues – Faculdade de Tecnologia de São José dos Campos – FATEC Prof. Jessen Vidal | Av. Cesare Mansueto Giulio Lattes, s/n, São José dos Campos/SP | CEP: 12.247-014 – Brazil | Email: alinerodrigues_1@msn.com Received: 08/02/2014 | Accepted: 09/30/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.389-394, Oct.-Dec., 2014

390

Rodrigues, A.C., Bernardi, H.H. and Otubo, J.

Table 1. Precipitates hardeners formed during the aging of maraging steels (Padial et al., 2000). Phase

Crystalline Structure

Ni3Mo

Orthorhombic

Ni3Ti

Hexagonal ordered

Ni3V

Hexagonal compact

Ni3W

Orthorhombic

Fe2(Mo, Ti)

Type hexagonal laves

FeMo

Tetragonal

FeTi

CsCl type cubic

Fe7Mo4

Hexagonal

R (Mo-Co-Cr)

Hexagonal rhombic

X (Fe-Cr-Mo)

Body-centered cubic

on maraging steels, depending on the alloy composition, are shown (Padial et al., 2000). Steel with martensitic structure, when subjected to a heat treatment at a constant temperature, suffers a growth of intermetallic precipitates uniformly in the matrix (Carvalho et al., 2013). The precipitate size is a contributing factor to the increased mechanical resistance of maraging steels. The smaller the precipitates and the more coherent with the matrix, the greater the hardness of the alloy, since there is less space between the particles; furthermore, they act as barriers to movement of dislocations (Padial et al., 2000). The alloying elements of maraging steel, such as molybdenum, titanium, nickel, cobalt, aluminum, and others, have a great influence on the mechanical properties of this alloy, in other words, they are responsible by precipitates hardeners (Carvalho et al., 2013). However, recent studies (Hu et al., 2008; Leitner et al., 2011; Mahmoudi et al., 2011; Mahmudi et al., 2011; Nili-Ahmadabadi, 2008; Schnitzer et al., 2010; Sha et al., 2013) are being conducted in maraging steels without cobalt addition, in order to reduce the production costs of this alloy. Maraging steels are considered expensive due to their materials’ preparation and expensive alloy elements such as nickel and cobalt processes. The elimination of the cobalt content and the substitution of nickel by cheaper elements, such as manganese, has been studied. Cobalt is used only to minimize the solubility of molybdenum. In the case of maraging steel studied, this element was replaced by chrome, which besides

improving the hardenability increases the corrosion resistance (Nili-Ahmadabadi, 2008). Maraging steels are mainly applied in the aeronautical, aerospace, nuclear and military industries, as they have great advantages such as good weldability, high strength, high yield strength, high fracture toughness, they support high working temperatures, have good machinability, good formability, among others, with great applicability matrices and tools (Lopes, 2007). High-tensile steels such as 300M and SAE 4340 have extensive application in aeronautics and aerospace industry, being mainly applied in the manufacture of structural components (Zang et al., 2013; Boakye-Yiadom et al., 2014). Maraging steels are being investigated because of the possibility of replacing the commercial alloy steels, 300M and 4340 for example, widely used in aeronautical industry (Carvalho et al., 2013). Thus, this study aims to examine the effect of aging heat treatment in non-commercial maraging steel (Co-free), in order to compare it to the 300M and 4340 steels.

EXPERIMENTAL The maraging steel used in this work was produced by vacuum induction melting. The ingots was hot forged and then hot rolled until we obtained bars of 20x20 mm of dimension and chemical composition: Fe – 0.014% C – 0.3% Mn – 3.9% Mo – 2.1% Cu – 0.19% Si – 11.8% Cr – 9.1% Ni – 1.0% Ti (wt.%). Its bar (20x20 mm) was separated into two parts resulting in two conditions: • Condition I (C-I): bar dimension of 20 x 20 x 10 mm; • Condition II (C-II): bars of 20 x 20 x 10 mm (~100 wt.), remelted (1 remelt) on an arc furnace, resulting in an ingot dimension of 16x10x95 mm. Both conditions (C-I and C-II) were solution treated at 1050°C during 1 hour, quenched in water at room temperature and then cold rolled to a plate of 3 mm in thickness, resulting in 66 and 77 % of area reduction (AR) for C-I and C-II, respectively. The deformed samples (AR = 66 and 77 %) were cut in several parts and each part was annealed at 550°C for different times, ranging from 10 to 960 min. Microstructural characterization was performed using a Scanning Electron Microscope (SEM) VEGA 3 TESCAN model, equipped with an energy dispersive spectrometer system (EDS). Vickers hardness

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.389-394, Oct.-Dec., 2014

Microstructural Analysis of Co-Free Maraging Steel Aged

391

and HV77% = 407), corresponding to the effect of work hardening. About thermal aging, there is a significant increase in hardness at relatively low heat treating times (Fig. 1). For both area reductions, increased hardness around 200 HV occurs with only 10 min thermal aging. He et al. (2002) observed that the heat treatment at 470°C performed in a maraging steel with 18%wt. Ni, after forging, has achieved a 90% increase in hardness, with only 15 min of aging. In the detail of Fig. 1, it is observed that the maximum hardness is obtained at 550°C/60 min for both ARs. On the other hand, the thermal aging at 550°C exhibits a decrease in hardness values after 240 min for deformed sample with AR = 66%, whereas a decrease it is observed after 90 min to condition AR = 77%. Therefore, it can be said that in smaller ARs, there is an increase in aging time of the alloy, delaying the onset of overaging. Based on the average hardness values, the processing) in the matrix, resulting in a lath martensitic structure (Mahmoudi et al., 2011; Hu and Wang, 2012), as shown in Figs. 2a and 2b. Co-free maraging steel is stable up to 16 hours of aging at 550°C.

testing was performed using a Future-Tech FM-7000 microindenter with a load of 200 g/15s, in which the results correspond to the mean of ten values taken on each specimen. The micrograph analyses and the Vickers’ hardness were undertaken in a longitudinal section of the samples (wire rolling direction).

RESULTS AND DISCUSSION Figure 1 shows the hardness variation as a function of heat treatment time for maraging steel with AR’s = 66 and 77 %. For solution treated (C-I and C-II) and deformed samples, the hardness values are represented at zero time in Fig. 1. Solution treated samples (C-I and C-II) showed average hardness values of 250 and 300 HV, respectively. Comparing the solution treated to the deformed condition, it is observed an increase around 50 % in average hardness values (HV66% = 425

700 AR = 66%

650

AR = 77%

600 550 680 670 660

450

650 400

Hardness (HV)

500

350 300 250

630 620 610 600 590 580

200

570

550°C

560

150 100

640

100

200

300

400

40 500

80 600

Time (min)

100 120 140 160 180 200 220 240 260

Time (min)

700

800

900

1000

1100

Figure 1. Annealing behavior of the maraging steel deformed with AR’s = 66% and 77%. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.389-394, Oct.-Dec., 2014

392

Rodrigues, A.C., Bernardi, H.H. and Otubo, J.

AR = 66%

20 μm

AR = 77%

(a)

20 μm

(b)

AR = 66%

(c)

AR = 66%

(e)

20 μm

AR = 77%

20 μm

AR = 77%

20 μm

(d)

20 μm

(f)

Matrix

Matrix P2

P3 P5

AR = 66%

(g)

5 μm

AR = 77%

(h)

5 μm

Figure 2. Microstructures of maraging steel under the conditions (a); (b) solution treated; (c); (d) deformed; (e); (f) aging at 550°C/1h (SEM-BSE) and (g); (h) aging at 550°C/16h (SEM-SE). J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.389-394, Oct.-Dec., 2014

Microstructural Analysis of Co-Free Maraging Steel Aged

Figure 2 shows the microstructure of the maraging steel at different conditions. It is observed for solution treated sample (C-I and C-II) that the solution heat treatment at 1050°C for 1h was effective in dissolving primary precipitates (formed during the early stages of maraging steel After thermal aging at 550°C for 1 hour (Figs. 2e and 2f), it can be seen that the microstructure is similar to deformed samples in AR = 66 and 77 % (Figs. 2c and 2d). This heat treatment time showed to be effective in relation to increased hardness (Fig. 1), due to presence of fine precipitates dispersed in the matrix, which contribute to the increase of this mechanical property (Hu and Wang, 2012). However, the technique of scanning electron microscopy, used in order to analyze the microstructure of this alloy, is not efficient to identify them in the matrix. Increasing the annealing time to 16 h (Figures 2g and 2h) promoted an increment in precipitation rate. These precipitates grew coarser in subsequent higher annealing times, but there still are fine precipitates elongated in the rolling direction, and few precipitates changed from elongate morphology to a spherical one. In this condition, it is observed some coalescence of precipitates having sizes of the order of 2 μm. Regardless of AR, the precipitates have similar morphologies. The EDS analysis (Table 2) showed that those precipitates, referred as P2 and P5, were depleted in Ni, and enriched in Mo and Ti (Figures 2g and 2h). In the analysis performed on the matrix — points 1 and 4 in Figs. 2g and 2h, respectively —, compared to the nominal composition of the alloy, it can be said that the rolling processes or thermal treatments did not significantly alter the composition of the material in terms of Mo, Ti, Cr and Ni elements. In relation to precipitates P2 and P5, a large increase of Mo and Ti was observed. This alloy contains low nickel, so it is not possible to stimulate the formation of Ni3Mo or Ni3Ti phases, which usually appear in the commercial maraging steels 250 and 350 (Padial et al., 2000). The presence of cobalt element in commercial alloys promotes the formation of Ni3Mo. On the other hand, the maraging steel study in this work does not have Co (Co-free) and low Ni content. However, the alloy phase formed in this alloy can be Fe2(Mo, Ti), FeMo, FeTi or Fe7Mo4 (Padial et al., 2000). The precipitate P3 features a larger enrichment of Ti in relation to P2 and P5, probably, the Ti-rich precipitates are carbides (TiC) or Fe2Ti, FeTi (Castanheira et al., 2006). The TiC is formed by carbon, which is characterized as

393

Table 2. Chemical compositions of the constituent phases (wt.%). Constituent

Composition

3.9

1.0

9.1

11.8

Matrix 1

4.6

1.2

9.2

12.1

Precipitate 2 (P2)

16.0

3.0

4.7

14.1

Precipitate 3 (P3)

7.6

12.5

6.4

9.9

Matrix 4

4.5

1.0

9.0

12.5

Precipitate 5 (P5)

14.5

2.6

5.4

14.8

an impurity of the material (Padial et al., 2000). Furthermore, other techniques are necessary, for example, an X-ray diffraction to determine the type of precipitate formed. The SAE 4340 and 300M steels are considered hightensile steels, low-carbon and low-alloy steels, being used in the aerospace industry due to their excellent mechanical properties. Commercial maraging steels are being studied in order to replace the 4340 and 300M steels, especially in the aeronautical and aerospace industry. According to studies by Cardoso et al. (2013b), the SAE 4340 and 300M steels can achieve hardness values of 250 HV and 350 HV, respectively, after drawing back. However, the maraging steel studied with forming processes and appropriate heat treatment can achieve hardness values around 670 HV, i.e., almost double the amount reported by the 300M steel.

CONCLUSION Based on this investigation, the following conclusions can be drawn: • The cold forming contributes to mechanical hardening, however an effective increase of hardness is obtained after thermal aging; • The hardness increased significantly with low time thermal aging at 550°C due to precipitation of the second phase particles. The maximum value obtained was of 650 HV at 550°C/60 min; • Based on the average hardness values, the Co-free maraging steel is stable for up to 16 hours of aging at

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.389-394, Oct.-Dec., 2014

394

Rodrigues, A.C., Bernardi, H.H. and Otubo, J.

550°C. It is possible to note a subtle decrease in the hardness values for both AR, due to coarsening of precipitates; • Appropriate thermal aging promoted the precipitation of the second phase particles, rich in Mo and Ti.

ACKNOWLEDGEMENTS The authors are thankful to FAPESP, CNPq, CNPq Universal (476030/2011-0), DCTA/IEAv, Villares Metals SA and Multialloy-Metais e Ligas Especiais Ltda.

REFERENCES Boakye-Yiadom, S., Khan, A.K. and Bassim, N., 2014, “A systematic study of grain refinement during impact of 4340 steel”, Materials Science and Engineering: A, Vol. 605, pp. 270-285. doi: 10.1016/j.msea. 2014.03.066.

Leitner, H., Schober, M., Schnitzer, R. and Zinner, S., 2011, “Strengthening behavior of Fe-Cr-Ni-Al-(Ti) maraging steels”. Materials Science and Engineering: A, Vol. 528, No. 15, pp. 5264-5270. doi: 10.1016/j.msea.2011.03.058.

Callister, W.D.Jr., 2012, “Fundamentos da Ciência e Engenharia de Materiais: uma abordagem integrada”, Rio de Janeiro: Editora LTC.

Lopes, J.C.O., 2007, “Os aços Maraging”. Ciência e Tecnologia dos Materiais, Vol. 19, No 1-2, pp. 41-44.

Cardoso, A.S.M., Abdalla, A.J., Lima, M.S.F., Bonjorni, F.M., Barbosa, M.J.R., Baptista, C.A.R.P. and Fanton, L., 2013a, “Study of Laser Welding and Heat Treatments Done in Different High Strength Steels: 4340, 300M, Maraging 300”, SAE Technical Paper Series, Vol. 36, pp. 1-5. doi:10.4271/2013-36-0510.

Mahmudi, A., Nedjad, S.H. and Behnam, M.M.J., 2011, “Effects of cold rolling on the microstructure and mechanical properties of FeNi-Mn-Mo-Ti-Cr maraging steels”. International Journal of Minerals, Metallurgy and Materials, Vol. 18, No 5, pp. 557-561. doi: 10.1007/ s12613-011-0477-y.

Cardoso, A.S.M., Abdalla, A.J., Ueda, M., Lima, M.S.P. and Bonjorni, F.M., 2013b, “Microstructural Characterization of the 4340 and 300M Steels After Laser Welding, Heat Treatment and Surface Plasma”, In: SAE, São Paulo.

Mahmoudi, A., Ghavidel, M.R.Z., Nedjad, S.H., Heidarzadeh, A. and Ahmadabadi, M.N., 2011, “Aging behavior and mechanical properties of maraging steels in the presence of submicrocrystalline Laves phase particles”. Materials Characterization, Vol. 62, No. 10, pp. 976-981. doi: 10.1016/j.matchar.2011.07.012.

Carvalho, L.G., Andrade, M.S., Plaut, R.L., Souza, F.M. and Padilha, A.F., 2013, “A dilatometric study of the phase transformations in 300 and 350 maraging steels during continuous heating rates”, Materials Research, Vol. 16, No. 4, pp. 740-744. doi: 10.1590/S151614392013005000069. Castanheira, M., Tikhomirov, V.V., Ozerskiy, A.D., Da Silva, J.E.R. and Lucki, G., 2006, “Influência dos Elementos de Liga e do Tratamento Térmico na Estrutura e Propriedades Mecânicas de um Aço Maraging sem Cobalto Resistente à Corrosão”, Anais do 17º CBECIMat, Brasil. He, Y., Yang, K., Qu, W., Kong, F. and Su, G., 2002, “Strengthening and toughening of a 2800-MPa grade maraging steel”. Materials Letters, Vol. 56, No. 5, pp. 763-769. doi: 10.1016/S0167577X(02)00610-9. Hu, Z.F., Mo, D.F., Wang, C.X., He, G.Q. and Chen, S.C., 2008, “Different behavior in electron beam welding of 18Ni Co-free maraging steels”. Journal of Materials Engineering and Performance, Vol. 17, No. 5, pp. 767-771. doi: 10.1007/s11665-007-9190-4. Hu, Z.F. and Wang, C., 2012, “Effect of Tube Spinning With Subsequent Heat-Treatment on Performance and Microstructure Evolution of T250 Maraging Steel”. Journal of Iron and Steel Research International, Vol. 19, No. 5, pp. 63-68. doi: 10.1016/S1006706X(12)60101-0.

Méndez, D., 2000, “Una revisión de los aceros Maraging”, Revista Ciência Abierta, No. 28, Chile. Nili-Ahmadabadi, M., 2008, “Improvement in mechanical properties of Fe-Ni-Mn maraging steel by heavy cold rolling”. International Journal of Modern Physics B, Vol. 22, No 18-19, pp. 2814-2822. doi: 10.1142/S0217979208047638. Padial, A.G.F., Monteiro, W.A., Andrade, A.H.P. and Rigo, O.D., 2000, “Microstrutural Analysis os 400 grade Maraging Steel After Thermomechanical Treatment”. Thermec 2000, Estados Unidos. Schnitzer, R., Schober, M., Zinner, S. and Leitner, H., 2010, “Effect of Cu on the evolution of precipitation in an Fe-Cr-Ni-Al-Ti maraging steel”, Acta Materialia, Vol. 58, No. 10, pp. 3733-3741. doi: 10.1016/j. actamat.2010.03.010. Sha, W., Chen, Z., Geriletu, X.X.X., Lee, J.S., Malinov, S. and Wilson, E.A., 2013, “Tensile and impact properties of low nickel maraging steel”. Materials Science and Engineering: A, Vol. 587, pp. 301-303. doi: 10.1016/j.msea.2013.08.076. Zang, G., Yang, X., He, X., Li, J. and Hu, H., 2013, “Enhancement of mechanical properties and failure mechanism of electron beam welded 300M ultrahigh strength steel joints”. Materials and Design, Vol. 45, pp. 56-66. doi: 10.1016/j.matdes.2012.09.004.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.389-394, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.403

Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients Hamid Farrokhfal1, Ahmad Reza Pishevar2

ABSTRACT: This paper concerns a numerical optimization method for designing airfoils based on adjoint method. The goal of present work is to reduce the compressibility drag or pitching moment of transonic airfoils without compromising on the lift coefficient. A new cost function based on this requirement is defined and the corresponding adjoint equations are discussed in details. At the end, by demonstrating some numerical results, we show that this technique is capable of converging to the optimum design point corresponding to the initial geometry of the airfoil. KEYWORDS: Optimization, Pitching moment, Transonic airfoils, Adjoint equations, Euler equations.

INTRODUCTION Improvement of aerodynamic performance of helicopter blade designs has been one of the most important research areas in rotorcraft aerodynamics. Low-pitching-moment airfoils found application primarily as helicopter rotor blades, but some attention has been given to the advantages of low pitching-moment sections for a “span-loader” vehicle. For such applications, a symmetric airfoil could conceivably be employed, but cambered airfoils can offer significant advantages (Barger, 1975). Therefore, designing cambered airfoils with low pitching-moment is attractive in helicopter blade design to achieve small control forces in rotor controls, and this task can be conducted via optimizing procedures. Also, shock waves in transonic flow not only increase the wave drag but also cause unfavourable flutter or buffet to fixed wings. Helicopter rotor blades in forward flight are also frequently exposed to strong unsteady shock waves at the blade tip region. These shock waves increase the required torque, and become a source of undesirable noise and vibration. Thus, elimination or possible reduction of the strength of these shock waves would be desirable for the enhancement of the performance of helicopters and airplanes. In order to improve airfoil performance under different flight conditions and to make the performance insensitive to off-design condition at the same time, Yu et al. (2010) have developed a multi-objective optimization approach considering robust design and applied it to airfoil tailoring. They used non-uniform rational B-spline (NURBS) representation, control points and related weights around airfoil as design variables. They could

1.Malek-Ashtar University of Technology – Isfahan – Iran 2.Isfahan University of Technology – Isfahan – Iran. Author for correspondence: H. Farrokhfal | Department of Mechanical & Aerospace Engineering | Malek-Ashtar University of Technology – 83145-115 | Shahinshahr – Isfahan – Iran | Email: Farrokhfal@mut-es.ac.ir Received: 08/04/2014 | Accepted: 10/14/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

396

Farrokhfal, H. and Pishevar, A.R.

obtain a set of non-dominated airfoil solutions with robustness, by adopting multi-objective genetic algorithm that is based on non-dominated sorting. They showed that the multi-objective robust design method makes the airfoil performance robust under different off-design conditions. In the gradient-based optimization design, the gradients of cost/constraint functionals with respect to design variables are important information in the design processes (Xie, 2002). A traditional approach to calculate gradients is the finite difference method which involves finite differencing performance functionals computed from the solutions to the governing equations with perturbed parameter values. The Finite difference method (FDM) with complex variables was explored by Anderson et al. (2001) to improve its accuracy. Even though it is very easy to implement FDM in program coding, its prohibitive computational costs (many times solving governing equations) motivate other time-efficient methods to calculate reduced gradients. A way of performing an efficient aerodynamic shape optimization is to regard the design problem as a control problem in which the control is the shape of the boundary. This approach to optimal aerodynamic design was introduced by Jameson (1988, 1990) who examined the design problem for compressible flow with shock waves, and devised adjoint equations to determine the gradient for both potential flow and also flows governed by the Euler equations (Jameson, 1995; Reuther and Jameson, 1994). Also, particular interest has been given to adjoint methods (Reuther et al., 2001) in which the gradient, regarding an arbitrarily large number of parameters, can be calculated with roughly the same computational cost as two flow solutions. Once the gradient has been calculated, a descent method can be used to determine a shape change which will make an improvement in the design. Based on the idea of adjoint method, an optimum aerodynamic design technique is presented by Ying et al. (2011), which can be applied to the optimum problems with a large number of design variables. The key of this method lies in that the optimization process is regarded as an unsteady evolution, i.e., the optimization is executed, simultaneously with solving the unsteady flow governing equations and adjoint equations. During the unsteady evolution, the airfoil surface is moving with time, and the dynamic grid technique is introduced in the grid generation to improve the computational efficiency (Ying et al., 2011).

Nadarajah and Tatossian (2008) also have presented an adjoint method for the multi-objective aerodynamic shape optimization of unsteady viscous flows. Their method is beneficial when both the pitching angle and free stream Mach number are sinusoidally varied. They showed that the multi-objective cost function was able to preserve both the lift and pitching moment while simultaneously decreasing the overall drag (Nadarajah and Tatossian, 2008). Gradient-based techniques are efficient direct aerodynamic shape optimization (ASO) methods for both incompressible and compressible flows, and many of them use continuous adjoint methods. More recently, the introduction of surrogate-based optimization (SBO) methods to ASO have been successful in reducing the total computational cost. The overall objective of using SBO methods is to reduce the number of evaluations of the high-fidelity models, and thereby making the optimization process more efficient. Leifsson and Koziel (2010) introduced a computational design methodology which exploits surrogates constructed using low-fidelity flow analysis models and shape-preserving response prediction technique and demonstrated that their approach allows a rapid design improvement of airfoils at a very low computational cost corresponding to a few evaluations of the high-fidelity model (Leifsson and Koziel, 2010). Then they replaced the direct optimization of an accurate, high fidelity airfoil model by an iterative re-optimization of a corrected low-fidelity model (Leifsson et al., 2011). Their low-fidelity model is based on the same governing fluid flow equations as the high-fidelity one, but uses coarser discretization and related convergence criteria. With these techniques, they succeeded to improve the overall robustness of the optimization algorithm. ASO problems in compressible viscous flow were also performed using simultaneous pseudo-time stepping (Hazra et al., 2005). Hazra et al. (2005) could optimize airfoil surface by use of a preconditioner for convergence acceleration which stems from the reduced sequential quadratic programming (SQP) methods. This paper describes the implementation of adjoint technique for designing transonic airfoils with minimum compressible drag or pitching moment coefficients while preserving the initial lift coefficient. For this purpose, the Euler and adjoint equations are solved sequentially on a two-block structured grid about the section. The finite-volume scheme of Jameson et al. (1981) is

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients

used to discrete governing equations. This scheme employs a cell

397

NUMERICAL METHOD

centered discretization technique. A multistage time-stepping algorithm is used to advance the solution in time. A key feature in this scheme is applying the dissipative terms at the end of

In order to apply the finite-volume method, Eq. 3 is written in the integral form as:

each time step for solving the flow and adjoint equations. The magnitude of the dissipative terms is adapted to the local

(4)

properties of the flow by means of a Jameson sensor (Nadarajah and Jameson, 2001), based on the local pressure gradient. Acceleration techniques are applied to obtain faster steadystate convergence. These methods include local time stepping, variable coefficient implicit residual smoothing, and multi grid technique.

(5)

GOVERNING EQUATIONS Let (u,v) denote the velocity components in the Cartesian coordinate system (x,y). Compressible Euler equations can be formulated as:

(1)

Where:

(2)

Defining the flux vecto r as

The computational domain is meshed by a multi-block structured quadrilateral grid. Applying Eq. 4 to a typical (i,j) cell, for a cell center algorithm we obtain a system of ordinary differential equations as:

, where

are the unit vectors in the (x, y) coordinate system, Eq. 1 can be written as:

Where is the cell area and represents the net convectional fluxes out of each cell. The convective fluxes can be determined by a second order central scheme. However, the resultant scheme is not dissipative, and therefore, undamped oscillations at odd and even mesh points can be developed during the computation. To suppress the tendency for odd-even decoupling and to prevent the appearance of oscillations in regions containing severe pressure gradients near shock waves and stagnation points, the finite-volume scheme is modified by the addition of artificial dissipative terms as below:

(6)

Where denotes the dissipative fluxes. Jameson has established that an effective form of dissipative terms for flows with discontinuities is a blend of second and fourth differences with coefficients which depend on the local pressure gradient (Jameson et al., 1981). Equation 6 can be integrated in time by a multistage Runge-Kutta scheme.

(3)

BOUNDARY CONDITIONS Equation 3 can be solved by the finite-volume method described in the proceeding section.

Special care must be given to the boundary conditions when solving Euler equations on a multi-block grid. When using more

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

398

Farrokhfal, H. and Pishevar, A.R.

than a single block for a two-dimensional body, the attachment boundary condition should be used for the neighboring cells at the block interfaces. This can be done by considering a layer of ghost cells around each block. The boundary condition at the airfoil surface is the tangency condition of flow, that is:

Where |(.) denotes the partial derivative at the frozen quantity (.) and G denotes the flow or design variables. From Eq. 9, the Lagrange multiplier σ can be obtained as:

(10) (7) Where is the velocity vector and is the unit vector normal to the airfoil surface. Far-field boundary condition is also applied at the outer surface of the domain. The treatment of the far-field boundary condition is based on the concept of Riemann invariants for a one-dimensional flow normal to the boundary (Jameson and Baker, 1983).

Substituting into Eq. 9 yields:

(11)

Where:

The airfoil lift, drag and pitching moment coefficients can also be written as:

DERIVATION OF THE INVISCID ADJOINT TERMS The main objective of the present study is to achieve a modified airfoil which minimizes the inviscid compressibility drag without compromising the pitching moment under the constraint of maintaining the desired lift level. For this purpose, using the Lagrange multiplier, the cost function is defined as:

(12)

(8) Here and are the weight coefficients by which the airfoil shape can be modified for a particular application. When is chosen to be much larger than , a reduction in the drag coefficient is the main priority of the designer while using a larger leads to an airfoil with smaller pitching moment coefficient. In this equation indicates the desired value of . The variation of cost function in relation to the angle of attack and other parameters can be written as:

Where is the reference point for computing the pitching moment and and are the components of unit vector normal to the airfoil surface. Taking variation from the steady state flow equations in the computational domain leads to:

(13)

Where: (9) Multiplying Eq. 13 by a vector of co-state variables ψ and integrating over the domain, J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients

(14)

399

Where from the tangent boundary condition of flow on the airfoil surface, the variation of F2 becomes:

Finally, this equation can be integrated by parts to give: (19) (15)

Thus, the variation in the cost function can be written as:

If the coordinate transformation is such that δS is negligible in the far field, then the boundary condition of adjoint Eq. 17 can be obtained as:

(20)

Where:

. Finally we have:

(16)

(21) and n1=0 on boundary B. Eliminating the term which contains δw from the above integrals, the adjoint equation can be obtained as:

(17)

Because n 1=0 on airfoil surface, the only remaining component of Fk is:

And the gradients can then be defined with respect to the design variables xi as:

(18)

(22)

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

400

Farrokhfal, H. and Pishevar, A.R.

NUMERICAL METHOD FOR ADJOINT EQUATIONS The semi discretized adjoint equations of Eq. 17 is obtained using the same approach used for the flow equations:

(25) We replace the gradient by a smoothed value to guarantee the generation of a sequence of smooth shapes. When smoothing is performed in the x direction, the smoothed gradient may be calculated from a discrete approximation to:

(26) (23)

Introducing the artificial dissipation term to the Eq. 23 is crucially important to avoid discontinuities and to keep vector variable ψ differentiable. The five stage modified Runge-Kutta time stepping scheme is used to march the adjoint equations to the steady-state limit. Further details of the procedure of obtaining Eqs. 17 to 23 can be found in Jameson (1990, 1995).

OPTIMIZATION ALGORITHM A simple optimization algorithm can be established by using a line search method. These methods require an algorithm to choose a direction p and search along this direction from the current design variables to obtain a new set of variables for the cost function value. Once the direction is chosen, then a step length λ is multiplied to the search direction to advance the optimization to the next iteration. A simple optimization algorithm can be defined by setting the search direction l=-G, to the negative of the gradient at every iteration:

where ε is the smoothing parameter. Finding the new position of the boundary mesh points is also necessary, in order to estimate the variation of internal nodes accordingly. Jameson (1986, 1990) introduced a grid perturbation method that modifies the current location of the grid points based on perturbations at the geometry surface. This allows the variation of the grid point location in the equation for gradient evaluation, to be substituted with the variation of the surface points. The Optimization procedure can finally be summarized as follows: a. Solve the flow equations for ρ, u1, u2, p and e. b. Solve the adjoint equations for ψ subject to appropriate boundary conditions. c. Evaluate the gradients and calculate the corresponding smoothed gradient ; d. Update the shape based on the direction of steepest descent ; e. Solve the adjoint equations again for determining the gradients and consequently for updating angle of attack by using the equation:

or:

(24) (27) With a line search method, the step size λ is chosen so that the maximum reduction of the objective function I(X) is attained. The search procedure used in this work is a descent method, in which small steps are taken in a direction defined by the smoothed gradient. X represents the design variable, and G= the gradient. Instead of making the step:

where δXs is the perturbed coordinates of nodes on the airfoil surface, obtained in step (d); and f. Return to step (a) until convergence is reached.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients

RESULTS

401

0.01 NACA0012 Mach=0.8 AoA=0.0 deg. NACA0012 Mach=0.8 AoA=1.25 deg. NACA0012 Mach=0.63 AoA=2.0 deg. RAE 2822 Mach=0.67 AoA=1.06 deg.

Norm of Residual (-)

0.008

0.006

0.004

0.002

500

1000

1500

Iteration (-) Figure 1. Convergence History of solutions.

-1.5

NACA 0012 M = 0.63 AoA = 2.0 deg.

-1

Computation Ref. 16

-0.5

Cp (-)

In this section, we first validate our flow solver code by simulating several benchmark problems for the transonic inviscid flow over airfoils. For this purpose, transonic flow over NACA 0012 and RAE 2822 airfoils are considered. The validation is performed by comparing the obtained results to the previously published results by Pulliam et al. (1983) and Cook et al. (1979). In all cases, the far boundaries are located 5 chords away from the center of the airfoil. The NACA airfoil is set to three flow conditions: I) M∞=0.63, α=2.0°; II) M∞=0.8, α=0.0°; III) M∞=0.8, α=1.25° and the free stream Mach number and the angle of attack for the RAE airfoil is IV)M∞=0.67, α=1.06°. The Courant number number was taken as 4.0 and five orders of magnitude reduction in the density residual is considered as the convergence criteria. In all cases, the convergence criterion is met in less than 1,500 time steps, as shown in Fig. 1. The computational grid consists of 128 × 33 nodes stretched near the leading and trailing edges. The grid is also clustered towards the airfoil surface. Computations were performed on a personal computer with a dual core CPU for which the computational time for the aforementioned cases is less than 2 minutes. The pressure coefficients Cp on the upper and lower surface of the airfoils are shown in Figs. 2 to 5 and compared with the other computational results. In these Figures, X has been normalized and denotes the distance from the airfoil centre. The agreement between the Euler calculation and the other reference data is obvious. Figures 6 to 8 display the pressure contour plots for case III and IV. The smoothness of rapid expansion near the nose region on the upper surface is notable. The results prove that the region near stagnation point is calculated without any excessive numerical difficulty. This problem is very important for the shape optimization procedure and can deteriorate the smoothness of surface. In all cases, the shock location is also computed accurately with a resolution of 2 to 3 grid points. Case III is a more difficult one, where a strong shock forms on the upper surface and a weak shock on the lower one. The grid is clustered near the shock position on the upper-lower surfaces of the airfoil for this case. The performance of the optimization algorithm for the cost function is examined and the effects of weight coefficients k1 and k2 are explored. As our first test case, we set k1=1.0 and

0 0.5 1 1.5

-0.4

-0.2

0.2

X (Normalized distance)

0.4

Figure 2. Comparison of pressure distribution for the case I.

k2=0.0, therefore, reducing the problem to a pressure drag minimization one. NACA 0012 airfoil is considered to be the initial geometry and the flow variables are set to M∞=0.75 and angle of attack α=3.0°. The lift and drag coefficients for the initial geometry are calculated as Cl=0.656 and Cd=0.0345. For the initial geometry, a strong shock is developed on the upper surface of the airfoil. The vertical coordinates of the boundary grid points are considered as the design variables and the aim of optimization process is to reach a new airfoil geometry

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

Farrokhfal, H. and Pishevar, A.R.

402

for which the drag coefficient is minimum while the lift coefficient is maintained at the same value of Cl=0.656. An implicit smoothing technique is also used to smooth out the gradients before modifying the location of each point on the airfoil surface. The convergence is quickly attained by this optimizing technique and only after four design cycles the drag coefficient is reduced to one third of its initial value under the fixed lift coefficient constraint Cl = 0.656 . Comparisons of the pressure distribution for the new geometry in this figure and for the initial NACA 0012 indicate that the upper surface

-1.5

NACA 0012 M = 0.80 AoA = 0.0 deg.

-1

Ref. 16

Cp (-)

-0.5

0.5

1.5

-0.4

-0.2

0.2

X (Normalized distance)

1.5

0.4

Figure 3. Comparison of pressure distribution for the case II.

-0.2

0.2

0.4

X (Normalized distance)

0.6

Computation Ref. 16

-0.5 0

0.4

Y (Normalized distance)

NACA 0012 M = 0.80 AoA = 1.25 deg.

-1

Cp (-)

-0.4

Figure 5. Comparison of pressure distribution for the case IV.

-1.5

0.2 0

-0.2

0.5

-0.4

1 1.5

RAE 2822 M = 0.75 AoA = 1.0 deg. Computation Ref. 17

-1

Computation

-0.5

Cp (-)

nodes are modified in a way that the local Mach number and the strength of the shock are attenuated on this surface. This, in turn, may lead to a lower wave drag but decreases the lift coefficient as well. As a result, the fall in the lift coefficient is compensated by modifying the shape of the trailing edge, as shown in Fig. 9 to recover the pressure difference near this region. Figure 9 illustrates that the final design shaped is achieved after 50 designed cycles. The upper surface shock disappeared from the pressure distribution results and the drag coefficient is reduced to Cd=0.0448 , i.e., one eighth of its initial value.

-0.6 -0.4

-0.2

0.2

X (Normalized distance)

0.4

Figure 4. Comparison of pressure distribution for the case III.

-0.8

-0.6

-0.4

-0.2

0.2

0.4

X (Normalized distance) Figure 6. Pressure contour plots for the case III.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

0.6

Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients

As can be seen in this Figure, the optimal shape is no longer symmetric, and a cusped shape is developed at the trailing edge. The convergence history of the gradient norm and the drag coefficient are shown in Figs. 10 to 11. These Figures indicate that the convergence rate of the procedure is fast and that the optimization is accomplished mostly in the early cycles. It is also required to show that the final result depends on the initial geometry. In order to do so, we repeat the computations

for a five digit NACA 23012 airfoil as the initial geometry and change the angle of attack to reach the same lift coefficient as in the previous example. To achieve this goal, the angle of attack must be changed to α=1.24°, as shown in Fig. 12. The initial drag coefficient is also calculated as Cd = 0.0433. Figure 13 illustrates that the optimization procedure converges to different solution for the final geometry and the pressure distributions and, therefore, the optimizing procedure is dependent to the initial

-1.5

1 0.8

-1

0.6

-0.5

0.4

Cp (-)

Y (Normalized distance)

403

0.2 0

0 0.5

-0.2 1

-0.4 -0.6-1

-0.8 -0.6 -0.4 -0.2

0.2

0.4

0.6

X (Normalized distance) Figure 7. Pressure contour plots for the case IV.

Figure 9. Inviscid drag minimization of NACA 0012, M=0.75, After 50 design iterations, Cl=0.656, Cd=0.0448, α=1.45º.

-1.5 0.004 -1

0.035 0.03

-0.5

0.025

Cd (-)

Cp (-)

Convergence History Drag Coeficient

0.5

0.02

0.015 0.01

0.005 5

Cycle (-) Figure 8. Inviscid drag minimization of NACA 0012, N=0.75, Cl=0.656, Initial Solution, Cd=0.0345, Cm=-0.0335, α=3.00º.

Figure 10. Convergence history of drag minimization of NACA 0012 airfoil.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

Farrokhfal, H. and Pishevar, A.R.

404

conditions. We also notice that in this drag minimization problem, the pitching moment of the optimal shape is unfavorably increased from Cm=-0.0558 to Cm=0.154. As our second example, the other extreme case of k1=0.0 and k2=1.0 is considered, i.e., the pitching moment coefficient is minimized under the constant lift constraint and without restricting the drag coefficient. The initial geometry and flow conditions are as in the previous example. As shown in Fig. 14, there is a 94% reduction in the pitching moment for the

optimal shape, while the lift coefficient remains almost constant. The drag coefficient has also shown a 23% increase. To minimize the drag coefficient under constant lift and pitching moment coefficient, the weight constants are set to k1=0.50 and k2=0.50. Figure 15 indicates the obtained results for the same initial airfoil geometry and the flow conditions. The obtained results revealed that the drag coefficient is reduced by 76% while the lift and pitching moment coefficients are remained constant. This is the optimal shape for an airfoil that can reduce the compressible

-1.5 Convergence History Norm of Gradient

0.25

-1 -0.5

Cp (-)

Cd (-)

0.2

0.15

0 0.5 1

0.1 0

Cycle (-) Figure 13. Inviscid drag minimization of NACA 23012, M=0.75, Final Solution, Cl=0.660 Cd=0.0436, α=-0.17º.

-1.5

-1

-0.5

Cp (-)

Figure 11. Decrement of gradient norm of drag minimization of NACA 0012 airfoil.

0.5

Figure 12. Inviscid drag minimization of NACA 23012 M=0.75, Cl=0.660, Initial solution, Cd=0.0433, Cm=-0.0558, α=1.24º.

Figure 14. Pitching moment minimization of NACA 0012, M=0.75, After 36 design iterations, Cl=0.656, Cd=0.0424, Cm=-0.0002, α=3.71º.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

Optimization of Airfoils for Minimum Pitching Moment and Compressibility Drag Coefficients

-1.5 -1

Cp (-)

-0.5 0 0.5 1

Figure 15. Optimization of NACA 0012, k1=0.50, k2=0.50,M=0.75, After 32 design iterations, Cl=0.656, Cd=0.0059, Cm=-0.0338, α=2.65º.

pressure drag at a given required lift without compromising the pitching moment coefficient. The test cases above show that the proposed cost function for the ASO is very efficient for designing airfoil with specific requirements in subsonic and transonic flows.

CONCLUSION An efficient methodology for ASO of airfoils in transonic or high subsonic regime was described. In this approach, the Euler equations are solved for a symmetric and asymmetric airfoil with a multi-block method. The results obtained from this approach for the pressure distribution on the airfoil surface were

405

validated and it was shown that the aerodynamic coefficients are estimated with a rather good accuracy. Then the prepared solver is employed to optimize the airfoils based on the adjoint method for minimum pressure drag, while maintaining the lift and pitching moment coefficients. Then, a new cost function was introduced for the problem of reducing the pitching moment and drag coefficients under the constant lift constraint. The procedure was first tested for the extreme case of reducing pitching moment under the constant lift constraint without restricting the drag coefficient which was led to a final shape with higher wave drag coefficient. However, by proper selection of cost function weight coefficients, it is possible to reduce the wave drag under constant pitching moment and lift coefficients. In this case, the final solution is an airfoil with a sharp trailing edge and a narrow leading edge. Drag minimization of airfoils, in the case of constant lift coefficient, has led to a cusped shape of trailing edge and relatively flat upper surface in both symmetric and cambered airfoils; but the final solution is different for two cases, especially for the shape of leading edge of the airfoils. These results indicate that the optimization procedure based on the adjoint method depends on the initial shape of airfoil. In other words, this technique can converge to a local extremum and not a global one. The main advantage of this method is that satisfactory results can be achieved by minimum computational efforts, particularly for the grid generation problem and the required central processing unit of computer (CPU) time. This feature becomes important where a low fidelity accurate case such as rotor blade flow solver is required for primarily design or optimization purposes. However, the proposed method ignores important phenomena such as the viscous flow or shock boundary layer interactions.

REFERENCES Anderson, W.K., Newman, J.C., Whitfield, D.L. and Nielsen, E.J., 2001, “Sensitivity analysis for Navier-Stokes equations on unstructured meshes using complex variables”, AIAA Journal, Vol. 39, No. 1, pp. 56-63. doi:10.2514/2.1270.

Jameson, A., 1986, “Multigrid algorithms for compressible flow calculations”, Proceedings of the 2nd European Conference on Multigrid Methods II, Lecture Notes in Mathematics, No.1228, pp. 166–201.

Barger, R.L., 1975, “Procedures for the design of low-pitchingmoment airfoils”, NASA TN D-7982.

Jameson, A., 1988, “Aerodynamic design via control theory”, Journal of Scientific Computing, Vol. 3, No. 3, pp. 233-260.

Cook, P.H., Mc Donald, M.A. and Firmin, M.C.P., 1979, “Aerofoil RAE 2822 pressure distributions, and boundary layer and wake measurement”, AGARD Advisory Report No. 138.

Jameson, A., 1990, “Automatic design of transonic airfoils to reduce the shock induced pressure drag”, In Proceedings of the 31st Israel Annual Conference on Aviation and Aeronautics, pp. 5-17.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

406

Farrokhfal, H. and Pishevar, A.R.

Jameson, A., 1995, “Optimum aerodynamic design using CFD and control theory”, AIAA Paper 95-1729-CP. Jameson, A. and Baker, T.J., 1983, “Solution of Euler equations for complex configurations”, AIAA, pp. 83-1929. Jameson, A., Schmidt, W. and Turkel, E., 1981, “Numerical solution of the Euler equations by finite volume methods using Runge-Kutta time-stepping schemes”, AIAA, pp. 81-1259. Hazra, S.B., Schulz, V., Brezillon, J. and , Gauger, N., 2005, “Aerodynamic shape optimization using simultaneous pseudotimestepping”, Journal of Computational Physics, Vol. 204, No. 1, pp. 46-64. doi:10.1016/j.jcp.2004.10.007. Leifsson, L., Koziel, S. and Ogurtsov, S., 2011, “Inverse design of transonic airfoils using variable-resolution modeling and pressure distribution alignment”, Procedia Computer Science, Vol. 4, pp. 12341243. doi:10.1016/j.procs. 2011.04.133.Leifsson, L. and Koziel, S., 2010, “Multi-fidelity design optimization of transonic airfoils using shape-preserving response prediction”, Procedia Computer Science, Vol.1, No. 1, pp. 13111320. doi:10.1016/j.procs.2010.04.146. Nadarajah, S. and Tatossian, C., 2008, “Multi-objective aerodynamic shape optimization for unsteady viscous flows”, Journal of Optimization and Engineering, Vol. 11, No. 1, pp. 67-106. doi: 10.1007/s11081008-9036-4.

Nadarajah, S. and Jameson, A., 2001, “Studies of the continuous and discrete adjoint approaches to viscous automatic aerodynamic shape optimization”, AIAA 2001-2530, AIAA 15th. Computational Fluid Dynamic Conference, Anaheim, Ca, June 11-14. Pulliam, T.H., Jespersen, D.C. and Childs, R.E., 1983, “An enhanced version of an implicit code for the Euler equations”, AIAA-83-0344. January 10-13, Reno, Nevada. Reuther, J.J., Jameson, A., Alonso, J.J., Rimlinger, M.J. and Saunders, D., 2001, “Constrained multipoint aerodynamic shape optimization using an adjoint formulation and parallel computers”, parts 1 and 2, Journal of Aircraft, Vol. 36, No. 1, pp. 51-74. Reuther, J. and Jameson, A., 1994, “Control based airfoil design using the Euler equations”, AIAA paper 94-4272-CP. Xie, L., 2002, “Gradient-based optimum aerodynamic design using adjoint methods”. Ying, G.Y., Feng, H. and Yu, S.M., 2011, “Aerodynamic airfoil design using the Euler equations based on the dynamic evolution method and the control theory”, Science China Physics, Mechanics and Astronomy, Vol. 54, No. 4, pp. 697-702. doi:10.1007/s11433-011-4287-z. Yu, L., Quan, C.X., Neng, L.Z. and Wu, X.J., 2010, “Multi-objective robust airfoil optimization based on non-uniform rational B-spline (NURBS) representation”, Science China Physics, Technological Sciences, Vol. 53, No. 10, pp. 2708-2717. doi: 10.1007/s11431010-4075-4.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.395-406, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.373

On Using Quasi-Newton Algorithms of the Broyden Class for Model-to-Test Correlation Jan Klement1

ABSTRACT: The correlation of a model with test results is a common task in engineering. Often genetic algorithms or adaptive particle swarm algorithms are used for this task. In this paper another approach is presented using two quasiNewton algorithms of the class defined by Broyden. A study was conducted with thermal models showing the performance of this approach. Comparing the results to other studies reveals that the approach reduces the number of iterations by several orders of magnitude; typical calculation times for model correlation times are reduced from the order of weeks and months to the order of hours and days. KEYWORDS: Thermal vacuum test, Thermal analysis, Correlation, Broyden, ESATAN, Thermica.

INTRODUCTION In every calculation, there are always discrepancies between the predictions of a mathematical model and their physical measurements. Refinement of the models can reduce the discrepancies, however, the exact values of many parameters are often unknown. Without perfect knowledge of the values of these parameters, a model refinement will never lead exactly to the measured results. Determining the exact values of all parameters individually is too costly or sometimes even impossible. Testing is often only possible at the system level, and extracting parameter values from the test results is a difficult task which is done by correlating the mathematical model with the measurement data. This means that the model parameters are changed via feedback from the measurement results, so that the discrepancies between the measurements and the models are minimized. Many methods have been developed and analyzed to perform model-to-measurement correlation (Jouffroy, 2007; De Palo et al., 2011; Momayez et al., 2009; Harvatine and De Mauro, 1994; Roscher, 2006; van Zijl, 2013; WenLong et al., 2011; Mareschi et al., 2005). Most methods are based on stochastic optimization algorithms and often require several hundred iterations to converge. In this paper, an approach is presented using Broyden’s class of methods (Broyden, 1965), which use considerably less iterations for the majority of cases.

1.Tesat-Spacecom GmbH & Co – Backnang – Germany Author for correspondence: Jan Klement | KG, Gerberstraße 49 | D-71522 Backnang – Germany | Email: Jan.Klement@tesat.de Received: 05/26/2014 | Accepted: 10/06/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

408

Klement, J.

THE THERMAL MODEL CORRELATION PROBLEM The method will be presented with a typical thermal model correlation as used in the space industry, but the procedure can also be applied to many other models as used in other industries and areas. Thermal models calculate temperatures at certain points depending on a set of parameters. Mathematically expressed, this means: tmdl = F(p),

(1)

where: F is the mathematical model with function F:Rk→Rm (k is the number of parameters, m is the number of results); tmdl is a vector containing the temperatures at the selected points; and p is a vector containing the parameters of the model, for example thermal conductivity, thermal capacity, thermal emissivity, and solar absorptivity. During the tests, the temperatures are measured at the same points as described in the model t mdl. The aim of the correlation is to find the set of parameters p corr for which the root sum square of the differences between the measured temperatures tmes and the calculated temperatures F(pcorr) is minimal. ||F(Pcorr) – tmes|| = min{||F(p) – tmes|||p ∈ P}

(2)

The advantages of this method are that a whole vector of information, in this case temperature differences, can be evaluated instead of a single scalar and that a set of simple functions are used instead of a complex function. One of the best methods to solve this type of problem is the multidimensional Newton method, which applies the following formula iteratively: pn+1 = pn – J(pn)-1 F(pn) (4) where: pn is the vector of parameters; J(pn) is the Jacobian matrix at pn A limitation of this method is that the Jacobian matrix can often only be calculated numerically, with the consequence that, for each individual parameter in pn, a separate calculation is needed. Fortunately, Broyden developed a class of methods to approximate the Jacobian matrix based on the results from the previous step. The methods are described in Broyden (1965) and therefore will not be explained here. Broyden’s first method (the “good Broyden method”) is chosen for the algorithm used in this paper as it performed better for Broyden’s practical problems. In addition, a self-developed method of the Broyden class is tested. This method updates each element bnij of the approximated Jacobian Matrix Bn with the following formula: ij bn+1 = (1+ kni bnij snj) bnij (5)

where: sn: = pn – pn-1 (6)

where P is the solution space for the parameter vector.

yex,n: = Bnsn (7)

THE METHOD

y is,n: = F(pn) – F(pn-1) (8)

In contrast to most correlation methods developed (Jouffroy, 2007; De Palo et al., 2011; Momayez et al., 2009; Harvatine and DeMauro, 1994; Roscher, 2006; van Zijl, 2013; WenLong et al., 2011; Mareschi et al., 2005), this method will not attempt to minimize the length of the vector ||F(p) – tmes||, but searches a root of the equation system:

ki =

F(pcorr) – (tmes) = 0.

(3)

yiis,n – yiex,n m

Σj=1 (bnij snj)2

(9)

CONSTRAINTS FOR THE METHOD Every optimization method has requirements which need to be fulfilled in order to be used effectively. Within

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

On Using Quasi-Newton Algorithms of the Broyden Class for Model-to-Test Correlation

this section, some of the more important requirements for practical model-to-test correlation, and the main practical consequences, are described. The requirements are important when choosing to use this method and also for the selection of the parameters and results to be used. MONOTONY AND DIFFERENTIABILITY An important restriction in the use of this method is that each result of the model is a differentiable monotone function of the parameters within the solution space. Most thermal models are monotone and differentiable functions of the parameters, but there are a few exceptions such as controlled heating systems or when the parameter is, for example, the orbit position of a spacecraft. For small changes the model is neither monotone nor differentiable due to numerical noise. If the function does not fulfill this requirement, within the solution space, the algorithm may be prevented from converging. OBSERVABILITY OF THE PARAMETERS Each of the parameters must have an effect on at least one result. This is not a specific issue to this particular method, but it is a general requirement for correlation. Although the requirement appears trivial, it is often the reason why the algorithm is unable to converge. For the presented method, a low degree of observability combined with numerical noise can lead to instability. Often, the relevance of parameters can be deduced through an understanding of the model, but for complex systems it is sometimes difficult to estimate the relevance of an individual parameter. Fortunately, the Jacobian matrix addresses some of these limitations. If a column of the Jacobian matrix consists solely of zeros, the corresponding parameter cannot be observed. In practical applications there are often many values close to zero, mainly due to small effects or numerical noise. If the relevance of the parameter is smaller than the accuracy of the measurement data, the parameter can then be considered as not observable. INFLUENCE OF THE PARAMETERS ON THE RESULTS Each result has to be influenced by at least one parameter. This is also not a general requirement for correlation, but it prevents model correlation. For the presented method,

409

numerical noise combined with a low level of influence on a result can lead to instability. Again, the Jacobian matrix is a good tool for estimating the influence of a parameter on results. If a row of the Jacobian matrix has only zeros or very small values, the corresponding result cannot be influenced. ACCURACY OF THE MEASUREMENT AND MODEL INACCURACIES Due to measurement errors and model inaccuracies, it is unlikely that the result vector will converge towards zero. If the measurement error is larger than the effect of a parameter, then the relevant parameter may not converge to a reasonable result. Model inaccuracies may, in addition, make it impossible to correlate it within the parameter space. CONSTRAINTS In most instances, each parameter has certain limits defined by the physics of the model. For example, thermal conductivity will never be zero or below and never exceed a certain value. As the method defined by Broyden is for unconstrained problems, the following algorithm is used: • Calculate pn+1 without considering the constraints; • If a parameter of pn+1 is outside the defined boundaries, scale the change of the parameters so that all parameters are within the boundaries; • If some parameters have already reached their constraints, calculate pn+1 for the condition that these parameters are fixed at their constraints.

LOAD CASES ANALYZED Two load cases were analyzed: a constructed simple model and a complex model of real piece of hardware. These models where build in Systema/THERMICA, a thermal analysis software specialized to simulate spacecraft (mostly compatible to ESATAN and similar to SINDA). The standard syntax for these models was used. THE CONSTRUCTED SIMPLE MODEL The thermal model is a steady state model consisting of four diffusive nodes and an environment node at 0°C. Each node

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

410

Klement, J.

has a surface of 0.1 m² with an emissivity of 1, which radiates to the environment. 10 W are dissipated in each node. Node 1 is connected to the environment node with a thermal conductance of 1 W/K. Nodes 1 to 4 are all connected to each other by 6 thermal linear links (GL) as shown in Fig. 1. In order for the temperatures at the nodes to correspond to the measured values, the parameter values of the thermal links need to be changed. Therefore, the equation system to be solved by the algorithm is:

parameters where chosen; they consist of thermal boundary conditions, thermal conductances, and emissivities for the infrared spectrum. Temperature readings were taken from 26 sensors at 6 test points over time, yielding 156 evaluated temperatures.

Tcalculated(GL) – Tmeasured = 0,

RESULTS OF THE UNDERDETERMINED SIMPLE MODEL Figure 2 shows the development of the root sum square of all temperature differences for both analyzed algorithms. The values of some significant points shown in this diagram are listed in Table 1. The model quickly converged toward 0, with both algorithms

where: Tcalculated is the function which represents the model; GL is the vector containing GL1 to GL6; and T measured is the vector of the calculated temperatures for GLi = (0.1 + i/100)W/K. This is assumed to be a measurement. The initial estimates for all GL are 0.5 W/K. Three load cases are derived from this problem: • An underdetermined system: All links are variable. There are six parameters (links) for four values (temperatures); • A determined system: Links 3 and 6 are constant (GL3 = 0.13W/K, GL6 = 0.16W/K). There are four parameters (links) for four values (temperatures); • An overdetermined system: Links 3, 5, and 6 are constant (GL3 = 0.13W/K, GL5 = 0.6W/K, GL6 = 0.16W/K). There are three parameters (links) for four values (temperatures), and GL5 is chosen so that it is impossible to solve the system. THE COMPLEX MODEL The complex model is a model of real spacecraft hardware, which was tested in a thermo vacuum chamber. In total, 13

1.00E+01

Broy den

1.00E+00

self -dev eloped method

RSS [K]

(10)

RESULTS

1.00E-01 1.00E-02 1.00E-03 1.00E-04 1.00E-05

Figure 2. Root sum square of the temperature differences over iterations of the underdetermined simple model. Table 1. Selected points of the algorithm convergence of the underdetermined simple model.

Algorithm 10W 0°C 0.1m² 0°C

1W/K

1 10W

GL4

GL2

GL3

Iterations for Jacobian matrix generation

0°C

2 GL5

GL1

0.1m²

0 1 2 3 4 5 6 7 8 9 10 11 12 1314

Iterations after Jacobian matrix generation

GL6

0.1m²

0°C

Broyden

10W

0.1m²

10W

0°C

Selfdeveloped method

Figure 1. Thermal test model schematics. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

Nº of iterations after Jacobian matrix generation

RSS/K

0 (initial)

4.261

5.921E-03

2.646E-05

0 (initial)

4.261

7.652E-04

5.385E-05

On Using Quasi-Newton Algorithms of the Broyden Class for Model-to-Test Correlation

reducing the RSS by approximately one order of magnitude every two to three iterations, until the numerical accuracy limit of 1E-5 was reached. After the 4th iteration, the self-developed algorithm was faster, but after the 9th iteration, the Broyden algorithm converged faster. In conclusion, the performance of both algorithms can be considered comparable for this case. RESULTS OF THE DETERMINED SIMPLE MODEL In the same way as shown for the underdetermined simple model, Fig. 3 and Table 2 show the root sum square development for the determined simple model. The model converged slower than for the underdetermined problem.

The Broyden algorithm needed an average of 11 steps per order of magnitude while the self-developed algorithm needed an average of only 5 of them. RESULTS OF THE OVERDETERMINED SIMPLE MODEL The overdetermined model converged, as expected, not to 0 but to a finite value (0.0375K), as is shown in Fig. 4 and Table 3. For this load case, the self-developed method converged considerably faster, with approximately 4 iterations per order of magnitude, while the Broyden algorithm used approximately 8 iterations per order of magnitude.

1.00E+01

Broyden

1.00E+01

411

Broyden Self-developed method

Self-developed method 1.00E+00 1.00E+00

RSS[K]

1.00E-01 1.00E-02

1.00E-01

1.00E-03 1.00E-04 1.00E-05

1.00E-02

1 5 7 9 13 17 21 25 29 33 37 41 45 49 53

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Iterations after Jacobian matrix generation

Iterations after Jacobian matrix generation Figure 3. Root sum square of the temperature differences over iterations of the determined simple model.

Figure 4. Root sum square of the temperature differences over iterations of the overdetermined simple model.

Table 2. Selected points of the algorithm convergence of the determined simple model.

Table 3. Selected points of the algorithm convergence of the overdetermined simple model.

Algorithm

Broyden

Selfdeveloped method

Iterations for Jacobian matrix generation

Nº iterations after Jacobian matrix generation

RSS/K

0 (initial)

3.442

1.024

9.487E-05

0 (initial)

3.442

5.813E-02

1.414E-05

Algorithm

Broyden

Selfdeveloped method

Iterations for Jacobian matrix generation

Nº iterations after Jacobian matrix generation

RSS/K

0 (initial)

3.442

1.869

3.760E-02

0 (initial)

3.442

1.869

3.757E-02

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

Klement, J.

RESULTS OF THE COMPLEX MODEL Figure 5 and Table 4 show the convergence of the algorithm for the complex model in a different way to the way that was shown for the simple model. The complex model is an overdetermined system and it is not expected to converge to 0. Therefore, the difference between the RSS of each iteration and the best RSS reached is used. This difference is standardized to the initial value. The best RSS was reached by restarting the algorithm near the optimum. As can be seen in the following figure, the optimum is nearly reached after a few iterations. In this steep initial part, both algorithms required approximately 2 iterations, on average, to reduce the target value by one order of magnitude. After reaching 0.25%, the Broyden algorithm stayed stable while the new algorithm converged slowly to a lower value.

100.00%

(RSS-RSSmin)/(RSSinitial-RSSmin)

412

Broyden Self-developed method

10.00%

1.00%

0.10%

0.01%

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Iterations after Jacobian matrix generation Figure 5. relative root sum square of the temperature differences over iterations of the complex model.

DISCUSSION OF THE RESULTS Both algorithms converged in all of the studied cases until either the numerical accuracy limit or a limit due to an over-constrained system was reached. The number of iterations can be approximated by the formula:

Table 4. Selected points of the algorithm convergence of the complex model.

Algorithm

Iterations for Jacobian matrix generation

m = 1 + k + r * c (11) where: m is the total number of iterations including the first; k is the number of parameters (iterations of Jacobian matrix generation); and r is a value specific to the problem; it depends mainly on the nonlinearity of the model and the interactions between the parameters. The total number of parameters is not very significant for this value. For the load cases studied, the value was between 2 and 10, on average, until a certain limit was reached. For similar load cases, it is not expected to increase significantly; c is the desired reduction in the order of magnitude of the difference of the root sum square and the minimal possible root sum square. It is defined by the following formula: c = log10(R0 – Rmin) – log10(Rfinal – Rmin) (12) where: R 0 is the initial root sum square of the deviation vector (F(po) – tmes);

Broyden

Selfdeveloped method

Nº iterations after Jacobian matrix generation

(RSSRSSmin) / (RSSinitialRSSmin)

0 (initial)

100.00%

0.81%

0.22%

0 (initial)

100.00%

0.19%

0.06%

R final is the final root sum square of the deviation vector (F(pfinal) – tmes); and Rmin is the minimum possible root sum square of the deviation vector (F(pfinal) – tmes). As can be seen in Fig. 6 the number of iterations needed for the presented algorithms (10 to 30) to the typical number of iterations needed for genetic (25600 (van Zijl, 2013), 843 to 33072 (Jouffroy, 2007)), adaptive particle swarm optimization (6000 (van Zijl, 2013)) or the stochastic design improvement (555, (Mareschi et al., 2005)), this approach is much faster. The number of parameters of the models used in these evaluations is comparable to the complex model used in this paper, but the

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

On Using Quasi-Newton Algorithms of the Broyden Class for Model-to-Test Correlation

413

Broyden

Complex model (2062 n, 13 p)

Self-developed method Stochastic methods

Underdet.simple model (5 n, 6 p) Det. simple model (5 n, 4 p) Overdet. simple model (5 n, 3 p) Jouffroy – GA (100 n, 5 p) Jouffroy – GA (100 n, 5 p) Jouffroy – GA (100 n, 21 p) van Zijl – GA (129 n, 16 p) van Zijl – APSO (129 n, 16 p) Mareschi, Perotto and Matteo SDI (111 n, 12 p) 1

100

1000

10000

100000

Nº of interations Figure 6. Comparison of the number of iterations needed for thermal model correlation (n=nodes of the model, p=parameters correlated).

number of nodes used is considerably lower. Unfortunately, the models and tools used for these methods are not accessible so that a direct comparison was not possible. For each iteration, the CPU time required for generating a new set of parameters is only a few milliseconds and, therefore, it is negligible compared to the minutes or hours required for solving the model. It can be concluded that the total CPU time is nearly proportional to the number of iterations, independent from the algorithm. Both algorithms studied showed similar performance. Therefore, both are considered suitable for model correlation. The new algorithm showed better performance for most situations.

CONCLUSION It has been shown that using root finding algorithms of the Broyden class for thermal mathematical model-to-test correlation is feasible and efficient. Although a direct comparison of the analyzed methods and a genetic algorithm was not possible,

the number of iterations needed for the tested algorithms is 20 to 1,000 times smaller than the number of iterations reported for genetic or adaptive particle swarm algorithms. In conclusion, this algorithm significantly reduces the cost of a thermal model correlation. The “good Broyden method” and the self-developed method delivered satisfactory results and are, therefore, both considered suitable. The self-developed method delivered slightly better results for most of the cases analyzed.

ACKNOWLEDGMENTS The laser communication terminal (LCT) project, which this publication is based on, is carried out on behalf of the Space Administration of the German Aerospace Center (DLR e.V.) with funds from the Federal Ministry of Economics and Technology, under the reference number 50-YH1330. The author is responsible for the contents of this publication and gratefully acknowledges their support.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

414

Klement, J.

REFERENCES Broyden, C.G., 1965, “A Class of Methods for Solving Nonlinear Simultaneous Equations”, AMS Mathematica of Computation, Vol. 19, No. 92, pp. 577-593. doi: 10.1090/S0025-5718-19650198670-6.

Momayez, L., Dupont, P., Popescu, B., Lottin, O. and Peerhossaini,

De Palo, S., Malost, T. and Filiddani, G., 2011, “Thermal Correlation of BepiColombo MOSIF 10 Solar Constants Simulation Test”, 25th European Workshop on Thermal and ECLS Software.

applthermaleng.2009.05.025.

Harvatine, F.J. and DeMauro, F., 1994, “Thermal Model Correlation Using Design Sensitivity and Optimization Techniques”, 24th International Conference on Environmental Systems and 5th European Symposium on Space Environmental Control Systems.

& EADS Astrium.

Jouffroy, F., 2007, “Thermal model correlation using Genetic Algorithms”, 21st European Workshop on Thermal and ECLS Software. Mareschi, V., Perotto, V. and Gorlani, M., 2005, “Thermal Test Correlation with Stochastic Technique”, 35th International Conference on Environmental Systems (ICES).

H., 2009, “Genetic algorithm based correlations for heat transfer calculation on concave surfaces”, Applied Thermal Engineering, Vol.

29,

No.

17-18,

pp.

3476-3481.

doi:

10.1016/j.

Roscher, M., 2006, “Genetische Methoden zur Optimierung von Satelliten Thermalmodellen bei EADS ASTRIUM“, Hochschule Wismar

van Zijl, N., 2013, “Correlating thermal balance test results with a thermal mathematical model using evolutionary algorithms”, Faculty of Aerospace Engineering, Delft University of Tecnology. WenLong, C., Na, L., Zhi, L., Qi, Z., AiMing, W., ZhiMin, Z. and ZongBo, H., 2011, “Application study of a correction method for a spacecraft thermal model with a Monte-Carlo hybrid algorithm”, Chinese Science Bulletin, Vol. 56, No. 13, pp. 1407-1412. doi: 10.1007/s11434-010-4053-z.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.407-414, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.339

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover Luca De Filippis1, Enrico Gaia2, Giorgio Guglieri1, Marco Re1, Claudia Ricco3

ABSTRACT: Human Machine Interface (HMI) design is a critical field of work because no general guidelines or rules have been assessed. In order to aid practitioners to design effective HMIs, different methodologies have been studied. To understand task objectives and plan goal-oriented actions, human operators exploit specific cognitive processes that have to be supported with advanced interfaces. Including cognitive aspects in HMI design allows generating an information flow that reduces user mental workload, increasing his/her situation awareness. This paper focuses on design and test of a Graphical User Interface (GUI) for the telenavigation of a space rover that makes the cognitive process of the user a priority in relation to the other development guidelines. To achieve this, a Cognitive Task Analysis (CTA) technique, known as Applied Cognitive Work Analysis (ACWA), is combined with a multi-agent empirical test to ensure the GUI effectiveness. The ACWA allows evaluating mission scenarios, i.e. piloting the rover on the Mars surface, in order to obtain a model of the human cognitive demands that arise in these complex work domains. These demands can be used to obtain an effective information flow between the GUI and the operator. The multi-agent empirical test, on the other hand, allows an early feedback on the user mental workload aiming to validate the GUI. The result of the methodology is a GUI that eases the information flow through the interface, enhancing operator’s performance. KEYWORDS: Cognitive engineering, Human machine interface, Space robotics, Space exploration, Graphical User Interface.

INTRODUCTION In today’s world, many systems are remotely operated or supervised by individuals who are decision makers. The planning, monitoring and controling of many of these systems are supported via visual display units. Applications that heavily depend on teleoperation and telenavigation are space exploration missions. Complex and dangerous conditions — deriving from environmental hazards — may occur to humans involved in these missions usually performed with supervised systems. As an example, Biesiadecki et al. (2007) state that “successful operation of the Mars Exploration Rover (MER) vehicles has depended on both manuallydirected and autonomous driving. The two methods are complementary and careful selection of the right technique leads to better overall performance”. The Graphical User Interface (GUI) design becomes a critical task in mission accomplishment, enhancing operator’s Situation Awareness (SA) and control capabilities. Existing design approaches cover only specific aspects of HMI design (Störrle, 2010), bringing to applicationoriented solutions which limit the development of complex architectures. Many approaches and tools to tackle individual problems in the interface design have been developed by Hashimoto et al. (2011), but any integrated solution addressing the whole design process has been defined. A big effort has been spent to design and implement displays and interfaces to operate rovers in space: as NASA Visual Environment for Remote and Virtual Exploration (VERVE), Predictive and Interactive Graphical Interface (PIGI) (Pedersen et al., 2010, 2012; Burridge and Hambuchen, 2009) or CliffBot Maestro (Norris et al., 2009).

1.Politecnico di Torino – Torino – Italy 2.Thales Alenia Space Italia S.p.A – Torino – Italy 3.Università degli Studi di Torino – Torino – Italy Author for correspondence: Giorgio Guglieri | Department of Mechanical and Aerospace Engineering, Politecnico di Torino | C.so Duca degli Abruzzi 24, Torino – 10129 – Italy | Email: giorgio.guglieri@polito.it Received: 02/19/2014 | Accepted: 09/26/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

416

•

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

Assessed modeling approaches are: Operator Function Model (OFM) (Chu et al., 1995), used in supervising, monitoring and controling of Unmanned Combat Aerial Vehicles (UCAVs) (Narayanan et al., 2000); Direct Manipulation Interfaces (DMI) proposed by Hutchins et al. (1985) with the aim of reducing gaps between the user’s goals and their knowledge of the system; Ecological Interface Design (EID) (Rasmussen and Vicente, 1989), a theoretical framework for designing interfaces in complex human-machine systems based on skills, rules, knowledge (SRK) taxonomy (Rasmussen, 1983) and the abstraction hierarchy (AH); Goals, Operators, Methods, and Selection Rules (GOMS) (Card et al., 1983) a formal predictive modeling technique for interface design based on cognitive problem solving behavior (Eberts, 1994); Cognitive Task Analysis (CTA) (Gordon and Gill, 1997).

Direct correlation between user’s SA and overall performance of the supervised system (Endsley, 2003) make the information flow through the GUI an essential feature in the designing process (Baxter, 2013). This paper proposes an integrated methodology in which SA requirements form the basis of the design phase. This integrated solution is achieved using a CTA approach — Applied Cognitive Work Analysis (ACWA) proposed by Elm (2003)—, coupled with direct testing methodology which involves user’s performance assessment (Endsley, 2000) as a feedback on design. This leads to the definition of an integrated approach able to obtain graphical interfaces for space applications that can ensure high standards in terms of users’ performance and SA. CTA methodology can capture expert’s knowledge in managing complex, dynamic and changing environments (Mast, 2014). Redding (1992) defines CTA “as an approach in determining the mental processes and skills required in performing a task and the changes that occur as the skill develops”. Many of the above-mentioned approaches cannot lead to an explicit representation of the operator goals (Okura et al., 2013) which is one of the major limitations of the OFM, or they are not effective in complex human-machine systems (Corujeira, 2013). The paper introduces a methodology to design GUI with small, manageable, engineering transformations, each requiring the skilled application of the methodology’s principles rather than requiring a design revelation at any point in the process. The design progress, therefore, occurs by generating artifacts

that capture the results of each of these intermediate stages. Each of these artifacts also provides an opportunity to evaluate completeness and quality of the analysis and design effort. The design process is then associated with an empirical test (Nielsen, 1994), designed to obtain information about user performance and SA. A comparative analysis is conducted introducing support of experts, in order to extract information about the user’s SA and Mental Workload (MW), obtaining a complete methodology that allows the practitioner to be supported from the design phase to tests and analysis. The remainder of this paper is organized as follows: next section presents an in-depth analysis of the ACWA methodology and covers all the design phases in which it is divided. Further sections describe the testing procedure to obtain all the information needed to correctly assess the GUI. Subsequently, the results of the testing phase is evaluated and further analysis on user perception and performance is made. The paper concludes with a discussion on the overall design approach.

DESIGN METHODOLOGY In order to model the cognitive process used to accomplish mission tasks and design a Human Machine Interface (HMI) able to support the user to reach mission goals, the ACWA methodology has been exploited. This approach is subdivided in four design processes and it starts with Functional Abstraction Network (FAN) definition to model the functional process required to perform the goal. The FAN captures the essential domain concept and the relationships between the problem-space and the domain practitioners. The next step is overlaying the Cognitive Work Requirements (CWRs) on the functional model as a way of identifying the cognitive demands which arise in the domain and require support. These cognitive demands may be successfully executed after identifying the Information and Relationship Requirements (IRRs). IRRs definition supports Representation Design Requirements (RDRs) which define the “shaping” of the information. The last step of the CTA approach used in this paper is the developing of Presentation Design Concepts (PDCs) to implement these RDRs, producing the correct information transfer to the users in order to fulfill their cognitive demands.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

An in depth analysis of each step of the methodology is made in the remainder of this chapter. However, it must be stressed that all the steps of the process are “open” in parallel, so as design thoughts for CWRs, IRRs, or PDCs occur, they can be recorded and aid in the definition of the FAN itself. Figure 1 represents the complete methodology flowchart, where ACWA steps are related to mockup generation and testing in a recursive process. FUNCTIONAL ABSTRACTION NETWORK FAN is a function-based goal-means decomposition of the domain. This step has its roots in the formal, analytic goal-means decomposition method pioneered by Rasmussen (1986) for representing cognitive work domains as an AH. The FAN is a structured representation of the functional concepts and their relationship used as the context for the information system to be designed. This produces a multilevel recursive means-ends representation of the work domain structure. In practice, building a FAN is an iterative process: the FAN starts from an initial base of knowledge regarding the domain that is gradually expanded using complementary techniques as observations, interviews or training (Elm et al., 2003). In order to obtain an artifact that reflects the process depiction without representing only its physical components, a “flow modeling” approach is introduced. This model is based on the definition of different shapes to represent different stages of the process: Sources, Sinks, Storage, Transport and Conversion shapes are used in order to represent the functional operation of both

abstract and relatively physical processes (Fig. 2). Figure 3 represents the primary goal formulated in the application here described, to start the development of the FAN. Then, complementary techniques, as face-to-face interview or verbal protocol techniques are used in order to expand and enrich the domain understanding and to evolve a functionbased model. Figure 4 shows the evolution of the FAN from goal 1 to goal 2. It expands the concept of map updating that is necessary to obtain a new map from the old one, merging new environment data obtained during the field-mapping process. The FAN obtained in this paper starts from a field mapping overall mission objective: this main goal is subsequently expanded through the definition of system deployment functions and the analysis of rover motion functionality. At the very low level of abstraction, the FAN leads to define the functions of gathering

Goal Desired objective of the process function Commodity that moves through the process Conversion/ Manipulation

Sources

Evaluation

Cognitive work Requirements

Testing Phases

Information and Relationship Requirements

Optimized GUI

Outcome

Figure 2. FAN Template.

Field G0 :Mapping Comodity that moves through the process

Old Map Functional Abstraction Network

417

New Map

Merge

Figure 3. FAN Main Goal.

G1: Successfully Detail Mapped Area Commodity: Map Data

GUI Mock UPS

Representation Design Requirement Presentation Design Concepts ACWA template

Figure 1. Design methodology.

Repository

Call back functions

Old Map Data

New Area Data

Comparison

Correlation, Integration

Updated Area

New Area

Figure 4. Main Goal Supporting Function. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

418

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

environment data, successfully recovering command information and evaluating operative mode constrains. As stated above, the scenario of the project is a field mapping oriented mission, all the other features and mission possibilities will be dropped to simplify the case of study. This assumption is needed to define a work domain not as complex as a whole mission definition, which allows researchers to describe it in minute details. For this reason, only the rover functions directly coupled with its motion capabilities and its deployment abilities are taken into account and investigated during this work. Furthermore, to ease the FAN depiction, the mission is intended to be in nominal mode; failure management will be only expected but it will be not investigated. The FAN representation of the work domain’s concept brings us to the second step of the ACWA process: deriving the CWR. COGNITIVE WORK REQUIREMENTS CWR represent the cognitive demands for each part of the domain model, i.e., all type of recognition, decision-making and problem-solving activities. In the methodology, the addition of CWRs to the design repository is described as “thickening” the analysis (Elm et al., 2003): in fact, each CWR is attached to a node of the FAN as an enrichment of the domain concept understanding. Based on the underlying premises of the CTA, these CWR center around either goal or functional process, monitoring for goal satisfaction and resource availability, planning or selecting among alternative options to achieve goals, and controlling the functional process. By organizing the specification of operator’s cognitive requirements around nodes in the FAN rather than organizing requirements around predefined task sequences, the representation helps to ensure a consistent, decision-centered perspective (Table 1). This perspective implies that the FAN nodes can be associated with different kind of CWRs: some of them can be found across a variety of domains, so a “template” of generic CWR can be tested for each node of the FAN, while others are uniquely coupled with the scenario features and demands. Thus, the FAN forms the basis for the structure of the cognitive demands reflected in the CWRs. For example, every goal node in the FAN has associated “goal monitoring” decisions; likewise, processes have associated “process monitoring” decisions and, similarly, there will always be some “feedback monitoring” decisions related to assessing whether actions are achieving the desired result. In many ways, this step is where the decision support system requirements are shaped: hence, good CWRs depiction

Table 1. Cognitive Work Requirements for Goal 1. CWR description

CWR-G1-1

Monitor field mapping progress with respect to the overall map dimension

CWR-P1-1

Monitor the presence of old map of the target area already available

CWR-P1-2

Select source between old and new map that maximize information accuracy

CWR-P1-3

Compare old and new data

CWR-P1-4

Monitor correct correlation of new mapped areas to already mapped ones

is essential to the final resulting GUI. A portion of the CWR obtained from the FAN are listed below: • Monitoring field-mapping progress regarding overall map dimension; • Monitoring successful positioning of the sensor for data acquisition; • Selecting the optimal motion of the rover to maximize sensor potential; • Choosing the best command sequence to move the rover to the acquisition site; • Monitoring current command input to avoid limit crossing; • Selecting the best navigation option to maximize rover response during mission; • Monitoring actual rover movement capabilities to maximize its potential; • Monitoring actual rover movement possibilities to perform defined action; • Monitoring rover system behavior to obtain information about its actual status; • Monitoring correct acquisition of sensed commands. INFORMATION AND RELATIONSHIP REQUIREMENTS While mental demands of every FAN node are gathered through CWRs definition, the information required for each decision to be made is still unidentified. IRRs are defined as the set of informative elements necessary to successfully resolve the associated CWRs. Thus, the focus of this step in the methodology is to identify the ideal and complete set of information for the

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

associated decision-making process. Therefore, IRR forms the basis for the visualization support to be designed in the next phase. However, information is not data: data can require complex transformations to become information. Thus, IRRs have a much deeper impact on the entire system architecture than merely the “look and feel” of the final GUI (Elm et al., 2003). It is important to note that IRRs are strongly related to mental demands fulfillment (CWRs) and are not limited by data availability in the current system. If the required data is not directly available, ACWA provides a logical basis to obtain that data (e.g. pulling data from a variety of databases, adding additional sensors, creating “synthetic” values); this leads to one of the assumptions of the project: some of the information obtained from IRR requires data not available with current state of the art sensors, but strongly needed for a successful space telenavigated mission. This approach is different in relation to the one that humanfactor engineers have had in the past (designing an interface after the system equipment has been specified). The standard approach, in fact, cannot give any profit in terms of understanding the user’s MW involved in the process of acquiring the correct SA. Consequently, “the ACWA approach is fundamentally broader in scope than other approaches to interface design that do not consider the impact of IRRs on system architecture specifications” (Vicente et al., 1996). It is easy to notice that a lot of information can be needed by a single CWR to be completely defined. Furthermore, high-level goals — as the ones in Table 2 — are dependent upon supporting information, successfully developed in lower level goals. A summary table was created to correctly evaluate all the information needed to satisfy the mental demands of our domain. In this table, the information are correlated to the supporting function, in order to define a hierarchy between them. Table 3 collects all the information which need to be provided to the user in order to trigger the cognitive processes related to each functional process. This information is, in many cases, the same for different cognitive processes — as it is shown on the second column of the table, where each number corresponds to a specific cognitive process. The table allows correlating each piece of information to the number of cognitive processes in which it is involved. Information involved in many cognitive processes need to have priority on the GUI, because the user will need them all the time to maintain its SA. Upon obtaining IRRs, the ensemble containing the FAN, the CWRs and their associated IRRs represent a solid basis for the development of the decision-making graphical support.

419

Table 2. Information and Relationship Requirement for Goal 1. IRR description for Goal 1 listed with respect to CWR

CWR-G1-1

Monitor field mapping progress regarding overall map dimension

IRR-G1-1.1

Deliver mapping progress status with respect to the whole area of interest

IRR-G1-1.2

Deliver typical rate of acquisition (design limits)

IRR-G1-1.3

Actual rate of acquisition with respect to typical rate

CWR-P1-1

Monitor the presence of old map of the target area already available

IRR-P1-1.1

Deliver selected area availability in storage

IRR-P1-1.2

Deliver data accuracy of old map

CWR-P1-2

Select source between old and new map that maximize information accuracy

IRR-P1-2.1

Rate of change between old map data and new map data

IRR-P1-2.2

Data already stored with respect to selected zone

CWR-P1-3

Compare old and new data

IRR-P1-3.1

Actual sensed data accuracy with respect to stored data accuracy

IRR-P1-3.2

Rate of change between stored and sensed data

CWR-P1-4

Monitor correct correlation of new mapped areas to already mapped ones

IRR-P1-4.1

Actual position of the sensing system

IRR-P1-4.2

Deliver nearest position available in the stored map

IRR-P1-4.3

Actual error in position

However, to obtain a Graphical Interface that communicates with operators without effort, the IRRs have to be converted into visual widgets. The last two steps of the ACWA, which are the RDRs and the PDCs, allow the methodology to bridge this gap. REPRESENTATION DESIGN REQUIREMENTS AND PRESENTATION DESIGN CONCEPTS RDRs define the goals and scope of the information representation, in terms of cognitive tasks it is intended to support. It also provides a description of the supporting

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

420

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

Table 3. Portion of Information and Relationship Requirement list. IRR

Goal supported by IRR

Command information: acceleration (actual and maximum)

4-5-6-7.5-8.2

Command information: acquisition sensor check up

7.5-8.2

Command information: command history

7.5-8.2

Command information: command verification

8.2

Command information: commanded torque

4-7.5-8.2

Command information: communication bandwidth

8.2

Command information: communication coverage

8.2

Command information: communication windows

8.2

Command information: electric current needed

4-5-6-7.5-8.2

Command information: maximum torque available

4-5-6-7.5-8.2

Command information: Power available to engines (actual and typical)

4-5-6-7.5-8.2

Command information: speed (actual and maximum)

4-5-6-7.1–7.2–7.5-8.2

Command information: Wheels Differential torque

4-7.5-8.2

Environment data: acquisition data rate (actual and typical)

8.1

Environment data: memory space available

8.1

Environment data: sensor status

7.3-8.1

Environment data: sun exposure (actual and typical)

8.1

Environment data: temperature (actual and typical)

8.1

Environment data: wind speed (actual and typical)

8.1

Guidance and Navigation data: Checkpoint position

3-4-5-6

information required to sustain the cognitive tasks. Furthermore, this step in the process begins shifting the attention from “what” is to be displayed (defined by FAN, CWRs and IRRs) to “how” to display it. It adds a more complete description of the behaviors and features needed to communicate the information effectively, as well as an allocation of the Information/Relationship Resources across the entire set of displays within the workspace. The visual framework of a Rover Simulation has been taken into account to obtain the RDRs and the PDCs. A software framework able to create 3D immersive virtual simulations has been used in order to develop a particular Rover Simulation application. The software was developed inside a Piedmont regional funded project, ended in May 2012, called STEPS (Sistemi e Tecnologie per l’EsPlorazione Spaziale). STEPS was a project supported by Regione Piemonte and carried out by Thales

Alenia Space Italia, Small and Medium Enterprises, Universities and public Research Centres belonging to the network “Comitato Distretto Aerospaziale del Piemonte”. The project objective was to develop hardware and software demonstrators for descent, soft landing and surface mobility of robotic and manned equipment during Moon and Mars exploration. The demonstrator was created in the Collaborative System Engineering (COSE) Centre, a Thales Alenia Space Italia facility in Torino which operates within the Engineering & Advanced Studies Directorate of the Domain Exploration and Science Italy. The Centre main mission is to research, develop, integrate and propose new methodologies and tools to enhance the system engineering capabilities and the multidisciplinary collaboration. Using the Virtual Rover Demonstrator it is possible to simulate a rover motion on a planetary surface.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

Three visual displays are used to obtain a wide screen virtual experience of the rover and the terrain (Fig. 5). Another benefit provided by RDRs is that, as long as the domain remains unchanged, the RDR serves as an explicit documentation of the presentation concept purpose, despite of the technologies available and used to implement it. As newer technologies become available, and as their interaction with human perception becomes better understood, the technologies used to implement the RDR can evolve. The RDRs are obtained splitting the IRRs into: • Essential information: it represents vital information, which needs to be on the central screen. It gathers information which have to be easily accessed by users; • Detailed information: represents information which may improve user’s awareness with an increase in their workload. This information is more accurate and detailed and can be used to improve awareness on a selected feature. The following data have been evaluated as part of essential information: user commands, speed, rover heading and “out of range” alert of these parameters. Other “alert signals” have to be fit into costless information: proximity hazard, pitch acceleration and roll acceleration. These alerts are needed because the rover stability can be heavily compromised if these parameters exceed their thresholds. This information represents the one related to a greater number of IRRs. On the other hand, in the detailed information, there are included the pitch and roll, angular velocity and speed of the wheels. It is easy to notice that the user may perform a proper telenavigation task by focusing attention on the central screen. The information on lateral screens plays a detailing role: if the user understands that something is not working as intended (e.g. from no-cost information) he/she will focus on lateral screens in order to obtain more detailed information. The set of information used on the lateral screen are obtained from the one that will affect a smaller number of functional processes. After the development of the RDRs, the last step in the methodology is obtaining explicit PDCs for the Interface (as GUI mock-ups). This final step requires the knowledge of the human perception and its interaction with the various presentation techniques and attributes. With the RDR as a guide, the sketches, drawings, and brainstorming concepts can all be resolved against the display’s intent and requirements. The issues of how it is perceived can best be done with empirical testing of prototypes and often requires considerable tuning and

421

Figure 5. Visual Display Framework.

adjustment to achieve the representation capabilities specified in the RDRs. During this phase, researchers found, for example, that the information panels couldn’t be put on the further edge (in relation to the central screen position) of both the side screens due to their dimensions. In fact, if positioned on the further edge, information becomes accessible only with a larger motion of the head, with a consequent degradation in driving performance. To avoid this issue, while maintaining a good distinction between essential and detailed information, researchers choose to insert the information panel on the nearest edge of the lateral screens. The availability of two different panels (both the side screens are used to represent detailed information) allows a further classification to aid users in the rationalizing process of the interface. Thus, researchers try to collect on the left screen the detailed information related to kinematic data and system parameters, while on the right screen were implemented navigation information. The final mock-up of the GUI obtained using this methodology is shown below (Figs. 6, 7 and 8): The availability of an older interface (Fig. 9) with the same purpose of the one created using the ACWA approach, allows researchers to enhance it, decreasing the time spent for this last step. All the improvements done on the interface are listed in Table 4. TEST DESCRIPTION An empirical test was developed in order to obtain information both on the user SA and on their performance in task fulfillment. Several methods of testing SA have been documented (Endsley et al., 2000; Endsley, 1995a; Endsley, 1995b), usually divided into: • Knowledge-based measurement techniques; and • Performance-based measurement techniques.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

422

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

Figure 6. Left screen Graphical User Interface.

Figure 9. Old Graphical User Interface for remote control.

Table 4. Graphical User Interface improvements.

Figure 7. Central screen Graphical User Interface.

Figure 8. Right screen Graphical User Interface.

Screen

Improvement

Lateral-left

Battery status bar added

Lateral-left

Wheel slip indicator added

Lateral-left

Artificial horizon added

Lateral-left

Angular velocity display added

Central

Directional compass added

Central

Ghosting function improved

Central

Torque-related color function added

Central

Rover speed indicator added

Lateral-right

Hexagon position modified

Lateral-right

Map position modified

Lateral-right

Rover heading function added in the map

The knowledge-based measurement techniques are founded on either simulations or real trials. The task is selected according to the level or type of SA being addressed by the experiments. The method identifies independent variables (i.e., the type of display for a GUI, the type of interaction device used to pilot a remote vehicle) and dependent variables, such as objective and subjective measures of testeing knowledge (i.e., understanding) and performances. Furthermore, there are several complex techniques which attempt to determine or model the subject’s knowledge of the situation at different times throughout simulation runs.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

For example, the Situation Awareness Global Assessment Technique (SAGAT) freezes the simulator screens at random times during the runs, and queries the subjects about their knowledge of the environment. This knowledge can be at several levels of cognition, from the most basic of facts to complicated predictions of future states. The performance-based measurement of SA has taken several forms. Some techniques measure the overall final performance of the human-in-the-loop system in any or all of its tasks. Alternatively, Testable Responses can be used in order to evaluate SA: the subjects will face predetermined situations during the simulation that require decisive and identifiable actions, if the subjects have the correct level of SA, they can correctly perform the required actions while, with low awareness, they cannot perform it. As a general comment, to provide a detailed assessment of the subject’s SA, the knowledge-based techniques are more accurate, as they measure these variables directly. Performance-based measurement can only make inferences based upon the particular information the subject acted upon, and how it was interpreted, thus these techniques are very useful when well-determined performance are assessed, while the knowledge-based techniques can be more accurate when a lot of different aspects are observed together, and multiple performance assessments can be made. Self-rating techniques are used in order to gain a subjective assessment of participant’s SA. Typically administered posttrial, self-rating techniques involve participants providing a subjective rating of their perceived SA via an SA related rating scale. As an example, the Situation Awareness Rating Technique (SART) is a subjective rating technique developed for the assessment of the pilot’s SA. The primary advantages of self-rating techniques are their ease of application (easy, quick and of low cost) and their non-intrusive nature (since they are administered post-trial) very useful for the early assessment in the design process. However, subjective self-rating techniques are heavily criticized for several reasons, including the various problems associated with the collection of SA data post-trial (correlation of SA with performance, poor recall) and also issues regarding their sensitivity. Another method used in order to estimate SA is based on observer’s ratings during or at the end of the trial. Observer rating techniques typically involve a Subject Matter Expert

423

(SME) observing participants performing the task under analysis and then providing an assessment or rating of each participant’s SA. The SA ratings are based upon observable SA related behavior exhibited by the participants during task performance. The main advantages associated to the use of observer rating scales to measure SA are their non-intrusive nature and their ability to be applied ‘in-the-field’. However, the extent to which observers can accurately rate participant SA is questionable, and also multiple SMEs may be required. A five points rating scale (1 = very poor, 5 = very good) and an additional ‘not applicable’ category can be used for each observable SA related behavior of the tester. The objectives of the testing phase are different: on one hand a first evaluation of the ACWA have to be made in order to assess its impact on the user perception, on the other hand the GUI itself has to be tested to verify its usability. To obtain a feasible and light testing methodology that can inspect these features, a brand new approach to the testing phase is developed. A performance-based concept is used in order to investigate the GUI usability while self-rating questionnaires are defined to obtain information about operator’s SA. These early tests in the overall design phase of the simulator cannot take advantage of the knowledge-based techniques: the overall behavior of the system is still undefined, thus it is impossible to define independent variable to be observed. While the self-rating questionnaires are administered post-trial in order not to influence the performance, the SMEs will focus their attention on the operator’s behavior during the trials and on their accord with the GUI. The coupling of the self-rating techniques with the observer rating techniques allows practitioners to enhance the testing phase thanks to the ease of application of these methodologies without incurring in the subjective limitations of the user self-assessment. Furthermore, this mixed technique offers the best tradeoff between result’s significance and test costs. The performance-based measurement obtained from the mixed (self-rating and observer) testing technique can be used as an indicator of the effectiveness of the ACWA definition of the interface. The GUI, defined through the ACWA, should enhance the operator’s awareness, easing his understanding of the process: if an increase in the user’s performances can be found during the test, the ACWA has a positive impact on the interface design.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

424

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

TEST PROCEDURE The testing procedure is hereafter shortly described. A number of selected users were expected to drive the rover in remote manual control (i.e., using the joystick) along an assigned path (defined by 6 waypoints, in alphabetical order from A to F, distributed on the map). Each test is completed when the rover reaches the final waypoint F, or when the battery completely discharges. Together with this main task, a second — and less important — goal has been defined to increase the user MW. This second objective referred to a completely different mental process regarding the one involved in the fulfillment of the main goal. To accomplish this, the users were asked to count the number of “skid events” — i.e., to count the number of bars which turn red, Fig. 10 — while performing the main task (to reach the final waypoint F). The SA assessment is based on expert observer’s ratings (during the test) and user’s self-ratings (at the end of the test). Prior to testing, a further questionnaire was given to the user: the aim of this “Relationship Questionnaire” (RQ) (Bartholomew & Horowitz, 1991) was to evaluate the attachment style of every participant. The RQ is a self-report utility which allows investigating the general orientation of our adult intimate relationships, regarding psychological and emotional intimacy. The test is arranged in four short self-descriptions, each of which summarizes the basic aspects of one of the four main patterns of attachment. This classification system allows the subject to be identified, as a prototype, within one of the four attachment styles: secure attachment style, dismissiveavoidant, preoccupied and fearful-avoidant. During the test, subjects have to choose the most representative prototype of themselves, furthermore they also have to assess the extent to which each of the four prototypes represents them, using a 1 (strongly disagreement) to 7 (total agreement) scale. This allows us to assess both the image that the subjects have of themselves and the image that they have of the others: obtaining, by the intersection of the data, a “model of self ” and a “model of others”, which may be positive or negative. In this way, the RQ investigates the hypothesis of Bowlby (1973), that attachment styles reflect the internal working models of self and others, which can be both positive and negative. It follows that the secure attachment, in which subjects are characterized by coherence, autonomy and self-confidence is derived from

Figure 10. Highlighted Skid events indicator on left screen.

the combination of a positive model of self and a positive model of others. The dismissive-avoidant attachment results from a positive self-evaluation associated to a negative representation of others: subjects tend not to be coherent, they sacrifice the intimacy and deny the importance of relationships, in “a kind of self-sufficiency affective and existential” (Bruni, 2004). The combination of a negative model of self and a positive model of the other develops instead a preoccupied attachment, in which the subjects, which are distinguished by their relentless pursuit and apprehension about the relations, safeguard their low selfesteem exposing the incoherence and idealization of their own relationships. Finally, fearful-avoidant attachment is derived from the combination of a negative model of self and a negative model of the other; fearful individuals avoid involvement with others for fear of being rejected, even though they wish it fervently, since, because of their low self-esteem, they do not feel worthy and expect others to be ill-disposed, unreliable and rejecting. Attachment styles highlighted by the RQ test also adapt, to some extent, to the relationship of “trust” that is established between the pilot and the interface of the vehicle driven. In addition to that, researches have extensively demonstrated that the different attachment styles are related to the ability to handle, emotionally and cognitively, different situations and stress that they can generate (Ricco, 2009). The attachment style will directly affect perception and, also, cognitive processes that can be used: consequently it has a key role in the efforts to improve the GUIs.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

Thus the RQ could give information about the attachment style required to improve the teleoperation: in fact, shifting from the concept of pilots to the idea of supervisors could change the desired psychological profile of the user. For example, pilots, most of the time, are very self-confident and self-centered people: this could lead to a difficult interaction with automated systems which require the confidence of the pilot to operate, exactly as it happens between people that need mutual trust in order to collaborate on a project. The tester experience started in a room next to the simulator where he has to fill the RQ isolated from the other users to avoid suggestions; the anonymous questionnaire, sealed inside a blank envelope, were collected by a dedicated practitioner. After the RQ collection, the tester could perform the driving session: in this phase three observers must fill out their questionnaire, looking at the user’s behaviors and at their performance (i.e., eyes motion, head motion or collecting defined data during the attempt) while maintaining absolute silence to avoid tester’s distraction. The last phase consists of the self-rating questionnaire: the tester answers the questions alone in the adjacent room then the anonymous questionnaires were collected by a dedicated practitioner. These questionnaires will explore the perception of the GUI from the operator point of view: some questions are related to parameter controlled by the observers (e.g., rover speed) in order to have a feedback on performance, while other are related to the cognitive processes that the operator have to exploit to correctly maneuver the rover. These questions can be used as a direct investigation on the ACWA effectiveness: if the operator cannot understand the GUI elements, it means that the process that he exploits is different from the one obtained from the ACWA that defines the GUI shape. Within three months, two different sessions of tests were scheduled. Both the sessions involved same users and same observers in order to avoid erroneous comparison. In the last test, the GUI was not modified, but a time restriction was given to the testers in order to increase their MW (Paas et al., 2004). The main task (to reach the final point “F”) and the secondary objective (to count the skid-events) were maintained, while the execution time was diminished by 20% in relation to the execution time measured during the first test. The aim of this second test was to increase the performance-based knowledge of the user’s understanding

425

process in a more dynamic and strained environment, and to verify the evaluations obtained in the previous test.

TEST The user should demonstrate their ability to detect, understand and respond to events while maintaining a good performance in both assigned tasks. To do this, they have to maneuver the rover through six waypoints positioned on Mars’ surface in order to reach the final waypoint F. While the user is performing the test, three observers will judge their behavior without interference. The observer’s ratings are divided into knowledge-based (e.g., distance and speed, randomly collected by the observer) and performance-based (e.g. time required to complete the path, number of collisions, average cross-track error in forward drive or the ability to drive straight, response to the presence of a random obstacle). When the driving test is completed, the user has to fill a self-rating questionnaire. The self-ratings give the assessment of the overall performance according to user’s opinion. General open comments are collected from both observers and testers, mainly related to the user’s SA and the HMI performance. In order to allow comparison within tests, the same questions made in the first test were used in this self-rating questionnaire. Furthermore, the same users were engaged for testing the new GUI. The number of users involved in the tests is limited (N = 7): this is acceptable because the number of operators is very limited in space applications. Space agencies invest many resources in training operators in order to employ highly specialized professionals. This leads to a deep knowledge of the system on which they are operating: therefore different applications employ users who possess a slightly different training. Thus, statistical survey cannot show direct correlation or general behavior because each system requires different skills and training. For this reason, the results of the tests performed to evaluate the developed GUI will not be statistically useful. EVALUATION Testing evaluation was made using two different modules: the first one, a self-rating questionnaire, was given to the user after the end of their performance, while the second one, the Observer-rating questionnaire, was filled during the test by three

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

426

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

observers. However, to avoid erroneous or misleading behavior, the observer could not interfere with the user during the tests. The self-rating questionnaire consisted in different questions about user’s perception of the overall test: they are asked to remember detailed parameters (e.g., the average speed maintained during the test or battery level at the end of the test) and to give feedback about their perception (judging several features of the GUI as the directional compass or the proximity hazard feature). The majority of the questions consisted in a multiplechoice answer to ease the user’s fulfillment possibility: an example is shown below (Table 5). The observer questionnaire tracked down the same parameters, e.g. rover speed, together with indicators that can help understanding user’s SA, such as coherence between user’s eyes motion and their task or their execution time (Table 6). The self-rating questionnaire and the observer questionnaire were then compared to obtain information about the overall SA of the user. The answer of observers can be used as neutral criteria to evaluate the personal, thus subjective, ratings of the users. In this way, pratictioners gathered information about performace, obtained from the observers annotations during the testers trial (e.g. average speed and execution time), and about interface effectiveness (thus the effectiveness of the cognitive approach was used to define the interface itself), obtained comparing the observer’s annotation and the subjective rating of the user. For example, the tester has to remember the level of charge of the battery at the end of their trial: the comparison between their answer and the actual data gathered by the observers allows obtaining information about the information transfer through the interface.

On the other hand, the comparison between the first test and the second one, which has an increased tester’s MW, may improve the knowledge about the cognitive process of testers. The increase in MW means a lowering in SA: this will affect tester capability of acquiring information from the system. This could highlight the cognitive processes of the tester: if the operator cognitive process and the cognitive approach used in the interface design exploits the same process this leads to a natural acquisition of the information by the operator (i.e. the tester has not to process the information to understand it). In this way, the test could also verify the results of the previous test about interface effectiveness. On the other hand, the RQ has been used to deepen the attachment style of the tester in order to obtain information about the tester’s individual difference that can significantly affect their performance. The RQ was submitted to the user before the start of the test and consists in four brief descriptions which summarize each of the main patterns of a different attachment style, from secure to fearful. TEST RESULTS In this section, the results of the two tests performed on the different GUIs have been reported. The following tables present the attachment style, the personal evaluation and the evaluation of the observer, for each of the seven users. In the second test (that was carried out with an increase MW), only two of the testers reach the final point F within the time restriction. Furthermore, both of them possess a secure attachment style (Table 7). This means that the secure attachment style, in which subjects are characterized by consistency, autonomy and self-confidence, may have a positive impact on the relationship of “trust” that is

Table 5. Self-rating questionnaire. How do you judge your performance in terms of cruise speed?

☐Not applicable

☐ Very poor

☐ Poor

☐ Average

☐ Good

☐ Very good

Looking at the map, how you judge red markers to detect the obstacles?

☐ Not applicable

☐ Very poor

☐ Poor

☐ Average

☐ Good

☐ Very good

☐ Good

☐ Very good

Table 6. Observer questionnaire. User eyes motion and coherence between current task and attention appointment.

☐ Not applicable

☐ Very poor

☐ Poor

☐ Average

Covered distance (number of waypoints)

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

established between the pilot and the interface of the vehicle driven. However, there are also other two secure testers that have not completed the task in the given time: this means that a further investigation has to be conducted in order to assess the real correlation between pilots’ attachment style and their performance. On the other hand, the results can give information about the overall effectiveness of the design methodology in terms of users’ performance and SA. Comparing testers’ answers given in the two tests with the data collected by observers, it is possible to understand if the tester’s perception of the GUI was coherent. Table 8 shows that users can evaluate and remember correctly the remaining battery level of charge at the end of the test. This means that also in a high MW environment (i.e. the second test), this feature enhance pilot’s SA and satisfy the pilot’s mental demands. Also, the future-position indicator (that was improved because of the IRR definition) was perceived correctly by the testers (Table 9).

427

From this comparison, it appears that the tester perception of the future position indication was better in the second test: this could be derived from the increase in their mental demands, thus the elements of the GUI that reduce their workload are more evident. Another element that has been evaluated is the map: in Table 10, the testers’ subjective assessment of the red marker used for obstacle detection is given. The results highlight a slightly increase in testers usage and perception of these elements in the second test. This leads toward two different assumptions: the first one is that, with an increase in the MW, testers have to improve their perception using the map elements more frequently, while the second one is that, during the first test, the performance pressure on testers was too low to force them to actively use the GUI at its maximum. In Table 11, the wheel skid counter has been reported: this was the secondary objective of the test to increase the MW of testers without conditioning their performance. If we compared this table to the observed eye motion coherence (Table 12), it is

Table 7. Task performance test number 2. Tester

Attachment

Collision (Test I)

Collision (Test II)

Time restriction

Secure

Yes

Reserved

Secure

Yes

Secure

Troubled

Reserved

Yes

Secure

Table 8. Battery level, test number 1 and 2. Tester

Tester (Test I)

Observer (Test I)

Tester (Test II)

Observer (Test II)

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

428

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

noticeable how testers with a poor coherence (i.e., tester 1 in test I, tester 2 in test II) cannot tell the right number of skid events. On the other hand, most of the participants have performed the secondary objective with a good level of accuracy: this means that the overall complexity of this task was correctly tuned in both of the tests. In addition to that, the MW perceived by testers were paired with their eyes movement coherence as detected by

observers (Table 12): it is easy to notice that testers subjected to high MW (average/high in the tester charts) most of the time have a low eyes motion coherence, which implies an overall difficulty in gaining the correct SA with regard to the current task. Furthermore, it seems that the time restriction has not influenced the perceived MW of the testers: most of the testers found the first test demanding, thus a further increase in their workload could have not been correctly perceived.

Table 9. Future position perception.

Table 10. Red marker obstacle detection on map.

Tester

Test I

Test II

Tester

Test I

Test II

Good

Poor

Good

Poor

Average

Good

Average

Good

Average

Good

Average

Poor

Good

Average

Good

Table 11. Wheel Skid counter. Tester

Tester (Test I)

Observer (Test I)

Tester (Test II)

Observer (Test II)

Table 12. Mental Workload and tester coherence. Tester

Tester (Test I)

Observer (Test I)

Tester (Test II)

Observer (Test II)

Average

Poor

Average

High

Very Poor

Low

Very Poor

Average

Very Good

Average

Low

Good

Low

Average

High

Very Good

High

Poor

High

Good

High

Very Poor

Average

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

Cognitive Based Design of a Human Machine Interface for Telenavigation of a Space Rover

CONCLUSION AND FUTURE IMPLEMENTATION This paper presents the development of a GUI for remote control of a rover in a planetary environment and the following testing phase to evaluate its performance regarding users’ SA and MW. The methodology used in order to define the GUI stands out against the other approaches, because it makes the cognitive process of the user a priority in relation to the other development guidelines. Furthermore, it allows, through an iterative process, to update the HMI based on the test results. This can be done because there is a strong interconnection between the physical architecture of the system and the HMI: if tests point out that something has to be modified, researchers can easily understand which information has to be changed and where this information is used. On the other hand, the multi-agent tests allow researchers to understand the lacks in the interface using few testers and link these results to a well-defined psychological profile. Furthermore the testing technique has proved as a feasible methodology to obtain information on the GUI effectiveness and on the operator performance in early phase design, while more complex and expensive techniques, as knowledge-based ones, cannot be used.

429

As research perspective: • The information which cannot be correctly understood by testers will be revised to improve the GUI during a second and more detailed design overview; • Involving observers is a useful improvement of test evaluation though it is important that parameters collected during tests are as objective as possible, precisely to prevent them from being distorted by factors purely individual. This will lead to the use of automatic functions that support observer rating (e.g. autonomous eyes motion capture or tracking of commands given by the pilot) in future testing phases; • A further and more complex (i.e. using intrusive testing methodologies) research about the connection between user’s attachment style and their ability to successfully teleoperate a system has to be made.

ACKNOWLEDGEMENT The authors would like to thank the management, the researchers and the engineering staff of Thales Alenia Space Italia (COSE-Center) for the essential technical support during implementation and testing of the HMI concept.

REFERENCES Bartholomew, K. and Horowitz, L.M., 1991, “Attachment styles among young adults: A test of a four-category model”. Journal of Personality and Social Psychology, Vol. 61, No. 2, pp. 226-244. doi:10.1037/0022-3514.61.2.226. Biesiadecki, J.J., Leger, P.C. and Maimone, M.W., 2007, “Tradeoffs Between Directed and Autonomous Driving on the Mars Exploration Rovers”. International Journal Of Robotic Research, Vol. 26, No. 1, pp. 91-104. doi: 10.1177/0278364907073777.

Chu, R.W., Mitchell, C.M. and Jones, P.M., 1995 “Using the operator function model and OFMspert as the basis for an intelligent tutoring system: towards a tutor/aid paradigm for operators of supervisory control systems”. IEEE Transactions on Systems, Man and Cybernetics, Vol. 25, No. 7, pp. 1054-1074. doi: 10.1109/21.391287. Corujeira, J., 2013, “A situation awareness interface for a bi-wheeled industrial hovercraft: design, development and evaluation”. Master Thesis, Universidade de Madeira, Funchal, Portugal.

Bowlby, J., 1973, Attachment and loss. Vol. 2, Separation. New York.

Eberts, R.E., 1994, “User Interface Design”. Engelwood Cliffs. Prentice Hall, NJ.

Bruni, R., 2004, “Nuovi scenari della genitorialità: la famiglia tra affetti e legami”. Istituto Italiano di Medicina Sociale-Azioni di sostegno alla genitorialità, pp. 145-184.

Elm, W.C., Potter, S.S., Gualtieri, J.W., Easter, J.R. and Roth, E.M., 2003. “Applied Cognitive Work Analysis: A Pragmatic Methodology for Designing Revolutionary Cognitive Affordances”, In: Hollnagel, E. (Ed.), Handbook for Cognitive Task Design, Lawrence Erlbaum Associates, London, pp. 357-382.

Burridge, R.R. and Hambuchen, K.A., 2009, “Using prediction to enhance remote robot supervision across time delay”. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). doi: 10.1109/IROS.2009.5354233. Card, S.K., Moran, T.P. and Newell, A., 1983. The psychology of human-computer interaction. Lawrence Erlbaum Associates Inc. NJ.

Endsley, M.R., 1995a, “Measurement of SA in dynamic systems”. Human Factors: The Journal of the Human Factors and Ergonomics Society, Vol. 37, No. 1, pp. 65–84. doi: 10.1518/001872095779049499. Endsley, M.R., 1995b, “Toward a theory of SA in dynamic systems”. Human Factors: The Journal of the Human Factors and Ergonomics Society, Vol. 37, No. 1, pp. 32–64. doi: 10.1518/001872095779049543.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

430

De Filippis, L., Gaia, E., Guglieri, G., Re, M. and Ricco, C.

Endsley, M.R., Sollenberger, R. and Stein, E., 2000, “Situation Awareness: a Comparison of Measures”. Proceedings of the Human Performance, Situation Awareness and Automation: User-Centered Design for the New Millennium. Gordon, S.E. and Gill, R.T., 1997, “Cognitive Task Analysis”, In: Zsambok, C.E., Klein, G. (Eds.), Naturalistic Decision Making, pp. 131-140. Hashimoto, S., Ishida, A., Inami, M. and Igarashi, T., 2011, “TouchMe: An Augmented Reality Based Remote Robot Manipulation”. The 21st International Conference on Artificial Reality and Telexistence, Proceedings of ICAT2011, Osaka. Hutchins, E.L., Hollan, J.D. and Norman, D.A., 1985, “Direct manipulation interfaces”. Human Computer Interaction, Vol. 1, No. 4, pp. 311-338. doi: 10.1207/s15327051hci0104_2. Narayanan, S., Ruff, H.A., Edala, N.R., Geist, J.A, Patchigolla, K.K, Draper, M. and Haass, M., 2000, “Human-integrated supervisory control of uninhabited combat aerial vehicles”. Journal of Robotics and Mechatronics, Vol. 12, No. 6, pp. 628-639. Nielsen, J., 1994, “Usability Inspection Methods”. Conference Companion on Human Factors in Computing Systems, pp. 413414. doi: 10.1145/259963.260531. Norris, J.S., Powell, M.W., Fox, J.M., Crockett, T.M. and Joswig, J.C., 2009, Cliffbot Maestro, NASA Tech Briefs. Okura, F., Ueda, Y., Sato, T. And Yokoya, N., 2013, “Teleoperation of mobile robots by generating augmented free-viewpoint images”. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 665-671. doi: 10.1109/IROS.2013.6696422.

Paas, F., Renkl, A. and Sweller, J., 2004, “Cognitive Load Theory: Instructional Implications of the Interaction between Information Structures and Cognitive Architecture”. Instructional Science, Vol. 32, No. 1-2, pp. 1-8. doi: 10.1023/B:TRUC.0000021806.17516.d0. Rasmussen, J., 1983, “Skills, rules, and knowledge; Signals, signs, and symbols, and other distinctions”.IEEE Transactions on Systems, Man and Cybernetics, Vol. 13, No. 3, pp 257-266. doi: 10.1109/ TSMC.1983.6313160. Rasmussen, J., 1986, “Information Processing and Human-machine Interaction: An Approach to Cognitive Engineering”. System Science and Engineering, Elsevier Science, New York. Rasmussen, J. and Vicente, J.K., 1989, “Coping with human error through system design: Implications for ecological interface design”. International Journal of Man-Machine Studies, Vol. 31, No. 5, pp. 517-534. doi:10.1016/0020-7373(89)90014-X. Ricco, C., 2009, “La matrice relazionale del sé: fallimento genitoriale e trauma evolutivo nei figli”. In: Borgogno, F., Vigna Taglianti, M. (a cura di), Identificazione e trasmissione psichica fra gli adulti e il bambino. Quaderni di Psicoterapia Infantile, Borla, Roma. Störrle, H., 2010, “Model Driven Development of User Interface Prototypes: An Integrated Approach”. Technical University of Denmark, Copenhagen, Denmark. Vicente, K.J., Christoffersen, K. and Hunter, C.N., 1996 “Response to Maddox critique”. Human Factors, Vol. 38, No. 3, pp. 546-549. doi: 10.1518/001872096778702051.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.415-430, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.399

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception Valentino Lau1, Fabiano Luis de Sousa1, Roberto Luiz Galski1, Evandro Marconi Rocco1, José Carlos Becceneri1, Walter Abrahão dos Santos1, Sandra Aparecida Sandri1

ABSTRACT: One issue the design team has to face in the process of building a new spacecraft, is to define its mechanical and electrical architecture. The choice of where to place the spacecraft´s electronic equipment is a complex task, since it involves simultaneously many factors, such as the spacecraft´s required position of center of mass, moments of inertia, equipment heat dissipation, integration and servicing issues, among others. Since this is a multidisciplinary task, the early positioning of the spacecraft´s equipment is usually done “manually” by a group of system engineers, heavily based on their experience. It is an interactive process that takes time and hence, as soon as a feasible design is found, it becomes the baseline. This precludes a broader exploration of the design space, which may lead to a suboptimal solution, or worse to a design that will have to be modified later. Recently, it has been shown the potential benefits of automating the process of spacecraft´s equipment layout using optimization techniques. In this paper, a prototype of an Excel® based tool for multidisciplinary spacecraft equipment layout conception is described. Provided the geometric dimensions, mass and heat dissipation of the equipment, and the available positioning area, the tool can automatically generate many possible trade-off solutions for the layout. It allows the user to set specific equipment to specific areas of positioning, and different combinations of objective functions can be used to drive the design. The features of the tool are shown in a simplified three dimensional problem. KEYWORDS: Layout optimization, Spacecraft, Conceptual design, Electronic equipment.

INTRODUCTION In the conceptual phase of the development of a new spacecraft, different candidate solutions for its electrical and mechanical architectures are assessed, in a search for one which would fit the spacecraft mission, within the constraints of cost and schedule. It is in this phase that the main features of its subsystems are defined, and where the systemic and multidisciplinary character of the design process becomes more relevant to the definition of its cost and performance. The assessment of different solutions for the mechanical and electrical architecture includes the positioning of the spacecraft’s equipment over its structure panels, aiming at satisfying mechanical and electrical requirements or constraints. A target position for the system’s mass center, preference of moment of inertia in a given direction, minimization of electromagnetic interference, avoidance of high heat dissipation due to equipment being positioned close to another, and minimization of cabling are examples of such concerns. The early positioning of the spacecraft’s equipment is usually done “manually” by a group of system engineers, heavily based on their experience. Coupled to an analysis stage, where the system’s performance and constraints are verified, the spacecraft’s equipment layout definition is an interactive process that takes time and hence, as soon as a good feasible design is found, it becomes the baseline. This reduces the exploration of the design space, and increases the probability that better designs are missed. Thus, increasing the creation of candidate solutions by numeric automatization of the

1. Instituto Nacional de Pesquisas Espaciais – São José dos Campos/SP – Brazil. Author for correspondence: Valentino Lau | Instituto Nacional de Pesquisas Espaciais | Avenida dos Astronautas, 1.758, CEP: 12.227-010 – São José dos Campos/SP – Brazil | Email: valentino.lau@inpe.br Received: 07/31/2014 | Accepted:10/17/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

432

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

search through the conceptual design space would increase the possibilities that better designs are found. The works of Ferebbe Jr. and Powers (1987) and Ferebbe Jr. and Allen (1991) are probably the firsts to propose numerical optimization methods for automating the process of determining the layout of equipment during the conceptual phase of spacecraft design. In a series of works, Teng et al. (2001), Sun and Teng (2003), Zhang et al. (2008) and Teng et al. (2010), studied the efficacy of the approach when applied to a spinning telecommunication satellite, considering also the influence of the application of different optimization methods. These works have in common the focus on placing the equipment driven by the system’s mass properties (position of mass center and magnitude and direction of principal axis of inertia) requirements, subject to geometric constraints. In Jackson and Norgard (2002), thermal issues and minimization of wiring between equipment were introduced as objectives to be considered in the search for candidate solutions in the design space. Thermal requirements are in fact one of the main drivers of the spacecraft layout design, and in the context of conceptual layout optimization they have been treated either by trying to meet requirements of equipment heat dissipation uniformity over the spacecraft’s structural panels (Jackson and Norgard, 2002; Hengeveld et al., 2011) or target temperatures on them (De Sousa et al., 2007). In the later work the problem was treated as fully multi-objective, that is, opposed to the usual approach of transforming it in mono-objective before optimization is performed, a set of trade-off solutions is the objective of the search. This provides more information about the design space, leaving for a posteriori analysis the choice of which solution will be implemented. Coupling optimization algorithms with Computer Aided Design (CAD) and engineering analysis packages, provides an efficient way to tackle the spacecraft equipment layout problem, as highlighted in the works of Baier and Pühlhofer (2003), Pühlhofer et al. (2004) and Cuco (2011). In the later one, a new methodology was proposed to address the problem. The Cuco’s methodology (Cuco, 2011; Cuco et al., 2014) considers the main drivers commonly used to define the equipment layout during the spacecraft’s conceptual design: • The position of the system’s center of mass; • The alignment and strength of the system’s main axis of inertia; • Avoidance of concentration of high heat dissipation equipment over the satellite panels; and • Equipment functional requirements.

The methodology of Cuco (2011), or different versions of it, may be implemented in different ways using commercial or custom made software. Cuco (2011) used modeFrontier® to couple Solidworks® , Matlab® , Excel® and an executable written in C, for such purpose. The advantage of using a software such as modeFrontier® as the core tool to implement the methodology is that, since it was specially developed to tackle optimization problems and act as an integrator of other CAD or Computer Aided Engineering (CAE) tools, it has readily available on its internal features different optimization algorithms and techniques to be used on the problem, and provides a user-friendly interface to integrate other tools and analyze the results. On the other hand, the user has limited or no access to changes on the workings of these tools, what may affect his/her ability to explore new ways of addressing the problem. In the context of a research tool for the exploration of different concepts and algorithms to address the spacecraft equipment layout optimization problem, Excel® would provide a convenient alternative, since it can be used as a platform where new optimization algorithms can be embodied, as a calculator for engineering analysis, data storage, visualization of results, as well as an integrator of CAD and CAE tools. It also has the advantage of being known and be available largely in the engineering community. In the present paper, an Excel® based tool for spacecraft equipment layout is presented. Built using Cuco’s (2011) methodology as the optimization framework, it can provide the spacecraft design team an efficient and easy way to explore the layout conceptual design space. Excel® was coupled to SolidWorks®, which is used to calculate design parameters and as a graphical interface, where candidate layout configurations can be visualized. The tool’s concept and early application tests were first presented in the 22th International Congress of Mechanical Engineering (COBEM 2013) (De Sousa et al., 2013). The present paper is an updated version of the former. It introduces new features such as integrated decision making criteria for selection of solutions on the approximate Pareto frontier, additional objective functions and three-dimensional (3D) capability. In the Sections that follow, the general spacecraft equipment layout optimization problem is formulated, a prototype of the layout tool is presented and a simplified 3D example of application is shown.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

SPACECRAFT EQUIPMENT LAYOUT PUT AS AN OPTIMIZATION PROBLEM The spacecraft’s layout problem can be tackled as a multidisciplinary multiobjective optimization problem and can be generally stated as: Minimize (1) Subject to: (2) (3) (4) where fi is a vector of I objective functions, xj is a vector of J design variables, gk and hl are vectors of K and L inequality and equality constraints, respectively, and x jmin and x jmax are the bottom and upper boundary constraints on the design variables. The objective functions encode the design requirements for the spacecraft, such as a target position for its mass center, whereas the constraints define the viable design space. For example, there must be no mechanical interference among the equipment. In the simplified case study showed further in the text these points will be made clear. The approach for the spacecraft equipment layout problem proposed by Cuco (2011) is used as the general framework to build the tool presented herein. The design variables are defined considering the faces of the panels where the equipment would be positioned and, over a given panel face, the local coordinate position of the equipment mass center being positioned on that panel. For a box-shaped equipment, the angle formed between the box edge and the panel axis is also a design variable. Hence, each equipment has 4 design variables: one index indicating in which panel’s face it is allocated, two local coordinates x and y of the position of the center of mass projected on the panel, and one orientation angle. For example, if there are 8 box-shaped equipment to be positioned and 2 panels available for positioning, there are 32 design variables for optimization. Design requirements

433

for the position of the system’s mass center, grouping of sets of equipment, avoidance of “hot spots” over the panels and, the alignment of the principal axis of inertia and the proportion of the principal moments of inertia are tackled by five objective functions, while geometric and functional requirements are take into account as constraints. In Cuco’s methodology, the result of the optimization is a set of candidate non-dominated solutions for the layout and its respective approximate Pareto frontier. The decision of which solution, or solutions, would be subject to a further analysis to become the baseline layout is left for the engineering team responsible for the layout design. The Cuco’s methodology framework embodies the basic aspects to be considered by any computational environment aimed at providing the system´s engineering team, a tool for the spacecraft´s equipment layout, during its conceptual design phase. It is flexible enough to accommodate design goals being treated either as objective functions or constraints. Adding to Cuco´s work, the tool presented herein also incorporates decision making criteria to help the design team choose one or more candidate solutions for further evaluation, after the approximates Pareto set and Pareto frontier are returned. For the general three-dimensional spacecraft layout problem, a practical tool must allow the exploration of any combination of equipment and positioning panel, as well as the possibility for the user to set a specific combination of equipment/panel. For example, it may be desirable that some equipment is positioned over a panel of the spacecraft with the least incidence of Solar thermal radiation. Moreover, functional aspects may require that some equipment is fixed in a given position, or that a set of them are positioned close to each other. All these features are implemented in the tool presented here. The main aspects considered for the conceptual spacecraft equipment layout are taken into account in the tool by 5 objective functions, which may be activated by the user independently. Mechanical interference between equipment is taken into account by constraints penalizing the objective functions, while parameterization of the design variables assures that the equipment remains inside the available positioning areas. The general optimization framework used for development of the tool is presented in Fig. 1. In the present version of the layout tool, the multi-objective optimization problem was posed as:

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

434

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

Input Data Geometric dimensions, mass and spatial coordinates of the spacecraft panels available for equipment positioning. Common sets of equipment, if applicable. Dimensions, mass, and heat dissipation of the equipment.

Optimization Loop Design Variables

Objective Functions and Constraints

Panel and Face for equipment positioning. Local coordinates of each equipment. Equipment orientation.

Objective functions for center of mass, heat density, common equipment set, alignment of principal axis of inertia and proportionallity of principal moments of inertia. Geometric and functional constraints.

Multiobjective Optimization Algorithm

Output Data Trade-off candidate solutions: Pareto set and Pareto frontier. Set of selected solutions from Pareto frontier, based on decision making criteria.

Figure 1. General optimization framework for the spacecraft equipment layout tool.

Minimize: (7) (5) (8)

(6) (9)

(6.1)

Subject to: (10) (11)

(6.2)

(12) (13)

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

→

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

f1 represents the goal of having the center of mass (CM) of the system, xi_CM_sys, as close as possible to a given target center of mass, x i_CM_target,. The parameters λ i, which may assume zero or one value, are used to disable or enable the CM coordinate components. For example, if the CM longitudinal component of an spacecraft is less constrained than the other components, then only the lateral components could be enabled to drive optimization. f2 is an object function devised to approximate the heat density over the spacecraft’s panels. This objective function is composed of two components. The first one, f2Global, measures how far the layout is from an ideal condition of uniformly heat distribution over the entire spacecraft. Npanel is the number of panels; Nequi,p is the number of equipment installed in panel p; Pi represents the heat dissipated by equipment i; Ap is the projected area of panel p; PTotal is the total heat dissipated; and ATotal is the total projected areas of the panels. The second Local component, f2,p , evaluates the heat dissipated by the equipment installed in panel p over discrete regions of this panel. The panel is divided in Ncell,p rectangular cells, each side of them with the size of rmin, which is the size of the smallest dimension, in contact with the panels, of all equipment divided by 2. ri,j is the distance between the center of the ith equipment to the center of jth cell, as seen in Fig. 2. Local Minimizing f2,p means that the standard deviation of the quantity is minimized, that is, the combined influence of all equipment over each panel cell would be the same.

14 Equipment 3

r 2,2

3 4 Equipment 1

53 Equipment 2

r 1,28

2 This would avoid “hot spots” over the panel. The ratio rmin /ATotal is applied for scaling compatibility of these two components. f3 represents the goal of minimizing the distance between equipment belonging to the same common set. We define here a common set, as a group of equipment that should be positioned near each other. di,j,k is the Euclidian distance between the geometric centers of equipments i and j belonging to a common set k, Nequi,k is the number of equipment in common set k and Nset is the number of common sets. f4 measures the alignment of the principal axis of inertia to the spacecraft global coordinate system; αi are the angles formed between the i-axis of the principal inertia and global coordinate systems, as shown in Fig. 3; and αi_target is a given target angle. Analogously to λi in f1, the parameters ρi, which are set to zero or one, are used to disable or enable angle components. The goal of f5 is to achieve a given proportionality between the principal moments of inertia. vinertia is a vector which components are the tree principal moments of inertia, and vtarget is a given target vector, which components are positive values, that keep a desired proportion. For example, in a spacecraft controlled by spin, the longitudinal moment of inertia should be larger than lateral ones, say n times, while lateral moments of inertia could be of the same order. Setting the longitudinal component of vtarget to n and the lateral components to 1 would represent this proportion. The vector norm of the cross product gives the area of the parallelogram formed by these two vectors. If they

ZGLOBAL ZPRINCIPAL

r 3,28 αZ

435

40 CM

YPRINCIPAL

αY YGLOBAL

XGLOBAL

Figure 2. Representation of how the distance between the equipment and the panel’s cells is considered in the heuristic used to calculate ∫2Local. Example with three equipment and 56 cells. Only distances for cell 28 are shown in the example, but all cells are considered when calculating the value of ∫2Local.

αX

XPRINCIPAL

Figure 3. Angles between principal axis of inertia and global axis.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

436

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

are aligned, what means that vinertia has the same proportion of vtarget, this area vanishes. On the other hand, if they are not aligned, a positive value is obtained. Normalization is used in order to keep f5 in the range of [0,1]. Tiface, si,l, si,2 and Iiθ are the design variables. The first defines the panel and the face where equipment i is installed. This is an integer variable corresponding to the index of an element in a list that contains all available panel faces, coded as the panel ID number with a signal, positive for top face and negative for bottom face. The number of available faces, Nface,i, can vary for each equipment i, since constraints may be applied to restrict equipment installation. The next two variables define the parameterized position of the geometric center of an equipment i over the panel. B, L and H are the equipment dimensions, while L1 and L2 are the panel dimensions, and θ is the equipment rotation angle, as shown in Fig. 4. The distances e1 and e2 are defined in Eq.14, and the relationship between the parameterized variables and the local coordinates D1 and D2 are presented in Eq.15. The values of the parametric variables si,1 and si,2 can vary in the range [0, 1]. This parameterization guaranties that the boundaries of the equipment always lies inside the area of the panel.

Iiθ is an integer variable used to evaluate the rotation angle θ as indicated in Eq.16. The number of increments N division,i is defined by the user. The angle θ varies in the range 0°≤θ≤180°.

(16)

Vinter is the total volume of mechanical interference among equipment and structure. The equality constraint (Eq. 13) is treated as a penalty for the objective functions when it is violated, using an exterior penalty method (Vanderplaats, 2007) approach.

DESCRIPTION OF THE SPACECRAFT EQUIPMENT LAYOUT CONCEPTUAL DESIGN TOOL The main components of the optimal layout tool are presented in Fig. 5. The Excel workbook consists of 7 worksheets and 3 main macros. From the Read Me worksheet, a description of all parameters used in this tool is presented. In the Control worksheet, parameters used to define and activate

(14)

(15)

L1 H

e2 CG

θ Y1

θ θ e1

X1 Z1 D1

Figure 4. Equipment position over the panel. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

L X2 X1

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

objective functions and constraints are entered. A specific geometric configuration, defined in the Panels and Equipment worksheets, is built in SolidWorks® , which is launched by clicking a macro button inside the Control worksheet. All design parameters calculated inside SolidWorks ® , are returned to the Control worksheet. In the Panels worksheet the geometric characteristics of each panel available for equipment positioning is entered. In the present version of the layout tool only rectangular panels are modeled. In the Equipment worksheet, the mechanical and thermal characteristics of the equipment are entered, together with the information of what subsystem they belong. In the current version of the layout tool, rectangular, cylinder and sphere solid shapes can be used to simulate the equipment. The solids may be assigned with different colors. In this worksheet, it can also be entered values for the design variables. In the Problem Description worksheet, a brief description of the objective functions, constraints and design variables being considered in the optimization problem is provided. In the Choice of Optimizer worksheet, the optimization algorithm to be used is chosen and information concerning its operational, such as parameters and stopping criteria, is entered. The optimization process is initialized from this worksheet, by clicking a macro button representing an available optimization algorithm. This calls a routine that embodies the algorithm and links it to other routines that launch and control SolidWorks ® . Finally, in the Results worksheet, the approximate Pareto set and Pareto frontier obtained during the search are presented. Different types of graphs available in Excel® may be used in order to show

the approximate Pareto frontier. For example, for problems with three objective functions, bubble or surface graphs may be used. In Fig. 6, screen prints of the seven worksheets are presented for illustration purposes. The macros for the optimization algorithms, objective

functions and routines that link Excel® to SolidWorks® are built using the VBA editor, in a modular approach, such that new

optimization algorithms or objective functions can be added or removed from the tool, as desired. In its present version, only a real coded implementation of the M-GEO optimization algorithm (Galski, 2006), was incorporated to the layout tool.

In Fig. 7 screen prints of the VBA editor and the SolidWorks® environment are shown.

The layout optimization process embodied in the layout tool just described is fully automatic. That is, once the “button” linked to an optimization algorithm is clicked in the Choice of Optimizer worksheet (for example, Play M-GEO in Fig. 3), the information on the Panels and Equipment worksheets are

accessed, SolidWorks® is launched and linked to Excel® , the optimization performed and the results sent to the Results

worksheet. The graph that plots the approximate Pareto frontier is also automatically updated. After the approximate Pareto frontier is retrieved, a particular layout solution may be visualized in SolidWorks® by selecting a solution ID and clicking in a macro button available in Result worksheet.

The automatic selection of particular solutions from the approximate Pareto frontier, based on dedicated decision making criteria that will be described ahead, is also calculated and listed in the Result worksheet.

Excel®

Worksheets: 1) Read Me 2) Control 3) Data Input for Panels 4) Data Input fot Equipment 5) Problem description 6) Choice Optimizer 7) Results

437

SolidWorks®

Visual Basic for Applications (VBA)

Calculate layout’s: 1) Center of Mass 2) Principal Moments of Inertia 3) Principal Axis of Inertia 4) Mechanical Interference Volume Visualizations of Equipment Layout Configuration

Figure 5. Main components of the spacecraft equipment layout tool. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

438

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

Figure 6. Screenshots of the layout tool Excel® worksheets.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

439

Figure 7. Screenshots of the VBA editor showing the M-GEO macro (left view) and SolidWorks® environment (right view).

THE OPTIMIZATION TOOL The optimization algorithm implemented in the layout tool so far is based in M-GEO (Galski, 2006), a multiobjective version of GEO evolutionary algorithm (De Sousa et al., 2003; De Sousa, 2002). In an early version of the layout tool, the canonical M-GEO was used (De Sousa et al., 2013). As in the original GEO, in the canonical M-GEO the design variables are codified in binary strings. However, it has been shown that for problems where the design variables are continuous, a real coded GEO may perform better than its canonical version (Mainenti-Lopes et al., 2008), what was also verified with real coded versions of M-GEO (Mainenti-Lopes et al., 2012, Mainenti-Lopes, 2013). It has also been showed that GEO can work successfully treating discrete variables directly (De Sousa and Takahashi, 2005). Because in the spacecraft equipment layout problem there is a mix of discrete (Iifaceand Iiθ) and continuous (si,1, si,2) design variables, was decided for the present version of the layout tool, to implement the M-GEO using the variables directly, instead of codifying them in binary strings. The main steps of the M-GEO algorithm as implemented in this work is described in Fig. 8. Because the number of non-dominated solutions found during the optimization search can become very large, the user of the layout tool can set the maximum number of non-dominated solutions desired to be stored in the computer’s memory and retrieved at the end of the search. Each time this number is exceeded, the “crowded distance” strategy proposed by (Deb et al., 2000) is used to select the point on the approximate Pareto frontier that is on its most crowded region, and it is removed from the solution set to be retrieved. For problems with a large number of non-dominated solutions, this approach helps the

user to keep the approximate Pareto set within a size more manageable for decision making analysis, while keeping on the solution set representative solutions of the entire approximate Pareto frontier. SELECTING CANDIDATE SOLUTIONS ON THE APPROXIMATE PARETO FRONTIER Though a multiobjective problem may be considered formally solved when the approximate Pareto set is found, from the practical point of view it is not over, since at least one of the non-dominated solutions has still to be choose to be implemented, or further investigated. Hence, some decision making criteria were included in the layout tool to help the designer in choosing solutions on the approximate Pareto Frontier (PF). Following The Smallest Loss Criterion, defined by Rocco et al. (2003) and used by Venditti et al. (2010) and Rocco et al. (2013), the solutions on the approximate Pareto frontier closest to its barycenter, calculated either considering all solutions on the frontier or only its edge values, and the utopian solution (the coordinates on the objective space that represents the optimal solution of each objective isolated), are used as references to choose solutions on the PF, as shown in Fig. 9. Since the edge solutions on the PF are the best solutions for each objective function, they are also candidate solutions to be further examined. In a problem with two objective functions, such as the hypothetical one shown in Fig. 9, there may be up to 5 solutions on the PF chosen by the criteria just outlined. It must be pointed out that the final choice of which solutions would be subject of further analysis and eventual implementation is always up to the designer. Automatic decision making strategies, such as the ones described above, should be used to help the decision making process and not as a substitute for the decision maker.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

440

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

Step 1. Initialize randomly the population of I design variables (species) and calculate the values of the J objective functions. Update the file of non-dominated solutions. Step 2. Calculate the values of the objective functions when, one at a time, the design variables are changed (mutated). With a random uniform pertubation for the discret variable and, for the continuous ones, with a Gaussian pertubation with zero mean and standard deviation σper equal to a given percentage of the variable’s design interval. Update the file of non-dominated solutions.

Step 3. Choose randomly one of the objective functions and set it as the reference. Step 4. For each design variable i attribute a “change fitness” CFi value equal to the value of the reference objective function choosen in Step 3, when the variable i is changed as in step 2. For minimization problems, sort of population of design variables in accordance to the value of CFi, such that the variable with least CF receives index ki=1, and the one with the highest, index ki=I. For maximization problems the sorting is done conversally. Step 5. Mutate one variable with probability Pi≈ki–.

Step 7. Initialize again population?

Step 6. Stopping criterion reached?

Yes

Yes Step 8. return the Pareto set and Pareto frontier.

Figure 8. Main steps of M-GEO multiobjective optimization algorithm as implemented in this work.

EXAMPLE OF APPLICATION A simplified three dimensional (3D) example is used for illustration of the tools features. It consists of placing 8 typical spacecraft equipment belonging to three different “common sets”, over two squared panels, each one with an area of 1 m 2. Only the panels’ top faces were selected as available for equipment installation. The equipment positions were defined using a total of 32 design variables. Optimization was performed using two different sets of objective functions. In the first run, the selected objective functions are the heat density (f2) and the common set distance (f3). In the second run, the same objective functions previously used are selected, and one additional objective function, the center of mass (f1), is included. The chosen target center of mass is located 0.3 m far from the panels’ top faces, with a height of 0.5 m from the lower edge of these panels. Therefore, two approximate Pareto

Non-dominated solution on the edge of the PF Barycenter of the PF considering all nonBarycenter of the PF domintaed solutions considering only the nondominated solutions on the edges of the PF Solution on the PF closest to the Solution on the PF closest to frontier’s barycenter the frontier’s barycentes (considering all (considering all solutions on solutions on the PF.) the PF). Solution on the PF closest Non-dominated to the utopian solution solution on the edge of the PF Utopian solution

Figure 9. Some criteria to select solutions on the approximate Pareto Frontier (PF) for further analysis. A hypothetical example with two objective functions is presented here. Circles are non-dominated solutions. Crosses are reference marks based on the criteria (see text).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

frontiers are calculated, one with two objective functions and other with three objective functions.

441

evident the convenience of using decision making techniques to help the design team select candidate solutions for further analysis. Using the Smallest Loss criterion described above and the edge solutions, five non-dominated solutions were

Common Set Distan ce F3 (m)

RESULTS OF SIMPLIFIED 3D CASE STUDY The values used for the geometric dimensions, mass, heat dissipation and common set for each equipment is shown in Table 1. The M-GEO algorithm was used for the optimizations, starting with initial configurations randomly generated. The number of model updates was selected as the stopping criterion. All the selected objective functions are evaluated in each model update. In both runs, a total of 500,000 model updates were evaluated, corresponding to 15,624 generations in M-GEO. The deterministic parameter τ was set to 20, the variable standard perturbation parameter σperc was set to 5%, and 5 re-initializations were used during optimizations. The stored non-dominated solutions were limited to a maximum of 100 solutions. In the first run, with two objective functions (f2, f3), 89 nondominated feasible solutions were recovered at the end of this search. The obtained approximate Pareto frontier is shown in Fig. 10. The large number of non-dominated solutions makes

4 F2

3.5 3

2.5

Ball

1.5

Ut 5

Bnd 9

F3 21

Heat Density - F2 (W/m2)

Figure 10. Optimization with two objective functions (F2, F3) Approximate Pareto frontier found using M-GEO.

Table 1. Geometric, mass, power and common set data of the equipment. Dimensions

Equipment

Mass (kg)

Heat Dissipation (W)

LX (m)

LY (m)

LZ (m)

Battery 1 Battery 2 Battery 3 Diplexer 1 Diplexer 2 PCDU Transponder 1 Transponder 2

4.000 4.000 4.000 0.750 0.750 18.800 2.800 2.800

4.71 4.71 4.71 1.30 1.30 44.43 29.20 29.20

0.166 0.166 0.166 0.156 0.156 0.45 0.156 0.156

0.229 0.229 0.229 0.21 0.21 0.265 0.21 0.21

0.095 0.095 0.095 0.025 0.025 0.225 0.094 0.094

Common Set

1 1 1 2 3 1 2 3

Table 2. Optimization using two objective functions - Selected solutions on the approximate Pareto frontier. Index of solution on the approximate Pareto frontier

f2 (W/m2)

f3 (m)

6.7826

3.7427

20.6329

1.5931

Bnd

11.5459

1.9071

Ball

6.9032

2.3249

7.1172

1.9220

Selection Criterion

Best value obtained for the thermal uniformity (f2) objective function. Best value obtained for the distance between equipment of the same common set (f3) objective function. Non-dominated solution closest to the barycenter calculated only considering the edges of the approximate Pareto frontier. Non-dominated solution closest to the barycenter obtained considering all solutions on the approximate Pareto frontier. Non-dominated solution closest to the utopian solution.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

442

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

picked from the frontier. They are presented in Table 2 and shown in Fig. 10. In Fig. 11, colors were used to distinguish equipment of common sets: green for set 1, blue to set 2, and orange for set 3. In the second run, with three objective functions (f1, f2, f3), 100 non-dominated feasible solutions were recovered. Figure 12 shows a plot of all these solutions. A colored scale was used to represent the f1 objective function. Table 3

presents the six selected solutions and Fig. 13 shows their layout configurations. The results shown on Figs. 10 and 12 shows clearly the capacity of the layout tool generate a great number of feasible non-dominated solutions for a 3D problem, starting from a completely random configuration. The two objective functions in the first optimization run are naturally competitive: the heat density f2 drives the layout to a spread equipment configuration to avoid “hot” spots, while the common set distance f3 drives

Best solution fordensity equipment Best solution for heat density (F2). Best solution for heat (F2). common set distance (F3)

Best solution for equipment common set distance (F3)

Best solution for heat (F2). Best solution fordensity equipment Best solution for heat density (F2). common set distance (F3)

Best solution for equipment common set distance (F3)

Best solution for heat density (F2).

Best solution for equipment common set distance (F3)

closest to barycenter Solution closest to barycenter Solution Solution closest to barycenter Solution closest to barycenter considering extreme considering extreme solutions (Bnd) considering allsolutions solutions(Bnd) (Ball) considering all solutions (Ball) closest to barycenter Solution closest to barycenter Solution Solution closest to barycenter Solution closest to barycenter considering extreme solutions considering extreme solutions (Bnd) considering all solutions(Bnd) (Ball) considering all solutions (Ball) Solution closest to barycenter considering extreme solutions (Bnd)

Solution closest to barycenter considering all solutions (Ball)

Solution Solution closest to utopian solution (Ut). closest to utopian solution (Ut). Solution Solution closest to utopian solution (Ut). closest to utopian solution (Ut).

Solution closest to utopian solution (Ut).

Figure 11. Optimization with Two Objective Functions - Layout configurations for selected solutions. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

equipment to dense clusters in order to reduce the distances between equipment. Plotting the results with these two functions as coordinate axis highlights this competitive behavior. While the mass center objective function may not conflict with heat density and the equipment common set ones, it drives the search towards equipment configuration which mass center

443

position is close to the desired one. Examining, in Figs. 11 and 13, the solutions selected using the automatic decision making criteria, it can be seen clearly that the edge criteria generate very different layout solutions, due to the fact that they represent trade-off solutions that privileges one of the objective functions. On the other hand, the solutions chosen using the barycenter

Common Set Distan ce - F3 (m)

7 Center of Mass DistanceF1 (m)

0.2971 0.2641 0.2311 0.1981 0.1651 0.1321 0.0991 0.0661 0.0331 0.0001

5 F2

Bnd Ball

3 Ut F3

2 1

Heat Density - F2 (W/m2)

Figure 12. Optimization with Three Objective Functions - Approximate Pareto frontier found using M-GEO.

Table 3. Optimization with Three Objective Functions - Selected solutions on the approximate Pareto frontier.

Selection Criterion

Index of solution on the approximate Pareto frontier

f1 (m)

f2 (W/m2)

f3 (m)

Best value obtained for the mass center (f1) objective function.

0.0001

20.6003

3.9522

Best value for the thermal uniformity (f2) objective function.

0.1981

6.7834

4.3506

Best value for the distance between equipment of the same common set (f3) objective function.

0.1457

8.5488

2.0540

Non-dominated solution closest to the barycenter calculated only considering the edges of the approximate Pareto frontier.

Bnd

0.0705

12.9192

3.7523

Non-dominated solution closest to the barycenter obtained considering all solutions on the approximate Pareto frontier.

Ball

0.0553

13.0616

3.8427

Non-dominated solution closest to the utopian solution.

0.0738

8.3236

2.4376

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

444

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

or utopian approach, are less dissimilar, but still provide a lot of information on alternative design solutions. Confirming what was observed previously for the two dimensional test example (De Sousa et al., 2013), it is noteworthy how the layout tool can provide potentially significant design gains. For example, in the present 3D application with three objective functions, a 38 % reduction on the value of objective function f3 is obtained

if solution Ut is chosen instead of solution F1. In a real design application this would mean a significant reduction on the cabling connecting the equipment, which can lead, for example, to cost savings and mitigation of integration problems. The processing time spent for running each optimization was approximately 3 hours and 32 minutes, in a PC with a Core i5 CPU, 2.5 GHz of clock and 4 GB of RAM memory.

solution heat density Best solution (F2). for heat density (F2). Best solution for center ofBest mass solution (F1). for Best center of massfor (F1).

Best solution heat density Best solution (F2). for heat density (F2). Best solution for center Best of mass solution (F1). for center of mass for (F1).

solution heat density Best solution (F2). for heat density (F2). Best solution for center ofBest mass solution (F1). for Best center of massfor (F1).

closest to barycenter Solution closest to barycenter Best solution for equipment Best solutionSolution for equipment considering extreme (Bnd) extreme solutions (Bnd) considering common set distance (F3)common set distance (F3) solutions closest to barycenter Solution closest to barycenter Best solution for equipment Best solution Solution for equipment solutions considering (Bnd)extreme solutions (Bnd) common set distance (F3) commonconsidering set distance extreme (F3) closest to barycenter Solution closest to barycenter Best solution for equipment Best solutionSolution for equipment considering extreme considering (Bnd) extreme solutions (Bnd) common set distance (F3)common set distance (F3) solutions

Solution closest to barycenter Solution closest to closest barycenter Solution to utopian solution (Ut). to utopian solution (Ut). Solution closest considering all solutions considering (Ball) all solutions (Ball) Solution closest to barycenter Solution closest to barycenter Solution closest to utopian Solution solution closest (Ut).to utopian solution (Ut). considering all solutionsconsidering (Ball) all solutions (Ball) Solution closest to barycenter Solution closest to closest barycenter Solution to utopian Solution solution closest (Ut). to utopian solution (Ut). considering all solutions considering (Ball) all solutions (Ball)

Figure 13. Optimization with Three Objective Functions - Layout configurations for selected solutions. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

A Multidisciplinary Design Optimization Tool for Spacecraft Equipment Layout Conception

CONCLUSIONS In this paper a tool for three dimensional multidisciplinary design conception of spacecraft equipment layout was presented. It is an evolution of an early prototype with 2D capability, which main features where presented in COBEM 2013 (De Sousa et al., 2013). The tool can be used either as a research bed for testing different candidate methodologies and optimization algorithms to the problem, as well as an operational tool to be used by an engineering design team. The choice of using Excel® as the main software platform over which the optimization tool is built, was based on the convenience of having a readily available and broadly known software, which could be easily used for data input, numerical calculations, output of results and integrator of CAD or CAE software. The tool uses Cuco’s multiobjective methodology (Cuco, 2011; Cuco et al., 2014) as the main framework for the layout optimization, which is performed by a customized implementation of the M-GEO (Galski, 2006) algorithm. The search for the optimal solutions, the approximate Pareto set, is performed from an initial completely random layout configuration. The user can select up to 5 different objective functions to guide the search. The user can also set which spacecraft panel’s faces are available for positioning a given set of equipment. Excel® was coupled to SolidWorks®, which is used to calculate design parameters and as a graphical interface,

445

where candidate layout configurations can be visualized. Results are automatically retrieved to a dedicated Excel® worksheet, becoming available to be further analyzed, either graphically or using internal Excel® features. The tool also embodies an automatic decision making procedure to select solutions on the approximate Pareto frontier, which, for a frontier with many non-dominated solutions, may help the user to decide which of them are more suitable to be further investigated. All these characteristics were exercised in a simplified three dimensional application example, which highlighted the potential benefits such a tool can provide. The Excel® based spacecraft equipment layout tool presented in this paper can be considered the first “operational” version of a tool which preliminary results were presented at COBEM 2013 (De Sousa et al., 2013). It was conceived to be continuously improved with new features, and short term goals in its development are the inclusion of new optimization algorithms and new objective functions to address additional engineering issues, as well as its application to a full real spacecraft layout problem. This would imply in a much larger design problem. For instance, for the service module of a middle size satellite of 500 kg, such as the MMP (Multi-Mission Platform) developed currently at INPE, the software would have to deal with around 88 design variables (for example, 22 equipment each one with four design variables). Moreover, there may be an increase in the number of constraints, depending on requirements posed on the positioning of some equipment.

REFERENCES Baier, H. and Pühlhofer, T., 2003, “Approaches for further rationalization in mechanical architecture and structural design of satellites”, In Proceedings of 54th International Astronautical Congress, Bremen, Germany. Cuco, A.P.C., 2011, “Development of a Multiobjective methodology for layout optimization of equipment in artificial satellites” (in Portuguese), Master dissertation, Postgraduate Course in Space Technology and Engineering, National Institute for Space Research (INPE). Cuco, A.P.C, De Sousa, F.L. and Silva Neto, A.J., 2014, “A multiobjective methodology for spacecraft equipment layouts”, Optimization and Engineering. doi:10.1007/s11081-014-9252. De Sousa, F.L., Muraoka, I. and Galski, R.L., 2007, “On the optimal positioning of electronic equipment in space platforms”, In Proceedings of the 19th International Congress of Mechanical Engineering, Brasilia, Brasil. De Sousa, F.L., 2002, “Otimização extrema generalizada: um novo algoritmo estocástico para o projeto ótimo”, (INPE-9564-TDI/836), Ph.D. Thesis in Computação Aplicada, Instituto Nacional de Pesquisas Espaciais, 142p.

De Sousa, F.L., Ramos, F.M., Paglione, P. and Girardi, R.M., 2003, “New stochastic algorithm for design optimization”, AIAA Journal, Vol. 41, No. 9, pp. 1808-1818. De Sousa, F.L. and Takahashi, W.K., 2005, “Generalized Extremal Optimization Applied to Three-Dimensional Truss Design”, Proceedings of the 18th International Congress of Mechanical Engineering (COBEM2005), CDROM, Ouro Preto, Brasil. De Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A., 2013, “A toll for multidisciplinary design conception of spacecraft equipment layout”, 22nd International Congress of Mechanical Engineering (COBEM, 2013), Ribeirão Preto, SP, Brazil. Deb, K., Agrawal, S., Pratap, A. and Meyarivan, T., 2000, “A Fast Elitist Nondominated Sorting Genetic Algorithm for multi-objective optimization: NSGA-II”. In: International Conference on Parallel Problem Solving From Nature, 6., Paris, France. Proceedings… Paris, France: Springer, pp. 849–858.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

446

Lau, V., Sousa, F.L., Galski, R.L., Rocco, E.M., Becceneri, J.C., Santos, W.A. and Sandri, S.A.

Galski, R. L., 2006, “Desenvolvimento de versões aprimoradas híbridas, paralela e multiobjetivo do método da otimização extrema generalizada e sua aplicação no projeto de sistemas espaciais”, (INPE-14795-TDI/1238), Ph.D. Thesis in Computação Aplicada, Instituto Nacional de Pesquisas Espaciais, São José dos Campos, Brazil, 279p. Hengeveld, D.W., Braun, J.E., Eckhard, A.G. and Williams, A.D., 2011, “Optimal placement of electronic components to mininize heat flux nonuniformities”, Journal of Spacecraft and Rockets, Vol. 48, No. 4, pp. 556-563. doi: 10.2514/1.47507. Jackson, B. and Norgard, J., 2002, “A stochastic optimization for determining spacecraft avionics box placement”, IEEE Aerospace Conference, Vol. 5, pp. 2373-2382. Ferebee Jr., M.J. and Powers, R.B., 1987 “Optimization of payload mass placement in a dual keel space station”, NASA Technical Memorandum 89051, March. Ferebee Jr. M.J. and Allen C.L., 1991. “Optimization of payload placement on arbitrary spacecraft”, Journal of Spacecraft and Rockets, Vol. 28, No. 5, pp. 612-614. doi: 10.2514/3.26288. Mainenti-Lopes, I., De Sousa, F.L. and Souza, L.C.G., 2008, “The Generalized Extremal Optimization With Real Codification”. Proceedings of International Conference on Engineering Optimization – EngOpt2008, Rio de Janeiro, pp. 01-05. Mainenti-Lopes, I, Souza, L.C.G. and De Sousa, F.L., 2012, “Design of a nonlinear controller for a rigid-flexible satellite using multiobjective Generalized Extremal Optimization with real codification”, Shock and Vibration, Vol. 19, No. 5, pp. 947-956. doi: 10.3233/ SAV-2012-0702. Mainenti-Lopes, I., 2013, “A Multiobjective Approach to the Optimization of Solar Sail Trajectories” (In Portuguese), Ph.D. Thesis, Pós-graduação em Engenharia e Tecnologia Espaciais, área Mecânica Espacial e Controle, INPE.

Pühlhofer, T., Langer, H., Baier, H. and Huber, M., 2004, “Multicriteria and discrete configuration and design optimization with applications for satellites”, In Proceedings of 10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Albany. Rocco, E.M., Souza, M.L.O. and Prado, A.F.B.A., 2003, “MultiObjective Optimization Applied to Satellite Constellations I: Formulation of the Smallest Loss Criterion”, Proceedings of the 54st International Astronautical Congress (IAC’03), Bremen, Germany. Rocco, E.M., Souza, M.L.O. and Prado, A.F.B.A., 2013, “Station Keeping of Costellations Using Multiobjective Strategies”, Mathematical Problems in Engineering, Hindawi Publishing Corporation, Vol. 2013, pp. 15. doi:10.1155/2013/476451. Sun, Z-G and Teng, H-F., 2003, “Optimal layout design of a satellite module”, Engineering Optimization, Vol. 35, No. 5, pp. 513-529. Teng, H-F, Sun, S-L, Liu, D-Q and Li, Y-Z., 2001, “Layout optimization for the objects located within a rotating vessel – a three-dimensional packing problem with behavioral constraints”, Computers and Operations Research, Vol. 28, pp. 521-535. Teng, H-F., Chen, Y., Zeng, W., Shi, Y-J. and Hu, Q-H., 2010, “A Dualsystem variable grain cooperative coevolutionary algorithm: satellitemodule layout design”, IEEE Transactions on Evolutionary Computation, Vol. 14, No. 3, pp. 438-455. Vanderplaats, G., 2007, “Multidiscipline Design Optimization”, Vanderplaats Research and Development Inc. ISBN 0-944956-04-1. Venditti, F.C.F., Rocco, E.M., Prado, A.F.B.A. and Suhkanov, A., 2010, “Gravity-assisted maneuvers applied in the multi-objective optimization of interplanetary trajectories”, Acta Astronautica, Elsevier Ltd., Vol. 67, No. 9–10, pp. 1255–1271. doi.:10.1016/j. actaastro.2010.06.022. Zhang B., Teng H-F. and Shi Y-J., 2008, “Layout optimization of satellite module using soft computing techniques”, Applied Soft Computing, Vol. 8, pp. 507-521.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.431-446, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.369

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study Arineiza C. Pinheiro1, Adenilso Simão1, Ana Maria Ambrosio2

ABSTRACT: The software in satellite applications has become increasingly larger, more complex and more integrated, so its verification and validation require exploration of new approaches. In this paper we present a Model-Based Testing (MBT) approach applied to the Communication Module of the ITASAT-1 university satellite. The models are Finite State Machines (FSM) representing the software behavior. In order to manage the difficulties to model the software behavior the approach employs the Conformance and Fault Injection (CoFI) testing methodology associated with the JPlavisFSM tool in the real context of a satellite’s critical software. The former advises the modularization of the modelling into different types of behavior into different FSMs, while the latter integrates several FSM-based methods to derive test cases, provides facilities to design and to check properties of the models and computes metrics. The main result of this case study was the evaluation of the drawbacks on the design of the testing models supported by CoFI and JPlavisFSM. The models, test sets, metrics with the application of our approach applied to the Communication Module are presented. The paper discusses the benefits as well as the points requiring new researches. KEYWORDS: Finite state machine, Model-based testing, Test-case generation methods, Testing methodology.

INTRODUCTION The increasing development of university satellites has allowed researchers to experiment new approaches in space area. In traditional satellite development, experimental methods are usually avoided to prevent risks and cost increase, whereas university satellites open an opportunity to explore new approaches using a real system (Alencar, 2013). The software for satellite applications has become increasingly more complex, as it includes more functions and is more integrated. The development of satellite-related software usually requires a series of rigorous tests, along with all the satellite development phases. The development tendency of this kind of software points out the use of formal models for the designing and testing. To adopt formal models, such as the Finite State Machines (FSMs), not only does it benefit the identification of ambiguities and gaps in the requirements (Morais and Ambrosio, 2010; Morais, 2011; Pontes et al., 2012), but it also makes the automatic generation of test cases feasible (Ambrosio et al., 2005; Romero et al., 2012) by applying the vast theory of automatic test case generation. The specification of test cases for embedded software in satellites should take into account characteristics such as: real-time requirements, integration of technologies and fault tolerance mechanisms. The occurrence of failures in this kind of software may cause large losses, so they should be thoroughly tested using systematic and rigorous approaches, in which

1. Instituto de Ciências Matemáticas e de Computação/USP – São Carlos/SP – Brazil 2.Instituto Nacional de Pesquisas Espaciais – São José dos Campos/SP – Brazil. Author for correspondence: Ana Maria Ambrosio | Instituto Nacional de Pesquisas Espaciais | Avenida dos Astronautas, 1.758, CEP: 12.227-010 –São José dos Campos/SP | Brazil | Email: ana.ambrosio@inpe.br Received: 04/15/2014 | Accepted: 10/20/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

448

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

various testing techniques are combined to exercise different aspects of the system. The Model-based Testing (MBT) is an approach that reduces cost with the automation of test cases generation and with the reuse of models created during the software designing and testing activities; besides that, it provides high fault detection and traceability. In MBT, the test designer can specify the test model using different modeling notations, such as FSMs, Labeled Transition Systems (LTS), Unified Modeling Language (UML) state machines, etc. FSM is a formal modeling technique, adopted due to its rigor and simplicity. FSM-based testing has been studied for several decades (Moore, 1956; Chow, 1978) and it has still presented contributions (Simão et al., 2009; Simão and Petrenko, 2010; Yin et al., 2010; Pedrosa and Moura, 2010; Hierons and Ural, 2010; Simão et al., 2005) such as a tool that integrates different FSM-based methods into a single software product named PLAVIS (PLAtform for Software Validation & Integration on Space Systems); similar work is given in Santiago et al. (2008). In MBT, supporting tool plays an essential role, since the cost of building the models for testing only should be balanced by the possibility of generating and executing larger test suites (Utting and Legeard, 2007). However, the difficulties still remain to build the models. In our work, we have applied the Conformance and Fault Injection (CoFI) methodology (Ambrosio, 2005) in order to guide the tester to create the models associated with the JPlavisFSM tool (Pinheiro, 2012). CoFI was firstly proposed to lead a tester to understand the problem well enough while creating a set of FSMs representing normal and abnormal behavior of software on-board a satellite. It defines detailed steps to apply FSM-based testing tools, in a systematic way and it has been applied in several cases (Ambrosio et al., 2007; Pontes et al., 2009; Anjos et al., 2011; Mattiello-Francisco et al., 2013). JPlavisFSM tool comprises not only four methods to automatically generate test cases starting from a given specification in FSM but also the Mutation Testing technique (Fabbri et al., 1994), that allows the adequacy evaluation of a test cases set. This paper presents a case study in which the CoFI methodology and the JPlavisFSM tool are combined to test the Communication Module of the ITASAT-1 satellite (Sato et al., 2011). All the steps applied to design the models have been illustrated. We employed the JPlavisFSM tool to the test case generation, exploring all the FSM-based methods available. The main result of this case study was the

identification of the drawbacks to create the testing models that is the key activity to successfully apply MBT techniques to satellite-related software. On the modelling work, we found that the previous knowledge on the testing methods theory had facilitated the modelling process, which took 16 days only. The length of time to understand the system under test, to do the modeling and to generate the test cases accounted for 33 days. We did not evaluate how the models impacted on the quality of the tests, as the tests were not executed against the implementation of the Communication Module software, because the implementation was not finished before the conclusion of this work. However, we evaluated the quality of the test sets by mutation analysis applied to FSM. The remainder of this paper presents the background, the CoFI and the JPlavisFSM tool, the case study, and finally, discusses the MBT applicability and the conclusions.

BACKGROUND In this section, we introduce the main concepts necessary for understanding the results presented in this paper. FINITE STATE MACHINES Finite State Machines (FSMs) have been widely used for modeling reactive systems, ranging from simple protocols to complex embedded systems (Lai and Leung, 1995). Among the main advantages of FSMs, there are the solid theoretical background, the expressiveness power and the existence of numerous methods to generate test cases. FSMs are hypothetical machines composed of states and events, which correspond to transitions between the states. A transition is associated to two kinds of events: input and output. When an input event occurs in a given state, the FSM responds with an output event and may move to another state (Gill, 1962). Formally, a Mealy FSM is defined as a tuple M = (S, s0, X, Y, D, λ, δ), where: S is the nonempty finite set of states; s0 is the initial state (s0 ∈S); X is the nonempty finite set of input symbols; Y is the nonempty finite set of output symbols; D ⊆ (S × X) is the specification domain; λ : D → Y is the output function; and δ : D → S is the transition function. Figure 1 shows an FSM example with 4 states and 8 transitions, where s0 is the initial state, S = (s0, s1, s2, s3), X = (a, b) and Y = (0, 1).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

449

• Completely specified – an FSM is completely speciﬁed, a/0

a/1

b/0

a/ b/1

b/0

a/0

b/1

•

Figure 1. Example of FSM.

Given state s and input x, the transition (s, x) is defined if (s, x) ∈D. An input sequence α = x1x2…xk is defined at state s if there exists a sequence of states s1, s2, …, sk+1, so that s = s1 and (si, xi) ∈D, δ (si, xi) = si+1, for 1 ≤ i ≤ k. The set of input sequence defined at state s is denoted ΩM(si). The output and transition functions are extended to defined input sequences in the usual way; given a state s and the empty sequence Є, we have that δ (s, Є) = s and λ (s, Є) = Є; given input sequence α and input x, we have that δ (s, αx) = δ (δ (s, α), x) and λ (s, αx) = λ(s, α)λ(δ (s, α), x). A set of input sequences is initialized if it contains the empty sequence. An input sequence α separates states s and s’, if λ(s, α) ≠ λ(s’, α). A set of sequences is an identifier for state s, denoted Id(s), if for each state s’ ≠ s, there exists a sequence in Id(s) which separates s and s’. A set containing one identifier for each state is harmonized, if for each pair of state s and s’, there exists an input sequence α ∈Id(s) ∩ Id(s’) which separates them. An input sequence α is a Unique Input/Output Sequence for state s, denoted UIO(s), if {α} is an identifier for s. A set of input sequence W is a characterization set for each pair of state s and s’, if there exists an input sequence in W which separates them. A preamble for state s is an input sequence α, so that δ(s0 , α) = s, i.e., it takes the FSM from the initial state to s. A state cover is a set containing preambles for each state. A transition cover is a set P of input sequences, so that, for each defined transition (s, x), there are α and αx in P, where α is a preamble for s. The FSM M is deﬁned over X. Based on this deﬁnition, some structural properties in FSM, which are important for test case generation methods, are listed:

•

or complete if all states (S) have a transition for each input event from the set of input symbols (X), so that D = (S × X). Otherwise, the FSM is partially speciﬁed, or partial; Strongly connected – an FSM is strongly connected if for every pair of states (si, sj) there is a path that leads si to sj, i.e., there is some input event sequence that performs a path of transitions with source in si and destination of sj. If all other states can be reached from the initial state, the FSM is initially connected; Deterministic – an FSM is deterministic when there is only one transition with a particular input event from any state that allows transition to a next state. Otherwise, the FSM is nondeterministic; Equivalent – one state si is considered equivalent to sj if there is no input event sequence which, when executed from the respective states, generates a different output sequence; Reduced – an FSM is reduced if there is no pair of equivalent states. Otherwise, it is unreduced.

FINITE STATE MACHINE (FSM)-BASED TESTING In the context of FSM-based testing, a finite sequence of input events is a test case, or just a test. A set of test cases is a finite set of sequences, so that there are no two sequences α and β ∈T, where α is a proper prefix of β. (There are two sequences α and β, α is prefix of β, if α ≤ β, such as αω = β, to any ω. And, α is a proper prefix of β, if α < β, such as αω = β). Given two FSMs M and I; two states s from M and t from I are distinguishable if there is an input sequence γ ∈ΩM(s) ∩ ΩI(t), called separating sequence, so that the same input event sequence generates different output sequences for both FSMs, i.e., λ (s, γ) ≠ λ (t, γ). Two FSMs I and M are distinguishable if their respective initial states are distinguishable. A fault domain Γ(X) is the set of all possible implementations of M defined over the set of input symbols X. Similarly, Γn(X) denotes the set of all FSM defined over the set of input symbols X with at most n states. The test set T is said n-complete, or just complete, for the specification M if for all I ∈Γn(X), so that I is distinguishable from M, there is at least one sequence α ∈T that produces different output sequences when applied to M and I in the respective initial states. In other words, a test set is complete when it is possible to distinguish the specified FSM with n states from all other distinguishable FSM with the same set of input symbols and at most n states.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

450

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

The main objective of the FSM-based testing is to compare a reduced FSM model M with n states to an implementation of the FSM model, which is assumed to be represented by an unknown FSM I. When a reduced FSM M which represents the correct version of the speciﬁed FSM is compared to FSM I, the following faults may be revealed (Chow, 1978): • Output Error: I and M differ in the output of a transition; • Transference Error: I and M differ in the ﬁnal state. When the final state achieved in both FSM is different after applying a test case; • Transition Error: it is the general term for output or transference error; • Missing States: I has fewer states than M; • Extra States: I has more states than M.

T against P. If a fault occurs, a defect was found. Otherwise, it is assumed that some defects already exist but cannot even be detected. The next step is to derive one or more products from the original artefact P, which gives the new products P1, P2, ..., Pn, also called mutants of P. Mutants are generated from mutation operators, which are dependent on the programming language and determine what types of syntactic changes can be made i n the artefact. The mutation operators’ aim: • Inducing simple syntactic changes based on the typical mistakes made by developers, such as changing the value of a constant, or • Forcing certain test goals, such as performing each node or arc of the product (Delamaro et al., 2007).

Several methods have been proposed for generating set of test cases which guarantee that the implementation does not contain any of the above listed faults, that is, which are complete regarding these faults. The completeness of the generated test case is proved by showing that no faulty implementation FSM would pass the test cases. The main differences among them are the cost to generate the sequences and the effectiveness (the power of the test cases to detect faults). One of the first methods to be proposed was the W (Chow, 1978), which is considered a precursor of the area, since most of the following methods are based on it: UIO (Sabnani and Dahbura, 1988), UIOv (Vuong et al., 1989), Wp (Fujiwara et al., 1991), HSI (Petrenko et al., 1993) and SPY (Simão et al., 2009).

The test set T is executed against each of the mutants generated. The main objective is to “kill” all mutants. It is expected that, due to the changes made, the mutants have different behavior from the original product, featuring a defect. Occasionally, when the artifact is a source code, a mutant and the original product can still have the same result for any test case in the set. In these cases, the tester has to check the code and determine whether there is equivalence between the products. If so, the mutant is called as equivalent. The equivalence between programs is an undecidable problem and, therefore, there is no automated solution to the problem and the manual intervention of the tester is necessary. However, when the product is a model, there is no equivalency problem, because it could be determined automatically due to the formality of the FSM. After the mutants’ execution and equivalence analysis, the mutation score is calculated. The aim is to determine the adequacy of test cases used in a range from 0 to 1, providing a quantitative measure of how efficient the test set is in revealing the difference between the mutant and the original product. The mutation score ms(P, T), to product P and the set of test cases T is calculated as:

MUTATION TESTING The Mutants Analysis is one of the most popular criteria of Defects-based testing, which aims at deriving test requirements from knowledge about typical mistakes made by designers or developers. Mutation Testing is a testing technique that can be widely employed as a way of evaluating the test cases generated. The criteria can be used as: • Black-box testing, when we consider the specification as a test artifact, or • White-box testing, when the artefact considered is the source code. In this context, the test artifact could be represented formally by a model, as FSM. The Mutants Analysis consists of assessing the adequacy of a test set T for the test artifact P. First, you run the test set

ms(P,T)=

DM(P,T) M(P)-EM(P)

where: DM (P, T): number of mutants killed by T; M (P): total number of mutants generated from P, and EM (P): number of generated mutants that are equivalent to P (Delamaro et al., 2007).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

The mutation score is obtained from the ratio between the number of mutants killed by the set T and the number of mutants that could be killed, given by the difference between the total number of mutants generated and the number of mutants classified as equivalent.

a/0

This section presents the main features of the JPlavisFSM tool and the CoFI methodology which were combined to guide the creation of FSMs, representing the system’s behavior. JPLAVISFSM TOOL In a joint cooperation project among Brazilian researchers, the test platform named PLAVIS was conceived. This platform integrated existing tools for automatic generation of test cases starting from FSM, which had been developed by PLAVIS-project’s members. PLAVIS was a web-based test platform providing different functions to support FSM-based testing (Simão et al., 2005). Years later, Pinheiro (2012) improved usability features in the previous platform, now named JPlavisFSM tool. The JPlavisFSM is a standalone software, whose graphical user interface (GUI) is imported from the open source tool JFlap (http://www.jflap.org/) to design the FSM. From the GUI, one can easily draw states and transitions, edit and adjust them before automatically generating a set of test cases as well as evaluating the adequacy of a set of test cases. The JPlavisFSM tool includes a function to analyze FSM structural properties namely complete, strongly connected, deterministic, etc. This kind of analysis is very important because the applicability of the test methods depends on presence or absence of some structural properties. For example, the W method requires a FSM with the properties of complete, strongly connected, deterministic, and reduced. One advantage is to reduce the need of theoretical knowledge, particularly for testers unfamiliar with such theory. Other functionalities were implemented to assist test case handling, such as test set execution, inclusion/exclusion and enabling/disabling. Test sessions can be created, as shown in Fig. 2. In this window, the user can visualize the FSM on the left and the generated test cases on the right side (each line has

b/0

a/ b/1

s1 a/0

a/1

b/0

b/1

THE JPLAVISFSM TOOL AND THE CONFORMANCE AND FAULT INJECTION METHODOLOGY

451

Figure 2. JPlavisFSM Tool.

one test case). At the bottom, there is an option that helps the tester to create particular test sets by himself. One of the best features of the JPlavisFSM’s test session is to support the construction of a test case set with a high probability of finding defects in the System Under Test (SUT). In a test session, the tester can import (or load) a set of test cases, which was automatically generated through one of the testing methods given by JPlavisFSM or manually generated by the tester. Once imported, the set of test cases can be executed (i.e., applied against the specification) and the mutation score obtained. The JPlavisFSM releases the following testing methods to generate test cases set: W, UIO, HSI and SPY. The W method generates test cases as follows. First, an initialized transition cover and a characterization set of the specification are determined. Secondly, for each sequence of the transition cover, each sequence of characterization set is appended. The UIO generates test cases as follows. First, an initialized transition cover is determined, as well as UIOs for each state. Secondly, for each sequence of the transition cover, the UIO of the state reached by that sequence is appended. HSI method generalizes the W method, and can be applied to partial FSMs. It first determines a transition cover and a set of harmonized state identifiers. Then, for each sequence of the transition cover, each of the sequence in the identifier for the state reached by that sequence is appended. The SPY method generalizes the HSI method; the main difference is that both the transition cover and the harmonized state identifiers are computed “on-the-fly”, i.e., during the execution of the method, trying to minimize the number of test cases which are required to obtain complete test cases. For the HSI and SPY methods two versions are provided: the original

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

452

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

one and the auto-completed; the latter auto-completes the FSM before applying the original method. New methods can be incorporated in JPlavisFSM as plugins or be imported at runtime (provided that the compatible input/output formats are used) as well as providing more flexibility to the tool and the possibility to explore all the other features for the test set analysis. Two different ways of evaluating a test cases set are provided: • Mutation Analysis; and • n-Completeness Analysis (Simão and Petrenko, 2010). The Mutation Analysis, provided by JPlavisFSM, allows a test set to be evaluated and also to be enhanced. This enhancement can be achieved by expanding the test set with test cases targeted at undetected mutations. The mutant operators (Fabbri et al., 1994) were designed according to the classes of faults detected on the FSM-based test models. When the user creates a test session, the mutants are automatically created and the user can access all the mutants to verify their statuses: alive, when the ongoing test fails to reveal the defect of the mutant; or dead, the other way around. This feature could help the user to upgrade test sets since this criterion guides the user to create tests that will run paths, which may contain defects. The n-Completeness analysis provides another way to verify the quality of a given test set, checking whether the test set is complete or not. COFI METHODOLOGY The CoFI methodology (Ambrosio, 2005) guides the functional testing activity. It was proposed considering the needs of space application’s validation and therefore defines steps for creating test cases for software embedded on-board of satellites in a systematic way in order to get as much reusability in testing as possible. The main objective is to help the tester to define FSM-based models which represent the behavior of the SUT. The SUT behavior’s decomposition into FSMs starts with the identification of services, which can be a system function from the user’s viewpoint. Each service shall be described by a set of FSMs that map different classes of behavior. This classification takes into account different kinds of input events: • Normal (absence of faults); • Exceptions (foreseen faults); • Unexpected inputs (inputs occurring when they are not expected); and • Hardware faults.

In short, the CoFI methodology consists of three steps: identification, modeling and test case generation. Figure 3 shows the three steps of CoFI methodology described hereafter. (A) Identification Based on a given specification, the tester has to understand properly the external behavior of the SUT, as in the blackbox testing. The information to accomplish this step can be extracted from textual specification documents, requirements, use-cases, sequence-diagrams, and also from interviewing experts in the SUT if necessary. In this step, it shall be identified: • Services that an user can recognize or execute in the SUT; • Events and actions: it is necessary to identify all the possible events and actions (or outputs) that can be, respectively, commanded and trigged (or observed). The selected events and actions will be abstracted as inputs and outputs in the FSMs; • Control and observation points: they are addresses or mechanisms to input data and to obtain responses from the SUT; • Physical faults: faults occurring in hardware. They are to be considered whenever the SUT includes embedded software in a hardware and this software must have mechanisms to treat the faults; • Facilities and constraints: the tools and theirs commands that supports the test execution against the SUT and the events that cannot be activated during the test execution. (B) Modeling In this step, for each service, the tester should develop four classes of behavior: • Normal; • Specified exceptions; • Sneak path; and • Fault tolerance.

IDENTIFICATION

S1 S2 ----

TESTE CASE GENERATION

MODELING

Figure 3. Main steps of the CoFI Methodology.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

TS1 TS2 ----

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

At least one FSM model shall be created representing the partial system behavior. The normal behavior model defines the sequence of events that the SUT normally expects when it is executed. The tester has to identify the normal mode the SUT will be subordinated to and the events and actions expected for this case. For the specified exceptions model the tester has to recognize the exceptions mentioned in the documents, as timeouts and unknown commands and thus define the events expected in this context, called exception events. The sneak path model deals with the correct inputs arriving in wrong time. The objective is to predict the unexpected behaviors, i.e., to describe what the machine should do if an input occurs in a state where it is not defined. The tester can use the normal model as a basis, write it in tabular form (event × state) and fill the blank cells with the corresponding expected behavior. In some cases, there is no possible behavior to model, and then the tester evaluates the best option for the SUT. The transitions trigged by the unexpected inputs should be designed based on the normal model. Finally, the fault tolerance model shall map the possible physical faults when the SUT implements mechanisms to tolerate hardware faults. For each physical fault, the tester has to adapt the normal model including the fault events. At the end of the modeling step, the tester has a set of models representing the SUT’s behavior, from which test cases may be automatically generated. (C) Test Case Generation This step can be supported by testing tools that automatically generate a set of test cases (also named test sequences). In the experiences about applying CoFI described in Ambrosio et al. (2007), Pontes et al. (2009), Anjos et al. (2011) and MattielloFrancisco et al. (2013), the Condado tool was used to automate the test cases generation. In our work, the JPlavisFSM tool was used because it provides five different testing methods, including some variations that can be applied in partial FSM. Moreover, it has facilities to analyze the models’ properties and the adequacy of test case sets.

CASE STUDY: ITASAT’S COMMUNICATION MODULE We now discuss the case study we have carried out applying CoFI methodology and JPlavisFSM tool in a MBT approach

453

for a real system. The SUT is the Communication Module’s software of the ITASAT-1 satellite. The ITASAT Mission (Sato et al., 2011) is part of a program funded by the Brazilian Space Agency (AEB) in the context of Action 4934 for Development and Launching of Small Technological Satellites developed by the Instituto Tecnológico de Aeronáutica (ITA) and other Brazilian universities. ITA is the responsible member for the project implementation and the Instituto Nacional de Pesquisa Espacial (INPE) provides technical consulting and laboratory infrastructure. The ITASAT-1 is a university satellite planned to execute experimental payloads namely, Digital Data Collection Transceiver, Heat Pipe Experiment, Micro Electrical Mechanical System for attitude determination and an Inter Satellite Link. The satellite architecture is composed of five service subsystems: Mechanical Structure (MSS), Thermal Control (TCS), Electric Power (EPS), Attitude Control and Data Handling (ACDH) and Telemetry & Telecommand (TMTC); the last three subsystems are shown in Fig. 4. The functions of the ACDH are commanded by the on-board computer (OBC). The Communication Module (CM), located in the ACDH subsystem, is in charge of the communication between the OBC and the TMTC. Figure 4 highlights the Communication Module and indicates that it is the System under Test (SUT), the focus of this study. The CM comprises not only two receivers and two transmitters in redundancy, but also the software in charge of receiving commands arriving from the ground stations and transmitting telemetry collected from satellite pieces of equipment to ground stations. The CM Software performs critical functions so it has to work properly for the space mission success. Its functions are to receive the telecommands from ground stations and to transmit the telemetries collected by the satellite’s payloads and sensors, to manage the payloads status and to turn off the pieces of equipment when a critical failure occurs leading the satellite into survival mode. Considering that the CM is one of the most critical parts of this satellite, it was chosen as case study to experimentally evaluate the proposed approach. The three main steps that guided this study are illustrated in Fig.5. (A) Identification The first activity was to study the Requirements Document of ITASAT-1 (Table 1 illustrates some of such requirements)

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

454

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

ACDH SUT

TMTC Receiver

S Band

Communication Module (CM)

Magnet ometer

3x Sun Sensor

CM Receiver

Diplex

CM Transmiter Transmiter

Hybrid

OBC Receiver

S Band

3x Air Coil

Communication Module (CM) CM Receiver

Diplex

P a y l o a d s

CM Transmiter Transmiter

EPS

Figure 4. ITASAT-1 satellite architecture.

CoFI Methodology

IDENTIFICATION CM Receiver and CM Transmitter

MODELING

TESTE CASE GENERATION

FSMs • Receiver • Transmitter

JPlavisFSM TS1 TS2

Figure 5. Development of the case study.

in order to understand the context and the behavior of the CM’s software. The studies lasted 12 days, five to understand the ITASAT-1 satellite documentation; five to the CoFI guidelines and two to identify the main elements required by CoFI, namely, services, events, actions, faults, and so on. We identified two services corresponding to the two main functions of the Communication Module: (S1) receive commands from ground and (S2) transmit the collected telemetries from on-board equipment. These functions are implemented respectively by the CM Receiver Software and by the CM Transmitter Software. The events and actions extracted from the requirements are those naming the transitions of the FSMs illustrated in Figs. 6 and 7. The events and actions are abstract representations, i.e., they

do not indicate a physical signal arriving. One example of a CM Receiver’s event is CmdOBCOK, that indicates the presence of one command to be executed. One action is StoreTC1wait, which indicates that the just-arrived telecommand shall be stored on board. Other examples are EndTimerB as event and StartTimerB as action to be performed by the CM Transmitter. For the CM Receiver Software, 12 events and 13 outputs were identified, while for the CM Transmitter Software, 19 events and 12 outputs were identified. Specified exceptions were identified in seven requirements, which are listed in Table 1. Concerning the physical faults, although they can occur, the Communication Module Software is not in charge of dealing with them, therefore, for testing purposes, the fault tolerance models were not created. (B) Modeling It was observed that on adopting the CoFI methodology’s guidelines the state explosion problem was avoided. First, we modelled the normal behavior of each service in the CM Receiver FSM (R_N) and the CM Transmitter FSM (T_N), as shown in Figs. 6 and 7, respectively. For the FSM models used here, namely the Mealy machine (Mealy, 1955), all the transitions should be represented by a pair of input-output, which means that each event should be

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

associated with an action. Since in this software test some events were not associated with an observed output (or action), the output “Terminate” was used in the model to indicate the end of a transition. Another modeling condiction was made up to represent one transition requiring the occurrence of two events or transitions with two outputs. Figure 6 illustrates the ACKandStartTimerC output, which, in practice, indicates two actions: “send an Acknoledge message” and “start the Timer C”. Regarding the specified exceptions, the FSMs R_Ex13, R_Ex2, R_Ex4 and R_Ex5 map the exceptions of the CM Receiver, and T_Ex1 and T_Ex2 map those of the CM Transmitter. Table 1 matches the FSMs to the ITASAT-1’s requirements they model. For sake of space, not all FSMs are shown in this paper. For more details see Pinheiro and Ambrosio (2013). The CM Software has to handle timers which trigger timeout events needed for the communication timing. On the first analysis, to handle timers seemed an issue, because classical FSM does not treat time events. Therefore, these events were abstracted away in the model. The CoFI

455

guides the definition of timers as external devices that are started by an action and produce an event to the FSM when the time is expired, indicating a timeout. For the normal behavior, the software only signalizes the start of a timer, as represented by StartTimerC and StartTimerA events. For the specified exception models, it was needed to represent events to indicate expired time, so we used the EndTimerC and EndTimerA events. The first CM Receiver’s FSMs were pointed out as non-deterministic by the JPlavisFSM, because the event INT2TC associated to StoreTC action occurred in several states. On further analysis of the real software’s behavior we defined two abstractions to solve this problem: • A finite buffer to store at most three TCs, called TC1, TC2 and TC3; and • Two actions to differentiate TC stored and TC sent. Then, to create several actions (outputs): StoreTC1wait, StoreTC1send, StoredTC2wait, and so on, instead of using only the StoreTC action.

Table 1. ITASAT-1 Requirements vs. Specified Exceptions Models. Specified exceptions #

Requirement

Description

FSM

FRq11152000-12

“When the OBC does not send a request in TC seconds, the CM receiver ﬁnalizes the communication routine”

R_Ex13

FRq11152000-13

“The request of the OBC shall be one of the following: the transmission of TC or TC error log from CM receiver to OBC, or the execution of a (correct) command from OBC.” Obs.: Terminate execution if the command is not OK.

R_Ex2

FRq11152000-16

“If the CPU does not send the acknowledge in TA seconds (value TBD), the CM receiver shall terminate the communication routine.”

R_Ex13

FRq11152000-27

“The CM receiver shall check the ﬁrst 16 bits, used for synchronization and packet addressing, and shall discard the TC packet if its address data does not match with the module address.”

R_Ex4

FRq11152000-29

“If the verification described in Frq11152000-2 (The CM receiver module shall check the mod the repeated data and EDAC field ) results in failure, the CM receiver shall store a TC error log (as specified in [AD3]) in the TC buffer”

R_Ex5

FRq11152000-52

“When any analog channel veriﬁed present unexpected results, the CM transmitter shall generate a pulse in the survival ﬂag pin (which is read by the CM receiver).

T_Ex1

FRq11152000-58

“When the OBC does not send a request in TC seconds, the CM transmitter ﬁnalizes the communication routine.”

T_Ex2

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

Wait1

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

ACK3

erA tTim Star and TC1

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

INT2TC / StoreTC2send

Buffer

Send2

NCPSpin / ACK andStartTimerC

C/ edT Stor

CmdOBCOK / ACK

INT2TC / StoreTC3wait Wait3 INT1SurvFlag / OffdeviceseBeaconON INT2TC / StoreTCFullwait INT2 DirectTC / Execute

NCPSpin / ACK andStartTimerC

erA tTim Star and TC2

Buffer

Send1

Wait2

INT2TC / StoreTC3entre

C/ edT Stor

CmdO BCOK / Term inate

ACK2

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

INT2TC / StoreTC1send

Send0

NCPSpin / ACK andStartTimerC

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

INT2TC / StoreTC2entre

A imer artT ndSt C3a C/T edT Stor

NCPSpin / ACK andStartTimerC

CmdOBCOK / ACK Store / NoTCmsg

ACK1

INT2TC / StoreTC2wait INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

CmdOBCOK / ACK

INT2TC / StoreTC1wait

Wait0

INT1SurvFlag / OffdeviceseBeaconON INT2TC / StoreTCFullwait INT2 DirectTC / Execute

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

CmdO BCOK / Term inate

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

CmdOBCOK / ACK

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

CmdO BCOK / Term inate

456

INT1SurvFlag / OffdeviceseBeaconON INT2 DirectTC / Execute

INT2TC / StoreTC3send

Buffer

Send3 Buffer

Figure 6. R_N: Finite State Machine representing the normal behavior of the CM Receiver.

EndTimerVand AnChVerOK / StartTimerV EndTimerB / Empty

/A CK

OB CO N Re qu es tF ro m CP U

PackSurvTM / SendToTrans

EndTimerB / Empty q4 q3

BeaconONandTMConfig / Config / AnChDtCollect

ct lle ico hD nC /A

Beacon OnandT MConfi g / Emp ty T M C Cm m / S OBCOF dC dO en F/ onfi BC dT OBCONR eqNotFro Empty gFr / S Ma q5 mCPU / om end ndA Empty OB Dt C C / To K K C A / C Tra P mCPU nsc U q2 estFro Co NRequ nfi O C B O g nsm Tra ty MTC p T / Em nfig Off Co con nfig / a e B tCo no eTM N O n co Bea

erB im dT En

BeaconONandTMnotConfig / ConfigTMTCTransm BeaconOFF / empty

OBCONResqNotFromCPU / Empty OBCOFF / Empty

EndTimerBandDB / StartTimerB

Figure 7. T_N: Finite State Machine representing the normal behavior of the CM Transmitter. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

After concluding the normal behavior modeling, we analysed all events presented in the normal model, in order to create the sneak path models. However, because of the logic implemented in the hardware design construction, events never occur in an inappropriate state, so sneak path models were not designed. The fault tolerant models were not designed as well, because the CM is not in charge of dealing with the hardware faults. In the end of the modeling step, we had eight FSMs, as shown in Table 2. The JPlavisFSM indicated incompleteness for all FSMs, consequently the traditional W method could not be applied. The completeness is a difficult property to be achieved when modelling the behavior of real systems, that is why the JPlavisFSM tool provides other methods, as HSI and SPY that do not require this propriety. Summarizing, the modeling steps accounted for 16 days, being 8 days to create the normal and specified exception models including reviews and re-work in order to adequate the models to precisely represent the software behavior; and generate the test cases through the methods. Three adequacies were considered: • Adequacy to the system – Do models represent the system? • Adequacy to the FSM concepts and generation methods – How can one abstract the model to achieve the FSM needs? • Adequacy to the final model – Have the final models achieved the system objectives and the methods needs? (C) Test Case Generation We applied all methods for the eight FSMs. The W method’s applicability is restricted to complete, deterministic and initially connected FSMs. As the FSMs were not complete we used the JPlavisFSM’s function to auto-complete it before using the W method. The auto-complete functionality adds Table 2. Summary of models design. FSM models by classes of behavior Services

Normal Behavior

Exception Specified

Sneak Path

Fault Tolerance

CM Receiver

CM Transmitter

457

all the missing transitions in a given state as self-loops, i.e., transition that starts and ends in the same state. Althought the auto-complete option leads to the generation of many test cases, which are not significant in practice, it allows the applicability of the W. One test case generated by W starting from the CM Receiver FSM normal behavior (Fig. 6) is given by the following sequence of inputs INT2TC, StoredTC, NCPSpin, and its corresponding expected output is StoreTC1wait,TC1andStartTimerA,ACKandStartTimerC. The UIO method requires a strongly connected FSM, whereas the SPY and the HSI can be applied into a non-completed FSM. Thus, these methods have a wider applicability in comparison to the others; moreover, the test set generated by them is smaller than those genereted by W and UIO methods. On the other hand, it is possible to apply the UIO Method because it requires a separation sequence defined for each state. One example of a test case generated by UIO method starting from the CM Transmitter FSM is BeaconONandTMnotConfig, OBCONRequestFromCPU, CmdOBC, EndTimerVAndAnChVerOK. The HSI and SPY are based on separating sequences by pair of states (see Background section). Thus, each pair of states must have a transition with an event in common with distinct outputs. In the FSMs of the CM Transmitter software no event occurs in more than one distinct state, so these methods were applied neither to T_N nor to T_EX1 and T_EX2. One test case generated by the SPY method is INT2TC, NCPSpin, StoredTC, INT2DirectTC, INT2TC, INT2DirectTC, INT2TC which starts with the same input as the example to W method, but explores a valid sequence of events. In general, all methods of the JPlavisFSM were applied to the created models. Table 3 shows the test sets generated by all methods for each FSM. The ‘-’ symbol indicates the method could not be applied and ‘*’ symbol indicates the use of auto-complete functionality. Particularlly, the original W method did not generate any test case for the non-complete FSM, so there is no column for it. The original W was executed only thanks to the auto-complete feature. This option increases the number of transitions, leading to a large number of test cases. The UIO method generated smaller sets, but the method does not guarantee the generation of complete sets. The HSI and SPY were successfully applied to CM Receiver FSMs because there was a separating sequence that distinguished their states. For CM Transmitter FSMs, it was not possible

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

458

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

to use the HSI and SPY methods, therefore using the autocomplete feature they could be applied. The auto-complete feature was executed in CM Receiver FSMs only to show the increase in the number of test cases, caused by the extra transitions artificially included in the FSM to enable the application of the methods. We have also explored the JPlavisFSM’s functionality to calculate the minimal test set composed by test cases produced by all the methods. After creating a test set comprising all test cases, all the prefixed test-cases are discarded and the minimal test set is generated by the tool. The test sets shown in the HSI* and SPY* columns were not considered, since the original methods, HIS and SPY, could be applied. Table 4 illustrates the number of test cases derived from each model (considering the total set and the minimal set), the number of mutants created to each model and the mutation score obtained with both sets of test case. The mutation scores for the minimal and the total sets were the same, since only prefixes are removed from the total set. The number of test cases (total and minimal sets) includes the tests generated by W*, UIO, HIS and SPY, but it does not include HIS* and SPY* to FSMs from CM Receiver. All the methods (except UIO) generate complete test sets, but it is a theoretical concept which assumes that the implementation can be represented by an FSM with the same number of states of the specification. In general, to apply FSM-based testing implies handling a very huge number of test cases, so the tester must either be prepared to automate the test execution or select and prioritize the test cases. In our approach, we suggest a test strategy that could be adopted: first is to select the smallest test set generated by original methods; then, if there is still enough time to the testing activity, other sets could be applyed to improve the system’s reliability. In this case study, the SPY and UIO methods should be selected for CM Receiver FSMs and CM Transmitter FSMs, because the Mutation Score of these sets were 1.0 for each one, and they are complete sets as well. The inputs and outputs represented in the models are abstracted to enable the creation of FSMs, this fact led to the generation of abstract test suites, so a post-processing of the test sets will be necessary before the test execution. This allows the test cases to be designed before defining some details of implementation and before the implementation is ready. Moreover, in the FSMs of the Comunication Module some

events shall be replaced by a set of others, as the case of the INT2DirectTC, which corresponds to 23 different telecommands.

DISCUSSIONS AND RELATED WORKS Various methods have been published to optimize the number of test cases to minimize the efforts required to obtain an acceptable level of software quality (Lai, 2002). However, these methods and test tools are yet rarely used in the software industry. Issues, such as ‘What are the hindrances Table 3. Number of tests generated by method for each Finite State Machine. FSM

UIO

HSI

HSI*

SPY

SPY*

R_N

234

R_Ex13

356

R_Ex2

201

R_Ex4

201

R_Ex5

201

T_N

316

184

T_Ex1

316

184

T_Ex2

335

158

Table 4. Results of the generated Test Sets. Number of test cases FSM

Mutation Analisys

Total Set

Minimal Set

# Mutants

Mutation Score

R_N

342

285

1852

1.0

R_Ex1

460

408

1720

1.0

R_Ex2

286

264

1320

1.0

R_Ex4

286

263

1357

1.0

R_Ex5

284

265

1357

1.0

T_N

593

432

775

1.0

T_Ex1

593

432

775

1.0

T_Ex2

573

454

591

1.0

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

to the adoption of the generation methods based on FSM in the industry?’ or ‘What is the gap between academic and industrial testing methods?’ have been investigated in order to ease the difficulties of employing FSMs in the real context. In this section we discuss these questions in general speaking and concerning our approach. The fault domain, introduced in Background section, is defined for general purpose FSM. That fault domain provides all possible faults that can happen in the model. However, it may be interesting to adapt the model to the real scenario that should be tested. In Srivastava and Singh (2009) a fault model based on FSM to embedded systems was proposed. The fault model provides three kinds of errors that could occur among the connections between the hardware and software components: unconnected outputs, when the implementation should activate an output but does not activate, in other words, there is no output after a transition happens; unconnected inputs, when an input event has no behavior effect, there is no transition to another state as expected; and redirected inputs, which occur when an input is redirected to an incorrect state and generates a wrong behavior. The requirements for the construction of the test set were defined according to the new fault domain. The main focus of Srivastava and Singh is on the definition of the domain applied to embedded systems; and a testing generation method is not applied. In Santiago et al. (2008) a test environment that supports the generation of test sequences based on Statecharts and FSMs is presented. The simplicity of the model is considered the main advantage of employing FSMs as a technique for modeling reactive systems. Methods such as W and UIO were applied in the context of embedded systems. Although the applicability of MBT has been widely investigated, there are some reasons for reluctance in adopting academic methods for FSM-based testing. We have evaluated some of the reasons presented in Lai and Leung (1995) and Lai (2002), such as feasibility, that concerns the use of the testing methods in real cases; extreme formalism, that gives the impression of being too academic; need for training or education in this area; and resistance to changing, because it is not necessary the use of formal methods to do tests, according to the current point of view of many industries acting in the satellite-based software sector. The difficulty in using MBT starts with the modeling phase. Identifying the system inputs and outputs, as well as the system behaviors, is not a trivial task. A broad knowledge about the system and about FSM is necessary to understand how a

459

non-formal specification could be transformed into a formal one. In our approach, CoFI methodology helped to deal with these difficulties. The tester is guided, step by step, to model the system behavior into several simple FSMs. Moreover, the JPlavisFSM tool facilitated fitting the models to the testing methods, automating the test cases generations, producing test sets’ metrics allowing the tester to choose the best testing set. The tester must yet interpret the generated test sets due to the abstraction of the model, though the practicality of MBT was improved by CoFI and JPlavisFSM. The formalism of the FSM-based testing methods is transparent to the JPlavisFSM tool user. The analysis of FSM properties is automatically generated. Some validations with the use of FSM-based testing techniques and CoFI methodology have been conducted (Anjos et al., 2011; Pontes et al., 2012; Mattiello-Francisco et al., 2013). In such cases the Condado (or ConData) (Martins et al., 1999) was used as a tool to automatically generate test cases. Condado tool does not have neither the facilities to FSM structural analysis nor different methods to be chosen, however the single FSM property it requires is the FSM be connected. In the approach presented here, the FSMs have more elaborated structures, so the facilities provided by the JPlavisFSM tool allowed improving the application of the CoFI. Besides that, it was applied to real embedded software of satellite application. Concerning the training aspect, JPlavisFSM tool can be used to teach how the generation methods work. The tool has the option to import extra methods. The user may implement his/ her own method and apply the n-Complete tool to analyze if the method is correct. The JPlavisFSM may be used to support the modeling phase too, since the GUI is simple, clean and it provides functionalities for analyzing FSMs. The usability improvements provided by JPlavisFSM and the guidelines of CoFI methodology were tentative of reducing the gap between the good results obtained in academia with testing methods and its use in practical cases of industry as discussed in Lai and Leung (1995).

CONCLUSION The gap between academic research and industrial practice is still large. Thus, it is important to analyze the performance of testing activity supported by FSMs in real systems, once FSMbased methods are richly explored in the academic context.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

460

Pinheiro, A.C., Simão, A. and Ambrosio, A.M.

This paper has reported on the applicability of the FSM-based testing in the real, developed Communication Module software of the ITASAT university satellite. As testing is a costly activity, automating the generation of test cases can help reduce the costs. In this sense, MBT is an approach that derives test cases from a formal model built to support the testing activity. On the other hand, real systems like ITASAT software should be tested taking into account requirements of timing and reactions to events. In this context, the systematic approach of CoFI methodology supported by the formalism implemented by JPlavisFSM could be an alternative to test the system. We have observed restrictions and benefits on applying this approach. Some observed limitations, which should be issues for new investigations, are: • The modeling phase is not trivial and some knowledge about FSM is required to start the process. The tester has to learn the basic concepts and structural properties of FSM; • The applicability of the generation methods is highly dependent on structural properties of FSM generated in modeling phase. The tester has to know the basic theory about the generation methods; • The abstraction of the system is required to build models, so the test set generated is an abstract set of inputs and outputs, which needs to be interpreted before being executed. So, the post-processing of test sets could aggregate an extra cost to the test execution phase. The observed benefits were: • The CoFI methodology guides the modeling phase, which helps the inexperienced testers start using the FSM technique;

• The FSM-based test generation was automated leading to

cost reduction of the test activity; • The JPlavisFSM tool eliminates the need of profound

theoretical knowledge about FSM structural properties and testing generation methods. The MBT still has some practical limitation, but the initiatives that were made in academia helps to reduce them. The definition of new generation methods which do not require properties, such as FSM minimality, has to be explored. At the same time, the industry needs to step forward on the use of formal methods. These initiatives will improve the efficiency and effectiveness of testing activity. As future work, we envision the following. The generated test sets must be post-processed and executed against the Communication Module implementation of ITASAT and the complexity of these activities should be analyzed against the obtained results.

ACKNOWLEDGMENTS The authors would like to thank professors Emília Villani, David Fernandez and Wilson Yamagutti for the opportunity of applying the proposed approach in the ITASAT project. The authors would also like to thank the financial support of FAPESP, CNPq and CAPES. The authors are very thankful to reviewers for their useful comments.

REFERENCES Alencar, W.A.F., 2013, “Model checking aplicado a software embarcado crítico do satélite universitário ITASAT”, Dissertação de mestrado, ITA, São José dos Campos, Brazil.

Aplicações Espaciais”, Tese de doutorado, INPE-13264-TDI/1031.

Anjos, J.S., Gripp, J., Pontes, R. and Villani, E., 2011, “Applying the CoFI testing methodology to a multifunctional robot end-effector”, São José dos Campos, SP, Brazil.

and Martins, E., 2007, “Designing fault injection experiments using

Ambrosio, A.M., Martins, E., Vijaykumar, N.L. and Carvalho, S.V., 2005, “Systematic generation of test and fault cases for space application validation”, In Proceedings of the 9th ESA Data System in Aerospace (DASIA), Edinburgh, Scotland. Noordwijk: ESA Publications.

Lecture Notes in Computer Science (LNCS) series. Springer Editors:

Ambrosio, A.M., 2005, “CoFI: uma abordagem combinando teste de conformidade e injeção de falhas para validação de software em

machines,” IEEE Transactions on Software Engineering, Vol. 4, No. 3,

206p. Ambrosio, A.M., Mattiello-Francisco, F., Santiago, V.A., Silva, W.P. state-based model to test a space software”, Third Latin-American Symposium on Dependable Computing (LADC), Morelia, México. Bondavali, A.; Brasileiro, F.; Rajsbaum, S. Springer, Berlin. pp. 170178. doi: 10.1007/978-3-540-75294-3_13. Chow, T.S., 1978, “Testing software design modeled by ﬁnite-state pp. 178–187. doi: 10.1109/TSE.1978.231496.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

FSM-Based Test Case Generation Methods Applied to Test the Communication Software on Board the ITASAT University Satellite: A Case Study

461

Delamaro, M.E., Maldonado, J.C. and Jino, M., 2007, “Introdução ao Teste de Software”, Elsevier.

NTE. INPE, São José dos Campos. Retrieved in March, 19th 2014, from http://urlib.net/8JMKD3MGP7W/3EUF338

Fabbri, S.C.P.F., Maldonado, J.C., Masiero, P.C. and Delamaro, M.E., 1994, “Mutation analysis testing for ﬁnite state machines”, In Fifth International Symposium on Software Reliability Engineering, Monterey, California, USA, pp. 220-229.

Pinheiro, A.C., 2012, “Subsídios para a aplicação de métodos de geração de casos de testes baseados em máquinas de estados”, Dissertação de mestrado. ICMC. USP. São Carlos, Brazil.

Fujiwara, S., Bochmann, G.Von, Khendek, F., Amalou, M. and Ghedamsi, A., 1991, “Test selection based on ﬁnite state models”, IEEE Transactions on Software Engineering, Vol. 17, No. 6, pp. 591603. doi: 10.1109/32.87284. Gill, A., 1962, “Introduction to the theory of ﬁnite-state machines”, McGraw-Hill, New York. Hierons, R.M. and Ural, H., 2010, “Generating a checking sequence with a minimum number of reset transitions”, Automated Software Engineering, Vol. 17, No. 3, pp. 217-250. doi: 10.1007/s10515009-0061-0. Lai, R. and Leung, W., 1995, “Industrial and academic protocol testing: the gap and the means of convergence”, Computer Network ISDN System, Vol. 27, No. 4, pp. 537–547. doi: 10.1016/01697552(93)E0110-Z. Lai, R., 2002, “A survey of communication protocol testing”, The Journal of Systems and Software, No. 62, pp. 21-46.

Pontes, R.P., Morais, M.H.E., Véras, P.C., Ambrosio, A.M. and Villani, E., 2009, “A comparative analysis of two veriﬁcation techniques for DEDS: Model checking versus model-based testing”, In 4th IFAC Workshop on Discrete Event System Design (DEDS), Valencia, Spain, pp. 70-75. Pontes, P.R., Véras, P.C., Ambrosio, A.M. and Villani, E., 2012, “Contributions of model checking and CoFI methodology to the development of space embedded software”, Empirical Software Engineering – An International Journal, Vol. 10, No. 2, Springer. Online FirstTM, doi: 10.1007/s10664-012-9215-y - ISSN 1382-3256. Romero, A.G., Ambrosio, A.M. and Souza, M.L.O, 2012, “Finite statemachine verification applied to hybrid systems”, 21st SAE BRASIL International Congress and Exhibition, Expo-Center Norte, São Paulo, Brazil. doi:10.4271/2012-36-0429. Sabnani, K. and Dahbura, A., 1988, “A protocol test generation procedure”, Computer Networks and ISDN Systems, Vol. 15, No. 4, pp. 285-297. doi: 10.1016/0169-7552(88)90064-5.

Martins, E., Sabião, S.B. and Ambrosio, A.M., 1999, “ConData: a Tool for Automating Specification-based Test Case Generation for Communication Systems”, Software Quality Journal, Vol. 8, No.4, pp. 303-319.

Santiago, V., Vijaykumar, N.L., Guimarães, D., Amaral, A.S. and Ferreira, E., 2008, “An environment for automated test case generation from statechart-based and ﬁnite state machine-based behavioral models”, In Proceedings of the 2008 IEEE International Conference on Software Testing Veriﬁcation and Validation Workshop. Washington, DC, USA: IEEE Computer Society, pp. 63-72.

Mattiello-Francisco, M.F., Ambrosio, A.M., Villani, E., Martins, E., Dutra, T. and Coelho, B., 2013, “An experience of the technology transfer of CoFI methodology to automotive domain”, LADC, Industrial track. BDBComp Biblioteca Digital.

Sato, L.S., Saotome, O., Timm, C., Fernades, D. and Yamaguti, W., 2011, “Itasat-1: Brazilian university microsatellite for payload test and validation in low earth orbit”, In Proceeding of the 8th Symposium on Small satellites for Eath Observation, Berlin, Germany.

Mealy, G.H., 1955, “A Method for Synthesizing Sequential Circuits”. Bell System Technical Journal, Vol. 34, No. 5, pp. 1045-1079.

Simão, A.S., Ambrosio, A.M., Fabbri, S.C.P.F., Amaral, A.S.M.S., Martins, E. and Maldonado, J. C., 2005, “Plavis/fsm: an environment to integrate FSM-based testing tools”. In Simpósio Brasileiro de Engenharia de Software - Sessão de Ferramentas. Universidade Federal de Uberlândia. Uberlândia, Brazil.

Moore, E.F., 1956, “Gedanken–experiments on sequential machines”, in Automata Studies. Princeton University Press, pp. 129-153. Morais, M.H.E and Ambrosio, A.M., 2010, “A new model-based approach for analysis and refinement of requirement specification to space operations”, In Proceedings of the 10th Conference on Space Operations. Huntsville, Alabama, USA. SpaceOps 2010. American Institute of Aeronautics and Astronautics (AIAA). doi: 10.2514/6.2010-2231. Morais, M.H.E., 2011, “Uma abordagem para a melhoria da qualidade de requisitos baseada em modelos de estados”, Dissertação de mestrado. INPE, 2011. São José dos Campos. Pedrosa, L. and Moura, A., 2010, “Generalized partial test case generation method”, In 4th International Conference on Secure Software Integration and Reliability Improvement Companion (SSIRI-C 2010). Washington, DC, USA: IEEE Computer Society, pp. 70-77. Petrenko, A., Yevtushenko, N., Lebedev, A. and Das, A., 1993, “Nondeterministic state machines in protocol conformance testing”, in Proceedings of the 6th International Workshop on Protocol Test systems VI (IFIP TC6/WG6.1). Amsterdam, The Netherlands, The Netherlands: North-Holland Publishing Co., pp. 363-378. Pinheiro, A.C. and Ambrosio, A.M., 2013, “Modelagem do Módulo de Comunicação do Satélite ITASAT – Segundo a metodologia CoFI”, Relatório de Pesquisa, sid.inpe.br/mtc-m19/2013/10.04.19.58-

Simão, A.S., Petrenko, A. and Yevtushenko, N., 2009, “Generating reduced tests for fsms with extra states,” In Proceedings of the 21st IFIP WG 6.1 International Conference on Testing of Software and Communication Systems and 9th International FATES Workshop, ser. TESTCOM ’09/ FATES ’09. Berlin, Heidelberg: Springer-Verlag, pp. 129-145. Simão, A.S. and Petrenko, A., 2010, “Checking completeness of tests for ﬁnite state machines”, IEEE Transactions on Computers, Vol. 59, No 8, pp. 1023-1032. doi:10.1109/TC.2010.17. Srivastava, S. and Singh, A., system using fault modeling”, ELECTRO.2009.5441142.

2009, “Testing of embedded pp. 177-180. doi:10.1109/

Utting, M. and Legeard, B., 2007, “Practical model-based testing: A tools approach”, 1st ed. Morgan Kaufmann. Yin, Y., Li, Z. and Liu, B., 2010, “Real-time embedded software test case generation based on time-extended EFSM: A case study”, In Proceedings of 2nd WASE International Conference on Information Engineering (ICIE’10), pp. 272-275. Vuong, S., Chan, W. and Ito, M., 1989, “The UIOv-Method for protocol test sequence generation”, In Proceedings of the 2nd International Workshop Protocol Test Systems, pp. 161-175.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.447-461, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.390

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry Fabiano Armellini1, Paulo Carlos Kaminski2, Catherine Beaudry1

ABSTRACT: Open innovation is the systematic integration of collaborative, sourcing and revealing practices into a firm’s business strategy. Its implementation does not happen at once, but through a journey. This paper investigates this subject in an emergent economy context, the Brazilian aerospace industry, presenting the critical analysis of a questionnaire-based survey performed by means of personal in-company interviews in 22 firms. It comprises a wide range of practices associated with open innovation, connected to a conceptual model. We find open innovation elements in the sample, with no open business strategy behind them, though. Deficiencies regarding funding, R&D maturity and intellectual property protection prevent the cluster from being fully open innovator. Nevertheless, the culture company in the sample is very prone towards openness. From that, we conclude that open innovation in the cluster is still “unfreezing”, but with great potential to emerge, once these problems are solved. KEYWORDS: Open innovation, Aerospace Industry, Brazil, Innovation management.

INTRODUCTION This work is about the applicability and importance of open innovation to the Brazilian aerospace industry. Open innovation is a term coined by Chesbrough (2003a) to designate a new mindset within industrial organization in which companies make use of “purposive inflows and outflows of knowledge to accelerate internal innovation, and expand the markets for external use of innovation, respectively” (Chesbrough, 2006). The subject of collaboration, networking and use of external sources for research and development (R&D) is not a novelty in the academic and specialized literature about innovation management. Freeman (1991), in a review paper in the early 1990’s, has shown evidence that the use of formal and informal R&D networks and other kinds of collaborative arrangements are in order since before World War II. Freeman also states that the growth of structured networks of innovators, claimed by Chesbrough (2003b) to be at the root of the open innovation era, dates from the 1980’s. Also within new product development (NPD) literature, review papers on the subject prior to the rise of open innovation (e.g. Brown and Eisenhardt, 1995; Liyanage et al., 1999; Krishnan and Ulrich, 2001) identify a clear outward tendency in their time. In spite of all that, the coining of the term “open innovation” has brought to the literature a binding perspective on a number of existing practices, and the necessity to structure such practices into firms’ strategies. According to the OECD (2008), the novelty of the open innovation approach lies on the systematic

1.Polytechnique Montréal – Montréal – Canada 2.Universidade de São Paulo – São Paulo/SP – Brazil Author for correspondence: Fabiano Armellini | Department of Mathematical and Industrial Engineering - Polytechnique Montréal | 2900 Boulevard Edouard-Montpetit-Montréal/QC H3T 1J4 | Canada | Email: fabiano.armellini@polymtl.ca Received: 08/28/2014 | Accepted: 10/28/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry

integration of such practices into firms’ strategy, as well as on the exploitation of the outbound (inside-out) process by firms. It is evident that the adoption of open innovation is different from firm to firm, and from industry to industry. So it should be, since the “optimum” level of openness depends on variables such as technology intensity, value chain position, and product development average lead times, to name a few (Chesbrough, 2003b; West et al., 2006). According to Chesbrough and Appleyard (2007), a good open business strategy “balances the tenets of traditional business strategy with the promise of open innovation”. In other words, it is about the decision of what shall and what shall not be disclosed. The adoption of open innovation does not happen at once, though. Enkel et al. (2011), for instance, developed a maturity level framework, which identifies five stages of adoption of open innovation: (1) Initial/arbitrary; (2) Repeatable; (3) Defined; (4) Managed; and (5) Optimizing. Chiaroni et al. (2011), in their turn, classify the level of adoption of open innovation by means of a three-axed framework, which they name “the open innovation journey”. In the first axe, the process of adoption of open innovation is split into three stages: the first is the “unfreezing” stage, which implies in establishment of a sense of urgency of change and a cultural shift towards openness, although not yet put in practice. The second, called the “moving” stage, concerns the actual implementation of change through the establishment of new procedures and patterns of behavior consistent to the new vision derived from the cultural shift. Finally, the third one is the “institutionalizing” stage, which the company achieves when open innovation is incorporated to its formal procedures and internal business process maps, with its own metrics and subjected to continuous improvement procedures. The second axe is with respect to the direction of openness, and adopts the terminology of Enkel et al. (2009): outside-in and inside-out, the first usually preceding the latter. The third axe regards the structuring of managerial levers for open innovation, namely: • Networking; • Organizational structures; • Evaluation processes; and • Knowledge management systems. From these two papers (Enkel et al., 2011; Chiaroni et al., 2011), one realizes that it is not just a matter of determining whether open innovation is adopted or not, but also of establishing

463

at which step one sits. This question can also be posed at the industry level. Another issue identified in the emerging open innovation literature is that although the concept has been presented as a general trend (sometimes even as a paradigm change), what is observed in practice is that open innovation publications are clustered into a few “open-dominated industry segments” (Chesbrough and Appleyard, 2007), where the evidences of openness are more clearly found. Another limitation is that most of the studies published so far analyze open innovation in the context of developed economies (especially the U.S. and the European innovation systems) as first noticed by West et al. (2006), which persists to date. That being said, this paper has, therefore, a double-edged purpose: the first one is to present the descriptive analysis of the results of a survey whose goal was to search for evidences of open innovation trends within the Brazilian aerospace cluster. The second goal is to answer, in an explorative basis, and based on the data from the same survey, the following two research questions: • Does open innovation makes sense for companies of the aerospace industry in Brazil? • How mature are companies in the cluster towards open innovation? In other words, the goal behind this work is to verify whether collaborative and outsourcing activities within this industry are indeed part of an open business strategy or not, and where companies sit within the open innovation journey framework. This is an important issue to investigate, since most open innovation publications ignore the importance of business models to date (West and Bogers, 2013). The relevance of this work lies on the previously mentioned lack of literature about open innovation in developing contexts. Moreover, there is a worldwide interest in Brazil since 2003, when Goldman Sachs issued a paper about the BRIC (Brazil-RussiaIndia-China) economies (Wilson and Purushothaman, 2003). One finds a number of publications which show evidences of openness within developing economy contexts, but there is no conclusive answer so far whether this issue should be managed in the same manner they are in developed countries. Moreover, the open innovation literature lacks studies about more traditional high-technology segments, since most of the evidence found is from industries where innovation dynamics is more intense, such as the information and communications

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

464

Armellini, F., Kaminski, P.C. and Beaudry, C.

technology (ICT) and the pharmaceutical sectors. The aerospace industry is an example of such an industry for which one finds few studies discussing open innovation. In order to fill both gaps, this paper presents the result of a research project which took place between 2011 and 2013 to investigate Brazilian aerospace companies in search of open innovation patterns within their R&D and NPD strategies. This work’s results and findings shall add to the current discussion concerning the dynamics of these “new entrants” in the global market. The structure of the remainder of the article will present an overview of the Brazilian aerospace industry and its journey towards openness. Afterwards, the research methodology employed to investigate the subject is presented, which includes a conceptual model and the structuring of a survey followed by the main results and a descriptive and critical analysis of the data obtained from the survey will be presented. Finally, some conclusions will summarize this paper’s contributions and makes suggestions for future investigation on the subject.

THE BRAZILIAN AEROSPACE INDUSTRY AND ITS PATH TOWARDS R&D OPENNESS A striking paradox in the Brazilian aerospace industry is that, although the country is proud to have Alberto Santos Dumont (1873-1932), a Brazilian-born inventor, among the pioneers of flight, the establishment of the industry has happened much later and did not occur spontaneously, but by means of government intervention. The creation of the Brazilian aerospace cluster in the city of São José dos Campos, strategically located in between São Paulo and Rio de Janeiro, the two largest metropolises in the country, is the cornerstone of the national aerospace. It was accomplished by the establishment, in the midtwentieth century, of the aeronautic research institute Centro Técnológico de Aeronáutica, nowadays called Departamento de Ciência e Tecnologia Aeroespacial (DCTA), and the Air-Forces engineering college Instituto Tecnológico de Aeronáutica (ITA) in the city (Ferreira, 2009; Gomes, 2012). Later on, in 1971, the establishment in the same city of the Brazilian space research institute (INPE – Instituto Nacional de Pesquisas Espaciais) consolidated São José dos Campos and surroundings as the heart of the aerospace cluster in Brazil. The history of most

Brazilian-owned aerospace companies is connected to these public institutes, as they are either spin-offs of these institutes, or have their founders coming from ITA, DCTA or INPE. That is the case of Embraer, Mectron and Avibras, to name a few. Aerospace comprises three segments: aeronautics, defense and space (IMAP, 2011). Undoubtedly, the main segment within the Brazilian aerospace in terms of revenues is aeronautics, as shown in Fig. 1. This is mostly due to Embraer, a global leader in the regional jets segment. In 2007, for instance, the aeronautic segment was responsible for 92.9% of the total revenues in the Brazilian aerospace market, of which 84.5% is attributable to Embraer (Ferreira, 2009). Not neglecting Embraer’s importance to the cluster, one may not reduce the Brazilian aerospace industry to one single company. There are several other niche markets exploited by a number of different companies acting in the country that must be taken into account. One such niche market is the helicopter segment exploited by Helibras, a subsidiary of EADS Eurocopter, the only helicopter manufacturer in Latin America. There are also a number of smaller aircraft manufacturers that develop and produce smaller vessels for agriculture, flight training and leisure. The defense and space segments in Brazil are highly segmented, with small and medium enterprises acting in very specific niche markets, related either to civil and military air control infrastructure, or to the modest but relevant Brazilian space program or yet to other initiatives that have the government as the demander and purchaser. In all three segments, though, the role of the government is central for funding and supporting product development projects, beyond its role as purchaser and final user. Local authorities acknowledge this situation by considering the aerospace industry as a strategic segment for the national development plan (Gomes, 2012). Although strategic, throughout their history, aerospace companies often find a hard time regarding public support to develop their own technologies and products. This is partly

Space 0.4% Defense 6.7%

Aeronautics 92.9%

Embraer 84.5%

Others 8.4%

Figure 1. Brazilian aerospace market share in 2007 (Ferreira, 2009).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry

due to the novelty of the regulatory framework for innovation in Brazil, whose cornerstone is the so-called ‘innovation law’ (federal law 10.973/04), issued only in 2004. As a result, despite the country having a distinguished competence in this industry globally, the cluster is very dependent on foreign knowledge, even for the prime contractor Embraer. When analysing Embraer’s supply chain, one realizes that 93% of its suppliers are located abroad due to the lack of qualification of local suppliers (Figueiredo et al., 2008). According to Cassiolato et al. (2002), the risk-sharing partnership model adopted by the company to develop its aircrafts since the mid-1990’s was the solution Embraer found to gain access to the technology within the systems provided by its suppliers. As for Helibras, since its establishment in the late 1970’s, its business model was basically based on the manufacture and sale of aircrafts developed by its controller Eurocopter. More recently, though, the company is engaged with R&D, thanks to some technology transfer agreements from Eurocopter, that put the Brazilian subsidiary in charge for the development of a couple of platforms. This R&D is being performed in collaboration with local universities and public research institutes (Caiafa, 2012). Finally, regarding the defense and space segments, the very structure of the market leads companies to collaboration, since their very small economies of scale and scope hinders them to invest in self-dependent R&D infrastructures. Collaboration is one of the few alternatives for survival in this scenario. One such example was the Atech-Omnisys joint venture for the development of the first 100% national S-band weather radar, which resulted in the spin-off Atmos Systems (Silveira, 2005). Another example was Mectron’s collaborative arrangement with Brazilian and South-African enterprises for the development of the fifth-generation missile A-Darter (Silveira, 2009). Recently, these segments are attracting the interest of new and bigger entrants, such as the French company Thales, which controls the Brazilian firm Omnisys, since the acquisition of 51% of the company in 2005 (Mileski, 2011). Likewise, the Swedish company Saab started operations in Brazil, through the launch of R&D facilities in 2011 (Pedroso, 2011), anticipating its victory in the fighter jet bid from the Brazilian Air Forces (Soto and Winter, 2013). Another important change in the segment’s panorama was Mectron’s acquisition by the Odebrecht group, the largest Brazilian company in the construction sector (Rolli, 2011).

465

All these recent changes motivated us to put forth the research project, which takes place within this dynamic environment, of alliances, partnerships, mergers and acquisitions, also characterized by secrecy and concealing due to the involvement of the military and matters of national sovereignty.

METHOD The literature exploring the adoption of open innovation practices in aerospace is rather sparse and close to inexistent. Therefore, this work aims at performing an extensive investigation of open innovation concepts, tools, practices and strategies to verify what is in order for aerospace product development. To that purpose, we performed a questionnairebased survey covering all relevant aspects related to an open innovation strategy. After gathering some information about the firm itself, the questionnaire inquires about technology innovation management in general (product and process innovation), using the OECD’s Oslo Manual framework (OECD and EUROSTAT, 2005). The most extensive part of the questionnaire, though, is about open innovation issues. Departing from the tripartite division of open innovation proposed by Enkel et al. (2009), we elaborated a list of the pertinent issues associated to each core processes, presented in Table 1. In order to analyze ‘when’ and ‘how’ open innovation occurs within product development process, we propose a conceptual model, based on the three-phased R&D framework found in the Frascati manual (OECD, 2002). This model, presented in Fig. 2, identifies the internal “products” within the three activities of the R&D framework, that is, “knowledge” as basic research’s main output, “technology” for applied research and “products” for development. The combination of such assets are the ideas, here defined as creative impulses, that allow the combination of existing data, information, knowledge and technologies into new knowledge, technologies, products and/or processes. This conceptual model, along with the issues identified in Table 1, serve as a guide for the elaboration of an extensive survey questionnaire that structured the interviews performed during data collection. Moreover, in order to cover for the open innovation journey, the questionnaire also inquires about open innovation culture

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

466

Armellini, F., Kaminski, P.C. and Beaudry, C.

(to evidence the “unfreezing” phase), formal business processes (for the “moving” phase) and open innovation metrics (for the “institutionalising” phase).

The result is a very comprehensive questionnaire of 71 questions, which took in average one hour and a quarter to be filled out during interviews.

Table 1. Open innovation issues. Core process

Issues associated

RESULTS AND ANALYSES

External knowledge/technology sourcing Integration of the customer and/or user in the innovation process Outside-in

The survey sample comprises a total of 22 Brazilian aerospace companies that presented R&D activities and high degree of maturity in the product development process (Oliveira and Kaminski, 2012). These companies provided detailed information between the years 2007 and 2011. Data collection was performed through personal interviews with R&D managers or directors responsible for the innovation process management within the company. All interviews took place inside the participating companies. The sample is representative of the population being studied. The official Brazilian aerospace catalogue CESAER (DCTA, 2011) comprises a list of 270 companies. However, when analyzing the catalogue, one realizes that this list includes companies that, although being part of the aerospace supply chain, are not indeed aerospace companies. After filtering out these

Integration of the supplier in the innovation process Licensing in Spin-ins, mergers and acquisitions (M&A) IP portfolio activity Licensing out

Inside-out

Spin-offs and divestments R&D outsourcing Co-development with other companies

Coupled

Collaboration with universities and other science and technology institutes (STI) Venture capital

Data Data

Lincensing-out

Information

Boundaries of the firm

Data

Information

Data

Information

Data

Information

Data

Information

Knowledge

Idea

Internal knowledge

Idea Idea

Information Data

Boundaries of the firm

Internal technologies

Basic research

Information

Applied research

Data

Lincensing-in External knowledge

Figure 2. Open product-development conceptual model. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

External technologies

Idea

Development

Goods, services and processes

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry

companies, one finds 150 aerospace companies. The catalogue includes a brief description of each company’s activities, which allows one to perform a second filtering, retaining only those engaged on R&D activities. By doing so, only 59 firms are left in the population. Therefore, the sample obtained represents about 37% (22/59) of the population of interest for this study. The aerospace industry comprises companies with varying technical expertise, since aerospace products embrace many different technologies. Looking up the distribution within the dataset, the largest proportion of firms lies in the electronic and avionic systems, followed by aircraft manufacturers/integrators and technical consulting firms. The others are specialized in satellite components, defense equipment, ICT/software, simulation equipment and mechanical machining. The small size of the population prevents us from disclosing the exact numbers of each category. One can also cluster the according to the firm value chain position, among four possible options: prime contractors, equipment manufacturers, subcontractors and final users. This research does not target final users. Regarding the other three positions, Fig. 3 shows the distribution considering three scenarios: • The entire aerospace industry (150 companies); • R&D-engaged firms (59 firms); and • This survey’s sample (22 firms). One realizes that the survey’s subset oversamples the final links of the chain (prime contractors and equipment manufacturers), even when compared with the R&D-engaged subset. Turning now to size representativeness, in this study a firm is considered a Small and Medium Enterprise (SME) if it has 500 employees or less worldwide (that is, accounting all plants around the world). The distributions are shown in Fig. 4. Following the trend observed in the value-chain position analysis, the sample studied in the survey focuses on larger firms, when compared both to the whole industry population or to the share engaged with R&D. Nevertheless, more than 3/4 of the sample is composed of SMEs, which means that their position is undoubtedly captured in the analyses that follow. INNOVATION MANAGEMENT The first section after the introductory questions to the characterization of the sample focuses on traditional innovation management indicators, as established by the Oslo Manual (OECD and EUROSTAT, 2005). Although this manual distinguishes four different types of innovation (product, process, organizational

1% 17%

467

81%

Whole industry Companies with R&D Sample

34% 9% 0%

Prime contractors

63% 50%

20%

41%

40%

60%

Equipment manufacturers

80%

100%

Subcontractors

Figure 3. Firm distribution with respect to value chain position. 92%

88%

12%

Whole insudtry Companies with R&D

77%

Sample 0% SME

20%

40%

23% 60%

80%

100%

Large firms

Figure 4. Distribution according to firm size.

and marketing), this survey concentrates only on technology product and process (TPP) innovation. All data covers the period ranging from 2007 to 2011. As one can infer from Fig. 5, no single company claimed to be engaged only in process innovation, which indicates that innovation within this industry (or at least within the sample under analysis) is product-oriented. Figure 6 confirms this finding, by showing that the impact of product innovation tends to be higher when compared with that of process innovation. This corresponds exactly to what Pavitt (1984) defines as a science-based industry in the taxonomy he proposed in the 1980’s. Besides product and process innovation indicators, a common metric for innovation performance are intellectual property (IP) protection tools. Table 2 shows the percentage of companies within the survey sample that used IP protection tools in the 2007-2011 period: the proportion of firms using formal methods (e.g. patents) is low while strategic methods, such as secrecy and complexity of design, are more frequent in the sample. This result follows a Brazilian tendency, partly due to well-known shortcomings in patent registration in the country (Cruz and Mello, 2006; Gosain, 2013), in addition to a world-level low-patent tendency observed in the aerospace industry globally. Moreover, Brazilian aerospace companies are in general not quite structured to deal with IP issues. As Fig. 7 shows, more

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

Armellini, F., Kaminski, P.C. and Beaudry, C.

468

Table 2. IP protection methods.

23%

68%

20%

40%

Process innovation only

None

60%

Type of protection 80%

100%

Product innovation only

Process Product

5% 0%

20%

60% 20%

Innovation at firm level

40%

Strategic methods

20% 35%

60%

Innovation at national level

80%

100%

23%

Innovation at world level

Figure 6. Highest innovation impact.

Sample

Patents and utility models

18%

Trademarks

23%

Registration of industrial designs

Secrecy

55%

Complexity of design

45%

Lead-time advantage

18%

Formal methods

Both

Figure 5. Types of innovation.

60%

IP protection method

0% No IP issues

50% 20%

IP issues manged informally

40%

14% 60%

distribuited IP management

5% 9%

80%

100%

TTO in the firm (not in the plant

TTO in the plant

Figure 7. IP management.

than 3/4 of the sample claimed either not to have any IP issues or to deal with them informally, when they occasionally occur. Only 28% of the respondents claimed to a formal structure to manage IP, and only half of it (14%) within a Technology Transfer Office (TTO) or equivalent (9% in the plant where the interview took place and 5% in the firm, at a different plant). Regarding at last the use of governmental support policies for innovation, Fig. 8 summarizes the adoption of a number of policies in the survey sample. The information that stands out is that more than half of the companies in the sample are users of non-refundable resources (grants) for innovation projects. Federal innovation agency FINEP and São Paulo state research agency FAPESP are the main culprits for this high rate, due to their well-established innovation support programs: “SubvençãoEconômica” from FINEP and “PIPE – Pesquisa Inovadora na Pequena Empresa” from FAPESP, the latter addressed to small enterprises (with less than 250 employees). As for the other indicators, the percentage is relatively low. This is partly due to the novelty of most programs, since they were in general established after the Brazilian innovation law, issued in 2004. PRODUCT DEVELOPMENT AND OPEN INNOVATION Recalling the conceptual model shown in Fig. 1, the questionnaire begins the open innovation section of the survey by inquiring where companies act across the R&D spectrum. As expected, all companies in the survey claimed to be committed to development activities. About 2/3 of the

Support for training Technology assistance programs

27% 5%

Loans for R, D & I

18%

Government venture capital 0% Grants for internal R&D projects

59%

Grants for collaboration with STI

23%

R&D tax credits

27% 10%

30%

50%

70%

Figure 8. Use of innovation public policies.

sample claimed to perform applied research internally. As for basic research, only 9% of the companies claimed to perform such activities in Brazil and other 9% claimed to perform it globally (in others plants) but not in Brazil. In order to perform this R&D, firms often find inspiration from others sources of knowledge. Figure 9 draws a radar diagram of the average importance attributed to a number of types of players that often contribute to R&D. Since the number of samples that claimed to be engaged to basic research was very low, the radar comprises and compares only applied research with development. The importance was given through a sevenpoint Likert scale with no central point. One can take interesting insights from the comparison between the two remaining shapes drawn in Fig. 9: • Internal R&D personnel are the most important, regardless of the R&D stage;

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry

• Firms from different industrial sectors are barely used as

sources of knowledge for aerospace companies, regardless of the phase; • Science and Technology Institutes (STI), such as universities and research labs are more important during research activities than for development; • Industry players, especially clients and suppliers are of great importance during development, but less important during the research phase. Among the same list, companies could name up to two players as the most useful for their respective R&D. Figure 10 shows the frequency at which each player was mentioned. Three players stand-out in this analysis: clients (59%), universities (50%) and suppliers (36%). Interestingly, the option “firms from other industries” is a choice that no respondent made, which corroborates the low importance of cross-industry collaboration for the segment. OPEN INNOVATION STRATEGY As mentioned in the introductory part of this paper, a recent review paper on open innovation research (West and Bogers, 2013) observed that there is a tendency on open innovation research to neglect the importance of open business models and strategies, in spite of their importance to distinguish open innovation from earlier research on interorganizational collaboration in innovation. Opposing such tendency, this research dedicates a full set of questions to investigate how openness is connected to the firm strategy and their respective business models. According to Dahlander and Gann (2010), there are two types of strategy for open innovation: pecuniary and nonpecuniary. Pecuniary strategies consist of external practices directly related to acquiring or selling companies, the first an inbound activity and the latter outbound. Figure 11 shows that a very low percentage of companies claimed to be engaged in this kind of practices in the 2007-2011 period, in both directions. The low percentages may be attributed to the low importance of such practices inside aerospace companies’ strategies, or may be due to the small time span of analysis (5 years). Merger and Acquisition (M&A), spin-offs and divestments in most cases are not part of the ordinary agenda of companies, and their lead-time is often greater than other open innovation practices. In order to cover for that, the survey also inquired about former involvement of the plant in M&A, divestment or spin-off processes, whose result is found in Fig. 12.

469

Applied research Development Internal R&D 7 Firm R&D Industry associations 6 (other plants) 5 Firms from 4 Other internal 3 other sectors sources 2 1 Universities Firms in the sector Gov labs

Clients

Private labs

Suppliers

Figure 9. Importance of players as sources of knowledge for R&D activities. Internal partner Universities

14% 50%

Gov labs 9% Private labs 14% Suppliers Clients Firms in the sector 9% Firms from other industry 0% Industry associations 0% 10%

36% 59%

30%

50%

70%

Figure 10. Top-of-mind partners for R&D.

With that approach, we found a larger M&A incidence within the sample. Therefore, even though 64% of the companies in the sample claimed they have never been involved with these matters, this result shows that this subject should not be disregarded when analyzing the Brazilian aerospace cluster, as 36% of the firms surveyed resulted from divestments, spin-offs or were merged or acquired by another company. Turning now to non-pecuniary strategies, these are connected to sourcing (inbound) and revealing (outbound) practices (Dahlanderand and Gann, 2010). One first example of such strategy is licensing. Figure 13 shows the percentage of companies in the sample that claimed to have performed licensing during the 2007-2011 period. The information that stands out is that not a single company in the sample claimed to have out-licensed their internal IP in this five-year period. In the inward direction, there were positive responses, but most of them are for the acquisition of specific development software tools or embedded software for the company’s own products. Figure 14 indeed shows that a high percentage of licenses (70% out of the 45%) is due to software development firms.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

470

Armellini, F., Kaminski, P.C. and Beaudry, C.

Spin-off and divestments Spin-ins and acquisitions 0%

23%

77%

18%

82% 20%

40% Yes

60%

Lincensing ou

80% No

100%

Figure 11. Strategic pecuniary practices related to the acquisition or selling of firms.

64%

18%

9% 9%

Licensing in 0%

100% 45%

55%

20%

40% Yes

Neiher

20%

40%

Resulted from divestiment or spin-out

60%

80%

Merged or acquired by greater company

Firms from other industries Clients

100%

When inquired about the importance of licensing activities for the company strategy, on a Likert-scale from 1 to 7 (where 1 means ‘not important’ and 7 means ‘very important’), the average remained below 2. Therefore, one concludes that firms in the sample do not adopt this kind of open strategy. Another non-pecuniary revealing strategy is the provision of R&D services to third-parties. Now this is a practice that is found to be of great importance among surveyed firms. As Fig. 15 shows, roughly 3/4 (73%) of the firms claimed to perform this kind of activity, and close to half (41%) of them stated to perform this practice with great intensity. The clients of these R&D services (Fig. 16) are mostly local, or located within the country, with low international incidence, which indicates that collaboration takes place with firms’ own networking links. Moreover, only 1/3 of such services are provided within collaborative arrangements in a strict sense, the remaining 2/3 are direct-contracted services for the development of parts for third-parties, in almost all cases, public institutes such as INPE, the DCTA or one of the many divisions of the Brazilian Air-Force Command (COMAER). This kind of practice, performed through direct-contracted development services, show a poor adherence of the open innovation model to the modus operandi of these companies. First, they are not performing R&D to develop their own products, but providing R&D services for a living, by outsourcing their own technical expertises to third parties. Besides, this kind

100%

30% 10% 20% 70%

Software suppliers

Both

Figure 12. Plants previously involved in M&A or divestment processes in the past.

80% No

Figure 13. Licensing in and out.

Firms in the sector 0%

60%

Private labs 0%

10% 20%

40%

60%

80%

Figure 14. Type of organizations from which IP is licensed.

of collaboration cannot be considered a partnership in the strict sense of the word; it is more akin to a supplier-customer relationship. In addition, there is not an open business strategy behind the decision of balancing the disclosure and revealing of internal knowledge and technologies. On the other hand, in line with Tranekjer and Knudsen (2012), the very provision of technical services to their customers grants these companies updated technical knowledge and privileged information about the market, which is crucial to their survival in the market, but also represents a source of opportunities to these companies to develop their own products. As a matter of fact, 68% of the R&D service providers in the sample claimed to use federal or state grants to finance the transformation of internal expertise into own products. What differs from the findings of Tranekjer and Knudsen (2012) in Denmark is that Brazilian aerospace R&D-service providers are not better product innovators in comparison to non-providers, since these firms often fail to accomplish the successful launching of such product innovations in the market. According to three R&D-service providers in the sample, Brazilian authorities lack uniformity in their purchasing policies of aerospace technologies, which makes it difficult for local companies for profiting from their own NPD initiatives, since opportunities identified today may not be confirmed as a purchase order when the product is ready for deployment.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry

27% 0%

20% Never

32% 40%

41% 60%

Rarely / sometimes

80%

93%

Locally (within 100km) Within the State of São Paulo (not locally)

100%

27% 47%

Within Brazil (not in São Paulo)

Oftne / all the time

Within the Mercosur (not in Brazil) Internationally (not in the Mercosur)

Figure 15. Proportion of firms providing R&D services by frequency.

Within collaborative arrangements

471

13% 27% 33%

0% 20% 40% 60% 80% 100%

Figure 16. Location of R&D service clients.

OPEN INNOVATION CULTURE Even though open innovation may not be a reality in many aspects of Brazilian aerospace firms’ behavior or strategy, its principles may yet be present in their internal culture. This is what Chiaroni et al. (2011) refer to as the “unfreezing” phase, which is addressed in the last part of the survey questionnaire. By means of a series of questions about the importance the company accords to a number of aspects, the survey is able to identify which open innovation practices are more related to the company routine, and therefore to its culture. The cultural facet is an aspect that open innovation research papers often lack, according to West and Bogers (2013). In Fig. 17, the level of importance attributed to each practice is shown in a 1-to-7 scale, since all questions were graded through a seven-point Likert scale (with no central point), and grouped by the three core processes (Enkel et al., 2009). As hinted from the results previously presented, there is a higher predominance of outside-in practices, followed by coupled activities and inside-out practices scored the lowest, with the exception of R&D services, as previously discussed. Performing a simple average of the importance attributed to each practice within the three processes shown in Fig. 17, one obtains overall indices of 3.3 for outside-in, 2.6 for coupled and 2.0 for inside-out practices. Not surprisingly, outside-in scored highest, followed by coupled, and inside-out with the lowest score, which is consistent to a number of open innovation publications (OECD, 2008; Gassmann et al., 2010; West and Bogers, 2013) that assert that there is a dominance of the outside-in process over the others. The final question to pose at this point is whether this scenario is consistent to a journey towards innovation openness. Since averages could be deceitful, one alternative is the creation of dummy variables to determine whether companies adopt each one of the three core-processes. To that effect, all three indices from all surveyed firms were put together, and the

median value (found to be 2.7) was set as the threshold for determining whether a particular open innovation process is part of the culture for each participating firm. Afterwards, firms were grouped according to the processes that were found to be present at their respective cultures. By doing so, it was found that there is a pattern of adoption of open innovation processes. A group of companies (about 28% of the sample) was found not to adopt any of the three processes, which were labeled as “closed innovators”. A second group of companies comprises those with one single process found within its culture (outside-in). This group corresponds to 5% of the sample. Among dual-process open innovators, two situations were found: outside-in is present in both cases, while the adoption of the second process differ: 14% of the sample adopts inside-out, while 43% of the companies, the largest group in this classification system, are coupled innovators. Finally, there is the full open innovators group that accounts for 10% of the sample, which adopt all three core processes. This classification system allows one the sketch the open innovation path illustrated in Fig. 18. Outside-in is the first process to be incorporated into a firm’s culture; pure inside-out or coupled innovators are inexistent in the sample. Companies only embody these core processes after they have mastered outside-in innovation. This finding is consistent to what Chiaroni et al. (2011) found in Italy. With respect to the process of adoption of openness within the open innovation journey framework, this survey’s questionnaire investigates the existence of a department formally responsible for open innovation processes and about the reference to open innovation activities in formal business process procedures of the firm. In both cases the incidence of positive answers was low (36%), even though the questions were structured in order to find the slightest evidence of these rather than fully operational open innovation management. The questionnaire also includes a few questions intended to verify whether “not-invented-here” (NIH) and the

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

472

Armellini, F., Kaminski, P.C. and Beaudry, C.

“not-sold-here” (NSH) syndromes (Chesbrough, 2003a) were present in the firm’s culture. According to the responses provided during the interviews, we found a very low presence of these syndromes in Brazilian aerospace companies: NIH affects only 4.5% of the companies in the sample, and NSH 9.1%. These answers clearly indicate a willingness to perform more collaboration and outsourcing than these firms actually do. All these results allow one to conclude that open innovation is indeed present within these firms’ culture, although not fully implemented in their strategies, internal procedures and market activity. It places the cluster in general in the “unfreezing” phase of the open innovation journey.

Open innovation is an emergent mindset observed in some technology segments, the so-called open-dominated segments (Chesbrough and Appleyard, 2007). Within this new mindset, firms become increasingly aware of both external knowledge for using in internal technologies and markets, as well as external opportunities for the use of internal knowledge in different markets. The body of knowledge built in the specialized literature, though, is mostly based on the case studies within these opendominated segments, often within developed economies. With that in mind, this research paper analyzed the applicability and relevance of open innovation within a different product development context in the Brazilian aerospace industry, a traditional high-tech industry within an emerging country. This industry is characterized by one large player (Embraer) and a large network of smaller companies, mostly concentrated in the cluster located in São José dos Campos and surroundings. Through an extensive survey, this research was able to raise some interesting data to add to the open innovation

CONCLUSIONS This paper aims to identify the level of integration of open innovation within the product development process.

External sourcing Early integration of customers Early integration of suppliers Spin-in and acquisitions Licensing in (not within collaboration)

5.2 4.5 3.4 1.5 1.7

R&D services (not within collaboration) Spin-outs and divestments Licensing out (not within collaboration)

3.5 1.3 1.0

Collaboration with other firms

4.0

Collaboration with STI R&D services (within collaboration) Licensing in (within collaboration) Licensing out (within collaboration)

3.7 2.8 1.5 1.0 1.0

Coupled activities

2.0

3.0

Inside-out activities

4.0

5.0

6.0

Outside-out activities

Figure 17. Open innovation practices as part of the day-by-day of Brazilian aerospace firms.

Closed innovators

Single-process open innovators

None of the processes 28%

Outside-in only 5%

Dual-process open innovators Outside-in and inside-out 14% Outside-in and coupled 43%

Figure 18. From closed to open: the journey of Brazilian aerospace firms. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

Full open innovators Outside-in, coupled and inside-out 10%

7.0

The Open Innovation Journey in Emerging Economies: An Analysis of the Brazilian Aerospace Industry

literature. This survey was performed by means of personal in-company interviews with 22 aerospace companies in the cluster, based on a questionnaire elaborated after the conceiving of a conceptual model to integrate open innovation to NPD, and a profound literature review for raising all pertinent issues within an open business strategy. By means of traditional innovation metrics, based on the Oslo Manual (OECD and EUROSTAT, 2005), we have found in general a product-oriented low to medium innovative performance, with low IP management structuring, lowadoption of public innovation policies and local impact. The prominent relation is the supplier-customer relationship, followed by the collaboration with STI. The latter, though, is much more relevant during research activities in comparison to development phases. Regarding, collaboration trends, we found that collaborative links are much more intense locally, and mostly with other aerospace companies. Besides, we found an awkward rate of R&D services provided through direct contract, which shows that many interviewed companies do collaborate, but not as a strategy for enhancing their own technologies and incomes, as advocates the open business model (Chesbrough, 2006; Chesbrough and Appleyard, 2007). In short, these results seem to lead to the conclusion that collaboration in this industry follows traditional standards, and that it is not fully adherent to an open business strategy, as defined by Chesbrough and Appleyard (2007). Analyzing the internal culture of the companies in the sample, though, we found that companies are willing to collaborate more and also to adopt more complex models for innovation and for sourcing and revealing outwards their own boundaries. However, institutional gaps in the Brazilian cluster prevent this willingness to become a reality. Among these gaps, the insight from some respondents point towards the lack of uniformity of the government in its role of purchaser and the prevention of companies from performing R&D for the development of their own products, which in part is due to the first gap. Comparing our findings to the open innovation journey framework proposed by Chiaroni et al. (2011), we found that the Brazilian aerospace cluster is indeed undergoing an openness process, but still in the “unfreezing” phase. In these stages, the sense of urgency for opening has been established, but institutional gaps prevents the clear adoption of an open business strategy. Through the analysis of the survey data, we can provide answers, on an exploratory basis, to the research

473

questions formulated in the beginning of this paper. On the one hand, yes, open innovation does make sense in the Brazilian aerospace cluster; however, firms in the cluster do not seem to be very mature in dealing with open innovation, as they are in general still framed in the unfreezing phase of adoption of the new mindset. Although this research focuses only one industrial segment (aerospace), from one single emerging economy country (Brazil), its results contribute for the general understanding of emerging economies in general, from an inductive perspective. The general panorama described in the data presented in this paper shows an industry with some very strong points from the knowledge and absorptive capacity perspective, but with structural gaps that prevents this potential to be fully developed. To change this scenario, effective government measurements are in order, to allow local companies to be able to develop their own technologies and products. For such, more than just offering funding alternatives, the government should revise its role as purchaser of aerospace technologies, and grant local companies a stable demand in the medium term that provides a stable ground for them to grow. On the other hand, there is also a step local companies should take, in order to organize further their R&D structures, so that they reach higher NPD maturity, with less informalities and more consistent strategies. Currently, the industry sits on a vicious cycle, where companies do not invest in R&D more because of lack of government support and the government do not support local companies because they lack R&D structure. As a final remark: one should be aware that all these results are based on a simple average of the results obtained from the interviews, accounted equally regardless of firms’ income. If we performed weighted averages with respect to income, we would find a much different scenario, because one player (Embraer), if present in the sample, would overshadow the whole sample, which is a problem normally found in studies that analyze the Brazilian aerospace industry. It is a common mistake to assume that the Brazilian aerospace industry resumes to Embraer alone.The focus intended for this paper is precisely the opposite. Regardless of having Embraer in the sample or not, what we desired to analyze here was the reality of the industry as a whole, including the needs of all those small players normally neglected in business sectors analyses.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

474

Armellini, F., Kaminski, P.C. and Beaudry, C.

REFERENCES Brown, S.L. and Eisenhardt, K.M., 1995, “Product development: past research, present findings and future directions”, Academy of Management Review, Vol. 20, No. 2, pp. 343-78. doi:10.2307/258850.

IMAP, 2011, “Aerospace global report 2011”, International Network of M&A Partners.

Caiafa, R., 2012, “R$200 milhões em investimento: Centro Tecnológico de Helicópteros em Itajubá”, Revista Tecnologia e Defesa, No. 130, São Paulo.

Krishnan, V. and Ulrich, K.T., 2001, “Product development decisions: a review of the literature”, Management Science, Vol. 47, No. 1, pp. 1-21. doi:10.1287/mnsc.47.1.1.10668.

Cassiolato, J. E., Bernardes, R. and Lastres, H., 2002, “Innovation systems in the South: a case study of Embraer in Brazil”, Paper prepared for UNCTAD-DITE investment policy and capacity-building branch, New York and Geneva, United Nations.

Liyanage, S., Greenfield, P. and Don, R., 1999, “Towards a fourth generation R&D management model – research networks in knowledge management”, International Journal of Technology Management, Vol. 18, Nos. 3-4, pp.372-93.

Chesbrough, H.W., 2003a, “Open innovation: the new imperative for creating and profiting from technology”, Harvard Business School Press.

Mileski, A.M., 2011, “O bom momento da Thales”, Revista Tecnologia e Defesa, No. 123, São Paulo.

Chesbrough, H.W., 2003b, “The era of open innovation”, MIT Sloan Management Review, Spring 2003, pp.35-41.

OECD, 2002, “Frascati manual – proposed standard practice for surveys on research and experimental development”, 6th edition, Organization for Economic Co-operation and Development.

Chesbrough, H.W., 2006, “Open business models: how to thrive in the new innovation landscape”, Harvard Business School Press. Chesbrough, H.W. and Appleyard, M.M., 2007, “Open innovation and strategy”, California Management Review, Vol. 50. No.1, pp. 57–76. Chiaroni D., Chiesa, V. and Frattini, F., 2011, “The open innovation journey: how firms dynamically implement the emerging innovation management paradigm”, Technovation, Vol.31, No. 1, pp. 34-43. doi: 10.1016/j. technovation.2009.08.007. Cruz, C.H.B. and Mello, L., 2006, “Boosting innovation performance in Brazil”, Economics Department Working Paper No. 532, Organization for Economic Co-operation and Development.

OECD, 2008, “Open innovation in global networks”, Organization for Economic Co-operation and Development. OECD and EUROSTAT, 2005, “Oslo manual: guidelines for collecting and interpreting innovation data”, 3rd edition, Organization for Economic Cooperation and Development. Oliveira, A.C. and Kaminski, P.C., 2012, “A reference model to determine the degree of maturity in the product development process of industrial SMEs”, Technovation, Vol. 32, No. 12, pp. 671-80. doi:10.1016/j. technovation.2012.08.001.

Dahlander, L. and Gann, D.M., 2010, “How open is innovation?”, Research Policy, Vol. 39, No. 6, pp. 699–709. doi:10.1016/j.respol.2010.01.013.

Pavitt, K., 1984, “Patterns of technical change: towards a taxonomy and a theory”, Research Policy, Vol. 13, No. 6, pp.343-73. doi:10.1016/00487333(84)90018-0.

DCTA, 2011, “CESAER: Catálogo de empresas do setor aeroespacial”, Departamento de Ciência e Tecnologia Aeroespacial.

Pedroso, R., 2011, “Governo quer transformar ABC paulista em polo da indústria de defesa”, Valor Econômico, October 20th.

Enkel, E., Bell, J. and Hogenkamp. H., 2011, “Open innovation maturity framework”, International Journal of Innovation Management, Vol. 15, No. 6, pp. 1161-89.

Rolli, C., 2011, “Odebrecht adquire controle da fabricante de mísseis Mectron”, Folha de São Paulo, 25 March.

Enkel, E., Gassmann, O. and Chesbrough, H.W., 2009, “Open R&D and open innovation: exploring the phenomenon”, R&D Management, Vol. 39, No. 4, pp. 311-16. doi: 10.1111/j.1467-9310.2009.00570.x. Ferreira, M.J.B., 2009, “Perspectivas do investimento em ciência – documento setorial: aeroespacial & defesa”, Vol. 12, Projeto PIB, Universidade Federal do Rio de Janeiro (UFRJ)/ Universidade Estadual de Campinas (UNICAMP). Figueiredo, P., Silveira, G. and Sbragia, R., 2008, “Risk sharing partnerships with suppliers: the case of Embraer”, Journal of Technology Management and Innovation, Vol. 3, No. 1, pp. 27–27. Freeman, C., 1991, “Networks of innovators: a synthesis of research issues”, Research Policy, Vol. 20, No. 5, pp. 499–514. doi: 10.1016/0048-7333(91)90072-X. Gassmann, O., Enkel, E. and Chesbrough, H.W., 2010, “The future of open innovation”, R&D Management, Vol. 40, No.3, pp.213-221. Gomes, S.B.V., 2012, “A indústria aeronáutica no Brasil: evolução recente e perspectivas”, in: Lage, F. (org.), “BNDES 60 anos: perspectivas setoriais”, Banco Nacional do Desenvolvimento, Vol. 1, Ch. 4, pp. 139-185. Gosain, R., 2013, “On the right track: the INPI ups its game”, World Intellectual Property Review, May-June 2013, pp. 56-57.

Silveira, V., 2005, “Atech constrói para SIVAM primeiro radar 100% nacional”, Gazeta Mercantil, 12 September, p.A-6, São Paulo. Silveira, V., 2009, “Mectron fornece mísseis para o Paquistão”, Valor Econômico, 17 April, p.B-9, São Paulo. Soto, B. and Winter, B., 2013, “Saab wins Brazil jet deal after NSA spying sours Boeing bid”, Reuters. Tranekjer, T.L. and Knudsen, M.P., 2012, “The (unknown) providers to other firms’ new product development: what’s in it for them?”, Journal of Product Innovation Management, Vol. 29, No. 6, pp. 986-99. doi:10.1111/j.1540-5885.2012.00974.x. West, J. and Bogers, M., 2013, “Leveraging external sources of innovation: a review of research on open innovation”, Journal of Product Innovation Management, Vol. 31, No. 4, pp. 814-831. doi:10.1111/ jpim.12125. West, J., Vanhaverbeke, W. and Chesbrough, H.W., 2006, “Open innovation: a research agenda”, in: Chesbrough, H.W., Vanhaverbeke, W. and West, J. (eds.), 2006, “Open innovation: researching a new paradigm”, Oxford University Press, Ch.14, pp. 285-307. Wilson, D. and Purushothaman, R., 2003, “Dreaming with BRICs: the path to 2050”, Goldman Sachs Global Economics Paper No: 99.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.462-474, Oct.-Dec., 2014

doi: 10.5028/jatm.v6i4.386

Development Status of L75: A Brazilian Liquid Propellant Rocket Engine Daniel Soares de Almeida1, Cristiane Maria de Moraes Pagliuco1

ABSTRACT: This paper provides an overview of the design and development process of the L75 Liquid Propellant Rocket Engine (LPRE). Being developed at Instituto de Aeronáutica e Espaço (IAE), it is the first Brazilian open-cycle liquid rocket engine pressurized by turbopump, designed to deliver 75 kN of thrust in vacuum as a cryogenic upper stage engine using liquid oxygen and ethanol. The Preliminary Design Review (PDR) was accomplished in December 2011 and December 2012 the project received from Agência Espacial Brasileira (AEB) a financial support, through an agreement with Fundação de Desenvolvimento da Pesquisa (FUNDEP), to proceeding with the development of models and tests. The main components of the engine, briefly described here, are thrust chamber assembly, gas generator, turbopump assembly, control system and ignition system. KEYWORDS: Liquid propellant rocket engine, Liquid propulsion, Ethanol.

INTRODUCTION The knowledge already acquired on technology of solid propellant rocket motors enables Brazil to have propulsion systems for small launch vehicles. In order to obtain commercially feasible vehicles, however, the inclusion of stages with liquid propulsion becomes a necessary condition. To obtain the technology and the know-how to design, manufacture, test and operate a Liquid Propulsion Rocket Engine (LPRE) is a goal to be achieved within the National Program of Space Activity (PNAE, 2012) in order to enable the development of satellite launch vehicles with higher performance and accuracy than solid propulsion. The development of a LPRE is based on foremost scientific methods and complex technological processes which require a heavy and long-term investment. The experience of other countries shows that the development cycle of a large liquid system is around five to ten years. At the same time, many activities are necessary to involve the broad participation of research centers, universities and industries in the solution of multi-disciplinary problems. The analysis of successful examples such as occurred in China, India, Japan and South Korea which had access to space after the first three — former Soviet Union, Germany, and the United States of America — shows that foreign partners had always been present in the initial stages of their development. It indicates that an investment in partnership is the key factor to overcome initial difficulties, mistakes and associated risks of the liquid propulsion technology (Niwa and Yoshino, 1997). The LPRE propels space vehicles better than any other type of chemical propulsion, giving generally a higher specific

1.Instituto de Aeronáutica e Espaço – São José dos Campos/SP – Brazil Author for correspondence: Daniel Soares de Almeida | Praça Marechal Eduardo Gomes, 50 - Vila das Acácias | CEP: 12228-904 – São José dos Campos/ SP – Brazil | Email: danieldsa@iae.cta.br Received: 06/27/2014 | Accepted: 10/30/2014

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

476

Almeida, D.S. and Pagliuco, C.M.M.

impulse compared with solid or hybrid rockets. Usually, the cryogenic propellants have the highest specific impulse in liquid rocket engines. Some of the most important advantages of LPREs are the very wide range of thrust values to fit specific applications, multiple ignition possibility, fast pulsing, long term operation time and ready to reuse; thrust variation upon command; controlling attitude changes and the capability to be checked and even fully tested before use. The restart of the LPRE thrust allows an accurate terminal flight velocity, which is important to achieve the required orbit. A remarkably high reliability has been achieved in the production of LPREs besides their lightweight structure, enabling more embedded payload for a mission. Most exhaust gases of LPREs using modern common propellants are non-toxic and environmentally friendly (Sutton, 2006). The cryogenic propellant liquid oxygen (LOx) and liquid hydrogen (LH2) produce the higher specific impulse available for all environmentally friendly bipropellant, which permit a considerable increase in cargo delivery for rocket launchers. However, cryogenic components have big issues with evaporation, which limits the duration of the mission flight. In addition to that, LH2 has the disadvantage of needing big tanks due to its very low density. The second more energetic and also non-toxic propellant pair is the LOx and kerosene, which has lower specific impulse but higher density and a prolonged service in space rather than LOx/LH2. LPRE with this propellant combination is suitable for both first and upper launcher stages. Moreover, although less energetic, the LOx and ethanol propellant pair has the advantages of its ready availability in Brazil, its chemical composition uniformity, being non-toxicity, its cleanliness, its low cost, and it does not suffer pyrolysis or produces significant soot at high temperatures inside the cooling channels of the regenerative cooled combustion chamber (Haeseler et al., 2000). Engines with higher specific impulse are, in order, those using LH2 and LOx as propellants and those operating in closed cycle. However, the technological challenges of using LH2 as propellant and the design of a closed cycle engine would make its development in the country still more complex than that of an open cycle engine using LOx and ethanol. There are several different designs whereby a turbine can be integrated into a LPRE, and this was identified as different engine cycles. The open cycle has a separate gas generator, where fuel

and oxidant are burned at a specific mixture ratio, resulting in the low temperature of the turbine inlet gases to prevent destruction of the turbine blades. This cycle is the simplest, and it often offers the lowest overall cost, provides a low engine structural mass due to lower internal pressure, but gives somewhat not higher performance than the closed cycle, which is more complex in design, manufacturing and testing. Figure 1 shows the L75 scheme, a LPRE with a typical open cycle (Torres et al., 2009).

THE L75 ENGINE CHALLENGE The current Brazilian Satellite Launcher Vehicle ( VLS-1) cannot meet the overall strategic objectives of the space sector. However, it is envisaged the possibility to significantly increase the payload capacity of VLS-1 through partial modification of the vehicle, especially with the replacement of the solid propellant motors by liquid propellant engines in an upper stage. To achieve this goal, several activities have been conducted in Instituto de Aeronáutica e Espaço (IAE) such as the specification, design and construction of liquid rocket engines. One of these projects is the L75 engine. A timeline of the project with the major milestones is: • 2008 – Beginning of the L75 project; • 2009 – System Requirement Review (SRR); • 2010 – 2012 – FUNCATE (Fundação de Ciência, Aplicações e Tecnologia Espaciais) contract: technical service to support development of the project - 20 specialists; • 2011 – Preliminary Design Review (PDR); • 2012 – Letter of exchange for Brazilian-German cooperation on the development and tests of a 75 kN engine; • 2013 – Fuel change from kerosene to ethanol; • 2013 – 2015 – FUNDEP agreement: to proceed with the development, manufacturing and test of some components. In 2011, a protocol of intention between Brazilian Space Agency (AEB) and the German Aerospace Center (Deutsches Zentrum für Luft-und Raumfahrt - DLR) was established with the objective of fostering cooperation in space activities. After a visit of a German delegation at IAE in April 2011, a letter of exchange was established in October 2012 for further detailing the envisaged activities within the cooperation. The conductance of a review of the L75 engine was identified as a first major step. Starting in 2012, meetings regarding a specific implementation

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

Development Status of L75: A Brazilian Liquid Propellant Rocket Engine

Figure 1. L75 scheme.

agreement between the IAE and the DLR to the development and tests of a 75 kN LOx-ethanol were organized. The L75 development is being based on the ECSS standards (ECSS, 2004), using development (DM), engineering (EM) and qualification (QM) models engine components, parts and subsystems in order to verify the conformance to requirements. A verification program was established through an engine development plan, according to the L75 engine requirements specifications, considering the verification methods, levels and models. For the L75 engine project, it is imperative to prioritize the development and mastery of critical space technologies, which are essential to the industrial progress, and the achievement of the necessary national autonomy in such a strategic activity. The allocation of specific system engineering requirements and needs to the L75 engine depends strongly on the type of agreement established between customer — IAE — and supplier — Brazilian enterprises — and the nature and level of complexity of the overall system subject of the agreement. The system engineering organization plans its activities in conformance with the project phases as defined by the management, in accordance to the European Cooperation for Space Standardization (ECSS) standards (ECSS, 2009). The L75 engine needs, program objectives and requirements identified in the present concept phase are

477

listed below. These requirements are preliminary in nature and not yet complete. The project needs and objectives identified by IAE are: • Consolidating the Brazilian space industry, by increasing its competitiveness and innovation capacity; • Encouraging funding of programs based on public or private partnership; • Enabling Brazilian suppliers for manufacturing, brazing and welding processes; • Encouraging the human resources development by training of experts needed in the Brazilian space activities, both domestically and abroad; • Expanding partnerships with other countries, by prioritizing joint development of technological and industrial projects of mutual interest; • Developing of the controlling, monitoring, transmission and the data log related to the project; • Developing of high complexity turbopump technology; • Developing of an upper stage engine for a launch vehicle capable to carry a payload; • Developing of staff and enterprises on liquid propulsion activities. Some tasks are not addressed to the L75 engine project, such as the launcher vehicle, the stage, the propellant tanks systems, the inert gas supply system, and the thrust vector control system. The critical technologies considered for the development of L75 engine are feasible materials for specific applications; bearings for the high rotation and low temperature operational conditions; seals for cryogenic application; welding and brazing process; development of cryogenic items as valves and regulators; regenerative thrust chamber; system feeding by turbopump; control system; assemblies and integrations; development of heat treatment after brazing for copper alloy and test facilities.

L75 ENGINE MAIN CHARACTERISTICS The main high-level function of the L75 engine is providing (75.0 ± 5.0) kN of nominal thrust in vacuum with the main operation conditions shown below:

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

478

Almeida, D.S. and Pagliuco, C.M.M.

• Specific impulse greater than 2,940 m/s in vacuum;

• Global mixture ratio in the range of 1.50 to 1.95;

• Total pressure inside of the combustion chamber in the

range of 5 to 7 MPa. The L75 engine nominal and extreme operational envelopes, a set of physical data in which the propulsion system, subsystem, or component is intended to operate, are shown in Fig. 2. The results obtained show that the L75 engine in the current configuration has a decent mixture ratio envelope, despite a small margin for thrust variation. The restrictive factors are the thrust chamber maximum mixture ratio, the gas generator maximum and minimum mixture ratios and the turbopump maximum fuel flow. The L75 engine (Fig. 3) thrust chamber consists of three main components, the combustion chamber, where the propellants undergo a set of chemical reactions at high pressure creating hot gases during the combustion process; the nozzle, where the hot gases are accelerated to supersonic velocities; and the injector head, where the propellants are introduced, broken up into small liquid droplets and distributed over the cross section of the burning region inside the combustion chamber. The thrust chamber cooling is a major challenge because the gas temperature exceeds the melting point of the metallic wall. To overcome this problem, the L75 engine uses a lightweight brazed and welded cooling jacket with a thin inner-wall construction. The turbopump is one of the most critical components during the development of a LPRE, since it is a high-precision and high-speed rotating machinery operating in critical conditions. The main goal of the turbopump is to feed the thrust chamber and gas generator with the necessary flow and pressure of fuel and oxidizer provided by the pumps, which are driven by the turbine. Its manufacturing is expensive, and its design is complex and demands a great effort of engineering. There are many different types of valves within the L75 engine project, such as fuel and oxidizer valves, vent valves, gas valves, pressure-regulating valves, check valves, thrust regulator, drain valves and safety valves. Most of the commercially available valves do not fulfill the engine requirements, are too heavy, or are not available for the Brazilian space program. For those reasons, most valves and regulators are being developed. The L75 engine will operate with a thrust regulator, which is capable to regulate the LOx pressure for the gas generator, modulating the turbine power, which reflects on the turbopump

Thrust [kN]

• Minimum mission burn time of 400 s;

P15 P8

P6 P5 P4

75 73 71

P16

P14

69 1,30

Nominal envelope Design point Extreme envelope

P13 1,35

1,40 1,45 1,50 Mixture Ratio (O/F)

P10 P9 P1

P11

P2 P12 1,55

1,60

Figure 2. L75 engine envelopes.

Valve Valve� assemblies� assemblies Thrust Thrust� chamber� chamber

GasGas�generator� generator Turbopump� Turbopump

Figure 3. L75 engine.

shaft speed and pumps power. It is also equipped with two main valve assemblies, one for LOx and another for ethanol. Both of them are very similar in configuration and their functions cover chilldown/filling bypass, gas purging, pressure safety relief, opening at the start-up sequence and closing at the shutdown sequence. The aim of the L75 ignition system is to achieve a safe, predictable, and reliable transition to nominal thrust in vacuum. The ignition system devices include a turbine starter, gas generator igniter, thrust chamber igniter and pyrotechnic actuators. The L75 engine igniter and the turbine starter are based on the operating principles of solid rocket motors, consisting of a metallic structure, an electro-pyrotechnic starter, a reinforcement

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

Development Status of L75: A Brazilian Liquid Propellant Rocket Engine

cartridge, the main cartridge propellant, and the thermal protection and nozzle. The control system of the L75 engine is responsible for the starting procedure and ensures a safe transition to nominal thrust in vacuum; to achieve the desired thrust by controlling the fuel and oxidizer flow rate and pressure drops; safe shutdown of the firing operation. Furthermore, additionally controlled features are the precision of the automatic thrust control, mixture-ratio control and condition monitoring (safety controls or health monitoring of the engine). A flow diagram of the L75 engine showing its components is given in Fig. 4.

479

The development life cycle phasing was elaborated according to ECSS system, but this system can be adapted to specific domains of application by use of tailoring activities necessary to meet specific project needs. The engine development is divided into five main phases: mission identification, feasibility (concept), preliminary design (development), detailed design (engineering) and qualification. Besides, the development models on thrust chamber, turbopump and gas generator (power pack), valves and regulators and control system level, two DM, one EM and two QM on engine level are considered the minimum necessary for engine development. The two qualification models will be subject to all essential modifications needed to meet the requirements of the engine specification. All engine components shall pass by acceptance tests before their integration to the engine. The selected basic approach is to minimize the overall technical risk and to reduce non-recurring cost during development by use of demonstrated hardware designs as much as possible.

DEVELOPMENT LOGIC The development and qualification philosophy dedicated to the L75 engine and its subsystems are presented in Fig. 5, where the hardware model flow is shown.

Ethanol LOx

Atm

Atm PSV 502

F-02

YV 20 NC Drain line Engine vehicle Atm GN 2

BVO– Oxidizer valves assembly F-02 – Oxidizer filter/Delay tank

Atm

FC – Mixture ratio regulator

BVO

BVGG

Chill-down line

IPS

VPE

NC Purge line R 25

BVGG– GG valves assembly

FO – Calibrated orifice

PC YV 19

BVC– Fuel valves assembly

Vehicle engine

YB 302.2

BVC

FO 206

R – Pressure regulator

YB – Event-driven burner Engine vehicle

FO 204

PSV – Pressure safety valve

VPE – Pre-stage throttle valve

Atm NC

YB 401

VD– Drain valve

YB 402

YB 302.1

IPS – Inter-propellant seal PC – Thrust regulator

GG – Gas generator

YV– Event-driven valve Liquid oxygen Ethanol Gaseous nitrogen Burnt gases

Figure 4. L75 engine flow diagram. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

480

Almeida, D.S. and Pagliuco, C.M.M.

OVERALL DEVELOPMENT STATUS Until the end of 2012, the L75 project was developed using, in addition to the IAE’s human resources, services based in a contracts with FUNCATE. This contract had a technical and service expertise as an objective to support technology development of the L75. In December 2012, the project received from AEB a financial support, through an agreement with FUNDEP, to proceed with the development of the first components, allowing proceeding with the needed tests to confirm the design parameters. This agreement seeks to give more flexibility to the development of the project and involved Brazilian industries in the space program. GAS GENERATOR The gas generator is a combustion device used in LPREs for creating gases in suitable conditions, which are used as a working fluid for driving the turbine of the turbopump. The basic elements of the gas generator (Fig. 5) are: injector element, cylindrical section, convergent section, in addition to fixing structural elements and interfaces. The igniter of the gas generator is part of the ignition system. The L75 gas generator

has a mass flow rate of 1.3 kg/s, a mixture ratio of 0.3 and gas temperature of 900 K. The first L75 gas generator was tested in April 2014 on IAE’s 20 kN test bench (Fig. 6). These tests are used in order to verify the overall component performance, the combustion process and to validate thermodynamic models used during the design. Figure 7 shows the gas generator relative chamber pressure versus the time during a hot test, demonstrating its stability. TURBOPUMP The L75 turbopump (Fig. 8) shaft is supported by two bearings, which are centrally mounted close to the fuel pump, so they can be lubricated and cooled by the fuel. Seals are also provided to limit leakage. Considering cavitation characteristics and the fuel rich gas on turbine side, the LOx pump is located at one end of the shaft and the fuel pump in the middle, connected by a spline between pumps and a shaft. The gas generator exhaust ducting leads to an impulse type single stage axial flow turbine with partial admission. Between the two pumps, an inter-propellant seal is adopted to avoid any interaction between propellants. The fluid enters the pumps at low pressure from the tanks. In this way, a tank pressurization system is necessary to maintain the required Net Positive Suction Head (NPSH) to prevent pump cavitation. The pump inlet pressure is usually minimized to reduce the tank size and weight. Once the fluid enters the pump at the inlet, the inducer adds energy to the fluid before passing to the impeller to add sufficient energy to the fluid to suppress cavitation. The rotor significantly adds more kinetic energy prior to directing the flow into the diffuser and the volute.

(a) Figure 5. Gas Generator.

(b)

Figure 6. (a) Gas generator hardware; (b) Gas generator hot test.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

Development Status of L75: A Brazilian Liquid Propellant Rocket Engine

The turbine’s (Fig. 9) main characteristics are power of 380 kW; inlet gas pressure of 5.0 MPa; outlet gas pressure of 0.4 MPa; gas flow rate of 1.3 kg/s; temperature in the blades of 800 K, rotation speed of 24,000 rpm. The oxidizer pump works with

481

inlet pressure of 0.4 MPa; outlet pressure of 7.6 MPa; and mass flow rate of 14.5 kg/s. The fuel pump works with an inlet pressure of 0.3 MPa; an outlet pressure of 10.6 MPa; and a mass flow rate of 9.4 kg/s (Fig.10).

Figure 7. Gas generator relative chamber pressure versus time during a hot test.

(a)

(a) Fuel pump

LOx pump rotor

Turbine

(b) Figure 8. (a) Turbopump; (b) Rotor assembly.

(b) Figure 9. (a) SAE 422 stainless steel turbine disk; (b) Finite element analysis of distribution of temperature on turbine disk. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

482

Almeida, D.S. and Pagliuco, C.M.M.

(a)

(b)

Figure 10. AA 7075 aluminum alloy high-speed rotor pumps (a) ethanol and (b) LOx.

(a) THRUST CHAMBER The main characteristics of the thrust chamber are: mass flow of 22.8 kg/s; mixture ratio of 1.7; temperature in the combustion chamber of 3400 K; and area expansion ratio of 150. Figure 11 shows the L75 thrust chamber in the short configuration for sea level hot tests. COMPONENTS The main characteristics of the L75 thrust regulator (Fig. 12), responsible for the control of the thrust level, are the nominal mass flow rate of 0.3 kg/s (LOx); the inlet pressure of 7.0 MPa and the outlet pressure of 6.0 MPa. L75 will operate with two main valve assemblies (Fig. 13) with the following parameters: LOx inlet pressure 7.5 MPa; nominal LOx mass flow rate 14.5 kg/s; ethanol inlet pressure 10.6 MPa; ethanol mass flow rate 9.4 kg/s. The ignition of the L75 engine is a challenging issue. The dominating criterion is reliability, repeatability and robustness of the ignition system (Jean and Dalbies, 2000). Several tests have already been executed with the thrust chamber and gas generator igniters and the turbine starter, as can be seen in Fig. 14.

(b) Figure 11. (a) thrust chamber; (b) equivalent stress in MPa (von Mises stress).

TEST FACILITIES Liquid rocket engines for launch vehicles as well as their subsystems need to be verified and qualified during cold and hot-runs. A high test cadence combined with a flexible test team helps to reduce the cost for test verification during development and qualification as well as during acceptance testing for

Figure 12. Thrust regulator valve.

production. The test facility allows to test subsystems in the same manner as during complete engine system tests and will therefore reduce development time and cost.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

Development Status of L75: A Brazilian Liquid Propellant Rocket Engine

The basic strategies for development and verification of the L75 engine are based on appropriate verification methods of the requirements, levels that the component occupies in the product tree, stage of development and definition for the development, engineering and qualification models. One or more models may be required to demonstrate compliance with specified requirements of the design, depending on the level of verification. Acceptance tests will be carried out on component, equipment and engine level. Prior to acceptance test, these items should be checked-out according to the defined assembly or integration procedures. To confirm adequacy of the design methodology of L75 engine, the construction of test facilities is in progress. The test benches in operation or under implementation are described hereafter.

483

HYDRAULIC TEST STAND The hydraulic test stand is a facility whose main objective is the characterization of LPRE components, using distilled water as working fluid. It consists basically of the specimens test area, the drive system consisting of pumps and electric motors, the distilled water storage tank, a filtration system, water cooling system and the data acquisition and control systems, which is able to achieve a water flow up to 30 kg/s, working pressure of 35 bar. The system power is 180 HP (132kW). Figure 16 shows the test area of the hydraulic test stand.

20 KN FIRING TEST STAND Designed for firing MFPLs up to 20 kN of thrust using pressurized tanks of LOx and ethanol as a propellant and with a combustion chamber pressure up to 100 bar. This test bench, which currently is used for testing the engines L5 and L15, was adapted for the L75 gas generator hot tests (Fig. 15).

PUMPS AND TURBINE COLD TEST STAND Pumps and turbine cold test stand will allow the testing of the L75 hydraulic pumps and turbine through the operation of individual pumps and turbine, control of components and data acquisition allowing the verification of functional requirements of these pumps and adequacy of the design of the component being tested. Figure 17 shows the scheme of the pumps and turbine cold test stand.

Figure 13. L75 main valve assembly.

Figure 15. 20 kN firing test stand.

Figure 14. Turbine starter hot test.

Figure 16. Hydraulic test stand. J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

484

Almeida, D.S. and Pagliuco, C.M.M.

Pumps

Gear box

Eletric motor

Valvwater reservois

Coriolis flowmeter

Valve Valve Valve

Filtro

enterprises and testing, but mainly driving the country on a new technological level to actually be able to open access to space. The next project critical activities are: manufacture of thrust chamber, turbopump, valves and regulators and cold and hot tests of these models. Usually, in a rocket engine development, from conception through qualification for flight, the cost of design and manufacture, which is the actual stage of L75 development, corresponds approximately to 20% of the total costs. Therefore, if Brazil indeed masters this long-term technology will be necessary more investments in human resources, infrastructure, manufacturing, etc.

Figure 17. Scheme of the pumps and turbine cold test stand.

FUTURE AND CONCLUSION.

ACKNOWLEDGES

The greatest merit of the L75 project is not only the development of the engine which is, in itself, already a great contribution to formation of a qualified team, implementation of infrastructure for design, manufacturing using qualified Brazilian

This paper was elaborated with support of a large number of members of the IAE and FUNDEP, and the authors wish to thank the team members. The authors thank AEB for the financial support.

REFERENCES ECSS – European Cooperation for Space Standardization, 2004, “ECSS-E-30 part 5.2 draft1 rev.1”, Propulsion for Launchers – Solid and Liquid. ECSS – European Cooperation for Space Standardization, 2009, “ECSS-M-ST-10C”, Project planning and implementation. Haeseler, D., Götz, A. and Fröhlichs, A., 2000, “Non-toxic propellants for future advanced launchers propulsion systems”, AIAA – 2000-3687. Jean, F. and Dalbies, E., 2000, “Development status of the VINCI engine for Ariane 5 upper stage”, AIAA/ASME/SAE/ASEEE 36th Joint Propulsion Conference and Exhibit, Hunstsville – USA. Niwa, M. and Yoshino, T., 1997, “Liquid Propulsion in Brazil”, COBEM 97.

PNAE – Programa Nacional de Atividades Espaciais 2012-2021: AEB Agência Espacial Brasileira, 2012, Ministério da Ciência e Tecnologia, Brasília, Brazil. Sutton, G.P., 2006, History of Liquid Propellant Rocket Engines, Virginia, USA: American Institute of Aeronautics and Astronautics, Inc., 911 p. Torres, M.F.C., Almeida, D.S., Krishna, Y.S.R., Silva, L.A. and Shimote, W.K., 2009, “Propulsão Líquida no IAE: Visão das atividades e perspectivas futuras”, Journal of Aerospace Technology and Management, Vol. 1, No. 1, pp. 99-106. doi: 10.5028/ jatm.2009.010199106.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.475-484, Oct.-Dec., 2014

485

ADHOC REFEREES Besides the participation of Editorial Board, the Journal of Aerospace Technology and Management had a collaboration of specialists as reviewers to evaluate the manuscripts. To them, the JATM thanks for the contribution in Vol. 6 (2014).

Adolfo Gomes Marto Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Carlos Ilário da Silva Embraer São José dos Campos/SP – Brazil

Eliton Souto de Medeiros Universidade Federal da Paraíba João Pessoa/PB – Brazil

Airton Nabarrete Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

Caroline Maria M. Mota Universidade Federal de Pernambuco Recife/PB – Brazil

Enda Bigarella Embraer São José dos Campos/SP – Brazil

Alexander Sukhanov Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Cayo Prado Fernandes Francisco Universidade Federal do ABC Santo André/SP – Brazil

Eusebio Valero Sanchez Universidad Politécnica de Madrid Madrid – Spain

Alexandre Nogueira Barbosa Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Cesar Pagan Universidade Estadual de Campinas Campinas/SP – Brazil

Fabiano L. Sousa Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Alfredo Rocha de Faria Instituto Tecnológico de Aeronáutica São José dos Campos/SP– Brazil

Claudia Gonçalves de Azevedo Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Filipe Pereira Pinto da Cunha e Alvelos Universidade do Minho Braga – Portugal

Aluisio Pinto Silva Faculdade de Engenharia de Guaratinguetá Guaratinguetá/SP – Brazil

Cristiano Alexandre V. Cavalcante Universidade Federal de Pernambuco Recife/PB – Brazil

Francisco das Chagas Carvalho Instituto de Ciência e Tecnologia São José dos Campos/SP – Brazil

Álvaro Martins Abdalla Escola de Engenharia de São Carlos São Carlos/SP – Brazil

Daniel Soares de Almeida Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Amilcar Porto Pimenta Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

Daniel Guerreiro e Silva Universidade Estadual de Campinas Campinas/SP – Brazil

Anderson Borille Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

Débora Albuquerque Vieira Universidade Federal da Paraíba João Pessoa/PB – Brazil

Annibal Hetem Jr Universidade Federal do ABC Santo André/SP – Brazil

Denilson Paulo S. Santos Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Antonio Alves Ferreira Júnior Instituto Nacional de Telecomunicações Santa Rita do Sapucaí/MG – Brazil

Diomadson Belfort Universidade Federal do Rio Grande do Norte Natal/RN – Brazil

Gerard Bobillot Office National d’Etudes et de Recherches Aérospatiales Châtillon – France

Antonio Mazzaracchio Sapienza University of Rome Rome – Italy

Don Fossey Roxel Propulsion Systems Kidderminster – UK

Gilson da Silva Instituto Nacional da Propriedade Industrial Rio de Janeiro/RJ – Brazil

Bento Silva de Mattos Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

Edson A. Araujo Querido Oliveira Universidade de Taubaté Taubaté/SP – Brazil

Gueorgui Smirnov Universidade Do Minho Braga – Portugal

Carlos Alberto Rocha Pimentel Universidade Federal do ABC Santo André/SP – Brazil

Eduardo dos Santos Pereira Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Guilherme Bertoldo Universidade Tecnológica Federal do Paraná Curitiba/PR – Brazil

Francisco José de Souza Universidade Federal de Uberlândia Uberlândia/MG – Brazil Francisco Santos Sabbadini Universidade do Estado do Rio de Janeiro Resende/RJ – Brazil Francisco Jasvier T. Salazar Faculdade de Engenharia de Guaratinguetá Guaratinguetá/SP – Brazil George dos Santos Marinho Universidade Federal do Rio Grande do Norte/RN Rio Grade do Norte/RN – Brazil

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.485-486, Oct.-Dec., 2014

486

Adhoc Referees – 2014

Hans Boden Royal Institute of Technology Stockholm – Sweden

Luciano Kioki Araki Universidade Federal do Paraná Curitiba/PR – Brazil

Renato Nunes Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Helen Stenmark EURENCO Karlskoga – Sweden

Luciene Dias Villar Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Ronaldo Rodrigo Ferreira Universidade Federal do Rio Grande do Sul Porto Alegre/RS – Brazil

Hilton Cleber Pietrobom Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Luis Cláudio Pimentel Universidade Federal do Rio de Janeiro Rio de Janeiro/RJ – Brazil

Rogerio Pirk Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Humberto Pontes Cardoso Equatorial Sistemas São José dos Campos/SP – Brazil

Luis Cláudio Rezende Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Rosamel Melita M. Riofano Escola de Engenharia de São Carlos São Carlos/SP – Brazil

Ijar Fonseca Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Marcelo Bender Perotoni Universidade Federal do ABC Santo André/SP – Brazil

Samuel Machado Leal da Silva Centro de Instrução de Guerra Eletrônica Brasília/DF – Brazil

Isabel Cristina dos Santos Universidade de Taubaté Taubaté/SP – Brazil

Marcilio Alves Universidade de São Paulo São Paulo/SP – Brazil

Sandro da Silva Fernandes Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

José Carlos Góis Universidade de Coimbra Coimbra – Portugal

Marcio da Silveira Luz Depto Ciência e Tecnologia Aeroespacial São José dos Campos/SP – Brazil

José Luis Gomes da Silva Universidade de Taubaté Taubaté/SP – Brazil

Marcio Teixera de Mendonça Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

Sara Silva Ferreira de Dafé Pontifícia Universidade Católica de Minas Gerais Contagem/MG – Brazil

Juan Manuel Pardal Universidade Federal Fluminense Niterói/RJ – Brazil

Marcos Loderllo Chaim Universidade de São Paulo São Paulo/SP – Brazil

Juan Pablo de Lima Costa Salazar Universidade Federal de Santa Catarina Florianópolis/SC – Brazil

Marcos V. T. Heckler Universidade Federal do Pampa Alegrete/RS – Brazil

Kleber Vieira Paiva Universidade Federal de Santa Catarina Joinville/SC – Brazil

Odenir de Almeida Universidade Federal de Uberlândia Uberlândia/MG – Brazil

Leonardo Henrique Gouvea Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Otávio Durão Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

Leonardo Sanches Universidade Federal de Uberlândia Uberlândia/MG – Brazil

Oyvind Hammer Johansen Chemring Nobel Saetre – Norway

Leticia Rittner Universidade Estadual de Campinas Campinas/SP – Brazil

Paul Bates Griffith University Queensland – Austrália

Luciano Bachmann Faculdade de Filosofia Ciencias e Letras de Ribeirão Preto Ribeirão Preto/SP – Brazil

Pedro Cobos Centro de Tecnologías Físicas Leonardo Torres Quevedo Madrid – Spain Pedro Teixeira Lacava Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.485-486, Oct.-Dec., 2014

Suzana França Dantas Daher Universidade Federal de Pernambuco Recife/PE – Brazil Tawfiqur Rahman Beihang University Beijing – China Ulisses T. Vieira Guedes Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil Vanderlei da Cunha Parro Instituto Mauá de Tecnologia São Caetano do Sul/SP – Brazil Volnei Tita Escola de Engenharia de São Carlos São Carlos/SP – Brazil Walter Miyakawa Instituto de Estudos Avançados São José dos Campos/SP – Brazil William G. Proud Imperial College London London – UK Willian Roberto Wolf Universidade Estadual de Campinas Campinas/SP – Brazil

INSTRUCTIONS TO AUTHORS (Revised in Dec., 2014) evaluators can accept the manuscript in the form it was submitted,

SCOPE AND EDITORIAL POLICY

they can reject it or request revisions. The manuscript that requires revision will be sent to the author that is supposed to submit a new

The Journal of Aerospace Technology and Management (JATM)

version and, in the case the author does not agree with the suggestions,

is the official publication of the Departamento de Ciência e Tecnologia

it is necessary to send a “letter to editor”, explaining the reasons. The

Aeroespacial (DCTA) and Associação Aeroespacial Brasileira (AAB), both

Editor will approve after verifying in the new version the adherence

in São José dos Campos, São Paulo State, Brazil. The journal is quarterly

to the reviewers’ suggestions or will send to another evaluation round

published (March, June, September and December) and is devoted to research and management on different aspects of aerospace technologies. The authors are solely responsible for the contents of their contribution. It is assumed that they have the necessary authority for publication. Topics of interest are: Acoustics, Aerodynamics, Aerospace meteorology, Applied computation, Astrodynamics, Ceramic materials, Circuitry, Composites, Computational fluid dynamics, Defense systems, Energetic materials, Fluid dynamics and turbulence, Guidance navigation and control, Management systems, Metallic materials, Photonics, Polymeric materials, Processing of aerospace materials, Propulsion and combustion, Radars and tracking systems, Robotics and automation, Structures,

if the changes have not been sufficiently addressed. Accepted manuscripts can be edited to comply with the format of the journal, remove redundancies, and improve clarity and understanding without altering meaning. Authors are also strongly advised to use abbreviations sparingly whenever possible to avoid jargon and improve the readability of the manuscript. All abbreviations must be defined the first time that they are used. The edited text will be presented to authors for approval.

MANUSCRIPT CATEGORIES

Synthesis and characterization of aerospace materials, Thermal sciences and Vibration and structural dynamics.

Editorial: Any researcher may write the editorial on the invitation of the Editor in Chief. Editorials should cover broad aspects of Aerospace Technology. Such manuscripts are not submitted to peer review.

PAPER SUBMISSION

Review articles: These should cover subjects that are relevant to

The manuscript should be digitalized using a Microsoft Word (.doc) software program and submitted electronically in English. PDF files are also accepted. See the instructions at http://www.jatm.com.br/ojs/index. php/jatm/about/submissions#onlineSubmissions. If there is any conflict of interest regarding the evaluation of the manuscript, the author must send a declaration indicating the reasons so that the review process occurs fairly. Papers already presented at conferences will be accepted if they were not published in complete form in the Annals of the conference or if they

the scope of the journal. Authors should bear in mind that they are expected to have expertise in the reviewed field. The article may be of any length required for the concise presentation of the subject. Original papers: These articles should report results of the scientific research. The article may be of any length required for the concise presentation and discussion of the data, but succinct papers are favored in terms of impact as well as in readability. Communications: They should report previous results of ongoing research and should not exceed eight pages.

are extended with additional results or new findings. These articles will be evaluated as the others. Articles from guest authors will be published after approval of one specialist associate editor. The JATM does not publish translated articles from other journals.

MANUSCRIPT STRUCTURE Whenever is possible, articles should include the following subsections, however articles from some areas should follow their usual

PEER REVIEW

format. Title and names of authors: The title should not contain

Manuscripts will be reviewed by at least two expert consultants,

abbreviations. All authors should be identified with full name, e-mail,

members of the Editorial Committee or external evaluators (ad

institution to which they are related, city, state, and country. One of them

hoc referees) in double blind peer review mode, ensuring complete

should be indicated as the author for correspondence and his/her full

anonymity. In case of disagreement on the results of the evaluation,

address is required.

the manuscript will be forwarded to a third reviewer, and it will

Abstract: They are limited to 250 words and structured into

be accepted for publication only if two approvals are received. The

objectives, methods, results, and conclusions. Citations or abbreviations

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.487-488, Oct.-Dec., 2014

(except internationally recognized abbreviations, such as weights, measures, and physical or chemical ones) are not permitted. Keywords: Three to six items that should be based on NASA Thesaurus volume 2 – Access Vocabulary. Introduction: It should set the purpose of the study, providing a brief summary (not a review) of previous relevant studies, and stating the new advances in the current investigation. The introduction should not include data or conclusions from the work being reported. A final sentence summarizing the novel finding to be presented is permissible. Methodology: The authors are free to use any structure in this section to fit the objectives of the work, they could also rename it (e.g. Numerical analysis, Case study, and so on), and in some cases it may be advisable to omit it. Clear and sufficient information to permit the study to be repeated by others should be briefly given. Standard techniques need only to be referenced. Previously published methods may be briefly described following the reference. Results: This section should be a concise account of the new information that was discovered, with the least personal judgment. Do not repeat in text all the data in the tables and illustrations, but briefly describe what these data comprise. Discussion: The discussion should include the significance of the new information and relevance of the new findings in light of existing knowledge. Only unavoidable citations should be included. Citations to review articles are not encouraged in this section. In some cases, it may be advisable to merge with the previous section (“Results and Discussion”). Acknowledgements: This section should be short, concise, and restricted to acknowledgements that are necessary. The financial support received for the elaboration of the manuscript must be declared in this item. References: Acceptable references include journal articles, numbered papers, books, and submitted articles, if the journal is identified. References must be restricted to directly relevant published works, papers, or abstracts that have been accepted for publication. References from private communications, dissertations, thesis, published conference proceedings, and preprints from conferences should be avoided. Selfcitation should be limited to a minimum. Authors are responsible for the accuracy and completeness of their references. References in text: The references should be mentioned in the text by giving the last name of the author(s) and the year of publication. Either use “Recent work (Smith and Farias, 1997)” or “Recently Smith and Farias (1997)”. With three or more names, use the form “Smith et al. (1997)”. If two or more references have the same identification, distinguish them by appending “a”, “b”, etc., to the year of publication. Standards should be cited in text by the acronym of entity followed by the number, and they do not need to appear in the reference list. Reference list: References should be listed in alphabetical order, according to the last name of the first author, at the end of the article. Only citations that appear in the text should be referenced. Unpublished papers, unless accepted for publication, should not be cited. Work that is accepted for publication should be referred to as “In press”. It is recommended that each reference contains the digital object identifier number (DOI). References retrieved from the Internet should be cited by

the last name of the author(s) and the year of publication, or n.d., if not available, followed by the date of access. Some examples of references are as the following ones: Coimbra, A. L., 1978, “Lessons of Continuum Mechanics”, Ed. Edgard Blücher, São Paulo, Brazil, 428 p. Alves, M.B. and Morais, A.M. F., 2009, “The management of knowledge and technologies in a Space Program”, Journal of Aerospace Technology and Management, Vol. 1, No. 2, pp. 265-272. doi:10.5028/ jatm.2009.0102265272 Paek, S.K., Bae, J.S. and Lee, I., 2002, “Flutter Analysis of a Wraparound Fin Projectile Considering Rolling Motion,” Journal of Spacecraft and Rockets, Vol. 39, No. 1, pp. 66-72. Bae, J.S., Kim, D.K., Shih, W.H., Lee, I. and Kim, S.H., 2004, “Nonlinear Aeroelastic Analysis of a Deployable Missile Control Fin,” Journal of Spacecraft and Rockets, Vol. 41, No. 2, pp. 264-271. Clark, J.A., 1986, “Private Communication”, University of Michigan, Ann Harbor. EMBRAPA, 1999, “Politics of R&D”, Retrieved in May 8, 2010, from http://www.embrapa.br/publicacoes/institucionais/polPD.pdf Silva, L.H.M., 1988, “New Integral Formulation for Problems in Mechanics” (In Portuguese), Ph.D. Thesis, Federal University of Santa Catarina, Florianópolis, S.C., Brazil, 223p. Sparrow, E.M., 1980a, “Forced Convection Heat Transfer in a Duct Having Spanwise-Periodic Rectangular Protuberances”, Numerical Heat Transfer, Vol. 3, pp. 149-167. Sparrow, E.M., 1980b, “Fluid-to-Fluid Conjugate Heat Transfer for a Vertical Pipe-Internal and External Natural Convection”, ASME Journal of Heat Transfer, Vol. 102, pp. 402-407. Tables: Tables should be numbered in order of appearance in the text, using Arabic numerals. Each table should have a title and an explanatory legend, if necessary. All tables must be referenced and mentioned in the text as “Table” and succinctly described in the text. Under no circumstances should a table repeat data that are presented in an illustration. Statistical measures of variation (i.e., standard deviation or standard error) should be identified, and decimal places in tabular data should be restricted to those with mathematical and statistical significance. Figures: All illustrations, line graphs, charts, schemes, photographs, and graphs should be referred as “Figure” and submitted with good definition. Number figures consecutively using Arabic numerals in order of appearance. References should be made in the text to each figure using the abbreviated form “Fig.”, except if they are mentioned in the beginning of the sentences. Captions should be descriptive and should allow the examination of the figures, without reference to text. The size of the figures (including frame) should be 8 cm (one column) or 17 cm (two columns) wide, with maximal height smaller than 22 cm. Equations: Type them on individual lines, identifying them by Arabic numerals enclosed in parenthesis. References should be made in the text to each equation using the abbreviated form “Eq.”, except in the beginning of the sentences, where the form “Equation” should be used.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.6, No 4, pp.487-488, Oct.-Dec., 2014

• LIVRE • PERIÓDICOS CAPES • PKP • REDALYC • SCOPUS • SOCOL@R • SUMÁRIOS.ORG • ULRICHSWEB

JATM IS SUPPORTED BY:

Journal of Aerospace Technology and Management

JOURNAL OF

AEROSPACE TECHNOLOGY AND MANAGEMENT Vol. 6 N. 4 Oct./Dec. 2014 ISSN 1984-9648 ISSN 2175-9146 (online)

www.jatm.com.br

V. 6, n. 4, Oct./Dec., 2014

Journal of Aerospace Technology and Management