Vol.5 N.3 - 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. 5 N. 3 Jul./Sep. 2013 ISSN 1984-9648 ISSN 2175-9146 (online)

www.jatm.com.br

V.5, n. 3, Jul./sep., 2013

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; SCOPUS - Elsevier; CAS - Chemical Abstracts Service; DOAJ - Directory of Open Access Journals; J-GATE - The e-journal gateway from global literature; LIVRE - Portal to Free Access Journals; GOOGLE SCHOLAR; SUMÁRIOS.ORG - Summaries of Brazilian Journals; EZB- Electronic Journals Library; ULRICHSWEB- Ulrich´s Periodicals Directory; CLASE/PERIÓDICA- Indice de Revistas Latinoamericanas en Ciencia; SOCOL@R- China Educational Publications; LATINDEX-Regional Cooperative Online Information System for Scholarly Journals from Latin America, the Caribbean, Spain and Portugal; REDALYC - Red de Revistas Científicas de América Latina y el Caribe, España y Portugal; EBSCO Publishing and PERIÓDICOS CAPES. 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 affiliated to ABEC - Brazilian Association of Scientific Editors.

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- 6493/5122 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

JATM is supported by:

Journal of Aerospace Technology and Management Vol. 5, n.3 (Jul./Sep. 2013) – São José dos Campos: Zeppelini Editorial, 2013 Quartely issued Aerospace sciences Technologies Aerospace engineering CDU: 629.73

Historical Note: JATM was created in 2009 after the iniciative of the diretor of Instituto de Aeronáutica e Espaço (IAE), Brigadeiro Engenheiro Francisco Carlos Melo Pantoja. In order to reach the goal of becoming a journal that could represent knowledge in science and aerospace technology, JATM searched for partnerships with others institutions in the same field from the beginning. 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. The copyright on all published material belongs to Departamento de Ciência e Tecnologia Aeroespacial (DCTA)

ISSN 1984-9648 ISSN 2175-9146 (online)

Journal of Aerospace Technology and Management Vol. 5 No. 3 - Jul./Sep. 2013 Editor 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 de Aeronáutica e Espaço 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

Marco A. Sala Minucci Vale Soluções em Energia São José dos Campos/SP – Brazil

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

Francisco Carlos M. Pantoja Diretoria de Engenharia da Aeronáutica Rio de Janeiro/RJ – Brazil

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

Carlos Antônio M. Kasemodel 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

Paulo Tadeu de Melo Lourenção EMBRAER São José dos Campos/SP – Brazil

Carlos de Moura Neto Instituto Tecnológico de Aeronáutica São José dos Campos/SP – Brazil

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

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

Acoustics

Applied computation

Ceramic Materials

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

José Márcio Machado Instituto de Biociências, Letras e Ciências Exatas São José do Rio Preto/SP – Brazil

José Maria Fonte Ferreira Universidade de Aveiro Aveiro – Portugal

Aerodynamics

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

Circuitry

SCIENTIFIC COUNCIL

ASSOCIATE EDITORS Bert Pluymers Katholieke Universiteit Leuven Leuven – Belgium

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

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

Astrodynamics

Antonio Sergio Bezerra Sombra Universidade Federal do Ceará Fortaleza/CE – Brazil

Altamiro Susin Universidade Federal do Rio Grande do Sul Porto Alegre/RS – Brazil

Antonio F. Bertachini Instituto Nacional de Pesquisas Espaciais São José dos Campos/SP – Brazil

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

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

Composites

Edson Cocchieri Botelho Faculdade de Engenharia de Guaratinguetá Guaratinguetá/SP – Brazil Flamínio Levy Neto Universidade de Brasília Brasília/DF – Brazil

Computational fluid dynamics

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

Defense Systems

Adam S. Cumming Defence Science and Technology Laboratory Salisbury/Wiltshire – England Wim P. C. de Klerk Netherlands Organisation for Applied Scientific Research Rijswijk/SH – Netherlands

Eletromagnetic Compatibility

Alain Azoulay École Supérieure d’Electricité Gif–Sur–Yvette – France Cynthia Junqueira Instituto de Aeronáutica e Espaço São José dos Campos/SP – Brazil

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

Guidance, Navigation and Control

Radars and Tracking Systems

David Murray–Smith University of Glasgow Glasgow – Scotland

Marc Lesturgie Office National d’Etudes et de Recherches Aérospatiales Palaiseau – France

Daniel Alazard Institut Supérieur de l’Aéronautique et de l’Espace Toulouse – France

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

Management Systems

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

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

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

Structures

Metallic Materials

José Rubens G. Carneiro Pontifícia Universidade Católica de Minas Gerais Belo Horizonte – Brazil

Photonics

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 Aeronáutica e Espaço São José dos Campos/SP – Brazil

Fluid Dynamics and Turbulence

Processing of Aerospace Materials

Vassilis Theofilis Universidad Politécnica de Madrid Madrid – Spain

Robotics and Automation

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

José Leandro Andrade Campos Universidade de Coimbra Coimbra – Portugal

Marcos Pinotti Barbosa Universidade Federal de Minas Gerais Belo Horizonte/MG – Brazil

Hugo H. Figueroa Universidade Estadual de Campinas Campinas/SP – Brazil

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

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

Valder Steffen Junior Universidade Federal de Uberlândia Uberlândia/MG – Brazil

Carlos Henrique Marchi Universidade Federal do Paraná Curitiba/PR – Brazil

Editorial Production Glauco da Silva 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

Helena Prado A.Silva Instituto de Aeronáutica e Espaço 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.5, No 3, 2013

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

CONTENTS Editorial 265 Spin-off Companies: A Great Challenge for the Brazilian Space Program Marcia Mantelli REVIEW ARTICLE 267 An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012 Simone Carvalho Rufino, Gilson da Silva, Koshun Iha ORIGINAL PAPERS 279 Experimental Study of Polyurethane-Based Fuels with Addition of Paraffin and Aluminum for Hybrid Rocket Motors Susane Ribeiro Gomes, Leopoldo Rocco Junior, José Atílio Fritz Fidel Rocco, Koshun Iha 287 Combustion Modeling of Aluminum Incorporated in Low-Explosive Formulations such as Solid Propellants Rene Francisco Boschi Gonçalves, Koshun Iha, José Atílio Fritz Fidel Rocco 293 Kick Solid Rocket Motor Multidisciplinary Design Optimization Using Genetic Algorithm Fredy Marcell Villanueva, He Linshu, Xu Dajun 305 Simulations of the Atmospheric Boundary Layer in a Wind Tunnel with Short Test Section Luciana Bassi Marinho Pires, Igor Braga de Paula, Gilberto Fisch, Ralf Gielow, Roberto da Mota Girardi 315 Verification of Response of Neutron Monitor for In-Flight Neutron Dosimetry Claudio Antonio Federico, Odair Lélis Gonçalez, Evaldo Simões da Fonseca, Karla Cristina de Souza Patrão, Marlon Antonio Pereira, Linda Viola Ehlin Caldas 323 Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software Ronaldo Rodrigues Ferreira, Rafael Baldiati Parizi, Luigi Carro, Álvaro Freitas Moreira 335 Analysis of Total Ionizing Dose Effects on 0.13 μm Technology-Temperature-Compensated Voltage References Thiago Hanna Both, Dalton Colombo, Ricardo Vanni Dallasen, Gilson Inácio Wirth 341 Nonlinear Characteristics of Revolute Joints with Clearance Liu Rong-qiang, Zhang Jing, Guo Hong-wei, Deng Zong-quan 349 Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane Bento Silva de Mattos, Ney Rafael Secco, Eduardo Francisco Salles 362 INSTRUCTIONS TO AUTHORS

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, 2013

CORRECTIONS J. Aerosp. Technol. Manag. Vol.5, No2, pp. 170, Apr.-Jun., 2013 Equation 1, instead of:

⎧ дρ ⎜ дt + · (ρv) = 0 ⎜ ⎜ ⎨ д (ρv) + · (ρvv) = − p + · ( = τ) ⎜ дt ⎜ д ⎜ τ · v)] + Q (ρE) + · [v (ρE + p)] = · [k T + ( = ⎩ дt ∆

∆

Insert:

⎧ дρ ⎜ дt + · (ρv) = 0 ⎜ ⎜ ⎨ д (ρv) + · (ρvv) = − p + · ( = τ) ⎜ дt ⎜ д ⎜ τ · v)] + Q (ρE) + · [v (ρE + p)] = · [k T + ( = ⎩ дt ∆

∆

J. Aerosp. Technol. Manag. Vol.5, No2, pp. 241, Apr.-Jun., 2013 doi number, instead of: doi: 10.5028/jatm.v5i2/193 Insert: doi: 10.5028/jatm.v5i2.193

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

Editorial Spin-off Companies: A Great Challenge for the Brazilian Space Program Marcia Mantelli1

pace science research and development is, by itself, a great challenge for research institutions throughout the world. This is especially true for developing countries such as Brazil, where the expenditures and number of institutions devoted to this field of study are limited. Additionally, because most funding for space research programs often originates from government sources, these are usually one of the first to be cut in times of economic distress. On the other hand, economy studies of the main space programs in the world show that the return on investment to the society in terms of new technologies greatly surpasses the research funds. Therefore, one of the greatest challenges for governments of developing countries with their own space programs is to create favorable conditions, so that the investments become profitable to the population, providing not only technological independence in the space field but also technological solutions to daily problems. Development of the heat pipe technology is a good illustration of the aforementioned tradeoff. Heat pipes are highly efficient, low cost passive heat transfer devices, used for the thermal management of equipment and panels in satellites. The first registered patents are concerned with application of these devices in domestic and bakery applications, dating back to the middle of the 20th century. Although very efficient, this technology was not largely employed until the Cold War, when the major countries intensified their space exploration

programs, leading to the need for reliable space-applicable technologies. In case of Brazil, heat pipe research and development started at the Instituto Nacional de Pesquisas Espaciais (INPE) in the late 1980s to provide thermal solutions for satellites. At that time, several cooperation agreements among Brazilian institutions devoted to spacerelated activities; Instituto Nacional de Pesquisas Espaciais (INPE), Departamento de Ciência e Tecnologia Aeroespacial (DCTA), Instituto de Aeronáutica e Espaço (IAE), etc., and Brazilian Universities; Universidade Federal do Rio de Janeiro (UFRJ), Universidade Federal de Santa Catarina (UFSC), Universidade de São Paulo (USP), etc., were established, aiming to promote research and development of spacecraft subsystems (satellites and rockets) as well as of onboard equipment and software. This effort allowed fulfillment of the objectives of the Brazilian Space Program and created a highly qualified workforce at INPE and other involved institutions. In this context, INPE and UFSC established their first common heat pipe development project. In the 1990s, the Brazilian Space Agency (AEB) was established and new contracts were signed between AEB and Brazilian Universities, allowing the establishment of reference laboratories and development of several productive research groups in Brazil. Although the Brazilian Space Program is still modest in comparison with other existing programs, several satellites have been successfully launched. For several years, programs

1.Professor at Department of Mechanical Engineering/Universidade Federal de Santa Catarina/Brazil, and coordinator of the research and development activities of the Laboratory of Heat Pipes. She was the first woman to graduate as a Mechanical Engineer from UNICAMP, Brazil in 1983 and received the degree of Master in Space Science by the Instituto Nacional de Pesquisas Espaciais, Brazil, where she worked from 1886 to 1999. She received her Ph.D. from the University of Waterloo, Canada in 1995. Approximately 50 research projects on the development of equipment for industry with heat pipes and thermosyphons were coordinated or are in progress, involving approximately 40 masters and doctoral students. She has published around 30 articles in specialized journals, and presented more than 100 conference articles. She also participates in editorial board and as a reviewer of several journals worldwide. She received, among others, the Claudia Award, Editora Abril, in the Sciences category in 2012. Email: marcia@labtucal.ufsc.br

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.265-266, Jul.-Sep., 2013

266

Mantelli, M.

such as “Microgravity” and “Uniespaço”, coordinated by AEB, have been providing space-related research opportunities for many different fields of science in Brazil. As a “side effect” of the Brazilian Space Program, many laboratories received a strong investment from the national industry, aiming the application of space technology to solve a large spectrum of practical problems. Additionally, in the last three decades, a highly qualified workforce in space-related subjects, mainly composed by engineers, was formed. Again, the heat pipe technology can be cited as an example, as this technology is being considered to be employed to thermal management in the petroleum, aeronautics, electronics and food industries, particularly when low cost and highly thermally efficient equipment are required. As the industry becomes increasingly convinced of the applicability of novel space-related technological solutions, a new problem emerges: who will produce the equipment to fulfill the needs of this emergent market?

As a developing country, Brazil has a growing number of highly qualified well-trained people, but it does not have a strong tradition in starting-up new technological companies, although incubators have being active for at least two decades. Furthermore, some of the start-up companies have a short life. Thus, the start-up and growth of new strong technologybased companies is a great challenge for governments of developing countries such as Brazil. In conclusion, the start-up of new technology-based companies is a “side effect” of space programs that should be taken into consideration when national space programs are being established. Considering the larger and recently established United States IT enterprises as an example, one can conclude that space program spin-off companies have large potential to be highly successful if their start-up process is well conducted.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.265-266, Jul.-Sep., 2013

doi: 10.5028/jatm.0501. doi: 10.5028/jatm.v5i3.242

An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012 Simone Carvalho Rufino1, Gilson da Silva1, Koshun Iha2

ABSTRACT: The purpose of this article is to present a study on the technological development of propellants, which are used in solid and hybrid rocket motor, that have been employing hydroxyl-terminated polybutadiene (HTPB) as a binder, for the past ten years. The results prove that major research conducted on propulsion technologies continues using HTPB as a binder, with China and the United States being the countries with greater publications, and Brazil appearing at the fifth position in the ranking. The scientific and technological information was collected from articles, conference papers, reviews (over the last ten years) and patents granted in the United States by scientific and USPTO patent databases. KEYWORDS: Propellant, Hydroxyl-terminated polybutadiene (HTPB), Patent, Article, Conference paper, Review.

INTRODUCTION This article presents a review on the technological development behavior of hydroxyl-terminated polybutadiene (HTPB)-based propellants used in solid and hybrid rocket motor during the last decade, using scientific and technological information collected from articles, conference papers, reviews and patents granted by the US patent and trademark office (USPTO) database. Companies and research institutions have been increasingly dedicated to technological forecasting as a tool for technology monitoring, investigation of new fields of investment and for minimizing the risks involved in research projects. The management of information and knowledge, in addition to excluding the subjectivity of decisions, provides technical support (Ortiz et al., 2002). They are looking for trend indicators such as the characteristics of scientific and technological production (year, country), key players involved in the market (competitors and/or potential partners), maturity of technology, development patterns, and correlated technologies, among others, which often indicate the length of active and emerging technologies. Papers are responsible for the dissemination of scientific research, and their results corroborate existing studies and inspiration for further research. Patents, in turn, are excellent indicators of innovation, as they represent the industrial application of the results of research and development; in other words, the ability of a country to transform scientific knowledge into technological product or results.

Universidade Federal do Rio de Janeiro – Rio de Janeiro/RJ – Brazil 1.Instituto Nacional da Propriedade Industrial – Rio de Janeiro/RJ – Brazil 2.Instituto Tecnológico de Aeronáutica – São José dos Campos/SP – Brazil Author for correspondence: Jules Ghislain Slama | Departamento de Engenharia Mecânica/COPPE/UFRJ/C.P. 68.503 | CEP 21.945-970 Rio de Janeiro/RJ – Author Brazil | for correspondence: Gilson da Silva | Instituto Nacional da Propriedade Industrial | Praça Mauá, 7 – Centro | CEP 20.081-240 Rio de Janeiro/RJ – Brazil |julesslama@yahoo.com.br Email: gilsondasilva@uol.com.br email: Received: 02/02/12 19/03/13 || Accepted: Accepted:17/05/13 Received: 30/10/12

o 3,X, Aerosp. Technol. Manag., São José dosCampos, Campos, Vol.X, No pp.1-12, XXX.-XXX., 2013 J.J.Aerosp. Technol. Manag., José dos Vol.5, pp.267-278, Jul.-Sep., 2013 J. Aerosp. Technol. Manag., São JoséSão dos Campos, Vol.X, No NX, pp.1-8, XXX.-XXX., 2013

268

Rufino, S.C., Silva, G. and Iha, K.

In the particular case of patents, they may not represent the real technological scenario, as, in general, these documents are considered strategic information for the country and the applicant chooses to make the patent application in secrecy, as it is in Brazil, and nobody can read the contents of this application (LPI nº 9279/96, cap. IX) (PN nº 1888/MD). For these reasons, the results in this article represent only the patents that are available to society from the United States (USPTO). Particularly, the choice by the U. S. patent database is due to the policy adopted by the United States to replace the shuttles used until then. Another important aspect that should be considered in the search for technological information, is that scientific documents represent more current information, once the publication of research in journals and congress occurs immediately. On the other hand, a usual patent is kept in secret for 18 months, except when an applicant requires an anticipated publish date.

PROPELLANTS Propellants are chemical compounds (normally, homogeneous) or mixtures (normally heterogeneous) that on ignition, exhibit self-sustained combustion and generate large volumes of hot gases at controlled, predetermined rates (Miskelly, 2004). The specific chemical composition depends on the desired combustion characteristics for a particular application (Beckstead et al., 2007), and propellants can be tailored so as to achieve desired burning rate characteristics (Miskelly, 2004). The main selection parameters are the specific impulse (Isp), the thrust, the flight speed, the simplicity of the technology, the storage ability, the safety and the reliability (Gascoin et al., 2012), and low pollutants in combustion. Propellants serve as a convenient, compact form of storing relatively large amounts of energy and working fluid for rapid release and enjoy wide utility in various industrial and military applications. Thus, propellants are generally employed in various situations requiring a readily controllable source of energy such as ballistic applications (e.g., for periods of time ranging from milliseconds in weapons to minutes for space vehicles) in which the generated gases function as a working fluid for propelling projectiles, for example rockets and missile systems, and also for pressurizing pistons and inflating containers (Miskelly, 2004).

Propellants can be divided basically into three groups: solid, in which all components used in the formulation are solid; hybrid, which usually has a solid fuel and a liquid or gas oxidizer; and liquid, which is characterized in that the fuel and oxidant are liquid, typically stored in separate reservoirs, and only react when mixed together inside the combustion c hamber (Paterlini et al., 2002). In cases where the HTPB is the most common binder for solid and hybrid propellants, the results of this article will not involve liquid propulsion. SOLID PROPELLANTS A solid rocket motor (SRM) or a composite propellant rocket motor is a rocket with a motor that uses solid propellants comprising a fuel and an oxidizer. The solid propellant is usually in the form of a propellant grain placed within the interior of the rocket motor (e.g. in the combustion chamber) and burned to produce hot gases, which, in turn, exit through the throat and nozzle of the rocket motor at a high velocity to provide a thrust that propels the rocket in the opposite direction (Petersen et al., 2012). It is widely used in tactical and strategic missiles (Rao et al., 2004) and considering the condition in which its constituent ingredients are interconnected, can be classified as homogeneous and heterogeneous (Beckstead et al., 2007). In a homogeneous propellant, the ingredients are linked chemically and the resulting physical structure is homogeneous throughout. Typical examples of homogeneous propellants are single-base (NC- nitrocellulose) or double-base (NC and NGnitroglycerine) propellants (Beckstead et al., 2007). In heterogeneous or composite propellants, the ingredients are physically mixed, resulting in a heterogeneous physical structure (Beckstead et al., 2007). These energetic materials are made by embedding finely divided crystalline oxidizer particles, such as ammonium perchlorate (AP), in a resinous or elastomeric matrix, based on HTPB, called a “binder”. The binder usually provides the fuel for the combustion reaction, although solid-reducing agents, such as aluminum (Al), are frequently included in the compositions (La Fuente, 2009). Rocket propellant formulations based on AP/Al/HTPB are extensively used, as they enable high performances to be achieved with a high specific gravimetric or specific impulse (Isp) and volumetric impulse (Ispv) (Bohn and Cerri, 2010). Formulations of solid propellants also contain other chemical ingredients such as ballistic modifiers, bonding agents, plasticizers, curing agents, stabilizers and crosslinking agents (Beckstead et al., 2007).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012

HYBRID PROPELLANTS Hybrid propulsion is a promising technology, as it is safer, can have a higher average specific impulse, and is environmental friendly. Hybrid propellants can be easier throttled and restarted than solid propellants and are less expensive and complex than liquid propellants in terms of development and management of the propulsive system (Guobiao et al., 2011; Gomes, 2012). A typical hybrid rocket motor (HRM) uses a liquid oxidizer and a solid fuel propellant. The motor structure consists of a liquid oxidizer storage, a feed system and a solid fuel thrust chamber (Guobiao et al., 2011). This technology involves the burning of the solid fuel contained in the combustion chamber with the injected oxidizer, either liquid or gaseous (heterogeneous combustion). The separation of the propellant components produces a very safe engine and by controlling the injection via a regulation valve, it is possible to adjust the thrust, to shut the engine down and to reignite it later if necessary (Gascoin et al., 2012). The choice of the oxidizer is more limited, as it should be non toxic, safe, and present a high burning rate. The three most frequently used oxidizers are oxygen (liquid LOx or gas GOx), hydrogen peroxide (H2O2), and nitrous oxide (N2O) (Gascoin et al., 2012). Limitations are still faced with regard to solid fuel, such as low regression rate and combustion inefficiency, causing unproven capability of large rocket operations. HTPB is also largely studied, because it is one of the most frequently used in solid propulsion but while offering good mechanical properties, it presents low regression rates. A recommended strategy for HTPB-based solid fuels is to increase regression rates by fluid dynamics means (swirling flows); high-energy metallic additives can be used to sensibly increase density and specific impulse while contrasting nozzle erosion and further augmenting regression rates (DeLuca et al., 2013). The addition of metal particles (Al, Li, Mg and B) as well as metal hydrides (AlH3 and others) is investigated as a means of increasing their combustive and thermomechanical properties (notable increase of regression rate), such as the flame temperature and the density (Gascoin et al., 2012). Different from SRM, the solid fuel regression rate in HRM is predominantly driven not by the chamber pressure but by the oxidizer mass flow rate. At a constant oxidizer mass flow rate for a simple cylindrical-grain motor, the thrust value descends with time into the firing, as the grain port diameter

269

(equivalent diameter if not a circle port) and area increase, thus reducing the mass flux- dependent fuel regression rate. The thrust affects the flight trajectory directly. Therefore, the optimal design of an HRM might be better established if undertaken in the context of meeting a rocket vehicle’s particular mission requirements (Guobiao et al., 2011). Studies show that the high-concentration H2O2 exhibits greater performance than N2O and is easier to handle than liquid oxygen in HRMs. The application of H2O2/HTPB propellant combination in the research field of HRM is becoming popular with the development of H2O2 concentration and handling technology. There are two standard concentrations for high concentration H2O2: 98 and 90%. Besides the initial grain shape parameters, the HRM and rocket vehicle’s performance are determined by four initial design parameter values, including initial thrust, chamber pressure, mixture ratio and nozzle expansion ratio (Guobiao et al., 2011). GREEN PROPELLANTS The great problem faced by most of the currently used solid rocket fuels is that they are still based on mixtures of AP, aluminum (Al) and resins, which produces hydrochloric acid (HCl) and aluminum oxides during combustion. Therefore, AP-based propellant formulations are not considered green propellants. The quest for more energetic propellants with reduced pollutant emissions has resulted in the use of several non-AP ingredients in solid propellants. The ingredients belong to a wide spectrum of chemical families, but mostly fall into one of the following four categories: • Nitramines: cyclotrimethylenetrinitramine (RDX), cyclotetramethylenetetranitramine (HMX), the known as CL-20; hexanitrohexaazaisowurtzitane (HNIW), hydrazinium nitroformate (HNF) • Azides: glycidyl azide polymer(GAP), 3,3′-bis (azidomethyl) oxetane (BAMO), 3-azidomethyl-3-methyloxetane (AMMO) • nitrate esters: NG, NC, 1,2,4-butane triol trinitrate (BTTN), metriol trinitrate (TMETN), diethylene glycol dinitrate (DEGDN) • Nitrates: ammonium dinitramide (ADN), ammonium nitrate (AN) (Beckstead, 2007; Badgujar et al., 2008). Cyclic nitramines such as RDX and HMX have received significant attention as energetic components of modern eco-friendly propellants by virtue of their positive heat of formation (+58.5 and +75 kJ/mol, respectively), superior

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

270

Rufino, S.C., Silva, G. and Iha, K.

chemical and thermal stability, as well as the presence of reduced pollutants in combustion products and non-smoky exhaust similar to that of double-based propellants. Since RDX and HMX contain a high concentration of hydrogen atoms, the combustion of nitramine produces lowmolecular-weight gases. Therefore, they produce relatively high specific impulses even though the adiabatic flame temperature is low. Furthermore, the combustion products are not only smokeless but also noncorrosive (Naya and Kohga, 2013). The use of HNF has also received attention due to its benefit in terms of clean combustion and superior heat of formation (-71 kJ/mol) in relation to AP oxidizer (Badgujar et al., 2008). In addition, the use of energetic azides, mainly GAP, in propellant formulations reduces the amount of flame and smoke in the exhaust gases, thus making the propellant formulations more eco-friendly (Badgujar et al., 2008). The use of ADN instead of ammonium perchlorate certainly increases the energy of the propellant formulation and the burning rate of propellants. ADN exhibits superior heat of formation (-151 kJ/mol) and is prepared from aliphatic monoisocyanate using stoichiometric quantities of nitronium tetrafluoborate and nitric acid in acetonitrile as the nitrating medium followed by ammonia treatment. The NH4+ or K+ salts of dinitramide are more stable compared with dinitramines. So, ADN combines the low sensitivity of ammonium salts with the high burning rates of nitramines. It has a low melting point, is slightly hygroscopic, and, hence, needs special process conditions (Badgujar et al., 2008).

the matter to be protected or the invention itself, and is, therefore, the field in which one should seek topics of interest. The search was done considering documents in which both keywords “propellant or propellants” and “polybutadiene or HTPB”, are present in the title, abstract, or keywords, in the Scopus database. In the Science Direct database, the search was done considering documents in which one of the words “htpb” or “polybutadiene” were present in the title, abstract or keywords and the word “propellant” was present in the whole text. The search in the Web of Science database was done considering documents in which one of the words “htpb” or “polybutadiene” were present in the title and the word “propellant” was present in the whole text. Document publication date was limited to the range from January 1, 2002 to December 31, 2012. Documents recovered, from reading the abstracts, were selected while excluding overlaps and considering only the ones related to studies on SRM and hybrid rocket motor (HRM), which used HTPB as a binder and investigated the properties of the propellant. The results were compiled in terms of number of documents, document type (article, conference paper and review), assignee countries, affiliation (universities and research institutes) and the main subjects or properties studied in these documents, considering the title and the abstract. The following documents recovered by this search strategy were excluded: those in which the compositions in which HTPB was blended with a single-base propellant were described; those in which ignition models and ignition delay were described; and those in which the compositions in which HTPB was used in waterjet motors were described.

METHODOLOGY Articles, conference papers, and reviews used in the present work were recovered using Scopus, Science Direct and Web of Science databases, which are the usual databases in a scientific community. Patents were recovered from the USPTO Patent Full Text and Image Database (PatFT). The choice of the USPTO database was based on the economic potential of the market in the United States, already recognized worldwide and supported by a large number of scientific publications in the area of interest of this study. In addition, this database is the one that allows free access to the search, specifically in the claims. In the case of patent documents, this field describes

RESULTS AND DISCUSSION From the search, 346 relevant documents were recovered and from that, 30 publications are related to green propellants (considering compositions that do not use AP as an oxidizer). The search also showed 30 publications that investigate the use of HTPB as a binder in hybrid propellants and these will be discussed later. Figure 1 presents the number of documents from 2002 to 2012 as well as their type (article, conference paper or review). Figure 1 shows a continuous growth rate of publications,

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012

271

20 10

1 3

0 Review Conference paper Article

1 3 17

1 4

3 21

1 5

Number of documents

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 1 1 1 1 3

20 15 10 5 0

2002

2003

University

2004

2005

2006

2007

Research Institute

2008

2009

2010

2011

2012

University and Research Institute

Figure 1. Number of documents, specifying each kind, that describe the use of HTPB as binder over the years from 2002 to 2012.

Figure 2. Number of documents of each affiliation and the kind of research institution from 2002 to 2012.

which is observed until 2010, followed by an expressive reduction in scientific publication between 2011 and 2012. Considering the type of documents, 279 were articles, 63 were conference papers and only 4 were reviews. As one can see, once the contribution toward the particular research area was provided by articles, the rate of decrease of the publications observed in the years 2011 and 2012 (Fig. 1) was directly related to the specific kind of document. One point that should be observed is that almost 44% of all scientific production focuses on the last four years of the period studied, which can suggest an increase in interest in that scientific research branch. Considering the affiliation of the research (Fig. 2), results showed that research institutes, which comprise private research institutes and government agencies, were responsible for about 50% of the publications made (142, in total period). The total number of publications made only by the universities (105, in total period) and the total number of publications that were the result of a partnership between the university and research institutes (99, in total period) are very similar, from which it is possible to observe that both units, responsible for that area of research in a country, although they were able to develop scientific research in this area alone, were open to working together. Considering the assignee countries, were selected from the results obtained the 5 countries that presented the most publications. Figure 3 shows that China occupies the first position, which corresponds to 172 publications in the last 10 years, followed by the United States which has 56, with less than half of the p ublications of China. India, Japan and Brazil

have, respectively, 34, 25, and 10 publications in this period. These results indicate the Chinese government’s effort toward the development of this technology that is most widely used in the area of propellants in agreement with the Chinese aerospace program. The United States occupies the second position, suggesting that investments are still done in this kind of propellant in order to improve properties, such as burning rate and flame structure. Brazil occupies a good position (fifth), showing the return on investments made in training developed in research institutes and universities. It might be observed, from Fig. 3, that there was an investment increase in propellant research in China, from 2002 to 2012. Considering the abstracts, publications were selected by the subject relevance and were grouped while considering the compositions and the properties investigated. In Figs. 4 and 5,

Number of documents

30 25 20 15 10 5 0

2002

2003 2004 2005 2006 China United States

2007 2008 India

2009 2010 2011 2012 Japan Brazil

Figure 3. Number of documents of each assignee country from 2002 to 2012.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

Rufino, S.C., Silva, G. and Iha, K.

“metal” represents aluminum and boron, and “energetic material” represents the explosives RDX, HMX, HNF, CL20, ONC, ADN, DNTF, TNAZ, and DNOAF. Figures 4 and 5, respectively, show the results considering the propellant compositions and the properties, investigated, during the whole period (2002–2012). In several publications (104, in total period), results showed that the other components used with HTPB in propellant composition were not specified in abstracts, and it was only described that this composition had these components. However, it is apparent that there has been a large investment in research on the properties of compositions with ammonium perchlorate, as oxidant, and HTPB, as binder (112, in total period). The combination of metal (Al or B) and AP to enhance properties in propellant HTPB, 31% HTPB blend, 3%

HTPB/energetic material, 5% HTPB/metal, 5% HTPB/AP/Al/energetic material, 4%

HTPB/AP, 34%

HTPB/AP/metal, 18%

HTPB: hydroxyl-terminated polybutadiene; AP: ammonium perchlorate; Al: aluminum

Figure 4. Percentage distribution of documents of the most relevant propellant compositions from 2002 to 2012.

Aging, 9% Plasticizer, 4%

Combustion properties, 9%

compositions represented approximately half of the studies done with AP/HTPB (60, in total period). With regarding to the properties investigated, results showed that the research was focused on the search for compositions with an improved burning rate (136, in total period) and better mechanical properties (100, in total period). Studies on the effect of the catalyst used in composition (64, in total period) and on the use of nano metals in the compositions (45, in total period) were also important. Studies on the effect of aging on composition properties were also representative (51, in total period), followed by the combustion prop erties (52, in total period) and also the thermal properties (60, in total period). Figures 6 and 7, respectively, consider the behavior of the focus of research over the years 2002–2012. Results showed that, in 2006, the studies on compositions HTPB/AP/Metal had an expressive increase, but the attention on these compositions decreased in 2008, when there was an unexpected increase in studies on compositions of HTPB/AP without metal. Even during 2009–2012, the research on HTPB/AP remained preferable in relation to HTPB/AP/metal, accompanied by the interest in the mechanical properties of these compositions, which were directly related to the energy and the performance of the propellant, as seen in Fig. 7. GREEN PROPELLANT COMPOSITIONS Based on results, 30 publications could be called “green propellant compositions”. Here, those compositions that do

Flame structure, 4%

Number of documents

272

Boding agent, 3%

Mechanical properties, 18%

Burning rate, 24%

Nano metal, 8% Catalyst, 11%

Thermal properties, 10%

Figure 5. Percentage distribution of documents of the most relevant properties investigated from 2002 to 2012.

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

HTPB

HTPB/AP

HTPB/AP/METAL

HTPB: hydroxyl-terminated polybutadiene; AP: ammonium perchlorate

Figure 6. Distribution of documents of the most relevant propellant compositions from 2002 to 2012.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012

273

18 16

Number of documents

14 12 10 8 6 4 2 0

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

Combustion properties

Burning rate

Thermal properties

Catalyst

Nano metal Mechanical properties

Aging

Figure 7. Distribution of documents of the most relevant properties investigated from 2002 to 2012.

not use AP as an oxidizer will be considered green propellant compositions; in other words, compositions employing nitrate ammonium as an oxidizer, Al or B as a metal, and explosives RDX, HMX, HNF, CL20, ONC, ADN, DNTF, TNAZ, GAP, and DNOAF as “energetic material”. Figure 8 shows the number of documents related to green propellant compositions from 2002 to 2012. The small number of green publications in relation to the total number of publications (346) indicates that, despite there being some worries in searching for nonpolluting technologies, this effect is still not significant in propellants, which can also indicate the level of difficulty in finding a suitable replacement for HTPB, in view of its characteristics (stability and/or compatibility with the other components in the propellant’s composition). However, from 2007, an increase was observed in research related to this matter, indicating a greater environmental awareness. Thus,

the total number of green publications, from 2007 to 2012, reached a level of approximately 9% of the total publications over the same period. Another point to be observed is that most of the publications in this area were proceeding from research institutes, with 15 publications. The universities showed a modest number of publications, with only 5 works reported. However, partnerships between research institutes and universities were stimulated, showing ten publications in the entire period. Most of the publications were articles (24, in total period) followed by conference papers (5, in total period) and only one was a review. Again, China was the most expressive assignee country with 14 publications in the entire period, which represents more than four times than each one of the other countries. China was followed by the United States, India, and Japan, each of which had three publications.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

Rufino, S.C., Silva, G. and Iha, K.

274

On the other hand, the fact that Brazil does not provide publications in the area of green propellants can indicate a simple lack of interest in this subject or a technological backwardness in relation to other countries, which would already be working on improving composition propellants in replacement of toxic and/or harmful to environment. Figure 9 shows the number of documents related to the most prominent assignee countries from 2002 to 2012. Results showed that the only country which had maintained its research on regular basis was China. Recently (2010 and 2011, respectively), India and Japan also reported their research, but the United States, since 2007, has not reported publications involving this kind of propellants. Figure 10 shows that studies conducted on HTPB/ energetic material compositions were the most expressive in the entire period, as they involve many different kinds

of explosives as oxidizers. When analyzed in isolation, each composition represents less than three publications in the entire period. For this reason, what can be seen from this figure is that HTPB/AN was the most important composition with eight publications. Similar behavior can be observed from 2008 until 2011, where there was an increase in research involving energetic materials as oxidizers. The combination of metal and HTPB/AN did not have an expressive representation, as it had only two publications, one with Al as metal and another with B, and, indeed, since 2006, there has been no publication in this area. The major property studied in publications was the burning rate, as observed in traditional compositions (Fig. 11), corresponding to 31% of all publications.

Number of documents

5 4 3 2 1 0

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Figure 8. Number of documents about green propellants from 2002 to 2012.

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 HTPB/AN

HTPB/AN/metal

HTPB/energetic material

Figure 10. Number of documents of most relevant propellant compositions from 2002 to 2012.

Number of documents

Mechanical properties, 11%

3 Catalyst, 14%

Combustion properties, 19%

Flame structure, 6%

Thermal properties, 19%

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 China

United States

India

Burning rate, 31%

Japan

Figure 9. Number of documents of each assignee country from 2002 to 2012.

Figure 11. Percentage distribution of documents of most relevant properties, for green propellants, investigated from 2002 to 2012.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012

Figure 15 shows the percentage distribution of documents with the most relevant properties for hybrid propellants over the entire period. One can see that the property which was mostly investigated was the regression rate of the propellant, which was directly related to the performance of the propellant that occupies the second priority position, as with the improve energy of formulations may be obtained, as observed before. Results revealed that the most relevant properties studied from 2009 to 2012 were the regression rates of compositions, followed by combustion properties and simulation studies. From 2003 to 2008, studies involving the performance of the system gained prominence followed by investigations about the regression rate of the systems. Results showed that when the composition interest was changed from the use of N2O to the use of H2O2 (Fig. 14), the focus of the studied properties changed to 8

Number of documents

7 6 5 4 3 2 1 0

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Figure 12. Number of documents about hybrid propellant compositions from 2002 to 2012.

Number of documents

HYBRID PROPELLANT COMPOSITIONS Based on the search, 30 publications that investigate the use of HTPB as a binder system in a hybrid propellant were recovered. Figure 12 shows an increase in this research area with a great difference in 2008. In 2009, the number of publications started falling and then, resumed increasing in 2012. Results showed that the number of publications, as articles, were 17, in the entire period, and they were very similar to those such as conference papers, which were 13 publications. This might suggest the interest in promoting the dissemination of this technology in congress and technological events. It can also be concluded that major research was conducted by the universities (15 documents, in the total period), which may corroborate the high number of publications in the congress. The number of publications from research institutes was six, which was very similar to those published in association with the university (9, in the total period). These results may indicate the good relations between research units. The major assignee countries were China (11, in the total period), the United States (10, in the total period), and Italy (3, in the total period), and this may indicate that China was searching for research in traditional areas, such as solid propellants. Indeed, China had also been making investments in new or alternative technologies in propulsion areas. As can be seen in Fig. 13, the United States had been studying these propellant compositions since 2002 and China started reporting its scientific results in this area only in 2005. In 2009, the United States decreased its investment and until 2010, it did not have publications in this area. In 2012, only China showed interest in studying this kind of compositions. Figure 14 presents the percentage distribution of the number of publications that investigated the major hybrid propellant compositions, which were HTPB/N2O, HTPB/ H2O2, HTPB/LOx, HTPB/GOx, and HTPB/GOx/AP/Fe2O3. It shows that HTPB/N2O, HTPB/H2O2, and HTPB/LOx were the most frequently used compositions reported in scientific publications, and also shows the technology evolution from 2002 to 2012. One also can observe the recent interest in HTPB/H2O2 propellants and that, in 2008, the system HTPB/N2O was intensely investigated but a fter 2009, there was no publication involving this system. M oreover, Fig. 14 shows that HTPB/LOx systems were investigated from 2003 to 2006 and then, they only returned to be reported in 2008.

275

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 China

United States

Italy

Figure 13. Number of documents of each assignee country from 2002 to 2012.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

276

Rufino, S.C., Silva, G. and Iha, K.

Number of documents

4 3

HTPB/N2O, 40% HTPB/GOx/ AP/FE2O3, 7% HTPB/GOx, 7%

HTPB/H2O2, 23% HTPB/LOx, 23%

2 1

2002

2003

HTPB/N2O

2004

2005

HTPB/H2O2

2006

2007

HTPB/LOx

2008

2009

HTPB/GOx

2010

2011

2012

HTPB/GOx/AP/FE2O3

HTPB: hydroxyl-terminated polybutadiene; AP: ammonium perchlorate; LOx: liquid oxygen; GOx: gas oxygen

Figure 14. Number of most relevant documents of propellant compositions and their percentage distribution from 2002 to 2012.

Simulation, 12%

Regression rate, 38%

PATENTS The database of the USPTO, as well as other patent offices in the world (Sipo – China, Paj – Japan, INPI – Brazil, etc.), has unrestricted public access. This research has focused only

Combustion properties, 12%

Design, 9%

on the USPTO’s database because of the American interest in changing its aerospace program, which until then was based on the use of shuttle. The patent documents were Performance, 29%

recovered using the USPTO Patent Full-Text and Image Database (PatFT). Only the documents in which the claims

Figure 15. Percentage distribution of documents of the most relevant properties, for hybrid propellants, investigated from 2002 to 2012.

were present were considered: both words “propellant or propellants” and “polybutadiene or HTPB”. In the searches, the date when the application was submitted to this patent office (application date) was limited from J anuary 1, 2002 to December 31, 2012.

regression rate, which can be related to an improvement in the energy of the propellant achieved, instead of performance, which was the focus property studied in formulations using N2O.

The recovered documents were filtered and considered only the ones related to studies on SRM and HRM, which used HTPB as a binder and investigated the properties of the propellant.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

An Overview of the Technological Progress in Propellants Using Hydroxyl-Terminated Polybutadiene as Binder During 2002–2012

Seven patent documents were recovered, which are compiled in Table 1. Almost all of them have the same International Patent Classification (IPC), C06B, that refers to “explosive or thermic compositions, manufactured thereof, use of single substances as explosives”. These results could corroborate the search strategy efficiency. In relation to the assignees, the results showed equilibrium between American universities and companies; the United States navy also has one patent open to the public. The date when the patents were granted (granted dates) can be divided into two groups: 2005–2006 and 2010–2012. Recent patent topics show the market interest in solid propellants even nowadays, adjusting formulations in order to get and also improve properties, such as self-extinguishing and burning rate.

FINAL CONSIDERATIONS Search results in the scientific publications databases showed that composite propellants which use HTPB as

277

a binder are still one of the most used and investigated propulsion technologies. Private research institutes and government agencies are in charge of about 50% of the publications made in this area. China was the major country in relation to investments in research, followed by United States. Brazil also occupied a good position (fifth) showing the return on investment in training conducted by research institutes and universities. Results also showed that studies on properties of compositions involving the use of AP as an oxidizer are still preferable, and major research is focused on investigating compositions which could have improved ballistic and mechanical properties. The use of catalysts and nano metals is also important in analyzing composition properties. In publications involving green propellants, the use of HTPB/AN represents the most important composition from 2008 until 2011, when an increase was observed in research involving energetic materials as an oxidizer in place of AN. China is also the main country in relation to the number of publications in the hybrid propellants field, which may

Table 1. Patents documents describing the use of HTPB as binder in propellant compositions. Patent number

Title

Assignee

IPC (2006.01)

Granted date

US8336287

Solid propellant rocket motor having self-extinguishing propellant grain and systems therefrom

University of Central Florida Research Foundation, Inc.

C06D5/00

December 5, 2012

US8114229

Self-extinguishable solid propellant

University of Central Florida Research Foundation, Inc.

C06B29/00

February 14, 2012

US8066834

Burn rate sensitization of solid propellants using a nano-titania additive

University of Central Florida Research Foundation, Inc.

C06B29/22

November 9, 2011

US7824511

Method of making GAP propellants by pre-reacting a metal fuel with isocyanate before mixing with binder and plasticizer

US NAVY

C06B45/10

November 2, 2010

US7011722

Propellant formulation

Alliant Techsystems Inc.

C06B29/22

March 14, 2006

US6913661

Ammonium nitrate propellants and methods for preparing the same

Universal Propulsion Company, Inc

C06B31/28

July 5, 2005

US6896751

Energetics binder of fluoroelastomer or other latex

Universal Propulsion Company, Inc

C06B21/00

May 24, 2005

Fonte: USPTO Patent Full-Text and Image Database (PatFT).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

278

Rufino, S.C., Silva, G. and Iha, K.

indicate that this country is not only looking for research in traditional areas, such as solid propellants, but also searching for new or alternative technologies in the propulsion area. Based on the number of publications in this period, it is possible to infer that the area of research in propellants still appears young, and the research conducted in it, continues being promising.

The very low number of granted patent documents reported in open patent databases (USPTO), which uses HTPB in propellants as a binder could be associated with the legal aspects that are specific for each country, where, in some cases, this type of document may be considered of interest for national defense, not being published, and for this reason, preventing public consultation.

REFERENCES Andrade, J., Iha, K., Rocco, J.A.F.F., Franco, G.P., Suzuki, N. and Suárez-Iha, M.E.V., 2007, “Determinação dos parâmetros c inéticos de d ecomposição térmica para propelentes BS e BD”; Eclética química, Vol. 32, pp.45-50. Beckstead, M.W., Puduppakkam, K., Thakre, P. and Yang, V. 2007, “Modeling of combustion and ignition of solid-propellant ingredients”, Progress in Energy and Combustion Science, Vol. 33, pp. 497-551 Bohn, M.A. and Cerri, S., 2010, “Ageing behaviour of composite rocket propellant formulations investigated by DMA, SGA and GPC”, NDIA 2010 Insensitive Munitions & Energetic Materials T echnology S ymposium (IMEMTS) on “International Progress in Insensitive Munitions and Energetic Materials”, NDIA Event 1550, Session 9A, Munchen, Germany. Cerri, S., Bohn, M.A. and Menke, K., 2009, “Ageing Behaviour of HTPB Based Rocket Propellant Formulations”, Central European Journal of Energetic Materials, Vol. 6, pp. 149-165. Badgujar, D.M., Talawar, M.B., Asthana, S.N. and Mahulikar, P.P., 2008, “Advances in science and technology of modern energetic m aterials: An overview”, Journal of Hazardous Materials, Vol. 151, pp. 289-305. DeLuca, L.T., Galfetti, L., Maggi, F., Colombo, G., Merotto, L, Boiocchi, M., Paravan, C., Reina, A., Tadini, P. and Fanton, L., 2013, “ Characterization of HTPB-based solid fuel formulations: Performance, mechanical properties, and pollution”, Acta Astronautica, Vol. 92, No. 2, pp. 150–162. Gascoin, N., Gillard, P., Mangeot, A. and Navarro-Rodriguez, A., 2012, “Literature survey for a first choice of a fuel-oxidiser couple for hybrid propulsion based on kinetic justifications”, Journal of Analytical and Applied Pyrolysis, nº94, pp. 1-9. Gomes, S.R., 2012, “Projeto e desenvolvimento de um motor foguete hibrido”, Ph.D. Thesis, Instituto Tecnológico de Aeronáutica, Brazil. Guobiao, C., Hao, Z., Dalin, R. and Hui, T., 2011, “Optimal design of hybrid rocket motor powered vehicle for suborbital flight”, Aerospace Sience and Technology, Vol. 25, No. 1, pp. 114-124.

La Fuente, J.L., 2009, “An analysis of the thermal aging behaviour in high-performance energetic composites through the glass transition temperature”, Polymer Degradation and Stability, Vol. 94, pp. 664-669. LPI nº 9279, de 14/05/1996, Lei da Propriedade Industrial. Ministério da Defesa, Portaria Normativa nº 1888/MD, de 23 de dezembro de 2010. Miskelly, H.L. Jr., 2004, “Destroying airborne biological and/or chemical agents with solid propellants”, US 6748868 B1. Muller, G., “Roadmapping”, Embedded Systems Institute. Disponível em: http://www.gaudisite.nl. Acesso em: nov. 2012. Naya, T. and Kohga, M., 2013, “Influences of particle size and content of HMX on burning characteristics of HMX-based propellant”, Aerospace Science and Technology, Vol. 27, No. 1, pp. 209-215. Ortiz, L.C., Ortiz, W.A. and Silva, S.L., 2002, “Ferramentas alternativas para monitoramento e mapeamento automatizado do conhecimento”, Ciência da Informação, Vol. 31, pp. 66-76. Paterlini, W.C., Botelho, E.C., Rezende, L.C., Lourenço, V.L. and Rezende, M.C., 2002, “Efeito da concentração do catalisador acetilacetonato ferco na cura de poliuretano a base de polibutadieno liquido hidroxilado (BLH) e diisocianato de isoforona (IPDI)”, Quim. Nova, Vol. 25, pp. 221-225. Petersen, E.L., Seal, S., Stephens, M., Reid, D.L., Carro,R., Sammet,T. and Lepage, A., 2012, “Self-extinguishable solid propellant”, US 8114229 B1. Subhananda Rao, A., Krishna, Y. and Nageswara Rao, B., 2004, “Comparison of fracture models to assess the notched strength of composite/solid propellant tensile specimens”, Materials Science and Engineering, Vol. A 385, pp. 429-439.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.267-278, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.228

Experimental Study of Polyurethane-Based Fuels with Addition of Paraffin and Aluminum for Hybrid Rocket Motors Susane Ribeiro Gomes1, Leopoldo Rocco Junior1, José Atílio Fritz Fidel Rocco1, Koshun Iha1

ABSTRACT: Experimental investigation was conducted to determine the relative propulsive and combustion behavior of several polyurethane-based solid-fuel formulations containing 30% w/w of paraffin or 10% of aluminum powder. In total, seven solid-fuel formulations were investigated, four containing 30% of paraffin and three with 10% of aluminum. The polyurethane was synthesized with pre-polymer technology. The oxidizer was gaseous oxygen, which was forced into the combustion chamber with axial and swirl methods. Firing tests with 7 configurations were performed. Thrust measurements indicated that the addition of paraffin increased thrust at about 57% and regression rates at about 70%. No relevant improvement in performance was obtained with aluminum addition. Specific impulse decreased when aluminum particles were added to the fuel. The mixture that produced the best ballistic parameters was polyurethane plasticized with castor oil and 30% w/w of paraffin with gaseous oxygen injected through a swirler. KEYWORDS: Hybrid rocket motor, Aerospace propulsion, Rocket motors.

INTRODUCTION Hybrid rocket engine (HRE) is a type of rocket motor that combines the advantages of both solid and liquid fueled rockets and avoids many of the underperformances. In a hybrid, the oxidizer is stored as liquid or gas in tanks and the fuel is stored as solid-fuel grain in the combustion chamber. Oxidizer is injected over the burning fuel surface, and the resulting gases are expelled out of a nozzle to produce thrust. Because the fuel and oxidizer are separate, and cannot easily mix (because they are at different phases), hybrid rockets have very little danger of exploding (Boardman, 2001). Hybrids may provide higher Isp than solid motors, and due to the high-density solid, they may present higher densityspecific impulse than liquid engines. Advantages such as fuel versatility — additives can be easily embedded in the fuel grain — potential environmental friendliness, and low cost due to high levels of safety and minimal failure modes are widely acknowledged. Classical hybrid rocket motors also present disadvantages yet to overcome, such as combustion inefficiencies due to poor mixing and mixture ratio shifting and mainly low regression of the fuel grain (Altman, 2001). Several methods for improving regression rates have been proposed and investigated. Approaches to increase regression rates usually employ advanced fuels into three categories: • Addition of energetic particles: A methodology to increase the regression rate in HREs has been to i ntroduce aluminum into the solid fuel (Chiaverini, 2007). At Penn

1.Instituto Tecnológico de Aeronáutica – São José dos Campos/SP – Brazil Author for correspondence: Susane Ribeiro Gomes | Instituto Tecnológico de Aeronáutica | Praça Eduardo Gomes, 50 | CEP 12.228-901 São José dos Campos/SP – Brazil | Email: susaneribeiro@gmail.com Received: 21/01/13 | Accepted: 19/06/13

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

280

Gomes, S.R., Rocco Junior, L., Rocco, J.A.F.F. and Iha, K.

State University, nanosized particles of aluminum powder with diameters of 70–150 μm were mixed into hydroxyl-terminated polybutadiene (HTPB) fuel grains and tested (Chiaverini et al., 2000; Evans et al., 2009). The results showed that the regression rates increased by 50% over the HTPB without aluminum. Researchers at University of Pennsylvania (UPENN) reported improvements of 60% on regression rate with the use of embedded nanoparticles (Risha et al., 2007). • Use of energetic polymers instead of conventional fuels such as HTPB; this polymer is widely used due to its strong mechanical and durability properties and the technology is well known due to the use in solid rocket motors (Altman and Humble, 1995; Altman and Holzman, 2007; Geisler et al. 2010). • Use of paraffin-based fuels due to the inherent masstransfer mechanism. Paraffin Fuels Researchers at Stanford University (Karabeyoglu et al. 2001) discovered that paraffin-based fuels exist which have regression rates that are three to four times than those of conventional hybrid fuels. This is largely due to the development of a thin liquid layer on the fuel grain surface which becomes instable: instability appears due to the incoming oxidizer flow pattern and liquid fuel droplets are injected into the boundary layer (Karabeyoglu et al. 2001; Karabeyoglu and Cantwell, 2002). This enhanced mass-transfer mechanism increases fuel mass flow without the blocking effect typical of gaseous fuel blowing. Paraffin presents poor mechanical properties; a binder is hence desirable to sustain loads on flight condition. Therefore, the purpose of this work was the development of a polyurethane (PU) binder, or polymeric matrix, with total substitution of HTPB pre-polymer for another polyol, or mix of polyols. This product is supposed to be applicable as a fuel binder or solid propellant binder formulations. The synthesized binder is embedded with paraffin and micron-sized aluminum particles. Synthesized fuels are tested to assess the improvement in combustion efficiency and regression rates. In addition to developing a potential high regression rate fuel, a second methodology was employed to improve combustion efficiency. Swirling type injectors are known to

increase heat transfer from the flame to the grain surface. This method has first-order effects on regression rates (Knuth et al., 1998; Carmicino and Sorge, 2005, 2007). The swirl injector was used in the second series of tests to evaluate the relative improvement in combustion efficiency and regression rates in comparison to the axial injector. The use of paraffin and swirl injectors yielded improvements in thrust up to 57%, and in regression rates up to 70% in relation to the standard PU fuel and axial injector.

METHODOLOGY This section outlines the steps taken in this work: synthesis of PU pre-polymer, addition of paraffin addition of aluminum particles, and firing tests. In the binder synthesis, HTPB-hydroxyl-terminated polybutadiene was replaced with a modified polytetramethylene ether glycol (PTMEG) pre-polymer with 4.3% of free NCO. The mixture was cured with a liquid amine (ETHACURE) and plasticized with castor or mineral oil. Finally, PU elastomers were obtained. A microcrystalline paraffin from petroleum (Petrobras 140/145-1) composed of saturated aliphatic hydrocarbons with melting point of 61.4°C and boiling point of 290°C, at standard conditions, was added to the synthesized PU. In this work, aluminum particles with diameters of 100 μm were added to three chosen fuel combinations in the proportion of 10% w/w. Firing tests with axial and swirl injectors were done and the ballistic results were obtained and evaluated. A side view of swirl injector is shown in Fig. 1, and a scheme of the flow entering the combustion chamber is presented in Fig. 2. EXPERIMENTAL SETUP The baseline engine design was developed from a need for simplicity and flexibility; therefore, a modular design was incorporated. The set could be assembled and disassembled in minutes, allowing the practice of several tests per session. The case and the flanges were made from stainless steel. The case was machined to fit between the steel flanges. The nozzle was adapted in the aft flange, to prevent nozzle escape. A hydraulic system was settled to perform the

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

Experimental Study of Polyurethane-Based Fuels with Addition of Paraffin and Aluminum for Hybrid Rocket Motors

281

Figure 1. Side view of the swirl injector. The holes are the oxidizer hose entrances.

(a)

(b) Figure 3. Side view of the labscale hybrid rocket motor on the test bench (a) and during test (b). Figure 2. Schematic diagram of the swirl injector cross-section.

thrust measurements. A view of the motor attached to the test bench is shown in Fig. 3. Compression fitting valves were used in the oxygen feed line. A squib was used to achieve ignition. A schematic diagram of the test facility is exposed in Fig. 4. Compression fitting valves were used in the oxygen feed line. A pyrotechnic method was used to achieve ignition. The combustion chamber is constructed of 316 L stainless steel, total length of 215 mm, and inner diameter of 68.3 mm. The fuel grain is 185 mm long, inner diameter of 20 mm. Pre-chamber is 15 mm long and post-chamber is 10 mm long. No thermal isolation was used in this project; it is intended to use a high temperature coating in future work. The experiments were performed in ambient conditions, with temperature of approximately 27°C and 1.024 MPA and humidity values varying between 40 and 80%.

PC Control

DAQ

Spark Plug

Needle vlv Solenoid vlv

Balance

Dataline Control line Oxigen flowline

Load Cell

Figure 4. Diagram of the experimental apparatus.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

282

Gomes, S.R., Rocco Junior, L., Rocco, J.A.F.F. and Iha, K.

Nozzle is made of graphite, throat diameter of 8 mm, convergence angle of 30° and divergence angle of 15°. The length was calculated accounting for the desired exit pressure as close as possible to the ambient pressure.

Results and Discussion Average regression rate is defined in Eq. 1 and oxidizer mass flux is experimentally defined in Eq. 2: _

. ravg = a Goxn

Gox =

(1)

. mox ⎛ Di + D f ⎞ ⎟ 2 ⎠

π ⎜⎝

(2)

Using experimental data, average regression rate is calculated as follows, Eqs. 3 and 4: Δm = π D22 − D12 ρ L = π ( D2 + D1 )( D2 − D1 ) ρ L 4 4

(

)

(3)

. D − D1 4 Δm r= 2 = (4) Δt π ( D2 + D1 ) ρ LΔt

The specific impulse is determined by dividing the total impulse by the weight of the propellant (fuel and oxidizer) consumed. The total impulse is calculated by numerically integrating the thrust over time for the duration of the test using a simple Riemann squares approximation implemented on matlab and then divided by the weight of the propellant used as shown in Eq. 5:

I sp =

∑

i = tb i =0

1 (T + T ) . (ti+∆t − ti ) (5) 2 i+∆t i g . mprop

where Isp is the specific impulse, Ti is the current thrust value, Ti + ∆t is the thrust at the next time step, ti is the current time, Ti + ∆t is the time at the next time step, g is the acceleration due to gravity, and mprop is the mass of the propellant consumed during the burn. The first series of synthesis provided three fuel formulations, the first was plasticized with mineral oil, the second with castor oil, and the third was plasticized with castor oil and embedded with 30% w/w of paraffin. To establish a baseline, commercial PU was acquired and compared. A series of tests with each one of the three synthesized fuel and the commercial PU were performed. The method of injection was axial to the grain, and the gaseous oxygen mass flow rate was kept constant at 40 g/s. The thrust results are presented in Fig. 5.

100

Thrust (N)

Axial, Comercial PU

Axial, PU with mineral oil

40 Axial, PU with castor oil

20 0

Time (s)

Axial, PU + 30% paraffin

Figure 5. Thrust data of polymeric fuels and paraffin wax mixture with axial injector and 20-second tests.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

Experimental Study of Polyurethane-Based Fuels with Addition of Paraffin and Aluminum for Hybrid Rocket Motors

The graph shows that the PU synthesized with mineral oil yielded the lowest thrust, followed by commercial and PU synthesized with castor oil, both with the approximate thrust of 62 N. Results obtained with the use of paraffin were the most promising reaching almost 100 N. Additional data are summarized in Table 1. Thrust presented the highest uncertainty of 4.9%; other variables showed smaller uncertainties. The chamber pressure was calculated as the mean value of the plateau pattern obtained in the pressure curves. It was observed that the binder that contained castor oil had better performance and higher stability. It increased Isp in about 15.6% and thrust in about 19.4%. The addition of 30% paraffin increased Isp in about 29.5% and thrust in about 56.9% and the chamber pressure achieved with the same oxidizer mass flow rate is 58.8% higher. Regression rates in relation to the PU synthesized with castor oil increased almost 70%. For the same oxidizer mass flow rate, pressure achieved was at least 70% higher when 30% w/w of paraffin was added. The next step was increasing the percentage of paraffin. Unfortunately, there were temperature issues on the surface of the axial injector and on the case. The swirl injector was used, and the test duration was reduced to 6 seconds. The second series of tests was performed with PU also synthesized with castor oil in the following proportions: • PU with 30% of paraffin • PU with 30% of paraffin and 10% of aluminum (w/w) • PU with 10% of aluminum

283

The results were the most promising so far; for more information, see Table 2. The highest regression rate was achieved in the configuration of 30% paraffin and 10% aluminum; however the highest thrust was obtained when no aluminum was added (PU + 30% paraffin). Specific impulse was the lowest when aluminum powder was present and the highest when no aluminum was added to the fuel. Previous analysis has shown that addition of aluminum to traditional hybrid solid fuels actually decreases the specific impulse because the molecular weight of the exhaust products is increased, which counteracts the raise in the temperature. Figure 6 shows the steadiness of thrust achieved with PU with 30% of paraffin tested with the swirl injector. Average thrust values were about 23% higher than both grains with aluminum particles. PU with 30% of paraffin ignites faster and presents the best performance. The other two cases show roughly the same constant thrust at about 88.9 N. Figure 7 shows a picture of the grain with 30% of paraffin after the firing test. Paraffin melted and later suffered condensation; one could deduce that a motor with paraffin or any other liquefying fuel will work differently if restarted. Another effect observed in this series of tests was the contraction of the fuel grain after firings, which is a dangerous effect, once combustion can occur at the outer region of the grain and jeopardize the mission.

Table 1. Experimental results for each fuel using of the axial injector. Fuel

P (atm)

O/F

Thrust (N)

Isp (s)

r (mm/s)

Injector

PU + paraffin 30% PU w/o castor oil PU with castor oil Commercial PU

13.98 8.11 8.94 8.97

1.63 3.72 3.23 2.89

98.5 52.7 62.9 62.6

177.9 121.1 140.0 137.5

0.97 0.50 0.53 0.53

Axial Axial Axial Axial

Table 2. Experimental results for each test configuration. Fuel

P (atm)

O/F

Thrust (N)

Isp (s)

r (mm/s)

Injector

PU + paraffin 30% PU + 10% Al PU + 30% paraffin + 10% Al

17.65 13.86 13.82

2.63 3.23 3.46

108.8 88.9 88.4

140.0 88.9 127.7

2.63 3.29 2.63

Swirl Swirl Swirl

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

284

Gomes, S.R., Rocco Junior, L., Rocco, J.A.F.F. and Iha, K.

120

Thrust (N)

100

Swril, PU + 10%Al

80 60

Swril, PU + 10% Al +30% paraffin

PU + 30% paraffin

20 0

4 5 Time (s)

Figure 6. Thrust data of polymeric fuels and paraffin wax mixture with swirl injector and 6-second tests.

Addition of aluminum and grain analysis Figure 8 shows the grain inside the stainless steel case prior to testing. This grain contains 10% of micron-sized aluminum powder and PU synthesized with castor oil.

This particular grain was hard to ignite; two squibs were necessary to achieve ignition. It is supposed that the oxide layer on the surface is responsible for this problem, once the particles were not stored in an oxygen-free controlled environment. Figure 9 shows the deposition of metallic aluminum on the rear section of the motor, around the nozzle. Metallic powder was also deposited on the injector head, as can be seen in Fig. 10. It is evident that the aluminum deposition on several parts of the engine might cause problems; the most obvious is the accumulation of material close to the nozzle, once a possible throat obstruction is likely to occur. Even though the aluminum amount used was the same with or without the paraffin addition, the deposition when paraffin was present was less significant.

(a)

(b) Figure 7. PU + 30% of paraffin grain seen from the injector (a) and from the nozzle (b).

Figure 8. PU with 10% of micron-sized aluminum powder before testing.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

Experimental Study of Polyurethane-Based Fuels with Addition of Paraffin and Aluminum for Hybrid Rocket Motors

285

Conclusion

(a)

(b)

Figure 9. Deposition of aluminum powder around the nozzle.

The embedding of paraffin increased the overall ballistic parameters of the motor. The use of the binder is essential to structure and strengthen the paraffin to endure flight condition loads. Mixing ratios of 30% w/w of paraffin wax in the binder structure yielded good results, particularly when swirl injectors are used. Several additional tests must be performed to address the best mixing ratio. The aluminum particles probably presented an oxide layer too thick to participate in the combustion process. In future work, aluminum particles with smaller diameters and stored in controlled environments must be used.

Acknowledgments

Figure 10. Deposition of aluminum powder on the injector.

The authors gratefully acknowledge financial support from CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico ) and AEB (Agência Espacial Brasileira).

References Altman, D., 2001, Rocket motors, hybrid. In Meyers, R.A. (Ed.), “Encyclopedia of Physical Science and Technology” (3rd ed., Vol. 14, pp. 303-321), Elsevier Science Ltd. Altman, D. and Holzman, A., 2007, Overview and history of hybrid rocket propulsion, Progress in Astronautics and Aeronautics Series (1st ed., Vol. 218, p. 650). AIAA.

Chiaverini, M., 2007, Review of solid-fuel regression rate behavior in classical and nonclassical hybrid rocket motors, in M.J. Chiaverini, K. K. Kuo (Eds.), “Progress in Astronautics and Aeronautics” (1st ed., Vol. 218, pp. 37-126), American Institute of Aeronautics and Astronautics. Chiaverini, M., Serin, N., Johnson, D., Lu, Y.C., Kuo, K. and Risha, G.,

Altman, D. and Humble, R., 1995, “Hybrid Rocket Propulsion Systems”, Space Propulsion Analysis and Design”, in R. Humble, G. Henry, W. Larson (Eds.). McGraw-Hill, New York, pp. 365-441.

2000, “Regression rate behavior of hybrid rocket solid fuels”, Journal

Boardman, T.A., 2001, “Rocket Propulsion Elements”, In G.P. Sutton, John Wiley & Sons.

Investigations at Penn State University’s High Pressure Combustion

of Propulsion and Power, Vol. 16, No. 1, pp. 125-132. Evans, B., Boyer, E., Kuo, K. K. and Risha, G., 2009, “Hybrid Rocket Laboratory: Overview and Recent Results”, 45th AIAA/ASME/

Carmicino, C. and Russo Sorge, A., 2005, “Role of injection in hybrid rockets regression rate behavior”, Journal of Propulsion and Power, Vol. 21, No. 4, pp. 606-612.

SAE/ASEE Joint Propulsion Conference and Exhibit, AIAA Paper

Carmicino, C. and Russo Sorge, A., 2007, “Performance comparison between two different injector configurations in a hybrid rocket”, Aerospace Science and Technology, pp. 61-67.

overview and solid rocket motor fundamentals, in R. Blockley, W.

2009-5349, Denver, Colorado, 2009. Geisler, R.L., Frederick Jr, R.A. and Giarra, M., 2010, Historical Shyy (Eds.), “Encyclopedia of Aerospace Engineering” (Vol. 9), John Wiley & Sons, Ltd. All rights reserved.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

286

Gomes, S.R., Rocco Junior, L., Rocco, J.A.F.F. and Iha, K.

Karabeyoglu, M.A. and Cantwell, B.J., 2002, “Combustion of liquefying hybrid propellants: part 2, Stability of liquid ﬁlms”, Journal of Propulsion and Power, Vol. 18, No. 3, pp. 621-630. Karabeyoglu,

M.A.,

“Development

and

Fuels”,

Cantwell, Testing

Proceedings

B.J. of

the

and

Altman,

Paraffn-Based 37th

D.,

2001,

Hybrid

Rocket

AIAA/ASME/SAE/ASEE

Joint Propulsion Conference and Exhibit, July 2001, AIAA Paper 2001-4503.

Knuth, W.H., Chiaverini, M.J., Gramer, D.J. and Sauer, J.A., 1998, “Experimental Investigation of a Vortex-Driven HighRegression Rate Hybrid Rocket Engine”, 34th AIAA/ASME/SAE/ ASEE Joint Propulsion Conference & Exhibit. AIAA. Risha, G.A., Evans, B.J., Boyer, E. and Kuo, K.K., 2007, Metals, energetic additives, and special binders used in solid fuels for hybrid rockets, in M. J. Chiaverini and K. K. Kuo (Eds.), “Progress in Astronautics and Aeronautics” (1st ed., Vol. 218, pp. 413456), American Institute of Aeronautics and Astronautics, Inc.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.279-286, Jul.-Sep., 2013

doi: 10.5028/jatm.0501. doi:10.5028/jatm.v5i3.240

Combustion Modeling of Aluminum Incorporated in Low-Explosive Formulations such as Solid Propellants Rene Francisco Boschi Gonçalves1, Koshun Iha1, José Atílio Fritz Fidel Rocco1

ABSTRACT: Aluminum that is incorporated in an e nergetic material such as a propellant plays a significant role in the combustion process by means of stabilization with regard to the burning and generation of additional energy. The use of simulation softwares to model the combustion mechanism and kinetic parameters of the elementary reactions that compose the oxidation were used as the pressure variation of the combustion chamber of a rocket motor conditions. The behavior of the molar fraction of the chemical species during the combustion and its posterior stabilization were observed. The systems submitted to higher pressures tend to stabilize more rapidly, according to the greater chemical speed of the elementary reactions. KEYWORDS: Aluminum, Combustion simulation, Chemkin.

INTRODUCTION FLAME STRUCTURE OF A COMPOSITE SOLID PROPELLANT GRAIN Most of the current available knowledge on chemical aspects involving combustion of composite solid propellants regards studies of flame structures like, for example, the spatial distribution of temperatures and concentrations of chemical species involved in the combustion. Data analysis of the flame structure can provide information on the composition (generated from reactions in the condensed phase) of the products that have originated from thermal decomposition on the burning surface of a solid propellant grain. Therefore, it is possible to identify the reactions that occur in the condensed phase, in addition to the reaction pathways involved. Furthermore, the chemical structure of the flame provides information on the mechanism and kinetics of the chemical reactions in the gas phase and subsequent processing of the products, with reactions considered responsible for the generation of heat from the gas phase. The development of combustion models for solid propellant formulations requires information on the reactions involved, in both the condensed and gas phases of the process. Without this information, it is impossible to develop a model that is able to predict actual parameters such as burning rate and other ballistic characteristics, which are intrinsic to the material considered Korobeinichev et al., 1974. In the last four decades, studies have been performed in order to meet the ever more comprehensive process of energetic material combustion in special formulations of solid

1.Instituto Tecnológico deRio Aeronáutica – São dos Campos/SP – Brazil Universidade Federal do de Janeiro – RioJosé de Janeiro/RJ – Brazil Author Boschi| Departamento Gonçalves | Departamento de Mecânica/COPPE/UFRJ/C.P Química, Instituto Tecnológico.de Aeronáutica | Praça Mal. Eduardo Gomes, 50 – Author for correspondence: correspondence: Rene Jules Francisco Ghislain Slama de Engenharia 68.503 | CEP 21.945-970 Rio de Janeiro/RJ – Vila das Brazil | Acácias | CEP 12.228-900 São José dos Campos/SP – Brazil | Email: renefbg@gmail.com email: julesslama@yahoo.com.br Received: 07/03/13 | Accepted: 12/06/13 Received: 02/02/12 | Accepted: 30/10/12

o 3, pp.287-292, Jul.-Sep., 2013 J. Aerosp. Technol. São Manag., Sãodos JoséCampos, dos Campos, Vol.5, J. Aerosp. Technol. Manag., José Vol.X, NoNX, pp.1-8, XXX.-XXX., 2013

288

Gonçalves, R.F.B., Iha, K. and Rocco, J.A.F.F.

ropellants (Barrere, 1968; Williams et al., 1969; Kuo et al., p 1984; Chiaverini et al., 1999; Gonçalves et al., 2009). However, the complete combustion of this process is unknown because, in practice, it involves many parameters such as steady-state combustion, erosive burning, combustion in transient state (ignition and combustion instabilities), and combustion of finely divided metal particles such as aluminum or boron (incorporated in the material) (Davenas, 2003). BURNING INSTABILITIES Currently, most of the research done in propulsion of solid propellant rocket engines (SPRE) is directed to the understanding of a phenomenon that is widely known as “burning instability.” This phenomenon directly affects the burning of such material due to the resonance characteristic of the combustion process. A great mass of information is obtained pertaining to flame structure, heat-release mechanisms, combustion response in acoustical terms (Shusser et al., 2000), and the development of the flow that is established from burning. The term “combustion response” refers to the response given by lowamplitude linear frequency caused by the combustion of solid propellants (harmonic oscillations in the combustion chamber pressure during combustion) (Brewster, 2000). However, despite these contributions, the existing burning models provide limited information about the fluid dynamics behavior in the regions where oscillations are observed in turbulent flow of gases inside the combustion chamber of the SPRE (Flandro et al., 2000). In the 1960s, when most research in this area was directed to formulations of composites based on ammonium perchlorate (AP), Summerfield et al., (1966) identified the mechanism by which this process occurred and proposed a burning model named granular diffusive flame. In this model, based on the assumption that both the fuel and the oxidant are converted from solid to gas on the burning surface and are diffused together forming the flame front, an understanding of the composite combustion process as a whole was initiated. However, parameters such as large variations in the burning velocity of solid propellants as a function of the pressure in the combustion chamber, distribution of the granulometric profile of the loads (AP, aluminum, and Al-additives), and the concentration of such fillers in the composite formulated required models broader than initially proposed by Summerfield et al., (1966).

In the early 1970s, Beckstead et al., (1970) presented a model called multiple flame, which was based on a complex interaction between the flame generated from the thermal decomposition of the oxidant and two other diffusive flames immediately above the interface between the “binder” (organic part of the propellant) and the oxidant (material responsible for supplying oxygen to the combustion process). This phenomenological interpretation of the combustion process of solid propellant was used for many years to explain the burning behavior of composites based on AP, and also in double-based ones (nitroglycerin/ nitrocellulose). Apparently, the diffusive flame generated from the thermal decomposition of AP is dominant in the mechanism present in Beckstead’s model and is related to the manner in which the AP decomposes, generating a large amount of chemical species derived from chlorine. These species (extremely reactive) allow a fast increase of the flame temperature and maintain the stability of the whole combustion process of the composite. To date, this phenomenological interpretation of Beckstead et al., (1970) remains valid. The use of computer packages for burning simulation of energetic materials, with an emphasis on aluminum oxidation, is the focus of this work.

METHODOLOGY The combustion modeling of solid propellant formulations can be performed through computer simulation using the software Chemkin (Aurora reactor, for the current work), developed by “Sandia National Laboratories, Livermore, CA” (Coltrin et al., 1991). This software is aimed at solving complex problems involving chemical kinetics using complex reaction pathways. The program’s architecture offers information on specific problems, as well as models of independent problems and some packages with preselected reaction models that take into account the equations of mass, energy, chemical species and, in some cases, momentum conservation. When solving a specific problem, there is a need to identify the chemical species involved in the process (combustion) and its properties, such as enthalpy (H), entropy (S) and heat capacity at constant pressure and volume (Cp, Cv) beyond the reaction pathways and reaction rates involved.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.287-292, Jul.-Sep., 2013

Combustion Modeling of Aluminum Incorporated in Low-Explosive Formulations such as Solid Propellants

COMBUSTION OF AN ALUMINUM PARTICLE OXIDIZED BY AIR Computer simulations provide a very convenient support relative to understand the physics of combustion of many chemical species, particularly aluminium, in processes which take into account the growth of the o xide layer that covers its surface. If the modeling is sufficiently complete, it is possible to obtain results (data) leading to the burning rates involved in aluminum combustion. Initial simulations of aluminum combustion (drops) were performed by Liang and Beckstead (1998). This two-dimensional model takes into account effects such as forced convection, radiation, growth of the oxide layer, transport properties variation and an extended condensation zone. The model considers a reaction mechanism involving nine steps, including two surface reactions, three gas-phase reactions, a dissociation reaction of the molten Al2O3, and three condensation reactions. In order to study the combustion of aluminum (drops) in an environment, for example, in the case of a rocket engine, these researchers (Liang and Beckstead, 1998) modified the kinetic model to consider the reaction of aluminum with H2O and CO2, as well as condensation reactions involving each of these oxidizers. Simultaneously, the thermodynamic properties were adjusted to follow changes in pressure. In this modeling, the process of aluminum combustion in vapor phase occurs in a homogeneous reactor where it is considered that throughout the control volume there is a perfectly homogenous mixture between fuel and oxidant chemical species (PSR – Perfect Stirred Reactor). The reaction mechanism for the oxidation of aluminum in the air consists of nine elementary reactions that are described next:

289

THERMAL DECOMPOSITION MODELING OF ALUMINUM In order to perform the thermal decomposition modeling of aluminum in a homogeneous reactor (transient state), the software Chemkin/Aurora model was utilized. In this type of reactor, the reacting species (in gaseous form) are homogeneously dispersed throughout the system volume, where the ignition subsequently occurs, resulting in an abrupt increase in temperature (due to heat release by the exothermic combustion reactions of alumina) and volume (due to high gas formation), in addition to changes in mole fractions of the chemical species considered for the modeling. The temperature used in the simulation (1700°C) resembles the observed temperature in a combustion chamber of a conventional rocket motor. Various values have been used to denote pressure, including 1, 10, 30, and 60 atmospheres, with the last value being the observed pressure in a combustion chamber of a conventional rocket motor. The mechanism used in modeling and Arrhenius data of each elementary reaction are depicted in Table 1. Table 1. Alumina combustion mechanism. Reactions

A (mol-cm-s-K)

Ea (kJ/mol)

Al+O2 = AlO + O

9.72E13

0.0

159.95

Al + O + M = AlO + M

3.0E17

-1.0

0.00

AlO + O2 = OAlO + O

4.62E14

0.0

19885.90

Al2O3 = AlOAlO + O

3.0E15

0.0

97649.99

Al2O3 = OAlO + AlO

3.0E15

0.0

126999.89

AlOAlO = AlO + AlO

1.0E15

0.0

117900.00

Al + O2 = AlO + O

(1)

AlOAlO = Al + OAlO

1.0E15

0.0

148900.00

AlO = Al + O

(2)

AlOAlO = AlOAl + O

1.0E15

0.0

104249.94

AlO + O2 = AlO2 + O

(3)

OAlO = AlO + O

1.0E15

0.0

88549.86

AlO2 = AlO + O

(4)

AlOAl = AlO + Al

1.0E15

0.0

133199.94

Al2O = Al + AlO

(5)

Al = Al

1.0E14

0.0

0.00

Al2O3 = Al2O3

1.0E14

0.0

0.00

Al2O2 = AlO + AlO

(6)

Al + H2O = H + AlOH

1.14E12

0.0

442.80

Al2O2 = AlO2 + Al

(7)

Al + H2O = AlO + H2

9.6E13

0.0

2868.60

Al2O2 = Al2O + O

(8)

AlOH = Al + OH

1.0E15

0.0

66431.80

O2 + M = O + O + M, where M is an inert molecule. (9)

AlOH = AlO + H

1.0E15

0.0

57725.20

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.287-292, Jul.-Sep., 2013

Gonçalves, R.F.B., Iha, K. and Rocco, J.A.F.F.

Figure 2 shows the simulation results depicting the pressure at 10 atmospheres. In the case where the pressure is 10 atmospheres, a higher reaction rate is initially noticed due to a sudden volume increase as a result of the system ignition. The mole fractions of intermediate species suffered additions, as in this case the higher pressure enables the occurrence of all intermediate reactions present in the mechanism. However, there is still not enough time to stabilize the system considered (2.10-4 s), residence time of chemical species in the combustion chamber in question, showing that the reaction does not occur fast enough to ensure burning stability and greater efficiency. Figure 3 presents the combustion process at 30 atmospheres. The speed of the process, in terms of the (volume versus time) curve slope, increased even more in this case, as there is a fast increase and stabilization of the volume, shortly after the ignition of the combustion process.

RESULTS AND DISCUSSION The combustion simulations of alumina were performed at different pressures (1, 10, 30, and 60 atmospheres). In Fig. 1 it is possible to observe the behavior of the molar fractions of the chemical species involved in the process, according to the elapsed time of combustion from ignition, and also the system volume variation as a function of time. The predominant chemical species in the simulation at 1 atmosphere are aluminum, gaseous oxygen, aluminum oxide, and elementary oxygen. Due to the low pressure at which the system is submitted, there is little formation of alumina and its intermediates. In this case, therefore, the predominant reaction is the capture of one of the oxygen atoms of O2 by the aluminum atom, with the formation of aluminum oxide (I) and elemental oxygen.

1.35

1.10

0.14

1.06 1.04

Volume (cm3)

0.12 Molar Fraction

Volume (cm3)

1.08

1.02 1.00 0.00E+00

0.10 0.08

1.25 1.20

0.06

1.15

0.04

1.10

0.02

1.05

5.00E -05

1.00E -04

1.50E -04

2.00E -04

1.00 -5.00E -05 5.00E -19 5.00E -05 1.00E -041.00E -04 1.50E -041.50E -04 2.00E -042.00E -04 -5.00E -05 5.00E -19 5.00E -05 Time (s)Time (s)

(a) AlO O Al O2 OAlO

0.041.10

AlO O Al O2 OAlO

0.14 0.12 Molar Fraction

1.35 0.14 1.30 0.12 1.25 0.10 1.20 0.08 1.15 0.06

0.021.05

0.10 0.08 0.06 0.04 0.02

0.001.00 -5.00E -05-05 5.00E -19-19 5.00E -05 -05 1.00E -04 -04 1.50E1.50E -04 -04 2.00E2.00E -04 -04 -5.00E 5.00E 5.00E 1.00E

0.00 -5.00E -05

5.00E -19

5.00E -05

1.00E -04

Time Time (s) (s)

Time (s)

(b)

Figure 1. Volume (a) and molar fractions of the chemical species (b) versus time from ignition to the combustion process of alumina with gaseous oxygen at 1 atmosphere.

1.50E -04

2.00E -04

Figure 2. Volume (a) and molar fractions of the chemical species (b) versus time from ignition to the combustion process of alumina with gaseous oxygen at 10 atmospheres.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.287-292, Jul.-Sep., 2013

0.10

0.08

0.06

0.04

0.02

(a)

Molar Fraction Volume (cm3)

0.12

0.00

Time (s)

2.00E -04

0.14

AlO O Al O2 OAlO

1.30

Molar Fraction

290

0.00 -5.00E -05

Combustion Modeling of Aluminum Incorporated in Low-Explosive Formulations such as Solid Propellants

The molar fractions of the chemical species suffered significant variations and this time the speed of the elementary reactions present in the mechanism is high enough to stabilize the system shortly after ignition. This stabilization ensures a better control of the burning process, providing stability and increasing combustion efficiency. Another simulation was performed, with the system subjected to a pressure of 60 atmospheres, shown in Fig. 4. The difference between the aluminum at 30 and 60 atmospheres in the combustion process is based on the speed at which the burning process stabilizes. At 60 atmospheres, the system volume increases instantaneously to its equilibrium (value greater than that achieved at lower pressures), after the ignition stimulation. In the case of chemical species, there is a mole fraction variation that is approximately its equilibrium value, which is reached almost i nstantly after ignition.

CONCLUSION The kinetics of alumina formation from aluminum was simulated at different pressures in a condition close to the one found in a combustion chamber of a solid propellant rocket motor, not considering the reacting flow to which the system is subjected in a continuous operation. An increase in the pressure at which the reactants are subjected is a d eterminant in the early establishment of a steady-state, in which there is a balance between the chemical species that react and their products. It was observed that the systems subjected to greater pressures tend to stabilize faster, due to the greater chemical speed of elementary reactions present in the combustion mechanism of aluminum, enabling a greater burning stability of the material and higher efficiency.

1.45 1.50

1.40 1.35

1.20 1.15 1.10 1.00

0.10

1.30 0.08 1.20

0.06 0.04

1.10

1.05 5.00E -19

1.40

0.02

5.00E -05

1.00E -04

1.50E -04

2.00E -04

-05 -05 5.00E -19 5.00E 5.00E -5.00E -050.00E+00

0.12

0.10 0.08

1.20 0.06 1.10

0.04 0.02

2.00E 2.00E -04 -04

O O2 Al2O3 AlOAl

0.10 0.08 0.06 0.04 0.02

1.00

2.00E -04

AlO Al OAlO AlOAlO

0.14

O O2 Al2 O 3 AlOAl

Molar Fraction

Molar Fraction Volume (cm3)

1.30

1.50E-04 -04 1.50E

(a) AlO Al OAlO AlOAlO

0.12

1.00E 1.00E-04 -04

Time (s) Time (s)

(a) 1.50 0.14

0.00 0.00E+00 5.00E -05 5.00E -19 5.00E -05 -5.00E -05

1.00E -04 1.00E -04

1.50E -04 1.50E -04

2.00E -04 2.00E -04

0.00 0.00E+00

5.00E -05

1.00E -04

1.50E -04

2.00E -04

Time (s)

TimeTime (s) (s)

(b)

Figure 3. Volume (a) and molar fractions of the chemical species (b) versus time from ignition to the combustion process of alumina with gaseous oxygen at 30 atmospheres.

0.14 0.12 0.10 0.08 0.06 0.04 0.02

1.00 0.00

Time (s)

1.40

O O2 Al2 O 3 AlOAl

Molar Fraction

Molar Fraction Volume (cm3)

Volume (cm3)

1.25

-5.00E -05

AlO Al OAlO AlOAlO

0.14 0.12

1.30

291

Figure 4. Volume (a) and molar fractions of the chemical species (b) versus time from ignition to the combustion process of alumina with gaseous oxygen at 60 atmospheres.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.287-292, Jul.-Sep., 2013

0.00 0.00E+00

292

Gonçalves, R.F.B., Iha, K. and Rocco, J.A.F.F.

In chemical propulsion systems, such as the SPRM, increased operating pressure in the combustion chamber

Acknowledgments The authors thank CAPES, CNPq, and FAPESP for the

leads to displacement of the equilibrium toward the p roducts, thus increasing the yield of the reaction.

financial support rendered.

References Barrere, M., 1968, “Solid Propellant Ignition: General Consideration”. Research Aerospatiale, pp. 15-28. Beckstead, M.W., Derr, R.L. and Price, C.F., 1970, “A Model of Composite Solid Propellant Combustion Based on Multiple Flames”, AIAA Journal, Vol. 8, pp. 2200-2207. Brewster, M.Q., 2000, “Solid Propellant Combustion Response: Quasi-Steady (QSHOD) Theory Development and Validation”, Progress in Astronautics and Aeronautics, AIAA, Vol. 185. Chiaverini, M.J., Harting, G.C., Lu, Y.C., Kuo, K.K., Peretz, A., Jones, H.S., Wygle, B.S. and Arves, J.P., 1999, “Pyrolysis behavior of hybrid-rocket solid fuels under rapid heating conditions”, Journal of Propulsion and Power, Vol. 6, pp. 888-895. Coltrin, M.E., Kee, R.J., Evans, G.H., Meeks, E., Rupley, F.M. and Grcar, J.F., 1991, “A Fortran Program for Modeling One-Dimensional Rotating-Disk/Stagnation Flow Chemical Vapor Deposition Reactors”, Sandia Report. Davenas, A., 2003, “Development of Modern Solid Propellants”. Journal of Propulsion And Power, Vol. 19, pp. 1108-1128. Flandro, G.A., Cai, W. and Yang, V., 2000, “Turbulent Transport in Rocket Motor Unsteady Flowfield”, Progress in Astronautics and Aeronautics, AIAA, Vol. 185.

Gonçalves, R.F.B, Rocco, J.A.F.F., Iha, K. and Machado, F.B.C., 2009, “Modelagem da combustão da dinitramida de amônio por simulação computacional”, Química Nova, Vol. 32. Korobeinichev, O.P., Anisiforov, G.I. and Tereshchenko A.G., 1974, “Study of high-temperature kinetics and mechanism of thermal decomposition of mixtures of ammonium perchlorate-polymeric binder-catalyst using the time-of-flight mass-spectrometer”, 12th Aerospace Science Meeting. Kuo, K.K., Gore, J.P. and Summerfield, M., 1984, “Transient Burning of Solid Propellant”, Progress in Astronautics and Aeronautics, Vol. 90, AIAA, pp. 599-651. Liang, Y. and Beckstead, M.W., 1998, Numerical Simulation of QuasiSteady, Single Aluminum Particle Combustion in Air, AIAA Paper 98-254. Shusser, M., Culick, F.E.C. and Cohen, N.S., 2000, “Combustion Response of Ammonium Perchlorate”, 36 th AIAA, SAE, ASME and ASEE Joint Propulsion Conference, Huntsville, A.L. Summerfield, M., Parker, K.H. and Most, W.J., 1966, “The Ignition Transient in Solid Propellant Rocket Motor”, Princeton University, Princeton, N.J., Aerospace and Mechanical Sciences Report 769. Williams, F.A., Barrere, M. and Hung, N.C., 1969, “Fundamental Aspects of Solid Propellant Rockets”, AGARDograph 116, Technivision Services, pp. 510-531, Slough, England.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.287-292, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.225

Kick Solid Rocket Motor Multidisciplinary Design Optimization Using Genetic Algorithm Fredy Marcell Villanueva1, He Linshu1, Xu Dajun1

ABSTRACT: In this paper, a multidisciplinary design optimization (MDO) approach of a solid propellant kick rocket motor is considered. A genetic algorithm optimization method has been used. The optimized kick solid rocket motor (KSRM) is capable of delivering a small satellite of 200 kg to a circular low earth orbit (LEO) of 600 km altitude. The KSRM should accelerate from the initial apogee velocity of 5000 m/s up to the orbital insertion velocity of 7560 m/s. The KSRM design variables and the orbital insertion trajectory profile variables were optimized simultaneously, whereas the mass characteristics of the payload deployment module were assigned. A depleted shutdown condition was considered, to avoid the necessity of a thrust termination device, resulting in a r educed total mass of the KSRM. The results show that the proposed optimization approach was able to find the convergence of the optimal solution with highly acceptable value for conceptual design phase. Keywords: Kick solid rocket motor, Multidisciplinary design optimization, Genetic algorithm.

INTRODUCTION Using small solid rocket motors (SRM) for space applications became very attractive, because of their advantages compared with liquid propellant driven rocket engines, e specially for insertion of small payloads into a circular Low Earth Orbit (LEO). Simplicity, reliability, and easy to fabricate and operate, are some of the main characteristics of the SRMs. In this article, a specific design used for insertion of a payload into a specified orbit commonly known as kick solid rocket motors (KSRM) are analyzed and discussed. Previous works were performed on analysis of several options of upper stages for different launch vehicles (LV) (McGinnis and Joyner, 2005), and in design and optimization of upper stages for transfer from LEO to geostationary earth orbit (Motlagh and Novinzadeh, 2012). Casalino et al., (2011) carried out an optimization of a hybrid upper stage configuration with a special emphasis on the grain geometry. He Linshu and Murad (2005) developed an improved method for conceptual design of multistage solid rockets based on depleted shutdown condition, which means that the SRM has to burn all-contained propellant, resulting in a reduced gross mass. The genetic algorithm (GA) global optimization method is increasingly being used in optimization of aerospace and propulsion systems (Kamran et al., 2009; Tedford and M artins, 2010; Riddle et al., 2007; Rafique et al., 2009). B ayley and Hartfield (2007) used

1.School of Astronautics – Beihang University – Beijing – China Author for correspondence: Fredy Marcell Villanueva | Beihang University | XueYuan Road No.37, HaiDian District, New Main Building B1011 | Beijing 100191 – China | Email: marcell385@yahoo.com Received: 16/01/13 | Accepted: 30/04/13

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

294

Villanueva, F.M., Linshu, H. and Dajun, X.

GA in multidisciplinary design optimization (MDO). Davis (2001), Goldberg (1989), and Fletcher (2000) provided a comprehensive description and step-by-step implementation of the GA in design and optimization of complex systems. The insertion process of a payload into the orbit of a typical solid propellant LV start after a ballistic flight phase, when the upper stage is in the apogee altitude, which c orrespond to the required circular orbital altitude. At this point, to reach the required orbital velocity, a kick impulse is necessary, and for this purpose, a KSRM is required. Its propellant should be burnt completely (depleted shutdown condition), avoiding the use of a thrust termination device, resulting in a reduced total mass of the KSRM. The insertion accuracy is maintained by using an attitude control system grouped in a payload deployment module (PDM). Thus, the objective of this article is to describe the optimization approach of a KSRM for insertion of a small payload into a prescribed circular LEO orbit by using a well-performed GA optimization method as a global optimizer.

OPTIMIZATION STRATEGY The MDO approach considered aims to find the KSRM optimal design and the corresponding trajectory to successfully accomplish the specified mission, which is an insertion of a small payload into a prescribed circular LEO orbit. To accomplish this task, a widely used GA optimization method is considered. The main advantage of this method relies on its independency of initial point to perform the optimization. The GA is considered to be a powerful heuristic global optimization tool in solving complex optimization problems that traditionally has been solved using the approximated analysis. An additional advantage of the GA is the ability to solve discrete and continuous variables. The operation sequence of the GA optimization method is shown in Fig. 1. Here, the procedure starts with the assignment of the design variables, the algorithm performs several operations as population initialization, selection,

Design Variables Population initialization Kick Solid Rocket Motor Multidisciplinary Design Analysis

Selection

Propulsion analysis

Crossover

Mass analysis Trajectory analysis

Mutation

Stopping criteria Yes Optimal Solution

Figure 1. Genetic algorithm optimization approach.

crossover and mutation until the optimal solution is reached, whereas all the considered constraints are satisfied. The drawback of the GA is its computational expense, because a large number of function evaluations are required before finding the optimal solution. The main characteristics of the GA are presented in Table 1. Table 1. Genetic algorithm characteristics.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

Variables

Characteristics

Generations

200

Population size

100

Stopping criteria

function tolerance 10e-6

Population type

double vector

Selection

stochastic uniform

Crossover

single point pc=0.8

Mutation

uniform pm=0.25641

Reproduction

elite count=2

Function evaluation

2000

Kick Solid Rocket Motor Multidisciplinary D esign Optimization Using Genetic Algorithm

KICK SOLID ROCKET MOTOR MODEL Kick solid rocket motor definition Mission analysis The mission is to deliver a small payload to a circular LEO orbit of hf=600 km altitude. An upper stage composed of a KSRM, a PDM of 100 kg, and, a small payload of 200 kg is considered for this research. The KSRM should accelerate the upper stage from an initial apogee velocity of V0=5000 m/s up to the insertion orbital velocity of Vf=7560 m/s, whereas the PDM containing the attitude control system maintains the upper stage witting the allowable insertion accuracy. KSRM design There can be many variants of design options; however, for this research effort, a classical configuration was considered. The rocket motor case is considered to be made of high-strength steel, titanium alloy for attachment parts, the ignition system is located in the forward central part of the chamber, and the nozzle has a thrust vector control (TVC) system. Grain characteristics For the present analysis, a hydroxyl-terminated polybutadiene (HTPB) grain is selected due to its high performance and suitable for space applications. The composite grain has an internal port and an equivalent length was adopted instead of a complex 3D geometry. This assumption considerably simplifies the model and is acceptable for the conceptual design phase. Propulsion analysis The propulsion analysis of the KSRM can be calculated using the classical approach. Sutton and Biblarz (2001) and He Linshu (2004a, 2004b) provided a detailed propulsion analysis including the essential parameters, like propellant mass flow rate, burn time, thrust, and nozzle parameters. In this analysis, a constant in time burning surface is considered; a grain geometry shape coefficient ks is used to represent the constant burning surface of the grain Sb as a function of the equivalent length of grain Lgn and diameter Dgn as follows:

ks =

295

Sb (1) Dgn Lgn

The mass of the grain is calculated from the design variables and propulsion analysis, and the burn time tb, mass . 2 the areks λgn Dgn of the grain mgn, and mass flow rate mgn =ofρ gn u Sgrain b = ρ gnu calculated as shown below:

mgn =

tb =

π 4

3 ρ gnηv λgn Dgn

π η v Dgn

(2)

(3)

4 u ks

. 2 mgn = ρ gnu Sb = ρ gnu ks λgn Dgn (4)

λgn =

Lgn

(5)

Dgn

where, u is the burning rate of propellant, ρgn is the density of grain, Lgn=Lm+0.3151Dm is the equivalent length of the grain, Dgn=Dm is diameter of the grain, Lm is the rocket motor cylindrical length, λgn is fineness ratio of the grain (grain length/ diameter), and ηv is the grain volumetric loading fraction. The nozzle throat area At, expansion ratio ε, and nozzle exit area Ae are calculated as follows:

At =

ε=

ρ gn u Sb Γ0 pc max

Rc Tc Γ0

(6)

γ −1 1 ⎡ ⎤ (7) ⎛ pe ⎞ γ 2γ ⎢ ⎛ pe ⎞ γ ⎥ ⎜⎜ ⎟⎟ 1 −⎜⎜ ⎟⎟ ⎝ pc ⎠ γ −1 ⎢⎢ ⎝ pc ⎠ ⎥⎥ ⎣ ⎦

Ae = At ε

(8)

γ +1

⎛ 2 ⎞ 2(γ −1) Γ0 = γ ⎜⎜ ⎟⎟ ⎝ γ +1⎠

(9)

where, Sb is the burning surface of grain, Rc=326 J/(kg.K) is gas constant, Tc=2790Kº is temperature in the combustion chamber, pe is exit pressure, pc is chamber pressure,

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

296

Villanueva, F.M., Linshu, H. and Dajun, X.

pcmax=1.1pc is the maximum value of chamber pressure, and γ=1.21 the specific heat ratio of the gas. The vacuum specific impulse Ivac , and the thrust T can be sp calculated as shown below:

p  I spvac = I spa + e   pc 

γ −1 γ

T = Ispvac m gn − pa Ae

Rc Tc (10) g 2 I spa (11)

where, pa is the atmospheric pressure, I asp is average specific . impulse, g is acceleration due to gravity, m gn is mass flow, and Ae is the nozzle exit area. Mass analysis The mass of the upper stage is represented by the following equations:

m0 = mKSRM + mPDM + mPAY .

mKSRM = mst + mgn

(12)

(13)

where, m0 is the upper stage mass, mKSRM is kick solid rocket mass, m PDM is payload deployment module mass, m PAY is payload mass, and m st is the structural mass of the KSRM. He Linshu (2004b) provided a detailed calculation of the KSRM structural mass, which is highly acceptable for conceptual design phase, and is shown in the following equation:

mst = mcy + mc1 + mc 2 + mq + m j1 + m j 2 + min , c1 + min , c 2 + min , cy + mnoz , ec + mnoz , sh

It is composed of the mass of the motor cylinder, mcy, motor dome ends, mc1 and mc2 , forward and aft skirts, mq , forward and aft joints, mj1 and mj2, forward and aft insulation liners, min,c1 and min,c2 , cylindrical section insulation liner, min,cy , nozzle expansion cone, mnoz,ez , nozzle spherical head, mnoz,sh , nozzle insulation, mnoz,in, igniter, mig , thrust vector control, mTVC , cables, mcab, and mass of attachment parts, map . The materials selected for the motor case are highstrength steel and titanium alloy, ethylene propylene diene monomer for chamber insulation, and carbon phenolic for the nozzle, whereas the factor of safety was taken as 1.5. The mass of cylindrical part of the chamber can be calculated by the following relation:

8σ b

(15)

where, Kcy=1.02 is the ratio of the case cylindrical length to rocket motor equivalent length (Lcy/Lm), which was obtained from statistics, f is factor of safety, Dch=Dm is diameter of chamber, and σb is the ultimate strength of the chamber material. The mass of the forward motor dome end mc1 can be calculated as follows:

mc1 =

πλ2e Dch3 f f p pc   1 1 −  2 2 2 8(λe − 1)σ cos θ 2  1 + (λe − 1)sin θ 2 

(16)

where, λe=2 is the chamber ellipsoid ratio, σ is strength ratio σ=σb/ρcl, where, ρcl is the density of closure material and θ2 taken as 60°. The relative pressure in the chamber fp is determined as shown below:

fp=

pc , max pc

(17)

The aft motor dome end is calculated as follows:

mc 2 =

πλ2e Dch3 f f p pc 8(λ2e − 1)σ cosθ 2

2 2   2 2 1 .  λe Dch − (λe − 1)d n −  2 2 2 2 λe Dch 1 + (λe − 1)sin θ 2  

(14)

+ mnoz ,in + mig + mTVC + mcab + map

3K cyπ f f p Dch3 pc λgn

mcy =

(18)

where, dn=0.4Dch is the diameter of closure rear end opening for nozzle. The mass of the forward and aft skirts of the SRM are calculated as follows:

mq = π Dch2 ρ qδ q

lq1 + lq 2 Dch

(19)

where, ρq is the density of skirt material, δq is thickness of skirts, and lq1 and lq2 are lengths of forward and aft skirts. The mass of the forward joint to the igniter mj1 and the aft joint to nozzle assembly mj2 can be calculated as shown below:

m j1 =

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

3 π f f p p c d n3

σj

(20)

Kick Solid Rocket Motor Multidisciplinary D esign Optimization Using Genetic Algorithm

σj =

σb j ρ j (21)

m j 2 = m j1

(22)

where, σj is the strain of joint, ρj is density of the joint material, and σbj is ultimate strain of joint. The forward and aft insulation liners min,c1, min,c2 and cylindrical section insulation liner min,cy can be determined using Eqs. 23 to 25.

min , c1 =

1 ρ inπDch2 Ra tb (23) 4

min ,c 2 =

ρ inπD Ra tb  λ D − (λ − 1)d n    (24) 4(λ − 1)  D  2 ch 2 e

2 e

2 ch

2 e 2 ch

where, ρin is the density of insulation material and Ra is the rate of ablation of the insulation material.

min ,cy = π Dch Lcy ρ in

(ε in cin ρ in + α gi ccy ρ cyδ cy )

αgi cin ρin

⎡ ⎤ 2ε in (ln θ p ccy ρ cyδ cy + α gi tb )α gi cin ρ in (25) . 1 − ⎢ 2 − 1⎥ ln θ p (ε in cin ρ in + α gi ccy ρ cyδ cy ) ⎣⎢ ⎦⎥ θp =

Tg − Tcy Tg − TI

(26)

where, Lcy is the length of case cylindrical section, εin is heat transfer coefficient of insulation material, cin is specific heat capacity of insulation, ccy is specific heat capacity of the cylindrical section, αgi is heat transfer coefficient from gas to insulation, ρcy is density of the case cylindrical section material, Tg is temperature of gas, Tcy is allowable temperature of the cylindrical section, and TI is the initial temperature of the cylindrical section. The mass of the nozzle expansion cone mnoz,ec, nozzle spherical head mnoz,sh, and nozzle insulation mnoz,in are expressed using the Eqs. (27), (28), and (29), respectively.

mnoz ,ec = ρ ec

 Ae   − 1 d t 4 sin β n  At 

4  d .  f × 0.67 6 e − 1 (1 − S )3 dt 

Pc maxσ ec   Eec 

(27)

297

where, ρec is the material density of the expansion cone, βn is nozzle expansion half angle, S is submerged coefficient of the nozzle, de is the nozzle exit diameter, dt is diameter of throat, σec is ultimate strength of the expansion cone material, and Eec is the elastic modulus of expansion cone material.

mnoz , sh = 3.656 ρ sh d t3

(28) where, ρsh is the material density of spherical head of nozzle. mnoz ,in =

innz

S nz

(29)

innz

where, ρinnz is the material density of the nozzle insulation, Snz is surface area of the nozzle, and δinnz is the thickness of the nozzle insulation. The igniter mass mig, can be calculated as follows: 1. 2

 7π d t2   mig =1.454  4 

(30)

The nozzle thrust vector control mass mTVC, is considered as shown below:

mTVC = 0.235mnoz (31) The mass of the cables mcab and attachment parts map are expressed in Eqs. (32) and (33). mcab = 1.3Lcy ρ c

(32)

where, ρc is the linear density of the cable (kg/m). And finally, the mass of the attachment parts is calculated as follows:

map =(6.13 . 10 -7 ) D m2 L1cy.148

(33)

Trajectory analysis For trajectory analysis, a 3 degree of freedom (3DOF) model has been developed and implemented using SIMULINK (Zipfel, 2007; Fleeman, 2001). This model uses the upper stage mass, including the KSRM, PDM and payload mass, and thrust as input parameters. The upper stage is considered as a point-mass in two-

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

298

Villanueva, F.M., Linshu, H. and Dajun, X.

dimensional coordinate systems (2D) over a spherical and non-rotating earth, where the Coriolis and centrifugal forces are not considered. The considered state variables are velocity, flight path angle, altitude, and mass, whereas the control variable is the programmed angle of attack. The trajectory analysis computes the state variables by solving the equation of motion presented in Eq. 34, and evaluating the constraint conditions at every phase of flight. Figure 2 illustrates the forces acting on the upper stage and below a set of governing equations of motion (He Linshu, 2004a; Xiao, 2001). dV T cos α = − g sinϑ dt m dϑ T sinα g cos ϑ V cos ϑ = − + dt mV V Re + h

nx =

T cos α ≤ nx max mg

(35)

ny =

T sin α ≤ n y max mg

(36)

The density variation with altitude can be represented as follows:

ρ = ρ0 e (− h/β) (37) The variation of gravity with altitude can be represented as follows:

dh = V sin ϑ dt Re dl V cosϑ = dt Re + h

(34)

µ ( Re + h) 2

(38)

where, ρ0 is the sea level density, β is the density scale height, and μ is the earth gravitational parameter. The required circular orbital insertion velocity Vf for a given final altitude hf can be represented as follows:

α =η + ϕ −ϑ η=

The axial and normal overload coefficients can be calculated in a body centered velocity coordinate systems (0, x, y) and represented as follows:

l Re

where, V is the velocity, m is vehicle mass, φ is pitch angle, θ is trajectory angle, ϑ is flight path angle, η is range angle, Re is radius of earth, h is height above the ground, and l is the range.

α V

Y y

mg η h

Re X

Figure 2. Forces acting on the upper stage.

Vf =

µ h f + Re

(39)

Orbit insertion profile sequence The trajectory optimization was performed considering different constraints that prevent the KSRM, its components, and the payload from failure. The trajectory is modeled considering a depleted shutdown condition at the time of insertion, as follows (He Linshu, 2004b; Qazi and He Linshu, 2005): Initial Condition For a typical solid propellant LV orbital payload insertion, the trajectory starts from the time where the upper stage is in the apogee altitude of h0=600 km, the initial condition is that the flight path angle should be ϑ0=0 degrees, and the initial velocity V0=5000 m/s, the angle of attack also should be zero degrees α0=0. Pitch over insertion maneuver After ignition, the upper stage accelerates by the power of the KSRM and flies following the orbital altitude. This phase

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

Kick Solid Rocket Motor Multidisciplinary D esign Optimization Using Genetic Algorithm

comprises up to the end of burning time (depleted shutdown condition). A second order curve is fitted from the initial time to the final insertion time; this curve represents the programmed angle for this phase. Insertion condition At this time all parameters should comply with the insertion accuracy requirements. The flight path angle at this time should approach zero degrees and within the boundary conditions, ϑf≈0, the programmed angle of attack is constrained to approach zero αf≈0, the orbital velocity should be Vf=7560 m/s corresponding to hf=600 km altitude, and the normal and axial acceleration should be less than its allowable maximum values. Orbit insertion profile formulation The variation of the flight path angle during insertion flight has substantial influence on the injection accuracy in orbit, acceleration loads, and final orbital velocity. It is influenced by a programmed angle of attack. Figure 3 explains the insertion maneuver, and the angle of attack is programmed using the following relations (He Linshu, 2004a; He Linshu, 2004b; Xiao, 2001):

α prog (t ) = α max sin 2 f (t ) f (t ) =

km =

π (t − t1 ) k m (t 2 − t ) + (t − t1 )

t m − t1 t2 − tm

(40) (41)

(42)

where, αmax is the maximum angle of attack, αprog (t) is the programmed angle of attack, km is the insertion maneuver variable, t is time of flight, tm is time corresponding to maximum angle of attack, t1 is time of start of insertion maneuver, and t2 is insertion time, coincident in value with the KSRM burning time tb.

DESIGN OPTIMIZATION PROBLEM Objective function For the present research effort, the objective is to minimize the KSRM mass mKSRM. The mathematical description of the objective function is as follows:

299

αmax

Figure 3. Programmed angle of attack variation.

min mKSRM = f ( X )

(43)

g j(X ) ≤ 0

(44)

hk ( X ) = 0

(45)

X lb ≤ X i ≤ X ub (46) where, X is the set of variables, Xlb is lower bound of variables and Xub is the upper bound of variables. Design variables The variables considered in the KSRM design and the insertion maneuver trajectory can be represented in Eq. (47) and listed in Table 2. X = [ Lm , Dm , pc , pe , k s , u , ρ gn ,η v ,α max , k m ] (47)

Table 2. Design variables.

Variables

Symbol

Units

Rocket motor cylindrical length

Rocket motor diameter

Chamber pressure

Nozzle exit pressure

Coefficient of grain shape

Grain burning rate

m/s

Grain density

pgn

kg/m3

Grain volumetric loading

ηv

Maximum angle of attack

αmax

X10

Insertion maneuver variable

deg

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

300

Villanueva, F.M., Linshu, H. and Dajun, X.

Design constraints To prevent the failure of the upper stage, KSRM and PDM, several constraints were selected, and are listed in Table 3.

Table 3. Design constraints.

Design sequence This section describes a step-by-step sequence for multidisciplinary design of the KSRM as follows:

Constraints

Value

Units

Orbit insertion velocity

Vf=7560±1

m/s

Final altitude

hf=600±0.1

Axial overload

nx ≤ 12

Normal overload

ny ≤ 2

Maximum angle of attack

αmax ≤ 10

deg

Orbit insertion angle

ϑf=0±0.02

deg

Step

Procedure

Reference

Fineness grain ratio

Define initial variables Χ

Eq. 47

λgn ≤ 2

Nozzle exit diameter

de ≤ 0.95 Dm

Define constraints C

Table 3

2 Eq. 2 Calculate grain mass mgn = ρ gnu Sb = ρ gnu ks λgn Dgn

KSRM total length (including nozzle)

LKSRM ≤ 1.8

Calculate burning time tb

C10

Burning time

tb ≤ 50

C11

Nozzle expansion ratio

ε ≤ 80

4 Calculate mass flow mgn = ρ gnu Sb = ρEq. gnu ks λgn Dgn

Calculate throat area At

Eqs. 6 – 9

Calculate specific impulse I vac sp

Eq. 10

Calculate thrust T

Eq. 11

Calculate KSRM mass mKSRM

Eqs. 13 – 33

Set constant values mpdm, mpAY

Mission

Calculate upper stage mass m0

Eq. 12

Trajectory conditions

Orbit Insertion Profile Sequence

Calculate αprog (t)

Eqs. 40 – 42

Calculate ρ(h)

Eq. 37

Calculate g(h)

Eq. 38

Calculate V(t), ϑ(t), h(t), l(t),

Eq. 34

Calculate overloads nx, ny

Eqs. 35, 36

Calculate final values Vf, hf, ϑf

Eq. 34

Check constraints

Table 3

Back to step 1

Eq. 3

V0 , h0 , ϑ0 , α 0

profile successfully reached the objective function. The optimized values of design variables were obtained, and these variables did not violate the considered design constraints. Table 4 shows the lower bound, upper bound, and optimized values of the design variables. There are several parameters that characterize the KSRM; however, only the most important were calculated from the optimized design variables and are presented in Table 5.

Table 4. Optimum values of variables.

Variables

Upper Bound

Optimized Value

0.65

0.90

0.8141

0.65

0.90

0.8140

ks u

m/s

ρgn

kg/m3

ηv

OPTIMIZATION RESULT The optimization results show that the optimized KSRM as well as the upper stage insertion trajectory

Lower Bound

αmax

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

deg

60e

70e

65.901e5

0.05e5

0.20e5

0.0581e5

1.50

1.90

1.5050

6.5e-3

8.5e-3

6.651e-3

1725

1735

1727.01

0.80

0.86

0.8176

7.5

9.0

8.3601

0.28

0.45

0.3222

Kick Solid Rocket Motor Multidisciplinary D esign Optimization Using Genetic Algorithm

Table 5. KSRM optimum values.

Symbol

Unit

Optimized Value

Upper stage mass

1061.81

Rocket motor mass

mKSRM

761.816

Grain mass

mgn

679.358

Structural mass

mst

82.488

Burning time

45.677 2544.12

KSRM total length

LKSRM

1.733

Nozzle throat diameter

0.062

Nozzle exit diameter

0.5409

Nozzle expansion ratio

6.5 6

30 Time(s)

600.05 600

76.13

Figures 4 and 5 illustrate the optimized insertion trajectory profile of the upper stage. From Fig. 4 it can be seen that the burning time of the KSRM is 45.67 seconds, the end of burning time shows coincidence with the required insertion parameters. Additionally, the figures evidenced that the payload is inserted at the required altitude of 600 km and the obtained circular orbital velocity is 7560 m/s. The insertion flight path angle is within the required accuracy, whereas the axial and normal overloads were maintained within the constraints limits. Sensitivity analysis Monte Carlo analysis is widely used in system and early stage of design. It provides a relatively accurate statistical evaluation of the response distribution under input uncertainties. Monte Carlo sensitivity analyses of the main KSRM parameters were conducted to investigate the effect of uncertainties of design variables over the expected result. For this analysis, a ±1% error was added to every optimized value of design variables. The results are presented in Table 6, and the scatter plot is shown in Fig. 6. The GA optimization method considered successfully reached the optimal solution, a population of 100 with 200 generation was sufficient to perform the present study; however, several trials had been carried out to obtain the desired accuracy of the optimal solution.

Altitude (km)

N.s/kg

599.95 599.9 599.85 599.8

30 Time(s)

0.2 Flight path angle (deg)

5.5

0.1 0 -0.1 -0.2 0

10 Angle of attack (deg)

Specific impulse

vac sp

7.5 Velocity (km/s)

Parameter

vac

301

8 6 4 2 0

Figure 4. Insertion trajectory profile of the upper stage.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

Villanueva, F.M., Linshu, H. and Dajun, X.

302

1000

1100 Gross mass (kg)

1120

Gross mass (kg)

1200

800 600 400 200

30 Time(s)

Velocity (km/s)

Thrust (KN)

400 600 Monte Carlo runs

800

1000

200

400 600 Monte Carlo runs

800

1000

200

400 600 Monte Carlo runs

800

1000

800

1000

7.6 7.65 7.5 7.45

30 Time(s)

600.2

Altitude (km)

12 10 8 6

600.1 600 599.5 599.8

30 Time(s)

0.8

40 39

0.6 Thrust (kN)

Normal overload

200

7.7

0.4 0.2 0

7.65

Axial overload

1040

1000

1060

1020

1080

38 37 36

30 Time(s)

Figure 5. Flight parameters of the upper stage.

200

400 600 Monte Carlo runs

Figure 6. Monte Carlo sensitivity analysis.

J. Aerosp. Technol. Manag., SĂŁo JosĂŠ dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

Kick Solid Rocket Motor Multidisciplinary D esign Optimization Using Genetic Algorithm

303

Table 6. Monte Carlo results.

Parameter

Unit

Minimum value

Maximum value

Gross mass

1030.6

1098.2

1063.9

Velocity

m/s

7515.1

7648.2

7579.2

Altitude

Thrust

599.86

600.12

36.383

39.184

Mean value

Standard deviation

600.00 37.846

11.771 0.0219 0.0473 0.497

trajectory profile c haracteristics required for insertion a small

CONCLUSION

payload into a circular LEO orbit of 600 km.

A GA based optimization approach has been applied to conceptual design and optimization of a KSRM. The advantage of the GA relied on its independency of initial point to calculate the optimum. A 2D dynamic model was developed to simulate the orbital insertion trajectory of the

The results of the KSRM parameters are shown in Tables 4 and 5 and its insertion trajectory profile is shown in Figs. 4 and 5, and Monte Carlo sensitivity analysis shown in Fig. 6 evidenced the validity of the used approach for the early stage of the design process.

upper stage composed of the PDM of 100 kg, the optimized a payload of 200 kg. A sensitivity analysis was conducted using Monte Carlo method to investigate the variation of the main parameters of the KSRM. The emphasis of this research was to find the optimal KSRM design and the upper stage insertion

ACKNOWLEDGMENTS Fredy Villanueva wishes to thank Beihang University and China Scholarship Council (CSC), for financial support.

REFERENCES Anderson M.B., Burkhalter, J. and Jenkins, R., 2001, “Multi-Disciplinary Intelligent Systems Approach to Solid Rocket Motor Design. Part I: Single and Dual Goal Optimization”, 37th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, AIAA, paper 2001-3599. Bayley, D.J. and Hartfield, R.J., 2007, “Design Optimization of a Space Launch Vehicles for Minimum Cost using a Genetic Algorithm”, AIAA 43rd AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, AIAA, paper 2007-5852. Casalino, L., Pastrone, D. and Simeoni, F., 2011, “Hybrid Rocket Upper Stage Optimization: Effects of Grain Geometry”, AIAA 47th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, AIAA, paper 2011-6024. Davis, L., 2001, “The Handbook of Genetic Algorithm”, Van Nostrand Reingold, New York. Fleeman, L.E., 2001, “Tactical, Missile Design”, AIAA, Reston. Fletcher, R., 2000, “Practical Methods of Optimization”, John Wiley & Sons. Goldberg, D., 1989, “Genetic Algorithm in Search, Optimization and Machine Learning”, first edition, Addison-Wesley–Longman Publishing Co., Reading, MA.

He Linshu, 2004a, “Launch Vehicle Design”, Beijing University of Aeronautics and Astronautics Press, Beijing. He Linshu, 2004b, “Solid Ballistic Missile Design”, Beijing University of Aeronautics and Astronautics Press, Beijing. He Linshu and Murad Y., 2005, “An Improved Method for Conceptual Design of Multistage Solid Rockets based on Depleted Shutdown”, Journal of Solid Rocket Technology, Vol. 28, No 2, paper 120-125. Kamran, A., Lian, G., Godil, J., Siddique, Z., Zeeshan, Q. and Rafique, A., 2009, “Design and Performance Optimization of Finocyl Grain”, AIAA Modeling and Simulation Technologies Conference 10 -13 A ugust 2009, Chicago, Illinois, AIAA, paper 2009-6234. McGinnis, P.M. and Joyner, C.R., 2005, “Upper Stage Propulsion Options for Various Launch Vehicle Architectures”, AIAA 41st AIAA/ ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, AIAA, paper 2005-4180. Motlagh, J.A. and Novinzadeh, A.B., 2012, “Solid Upper Stage Design Process using Finite Burn Maneuvers for Low Earth Orbit– Geosynchronous Earth Orbit Transfer Phase”, Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

304

Villanueva, F.M., Linshu, H. and Dajun, X.

Qazi, M. and He Linshu, 2005, “Rapid Trajectory Optimization u sing Computational Intelligence for Guidance and Conceptual Design of Multistage Space Launch Vehicles”, AIAA Guidance, Navigation, and Control Conference and Exhibit, AIAA, paper 2005-6062. Rafique, A.F., He Linshu, Zeeshan, Q., Kamran, A., Nisar, K. and Wang, X.W., 2009, “Integrated System Design of Air Launched Small Space Launch Vehicle using Genetic Algorithm”, 45th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, AIAA, paper 2009-5506. Riddle, D.B., Hartfield, R.J., Burkhalter, J.E. and Jenkins, R., 2007, “ Genetic Algorithm Optimization of Liquid Propellant Missile Systems”, AIAA Aerospace Sciences Meeting and Exhibit, AIAA, paper 2007-0362.

Sutton, G.P. and Biblarz, O., 2001, “Rocket Propulsion Elements”, Wiley-Interscience, New York. Tedford, N.P. and Martins, J.R., 2010, “Benchmarking Multidisciplinary Design Optimization Algorithms”, Optimization and Engineering, Vol. 11, No 1, paper 159-183. Xiao, Y.I., 2001, “Rocket Ballistics and Dynamics”, Beijing University of Aeronautics and Astronautics. Zipfel, P.H., 2007, “Modeling and Simulation of Aerospace Vehicle Dynamics”, second edition, AIAA Education Series, AIAA, Reston.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.293-304, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.190

Simulations of the Atmospheric Boundary Layer in a Wind Tunnel with Short Test Section Luciana Bassi Marinho Pires1, Igor Braga de Paula2, Gilberto Fisch3, Ralf Gielow2, Roberto da Mota Girardi4

ABSTRACT: This article presents a study of three different passive devices (spires, screens, and a carpet) separately and in various combinations, to simulate the atmospheric boundary layer (ABL) in a wind tunnel with a short test chamber (465×465x1200mm). The influence of distances between these devices on the formation of the ABL is established, and optimization of variation of thicknesses of the screens (thin, medium, and coarse) on pressure loss is explored. The results obtained in this work gave support for the analysis of the atmospheric flow and turbulence at Alcantara Launching Center (ALC) in order to launch Brazilian space vehicles under safe conditions. The results show that the “spires” and the thin screen are the devices that require the least area to form an ABL in a test chamber. The physical proximity of two devices (the spires and the medium screen) also influences the size of the ABL formed, which varies from 180 to 200 mm. The power law exponent ranged from 0.12 up to 0.14 after the insertion of a carpet. Keywords: Passive methods, Spires, Carpet, Screens, Wind power law.

INTRODUCTION The first boundary layer (BL) wind tunnel was built in 1965 by Davenport (1967) at Western University (formerly University of Western Ontario), Ontario, Canada. The studies of Jensen and Franck (1963, 1965) and Franck (1963) in wind tunnels concluded that it would be possible to obtain an appropriate scale of the natural wind structure by covering a considerable length of the wind tunnel’s floor with a material of suitable roughness. However, the disadvantage of this process is that it requires a length of about 25 m to form a BL with 60 to 120 cm height, which is possible only in tunnels with a long test chamber (Blessmann, 1973). Thus, improved techniques for reproduction of the main characteristics of natural winds, as well as the formation of the atmospheric boundary layer (ABL), are needed. These techniques will permit shorter test chambers, so that existing aeronautical tunnels could be used for atmospheric simulations of meteorological interest, with the advantage of flow control and improved data collection. The objective of the current article is to provide directions to fulfill the challenges of adaptation of aeronautic wind tunnels for meteorology purposes, cataloging the steps of modification and their expected results. This work was done at Departamento de Ciência e Tecnologia Aeroespacial (DCTA) wind tunnel to give support for the analysis of the atmospheric flow and turbulence at Alcantara Launching

1.Instituto Nacional de Pesquisas Espaciais – São José dos Campos/SP – Brazil 2.Pontifícia Universidade Católica do Rio de Janeiro – Rio de Janeiro/RJ – Brazil 3.Instituto de Aeronáutica e Espaço – São José dos Campos/SP – Brazil 4.Instituto Tecnológico de Aeronáutica – São José dos Campos/SP – Brazil Author for correspondence: Luciana Bassi Marinho Pires | Instituto Nacional de Pesquisas Espaciais – INPE/CPTEC | Avenida dos Astronautas, 1.758 – Jardim da Granja | CEP 12.227-010 São José dos Campos/SP – Brazil | Email: lubassimp@gmail.com Received: 13/11/12 | Accepted: 15/05/13

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

306

Pires, L.B.M., Paula, I.B., Fisch, G., Gielow, R. and Girardi, R.M.

Center (ALC). This is an ongoing project that aims to develop technical tools for Aerospace Meteorology associated with the launching of the Satellite Vehicle Launcher (VLS) within the Brazilian Space Program (Pires et al., 2010). There are several studies using the wind tunnel for simulating the characteristics and behavior of the atmosphere. Some examples are the works of Novak et al., (2000) analyzing the turbulent structure of the atmosphere within and above canopies; Kwon et al., (2003) simulating the atmospheric wind field over a complex topography in order to plan the Naro Space Center in South Korea; Mavroidis et al., (2003), studying pollutant dispersion fields immersed in obstacles; Cao and Tamura (2006), simulating the turbulent BL flow over 2D steep, smooth, and rough hills in Tokyo. Kozmar (2009) reproduced the BL at different scales inside the wind tunnel to address the proper choice of simulation length scale and concluded that the length-scale factor does not influence the generated ABL models when using similarity criteria; and recently, Carpentieri et al., (2012) worked to quantify the mean and turbulent flow in geometries of real street canyon intersections, focusing on the area surrounding the intersection between Marylebone Road and Gloucester Place in Central London, UK. Recently, Avelar et al., (2012) did a detailed analysis of the atmospheric flow at ALC using a large wind tunnel. They have constructed the ABL profile using a combination of barrier and small blocks of wood. As all of them used wind tunnels for their experiments, new techniques and methods to develop the ABL are necessary in order to use different sizes of wind tunnel section tests. The methods of simulating the ABL in wind tunnels are divided into passive and active types. The passive methods use barriers such as grilles, flat plates, triangular plates, writing pads, spires, carpets, etc. The active methods are those that use air jets to form a wall of fluid. The objective of this work is to present some of the passive devices (screens, spires, and carpet) used for the formation of the BL in wind tunnels. It will be done analyzing the optimizations of the formation of the ABL with the use of such devices as a function of both the distance among them and their characteristics (grids with larger or smaller spacing) based primarily on the mean velocity profile and turbulence intensity profiles. The main focus was on rural terrain. However, additional studies could investigate whether this approach works well when applied to other types of terrains (e.g., suburban, urban) with different geometries.

MATERIAL AND METHODS This work was carried out in the commercially available open subsonic wind tunnel Plinth & Partners LDD Wokingham Berkshire England (Serial No. 44/5065 TE) at the Kwein Lien Feng Laboratory at Instituto Tecnológico de Aeronáutica (ITA)/ DCTA, São José dos Campos, SP, Brazil (Fig. 1a). Its test section is square with cut corners (465x465 mm) with a length of 1200 mm. For this study, a channel apparatus with a width of 410 mm and length of 1200 mm, consisting of an uncovered wooden frame, with free edges and side walls parallel to each other and perpendicular to the floor of the wind tunnel, was used to extend the test section for the formation of the ABL. The walls of the tunnel and of the “channel” (Fig. 1b) do not coincide, to minimize the lateral BL of the tunnel. An automatic positioner device with an arm capable of moving in three directions perpendicular to each other and with an accuracy of tenths of a millimeter was used, coupled to a computer, with the data collection points inserted by automatic handling equipment or manually (Fig. 1b). The positioner consists of the following elements: (i) positioner, (“traversing”) from Dantec Dynamics serial code 9057 h 0123, Denmark, (ii) controls for the positioner, Dantec type 57 b 100; and (iii) command manual positioner, serial code 9055X530. The same configuration was utilized by Roballo (2007). The atmospheric flow is simulated by 22 kW (30 hp) electric fans for the range of the windspeeds. The maximum windspeed was 33 m/s (120 km/h). For initial adjustments, passive devices — screen and spires — were used and for fine tuning a wrinkled carpet was added. It is important to note that the insertion of screens within wind tunnels is usually done with a primary objective of reduction of turbulence of flow, because they have a tendency to uniform the flow and to break the big vortices into smaller ones that decay quickly (energy cascade process). However, in this study, the screens were used to accelerate the formation of the ABL in the test section. Three types of screens with different meshes were used: (i) nylon square mesh with spacing of 2×2 mm2 and 0.4 mm in diameter, called “thin screen,” (ii) metal square mesh with spacing of 5.5×5.5 mm2 and 1.0 mm diameter, called “medium screen”; and (iii) metal mesh type “beehive” with a spacing of 19.0×17.0 mm2 and 0.5 mm in diameter, called “coarse screen” (Fig. 2). All these screens were available in local shops. The screens were assembled in a rigid frame made of steel stripes

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

Simulations of the Atmospheric Boundary Layer in a Wind Tunnel with Short Test Section

(a)

307

(b)

Figure 1. Wind tunnel of Prof. Kwein Lieng Feng Laboratory (ITA) (a) and automatic positioner and extension attached (b).

of 2 mm thick. Their tension was adjusted with small plates used to tighten the mesh to the frame. This was necessary in order to avoid the screen bending from aerodynamic loads during measurements. The mechanism created a small step of approximately 3 mm (approximately 1% of the BL thickness) at the basis of the BL. This height is considered small for the current experiments. For instance, the carpet used to form the ABL had nearly the same height as the step, but a much longer length. Thus, the influence of such a small step on the ABL was assumed to be negligible. The spires consist of triangular plates arranged in the test chamber entrance which, combined with the surface roughness, are used to generate the windspeed profile of the BL. The dimensions of the spires depend on the desired BL and the dimensions of the tunnel. It is recommended that the distance between two spires should be half of the height of one spire, and their heights should be smaller than the height of the wind tunnel (Pires, 2009). It is assumed, according with Blessman (1973), that the BL real δ is around 280 m. The tunnel test chamber used has a height of 460 mm and the reduction of scale with respect to the actual size was 1:1000. Therefore it should produce a ABL as close as possible to 280 mm, which was also checked by numerical simulations

Thin screen

Medium screen

(Pires, 2009). Details regarding the calculation of the quantity and characteristics of spires that need to be used for different wind tunnel dimensions can be found in Blessmann (1973) and Pires (2009). For this work, three spires, each with heights of 307.7 mm and base width of 32.6 mm, were used. The spires (Irwin, 1981) were produced using steel plates, with each spire consisting of two plates cut and folded together by spot welding (Fig. 3). Hot-wire measurements were performed with a singlewire BL probe (“Dantec 55P15”) of the constant-temperature hot-wire anemometer (“DANTEC Streamline bridge”). The calibration of the probe was performed using a DANTEC calibration unit. The equipment enables a precise adjustment of flow velocity at the exit of a nozzle. The dynamic pressure of the jet was measured with a standard Betz type manometer, and the respective velocities were correlated with

Coarse screen

Figure 2. Screens used for the formation of atmospheric boundary layer.

Figure 3. Devices used: carpet, screen, and spires. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

308

Pires, L.B.M., Paula, I.B., Fisch, G., Gielow, R. and Girardi, R.M.

the hot-wire signal by the King’s law. The procedure resulted in calibration errors below 0.5%, which are typical of hot-wire applications. During the measurements, the hot-wire signal was acquired with an AD card (NI-PCI-MIO-16) with a resolution of 16 bits. To avoid clipping effects, the gain applied to analog signal was fixed; this kept the maximum voltage always below 80% of the range of the AD card. The DC and AC parts of the signal were not split prior to the acquisition; therefore no high-pass filtering was applied. Only low-pass filters were used to avoid aliasing effects. The positioning of the probe was done with a DANTEC traverse mechanism, which provides a resolution of 6.25 μm in all three axes.

y (mm) 20 mm

One important issue of BL measurements is the definition of the surface location. In the current experiments, the initial position was set by moving slowly the probe toward the wall until it visually touched the surface. Afterward, a correction based on the extrapolation of the mean flow data was applied. The procedure resulted in a nearly constant shift of approximately 1 mm. The experimental apparatus is shown in Figs. 4 and 5. It uses an x (longitudinal), y (lateral), and z (vertical) coordinate system to describe the positions of measurements. Hot-wire anemometers of constant temperature were used to measure the flow. These anemometers utilize a several m illimeter long tungsten

mesh

screen

410 mm x (mm)

spires

carpet

300 mm

900 mm

Figure 4. View of the arrangement of the experimental apparatus with the x (longitudinal) and y (lateral) coordinates.

y (mm)

-70 -50 -30 -10 1120 10 30 50 70

1220

Figure 5. View of the mesh where measurements were taken. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

1320

1420

1520

1620 x (mm)

Simulations of the Atmospheric Boundary Layer in a Wind Tunnel with Short Test Section

wire with a diameter of around 4 μm. The system functions as one of the resistors in a Wheatstone bridge, allowing measurements with high spatial and temporal resolutions. Further details about this technique can be seen in Roballo (2007). For data acquisition, we used the Labview computer program. The positions on the longitudinal and lateral axes are shown in Fig. 5. The vertical profile was performed at the heights 1, 3, 6, 10, 20, 30, 40, 50, 60, 80, 100, 130, 160, 190, 220, and 260 mm, using a scale factor of 1:1000. The experimental uncertainty in the measurement of pressure and temperature is 0.1 and 0.4% respectively. This result is an uncertainty of 3.1% in the estimation of the Reynolds number, with the confidence interval of 95%. More details can be found in Roballo (2007). Initially, the spires and coarse, medium, and fine screen devices were tested separately. Afterward, combinations of these devices were tested. Then, an optimization of the positions of spires and screen was done with distances of 150 and 300 mm from the screen. Finally, the carpet was inserted to perform the fine tuning of the desired roughness. The metric used to optimize the simulation of rural land was by the coefficient α of the wind power law, which was assumed equal to 0.15.

RESULTS AND DISCUSSION Table 1 presents the values of the coefficient α of the power law versus the Reynolds number (Re) for the four devices (spires and fine, medium, and coarse screens) individually deployed, corresponding to the positions x=1120 mm and x=1620 mm grid points. Re was calculated for an ABL height of 280 mm as

309

suggested by Pires et al., (2009), and the windspeed used was u∞. The largest pressure loss occurred for the fine screen, with a dynamic pressure of about 28.8 mm H2O and a Re of 3.81×105. The pressure losses measured by the Pitot tube were smaller for the coarse screen (v=27.6 m/s). The spires presented the second highest pressure loss with Re equal to 4.16×105, which is not very relevant, since it will be coupled to another device for the generation of the ABL. The ideal configuration causes the smallest pressure loss and generates the ABL in the shortest possible distance of the test section. Note that the coefficient α is independent of Re or the longitudinal position (x) and is always smaller than 0.1 (ranging from 0.05 up to 0.09). This small value of α is due to absence of other passive devices to fine tune the generation of ABL. The ABL was fully developed at x=1120 mm, showing nonsignificant differences with the combination of spires and thin screen, due to different gaps between the spires and the screen (150 mm or 300 mm). The combination of the spires with the medium screen moved the ABL to the position x=1420 mm, with the highest mean square deviation for spires 300 mm distant from the screen reaching 3.0 m/s near the surface and 0.5 m/s at the height of 260 mm. For the combination of the spires with coarse screen, the ABL was not developed inside the test chamber section, presenting an irregular windspeed profile, requiring a longer test section for its formation. Table 2 summarizes the values of α and Re for spires combined with the fine and medium screens. The higher values of Re, of about 4.8×105, were obtained for the medium screen placed further away (300 mm) from the spires. Spires with the thin screen display lower values of Re (3.9×105) for both small and greater separations. The largest coefficient α occurs for spires close (150 mm) to medium screens, with values around 0.10.

Table 1. Re and α for isolated spires or screens.

Devices

Spires Thin screen Medium screen Coarse screen

Positions (mm)

x=1220 x=1620 x=1220 x=1620 x=1220 x=1620 x=1220 x=1620

Windspeed (m/s) Dynamic pressure (mm H2O)

24 24 22 22 27 27 27.6 27.6

32.8 32.8 28.8 28.8 41 41 44 44

0.06 0.09 0.02 0.07 0.07 0.04 0.06 0.05

4.16×105 4.16×105 3.81×105 3.81×105 4.68×105 4.68×105 4.78×105 4.78×105

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

310

Pires, L.B.M., Paula, I.B., Fisch, G., Gielow, R. and Girardi, R.M.

Table 2. Values of Re and α for spires combined with fine and medium screens. Devices

Spires+thin screen Spires+medium screen Spires close+thin screen Spires close+medium screen

Positions (mm)

Windspeed (m/s)

x=1120 x=1220 x=1320 x=1420 x=1520 x=1120 x=1220 x=1320 x=1420 x=1520

22.5 22.5 28 28 28 23 23 24 24 24

0.03 0.03 0.06 0.05 0.04 0.08 0.07 0.11 0.11 0.10

3.9×105 3.9×105 4.8×105 4.8×105 4.8×105 3.9×105 3.9×105 4.2×105 4.2×105 4.2×105

Gray: position where the ABL is formed.

Measurements were carried out at several locations along the stream and spanwise directions, as shown schematically in Fig. 5. The variations of mean and turbulent flow profiles along the spanwise direction were used as a criterion to ensure that the

ABL was developed. Figure 6 shows an example of developed BLs, measured with medium screens. No significant variation of the profiles is present in the results, suggesting the presence of a welldeveloped flow at position x=1420.

300

300 mm

0.9 0.8

Z (mm)

(Z/δ)

0.7 0.6 0.5 0.4

200 150 100

0.3 0.2

0.1 0 0.65

0.7

0.75

0.8 0.85 U(z)/U(δ)

0.9

0.95

0 0

0.04

0.06 ITU/U(δ)

0.08

0.1

0.12

0.02

0.04

0.06 ITU/U(δ)

0.08

0.1

0.12

250

0.8

Z (mm)

0.7

(Z/δ)

0.02

300

150 mm

0.9

200

0.6

150

0.5 0.4

100

0.3 0.2

0.1 0 0.65

y=0 y = 10 y = 30 y = 50 y = 70

250

0.7

0.75

0.8 0.85 U(z)/U(δ)

0.9

0.95

Figure 6. Nondimensional windspeed and turbulence profile at x=1420 mm. Cases with medium screen (see Table 2). The turbulence level is normalized with respect to the velocity at top of the boundary layer. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

Simulations of the Atmospheric Boundary Layer in a Wind Tunnel with Short Test Section

Figure 7 shows the windspeed profile for a medium screen positioned at 150 mm (named as near) and 300 mm (far) from the spires, with and without the carpet for fine adjustment. Note that the drag is higher with the carpet, and the ABL is better formed with the carpet and a medium screen positioned at a 300 mm distance from the spires. Also,

300

Height (mm)

250

for spires further away from the screen (Fig. 7a), the velocities are greater for low heights, 15 and 25 m/s, respectively. Figure 8 shows that when the carpet is used in conjunction with screens and spires, the windspeed profiles become more convex, with the greatest convexity noted for the smallest separation between the spires and screens (150 mm).

Medium screen distant 300 mm

Medium screen distant 150 mm 300

with carpet without carpet

250

200

150

100

0 0

311

0 0

Windspeed (m/s)

Figure 7. Windspeed profile at x=1420 mm.

300 250

300

Spires 300 mm from screen Spires 150 mm from screen

250

Height (mm)

200 150 100

150 100 50

50 0 15

200

20 25 Windspeed (m/s)

0 0

0.02 0.04 0.06 0.08 0.1 ITU/U(z)

0.12 0.14 0.16 0.18

Figure 8. Windspeed profile and turbulent intensity at the point x=1420 mm with medium screen and carpet. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

312

Pires, L.B.M., Paula, I.B., Fisch, G., Gielow, R. and Girardi, R.M.

Table 3. Values of Re and α for tests with the carpet. Devices

Positions (mm) Windspeed (m/s)

Without carpet Without carpet, with spires close to the screen With carpet With carpet, with spires close to the screen

x=1420 x=1420 x=1420 x=1420

Thus, the height of the ABL depends on the distance separating the devices. Adjusting the data to the theoretical power law profile, with α=0.15, with the spires 300 mm away from the screen, an ABL around 200 mm was obtained, while for the spires closer to the screen (distance of 150 mm), the result was an ABL of 180 mm. The intensity of turbulence showed nonsignificant differences between the two cases. Table 3 summarizes that Re is not sensitive to the inclusion (or not) of the carpet, varying from 4.2×105 to 4.9×105. However, the inclusion of the carpet resulted in values of α closer to the ideal 0.15. The height of the ABL increased from 180 to 200 mm as the distance between the spires and the medium screen increased (from 150 to 300 mm). Thus, the mean flow characteristics were apparently met with the combination of spires, medium screen, and carpet. In order to evaluate the results with respect to real ABLs, the mean flow profile was compared to theoretical power law and log-law profiles (Fig. 9). This last one is given by the relation U(z)/u*=1/0.4 log[(z–zd)/z0] (Blessman, 1973), where u* is the friction velocity, z0 is the roughness factor, and zd is the zero-plane displacement for very rough surface. Here zd is

28 24 25.5 25.5

Alpha

Height of ABL (mm)

0.05 0.11 0.14 0.12

4.9×105 4.2×105 4.4×105 4.4×105

– – 200 180

assumed to be zero due to the flat terrain conditions and u* and z0 were obtained using the Clauser method. The results of Fig. 9 show a good agreement of the experimental data with the theory. This confirms a good modeling of the mean flow from the ABL at the wind tunnel. The turbulence intensity is also an important feature of the ABL; therefore its intensities measured in the current experiments were compared to a benchmark suggested by ESDU (Engineering Science Data Unit). Such a benchmark consist of an empirical correlation based on a large experimental database of ABLs measured at height up to 100 m. The expression for this correlation can be found in ESDU (1985) and Liu et al., (2003). Therefore, the current data had to be scaled to the real ABL in order to allow a proper comparison. The results in Fig. 10 show a good agreement between the correlation and the experiments, especially for the streamwise stations where the BL was fully developed (x=1420 and 1520). At these stations, all experimental results lie between the limit of ±15%, represented in the figure

100

(x=1320) (x=1420) (x=1520)

90 80 70

Power law - α=0.15 Log law Experiments

Z (m)

60 50 40

U+

30 20 10

0 0 5 101

102

103 +

Figure 9. Windspeed profile in comparison to theoretical power law (α =0.15) and log law profiles.

0.1

0.2

0.3

0.4

0.5

Figure 10. Turbulent intensity at the first 100 m (equivalent to the boundary layer scaling) and comparison with the correlation suggested by ESDU 7430 and 7431 (Engineering Science Data Unit). The dotted lines represent the ±15% deviation from the correlation.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

Simulations of the Atmospheric Boundary Layer in a Wind Tunnel with Short Test Section

313

to point out that for other terrain conditions, especially those at

Dimensionless spectrum

100

urban regions, this might not be the case.

10-1

CONCLUSION The ideal situation is the formation of the ABL in the

10-2

10-3 10-2

shortest possible extension of the wind tunnel, with the lowest

von Kármán

pressure loss possible, to obtain a large Re, similar to the values

vonPresent Kármánexp (Z/δ=0.1) Present exp (Z/δ=0.1)

10-1 100 101 Dimensionless frequency

between 106 and 107 observed in nature. An ABL can be formed 102

by combinations of passive devices, which do not require a long test section. So, in some situations, existing aeronautical wind tunnels can be used for atmospheric simulations, in particular

Figure 11. Comparison of the dimensionless spectrum obtained at z/δ =0.1 and the von Kármán spectrum.

those with small values of the coefficient α in the power law. The lowest pressure loss was obtained with the coarse

by dotted lines. This suggests that the characteristics of the BL turbulence were also well modeled at the wind tunnel. In order to complete the assessment of wind tunnel data, the spectrum of flow fluctuations was compared with the vonKármán spectrum (ESDU, 1985), and the results are displayed in Fig. 11. According to ESDU, von-Kármán spectrum is a good model to represent the main features of the longitudinal component of turbulence in ABL. The formula of the dimensionless spectrum is fSu/σ2=4Xu(z)/[1+70,78Xu(z)2]5/6, where Su is the spectral density function, f the dimensional frequency, σ2 is the variance, and Xu is the dimensionless frequency. The last term is given by fL(z)/U(z), L(z) being the integral scale calculated using the autocorrelation spectra. A qualitatively good agreement is found between the experiments and the von Kármán spectrum. The region of constant slope that is equivalent to the -5/3 exponential decay in the inertial subrange of Kolmogorov’s cascade is nearly the same in both cases. Unfortunately, a more quantitative comparison could not be made with the current data because the period set for data acquisition was not long enough to enable the use of averaging procedures. Nevertheless, it is clearly observable that the main features of both

screen alone, while spires followed by a thin screen resulted in the highest pressure loss. The medium screen generated a well-developed ABL at the mesh point x=1420 mm, while the use of thin screen moved the ABL to the position x=1120 mm. In view of these findings, if the length of the wind tunnel is critical (being short) and the pressure loss is not so important; the best choice would be the thin screen. The distance between spires and screen was the other factor examined. A lower pressure loss, with a larger Re equal to 4.8×105, was observed with the medium screen more distant from the spires, while for the thin screen Re was equal to 3.9×105. However, higher values of α (equal to or greater than 0.10) were found for all cases of spires closer to the screens. The insertion of a carpet caused a sharp increase of α, which became closer to the desired value of 0.15. The ratio of pressure loss between the distances of the spires and the screen remained the same, but demonstrated a smaller drop with the increase of the distance between the devices; however, for the same screen, the height of the ABL was increased with a smaller distance between the screens and spires, resulting in 180 mm for the distance of 150 mm, and 200 mm for 300 mm. These findings indicate that this technique is a suitable

spectra are the same. Therefore, it is reasonable to assume that

tool for generation of an ABL in short-chamber wind

mean flow and the turbulence characteristics of the ABL, expected

tunnel experiments. The main characteristics of wind

to occur at the coast close to ALC, could be well simulated in a

in an ABL seem to be well reproduced in the current

short wind tunnel with the use of passive devices. It is important

experiments. This is confirmed by comparisons with

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

314

Pires, L.B.M., Paula, I.B., Fisch, G., Gielow, R. and Girardi, R.M.

established benchmarks from the literature. However, an important caveat is that the technique employed here

ACKNOWLEDGMENTS

seems to be limited to the reproduction of ABLs of nearly flat terrains, which have very small values of power law

Authors acknowledge a doctoral fellowship (CNPq

exponent. For regions with higher terrain roughness, such

141861/2006-1) and a Research Scholarship 302117/2004-0, both

as forests and urban areas, the ABL wind tunnels clearly

granted by CNPq (Conselho Nacional de Desenvolvimento

are the most suitable tools.

Científico e Tecnológico, Brazil).

REFERENCES Avelar, A.C., Brasileiro, F., Marto, A.G., Marciotto, E. and Fisch, G., 2012, “Wind Tunnel Simulation of the Atmospheric Boundary Layer for Study of the Wind Pattern at the Alcantara Space Center”, Journal of Aerospace and Technology Management, Vol. 4, No. 4, doi: 10.5028/jatm.2012.04044912.

Kwon, K.J., Lee, J.Y. and Sung, B., 2003, “PIV Measurements on the Boundary Layer Flow around Naro Space Center”, Proceedings of the 5th International Symposium on Particle Image Velocimetry, pp. 22-24.

Blessmann, J., 1973, “Simulation of the Natural Wind Structure in an Aerodynamic Wind Tunnel” (in Portuguese), Ph.D. thesis, Instituto Tecnológico de Aeronáutica (ITA), São José dos Campos, S.P., Brazil, 169 p.

Liu, G., Xuan, J. and Park, S., 2003, “A new method to calculate wind profile parameters of the wind tunnel boundary layer”, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 91, pp. 1155-1162.

Cao, S. and Tamura, T., 2006, “Experimental Study on Roughness Effects on Turbulent Boundary Layer Flow over a Two-Dimensional Steep Hill”, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 94, pp. 1-19. doi: 10.1016/j.jweia.2005.10.001.

Mavroidis, I., Griffiths, R.F. and Hall, D.J.H., 2003, “Field and Wind Tunnel Investigations of Plume Dispersion around Single Surface Obstacles”. Atmospheric Environment, Vol. 37, pp. 2903-2918, doi: 10.1016/S1352-2310(03)00300-5.

Carpentieri, M., Hayden, P. and Robins, A.G., 2012, “Wind Tunnel Measurements of Pollutant Turbulent Fluxes in Urban Intersections”, Atmospheric Environment, Vol. 46, pp. 669-674. doi: 10.1016/j. atmosenv.2011.09.083.

Novak, M.D., Warland, J.S., Orchansky, A.L., Ketler, R. and Green, S., 2000, “Wind Tunnel and Field Measurements of Turbulent Flow in Forests. Part I: Uniformly Thinned Stands”, Boundary Layer Meteorology, Vol. 95, pp. 457-495, doi: 10.1023/A:1002693625637.

Davenport, A.G., 1967, “The Dependence of Wind Loads on Meteorological Parameters”, Proceedings of the International Seminar on Wind Effects on Buildings and Structures, Ottawa, pp. 19-82. ESDU, 1985, “Characteristics of atmospheric turbulence near the ground. Part II: Single point data for strong winds (neutral atmosphere)”, ESDU International, Item No. 85020, London.

Pires, L.B.M., 2009, “Study of the Internal Boundary Layer Developed on Coastal Cliffs with Application to the Alcântara Launching Center, Brazil” (in Portuguese), Ph.D. thesis, Instituto Nacional de Pesquisas Espaciais (INPE), São José dos Campos, SP, Brazil. 165p, (INPE-16566-TDI/1562), http://pct.capes.gov.br/teses /2009/33010013003P8/TES.PDF.

Franck, N., 1963, “Model law and Experimental Technique for Determination of Wind Loads on Buildings”, in Symposium on Wind Effects on Buildings and Structures, Vol. 16, National Physical Laboratory, Teddington, UK, pp. 181-196.

Pires, L.B.M., Souza, L.F., Fisch, G. and Gielow, R., 2009, “La Influencia de la Altura de la Capa Límite Oceánica en la Región del Centro de Lanzamientos de Alcántara en Brasil”, Información Tecnológica, Vol. 20, pp. 119-128, doi: 10.1612/inf.tecnol.4071it.08.

Irwin, H.P.A.H., 1981, “The Design of Spires for Wind Simulation”, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 7, pp. 361–366, doi: 10.1016/0167-6105(81)90058-1.

Pires, L.B.M., Roballo, S.T., Fisch, G., Avelar, R.M. and Gielow, R., 2010, “Atmospheric Flow Using the PIV and HWA Techniques”, Journal Technology and Management, v.2, pp.127-136, jatm.2010.02027410.

Jensen, M. and Franck, N., 1963, “Model Scale Tests in Turbulent Wind”, Part I, Danish Technical Press, Copenhagen. Jensen, M. and Franck, N., 1965, “Model Scale Tests in Turbulent Wind”, Part II, Danish Technical Press, Copenhagen. Kozmar, H., 2009, “Scale Effects in Wind Tunnel Modeling of an Urban Atmospheric Boundary Layer”, Theoretical and Applied Climatology, Vol. 100, pp. 153–162, doi: 10.1007/s00704-009-0156-3.

A.C., Girardi, Measurements of Aerospace doi:10.5028/

Roballo, S.T., 2007, “Estudo do Escoamento Atmosférico na CLA, através de Medidas em Torre Anemométrica e em Túnel de Vento” (in Portuguese), Master thesis, Instituto Nacional de Pesquisas Espaciais (INPE), São José dos Campos, SP, Brazil, 137 p, (INPE14824-TDI/1264).http://mtc-m17.sid.inpe.br/col/sid.inpe.br/ mtc-m17@80/2007/06.12.18.07/doc/ publicacao.pdf.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.305-314, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.222

Verification of Response of Neutron Monitor for In-Flight Neutron Dosimetry Claudio Antonio Federico1, Odair Lélis Gonçalez 1, Evaldo Simões da Fonseca2, Karla Cristina de Souza Patrão2, Marlon Antonio Pereira1, Linda Viola Ehlin Caldas3

ABSTRACT: In this work, we present the results of verification of special neutron monitoring equipment for use in aircrew dosimetry inside aircraft. The equipment was precalibrated with conventional neutron sources in a National Standard Laboratory of the Institute for Radioprotection and Dosimetry (IRD, Brazil) and verified in a Cosmic Energy Reference Field (CERF), a special field from CERN (Centre Européan de Recherche Nucléaire), a facility that reproduces the shape of neutron field encountered in the atmosphere at aircraft altitudes. The equipment consists of a special neutron probe from Thermo Scientific with response up to 5 GeV neutrons and was verified with respect to ambient dose equivalent rate linearity and angular incidence. The results show the adequacy of the equipment for this type of measurement and the feasibility to use conventional neutron sources to calibrate this specific equipment, in the absence of access to the CERF field. Keywords: Radiation dose, Aircrew, Neutron dose.

INTRODUCTION In recent decades, the problem of controlling the level of ionizing radiation dose received by aircrew and sensitive equipment has received great attention and motivated several studies in specialized international literature (FAA, 1990; Wilson et al., 1998; Bartlett, 2004; Edwards et al., 2004; Hajek et al., 2004). In Brazil, these studies were initiated by the Brazilian Air Force (FAB) in conjunction with research institutions from the National Commission on Nuclear Energy (CNEN) (Federico et al., 2010a, 2010b, 2012). The reason for this concern is the fact that the dose rate arising from cosmic radiation (CR) undergoes a considerable increase with altitude, and consequently, aircrew frequently exceed the annual effective dose limit proposed by international organizations for planned exposures to members of the public (ICRP, 1991), which is 1 mSv. The International Commission on Radiological Protection (ICRP) recognizes the need to control exposure for the group of professional flight crew and pilots (ICRP, 1998) because this group is exposed to radiation levels that are comparable with or greater than the average levels of radiation received by professionals working with radiation in medicine and technology. The measurement quantity recommended for aircraft is the ambient dose equivalent (H*(10)), and its employment for purposes of dosimetric control is made by conversion factors for effective dose recommended in the literature, so that the results obtained through the calibrated equipment

1.Instituto de Esudos Avançados – São José dos Campos/SP – Brazil 2.Instituto de Radioproteção e Dosimetria – Rio de Janeiro/RJ – Brazil 3.Instituto de Pesquisas Energéticas e Nucleares – São Paulo/SP – Brazil Author for correspondence: Claudio Antonio Federico | Instituto de Estudos Avançados | Trevo Coronel Aviador José Alberto Albano do Amarante,1 | CEP 12.228-001 São José dos Campos/SP – Brazil | Email: claudiofederico@ieav.cta.br Received: 17/12/12 | Accepted: 22/05/13

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

Federico, C.A., Gonçalez, O.L., Fonseca, E.S., Patrão, K., Pereira, M.A. and Caldas, L.V.E.

can be compared to the limits applicable to the crew of the aircraft. Due to the large contribution of the neutron component in ambient dose equivalent incident in aircrews and in singleevent effects in aircraft avionics, it is necessary to calibrate the response of the instrumentation used in relation to the field of neutrons present at flight altitudes. This calibration is not a conventional procedure due to higher energy of the field and the lack of facilities with this capability. The neutron radiation field produced by CR covers energy up to the order of hundreds of MeV, presenting more pronounced peaks in the region of thermal neutrons (about 0.023 eV), neutron evaporation (about 1 MeV), and neutrons from spallation processes in atmospheric constituents (peak around 100 MeV). In special, the thermal peak is not very significant in the atmosphere, due to neutron absorption on nitrogen, but can be greatly enhanced by the presence of thermalizing materials within aircraft, such as, the fuel. The characteristics of this type of field are very different from those obtained in neutronic fields emitted by conventional sources of radionuclides, accelerators, or reactors, forcing the use of specific and characterized fields, with broad spectra, created specifically to simulate those obtained at the sites of measurement (Schuhmacher, 2004), for verification or calibration of measuring instruments. There is a special arrangement in the laboratories of CERN (Centre Européan de Recherche Nucléaire) called CERF (CERN-EU high energy Reference Field), Prevessin, France, designed to reproduce, with reasonable proximity, the existing field on aircraft flight altitudes. While this field cannot be treated as a metrological standard, it is recognized as amongst “the best efforts” in the world to reproduce this kind of neutron field and is widely used to verify the instrumentation used for aeronautical dosimetry purposes.

acquisition. The internal proportional counter gives the results from low LET components of the field, and the system was designed and calibrated to provide results directly of the operational quantity ambient dose equivalent (H*(10)) and can be operated directly or programmed to acquire measurements of dose rate at time intervals set by the user, which can vary from 1 second to several hours. The system has a volatile internal memory that lets you store the results of up to 256 measuring points, which can be collected by the user later. Each measurement point consists of a record, with information covering the start and finish times of each measurement interval, the result of the proportional internal probe measurement unit, the measurement result of the optional external probe, the measurement unit, the measurement interval (in seconds), and the type of external coupled probe. For the measurements reported here, the THERMO FHT-762 (WENDI-II) was coupled to the external probe system, which consists of a 3He proportional type neutron probe, covered with a layer of tungsten and an external polyethylene layer, which allows obtaining the response to neutrons with energies up to about 5 GeV, and which differs greatly from conventional neutron detectors that typically have an adequate response to energies up to at most about 20 MeV. The efficiency curve for ambient dose equivalent as a function of neutron energy is shown in Fig. 1, compared

Relative response per unit of ambient dose equivalent (normalized for the 252Cf neutron spectrum)

316

WENDI-II (Side) Eberline Hankins-NRD Anderssn-Braun (Side) 1

0.1

EXPERIMENTAL PROCEDURE THE DETECTION SYSTEM The FH THERMO detection system consists of a Thermo Electron Eberline acquisition, model FH40G-10, with remote programming capability through connection to a computer, coupled to an internal proportional counter, and allows the connection of an external probe with simultaneous

0.1

100

1000

Energy (MeV)

Figure 1. Response curve depending on the incident neutron energy to the probe FHT-762 (WENDI-II), compared with that obtained for other conventional monitor type Eberline Hankins-NRD and Andersson-Braun (extracted from Thermo-Scientific, 2009).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

Verification of Response of Neutron Monitor for In-Flight Neutron Dosimetry

317

Table 1. Main characteristics of FHT-762 probe (ThermoScientific, 2009). Parameters

Value

Measurement (0.001 to 1 × 106) µSv/h interval Sensitiveness

0.84 cps/(µSv/h)

Linearity Energy interval Angular dependence Pressure interval

-9% to +11% 25 meV to 5 GeV

Table 2. Source of uncertainties at the Neutron Laboratory setup of the Ionizing Radiation Metrology National Laboratory.

Observations Factor

Uncertainty (%) Am/Be

241

Referenced to neutrons of 252Cf Response curve according to ICRP 74 (1997)

±20% 500 to 1500 hPa

Emission rate Converting coefficient Time Scattering Positioning

1.1 4 1 1 0.6

252

2 1 1 1 0.6

with the curve obtained in other conventional equipment (Thermo-Scientific, 2009). The main features of neutron measurement system are presented in Table 1 and the same is calibrated in the field relative to a neutron source from 241Am/Be.

from the value of the emission rate of the neutron source, determined by means of a bath of manganese sulfate. The bath of manganese sulfate, which is the primary standard for determining the rate of emission of neutron sources, is traceable to International Metrological System through the “K9” key comparison (Roberts et al., 2011), coordinated by the Bureau International des Poids et Mesures (BIPM).

CONVENTIONAL SOURCES Initially the neutron probe was recalibrated at the Neutron Laboratory of the Ionizing Radiation Metrology National Laboratory, Brazil, using standard sources of 241Am/Be and 252 Cf. The traditional calibration protocol of the laboratory was modified to suit the peculiarity in the use of equipment that operates with low dose rates, present in aircraft flight altitudes. The equipment was exposed to different distances from the standard sources, ranging from about 0.86 m up to 4.7 m, where the ambient dose equivalent of the neutron field is known with a combined average uncertainty varying from around 9 to 2% for sources of 252Cf and 241Am/Be respectively. Such uncertainty is a combination average of the sources of uncertainty given in Table 2. In the calibration process, we used a shadow cone constructed in accordance with the Safety Report Series 16 (IAEA, 2000), and for each measurement point (distance from source) calibration was performed in two stages, with and without using the shadow cone, so as to eliminate the contribution of the components originating from neutrons scattered in the environment, according to the methodology used in Federico et al., (2010a). The quantity used for calibration was the ambient dose equivalent, whose conventional true value was obtained

The Centre Européan de Recherche Nucléaire high energy Reference Field The CERF field is produced by a hadron beam, comprising 34.8% protons, 60.7% pions, and 4.5% kaons, accelerated to an energy of 120 GeV and incident on a copper target (Mitaroff and Silari, 2002). The acceleration of the beam until the energy extraction occurs on an acceleration ring of 6.8 km in circumference, having a 48s cycle between each pulse of particles to be accelerated. Due to the dispersion occurred in the acceleration process, the extracted beam has an average duration of 9s, with the pulse shape shown in Fig. 2 for average dose rates during the pulse varying between 11 and 1300 μSv/h, where the intensity of the pulse can be controlled by the user from the main control room, through a set of collimators that are adjustable remotely. Two neutron fields are available with different characteristics. The first one uses the neutrons generated on the target and passing through an iron platform according to the schematic drawing shown in Fig. 3. The distribution of neutron fluence as a function of the energy was evaluated by Mitaroff and Silari (2002) and, for the field in iron platform, the shape of spectra extends from thermal up to 600 MeV, with the predominance of the neutron energy in the range of 0.1–1 MeV.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

318

Federico, C.A., Gonçalez, O.L., Fonseca, E.S., Patrão, K., Pereira, M.A. and Caldas, L.V.E.

1400 11 µSv/h 65 µSv/h 260 µSv/h 490 µSv/h 1300 µSv/h

Count rate (cps)

1200 1000 800 600 400 200 0 0

8 10 12 Time (s)

The second field, of the most interest to this work, utilizes neutrons generated on the target and passing through a concrete platform, whose spectrum in the position above the concrete platform extends from thermal neutrons up to 500 MeV, with the presence of pronounced peaks characteristic of processes of evaporation and spallation in atomic nuclei (from 1 up to hundreds of MeV) and with great similarity, in shape of the spectrum, with the typical spectrum of neutrons in aircraft flight altitudes (Fig. 4). The detectors were irradiated at predetermined positions on the platforms of concrete and iron, where the particle flux and spectra can be calculated from the number of counts (PIC counts) obtained in one beam monitor chamber, placed at the exit of the hadrons extraction channel (Fig. 3).

Figure 2. Hadrons pulse shape for dose rates used.

RESULTS

Concrete platform Iron platform Beam line Hadrons 120 GeV Target positions

Beam monitor chamber

0.14

180

0.12

Conventional true value - H*(10) ( µ Sv/h)

Neutron flux (n/(cm2.s.lethargy))

Figure 3. Schematic drawing of the layout of irradiation, with the positions of the beam, target, and measurement sites.

The results of measurements with the source 241Am/Be and 252Cf are shown in Figs. 5 and 6 respectively, along with the lines resulting from a linear fit to experimental data, where the uncertainties in the measurements and in the conventional true value were considered as weight factor in the fitting process. The conventional true value is the best estimative of the true value of the quantity, given by the laboratory maintainer, based on metrological references, as explained previously. During the verification process of the response of the instrument in the CERF, it was observed that the

160

0.10

140

0.08

120

0.06

100

0.04 0.02 0.00

-7

-6

-5

-4

-3

-2

-1

10 10 10 10 10 10 10 10 10 10 10 10 10 Energy (MeV/n) Figure 4. Typical shape of the neutron flux spectrum at 35,000 ft altitude, made using EXPACS code (Sato and Niita, 2006).

80 60 40 20 0

40 60 80 100 120 140 160 180 Measured value - H*(10) ( µ Sv/h)

Figure 5. Measurements obtained with the 241Am/Be source, with straight line fitted to experimental data.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

Verification of Response of Neutron Monitor for In-Flight Neutron Dosimetry

background (BG) radiation in the place is variable and depends, among other factors, on the opening of collimators which are used to control the rate of the desired dose. Through the FH40G-10 detector and its neutron probe, we obtained the BG neutron field in some configurations and an empiric

100 80 60

Conventional true value - H*(10) (µ Sv/h)

40 20 0

40 60 80 Measured value - H*(10) (µSv/h)

100

Figure 6. Measurements obtained with the 252Cf source, with straight line fitted to experimental data.

1.4 Average background (microSv/h)

function was adjusted to express the BG to be subtracted from experimental measurements, as a function of the opening of collimators for each measurement, as can be seen in Fig. 7. The function type and the coefficients obtained are presented in Table 3. The verification of the response of the detector FH40G-10 with the neutron probe (FHT-762) was made by means of comparison with the conventional values of the ambient dose equivalent of the neutron field, for the CERF fields from concrete and iron platforms. The results are shown in Figs. 8 and 9, in which it can

250 200 150 100 50 0 0

1.3

50 100 150 200 Measured value - H*(10) (µSv/h)

1.2

250

1.1 1.0

Figure 8. Response curve as a function of dose rate for the probe FHT 762, in the field of concrete, with straight line fitted to experimental data.

0.9 0.8 0.7 0.6

350

0.5 0.4

100

200 300 400 Beam area (mm2)

500

600

700

Figure 7. Evaluation of the background component in the neutron field Centre European de Recherche Nucleaire high energy Reference Field. Table 3. Empirical function and coefficients from background adjust. Function type

Y=A1*exp(-x/A2) + A3

Conventional true value - H*(10) (µ Sv/h)

120

319

300 250 200 150 100 50 0 0

Parameters

Value

Uncertainty

0.62

0.22

0.587

0.017

100

150

200

250

300

350

400

Measured value - H*(10) (µSv/h)

Figure 9. Response curve as a function of the dose rate for the FHT 762 probe, in the field of iron, with straight line fitted to experimental data. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

Federico, C.A., Gonçalez, O.L., Fonseca, E.S., Patrão, K., Pereira, M.A. and Caldas, L.V.E.

also be observed the lines resulting from linear fittings to the experimental data, where the uncertainties in the measurements were considered as weighting factors in the fitting process. The conventional value of the ambient dose equivalent was previously obtained by Mitaroff and Silari (2002). Table 4 presents the coefficients obtained for each linear fitting performed by the least squares method, where one can observe that the value of the constant term for the four-neutron radiation fields used are compatible to zero, with a confidence interval of up to two standard deviations. It is observed in the same table an underestimation of 4% on the value indicated by the instrument for 241Am/Be neutron field and 16% for the 252Cf neutron field. This underestimation of the real value can be explained by the dependence of the response to the energy of the neutrons reported by the equipment manufacturer, considering that the average energy of the 241Am/Be neutrons is approximately 4.5 MeV and the 252 Cf ones approximately 1 MeV. We can suppose that, for a good calibration, the relation between the measured values and the reference values is linear, without any constant off-set. Indeed, for the CERF field in concrete platform, which is of particular interest in this work, the constant coefficient obtained in the fitting is compatible to zero, which indicate that there is no systematic deviation of the measured values in relation to the reference values of neutron field. Therefore, a new fitting assuming the constant coefficient as null was accomplished. The uncertainty σA obtained previously is now encompassed in the uncertainty of the calibration factor (angular coefficient). This new fitting obtains the value of 1.022±0.005 for the slope, with an adjusted R-square equal to 0.998, which indicates that the instrument readings should be corrected by +2.2% with a statistical uncertainty in coefficient of 0.5%.

This uncertainty, obtained from the linear fitting of the instrument readings on CERN standard field values, is of statistical nature because it is only associated with the instrument reading errors. But, Mitaroff and Silari (2002) reported an uncertainty of 2% in the value of the standard field used in this calibration that affects in the same way all calibration data. Therefore, we must consider this as a systematic error of statistical nature in the calibration procedure, which could be taken into account as a covariance. Considering that the calibration function is linear and that we considered only the variations of statistical nature in obtaining the calibration coefficient by the linear fitting, the uncertainty of the field must be quadratically added to the uncertainty of the calibration coefficient value in order to calculate the standard deviation of the calibration coefficient. Thus, the calibration coefficient for the THERMO FH detector system for the neutron component is 1.022±0.021. We also evaluated the dependence of the response with the orientation of the detector positioning on the concrete platform of the CERF field, using the positions of standing, lying, and upside down, where differences below 1% were observed for the equipment in question, as can be seen in Fig. 10. We may therefore disregard the influences of the positioning direction. Comparing the results obtained in this study with measurements made by Mayer et al., (2007) for this same type of equipment, it is observed that the author noticed a difference of -11.9% in the response of the detector to the

1.006 Side incidence relative response

320

Table 4. Results from the linear fitting to the experimental data.

Source

Constant coefficient

Angular coefficient

σA

σB

Am/Be

0.14

0.25

1.040

0.020

0.33

0.21

1.160

0.022

0.05 0.12

0.91 0.39

0.782 1.021

0.018 0.006

241

252

CERF (iron) CERF (concrete)

1.004 1.002 1.000 0.998 0.996 0.994 0.992 0.990 0.988 From up

From down Incidence direction

From side

Figure 10. Response curve as a function of preferential direction of incidence for FHT 762 probe.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

Verification of Response of Neutron Monitor for In-Flight Neutron Dosimetry

field in the CERF concrete platform regarding calibration performed for the field of 252Cf. In this study, the difference between the response of the CERF field in concrete platform in relation to the response obtained for the field of 252Cf is about 13.9%, with an uncertainty of about 2.2%, confirming the result obtained by Mayer et al., (2007) within one standard deviation. Similarly, Yasuda and Yajima (2010) made tests with the same type of monitor, calibrated for neutron from 241Am/Be sources and found differences in the H*(10) rate measured in relation to the H*(10) calculated rate of about 3%, consistent with the underestimation of 2.2%, with an uncertainty of about 2.1%, found in this study.

CONCLUSION The results in terms of rate of environment equivalent dose indicate that the detector responds appropriately to high-energy neutrons, which can be used for radiation measurements on aircraft, with a small correction in the value read from the instrument. By the results obtained in this work and by Yasuda and Yajima (2010), one can observe that the difference between the response of H*(10) of such equipment to the CR field and the response calibrated to

321

conventional sources of 241Am/Be is about -2 to -3%, and this type of source can be used for calibration of this instrument, in the case of inability to use the CERF field. This is a very important result due to the difficulties related on support for this type of calibration in CERF field, especially for people from outside of Europe. Unfortunately, Mayer et al., (2007) show that the same is not the case for other types of conventional neutron detectors whose responses are hardly suitable for energy greater than 16 MeV, being therefore, for the most part, unsuitable for H*(10) measurements of neutrons from CR. The characterization of this equipment to the field of CR at aircraft flight altitude is an important step in the consolidation of a Brazilian group with training, experience, and proper equipment to perform such measurements in flight, which does not currently exist in Brazil, nor possibly elsewhere in Latin America.

ACKNOWLEDGMENTS We thank FINEP and CNPQ for partial financial support; to CERN, for allowing the verification of instruments in the CERF field; and to the Brazilian Air Force, which supports this research in IEAv.

REFERENCES Bartlett, D.T., 2004, “Radiation protection aspects of the cosmic radiation exposure of aircraft crew”, Radiation Protection Dosimetry, Vol. 109, No. 4, pp. 349-355.

By Aircrew of DCTA’s Flight Test Special Group”, Journal of Aerospace Technology and Management, Vol. 2, pp. 137-144.

Edwards, R.E., Normand, E., Dyer, C.S., 2004, “Technical Standards for atmospheric radiation single event effects (SEE) on avionics electronics”, IEEE Radiation effects Data Workshop Record, IEEE 04TH8774, 2004.

Federico, C.A., Gonçalez, O.L., Sordi, G.M.A.A. and Caldas, L.V.E., 2012, “Effects of Cosmic Radiation in Aircrafts: A Discussion about Aircrew over South America”, Journal of Aerospace Technology and Management, São José dos Campos, Vol. 4, No. 2, pp. 219-225, Apr-Jun.

FAA, 1990, Federal Aviation Administration, U.S. Department of Transportation, Radiation Exposure of Air Carrier Crewmembers. FAA Advisory Circular n. 120-152.

Hajek, M., Berger, T. and Vana, N., 2004, “A TLD-based personal dosemeter system for aircrew monitoring”, Radiation Protection Dosimetry, Vol. 110, No. 1-4, pp. 337-341.

Federico, C.A., Gonçalez, O.L., Fonseca, E.S., Martin, I.M. and Caldas, L.V.E, 2010a, “Neutron Spectra Measurements in the South Atlantic Anomaly Region”, Radiation Measurements, Vol. 45, No. 10, pp.1526-1528.

IAEA, 2000, Calibration of Radiation Protection Monitoring Instruments, Safety Report Series 16, International Atomic Energy Agency, Vienna.

Federico, C.A., Pereira, H.H.C., Pereira, M.A., Gonçalez, O.L. and Caldas, L.V.E., 2010 b, “Estimates of Cosmic Radiation Dose Received

ICRP, 1991, International Commission on Radiological Protection. Recommendations of the international commission on radiological protection, Publication 60, Annals of the ICRP, Vol. 21, No. 1-3.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

322

Federico, C.A., Gonçalez, O.L., Fonseca, E.S., Patrão, K., Pereira, M.A. and Caldas, L.V.E.

ICRP, 1997, International Commission on Radiological Protection. Conversion coefficients for use in radiological protection against external radiation, Publication 74, Annals of the ICRP, Vol. 26, No. 3. ICRP, 1998, International Commission on Radiological Protection. General principles for the radiation protection of workers, Publication 75, Annals of the ICRP, Vol. 27, No. 1. Mayer, S., Forkel-Wirth, D., Fuerstner, M., Menzel, H.G., Mueller, M.J., Perrin, D., Theis, C. and Vincke, H., 2007, “Response of neutron detectors to high-energy mixed radiation fields”, Radiation Protection Dosimetry, Vol. 125, No. 1-4, pp. 289–292.

Measurements of Neutron Source Emission Rate (1999-2005): CCRI(III)-K9.AmBe”, Metrologia, 48, Tech. Suppl., 06018. Sato, T. and Niita, K., 2006, “Analytical Functions to Predict CosmicRay Neutron Spectra in the Atmosphere”, Radiation Research, Vol. 166, pp. 544-555. Schuhmacher, H., 2004, “Neutron Calibration Facilities”, Radiation Protection Dosimetry, Vol. 110, No. 1-4, pp. 33-42. Thermo-Scientific,2009, FHT 762-WENDI-2 Datasheet. (n.42540/85 TD-E).

Mitaroff, A. and Silari, M., 2002, “The CERN–EU High-Energy Reference Field (CERF) Facility for Dosimetry at Commercial Flight Altitudes and in Space”, Radiation Protection Dosimetry, Vol. 102, No. 1, pp. 7-22.

Wilson, J.W., Goldhagen, P., Maiden, D.L. and Tai, H., 1998, “High altitude radiations relevant to the high speed civil transport (HSCT)”. NASA Technical Documentation, NASA.

Roberts, N.J., Jones, L.N., Králik, M., Park, H., Choi, K.O., Pereira, W.W., da Fonseca, E.S., Cassette, P., Dewey, M.S., Moiseev, N.N. and Kharitonov, I.A., 2011, “International Key Comparison of

Yasuda, H. and Yajima, K., 2010, “Characterization of Radiation Instruments at the Summit of Mt. Fuji”, Radiation Measurements, Vol. 45, pp. 1600-1604.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.315-322, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.224

Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software Ronaldo Rodrigues Ferreira1, Rafael Baldiati Parizi1, Luigi Carro1, Álvaro Freitas Moreira1

ABSTRACT: This paper discusses how compiler optimizations influence software reliability when the optimized application is compiled with a technique to enable the software itself to detect and correct radiation-induced control-flow errors. Supported by a comprehensive fault-injection campaign using an established benchmark suite in the embedded systems domain, we show that the compiler is a non-negligible source of noise when hardening the software against radiationinduced soft errors. Keywords: Compilers, Radiation effects, Single event upsets, Software reliability, Software engineering.

INTRODUCTION Compiler optimizations are taken for granted in modern software development, enabling applications to execute more efficiently in the target hardware architecture. Modern architectures have complex inner structures designed to boost performance, and if the software developer were to be aware of all those inner details, performance optimization would jeopardize the development processes. Compiler optimizations are transparent to the developer, who picks the appropriate ones to the results s/he wants to achieve, or, as it is more common, allowing this task to be performed by the compiler itself by flagging if it should be more or less aggressive in terms of performance. Industry already offers microprocessors built with 22 nm transistors, with a prediction that by 2026, the size of the transistor will reach 5.9 nm (ITRS, 2012). This aggressive technology scaling creates a big challenge concerning the reliability of microprocessors using newest technologies. Smaller transistors are more likely to be disrupted by transient sources of errors caused by radiation, known as soft-errors (Borkar, 2005). Radiation particles originated from cosmic rays when strikes a circuit induces bit flips during software execution, and because transistors are becoming smaller in size, there is a higher probability that these transistors will be disrupted by a single radiation particle with smaller transistors requiring a smaller amount of charge to disrupt their stored logical value. The newest technologies are so

1.Universidade Federal do Rio Grande do Sul – Porto Alegre/RS – Brazil Author for correspondence: Ronaldo Rodrigues Ferreira | Instituto de Informática, Universidade Federal do Rio Grande do Sul | Avenida Bento Gonçalves, 9500, Campus do Vale – Bloco IV – Agronomia | CEP 91.509-900 Porto Alegre/RS – Brazil | Email: rrferreira@inf.ufrgs.br Received: 08/01/13 | Accepted: 02/05/13

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

324

Ferreira, R.R., Parizi, R.B., Carro, L. and Moreira, Á.F.

sensitive to radiation that their usage will be compromised at the sea level, as predicted in the literature (Normand, 1996). Rech et al., (2012) have shown that modern graphics processing unit (GPU) cards are susceptible to such an error rate that makes their usage unfeasible in critical embedded systems. However, industry is already investing in GPU architectures as the platform of choice for high performance and low power embedded computing, such as the ARM Mali® embedded GPU (ARM, n.d.). The classical solution to harden systems against radiation is the use of spatial redundancy, i.e., the replication of hardware modules. However, spatial redundancy is prohibitive for embedded systems, which usually cannot afford extra costs of hardware area and power. The increase on power is a severe problem, because it is expected that 21% of the entire chip area must be turned off during its operation to meet the available power budget, and an impressive chip area of 50% at 8 nm (Esmaeizadeh, 2011). This creates the dark silicon problem (Esmaeizadeh, 2011), i.e., a huge area of the circuit cannot be used during its lifecycle. This problem gets worse when the microprocessor has redundant units, because system’s reliability could be compromised if redundant units were turned off. The current solution to this problem is to use radiation-hardened microprocessors, which are designed to endure radiation. The problem with this approach is the low availability, high unit pricing, and International Traffic in Arms Regulations (ITAR) restrictions of those radiation-hardened components. For instance, a 25 MHz microprocessor has a unitary price of $ 200,000.00 (Mehlitz and Penix, 2005). This high unit pricing makes the use of radiation-hardened microprocessors unfeasible for embedded systems used in aircrafts, not to say about cars and low-end medical devices, such as pacemakers. For these critical embedded systems, where cost and ITAR restrictions are hard constraints, a cheaper, but yet effective approach for reliability against radiation, is necessary. Software-Implemented Hardware F ault-Tolerance ( SIHFT) (Goloubeva, 2006) is an approach for radiation reliability that adds redundancy in terms of extra instructions or data to the application, keeping the hardware unchanged. SIHFT techniques work by modifying the original program by adding checking mechanisms to it. SIHFT techniques are classified either as control-flow or as data-flow. The former is

designed to detect when an illegal jump has occurred during application execution to possibly proceed with the resolution of the correct jump address or at least signaling that such an error has occurred. The latter checks if a data variable being read is correct or not. While the effects of data-flow SIHFT m ethods are clear (usually, the duplication of program variables or the addition of variable checksums solves the problem), the impacts of the control-flow ones, are yet not well understood. Because the control-flow methods modify the program’s control-flow graph (CFG), which happens to be the same artifact used by compiler optimizations, the efficiency of control-flow reliability techniques might be influenced by the optimizations in an unpredictable way. In this paper, we evaluate how the cumulative usage of compiler optimizations influence reliability of applications hardened with the state-of-the-art Automatic Correction of Control-flow Errors (ACCE) (Vemu, Gurumurthy and Abraham, 2007) control-flow SIHFT technique, which is selected, because it is the current most efficient method in terms of reliability, attaining an error correction rate of ~70%. The application set we use in this paper is drawn from the MiBench suite (Guthaus et al., 2001). RADIATION EFFECTS ON SOFTWARE R ELIABILITY Highly energized radiation particles are known hazard sources in electronics since the 1970s (Binder et al., 1975), as well as the mitigation schemes for such sources. Single-Event Transient (SET) is the observed physical effect of radiation on electronics, corresponding to voltage glitches in circuitry, which by itself does not incur on system hazards. System hazards originate when SETs are caught up by memories (e.g., SRAM’s) and sequential logic (e.g., registers), thus, becoming a Single-Event Upset (SEU). An SEU is a non-permanent damage to the systems (i.e., transient) that results in a bit with logical value 1 that flips to 0 and vice-versa. Mitigation approaches might be as follows: • Substrate and gate-level: The reduction of charge

generation and collection. • Hardware design: The modification of circuit response

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

through the addition of logical elements or even the storage of data on spatially separated nodes.

Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software

• System level: The addition of redundancy at system level,

e.g., software hardening. Definitely, the system level mitigation approach is the most feasible for components off-the-shelf and to overcome ITAR restrictions. A bit flip caused by a radiation-induced SEU can compromise the software in two different ways. Firstly, an SEU can corrupt data, i.e., the values of program variables. Secondly, an SEU can corrupt control, i.e., the program flow of execution. To illustrate these two situations, consider the program presented in Fig. 1, which corresponds to the Bubble Sort algorithm. Bubble Sort is a naïve O(n2) solution for sorting an array of arbitrary numbers. The Bubble Sort algorithm is divided in labeled regions named basic blocks (identified by the gray and white regions in the Bubble Sort source code). A basic block is a region of a program where the contained program instructions does not contain any branch, i.e., iteration loops (e.g., for and while commands), if-conditionals, function call, and return. Therefore, a basic block only contains variable assignments and logical evaluations. The CFG of a program P is a graph GP=(V, E), where the set V of vertices contains the program’s basic blocks and the set E of edges contains the transitions in the execution flow. To illustrate this, the CFG of the Bubble Sort algorithm is presented in Fig. 2. An executed branch of the program P (represented by an arrow in Fig. 2) is said to be legal, if and only if, it is an element

Algorithm Bubblesort(input: n, V) def n : number of values to sort; def V[n] : array of size n; def temp, i, j: integer variables; 1. i := n - 1; 2. while ( i >= 1 ) do 3. j := 0; 4. while ( j < i ) do 5. if ( V[j] < V[j+1] ) 6. temp := V[j]; 7. V[j] := V[j+1]; 8. V[j+1]:= temp; 9. end if 10. j := j + 1; 11. end while 12. i := i - 1; 13. end while 14. return V; Figure 1. Bubble Sort algorithm with explicit basic-blocks, which are represented by the grouped numbered lines.

325

of the set E of GP and the condition to execute it is satisfied (e.g., the blue arrow in Fig. 2); it is wrong if the executed branch is an element of the set E, but its condition to execute cannot be satisfied (e.g., the purple arrow in Fig. 2); and an executed branch is wrong if it is not an element of the set E (e.g., the red arrow in Fig. 2). A control-flow error (CFE) occurs when either a wrong or illegal branch is executed. Notice that in these two cases, a CFE cannot exist if the program execution is not corrupted, i.e., an illegal branch cannot exist in a correct program execution, because it only executes branches from the set E; a wrong branch cannot exist, because it is always possible to satisfy the logical conditions of all branches if program execution is correct. A CFE can be created by a radiation-induced SEU in any of the following three scenarios: • A non-branch instruction being executed changes into a non-valid branch, i.e., the operation code data is corrupted. • The target address of a valid branch is corrupted. • One of the variables composing a logical expression that activates a branch is corrupted. Scenarios (1) and (2) leads to an illegal branch, and scenario (3) leads to a wrong branch. A data-flow error (DFE) is caused by a radiation-induced SEU that corrupts variables within a basic block. A DFE might lead to erroneous results or even to a CFE, in case the corrupted variable is used in a logical expression controlling a branch. The focus of this paper is CFE’s. For an extensive review of mitigation techniques of CFE and DFE, interested readers may refer to the work of Goloubeva et al., (2006).

3 0

Figure 2. Control-flow graph of the Bubble Sort algorithm. The blue arrow is a legal branch (together with the black arrows), the purple arrow is a wrong branch, and the red arrow is an illegal branch. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

326

Ferreira, R.R., Parizi, R.B., Carro, L. and Moreira, Á.F.

The detection of transient CFE was established in the literature with techniques that check assertions during runtime. The general idea is to compute signatures identifying each basic block, and checking the signatures generated during compilation and runtime. If they do not match, an error is signaled. CFE’s were first identified by the usage of watchdog processors, which are intrusive in the hardware design ( Saxena and McCluskey, 1990). Lately, techniques based on the signature checking scheme in software, such as the Control-flow Checking Approach (CCA) (Kanawati et al., 1996), were identified, but with a coverage rate of only 38% and a performance overhead of 50%. Advances in the signature checking method offered some improvements on coverage and performance, such as the Control-Flow Checking by Software Signatures, which incurs in 50% of overhead in execution time and program size (Oh et al., 2002). The most efficient technique of signature checking capable of correcting errors is the ACCE (Vemu et al., 2007), which incurs in approximately 20% of overhead in execution time to produce an average 70% of correct answers in fault-injection campaigns. However, ACCE is not capable of correcting errors that occur within a basic block, i.e., in the data flow; hence, the use of complementary techniques is required. When ACCE is enhanced with data-flow correction, its coverage rate achieves the average of 91.6% (Vemu et al., 2007). Because the CFG is one of the most important software artifacts used by compilers when analyzing and modifying programs, it is important to measure how the compiler impacts the reliability of the software mitigation techniques for radiation-induced CFE. The understanding of these impacts is imperative to employ software mitigation techniques in real systems. This paper presents a study using the ACCE mitigation technique, which is briefly reviewed in the next section. AUTOMATIC CORRECTION OF CONTROL-FLOW ERRORS ACCE (Vemu et al., 2007) is a software technique for reliability that detects and corrects CFE’s due to random and arbitrary bit-flips that might occur during software execution. The hardening of an application with ACCE is done at compilation, because it is implemented as a transformation pass in the compiler. ACCE modifies the applications’ basic blocks with the insertion of extra instructions that perform

the error detection and correction during software execution. In this section, we briefly explain how the ACCE works in two separate subsections, one dedicated to error detection and the other to error correction are discussed in the subsequent subsections. The reader should refer to the ACCE article for a detailed presentation and experimental evaluation (Vemu et al., 2007). The fault model that ACCE assumes is further described in the “Fault Model and Methodology” section. Control-Flow Error Detection ACCE performs online detection of CFE-s by checking the signatures in the beginning and in the end of each basic block of the CFG, thus, ACCE is classified as a signature checking SIHFT technique as termed in the published literature. The basic block signatures are computed and generated during compilation; the signature generation is critical, because it requires computing non-aliased signatures between the basic block, i.e., each block must be unambiguously identified. In addition, for each basic block found in the CFG, two additional code regions are added, the header and the footer. The signature checking during execution takes place inside these code regions. Figure 3 shows two basic blocks (labeled as N2 and N6) with the additional code regions. The top region corresponds to the header and the bottom to the footer. Still, at compilation, the ACCE creates two additional blocks for each function, namely the function entry block and the Function Error Handler (FEH). For instance, Fig. 3 depicts a portion of two functions, f1 and f2, both owning entry blocks labeled as F1 and F2, and FEHs, labeled as FEH_1 and FEH_2, respectively. Finally, ACCE creates a last extra block, the Global Error Handler (GEH), which can only be reached from a FEH block. The role of these blocks will be presented soon. At runtime, the ACCE maintains a global signature register (represented as S), which is constantly updated to contain the signature of the basic block that the execution has reached. Therefore, during the execution of the header and footer code regions of each basic block, the value of the signature register is compared with the signatures generated during compilation for those code regions, and if those values do not match, a CFE is detected and then the control should be transferred to the corresponding FEH block of the function where the execution takes place at that time. The ACCE also maintains the current function register

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software

(represented as F), which stores the unique identifier of the function currently being executed. The current function register is only assigned at the extra entry function block. This process encompasses the detection of an illegal and erroneous branch due to a soft error. Figure 3 depicts an example of the checking and update of signatures performed in execution time that occurs in a basic block. In this example, the CFE occurs in the block N2 of function F1, where an illegal jump incorrectly transfers the control flow to the basic block N6 of function F2. When the execution reaches the footer of the block N6, the signature register S is checked against the signature generated at compilation. In this case, S = 0111 (i.e., the previous value

assigned in the header of the block N2). Thus, the branch test in the N6 footer will detect that the expected signature does not match with the value of S, and thus, the CFE must be signaled (step 1 in Fig. 3). In this example, the application branches to the address f2_err, making the application enter the FEH_2 block (because the error was detected by a block owned by the function F2, the FEH invoked is the FEH_2). At this point, the CFE is detected and ACCE can proceed with the correction of the detected CFE. Control-Flow Error Correction The correction process starts as soon as an illegal jump is detected by the procedure described in the last subsection,

F=1 br err_fla==1,fl_sxy

F=2 br err_fla==1,fl_sxy

...

... N2

br S! = 1110, f2_svv S=S XOR 1011

br S! = 1110, f2_svv S=S XOR 1000

CFE

[S=0111]

[S=0111] br S! = 0110, f1_svv

...

br S! = 0110, f2_svv S=S XOR 1010 ...

5 FEH_2

FEH_1

327

brF!=1, error_handler err_flag=0 num_err=num_err+1 br num_err > thresh, exit ... br S == 0111, jmp N2 ... jmpf1_svv

brF!=2, error_handler err_flag=0 num_err=num_err+1 br num_err > thresh, exit ... br S == 0110, jmp N6 ... jmpf2_svv 2 GEH err_flag=1 br F == 1, F1 br F == 2, F2 num_err=num_err+1 br num_err>thresh, exit jmp error_handler

Figure 3. Depiction of how the control is transferred from a function to the basic blocks that ACCE has created when a control-flow error occurs during software execution. In this figure, there is a control-flow error (dashed arrow) causing the execution to jump from the block N2 of function F1 to the block N6 of function F2.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

328

Ferreira, R.R., Parizi, R.B., Carro, L. and Moreira, Á.F.

with the control flow transferred to the FEH corresponding to the function where the CFE is found. The FEH checks if the illegal jump was originated in the function it is r esponsible to handle its detected errors by comparing the value of the function’s identifier (F1 or F2, in the example of Fig. 3) with the current function register F. If the error occurred in the function stored in the F register, FEH evaluates the current value of the signature register and then transfers the control to the basic block that is the origin of the illegal jump (this origin is stored in the S register). On the other hand, if the illegal jump is not originated in the function where the d etection has occurred, the FEH then transfers the control flow to the GEH. In this case, the GEH is responsible for identifying the function where the CFE has occurred and to transfer the control flow back to this function, so that the error is correctly treated by the function’s FEH. The GEH searches the function where the error has occurred and transfers the control to its entry block, which then sends the control flow to the proper FEH so that the error can be corrected, i.e., branching the control to the basic block where the CFE has occurred. Recalling the example depicted in Fig. 3, after the CFE is detected and the control is transferred to FEH_2 (step 1), the F register is matched against the function identifier of the function from where the control originated. However, because the CFE originated in the basic block N2 of the function F1, F = 1. Therefore, FEH_2 is not capable of finding the basic block where the CFE originated, and then it transfers the control to the GEH so that the correct FEH can be found (step 2). The GEH searches for the function identifier stored in F, until it finds that it should branch to F1 (step 3). Upon reaching the entry block F1, the variable err_flag=1, because it is assigned to 1 in the GEH, meaning that there is an error that should be fixed, thus, the control branches to FEH_1 (step 4). Now, because F=1, FEH_1 knows that it is the FEH capable of handling the CFE and, as such, it sets the variable err_flag to 0. Finally, it searches for the basic block that has the signature equal to the register S. Upon finding it, the control branches to this basic block, i.e., N2 in Fig. 3 (step 5). This last branch restores the control flow to the point of the program right before the occurrence of the CFE. Notice that inside all the FEH and the GEH, there is the variable num_error, counting how many times the control has passed through a FEH or a GEH. This acts as a threshold for the number of how many times the correction must be attempted, which is

necessary to avoid an infinite loop in case the registers F or S get corrupted for any reason. This process concludes the correction of a CFE with the ACCE. FAULT MODEL AND METHODOLOGY The fault model we assume in the experiments is the single bit flip, i.e., only one bit of a word is changed when a fault is injected. The ACCE is capable of handling multiple bit flip as long as the bits flipped is within a same word. Because the fault injection, as it will be discussed later, guarantees that the injected fault ultimately turned into a manifested error, it does not matter how many bits are flipped, i.e., there is no silent data corruption, meaning the faults that cause a word to change its value neither change the behavior of the program nor its output. This could happen in the case that the fault flipped the bits of a dead variable. The ACCE technique was implemented as a transformation pass in the Low Level Virtual Machine (LLVM; Lattner and Adve, 2004) production compiler, which performs all the modifications in the CFG using the LLVM Intermediate Representation (LLVM-IR). The LLVM was selected as our compilation platform, because of its increased use in the industry, accompanied with a very detailed documentation and quality of its source code. The ACCE transformation pass was applied after the set of compiler optimizations, because executing in the opposite order, a compiler optimization could invalidate the ACCE generated code and semantics. Table 1 presents the LLVM optimization passes used in the experiments. Because the ACCE is a SIHFT technique to detect and correct CFE-s, the adopted fault model simulates three distinct control-flow disruptions that might occur due to a CFE. Remember that a CFE is caused by the execution of an illegal branch to a possibly wrong address. The branch errors considered in this paper are as follows: • Branch creation: The program counter is changed,

transforming an arbitrary instruction (e.g., an addition) into an unconditional branch. • Branch deletion: The program counter is set to the next

program instruction to execute independently if the current instruction is a branch. • Branch disruption: The program counter is disrupted

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

to point to a distinct and possibly wrong destination instruction address.

Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software

We implemented a software fault injector, using the GDB (GNU Debugger), in a similar fashion as implemented by Krishnamurthy et al., (1998), which is an accepted fault-injection methodology in the embedded systems domain, to perform the fault-injection campaigns. The steps of the fault-injection process are the following: • The LLVM-IR program resulting from the compilation with a set of optimization and with ACCE is translated to the assembly language of the target machine.

329

IMPACT OF COMPILER OPTIMIZATIONS ON SOFTWARE RELIABILITY This section studies the impacts on software reliability when an application is compiled with a set of compiler o ptimizations and further hardened with the ACCE method. Throughout this section, the baseline for all comparisons is an application compiled with the ACCE method without any other compiler optimization. The ACCE performs detection and correction of CFE-s,

• The execution trace in assembly language is extracted

from the program execution with GDB. • A branch error (branch creation, deletion, or disruption)

Table 1. Set of Low Level Virtual Machine optimization passes used for experimental evaluation in this paper.

is randomly selected. On an average, each branch error

-adce

-loop-reduce

accounts for 1/3 of the amount of injected errors.

-always-inline

-loop-rotate

-argpromotion

-loop-simplify

-block-placement

-loop-unroll

-break-crit-edges

-loop-unswitch

-codegenprepare

-loweratomic

-constmerge

-lowerinvoke

-constprop

-lowerswitch

-dce

-mem2reg

-deadargelim

-memcpyopt

-deadtypeelim

-mergefunc

-die

-mergereturn

-dse

-partial-inliner

-functionattrs

-prune-eh

-globaldce

-reassociate

-globalopt

-reg2mem

-gvn

-scalarrepl

-indvars

-sccp

-inline

-simplifycfg

-instcombine

-simplify-libcalls

-internalize

-sink

-ipconstprop

-sretpromotion

-ipsccp

-strip

-jump-threading

-strip-dead-debug-info

-lcssa

-strip-dead-prototypes

-licm

-strip-debug-declare

-loop-deletion

-strip-non-debug

-loop-extract

-tailcallelim

-loop-extract-single

-tailduplicate

• One of the instructions from the trace obtained in step

2 is chosen at random for fault injection. In this step, a histogram of each instruction is computed because instructions that execute more often have a higher probability to be disrupted. • If the chosen instruction in step 4 executes n times,

choose at random an integer number k with 1 ≤ k ≤ n. • Using GDB, a breakpoint is inserted right before the k-th

execution of the instruction selected in step 4. • During program execution, upon reaching the breakpoint

inserted in step 6, the program counter is intentionally corrupted by flipping one of its bits to reproduce the branch error chosen in step 3. • The program continues its execution until it finishes.

A fault is only considered valid, if it has generated a CFE, i.e., silent data corruption and segmentation faults were not considered to measure the impacts of the compiler optimizations on reliability. All the experiments in this paper were performed in a 64-bit Intel Core i5 2.4 GHz desktop with 4 GB of RAM and the LLVM compiler version 2.9. For all program versions, where each version corresponds to the program compiled with a set of optimizations plus the ACCE pass, 1,000 faults were injected using the aforementioned fault-injection scheme. In the experiments we considered ten benchmark applications from the MiBench (Guthaus et al., 2001) embedded benchmark suite as follows: basicmath, bitcount, crc32, dijkstra, fft, patricia, quicksort, rijndael, string search, and susan (comprising susan corners, edge, and smooth).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

330

Ferreira, R.R., Parizi, R.B., Carro, L. and Moreira, Á.F.

thus all data discussed in this section considers the

is the case of the crc_32 one, where the RIP is within the ±5%

correction rate as the data to compute the efficiency

interval around the baseline.

metric. In this analysis, we use 58 optimizations

Figure 5 depicts the RIP of a selected subset of the 58

provided by the LLVM production compiler. Finally, the

LLVM optimizations, making it clear that even within a

results were obtained using the fault model and fault

small subset, the variation in the RIP for reliability is far from

injection methodology described in the section “Fault

negligible. For instance, the always-inline LLVM optimization

Model and Methodology”.

has an error correction RIP interval of [-4.55%, +9.24%].

The impact of the compiler optimizations when compiling

Usually compiler optimizations are applied in bulk, using

for reliability is measured in this paper using the metric

several of them during compilation. Therefore, it is also

Relative Improvement Percentage (RIP; Pan and Eigenmann,

important to examine if successive optimization passes could

2006). The RIP is presented in Eq. 1, where Fi is a compiler

compromise or increase software reliability of a hardened

optimization, E(Fi) is the error correction rate obtained for

application. Figure 6 presents the error correction rate RIP,

a hardened application compiled with Fi, and EB is the error

where the hardened application was compiled with a subset of

correction rate obtained for the baseline, i.e., the application

the 58 LLVM optimizations. In this experiment, we used six

compiled only with ACCE and without any optimization.

sizes of subsets: 10, 20, 30, 40, 50, and 58. The RIP shown in Fig. 6 is the average of five random subsets, i.e., it is an average

RIPB ( Fi ) =

E ( Fi ) − E B . 100% EB

(1)

of distinct subsets of the same size. Taking the average and picking the optimizations at random, reproduces the effects of indiscriminately picking the compiler optimizations, or at

Figure 4 shows a scatter plot of the obtained RIP for each

least, selecting optimizations with the object of optimizing

application, with each of the 58 LLVM optimizations being

performance without previous knowledge of how the selected

a point in the y-axis. Each point represents the hardened

optimizations together influences the software reliability.

application compiled with a single LLVM optimization at

It is possible to see that the cumulative effect of compiler

a time, with each application compiled with 58 distinct

optimizations in the error correction RIP is in most of the

optimizations. Figure 4 shows that several optimizations

cases deleterious, but for a few exceptions. Figure 6 confirms

increase the RIP considerably, sometimes reaching a RIP of ~10%. This is a great result, which shows that reliability can be increased for free by just picking appropriate optimizations that facilitates for ACCE the process of error detection and correction. However, we also find that some optimizations totally jeopardize reliability, reaching a RIP of -73.27% (bottom filled red circle for bitcount). It is also possible to gather evidence that the structure of the application also influences how an optimization has an impact on the RIP of reliability. Let us consider the blockplacement optimization, which is represented by the white diamond in Fig. 4. In the case of the qsort application, block-placement has a RIP of -42.75% and a RIP of +11.68%. The reader can notice that other optimizations also show

that some applications are less sensitive to the effects of compiler optimizations, e.g., the crc32 has its RIP within the interval [-1.11%–0.73%]. On the other hand, basicmath, bitcount, and patricia, are jeopardized. It is interesting to notice that the RIP in case of picking a subset of optimizations is not subject to the much severe reduction that was measured when only a single optimization was used (Fig. 4), providing an evidence that the composition of distinct optimization may be beneficial for reliability. Based on the data and experiments discussed in this section, it is clear that selection of compiler optimizations requires the software designer to take into the consideration that some optimizations may not be adequate in terms of reliability for a given application. Moreover, data also shows that a given optimization is not only by itself a source of

this behavior (increasing RIP for some applications and

reliability reduction; reliability is also dependent of the

decreasing it for others). It also happens that some hardened

application being hardened, and how a given optimization

applications are less sensitive to compiler optimizations, as it

facilitates or not the work of the ACCE technique.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software

331

Relative Percentage RelativeImprovement Improvement Percentage

20.00 20.00 10.00 10.00 0.00 0.00 -10.00 -10.00 -20.00 -20.00 -30.00 -30.00 -40.00 -40.00 -50.00 -50.00 -60.00 -60.00 -70.00 -70.00 -80.00 -80.00

ath cm i s ba

ou bitc

rt qso

stra

dijk

pat

ric

s ng_ stri

ch ear

an sus

j da rin

fft

_32

crc

Figure 4. Relative Improvement Percentage for the error correction rate of applications hardened with ACCE under further compiler optimization. Each hardened application was compiled with a single optimization at a time, but all applications were compiled with the 58 Low Level Virtual Machine optimizations, thus, each hardened application has 58 versions. The baseline (Relative Improvement Percentage = 0%) is the error correction rate of the hardened application compiled without any Low Level Virtual Machine optimization.

Relative Improvement Percentage

always-inline

inline

loop-unroll

tailduplicate

10.00 8.00 6.00 4.00 2.00 0.00 -2.00 -4.00 -6.00

ath icm bas

t oun bitc

rt qso

stra

dijk

ric pat

se ng_ stri

h arc

an sus

j da rin

fft

_32

crc

Figure 5. Relative Improvement Percentage of a selected subset of the 58 Low Level Virtual Machine optimizations. The baseline (Relative Improvement Percentage=0%) is the error correction rate of the hardened application compiled without any Low Level Virtual Machine optimization. J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

Relative Improvement Percentage

332

Ferreira, R.R., Parizi, R.B., Carro, L. and Moreira, Á.F.

# of optimizations

5.00 0.00 -5.00 -10.00 -15.00

th ma c i s ba

t oun bitc

rt qso

stra

dijk

pat

rici

s ng_ stri

ch ear

an sus

j da rin

fft

_32

crc

Figure 6. RIP of random subsets of the 58 Low Level Virtual Machine optimizations with a varying number of optimizations for each different subset: 10, 20, 30, 40, 50, and 58 optimizations. The Relative Improvement Percentage for each subset was measured taking the average of six random subsets for each subset size. Hence, distinct possible optimizations for subsets were considered. The baseline (Relative Improvement Percentage=0%) is the error correction rate of the hardened application compiled without any Low Level Virtual Machine optimization.

RELATED WORK Much attention has been devoted to the impact of compiler optimizations on program performance in the published literature. However, the understanding of how those optimizations work together and how they influence each other is a rather recent research topic. The Combined Elimination (CE) (Pan and Eigenmann, 2006) is an analysis approach to identify the best sequence of optimizations for a given application set using the GNU Compiler Collection (GCC). The authors discuss that simple orchestration schemes between the optimizations can achieve near-optimal results as if it has performed an exhaustive search in all the design space created by the optimizations. CE is a greedy approach that first compiles the programs with a single optimization, u sing this version as the baseline. From those baseline versions, the set of RIP is calculated, which is the percentage that the program’s performance is either reduced or increased. With the RIP at hand for all baselines, the CE starts removing the optimizations with negative RIP, until the total RIP of all optimizations applied into a program do not reduce. CE was evaluated in different architectures, achieving an average RIP of 3% for the SPEC2000, and up to 10% in case of the Pentium IV for the floating point applications.

The Compiler Optimization Level Exploration (COLE) (Hoste and Eeckhout, 2008) is another approach to achieve performance increase by selecting a proper optimization sequence. COLE uses a population-based multi-objective optimization algorithm to construct a Paretto optimal set of optimizations for a given application using the GCC compiler. The data found with COLE give some insightful results about how the compiler optimizations behave when they are applied with several of them at the same time. For instance, 25% of the GCC optimizations appear in at most one Paretto set, and some of them appear in all sets. Therefore, 75% of all the optimizations do not contribute to improve the performance, meaning that they can be safely ignored! COLE also shows that the quality of an optimization is highly tied with the a pplication set. The Architectural Vulnerability Factor (AVF) (Mukherjee et al., 2003) is a metric to estimate the probability that the bits in a given hardware structure will be corrupted by a soft error when executing a certain application. The AVF is calculated as the total time the vulnerable bits remains in the hardware architecture. For example, the register file has a 100% AVF, because all of its bits are vulnerable in case of a soft error. The AVF metric is highly influenced by the application due to l iveness of program’s variables. For instance, a dead variable has a 0%

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

Compiler Optimizations Impact the Reliability of the Control-Flow of Radiation-Hardened Software

AVF, because it is not used in a computation. The impact of the GCC optimizations in the AVF metric is evaluated by trying to reduce the AVF-delay-square-product (ADS), introduced by the authors (Jones et al., 2008). The ADS considers a linear relation of the AVF between the square of the performance in cycles, clearly prioritizing performance over reliability. It is reported that the -O3 optimization level is detrimental both to the AVF and performance, because the benchmarks that are considered (MiBench) have increased the number of loads executed. Again, it was found that the patricia application was the one with the highest reduction in the AVF at 13%. Bergaoui and Leveugle (2011) analyzed the impact of compiler optimizations on data reliability in terms of variable liveness. Liveness of a variable is the time period between the variable that is written and it is last read before a new write operation. The authors concluded that the liveness is not related only with the compiler optimization, but it also depends on the application being compiled, which is in accordance with the discussion of this paper. This paper shows that some optimizations tend to extend the time a variable is stored in a register instead of memory. The goal behind this is obvious, i.e., it is much faster to fetch

333

CONCLUSION In this paper, we characterized the problem of compiling embedded software for reliability, given that compiler optimizations impact the coverage rate. The study presented in this paper makes clear that selecting optimizations indiscriminately, can decrease software reliability to unacceptable levels, probably avoiding the software to be deployed as originally planned. Embedded software and systems deployed in space applications must always be certified with evidence that they support harsh radiation environments, and given the increasing technology scaling, other safety critical embedded systems might have to tolerate radiation-induced errors in a near future. Therefore, the embedded software engineers must be very careful while compiling the safety critical embedded software. Future research work is focused on the formalization of the ACCE transformation pass to generate automatic proofs about the correctness of programs compiled with the ACCE. This step is important to allow the certification of software hardened with ACCE.

the value of a variable when it is in the register than in the memory. However, the memory is usually more protected than registers because of cheap and efficient error correction code (ECC) schemes, and thus, thinking about reliability, it is not a good idea to expose a variable in a register for a longer time. The solution to that could be the application of ECC, such as Huffman to the program variables itself. Decimal Hamming (DH) (Argyrides et al., 2011) is a software technique that performs this for a class of programs where the program’s output is a linear function of the input. The generalization of the efficient data-flow SIHFT techniques, such as DH (i.e., ECC of program variables) is still an open research problem.

ACKNOWLEDGEMENTS This work is supported by the CAPES foundation of the Ministry of Education, CNPq research council of the Ministry of Science and Technology, and FAPERGS research agency of the State of Rio Grande do Sul, B razil. R. Ferreira was supported with a doctoral research grant from the Deutscher Akademischer Austauschdienst (DAAD) and from the Fraunhofer-Gesellschaft, Germany.

REFERENCES Argyrides, C., Ferreira, R., Lisboa, C. and Carro, L., 2011, “Decimal Hamming: A Novel Software-Implemented Technique to Cope with Soft Errors”, Proceedings of the 26th IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems, pp. 11-17, doi: 10.1109/DFT.2011.35 ARM, n.d. 2012, “ARM Mali Graphics Hardware”, Retrieved in December 21, 2012, from http://www.arm.com/products/ multimedia/mali-graphics-hardware/index.php.

Bergaoui, S. and Leveugle, R., 2011, “Impact of Software O ptimization on Variable Lifetimes in a Microprocessor-Based System”, Proceedings of the 6th IEEE International Symposium on Electronic Design, Test and Application, pp. 56-61, doi: 10.1109/DELTA.2011.20 Binder, D., Smith, E.C. and Holman, A.B., 1975, “Satellite Anomalies from Galactic Cosmic Rays”, IEEE Transactions on Nuclear Science, Vol. 22, No. 6, pp. 2675-2680, doi: 10.1109/ TNS.1975.4328188

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

334

Ferreira, R.R., Parizi, R.B., Carro, L. and Moreira, Á.F.

Borkar, S., 2005, “Designing Reliable Systems from Unreliable Components: The Challenges of Transistor Variability and Degradation”, IEEE Micro, Vol. 25, No. 6, pp. 10-16, doi: 10.1109/ MM.2005.110

Lattner, C. and Adve, V., 2004, “LLVM: A Compilation Framework for Lifelong Program Analysis & Transformation”, Proceedings of the International Symposium on Code Generation and Optimization, pp. 75-86, doi: 10.1109/CGO.2004.1281665

Esmaeizadeh, H., Emily, B., Renee, A. and Sankaralingam, K., 2011, “Dark Silicon and the End of Multicore Scaling”, IEEE Micro, Vol. 32, No. 3, pp. 122-134, doi: 10.1109/MM.2012.17

Mehlitz, P.C. and Penix, J., 2005, “Expecting the unexpected – radiation hardened software”, NASA Ames Research Center, pp. 10.

Goloubeva, O., Rebaudengo, M., Sonza Reorda, M. and Violante, M., 2006, “Software-Implemented Hardware Fault Tolerance”, Ed. Springer, New York, NY, USA, p 228. Guthaus, M.R., Ringenberg, J.S., Ernst, D., Austin, T.M., Mudge, T. and Brown, R.B., 2001, “MiBench: A Free, Commercially Representative Embedded Benchmark Suite”, Proceedings of the IEEE International Workshop of Workload Characterization, pp. 3-14, doi: 10.1109/WWC.2001.990739 Hoste, K. and Eeckhout, L., 2008, “Cole: Compiler Optimization Level Exploration”, Proceedings of the 6th Annual IEEE/ACM International Symposium on Code Generation and Optimization, pp. 165-174, doi: 10.1145/1356058.1356080 ITRS, 2012, “ITRS 2009 Roadmap”, International Technology Roadmap for Semiconductors. Jones,T.M., O’Boyle, M.F.P. and Ergin, O., 2008, “Evaluating the Effects of Compiler Optimisations on AVF”, Proceedings of the Workshop on Interaction Between Compilers and Computer Architecture, 6p. Kanawati, K., Krishnamurthy, N., Nair, S. and Abraham, J.A., 1996, “Evaluation of Integrated System-level Checks for On-Line Error Detection”, Proceedings of the 2nd International Computer P erformance and Dependability Symposium, pp. 292-301, doi: 10.1109/ IPDS.1996.540230 Krishnamurthy, N., Jhaveri, V. and Abraham, J.A., 1998, “A Design Methodology for Software Fault Injection in Embedded Systems”, Proceedings of the Workshop on Dependable Computing and its applications, pp. 12

Mukherjee, S.S., Shrewsbury, M.A., Weaver, C., Emer, J. and Reinhardt, S.K., 2003, “A Systematic Methodology to Compute the Architectural Vulnerability Factors for a High-Performance Microprocessor”, Proceedings of the 36th Annual IEEE/ACM International Symposium on Microarchitecture, pp. 29-40, doi: 10.1109/MICRO.2003.1253181 Normand, E., 1996, “Single Event Upset at Ground Level”, IEEE Transactions on Nuclear Science, Vol. 43, No. 6, pp. 2742-2750, doi: 10.1109/23.556861 Oh, N., Shirvani, P.P. and McCluskey, E.J., 2002, “Control-flow Checking by Software Signatures”, IEEE Transactions on Reliability, Vol. 51, No. 1, pp. 111-122, doi: 10.1109/24.994926 Pan, Z. and Eigenmann, R., 2006, “Fast and Effective Orchestration of Compiler Optimizations for Automatic Performance T uning”, Proceedings of the International Symposium on Code Generation and Optimization, pp. 319-332, doi: 10.1109/CGO.2006.38 Rech, P., Aguiar, C., Ferreira, R., Silvestri, M., Griffoni, A., Frost, C. and Carro, L., 2012, “Neutron-Induced Soft Errors in Graphic Processing Units”, IEEE Radiation Effects Data Workshop, pp. 1-6, doi: 10.1109/REDW.2012.6353714 Saxena, N. and McCluskey, E., 1990, “Control Flow Checking using Watchdog Assists and Extended-Precision Checksums”, IEEE Transactions on Computers, Vol. 39, No. 4, pp. 554-559, doi: 10.1109/12.54849 Vemu, R., Gurumurthy, S. and Abraham, J.A., 2007, “ACCE: Automatic Correction of Control-Flow Errors”, IEEE International Test Conference, pp. 1-10, doi: 10.1109/TEST.2007.4437639

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.323-334, Jul.-Sep., 2013

doi: 10.5028/jatm.0501. doi: 10.5028/jatm.v5i3.227

Analysis of Total Ionizing Dose Effects on 0.13 μm Technology-TemperatureCompensated Voltage References Thiago Hanna Both1, Dalton Colombo1, Ricardo Vanni Dallasen1, Gilson Inácio Wirth1

ABSTRACT: The purpose of this work is to briefly discuss the effects of the total ionizing dose (TID) on MOS devices in order to estimate the results of future irradiation tests on temperature-compensated voltage references that are implemented on a mixed-signal chip fabricated using IBM 0.13 µm technology. The analysis will mainly focus on the effects of the parametric variations on different voltage references. Monte-Carlo analyses were performed in order to determine the effects of threshold voltage shifts in each transistor on the output voltage. KEYWORDS: Ionizing dose, Radiation, TID, Voltage reference.

INTRODUCTION Space and high-altitude aeronautical applications of electronics are exposed to a continuous action of ionizing radiation. The interaction between radiation and the structure of semiconductor devices results in undesired effects, which may be either transient or accumulated. Predicting these effects and accurately addressing solutions that guarantee operation under radiation is a challenge for the aerospace industry. The transient effects occur when an energetic particle interacts with a sensitive region of the structure, resulting in a generation of charge that produces an electric pulse at a circuit node (Single Event Effect–SEE). Transient effects may result in hard errors, which are permanent, destructive, faults (Sexton, 2003); soft errors, which are non-destructive faults that affect circuit operation (Karnik and Hazucha, 2004); or disturbances that do not cause a fault. Accumulated effects can be categorized into displacement damage (DD) and total ionizing dose effects (TID). Displacement damage occurs when the collision of a particle results in non-ionizing energy loss to the medium. This collision d isplaces an atom from the lattice, damages the crystalline structure of the silicon, and creates a defect that degrades electric parameters of the device (Srour, 2003). The TID usually result from accumulated charge in the dielectric and at the interface between the dielectric and the semiconductor, known, respectively, as oxide-trapped charge and interface-trapped charge.

1.Universidade FederaldodoRio Riode Grande do–Sul Porto Alegre/RS – Brazil Universidade Federal Janeiro Rio– de Janeiro/RJ – Brazil Author Hanna Slama Both | Universidade Federal do Rio Grande do Sul | Avenida Bento Gonçalves, Campus do Vale, Bloco IV | Author for correspondence: Thiago Jules Ghislain | Departamento de Engenharia Mecânica/COPPE/UFRJ/C.P . 68.503 9500, | CEP 21.945-970 Rio de Janeiro/RJ – CEP: 91.509-900 Porto Alegre/RS – Brazil | Email: thboth@gmail.com Brazil | email: julesslama@yahoo.com.br Received: 19/01/13 | Accepted:19/04/13 Received: 02/02/12 | Accepted: 30/10/12

o 3, pp.335-340, Jul.-Sep., 2013 J. Aerosp. Technol. São Manag., Sãodos JoséCampos, dos Campos, Vol.5, J. Aerosp. Technol. Manag., José Vol.X, NoNX, pp.1-8, XXX.-XXX., 2013

336

Both, T.H., Colombo, D., Dallasen, R.V. and Wirt, G.I.

Ionizing radiation generates electron-hole pairs in the oxides of semiconductor devices. These electron-hole pairs suffer from an initial recombination that is a f unction of the electric field applied and the energy and type of the incident particle. After this initial recombination, the electrons, which have a higher mobility inside the insulator, escape from the oxide in a few picoseconds. The holes, on the other hand, have lower mobility and are eventually trapped in oxide traps, border traps or interface traps. The detailed mechanism of hole trapping in oxides can be found elsewhere (Oldham and McLean, 2003). In both NMOS and PMOS transistors, oxide-trapped charge is typically net positive, whereas interface-trapped charge is typically net positive for PMOS transistors and negative for NMOS transistors. The consequence of this trapped charge is the degradation of electric parameters of the semiconductor devices, such as the threshold voltage, the leakage current, and the carrier mobility. It is also reported in the works of Fleetwood and Scofield (1990) and Fleetwood et al., (1994) that TID effects increase the 1/f noise in MOS devices, as shown in Fig. 1. It should be noted that the spikes in Fig.1 are related to the power-line fundamental frequencies and their harmonics. The 1/f noise in MOS devices is associated with charge carrier trapping near or at the interface between the semiconductor and the insulator. Charge trapping causes

TEMPERATURE-COMPENSATED VOLTAGE REFERENCES Voltage references are building blocks that are p resent in a large variety of circuits; for instance, v oltage regulators, comparators, and data c onverters. A v oltage reference circuit should provide a stable o utput voltage despite variations in temperature, power s upply, and p rocess. Since voltage reference circuits can limit the accuracy of these applications, it is important to verify the impact of TID on the generated output voltage (V REF).

200 krad

10-12

100 krad 30 krad

Irradiate

PRE

Noise Power, K (10-9 V2)

Power Spectral Density (V2/Hz)

500 krad

carrier number fluctuations and mobility fluctuations in the channel region of transistors. The activity of these traps results in d iscrete variations of the signal, known as R andom Telegraph Signal (RTS), which, in MOS devices, is the major source of 1/f noise. It has also been reported that the annealing of the device after irradiation, for relatively thick oxides, results in a reduction of 1/f noise (Meisenheimer and Fleetwood, 1990), as observed in Fig. 2. The degradation of electric parameters of the device and the increase in the 1/f noise may either result in malfunction or fully disable an electronic system. The purpose of this work is to review the TID effects on MOS devices in order to estimate the effects of an irradiation test on different temperature-compensated voltage references that are integrated in one chip using IBM 0.13 µm technology. Monte-Carlo analyses were performed in order to identify how parametric variations affect different topologies.

10-13

10-14

Vd=100mV tox=48nm Vg-Vt=3.0V

100

101 Frequency (Hz)

102

Figure 1. 1/f noise spectra for an irradiated NMOS transistor (oxide thickness of 48 nm). Adapted from Meisenheimer and Fleetwood, 1990.

80ºC Anneal

6.0 4.0 2.0 0.0 PRE

Vg-Vt=3.0V Vd=100mV f=10Hz tox=48nm

101

102

103

104

105

106

Time (s) Figure 2. Noise power response during irradiation and annealing of an NMOS transistor (oxide thickness of 48 nm). Adapted from Meisenheimer and Fleetwood, 1990.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.335-340, Jul.-Sep., 2013

Analysis of Total Ionizing Dose Effects on 0.13 μm Technology-temperature-compensated Voltage References

The traditional implementation for voltage reference is the bandgap circuit, where the generated VREF is the bandgap voltage of silicon extrapolated to absolute zero Kelvin (i.e., ~1.2 V). Due to its stability regarding temperature and process, bandgap references have been used for the past forty years. Bandgap references generate a temperature-compensated VREF through a balanced sum of the diode voltage and the thermal voltage. The diode voltage (or base-emitter voltage) has a negative temperature coefficient (TC), while the t hermal voltage has a positive TC. The thermal voltage is given by k∗T/q, where k is the Boltzmann constant, T is the temperature and q is the electron charge. Another topology of voltage reference that is widely used in the industry is the threshold-voltage-(VTH)-based reference. This topology of reference generates VREF that is equal to the threshold of a transistor extrapolated to absolute zero Kelvin (Colombo et al., 2011). The thresholdbased voltage reference has gained increasing importance recently because of its ability to operate under low supply voltages (e.g., less than 1 V). Different from bandgap circuits that use diodes or BJTs, VTH-based references use the gatesource voltage (VGS) of MOS transistors to generate a voltage with negative a TC. Bandgap and threshold-voltage-based references are two of the most frequently used techniques to generate temperature-compensated VREF and, consequently, they are chosen as the case study for our project. An integrated circuit with five different voltage references was designed. In the fabricated chip, there are four references based on VTH, shown in Figs. 3 to 5. and one bandgap-based reference, shown in Fig. 6.

VDD M1

M4 R1

M10

M11

R3 R2

R2 M2

M4 R1 M6

Figure 4. Simulated voltage reference 2 (Colombo et al., 2011).

Ip VREF

M2 M7

MR1

Current source subcircuit

Figure 3. Simulated voltage reference 1 (Colombo et al., 2011).

VREF

Reference 2 is the simplest V TH-based voltage r eference that can be implemented. Its output voltage is given by the sum of VGS of M6 and the voltage across R2, which is proportional to the temperature. Reference 1 is similar to reference 2, but the temperature compensation is done by adding currents with opposite TC instead of voltages. Reference 3 and 4 are alternative V TH-references that can be implemented with either one or none resistors, respectively. These architectures are appropriate for applications with requirement of a low silicon area. Reference 5 is a bandgap reference that uses a PMOSFET acting as a diode to generate V REF. Further information regarding the designed voltage references can be found elsewhere ( Colombo et al., 2012; Banba et al., 1999; Ueno et al., 2009). These circuits were s imulated u nder TID effects.

337

Bias voltage subcircuit

Figure 5. Simulated voltage reference 3 (using a resistor rather than the MR1 NMOS transistor) and voltage reference 4 (using the MR1 NMOS transistor) (Ueno et al., 2009).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.335-340, Jul.-Sep., 2013

338

Both, T.H., Colombo, D., Dallasen, R.V. and Wirt, G.I.

M6 VDD

IBIAS

M11

M12

M10

VREF

RESULTS AND DISCUSSION The output voltage obtained for the five voltage references simulated under 0 krad (Si), 50 krad (Si), and 70 krad (Si) is presented in Figs. 7 to 11. The variation of the standard deviation as a function of temperature and total dose is presented in Tables 1 to 3.

RPTAT M14

gnd

0.324

Figure 6. Simulated voltage reference 5 (Colombo et al., 2012).

Voltage (V)

0.32 0.318 0.316 0.314 0.312 -40

-20

0 20 40 Temperature (ºC)

Figure 7. Output voltage for reference 1.

0.33

0 krad(Si) 50 krad(Si) 70 krad(Si)

Voltage (V)

0.328 0.326 0.324 0.322 -40

-20

0 20 40 Temperature (ºC)

Figure 8. Output voltage for reference 2.

0 krad(Si) 50 krad(Si) 70 krad(Si)

0.696 0.695 Voltage (V)

SIMULATIONS In order to evaluate total ionizing dose effects on each voltage reference, SPICE simulations were performed, applying threshold voltage shifts to each transistor of the circuit. Data regarding threshold voltage shifts due to total ionizing dose effects on 0.13 µm MOS transistors were obtained from (Haugerud et al., 2005). The voltage references were simulated for different bias conditions and under different temperature conditions. A total of 10,000 Monte-Carlo simulations were performed for each voltage reference. Process variability, provided in the IBM 0.13 µm commercial design kit, was also included in thesimulations, in an attempt to evaluate TID effects for different circuit conditions. For each of the 10,000 Monte-Carlo simulations, threshold voltage shifts due to process variability and also due to total ionizing dose effects were pseudo-randomly selected from normal distributions and assigned to each transistor of the circuit, meaning that for each simulation every transistor has its own threshold voltage shifts due to TID and also due to variability. It is important to state that data regarding total ionizing dose effects on 0.13 µm technology obtained from (Haugerud, et al., 2005) were not obtained for the 0.13 µm IBM technology used for integration and simulation. For references 1, 2, 3, and 4, DC simulations were performed for temperature conditions ranging from -40°C to 80°C, at a fixed supply voltage of 1.2 V. For reference 5, DC simulations were performed for temperature conditions ranging from -40 to 80°C, at a fixed supply voltage of 2.5 V. All simulations were performed while considering a pre-irradiation situation (0 krad) and under total doses of 50 krad (Si) and 70 krad (Si).

0 krad(Si) 50 krad(Si) 70 krad(Si)

0.322

0.694 0.693 0.692 0.691 0.69 0.689 -40

-20

0 20 40 Temperature (ºC)

Figure 9. Output voltage for reference 3.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.335-340, Jul.-Sep., 2013

Analysis of Total Ionizing Dose Effects on 0.13 μm Technology-temperature-compensated Voltage References

339

Table 1. Standard deviation of the output voltage for 0 krad (Si).

Voltage (V)

0.825

0 krad(Si) 50 krad(Si) 70 krad(Si)

0.82

0.815 -40

-20

0 20 40 Temperature (ºC)

Temperature

Voltage references

-25°C

0°C

25°C

50°C

1 (mV)

14.7

14.9

15.2

15.4

2 (mV)

14.6

14.9

15.2

15.6

3 (mV)

27.8

28.2

28.8

29.2

4 (mV)

32.0

32.3

32.7

33.1

5 (mV)

23.8

25.5

27.2

29.2

Figure 10. Output voltage for reference 4. Table 2. Standard deviation of the output voltage for 50 krad (Si).

Output voltage for Reference 5

Voltage (V)

1.48

0 krad(Si) 50 krad(Si) 70 krad(Si)

1.478 1.476

Temperature

Voltage references

-25°C

0°C

25°C

50°C

1 (mV)

18.0

18.2

18.5

18.8

2 (mV)

16.2

16.5

16.9

3 (mV)

28.1

28.6

29.1

29.6

1.474

4 (mV)

27.9

28.2

28.6

29.1

1.472

5 (mV)

25.5

21.2

29.0

31.0

-40

-20

0 20 40 Temperature (ºC)

80 Table 3. Standard deviation of the output voltage for 70 krad (Si).

Figure 11. Output voltage for reference 5.

The simulation results indicate that for the 0.13 µm technology voltage references, total ionizing dose is not a major concern. Compared with pre-irradiation simulations, both 50 and 70 krad simulations for the five voltage references showed shifts of a few millivolts in the average output voltage. The standard deviation compared with the pre-irradiation, which considers only process variability, increased by a few millivolts under total ionizing dose for voltage references 1, 2, 3 and 5. On the other hand, voltage reference 4 experienced a decrease of the standard deviation under TID. This reduction was also observed when threshold voltage shifts due to total ionizing dose were applied only to the NMOS transistor MR1, as shown in Fig. 5. The mechanism responsible for this reduction of the standard deviation, however, was not identified. The observed overall reduction in the standard deviation between 50 krad and 70 krad was expected due to the fact that the standard deviation of the threshold voltage shift data from Haugerud et al., (2005) is reduced between 50 krad and 70 krad, regardless of the fact that the average shift increases.

Temperature

Voltage references

-25°C

0°C

25°C

50°C

1 (mV)

16.6

16.7

17.0

17.2

2 (mV)

15.6

15.9

16.2

16.5

3 (mV)

28.0

28.5

29.0

29.5

4 (mV)

27.8

28.1

28.5

28.9

5 (mV)

24.7

26.3

28.1

30.0

The average output voltage for all voltage references also presented a small variation between 50 and 70 krad compared with the variation observed between 0 rad and 50 krad. This is expected due to similar threshold voltage shifts for such doses. The results obtained in this work indicate that these voltage references are functional for doses of approximately 70 krad (Si), and are hence suitable for most space applications. The standard deviation increase of the output voltage due to TID effects, however, should also be considered in o rder to guarantee proper circuit operation. It is necessary to observe, however, that the data regarding

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.335-340, Jul.-Sep., 2013

340

Both, T.H., Colombo, D., Dallasen, R.V. and Wirt, G.I.

threshold voltage shifts due to TID, employed in simulation, were not obtained for the 0.13 µm IBM technology. Further irradiation tests are necessary to confirm the simulation results. In addition, the simulations performed in this work did not take circuit layout into consideration. This is an important issue considering that circuit layout techniques, such as the layout of an enclosed gate for NMOS transistors to reduce leakage current beneath the bird’s beak region (Mavis and Alexander, 1997), affect radiation tolerance.

CONCLUSIONS The five voltage references simulated in this work presented tolerance to total ionizing dose effects for

doses of approximately 70 krad (Si). For the simulated 0.13 μm technology, parametric shifts due to TID e ffects resulted in shifts of a few millivolts in the output voltage of voltage references, and in the standard deviation of the output voltage. Despite the positive results in simulations, irradiations tests are required, as circuit layout, for instance, was not accounted for during simulation.

ACKNOWLEDGMENTS The authors would like to acknowledge the support rendered by CNPq, FAPERGS, and AEB for partially funding this research.

REFERENCES Banba, H., Shiga, H., Umezawa, A., Miyaba, T., Tanzawa, T., Atsumi, S. and Sakui, K., 1999, “A CMOS bandgap reference circuit with sub-1-V operation”, IEEE Journal of Solid-State Circuits, Vol. 34, pp. 670-674. Colombo, D., Fayomi, C., Nabki, F., Ferreira, L.F., Wirth, G. and Bampi, S., 2011, “A Design Methodology Using the Inversion Coefficient for Low-Voltage Low-Power CMOS Voltage References”, Journal of Integrated Circuits and Systems, Vol. 6. Colombo, D., Werle, F., Wirth, G. and Bampi, S., 2012, “A CMOS 25.3 ppm/°C bandgap voltage reference using self-cascode c omposite transistor”, IEEE Third Latin American Symposium on C ircuits and Systems (LASCAS), pp. 1-4. Fleetwood, D.M. and Scofield, J.H., 1990, “Evidence that similar point defects cause 1/f noise and radiation-induced-hold trapping in m etaloxide-semiconductor transistors”, Physical Review Letters, Vol. 64, pp. 579-582.

Karnik, T. and Hazucha, P., 2004, “Characterization of soft errors caused by s ingle event upsets in CMOS processes”, IEEE Transactions on D ependable and Secure Computing, Vol. 1, pp. 128-143. Mavis, D.G. and Alexander, R., 1997, “Employing radiation hardness by design techniques with commercial integrated circuit processes”, Digital Avionics Systems Conference, 16th DASC., AIAA/IEEE, Vol. 1, pp. 15-22. Meisenheimer, T.L. and Fleetwood, D.M., 1990, “Effect of RadiationInduced Charge on 1/f Noise in MOS Devices”, IEEE Transactions on Nuclear S cience, Vol. 37. Oldham, T.R. and McLean, F.B., 2003, “Total Ionizing dose effects in MOS oxides and devices”, IEEE Transactions on Nuclear Science, Vol. 50, pp. 483-499. Sexton, F.W., 2003, “Destructive single-event effects in s emiconductor devices and ICs”, IEEE Transactions on Nuclear Science, Vol. 50, pp. 603-621.

Fleetwood, D.M. Meisenheimer, T.L. and Scofield, J.H., 1994, “1/f noise and radiation effects in MOS devices”, IEEE Transactions on Electron Devices, Vol. 41, pp. 1953-1964.

Srour, J.R., 2003, “Review of displacement damage effects in silicon devices”, IEEE Transactions on Nuclear Science, Vol. 50, pp. 653-670.

Haugerud, B.M. Venkataraman, S., Sutton, A.K., Prakash, A.P.G., Cressler, J.D., Guofu Niu, Marshall, P.W. and Joseph, A.J., 2005, “The impact of substrate bias on proton damage in 130 nm CMOS technology”, IEEE Radiation Effects Data Workshop Proceedings, pp. 117-121.

Ueno, K., Hirose, T., Asai, T. and Amemiya, Y., 2009, “A 300 nW, 15 ppm/°C, 20 ppm/V CMOS Voltage Reference Circuit Consisting of Subthreshold MOSFETs”, IEEE Journal of Solid-State Circuits, Vol. 44, pp. 2047-2054.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.335-340, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.236

Nonlinear Characteristics of Revolute Joints with Clearance Liu Rong-qiang1, Zhang Jing1, Guo Hong-wei1, Deng Zong-quan1

ABSTRACT: The nonlinear contact characteristic of revolute joint with clearance can degrade the performance of deployable structure. In this article, tensile and compressive tests are adopted to investigate the accuracies of the simplest conformal model and nonconformal model for contact, which are used to calculate the revolute joint deformation. In order to study the applicability of the two models, different clearances of joints are introduced in the tests. The results of the two contact models do not well agree with the experimental results. A model for the calculation of contact force and deformation of revolute joint, considering geometric constraints and cylinder contact characteristic, is presented. By comparison with the simplest conformal model and nonconformal model for contact, the proposed model is more accurate to calculate the nonlinear contact of revolute joint. The theoretical and experimental analysis of revolute joint with clearance is helpful to improve the reliability of deployable structure simulation. KEYWORDS: Nonlinear stiffness, Revolute joint, Clearance, Conformal contact.

INTRODUCTION The deployable structures have always been used to reduce the packaging volume of the spacecraft. Many appendage structures (Larsen et al., 2009; Meguro et al., 2009), such as communication antennas, have been successfully deployed in space. The truss elements of deployable structures are constructed by revolute joints. In order to achieve connection and large-scale rotation of truss elements, there are small clearances in the joints. It affects the performance of deployable structures and causes noise, wear, and vibrations (Ingham and Crawley, 2001; Schwab et al., 2002; Parenti-Castelli and Venanzi, 2005; Flores et al., 2006a,b, 2010; Qi et al., 2010). The presence of clearance in the revolute joints often degrades the structure stiffness and has a significant effect on the dynamic behavior of deployable structures. Formulation for the contact of clearance joint is very important for researchers to study the nonlinear dynamics of structures and simulate the multibody system precisely (Liu et al., 2006). So it is necessary to analyze and calculate the stiffness of clearance joint. The heart of the revolute joint formulation is the intermittent and continuous contact simulation of joints. Various types of continuous models have been proposed by many researchers. There are two main types of contact models – Kelvin-Voigt viscoelastic model and Hertz contact model (Johnson, 1985), which are based on elasticity theory. The models have been applied in joint deformation calculation and impact simulation. Dubowsky and Freudenstein (1971) proposed deformation expression of a pin inside a cylinder. Energy dissipation characteristic of impact is simplified into the product of the rebound force and the coefficient of restitution (Bengisu et al., 1986; Ravn, 1998). Rhee and Akay (1996) studied the

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

342

Rong-qiang, L., Jing, Z., Hong-wei, G. and Zong-quan, D.

response of four-bar mechanism with clearance joint, including the impact force and friction. Tian et al., (2009) deduced the lubricant force in lubricated joints based on Reynolds’ equation. In these plane mechanism studies, the Hertz law expression about two spheres has been used to calculate the contact force in revolute joint (Lankarani and Nikravesh, 1990; Jia et al., 2002; Khemili and Romdhane, 2008; Shi and Jin, 2008; Erkaya and Uzmay, 2012; Olyaei and Ghazavi, 2012). It is difficult to obtain accurate contact deformation and load of two cylinders by using sphere contact model and nonconformal contact model. So the rationality of the nonconformal contact model and conformal contact model has to be testified. The main methods of building the contact model are the elastic theory and discrete method. Based on the contact force model with hysteresis damping given by Lankarani and Nikravesh (1990), Flores and Ambrosio (2004), and Flores et al., (2006a,b) gave the normal contact model of planar revolute joints and compared different contact models. The contact area and load can be separated into its normal and tangential components (Abdo and Shamseldin, 2005). Discrete method is usually achieved by using finite element method (FEM), which has been used to give the force–displacement relation of sphere and the parameters influence on contact stress (Zhang and Vu-Quoc, 2001; Knight et al., 2002). However, it will take a long time to simulate the contact of joint with FEM. Because a simple and accurate calculation method has not yet been developed, the particular focus in this article lies on the evaluation and formulation of the relationship among contact force, depth, and width of revolute joint, which is based on joint experiment, geometric constraint, and elastic theory. The layout of this article is as follows: in Hertz model that has been widely used to calculate the revolute joint contact in mechanism and conformal model that has been most studied are presented. In revolute joint experiment is achieved. The results of the two models are compared with the experimental results when the clearance size changes. The new model of revolute joint with clearance is proposed, which is more precise than Hertz model and conformal model in a number of cases.

in revolute joints to allow relative rotation between the clevis and tang. The contact force and the deformation of revolute joint with clearance are usually calculated by Hertz model. In this section, the contact model of two cylinders is presented first, which is based on plane strain theory. Figure 1b depicts the contact of revolute joint, in which the radial clearance, ΔR, is defined as difference between the pin and tang radius, R1 and R2, respectively. It is assumed that the two contact bodies are isotropic materials. The Young’s modulus and Poisson’s ratio of tang and pin are represented by E1, E2, υ1, and υ2, respectively, a is the contact width and ε is the semiangle of contact corresponding to the contact relationship. According to the Hertz laws for two-dimensional contact of cylindrical bodies (Johnson, 1985), two cylinder contact can be simplified from the three-dimensional contact into two-dimensional contact when their axes parallel. The relationship of contact force and width is obtained based on the constraints of contact. The sign of the curvature is related with the shape of the contact surface. When the contact surface is concave or saddle shaped, the radius curvature sign of the contact solid is negative. So the relative radius of the pin and the tang contact can be expressed as R=

The equivalent Young’s modulus can be written 1 1- u 1- u = + E E1 E2 2

(2)

The contact pressure distribution is given by

p( x ) =

2P 2 2 12 (a −x ) (3) π a2

Clevis Pin Tang

Two existing models for revolute joint with clearance Nonconformal contact model The revolute joint is composed of clevis, tang, and pin, shown in Fig. 1a. Some amount of clearance always exists

R1 R2 R2 - R1 (1)

O R2

ΔR

R1 ε a

(a)

(b)

Figure 1. Model of revolute joints with clearance: (a) The revolute joint (Lake and Lee, 1996) and (b) contact model for revolute joint with clearance.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

Nonlinear Characteristics of Revolute Joints with Clearance

πa2 E P= 4R (4) Because a can be expressed by δ , as

a = δ R (5) where δ is the mutual approach of distant points on the pin and the hole of the tang. So the contact load can be expressed by the contact depth

P = π Eδ (6) 4 In Hertz laws (Johnson, 1985), three assumptions need to be satisfied: surfaces are continuous, nonconforming, and frictionless; the strain is small; each solid could be considered as an elastic halfspace. However, the contact pair of revolute joints with clearance has conformal surfaces that cannot satisfy the assumption of Hertz theory. So the conformal model must be considered. Conformal contact model In the previous part, the nonconformal contact model is expressed to calculate the contact force and penetration depth. Because the clearance in revolute joint is small, the contact of revolute joint is conformal contact. The normal contact of conformal surfaces has also been considered by Steuermann and Persson (Johnson, 1985). The profiles are represented as a polynomial to achieve a required degree approximation. So the initial separation of the contact bodies’ boundary can be expressed as h = A1x 2 + A2 x 4 + A3 x 6 + ... + An x 2n + ... (7)

The corresponding Hertz approximation can also be used to express the compression by δ=

( R1 - R1 1

(

where P is the contact load under the per unit length of the contact cylinder. The contact load P can be expressed as (Johnson, 1985)

343

a 2 (12)

The total load can be written

Pn =

4nEaδ 2n + 1 (13)

It is clear in Eq. 13 that the stiffness of the contact pair is increased when the number of the polynomial is improved. So the simplest conformal model can be helpful to evaluate the error of other number conformal model. Based on Eqs. 12 and 13, the contact load can be written 4E 32 ⎛ 1 1 ⎞ P= δ ⎜ − ⎟ 3 ⎝ R1 R2 ⎠

−

1 2

(14)

Revolute joint experiment and comparison The experiment, which is shown in Fig. 2, is set up to evaluate the accuracy of nonconformal model and conformal model in calculating the stiffness of revolute joint. The joint was fixed between the base and the moving beam. The moving beam moved in the vertical direction. So the load applied on the joint was also in the vertical direction. The displacement of the moving beam and force along the joint can be collected by encoder and sensor, which correspond to the variable δ and P. The Young’s modulus of revolute joint is 2.06×105 MPa. The experiment introduces three joints with different radial clearances which are 0.067, 0.094, and 0.104 mm, respectively. The contact length is taken as one

The external normal load and compression between centers of the contact bodies can be expressed as Pn =

4An Ena 2n + 1 2 ⋅ 4 ⋅⋅⋅ 2n (2n + 1) 1 ⋅ 3 ⋅⋅⋅ (2n − 1)

2 ⋅ 4 ⋅⋅⋅ 2n δ= Aa 1 ⋅ 3 ⋅⋅⋅ ( 2n − 1) n

(8) Moving Beam

(9)

The second-order profiles assumed in this model is corresponding to n=1 in the Hertz theory. So the simplest initial separation in the revolute joint contact can be given by h = A1x 2 (10)

Joint Moving Direction

Base

(a) Revolute joint installation

The compression can be expressed as δ = 2A1a 2 (11)

(b) Revolute joint

Figure 2. Revolute joint experiment.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

344

Rong-qiang, L., Jing, Z., Hong-wei, G. and Zong-quan, D.

unit length for the comparison of the experimental data with other model results, because the coaxially usually cannot

Based on Eqs. 15 and 16, the distance of points having the same x coordinate separately in different circles can be written as

be guaranteed in experiment. The comparisons of the joint

.. θ . cosθθ2 − hh=h=-= Δ - ΔR-R Δ+R+R+R2.2R. 2cos 2 −RR 1 1 cos 1 (17)

deformation among nonconformal and conformal model results with experimental results demonstrate the suitability of the models, shown in Fig. 3. As shown in Fig. 3, it is found that the results based on

Because the two points G1 and G2 have the same x coordinate, θ1 and θ2 satisfy

R1 . sin 1 = R2 . sin

nonconformal model and conformal model are not close to

(18)

3 5 ×10

the experimental results. The average errors of the two m odels are more than 50% and 30% separately. The nonconformal

Hertz Theory Conforming Model Experiment

model results based on Hertz theory have large errors in small clearance and small errors in large clearance. However,

P (N)

conformal model is closer to the experiment results compared

with Hertz model when the clearance is small. Because the Hertz model is base on nonconformal contact, it is more

applicable to large clearance contact. The conformal model is

0 0.01

more suitable for small clearance contact.

0.02

0.03 δ (mm)

0.04

(a) ΔR=0.067 mm

Nonlinear contact model based on the geometric constraint

4.5

In the previous comparison, it is obvious that Hertz theory cannot solve the contact of revolute joint with clearance exactly,

×10

Hertz Theory Conforming Model Experiment

3.5

which is based on the assumption of nonconformal contact and P (N)

the simplest conformal equation. So the dynamics simulation of multibody system cannot be performed conveniently and accurately. A new contact model is proposed in this section and

2.5 1.5

is proved by comparing with above models. 0.5

Geometric constraints

0.015 0.02 0.025 0.03 0.035 0.04 δ (mm)

(b) ΔR=0.094 mm

To find the relationship between the contact depth and the width, the new model must satisfy the geometric constraint.

2.5

Based on the model of Fig. 1, the contact deformation and other assistant parameters are shown in Fig. 4.

1.5

P (N)

the pin and hole can be written as:

⎧⎪ x1 = R1 . sinθ1 ⎨ (15) ⎪⎩ y1 = R1 . cosθ1

0.5 0

(16)

where θ1 and θ2 are the center angles of the G1 and G2.

Hertz Theory Conforming Model Experiment

The coordinates of the points G1 and G2 in the border of

⎧⎪ x 2 = R2 . sin θ2 ⎨ ⎪⎩ y 2 = −ΔR + R2.cosθ2

×10

0.008

0.012

0.016 0.02 δ (mm)

0.024

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

Nonlinear Characteristics of Revolute Joints with Clearance

Considering Eq. (18) and trigonometric relationship sin q2 + cos 2 q2 = 1 , h in Eq. (17) can be redefined into 2

θ1 = ∆ +

2 2

R R11 . .



. .

− θ1

− 

R22 R

θ1



(19)

When θ1 is equal to contact semiangle ε, h(θ1) corresponds to the distance of the edge points H1 and H2. Due to the deformation of the hole of the tang and the pin, the total displacement of the tang is equal to δ and h(ε), which can be found in Fig. 4. Based on Eq. 19, δ can be written as 2

 R1 .



(20)

345

Based on the relationship of contact width with pin radius and contact semiangle a = R1.sinε which is shown in Fig. 4, Eq. 20 can be expressed as δ = −ΔR + R22 − a 2 − R12 − a 2

(21)

Eq. 21 can also be expressed as

(δ + ΔR )(

R22 − a 2 + R12 − a 2

)

R22 − R12

(22)

So 2 2

R − R12 = R22 − a 2 + R12 − a 2 (23) (δ + ΔR ) 2 2 By adding the Eq. 23 to Eq. 21, R1 − a can be e liminated.

The contact width a can be given by a = R22 −

 1  R22 −R12  + δ +∆R   4  δ + ∆R 

(24)

Approximate model for revolute joint with clearance The difficulty of establishing the contact model of revolute

The Hole of theTang Deformation of the Hole

δ ΔR

joints with clearance is how to determine the distribution of contact pressure and stress in the contact body when the geometry constraint is satisfied. In this article, the research

is limited to the case that deformations of contact bodies are

O1 O

elastic and there is no friction. Based on the assumptions that:

the shape of the contact area of the pin and role satisfies the

θ2 θ1 G1

H2 h

geometric constraint given by Eq. 24; based on the half-space

theory, the inner stress distribution of contact bodies has

been considered; the contact pressure is assumed as ellipse distribution of Hertz theory. The deformation of the pin can

(a) O2 O1

be given by (Johnson, 1985)

δ b = 2P

∆R

Because the clevis and tang contact with pin at relative

Deformation of the Hole

θ2

direction, the total deformation of contact pair of clevis and pin is

The Hole of theTang

θ1

δ = 2δ b (26)

G1 G2

1 −ν 2   4R   2ln  1  1 (25) πE   a  

Substituting Eq. 21 into Eq. 26, the contact load on per

H2 h(e ) = δ h

unit length of cylinder can be expressed as

(b)

Figure 4. The geometry model of revolute joint with clearance: (a) integral geometry model and (b) A quarter of the geometry model.

−ΔR + R22 − a 2 − R12 − a 2 4 (1 − ν ) ( 2ln ( 4R1 / a ) − 1) (27)

πE

The relationship of contact depth and force can be expressed by

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

346

Rong-qiang, L., Jing, Z., Hong-wei, G. and Zong-quan, D.

π Eδ ⎛⎜ 2ln ⎛⎜ 4R1 / R 2 - 1 ⎛⎜ R22 -R12 +δ +ΔR ⎞⎟2 ⎞⎟ -1 ⎞⎟ P= 2 2 ⎜ ⎟ ⎟ ⎜ 4 (1 -ν

) ⎜⎝

⎝

4 ⎝ δ +ΔR

⎠

Figure 6 shows the penetration depth of simplest conformal

−1

⎟ (28) ⎠ ⎠

and nonconformal contact when the radial clearance is 0.067 mm. It is clearly shown that the nonconformal and simplest conformal contact depth cannot be very large, because the

Comparisons of three models and experiment

contact depth has linear relationship with the square of

for revolute joints with clearance To verify the new contact model for revolute joint, the comparisons among the Hertz model, conformal model, and new model (experiment) are conducted. The results are presented in Table 1. The normal load P versus the normal displacement δ curve is shown in Fig. 5 which is obtained from the Hertz model given by Eq. 6, new model expressed by Eq. 28, simplest conformal model

contact width. The maximum displacement increases rapidly in the new model when the contact width is close to the radius of the surface. However, the penetration depth of conformal model and corresponding Hertz model always rise steadily with the contact width increase and is hard to reach a large value when the contact width approaches to the radius of the contact surface.

written as Eq. 14, and experimental data. Observed from Fig. 5, it is found that the new contact model is closer to the experimental results than Hertz model and conformal model. The error of Hertz model

Conclusions and future works

decreases with the increase of the clearance. The errors of simplified conformal model and new model are small

This article investigates the conformal contact model

compared with Hertz model when the clearance is small.

of revolute joint with clearance. The revolute joint

Because the contact of clearance revolute joint is close to

experiment is conducted to evaluate the nonconformal

conformal contact, the new model is more suitable for

and the simplest polynomial fitting conformal model. By

calculating the contact of revolute joint.

comparing the two models with experimental results, it

Table 1. The comparison of the Hertz and conformal model with experimental results. ΔR (mm)

0.067

0.094

0.104

δ (×10–2mm)

Experiment

Hertz theory

Conformal model

p -value (N)

Relative error (%)

p -value (N)

Relative error (%)

0.981

500

871

74.20

494

-1.20

1.971

1000

1751

75.10

1410

41.00

3.253

2000

2889

44.45

2988

49.40

4.484

3000

3983

32.77

4836

61.20

1.156

500

1026

105.20

527

5.40

2.308

1000

2054

105.40

1502

50.20

3.232

2000

2871

43.55

2483

24.15

4.484

3000

3983

32.77

4056

35.20

0.734

500

653

30.60

252

-49.60

1.436

1000

1275

27.50

696

-30.40

2.512

1640

2231

36.04

1607

-2.01

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

Nonlinear Characteristics of Revolute Joints with Clearance

80 60 50 40

δ /mm

Relative Error of P (%)

0.9 0.8 0.7 0.6

Hertz Theory Conformal Model New Model

30 20 10 0 -10

(a) ΔR=0.067 mm

Relative Error of P (%)

120

60 40 20

Relative Error of P (%)

40 30

0.01156 0.02308 0.03232 δ (mm)

0.04484

(b) ΔR=0.094mm

20 10 0 - 10 - 20 - 30

Hertz Theory Conformal Model New Model

- 40 - 50

0.00734

0.01436

1.5 a/mm

2.5

is found that the Hertz model is just suitable to solve the revolute joint contact when the contact load is not very large. The stiffness calculated by the simplest conformal model is higher than the nonconformal model and it is more accurate when the clearance is not small. The new model is built to solve the conformal contact problem which is based on geometric restriction, the half space theory, pressure distribution of Hertz theory, elastic deformation, and frictionless. When the clearance is large, the errors of conformal model and new model are close to each other by comparing these models with experiment. The error of new model is the smallest when the clearance is not very large. And it has not changed rapidly with the increase of clearance. So the new model is more simple and reliable to calculate the contact of revolute joint. Future work focuses on the dynamic characteristics of the revolute joints. The friction and damping will be considered. The corresponding dynamic experiment will be conducted to get the dynamic characteristic.

0.5

Figure 6. Relation of the contact width and the penetration depth.

Hertz Theory Conformal Model New Model

100

Hertz Theory New Model

0.5 0.4 0.3 0.2 0.1 0

0.00981 0.01971 0.03253 0.04484 δ (mm)

347

0.02512

δ (mm)

Acknowledgments

The authors acknowledge the support from the Natural Science Foundation of China (Project No. 50935002&11002039).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

348

Rong-qiang, L., Jing, Z., Hong-wei, G. and Zong-quan, D.

References Abdo, J. and Shamseldin, E., 2005, “Modeling of Contact Area, Contact Force, and Contact Stiffness of Mechanical Systems with Friction”, 2005 ASME International Mechanical Engineering Congress and Exposition, Orlando, pp. 1-8. Bengisu, M.T., Hidayetoglu, T. and Akay A., 1986, “A Theoretical and Experimental Investigation of Contact Loss in the Clearances of a FourBar Mechanism”, ASME, Vol. 108, pp. 237-244. Dubowsky, S. and Freudenstein, F., 1971, “Dynamic analysis of mechanical systems with clearances, Part 1: Formulation of dynamic model”, Journal of Engineering for Industry, Vol. 93, pp. 305-309. Erkaya, S. and Uzmay, I., “Effects of balancing and link flexibility on dynamics of a planar mechanism having joint clearance”, Sharif University of Technology Scientia Iranica, doi: 10.1016/j. scient.2012.04.011. Flores, P. and Ambreósio, J., 2004, “Revolute joints with c learance in multibody systems”, Computers and Structures, Vol. 82, pp. 13591369. Flores, P., Ambrósio, J., Claro, J. C. P. and Lankarani, H. M, 2006a, “Dynamics of Multibody Systems with Spherical Clearance Joints”, Transactions of the ASME, Vol. 1, pp. 240-247. Flores, P., Ambrósio, J., Claro, J.C.P., Lankarani, H.M. and Koshy, C.S., 2006b, “A Study on Dynamics of Mechanical Systems Including Joints with Clearance and Lubrication”, Mechanisms and Machine Theory, Vol. 41, pp. 247-261. Flores, P., Leine, R. and Glocker. C., 2010, “Modeling and Analysis of Planar Rigid Multibody Systems with Translational Clearance Joints Based on the Non-smooth Dynamics Approach”, Multibody System Dynamics, Vol. 23, pp. 165-190. Ingham, M.D. and Crawley E.F., 2001, “Micro Dynamic Characterization of Modal Parameters for a Deployable Space Structure”, AIAA, Vol. 39, pp. 331-338. Jia, X., Jin, D., Ji, L. and Zhang, J., 2002, “Investigation on the Dynamic Performance of the Tripod-ball Sliding Joint with Clearance in a Crank-sliding Mechanism. Part 1. Theoretical and Experimental Results”, Journal of Sound and Vibration, Vol. 252, pp. 919-933. Johnson, K.L., 1985, “Contact Mechanics”, Cambridge University Press, Cambridge, 99 p. Khemili, I. and Romdhane, L., 2008, “Dynamic analysis of a flexible slider-crank mechanism with clearance”, European Journal of Mechanics A/Solids, Vol. 27, pp. 882-898. Knight, M.G., de Lacerda, L.A., Wrobel, L.C. and Henshall, J.L., 2002, “Parametric Study of the Contact Stresses Around Spherical and C ylindrical Inclusions”, Computational Materials Science, Vol. 25, pp. 115-121. Lake, M.S. and Lee, D., 1996, “A revolute joint with linear loaddisplacement response for precision deployable structures”, The 37th

AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Salt Lake City, pp. 1-14. Lankarani, H.M. and Nikravesh, P.E., 1990, “A contact force model with hysteresis damping for impact analysis of multibody systems”. Journal of Mechanical Design, Vol. 112, pp. 369-376. Larsen, J.J., Jenkins, C., Banik, J. and Murphey, T., 2009, “Continuum Structural Representation of Flexure and Tension Stiffened 1D Spacecraft Architectures”, 50th AIAA/ASME/ASCE/AHS/ ASC Structures, Structural Dynamics, and Materials Conference, California, pp. 1-13. Liu, C.S., Yang, L. and Zhang, K., 2006, “Normal Force-Displacement Relationship of Spherical Joints with Clearances”, Transactions of the ASME, Vol. 1, pp. 160-167. Meguro, A., Shintate, K., Usui, M and A. Tsujihata., 2009, “In-orbit Deployment Characteristics of Large Deployable Antenna Reflector Onboard Engineering Test Satellite VIII”, Acta Astronautica, Vol. 65, pp. 1306-1316. Olyaei, A.A. and Ghazavi, M.R., 2012, “Stabilizing slider-crank mechanism with clearance joints”, Mechanism and Machine Theory, Vol. 53, pp. 17-29. Parenti-Castelli, V. and Venanzi, S., 2005, “Clearance Influence Analysis on Mechanisms”, Mechanism and Machine Theory, Vol. 40, pp. 1316-1329. Qi Z.H., Xu, Y., Luo, X. and Yao, S., 2010, “Recursive Formulations for Multibody Systems with Frictional Joints Based on the Interaction Between Bodies”, Multibody System Dynamics, Vol. 24, pp. 133166. Ravn, P., 1998, “Continuous Analysis Method for Planar Multibody Systems with Joint Clearance”. Multibody System Dynamics, Vol. 2, pp. 1-24. Rhee, J. and Akay, A., 1996, “Dynamic response of a revolute joint with clearance”, Mechanisms and Machine Theory, Vol. 31, pp. 121-134. Schwab, A.L., Meijaard, J.P. and Meijers, P., 2002, “A Comparison of Revolute Joint Clearance Models in the Dynamic Analysis of Rigid and Elastic Mechanical Systems”, Mechanism and Machine Theory, Vol. 37, pp. 895-913. Shi, B. and Jin, Y., 2008, “Dynamic analysis of the reheat-stop-valve mechanism with revolute joint in consideration of thermal effect”, Mechanism and Machine Theory, Vol. 43, pp. 1625-1638. Tian, Q., Zhang, Y., Chen, L. and Flores, P., 2009, “Dynamics of spatial flexible multibody systems with clearance and lubricated spherical joints”, Computers and Structures, Vol. 87, pp. 913-929. Zhang, X., and Vu-Quoc, L., 2001, “Modeling the Dependence of the Coefficient of Restitution on the Impact Velocity in Elasto-Plastic Collisions”, International Journal of Impact Engineering, Vol. 27, pp. 317-341.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.341-348, Jul.-Sep., 2013

doi: 10.5028/jatm.v5i3.223

Optimal Design of a High-Altitude SolarPowered Unmanned Airplane Bento Silva de Mattos1, Ney Rafael Secco1, Eduardo Francisco Salles1

ABSTRACT: This paper describes a multi-disciplinary design and optimization framework tailored for the conceptual development of high-altitude solar-powered unmanned aerial vehicles. The aircraft baseline configuration that the framework is able to handle is very similar to that of Zephyr, which is developed by the UK based company QinetiQ. The disciplines of aerodynamics, structures, stability, weight, and systems were considered and integrated into a modeFrontier® workflow, capable of providing a relatively simple sizing, but highly realistic airplane. Keywords: Airplane design, Solar energy, Multi-disciplinary design and optimization, Airplane stability and control.

INTRODUCTION The present work is concerned with the optimal design of unmanned high-altitude long-endurance (HALE) solar-powered airplanes. In order to accomplish this, a multi-disciplinary design framework was elaborated employing the modeFrontier® (ESTECO, 2011) commercial optimization package. A brief technical retrospective of the development on solar-driven aircraft is provided in this section as well as the reasoning to address the utilization of such kind of airplane. The first flight of a solar-powered aircraft took place on November 4th, 1974, when the remotely controlled Sunrise I, designed by Robert J. Boucher of AstroFlight, Inc., flew after a catapult launch (Noth, 2008). The flight lasted 20 minutes at an altitude of around 100 m (Noth, 2008). AeroVironment, Inc. was founded in 1971 by the ultralight airplane innovator Paul MacCready. Following the AstroFlight’s airplane debut, AeroVironment undertook a more ambitious project to design a human-piloted, solarpowered aircraft. The firm initially based its design on the human-powered Gossamer Albatross II and scaled it down to three-quarters of its previous size for solar-powered flights. This was easier to accomplish because, in early 1980, the Gossamer Albatross aircraft had participated in a flight research program at NASA Dryden Research Center. Some of the Albatross flights were performed employing a small electric motor. The scaleddown aircraft was designated Gossamer Penguin (Fig. 1). It had a 71-ft wingspan compared to the 21.64-m (96-ft) span of the Gossamer Albatross (MacCready et al., 1983). Penguin’s Empty Weight was just 31 kg, it had a low-power requirement and, therefore, it was an excellent test bed for testing of solar power systems (Noth, 2008).

1.Instituto Tecnológico de Aeronáutica – São José dos Campos/SP – Brazil Author for correspondence: Bento Silva de Mattos | Instituto Tecnológico de Aeronáutica | Praça Marechal Eduardo Gomes, 50 – Vila das Acácias | CEP 12.228-900 São José dos Campos/SP – Brazil | Email: baviator@gmail.com Received: 17/12/12 | Accepted: 10/04/13

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

350

Mattos, B.S., Secco, N.R. and Salles, E.F.

Figure 1. The first piloted solar plane, the Gossamer Penguin.

AstroFlight, Inc., of Venice, California, provided the power plant for the Gossamer Penguin, an Astro Cobalt 40 electric motor. This engine was designed to run at between 10,000 and 15,000 rpm, but the 11-ft diameter propeller of Penguin rotates between 120 and 130 rpm. The speed reduction was achieved with a three-stage gearbox with two timing pulley stages providing a reduction ratio of 5/1 each and a final chain reduction stage delivering a 5.12/1 ratio. Robert Boucher, designer of the Sunrise II, worked as a key consultant for both aircraft and for the Solar Challenger. The power source for the initial flights of the Gossamer Penguin consisted of 25 nickelcadmium batteries and panels of 3,640 solar cells delivering 541W. The battery-powered flights took place at Shafter Airport near Bakersfield, California. Paul MacCready’s son, Marshall, was the test pilot. He was 13 years old and weighed approximately 80 lb (36 kg) at that time. The main targets of the test flights were the determination of required power to fly the airplane, the optimization of the airframe/propulsion system, and pilot training. Flights took place on April 7th, 1980, and a short solar-powered one on May 18th. However, the official project pilot was Janice Brown, a Bakersfield school teacher who weighed in at slightly less than 45 kg and was a charter pilot with commercial, instrument, and glider ratings. She checked out in the plane at Shafter and made about 40 solar-powered flights. Wind direction, turbulence, convection, temperature and radiation at Shafter in midsummer proved to be less than ideal for Gossamer Penguin because of crosswinds at takeoff and decreased solar power

output due to higher ambient temperatures. Consequently, the project moved to Dryden in late July, although conditions there also were not ideal. Nevertheless, Janice finished the testing, and on August 7th, 1980, she flew a public demonstration at Dryden in which she was able to cover 3.14 km in 14 minutes and 21 seconds. This was significant as the first sustained flight of an aircraft relying solely on direct solar power rather than batteries. It provided the designers with practical experience for developing a more efficient solar-powered aircraft, taking into account that the Gossamer Penguin was fragile and had limited controllability. The Penguin flew preferably in the morning, when usually wind is minimal and consequently the turbulence level is still low. On the other hand, at this part of the day, the angle of the sun was low, requiring a tilting solar panel to capture properly the solar radiation (Noth, 2008). From the lessons learned with Gossamer Penguin, AeroVironment engineering staff designed Solar Challenger, a piloted, solar-powered aircraft that was capable of highaltitude and long-endurance flights (Noth, 2008). The last big step in aircraft powered by solar energy is the first flight of twin-engine Solar Impulse airplane on December 3rd, 2009 (Solar Impulse, 2013). Solar Impulse is a long-range solar powered plane project currently being undertaken at the École Polytechnique Fédérale in Lausanne, Switzerland. The project is promoted by Bertrand Piccard, who piloted the first balloon that was able to circumnavigate the world nonstop. This project is expected to repeat that milestone in aviation history by employing a solar-powered airplane. The first aircraft, officially named HB-SIA, is able to accommodate a single pilot, capable of taking off under its own power, and is intended to remain airborne up to 36 hours. Building on the experience of this prototype, a slightly larger follow-on design (HB-SIB) is planned to make circumnavigation of the globe in 20-25 days. On April 7th, 2010, the HB-SIA underwent an extended 87 minutes test flight (Fig. 2). In contrast to earlier tests, the April flight reached an altitude of 1,200 m (3,937 ft). In February 2012, Solar Impulse pilot Andre Borschberg spent more than 60 hours straight at the controls of the Solar Impulse flight simulator (Fig. 3). Granted, he was able to get some sleep, sometimes napping for a whole 20 minutes at a time (Solar Impulse, 2013). At that time, Borschberg was approaching the end of a 72-hour stint in the simulator, running through a series of tests and challenges to prepare for what lies ahead

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane

351

Figure 2. Solar Impulse flight on April 7th, 2010.

simultaneously (Romeo and Frulla, 2004). This means a user community of 8.5 million per unit (although this does not take into account data transmission). The UAV platform of major interest to the present work is one that is capable to reach stratospheric altitudes and long endurance flights (HALE category) and powered by solar energy. In this context, the Zephyr, developed by the UK based QinetiQ (QinetiQ, 2013), is a configuration that is very similar to that envisaged by the authors. The airplane carries a small payload and has no need for sophisticated assisted takeoff systems or prepared airfields (Fig. 4). The Zephyr is a lightweight airplane

Figure 3. Solar Impulse flight simulator.

Figure 4. QinetiQ’s Zephyr during an assisted take off run (Photo: QinetiQ, 2013).

when he attempts to fly around the world in a solar airplane in 2014 (Solar Impulse, 2013). In recent years, European countries are taking a more central role in the development of High Altitude/Long Endurance solar-powered unmanned aerial vehicles (UAVs). Several projects are either ongoing or being planned for advanced solar powered vehicles. In the 2000–2003 period, a team from the Politecnico de Torino, in Italy, together with a team from the University of York in the UK, developed a concept for the Heliplat – a Very-Long Endurance Solar Powered Autonomous Aircraft (VESPAA) (Romeo and Frulla, 2004). Heliplat and other VESPAA UAVs could play the role of a “pseudo satellite”, with the advantages of being closer to the ground, more flexible and at a cost lesser than an actual satellite. Heliplat-like HALE flying above a major city will be able to cover a circular area of 1,000 km diameter, and process a predicted 425,000 cell phone conversations

that uses a combination of a solar array and batteries to stay airborne. The plane weighs just 31 kg and has a wingspan of about 18 m (QinetiQ, 2013). The vehicle can circle over a particular area for extended periods. The military uses the vehicle for reconnaissance and communications platforms. Civilian and scientific programs use it for Earth observation. During the day, Zephyr uses its state-of-the-art solar cells spread across its wings to recharge high-power lithium-sulphur batteries and drive two propellers. At night, the energy stored in the batteries is sufficient to keep Zephyr airborne. The batteries are supplied by Sion. In 2008, the airplane registered a recorded flight, staying airborne for a total of 82 hours and 37 minutes, which is a very impressive number for a solar powered air vehicle (QinetiQ, 2013). The Zephyr went for two straight nights without stopping for recharging and only relying on its solar powered batteries. The airplane is able to climb to altitudes up to 18 km (58,000 ft) (QinetiQ, 2013).

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

352

Mattos, B.S., Secco, N.R. and Salles, E.F.

Missions conceived for Zephyr include forest and ocean monitoring, surveillance, and communication relay. Some highlights of Zephyr configuration: • Zephyr’s low-weight structure allows it to travel non-stop over long distances or to remain airborne in a desired station for a considerable long time. • The unique propeller design provides the aircraft with a higher power-to-weight ratio. • The aircraft flies day and night powered by solar energy – recharging its batteries along the day. • No complex launching mechanisms – the aircraft is launched manually by a ground crew of three people. • Silent flight – neither noise nor pollutant emissions. • The wing design takes into account thermal air currents to lift the aircraft to higher altitudes.

METHODOLOGY Before starting with the design of any solar-powered aerial vehicle, it is important to establish how the solar energy will be converted into vehicle motion (propulsion) and how to provide power for the airplane systems. Usually, the wing is an appropriate place to be covered with solar cells, which collect solar radiation and transform it into electrical energy. Batteries, propulsion and systems operate under different conditions and thus a converter package is necessary. An appliance called Maximum Power Point Tracking (MPPT) enables the extraction of maximum power of solar cells and supplies the aircraft systems. During night, there is considerably less energy from solar cells and battery shall fulfill all power needs. Some modern solar cells are able to convert the infrared night radiation into electric power, enabling the batteries continuously to be loaded. This kind of solar cell was not considered in the present work. Figure 5 provides a scheme of the energy management inside the aircraft. Solar Panels

Night

Converter

Engines, electronic equipament, others

Batteries

Day Figure 5. Solar energy conversion system to power engine and other aircraft systems.

SolAR iRRAdiATion modelS For a better understanding of the power system, it is necessary to describe the solar radiation model and the systems responsible for its transformation into useful energy. In order to proceed with the calculation of the solar panel area required on the airplane, it is necessary to estimate the available solar radiation in a given location and date. For this, the relative movement of the Earth around the Sun and of our planet around its axis must be well understood. The Earth has an elliptical orbit around the Sun, with an average distance of 150 million km. Period of revolution is of 365 days and 4 hours. On January 1st, the nearest distance between our planet and the Sun takes place (perihelion); and on July 1st, the remotest position, the distance is 3.3% farther than that of the closest position (aphelion). Considering that the incident solar radiation is inversely proportional to the square of the distance, the perihelion receives around 7% more radiation than when the Earth is at the aphelion. Another four points also are very important in the astronomy: the autumn and spring equinoxes; and summer and winter solstices (Fig. 6).

Figure 6. Reference points of Earth’s orbit.

The perihelion and the aphelion are the days of most and least intense radiation, respectively. The summer and winter solstices are the days with the longest and shortest time of solar radiation exposure. Being conservative, it is considered that the aircraft should able to fly in any condition and so the point of project is the critical day of the year, in which it receives the least radiation along the day.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane

14000 Accumulated Energy (Wh/m2)

The solar irradiance models found in the literature are based on data collected in the Northern hemisphere, mainly in the United States and Europe. For the present work, the model R.SUN was employed (IET, 2013). This sophisticated model was distributed by the Photovoltaic Geographical Information System (PVGIS) of the European Union. PVGIS provides online interactive maps of photovoltaic potentials in Europe and Africa. Figure 7 shows such a map considering the yearly sum of solar electricity generated by 1 kWp system with optimallyinclined modules. R.SUN was developed for the European and Africa environment and it is able to predict the direct and diffuse irradiation and that reflected by the Earth’s surface. This last part of the solar irradiation is not taken into account in the present work, since the solar panels are located on the wing upper side of our solar-powered UAV.

353

12000 10000 8000 6000 4000 2000 0 1/Jan

14/Mar

26/May

7/Aug

19/Oct

31/Dec

Figure 8. Yearly variation of accumulated daily radiation.

the irradiation model. The power demand should be easily achieved in the inter-tropical region, where Brazil is located, since the incident radiation is larger there. eleCTRiCAl loAd In the preceding paragraphs, the available power from solar irradiation was outlined. In this section, the concern is with the electrical power demand for the aircraft. Power required for the payload package can vary. For this reason, an electrical demand of 100 and 200 W was considered. Power required for the control systems actuators is estimated to be 25.5 W. The engine power can be modeled in the following way (Eq. 1; Anderson, 1989):

Figure 7. Photovoltaic solar electricity potential in European countries (Source: Photovoltaic Geographical Information System of European Union).

R.SUN is able to estimate the solar irradiation according to latitude, altitude, date and time of the day. In this way, it can be computed the radiation accumulated during an unclouded day. From this estimation, an example of density of energy per area is shown in Fig. 8. According to the R.SUN model, for an aircraft at 40ºN latitude and flying 16 km above sea level, the critical day is December 21st. In this day, the distribution of radiation is shown in Fig. 6. The chosen latitude is the reference for

CD C1.5 L

√(mg)S 2ρ 3

(1)

QinetiQ’s Zephyr presents an estimated electrical load of 174 W. This figure occurs close to the stall condition, being necessary to adopt a safety margin. There are some power losses in the power plant system that increase the required power to operate the airplane. Such losses are caused by the propeller, the control box, the engine, and the gearbox, which present efficiencies of 87, 95, 95, and 95%, respectively (Noth, 2008; Roskam, 2000). Taking into account such losses, the required weight-to-power ratio tops 100 kg/hp. Thus, the cruise is not the most critical condition to define the required power. Some critical maneuvers must be addressed for such purpose. The airplane must be able to handle gusts and other severe operating conditions. However, in the present work, the used approach was based on similar weight-to-power ratios at cruise

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

354

Mattos, B.S., Secco, N.R. and Salles, E.F.

Table 1. Weight to nominal power ratio for some solarpowered airplanes (Noth, 2008). Airplane

weight-to-power ratio (kg/hp)

Solar challenger

60.98

Pathfinder

29.53

Pathfinder plus

23.77

Centurion

27.46

Zephyr (estimated)

22.37

of some solar-powered planes (Table 1). The adopted value was slightly higher than that one of Zephyr: 25 kg/hp. Typical cruise power for electrical driven airplanes is 56% of the engine nominal power. Thus, for the Zephyr airplane, this represents an increase of 140% and the adopted figure for the present work airplane was of 17.9 kg/hp. ACTUAToRS Actuators are employed in control surfaces like ailerons, rudder, and elevators. The airplane under development is configured with no high-lift devices and presents no landing gear. German company Volz is the actuator supplier for the Zephyr UAV (Volz Servos, 2013). The temperature of operation for usual actuator varies from 30 to 70ºC. However, in the tropopause, the typical temperature is -56.5ºC. For this reason, they developed a special actuator

Figure 9. Volz actuator DA 13-05.

able to withstand temperatures below -70ºC – model DA 13-05-EXP 11859 (Fig. 9). All actuators of the DA-13-05 Volz family have the same size (37.4x34.9x13.3 mm), operating voltage (4.8 to 5 V), angle (130°), and casing mass (19 g) but they differ by available output torque, device rotating speed and electrical current (power) (Volz Servos, 2013). Zephyr’s actuator is the most advanced available on market in its class. However, there is no public data available and therefore we adopted the specification of the DA 13-05-60 device. SolAR pAnelS And BATTeRieS The solar panels are extremely important components, considering that they are the unique source of energy, except the charged batteries in the first day of operation. The impact of solar panels on the project is unquestionable. Their development was recorded to be slow and always has been a critical bottleneck for this kind of airplane. For example, Pathfinder and Pathfinder Plus were developed with the same characteristics, at the same time and using the same technology, and by the same team. However, they differ in service ceiling, which is about 30,000 ft for the Plus version thanks to increased efficiency of solar panels. Several characteristics of state-of-art solar panels were compared, among them the efficiency and weight per area. Flexible panels (Fig. 10) were then selected. Both specific weights are very similar and they can be considered as ultralight panels. Efficiency is the factor that was taken into account for panel selection as well. In a conservative way, in this study we consider ISA+15 atmosphere conditions, as the efficiency increases with decreasing temperature. As result, the panel

Figure 10. High-flexible and low-weight solar panels.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane

with Copper indium gallium selenide (CIGS) technology produced by Swiss company Flisom was selected, providing maximum efficiency of 18.8% (Flisom, 2013). The most used batteries for portable equipment have lithium in their composition. However, a new generation of batteries based on chemical reaction between lithium and sulfur is available, but the concept is experimental stage. The advantages are weight and volume reduction for the same charge capacity. Lithium-sulfur batteries made by Sion Power are very advanced and some versions have specific energy above 265 Wh/kg (Volz Servos, 2013), value that was adopted in the present study. Sion Power lithium sulfur (Li-S) batteries were fitted into the Zeyphr airplane in 2010. The batteries played a critical role in the QinetiQ Zephyr smashing the world record for the longest duration unmanned flight. As a result of an intensive joint development effort between Sion Power and QinetiQ, the Zephyr flight exceeded 336 hours (14 days) of continuous flight, significantly surpassing the previous official record of 30 hours 24 minutes set by Northrop Grumman’s RQ-4A Global Hawk in March 2001 (Sion Power™, 2013). The Zephyr’s world record flight was completed on July 23, 2010, at the US Army’s Yuma Proving Grounds in Yuma, Arizona (Sion Power™, 2013). Optimization framework The simulations for the airplane design under consideration were carried out by using the commercial software package modeFrontier® V4.3 (ESTECO, 2011). The optimization framework modeFRONTIER provides a multi-objective optimization and design environment and it easily enables the coupling of CAD/computer aided engineering (CAE) tools, finite element structural analysis and computational fluid dynamics (CFD) software. It is developed by ESTECO SpA and provides an environment for product engineers and designers (ESTECO, 2011). ModeFRONTIER® is a Graphical User Interface (GUI) driven software written in Java that wraps around the CAE tool, performing the optimization by modifying the value assigned to the input variables, and analyzing the outputs as they can be defined as objectives and/or constraints of the design problem. In the present work, the airplane configuration that was select for study is very similar to that of the monoplane Qinetiq Zephyr, i.e., featuring a rectangular wing with a conventional tail, which is connected to the wing by a boom. The two engines are located at the inner wing.

355

The design variables of the present optimization framework are given in Table 2. No geometric wing twist was considered. Two electrical power requirements related to two different payloads were simulated: 100 and 200 W. Table 2. Design variables. Variable

Description

ARw

Wing aspect ratio

ARHT

Horizontal stabilizer aspect ratio

ARVT

Vertical stabilizer aspect ratio

Wing area

CVH

Horizontal tail volume coefficient

CVT

Vertical tail volume coefficient

Location of the break station of the wing

Horizontal tail arm

Vertical tail arm

Φw

Outer wing dihedral angle

TConfig

Tail configuration (=0: conventional; =1: “T” tail)

It is reasonable to assume that a lighter airplane will be cheaper. Airplane size is proportional to weight, and if there are two airplanes, one small and one large, with the same payload, the smaller one is structurally more efficient, considering it requires a lighter structure to perform the same mission. Thus, airplane weight is an excellent indicator of configuration efficiency and it was considered the objective to be minimized in the mono-objective simulations. The constraints that were taken into account are: handling qualities, rudder shadowing at stall condition of the horizontal tail, and the area of solar panels, which have to be smaller than the available area on the wing. Multi-objective simulations were run as well. In this case, weight and energy surplus were taken as objectives and an electrical demand of 200 W for the payload was considered. Disciplines The airplane under consideration in the present study must operate at high altitudes and its loiter is performed at very low speeds, which implies a low Reynolds number relative to wing mean aerodynamic chord. Typical Reynolds number

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Mattos, B.S., Secco, N.R. and Salles, E.F.

356

lies in the 150,000 – 200,000 range. Due to the low speed on station, high values of lift coefficient (Cl) are experienced by the airplane wing. Taken into account that energy supply is critical, a low-drag configuration is mandatory. Based on Zephyr configuration, a typical Cl of 0.83 at 17 km of altitude was estimated, considering a speed of 60 km/h. In this condition, almost 80% of span operates at Cl > 0.8, as shown in Fig. 11. Thus, the biggest aerodynamic issue was to find out a lowdrag airfoil for high local lift coefficients. Due to low Reynolds number, a laminar airfoil was considered, with “drag bucket” covering the desired Cl range (Fig. 12). The NACA 6-digit family 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0.2

0.4

y/(b/2)

0.6

0.8

to experimental results published by Abbott and Doenhoff (Abbott and Doenhoff, 1959) (Fig. 13). Table 3 contains some figures obtained with XFOIL and from experimental data. The correction for the polar curve was obtained by fitting a parabola to the non-laminar portion of the polar curve (Eq. 2): Cd = Cd, XFOIL + 0.0006

(2)

Where Cd is the drag coefficient.

Figure 11. Lift coefficient distribution along wingspan for a Zephyr-like airplane. Calculation performed with Athena Vortex Lattice (Sion Power™, 2013).

Cd: drag coefficient; Cl: lift coefficient.

Figure 13. Adjusted polar curve for NACA 63412 airfoil according to experimental data.

0.0200 0.0180

NACA 2415

0.0140

NACA 66-415

Drag Coefficient (Cd)

0.0160 0.0120

Table 3. Abbott and XFOIL correlation.

0.0100 0.0080 0.0060 0.0040

DRAG BUCKET

0.0020 0.0000

-1.0

-0.5

0.0 0.5 Lift coefficient (Cl)

1.0

1.5

Figure 12. Typical drag polar of laminar airfoils.

offers a variety of laminar airfoils and therefore the NACA 63412 profile was selected to compose the wing. It combines all desired characteristics for the solar-powered UAV under consideration: high maximum lift coefficient (Clmax); moderate to low value for the moment coefficient (Cm); and 12% relative thickness, an adequate one. All the polar curves were computed with the panel code XFOIL (Drela, 2000) and adjusted to fit

Source

Cd,laminar

Cdm

Clm

Abbott

0.0051

0.0063

0.214

0.00732

XFOIL

0.0045

0.0061

0.177

0.00557

The NACA 63010 airfoil was chosen for the horizontal and vertical tailplanes (Fig. 14). Aerodynamic calculations for the complete configuration were performed with the incompressible Athena Vortex Lattice (AVL) vortex-lattice code (Youngren and Drela, 1988). AVL handles lifting surfaces, fuselages, and booms. All required aerodynamic and stability derivative coefficients can be quickly calculated with AVL (Youngren and Drela, 1988). Concerning load calculation, there is no applicable certification regulation for this kind of aircraft. Estimations that were carried out for the V-n diagram elaboration brought no realistic results. In this effort, FAA FAR-23 and JAR-VLA

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane

regulations were considered for a Zephyr-like airplane. For this reason, the present work considered a maximum a load factor of 3.8 for the structural sizing (Fig. 15).

357

analysis employed in the present study were very simple, as described by Megson (1999). The bending moment is supported by stringers only and torsion loads, by both stringers and skin. The structural layout is shown in Fig. 16, extending from leading edge to 25% of chord, where aerodynamic center is localized. The thickness of the wing skin is the same along the entire span, but stringers may vary for each wing station.

Cd: drag coefficient; Cl: lift coefficient.

Figure 14. Adjusted polar curve for NACA 63010 airfoil according to experimental data.

Figure 16. Wing structural layout.

Maneuver

3 Gust

2 nz

1 0

-1 -2

Equivalent Airspeed (m/s)

Nz: load factor.

Figure 15. Estimate V-n diagram for the Zephyr airplane.

The cruise speed defined by the minimum required engine power for leveled and straight flight. The upper limit of the airplane dive speed was set to be 25% above the cruise speed. The maximum load factor 3.8 was applied for wing sizing; horizontal and vertical tail stabilizers sizing considered the cruise condition, CLmax, and control surface deflection of 15º; and boom was sized considering simultaneous tails forces. Fuselage and engines were not calculated. The methods of

Ribs were not computed but it was assumed putting three of them per meter, with constant thickness. The boom has a cylindrical form with constant radius and thickness, approximated by thin-walled tube. It can easily withstand the existing loads for this kind of airplane. Its deflection was a constraint and used as sizing criteria. Carbon fiber was the standard material for solar airplanes, except skin, which is built of Mylar. Biaxially-oriented polyethylene terephthalate (BoPET) is a polyester film made from stretched polyethylene terephthalate (PET) and is used for its high tensile strength, chemical and dimensional stability, transparency, reflectivity, gas and aroma barrier properties, and electrical insulation. A variety of companies manufacture BoPET and other polyester films under different brand names. In the UK and the United States, the most well-known trade names are Mylar, Melinex and Hostaphan (Staugaitis and Kobren, 1996). Densities of the selected materials for the airplane structure are well known and thus the calculation of the mass distribution is an easy task. The fuselage and payload had fixed weight; the weight of batteries depends on the capacity to supply all systems during night; and engine weight was estimated based on methodology described by Noth (2008). Thus, the weight and moment of inertia of each component can be estimated. These calculations are performed through an iterative process

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

358

Mattos, B.S., Secco, N.R. and Salles, E.F.

until convergence is reached according to some stopping rules. Demand for electric power depends strongly on the drag, which, in turn, depends on the lift coefficient. Stability and control is a top issue in aircraft design, particularly hard for autonomous airplanes. Unplanned situations may happen and, in this case, there is no pilot to control and stabilize the airplane. Although an on-board computer can control an unstable airplane, this paves the way for accidents. For this reason, flying qualities levels 1 or 2 were imposed, according to MIL-F8785C (Department of Defence of United States of America, 1980), adapted for UAVs (Peters and Andrisano, 1997). This leads to the implementation of lateral-directional analyses in the design framework, addressing short-period, phugoid, roll, divergent spiral and Dutch-roll characteristics.

RESULTS The modeFrontier® workflow for the task under consideration is displayed in Fig. 17. There are ten design variables in total. They are related to the wing, horizontal tail (HT), and vertical tail (VT). The AVL panel code was employed to calculate the aerodynamic coefficients and stability derivatives. Some MATLAB® routines then evaluated the CLmax and dynamic behavior of the individuals that arise in the simulation process from data provided by AVL.

Wing

The simulations were run on a desktop computer fitted with a single Intel Core2Quad Q6600 2.40 GHz processor, and 8.0 GB of installed Random Access Memory (RAM). The initial population was generated with the SOBOL algorithm, available in the modeFrontier® framework. SOBOL guarantees that any region of the design space does not become saturated, assuring that the entire domain be fulfilled with a great variety of individuals. The multi-objective MOGA-II genetic algorithm (ESTECO, 2011) was chosen for running the simulations. MOGA-II is an efficient multi-objective genetic algorithm that uses a smart multi-search elitism. This new elitism operator is able to preserve some excellent solutions without bringing premature convergence to local-optimal frontiers. The efficiency of this algorithm has been orderly proved on six well known test functions for multi-objective optimization. For simplicity, MOGA-II requires only very few userprovided parameters; several other parameters are internally settled in order to provide robustness and efficiency to the optimizer. The algorithm attempts a total number of evaluations that is equal to the number of points in the Design of Experiment (DOE) table (the initial population) multiplied by the number of generations. mono-oBJeCTiVe SimUlATion The first runs revealed that a large number of individuals violated the posed restrictions. It is easy to understand, because in first generations genetic algorithm searches in all

Short period Short period level

Horizontal tail Vertical tail

ΦW SW ARW yq

ARHT LH CHT

ARVT LV CVT

Phugoid Phugoid level

Desgin variables

Dutch roll Dutch roll level Roll

Stabilizers combination constraint

Divergent spiral

Evaluate stabilizer combination (blanketing, etc...) DOE

MOGA II

Geometry

RestEmp

AVL 1

Roll level Divergent spiral level

AVL 2 Flight quality constraints

AVL input

AVL run

Loads

Extract AVL calculated parameters from output file

Figure 17. ModeFrontier® workflow for solar-powered unmanned aerial vehicles optimization.

END

Flight quality

Weight Weight minimization

Available area for Required area solar panels for solar panels constraint

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Stability

Structural sizing Inertias

Loads

Energy surplus

Energy susplus constraint Energy surplus Maximization

Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane

considered domain and more advanced generations are next to optimum. So, algorithm naturally reduced the domain to small range that offers good individuals, except mutations. For the simulation tasks considering payload electrical demand of 100 W, 20 generations revealed to be enough to provide satisfactory convergence for obtaining an optimal airplane configuration (Table 4). The first runs recorded a large number of individuals that violated the prescribed constraints. Along the convergence process, the optimization framework naturally narrowed the band to generate individuals and most of them fulfilled the imposed constraints (Fig. 18). In other hand, mutation may generate unfeasible individuals along the simulation process.

359

Calculation error (too ugly) Handling characteristics (HC) 12% 16%

All three constraints 1%

OK 6%

Power supply + HT location 7% HC + HT location 7%

HC + power supply 9%

32% Power supply HT location 10%

HT: horizontal tail; HC: handing characteristics.

Figure 19. Constraint violators pinpointing.

Table 4. Information for the standard demand simulation. Variable

Value

Generations

Number of individuals

400

Repeated individuals

Total processing time

7h51m25s

Average processing time for a single individual

1m19s

Average processing time for four individuals

5m15s

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

Feasible Unfeasible Error

1 2 3 4 5 6 7 8 9 10 11 12 13 14151617 1819 20

The area and weight are consistent with data from Zephyr, but its required power supply is unknown. However, the wing loading of the best individual is very close to that of Zephyr. Considering only the best individuals that did not violate restrictions, it is possible to state that a minimum wing area of 16 m2 is required for obtaining a feasible configuration. Larger areas can feed systems with greater capacity during the day and must be turned off at night because the battery does not have sufficient storage capacity. For example, a system that requires 350 W can work under daylight conditions and must be turned off at night for airplanes with wing area set to 20 m2. The day chosen as the project point is very restrictive. If the same airplane flew in the summer, an 1,800 W of power could be generated during the day. An overview of the best individual from the simulation is provided by Fig. 20. The optimal airplane presents 20 m2 of wing area and wing aspect ratio of 25 (Table 5). All handling qualities were level 1 on station.

Generation

Figure 18. Feasible individual progression along the simulation process.

The violation of power supply constraint is strongly associated with small wing area, while most of other violations are related to tail configuration parameters. Figure 19 provides an overview of the constraint violation by individuals for the entire simulation.

Figure 20. Optimal airplane configuration from the monoobjective simulation.

J. Aerosp. Technol. Manag., SĂŁo JosĂŠ dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Mattos, B.S., Secco, N.R. and Salles, E.F.

Table 5. Optimal airplane from mono-objective simulation.

20 m2

0.40

ARw

ARHT

outer wing dihedral angle

14°

ARV

1.75

4.25 m

Airplane gross weight amounts 30.1 kg. The weights of the airplane components are consistent and the battery weight fraction is huge (Fig. 21), revealing that this equipment presents an issue critical for the design of a solar-powered airplane. Typically, a battery generates 350 Wh/kg (265 Wh/kg considering the wrapper), but further development can increase this figure to 600 Wh/kg, which would have great benefit for the configuration of solar-powered aircraft.

6500 6100

4.75 m

Sw: wing area; Yq: break position; ARw: wing aspect ratio; ARHT: horizontal tail aspect ratio; ARv: vertical tail aspect ratio; Lv: vertical tail lever; LH: horizontal tail lever.

between 30 and 60 m2. After 40 generations with 30 individuals each, the simulation was stopped. Figure 22 shows the feasible individuals of the present simulation and Fig. 23, its Pareto front.

5700 5300 Power (W)

360

4900 4500 3700 3300 2900 2500 30

70 90 Weight (kg)

110

130

Figure 22. Feasible individuals from the multi-objective simulation.

Tail boom

5700

Fuselage + Payload

Power (W)

Wing

Batteries (65%)

4100

4700

3700 Engine

2700 35 EH: horizontal tail; EV: vertical tail.

Figure 21. The above pie chart provides a weight breakdown of the optimal airplane.

As seen before, batteries are critical for design of solar powered airplanes. If the electrical power requirement for the payload is increased, the structural weight is also increased. This directly affects the weight of the engine and the required thrust to propel the airplane. Multi-objective optimization Weight minimization and power surplus maximization were taken as objectives for this task. Wing area was allowed to vary

55 65 Weight (kg)

Figure 23. Pareto front of the multi-objective simulation.

Here, the available power from the optimal individuals is considerably higher than that of the optimal airplane from the previous simulation. This can be credited to the lower boundary that was set for the wing area, which was 50% higher than that of the optimal airplane from the mono-objective simulation. The configuration from the Pareto front with the lowest available power weighs just 35.21 kg, slightly over the weight of the optimal airplane from the previous simulation. Table 6 contains some characteristics of this airplane.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

Optimal Design of a High-Altitude Solar-Powered Unmanned Airplane

Table 6. Characteristics of the optimal lowest power airplane from the multi-objective simulation. Sw ARw Outer wing dihedral angle LV

30 m 21

12° 4.75 m

ARHT

0.55 3.25

ARV

1.50

6.50 m

Sw: wing area; Yq: break position; ARw: wing aspect ratio; ARHT: horizontal tail aspect ratio; ARv: vertical Tail aspect ratio; Lv: vertical tail lever; LH: horizontal tail lever.

CONCLUDING REMARKS The development of solar-powered airplanes in the past was directly affected by the technology of solar panels. Nowadays, alongside with the solar panel efficiency issue, energy storage became a major concern for the design of solar air vehicles. In addition, considering that there are few airplanes flying in altitudes from 15 to 18 km, there are also few aircraft systems manufacturers able to deliver products capable of working under severe environmental conditions like the ones found in such altitudes.

361

The application of optimization tools proved to be quite appropriate, as provided in less than one day running a very large variety of airplanes, for which one can identify the strengths, weaknesses, and typical values of the variables for a good design. The case study proposed is an excellent example of this, because there are few semi-empirical mathematical models and built aircraft data, and even then with the application of simple theories, it was possible to design and verify the consistency with the few existing models. Future work will encompass the wing and tail airfoils as design variables.

ACKNOWLEDGMENTS The authors thank Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) for its support by the project 2007/00305-5.

REFERENCES Abbott, I.H. and Doenhoff, A.E., 1959, “Theory of wing sections”, Dover Publications, Mineola, NY, USA. 693 p. Anderson, J.D., 1989, “Introduction to Flight,” 3rd Edition, McGrawHill, USA. Department of Defence of United States of America, 1980, “MILF-8785C: Military Specification Flying Qualities of Piloted Airplanes”, Washington, DC. Drela, M., 2000, “X-FOIL subsonic development system” [Internet], Massachusetts Institute of Technology (MIT).Available from: http://web. mit.edu/drela/Public/web/xfoil/. Accessed: June, 2006. ESTECO s.p.a., modeFrontier 4 User Manual, 2011. ®

flying qualities requirements for light weight unmanned aircraft,” AIAA Guidance, Navigation, and Control Conference, New Orleans, LA, Aug. 11-13. QinetiQ [Internet]. Available from: http://www.qinetiq.com/. Accessed: April 26, 2013. Romeo, G. and Frulla, G., 2004, “Heliplat: High Altitude Very-Long Endurance Solar Powered UAV for Telecommunication and Earth Observation Applications”, The Aeronautical Journal, Vol. 108, pp. 277-293. Roskam, J., 2000, “Airplane Design, Part VI: Preliminary Calculation of Aerodynamics, Thrust and Power Characteristics”, The University of Kansas, Lawrence. DARcorporation.

Flisom – Flexible Solar Modules [Internet]. Available from: http://www. flisom.ch. Accessed: April 26, 2013.

Sion Power™ [Internet]. Available from: http://www.sionpower.com/. Accessed: April 26, 2013.

Joint Research Centre, Institute for Energy and Transport (IET), “Solar Radiation and GIS” [Internet]. Available from: http://re.jrc.ec.europa.eu/ pvgis/solres/solmod3.htm. Accessed: April 26, 2013.

Solar Impulse [Internet]. Available from: http://www.solarimpulse.com/. Accessed; April 26, 2013.

MacCready, P.B., Lissaman, P.B.S., Morgan, W.R., and Burke. J.D., 1983, “Sun-Powered Aircraft Designs”, Journal of Aircraft, Vol. 20, No 6, pp. 487-493. Megson, T.H.G., 1999, “Aircraft structures for engineering students”, 3 Edition, Butterworth Scientific, Oxford.

Noth, A., 2008, “Design of Solar Powered Airplanes for Continuous Flight,” Ph.D. Thesis, École Polytechnique Fédérale de Lausanne, Swiss. Peters, M.E. and Andrisano, D., 1997, “The determination of longitudinal

Staugaitis, C. and Kobren, L., 1996, “Mechanical and Physical Properties of the Echo II Metal-Polymer Laminate,” NASA TN D-3409, NASA Goddard Space Flight Center. Volz Servos [Internet]. Available from: http://www.volz-servos.com/en/ index.php?m=m. Accessed: April 26, 2013. Youngren, H. and Drela, M., 1988, “Athena vortex lattice (AVL): an extended vortex-lattice model for aerodynamic analysis, trim calculation, dynamic stability analysis, and aircraft configuration development” [Internet]. Available from: http://web.mit.edu/drela/Public/web/avl/. Accessed: February, 2007.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.349-361, Jul.-Sep., 2013

INSTRUCTIONS TO AUTHORS (Revised in March, 2013)

SCOPE AND EDITORIAL POLICY

PAPER SUBMISSION

The Journal of Aerospace Technology and Management (JATM) is the official publication of the Departamento de Ciência e Tecnologia Aeroespacial (DCTA), in São José dos Campos, São Paulo State, Brazil. The journal is quarterly 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. When submitting the contribution, authors should classify it according to the area selected from the following topics:

The manuscript should be digitalized using a Microsoft Word (.doc) software program and submitted electronically in English. 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. After submitting the manuscript, the corresponding author will receive an e-mail with the Term of Copyright Transfer, in which the author agrees to transfer copyrights to the DCTA in case of acceptance for publication, thus being forbidden any means of reproduction (printed or electronic) without previous authorization of the Editor in Chief. If the reproduction is allowed, it is mandatory to mention the JATM. The author also declares that the manuscript is an original paper, its content is not being considered for publication in other periodicals and that all co-authors participated satisfactorily in the paper elaboration as to make public the responsibility for its content. The declaration must be printed, signed by the main author, and sent back by mailing to the following address: Instituto de Aeronáutica e Espaço/ATTN: Helena Prado/ Praça Marechal Eduardo Gomes, 50 – Vila das Acácias – CEP: 12228-901 – São José dos Campos/SP, Brazil. 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 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.

• Acoustics • Aerodynamics • Aerospace Meteorology • Applied Computation • Astrodynamics • Ceramic Materials • Circuitry • Composites • Computational Fluid Dynamics • Defense Systems • Eletromagnetic Compatibility • 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

PEER REVIEW

• Structures • Synthesis and Characterization of Aerospace Materials • Thermal Sciences • Vibration and Structural Dynamics.

Manuscripts will be reviewed by at least two expert consultants, members of the Editorial Committee or external evaluators (ad hoc referees) in double blind peer review mode,

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.362-364, Jul.-Sep., 2013

ensuring complete anonymity. In case of disagreement on the results of the evaluation, the manuscript will be forwarded to a third reviewer, and it will be accepted for publication only if two approvals are received. The evaluators can accept the manuscript in the form it was submitted, 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 version and, in the case the author does not agree with the suggestions, it is necessary to send a “letter to editor”, explaining the reasons. The Editor will approve after verifying in the new version the adherence to the reviewers’ suggestions or will send to another evaluation round 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 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. Review articles: These should cover subjects that are relevant to 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. Thesis abstracts: The journal welcomes recent Masters and PhD thesis abstracts for publication. Such contribution will not be submitted to peer review.

MANUSCRIPT STRUCTURE Whenever is possible, articles should include the following subsections, however articles from some areas should follow their usual format. Title and names of authors: The title should not contain abbreviations. All authors should be identified with full name, e-mail, institution to which they are related, city, state, and country. One of them should be indicated as the author for correspondence and his/her full address is required. Abstract: They are limited to 250 words and structured into objectives, methods, results, and conclusions. Citations or abbreviations (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.

J. Aerosp. Technol. Manag., São José dos Campos, Vol.5, No 3, pp.362-364, Jul.-Sep., 2013

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. Self-citation 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, 428p. 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 constructed using the table feature in the word processor or using a spreadsheet program, such as Microsoft Excel. They 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. Authors should take notice of the limitations set by the size and layout of the journal. Therefore, large tables should be avoided. 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.5, No 3, pp.362-364, Jul.-Sep., 2013

JATM is supported by:

Journal of aerospace technology and management

JOURNAL OF AEROSPACE TECHNOLOGY AND MANAGEMENT Vol. 5 N. 3 Jul./Sep. 2013 ISSN 1984-9648 ISSN 2175-9146 (online)

www.jatm.com.br

V.5, n. 3, Jul./sep., 2013

Journal of Aerospace Technology and Management