Aplicação de técnicas de otimização à engenharia de confiabilidade

• Ano de defesa: 2008
• Autor(a) principal: Marcia de Fatima Platilha Salgado
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Engenharia Elétrica; Confiabilidade (Engenharia); Engenharia elétrica
• Link de acesso: http://hdl.handle.net/1843/BUOS-8CDG4N

This thesis aims to present Traditional as well as Computational Intelligence Based techniques for reliability modeling and analysis of repairable and non-repairable systems. The more recent applications of Computational Intelligence techniques to reliability engineering arebriefly presented. In order to demonstrate the applicability of Computational Intelligence to reliability optimization problems, evolutionary and immune algorithms are employed on theoptimization of generic systems focusing on decision variables such as system design, components reliability and redundancy level besides the costs involved on those decisions. From the reliability engineering perspective systems performance can be measured in terms of Mean Time To Failure (MTTF) in case of non-repairable systems, or in terms of Mean Time Between Failures (MTBF) in case of repairable systems. From the maintainability engineering perspective, a measure ofinterest is the Mean Time To Repair (MTTR) which is applicable only to repairable systems. All of them are the mean values of probability distributions when stochastic models are employed. The Availability is another measure often used to evaluate repairable systems performance which can be estimated from the MTBF and the MTTR. Once system reliability, maintainability and availability measures have been specified, other variables of interest can also be considered such as investment or costs associated to the design and the maintenance strategies, or froma risk perspective, the failure losses. Such measures might be deduced through a reliability modeling and analysis process and from life cycle cost studies. Once one system has been modeled and the variables of interest have been defined, an optimization problem can formulated, i.e.,objectives and constraints are represented mathematically. The characteristics of the objective and constraints functions are studied so that more suitable optimization techniques can be chosen. In practice, multi-objective formulations (multi-criteria or multi-attribute) are preferred since they make possible to turn the decision-making process more accurate once simultaneous contradictory objectives are involved, e.g., the system performance level must be maximized while the involved costs must be minimized (cost-benefit analysis). The study of ptimizationtechniques is extremely important once the search for feasible solutions which maximize the performance and minimize the costs associated to a system are intrinsic goals of the reliability and maintainability engineering. Optimization problems are generally defined in the form non- linear mixed-integer programming. Deterministic techniques are not efficient on solving this class of problems since they are combinatorial what make difficult to effectively solve them in polynomial-time by none of exact methods. This type of problem is known in the literature asNP-hard (nondeterministic polynomial-time hard). Stochastic techniques are more suitable in these cases, what motivates the use of Computational Intelligence techniques.