Uma abordagem ACO para a programação reativa da produção

• Ano de defesa: 2010
• Autor(a) principal: Fonseca, Marcos Abraão de Souza
• Orientador(a): Kato, Edilson Reis Rodrigues
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
• País: BR
• Palavras-chave em Português: Inteligência artificial; Programação da produção; Sistemas flexíveis de manufatura; Programação reativa da produção; Ant Colony; Optimization; ACO
• Palavras-chave em Inglês: Reactive production scheduling; Flexible manufacturing systems; Ant Colony optimization; ACO; Graph representation
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/454

In the context of automated manufacturing systems, combinatorial optimization problems, such as determining the production schedule, have been focused in many studies due to the high degree of complexity to their resolution. Several studies point to use of metaheuristics for the problem dealt, where different approaches perspectives have been proposed in order to find good solutions in a short time. In this paper, we propose an approach based on Ant Colony Optimization metaheuristic (ACO) for the reactive production scheduling problem in an FMS aiming the combination of problem characteristics with metaheuristic characteristics. For this, the problem is addressed from two perspectives, based on modeling and the search method. The problem representation is characterized by a description of the problem at the operations level, since the production schedule is included in this context. On the model is applied a constructive search method based on ACO that using the collaboration principle, establishing a relationship between operations so that it lead the search for promising regions of the solution space. The goal of this work is to obtain a reactive programming in acceptable response time in order to minimize the makespan values. Experimental results showed an improvement of the results obtained so far by other approaches.