Evolução diferencial melhorada implementada em processamento paralelo

• Ano de defesa: 2014
• Autor(a) principal: Brandão, Milena Almeida Leite
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Engenharia Mecânica; Engenharias; UFU
• Palavras-chave em Português: Otimização; Evolução diferencial melhorada; Computação paralela; Robô manipulador; Grandes sistemas lineares; Otimização combinatória; Automação industrial; Sistemas lineares; Optimization; Improved differential evolution; Parallel computing; Robot manipulator; Large linear systems; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/14760; https://doi.org/10.14393/ufu.te.2014.129

The aim of this work is to present an improvement of the heuristic optimization method of Differential Evolution proposing modifications to its basic algorithm by using the concept of evolution with Shuffled Sets. The method called Improved Differential Evolution (IDE) was developed, and implemented for parallel computing (IDEP), making it suitable for solving complex optimization problems. The algorithm developed is adapted to work with multiple objective optimization problems and in the presence of constraints. Some test functions are solved by using the IDEP method in order to validate the algorithm. The IDEP methodology is utilized to obtain the optimal design of a robot manipulator with three rotational joints (3R) taking into account the characteristics of its topology. For this purpose, a multiple objective optimization problem is formulated to obtain optimum robot geometrical parameters, considering the maximization of the volume of the workspace, the rigidity and the optimization of its dexterity. Finally, the IDEP algorithm is applied to solve large linear systems, rewritten as a residues minimization problem. All the results obtained with the developed algorithm are compared with the solutions calculated through other methodologies in order to prove its efficiency and the relevant gain in terms of computational time.