Um algoritmo evolucionário para o problema dinâmico de localização de facilidades com capacidades modulares

• Ano de defesa: 2017
• Autor(a) principal: Silva, Allyson Fernandes da Costa
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Brasil; UFRN; PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA ELÉTRICA E DE COMPUTAÇÃO
• Palavras-chave em Português: Localização dinâmica de facilidades; Capacidade modular; Metaheurística híbrida; Algoritmo genético; Variable neighborhood search; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA E DE COMPUTAÇÃO
• Link de acesso: https://repositorio.ufrn.br/jspui/handle/123456789/24220

Location problems aim to determine the best positions where facilities should be installed in order to meet existing demands. Due to its wide applicability, several characteristics have already been appended to the models to better represent real situations. One of them generalizes classical models to the case that location decisions should be taken periodically. Another allows models to deal with capacity sizing as a problem variable. The Dynamic Facility Location Problem with Modular Capacities unifies these and other characteristics present in location problems in a single and generalized model. This problem was recently formulated in literature where an exact approach was introduced and applied to instances of a case study in the context of the forestry sector. We present an alternative method to solve the same problem. The method chosen uses a Genetic Algorithm metaheuristic framework and hybridizes it with a Variable Neighborhood Descent routine with three neighborhoods adapted from others applied to location problems. Experiments attested the effectiveness of the hybrid metaheuristic developed in comparison to the use of those methods purely. Compared to the exact approach, the heuristic proved to be competent by finding solutions up to a gap of 0,02% to the global optimum in the majority of the instances tested.