Uma Math-Heurística Para O Problema De Dimensionamento De Lotes Com Máquinas Paralelas E Setup Carry-Over

• Ano de defesa: 2018
• Autor(a) principal: Fernandes, Carla Cristina Doescher [UNIFESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Paulo (UNIFESP)
• Palavras-chave em Português: Lot Sizing; Setup Carry-Over; Parallel Machine; Multiple Itens; Dimensionamento De Lotes; Múltiplos Itens; Máquinas Paralelas
• Link de acesso: https://sucupira.capes.gov.br/sucupira/public/consultas/coleta/trabalhoConclusao/viewTrabalhoConclusao.jsf?popup=true&id_trabalho=6383839; https://repositorio.unifesp.br/handle/11600/52314

This Master’s Dissertation deals with the Lot Sizing Problem (LSP) in Parallel Machines with multiple items are considered and different machines that produce the same items and that have capacity restrictions. The items can be produced in any machine and at the beginning of the production of each item the time and cost of setup of the used machine is incurred. In addition, it is considered the possibility of taking advantage of setup at the end of each period: the setup carry-over, for cost reduction and, implicitly, for the definition of a production planning that reduces possible environmental impacts caused by the preparation of machines. Despite several studies in the literature for LSP in parallel machines, few of them introduce the efficient solution methods to solve it considering the setup carry-over. In this sense, to solve this problem efficiently, in this Dissertation, it is proposed its study from the point of view of modeling and efficient solution method. First, three mixed integer programming models were adapted for the LSP in Parallel Machines. After experiments, two models were defined that were used for the proposed solution method based on modeling (math-heuristics). The math-heuristic based on the well-known heuristic relax-and-fix with local branching constraints was modeled whose structure was inspired by the Greedy Randomized Search Procedures (GRASP). Experiments, in which the solutions obtained by the CPLEX solver and the math-heuristic are compared, demonstrated the efficiency of the proposed method.