Construção de algoritmos de otimização para o planejamento do deslocamento de bobinas em uma fábrica de tubos flexíveis

• Ano de defesa: 2024
• Autor(a) principal: Letícia Diniz da Cruz
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA ELÉTRICA; Programa de Pós-Graduação em Engenharia Elétrica; UFMG
• Palavras-chave em Português: Busca local; Heurísticas; Otimização combinatória; Planejamento da produção; Programação linear inteira; Heurística; Programação linear
• Link de acesso: http://hdl.handle.net/1843/73610

In a flexible pipe factory, due to a series of restrictions related to product delivery dates, machine allocation, physical limitations of equipment and available space, it is necessary to plan how pipes in production should move between the machines responsible for construction of each layer of flexible pipes. The layers are defined according to customer specifications and for this reason there is no single continuous production line. To ensure operational efficiency, it is necessary that all stages of the production process are carried out with minimum delay, maximum advance and reduced use of resources. The displacement of flexible pipes is carried out by moving the reels that contain them and, for this reason, in this work two optimization strategies are proposed for planning the movement of reels. The first is aimed at daily factory planning, using algorithms based on meta-heuristics and mathematical models and with less rigid execution time restrictions. The second is intended for real-time replanning after unforeseen events and was designed based on local search heuristics that should be able to find good quality solutions in low computational time. Multiple heuristics were developed for each strategy, but all heuristics, both for planning and replanning, consider the same restrictions, objective functions and planning horizon, differing only by the available execution time. The proposed algorithms were tested in real instances of an industrial plant currently in operation and are capable of finding results in accordance with the factory's expectations in up to four hours for planning and in up to five minutes for replanning. Among the proposed heuristics, the one that achieved the best results for the planning problem was developed from the VNS meta-heuristic and for replanning, a deterministic local search heuristic.