Heurísticas para o Problema de Dimensionamento de Lotes com Máquinas Paralelas Flexíveis

• Autor(a) principal: Catelan, M. C. F. [UNESP]
• Data de Publicação: 2020
• Outros Autores: Araujo, S. A. De [UNESP], Fiorotto, D. J., Carvalho, D. M. [UNESP]
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• DOI: 10.5540/tema.2020.021.02.0003130313
• Texto Completo: http://dx.doi.org/10.5540/tema.2020.021.02.0003130313; http://hdl.handle.net/11449/212216
• topic: lot sizing problem; machine flexibility; heuristics; problema de dimensionamento de lotes; flexibilidade de máquinas; heurísticas

This work is related to the lot sizing problem with flexible parallel machines, which basically consists in determining the quantity of items to be produced, on each of the machines, in a finite time horizon, satisfying a given demand. This problem has an economic origin and involves costs of production, inventory and setup. In the standard problem, each item can be produced on any of the machines, i.e., we have the total machine flexibility. However, it is not always feasible to have total machine flexibility due to the costs. Therefore, it may be important to implement only a limited amount of flexibility. The investment of upgrading a machine to produce a specific product becomes a binary decision variable and there is a global budget on investment decisions. A mathematical formulation to this problem was proposed in the literature and the computational results obtained throwgh a combinatorial optimization package showed that solve the formulation is very difficult, especially for instances with many items. So in this work we have proposed three different heuristics for this problem, in order to find good solutions in low computational times. Experiments were performed to compare the heuristics with the resolution of the model via optimization package. We have seen that by increasing the number of items and machines, the results with better quality were obtained by only one of the heuristics. In terms of computational times, all the heuristics achieved inferior times in average over CPLEX.