Problema da programação da produção de bebidas à base de frutas: modelos de otimização e métodos heurísticos

Abstract

The present work aims to study the problem of the production scheduling optimization of fruit-based beverage, such as nectars and fruit flavoured drinks. This is a lot sizing and scheduling problem in two main production stages (beverage preparation and bottling) found in typical companies of the sector. The main characteristics from this production process, that differs it from other lot sizing and scheduling the production of beverages problems, such as soft drinks and beer, is the existence of mandatory temporal cleanings in the process after a determined time without any cleaning and the presence of inventory buffers in the second stage, which impacts the synchronization between the two production stages. As far as the research is concerned, this optimization problem has not yet been addressed in industrial engineering and operational research literature. The objective of this thesis is to propose approaches with effective solutions for this problem. With these approaches, it is evaluated how to optimize the production scheduling in common factories at the sector, and with that, it is intended to contribute to the lot sizing and scheduling problems literature. A new mathematical model of optimization is proposed to represent the integrated two-stage lot sizing and scheduling problem with temporal cleaning for the fruit-based beverage production. Heuristic methods based on the decomposition of the problem into stages and on approximated models for the problem are also proposed. In addition, the branch-and-cut method from CPLEX solver and the relax-and-fix and fix-and-optimize heuristics are used to solve the integrated model. It also explores fix-and-optimize heuristics with neighborhood search. Computational tests are performed with instances based on real data from a typical company at the sector. The results show that the proposed optimization approaches provide good solutions to the problem and are effective in solving it in real situations.