Análise de objetivos e meta-heurísticas para problemas multiobjetivo de sequenciamento da produção

Abstract

The sequencing of the production is an important process in decision-making and it is used in industries in order to allocate tasks to resources. Given the relevance of this kind of problem, the research in production scheduling is necessary. This study involves the process of optimization in the following problems: single machines, parallel identical machines, parallel identical machines with release time, unrelated parallel machines with setup time dependent on the sequence and on the machines, and flow shop which is flexible with setup time dependent on the sequence and stages. Moreover, multiple and conflicting objectives must be optimized at the same time in production scheduling and the literature has been showing progress in this sense. The present study analyses the commonly adopted objectives and suggests a set of objective pairs. Correlation analysis and aggregation trees are used here to indicate possibilities of aggregation among the conflicting objectives. Metaheuristics are commonly used to solve the sequencing problems addressed in this study and two of them, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Predator-Prey(PP), are applied to the proposed multiobjective problems in order to study their adjustments to the new cases. The NSGA-II is one of the most used genetic algorithms in sequencing problems. The PP is a recent evolutionary approach to scheduling problems, where each Predator is responsible for dealing with just one objective. A generalization of the PP technique, in which Predators considered both objectives using weights, is also proposed. In addition, the influence of the adoption of local search on these techniques is analyzed. Computational experiments adopting the hypervolume as a performance measure were conducted aiming at evaluating the computational techniques considered in this study and their variants.