Genetic algorithms for the scheduling of multiproduct batch plants within uncertain environment

Abstract

This study addresses the problem of batch plant scheduling. In addition uncertainty on product demands is considered through probabilistic-based methods. In the resulting two-stage stochastic programming problem, the objective is to maximize an Expected Profit Value (EPV) while respecting a constraint forcing the makespan to be lower than a time horizon. A Genetic Algorithm (GA) is proposed for the solution of a multiproduct example. The variable encoding requires special attention. Computational tests are first carried out with a deterministic model to validate the GA efficiency. Then, different runs with different scenario sets highlight the existence of various solution classes, characterized by specific numbers of batches manufactured for each product. Further analysis finally enables to discuss if each schedule is really the best-fitted to the scenario set for which it has been determined.