Multi-population Cooperative Multi-Objective Evolutionary Algorithm for Sequence-Dependent Group Flow Shop with Consistent Sublots.
Yuanyuan Zhang,Junqing Li,Ying Xu,Peiyong Duan
DOI: https://doi.org/10.1016/j.eswa.2023.121594
IF: 8.5
2024-01-01
Expert Systems with Applications
Abstract:Inspired by the production model of pressure vessels for spacecraft, i.e., tanks and cylinders, this study addresses the sequence-dependent group flow shop scheduling problem with consistent sublots (SDGFSP_CS) to minimize makespan and total energy consumption. In the problem under consideration, there are several coupling sub-problems, namely, the group sequencing, job sequencing, lot assignment, and machine speed assignment. To solve these problems, a multi-population cooperative multi-objective evolutionary algorithm (MPCMOEA) is proposed. In the MPCMOEA, a hybrid initial method that combines two problem-specific heuristics is designed to generate high-quality initial solutions. Then, considering the problem features, a cooperative mechanism considering the co-evolution of multi-population and the archive set is designed to accelerate the optimization process. In the co-evolutionary stage, to deepen the exploitation ability of local search, an enhanced search with multiple problem-specific operators is implemented. Furthermore, a re-initialization method is developed to improve the global search abilities. Finally, 27 different scale instances are generated for a series of numerical experiments. For the hypervolume and inverse generational distance metrics, MPCMOEA gets 20/27 and 21/27 optimal values, respectively. It verifies that the MPCMOEA outperforms efficient algorithms in terms of the diversity and convergence performance.