An Enhanced Estimation of Distribution Algorithm with Problem-Specific Knowledge for Distributed No-Wait Flowshop Group Scheduling Problems
Zi-Qi Zhang,Yan-Xuan Xu,Bin Qian,Rong Hu,Fang-Chun Wu,Ling Wang
DOI: https://doi.org/10.1016/j.swevo.2024.101559
IF: 10.267
2024-01-01
Swarm and Evolutionary Computation
Abstract:With the trend of economic globalization, distributed manufacturing widely exists in modern manufacturing systems. As an extension of the distributed flowshop scheduling problem, the distributed no-wait flowshop group scheduling problem with sequence-dependent setup times (DNFGSP_SDSTs) is investigated in this article. To address DNFGSP_SDSTs with the criterion of minimizing makespan, this study proposes an enhanced estimation of distribution algorithm·(EEDA) with problem-specific knowledge. First, a mixed integer linear programming(MILP) model of DNFGSP_SDSTs is established. Second, based on the characteristics of DNFGSP_SDSTs, five problem-specific properties about local search operators are derived as prior knowledge to reduce computational cost. Third, two NEH-based two-stage heuristics are presented to construct a high-quality population with diversity. Fourth, a probability model with problem-specific knowledge and a family-based updating mechanism are developed to accumulate valuable pattern information from high-quality solutions, while a sampling strategy is designed to generate new populations with the accumulated information. Fifth, several local search operators are devised to refine the obtained solutions. Furthermore, perturbation and reinitialization methods are developed to avoid premature convergence. Finally, the validity of the MILP model is verified by using the Gurobi solver. The parameters of EEDA are tuned through a design of experiments. The effectiveness of key components in EEDA is confirmed through extensive experiments, and the computational comparisons with the state-of-the-art algorithms indicate the effectiveness of the proposed EEDA for solving DNFGSP_SDSTs.