Multi-objective evolutionary algorithms with heuristic decoding for hybrid flow shop scheduling problem with worker constraint

The classical hybrid flow shop scheduling problem (HFSSP) considers the operation and machine constraints but not the worker constraint. Acknowledging the influence and potential of human factors as a key element in improving production efficiency in a real hybrid flow shop, we consider a new realistic HFSSP with worker constraint (HFSSPW) and construct its mixed integer linear programming model. Seven multi-objective evolutionary algorithms with heuristic decoding (HD) (MOEAHs) are proposed to solve the HFSSPW. According to list scheduling, we first present four HD methods for four MOEAHs, and these methods incorporate four priority rules of machine and worker assignments. The earliest due date (EDD) rule is further introduced into the HD methods for the other three MOEAHs. The developed model is solved using CPLEX based on 20 loose instances under a time limit, and the four proposed MOEAHs are evaluated by comparing them with the results from CPLEX and two best-performing algorithms in the literature. The computational results reveal that the proposed MOEAHs perform excellently in terms of the makespan objective. Additionally, comprehensive experiments, including 150 tight instances, are conducted. In terms of solution quality and efficiency, the computational results show that the proposed MOEAHs demonstrate highly effective performance, and integrating EDD into the HD can substantially enhance algorithm performance. Finally, a real-life problem of the foundry plant is solved by MOEAHs and the scheduling solutions totally meet the delivery requirement.

Automated summary

This study focuses on a new realistic hybrid flow shop scheduling problem with worker constraint (HFSSPW) and proposes seven multi-objective evolutionary algorithms to solve the problem, incorporating the earliest due date (EDD) rule into the heuristic decoding methods. The computational results demonstrate the excellent performance of the proposed algorithms in terms of makespan objective.