Knowledge-Guided Multiview Hierarchical Evolutionary Algorithm for Flexible Job Shop Scheduling With Finite Skilled Workers

Abstract

This work addresses the flexible job shop scheduling with finite skilled workers, extending classical flexible job shop scheduling by incorporating operation decomposition, finite worker, and worker transfer. These new problem features significantly increase the complexity of solving, as several operations requiring multiple workers can lead to worker competition, causing delays in other operations that depend on the same workers. Previous studies focused on either operation decomposition or worker transfer but did not address the issue of worker competition. To tackle this challenging optimization problem, we propose a knowledge-guided hierarchical evolutionary algorithm (KHEA) with multiview cooperative neighborhood search. The key contributions of this work are as follows: 1) a hierarchical solving framework is proposed to reduce the solving difficulty. This problem is decomposed into three levels. The first level ignores the worker assignment and the second level starts optimizing it. The final level then refines the global solution; 2) a knowledge-guided crossover operator with a feedback schema is designed to improve the efficiency of crossover operations; and 3) a multiview cooperative neighborhood search strategy is proposed to reduce the idle time caused by worker competition. This involves designing a new disjunctive graph that accounts for worker competition to identify the critical path. The information from both machine-view and worker-view Gantt charts is cooperatively utilized to minimize idle time. Our method, KHEA, was tested on two benchmarks across 28 instances and 16 large-scale instances, with equal running time for comparisons. Compared to state-of-the-arts, KHEA obtains significant superiority.