Tackling dual-resource flexible job shop scheduling problem in the production line reconfiguration scenario: An efficient meta-heuristic with critical path-based neighborhood search

Abstract

Addressing diverse production demands, companies must frequently reconfigure the production line to manufacture various customized products. However, production line reconfiguration requires reasonable scheduling of workers and auxiliary resources to ensure the debugging of different machines. Therefore, this paper defines the dual-resource flexible job shop scheduling problem in the production line reconfiguration (DRFJSP-PLR) scenario to minimize makespan. While traditional single-resource scheduling methods inadequately tackle the dual-resource cooperative constraints, struggle to guarantee solution quality. Hence, a mixed integer linear programming (MILP) model is developed, addressing the lack of rigorous mathematical characterization in prior methods. Based on this, a rule-guided exemplar learning genetic algorithm with neighborhood search (RgELGA_NS) is proposed. The main innovations include: (a) a rule-guided initialization approach is designed to enhance the initial population quality. (b) an exemplar learning strategy is adopted to select crossover individuals to reduce the destruction of inferior solutions to superior ones. (c) a neighborhood search operator considering resource cooperation based on critical path is presented, which significantly augments the population local exploitation ability. Experimental results on 60 instances demonstrate that the MILP model can effectively solve small- and medium-sized problems, and RgELGA_NS can obtain near-optimal solutions for different scale problems. Compared to other meta-heuristics, our algorithm exhibits superior convergence and stability, achieving the best scheduling schemes on 93.33% instances.