A matheuristic-based self-learning approach for distributed heterogeneous assembly flowshop scheduling with multiple assembly factories and make-to-order delivery

Concerns about mass personalized customization and customer services have highlighted the importance of make-to-order delivery in distributed manufacturing areas. These make-to-order delivery services are deeply intertwined with distributed assembly scheduling, where variations in customer demand significantly influence production costs and efficiency. To address this, we propose the distributed heterogeneous assembly flowshop scheduling with multiple assembly factories and make-to-order delivery. Our approach begins with a mixed-integer linear programming model aimed at minimizing the tardiness cost. Subsequently, a hybrid algorithm, incorporating mathematical programming, iterated greedy technique, and self-learning strategy, is designed to solve the model, and termed matheuristic-based self-learning iterated greedy (MSIG) algorithm. This algorithm features a matheuristic-based decoding mechanism and a problem-specific NEH heuristic to generate high-quality initial solution. The nested greedy phase involves the extraction of both customers and products to refine solution quality. Furthermore, the local search phase incorporates knowledge-based operators, rule-based operator candidate sets, and a self-learning selection strategy to enhance the algorithm’s exploratory capabilities. Finally, through comprehensive comparisons with nine existing heuristics and six state-of-the-art meta-heuristics, the superiority of the MSIG algorithm and the efficacy of its components are conclusively demonstrated.