Scheduling for remanufacturing systems with finite buffers: A regional co-evolutionary algorithm integrating deadlock repair mechanism

Remanufacturing is a core pillar of the circular economy. However, in actual production environments, buffer capacities within workshops are inherently finite. Ignoring this constraint can lead to production blockages that disrupt normal operations. To more accurately reflect this practical production constraint, this study explores the integrated remanufacturing system scheduling problem with finite buffers (RMSSP-FB), encompassing three integrated stages: parallel disassembly, job-shop-type reprocessing, and parallel reassembly. A mathematical model aiming to minimize the makespan and workload imbalance is established under strict finite buffer constraints, with its validity confirmed using OR-Tools. To solve the problem, a knowledge-enhanced regional co-evolutionary algorithm (KERCEA) is proposed. A five-layer encoding scheme is designed to balance the search space dimension with solution representation. During the decoding phase, deadlock triggering conditions are identified through theorem derivation, and a dedicated deadlock detection and repair mechanism is developed to guarantee the feasibility of the generated schedules. Additionally, heuristic initialization strategies are designed based on problem-specific knowledge, and a regional co-evolutionary framework is constructed to balance global exploration and local exploitation. Domain knowledge is deeply integrated into evolution via specialized local search operators, namely critical-path-based backtracking repair search and imbalance-driven workstation transfer search. Finally, comparative tests against similar algorithms on 36 multi-scale instances show that KERCEA has significant advantages in solution quality, diversity, and convergence. Furthermore, data analysis confirms that the algorithm significantly reduces the system's average blocking time.