An improved deep Q-network for dynamic flexible job shop scheduling with limited maintenance resources
针对有限维护资源下的动态柔性作业车间调度问题，提出结合深度Q网络与局部搜索的混合算法，有效优化完工时间，适用于动态制造环境。
This study addresses the Dynamic Flexible Job Shop Scheduling Problem with Limited Maintenance Resources (DFJSP-LMR), a critical challenge in modern manufacturing systems. Traditional Flexible Job Shop Scheduling (FJSP) models often fail to account for machine deterioration, maintenance constraints, and dynamic events such as machine breakdowns and urgent tasks. To bridge this gap, we propose a novel mathematical model that integrates preventive maintenance (PM) and limited maintenance resources into the scheduling framework, aiming to minimise the makespan. The model considers machine reliability, maintenance activities, and resource constraints, reflecting real-world production environments. To solve this NP-hard problem, we introduce a hybrid algorithm combining Deep Q-Network (DQN) with a local search (LS) mechanism, enhanced by a Sliding Time Window Algorithm (STW) to manage maintenance resource allocation efficiently. The proposed DQN-LS algorithm is validated through extensive computational experiments, demonstrating superior performance compared to traditional scheduling rules and DQN without LS. The results highlight the algorithm's effectiveness in optimising makespan while balancing maintenance and production scheduling, offering valuable insights for both theoretical research and practical applications in dynamic manufacturing systems.