Resilience-driven maintenance scheduling methodology for multi-agent production line system

In recent years, the public has been paying ever greater attention to problems related to resilience, since resilience presents the recovery ability to increasing complexity and vulnerability of disturbances. Many researches have been performed to address resilience of networks disruptions in manufacturing. However, a systematic method to model and analyze maintenance scheduling in disturbed production line system is not well developed. This paper considers a resilience-driven maintenance scheduling methodology under maintenance resource constraints for a simplified production line system, which consists of an upstream production machine, a downstream production machine and an intermediate buffer. The machines with degradation quality states represented by multiple decreasing yield levels are modeled as semi-Markov decision processes. A hierarchical and policy-coupled methodology based on reinforcement learning is used to determine maintenance policy of the system. The numerical results show that the application of the methodology to the aforementioned system can minimize the total cost and converge to the approximate optimal solution.