Comparative analysis of Deep Reinforcement Learning configurations in Flow Shop for enhanced Maintenance Management

Optimising maintenance scheduling in flow shop settings is a significant problem in achieving the goal of efficiency in industrial environments, necessitating novel solutions. This paper presents a complete multi-method approach to overcoming the complexities of maintenance management in flow shop production systems, combining Deep Reinforcement Learning (DRL) with advanced simulation approaches. We especially look into the effect of different configurations on the performance of a DRL-trained model tasked with maintenance decision-making. Our methodology comprises developing and comparing two distinct DRL configurations. We conducted rigorous simulation-based studies to assess the effectiveness of each DRL configuration in managing maintenance schedules under varied production needs and machine failure rates. The comparison research provides findings about the trade-offs between short-term efficiency and long-term sustainability in maintenance planning, emphasising sophisticated DRL techniques' ability to adaptively balance these goals. Our findings show that a multi-method approach combining DRL and simulation can provide a versatile and powerful tool for enhancing maintenance procedures in flow shop environments. By demonstrating the benefits and limitations of various DRL setups, the research adds vital perspectives to the ongoing development of intelligent production management systems, opening the path for more resilient and efficient manufacturing operations.