Self-repair of smart manufacturing systems by deep reinforcement learning

Smart manufacturing is built on complex decision-making in real-time based on data from networked machines and sensors. As a potential enabler of smart manufacturing, reconfigurability enhances adaptability to demands and enriches the utility of the collected data. This study focuses on a synergistic combination of advanced manufacturing technologies in a highly diverse market environment, to identify deficient components by inferring from changes in product quality and to sustain the operation of the manufacturing system by creating multiple-level self-repair strategies. Deep reinforcement learning is then used to select the strategy based on system status and performance.