Joint Optimization of Maintenance and Quality Inspection for Multi-Stage Manufacturing System Based on Genetic Reinforcement Learning

Multi-stage manufacturing systems (MMSs) are facing previously unheard-of opportunities for technological innovation due to the global wave of new-generation information and communication technologies and artificial intelligence. The development and use of intelligent manufacturing are at the heart of this shift. Nevertheless, most current research uses conventional techniques for optimization analysis, and the joint optimization of production activities for MMSs is insufficient. In this research, the Genetic reinforcement learning (GRL) method is utilized to simultaneously optimize system maintenance and quality inspection, considering the effect of buffer stock and the relationship between machine reliability and product quality. First, the manufacturing system's components are investigated, and models for machine degradation, machine machining quality, and buffer inventory are created. A joint optimization model based on GRL is then built following the manufacturing system's production process, and two methods for agent-environment interaction are proposed. The GRL algorithm suggested in this research has excellent adaptability and generalization ability compared to the Double Deep Q Network (DDQN) and genetic algorithm (GA). What is more, the superiority of the proposed algorithm is demonstrated by contrasting the strategies made following algorithm training with the control operation. Finally, pertinent recommendations for production management are obtained through comparison and agent learning.