A Rule and Network Dual-Driven Dynamic Scheduling Approach for Complex Flexible Manufacturing Job Shop

Traditional manually designed dispatching rules (DRs) cannot cope with the variety and rapid changes in production environments due to their inflexibility and lack of adaptability. Meanwhile, the development in manufacturing intelligence and informatization allows for storing a vast amount of scheduling data during production. Consequently, a novel approach emerges in constructing scheduling models and decision-making strategies based on historical data. Following this idea, this paper proposes a dual-driven dynamic scheduling approach for complex flexible manufacturing job shops, utilizing a combination of manually designed DRs and backpropagation neural network (BPNN) technology. Specifically, six different production scenarios are designed to optimize different objectives, simulating the actual complex manufacturing environment. Then, a real-time scheduling mapping network model is developed with the production states as input and selected DRs as output by training the BPNN. The results demonstrate the effectiveness and practicality of the proposed method in resolving dynamic scheduling issues of complex flexible manufacturing job shops, surpassing traditional manual DRs.