Dynamic scheduling for manufacturing workshops using Digital Twins, Competitive Particle Swarm Optimization, and Siamese Neural Networks

Flexible manufacturing workshops often encounter scheduling challenges due to complex processes and cumbersome procedures. To address these issues, a dynamic scheduling method is proposed. Initially, a discrete manufacturing workshop scheduling problem model is developed, considering the unique characteristics of the workshop. Digital Twin technology and a Competitive Particle Swarm Optimization algorithm are then integrated to create the scheduling model. Finally, Siamese Neural Networks are incorporated to form a dynamic scheduling mechanism that optimizes disturbance scheduling. The research model demonstrated a quick convergence, efficiently searching for the optimal fitness value using both the Sphere and Griewank functions. In the scheduling objective function test, the model achieved a maximum completion time of 244.8 minutes, the shortest time compared to similar technologies. In Siamese Neural Network experiments, the model successfully suppressed the influence of disturbances, maintaining optimal scheduling performance. Without adjustments for disturbances, the maximum completion time was 58.5 minutes. After optimization, it decreased to 54.2 minutes. These results demonstrate the effective application of the proposed technology in workshop scheduling. The findings provide valuable technical insights for the application of intelligent technologies in workshop scheduling optimization.