Intelligent scheduling optimisation for whole-vehicle stamping production via proximal policy optimisation deep reinforcement learning
提出一种结合图神经网络与近端策略优化算法的整车冲压生产调度方法，通过动态选择调度规则和混合策略提升设备利用率和生产效率，解决传统方法处理动态任务时的局限性。
The complex scheduling problem in whole-vehicle stamping production is a key challenge in the field of resource allocation. Due to the diversity and dynamic nature of shop-floor scheduling tasks, stamping production is regarded as a typical flexible manufacturing resource, and the rationality of its scheduling plan directly affects equipment utilization and production efficiency. This paper proposes a whole-vehicle stamping production scheduling algorithm based on Graph Neural Network and Proximal Policy Optimization (GPVSM). This method adopts a GNN-based stamping production task resource construction method, combined with PPO algorithm for model training to optimize scheduling decisions. The main contributions include: (1) designing a scheduling rule combinatorial optimization method based on PPO deep reinforcement learning, which dynamically selects optimal rules through GNN; (2) adopting a hybrid scheduling rule strategy that integrates multiple rule advantages to improve efficiency; (3) designing a hyperparameter optimization method based on orthogonal experiments, systematically evaluating combinations to select optimal configuration. The innovation of this paper lies in combining deep reinforcement learning with whole-vehicle stamping production task scheduling, overcoming the limitations of traditional stamping task scheduling methods in handling dynamic tasks, and providing an efficient and intelligent solution for whole-vehicle stamping production scheduling.