Co-evolutionary Mechanism of Resource-Information Flows in Complex Industrial Systems: An AI-Enabled Framework for Manufacturing

Abstract

Complicated manufacturing systems need to coordinate physical resources and information flow in a changing production environment to ensure the stable operation of the production process. Industrial Internet of Things (IIoT) technology provides important support for equipment perception and data transmission, but existing research usually optimizes resource scheduling and information transmission separately, leaving the bidirectional interaction between manufacturing-state changes and information freshness insufficiently represented in the decision process. Thus, this paper proposes a resource-information co-evolution framework (STCF) for artificial intelligence-driven manufacturing systems, in which high-order manufacturing-state representation, information freshness, and resource-information decision-making are jointly modeled and optimized. Specifically, a dynamic heterogeneous hypergraph is constructed to represent many-to-many relationships among machines, products, operators, and sensors and to characterize their time-varying resource-information dependencies. Subsequently, a Proximal Policy Optimization (PPO)-based multi-agent deep reinforcement learning mechanism jointly adjusts production-resource allocation and information-acquisition strategies according to changes in resource states and information freshness. Experiments are conducted on two publicly available industrial datasets, including the AI4I 2020 Predictive Maintenance Dataset and the NASA C-MAPSS dataset. The results show that STCF reduces total energy consumption by 14.2% and decreases the average Age of Information (AoI) by 21.8% compared with the genetic algorithm baseline. Furthermore, on AI4I 2020, the proposed framework improves the Resilience Score by approximately 18.3% compared with the standard DQN baseline under dynamic equipment conditions. These results indicate that explicitly coordinating resource scheduling and information acquisition can improve the balance among resource efficiency, information freshness, and operational resilience in complex manufacturing systems.