Multi-Domain and Multi-Scale Digital Twin Dynamic Modeling for Intelligent Workshop State Awareness

To address the challenges of complex, dynamic, and multi-source heterogeneous intelligent workshop state awareness and modeling, this paper proposes a multi-domain, multi-scale digital twin dynamic modeling framework. The approach integrates cross-domain abstraction, adaptive multi-scale feature extraction, and a graph neural network-based state mapping reasoning mechanism, achieving unified modeling across temporal, spatial, and data dimensions. Experiments based on the SECOM industrial dataset demonstrate that the proposed model achieves superior performance in state prediction and feature separability, showing clear advantages over traditional machine learning models. The results provide strong theoretical and experimental support for real-time monitoring, anomaly detection, and decision optimization in smart manufacturing.