To meet the demand for intelligent monitoring and diagnostic decision-making of complex equipment, this paper proposes a digital twin enabled fault diagnosis method. Based on the five-dimensional digital twin model, a digital twin framework for complex equipment is established, incorporating the physical entity, virtual entity, digital twin data, services and connections. To address the challenges posed by multi-source data, strong temporal variability and nonlinear cross-channel dependencies in fault diagnosis of complex equipment, an improved transformer model is developed to enhance multi-channel temporal feature extraction through a redesigned patch construction mechanism. A digital twin system is further implemented on the Unity platform, where the diagnostic model is embedded via the Sentis inference mechanism to enable real-time data mapping, in-system inference and interactive visualization. Experimental validation using Permanent Magnet Synchronous Motor (PMSM) multi-channel current data demonstrates that the proposed model achieves a diagnostic accuracy of 99.8 percent, outperforming the baseline Deep Convolutional Neural Networks with Wide First-layer Kernel (WDCNN) by 2.16 percent. The results confirm that the proposed method can effectively characterize key features within multi-channel operational signals and support real-time diagnostic capabilities in digital twin environments, offering a scalable solution for intelligent fault diagnosis of complex equipment.
He et al. (Thu,) studied this question.