A digital twin–driven fault diagnosis method with optimized feature matching for high-speed train gearbox gears

Abstract Data-driven approaches have been widely applied to fault diagnosis of high-speed train gearbox gears; however, under complex operating conditions, fault-state data are difficult to obtain, which significantly limits the performance of such methods in practical industrial scenarios. To address this challenge, this study proposes A digital twin-driven fault diagnosis method with optimized feature matching for high-speed train gearbox gears digital twin-driven fault diagnosis (DTFD). In the physical space, a series of experimental test rigs are constructed to collect gearbox vibration signals, while in the digital space, a structurally coupled dynamic model of the high-speed train transmission gearbox is established based on multibody dynamics, enabling accurate characterization of typical gear fault mechanisms and generation of abundant fault-related vibration data. Furthermore, a data interaction network is developed to achieve cross-space sharing of fault-state information through optimal feature matching, effectively integrating incomplete physical data with augmented digital data to support high-accuracy gearbox fault diagnosis. Experimental results demonstrate that the proposed DTFD method achieves an average diagnostic accuracy of 96%, outperforming the best comparative method by approximately 3%, thereby verifying its effectiveness and robustness.