An optimized dynamic model improved deep discriminative transfer learning network for fault detection in rotation vector reducers

In recent years, recognized fault types using artificial intelligence (AI) models has gradually become one of the mainstream directions in the field of mechanical fault diagnosis. However, it is very difficult to obtain relatively complete fault samples, which limits the application of AI models for complex mechanical systems, such as rotation vector (RV) reducers. To address this issue, physical model-based fault sample generation methods attracted many attentions but still an open problem: the difference in fault samples between numerical simulation and measurement of a physical system needs to be well decreased. Therefore, this article proposes an improved deep discriminative transfer learning network (IDDTLN) for RV reducer fault diagnosis. The presented network is driven by an optimized dynamic model. Firstly, the measurement normal signal and whale optimization algorithm (WOA) are utilized to update the dynamic model parameters of the RV reducer. Secondly, the updated numerical model is employed to calculate simulation fault samples. Finally, simulation samples and unlabeled measurement fault samples are used to train IDDTLN. The trained IDDTLN can accurately identify the unknow measurement fault samples. The data obtained from RV reducer test rigs are utilized to explore the feasibility of the proposed method, and the classification accuracies reach 99.8 %.