Cross-device fault diagnosis of rolling bearings using domain generalization and dynamic model

Over the past few years, intelligent fault diagnosis technology has been widely applied and has achieved good results. However, these methods cannot effectively diagnose faults across devices. Existing transfer learning methods such as domain generalization (DG) can solve this problem, but these methods rely on multiple source domains to train the network, which limits their practical application. In response to this issue, this study proposes a simulate data-driven method for cross-device fault diagnosis method and a DG network named Adversarial Domain-Invariant Feature Exploration (ADIFEX). This method only requires a set of actual data to perform cross-device fault diagnosis tasks that are not visible in the target domain. ADIFEX extracts features of different health states of bearings by learning domain invariant features from the simulation model data and actual data and generalizes them to invisible target domains. Experiments using multiple sets of data verify that the proposed method has better performance in cross-device fault diagnosis with unknown target domains than other methods.