Single-Source Domain Generalization Fault Diagnosis of Wheel-Set Bearings Based on Flow Model and Contrastive Learning

Domain generalization methods have demonstrated their effectiveness in identifying mechanical faults under previously unencountered working conditions. However, the majority of these approaches necessitate data from multiple source domains to train the model. Nevertheless, the collection of monitoring data of each health state under different working conditions is unpractical for wheel-set bearings. Aiming at the difficulty of obtaining multi-source domain data, we present a single-source domain generalization model that leverages flow modeling and contrastive learning techniques for health state identification of wheel-set bearings under diverse operating conditions. The proposed model employs flow model as a domain generation module to produce datasets within an extended domain. Then, domain-invariant features are extracted from the source and the generated domains. The diversity of generated dataset and the validity of domain-invariant features are guaranteed by a strategy of adversarial contrastive learning. Finally, single-source domain generalization diagnosis experiments carried out on a wheel-set bearing dataset verify the good performance of the presented method over traditional domain generalization methods.