Cross-Domain Fault Diagnosis of Rolling Bearings Using Domain Adaptation with Classifier Discrepancy

When diagnosing rolling bearing faults based on data-driven methods, the discrepancy in data distribution under different operating conditions may result in severe degradation of model diagnosis performance. To handle this issue, a cross-domain fault diagnosis method of rolling bearing based on domain adaptation with classifier discrepancy was proposed. Firstly, the convolutional neural network was used to extract the features of the labeled source domain samples and the unlabeled target domain samples. Then, the features were classified by two fully connected classifiers. Finally, the classification loss, the maximum mean discrepancy loss and the classifier discrepancy loss were optimized step by step to align the domain distribution discrepancy between the source domain and the target domain so as to implement the fault diagnosis of unlabeled target domain samples. The experimental results showed that the proposed method has a higher fault diagnosis accuracy rate than the mainstream domain adaptation methods, which verifies its rationality and feasibility. © 2021, Editorial Department of Journal of Northeastern University. All right reserved.