A novel deep convolution multi-adversarial domain adaptation model for rolling bearing fault diagnosis

The traditional rolling bearing fault diagnosis methods are difficult to achieve effective cross-domain fault diagnosis. Therefore, a novel deep convolution multi-adversarial domain adaptation (DCMADA) model for rolling bearing fault diagnosis is proposed, which includes a feature extraction module, a domain adaptation module, and a fault identification module. In the feature extraction module, an improved deep residual network (ResNet) is used as the feature extractor to extract the transferable features from the raw vibration signals. In the domain adaptation module, the marginal and conditional distributions are adjusted using multi-kernel maximum mean discrepancy (MK-MMD) and multiple domain discriminators in the source and target domains, and an adaptive factor is designed to dynamically measure the relative importance of these two distributions. In the fault identification module, the classifier uses the extracted domain-invariant features to complete cross-domain fault identification. Experiments show that the model has superior transfer capability in cross-domain bearing fault diagnosis.