Imbalance domain adaptation network with adversarial learning for fault diagnosis of rolling bearing

Domain adaptation has been widely used in industrial fault diagnosis. However, it requires a balance of data distribution between the source domain and the target domain. Unfortunately, cross-domain balanced distribution is not common in actual applications, bringing difficulties to the practical application of domain adaptation. In the article, we focus on this difficult problem and propose a new imbalance domain adaptation network with adversarial learning (IDAL). The model applies adversarial learning to data augmentation of the target domain and uses the domain adaptation based on a neural network to narrow the feature distribution discrepancy between the source and target domains. Ultimately, the parameters are transferred to the target domain and fine-tuned with a small number of labeled samples so as to achieve fault diagnosis. The accuracy of the proposed method on two data sets is more than 98%. It is worth noting that common deep learning networks can be embedded in IDAL, so the model can be widely used in different industrial scenarios.