Joint Discriminative Adversarial Domain Adaptation for Cross-Domain Fault Diagnosis

The automatic feature extraction capability of deep learning has led to its extensive usage in fault diagnosis applications. In engineering scenarios where the distribution between training and test sets is inconsistent, deep domain adaptation (DA) methods are commonly used to solve cross-domain fault diagnosis problems. Despite achieving good performance for cross-domain diagnosis, there are some limitations to DA models. First, most existing research has only focused on domain alignment between source and target domains while neglecting class information, which can result in incorrect alignment between classes of the two domains. Second, target samples that are distributed close to the boundaries of the clusters are easily misclassified by the classification decision boundary learned from the source domain. To address these issues, joint discriminative adversarial DA (JDADA) is proposed in this article. The proposed method combines domain alignment and class alignment by introducing a class alignment module into the domain adversarial network. Furthermore, the discriminative discrepancy module is proposed to compact features of the same class and separate features of different classes to extract more discriminative features. In addition, we propose a new pseudolabeling strategy to address the problem of target training samples without labels. The proposed method is evaluated on the gearbox dataset and bearing dataset, and the results demonstrate its effectiveness and superiority over state-of-the-art DA methods. Specifically, JDADA achieves up to 5.0% accuracy improvement on the gearbox dataset and 3.4% accuracy improvement on the bearing dataset.

Automated summary

This article proposes a joint discriminative adversarial DA (JDADA) method for solving cross-domain fault diagnosis problems. The method combines domain alignment and class alignment by introducing a class alignment module and a discriminative discrepancy module to extract more discriminative features. In addition, a new pseudolabeling strategy is proposed to address the problem of target training samples without labels.