Scholarlink

<p>Most transfer learning-based bearing fault diagnosis methods under variable working conditions typically focus on domain alignment, while neglecting the inter-class separability among samples. This oversight leads to the degradation of diagnostic accuracy in the target domain. To address this issue, a novel fault diagnosis method based on contrastive learning and domain adaptation network (CDAN) is proposed. A feature contrast module, guided by a novel global contrastive loss (GCL), is constructed to quantify the similarity between different feature distributions in order to enhance the inter-class separability between different samples. Concurrently, an adversarial domain adaptation module is utilized to learn the discriminative features shared between domains, aligning the data distributions of the source and target domains. Furthermore, an adaptive factor <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\eta \)</EquationSource> </InlineEquation> is designed to dynamically balance the relative importance between domain alignment and classification performance, mitigating the adverse impacts caused by overly large or small loss terms. Experimental results on the CWRU and PU bearing datasets validate the effectiveness and superiority of the proposed method, achieving average diagnostic accuracies of 99.64% and 80.25%, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bearing fault diagnosis method based on contrastive learning and domain adaptation under variable working conditions

  • 期刊论文

  • 发表日期：

    2026年1月17日

查看全文
  • Xiaolei Pan,
  • Ao Shen,
  • Hongxiao Chen,
  • Kunyi Wu

摘要

Most transfer learning-based bearing fault diagnosis methods under variable working conditions typically focus on domain alignment, while neglecting the inter-class separability among samples. This oversight leads to the degradation of diagnostic accuracy in the target domain. To address this issue, a novel fault diagnosis method based on contrastive learning and domain adaptation network (CDAN) is proposed. A feature contrast module, guided by a novel global contrastive loss (GCL), is constructed to quantify the similarity between different feature distributions in order to enhance the inter-class separability between different samples. Concurrently, an adversarial domain adaptation module is utilized to learn the discriminative features shared between domains, aligning the data distributions of the source and target domains. Furthermore, an adaptive factor \(\eta \) is designed to dynamically balance the relative importance between domain alignment and classification performance, mitigating the adverse impacts caused by overly large or small loss terms. Experimental results on the CWRU and PU bearing datasets validate the effectiveness and superiority of the proposed method, achieving average diagnostic accuracies of 99.64% and 80.25%, respectively.