Scholarlink

<p>Bearings are critical components in mechanical systems, directly affecting the efficiency and safety of equipment. Existing fault diagnosis methods rely on signal processing and empirical rules, which often require manual intervention and complex feature extraction, leading to potential information loss. To overcome these limitations, this paper proposes an innovative hybrid fault diagnosis model that combines temporal convolutional network (TCN) and Transformer in a parallel architecture. This structure leverages the advantages of TCN in handling long-range time dependencies, while the self-attention mechanism of transformer captures global features. This effectively addresses the shortcomings of traditional methods in feature extraction. Experimental evaluations on the CWRU dataset show that the proposed hybrid model significantly outperforms traditional methods and standalone deep learning models in classification accuracy, generalization, and robustness. Furthermore, when applied as a pre-trained model to a new environment, it maintains high diagnostic accuracy, even with very limited training data. This effectively solves the bearing fault diagnosis problem under limited sample conditions and makes a meaningful contribution to the field of rolling bearing fault diagnosis.</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

A novel bearing fault diagnosis method using a hybrid TCN-transformer architecture: A deep learning approach

  • 期刊论文

  • 发表日期：

    2025年7月1日

查看全文
  • Liang Jiang,
  • Jun Chen,
  • Haixiao Cao,
  • Peng Li,
  • Keqing Wang

摘要

Bearings are critical components in mechanical systems, directly affecting the efficiency and safety of equipment. Existing fault diagnosis methods rely on signal processing and empirical rules, which often require manual intervention and complex feature extraction, leading to potential information loss. To overcome these limitations, this paper proposes an innovative hybrid fault diagnosis model that combines temporal convolutional network (TCN) and Transformer in a parallel architecture. This structure leverages the advantages of TCN in handling long-range time dependencies, while the self-attention mechanism of transformer captures global features. This effectively addresses the shortcomings of traditional methods in feature extraction. Experimental evaluations on the CWRU dataset show that the proposed hybrid model significantly outperforms traditional methods and standalone deep learning models in classification accuracy, generalization, and robustness. Furthermore, when applied as a pre-trained model to a new environment, it maintains high diagnostic accuracy, even with very limited training data. This effectively solves the bearing fault diagnosis problem under limited sample conditions and makes a meaningful contribution to the field of rolling bearing fault diagnosis.