Scholarlink

<p>Rolling bearings are critical components in industrial equipment, and their reliable operation is essential for ensuring system efficiency and safety. However, existing fault diagnosis methods face major challenges, including limited availability of labeled data and difficulties in extracting multi-scale features from complex vibration signals. To address these issues, this study proposes an IIoT-based bearing fault diagnosis framework via Generative Models and Attention Mechanism (GMAM). The GMAM model employs a conditional variational autoencoder (CVAE) to generate diverse and class-consistent synthetic samples, effectively mitigating the limitations of data scarcity. In parallel, a multi-head self-attention (MHSA) mechanism is integrated to capture long-range dependencies and enhance multi-scale feature extraction. Extensive experiments conducted on two real-world bearing datasets demonstrate the robustness and generalization ability of the proposed model, which consistently outperforms state-of-the-art approaches under varying operating conditions and noise levels.</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

Industrial Internet of Things-based Rolling Bearing Fault Diagnosis Using Generative Models and Attention Mechanism

  • 期刊论文

  • 发表日期：

    2025年7月18日

查看全文
  • Jiamao Yu,
  • Hexuan Hu

摘要

Rolling bearings are critical components in industrial equipment, and their reliable operation is essential for ensuring system efficiency and safety. However, existing fault diagnosis methods face major challenges, including limited availability of labeled data and difficulties in extracting multi-scale features from complex vibration signals. To address these issues, this study proposes an IIoT-based bearing fault diagnosis framework via Generative Models and Attention Mechanism (GMAM). The GMAM model employs a conditional variational autoencoder (CVAE) to generate diverse and class-consistent synthetic samples, effectively mitigating the limitations of data scarcity. In parallel, a multi-head self-attention (MHSA) mechanism is integrated to capture long-range dependencies and enhance multi-scale feature extraction. Extensive experiments conducted on two real-world bearing datasets demonstrate the robustness and generalization ability of the proposed model, which consistently outperforms state-of-the-art approaches under varying operating conditions and noise levels.