Scholarlink

<p>This paper proposes a fault diagnosis method for rolling bearings based on bidirectional two-dimensional principal component analysis (B2DPCA) and deep separable convolution (DSC). Firstly, vibration signals from various faulty bearings are obtained and analyzed using continuous wavelet transform to extract fault features as time-frequency images, containing time-frequency fault information. Secondly, the B2DPCA method is applied to reduce the dimensionality of these time-frequency images, resulting in reconstructed images that maintain or even improve diagnostic effectiveness while having significantly smaller sizes. Finally, a lightweight fault diagnosis model based on DSC is constructed, which has a much smaller number of parameters compared to a conventional CNN. The reconstructed time-frequency images are used to train this model, successfully achieving fault diagnosis. In an experiment recognizing faults in rolling bearings under strong noise, two types of faults with three different severity levels are considered, resulting in an accuracy of 97.60 % for the B2DPCA-DSC method.</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

Fault diagnosis of rolling bearings based on block 2D principal component analysis and depth separable convolution model

  • 期刊论文

  • 发表日期：

    2025年8月2日

查看全文
  • Zhuofei Xu,
  • Dong Liu,
  • Yongfang Zhang,
  • Meng Wang,
  • Qing Huang,
  • Xu Wang

摘要

This paper proposes a fault diagnosis method for rolling bearings based on bidirectional two-dimensional principal component analysis (B2DPCA) and deep separable convolution (DSC). Firstly, vibration signals from various faulty bearings are obtained and analyzed using continuous wavelet transform to extract fault features as time-frequency images, containing time-frequency fault information. Secondly, the B2DPCA method is applied to reduce the dimensionality of these time-frequency images, resulting in reconstructed images that maintain or even improve diagnostic effectiveness while having significantly smaller sizes. Finally, a lightweight fault diagnosis model based on DSC is constructed, which has a much smaller number of parameters compared to a conventional CNN. The reconstructed time-frequency images are used to train this model, successfully achieving fault diagnosis. In an experiment recognizing faults in rolling bearings under strong noise, two types of faults with three different severity levels are considered, resulting in an accuracy of 97.60 % for the B2DPCA-DSC method.