Scholarlink

To address issues such as a lack of prior knowledge, noise interference, and difficulties in early fault feature extraction, a rolling bearing fault diagnosis technique utilizing an optimized maximum correlated kurtosis deconvolution (MCKD) and multi-scale depth-wise separable convolutional neural Network (MDSCNN) is proposed in this study. The parameters of MCKD are optimized adaptively by using the sparrow optimization algorithm, and the optimized MCKD is employed to filter the condition monitoring (CM) signal for fault feature enhancement. The filtered signals are then used as the input to train the MDSCNN and for fault classification. The validity of the proposed algorithm is examined using a published bearing dataset where an accuracy as high as 99.87% is achieved. Furthermore, the superiority of the proposed algorithm in bearing fault diagnosis is verified by comparing the diagnostic result with those using other commonly employed techniques in a comparison study.

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 Bearing Fault Diagnosis Technique Based on an Optimized MCKD and Multi-scale DSCNN

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hongjiao Luo,
  • Yajun Shang,
  • Kailin Jiang,
  • Yiming Chen,
  • Tian Ran Lin

摘要

To address issues such as a lack of prior knowledge, noise interference, and difficulties in early fault feature extraction, a rolling bearing fault diagnosis technique utilizing an optimized maximum correlated kurtosis deconvolution (MCKD) and multi-scale depth-wise separable convolutional neural Network (MDSCNN) is proposed in this study. The parameters of MCKD are optimized adaptively by using the sparrow optimization algorithm, and the optimized MCKD is employed to filter the condition monitoring (CM) signal for fault feature enhancement. The filtered signals are then used as the input to train the MDSCNN and for fault classification. The validity of the proposed algorithm is examined using a published bearing dataset where an accuracy as high as 99.87% is achieved. Furthermore, the superiority of the proposed algorithm in bearing fault diagnosis is verified by comparing the diagnostic result with those using other commonly employed techniques in a comparison study.