Scholarlink

Rotating machinery is widely used in the industry, but its poor working environment is easy to cause failures. Traditional fault diagnosis methods rely on signal processing technologies (such as Fourier transform and wavelet transform). With the development of depth learning, depth learning is also applied to fault diagnosis problems. This paper presents a hybrid model (DWT-1DCNN-LSTM) combining DWT, 1D-CNN and LSTM for fault diagnosis of rotating machinery. Firstly, the vibration signal is decomposed into low-frequency and high-frequency components by DWT to enhance the time-frequency characteristics; Then, the local feature is extracted by using 1Dand the temporal dependency is captured by LSTM; Finally, the soft max classifier is combined to realize fault classification. The experiment uses CWRU and HUST bearing data sets, and the results show that the method has a high accuracy on CWRU and HUST data sets, and its universality is verified. In addition, in the test with white noise (50–30 dB), the accuracy of the model fluctuated slightly and showed strong robustness.

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

Research on Fault Diagnosis of Rotating Machinery Based on DWT and 1D-CNN-LSTM

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Wei Zhang,
  • Jiangui Li,
  • Zhangjie Li,
  • Chengwen Sun,
  • Xiaolei Shi,
  • Huiliang Liu

摘要

Rotating machinery is widely used in the industry, but its poor working environment is easy to cause failures. Traditional fault diagnosis methods rely on signal processing technologies (such as Fourier transform and wavelet transform). With the development of depth learning, depth learning is also applied to fault diagnosis problems. This paper presents a hybrid model (DWT-1DCNN-LSTM) combining DWT, 1D-CNN and LSTM for fault diagnosis of rotating machinery. Firstly, the vibration signal is decomposed into low-frequency and high-frequency components by DWT to enhance the time-frequency characteristics; Then, the local feature is extracted by using 1Dand the temporal dependency is captured by LSTM; Finally, the soft max classifier is combined to realize fault classification. The experiment uses CWRU and HUST bearing data sets, and the results show that the method has a high accuracy on CWRU and HUST data sets, and its universality is verified. In addition, in the test with white noise (50–30 dB), the accuracy of the model fluctuated slightly and showed strong robustness.