Fault Diagnosis for Rolling Bearings Using Time-Frequency Energy Matrix Difference and CNN

At present, The fault diagnosis of rolling bearing is mainly based on time-frequency. However, due to the large amount of rolling bearing fault data, the data processing speed and fault classification accuracy are not good enough. in this paper, a method combing wavelet time-frequency transform and...

Full description

Saved in:
Bibliographic Details
Published in2023 9th International Conference on Mechanical and Electronics Engineering (ICMEE) pp. 317 - 321
Main Authors Li, Lin, Yuan, Xiaoxi, Yang, Congkun
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.11.2023
Subjects
Bearing fault diagnosis
CNN
Employee welfare
energy matrix difference
Fault diagnosis
Neural networks
Representation learning
Rolling bearings
Time-frequency analysis
wavelet transform
Wavelet transforms
Online AccessGet full text

Cover

More Information
Summary:At present, The fault diagnosis of rolling bearing is mainly based on time-frequency. However, due to the large amount of rolling bearing fault data, the data processing speed and fault classification accuracy are not good enough. in this paper, a method combing wavelet time-frequency transform and CNN classification is proposed for the fault diagnosis of rolling bearing. Firstly, the method transforms the time-frequency matrix into an energy matrix, which effectively reduces the data dimension. Then, the energy matrix is differentiated along the time axis in order to highlight the variation of energy information in the time domain. Finally, these features are fed into a suitable CNN model for fault diagnosis of rolling bearings. By experiments, the effectiveness and superiority of the differential method are verified.
DOI:10.1109/ICMEE59781.2023.10525623