Research on fault diagnosis of rolling bearing based on invariant moments of three-dimensional vibration spectrogram

Abstract

Fault diagnosis of rolling bearings is a key issue in the field of engineering. To solve the problem that the accuracy of the current fault diagnosis of rolling bearings is not high and the model construction time is long, This paper proposed a new fault diagnosis method for rolling bearings based on invariant moments of three-dimensional vibration spectrogram. The pseudo-Wigner-Ville distribution time-frequency analysis method was adopted to generate vibration spectrum images of the rolling bearings by means of signal processing. This method extracts the point cloud three-dimensional invariant moments of the vibration spectrogram as the characteristics of the failure mode, and realizes the bearing fault identification with the BP neural network. The experimental results show that the proposed method not only has better recognition rate than the feature extraction method of the two-dimensional Hu invariant moment, but also can effectively identify and classify faults such as inner ring and outer ring, which has strong application value in the fault diagnosis of bearings and other rotating machinery.