Rolling bearing fault diagnosis method based on total variation de-noising and fast spectral correlation

Feature extraction plays a crucial role in rolling bearing fault diagnosis. However, vibration signals measured are complex and non-stationary inherently, and pulse features of faulty bearings are often submerged in noise. Here, in order to effectively extract the fault information of rolling bearings under strong background noise, a fault feature extraction method based on combination of total variation de-noising and fast spectral correlation (TVD-FSC) was proposed. Firstly, the total variation de-noising method was used to de-noise vibration signals and improve their signal-to-noise ratios. Then, the fast spectral correlation analysis was performed for de-noised signals to correctly identify fault feature frequencies of a bearing. Simulation and test results showed that the proposed method can be used to effectively extract the weak fault feature information of rolling bearings, and its analysis results are better than those using the direct fast spectral correlation method and the one combining the wavelet threshold de-noising with the fast spectral correlation, respectively; it provides an effective way for extracting weak fault features of rolling bearings. © 2019, Editorial Office of Journal of Vibration and Shock. All right reserved.