A Feature Extraction Algorithm for Rolling Bearing Faults and Its Application

Abstract

Focusing on the difficulty of completely extracting the surface damage caused by rolling bearing lubrication failure, an algorithm for extracting bearing lubrication fault is proposed, which is based on periodic optimum singular value decomposition (O-SVD) cascaded fast spectral correlation (FSC). Initially, conventional T-SVD with energy leakage defects was modified into O-SVD, which was used as the preprocessing unit for signal processing. Then, FSC calculation was performed on the reconstructed signals, eventually obtaining enhanced envelope spectrum with obvious features that could well preserve local details. Simulation and experimental results show that the proposed algorithm allows rather complete extraction of slight fault features resulting from poor lubrication under small sampling length and low signal-to-noise ratio (SNR), and has good applicability in extracting compound and composite fault features. The extracted signals have advantages over existing algorithms regarding fault resolution and local details preservation.

Keywords

• Rolling bearing
• singular value decomposition
• fast spectral correlation
• slight fault
• feature extraction
• underlubrication fault