Compound fault features separation of rolling element bearing based on the wavelet decomposition and spectrum auto-correlation

Abstract

In order to separate the compound fault features from a single-channel vibration, the combination of the wavelet decomposition and spectrum auto-correlation method is proposed, based on the wavelet frame theory. Decomposing the compound fault deduced vibration with orthogonal wavelet basis functions, the spectrum auto-correlation method is applied to sub-signals that reconstructed with different scales vibration respectively. Eliminating the tail phenomenon which existed in the result of time domain auto-correlation, the proposed method possesses a more powerful anti-noise capability and highlights the fault deduced impulse feature with primary energy. Based on the analysis of vibration collected on the test rig of 6220 rolling element bearing with inner and outer race defect, the efficiency of the proposed procedure is validated as well. It is shown that the lesser powerful fault induced impulsive feature is restrained in any decomposed sub signals, which actualizes the separation of the compound fault features. Compared with the combination of wavelet and envelope analysis, the proposed method is more powerful with efficient features separation effect and is valuable for the engineering application. © 2013 Journal of Mechanical Engineering.