A novel method of fault diagnosis for rolling element bearings based on the accumulated envelope spectrum of the wavelet packet

Envelope analysis is an effective technique of fault diagnosis for rolling element bearings (REBs). However, envelope analysis needs to select an appropriate frequency band of the signal. Numerous selection methods have been investigated, such as spectral kurtosis (SK) and wavelet packet kurtogram (WPK), which are based on kurtosis. Nevertheless, existing approaches are sometimes unable to identify bearing faults due to the contamination of discrete and random noises. In this paper, a novel method of fault diagnosis for bearings is proposed, which accumulates all or part of sub-band signals’ envelope spectrums at a given level in wavelet packet rather than demodulates one selected frequency sub-band signal. Two simulated signals and a real signal are analyzed to test the performance of the novel method. In the first case, stochastic impulses are added into the simulated outer race fault signal. The analysis shows that the novel approach is more robust to stochastic impulses than the other two methods such as SK and WPK. In the second case, the simulated signal containing two outer race faults from different REBs is analyzed using the three methods, respectively. The results show that the novel approach captures more useful information and detects the two different REBs’ faults contained in the same signal more effectively. Furthermore, the effectiveness of the novel method is validated by identifying the characteristic fault frequency of the real REB.