Bearing Fault Diagnosis Method Based on Hilbert Envelope Demodulation Analysis

Abstract

A digital envelope analysis method based on Hilbert transform is proposed to solve the problems of bearing fault signals such as non-stationary, faintness and non-linearity. This method firstly compares the power spectrums between the fault signal and the normal signal. In order to improve the signal-to-noise ratio, a frequency band with a large difference between the two power spectrums is selected. Secondly, the frequency band is used to remove the high frequency carrier and retain the low frequency part which contains fault information only. In order to avoid the wrapping effect in the circular convolution, the tail of the frequency band is filled with zero to double its length. Then the inverse Fourier transform and the Fourier transform are applied to the transformed signal. Finally, the fault diagnosis is made for its envelope spectrum. This method is applied to the bearing fault signal of case Western Reserve University in the United States. The results show that this method can diagnose the fault type and fault location well and has high sensitivity.