Optimal Hermitian Wavelet Filtering and Envelope Spectrum Based Bearing Defect Diagnosis

Abstract

Local faults in rolling bearings often give rise to periodic impulses in vibration signal. The periodic impulses can be extracted by resonance demodulation technique. Therefore, the resonance demodulation technique is a popular technique of rolling bearing defect recognition and diagnosis. However, the selection of optimal band-pass filter parameter is usually more difficult. The proposed method is the union of the continuous Hermitian wavelet transform as time frequency tool and the spectral kurtosis as transient impulse indicator. Firstly, continuous Hermitian wavelet transform is utilized to process the vibration signal and the wavelet amplitude and phase diagram are obtained. Secondly, the spectral kurtosis is calculated to select the best band-pass filter parameters. The proposed procedure and spectral kurtosis are used to optimize the parameters of continuous Hermitian wavelet for bearing defect detection. The bearing failure test results show that the proposed method can effectively extract the transient features and diagnose the bearing fault.