Feature Extraction for Bearing Fault based on Singular Value Product Sequence of Sliding Hankel Matrix

In order to solve the difficulty of frequency baud selection and the problem of in-band noise, a bandpass filtering method based on singular value product sequence of sliding Hankel matrix is proposed. In this work, a singular value product sequence is introduced to locate the fault impact. With this sequence, the resonance frequency can be obtained through spectrum analysis. Then, the optimal band-pass filter is designed with the resonance frequency and the bandwidth is optimized by information entropy. Autocorrelation analysis is performed ou the filtered signal to further suppress the interference in the frequency band. Eventually, the envelope spectrum is applied to fault feature extraction. The performance of the new method is evaluated by the simulated signal and fault bearing information.