Abstract:

Aiming at the feature extraction of Morlet wavelet transform results,the wavelet coefficient matrixobtained by the continuous Morlet wavelet transform is decomposed by singular value decomposition (SVD).Therelationship among the singular value,the feature signal and the noise in the Morlet wavelet transform results isanalyzed.Based on this relationship,the effective feature information of wavelet transform results can be clearlyextracted by selecting appropriate singular values for SVD reconstruction.Further calculation is carried out to obtainthe frequency- energy spectrum,and the shock feature can be extracted according to the peak position of this spec-trum.Finally,the proposed method is applied to the fault feature extraction of bearing vibration signals and is com-pared with other methods.The results show that the proposed method can extarct the distinct fault waveforms andachieve a very good effect on fault feature extraction at a low signal- to- noise ratio(SNR).

Key words: feature extraction, Morlet wavelet, singular value decomposition, frequency- energy spectrum