Envelope analysis by wavelet-filter based spectral kurtosis for bearing health monitoring

Envelope analysis has been widely used in health monitoring and fault diagnosis of rolling element bearings. However, the manual frequency band selection limits its accuracy in extracting transient features from a noisy signal and identifying defects, especially for detecting an incipient defect of the bearing. Because the complex Morlet wavelets have the similar time-frequency structure with the impact energy generated by faulty bearings, we propose an automatically frequency band selection method based on the Morlet wavelets filter and spectral kurtosis. The vibration signals of two bearings are analyzed using the proposed method. And the results indicate that, comparing with the original envelope spectrum and the FIR filter based kurtogram, the proposed approach is effective and practical.