Enhanced bearing fault diagnosis using adaptive stochastic resonance

摘要

Signal filtering approaches are always used to attenuate noise and enhance useful signal in bearing fault diagnosis. Stochastic resonance (SR) is a nonlinear filter that can utilize noise to enhance weak periodic signal. This study proposes an adaptive SR method which can automatically enhance the bearing fault characteristic frequency (FCF). First, a second-order SR filter is utilized to purify the demodulated vibration signal. Second, a new criterion that measures both the power spectrum kurtosis and the correlation coefficient is proposed to tune the filter parameters in the SR procedure. Finally, the FCF is enhanced, which facilitates bearing fault identification. Experimental verification is conducted to evaluate the effectiveness of the proposed method.