A method for fault feature extraction of rolling bearings based on generalized demodulation

In order to solve the problem that iterative generalized demodulation spectrum aliasing leads to large error in extracting fault characteristics of rolling bearing with variable speed, an variational nonlinear chirp mode decomposition and generalized demodulation method for fault diagnosis of rolling bearing was proposed. First the rapid spectral kurtosis algorithm was used to determine the fault signal resonance band. A band-pass filter was use to intercept signal resonance band. Then the envelope of the signal was solved, and it is decomposed by variational nonlinear chirp model decomposition. The instantaneous frequency of the signal was estimated, and the instantaneous frequency was used as a phase function to perform generalized demodulation of the signal until the signal was completely decomposed. Finally after all the demodulation signals were summarized, the result of the summation was processed with Fast Fourier Transform(FFT) to extract fault characteristic frequency. Simulation signals and measured signals were used to compare the proposed method with the iterative generalized demodulation. The results show that the proposed method has better noise resistance than the latter, and the maximum error of extracted features is reduced by 10.5hz and 5.75 Hz, respectively. © 2020, Editorial Office of Journal of Vibration and Shock. All right reserved.