Adaptive demodulation synchro-extracting transform for bearing time-varying fault feature extraction

Since bearing fault-induced impulses are time-varying under variable rotational speeds and often polluted by background noise and other interference components, it is still a challenging task for bearing feature extraction and diagnosis. As such, a novel bearing nonstationary fault feature extraction technique, termed adaptive demodulation synchro-extracting transform (ADSET) is developed in this paper. In the developed technique, the reset criterion is defined and introduced into the generalized demodulation transform (GDT) to transform the time-varying fault characteristic frequency (FCF) with strong anti-noise property; In addition, the local maximum criterion is developed and introduced into the synchro-extracting transform (SET) to extract fault-related harmonics and eliminate other interference components; Finally, the diagnosis index model is constructed to detect bearing fault location. The performance of the developed technique is verified using both simulated and different experiment signals, and the results show that rolling bearing fault is quantitatively characterized and accurately detected without rotational speed measurement.