Fault Feature Enhancement and Extraction using Generalized Demodulation Guided by Ridge Scale Factor Probability Distribution

Shock features induced by bearing early failure are often weak and drowned by the strong background noises; thus, a generalized demodulation (GD) guided by the probability distribution of ridge scale factors is proposed in this paper for the enhancement and extraction of bearing fault features. GD is usually used to process vibration signals of rotating machine working under non-stationary conditions. GD can help to increase the energy concentration of the non-stationary signal, but it depends on an accurate ridge estimation to construct the demodulator. Multiple TF ridges of bearing vibrations are first extracted by a peak search method from the time-frequency representation (TFR) obtained by short-time Fourier transform (STFT). Then, the probability distribution of the scale factors of TF ridges is used to select the optimal ridge to construct demodulators for the GD. Finally, By shifting the extracted components on the TFR, the enhanced extracted signal components can be obtained. Experimental signal processing results show that the method can accurately extract the fault features submerged by noise.