Rolling Bearing Fault Feature Extraction Based on EMD-CEP-SR Algorithm

Rolling bearing is one of the most important parts of rotating machinery, For the problem that the fault characteristics of rolling bearing vibration signals is difficult to identify due to strong noise and discrete frequency component interference, a method based on Empirical Mode Decomposition (EMD)-Cepstrum Editing Procedure (CEP)-Stochastic Resonance (SR) gearbox rolling bearing fault diagnosis methods. First, the sampled signal is pre-processed by EMD noise reduction based on the correlation number and cliff criterion to highlight the high frequency resonance components; then, the pre-whitening algorithm is used to process the noise-reduced vibration signal to enhance the impact characteristics and generate a signal containing white noise and bearing fault information; finally, the normalized stochastic resonance is combined with the bearing envelope signal as the input of the detection model to enhance the frequency components of bearing fault features. Realize the feature origin of the fault. The feasibility of the algorithm in the fault diagnosis of rolling bearings in gearboxes was verified by simulating and analyzing the rolling bearing data from Case Western Reserve University. Moreover, the local spectral cliffness and local signal-to-noise ratio at the bearing fault feature frequency were used as indicators to compare with the EMD-SK method and the CEP-SR method. The results show that, compared with the two algorithms, EMD-CEP-SR has stronger anti-noise and more superiority.