Rolling bearing fault diagnosis based on generalized multiscale mean permutation entropy and GWO-LSSVM

Faults in rolling bearings are usually observed through pulses in the vibration signals. However, due to the influence of complex background noise and interference from other machine components present in measurement equipment, vibration signals are typically non-stationary and often contaminated by noise. Therefore, in order to effectively extract the random variation and non-linear dynamic variation characteristics of vibration signals, a new method of rolling bearing fault diagnosis based on generalized multiscale mean permutation entropy (GMMPE) and grey wolf optimized least squares support vector machine (GWO-LSSVM) is put forward in this paper. Based on the multiscale permutation entropy (MPE), the multiscale equalization is firstly used to replace the coarse grained process, and the value of mean is extended to variance to avoid the dynamic mutation of the original signal. Finally, the parameters of LSSVM are optimized by the grey wolf optimization algorithm to achieve accurate identification of fault modes. The results of simulation and experiment show that applying the proposed GMMPE to rolling bearing fault feature extraction is feasible and superior, and the method based on GMMPE and GWO-LSSVM has better noise robustness, which can effectively achieve rolling bearing fault diagnosis.