Fault Diagnosis of Rolling Bearings in VMD and GWO-ELM

Abstract

Strong noise, nonlinearity, and complex frequency are the remarkable symbols of vibration signals when rolling bearing failure. So a new method that improved the accuracy of rolling bearing fault diagnosis is proposed based on variational mode decomposition (VMD) and grey wolf algorithm optimized extreme learning machine (GWO-ELM). The bearing vibration signal is decomposed by VMD, and decomposed Intrinsic Mode Component (IMF) is used for fuzzy entropy calculation to construct a multi-scale complexity eigenvector. The fault pattern recognition is carried out in the extreme learning machine afterward. The parameter selection problem in VMD and fuzzy entropy algorithm is analysed through the experimental fault data of the rolling bearing, and the noise robustness of the experimental data is verified. Through fault data of rolling bearings on site, the comparison of the results shows that the GWO-ELM model can effectively identify the abnormal conditions, and has a faster training and learning speed in diagnosing the faults.