Rolling Bearing Fault Diagnosis Method Based on GWO-VMD-SVM

As one of the most critical components of high-speed locomotives, wheelset bearings have received increasing attention in recent years. However, the non-smooth vibration signal and heavy background noise make it difficult to mine the weak fault features. In this paper, we take bearings as the research object and investigate the feature extraction and pattern recognition methods for rolling bearings. From the perspective of shallow machine learning, to address the limitations of traditional signal processing methods and the performance of Variational Mode Decomposition (VMD) and Support Vector Machine (SVM) which are easily affected by parameters, a bearing fault diagnosis method using Grey Wolf Optimization (GWO) algorithm to improve the VMD and SVM parameters is proposed. The results show that the GWO-VMD-SVM method improves the accuracy of bearing fault diagnosis compared with EEMD-SVM and VMD-SVM.