Early fault signals of rolling bearings are easily interfered by noise,and fault impact components are difficult to be extracted and fault identification is difficult.In order to extract characteristic parameters from multi-angle fault bearing vibration signals,the vibration signals were decomposed into several intrinsic mode functions (IMFs) using variational mode decomposition (VMD),and IMF components were screened based on envelope entropy,correlation coefficient and kurtosis.The time domain and frequency domain features of selected IMF,VMD energy entropy of signal and energy ratio of each IMF were extracted to form feature vectors,and the fault information was reflected from time domain,frequency domain and energy perspective.The sparrow search algorithm (SSA) was used to optimize SVM parameters and determine the optimal parameters to overcome the problem of parameter selection.The sample feature vectors were input into SSA-SVM for fault classification.The experimental data of bearing faults show that the average accuracy rate of fault identification by this method is above 98.71%.Compared with single domain feature,this method has better recognition effect on fault type and damage degree.