In order to solve the problem of bearing fault diagnosis with many fault classification model parameters and weak generalization performance,low fault recognition rate and slow speed,a fault diagnosis method based on deep learning model ECA-ResNet,complete ensemble empirical mode decomposition with adaptive noise and the support vector machines optimized by sparrow search algorithm(SSA-SVM) was designed.ECA-ResNet was used to model the bearing signal to extract the frequency domain fault characteristics;the frequency domain characteristics,the energy entropy extracted by using CEEMDAN and the time domain characteristics of the traditional signal were combined to form a feature matrix;the fault types were identified by using SSA-SVM.The results show that compared with the traditional fault feature extraction method,by using the proposed bearing fault diagnosis method,a good diagnosis effect can be get,and the bearing fault recognition rate and classification speed are high.