Electric motors are the core equipment of industrial production, and rolling bearings are the key parts that are most prone to failure during the operation of electric motors. In order to accurately diagnose bearing faults and improve equipment reliability, this study extracts features from motor vibration signals through ensemble empirical mode decomposition, and classifies signal features using support vector machines. In addition, an optimized GWO is introduced to improve the hyperparameter settings of the support vector machine model, enhancing the fault classification ability, and ultimately constructing a new diagnosis model. The new model had the highest fault classification accuracy of 96.6%, the highest precision of 94.58%, the highest F1 score of 95.18%, and the shortest running time of 8.07 seconds. In addition, its MSE, RMSE, and MAE for outer ring fault detection were the lowest, at 0.072, 0.268, and 0.189, respectively, with a diagnosis time of 7.33 seconds, significantly better than comparison models. From this, the model can enhance the diagnosis accuracy and efficiency, and also provide an effective solution for motor bearing fault diagnosis in industrial applications.