Aiming at the problem of motor failure caused by complex factors interacting and interfering during the operation of pure electric mining trucks, a method based on principal component analysis (PCA) and random forest (RF) is proposed for predictive diagnosis. A dataset is constructed based on the actual collected motor failures of electric mining trucks, and the eigenvalue extraction and dimensionality reduction of the failure data are carried out using principal component analysis to reduce the dimensional redundancy of the data; the random forest prediction model is used to train and test the dimensionality-reduced data, and to predict the motor failure categories. The results show that the accuracy of motor fault type diagnosis using PCA-RF method reaches more than 97%, which is significantly improved compared with the accuracy of the method without dimensionality reduction processing. The accuracy of the above method for motor fault diagnosis of electric mining trucks is confirmed.