Abstract: A method for predicting the residual life of coated self-lubricating spherical bearings based on convolutional neural network (CNN) and long-short term memory neural network (LSTM) was proposed. Firstly, the failure features of the friction torque signal of the spherical bearing was extracted by CNN. Then the torque signals processed by principal component analysis (PCA) and filtering were input into LSTM neural network for training to obtain the life prediction model of coated self-lubricating spherical bearings, which enabled accurate predictions of the bearing residual life. Finally, based on the accelerated life tests, the reliability of coated self-lubricating spherical bearings was evaluated using a two-parameter Weibull distribution model. The results indicate that coated self-lubricating spherical bearings can maintain long-term stable work at high reliability levels (90%) under light load and low frequency.