Aiming at the problem that the fault diagnosis rate is low due to the lack of negative samples of mine ventilator bearings in practical applications,a fault diagnosis method for mine ventilator bearings based on deep transfer learning was proposed.Combining convolutional neural network (CNN) with bidirectional gated recurrent unit (BiGRU) and replacing the Softmax layer of CNN with random forest (RF) classifier,a CNN-BiGRU-RF diagnostic model was built to extract deeper bearing fault features for fault identification.The source domain data were used to train the model to determine the model structure parameters.Finally,transfer learning was introduced to migrate the model to the target domain,the labeled data in the target domain were used to fine-tune model parameters,and a target domain diagnostic model was built for fault classification.The experimental results show that the average fault identification accuracy of the proposed method is more than 94% when the negative samples of the mine ventilator bearing are sparse.Compared with other methods,the proposed method has better diagnostic accuracy and generalization ability.