It is difficult to obtain bearing fault data under actual operating conditions, so a small number of data samples are captured, which leads to over-fitting problems in model training, and the trained model can only diagnose the fault under current operating conditions. In order to improve the adaptability and accuracy of bearing fault diagnosis, the bearing fault diagnosis method based on transfer ensemble learning is proposed in this paper. Firstly, the method completes model training on public datasets. Secondly, through the transfer of task domain and feature space, the problem of poor model adaptability is solved. Finally, the voting mechanism in ensemble learning is reconstructed to improve the model's ability to diagnose bearing fault under actual conditions. The experimental results show that the proposed algorithm has better bearing fault diagnosis ability compared with similar methods.

Fault Diagnosis; Transfer Learning; Ensemble Learning; Voting Mechanism