Deep Transfer Learning in Inter-turn Short Circuit Fault Diagnosis of PMSM

摘要

This paper studies the inter-turn short circuit fault identification method of permanent magnet synchronous motor (PMSM) in the case of small fault dataset, and proposes an efficient and accurate inter-turn short circuit fault identification method based on transfer learning and one-dimensional convolution neural network(1d-CNN). Firstly, the 1d-CNN is pre-trained on the big data simulation dataset. Then the pre-trained network is applied to a small real dataset sample by using the transfer learning method, optimizing the network based on L1 regularization, and cost-sensitive loss function strategy. To verify the effectiveness of the designed deep model, this method is compared with other deep learning methods, the test results show that the accuracy of this method is up to ninety-eight percent on the small sample dataset, and it has lower data dependence than the compared methods.