Bearing Fault Diagnosis Based on 1D-CNN-LSTM-TL

Abstract:
The traditional rolling bearing fault diagnosis method has weak feature extraction ability, less labeled sample data in the actual working environment, and poor generalization of robustness after using transfer learning. Therefore, this paper proposes a bearing fault diagnosis model based on a One-dimensional Convolutional Neural Network (CNN), Long-Short Term Memory (LSTM) and Transfer Learning. The model uses a convolutional neural network to extract features of rolling bearing data and combines an LSTM layer to process the time series features of rolling bearing data. Finally transfer this fault diagnosis model to other fault diagnosis systems through migration learning without freezing any parameters. This paper uses 1D-CNN, 1D-CNN-LSTM, and 1D-CNN-LSTM-TL models to classify and recognize four different working states of rolling bearings. The results show that the model proposed in this paper has strong robustness and generalization.