Rolling bearing fault diagnosis based on channel attention multi-scale convolution-BiLSTM network

In recent years, deep learning has been used in rolling bearing fault diagnosis widely due to its strong feature learning ability. However, the traditional existing deep learning models lack the ability of feature learning because of their shallow network structure. In view of the incomprehensive extraction of CNN feature information and the lack of feature learning ability of LSTM network, a MCNN-BiLSTM network model based on channel attention is proposed in this paper. Firstly, multi-scale feature information is obtained through multi-scale convolution module with channel attention, and then BiLSTM network is used to obtain the dependency relationship feature through forward and backward directions. Finally, the effectiveness of the method is verified by CWRU datasets, and the experimental results show that this method has strong generalization ability, significantly better diagnostic accuracy than existing methods and higher diagnosis efficiency.