Improved Dual-channel CNN-BiLstm rolling bearing fault diagnosis study

Inhalt:
Aiming at the problem that the single-scale network model cannot completely extract the feature information and the fault data has time series, an improved Dual-channel CNN-BiLstm rolling bearing fault diagnosis method (Dual-CNN-BiLstm) was proposed in this paper. Firstly, s-transform was used to process the signal as the input of the model. A two-channel network model was used to extract rolling bearing fault features, a large convolution kernel was used to improve the anti-interference ability of the model, and BiLstm was used to extract time series attributes in timefrequency features. At the same time, the attention mechanism module is introduced to pay attention to the regions with large differences and extract fault information to the maximum extent. Finally, a particle swarm optimization algorithm is introduced to optimize the classifier to improve the accuracy of fault identification. To verify the validity of this model, experimental comparisons with other variants of this model on a public data set at Case Western Reserve University show that all of the models presented in this paper have better fault identification performance.