A rolling bearing fault diagnosis model based on WCNN-BiGRU

Aiming at problems such as poor diagnosis effect of simple models and complex structure of hybrid models in the fault diagnosis of rolling bearing vibration signals, this paper establish a fault diagnosis model, which combines WCNN with BiGRU. First, the model extracts feature of the raw data through a one-dimensional wide convolutional network. Then, the reduced-dimensional fault sequence is sent to the BiGRU for further characteristics extracting. Finally, a classifier is used to categorize the fault diagnosis result. A comparison between our model and other models is given on the CWRU dataset. The experimental results show that the model proposed in this paper has a higher fault diagnosis rate than MLP, GRU, VGGNet-11, CNN-GRU, WDCNN, and with a fewer parameters. Besides, the model is applied to the data generated by the PT300 bearing fault simulation platform for fault diagnosis. Results show that the average accuracy of our model on both datasets exceeds 99%.