Dynamic Temporal Convolutional Networks for Fault Diagnosis of Rolling Bearings

Since rolling bearing plays a crucial role in rotating machinery, the fault diagnosis research of it has been widely concerned in recent years. However, the collected fault signals often contain a lot of noise interference due to the complex and changeable operating environment. The anti noise capability of traditional deep learning methods is weak, resulting in poor feature extraction performance. Therefore, the dynamic temporal convolutional network (DyTCN) is presented for rolling bearing fault diagnosis in this article. Specifically, the causal dilation convolution is first designed to acquire fault feature of vibration signal by enlarging receptive field of the network. Then, the dynamically adjusted convolution kernel parameters are used to achieve adaptive acquisition of dynamic features and improve the robustness. Extensive experimental results on real dataset demonstrate that DyTCN has excellent accuracy in fault diagnosis.