An Improved Residual Network for Bearing Fault Diagnosis in Strong Noise Background

In bearing fault diagnosis, the residual network has achieved certain results. However, in strong noise environment, the fault diagnosis accuracy of traditional residual network is not high. Therefore, this article proposes an improved residual network with multi-channel (MCRN). In MCRN, each layer establishes a directly connection with input data, and forming a multi-channel feature extraction mode. Then, we design the multi-channel integration algorithm to extract fault features of signals. Thereby MCRN can obtain more complete fault information from input data. Experiments are conducted on the IMS datasets, and the effectiveness of MCRN is demonstrated. Compared with traditional methods, when the signal-to-noise ratio reaches −4 db, the accuracy of MCRN keeps over 95%.