Fault diagnosis on bearing of electric motor based on DRSN-BIGRU using stator current signals

Abstract To address the challenges of low bearing fault diagnosis accuracy caused by high noise in motor stator current signals and the weak mapping of fault information, a diagnostic model integrating a Deep Residual Shrinkage Network (DRSN) and a Bidirectional Gated Recurrent Unit (BiGRU) is proposed. The model employs DRSN to eliminate noise from the fault current signals and extract deep features, while the BiGRU network performs temporal modeling on the extracted features to further enhance diagnostic accuracy. Experimental results show that the model achieves a fault diagnosis accuracy of up to 97.77%.