A Fault Diagnosis Method Based on Dual-discriminator Against Mechanism for Rolling Bearing

Abstract In practical applications, changes in the operating conditions of mechanical equipment render the original fault diagnosis models less applicable to new operating environments. To address this issue, this paper introduces an innovative approach for rolling bearing fault diagnosis. Initially, the white whale algorithm (WWA) optimization is employed to minimize the envelope entropy of the decomposition mode components obtained from variational mode decomposition. The mode component with the lowest envelope entropy is then used for feature extraction. Subsequently, a specialized network architecture is designed, featuring two discriminators that engage in an against mechanism to determine the optimal network configuration (DCGN). The comparative experimental results indicate that this method achieves a high classification accuracy of up to 94.6% in the cross-condition diagnosis task for rolling bearings, thereby validating the effectiveness and reliability of this scheme.