Trusted multi-source information fusion for fault diagnosis of electromechanical system with modified graph convolution network

Vibration, current, and acoustic signals have different advantages and characteristics in fault diagnosis. Although a few researches have explored their fusion methods and applied them to fault diagnosis fields in recent years, it remains a knotty problem whether the classification results are trustworthy or not. Therefore, in order to facilitate trusted multi-source information fusion learning and deep sensitive fault feature mining, a modified graph convolution network-trusted multi-source information fusion (MGCN-TMIF) framework is designed. First, the modified graph convolution network is used to deeply mine the relationship between samples through the original signals to obtain the nonlinear evidence. Second, the nonlinear evidence is combined with the Dirichlet distribution to obtain the classification probability distribution. Finally, the evidence is integrated by the reduced D-S evidence theory (DST) to obtain the trusted fusion results. The effectiveness of MGCN-TMIF is verified by experimental-level and industrial-level electromechanical coupling equipment datasets, and the results demonstrate the classification accuracy of the proposed method up to 100 %. The proposed fusion diagnosis method is also verified to have high noise robustness performance through anti-noise experiments.