A New Fault Identification Method Based on Combined Reconstruction Contribution Plot and Structured Residual

Abstract

The existing contribution plot-based reconstruction fault identification methods suffer from low identification accuracy and serious trailing effect. In this paper, a new fault identification method is proposed based on a combination of reconstruction contribution plot and structured residual method. The fault direction vector is calculated by utilizing the structured residual method. The reconstruction contribution plot utilizes the obtained fault direction vector to accurately localize the fault variable, where the fault source can be accurately localized subsequently. The experimental results show that, compared with the traditional PCA and PPCA (PCA based on probability) reconstruction contribution method, this algorithm can accurately identify the fault variables, and reduce the influence of the fault variables on the non-fault variables.