Contribution Analysis of Sensor Fault Diagnosis Based on Improved Relative ReconstructionAbstractFigures from this paper

Abstract

In the use of the principal component analysis method to fault diagnosis of multiple sensors, when a fault happened, the fault data is reconstructed, the reconstructed contribution rate determined which sensor was failed. But the contribution rate contained variables that was independent of the direction of the fault and the cross terms with other fault directions. There was a small difference in the contribution value of the failure data, which could lead to false report. In order to solve the above problems, after the principal component analysis of the failure data, the improved relative reconstruction contribution rate was calculated, and the correct diagnosis rate was improved by the simulation experiment.