Comparative Study of Fault Detection Algorithm Based on Multivariate Statistical Analysis

This paper has studied several methods based on multivariate statistical analysis, DPCA, CPCA, and MBPLS, which are all the extension of PCA, mainly introducing the principles and steps. At the same time, a method for determining the threshold in practice is proposed in this paper. Besides, we verify the effectiveness of the detection method by the data of train suspension system from simulation experiment. And then we make a comparative analysis of the results through the effect and time. According to the results, we can find it is obvious that CPCA and MBPLS are superior to DPCA in detecting faults.