Generalized CCA with Applications for Fault Detection and Estimation

Abstract

Canonical correlation analysis (CCA) is a well-established multivariate analysis method for finding the relationship between two data sets, which has been explored for fault detection recently. In this paper, we revisit the generalized canonical correlation analysis (CCA) form and discuss its applications for fault detection and estimation. The motivation of using CCA for fault detection is to reduce process uncertainty by taking the correlation coefficients into account. Then, the fault detectability in terms of fault detection rate is increased. Finally, the generalized CCA-based fault detection method is validated on the benchmark, which is a simulation of high-speed trains traction drive control system. The achieved results show that the proposed method is able to successfully detect the faults.