Improved kernel principal component analysis for fault detection

This paper improves kernel principal component analysis (KPCA) for fault detection from two aspects. Firstly, a feature vector selection (FVS) scheme based on a geometrical consideration is given to reduce the computational complexity of KPCA when the number of samples becomes large. Secondly, a KPCA plus Fisher discriminant analysis (FDA) scheme is adopted to improve the fault detection performance of KPCA. Simulation results are given to show the effectiveness of these improvements for fault detection performance in terms of low computational cost and high fault detection rate.