Online reduced kernel principal component analysis for process monitoring

Abstract

Kernel principal component analysis (KPCA), which is a nonlinear extension of principal component analysis (PCA), has gained significant attention as a monitoring method for nonlinear processes. However, KPCA cannot perform well for dynamic systems and when the training data set is large. Therefore, in this paper, an online reduced KPCA algorithm for process monitoring is proposed. The process monitoring performances are studied using two examples: a numerical example and Tennessee Eastman Process (TEP). The simulation results demonstrate the effectiveness of the proposed method when compared to the online KPCA method.