The fault monitoring and diagnosi based on KPLS

Abstract

In this paper, a novel fault monitoring and diagnosis approach based on kernel partial least squares(KPLS) is introduced. Unlike other nonlinear least squares (PLS) techniques, KPLS does not consider any nonlinear systems optimization procedures and has the characteristics similar to that of linear PLS. In this paper, KPLS provides good monitoring performance by finding those latent variables that present a nonlinear correlation with the response variables and at the same time improve model understanding. Simulation results show the proposed method can effectively capture the nonlinear relationship among variables and improve diagnosis performance.