Nonlinear dynamic process monitoring based on kernel partial least squares

Nonlinearity and dynamic are two typical behaviors that widely present in industrial processes. The monitoring performance of multivariable statistical process control techniques will be degraded if those two behaviors are not well addressed. In this paper, a kernel partial least squares (KPLS) based nonlinear state space model is proposed to model the process, which can handle the nonlinear and dynamic data behaviors simultaneously. Due to the non-Gaussian distribution of the nonlinear scores in the KPLS model, support vector data description is introduced for modeling and the corresponding statistic is constructed for monitoring. Two case studies are provided for performance evaluation of the proposed method.