Fault tolerant SVM based nonlinear Model Predictive Control

Abstract

This paper describes a nonlinear Model Predictive Controller (MPC) which uses Support Vector Machines (SVM) as a model of the plant. Both batch and on-line trained SVM are presented. The fault tolerance capabilities of this kind of nonlinear MPC are discussed. A strategy for achieving fault tolerance in this nonlinear MPC based on on-line trained SVM is presented. Results obtained under simulation in a continuous stirred tank reactor are discussed.