Stability analysis of the RBF-ARX model based nonlinear predictive control

This paper gives stability analysis of the nonlinear predictive control strategy based on the off-line identified RBF-ARX model which is a pseudo-linear time-varying ARX model with system working-point dependent Gaussian RBF neural network style coefficients. The predictive controller doesn´t require on-line parameter estimation; it may be applied to a class of smooth nonlinear processes whose working-point varies over a wide range. Stability analysis of the nonlinear predictive controller is given both in unconstrained case and in case of a posterior input constraint. An industrial experiment result of the predictive control design is also revealed for illustrating its effectiveness and feasibility.