An approach to the closed loop identification of the Wiener systems with Variable Structure Controller using an Hybrid Neural model

The majority of nonlinear systems encountered in the process industry, such as distillation columns, valve for fluid flow control and pH processes can be successfully modelled by the Wiener model which is a linear dynamic subsystem combined with nonlinear static subsystem. This paper proposes an approach to the closed loop Wiener models identification. The Wiener system to be identified is in closed loop with variable structure controller. This controller results in high performance systems that are robust opposite to parameter uncertainties and noise. Furthermore, the control signal is very rich in commutations and is very interest for identification. The proposed identification approach here consists to model the Wiener system by an hybrid neural model which is composed of an ARMA model and a neural network (NN). The NN is used to approximate the nonlinear part. The ARMA model parameters are estimated by a least mean square algorithm whereas The NN is learned by the back-propagation algorithm. To confirm the validity of our approach, a simulation example concerning the identification of a valve for fluid flow control, is provided