A novel method to identify nonlinear dynamic systems

This paper presents a new method for identifying a nonlinear system using the Hammerstein model. Such model consists of static nonlinear part and linear dynamic part in a cascading structure. The static nonlinear part is modeled by a fuzzy neural network (FNN), and the linear dynamic part is modeled by an auto-regressive moving average (ARMA) model. Based on our approach, a nonlinear dynamical system can be divided into two parts, a nonlinear static function and an ARMA model. Furthermore, a simple learning algorithm is developed for obtaining the parameters of FNN and ARMA model. In addition, the convergence analysis for the cascade model (FNN+ARMA) is also studied by the Lyapunov approach. A simulation result is given to illustrate the effectiveness of the proposed method. Simulation result also demonstrates that this approach is useful for systems with disturbance input.