Neuro-fuzzy system based identification method for Hammerstein processes

Hammerstein model, which consists of the cascade structure of a static nonlinearity followed by a linear dynamic part, can effectively describe the nonlinear dynamics of many industrial processes. To circumvent the open problems in existing identification methods of Hammerstein processes, Sung developed a new system identification method, which completely separates the identification problem of the linear dynamic part from that of nonlinear static part using a special test signal. However, the polynomials are employed to approximate the nonlinear static function with some conditions that may limit its practical applications. To alleviate this problem, neuro-fuzzy system is employed in this paper to describe the nonlinear static function of the Hammerstein model without any prior knowledge and restriction on static nonlinear function. Furthermore, a non-iterative algorithm is proposed to obtain the neuro-fuzzy system based nonlinear static model. Literature examples are used to illustrate the performance and applicability of the proposed Hammerstein model.