Identification Algorithm of Hammerstein Nonlinear System Using Neural Fuzzy Network and State Space Model

Abstract

This paper discusses a two-stage parameter identification of Hammerstein nonlinear system by using neural fuzzy network and state space model. The developed Hammerstein systems are characterized by static nonlinear subsystems according to neural fuzzy network followed by a dynamic linear block modeled using state space model, and input test signals which is composed of binary signals and random signals are applied to parameter separation identification of the Hammerstein system, that is, estimate separately the underlying nonlinear block parameter and dynamic linear block parameter. To begin with, the parameters of linear block are estimated by way of recursive least square algorithm based on the input-output data of binary signals measurement. Moreover, Taylor series expansion theory and clustering algorithm are used to identify static nonlinear block parameters using random signals. Experiments show that the algorithm can identify the Hammerstein nonlinear system effectively and acquire excellent identification accuracy.