Neuro-fuzzy based identification of Hammerstein OEAR systems

Abstract

This paper considers the parameter identification of neuro-fuzzy based Hammerstein output error auto-regressive (OEAR) systems by combining multiple signal source separation principle and auxiliary model identification idea. The unmeasurable internal variable is replaced by the correlation function of input and output data, then correlation analysis method is adopted to identify the parameters of linear part. In order to solve the parameter identification of the nonlinear part and the noise model, this paper presents a recursive generalized least squares algorithm based on auxiliary model. The convergence analysis in stochastic process theory shows that the parameter estimation error converges to zero under the persistent excitation condition. Examples results indicate that the proposed algorithm has significant advantages of good recognition accuracy to noise disturbance.