Nonlinear System Identification with Internal Variable Estimation

Abstract

Abstract The paper deals with the identification of nonlinear dynamic systems from samples of inputs and outputs in the presence of additive noise. A new iterative method is presented for the identification using block-oriented models with immeasurable or unobservable internal varia-ble 8 The models consist of various combinations of no-memory gains (of an assumed polynomial form) and linear dynamic systems, respectively. The coefficients of the polynomial nonlinearities and the parameters of linear system pulse transfer functions are adjusted simultaneously to minimize the mean square error criterion. A special mathematical form of these block-oriented nonlinear models enables resulting regression equations for optimal parameters estimations to be linear in all the parameters. Digital computer simulations are included to demonstrate the feasibility of proposed iterative method.