Continuous-Time Nonlinear Block-Oriented Dynamic System Identification from Sampled Step and Step-Like Responses

Abstract

In this paper an approach to continuous-time nonlinear block-oriented dynamic system identification containing unknown time delay is presented. In the presented approach the excitation being a step or step-like signal is generated from D/A converter equipped with zero-order hold filter. Data used for model identification are sampled step or step-like responses. Parameters of the corresponding models are obtained by an optimisation algorithm utilizing ideas of randomized search. The presented discussion is illustrated by simulation experiments devoted to model identification of Wiener, Hammerstein, Hammerstein-Wiener and Wiener-Hammerstein systems.