Identification of Nonlinear Systems Described by Hammerstein Models

Abstract

This paper deals with a method for identification of nonlinear systems suitable to be described by Hammerstein models consisting of a static nonlinearity followed by an ARX linear model. The estimation of the static nonlinearity is carried out supplying the system with a sequence of step signals of various amplitude and determining the corresponding steady-state responses. The estimation of the parameters of the ARX linear system is carried out by means of a least square estimator using data generated supplying the system with a Pseudorandom Binary Sequence (PRBS). The method in question is able to identify static nonlinearities of general type, also with hysteresis and/or discontinuities. Simulation results confirm the validity of the method and its capabilities with respect to some other methods.