Identification of Wiener model using step signals and particle swarm optimization

Abstract

This paper develops a new approach to identify Wiener model. Firstly, a sequence of step signals with various amplitudes are supplied to the system. The structure of the static nonlinear function is obtained from the input step signals and their corresponding steady-state responses. Once the structure of the nonlinearity is determined, the parameters describing the nonlinearity are estimated through minimizing an objective function. Secondly, the parameters of the linear dynamic subsystem are estimated using random input from the view point of optimization. Particle swarm optimization is used to solve the two optimization problems involved in parameter estimation of static nonlinear function and linear dynamic subsystem. The proposed method makes the identification problem of nonlinearity separate from that of linear part and simplifies the identification procedure significantly. Also, it does not require any structure information about the static nonlinear function. Two examples are given to validate the effectiveness of the proposed method.