Identification of hysteretic systems using the differential evolution algorithm

Abstract

In system identification, it is well-known that any non-linear in the parameters system, possesses a complicated error surface with many local minima. In this case the conventional algorithms, such as those based on least squares methods are not applicable, and an optimisation scheme is required. The structure of some hysteretic models contains unmeasured states and parameters which enter the equations in a non-linear way that are not differentiable. Therefore, to identify this class of models an optimisation scheme which, does not depend on gradient information, is required. This paper features a study of the application of the differential evolution optimisation algorithm to the Bout-Wen class of hysteretic systems. This study is based on simulated data.