Modelos autorregressivos para representação de sistemas com histerese

Abstract

Hysteresis is a highly nonlinear phenomenon commonly found in electromagnetic and electromechanical devices which involve memory efects among the output and the history of the process variables. In sensors and electromechanical actuators, this efect is related to the natural memory of inelastic behavior, being the restoring force dependent on the history of deformation. Due to such severe nonlinearity, problems in modeling and control of systems with hysteresis are common and have been exhaustively studied. In the modeling context, the Bouc-Wen model is seen as one of the classic models, being still used in current research. The LuGre model, initially proposed for modeling friction is also used to model hysteresis. However, the use of these models in a feedforward control scenario is a hard task, since its inverse model is not easily obtained. In this sense, auto regressive polynomial models appear as a feasible alternative. However, the reason why there is hysteresis in such model is still not well explained. This thesis presents suficient conditions for hysteresis in polynomial models. It is shown that using multi functions of the first diference of the input are suficient to occur hysteresis in such models. Such conditions are related to the model equilibria, to the forcing function and to certain term clusters in the polynomial models. Besides, a technique for structure selection based on term cluster is proposed in order to identify model structures for systems with hysteresis. The main results of this thesis are used in the identifcation and analysis of nonlinear models estimated fromdata produced by magneto-rheological damper (MRD) models (Bouc-Wen and LuGre model).