Identification of an Experimental Hybrid System

In this paper we present an experimental study in the identification of an industrial hybrid system. Piecewise ARX models, that consist of a number of ARX models, together with the partition of the regressor space into regions where each of the models is valid, were identified. Effects of dry friction, and mechanical constraints in the experimental setup are demonstrated, and their influence on the identification procedure is discussed. Comparison of the simulated responses of the identified models with the responses of the real system shows that the obtained models are able to describe relevant aspects of the dynamics of the experimental setup. Ways to improve the identification procedure are proposed.