Model structure determination in parametric model identification

The skill of choosing a model structure is one of the important elements in the procedure of parametric model identification. If we do not know the true structure of the system under consideration we usually search the optimal model structure using trial and error technique assuming a criterion of the model quality. The investigation of a suitable model may be done by an "exhaustive search" where all models with admissible structures are estimated or by a certain way of an "integer programming" where a strategy of handling the search is proposed to decrease the number of estimated models. This paper presents a new algorithm for model structure selection in identification of MISO systems. Simulation examples show efficiency of the method.