A reduced order discrete-time approximation of continuous-time models

In this paper, a technique to obtain a reduced order discrete approximation of a continuous-time model is proposed. This technique is based on the search of the best fit of the frequency response at a finite and probably large set of frequency points. The approximation uses orthogonal bases, where prior knowledge about the underlying system dynamics can be integrated. This strategy can be combined with an iterative procedure to obtain a discrete-time model of increased accuracy. The proposed technique also allows to emphasize the accuracy of the model in prescribed frequency bands, depending on the particular application.