Model Reduction Via Limited Frequency Interval Gramians

Abstract

An improved frequency domain interval Gramian-based model reduction scheme for discrete time systems is presented. It is first shown that two of the main results presented in the model reduction method of are incorrect. Improved methods which overcomes these shortcomings are then presented. Improved methods not only yields stable reduced-order models but also have easily computable frequency response error bounds. The method is further extended to 2-D separable denominator system approximation. The simulation results show the effectiveness of the proposed scheme.