MODEL-REDUCTION VIA FREQUENCY WEIGHTED BALANCED REALIZATION

Abstract

We propose a model reduction method for multi-input multi-output linear time-invariant systems based on frequency weighting and balanced realization. We show that with both input-weighting and output-weighting, the method yields stable reduced order model, provided that the full order model is stable. Examples show the effectiveness of frequency weighting. Finally, our results are compared with those obtained by other frequency weighted model reduction methods.