Improved Frequency Weighted Model Order Reduction Techniques

Abstract

The main drawback of the existing pioneer frequency weighted model reduction technique by Enns was to yield an unstable reduced-order model from the original stable system. Many existing techniques address this limitation and preserve stability in reduced models but at the cost of large approximation error. In this paper, three frequency weighted model reduction techniques are proposed for continuous-time systems. Using these techniques, comparatively low approximation error is achieved as compared to existing stability preserving frequency weighted model reduction techniques along with a satisfactory error bound for stable reduced-order models. Different examples are also presented which shows the efficiency of the proposed work.