Accuracy Enhancing Model Reduction Technique for Weighted and Limited Interval Systems with Error Bound.

Existing frequency-weighted and frequency-limited intervals-based model reduction schemes can generate unstable reduced-order models for the continuous time of some stable linear models. Many researchers have offered various solutions to retain the reduced-order model's stability to rectify these main limitations. Conversely, under certain circumstances, existing methods can often yield an unstable reduced model and generate a significant variance from the original model, creating a large approximation error. In this article, the new framework is proposed based on the weighted and limited-intervals Gramians for the stable continuous-time systems. The suggested framework ensures the ROM's stability and ensures that the low frequency-response approximations error is achieved in the desired weights and limited-intervals; furthermore, the proposed framework offers an a priori error-bound expression, which is easy to calculate. The proposed framework yields steady and accurate results compared to conventional schemes, which demonstrate the effectiveness of the proposed framework. [ABSTRACT FROM AUTHOR]