SVD-based Frequency Weighted Model Order Reduction of Continuous-time Systems

Some conventional frequency-weighted model reduction methods yield unstable reduced-order models (ROM) even for originally stable systems. Further, in some instances, existing approaches give a stable reduced model that differs significantly from the actual model, resulting in considerable approximation error. To overcome the above issues, this work presents a new frequency-weighted model reduction method using an innovative structure of frequency-weighted Gramians. The suggested approach ensures ROM stability and minimal inaccuracy in frequency-response approximations. The numerical examples support the effectiveness of the proposed technique and compared to conventional methods, the suggested approach offers consistent results, illustrating its usefulness.