Model reduction for a class of linear dynamic systems (Q1061660)

A model reduction technique to remedy the singularity of reduced-order models is proposed. The approach adopted is based on the least-square fitting of time-moments of the system. The proposed method is also available to stabilize unstable reduced models. This method is superior to the existing techniques as shown in the numerical examples. It will be proved that this method always gives a stable reduced model assuming the original system is stable.