A new method for model reduction of LTI interval systems is presented where the denominator and numerator of the reduced order model are synthesized separately. The denominator of the reduced interval model is calculated using Pole clustering technique and numerator by interval matching of time moments and markov parameters with those of the original system. The method has the flexibility of having better approximation in the transient portion of the time response(high frequency portion of frequency response) and/or steady state part of the time response(low frequency portion of the frequency response) by interval matching of appropriate number of initial time moments and markov parameters. The method is simple, efficient and stability preserving. A couple of numerical examples are included to support the proposed method.
Model reduction, Interval systems, Pole clustering, Time moments, Markov parameters
