A New Method for Model Order Reduction

Abstract

A new frequency domain method is proposed for reduced order modelling. In addition to using time moments that determine the low frequency matching, consideration of the Markov parameters is required to give better, accurate initial transient response matching and in some cases to provide stable models. We define a new set of parameters called the approximate generalised Markov parameters (AGMPs) and use them in conjunction with the approximate generalised time moments (AGTMs) to obtain reduced order transfer function models for high order systems. The method is applied to some typical examples from the literature.

Indexing terms:

• Large-scale systems
• Model order redaction
• Linear systems