An Adaptive Method for Interpolating Reduced-Order Models Based on Matching and Continuation of Poles

Abstract

This work presents an adaptive parametric model order reduction method based on interpolating poles of reduced-order models. The prerequisite of this method is correct and efficient pole matching. We propose to use a branch and bound method to deal with combinatorial explosion in pole matching. A continuation technique is introduced not only to further ease pole-matching, but also to guide the generation of a small set of reduce-order models that represent the parameter space. This method has several advantages over classical projection-based methods, e.g., compatibility with any model order reduction method, constant size of the parametric reduced-order model with regards to the number of parameters, and capability to deal with complicated parameter dependency.