On shift selection for Krylov subspace based model order reduction

Abstract Mechanical systems are often modeled with the multibody system method or finite element and numerically described of differential equations. Increasing demands on detail resulting high complexity these make use model order reduction inevitable. Frequently, moment matching based Krylov subspaces is used for reduction. There, transfer functions full reduced matched at distinct frequency shifts. The selection shifts, however, not trivial. In this contribution we suggest an algorithm that evaluates increasing number shifts iteratively until a approximates in subspace very low approximation error found. Thereafter, projection matrix spans decomposed singular value decomposition only most important directions retained. way, small models good properties do exceed predefined bound can be found low-error given generated. evaluation more than necessary further by means novelty contribution. paper, novel approach extensively studied and, furthermore, applied to numerical example industrial helicopter model.