Computational costs for large structure frequency response methods

Abstract

The approach most often used for computing frequency response of complex structures over large frequency ranges is a modal approach. Global modes with natural frequencies up to a cutoff frequency are obtained, and the frequency response problem is projected onto the truncated modal subspace. Recently another method known as Adaptive Multi-Level Substructuring has been developed, in which frequency response \s represented in terms of modes of many substructures on multiple levels. An adaptive procedure is used to identify an optimal set of substructure modes for each frequency, for representing response with a userspecified error tolerance. This paper examines the relative computational costs of the two approaches for an example problem. It is found that the Adaptive Multi-Level Substructuring approach requires much less computer time than the global modal approach, as it is implemented in commercial NASTRAN software.