Application of a model order reduction method based on the Krylov subspace to finite element transient analysis imposing several kinds of boundary condition

Abstract

Model order reduction (MOR) via Krylov subspace (KS-MOR) is one of projection-based reduction method for spatially discretized time differential equation. This paper presents a treatment of KS-MOR incorporating with finite element method for structure dynamics. KS-MOR needs basis vectors for the projection into Krylov subspace. In this context, Arnoldi and/or Lanczos method are typical techniques to generate basis vectors, and these techniques requires the information of right hand side (RHS) vector, which is the loading pattern vector in structure dynamics. In this study, we propose a treatment of Dirichlet boundary problem by generating an equivalent blocked system equation including three RHS vectors. In order to solve the multiple RHS vector problem, Block Second Order Arnoldi (BSOAR) is utilized in this paper. After projection, time integration of the projected small system equations was performed by the conventional Newmark-β method. In order to show the performance of KS-MOR, several numerical simulations are conducted. The numerical results show less than 1% of the original degrees of freedoms (DOFs) are necessary to get the accurate results for all of our numerical examples, and the CPU time is less than 2% of the conventional FE calculation.