Transactions of the Japan Society for Computational Engineering and Science
Online ISSN : 1347-8826
ISSN-L : 1344-9443
Preconditioner Using Eigenmode Deflation for Parallel Finite Element Analysis Based on Domain Decomposition
Takumi MURAINaoto MITSUMENaoki MORITA
Author information
JOURNAL FREE ACCESS

2024 Volume 2024 Issue 1 Pages 20241011

Details
Abstract

The Deflated Conjugate Gradient (DCG) method represents is one of preconditioned conjugate gradient method that utilize of known linearly independent vectors. The study proposes subdomain eigenmode deflation preconditioning that employs low-order eigenmodes, obtained from eigenvalue analysis within partitioned domains of parallel finite element analysis, as a set of deflation vectors for the DCG method. The parallel performance of the proposed method is evaluated by the number of iterations required for convergence and the total computation time through its application in structural analysis. Compared to the standard conjugate gradient method, the proposed method reduces the total computation time by up to 85 %.

Content from these authors
© The Japan Society For Computational Engineering and Science
Previous article
feedback
Top