Performance Evaluation of the Deflated CG Method Using Subdomain Eigen-mode in Parallel Finit Element Method

Abstract

The deflated conjugate gradient (CG) method is one of the linear solvers for numerical simulations, which can improve the convergence rate since the condition number would be reduced by inputting a known independent basis of the solution space. In this study, we measured the number of iterations and computation time for structural analysis using the deflated CG method with subdomain eigen-mode deflation. As the results, when the appropriate number of known basis are input, the proposed method leads better convergence and computational performance.