Grant-in-Aid for Scientific Research (B)
|Allocation Type||Single-year Grants |
|Research Institution||KYUSHU UNIVERSITY |
KANAYAMA Hiroshi KYUSHU UNIVERSITY, Faculty of Engineering, Professor, 大学院・工学研究院, 教授 (90294884)
FUJINO Seiji KYUSHU UNIVERSITY, Computing and Communications Center, Professor, 情報基盤センター, 教授 (40264965)
SUZUKI Atsushi KYUSHU UNIVERSITY, Faculty of Mathematics, Research Associate, 大学院・数理学研究院, 助手 (60284155)
TAGAMI Daisuke KYUSHU UNIVERSITY, Faculty of Engineering, Research Associate, 大学院・工学研究院, 助手 (40315122)
OGINO Masao KYUSHU UNIVERSITY, Faculty of Engineering, Research Associate, 大学院・工学研究院, 助手 (00380593)
塩谷 隆二 九州大学, 大学院・工学研究院, 助教授 (70282689)
|Project Period (FY)
2003 – 2005
Completed (Fiscal Year 2005)
|Budget Amount *help
¥15,200,000 (Direct Cost: ¥15,200,000)
Fiscal Year 2005: ¥4,700,000 (Direct Cost: ¥4,700,000)
Fiscal Year 2004: ¥4,800,000 (Direct Cost: ¥4,800,000)
Fiscal Year 2003: ¥5,700,000 (Direct Cost: ¥5,700,000)
|Keywords||domain decomposition method / large scale problem / eddy current problem / thermal convection problem / Stokes equation / iterative solver / balancing preconditioner / parallel computation / 共役勾配法 / BDD法 / 前処理付反復法 / 近似逆行列前処理 / 合同な部分領域への領域分割 / 有限要素法コードのベクトル化|
A.Numerical computations of large scale magnetic problems by finite element method:
First, we have developed computational codes for time-harmonic eddy current problems and nonlinear magnetostatic problems. To make the codes efficient, then we have considered the reduction of their computational costs, and the enhancement of applicable problems, for example, a permanent magnet or a moving obstacle.
B.Numerical solver for linear systems with symmetric coefficient matrices:
We have considered the convergence of MRTR method based on the minimization of residuals. Moreover, we have developed several precondisioners, which come from robust incomplete Cholesky factorization, and have shown their efficiency by appling to some real problems.
C.Application of a substracturing method to discretized Stokes equations:
We have established a coarse space of a balancing precondisioner for applying to discretized Stokes equations.
D.Error estimates of finite element methods for thermal convection problems:
We have established error estimates of a class of finite element methods for thermal convection problems with variable coefficients.
E.Realization of highly efficient BDD methods for large scale structural analysis:
We have developed incomplete BDD methods, improved its parallel efficiency, and realized nonstationary structural analysis with about 10 million DOF problems.