A Scalable Balancing Domain Decomposition Based Preconditioner for Large Scale Heat Transfer Problems

概要

論文の詳細を見る
An efficient and scalable Balancing Domain Decomposition (BDD) type preconditionerfor large scale linear systems arising from 3-dimensional heat transfer problems ispresented. The new method improves parallel scalability of BDD by employing anincomplete balancing technique to approximate a coarse space problem and a diagonalscaling to precondition the local fine space problems instead of the Neumann-Neumannpreconditioner. It may increase the number of iterations but reduces the computationcosts of the precondition process for each iteration. Consequently, total computationtime and required memory are expected to be reduced. The convergence estimates mayalso be independent of the number of subdomains. We have implemented thisalgorithm on the parallel processors and have succeeded in solving some illconditionedlarge scale heat transfer problems.
一般社団法人日本機械学会の論文
2006-05-15