抄録
The finite elements are widely applied in various fields of engineering sciences. For large-scale problems, iterative methods are normally implemented to solve a set of discretized linear equations. However, it is known that the convergence rate declines as the iteration loop is processed. Geometric multigrid methods are known to increase this convergence rate by implementing mesh data at several levels. At the same time, in order to carry out computer simulations of complex large-scale problems, parallel computers are necessary to store huge amounts of data and to solve such a problem in a given practical time. In the past, the authors have reported a tool called the Parallel Mesh Relocator which combined the functions of partitioning and refining of a given finete element mesh. In this study, an algorithm is reported which tightly couples this PMR and multigrid solver, enhancing the large-scale parallel finite element analyses.
