On a parallel multilevel solver for linear elasticity problems (Q2760351)
From MaRDI portal
| File:Ambox important.svg | This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use this page instead for the normal view: On a parallel multilevel solver for linear elasticity problems |
scientific article; zbMATH DE number 1684509
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | On a parallel multilevel solver for linear elasticity problems |
scientific article; zbMATH DE number 1684509 |
Statements
19 December 2001
0 references
linear elasticity
0 references
additive multilevel iteration method
0 references
convergence
0 references
non-uniform prismatic mesh
0 references
elliptic operator
0 references
hierarchical basis
0 references
stiffness matrices
0 references
splitting
0 references
iterative two-level preconditioner
0 references
optimality conditions
0 references
parallelization
0 references
bridge foundation modeling
0 references
Cray T3E-600
0 references
Sun ES/4000
0 references
0 references
0 references
0.9341097
0 references
0.93306494
0 references
0.9308462
0 references
0 references
0.91952324
0 references
0.9172065
0 references
On a parallel multilevel solver for linear elasticity problems (English)
0 references
The author applies the additive multilevel iteration method to parallelly solving large scale linear elasticity problems. The method is based on a finite sequence \(\{T^i\}\) of three-dimensional meshes consisting of prisms. The meshes are constructed as tensor product of a fixed one-dimensional grid \(T_z\) and a sequence of nested two-dimensional triangulations \(T_{xy}^i\). The linear elasticity operator is approximated by a simpler elliptic operator splitted into an \(xy\)-plane part and a \(z\)-direction part. By means of a hierarchical basis related to \(\{T^i\}\), the author introduces stiffness matrices \(Q^i\) reflecting the splitting. Then an \(i\)-dependent iterative two-level preconditioner to \(Q^i\) as well as its multilevel modification are presented, and optimality conditions for the proposed method are investigated. After a brief comment on parallelization, attention is paid to model problems arising from bridge foundation modeling. Finally, the author describes the parallel performance of the solver on Cray T3E-600 and Sun ES/4000 computer systems.
0 references