IDEMAS: Industrial demonstration of accurate and efficient multidimensional upwind and multigrid algorithms for aerodynamic simulation on unstructured grids

• Unité de recherche: PCRD EU
• Numéro de projet: 97.0427-1
• Titre du projet anglais: IDEMAS: Industrial demonstration of accurate and efficient multidimensional upwind and multigrid algorithms for aerodynamic simulation on unstructured grids

Mots-clé (Anglais)

Parallel computing; computational fluid dynamics; domain decomposition methods;

Autre Numéro de projet (Anglais)

EU project number: BRPR-CT97-0591

Programme de recherche (Anglais)

EU-programme: 4. Frame Research Programme - 2.1 Industrial and materials technologies

Description succincte (Anglais)

See abstract

Partenaires et organisations internationales (Anglais)

Coordinator: Von Karman Institute for Fluid Dynamics (B)

Résumé des résultats (Abstract) (Anglais)

In this project, we have developed, analyzed and implemented parallel methods for the solution of large-scale linear systems arising from computational fluid dynamics (cfd) problems. The techniques adopted are based on domain decomposition methods. In particular, we have developed and implemented a parallel matrix-free schur complement (sc) solver. The starting point was a code where parallelism was achieved by a block-jacobi procedure and anributing each jacobi block to a processor. We have enhanced the linear solver by implementing both parallel matrix-free sc based solvers and parallel preconditioners based either on multilelevel sc techniques or on schwarz iterations. These different techniques have been tested and compared on realistec aeronautical problems.
Moreover, since the performance of these preconditioners degradates as the number of processor grows, we have analyzed the effects of a coarse correction that is formed using algebraic procedures. Using this procedure we avoid to build a coarse grid, step that can be fairly expensive for unstructured 3d meshes.
Numerical experiments confirm the good properties of the proposed methods. All the tests have been run in a parallel environment using the mpi library.

Références bases de données (Anglais)

Swiss Database: Euro-DB of the
State Secretariat for Education and Research
Hallwylstrasse 4
CH-3003 Berne, Switzerland
Tel. [phone]
Swiss Project-Number: 97.0427-1