A domain decomposition method with Lagrange multipliers for the massively parallel solution of large-scale contact problems - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2002

A domain decomposition method with Lagrange multipliers for the massively parallel solution of large-scale contact problems

Résumé

Predicting the contact that can develop between various areas of the same system or different subsystems is of significant importance to many applications in civil, mechanical, and biomechanical engineering. These include, to name only a few, the investigation of the failure of bridge joints, the optimization of metal forming processes, the crash analysis of automotive systems, and the solution of biomechanics problems involving human joints, implants, or teeth. The numerical simulation of such problems is known to be computationally intensive, particularly when the underlying finite element models are well resolved, and the timescales are rather long. For this reason, the massive parallelization of contact algorithms has received a well-deserved attention. Currently, most contact codes rely on the penalty method and explicit time-integration for enforcing contact. For this reason, and because the explicit architecture is well suited for massively parallel processing, parallelizing the contact enforcement step in these codes was never a major obstacle. Nevertheless, while explicit solution algorithms can deliver parallel scalability, they cannot achieve numerical scalability. This is because their time-steps, which are restricted by a CFL condition, typically decrease with the mesh size. Hence, the implicit architecture is a necessary condition for achieving numerical scalability. However, even when an implicit scheme can operate with a time-step that is independent of the mesh resolution, it can deliver a scalable performance only when the underlying equation solver can achieve numerical as well as parallel scalability. The objective of this paper is to report on such a solver which is based on a domain decomposition method with Lagrange multipliers. More specifically, our new algorithm integrates various recent developments in the FETI method. We highlight its performance with the solution on massively parallel processors with shared as well as distributed memory of several large-scale contact problems from the automotive industry.
Fichier non déposé

Dates et versions

hal-00325937 , version 1 (29-08-2016)

Licence

Paternité - Pas d'utilisation commerciale - Pas de modification

Identifiants

  • HAL Id : hal-00325937 , version 1

Citer

Karim-Frédéric Traore, Charbel Farhat, Michel Lesoinne, David Dureisseix. A domain decomposition method with Lagrange multipliers for the massively parallel solution of large-scale contact problems. Fifth World Congress on Computational Mechanics (WCCM V), Jul 2002, Vienne, Austria. ⟨hal-00325937⟩
481 Consultations
1 Téléchargements

Partager

Gmail Facebook X LinkedIn More