2014SNA + MC 2013 - Joint International Conference on Supercomputing in Nuclear Applications + Monte Carlo
|Number of page(s)||7|
|Section||4. Advanced Parallelism and HPC Strategies: c. Advanced HPC Strategies for Applications|
|Published online||06 June 2014|
Hybrid parallel strategy for the simulation of fast transient accidental situations at reactor scale
1 CEA, DEN, DANS, DM2S, SEMT, DYN, Gif sur Yvette, France, F-91191
2 LaMSID, UMR EDF-CEA-CNRS 2832, Clamart, France, F-92141
3 CEA, DEN, DANS, DM2S, STMF, LATF, Gif sur Yvette, F-91191
4 EDF R&D, Analysis in Mechanics and Acoustics, Clamart, France, F-92141
5 EDF, Service Etudes et Projets Thermiques et Nucléaires, Villeurbanne, France, F-69628
6 INRIA, EPI MOAIS, Laboratoire d’Informatique de Grenoble, Montbonnot Saint Martin, France, F-38330
* Corresponding Author, E-mail: email@example.com
This contribution is dedicated to the latest methodological developments implemented in the fast transient dynamics software EUROPLEXUS (EPX) to simulate the mechanical response of fully coupled fluid-structure systems to accidental situations to be considered at reactor scale, among which the Loss of Coolant Accident, the Core Disruptive Accident and the Hydrogen Explosion.
Time integration is explicit and the search for reference solutions within the safety framework prevents any simplification and approximations in the coupled algorithm: for instance, all kinematic constraints are dealt with using Lagrange Multipliers, yielding a complex flow chart when non-permanent constraints such as unilateral contact or immersed fluid-structure boundaries are considered. The parallel acceleration of the solution process is then achieved through a hybrid approach, based on a weighted domain decomposition for distributed memory computing and the use of the KAAPI library for self-balanced shared memory processing inside subdomains.
Key words: Large scale fluid-structure systems / fast transient accidental situations / hybrid parallel solver
© Owned by the authors, published by EDP Sciences, 2014