L’objectif de cette rencontre est de créer ou renforcer les liens entre des jeunes chercheuses et jeunes chercheurs francophones qui travaillent sur des sujets de recherche en lien avec l’étude mathématique, l’analyse et la simulation numériques de problèmes de propagation d’ondes. En plus des intervenant·e·s, les journées seront ouvertes à d’autres participant·e·s, tout en gardant un petit groupe d'une trentaine de personnes afin de maximiser les interactions.
Les inscriptions sont désormais closes. Pour toute question, n'hésitez pas à nous contacter.
14:00 — 14:30
14:30 — 16:15Théophile Chaumont-Frelet
14:30 — 14:45Jérémy Heleine
14:45 — 15:30Zakaria Kassali
15:30 — 16:15Tuan Anh Vu
16:15 — 16:45
16:45 — 18:15Théophile Chaumont-Frelet
16:45 — 17:30Morgane Steins
17:30 — 18:15Hadrien Montanelli
19:30 — 23:59
10:15 — 10:45
10:45 — 12:15Marcella Bonazzoli
10:45 — 11:30Nouha Jenhani
11:30 — 12:15Tiphaine Delaunay
12:15 — 14:00
14:00 — 15:30Pierre Marchand
14:00 — 14:45Pierre Amenoagbadji
14:45 — 15:30Alice Nassor
15:30 — 16:00
16:00 — 17:30Pierre Marchand
16:00 — 16:45Margot Sirdey
Three dimensional electromagnetic waves simulation is an important issue in many applications. Industrial simulation frequently involves the solution of a large linear system. When resorting to direct methods (LU decomposition) the necessary memory for the inversion of the matrix increases very quickly with the size of the computational domain. A natural alternative is to use an iterative method such as a GMRES method or a domain decomposition method. However, classical methods (Finite Elements, Finite Volumes, Finite Differences) are not adapted to Krylov-type methods whereas Trefftz methods are. These methods can be interpreted as a Discontinuous Galerkin method whose basis functions are local solutions of the studied equation or as a domain decomposition method. An iterative Trefftz solver whose solution is computed thanks to a preconditioned GMRES method using domain decomposition has been developed. However, iterative Trefftz methods based on a plane wave approximation are ill-conditioned. In this talk, improvements of Cessenat and Després preconditioner and a performant basis reduction will be presented. A special attention will be paid to the memory consumption. A matrix-free strategy allows to avoid the assembly of the matrix associated to the linear system. Also, a new global preconditioner will be presented as an efficient alternative to the Cessenat and Després one. The efficiency of the method will be illustrated by several numerical experiments achieved in HPC context and by comparisons with classical methods.
16:45 — 17:30Philippe Marchner
9:00 — 9:45Jérémy Heleine
9:00 — 9:45Rose-Cloé Meyer
9:45 — 10:00
10:00 — 11:45Jérémy Heleine
10:00 — 10:45Zoïs Moitier
10:45 — 11:30Igor Chollet
11:30 — 11:45Pierre Marchand
12:00 — 13:00
