MPI Kolloquiumsreihe: Prof. Dr. Axel Klawonn, Robust and parallel scalable domain decomposition methods

20041 1574262604

MPI Kolloquiumsreihe / FMA Kolloquium: Prof. Dr. Axel Klawonn, Robust and parallel scalable domain decomposition methods forproblems in solid mechanics

  • Datum: 19.12.2019
  • Uhrzeit: 17:00 - 18:00
  • Vortragende(r): Prof. Dr. Axel Klawonn
  • Universität zu Köln, Mathematisches Institut, Numerische Mathematik und Wissenschaftliches Rechnen
  • Ort: Max-Planck-Institut Magdeburg
  • Raum: Großer Seminarraum "Prigogine"
  • Gastgeber: veranstaltet vom Max-Planck-Institut Magdeburg und der Fakultät für Mathematik an der Otto-von-Guericke-Universität Magdeburg
  • Kontakt: sek-csc@mpi-magdeburg.mpg.de

Das Max-Planck-Institut Magdeburg lädt Sie herzlich zu seiner öffentlichen Kolloquiumsreihe ein. Hochrangige WissenschaftlerInnen verschiedener Fachgebiete aus renommierten Forschungseinrichtungen aus Deutschland und weltweit präsentieren ihre Forschungsarbeit.

Abstract

The discretization of problems in solid mechanics with finite elements can lead to very large linear or nonlinear systems of equations. In this talk, domain decomposition methods for the solution of these problems will be considered. Here, domain decomposition methods are preconditioners for Krylov space methods and some of them are highly scalable for up to several hundred thousands of cores. FETI-DP and BDDC methods are examples for such methods and they will be described in more detail in this talk together with examples of their parallel scalability. Another issue in the solution of discretized problems from solid mechanics is the robustness of the iterative solvers with respect to certain parameters, e.g., discontinuities in the material coefficients which occur when composite materials are considered. If time allows, new theoretical approaches to adapt some important components of the domain decompositon preconditioners to the specific problem will be discussed as well and their robustness for composite materials will be numerically demonstrated.
 
loading content
Zur Redakteursansicht