Adaptive multipreconditioned FETI: scalability results and robustness assessment

Christophe Bovet (1, 2) , Augustin Parret-Fréaud (3) , Nicole Spillane (4) , Pierre Gosselet (2)

1 ONERA - The French Aerospace Lab [Châtillon]
2 LMT - Laboratoire de Mécanique et Technologie
3 Safran Tech
4 CMAP - Centre de Mathématiques Appliquées - Ecole Polytechnique

The purpose of this article is to assess the adaptive multipreconditioned FETI solvers (AMPFETI) on realistic industrial problems and hardware. The multi-preconditioned FETI algorithm (first introduced as Simultaneous FETI [1]) is a non-overlapping domain decomposition method which exhibits good robust-ness properties without requiring the explicit knowledge of the original partial differential equation, or any a priori analysis of the algebraic system through eigenvalues problems. Multipreconditioned FETI solves critical problems in significantly fewer iterations than classical FETI but each iteration involves a larger computational effort. An adaptive strategy (known as the adaptive mul-tipreconditioned conjugate gradient algorithm [2]) has been proposed to achieve balance between robustness and efficiency and we will observe that it provides an efficient solver for the problems considered here.

Mots clés: Domain decomposition; FETI; Krylov subspace methods; multipreconditioning; adaptive multipreconditioning

Domaines: Analyse numérique [math.NA]; Calcul parallèle, distribué et partagé [cs.DC]; Informatique [cs]; Mathématiques [math]; Matériaux et structures en mécanique [physics.class-ph]

Christophe Bovet, Augustin Parret-Fréaud, Nicole Spillane, Pierre Gosselet. Adaptive multipreconditioned FETI: scalability results and robustness assessment. Computers & Structures, 2017, 193, pp.1-20. ⟨10.1016/j.compstruc.2017.07.010⟩. ⟨hal-01458725⟩