ARAS2 preconditioning technique for CFD industrial cases

International audience; A two-level preconditioning technique based on the Aitken’s acceleration of the convergence of the Restricted Additive Schwarz (RAS) domain decomposition method is derived. When it is applied to linear problems, the RAS has a pure linear rate of convergence/divergence that can be enhanced with optimized boundary conditions giving the ORAS method based on the underlying PDE. The RAS method’s linear convergence allows its acceleration of the convergence by the Aitken’s process. In this new two level algebraic preconditioner technique named ARAS2, the coarse grid operator uses only parts of the artificial interfaces contrary to the patch substructuring method. In this way, it can be seen as similar as the SchurRAS method but it differs because the discrete Steklov-Poincar operator connects the coarse artificial interfaces of all the subdomains. Numerical results of the good properties of the ARAS2 preconditioning are provided on industrial problems with no knowledge of the underlying equations.