Title: Heterogeneous Domain Decomposition for Boundary Layer Problems


		     O.Dia, M. Garbey and Y.Jobic
		       Dept. of Computer Science
		         University of Houston
		   Email: mailto:garbey@tlc2.uh.edu


Abstract:
 we propose an heterogeneous domain decomposition solver for 
Navier-Stokes flow in pipe. The main target application is the computation of 
blood flow in large arteries. The domain of computation is decomposed into a 
regular domain with Cartesian grids and several boundary layer domains that 
fit the boundaries with local orthogonal meshes. The domain decomposition is 
motivated by the physic and/or singular perturbation analysis for large high 
Reynolds numbers. We have then very different type of meshes between the 
regular domain and the boundary layer domain. The numerical efficiency of the 
domain decomposition is the consequences of few factors such as:
(1) each sub-domain can use a fast solver that takes full advantage of 
either the stretching of the mesh in one space direction for boundary 
layer domains, or the regular data structure with Cartesian grids used for 
the main part of the flow.
(2) simplicity of the implementation, grid generation, and memory 
allocation due to the use of the additive Schwarz method for the iteration 
process between overlapping non matching grids.
(3) fast convergence of the domain decomposition algorithm thanks to the 
use of an acceleration procedure to speed up the convergence of the 
Schwarz method.
 We will discuss the numerical accuracy and robustness of this 
heterogeneous domain decomposition technique.