Paul F. Fischer
(Mathematics and Computer Science Division, Argonne National Laboratory)
Spectral Element Methods for Transitional Flows
Abstract
We consider spectral element solutions of the unsteady incompressible
Navier-Stokes equations in general two- and three-dimensional domains.
The spectral element method is based upon rapidly convergent, high-order
weighted residual techniques employing tensor-product polynomial bases
of degree N in each of K deformed quadrilateral or hexahedral elements.
Because of its low numerical dissipation and dispersion, the method is
well suited to transitional Reynolds number applications where the physical
viscosity is small. For high Reynolds numbers, the Galerkin formulation
is stabilized using a recently developed filter that is local and preserves
both interelement continuity and spectral accuracy. Time advancement is
based on consistent, high-order, operator splittings that permit large
time steps (typ. a convective CFL of 1--5) and yield independent convective,
viscous, and pressure subproblems, the latter of which are solved using
Jacobi and Schwarz-preconditioned CG, respectively. The code scales to
1000s of processors, achieving performance in excess of 400 GFLOPS on 4096
processors of the Intel ASCI Red machine at Sandia National Laboratory.
We present several example applications that demonstrate current capabilities,
including flow in an arteriovenous graft model at Reynolds number 2700,
turbulent bouyancy driven convection at Rayleigh number 10^6, and hairpin
vortex formation in the wake of a roughness element for a boundary layer
at Reynolds numbers 500--3500. |
|