[en] Stability and convergence of the interior penalty (IP) discontinuous Galerkin method are closely
related to the penalty coefficient sigma_f . Whereas Shahbazi  has derived nearly optimal values of sigma_f for elliptic problems (assuming a constant viscosity parameter), we propose a generalisation of this definition, in order to take into account mesh anisotropy on the one hand, and strong variations of the diffusivity on the other hand, typical for Reynolds-averaged Navier-Stokes models, in particular the Spalart-Allmaras model. The adequacy of this new definition is illustrated by the application to benchmark 2D computations. Finally, a comparison with two state of the art finite volume solvers is presented for a 3D high-lift cascade flow.
Fonds de la Recherche Scientifique (Communauté française de Belgique) - F.R.S.-FNRS