LU-SSOR / any experience?

A. Pechloff

Hello.

Concerning 3D compressible turbulent Navier-Stokes flow about aircraft wings, I'm employing an implict LU-SSOR method (finite volume) coupled with a multigrid algorithm. The Spalart- Allmaras tubulence model is implemented. In order to resolve the boundary layer an off body distance of 10^-5 root chord length is selected for the first grid line, resulting in high aspect ratio cells surrounding the wing.

My problem:

In order to obtain a convergent steady state solution (in single grid mode) the implicit dampening parameter (omega) has to be set to a very high value (>20) . In the literature one usually finds values smaller than <2. For my case (>20) the time step of the LU-SSOR scheme is reduced dramatically, resulting in slow convergence. When reducing omega, convergence accelerates, but sudden instabilities in the wing wake occur down the line (residual 10^-3), destroying the solution. I compensate for this by employing the multigrid method. As the grid is restricted, omega can be reduced.

My Questions:

1. Has anyone experienced similar problems, when coupling the S/A turbulence model with LU-SSOR and Multigrid?

2. I'd like to implemented a cell adaptive or flow field adaptive omega. Has this ever been done before (literature)?

3. Would artificial dissipative terms be helpful in stabilizing the solution, eventhough this is not an explicit scheme?

4. Is it wise to solve the S/A turbulence model in the same way as the RANS equations (Roe/Chakravarty)?

Your help is appreciated.

Alexander.