Dear Friends,

I am developing a preconditioned solver for inviscid and viscous flows. I am using the Weiss-Smith preconditioner with Roe scheme for the simulations. In tesing the code with inviscid flows past a bump (Ni's classical test case) and the flow past a cylinder, I found that the pressure contours were quiet noisy for M=0.0005, but the velocity contours were good. However, the pressure and velocity contours for the higher mach numbers like M=0.005 and 0.05 were good. The problem is noise in pressure contours when the mach number goes of the order of 1e-4 or below, though the velocity contours and surface pressure distributions are really good.

Any advice and suggestions will be appreciated

Thanks in advance

Ganesh

Dear Ganesh,

For preconditioned Roe scheme, dissipation terms are modified only. I hope you are OK with this. You must obtain smooth pressure (and other flow variables)distributions at mach numbers down to 10^-6. For external flows, the use of modified form of characteristic BC are essential. However, modified non-characteristic BCs work very well as well (if BC are placed far away from the model (~20 chords)). Turkel's article has a simple form for BCs in case of both external/internal flows.

Regards.

This is an indication that preconditioning is not working at the lower mach numbers.

You can try the preconditioned roe flux given by viozat

http://citeseer.ist.psu.edu/312585.html

They also give a different perspective on preconditioning in terms of asympotic behaviour of compressible flows at low mach numbers. Examples of noisy pressure contours are also given in this paper.