Turbulent flat plate validation, incorrect convergence
 

Hi everyone,

I've been trying to validate SU2 against OpenFOAM using a turbulent flat plate in accordance with http://www.grc.nasa.gov/WWW/wind/val...rb/fpturb.html . I created a mesh using blockMesh and then converted the mesh into a SU2 mesh such that a validation of the two codes could be made.

I received convergence in both cases, but in the case of SU2 I found that the converged solution is quite far from the experimental values and the results of OpenFOAM. I believe the mesh to be reasonably good (y+=~1), so I'm quite at loss what causes the discrepancy in the solution for SU2. In the SU2 case I'm using the SA turbulence model but it shouldn't influence the result this much I believe. Is there something else I could be missing?

My case containing the SU2 and OpenFOAM cases can be found here
https://dl.dropboxusercontent.com/u/...Roe_tf_SU2.zip

Hi Magnus,

Thanks for getting in touch about this. We expect that SU2 should match the experimental results quite well, and we have performed a similar validation case here: http://adl.stanford.edu/docs/display...ent+Flat+Plate.

After a quick check of your config file, it looks like you are using JST for computing convective fluxes. Have you tried computing the flow using 2nd-order Roe? Using the Roe method should automatically apply the appropriate level of dissipation in the boundary layer. Alternatively, you might try to reduce the higher-order dissipation coefficient for JST to a lower level (final value in the option AD_COEFF_FLOW= ( 0.15, 0.5, 0.02 )). You might also consider applying the exact settings from the config file used in the tutorial case linked above.

Hope this helps,
Tom

Hi Tom,

Thank you for your reply! I first tried changing the convective flux scheme into Roe 2nd-order but I received the same convergence.

I then took the test case file and only changed the boundary conditions to fit the Wieghardt ones, this produced better results, but still there is quite some discrepancy against the experimental values and the OpenFOAM values.

http://i.imgur.com/B0CFj5V.png

I also provide the residuals here,

http://imgur.com/B0CFj5Vhttp://imgur.com/B0CFj5V,qRn5mGd#1http://i.imgur.com/qRn5mGd.png

The question at hand is whether the discrepency in the solution with the experimental values might be due to the SA turbulence model or something different? The y+ values are around 2, which I believe to be sufficiently small?

I uploaded the configuration file and the plots in the post if you'd want to take a look at them.

http://imgur.com/qRn5mGd

I see... Are you using the most recent version of the code available on GitHub (https://github.com/su2code)? I would recommend updating to that version and perhaps then you could give the following two things a try:

1. We also have the SST turbulence model available, and you could try it instead of the S-A model by choosing 'KIND_TURB_MODEL= SST' in the configuration file.

2. As we do not have any wall functions available at the moment, you might try with a mesh that has y+ ~ 0.5 to see how the better resolution near the wall affects the results.

All the best,
Tom