[Zhen ZHANG]

Hi all,

I am trying the SST turbulence model with an inlet profile and the nondimensional treatment (FREESTREAM_VEL_EQ_ONE).

I am wondering whether the TKE and Omega in the inlet profile should be nondimensionalized when making the profile. With the dimensional values, the solver does not converge well.

The inlet profile for the SA model seems not to be affected by the nondimensional treatment, is that right?

Thanks

Zhen

[TKatt]

Hi,
from a short look into the code I would say they have to be nondimensional. Which your findings support.



A way to check could be via the automatic example_file writing when no inlet is present. If those values are non-dim then you can be pretty sure that you have to feed non-dim values as well.


The Treatment for SA and SST looks very similar as well ... and there is non-Dimensionalization for nu-tilde so that should apply as well. For the non-dim Factor please see the the Reference value section which is written in the screen output at simulation beginning



Good point btw, I'll take a look how and where we could document that .


Best, Tobi

[Zhen ZHANG]

Hi Tobi,

Thank you for the reply.

I checked the CSolver::LoadInletProfile and find the following:

//for SA
columnValue << config->GetNuFactor_FreeStream()*config->GetViscosity_FreeStreamND()/config->GetDensity_FreeStreamND() <<"\t";
//for SST
columnValue << config->GetTke_FreeStream() << "\t" << config->GetOmega_FreeStream() <<"\t";

So in the example_file the TKE and omega are dimensional, while the SA nu_tilde is nondimensional. I agree that a document is helpful.

Zhen

[TKatt]

Nice find!


Also dont hesitate to change bits of the code, compile and see how the results turn out. In case that helps you could even open a new PR on github (but I recall you mentioned some difficulties with working on github).


I am currently also debugging some things for multiple inlet markers with specified inlet profiles (inc flow + sst + species transport) so I hope to fix that any problems soon.



In any case I'll let you know here

[Zhen ZHANG]

Thanks and good luck!

[TKatt]

Hi Zhen,


the issue will be fixed in develop with https://github.com/su2code/SU2/pull/1427
(Edit: it is now merged into develop, so I hope this fixes it for you)




So what made me state that the values from the inlet file have to be non-dimensional (in case of a non-dim simulation of course) is that after reading the inlet values they are never non-dimensionalized... so they would have worked in case they were non-dimensionalized. The problem is now, that for SST, TKE and Omega are written dimensional for the example file but the solver expects non-dimensional values.


I now changed it that all values in the inlet file are dimensional (and the example writes dimensional values only) and the values are non-dimensionalized directly when they are used in the inlet. That is consistent with the mean flow quantities.


If you can test it and tell me whether that worked for you, that would be awesome. Note that I only tested incompressible flow, but it should have fixed compressible as well

[Zhen ZHANG]

Hi Tobi,

Sorry for late reply. I pulled the develop branch to test it, but the compile failed.

../SU2_CFD/src/numerics/turbulent/turb_sources.cpp:764:62: error: invalid initializer for array member ‘su2double CSourcePieceWise_TurbSST<CEulerVariable::CIndices<short unsigned int> >::Jacobian_Buffer [4]’
idx(val_nDim, config->GetnSpecies()) {

I think the ref values for nu_tilde, TKE, and Omega are correct. And I believe that length_Ref/VelocityRef = ViscRef/(rhoRef*VelocityRef^2). So I think this will work well with my compressible problem.

Zhen

[TKatt]

Do you use gnu/4.8.5 by any chance? I suspect the

Jacobian_Buffer[4] = {0.0};

part to be the baddy. There was a recent fix https://github.com/su2code/SU2/pull/1421 (There an issue is linked with a similar error message). "Newer" compilers (e.g. 5.3.0 works) do not have that problem.



So a quick test would be to just make it

Jacobian_Buffer[4];

and try to compile again. If that helps either you can push a small fix that moves this initial zeroing to the constructor or I can do it. Note though without zeroing in the constructor you might get bad results but for a compilation stuff the removal test should be fine

[TKatt]

Little update:

If you now pull the very latest my suggested change is already incorporated.
See https://github.com/su2code/SU2/pull/1445


So I hope this fixes it for you. It would be great to hear if the compiling and inlet files now work for you

[Zhen ZHANG]

Hi Tobi,
Thank you for your help.

Yes, I am using 4.8.5. I tried several times these days but cannot pull the code from Github. So I just modified the code to

Jacobian_Buffer[4];

as you suggested.

Now both the compiling and inlet profile work well for me. The dimensional value in the example_inlet.dat is very convenient.

Thanks again
Zhen

[TKatt]

Awesome, That is good to hear that it now works for you Thanks for your debugging work !