[robertjenkins]

Hi

I'm trying perform an analysis on the NACA63-451 foil for my dissertation, this includes capturing the boundary layer separation. Using the un-altered SST model it delays the separation point at high AoAs. Using Ansys student, if the a1=0.28 (instead of 0.31) this correctly predicts the separation point (and hence Cl/Cd). For various licence limitations I need to repeat this on SU2.

I've tried changing the constant in the source code in this file: CTurbSSTSolver.cpp
Via file path: su2code-SU2-f999853\SU2_CFD\src\solvers

This made zero difference to the transient and converged part of the solution, my supervisor confirmed to me it should produce a similar effect to the Ansys change.

Just wondering if the constants are defined elsewhere that I am missing? I've tried looking through the source code as much as I can and attempted to google for solutions, but a bit stumped now. Sorry if I'm just missing it completely!

Many thanks
Robert

[bigfootedrockmidget]

That is indeed the location where the constants of the SST model are defined. However, the SST models in Ansys Fluent and SU2 are different. SU2 has the SST-1994 model implemented and Fluent the SST-2003 model:


https://turbmodels.larc.nasa.gov/sst.html


One difference is in the treatment of the scaling of the turbulent viscosity with strain (Fluent) instead of vorticity (SU2), and this is the term that involves a1 (the Bradshaw correlation). In a separating boundary layer, strain will be larger than vorticity, and in this case, SU2 probably predicts a vorticity that is below the threshold to activate the production limiter (and a1 is effectively not used in the computations).

[robertjenkins]

Thank you very much, that makes sense. How hard would it be to implement the 2003 model into the SU2 SST source code myself? Following the link you provided it suggests that the differences are just alterations to the equations/constants?


Thanks in advance.

[bigfootedrockmidget]

We found out about this in a project involving species transport. The exact model from Fluent was implemented (which is also slightly different than the SST-2003 model).
The implementation was saved in a draft pull request here:
https://github.com/su2code/SU2/pull/...bcf54949daa914


There is some interest in implementation of several versions,
https://github.com/su2code/SU2/issues/1364


If you are going to implement and validate the SST-2003 model, please consider joining the developers meetings Wednesdays at 16:00 CET and contributing your code to SU2.

meet.jit.si/SU2_DevMeeting

[robertjenkins]

Thank you much appreciated, I'll take a look at this and try implement it. I'll compare it to my Ansys solutions as well.



As much as I would love to help. I really have zero experience with this kind of stuff so wouldn't be much help at all, sorry!

[robertjenkins]

Hi



I thought I'd add to this here, I implemented the draft pull request you suggested into the SU2 7.2.1 source code and compiled it without issues. Although it didn't have any noticable affect on the solution for some reason. Any idea of why this may be?

[bigfootedrockmidget]

Hi,

you could write some output to the screen during computation to check if the constant a1 is used in the production or not. Also, did the turbulence quantities converge completely? They tend to converge a little slower than the flow quantities.

[robertjenkins]

Hi, sorry for the delayed reply. In hindsight I didn't really check the turbulent quantities, Cl and Cd has converged to within about +/- 0.5% and I figured they weren't about to massively change from 1.5ish to 1.3ish like required. I may have been wrong.

Unfortunately since this is for my 4th year main project, I don't really have time to investigate further. My supervisor suggested/recommended to just continue with the SST default model.

I really appreciate the help though, thank you.

[keerthi_durga]

Originally Posted by bigfootedrockmidget

We found out about this in a project involving species transport. The exact model from Fluent was implemented (which is also slightly different than the SST-2003 model).
The implementation was saved in a draft pull request here:
https://github.com/su2code/SU2/pull/...bcf54949daa914


There is some interest in implementation of several versions,
https://github.com/su2code/SU2/issues/1364


If you are going to implement and validate the SST-2003 model, please consider joining the developers meetings Wednesdays at 16:00 CET and contributing your code to SU2.

meet.jit.si/SU2_DevMeeting

Hello Sir,
This is regarding few things which I couldn't get in SU2 7.5 version. Thing is I'm currently working on CFD of flat plate having adverse pressure gradient leading to separation. I tried using SST 2003m and also other variants but the results were not satisfactory. I tried using Ansys Fluent and for k epsilon model with Non equilibrium wall function, I got Proper results. Thus I wanted to know, if SU2 has any implementation of k epsilon model as it is know for its robustness and also Wall functions. There is wall function marker present in SU2 but I am unable to use the non equilibrium wall function. And It shows Non equilibrium wall model can only be implemented for LES. But is LES already implemented in SU2?
It would be great if you could help me out in this.
Thank You!

[bigfootedrockmidget]

Hi,

The SST model was created specifically with the idea to improve performance for flows with adverse pressure gradients, so performance for such cases should generally be good. I do not think there are cases where k-epsilon will be consistently more accurate than SST, especially not when there are adverse pressure gradients.


The model seems to work fine for the Stanford diffuser (Buice & Eaton. It might be that your mesh is just too coarse.

[keerthi_durga]

Thank you sir, will try doing the mesh finer near wall region.
And sir is LES for FVM already implemented in SU2?