[WIP] Inconsistencies and improvements to SST model #2329
Conversation
- Added production terms to SST
- Added user-defined production limiter constant for TKE
rois1995
changed the title
- Updated boundary conditions as in TMR page
Common/include/CConfig.hpp
Outdated
| nPrandtl_Lam, /*!< \brief Number of species | ||
| addDoubleOption("FREESTREAM_TURB2LAMVISCRATIO", TurbIntensityAndViscRatioFreeStream[1], 10.0); Prandtl number. */ |
There was a problem hiding this comment.
Was this intentionally commented, or...? If it is used, I guess this should go to line 872
| nPrandtl_Lam, /*!< \brief Number of species | |
| addDoubleOption("FREESTREAM_TURB2LAMVISCRATIO", TurbIntensityAndViscRatioFreeStream[1], 10.0); Prandtl number. */ | |
| nPrandtl_Lam, /*!< \brief Number of species laminar Prandtl number. */ | |
| addDoubleOption("FREESTREAM_TURB2LAMVISCRATIO", TurbIntensityAndViscRatioFreeStream[1], 10.0); /*!<\brief Freestream mu_turb to mu_lam viscosity ratio */ |
There was a problem hiding this comment.
This is just a wrong copy-paste from me. It should not be there in the first place.
|
Great contribution, Thanks @rois1995 ! |
|
If you are looking into robustness aspects too you should get in touch with @emaberman and @YairMO, seems like they have some good ideas and between the free time of 3 people a lot more can get done :) |
|
Hi, Regarding the cross-diffusion term (CD) that appears in Omega source term (residual). The SST model (1994/2003) is a high-Reynolds-number model. Namely, It can not predict correctly the sub-layer region (especially the correct profile of the TKE). Therefore, only a positive contribution is required. Moreover, since the SST model was design as a k-w and k-epsilon blending, the CD term "belongs" only to the k-epsilon "branch", that is why the CD term include the factor "1-F1". However, it may happen, that the factor "1-F1" is not a 100% safe guarantee. It may happen that "1-F1" is not zero in region where the CD term is negative (this happen due numerical errors). To avoid such a situation, it is a good idea to clip the CD term with zero. Otherwise, severe numerical robustness issues may rise. |
- Given option for cross diffusion limiting in W residual
|
Hi, The use of an Omega production limiter (about the cross-diffusion term) is correct for low-Reynolds-number (LRN) models (the approach described by Peng et al. is very naive; there are other more rigorous treatments). For high-Reynolds-number (HRN) models, the clipping should be zero, keeping the cross-diffusion term positive; thus, the current implementation is correct. Indeed, it is not exactly as it appears in Menter's original publication. The factor (1-F1) aimed to promise that the cross-diffusion term will be activated only outside the boundary layer, where it is positive (the cross-diffusion term switches its sign deep inside the boundary layer). This was also recognized by Peng et al. (first paragraph above Eq. 17). However, it may happen that the factor (1-F1)=1 where the cross-diffusion term is negative. Usually, it may happen at the wake, very near the airfoil trailing edge, where the upper and lower boundary layers merge. It is due to the imperfection of the F1 function. To summarize, the current implementation is correct, and it is perfect for HRN models. |
|
For the sake of clarity, "current implementation" refers to the current treatment of the production code |
|
What YairMO is saying, is that allowing negative cross diffusion values is incorrect for high Reynolds models and should not be an option, this is a fix used for low Reynolds models only |
|
Hi @YairMO, Hi @emaberman , thank you very much for your comments. I haven't found any suggestion in literature to clip to only positive values the cross-diffusion term in the w-equation. I understand that it might be more robust, but it is not the standard implementation of the SST model, which is the first thing that we need to achieve. Only then we can build on top of that to improve the robustness of SU2. Nevertheless, I tried the SWBLI test case and I compared the results across 6 different combinations: 1- develop branch, no changes When my branch is used, then the changes to the supersonic inlet BC are already in place. I haven't achieved convergence with 1, 2 and 3. More precisely, 1 diverged right away (after 30 iterations), while 2 and 3 gave "FGMRES - Orthogonalization Failed" after 900ish iterations. Here you can see the residuals for the different combinations.
Unfortunately I will be busy with the AIAA Conference next week, thus I don't know how much I will be able to work on this. The next test case will be the 2D airfoil near-wake from TMR. |
|
Hi rois1995, First of all, enjoy your time in Las Vegas. Any paper that you are presenting? As for our discussion about the cross-diffusion term, I've emailed the "source" (Menter). I believe he will make it clear. |
|
The wall distance is correctly computed. The nu_tilde, however, is higher on the symmetry plane, just like the Mach number. And, again, it happens only at the lower symmetry boundary.
|
|
Based on your input (unexpected generation of the eddy viscosity at the lower symmetry BC), I would suspect the wall distance calculation. Can you tell how the wall distance is calculated in the cells above the symmetry section? Consider a cell above the symmetry section. Is its wall distance taken relative to the symmetry part or relative to the wall part? |
|
It is taken relative to the wall part (which you can see as a straight white line. I don't know if the increase in nu_tilde causes an increase of the vertical velocity on the symmetry or it's the opposite.
|
|
Yes, the WD is as it should be; thank you. I also agree: who comes first, the egg or the chicken? I'll try to run this case using my code. |
|
I think the problem is in the fix for the symmetry boundary in PR #2194. I tried with the develop at the last commit (LC), at the commit before the implementation of the correction (BC) and at the commit of the implementation (AC). It is clearly visible that something is happening with the symmetry BC correction. I will look further into it
Also the profiles of the Mach number and nu_tilde are the same between LC and AC. With the BC instead the Mach number and nu_tilde are correct. |
|
I tried without the MUSCL reconstruction and with the AC commit and both Mach and nu_tilde profiles are correct at the symmetry plane. @bigfooted can you comment on this please? |
|
I think I found out the problem. In the correction of the gradients on the symmetry plane, somehow the gradient of the velocity along the normal direction was still not 0. By fixing it then the Mach number and nu_tilde profile are correct. I don't know if I am still missing the correction in the tangential direction, but at least the solution improves. Here you can see the same graph as before but without the line of the last commit. I have added, instead, a try with the Weighted Least Square (LS) and a first order (FO) (no MUSCL). In the end I have added the line for the correction implemented (Corr Grad). We can see that the behaviour before the implementation of the gradient correction is recovered.
|
| // } | ||
| // } | ||
|
|
||
| // Do I really need all of the previous code? Can't I just merge the two for loops? Maybe I am missing the tangential direction |
There was a problem hiding this comment.
Yes what you are doing neglects that velocity is a vector quantity
There was a problem hiding this comment.
The previous implementation was actually correct, I'll revert to that one. Nevertheless something is off in the gradient computations since without MUSCL reconstruction I don't have these problems.
There was a problem hiding this comment.
Can you try the coarser mesh and converging it very well to check if this problem still happens? In your previous plots the residuals are not dropping a lot.
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
rois1995
force-pushed
the
feature_SSTMod
branch
from
a784a4f
to
317f563
Compare
|
I tried running the coarsest mesh using the develop branch on the commit before the symmetry correction (BC, bb0f9de) and at the symmetry correction (AC, 77fed34). For lines blue and yellow I tried without NK for the first 1000 iterations and then with NK. Convergence is good but somewhat different. No strange behaviour on the Mach number and nu_tilde are present. Lines purple and red are with NK from the start. No convergence at all on the AC commit and bad solutions at the symmetry planes. Everything normal on the BC commit.
I will try changing the cell spacing in the normal and tangential direction separately and check what brakes it. |
|
I've tried reducing the residuals up to -11.5 for the logarithm of the density residual. I've started on the coarsest grid as you saw in the previous plot. I've proceeded on the L2 mesh keeping either the tangential or the first cell spacing constant, using as restart the converged solution on the L1 mesh. The results are essentially the same as with L1 mesh. When using the NK solver straight away the AC branch does not converge, while the BC one easily converged. If I remove the NK solver, both converge. The same has happened on the L3 grid (restart from converged L2 results). Here you can see the residuals for the L2 grid (either the case, he outcome is the same)
No strange behaviors of Mach number and nu_tilde have been seen on converged solutions. On not-converged ones, instead, the problem is right at the symmetry planes. |
|
Thanks for testing, part of the NK options is a period where instead of NK we use the normal quasi-Newton approach. |
|
But what do you think of the changes between the two commits? It doesn't seem that the convergence is improved when using the new symmetry BC implementation. |
|
Could you please share your L3 grid (together with your input file)? |
|
The changes to the symmetry BC were to improve the accuracy, especially when there are walls perpendicular to the symmetry plane. |
|
@YairMO in the zip file you can find the .geo file to create the mesh via gmsh, the L3 mesh and the configs file used for simulations with SA. |
|
If you see the stagnation in residuals on the L1 meshes while not using the NK solver (line blue and yellow), it is due to a single point in the grid that you can see here. I checked the mesh quality in that point and it doesn't seem to be bad, or at least not worse than the other points.
|
|
Hi, I'm using SU2 8.1.0 but with some differencing in SA implicit treatment (underwork).
|
|
Impressive! Can you share some info on your implementation? What about the CFL values and the computational time? One more thing: in another PR (if I recall correctly it was the one for the dimensionless limits to SST) you mentioned the VFE2 test-case. Do you have access to shareable CAD files and meshes? |
|
Thank you. As for the VFE-2 mesh, I've obtained this mesh from someone else. Please let me check with him first (I believe there will be no problem. Sorry). |
|
Sure, thank you for checking! Regarding the SWBLI case, if I understood correctly, you have used your code with NK solver from the start and you changed the number of linear solver iterations, right? Can you also try with SST or is your implementation different only for SA? |
|
Yes, your understanding is correct. |
|
Here are the results obtained using the SST2003 model
|
|
It seems that yoru implementation affects also the SST model, is it correct? Nevertheless, huge improvement wrt both this branch and develop branch results (density residual stagnates around -10)! |
|
Yes, it also affects the SST model. Thanks |
|
Hi Rois1995, I have permission to share the VFE-2 grid; please send me your private email. |
|
That's perfect, thank you! [email protected] |
|
I have a few updates. I found out that most of my FGMRES problems in this branch where related to the fact that I was using single precision for the linear system. You can see it in this graphs for the residuals of Rho.
The first two curves are in mixed-precision and they stop long before reaching the minimum residual required due to divergence of the linear solver. Moreover, it seems that the 2003 model here implemented (which, simply put, considers the full reynolds stress tensor for the computation of the production of k and always considers k in the stress tensor and in the thermodynamic variables) has faster convergence and increases the recirculating zone (maybe due to the reduction of turbulence kinetic energy) which is a feature that has also been seen in [DOI:10.1017/aer.2020.93].
There are some differences between the use of the TMR boundary conditions and the ones used before. The results improve with respect to the develop in the prediction of the SFC distribution, which has results completely off from the V&V page of SU2
I may keep on checking if the implementation of the v2003 model is correct or not by searching for other test cases (probably coming from DOI:10.1017/aer.2020.93). |
Proposed Changes
Hi everyone,
I have found some inconsistencies with respect to the literature on the implementation of the Menter's SST model. I would like to use this branch as test bench for any corrections/improvements made to the SST model.
Implementation errors found:
positive portion of the cross-diffusion term in Eq (A2)" pag. 1604. Moreover, a clipping has been introduced for large negative values of this term, as suggested in Peng, Shia-Hui, Peter Eliasson, and Lars Davidson. "Examination of the shear stress transport assumption with a low-Reynolds number k-omega model for aerodynamic flows." Eq 17.
Changes to SST model proposed:
I've seen the proposed changes to the lower limits of k and w in #2323 and I tried implementing it in my branch. However, if the implementation proposed in the respective PR is preferred then I will change mine.
Related Work
#2323 #1851
PR Checklist
Put an X by all that apply. You can fill this out after submitting the PR. If you have any questions, don't hesitate to ask! We want to help. These are a guide for you to know what the reviewers will be looking for in your contribution.
pre-commit run --allto format old commits.