hi 1/153!
thanks for your comments - no prob about your english! its good!! i have thought about this overnight, and actually gone back to the theory and a couple papers to as well, and i think the confusion comes in because of a couple idiosyncracies in the models themselves and so terminology used in OF ... 1) high Re models = normal y+ rules, wall functions required 2) low Re models = normal y+ rule, no wall functions, specify explicit BC's 3) SPECIAL = Menter's k-omega SST = either high Re with wall functions, or low Re BUT THEN you need to use a "wall function" for blending the viscosity from the k-omega implementation to the k-e implementation. Currently this is nutUSpaldingWallFunction (formulation based on U, OF2.1.1), previously nutSpalartAllmarasWallFunction (older releases). You also should use the kqRWallFunction for k and omegaWallFunction for omega, although you can use explicit BC's as for the other low Re models. I think the reason for the 'wallfunctions' needed in SST are because in the revised Menter model (2003), the model constants are blended from the k-omega implemetation to the k-e implementation, and the omega-equation is blended into the epsilon-equation, which may not be done in OF. Perhaps the wallfunctions for nut, k and omega blending are in fact work-arounds for this (i havent checked the code for omega, so i dont know if omegaWallFunction is in fact the omega-equation blending used in the 2003 model. The confusion for me has been as a result of the reference to the blending functions as 'wall functions' - a term which one doesnt associated with low Re models! 4) FINALLY, with the low Re models, in order to allow for the calculation of yPlusRAS, instead of using 'calculated' for nut, you need to use nutLowReWallFunction, which as Roby mentioned above, is actually just a placeholder, and equivalent to ( fixedValue uniform 0 ) but allows OF to calc yPlus values. Unfortunately, i am still getting terrible convergence with k-omega SST, but as far as i can tell, the above is how its supposed to work in OF! Anyway, those in the know - please correct me where i am wrong!!! EDIT: oops sorry, reference to Roby above was incorrect, i was reading another thread on this where the poster was AlmostSurelyRob - who i meant to reference, not Roby above |
This thread turns out to be very interesting and it contains good explanations.
I arrive here because I want to know how OF manages the heat flux at the wall in a turbulent flow. I've found that Fluent includes the Energy into it's wall functions ( http://combust.hit.edu.cn:8080/fluen...e430.htm#stdwf ). From what I know so far and from the source code I've read, it seems that OF doesn't make any adjustment of the thermal boundary layer. So my question is: does the temperature field have meaning in case of a coarse mesh (y+ = 30). I feel that my question is stupid because the thermal boundary layer can be quite different from the momentum boundary layer but I still want to get some point of view. Does any of you have an idea of how to get the heat flux at a wall in the case of a turbulent flow ? For example an airfoil of 1m at Mach 0.3 ? Also, can you confirm that OpenFoam doesn't have any friction heating ? (edit, it seems to be included in the solver itself but I don't know if this treatment include the boundary layer). Ps. I mostly use rhoSimplecFoam. |
I also wonder, does the thermal boundary layer works like the momentum boundary layer, I mean: is it possible to make a very fine mesh and so avoid any wall function ?
For example, is it possible to make a very fine mesh y+ < 1, and then simply compute the heat flux using the gradient of temperature ? |
Terrible Convergence for k-omega SST
Hi Jonathan,
I have also gotten terrible residuals for k-w-SST. Then I went to the motorbike tutorial in OF-2.2.0 which should be the Meca of a fine tuned model, and the residuals look amazingly crappy (granted it only iterates to 500 steps - but doesn't look as if it gets any better after). I found after a bit of poking around the model, that the quality of the residuals is highly dependent on your mesh. For example, I got much better results by sending the layer insertion to hell and just leaving the elements from the snap alone. Here is a thread I started, asking about the convergence for the motorbike tutorial (OF-2.2.0): http://www.cfd-online.com/Forums/ope...mega-mine.html Quote:
|
I answer part of my question: I've just found the Foam::compressible::alphatJayatillekeWallFunctionF vPatchScalarField
Quote:
|
enhanced wall treatment
Hi,
I read the thread because I'm facing some problems regarding the wall treatment in openfoam. I found a presentation which could solve some problems: http://www.opensourcecfd.com/confere...9_Kurenkov.pdf Here, for k-epsilon and k-omegaSST automatic wall treatment was implemented. Does anyone have some experience with that? Maybe can help to implement or did something similar already? Best regards, Andy |
hi all
Is there any low-Re capable turbulent model in OF for "compressible,internal flow in a tube"? Could introduce them to me? Thanks so much. |
Quote:
Hi Jonathan, I'm interested in what you had to say about the Low Reynolds wall function (LowReWallFunction). Do you implement it it on a low Reynolds mesh? with a very small y+? I have been using the LowReWallFunction in 'nut' with the corresponding transport equations like kappa and omega (k-Omega SST) and thought 'if I'm applying a wall function k and W also require wall functions'. Is this correct because I can't see how it would really work. Any help would be greatly appreciated. Regards David |
hi Crossley
my understanding is that LowReWallFunction was instituted by the OF developers to be used as the BC for nut with low Reynolds turbulence models on y+ ~1 meshes. If you look at the header of the code, it says: Code:
Class Quote:
in terms of the 'correct' setup for kOmegaSST, there are a few other threads where this is discussed, and perhaps they might help as well. hope this helps, cheers jonathan |
Hello guys,
I read trough all the discussion, it was very useful to understand the general topic of wall function. However I got a bit lost and at the end I have probably a very stupid question. Which is the wall function that I should use for high Reynolds? I am using compressible k-e with Y+ above 30. Can you provide me the name of the wall function? I was a Fluent user so I am used to the Fluent nomenclature, in this case I was used to employ the standard wall function. Thanks |
Hi
those are different for different variables.for k use:"kqRWallFunction" for epsilon:"epsilonWallFunction" look into a turbulent case in tutorial folder. |
Shall I define also the wall function for the mut?
|
yes,you can use "xxx" for example in return of type of mut or anything else and compiler says you what types are there for using.like mutkWallFunction,mutUWallFunction,mutUSpaldingWall Function,...
|
I was searching everywhere for the way to get kw-sst working as low-re formulation and your post got me on the right way with the NutUSpaldingWallFunction. Thanks alot.
|
if its so, add a reputation power for me :)
and you remembered me about it to write in my thesis mention to it. do you know a good reference that has described mutUSpaldingWallFunction? |
damping functions within the sst model
1 Attachment(s)
this is quite an interesting thread, with some good explanations
i have a question regarding the damping function the sst model uses. so it says earlier in this post that the sst model can be used as a high Re model or a low RE model. I'm simulating the propagation of a pressure wave within a pipe.I have created a simulation using the RNG k-epsilon model and got reasonable results, using a mesh with a y+ in the range 30-300. then i applied the SST k-omega model. i got a slightly better result that the RNG model. I did not refine the mesh, so i applied the SST model with wall functions. my question - in the SST model are the damping functions also applied when the model is being used as a high Re model? or how can i explain the improved results? I've attached a graph of my results some advice would be appreciated |
Rough wall example
Quote:
Thank you very much for your help. |
Hey guys,
first of all I'm a nooby in OF so keep it in mind please :D my question: how can I have different wall treatment in one case. Lets say wall 1 should have a high Reynolds number treatment and wall 2 should have a low one. - What exactly do I have to choose for a turbulence model in RASProperties file? RASModel: kEpsilon (for high Re) or LaunderSharmaKE (for low Re)? - What should I put instead of epsilonWallFunction in the following code segment in case of a low Reynolds treatment? wall_high_Re { type epsilonWallFunction; value uniform 0.02911; } wall_low_Re { ? } thx |
Hi Mo,
As I understand it, your model must be capable of doing this. I am not sure, if the low-Re k-epsilon models in openfoam (such as the LaunderSharma you mentioned) work with high-Re boundary conditions at all. I guess the k-omega-SST could do this. If the model supports it, you just need to use: wall_low_Re { type fixedValue; //(<- most of the low_re conditions are fixedValue, depends on the model value XXX; //whatever fits to your model, LaunderSharma needs a boundary condition for the modified epsilon-tilda to be zero at walls } Have a look here for all the boundary conditions: http://www.tfd.chalmers.se/~lada/pos...ndium_turb.pdf |
Quote:
from what I know so far, in order to have an enhanced-wall-function-like feature in openFoam, nutUSpaldingWallFunction (unified wall function based on Spalding's law) can be used together with kqRWallFunction and EpsilonWallFunction. And I saw from other threads stating that kLowReWallFunction and epsilonLowReWallFunction can be used in kEpsilon model, which I doubt it, as those 2 wall functions are actually created for v2f-turbulence model. I'm still new to openFoam and correct me if I'm wrong. |
All times are GMT -4. The time now is 12:11. |