Documentation of rough wall boundary conditions

Miscellaneous discussion topics about Code_Saturne (development, ...)
Post Reply
jcharris
Posts: 19
Joined: Tue Apr 16, 2019 7:28 pm

Documentation of rough wall boundary conditions

Post by jcharris »

If I understand correctly, with a k-epsilon approach (iturb=20), the default wall-model is a two-scales log layer (iwallf=3). However, the documentation does not appear to correspond exactly to what is in the code.

In particular, it seems that Eq. I.5.2 is not correct (which should instead correspond to line 775 of <https://github.com/code-saturne/code_sa ... clptrg.f90>)?

(Also, there is apparently a missing reference at the top of pg. 53 in the theory guide for v6.0 <https://www.code-saturne.org/cms/sites/ ... theory.pdf>.)
jcharris
Posts: 19
Joined: Tue Apr 16, 2019 7:28 pm

Re: Documentation of rough wall boundary conditions

Post by jcharris »

To try again on this...  For something simple... I assume that the C_log that appears in Eq. I.5.25 should be removed and was a typo?

For default k-epsilon, for example, looking through the code, the friction velocity is called uk, and is determined from the value of k, using the theoretical log-layer value?  (which depends on itytur, which the online documentation claims is now deprecated?)  It would be nice to have a reference of those outside of Code_Saturne who have used that -- does anyone know what this missing reference in section 5.4.1 is supposed to be? (And is this blending actually used for rough boundaries?)

For k, then, the actual boundary condition seems to be a Dirichlet boundary condition, imposing that k = the value of k at the adjacent cell?  Is this correct?  If so, is this mathematically the same as a Neumann condition with a derivative of zero?

For epsilon, this appears to be a Neumann condition; is there any documentation of this without going through the code line by line?  In particular, it's not clear why the derivative depends on "ydep = distbf*0.5d0+rugd", when distbf is the distance between the cell face and the cell center?  Where did the factor 1/2 come from?
Yvan Fournier
Posts: 3322
Joined: Mon Feb 20, 2012 3:25 pm

Re: Documentation of rough wall boundary conditions

Post by Yvan Fournier »

Hello,

I am not too well versed in the intricacies of the various wall layers, but do know that there has been recent unification work between the smooth and rough wall laws, and the documentation may be lagging.

Do not hesitate also to point to documentation issues in the internal bug-tracker (I usually recommend the GitHub one to those who have access to both, but with the internal GitLab tracker, I can assign issues to some colleagues who often forget to visit this Forum or the external tracker).

I reminded Martin to take a look, but if you have no news shortly after the holidays, I'll re-post on the bug-tracker myself.

Best regards,

Yvan
jcharris
Posts: 19
Joined: Tue Apr 16, 2019 7:28 pm

Re: Documentation of rough wall boundary conditions

Post by jcharris »

Ok, thanks for the response. If it's still a question at some point in January, I'll come back for more info.

That said, given that wall functions are not visible from the GUI, and the GUI has been promoted for some years now, it seems like a reasonable assumption that the options available by default, even if not well documented, are "correct" for some range of parameters, right? If documenting is not possible, would it be possible that someone can confirm that the boundary conditions (i.e., inlet / outlet / wall) have been validated or are implemented as intended, or is it a bit still the "Wild West" when using the code? (Or is there a list somewhere of what options are trusted?)

And to profit from the public forum -- are there others that have used rough wall boundary conditions for applications?
Martin FERRAND
Posts: 40
Joined: Wed Mar 14, 2012 10:06 am

Re: Documentation of rough wall boundary conditions

Post by Martin FERRAND »

Hello,

Indeed eq I.5.25 has a typo (no Clog). I will correct that.

Concerning the two velocity scale, yes there are other references you can read (Craft, and also PHD of Jean-Francois Wald).

The derivation of the BCs of k and epsilon is detailed in appendix (which help reading the code). Indeed the condition on k is a Dirichlet which mimic an homogeneous Neumann for steady flow. I have no more explanation that “it was found to be more stable”. If you want to experiment a true Neumann condition you can do (I did some tests in that direction) but changing this is really hazardous for us because it requires to relaunch all the VnV process (long task).

Finally, all options are supposed to be ok (otherwise an issue should be open on one of our bug tracker). What is in the GUI is the “standard way of doing”, what is not in the GUI is supposed to be “advanced options” or “new options”.
If you want to get access to what is validated, from the EDF network you can get the validation report (not open).

Best regards
Martin
Post Reply