code_saturne User's Forum

Dear support and users,
I have another question regarding the mesh quality: do you have some recommendations about the weighting coefficient and the offset ? (for example, in which range should my coefficients be?)
 
Best Regards,
 
Stéphane

Hi Stéphane,
Short answer with order of importance (IMHO) and impact: - face warping angle = 0          : integration involving a face normal (divergence, gradient, ...) - non-orthogonality angle = 0  : variable extrapolation within a cell - offset coefficient = 0              : gradient calculation, and more generally variable extrapolation within a face - weighting coefficient = 0.5    : variable interpolation at a face (between two cells) Long answer: it depends !
It usually depends on where the cells of worst quality are in the mesh. You ought to avoid to have bad quality in a region of interest. You may have, for example, a non-orthogonality angle up to 80 degrees with good results. Some models are more prone to blow up than others like LES simulations, particles tracking (but you may use the "cut warped-faces" feature if needed), ...
There are also other criteria such as cell expansion ratio (length / height) that can be of interest. As for the range of the one you wrote about, it's really difficult to give some as it depends on where the region of bad quality is...
David
ps : I haven't forgotten the example I owe you for fans modeling... I just need some time to find a proper one at the office ;)

Hi David,
Thank you for those explanations. I totally understand it depends on the concerned region, and on the geometry as well. My question was regarding the regions of interest, but, again, I understand it can be difficult to give general guidelines.
 
/Stéphane

I'm afraid I won't dare to give you explicit values... too risky and difficult :)
Perhaps someone else can share his/her experience in this thread.

Is the aspect ratio of cells something important regarding the mesh quality?

It can be. The gradient calculation can lead to somehow bad results if the aspect ratio is too big.
Let's say that it's (again) strongly depend on the flow pattern. If you have the larger dimension of the cell aligned with the flow pattern (like a thermal stratification), then it's not a problem... Otherwise, you never know when the aspect ratio is too high; we've already run successfully simulations with aspect ratio of 1/1000 in the near wall regions (LES simulations).

Another thing to avoid is "shaking" meshes (don't know if this is the right translation...). These are meshes with a succession of short - large - short - large - short - ... cells. It usually leads to bad interpolation for face values (because we are cell-centred).

Hello,
Thanks for the tips again!
I have one question regarding the head loss coefficient:
Is it possible to add the viscous loss term when defining the head loss coefficient?
Regards,
Stéphane