CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   Gambit meshing (

lapin_tout_fou August 27, 2009 05:09

Gambit meshing
Hi everyone,

I'm dealing with a meshing problem. Well I generated a mesh with Gambit. The problem is that the equiangle skew is higher than 0.95 in some of the cells. I'd like to reduce it for these cells? How can I proceed without regenerating a complete mesh on the volumes concerned? Can I move the points that compose the cells easily or do I have to regenarate the mesh by changing some parameters in the defauts menu?
I don't know if it's because of that, but Fluent never converges. First the skewness is higher than 0.98 in some cells (can I reduce that in Fluent) and I always have divergence after 5 or 6 iterations with the warning "divergence in AMG solver: pressure, temperature, or k or what else". Maybe by adapting the iso-values of cell skewness my case could iterate but I don't know how to use it.
Maybe you have other leads. Thank you for your answer.

-mAx- August 27, 2009 08:39

Salut Thomas,
Alors en anglais ša donne:
Try to reduce equiangle skew below 0.9
I am afraid you cannot "easily" fix those bad cells without remeshing the volume, except if the nodes of those cells belong to a surface.
Hence you can here move the node on a surface.. it will remesh the cells which have this node.
It was the gambit side.
On the fluent side, is the check mesh ok?
Divergence in AMG solver direct after the firsts iterations sounds like poor initialization.
So try another initial guess, or try other solvers.
But once again I would try to go back in gambit and remesh the volume.
Check why you get some bad cells
Bon courage

lapin_tout_fou August 27, 2009 09:44

Salut Max

Alors tu es Suisse, comme un joueur de tennis devant qui je suis admiratif!
I'd like to reduce equiangle skew below 0.9.
The nodes of the cells don't belong to a surface, unfortunately.
I think that a mesh was created on surfaces and some edges and that the mesh was generated on volumes through these surfaces. And it appears in Gambit that some volumes skewnesses are higher than 0.9 but less than 0.96. What I don't understand is that, I think, there's a slight difference between the skewness in Fluent and in Gambit.
The skewness exceeds 0.98 in two cells in thread ID=101 in Fluent and is different of the max skewness in Gambit. Why?
The check mesh is OK. No errors appeared. But the case check told me the skewness is too high in only 2 cells. That's frustrating!!!
I did the manipulation "combining skewness-based smoothing and face swapping" described in 23.11.3 of Fluent help. Should I do some Laplacian Smoothing (23.11.1)?
Actually I'm working on a mesh that's supposed to work as there were previous Fluent calculations. Maybe the initialization is bad as you mentionned.
I am not realy familiar with Gambit and Fluent, but I think I'm getting better every day. The fact is I need to know why I do this or that manipulation.

Well thanks for your help.

-mAx- August 27, 2009 10:03

Have you seen where are the bad cells?
you can click on the examine Mesh Icon (bottom right), and set the lower value with 0.9
Click update, it will show you all the cells which skewness is between 0.9 and 1.
If they are concentrate at the same location, it means some entity enforces the cells to be disturbed (for instance a very small edge).
Else you may play with size functions, etc...
This is how I fix those issues.

lapin_tout_fou August 27, 2009 10:17

I've already examine the mesh with Gambit with the mesh icon. I saw the cell which skewness is between 0.9 and 1. Some of them are isolated but others are concentrated at the same location. If I play with size functions on edges I meshed, do I need to regenerate the complete mesh volume or is it automatically done when you modify the meshed edge?

Thank you again for your time, -mAx-

-mAx- August 27, 2009 10:43

ok, check the geometry around the location where all the cells seem to converge.
Is there a small edge, a gap, a small angle?
you may post a picture with a big zoom on this area.
To your other question, if you apply a size function, you will have to remesh your geometry

All times are GMT -4. The time now is 10:47.