CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Mesh Generation (

Dhananjay May 3, 2002 16:38

Mesh Generation
Hello all,

I am trying to generate a mesh for a fairly complicated 3-D geometry in GAMBIT. However I seem to have a problem with the solution (reversed flow) and need to revise the mesh.

I have a couple of questions about meshing in GAMBIT -

a)Is there any rule of thumb for selecting between unstructrured and structured meshes?

b)Has anyone used the size function option in GAMBIT and found it to be effective?

c)How much skewness is too much skewness?

Thanks and looking forward to your replies.


heıdarı May 4, 2002 01:57

Re: Mesh Generation
the response to your thırd questıon ıs ın the gambıt's help,so just read ıt.

Jonas Larsson May 4, 2002 05:11

Re: Mesh Generation
Unstructured/structured: Use structured meshes whenever possible. In general you will get better accuracy and better convergence with a structured hex-mesh, especially if you can align it with the flow direction fairly well.

Size-function: Haven't tried it yet.

Skewness: Haven't used gambit for some time now but if I remember correctly the "skewness measure" in Gambit should be better than about 0.7 for the grid to work well in FLUENT.

Christian May 6, 2002 07:53

Re: Mesh Generation
Unstructured/structured mesh: I have heard it before and I have experienced it, but I have never seen a mathematical description of this. Do you have a reference ?

Cheers Christian

Jonas Larsson May 6, 2002 08:37

Re: Mesh Generation
I can't give you a reference from memory... this has been discussed several times on these forums though. I remember that Hrvoje wrote several nice posts about this. Here is one example from the archives:

Steven Smith May 9, 2002 10:26

Re: Mesh Generation
Regarding your skewness question, I normally continue working with tet meshes until I get the maximum volume skewness under 0.9 I'm not positive it's necessary, but I do check both equiangle and equisize skewness. Maximum skewnesses less than 0.9 are good, and allow more aggressive use of the relaxation parameters. If your maximum skewnesses is above 0.9, you might have to spend a lot of time babysitting the solver to prevent divergence.

Volker Pawlik May 14, 2002 03:38

Re: Mesh Generation
Concerning skewness: I observed divergence for grids where one (!) or more cells > 0.97 are found.

So I think even the advice is to avoid cells > 0.9 for tetrahedrals I think the ultimate limit is 0.97 for the fluent solver. Especially if the cell is situated in zones of low (!) massflow, bad skewness helps to fortify gradients and causes numerical diffusion.

Dhananjay May 14, 2002 08:22

Re: Mesh Generation
Thank you all for your replies!

edvin May 14, 2002 10:55

Re: Mesh Generation
try to use structed mesh if possible. you can get the same solution with lower number of cells with structed mesh and it is easy to control the mesh.

the unstructed mesh is useful when you have a complex geometry. you can apply it to almost all geometries.

and about skewness: there are several skewness types. look at gambit tutorial.

All times are GMT -4. The time now is 02:09.