CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [GAMBIT] Meshing problems (

iHal February 18, 2010 20:58

Meshing problems
Hello everybody. I'm having some trouble making an unstructured mesh. I have 13 3D spheres contained in a cube. I subtracted my sphere volume from the cube but kept the sphere, as well of course keeping the cube. I want to mesh the spheres and the cube with an unstructured tetrehedral mesh. The spheres appear to be meshing fine according to the transcript.
But the cube meshes its 6 faces and then slows down and fails to mesh all the elements within itself. :confused: proably something to do with the spheres..

Gambit prompts the reason for this as "Tetrahedral meshing has failed for volume volume _; This is usually caused by problems in the face meshes. Check the skewness of your face meshes. Check the skewness of your faces meshs and make sure their sizes aren't too large in small gaps."

I didn't use face meshes as i had the meshs Tet/hybrid and TGrid from volume meshing of the geometry (the geometry which i made up in gambit). I have changed the growth rate, max size and spacing but continues to fail. I checked the amount of skewness from the examine mesh tab and i'm not sure as im a beginner of the software how to fix skewness in that tab or by some other means if thats indeed the problem.
Can somebody please help me?

-mAx- February 19, 2010 02:26

If you mesh your volume, Gambit will mesh automatically the surfaces first.
So examine you surface mesh Click on the bottom right's icon, choose 2D and range
Enable worst element, it will show you where is your problem.
Then correct your problem.
Question: Why are you keeping the spheres if you have substrated them from the cube?
If you still have problem, post a picture

iHal February 19, 2010 18:36

Lack of dynamic memory
Thanks mAx for replying to my thread.:) I was wrong in saying that it was a face mesh issue as i kept interrupting the meshing when it started to freeze. In answer to your question the cube is an enclosure which will have its top and bottom faces at different temperatures (im using this in fluent as a heat transfer problem). The spheres will denote a conducting material so i need to have both the cube and spheres seperate. The energy equation will be employed later on in fluent.

I got somebody to look at it and the reason why it isn't working is because I don't have enough "dynamic memory" with my computer for Gambit to mesh the geometry. I just deleted processes occupying too much disk space and I have gone into to task manager to temporaily stop processes running in the background using memory. Also i made sure im not running my computer on power saver. I can run the software now with just a single sphere in the cube with a high max size but what i want to do is model and mesh a number of them. Any solutions/advice to my computer's memory limitations in this instance would be appreciated.

If it helps I have set the growth rate and the max size to 200 already for a cube of 100 height, depth and width. The sphere(s) are 10 in diameter. Gambit is prompting me to have an additional 20MB of working memory for a cube containing 13 spheres.

-mAx- February 22, 2010 02:17

Ok, if you compute also solid conduction, you are right.
Reagarding your Gambit problem, I assume you are attempting to mesh your volume on the fly, without any volume decomposition (split).
Try to split your volume into smaller and basic volumes. it will ask less memory while meshing smaller volumes, and you will have more control on your mesh.

iHal February 23, 2010 19:08

Volume decomposition
3 Attachment(s)
Thanks for replying to my thread again mAX.

I not sure how to go about volume decompostion, I looked over some gambit tutorials but it seems tough for what I'm doing as I have spheres (see attached pictures) scattered throughout my biggest volume. I found something similiar whereby a single cylinder is centered in a cube along the y-axis of a single sphere. The cylinder is half the radius of the sphere. The next thing the author says he vaguely did was "split the sphere with cylinder volume and perform an additional split using the diagonal face."I not too sure how to do this.After that the cooper scheme will be applied where the author said to mesh the volumes by manually select the source faces I've split the volume in this way below, I'm not sure if it is right. Could you tell me how to do these remaining things or refer me to some literature which could help.

Thank you for your help so far it means alot, this is part of a project I'm doing.

-mAx- February 24, 2010 02:09

I don't see the interest of spliting your geometry with 2 cylinders.
Anyway, if your model enables it, you can mesh only 1/8 of your geometry (it seems it is symmetrical, isn't it?)
To proceed, (regarding the green color of your geometry, I assume it is a real one) create (xy)-plane, (xz)-plane, and ((yz)-plane. As your geometry is centered, you can split your geometry with those 3 planes (you will use 3 times split operation, one per plane).
If it's successfull you can try to mesh your volume on the fly. If you still have problem, you need to split the volume again (for instance with a sphere surrounding the others).
Regarding litterature, I would recommend you the tutorial #3 of gambit: MODELING A THREE-PIPE INTERSECTION (3-D)

dynamicdom February 25, 2010 09:21

hello All,

I'd like like to do an meshing analyzation of a steam turbine gasket. That means, I would like to do do several meshes in the ANSYS 12 mesher.

The meshes should be from very fine to very coarse. The cells are triangles. For each mesh I will start a calculation in FLUENT.

The aim is to see how the results of the calculations diverge if I change the mesh.

How can I create several meshes out of one?
Or how can I change one mesh and save the different calculations in FLUENT?

It'd be nice if anybody could help me.


All times are GMT -4. The time now is 12:07.