CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] Why Icem can't craeate Unstructured 2d Mesh on Circular Geometry (

Jvb December 29, 2011 17:30

Why Icem can't craeate Unstructured 2d Mesh on Circular Geometry
1 Attachment(s)
I'm a new Icem User and I'm trying to create an unstructured 2d mesh on circular annulus geometry. But I obtain only bad results like this in picture.
I do this action:
-Create annulus Geometry
-Create Body in annulus geometry
-Set Global Mesh Parameter
-Set Curve Mesg Setup
-Compute Mesh -> Surface Mesh Only -> Apply

PSYMN December 30, 2011 20:26

Is there a surface? Which surface meshing method are you using?

Each surface mesher has its own requirements (as shown in the tutorials).

ICEM CFD is very capable of meshing a torus, but it needs the right inputs.

Jvb December 31, 2011 08:23

Unstructured 2d Mesh
1 Attachment(s)
Thank you for your interest!
This is a simple 2D personal exercise model and there is no surface, only two points two circles and and a body between the circles. I don't create any surface because in blocking models is not necessary and I think that the behavior of the mesher is the same. I want to create an "equivalent" unstructured mesh of this block mesh.
I use All Tri Surface Mesh and set the mesh of circles in an arbitrary number of points.
Sorry but i'm a beginner and I'm in the "experimentation period" with Icem.
Thank you so much for your Help....and HAPPY NEW YEAR!!!

Jvb December 31, 2011 10:32

Another unstructured 2d Mesh
1 Attachment(s)
I made this mesh with the same procedure described before.
- Make Point
- Make Line
- Make Body inner Line
- Set Global Mesh Parameter (Scale and Max Element)
- Set Curve Mesh Setup
- Compute Mesh -> Surface Mesh Only -> IT WORKS!

And the mesh has no problem in Fluent!
This is the reason that I don't understand why with an annulus this procedure don't works.
Thanks a lot for your support.

PSYMN December 31, 2011 12:13

The patch dependent mesher uses a concept of "loops". It is actually a "recursive loop mesher" if you wanted to dig into the algorithm further.

All you need to understand is that it is pretty easy to figure out the loop when you have a single chain of curves in a flat plane. But with your torus, you have two separate circles without any way for it to understand that the one is within the other and no where to set the mesh size between the two. There are various ways to sort this out. You could draw a line between the inner and outer circle, but this would force a line of nodes in the mesh and may not be desirable.

The easier way is to create a surface from the curves and then build topology. The topology information tells the mesher that the surface is bounded by both curves and it can easily figure out the loops from there.

The material point is irrelevant to this method. Material points are for volume mesh flood fill after using a top down mesher like octree. The shells of the patch dependent method are put in the part of the surface or curves.

PSYMN December 31, 2011 12:15

One other thought... If you have already blocked this out with hexa and just want to replace that with an unstructured mesh, you can simply convert your hexa blocks to unstructured... This is done under blocking => Edit blocks => Change block type => free. Select the blocks you want to convert. Then go in again and set the free mesh type to what ever you want (all tri, quad dominant, etc.)

Jvb January 2, 2012 12:41

2 Attachment(s)
Thank you for your hints!!!
I trust both the solutions. I think that blocking free results more useful than all unstructured mesh creation because there is the possibility to reserve some area to structured mesh (I.E. for Boundary Layer Calculation). I attach both my meshes.
I should study in deep the concepts of "patch dependent", "Recursive loop meshers", "Top down meshers -> Octree".
Have you suggestion on inherent literature? Do you think that Icem User Guide or Icem Tutorial are usefu to explain this conceptsl?
Thanks a lot.

All times are GMT -4. The time now is 13:52.