CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] help with missing elements of curved surface (

siw September 3, 2009 04:23

[ICEM] help with missing elements of curved surface
2 Attachment(s)

Can anyone tell me how to get my Octree mesh to correctly mesh a curved surface?

These images show the mesh and how the surface mesh is okay on the lower 2 curved surfaces but misses elements on the upper curved surface - even though they are of the same curvature. How can this be fixed? I always run the Build Topology before starting. I've tried switching on the curvature/proximity but this does not fix it. I've also tried Thin Cuts but again it's still missing elements.

I don't know what else to try.


PSYMN September 4, 2009 08:54

Edge criterion.
Octree Tetra’s fundamental approach of starting with the volume and fitting to the surface is both its strength and its Achilles’ heel.

In this case, the mesh size and feature size are close enough that the process of deciding if a node should be moved to the surface or if the edge should be split and the new node moved is causing the pain. The difference between the two fins is due to the way they passed thru the background grid at their respective locations…

You can sort this out with Mesh tab => Global Mesh Setup => Volume Meshing parameters. Set the Mesh Method to Robust Octree and go down to "edge criterion".

You can look up what this does in the help. 0.2 is the best default for most users, but my experience is that setting this to a smaller number, like 0.02, will sort this problem out for you.

siw September 7, 2009 05:47

2 Attachment(s)
Thanks for the information.

I tried an edge criterion of 0.02 and the results can be seen in the first image. So I then tried a value of 0.01 and the results can be seen in the second image. So the problem still occurs. When I tried the 0.01 value I think ICEM did not like something because in the Messge Window red text appeared saying "14 more messages - not printed. All messages saved to file ./ERROR_LOG2.tmp". So I guess reducing the edge criterion further will not help.

Am I just going to have to except this as an inabiliy of ICEM to mesh a curved surface and let CFX solve the flowfield?


PSYMN September 7, 2009 15:23


Well, I guess you should change it back to 0.2 (or 0.1)

Next I would check the surfaces, make sure they are complete and match up with the curves...

Also, check the sizes set on the edge surface to make sure it is the same all the way around...

If that still doesn't do it, I would happily take a look at the model for you and see if I can figure it out.

siw September 8, 2009 02:57

Again, I appreciate your comments and grateful that you'll take a look at the mesh for me.

Some other things I've tried but have not fixed the problem are:

1) I put ORFN points in each fin.
2) I put a Density line across the curved fin tip.
3) I created a new Part using the curves of the fin tip and then used the Thin Cuts with these and the fin sides and fin edges.
4) Reduced the element size of the fin edges.
5) I have set Curve Meshes to the curved fin tips and specified both a Maximum Size and a Number of Node (on separate attempts though).

Just some extra information that may help.

1) The CAD file was created in Solid Edge version ST and saved as a *.par file and imported using the Workbench Reader option.
2) I only have the Tetra/Prism licence.
3) All the element sizes are 1, 1/2, 1/4, 1/8, 1/16 etc. You have given me a detailed description in another topic about how ICEM uses the factor of 2 values rather than values with units (m, mm, ft, in etc). But to a non-expert like myself who has mostly used CFX-Mesh this takes a lot of getting used to and seems less flexible.
4) Prism values have been set but I'm only at the early stage of computing an Octree mesh without the Create Prism Layer option checked.
5) After the Octree has been made and smoothed to a Quality > 0.3 I'll use the Octree surface mesh for a T-Grid volume mesh (may need some extra smoothing) and then finally I'll make the prism layer and split that.
6) The size of the air domain and the position of the wing etc within it have not been optimized.

Do you have an e-mail address I can send the ICEM files to as I'd rather not attach them here so anyone can download them?


All times are GMT -4. The time now is 13:17.