CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ANSYS Meshing] bad mesh (

hadrien51 October 2, 2012 08:50

bad mesh
4 Attachment(s)
Hi Everybody
I have a problem with a simple stupid mesh.
It s like a cylinder defined by curves and surfaces.
the surfaces are not touching the edges, I don't know if could be the error
When I produce the mesh, Ansys don't respect the curves and the edges are horrible.

I made lot of thing before posting this thread.
-I clean the geometry
-I set up the max size element to a value and put 0 for the all curves size element
-I set up the max size element to a value and put the same value for all curves size element
-I set up the max size element to a value and put a fine value for all curves size element
I am doing something wrong because ansys have to be capable to mesh this !
any advice ?
thank you

siw October 2, 2012 11:02

You don't say but that looks like ICEM with an octree mesh.

Did you run Build Topology to sort out the connectivity of the geometry?

hadrien51 October 2, 2012 11:10

yes sorry i forgot to tell that is a octree mesh.
yes also I did Build Topology

siw October 2, 2012 13:25

This is a bit of a longshot, but maybe decrease the Triangulation Tolerance by an order of magnitude (but not too many orders).

hadrien51 October 2, 2012 17:41

hi siw and thank you
what do you mean with "Triangulation Tolerance"
this is an option in octree mesh ?
I saw the post above mine "Octree not respecting curve parameters"
posted by you and now I am trying to use patch dependant but it is not so trivial.
So what do you think ? it is better to follow with octree mesh or passing by patch dependent and delauney
thank you Siw

siw October 3, 2012 02:20

Tri Tol can be found in Settings > Model, but as I said that's a bit of a longshot.

I use the octree method by default as the first step in making a hybrid mesh, I only ever use is as a stepping stone to getting a surface mesh. I then always replace the octree volume mesh by a Delaunay mesh. I've had less success starting with surface mesh - the reason for needing it the other pot is as I need to use autoblock aswell so I need the unstructured tri cells to conformal with the autoblock nodes on adjoining surfaces.

Maybe Far's topic could help

The other alternative is to upload your files and maybe one of the ICEM experts can resolve it.

hadrien51 October 3, 2012 06:33

thank you siw for your wise advice !
I will post my last mesh way if it can help some one

PSYMN October 3, 2012 13:25

Yes, SIW is right... reduce the tri-tolerance by a few orders of magnitude... It will render your geometry more cleanly so Octree has a chance of capturing it.

Not a long shot at all.

If that fails, then try adjusting edge criterion, but it should be enough.

hadrien51 October 4, 2012 10:09

Thank you SIW and PSYMN
It is working perfectly
I decreased by 10 the tri-tolerance and the octree mesh was perfect !
after delaunay and after prism layer
thank you for the support !

All times are GMT -4. The time now is 02:26.