CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Native Meshers: snappyHexMesh and Others (
-   -   snappyHexMesh - geometry edges (

johannesk April 8, 2009 04:46

snappyHexMesh - geometry edges
Hi Everybody,

My name is Johannes, I'm new to OpenFOAM (well 3months experience by now). I'm working on my diploma thesis in mechanical engineering at the University of Karlsruhe, Germany.

I'm trying to simulate a test rig with a big complex geometry (10mio cells is still a coarse mesh). At the moment I'm using a .stl file generated with ICEMCFD. I succeed in generating a mesh with snappyHexMesh from the stl-file, but it does not respect edges.

Do I need an eMesh file for the features list in the castellatedMeshControls subdirectory? The manual says, I can then give the level of refinement for the lines/edges. Does this mean the mesh is splitted at the edge, and will therefore respect it, even after snapping to the surface? I guess I can then also play around with the quality settings?
I have not yet been successful in exporting an eMesh file from ICEM. If the emesh file is the way to go, can someone give me a hint on how to get an emesh file from ICEM?

Thanks in advance.

wolle1982 April 9, 2009 05:52

hi johannes,

I'm also struggling with the same problem. but as i see so far, there is currently no way how to keep the edges "sharp" without being displaced. you also could get this information out of the other threats here.

paulo April 10, 2009 06:25

We also have problems here with edges (We work with wind blades). If you run the motorbike example, we can see the problem in the front of it.

We minimized the problem defining a refinement region and increasing the refinement level there. The results got better (the problem still remained).

Hope this helps.



johannesk April 11, 2009 19:43

too bad

I'm currently trying something different. Basically I'm avoiding stl-geometry info in places where I don't want to have a mesh refinement. Which means I have to generate a blockMesh that follows my external geometry, only using a stl geom where necessary.


lhcamilo April 12, 2009 05:18

Snappy can be rather tricky, I was working on a two-wing spoiler a while ago and snappy had problems attaching a boundary layer to the leading edge of the wings. After some tweaking I noticed a few things

  1. The surface levels needed to be the same, as in (n m) where n=m ... the BL would not attach well if the values were different n != m
  2. Create as many patches as possible. As I said before My structure was a spoiler composed of two wings [and a side plate on each end]. If I tried to use Snappy on the whole structure the BL would not attach at the LE. Only when I created a patch [actually STLs] for each wing with Salome, then the BL issue was solved. This may be a little impractical for very complex structures
  3. As a last resort try refinement regions.
hope this helps

bastil April 12, 2009 12:35


it has been discussed some time ago that feature capturing is in development and will be containe din further releases. Maybe Eugene can tell us more?

Regards BastiL

openfoam_user September 23, 2009 08:59

Dear snappyHexMesh-developpers,

When will the feature for capturing edges be available ?



openfoam_user January 13, 2010 04:46

Dear all,

someone has news about the development or not of the feature for capturing edges with snappyHexMesh ?

For some kind of geometries such as propeller blades having sharp edges is very important. Maybe someone has found another solution to overcome this problem.

Best regards,


bastil January 13, 2010 18:13

Well problems here, too. Apart from that snappy works fine for us...
I see so many people interested in this feature so I would like to propose the following:
All interested persons/companies give an development contract for this feature to someone capable to develop this (e.g. OpenCFD). Costs can be shared if so many people are interested in this feature still not present in the code. I guess funding is the main reason why this is not yet in. Who would in general be interested in taking part?
Another option is using the ICON implementation which - unfortunately - is not freely available. Has anyone experience with this?


All times are GMT -4. The time now is 17:15.