CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] Uncertainty about the mesh (

lost.identity June 7, 2009 12:37

Uncertainty about the mesh

For the last 6 months I've been using ICEM to mesh a simple convergent-divergent nozzle. As you could imagine this is a fairly simple problem but I can't seem to get the correct results from my CFD simulations (even after 6 whole months working on this).

Of course the problem might be with the CFD code, but I still have some doubts on the mesh and I want to remove these uncertainties once and for all.

The critical bit of my geometry is the nozzle itself and I'm attaching a picture here to show it.

1) I notice that the surfaces at the curved edge do not pass through all the points (the curves do however). I'm not sure whether this is really an issue.

2) Let say I have two curves which are horizontal and joined to each other. Also lets say I have two edges corresponding to these two curves. Now when I do the association, would it matter if I associate one curve to the corresponding edge and the second curve to its edge OR can I just select the two edges and then select the two curves?

3) To me the mesh is not very good because the mesh lines suddenly changes from horizontal to curved when it encounters the curvature. I tried to smoothen the mesh but did not achieve anything good. Is there another way to make this better?

4) Since my problem is axisymmetric I'm only considering the top half of the geometry and a segment of 10 degrees. The meshes at the centreline are not very good with 2x2x2 determinants close to 0.04. Surely this must be the case for all such segments.

I know that this is a long post but I'd be grateful if anyone can point out anything obviously wrong with the mesh. Cheers.

fauka June 10, 2009 06:19

1 Attachment(s)

1. I think the surfaces are not matching with the curves just because of the problem in the graphics. Otherwise the mesh will be very bad near the boundaries.

2. Either ways is ok I believe.

3. I really dont see any big problem with the mesh. But you can try the one shown below.

4. s, you can increase the angle to get good quality mesh.

please correct me if I am wrong

lost.identity June 10, 2009 08:57

Hi, Thanks for the reply. I'll try to mesh it as you've shown me.

The edges which I associated to the curve are linear. I read somewhere today that says that the edges can be edited so that they can become a spline. My question is do I need to edit the edge to make it a spline? I thought ICEM will associate a linear edge to a circular curve, for example, without having to make it a spline.

Rather than using the in-house CFD code I've decided to test it using CFX now just to see what's happening. I've got some issues with setting the periodic boundary conditions in CFX (I've posted this on the CFX sub-forum).

fauka June 11, 2009 02:01

Associating the edge with a curve is sufficient to make the edge represent the curvature. But if the curvature complex with many twists, it will be a good idea to split the curve and make it a spline.

PSYMN June 12, 2009 12:26

Just one more thing...
Or maybe just throw some linear splits in there to better fit the curvature...

But that is not needed for this model. Association is sufficient.

If you prefer to see the edges curved, you can display it that way using the display options accessed by right clicking on Blocking => Edges in the model tree.

All times are GMT -4. The time now is 01:32.