CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] Simple Geometry Poor Results (

tav98f August 2, 2010 12:00

Simple Geometry Poor Results
2 Attachment(s)
I have been trying to mesh a tube that starts with a circular opening and goes to an eliptical outlet. I have been getting very poor results.(See picture) :(

Attachment 4284

It has a length of about 0.1 meters and an average hydraulic diameter of about 0.0175 meters. There is more to my geometry below the tube but I am confident it is not effecting this tube part.

I have provided the parameters used to mesh my geometry (see other picture). I have tried with different volume and surface meshing methods but I have had very similar results.

Attachment 4283

I have not smoothed my mesh while volume meshing so that I could show you more clearly what is happening. After volume meshing and taking a screen shot I smoothed the mesh for 10 iterations up to .2 and 10 iterations up to .3 and took another screen shot. After the smoothing my elements are better but the overall mesh is still terrible because of the two bands of very small cells.

Any insight as to why it is behaving this way would be very helpful.

OzMantle August 2, 2010 19:47

2 Attachment(s)
I think is better if you try to make an hex mesh with O-grids. First create a block that contains yoru geometry, the split them with an O-grid (you must see that ICEM creates a little block inside the big one that you create). Those little block will be respected for the software in the mesh creation, so you may expect that the elements in the little block have a high quality (over 0.85), then just Associate the edges of the big block to the curve (on each side of your geometry) and finally move some vertex for better results. You finally obtain something like this. :cool:

tav98f August 3, 2010 12:40

Thank you for your reply. I will try to figure out exactly how to do this. I have not worked with an O-grid before and I will first try the tutorials on o-grids before I proceed. I will let you know if I get somewhere with this. :)

PSYMN August 3, 2010 22:22

Edge criterion.
Right, Hexa with an Ogrid is certainly better for this portion of the geometry... However, you mentioned there was more and perhaps the rest is too complicated, so I will tell you how to proceed with Tetra...

What you are seeing is over refinement. Since it is an Octree process, if it feels it needs to refine once more, it has to cut the mesh down to half the size, and you notice it right away. It could be caused by two of your settings.

There is a slight chance that your curvature and proximety refinement is the cause. Try turning that off. Just set the size you want on the surface and mesh it like that.

But most likely, you can keep your curvature refinement on and increase your edge criterion from 0.05 to 0.2. This number kicks in during the portion of the octree process when the mesh is fitting to the geometry. The algorithm checks each node to see if it should be projected to the surface or if the edge should be segmented to create a new node. If the node is within this criterion of the surface (in your case 0.05), then it moves the node. If the node is further from the surface (say 25% of the edge passed thru the surface) then it would split the edge (refining) and try again with the new node. If the new node was within that 5% of the surface it would move it, but since you set such a small criterion, it will split again. Because your model is tapered, it slopes relative to the original octree mesh and you see zones where more refinement was necessary.

Anyway, the short version is increase your edge criterion up to 0.2 and you should be fine.

tav98f August 4, 2010 13:36

Right Again
Wow! That worked like a charm.:) I am ashamed that I spent this much time trying to figure this out before I asked for help. I was begining to think that there was problems with my goemetry. My time spent looking into O-grid may be useful for other projects but tetra is definately the way to go with this one.

Thank you for the reply.

All times are GMT -4. The time now is 12:30.