Hello,

I'm trying to mesh a 3-D segment using ICEM. I can get the mesh, but when I check the quality of the mesh it is very distorted because of the curvature in it.

I'm using Top-Down approach using Hexa meshes. I tried to use O-grid meshing but it didn't work for my geometry for some reason.

This must be a trivial issue for someone more experienced in mesh generation, but I have no clue on how to fix this.

The link below shows the problematic elements in the mesh. Any suggestions is much appreciated.

http://tinypic.com/view.php?pic=24pa2id&s=5

Thanks

James

To be sure, that is one of the nastier type of geometry to mesh with hexa. The best solution is, if you can, to change the vertical line going into the acute angle so that it is no longer vertical. Not knowing what your geometry is like, that may not, of course be possible. If your solver will handle a triangular block (i.e. some slender wedge-shaped cells), it may be best to collapse the top face, in order to get a triangular topology that fits what is more or less triangular geometry. Failing that, try a quarter O-block.

Thank you for the quick response. I have tried your first suggestion (slanting the vertical line) and it worked! I do not get the bad elements as I did before.

I have another problem, and I think this is because of the inherent issues with hexa meshes for my geometry. The link below shows the geometry from a side, as you can see it is an angular segment. When I plot the determinants it highlights that the elements and the edge have nearly 0 discriminant.

http://tinypic.com/view.php?pic=2jikc1&s=5

Is there any way to resolve this using hexa meshes? BTW what is a Quartet O-grid? Is it available in ICEM? Someone suggested using Delaunay triangles and using a free tool from the internet to do the meshing. Again, I'm not sure how I'd achieve that.

Apologies if my questions are really obvious (I'm a complete newbie to ICEM and mesh generation).

It looks like you are working with an angular segment in order to do an axi-symmetric analysis. The simple answer here is that if you are using a solver that can do 2-dimensional axi-symmetric flow, then it is much better to do a planar 2D mesh on the X-Y plane and tell the solver that it is really axi-symmetric. Many solvers (Fluent, CFD-Ace, CFD-Fastran, Cobalt, Adapco Comet, etc.) can do this, although some, such as CFX, Star-CD (?) and OpenFOAM cannot. Failing that, just work with the segment and hope that the mesh quality issue is not critical.

It is easy to do a quarter O-grid in Icem. You can either split the block manually to do this, or you can create a blocking topology that would have a wedge-shaped block, then delete this block and build a quarter O-block from 6 vertices (look through the hexa menus, it is there).