[somas123]

The actual geometry is a hollow cylinder and I'll be using the axisymmetric condition, hence I've taken only half part of the cylinder and rotated it by 5 degrees. I've highlighted the problem areas in the attached pic. The initial problem was that at these highlighted corners, 7-node hexa elements were created. So CFX wasn't accepting the mesh. Hence I used the divide 7 node hexas into pyramids option. Now I'm able to import the mesh but CFX generates extra surfaces (which are the faces of the mesh element) at those corner nodes. Because of these extra surfaces I'm unable to use the symmetry BC in my simulation. I even tried assigning the symmetry BC to these extra surfaces but the solver still gave an error. What am I supposed to do?

[JuPa]

A nice trick is to replace the axisymmetric "pizza slice" with a "wedge". The advantage of this is you completely avoid pyramid and triangular cells. You can get a high quality hexahedral mesh.

So instead of the geometry finishing to a point like it does in your example, make it finish to a tiny flat surface.

This paper did exactly what I'm referring to: http://www.tandfonline.com/doi/abs/1...32.2015.994459

[ghorrocks]

What meshing software did you use to generate this mesh?

I do not understand the original problem. What is a 7 node hexa? Do you mean a 8 node hex with one degenerate edge where two nodes are coincident?

If you take a quad meshed surface and rotate it about one edge you get 8 node hex for most of the geometry and 6 node wedges along the rotation edge.

You have to be careful with Mr CFD's suggestion. While that approach does eliminate the problem at the rotation axis, it does change the geometry. Whether it is significant or not depends on what you are doing.

[somas123]

Originally Posted by ghorrocks

What meshing software did you use to generate this mesh?

I do not understand the original problem. What is a 7 node hexa? Do you mean a 8 node hex with one degenerate edge where two nodes are coincident?

If you take a quad meshed surface and rotate it about one edge you get 8 node hex for most of the geometry and 6 node wedges along the rotation edge.

You have to be careful with Mr CFD's suggestion. While that approach does eliminate the problem at the rotation axis, it does change the geometry. Whether it is significant or not depends on what you are doing.

I used ICEM CFD (ANSYS 15 version). In the 2nd image you can see that the extra surfaces formed are basically 2d mesh elements. That element's one side is on the corner (located on the axis of symmetry) of the geometry whereas the other edges are inside. Therefore a total of 7 nodes are available for the element instead of six or eight. I am getting such elements at the corners because of the O-grid that I have used to specify first cell height. CFX wasn't accepting the mesh because of these 7 node elements. Hence I had to divide them into pyramids and now it is accepting the mesh but it shows these extra surfaces which are clubbed under primitive 2D.

[ghorrocks]

In that case, in ICEM try defining the elements which are at the end of the sweep as parts, and make sure the correct element faces are in the parts and it should flow through to CFX.

In ICEM if you do a rotate mesh operation it should generate correct wedge elements on the rotation axis. This would avoid this problem entirely.

[somas123]

Originally Posted by ghorrocks

In that case, in ICEM try defining the elements which are at the end of the sweep as parts, and make sure the correct element faces are in the parts and it should flow through to CFX.

In ICEM if you do a rotate mesh operation it should generate correct wedge elements on the rotation axis. This would avoid this problem entirely.

I have generated my mesh using a similar method. First I created a 2d block by using sweep method. Then I meshed only the surface, to which the 2d block was associated, using pre-mesh. After this, I converted the 2d block to 3D using rotate option and then updated the mesh. The main reason why I can't get a wedge at that corner is because the O-grid is dividing the block at 45 deg. at that corner node.

[somas123]

Is there any method by which I can convert the mesh in those problematic areas to tet mesh and then merge it with the hex mesh in other areas?

[ghorrocks]

ICEM can handle doing that, and much more. There is not much mesh editting you cannot do in ICEM. But ICEM is a complex package and takes a while to get used to. Make sure you have done the tutorial examples (from the ANSYS Community website).