Hi

I am trying to generate a structured mesh in a small port using Hexa in ICEM CFD 5.1. If I create a block along the length of the pipe the cells at the four corners are highly skewed. Generating an o-grid is no better as the cells in the corners of the blocks at the end are also skewed. The amount of skew is unacceptable for my in-house solver.

I have tried to use the smoothing functions but the program just seems to hang. The manual does not say how to actually smooth the grid, just that it is possible.

Is there another solution or can somone explain how to smooth the grid? The in-house solver I am using can only accept a structured, cartesian, rectilinear grid.

Thanks

Lewis

in http://www-harwell.ansys.com/demoroom/, there're many demo swf files to show how to create mesh by tetra/hexa/hybrid. I've learned a lot from it! I am sure you can move the vertice to adjust the skewness (please refer to the turbomachinery/Demo2 case).Hope it will help you.

You should be able to smooth the angle/determinant in ICEMCFD5.1. I've no problem doing that so far eventhough it takes quite some times to finish. If you specify more than 10 smoothing steps, then it will take you ages esp when you use multiblock.

Adjusting the block topology yourself will make things easier.

1. Locate the problem area by checking the angle/determinant/skewness etc in the "premesh quality".

2.Use "Move Vertices" to adjust your block at the problem area & fix whatever axis if necessary.

Cheers, TB

What type of smoothing do you use? Standard smoothing for hexa grid is not very useful - program behaves like it is hang... Try to use hexa smoothing alone(orthogonal options for both surface and volume parts), not general smoothing. But it sometimes produces not so good results. Also try maniplulate vertices and block smoothing.

Moving the block verticies does not improve the mesh as there is a square block in a pipe; the verticies must be on the 'curve'. I have been trying to smooth in Hexa and, as you say, it seems to hang even after 6 hours on a dual 3Ghz Xeon machine with a 0.5 million cell mesh. I tried to smooth the mesh in Med (I am using the old gui) and that did not seem to make the mesh any better.

What I really want to do is to move the nodes in the corners of the slightly into the centre of the pipe to lessen the skew. Is this possible?

PS Thanks for the links to the tutorials Kiddo - they are good!

Even in old GUI of 5.1(MED) you can use orthogonal hexa smoothing - in utilities menu(as I remember) . Can you send some screenshots of your mesh?

Hi,

Are you using O-grid topology? I can get the min face angle up to around 40 deg using multiblock for a round section, which must be quite sufficient for most engineering problem.

Here is a screenshot of the current mesh http://www-staff.lboro.ac.uk/~ttljb/icemskewedmesh.jpg The skew in the four corners is too much for my solver.

TB; I have tried an o-grid but the blocks at the top and the bottom have the same problem. The blocks along the main part are fine with an o-grid.

Thanks for all you help!

Lewis

Can you post a 3D drawing for your geometry? I don't get why you do your blocking this way. On the way, your center block is still looked like a H-grid to me. You may have to think of dividing that into 5 block ( one in the middle and 4 others around the center block)

Cheers, TB