Learning Forum

[ishan nande]

I have an axisymmetric case for which block-structured mesh needs to be generated in ICEM. I start from 2D blocks, mesh them and rotate the 2D blocks to convert them to 3D. This will generate singularity at the origin and to avoid this we need to generate O-Grids. There seems to be no way to handle this scenario in a single file. The way I am doing this is by creating the OGrid blocks separately and then merging them with the parent blocks. This is painstaking and the spacing of the OGrid blocks is distorted.

[ishan nande]

A little clarification and more details. My final mesh will be close to 50 million or even more. I created a script which takes care of this rotation and blocks merging. I specifically went  for this approach because I can then run this script on the server and go home, come in morning and have my .msh file ready. It works but it's not straightforward. Thank you for reading.

[Rob]

If you're using circumerential sweeps why not use Ansys Meshing and then pave the centre block? O-grids were great for structured multiblock but not all that common now.  Similarly, with what you're proposing the cell quality at the outer circumference may not be ideal: there's a reason we tend to favour poly mesh for many applications. 

[ishan nande]

I unfortunately cannot use non-hex cells for my application. Will ANSYS meshing create hex cells only ? I have a student license with access to 16 cores. I was originally creating only a quarter portion when I went from 2D to 3D. The model quickly becomes too complicated when you try to mesh it in 3D. 

[Rob]

If you have parallel it's not Student, possibly Research?  Ansys Meshing may add a few tri/prism elements but that shouldn't be an issue for the Ansys solvers. 

[ishan nande]

I think it's research. The solver is not any of the ANSYS products.  The solver is for LES and I really want to avoid any kind of non-hex cells. With regards to the approach you mentioned, is there any specific tutorial or video you can please point me to ? Also, can ANSYS meshing be automated through a Python script ? I unfortunately cannot visualize large cells. Ideally, I would want to generate a 2D mesh, sweep it radially and then generate the centre OGrid-style mesh. I can then write a shell script to take care of post-ANSYS operations.

Thank you for reading. 

[Rob]

I'm not sure Ansys Meshing is, but Workbench input and output can be controlled from other tools (Twin Builder for example). 

[ishan nande]

Thank you. I will look into this. One last question regarding ICEM. When I rotate the 2D blocks to 3D, the edge parameters seem to lose their grading. The 2D blocks are created inYZ plane and the rotated along X axis. The grading in X direction seems to be completely ignored when the rotation happens.

 Any tips on why this could happen ?

[Rob]

If the surface is on the yz plane, won't rotating on the x-axis smear the surface? If the system uses the x-axis as the centreline shouldn't the blocks be created in xy plane?  It's been a long week, so apologies if I've missed something. 

[ishan nande]

Yes!! Sorry. My mistake. The blocks are made in the XY plane and then rotated along the X axis.  This is when the edge parameters seem to lose the node grading information. 

[Rob]

Not sure, I used ICEM CFD's predecessor in the 90s, but not the current form. 

When you revolve the block does it ask for an edge size or element count? Either way, is that set on the inner or outer annular face? Have you checked the tutorials?

[ishan nande]

Yes. I did. There  are no 2D to 3D rotation tutorials. ICEM asks for angle of rotation of each copy, nodes per copy and number of copies(attached image for reference). 

https://pasteboard.co/VJy9l12NvS6f.jpg

[ishan nande]

I figured out what the issues is. When the rotation of blocks is done, the linked spacing feature is not updated. In the image below, the selected edge's end spacing is linked to first spacing of edge 43 49. When the block rotation takes place this linked information is lost. Maybe I am doing something wrong or the issue is case specific but this should not happen. Thanks a lot for the help.