CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] Multiblock plot3d (

papis March 31, 2013 18:41

Multiblock plot3d
2 Attachment(s)
Hi to all,

I am opening an new thread about this issue I have, with a plot3d mesh.
I have a grid of wind turbine blade of approximately ~5million nodes.The grid is in plot3d format and is split in 144 blocks with no Iblank arrays.

I use ICEM CFD in order to open the mesh apply boundary conditions/minor changes and export it in the format that I want.

When the mesh is imported ICEM asks if I want to recompute mesh topology or use basic topology where it's domain it's unattached to the other.

At first i chose recompute mesh topology.What ICEM did was to remove the double interblock boundaries.The way the plot3d mesh is structured each x,y,z coordinate is repeated at interblock boundaries and thus double nodes/lines/shells are created pointing to the same x,y,z coordinates.

By recomputing mesh topology the double nodes/lines/shells were removed.However the interblock boundaries remained(I will explain later about that) and also the check mesh utility identified problematic edges at interblock boundaries.

By using the basic topology where each domain is assumed unattached to another the check mesh utility showed no problems.However at each interblock boundary I had double nodes shells etc..

My main issue about ICEM recognizing the inter-domain boundaries as real boundaries is that I have to create a part (interior/wake bc) which is not easy/error-proof when you have 144 blocks and mesh with first-layer 1e-06 with 5 million nodes.I need a systematic way of doing that or way to make ICEM understand that the domains should be merged thus leaving only the real boundary shells(which I assume is the correct way).

Another thing that I tried is splitting the mesh to 144 single plot3d grids.I thought if I load them one by one maybe something would change.Indeed when I loaded the second block ICEM asked if it should merge the mesh or replace it.I chose merge and wow the interblock boundary disappeared without asking anything about topology. However even this workaround was not plausible because the 5th block didn't have any actual connection with the 4th and when I loaded the 5th everything else disappeared (also I didn't liked the idea of loading 144 files by hand)...(I hope i didn't confuse you)

Does anybody have an idea of how to address this problem?I think I am missing something because identifying domains that are attached and merging them should be a trivial task for ICEM (or not?).

Also I would like to point out that because the mesh is imported no Geometry is present.

Any idea would be appreciated

Thank you very much for your time

I have attached a picture of the mesh.the lines show the interblock boundaries and the shells are the the surface bound cells.

Bretert July 16, 2014 11:32

Dear Papis,

have you already solved your problem? I have met the same problem and I am wondering if you have found a solution.

papis July 16, 2014 11:42

Hi Bretert

No I haven't found something that makes icem disregard the double nodes edges.However, what I did was to create a different part for the interblock boundaries and apply wake-like boundary conditions to that part(wake like in the sense as in 2D structured grids double sided wake line).

What helped me a lot in creating these parts(because if you have lot's of blocks it will be problematic) is the hotkeys "l","k","r" . When creating a part you select the first element and then move using those hotkeys since using the mouse is impossible at least at the beginning. "l" select elements up to an angle,"k" selects element by directions and "r" up to a curve.

Another thing that I found is that gridgen by pointwise can actually merge those different domains however I don't think that's a proper solution since it doesn't apply to ICEM.

Good luck

All times are GMT -4. The time now is 00:09.