[BrolY]

Here is an email exchange I had with Simon Pereira:

Questions:
I have a few questions for you regarding the blocking in ICEM CFD. Our current version of ICEM is 12.0.1.
1) Our geometry has a plane of symmetry. So, we created the blocking for the first half part of this geometry, and we want to use the symmetry option for moving our blocking to the second half part of the geometry.
But, when we associated the symmetric blocking to the 2nd geometry, it didn't work very well. The blocks or the edges are not associated with the right part or curves... Do you have a solution for this issue?

2) We created an interface to close the mesh of the first part of the geometry. When we create the whole mesh thanks to the symmetry option, the two parts have these interfaces which prevent the fluid from moving from one part to the other.
We tried to delete these interfaces, but the final mesh wasn't very good. Do you know how to handle with that?


Answers:
1) It could be that your blocking is mirrored but you still have differences due to grouping of curves. For instance, on the initial side, you associated one edge to three curves. This would automatically group those three curves (concatenate) into a single curve and then make the association. When you mirror that blocking feature, the edge tries to associate with a curve, but doesn't know to group the three curves. In these cases, the simplest sure fix is a manual one. Right click on edges (in the model tree) to select the "Show associations" option. This will create little green arrows that shows what blocking is associated to what curves. Then use "associate edge to curve" and correct any problems that you see.


2) After mirroring, your symmetry plane is in the way... First thing to check is if the nodes automatically merged. Right click on verts in the model tree and turn on the option for "numbers". Look at the symmetry vertices to check for numbers. If you see two numbers overlapping, you will need to merge them together (using merge vertex). Once they are merged, if the volume materials on either side are the same, the face edges should turn blue and the shells should no longer form. If it still gives you symmetry shells, you can go into the blocking (tab) => Edit Associations => Face to Surface and select "none". Then remove the associations from the symmetry faces (you could select individually, or from corners, or by Part).


Just one comment about your process (sorry if it seems obvious to you, but your first question made me wonder)... If your model is truly symmetrical, then it is better to simply use that symmetry boco (if your bocos are symmetrical) or perhaps mirror the mesh and delete the symmetry shell elements (if your bocos are not symmetrical). When you mirror blocking, you create a bit of a headache with the index control, so it is only worth doing if your model is somewhat similar, or mostly topologically symmetrical, but the geometry is not truly symmetrical. Then you can take advantage of the elastic nature of the blocking to fit to the actual geometry and save yourself the time of blocking it. You could also add (or remove) holes, etc. that are not symmetrical.

[PSYMN]

Thanks for posting this so others could benefit.

Simon

[ringtail]

what is "boco"?
Maybe this question is simple, but I'm a beginner of ICEM CFD.
thanks

[PSYMN]

BOCO is short for "boundary condition"

Instead of actually mirroring the blocking, if the boundary conditions are symmetric, you could just apply a symmetry boundary condition and get the same solution with half the mesh.

However, if the model is symmetrical, but the boundary conditions are not, then you must at least mirror the mesh before you apply your boundary conditions.

It is only worth mirroring the Hexa blocking if your model is only mostly symmetrical. You could mirror the blocking to save time on the second half, but then continue to fit the second half of the blocking to the geometry, perhaps adding more splits, drilling holes, extruding faces, etc. to make any "non-symmetrical" adjustments. The flexibility of blocking makes it perfect for this.

[BrolY]

Hi all,
First of all, our BOCO are not symmetrical, only our geometry is. So we created the blocking for the first half of the geoemtry and we rotate our blocking on the second half.
But the issue is that the blocking isn't fitting well the second half of the geometry. There are too many edges and curves to associate themselves manually. Is there another solution?

For the interface, we merged the nodes but it created pyramids along the interfaces. How can we avoid that?

Thanks,

AB

[PSYMN]

Pyramids? I thought this was a hexa model. You should only get pyramids as a way to connect hexas (or prism sides) with tetras (or prism tops).

If your geometry is symmetrical, I would just mirror the mesh. If you have some differences in the part naming, there is an option under "Edit Mesh (tab) => Repair mesh" to associate the mesh with the underlying geometry.

After you mirror, the nodes should be lined up (hexa elements on both sides). Just do a simple "Edit Mesh => Merge Nodes => With a tolerance" and set a very small tolerance (much smaller than the mesh size). Make sure to turn on the ignore projections option. This will merge nodes on both sides of the mirror so that the hexas are all connected. Next, you will get rid of your symmetry quads. I would do this with "Edit Mesh => Delete elements" then use the selection tool bar to select items in a part (hot key P), and pick that symmetry part.

That should be all you need to do.

If that doesn't work for you, then I am missing some key bit of info. If you can send me your model and I will figure it out for you. It is probably something simple.

Simon

[BrolY]

I think I didn't explain our problem very well:

When we mirrored the blocking, the association didn't work well when we displayed the mesh with "the pre mesh" option. One way to prevent this issue would be to associate the new blocks with the geometry manually, but we would like to avoid it because they are too many curves and edges in our model. Do you know how to do that?

Thanks !

P.S: we had pyramids on one of our test when the premesh was converted in unstructured mesh. But this is not a matter anymore

[BrolY]

Our geometry was fully symmetrical, so we only wanted to create the blocking for half of the geometry, and then rotate the blocking. But it didn't work because the association was very bad: the rotated blocking was still associated with the first half part of the geometry.

But, hopefully, we found a way to fix this issue:
1) we rotate the blocking
2) update association (only vertices and edges, NOT faces)
3) meshing parameters with "keep counts" options

And the pre-mesh worked well.

[camoesas]

Originally Posted by PSYMN

Pyramids? I thought this was a hexa model. You should only get pyramids as a way to connect hexas (or prism sides) with tetras (or prism tops).

If your geometry is symmetrical, I would just mirror the mesh. If you have some differences in the part naming, there is an option under "Edit Mesh (tab) => Repair mesh" to associate the mesh with the underlying geometry.

After you mirror, the nodes should be lined up (hexa elements on both sides). Just do a simple "Edit Mesh => Merge Nodes => With a tolerance" and set a very small tolerance (much smaller than the mesh size). Make sure to turn on the ignore projections option. This will merge nodes on both sides of the mirror so that the hexas are all connected. Next, you will get rid of your symmetry quads. I would do this with "Edit Mesh => Delete elements" then use the selection tool bar to select items in a part (hot key P), and pick that symmetry part.

That should be all you need to do.

If that doesn't work for you, then I am missing some key bit of info. If you can send me your model and I will figure it out for you. It is probably something simple.

Simon

HI Simon, Hi Everybody,

I am still having problems mirroring my mesh the way you described it here.

1. I am mirroring the mesh (Merge Nodes: automatic; Delete duplicate elements)
2. Merge Nodes: Like you have described it: (ignore projection, tolerance 0.01)
3. Delete Symmetry Plane

But Check Mesh still gives me:
- uncovered Faces
- multiple Edges
- non manifold vertices

I ve tried a lot like varying the tolerance, mirroring with and without symmetry Plane but it still doesnt work. Maybe somebody see the mistake??

Thanks

[BrolY]

Non manifold vertices are usually due to association problems.
Be sure you have associated all the edges at the interface where you merge the nodes.

Multiple edges is only a possible problem. If you have a T junction for example, it's normal to have multiple edges.

Uncovered faces means some shells doesn't belong to any parts. Some Fix the trouble and adds the shells to the right part.

P.S : post some pictures, it might help

[camoesas]

HI Broly,

Here is my case. As you can see its an simple cube on sticks. I dont know what screenshots could be of any help. So if you need additional information please let me know.
If I choose add uncovered edges to Fluid I get a Wall in my Fluid. The Symmetry plane is back just with a different name....

Now I am getting additional error messages when checking the mesh:

- probleme volume elements
- duplicate elements
- volume orientation


I am on the right way but in the wrong direction...

edit: blue is the symmetry plane, pink is the cube.

[BrolY]

Don't add the elements to the fluid ! If the uncovered elements are attached to the symmetry plan, then add the elements to the symmetry plane part.

if duplicate elements appears, delete them.

[camoesas]

HI Broly,

but id like to delete the symmetryPlane part. Somehow I have to add the former symmetryplane elements to the inner fluid. If i add them to the (now empty) symmetryplane part I simply revive it.?

[BrolY]

Post an image with the location of the uncovered faces.

[camoesas]

HI BrolY,

Here is a snapshot of my unconected vertices. They are exactly were there was the symmetry plane.

[BrolY]

Unconnected vertices is not an issue.
Post the same picture with the uncovered faces.

[Far]

if possible post .tin and .blk

[camoesas]

Originally Posted by BrolY

Unconnected vertices is not an issue.
Post the same picture with the uncovered faces.

Sorry I mixed it up. At the moment I am not in office I will do so next week.

Thanks

[camoesas]

HI Broly,

Here are my tin and blk files:

[Far]

Steps are :

1. Mirror geometry
2. Mirror blocking
3. Update association
4. Edge to surface association

Make sure surfaces are also copied and rename them as deemed necessary.

Both files are attached: 1. Before mirror 2. After mirror (Blocking is bit different)


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08


Shot at 2012-08-08