Is there a way to coarsen only a part of interior Hexa elements?
 

Hello everyone,
I have used blocking ogrid to create a mesh for a cylinder, consisting of hexa and quad elements only.
Since there are many curves on the surface (wall) of my geometry, I chose a great number for edge nodes to make the mesh follow the geometry properly.

However, I don't need these fine elements deep inside the interior of the volume, leading to my question:
Is there a way to only coarsen (somehow combine) inside hexa elements to reduce computation time and cost, while keeping them finer on the way to near the surface?

Any help would be greatly appreciated.
Attachment 69636

Hi atafiroozi,

Yes, somewhat.. There are so called refinement levels which you can assign to specified blocks. each additional level splits the underlying mesh an additional time. In the Meshing Parameter Tab, you'll find a refinement menu. It's in the same tab where you find the edge parameter setter. You'd need to set a higher refinement level at the circumference than in the center. However, this technique might be not appreciated by your solver... as you can see here: https://www.cfd-online.com/Forums/an...-elements.html

Therefore a more complex blocking would be my suggestion which leaves fewer projections from the edge to the center. There is a technique called nesting, or clamping - basically an internal c-grid - to map 1 edge onto 3 edges. You might find more information here: http://blog.gridpro.com/nesting/

In your case you'd need to create a blocking such as the following drawing. Starting from you current o-grid, you'd need to insert a c-grid in each quadrant separately, as indicated in the drawing.
Attachment 69681
Of course this blocking will need some attention and testing on vertices positions to get good cell qualities.


Best regards,
Sebastian

Hi Sebastian,
Thanks for your reply.

But I don't quite get the drift about the c-grid making procedure.
I mean I couldn't find related command option, is it under the "Split Block" menu?

Would you elaborate on using this method in ICEM?
Thanks in advance. :)

Well, you have used the o-grid tool at least once to create your initially shown mesh, did you? Are you confused with the term "c-grid"? A c-grid is just a quarter of an o-grid. It's used in cases where the "o"-layer is not a closed loop anymore. However, you can create it with the o-grid tool, too.

The key idea of the o-grid tool is, that you selection determines which edges/faces are to be extruded into blocks of the o-grid.

If your mesh is 2D you'd need to create 4 additional o-grid, each being a c-grid in its respective quadrant.
Each of these o-grid is created by selecting the one block of the respective quadrant. Deselect any other automatically added blocks. Then, add the circumferential edge to the selection. That way, each edge but the outer will be extruded inside, leaving you with a c-grid.
In 3D you need to select the top and bottom face of that quadrant prism to prevent them from extruding.

I strongly suggest you to have a read of the sticky thread in this subforum https://www.cfd-online.com/Forums/ansys-meshing/116369-before-you-start-thread-see-if-you-can-find-answers-here.html

You'll find a guide called Simon's Tips & Tricks. Page 76 and following will be most interesting for you.


Best regards,
Sebastian

CGrid is OGrid, when you select one face or edge. The meshing shape will look C

Thank You Sijal, I get it now.

Dear Sebastian,
My case is 3D and I've chosen these three faces:
Attachment 69730
After creating one c-grid in upper quadrant pre-mesh looks like this, which is not desired:
Attachment 69728

Actually what I want is to unlink these two edges shown between arrows, in order to give them each independent node numbers (concentrated mesh in upper edge and coarse mesh in near-core one):
Attachment 69729


Now I know this is a job for so-called c-grid, but I couldn't find the right way to do this so far. :confused:

It would be wonderful if you'd guide me further on doing this. :o

Explain, why your c-grid mesh looks undesired? What makes you think so? What did you expect? What steps did you try to resolve this?


From my point of view, this is exactly the structure i suggested. It seems, you just haven't moved any vertices jet. Make the c-layer blocks as big as its enclosed block. Then assess again.



You might need to rethink about the implications of structured meshes. Parallel edges will always have the same number of nodes. The proposed c-grid reduces this dependency. Further reduction of the dependence of the core edge to the perimeter edge could be the introduction of further layers in the cgrid, making the edge which is associated to the core shorter and shorter, while the c-layer edges fill the circumference.


If this gets to complicated to you, you might should think about a swept mesh. A swept block has a top (and bottom) face which are meshed unstructuredly. This top face mesh will then be extruded in layers to fill the block. For example use a all quad face with really independent edge counts. You'll probably still have a structured mesh in flow direction.

Thank you Sebastian,
Your comments were helpful.