CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ANSYS Meshing] block strategy (http://www.cfd-online.com/Forums/ansys-meshing/110384-block-strategy.html)

yorelchr December 10, 2012 07:01

block strategy
 
Hello everyone,


I have few question about block strategy :
I have a geometry in which I have splitted the block in 4 blocks (4 blocks in "vertical" direction and 5 blocks in "horizontal" direction (but the horizontal split is not of interest in my question) .
For the first 3 ones, I want the edges to have the same number of nodes and it's ok.
For the last one, I want it to have different number of nodes, but since the edges of the 3rd and 4rth are adjacent, I don't know how to do that.
So I tried in a different way : creation of 1 block, splitted in 3 blocks. Then I created a 4rth block, but when I want to split it horizontally for creating my 5 blocks, it automatically "transfers" the split to the 3 vertical blocks. So it means that even if I created the 4rth block in an independent way (not by splitting), I have no independence for cplitting it, for meshing?
Thank you for giving me some clue.
Do I have for example to create 2 bodies? but the edges will still be in common...

yorelchr

yorelchr December 10, 2012 12:59

Quote:

Originally Posted by yorelchr (Post 396627)
Hello everyone,


I have few question about block strategy :
I have a geometry in which I have splitted the block in 4 blocks (4 blocks in "vertical" direction and 5 blocks in "horizontal" direction (but the horizontal split is not of interest in my question) .
For the first 3 ones, I want the edges to have the same number of nodes and it's ok.
For the last one, I want it to have different number of nodes, but since the edges of the 3rd and 4rth are adjacent, I don't know how to do that.
So I tried in a different way : creation of 1 block, splitted in 3 blocks. Then I created a 4rth block, but when I want to split it horizontally for creating my 5 blocks, it automatically "transfers" the split to the 3 vertical blocks. So it means that even if I created the 4rth block in an independent way (not by splitting), I have no independence for cplitting it, for meshing?
Thank you for giving me some clue.
Do I have for example to create 2 bodies? but the edges will still be in common...

yorelchr


Hi !!!
no clue from anybody? I really need your help: there must be a way to have adjacent blocks that are independent no? any idea, even if not "sure" would be welcome....it's been days I've been trying everything I could think about.....but no way....

energy382 December 10, 2012 13:02

i guess you're using icem hexa?! if so, can you provide your .tin and your blocking files or some pics? should be an easy one to fix.

yorelchr December 10, 2012 13:28

blocking strategy
 
2 Attachment(s)
Quote:

Originally Posted by energy382 (Post 396747)
i guess you're using icem hexa?! if so, can you provide your .tin and your blocking files or some pics? should be an easy one to fix.

Hi energy382

thank you very much !!
I send you the .tin and the block file.
My problem: for the "down" part (the block close to the axis of the geometry) , I'm trying to use the O-Grid, so that I can join 3 vertices with the 3 points on the curve at the top of this last block, and the 4th vertex with the 4th point (the one on the axis). then, if for example, if I put 10 nodes on the block before (adjacent to the last one), I can put 5 and 5 on the curve containing the 3 points .... Maybe I could also do it without O-grid but I need some more vertices I guess...anyway, the problem is that when I put the 10 nodes on the adjacent block, then I have 20 nodes on the curve containing the 3 points...
Maybe my explanation is not really clear. Have a look at the file I send you and tell me if you get my problem. Thank you very much !!!

PS: camera1.zip is the block file; camera.zip the geometry, but you should rename it camera.zipx for opening it. Couldn't send it as zipx since the format is not recognized by the cfd site. Tell me if it's ok.

energy382 December 10, 2012 13:41

Ok. Unfortunately, I'm out of office right now, but I'll take a look at it tomorrow.

PSYMN December 11, 2012 12:52

Quote:

Originally Posted by yorelchr (Post 396752)
I send you the .tin and the block file.

I tried to take a look, but the tetin file wouldn't unzip for me (strange because the blk file in the same archive came out no problem).

Anyway, this makes it more awkward to see what is going on...

Generally speaking, blocks are mapped, which means that the number of nodes on either side must match. If you have a series of connected blocks, they must all match (parallel edges must have the same number of nodes).

There are several ways to get out of this.

1) you could use "Pre-Mesh Parameters => Refinement" to make one block coarser or finer than the block next to it. Not all solvers are ok with the hanging nodes that result.

2) you could "Edit Block => Change Block Type" so the block was no longer mapped, in which case, you would be free to change the edge parameter.

3) you could try a different topology (such as a quarter ogrid or wedge block) to add or remove nodes from a blocking. Really, you are just changing what is parallel.

In your particular case, I wasn't exactly sure what you want to do, hence the generality instead of a specific solution.

energy382 December 13, 2012 09:18

1 Attachment(s)
the blocking was wrong. you have to block such a topology in the way i did.

you've to convert the degenerated blocks to y-blocks and adjust edge parameters.

yorelchr December 13, 2012 11:11

Quote:

Originally Posted by energy382 (Post 397339)
the blocking was wrong. you have to block such a topology in the way i did.

you've to convert the degenerated blocks to y-blocks and adjust edge parameters.


Hi energy382,
Thank you very much for your help.
The problem in doing so is that the cells on the axis will be really "degenerated".
In fact, I have tried with O-Grid (see PSYMN previous advice) and now I think it's ok for the meshing. But when I try to read the mesh in Fluent, I have an error message. I'm working on it.

energy382 December 13, 2012 11:16

no, you have to convert this degenerated blocks into quarter o-grids (y-grids).

blocking/edit block/ convert block type and then choose y-block and mark all degenerated blocks (the blocks with 6 vertices)

yorelchr December 13, 2012 11:16

Quote:

Originally Posted by PSYMN (Post 396954)
I tried to take a look, but the tetin file wouldn't unzip for me (strange because the blk file in the same archive came out no problem).

Anyway, this makes it more awkward to see what is going on...

Generally speaking, blocks are mapped, which means that the number of nodes on either side must match. If you have a series of connected blocks, they must all match (parallel edges must have the same number of nodes).

There are several ways to get out of this.

1) you could use "Pre-Mesh Parameters => Refinement" to make one block coarser or finer than the block next to it. Not all solvers are ok with the hanging nodes that result.

2) you could "Edit Block => Change Block Type" so the block was no longer mapped, in which case, you would be free to change the edge parameter.

3) you could try a different topology (such as a quarter ogrid or wedge block) to add or remove nodes from a blocking. Really, you are just changing what is parallel.

In your particular case, I wasn't exactly sure what you want to do, hence the generality instead of a specific solution.


Thank you PSYMN. I insisted with O-Grid and I think it is ok now. But problems when reading the mesh with Fluent.
- What is a "quarter" O-grid?
- I wanted to try your 2nd advice but the type was already on Swept. I don't even know what it means. Can it be a problem if all my blocks are like this? I can't read the mesh in Fluent and I'm trying to understand why...

yorelchr December 13, 2012 11:19

Quote:

Originally Posted by energy382 (Post 397358)
no, you have to convert this degenerated blocks into quarter o-grids (y-grids).

blocking/edit block/ convert block type and then choose y-block and mark all degenerated blocks (the blocks with 6 vertices)

ok, I get it !!!!

thank you

PS: with just O-grid I think the result is not bad either but maybe less "elegant" :-)

yorelchr December 13, 2012 11:57

2 Attachment(s)
Quote:

Originally Posted by yorelchr (Post 397360)
ok, I get it !!!!

thank you

PS: with just O-grid I think the result is not bad either but maybe less "elegant" :-)


here is the mesh generated with the O-Grid.


All times are GMT -4. The time now is 17:42.