CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (https://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ICEM] suggestion for blocking strategy and meshing a corner (https://www.cfd-online.com/Forums/ansys-meshing/125794-suggestion-blocking-strategy-meshing-corner.html)

Daniel_Khazaei October 31, 2013 07:44

suggestion for blocking strategy and meshing a corner
 
3 Attachment(s)
Hello

I am trying to mesh a converging channel with a slender obstacle inside it. (you can see the geometry in the attachment)

I have already used split command to split the initial block as you can also see in the attachment.

This way, I have managed to get an initial mesh which is not so bad, but I can see irregularities in triangular block which is shown with a star in the picture.

My question is how can I mesh the triangular block in a better way? Is something wrong with my blocking strategy?


thanks in advance

regards

khoopes October 31, 2013 11:43

You could try a Y-Block for that area. Search the ICEM help for y-block and it shows the details.

Daniel_Khazaei November 1, 2013 05:41

2 Attachment(s)
Quote:

Originally Posted by khoopes (Post 460046)
You could try a Y-Block for that area. Search the ICEM help for y-block and it shows the details.

I have tried your suggestion (attachment)

using Y-Block has made an unnecessary node density in a region which you can see in the attachment.

furthermore, I have tried this mesh with fluent and the solution was diverged. (I could get convergence with my old mesh)

Am I doing something wrong here?

thanks in advance

khoopes November 1, 2013 09:26

The elements there do look like they are of pretty poor quality. I think you could move the vertex at the center of the Y to straighten them up. I would use the ICEM mesh metrics.

Why don't you want to use your first mesh? How does the simulation compare to experiment? Are you planning on adding near wall refinement later?

Daniel_Khazaei November 1, 2013 09:42

Quote:

Originally Posted by khoopes (Post 460148)
The elements there do look like they are of pretty poor quality. I think you could move the vertex at the center of the Y to straighten them up. I would use the ICEM mesh metrics.

Why don't you want to use your first mesh? How does the simulation compare to experiment? Are you planning on adding near wall refinement later?

Actually near wall refinement is the main reason which leads me to change my older mesh.

This geometry is fluid part of fluid-solid interaction problem, so I do not have a separate results to check for.

The paper I am working on it uses a time step of 0.1 sec that leads me to think that they have used a coarse mesh. It's a benchmark case to validate a coupling algorithm.

can you tell me how can I add wall refinement or is there any tutorial for that?

thank in advance

khoopes November 1, 2013 10:13

The hex mesh will be really easy to add near wall refinement, you will just change the edge spacing, making sure to propagate your changes to all parallel edges. Tutorial 1 step 6, and Tutorial 2 "Refine the Mesh using Edge Parameters" go over it. These section numbers correspond to the ICEM 14.5.7 help.

Daniel_Khazaei November 2, 2013 11:30

1 Attachment(s)
Quote:

Originally Posted by khoopes (Post 460154)
The hex mesh will be really easy to add near wall refinement, you will just change the edge spacing, making sure to propagate your changes to all parallel edges. Tutorial 1 step 6, and Tutorial 2 "Refine the Mesh using Edge Parameters" go over it. These section numbers correspond to the ICEM 14.5.7 help.

I am trying to produce a mesh like the one in the attachment, is it possible in ICEM?

cesarcg November 4, 2013 15:54

Sure it is possible using the Y-Block and adjusting the vertices of the blocking as convenient, as suggested previously by @khoopes. However, that strategy will impose a very dense distribution in some undesired areas. In my opinion, I think it is possible to mesh the geometry without the need of that Y-Block.

Regards.

diamondx November 5, 2013 11:33

you need a y-grid. yo cannot do that in 2d, so create your blocking, extrude 3d, convert that block into a y-grid, and convert the hole blocking to 2d again ! i hope i was clear !

Daniel_Khazaei November 6, 2013 09:04

Quote:

Originally Posted by diamondx (Post 460681)
you need a y-grid. yo cannot do that in 2d, so create your blocking, extrude 3d, convert that block into a y-grid, and convert the hole blocking to 2d again ! i hope i was clear !

Thanks for the tip, but would you please tell me is there any difference between Y-block and y-grid?


regards

diamondx November 6, 2013 09:56

my mistake, I wanted to say y-block !

Daniel_Khazaei November 6, 2013 10:11

Quote:

Originally Posted by diamondx (Post 460851)
my mistake, I wanted to say y-block !

I have tried Y-block method as khoopes suggested, but it will produce an unnecessary mesh density in the regions near the Y shape. ( as you can see in the 3rd post )

diamondx November 6, 2013 14:14

have you tried something just like this :

https://dl.dropboxusercontent.com/u/...suggestion.png

Daniel_Khazaei November 6, 2013 15:42

Thanks for the deep recommendation man ;)

I will try that and come back to you.

Daniel_Khazaei November 7, 2013 06:55

1 Attachment(s)
Quote:

Originally Posted by diamondx (Post 460886)
have you tried something just like this :

I tried to mesh like the picture you have provided, but when I test the mesh in fluent the solution does not converge. (huge oscillation in residual)

I have attached the geometry here, would please tell me how do you deal with geometries like this?

Far November 7, 2013 09:39

1 Attachment(s)
try this...and let me know if it is ok or not

Daniel_Khazaei November 7, 2013 12:32

1 Attachment(s)
Quote:

Originally Posted by Far (Post 461021)
try this...and let me know if it is ok or not

Can you tell me how did you create a boundary layer mesh around the block?

I only need that for bottom walls and I have done a few merged command to get a mesh like the one in the attachment. ( I done it on your files )

However I have a few problems on the merged edges. After the merging process I have negative quality, and fluent reports 0 orthogonal quality.

micro11sl November 7, 2013 18:43

5 Attachment(s)
Hi,
I just find I have done something similar before, see my attachment, extracting from one of my report.

Regards
Sheng

Daniel_Khazaei November 7, 2013 19:08

Quote:

Originally Posted by micro11sl (Post 461090)
Hi,
I just find I have done something similar before, see my attachment, extracting from one of my report.

Regards
Sheng

would please tell me how did you managed to get the final mesh? (first image)

Far November 7, 2013 22:51

Steps are:

1. Make the initial block

2. Make one horizontal spilt

3. Make 4 vertical splits. One is at the end of ramp and other few units before it.

4. Delete the block in lower right corner

5. merge vertices at the end of ramp and now you have wedge block

6. Go to y-block command apply on it

7. Make associations

8. Again select all blocks, faces on two sides of your geometry (symmetry) and other faces where you don't want O-grid

9. Smooth mesh in wedge block, but be careful using parameters for smoothing.

micro11sl November 8, 2013 06:52

Quote:

Originally Posted by Far (Post 461102)
Steps are:

1. Make the initial block

2. Make one horizontal spilt

3. Make 4 vertical splits. One is at the end of ramp and other few units before it.

4. Delete the block in lower right corner

5. merge vertices at the end of ramp and now you have wedge block

6. Go to y-block command apply on it

7. Make associations

8. Again select all blocks, faces on two sides of your geometry (symmetry) and other faces where you don't want O-grid

9. Smooth mesh in wedge block, but be careful using parameters for smoothing.

If you use 2D blocking, Y Block from the panel cannot be used. It is tricky to do it. You need go to the Ogrid splitting, now selecting two edges to create a Y block. Remember, when you select no edge, you get an Ogrid. When you select one edge, you get C grid. When you select two edges, you get Y grid. When you finish the Y block splitting, the raw blocking may have redundant parts, so use merge block etc to make it nice.:)

Daniel_Khazaei November 10, 2013 06:57

3 Attachment(s)
Quote:

Originally Posted by Far (Post 461102)
Steps are:

1. Make the initial block

2. Make one horizontal spilt

3. Make 4 vertical splits. One is at the end of ramp and other few units before it.

4. Delete the block in lower right corner

5. merge vertices at the end of ramp and now you have wedge block

6. Go to y-block command apply on it

7. Make associations

8. Again select all blocks, faces on two sides of your geometry (symmetry) and other faces where you don't want O-grid

9. Smooth mesh in wedge block, but be careful using parameters for smoothing.

I have tried your method, but with out boundary layer part which I was not able to produce like the one you have provided.

The convergence is now smoother and the minimum quality is equal to 0.892.

but mesh density near the corner is a problem in my unsteady simulation.

I have to decrease the time step just because of that region!

Far November 10, 2013 08:10

this is almost similar to what i did. Only difference is that I made one extra split before Y-grid. Now you can apply ogrid to improve solution accuracy and convergence. Or you can make ogrid first and then merge vertices and Yblock.


why you are solving 3d problem?

Daniel_Khazaei November 10, 2013 08:39

Quote:

Originally Posted by Far (Post 461402)
this is almost similar to what i did. Only difference is that I made one extra split before Y-grid. Now you can apply ogrid to improve solution accuracy and convergence. Or you can make ogrid first and then merge vertices and Yblock.


why you are solving 3d problem?


It is not actually a 3D problem, I am using OpenFOAM and as you know it only works in 3D. So any geometry must have 3 dimensions. However if we only need to solve a problem in 2 dimensions, defining the planes perpendicular to the third direction as "empty" type, will do the job.

Daniel_Khazaei November 11, 2013 15:22

2 Attachment(s)
Quote:

Originally Posted by Far (Post 461402)
this is almost similar to what i did. Only difference is that I made one extra split before Y-grid. Now you can apply ogrid to improve solution accuracy and convergence. Or you can make ogrid first and then merge vertices and Yblock.


why you are solving 3d problem?

I have tried to add the boundary layer part to my previous mesh. Is there any suggestion to improve it even further?


thanks in advance

Far November 11, 2013 22:21

you did first O-grid and Y-grid? Seems so!!!. Now its perfect :D

Daniel_Khazaei November 12, 2013 05:26

Quote:

Originally Posted by Far (Post 461616)
you did first O-grid and Y-grid? Seems so!!!. Now its perfect :D

yes...

1- I have done vertical and horizontal splits as you suggested.

2- I have removed the block at lower right corner.

3- I have applied the O-grid.

4- a little playing with merge & move vertices to make it nice.

5- Then I have applied Y-block


many thanks for your help;)


All times are GMT -4. The time now is 06:51.