[Far]

In next try created blocking for next set of circles. And also created buffer zone between them to uncouple the mesh distribution. But do not know why ICEM is not respecting the buffer zone and is forcing the unwanted mesh distribution.

[despaired student]

Hi,
I do the tutorials you posted but it will surely take some time. The model looks good - I'll try to do one myself. I hope you are right and it's no magic . If questions/problems appear, well I hope there won't be many...

[Far]

I guess we should play with index control to make the appropriate edges parallel.

[energy382]

I didn't have the time to work out the details of the blocking (e.g. edge params, move verticies.....), but you can see my blocking in the attached pics (there may be even a better and easier way to create a blocking for this topology ---> time!!)

min. quality ~0,4 and min determinante 3x3 ~0,6 should be a good starting point to do some additional work on that blocking like moving nodes and smoothing (i prefer 10x ortho, 10x laplace and again 10x ortho, but that's your choice).

[despaired student]

Hi,

@Far: The problem with the buffer-zone is due to the unequal distribution of nodes. On the outer side the node-distance under the circles is smaller than on the inner side (third of your pictures -> left side) because there is no circle oppositional to this one. So there are a lot of nodes on the outer side that need oppositional nodes on the inner side. They automatically choose the nearest ones on the inner side. This normalises only when on the inner side a bunch of nodes appear (third picture -> right side).
I think that is the reason for the odd-looking buffer-zone...?

@ Far: How did you create these round edges (picture) ?

I still have a problem with my 2D-Meshing. It is not possible to create a new 2D-Block by "Create Block>Create Block" without deleting the old Blocks? I have do copy the block, right?

My first blocking-try looked more or less similar to energy's try but I would like to comprehend how you, Far, created your Blocking.

[Far]

I am trying to make a recording and then will upload on youtube. Wait one day.

[Far]

The problem with the buffer-zone is due to the unequal distribution of nodes.

I know this, but don't know how to control this in ICEM. I have created these type of grids in GAMBIT and Gridgen without any difficulty, but having difficulties in controlling these aspects in top down mesher (ICEM).

@ Far: How did you create these round edges (picture) ?

from blocing >>>>>>>edit edge >>>>> I use spline option (see Figure below)

I still have a problem with my 2D-Meshing. It is not possible to create a new 2D-Block by "Create Block>Create Block" without deleting the old Blocks? I have do copy the block, right?

I am creating one bigger block and then subtracting portion by portion to fit into the particular geometry part.

[Far]

I am trying to make a recording and then will upload on youtube. Wait one day.

Here it is http://www.youtube.com/watch?v=roZNU...ature=youtu.be

[PSYMN]

Behind the scenes, all splits must propagate across mapped blocks... When you only split the visible blocking, we call those hidden splits "implied splits". Later, when you split across another index, it may simply expose those implied splits, or it may create a new split (with associated implied splits)... This becomes difficult to control.

For my own personal sanity, I would have split to capture the holes using full splits. Then I recycle as many splits as possible. A split across the block may be the left side of several holes, and then go thru between several holes and then be on the right side a the next ring... But at least all the splits are radial.

I am not clear what you mean by a buffer zone... You could create a buffer zone by deleting a block "permanently". If you don't delete permanently, then the mesh still carries across the structured VORFN blocks. Another way to break the propagation is to change the block type from structured to "Free"... The paved blocks do not force propagation across them.

[PSYMN]

Oh, you might just want "Linked Bunching". Look into the help to see exactly how to use this (found under Edge Params), but it will let you match distributions across indices where you have differing numbers of edges...

[Far]

Oh, you might just want "Linked Bunching".

I tried it, but it is not working. smaller edges on one side are directly linked to larger edges on the other side, therefore linked bunching is not working. It is also strange all edges have some parallel edges on other side, remaining edges (since for more circles I have more edges) have no corresponding edge on other side and therefore their meshing converge to single node on the buffer zone.

For my own personal sanity, I would have split to capture the holes using full splits. Then I recycle as many splits as possible. A split across the block may be the left side of several holes, and then go thru between several holes and then be on the right side a the next ring... But at least all the splits are radial.

Like this?

[despaired student]

Hi,

first of all, a thank-you for your help, energy, Far and PSYMN.
I did a quite O.K. - well I would say so - blocking myself. There are still some "minor" problems remaining and the way I chose isn't the most efficient one. I hope you know why some of the problems appeared, how to solve the and maybe you can give me some feedback about my blocking-strategy.

I encircled the critical regions in the attached picture - I tried to vary the number of nodes but in these regions it didn't work? And there is also a curiosity with the middle circle.


@Far: In your video you created 2 O-Grids. The outer one you used for your splines. Did you make this to express the boundary layer around the membranes?


Mistakes in the blocking-strategy:
1. For each ring I generated a new block from my initial 2D-Block with "Transform
Blocks >> Translate Block". Then I split the block in 15 blocks to fit in the
ring-segments.
For the next ring I repeated this process.
==> in this case the adjacent blocks in neighboring rings aren't connected, right?
They are only connected within the ring.
--> a better way is to use one block, split it vertical and horizontal to not lose the
connection across the rings. (like you did Far?)
2. Is it recommendable to draw the edges through the rings (pic-2) so that in every ring
will be the same amount of blocks or could for example three blocks end on one
block (pic-1)?


Kind regards

[Far]

good.

@Far: In your video you created 2 O-Grids. The outer one you used for your splines. Did you make this to express the boundary layer around the membranes

I had no idea about the important flow areas, so just created the one o-block to capture the boundary layer around the pipes. Interior O-block was created with the intention to improve the mesh quality as well as to capture the flow (if any).

a better way is to use one block, split it vertical and horizontal to not lose the connection across the rings. (like you did Far?)

Yes, otherwise you will have problems of disconnected blocks. As ICEM is top-down approach, so one should come from the bigger one which fits the whole geometry to be meshed.

Is it recommendable to draw the edges through the rings (pic-2) so that in every ring
will be the same amount of blocks or could for example three blocks end on one
block (pic-1)?

This should not be the problem if mesh is looking nice. I feel that you have some association problems in outer most blocks.

[Far]

Topology can be improved further (alteast 10 times ). Lets work on it.

[despaired student]

Topology-improvement at least 10 times...I expected more

[Far]

Ok, set the goal to 50 times and lets start the race.

[despaired student]

I'll try to complete an improved version tomorrow. Today I have only 10 minutes left till the computer-pool closes...

Thanks for the quick answer, Far!

[Far]

Here are some problem areas which can be rectified easily (further splits)



Once 2-D topology is perfect then we shall move towards 3-D blocking.

[despaired student]

right, the area at the right down corner should be solid.

[despaired student]

and I didn't use part mesh setup to set the number of nodes...so this maybe explains some differences in the regions.