CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   Aligned edges in ICEM (

RichardG June 3, 2010 07:24

Aligned edges in ICEM
1 Attachment(s)
Iīd like to ask for help for a problem with ICEM:

When I activate the option "Projected Mesh Shape" for the view of the edges, the projected edges will be displayed. However at the borders of some splitted blocks the edges between two vertices are still straight, though they should be aligned to the curved geometry. In some cases this can result in meshing errors (cf. attached picture).

Unfortunately I couldnīt find out, why they remain straight. Apparently the option "Blocking - Associate - Associate Face to Surface - Link shapes" doesn't seem to work.

Therefore I want to ask: Can someone explain me, why these edges won't become aligned and how I can change this?

It would be nice, when someone could help me.

Thank you in advance,

PSYMN June 6, 2010 08:02

I assume you are referring to the automatic alignment? I am not sure why it wouldn't align, but with Hexa, you usually have options.

If you look under "edit Edges", you will find two options.

I would start with "link shape". It asks for a factor, this is because it maintains radius of curvature, so if the radius of the top curve was 1 and the radius of the bottom edge was about 1.2, then you would pic the top edge (usually associated to a curve or at least a surface) as your source, and set the "factor" to 1.2. Get it? Play with the factor a bit and you will.

You can also do it manually. Orient the screen so that the edge edge is the in the plane of the screen, then use the "split edge" options. You could split with linear splits or a spline split and control that edge to get the exact shape you want.

The faces usually just interpolate based on the edge shapes. If they have trouble because that is not enough, then you should try the link face shape thing... But that doesn't do much good if the edges are out of alignment.

RichardG June 7, 2010 08:08

Hello PSYMN,
Thank you very much for your answer and for your explanations concidering the "Link face"-option.

I think, the variant with "Edit edge - Link edge" is the best compromise. Therewith I can align the edges as I want. The disadvantage is, that I've to do it for every individual edge, which crosses another one. At a complex model it could be a lot of work.

I came to realize, that the "wrong" aligned edges will occur, when a block is splitted more than one time in the same direction. After the first split, the alignment is correct. But after the second split, the alignment of the edges seems to be inconsequential.

Best regards,

PSYMN June 7, 2010 08:55

I'll see if I can fix that for R13...

PSYMN June 8, 2010 17:27

5 Attachment(s)
Actually, I checked and I had already reported this and it may have already been taken care of in 13.0. I would like to make sure and test on your model if possible.

In the mean time, here are the two tests I just ran.

First, with no splits, the interpolation was fine... (picture not shown since I can only post 5 images)

Then I add a split and it is still good.
Attachment 3679

Then I add a second split and it is still good.
Attachment 3680

So then I tried to make a model that looked more like yours. After the first split, all was well...
Attachment 3681

So I tried the second, still good...
Attachment 3682

So I split it a bunch of times, no problem.
Attachment 3683

Hopefully this is all sorted out now.

Nenad July 27, 2010 07:23

1 Attachment(s)

I have problem to edit edges. I need edge to be concentric to curve, like on picture, but I can't do that with split edge, like it's shown on this video
What might be a problem here?


PSYMN July 27, 2010 08:25

Link Shape...
The image looks like it is using "Link Shape".

Link shape asks for a reference edge (usually one that is getting is shape by projecting to a curve or surface) and a factor.

It works by maintaining radius of curvature. I think there are several images in the help to explain how to use the factor, but basically, if the new edge was 20% further from the "center" of the curve than the reference edge, you would use a factor of 1.2 to increase the radius of curvature by 20% and keep things concentric.

If you use split edge, then I would suggest the spline option that lets you just eye ball it...

Nenad July 28, 2010 05:29

1 Attachment(s)
I can't figure what to do. I want to project selected edge to the selected curve, and then by using link edge, I will try to shape lower edges. I don't know how good it will look, but that's my only solution, at least for now. Still, I don't know how to match this edge with this curve. I tried by making 30 points on the curve, and then using option "linear" in split edge, but I'm not sure if that's good solution.

Nenad July 28, 2010 08:52

2 Attachment(s)
I can't manage to make constant distance between edges, while changing direction. What is the best way to do that?


PSYMN July 29, 2010 09:32

What is the problem?

Your pics look fine. What don't you like? Is it the surface projected edge on the top surface? Maybe show me the premesh of the problem area...

Nenad July 29, 2010 09:48

They look fine, because I didn't make that :) I am just wondering how to do exactly the same thing with edges

PSYMN July 29, 2010 10:24

The image you sent yesterday was done by selecting that edge and using "edge params" to set a number of nodes. It needs this number for the automatic edge split... Then the user went to Edit Edges => Split edge => Automatic Linear.

That is all.

Have you tried that?

Are you aware that the edge will bend to fit even if it doesn't look that way at first? (you may not even need to do the Automatic linear) Have you tried just generating the premesh to see if it is already fine? How about showing the edges curved instead of straight (right click on edges in the model tree and select the option for "Projected mesh Shape". Again, this will be sensitive to the edge params on each edge so use the "Premesh Params => Update Sizes" (or at least set the edge params manually) first.

Can you show your problem?

Nenad August 9, 2010 06:07

1 Attachment(s)
Thanks lot for those answers, they were useful. Pre-mesh image starts to look good, and I hope that mesh will be good also :) But now I have new problem with one block. Number of nodes on these edges is always the same. I need 5 nodes on vertical edges, and 25 on horizontal, but whenever I change number of nodes on vertical edge it automatically changes number on horizontal edge (and vice versa).

Thank You for help!

PSYMN August 9, 2010 13:48

Not sure how you did that but you must have manipulated the blocks (collapsed corners, etc.) until you have made those two parallel some how.

Turn on "copy to parallel" so the parallel edges appear as red, then follow it around to see how it connects...

If you had done regular top down blocking, this couldn't have happened.

Nenad August 17, 2010 10:41

I have made some changes in blocking, and resolved that problem. Now, the only problem (hopefully) I have left is pre-mesh quality. What is the best solution for these bad elements?

Thank You for help

PSYMN August 17, 2010 14:52

In most cases, bad quality is due to one of two things...

1) Less than optimal blocking (topology)

2) Poor edge distributions.

I think your problem is #1). I would guess that if you just imagined that each of those blocks was a single element, the quality would still be poor. You need a better way to fit the topology to the geometry in those regions. Unfortunately, that comes with practice and is hard to pass via text.

I would need a better look to properly diagnose, but from here it looks like...

A) Where your Ogrid passes from the pipe into the block, I think you have too much going on... You would be better off with a simple puncture. I think I had to fix that on someone elses model, but I can't recall whose (or I would include a link).

B) The other case is where you have tried to fit a block to a shape that is too sloped... I would need to look closer to suggest how to fix it.

Nenad August 18, 2010 13:45

I have made some changes in blocking, and fixed that bad angle elements. But the other section is still bad.

Nenad August 24, 2010 09:38

I've started from the beginning, and you were right when saying that I imagined those blocks as single elements. Now, I was looking only for topology of this model and made different blocking strategy. But I still have problems with the nozzle, cause every blocking I have tried here created bad elements. Everything else in this model is fine (angle>18, determinant >0.2)

Ahmed Saeed Mansour August 8, 2016 10:04

1 Attachment(s)
Any ideas for this impeller?

skmaperu March 11, 2017 14:18

icem edge adjusments
1 Attachment(s)
Hallo everyone,

can some one suggest me how to get equal spacing and concentric edges in that case ?

Thanks in Advance

All times are GMT -4. The time now is 08:14.