CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] width in curve meshing (

sadjad.s June 25, 2013 04:55

width in curve meshing
Hi mates,
In this thread, Simon has given a tip of making boundary layer around 2d object.
He used width to make boundary layer.
But what is "width"?:confused:
I can only find "Tetra width", as seen in the picture.
And also why he didn't use Number of Layers to make that boundary layer?

Far June 25, 2013 08:47

sadjad.s: May be by setting tetra width simon was able to generate tetra mesh with some layers packed into boundary layer. he also said that prism in 2d failed due to the way geometry is assembled. Finally he shown the way to work with HEXA.


Originally Posted by PSYMN (Post 265193)
Ok, so I took a quick look. There are two things necessary for patch dependent mesh.

1) you must have curves around (and attached) to the perimeter of every surface.

2) you must have mesh size set on those curves.

The algorithm works by first creating line elements along the curves around each surface (we call these loops). If you delete a curve between two surfaces (topologically attached), it becomes dormant and the loops are joined into one larger loop. Ignore size works the same way (absorbs small loops into larger neighbors). After meshing the perimeter, it does a recursive loop algorithm to mesh the surfaces. Because all the loops share curves, the final mesh comes out already connected between all the surfaces, etc.

In your case, you had problems with 1 and 2. You had curves, but they were not shared between surfaces. By turning on color by count (right click display option under "curves" in the model tree), I could see that your seams were "yellow" and therefore not connected. So I used "Geometry (tab) => Repair Geometry => Build Diagnostic Topology" to tie those surfaces and curves together. Next, I saw that some curves had sizes set others did not. For the ones that didn't it tries to use a single element along those curves and just can't generate a good mesh. So I just set all the curve sizes to 0.01. That worked and I got my first mesh.

Next I wanted to improve the mesh a bit. For the radial curves, I setup biasing from 0.01 to 0.02 with a growth ratio of 1.2. That gave me this...
Attachment 3953

But I wanted some inflation... You can use ICEM CFD Prism with the "BLAYER2D" option, but with your surface patches like this, that failed for me. Rather than figure out how to get it to work, I just set the width on the "BLADE" curves to 2 and then regenerated to get this mesh. (I will send the tetin file).
Attachment 3954

But then I figured I should show you the hexa way. It took about 5 minutes. I will send you the replay file.
Attachment 3955

Then I smoothed it. I used 10 iterations of orthogonality followed by 10 iterations of Sorenson/THomas&Middlecoff (new in 13.0)

Attachment 3956

sadjad.s June 25, 2013 09:19

Hi mate,
I think i got the answer via try and error.
In 2d, there is no difference between "No. of Layers" and "Tetra width".
As you see in the first picture, i used No. of layers of 10 to make boundary layer. It did it successfully.

In the next try, i used Tetra width to make 8 layers:

And if you determine both parameters, mesher will ignore Tetra width for 2d cases:

Far June 25, 2013 10:38

Did you turn on blayer2d?

Far June 25, 2013 11:43


I tried both options

1. no of layers and

2. tetra width.

both have same effect, but no of layers overrule tetra width.

sadjad.s June 25, 2013 14:09


Originally Posted by Far (Post 435889)
Did you turn on blayer2d?

to whom may see this thread in future, there is no need to turn on Blayer2D option in prism.
Actually in 2d there is no need to include prism section.
All is done by curve mesh setup.

All times are GMT -4. The time now is 15:02.