CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] ~Howdy (

Nicolastheterminator September 4, 2012 10:44

I am new to ICEM CFD so forgive me if this is a relatively basic question. How do you know how many layers you need for a particular boundary layer? Do you use y+ or do you just simply guess and do it by trial and error?


diamondx September 4, 2012 11:48

i use y+. there is a y+ calculator in the website:
Here is the link :

Nicolastheterminator September 5, 2012 10:28

Hey diamondx,
Cool, thanks for directing me to the calculator. I am working on hexa-meshing now, and am about to tackle a complex geometry. How would you recommend me to tackle?
Is it possible to work on each part one by one? So for instance isolating each piece, blocking it and meshing it and then putting all the pieces together?
Or is it better simply to make one big block and carve a topology out of that?

Thanx for your reply,

stuart23 September 5, 2012 10:37

Hi Andris,

The benefit of structured mesh is that it is aligned and conformal to the geometry. By splitting your domain into separate pieces and meshing each individually, you will end up with non-conformal interfaces, thus negating the benefit structured mesh. I cannot think of any reason why it may be a better option to do this (though there probably are some).

Blocking is a skill and artform that is built up with practice. If you try the tutorials, and have a look at some of the older threads in this forum, you will see some great examples of how to block geometries, and hopefully find something that is similar to your project. If you still can't figure it out, post it up here and we can help you out.


Nicolastheterminator September 6, 2012 10:00

1 Attachment(s)
I've just done an unstructured mesh on my model, but I get these low quality elements at the the trailing edge, i adjusted the edge criterion to 0.001 but the problem is still there?I have read other posts about this but I don't quite understand how to get it done.
Many thanks, :D

stuart23 September 6, 2012 20:18

What meshing algorithm did you use? And have you tried smoothing yet?

Nicolastheterminator September 7, 2012 02:25

I used Octree, then deleted the quads, smoothed the tri with Laplace, and then meshed with Delaunay which i smoothed by freezing the tris without Laplace. For prism settings i used 0.25 ortho, 0.5 fillet, 180 angle, 3 layers, and initial height set to 0 and a height expansion ratio of 1.2.
The geometry is split into 3 parts, inlet, outlet and wall. No matter what way I mesh, whether hexa or unstructured, I get the same problem of low quality elements at this trailing edge.
Thanks for your help

Nicolastheterminator September 7, 2012 02:26

After which I smoother everything whilst freezing the prisms and the smoothed the prisms at 0.01. This does not make a difference, the low quality elements continue to stay there.

yonchong September 7, 2012 09:24

You might want to try adding a finer density to the trailing edge. You can either use the trailing edge curve or create a new one some distance off of the edge.

Then do

Mesh -> Create Mesh Density

Another way is to create a new part and put the trailing edge curve to the part. Then you can change the max size from the Mesh -> Part Mesh Setup.

Also this could be another time saving tip. Rather than creating a volume and shell mesh with the Octree and delete the volueme to have just the shells, try surface mech generation with Patch Independent option. It does exactly the same operation but you don't have to delete the volume mesh manually.

Nicolastheterminator September 7, 2012 09:51

2 Attachment(s)
I have tried to putting a mesh density at the trailing edge zone and I still have the same problem, I have attached 2 snapshots, one before prism, and one after the prisms. Also I get a much bigger mesh expansion ratio :(

Nicolastheterminator September 7, 2012 09:53

1 Attachment(s)
Here is another cut-plane with the prisms

yonchong September 7, 2012 10:26

Other people might disagree with me but I think that might be as good as you will get with the tet and prism.

If I were you I will try to run that with a CFD code and see what happens.

By the way, is there a particular reason why you want to run with a tet/prism mesh for this configuration as blocking seems to be suitable for this case?

diamondx September 7, 2012 15:38

yonchong is right, sometimes there is nothing more you can do expect hoping that your solver handles that.
about structured mesh, if you don't mind. We can help you get a descent structured mesh if you are not familiar with blocking in ICEM.

Andris September 8, 2012 14:13

3 Attachment(s)
Hi guys,
Well the reason for unstructured mesh was just to give it a go and see if I can get a good mesh quality.
Thank you for your advice, I tried giving blocking a go and here are my results, I cant seem to get the bifurcation quite right as I dont know which way to move the O-grid.
Many thanks for your help :D

raghav September 13, 2012 00:56

Reg. Geometry
Can u please share the geometry?


Andris September 13, 2012 05:11

1 Attachment(s)
I've zipped the model, the geometry format is in the form of an stl file.

PSYMN September 13, 2012 15:05

Search CFD online for posts about a "trailing edge surface"...

This is a way to improve quality by letting the prisms run back from the cusp rather than wrapping around it.

raghav September 14, 2012 05:36

ICEM - Hexa
1 Attachment(s)
PFA of the completed hexa mesh for your geometry. Quality is maintained at .32. and 8 elements are placed to capture the boundary layer. Randomly i have taken 8 elements. Hope this helps you :)

Andris September 14, 2012 05:56

Hi Raghav,
Many thanks for the document, it has helped me understand how to properly block the geometry. It was very kind of you.
Simon, is the idea to create a region at the trailing edge, and exclude it from the prism settings in part mesh setup?

PSYMN September 14, 2012 08:44

No, the idea is to create a trailing edge surface so the prisms can follow it out into the wake rather than try to wrap around the cusp... I give details and pics in other posts.

All times are GMT -4. The time now is 10:20.