CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ICEM] Problem with Prisms (http://www.cfd-online.com/Forums/ansys-meshing/106568-problem-prisms.html)

Andris September 1, 2012 13:00

Problem with Prisms
 
Hi all,
I am doing a tetra/prism mesh. I have a high quality tetra volume mesh using Delaunay with no elements under a 0.2 quality and a mesh expansion ratio under 20.
Yet when I insert the prism layer it goes preety crazy, mesh expansion ratio jumps to 100 and stays very high regardless of smoothing. I have fiddled around with the prism properties but have to no avail found the problem.
Could someone please help?
Regards,
Andris

stuart23 September 2, 2012 11:59

Andris,

For all CFD purposes, it is most efficient to have very high aspect ratio prisms. If you think about the velocity gradients, the biggest one is in the direction normal to the wall, therefore this is where you want to have your smallest Δx (cell edge length).

If you also have large velocity gradients parallel to the wall, you must capture these using surface sizings before you inflate the prisms.

Andris September 3, 2012 04:47

Hi Stuart,
I don't quite understand what you are saying. I understand that the velocity change is highest near the wall due to the no-slip condition and that is why I am putting prisms with the tetra-hedral mesh. And I understand that they should have a good aspect ratio as the computation of these volumes greatly influences the result.
My methodology was to have a tetra-mesh of very good quality before inserting the prisms.
Are you suggesting that I smooth this tera mesh even further to get a better prism structure by decreasing the cell edge length when I do Laplacian smoothing?
Many Thanks for your help,
Regards,
Andris

stuart23 September 3, 2012 06:30

Andris, near the surface, the velocity gradient is in one direction only, normal to the wall, so this is the only direction that you need lots of resolution. Therefore having high aspect ratio prisms is fine. I have attached a (generic) picture of what I think is a good boundary layer mesh.

Source:http://www.smr.ch/local/doc/virtual_aircraft/d2d/user_manual/ar01s03.xhtml

http://www.smr.ch/local/doc/virtual_...-naca-comb.png

siw September 3, 2012 06:56

stuart23, just out of interest would you not think from those pictures that the difference in area of adjacent cells is too large at the trailing edge point between the last quad and first tri elements? I've always been lead to keep an epansion rate of 1.2 or less.

Andris September 3, 2012 06:57

Dear Stuart,
How do I get these high aspect ratio prisms? All I'm doing is selecting the layers, choosing an exponential expansion and floating the inital height.
Many Thanks,
Regards,
Andris

stuart23 September 3, 2012 07:10

@siw: i tend to agree.

It depends on the Re of the simulation to the size of the boundary layer, however I think that the last layer of prisms are too big, and it would have been better to switch to tets sooner. Meshes with a smooth transition between prism and tet are more accurate, and I think that the last layer of prisms are too big for the first layer of tets. It's not practical to solve the boundary layer like this and then have such small tets. The only reason I could maybe think of this is if you are using a hybrid LES model, and therefore need to model eddies away from walls, but this still looks wrong!

Having said that, it is still a good example of high aspect ratio prisms, just try not to look at the tets!!!!!

Stu

stuart23 September 3, 2012 07:13

@Andris, when you inflate your prisms, you can specify them however you want (first layer height / total height / growth rate etc etc), then once you want to smooth, freeze the prisms so they are not smoothed.

Stu

Andris September 3, 2012 07:45

Hi Stuart, here are photos of what I have done so far, i' think i am more or less happy with it, but i would appreciate your feedback
Many Thanks,
Andris

http://www.blog.co.uk/media/photo/1st/6583675#content
http://www.blog.co.uk/media/photo/2nd/6583676


All times are GMT -4. The time now is 23:53.