When will Ansys introduce polyhedral meshing?
 

Hi,

I am using ICEM CFD for preprocessing. For complex geometries, more time is required to get good quality mesh and also number of elements are going high. But I came to know that polyhedral meshing can reduce the mesh count upto one fourth of tet elements with same accuracy in results. So there is any plan to implement the polyhedral meshes in on coming versions?

I know that the solvers (Fluent & CFX) are not that much good in handling polyhedral meshes. But what is Ansys future vision in this area?

I am not offensive. Just want to know....

Polyhedral Panacea
 

Ah, Polyhedral mesh... The panacea for all your meshing troubles...:rolleyes: Now keep in mind that I am here as a free agent and my opinions are my own and not necessarily the opinions of my company, actually, I don't know if the company even has an official position on Polyhedra, but never mind that...;)

Basically, Polyhedral mesh are the inverse of tetra mesh... Check out this website... http://www.cs.cornell.edu/home/chew/Delaunay.html for a interactive graphical experience.

Here is another shot from in CFX Prep/Post that should help illustrate.

Attachment 2702

I don't really believe they are any easier to generate than Tetrahedral mesh. However, without much in the way of quality metrics for them, it is harder to know when they are bad. You will have just as hard a time squeezing good polyhedrals into a tight gap as you would fitting tetra in there. And, in the same way that a bad tetra mesh can cause trouble with convergence or simply give an unreliable solution, so can polyhedra.

Companies that make it sound like the magic meshing tool are really just taking advantage of fact that you don't have a good way to evaluate the meshes. I really think it is more marketing than substance and am interested to hear what others on CFD online think about this... I know that some meshing companies require all concave polyhedra, so I am not sure how that additional requirement affects the ease of meshing.

I have heard of poly's being better than the inverse elements in certain locations where several poor elements could possibly combine to form a relatively decent poly... But you would get that same benefit by running CFX because it solves that way naturally.

Speaking of which, there are some solver advantages, particularly for strong swirl (due to more faces in more directions), etc. So the CFX solver has been a node centered solver for over a decade (long before it was cool:cool:), and the FLUENT solver was not limited by the number of element faces, so they had no problem adding a function to invert any mesh and run on the polyhedra (if that is what you would prefer).

In ICEM CFD, you can view the elements as polyhedra, but it is just visual. We don't have mesh editing tools to work on them (does anyone?) and we don't have a lot of good metrics for them (Mesh expansion is a polyhedral metric that was added at 12.0 for our CFX customers. keep in mind that a polyhedral mesh has a higher rate of volume change than its inverse tetrahedral mesh) and when you export the mesh, it must be as regular elements.

Here are some screen shots... Note that the inverse of tetras is a polyhedral bubble, but the inverse of hexas or prisms are still hexas or prisms.

Attachment 2699

Attachment 2700

Attachment 2701

Now I have heard that there maybe some memory advantages to generating polyhedra directly, and we are looking into doing that in ANSYS Meshing.

Hi Simon,

I can understand that this is open forum and your replies are great help to users. Really I learned many things from your replies. My intention is just want to know about polyhedral meshing.

Your reply clarifies that polyhedral meshing will not be available in upcoming ICEM CFD release.


Thanks for your kind reply.:)

Polyhedral Panacea
 

Right, sorry if my reply was a bit excessive ;) I tend to get a bit carried away about this topic.

We don't plan to add it for 13.0. (primarily because we don't think actually solves all meshing problems as advertised and would rather work on other things).

We probably will add it to ANSYS Meshing at some point though because it does do some things well.

Very interesting! Not excessive. More is better than less for learning purposes. :)

A little update please?
 

Hello jsm, PSYMN and Jade M,

I came accross this post looking for info on ICEM and polyhedral grids, but noticed that the last entry dates from 2010! :eek: What is the current state of this subject? Is ICEM now planning to include a polyhedral mesher in upcoming releases (i.e. 15.0)?

Thanks in advance,

MDB