CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Hex and Tet (http://www.cfd-online.com/Forums/cfx/22491-hex-tet.html)

hagupta April 25, 2006 08:48

Hex and Tet
 
Hello all,

For a given simulation, does the CPU time depend only on the number of nodes. Does the number of elements have no role to play?

For a geometry, lets say there are 1) 10K nodes, 90K elements for tet meshing 2) 10K nodes, 10K elements in hex meshing.

Since the number of nodes is the same, will the CFX solver take the same time for solving?


jemteo April 26, 2006 18:42

Re: Hex and Tet
 
I'm not sure for 10K nodes, 90K elements for tet meshing. but comparing tet and hex, hex are mathematically cheaper. Tet will tend to have extreme geometrical disortions which might make it more difficult to converge. Hex is also favoured becuase it does not have the false 'anisotropy' that tet has. i came across a website that showed the effects tet had on flow velocity as compared to hex. these are effects due to element usage and not the physical problem itself.

someone please correct me if necessary


Rui April 27, 2006 05:27

Re: Hex and Tet
 
Hi,

This may interest you: ANSYS CFX Validation Report, Comparison of accuracy for various element types: laminar flow ( http://www-waterloo.ansys.com/commun...orts.asp?id=13 )

hagupta April 27, 2006 08:50

Re: Hex and Tet
 
Thanks for the replies.

So is it ok if I say that for the same number of nodes, as hex gives me a better quality, the hex will converge faster although the time per iteration will remain the same ?

Rui, could you paste that comparison report at the CFX community site here. Our institution has not yet registered at the community site, so I do not have access to community support.

Terry April 27, 2006 10:06

Re: Hex and Tet
 
Hi,all

For a strong recirculation zone, can we still get significant benefits from hex type mesh?

Regards!

Terry

Rui April 27, 2006 12:42

Re: Hex and Tet
 
I'm sorry, but I don't think I'm allowed to post it. And I strongly advice you to register at the CFX Community web site (if you have CFX it's free :) ), even if since Ansys acquired CFX the site has become worst and worst, it may be helpful.

About the time per iteration being the same for tet and hex meshes, I'm sure there are a lot o people here who know much more about this than me, but I think as for tet meshes you'll have a higher number of elements and therefore also a higher number of integration points, the number equations and unknows and the size of the matrix will be same, but the equations will be more complex and you will have less zeros on the matrix, which will lead to a greater computational effort

Take also a look at this topic: http://www.cfd-online.com/Forum/cfx_....cgi/read/6514

hagupta April 27, 2006 13:43

Re: Hex and Tet
 
Thanks Rui

So, except for the efforts in generating a hex mesh, hex meshes score over tets in all respects, this is what I have gathered from everyone.

Regarding the CFX community registration, I am trying to get it done.

Robin April 27, 2006 16:35

Re: Hex and Tet
 
CFX does not use tetrahedral control volumes, so what you read regarding tet's would not apply. The elements are tetrahedral, but the control volume is build up around the node by adding together the sector of each element connected to the node. The actual control volume is therefore a polyhedral element.

Regarding run-time and memmory usage, tet's tend to be more memory intensive due to the number of integration connections to neighboring elements. The element count will influence the assembly time, since assembly is done on a per element basis. Anyway, generally Hexehedral and prismatic elements are more efficient than tet.

That said, if it takes you a week to get a good hex mesh vs. an afternoon for an unstructured mesh, you may as well run the unstructured mesh and save yourself the man hours. I generally recommend using hexes where they are easy to add. In the unstructured bodies, inflate the mesh as far as you can to take advantage of the prisms (CFX-Mesh does this very nicely). You can end up with a larger mesh with prisms going a long way into the interior, which will run faster than a similar mesh with more tets and fewer nodes!

Regards, Robin

TB April 27, 2006 21:03

Re: Hex and Tet
 
If you have dirty geometry, I'll recommend ICEM HEXA as I always find it easier to clean the geometry in ICEM (I may be wrong as I seldom use CFX-mesher after it has been merged into workbench). If tet or poly (according to Robin) is required, I will use CFX mesher and avoid ICEM Tet tool. ICEM Tet tool failed me a couple of times when I tried to create inflation layers on a complicated cad geometry.


All times are GMT -4. The time now is 11:24.