CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   aspect ratio in a domain (http://www.cfd-online.com/Forums/main/92299-aspect-ratio-domain.html)

rakadit September 9, 2011 08:41

aspect ratio in a domain
 
Dear ALL!

Please clarify ,in case of variable meshing in a domain, what can be maximum acceptable aspect ratio ?what are other parameters to assess mesh quality in this case?

regards,

jonny_b October 26, 2011 12:32

I actually was discussing this with an ANSYS rep this past week and here are some target values to hit:

The max aspect ratio is dependent on the precision of the solver:
Single precision - max aspect ratio : 1000 or so
Double precision - max aspect ratio : 5000 - 10,000

Another metric to watch out for is expansion ratio - or the change in volume of adjacent cells. In my opinion I would try to keep the expansion ratio less than 20%

Finally the skewness angle is another important metric to watch. This measures essentially the orthogonality of the adjacent faces of a single cell. This metric is repersented differently depending on the mesher or solver calculating the value so you will need to consult the documentation for the software. Essentially you want two faces of a cell to be orthogonal (the angle between faces should be around 90 degrees) but this is typically very difficult to achieve for all cells in a mesh and the angle between adjacent faces can easily fall to much lower values. Try to keep the minimum skewness angle between faces larger than about 15 deg. Again take note of how the mesher /solver calculates skewness and if in doubt contact a rep for that software for more clarification.

Hope this helps and happy meshing!! It's always the hardest part of performing a CFD calculation and will give you many grey hairs over your career!!!

Far October 26, 2011 14:47

These values are good enough for the two equation eddy viscosity models. Requirements are much tighter for the Reynolds stress models ( should be less than 100)

jonny_b October 26, 2011 17:26

Yeah that's a good point, I'm sure the mesh quality requirements can be different depending on the types of turbulence models used.


All times are GMT -4. The time now is 18:31.