CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   aspect ratio (

Dougal McQueen August 5, 2004 10:36

aspect ratio
Hi, I am simulation atmospheric flow. To keep my yplus values down I need quite a small first cell height. However I want my domain top be as "long" as possible. Is there a prescribed or a recommended limit to the aspect ration?



Glenn Horrocks August 5, 2004 18:16

Re: aspect ratio
Hi Dougal,

It is problem dependant, but as a general guide hex and prism cells (as long as they are fairly square) can go to about 100:1 if you must, and normal tet cells should not really go past about 20:1.

Don't forget another very important guide is the volume ratio between adjacent cells. Aim for better than 1.5, but 2.0 is OK if you must. Much higher than that can cause problems.

Again, note this is all problem dependant. If your flow is aligned with the elements then the restrictions are less severe, if you have cross-element flows (including separations and the like) the restrictions are more severe. You will need to do a mesh independance check to confirm your mesh is OK.

On another matter, I am curious as to what you are modelling and how you are doing it. Are you modelling steady state or transient? I have always had problems getting convergence with atmospheric scale models, especially with steady state models. I think it is because the default k-e turbulence model is not averaging out the turbulence time scales properly so I was forced to use transient approach in a LES-like technique. What are your comments?

Glenn Horrocks

Dougal McQueen August 6, 2004 04:44

Re: aspect ratio
The physics of my problem are actually very simple. I am trying to model two dimension flow across a transition in surface roughness and heat flux (i.e. flow across the coastal discontinuity). This is for offshore wind farms. The main constraint on my domain is that it extends from the coast (upstream a little of the transition) out to the furthest extent of a particular wind farm (~10km). The other constraint is to keep the y+ values within a sensible range (i.e. less than about 1000, at present I'm getting about 600). I also have the problem of trying to obtain a static profile by developing a profile over a very long range (any comments on this would be useful). I have been using an aspect ratio of 25, I will try 100 and see what difference I get (of course this means a much longer domain for computing power so I'd like to puch this limit. I have been trying to use steady state simulations, these work fine for the problem without heat flux and boyancy considerations. As soon as I introduce a heat flux the results go a bit mental. I have tried to model a change in surface temperature and the results seem to be alright, although not a lot of change in the wind shear profile is observed. A change in heat flux works with transient simulation, the only problem is what time frame to use (presently I have been trying 10 steps of 6 minutes), as this has to be long enough for the boundary layer to adjust. Of course the physical system is not static for that long so there are other problems with that as well.

Let me know what you think, and just ask for more clarification if you need it.



Robin August 6, 2004 11:21

Re: aspect ratio
Near the wall, you can get away with aspect ratio's as high as 1000x, since the flow is well aligned. If you have convergence difficulties with these high aspect ratio's and cannot get around it by modifying your mesh, you can always run double precision. With double precision on, there are virtually no limits to aspect ratio.

Regards, Robin

All times are GMT -4. The time now is 18:51.