CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Flow in a cavity (http://www.cfd-online.com/Forums/cfx/19590-flow-cavity.html)

 Herry April 25, 2003 06:41

Flow in a cavity

HI ! does anyone knows how to model a laminar flow in a cavity for 2Dimension. my tutor say i have to put the 'z' walls as a symetric wall/boundary. i am using CFX 5.5.1. could you one of you give me the hint?

The other think i tries to model this flow but however, i do not get the secondary vortex, only a main vortex.

anyway, thanks for your help whoever are you. i am just a new comer in fluid dynamics.

 Glenn Horrocks April 27, 2003 18:30

Re: Flow in a cavity

Hi Herry,

Laminar flow in a 2D cavity is a good validation case to get you started in CFD. The laminar flow case is fairly easy to get a good solution for low Reynolds numbers (less than 1000), and a bit trickier for laminar flow above this Reynolds numbers. I used these cases as assignments in a CFD course at university I used to tutor. Getting good answers in the turbulent regime is quite a lot harder, and requires a high level of understanding of what's going on - if you're still at school I think getting a good answer in the laminar case should definitely be worth 10/10!

Anyway, typical things to look for in the laminar case to improve accuracy include:

1) Is the problem properly posed - dimensions correct? Flow driving velocity correct? Correct density and viscosity? Are you using the correct physics? If laminar then this is pretty simple.

2) Is it properly converged - all residuals to less than about 1E-4, and balances to better than 0.1%?

3) Is your mesh fine enough? Say run a 10x10 mesh, then a 20x20 mesh and if your solution does not change your solution is "mesh independant". If it does change run a 40x40 mesh, and keep refining the mesh until you do achieve mesh independance.

4) Are you running an accurate enough discretisation scheme? You are probably using the default first order scheme (Upwind). The "High Resolution" scheme is more accurate, but can be more difficult to converge. I assume you are not doing a transient simulation so I will not discuss these issues.

Anyway, I hope that helps.

Regards, Glenn

 All times are GMT -4. The time now is 10:53.