Free surface vortex flow
 

Hi my name is Guillaume

I am using CFX 5.7.1 to model a gas entraining vortex above an intake pipe. So the phenomenon is free surface and presents a high swirling pattern. The geometry is a tank with a tangential injection.

When I simulate a high injection flow rate (therefore high circulation), I obtain the vortex. But when I use a small flow rate, I don't obtain any vortex, whereas the experiment shows a very strong gas entraining vortex.

I tried lots of things: mesh resolution, timestep, underelaxation, turbulence model (k-e, SSG), inhomogeneous/homogeneous. The free surface stays perfectly flat.

Any ideas? Any similar problems already experienced?

Cheers Guillaume

Re: Free surface vortex flow
 

Hi Guillaume,

It is most likely due to grid resolution. The solution requires very large gradients within a very small region. Keep refining the mesh around the core until you resolve the features.

You might also run it laminar, as the turbulence level would be very low, at least in the inlet section. Having turbulence on will introduce additional turbulent viscosity and further damp out your vortex.

Regards, Robin

Re: Free surface vortex flow
 

Hi,

I am not 100% sure of your physical model but from what I gather there is a liquid in the bottom and swirling gas in the top??

If this is the case then the model for the momentum (and possibly turbulence quantity) transfer at the interface is critical.

The free surface model assume a homogeneous field for the 2 phases but never occuring in the same space. This essentially "locks the two velocity fields as one" at the interface. This is for many applications not appropriate and you have to model the shear stress, transport and generation of turbulence, etc at the interface??

Think about it and let us know what the appropriate physics if before we worry about refining the mesh! ;-)

Regards...............Bak_Flow

Re: Free surface vortex flow
 

Hi,

The model is a swirling liquid at the bottom entraining a static gas. The flow is physically laminar. So I don't have any turbulence model.

I agree with you, I don't want to refine the mesh anyway, because I am already hitting the limit of my computer in terms of mesh refinement, and I surely don't want to run forever...

I agree with the fact that to capture smaller gradients, you have to refine your mesh. However I can't refine my mesh more, so I think about playing with the discretisation model.

Running it in inhomogeneous may not be that suitable. We already got some vortices with the homogeneous model for other geometries, but when we solved iot with the inhomogeneous model, the solver was really more unstable. Moreover, it takes a long time to solve the 2 velocity fields.

I was thinking of a timestep issue. The results change a lot when we change it. I am not quite sure of what timestep to employ. The auto timescale doesn't seem to help.

Cheers for your help guys, Guillaume