[lasherwc]

I'm simulating a tank with a drain using interFoam. Water flows into the tank over lips at the edge at a known flowrate. The goal is to measure the height of the water and plot height vs. flowrate.

I built a very coarse unstructured mesh using Salome (Netgen) and got decent results, although the mesh is too coarse. There is an air core vortex that forms spontaneously, then shrinks and expands, just like in the experiment, and the water level is pretty close to what is measured. See pics attached.

When I refined the mesh the air core vortex eventually dissipated. When I meshed it using snappyHexMesh the air core vortex never formed. I have tried everything I can think of and can't get an air core vortex on a really good mesh.

I know that the flow must be "perturbed" to get the vortex to form based on experiments we did. I am assuming the really coarse Netgen mesh produces enough "wiggles" to kick the vortex off. I tried changing the div of U to linear hoping it would introduce wiggles and a vortex didn't form. I have played around with mesh density and convergence criteria and can't get the vortex on a good mesh. I tried several different turbulence models including kOmegaSST, kEpsilon, realizable, and LES. None of these gave me a vortex (I used kOmegaSST in the original model).

I added a tangential component to the incoming flow but that died off before it reached the drain. I created a model with an initial swirl and got an air core vortex to start, but it eventually died off. I put turning vanes in the tank which caused swirl but no dip in the free surface (ie, no air core).

Any thoughts on this? I believe it is real and important to simulate because it affects separation at the exit which determines the height of the water for a given flowrate.

[kishpishar]

Hi Bill,


This is a very interesting observation about the vortex formation simulation with tetrahedral meshes. Were you eventually able to obtain the free surface vortices in simulations on hexahedral meshes?


Thanks & regds
Kumar

[ves]

You have very bad mesh with high skewness. You must buid mesh with skewness 0.5-0.7. You may receive best result with cartesian cut-cell mesh or poly-hex core method (best mesh for free surface flow - cartesian cells) in Fluent meshing. If you send me geometry in Parasolid x_t I may build good mesh in Ansys fluent meshing and convert it to OpenFoam

[ves]

You must drawing pipe 10 diameters long fo corect implementation of boundary condition

[lasherwc]

I was able to get good swirl using a high-quality mesh. The system I was modeling is described in ANSI standard Z1034 2015. The model consisted of a square tank with 10 jets on each side. The jets produce a high amount of vorticity which allows swirl to develop.

I believe the original mesh (which I know is very low quality) produces vorticity due to jumps in mesh size and skewness. The inlet BC was a low speed "trickle" of water, as opposed to high-speed jets of the final model, so there was little vorticity inherent in the problem. When I made the mesh higher quality the vorticity introduced by the bad mesh went away.

What surprises me is how much vorticity is required to get the swirl. I presumed that the bad mesh was generating some instabilities and thought all I had to do was "tickle" the flow to get the vortex going. That is not the case - the amount of vorticity that has to be introduced into the system is quite high.

I've posted a bit more information on this on my website - see the last article on this page: https://cfdsolutionsllc.com/fascinating-findings

[kishpishar]

Thank you Bill! That was very informative.

[RicardoLB]

Hello,
I am doing some work on irrotational vortices in tanks with tangential inlets and a bottom drain in the centre, and at least in simpleFoam I manage to simulate the vortex with the realizable k-epsilon model (SST k-omega, for example, does not work at all). I am currently trying to get the same case running in interFoam to model the bending of the water surface, but I the irrotational vortex doesn't develop as it should. What turbulence model did you use?

Greetings,
Ricardo

[lasherwc]

Hi Ricardo:

I used the k-omega SST model. I did try an inlet flow with a tangential component at one point and was not able to get any swirl. I suspect that the tangential velocity was too low. You might want to try very small inlets with a high velocity. I think you need to generate significant shear, not just swirl, to get it to work.

Hope this helps.

Bill

[RicardoLB]

Hello Bill,
Thanks for your quick answer. I am comparing my results with experimental values, so I have to respect the inlet speeds. But you are right, if you reduce the area of the inlets the forced vortex in the outer sections has higher velocities. From my observations, yes, the tangential velocity helps to create the swirl, but the flow rate going through the outlet is ultimately responsible for the strength of the irrotational vortex. I just extended the outlet section, as Ves suggests, and the results look promising.

Greetings,
Ricardo

[lasherwc]

Good to hear!