I want to know whether it's possible to simulate 3D free surface problems with CFX. Thank you!

Alan

Yes.

Yes. I have done quite a few cases of weirs, piles, FS along boat hulls etc. Some CFX papers have been presented at conferences, so you should check their website.

Hi Holidays! Can you tell me where to find such papers on free surface? What is the website?

Hi Holidays, have you ever tried the full 2 phase approach when modelling the free surface ? I've tried the homogeneous model, and that worked well, but have never managed to get the non homogeneous model to work. I didn't spend a long time trying, just playing with time step sizes etc. Have you tried this and is there a trick to getting this model to work ? Bob

Good point Bob!

Our experience is the same: the 2 or even 3 phase model is pretty much useless for free surface flow. The model is simply not able to create stable solutions. And this is to bad as many interesting problems cannot be solved with the homogeneous model (e.g free buble columns). I think we in total have spendt ½ year trying, but have decided to abandone this approach and find another way.

Regards

Michael

Damn, I was hoping that your response was "... this is the way to do it !!" What did you try ? I played with timesteps, but due to time restrictions not much else (oh and differencing scheme). Niether had any effect. I even tried interpolating the results from the homogeneous model (not sure if that is a sound approach - didn't work anyway !). Did you speak to the support desk about this ? My problem was water past a bluff body. The homogeneous approach was fine, but I wanted to model this as non homogeneous then move onto more complex problems, possibly even look at a breaking wave if possible ? Bob

I'm sorry to disappoint you!

I think we have covered proberly all the combinations of timestep, grid resolution, itterations, differecing scheme, turbulences models available in 4.2 - 4.4 and 5.5. without succes.

I had hoped - like you - that it was because that we had not found the right combination that it did not work. However, sooner og later the simulation breaks down due to instabilities. I will not rule out that we are to stupid and should not be allowed to work with CFD. On the other hand, the homogeneous model works fine at almost any setup - so what is the reason?

We can start a simulation and run it for a period, but at a certain point the velocities grows exponetial in a region causing some very urealistic flow fields.

My personal opinion is that the instabilities originates from the air above the water surface. It seems that it is very diffucult to get the 2 momentum equations, interphase momentum transfer and the overall mass balance to converge at the same time for this area. Small variations in the volume fraction of water above the surface generates huge air velocities which again moves the water around.

It should be said that a 2-phase simulation with a degassing boundary works fine and is fairly stable - its is only when we have a situations where we have a large "air pocket" above the watersurface, that it fails.

A strange experience was that when we made a small cylindercal tank with a free surface and with an air inlet in the bottom, this worked fine. However when we scaled the geometry up but keept the grid size constant - the model became unstable.

I once did get it to work in 4.3, when I simulated waves over a hydrofoil. However I needed timessteps in the order of 1/1000 second (and 30 itterations so far as I remember). So it was not very practical.

Also I think that the justification of the homogeneous model in CFX is that it is so damned difficult to get the "true" 2-phase model to work for free surface flow. There is an free surface algorithm integrated into the homogeneous model which sharpens the interface. I would guess that this is undesirable feature when you want to simulate the air-entrainment into breaking waves.

We have also studied wave propagation around marine strutures comparing the homogeneous model in 4.4 and 5.5.1, where the 4.4 version seems to dampen out the waves, the 5.5.1 version works much better (but we are not finished evaluating the model yet).

Regards

Michael