Learning Forum

[Pollovr]

Hello,

I'm currently performing a classical sloshing problem: water and air inside a tank accelerated under resonance; that gives away a free surface that is exponentially growing. So far it runs perfectly when I insert a submerged vertical baffle in the middle of the tank, but when I set up the same tank but rotate the baffle 45º, the solution get's oddly unstable (check the video)

45.mp4

[YasserSelima]

Under relaxation factors should not affect your solution. If they do, the solution did not converge.nincreasing URFs helps the solution to converge more quickly, but there is a risk of divergence. I am currently have no optimum criteria for the URFs, but I do change them between iterations using a UDF. I make them small in the first couple of iterations and increase them when the residuals decrease. Not sure if this is the optimum or not, but this decreases my solution time slightly. nRegarding your case, As you are interested in the transient solution, here is my advice. Decrease the time step and increase the number of iterations in the time step ... make sure that the residuals converge ... they should be almost straight lines before moving to the next time step.n

[Pollovr]

Are you sure? I tough that since it was a Transient formulation with no iteration (NITA) it would because there is no convergence term, since it's not iterating. So far I've tried reducing the mesh size and the time-step, but it simply didn't work.nI'm just concerned about the fact that since I'm not iterating over a solution, URF would have an effect because it's a transient simulation.n

[YasserSelima]

No outer iterations, but there are inner iterations to solve the conservation equations. Check the theory guide.n

[YasserSelima]

Increase the number of iterations. The solution in your first video is not converged.n

[Pollovr]

How do I Increase the number of inner iterations?n

[YasserSelima]

I just checked, I have no control over it!!n

[YasserSelima]

 in order to preserve overall time accuracy, you do not really need to reduce the splitting error to zero, but only have to make it the same order as the truncation error ... So, it keeps iterating internally until the error in the same order of truncation error. Regardless of the the URF. n

[Rob]

Sounds like you need to reduce the time step, in transient that's the main control as opposed to URFs that are more useful in steady. Watch the flow field as it develops, what is different in the result in vertical v 45 degree baffle? Chances are something changes in the flow field to trigger the failure, this could be a wave-surface interaction that's happening more rapidly than the solver is set for. Note, in 2021R1 there are some more VOF stability tools so have a look in the update documentation. n

[Pollovr]

I have already changed the time step, I tried going from 5 (mm) elements to 2,5 (mm) and reducing by 2 the time steps from 0,1 (ms) to 0,05 (ms) but the same problem raised. The problem is that at the beginning the residuals for continuity skyrocket and the simulation is unstable from the beginning (as compared to the one with the baffle at 90º)nArrayWhat do you mean by the flow field? If you can confirm me that the solution doesn't get tamed or damped when using the URF it means that the free surface (the elevation for example, which is a parameter that I'm studding) is the same, it just helps the inner convergence, but the results are valid.nIdeally, I would be able to do the same without URF but if using them won't affect the final results it shouldn't be a problem.n

[Rob]

Try 0.025s and see what happens: you may find the flow passes through a cell more quickly than you realise. Staff aren't permitted to open attachments so please post a few screen shots. How's the mesh quality (cell quality and resolution) looking?n

[Pollovr]

This is what happens without a baffle (pretty standard):nWhen you add a 90º baffle the same oscillation happens, but with no magnification, the oscillation gets stabilized really fast:nWith baffles at angles of 30º,60º,120º or 150º same things happens, but less damped, it takes more time to stabilize the wave and the elevation is higher. But when I tried doing it for 45º:nIt happened the same thing for the 135º angle, the problem I believe was a great mass imbalance. The setting where the same for all of the simulations. I added a URF of 0.8 for the pressure and the 45º worked perfectly, but the 135º, with a value of 0.7 is still giving me a bit of a problem:nI feel it's almost there but in the centre of the free surface... I don't know if the simulations are gonna give the same results with different URF, that is why I asked if these values should affect the final results. n

[Pollovr]

When you say 0.025 you mean (ms) or (s)?n

[Rob]

Sorry, ms, I misread the units in your post. nThose results look sensible depending on what you set the left and right boundaries to. Those boundaries are also far too close to the baffle: you're going to artificially force the result with the boundary settings. n