Hello Everyone

I am modeling an oil pan with 2 fluids using VOF + GeoReconstruct + URelax=1 + Courant=0.25 + PISO + SkewCorrect with 1 iteration + Body Weighted DS + 1st order for everything. I have tried to reduce the time step and underrelaxation factors for fractions and pressure and momentum but I still find that my continuity diverges rapidly especially when the solution goes onto the next time step (first time period converged). The main fluid is set to be oil and the secondary one is air. I also tried reducing UR for k, e, momentum and Pressure without success.

The upper fluid is air and the fluid at the bottom is engine oil. By applying 3 acceleration components x, y, z, simulating the move of the vehicle around a curved road with radial and axial accelerations, the two fluids are set to "move" around the oil pan (which is just closed and has no inlets/outlets). The highest acceleration is 2.2g and the time step is now 0.00001 sec. I don't think that I can lower it any further!!!

Any ideas anyone? Anyone knows a majic setup that works??!! Thanks in advance.

Al

Hi,

the vof model is quite hard to converge.

Try a simplier reconstruction model (the explicit or implicit one). They are more diffusive so your interface will be less sharp but they are more stable.

I hope this could help you

Best regards

Alain

In addition to Alain's suggestion,

Did you give a sufficient grid number to your system? especially around the air/oil interface

Sincerely Yours

Thanks guys for your suggestions. My mesh size has 1.77M cells in it (v. large for a transient case) with a maximum angle skew of 0.8 (good mesh). I did use the explicit reconstruct method at one stage and again it diverged rapidly. However, I will see if the implicit one would do but in any case, all these treatments are still VOF ones and you can only get to them by switching VOF on which means that I still have to use VOF.... I will also try to adapt the mesh across the fractions interface using gradient adaption and hope for the best.

The strange thing is, I did ask other people whom have done similar problems and they are telling me that they had to abandon Fluent for Star-CD when solving for such a problem and again it was also modeled as a moving mesh rather than that with accelerations.

I have been a keen user of Fluent across all types of problems but this one has certainly been the toughest one to converge and to defy all my attempts!

Thanks to you all.

Al

Hi Al,

I've found with VOF problems that starting with an "easier" method (ie not geo recon. which I'm not sure I've ever got working) and setting the UR for volume fraction to around 0.5 helps alot, although your problem sounds fairly tricky!

Have you checked the y+ values? Have you also looked at the solution after that first converged timestep? Maybe something weird is going on. Are the initial patched conditions correct? (Have you checked by plotting contours or something similar?)

Trac.

Another trick you could use is lowering the time step to 1E-08 or lower, and then increase every 100 timesteps or so. Seldom have I run solution with a time step greater than 5E-06.

Hi Al:

I saw a presentation at a conference recently involving the use of Fluent for a fuel tank filling application. I don't recall the specific details but I do remember that it took over a month to converge!

At the same conference I saw a very impressive "dam break" animation by AEA Technology. Apparently they have implemented a new VOF capability into CFX-5, which is supposed to be released later this year. They claim that it converges reasonably well within a few days in 2nd order. You might want to call AEA and check it out.

Good luck.

Mic