Is the adaptive mesh refinement suitable for interface tracking process?
 

Hello, Foamers,

recently, i use the interFoam to simulate the droplet impacting walls. First I use the axial simulation, which is 2d. Then I did a 3d simulation with adaptive mesh
refinement. the two results are different. It seems that AMR's results lag behind the 2d simulations. for reference, please see the attached.

I guess it is due to accumulated error the refinement casued. Thus, I was wondering If AMR is suitable for tracking interface in multiphase simulation? or is there method to improve this situation?

any hint is welcomed.

Hi,

• is your 2D simulation really similar to the 3D in terms of physics?
• Once I added the functionality for interface refinement to the library (https://www.youtube.com/watch?v=Zn2Qxb1DYzQ)
• Temporary link: https://2019.holzmann-cfd.com/en/ope...dlrefinefvmesh

Yes, I think the physics is tha same. the droplet impact on flat wall at moderate velocity. It should be axial sysmetric. So I guess that the axial sysmetric simulation shoud be the same as the 3d simulation. but the result is different. I don't know why.

A common misconception, maybe, using AMR is that it will work flawlessly out of the box. Often a bit of tuning -- of AMR parameter choices, numerical settings, etc -- is required. The problem at hand though seems to be a difference between 2d and 3d simulations. I think it is unwise to assume that a (axisymmetric) 2d simulation will automatically reproduce 3d results without showing this to be true; it is likely that the 2d case could be tuned to match the 3d one but, again, a match the first try seems to me unlikely.

So, thinking "out loud", I would think about it like this :

A 3d, static (high resolution), as the "truth" case.
A 3d, dynamic simulation with AMR to match 3d, static case (and show it matches static case)
A 2d, static/dynamic (axisymmetric) simulation to match 3d (static or amr).

In my experience it has been hard to convince people of the validity of AMR results, so such rigor may not be necessary.

As far as interface capturing goes, I have had great success in the past using isoAdvector (https://github.com/isoAdvector/isoAdvector) and also some moderate success with a coupled level-set VOF (CLSVOF) method (e.g. http://www.tfd.chalmers.se/~hani/kur...ankarMenon.pdf).

Caelan

Thank, in fact i found a paepr 'Investigation of effects of receding contact angle and energy conversion on numerical prediction of receding of the droplet impact onto hydrophilic and superhydrophilic surfaces'
In this paper, the author encountered the same situation ans he gave out an explanatoin that the difference between 2d and 3d simulation is due to parasitic current, which is caused by vof method itself.
Thank you for the suggestion. I will try the isoadvector and coupledlevelset method.

From my understanding, surface tension acts generally weaker in 2D than in 3D. This is related to the physical principles of the process itself. Either implementing surface tension based on local curvatures or in tangential cohesion, you will end up having a certain pulling force magnitude that arises from the superposition of curvatures on the three planes XY, XZ and YZ, respectively. Think about the natural evolution of torus of water... and now, what do you expect to obtain when simulating the 2D cross section, which is initially two separate circles ?