[Santhosh91]

Hello all,


Hope somebody can help me out on this, I have been struggling for months.


I am performing 3D simulation using interFoam to model a two phase flow in a T-junction geometry. As you can see in the attached picture, emulsions are produced. I am trying to reach mesh independency with an uniform meshing (I set it up with blockMesh and it feels natural because of the geometry to limit numerical diffusion).


The issue is that when I am comparing the velocity profile inside one emulsion downstream between two consecutive meshes, it never matches or fall below the usual 5% to ensure ''mesh independency''. I tried mesh with about 3 millions, 5 millions cells but still not observe matching velocity profile especially near the emulsion border where there is the interface ...


I tried to use refinement near the walls but the solution is completely different, the jet penetrates inside the main channel which is not logical compare to what is observed in experiment. I feel like it is because of numerical diffusion. Because of that I kept the meshing uniform.


Has anyone encountered this type of problem ? Or anyone have any suggestions to make ?



Thanks a lot for your time,
Sincerely,
Santhosh

[LuckyTran]

You should have some wall refinement but walls are not what is limiting convergence for this type of problem. You won't achieve mesh independence in the velocity profile until your cells are much smaller than the thickness of the interface; and you have a long way to go until you're there. You can, however, converge with respect to the slug travel time.


I do recommend that you continue to refine the uniform sizing of the mesh, but do have some wall layers because those are always needed if you have no slip BCs.

[Santhosh91]

Hello Lucky,


Thanks a lot for the feedback.
My meshing is about 2micrometer in terms of cells length while the interface thickness is about 1-2micrometer so that makes sense. Do you think a factor 10 (meaning cell length of about 0.1/0.2 micrometer would be sufficient (with a usual Courant number for this type of problem of 0.25) ?



Also for the wall layers, is the only way to do it in openFoam with snappyHexmesh ?


Finally I am trying to get useful datas from these simulations, when I look at the color patterns of velocity and vorticity and the global streamlines, it looks alike, but the specific datas values does not fall under the 5% requirement. My emulsion length is matching too and probably the slug travel time but do you think there is a way to ''reasonably'' retrieve good datas about velocity, vorticity keeping this meshing ?


Sorry for all these questions and thanks again for the previous feedback. It is really helpful to orient me in the right direction.


Sincerely,
Santhosh