[Pedro24]

Hello ,

I have a problem with the VOF part of fluent 6.3.

I want to model a taylor flow (slow flow) into a horizontal mini-channel ( width = 2mm, no gravity).

In the initial condition, the tube is full-filled with oil and i drain from one side. During the simulation, the tube is filled from the other side by air.

So my boundary conditions are :
Input = Pressure Inlet
Output = Velocity inlet (but negative because it's a drain)
One side = Wall adhesion
Other side = Symmetry

I study at the end of the simulation the liquid dynamic film let on the surface tube. A theory (Bretherton 1961) is able to give explicitly the thickness of this film under special conditions (capillarity number low, no gravity).

The problem is following :

If i use a mesh not well refined, the results are wrong because there is not enough cells into the film (less than 2).
But when i refined the mesh (with irregular mesh or regular) i have another problem. The thickness of the film become dependant with the mesh. If i change the width or the placement of my cells, the thickness will be modified. I noted that the interface between the two phases tends to be close to the nodes of my mesh.
I can't have interface located between two nodes.
I precise that my convergence conditions are good and that i tried all the schemes proposed (Geo-R, QUICK, CICSAM....). CICSAM seems to be the better one because oil and air have a big viscosity difference.
I use another VOF-code which doesn't have the problem (and do not use these schemes but simpler ones) but i try to understand why the schemes used by Fluent can't locate the interface between two nodes (in the dynamic film of course, because in the front, Fluent has no problem to locate the film at the intermediate between two nodes).

Sorry for the long text,

If you have an idea ?

Pedro