[ThomasN]

Hello,

I know there are many discussions about this theme and it's problems but no approach about the contact line velocity.

The results of my two phase flow high dynamic simulation depend extremely on the value of the contact angle at the walls. So I want to use an approach for the dynamic contact angle. Everything works fine but I have to read the velocity of the moving contact line. (Currently I define a velocity u v w and the contact angle changes in response to the orientation of the contact line (normal of the free surface) and the velocity vector )

So there is the problem that a no slip boundary has no velocity. Therefore I want to use the velocity in a certain distance of the wall. Does anybody have an idea how to read the value of a variable with the coordinates x,y,z in CEL without defining a Monitor Point? Is there a way to implement something like that:

interface_vel=Oil.velocity (x, y, 0.001[m]) ?

Thanks!

[ghorrocks]

You have found the well known issue that a contact line at a wall boundary should not move (because of the no slip condition), but it clearly does. This paradox is glossed over in just about all CFD codes and just ignored - I know CFX and Fluent have nothing to deal with this and I strongly suspect Comsol, StarCD and the others do not either.

The result of this is that simulated contact lines do not converge with grid refinement. The results they give are vaguely correct, but if you trying to be accurate it is impossible.

I cannot see how your suggestion will improve things either. Using a fudge to avoid a fundamental limitation in Navier Stokes multiphase simulation methodology sounds unlikely to be successful to me.

Can you describe what you are simulating? It is unusual to have simulations where contact angle is important but the length scale is larger than millimetres.

My recommendation would be to move to a simulation software which does not have fundamental limitation on moving contact lines. Unfortunately I cannot name one - but you might have some luck with a lattice Boltzmann solver (which does not use the Navier Stokes equations).

[ThomasN]

Hello Glenn,

I'm not allowed to explain it in detail because it is not published for now. But I simulate a two phase flow between two rotating discs (rotating couette flow) where the gap between the discs is about 0,1mm and the rotating speed is 3000min^-1. The Domain size is 20mm. So the effects of the contact angle are significant.

I'm working on this problem with cfx for a long time and i have to solve it with this code (the results aren't that bad with a constant contact angle, I only want to improve them)
I know some approaches where they use the contact line velocity in a certain distance from the wall or the whole drop velocity.
Is it possible to use a velocity of a cell with specified coordinates for the calculation of another cell? So I could imagine that the cells at the wall use the velocity of the cells in a certain distance of the wall to determine the contact angle in response to the orientation of the contact line and the velocity vector.

Thomas

[ghorrocks]

I cannot think of an easy way to implement the velocity condition you request. It will probably have to be done in fortran and it does not sound straight forward at all.

So my warning is simply that you if your results are close, you are going to have problems getting much better than that. For a simple example, do a simulation of flow in a capillary tube where the flow is pulled by a meniscus working its way along the capillary. This is a good example as it has an analytical solution. You will find that as you refine the mesh your results initially get more accurate, but then get worse and do not converge on the analytical solution.

[ThomasN]

Ok thanks for the advices. I will try it. At the moment I experiment with a drop on a surface.

[ghorrocks]

If the free surface interface does not move then the inconsistency does not occur. So if your drop is stationary it will run just fine. You need a moving interface for the problem to show up.

[ThomasN]

The drop and the Interface are moving... the convergence is quite good.