[AlmostSurelyRob]

Dear All,

I have recently encountered an issue whilst trying to model simple 2D case of a channel flow and I was wondering if you could give me some insights about it.

The minimal example illustrating my point is the following:
2D geometry
Left, Top and Bottom boundaries are Walls. Right boundary is a pressure boundary.
Gravity
Transient simulation
Initial condition Velocity = 0


I have tried three different initialisations for two phase flow:
1) Full air
2) Full water
3) 50/50 water/air stratified (air at the top).

I have run these cases in a number of codes (Fluent, openFoam, STAR-CD) and all of them give a bit strange results.

In case 1) and 2) I expect a no flow solution. Fluent and openFoam meet my expectations. STAR-CD results seem to be affected by a choice of reference density. It's probably because hydrostatic pressure at the outlet is calculated with the use of reference density.


With 3) the situation is less obvious to me.

All the codes give me some sort of backflow. It usually looks a little bit like a dam break problem but occurring on a pressure boundary. So either water seems to be 'sloshing down' or the air gets pushed away by a fluid moving from outside the domain.


My conclusion for now is that with two-phase flow the outlet conditions matter. So it generally depends on a particular application.

But is there any other way to model outlet? How would you implement/model the outlet conditions with two phase flow. I am aiming at modelling long channels and ideally I would like to minimize the effect of outlet. Is there any expedient way of doing it?


e.g.
STAR-CD offers a mean pressure boundary. So only the mean pressure is specified and the cell values are adjusted to fit it. This seems to be solving my problem, but there is still a small back-flow occurring in the above example.