[danylo]

Hello everyone!

I am trying to simulate the freezing of water, using a customized version of rhoPimpleDyMFoam on OpenFOAM 5. Now I am running into a problem right from the start, when trying to simulate the density based heat convection.

The cooling comes from the side wall on the right-hand side and the top patch is a surface open to the atmosphere. Now the usual boundary conditions for free surface:
p - totalPressure 1e5
U - pressureInletOutletVelocity (0 0 0)
result in a weird verlocity field, where the highest velocity is in the top corner on the cold side. I think the reason for this might be, that OpenFOAM treats this patch as an In-/Outlet where the sum of velocitys results in (0 0 0). Meaning incoming flow on one side directly results in outgoing flow on the other side, which does not seem right to me. The further the simulation run goes, the more this flow in the corner develops and disrupts my desired convection flow.

I managed to achieve the convection I am aiming for, using a pressure reference cell somewhere in the top layer (1e5) and following BC´s:
p: zero Gradient
U: slip
This runs well in the beginning (see screenshots), but eventually results in a rise in total pressure in the system, followed by velocity instability due to the unchanged pressure of the reference cell.

Any idea what boundary conditions I could use to solve these problems? Or if there is any possibility of working with a reference pressure, while simultaneously limiting the system pressure?

Thank you all

P.S.
BC´s for the other walls:
p - zeroGradient
U - noSlip
frontAndBack are empty

[danylo]

Found the reason for this weird behaviour.

It was due to how Openfoam calculates the the gravitational effect of having different densities. Instead of increasing the pressure at the bottom, it lowered the pressure at the top so a gradient with the top boundary was created.

Set my upper boundary on the y-axis and now it runs fine.