[Edy85]

I'm a student of Nuclear Engineering and I'm going to graduate in the next months. My BIG problem is to complete my degree thesis.

I'm using the CFX (ANSYS v12.1) to simulate a 2D fluid flow in a cyrcuit like this:

vertical channel (the water enters in the domain and goes up) --> horizontal channel --> vertical channel (the water goes down and exit from the domain)
In the middle of the horizontal channel there is a hole which "looks up" in another little vertical channel where there is air at ambient pressure and so a free surface between water and air appears.


I'm interesting in calculating only the velocity profile of the water in the circuit.

Boundary conditions:
inflow: bulk mass flow rate 0.1 kg/s
outflow: bulk mass flow rate 0.1 kg/s
ambient: Opening, 0 Pa (relative pressure) <-- it refers to the hole where air comes in

I use a structured mesh which works good with the single phase problem (only water and boundary condition at the top imposed as wall instead of opening).

I've imposed as initial conditions: a stationary water in the circuit and a volume fraction of air equal to 1 in the little "up channel"; a initial pressure due to the water column; initial velocities components equal to 0.

I have adopted the standard free surface model.

PROBLEM: the simulation doesn't converge. The residuals are around 10^-3.

My questions:
1) do I have to use the dishomogeneous multiphase model or the homogeneous one? When the situation will become stationary i'll have a "clean" free surface but during the simulation there seem to be significant splashes! If i use the dishomogeneous one, do i have to set air as "disperse fluid"?

2) I've tried different time steps, also during the same simulation but nothing good.

3) i accept all the suggestions!

THANKS!

Ettore

[ghorrocks]

Convergence FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Also, you might not need a multiphase model for this. Can you replace the free surface with a pressure opening?

[Edy85]

Thank you very much for the answer.

I've tried with a coarser mesh and i've reached the convergence (almost 10^-6). But in the post processor i can't see a well defined free surface.

Anyway now i'm trying with a finer mesh starting from the solution of the coarser mesh. I'm seeing that the convergence is not as good as the last one, but maybe the solution is ok.

Is it possible to create a model with a single phase (water) but with a non completely filled domain? I mean: now I have to simulate with a starting level of water lower than before (now the "box" is not full of water but it's filled only for a half) and i expect a steady state condition with a free surface not horizontal. (I think in the transitory there will be some splashes).

How can i model this situation? Multiphase? Disomogeneous? Which model of turbolence? (my maximum velocities are less than 1 m/s!)

THANK YOU

[ghorrocks]

You would have to explain what you are doing better. Does the free surface move? What are you trying to get out of the analysis?

[Edy85]

I start from a certain level of water with an horizontal free surface with the air. Than at t=0 I have a water mass flow at the inlet and the same mass flow at the outlet. So the level of water, that was horizontal and not moving, starts to move and when the steady state is reached I expect a free surface which is sloped. This is a consequence of the geometry of my circuit: vertical channel (the water goes up) --> horizontal channel (the water goes from the left to the right) --> vertical channel (the water goes down and exits from the circuit).

My aim is to get the field velocity and the shape of the free surface from the analysis.

I don't know which model I should use.

Thanks

[ghorrocks]

If the free surface is coherent (ie not foamy) then use the homogenous with free surface. The tutorial example free surface flow over a bump shows how to set up free surface simulations.