Hello everybody,

I'm trying to simulate the backwashing of a rapid filter for water treatment. In the filter water is flowing upward in a cylinder and then over the edge of a funnel in the middle of that cylinder. The cylinder is open to the atmosphere. The velocity of the water at the bottom velocity inlet is about 0.01 m/s.

I'm using an axis-symmetric 2D-model, VoF geo-reconstruct and k-epsilon standard.

I get the simulation started by lowering the under-relaxation values. The resulting boundary between water and air corresponds to experimental data.

My problem is the flow path of the water. In an unsteady simulation, there is a huge eddy below the funnel and the path lines from the velocity inlet at the bottom of the model go up and then sideways at a right angle which is not the case in the experimental filter. When I switch to steady simulation, path lines go completely havoc. Those near the filter wall detach from the water phase and go right upward into the atmosphere and path lines from the atmosphere pressure inlet cross into the water-phase.

My questions: Has anybody done similar simulations or experienced likewise problems with the VoF-model? What possible changes of parameters could help?

Thanks, Andreas

Hi, I am trying to do the same thing as you. But for me the problem is that I don't know how to simulate the face open to the atmosphere. For the VOF model I think you may try to decrease the under relaxation factors to 0.2. Try also to use a body force weighted model for the pressure (solve_>controls_>solutions)vith PISO coupling and second order for momentum,... and so one.

I read it on the tutorial.

Please could you also send me your file because my case doessn't work too at marion.roux@enstimac.fr

Thank you very much

Marion

Thanks!

For the atmosphere I have a Pressure Inlet Boundary. Works well.

Using the body force model worked for the eddy, it's much smaller now. Only now, the pathlines near the walls go zigzag into the walls. I use the standard wall function.

Anybody with any ideas? Maybe another viscuosity model than k-epsilon? Or could refining the the grid help?

Andreas