[Jorge22]

I am trying to simulate the flow in a mixing tank with a rotating impeller. There is a free surface in the top, so I use a homogeneous model of water and air. I have done the tutorial of CFX that deals with a mixing tank and since my problem is similar, i don't think the physics is the problem. However since the geometry step is skipped in the tutorial, i think i might be missing something there. My geometry process is as follows:
-Draw the tank
-Draw the impeller (in a different geometry file)
-Open the tank geometry file and import the impeller.
-Translate and rotate the impeller into an appropiate location.
-Set both bodies as fluids.
-Continue meshing.

The residuals initially go down smoothly but then the solution diverges (see attached image)

http://postimg.org/image/bs2bs8m0b/

Things i have tried:

-Using other turbulence models besides k-e (as in the tutorial).
-Using lower or higher timesteps.
-Refining the mesh.
-Doing a transient run.
-Using a lower rotational speed (currently using 100 RPM).

Please, anyone has an idea what might be wrong?

[ghorrocks]

What happens if you keep running it? Sometimes there is a temporary increase in residuals as the flow establishes and it just needs time to work its way through.