Bad pressure distribution in pump
 

Hello everyone,
I use CFD to calculate radial force in cetrifugal pump. I made geometry of old "test stand" used to calculate radial force. I have data from scientific research (capacity, static pressure in inlet and outlet). I want to confront them with CFD results. My problem is I get bad results, which are not connected with scientific data, eg. force decreases with capacity lower than in BEP (0,5*Qn), so it's not physical. It looks that I get bad pressure distribution.
see the pictures of model:
http://http://www.picshot.pl/public/view/full/301733http://www.picshot.pl/public/view/full/301733
http://www.picshot.pl/pfiles/301734/volute.jpg
Blue colour are interfaces, red are rotating walls in stationary domain, purple - LE and TE. Volute geometry is very simple-similar to rectangle.
When I print forces in Fluent when residuals are about 2e-3 I get eg 200N, but when my residuals are about 1e-3 the force can be twice lower. It looks that pressure distribution is not stable(?), but residuals are quite stable. Usually I can't get converge lower than 1,5e-3 for second order upwind. I used k-e, k-e realizable, k-w SST but it not helped. In inlet I use BC mass flow inlet and in outlet outflow. I tried also use pressure BC in outlet and inlet, but I didn't get capacity which I should get.
I've meshed impeller domain in TurboGrid. Minimum orthogonal quality is 0,15. I made volute grid in ICEM (hexa) and it has min quality 0,45. So mesh is acceptable. Hub and inlet_and_shroud domain were made in Ansys Meshing. I tried 2 grids, bigger has about 4,5 mln nodes and it not helped.
I don't know what am I supposed to do?
Are my interfaces in bad position (see the pictures)? Should I place them in other place or way? Impeller diameter is 268 trailing edge diameter is 264 and tongue diameter in volute is about 290.
Why I get pressure distribution completety not connected with scientific data.
I'd be very grateful for help because I don't know what should I do....

BW,
Marek

I think that concerning the BC it is better for inlet Velocity or mass (the same) and for outer Pressure.

Thanks for advice. I've moved interface of impeller farther into volute and now it looks better even with low quality mesh