[indy07cz]

Hello guys,
I have an issue with deciding which turbulent model is appropriate for my case. I'm solving a real case which is T-junction for turbines situated in dam. Aim of the simulation is determine minor head loss coefficient (K) for both two legs of junction. Case is fully turbulent with high Reynolds number (3e5-3e6). The K coefficient is obtained from pressure drops behind the T-junction. I made several simulations to calculate friction coeff. so I'm able to separate losses from friction to get minor coeff.

But problems occurs when I try to compare results from k-epsilon and k-omegaSST model. Both models converge fine watching pressures in legs (but k-omegaSST is more transient), but pressure drop from k-epsilon is about 30% higher than from SST just for the T2 leg (see the picture), T1 leg is equal. Also in T2 leg the flow for k-epsilon is still swirling and for SST is developed and calm. I'm a self-teacher a bit in these so I can't decide which one is more accurate. I don't even have any measurements in-situ.
Mesh is quite fine, y+ about 50 for both models. I think that in this case is not necessary to solve viscous layer, but maybe you will have another opinion.
I would be very grateful for some kind of information particulary for this kind of flow.
Thank you