Why so different 'htc' results...?

azt

I have run a couple of simulations of a rotating heated disc with channels, using periodic boundary conditions and 'Tbulk for HTC=293K'...same mesh, same conditions, etc. In the treshold of laminar-to-turbulent conditions AROUND THE DISC I am using 2 equation model (k-epsilon) and SSG model. When I postprocess the results, the heat transfer coefs. on the 'free surfaces (upper and lower)' are close enough between the two models. However, the averaged heat transfer 'within' the channels is very different for each model, with k-epsilon model giving a high average 'htc' (40 W/m^2/K) and SSG a low average 'htc' (27 W/m^2/K)...why is so high this difference (I dare to ask...)

I am quite surprised and worried, since at this point I have no exact clue of what the difference's source is. I am in the hunt of some papers about the performance of these two models in heated rotating flows... but if anyone has any idea to share about this, please write.

The RMS errors for the 2 solutions are below 1e-4 for all components (in some cases 1e-6). The distribution of local 'htc' within the channel is also surprisingly different... although I am not interested in 'local' values... but in average values. (fyi:the channels contains some ribs).

Cheers,