Separated Flow on Bluff Body

GPB

Hi, I'm comparing CFD predictions of flow over a cube against an equivalant set of measured data. I've tried standard k-e, Realz k-e, RNG and RSM models. For each I have tried Standard WF, Non-Equil WF and Enhanced Wall treatment. I am finding it impossible to accurately predict the separation bubble. Upstream the V profiles are o.k. but in most cases (except standard k-e which is too small)the separation bubble is prediced as too big, in some cases not even reattaching on the upper surface of the cube! This is a 3-D steady state calculation. I can't understand why the RSM is so inaccurate, and why FLUENT fails so badly in predicting such an underlying flow regime. I'm running out of time and ideas. Anyone got any thoughts/experience on this. Thanks.

Regards, G.