[RanSoffer]

I'm simulating flow above a wavy wall, and I'm trying to find a RANS model that can accurately predict the separation and reattachment. When comparing RNG, realizable k-epsilon, and SST models, I've noticed they all have issues, especially SST, in predicting the velocity around the separation zone and the locations of separation and reattachment.

I'm considering adjusting the empirical constants in these models to see if I can achieve better results. However, I haven't found any literature discussing this approach in depth, nor have I come across detailed explanations on which constants could be adjusted to control the shear layer size, separation points, or reattachment locations.

Does anyone have experience with this? Or could someone point me to papers or books that delve deeply into tuning these constants in turbulence models?