Abstract

In turbomachinery CFD simulations, Reynolds-Averaged Navier-Stokes (RANS) based
approaches (steady and time accurate) have a tendency to under predict the radial mixing in the hub
and shroud regions where the flow field is affected by secondary flow features. RANS predictions
typically yield more coherent features, and hence more pronounced variations than experimental
data. This poor prediction of radial mixing has a deleterious impact on performance predictions.
The hybrid RANS/LES methods (i.e. various DES formulations), with RANS modeling of the flow
near the wall, and eddy-resolving modeling away from the wall, help to better represent the radial
mixing in turbine flows. Previous work with DES modeling, applied to a linear compressor cascade,
have demonstrated that the prediction of compressor endwall flows can be notably improved when
stochastic unsteadiness, outside of the boundary layer region, is resolved. In the present paper,
DES modeling is applied to predict radial profiles and other measured quantities from a two-stage
high-pressure-turbine rig. The DES-based radial profiles are shown to better match the
experimental data compared to RANS and time-accurate RANS results. The improvement is found mainly on the turbulence mixing prediction near the endwalls, i.e, hub and tip regions. The DDES prediction is useful for improved airfoil designs.