The prediction of the flow in a gas turbine exhaust diffuser of a combined cycle power plant is particularly difficult as maximum performance is obtained with highly loaded diffusers, which operate close to boundary layer separation. CFD (computational fluid dynamics) simulations then need to cope with complex phenomena such as smooth wall separation, recirculation, reattachment, blockage and free shear layer mixing. Recent studies based on the RANS (Reynolds-Averaged-Navier-Stokes) approach demonstrate the challenge for two-equation turbulence models to predict separation and mixing of the flow correctly in such highly loaded diffusers and identify that more accurate methods are needed. Hence, the application of a hybrid Scale-Adaptive Simulation (SAS) is investigated and the CFD results are compared with experimental results from an in-house test rig.
In the present study the flow in a model exhaust diffuser (for heavy-duty gas turbine diffuser applications typical Reynolds number 1.5×106 and inlet Mach number 0.6) is examined with unsteady RANS (URANS) simulations with the SST (Shear Stress Transport) model as well as a hybrid Scale-Adaptive Simulation (SAS) model. The SAS model switches from URANS to a mode similar to a Large Eddy Simulation (LES) in unsteady flow regions to resolve various scales of detached eddies. The current study shows that with the SST model similar results are obtained with RANS and URANS simulations, whereas the more complex SAS model leads to a much better resolution of the unsteady fluctuations. However, the time-averaged results of the SAS calculations overpredict the blockage of the separation and hub wake. This results in an underprediction of the pressure recovery and the mixing of the flow compared to the simpler two-equation models and also compared to experimental results.