In this work the flow regime within a generic turbine cooling system is investigated numerically. The main objective is to validate the performance of various turbulence models with different complexity by comparing the numerical results with experimental data. To maximize surface heat transfer rates, present-day cooling systems of high pressure turbines have highly complex shapes generating high turbulence levels and flow separations. These flow structures lead to higher requirements of CFD-techniques for sufficient prediction. To simulate complex flows in the industrial design process, Reynolds averaged Navier-Stokes (RANS) techniques are applied instead of computationally expensive LES and DNS simulations. Therefore, higher order turbulence models are necessary to predict flow field and heat transfer performance in such complex motion. The DLR standard flow solver for turbomachinery flows, TRACE, is used to solve the RANS equations. Four turbulence models have been analysed: the one equation model of Spalart and Allmaras, the two equation kω model of Wilcox, the two equation kω SST model of Menter and the anisotropy resolving Explicit Algebraic Reynolds Stress model (EARSM) of Hellsten. The investigated cooling geometry consists of a two-pass smooth channel with a 180 degree bend. At the DLR institute of propulsion technology PIV measurements in a rotating cooling channel test bed for Rotation numbers up to 0.1 have been performed. This work uses the experimental data for Re = 50,000 and Ro = 0 without rotation for comparison. For all models adiabatic and diabatic calculations have been performed. In order to accurately apply the turbulence models, a study concerning the turbulent boundary conditions has been performed prior to the calculations. The results obtained through RANS simulations are presented in comparison with the experiments along planes in the flow direction and in the orthogonal direction to study the velocity field, the shape and size of the separation bubbles and the wall shear stress. The EARSM predicts the flow field overall more accurately with improved agreement between all relevant parameters compared to the other models. The diabatic simulations reflect the adiabatic results. However, it can be noticed that higher complexity in turbulence modelling is related to increased heat transfer. Our work confirms the EARSMs ability to predict complex flow structures better than the more elementary approaches.

This content is only available via PDF.
You do not currently have access to this content.