| description abstract | This study is carried out in the context of hot streak flows in high-pressure turbines, for which a correct prediction of the temperature evolution is required. The present work particularly focuses on the turbulence anisotropy analysis of a swirled hot streak flow in a bent channel representative of a nozzle guide vane (NGV) passage of a high-pressure turbine. Large-eddy simulations are conducted with the in-house solver IC3 in order to measure and characterize the anisotropy of turbulence. Moreover, to evaluate turbulence modeling, steady simulations of the bent channel are performed with the elsA software, which solves the Reynolds-averaged Navier–Stokes (RANS) equations. LES is first used to complete a turbulent kinetic energy (TKE) budget that enables to understand the energetic transfers associated with turbulence. This budget reveals two distinct zones where turbulence activity is impacted when the curvature is reached. The analysis of the anisotropy of turbulence based on two metrics highlights a misalignment of the Reynolds stress tensor and the mean strain-rate tensor (Schmitt's criterion), and a strong anisotropy developing inside the bent duct (Lumley's analysis) that may cause the failure of the classical RANS turbulence models based on Boussinesq's hypothesis. To check this hypothesis, RANS is positioned against LES with different turbulence models that accounts or not for the anisotropy of turbulence. Both turbulence activity (TKE budgets, Lumley's analysis) and aerothermal fields (radial distributions) are compared. Results show that Explicit Algebraic Reynolds Turbulence Models (EARSM) enable to better account for the anisotropy of turbulence, which in turn promote a better prediction of temperature, both in terms of intensity and trajectory. | |
