Influence of RANS Turbulent Inlet Set-Up on the Swirled Hot Streak Redistribution in a Simplified Nozzle Guide Vane Passage: Comparisons With Large-Eddy Simulations

Abstract

Abstract A high-pressure turbine is a complex component of a gas turbine from a thermo-mechanical point of view. In modern lean-burn combustion chambers, this complexity is enhanced with the presence of hot streaks together with swirl components and high levels of turbulence coming from the combustion chamber. These radial and azimuthal velocities and temperature distortions have a substantial impact either on the aerodynamics inside the high-pressure turbine or on the aero-thermal behavior of the vanes and blades. It is thus clear that the early development of a high-pressure turbine using numerical simulations must take into account swirled hot streaks. From a practical point of view, imposing the proper turbulence at the inlet boundary condition is not easy in classical Reynolds-Averaged Navier-Stokes (RANS) methods, where all the turbulence is modeled. This paper studies the redistribution of a swirled hot streak in the simplified case of a bent duct. This work focuses on turbulence modeling. High-fidelity Large-Eddy Simulation (LES) results are used as reference data to validate different RANS set-ups to predict the hot streak redistribution in terms of migration and diffusion. Results show that imposing the turbulent quantities from a LES causes an immediate destruction of the swirl components and a too-high total temperature diffusion in a RANS approach. It is found that the turbulent length scale, expressed in terms of μT/μ, plays a significant role in the aerodynamic and aero-thermal behavior of the flow. The optimal range for the value of μT/μ differs from what is encountered in the literature on a high-pressure turbine configuration. Also, imposing quantities at the inlet consistent with the LES or measurements does not improve the prediction of the trajectory of the swirl jet or the total temperature distribution. The anisotropy of turbulence is suspected to explain the failure of the usual RANS models.