Abstract
The highly-loaded, linear low-pressure turbine cascade T106C is a popular validation test case for direct numerical and large-eddy simulation solvers. In most previous high-fidelity simulations, the cascade’s geometry has been simplified by considering only a fraction of the spanwise extent. A periodic boundary condition in the spanwise direction is then typically used to analyze the flow in the mid-span region. However, it has proved challenging to numerically reproduce experimental data of the blade loading at engine-relevant conditions of isentropic exit Reynolds number of 80,000 and isentropic exit Mach number of 0.65 in the absence of a turbulent inflow. The discrepancies between high-fidelity simulations and experiments have mostly been attributed to an insufficient spanwise domain extent in the numerical setup. The present paper resolves these differences by including spanwise end-walls in the numerical setup using a dedicated high-fidelity computational fluid dynamics solver. Both the blade loading and the wake loss are thus adequately predicted by the present numerical study, which had not previously been achieved in the literature. Furthermore, a series of highly-resolved large-eddy simulations have been performed proving that spanwise periodicity cannot capture the underlying flow physics determining the blade’s mid-span performance at the considered operating point. For these cases, grid resolution and spanwise domain extent were systematically varied to highlight their combined effect on the blade loading along the suction side and wake loss prediction. Finally, a rigorous comparison between different flow solvers and numerical setups summarizes best-practice guidelines for conducting high-quality, scale-resolving simulations of highly-loaded low-pressure turbine blades.
