Large Eddy Simulation of an Experimental Turbine Stage: Importance of Spatio-Temporal Resolution Requirements for Aerodynamic Performance Prediction

The National Experimental Turbine (NExT) program is a government-industry-university collaborative effort involving the design, development, manufacturing, and testing of a single stage research turbine [1]. The primary goal of the effort is to develop a modern research turbine for U.S. manufacturers and institutions that provides a platform for acquiring detailed data to be used for new design method development and new design concept validation. The present work relates to CFD modeling of this 1-stage turbine.

In 2023, wall-resolved LES remains challenging from a compute resource requirement standpoint for turbomachinery systems where chord Reynolds numbers in excess of 10⁷ arise. Recent work [2] suggests that meshes and discretization accuracies less than conventionally deployed for LES - indeed RANS-like - can provide suitable accuracy for bulk performance parameters including stagnation pressure losses, loading and off-design operation. To explore this, we developed a sequence of meshes starting from RANS-like, (nominally 2x10⁸ cells for the 3:2 vane:blade rotor-stator analysis). The meshes were refined locally based on the ratio of resolved to unresolved turbulence kinetic energy estimates, and surface and wake spectra. Through comparisons to available loading and loss profiles, and RANS simulations, we assess the quantitative accuracy of comparatively coarse LES modeling and make recommendations for practitioners who wish to pursue LES in these systems using conventional numerics.