Numerical Design of Experiments for Repeating Low-Pressure Turbine Stages Part I: Computational Opportunities and Methodology

Abstract

Abstract. The complex transitional and turbulent nature of unsteady flows seen in low-pressure turbines often demands high-order methods such as large eddy simulations (LES) for accurate predictions of turbine efficiency and loss generation. This study introduces a new framework that integrates high-fidelity simulations into unsteady Reynolds-averaged Navier–Stokes (URANS)-based design cycles, via use of cutting-edge numerical tools leveraging modern high-performance computing architectures. This multifidelity simulation framework consists of the capability to perform simultaneous LES and URANS simulation and a multifidelity reconstruction method to correct the lower-fidelity trendlines. Modern supercomputing hardware nodes typically connect multiple graphics processing units (GPUs) to multicore central processing units (CPUs). This node layout is taken advantage of as part of the novel multifidelity approach by using both the CPUs and GPUs concurrently, thereby increasing the utilization of modern supercomputing architectures. In this framework, LESs are executed on the GPUs, while the otherwise idling CPU cores are used for multiple URANS calculations providing high-fidelity and low-fidelity results concurrently. Ultimately, the low-cost trend predictions of URANS are combined with the accuracy of LES to create a multifidelity dataset spanning the entire Reynolds number regime of interest via an established multifidelity reconstruction method. The multifidelity reconstructions, validated against unseen LES data, prove to be superior to URANS at all operating conditions. This approach thus allows for highly accurate and fine-grained parametric sweeps requiring a minimal number of costly LESs.