Wall-Modeled Large Eddy Simulation of Full Annular Combustor Leveraging GPU Technology

Abstract As part of sustainability initiatives, gas turbine industry has been focusing more on fuel-efficient engines with a reduced carbon footprint. These initiatives present new challenges for simulation engineers due to limitations in computational hardware and conventional simulation workflows. Combustor design involves detailed physics, such as turbulent flow with a wide range of scales, spray particles, evaporation, etc. within complex geometries. To meet design timelines, assumptions like periodic sectors and reduced geometry are often used, which can affect accuracy due to the high computational cost of combustion design systems. The integration of Large Eddy Simulation (LES) with Graphics Processing Units (GPUs) is increasingly appealing to the engineering community due to its significant computational speed and accuracy advantages. This study aims to evaluate the combination of LES and GPU framework on gas turbine combustion simulation. The highly parallel top-down mesh technology, known as Rapid Octree (RO), is used to generate computational grids. Ansys Fluent Native GPU Technology combined with optimized LES numerics, flamelet generated manifold (FGM) model, and Lagrangian particle tracking is used. The simulation results are compared to available experimental data of Energy Efficient Engine (EEE) and previous studies, with the focus on both accuracy and speed. Additionally, the effects of the combustor coupled with the first vane are investigated under Sea Level Takeoff (SLTO) conditions. In overall, good agreement between simulation results and experimental measurements are observed for both single sector and full annular combustor cases, with total computational time ranging from 8–15 hours in different computational scenarios based on the used hardware.