Jet engine design relies on statistically averaged turbulence models. Known as RANS, these models constitute the most accurate, computationally affordable approach available, but have important limitations, particularly in the area of very complex geometries and off-design conditions. Industry would therefore like to complement their simulation tools with more accurate approaches such as large-eddy simulations (LES), which compute the largest turbulent flow features directly and models the smallest ones. Low-order finite-volume methods, typically tailored for RANS, continue to be used to solve for geometrical complexity. But in academia, LES is typically performed using much more accurate, yet geometrically inflexible, discretization methods.
Recently, unstructured, discontinuous, high-order methods have emerged that bridge the gap between academic accuracy and industrial geometrical complexity, thereby providing a much better suited discretization for industrial LES. This project represents the first use of such a high-order unstructured method for the LES of an actual machine, a transonic high-pressure multistage axial compressor. The computational approach used for these simulations is the discontinuous Galerkin flow solver Argo.
The first objective is to vastly improve the resolution and accuracy with respect to previous RANS and LES studies, providing an unprecedented and reliable insight into flow, particularly features such as leakage flows and stall, which are notoriously hard to capture. This new approach will allow industry to reliably predict jet engine performance, before constructing costly prototypes. Furthermore, this project will provide important information for future uptake of LES in industry and constitute a milestone and reference in their development.