Large Eddy simulation and data-driven surrogate modeling of high-pressure cryogenic oxygen jets in crossflow

In this work, the mixing characteristics of a transcritical liquid oxygen jet injected into a vitiated crossflow are investigated at an operating pressure of 100 bar. A comprehensive parametric analysis is performed using Large Eddy Simulations (LES) coupled with tabulated thermochemistry to decouple the effects of the momentum flux ratio (J) and the jet Reynolds number (Rej). Results indicate that while J primarily governs global jet penetration, Rej significantly affects the near-field topology. At low Reynolds numbers, thermal expansion effects dominate over inertia, leading to a pronounced radial spreading of the dense core, whereas higher Reynolds numbers extend the liquid-like core length. To enable rapid design space exploration, the high-fidelity database is leveraged to train a resolution-invariant surrogate model based on Fourier Neural Operators (FNO). The proposed model accurately infers the time-averaged scalar field and jet trajectory for unseen operating conditions, capturing the non-linearities more effectively than classical empirical correlations. The surrogate demonstrates a speed-up of several orders of magnitude compared to the CFD solver, suggesting its potential as a tool for the preliminary design of high-pressure propulsion systems.