Development Of A Nonlinear Reduced-Order Model For Spatio-Temporal Urban Canyon Airwakes

Abstract

An improved nonlinear reduced-order model (ROM) has been developed for application in Advanced Air Mobility (AAM) flight within urban canyon turbulent airwakes. The new ROM outputs time-accurate three-dimensional (3D) flowfields from a fluid mask and a timestamp as input. The ROM accurately predicts the training dataset, concurrently providing accurate interpolated intermediates. This capability has been validated through comparisons with both time-averaged and instantaneous flowfields, and is consistent with the lattice-Boltzmann model (LBM) simulations serving as truth data as well as experimental flow visualization. Furthermore, a diffusion model is employed to reintegrate the missing fine-scale turbulence into the ROM, thereby enhancing the fidelity of the predictions.