Aircraft post-upset flight risk region prediction for aviation safety management

Flight trajectory prediction is a vital tool for enhancing the national airspace system (NAS) safety management, especially with the rapid increase in flight density. In-flight uncertainties during aircraft upset events significantly impair the flight path trajectory prediction, hence a robust uncertainty quantification study is needed for realistic flight path risk region construction upon such upset events. The NASA transport class model (TCM) is implemented as a high-fidelity flight dynamics simulator to mimic post-upset aircraft response. Take-off and high-altitude stall scenarios are considered, due to their major contribution to aircraft loss of control in-flight incidents, where stochasticity is introduced in the TCM upset-triggering parameters. Automated recovery algorithm is developed and applied into TCM framework to control the rate of elevator surface deflection and/or throttle level, leading to flight path nominal conditions recovery. Monte Carlo simulations are performed to estimate a stochastic risk region in terms of confidence ellipsoid for both non-recovery and recovery simulated cases, where a relatively larger uncertainty level is observed during the recovery process. Additionally, a data-driven deep-learning surrogate model is developed to enhance the computational feasibility of such risk region estimation, which is essential for in-situ NAS safety assessment. Finally, the wind speed effect on the risk region and flight dynamics response prediction during high-altitude post-upset recovery cases is investigated.