Using Multiple Information Sources to Construct Stochastic Databases to Quantify Uncertainty in Certification Maneuvers

Understanding how a new aircraft configuration will perform during certification maneuvers is important during the conceptual design phase. Most current methodologies require a complete aerodynamic database be constructed without a way to determine how the uncertainty in the database affects the resulting maneuvers. Building upon previous work in combining multiple information sources to build a stochastic aerodynamic database for the configuration of interest, this paper contains a methodology to estimate what location in the flight envelope has the dominant impact on the uncertainty in the quantity of interest. This is done by estimating the aerodynamic characteristics at a set of Latin Hypercube Sampling locations. The maneuver is then run multiple times using deterministic samples of the stochastic database incorporating each of the LHS locations separately. Fitting a Gaussian Process between the vehicle conditions (angle of attack (α), Mach number, control surface deflection, etc.) and the quantity of interest (a measure of the uncertainty in the maneuver), we found the best results could reduce the 90% Confidence Interval of the time to descend in the emergency descent case by approximately 70% when adding 5 new high fidelity points.