A Multifidelity Approach for Uncertainty Propagation in Flight Dynamics

Abstract

This paper is concerned with the estimation of failure probabilities and statistical moments in a flight dynamics context with multifidelity Monte Carlo techniques. A propulsion-based emergency autoland scenario for the ATTAS research aircraft is considered, where stochastic wind disturbances complicate the landing. All statistical quantities of interest are inferred from samples of the wind shear and the corresponding aircraft touchdown parameters. The underlying Monte Carlo methodology ensures un-biased and reliable estimation of statistical quantities of interest. The advantages and limitations of using black-box surrogates as low-fidelity models are revealed through a series of numerical examples. In particular, significant computational gains can be achieved in some cases, however, advanced surrogate modeling may be required for some solution parameters.