To estimate accident probabilities for commercial airlines, a sampling method called subset simulation is used at the Institute of Flight System Dynamics. Thereby, a physical model is utilized that considers a set of contributing factors and returns a so-called incident metric which describes the criticality of the particular flight with respect to a specific accident category. For accident probability estimations, multiple high-dimensional samples, i.e. virtual flights are generated and the physical model is applied for them. The goal of this paper is to integrate vine copula dependence models into the subset simulation implementation. These mathematical dependence models very flexibly describe high-dimensional and non-linear dependencies between the contributing factors. Eventually, the prevailing dependence structures of the contributing factors are recognized during the generation of the samples leading to more realistic results.     «

To estimate accident probabilities for commercial airlines, a sampling method called subset simulation is used at the Institute of Flight System Dynamics. Thereby, a physical model is utilized that considers a set of contributing factors and returns a so-called incident metric which describes the criticality of the particular flight with respect to a specific accident category. For accident probability estimations, multiple high-dimensional samples, i.e. virtual flights a...     »