An Application of D-vine Regression for the Identification of Risky Flights in Runway Overrun

In aviation safety, runway overruns are of great importance because they are the most frequent type of landing accidents. Identification of factors which contribute to the occurrence of runway overruns can help mitigate the risk and prevent such accidents. Methods such as physics-based and statistical-based models were proposed in the past to estimate runway overrun probabilities. However, they are either costly or require experts' knowledge. We propose a statistical approach to quantify the risk probability of an aircraft to exceed a threshold at the speed of 80 knots given a set of influencing factors. This copula based D-vine regression approach is used because it allows for complex tail dependence and is computationally tractable. Data obtained from the Quick Access Recorder (QAR) for 711 flights are analyzed. We identify 41 flights with an estimated risk probability > 0.001 for a chosen threshold and rank the effects of each influencing factor for these flights. Also, the complex dependency patterns between some influencing factors for the 41 flights are shown to be non symmetric. The D-vine regression approach, compared to physics-based and statistical-based approaches, has an analytical solution, is not simulation based and can be used to estimate very small or large probabilities efficiently.      «

In aviation safety, runway overruns are of great importance because they are the most frequent type of landing accidents. Identification of factors which contribute to the occurrence of runway overruns can help mitigate the risk and prevent such accidents. Methods such as physics-based and statistical-based models were proposed in the past to estimate runway overrun probabilities. However, they are either costly or require experts' knowledge. We propose a statistical approach to quantify the ris...     »