Analysing the risks of an unmanned aircraft accident near airport using a Bayesian network

Abstract

As drones and unmanned aerial vehicles (UAVs) are used in different scenarios, a variety of potential risks and safety challenges have arisen. One of the threats is that an increasing number of UAV encounter events are found near the airport in recent years, which pose significant dangers to manned aircraft and result in accidents. However, only a few studies examine the impacts of these events and propose effective countermeasures to enhance safety. To unveil the risks of UAV risk events (incidents or accidents) and examine the mechanism with risk factors, this study uses a tree-augmented naive Bayes network (TAN-BN). This method analyses the relationships among risk factors and UAV accidents/incidents to assess the efficacy of risk mitigation measures. Environmental, technological and human factors are simultaneously considered in constructing the Bayesian network. The analysis results reveal 12 specific risk factors that are significantly associated with UAV accidents/incidents in UAV operation scenarios, among which flight control system failure (the most critical factor), remote communication failure, other aircraft approaching, loss of electrical power, adverse weather, electromagnetic interference, operational errors, violations and risk factors leading to loss-of-control in flight are recognised as the most prominent factors. Based on these, five targeted risk mitigation measures are comprehensively implemented and evaluated. Moreover, a case study using UAV operation data near the Guanghan airport is introduced to justify the generalisability of the proposed TAN-BN model and the effectiveness of risk mitigation measures.