Risk assessment of main accident causes at highway-rail grade crossings

Highway-rail grade crossings (HRGCs) are considered the most dangerous part of the transportation network. Current research focuses on assessing the risk of HRGCs to assist decisions on allocating safety improvement resources for HRGCs. However, due to the lack of accident cause analysis, these studies cannot determine how accidents occurred and thus cannot support the development of targeted prevention strategies. In this study, the main causes of accidents at HRGCs were divided into four categories: inattentiveness, misjudgment, violation, and deliberately disregarded warnings. Four data-driven Bayesian network models were constructed with different types of accident causes as class nodes, and various physical and operational characteristics of HRGCs as feature nodes. The HRGC accident risk caused by different causes can be assessed by these models. Their effectiveness was demonstrated through extensive experiments on a large-scale real-world dataset. Besides, the main occurrence scenarios and key influencing variables of the four accident causes are inferred and compared. Significant differences in some details were discovered. For example, although the crossing accident rate is higher in urban, accidents caused by inattentiveness are more likely to occur in rural areas. In contrast to other causes, inattentiveness is more affected by train speed, even exceeding traffic volume.