Explainable Artificial Intelligence Model for Track-Related Rail Accident Severity Analysis of Rail Transports of Hazardous Materials

The transportation of hazardous materials (hazmat) by rail presents significant safety challenges, particularly the potential for hazmat release accidents. While extensive research has evaluated the probabilities and consequences of hazmat rail transportation accidents, the complex relationship between accident severity and its influencing factors remains unknown in accidents caused by track-related factors. This study aims to investigate the track-related safety risk in the rail transport of hazmat from the perspective of accident severity. Explainable artificial intelligence techniques are employed to interpret modeling outputs and identify key contributing factors based on real-world rail accident records involving rail infrastructure statistics and external environmental conditions. The quantitative findings indicate that the position of the first car derailed, train speed, and the number of cars carrying hazmat are critical determinants of severity in track-related hazmat rail transport accidents. Interactions among infrastructure and train- and climate-related factors may collectively intensify the severity of railway accidents. Moreover, temperatures below −7°C (20°F) exacerbate the severity of accidents, likely due to an increased occurrence of track defects. Accidents involving hazmat releases are likely to be more severe when the first derailed car is near the front of the train. This effect is reduced if this position exceeds approximately 40. These insights offer valuable implications for infrastructure resilience and risk-informed decision-making, supporting more effective safety policies for hazmat transportation by rail.