A multi-scale network approach for analyzing railway operational accidents

Learning from historical railway operational accidents is critical for enhancing railway safety. The interactions among their associated hazards give rise to an accident causation network. This network provides a structured basis for deriving valuable insights into railway operational accidents. In this paper, a novel multi-scale network approach to explore railway operational accidents is proposed, aiming to uncover the underlying features of accidents by analyzing hazards at multiple levels. This approach serves as a powerful complement to existing network topology-based approaches for railway accident analysis. Its originality is distinguished by the introduction of a multi-scale perspective into railway accident analysis, by means of a multi-scale network and its corresponding tailored topological indicators. To facilitate constructing the multi-scale network from accident investigation reports, a modelling method is proposed. The outcomes of the multi-scale network approach provide a decision-making basis for railway operators to formulate multi-granularity accident prevention strategies. An application to real UK accidents demonstrates the effectiveness of the approach in revealing potential accident features and aiding the formulation of multi-granularity accident prevention strategies.