Understanding railway operational accidents using network theory

Abstract Learning from past accidents in railway operations is valuable for ensuring the future safety of railway operations. Railway operational accidents are of different types, such as collisions and derailments. Different types of railway operational accidents are related to each other due to the interactions between hazards leading to accidents. It is useful to explore the nature of accidents as a set. In this paper, a new network theory-based approach to understanding railway operational accidents is proposed, which aims to reveal latent patterns of hazards from an overall high-level perspective. This approach serves as a complement to conventional network theory-based analyses. Its originality is in the customization of a topological analysis for studying accidents, with several tailored indicators adapting to the characteristics of railway operational accidents. It also provides a practical way to extract and construct the accident causation network from numerous accident investigation reports. The outcomes of this approach could assist railway operators in formulating more targeted accident prevention strategies and approaches. The method has been applied to real railway operational accidents in the UK. The results show that the proposed approach is effective and practical in terms of capturing important causes of accidents and revealing latent rules of railway operational accidents.