Natural hazard-triggered technological accidents (Natech) add a layer of complexity to disaster risk management. They involve the combination of both natural and anthropogenic hazards. Noticeably, there is an uneven distribution in the extent of research dedicated to the different natural hazards as causes of accidents. Extreme weather events have been overshadowed, despite low temperatures being the third cause of Natech accidents in Europe, ranking only behind lightning and floods. The present study aims to comprehensively analyse the role of safety barriers within the context of Natech events triggered by cold waves. Data selected from established databases are investigated through Unsupervised Machine Learning and Bayesian Networks. This approach facilitates identifying and quantifying the relationships between undesired event features and safety barrier failures. The model obtained is designed to predict the behaviour of safety barriers, providing a dynamic and updatable structure adaptable to future developments and integrable with additional available information. This analysis provides valuable lessons to prevent the recurrence of similar events in the future and a robust foundation for the proposition of targeted safety barrier protection programs.
