Comparison of Fault Tree and Bayesian Networks for modeling safety critical components in railway systems

In spite of reliable signaling and train protection and warning systems, trains are still passing red signal seven in modern railway systems. These so-called SPAD events can lead to train derailment, head on collisions with other trains, collisions with infrastructure and other adverse consequences. The classical way of modeling such events is by means of Fault Tree (FT) analysis. However, the FT methodology has limitations when modeling complex systems. This motivates an investigation into the use of Bayesian Networks (BN) for modeling and analyzing SPAD and other safety critical events in railway systems. BN allows combining systematic, expert and factual knowledge about the system and is a flexible and compact form of system representation. In this paper, it is studied by means of the SPAD example whether the use of BN provides significant advantages over the FT methodology for modeling safety risks in railway systems. The causes of train derailment due to SPAD are summarized and the FT and BN methods are compared with respect to different modeling and analysis aspects that are relevant for railway systems.