Critical Scenario Risk Modeling and Analysis: A Case Study of Autonomous Rail Transit Trains

In recent years, the development direction of next-generation railways has been to enhance the autonomy of driverless trains. Although there are anticipated benefits in reducing human errors and significantly improving overall efficiency and safety levels, the development of driverless trains will undoubtedly introduce new risks to the braking process. This paper aims to identify the sensitive factors that increase the risk of train accidents and predict the dynamic changes in these risk conditions. First, through literature review and expert consultations, the paper analyzes risk scenarios directly related to train operation safety and key factors leading to a decline in braking performance. It then models the sequence of events leading to risks using a hybrid causal logic approach. Next, addressing issues such as structural redundancy and the lack of consideration for temporal dynamics in traditional HCL methods, the study introduces dynamic Bayesian networks for model transformation. Finally, the Expectation-Maximization algorithm in GeNIe software is used to calculate the state transition and conditional probabilities of the model, predicting the overall trend of train accident risk and analyzing the most sensitive factors leading to risks at different time intervals. These findings are significant for enhancing the autonomy and safety of driverless trains.