Generalising Event Trees Using Bayesian Networks with a Case Study of Train Derailment