Quantitative Risk Analysis Methodology Based On Hybrid Causal Logic And Its Application

As a safety-critical system, the Chinese Train Control System (CTCS) can be closely related to the trains’operational safety, which has been responsible for the high-speed and effective operation. In this paper, quantitative risk analysis model has been conceived and constructed as well as quantitative analysis methodology, which not only can be exploited to demonstrate risk transmission mechanisms and accident evolution paths, but also to evaluate the performance of deterministic and non-detreministic factors. In ddition, accident occurrence probabilities in special risk scenario could be predicted.First of all, in order to detailly describe the causes and consequences of accidents, the risk analysis model based on hybrid causal logic has been designed. Combined forward reasoning model from risks to accidents with backward reasoning model from consequences to causes, it can provide technical supports for safe operation as well as risk mitigation measures. In addition, the deterministic causal relationships have been identified for equipment and non-deterministic ones for human or organizational factors. Event tree exploited for graphical presentations of risk evolution and calculation of accident probabilities is located at the top, and the middle layer for fault tree is developed for the pivotal events in event trees as well as its occurrence probabilities. Located at the bottom, Bayesian network is applied for modeling the events with variability, complexity and dependence.After that, the quantitative analysis methodology has been conceived on the basis of risk analysis model. The analysis method for human performance based on Bayesian network is put forward. What’s more, human error model can be constructed. The prior probability processing method based on evidence theory is given to make up for the inadequacy and lack of data. The conditional probability allocation algorithm based on fuzzy logic is proposed for reduction of sujective blindness and automatic allocation. Besides, the prerior probabilities has been calculated with Bayes theory using GeNIe software. The influences of organizational factors can be assessed with ω method. By sensitivity analysis, the contributors of risk influencing factors can be evaluated.Finally, the risk analysis model has been applied to analyze the on-board ATP (Automatic Train Protection) equipment. With special risk scenario, the probability of driver error has been evaluated as well as the influence to ATP failure of organizational maintenance activities. As an example, the accident occurrence probabilities caused by SDU failure has been predicted.