Abstract: In order to evaluate the risk of accidents caused by hazards in train control system more scientifically, Bayesian network technology is introduced to figure the causal relationship among hazards, risks and consequences in the accident. Potential hazards, probability of occurrence of accidents and potential severity are identified. By combining with the advantage of Bayesian network processing incomplete data, the risk assessment model based on Bayesian network is established to calculate the Tolerable Hazard Rates（THRs） of accidents. And they are compared with the required safety standards to determine whether the system meets the safety requirements or objectives. The model integrating the railroad grade 1 accident/incident database is used to assess U.S. train protection warning system. THRs of accidents caused by four initial hazards are less than specified value. The validity of this model provides a new thought of train for the concrete implementation method of risk assessment on train control system.

Key words: train control system, Bayesian network, hazard, risk assessment

摘要： 为科学评估列车运行控制系统内各危险导致的事故风险，用贝叶斯网络描述危险、风险和事故后果间的因果关系。通过识别系统中的潜在危险、危险导致事故的发生率和严重程度，结合贝叶斯网络处理不完备数据的优势，建立基于贝叶斯网络的风险评估模型，计算危险导致事故发生的可容忍危险率，判断系统能否满足安全要求并达到设定的安全目标。以美国的列车保护警报系统和I级铁路事故/事件数据库为例，利用该模型进行风险评估，结果表明4个初始危险导致的事故可容忍危险率小于规定值，验证了模型的有效性，为列控系统风险评估的具体实施方法提供了新思路。

关键词: 列控系统, 贝叶斯网络, 危险源, 风险评估