针对核电厂改进的辅助给水系统（ASG），研究了基于贝叶斯网络的系统安全评价方法。通过故障模式与影响分析构建ASG系统可靠性框图，根据可靠性框图与贝叶斯网络的映射关系，建立ASG系统贝叶斯网络，利用网络推理计算事件发生概率，对系统进行了可靠性分析。研究结果表明，该方法通过贝叶斯网络的正向推理、反向推理和双向推理，能够计算任意节点事件的发生概率，表征设备多态失效模式和获得更多评估信息，与传统的故障树方法相比，具有更强的分析能力和更接近实际，适用于核电厂安全系统的可靠性评价，且具有一定的优越性。

With respect to the improved Auxiliary Feedwater System(ASG) of Nuclear Power Plant(NPP), this paper studies the safety evaluation method based on the Bayesian Network(BN). The first step is to set up the Reliability Block Diagram(RBD) of ASG through the Failure Mode and the Effect Analysis(FMEA). Then the BN of the ASG is built in terms of the mapping relation between RBD and BN. Based on the network reasoning, the event occurrence probability is calculated and the system reliability is analyzed. It is indicated that through the forward reasoning, the backward reasoning and the bidirectional reasoning, the occurrence probability of an arbitrary node can be calculated and more information about the reliability evaluation can be obtained. Compared with the traditional Fault Tree Analysis(FTA), this method enjoys a more powerful analysis ability and its results are more closely verified by practice. It can be applied to the reliability analysis of the safety system in NPP.