We propose a method based on the Noisy-OR gate Bayesian network to address cases of insufficient sample data. First, a fault tree model of gas pipelines was established. Mapping this model to the Bayesian network (BN), the failure probability was 0.074 according to a traditional BN and fault tree analysis (FTA). By applying the Noisy-OR gate to determine the conditional probability of related nodes, the failure probability of the system was 0.058. Compared with FTA and the BN, this approach could more precisely determine minimum cut sets and diagnose risky factors. The combination of the BN and a Noisy-OR gate is an alternative method for evaluating the reliability of gas pipelines, and this approach can provide a relatively realistic analysis in other evaluation fields because it considers other influencing factors. The findings of this study can aid decision-making and prevent accidents from occurring.

我们提出一种基于Noisy-OR门贝叶斯网络的方法来解决样本数据不足的情况。首先，建立了天然气管道故障树模型。将该模型映射到贝叶斯网络（BN），根据传统的BN和故障树分析（FTA），故障概率为0.074。通过应用Noisy-OR门确定相关节点的条件概率，系统的故障概率为0.058。与FTA和BN相比，此方法可以更精确地确定最小割集并诊断风险因素。BN和Noisy-OR门的组合是评估天然气管道可靠性的另一种方法，并且该方法可以考虑其他影响因素，因此可以在其他评估领域提供相对现实的分析。