基于CREAM和贝叶斯网络的航空维修人为差错概率预测

航空维修差错不仅严重威胁着飞行安全，同时也会增加航空公司的维修成本。针对航空维修人员发生差错成因的复杂性以及历史事故数据缺乏的情况下，将人因可靠性与失误分析方法（CREAM）和贝叶斯网络（BN）相结合，提出一种改进的维修差错分析模型。根据维修任务构建相应的贝叶斯网络模型，为各子节点设置条件概率表（CPT）；基于维修基地的实际维修环境，对行为形成因子（PSFs）进行评估，得到共同绩效条件（CPCs）的水平；利用各CPC因子下各个行为功能失效模式的权重因子，对各认知活动进行失效概率的修正；将修正概率作为贝叶斯网络根节点的输入，利用推理机制，得到差错发生概率。通过案例分析和计算，验证了所述方法的可行性和有效性。

Prediction of human error probability in aviation maintenance based on CREAM and Bayesian network

Aviation maintenance errors could not only threaten fight safety, but also increase the maintenance cost of airlines. Due to the complexity in causes of human error by aviation maintenance personnel and the lack of enough historical accident data, an improved model on error analysis of aviation maintenance was proposed based on cognitive reliability and error analysis method(CREAM) and Bayesian network. Firstly, according to maintenance tasks, the Bayesian network model of maintenance error was constructed, and the conditional probabilities table(CPT) of each child node was determined. Secondly, based on the practical maintenance environment of certain maintenance base, the performance shaping factors(PSFs) were assessed, and the levels of common performance conditions(CPCs) were obtained. Then, the weighting factors of each behavioral function failure mode under each CPC factor were used to revise the failure probabilities of cognitive activities. Finally, the revised probabilities were regarded as the input of root node in Bayesian network, and the maintenance error probability was calculated by using reasoning mechanism. The feasibility and effectiveness of the method were validated by the case study.