实验贝叶斯估计在人为差错致因量化中的应用

Application of Bayesian Estimation in Quantification of Human Errors Causality

研究航空安全的人为差错定量化致因分析问题。以飞行事故及其事故征候为研究对象,通过飞行测录的历史数据分析,提出了一种基于贝叶斯的人为差错定量分析方法。利用Dirichlet分布描述的历史数据作先验信息,利用Gibbs抽样进行贝叶斯估计得到人为差错概率,确定事故征候的风险水平,依据风险水平划分为低风险事件和高风险事件,寻找潜在风险事件,从而实现定量分析。采用改进方法有助于正确判定飞行安全及其风险程度。

In this paper, we discussed the human errors causality in flight safety. Accidents and their accident symptoms were used as the study materiels, and based on historical data analysis in civil flight, Bayesian method was proposed for quantitative analysis of human errors. Firstly, the prior information based on historical data recorded in civil flight was described by Dirichtet distribution. And then the probability of human errors in Bayesian estimation was obtained to achieve the quantitative analysis by Gibbs sampling. Furthermore, the risk of the accident symptoms was determined, as well as the potential risk events were required to be identified. Finally, based on our questionnaire survey, the low risk events and high risk events were correctly determined in events listing. The method in the civil aviation flight helps to correctly determine the affected flight safety and the level of risk.