民航空管威胁、差错与意外状态相关性的贝叶斯网络方法研究

On the correlation among threats,errors,and undesired states in air traffic control based on a Bayesian network model

提出了一种基于贝叶斯网络模型研究民航空管单位运行中威胁、差错和意外状态相关性的方法。该模型中,将威胁、差错中发生频率最高,且直接导致意外状态发生的6种类型信息作为研究对象,即内部程序威胁、相邻单位威胁、空中交通威胁、设备处理差错、程序差错和通讯差错,并以此为基础构建贝叶斯网络。探讨了空中意外状态与各类威胁及差错的关系,提供了基于威胁及差错发生的先验概率,获得意外状态与各类威胁及差错间关系的后验概率的计算方法。应用贝叶斯网络方法研究了威胁、差错以及意外状态各类安全信息的相关性,结果表明:1)基于意外状态后验概率得到的威胁和差错等致因因素严重度排序与基于先验概率得到的排序明显不同,说明后验概率应作为判定威胁和差错影响意外状态程度的主要指标;2)空中意外状态发生时,内部程序威胁与空中意外状态的发生相关性为77%,属最为敏感的致因因素;同时空中威胁的后验概率也达到72%,属高相关性致因因素,应在运行管理中重点管控。

This paper is aimed to introduce a Bayesian network model developed by the author to investigate the correlation among the threats, errors, and undesired states in the air traffic control. In this model, we have taken the Bayesian Network （BN） as the theoretical foundation for analyzing the quantitative method on the database of Threat and Error Management （TEM）. For our research purpose, we have adopted six types of safety states most frequently to be dealt with （that is, the interior procedure threat, the adjacent units threat, the air traffic threat, equipment operating error, procedure error, and communication error） to establish our BN network model. While dis- cussing the relations among the undesired states ~in the air and allthreats and errors concerned, we have established a risk assessment model with reference to the Bayesian Network. All this has enabled the given model we have developed to provide the predictive probabil- ity in advance as for the incoming threats and errors so as to find nee- essary means to work out the corresponding subsequent measures to deal with the undesired states and all other threats and errors. Thus, we have thoroughly founded our research results on a database of TEM and the proper application of Bayesian network. Thus, the above re- search results of ours make us draw the following conclusions： （1） There actually exist significant gaps among the results based on the study of the resultant probability and the anticipating probability. While the resultant probability mainly indicates the probability of threats and errors likely to come about under the premise of accident status, it allows the flight control workers to use al/the information of forecasting as the main indicators to determine the likely events due to the threat and error factors under control. （2） If any undesired states in the air occur, the posterior probability of the interior progress threats can be made as high as to 77 %, which can be thought of as the most effective factor in operation. Meanwhile, the posterior prob- ability of the air threats can also be as high as to reach 72 %, which means it is also a major influential factor to the incidence of undesired states. However, the interior progress threats and air threats should be emphatically controlled in the civil aviation operation.