飞机冲偏出跑道人为差错量化分析模型

A quantification model for controlling human errors resulting from the runway excursion of the aircraft

为了分析飞机冲偏出跑道的人为差错,找到重点防控对象,提出了冲偏出跑道人为差错量化分析模型。首先应用STPA方法对飞机滑行、起飞及着陆程序进行建模,构建安全控制结构图,找出飞机冲偏出跑道的人为差错。然后将它作为BP神经网络的输入层节点,构建24-6-3的3层BP神经网络模型,从ASRS系统中抽取民航领域近5年的冲偏出跑道不安全事件报告作为数据源,根据发生阶段的不同将不安全事件报告分为滑行阶段、起飞阶段、着陆阶段3类,最后将数据源输入BP神经网络模型中进行训练,并利用测试数据检验模型的合理性,准确率为86. 7%。结果表明,特殊情况下机组人员操作不熟练、刹车系统/方向控制错误等是构成所有阶段冲偏出跑道的主要人为差错,飞机冲偏出跑道的发生阶段不同,影响因素的影响程度也不同。

The present paper intends to propose a new analysis model in hoping to quantify human errors of the aircraft runway excursion by identifying and determining the key prevention and control targets for the flight safety.To achieve the said purpose,we have first of all managed to construct a safety control structure diagram through the system theoretic process analysis(STPA)according to the operational procedures for the aircraft taxiing,taking-off and landing based on the observation of the interactive mechanism of the components of the system.All the above mentioned operational procedures should help us to find the 4 types of human errors,that is,the operational errors,the decisionmaking errors,the inefficient skill-taking errors along with the perceptual errors,for a total number of 24 individual factors.And,secondly,attempts have been made to classify the runway excursion of the aircraft into 3 types according to the different stages of the occurrence.That is,the output layer node,the human errors known as the input layer nodes to construct the 24-6-3 BP neural network.And,from the point of view of the aviation safety reporting system(ASRS),we have managed to extract the causes leading to such errors from the reports of such reckless incidents in the civil aviation area that have taken place for the recent 5 years as a data source by converting them into the input vector and output vector of the BP neural network model.The said BP neural network model has usually been supposed to be adopted to train the data source whereas the testing data are supposed to be used to test the rationality of the model.What is more,the aforementioned testing data can also be used to find the accuracy which stands about 86.7%.The results of our examination show that unskilled operation in the special case and wrong brake system/direction control are to be the main human errors responsible for the runway excursion in all stages.And,thanks to the systematic nature of the system theoretic processing analysis and the practicability of the 3-layer BP neural network,the model we have proposed proves to be in a position to quantify the human errors of the aircraft at a specific airport and the scene of the aircraft veer off or overrun off the runway sphere.What is more,it is also proven to be convenient for the purpose of practical application.