基于Logistic方法的鸟撞对飞行安全的影响分析

Influence Analysis of Bird Strike on Flight Safety Based on Logistic Method

为了掌握鸟撞事件中各类因素对飞行安全影响的规律,采用无序多分类Logistic回归,对鸟撞事件中影响飞行的因素进行分析。数据来源为美国联邦航空局公开的1990~2015年飞机遭遇野生动物(主要为鸟)撞击的数据库。对数据库中的数据进行了简单的剔除和筛选,确定了包含13种影响因素、4种对飞行产生的影响在内的44837条记录;将确定的13个影响因素作为回归变量,4种影响作为响应变量。采用方差扩大因子法,借助SAS软件对回归变量进行多重共线性诊断。结果表明:回归变量间并无明显的共线性关系。应用SAS软件进行最大似然估计,进而对得到的参数估计结果进行分析;获得显著影响飞行的因素为鸟的数量及大小、飞行高度、飞行速度、飞机质量、发动机数量、季节、飞行阶段、日中时间、飞行员被警告与否。为航空公司和机场等相关单位的决策制定提供一定的指导。

In order to understand the influence of various factors in bird strike events on flight safety,the multinomial Logistic Regression was used to analyze the factors affecting flight in bird strike events. The data came from the Federal Aviation Administration’s database of aircraft strikes with wild animals(mostly birds)from 1990 to 2015. The data in the data-base were simply eliminated and screened,and 44837 records including 13 kinds of influencing factors and 4 kinds of influencing factors on flight were determined. The determined 13 influencing factors were taken as regression variables,and 4 kinds of influences were taken as response variables. The multicollinearity of regression variables was diagnosed by using SAS software with the method of variance inflation factor. The results show that there is no obvious collinear relationship among the regression variables. The maximum likelihood estimation is carried out by using SAS software,and then the parameter estimation results are analyzed. The factors that significantly affected flight were number and size of birds,flight altitude,flight speed,mass of aircraft,number of engines,season,flight stage,time of day,and whether the pilot had been warned. It provides certain guidance for airlines and airports and other related enterprises while making decision.