基于贝叶斯网的航班过站时间动态估计  

Dynamic Estimation About Turnaround Time of Flight Based on Bayesian Network

作　　者：丁建立[1] 赵键涛 曹卫东[1] 

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300

出　　处：《南京航空航天大学学报》2015年第4期517-524,共8页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：国家自然科学基金(60879015)资助项目;中国民航科技重大专项(MHRD201241)资助项目;中央高校基本科研业务费(3122014P004)资助项目

摘　　要：一架飞机每天要执行多个航班,从而形成航班链。前序航班进港后,若估计出飞机在机场的过站时间,后续航班的离港时间便可较准确给出。文中选取了对航班过站时间影响较为显著的几个因素,运用历史数据,采用最大似然估计进行贝叶斯网参数学习并获得不同情况下过站时间的估计值。同时,利用贝叶斯网增量学习的特性,运用航班增量数据基于贝叶斯估计修正贝叶斯网参数,并用新的学习结果更新过站时间估计值。实验数据表明,所提出的方法能较好地对飞机过站时间进行估计。最后,对影响过站时间的各因素进行了灵敏度分析对比。An aircraft needs to perform several flights one day, thus forming a flight chain. After the for- mer flight arrives the estimate departure time of next flight could be obtained if the approximate turn- around time is acquired. This paper selects several notable factors which affect the turnaround time. Firstly, the Bayesian network is used to acquire estimate turnaround time by learning the parameters through maximum likehood estimation based on historical data. Secondly, the incremental learning property of Bayesian network is used to revise the parameters of the model based on Bayesian estimation using the increased flight data and the turnaround time is updated by the new results. The experimental data indicate that the proposed method has good performance on estimating the turnaround time. Final- ly, the sensitivity analysis and comparison of the factors influencing turnaround time are carried out.

关 键 词：航空运输 过站时间估计 贝叶斯网 增量学习 灵敏度分析 

分 类 号：U8[交通运输工程]