关键词: 航空运输, 离港航班可变滑行时间, 贝叶斯网估计, 航班离港滑行过程, 增量学习, 期望优化算法

Abstract: Estimated Taxi-Out Time（EXOT） is defined by the Airport-Collaborative Decision Making（A-CDM）, and for improving the efficiency of aircraft department, the influencing factors related to EXOT are analyzed by relevant provisions in A-CDM. A dynamic estimation model of EXOT based on Bayesion network is established according to the analysis of historical data and suggestion of civil aviation experts. Bayesion network is an effective tool to do the data analysis and uncertainty inference by using probability statistics knowledge in complex areas. The control of airport surface operation can be achieved by dynamically modulating the Bayesion network model with incremental learning capacity. Taking a large domestic hub airport as an example, the Expectation Maximization（EM） algorithms can be used to solve the problem of data missing at random, verifying the validity of this model. After comparing the experimental results with the data of actual surface operation, this model can estimate EXOT effectively in a high confidence.

Key words: air transportation, Estimated Taxi-Out Time（EXOT）, Bayesion network estimation, department aircraft taxing process, incremental learning, expectation maximization