Research On Change Point Detection And Recognition Of Aircraft Trajectory

International Civil Aviation Organization(ICAO)predicted that aviation emissions in 2050 will be about 300% higher than that in 2010.As a result,civil aviation energy-saving and emission-reduction become the frontier issue and focus of national struggles in the international climate negotiations.Civil aviation energy-saving and emission-reduction require effective means of active verification and monitoring.According to the aircraft energy consumption model in different flight phases,as well as aircraft trajectories,active monitoring of aviation emissions can be realized.Therefore,flight phase segmentation based on aircraft trajectories is one of the core technologies to realize the active monitoring task of airline emissions.Firstly,the segmentation model of typical aircraft flight phases was established,and those parameters,such as the flight altitude,ground speed and lift rate were extracted,which reflected the changing characteristics of different flight phases.Aircraft trajectory information was obtained via the Automatic Dependent Surveillance-Broadcast(ADS-B)system,and a dynamic data extraction process was established.Secondly,aiming at the flight phase segmentation of real-time aircraft trajectory,based on the aircraft altitude,ground speed and lift rate and other dynamic information that provided by civil aviation surveillance system,a method which calculated the Hotelling’s T-square statistic of sequential double sliding window data was proposed for detecting and estimating the change points.A change point was detected when the Hotelling’s T-square statistic exceeded the threshold,followed by Maximum Likelihood Estimation(MLE)method,which was used to determine the position of the change point.Finally,two optimization methods were proposed to improve the performance of change point detection.On the one hand,by normalizing the constant coefficient of simplified Hotelling’s T-square expression,the significant effect of window length on Hotelling’s T-square statistic was eliminated,and false alarm reduced.On the other hand,diagonal loading method was used to solve the statistical mismatch problem caused by the significant gap of different variables’ variance,as well as making the Hotelling’s T-square statistic more robust.The proposed methods were validated effective via different airline data of B737-800 aircrafts.