Research On Cluster Analysis And Anomaly Detection Of Aircraft Flight Trajectory In Terminal Area

In the busy terminal control area,the controller needs to issue the corresponding control instructions for the aircraft according to the flight situation,airspace route structure and control intention of the aircraft in the control area,allocation,mobilization,sequencing and safety interval.Depending on the impact of this high load and empirical air traffic control mode,the operation efficiency of the aircraft in the terminal area is not only low it limits the real-time control decision-making and greatly increases the aircraft safety risk.Therefore,considering that the aircraft flight trajectory,as a macro form to represent the aircraft operation state,is the key point to mine the abnormal trajectory from the mass trajectory set based on the pattern of trajectory data,this paper takes this as a breakthrough to study the aircraft flight trajectory distribution rules in the terminal area and conduct real-time abnormal detection,so as to reduce the workload of controllers and ensure the aircraft flight purpose of safety.According to the characteristics of analytical data in surveillance database,data preprocessing and trajectory compression are carried out to obtain the flight reference trajectory;meanwhile,the multi-dimensional characteristic attributes of trajectory data are fully considered,and the trajectory similarity model based on multi-dimensional characteristics is constructed.On this basis,the improved clustering method is adopted to analyze the distribution law of trajectory,and the results show that the improved method improves the performance of the algorithm The ability of aircraft flight pattern mining is proved,and automatic classification of flight trajectory is realized.Further,on the basis of summarizing the existing flight trajectory prediction models,this paper proposes a real-time prediction model of the future flight trajectory of aircraft by combining the energy altitude of aircraft and convolution neural network algorithm.At the same time,the influence of the network model structure is discussed,and the best flight trajectory prediction model suitable for this paper is obtained.The prediction results are compared with the three-dimensional position information predicted by BP neural network.The results show that the flight trajectory prediction model established in this paper has good accuracy and real-time performance.The definition of abnormal trajectory of aircraft in terminal area is given.The entropy of trajectory information and flight distance are used as the characteristic indexes to measure the trajectory deviation from the center trajectory.Combined with statistical theory,an abnormal trajectory recognition method is proposed.Simulation tests are carried out on horizontal section and vertical section respectively,and then the final abnormal aircraft is determined by combining the abnormal detection results on the two sections.Through the evaluation index of anomaly detection method,the best threshold value of anomaly detection is obtained;the final test results show that the combination of flight trajectory prediction model and anomaly detection model can effectively realize the real-time anomaly detection and recognition of aircraft.