The short-term traffic flow prediction can help to reduce flight delays and optimize resource allocation. Using chaos dynamics theory to analyze the chaotic characteristics of en-route traffic flow is the basis of short-term en-route traffic flow prediction and ensuring the orderly and smooth state of the en-route. This paper takes the time series of en-route traffic flow extracted from Automatic-Dependent Surveillance Broadcast (ADS-B) measured data as the research object, uses the improved C–C method to reconstruct the phase space, and uses the improved small data volume method to calculate the Lyapunov index to identify the chaos phenomenon of en-route traffic flow. In order to avoid the interference of chaos phenomenon on traffic prediction, the Wavelet Neural Network (WNN) model is established to predict the traffic flow at en-route points. The experimental shows that when the number of iterations is 10,000, the average accuracy of WNN prediction is 0.87173, and the average running time is 6.9335334[Formula: see text]s. According to the experimental results, it can be seen that the smaller number of iterations has more advantages in running time, which greatly reduces the overall running time. At the same time, it indicates that appropriately increasing or reducing the number of iterations in this experiment has little effect on the results.

In order to efficiently mine the flight trajectory information in the terminal area and grasp the spatial distribution characteristics of the approaching traffic flow in the terminal area, this paper develops a flight trajectory data attribute correlation analysis model, and establishes a flight trajectory feature extraction model based on the data attribute correlation and improved k-means. Using the t-distributed stochastic neighbor embedding (t-SNE) method and density peaks clustering approach (DPCA), the prevailing traffic flow is then extracted. Finally, the real flight trajectory data in the terminal area are verified by analyzing the model parameters. The findings indicate that 788 flight trajectories are compressed to 100 trajectory points, divided into seven clusters, and seven prevailing traffic flows are extracted. Compared with the traditional flight trajectory clustering method, the accuracy of the method is improved by compressing the flight trajectory data scale and reducing the dimensionality. Simultaneously, the application of the DPCA method achieves a more detailed recognition of the flight trajectory.

An accurate recognition method of the air traffic flow pattern is proposed based on historical flight trajectory data to understand the spatial distribution of air traffic flow and improve the airspace utilization in the terminal area. Due to the high dimensionality of flight trajectory data, we establish a trajectory similarity model based on the geodesic distance and use an improved spectral clustering algorithm to classify the flight trajectory sample data. An improved algorithm based on the minimum spanning tree is presented to extract the skeleton similarity between the tracks and obtain the model of the prevalent traffic flow. The experimental results show that the method can accurately divide 1070 flight paths into 5 categories and extract 7 prevalent traffic flows, exhibiting strong robustness to abnormal trajectories and noise.

为有效掌握空中交通流的分布规律,提高飞行轨迹聚类效率与质量,提出了一种精确度高、运算快、自主识别异常轨迹的飞行轨迹聚类方法.首先,改进均匀参数化法降低了飞行轨迹数据规模.其次,提出一种基于核主成分分析(kernel prin-cipal component analysis,KPCA)和飞行轨迹降维方法,突出不同类点之间的差异.最后,采用基于密度空间聚类(density-based spatial clustering of applications with noise,DBSCAN)算法剔除飞行干扰轨迹并完成聚类.实验表明,该方法在简化数据预处理的条件下,对1243条飞行轨迹实现准确聚类,划分识别出6个类别,保持较高的聚类质量并识别异常轨迹.相较于其他聚类方法,该方法简化了聚类前对飞行轨迹的预处理,提高了聚类效率,聚类效果更加准确,并能够识别异常轨迹.

In order to relieve the pressure of airspace and optimize the allocation of airspace resources, the traffic prediction method is studied. This paper proposes a least squares support vector machine (LSSVM) prediction method based on the particle swarm optimization (PSO) algorithm, which optimizes the relevant parameters in the least squares support vector machine by the particle swarm algorithm to construct a PSO-LSSVM waypoint traffic prediction model. The ADS-B data of Dawangzhuang waypoint was used as the base statistical traffic for the simulation test, and the root mean square error of the prediction results were 1.3236 and 1.2898, and the average relative errors were 0.1201 and 0.2703, which verified the correctness and applicability of the method.

From 2015 to 2019, the total turnover of China's civil aviation transportation increased from 85.17 billion to 129.33 billion ton kilometers, with a compound annual growth rate of 8.7 %. With the growth of air traffic flow, the technology of airspace security risk assessment needs to be improved urgently. In order to deeply excavate the data information of aviation security incidents and explore the law and characteristics of hidden accident causes, on the basis of constructing a knowledge graph of aviation safety events, the method of correspondence analysis is used to study the internal relations between the constituent elements of aviation safety events, explores the seven types of aviation security incidents and the correlation degree between time and area, distinguishes and marks the relationship according to the results of correspondence analysis, visualizes the information data of aviation security incidents, obtains the knowledge Graph of aviation security incidents, and provides auxiliary decision-making for airspace security risk assessment.

In order to deeply analyze the spatial characteristics of the traffic flow in the airport scene, reveal the mutual influence and evolution process of various parameters in the traffic flow. This article takes Tianjin Binhai International Airport as an example, draws on the three-phase traffic theory and related analysis methods in ground transportation, and builds taxiway traffic flow based on aircraft ADS-B historical data using the cell transmission model (CTM) method Cell transmission model, and the simulation of the phase change process of taxiway traffic flow on the model scene is realized by MATLAB, and the error between the simulation results and the real data is not more than 0.02 flights/min. Finally, the operating state of the traffic flow on the airport scene is divided, and the cause and evolution mechanism of its formation is analyzed for different traffic flow states. It can be seen from the experimental results that the study on the characteristics of the traffic flow on the airport surface can effectively optimize the arrival and departure procedures, improve the efficiency of airport control, and has great application value for traffic flow management.

Flight number is the main identification mark for aircraft operation management in the control area, but similar flight numbers may cause control risk problems. This paper establishes an efficient method for identifying similar flight numbers by studying the rules of identifying similar flight numbers. Based on the relevant regulations of China Civil Aviation Administration and the identification rules of the experimental analysis in this paper, the procedure of identifying similar flight numbers is optimized and improved. Similar flight number identification method uses specially optimized LD algorithm to better adapt to the screening of similar flight numbers, and to improve the accuracy and effectiveness of similar flight number screening. Finally, by experiments and horizontal comparison analysis with other similarity calculation methods, it is proved that this method can effectively detect similar flight numbers and achieve the objective of assisting air traffic controllers to complete their work safely and efficiently.