| Aviation service system plays a very important role in the field of economic development and civil aviation services.With the increasing civil aviation business,flight delay and airspace efficiency have become the key issues studied and solved by the Civil Aviation Administration in recent years,which has brought huge economic impact to airlines and related industries.The low efficiency of the aviation system will bring various implicit and explicit losses to individual passengers and airlines.Therefore,it is necessary to analyze and optimize the network construction of the aviation system.At present,the route planning of the aviation system basically relies on manual planning and experience,and its antiinterference ability is poor in extreme weather and other situations.At the same time,with the increase of the number of aircraft and airspace flow,the traditional air traffic flow allocation and control become more and more complex,and the traditional manual route allocation method is no longer applicable.In addition,the influencing factors of flight delay have high complexity and non-linear relationship.The different conditions of various regions and airports and even the arrangement deviation of airports or airlines have a certain impact on flight delay,which also makes the prediction more difficult.In this paper,the network analysis of the aviation system is established,including the impact of complex network indicators,node delay prediction,and the optimization of node routes when the network has delay propagation problems.This paper focuses on three aspects:1.The complex network is used to analyze the flight data of civil aviation of China obtained from the Internet and build an aviation directed network.Analyze the construction of directional network,calculate the network centrality index,and study the impact of different centrality indexes on flight punctuality.Count the airport indicators including network indicators,and calculate the correlation between different indicators and punctuality.The first is to study the positive and negative factors that affect the punctuality of flights.The study of these factors will provide some reference for the practical application of improving flight punctuality.The delay prediction of specific flights is particularly important for airline company arrangement,airport resource allocation,various insurance company strategies and personal arrangement.2.In view of the limitations of the existing delay prediction models,this paper proposes a flight delay prediction model with more generalization ability and the corresponding machine learning classification algorithm.The model fully excavates the space-time characteristics of higher latitudes,such as the influence of previous flights,the situation of departure and landing airports,and the overall situation of flights on the same route,and has good generalization ability.The machine learning process is trained with historical data and tested with the latest actual data.The test results show that the model and algorithm can provide an effective method for flight delay classification and prediction.3.Using reinforcement learning method,this paper designs an algorithm model for path planning of multiple flight targets in unstructured three-dimensional airspace.Specifically,based on the method of in-depth reinforcement learning,first simulate the aircraft flow environment and conflict in the airspace,and then train the model through the reinforcement learning method combined with the figure.Through the learning model,multiple flights can be guided to the required course,which can prevent 100%potential conflict and 90.2%potential early warning under the normal airspace flow density.It provides further reference for network optimization and changing from manual distribution to reinforcement learning model to control Airspace Traffic. |
