General Aviation Safety Monitoring And Early Warning Based On Real-time Sensor Data

With the gradual opening of low-altitude airspace,the development prospect of general aviation is becoming more and more open,which corresponds to higher and higher safety requirements.However,unlike traditional civil aviation large transport airliners,general aviation is limited by equipment conditions,operating conditions and other adverse factors,and its safety is difficult to guarantee.Flight accidents occur from time to time,and there has been an increasing trend in recent years.In order to reduce the accident rate of general aviation,it is necessary to comprehensively analyze various factors in the air traffic operation process.In view of the above problems,this thesis has carried out the following research:Firstly,data collection and processing analysis are carried out.By installing comprehensive sensors on the navigation aircraft,real-time monitoring of the flight of the navigation aircraft is carried out and operational data of the aircraft is collected,the collected flight data is processed and analyzed,and compared with the data obtained by the SD card.The experimental results show that the missing values of position data,magnetic field data,six-axis acceleration data,heading data,and linear acceleration data are less than 3.5%,and the maximum missing value is 3.46%,which is magnetic field data,The minimum missing value of 0 is the position data;The measured data delay is less than 3%,of which the highest value is heading data,and the delay ratio is 2.81%;The minimum is six-axis acceleration and linear acceleration,and the delay ratio is 0.18%.Secondly,in view of the situation that there is usually no angle of attack sensor on the general aviation aircraft,an unscented Kalman filtering method is built based on the aircraft dynamics model to estimate the angle of attack,and the idea of heart-to-heart covariance matching is introduced in the filtering to judge whether the filtering is divergent,so as to avoid the filtering divergence and improve the filtering operation rate.Based on the aircraft operation data,the filtering estimation is carried out,and the error is 0.1 °,accounting for88.77%;The error exceeds 0.2 °,accounting for 3.81%;0.1% for errors exceeding 0.3 °;The average error is 0.05 °.Finally,the over-limit events of the general aviation aircraft are studied and analyzed,the risk factors are quantified,and the early warning models are established for different overlimit events.For overrun events with clear warning threshold,such as route deviation overrun and altitude deviation overrun,early warning shall be conducted by establishing a warning threshold database and corresponding warning model;For overrun events without clear warning threshold,such as angle of attack overrun,slope overrun,etc.,the warning model is constructed by three different algorithms,namely KNN algorithm,GBDT algorithm and XGboost algorithm,and the best model is selected by evaluating different models;It is found that the KNN model for overrun warning of the wrong runway is the most suitable,and the XGboost model for other overrun events is the most suitable.