Comprehensive Analysis And Modeling Of Aircraft Flight Safety In Approach And Landing

With the rapid development of the civil aviation, the number of aircraft and the density of routehave increased rapidly, which made it particularly important to ensure flight safety. Various factorswhich may cause accidents make descent and approach an accident-prone phase all around the flight.Currently, two main ways to prevent flight accidents and control flight risk are safety training afteraccident analysis and risk evaluation based on flight safety model, but both ways don’t make athorough research on how accidents happen. In this paper, aircraft attitude control of the descent andapproach phase is taken as a background and the factors which may affect flight safety are analyzed.A dynamic control model of the aircraft attitude is established which really provides a reference forflight risk control.Factors which affect flight safety are numerous. The factors that affect flight safety of descentand approach phase are summarized through theoretical analysis as follows: flight conditions, aircraftstatus and pilot’s operations. Pilot operations are the main factors. The validity of theoretical analysisis proved through the data analysis of QAR flight data. According to the fact that factors of flightconditions change slowly, a way to classify the flight conditions is proposed, under which furtheranalysis and modeling can be taken.On the basis of the classification of flight conditions, a method to establish the prediction modelof aircraft’s attitude with BP neural network is proposed. Take aircraft status and pilot’s operations asthe input of the training samples and the aircraft attitude angles as output. Train the BP network andget a static prediction model. Three ways are proposed to change the network structure to improve theaccuracy of the model. Contrast experiments are taken and they really optimize network performance.The result indicates that the change of the aircraft’s attitude is in line with the method of "man-machine-environment".Improvements based on the static model are made to achieve aircraft attitude angle’s control.Time information of the original data in the training sample is reserved and a dynamic neural networkidentification model based on time series is established. Take the dynamical identification model as acontrol object and control attitude angles with neural network PID control which can achieve aclosed-loop control of aircraft attitude angles. A dynamic attitude control model of "man-machine-environment" is established. This model provides a verification platform for aircraft’s attitude control,which presents an effective analysis way to prevent risk of flight accidents.