Research On Flight Quality In Landing Phase Based On Time Series Analysis

As an important process of the growth of route transport pilots,flight training is not only an important stage for flight students to learn flight technology,but also an important part of ensuring flight safety.In order to solve the problem of relying on the subjective scoring of instructors in the traditional flight training quality assessment,this thesis takes the complex landing stage as the research object,combined with the flight data(airborne SD card data)recorded by the airborne Garmin 1000 integrated avionics system,and quantifies the flight training quality in the landing stage by using the method of time series similarity measurement according to the characteristics of flight time series data.Then,based on XG-Boost,a flight training control quality evaluation model in the landing stage was established.The research content of this thesis can be divided into the following aspects.First: Firstly,the flight training situation is analyzed,and the rectangular take-off and landing route is taken as an example to introduce the flight stages in flight training.According to the operational characteristics of the landing stage,this thesis divides the landing stage into three parts: leveling completion stage,flat floating stage and landing run stage,summarizing the key points of the landing stage and providing reference for the landing trainee landing training.Second: The characteristics and composition of airborne SD card data are introduced,and the data are preprocessed by least squares method and linear interpolation method respectively when there are wild values and missing values in the process of recording flight data.According to the characteristics of the flight parameters in the landing stage,the data extraction of the landing stage was completed by constructing a decision tree model.Third: By referring to the flight technical error,the assessment requirements of the flight training landing stage and the opinions of the flight instructor,the five flight parameters of altitude,speed,pitch attitude,roll angle and track were selected for research.According to the time characteristics of flight data,the method of time series similarity measurement is adopted,and the target curve and the curve to be evaluated are compared after the dimensionality reduction of the flight parameter curve,and after analysis,three measurement indicators are proposed: angular distance measurement,time difference measurement and function value difference measurement.Fourth: A time series similarity metrics model is constructed based on similarity metrics.Through the cluster analysis of the sample data of flight trainees,the flight training quality level classification in the landing stage was determined,and the label data of flight trainees was marked.The XG-Boost model is used to train the model with the training set data,and the flight training quality classification evaluation model in the landing stage is obtained,and the flight quality division of the flight trainees in the test set is completed.