Research On The Recognition And Prediction Of Abnormal Events In Flight Training

With the continuous development of the civil aviation industry in recent years,the demand for flight training has continued to increase,and the safety requirements for flight training have continued to increase.During the flight training process,a huge amount of flight data will be generated.Research based on the collected flight data will help ensure the safety of flight training and improve the quality of flight training.In response to the above problems,this paper analyzes the collected flight training data to study the safety of flight training.Through in-depth analysis of a large amount of data generated during the flight of a training aircraft,use the collected data to identify abnormal events in flight training,combine neural networks to predict flight training abnormal events,and finally develop flight training abnormal event identification prediction system.The specific research contents are as follows:First of all,the specific content of the collected flight training data files is introduced in detail,and the meaning of the data fields in the research files and the internal correlation between each field are analyzed to pave the way for the next step of research.Next,the content of the main abnormal events in flight training is introduced,how the abnormal events occur,and the specific correspondence between the occurrence of the events and the data in the file.Afterwards,build a flight training expert knowledge base,analyze and study the collected flight training data,and clarify the cause,time and specific content of the event.Secondly,methods such as data cleaning and feature engineering processing of flight training data ensure the availability of data and pave the way for subsequent prediction research.Based on the analysis and research of the above flight training data,the neural network method is used to predict the abnormal events of flight training.Construct multiple CNN-BILSTM prediction models based on the attention mechanism,multiple linear regression models based on incremental learning,and CNN-BILSTM-Attention models based on incremental learning,and compare the advantages and disadvantages of each model,and The CNN-BILSTMAttention model based on incremental learning is selected as the prediction model in the system.Finally,according to the software engineering process development method,the development of flight training abnormal event identification and prediction system is realized,and the reliability and usability of the system are verified by experiments.Through the research and development of the flight training abnormal event identification and prediction system,the flight training abnormal event can be identified,predicted and analyzed.Experiments show that the system can effectively identify and predict abnormal events in flight training,and can effectively use the collected flight training data,which will help the follow-up flight training safety research and ensure the safety of flight training.