Aviation Safety Assessment Based on Gray Correlation Analysis and BP Neural Networks

The purpose of this paper is to provide a comprehensive assessment of aviation safety through gray correlation analysis and BP neural network model. Firstly, using Quick Access Recorder (QAR) data, the data are cleaned and reliability studied by 3σ law and normal distribution sample processing, and the importance of key data items is analyzed by using Random Forest and XGBoost models. Secondly, based on the change of the joystick process, the indicators with significant influence on the flight status were screened out, and the BP neural network model was constructed to establish the mapping relationship between the independent variables and the flight maneuver quantization. In addition, in the analysis of overrun conditions, this paper applies factor analysis to explore the basic characteristics of different overrun conditions. Then, through the gray correlation analysis method, to overcome the inconsistency between the quantitative and qualitative results of the traditional system analysis method, the regression prediction is carried out in combination with the neural network model, and the flight skill level of different pilots is finally assessed. Finally, based on the model of BP neural network, a real-time automated warning mechanism is established to discriminate when the QAR value is likely to have an accident through the simulation results and provide corresponding warning hints. This study not only provides new ideas and methods for aviation safety assessment, but also provides data support and decision-making basis for future flight safety management practices.