Research On Data-oriented General Aviation Risk And Control Modeling

High level of aviation safety has always been the ultimate mission pursued by the aviation industry.Aiming to accomplish this mission from aviation accident event analysis,this thesis propose a method focusing on aviation accident important events recognition;importance analysis;accident parameters prediction,finally general aviation risk quantification model and optimized safety resource allocation model are been established based on general aviation accident data.Core contents are as follow:A statistical analysis based model was proposed to identify the differences and unique feature from general aviation and general transportation aviation at the level of accident event;A method based on the multivariate statistical method to measure importance between aviation events are proposed;A general aviation risk quantification model based on the idea of a risk matrix is established;An accident parameters prediction model based on extreme learning machines is established to satisfy the requirements of future aviation industry risk analysis and optimization;A risk control model based on the allocation of safety guarantee resources is proposed by introducing the function of the proportion of accidents and the proportion of resources input.Based on the data of the NTSB Aviation Accident Database in the last 11 years,the quantitative values of the importance of different accident events have been calculated,and the important events in general aviation have been found out according to the proposed research ideas and methods;Using the proposed forecasting model applicable to the aviation industry,the accident parameters for the next year was successfully predicted;The risk value of the entire industry has been calculated based on forecasted data and historical data;Finally,according to the resource allocation model,the resource ration in the example is reasonably allocated,which significantly reduces the risk of the entire industry.Experimental results prove that the analysis method and model proposed in this thesis can effectively calculate importance value and provide risk reduction strategies under the input of real data sets.