Research On Risk Assessment Of Civil Aircraft Operation Based On Data Drive

Civil aircraft operation safety is the most important issue for civil aviation.The focus of strengthening civil aviation safety management is to accurately analyze risks and effective risk control.Based on the characteristics of civil aircraft operation,the main findings of the study include the data to characterize the operating status of the data,and adopts data-driven related technologies to deal with the non-linear problems between the related data obtained during the operation of civil aircraft and the operational risks,as well as the time variation caused by the cumulative effect of time problem.The specific contents studied in this article include:From the perspective of complex system risk identification and risk analysis,the feasibility and research ideas of applying data-driven related technologies to civil aircraft systems are proposed,and a data-driven framework for civil aircraft operation risk assessment is established according to the characteristics of civil aircraft operation.Aiming at the characteristics of strong non-linear and noisy data during the operation of civil aircraft,the risk measurement and risk identification were combined to build a Copula-SVM model for civil aircraft operation risk identification,and a support vector machine(SVM)correlation between imbalanced data sets with "thick tail" characteristics.Using the operational data collected from a certain aircraft and the annual fault database as sample data,the accident rate of a certain aircraft operation as the research object,select the failure rate,unplanned replacement rate(URR),average the mean time between failures(MTBF)into internal characteristic index.Weather,machinery,units,ground support and others are external characteristic indexes.The accuracy of the proposed model is higher than 80%,which verifies the feasibility of the model.For the operation of civil aircraft facing harmful consequences,a fault tree(FT)mapping to artificial neural network(ANN)method is adopted to realize dynamic security risk analysis.The hazardous consequences of a certain type of aircraft "civilian fire uncontrollable" is taken as a case.The fault tree is analyzed for this dangerous consequence,the fault tree is established and mapped one by one,and the artificial neural network ANN is used for multiple trainings.Comparing the obtained results with the FT results,it is found that the matching rate between ANN and FT results are better,which shows that the ANN model mapped with FT is an effective risk assessment technique.Civil aircrafts need to deal with models with strong non-linear data,and also need to consider time accumulation.Therefore,the Long Short-Term Memory(LSTM)algorithm and the Genetic Algorithm(GA)are combined to establish a dynamic risk prediction model.The method proposed in the article is compared with the traditional pure supervised learning method and the latest research in recent years.It is found that the accuracy rate of civil aircraft operation risk dynamic evaluation using this method is better than the traditional supervised learning model method.It can be seen that the model proposed in this chapter is effective,can improve the reliability of civil aircraft operation,and has strong application value.