基于贝叶斯网络的民航事故人为因素分析模型（外文翻译）

附录A外文文献Human Factor Analysis Model of Civil AviationIncidents Based on Bayesian NetworkAuthor informationMing Ming Zhang;Yan Yang Wang;Min Luo;Wen Zhong Tang BeihangUniversity;China Academy of Civil Aviation Science and TechnologyAbstractTo generate a model which can provides detailed data analysis supportivein aviation incident analysis,a human factor analysis model based on Bayesiannetwork theory is established.This model is a Bayesian network which usesthree layers nodes to represent causality between human factors and incidents.The specific impact degree of human factors on aviation incidents isrepresented by conditional probability parameters of the model.The modelstructure,constructed by combing hill-climbing search method with CH scorefunction,coincides with the actual data.This model is useful in aviationincident analyses and deductions.Keyword Artificial Intelligence(AI),Bayesian Network (BN),HumanFactors,Civil Aviation Safety,Incident Analysis-1-IntroductionStatistic information of aviation accidents investigation shows that nearly80%accidents were caused by human factors so it's important to do researchon human factors in order to improve aviation safety and upgrade aviationquality.Analysts are trying to find out the human error factors in accidents togive recommendations to prevent the same accidents from happening again.Generally,the theories or models used in human factors analysis are SHELmodel,Reason model,PEAR model and accident chain theory,etc.SHEL andPEAR model all show the various human factors as a simple human factoranalysis framework.Reason model develops a schematic interaction betweenindividual error and latent factors such as faulty top layer managementdecisions.Accident chain theory is that accident is caused by various factorswhich like chain links and any small error which should not be ignored.Theseanalysis models can provide a macroscopic way of research and relationshipsbetween human factors and aviation incidents.However,they can hardlyprovide a more detailed,objective data analysis support,which means that theimpact degree of human factors to accidents is not available under thesemodels.This paper based on Bayesian network theory,employed the layerstructure of human error factor model,developed Bayesian network humanfactor analysis model.By using this model to analyze civil aviation incidentsstatistic data,detailed impact degree of every human factor to incidents isderived as Bayesian network parameters (conditional probability)to helpdiagnose civil aviation incidents.This model can be used in aviation incidentsderivation and deduction.Bayesian NetworkA Bayesian network is a binary group (S,P)in which S stands for adirected acyclic graph and the node of the graph stands for stochastic variable,the edge between nodes stands for the dependency relationship between thevariables 7.P is conditional probability distribution p(r m(r),which stands for-2the strength of dependency relationship among nodes.x(x)is the set o