Fault Tree And Bayesian Network Model For Fatal And Injury Crashes On Mountainous Freeways

Due to its varied topography and changeable environment,the mountainous freeway has the characteristics of frequent traffic crashes and high crash severity.The safety situation is severe.It is of great significance to deeply analyze the occurrence mechanism of fatal and injury crashes on mountainous freeways and identify important crash inducing factors for safety improvement and treatment.Therefore,based on fault tree and Bayesian network model,this paper analyzed the causes of fatal and injury crashes on mountainous freeways.According to four aspects of people,vehicles,roads and environment,the possible factors causing fatal and injury crashes on mountainous freeways were summarized.In the light of the significance test,the pareto diagram method and the regressive analysis,the main factors causing crashes were identified,and the factors that are irrelevant and have very little influence were eliminated.Based on statistical analysis of the data,the basic probability of main factors causing crashes was given.Taking the fatal and injury crash on mountainous freeways as the top event,the driver,vehicle,road and traffic environment factors as intermediate events,and the 20 specific crash factors as basic events,the fault tree model for fatal and injury crashes on mountainous freeways was built,and its qualitative and quantitative analysis were carried out.In the qualitative analysis,the minimum cut sets were solved and 25 ways to crashes were found.In the quantitative analysis,the importance of each basic event and the importance of minimum cut sets were calculated,and the influence degree of each factor on crashes was known.Given the quantitative analysis method from the top event to the basic event,five major crash chains were found.Given the quantitative analysis method from the basic event to the top event,the proportion of driver,vehicle,road and environment factors in the cause of crashes was obtained.The events and logic gates of the fault tree were transformed into the nodes and conditional probability tables of the Bayesian network respectively,and the Bayesian network model for fatal and injury crashes on mountainous freeways was built.The bidirectional reasoning of Bayesian network model was carried out.Through the posterior probability reasoning,the probability of each factor under crash condition was obtained.Based on the results of posterior probability reasoning,the ranking of posterior importance of each minimal cut set was obtained.The sensitivity analysis was used to get the sensitivity of each factor to the crash.The most probable explanation problem was solved and the most probable combination of factors causing crashes was obtained.Considering the polymorphism of events,the correlation between various factors and the uncertainty of logical relationship among factors,the Bayesian network model structure was adjusted,and the conditional probability table was changed by expert evaluation method,thus an improved Bayesian network model for fatal and injury crashes was built.The bidirectional reasoning of the improved model was carried out.The weaknesses in different crash states were diagnosed by posterior probability reasoning.Through reliability importance analysis,criticality importance analysis,sensitivity analysis and most probable explanation reasoning,the changes of the diagnosis results after the improvement of the model were understood,and further the main causes of crashes were explored.A decision node and a utility node were introduced into the improved model to form a decision Bayesian network,which provided decision support for the safety improvement of mountain freeways.The analysis contents of the three models were summarized,and the similarities and differences of fault tree model and Bayesian network model in the cause analysis of crashes were deeply discussed.