Traffic Safety Assessment Of Urban River-crossing Tunnels Based On Bayesian Network Analysis

With the process of urbanization,more and more urgent traffic demand to break the geographical barriers has been put forward at a very high demand.Because of its unique influence by natural conditions and small influence on the ground traffic,tunnels have become the hub choice of the city in priority to bridges.Many experts and scholars have not yet realized the particularity of traffic safety in tunnels,thus they still include them in the road traffic safety research system.Researches on tunnel safety are relatively scarce.As a special part of urban tunnels,the research on urban cross-river tunnels is even more scarce.Considering the three factors of human-vehicle-environment,the factors influencing the safety of urban river-crossing tunnels are various,and researches on individual factors are more complex.In this paper,Bayesian network analysis is chosen as the system research method and driving simulator is used as the data acquisition method.Models of the 12 river-crossing tunnels in Shanghai city are established in the driving simulator.With the Bayesian network analysis tool GeNIe,structure learning and parameter learning are conducted using a large number of data to establish a more optimal analysis model of the interaction between the various factors.Finally,the results of model analysis are applied to the Bayesian network prediction,diagnosis,and sensitivity analysis.And the interaction between the factors and the impact of different factors on the final evaluation index of the accident rate of intensity are achieved.As a result,the accident rate is lowest when the vehicle speed is in state of "state 3" and "state 4".In order to reduce the tunnel accident rate,the majority of the speed(64.8%)should be in the "state 4" level,and steering wheel control(69.5%)in the "state 3" level.Further sensitivity analysis is carried out to find the nodes which are more sensitive to the accident rate.The most sensitive factors include "speed","tunnel section","vertical slope","steering wheel","radius".According to the combination of different factors to diagnose the more serious combination of the main factors,main influencing factor combinations of tunnel accidents are found out.The research results not only simplify the scope of traffic regulation in tunnels,but also provide guidance for the improvement of related design codes.