Resumen
To alleviate the traffic congestion of urban expressways, much previous studies has focused on the analysis of basic traffic parameters, while in reality, congestion is a random event and drivers pay more attentions to the probability of congestion occur and the time it will last, therefore, the purpose of this report is to establish a risk assessment model base on the distributions of capacity and congestion duration (here capacity is a stochastic concept which reflect the probability of congestion occur). Firstly, a criterion for identifying the traffic state was presented, compared with the speed threshold method, it's more comprehensive and reasonable. Then the survival analysis approach was applied to estimate the statistical distributions of capacity and congestion duration. It turned out that Weibull and Loglogistic distribution best fit the distribution curves respectively, on the basis of this, a risk assessment model had been established, this model could be used to assess the reliability of bottlenecks, analysis the influences of different factors on the traffic congestion (morning and evening peak, weekday and weekend, etc.) and applied to the traffic control systems (ramp metering, VMS guidance, etc.). An experiment had been conducted base on the data from Shanghai inner ring expressways to estimate the parameters of risk assessment model, the results indicated that for the studied road sections, the shape parameters of model seems to be around a constant value, while the scale parameters can be much different and this may mostly due to different traffic demand, control conditions and travel purpose for specific road section.