Abstract:One of the important factors causing urban expressway congestion was traffic flow breakdown at bottlenecks, which resulted from mutual effects of traffic demand, driving behavior and road geometry. Aiming at probabilistic nature of breakdown at urban expressway recurring bottlenecks, survival analyses models of breakdown were proposed based on the detector data. Firstly, breakdowns were identified according to speed and density thresholds. Secondly, traffic parameters during and before breakdown were counted, and relationship between probability of breakdown and bottleneck capacity was described by survival function. Finally, factors of bottleneck breakdown were analyzed by COX regression models. 277 breakdown events were collected from three typical two-lane recurring bottlenecks (Guangzhong on-ramp in inner-circle, Wuning on-ramp in inner-circle and outer-circle) at inner-ring expressway in Shanghai. Analyses results showed when the probability of breakdown was less than 50%, the single lane capacities of the three bottlenecks were 1341veh/h, 1552veh/h and 1662veh/h respectively. It can effectively protect breakdown of bottlenecks to adjust or control speed of main lane and ramp volume. The probabilistic capacity of bottleneck would be quantified more precisely according to the proposed method, which also could supply the scientific basis for active traffic management strategies design.