Crash Risk Analysis And Proactive Safety Design For Expressway Weaving Areas

Expressway is a special road system that has large traffic capacity and allows motor vehicles to move quickly.It generally includes urban expressways and freeways,undertaking medium and long-distance transportation service in cities.With the rapid growth of motor vehicle ownership in China,expressway has been recognized as a significant growth source of traffic accidents,especially the weaving area.In the weaving area,lane changing maneuvers are frequent,acceleration and deceleration are concentrated,traffic flow states and crash risk mechanisms are extremely complex.Previous studies have proved that traffic organization design has significant effects on traffic safety in the weaving area,any unreasonable design is easy to induce dangerous driving behaviors and increase crash risk.Therefore,proactive traffic safety design is one of the most effective ways to improve traffic efficiency and safety in the weaving area.However,the current safety analysis and optimization for the weaving area have suffered from several limitations.First,most of existing studies developed the crash frequency model to establish the statistical relationship between expected crash frequency,traffic flow,road design and environment.It is difficult to reflect the effects of changing traffic flow characteristics and road traffic environment on crash occurrence.Second,existing studies shed light on the relationship between traffic flow states and crash risk,ignoring the differential impacts of traffic organization design on traffic flow states.As a result,the correlation mechanism between crash risk,traffic flow and traffic organization design cannot be established.Third,previous studies only considered one objective（i.e.traffic capacity or traffic safety）to optimize traffic organization design of the weaving area,ignoring the contradictory relationship between different optimization objectives.The research on the optimization of traffic organization design from the traffic safety aspect is not comprehensive.This thesis was sponsored by the National Natural Science Foundation of China“Proactive safety design of urban road design based on multi-source data”.This paper systematically studied crash risk model and proactive safety design for expressway weaving areas.The main research work is as follows:（1）Analysis of the relationship between traffic organization design and traffic safety in the weaving areaThe common types of expressway weaving areas in China were summarized,and four types（Type A₁,A₂,B₁ and B₂）were defined according to the merge and diverge behaviors.The traffic organization design characteristics of different weaving types,including ramp spacing,number of lanes and gore areas,etc.were compared in detail.The contingency table chi-square test and nonlinear canonical correlation analysis were used to analyze the correlation between traffic organization design and different crash types.The results indicate that weaving type,number of lanes on the exit ramp,ramp spacing,gore areas,speed limit,the shoulder and median width are significantly associated with crash types.On this basis,random forest algorithm was used to rank the importance of each design factor.The factors related to weaving type and ramp spacing were found to be crucial for traffic safety in the weaving area.（2）Analysis of the relationship between traffic flow states and traffic safety in the weaving areaThe vehicle trajectory data of four types of the weaving area were collected.The traffic flow characteristics of the four weaving types,including average speed,headway,lane changing position and lane changing gap,were compared in detail.Based on the high-resolution traffic flow data of crash and non-crash samples,an attention convolution embedded clustering algorithm（ACEC）was developed to extract the spatiotemporal characteristics of traffic flow in the weaving area,and then classify traffic flow into seven states.The common traffic flow states for different types of the weaving area were identified.On this basis,the Logistic regression models were developed to quantitatively analyze the relationship between traffic flow states and crash risk,identify dangerous traffic flow states,explore the heterogeneity of crash occurrence under different traffic flow states.（3）Analysis of the safety impacts of traffic organization design factors based on traffic flow statesThe Bayesian Networks（BNs）were developed to explore the relationship between key design factors（i.e.weaving types,ramp spacing）,traffic flow states and crash risk of different crash types（i.e.rear-end,sideswipe,hit-object）.The model results indicate that weaving type has significant effects on the relationship between traffic flow states and crash risk,and the safety impacts of ramp spacing were mediated by traffic flow states.Based on the BNs,the heterogeneity of dangerous traffic flow states and crash mechanisms under different weaving types and ramp spacings were analyzed.At the same time,the deep convolution generative adversarial network（DCGAN）was adopted to improve the balance between the samples of different traffic states.Then the discriminative models for traffic states were developed by three traditional methods,i.e.fisher discriminant analysis,k-nearest neighbor and long short-term memory,of which discrimination accuracy was compared in this study.（4）Crash risk model and proactive safety design for the weaving areaThe multilevel Logistic regression model was developed to predict crash risk in the weaving area,establishing the correlation mechanism of crash risk,traffic flow states and traffic organization design.In the modeling process,the hierarchical structure was used to incorporate the heterogeneity under different traffic flow states,and the cross-level effects were used to describe the interactions between traffic organization design and traffic flow characteristics.The results suggest that weaving type,basic number of lanes,short ramp spacing,gore areas and out shoulder width are significantly associated with crash risk in the weaving area.The effects of these design factors are not fixed but vary with traffic conditions.On this basis,the influencing mechanism of each traffic organization design factor on crash risk was explained by the sensitivity analysis.With consideration of traffic capacity and Los of service constraints of weaving areas,the design standard curves of weaving areas based on the crash risk model were drawn,and the selection procedure for three key design factors（i.e.weaving type,basic number of lanes and short ramp spacing）was proposed from the traffic safety aspect.（5）A case study of the traffic safety optimization method for traffic organization designBased on the vehicle trajectory data,the VISSIM simulation models for the four types of the weaving area were developed.The car following and lane changing parameters in the models were modified by the genetic algorithm.Based on the simulation models,the impacts of the actual and optimized traffic organization design on traffic safety and traffic capacity were compared,and the effectiveness of the proposed traffic safety optimization method was evaluated.