| With the rapid development of social economy,in recent years,the numbers of motor vehicle drivers and ownership,and the highway mileage in China have all increased rapidly.Among them,the numbers of drivers and highway mileage have ranked first in the world.As a high-grade road,the highway provides fast and efficient passenger and freight transportation,while at the same time brings traffic safety problems.The highway has a relatively high traffic accident rate and fatality rate(number of deaths/injuries)due to high traffic volume,fast driving speed,and large proportion of trucks.Against the background of continuous increase in highway mileage and passenger and freight traffic,the traffic safety pressure is constantly increasing,the traffic safety situation is not optimistic,and improvement of traffic safety is imperative.In order to build an intrinsically safe highway transportation system,it is necessary to carry out in-depth research on the mechanism of highway traffic accidents,and measure and analyze the impact of people,vehicles,roads,environment and other factors on the highway accident risk and the degree of casualties,and thus to formulate economic and effective traffic safety improvement measures and establish a scientific and effective highway traffic safety evaluation technology system.To this end,based on the road,traffic and accident data of Kaiyang Highway in Guangdong Province in 2014,the following research work is carried out:(1)Based on the principles of plane curvature,longitudinal slope and traffic composition homogeneity,the whole highway is divided into 154 road sections.The accident is correlated with the road section through the accident location in the accident record.In view of the spatial correlation between the accident risks of adjacent road sections,the Bayesian conditional autoregressive model is used to predict and model the annual accident number of road sections,and analyze the influence of roads,traffic and other factors on the annual accident data of road sections.The random parameter with conditional autoregressive prior indicates the common influence of unobserved risk factors on the accident risk of adjacent road sections.At the same time,the fitting performance of conditional autoregressive model,traditional Poisson model and Bayesian hierarchical model in the prediction of accident frequency is compared.(2)The number of annual accidents of the road segment is divided into monthly numbers according to the time of the accident,and the impact of the time change of traffic factors on the frequency of accidents is further analyzed.Random effect model,lag-1 autoregressive model(AR-1),and both(REAR-1)for monthly accident frequency and road and traffic factors of related road sections are established,and the time correlation between the number of incidents in adjacent months of the road sections is explained through the road random effects term,lag-1 random terms,and the combination of the two.The fitting performance of the three models and the Poisson model in the prediction of monthly accidents is compared.(3)Based on the data of the accident record,the road and traffic information are correlated with the accident location and time to construct an accident casualty analysis data set.On the basis of the traditional ordered Logit model,a random term with conditional autoregressive prior is added to analyze the spatial correlation between the casualties of adjacent accidents,so as to establish a spatial ordered Logit model to analyze the influence of drivers,vehicles,roads,traffic,environment and characteristics of the accident on the degree of casualties.On this basis,the formulas for calculating the marginal effects of continuous variables and 0-1 variables are derived,in order to quantify the influence of significant factors on the probability of occurrence of various casualties.At the same time,the fitting performance and classification accuracy of the spatial ordered Logit model and the traditional ordered Logit model are also compared.(4)The number of annual accidents in the road segment is divided into the number of non-injury accidents and the number of casualty accidents according to the degree of accident casualties,and a multivariate conditional autoregressive model is established for joint prediction,in order to analyze the impact of the road and traffic factors on the frequency of accidents at various levels of casualties,and at the same time analyze the correlation between the spatial correlation of these two types of accident frequencies and the degree of casualties.By comparing the fitting performance of the multivariate Poisson log-normal model and the univariate conditional autoregressive model,the superiority of the multivariate conditional autoregressive model is demonstrated.The above-mentioned highway traffic safety analysis results show that there is a significant spatiotemporal correlation in the accident data.The analysis of the spatiotemporal correlation through the Bayesian spatiotemporal model helps to improve the model’s fitting prediction performance.The curvature,slope and traffic composition of the road section have a significant impact on the frequency of accidents.Vehicle type,weather conditions,accident season,traffic composition,rescue duration and accident type all have a significant impact on the accident casualties.The curvature only has a significant effect on the frequency of accidents without casualties,and the slope only has a significant effect on the frequency of casualty accidents.Traffic composition affects the frequency of these two types of accidents in opposite directions.(5)Based on the above safety analysis results,combined with the current situation of Kaiyang Expressway traffic enviroment,from the aspects of safety education,safety warning,traffic police law enforcement,road facilities,accident response mechanism,safety technology development,etc.,suggestions on highway traffic safety improvement measures are provided.The research results of this paper provide theoretical basis and technical support for understanding the mechanism of highway traffic accidents and predicting the development trend of traffic safety,as well as have important reference value for the practice of highway traffic safety management. |
PDF Full Text Request