山区高速公路桥隧道段事故预测

Accident prediction of bridge and tunnel section of expressway in mountainous area

为准确预测山区高速公路桥隧道段的交通事故数,首先通过探讨桥隧道段组合结构安全长度,考虑驾驶员识别视距与通风照明需求后对桥隧道段的定义重新进行诠释。然后,选取我国西南山区某高速公路符合桥隧道段定义的2条路段作为研究对象,以该路段2013—2017年的事故数据为样本,采用路段长度一致法将路段划分为1 km的路段单元。最后,考虑桥隧道段风险特性,从构造物设计、道路线形、交通流参数及路面状况4个方面选取13个解释变量,以事故率与事故数作为被解释变量,通过K-S检验与Q-Q图检验分析选用事故数作为最终的被解释变量,并选用负二项模型进行解释变量显著性分析,进一步确定显著性解释变量并对所选路段进行事故预测。结果表明:隧道长度占比、相邻路段纵坡差、抗滑性能指数及年平均日交通量是影响桥隧道段事故发生的显著影响因素,采用负二项回归模型进行事故预测拟合程度较高,预测结果与实际平均单元事故起数误差仅为1.5%,模型具有较好的准确性。

To accurately predict the number of traffic accidents in the bridge and tunnel section of the expressway in mountainous areas,firstly,through the research on the relevant specifications of bridges,tunnels,and tunnel groups,this paper combs the minimum length requirements of each structure of the bridge and tunnel section,and reinterprets the definition of the bridge and tunnel section after considering the driver’s identification sight distance and tunnel ventilation and lighting requirements.Then,two sections of an expressway in a southwest mountainous area of China that meet the definition of the bridge-tunnel section are selected as the research object.Taking the accident data of this section from 2013 to 2017 as the sample,after comparing and analyzing the two-section unit division methods of section length consistency method and non-section length consistency method,it is decided to use the section length consistency method to divide the research section into 1 km section units.Finally,based on the risk characteristic analysis of the bridge-tunnel section,four types of potential influencing factors of accidents in the bridge-tunnel section and a total of 13 explanatory variables are selected,including structure design,road alignment,traffic flow parameters and pavement condition indicators,and the accident rate and the number of accidents are taken as the explanatory variables,Through K-S test and Q-Q chart test and analysis,it is decided to select the number of accidents as the final explained variable.According to the existing research results,it is decided to select the negative binomial model to model and analyze the total number of accidents,then determine the significant explanatory variables and predict the accidents of the selected road sections.The research results show that the proportion of tunnel length,longitudinal slope difference of adjacent sections,anti-sliding performance index,and AADT(Annual Average Daily Traffic)are significant factors affecting the accidents in the bridge tunnel section.The accident prediction using the negative binomial regression model has a high degree of the fitting.The error between the prediction results and the actual average number of unit accidents is only 1.5%,and the model has good accuracy,It can provide effective support for accident prediction of highway bridge and tunnel sections in mountainous areas.