华南理工大学学报(自然科学版) ›› 2022, Vol. 50 ›› Issue (7): 76-84.doi: 10.12141/j.issn.1000-565X.210658

• 交通运输工程 • 上一篇    下一篇

平凸曲线组合均衡性对公路安全性的影响

王晓飞1 李思雨1 陈迷2 申天杰1,3 刘永4 肖鹏2   

  1. 1.华南理工大学 土木与交通学院,广东 广州 510640
    2.广州公路工程集团有限公司,广东 广州 510599
    3.广东省现代土木工程技术重点实验室,广东 广州 510640
    4.广州市市政工程设计研究总院有限公司,广东 广州 510062
  • 收稿日期:2021-10-14 出版日期:2022-07-25 发布日期:2021-12-17
  • 通信作者: 王晓飞(1980-),女,博士,副教授,主要从事公路路线及交通安全研究。 E-mail:xiaofeiw@ scut.edu.cn
  • 作者简介:王晓飞(1980-),女,博士,副教授,主要从事公路路线及交通安全研究。
  • 基金资助:
    广东省自然科学基金资助项目(2022A1515011974);广东省现代土木工程技术重点实验室资助项目(2021B1212040003);国家自然科学基金资助项目(51878297)

Influence of the Combination Equilibrium of Horizontal and Crest Vertical Curves on Highway Safety

WANG Xiaofei1 LI Siyu1 CHEN Mi2 SHEN Tianjie1,3 LIU Yong4 XIAO Peng2   

  1. 1.School of Civil Engineering and Transportation,South China University of Technology,Guangzhou 510640,Guangdong,China
    2.Guangzhou Highway Engineering Group Co. ,Ltd. ,Guangzhou 510599,Guangdong,China
    3.Guangdong Provincial Key Laboratory of Modern Civil Engineering Technology,Guangzhou 510640,Guangdong,China
    4.Guangzhou Municipal Engineering Design and Research Institute Co. ,Ltd. ,Guangzhou 510062,Guangdong,China
  • Received:2021-10-14 Online:2022-07-25 Published:2021-12-17
  • Contact: 王晓飞(1980-),女,博士,副教授,主要从事公路路线及交通安全研究。 E-mail:xiaofeiw@ scut.edu.cn
  • About author:王晓飞(1980-),女,博士,副教授,主要从事公路路线及交通安全研究。
  • Supported by:
    the Natural Science Foundation of Guangdong(2022A1515011974);the Foundation of Guangdong Provincial Key Laboratory of Modern Civil Engineering Technology(2021B1212040003);the National Natural Science Foundation of China(51878297)

摘要:

为深入分析平纵线形组合均衡性与道路安全性的定量关系,针对“平曲线+凸竖曲线”线形组合(以下简称平凸曲线),采集了美国华盛顿4条州际道路477 km的道路线形及2011年至2018年的交通数据和事故数据,作为本研究的训练样本和测试样本。根据平纵线形组合特性,提出错位值、平曲线半径、竖曲线半径、平曲线长度和竖曲线长度为线形组合均衡性表征指标,采用决策树、随机森林和极端随机树3种机器学习模型,分析平凸曲线均衡性指标对亿车千米事故率的影响,其中随机森林模型的预测和拟合精度最高。基于随机森林模型进行敏感性分析和数值分析,结果表明:当平曲线半径大于2.8 km或竖曲线半径大于58 km时,平、竖曲线半径的增大对线形安全性影响较小。文中同时研究了平曲线半径较小时均衡性表征指标与事故之间的关联性,并推荐安全性较高的取值范围。研究结论可为后续平纵线形组合的定量优化设计和安全性改善提供参考。

关键词:  公路线形, 道路安全, 平凸曲线, 组合均衡性, 机器学习

Abstract:

To throughly analyze the quantitative relationship between the equilibrium of horizontal and vertical alignment combination and road safety, aiming at the “horizontal curve(HC)+crest vertical curve(CVC)” (referred to as HC-CVC) alignment combination, this study collected the road alignment (a total of 477 km), the traffic data and accident data from 2011 to 2018 of four interstate roads in Washington, D.C. as training data and test data. According to the characteristics of horizontal and vertical alignment combination, the paper suggested to consider the dislocation value, the horizontal curve radius, the vertical curve radius, the length of horizontal curve and the length of vertical curve as variables to characterize the equilibrium of horizontal and vertical alignment combination. Three machine learning models, namely, Decision Trees, Random Forests and Extremely Randomized Trees, were applied for model training to analyze the influence of HC-CVC combination on the accident rate per 100 000 000 vehicle kilometers. The prediction and fitting accuracy of Random Forests is the highest among all models. What’s more, sensitivity analysis and numerical analysis based on Random Forests model show that: when the horizontal curve radius is greater than 2.8 km or the vertical curve radius is greater than 58 km, the increase of horizontal and vertical curve radius has little impact on the road safety. At the same time, this paper also studied the correlation between variables and accident and suggested the value range of the variables when the horizontal curve radius is small. The research conclusions can provide reference for the subsequent quantitative optimization design and safety improvement of horizontal and vertical alignment combination.

Key words:  highway alignment, road safety, combination of horizontal and crest vertical curves, equilibrium of horizontal and vertical alignment combination, machine learning

中图分类号: