Accident prediction model of mountainous freeway based on crash modification factors

doi:10.16265/j.cnki.issn1003-3033.2023.01.0472

Abstract

Abstract:

In order to quantitatively analyze the influence of mountain highway alignment index on the occurrence and severity of traffic accidents, the concept of CMF was introduced, and a highway traffic accident prediction model based on CMF was proposed. The accident data were verified to obey the zero-inflated negative binomial distribution, and ten basic accident prediction models for different alignment combinations of road sections were determined. With the help of the dominance ratio analysis method, ten crash modification factor models of different alignment combinations with respect to the basic prediction model were established. The factors having significant influence on accidents of different severity were identified by elasticity analysis, and the safety effects of highway alignment index and combination of alignment were analyzed. The results show that: CMF can quantitatively reflect the risk effect of different alignment combinations. The higher the value, the higher the accident rate and accident severity may be. In the mountain highway, the combination of horizontal curve and gentle slope is relatively safe, while the combination of horizontal curve and steep slope and the combination of horizontal curve and vertical curve have higher accident risk. The risk of combination of horizontal curve and concave vertical curve is slightly lower than that of combination of horizontal curve and convex vertical curve. Controlling vertical slope and reducing the combination of horizontal curve and vertical curve are beneficial to highway safety in mountainous areas.

Key words: crash modification factors (CMF), mountainous freeway, accident prediction model, accident severity, zero inflated negative binomial (ZINB)

MENG Xianghai, YOU Bingyu, QIU Zhixiong, LI Zhixiao, ZHANG Mingyang. Accident prediction model of mountainous freeway based on crash modification factors[J]. China Safety Science Journal, 2023, 33(1): 10-17.

Figures/Tables 10

Tab.1

Tab.2

Tab.3

Tab.4

Tab.5

Fig.1

Fig.2

Tab.6

The most dangerous and safest alignment conditions of each section

路段类型	最危险线形条件	最安全线形条件
C_G (G≤1%)	CR-500 m,CL- 800 m,G-0.5%	CR-5 000 m,CL-800 m, G-0.11%
C_G (G>1%)	CR-500 m,CL- 600 m,G-4.59%	CR-5 011 m,CL-800 m, G-1.21%
S_CV、 $S S V$	基准条件	SVR-5 000 m,G-5.09%
C_SV	CR-500 m,CL- 600 m,GD-5.45%	CR-4 978 m,CL-800 m, GD-0.2%
$C C V$	CR-500 m,CL- 600 m,GD-6.24%	CR-8 111 m,CL-1 000 m, GD-0.32%

Tab.6

Tab.7

Fig.3

References 17

[1]	World Health Organization. Global status report on road safety 2016: supporting a decade of action[R]. 2016.
[2]	PARK J, ABDEL-ATY M. Evaluation of safety effectiveness of multiple cross sectional features on urban arterials[J]. Accident Analysis & Prevention, 2016, 92: 245-255. doi: 10.1016/j.aap.2016.04.017
[3]	AL-MARAFI M N, SOMASUNDARASWARAN K, AYERS R. Developing crash modification factors for roundabouts using a cross-sectional method[J]. Journal of Traffic and Transportation Engineering: English Edition, 2019, 7(3): 362-374.
[4]	LA TORRE F, MEOCCI M, DOMENICHINI L, et al. Development of an accident prediction model for Italian freeways[J]. Accident Analysis & Prevention, 2019, 124: 1-11. doi: 10.1016/j.aap.2018.12.023
[5]	MATARAGE I C, DISSANAYAKE S. Quality assessment between calibrated highway safety manual safety performance functions and calibration functions for predicting crashes on freeway facilities[J]. Journal of Traffic and Transportation Engineering: English Edition, 2020, 7(1): 76-87.
[6]	DAS S, GEEDIPALLY S R, FITZPATRICK K. Inclusion of speed and weather measures in safety performance functions for rural roadways[J]. IATSS Research, 2021, 45(1): 60-69. doi: 10.1016/j.iatssr.2020.05.001
[7]	WEN Xiao, XIE Yuanchang, JIANG Liming, et al. On the interpretability of machine learning methods in crash frequency modeling and crash modification factor development[J]. Accident Analysis & Prevention, 2022, 168: DOI:10.1016/j.aap.2022.106617. doi: 10.1016/j.aap.2022.106617
[8]	涂圣文, 王冰, 邓梦雪, 等. 考虑平纵组合的事故预测模型在双车道公路线形安全分析中的应用[J]. 公路, 2019, 64(7):196-203.
	TU Shengwen, WANG Bing, DENG Mengxue, et al. Application of accident prediction model in safety analysis of two-lane highway alignment considering horizontal and vertical combination[J]. Highway, 2019, 64(7):196-203.
[9]	周荣义, 林金玉, 刘勇. 危险货物道路运输风险概率模型及应用[J]. 湖南科技大学学报:自然科学版, 2019, 34(1):21-27.
	ZHOU Rongyi, LIN Jinyu, LIU Yong. A risk probability model for road transportation of hazardous materials and Its application[J]. Journal of Hunan University of Science and Technology: Natural Science Edition, 2019, 34(1): 21-27.
[10]	林小媛, 朱顺应, 李维吉, 等. 基于成本-效益的超多车道高速公路左侧硬路肩设置条件[J]. 公路工程, 2022, 47(5): 63-70.
	LIN Xiaoyuan, ZHU Shunying, LI Weiji, et al. Setting criterion of left hard shoulder in super multi-lane expressway with limited width[J]. Highway Engineering, 2022, 47(5): 63-70.
[11]	戢晓峰, 李德林, 杨文臣. 山区二级公路交通事故致因的时间演化机制[J]. 中国安全科学学报, 2019, 29(4):31-36. doi: 10.16265/j.cnki.issn1003-3033.2019.04.006
	JI Xiaofeng, LI Delin, YANG Wenchen. Temporal evolution analysis on causes of traffic accidents occurring on secondary highways in mountain areas[J]. China Safety Science Journal, 2019, 29(4): 31-36. doi: 10.16265/j.cnki.issn1003-3033.2019.04.006
[12]	戢晓峰, 吴亚欣, 郝京京, 等. 平纵组合路段事故严重程度致因辨识模型[J]. 交通运输系统工程与信息, 2020, 20(6):197-204.
	JI Xiaofeng, WU Yaxin, HAO Jingjing, et al. Severity factors analysis model of traffic accident on combined horizontal and vertical alignments[J] Journal of Transportation Systems Engineering and Information Technology, 2020, 20(6):197-204.
[13]	李明, 王永岗, 张巍, 等. 山区高速公路交通死亡事故显著影响因素鉴别[J]. 中国安全科学学报, 2015, 25(5): 126-130.
	LI Ming, WANG Yonggang, ZHANG Wei, et al. Identifying significant factors influencing occurrence of fatal crash on mountainous freeway[J]. China Safety Science Journal, 2015, 25(5): 126-130.
[14]	王磊, 吕璞, 林永杰. 高速公路交通事故影响因素分析及伤害估计[J]. 中国安全科学学报, 2016, 26(3):86-90.
	WANG Lei, LYU Pu, LIN Yongjie. Traffic accidents on freeways: influencing factors analysis and injury severity evaluation[J]. China Safety Science Journal, 2016, 26(3):86-90.
[15]	李庚凭. 基于有序Logit和多项Logit模型的高速公路交通事故严重程度预测[D]. 长安: 长安大学, 2018.
	LI Gengping. Highway traffic accident severity prediction based on ordered logit and multinomial logit model[D]. Chang'an: Chang'an University, 2018.
[16]	段萌萌, 唐伯明, 刘飞. IHSDM在山区高桥隧比高速公路事故预测中的适用性研究[J]. 公路工程, 2018, 43(3): 70-76.
	DUAN Mengmeng, TANG Boming, LIU Fei. Study on the applicability of IHSDM in accident prediction of freeway of high proportion of bridges and tunnels in mountain area[J]. Highway Engineering, 2018, 43(3): 70-76.
[17]	WILSON P. The misuse of the Vuong test for non-nested models to test for zero-inflation[J]. Economics Letters, 2015, 127: 51-53. doi: 10.1016/j.econlet.2014.12.029

路段单元名称	符号	个数
直线+纵坡	S_G	194
直线+凹型竖曲线	S_SV	140
平曲线+凸型竖曲线	C_CV	104
直线+凸型竖曲线	S_CV	109
平曲线+纵坡	C_G	207
平曲线+凹型竖曲线	C_SV	151

变量类型	变量名称及单位	平均值	标准差	最小值/最大值
因变量	路段单元年事故数N	0.738	1.79	0/17
自变量	年平均日交通量(Annual Average Daily Traffic,AADT)/(Veh·d^-1)	11 344	12.43	8 675/17 643
	大型车比例(Proportion of Large Vehicles,PHV)	0.52	0.12	0.23/0.75
	平曲线半径(Curve Radius,CR)/km	0.825	20.32	0/8.11
	平曲线偏角(Deflection Angles,DA)/rad	0.48	2.52	0/5.63
	平曲线长度(Curve Length,CL)/km	0.388	0.95	0/0.18
	纵坡坡度G/%	-0.22	1.91	-5.036/5.092
	纵坡坡度差(Gradient Difference,GD)/%	-0.002 3	1.29	-5.61/5.092
	凸型竖曲线半径(Convex Vertical Radius,CVR)/km	4.824	66.456	0/200
	凹型竖曲线半径(Sunken Vertical Radius,SVR)/km	5.919	88.065	0/300
	竖曲线长度(Vertical Curve Length,VCL)/km	0.101	0.44	0/0.941

变量名称	参数值	标准误	z值	p值	95%置信区间
ln(AADT)	1.09	0.03	3.144	0.000	1.04	1.15
G/%	0.073	0.007	5.7	0.000	0.066	0.089
ln(1 800/CR)	0.561	0.002	11.34	0.000	0.502	0.63
1/CR×1/CL	1.293	0.056	4.463	0.000	0.896	1.504
截距项_cons	-0.939	0.009	7.7	0.001	-1.048	-0.894
过离散参数α	1.86	0.13	—	—	1.708	1.976

线形	基本事故预测模型
C_G (G≤1%)	N_FI=exp[1.09ln(AADT)+0.073G+0.561ln(1 800/CR)+1.293(1/CR)(1/CL)-6.939]
C_G (G≤1%)	N_PDO=exp[1.67ln(AADT)+0.071G+0.421ln(1 800/CR)+1.208(1/CR)(1/CL)-3.299]
C_G (G>1%)	N_FI=exp[1.01ln(AADT)+0.07G²+0.413ln(900/CR)+1.127(1/CR)(1/CL)+0.67]
C_G (G>1%)	N_PDO=exp[1.08ln(AADT)+0.047G²+0.39ln(900/CR)+1.084(1/CR)(1/CL)+0.556]
S_CV、S_SV	N_FI=exp[1.53ln(AADT)+0.052G²+2.317] N_PDO=exp[1.795ln(AADT)+0.044G²+1.873]
C_SV	N_FI=exp[1.05ln(AADT)+298.301(A/CR)+22.091(A/CL)+0.327ln(1 000/CR)+8.702(1/CR)(1/CL)-3.28]
C_SV	N_PDO=exp[1.327ln(AADT)+277.903(A/CR)+20.023(A/CL)+0.307ln(1 000/CR)+8.204(1/CR)(1/CL)-2.936]
C_CV	N_FI=exp[1.13ln(AADT)+380.14(A/CR)+0.263ln(2 000/CR)+14.901(1/CR)(1/CL)-1.462]
C_CV	N_PDO=exp[1.302ln(AADT)+346.992(A/CR)+0.259ln(2 000/CR)+12.88(1/CR)(1/CL)-1.202]

线形	CMF模型
C_G (G≤1%)	CMF_FI=exp[0.073G+0.561ln(1 800/CR)+1.293(1/CR)(1/CL)]
C_G (G≤1%)	CMF_PDO=exp[0.071G+0.42ln(1 800/CR)+1.208(1/CR)(1/CL)]
C_G (G>1%)	CMF_FI=exp[0.07G²+0.413ln(900/CR)+1.127(1/CR)(1/CL)]
C_G (G>1%)	CMF_PDO=exp[0.047G²+0.39ln(900/CR)+1.084(1/CR)(1/CL)]
S_CV、S_SV	CMF_FI=exp[1.53ln(AADT)+0.052G²] CMF_PDO=exp(0.044G²)
C_SV	CMF_FI=exp[298.301(A/CR)+22.091(A/CL)+0.327ln(1 000/CR)+8.702(1/CR)(1/CL)]
C_SV	CMF_PDO=exp[277.903(A/CR)+20.023(A/CL)+0.307ln(1 000/CR)+8.204(1/CR)(1/CL)]
C_CV	CMF_FI=exp[380.14(A/CR)+0.263ln(2 000/CR)+14.901(1/CR)(1/CL)]
C_CV	CMF_PDO=exp[346.992(A/CR)+0.259ln(2 000/CR)+12.88(1/CR)(1/CL)]