基于线形指标的山岭重丘区高速公路事故预测模型

An Accident Prediction Model for Expressway in Mountainous Hilly Area Based on Alignment Indexes

依据我国山岭重丘区高速公路几何线形和交通事故数据,建立了基于交通流量和几何线形指标的高速公路基本路段事故预测模型。首先,基于几何线形条件对基本路段进行了划分,确定了路段单元。其次,分析并确定了理想线形条件的范围,建立了理想线形条件下的基本事故率预测模型。再次,应用BP神经网络与敏感性分析相结合的方法,确定出了对事故发生有突出影响的道路纵坡、平曲线半径和直线段长度3个线形指标,并确定了上述线形指标的事故率修正系数。依据基本事故率预测模型及事故率修正系数即可进行事故预测。模型验证结果表明:该模型能够对路段单元进行事故预测,事故总体预测值与实际值的相对误差在-5.85%^-7.87%之间。

Two accident prediction models for freeways in mountainous and rolling areas were proposed by analyzing the geometric alignment and accident data. Firstly, a series of subsections were obtained by segment division based on geometric alignment conditions. Then the ideal alignment conditions were defined and the accident rate prediction models with base conditions were determined. The three alignment indexes of longitudinal gradient, radius of horizontal curve and length of tangent which have significant impact on occurrence of accident were selected by BP network and sensitivity analysis method. Then the corresponding modification factors of the three alignment indexes were determined. The accident can be predicted by the accident rate prediction model with base conditions and modification factors. The result of model validation shows that the relative error between predicting crash rate and actual value is between -5.85 percent and -7.87 percent which suggests that the models of this paper can be used to predict accident of freeway subsections.