引入路表性能的高速公路负二项事故预测模型研究

发布时间：2018-06-11 21:04

本文选题：高速公路 + 几何线形条件　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：道路条件是交通事故发生的重要原因。其中,交通量和几何线形是造成事故的关键道路交通条件,此方面相关的研究已经取得很多成果。除此之外,道路路表性能也是影响交通事故的重要道路条件,但由于路表性能数据的难获得,此方面相关的研究进展不多。目前,大数据的发展和道路检测技术的规范化使得路表性能数据已可大量获取,由此本文构建包含道路交通条件和路表性能条件的高速公路事故预测模型。常用路表性能指标有五类,分别为路面行驶质量指数、路面损坏状况指数、路面结构强度指数、路面抗滑性能指数及路面车辙深度指数,本文路表性能指标均从0至100取值,分别表示理论上的最差和最好的路表性能。本文通过分析高速公路的道路交通条件和路表性能条件,建立交通事故预测模型。首先,以辽宁省沈山高速公路作为本文研究对象,对路段单元的交通量数据、道路几何线形数据、路表性能指标数据和交通事故数据进行整理分析;分析常用路段划分方法,确定用同质法划分沈山高速公路路段,为事故预测提供基本研究单元。其次,对常用事故预测模型的主要型式及适用性进行了分析,根据高速公路交通事故数据的统计分布特征,确定了本文选用的事故预测模型形式为负二项分布;选择了对道路交通安全有显著影响的交通量、道路几何线形指标和路表性能指标作为事故预测模型的自变量,路段每年事故率作为因变量,依据95%置信水平,构建了本文的负二项分布事故预测模型。然后,依据敏感性分析的原理,计算各个事故影响因素的相对敏感性系数,根据结果分析了变量的影响程度;再分析研究道路几何线形指标和路表性能指标与事故率的关系,得出指标的优化值。最后,以沈山高速公路部分数据为例,对本文的负二项事故预测模型进行验证研究,计算所建模型的预测精度;以广东省粤赣高速公路作为案例,对本文的事故预测模型进行应用,并通过实际事故数验证模型精度,表明本文所建负二项模型具有一定的适应性,然后对广东省粤赣高速公路未来年事故数进行预测。
[Abstract]:Road condition is an important cause of traffic accidents. Among them, traffic volume and geometric alignment are the key road traffic conditions, and many achievements have been made in this field. In addition, the performance of road surface is also an important condition affecting traffic accidents. However, due to the difficulty of obtaining the performance data of road surface, there is not much research progress in this area. At present, with the development of big data and the standardization of road detection technology, the road table performance data can be obtained in large quantities. Therefore, this paper constructs a highway accident prediction model which includes road traffic conditions and road surface performance conditions. There are five kinds of road surface performance indexes in common use, which are road quality index, pavement damage condition index, pavement structure strength index, pavement anti-skid performance index and rutting depth index. The performance index of this paper is from 0 to 100. Represents the worst and best performance in theory, respectively. In this paper, the traffic accident prediction model is established by analyzing the road traffic condition and the performance condition of highway surface. First of all, taking Shenshan Expressway in Liaoning Province as the research object, the traffic volume data, road geometry line data, road surface performance index data and traffic accident data are analyzed. The homogeneous method is used to divide the section of Shenshan Expressway, which provides the basic research unit for accident prediction. Secondly, the main types and applicability of common accident prediction models are analyzed. According to the statistical distribution characteristics of expressway traffic accident data, the accident prediction model selected in this paper is negative binomial distribution. The traffic volume, road geometric line index and road surface performance index, which have significant influence on road traffic safety, are selected as independent variables of accident prediction model, and the accident rate of road sections is taken as dependent variable, according to 95% confidence level, the traffic volume, road surface performance index and road surface performance index are selected as independent variables of accident prediction model. The negative binomial distribution accident prediction model is constructed in this paper. Then, according to the principle of sensitivity analysis, the relative sensitivity coefficient of each accident influencing factor is calculated, the influence degree of the variable is analyzed according to the result, and the relationship between the road geometry linear index and the road surface performance index and the accident rate is analyzed. The optimum value of the index is obtained. Finally, taking some data of Shenshan Expressway as an example, the negative binomial accident prediction model of this paper is verified and studied to calculate the prediction accuracy of the established model, and the Guangdong Guangdong-Jiangxi Expressway is taken as a case study. The application of the accident prediction model in this paper is carried out, and the accuracy of the model is verified by the actual accident number, which shows that the negative binomial model established in this paper has certain adaptability, and then forecasts the number of accidents in the future of Guangdong Guangdong province expressway.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.3

【参考文献】