【作者】 毛建民；

【导师】 赵建有；

【作者基本信息】 长安大学 ， 载运工具运用工程， 2010， 硕士

【摘要】 随着公路交通业的不断发展,我国已成为世界上隧道最多、最复杂、发展最快的国家之一。然而,公路隧道的交通安全形势不容乐观,隧道已经成为交通事故的主要空间分布点和事故黑点,且具有事故危害程度大、事后处理困难、容易诱发二次交通事故等特点,提高公路隧道运营安全管理水平仍然是我国公路交通管理者面临的重要课题。本文在阐述公路隧道交通特点、事故形态的基础上,从公路隧道交通事故的直接原因、间接原因和基础原因三方面对影响公路隧道交通事故的因素进行系统分析,阐述影响因素与交通事故的关系,为公路隧道交通事故预测模型的建立奠定基础。运用灰色关联度理论对公路隧道交通事故的影响因素指标进行分析,确定相关度较高的七项指标(GDP、隧道长度、客运周转量、货运周转量、民用客车拥有量、民用货车拥有量、机动车驾驶人数)作为最终的公路隧道交通事故预测模型的影响因素指标。在分析国内外常用的交通事故预测方法及其特点的基础上,针对公路隧道交通系统是一个动态时变参数系统的特点和神经网络在解决复杂非线性系统方面的优势,分析了神经网络应用于公路隧道交通事故预测的可行性,确定利用BP神经网络对公路隧道交通事故进行预测。在介绍了BP神经网络用于公路隧道交通事故预测的原理和步骤后,探讨了建立基于BP神经网络的预测模型的关键技术,包括网络输入输出变量的选取、样本的选取与预处理、初始权值和阈值的选取、隐层节点数的确定、神经元函数、训练算法与参数的选取等,最后以我国1995-2008年公路隧道交通事故统计数据为例对模型进行训练和检验,结果表明此模型具有很高的预测精度,可用于公路隧道交通事故预测。最后,根据致因分析和预测的结论,提出了公路隧道交通事故预防对策。分别从事故发生前(减少事故发生的可能性)和事故发生后(减轻事故发生后引起的后果)两方面提出了预防公路隧道交通事故对策体系。更多还原

【Abstract】 With the rapidly development of transportation, our country has the most complicated tunnels with the fastest developing speed in the world. However, the traffic safety situation of tunnel is not optimistic, tunnel has become a major distribution point of traffic accidents and accident black spots, and has a major accident hazard degree, aftertreatment dealing with hardly and causing secondary accidents easily etc., so improving tunnel safety management level in China is still an important topic which facing the tunnel transportation management.Based on presenting traffic accident characteristics and accident forms of tunnel, this paper analyses the influence factor systematically from direct cause, indirect causes and basic causes of tunnel accidents, expatiates the relationship between traffic accidents and factors for building up a forecasting model for tunnel traffic accidents.Using grey relational theory to analyze the influence causes, choosing seven higher correlation indicators (GDP, the tunnel length, passenger turnover, freight turnover, the number of civil passenger, freight vehicles and driver) as the final indexes of the forecast model.After analyzing the methods of traffic accident forecast and its characteristics, according to the characteristic that the traffic system is dynamic and time-variance and the advantage that ANN has in solving complicated and nonlinear problems, analysis the feasibility of BP Neural networks using to forecast of tunnel traffic accidents. And confirm using it for forecasting tunnel traffic accidents.This dissertation builds up a BP Neural networks model of tunnel safety prediction, and discusses some key technique and means applying the model, including the selection of input-output variables, the selection and pretreatment of swatches, conforming the number of the nodes in hidden layers, the selection of the initial weight and value, the selection of activation function, training arithmetic as well as parameter. At last, through the example of the tunnel safety prediction by the data of our country’s tunnel accidents from 1995 to 2008, it validates that this method opens out the relation of the tunnel safety and its influence factors in some error bound, and could be applied to forecast tunnel traffic accidents.Finally, on the basis of above studied, this paper establishes the prevention system of tunnel traffic accidents, and puts forward preventive countermeasures from beforehand (reducing the possibility of accidents) and afterwards (alleviating the loss of accidents) for tunnel traffic safety.更多还原