【作者】 高清平；

【导师】 晏启鹏；

【作者基本信息】 西南交通大学 ， 交通工程， 2010， 博士

【摘要】 危险货物是工业生产的原料、燃料或产品,现代社会不可缺少。但大多数危险货物往往不是在它们的生产地点被使用,而是需要经过较长距离的运输才被使用或处置,危险货物的公路运输量巨大。然而,危险货物容易在生产、贮存、运输中引起泄漏、燃烧、爆炸、中毒等灾害事故,对路径两侧影响范围内的财产和环境产生严重破坏,造成人身伤害、死亡以及人群疏散和交通中断。因此,危险货物运输的风险控制和安全保障问题成为当前社会各界关注的重点问题之一。而不确定条件下危险货物运输的风险分析、路径选择和应急设施选址优化,是保障危险货物运输安全的重要手段之一本论文构建了不确定条件下危险货物运输相关路径选择、网络优化和设施选址的理论框架,为危险货物运输的一系列决策问题提供了理论支持。本论文的研究内容主要包括以下几个方面：(1)通过回顾危险货物运输风险度量和风险分析的既有文献,总结了既有研究存在的问题。应用既有的研究成果,采用模糊集理论和模糊逻辑进行危险货物运输的风险分析,分别引入了用非参数核密度估计方法估计危险货物运输的事件率的概率分布,以及用概率测度-可能性测度转换公式建立风险参数的模糊数。分析了贝叶斯方法具有充分利用先验信息和样本信息的特点,提出采用贝叶斯方法估计危险货物运输的风险参数。根据历史经验数据和相关专家知识确定随机变量的先验分布,结合新样本信息,最后估计危险货物运输事故次数等风险参数。介绍了粗糙集理论的基本方法,阐述了粗糙集理论对危险货物运输风险分析的适用性。通过粗糙集理论的属性重要度方法,识别出影响危险货物运输事件发生的主要因素和次要因素。通过粗糙集理论的规则推理方法,揭示了危险货物运输事件类型和严重程度与多个影响因素之间的“结果-原因”关系。(2)基于路段走行时间、风险的正态分布假设和对数正态分布假设以及既有的路径属性简单累加形式,给出了路径属性概率分布的递推形式。基于路径属性的概率分布,考虑时间窗约束和属性可靠度约束,采用占优准则最终确定占优路径。针对城市空间发展和危险货物运输冲突的情形,通过个人风险值和社会风险值设定的阈值推算危险货物运输的运输量限制。进一步,依据运输量限制,引入最小费用最大流方法求解从出发地到目的地的最小风险配流方案。(3)引入经典的TSP问题,分析了路径属性均值和方差的既有递推公式,推导了考虑走行风险和服务风险的风险属性值递推公式。为求解随机、动态、带时间窗约束的TSP问题,改进了Tsung-Sheng Chang等人的既有求解算法,设计了求解多目标的随机、动态、带时间窗旅行商问题的启发式算法。针对更为复杂的情形,引入VRP模型求解。(4)针对现实运输网络中,各个路段各种危险货物的单位运输风险和运输费用具有随机性且服从一定概率分布,基于双层规划理论构建了随机风险和费用条件下运输管理者和运输者相互作用的随机双层规划模型。设计了基于随机模拟的遗传算法求解模型,并进行了数值仿真研究。基于图论和不确定规划理论,构建了不确定条件下危险货物运输网络设计的最小风险树模型,第一阶段求解最小风险树问题,第二阶段通过在最小风险树上添加适当的路段,使运输风险增加不多的条件下运输费用显著减小,从而求得令运输管理者和运输者满意的危险货物运输方案。(5)构建了反应时间服从对数正态分布条件下,应急救援设施的的选址优化模型。不仅考虑反应时间的随机性,而且以需求点的受威胁人口数量和事件概率作为权重,增加救援设施的工作量约束和到达时间约束。模型以各救援设施覆盖的需求点全部收益期望值最大作为目标,以各救援设施的位置和各救援设施分别覆盖哪些需求点作为决策变量。结合危险货物运输应急设施选址的实际需要,采用弧覆盖模型进行选址决策,同时引入部分覆盖、覆盖衰减函数拓展了经典覆盖的概念,构建了危险货物运输应急反应设施选址优化的覆盖逐步衰减的弧覆盖模型。更多还原

【Abstract】 Hazardous materials are industrial raw material, energy resources and products and indispensable to our modern society. Most hazardous materials are not used at their point of production, they are transported over considerable distances, and their road freight volumes are high. However, incidents involving hazardous materials cargo can lead to severe consequences characterized by fatalities, injuries, evacuation, property damage, environmental degradation, and traffic disruption. So, risk control and safety guarantee of hazardous materials transportation have aroused people’s considerable attention. And the path selection, network design, and emergency response facility location for hazardous materials transportation under uncertain conditions are important means for the transportation safety.The theoretical framework of path selection, network design, and emergency response facility location for hazardous materials transportation is constructed in this dissertation, in order to provide theoretical support for hazardous materials transportation decision problems. The contributions of this dissertation include:(1) The existing literatures on risk measurement and risk analysis are reviewed, and the existing problems are discovered.According to the existing research achievements, stochastic theory and fuzzy logic are presented to assess the risk of hazardous materials transportation. The nonprarametric kernel density estimation is applied to estimate the probability density distributions of that accident rates, and probability-possibility transformation method is introduced to obtain the fuzzy numbers of risk parameters.Based on the advantages of the Bayesian method integrating prior information sample information, Bayesian method is presented to assess the risk parameters of hazardous materials transportation. The prior distributions of random variables are determined by historical data and expert knowledge, and the risk parameters such as accident frequency are assessed with given new sample information.Rough set theory is introduced and it’s applicability to risk analysis of hazardous materials transportation is discussed. Main factors and secondary factors affecting the incident occurrence are distinguished with attribute significance method of rough set theory. And the relations between factors and incidents are revealed with the rule reasoning method of rough set theory.(2) Based on the existing simple accumulation formula and the assumption that the link attributes follow normal distributions and lognormal distribution respectively, the recurrence formula of the probability density distributions for route attributes such as route risk and route travel time are developed. Considering the probability density distributions for route attributes, the time windows constraints and route attribute reliability constraints, the nondominated paths are achieved by means of three dominance criteria. As for the case where the mutual restriction of urban development and hazardous materials transportation exist, the maximal amount of hazardous materials that are allowed to be transported on each road link can be achieved by individual risk and societal risk threshold. Ulteriorly, according to the freight volume limitations, the algorithm of the minimum-cost maximum flow is used to achieve the optimal flow distribution solutions.(3) The traveling salesman problem is applied to solve the hazardous materials distribution problem. The convolution-propagation formulas of means and variances for the route attribute distributions are analyzed and then the convolution-propagation formulas for route risk distributions considering both traveling risk and service risk are deduced. In order to solve the stochastic dynamic hazardous materials distribution problem, the algorithm presented by Tsung-Sheng Chang is improved and a new heuristic algorithm is developed. As for complicated environments with multi-vehicle and multi-goods, the vehicle routing problem model is used.(4) In a real-life transportation network, the transportation risk and cost of each road link are random variables and follow certain distributions. A stochastic bi-level programming model for network design is developed, which takes into account the interaction between the traffic manages and the hazardous materials transportation carriers. A genetic algorithm based on simulation is designed to solve the problem and a numerical example is presented. Base on graph theory and uncertain programming theory, a minimum risk tree model with uncertainty for hazardous materials transportation is developed. In the first phase the minimum risk tree is found and in the second phase the tree network is extended by adding links, which reduces the total cost and increases risk not rapidly and satisfactory solutions can be achieved.(5) Based on the assumption of lognormal distributions of response times, an optimization model for emergency response facility location is developed, which takes into account response time uncertainty, the number of the people exposed to the danger, the incident probability, the equal workload constraints, and the response time constraints. The objective function is to maximize the total expected demand covered, and the decision variables are the selected location sites and the respective response facilities for every demand nodes. In order to meet the real-life need of emergency response facility locations, a maximal arc-covering model is used and extended with partial coverage and gradual covering decay, and then a maximal arc-covering model with gradual covering decay is developed for emergency response facility location.更多还原