【作者】 王绍仁；

【导师】 马祖军；

【作者基本信息】 西南交通大学 ， 物流工程， 2010， 博士

【摘要】 自然灾害中以大规模地震的破坏性较大,使建筑物、通信设施和道路受到严重破坏,导致通往灾区的交通中断,基础设施被毁,在短时间内很难恢复正常,给人类造成了巨大的人员伤亡和财产损失。应对地震灾害的挑战,需要建立一套科学、高效的地震灾害应急物流系统,从而在最短的时间内实现应急救灾物资供应及配送,以最大限度地降低地震灾害造成的损失和危害后果。如何定位应急设施?如何优化应急物资运输路线?这两个问题是震灾应急物流系统优化中的两个关键问题,彼此之间存在相互依赖、相互影响的关系。政府在地震发生后一般会立即启动应急预案,果断地采取了一系列有力的措施来开展紧急救灾,但如果缺乏合理规划的应急物流系统方案,就可能会出现有些灾区重复配送,致使供给过剩,而有些灾区又未接收到应急物资,导致应急物资配送效率低下。因此,对地震灾害应急物流系统优化中设施定位配给问题和车辆路线安排问题的集成优化研究具有重要的理论和实践意义。具体如下：(1)针对震后初期紧急响应阶段空运应急物流系统优化问题,提出了一个以系统总耗时最少为目标的单级设施LRP模型,据此对震害初期进行应急配送中心定位-分配和运输路线安排决策。设计基于两阶段分解的“三角”启发式算法,并与改进的遗传算法进行对比研究,结果表明“三角”启发式算法性能优越,具有较高的运算效率。(2)考虑震后空运应急物流系统中路网中断情况下,以及由于飞机数量有限,每架飞机可能被重复使用,对初期LRP模型进一步改进,建立一个两级设施LRP模型,得出震后应急救援过程中救灾物资集散点和应急配送中心的定位以及运输机、直升机路线安排的联合方案。根据该模型的特点,提出了一种改进的遗传算法,采用特定实值编码、罚函数法和物资需求量分割策略处理模型中的约束条件。给定问题规模下进行算例研究,表明该模型和算法可以有效解决震后应急物流系统中的LRP。(3)针对震后应急物流系统中LRP,考虑系统中的时效性、路网连通性、需求模糊性等特点,建立了一个带时间窗的多式联运模糊LRP多目标优化模型,据此进行救援过程中的灾区外围应急物资集散点和灾区应急配送中心定位以及应急物资运输路线安排的联合决策。针对该模型的特点,提出了一种基于权重系数变换的改进遗传算法,为防遗传算法过早收敛问题,使用了随机遍历抽样法、重组策略和变化变异率法,并通过特定实值编码、罚函数法和物资需求量分割策略处理模型中的约束条件。最后,通过算例分析验证了该模型和算法的有效性。(4)考虑震后不同阶段救灾物资需求强度的不同,应急物资配送时间和成本因素的权重也会随之发生变化,建立了一个多阶段模糊LRP多目标优化模型,据此进行救援过程中不同阶段灾区外围救灾物资集散点和灾区应急配送中心的定位以及救援物资运输路线安排的联合决策,在模糊优化过程中,每条路径(车辆)拜访的最后救援点,可能出现未满足需求,需要采取紧急配送策略以应对未满足救援点的需求。根据该模型的特点,提出一种基于多阶段并行计算和权重系数变换法的改进遗传算法,采用特定实值编码、罚函数法和物资需求量分割策略处理模型中的约束条件,并使用随机遍历抽样法、重组策略和变化变异率法以防遗传算法过早收敛问题。算例分析表明,该模型和算法可以较好地解决震后应急物流系统中的LRP。(5)考虑震后救灾物资需求的优先级,建立了一个不同运输方式的多阶段LRP多目标优化模型,据此得出不同阶段应急配送中心定位以及救援物资运输路线安排决策方案。根据该模型的特点,提出一种基于多阶段并行计算和权重系数变换法的改进遗传算法,并运用罚函数法处理模型中的约束条件,算例分析表明了该模型和算法的有效性。更多还原

【Abstract】 Among the most destructive natural disasters, the earthquake always results in the disruption of transportation and communication facilities, abundant deaths and economic losses in the affected area. Transportation paths are unlikely to recover in a short time after the disaster. Thus, earthquake relief response requires a scientific and effective emergency response logistics system to eliminate the losses and damage. Two inter-connected key problems in the emergency response logistics system optimization are how to locate the emergency facilities and optimize the routes to affected area.After the earthquake, the authorities immediately activated emergency response plans and resolutely adopted a series of effective measures to carry out emergency relief. But the lack of rational planning of the emergency logistics system program always led to iterative distribution, oversupply, missed deliveries, and inefficient distribution of relief resources. Consequently, the integrated optimization of LRP on emergency logistics system in earthquake disaster was studied with its great theoretical and practical significance as follows:(1) For air transport emergency logistics system optimization in initial phase of post-earthquake response, a single echelon LRP model was formulated to minimize the total time elapse of the entire system in relief response. This model assisted in distribution center location-allocation and route planning at the initial response stage. A two phase decomposed triangle based heuristics was generated and proved its ascendency and computational efficiency compared with improved genetic algorithm.(2) Breakage of road network, a two-echelon LRP model was formulated to determine the locations of distribution centers of relief commodities and relief distribution centers, as well as the relief airplane routes during relief process. According to the characteristics of the model, an improved genetic algorithm (GA) was proposed. And a special real-valued coding scheme, punishment function method and demand split strategy were adopted to deal with restriction in the model. The results of a numerical example show that the proposed model and algorithm can resolve the facility location-allocation and cargo-transport plane, helicopter routing problem in post-earthquake emergency logistics systems in an efficient manner.(3) Considering timeliness, connectivity of road networks, multi-modal transportation and uncertain demand in the system, a fuzzy LRP model multi-objective was developed to determine the locations of distribution centers of relief commodities and relief distribution centers, as well as scheduling the routes of emergency vehicle during relief process. According to the characteristics of the model, an weighted coefficient transformation based GA was proposed. To overcome the premature problems of GA, stochastic selection, regrouped strategy and changing mutation probability are used, and a special real-valued coding scheme, punishment function method and demand split strategy were adopted to deal with restrictions in the model. The validity of the model and algorithm was demonstrated by a numerical example.(4) With the changing need intensity of rescue resources at different stages, a weighted coefficient of the dispatching time and the cost was changing along. A fuzzy dynamic two-echelon optimization model with multi-objective was developed to determine the locations of distributing centers of relief commodities around the disaster area and relief distribution centers in the disaster area, as well as the relief vehicle routes in each period during relief process. The last rescue point of every route was likely to not satisfy needs in fuzzy optimization. Therefore, an emergent distribution strategy needs to be adopted. A weighted coefficient transformation and multi-period parallel computation based improved GA was proposed, and a special real-valued coding scheme, punishment function method and demand split strategy were adopted to deal with restrictions in the model. The results of a numerical example show that the proposed model and algorithm are effective for resolving the joint decision-making of LRP in post-earthquake emergency logistics systems.(5) Considering the priorities of relief items, multi-commodities and multi-modal transportation, a stochastic demand LRP model with multi-objective was formulated to determine the locations of relief distribution centers in the disaster area, as well as the relief vehicle routes in each period during relief process. According to the characteristics of the model, a multi-period parallel computation and weighted coefficient transformation based GA was proposed, and a punished function method was adopted to deal with restriction in model. The results of a numerical example indicated the effectiveness of the proposed model and algorithm.更多还原