【作者】 王晓博；

【导师】 李一军；

【作者基本信息】 哈尔滨工业大学 ， 管理科学与工程， 2008， 博士

【摘要】 随着国际互联网的发展和信息技术的日新月异,电子商务正越来越深入影响着当今经济生活。如何合理地对设施选址、需求分派、运送方式和路线选择等进行决策,建立一套高效率的配送系统,从而降低配送成本,提高服务质量,成为电子商务物流业发展面临的重要问题。因此,具有较高的研究价值和实用价值。首先,本文研究电子商务配送中心选址模型与评价方法。依据电子商务下配送网络的不同和配送战略的侧重点不同,构建了基于时间-成本的分散型配送中心选址模型和基于服务-成本的混合型配送中心选址模型。针对分散型选址模型的约束条件和0-1变量多的特点,采用基于分解-过滤的启发式算法求解。针对混合型选址模型节点多和备选地址不确定的特点,用基于模糊c均值聚类法和扫描法的三阶段启发式算法进行求解。由于配送中心选址是一项复杂的系统工程,还需进行定性的选择。为此,建立多准则多层次模糊评价模型,模型中准则层权重是运用基于按比例分配的模糊层次分析法获得,子指标的权重通过计算模糊评价值的期望值法获得,选址方案评价指标权重是准则层权重和对应准则下子指标权重的组合。方案确定是为了决策,依据定量计算和定性分析的结果,采用协调分析法确定最优选址方案。其次,研究了电子商务下车辆调度优化问题。为满足电子商务客户多样化和个性化的需求,分别建立多约束条件的车辆调度模型、一体化配送与集货车辆调度模型以及有时间窗的车辆调度模型。针对多约束条件模型的特点,设计基于改进的顺序交叉算子和引入爬山算法的混合遗传算法,通过对仿真实例的计算,无论在寻优结果,还是在算法的稳定性上均好于标准遗传算法和改进遗传算法。对于多个配送中心的一体化配送与集货模型,采用对混合遗传算法求得的精英种群进行禁忌搜索的混合遗传启发式算法。仿真实例计算表明该算法好于单独使用遗传算法或是禁忌搜索算法。由于,有时间窗的车辆调度问题中产生的等待费用和延迟费用影响配送成本,为此,在考虑配送线路前提下;将最小费用作为优化目标,并设计了改进两阶段算法对该问题求解。通过对仿真实例的计算,表明该算法具有简单、清晰、灵活的特点,并为大规模解决实际问题提供思路。最后,研究了电子商务下一体化配送与集货以及有时间窗的定位-运输路线安排问题。由于传统的多级分解算法易陷入局部最优解,而不是全局最优。因此,本文从整体上设计混合启发式算法求解一体化配送与集货的定位-运输路线安排问题。首先,采用基于分层聚类算法和改进重心法构造弱初始可行解;其次,使用改进插入法生成强初始可行解;最后,设计了基于4种邻域操作、利用能力约束条件控制配送中心的启用与客户点的插入的禁忌搜索算法进行优化求解。仿真实例计算显示本算法的良好寻优性能,具有很高的收敛速度。对于有时间窗的定位-运输路线安排问题,设计了混合遗传模拟退火算法求解。采用基于向量的混合编码,引入个体数量控制选择策略;对线路子串用改进的最大保留交叉操作,对中心选址定位子串用单点交叉操作;采用全局最优基因保护策略,使用自适应变异算子;利用模拟退火算法的Boltzmann机制,控制交叉、变异操作。仿真实例计算证明本算法,无论在寻优结果,求解质量上、计算效率上,以及算法稳定性上均好于单独应用遗传算法或是模拟退火算法。通过对文中所给出的模型和算法的实例分析,证明了这些模型和算法的有效性和实用性。更多还原

【Abstract】 With the development of internet and information technique, e-business affects economic life more and more. It becomes the important problem facing with e-business logistics industry that how to make decisions on logical facility location, demand assignment, transportation mode and routing selection, establish high efficient distribution system so as to reduce the distribution cost and improve service quality.First of all, the study research on location model and appraisal method of the logistics distribution center under e-business. According to the different distribution network and strategy, establish the scattered distribution center location model based on time-cost and hybrid distribution center location model based on service-cost. For the restrain condition of scattered distribution center and the characteristics of multi-variable with 0-1, solve the problem adopting the heuristic algorithm based on decomposition-filtration. Considering the characteristics of more node of hybrid location model and uncertainty of standby location center, solve the problem adopting three-stage heuristic algorithm based on fuzzy C-means clustering algorithm and scanning method. Because distribution center location is complex system engineering, it needs to make decisions with qualitative method. Therefore, establish fuzzy appraisal model with multi-criterion and multi-hierarchy. The weight of criterion in model can be gotten based on fuzzy AHP of distribution with proportion. The weight of sub-index can be gotten through calculating fuzzy appraisal value and expected value. The weight of appraisal index of location scheme is the combination of the weight of criterion and the weight of sub-index under the corresponding criterion. The aims of confirming scheme is to make decisions. According to the result of quantitative calculation and qualitative analysis, confirm the optimal scheme using coordinated analysis.Secondly, research on the optimization problem of vehicle scheduling problem under e-business. In order to satisfy with the individual and various demand of customer under e-business, respectively establish vehicle scheduling problem model with multi-restrain condition, vehicle scheduling with picking- delivery model and vehicle scheduling model with time-windows. According to multi-restraint condition and the characteristics of model, design hybrid genetic algorithm based on improved ordered crossover operator and introduced climbing algorithm. Emulation and calculation prove that it is better than improved genetic algorithm and standard genetic algorithm from the side of not only the stability of algorithm but also optimization result through. To picking-delivery model with multi-distribution center, stock elite adopting genetic algorithm take the hybrid genetic algorithm with taboo searching algorithm. The emulation and calculation proves that it is better than only using genetic algorithm and taboo searching algorithm. For waiting expense and delaying expense affecting distribution cost in vehicle scheduling problem with time windows, the minimum expense is the optimal aim considering distribution route. And design improved two-stage algorithm to solve the problem. The emulation and calculation prove that this algorithm has the characteristics of simple, clear and flexible and it can offer the thought to settle the practical problem in scale.Finally, the study proposes the location routing problem model with time windows of picking-delivery under e-business. Considering that the traditional multi decomposition algorithm is easy to get into partial optimization solution, and not overall optimization solution, the study deign hybrid heuristic algorithm to solve the location routing problem model of picking-delivery as a whole. Firstly, establish weak initial feasible solution based on hybrid clustering algorithm and improved center method. Secondly, create strong initial feasible solution based on improved insertion method. Lastly, optimize the solution with taboo searching algorithm. The emulation and calculation proves that this algorithm has good searching performance and high constringency speed. For location routing problem with time windows, solve the problem using hybrid genetic simulating annealing algorithm as a whole. Adopt the hybrid code based on vector. Introduce the individual amount control selection strategy. To route and location sub-branch, execute the operation of keeping the operation maximum reservation crossover point and single crossover point. Adopt the optimization gene protection strategy as a whole. And control crossover and variation operation using the Boltzmann mechanism of adaptive mutation operator and simulated annealing algorithm. The emulation and calculation prove that this algorithm is better than only using genetic algorithm and simulated annealing algorithm from the side of searching result, solving quality, calculation efficiency and algorithm stability.According to model and algorithm, the study offers the case analysis. The emulation and calculation prove that the model and algorithms is effective.更多还原