【作者】 王建华；

【导师】 李南；

【作者基本信息】 南京航空航天大学 ， 管理科学与工程， 2010， 博士

【摘要】 汽车是改变世界的机器,推动社会进步的车轮,经济持续增长的发动机和产业结构升级的推进器。在21世纪全球化竞争市场中,为了提高系统运作效率和客户服务水平,汽车制造业需要逐步从传统的面向库存生产转向面向订单生产,汽车制造业供应链需要逐步实现精益化和敏捷化。经过长时间的市场竞争和协调发展之后,汽车制造业供应链逐步采用各种利益风险共担的协调策略、库存管理技术以及信息共享技术,使得供应链可以在近似JIT方式下保持高效的运作。要在此基础上进一步提升供应链运作效率,实现供应链的敏捷性,则需要在供应链成员更精确的运作数据基础上进行供应链生产和运输调度优化。本文对汽车制造业供应链敏捷化策略和关键支撑技术进行了分析和探讨,基于调度决策问题特性构建了通用的供应链结构框架模型并进行供应链结构形态分析,并在此基础上,对敏捷供应链的三类调度决策问题的建模和优化求解方法进行重点研究。首先研究了应对特定市场需求的敏捷供应链静态调度决策的建模和优化问题。静态调度中决策主体需要根据成员企业可调度时段、作业参数以及相关费率进行最终产品及其组成部件的生产和运输任务分配,实现供应链以最小的总成本准时响应市场需求。该部分进行的时间槽编码、遗传算子设计以及贪婪-序解码技术在解决静态调度决策优化的同时,为后续两类调度问题奠定基础。然后研究了应对市场需求发生变化之后的敏捷供应链动态调度决策的建模和优化问题。动态调度中决策主体需要根据需求增加或降低,以及在决策时刻原调度计划执行状态的基础上进行供应链系统的重构,制定最优的生产和运输调度计划。该部分进行的调度分离界面、关键路径分析以及时间槽集合划分在解决动态调度决策优化的同时,为应急调度优化奠定基础。最后,在静态和动态调度研究成果的基础上,研究了供应中断情况下的敏捷供应链应急调度决策的建模和优化问题,供应中断具有两种形态:生产中断和运输中断,应急调度在原调度计划及其执行情况的基础上,构建出两种中断形态通用的调度决策模型和优化求解算法。且对三类调度决策问题模型和优化算法的有效性使用算例及对比仿真实验进行了验证。本文的主要创新点在于:(1)提出了时间槽的概念及其划分方式。使用时间槽表示供应链企业可用调度时段可以将连续调度决策问题转化为离散调度决策问题,在提高求解精度的基础上简化了问题求解的难度。(2)构建了基于调度的供应链结构框架模型。现实汽车制造业供应链系统结构形式多种多样,基于调度问题特征构建一种通用的供应链结构框架模型是实现调度决策优化的前提条件。(3)提出了调度分离界面的概念以及对应的时间槽分类数学规划模型,为实现敏捷供应链调度优化提供了概念集合及计算基础。在动态调度和应急调度中,需要对原调度方案中的时间槽根据当前执行情况进行区别处理,调度分离界面是实现对其区别处理的关键。(4)设计基于贪婪-序解码技术的混合遗传算法,有效解决了汽车制造业敏捷供应链三类调度决策问题的优化求解。调度决策数学规划模型中存在大量时间和数量约束,贪婪-序解码技术可以在特定序列时间槽编码构成的染色体中快速获取满足这些约束的、唯一的可行调度方案,保证遗传算法以较快的速度获得优化解。更多还原

【Abstract】 Automotive is the machine to change the world, the wheel to promote the social go ahead, the engine to ensure the continuous growth of economy and the propeller of the upgrading of the industrial structure. In the 21st century global competition market, in order to improve the operational efficiency and customer service levels, the automotive industry need to gradually change its production style from the traditional Make-to-Stock to Make-to-Order, and the automobile manufacturing supply chain should try to make the progressive realization of lean and agile. After long time of market competition and coordinating development, supply chain enterprises gradually adopt a variety of coordination policies with benefit and risk sharing, inventory management technologies and information sharing technologies which make the supply chain running in a JIT style. Now, in order to further improve supply chain operation efficiency and achieve supply chain agility, the supply chain needs to make the optimization of production and transportation scheduling on the precise operational data of its members.In this thesis, agile strategies and key support agile techniques are analyzed and discussed for automobile manufacturing supply chain, a supply chain structure framework model is built based on the characteristics of scheduling decisions and morphological analysis of the model is analyzed, and then the modelling and optimization methods of the three types of scheduling decision problems of agile supply chains are studied. Firstly, the static scheduling problem is researched where the supply chain needs to deal with a definite demand. the static scheduling needs to decide all of the parts’production and transportation planning which can ensure the supply chain to meet the demand in time with the lowest operational cost based on the schedulable time, operation parameters and the cost rates of the members. The technoloyies, such as time-slot coding, genetic operators, greedy-sequence decoding, which are designed in this section not only resolve the optimization of static scheduling problem, but also firm the strong groundworks for the following two scheduling problems. Then the dynamic scheduling problem is researched where the supply chain needs to change its old schedule because of the old demand quantity is uncorrect and changed now. In the dynamic scheduling process, the decision agent needs to rebuild the supply chain and its optimal production and transportation planning according to the demand changing quantity and the executive situation of the old schedule. The technologies, such as scheduling split interface, key path analysis and time-slot sets classification, which are designed in this section not only resolve the optimization of dynamic scheduling problem, but also firm a strong groundworks for the urgent scheduling problem.. Finally, the urgent scheduling problem is researched where the supply chain need to rescheduling of its old schedule for supply interruptions. There have two kinds of supply interruption: production breakdown and transportation interruption. A common decision model and its optimization algorithm are built based on the original schedule and its exective status. Also, three experiments are proposed to validate the scheduling problems’models and its optimization algorithms.The primary innovations are as follows:(1) The concept of time slots is promoted and its division procedures are designed. Using time-slot to represent corps’available scheduling period changes a continous scheduling decision problem to a descrete one, reduces the optimization difficulty and increases the scheduling precision at the same time.(2) The structure framework model of supply chain based on scheduling is designed. The structures of actual automotive manufacturing supply chains are various, so to construct a common structure framework model of supply chain based on scheduling decisions characteristics is precondition of supply chains’scheduling optimizatioin.(3) The concept of scheduling split interface is defined and the mathmatic models for all of the time slots are set up which give the basis of sets and calculation for optimizing the agile supply chain schedule. In the processes of the dynamic scheduling and the urgent scheduling, the time slots in the origin scheduling should be processed at different manners according to its status, and the scheduling split interface is the key element of the judgement.(4) The hybrid genetic algorithms integrating Greedy-Sequence Decoding Motheds(GSDM) are schemed out which resolve the three scheduling problems of anutomotive manufacturing agile supply chain efficiently. There are a lot of time and quantity constraints in the scheduling decision mathmatic models, the GSDMs can help the HGAs decode exclusive feasible scheduling solutions from a chromosome which is made up of stochastic time-slot codes sequences, and get the optimal solution quickly.更多还原