【作者】 周鹏飞；

【导师】 康海贵；

【作者基本信息】 大连理工大学 ， 港口、海岸及近海工程， 2006， 博士

【摘要】 伴随着全球经济一体化和国际贸易的迅猛发展,集装箱吞吐量出现了高速增长的势头。作为资本密集型的经济实体,集装箱码头的资源有效管理和调度是箱流畅通的必要保证,更关系到企业成本控制和客户需求的满足,进而影响码头企业的竞争力。本文以岸桥-集卡-轮胎式龙门吊装卸工艺的集装箱码头为研究对象,充分考虑了集装箱码头调度问题的动态性、不确定性和复杂性,通过理论分析和应用研究相结合的方法对面向不确定环境的集装箱码头调度问题的基础理论、应用模型和求解算法进行了较为系统地研究。论文涉及管理学、运筹学、港口规划、信息科学等相关学科,对提高集装箱码头企业自身的竞争力、增加企业收益和客户满意度具有重要的现实意义和理论价值。本文的主要研究内容如下: 1.以面向不确定环境的集装箱码头优化调度研究为背景,针对当前不确定性测度问题可加性和完备性刻画不足的现状,提出了一种能够刻画不确定性因素可加性和完备性的未确定规划方法。在未确定规划三个公理化假设的基础上,界定了未确定测度空间、未确定乘积测度空间、未确定变量、未确定变量的分布函数和分布密度函数、未确定变量的独立性、期望值算子等概念,并对其性质及联合未确定分布函数的确定进行了分析和论证,同时讨论了未确定变量模拟方法。较全面地分析了面向不确定环境的集装箱码头优化调度的方案选择和集装箱码头调度问题中不确定性因素的处理方法,讨论了集装箱码头调度常用的未确定规划模型包括:期望值模型、机会约束规划、机会相关规划,同时提出模型中未确定函数的神经网络模拟方法,并给出未确定规划模型的混合智能求解算法。该方法虽然是针对集装箱码头优化调度问题提出的,但对其他面向不确定环境的规划问题同样适用。 2.在对集装箱泊位-岸桥分配和调度问题分析的基础上,基于机会约束的未确定规划的思想建立了集装箱泊位岸桥分配模型。根据模型最优解的特征,设计了求解模型的集束搜索算法和遗传算法。将遗传算法与神经网络结合,并用神经网络模拟未确定函数,实现了求解该问题的混合智能算法。 3.根据堆场管理的要求将堆场箱位分配问题分解为堆场区段分配和区段内具体箱位确定两部分。根据堆场区段的分配标准分别建立了以平衡堆场段间计划作业量为目标的堆场区段分配未确定期望值模型Ⅰ,以同组箱同区段摆放和堆场区段距相应泊位较近为目标的堆场区段分配未确定期望值模型Ⅱ。建立了区段内具体箱位分配的未确定机会相关规划模型。提出了基于未确定变量机会相关的集装箱堆场进、出口箱的翻箱量计算方法,进而建立了以同组进、出口箱靠近摆放和翻箱量最小为目标的0.1规划模型。基于模型最优解特征设计了求解模型的近似启发式算法。 4.在对岸桥-堆场龙门吊-集卡作业流程分析的基础上,将岸桥-堆场龙门吊-集卡的作业调度问题分为龙门吊在堆场段间的分配和岸桥-龙门吊-集卡的作业序列优化两个子问题。建立了面向不确定环境的龙门吊堆场段间调度的未确定期望值规划模型,合理地在堆场段间调度龙门吊可以使规划时段结束时各堆场段未完成的作业量最少,设计了求更多还原

【Abstract】 With the rapidly increasing of foreign trade and the development of global economy integration, container turnover becomes rapidly increasing. As a capital-dense entity, container terminal’s effective resource management and allocation are the effective guarantee for expedite container flow. And they also affect firms’ operation cost and the custom demand, all of which contribute to container terminal’s competition. Considering the container terminal’s uncertainty, complexity and dynamics in resource allocation, the paper systemically lays emphasis on the theory and method of resource allocation. The study has realistic value, for example, it is beneficial to improve container terminal’s service level and increase corporation income. The researches of resource allocation under uncertainty circumvents are summarized as follows.1. Due to the shortcoming in the depiction of additivity and entirety of uncertain measure problems, the paper proposes a new uncertain measure denotation, in which additivity index (λ) and the entirety coefficient (κ) depict uncertain problems’ characters. On that basis, the paper puts forward some related notions, such as uncertainty measure space, uncertainty product measure space, uncertainty variable, distribution function and distribution density function of uncertainty variable, the variable expected value, uncertainty variable independence, and uncertainty variable simulation are suggested. And the characteristics of related concepts and the determination of a joint uncertainty distribution function are discussed and proofed. Then a simple berth allocation problem with uncertainty variable in container terminal is proposed. Various uncertainties in resource allocation problem are analyzed, and the classification of uncertainties, the mathematical description of uncertainties, the mathematical models for resource allocation and the optimization methods are studied in details. Further, the program of uncertain neural network optimal model in the uncertain environment is given, and the intelligent algorithm of the uncertain program is given as well.2. Based on analysis of the berth and quay crane allocation problem in container terminal, the paper proposes a berth and quay crane allocation uncertainty programming model in uncertain chance constraint. Considering the characteristics of its optimal solution, the author puts forward improved beam search algorithm and genetic algorithm to solve the above model. Genetic algorithm and neural network are combined to simulate uncertainty function and a mixed intelligent algorithm is proposed for the model.3. Considering the characteristics of container storage management, storage space allocation problem is divided into two steps: storage block allocation and container location determination. According to storage block allocation criterion, uncertainty expected value model I and II are proposed. Model I aims to balance the workloads among storage blocks, and model II is to stack the same group containers as on the same block as possible and to decrease the distance of conveying containers between the storage and quay area. The method dependent of uncertainty chance to calculate reshuffle is discussed. Then, dependent of chance uncertain 0-1 programming model of container location determination in a storage block is proposed to put the same group containers as close as possible and decrease total reshuffles of inbound and outbound flow.4. Based on the analysis of the workflow of quay-cranes, Rubber Tyred Gantry Cranes (RTGC) and trucks in container terminal, the equipment allocation problem is split into RTGCassignation among blocks and operation sequence optimization of quay-cranes, RTGCs and trucks. The uncertainty expected value programming model is proposed for RTGC assignation to reduce undone planning jobs by the end of planning period. And the beam search algorithm is discussed to solve the proposed model. The operation sequence optimization is formulated by an uncertain program model based on chance constraint, which could reflect risk preference and experience of decision-makers. Considering the model’s complexity, the genetic algorithm and beam search algorithm are discussed to solve the model. And the functions dependent of chance is treated by direct predigestion and neural network simulation method respectively.5. Container terminal is a dynamic and cooperative system. The flexibly, robustly, dynamically, synthetically design and development of Multi-agent system is the necessity for container terminal allocation, through which the integrative performance can be improved and copes with the shortcoming of knowledge and information. To realize the dynamic, cooperative and integrative allocation of container terminal and to decrease the uncertainty troubles, based on uncertainty program, the paper puts forward distributed and intelligent Multi-agent system, in which agent selection, MAS architecture, communication, conflict and cooperation are discussed. In this system we use agent to integrate some optimization technologies and make them get mutual advantages, through which optimize the container terminal production process. The prototype system is developed by C# language.更多还原