【作者】 尚晶；

【导师】 徐长生；

【作者基本信息】 武汉理工大学 ， 物流技术与装备， 2011， 博士

【摘要】 集装箱码头取得竞争优势的一个有效途径就是优化其装卸资源的配置与调度。集装箱船舶的在港时间和集装箱码头的吞吐量取决于岸桥的作业效率,而堆场作业机械和水平运输机械的合理配置与调度又直接影响岸桥作业效率的充分发挥。目前,集装箱码头生产过程中,设备资源的配置、调度以及路径优化等方面还大多采用经验管理,而且与集装箱码头生产调度相关的学术研究中涉及双小车岸桥和双40英尺岸桥等新型设备的还非常少见。因此,本文在综合分析相关调度优化理论与技术的基础上,结合新型岸桥的作业特点,对集装箱码头的机械配置与调度策略进行研究,主要内容如下：(1)以岸桥-轮胎式龙门吊-集卡的配置为研究对象,分两个阶段建立龙门吊配置多目标混合整数规划模型和采用全场调度模式的集卡配置多目标混合整数规划模型。设计多目标遗传算法对两个模型进行求解,并通过多个算例对所设计的模型与算法的可行性与有效性进行验证。(2)提出一种集装箱码头机械配置的仿真优化方法。采用排队网络模型对集装箱装卸作业流程进行定性描述,并采用计算机仿真方法得出系统的统计性能分析结果。并在分析岸桥作业效率、龙门吊作业效率、集装箱堆存地点分布情况、双40英尺岸桥双箱作业率、进出口作业量均衡程度等影响机械配置方案的多个因素的基础上,引入试验设计方法,得出在不同因素水平下的最优龙门吊和集卡数量,再通过回归分析方法对龙门吊和集卡数量配比与这些因素间的相关性进行定量研究。(3)建立龙门吊在堆场不同箱区之间进行分配的一种多目标混合整数规划模型,设计一种基于网络流的启发式求解算法。根据各箱区内的集装箱堆存情况和岸桥的作业计划对单台龙门吊在其负责的箱区内的作业路径进行优化,建立以最小化龙门吊移动时间为目标的整数规划模型,综合运用枚举法、表上作业法和动态规划方法设计一种求解算法,采用该算法可得出所求问题的全局最优解。(4)采用滚动调度法建立一种综合考虑新型岸桥作业特点的集卡全场调度模型,设计基于时间和距离加权和的启发式算法和改进遗传算法,其中改进遗传算法的求解质量较高。针对集卡调度动态信息的随机性和模糊性,应用强化学习方法研究集卡的调度策略。采用Q学习算法进行集卡调度策略优化,采用小脑模型关节控制器进行Q函数的泛化和逼近,通过计算机仿真验证Q学习算法解决集卡调度问题的有效性。本文在研究过程中综合运用多目标决策理论、多种运筹学经典算法、遗传算法、启发式算法、计算机仿真、强化学习等多种理论与方法,对集装箱码头生产调度优化问题的研究具有一定的理论指导意义。利用本文研究方法和结论,可以根据生产数据库中的大量数据产生调度方案,为集装箱码头生产调度策略的制定提供科学的手段,具有较大的实际应用价值。更多还原

【Abstract】 An effective way of getting advantage in the competition among container terminals is optimizing the configuration and scheduling of terminals’resources. The turnaround time of container ships and the throughput of container terminals depend on the efficiency of quay cranes, while reasonable configuration and scheduling of yard handling machines and horizontal transport machines directly affects the efficiency of quay cranes. During the operating process of terminals, operators often rely on their experiences to configure, dispatch and route the handling facilities, and the research on terminals’ operation management which involve new equipments such as dual-trolley quay crane and twin 40-ft quay crane is rare. Therefore, this thesis studies the configuration and scheduling strategies of equipments in container terminal on the basis of integrating relevant scheduling optimization theories and technologies and considering the working characteristics of new type quay cranes. The main contents and results are as follows.(1)Aiming at the configuration of quay crane, rubber-tyred gantry crane (RTG) and container truck, it is divided into two phases to build multi-objective mixed integer programming models of RTGs and container trucks adopting the dual-cycle dispatching mode. An improved multi-objective genetic algorithm is designed to solving the two models. The viability and effectiveness of the models and algorithm are tested by several examples.(2)A simulation optimization method of equipment configuration in container terminal is presented. Queuing network model is used to qualitative describe the operation process of container terminal. Computer simulation is used to obtain the system’s statistics performances. Many factors affecting the schemes of equipment configuration are analyzed, such as quay crane’s working efficiency, RTG’s working efficiency, container’s stacking location distribution, ratio of handling two 40-ft containers simultaneously for twin 40-ft quay crane, balanced degree of import and export container, and so on. Introducing experimental design method, the optimized quantities of RTGs and container trucks at different levels of factors are obtained. Quantitative research on the relevance of these factors and the quantities of RTGs and container trucks is carried on through regression analysis.(3)Multi-objective mixed integer programming models to dispatch RTGs among different yard blocks is build and a new heuristic algorithm based on network flow is designed. Routing optimization of one RTG taking charge of container handling in one block is deliberated according to container’s stacking plan and quay crane’s working plan. An integer programming model aiming at minimizing RTGs’moving time is build and an algorithm applying enumeration method, table dispatching method and dynamic programming is proposed, which can reach the global optimum.(4)A dual-cycle container truck dispatching model is established, which utilizes rolling scheduling method and takes the working characteristics of new type quay cranes into consider. An improved genetic algorithm and a heuristic algorithm based on weighted sum of time and distance are designed, in which the improved genetic algorithm has better solution quality and efficiency. Because of the randomness and fuzziness of container truck’s dynamic scheduling information, reinforcement learning method is used to solve the dispatching problem of container trucks. A reinforcement learning model based on Q-learning was designed. The Q-function was generalized and approximated using the cerebella-model-articulation-controller (CMAC) neuron network. Simulation results showed that Q-learning algorithm could optimize the truck scheduling strategies efficiently.Many methods, such as multi-objective decision theory, claddical algorithms, genetic algorithm, heuristic algorithm, computer simulation and reinforcement learning, are applied in the research on scheduling optimization in container terminal. Using the methods and results in this thesis and mass data in the production database, scheduling schemes are created to provide scientific approaches for decision-making in container terminal. This thesis has particular theoretical significance and practical application value for the research on production scheduling optimization in container terminal.更多还原