【作者】 朱卫锋；

【导师】 费奇；

【作者基本信息】 华中科技大学 ， 系统工程， 2004， 博士

【摘要】 一个理想的物流问题是一个网络流问题,它由地理上分布的许多实体组成。在物资调运领域多年的信息系统开发经历,以及对系统复杂性的关注,引发了我们对更为一般的物流系统的思考,并最终提出了 “复杂物流系统”的概念。复杂物流系统是由若干供方、需方等实体构成的供应链物流系统。它通常处在一个不确定的环境中,包括顾客需求、外部供应、交纳周期等,受很多随机因素的影响,具有多目标、多因素、多层次的特点,是以离散事件系统为主的复杂的连续-离散事件混合系统,而不是单纯的离散事件系统。该类系统的复杂性包括结构复杂性、不确定性、信息的失真与放大、组织机构的动态性等,具有分布性、自治性、交互性、动态性等特征。库存控制问题是复杂物流系统管理的重要内容,其优化目标包括成本优化和时间优化。采用何种库存控制策略应对复杂物流系统的众多不确定性,是库存管理工作的重点和难点问题。系统仿真方法是物流系统研究的一种常用理论和方法。相对解析法而言,它是一种间接的系统优化方法。因此,研究如何构造具有决策支持和仿真演练功能的复杂物流系统仿真工具,具有重要的理论意义和应用价值。本文的重点在于建立一套比较完整的关于复杂物流系统建模与仿真的理论和方法体系,包括复杂物流系统的描述方法、仿真体系结构、库存模型构造、仿真模型构造等主要内容。具体来说,包括下面一些内容:应用有色Petri网建模方法,建立了主体BDI(信念Belief、目标Desire、意图Intention)的Petri网模型;在此基础上,结合现有的供应链BDSPN(Batch Deterministic and Stochastic Petri Nets,批量确定随机Petri网)模型,分别建立了库存管理策略和复杂物流系统的BDI-BDSPN模型。基于HLA/RTI(High Level Architecture/Run Time Infrastructure),提出了一种分布式多代理仿真体系结构,并用有色Petri网描述分布式仿真系统的运行过程以及内部结构关系;结合一般的集中式仿真体系结构,提出了一种集成的复杂物流系统仿真体系结构,探讨了集中式仿真与分布式仿真之间的动态衔接和信息交互关系。运用模糊集合理论建立了顾客需求模糊情况下的复杂物流系统模糊分析模型,设计了该模型的计算机求解算法。该模型用来决定组成复杂物流系统的各个实体所采用的最佳订货点量;在总结了不确定性程度的度量方法后,提出了保管缺货率的概念,并分析了保管缺货率对订货点量决策的影响规律。 <WP=5>建立了复杂物流系统仿真模型,给出了该模型的计算机实现算法,该模型用来模拟复杂物流系统的行为与性能,在仿真策略方面提出了基于时间步进的事件调度法仿真策略,仿真模型则主张在复杂物流系统Petri网模型的基础上采用实体流图法构造;在此基础上,运用我们开发的复杂物流系统集中式仿真工具CLCSim,研究了顾客需求不确定性对复杂物流系统行为与性能的影响规律,并探讨如何运用CLCSim进行订货点量的仿真调整实验。以敏捷后勤仿真系统的构造为例,探讨了复杂物流系统建模与仿真理论在具体环境中的应用问题。说明了该理论的有效性和重要应用前景。复杂物流系统建模与仿真理论研究涉及建模方法、仿真技术、数据库技术、决策理论、软件工程等多个领域的内容。其进一步的完善和发展也还存在很多工作,如在分布式仿真系统构造与实现方面,可以作进一步的探索。但总体来说,该理论可以对大型水电工程、灾难救助、国民经济动员等众多领域中的物资调运仿真演练系统构造具有重要的借鉴意义,其应用前景是广阔的。更多还原

【Abstract】 An ideal logistics problem is considered as a network flow problem consisting of many geographically distributed entities. Many years’ development experiences on information systems in materials dispatching domain and continuous considerations in complexity foment us thinking about the more general logistic systems. As a result, the concept of Complex Logistics (CL) is proposed. CL is a supply chain logistics consisting of several supplier entities, demand entities and so on. It is always in an uncertain environment, including customer demand, external supply, lead time and so on, influenced by many random parameters, having multi objectives, factors and levels. It is a complex mixed system which has both characteristics of continuous systems and those of discrete event systems. The complexities of CL comprise of structure complexity, uncertainty, information contortion and amplification, dynamic organization and so on, with distributional, autonomous, interactive and dynamic features.Inventory control is the main problem of CL. The goal of CL optimization is cost optimization and time optimization. Which inventory control policy should be adopted to deal with uncertainty in CL is an important and difficult problem to inventory managers. System simulation is a familiar theory and method for logistics study. Comparing with resolution method, it is an indirect system optimization method. Therefore, the research on how to construct a CL simulation tool having such functions as decision supporting, simulating, and training has very important theoretical significance and practical value. The emphasis of this paper is to build a complete methodology of CL modeling and simulation, including system depicting methods, simulation architecture, inventory models, simulation models construction and so on. Specifically, this work is outlined as follows: Using Colored Petri-Net, the Petri-Net model of agent’s BDI (Belief, Desire and Intention) is built. Combining BDSPN (Batch Deterministic and Stochastic Petri Nets) an existing Petri-Net modeling method for supply chains, the Petri-Net models of inventory management policy and CL, called BDI-BDSPN, are constructed. An architecture for distributed multi-agent based simulation is proposed based on HLA/RTI(High Level Architecture/Run Time Infrastructure), where Colored Petri-Net is used to depict the run process and internal structural relationship of distributed simulation systems. Combining the general centralized simulation architecture, an integrated CL simulation architecture is proposed, where the dynamic connection and information interaction between centralized simulation and distributed simulation is discussed.CL fuzzy analytical model when customer demands are fuzzy is built using fuzzy sets and the relative computer solving arithmetic is designed. CL fuzzy analytical model is <WP=7>used to determine the optimal order-up-to levels of entities in CL. After summarizing uncertainty measurement methods, the concept of holding-shortage coefficient is proposed. Then how this holding-shortage coefficient affecting order-up-to level decision is also analyzed. CL simulation model is built and the relative computer implementation arithmetic is given too. The model is used to simulate CL behavior and performance in an uncertain environment. The simulation policy of CL simulation is based on Event Scheduling (ES) where time advances equally, i.e. simulation advances in equal increments. Entity Flow Diagram is used to erect CL simulation model based on CL Petri-Net model built before. After that, CLCSim, a centralized simulation tool for CL newly developed by us, is used to research on how customer demand uncertainty affecting CL behavior and performance. After that, how to simulate and adjust the order-up-to levels of entities in CL using CLCSim with experiments is also described.According to the methodology of CL modeling and simulation, an agile logistics system is constructed as a case. In addition, it illustrates the methodology is effective and applicable.The met更多还原