【作者】 郭方铭；

【导师】 王乘；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2010， 博士

【摘要】 本文首先在对传统项目进度计划方法进行阐述的基础上,分析其存在的主要优缺点,通过与传统的项目计划管理方法相对比,突出关键链项目管理方法能有效地管理项目中的不确定性,缩短项目周期,提高项目效率。由于关键链方法充分考虑了人的主观行为因素,该方法较之传统方法更具实用性。在关键链技术研究中,主要方向是关键链的识别及缓冲区管理。考虑到现有的多数关键链识别方法缺乏对于项目活动工期随机性的计算,提出了一种基于统计学理论的关键链识别方法,能够较好的控制项目工期及兼顾项目计划的实用性。而现有的缓冲区大小确定方法未能充分考虑到项目中各个活动自身的特性,为此提出了一种自适应的缓冲设置方法。通过试验证明,该方法能够有效地缩短项目中的汇入缓冲大小,并能有效避免因为缓冲设置产生新的资源冲突。在以上研究的基础上,对多项目计划管理的约束情况进行了研究和分析,将关键链技术引入到多项目计划与进度管理中,建立了基于关键链技术的多项目计划调度模型,并提出了相应的目标函数。为求解该目标函数,在微粒群算法的基础之上,设计了一种混合遗传操作的微粒群算法,采用了一种新的微粒编码方式。该编码方式采用随机优先权和延迟时间作为粒子的基因,每个基因的随机性保证了初始种群可以在可行解空间内均匀分布,而且该基因携带的遗传信息,可以保证在后续的算法过程中可以找到能够使目标函数最短的子项目优先值并遗传下去。在每次迭代之后,优秀个体将存入记忆库,同时随机产生新个体加入到新种群中,这部分新产生的个体在保持群体多样性的同时,也降低了算法过早收敛的可能性,另一方面又利用了记忆库信息,维持了种群的整体质量。为测试该算法的有效性,提出了一种多项目实例生成方法,该方法从标准的PSPLIB库中选用已有的单项目实例,按照给定的参数,生成所需的多项目实例。再对生成的多项目实例采用本文的算法进行仿真计算,通过对仿真计算结果的对比分析,说明了算法的有效性。最后,结合某大型空调生产企业的计划调度问题进行实例应用,针对该空调生产企业项目计划管理中的主要问题,把本文提出的多项目计划方法应用到该企业的实际生产调度过程中,实践表明,该方法可有效提高企业生产计划的执行效率。更多还原

【Abstract】 This paper, based on elaborating the traditional project schedule planning methods, analyzes their main advantages and disadvantages, and highlights the significant advantages of critical chain project management method in contrast with traditional project management methods on many aspects, which especially are that the critical chain method can effectively manage the uncertainty of project, shorten project cycle and improve project efficiency. The critical chain method, fully considering the factor of people’s subjective behaviors, is therefore more practical than traditional methods.Critical chain identification is an important basis for critical chain project management in the study of critical chain technologies. However, the majority of existing critical chain identification methods doesn’t take the randomness of project activities duration into account effectively. Therefore, this paper proposes a critical chain identification method on basis of statistics theory. Experiments show that this method works well and can control the project duration while taking into account the project practicality.Buffer management is an effective way among the critical chain technologies to cope with uncertainty, but the existing buffer size determination methods don’t fully consider the characteristics of various activities in project. Thus this paper proposes a self-adaptive method for setting buffer. Experiments show that this method can effectively reduce the buffer size imported in the project and can effectively avoids new resource conflicts generated by setting buffer.On the basis of the above, this paper conducts a research and analysis of limit situation of multi-project schedule management, introduces critical chain theory to multi-project planning and schedule management, then establishes a critical chain method-based multi-project planning and scheduling model and at last proposes the corresponding objective function.To solve the objective function, this paper, based on Particle Swarm Optimization (PSO), designs a PSO algorithm mixed with genetic manipulation, which uses a new particle coding mode. This coding mode takes random priority and delay time as gene particles. The randomness of each gene ensures initial population a homogeneous distribution within feasible solution space and the genetic information this gene carries ensures that sub-project priority value, which makes the objective function the shortest, can be found and be hereditary in the subsequent algorithm process. After each iteration, outstanding individuals will be deposited into the memory, while new individuals will be generated randomly and become a member of new populations. On one hand, these newly-generated individuals can maintain population diversity and at the same time reduce the possibility of premature convergence of the algorithm; on the other hand, the memory information can be used to maintain the overall quality of population.To test the effectiveness of the algorithm, this paper puts forward a multi-project instances generation method, which selects existing single-project instances from the standard PSPLIB to generate necessary multi-project instances in accordance with the given parameters. Then, this paper uses the algorithm mentioned above to conduct a simulating calculation for these generated multi-project instances and further illustrates the effectiveness of the algorithm by analyzing and comparing the simulating calculation result.At last, this paper puts forward the major problems in the project planning management of a large air conditioner manufacturer, combined with analysis of planning and scheduling problems in this manufacturer. At the same time, the proposed multi-project schedule management method is applied to the actual production process of the manufacturer, which obtains a good result.更多还原