考虑重叠并行性的项目间知识转移对项目集聚类的影响研究

The Impact of Knowledge Transfer among Overlapped Projects on the Program Clustering

项目并行执行和知识转移是多项目管理的重要特征,在重叠并行过程中,一个项目积累的知识和经验会转移到其他项目,由此可以提升整个项目集管理的绩效。本文针对研发项目的特点,从重叠并行性的视角分析知识在项目间的转移和学习过程。首先,分析了重叠并行情况下的项目间知识转移与知识获取过程,然后,采用依赖结构矩阵(DSM)方法描述重叠对知识转移的影响,构建了项目的知识成熟度和知识获取能力模型。在此基础上,建立项目间因知识转移而产生的依赖关系强度模型,并量化分析了由于学习效应而导致的项目时间压缩的相对量。进一步,本文采用两阶段聚类算法,构建最大化“增加的平均类内外压缩时间比”聚类目标函数。最后,通过算例验证本文构建的模型和方法,计算结果表明,聚类后的项目集可以增加类内项目间的依赖关系强度,缩短项目集的总持续时间并降低总协调成本,提高项目集管理的绩效。

Concurrent execution and knowledge transfer among projects are fundamental characteristics of multi-project management,especially for the new product development(NPD)project.The accumulated knowledge and experience can be transferred from one project to other projects in the overlapping and concurrent process,thus improving the performance of the program.However,existing research on multi-project management don't take the influence of knowledge transfer and learning among projects into account,as well as the clustering analysis for multi-project.Using the dependency structure matrix(DSM),knowledge transfer theory and two-stage clustering algorithm,a new multi-project management method is proposed.It attempts to solve three key problems:1)how to measure the knowledge maturity and acquirement model of projects;2)how to analyze the interaction relationship among projects through knowledge transfer and learning;3)how to cluster projects into programs via the improved two-stage DSM clustering algorithm.Hence,the knowledge transfer and learn-ing process among projects is analyzed from the perspective of overlapping which is the basic characteristic of NPD projects.First,the knowledge transfer and learning among projects is analyzed in the overlapping and concurrent process.Then,the knowledge maturity model is built for transferring knowledge projects and knowledge acquirement model for receiving knowledge projects in the concurrent process using the DSM method.Next,the interdependency strength models resulted from knowledge transfer among projects is established.Additionally,by taking account the"knowledge transfer-learning"process,quantitatively measure model for the shorten time is built using the improved learning curve model.Further,a two-stage program DSM clustering algorithm with maximizing"Added Internal and External Average Crashed Time Ration"criterion is defined as an objective function.Finally,an industrial example is provided to illustrate the proposed model and method.The results indicate that the clustered program can enhance the dependency strength and shorten the time of the program,reducing the total coordination costs,and significantly improve the performance of program management.