Abstract: Today’s project management technology has become increasingly mature, and various project management software also provides a more reliable analysis tool for project management. However, even so, project delays and overruns are still common. Based on this, a critical chain project management method that considers project resource constraints and decision-maker behavior characteristics is proposed, and buffer is the core concept of critical chain project management. By extracting the safe time of each activity and concentrating on it at the end of the chain to form a buffer, it can effectively avoid the student syndrome and Parkinson’s law of project members, and reduce the waste of safety time caused by the influence of factors such as multitasking and resource constraints. Effective monitoring of the buffer area in the project network can absorb various uncertainties in the project execution process, shorten the project duration under the premise of ensuring the probability of project completion, and realize the overall risk sharing of the project. However, the traditional buffer monitoring method mainly focuses on the project buffer consumption of the critical chain, ignoring the subsequent consumption information of the parallel feeding buffers, but the management efficiency and on-time completion rate of the project schedule are still difficult to guarantee.
From the perspective of buffer management, this paper proposes a project mixed buffer dynamic monitoring method based on parallel-link trend prediction, so as to comprehensively and systematically control the project execution process. Firstly, the gray neural network monitoring and forecasting model of project buffer and feeding buffer is established by referring to the idea of gray neural network, and the follow-up trend information is quantitatively predicted according to the buffer data of parallel critical chain and non-critical chains. Secondly, considering the relationship between the actual remaining buffer and the predicted buffer consumption and the different characteristics of project buffer and feeding buffer, the monitoring trigger granularity of the two types of buffers is set differently to form a mixed buffer dynamic monitoring system based on parallel link buffer prediction. Finally, the validity of the proposed method is verified by Monte Carlo simulation experiments. The results show that the duration and cost of this method are lower than those of the traditional buffer monitoring methods, and that this method also reduces the fluctuation of the project duration during the execution process and achieves the optimization of both the project duration and cost while ensuring the completion probability.
According to the results of the simulation experiment, it can be concluded that the buffer monitoring method proposed in this paper takes into account the actual implementation of the critical chain and the non-critical chain of the project, it not only pays attention to the project buffer information, but also considers the feeding buffer information, and adopts different monitoring trigger granularity according to the characteristics of different buffers, allowing decision makers to monitor the project chain in a differentiated and reasonable manner while grasping the development trend of the project. It can overcome the false early warning caused by one-sided project buffer monitoring, and make full use of the future buffer development trend of critical chains and parallel non-critical chains information. Therefore, the proposed method can provide managers with more comprehensive project decision-making information, which is beneficial for project managers to actively and differentially control the different buffers in the project execution process, and reduce unnecessary construction delays and resource consumption.
The focus of the next research is to consider the introduction of resource early warning mechanism, the different impacts of bottleneck resources and non-bottleneck resources on buffer consumption, and combine resource buffer early warning with mixed buffer monitoring of parallel links to build a resource-buffer integrated schedule control system. In addition, considering the information flow of different activities of the project will also affect the function of resource early warning and buffer monitoring, so the relevant methods to describe the information flow of the project network and the consideration of the influence of information factors in the process of resource early warning and buffer monitoring could also be the next research direction.

Key words: critical chain project management, parallel-link, trend prediction, mixed buffer monitoring, monitoring granularity

摘要： 针对现有缓冲监控方法只关注关键链的项目缓冲消耗而忽略并行链路接驳缓冲消耗及其趋势信息的不足,提出了一种基于并行链路趋势预测的项目混合缓冲动态监控方法,以对项目的执行过程进行全面而系统的控制。首先,建立项目缓冲和接驳缓冲的灰色神经网络监控预测模型,根据并行关键链和非关键链的缓冲数据对后续趋势信息进行定量预测。其次,综合考虑实际剩余缓冲与预测缓冲消耗之间的关系以及项目缓冲和接驳缓冲的不同特点,对两种类型缓冲的监控触发粒度进行了差别设置,形成基于并行链路缓冲预测的混合缓冲动态监控体系。最后,通过蒙特卡洛模拟实验对本文的模型体系进行验证。实验结果表明本文方法产生的工期和成本均低于传统的缓冲监控方法,同时本文方法还减少了项目在执行过程中工期的波动,在保证完工概率的前提下实现了项目工期与成本的双重优化。

关键词: 关键链项目管理, 并行链路, 趋势预测, 混合缓冲监控, 监控粒度