Abstract: Whether project buffer should be centralized or decentralized regarding project characteristic isa core issue of project buffer management. Classic buffer management researchtries to get more optimized results by increasing the monitoring frequency or monitoring difficulty of the project, while ignores research on whether the project buffer should be centralized at the end of the project or distributed at the end of each activity. The existing buffer monitoring methods mainly focus on the centralized buffer management, that is, the buffer is centralized in the tail and then monitored. In addition, the activity flexibility reflects the ability of a project activity to start in time after the completion of its pre activity. Previous studies usually use percentage to describe this ability. However, this description is often difficult to be quantified. For example, for a 10 day activity, its flexibility to start is 50%. In the actual project planning and monitoring process, this flexibility to start expressed in probability is not easy to be operated. The proposed method therefore studies and compares these two buffer allocation models considering the activity flexibility, and discusses the conditions that need to be met for 100% starting flexibility. The simulation results demonstrate that the centralized buffer management can better grasp the possible early completion opportunities of the project activities by making the resources in place ahead of time. It indicates that the project duration is reduced by 8.67%, while the total cost is only increased by 2.02%, implying that the centralized model can effectively reduce the project duration. However, the decentralized buffer management model is an extension of existing buffer management research. For this model, the inventory cost does not increase compared with the centralized model, considering the flexibility of the project. The experimental results show that the project can obtains the optimal buffer dispersion degree of the project when the activity completion rate falls between 50% and 60% of the probability distribution function. For the first time, this paper combines resource planning with the flexibility of project activities, redefines the concept of flexibility of project activities and makes a quantitative study on it.

Key words: critical chain project management, start-up flexibility, resources in place, centralized buffer management, decentralized buffer management