Abstract: The large-scale projects always have an enormous number of data and tasks, and that makes the project manager easier to unconsciously ignore some important information. The traditional buffer sizing methods adopted by the project manager only consider the information of the activities on the critical chain, which usually cause an inaccurate buffer size for lacking of information of non-critical activities. What is more, the project manager and workers always have a game on striving for the report time, which results in large amount of time waste. Hence, a new buffer sizing method is proposed by considering the schedule fractal dimension, the project link activities and the correlation between different levels of activity chains.Firstly,the complexity of projects and the correlation of the different level of activity paths are described based on the Fractal Dimension theory. Secondly, considering the project manager's decision-making power and risk adverse index, a balanced time-cut position between the manager and workers is identified based on the game theory. Finally, the aforementioned information is incorporated to the logistic growth model and the buffer size of each dimension chain is calculated.One of the advantages of the logistic growth model is that it sets anupper limit for the buffer size, overcoming the problem of liner increasing as the scale of the project grows and the number of activities increases. The improved project buffer size is reasonably calculated by the growth function $f\left(C \right)=\frac{K}{{1 + \left({\frac{K}{{{f_0}}}-1} \right){e^{-pC}}}}$, where F(C) represents the needed buffer, f₀ denotes the initial buffer size, K is the most optimistic buffer size, p represents the speed of buffer increasing and the C is the schedule fractal dimension.
To validate the proposed method, the management processis simulated by Matlab software and compare the experiment results with three classic methods(C&PM, RSEM and APD). The results indicate that the proposed method can effectively shorten the project duration (4.7%) and reduce the total cost of the project (34.9%).Our research has filled gaps in the critical chain field by determining the buffer with an in-depth analysis of the schedule fractal dimension of the project network. In practice,the model constructed in this study can be revised and improved to provide a decision support in solving complicated problems about buffer decision inlink information missing and multi-loop network projects, making a contribution to the studies of critical chain project management (CCPM) by providing a more reasonable buffer size.

Key words: project management, critical chain, correlation dimension, buffer sizing, chain