Abstract: Fractures and their combinations (i.e.,rock mass structures) are key factors influencing the integrity, permeability, physical and mechanical properties, and stability of rock masses. Accurate characterization of rock mass structure is therefore essential for understanding the engineering behavior of rock masses. The 3D fracture network model, capable of representing the spatial distribution of fractures inside a rock mass accurately and comprehensively, has become a widely used method for characterizing rock mass structures. However, accurately modeling fractures faces two major challenges: first, it is difficult to measure fracture diameter directly, as theoretical assumptions and trace length corrections introduce errors in estimating this parameter; second, error propagation leads to low accuracy in estimating the three-dimensional fracture density. To address these issues, this paper proposes a 3D fracture network modeling method for rock masses based on the flower pollination optimization algorithm. By constructing a suitable objective function, the method iteratively optimizes the 3D fracture network model to obtain optimal solutions for fracture geometric parameters, enabling refined characterization of rock mass structure. To validate the effectiveness of the proposed method, a simulation case was designed with fracture diameters following log-normal and negative exponential distributions. The computed fracture parameters were compared with initially set values and with results from Zhang's method,demonstrating that the relative error of the proposed method is significantly smaller. Finally, the method was applied to the Three Gorges Project to construct a fracture network model for the underground powerhouse. This approach allows for detailed characterization of rock mass structure and provides a foundation for further research on the mechanical properties, permeability, and stability of engineering rock masses.