Analysis of fracture spatial distributions and fast reconstruction of discrete fracture networks model based on non-parametric kernel density estimation method

Fracture spatial distributions significantly impact the mechanical properties of rocks and play a vital role in subsurface flow and transmission. However, many studies generate random geometrical distributions for fractures, leading to unrealistic subsurface models. This paper describes a new method for analyzing and modeling fracture spatial distributions based on borehole and outcrop observations. The center of gravity of the fracture is used to define the fracture position, and the non-parametric kernel density estimation is used to analyze the cluster distribution of the fractures in space. Then, the Fast Fourier Transform (FFT) is utilized to rapidly determine the fracture position. Finally, the length and occurrence of fractures are generated by the Monte Carlo method, and the discrete fracture network (DFN) is reconstructed in Matlab. This approach is applied to three field examples. Numerical simulations demonstrate that the generated fracture morphology exhibits flow behavior similar to the original fracture network. Furthermore, it is found that the fractal dimension of the fracture center point (C) can characterize the spatial distribution of fractures and shows a logarithmic normal relationship with fracture permeability. This modeling method accelerates the speed of complex DFN reconstruction and improves the ability to quantify the permeability of fractured rock mass.