摘要
The simulation of fracture distribution is an important problem in various fields of geosciences. To simulate the spatial distribution of fracture networks, this paper proposes a geostatistical method in consideration of their directions (strikes and dips). Fracture locations are generated randomly, following fracture density values assigned by sequential Gaussian simulation method. Fracture direction is divided into n equal (or unequal) groups, and sample fracture directions are assigned to its corresponding group. Then sample fracture directions are transformed into indicators consisting of n binary variables, where 1 and 0 represents belonging to and not belonging to this group. For calculation convenience, the indicator number is reduced by using the principal component analysis. Then ordinary kriging is employed to estimate the distributions of these principal components. The results are inversed to the original indicator form, and the biggest one is assigned as 1 while the others areassigned as 0. Fracture directions are generated randomly by using the cumulative distribution function of the biggest group. Based on these simulated results, fracture elements can be determined with location and direction. At last, fracture elements within the angle and distance tolerances are connected to be one fracture. The case study of the fracture data in Gejiu dolomite of southeast Yunnan Province shows that the combination of sequential Gaussian simulation, ordinary kriging and principal component analysis can provide a reasonable simulation result for locations and directions of fractures.