Evaluation of the structural similarity of fractured rock masses based on multiple fracture parameters

Evaluation of the structural similarity or structural domains is a fundamental step for characterizing fractured rock masses. In this research, the division of rock masses into structural domains is treated as a multi-parameter clustering problem. The fuzzy spectral clustering method is used and proved to be an effective approach for measuring the structural similarity of jointed rock masses, in which six fracture parameters (i.e., orientation, aperture, roughness, fracture intensity P₂₀, filling, and groundwater condition) are selected. The results suggest that the study area, located at the Songta dam site of China, could be grouped into three structural domains based on over 1300 fractures collected from 6 adjacent exploration tunnels. The results of the proposed method are validated by the Kolmogorov-Smirnov statistical test through quantifying the degree of similarity between fracture parameters collected from different areas. The results show that the fuzzy spectral clustering could provide reliable results which agree well with those obtained by the Kolmogorov-Smirnov test. Moreover, compared with the K-means algorithm, the new method could perform better in identifying structural domains.