基于谱聚类算法的岩体结构面产状优势分组

结构面分组是开展岩体工程稳定性分析的基础，为此，采用谱聚类算法根据岩体结构面产状信息将结构面进行优势组划分。与目前广泛使用的K均值聚类相比，该算法能够收敛到全局最优。选取结构面法向量所夹锐角的正弦值平方作为结构面间的相似度量准则，应用谱聚类算法进行优化求解；同时，引入Silhouette指标对聚类有效性进行评价，以确定最佳分类数目。利用谱聚类方法对人工生成结构面数据进行计算的结果验证了该方法的可靠性。最后，将该算法应用于三山岛金矿岩体结构面的优势组划分，取得了理想的分类效果，为进一步岩体稳定性分析提供了可靠的数据基础。

Dominant Grouping of Rock Mass Structural Space Orientation Based on Spectral Clustering Algorithm

The identification of rock discontinuity sets is the foundation of stability analysis of rock masses in engineering practices.Therefore，a spectral clustering algorithm was used in this paper to classify the discontinuity data sets based on discontinuity orientations.Compared with the K-means clustering algorithm widely used at present，this algorithm can converge to the global optimum.The square value of the sine of acute angle between normal unit vectors of structural planes was selected as similarity criterion and then the spectral method was used to group discontinuity data.Meanwhile，the Silhouette index was introduced to evaluate the clustering validity to determine the optimal clustering number.The test of artificial discontinuity data shows that the method is reliable. Finally，the algorithm was applied to group the discontinuity data in Sanshandao gold mine.The results show that the algorithm can achieve ideal classification results，which provides reliable data base for further analysis of rock mass stability.