基于混合聚类的岩体结构面优势分组方法

提出一种基于凝聚层次法和模糊C均值法的混合聚类法,用于对岩体结构面的优势组划分。该方法将结构面投放到在单位球面上,并使用欧式距离作为极点的相似性度量准则。先剔除结构面数据中的孤值产状,然后用凝聚层次法得到初步聚类结果,并将其作为FCM法的初始聚类中心,最后用FCM法划分优势组。通对人工生成产状样本的分组,验证了该法的正确性。将该方法应用于大藤峡坝址区实测的结构面数据的划分。在实测数据中寻找到两个孤值产状,成功将大藤峡D1y^1-3地层岩体结构面划分为两组,得到了符合实际的分组结果。

Dominant partitioning method of rock mass discontinuities based on hybrid clustering

A hybrid clusting analysis method based on agglomerative hierarchical algorithms and fuzzy C-means algorithm is introduced for partitioning discontinuities data of rock mass.The method projects the discontinuities data onto the unit sphere and uses Euclidean distance as the similarity criterion for the poles.Firstly,the method eliminates solitary occurrences in the discontinuities data,and then gets the initial clustering result by the aggregation hierarchy method,which was taken as the initial cluster centroids of FCM method,and finally the dominant group was divided by FCM method.The validity of this method is verified by grouping the artificially generated occurrence samples.Moreover,the method is applied to the analysis of the discontinuities data collected from the Datengxia dam site and two solitary occurrences are identified in the discontinuities data,and the discontinuities data of D1y^1-3 strata in Datengxia dam site are successfully divided into two groups,from which more realistic grouping result is achieved.