关键词: 岩体结构面分组, 模糊C均值聚类, 改进粒子群算法, 聚类有效性

Abstract: Rock mass structural plane grouping is a basic work in rock mechanics analysis. Compared with the general clustering analysis, the special characteristics of rock mass structural plane grouping make it hard to apply the Euclidean distance measuring the similarity of structural planes. In this paper, the Euclidean distance is used to measure the similarity degree between structural planes by simple mathematical treatment, and the grouping of rock mass is studied using the fuzzy c-means (FCM) clustering algorithm. An efficient modified particle swarm optimization algorithm is employed to optimize the determination of FCM initial cluster center. The validity index of cluster is used to obtain the optimal cluster number. This method is verified by simulated data, the number of grouping is discussed, and the general determination method of grouping number is given. Finally, the presented grouping method is applied to group the structural planes of rock mass of a dam foundation. The results show that the grouping method for rock mass structural planes is easy to perform, which can be used as a useful method for practical engineering.

Key words: rock mass structural plane grouping, fuzzy c-means clustering, particle swarm algorithm, cluster validity