基于GA和FCM的岩体结构面的混合聚类方法

提出了一种基于遗传算法(GA)和模糊C均值(FCM)算法的岩体结构面混合聚类方法.利用GA的全局搜索性能,求得初始聚类中心;在此基础上利用FCM算法,根据精度要求再作进一步求解.该方法避免了人为划定分类界限的主观性,消除了FCM聚类算法的局部最优的弱点,解决了采用普通遗传算法聚类时搜索速度和聚类精度的矛盾.结合实测数据,对应用该方法进行结构面组识别的步骤、参数选取、分组有效性、优势方位的判定进行了分析和讨论.

Hybrid Cluster Analysis Method Based on GA and FCM for Automatically Identifying Joint Sets

A hybrid cluster analysis method based on genetic algorithm (GA) and fuzzy C-means (FCM) algorithm is introduced for the automatic identification of joint sets. The initial cluster centers for FCM are obtained by GA, and then the optimal cluster results can be calculated by FCM on the basis of the work in the first stage. This method eliminates the local optimality disadvantage of FCM and the subjectivity of traditional methods such as pole and contour plots for classifying joints into sets and resolves the conflict between search speed and cluster precision by general GA. The analysis steps, parameters selection, cluster validity and dominant direction determination for identification of joints sets using the hybrid cluster analysis method are discussed based on joint survey data sets.