An improved K-means clustering method for the automatic grouping of discontinuity sets

Abstract

An improved K-means dynamic clustering method has been proposed in this paper for identifying and delineating groups (sets) of discontinuity data. Although the number of the desired clusters has to be defined in advance and is somewhat subjective, this drawback can be overcome by an improved classical algorithm to refine the orientation of the initial cluster centers. An iterative and automatic procedure for cluster center estimation is developed, which is called by the main program to alleviate the influence of initial cluster center over the results. Besides, we have suggested using the acute inclination angle between any discontinuity surfaces as a distance measurement. Since the method is frilly automatic, it is not necessary to construct density contour plots and no free parameters are required to be optimized. The method can be easily implemented on the MATLAB platform and readily used in any stages of geological survey. A case study which demonstrates the effectiveness of the dynamic clustering method is presented in this paper. Copyright 2010 ARMA, American Rock Mechanics Association.