The dominant grouping of the discontinuities in a rock mass is crucial to investigating the rock structure and stability. The traditional grouping methods are mostly based on the discontinuity orientation. However, other discontinuity parameters nonnegligibly impact the properties of a rock mass. If two discontinuities have the same orientation but differ otherwise, their mechanical and hydraulic properties would differ. In the present study, orientation, trace length, opening degree, and undulation of discontinuities were used to develop a multi-parameter dominant discontinuity grouping method that utilizes an improved ISODATA algorithm. The developed method uses several indicators, such as the weighted Euclidean distance and standard deviation, to iteratively update the clustering centers of the discontinuities. A clustering validity index was introduced for assessment and optimization of the discontinuity grouping results, and an adaptive grouping model that considers the weight of each grouping parameter was developed. By grouping discontinuities generated by Monte Carlo stochastic simulation, defects existing in the grouping based on the orientation only were illustrated, and the rationality of the established adaptive grouping model was verified. The engineering practicability of the method was further verified by using it to group discontinuities measured in the Dongsheng Quarry in Jingyuetan, Changchun, China.
A New Clustering Validity Index based on K-means Algorithm
