A clustering algorithm based on differential evolution for the identification of rock discontinuity sets

Rock discontinuities significantly influence the deformation as well as strength of rock masses. Identification of rock discontinuity sets is one of the fundamental issue in rock mechanics. In this study, a new clustering method is developed to automatically identify rock discontinuity sets. The method is established on account of differential evolution, which is a robust and global optimization algorithm. An improved encoding approach was used to realize the full automation of algorithm. The main parameters of the algorithm are determined by self-adaptation techniques. The performance of the new algorithm was studied using an artificial data set. The clustering results demonstrate that the new algorithm could well identify discontinuity sets. Furthermore, the new algorithm is applied to analyzing discontinuity data collected at an underground cavern site, and satisfactory result is obtained. Additional advantage is that the method is totally automatic, without selecting proper initial cluster centers and specifying the number of discontinuity sets.