A new iteration clustering method for rock discontinuity sets considering discontinuity trace lengths and orientations

Abstract

Discontinuity clustering is an important and preliminary step in estimating discontinuity properties. To group discontinuities quickly and accurately, in this study, a new iteration clustering method (NICM) based on the data of discontinuity orientations and trace lengths is proposed, with its corresponding algorithm code developed in MATLAB. The method assumes that the trace lengths of the discontinuities obey a specific distribution, through a redistribution of the discontinuities whose membership degree values are unclear to adjust the probability density function of the trace lengths to make them more closely obey the specific distribution. The membership degree value calculation and cluster code determination are two main core modules of NICM. The characteristics of the method were analyzed with three types of numerical models in which the data are synthetic and the clustering results can be clearly grasped. In contrast, the NICM and fuzzy C-means method were applied to group the discontinuities of a rock slope at the Changhe Dam located in Sichuan Province, southwestern China, at the same time. The results clearly demonstrate that (a) a 10% enhancement in the precision can be expected by considering the distribution of the trace lengths; (b) the assumed specific distribution of the trace lengths does not have an obvious effect on the precision improvement; and (c) the problem of hard boundaries can be effectively alleviated by using the method.