文献类型：期刊文献

中文题名：基于优化的K均值算法对岩体结构面产状聚类分析

英文题名：Cluster Analysis of Rock Mass Discontinuity Based on Optimized K-Means Algorithm

作者：沙鹏[1];倪吉吉[1];汪自学[1];黄曼[1];黄永亮[2]

机构：[1]绍兴文理学院土木工程学院,浙江绍兴312000;[2]浙江有色地质环境研究院,浙江绍兴312000

年份：2023

卷号：43

期号：2

起止页码：16

中文期刊名：绍兴文理学院学报

外文期刊名：Journal of Shaoxing University

收录：国家哲学社会科学学术期刊数据库

基金：浙江省重点研发计划项目“交通基础设施全生命周期地质灾害预警与防治技术装备研发”(2020C03092);浙江省基础公益研究计划项目“岩质边坡灾害智能识别与动态预警方法研究”(HZ21D020001).

语种：中文

中文关键词：K均值聚类;结构面;产状;岩体力学;最小误差平方和

外文关键词：K-means clustering;structural plane;occurrence;rock mass mechanics;minimum sum of squared errors

中文摘要：为实现对岩体结构面子集的智能划分,采用改进的K均值算法,将代表结构面三维坐标点通过等面积投影到单位圆平面,用二维点划分子集,并在初始聚类中心选择阶段,比较每一个数据点作为中心时的误差平方和,选取误差平方和最小或能最大限度减小误差平方和的点作为初始聚类中心,从而避免了K均值对初始聚类中心敏感的问题.通过对人工模拟结构面数据进行聚类,证明了该法的可靠性.将该方法应用到实测结构面数据的聚类分析中,结构面被分成两组,每组的结构面个数为132和121,对应的优势产状为268.16°∠48.61°和135.13°∠62.00°.

外文摘要：In order to divide the rock mass structural planes into subsets intelligently,the paper adopted an improved K-means algorithm to project the three-dimensional coordinate points of the representative structural planes onto the unit circle plane through equal area,divide the subsets with two-dimensional points,and place them at the initial cluster center.In the selection stage,the sum of squares of error was compared when each data point was used as the center,and the point with the smallest sum of squares or the point that can minimize the sum of squares as the initial clustering center was selected,with the K-means sensitive to the initial clustering center problem avoided.The reliability of the method is proved by clustering the artificially simulated structural plane data.This method is applied to the cluster analysis of the measured structural plane data.The structural planes are divided into two groups,the number of structural planes in each group being 132 and 121,and the corresponding average occurrences 268.16°∠48.61°and 135.13°∠62.00°.