坝肩岩体质量LDA-KNN分类模型  

The LDA-KNN model to classify quality of rock mass of a dam abutment

作　　者：荀鹏 李娟 魏玉峰[1] 李常虎 范文东 XUN Peng;LI Juan;WEI Yufeng;LI Changhu;FAN Wendong(State Key Laboratory of Geohazard Prevention and Geoenvironment Protection,Chengdu University of Technology,Chengdu 610059,China;Sichuan Institute of Energetical and Geological Survey,Chengdu 610070,China;PowerChina Northwest Engineering Corporation Limited,Xi’an 710065,China)

机构地区：[1]地质灾害防治与地质环境保护国家重点实验室(成都理工大学),成都610059 [2]四川省能源地质调查研究所,成都610070 [3]中国电建集团西北勘测设计研究院有限公司,西安710065

出　　处：《成都理工大学学报（自然科学版）》2024年第2期281-290,302,共11页Journal of Chengdu University of Technology: Science & Technology Edition

基　　金：国家自然科学基金(42072303)。

摘　　要：工程岩体质量分级评价对工程的安全、设计、经济效益等有重要影响。针对当前岩级划分方法中存在不确定性,人为因素干扰和忽视了传统定性分级中对岩体质量评价的重要性等问题,本文通过在工程实际中搜集样本建立数据库,从工程的实际需求出发,选择岩体完整性系数(K v)、结构面间距(D)、岩石质量指标(RQD)等合适的评价指标,通过引入LDA(Linear Discriminant Analysis)降维方法和K近邻分析(K-Nearest-Neighbor,KNN)相结合的多分类模型,实现了岩体的非线性分级预测。通过定性定量相结合实现了岩体多因素,多指标的综合分级,并解决了多指标判断时信息冗余,复杂程度高的问题。与其他判别方案相比较,模型得出的结果准确率高,符合工程实际,减少了人为因素的影响,体现出较强的预测判别能力。该研究为水电站大坝坝肩处的平硐岩体质量划分提出了一种可行的预测方案。The classification and evaluation of the engineering quality of rock mass has an important impact on the safety,design,and economic benefits of construction projects.In this study,we responded to the problems posed by uncertainty,interference by human factors,and the neglect of traditional qualitative grading in prevalent models used to assess the quality of the rock mass.We collected samples of drift rock at the abutment of a hydropower dam to establish a database of rock mass,selected the spacing between structural planes(D),coefficient of integrity of the rock(K v),and rock quality index(RQD)as indices of evaluation based on an analysis of actual projects,and introduced the linear discriminant analysis-based method of dimension reduction and the K-Nearest-Neighbor multi-level classification model to develop a nonlinear model to classify rock mass.A combination of qualitative and quantitative analyses was used to comprehensively classify the rock mass based on multiple factors and indicators,and to solve the problems of the redundancy of information and the complexity of judgment based on multiple indicators.The proposed model generated more accurate results than previously developed schemes that were in line with engineering practice,reduced the influence of human factors,and exhibited a strong predictive capability.

关 键 词：岩体结构 岩体质量分级 线性降维 K近邻算法 分类模型 

分 类 号：TV223.1[水利工程—水工结构工程]