摘要
岩爆受多种条件因素影响,需要岩爆预测指标体系的完整性。取弹性变形能指数W_(et)、应力系数σ_(θ)/σ_(c)、脆性系数σ_(c)/σ_(t)、单轴抗压强度σ_(c)、单轴抗拉σ_(t)和围岩切向应力σ_(θ)等指标构建岩爆倾向性预测指标体系。采用主成分分析(PCA)法对指标降维处理、提取指标数据主要信息,得到3个线性无关的主成分输入向量。将处理后的指标作为输入向量对3种不同自组织特征映射神经网络(SOFM)模型进行训练与测试,比较3个模型的方差和竞争层神经元聚类效果,得到输出层神经元个数为16的最优预测模型。最后引入12组国内外工程实例与多维正态云模型、模糊综合评价法、灰类白化权函数聚类法、Russense判据所预测的岩爆烈度等级进行比较,研究表明:基于PCA-SOFM岩爆烈度等级预测结果与工程实际情况吻合度高,可行性较好,为岩爆烈度等级预测提供一种新的研究方法。
Rockburst is affected by many conditional factors,requiring the integrity of rockburst prediction index system.Prediction index system of rockburst intensity grade is constructed by selecting the elastic deformation energy index W_(et),stress coefficientσ_(θ)/σ_(c),brittleness coefficientσ_(c)/σ_(t),uniaxial compressive strengthσ_(c),uniaxial tensionσt,and tangential stress of surrounding rockσ_(θ).PCA method is used to reduce the dimension of the index,extract the main information of the index data,and obtain three linearly independent principal component input vectors.The processed indexes are used as input vectors to train and test three different SOFM models.The variance of the three models and the clustering effect of neurons in the competitive layer are compared,and the optimal prediction model with 16 neurons in the output layer is obtained.Finally,12 groups of engineering examples at home and abroad are introduced to compare the rock burst intensity grade predicted by multidimensional normal cloud model,fuzzy comprehensive evaluation method,grey whitening weight function clustering method and Russense criterion.The research shows that the prediction result of rock burst intensity grade based on PCA-SOFM is highly consistent with the actual situation of the project.The model is feasible and has certain engineering applicability,It provides a new research method for the prediction of rockburst intensity.
作者
陈则黄
李克钢
李明亮
秦庆词
毛明发
Chen Zehuang;Li Kegang;Li Mingliang;Qin Qingci;Mao Mingfa(School of Land and Resources Engineering,Kunming University of Science and Technology,Kunming 650093,P.R.China;Yunnan Key Laboratory of Sino-German Blue Mining and Utilization of Special Underground Space,Kunming 650093,P.R.China;Yunnan Chihong Zn&Ge,Co.,Ltd.,Qujing,Yunnan 655011,P.R.China)
出处
《地下空间与工程学报》
CSCD
北大核心
2022年第S02期934-942,951,共10页
Chinese Journal of Underground Space and Engineering
基金
国家自然科学基金(41672303,51934003)
云南省高校科技创新团队支持计划
关键词
岩石力学
主成分分析法
自组织特征映射神经网络
岩爆烈度等级预测
欧氏距离
rock mechanics
principle component analysis(PCA)
self-organizing feature mapping neural network
prediction of rockburst intensity grade
Euclidean distance
