| 矿柱稳定性判别的ICA-RoF模型及其工程应用 |
| |
| 引用本文： | 肖屈日,赵国彦,刘建,简筝.矿柱稳定性判别的ICA-RoF模型及其工程应用[J].中国地质灾害与防治学报,2019,30(4):116-122. |
| |
| 作者姓名： | 肖屈日 赵国彦 刘建 简筝 |
| |
| 作者单位： | 中南大学资源与安全工程学院;长沙矿山研究院有限责任公司 |
| |
| 基金项目： | 国家重点研发计划项目(2018YFC0604606);国家自然科学基金资助项目(51774321) |
| |
| 摘 要： | 为准确判别矿柱稳定性情况,综合考虑矿柱形状特征量、力学状态量和力学极限量3类指标,选取矿柱宽度、矿柱高度、矿柱宽高比、矿柱约束、矿柱摩擦系数、矿柱应力、矿岩单轴抗压强度、矿柱强度共8个特征作为识别指标,利用独立成分分析旋转森林(ICA-RoF)算法逆构特征指标与矿柱状态之间的非线性映射关系,建立一种基于ICA-RoF算法的矿柱稳定性判别模型。结合工程实例,以150组矿柱样本数据进行训练,采用40次5折交叉验证算法获得最佳模型参数,以剩余12组样本数据对该模型进行检验,并与主成分分析旋转森林算法(PCA-RoF)、CART决策树算法(CDT)和高斯过程分类算法(GPC)进行比较。研究结果表明:ICA-RoF判别模型精度高、泛化能力强,在显著性水平α=0.05的情况下,ICA-RoF明显优于PCA-RoF、CDT和GPC。
|
| 关 键 词： | 安全工程 硬岩矿柱 状态识别 旋转森林 主成分分析 |
The ICA-RoF model of pillar stability recognition and its engineering application |
| |
| XIAO Quri,ZHAO Guoyan,LIU Jian,JIAN Zheng.The ICA-RoF model of pillar stability recognition and its engineering application[J].The Chinese Journal of Geological Hazard and Control,2019,30(4):116-122. |
| |
| Authors: | XIAO Quri ZHAO Guoyan LIU Jian JIAN Zheng |
| |
| Institution: | (School of Resources and Safety Engineering,Central South University,Changsha,Hunan 410083,China;Changsha Institute of Mining Research Co.,Ltd,Changsha,Hunan 410012,China) |
| |
| Abstract: | In order to accurately identify pillar stability,three types of variable with respect to the pillar,such as shape feature,mechanical state and mechanical limit,were took into account,and a total of eight characteristics were selected as identification indicators,which include pillar width,pillar height,pillar width/height ratio,average pillar confinement,pillar friction term,pillar stress,unconfined compressive strength of intact sample of pillar material and pillar strength.The Independent Component Analysis Rotational Forest(ICA-RoF)algorithm was employed to search for the nonlinear relationship between indicators and pillar stability state,and the ICA-RoF model of pillar stability recognition was established.Combined with engineering examples,the sample data containing 150 pillars were utilized in training the model and optimizing its parameters based on forty runs of 5-fold cross-validation,then the model was tested with the remaining 12 samples,as well as compared with Principal Component Analysis Rotation Forest(PCA-RoF),CART Decision Tree(CDT)and Gaussian Process Classification(GPC).The results indicated that high precision and remarkable generalization were achieved by ICA-RoF recognition model,which was significantly better than PCA-RoF,CDT and GPC(significance level,i.e.α=0.05). |
| |
| Keywords: | safety engineering hard rock pillars failure recognition rotation forest ICA |
| 本文献已被 CNKI 维普 等数据库收录！ |
| 点击此处可从《中国地质灾害与防治学报》浏览原始摘要信息 |
| 点击此处可从《中国地质灾害与防治学报》下载免费的PDF全文 |
|
