摘要
为了提高岩爆预测模型的精度,以围岩洞壁最大切向应力(MTS)、岩石单轴抗压强度(UCS)、岩石单轴抗拉强度(UTS)、应力系数(SCF)、脆性系数(BI)、岩石弹性能指数(EEI)等参数作为预选预测指标。运用修正散点图矩阵分析指标间、指标与岩爆等级间的关系,筛选指标集中的离群值,确定构成岩爆预测的指标体系。引入并优化随机森林算法,采用Randomize Search CV和Grid Search CV方法寻求最优超参数,运用优化后模型对岩爆实例进行岩爆倾向性等级预测,并将预测结果与神经网络模型(ANN)、支持向量机模型(SVM)、XGBoost模型结果进行分析对比。研究表明:修正散点图矩阵对筛选多维岩爆数据离群值是有效的,优化后的Random Forest模型的预测准确率为92.6%,为岩爆倾向性分级提供一种新的方法。
In order to improve the accuracy of rockburst prediction model,the parameters such as maximum principal stress of surrounding rock wall,uniaxial compressive strength of rock,uniaxial tensile strength of rock,stress concentration factor,brittleness index and rock elastic energy index.The modified scatter diagram matrix is used to analyze the relationship between the indexes and the rockburst grade,to screen the outliers in the index set,and to determine the index to form the rockburst prediction index system.The random forest algorithm is introduced and optimized,and the Randomize Search CV and Grid Search CV methods are used to find the optimal parameters.The optimized model is used to predict the rockburst tendency grade of rockburst examples,and the predicted results are analyzed and compared with the results of neural network model(ANN),support vector machine model(SVM)and XGBoost model.The results show that the modified scatter graph matrix is effective for multi-dimensional data analysis,and the accuracy of the optimized Random Forest model is 92.6%.It provides a new method for rockburst tendency.
作者
刘剑
周宗红
LIU Jian;ZHOU Zonghong(School of Land and Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China)
出处
《有色金属工程》
CAS
北大核心
2022年第3期120-128,共9页
Nonferrous Metals Engineering
基金
国家自然科学基金资助项目(51864023,51264018)。
关键词
岩爆灾害等级预测
修正散点图矩阵
指标优选
优化随机森林模型
rockburst disaster grade prediction
scatter plot matrix
index optimization
optimized random forest model
作者简介
刘剑(1995-),男,硕士,主要从事采矿与岩石力学研究;通信作者:周宗红(1967-),男,教授,博导,主要从事于采矿工程与岩石力学教学科研工作。
