多轮优化XGBoost-T模型预测爆堆大块率  

Large Block Rate Prediction of Blast Piles Based on Multi-Round Optimized XGBoost-T Model

作　　者：陈立军 CHEN Lijun(China Railway 19th Bureau Group Mining Investment Co.Ltd.,Hulunbuir Inner Mongolia 021400,China)

机构地区：[1]中铁十九局集团矿业投资有限公司,内蒙古呼伦贝尔021400

出　　处：《铁道建筑技术》2025年第2期214-218,共5页Railway Construction Technology

基　　金：岩土力学与工程国家重点实验室开放基金课(No.Z020017);中央高校基本科研业务费专项资金资助项目(N2101041)。

摘　　要：针对露天矿现场台阶爆破后爆堆大块率过高问题,系统收集乌努格吐山铜钼矿30组爆破相关数据,形成以7种预测输入数据和1种输出数据的特征预测数据集。提出通过参数循环调优与交叉验证网格搜索法结合,使用python语言搭建XGBoost-T矿岩爆堆大块率预测模型。利用特征预测数据集80%的数据进行训练,20%数据用于验证,将XGBoost-T模型预测结果同XGBoost模型、最小二乘支持向量机模型(LSSVM)和随机森林(RF)回归预测法对比,并采用RMSE、MAE和R23个评价指标验证模型效果。研究结果表明,XGBoost-T模型预测效果明显高于其他几种模型,其中RMSE、MAE和R2分别为0.1152、0.1387和0.9802,证明了XGBoost-T模型具有较高的预测水平,适用于爆破大块率问题的研究。To address the issue of high block rate of blast piles after bench blasting at open-pit mines,a dataset of 30 groups of blasting-related data was collected from Ulugtau copper-molybdenum mine.This dataset consists of seven predictive input features and one output feature,forming a predictive feature dataset.A prediction model for the large block rate of blast piles,named XGBoost-T,was developed using Python by combining parameter tuning through iterative adjustment with grid search cross-validation.Eighty percent of the predictive dataset was used for training,while the remaining 20%was used for validation.The prediction results of the XGBoost-T model were compared with those of the XGBoost model,the Least Squares Support Vector Machine(LSSVM),and the Random Forest(RF)regression prediction method.Three evaluation metrics-Root Mean Square Error(RMSE),Mean Absolute Error(MAE),and the coefficient of determination(R2)-were applied to assess model performance.The results indicate that the XGBoost-T model significantly outperforms the other models,with RMSE,MAE,and R2 values of 0.1152,0.1387 and 0.9802,respectively.This demonstrates the high predictive capability of the XGBoost-T model,making it suitable for studying the issue of large block rate of blast piles.

关 键 词：露天矿 机器学习 台阶爆破 大块率 XGBoost模型 

分 类 号：TD235[矿业工程—矿井建设]