基于灰狼算法改进随机森林算法的爆破振动速度预测研究
Study on prediction of blasting vibration velocity based on Grey Wolf Algorithm improved Random Forest Algorithm
胡学敏 1曾晟 2宋良灵3
作者信息
- 1. 南华大学资源环境与安全工程学院;湖南有色金属职业技术学院资源环境系
- 2. 南华大学资源环境与安全工程学院
- 3. 南华大学土木工程学院
- 折叠
摘要
为改善露天矿爆破作业过程中的峰值质点速度在采用传统经验公式和单一仿生算法时预测精度不够的问题,引入灰狼算法(GWO)优化了随机森林算法(RF)中决策树的棵数和层数2 个超参数,成功构建了基于灰狼算法改进的GWO-RF爆破振动速度预测模型.结合某爆破工程69 组爆破监测数据,以爆心距、最大段药量、总装药量、微差时间、炮孔数、孔距、孔深及排距为输入参数,运用GWO-RF预测模型和RF模型进行爆破振动峰值速度预测对比.结果表明:GWO-RF组合算法能够考虑更多符合实际的爆破振动速度影响因素,GWO-RF组合算法误差率比RF算法提高了37.83 百分点;GWO-RF组合算法的爆破振动速度预测精度达到97.72%.说明GWO成功优化了RF中决策树的2 个超参数,也证明GWO-RF组合算法能很好进行露天矿爆破振动速度预测.
Abstract
In order to improve the prediction accuracy of peak particle velocity during open-pit blasting opera-tions,which is insufficient when using traditional empirical formulas and single bionic algorithms,the Grey Wolf Algo-rithm(GWO)is introduced to optimize 2 hyperparameters,the number and depth of decision trees in the Random Forest Algorithm(RF).This successfully constructs the GWO-RF blasting vibration velocity prediction model.By combin-ing 69 sets of blasting monitoring data from a blasting project,input parameters such as blasthole distance,maximum interval charge,total charge,millisecond delay,number of blastholes,hole spacing,depth,and role spacing are used to compare the prediction of peak vibration velocity using the GWO-RF model and the RF model.The results show that the GWO-RF combined algorithm can consider more practical factors affecting blasting vibration velocity and improves the error rate by 37.83 percentage points compared to the RF;the prediction accuracy of blasting vibration velocity using the GWO-RF combined algorithm reaches 97.72%.This indicates that the GWO successfully optimizes the 2 hyperparameters of decisiontrees in the RF model and demonstrates that the GWO-RF combined algorithm can be used for accurate prediction of blasting vibration velocity in open-pit mining.
关键词
露天开采/爆破振动/速度预测/随机森林算法/灰狼算法Key words
open-pit mining/blasting vibration/velocity prediction/Random Forest Algorithm/Grey Wolf Algo-rithm引用本文复制引用
基金项目
湖南省研究生科研创新项目(QL20210225)
出版年
2024
NETL
NSTL
维普
万方数据