img

QQ群聊

img

官方微信

高级检索

黄金科学技术 ›› 2024, Vol. 32 ›› Issue (4): 620-630.doi: 10.11872/j.issn.1005-2518.2024.04.022

• 采选技术与矿山管理 • 上一篇    下一篇

基于PSO-XGBoost的露天矿山PPV预测模型研究

李振阳1(),张宝岗1,熊信2,杨承业2,白玉奇1   

  1. 1.北京奥信化工科技发展有限责任公司,北京 100040
    2.中南大学资源与安全工程学院,湖南 长沙 410083
  • 收稿日期:2024-01-11 修回日期:2024-06-04 出版日期:2024-08-31 发布日期:2024-08-27
  • 作者简介:李振阳(1993-),男,河北邯郸人,工程师,从事金属矿床开采及爆破工艺研究工作。735424518@qq.com
  • 基金资助:
    中南大学研究生自主探索创新项目“矿山多装备智能体协同调度优化及孪生仿真研究”(1053320220742)

Research on PPV Prediction Model of Open-pit Mine Based on PSO-XGBoost

Zhenyang LI1(),Baogang ZHANG1,Xin XIONG2,Chengye YANG2,Yuqi BAI1   

  1. 1.Beijing Aoxin Chemical Technology Co. , Ltd. , Beijing 100040, China
    2.School of Resources and Safety Engineering, Central South University, Changsha 410083, Hunan, China
  • Received:2024-01-11 Revised:2024-06-04 Online:2024-08-31 Published:2024-08-27

摘要:

爆破振动的质点峰值速度(Peak Particle Velocity,PPV)是评估爆破开采设计参数合理性的重要指标。为实现一种PPV有效预测模型,借助粒子群优化算法(Particle Swarm Optimization,PSO),结合优化极端梯度提升树(Extreme gradient boosting,XGBoost),构建了一种参数自优化的PSO-XGBoost预测模型。以LK露天铜钴矿为研究对象,选取最大单段炸药量、总炸药量、测量的爆心距、炮孔平均进深尺度和高程差共5个影响因素指标作为研究参数,通过现场收集187次爆破作业实测数据,进一步开展PPV的PSO-XGBoost预测研究,并与传统XGBoost模型、SSA-XGBoost优化模型及萨道夫斯基经验公式的PPV回归预测进行对比分析,最后利用Shapley Additive Explanatory(SHAP)方法开展影响PPV预测结果的敏感性因素分析。结果表明:PSO-XGBoost预测模型的预测评价指标计算结果最优(R2=0.921,RMSE=0.0752,MAE=0.0717,MBE=0.0683),其对PPV的预测结果明显优于传统XGBoost模型、SSA-XGBoost混合优化模型及萨道夫斯基经验公式,同时,敏感性分析得到总炸药用量是影响PPV预测结果的重要参数。进一步说明PSO-XGBoost预测模型可处理多因素的非线性特征,利用PSO-XGBoost预测模型能够更好地结合非线性、离散数据,建立一种可靠、简单有效的PPV预测模型。研究结果可为露天矿山爆破振动快速预测提供参考。

关键词: 露天矿山, 混合优化算法, 机器学习, 爆破振动, 回归分析

Abstract:

The peak particle velocity(PPV) resulting from blasting vibration serves as a crucial metric in assessing the efficacy of blasting and mining design parameters.To enhance the accuracy of PPV predictions,a novel parameter self-optimizing PSO-XGBoost prediction model is introduced,leveraging the Particle Swarm Optimization (PSO) algorithm in conjunction with optimized Extreme Gradient Boosting (XGBoost).The research focuses on the LK open-pit copper-cobalt mine and examines five influencing factors,namely the maximum single explosive charge,total explosive charge,measured distance to blast center,average depth of blastholes,and elevation difference,as study parameters.A total of 187 sets of measured data from on-site blasting operations are gathered for further investigation into predicting PPV using the PSO-XGBoost model.The findings indicate that the PSO-XGBoost prediction model produces superior prediction evaluation metrics (R2=0.921,RMSE=0.0752,MAE=0.0717,MBE=0.0683), compared to alternative models,including the traditional XGBoost model,the SSA-XGBoost hybrid optimization model,and the Sariakaliski empirical formula.The sensitivity analysis reveals that the total explosive charge,significantly influences PPV prediction outcomes,underscoring the importance of using an appropriate amount of explosives to avoid energy inefficiency in blasting activities. Furthermore,the research demonstrates that the PSO-XGBoost prediction model,is capable of effectively addressing the nonlinear attributes of various factors,as well as integrating nonlinear and discrete data to develop a dependable,straightforward,and efficient PPV prediction model.This study offers valuable insights for promptly predicting blasting vibration in open-pit mining and assessing the impacts of blasting activities.

Key words: open-pit mine, hybrid optimizing algorithm, machine learning, blasting vibration, regression analysis

中图分类号: 

  • TD235

图1

PSO算法原理"

表1

XGBoost模型计算参数"

参数参数说明
n_estimator该参数决定了模型的预测能力,其值越大则模型的学习及数据分析能力越强,预测、结果越精确,但参数过大将浪费计算资源
learning_rate该参数决定了迭代的步长,参数过大将影响运行的准确率,过小则影响运行的速度
max_depth该参数表示为树的最大深度,用来控制算法的过拟合,其值越大则模型的学习将更具体
gamma该参数指定了决策树节点分裂所需的最小损失函数下降值,其值越大,算法越保守

图2

PSO-XGBoost预测算法流程"

图3

研究区域与所选参数"

表2

现场实测数据(部分)"

序号

最大单段

炸药量/kg

总炸药量/kg爆心距/m孔深/m高程差/m

PPV

/(cm·s-1

122256411054.1
215231561232.3
318245481163.4
419243491133.7
517235621362.9
615240541243.1
???????
18116247701252.4
18219254411344.2
18315242471333.3
18415231551232.3
18516248561132.5
18613268431364.0
18716246491133.0

图4

数据集相关性矩阵分析注:Corr和 P_value为2个参数之间的相关性。其中,Corr值越大表示2个参数之间的相关性越强,Corr的正负值表示正负相关性;P_value<0.05表示2个参数之间存在统计学意义上的相关性"

图5

优化适应度收敛"

表3

最优计算参数"

参数参数取值
n_estimator6.56e+02
learning_rate3.04e-02
max_depth3.49
gamma6.95e-03

图6

不同模型回归预测结果比较"

图7

SHAP敏感性分析"

Deng Hongwei, Luo Liang,2023.PPV prediction model based on random forest optimized by SMA algorithm[J].Gold Science and Technology,31(4):624-634.
Dong Donglin, Zhang Longqiang, Zhang Enyu,et al,2023.A rapid identification model of mine water inrush based on PSO-XGBoost[J].Coal Science and Technology,51(7):72-82.
Fan Yong, Pei Yong, Yang Guangdong,et al,2022.Prediction of blasting vibration velocity peak based on an improved PSO-BP neural network[J].Journal of Vibration and Shock,41(16):194-200.
Gou Qianqian, Zhao Mingsheng, Chi En’an,et al,2018.Prediction and application of evaluation factors in blasting vibration based on PCA-BP neural network[J].Mining Research and Development,38(12):97-102.
Guo Xiaoqiang,2021.Study on the determination of the limit distance of boundary borehole based on blasting vibration[J].Mining Research and Development,41(6):26-30.
He Li, Liu Yihe, Li Linna,et al,2022.Prediction of mine Blasting vibration velocity of mines based on particle swarm -least square support vector machine model[J].Metal Mines,51(7):145-150.
Hong Z X, Tao M, Liu L L,et al,2023.An intelligent approach for predicting overbreak in underground blasting operation based on an optimized XGBoost model[J].Engineering Applications of Artificial Intelligence,126(2):1-16.
Hu Xiaobing, Chen Zhiyuan, Wei Geping,et al,2020.Blasting vibration prediction system based on BP neural network[J].Mining Research and Development,40(9):154-158.
Khandelwal M, Armaghani D J, Faradonbeh R S,et al,2017.Classification and regression tree technique in estimating peak particle velocity caused by blasting[J].Engineering with Computers,33:45-53.
Li D T, Yan J L, Zhang L,2012.Prediction of blast-induced ground vibration using support vector machine by tunnel excavation[J].Applied Mechanics and Materials,170/171/172/173:1414-1418.
Ma C L, Wang W H, Wang S X,et al,2023.Prediction of shear strength of RC slender beams based on interpretable machine learning[J].Structures,57:105171.
Sadovsky M A,1952.Mechanical action of blast waves on data of experimental studies[J].Physics of Explosion,12(1):70-110.
Shi Chunyu, Zhang Zhihong, Zhou Jie,et al,2023.Study on blasting vibration transmission characteristics of rock mass of structural planes with different angles [J].Mining Research and Development,43(5):184-189.
Shirani Faradonbeh R, Jahed Armaghani D, Abd Majid M Z,et al,2016.Prediction of ground vibration due to quarry blasting based on gene expression programming:A new model for peak particle velocity prediction[J].International Journal of Environmental Science and Technology,13(6):1453-1464.
Trabi B, Bleibinhaus F,2023.Blast vibration prediction[J].Geophysical Prospecting,71(7):1312-1324.
Tyagi P, Sharma A, Semwal R,et al,2023.XGBoost odor prediction model:Finding the structure-odor relationship of odorant molecules using the extreme gradient boosting algorithm[J].Journal of Biomolecular Structure and Dynamics,12(2):1-15.
Wang Ziyi, Wu Guiyi, Luo Chang,et al,2023.Study on vibration response and stability of steep slope under multiple blasting vibration[J].Blasting,40(3):158-169.
Xu Xiangdong, Lu Yu,2023.Study on empirical formula of blasting vibration propagation reflecting the influence of height difference[J].Automation and Instrumentation,(6):63-66.
Xuan G Q, Wu X Y,2023.Support vector regression optimized by black widow optimization algorithm combining with feature selection by MARS for mining blast vibration prediction[J].Measurement,218:113106.
Yu Z, Li C Q, Zhou J,2023.Tunnel boring machine performance prediction using supervised learning method and swarm intelligence algorithm[J].Mathematics,11(20):4237.
Zeng Xiaohui, Zhang Xuemin, Dai Bin,et al,2023.Prediction of tunnel blasting vibration velocity considering influence of number of free surfaces and resistance line[J].China Work Safety Science and Technology,19(6):83-89.
Zhang Xiliang,2020.Ensemble Learning Model and Engineering Application of Environmental Effect Prediction of Rock Mass Blasting[D].Hefei:University of Science and Technology of China.
Zhou J, Qiu Y G, Khandelwal M,et al,2021.Developing a hybrid model of Jaya algorithm-based extreme gradient boosting machine to estimate blast-induced ground vibrations[J].International Journal of Rock Mechanics and Mining Sciences,145(1):104856.
Zou Ping, Wang Liang, Dai Yong,et al,2023.Establishment and application of blasting vibration prediction system based on SSA-XGBoost [J].Blasting,40(3):199-205.
邓红卫,罗亮,2023.基于SMA算法优化随机森林的PPV预测模型[J].黄金科学技术,31(4):624-634.
董东林,张陇强,张恩雨,等,2023.基于PSO-XGBoost的矿井突水水源快速判识模型[J].煤炭科学技术,51(7):72-82.
范勇,裴勇,杨广栋,等,2022.基于改进 PSO-BP 神经网络的爆破振动速度峰值预测[J].振动与冲击,41(16):194-200.
苟倩倩,赵明生,池恩安,等,2018.基于PCA-BP神经网络在爆破振动评价要素中的预测及应用[J].矿业研究与开发,38(12):97-102.
郭晓强,2021.基于爆破振动的极限边孔排距的确定方法研究[J].矿业研究与开发,41(6):26-30.
何理,刘易和,李琳娜,等,2022.基于粒子群—最小二乘支持向量机模型的矿山爆破振动速度预测[J].金属矿山,51(7):145-150.
胡晓冰,陈志远,魏格平,等,2020.基于BP神经网络的爆破振动预测系统[J].矿业研究与开发,40(9):154-158.
史春宇,张志鸿,周杰,等,2023.不同角度结构面岩体的爆破振动传播特性研究[J].矿业研究与开发,43(5):184-189.
王子一,吴桂义,罗畅,等,2023.多次爆破振动下陡边坡振动响应及稳定性研究[J].爆破,40(3):158-169.
徐向东,陆瑜,2023.反映高差影响的爆破振动传播经验公式研究[J].自动化与仪器仪表,(6):63-66.
曾晓辉,张学民,戴斌,等,2023.考虑自由面数量和抵抗线影响的隧道爆破振速预测[J].中国安全生产科学技术,19(6):83-89.
张西良,2020.岩体爆破环境效应预测的集成学习模型及工程应用[D].合肥:中国科学技术大学.
邹平,王亮,戴勇,等,2023.基于SSA-XGBoost的爆破振动预测系统的构建与应用[J].爆破,40(3):199-205.
[1] 李地元, 杨博, 刘子达, 刘永平, 赵君杰. 基于集成树算法的岩石黏聚力和内摩擦角预测方法[J]. 黄金科学技术, 2024, 32(5): 847-859.
[2] 李炎, 王建国, 魏生云, 李国璋, 胡建, 王志男. 西藏德新铅多金属矿床地球物理与地球化学综合找矿研究[J]. 黄金科学技术, 2024, 32(3): 400-415.
[3] 李祥龙, 余林, 黄原明, 陈浩, 赵艳伟. 基于VMD-HHT的井下预裂爆破振动效应分析[J]. 黄金科学技术, 2024, 32(3): 501-510.
[4] 姜志宏, 陈澳. 融合全监督学习的半监督矿石粒度预测算法[J]. 黄金科学技术, 2024, 32(3): 539-547.
[5] 代树红, 张战军, 柳凯, 郑昊, 孙清林. 基于PEMD-MPE算法的露天矿爆破振动信号降噪方法[J]. 黄金科学技术, 2024, 32(1): 82-90.
[6] 凡兴禹, 王雪林. 基于改进XGBoost算法的深部巷道松动圈智能预测研究[J]. 黄金科学技术, 2024, 32(1): 109-122.
[7] 海龙, 冯丽鑫, 谭世林, 吕勇博. 根土复合体加固露天矿山排土场边坡研究[J]. 黄金科学技术, 2023, 31(6): 911-918.
[8] 许方颖, 邹艳红, 易卓炜, 杨福强, 毛先成. 基于非均衡数据的ADASYN-CatBoost测井岩性智能识别——以胶西北招贤金矿床为例[J]. 黄金科学技术, 2023, 31(5): 721-735.
[9] 邓红卫, 罗亮. 基于SMA算法优化随机森林的PPV预测模型[J]. 黄金科学技术, 2023, 31(4): 624-634.
[10] 王鹏飞,毕林,王李管. 露天矿无人驾驶矿卡速度规划研究[J]. 黄金科学技术, 2022, 30(3): 460-469.
[11] 吴钦正,李润然,李桂林,李金平,尹延天,徐帅. 基于JKSimBlast软件的露天矿爆破毫秒延期时间优化[J]. 黄金科学技术, 2021, 29(6): 854-862.
[12] 胡建华,郭萌萌,周坦,张涛. 基于改进迁移学习算法的岩体质量评价模型[J]. 黄金科学技术, 2021, 29(6): 826-833.
[13] 张美道,饶运章,徐文峰,王文涛. 全尾砂膏体充填配比优化正交试验[J]. 黄金科学技术, 2021, 29(5): 740-748.
[14] 景岳,王少锋,鲁金涛. 矿岩开挖松动区厚度预测及非爆机械化开采判据[J]. 黄金科学技术, 2021, 29(4): 525-534.
[15] 刘奇,岑佑华,刘东锐,罗卫兵,徐喜. 基于静态沉降试验的全尾砂浓密技术参数预测[J]. 黄金科学技术, 2021, 29(2): 266-274.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!