基于Stacking集成算法的岩爆等级预测研究

doi:10.13722/j.cnki.jrme.2021.0831

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (506 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要岩爆是地下工程面临的一个巨大灾害，岩爆预测可以降低岩爆带来的危害。机器学习是岩爆预测方法的研究热点和发展方向，但现阶段各机器学习算法表现不同，并且相互独立工作，没有融合，不能优势互补，导致各机器学习算法的准确率、泛化性和稳定性较低。本文采用Stacking集成算法，融合现阶段使用较多的8个机器学习算法(4个集成算法和4个基本算法)，充分发挥各算法的优势，实现优势互补；为保证新特征信息具有多样性，结合各种机器学习算法的原理和岩爆样本库的特点，提出3组考虑多个岩爆预测指标的Stacking集成算法，每组算法拥有不同基模型和多个元模型，解决了传统Stacking集成算法接受特征信息受限和元模型选择困难的难题。对比分析各组Stacking集成算法与独立算法的准确率、精确率、召回率和F1值，结果表明构建的Stacking集成算法可以有效融合各机器学习算法，预测性能显著提升。在3组Stacking集成算法中，Stacking集成算法二的基模型由Random Forest Classifier，Extra Trees Classifier，Gradient Boosting Classifier和LGBMClassifier四个预测准确率较高的集成算法构成，基模型提供的新特征信息精度最高，各元模型预测性提升最显著，表现最优，其中该组元模型中SVC算法和LGBMClassifier算法的准确率在3组元模型中最高。采用不同组表现较好的元模型投票预测岩爆，Stacking集成算法可以学习到更多特征信息，测试集的准确率最终稳定在94.12%，优于独立机器学习和常用理论判据的预测性能(最高准确率分别为91.18%和53.9%)。最后，采用建立的Stacking集成算法对终南山隧道竖井进行岩爆预测，预测结果与实际情况较为一致。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	刘德军1
	2
	戴庆庆1
	左建平1
	2
	商奇1
	陈国亮1
	郭艺豪1

关键词 ：岩石力学, 岩爆预测, Stacking集成算法, 算法融合, 机器学习

Abstract：Rockburst is a huge disaster faced by underground engineering，and rockburst prediction can reduce the harm caused by rockburst. Machine learning is the research hotspot and development direction of rockburst prediction methods，However，at present，each machine learning algorithm performs differently，and works independently of each other. Without fusion，they cannot complement each other?s advantages，resulting in low accuracy，generalization and stability of all machine learning algorithms. In this paper，eight machine learning algorithms(four ensemble algorithms and four basic algorithms) which are widely used at present are integrated by the Stacking integration algorithm，so as to give full play to the advantages of each algorithm and realize the complementary advantages. In order to ensure the diversity of new feature information，combined with the principles of various machine learning algorithms and the characteristics of rockburst sample library，three sets of Stacking integration algorithms with multiple rockburst prediction indices were proposed，each of which had different base models and multiple meta-models. The difficulty of accepting feature information and selecting meta-models in traditional Stacking integration algorithm is solved. The accuracy，accuracy，recall，and F1 values of each Stacking algorithm and independent algorithm were compared and analyzed，the results show that the Stacking algorithm can effectively integrate all the machine learning algorithms，and the prediction performance is significantly improved. Among the three sets of Stacking algorithms，the base-model of Stacking algorithm 2 is composed of Random Forest Classifier，Extra Trees Classifier，Gradient Boosting Classifier，and LGBMClassifier integration algorithms with high prediction accuracy，and the precision of new feature information provided by the base-model is the highest，and the prediction improvement of each meta-model is the most significant. Among the three sets of Stacking algorithms，the accuracy of the SVC algorithm and the LGBMClassifier algorithm in the meta-model of Stacking algorithm 2 is the highest in the three sets of meta-models. The Stacking ensemble algorithm can learn more feature information by using different groups of meta-models with good performance to vote for rockburst prediction. The accuracy of the test set is finally stable at 94.12 %，which is better than the prediction performance of independent machine learning and common theoretical criteria(the highest accuracy is 91.18 % and 53.9 %，respectively). Finally，the rock burst prediction of Zhongnanshan tunnel shaft is carried out by the established Stacking integration algorithm，and the prediction results are consistent with the actual situation.

Key words： rock mechanics rockburst prediction Stacking integration algorithm algorithm fusion machine learning

引用本文:

刘德军1，2，戴庆庆1，左建平1，2，商奇1，陈国亮1，郭艺豪1. 基于Stacking集成算法的岩爆等级预测研究[J]. 岩石力学与工程学报, 2022, 41(S1): 2915-2926.
LIU Dejun1，2，DAI Qingqing1，ZUO Jianping1，2，SHANG Qi1，CHEN Guoliang1，GUO Yihao1. Research on rockburst grade prediction based on stacking integrated algorithm. , 2022, 41(S1): 2915-2926.

链接本文:

http://rockmech.whrsm.ac.cn/CN/10.13722/j.cnki.jrme.2021.0831 或 http://rockmech.whrsm.ac.cn/CN/Y2022/V41/IS1/2915

[15]	鲍跃全，李惠. 人工智能时代的土木工程[J]. 土木工程学报，2019，52(5)：1-11.(BAO Yuequan，LI Hui. Civil engineering in the era of artificial intelligence[J]. China Civil Engineering Journal，2019，52(5)：1-11.(in Chinese))
[25]	ZHAO H B. Rockburst prediction using evolutionary support vector machine[C]// Progress in Safety Science and Technology. [S. l.]：[s. n.]，2005：494-498.
[55]	BREIMAN L. Stacked regressions[J]. Machine Learning，1996，24(1)：49-64.
[7]	HOEK E. Underground excavations in rock[M]. London：The Institute of Mining and Metallurgy，1980：27
[47]	赵洪波. 岩爆分类的支持向量机方法[J]. 岩土力学，2005，26(4)：642-644.(ZHAO Hongbo. Support vector machine method for rockburst classification[J]. Rock and Soil Mechanics，2005，26(4)：642-644.(in Chinese))
[6]	陶振宇. 高地应力区的岩爆及其判别[J]. 人民长江，1987，18(5)：25-32.(TAO Zhenyu. Rock burst in high ground stress area and its identification[J]. People?s Yangtze River，1987，18(5)：25-32.(in Chinese))
[56]	GRANDVALET Y. Bagging equalizes influence[J]. Machine Learning，2004，55(3)：251-270.
[1]	徐成光. 岩爆预测及防治方法综述[J]. 现代隧道技术，2005，42(6)：81-85.(XU Chengguang. Summary of rockburst prediction and prevention methods[J]. Modern Tunnelling Technology，2005，42(6)：81-85.(in Chinese))
[11]	李明亮，李克钢，刘月东，等. 基于变异系数与序关分析法-多维正态云模型的岩爆预测[J]. 岩石力学与工程学报，2020，39(增2)：3 395-3 402.(LI Mingliang，LI Kegang，LIU Yuedong，et al. Rockburst prediction based on coefficient of variation and sequence analysis method-multi-dimensional normal cloud model[J]. Chinese Journal of Rock Mechanics and Engineering，2020，39(Supp.2)： 3 395-3 402.(in Chinese))
[16]	ZHOU J S，CAI Z Y，DESTECH PUBLICAT I. Overview of data mining based on machine learning[C]// Progress of the International Conference on Computer Science and Communication Engineering (CSCE). [S. l.]：[s. n.]，2015：494-498.
[21]	ZHOU J，LI X，MITRI H S. Classification of rockburst in underground projects：comparison of ten supervised learning methods[J]. Journal of Computing in Civil Engineering，2016，30(5)：04016003.
[26]	ADOKO A C，GOKCEOGLU C，WU L，et al. Knowledge-based and data-driven fuzzy modeling for rockburst prediction[J]. International Journal of Rock Mechanics and Mining Sciences，2013，61(4)：86-95.
[31]	白明洲，王连俊，许兆义. 岩爆危险性预测的神经网络模型及应用研究[J]. 中国安全科学学报，2002，12(4)：68-72.(BAI Mingzhou，WANG Lianjun，XU Zhaoyi. Neural network model and application research on rockburst risk prediction[J]. Chinese Safety Science Journal，2002，12(4)：68-72.(in Chinese))
[36]	李丽娟. 金川矿山深部岩石岩爆倾向性研究[硕士学位论文][D]. 长沙：中南大学，2009.(LI Lijuan. Research on rock burst tendency of deep rocks in Jinchuan mine[M. S. Thesis][D]. Changsha：Central South University，2009.(in Chinese))
[37]	刘章军，袁秋平，李建林. 模糊概率模型在岩爆烈度分级预测中的应用[J]. 岩石力学与工程学报，2008，27(增1)：3 095-3 103.(LIU Zhangjun，YUAN Qiuping，LI Jianlin. Application of fuzzy probability model in rockburst intensity classification prediction[J]. Chinese Journal of Rock Mechanics and Engineering，2008，27(Supp.1)： 3 095-3 103.(in Chinese))
[39]	孙臣生. 基于改进MATLAB-BP神经网络算法的隧道岩爆预测模型[J]. 重庆交通大学学报：自然科学版，2019，38(10)：41-49. (SUN Chensheng. Tunnel rockburst prediction model based on improved MATLAB-BP neural network algorithm[J]. Journal of Chongqing Jiaotong University：Natural Science，2019，38(10)：41-49.(in Chinese))
[41]	王万德，张延新. 深部开采岩爆预测的神经网络方法[J]. 河北科技师范学院学报，2007，21(2)：35-38.(WANG Wande，ZHANG Yanxin. Neural network method for prediction of rock burst in deep mining[J]. Journal of Hebei Normal University of Science and Technology，2007，21(2)：35-38.(in Chinese))
[43]	许梦国，杜子建，姚高辉，等. 程潮铁矿深部开采岩爆预测[J]. 岩石力学与工程学报，2008，27(增1)：2 921-2 928.(XU Mengguo，DU Zijian，YAO Gaohui，et al. Rockburst prediction in deep mining of Chengchao Iron Mine[J]. Chinese Journal of Rock Mechanics and Engineering，2008，27(Supp.1)：2 921-2 928.(in Chinese))
[46]	张乐文，张德永，邱道宏. 基于粗糙集的可拓评判在岩爆预测中的应用[J]. 煤炭学报，2010，35(9)：1 461-1 465.(ZHANG Lewen，ZHANG Deyong，QIU Daohong. Application of extension evaluation based on rough set in rockburst prediction[J]. Journal of China Coal Society，2010，35(9)：1 461-1 465.(in Chinese))
[10]	王元汉，李卧东，李启光，等. 岩爆预测的模糊数学综合评判方法[J]. 岩石力学与工程学报，1998，17(5)：15-23.(WANG Yuanhan，LI Wodong，LI Qiguang，et al. Comprehensive fuzzy mathematics evaluation method for rockburst prediction[J]. Chinese Journal of Rock Mechanics and Engineering，1998，17(5)：15-23.(in Chinese))
[20]	DONG L J，LI X B，PENG K. Prediction of rockburst classification using Random Forest[J]. Transactions of Nonferrous Metals Society of China，2013，23(2)：472-477.
[30]	SU G S，ZHANG K S，CHEN Z. Rockburst prediction using gaussian process machine learning[C]// Proceedings of the International Conference on Computational Intelligence and Software Engineering. [S. l.]：[s. n.]，2009：1-4.
[40]	王吉亮，陈剑平，杨静，等. 岩爆等级判定的距离判别分析方法及应用[J]. 岩土力学，2009，30(7)：2 203-2 208.(WANG Jiliang，CHEN Jianping，YANG Jing，et al. The distance discriminant analysis method and its application for judging rockburst grade[J]. Rock and Soil Mechanics，2009，30(7)：2 203-2 208.(in Chinese))
[8]	KIDYBI?SKI A. Bursting liability indices of coal[J]. International Journal of Rock Mechanics and Mining Sciences and Geomechanics Abstracts，1981，18(4)：295-304.
[18]	PU Y，APEL D B，XU H. Rockburst prediction in kimberlite with unsupervised learning method and support vector classifier[J]. Tunnelling and Underground Space Technology，2019，90：12-18.
[28]	SUN J，WANG L G，ZHANG H L，et al. Application of fuzzy neural network in predicting the risk of rock burst[J]. Procedia Earth and Planetary Science，2009，1(1)：536-543.
[38]	秦乃兵. 用模糊数学方法对岩爆进行预测预报[J]. 工程力学，2001，18(增)：734-738.(QIN Naibing. Using fuzzy mathematics method to forecast rockburst[J]. Engineering Mechanics，2001，18(Supp.)：734-738.(in Chinese))
[48]	田睿. 基于机器学习的岩爆烈度等级预测模型研究与应用[博士学位论文][D]. 包头：内蒙古科技大学，2020.(TIAN Rui. Research and application of rockburst intensity prediction model based on machine learning[Ph. D. Thesis][D]. Baotou：Inner Mongolia University of Science and Technology，2020.(in Chinese))
[4]	张传庆，卢景景，陈珺，等. 岩爆倾向性指标及其相互关系探讨[J]. 岩土力学，2017，38(5)：1 397-1 404.(ZHANG Chuanqing，LU Jingjing，CHEN Jun，et al. Discussion on rockburst tendency indexes and their correlation[J]. Rock and Soil Mechanics，2017，38(5)： 1 397-1 404.(in Chinese))
[14]	杨莹春，诸静. 一种新的岩爆分级预报模型及其应用[J]. 煤炭学报，2000，25(2)：169-172.(YANG Yingchun，ZHU Jing. A new rockburst classification prediction model and its application[J]. Journal of China Coal Society，2000，25(2)：169-172.(in Chinese))
[24]	宫凤强，李夕兵，张伟. 基于Bayes判别分析方法的地下工程岩爆发生及烈度分级预测[J]. 岩土力学，2010，31(增1)：370-379. (GONG Fengqiang，LI Xibing，ZHANG Wei. Rockburst occurrence and intensity classification prediction of underground engineering based on Bayes discriminant analysis method[J]. Rock and Soil Mechanics，2010，31(Supp.1)：370-379.(in Chinese))
[34]	葛启发，冯夏庭. 基于AdaBoost组合学习方法的岩爆分类预测研究[J]. 岩土力学，2008，29(4)：943-948.(GE Qifa，FENG Xiating. Research on rockburst classification prediction based on AdaBoost combined learning method[J]. Rock and Soil Mechanics，2008，29(4)：943-948.(in Chinese))
[44]	薛杨朔，孙志国，于全有. BP神经网络动态预测方法在巷道底板岩爆中的应用[J]. 煤矿安全，2013，44(3)：150-152.(XUE Yangshuo，SUN Zhiguo，YU Quanyou. Application of BP neural network dynamic prediction method in roadway floor rockburst[J]. Coal Mine Safety，2013，44(3)：150-152.(in Chinese))
[54]	GILLET A， BROSTAUX Y， PALM R. Main models used in logistic regression[J]. Biotechnologie Agronomie Societe Et Environnement，2011，15(3)：425-433.
[2]	周青春，李海波，杨春和. 地下工程岩爆及其风险评估综述[J]. 岩土力学，2003，24(增2)：669-673.(ZHOU Qingchun，LI Haibo，YANG Chunhe. Summary of underground engineering rock burst and its risk assessment[J]. Rock and Soil Mechanics，2003，24(Supp.2)：669-673.(in Chinese))
[12]	刘鹏，余斌，曹辉. 基于多维正态云-CRITIC模型的岩爆倾向性综合评价方法[J]. 岩石力学与工程学报，2020，39(增2)：3 432- 3 439.(LIU Peng，YU Bin，CAO Hui. Comprehensive evaluation method of rockburst tendency based on multi-dimensional normal cloud-CRITIC model[J]. Chinese Journal of Rock Mechanics and Engineering，2020，39(Supp.2)：3 432-3 439.(in Chinese))
[22]	汤志立，徐千军. 基于9种机器学习算法的岩爆预测研究[J]. 岩石力学与工程学报，2020，39(4)：773-781.(TANG Zhili，XU Qianjun. Research on rockburst prediction based on 9 machine learning algorithms[J]. Chinese Journal of Rock Mechanics and Engineering， 2020，39(4)：773-781.(in Chinese))
[32]	蔡嗣经，张禄华，周文略. 深井硬岩矿山岩爆灾害预测研究[J]. 中国安全生产科学技术，2005，1(5)：19-22.(CAI Sijing，ZHANG Luhua，ZHOU Wenlue. Research on the prediction of rock burst disasters in deep hard rock mines[J]. China Safety Science and Technology，2005，1(5)：19-22.(in Chinese))
[42]	王心飞. 深埋隧道稳定性分析的智能化及非线性研究[博士学位论文][D]. 重庆：重庆大学，2006.(WANG Xinfei. Intelligent and nonlinear research on stability analysis of deep-buried tunnels[Ph. D. Thesis][D]. Chongqing：Chongqing University，2006.(in Chinese))
[52]	张旭东，钱仲文，沈思琪，等. 一种基于LSTM与LGBM的电力负荷预测算法[J]. 系统工程，2019，37(1)：152-158.(ZHANG Xudong，QIAN Zhongwen，SHEN Siqi，et al. A power load forecasting algorithm based on LSTM and LGBM[J]. Systems Engineering，2019，37(1)：152-158.(in Chinese))
[3]	陈结，高靖宽，蒲源源，等. 冲击地压预测预警的机器学习方法[J]. 采矿与岩层控制工程学报，2021，3(1)：57-68.(CHEN Jie，GAO Jingkuan，PU Yuanyuan，et al. Machine learning method for forecasting and early warning of rock burst[J]. Journal of Mining and Strata Control Engineering，2021，3(1)：57-68.(in Chinese))
[5]	RYDER J A. Excess shear stress in the assessment of geologically hazardous situations[J]. Journal of the Southern African Institute of Mining and Metallurgy，1988，88(1)：27-39.
[9]	WANG C，WU A，LU H，et al. Predicting rockburst tendency based on fuzzy matter-element model[J]. International Journal of Rock Mechanics and Mining Sciences，2015，75(4)：224-232.
[13]	胡建华，尚俊龙，周科平. 岩爆烈度预测的改进物元可拓模型与实例分析[J]. 中国有色金属学报，2013，23(2)：495-502.(HU Jianhua，SHANG Junlong，ZHOU Keping. Improved matter-element extension model and case analysis for rockburst intensity prediction[J]. The Chinese Journal of Nonferrous Metals，2013，23(2)：495-502.(in Chinese))
[17]	ZHOU J，LI X B，SHI X Z. Long-term prediction model of rockburst in underground openings using heuristic algorithms and support vector machines[J]. Safety Science，2012，50(4)：629-644.
[19]	苏国韶，宋咏春，燕柳斌. 岩体爆破效应预测的一种新方法[J]. 岩石力学与工程学报，2007，26(增1)：3 509-3 514.(SU Guoshao，SONG Yongchun，YAN Liubin. A new method for predicting the effect of rock mass blasting[J]. Chinese Journal of Rock Mechanics and Engineering，2007，26(Supp.1)：3 509-3 514.(in Chinese))
[23]	田睿，孟海东，陈世江，等. 基于机器学习的3种岩爆烈度分级预测模型对比研究[J]. 黄金科学技术，2020，28(6)：920-929. (TIAN Rui，MENG Haidong，CHEN Shijiang，et al. Comparative study on three rockburst intensity grading prediction models based on machine learning[J]. Gold Science and Technology，2020，28(6)：920-929.(in Chinese))
[27]	AFRAEI S，SHAHRIAR K，MADANI S H. Developing intelligent classification models for rock burst prediction after recognizing significant predictor variables，Section 2：Designing classifiers[J]. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research，2019，84：522-537.
[29]	LIU R，YE Y，HU N，et al. Classified prediction model of rockburst using rough sets-normal cloud[J]. Neural Computing and Applications，2019，31(12)：8 185-8 193.
[33]	陈海军，郦能惠，聂德新，等. 岩爆预测的人工神经网络模型[J]. 岩土工程学报，2002，24(2)：229-232.(CHEN Haijun，LI Nenghui，NIE Dexin，et al. Artificial neural network model for rockburst prediction[J]. Chinese Journal of Geotechnical Engineering，2002，24(2)：229-232.(in Chinese))
[35]	贾义鹏，吕庆，尚岳全. 基于粒子群算法和广义回归神经网络的岩爆预测[J]. 岩石力学与工程学报，2013，32(2)：343-348.(JIA Yipeng，LU Qing，SHANG Yuequan. Rockburst prediction based on particle swarm optimization and generalized regression neural network[J]. Chinese Journal of Rock Mechanics and Engineering，2013，32(2)：343-348.(in Chinese))
[45]	张俊峰. 大相岭隧道岩爆灾害分阶段预测与控制技术研究[硕士学位论文][D]. 成都：西南交通大学，2010.(ZHANG Junfeng. Research on staged prediction and control technology of rockburst disaster in Daxiangling Tunnel[M. S. Thesis][D]. Chengdu：Southwest Jiaotong University，2010.(in Chinese))
[49]	王东. 基于多个元模型的Stacking算法研究与应用[硕士学位论文][D]. 广州：华南理工大学，2020.(WANG Dong. Research and application of stacking algorithm based on multiple metamodels[M. S. Thesis][D]. Guangzhou：South China University of Technology，2020.(in Chinese))
[50]	李杨，陈子彬，谢光强. 一种基于ExtraTrees的差分隐私保护算法[J]. 计算机工程，2020，46(2)：134-140.(LI Yang，CHEN Zibin，XIE Guangqiang. A differential privacy protection algorithm based on ExtraTrees[J]. Computer Engineering，2020，46(2)：134-140.(in Chinese))
[51]	PUNMIYA R，CHOE S. Energy theft detection using gradient boosting theft detector with feature engineering-based preprocessing[J]. IEEE Trans Smart Grid，2019，10(2)：2 326-2 329.
[53]	罗森林，赵惟肖，潘丽敏. 结合加权KNN和自适应牛顿法的稳健Boosting方法[J]. 北京理工大学学报，2021，41(1)：112-120. (LUO Senlin，ZHAO Weixiao，PAN Limin. Robust boosting method combining weighted KNN and adaptive Newton method[J]. Journal of Beijing Institute of Technology，2021，41(1)：112-120.(in Chinese))
[57]	王羽，许强，柴贺军，等. 工程岩爆灾害判别的RBF-AR耦合模型[J]. 吉林大学学报：地球科学版，2013，43(6)：1 943-1 949. (WANG Yu，XU Qiang，CHAI Hejun，et al. RBF-AR coupling model for engineering rockburst disaster identification[J]. Journal of Jilin University：Earth Science，2013，43(6)：1 943-1 949.(in Chinese))