In order to enhance the precision of short-term rockburst risk prediction during the excavation of deep hard rock,a prediction methodology utilizing microseismic(MS) information was investigated.An analysis was conducted on the correlation between MS parameters and rockburst risk levels using 103 sets of MS sample data.Six MS parameters were identified as predictive indices:The number of MS events (N),MS energy (E),MS apparent volume (V),event rate(NR),energy rate(ER),and apparent volume rate(VR).We introduce a novel approach for forecasting short-term rockburst risk levels utilizing the CatBoost integrated learning algorithm.The model parameters of CatBoost were optimized using particle swarm optimization(PSO).Following the construction of the model and the assessment of various performance metrics,the proposed method demonstrated superior test accuracy,reaching up to 90%,compared to other models employed in this study,including CatBoost,random forest,XGBoost,backpropagation neural network,and logistic regression algorithms,the proposed method demonstrated improvements of 9%,4%,9%,19% and 14%,respectively.Subsequently,the method was applied to seven challenging hard rock engineering cases,including the Qinling tunnel,the Xinjiang Ashele copper mine,the New Jersey hydroelectric tunnel in Pakistan,and the Jinping Ⅱ hydropower station,for validation purposes.The predicted outcomes were in alignment with the actual results.In comparison to similar methodologies,the prediction accuracy and engineering applicability of this model were superior,offering a scientific reference for short-term rockburst risk level prediction during deep underground engineering construction.
岩爆是在高应力环境下地下围岩体的一种动态失稳现象,通常储存在岩体中的应变能会以突发且剧烈的方式释放出来,并伴有岩石碎片的剥离和崩出现象(李夕兵等,2011,2019)。随着地下工程持续向深部推进,深部“三高一扰动”的复杂环境和岩体自身性质等因素使得岩爆发生的频率和风险增大,往往导致工程项目暂停、人员伤亡和设备损坏,并带来巨大的经济损失(蔡美峰等,2013;冯夏庭等,2013;钱七虎,2014;江飞飞等,2019;Roohollah et al.,2020),因此,及时准确地预测岩爆的发生,并采取措施进行预警和防范,对于确保地下工程施工安全至关重要。
历史上,最早的岩爆记录可追溯至1738年英国莱比锡煤矿的岩爆事件,此后,波兰、加拿大和德国也相继发生不同程度的岩爆灾害,我国同样饱受岩爆灾害困扰(周煦桐,2021)。国内外学者对岩爆的成因机制、影响因素及防控措施进行了深入研究,并积极探索岩爆预测方法。岩爆预测可划分为长期风险预测和短期风险评估2类(Liang et al.,2020),长期预测主要针对工程勘察和设计阶段,而短期评估对于施工过程中的实时有效预警和防治岩爆灾害具有重要意义。
尽管现有模型在一定程度上提高了岩爆预测的准确性,但多数模型由于其结构的复杂性和容易陷入局部最优的特性,使得预测结果解释性较差,因此,探究新的预测理论和方法用于短期岩爆预测显得非常必要。在各机器学习算法中,分类增强算法(CatBoost)是在梯度提升决策树(GBDT)框架下的一种改进的提升算法,以对称决策树作为基础学习器。与传统算法相比,该算法在处理分类特征、梯度偏差和预测偏差等方面具有较好的性能,有效提高了其精度和泛化性(Samat et al.,2021)。为提高短期岩爆预测的准确率,本文提出了基于粒子群(PSO)优化CatBoost的集成学习算法,构建了PSO-CatBoost岩爆短期预测模型,并将其应用于多个硬岩岩爆工程案例中,验证模型的性能,以期为深部地下工程灾害预防和安全施工提供参考。
1 数据收集与指标选取
1.1 短期岩爆数据库
根据Feng et al.(2019)和孙嘉豪等(2024)的研究成果,微震事件数(N)反映了岩体在高应力作用下受压破裂产生的微裂隙数量和密度,微震能量(E)和视体积(V)反映了微裂隙的强度和大小,微震事件率(NR)、能量率(ER)和视体积率(VR)反映了微裂隙随时间效应的变化程度。这些评价指标与岩爆灾害形成过程中的微裂缝活动密切相关,综合反映了岩体总破裂时间、强度和变形积累,能够较好地表征岩爆发展趋势,反映岩体应力状态和破裂情况,从而实时预测岩爆的发展趋势和严重程度,已被广泛应用于短期岩爆预测(Liu et al.,2021;Jin et al.,2022;Ma et al.,2023)。本文在现有研究的基础上,考虑将N、E、V、NR、ER和VR这6项微震参数作为短期岩爆评价指标。根据现场破坏特征和相关研究(吴顺川等,2019),可将岩爆烈度划分为无岩爆、轻微岩爆、中等岩爆和强岩爆4个等级,用标签N(None)、L(Light)、M(Moderate)和H(High)分别代表这4个岩爆等级。
国内外众多地下岩土工程均发生过多起岩爆事故,本研究搜集了包括矿山和隧道等在内的103组不同岩土工程项目的真实岩爆案例数据来构建微震样本数据库(冯夏庭等,2013;Feng et al.,2019;Liu et al.,2021)。其中,无岩爆、轻微岩爆、中等岩爆和强岩爆分别有37组、27组、27组和12组,各自占到总样本数的35.9%、26.2%、26.2%和11.7%,各类别样本数量分布情况如图1所示。每组数据均包含以上6个参数,这些样本能够很好地表示地下硬岩工程实施中的岩爆微震数据变化特征,部分样本数据如表1所示。
CatBoost是在GBDT框架下针对XGBoost和轻量级梯度提升机(LightGBM)模型进行优化的一种改良Boosting集成算法。相比传统算法,该算法利用完全对称二叉树进行梯度增强,通过克服梯度偏差和预测偏移的困难来减少发生过拟合的情况(Huang et al.,2019)。主要的创新措施包括:
CaiMeifeng, JiDong, GuoQifeng,2013.Study of rockburst prediction based on in-situ stress measurement and theory of energy accumulation caused by mining disturbance[J].Chinese Journal of Rock Mechanics and Engineering,32(10):1973-1980.
[2]
CaiW, DouL M, ZhangM,et al,2018.A fuzzy comprehensive evaluation methodology for rock burst forecasting using microseismic monitoring[J].Tunnelling and Underground Space Technology,80:232-245.
[3]
ChenY Y, XiaoP W, LiP,et al,2022.Formation mechanism of rockburst in deep tunnel adjacent to faults:Implication from numerical simulation and microseismic monitoring[J].Journal of Central South University,29(12):4035-4050.
[4]
FengG G, XiaG Q, ChenB R,et al,2019.A method for rockburst prediction in the deep tunnels of hydropower stations based on the monitored microseismicity and an optimized probabilistic neural network model[J].Sustainability,11(11):3212.
[5]
FengG L, ChenB R, XiaoY X,et al,2022.Microseismic characteristics of rockburst development in deep TBM tunnels with alternating soft-hard strata and application to rockburst warning:A case study of the Neelum-Jhelum hydropower project[J].Tunnelling and Underground Space Technology,122:104398.
[6]
FengG L, FengX T, ChenB R,et al,2014.A microseismic method for dynamic warning of rockburst development processes in tunnels[J].Rock Mechanics and Rock Engineering,48(5):2061-2076.
[7]
FengXiating, ChenBingrui, ZhangChuanqing,et al,2013.Mechanism,Warning and Dynamic Control of Rockburst Development Processes[M].Beijing:Science Press.
[8]
JiangFeifei, ZhouHui, LiuChang,et al,2019.Progress,prediction and prevention of rockbursts in underground metal mines[J].Chinese Journal of Rock Mechanics and Engineering,38(5):956-972.
[9]
HeH, DouL M, GongS Y,et al,2019.Microseismic and electromagnetic coupling method for coal bump risk assessment based on dynamic static energy principles[J].Safety Science,114:30-39.
[10]
HuL, FengX T, YaoZ Z,et al,2023.Rockburst time warning method with blasting cycle as the unit based on microseismic information time series:A case study[J].Bulletin of Engineering Geology and the Environment,82(4):121.
[11]
HuangG M, WuL F, MaX,et al,2019.Evaluation of CatBoost method for prediction of reference evapotranspiration in humid regions[J].Journal of Hydrology,574:1029-1041.
[12]
JinA B, BasnetP M S, MahtabS,2022.Microseismicity-based short-term rockburst prediction using non-linear support vector machine[J].Acta Geophysica,70(4):1717-1736.
[13]
LiXibing, GongFengqiang, WangShaofeng,et al,2019.Coupled static-dynamic loading mechanical mechanism and dynamic criterion of rockburst in deep hard rock mines[J].Chinese Journal of Rock Mechanics and Engineering,38(4):708-723.
[14]
LiXibing, YaoJinrui, GongFengqiang,2011.Dynamic problems in deep exploitation of hard rock metal mines[J].The Chinese Journal of Nonferrous Metals,21(10):2551-2563.
LiangWeizhang, ZhaoGuoyan,2022.A review of long-term and short-term rockburst risk evaluations in deep hard rock[J].Chinese Journal of Rock Mechanics and Engineering,41(1):19-39.
[17]
LiuG F, JiangQ, FengG L,et al,2021.Microseismicity-based method for the dynamic estimation of the potential rockburst scale during tunnel excavation[J].Bulletin of Engineering Geology and the Environment,80(5):3605-3628.
[18]
LiuJ P, FengX T, LiY H,et al,2013.Studies on temporal and spatial variation of microseismic activities in a deep metal mine[J].International Journal of Rock Mechanics and Mining Sciences,60:171-179.
[19]
LiuJ X, YangS, LiL,et al,2024.Precursory characteristics of large-scale rockmass collapse based on multiple microseismic parameters[J].Journal of Central South University,31(3):878-889.
[20]
MaK, ShenQ Q, SunX Y,et al,2023.Rockburst prediction model using machine learning based on microseismic parameters of Qinling water conveyance tunnel[J].Journal of Central South University,30(1):289-305.
[21]
MaT H, TangC A, TangS B,et al,2018.Rockburst mechanism and prediction based on microseismic monitoring[J].International Journal of Rock Mechanics and Mining Sciences,110:177-188.
[22]
QianQihu,2014.Definition,mechanism,classification and quantitative forecast model for rockburst and pressure bump[J].Rock and Soil Mechanics,35(1):1-6.
[23]
RoohollahS F, TiAbbas, LuisR E S,et al,2020.Rockburst assessment in deep geotechnical conditions using true-triaxial tests and data-driven approaches[J].International Journal of Rock Mechanics and Mining Sciences,128:104279.
[24]
SamatA, LiE Z, DuP J,et al,2021.GPU-accelerated CatBoost-forest for hyperspectral image classification via parallelized mRMR ensemble subspace feature selection[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,14:3200-3214.
[25]
SrinivasanC, AroraS K, YajiR K,1997.Use of mining and seismological parameters as premonitors of rockbursts[J].Internation Journal of Rock Mechanics and Mining Sciences,34(6):1001-1008.
[26]
SunJiahao, WangWenjie, XieLianku,2024.Short-term rockburst prediction model based on microseismic monitoring and probability optimization naive Bayes[J].Rock and Soil Mechanics,45(6):1884-1894.
[27]
TangS B, LiuY H, XuHR,et al,2023.Review for the microseismic source location in surrounding rock of deep-buried tunnel[J].Journal of Central South University,30(12):4182-4196.
[28]
WuShunchuan, ZhangChenxi, ChengZiqiao,2019.Prediction of intensity classification of rockburst based on PCA-PNN principle[J].Journal of China Coal Society,44(9):2767-2776.
[29]
XuS D, YangX L, ZhangF F,et al,2022.Application of a microseismic method of rock burst risk assessment under blasting mining in Ashele copper mine[J].Shock and Vibration,2:1-13.
[30]
YinX, LiuQ S, HuangX,et al,2021.Real-time prediction of rockburst intensity using an integrated CNN-Adam-BO algorithm based on microseismic data and its engineering application[J].Tunnelling and Underground Space Technology,117:104133.
[31]
ZhouXutong,2021.Prediction of Rockbursts Based on BP Neural Network[D].Xiangtan:Xiangtan University.
[32]
ZhangJun, YangQingping, LiuFangfang,et al,2023.Design and research of distributed microseismic monitoring system in deep well large-scale mining mine[J].Gold Science and Technology,31(4):659-668.