基于数据预处理技术并考虑围岩应力梯度影响的隧洞岩爆预测

夏元友; 张宏伟; 吝曼卿; 阎要锋

doi:10.11779/CJGE20220701

基于数据预处理技术并考虑围岩应力梯度影响的隧洞岩爆预测 English Version

1.
武汉理工大学土木工程与建筑学院, 湖北武汉 430070
2.
武汉工程大学资源与安全工程学院, 湖北武汉 430074

基金项目:

国家自然科学基金项目 42077228

国家自然科学基金项目 52174085

详细信息

作者简介:
夏元友(1965—)，男，教授，博士生导师，主要从事岩土工程教学和科研工作。E-mail: xiayy1965@whut.edu.cn

中图分类号: TU431
计量
- 文章访问数: 0
- HTML全文浏览量: 0
- PDF下载量: 0
出版历程
- 收稿日期: 2022-05-31
- 网络出版日期: 2023-10-16
- 刊出日期: 2023-09-30

Prediction of tunnel rockbursts based on data preprocessing technology considering influences of stress gradient of surrounding rock

1.
School of Civil Engineering and Architecture, Wuhan University of Technology, Wuhan 430070, China
2.
School of Resources and Safety Engineering, Wuhan Institute of Technology, Wuhan 430074, China

摘要

摘要: 针对目前岩爆预测研究通常忽视岩爆数据集存在离群样本、缺失值与样本不平衡性问题以及围岩应力梯度的影响，提出一套完备的岩爆数据预处理流程，引入可间接表征围岩应力梯度的洞径指标，建立了隧洞岩爆多因素综合预测模型。在数据采集阶段，考虑隧道与采场及隧洞群受力条件差异，从岩爆数据库中分离出隧洞岩爆样本共306例。在岩爆预测指标选取阶段，选取隧洞洞径D₀、围岩最大切向应力 ${\sigma _{\theta \max }}$ 、岩石单轴抗压强度 ${\sigma _{\text{c}}}$ 、岩石抗拉强度 ${\sigma _{\text{t}}}$ 、弹性能变形指数W_et共5个指标。在数据预处理阶段：针对缺失值，引入随机森林多重插补法（MI-RF）对岩爆样本进行补全；针对离群样本，引入最近邻（KNN）、孤立森林（Isolation Forest）、局部异常因子（LOF）3种无监督算法综合评估岩爆数据集并剔除离群样本；针对样本不平衡，引入自适应综合过采样（ADASYN）算法扩容少数类样本。在模型验证阶段：采用支持向量机（SVM）、随机森林（RF）、梯度提升树（GBDT）、自适应提升树（AdaBoost）、极限梯度提升树（XGBoost）5类算法构建岩爆预测模型。模型预测结果表明：基于数据预处理并考虑洞径指标的5类模型皆为同类算法模型中的最优；在不进行数据预处理的条件下，考虑洞径指标模型要优于不考虑洞径指标的同类算法模型。
- 地下工程 /
- 岩爆预测 /
- 数据预处理 /
- 围岩应力梯度 /
- 洞径
Abstract: As the current rockburst prediction investigation frequently ignores outliers, missing values, sample imbalance in the rockburst dataset and the influences of surrounding rock stress gradient, a complete preprocessing process of rockburst data is proposed, and the hole diameter index that indirectly represents the stress gradient of surrounding rock of tunnel is employed to establish the multi-factor comprehensive prediction model for tunnel rockbursts. At the stage of the data collection, considering the variation in stress conditions between the tunnel, stope and tunnel group, 306 samples of rockbursts in tunnels are isolated from the rockburst database. At the stage of determining prediction index, five indices are selected including the hole diameter (D₀), the maximum tangential stress ( ${\sigma _{\theta \max }}$ ), the uniaxial compressive strength ( ${\sigma _{\text{c}}}$ ), the uniaxial tensile strength of the rock (σ_t) and the elastic energy deformation index (W_et). At the stage of the data preprocessing, the multiple imputation method of random forest (MI-RF) is introduced to fill in the missing values. Three unsupervised algorithms including the K-nearest neighbor (KNN), the isolation forest (IForest) and the local outlier factor (LOF) are introduced to comprehensively evaluate the rockburst dataset and removed outliers. The adaptive comprehensive oversampling (ADASYN) algorithm is introduced to expand the number of minority samples. At the stage of the model validation, five types of models including the support vector machine (SVM), the random forest (RF), the gradient boosted decision trees (GBDT), the adaptive boosting algorithm (AdaBoost) and the extreme gradient boosting algorithm (XGBoost) are adopted for comparison. The results demonstrate that the aforementioned models based on the data preprocessing and the hole diameter index are all the best among similar algorithm models. Without the data preprocessing, the model considering the hole diameter index is better than those without considering the hole diameter.
- underground engineering /
- rockburst prediction /
- data preprocessing /
- stress gradient of surrounding rock /
- hole diameter

HTML全文

图 1 围岩切向应力分布

Figure 1. Distribution of tangential stress around surrounding rock

下载: 全尺寸图片幻灯片

图 2 岩爆案例的五因素统计学关系图例

Figure 2. Statistical relationship of five factors for rockburst samples

下载: 全尺寸图片幻灯片

图 3 岩爆数据预处理流程

Figure 3. Flow chart of data preprocessing of rockbursts

下载: 全尺寸图片幻灯片

图 4 缺失岩爆数据分布图

Figure 4. Distribution of missing rockburst data

下载: 全尺寸图片幻灯片

图 5 岩爆插补数据分布

Figure 5. Distribution of imputation data of rockburst

下载: 全尺寸图片幻灯片

离群样本散点图( ${\sigma _{\text{c}}}$ 与 ${\sigma _{\text{t}}}$ )

Outlier scatter ( ${\sigma _{\text{c}}}$ _VS ${\sigma _{\text{t}}}$ )

下载: 全尺寸图片幻灯片

图 7 隧洞岩爆预测流程

Figure 7. Flow chart of predicting tunnel rockburst

下载: 全尺寸图片幻灯片

表 1 隧洞岩爆案例集

Table 1 Dataset of tunnel rockburst

样本编号	D₀/ m	${\sigma _{\theta {\text{max}}}}$ / MPa	${\sigma _{\text{c}}}$ / MPa	${\sigma _{\text{t}}}$ / MPa	W_et	岩爆烈度
1	9.58	30	88.7	3.7	6.6	3
2	10.83	30	88.7	3.7	6.6	3
3	5.00	90	170.0	11.3	9.0	4
…	…	…	…	…	…	…
304	4.72^*	60	86.0	7.14	2.85	2
305	4.72^*	60	145.2	9.3	3.5	2
306	4.72^*	60	136.8	10.4	2.12	2
注：“*”表示D₀为换算值；第七列中“1”为无岩爆，“2”为弱岩爆，“3”为中等岩爆，“4”为强岩爆。

下载: 导出CSV

表 2 换算系数

Table 2 Conversion factors

断面形状	椭圆	拱形	正方形	正梯形	长方形	单边斜梯
换算系数	1.05	1.1	1.15	1.2	1.2	1.25

下载: 导出CSV

表 3 插补数据集的总体R²检验

Table 3 R²-tests for imputation dataset

项目	估计值	估计值95% 置信上限	估计值95% 置信下限	评价标准
R²	0.52	0.60	0.42	相关性较强

下载: 导出CSV

表 4 岩爆样本的离群样本分数

Table 4 Outlier scores for rockburst samples

样本编号	1	2	3	...	304	305	306
分数	-0.51	-0.55	0.97	...	0.12	0.08	-0.08
识别	0	0	1	...	0	0	0
注：“0”表示正常样本，“1”表示离群样本。

下载: 导出CSV

表 5 考虑数据预处理及洞径指标的岩爆预测

Table 5 Rockburst prediction considering data preprocessing and hole diameter index

训练轮次	预测准确率/%
训练轮次	SVM	RF	GBDT	Adaboost	XGBoost
1	57.1	57.1	57.1	57.1	53.6
2	67.9	67.9	74.9	67.9	67.9
3	42.9	50.0	50.0	50.0	50.0
4	67.9	71.4	71.4	71.4	71.4
5	71.4	71.4	67.9	71.4	67.9
6	67.9	78.6	75.0	71.4	71.4
7	60.7	67.9	50.0	53.6	53.6
8	75.0	78.6	75.0	78.6	78.6
9	82.1	85.7	82.1	89.3	85.7
10	64.3	67.9	71.4	67.9	67.9
平均值	65.7	69.7	67.5	67.9	66.8

下载: 导出CSV

表 6 考虑洞径指标但不考虑数据预处理的岩爆预测

Table 6 Rockburst prediction considering hole diameter index without data preprocessing

训练轮次	预测准确率/%
训练轮次	SVM	RF	GBDT	Adaboost	XGBoost
1	72	80	72	72	68
2	52	48	52	60	56
3	56	72	76	72	72
4	48	48	56	48	44
5	52	64	56	56	52
6	60	64	64	64	64
7	68	52	60	68	68
8	64	64	60	52	64
9	72	68	72	68	64
10	60	72	76	60	72
平均值	60.4	63.2	64.4	62.0	62.4

下载: 导出CSV

表 7 不考虑数据预处理及洞径指标的岩爆预测

Table 7 Rockburst prediction without data preprocessing and hole diameter index

训练轮次	预测准确率/%
训练轮次	SVM	RF	GBDT	Adaboost	XGBoost
1	72	72	72	72	68
2	60	56	60	56	56
3	64	64	56	68	60
4	40	40	48	32	44
5	48	52	52	48	48
6	52	52	52	56	56
7	56	56	64	68	56
8	56	68	64	52	60
9	52	64	68	64	64
10	60	72	64	72	64
平均值	56.0	59.6	60.0	58.8	57.6

下载: 导出CSV

表 8 是否考虑数据预处理的多模型预测准确率对比

Table 8 Comparison of model prediction accuracy with and without data preprocessing

条件	平均预测准确率/%
条件	SVM	RF	GBDT	Adaboost	XGBoost
数据预处理考虑洞径	65.7	69.7	67.5	67.9	66.8
原始数据集考虑洞径	60.4	63.2	64.4	62.0	62.4
差值	+5.3	+6.5	+3.1	+5.9	+4.4

下载: 导出CSV

表 9 是否考虑洞径指标的多模型预测准确率对比

Table 9 Comparison of model prediction accuracy with and without hole diameter index

条件	平均预测准确率/%
条件	SVM	RF	GBDT	Adaboost	XGBoost
原始数据集考虑洞径	60.4	63.2	64.4	62.0	62.4
原始数据集不考虑洞径	56.0	59.6	60.0	58.8	57.6
差值	+4.4	+3.6	+4.4	+3.2	+4.8

下载: 导出CSV

表 10 地下洞室的岩爆特征参数

Table 10 Rockburst characteristics of underground caverns

开挖段	主厂房岩爆特征参数
开挖段	D₀/ m	${\sigma _{\theta \max }}$ / MPa	σ_c/ MPa	σ_t/ MPa	W_et	实际烈度
中导洞Ⅰ区	11.2^*	90	97.3	6.03	6.6^#	3~4
扩挖Ⅱ区	14.8^*	90	97.3	6.03	5.5^#	3~4
扩挖Ⅲ区	17.7^*	90	97.3	6.03	6.6^#	3~4
注：“*”表示D₀为换算值；“#”表示W_et为插补值。

下载: 导出CSV

表 11 地下洞室的岩爆预测

Table 11 Prediction of rockburst for underground caverns

开挖段	主厂房岩爆倾向性预测
开挖段	SVM	RF	GBDT	Adaboost	XGBoost
中导洞Ⅰ区	3	4	4	4	4
扩挖Ⅱ区	3	3	3	4	3
扩挖Ⅲ区	3	3	4	4	4

下载: 导出CSV

表 12 苍岭隧道的岩爆特征参数

Table 12 Rockburst characteristics of Cangling Tunnel

区段	D₀/ m	${\sigma _{\theta \max }}$ / MPa	σ_c/ MPa	σ_t/ MPa	W_et	实际烈度
1	11.6^*	32.8	160	6.6	4.6	2
2	11.6^*	44.8	160	6.8	4.9	2
3	11.6^*	50.9	160	7.5	5.3	3
4	11.6^*	44.8	160	6.7	4.8	2
注：“*”表示D₀为换算值。

下载: 导出CSV

表 13 苍岭隧道的岩爆预测

Table 13 Prediction of rockburst for Cangling Tunnel

区段	SVM	RF	GBDT	Adaboost	XGBoost
1	2	2	2	2	2
2	2	2	2	2	2
3	4	3	3	3	3
4	2	2	2	2	2

下载: 导出CSV

参考文献(26)

[1]	FENG X T, LIU J, CHEN B, et al. Monitoring, warning, and control of rockburst in deep metal mines[J]. Engineering, 2017, 3(4): 538-545. doi: 10.1016/J.ENG.2017.04.013
[2]	宫凤强, 潘俊锋, 江权. 岩爆和冲击地压的差异解析及深部工程地质灾害关键机理问题[J]. 工程地质学报, 2021, 29(4): 933-961. https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202104003.htm GONG Fengqiang, PAN Junfeng, JIANG Quan. The difference analysis of rock burst and coal burst and key mechanisms of deep engineering geological hazards[J]. Journal of Engineering Geology, 2021, 29(4): 933-961. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202104003.htm
[3]	张传庆, 俞缙, 陈珺, 等. 地下工程围岩潜在岩爆问题评估方法[J]. 岩土力学, 2016, 37(增刊1): 341-349. https://www.cnki.com.cn/Article/CJFDTOTAL-YTLX2016S1046.htm ZHANG Chuanqing, YU Jin, CHEN Jun, et al. Evaluation method for potential rockburst in underground engineering[J]. Rock and Soil Mechanics, 2016, 37(S1): 341-349. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YTLX2016S1046.htm
[4]	徐琛, 刘晓丽, 王恩志, 等. 基于组合权重-理想点法的应变型岩爆五因素预测分级[J]. 岩土工程学报, 2017, 39(12): 2245-2252. doi: 10.11779/CJGE201712013 XU Chen, LIU Xiaoli, WANG Enzhi, et al. Prediction and classification of strain mode rockburst based on five-factor criterion and combined weight-ideal point method[J]. Chinese Journal of Geotechnical Engineering, 2017, 39(12): 2245-2252. (in Chinese) doi: 10.11779/CJGE201712013
[5]	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: 224-232. doi: 10.1016/j.ijrmms.2015.02.004
[6]	贾义鹏, 吕庆, 尚岳全, 等. 基于证据理论的岩爆预测[J]. 岩土工程学报, 2014, 36(6): 1079-1086. doi: 10.11779/CJGE201406013 JIA Yipeng, LÜ Qing, SHANG Yuequan, et al. Rockburst prediction based on evidence theory[J]. Chinese Journal of Geotechnical Engineering, 2014, 36(6): 1079-1086. (in Chinese) doi: 10.11779/CJGE201406013
[7]	高玮. 基于蚁群聚类算法的岩爆预测研究[J]. 岩土工程学报, 2010, 32(6): 874-880. http://www.cgejournal.com/cn/article/id/13417 GAO WEI. Prediction of rock burst based on ant colony clustering algorithm[J]. Chinese Journal of Geotechnical Engineering, 2010, 32(6): 874-880. (in Chinese) http://www.cgejournal.com/cn/article/id/13417
[8]	PU Y, APEL D B, WANG C, et al. Evaluation of burst liability in kimberlite using support vector machine[J]. Acta Geophysica, 2018, 66: 973-982.
[9]	WANG C, CHUAI X, SHI F, et al. Experimental investigation of predicting rockburst using Bayesian model[J]. Geomechanics & engineering, 2018, 15(6): 1153-1160.
[10]	谭文侃, 叶义成, 胡南燕, 等. LOF与改进SMOTE算法组合的强烈岩爆预测[J]. 岩石力学与工程学报, 2021, 40(6): 1186-1194. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202106010.htm TAN Wenkan, YE Yicheng, HU Nanyan, et al. Severe rock burst prediction based on the combination of LOF and improved SMOTE algorithm[J]. Chinese Journal of Rock Mechanics and Engineering, 2021, 40(6): 1186-1194. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202106010.htm
[11]	YIN X, LIU Q, PAN Y, et al. Strength of stacking technique of ensemble learning in rockburst prediction with imbalanced data: Comparison of eight single and ensemble models[J]. Natural Resources Research, 2021, 30(2): 1795–1815.
[12]	夏元友, 刘昌昊, 刘夕奇, 等. 均布与梯度应力加载路径下岩爆破坏特征试验[J]. 中国安全科学学报, 2020, 30(5): 149-155. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202005025.htm XIA Yuanyou, LIU Changhao, LIU Xiqi, et al. Experimental study on rockburst characteristics under uniform and gradient stress loading paths[J]. China Safety Science Journal, 2020, 30(5): 149-155. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202005025.htm
[13]	何佳其, 吝曼卿, 刘夕奇, 等. 引入梯度应力的岩爆预测方法[J]. 岩土工程学报, 2020, 42(11): 2098–2105. doi: 10.11779/CJGE202011015 HE Guiqi, LIN Manqing, LIU Xiqi, et al. New method for introducing gradient stress into rock-burst prediction[J]. Chinese Journal of Geotechnical Engineering, 2020, 42(11): 2098-2105. (in Chinese) doi: 10.11779/CJGE202011015
[14]	刘长武, 曹磊, 刘树新. 深埋非圆形地下洞室围岩应力解析分析的"当量半径"法[J]. 铜业工程, 2010(1): 1-5. https://www.cnki.com.cn/Article/CJFDTOTAL-TYGC201001003.htm LIU Changwu, CAO Lei, LIU Shuxin. Method of "equivalent radius" for the analyzing rock stress of high-buried non-circular underground chambers[J]. Copper Engineering, 2010(1): 1-5. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-TYGC201001003.htm
[15]	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.
[16]	李准, 苗勇刚, 夏志远. 深部矿山巷道岩爆倾向性预测及防治技术研究[J]. 有色金属(矿山部分), 2021, 73(3): 41-47, 60. https://www.cnki.com.cn/Article/CJFDTOTAL-YSKU202103008.htm LI Zhun, MIAO Yonggang, XIA Zhiyuan. Study on rockburst tendency prediction and prevention technology of deep mine roadway[J]. Nonferrous Metals (Mining Section), 2021, 73(3): 41-47, 60. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YSKU202103008.htm
[17]	QIU D H, LI T B, XUE Y G, et al. Rockburst prediction based on distance discrimination method and optimization technology-based weight calculation method[J]. Electronic Journal of Geotechnical Engineering, 2014, 19: 1843-1854.
[18]	王庆武. 拉林铁路桑珠岭隧道地应力场分析及岩爆预测研究[D]. 成都: 成都理工大学, 2017. WANG Qinwu. Analysis of Geostress Field and Rockburst Prediction in the Sang Zhuling Tunnel of Lasa-Linzhi Railway Engineering[D]. Chengdu: Chengdu University of Technology, 2017. (in Chinese)
[19]	易迪青. 沪昆高铁雪峰山隧道岩爆灾害评估与防治[D]. 北京: 北京交通大学, 2014. YI Diqing. Rock Burst Disaster Assessment and Prevention in Xuefeng Mountain Tunnel of Shanghai-Kunming High-Speed Railway[D]. Beijing: Beijing Jiaotong University, 2014. (in Chinese)
[20]	XUE Y G, BAI C H, KONG F M, et al. A two-step comprehensive evaluation model for rockburst prediction based on multiple empirical criteria[J]. Engineering Geology, 2020, 268: 105515.
[21]	WU S C, WU Z G, ZHANG C X. Rock burst prediction probability model based on case analysis[J]. Tunnelling and Underground Space Technology, 2019, 93(03069): 1-15.
[22]	JIA Q J, WU L, LI B, et al. The comprehensive prediction model of rockburst tendency in tunnel based on optimized unascertained measure theory[J]. Geotechnical and Geological Engineering, 2019, 37: 3399-3411.
[23]	XUE Y G, LI Z Q, LI S C, et al. Prediction of rock burst in underground caverns based on rough set and extensible comprehensive evaluation[J]. Bulletin of engineering geology and the environment, 2019, 78(1): 417-429.
[24]	LIU F T, TING K M, ZHOU Z H. Isolation-based anomaly detection[J]. Acm Transactions on Knowledge Discovery from Data, 2012, 6(1): 1–39.
[25]	HE H, BAI Y, GARCIA E A, et al. ADASYN: Adaptive synthetic sampling approach for imbalanced learning[C]// 2008 IEEE International Joint Conference on Neural Networks, HongKong, 2008.
[26]	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. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201302015.htm JIA Yipeng, LÜ Qing, SHANG Yuequan. Rockburst prediction using particle swarm optimization algorithm and general regression neural network[J]. Chinese Journal of Rock Mechanics and Engineering, 2013, 32(2): 343-348. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201302015.htm