Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2021, Vol. 55 Issue (3): 448-454    DOI: 10.3785/j.issn.1008-973X.2021.03.004
机械工程     
基于马尔可夫过程和深度神经网络的TBM围岩识别
毛奕喆(),龚国芳*(),周星海,王飞
浙江大学 流体动力与机电系统国家重点实验室，浙江 杭州 310027
Identification of TBM surrounding rock based on Markov process and deep neural network
Yi-zhe MAO(),Guo-fang GONG*(),Xing-hai ZHOU,Fei WANG
State Key Laboratory of Fluid Power and Mechatronic Systems, Zhejiang University, Hangzhou 310027, China
 全文: PDF(1291 KB)   HTML
摘要：

为了实现隧道围岩的实时识别，基于马尔可夫过程和深度神经网络模型，提出将先验围岩信息和掘进参数结合，作为深度神经网络输入的隧道掘进机（TBM）围岩实时识别方法. 根据施工现场地质勘探资料，用马尔可夫过程的隧道围岩分类方法预测隧道沿线的围岩分布概率；将该围岩分布概率作为先验围岩信息，结合TBM掘进参数作为神经网络输入，真实围岩类别作为输出，训练深度神经网络以实现对TBM前方围岩的实时识别. 使用工程现场数据进行对比实验，结果表明，所设计的深度神经网络模型的围岩总体识别率高于96%. 相比于仅将掘进参数作为输入，当结合先验围岩信息和掘进参数作为输入时，模型围岩识别率提高6%以上.
关键词： 隧道掘进机（TBM）;  马尔可夫过程;  深度神经网络;  围岩识别;  预测    
Abstract:

In order to realize the real-time identification of tunnel surrounding rock, based on the Markov method and the deep neural network model, a real-time identification method of tunnel boring machine (TBM) surrounding rock was proposed, which combined the prior surrounding rock information and the excavation parameters. The Markov process method of tunnel surrounding rock classification was used to predict the distribution probability of surrounding rock along the tunnel according to the geological exploration data of the construction site. The surrounding rock distribution probability was used as the prior rock information. The prior rock information and the TBM excavation parameters were used together as the input of the neural network, and the real surrounding rock category as an output. A deep neural network was trained to achieve real-time identification of the surrounding rock in front of the TBM. Experiments on practical engineering data showed that the overall recognition rate of the surrounding rock of the designed deep neural network model was above 96%. Compared with taking the excavation parameters as input, when the priori surrounding rock information and the excavation parameters were combined as input, the recognition rate of the surrounding rock of the model was increased by more than 6%.
Key words: tunnel boring machine (TBM)    Markov process    deep neural network    surrounding rock identification    prediction
收稿日期: 2019-11-10 出版日期: 2021-04-25
CLC:  TN 137  
基金资助: 国家重点研发计划资助项目（2018YFB1702503，2017YFB1302602，2017YFB1302604）
通讯作者: 龚国芳     E-mail: 21825058@zju.edu.cn;gfgong@zju.edu.cn
作者简介: 毛奕喆（1995—），男，硕士生，从事大型隧道掘进装备智能化研究. orcid.org/0000-0002-8523-8514. E-mail： 21825058@zju.edu.cn
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
毛奕喆
龚国芳
周星海
王飞

引用本文:

毛奕喆,龚国芳,周星海,王飞. 基于马尔可夫过程和深度神经网络的TBM围岩识别[J]. 浙江大学学报(工学版), 2021, 55(3): 448-454.

Yi-zhe MAO,Guo-fang GONG,Xing-hai ZHOU,Fei WANG. Identification of TBM surrounding rock based on Markov process and deep neural network. Journal of ZheJiang University (Engineering Science), 2021, 55(3): 448-454.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2021.03.004    或    http://www.zjujournals.com/eng/CN/Y2021/V55/I3/448

图 1  引松工程TBM施工总体图
图 2  引松工程围岩各级占比
图 3  含有异常值的TBM推进速度随时间变化曲线
图 4  去除异常值的TBM推进速度随时间变化
图 5  小波变换前、后推进速度的比较
图 6  DNN拓扑结构
图 7  根据17个观测点数据预测的围岩概率分布
图 8  5折交叉验证的过程
模型 围岩级别 P R F1
DNN II 0.99 0.97 0.98
III 0.96 0.98 0.97
IV 0.95 0.93 0.94
V 0.91 0.90 0.91
AVG 0.96 0.96 0.96
SVC II 0.97 0.93 0.95
III 0.93 0.92 0.92
IV 0.84 0.84 0.84
V 0.80 0.88 0.84
AVG 0.88 0.89 0.89
AdaBoost II 0.84 0.89 0.87
III 0.76 0.84 0.80
IV 0.62 0.51 0.56
V 0.73 0.72 0.72
AVG 0.74 0.74 0.74
表 1  DNN、SVC、AdaBoost围岩预测结果比较
图 9  有、无马尔可夫先验围岩信息作为输入DNN预测结果对比
数据集 围岩级别 P R
训练集 II 0.9976 0.9953
III 0.9754 0.9794
IV 0.9471 0.9542
V 0.9247 0.8677
AVG 0.9652 0.9652
测试集 II 0.9955 0.9911
III 0.9750 0.9744
IV 0.9361 0.9499
V 0.8894 0.8360
AVG 0.9595 0.9595
表 2  有马尔可夫先验围岩信息的DNN模型预测结果
数据集 围岩级别 P R
训练集 II 0.7747 0.7081
III 0.9580 0.9315
IV 0.8577 0.9479
V 0.8074 0.6884
AVG 0.8985 0.8985
测试集 II 0.7838 0.6829
III 0.9489 0.9325
IV 0.8657 0.9435
V 0.8000 0.7028
AVG 0.8967 0.8967
表 3  无马尔可夫先验围岩信息的DNN模型预测结果
1 杨华勇, 周星海, 龚国芳 对全断面隧道掘进装备智能化的一些思考[J]. 隧道建设, 2018, 38 (12): 1919- 1926
YANG Hua-yong, ZHOU Xing-hai, GONG Guo-fang Perspectives in intelligentization of tunnel boring machine (TBM)[J]. Tunnel construction, 2018, 38 (12): 1919- 1926
doi: 10.3973/j.issn.2096-4498.2018.12.001
2 HWANG J H, LU C C A semi-analytical method for analyzing the tunnel water inflow[J]. Tunnelling and Underground Space Technology, 2007, 22 (1): 39- 46
doi: 10.1016/j.tust.2006.03.003
3 SCHEPERS R, RAFAT G, GELBKE C, et al Application of borehole logging, core imaging and tomography to geotechnical exploration[J]. International Journal of Rock Mechanics and Mining Sciences, 2001, 38 (6): 867- 876
doi: 10.1016/S1365-1609(01)00052-1
4 ASHIDA Y Seismic imaging ahead of a tunnel face with three-component geophones[J]. International Journal of Rock Mechanics and Mining Sciences, 2001, 38 (6): 823- 831
doi: 10.1016/S1365-1609(01)00047-8
5 JETSCHNY S, BOHLEN T, KURZMANN A Seismic prediction of geological structures ahead of the tunnel using tunnel surface waves[J]. Geophysical Prospecting, 2011, 59 (5): 934- 946
6 LEE I M, TRUONG Q H, KIM D H, et al Discontinuity detection ahead of a tunnel face utilizing ultrasonic reflection: laboratory scale application[J]. Tunnelling and Underground Space Technology, 2009, 24 (2): 155- 163
doi: 10.1016/j.tust.2008.06.001
7 吴俊, 毛海和, 应松, 等 地质雷达在公路隧道短期地质超前预报中的应用[J]. 岩土力学, 2003, 24 (Suppl.1): 154- 157
WU Jun, MAO Hai-he, YING Song, et al Application of ground radar to short-term geological forecast for tunnel construction[J]. Rock and Soil Mechanics, 2003, 24 (Suppl.1): 154- 157
8 RISSAFI Y, TALBI L, GHADDAR M Experimental characterization of an UWB propagation channel in underground mines[J]. IEEE Transactions on Antennas and Propagation, 2011, 60 (1): 240- 246
9 刘新荣, 刘永权, 杨忠平, 等 基于地质雷达的隧道综合超前预报技术[J]. 岩土工程学报, 2015, 37 (2): 51- 56
LIU Xin-rong, LIU Yong-quan, YANG Zhong-pinng, et al Synthetic advanced deologicall prediction technology for tunnals based on GPR[J]. Chinese Journal of Geotechnical Engineering, 2015, 37 (2): 51- 56
10 ERHARTER G H, MARCHER T, REINHOLD C Application of artificial neural networks for underground construction – chances and challenges – insights from the BBT exploratory tunnel Ahrental Pfons[J]. Geomechanics and Tunnelling, 2019, 12 (5): 472- 477
doi: 10.1002/geot.201900027
11 ERHARTER G H, MARCHER T, REINHOLD C. Comparison of artificial neural networks for TBM data classification [C]// Proceedings of the 14th International Congress on Rock Mechanics and Rock Engineering. [S.l.]: CRC Press, 2019: 2426–2433 .
12 ZHANG Q, LIU Z, TAN J Prediction of geological conditions for a tunnel boring machine using big operational data[J]. Automation in Construction, 2019, 100: 73- 83
doi: 10.1016/j.autcon.2018.12.022
13 朱北斗, 龚国芳, 周如林, 等 基于盾构掘进参数的BP神经网络地层识别[J]. 浙江大学学报: 工学版, 2011, 45 (5): 851- 857
ZHU Bei-dou, GONG Guo-fang, ZHOU Ru-lin, et al Identification of strata with BP neural network based on parameters of shield driving[J]. Journal of Zhejiang University: Engineering Science, 2011, 45 (5): 851- 857
14 JUNG J H, CHUNG H, KWON Y S, et al An ANN to predict ground condition ahead of tunnel face using TBM operational data[J]. KSCE Journal of Civil Engineering, 2019, 23 (7): 3200- 3206
doi: 10.1007/s12205-019-1460-9
15 NIE S W, XUE L, JIA G P, et al. Identification of surrounding rock in TBM excavation with deep neural network [C]// 2019 2nd International Conference on Artificial Intelligence and Big Data. Guangzhou: [s.n.], 2019: 251-255.
16 MOAVENZADEH F, EINSTEIN H H, MARKOW M J, et al. Tunnel cost model: a stochastic simulation model of hard rock tunneling. volume 1. summary report [R]. Massachusetts: [s.n.], 1974.
17 CHAN M H C. A geological prediction and updating model in tunneling [D]. Massachusetts: Massachusetts Institute of Technology, 1981.
18 IOANNOU P G Geologic prediction model for tunneling[J]. Journal of Construction Engineering and Management, 1987, 113 (4): 569- 590
doi: 10.1061/(ASCE)0733-9364(1987)113:4(569)
19 刘东海, 周云晴, 王帅, 等 岩性Markov预测下的长隧洞TBM施工进度随机仿真分析[J]. 系统仿真学报, 2009, 21 (2): 558- 562
LIU Dong-hai, ZHOU Yun-qing, WANG Shuai, et al Stochastic simulation analysis of TBM construction progress of long tunnel under the prediction of lithology Markov[J]. Journal of System Simulation, 2009, 21 (2): 558- 562
20 徐琛, 胡程科, 刘晓丽, 等 基于Markov过程与信息熵法的长隧道勘探位置优化[J]. 地下空间与工程学报, 2018, 14 (6): 1611- 1617
XU Chen, HU Cheng-ke, LIU Xiao-li, et al Location optimization of long tunnel exploration based on the Markov method and information entropy[J]. Chinese Journal of Underground Space and Engineering, 2018, 14 (6): 1611- 1617
21 王芳芳, 李景叶, 陈小宏 基于马尔科夫链先验模型的贝叶斯岩相识别[J]. 石油地球物理勘探, 2014, 41 (9): 183- 189
WANG Fang-fang, LI Jing-ye, CHEN Xiao-hong Bayesian facies identification based on Markov chain prior model[J]. Petroleum Geophysical Exploration, 2014, 41 (9): 183- 189
22 赵海雷, 陈馈, 周建军, 等 引松供水4标TBM连续穿越灰岩的施工技术研究[J]. 隧道建设, 2017, 37 (3): 354- 362
ZHAO Hai-lei, CHEN Kui, ZHOU Jian-ju, et al Research on construction technologies of TBM continuous boring through the limestone section of bid section No. 4 of Songhua river water conveyance project[J]. Tunnel Construction, 2017, 37 (3): 354- 362
doi: 10.3973/j.issn.1672-741X.2017.03.015
23 段理文. TBM操作参数智能决策方法研究[D]. 杭州: 浙江大学, 2019.
DUAN Li-wen. Research on intelligent decision method of TBM operating parameters [D]. Hangzhou: Zhejiang University, 2019.
[1] 许佳辉,王敬昌,陈岭,吴勇. 基于图神经网络的地表水水质预测模型[J]. 浙江大学学报(工学版), 2021, 55(4): 601-607.
[2] 于勇,薛静远,戴晟,鲍强伟,赵罡. 机加零件质量预测与工艺参数优化方法[J]. 浙江大学学报(工学版), 2021, 55(3): 441-447.
[3] 黄发明,曹中山,姚池,姜清辉,陈佳武. 基于决策树和有效降雨强度的滑坡危险性预警[J]. 浙江大学学报(工学版), 2021, 55(3): 472-482.
[4] 苗发盛,吴益平,李麟玮,廖康,薛阳. 基于Boosting-决策树C5.0的岩体结构面粗糙度预测[J]. 浙江大学学报(工学版), 2021, 55(3): 483-490.
[5] 刘芳,汪震,刘睿迪,王锴. 基于组合损失函数的BP神经网络风力发电短期预测方法[J]. 浙江大学学报(工学版), 2021, 55(3): 594-600.
[6] 陈世达,刘强,韩亮. 降低分布式训练通信的梯度稀疏压缩方法[J]. 浙江大学学报(工学版), 2021, 55(2): 386-394.
[7] 段有康,陈小刚,桂剑,马斌,李顺芬,宋志棠. 基于相位划分的下肢连续运动预测[J]. 浙江大学学报(工学版), 2021, 55(1): 89-95.
[8] 陈纬奇,王敬昌,陈岭,杨勇勤,吴勇. 基于深度神经网络的多因素感知终端换机预测模型[J]. 浙江大学学报(工学版), 2021, 55(1): 109-115.
[9] 李文书,邹涛涛,王洪雁,黄海. 基于双尺度长短期记忆网络的交通事故量预测模型[J]. 浙江大学学报(工学版), 2020, 54(8): 1613-1619.
[10] 江冬冬,李道飞,俞小莉. 基于驾驶员需求转矩预测的模型预测控制能量管理[J]. 浙江大学学报(工学版), 2020, 54(7): 1325-1334.
[11] 王晨霖,杨洁,居文军,顾复,陈芨熙,纪杨建. 基于智能家电的短期电力负荷预测与削峰填谷优化[J]. 浙江大学学报(工学版), 2020, 54(7): 1418-1424.
[12] 闫旭,范晓亮,郑传潘,臧彧,王程,程明,陈龙彪. 基于图卷积神经网络的城市交通态势预测算法[J]. 浙江大学学报(工学版), 2020, 54(6): 1147-1155.
[13] 刘创,梁军. 基于注意力机制的车辆运动轨迹预测[J]. 浙江大学学报(工学版), 2020, 54(6): 1156-1163.
[14] 江金生,任浩然,李瀚野. 基于卷积自编码器的地震数据处理[J]. 浙江大学学报(工学版), 2020, 54(5): 978-984.
[15] 杨萍,王丹,康子健,李童,付利华,余悦任. 基于模式识别和集成CNN-LSTM的阵发性房颤预测模型[J]. 浙江大学学报(工学版), 2020, 54(5): 1039-1048.