基于深度学习技术的公路隧道围岩分级方法
|
摘要
通过深度学习技术提取公路隧道掌子面图片中的围岩分级相关信息。训练以掌子面图片和特征标签为数据集的深度卷积神经网络模型,识别围岩的节理、裂隙、破碎程度、粗糙程度、光滑程度、泥夹石和涌水等分布式特征;结合深度学习技术和岩体裂隙图像智能解译方法统计围岩节理组数和间距来描述结构面完整程度;再利用色彩模型确定岩石种类描述出岩石坚硬程度;最后将围岩分级各判别因子转换为BQ值进行分级,获得围岩分级最终结果。结果表明:深度学习模型适用于识别围岩不同形态特征,利用图像识别技术获取的围岩分级参数能够实现对公路隧道围岩等级的综合判定。该处理结果与传统BQ分级结果相吻合,验证了深度学习围岩分级的可行性和准确性。
By extracting the relevant information of surrounding rock classification of road tunnel face using the deep learning technology, a multilayer convolution neural network model is established to recognize the distributive features of surrounding rock including joints, cracks, broken situations, rough degrees, smooth degrees, mud stone and water burst, etc. The deep learning AlexNet model is modified to count the number and spacing of rock joints. The deep convolution is used to extract different rock boundaries, and the specific species of rock are determined by the comprehensive color model. The degree of development of structural plane, rock hardness, structural plane roughness, groundwater development, structural types and degree of grade factors of the surrounding rock classification are qualitatively described for the results of the surrounding rock classification so as to obtain the final results of the surrounding rock classification. The results show that the deep learning model is applicable to identify different morphological characteristics of the surrounding rock. Based on the Matlab interface technology, image recognition technology, boundary extraction technology and HIS color model, the comprehensive judgement of surrounding rock classification of highway tunnels is realized. In order to verify its feasibility and accuracy, the classification results of the deep learning technology are compared with those of the traditional BQ classification.
By extracting the relevant information of surrounding rock classification of road tunnel face using the deep learning technology, a multilayer convolution neural network model is established to recognize the distributive features of surrounding rock including joints, cracks, broken situations, rough degrees, smooth degrees, mud stone and water burst, etc. The deep learning AlexNet model is modified to count the number and spacing of rock joints. The deep convolution is used to extract different rock boundaries, and the specific species of rock are determined by the comprehensive color model. The degree of development of structural plane, rock hardness, structural plane roughness, groundwater development, structural types and degree of grade factors of the surrounding rock classification are qualitatively described for the results of the surrounding rock classification so as to obtain the final results of the surrounding rock classification. The results show that the deep learning model is applicable to identify different morphological characteristics of the surrounding rock. Based on the Matlab interface technology, image recognition technology, boundary extraction technology and HIS color model, the comprehensive judgement of surrounding rock classification of highway tunnels is realized. In order to verify its feasibility and accuracy, the classification results of the deep learning technology are compared with those of the traditional BQ classification.
引文
[1]方昱,刘保国,刘开云.隧道围岩分级的遗传–支持向量分类耦合模型[J].铁道学报,2013,35(1):108–114.(FANG Yu, LIU Bao-guo, LIU Kai-yun. Tunnel surrounding rockgradingmodelbasedongenetic-supportvector classificationcouplingalgorithm[J].JournaloftheChina Railway Society, 2013, 35(1):108–114.(in Chinese))
[2]段林娣,宋成辉.应用BP神经网络进行隧道围岩快速分级[J].中国安全科学学报,2010,20(2):41–45.(DUAN Lin-di, SONG Cheng-hui. Application of BP neural network intherapidclassificationofsurroundingrock[J].China Safety Science Journal, 2010, 20(2):41–45.(in Chinese))
[3]闫明.基于系统聚类分析的公路隧道围岩分级方法研究[D].吉林:吉林大学,2013.(YANMing.Studyon classificationmethodofroadtunnelsurroundingrock accordingtothehierarchicalclusteranalysis[D].Jilin:Jilin University, 2013.(in Chinese))
[4]白明洲,许兆义,王连俊,等.隧道围岩分级的模糊信息分析模型及应用研究[J].铁道学报,2001,23(6):85–88.(BAI Ming-zhou, XU Zhao-yi, WANG Lian-jun, et al. Study andapplicationonfuzzyinformationanalysismodelin tunnel surrounding rock mass classification[J]. Journal of the China Railway Society, 2001, 23(6):85–88.(in Chinese))
[5]陈安凡.公路隧道围岩分级专家系统研究[D].重庆:重庆大学,2013.(CHENAn-fan.Researchonroadtunnel surrounding rock classification expert system[D]. Chongqing:Chongqing University, 2013.(in Chinese))
[6]牛文林,张广洋,熊国斌,等.基于ASP.NET技术的围岩分级系统开发[J].现代隧道技术,2011,48(1):12–16.(NIU Wen-lin, ZHANG Guang-yang, XIONG Guo-bin, et al.RockmassclassificationsystembasedonASP.NET[J].ModernTunnellingTechnology,2011,48(1):12–16.(in Chinese))
[7]何发亮,谷明成,王石春.TBM施工隧道围岩分级方法研究[J].岩石力学与工程学报, 2002, 21(9):1350–1354.(HE Fa-liang,GUMing-cheng,WANGShi-chun.Studyon surrounding rock mass classification of tunnel cut by TBM[J].Chinese Journal of Rock Mechanics and Engineering, 2002,21(9):1350–1354.(in Chinese))
[8]王乐华,李建荣,李建林,等. RMR法评价体系的修正及工程应用[J].岩石力学与工程学报, 2013, 32(增刊2):3309–3316.(WANGLe-hua,LIJian-rong,LIJian-lin,etal.Correction of RMR system and its engineering application[J].Chinese Journal of Rock Mechanics and Engineering, 2013,32(S2):3309–3316.(in Chinese))
[9]张前进.基于TSP多波关系的围岩稳定性等级判定方法研究[D].北京:中国地质大学,2013.(ZHANGQian-jin.Researchonthemethodofgradedeterminationfor surroundingrockstabilitybasedonTSPmultiwave relation[D]. Beijing:China University of Geosciences, 2013.(in Chinese))
[10] HINTON G, DENG L, YU D, et al.Deep neural networks for acoustic modeling in speech recognition:the shared views of fourresearchgroups[J].IEEESignalProcessMag,2012,29(6):82.
[11] HINTON G E. Learning multiple layers of representation[J].Trends in Cognitive Sciences, 2007, 11(10):428–434.
[12] HINTON G E. Deep belief networks[J]. Scholarpedia, 2009,4(5):5947.
[13] JIA Yang-qing, SHEL Hamer E, DONAHUE J, et al. Caffe:Convolutionalarchitectureforfastfeatureembedding[C]//ProceedingsoftheACMInternationalConferenceon Multimedia. ACM, 2014:675–678.
[14]AlexKrizhevsky.ImageNetClassificationwithDeep ConvolutionalNeuralNetworks[C]//AdvancesinNeural Information Processing Systems, 2012:1097–1105.
[15]刘春,王宝军,施斌,等.基于数字图像识别的岩土体裂隙形态参数分析方法[J].岩土工程学报,2008,9(9):1383–1388.(LIUChun,WANGBao-jun,SHIBin,etal.Analyticmethodofmorphologicalparametersofcracksfor rock and soil based on image processing and recognition[J].ChineseJournalofGeotechnicalEngineering,2008,9(9):1383–1388.(in Chinese))
[16]GBT50218—2014《工程岩体分级标准》[S].2014.(GB T50218—2014Standardforengineeringclassificationof rock mass[S]. 2014.(in Chinese))
[2]段林娣,宋成辉.应用BP神经网络进行隧道围岩快速分级[J].中国安全科学学报,2010,20(2):41–45.(DUAN Lin-di, SONG Cheng-hui. Application of BP neural network intherapidclassificationofsurroundingrock[J].China Safety Science Journal, 2010, 20(2):41–45.(in Chinese))
[3]闫明.基于系统聚类分析的公路隧道围岩分级方法研究[D].吉林:吉林大学,2013.(YANMing.Studyon classificationmethodofroadtunnelsurroundingrock accordingtothehierarchicalclusteranalysis[D].Jilin:Jilin University, 2013.(in Chinese))
[4]白明洲,许兆义,王连俊,等.隧道围岩分级的模糊信息分析模型及应用研究[J].铁道学报,2001,23(6):85–88.(BAI Ming-zhou, XU Zhao-yi, WANG Lian-jun, et al. Study andapplicationonfuzzyinformationanalysismodelin tunnel surrounding rock mass classification[J]. Journal of the China Railway Society, 2001, 23(6):85–88.(in Chinese))
[5]陈安凡.公路隧道围岩分级专家系统研究[D].重庆:重庆大学,2013.(CHENAn-fan.Researchonroadtunnel surrounding rock classification expert system[D]. Chongqing:Chongqing University, 2013.(in Chinese))
[6]牛文林,张广洋,熊国斌,等.基于ASP.NET技术的围岩分级系统开发[J].现代隧道技术,2011,48(1):12–16.(NIU Wen-lin, ZHANG Guang-yang, XIONG Guo-bin, et al.RockmassclassificationsystembasedonASP.NET[J].ModernTunnellingTechnology,2011,48(1):12–16.(in Chinese))
[7]何发亮,谷明成,王石春.TBM施工隧道围岩分级方法研究[J].岩石力学与工程学报, 2002, 21(9):1350–1354.(HE Fa-liang,GUMing-cheng,WANGShi-chun.Studyon surrounding rock mass classification of tunnel cut by TBM[J].Chinese Journal of Rock Mechanics and Engineering, 2002,21(9):1350–1354.(in Chinese))
[8]王乐华,李建荣,李建林,等. RMR法评价体系的修正及工程应用[J].岩石力学与工程学报, 2013, 32(增刊2):3309–3316.(WANGLe-hua,LIJian-rong,LIJian-lin,etal.Correction of RMR system and its engineering application[J].Chinese Journal of Rock Mechanics and Engineering, 2013,32(S2):3309–3316.(in Chinese))
[9]张前进.基于TSP多波关系的围岩稳定性等级判定方法研究[D].北京:中国地质大学,2013.(ZHANGQian-jin.Researchonthemethodofgradedeterminationfor surroundingrockstabilitybasedonTSPmultiwave relation[D]. Beijing:China University of Geosciences, 2013.(in Chinese))
[10] HINTON G, DENG L, YU D, et al.Deep neural networks for acoustic modeling in speech recognition:the shared views of fourresearchgroups[J].IEEESignalProcessMag,2012,29(6):82.
[11] HINTON G E. Learning multiple layers of representation[J].Trends in Cognitive Sciences, 2007, 11(10):428–434.
[12] HINTON G E. Deep belief networks[J]. Scholarpedia, 2009,4(5):5947.
[13] JIA Yang-qing, SHEL Hamer E, DONAHUE J, et al. Caffe:Convolutionalarchitectureforfastfeatureembedding[C]//ProceedingsoftheACMInternationalConferenceon Multimedia. ACM, 2014:675–678.
[14]AlexKrizhevsky.ImageNetClassificationwithDeep ConvolutionalNeuralNetworks[C]//AdvancesinNeural Information Processing Systems, 2012:1097–1105.
[15]刘春,王宝军,施斌,等.基于数字图像识别的岩土体裂隙形态参数分析方法[J].岩土工程学报,2008,9(9):1383–1388.(LIUChun,WANGBao-jun,SHIBin,etal.Analyticmethodofmorphologicalparametersofcracksfor rock and soil based on image processing and recognition[J].ChineseJournalofGeotechnicalEngineering,2008,9(9):1383–1388.(in Chinese))
[16]GBT50218—2014《工程岩体分级标准》[S].2014.(GB T50218—2014Standardforengineeringclassificationof rock mass[S]. 2014.(in Chinese))