Page 15 - 2024年第55卷第7期
P. 15
[11] 谭震,郭新蕾,李甲振,等.基于多尺度卷积神经网络的管道泄漏检测模型 研 究 [J].水 利 学 报,2023,
54(2):220 - 231.
[12] DONGLJ,TANGZ,LIXB,etal.Discriminationofminingmicroseismiceventsandblastsusingconvolutional
neuralnetworksandoriginalwaveform [J].JournalofCentralSouthUniversity,2020,27(10):3078 - 3089.
[13] JOHNSONSW,CHAMBERSDJA,BOLTZM S,etal.Applicationofaconvolutionalneuralnetworkforseismic
phasepickingofmining - inducedseismicity [J].GeophysicalJournalnternational,2021,224(1):230 - 240.
[14] ZHAOH,CHEN B.Data - drivenmodelforrockburstprediction[J].MathematicalProblemsinEngineering,
2020,2020:5735496.
[15] 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.
[16] 杨小彬,裴艳宇,程虹铭,等.基于 SOFM神经网络模型的岩爆烈度等级预测方法[J].岩石力学与工程
学报,2021,40(S1):2708 - 2715.
[17] 张昱,张明魁.基于 GD - DNN模型的岩爆烈度等级预测方法与应用[J].科学技术与工程,2023,23(27):
11835 - 11840.
[18] CAIW,DOUL,ZHANGM,etal.Afuzzycomprehensiveevaluationmethodologyforrockburstforecastingusing
microseismicmonitoring[J].Tunnelling& UndergroundSpaceTechnology,2018,80(10):232 - 245.
[19] 胡磊,冯夏庭,毕鑫,等.基于微震信息和深度卷积神经网络的隧洞岩爆预警方法:中国,201910560076.
2[P].2019 - 10 - 01.
[20] 马佳骥,马春驰,曾俊,等.基于 LSTM 多变种模型的岩爆微震参数预测研究[J].地下空间与工程学报,
2022,18(5):1481 - 1494.
[21] 张航.基于深度学习的隧道微震信号处理及岩爆智能预警研究[D].成都:成都理工大学,2020.
[22] MAK,SUNX,ZHANGZ,etal.Intelligentlocationofmicroseismiceventsbasedonafullyconvolutionalneural
network (FCNN)[J].RockMechanicsandRockEngineering,2022,55(8):4801 - 4817.
[23] MAK,SHENQ,SUNX,etal.Rockburstpredictionmodelusingmachinelearningbasedonmicroseismicpa
rametersofQinlingwaterconveyancetunnel [J].JournalofCentralSouthUniversity,2023,30(1):289 - 305.
[24] 刘福生.基于微震监测的岩爆风险判别及预测预报[D].西安:西安理工大学,2020.
[25] FRANCOISC,张亮.Python深度学习[M].北京:人民邮电出版社,2018.
[26] 张平.图解深度学习与神经网络:从张量到 TensorFlow实现[M].北京:电子工业出版社,2018.
Areal - timerockburstpredictionmodelforQinlingTunnelbasedonthe
characteristicparametersofmicroseismicmonitoring
2
1,2
1
HUJing,LIUShen,CHENZuyu
(1.ChinaInstituteofWaterResourcesandHydropowerResearch,Beijing 100048,China;
2.InstituteofGeotechnicalEngineering,ZhejiangUniversity,Hangzhou 310058,China)
Abstract:Rockburstisoneofthemaindisastersintheconstructionofdeepearthengineering,andmicroseismic
monitoringisthemainmethodofshort - termpredictionforrockburst.Inordertosolvetheproblemthattheshort -
termpredictionofrockburstmainlydependsonexperience ,adatabasewasestablishedconsistingofthemicroseis
micmonitoringandtherockburstrecordeventsatQinlingTunnelofthewaterdiversionprojectfromHanjiangRiver
toWeiheRiver.Basedonallcharacteristicparametersduringaperiod,e.g.,energy,position,andmagnitudeof
themicroseismicevent ,areal - timepredictionmodelforrockburstwasestablishedusingtheconvolutionalneural
network.Besides,theinfluenceofworksurfacepositiononrockburstwasalsoconsidered.Accordingtothetrain
ing ,validationandtestresults,themodelstructure,selectionoflookbacktime,predictiontimeandtrainingep
ochwereoptimized.Themodelcouldreasonablydescribetheinfluenceofthedistributionfeatureofmicroseismic
andtheconstructionprogressontherockburstprobability.Basedonthetestresults ,themodelcanpredictthe
probabilityofrockburstinthenext48hourscontinuously.Theaccuracyofthepredictionismorethan80%,which
providesaneffectivetechnicalwayforreal - timepredictionofrockburst.
Keywords:microseismicmonitoring;rockburst;prediction;convolutionalneuralnetwork;characteristicparameter
(责任编辑:李 娜)
— 7 6 7 —