Geotechnical Aspects of Underground Construction in Soft Ground - C. NG., Et Al., (CRC, 2009) BBS PDF

GEOTECHNICAL ASPECTSOF UNDERGROUNDCONSTRUCTIONINSOFT GROUND
PROCEEDINGS OF THE 6TH INTERNATIONAL SYMPOSIUM (IS-SHANGHAI 2008),

SHANGHAI, CHINA, 1012 APRIL 2008
Geotechnical Aspectsof
UndergroundConstruction
inSoft Ground
Editors
C.W.W. Ng
Hong Kong University of Science and Technology,
Hong Kong Special Administrative Region
H.W. Huang& G.B. Liu
Tongji University, Shanghai, China
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business
2009Taylor & FrancisGroup, London, UK
Typeset byCharonTecLtd(A MacmillanCompany), Chennai, India
PrintedandboundinGreat BritainbyCromwell PressLtd, Trowbridge, Wiltshire.
All rightsreserved. Nopart of thispublicationor theinformationcontainedhereinmaybereproduced, stored
inaretrieval system, or transmittedinanyformor byanymeans, electronic, mechanical, byphotocopying,
recordingor otherwise, without writtenprior permissionfromthepublishers.
Althoughall careistakentoensureintegrityandthequalityof thispublicationandtheinformationherein, no
responsibilityisassumedbythepublishersnor theauthor for anydamagetothepropertyor personsasaresult
of operationor useof thispublicationand/or theinformationcontainedherein.
Publishedby: CRC Press/Balkema
P.O. Box447, 2300AK Leiden, TheNetherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com www.taylorandfrancis.co.uk www.balkema.nl
ISBN: 978-0-415-48475-6(Hardback)
ISBN: 978-0-203-87998-6(eBook)
Geotechnical Aspects of Underground Construction in Soft Ground Ng, Huang & Liu (eds)
2009 Taylor & Francis Group, London, ISBN 978-0-415-48475-6
Tableof Contents
Preface XIII
Sponsors XV
Special lectures
ProcessesaroundaTBM 3
A. Bezuijen &A.M. Talmon
Supportingexcavationsinclay fromanalysistodecision-making 15
M.D. Bolton, S.Y. Lam &A.S. Osman
Overviewof Shanghai YangtzeRiverTunnel Project 29
R. Huang
Undergroundconstructionindecomposedresidual soils 45
I.M. Lee & G.C. Cho
General reports
Safetyissues, riskanalysis, hazardmanagement andcontrol 67
C.T. Chin & H.C. Chao
Calculationanddesignmethods, andpredictivetools 77
F. Emeriault & R. Kastner
Analysisandnumerical modelingof deepexcavations 87
R.J. Finno
Constructionmethod, groundtreatment, andconditioningfor tunneling 99
T. Hashimoto, B. Ye & G.L. Ye
Physical andnumerical modelling 109
P.L.R. Pang
Casehistories 121
A. Sfriso
Theme 1: Analysis and numerical modeling of deep excavations
Optimizationdesignof compositesoil-nailinginloessexcavation 133
G.M. Chang
Three-dimensional finiteelement analysisof diaphragmwallsfor top-downconstruction 141
J. Hsi, H. Zhang &T. Kokubun
Numerical evaluationof dewateringeffect ondeepexcavationinsoft clay 147
L. Li & M. Yang
Analysisof thefactorsinfluencingfoundationpit deformations 153
Y.Q. Li, K.H. Xie, J. Zhou & X.L. Kong
V
Constructionmonitoringandnumerical simulationof anexcavationwithSMW
retainingstructure 159
Z.H. Li & H.W. Huang
A simplifiedspatial methodologyof earthpressureagainst retainingpilesof pile-row
retainingstructure 165
Y.L. Lin & X.X. Li
Considerationof designmethodfor bracedexcavationbasedonmonitoringresults 173
H. Ota, H. Ito, T. Yanagawa, A. Hashimoto, T. Hashimoto &T. Konda
Groundmovementsinstationexcavationsof Bangkokfirst MRT 181
N. Phienwej
Numerical modellingandexperimental measurementsfor aretainingwall of
adeepexcavationinBucharest, Romania 187
H. Popa, A. Marcu & L. Batali
3Dfiniteelement analysisof adeepexcavationandcomparisonwithinsitumeasurements 193
H.F. Schweiger, F. Scharinger & R. Lftenegger
Theeffect of deepexcavationonsurroundinggroundandnearbystructures 201
A. Siemi nska-Lewandowska & M. Mitew-Czajewska
Multi-criteriaprocedurefor theback-analysisof multi-supportedretainingwalls 207
J. Zghondi, F. Emeriault & R. Kastner
Monitoringandmodellingof riversidelargedeepexcavation-inducedgroundmovementsinclays 215
D.M. Zhang, H.W. Huang &W.Y. Bao
GPSheight applicationandgrosserror detectioninfoundationpit monitoring 223
H. Zhang, S.F. Xu &T.D. Lu
Studyondeformationlawsunder theconstructionof semi-reversemethod 227
J. Zhang, G.B. Liu &T. Liu
Comparisonof theoryandtest onexcavationcausingthevariationof soilmassstrength 235
J. Zhou, J.Q. Wang & L. Cong
Theme 2: Construction method, ground treatment, and conditioning for tunnelling
Tenyearsof boredtunnelsinTheNetherlands: Part I, geotechnical issues 243
K.J. Bakker &A. Bezuijen
Tenyearsof boredtunnelsinTheNetherlands: Part II, structural issues 249
K.J. Bakker &A. Bezuijen
Theinfluenceof flowaroundaTBM machine 255
A. Bezuijen & K.J. Bakker
Mechanismsthat determinebetweenfractureandcompactiongroutinginsand 261
A. Bezuijen, A.F. van Tol & M.P.M. Sanders
Researchof non-motor vehicle-rail transit-tubeinterchangingtransport systempattern 269
A.Z.G. Deng & Q.H. Zhang
Shotcreteexcavationsfor theMunichsubway Comparisonof different methods
of facesupport insettlement sensitiveareas 275
J. Fillibeck & N. Vogt
Fracturingof sandincompensationgrouting 281
K. Gafar, K. Soga, A. Bezuijen, M.P.M. Sanders &A.F. van Tol
VI
Historical casesanduseof horizontal jet groutingsolutionswith360
distributionand
frontal septumtoconsolidateveryweakandsaturatedsoils 287
G. Guatteri, A. Koshima, R. Lopes, A. Ravaglia & M.R. Pieroni
Theeffectsof sampledimensionandgradationonshear strengthparametersof
conditionedsoilsinEPBM 295
M. Hajialilue-Bonab, M. Ahmadi-adli, H. Sabetamal & H. Katebi
Experimental studyoncompressibilitybehavior of foamedsandysoil 301
M. Hajialilue-Bonab, H. Sabetamal, H. Katebi & M. Ahmadi-adli
Studyonearthpressureactinguponshieldtunnel lininginclayeyandsandygroundsbased
onfieldmonitoring 307
T. Hashimoto, G.L. Ye, J. Nagaya, T. Konda & X.F. Ma
Thedouble-o-tubeshieldtunnel inShanghai soil 313
C. He, L. Teng & J.Y. Yan
Frozensoil propertiesfor crosspassageconstructioninShanghai YangtzeRiverTunnel 319
X.D. Hu &A.R. Pi
Theinfluenceof engineering-geological conditionsonconstructionof the
radioactivewastedump 325
J. Kuzma & L. Hrustinec
Critical ventilationvelocityinlargecross-sectionroadtunnel fire 331
Z.X. Li, X. Han & K.S. Wang
MetrotunnelsinBuenosAires: Designandconstructionprocedures19982007 335
A.O. Sfriso
Studyontheearthpressuredistributionof excavationchamber inEPB tunneling 343
T.T. Song & S.H. Zhou
Backfill groutingresearchat GroeneHartTunnel 349
A.M. Talmon &A. Bezuijen
Longitudinal tubebendingduetogrout pressures 357
A.M. Talmon, A. Bezuijen & F.J.M. Hoefsloot
Theme 3: Case histories
Tunnel facestabilityandsettlement control usingearthpressurebalanceshield
incohesionlesssoil 365
A. Antiga & M. Chiorboli
Displacementsandstressesinducedbyatunnel excavation: Caseof BoisdePeu(France) 373
S. Eclaircy-Caudron, D. Dias & R. Kastner
Shieldtunnelingbeneathexistingrailwaylineinsoft ground 381
Q.M. Gong & S.H. Zhou
Casehistoryonarailwaytunnel insoft rock(Morocco) 385
A. Guiloux, H. Le Bissonnais, J. Marlinge, H. Thiebault, J. Ryckaert, G. Viel,
F. Lanquette, A. Erridaoui & M.Q.S. Hu
Observedbehavioursof deepexcavationsinsand 393
B.C.B. Hsiung & H.Y. Chuay
Environmental problemsof groundwater aroundthelongest expresswaytunnel inKorea 399
S.M. Kim, H.Y. Yang & S.G. Yoon
VII
Measurementsof grounddeformationsbehindbracedexcavations 405
T. Konda, H. Ota, T. Yanagawa &A. Hashimoto
Researchontheeffect of buriedchannelstothedifferential settlement of building 413
D.P. Liu, R. Wang & G.B. Liu
Performanceof adeepexcavationinsoft clay 419
G.B. Liu, J. Jiang & C.W.W. Ng
Deformationmonitoringduringconstructionof subwaytunnelsinsoft ground 427
S.T. Liu & Z.W. Wang
Theconstructionandfieldmonitoringof adeepexcavationinsoft soils 433
T. Liu, G.B. Liu & C.W.W. Ng
Excavationentirelyonsubwaytunnelsinthecentral areaof thePeoplesSquare 441
Y.B. Mei, X.H. Jiang, Y.M. Zhu & H.C. Qiao
Thebenefitsof hybridgroundtreatment insignificantlyreducingwall movement:
A Singaporecasehistory 447
N.H. Osborne, C.C. Ng & C.K. Cheah
3Ddeformationmonitoringof subwaytunnel 455
D.W. Qiu, K.Q. Zhou, Y.H. Ding, Q.H. Liang & S.L. Yang
Challengingurbantunnellingprojectsinsoft soil conditions 459
H. Quick, J. Michael, S. Meissner & U. Arslan
Supervisionandprotectionof Shanghai MassRapidLine4shieldtunnelingacross
theadjacent operatingmetroline 465
R.L. Wang, Y.M. Cai & J.H. Liu
KowloonSouthernLink TBM crossingover MTRTsuenWanLinetunnelsinHKSAR 471
K.K.W. Wong, N.W.H. Ng, L.P.P. Leung &Y. Chan
Applicationof pileunderpinningtechnologyonshieldmachinecrossingthroughpile
foundationsof roadbridge 477
Q.W. Xu, X.F. Ma & Z.Z. Ma
Characteristicsof tunneling-inducedgroundsettlement ingroundwater drawdownenvironment 485
C. Yoo, S.B. Kim &Y.J. Lee
Effect of long-termsettlement onlongitudinal mechanical performanceof tunnel insoft soil 491
H.L. Zhao, X. Liu, Y. Yuan &Y. Chi
Theme 4: Safety issues, risk analysis harzard management and control
Researchonstochasticseismicanalysisof undergroundpipelinebasedon
physical earthquakemodel 499
X.Q. Ai & J. Li
Riskassessment for thesafegradeof deepexcavation 507
X.H. Bao & H.W. Huang
Multi-factorsdurabilityevaluationinsubwayconcretestructure 513
C. Chen, L. Yang & C. Han
Theuseof artificial neural networkstopredict groundmovementscausedbytunneling 519
I. Chissolucombe, A.P. Assis & M.M. Farias
Researchandapplicationof roadtunnel structural optimization 525
W.Q. Ding &Y. Xu
VIII
Floor heavebehavior andcontrol of roadwayintersectionindeepmine 531
B.H. Guo &T.K. Lu
Squeezingpotential of tunnelsinclaysandclayshalesfromnormalizedundrained
shear strength, unconfinedcompressivestrengthandseismicvelocity 537
M. Gutierrez & C.C. Xia
Frameworkof performance-basedfireprotectiondesignmethodfor roadtunnel 545
X. Han & G.Y. Ding
Predictionof surfacesettlementsinducedbyshieldtunneling: AnANFISmodel 551
J. Hou, M.X. Zhang & M. Tu
Experimental studiesof ageological measuringsystemfor tunnel withultrasonictransducer 555
D.H. Kim, U.Y. Kim, S.P. Lee, H.Y. Lee & J.S. Lee
Performancereviewof apipejackingproject inHongKong 561
T.S.K. Lam
Geotechnical control of amajor railwayproject involvingtunnel worksinHongKong 567
W. Lee, S.S. Chung, K.J. Roberts & P.L.R. Pang
Researchonstructural statusof operatingtunnel of metroinShanghai andtreatment ideas 573
J.P. Li, R.L. Wang & J.Y. Yan
Maximisingthepotential of straingauges: A Singaporeperspective 579
N.H. Osborne, C.C. Ng, D.C. Chen, G.H. Tan, J. Rudi & K.M. Latt
Discussionondesignmethodfor retainingstructuresof metrostationdeepexcavationsinShanghai 587
R. Wang, G.B. Liu, D.P. Liu & Z.Z. Ma
Riskanalysisfor cutterheadfailureof compositeEPB shieldbasedonfuzzyfault tree 595
Y.R. Yan, H.W. Huang & Q.F. Hu
Riskassessment onenvironmental impact inXizangRoadTunnel 601
C.P. Yao, H.W. Huang & Q.F. Hu
Riskanalysisandfuzzycomprehensiveassessment onconstructionof shield
tunnel inShanghai metroLine 607
H.B. Zhou, H. Yao &W.J. Gao
Theme 5: Physical and numerical modelling
Tunnel behaviour under seismicloads: Analysisbymeansof uncoupledandcoupledapproaches 615
D. Boldini &A. Amorosi
Investigatingtheinfluenceof tunnel volumelossonpilesusingphotoelastictechniques 621
W. Broere & J. Dijkstra
Assessment of tunnel stabilityinlayeredground 627
P. Caporaletti, A. Burghignoli, G. Scarpelli & R.N. Taylor
Reinforcingeffectsof forepolingandfaceboltsintunnelling 635
K. Date, R.J. Mair & K. Soga
Mechanical behavior of closelyspacedtunnelslaboratorymodel testsandFEM analyses 643
J.H. Du & H.W. Huang
Stabilityanalysisof masonryof anoldtunnel bynumerical modellingandexperimental design 649
J. Idris, T. Verdel & M. Alhieb
IX
Excavationwithstepped-twinretainingwall: Model testsandnumerical simulations 655
N. Iwata, H.M. Shahin, F. Zhang, T. Nakai, M. Niinomi &Y.D.S. Geraldni
Stabilityof anunderwater trenchinmarineclayunder oceanwaveimpact 663
T. Kasper & P.G. Jackson
A studyonbehavior of 2-Archtunnel byalargemodel experiment 669
S.D. Lee, K.H. Jeong, J.W. Yang & J.H. Choi
Behavior of tunnel duetoadjacent groundexcavationunder theinfluenceof
pre-loadingonbracedwall 677
S.D. Lee & I. Kim
Twodistinctiveshear strainmodesfor pile-soil-tunnellinginteractioninagranular mass 683
Y.J. Lee & C.S. Yoo
Stabilityanalysisof largeslurryshield-driventunnel insoft clay 689
Y. Li, Z.X. Zhang, F. Emeriault & R. Kastner
Effectsof soil stratificationonthetunneling-inducedgroundmovements 697
F.Y. Liang, G.S. Yao & J.P. Li
Centrifugemodellingtoinvestigatesoil-structureinteractionmechanismsresultingfrom
tunnel constructionbeneathburiedpipelines 703
A.M. Marshall & R.J. Mair
Groundmovement andearthpressureduetocircular tunneling: Model testsand
numerical simulations 709
H.M. Shahin, T. Nakai, F. Zhang, M. Kikumoto, Y. Tabata & E. Nakahara
Analysisof pre-reinforcedzoneintunnel consideringthetime-dependent performance 717
K.I. Song, J. Kim & G.C. Cho
Vault temperatureof vehiclefiresinlargecross-sectionroadtunnel 725
K.S. Wang, X. Han & Z.X. Li
Effectsof different benchlengthonthedeformationof surroundingrockbyFEM 729
X.M. Wang, H.W. Huang & X.Y. Xie
Theeffectsof loadedboredpilesonexistingtunnels 735
J. Yao, R.N. Taylor &A.M. McNamara
3DFEM analysisongrounddisplacement inducedbycurvedpipe-jackingconstruction 743
G.M. You
Theme 6: Calculation and design methods, and predictive tools
Calculationof thethreedimensional seismicstressedstateof MetroStationEscalatorOpen
LineTunnels system, whichislocatedininclinedstratifiedsoft ground 751
R.B. Baimakhan, N.T. Danaev, A.R. Baimakhan, G.I. Salgaraeva, G.P. Rysbaeva,
Zh.K. Kulmaganbetova, S. Avdarsolkyzy, A.A. Makhanova & S. Dashdorj
A complexvariablesolutionfor tunneling-inducedgroundmovementsinclays 757
H.L. Bao, D.M. Zhang & H.W. Huang
Simulationof articulatedshieldbehavior at sharpcurvebykinematicshieldmodel 761
J. Chen, A. Matsumoto & M. Sugimoto
Deformationandporepressuremodel of thesaturatedsiltyclayaroundasubwaytunnel 769
Z.D. Cui, Y.Q. Tang & X. Zhang
Analytical solutionof longitudinal behaviour of tunnel lining 775
F.J.M. Hoefsloot
X
Designof tunnel supportingsystemusinggeostatistical methods 781
S. Jeon, C. Hong & K. You
Comparativestudyof softwaretoolsontheeffectsof surfaceloadsontunnels 785
D.K. Koungelis & C.E. Augarde
GeologicModel TransformingMethod(GMTM) for numerical analysismodelingin
geotechnical engineering 791
X.X. Li, H.H. Zhu &Y.L. Lin
Reviewandinterpretationof intersectionstabilityindeepundergroundbased
onnumerical analysis 799
T.K. Lu, B.H. Guo, L.C. Cheng & J. Wang
Analysisof surfacesettlement duetotheconstructionof ashieldtunnel insoft
clayinShanghai 805
Z.P. Lu & G.B. Liu
Urbantunnelsinsoil: Reviewof current designpracticeinBrazil 811
A. Negro
A studyonloadsfromcomplexsupport systemusingsimple2Dmodels 817
Z. Shi, W. Bao, J. Li, W. Guo & J. Zhu
Groundreactionduetotunnellingbelowgroundwater table 823
Y.J. Shin, J.H. Shin & I.M. Lee
Basal stabilityof bracedexcavationsinK
0
-consolidatedsoft claybyupper boundmethod 829
X.Y. Song & M.S. Huang
Analytical twoandthreedimensionmodelstoassessstabilityanddeformationmagnitudeof
undergroundexcavationsinsoil 837
L.E. Sozio
Dynamicresponseof saturatedsiltyclayaroundatunnel under subwayvibrationloading
inShanghai 843
Y.Q. Tang, Z.D. Cui & X. Zhang
Lateral responsesof pilesduetoexcavation-inducedsoil movements 849
C.R. Zhang, M.S. Huang & F.Y. Liang
Elastic-plasticanalysisfor surroundingrockof pressuretunnel withliningbasedonmaterial
nonlinear softening 857
L.M. Zhang & Z.Q. Wang
Modificationof keyparametersof longitudinal equivalent model for shieldtunnel 863
W. Zhu, X.Q. Kou, X.C. Zhong & Z.G. Huang
Author Index 869
XI
Preface
Under theChairmanshipof Professor K. Fujita, thefirst symposiumpurposely addressinggeotechnical issues
relatedtoundergroundconstructioninsoftgroundwasheldin1994, priortothe13thInternational Conferenceon
Soil MechanicsandGeotechnical EngineeringheldinNewDelhi. Followingthesuccessof thefirstsymposium,
Professor R. Mair succeeded the Chairmanship of TC28 and he initiated a series of three-day International
SymposiaonGeotechnical Aspectsof UndergroundConstructioninSoft Groundincludingtechnical sitevisits
toundergroundconstructionprojects. Intotal, four three-dayInternational Symposiahavebeenheldverythree
yearssince1996.TheseincludetheonesheldinLondon, UK (1996), inTokyo, J apan(1999), inToulouse, France
(2002) andinAmsterdam, theNetherlands(2005).
This volume includes a collection of four invited special lectures delivered by Dr A. Bezuijen
(TheNetherlands), Mr HuangRong(China), Professor M.D. Bolton(UK) andProfessor I.M. Lee(Korea). The
titlesof theirlecturesareProcessesaroundaTBM, Overviewof Shanghai Yangtzerivertunnel project, Sup-
portingexcavationsinclay fromanalysistodecision-makingandUndergroundconstructionindecomposed
residual soils, respectively.
Inaddition, thisvolumecontains112papersgroupedunder sixthemesincluding(i) Analysisandnumerical
modelling of deep excavations; (ii) Construction method, ground treatment, and conditioning for tunnelling;
(iii) Casehistories; (iv) Safetyissues, riskanalysis, hazardmanagementandcontrol; (v) Physical andnumerical
modelling and (vi) Calculation and design methods, and predictivetools. Six general reports discussing and
commentingpapersgroupedunder thesixthemeswerecontributedorallyduringtheSymposiumbyProfessor
RichardFinno, Professor Tadashi Hashimoto, Mr Alejo Sfriso, Dr C.T. Chin, Dr RichardPangandProfessor
RichardKastner, respectively. Thewrittenversionsof their sixgeneral reportsarealsoincludedinthisvolume.
Y.S. Li
Chairman of the Symposium
C.W.W. Ng, H.W. HuangandG.B. Liu
Vice-Chairmen of the Symposium and Editors
XIII
Sponsors
Organized by:
Tongji University
Under the auspices of:
ISSMGE
Technical Committee28of theInternational
Societyof Soil MechanicsandGeotechnical
Engineering(ISSMGE)
Supported by
ChinaCivil EngineeringSociety
ChineseSocietyfor RockMechanics
andEngineering
Geotechnical Division, theHongKong
Institutionof Engineers
HongKongGeotechnical Society
HongKongUniversityof Science
andTechnology
XV
ScienceandTechnologyCommissionof
Shanghai Municipality
Shanghai ChangjiangTunnel & Bridge
Development Co., Ltd.
Shanghai Societyof Civil Engineering
XVI
Special lectures
ProcessesaroundaTBM
A. Bezuijen&A.M. Talmon
Deltares and Delft University of Technology, Delft, The Netherlands
ABSTRACT: Processesthatoccur aroundaTBMduringtunnellinghavebeeninvestigatedwhiletunnellingin
saturatedsand. Theporepressureinfront of theTBM increasesduetoalackof plasteringduringdrilling. This
hasconsequencesfor thestabilityof thetunnel face, or thesoil infront of thetunnel. A bentoniteflowislikely
alongsidetheTBMfromthetunnel face, and/or groutflowfromtheback. Itseemsthatvirtuallynoinvestigation
hasbeenmadeof thispart of theTBM, but it isimportant tounderstandthevolumelossthat occursarounda
tunnel. TheliningisconstructedbehindtheTBM andthetail voidgrout isapplied. Pressuresmeasuredinthe
tail voidgrout will bediscussed, aswell astheconsequencesfor loadingonthesoil andthelining. Most of the
resultsdescribedarebasedonfieldmeasurementsperformedat varioustunnelsconstructedintheNetherlands.
1 INTRODUCTION
Dutch experience of using TBM tunnelling is rela-
tively recent. Thefirst TBM tunnel was constructed
intheNetherlands between1997and1999(theSec-
ond Heinenoord Tunnel). In the early 1990s, Dutch
engineers were uncertain whether the soft saturated
soil in the western parts of their country was suit-
ablefor TBM tunnelling. Thedecisionwas therefore
taken to include a measurement programme in the
first tunnelling projects. An overview of this pro-
grammeandsomeresultsarepresentedby Bakker &
Bezuijen(2008). Intheprogramme, results fromthe
measurements werepredictedusingexistingcalcula-
tionmodels. Themeasurement results wereanalysed
at alater date, anddiscrepancieswiththepredictions
wereexplainedwherepossible.
An important part of the measurement and anal-
ysis programme was dictated by the processes that
occur around theTBM. This paper deals with some
of these processes. It does not cover all aspects of
TBM tunnellingasthiswouldnot fit withinthelimits
of this paper (seeBezuijen& vanLottum, 2006, for
moreinformation).Thepaper focusesoncertainareas
whereideasconcerningthemechanismsinvolvedhave
changed over the last decade, and where a better
understandingisnowapparent.
In order to structure this paper, we walk along
theTBM. Westart with aprocess at thefront of the
TBM:thecreationandstabilityof thetunnel faceunder
theinfluenceof excessporepressures. Thepaper then
discusses what happens next to the TBM. The last
part of the paper deals with the tail void grout that
isinjectedattheendof theTBM. Thepaper describes
thecurrent stateof theart of theseprocesses, anddis-
cusses howknowledgegained about theseprocesses
mayinfluencethedesignof aTBMtunnel insoftsoil.
2 PORE PRESSURESINFRONT OF A TBM
2.1 Flow in coarse and fine granular material
DuringTBM tunnelling, it isessential that thetunnel
faceisstabilisedby pressurisedslurry (slurry shield)
or muck (EPB shield). Thepressuremust beadapted
to theground pressureto stabilisethefront. If pres-
sureistoolow, thiswill leadtoaninstabletunnel front
resultingincollapseof thetunnel face. If pressureis
too high, a blow-out will occur. Various calculation
methods have been proposed to calculate the stabil-
ity of thetunnel face. Most of thesemethods do not
taketheinfluenceof porewater flowintoaccount. It
is assumed that the bentonite slurry or muck at the
tunnel facecreates aperfect seal that prevents water
flowfromthefaceintothesoil. Experiencewithtun-
nels built in areas wherethesubsoil contains gravel
hasshownthat thebentoniteslurrycanpenetrateinto
thesubsoil overmorethan7m(Steiner, 1996). Steiner
advises that thesandandfines shouldberetainedin
theslurry(insteadof removingthemintheseparation
plant),andthatsawdustshouldbeusedinthebentonite
(Steiner, 2007). Anagnostou&Kovari (1994) propose
acalculationmethodforsuchasituation.However,this
methodonlytakestheviscousbehaviour of theslurry
into account, and not the stiffening that occurs dur-
ing standstill. Theresults of this calculation method
may therefore lead to the prescription of bentonite
3
Figure1. Measuredexcessporepressureinfrontof aslurry
shieldandapproximation.
with viscosity that is too high (Steiner, 2007). The
stateof theart for such asituation involving coarse
granular material is still trial and error, but thetrial
canbeperformedinthelaboratory toavoiderrors in
thefield.
Usual tunnellingconditionsintheNetherlandsare
asaturated sandy soil in medium-finesand. In such
soil conditions, thegroundwater flow influences the
plastering. There will be virtually no plastering of
thetunnel faceby thebentoniteor themuck during
drilling, becausethegroundwater infrontof theTBM
prevents water inthebentoniteslurry or muck flow-
ing into the soil. Plastering will only occur during
standstill of theTBM process.
Figure1showsmeasuredporepressureinfront of
aslurry shieldas afunctionof thedistancefromthe
TBM front. Plasteringoccursduringstandstill, result-
inginapressureof 120kPa(thehydrostaticpressure).
Higher porepressuresweremeasuredduringdrilling,
becausetheTBMscutter headremovesacakebefore
it canformat thetunnel face.
Figure2showsthesamephenomenonmeasuredin
frontof anEPB shield. Here, onlythepressureduring
drillingwasrecorded.
Bezuijen (2002) shows that theamount of excess
pore pressure measured in the soil in front of the
TBM (apart frompressure at the tunnel face) also
depends onsoil permeability, thequality of theben-
tonite or muck, and the drilling speed. Where EPB
drilling takes place in sand with a low permeability
(k=10
5
m/s), theporepressuremeasuredinsandin
front of theTBM is virtually equal topressureinthe
mixing chamber. Thepressureis lower in sand with
higher permeability (k=3 10
4
m/s), becausesome
plasteringof thefaceoccursduringdrilling. Soil per-
meabilityalsoinfluencesthefoamproperties. Muckin
themixingchamberwill bedryerinsandwithahigher
permeability. Where the permeability of the sand is
lower, thewater content inthemuckisnearlyentirely
Figure 2. Measured excess pore pressure in front of an
EPBshield() andapproximation(BotlekRail Tunnel, MQ1
South). Relativelyimpermeablesubsoil.
determinedby water inthesoil andmuchlessby the
foamproperties(alsoseeBezuijen, 2002).
Figure1andFigure2alsoshowatheoretical curve
(Bezuijen, 2002):
Where
0
is thepiezometric headat thetunnel face,
the piezometric head at a distance x in front of
thetunnel face, and R theradius of thetunnel. This
relationship is valid for situations wheretheperme-
ability of soil around the tunnel is constant. In the
Netherlands, thesandy layers usedfor tunnellingare
sometimes overlain with soft soil layers of peat and
clay withalowpermeability. Insuchasituation, the
pressure distribution in the soil can be evaluated as
asemi-confinedaquifer. This is describedby Broere
(2001).
2.2 Influence on stability
Bezuijenet al (2001) andBroere(2001) haveshown
thatthegroundwater flowinfrontof theTBMimplies
that a larger face pressure is necessary to achieve a
stablefront.AccordingtoBezuijenetal (2001),thedif-
ferenceis approximately 20kPafor a10-m-diameter
tunnel constructed in sand, wherethetop is situated
15mbelowthegroundsurface.
Knowledge of this groundwater flow appeared
essential during the Groene Hart Tunnel (GHT)
project, not toprevent collapseof thetunnel facebut
toprevent aformof blow-out (Bezuijenet al, 2001).
Thistunnel entersadeeppolderwherethepiezometric
headinthesandlayersunderneaththesoftsoil layersis
higherthanthesurfacelevel (seeFigure3).Asaresult,
theeffectivestresses beneath thesoft soil layers are
extremely small. Thecalculatedexcessporepressure
4
Figure3. Geotechnical profileGHT tunnel inpolder. Tun-
nel isdrilledfromright toleft inthispicture.
in the sand layer induced by the tunnelling process
couldcausefloating of thesoft layers. Thecontrac-
tor madedetailednumerical calculations(Aimeet al,
2004). As aresult of thesecalculations, atemporary
sanddamwasconstructedatthepointwherethetunnel
enteredthepolder. This damdeliveredthenecessary
weight to prevent lifting of the soft soil layers due
to excess pore pressure generated at the tunnel face
duringdrilling.
3 FLOWAROUNDTHETBM
3.1 Calculation model
Until recently, only limited attention has been given
to pressure distribution and flow around the TBM
shield. It was assumed that the soil was in contact
withtheTBMshieldacrosstheshield. Duringdrilling
of theWestern Scheldt tunnel, however, it appeared
thattheTBMdeformedatlargedepthsandhighwater
pressures(thetunnel isconstructedupto60mbelow
the water line). This could not be explained by the
concept of a TBM shield in contact with the soil.
Furthermore, tunnelling technology has advanced to
a level where the ground loss due to tunnelling is
lessthanthevolumedifferencecausedbytaperingof
theTBM. TBMs areusually tapered, with a slightly
larger diameter at the head compared with the tail.
ThisallowstheTBM tomanoeuvreandtodrill witha
certaincurvature.Table1showsthevolumedifference
duetotaperingfor differentTBMs.
Thevolumelossesmeasuredduringtheseprojects
varied, but negative volume losses were sometimes
measuredinall theprojects(therewasactuallyheave).
It is clear that themeasuredvolumeloss canbeless
than the volume loss due to tapering. This leads to
Table 1. Percentage of tapering of the TBM in 3 tunnel
projectsinTheNetherlands.
Tunnel project Tapering%
SecondHeinenoord 0.95
Botlek 0.77
Sophia 0.79
theidea(Bezuijen, 2007) that thesoil is not in con-
tact with theTBM all over theTBM. Overcutting at
the tunnel face can lead to bentonite flow over the
TBMshieldfromthefacetowardsthetail. Groutpres-
sureduringgrout injectionisusuallyhigher at thetail
than the soil pressure. The soil is therefore pushed
away fromthe TBM, and grout will flow fromthe
tail over theshield. It is possibleto describeflowon
theshield, if it isassumedthat boththebentoniteand
thegrout areBinghamliquids, that theyieldstressis
dominant intheflowbehaviour, andthat thereislin-
ear elastic soil behaviour. A moreor less conceptual
model is developed, assumingacylindrical symmet-
rical situationaroundthetunnel axis. Changes inthe
soil radius for such a situation can be described as
(Verruijt, 1993):
WhereL is thechangein pressure, Lr thechange
inradius, r theradiusof thetunnel andthegrout, and
G theshear modulusof thesoil aroundthetunnel.
The flow around the TBM shield can be
describedas:
WhereLP is thechangeinpressuredueto theflow,
Lx alengthincrementalongtheTBM, s thegapwidth
betweenthetunnel andthesoil, and
theyieldstress
of thegrout aroundtheTBM. is acoefficient indi-
cating whether there is friction between the soil or
bentoniteandthegrout (=1) only, or alsobetween
theTBM andthegrout or bentonite(=2). Viscous
forcesareneglectedinthisformula.Thisispermissible
duetothelowflowvelocitiesthat canbeexpected.
Withno grout or bentoniteflowaroundtheTBM,
taperingwill leadtoaneffectivestressreductionpro-
ceedingfromtheTBMs faceto thetail accordingto
equation(2).Thegroutandbentoniteflowwill change
thispressuredistribution.Inordertocalculatethepres-
suredistributionunder flow, theflowdirectionof both
the bentonite and the grout must be known. These
flow directions can vary during the tunnelling pro-
cess(Bezuijen, 2007). Onaverage, however, theTBM
5
Table2. Inputparametersusedincalculationwithbentonite
andovercutting.
LengthTBM shield 5 m
Diameter 10 m
Diameter reduction 0.2 %
Overcutting 0.015 m
Asymmetric(1) or symmetric(2) 2
Grainstress 150 kPa
Grout pressure 400 kPa
Porepressure 200 kPa
Pressureontunnel face 250 kPa
Shear modulus(G) 90 MPa
Yieldstressgrout 1.6 kPa
Yieldstressbentonite 0.01 kPa
Figure4. PressuresandgapwidthalongaTBM.Groutpres-
suresandbentonitepressures. ParametersseeTable2. Plots
show pressures and gap width for the bentonite and grout
pressureseparatelyandthecombinedresult.
advances andthereforethebentoniteandgrout front
must also advanceinthesamedirectionto achievea
stablesituation. This means that grout andbentonite
onlymovewithrespecttothesoil, andnotwithrespect
totheTBM.Therefore=1forboththebentoniteand
thegrout. Theresult of anexamplecalculationusing
theparametersgiveninTable2isshowninFigure4.
The Figure shows that the gap width for a com-
pletelystiff soil masswouldincreasefrom0.015mat
thefrontto0.025matthetail of theTBM. If therewere
onlygrout pressures, thegapwidthwouldbe0.028m
at thetail of theTBM, dueto thegrout pressurethat
islarger thanthetotal stress. However, thegapwould
closeat3.4mfromthetail. If theinfluenceof theben-
toniteisincluded, thereisstill agapwidthof 0.01mat
thetunnel face(5mfromthetail).Thelinethroughthe
trianglespresentsthegapwidthduetothecombined
effects of both thebentoniteand thegrout. Theplot
abovepresentsthepressuresinthesameway.
3.2 Consequences and status
The model shows that the volume loss is not deter-
mined by tapering of the TBM (as suggested for
exampleby Kasper & Meschke, 2006), but is influ-
encedbythepressuredistributionof thebentoniteand
grout. Withsufficient grout pressure, it ispossibleto
haveanegative volumeloss (thesurfacelevel rises
aftertheTBMpasses). Italsoexplainsthatbentoniteis
sometimesfoundinthetail void, andgroutisfoundin
thepressurechamber. Thefirst situationoccurswhen
bentonitepressureisrelativelyhighandgroutpressure
is low(wewill seethat it is quitedifficult to control
grout pressure, especially during ring building). The
second situation occurs when grout pressures in the
tail voidarerelatively high(whichmay occur during
drilling).
Contrary, however, to themodel describedfor the
pore pressures in front of the TBM and the grout
pressure, tobedescribedinthefollowingsections, the
experimental evidenceforthismodel isstill limited.To
our knowledge, pressuredistributionaroundtheTBM
shieldhasneverbeenmeasured.Theshieldwasperfo-
ratedduringconstructionof theWesternScheldttunnel
butnogroutwasfoundbetweentheshieldandthesoil
(Thewes, 2007). Thefactthatnogroutwasfounddur-
ingthisinvestigationmaybecausedbythefactthat, in
reality, theTBMwill notbeplacedassymmetricallyin
thedrilledholeassuggestedinthissimplemodel.The
TBM must bein contact with thesoil at somepoint
tomaintainmechanical equilibrium. Therewill beno
grout aroundtheshieldat that location.
Guglielmetti (2007) rightfully argues that more
research is needed in this field, because: The topic
(flow of bentonite and grout around the TBM) is
definitely one of the most important in the field of
mechanisedtunnelling, beingthemanagement of the
voidaroundtheshieldof aTBM asoneof themajor
sourcesof concernfor bothdesignersandcontractors
involvedinurbantunnellingprojects.
Thereis someevidencefromtheresults of exten-
someter measurementscarriedout at theSophiaRail
Tunnel. The results of the extensometers (shown in
Figure5) arepresentedinFigure6duringpassageof
6
Figure5. SophiaRail Tunnel,soil stratificationandlocation
of extensometersat themeasurement location(pictureArne
Bezuijen).
Figure 6. Extensometer results. The vertical line shows
when the tail of theTBM passes. Soil above theTBM is
alreadycompressedbeforethetail passes.
theTBM. Theresultsshowthat thereisinitiallysome
extensionof thesoil infrontof theTBMduetotherela-
tivelylowstressesatthetunnel face. However, thesoil
abovethetunnel (seetheextensometer at 12.5m) is
compressedseveral rings beforethetail of theTBM
passes (thevertical line) indicating heave, and there
is thereforeno settlement dueto thetapering. When
theTBMhaspassed, theextensometer at12.5mfol-
lowsthecourseof thegroutpressuresmeasuredaround
thelining. Thiswill bediscussedinmoredetail inthe
nextsection, andshowsthatachangeingroutpressure
indeedleadstoachangeinsoil deformation.
We are currently working on the possibility of
measuringpressuresaroundtheshield.
4 TAIL VOIDGROUTING
4.1 Introduction tail void grouting
Coming at the end of theTBM, the tail void grout-
ingprocess is important. Theprocess determines the
loadingonthesoil andonthelining.
Thepressuredistributioncausedbytail voidgrout-
inghasbeenstudiedduringconstructionof theSophia
Rail Tunnel (Bezuijen et al, 2004) and the Groene
Hart Tunnel. Here, wewill describethefundamental
mechanismsusingmeasurementsfromtheSophiaRail
Figure 7. First tube Sophia Rail Tunnel: drilling velocity
andmeasuredgrout pressuresat theright sideof thetunnel
asafunctionof time.
Tunnel, astheyhaveprovidedthemost completedata
set until now.
Thestudyof groutpressureswasinitiatedbyearlier
measurements performed at the Second Heinenoord
Tunnel and the Botlek Rail Tunnel. These measure-
ments didnot matchthegenerally acceptedassump-
tionatthattime atleastintheNetherlands thatthe
vertical pressuregradient inliquidgrout must bedic-
tatedbythedensityof thegrout, andthat thepressure
distribution after hardening must reflect the K
0
(the
ratiobetweenthehorizontal andvertical soil pressure).
Inreality, thevertical pressuregradientwaslower and
theinfluenceof K
0
couldnot bedetected.
4.2 Measurements
TheSophiaRail Tunnel wasconstructedinsandysub-
soil overlainwithsoft soil layers (seeFigure5). The
water tableis closeto thesurface. During construc-
tionof theSophiaRail tunnel, tworingsinthelining
wereeachequippedwith14pressuresensors.Thepres-
sures measuredwithoneof theseinstrumentedrings
areshowninFigure7.
These measurements are discussed in detail in
Bezuijenet al (2004): wewill onlydescribethemain
phenomena here. The upper plot in Figure 7 shows
thedrillingvelocity, whendrillingoccurs, andwhen
therewasastandstill for ringbuilding. It canbeseen
thatanincreaseinpressureismeasuredassoonasthe
7
pressuregauges(builtintotheliningelements) moved
fromthegreaseinto thegrout. Pressureincreases as
longasdrillingcontinues, anddecreaseswhendrilling
stopsduringringbuilding.
4.3 Grout pressures
The mechanism that leads to these pressure varia-
tionsisexplainedinBezuijen&Talmon(2003). Grout
bleeding or consolidation of the grout leads to a
volume loss of grout. Experiments showed that this
volumelossisbetween3%and8%, dependingonthe
typeof grout(Bezuijen&Zon, 2007).Thisconsolida-
tionleadstostressreductionintherelativelystiff sand
layer. Thisstressreductionismeasuredasareduction
of groutpressure.Theeffectivestresseswill ultimately
beverysmall: theminimumstressthatisnecessaryto
keeptheholeinthegroundopen. Leca& Dormieux
(1990)calculatethisforatunnel openinginsand.They
calculatethatacylindrical cavityinthegroundremains
open when effective stresses of only a few kPa are
applied.
Theconsequenceisthatgroutpressuresaroundthe
liningwill decreasetovaluesthat areonly afewkPa
abovetheporewater pressure. Itisthereforeclear that
theoriginal K
0
cannolongerbefoundinthegroutpres-
sures. Thepressuredecreaseduetovolumelossinthe
grouthaschangedtheoriginal stressstate, andunload-
ingof thesoil leadstomuchlower stresses. Sincethe
stresses in the sand around the tunnel decrease, the
sandreactionwill bethereactionof averystiff mate-
rial. Only asmall volumedecreasein thegrout will
leadtoalargedecreaseinstresses. Calculationmeth-
odsquiteoftenstill usetheoriginal in-situstressesto
calculateloadingonthelining. For atunnel insand,
thisleadstoacalculatedloadingthatismuchtoohigh,
asshownbyHashimotoet al (2004).
For slow hardening or non-hardening grouts, the
strengthincreaseinthegroutiscausedbygroutbleed-
ing or consolidation. It should be realised that this
strength increase is only present when the tunnel is
drilledthroughapermeablesoil. Whendrillingtakes
placethrough less permeablesoils such as clay, this
consolidationwill bemuchlowerandthegroutwill be
inliquidformoveragreaterpartof thetunnelslength.
Thishasconsequencesfor loadingonthelining, aswe
will discusslater.
4.4 Pressure gradients
Thevertical pressuregradient over thetunnel lining
isimportantwhencalculatingthelongitudinal loading
onthelining. Thevertical pressuregradient that was
measuredduringconstructionof thefirsttunnel tubeof
theSophiaRail Tunnel isshowninFigure8.Thepres-
suregradient starts at nearly 20kPa/manddecreases
to values under the pore water pressure gradient of
Figure8. First tubeSophiaRail Tunnel: pressuregradient
over thetunnel liningat onelocation, andpumpactivity for
oneof theinjectionpoints(A1) asafunctionof time.
9.81kPa/m. The tail void grout used for this tunnel
had adensity of 2190kg/m
3
. If thevertical pressure
were to increase with depth in accordance with this
density, thepressuregradient should be21.5kPa/m.
Results showed that themeasured vertical density is
alwayslower. Thisisbecausethegrout isaBingham
liquid, with aviscosity and ayield stress. Thegrout
hastoflowdownwardsif moregroutisinjectedinthe
upper half of thetunnel. This downward flow needs
adriving forceto overcometheyield stress, and the
pressuregradient will thereforebeless thanthegra-
dient that iscalculatedfromthedensity. Talmonet al
(2001) developedanumerical programtocalculatethe
pressuredistributioninthetail voiddueto injection.
Weonlydescribesomeof theconsequenceshere. If the
viscosityisnottakenintoconsideration, themaximum
pressuregradient (dP/dz) that canbeexpectedis:
Where
gr
isthedensity of thegrout, g theaccelera-
tionof gravity,
theyieldstrengthof thegrout, ands

thewidthof thetail voidgapbetweenthesoil andthe
lining. If theyieldstressinthegroutislow, thevertical
pressuregradient is determined by thegrout density
(21.5kPa/mfortheSophiaRail Tunnel, slightlyhigher
thanthemaximumvaluemeasuredinFigure8). Con-
solidationor hardeningof thegrout leadstoahigher
yieldstress, andthustoalower gradient.
A complicating factor is that themaximumshear
stress that canbedevelopedis avector. If themaxi-
mumshear stressisdevelopedinonedirection, there
will be no shear stress perpendicular to that direc-
tion. Whendrillingstartsfor anewringandthegrout
pumpsareactivated, theelasticsoil reactionwill lead
toanincreaseof thetail voidandgrout will therefore
flow backwards fromtheTBM. Ring shear stresses
barelydevelopinthissituation, andthevertical gradi-
entsthereforeincreaseduringdrilling. Theydecrease
againwhendrillingstops(Figure8).
8
Further from the TBM, the vertical gradients
decrease and become equal to the gradient accord-
ing to the buoyancy forces. This has to be the case,
because the total force on the lining far away from
theTBM must be zero. The vertical pressure gradi-
entthereforecompensatesfor theweightof thelining.
Asaresult, thegradient becomeslower thanthegra-
dient in the pore water. This is because the average
density of theliningislower thanthedensity of pore
water. Oneremarkableresult isthat thevertical pres-
suregradientatsomedistancefromtheTBM(at12:00
inFigure8, 5ringsbehindtheTBM) decreasesduring
drilling. Theflownolonger has any influenceat this
point, but drilling and grout injection lead to higher
gradientsinthefirstpartof theliningandthereforeto
higher buoyancy forces. Thefirst rings havetheten-
dencytomoveupwards, whichmust becompensated
by theTBM and the rings further away. This partly
compensates for theweight of therings further from
theTBM, so that theeffectiveweight of theserings
andalsothevertical gradient isless.
5 INFLUENCE ONPORE WATER PRESSURES
Section2.1describeshownoplasteringoccursat the
front when drilling takes place in fine to medium-
finesaturatedsand, becausethebentonitefiltercakeis
destroyedbythecuttingwheel beforeitisabletoform.
Asaresult, water flowsfromthetunnel faceintothe
soil. Section 4.3 describes how consolidation of the
grout alsoleadstoawater flowfromthetunnel lining
into thesoil, becausewater expelled fromthegrout
will flowintothesurroundingsoil. A grout cakewill
formhowever, because the consolidated grout is no
longer disturbed. It isthereforereasonabletoassume
that examination of thevariation in porepressurein
soil nexttoatunnel under constructionwill showpore
pressures that aredominatedby pressures existingat
thetunnel face. This theory was testedat theGroene
Hart Tunnel. Pore pressure transducers (PTTs) were
installedascloseas0.75mfromthetunnel lining.The
PTTswereplacedinoneplane, withthegroutpressure
gaugesonRing2117of thetunnel (seeFigure9).
Figure10showsthemeasurementresults.Thegrout
pressuregauges on Ring 2117 giveno signal before
theyareinthegrout. ThePPTsshowaslight increase
during drilling due to the excess pore pressure gen-
erated at the tunnel face. As drilling stops, the pore
pressurereducestothehydrostaticpressure. Thevar-
iousconstructioncyclescanbeseen. Thereisasharp
increaseingrout pressurewhenRing2117leavesthe
TBM, followedby adecreaseduetoconsolidation. It
isremarkablehowever that thishasvirtuallynoinflu-
ence on the measured pore pressures at less than a
metrefromthesegauges. Thisresult isconfirmedby
numerical calculations.Thequantityof waterexpelled
Figure9. Positionof porepressuregaugesandgrout pres-
suregaugesat ring2117of theGHT.
Figure10. Porepressuresandgrout pressuresmeasuredat
GHT (alsoseetext).
fromthegrout isfar lessthanthewater flowfromthe
tunnel face. Thelatter dominatestheporepressures.
The measurements show another remarkable fea-
ture. Grout pressuregauge05followsthewater pres-
sureafter3.20:00, butthisisnotthecaseforgauge03.
Thismayindicatethat thereisnosealing grout layer
aroundgauge05, sothat it ispossibletomeasurethe
porewater pressure.
6 LOADINGONTUNNEL LINING
WehaveseeninSection4.4thatvertical pressuregra-
dients exist in the zone where the grout is not yet
consolidatedor hardenedwhicharehigher thancorre-
spondstotheweightof thelining.Measurementsatthe
SophiaRail Tunnel showedthatthegradientdecreases
moreor less linear withthedistance(seeFigure11).
Asaresult, thatpartof theliningispressedupwardsby
9
Figure11. Exampleof gradient inthegrout pressureas a
function fromthe distance (0 on the X-axis represents the
point wheretheliningis moreor less fixed. TheTBM is at
9m). ResultsfromSophiaRail Tunnel (Bezuijenetal, 2004).
Figure12. Calculated shear forceand moment in thelin-
ing, anddisplacement wherethegrout hasnot yet hardened.
Calculatedmomentsaredividedby10.
thebuoyancyforces. It isnecessarytomobiliseshear
forcesfromtheTBMtoachieveastabletunnel lining.
Thiswill leadtomomentsinthelining.
Bezuijen & Talmon (2005) have shown that the
momentsintheliquidgrout zoneincreasebackwards
fromtheTBM (see Figure 12). A positive moment
meansherethattheforceonthelower partof thetube
is higher than on the upper part. At the TBM, this
moment iscreatedbytheTBM itself. Thisisbecause
facepressureishigher at thebottomduetolarger soil
stresses.
At theGroeneHartTunnel thebendingmoment in
thelining was measured for a largedistancebehind
theTBM using strain gauges installed in the lining
segments.Thereisanincreaseinthemomentforafew
rings, inaccordancewiththecalculations previously
mentioned. Thereissubsequentlyadecrease, withthe
momentsbecomingnegativeatagreaterdistancefrom
thetunnel. Bogaards& Bakker (1999) andHoefsloot
(2008) arguethat theremainingbendingmoment isa
result of thestaged construction of thetunnel. They
developedacalculationmodel totakeintoaccountthe
different stagesinconstruction.
Figure13. Boundaryconditionfor beamcalculation.
200
150
100
50
0
50
100
150
200
20 0 20 40 60 80 100 120 140 160 180
Disance behind TBM [m]
Analytical model
Measurement Ring
2117
B
e
n
d
i
n
g

m
o
m
e
n
t

[
M
N
m
]
Figure14. Bendingmoment ring2117, measurement and
calculation. GroeneHartTunnel (Hoefsloot, 2008).
However, Talmon (2007) has shown that such a
staged calculationisnot necessary tofindthesame
results.
AccordingtoTalmon, thenegativemomentappears
at somedistancefromtheTBM becausethereaction
forceto compensatethebuoyancy in thefluid grout
zoneissituatedfurther fromtheTBM thanthebuoy-
ancy force itself. The tunnel lining is pushed a bit
higher inthesoil thanintheequilibriumsituationfar
behindtheTBM.
HoefslootandTalmonbothmodel thetunnel lining
as abeamonanelastic foundation, except for lining
elements insidetheTBM and lining elements in the
liquidgrout zone, seeFigure13. Theexact boundary
conditionsandthetransitionbetweenliquidandsolid
grout arestill thesubject of debate.
Althoughexamplecalculationshavebeenpresented
that show good correlation with measurements (see
Figure14), therearestill uncertainties withthis type
of calculationthat needfurther research:
An important input parameter is themoment and
shear forcethat istransferredfromtheTBM tothe
lining. Whilethemoment canbederivedfromthe
jackforces, theshear forceisnot determined.
Withgenerally-acceptedparameters for thelining
stiffnessandthesoilselasticparameters, thecalcu-
latedmovement of theliningismuchsmaller than
themeasuredmovement.
The grout pressures are only measured when the
groutismoreorlessintheliquidphase.Thisresults
10
Table 3. Specification of grout mixtures used in fracture
tests(WCR=water-cement ratio). CoclayD90Caactivated
bentoniteisused.
Bentonite k
Mixture WCR % (m/s)
1 1 7 5.10
8
2 10 7 6.10
0
in loading on the tunnel lining as shown in Fig-
ure11. However, loadingonthelininginsituations
wherethegrouthashardenedislessknown.Thisis
becausetheinstruments usedwerenot suitableto
measurepressureswhengrout hashardened.
Conclusions that can be drawn fromthis type of
calculationsare:
Thelengthof theliquidgrout zoneandthedensity
of the grout are extremely important parameters
whencalculatingbendingmomentsinthelining. If
thislengthistoolong, loadingwill betoohighand
tunnellingwill notbepossible(alsoseeBezuijen&
Talmon, 2005).
The shear force that is exerted on the lining by
theTBM isanimportant parameter. It istherefore
worthwhiletomeasurethisshear force.
7 COMPENSATIONGROUTING
Grout consolidation also appeared to be impor-
tant whendescribingcompensationgrouting. Exper-
iments (Gafar et al, 2008) showedthat thefracturing
behaviour in compensation grouting depends on the
specification of the grout. If more cement is added,
thepermeability of thegrout is higher andtherewill
bemoreconsolidationandleak-off duringgroutinjec-
tion. Gafar et al describe how this influences the
fracturingbehaviour. Recenttestscarriedoutaspartof
theresearchprojectoncompensationgroutingpresent
proof of thesuggestedgroutconsolidationmechanism.
AtDelftUniversity,thedensityof groutbodiesmadein
twocompensationgroutingexperimentswasanalysed
inaCT-scan. SuchaCT-scancanbeusedtodetermine
thedensity of thematerial tested. Thegrout mixtures
usedintheexperimentsareshowninTable3.
Theresultsof theCT-scansareshowninFigure15
and Figure16. Theresults of thefirst grout mixture
clearlyshowanincreaseindensityat theboundaryof
thegrout body. Grout at theboundary of thesample
isconsolidated. Thegroutbodymadewiththesecond
mixturehas amoreconstant density across thefrac-
ture(themiddlesection). In thesecond experiment,
theCT-scanwasperformedwhilethegrout bodywas
still in the sand. The more homogeneous density of
Figure 15. Density measured with a CT-scan. Raw data
(inset) anddensity. Correctionfor beamhardeningeffectand
calculated value of the density of the grout along the line
shownininset. Mixture1inTable3.
Figure 16. Grout density in a fracture measured with a
CT-scan. Mixture2inTable3.
the grout body in the second test is understandable
if thepermeabilities of thegrout areconsidered. The
lower permeabilityof thesecondgrout sampleresults
in much less grout consolidation within the limited
injectiontime. Thegroutdensityinthefracturethere-
foredoesnot increaseat theboundaryof thegrout as
isthecasefor mixture1.
Thepermeabilitiesweredeterminedusingthepro-
ceduresuggestedby McKinley andBolton(1999), a
formof oedometertestwithdrainageononeside.This
procedurecan also beused to test theconsolidation
properties of tail void grout. However, thethickness
of thegroutlayer inthetestshouldbeidentical tothat
inthefield. Thisistoavoidscalingeffectsthat occur
11
becausehardeningof thegrout is independent of the
samplesize(Bezuijen& Zon, 2007).
8 DISCUSSION
The research described above has increased under-
standing of the processes that occur around aTBM
duringtunnelling. Thishasalreadyhadconsequences
for practical aspects of tunnelling. Examples arethe
excessporepressuresinfront of theTBM: extrasand
was added locally above the planned tunnel trajec-
toryof theGroeneHartTunnel toprevent ablow-out,
and the grout was changed in a tunnel project in
Londonwhereit appearedthat theliquidzoneof tra-
ditional grout for a tunnel drilled in clay with no
possibility of consolidation was too long to achieve
thedesireddrillingspeed.However,theauthorsbelieve
that theresultscanmakeanevengreater contribution
toimprovingshieldtunnelling. Knowledgeabout the
influenceof excessporepressuresonfacestabilitycan
improvedefinitionof thepressurewindowat thetun-
nel face, sopreventingablow-out duetoexcessively
highpressuresandinstabilitycausedbypressuresthat
aretoolow. IncombinationwithresearchonEPBtun-
nellinginclay (Merrit & Mair, 2006), foamresearch
for EPB tunnellinginsandcanleadto better control
of theEPBprocess. Ithasalreadybeendiscussedhow
flowaroundtheTBMisimportantforTBMdesign,and
that moreexperimental evidenceisneeded. Research
into grouting can lead to smaller settlement troughs
and optimisation of loading on the lining. This last
aspect mayleadtocheaper liningconstruction.
Theresultsmust bediscussedwithtunnel builders
and contractors if improvements to the shield tun-
nellingprocess areto beachieved. Discussionabout
certain aspects has already started, but wehopethat
this paper will stimulate the involvement of more
parties.
9 CONCLUSIONS
To understandtheprocesses that areimportant when
tunnelling with aTBM, theflowprocesses around a
TBMmustbeconsidered: groundwaterflowatthetun-
nel face,bentoniteandgroutflowaroundtheTBM,and
grout flowandgrout consolidationaroundthetunnel
lining.Theresearchdescribedinthispaperhasbrought
about progress with regard to these flow processes
duringtunnellinginsoft ground:
The groundwater flow at the tunnel face is
described.
Themuck inthemixingchamber isdescribedasa
functionof drillingspeedandpermeability.
A conceptual model for theflowof bentoniteand
grout has been developed. Although this model
must still beverifiedusingtheresultsof measure-
ments, it showssomepromisingresults.
Considerableinformationhasbeenobtainedabout
the grouting process and the resultant lining
loading.
Although not unusual, it is interesting to see that
this research also raises new questions: what is the
exact positionof theTBM duringthetunnellingpro-
cess, whatistheinteractionbetweentheTBM andthe
lining, are the predicted pressures around theTBM
correct, andwhat aretheconsequencesfor our design
methods?Eveninarelativelysimplebeamcalculation
for calculatingloadingonthelininginalongitudinal
directionit appearsthat uncertaintiesintheboundary
conditions determinetheoutcomeof thecalculation.
As long as these uncertainties remain, more sophis-
ticated numerical calculations will present the same
uncertainties.
ACKNOWLEDGEMENTS
Theresearchdescribedinthispaper wassponsoredby
COB, the Dutch Centre for Underground Construc-
tion, andDelft Cluster. Wewouldliketo thank these
organisationsfor givingustheopportunitytoperform
thisresearch. Wealsowishtothanktheprojectorgan-
isationsof thedifferent tunnelsfor givingpermission
tousetunnellingdatainour research. Andlastbutnot
least, wewouldliketo thank our fellowmembers in
theCOB committeesfor their stimulatingdiscussions
onthevarioussubjects.
REFERENCES
Aime, R., Aristaghes, P., Autuori, P. & Minec, S. 2004. 15m
Diameter Tunneling under Netherlands Polders. Proc.
Underground Space for Sustainable Urban Development
(ITA Singapore), Elsevier.
Anagnostou, G. & Kovri, K. 1994. The face stability of
Slurry-shield-drivenTunnels.Tunelling and Underground
Space Technology 9(2): 165174.
Bakker, K.J. & Bezuijen, A. 2008. 10 years of bored tun-
nels in theNetherlands. Proceeding 6th Int. Symposium
on Underground Construction in soft Ground, Shanghai.
Bezuijen, A. 2002. Theinfluenceof soil permeabilityonthe
properties of afoammixturein aTBM. 3rd. Int. Symp.
on Geotech. Aspects of Underground Construction in Soft
Ground, IS-Toulouse.
Bezuijen, A. 2007. Bentoniteandgrout flowaroundaTBM.
Proc. ITA 2007, Prague.
Bezuijen,A., Pruiksma, J.P. &Meerten, H.H. van. 2001. Pore
pressures in front of tunnel, measurements, calculations
and consequences for stability of tunnel face. Proc. Int.
Symp. on Modern Tunneling Science and Techn. Kyoto.
12
Bezuijen, A. &Talmon, A.M. 2003. Grout thefoundationof
aboredtunnel. Proc ICOF 2003 Dundee.
Bezuijen,A.,Talmon,A.M.,Kaalberg,F.J.&Plugge,R.2004.
Fieldmeasurements of grout pressures duringtunneling
of theSophiaRail tunnel. Soils and Foundations 44(1):
4150.
Bezuijen, A. & van Lottum, H. (eds). 2006. Tunnelling A
Decade of Progress. GeoDelft 1995-2005, Taylor and
Francis/Balkema, Leiden, ISBN0415391134.
Bezuijen, A. & Zon, W. van der. 2007. Volumechanges in
groutusedtofill upthetail void. Proc. ITA 2007, Prague.
Bogaards, P.J. & Bakker, K.J. 1999, Longitudinal bending
momentsinthetubeof aboredtunnel. Numerical Models
in Geomechanics Proc. NUMOGVII: 317321.
Broere, W. 2001. Tunnel Face Stability & New CPT Applica-
tions. Ph.D. Thesis, Delft Universityof Technology, Delft
UniversityPress.
Gafar, K., Soga, K., Bezuijen, A., Sanders, M.P.M. & van
Tol, A.F. 2008. Fracturingof sandincompensationgrout-
ing. Proceeding 6th Int. Symposium on Underground
Construction in soft Ground, Shanghai.
Guglielmetti, V. 2007. Tunnels and Tunnelling International,
October, P32.
Hashimoto, T., Brinkman, J., Konda, T., Kano, Y. &
Feddema, A. 2004. Simultaneousbackfill grouting, pres-
suredevelopment in construction phaseand in thelong
term. Proc. ITA Singapore.
Hoefsloot, F.J.M. 2008. Analytical solution longitudinal
behaviour Tunnel lining. Proceeding 6th Int. Symposium
Kasper, T. & Meschke, G. 2006. On the influence of face
pressure,groutingpressureandTBMdesigninsoftground
tunnelling.Tunn. and Undergr. SpaceTechn. 21: 160171.
Leca, E. &Dormieux, L. 1990. Upper andlower boundsolu-
tions for thefacestability of shallowcircular tunnels in
frictional material. Gotechnique 43: 519(inFrench).
Merritt, A.S. & Mair, R.J. 2006. Mechanics of tunnelling
machine screw conveyors: model tests. Geotechnique
56(9): 605615.
McKinley, J.D. &Bolton, M.D. 1999.Ageotechnical descrip-
tion of fresh cement grout Filtration and consolida-
tion behaviour. Magazine of Concrete Research 51(5):
295307.
Steiner, W. 1996. Slurry penetration into coarse grained
soils and settlements froma large slurry shield tunnel.
Proc. Geotech. Aspects of Underground Construction in
Soft Ground, London, Mair andTaylor (eds). Balkema,
Rotterdam, ISBN9054108568: 329333.
Steiner, W. 2007. Privatecommunication.
Talmon, A.M. 2007. Notes on analytical beam model. Delft
Hydraulicsreport Z3934/Z4145.
Talmon, A.M., Aanen, L. Bezuijen, A & Zon, W.H. vander.
2001. Grout pressures around a tunnel lining Proc. Int.
Symp. on Modern Tunneling Science and Techn. Kyoto.
Thewes, M. 2007. Privatecommunication.
Verruijt, A. 1993. Soil Dynamics. Delft University of Tech-
nology, b28.
13
Supportingexcavationsinclay fromanalysistodecision-making
M.D. Bolton& S.Y. Lam
University of Cambridge, UK
A.S. Osman
Durham University, UK
ABSTRACT: FiniteElementAnalysis(FEA) isusedtocalibrateadecision-makingtool basedonanextension
of theMobilizedStrengthDesign(MSD) methodwhichpermits thedesigner anextremely simplemethodof
predicting ground displacements during construction. This newly extended MSD approach accommodates a
number of issueswhichareimportant inundergroundconstructionbetweenin-situwalls, including: alternative
base heave mechanisms suitable either for wide excavations in relatively shallow soft clay strata, or narrow
excavationsinrelativelydeepsoft strata; theinfluenceof support systemstiffnessinrelationtothesequenceof
proppingof thewall; andthecapability of dealingwithstratifiedground. Thesedevelopmentsshouldmakeit
possiblefor adesignengineer totakeinformeddecisionsontherelationshipbetweenpropspacingandground
movements, or theinfluenceof wall stiffness, or ontheneedfor andinfluenceof ajet-groutedbaseslab, for
example, without havingtoconduct project-specificFEA.
1 INTRODUCTION
TheMobilizableStrengthDesign(MSD) methodhas
developed following various advances in the use of
plasticdeformationmechanismstopredictgrounddis-
placements: (MilliganandBransby, 1975; Boltonand
Powrie, 1988; Boltonetal. 1989, 1990a, 1990b). MSD
isageneral, unifieddesignmethodology, whichaims
tosatisfy bothsafety andserviceability requirements
inasinglecalculationprocedure,contrastingwithcon-
ventional design methodology which treats stability
problems and serviceability problems separately. In
theMSD method, actual stress-strain datais used to
select a design strength that limits ground deforma-
tions,andthisisusedinplasticsoil analysesthatsatisfy
equilibriumconditions without the use of empirical
safetyfactors.
Simpleplasticmechanismsareusedtorepresentthe
workingstateof thegeotechnical system. Themecha-
nismsrepresentboththeequilibriumanddeformation
of thevarioussoil bodies, especially at their junction
with the superstructure. Then, raw stress-strain data
fromsoil testsonundisturbedsamples,takenfromrep-
resentativelocations, areuseddirectlytolinkstresses
and strains under working conditions. Constitutive
lawsandsoil parametersareunnecessary.
The MSD approach has been successfully imple-
mentedfor shallowfoundations (OsmanandBolton,
2005), cantilever retainingwalls(OsmanandBolton,
2005), tunneling-induced ground displacements
(Osman et al. 2006) and also the sequential con-
struction of braced excavations which induce wall
displacements andgrounddeformations (Osmanand
Bolton, 2006).
Consider theimpositionof certainactionsonasoil
body, duetoconstructionactivitiessuchasstressrelief
accompanyingexcavation, or to loads appliedinser-
vice. The MSD method permits the engineer to use
simplehand calculations to estimatetheconsequen-
tial ground displacements accounting for non-linear
soil behavior obtainedfromasinglewell-chosentest
of theundisturbedsoil.
TheMSDapproachfirstlyrequirestheengineer to
representtheworkingstatesof thegeotechnical system
byagenericmechanismwhichconveysthekinematics
(i.e. thepatternof displacements) of thesoil duetothe
proposedactions.Analysisof thedeformationmecha-
nismleadstoacompatibilityrelationshipbetweenthe
averagestrainmobilizedinthesoil andtheboundary
displacements.
Theaverageshearstrengthmobilizedinthesoil due
totheimposedactionsisthencalculated,eitherfroman
independent equilibriumanalysisusingapermissible
stressfield(equivalenttoalower boundplasticanaly-
sis),orfromanequationbalancingworkandenergyfor
thechosenmechanism(equivalent toanupper bound
plasticanalysis).
Thelocationof oneor morerepresentativesoil ele-
ments is then selected, basing this judgment on the
soil profile in relation to the location and shape of
15
the selected mechanism. The centroid of the mech-
anism can serve as a default location if a single
locationistobeemployed. Stress-strainrelationships
are then obtained fromappropriate laboratory tests
on undisturbed soil samples taken fromtheselected
locations and carried out with precise strain mea-
surements. Equivalentin-situtestssuchasself-boring
pressuremeter tests can alternatively be carried out.
Themodeof deformationinthesoil testsshouldcor-
respondascloselyaspossibletothemodeof shearing
intheMSDmechanism. Otherwise, anisotropyshould
somehowbeallowedfor.
Finally, the mobilized shear strength required for
equilibriumunderworkingloadsissetagainsttherep-
resentativeshear stress-straincurveinorder toobtain
the mobilized soil strain, and thereby the boundary
displacementsof thesimplifiedMSDmechanism.
2 MSDFOR DEEP EXCAVATIONPROBLEM
OsmanandBolton(2006) showedfor anin-situwall
supporting a deep excavation that the total defor-
mation could be approximated as the sum of the
cantilever movementprior topropping, andthesubse-
quent bulging movement that accretes incrementally
witheverysequenceof proppingandexcavation.
A methodfor estimatingthecantilever movement
had been suggested earlier in Osman and Bolton
(2004). Itbeginsbyconsideringthelateral earthpres-
sure distribution for a smooth, rigid, cantilever wall
rotating about a point some way above its toe, in
undrained conditions. A simple mobilized strength
ratioisintroducedtocharacterizetheaveragedegree
of mobilizationof undrainedshear strengththrough-
out the soil. By using horizontal force and moment
equilibriumequations, thetwo unknowns theposi-
tion of the pivot point and the mobilized strength
ratio are obtained. Then, a mobilized strain value
isreadoff fromtheshear stress-straincurveof asoil
element appropriatetotherepresentativedepthof the
mechanismatthemid-depthof thewall. Simplekine-
matics for a cantilever wall rotating about its base
suggeststhattheshearstrainmobilizedintheadjacent
soil isdoubletheangleof wall rotation. Accordingly,
fortheinitial cantileverphase, thewall rotationisesti-
matedasonehalf of theshearstrainrequiredtoinduce
thedegreeof mobilizationof shear strengthnecessary
to hold the wall in equilibrium. Osman and Bolton
(2004) usedFEA toshowthatcorrectionfactorsupto
about 2.0 could beapplied to theMSD estimates of
thewall crestdisplacement, dependingonavarietyof
non-dimensional groupsof parametersignoredinthe
simpleMSDtheory, suchaswall flexibilityandinitial
earthpressurecoefficient prior toexcavation.
A typical increment of bulging, ontheother hand,
was calculated in Osman and Bolton (2006) by
consideringanadmissibleplasticmechanismfor base
heave. Inthis case, themobilizedshear strengthwas
deduced fromthe kinematically admissible mecha-
nismitself, usingvirtual work principles. Theenergy
dissipatedbyshearingwassaidtobalancethevirtual
loss of potential energy dueto thesimultaneous for-
mation of a subsidence trough on the retained soil
surface and a matching volume of heave inside the
excavation.Themobilizedstrengthratiocouldthenbe
calculated,andthemobilizedshearstrainreadoff from
thestress-straincurveof arepresentativeelement, as
before. Thedeformation is estimated using therela-
tionshipbetweentheboundarydisplacementsandthe
averagemobilizedshear strain, inaccordancewiththe
original mechanism.
TheMSD solutions of Osman and Bolton (2006)
comparedquitewell withsomenumerical simulations
usingtherealisticnon-linear MIT-E3model, andvar-
ious case studies that provided field data. However,
theseinitial solutions arecapableof improvement in
threewaysthatwill contributetotheir applicabilityin
engineeringpractice.
1 Theoriginal mechanismassumedarelativelywide
excavation, whereascut-and-cover tunnel andsub-
wayconstructionsarelikelytobemuchdeeperthan
their width. TheMSDmechanismthereforeneeds
tobeadaptedforthecaseinwhichtheplasticdefor-
mationfieldsfor thesidewallsinterferewitheach
other beneaththeexcavation.
2 The structural strain energy of the support sys-
temcanbeincorporated. Thiscouldbesignificant
when the soil is weak, and when measures are
taken to limit base heave in the excavation, such
asby basegroutingbetweenthesupportingwalls.
In this case, the reduction of lateral earth pres-
suredueto grounddeformationmay berelatively
small, and it is principally the stiffness of the
structural systemitself that limitsexternal ground
displacements.
3 Progressively incorporating elastic strain energy
requiresthecalculationproceduretobefullyincre-
mental, whereas Osman and Bolton (2006) had
been able to use total energy flows to calculate
theresults of each stageof excavation separately.
A fullyincremental solution, admittinggroundlay-
ering, will permit the accumulation of different
mobilized shear strengths, and shear strains, at
different depths in theground, thereby improving
accuracy.
It istheaimof thispaper tointroduceanenhanced
MSDsolutionthat includesthesethreefeatures. This
is thencomparedwithexistingFEA of bracedexca-
vations which featured a range of geometries and
stiffnesses. It will be suggested that MSD provides
theideal meansof harvestingFEA simulationsfor use
indesignanddecision-making.
16
3 PLASTIC FAILURE MECHANISMS
Limitequilibriummethodsareroutinelyusedinstabil-
ity calculationsfor soft clay whichisidealised, unre-
alistically, as rigid-plastic. Slip surfaces areselected
astheassumedfocusof all plasticdeformations. Fail-
uremechanisms shouldbekinematically admissible,
meaningthat unwantedgapsandoverlapsshouldnot
be produced. Furthermore, in the case of undrained
shearingof clays, aconstant-volumeconditionshould
be respected at every point. A consequence is that
undrainedplane-strainfailuremechanismsmustcom-
priseonlyslipplanesandslipcircles.Thesoil onsuch
failuresurfacesistakentomobilizeitsundrainedshear
strength divided by a safety factor, to maintain the
mechanismin limiting equilibriumunder the action
of gravity, andany other appliedloads. Calculatedin
thisway, thesafetyfactorliterallyoffersanestimateof
thefactor bywhichthestrengthof thesoil wouldhave
to drop before the soil construction would collapse.
Suchestimatesmighterr either onthehighsideor the
lowside, dependingontheparticularassumptionsthat
weremade.
Inthecaseof baseheaveinbracedexcavations,plas-
tic solutions werederivedfromslip-linefields based
onthemethodof characteristics. Suchsolutionscom-
prise both slip surfaces, as before, and plastic fans
which distribute plastic strains over a finite zone in
the shape of a sector of a circle. Notwithstanding
these zones of finite strain, the additional presence
of slipsurfaces still restricts theapplicationof these
solutions to the prediction of failure. Furthermore,
no such solution can be regarded automatically as
anaccuratepredictor of failure, notwithstandingtheir
apparent sophistication. All that can be said is that
theywill leadtoanunsafeestimateof stability. Their
useinpracticecanonly bejustifiedfollowingback-
analysisof actual failures, whether inthefieldor the
laboratory.
Two typical failure mechanisms as suggested by
Terzaghi (1943) and Bjerrum and Eide (1956) are
showninFigure1. They haveeachbeenwidely used
for the design of multi-propped excavations. Terza-
ghi (1943) suggestedamechanismconsistingof asoil
columnoutsidetheexcavationwhichcreatesabearing
capacity failure. Thefailureis resistedby theweight
of acorrespondingsoil columninsidetheexcavation
andalso by adhesionactingalongthevertical edges
of themechanism. BjerrumandEide(1956) assumed
that thebaseof theexcavation could betreated as a
negativelyloadedperfectlysmoothfooting. Thebear-
ingcapacityfactorsproposedbySkempton(1951) are
useddirectlyinthestabilitycalculationsandaretaken
as stability numbers, N=H/c
u
. Eide et al. (1972)
modifiedthisapproachtoaccount for theincreasein
basal stabilityowingtomobilizedshearstrengthalong
theembeddedlengthof therigidwall.
Figure1. Conventional basal stabilitymechanismandnota-
tion(after Ukritchonet al. 2003).
ORourke(1993) further modifiedthebasal stabil-
itycalculationsof BjerrumandEide(1956) toinclude
flexureof thewall belowtheexcavationlevel. It was
assumed that the embedded depth of the wall does
not changethegeometry of thebasal failuremecha-
nism. However, anincreaseinstabilitywasanticipated
duetotheelasticstrainenergystoredinflexure. This
gavestabilitynumbersthatwerefunctionsof theyield
momentandassumedboundaryconditionsatthebase
of thewall.
Ukritchonet al. (2003) usednumerical limit anal-
ysis to calculate the stability of braced excavations.
Upper and lower bound formulations are presented
basedonSloanandKleeman(1995) andSloan(1988),
respectively.Thetechniquecalculatesupperboundand
lower boundestimates of collapseloads numerically,
by linear programming, while spatial discretization
and interpolation of the field variables are calcu-
lated using the finite element method. No failure
17
Figure2. Incremental displacements inbracedexcavation
(after ORourke, 1993).
mechanismneed beassumed and failureboth of the
soil andthewall aretakencareof. However, bothsoil
andwall areagainassumedtoberigidperfectlyplas-
tic so thefailuremechanismincludes aplastic hinge
at thelowest level of support.
All these collapse limit analyses provide useful
guidanceonthepossiblegeometryof plasticdeforma-
tion mechanisms for serviceconditions. But thekey
requirement for MSD mechanisms is that displace-
ment discontinuities (slip surfaces) must beavoided
entirely. Inthat way, small but finitegrounddisplace-
mentsareassociatedateveryinternal pointwithsmall
but finitestrains.
4 WALL DEFORMATIONS
Consider now the deformations of a multi-propped
wall supportingadeepexcavationinsoft, undrained
clay. At eachstageof excavationtheincremental dis-
placementprofile(Figure2)of thegroundandthewall
belowthelowest propcanbeassumedtobeacosine
function(ORourke, 1993) asfollows:
Herew is theincremental wall displacement at any
distance y below the lowest support, w
max
is its
maximumvalue, and l isthewavelengthof thedefor-
mation, regardedasproportional tothelengths of the
wall belowthelowest level of current support:
ORourke(1993) definedthewavelengthof thedefor-
mationasthedistancefromthelowestsupportlevel to
thefixedbaseof thewall. OsmanandBolton(2006)
suggestedadefinitionforthewavelengthof thedefor-
mationbasedonwall endfixity. For wallsembedded
into a stiff layer beneath the soft clay, such that the
wall tip is fully fixed in position and direction, the
wavelength was set equal to thewall length (=1).
For short wallsembeddedindeepsoft clay, themaxi-
mumwall displacementoccursatthetipof thewall so
thewavelengthwastakenastwicetheprojectingwall
length(=2). Intermediatecasesmight bedescribed
asrestrained-endwalls(1--2).
However, these definitions applied only to very
wideexcavations. Whenanarrowexcavationis con-
sidered, thewavelength will belimited by thewidth
of theexcavation. Inaddition, inthecaseof thepar-
tiallyrestrainedwall, thedepthof arelativelystiff soil
stratummay also limit thedepth of thedeformation
pattern.
5 GEO-STRUCTURAL MECHANISMS
Anincremental plastic deformationmechanismcon-
forming to Equation 1 was proposed by Osman and
Bolton (2006) for an infinitely wide multi-propped
excavation in clay. In this mechanism, the wall is
assumed to be fixed incrementally in position and
directionatthelowestprop, implyingthatthewall has
sufficient strength to avoid the formation of a plas-
tichinge. Thewall andsoil aredeformingcompatibly
andthesoil deformationalsofollowsthecosinefunc-
tionof Equation1.Thedimensionsof thismechanism
dependonthewavelengthl.
Figure3(a) showsthecompletedisplacement field
for the mechanismproposed by Osman and Bolton
(2006).Thesolutionincludesfourzonesof distributed
shear whichconsist of acolumnof soil adjoiningthe
excavationabovethelevel of thelowestprop,acircular
fanzonecentredatthelowestprop,anothercircularfan
zonewithits apex at thejunctionof thewall andthe
excavation surface and a 45 degree isosceles wedge
below the excavation surface. It is required that the
soil shearscompatiblyandcontinuouslywithnorela-
tiveslidingattheboundariesof eachzone. Thedotted
lineswitharrowsshowthedirectionof theflow.Along
each of theselines thedisplacement is constant and
is given by the cosine function of Equation 1. It is
assumedthat thezoneoutsidethedeformationzones
is rigid. This mechanismis simple and neat, but it
onlyappliestoverywideexcavations. Inthecaseof a
narrowexcavation, thewidthof thetriangular wedge
couldbebiggerthantheactual widthof theexcavation.
Inviewof this, anewmechanismfor narrowexcava-
tions is proposed in Figure 3(b). The mechanismin
the passive zone (zone EFHI) is replaced. The new
mechanismmeets theconditionfor undrainedshear-
ing, which means that thevolumetric strain remains
zerothroughout thezone.
The following solution approach is an extension
of Osman and Bolton (2006). In their original solu-
tion, soilsareassumedtobehomogenous.Theaverage
shear strain increment in each zoneis calculated by
taking the derivative of the prescribed displacement
18
equation. Then, theundrainedshear strength(c
u,mob
)
mobilized at any location for any excavation height-
was expressed using a single mobilization ratio
( =c
u,mob
,c
u
) tofactor thestrengthprofile. Withthe
use of the virtual work principle, the plastic work
donebyshearingof thesoil wasequatedtothevirtual
change of gravitational potential energy of the soil.
A factor can then be found so that a correspond-
ing mobilized shear strain can be read off fromthe
chosenstress-straincurve. Theincremental displace-
mentcanthenbecalculatedbythecorrelationbetween
theaverageshearstrainincrementandtheincremental
wall displacement.
Thisapproachofferedastraightforwardwaytoesti-
matethebulgingdisplacement of theretainingwall.
However, theapproachrequiresrefinementinorder to
includesomeadditional features that may besignifi-
cant indeepexcavations. Firstly, theapproachdidnot
consider theelasticstrainenergystoredinthesupport
system. Secondly, itiscommontofindanon-uniform
soil stratum with undrained shear strength varying
irregularly withdepth. Furthermore, thegeometry of
thedeformationmechanismchangesastheconstruc-
tion proceeds, so the representation of mobilization
of shear strength through the whole depth, using a
singlemobilizationratio, is only aroughapproxima-
tion. Inrealitytherewill bedifferencesinmobilization
of shear strength at different depths for calculating
incremental soil displacement. Lastly, the original
mechanismof Osman and Bolton (2006) shown in
Figure 3(a) only applied to wide excavations; nar-
row excavations called for the development of the
alternativemechanismof Figure3(b).
Inviewof theseissues, anewfullyincremental cal-
culationmethodhasbeenintroduced, allowingfor the
storageof elasticstrainenergyinthewall andthesup-
port system, andrespectingthepossibleconstriction
of theplastic deformations dueto thenarrowness of
anexcavation.
5.1 Deformation pattern in different zones
FromFigure 3, the soil is assumed to flow parallel
to thewall at theretainedsideabovethelevel of the
lowestsupport(zoneABDC) andtheincremental dis-
placement at anydistancexfromthewall isgivenby
thecosinefunctionof Equation1, replacingybyx.
By taking the origin as the top of the wall, the
deformation pattern of retained soil ABDC is given
inrectangular coordinatesasfollows:
Hard stratum
(a) Incremental displacementfield for wide excavation
Hard stratum
(b) Incremental displacement field for narrow excavation
0 1 2 3 4 5 6 7 8 9 10
5
4
3
2
1
0
1
2
3
4
5
F I
E H
t
H
h
s
B
L
Excavation
depth
l
dw
max
dw
max
dw
max
dw
max
A B
C D
F
H
I
E
l
w
max
2 ave

t
H
h
s
L
l
dw
max
dw
max
dw
max
A B
C D
F
E
I
H
Excavation
depth
l
B
l
w
max
2
l
w
max
ave 2 . 2
ave
Figure3. Incremental displacement fields.

19
Infanzone, CDE, bytakingtheapexof thefanzone
astheorigin
For fanzoneEFHinverywideexcavationsasindi-
catedinFigure3(a), bytakingthejunctionof thewall
andthecurrent excavationlevel astheorigin:
For thetriangular zoneFHI in very wideexcava-
tions, againtakingthejunctionof theexcavationand
thewall astheorigin:
For narrowexcavationsasshowninFigure3(b), a
rectangular zone EFHI of 2D shearing is now pro-
posed. The origin is taken as the mid-point of FE,
mid-wavelengthintheexcavation, at thewall.
In order to get moreaccuratesolutions, it is sup-
posed that the soil stratumis divided into n layers
of uniformthickness t (Figure4). Theaverageshear
straind(m,n) is calculatedfor n layers inm excava-
tionstages. Theincremental engineeringshear strain
ineachlayer iscalculatedasfollows:
In order to get a better idea of the deformation
mechanism, the relationship between the maximum
incremental wall displacement andtheaverageshear
strainmobilizedineachzoneof deformationshould
l
A B
C D
F
E
H
I
s
1
s
1
s
1
s
1
DSS
DSS
PSA
PSP
Layer 1
Layer 2
Layer 3
Layer 4
Layer n
Layer (n-1)
Figure4. Mobilizableshear strengthprofileof anexcava-
tionstageinanlayeredsoil.
beobtained. Ontheactivesideof theexcavation, the
spatial scaleis fixed by thewavelength of deforma-
tion l, and all strain components areproportional to
dw
max
/l. Theaverageengineering shear strain incre-
ment
mob
mobilized in the deformed soil can be
calculatedfromthespatial averageof theshear strain
increments in the whole volume of the deformation
zone.Forawideexcavationi.e.Figure3(a),theaverage
shear strainis equal to 2dw
max
/l. For anarrowexca-
vation, theaverageshear strain (
ave
) of activezone
ABCDandfanzoneCDEis2dw
max
/l and2.23dw
max
/l,
respectively, while
ave
in the passive zone EFHI
dependsbothonthewavelengthl of thedeformation
and the width B of the excavation. The relationship
betweenthenormalizedaverageshear straininEFHI
andtheexcavationgeometryisshowninFigure5.The
correlationsareasfollows:
Apartfromthefirstexcavationstage,all subsequent
deformation mechanisms must partially overlay the
previous ones (Figure6). Dueto thenon-linearity of
soil itisimportanttocalculatetheaccumulatedmobi-
lized shear strain in each particular layer of soil in
order tocorrectlydeducethemobilizedshear strength
of that layer. This is doneby anareaaveragemethod
describedasfollows:
20
Figure 5. Correlation between normalized average shear
strainandexcavationgeometryfor anarrowexcavation.
A B
C D
F
E
H
I
C" D"
F"
E"
H"
I"
n th Layer
Deformation
mechanismof
Excavation stage
m-1
Deformation
mechanismof
Excavation stage
m
Enlarged strip of the nth layer
A(m-1,n)
A(m,n)
Figure6. Overlappingof deformationfield.
where(m, n) isthetotal shear strainof thenthlayer
in the mth excavation stage, and A(m, n) is the area
of deformationinthenthlayer inthemthexcavation
stage.
Withthehelpof somesuitablestress-strainrelation
for the soil (discussed in the following section), the
mobilizedstrengthratio(m,n) at excavationstagem
for soil layer ncanbefound(Figure7).
Figure7. Typical stressstrainrelationshipof soft clay.
5.2 Shear strength mobilized in mechanism
In soft clay, the undrained shear strength generally
varieswithdepth, andwithorientationof shear direc-
tion. The strength matrix c
mob
(m,n) mobilized for
excavationstagem for layer n canbeexpressedusing
a matrix (m,n) on the appropriate undrained shear
strengthprofile. Regardingorientation, anisotropy of
soft soil canbeasignificant factor for excavationsta-
bility. For example, Clough&Hansen(1981) showan
empirical factor basedontheobservationthat triaxial
extension tests give roughly one half the undrained
shear strength of triaxial compression, with simple
shear roughly half way between. Figure4 shows the
orientation of the major principal stress direction
within thevarious zones of shearing in theassumed
plasticmechanismforwideexcavations, andindicates
withacodethesoil test configurationthat wouldcor-
rectlyrepresent theundrainedshear strengthof at the
specific orientation. For locations marked DSS the
assume directions of shearing are either vertical or
horizontal, so the ideal test on a vertical core is a
direct simple shear test. In areas marked PSA and
PSP, shearing takes place at 45 degrees to the hori-
zontal and thesezones arebest represented by plain
strainactiveandpassivetests, respectively. Sincethe
undrainedshearstrengthof thedirectsimplesheartest
isroughlytheaverageof thatof PSA andPSP, therel-
ativeinfluenceof thePSA andPSP zones is roughly
neutral withrespecttodirectsimpleshear.Asaresult,
thedesignmethodfor bracedexcavationcanbest be
basedontheundrainedshearstrengthof adirectsimple
shear test. A similar decisionwasmadebyORourke
(1993).
Theequilibriumof theunbalanced weight of soil
insidethemechanismis achievedby mobilizationof
shearstrength.Foreachexcavationstage,mobilization
21
of shear strength of each layer is considered by the
following:
where c
u,mob
(m,n) is the mobilized undrained shear
strengthfor layer n inexcavationstagem; c
u
(n) isthe
undrainedshear strengthfor layer n; and(m,n) isthe
mobilized strength ratio for excavation stage m and
soil layer n.
5.3 Incremental energy balance
By conservationof energy, thetotal loss of potential
energy of thesoil (LP) must balancethetotal dissi-
patedenergy duetoplastic shearingof thesoil (LD)
andthetotal storedelasticstrainenergyinbendingthe
wall (LU).
Thepotential energy loss ontheactivesideof the
wall andthepotential energygainof soil onthepassive
sidecanbeestimatedeasily. Thenet changeof gravi-
tational potential energy (LP) isgivenby thesumof
thepotential energychangesineachlayer:
wheredw
y
(m, i) isthevertical componentof displace-
ment of soil intheithlayer for themthconstruction;
sat
(m, i) isthesaturatedunit weight of soil intheith
layer for themthconstruction.
Sincetherearenodisplacementdiscontinuities, the
total plasticwork donebyshearingof soil isgivenby
summingtheinternal dissipationineachlayer:
wherec
u
(m,i) is theundrained shear strength of soil
in the ith layer for the mth construction; d(m,i) is
theshear strain increment of soil in theith layer for
themthconstruction;andthecorrespondingmobilized
strengthratioisgivenby:
Thetotal elasticstrainenergystoredinthewall,LU,
canbeevaluatedbyrepeatedlyupdatingthedeflected
shapeof thewall. It isnecessarytodothissinceU is
aquadraticfunctionof displacement:
whereE andI aretheelasticmodulusandthesecond
moment of areaper unit length of wall, and s is the
lengthof thewall inbending. L canbethelengthof
wall s belowthelowest prop.
Byassumingthecosinewaveformequation(Equa-
tion1), thestrainenergy termcanbeshowntobeas
follows:
wherel isthewavelengthof deformation, dw
max
isthe
maximumdeflection of the wall in each excavation
increment.
5.4 Calculation procedure
Thefollowingcalculationprocedureisprogrammedin
Matlab2006b.
1 At eachstageof excavation, amaximumdeforma-
tionw
max
,whichisboundedbyanupperandalower
bound, isassumed. Thesoil stratumisdividedinto
n layers. Theareasonboththeactivesideandthe
passivesideineachlayer arecalculated.
2 For eachlayer, withthehelpof thenumerical inte-
gration procedurein Matlab, themobilized shear
strain and the change in PE on both active and
passivesidesindifferentzonesiscalculated.(Equa-
tion18)Thetotal mobilizedshear strainisupdated
accordingEquation15.
3 Withtheuseof asuitablestress-straincurve(Fig-
ure 7), the mobilizable strength ratio can be
found.
4 Total changeinPEandtotal energydissipationand
elasticbendingenergyinthewall canbecalculated
byEquations18, 19& 21, respectively.
5 By considering the conservation of energy of a
structure in statical equilibrium, the sumof total
energy dissipationandelastic strainenergy inthe
wall balances thetotal changeinPE. To facilitate
solvingthesolution, anerror termisintroducedas
follows:
6 Whentheerror issmaller thanaspecifiedconver-
gencelimit, theassumeddeformationis accepted
asthesolution; otherwise, themethodof bisection
isemployedtoassumeanothermaximumdisplace-
ment andtheerror termis calculatedagainusing
steps1to5.
7 Then, the incremental wall movement profile is
plottedusingthecosinefunctionof equation
8 Thecumulativedisplacementprofileisobtainedby
accumulatingtheincremental movement profiles.
22
6 VALIDATIONBY NUMERICAL ANALYSIS
Thefiniteelement method can provideaframework
for performingnumerical simulations to validatethe
extendedMSDmethodinevaluatingtheperformance
of bracedexcavations. However, finiteelement anal-
ysisof retainingwallsispotentiallyproblematic. One
themost difficult problems is theconstitutivemodel
usedfor thesoil. Thestress-strainrelationshipcanbe
verycomplicatedwhenconsideringstresshistoryand
anisotropyof soil (Whittle, 1993).
The validation of the extended MSD method is
examinedbycomparingitspredictionswithresultsof
comprehensiveFE analyses of aplanestrain braced
excavation in Boston Blue Clay carried out by J en
(1998). In these analyses, the MIT-E3 constitutive
model isused(Whittle, 1987). Themodel isbasedon
ModifiedCamclay(RoscoeandBurland1968). How-
ever, several modificationshadbeenmadetoimprove
thebasiccritical stateframework. Themodel cansim-
ulatesmall strainnon-linearity, soil anisotropyandthe
hysteretic behaviour associatedwith reversal of load
paths. Whittle(1993) alsodemonstratedtheabilityof
themodel toaccuratelyrepresentthebehaviour of dif-
ferent clays when subjected to a variety of loading
paths.
J en(1998) extendedtheuseof theMIT-E3model
for analyzing cases of deep excavation in a great
varietyof situations. Sheconsideredtheeffectof exca-
vationgeometrysuchaswall length, excavationwidth
anddepthof bedrock, theeffectof soil profilesuchas
c
u
/OCR ratioandlayeredsoil, andtheeffect of struc-
tural stiffnesssuchaswall stiffnessandstrutstiffness.
Thisprovidesavaluabledatabasefor validationof the
extendedMSDmethod.
6.1 An example of MSD calculation
ThefollowingexampleshowstheextendedMSDcal-
culationof wall deflections for a40mwall retaining
17.5mdeepand40mwideexcavation(Figure8).The
constructionsequencecomprisesthefollowingsteps:
1 The soil is excavated initially to an unsupported
depth(h) of 2.5m.
2 Thefirst support isinstalledat thegroundsurface.
3 Thesecondlevel of props is installedat avertical
spacingof 2.5m, and2.5mof soil isexcavated.
Theundrainedshearstrengthof thesoil isexpressed
bytherelationshipsuggestedbyHashashandWhittle
(1996) for BostonBlueClay(BBC) asfollows:
Thecantilevermodeof deformationandthebulging
movementsarecalculatedseparatelyusingthemech-
anismof Osman & Bolton (2006) and the extended
Retaining wall
(EI=1440,77 and 19MNm
2
/m)
OCR=1
BBC
properties
C=37.5, 50 and 100m
L=20, 25, 30, 35 and 40m
B/2=15,20,25 & 30m
C
L
h=2.5m
s=2.5m
Figure8. Scopeof parametricstudytoexamineexcavation
widtheffect.
Figure 9. Stress-strain response for Ko consolidated
undrained DSS tests on Boston blue clay (After Whittle,
1993).
MSDmethodasdescribedabove.Thetotal wall move-
mentsarethenobtainedbyaddingthebulgingmove-
ments to the cantilever movements to the cantilever
movement accordingtoCloughet al. (1989).
6.1.1 Cantilever movement
By solving for horizontal force equilibrium and
moment equilibriumabout the top of the wall, the
mobilizedshear stress(c
mob
) isfoundtobe11.43kPa.
Themobilizedstrengthratio is0.2886.Withthehelp
of directsimpleshearstress-straindataforBostonblue
claybyWhittle(1993) (Figure9), themobilizedstrain
is read off for an appropriatepreconsolidation pres-
sure
p
andanappropriateOCR. Themobilizedshear
23
Figure10. Wall deflectionsfromMSDwithdifferentexca-
vationdepths.
strain(
mob
) isfoundtobe0.2%. By consideringthe
geometrical relationship, thewall rotationisfoundto
be 0.1%. The displacement at the top of the wall is
foundtobe39mm.
6.1.2 Bulging movement
The first support is installed at the top of the wall.
The length of the wall below the support is 40m.
By adopting the iterative calculation procedure, and
usingthedeformationmechanismforanarrowexcava-
tion, thebulgingmovementateachstageof excavation
canbeobtained. Then, theincremental bulgingmove-
ment profileineachstageisplottedusingthecosine
function, using the maximumincremental displace-
ment in each stage together with the corresponding
wavelengths. The total wall movement is obtained
by accumulating cantilever movement and the total
bulgingmovement. Figure10shows thefinal defor-
mationprofileof theaccumulatedwall movement of
anexcavationwithawidthof 40m.Themaximumwall
deflectionatanexcavationdepthof 17.5mis115mm.
The position of the maximumwall displacement is
locatedat 0.75L, whereL isthelengthof thewall.
6.2 Effect of excavation width
The effect of excavation width on predicted ground
movements is thefocus of this section. Underground
transportation systems may have excavation widths
ranging from 25m (a subway station) to 60m (an
undergroundhighway). Themost widely useddesign
charts generally incorporatetheeffect of excavation
width in estimation of factor of safety against base
heave(BjerrumandEide, 1956) or asamultiplication
factor in estimating themaximumsettlement (Mana
andClough, 1981).
Thescopeof theexcavationanalysesareshownin
Figure8. Intheanalyses, theexcavationwas carried
out inundrainedconditions inadeposit of normally
consolidated Boston Blue Clay with depth C taken
to be 100m. A concrete diaphragm wall of depth
L =40m, and thickness 0.9m, supported by rigid
props spaced at h=2.5m, was used for supporting
thesimulatedexcavation. Theexcavationwidthvaries
from20mto 60m. The wavelength of deformation
is chosen according to the l =s rule, where was
takento be1.5ands is thelengthof wall belowthe
lowest prop. ComputedresultsbyJ en(1998) areused
forcomparison. Full detailsof theanalysisprocedures,
assumptionsandparametersaregiveninJ en(1998).In
thefollowingsection, onlyresultsof wall deformation
will betakenfor comparison.
Figure11(a) and(b) showthewall deflectionpro-
filewithdifferent excavationwidthsat anexcavation
depthof 17.5m, as calculatedby theextendedMSD
method and the MIT-E3 model. Figure 11(a) shows
that theexcavationwidthdoesnot haveany effect on
the deflected shape of the wall as calculated by the
extendedMSDmethod.Figure11(b),simularly,shows
alimitedeffectonthedeflectedshapeof thewall bythe
MIT-E3model. WhiletheMSD-predictedmaximum
wall deflectionincreasesbyafactorof 1.5asthewidth
isincreasedfrom30mto60m, theMIT-E3computed
maximumwall deformation increased by afactor of
1.6withthesameincreaseinexcavationwidth.
6.3 Effect of bending stiffness of the wall
In general, structural support to excavations is pro-
videdbyawall andbracingsystem. Soldier pilesand
lagging, sheet piles, soil mixandsoldier piles, drilled
piers(secantpiles),andreinforcedconcretediaphragm
wallsareexamplesof wall typesthathavebeenusedto
supportexcavations. Thevarioustypesof wall exhibit
asignificant rangeof bendingstiffnessandallowable
moment. Supportwallscomposedof soldier pilesand
sheet pilesaregenerallymoreflexibleandcapableof
sustaining smaller loads than the more rigid drilled
piersandreinforceddiaphragmwalls.
The preceding sections have all assumed a 0.9m
thick concrete diaphragmwall with elastic bending
stiffness EI =1440MNm
2
/m. Although it is possi-
ble to increase this bending stiffness by increasing
the wall thickness and reinforcement, or by using
T-panels(barettes), most of thewallsusedinpractice
havelower bendingstiffnesses. For example, thetypi-
cal bendingstiffnessof sheetpilewallsisintherange
of 50to80MNm
2
/m. Thissectionassessestheeffect
of wall bending stiffness on the excavation-induced
displacements.
24
Figure 11. Wall deflection profile of different excavation
widthsat H=17.5m.
Excavationinsoft clay withawidthof 40msup-
portedbyawall of length25mandof variousbending
stiffness(EI =1440, 70and20MNm
2
/m) arestudied.
Results generatedby theMSD methodandFEA are
compared. Figure12(a)and(b)presentsthedeflection
profilesof theexcavationspredictedbyextendedMSD
and theMIT-E3 model, respectively. As thebending
stiffnessof thewall decreases, thereisnopronounced
changeintheoverall shapeof thewall; themaximum
wall deflection increases and its location migrates
towardstheexcavatedgrade. AtH=12.5m, themax-
imumwall displacement is 47mmfor the concrete
diaphragmwall withthemaximumdeflectionlocated
at 7.5mbelow theexcavation level, whiletheresult
for the most flexible sheet pile wall shows 197mm
of maximumwall deflectionoccurringat5.5mbelow
theexcavationlevel. Inadditional to this, changes in
Figure12. Deflectionprofilesof wallswithvariousbending
stiffnesses.
wall stiffness also affect the transition froma sub-
gradebendingmodetoatoekicking-outmode.Asthe
wall stiffness decreases, theinfluenceof embedment
depth reduces, and hencethetendency for toekick-
out to occur is less. Again, a fairly good agreement
can beseen when comparing extended MSD results
and numerical results by theMIT-E3 model, though
kinks areusually foundat thewall toeinthenumer-
ical predictions, which implies localization of large
shear strainsdevelopedbeneaththewall toe.
6.4 Effect of wall length
Wall length is oneof thegeometrical factors affect-
ingthebehaviour of asupportedexcavation. Previous
analysesweredonebyOsmanandBolton(2006). The
studies showedthat thewall endconditionshouldbe
assumedtobefreefor short walls(L =12.5m) since
theclayisverysoftatthebaseandtheembeddedlength
isnot longenoughtorestrainthemovement at thetip
of thewall (kick-out modeof deformation). For long
walls (L =40m), the embedded depth was assumed
25
Figure13. Wall deflection profileof excavation with dif-
ferent support wall lengths, byExtendedMSDmethod.
to besufficient to restrain themovement at thewall
base (bulging model of deformation). However, the
effect of structural stiffnesswasnot consideredinthe
old MSD method, though similar observations were
madebyHashashandWhittle(1996)intheirnumerical
analyses.
Inthissection, theeffectof wall lengthwill becon-
sidered. Excavationswithwidthsof 40msupportedby
a 0.9mthick concrete diaphragmwall with varying
length (L =20, 25, 30 and 40m) are studied. Fig-
ure13 shows thewall displacement profiles against
depth with different lengths of wall. For H7.5m,
thedeflectedwall shapesarevirtuallyidentical for all
four wall cases of wall length. This agrees with the
conclusionmadeby Hashash(1992) that wall length
hadaminimal effectonpre-failuredeformations.AsH
increasesto10m, thetoeof the20mlongwall begins
tokick out withmaximumincremental deformations
occurring at the toe of the wall. The movements of
the25, 30 and 40mlong walls arequitesimilar. At
H=12.5m, the toe of the 20mand 25mlong wall
kickout, whilethetwolongerwalls(L =30and40m)
continue to deformin a bulging mode. The differ-
enceindeformationmodeshapedemonstratesthatthe
wall lengthhas asignificant influenceonthefailure
mechanismfor abracedexcavation.
Figure15showsasimilar set of analysesby using
theMIT-E3model.Similarobservationsaboutthewall
shapecanbemade.
Figure 14 summarizes the variation of the nor-
malizedexcavation-induceddeflection(w
max
,H) with
thewidth to length ratio (B/L) for different bending
stiffnessesof thesupport wall, for H=17.5m.
Figure 14. Variation of maximum wall deflection with
widthtolengthratioof wall.
Figure15. Wall deflection profileof excavation with dif-
ferent support wall lengths, by MIT-E3 method (After J en
(1998)).
For aflexiblewall (EI =12.3MNm
2
/m), thenor-
malized maximumwall deflection increases linearly
as theB/L ratio increases from1 to 2. Thegradient
changes andw
max
/H increases in agentlefashion as
the B/L ratio increases from2 to 2.8. For a rigidly
supportedwall, theincreaseinw
max
/Hratioislesssig-
nificant astheB/L ratioincreases. Inother words, the
maximumwall deflectionislesssensitivetoachange
of B/L ratiofor arigidwall.
26
Figure 16. Wall deflection profiles of excavation with
different depthstothefirmstratum.
6.5 Effect of the depth of bearing stratum
Thedepthto bedrock, C, is animportant component
of the excavation geometry. The preceding analyses
haveassumedadeepclay layer withbedrock located
at C=100mwhichrepresentsapractical upper limit
on C. In practice, however, the clay layer is usually
less than100mdeep. Thefollowingresults focus on
the discussion of the geometrical parameter C. The
analysis involves planestrainexcavationinnormally
consolidated Boston blue clay supported by a 0.9m
thickconcretediaphragmwall withrigidstrutsupports
spacedat 2.5m.
The wall deflection profiles for excavations pre-
dictedby bothMSD andMIT-E3withtwo depths of
theclaystratum(C=35mand50m) arecomparedin
Figure16(a) and(b).
Ingeneral, thedepthof thefirmstratumwouldonly
affect wall deformations below theexcavated grade,
hencethelargest effectscanbeseenat thetoeof the
wall. For situationswherethewavelengthof deforma-
tionisrestrictedlessby excavationwidththanby the
depth of thefirmstratum(B>C), themagnitudeof
maximumwall deflectionincreaseswiththedepthof
thefirmstratum(C). TheMSD methodpredicts that
thekick-out displacement of thewall toeis limited
by the restriction of developing a large deformation
mechanism. As a result, the maximumwall deflec-
tionis also limited. Theincreaseinincremental wall
deflectiondecreasesinlaterstagesof excavationwhen
H increases from12.5mto 17.5mdueto thereduc-
tionof wavelengthof deformation. Ontheother hand,
whenthedepthof thefirmstratumismuchlargerthan
thewidthof theexcavation(B-C), thedepthof the
bed rock has a minimal effect on the magnitude of
wall deflection. Results by FEA by J en (1998) (Fig-
ure16(b)) alsoshowedthesameobservation. Despite
theshortcomingof MSDnot beingabletomodel the
correctshapeof wall, themaximumwall deflectionis
predictedreasonablywell. Thenet differenceinmax-
imumwall displacement between MSD and the full
FEA isgenerallylessthan20%.
7 CONCLUDINGREMARKS
AnextendedMSD methodis introducedto calculate
the maximumwall displacement profile of a multi-
proppedwall retaininganexcavationinsoft clay. As
withtheearlierMSDapproach, eachincrementof wall
bulging generated by excavation of soil beneath the
current lowest level of support is approximated by a
cosinefunction. Thesoil isdividedintolayersineach
of whichtheaverageshear strainincrementsarecom-
pounded so that themobilized strength ratio in each
layercanbetrackedseparatelyasexcavationproceeds,
usingstress-straindatafromarepresentativeelement
test matchedtothesoil propertiesat mid-depthof the
wall. Theincremental lossinpotential energyassoci-
atedwiththeformationof asettlement trough, dueto
wall deformationandbaseheave, canbeexpressedas
afunction of thoseground movements at any stage.
By conservationof energy, this must always balance
theincremental dissipationof energythroughshearing
andtheincremental storageof elasticenergyinbend-
ingthewall. Byaniterativeprocedure, thedeveloping
profileof wall displacementscanbefound.
A reasonableagreement is found between predic-
tionsmadeusingthisextendedMSDmethodandthe
FEA resultsof J en(1998) whocreatedfull numerical
solutions usingtheMIT-E3soil model. Inparticular,
theeffectsof excavationwidth, wall bendingstiffness,
wall length, andthedepthof theclaystratum, wereall
quitecloselyreproduced.
It is important to drawtheright lessons fromthis.
Theexcellent work at MIT over many years, on soil
element testing, soil constitutive models, and Finite
27
ElementAnalysis, haveprovideduswiththemeansto
calibrateaverysimpleMSDpredictionmethod. This
was basedonanundrainedstrengthprofile, asingle
stress-strain test, and a plastic deformation mecha-
nism. Wereit not for themultiplelevel of props the
calculationof grounddisplacementscouldbecarried
out inhardly moretimethanis currently requiredto
calculateastabilitynumber or factor of safety. Allow-
ingfortheneedtorepresentvariouslevelsof props,the
calculationsthencall for aMatlabscript or aspread-
sheet, andthewholeprocessmight takehalf aday to
complete.
Anengineer canthereforeanticipatethatimportant
questionswill becapableof approximatebut reason-
ablyrobust answersinasensibleindustrial timescale.
For example:
Will apropspacingof 3mbesufficient for awall
of limitedstiffnessandstrength?
Shouldthebaseof thewall befixedbyjet-grouting
prior toexcavation?
Will aparticular construction sequencecausethe
soil tostrainsomuchthat it indulgesinpost-peak
softening?
Is it feasibleto propthewall at sufficiently close
spacingstorestrictstrainsintheretainedgroundto
valuesthatwill preventdamagetoburiedservices?
Thismayleadengineerstotakesoil stiffnessmore
seriously, andtorequest accuratestress-straindata. If
so, inadecadeperhaps, our Codes of Practicemight
beupdatedtonotethatMSDfordeepexcavationspro-
videsapractical wayof checkingfor theavoidanceof
serviceabilitylimit states.
ACKNOWLEDGEMENT
Theauthorswouldliketoacknowledgetheearmarked
research grant #618006 provided by the Research
Grants Council of the HKSAR Government, and
alsothePlatformGrant (GR/T18660/01) awardedby
theUK EngineeringandPhysical Sciences Research
Council.
REFERENCES
Bjerrum, L. &Eide, O. 1956. Stabilityof struttedexcavations
inclay. Geotechnique 6: 115128.
Bolton, M.D. & Powrie, W., 1988. Behaviour of diaphragm
walls retaining walls prior to collapse. Geotechnique
37(3): 335353.
Bolton, M.D., PowrieW. & Symons, I.F. 1989. Thedesign
of stiff in-situwallsretainingoverconsolidatedclay part.
Ground engineering 22(8): 4448.
Bolton,M.D.,Powrie,W.&Symons,I.F.1990a.Thedesignof
stiff in-situwallsretainingover-consolidatedclay part 1,
short termbehaviour. Ground Engineering 23(1): 3439.
Bolton,M.D.,Powrie,W.&Symons,I.F.1990b.Thedesignof
stiff in-situwallsretainingover-consolidatedclay, partII,
longtermbehaviour. Ground engineering 23(2): 2228.
Clough, G.W. & Hansen, L.A. 1981. Clay anisotropy and
braced wall behavior. ASCE Journal of Geotechnical
Engineering 107(7): 893913.
Clough, G.W., Smith, E.W. &Sweeney, B.P. 1989. Movement
control of excavationsupport systemby iterativedesign.
Foundation Engineering Current Principles and Practices
Vol. 2ASCE, NewYork, NY: 869882.
Eide, O., Aas, G. & J osang, T. 1972. Special applicationof
cast-inplace walls for tunnels in soft clay. Proceedings
of the 5th European Conference on Soil Mechanics and
Foundation Engineering, Madrid, Spain, 1: 485498.
Hashash, Y.M.A. & Whittle, A.J. 1996. Ground movement
predictionfordeepexcavationsinsoftclay.ASCE Journal
of Geotechnical Engineering 122(6): 474486.
J en, L.C. 1998.The design and performance of deep excava-
tions in clay. PhDthesis, Dept. of Civil andEnvironmental
Engineering, MIT, Cambridge, Mass.
Mana, A.I. & Clough, G.W. 1981. Predictionof movements
for braced cut in clay. J. Geo. Engrg. ASCE 107(GT8):
759777.
Milligan, G.W.E. & Bransby, P.L. 1976. Combined active
andpassiverotational failureof aretainingwall insand.
Geotechnique 26(3): 473494.
ORourke, T.D. 1993. Base stability and ground move-
ment prediction for excavations in soft clay. Retaining
Structures, ThomasTelford, London: 131139.
Osman, A.S. &Bolton, M.D. 2004A newdesignmethodfor
retaining walls in clay. Canadian Geotechnical Journal
41(3): 451466.
Osman, A.S. & Bolton, M.D. 2005Simpleplasticity-based
predictionof theundrainedsettlementof shallowcircular
foundationsonclay. Geotechnique 55(6): 435447.
Osman,A.S. &Bolton, M.D. 2006Groundmovementpredic-
tionsforbracedexcavationsinundrainedclay.ASCEJour-
nal of Geotechnical and Geo-environmental Engineering
132(4): 465477.
Osman, A.S., Bolton, M.D. & Mair, R.J. 2006 Predicting
2Dgroundmovementsaroundtunnelsinundrainedclay.
Skempton, A.W. 1951. Thebearingcapacity of clays. Proc.,
Building Research Congress, London: 180189.
Sloan, S.W. 1988. Lower bound limit analysis using finite
elementsandlinear programming. International Journal
for Numerical and Analytical Methods in Geomechanics
12(1): 6177.
Sloan, S.W. & Kleeman, P.W. 1995. Upper bound limit
analysis using discontinuous velocity fields. Computer
Methods in Applied Mechanics and Engineering 127:
293314.
Terzaghi, K.1943. Theoretical soil mechanics, J ohnWiley&
Sons, Inc., NewYork, N.Y.
Ukritchon, B., Whittle, A.J. & Sloan, S.W. 2003. Undrained
stability of braced excavations in clay. ASCE Journal of
Geotechnical and Geoenvironmental Engineering 129(8):
738755.
Whittle,A.J. 1987.A constitutive model for overconsolidated
clays with application to the cyclic loading of friction
piles. PhD thesis, Dept. of Civil and Environmental
Engineering, MIT, Cambridge, Mass.
Whittle, A.J. 1993. Evaluation of a constitutive model for
overconsolidatedclays. Geotechnique 43(2): 289313.
28
Overviewof Shanghai YangtzeRiverTunnel Project
R. Huang
Commanding Post of Shanghai Tunnel & Bridge Construction, Shanghai, P.R. China
ABSTRACT: Inthepaper, anintroductionof theconstructionbackgroundandscaleof Shanghai YangtzeRiver
Tunnel andBridgeProject andnatural conditionsof Shanghai YangtzeRiverTunnel constructionaregiven. The
overall designconcept andsomecritical technical solutionssuchassegment structureof largediameter bored
tunnel, water proofingof segment under highdepthandwater pressure, longtunnel ventilationsystemandfire
fightingsystemaredescribed. Characteristicsof twomixedTBM withadiameter of 15,430mmaredescribed.
Theoverall constructionmethodsof tunnel, andcritical technical solutionsandriskprovisionmeasuresfor large
andlongriver-crossingtunnel suchasthefrontsurfacestabilityforboredtunnel construction, floatingresistance
of largediameter tunnel, longdistanceconstructionsurvey, synchronousconstructionof internal structure, and
crosspassageconstructionof fresh/saltyalternatinggeological/environmental conditionarediscussed.
1 INTRODUCTION
Shanghai YangtzeRiver Tunnel andBridgeproject is
located at the South Channel waterway and North
Channel waterway of Yangtze River mouth in the
northeast of Shanghai, which is a significant part
of national expressway, as shown in Figure 1. It is
anextremely major transport infrastructureproject at
seashore area in China at Yangtze River mouth and
alsothelargesttunnel andbridgecombinationproject
worldwide. Thecompletionof theproject will further
promotethedevelopmentspaceforShanghai, improve
the structure and layout of Shanghai traffic system,
develop resources on Chongming Island, acceler-
ate economic development in the north of J iangsu
Province, increasetheeconomy capacity of Pudong,
accelerate the economy integrity of Yangtze River
Delta, boomthe economic development of Yangtze
River areaand even thewholecountry and upgrade
thecomprehensivecompetenceof Shanghai inChina
andevenintheglobal economy.
ShanghaiYangtzeRiverTunnel andBridge(Chong-
ming Crossing) alignment solution is the planned
westernsolutionwhichisimplementedfirstlybasedon
theShanghai overall urbanplanning, andcomparison
betweeneast andwest alignment andincombination
of variousaspects. Thewesternalignment startsfrom
Wuhaogou in Pudong, crossingYangtzeRiver South
Channel waterwaytoChangxingIslandandspanning
Yangtze River North Channel waterway to east of
ChongmingIsland.
YangtzeRiver beginstobedividedinto3levelsof
branchesandhave4mouthsflowingintothesea: The
Figure1. Sitelocationof ChongmingCrossing.
South Channel waterway is mixed river trench. The
intermediate slow flow area forms Ruifeng shoal
whichisrelatively stablefor alongtime. Thenatural
water depthmakesit asthemainnavigationchannel.
However, the North Channel waterway is located in
the middle part of river, which is influenced by the
south part and branch transition into North Channel
waterway. So the trench varies alternatively and the
river mapisnot asstableasSouthChannel waterway.
Therefore, after iterativediscussionbyseveral parties,
finally thesolution of SouthernTunnel & Northern
Bridge isselected. Thetotal project is25.5kmlong,
among which 8.95kmis tunnel with adesign speed
of 80km/hand9.97kmisbridgeand6.58kmisland
connectionwithadesignspeedof 100km/h, asshown
29
Figure2. Diagramof Shanghai YangtzeRiverTunnel andBridge.
in Figure 2. The total roadway is planned as dual
6lanes.
2 CONSTRUCTIONBACKGROUNDAND
PLANNING
The planning study of Shanghai Yangtze River Tun-
nel and Bridge Project (Chongming Crossing) was
incepted from90s of last century. The preliminary
preparatoryworkhaslasted11years. InMay1993, the
National ScientificCommitteeheldtheYangtzeRiver
mouthcrossingsignificanttechnical-economical chal-
lenges earlystageworkmeeting.Afteroneyearspe-
cial investigation, thePreliminarystudyreportof sig-
nificanttechnical challengesof YangtzeRiver Cross-
ing wasprepared.Thepre-feasibilitystudyreportwas
prepared in March 1999. InAugust 2001, the inter-
national concept competition was developed and the
Southern Tunnel & Northern Bridge solution was
defined. TheNational PlanningCommitteeapproved
theprojectproposal inDecember2002.Thefeasibility
study report was approvedby theNational Develop-
ment andReformCommitteeinNovember 2004. The
preliminary design was approved by theMinistry of
CommunicationinJ uly 2005andtotal investment of
12.616billionRMB wasapprovedfor theproject.
For the project construction investment, 5 billion
wasfundedbyShanghai ChengtouCorporation(60%)
andShanghai RoadConstructionCooperation(40%),
and7.6billionwasfinancedfromBankConsortium.
Basedonthecharacteristics of thenational major
project, Commanding Post of Shanghai Tunnel &
BridgeConstructionwasestablishedwithapproval of
Shanghai Municipal Committee. Thepost is directed
bythevicemajor andcomposedof staff fromPudong
NewArea, ChongmingCountyandother committees
andbureaus. Themainresponsibilityistomakedeci-
siononsignificantproblemsandcoordinateimportant
items. Inorder toimprovethedepthof dailymanage-
ment, officewas set upunder thecommandingpost,
workingtogether withestablishedShanghai Yangtze
River Tunnel and BridgeConstruction Development
Co., Ltd. whichismainlyinchargeof theimplemen-
tation of theproject and daily work of commanding
post and performs the investment management on
behalf of theclient. Thespecific work is responsible
for thefinancing, investment, construction, operation
and transfer of the project. To detail the technical
assurancemeasures, theclients sets up thetechnical
consultant teamwhich provides theoretical support,
technical assistanceandconsultancyserviceforsignif-
icant technical challengesduringtheimplementation.
Meanwhile, theteamisinvolvedintheinvestigationof
significanttechnical solutions, reviewof construction
methodstatementandtreatmentof technical problems
to ensure the high quality and safety. International
well-knownconsultancy companies areentrustedfor
thepurposeof applicationof state-of-art philosophy,
mostsuccessful experience, optimal conceptandmost
maturemanagementtomaketheYangtzeRiverTunnel
andBridgeProject asCenturyEliteProject.
The project finally initiated on 28th, December
2004andplannedto beopento traffic inJ uly 2010.
Themaincivil structureof thebridgeisplannedtobe
closedinJ une2008, andtunnel inApril 2009.
3 NATURAL CONDITIONSOF TUNNEL
PROJ ECT
3.1 Environmental conditions
Shanghai YangtzeTunnel Project starts fromWuhao-
gou of Waigaoqiao in Pudong NewArea, connected
withShanghai mainfast roads suchas MiddleRing,
OuterRingandSuburbRingthroughWuzhouAveneu,
crossing southern water area and lands on Changx-
ingIsland400mwestof XinkaiheHarbour, connected
withChangxingIslandroadnetthroughPanyuanInter-
change.Themainbuildingonlandisthefloodpreven-
tionwall onPudongsideandChangxingIsland.Others
arefarmfields.Theriver-crossingsectionismainlythe
30
Figure3. Longitudinal profileof tunnel.
southernwater wayfor navigationwhichisanimpor-
tantpassagefor connectingYangtzeWaterswithother
seashoreareainChinaandoceansworldwide.
Therearetwoseacablesarrangedalongthebored
tunnel axis with a depth of 3mbelow natural river
bed. One cable is basically located at the west side
of thetunnel andgoes into theriver near Wuhaogou
on Pudong side, which is about 1,500maway from
thetunnel. It becomes closer to thetunnel gradually
tothenorthandcrossesthetunnel toitseast at 240m
fromChangxingIslandandlandsonChangxingIsland
at 350mwest of XinkaiheHarbour. Theother cable
goes into theriver near Wuhaogou, 1,300away from
thetunnel. Thenit turnstoNE first andN at 2,600m
wayfromPudongLandConnections, almostidentical
with the tunnel alignment. And it changes fromthe
west of tunnel to east of tunnel gradually and lands
on Changxing Island about 300mwest of Xinkaihe
Harbor.
Furthermore, two sunkenboats closeto Chainage
XK2+350 and XK1+500 have been salved before
boredtunnel construction. Earthwasalsofilledback
atcorrespondinglocations; however, theremaybestill
someremains.
3.2 River regime and hydrological conditions
Atthemouthof YangtzeRiveritistideareawithinter-
mediatelevel. Outsideof mouthisregularhalf daytide
andinsideisirregular half dayshallowtideduetothe
changeof tidewave.Averagefloodtidetimeis5hand
averageebbtidetimeis7h, sototal timefor ebband
fluxis12h. Theaveragecurrencyflowis1.05m/sfor
flood tide during flood season and 1.12m/s for ebb
tide. Themaximal flowfor floodtideis1.98m/sand
2.35m/sfor ebbtide.
The underground water type in the shallow stra-
tumat tunnel siteis potential water, whichhas close
hydraulicrelationwithriverwater.Thepotential water
level is mainly influenced by theYangtzeRiver flux
andebb. Theaveragewater level for Waigaoqiaoand
ChangxingIslandis2.8mand2.4m, respectively.
In the stratum and at site area, the con-
finedwater is rich. At most area, theconfinedwater
is directly continuous. The confined water level is
between 4.15mand 6.76m. Furthermore, slight
confined water distributes in
2
, which has certain
hydraulicrelationswithconfinedwater in.
3.3 Geological conditions
Therelief of onshoreareaof theprojectisrivermouth,
sandmouth,sandislandwhichiswithinthemajorfour
relief units in Shanghai. Theground surfaceis even
withanormal elevationof 3.5m(WusongElevation).
Thewater areaisclassifiedasriver bedrelief.
Theprojectsitehasaseismicfortificationintensity
of 7, classified as IV site. The stratum
3
and
2
sandysilt distributingonPudonglandareaisslightly
liquefied.
Main geological layers (refers to Figure 3) TBM
crosses are:
1
grey muddy clay,
1
grey muddy
clay,
2
grey clayey silt withthinsilty clay,
3
silty
clay,
3
tlens,
11
grey clay silt,
12
grey sandy
silt, etc. Unfavorablegeological conditions areexpe-
riencedalongtheaxisof thetunnel, suchasliquefied
soil, quick sand, piping, shallowgas (methane), lens
andconfinedwater, etc.
4 TUNNEL DESIGNSOLUTION
4.1 Scale
Shanghai Yangtze River Tunnel is designed as dual
6lanesexpressway, andrail traffic provisionismade
belowtheroaddeck. Seismic fortificationlevel is 7.
31
Figure4. Crosssectionof boredtunnel.
Design service life is 100 years. The project con-
sists of landconnections of Pudongside(657.73m),
river-crossingtunnel (east tube7,471.654mandwest
tube7,469.363m) andlandconnections on Changx-
ingIsland(826.93m). Total lengthis8,955.26mand
investmentis6.3billionRMB.Theriver-crossingpart
istwin-tubeboredtunnel.
4.2 Tunnel alignment
Thelongitudinal profileof boredtunnel isinashapeof
W withalongitudinal slopeof 0.3%and0.87%. The
landconnectionshavealongitudinal profileof 2.9%.
The minimal curvature radius of horizontal plane is
4,000mandvertical profile12,000m.
4.3 Building design
4.3.1 Cross section of bored tunnel
Basedonstructural limitof trafficpassageandequip-
ment layout requirement, the internal diameter of
lining for bored tunnel is determinated as 13.7m
considering the fitted tolerance of lining at curved
section,constructiontolerance,differential settlement,
and combining the design and construction experi-
ence. Onthetopof tunnel, smokedischargeductsare
arrangedforfireaccidentwithanareaof 12.4m
2
.Each
tunnel hasthreelaneswithastructural clear widthof
12.75mandroadlaneclear height of 5.2m. Thecen-
tral partbelowroaddeckisfor rail trafficprovisionin
future. Ontheleft side, besidetheburiedtransformer
arrangement, it alsoservesasmainevacuationstairs.
The right side is cable channel, including provision
spacefor 220kV power cable, asshowninFigure4.
4.3.2 Cross-section of land connections
Working shaft is underground four-floor building:
1 is for ventilation pipe and pump plant for fire
fighting; 2 is for road lanewith cross over; 3 is
for rail traffic provisionandpower cablegallery and
4isfor wastewater pumpplant.
The cut-and-cover is designed with a rectangular
shape consisting of two tubes and one cable chan-
nel. 3lanes arearrangedineachtube. Thestructural
limitis13.25minwidthand5.5minheight, asshown
inFigure5. Upper areawithaheight of 0.6mis for
equipmentprovision.Theupper partof central gallery
is for cablechannel, middlepart for evacuation and
lowerpartforpipeditch.Ventilationshaftandbuilding
for equipments arearrangedabovethecut-and-cover
tunnel closetotheworkingshaft.
Theapproachconsists of light transitionzoneand
openramp.Thestructural limitof crosssectionisiden-
tical withthatof cut-and-covertunnel. Bothsideshave
aslopesectionwithaslopeof 1:3withgreenplanting
for protection. Thelighttransitionzoneisdesignedas
steel archstructure.
4.4 Structural design
4.4.1 Structural design of bored tunnel
The external diameter of bored tunnel lining is
15,000mm and internal diameter 13,700mm, as
shown in Figure6. Thering width is 2,000mmand
thicknessis650mm. Precastreinforcedconcretecom-
montaperedsegments areassembledwithstaggered
joint. ConcretestrengthclassisC60andseepageresis-
tance class is S12. The lining ring consists of 10
segments, i.e. 7 standard segments (B), 2 adjacent
segments (L), and 1 key segment (F). According to
the different depth, segments are classified as shal-
lowsegments, middle-deepsegments, deepsegments
andextremelydeepsegments. Skewboltsareusedto
connect segmentsinlongitudinal andcircumferential
direction.38M30longitudinal boltsareusedtocon-
necttherings. 2M39circumferential boltsareused
toconnectthesegments. Shearpinsareaddedbetween
32
Figure5. Cross-sectionof cut-and-cover.
Figure6. Liningstructure.
liningringsatshallowcoverarea, geological condition
variationareaandcrosspassagetoincreasetheshear
strengthbetweenringsat special locationandreduce
thestepbetweenrings.
4.4.2 Structural design of land connections
Theworking shaft and cut & cover tunnel sharethe
same wall. The thickness of diaphragmof working
shaft is 1,000mm, and the inner wall is 500mm,
1,200mm, respectively. For thecut-and-cover tunnel,
the thickness of diaphragm is 1,000mm, 800mm,
and600mmrespectivelydependingontheexcavation
depth. Theinner structurethicknessis600mm.
For theopen cut ramp, thebottomplatestructure
thicknessisaround5001,100mm. Under thebottom
plate, boredpilesarearrangedasup-liftingresistance
piletofulfil thestructural floatingresistancerequire-
ment. Theslopeuses in-situcast reinforcedconcrete
grid and fill earth and green planting in thegrid for
protection.
4.5 Structural water-proof and durability design
4.5.1 Requirement and standard
Fortheboredtunnel andworkingshaft, thewaterproof
standardof slightlyhigherthanlevel II isrequired. For
33
(a) (b)
Figure7. Segment joint water proofingsketch.
theentiretunnel, theaverageleakageshould beless
than0.05L/m
2
d. For eachrandom100m
2
, theleak-
ageshouldbelessthan0.1L/ m
2
d. Theinner surface
wet spots shouldnot bemorethan4of total inner
specificsurfacearea. Ineachrandom100m
2
, thewet
spots shouldnot bemorethan4locations. Themax-
imal areaof individual wet spot should not belarge
than0.15m
2
.
Thechloridediffusioncoefficient of concretelin-
ing structure of bored tunnel is not more than
1210
13
m
2
/s. Concreteseepageresistanceclassis
notlessthanS12.Furthermore,itisrequiredthatunder
1MPawaterpressurewhichisequivalentto2timesof
waterpressureforthetunnel withthelargestdepth, no
leakageisoccurredwhentheliningjoint opens7mm
andstaggers 10mm. Thesafety servicelifeof water
proof material is100years.
The seepage resistance class of onshore tunnel
structureisnot lessthanS10.
4.5.2 Water proofing design
Thesegmentjointwaterproof arrangementconsistsof
EPDMrubberstripwithsmall compressivepermanent
deformation, small stress relaxation and good aging
resistanceperformanceand hydrophilic rubber strip,
asshowninFigure7.
Thedeformationjoint at cut-and-cover tunnel uses
embedded water stop gasket, outer pastegasket and
inserted sealing glue forming enclosed system. The
top plate uses water proof paint as outer water
proof layer.
4.6 Tunnel operation system
4.6.1 Ventilation system
The road tunnel uses jet fan induced longitudinal
ventilationcombinedwithsmokeventilation.
Thelongitudinal ventilationareaintunnel is82m
2
.
J et fansaresuspendedabovethedecklaneandbelow
thesmokedischargeduct, supportinginducedventi-
lation in normal operation and congested condition.
78jet fanswithadiameter of 1,000mmarearranged
ineachtubefromPudongaccesstoChangxingIsland
access, every3asagroup.
VentilationshaftsarearrangedonPudongsideand
ChangxingIsland, respectively, housinglargeventila-
tion machineandspecial smokedischargeaxial fan.
Thefans areconnectedwithmaintunnel throughair
inlet and ventilation duct. During normal operation
and congested condition, the ventilation machine is
turnedontodischargethepollutedair inthetunnel. 6
largeaxial fanswithacapacityof 75m
3
/s 150m
3
/s
arehousedintheworkingshaftonChangxingInsland
andPudong, respectively.
For normal operation of lower rail traffic, piston
ventilationmodeisused.
4.6.2 Water supply and drainage system
The fire water, washing waste water, and structural
leakagearecollectedby thewastewater sumpat the
lowest point of river. Sumpis arrangedat upper and
lower level, respectively. The lower waste water is
drainedby therelay of upper sump. Theupper sump
is arranged on two sides of rail traffic area, housing
34
four pumpswhichareusedalternativelyunder normal
operation and turned on entirely during fire fight-
ing. For lower level, 4 sumps with a dimension of
1,0001,000550mmare arranged at the lowest
point of tunnel whereSGI segment isusedandabove
thesumpwater collectiontrenchwithalengthof 7m
andawidthof 1misarranged. Onewastewater pump
is placed in each pit which are used alternatively at
normal conditionandthreeareused, onespareduring
firefighting.
At each access of tunnel, one rain water sump is
arrangedto stopwater anddrainit out of thetunnel.
Therainamount isdesignedbasedonareturnperiod
of 30yearsfor rainstorm.
4.6.3 Power supply system
Theelectricityloadintunnel isclassifiedasthreelev-
els: level I is for ventilation fan, valve, water pump,
lightingandmonitoring& control systemanddirect
currentscreen, etc; level II isfortunnel inspectionand
repair,andventilationfanintransformerplant;level III
isfor air conditioningcoldwater machines.
OnPudongsideandChangxingIsland, two trans-
formers arearranged. Two independent 35kV power
circuits are introduced respectively and can be used
as spare power for the other through two connec-
tioncables. Eachrouteensurestheelectricity loadof
level I and II in the tunnel. For the dynamical and
lighting load far away fromtransformers, thepower
is supplied through 10kV power supply network in
thetunnel andembeddedtransformersunderneaththe
tunnel toensurethelongdistancepower supplyqual-
ity andreduceenergy losses. 6kV power is supplied
for theconcentratedventilationfan. Lightingelectric-
ityissuppliedbyindependent circuit inpower supply
system.
4.6.4 Lighting system
Light belt isusedfor lightinginthetunnel. At portal
area, natural light transitionandartificial light com-
binationisusedfor lighting. Fluorescencelampisthe
mainlightsourceinthetunnel. Strengtheninglighting
uses thehigh pressuresodiumlamp. Takingaccount
of the energy consumption, the application research
of LEDwithhighpower isbeingdeveloped. Theshift
timefor emergency lightinginthetunnel shouldnot
belarger than0.1sandtheemergencytimeis90min.
4.6.5 Monitoring and control system
Thecomprehensivemonitoringsystemconsistsof traf-
ficmonitoringsystem, equipmentmonitoringsystem,
CCTV monitoring system, communication system,
fire automatic alarming system, central computer
management system, monitoring and control center.
Equipmentmonitoringsystemisclassifiedasventila-
tionsubsystem, water supplyanddrainagesubsystem,
lightingsubsystem, andelectrical monitoringsubsys-
tem.Monitoringsystemhasaccessprovisionforhealth
monitoringsystem, andexpressway net traffic moni-
toring emergency center, rail traffic monitoring and
220kV, etc.
The information collected by the tunnel monitor-
ingsystem, bridgemonitoringsystem, andtoll station
systemis transferred to the monitoring and control
center inthetunnel andbridgeadministrationcenter
on Changxing Island. Furthermore, one administra-
tioncenter is arrangedat WuhaogouonPudongside
assistingthedaily management andemergency treat-
ment, establishingthethreelevel frameof monitoring
andcontrol center administrationcenter outfield
equipment.
4.7 Fire-fighting system
The fire fighting sytemdesign cosists of balanced
and redundant design of safety and function for the
entire tunnel structure, building, water supply and
drainageandfirefighting, emergencyventilationand
smoke discharge, lighting, power supply and other
subsystems. Thedetailsareasfollows:
Crosspassageisarrangedevery830mconnecting
theupchainageanddownchainagetunnel for pas-
senger evacuationwithaheightof 2.1mandwidth
of 1.8m. Three evacuation ladders are arranged
betweentwo cross passages connectingtheupper
andlower level.
The passive fire proof design uses the German
RABT fireaccident temperaturerisingcurve. The
fire accident temperature is 1,200
C. Fire proof
inner lining which ensures the surface tempera-
tureof protectedconcretesegmentisnotmorethan
250
C within 120 minutes is selected to protect

thearchabovesmokeduct, smokeduct andcrown
above the finishing plate. For rectangular tunnel,
fireproof material whichensuresthestructuretop
platesafety within 90 minutes is selected to pro-
tect thetopplateand1.0mbelowthetopplate. To
ensurethepassengerevacuation,fireproof bursting
resistancefibreis mixedintheconcretebulkhead
to achieveno damageof structurewhenstructure
isexposedtofirefor 30minutes.
Theventilation systemis designed based on only
onefireaccidentinroadtunnel andrail trafficarea.
The marginal arch area of bored tunnel is used
for smoke duct. Special smoke ventilation valve
is arranged every 60mfor the smoke ventilation
incaseof fireaccidents onroadlevel. Whenfire
accidentoccursinlowerlevel, ventilationfaninthe
working shaft is turned on to ventilatethesmoke
tothesideof firesourcewhilepassengersevacuate
towardsthefreshair.
The emergency lighting is arranged on two sides
withthesametype. Asthebasic lighting, inserted
intothebasic lightinguniformly. Meanwhile, nor-
mal lightingandemergencylightingareinstalledin
35
thecablepassage. Evacuation guidancesigns are
arranged on the two sides of road, cross passage
andsafetypassage. Emergencytelephoneguidance
signsarearrangedabovethetelephonesintunnel.
Fire water supply at both ends of tunnel is from
theDN250water supplypipeintroducedfromtwo
different municipal water pipeswithout firewater
pond. One fire fighting pump plant is arranged
in working shaft on Pudong side and Changx-
ing Island, respectively. The fire hydrant system
is continuous in the longitudinal evacuation pas-
sage. Firehydrant groupisarrangedevery50mat
onelanesideineachtunnel andfireextinguisher
groupevery 25m. Foam-water sprayingsystemis
usedinthetunnel whichcanprovidedfoamliquid
continuouslyfor 20minandarrangedevery25m.
The communication and linkage of each sub-
systemof comprehensivemonitoring and control
system can realize the monitoring, control and
test of thewholetunnel suchas fan, water pump,
electrical and lighting equipment. Fire automatic
alarming systemcan detect the possible hazards
suchas firefast, real-timeidentify andalarmand
has the function of passage alarming and tunnel
closed. Furthermore, corresponding equipments
canbeautomaticallyactivatedtoextinguishthefire
at earlytimeandorganizethehazardpreventionto
reducethelosstotheminimumextent.
5 +15, 430MM SLURRY MIXEDTBM
Two large slurry pressurized mixed shield machines
withadiameter of 15.43mareusedfor theconstruc-
tionof 7.5mlongboredtunnels.
5.1 TBM performance and characteristics
TheTBM consistsof shieldmachineandbackupsys-
temwithatotal lengthof 13.4mandweightof 3,250t,
including cutter head system, shield body, tailskin,
main drive, erector, synchronized grouting system,
transportsystem,guidancesystemanddataacquisition
systemandslurrysystem.
The TBM has excavation chamber and working
chamber.Duringadvancing,theairbubbleinthework-
ing chamber is adjusted through the control unit to
stabilize the slurry level thus balance the water/soil
pressureinexcavationchamber, asshowninFigure8.
Thethrustsystemconsistsof 19groupsthrustcylin-
ders with a total thrust force of 203,066kN. Cutter
head is drived by 15 motors with 250kW power, so
thetotal power is 3,750kW. Installation position for
twosparemotorsisalsoprovided. Tailskinseal struc-
tureiscomposedof threerowssteel wirebrushesand
onesteel platebrush, forming3greasechambers. The
erector systemis centrally supportedwith6freedom
Figure8. Bulkheadof MixshieldTBM.
degrees. Vacuumsuction plated is used to grasp the
segment. 6-point grouting is used for simultaneous
grouting.
Backupsystemconsistsof 3gantries:gantry1hous-
ing thepower equipment and control system, gantry
2 housing 3 cranes and bridge section for segment,
road element, and other construction material trans-
port, gantry 3 is pipelaying gantry for carrying the
extensionof thedifferent servicessuchascablehose,
slurry, air andindustrial water pipes.
Excavated soil is transported from excavation
chamber to theslurry treatment plant (STP) through
theslurry pipeintheslurry circulationsystem. After
separationbythetreatment equipment, excavatedsoil
withlargesizeisseparatedandthentherecycledslurry
ispumpedbackintoexcavationchamber andworking
chamber.
5.2 Adaptability to the large, long and deep
characteristics
For theTBMconstruction, firstlytheprojectandcrew
safetyshouldbeensured.Thekeyforsafetyof TBMis
toprotectthecutterheadandtailskin,mainlyincluding
cutter headdesign, mainbearingseal andtailskinseal
assurance. Furthermore, the maintenance and repair
of theseparts arerisk and difficult to access, so the
inspectionandpossibility for maintenanceincaseof
failuremust beconsidered.
5.2.1 Cutter head and cutting tools
Cutter head is for soft ground and can berotated in
two directions. The cutter head is pressure resistant
steel structureandspecificwearprotectionisarranged
for theperipheryarea. Special wear protectionisalso
designedfor cuttingtools.
The closed type cutter head is designed with 6
main arms and 6 auxiliary arms, 12 large material
openingand12small material opening. Theopening
ratioisaround29%. 209cuttingtoolsarearrangedon
36
Figure9. Mainbearingseal arrangement.
thecutter head, amongwhich124fixedscrapers, 12
buckets, 2copycutters, 7replaceablecenter toolsand
64replaceabletools.
Thescrapersarecustommadelargetoolswithfea-
tures of 250mm width, wear-resistance body and
highquality carbidealloy cuttingedges whoseangle
matchestheparameterof excavatedground.Thescrap-
ers at theedgeareusedto removetheexcavatedsoil
at edgeand protect thecutter head edgefromdirect
wear.Copycuttercanautomaticallyextendandretract.
Themultipleover-cut areas can besetup in thecon-
trol cabinandcorrespondingcuttingtoolspositionare
displayed. Thereplaceablecuttingtools havespecial
seal topreventtheslurryatthefrontsurfaceenter into
thecutter headchamber. Duringoperation, thework-
ers can enter thecutter head chamber to replacethe
cutting tools under atmospheric condition with high
safety, goodoperationpossibilityandlowrisk.
In order to avoid clogging at cutter head center,
theopeningat center isdesignedaschutetoeasethe
material flowing. Meanwhile, one bentonite hole is
arrangedat each openingto easeflushingin caseof
clogging.
5.2.2 Main bearing seal
Two sets seal systemarearrangedfor themainbear-
ing seal design. The outer seal is for the excavation
chamber sideandinner seal for theshieldbody with
normal pressure. The special seal combination can
bear apressureof 8.5bar.
Outer seal istoseparatethemainbearingandexca-
vation chamber. Seal type is axial seal with large
diameter, totally 4 lip seals and one pilot labyrinth,
thusforming4separateareas, asshowninFigure9.
Theinner seal onethegear boxsideisspecial axial
seal whichcancarrythepressureof gear chamber.
The seal systemhas grease lubrication and leak-
agemonitoringsystemwhichcanmonitor thegrease
amountbypressureandflowmonitoring.Theseal sys-
temhasbeenprovedsuccessfullyinmanyprojectsfor
several yearsandbecomeastandardconfiguration.
5.2.3 Tailskin
Thetailskin is sealed off by 3 rows steel wirebrush
and 1 steel plate brush, as shown in Figure 10.
Figure10. Tailskinstructure.
Furthermore, 1 emergency seal is arranged between
the3rdrowsteel wirebrushandthesteel platebrush.
Theemergencyseal hasthefunctiontoprotectthering
buildingareafromwater ingress whilechangingthe
first threesteel brushseals. Duetonopractical appli-
cationreferencesof thistechnology, modelingtesthas
beencarriedoutfor theemergencyseal installationto
confirmthereliabilityof theemergencyseal whenthe
inflatableseal ispressurizedto1MPa.
Simultaneous grouting lines are arranged at the
tail skin, including one standard grout pipeline and
one spare pipeline for filling the annulus gap out-
side the segment after excavation. Furthermore, 19
chemical grout pipesareaddedfor special hardening
grout (simultaneous slurry penetrating into cement)
or polyurethanefor leakageblock inemergency con-
dition. 193greasepipes havethefunctionof steel
wirebrushlubricationandtail skinsealing. Theseal
systemiscontrolledfromthecabinetinautomaticand
manual modesthroughtimeandpressurecontrol.
Furthermore, freezingpipelinesarearrangedatthe
tailskintoeasethegroundtreatmentbymeansof freez-
ing measures in caseof leakageand ensuretheseal
treatment andrepair safety.
5.2.4 Man lock and submerged wall
Duringlongdistanceadvancing, thereisapossibility
of operation failure of mixing machine due to large
obstacles blockingsuchas stones, mainbearingseal
replacement due to wear, submerged wall closed or
leakageexaminationintheair bubblechamber. These
maintenance and repair work need workers access
theair bubblechamber withapressureupto 5.5bar.
Therefore, two manlocks arearrangedto ensurethe
maintenanceandrepair workerscanaccess.
Themainchamber of manlockcanhouseone1.8m
stretcher.Underpressure-reducingcondition,themed-
ical staff can access themain chamber and organize
rescueincaseof emergency.Meanwhile,theotherman
lock can transport thetools, material and equipment
fromTBM totheair bubblechamber.
Themanlockisequippedwithpoisonousgasdetec-
tion systemwhich can take the sample of enclosed
gas in the man lock. The system information will
be displayed at the working position where outside
37
staff is. The man lock also provides the flange con-
nection. Once the rescue and injuries enters into
temporary rescue chamber, the temporary chamber
can be disassembled fast and transported out of the
tunnel, connectedwithlargemedical chamber for the
convenienceof medical worktorescue.
The submerged wall uses hydraulic drive and is
equipped with air pressure seal strip. When normal
operationintheworkingchamber isneeded, thesub-
mergedwall canbeclosedthustheexcavationchamber
andworkingchamber canbeseparated, andthenthe
valvecanbeopenedfor reducingthepressure. At this
time, pipefor supplementingslurry whichpenetrates
workingchamber canmaintaintheslurry pressurein
theexcavationchamber.
6 TUNNEL CONSTRUCTIONMETHOD
6.1 Overall arrangement and time schedule
Basedontheoverall programming, theconstructionof
working shafts, bored tunnel, synchronous construc-
tionof roadstructure,operationequipmentinstallation
andcommissioningarethemainworksandsecondary
works such as receiving shaft and crosspassage in
parallel.
InMay2006, thelaunchingshaftandonshorestruc-
turesonPudongsidewerecompletedandsiteassembly
of twoTBMsstarted. Theeasttunnel startsadvancing
inSeptember2006, whilewesttunnel inJ anuary2007.
During construction of these two tunnels, the pre-
fabricatedroadelement erectionandTBM advancing
are synchronous, which on one hand resist the tun-
nel floatingduringconstructionstageandontheother
handprovidespecial truckpassagefor segments, pre-
fabricatedroadelements andmaterials to realizethe
fast boredtunnel construction. Inparallel withbored
tunnel construction, theroaddeckstructureconstruc-
tionisalsocarriedout200250mbackfromsegment
erectionandtopsmokeduct will start constructionin
J anuary, 2008, forminggraduallyworkingflowintun-
nel. After west tubeTBM advancing 3km, the first
crosspassage started construction in October, 2007.
After thetunnel is through, final connectionwork of
working shaft and road structure is carried out and
operationequipmentandfinishingandpavementwork
will start.
6.2 Main critical technical issues during bored
tunnel construction
6.2.1 TBM launching and arriving technology
6.2.1.1 TBM launching
(1) Tunnel eyestabilization
3-axial mixing pile and RJ P injection procedure is
used surrounding the working shaft to stabilize the
ground forming a stabilized area of 15min length.
6dewateringwellsforbearingwateraresupplemented
Figure11. Water stoptanksketch.
beyondthetreatedgroundareaandholesareboredfor
groutingtheannulus to ensurethesafety duringtun-
nel gateremoval.Thesethreemeasuresapplicationhas
achievedgoodperformance. DuringTBM launching,
thetreatedsoil isstable.
(2) Tunnel annulusseal
Thediameterof tunnel eyeisupto15,800mm.Topre-
vent theslurry entersintotheworkingshaft fromthe
circularbuildgapbetweentunnel eyeandshieldorseg-
mentduringlaunchingthusaffecttheestablishmentof
front facesoil andwater pressure, goodperformance
seal water stoppingfacilityisarranged. Thefacilityis
aboxstructurewith2layerswaterstoprubberstripand
chainplateinstalled, asshowninFigure11. Theout-
sidechainplateis adjustablewith50mmadjustment
allowance. Furthermore, 12grout holes arearranged
uniformly alongtheoutsidebetweentwo layer water
stoponthebox for thepurposeof sealingincaseof
leakage at the tunnel eye. The outer end surface of
water stopfacilityshall bevertical tothetunnel axis.
(3) Backsupport forTBM
Theback up shield support includes 7 rings, among
which-6is steel ringcomposedof 4largesteel seg-
ments with high fabrication quality to ensure the
circularityandstiffnessof thereferencering, asshown
inFigure12. After precisepositioningof steel ring, it
issupportedontheconcretestructureof cutandcover
tunnel by19steel strutswithalengthof 1.2m. Other
6 minus closed rings segments are assembled with
staggered joint. Inserts are embedded on the inside
andoutsidesurface. After eachringbuilding, thecir-
cumferential ringandlongitudinal ringareconnected
withsteel platetoimprovetheintegratestiffnessand
ensurethecircularityandringplaneevenness. Mean-
while, the circumferential plane of each minus ring
shall bevertical tothedesignaxis.
6.2.1.2 TBM receiving
(1) Arrangement inreceivingshaft
BeforeTBMreceiving, thediaphragmbetweenreceiv-
ing shaft and cut & cover tunnel and thediaphragm
inthereceivingshaft betweenupchainageanddown-
chainagetunnel shall becompletedtomakethereceiv-
ing shaft as an enclosed shaft structure. Then MU5
cementmortariscastintheworkingshaftwithaheight
38
Figure12. BacksupportsforTBM.
of 3mhigher thantheTBM bottom. Steel circumfer-
ential plateisarrangedalongthesteel tunnel annulus.
Theinner diameter of steel plateis 5cmlarger then
TBM. 18 grout holes are arranged surrounding the
tunnel annulus and inflatable bag is installed in the
tunnel eye.
(2) TBM arriving
Whenthecuttingsurfaceof TBM isclosetothecon-
cretewall of tunnel eye, advancing is stopped. Then
pumpwater inthereceivingshaft totheunderground
water level. Meanwhile, inject doublegrout into the
annulus 30mback fromtailskin through the preset
grout holeonthesegment tostabilizetheasbuilt tun-
nel andblockthewater/soil seepagepassagebetween
untreatedgroundandTBM.
After above work, the TBM starts excavation of
C30glass fibrereinforcedconcreteandaccesses the
working shaft. The cutting surface accesses into the
working shaft and thecutter head will cut theMU5
cementmortardirectlyandsitonthemortarlayer.Dur-
ingaccessinginto theworkingshaft, polyurethaneis
injectedthroughthechemical groutingholes.
(3) Tunnel eyesealingandwater pumping
When2/3of TBM accessesthereceivingshaft, water
pumping is started. After pumping the water in the
workingshaft, continuetheTBMadvancingandinject
the grout timely. When theTBM is in the working
shaft, fill air intheinflatablebagintimetomakethe
inflatedbagseal thecircumferential gap. Meanwhile,
groutingisperformedthroughthe18holesonthetun-
nel annulus. Grout material ispolyurethane. After the
gapisfullyfilledwiththegrout, theair ininflatedbag
could bereleased slowly under closeobservation. If
any water leakageisobserved, thepolyurethaneshall
beinjectedagainfor sealing.
Whenthetunnel gateringisout of thetailskin, the
weldingworkbetweenringsteel plate, seal steel plate
andembeddedsteel platesshall bedoneimmediately
tofill thegapbetweentunnel gateringandtunnel.
6.2.2 TBM advancing management
6.2.2.1 Mainconstructionparameters
DuringTBM construction, theconstruction parame-
tersshall bedefinedandadjustedbasedontheoretical
calculation and actual construction conditions and
monitoreddatatorealizedynamical parametercontrol
management.
Theadvancingspeedat beginningandbeforestop
shall not be too fast. The advancing speed shall be
increasedgraduallytopreventtoolargestartingspeed.
Duringeachringadvancing, theadvancingspeedshall
beas stableas possibletoensurethestability of cut-
tingsurfacewater pressureandsmoothness of slurry
supplyanddischargepipe. Theadvancingspeedmust
be dynamically matching with the annulus grout to
fill the build gap timely. Under normal boring con-
dition, the advancing speed is set as 24cm/min. If
obstacles varying geological conditions are experi-
encedat thefront face, theadvancingspeedshall be
reducedapproximatelyaccordingtoactual conditions.
Basedonthetheoretical excavationamount calcu-
latedfromformulaandcomparedtoactual excavated
amount whichiscalculatedaccordingtothesoil den-
sity, slurry discharge flow, slurry discharge density,
slurrysupplydensityandflow, andexcavationtime, if
theexcavationamountisobservedtoolarge, theslurry
39
density, viscosityandcuttingfacewater pressureshall
becheckedtoensurethefront surfacestability.
In order to control theexcavated soil amount, the
flowmeteranddensitymeterontheslurrycircuitshall
becheckedperiodically.Theslurrycontrol parameters
are: density =1.151.2g/cm
3
, viscosity=1825s,
bleedingratio-5%.
Single type grout is used to inject at 6 locations,
whichiscontrolledbybothpressureandgroutamount.
Thegroutpressureisdefinedas0.450.6Mpa.Actual
groutamountisaround110%of theoretical buildgap.
20h-shear strength of grout shall not be less than
800Paand28daystrengthshall beabovetheoriginal
soil strength.
6.2.2.2 Shallowcover construction
At the launching section, the minimumcover depth
is 6.898m, i.e. 0.447D, which is extremely shallow.
Toensurethesmoothadvancing, 12msoil isplaced
abovethetop. Meanwhile, inorder to prevent slurry
blow-out,leakage-blockingagentismixedintheslurry
andsurfaceconditioniscloselymonitored.
6.2.2.3 Crossingthebankof YangtzeRiver
Before theTBM crossing, the terrain and land fea-
tureintheconstructionsurroundingareaarecollected,
measuredandphotographedfor filing. 155monitor-
ingpointsarearrangedalongthebankin7monitoring
sections. During TBM crossing, the pressure is set
accordingtothewater pressureat excavationsurface
calculatedfor eachring. Theslurry parameter isalso
adjustedtimelybasedonthesurfacemonitoringinfor-
mation. Greaseinjectionattail skinisperformedwell
to avoidleakageandsynchronous grout amount and
qualityarestrictlycontrolled.
6.2.2.4 Adversegeological condition
(1) Shallowgas
When the TBM is crossing the deposit on Pudong
side, methane gas may be experienced in the shal-
low area. At this time, the ventilation in the tunnel
shall be increased to ensure good ventilation condi-
tionsof TBM. Theconcentrationtest of methaneand
combustiblegasarecarriedout.
(2) Lens
Prior to the construction, geological investigation is
carriedouttolearnthegeneral locationof prism. Dur-
ingconstruction, theTBMissetwithsuitablespeedto
crossthestratumasfast assafelypossible.
(3) Boredhole
Due to the tunnel alignment adjustment, 9 geologi-
cal bored holes will be experienced along theTBM
advancing. Duringcrossing, slurrywithlargedensity
is used and polyurethaneis injected surrounding the
tunnel after crossing.
6.2.3 Quality assurance technical measures for
large tunnel
6.2.3.1 Segment prefabrication
Nine sets steel formwork with high preciseness are
used for segment prefabrication to fulfill thetechni-
cal requirement to segment such as allowable width
tolerance0.40mm, thicknesstolerance+3/1mm,
arclength1.0mm, circular surfaceandendsurface
plainness 0.5mm. In order to control prefabrica-
tion preciseness strictly and ensure the production
quality, special laser survey system is introduced
to conduct accurate measurement of segment pro-
filedimensionbesidetraditional surveymeasurement
toolsandsegment trial assembly.
Fly ash and slag are mixed in the concrete
for segment prefabrication. Strictly concretecasting,
vibrating and curing procedures are used to control
cracks andachievethewater proofinganddurability
requirement.
6.2.3.2 Segment assembly
Thesegment assembly shall satisfy thefitted tunnel
designaxisrequirementbysegmentselection(rotation
angle) andmeanwhilemakethelongitudinal joint not
onthesameline. Duringthewholeassemblyprocess,
forstraightline,theprincipleistoerectonleftandright
at intervals. For curvedsection, thesuitablesegment
rotationangleshall beselectedbasedonTBMattitude,
andsegment lippingdata.
Secondly, therelativedimensionbetweensegment
andshieldshall becheckedtocorrect thepositioning
of eachringsegment.
Then, eachsegment buildingshall beclosely con-
tacted. TheringplaneandT joint shall beeven.
Finally, strictly control the lipping of ring. When
the segment lipping exceeds the control value, the
rotational angle of segment shall be adjusted timely
to ensuretheverticality betweensegment andtunnel
axis.
6.2.3.3 Floating-resistanceof tunnel
Due to the tunnel diameter up to 15m, the floating
resistance and deformation control during construc-
tion for large diameter tunnel are very challenging.
Thetechnical measureis mainly to improvethesyn-
chronous grouting management. Mortar type grout-
ingmaterial withcementationproperty is injectedat
multi-points. Furthermore, groutpackagewithcertain
strengthshall beformedsurroundingthetunnel timely
toresist thetunnel upfloating. Meanwhile, thetunnel
axisshall bestrictlycontrolledduringconstructionand
thetightconnectbetweensegmentsshall beimproved
toachievethetunnel-floatingresistance.
6.2.3.4 Grounddeformationcontrol
The ground settlement during TBM construction is
mainly contributed by the front surface slurry pres-
suresetup, annulusgroutingandshieldbody tamper.
40
Figure13. STP systemflowchart.
Therefore, thegroundsettlementvariationcandirectly
reflect theTBM construction parameters setup. The
crewcancorrect theconstructionparameter basedon
settlement monitoringtoincreasethedeformation.
6.2.4 Back-up technology for long distance TBM
construction
6.2.4.1 Slurrytreatment andtransport
Theslurry separation systemconsists of subsystems
of treatment, conditioning, new slurry generation,
slurry discharge and water supply; with a capacity
of 3,000m
3
/h to fulfill the advancing requirement
of 45mm/min, as shown in Figure 13. Based on
thegeological conditionsalongthetunnel alignment,
thetreatment systemselects 2 level treatment meth-
ods. The initial treatment uses two rolling shieve to
separatesoil withasizeof larger than7mm. For sec-
ondary treatment, firstly grain with a size of large
than 75 is separated by 4750mmcyclones and
thengrainwithasizeof largethan40misseparated
by 12300mmcyclones. Theslurry spilledat the
topof cycloneistransportedtoconditioningtank for
reuse. After adjustment, thedensityof suppliedslurry
is 1.051.35g/cm
3
. Themaintained optimal valueis
between 1.20 and 1.30 and d50 is between 40 and
50m. The STP systemcirculation efficiency is up
to70%. Dischargedslurryandwasteistransportedto
thebardgeatriversidebypipesandtrucks. Theslurry
supply pipehas adiameter of 600mmanddischarge
pipe500mm.Toensurethelongdistanceslurrysupply
velocity of 2.5m/s anddischargevelocity of 4.2m/s
toavoidslurrysettlementinpipeandmaintainnottoo
highpressureinthepipe, onerelaypumpisarranged
every1km. Thepressureat pumpoutlet iscontrolled
within10bar.
6.2.4.2 Axiscontrol andconstructionsurvey
guidance
Static measurement withGPS control net is usedfor
surfacecontrol survey. For elevationcontrol, GPSele-
vationfitmethodisusedfor elevationtransfer. Partof
basic traversemark every 500mis selected as main
traverse. Inthetunnel, level II subtraverseisusedfor
theplanecontrol, i.e. construction traverseand con-
trol parallel traverse. Thecontrol mark has aspacing
of 600900m. Theelevationcontrol survey intunnel
useslevel II. Thefixedlevel mark isarrangedwitha
spacingof 80m.
6.2.4.3 Constructionventilationandfireprotection
Duetothelargediameter, longdistanceandW lon-
gitudinal slope, especiallywhentheTBMisadvancing
withaupgradingslope, theheat andhumidity gener-
atedattheworkingfacecannotbedischargednaturally
thus concentrate at the working face in a shape of
fog. Meanwhile, heavy trucks for constructionmate-
rial transport also causealargeamount of wasteair
41
inthetunnel. Badenvironment will haveunfavorable
influenceonTBMequipmentandcrew, andalsoaffect
thesmoothprogressingof surveyactivity.
During construction stage, 2 special axial fans
(SDF-No18) are arranged on the surface to provide
fresh air to thespacebelowroaddeck in thetunnel,
then the relay fan and ventilation systemequipped
on the gantry will transport the fresh air to work-
ing surface. Meanwhile, other ventilation equipment
on the gantry provides fresh air to main secondary
equipments of TBM such as transformer, hydraulic
equipment andelectrical installations.
Adequate fire extinguishers are arranged in the
shield and gantry and also oxygen, poisonous gas
protection mask are equipped. Fire extinguisher is
equipped on each transport truck. Safety staff is
equippedwithportablegasanalysisdevicefor check
theair qualityintunnel everyday.
6.2.4.4 Material transport
Segment, grout and prefabricated elements, etc are
transportedtotheworkingareabyspecial trucksfrom
ramparea, throughcut & cover tunnel androaddeck
which is constructed synchronously. Truck transport
can avoid the derailing problems during traditional
electrical truck transport. Furthermore, thetruck has
twolocos, sothetransport efficiencyishigh.
Prefabricated road element is transported to the
gantry 2by trucks andthenliftedanderectedby the
craneonthebridgebeam. Segmentsaretransportedto
gantry2andthentransferredtothesegmentfeeder by
thecraneonthebridgebeamandthentransportedto
erectionarea.
6.2.5 Critical equipment examination and
replacement technology
6.2.5.1 Mainbearingsealing
Four supersonic sensors are installed in the seal
arrangementfor monitoringthemainseal wear condi-
tion. Oncetheabrasionreachescertainvalueorgrease
leakageismonitoredinthetank, it indicatesthemain
seal needstoberotatedtoanother oritentation.
Oncetheseal wearisobservedbeyondpresetvalue,
thesurfacecouldbemovedtoensurethereplacement
of mainbearingseal. Duringreplacing, theslurry in
thechambermustbedrainedandprovideeffectivesup-
port to excavation face. Theoperation staff shall go
totheslurry camber toreplacetheseal under certain
pressure.
6.2.5.2 Abrasionmeasurement andreplacement of
cuttingtools
Thesystemwill beinstalledon8selectedscraperposi-
tions as well as on two bucket positions. It will be
connected to a plug at the rear of the cutting wheel
to allowfor simpleconditiondiagnosis fromaread-
outdevice. Conductor loopisembeddedinthedevice.
Thewear condition of cuttingtools can beindicated
bycheckingtheclosed/openstatusof loops.
Theworker accesses thecuttingwheel arms from
thecenter of themain drive. Theworker installs the
lowering/ lifting frame (with bolts) and screws it to
the fixing plate of the tool. The fixing plate is then
unscrewed. Theworker will thenlower thetool using
the frame (with bolts). The pressure-tight gate will
be closed down. The worn out tool shall be then
exchangedwithanewone. Thetool will beliftedto
positionbehindthegate.Thegatewill beopened.Then
the tool will be put in its final position. The fixing
plateisthenscrewedtothetool support. Theframeis
transferredtothenext tool.
6.2.5.3 Tail seal andsteel wirebrushreplacement
When the leakage is experienced at tail skin, and
steel brushisdefinedtobereplacednecessarily, open
the emergency sealing and erect special segments.
Strengthenthesurroundingsoil attail skinwithfreez-
ing method and then replace first 2 or 3 rows steel
brush.
6.3 Synchronous construction of road deck
The synchronous construction of road structure
includes erectionof roadelement, segment roughen-
inganddrillingfor insertingrebar, prefabrication of
two sideballast, insitu cast corbel and road deck on
twosides.Accordingthevariationandtrendof asbuilt
ringdeformationandsettlement, andtheconstruction
progress of 12m(6 rings) per day and based on the
requirement of deformation joint arrangement every
30m, the construction is organized and arranged as
flowingoperationevery15m. AsshowninFigure14,
thebasicconstructionprocedureisasfollows:
Roadelement installation, 25ringslater thanseg-
ment erection.
Segment rougheningincludesthejunctionsurface
between ballast and segment and segment inner
surfaceat corbel. Theinsert bar placing includes
the+16bar at ballast and+20bar at corbel. The
rougheningworksatballastpositioniscarriedoutat
gantry2, andtherougheningoperationplatformat
corbel isfixedtogantry2. Insertbar placingisfol-
lowinggantry3.Theroughingmachineisequipped
withdust suctionfacility whichcaneliminatethe
dust tomaximumextent.
Reinforcement placing, formwork erection and
concretecasting for ballast is carried out at 15m
behindthegantry3and15mmorebehindfor cor-
bel,andthenanother15mforroaddeck.Roaddeck
concretecastingworksarelocatedat250m300m
fromthesegment erectionarea. After casting, the
curing with frame lasts 3 days and formwork is
removedonthe4thday. After 28days curing, the
roaddeckcanbeopentotraffic. Duringcuring, the
42
Figure14. Synchronousconstructionflowchart.
roaddeckareaisseparated. Concretemixingtruck
isusedfor concretecasting.
6.4 Cross passage construction
Thecross passagewhichconnects thetwo maintun-
nelshasalengthof around15manddiameter of 5m.
Theconstructionwill beby freezingmethodfor soil
strengtheningandminingmethodfor excavation.
Thefreezingholes arearrangedas insideandout-
side rows which are drilled from two sides. The
freezingisdonefromonesideorbothsides. Insiderow
holes aredrilledfromupchainagetunnel, 22in total
and outsiderowholes aredrilled fromdownchainge
tunnel, 18intotal.
Miningmethodwill beusedfor excavationbyarea
division. Firstly, pilot with a horn opening is exca-
vated, andthenthecrosspassageisexcavatedtodesign
dimension. Thefullsectionexcavationis donewitha
stepof 0.6mor 0.8m.
Whenthemainstructureconcretestrengthreaches
75%, enforced thawing will be carried out. The hot
brinefor thawing circulates in thefreezing pipeand
the frozen soil is thawed by section. Based on the
informational monitoringsystem, thesoil temperature
andsettlementvariationismonitored. Groutingpipeis
arrangedat shallowanddeeper areafor densegrout-
ing. Theoverall principlefor thawing is to thaw the
bottompart, thenmiddlepart, andlastly thetoppart,
as shown in Figure 15. When thawing by section is
doneinsequence, onesectionisbeingthawedandsub-
sequentsectionsmaintainthefreezingfor thepurpose
of maintaining thecross passagestructureand main
tunnel asanintegratedpart thussettlement avoidance
beforethesectiongrouted.
6.5 Land connections construction
Theprofiledimensionof workingshaftis22.449m,
with a depth of 25m. 1.0mthick diaphragmwith a
depthof 45misusedfor retainingstructure. Opencut
isusedfor excavation. Thesupport systemconsistsof
5layersreinforcedconcreteand1layer steel support.
Insidethepit, 3mbelowthebottom, injectionisdone
interval tomakethestrengthnot lower than1.2MPa.
13.516.0moutsidetheworkingshaft is treated. For
43
Figure15. Dividedthawingareaof crosspassage.
diaphragmat theTBM accessing into the receiving
shaft, GFPR isusedinsteadof normal reinforcement
so that theTBM can cut the retaining wall directly
and thus avoid the reinforcement cutting and tunnel
eyeconcreteremoval, whichsimplifies theconstruc-
tionprocedure, acceleratesconstructionprogressand
reducestheconstructionrisk.
The excavation depth of pit for Pudong cut-and-
cover is 23.19.9m, and Changxing cut-and-cover
17.2m8.4m. According to the excavation depth,
diaphragmwiththicknessof 1.0m, 0.8mand0.6mis
selectedrespectively.Thesupportsystemiscomposed
of reinforcedconcretesupport andsteel support. 3m
underneath the pit bottomis strengthened by rotat-
ing injection and also thejunction between working
shaft andcut-and-cut outsidethepit toensurethepit
excavationstability.
Therampisopencutwithaslopeof 1:3. Theslope
is protectedthroughgreenplantinginthereinforced
concretegridwhichisanchoredinsoil by anchorsto
prevent fromsliding. Inorder to avoidslopesliding,
theslopeisstrengthenedby cement mixedpileswith
adiameter of 700mm.
7 CONCLUSION
During the process from planning to implementa-
tion, Shanghai YangtzeRiver Tunnel hasexperienced
various challenges. Technical support of tunnel con-
struction fromChina and abroad is provided. With
indenpendently developed and owned IPR and fea-
turedTBM tunnel constructiontheory andcoretech-
nologyisestablished, formingthecoretechnologyof
large and long river-crossingTBM tunnel in China.
Special technical issuessuchasliningstructuredesign
of extremely large tunnel, long distance TBM con-
struction and hazard prevention systemfor long and
largetunnel achievetobeinternally state-of-art. Rel-
evant standards, codes, guidance, specification and
patenttechnologyaredevelopedtoimprovethetechni-
cal systemof tunnel constructioninChinaandupgrade
theinternal competenceof tunnel engineering.
REFERENCES
Cao,W.X. etal. 2006. Shanghai YangtzeRiverTunnel Project
design. Shanghai Construction Science andTechnology 5:
26.
Chen, X.K. & Huang, Z.H. 2007. Shanghai Yangtze River
Tunnel TBMcuttingtoolsweardetectionandreplacement
technology. The 3rd Shanghai International Tunneling
Symposium Proceedings: Underground project construc-
tion and risk provision technology: 152157. Tongji
UniversityPublicationCompany.
He, R. & Wang, J.Y. Shanghai Yangtze River Tunnel syn-
chronousconstructionmethodstatement. The 3rd Shang-
hai International Tunneling Symposium Proceedings:
Underground project construction and risk provision tech-
nology:168177.Tongji UniversityPublicationCompany.
Sun, J. & Chen, X.K. 2007. Discussion of TBM selection
for Shanghai Yangtze River Tunnel. The 3rd Shanghai
International Tunneling Symposium Proceedings: Under-
ground project construction and risk provision technol-
ogy: 9198. Tongji UniversityPublicationCompany.
Yu, Y.M. & Tang, Z.H. 2007. Shanghai YangtzeRiver Tun-
nel construction survey technology. The 3rd Shanghai
Zhang, J.J. et al. 2007. Shanghai YangtzeRiver Tunnel TBM
launching construction technology. The 3rd Shanghai
44
Undergroundconstructionindecomposedresidual soils
I.M. Lee
Department of Civil Engineering, Korea University, Seoul, Korea
G.C. Cho
Department of Civil and Environmental Engineering, KAIST, Daejeon, Korea
ABSTRACT: Largescaleundergroundconstructionprojects, includingsubway constructionprojects insix
major cities, have been ongoing in Korea, where residual and granite soils are the most common soil type.
Characteristicsof decomposedgranitesoilsaredifferentfromthoseof puresandand/orclay.Thispaperpresents
an overview of geotechnical aspects of underground construction in urban areas where mostly decomposed
residual soils arepresent, focusing on mechanical properties, apparent earth pressure, effect of groundwater,
and effect of spatial variability in geotechnical properties. Although several important aspects are theoreti-
cally, numerically, and experimentally discussed herein, it remains a challenge to fully understand residual
soils, particularly in relation to the practice of underground construction, because of their complexity and
richness.
1 INTRODUCTION
Large scale subway construction projects have been
ongoing in six major cities in Korean peninsular.
Inparticular, undergroundconstructionwork of sub-
way line No.9 is being carried out under the Seoul
Metropolitan Government along with extension of
subway lines No.7 and No.3. Construction of anew
subway line(LineNo.2) will belaunchedinIncheon
thisyear andwill befinishedbeforetheAsianGames
areheldin2014.
Residual andgranitesoilsarethemostcommonsoil
typeinKorea. Characteristicsof decomposedgranite
soilsaredifferentfromthoseof sandand/orclay.Their
mechanical properties andbehaviors vary depending
on the parent rock types and weathering processes.
Moreover, theprofileof thegroundinKoreais gen-
erally not uniform, isotropic, or homogeneous; mul-
tilayeredconditions arecommon, withgroundbeing
composedof successivelayersof fill and/or sedimen-
tary layers, weatheredresidual soils, andsoft tohard
rock. Therefore, conventional/classic soil mechanics
cannot bedirectlyappliedtothesegroundconditions.
Inordertoprovidedataandmethodologiestoenhance
undergroundconstructioninareascharacterizedbythe
predominant presenceof decomposed residual soils,
thispaperpresentsanoverviewof geotechnical aspects
of underground construction in urban areas where
decomposedresidual soilsarethemaingroundcom-
ponent, focusing on mechanical properties, apparent
earth pressure, effect of ground water, and effect of
spatial variabilityingeotechnical properties.
2 MECHANICAL PROPERTIESOF RESIDUAL
SOILS
2.1 General
Wehaveintensively studiedthecharacteristics of the
following two residual soils: Shinnae-dong and Poi-
dong soils. The characteristics of the two soils are
documentedinTable1(i.e., thetypeof mineral, com-
pactionproperties, plasticity, andsoil classification).
The parent rock of the Shinnae-dong soil is a gran-
itewhilethat of thePoi-dongsoil is abandedbiotite
gneiss.TheShinae-dongsoil isclosertoacohesionless
soil, predominantly consisting of primary minerals,
with only about 10%of fineparticles. On theother
hand, thePoi-dongsoil showsclay-likecharacteristics
due to a large percentage of fine particles and sec-
ondaryminerals. Theparticlesizedistributionsof the
twosoilsareshowninFigure1.
Lee (1991) studied the behavior of a Bulamsoil
in his Ph.D. dissertation. The characteristics of the
Bulamsoil are similar to those of the Shinna-dong
soil, sincetwoareas arevery closetoeachother and
share the same rock origin. Kim(1993) studied the
mechanical behavior of Andong and Kimchun soils.
Thesesoils areclassified as SM in theUnified Soil
45
Table1. Characteristicsof tworesidual soils.
Characteristics Poi-dong Shinnae-dong
Primaryminerals
Quartz (%) 17.7 33.3
Feldspar (%) 15.0 50.0
Mica(%) 9.8 9.0
Secondaryminerals
Kaolinite(%) 23.5 6.0
Illite(%) 20.7
Vermiculite(%) 8.4 2.0
Chlorite(%) 4.5
Montmorilonite(%)
Porosity 0.409 0.358
Maximumdrydensity(kN/m
3
) 16.68 18.64
Percent passing#200sieve(%) 47.36 10.05
Plasticity Nonplastic
LI 34.0
PL 19.84
PI 14.16
Specificgravity 2.74 2.65
USCS SC SW-SM
Note: LI =liquidityindex; PL =plasticlimit; PI =plasticity
index; USCS=UnifiedSoil ClassificationSystem.
Figure1. Particlesizedistributionof tworesidual soils.
ClassificationSystemwiththepercentpassinga#200
sieve being 1417%, and are also included in the
followingdiscussiononmechanical characteristicsof
residual soils.
2.2 Strength characteristics
Figure2presentsasummaryof thepeakinternal fric-
tionangleof eachresidual soil. Theinternal friction
angledecreaseswithanincreaseof finecontents.
Becauseof capillarity, partial saturationaffectsthe
strengthof residual soils. Leeet al. (2005) performed
triaxial teststoobtainthestrengthpropertiesof unsat-
uratedresidual soils. Figure3presentstypical results
of failureenvelopes at different matric suctions. The
internal frictionangleaswell astheapparentcohesion
increasewith an increasein thematric suction (i.e.,
whenthesoil isunsaturated).
Figure2. Peak internal frictionangleversus finecontents
passingthe#200sieve.
Figure3. Failureenvelopesat different matricsuctions.
Lee et al. (2002) performed shear tests on unsat-
urated residual soils and found that the apparent
cohesioncanincreasefrom20kPaat asaturatedcon-
dition(i.e., matricsuction=0kPa) upto200kPaatan
unsaturatedcondition(i.e., matricsuction=400kPa).
As an unsaturated soil is re-saturated, its appar-
ent cohesion can beeliminated. Thus, during tunnel
construction, cohesion loss can be induced by re-
saturation(e.g., seepagehindrance, drainageclogging
and groundwater change), and may result in tunnel
faceinstability. Utilizingthelimit equilibriumanaly-
sisproposedbyLeca&Dormieux(1990), asshownin
Figure4, therequiredsupportpressuretostabilizethe
46
Figure 4. Collapse mechanismof a tunnel face with two
conical blocks.
Figure5. Groundconditionfor limit equilibriumanalysis.
tunnel facecanbecalculatedbyreplacingthespecific
apparentcohesionwithavalueof zero.Asanexample,
thegroundconditionof asiteisshowninFigure5.The
requiredsupportpressure
increasesfromzeroupto
=9.0kPaas thecohesiondecreases from120kPa

to0kPa. Theresultsshowthat theapparent cohesion
isakeyfactor intunnel facestability.
2.3 Deformation characteristics
Oneof themost difficult tasks ingeotechnical engi-
neering is estimating the deformation-related soil
properties properly. Typical material properties com-
monly usedat thedesignstageinKoreaaresumma-
rizedinTable2.
Cho et al. (2006) proposed an analytical method
toestimatesoil parametersfromrelativeconvergence
measurements. Asanexample, thegroundconditions
of theBusansubway siteareshowninFigure6. Ini-
tial estimates aretakenfromTable2. Thecrownand
sidewall convergencedatameasuredby atapeexten-
someter, presentedinFigure7, areusedas observed
values. Resultsobtainedfromback-analysesaresum-
marizedinTable3alongwithinitial estimates. There
isalargediscrepancybetweentheinitial inputproper-
ties and theproperties obtained fromback-analyses.
In particular, the initial inputs of Youngs modulus
and earth pressure coefficient at rest of the residual
Table2. Typical material properties.
E c
Layer (MPa) j (kN/m
3
) K
0
(kPa) (
)
Filling 19.6 0.35 18.6 0.5 0 35
Residual soil 29.4 0.33 18.6 0.5 49.1 35
Weather rock 196 0.23 21.6 0.5 98.1 35
Soft rock 981 0.2 23.5 0.7 196 40
Notation: E=Youngsmodulus, j=Poissonsratio, =unit
weight of soil, K
0
=lateral earth pressure coefficient,
c=apparent cohesion, and =internal frictionangle.
Figure6. Groundconditionsof theBusansubwaysite.
Figure7. Relativeconvergencebehindtunnel face.
Table3. Resultsof parameter estimation.
Properties E
r
(MPa) E
w
(MPa) K
ow
Initial input 29.4 196 0.5
Back-analysis 83.3 210 0.76
Notation: E
r
andE
w
=Youngsmoduli of theresidual soil and
weathered rock, respectively; and K
ow
=theearth pressure
coefficient of theweatheredrock.
47
Figure 8. Apparent earth pressure distribution for braced
andanchoredwalls.
soil aretoosmall. Thisexampleclearly demonstrates
theimportanceof theobservational methodintunnel
engineering.
3 APPARENT EARTHPRESSURE
3.1 Apparent earth pressure in braced and
anchored walls
Designof aninsituwall systemrequiresalateral earth
pressuredistributionbehindthewall toestimatesup-
portloadsandwall bendingmoments. Oneof themost
well-knownapparent earthpressuresisthat proposed
byPeck(1969); however, hissuggestionisonlyappli-
cabletoeither groundthat ispurely sandand/or clay,
and cannot bedirectly applied to cohesivesoils that
havecohesionaswell asaninternal frictionangle, or
tomultilayeredgroundconditions.
ManyKoreanresearchershaveattemptedtocollect
fielddatatoproposetheearthpressureinmultilayered
ground (for example, Lee & J eon 1993, Yoo 2001).
Figure8presentsatypical set of results, showingthe
apparent earth pressure distribution of a 33mdeep
excavation site along with the distribution proposed
by Peck (1969) (Yoo 2001). The actual (measured)
earthpressureisabout 68to83%of Peaks. Figure9
shows the maximumearth pressures obtained from
62 excavation sites. In this data, the weighted aver-
agevalues of theinternal friction angleand theunit
weight of soil areusedfor themultilayer ground. The
averagevalueof themeasured apparent pressures is
approximately 75% of Pecks suggestion (i.e., earth
pressure=0.65K
a
H).
Figure9. MaximumapparentearthpressureversusK
a
H.
3.2 Apparent earth pressure in a vertical shaft
Itiswell knownthattheearthpressureactingonaver-
tical shaftislessthanthatonaretainingwall, because
of thethreedimensional archingeffect. Theexisting
equations of earthpressures actingonvertical shafts
consider only either purely cohesionless or cohesive
soils. These solutions are not directly applicable to
estimationof earthpressuresformulti-layeredground.
Leeet al. (2007) proposedanequationtoestimate
earth pressures in multi-layered ground, assuming
that the failure shape is conical, as shown in Fig-
ure 10(a). For equilibriumof horizontal forces and
vertical forces, as shown in Figure 10(b), the earth
pressure(P
i
) canbeexpressedasfollows
and
where K
w
=the coefficient of radial earth pres-
sure, =thecoefficient of tangential earthpressure,
and =the wall friction angle. Figure 11 presents
schematic drawings of a construction site in multi-
layeredgroundandthreevertical shaftsalongwiththe
locations of measuring instruments. Earth pressures
aremeasured at different shafts. Theearth pressures
calculated fromEq. (1) arecompared with themea-
sured values, as shown in Figure 12. The measured
earth pressures areeven smaller than thoseobtained
fromthetheoretical equation.
48
Figure10. Derivationof earthpressureinvertical shaft.
4 EFFECT OF GROUNDWATER
4.1 Effect of seepage pressure
Unexpected groundwater inflow and seepage forces
oftencausetunnel failures duringconstruction. Shin
et al. (2006) presentedandreviewedfivecave-incol-
lapses that occurredwhileconstructingLineNo.5of
theSeoul Metropolitansubway. Figure13shows the
general features of the collapses and failure details
aresummarizedinTable4. A comprehensivereview
Figure 11. Schematic drawings of a construction site in
multi layeredground Threeshafts withlocations of mea-
suringinstruments.
on such collapse mechanisms reveals the following
commonfeatures:
1. Thin soil/rock cover and/or mixed faced ground
conditionsincludingdecomposedgranitesoils;
2. Collapse initiated at the tunnel shoulder during
excavationof theupper half of thetunnel section;
and
3. A considerableamountof groundwaterinflowwith
soil.
Inparticular, it is observedinthesesites that tun-
nel facecollapsesalwaysoccurredalongwithseepage
aheadof thetunnel face.
Leeet al. (2003) modified theupper bound solu-
tionoriginallyproposedbyLeca&Dormieux(1990),
takingintoaccountseepageforcesinastabilityassess-
mentof atunnel face(refertoFigure4).Thehorizontal
componentsof seepagepressuresactingonthetunnel
face,
S.F
, canbesimplyconsideredasanexternal load
49
Figure12. Earthpressuresmeasuredat vertical shafts.
Figure 13. Cave-in collapses in the Seoul Subway Line
No.5.
in theoppositedirection of thesupport pressure
T
.
A modifiedupper boundsolutionwithconsideration
of seepageforcesbecomes
whereP is thesurcharge,
c
is theunconfined com-
pressivestrengthof thesoil,
T
istherequiredsupport
pressureapplied to thetunnel face,
S.F
is theseep-
agepressureactingonthetunnel face, K
p
isRankines
earthpressurecoefficient for passivefailure, isthe
unit weight of soil, D is thetunnel diameter, and N
s
andN
aretheweightingcoefficients, respectively.
When tunnel excavation is performed below the
groundwater level, thestressconditioninfront of the
tunnel face becomes the summation of the effective
stress and the seepage pressure. The effective stress
canbecalculatedbytheupperboundsolutionwhilethe
seepagepressurecanbeobtainedfromnumerical anal-
yses. Theeffectivesupport pressureat thetunnel face
canbeobtainedbyEq. (4) withuseof thesubmerged
unit weight
sub
insteadof .
As an example analysis, consider a virtual tun-
nel with a diameter D driven horizontally under a
depth C, as shown in Figure 14. Ground material
properties used for the analysis are c=0,
=35
,
and
sub
=5.4KN/m
3
. Figure15showsthetotal head
distribution around the tunnel face, determined by
seepageanalyses, andthefailurezone, estimatedfrom
alimit equilibriumanalysis. Theseepagepressureis
calculatedbyusingJ =i
w
A, wherei isthehydraulic
gradient andA isthearea. Thetotal support pressure
isthenobtainedbysumminguptheeffectivesupport
pressureandtheseepagepressure. Figure16showsthe
supportpressurechangewithvariationof theH/Dratio
(For thecaseof adry condition, thedry unit weight
d
=15.2KN/m
3
isusedfor theanalysis). Theresults
suggestthatthetotal supportpressureislittleaffected
by the tunnel depth and increases significantly with
50
Table4. Collapsemechanisms, damage, andremedial work.
Case Failuremechanism Damage Remedial work
A 17Nov1991, 18:50: blasting nohumancasualties soil dumpingimmediatelyafter
21:05: total collapse(1,000m
3
) 2-1aneroadcollapsed collapse
thinweatheredrockcover stopof gassupply faceshotcreteandinvert Close
inflowof soil andgroundwater (5,000households) (t=1.52.0m)
weatheredgranite(WG) at theface damagetolightingpoles cement milkgroutingandcurtain
veryclosetoanexistingstream andtrafficlight poles wall grouting
B 27Nov1991, 10:40: blasting nohumancasualties grouting: cement mortal cement
16:00:rockfallsat theface 80householdsevacuated milkchemical grout
22:00: soil andwater inflow(D=25m) electricityandwater loweringof groundwater
28Nov1991, maincollapsed level(3m/day)
03:20:additional collapse(D=20m) (3-storybldg) slipped pumping/forepolingfor
WGat theface intocrater re-excavation
permeability: (1.010
4
remedial cost: $4.5million
2.010
5
cm/sec)
C 11Feb1992, roadheader excavation nohumancasualties dumpingsoilsintocollapsed
04:30: raveling 4-laneroadcollapsed area(240tons)
significant inflowof ground serviceculvert (6m3m) dumpingready-mixed-
water (100130/min exposed(including154kv concrete(105tons)
about 4.5tonof soil flewinto electricitycable) grouting
face: D=10m passengersdelay
heavilyWGat theface
D 7J an1993, 03:30: blasting nohumancasualties grouting: cement mortal, cement
collapseafter removingmaterials 2-laneroadcollapsed milkandLW(160holes)
collapsesize: 0.7m1.2m supplystopof water mortal injectionbeneathsewer culvert
startedat theleft sideof crown main(+=200mm) reduceinflowof groundwater
soil inflow: 900m
3
, groundwater: sewer culvertswerebroken usingchemical grouting
300/min 40householdsevacuated
WG& DGSat theface
E 1Feb1993, ringcut nohumancasualties soil dumpingintocollapsed
08:30: rockfall andcollapse 6itemsof excavation area(5,500m
3
)
(oval shapeD=1030m) equipment wereburied chemical grouting(76holes)
inflowof soil withgroundwater compactiongrouting(300holes)
alluvium& DGSat theface remedial cost: $1.7million
beneathanexistingstream
Figure14. Dimensional conditionfor seepageanalysis.
anincreaseinthegroundwater level ratio. Asthetotal
support pressure is related to the tunnel face stabil-
ity, theseepageforceseriouslyaffectsthetunnel face
stability. While the effective overburden pressure is
reduced slightly by the arching effect during tunnel
excavation, theseepagepressureremainsat thesame
level duringtunnel excavation. Thisexplainswhythe
Figure15. Hydraulicheaddistributionandfailurezone.
effectof seepageplaysanimportantroleintunnel face
stabilityproblems.
4.2 Particle transport characteristics of granite
residual soils
A soil issaidtobeinternallystableif itisself-filtering
andif its fineparticles do not move/migratethrough
51
Figure16. Changeof supportpressurewithvariationof the
H/DRatio.
Figure17. Schematicdrawingsof experimental set-up.
thepores of its owncoarser fraction. Previous inves-
tigations into the internal stability of cohesionless
soils suggest that soils with auniformity coefficient
(C
u
)>20andwithconcaveupwardgrainsizedistri-
butionstendtobeinternallyunstable(Leeetal. 2002).
Most residual soils in Korea, including those listed
inTable1, haveuniformitycoefficientsmuchgreater
than20, suggestingthat theyareinternallyunstable.
Lee et al. (2002) studied the nature of particle
transport and erosion in residual soils. Two types
of residual soils introduced in Section 2 are used:
Shinnae-dongsoil andPoi-dongsoil.Theexperimental
setupisshowninFigure17. Inselectedexperiments,
acylindrical hole7mmindiameter isdrilledintothe
compacted specimens to induce erosion only in the
holeandsimulatesurfaceerosionof thesoils.Anelec-
tronicpumpisusedtoachieveaconstant flowrateof
the influent fromthe water tank. The effluent from
Figure18. Cumulativemass versus (a) timeand (b) pore
volumefor Shinnae-dongsoil.
the cell is characterized with respect to its turbidity
(intermsof nephelometricturbidityunits, NTUs) and
particlesizedistribution.
The cumulative mass of particles in the effluent
eroded fromthebasesoils is plotted with respect to
timeinFigures18(a) &19(a) for thetworesidual soil
types. Someimportant differencescanbeobservedin
theinternal erosionbehaviorof thetwosoil types.The
Shinnae-dongsoil exhibitsalmostthesamerateof ero-
sionduringtheinitial stagesof theexperiment for the
threedifferent flow rates used (Figure18a). Particle
redepositioninthesoil sampleappearstocompensate
fortheincreasederodibilityathigherflowrates.Thisis
evenmoreapparentwhenthecumulativemassisplot-
tedintermsof porevolume(Figure18b). Itisseenthat
at lowflowratestheerosionratesarehigher, because
of thereducedparticledeposition.Thereappearstobe
amaximumlimit for thecumulativemassof internal
erosionforeachflowrate, beyondwhichthesoil pro-
tecteditself fromfurthererosion, perhapsthroughthe
formationof afilterbridge. Fortherelativelycohesive
Poi-dong soil (Figure 19a & 19b), self-protection
dueto particleredepositionis not apparent. Thereis
nocaponthemaximumerodedquantitiesduringthe
periodof testing.
Incontrasttotheinternal erosionbehavior, thesur-
face erosion fromthe two samples (as observed in
experiments where erosion is induced in a cylindri-
cal hole) follows almost linear trends, with therates
52
Figure19. Cumulativemass versus (a) time, and(b) pore
volumes, for Poi-dongsoils.
of erosion increasing as theflowrateincreases. The
ratesof erosionarealsoconsiderablyhigherthanthose
obtainedintheinternal erosionexperimentsdiscussed
above.
Particle transport characteristics of residual soils
mightbeamongthefactorsthatresultintheinstability
of undergroundstructures.
4.3 Difficulties in penetration grouting
Tunnelling works in soft ground frequently require
grouting technology, either to prevent groundwater
or to improve mechanical properties of the ground.
However, grouting is not availablein many cases in
decomposedresidual soilsduetolowgroutability.Bur-
well definesthegroutability(N) of suspensiongrouts
bythefollowingsimpleequation(Kimet al. 2007):
where D
15
is the particle size of base soils corre-
sponding to 15% finer and d
85
is the particle size
of grouts corresponding to 85%finer. If N is larger
than 25, grout can be successfully injected into the
soil formation. However, Burwell notes that even in
Figure20. Grain-sizedistributionof soilsandgrouts.
Table5. Soil andgroutpropertiesusedinchamberinjection
tests.
D
10
D
15
d
85
d
95
Material (mm) (mm) (m) (m) N
Soil A 0.60 0.64 32
Soil B 2.10 2.22 111
Finecement 16 27
Quicksettingagent 37 70
Finecement+Quick 20 39
settingagent
caseof N >25, thefollowingrequirement shouldbe
additionallysatisfiedfor thesoil tobegroutable:
where D
10
is the particle size of base soils corre-
spondingto 10%finer andd
95
is theparticlesizeof
groutscorrespondingto95%finer. Kimet al. (2007)
performedpilot-scalechamberinjectionteststoinves-
tigatethegroutabilityof twogranular soilsthatsatisfy
thegroutabilitycriteriaproposedbyBurwell.
Thegrain-sizedistributionsof soilsandgroutsare
shown in Figure20 and their properties aresumma-
rizedinTable5.Theexperimental set-upforpilot-scale
chamber injection tests is shown in Figure21. Typi-
cal resultsof theexperimentsareshowninFigure22.
AlthoughtheNvalueof thesoil A (N=32) isgreater
than25,thegroutcouldnotbesufficientlyinjectedinto
soil A.Meanwhile,groutabilityisfairlygoodforsoil B
(N=111).Theseresultssuggestthattheconsideration
of filtration phenomena is indispensable to reason-
ablyevaluatingthepotential of groutpenetration. The
Nvalueof theShinnae-dongsoil showninFigure1is
2.6andthat of thePoi-dongsoil is 0.2. Thesevalues
meanthatpenetrationgroutinginthesesoilsisalmost
impossible. Therefore, finding an appropriategrout-
ingmethodhaspresentedaconsiderablechallengein
graniteresidual soils.
53
Figure 21. Experimental set-up for pilot-scale chamber
injectiontest.
Figure 22. Maximum injection volume with injection
pressure.
4.4 Ground reaction curve with consideration of
seepage forces and grouting
Theoretical analyses of seepagearound tunnels sug-
gest that a loss in hydraulic heads occurs at the
shotcrete lining and concentration of seepage force
at theshotcreteliningintheradial directioninduces
unfavorablegroundreaction(Shin, 2007).Thus, when
seepageproblems areanticipated during tunnel con-
struction, proper groutingaroundtunnelscanprovide
effective reduction of seepage force acting on the
shotcrete lining and also increases the stiffness and
strength of the surrounding ground. When grouting
Figure 23. Hydraulic head at soil-grouting interface
dependingonthepermeabilityratio.
is applied around the tunnel, a loss of hydraulic
heads occurs inthegroutingzonearoundthetunnel;
thisreducestheseepageforceactingontheshotcrete
lining, andresultsinafavorablegroundreaction.
FollowingDarcyscontinuityequation,thehydraulic
head acting on the soil-grouting interface can be
writtenas
whereH
I
isthetotal headatthesoil-groutinginterface,
H
T
isthetotal headof asite, L
g
isthethicknessof the
grouting, L
s
is thelength across which water travels
throughthesoil media, and isthepermeabilityratio
betweenthesoil andgroutingarea(i.e., =K
g
,K
s
).
Figure23showsthevariationof thehydraulicheadat
the soil-grouting interface with different permeabil-
ityratios. Asthepermeabilityratiodecreasesandthe
groutingthicknessincreases, thehydraulicheadacting
ontheinterfaceincreases.
Finiteelement analyseswereperformedinorder to
exploretheeffect of groutingonthegroundreaction
with consideration of seepage. Seepageforceacting
onthegrouting-soil interfacecanbemodeledbyfully
coupled mechanical-hydraulic analyses, as shown in
Figure24. Material propertiesusedinthisanalysisare
summarizedinTable6. It wasassumedthat shotcrete
isnotapplied, thegroundwaterflowisinasteady-state
condition, thegroutingthickness is 1m, andtheper-
meability ratio is =0.1. Four cases weresimulated
numerically: 1) Grouting with seepage; 2) Grouting
without seepage; 3) No grouting with seepage; and
4) No grouting without seepage. Figure 25 presents
theeffectof groutingandseepageforceontheground
reaction curve as given by the numerical analysis
results. The caseof seepageforce without grouting
54
Figure24. Finiteelement model for seepageforceanalysis
withconsiderationof grouting.
Table6. Material propertiesusedinnumerical simulation.
E c
(MPa) j (kN/m
3
) K
0
(kPa) (
)
Weatheredsoil 50 0.35 18.64 0.5 10 35
Groutedzone 500 0.33 18.64 0.5 100 35
Figure25. Effect of groutingandseepageforceonground
reactioncurve(i.e., =0.1).
yieldsaveryunfavorablegroundreactioncurve,which
induces alargedeformation and requires substantial
internal support. However, if theseepageforceisnot
considered, thegroundreacts almost elastically even
though grouting is not applied. This means that the
seepage force significantly affects the ground reac-
tion behavior. In thecasewheregrouting is applied,
unfavorablegroundreactionsinducedby theseepage
forcecouldbeconsiderablyreduced.
5 CHARACTERIZATIONANDMODELINGOF
GROUTEDRESIDUAL SOIL
5.1 Experimental study on time-dependent
characteristics of grouted residual soil
InKorea, inconventional tunnellinginresidual soils,
pre-reinforcement(groutinjection)istypicallyapplied
ahead of the tunnel face to enhance the construc-
tion safety. In addition, a 1 to 2 day time interval
is given between one face and the next face. Dur-
ing this time interval, it is known that changes in
the material properties occur due to effects of the
curing of the grouting material. However, the stiff-
ness and strength at 28 curing days after the grout
injection aregenerally applied as thematerial prop-
erties for pre-reinforced zones in the design stage
without consideringtheeffect of thetime-dependent
behavior of the injected grout material. Thus, this
paper present anewmethodtocharacterizethetime-
dependent behavior of pre-reinforced zones around
a large-section tunnel in residual soil using elastic
wavesandtoconsider time-dependent characteristics
innumerical modelingfortunnel design(Song, 2007).
Figure26presentsschematicdrawingsof theexper-
imental setupfor investigationof thetime-dependent
characteristics of grouted residual soils: (a) Setup
for elastic wave measurements; (b) Setup for shear
strength parameter measurements. Bimorph bender
elementswereinstalledinthetestingdeviceandused
tosendandreceiveP- andS-waves(Figure26a). The
specimenswerepreparedbymixingaresidual soil with
5%cement (by weight; thecement-water ratio is the
sameasthat usedinthefield).
Figure27showstypical resultsfor theelasticwave
velocity according to thecuring timewhen thenor-
mal stress is
n
=160kPa. Theresults showthat the
wave velocity increases drastically according to the
curingtimeandisalmostconstantafter7days. P-wave
velocity is faster than S-wavevelocity and Poissons
ratio can be readily determined fromthe two wave
velocities.
Figure 28 shows the time-dependent characteris-
tics of shear strength parameters obtained fromthe
direct shear test. As shown in Figure28(a), thefric-
tion angle does not change in accordance with the
curingtime. Ontheother hand, it isapparent that the
cohesion increases with thecuring time; after acer-
tainamountof curingtimethecohesionconverges, as
shown in Figure 28(b). It is deduced that the bond-
ing of cement increases thecohesion and, after with
55
Figure 26. Experimental setup for investigation of
time-dependent characteristicsof groutedresidual soils.
0
200
400
600
800
1000
1200
1400
0 5 10 15 20 25 30
Curing time (Days)
V
e
l
o
c
i
t
y

(
m
/
s
e
c
)
P-wave
S-wave
Figure27. Elastic wavevelocity accordingto curingtime
(
n
=160kPa).
theelapseof time, thecohesionmaintains auniform
valuewiththeendof cementation. Theearlystageof
this phenomenon is controlled by thenormal stress,
but ascuringtimeincreasesthecementationcontrols
thefrictionangleandcohesion.
Thewavevelocityandcohesionof groutedresidual
soils can be respectively correlated with the curing
timeasfollows:
0
10
20
30
40
50
0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7
Curing time (Days)
F
r
i
c
t
i
o
n

a
n
g
l
e

(
)
(a) Friction angle
0
50
100
150
200
250
Curing time (Days)
C
o
h
e
s
i
o
n

(
k
P
a
)
(b) Cohesion
Figure28. Time-dependentcharacteristicsof shearstrength
parameters.
where , , A, and B are the fitting parameters and
t is the curing time. These fitting parameters can
be determined by best-fitting the experimental data
with Eq. (8) and Eq. (9). Also, the shear strength
and strength parameters (i.e., the cohesion and fric-
tion angle) can be uniquely correlated to the elastic
wavevelocities.
5.2 Numerical simulation of time-dependent
characteristics of grouted residual soil
The construction of underground space in residual
soil entails many risk factors such as difficulties in
predicting arching effects and determination of var-
ious uncertain underground properties. Researchers
havesuggestedvarious techniques for auxiliary sup-
portsystemssuchasthereinforcedprotectiveumbrella
method (RPUM), which has the advantage of com-
bining a modern forepoling systemwith a grouting
injection method (Barisone, 1982). This method is
used for pre-reinforcement design before the under-
groundexcavation: not only for small sectiontunnel-
ingwithinweatheredandcrashedzones, but alsofor
largeundergroundspaces. Inaddition, todecreasethe
riskof acollapseorfailureinlargeexcavationcaverns,
researchers have developed various techniques and
constructionmethods. Someexamplesinclude: atun-
nelingmethodusinganadvancedreinforcingsystem
whereadoublesteel pipeis usedfor water-proofing
56
Figure29. 3D tunnel model and time-dependent material
propertiesof thepre-reinforcedzoneafter 12mexcavation.
andaurethaneinjectionisusedfor reinforcement; the
Trevi jetmethod, whichinvolvesconstructinganarch-
shell structure around a tunnel crown with cement
grout; and steel pipereinforced multi-step grouting,
whereabeamarch is constructed around thetunnel
crownwithlargediameter steel pipes, andmultilayer
cement groutinginjectionisemployed.
Threedimensional FE analyseswereperformedto
examinethetime-dependent behavior of thegrouted
zone. Theresultsobtainedfromlaboratorytestswere
appliedto anumerical simulationof atunnel, taking
into account its construction sequence. Figure 29(a)
showsasimulated3D four-lanetunnel model, where
the same stress state and stress level as used in the
experiment were assumed. Figure 29(b) shows the
time-dependent elastic modulus andcohesionvalues
obtainedfromtheexperimental studyaswell asthose
usedinthenumerical analysis.
The time-dependent behavior of a pre-reinforced
zonecanbemodeledusingthefollowingprocedure.
Thematerial properties(i.e., stiffnessandstrength) of
thepre-reinforcedzoneareconsideredastheboundary
conditions fromDay 1toDay 28. Theregisteredini-
tial boundaryconditionsareappliedtoapre-assigned
mesh in pre-reinforcement construction. Thebound-
aryconditionsarethenupdatedaccordingtothefield
constructionsequence.
For a quantitative analysis, the displacements of
each casearenormalized with theresults of apipe-
onlycase. Figure30(a) showsthenormalizedvertical
0
0.2
0.4
0.6
0.8
1
1.2
0 5 10 15 20 25 30
Excavation Length (m)
N
o
r
m
a
l
i
z
e
d

V
e
r
t
i
c
a
l

D
i
s
p
l
a
c
e
m
e
n
t
1D Stiffness and Strength
Time-Dependent Stiffness and Strength
(a) Vertical displacement on a tunnel portal
0.80
0.85
0.90
0.95
1.00
0 5 10 15 20 25
Excavation Length (m)
N
o
r
m
a
l
i
z
e
d

H
o
r
i
z
o
n
t
a
l

D
i
s
p
l
a
c
e
m
e
n
t
Time-Dependent Stiffness and Strength
(b) Horizontal displacement on a tunnel face
Figure30. Variationof normalizeddisplacement.
displacementattheportal.Thetrendof thenormalized
vertical displacement curve for the time-dependent
condition is similar to that of the one day curing
casewithintheinitial excavationsection(-8m). As
the excavation continues, the results of the time-
dependent condition become similar to those of the
23 days curing case, until vertical displacement
eventuallyconverges.Thestiffnessandstrengthof the
pre-reinforcedzonefor the12dayscuringcaseare
roughly3050%of thoseof the28dayscuringcase.
Inotherwords,areductionof thematerial propertiesof
thepre-reinforcedzonemakesitpossibletomodel the
time-dependenteffectof thepre-reinforcedzoneonthe
global tunnel behavior uponinitial tunnel excavation.
Figure 30(b) shows the normalized horizon-
tal displacement at the tunnel face. It is found
that the normalized horizontal displacement for the
time-dependent condition varies within a range of
0.940.98, whichisverysimilartothatof othercases
during excavation. Therefore, pre-reinforcement can
be considered for prevention of collapse rather than
as a means of displacement reduction control at the
tunnel face. Thus, it can beconcluded that grouting
reducesthehorizontal displacementbyapproximately
26%at thetunnel facewiththepre-reinforcement
method.
57
Figure31. Diagramof simplifiedpre-reinforcedzone.
An analysis method combining experimental
and numerical procedures that consider the time-
dependent effect onthepre-reinforcedzoneontunnel
behavior will provideareliableand practical design
basisandmeansof analysisfor tunnelsinsoftground.
5.3 Determination of equivalent design parameters
for the pre-reinforced zone in residual soil
The design and analysis of pre-reinforcement tech-
niquesrequiredesignassumptionsthatareproblematic
at best, resulting in increased uncertainty in tunnel
design.Thepre-reinforcementeffectistypicallymod-
eledbysimulatingtheconstructionsequenceof setting
thereinforcedzoneandthenincreasingthestiffness,
thereby obviating theneed for complex modeling of
eachbulbandsteel pipe. However, thisapproachhas
beenfoundtobeunsuitablewhenthecenter-to-center
distancebetweenpipesislarger thantheexpansionof
thegroutbulb.Thisapproachassumesthatthestiffness
of thepre-reinforcedzoneisthesameasthestiffness
of the grout bulb, which may be either completely
hardened or arbitrary, and that the pre-reinforced
zonebecomes twotofour times stronger thanbefore
reinforcement. Therefore, this study presents a new
technique for determining a reasonable equivalent
parameter of thepre-reinforcedzone.
A pre-reinforced zone consists of ground, grout
bulbs,andsteel pipes.Itmaybesimplified,asshownin
Figure31, toaconditionwherethespacebetweenthe
pipesiswiderthanthegroutexpansionrange. Insandy
soil andweatheredsoil, thebulbmay becylindrical.
Weassessedfivecasescomprisingvariousconditions
to model thepre-reinforced zone, as summarized in
Table7.
For thestrengthening of thegrout bulb reinforce-
ment, wefollowedthemethodof Kikuchi etal. (1995).
Thegrout injection consequently produced atenfold
increaseinthestiffnessof theweatheredsoil. Table8
shows the equivalent design parameters for the var-
ious compositions of the ground, the bulb, and the
steel pipes. A precisely-simulatedmodel (Figure32a)
Table7. Summaryof equivalent designmethods.
Table 8. Equivalent design properties used for numerical
modeling.
Equivalent designproperties E
eq
(MPa) C
eq
(kPa)
Case1 1748.21 1065.04
Case2 107.35 91.21
Case3 117.81 141.22
Case4 315.22 216.73
Case5 490.35 588.42
iscomparedwithvariousequivalent stiffnessmodels
(Figure32b).
Figure 33 shows the vertical displacement at the
tunnel crown, the horizontal displacement at the
springline, and ground surfacesettlement. To obtain
thesevalues, weaveragedthefivenodeslocatedatthe
center andleft areasof thetunnel crownat adepthof
4.5mfromtheportal. TheDRM/DEM parameter is
hereindefinedas thefractionof thedisplacement of
thereferential model (DRM) to thedisplacement of
theequivalent model (DEM):
The precisely-simulated model is represented by a
valueof 0%.
Fromthevertical displacement results, case1and
case2givetheclosestresulttotheprecisely-simulated
model in weathered soil. However, general methods
tend to overestimatetheeffect of pre-reinforcement.
Although there is only a slight difference between
Case 2 and Case 3, Case 2 predicts a similar
horizontal displacementatthespringlinerelativetothe
58
Figure32. 3DFE analysismodel for comparisonof equiv-
alent model andpreciselymodeled.
precisely-simulated model and the design is rather
safe. Theequivalent model offers asatisfactory pre-
diction of the ground surface settlement, as the
DRM/DEMvaluerangesfrom0.55%to0.7%inCase
1andCase2.
Case 1 exemplifies a proper equivalent modeling
technique for simulating the pre-reinforcing mecha-
nismin residual soils. Case 2, a SPSS in which the
stiffnessof thebulbandsteel pipesarecoupledinpar-
allel andthenconnectedtothestiffnessof theground
in series, exemplifies a simple equivalent modeling
technique that predicts the vertical displacement at
thetunnel crown, thehorizontal displacement at the
springline, andthegroundsurfacesettlement.
Whentheground, grout bulbs, andsteel pipes are
regarded as an individual support system, the pre-
reinforced zone is not a series or parallel stiffness
systembut aseries-parallel compound stiffness sys-
tem. Thus, asmall degreeof stiffnesssupportreceives
thelargest stress; moreover, alinear combination of
largestiffness supports resists ground displacement.
The SPSS explains the failure mechanism of the
Figure33. Comparisonbetweenequivalent cases.
pre-reinforced zone: namely, theruptureof thesteel
pipes and thegrout bulb follows theyielding of the
ground. TheSPSS shouldbeusedinvarious studies
relatedtotheanalysisof pre-reinforcedtunnels.
59
6 EFFECT OF SPATIAL VARIABILITY
The mean value of the measurements is often used
for design parameters even if there is a noticeable
variation. Theeffect of thevariation of geotechnical
parameters on tunnel safety or deformation is rarely
studiedorconsideredusingastatistical concept.There
are two kinds of sources for variation in the design
parameters: spatial distributionanduncertainty.
Uncertainty means that the material property of
the soil/rock has a characteristic unreliability. In
reliability-based designs, theuncertainty of thegeo-
material is significant for a specific site character-
ization. There are three major geotechnical uncer-
tainties governing the variability of geoproperties:
inherent soil characteristics, measurement errors, and
transformationfallacies(Phoon& Kulhawy, 1999).
The spatial distribution of soil has been consid-
eredforecological andenvironmental modeling(J ury,
1985) andvariouscharacterizationmethodshavebeen
suggested by Cho et al. (2004). Likewise, the spa-
tial distribution of the geoproperties is important
for the mechanical behavior of underground struc-
tures surrounded by variable soils. The soil itself
is not an isotropic material, but an anisotropic and
non-homogenous material. These characteristics are
inducedby theformationprocess andgroundstress.
Spatial distributiontakesthemacroscaleintotherange
of interest.Thus, itisexpectedthat, amongothersoils,
weatheredresidual soilshavehighspatial variabilityof
their geopropertiesduetotheir originandweathering
process.
Theeffectof thespatial distributiononthegeotech-
nical parameters of tunnel deformation is studied
through numerical analyses based on statistical con-
ceptsasshowninFigure34. Thegeotechnical param-
eters that cause the largest deformation of tunnels
whenthegroundmaterial followstheMohr-Coulomb
model andtheexpecteddisplacement variationchar-
acteristicsfor eachgeotechnical designparameter are
presentedinthisstudy.
Thecoefficient of variation (COV) has long been
commonlyusedtoquantifythevariabilityof soil and
rock properties (Harr, 1987). TheCOV is definedas
thestandarddeviation() dividedbythemean(j) of
theparameter:
Eachmaterial property of theresidual soil canbe
regardedas anormal randomvariablethat has acer-
tainprobabilisticerror. Inparticular, ithasbeenshown
that thespatial distribution of thefriction angleand
unit weight follows anormal distributionparameter-
ized with the mean and COV (Lumb, 1966; Hoeg
& Murarka, 1974; Lacasse& Nadim, 1996; Low &
Figure34. Numerical model of spatial distributionof geo-
property(cohesion, Range=3R, COV=40%).
Table9. Representativecoefficient of variation (COV) of
thegeotechnical parameters.
Properties COV(%) Reference
ElasticModulus 1545 Harr, 1987; Phoonand
Kul-hawy,1999
FrictionAngle 2432 Schultze, 1972
Cohesion 4068 Schultze, 1972; Tan
et al. 2000
Tang, 1997; Phoon & Kulhawy, 1999; Tanit et al.,
2004). Thus, it is assumed herein that all geotechni-
cal parameters, i.e., constitutive components of the
Mohr-Coulombmodel, havenormal distributionchar-
acteristics.TherepresentativeCOVof thegeotechnical
parametersissummarizedinTable9.
Thebulk modulus (B) andshear modulus (G) are
importantmaterial parametersandcanbealternatively
changedbytheelasticmodulus(E)andPoissonsratio,
respectively.Thenormallydistributedelasticmodulus
generatedwithaspecific COV isconvertedtoabulk
modulusandshearmodulustosimulatethespatial dis-
tribution effect of the elastic modulus on the tunnel
deformation.
Thenumerical analysisresultsareshowninFigure
35(a). A COV of 20%shouldbeconsideredacritical
variation of theelastic modulus effect on thetunnel
deformation, asaCOV of morethan30%causesarel-
atively significant deformation variation. Theelastic
modulusisaninfluential geopropertyintunnel defor-
mation. Thelargest NDVR is morethan0.075when
COV =40%andrange=3R. Thedisplacement vari-
ationinducedbytheelasticmodulusvariationcanbe
predictedfromFigure35(a), dependingontherange
fromthetunnel center.
60
Figure35. Effect of spatially distributedgeoproperties on
tunnel deformation.
Theeffectof spatial distributioninthefrictionangle
ontunnel deformationis showninFigure35(b). The
NDVR at the tunnel crown and springline reaches
0.0770.078whenthevariationrangeis 3R andthe
COV of thefrictionangleis40%. Therefore, thefric-
tionanglevariability is themost high-rankingfactor
for thecalculationof deformationinthetunnel com-
pared with other geotechnical parameters when the
Mohr-Coulomb material model is used. Thegeneral
COV of thefrictionangleis small at 12%(Schultze,
1972). Therefore, arangeof 10%20%variation of
the friction angle is critical for tunnel deformation
characteristics.Itcanbeseenthatasthevariationrange
increases, theslopeof thecurvatureincreases.
The COV of cohesion is near 40% (Fredlund &
Dahlman, 1972).AsshowninFigure35(c), theNDVR
enlarges inall cases as theCOV increases. However,
it canbeobservedthat thevariationof cohesionhasa
minor effect ondeformationinthetunnel. Themaxi-
mumNDVRinducedbythecohesionvariationislower
than0.00006whenthevariationrangeis 3R andthe
COV is40%. Fromtheanalysisresults, thespatial dis-
tributioneffect of cohesionontunnel deformationis
smaller than that of theelastic modulus and friction
angle. Therefore, thespatial distribution of cohesion
is not acritical parameter for thecharacterizationof
tunnel behavior.
7 CONCLUSIONS
Wepresentedanoverviewof geotechnical aspects of
undergroundconstructioninurbanareaswheremostly
decomposed residual soils are present, focusing on
mechanical properties, apparent earthpressure, effect
of ground water, and effect of spatial variability in
geotechnical properties.
The strength and deformation characteristics of
residual soils areaffectedby particle/poresize, fines
content, mineralogy, andunsaturation, amongothers:
Theinternal frictionangledecreaseswithanincrease
of fine content and partial saturation increases the
strengthof unsaturatedresidual soilsduetocapillarity.
Results of back analyses fromactual measurements
show that geotechnical properties of residual soils,
which are commonly used for design and modeling
inKorea, areinappropriateinthisregard, thusdemon-
stratingtheimportanceof theobservational methodin
tunnel engineering.
As an unsaturated soil is re-saturated, its appar-
ent cohesion can beeliminated. Thus, during tunnel
construction, cohesion loss can be induced by re-
saturation(e.g., seepagehindrance, drainageclogging
andgroundwaterchange)andmayresultintunnel face
instability.
The lateral earth pressure coefficient of residual
soils is smaller than the coefficient commonly sug-
gested in the literature (e.g., 7080% of Pecks
suggestion).
Unexpectedgroundwaterinflowandseepageforces
often cause tunnel failures during construction.
Several case histories suggest that tunnel face col-
lapses always occur alongwithseepageaheadof the
tunnel face. Analytical results suggest that the total
61
support pressureis littleaffectedby thetunnel depth
and increases significantly with an increase in the
groundwater level ratio. Asthetotal support pressure
isrelatedtothetunnel facestability, theseepageforce
seriouslyaffectsthetunnel facestability.
Particle transport characteristics of residual soils
mightbeamongthefactorsthatresultintheinstability
of undergroundstructures.Their phenomenaarecom-
plicatedandareinvolvedinerosionversusself-healing
(redeposition) processes.
Whilegroutabilityisaffectedbytheporesizeof the
residual soils, proper selectionof agroutingmethod
has proved a difficult task in granite residual soils.
When seepage problems are anticipated during tun-
nel construction, proper groutingaroundtunnels can
achieveeffectivereductionof seepageforceactingon
theshotcreteliningandcanincreasethestiffnessand
strengthof thesurroundingground.
An analysis method combining experimental
and numerical procedures that consider the time-
dependent effect onthepre-reinforcedzoneontunnel
behavior will provideareliableandpractical de-sign
basisandmeansof analysisfor tunnelsinsoftground.
It is expected that, among other soils, weathered
residual soilshavehighspatial variabilityof geoprop-
erties, becauseof their originandweatheringprocess.
The numerical results show that tunnel deformation
increaseswithanincreaseinthespatial variabilityof
geotechnical designparametersandisacceleratedwith
anincreaseintunnel size.
Research on decomposed residual soils has been
conductedthroughout thelast fewdecades. Nonethe-
less, thecurrentincreaseof undergroundconstruction
projects inurbanareas requires better understanding
of residual soilsfor safer undergrounduse.
ACKNOWLEDGEMENTS
ThisstudywasfundedbytheKoreaInstituteof Con-
struction and Transportation Technology Evaluation
andPlanningunder theMinistry of Constructionand
TransportationinKorea(Grant No. 04-C01).
REFERENCES
Barisone, G., Pigorini, B. & Pelizza, S. 1982. Umbrellaarch
method for tunnelling in difficult conditions Analysis
of Italiancases. Proceedings of the 4th Congress Interna-
tional Association of Engineering Geology. NewDelhi, 4:
1527
Cho, G.C., Lee, J.S. & Santamarina, J.C. 2004. Spatial vari-
abilityinsoils: Highresolutionassessmentwithelectrical
needleprobe. Journal of Geotechnical and Geoenviron-
mental Engineering. ASCE 130(8): 843849
Cho, K.H., Choi, M.K., Nam, S.W. & Lee, I.M. 2006.
Geotechnical parameterestimationintunnelingusingrel-
ativeconvergencemeasurement.International Journal for
Numerical and Analytical Method in Geomechanics 30:
137155
Fredlund, D.G. & Dahlman, A.E. 1972. Statistical geotech-
nical properties of glacial lake edmonton sediments, in
Statistics and Probability in Civil Engineering. Hong
KongUniversity Press, distributedby OxfordUniversity
Press. London.
Harr, M.E. 1987. Reliability Based Design in Civil Engineer-
ing, Dover Publications, INC.
Hoeg, H. & Murarka, R.P. 1974. Probabilistic analysis
and design of aretaining wall. Journal of Geotechnical
Engineering Division. ASCE 100(3): 349366
J ury, W.A. 1985. Spatial variability of soil properties.In:
Hern, S.C., Melancon, S.M. (Eds.), Vadose Zone Modeling
of Organic Pollutants, Lewis, Chelsea, MI, 245269
Kikuchi, K., Mito, Y. & Adachi, T. 1995. Case study on
themechanical improvement of rockmassesbygrouting.
Rock Foundation. Balkema393398
Kim, J.S., Choi,Y.K., Park, J.H.,Woo, S.B. &Lee, I.M. 2007.
Effectof viscosityandcloggingongroutpenetrationchar-
acteristics.Journal of Korean Geotechnical Society 23(4):
513
Kim, Y.J. 1993. Mechanical characteristics of decomposed
Koreangranites. Ph.D. Thesis, Seoul: KoreaUniversity
Lacasse, S. & Nadim, F. 1996. Uncertaintiesincharacteriz-
ingsoil properties.Uncertainty 96, Geotechnical Special
Publication, ASCE 58(1): 4975
frictional material. Geotechnique 40(4): 581606
Lee, C.K. & J eon, S.K. 1993. earthpressuredistributionon
retainingwallsintheexcavationof multi-layeredground,
Journal of Korean Geotechnical Society 9(1): 59
Lee, I.K. 1991. Mechanical behavior of compacteddecom-
posedgranitesoil. Ph.D. Thesis, London: CityUniversity.
Lee, I.M., Moon, H.P., Lee, D.S., Kim, K.R. & Cho, M.S.
2007. Earth pressureof vertical shaft considering arch-
ingeffect inlayeredsoils. Journal of Korean Tunnelling
Association 9(1): 4962
Lee, I.M., Nam, S.W. & Ahn, J.H. 2003. Effect of seepage
forces on tunnel face stability, Canadian Geotechnical
Journal 40: 342350
Lee, I.M., Park, Y.J. & Reddi, L.N. 2002. Particletransport
characteristicsandfiltrationof graniticresidual soilsfrom
theKoreanpeninsula.Canadian Geotechnical Journal 39:
472482
Lee, I.M., Sung, S.G. &Cho, G.C. 2005. Effectof stressstate
ontheunsaturatedshear strengthof aweatheredgranite.,
Canadian Geotechnical Journal 42: 624631
Lee, S.J., Lee, S.R. & J ang, B.S. 2002. Unsaturated shear
strengthcharacteristicsof weatheredgranitesoils.Journal
of Korean Geotechnical Society 22(1): 8188
Low, B.K. & Tang, W.H. 1997. Reliability analysis of rein-
forcedembankmentsonsoft ground. Canadian Geotech-
nical Journal 34(5): 672685
Lumb, P. 1966. The variability of natural soils. Canadian
Geotechnical Journal 3(2): 7497
Peck, R.B. 1969, Deep excavations and tunneling in soft
ground, Proc., 7th Int. Conf. on Soil Mech. And Found.
Engrg., State-of-the-Art Rep., State-of-the-Art Vol:
225290
Phoon, K.K. & Kulhawy, F.H. 1999. Characterization of
geotechnical variability. Canadian Geotechnical Journal
36(4): 612624
62
Schultze, E. 1972. Frequency distributions andcorrelations
of soil properties. In Statistics and Probability in Civil
Engineering, HongKongUniversityPress, distributedby
OxfordUniversityPress, London.
Shin, J.H., Lee, I.K., Lee, Y.H. & Shin, H.S. 2006. Lessons
fromserial tunnel collapses during construction of the
Seoul SubwayLine5.Tunnelling and Underground Space
Technology 21: 296297
Shin,Y.J. 2007. Elasto-plasticgroundresponseof underwater
tunnels consideringseepageforces, Ph.D. Thesis, Seoul:
KoreaUniversity
Song, K.I., Kim, J. & Cho, G.C. 2007. Numerical anal-
ysis of pre-reinforced zones in tunnel considering the
time-dependentgroutingperformance. Journal of Korean
Tunnelling Association 9(2): 109120
Tanit, C. 2004. Reliability-based design for Internal sta-
bility of mechanically stabilized earth walls. Journal of
Geotechnical and Geoenvironmental Engineering. ASCE
130(2): 163173
Yoo, C. 2001. Behavior of bracedandanchoredwallsinsoils
overlyingrock. Journal of Geotechnical and Geoenviron-
mental Engineering. ASCE 127(3): 225233
63
General reports
Safetyissues, riskanalysis, hazardmanagement andcontrol
C.T. Chin& H.C. Chao
Moh and Associates, Inc., Taipei, Taiwan
ABSTRACT: Theoccurrenceof hazardeventsingeotechnical practiceisoftenassociatedwithgeotechnical
uncertainties. Toprevent hazardeventsfromhappeningor reducetheimpactsof their consequence, sourcesof
geotechnical uncertainties needtobeidentifiedandtreated, andappropriatecontrol measurehas tobeimple-
mentedthroughout aproject. Theprobabilistic methodsarethedevicesdealingwithuncertaintiesandtherisk
managementisthetechniquetofacilitateachievingthegoal of projectsafety.Thepaperspresentedinthistheme
aresummarizedandbrieflydiscussedinthreecategories riskanddecisionanalysis, geotechnical control, and
analysisandcontrol of groundresponse. Inriskanalysis, geotechnical uncertaintiesneedtobetreatedexplicitly.
Resultsof riskanalysis, qualitativelyor quantitatively, not onlyprovideabaselinefor decisionmakingbut also
insightstotheproblemof concern. Inorder tomakebest useof availableanalysistools, morecasestudiesare
needed.
1 INTRODUCTION
1.1 Geotechnical uncertainty
Uncertainty is thelack of certainty, astateof having
limited knowledge where it is impossible to exactly
describe existing state or future outcome (Hubbard,
2007). If uncertainty requires tobetreated, probabil-
ity is thedevice. Therearetwo kinds of probability:
relative frequency and subjective, degree-of-belief
probability.Theprobabilityof anuncertaineventisthe
relativefrequency of occurrencewhen it is obtained
throughrepeatedtrialsor experimental sampling. The
subjective, degree-of-belief probability comes from
judgment wheretheprobabilityof anuncertainevent
isthequantifiedmeasureof onesbelief orconfidence,
accordingtotheinformationavailableandonesstate
of knowledgeat thetimeit isassessed(Vick, 2002).
In engineering practice, although uncertainty is
surelyacertaintyandmaycreatesomeimpact,itseems
theexistenceof which is not abother to many. As a
matter of fact, most of theengineers get alongquite
well without explicitly using the necessary device,
probability, to manageit. This probably attributes to
thecustomary practiceengineers have. By using the
establishedstandards, codes, factorsof safety, design
criteria,orproceduresinwhichtheuncertaintiessome-
how have been accounted for or somewhere hidden
behind, deterministic method by which answers are
either correct or wrongcancomfortablybeappliedin
solvingproblems.
However, the material in geotechnical engineer-
ing is widely known for its significantly abundant
Figure 1. Categories of uncertainty in soil property data
(Christianet al. 1994).
uncertainties. Sources of uncertainties can be
unknownpresenceof geologicdefects,uncertainvalue
of soil properties, limited knowledge to the mech-
anisms and processes, and much more. One of the
examples as presented by Christian el al. (1994) is
demonstratedin Figure1, in which theuncertainties
associated with the characterization of soil proper-
ties areattributeto systematic error anddatascatter.
Systematicerror resultsfromstatistical error inmean
valuearisingfromlimitednumbersof measurements
or bias inmeasurement procedures likethoseassoci-
atedwithfieldpermeabilitytestor fieldvantest. Data
scatter, ontheother hand, isaresultof randomtesting
errorsor actual spatial variationinthesoil profile.
Another sourceof geotechnical uncertaintyismod-
eling. A model is an appropriate simplification of
reality. Goodmodelingskill isreflectedintheability
toidentifytheappropriatelevel of simplification to
recognizewhatfeaturesareimportantandwhatarenot.
67
Very often engineers are unaware of the simplifica-
tions that they have made and problems may arise
preciselybecausetheassumptionsthathavebeenmade
are inappropriate in a particular application (Wood,
2004).
Thetypesof geomechanical model canbedivided
into thefollowingcategories: empirical model, theo-
retical model, analytical model, numerical model and
constitutive model. Model uncertainty is the extent
to which amodel incarnates auniquely correct rep-
resentation of the physical process it seeks to emu-
late. Model uncertaintiesarisefromitsrepresentative
degreeto thereal fieldprocesses, andfor aphysical
processdifferent modelsandoperationcanalwaysbe
found.
In conventional approach, effects of uncertainty
are generally accounted for, consciously or not, by
standards, codes, designcriteria, or factors of safety.
Theunderlying uncertainties may havealready been
consideredorevaluatedsomehowintheprocessdevel-
oping them. This simple strategy seems work quite
well forworkshavingbeenencounteredbeforeandlot
of experience accumulated. However, when circum-
stances areuniqueand uncertainties arenot routine,
whichareoftenthecasesingeotechnical engineering
practice, aforementionedproceduresor toolsmaynot
befullyapplicableanymore.
1.2 Geotechnical safety
Theoccurrenceof hazard event in major geotechni-
cal engineering projects such as tunneling or deep
excavationoftendrawsalotof attentionfromthegen-
eral public. Potential consequencesincludesignificant
financial loss to the client and contractor, schedule
delayandlossof confidencetothegeneral public, and
casualty. Thus, itisobviousthatsafetyisanimportant
issuethatrequiresspecial carestoensuretheobjectives
of aproject canbeachieved.
Safetyitself isinessencenotameasurablequantity.
Inpractice,itisevaluatedthroughthesafetyindicators.
Someof thephysical characteristicssuchasthesizeof
cracks, deformationsanddifferential settlements, are
selected to serve as safety indicators for evaluating
the safety status. This process requires analysis and
interpretation, and judgment plays an important role
init. If theuncertainty istobeaccountedfor andthe
effectsareaccommodated,sourcesof uncertaintymust
beunderstoodfirst. This is whereprobability comes
in. One of the most important tools in dealing with
probability is judgment. J udgment provides aninter-
pretativeframeworkthathelpsguidehowuncertainties
arecomprehendedandsubsequently managed. When
astructureis saidto besafeduringadjacent excava-
tion, itmeanstheassessorholdsomesufficientdegree
of belief.
In past decades, there has been a greater aware-
ness on need to treat and manage the geotechnical
Figure2. Geotechnical safety.
uncertaintyinarational andexplicit way. Inresponse
to this demand, active discussion and development
have been seen on topics that may be grouped into
fivecategories: soil variabilitycharacterization, relia-
bility analysis, newgenerationdesigncode, observa-
tional method, andrisk assessment andmanagement.
Advancementsinthesedisciplines, ontheother hand,
areintendedto address better issues ongeotechnical
safety. Theseconceptsareconveniently illustratedby
Figure2.
1.3 Risk assessment
Traditionally, risks weremanaged indirectly through
the engineering decisions taken during the project
development.Moreoftenthannot,theinformationand
knowledge used behind the decisions are inexplicit
and not easy to trace. To improve this, it is neces-
sary tointroduceformal risk management technique.
Inthispractice, riskisdefinedastheproductof failure
probabilityandconsequences:
Where, P
f
is probability of failure and C
f
is conse-
quencesof failure.
By definition, risk assessment must account for
both the probability and extent of adverse conse-
quencesof hazardeventsarisingfromagivenactivity.
Thus, risk assessment involves identificationof haz-
ardeventsandqualitativeor quantitativedescriptions
of risks. Thescopeof riskanalysiscontainstheentire
process of the causes and effect of adverse events.
Suchaprocessiscomposedof threesequential parts,
an initiator that starts it, response to the initiator
and the consequences (Vick, 2002) as illustrated in
Figure3.
Initiatoristhecausethatsetsapotential failurepro-
cessinmotion. Responseistheeventdirectlyresulted
fromthe initiator, and also called the hazard event
in some occasions. Response leads to failure if the
68
Figure3. Componentsof riskanalysis(Vick, 2002).
resistanceor capacity of astructureis unableto sus-
tain theeffects. In risk analysis, failureis treated as
consequenceandneedstobeaphysically observable
conditionlikeretainingwall collapse, utility pipeline
broken, roadwaydamageet al. Generally, failurecan-
not be evaluated meaningfully if it is not converted
to something measurable such as lives lost or dol-
lar cost. It is also important to realize there are
uncertaintiesassociatedwiththeresponse. Asshown
in Figure 3, probabilistic methods such as Bayes
approach, reliabilityanalysis, subjectivejudgmentand
othersareintroducedasthetoolsfor evaluatingthese
uncertainties.
As shownalso inFigure3, initiator, responseand
consequence are linked together by decomposition
techniquessuchaseventtreeandfaulttree. Eventtree
and fault treefacilitateenvision how failureprocess
occurs. As such, they are basic tool of risk identi-
fication and analysis. Figure 4(a) shows a two-level
of event tree and fault tress. The event starts with
initiator event I. If it occurs, the next event to hap-
pen is R
1
, followed by R
2
or R
3
or both. This event
tree contains two failure modes IR
1
R
2
and IR
1
R
3
.
Figure 4(b) presented the same failure progress by
Fault tree. Event uses a bottomup or forward
approach beginning with the initiator and taking it
to consequences. This captures failure process by
expressing its logical order. Fault tree, on the other
hand, usesatopdownorbackwardapproachstart-
ing fromthe top event and identifying the possible
eventsinthe2ndlevel throughasearchprocess. The
and andor logic gatesinstrumentedinFault tree
allowcomputations usingBooleanalgebraincoping
withcomplexproblem.
Figure4. Event treeandFault treeformat.
After riskisidentified, therestof workinriskanal-
ysis is to evaluate the probability and consequences
suchthat thecorrespondingrank-order canbedeter-
mined based on predefined criteria. Subjective,
degree-of-belief probability often areassessed using
expert investigation method incorporated with other
techniquessuchasDephi method, AnalyticHierarchy
Process(AHP), Fuzzysettheoryinriskanalysis. One
of thecontroversial issuesariseshereintheprocess
quantification of risks. According to Vick (2002),
qualitative approaches allows us to discern that one
thingismorelikelyor lesslikelythansomeother, but
animportantproperty byhowmuchmoreorless in
thequalificationprocessismissing. Other thanit, the
depthof theinsight totheproblemcanbereachedis
consideredinproportiontotheeffortsmadeinthepro-
cess of risk analysis wherenumerical quantification
usuallyrequiresmore.
Risk responsemeasureandreductionstrategy can
bedevelopedbasedonriskacceptablecriteriadefined
in risk policy (ITA, 2004). For risks with high rank-
order, itmaybenecessarytodevelopdesignoralterna-
tivesindeterminingriskreductionmeasure. Decision
analysis is essentially acomparativeapproach based
ondecisionrules andresults of risk analysis onvar-
ious alternatives. Similar techniques in risk analysis
69
0 1 2 3 4
Factor of safety, FS
P
r
o
b
a
b
i
l
i
t
y

D
e
n
s
i
t
y

F
u
n
c
t
i
o
n
FS=1.50
FS=2.00
(P
f
)
1
=0.6%
(P
f
)
2
=7.4%
Design alternative 1:
low uncertainty
Design alternative 2:
high uncertainty
Figure5. Effect of uncertainty.
can be applied to evaluate the risk associated with
each of the alternative. By comparing the results of
risk analysis, thealternativehaving thegreatest risk
effectiveness can be identified. Results of decision
analysis providea solid baselinefor decision maker
tomakedecisionwhendealingwithrisk.
1.4 Deterministic thinking versus probabilistic
thinking
Inengineering, adeterministicsystemisasystemthat
givenaparticularinput,itwill alwaysproducethesame
output, andtheunderlyingmachinewill always pass
through thesamesequenceof states. In this system,
results of thinking areeither right or wrong, mathe-
matic algorithms used are either correct or not, and
thereisnoneedtobebotheredbyuncertainty.
However,uncertaintyinfactexistsandtheeffectsof
whichmaybesignificant inmanyoccasionsandcan-
notjustbeneglected.Ashasbeenmentionedearlierin
this report, uncertainties ingeotechnical engineering
practicearemanagedthroughstandards, codes, design
criteria, factors of safety or established procedures.
Withsuchstrategies, thedeterministicthinkingworks
and functions well for situations where experience
androutineworksdominate. For uniquesituationsor
non-routineconditionsor inriskanalysis, uncertainty
becomes an unavoidableissuethat requires using of
the technique of probabilistic methods in obtaining
appropriatesolution.
Figure5demonstrates theeffect of uncertainty in
design results for two alternatives based on differ-
ent design models. The computed factors of safety
for alternative1andalternative2are1.50and2.00,
respective. In a deterministic system, the factor of
safety for alternative2is apparently higher, andthis
alternativeisalmost certaintobeselectedasthefinal
scheme. However, if theuncertaintiesassociatedwith
bothmodelsareconsidered, thefailureprobabilityof
alternative2is apparently higher eventhecomputed
factor of safetyseemsbetter.
Table1. Riskanddecisionanalysis.
ID Topic
IS-022 Riskanalysisandfuzzycomprehensive
assessment onconstructionof shield
tunnel inShanghai metroline
IS-294 Riskassessment onenvironmental impact in
XizangRoadTunnel
IS-293 Riskanalysisfor cutterheadfailureof composite
EPB shieldbasedonfuzzyfault tree
IS-070 Riskassessment for thesafegradeof deep
excavation
IS-055 Multi-factorsdurabilityevaluationinsubway
concretestructure
IS-383 Researchonstructural statusof operatingtunnel
of metroinShanghai andtreatment ideas
2 REVIEWOF PAPERS
2.1 Risk and decision analysis
As listed in Table 1, the papers related to risk and
decisionanalysisaregroupedintothissection.Among
thesepapers, IS-022, IS-294andIS-293discuss tun-
nelingriskanalysis, andIS-070introducesriskanaly-
sis for deepexcavation, andIS-055is about decision
analysis. Whiletheothers apply formal risk or deci-
sion analysis techniques in the evaluation approach,
theapproachpresentedinpaper IS-383canbetaken
as an informal risk management approach. As have
beenusedwidelyingeotechnical practice, risksasso-
ciated with the identified hazard events were not
presented. Nevertheless, thecauses of hazard events
were identified and the hazard prevention measures
weredeveloped.
Paper IS-022presentedarisk analysisapproachin
whichriskswereidentifiedthoroughthework break-
downandfaulttreemethod.Theassociatedrank-order
of theidentifiedriskswereanalyzedbyfuzzycompre-
hensive evaluation method. The expert investigation
method and AHP were used to determine the risk
indicators weight, evaluatetheprobabilitiesandcon-
sequences of the identified risks. The membership
function was applied to determine the membership
degreevalueof eachof theriskevents.Thecaseinves-
tigated in this paper was Shanghai metro, which is
a32.2kmlongshieldtunnel. For theconvenienceof
risk analysis, themetro linewas dividedinto 12sec-
tions. Resultsof theevaluationshowedtheidentified
riskeventsincludeobstruction, tunnel collapse, quick-
sand, groundwater ingress, ground settlement, et al.
The risk levels with respect to each of the sections
werefrommediumtosignificant. Thepossiblefinan-
cial lossisbetween100,000RMBto10,000,000RMB.
Basedontheriskclassificationcriteriaof thiscase, the
consequencesaregreat but compensable.
70
Paper IS-294 presented the risk analysis for the
constructionof Shanghai XizangRoadTunnel, ariver-
crossing tunnel built for the 2010 Shanghai Expo-
sition. Other special features included up-cross and
down-cross existing metro lines. Risk analysis was
based on expert investigation and confidence index
methods. The computer software TRM 1.0 devel-
oped by theTongji University was used as the tool
to analyzetherisk dataand thereafter determinethe
correspondingrisk level. Inthecaseinvestigated, the
work of risk analysis was focused on the environ-
mental impacts as a result of tunnel construction to
thesurroundingbuildings, roads, andutilitypipelines.
Basedontheanalysisresults, riskmitigationmeasures
focusedongroundsettlement control wereproposed.
Theauthorsalsopointedoutthattheriskisdynamicin
nature, whichrequiresconstantandcycledassessment
andtreatment.
Paper IS-293 presented for tunneling in adverse
groundcondition, thecutterheadof shieldmachineis
exposed to high risk of failure as a result of heavy
demand in machine operation. This paper presented
arisk analysis approach for cutterhead failureusing
the fault tree analysis method and fuzzy set theory.
In this study, risk identification was conducted by
thefault treeanalysis method. Results of theidenti-
fication indicated three major hazard events cuter
disk failure, cutter failure and other systemcompo-
nents failure will leadto cutterheadfailure. Expert
investigationmethodandfuzzyset theorywereintro-
ducedtoevaluatetheprobabilitiesandconsequences
of theoccurrences for thesehazardevents. Basedon
theresults of risk analysis, remediationmeasures for
thehazardeventswithhigherrisklevel weredeveloped
andpresented.
HistoricdatainpaperIS-070showedtheproportion
of accidentsindeepexcavationprojectsresultedfrom
designandconstructionrelatedproblemsis87%based
on344cases investigated. Becauseeffectivetools in
dealingwithgeotechnical uncertaintyandapplicabil-
ityof currentanalytical theoriesarelimited, engineers
tendtobeconservativeconsciously or not indealing
withgeotechnical problems. This is usually reflected
in high construction costs. To improvethis, aFuzzy
synthetic evaluation process was used in this study
for risk analysis. Thismethodfollowstheprocedures
in which the risk factors were identified first, and
expert investigationmethod, Analytic Hierarchy Pro-
cess(AHP) methodandDelphi methodwereadopted
todeterminetheweightfor eachof thefactorsandthe
safetyindex,andthesafetyrankingorderof anexcava-
tionproject canthusbedecided. Thecasevisitedwas
theShanghai international passenger transport center
project. The size of the excavation area, hydrogeol-
ogycondition, designfeature, constructionaspectand
surroundingenvironment areselectedas theprimary
factorsaffectingthesafetyof deepexcavation. Results
of the evaluation indicated that the level of safety
of the project as a whole was acceptable. However,
size of excavation area and hydrogeology condition
wereidentifiedasthefactorshavingmoresignificant
impactonthesafetyof theexcavationactivities. Mea-
sures to prevent hazard events fromhappening were
thusdevelopedbasedontheresultsof riskassessment.
Paper IS-055 presented a decision analysis
approach for determining the optimal design alter-
native for the concrete mixture of a subway tunnel.
Factors having significant influence on the durabil-
ity of subway concrete structure were identified to
be stray current corrosion, chloride ions ingression,
sulfate attack and carbonation. While the durability
attribution fromeach of the factors was determined
individually, thecombinedinfluenceof thesefactors
wasvague. Toobtaintheoptimal designscheme, five
designalternativesweredeveloped. Expert investiga-
tion method, AHP and Multiple Attribute Decision
Model incorporatedwithfuzzysettheorywereapplied
inassessingthejointinfluenceof thesefactors.Results
of the evaluation provided a baseline for decision
maker inselectingtheadequatedesignscheme.
PaperIS-383presentedthehealthdiagnosticresults
and hazard prevention or reduction measures for an
operating tunnel of Shanghai metro. Results of the
diagnosisshowedthetunnel structureissufferedfrom
detrimental events such as ingress of groundwater,
crackandconvergenceof tunnel lining, andsettlement
alongthetunnel alignment. Thecausesof thesedetri-
mental eventswereidentifiedtobeadversegeological
condition, sequela of the accidents during construc-
tion, local groundsubsidence, constructionactivities
in the proximity, etc. To prevent these detrimental
events fromgetting worseand reducetheassociated
riskstothetunnel andsurroundingenvironment,metro
passengersandthirdparties, control measuresinterms
of regulation, monitoringschemeswereproposed.
2.2 Geotechnical control
The geotechnical control is a process exercised
throughout the planning, design and construction
phasesof aprojecttofacilitateachievingprojectobjec-
tives. The control measures applied include regular
audit inspection, site supervision and risk manage-
menttoensureadequatedesignstandardsandeffective
safety protection beperformed such that theproject
can be completed in a manner of optimization. As
listedinTable2, twoof thepapersweregroupedinto
thissection.Thepaper IS-372introducedametrorail-
wayunderconstructionandthepaperIS-380presented
a retrospect study of a cable duct crossing project.
Bothof thecasesreportedwereconstructionprojects
inHongKong.
Paper IS-372 presented a geotechnical control
process exercised for a HK$8.3 billion subway
71
Table2. Geotechnical control.
ID Topic
IS-372 Geotechnical control of amajor railwayproject
involvingtunnel worksinHongKong
IS-380 PerformanceReviewof aPipeJ ackingProject
inHongKong
construction project in Hong Kong. This project is
packedintothreedesign-builtcontracts.Thecontentof
theprojectincludestheconstructionof shieldtunnels,
cut-and-cover tunnels and underground station. The
Geotechnical EngineeringOffice(GEO), Civil Engi-
neering and Development Department of the Hong
KongSpecial AdministrationRegion(HKSAR) is in
chargeof controllingthebuildingordinanceandreg-
ulations and issuing technical standards. Under of
auditingof GEO, theprivateowner of thisproject, the
KowloonCantonRailway corporation(KCRC) com-
mitted to follow theJ oint Codeof Practicefor Risk
Management of TunnelingWorksfor implementation
of risk management process. The geotechnical con-
trol process startedfromtheplanningstageandwill
be kept in effect throughout the construction stage.
Because the project is considered a private project,
thetunnel worksmaybeexemptedformtheadminis-
trativeprocedures. Thus, areviewpanel was formed
withinGEOintheplanningstage.TheKCRCdemon-
strated to GEO they had met the requirements for
instrument of exemption (IoE) in aspects including
risk management, design, and construction. TheIoE
wasissuedafter thedocument met withthespecified
requirements. Under the IoE, the KCRC is required
to appoint authorized personnel, employ assurance
system and control scheme, prepare site supervi-
sion plan, and keep appropriate records and reports
for regular GEO inspection during the construction
stage. Monthly meeting arescheduled between rele-
vantparties.Whensignificantchangesoccurindesign
or working methods, the KCRC needs to report to
GEO. The geotechnical control process is currently
inprogress.
Paper IS-380 review the performance of a cable
ductcrossingconstructionprojectinHongKong.With
theproject, thegeotechnical control processwasexer-
cised by the GEO, HKSAR. This project involved
theconstruction of a222mlong and 1.95mdiame-
ter tunnel to serveas cableduct crossingahighway,
two MRT tunnels, and twoAirport express link tun-
nels.Thiscableductwasconstructedwithpipejacking
method.Priortothecommencementof works,acondi-
tionsurveywasconductedforexistingutilitypipeline.
Duringconstruction, anautomaticmonitoringsystem
wasusedtotakereadingsfor monitoringrailwaytrack
settlement. Theworkof monitoringwascontrolledby
asystembasedonalert level, actionlevel andalarm
level management.Specifiedresponsescorresponding
toeachlevel of thesystemwerealsodefinedbeforethe
constructionstarted. Duringtheconstruction, aclose
to action level reading on the track settlement was
recorded. Under theframework of geotechnical con-
trol process, an urgent meeting was held among the
client, design and contractor. After the meeting, the
type of lubricant used for filling the gap between
thetunnel liningandsurroundingsoilswasreplaced.
Withthecontrol process, themaximumsettlement of
railway tracks was controlled within the maximum
allowable range throughout the whole construction
phase.
2.3 Ground response, analysis and design
Most of theengineeringdesignsarecomposedof two
major components determining the imposed load
and computing the resistance or capacity. Factor of
safety in a design problem is defined as the ratio
between the resistance and load. Because most of
thegeotechnical engineering projects areperformed
underground, theconstructionactivitiesinvolvingadd
or remove loads to the ground will inevitably result
in stress and strain redistribution. When theresulted
stresses exceed the resistance or capacity of the
ground,failurecouldoccur.However,groundresponse
beforereachingfailurestatemay drawmoreconcern
for underground construction project. In urban area,
underground construction induced ground deforma-
tion make the construction itself and the buildings,
utilitypipelinesandinfrastructuresadjacenttothecon-
struction siteexposed to therisk of being damaged.
Resultsof groundresponsepredictionandmonitoring
canbeusedas abaselinefor developingcontrol and
protectionschemes.
Thepapersinthisgrouparefurtherdividedintotwo
sub-groups groundresponse, analysis anddesign
for discussionpurposes.
2.3.1 Ground response
Table3liststhepapersrelatedtotheissueof ground
response. Paper IS-048 introduced a prototype and
laboratory scale non-destructive scanning technique
designed for detecting crack or cavity ahead of the
frontendof ashieldmachine. Paper IS-369presented
an analytical approach for evaluating the squeez-
ing potential of soft rock. Paper IS-247 presents a
numerical investigation for the floor heave behavior
at theT-sectionof adeepminingtunnel usingthree-
dimensional finiteelementmethod. Paper IS-376dis-
cusses the application of strain gauge in measuring
strut load in deep excavation project in Singapore.
Suggestionsfor maximizingitseffectivenessarealso
proposed. Paper IS-014andIS-339present theappli-
cation of fuzzy set theory and neural network in
predictinggroundsettlement inducedbytunneling.
72
Table3. Groundresponse.
ID Topic
IS-048 Experimental studiesof ageological measuring
systemfor tunnel withultrasonictransducer
IS-369 Squeezingpotential of tunnelsinclaysand
clayshalesfromnormalizedundrainedshear
strength, unconfinedcompressivestrength
andseismicvelocity
IS-247 Floor heavebehavior andcontrol of roadway
intersectionindeepmine
IS-376 MaximisingthePotential of StrainGauges:
A SingaporePerspective
IS-014 Predictionof surfacesettlement inducedby
shieldtunneling: anANFISmodel
IS-339 Theuseof artificial neural networkstopredict
groundmovementscausedbytunneling
Paper IS-048 presented an experimental model
developed for detecting multiple reflection sources
basedonrotational scanningtechniqueandultrasonic
wavereflectionmethod. Intheproposedtest config-
uration, theground was simulated by plaster blocks,
amongwhichthemaximumdimensionisof 1000mm
in height, 1200mmin width and 150mmin thick-
ness. Horizontal andinclinedcracksaresimulatedby
drilledholes. Various signal process techniques such
as stacking, signal compensation, and demodulation
wereappliedto obtainultrasonic imageof thespace
aheadof thescanningtransducer. Resultsof theexper-
imental evaluationshowedthat theproposedmethod
is ableto identify thelocation of multiplereflection
sources.
The tunnel squeezing phenomenon was first
describedbyTerzaghi (1946) whoassociatedsqueez-
ing mainly with clayrich rocks. One of the first
stability criteria to predict squeezing was developed
by Peck (1969) for tunnels in clays based on Broms
and Bennermarks (1967) stability criteria for open
excavation. For tunnel in rocks, most of thesqueez-
ingcriteriaproposedareempirical or semi-empirical
suchas Singhet al. (1992), Goel et al. (1995, 2000),
J ethwael al. (2000) andmore. Themainchallengein
useof thesesemi-empirical approachesisinthedeter-
minationof therock massstrength. Inaddition, most
of theproposedmethodologiesdevelopedthusfar are
mainlyfor claysor hardrocks. Fewstudieshavebeen
proposedfor intermediatematerial suchashardsoils
or soft rocks.
In paper IS-369, simple methods to evaluate the
squeezingpotential of intermediatesoil-rockmaterial
based on undrain shear strength, unconfined com-
pressivestrengthandP-wavevelocity was proposed.
Evaluation of field measurement and other empiri-
cal tunnel squeezing criteriawas also performed for
comparativepurpose.
For largecrosssectiontunnel, thegrounddeforma-
tion resulted fromexcavation work is usually much
moresignificant at thetunnel intersectionthanat the
regular part. Thus, special measures controllingpos-
sibleground movement that might bedetrimental to
the safety is necessary. Paper IS-247 presented an
investigationof thefloor heaveataT-sectionof aven-
tilationtunnel atthedepthof GL. 990mof Tongkou
colliery usingthree-dimensional finiteelement mod-
eling. Basedontheevaluationresults, suggestionsfor
controllingfloor heaveweregiven.
Instrumentation plays an important role in ensur-
ing that construction control is maintained during
excavation. Comparison between themonitored data
anddesignpredictions provides theopportunities for
verifying design results and refining design meth-
ods. Experiences indicate in most of the excavation
project, someof theinstrumentationreadingsaregen-
uineloadconditionswhilesomearenot. Itisimportant
togainclear understandingof theimpact of construc-
tionactivitiesanddatainterpretationandmakeeffort
tomaximizethequalityof thedata.
Accordingto paper IS-376, straingaugeis widely
used in measuring the change of strut load in deep
excavationinSingaporebecauseof theacceptablereli-
ability andlowcosts comparedto other instruments.
However, theperformanceof straingaugecaneasily
beaffectedbythelocationwhereitisinstalled, electro-
magneticinterference, temperature, preloading, weld-
ingandother constructionactivities. Tominimizethe
influence of these factors, measures such as use of
loadcell asacrossreference, protectionagainst con-
struction induced disturbance, and use of real time
systemwereproposed. For datainterpretation, skilled
personnel fullyawareof thedesignpredictionsfor the
excavation, the excavation process and the potential
impact of the excavation on the readings should be
assigned.
Duetothecomplexity andvariationof theground
composition, accuratepredictionof tunnelinginduced
ground settlement based on conventional geotechni-
cal approachessuchasempirical methods, analytical
methodsor numerical methodsisconsideredbymany
asachallenge. Empirical approachiseasy tousebut
oftenprecludestheconsiderationof groundresistance
anddeformability parameters. Analytical andnumer-
ical approach are generally using simplified ground
parameters for practical purposes rather than incor-
porating the complete ground conditions along the
wholerangeof atunnel alignment intothemodeling
approach. The applicability of available constitutive
models and simulation of three dimensional ground
responsewithin2dimensional spacefor groundset-
tlement predictionis controversial tomany. Analter-
nativeinsolvingtheproblemsmentionedaboveisthe
artificial intelligencetechnique. Artificial neural net-
work and fuzzy logic wereintroduced into this field
for predictioninrecent years.
73
Figure6. Conceptual model of aneural network withtwo
inputsandoneoutputs.
An artificial neural network (ANN), often just
called a neural network (NN), is a mathematical
model based on biological neural networks. A con-
ceptual model is showninFigure6. It consists of an
interconnected group of artificial neurons and pro-
cessesinformationusingaconnectionist approachto
computation. In most cases an ANN is an adaptive
systemthat changesitsstructurebasedonexternal or
internal information that flows through the network
duringthelearningphase. Inmorepractical termsneu-
ral networks arenon-linear statistical datamodeling
tools. They can be used to model complex relation-
shipbetweeninputsandoutputsor tofindpatternsin
data. Oneof themostimportantpropertiesof ANNsis
their abilitytolearnfromenvironmentandtoimprove
their performance with such learning. The learning
occurs when the NN reaches a generalized solution
for aparticular caseof problems.
Fuzzylogicisaformof multi-valuedlogicderived
fromfuzzy set theory to deal with reasoning that is
approximaterather than precise. J ust as in fuzzy set
theory theset membershipvalues canrangebetween
0and1,infuzzylogicthedegreeof truthof astatement
canrangebetween0and1andisnotconstrainedtothe
twotruthvalues, trueandfalse, asinclassicpredicate
logic.Whenlinguisticvariablesareused,thesedegrees
maybemanagedbyspecificfunctions.
The Adaptive Neural-Fuzzy Inference System
(ANFIS) isahybridintelligent systemcombiningthe
abilityof aneural networktofuzzylogic. Anillustra-
tivearchitect of ANFIS is given in Figure7. It uses
agiven set of input/output datato construct afuzzy
inferencesystemwithinwhichthemembershipfunc-
tion parameters are tuned or adjusted using either a
backpropagation algorithmalong or in combination
with aleast squaretypeof method. This adjustment
allows thefuzzy systemto learn fromdataand thus
it alsohasthepotential inprediction. Thispaper pre-
sentedanapplicationof anANFISbasedmodel inthe
predictionof shieldtunnel inducedsettlement.
Paper IS-339 presented an application of ANN in
predictinggroundsettlement inducedbytunnel exca-
vation. ThecasevisitedistheMetro-DF inthecityof
Brasiliaandthetunnel wasconstructedusingtheNew
Figure7. Architectureof ANFIS.
Table4. Input andoutput parameters.
Input parameters, x
i
Output parameter, y
Groundsettlementsalongthe Groundsettlement
centerlineof tunnel alignment alongthecenterline
0m, 5m, 10m, 20mand of tunnel alignment
30mrelativetofront endof +5mrelativeto
shieldmachine front endof shield
tunnel lininginstalledper dayin machine
termsof thenumber of rings
Austrian Tunneling Method (NATM). Two data sets
wereused for training theANN. Oneof thedataset
wasobtainedfromthemonitoreddataduringthetun-
nel constructionandtheother was establishedbased
on results of finite element modeling. Validating or
testingof theANN was performedafter thecomple-
tionof trainingphase. Results of theevaluationwith
monitored data show the average correlation coeffi-
cientsbetweenthepredictedandmeasuredvaluewere
about0.99and0.95atthecompletionof trainingphase
and at the validating phase, respectively. Although
the prediction with finite element method exhibited
better correlation between the results obtained from
thetrainingandvalidatingphases, precisionwas not
always achieved. Theauthors attributed thecauseof
lacking precision to inadequate constitutive model,
difficulty insimulatingthereal tunnel geometry, and
simulatethethreedimensional physical realityintwo
dimensional space. Based on the evaluation results,
the authors concluded that theANN is an effective
computational tool inpredictingsettlementinducedby
tunnelingwhengoodset of trainingdataisavailable.
Paper IS-014 presented the application of an
ANFIS model inthepredictionof groundsettlement
induced by shield tunneling. The case visited is the
Shanghai No.2subwayproject.Datasetobtainedfrom
field measurements was used in model training and
validating.
Results of the evaluation showed that theANFIS
predictionsareingoodagreement withthemeasured
data with relative error within the range from 2%
to 7%. For this particular case visited, the author
74
Table5. Analysisanddesign.
ID Topic
IS-083 Researchandapplicationof roadtunnel structure
optimization
IS-374 Frameworkof performance-basedfireprotection
designmethodfor roadtunnel
IS-125 Discussionondesignmethodfor retaining
structuresof metrostationdeepexcavation
inShanghai
IS-234 Researchonstochasticseismicanalysisof
undergroundpipelinebasedonphysical
earthquakemodel
also compared the ANFIS predicted results with
those derived fromPeck approach (1969), Pi-sigma
approach (Gupta and Rao, 1994) and Back Propa-
gation based neural network method. According to
theauthors, theANFISpredictionexhibitedrelatively
better accuracy and stablein terms of thecomputed
results.
2.3.2 Analysis and design
Thepapersrelatingtoanalysisanddesignaregrouped
inthissectionaslistedinTable5.
Accordingtopaper IS-083, conventional roadtun-
nelsaredesignedbasedonpassiveanalysisapproach.
This method follows the procedure (1) alternatives
development (2) computationandanalysis(3) select-
ing the best design scheme among alternatives. The
advantageof this methodis conceptually easy. How-
ever it is time consuming and may not be able to
identify the best alternative if it is not included in
the scheme developed. Thus, it is difficult to eval-
uate the performance of the structure. To improve
this, anoptimizationmethodisproposedinthispaper.
In this method, the dimensions of the tunneling are
treatedasinputvariables, andthestressof thespringis
theobjectivefunction. Algorithmof complexanalysis
programmedby C++ languagewasusedintheopti-
mizationprocess. Resultsof comparisonshowedboth
thestressdistributionalongthetunnel andtherelating
structural costwerereducedsignificantlybasedonthe
resultsevaluationfor thismethod.
Currently, performancebaseddesigncodehasbeen
usedinfireprotectiondesignfor buildings. However,
the design code for road tunnels is still prescrip-
tivebased. In viewof this, Paper IS-374 proposed a
frameworkof performancebaseddesigncodeforroad
tunnelswithlargecrosssections.
Although there is no significant variation in the
construction methods and geological condition for
Shanghai metro, results of diaphragmwall and strut
designinterms of diaphragmdepth, thickness, rebar
content, andstrut loadindeepexcavationareappar-
ently variedevenfor projectswithsimilar excavation
depth.ThepaperIS-125presentstheresultsof investi-
gationforprojectswithexcavationdepthfrom14.92m
to17.28m.Thecomparisonof thestrutloadsbetween
thedesignvalueandfieldmeasurement inShanghai
wasalsopresented. Causesof thedeviationwereiden-
tifiedtoberesultsof differentcomputationtools, anal-
ysisparameters, andlackingcommunicationbetween
designer and contractor. Over-design and unneces-
sary cost were often the consequences. Suggestions
ontheissues relatingto activeearthpressurecoeffi-
cients, vertical springcoefficient, k
v
, under thetoeof
thediaphragmwall, equivalent subgradecoefficient,
k
h
, strut load, diaphragmthickness, and monitoring
schemewereproposed.
3 CONCLUSIONANDREMARK
3.1 Treatment for geotechnical uncertainty
Inconventional approach, geotechnical uncertaintyis
managed through codes, standards, design criteria,
establishedproceduresandother devices. Uncertainty
is recognized somewhat indirectly and inexplicitly.
Engineering judgment plays an important role and
has beenusedcommonly indefiningandsettingthe
aforementioned management devices. For risk man-
agementandotheruniqueornon-routinegeotechnical
problems, judgment still plays an important rolebut
geotechnical uncertaintyneedstobetreatedexplicitly
for requirements such as safety evaluation. Proba-
bilistic methods includingBayes theorem, reliability
method, subjectiveprobability and moreprovidethe
necessarytoolsincopingwithit.
3.2 Qualification and quantification for risks
Expert investigationmethodwasusedextensivelyfor
risk analysis indeterminingtheprobability andcon-
sequencesof riskevents.Thismethodrequiresexperts
toexpresstheirjudgmentinqualitativetermsbasedon
predefinedcriteria, all informationavailable, andthe
uncertaintiesof therisktheyperceive.Theapplication
of elicitationtechniquesbasedoncognitiveprocessis
necessaryinfindingthetruebelief of theexperts. Itis
alsoimportanttoreducetheextentof whatconsidered
unavoidablebiaswhenusingelicitationtechniquesin
expert investigation.
Uncertainty is usually expressed qualitatively and
the use of verbal terms like possible, probable,
likely or unlikely seems to work well. However,
numerical quantificationprovidesthemissinglinkfor
qualitativeapproach. Theextent to which theuncer-
tainty of an event is greater than the other can be
determinedthroughquantitativeapproach. Itisimpor-
tant to note that probability, qualitatively or not, is
only averbal description or theoutput of numerical
quantification process. Theinsight obtained through
75
theprocess is what really matters. Theinsight helps
us understand the true meaning of the probability
associatedwiththeproblemandwhat behindit.
3.3 Decision analysis
Decision analysis is a technique developed on the
framework of risk analysis and comparative skills.
Throughthistechnique, arelativelybetterdesignalter-
nativecan bedetermined. Nevertheless, it shouldbe
notedthat decisionanalysis only provides abaseline
for decisionmaker. A well-informeddecisioncomes
about by consideringrisk magnitudes, risk reduction
measures,andfeasibilityinperformingthesemeasures
other thantheresultsof decisionanalysis.
3.4 Need for more case studies
Nonewtheory insoil mechanicscanbeacceptedfor
practical use without ample demonstration by field
observations that it is reasonable accurate under a
variety of conditions (Terzaghi and Peck, 1948). To
interpret theresults of observation, it is necessary to
incorporatejudgment. J udgment, or subjectiveproba-
bilityplaysanimportantroleforriskmanagementand
hazardcontrol. Theacquisitionof judgment relieson
experience and knowledge that can be derived from
field observation or diligent study of published case
histories. Becausetheopportunityof performingper-
sonal observationislimitedfor mostincontemporary
geotechnical practice, studyof casehistoriesbecomes
theprimarysourcefor accessinganddevelopingones
ownexperience.
Forriskmanagementandhazardcontrol, morecase
studies are needed. Geotechnical engineers should
makebestuseof availabletoolsandpresenttheircases
for further studies.
PAPER INTHEME 4
Ai, X.Q. &Li, J. 2008. Researchonstochasticseismicanaly-
sisof undergroundpipelinebasedonphysical earthquake
model.
Bao, X.H. &Huang, H.W. 2008. Riskassessmentforthesafe
gradeof deepexcavation.
Chissolucombe, I., Assis, A.P. &Farisa, M.M. 2008. Theuse
of artificial neural networkstopredictgroundmovements
causedbytunneling.
Cong, C. &Linde,Y. 2008. Multi-factordurabilityevaluation
insubwayconcretestructure.
Ding, W.Q. & Xu,Y. 2008. Researchandapplicationof road
tunnel structural optimization.
Guo, B.H. &Lu,T.K. 2008. Floorheavebehaviorandcontrol
of roadwayintersectionindeepmine.
Gutierrez, M. &Xia, C. 2008. Squeezingpotential of tunnels
inclaysandclayshalesfromnormalizedundrainedshear
strength, unconfined compressive strength and seismic
velocity.
Han, X. & Ding, G.Y. 2008. Framework of performance-
basedfireprotectiondesignmethodfor roadtunnel.
Hou, J., Zhang, M.X. & Tu, M. 2008. Predictionof surface
settlementinducedbyshieldtunneling: anANFISmodel.
Kim, D.H., Kim, U.Y., Lee, S.P., Lee, H.Y. & Lee, J.S. 2008.
Experimental studies of a geological measuring system
for tunnel withultrasonictransducer.
Lam, T.S.K. 2008. Performance review of a pipe jacking
project inHongKong.
Lee, W., Chung, S.S., Roberts, K.J. & Pang, P.L.R. 2008.
Geotechnical control of amajor railwayprojectinvolving
tunnel worksinHongKong.
Li, J.P., Wang, R.L. & Yan, J.Y. 2008. Research on struc-
tural statusof operatingtunnel of metroinShanghai and
treatment ideas.
Osborne, N.H., Ng, C.C., Chen, D.C., Tan, G.H., Rudi, J. &
Latt, K.M. 2008. Maximising the potential of strain
gauges: A Singaporeperspective.
Wang, R., Liu, G.B. & Liu, D.P. 2008. Discussionondesign
method for retaining structures of metro station deep
excavationinShanghai.
Yan, Y.R., Huang, H.W. & Hu, Q.F. 2008. Risk analysis for
cutterheadfailureof compositeEPBshieldbasedonfuzzy
fault tree.
Yao, C.P., Huang, H.W. &Hu, Q.F. 2008. Riskassessmenton
environmental impact inXizangRoadTunnel.
Zhou, H.B., Yao, H. & Gao, W.J. 2008. Risk analysis
andFuzzycomprehensiveassessment onconstructionof
shieldtunnel inShanghai Metroline.
REFERENCES
ABI/BTS. 2004. A J oint Codeof Practicefor theProcure-
ment, DesignandConstructionof TunnelsandAssociated
Underground Structures, London: The Association of
BritishInsurer, TheBritishTunnelingsociety.
Douglas Hubbard Howto MeasureAnything: Finding the
Value of Intangibles in Business, J ohn Wiley & Sons,
2007.
David Muir Wood. 2004. Geotechnical Modeling, Spon
Press, NewYork.
Gupta, M.M. & Rao, D.H. 1994. Ontheprinciplesof fuzzy
neural network. FuzzySetsandSystems, 61(1): 18.
Huang, H.W. etal. 2006. Guidelinesof riskManagementfor
MetroTunnellingandUndergroundEngineeringWorks,
Tongji University.
Internatinal tunnel Association. 2002. WorkingGroupNo. 2,
Guidelinesfor tunnelingriskmanagement, Balkema.
Peck, R.B. 1969. Deep excavation and tunneling in soft
ground. Proceedingsof the7thInternational Conference
onsoil MechanicsandFoundationEngineering, Mexico.
StevenG. Vick. 2002. Degrees of Belief-SubjectiveProba-
bilityandEngineeringJ udgment, ASCE Press, American
Societyof Civil Engineers, Reston, Virginia, USA.
Terzaghi, K. &Peck, R. 1948. Soil mechanicsinengineering
practice, Wiley, NewYork.
Whitman, R.V. 2000. Organizing and Evaluating Uncerat-
intyinGeotechnical Engineering,J ournal of Geotechnical
andGeoenvironmental Engineering,Vol. 126, No. 7, J uly,
2000.
76
Calculationanddesignmethods, andpredictivetools
LGCIE, INSA-Lyon, F-69621, France
ABSTRACT: This general report covers 19papers that areincludedinsession6of thesymposium, related
to thedesignor calculationmethods andpredictivetools for tunnelinganddeepexcavations. For this report,
thepapers havebeenclassifiedin3mainsubjects: i) excavations, ii) tunneling, iii) general papers ondesign
methods andtools. Thereareagreater number of papers concerningtunnelling, coveringalargenumbers of
subjects, subdividedinthefollowingtopics: T.B.M. simulation, Groundreactioncurve, Longitudinal behaviour
of segmentedlining, Settlement troughs, Effect of vibrations.
1 INTRODUCTION
This session with a very broad theme, contains 19
papers. Tables1and2presentaclassificationof these
papers, by countriesandby themes. Therearepapers
from8 countries, but more than half of the papers
(11/19) arefromChina.
Table1. Classificationbycountries.
Countries Number of papers
Brazil 2
China 11
J apan 1
Kazakstan 1
Korea 1
Netherlands 1
UK 1
USA 1
Table2. Classificationbysubjects.
Number
Topics Sub-topic of papers
Excavations 3
Tunnels TBM simulation 1
Groundreactioncurve 2+2(rock)
Settlement trough 2
Longitudinal behaviour 2
Effect of vibrations 3
General 4
Thecontent of thepapers can bebroadly divided
intocalculationanddesignof tunnellingworks, exca-
vations, andmoregeneral papers. 3papers deal with
excavation, considering3different aspects: basal sta-
bility, strut loads, and effect on nearby piles. There
areagreater number of papersconcerningtunnelling,
covering a large number of subjects. Among these
12 papers, 1 reports onTBM numerical simulation,
2present analytical methods for thegroundreaction
curve, 2 deal with the assessment of the settlement
trough, 2 consider the problem of the longitudinal
behaviour of thesegmentedtunnel lining, 3report on
theeffectof vibrationsor ontheseismicresponse, the
2remainingconcerningmoreproblemsrelatedtorock
tunnels. Finallythereare4general papersconcerning
thesimulationtoolsor presentinganational report.
Consideringthislargenumber of subjects, it isnot
reallypossibletohighlight amainemphasis, someof
thepapers covering very narrowsubjects and others
concerningverygeneral topics.
In thefollowing sections of this report, themajor
findingsandkeyfeaturesof eachpaper arepresented
andbrieflydiscussed.
2 EXCAVATIONS
Song and Huang studiedthebasal stabilityof anexca-
vationinsoft clay by anupper boundapproach. The
failuremechanismconsidered(Figure1) is basedon
theclassical Prandtl failuremechanism.Theiroriginal
contributiontothisproblemistoconsider thedepen-
denceof theshorttermshearresistanceof softclayson
thelocal orientationof thefailuresurface. They pro-
poseananalytical upper boundsolutionbasedonthis
77
H
D
T
B B1
i

R
Figure1. Definitionof geometricparameters.
0. 3 0.4 0.5 0.6 0.7 0. 8 0. 9 1
Anisotropy Ratio
0.6
0.8
1
1.2
1.4
S
a
f
e
t
y

F
a
c
t
o
r
D=6m
D=8m
D=10m
D=12m
Figure 2. Influence of D/H on the factor of safety
( =18kN/m
3
, undrained shear strength S
uv
(z)=0.33
v
,
width of the excavation B=15m, depth of the excavation
H=12m).
kinematical mechanismandontheequationproposed
by CasagrandeandCarillo (1944) for describingthe
anisotropyof shear strength:
whereS
uh
andS
uv
areobtainedby undrainedtriaxial
compressionandextensiontests.
Figure2presentstheresultsof aparametric study
ontheevolutionof thesafetyfactorwiththeanisotropy
ratio, for different values of the embedment depth.
It appears clearly that theanisotropy ratio has more
influencethantheembedment depthwhichincreases
onlyslightlythesafetyfactor.Theauthorsstudiedalso
the influence of the depth of the bedrock, when the
bedrocklimitstheextensionof thefailuremechanism.
Thisapproachcomparedwell witha2DFEanalysis
of adeepexcavationinBostonBlueClayapresentedby
Figure3. Comparisonof freeheadpilegroup.
HashashandWhittle(1996) usinganadvancedeffec-
tivestress soil model, MIT-E3. Theauthors analysed
also a case of failure in Shanghai where the stan-
dardcodesledtosafety factorsof morethan1.4and
wherethis approach leads to asafety factor of 0.97,
explainingthebasal instability.
Zhang and co-authors present a method for esti-
matingtheresponseof pilestolateral soil movements
inducedbyanearbyexcavation.
Forasinglepile,themethodisbasedontheclassical
two-stageapproach(Poulos& Chen1997):
in afirst step, thefree-field soil movement must
bedeterminedeither bymeasurement or bycalcu-
lation;
inasecondstep, thesesoil movementsareimposed
to the piles through aWinkler subgrade reaction
model: thepileis representedby anelastic beam,
the pile-soil interaction is modeled using linear
elastic soil springs, the effect of axial load on
thepileisignored. TheWinkler subgradereaction
equationissolvedbyaFDapproach, permittingto
takeintoaccount heterogeneoussoils.
This classical method has been extended by the
authorstopilegroups. Inthecaseof pilegroups, the
shielding effect of piles is modelled by superposing
to the free field soil movement the reduction of the
displacement dueto neighbouringpiles. This shield-
ingeffect is calculatedusinganattenuationfunction
basedonsimplifiedMindlinsequation.
Theauthorspresentacomparisonof their approach
withcentrifugemodel tests andfiniteelement simu-
lationpublishedbyLeunget al (2000).
A comparisonof calculatedandmeasuredbending
moments is given on Figures 3 and 4 in thecaseof
78
Figure 4. Comparison of front pile in capped head pile
group.
Figure5. Exampleof strut system.
pilegroups: thissimplifiedmodel fitsquitewell with
theexperimental resultsforfreeheadedpilesbutthere
aresomedifferencesinthecaseof cappedpiles. The
authors explainthat this differencecouldbereduced
by using a non-linear elastic spring hypothesis, but
perhapsisit duetotheWinklershypothesisitself.
Finally, itmustbenotedthatthismethodisnotspe-
cifictoexcavations, andcouldbeusedforotherworks
inducinglateral soil movements, suchastunnellingor
embankmentsonsoft clays.
Shi and co authors presentamethodfor estimating
byback-analysisthestrutloadsinacomplexconcrete
strutsystem, suchasthesystempresentedonFigure5.
The proposed method is based on measured dis-
placements of thewall inthehorizontal planeof the
Figure6. ForcesactingontheTBM.
strut system, the load distribution between the dif-
ferent struts at the same level being calculated by
back-analysis.
Astherearenodirectloadmeasurements, themag-
nitudeof theloadsobtainedbythisapproachdepends
stronglyontheapriori hypothesisconcerningthedis-
tributionof soil pressureactingonthewall. Therefore
thismethodcanonlygiveanindicationontherelative
distributionof theloads.
3 TUNNELING
3.1 T.B.M. simulation
Chen and co-authors present aninterestingpaper on
thebehaviourof aTBMwhenfollowingacurvedalign-
ment.Theyproposeacomprehensivenumerical model
of an articulated shield which is an extension of a
kinematic shield model proposed for a single circu-
lar shield(Sugimoto & Sramoon2002). Their model
isfocusedonthetunnel boringmachine, considering
all theforcesactingontheshield, suchasfor example
(Figure6)
thedifferent jackthrust forces,
thepressureactingontheface,
theforcesactingontheshieldperiphery.
Theselatterforcesrepresenttheinteractionbetween
theshieldandthesurroundingsoil.Theyaresimulated
by aspringmodel. Therefore, inthis model, thetun-
nellingoperationisseenmainlyfromthepointof view
of theTBM.
Thesimulationof theTBMbehaviourisobtainedby
imposingtothemodel themainoperationparameters
of theshield.
A comparisonbetweenanobservedandsimulated
behaviour is presentedinFigure7. Theactual shield
trajectory, inthevertical andhorizontal plane, andin
79
29.6
29.7
29.8
29.9
20 40 60 80 100
Distance (m)
x

(
m
)
-6940
-6920
-6900
-6880
-8740 -8720 -8700 -8680 -8660 -8640
z (m)
y

(
m
)
Observed
Simulated
start
end
start
end
Vertical plane
Horizontal plane
R =20 m
Figure7. Simulatedandobservedbehaviour.
extremeconditionsof asharpcurve, iswell simulated
bythisverycomprehensivemodel.
Thismodel givesalsothefieldof soil pressureact-
ing on theshield, derived fromthespring model, as
shownonthisfigure. Inafurtherstep, itcouldbeinter-
estingtousethesecalculatedcontactstressesbetween
theshieldandthesoil inacontinuummodel of thesoil
massfor modellingthesoil deformationandtocheck
if thesecalculated contact stresses lead to arealistic
simulationof observeddisplacements andsettlement
troughs.
3.2 Ground reaction curve
Shin and co-authors proposean analytical model of
the ground reaction curve taking into account the
seepageforces.Thefollowingclassical hypothesesare
adopted:
Thetunnel isboredinaninfinitesoil masssubjected
toahydrostaticinsitustress,
Thesoil is linear elastic perfectly plastic withthe
Mohr-Coulombyieldcriteria,
Radial seepage forces are taken into account, as
indicatedintheequilibriumequation:
The hydraulic gradients are calculated separately,
consideringasteadystateof seepage.
Basedonthesehypotheses, theauthorsproposean
analytical solutionof theelasto-plasticstateof stress,
expressed in terms of stress state and displacement.
Duetotheseassumptions, thismodel ismoreadapted
todeeptunnels.
Theauthors present an application of their model
for a50mdeeptunnel, withadiameter of 5meters.
Different cases are examined: fully drained or with
Figure8. Effect of radial seepageon theground reaction
curve.
asemi impervious lining, havingthesamemechani-
cal propertiesthanthesurroundingsoil.A comparison
betweenthedrysoil caseandthefullydrainedcaseis
presentedinFigure8:
in dry conditions, the pressure decreases rapidly
with the convergence, as the soil resistance is
mobilised,
when fully drained seepage is considered, there
is a marked increase of the convergence for a
given internal pressure, as the seepage forces do
not depend on the soil convergence and remain
constant.
Despitethediversesimplificationsof suchanana-
lytical model, this paper gives interestingindications
ontheinfluenceof seepageforces andshows clearly
that they should be taken into account for mod-
elling the ground reaction curves and for assessing
thestabilityof theexcavation.
Inasecondpaper ongroundreactioncurves, Sozio
presentsa2Dor 3Danalytical model representingthe
tunnel andthesoil cover by athick sphereor athick
cylinder (Figure 9). The soil model is the classical
linear elastic-perfectlyplasticMohr-Coulombmodel.
The originality of this model is that the gravity
forcesareemulatedbyradial bodyforces.Thisenables
totakeintoaccount alimitedcover depth, withasur-
faceloadandaninternal pressure.Theauthorproposes
to use this 3D model for a preliminary assessment
of thestability of theunlined length of atunnel, the
problembeing to estimate the radius of the sphere
equivalent tothetunnel unsupportedheading.
Such analytical models are generally based on
restrictive hypotheses. It is the case for this model,
butithastobehighlightedthatinhispaper, theauthor
indicates very clearly thelimitations of theproposed
models. Itshouldbeinterestingtocomparethismodel
totheclassical approachbasedontheassumptionof a
tunnel boredinaninfinitesoil mass.
80
Figure9. Representationof thesoil tunnel interaction.
ThepaperspresentedbyZhang &Wang andLu et al
concernmorespecificallydeeprocktunnels.
Zhang &Wang studythegroundreactioninthecase
of apressuretunnel, therockmassbeingnotunloaded
but loaded by the internal pressure. In this specific
situation, quitefar fromurbantunnelsinsoft ground,
thesofteningof therockconsideredbytheauthorscan
leadtoabrokenzonearoundthetunnel.
Lu et al studied by 3D numerical simulations the
stability of different types of intersections between
deepminetunnelsandtheinfluenceof theconstruction
sequences. Themethodof simulationisnot precisely
described. If this study is not directly applicable to
shallowtunnels in soft ground, someof their results
can be considered froma qualitative point of view
suchas thefact that excavatingtowards theintersec-
tionappearsmoredangerousthanexcavatingfromthe
excavation.
3.3 Longitudinal behaviour of segmented lining
The paper proposed by Hoefsloot is based on the
observationthatthestagedconstructionof segmented
tunnel liningsinducesapermanentlongitudinal bend-
ingmomentinthelining. Basedonsolutionsproposed
byBogaards&Bakker(1999), andBakker(2000), the
author proposesananalytical solutionbyconsidering
thesegmented tunnel lining as abeamon an elastic
foundation.
Thelongitudinal loadingscheme(Bendingmoment
andshearforcefromjackforces, shearforcefromsteel
brushes, weight of lining segments, uniformly dis-
tributedloadof limitedlength: backuptrain)advances
withtheprogressof theTBM(Figure10).Thisanalyt-
ical model hasbeenbuiltinaspreadsheet, andverified
usingPLAXIS2D.
The result of this model is compared with strain
measurements made in the lining of the GROENE
HART tunnel in the Netherlands. As illustrated on
Figure11,abendingmomentisinducedintheliningby
D
br

l
i
l
u
x
q
w
M
jack
D
jack
q
l
q l
Figure10. Model usedbyHoefsloot torepresent thestage
constructionof thetunnel lining.
Figure11. Groenehardt tunnel evolutionof thebending
moment withtheadvanceof theTBM.
theadvancementof theTBMandbecomespermanent
afterabout60meters. Despitethesimplehypothesisof
aspringmodel,theevolutionof thebendingmomentis
quitewell modelled. Neverthelessthisresult hasbeen
obtainedbyadjustingsomeparameterswhicharedif-
ficult to assess, such as the lining bending stiffness
andtheeffect of grouting(Talmonet al. 2008).
Finally, this analytical model, validated on field
measurements, shows that thestagedconstructionof
thesegmental lininginastraightalignmentresultsina
permanent longitudinal bendingmoment, that should
beconsideredinthedesignof theliningandfor the
installationof thesegments.
A second paper on the longitudinal behaviour of
segmented lining is presented by Zhu et al. The
authors examinetheproblemof theactual longitudi-
nal stiffnessof segmentedliningswhichisoneof the
parameterswhichwasnecessarytoadjustinthemodel
proposedbyHoefsloot.
Theassessment of thelining stiffness is based on
a 3D numerical model, composed of shell elements
(Figure12) andjointswithshear andnormal stiffness
at all theinterfaces betweentheindividual elements.
Thecompletenumerical model, loadedasacantilever
beam, is comparedtoasimplifiedequivalent contin-
uous model, which is not precisely described in the
paper.
The stiffness deduced fromthe numerical model
appears on their example to depend on the segment
length and to be lower than the stiffness obtained
81
Figure12. Liningmodel.
2.U
0
U
0
/2
U
0
/2
Figure13. ConvergenceschemeproposedbyHeli Baoetal.
by the equivalent continuous model. But both mod-
els are based on hypothesises concerning the joint
behaviour, which need to be measured or assessed
based on the observation of the actual behaviour of
full scalesegmentedlinings.
3.4 Settlement troughs
Thereis generally alargeconsensus about theuseof
theGaussiantypecurvefor describingthesettlement
trough.
Thedirectestimationthroughelasticcalculationsor
by numerical simulations oftenleads to larger settle-
menttroughsthanobserved.Theresultdependsinfact
ondifferent assumptions, onebeingtheconvergence
profileof thegroundaroundthetunnel.
Heli Bao and his co-authors present an analyti-
cal solution, using conformal mapping of an elastic
half space. In order to fit with the observed settle-
ment troughs, they proposeanelliptical convergence
shape, basedonthesolutionproposedbyPark(1974),
asindicatedonFigure13.
This approachis comparedwiththeobservedset-
tlements duringtheconstructionof a6.2mdiameter
tunnel inShanghai. Thecalculatedsettlement trough
fitsquitewell withtheobservedone. Butnoindication
is givenconcerningtheassessment of themagnitude
of theconvergence, theauthorsindicatingsimply the
gapbetweentheTBM andthelining, gapwhichisin
fact certainlypartiallyfilledbygrouting.
Intheirpaper, Zu and Liu compareddifferentmeth-
odsforsettlementtroughassessment: Pecksempirical
approach(Peck1969), stochasticmediumtheory, and
thesolutionproposedbyVerruijt andBooker (1996).
-30 -20 -10 0 10 20 30
0
20
40
60
80
100
120
Distance to the center of tunnel (m)
s
u
r
f
a
c
e

s
e
t
t
l
e
m
e
n
t
(
m
m
)
Peck's method
Stochastic medium theory
Verruijt and Booker's method (v=0.1)
Verruijt and Booker's method (v=0.2)
Figure 14. Settlement trough calculated using 3 different
methods(ZuandLiu).
TheexamplepresentedonFigure14exhibitsverylarge
differences, but it is in fact an extremecase, with a
coverdepthof lessthanonemeterforatunnel 6meters
indiameter.
Theypresentalsosomeobservedsettlementtroughs
fromShanghai metroline7construction.Itwouldhave
been interesting to have more details on the tunnel
worksleadingtothesesettlementtroughsandtocom-
paretheobservedsettlementtroughstothecalculation
methodspresentedinthefirst part of thepaper.
3.5 Effect of vibrations
Intheir 2complementarypapers, Cui and co-authors
investigate experimentally the dynamic loading and
the development of pore pressure of saturated silty
claysnearShanghai subwayLineNo.2, duringthepas-
sageof metrotrains. Ontheobservedsite, settlements
exceeding20cmwhereobserved, butnodetailsonthe
evolutionof thesesettlementsisgiven.
Theexperimental study is basedonfieldobserva-
tionsanddynamictriaxial tests.
Based on in situ measurements, the authors pro-
poseanexperimental lawof attenuationof thedynamic
loadingwiththedistancetothetunnel.Thislaw, dueto
thepolynomial approximationadopted, has certainly
alimited domain of validity, and cannot beextrapo-
lated outsidetherangeof distancecorresponding to
themeasurement points.
They studied also the development of pore pres-
surewiththedynamic loading, bothinthefieldand
by triaxial tests; but the relation of field measure-
mentsresultswithdynamictriaxial testsisnot clearly
described.
It would have been interesting to have more
details concerning the long termevolution of pore
pressureinsitu, combinedalso withtheevolutionof
soil deformationsduringcyclingloadingtests,inorder
toexplaintheobservedlargesettlements.
Baimakhan and co-authors propose a coupled
approach(analytical andnumerical) todeterminethe
82
stresses induced by earthquakes on the lining of
tunnelsof subwaylines.
Usingaconceptof homogeneousanisotropicelastic
mediumthey consider theeffect of thesuccessionof
different soil or rocklayers.
They analyse in a more specific way the case of
tunnel or gallerieswithalongitudinal axismakingan
anglewiththemajor directionof anisotropy.
4 GENERAL PAPERSONDESIGN
METHODSANDTOOLS
Koungelis &Augarde compare, onanacademicexam-
ple, theresultsgivenby2differentFE codes, Strand7
andPlaxis.Theyinvestigatetheeffectof surfaceload-
ingonwishedinplacetunnelsinsoftgroundassuming
planestrainconditions.
The initial conditions and soil characteristics are
thesameinall their simulations, except for dilatancy.
Thetwo meshes used in their comparativestudy are
quitedifferent: inPlaxisthemeshismorerefinedand
consistsof fifteen nodedtriangular elements, andin
Strand7themeshiscoarser, andmoreover consistsof
simple6nodedtriangular elements.
Theauthorscomparethechangesinhorizontal and
vertical diametersfor different positionof thesurface
load. There are actually small differences although
thereis agreat differencebetweentherefinement of
themeshesandthetypeof triangular elements.
Although no indication is given in thepaper, one
can suppose that in this example, plastic zones are
certainly very limited or absent around the tunnel.
Therefore, this paper compares essentially theinflu-
enceof themeshrefinementandof thetypeof element
inalinearelasticcase,whichexplainsthefactthatonly
minordifferenceisnoted.Suchcomparisonsshouldbe
extended to less academic situations, wherethetun-
nel construction is simulated and wheresmall strain
behaviour ismodelledor wherelargeplasticzonesare
mobilised.
Jeon and coauthors present an interesting com-
munication on the use of geostatistical methods for
assessingthespatial distributionof therockmasschar-
acteristicnamedRMR.Theycompareamethodnamed
SIS(J uanget al., 2003, Fenget al., 2006) tothemore
classical kriging (Marinoni, 2003; Pardo-Igzquiza
andDowd, 2005).
Theproblemconsideredhere, ishowtoassess, on
the base of limited bore holes, the ground charac-
teristics along thetunnel alignment. Theapplication
presented concerns deep rock tunnels, where geo-
statistical estimations of RMR around a tunnel are
compared.
Compared to kriging which gives a deterministic
value at each point considered, SIS gives a statisti-
cal distribution of theunknown valuewith different
Figure15. Distributionof estimatedvalue.
characteristics of this statistical distribution as illus-
tratedonFigure15.
Such a distribution is important information for
tunnelling projects, based on a limited number of
investigationpoints, inorder toevaluatethelimitsof
thedesignandfor riskassessment.
One of the difficulties for multiplying 2D or 3D
numerical simulations on tunnelling projects is that
thepre-processingtasksaremuchtimeconsuming.
Li and co-authors proposeintheir paper amethod-
ologytoderiveaFEMmodel fromnumerical geologi-
cal models, whicharemoreandmoreusedintheframe
of largegeotechnical projects.
The figures presented Figure 16 show some of
thestages, beginningfromthegeological model and
ending to a 3D finite element mesh and where the
soil characteristics areimported fromthegeological
model.
Insuchanapproach, couplingthegeological model
withthegeostatistical methodpresentedintheprevi-
ouspaper couldbecertainly avery powerful tool for
helpingthedesigner totestdifferenthypothesisof the
soil parameters.
Negro presentstheresultsof acomprehensivesur-
vey of current designpracticeinBrazil. Theanalysis
of this survey is basedontheanswers of 20experts.
Thetopics of thesurvey concernthemainaspects of
tunnel design:
Tunnel headingstability,
Settlement,
Damagetoexistingstructures,
Liningdesign,
Account of groundwater loading,
2D/3DFEM or FDM models,
Soil modelsinFEM/FDM models,
Soil investigations,
Monitoring.
Theresults of this survey areclearly summarized
andanalysedinthepaper byNegro, therefore, inthis
report aregivenonlysometypical examples.
Fromthissurvey, it resultsthat thetypical scenario
for tunnel projects inBrazil is thefollowing: tunnels
withequivalentdiameter larger than6m, drivenunder
83
Figure16. Somestages fromthegeological model to the
3DFE model.
Figure17. Soil modelsusedinnumerical analyses.
mixedfacecondition, incohesivesoils, belowwater
table, andusingsprayedconcreteaslining(NATM).
Concerning the assessment of tunnel face stabil-
ity, the survey shows that there is a large range of
methodscurrentlyused, andit isnoticedthat someof
thesemethodssuchaslimit equilibrium, upper bound
solutionsor empirical methodscanbeunsafe.
It is highlighted also that Practitioners arein fact
unhappywiththeavailablemethodsfor stabilityanal-
ysis, which certainly explains this broad range of
methodsusedinpractice.
Another interestingresultconcernstheconstitutive
modelsusedinnumerical analysis(Figure17).
A largemajority still usethelinear elastic/plastic
Mohr Coulombmodel whichis well known, but cer-
tainlyoftennotadaptedtoshallowurbantunnelswhere
thelimitationof soil movementsleadtosmall strains.
Very few or no plasticity will be mobilised, and the
model will be equivalent to a simple linear elastic
model which very poorly describes the small strain
behaviour of thesoil. It canbeaddedthat thissimple
model canbeunsafeincoupledanalysis, asthevolume
changesinthesoil arenot correctlydescribed.
Thisexampleshowsthat, withthelargeavailability
of 2D or 3D Finiteelement or finitedifferencepow-
erful codes, thereiscertainlyaneedfor clarifications
concerningthetypesof soil modelstobeusedindif-
ferent situationsandalsoaneedfor disseminationof
thisknowledge.
Therichconclusionsof suchasurveycouldbecer-
tainlyagoodbasefor developingandenhancinggood
practicesinthefieldof tunnel design. For thisreason
TC28has proposedto launchnational surveys based
ontheexampleproposedbyNegro, tobecollectedand
analysedfor thenextTC28symposium.
5 CONCLUSION
After excludingthe4general papers, the15remain-
ingpapersallocatedtothissession, wererelatedto9
different specifictopics.
This highlights that tunnels and deep excavations
arecomplexworks, withstronginteractionwiththeir
environment, and that there is obviously a demand
for simplecalculationtoolsaddressingspecificprob-
lems, easytouse, especiallyatthepreliminarydesign
phases.
Concerning the calculation methods, it can be
noticed that 9 papers concern analytical or mixed
approaches as only 5 concern numerical methods.
Therefore, considering the limits of the analytical
approaches, duegenerallytotherestrictivehypotheses
necessary to obtainaclosedformsolution, it should
becertainly useful nowto developsimplenumerical
tools, easy to useand timesaving, dedicated to lim-
itedspecificproblems. Thesetoolscouldbebasedon
existingcodesandthereforeabletotakeintoaccount
advanced soil models and realistic geometries. Such
tools, after acomprehensiveevaluationof their limits,
couldbecertainlyuseful for practitioners.
Another way to beconsidered is thedevelopment
of easytousepre-processingtools, suchastheexam-
ple presented in this session, to facilitate the use of
complex3Dmodels.
Andfinally, it shouldbestressedthat all thesecal-
culation methods have to be validated carefully and
in a scientific way against comprehensive measure-
ments, andthatthelimitationsof thesemodelsshould
beclearlyindicated.
84
LIST OF PAPERSWITHINSESSION
Baimakhan, R.B., Danaev, N.T., Baimakhan, A.R.,
Salgaraeva, G.I., Rysbaeva, G.P., Kulmaganbetova, Zh.K.,
Avdarsolkyzy, S., Makhanova, A.A. & Dashdorj, S. Cal-
culation of the three dimensional seismic stressed state
of MetroStationEscalatorOpenLineTunnelssystem,
whichislocatedininclinedstratifiedsoft ground.
Bao,H.,Zhang,D.&Huang,H.AComplexVariableSolution
forTunneling-InducedGroundMovementsinClays.
Chen, J., Matsumoto,A. &Sugimoto, M. Simulationof artic-
ulatedshieldbehavior at sharpcurvebykinematicshield
model.
Cui, Z.D., Tang, Y.Q. & Zhang, X. Deformation and pore
pressuremodel of thesaturatedsiltyclayaroundasubway
tunnel.
Hoefsloot, F.J.M. Analytical solution of longitudinal
behaviour of tunnel lining.
J eon, S., Hong, C. & You, K. Design of tunnel supporting
systemusinggeostatistical methods.
Koungelis, D.K. & Augarde, C.E. Comparative study of
softwaretoolsontheeffectsof surfaceloadsontunnels.
Li, X.X., Zhu, H.H. & Lin, Y.L. Geologic model transform-
ingmethod(GMTM) for numerical analysismodelingin
geotechnical engineering.
Lu, T.K., Guo, B.H., Cheng, L.C. & Wang, J. Review and
interpretation of intersection stability in deep under-
groundbasedonnumerical analysis.
Lu, Z.P. &Liu, G.B.Analysisof surfacesettlementduetothe
constructionof ashieldtunnel insoft clayinShanghai.
Negro, A. UrbanTunnelsinSoil: Reviewof Current Design
PracticeinBrazil.
Shi, Z., Bao, W., Li, J., Guo, W. & Zhu, J. A study onloads
fromcomplexsupport systemusingsimple2Dmodels.
Shin, Y.J., Shin, J.H. & Lee, I.M. Ground Reaction due to
TunnellingbelowGroundwaterTable.
Song, X.Y. & Huang, M.S. Basal Stability of BracedExca-
vations in K0-consolidated Soft Clay by Upper Bound
Method.
Sozio, L.E. Analytical Two and Three Dimension Mod-
els to Assess Stability and Deformation Magnitude of
UndergroundExcavationsinSoil.
Tang,Y.Q., Cui, Z.D. &Zhang, X. DynamicResponseof Sat-
uratedSiltyClayaroundaTunnel underSubwayVibration
LoadinginShanghai.
Zhang, C.R., Huang, M.S. & Liang, F.Y. Lateral Responses
of PilesduetoExcavation-InducedSoil Movements.
Zhang, L.M. & Wang, Z.Q. Elastic-plastic analysis for sur-
rounding rock of pressure tunnel with lining based on
material nonlinear softening.
Zhu,W.,Kou,X.,Zhong,X.&Huang,Z.Modificationof Key
Parametersof Longitudinal Equivalent Model for Shield
Tunnel.
REFERENCES
Bakker, K.J. 2000. Soil Retaining Structures. Rotterdam:
Balkema.
Bogaards, P.J. & Bakker, K.J. 1999. Longitudinal bending
in Geomechanics Proc. NUMOGVII: p. 317321.
Casarande,A. &Carillo, N. 1944. Shearfailureof anisotropic
soil. J. of the Boston Society of Civil Engineers, 31(4).
Feng, Y., Tang, S. & Li, Z. 2006. Application of improved
sequential indicator simulation to spatial distribution
of forest type. Forest Ecology and Management 222:
391398.
prediction for deep excavations in soft clay. Journal of
Geotechnical and Geoenvironmental Engineering,ASCE,
122(6): 474486.
J uang, K., Chen, Y. & Lee, D. 2003. Usingsequential indi-
cator simulation to assess theuncertainty of delineating
heavy-metal contaminatedsoils. Environmental Pollution
127: 229238.
Leung, C.F., Chow, Y.K. & Shen, R.F. 2000. Behavior of
pilesubjectedtoexcavation-inducedsoil movement.Jour-
nal of Geotechnical and Geoenvironmental Engineering,
126(11): 947954.
Marinoni, O. 2003. Improving geological models using a
combinedordinaryindicatorkrigingapproach.Engineer-
ing Geology 69(12): 3745.
Pardo-Igzquiza, E. & Dowd, P.A. 2005. Multipleindicator
cokrigingwithapplicationtooptimal samplingfor envi-
ronmental monitoring. Computers & Geosciences 31(1):
113.
Park, K.H. 2004. Elastic solution for tunneling-induced
ground movements in clays. International Journal of
Geomechanics 4(4): 310318.
Peck, R.B. 1969. Deep excavations and tunneling in soft
ground. Proceeding of 7th international conference on
soil mechanics and foundation engineering. MexicoCity:
Stateof theArt Report.
Poulos, H.G. & Chen, L.T. 1997. Pile response due to
excavation-induced lateral soil movement. Journal of
Geotechnical and geoenvironmental engineering, ASCE,
123(2): 9499.
Sugimoto, M. & Sramoon, A. 2002. Theoretical model of
shield behavior during excavation I: Theory. Journal of
138155.
Talmon, A.M., Bezuijen, A. & Hoefsloot, F.J.M. 2008. Lon-
gitudinal tubebendingduetogrout pressures. Shanghai:
TC28.
Verruijt, A. & Booker, J.R. 1996. Surface settlements due
to deformation of a tunnel in an elastic half plane.
85
Analysisandnumerical modelingof deepexcavations
R.J. Finno
Department of Civil and Environmental Engineering, Northwestern University, Evanston, IL, USA
ABSTRACT: Thesixteenpapers comprisingthegeneral themeAnalysis andNumerical Modelingof Deep
Excavationsof the6thInternational SymposiumonGeotechnical Aspectsof UndergroundConstructioninSoft
Groundaresummarizedherein. General characteristicsof all papersarepresentedasarebrief summariesof each
paper. Mostpapersincludedpresentationof resultsof finiteelementsimulationsandattendantcomparisonswith
variousaspectsof observedfieldperformance. Someof thepitfallsfor makingcomparisonsbetweennumerical
resultsandfieldobservationsof deepexcavationperformancearediscussedbriefly. Inparticular, theeffectsof
modelingconstructiondetailsandselectionof appropriateconstitutivemodelsarepresented. Recommendations
aretenderedregardingtheessential informationthatshouldbeconveyedinpapersthatpresentresultsof numerical
calculations.
1 INTRODUCTION
Sixteen papers concerning analysis and numerical
modeling of deep excavations werepublished in the
Proceedings of the 6th International Symposiumon
Geotechnical Aspectsof UndergroundConstructionin
SoftGround.Of thesepapers,tenwerepresentedorally
atthesymposium.Anoverviewof thepapersismadeto
provideasnapshotof thestateof thepracticeasregards
to this topic. All papers are summarized and trends
in the contents are discussed. Because most papers
includedpresentationof resultsof finiteelementsimu-
lationsandattendantcomparisonswithvariousaspects
of observed field performance of deep excavations,
general commentsregardingfactorsthatcanbeimpor-
tant inorder to makeanaccurateprediction. Finally,
recommendationsaretenderedregardingtheessential
information that should be conveyed in papers that
present resultsof numerical calculations.
2 OVERVIEW
The sixteen papers covered the broad topics sum-
marized in Table 1. The classification is somewhat
arbitrary andseveral papers couldhavefit into more
thanonetopic. Itisclear, however, thatthemajorityof
papers explicitly included comparisons of computed
resultsandsometypeof performancedata.
Of these16papers, resultsof finiteelement analy-
seswerereportedintenof them. Table2summarizes
theFE codesthat wereused. Theseresultsagreewith
the authors experience that the commercial codes
Table1. Topicsandnumber of papers.
Topic No. of papers
Numerical analysisandmeasurements 7
Numerical analysis 3
Back-analysis 2
Measurements 1
Design 1
Stresspath 1
Earthpressure 1
Table2. Summaryof finiteelement codesusedinpapers.
Analyses Number andcodeused
Three-dimensional simulations 2-Plaxis3DFoundations
2- FLAC3
2-Researchcode(?)
Planestrainsimulations 6-Plaxis
1-GeoTunnel
1-Researchcode(?)
FLAC andPLAXISaremost commonlyusedinboth
geotechnical practiceandresearch. Theeaseof useof
thesecodes has progressedto thepoint wherethree-
dimensional analyseshavebecomemorecommon, as
suggested by thenumber of such analyses presented
inthepapersinthissession.
Whiletheuseof finiteelementanalyseshasbecome
more common in practice, likely as a result of the
87
Table3. Constitutivemodel summary.
Model No. of applications
Mohr-Coulomb 6
HardeningSoil 3
Duncan-Chang 1
improvedi/oof commercial codes, theaccuracyof the
resultsdependsthefaithful representationof activities
that inducestresschangesinthegroundduringexca-
vation and on theconstitutiveresponses assumed in
theanalyses. Simplemodels areeasy to use, but are
limited in thetypes of computed responses that will
agreewithobservations.
A keyaspect toapplyingfiniteelement analysisto
practical problems in geotechnics remains theselec-
tion of thesoil model and its individual parameters.
Of thetenpapersthatpresentedresultsof FEanalyses,
sixassumedsoil respondedasaMohr-Coulombmate-
rial, as shown inTable3. This assumption limits the
predictivecapabilitiesof aFEsimulationof deepexca-
vationsortunnelinginthatelasticresponseisassumed
until thesoil fails. Presumably, thischoicewasmade
becauseof lack of detailed laboratory or field char-
acterizationsof thesoilsconsideredbytheauthorsin
their papers. At thecost of simplicity, theHardening-
Soil and Duncan-Chang models are non-linear and
can account for different responses in loading and
unloading. But thesemodels also arelimitedintheir
predictivecapabilitiesinthat they donot account for
theincremental non-linearityandsmall strainstiffness
responsesthatall soilsexhibit. Inanycase, thefactors
leadingtotheselectionof asoil model shouldbedis-
cussedinapaper to put theresults incontext. Inthe
same spirit, the parameters and a rationale for their
selectionalsoshouldbeincludedinapaper.
3 SUMMARY OF PAPERS
3.1 Numerical analysis and measurements
Li and Huang presented Construction monitoring
andnumerical simulationof anexcavationwithSMW
retainingstructure. A SMW retainingstructurewas
used to support two long excavations in Shanghai.
Bearing and deformation mechanisms of the SMW
were analyzed briefly and the structural analysis of
SMWwasdiscussed. Basedonthein-situexcavation
procedures, theauthorssimulatedconstructionof the
wall numerically using the FE code FLAC3D. They
representedsoil behaviorwithaMohr-Coulombmodel
andconsideredtwocases. Case1wasthetypical con-
structionsituationatthesitewhereinthesupportswere
installed in a timely fashion. Case 2 considered the
situationwhereinthesupportswerenot installedina
timelyfashioninthelateral direction, therebyleaving
excessivelylargeamountsof wall without lateral sup-
port. They compared the computed deformations of
theretainingstructure, thehorizontal displacement at
thetopof SMWandtheaxial forcesof steel pipesup-
portswiththefieldobservationdataforbothinstances.
The authors concluded good agreement was shown
between thecomputed and observed results. Froma
practical point of view, they showed that thenormal
constructionsequencingincase1resultedinastable
andsafeexcavation; theaxial forces werelower than
the alarmvalues and the displacement due to exca-
vation werewithin thepermissiblerange. In case2,
however, thecomputedresultsshowedtheexcavation
wasclosetobecomingunstable, andmeasureshadto
betakentoprotecttheretainingstructurefromfailure.
Popa et al. presented Numerical modeling and
experimental measurements for aretainingwall of a
deepexcavationinBucharest, Romania. They sum-
marized a case history of a diaphragm wall for a
deep basement of a new building in the center of
Bucharest. Theexcavationimpactedanumber of his-
toric structures, leading to the use of top-down
techniques to support theexcavation. Thenumerical
results obtained by plane strain FE simulation were
compared with measurements recorded during con-
struction. Soil behavior was assumed to bethat of a
Mohr-Coulomb material. The computed lateral dis-
placementswere15%and75%of theobservedvalues,
depending on if the comparisons were made in an
area with or without a grouted wall not explic-
itly modeled in theFE simulations adjacent to the
diaphragmwall.
Schweiger etal. presented3Dfiniteelementanal-
ysisof adeepexcavationandcomparisonwithinsitu
measurements.Thepaper describestheresultsof FE
analyses usingPlaxis of adeepexcavationproject in
clayeysilt inSalzburg. Theexcavationwassupported
byadiaphragmwall, ajetgroutpanel andthreelevels
of struts. Thesoil responses wererepresentedby the
Hardening-Soil model. Becauseof insufficient infor-
mationavailableat thetimeof designonthematerial
properties of the jet grout panel, the authors varied
itsstiffnessinaparametricstudy. Theeffect of taking
intoaccountthestiffnessof acrackeddiaphragmwall
on the deformations also was investigated. In some
of the 3D calculations, the authors simulated non-
perfectcontactbetweenthediaphragmwall andastrut
by means of anon-linear behaviour of thestrut. The
evaluationof theresults andcomparisonwithinsitu
measurements showed that analyses which took into
account the reduced stiffness of the diaphragmwall
duetocrackingachievedthebest agreement withthe
measurements. Furthermoresettlements of buildings
could be best reproduced by the three-dimensional
model, althoughthepredictedsettlementswerenot in
goodagreement withtheobservations.
88
Zhang and Huang presented Monitoring and
modeling of riversidelargedeep excavation-induced
ground movements in clays. They discussed adeep
excavationlocatedat theShanghai international pas-
sengercenterthatwas800mlongand100150mwide
with the depth of 13m. The south side of the deep
excavation was within 4.6mof aparallel flood wall
of the Huangpu River. The north side of the exca-
vation was 5mfromahistoric building. Becauseof
thedifferences in theconditions on thetwo sides of
the excavations, Plaxis FE analyses were conducted
which explicitly included both sides of the excava-
tion, ratherthanacenterlinesymmetriccondition. Soil
responses were assumed to correspond to that of a
Mohr-coulombmaterial. Computeddifferencesof lat-
eral wall movementsoneachsidedifferedbyasmuch
as 50%, as was verified by field observations made
duringconstruction.
Hsi et al. presentedThree-dimensional finiteele-
ment analysis of diaphragmwalls for top-downcon-
struction. They discussed theTugun BypassTunnel
inGoldCoast, Australia. Thetunnel wasconstructed
using diaphragmwalls with the top-down cut-and-
cover method to allow simultaneous construction of
an airport runway extension above the tunnel, and
excavation of the tunnel beneath. The tunnel was
built indeepdeposits of saturated, alluvial andestu-
arine soils with the toes of the walls founded in
soil deposits. Therewas apotential risk for differen-
tial settlements between the diaphragmwall panels,
caused by the runway fill placed over the tunnel
roof duringexcavation. Three-dimensional numerical
modelingwasundertakenwithPlaxis3DFoundation
to predict the differential settlements of the tunnel
arising fromthe variable subsurface profile, staged
excavationanddewatering, non-uniformloadingand
soil-structureinteraction. Soil wasassumedtobehave
as a Hardening-Soil material. Settlements measured
after construction were within the range of those
computedwiththefiniteelement simulations.
Phienwej presentedGroundmovementsinstation
excavationsof Bangkok first MRT. Thecharacteris-
tics of thelateral movements of thediaphragmwalls
at excavations for 18 stations of the first Bangkok
undergroundMRT linewereevaluated. Threemodes
of deflectedshapesof thewallswereobservedat dif-
ferent excavation depths, namely a cantilever mode
andbracedmodeswithabulgeinsoftclayandabulge
instiff clay. Theratioof maximumlateral wall deflec-
tionasafunctionof excavationdepthandtheratioof
groundsurfacesettlementtoexcavationdepthandthe
normalizedgroundsurfacesettlement variedwiththe
modeof wall deflection. UndrainedundrainedYoungs
moduli for aMohr-Coulombconstitutiveresponsefor
different soil layers were back-calculated fromwall
movement dataof threeselectedstationsusingthe2-
D Plaxis FE code. Themodulus values, which were
higher than those commonly obtained fromconven-
tional triaxial tests, canbeusedasguidelinefor future
excavationsinBangkok.
Ota et al. presented Consideration of design
method for braced excavation based on monitoring
results. They comparedobservedanddesignvalues
of wall deflections at several cut-and-cover exca-
vations through soft and sensitive clay ground at
theOsakaSubway LineNo.8. A beam-springmodel
was employed in the braced design method which
accounted for the characteristics of the Osaka soft
ground. Whiletherewasgoodagreementbetweenthe
observed data and design values in past results, the
observed wall deflections in this study were larger
thanthat expectedfor constructionsites whereinthe
excavations encountered 10 to 20mthick, soft and
sensitiveclaylayer.Theauthorsdiscusshowtheyeval-
uatedthehorizontal coefficient of subgradereaction
k
h
ontheexcavationsideof softclaylayer.Theauthors
makenewrecommendationregardingselectionof k
h
,
andshowthatthecalculatedwall movementswiththe
revisedvaluesagreewiththeobservations. Theserec-
ommendationsareapplicabletothesoft andsensitive
Osakaclays.
3.2 Numerical analysis
Li andYangpresentedthepaperNumerical evaluation
of dewateringeffect ondeepexcavationinsoft clay.
They described a FLAC3D analysis that modeled
top-downconstructionof a33.7mdeepunderground
transformersubstationinthedowntownareaof Shang-
hai. Therearebothunconfinedandconfinedaquifers
onthesiteof thisproject anddrainagebydesiccation
inthefoundationpit wasadopted. AssumingaMohr-
Coulomb soil response, the effective stress methods
of analysisincorporatedexcavationanddewateringof
thefoundationpitaspartof thesimulationof construc-
tion activities. Thecomputed wall deflections, basal
heavesandsurfacesettlementsbasedonanalysesthat
didnot consider dewateringwerecomparedto those
that did. Results of analyses that considered leakage
throughthewall andleakagebetweentheaquifersare
presentedaswell.Theanalysisshowsthatalthoughthe
computed differences in lateral wall movement and
basal heaveweresmall, dueto thelowpermeability
of thesoil, dewateringincreasedtheamount of com-
putedsurfacesettlements as aresult of drawdownof
thewater outsidethewallsof theexcavation.
Li et al. contributedthepaper Analysisof thefac-
tors influencing foundation pit deformations. They
presented results of FE computations based upon
3-D Biots consolidation theory, assuming the soil
responded as a nonlinear Duncan-Changs material.
The finite element equations explicilty considered
thecoupling of groundwater seepageand soil skele-
ton deformation during excavation. They presented
89
resultsthat showedtheindividual effectsof theinflu-
enceof soil permeability, rigidityandlevelsof lateral
supports, rigidity of retaining wall and excavation
durationongroundsurfacesettlement, wall horizontal
displacement andbasal heaveof anexcavation.
Siemiska-Lewandowsk and Mitew-Czajewska
presented the paper The effect of deep excavation
onsurroundinggroundandnearby structures. They
describedproblemsrelatedwiththeconstructionof a
29mdeep excavation of Nowy Swiat Station (S11)
of 2nd metro line in Warsaw. A critical section of
theproject consistedof 7stationsand6runningtun-
nels 6kmlength in total. Running tunnels will be
constructed usingTBM while the stations are to be
constructed using cut and cover techniques. Deep
excavation will be made with diaphragmwalls sup-
ported by several levels of slabs and struts. They
presentedresults of 2-D Plaxis FE analyses interms
of ground surfacesettlements, displacements of sur-
rounding foundations and lateral wall movements,
assumingthesoil behavesasaMohr-Coulombmate-
rial. Additionally, settlements of the surface were
calculatedabovetheTBM (runningtunnels). Result-
ingvaluesof settlementsinbothcaseswerediscussed,
andformedthebasisof designpredictionsthatwill be
verifiedduringconstruction.
3.3 Back analysis
Zghondi et al. presented the paper, Multi-criteria
procedure for the back-analysis of multi-supported
retaining walls. They described a numerical back-
analysis procedurefor multi-supported deep excava-
tions basedontheoptimizationof several indicators,
taking in account the forces in the struts and the
differential pressures derivedfromthewall displace-
ment. The evaluation of the procedure is based on
results of 1g small scale laboratory experiments on
semi-flexibleretainingwalls embeddedinaSchnee-
belli material. Theproposednumerical procedurewas
appliedtoanexcavationwith2levelsof strutswithlow
stiffness.TheoptimizedHardeningSoil Model param-
etersformthebasisof calculationsof responseof 14
different tested configurations. Theresults arecom-
paredwiththeclassical methods, SubGradeReaction
Method, FiniteElement analysiswithMohr Coulomb
model and with the back-analysis using Hardening
Soil Model parametersbasedonbiaxial testsresults.
Zhangetal. contributedthepaper Studyondefor-
mation laws under the construction of semi-reverse
method. Taking a 24.1-m-deep foundation pit of
Shanghai Metro Line1 which uses thesemi-reverse
construction process of three open excavating-one
tunneling as anexample, they determineddeforma-
tionlawsof afoundationpitunder theconstructionof
asemi-reversemethodbasedonanalysisof fieldmon-
itoringdataandforwardandback analyses methods.
They employedPlaxis v8andassumedthesoil acted
as a Hardening-Soil material in their computations.
Theauthors stated that results of this approach indi-
catedthatthesemi-reversemethodisaneffectiveway
toimproverigidityof theexterior support, control the
deformation of excavation, and ensure safety of the
surroundingbuildingsandpipelines.
3.4 Measurements
Zhanget al. contributedthepaper GPSheight appli-
cation and gross error detection in foundation pit
monitoring. The authors introduced a deformation
monitoring model that combined traditional survey
technology and GPS measurements. They illustrated
foundationpit deformationmonitoringbasedontheir
experienceof deepfoundationpit constructionof an
underground tunnel in Lishui Road, Hangzhou city.
WhenanalyzingGPSheightconversiontoimprovethe
reliability of theGPS datum, they employedDixons
test in the GPS datummark to determine potential
height anomalies. The authors concluded that this
approachisaconvenientwaytosearchanddeleteraw
datathat includesgrosserrors.
3.5 Design
Changcontributedthepaper Optimizationdesignof
compositesoil-nailing in loess excavation. Excava-
tions through loess haveuniquecharacteristics com-
pared with theothers dueto its structural properties
andcollapsibility. Toevaluatethemechanismsof sup-
port and to develop reasonable methods to design
compositesoil-nailinginloessexcavation, theauthors
usedresultsof finiteelement analysistodesignasoil
nail support system. Their optimizationdesignmeth-
odsarebasedontheresultsof finiteelement analysis
apparently assuming Mohr-Coulomb soil responses.
Theyconductedthesimulationstodeterminetheregu-
larityof deformationandthesafetyfactor, asfunctions
of selecteddesignvariables.Theauthorsjustifiedtheir
methodsby reportingthat thelateral deformationsof
theexampleexcavationwerelimitedto16mm.
3.6 Stress path
Zhouetal. contributedthepaper Comparisonof the-
oryandtestonexcavationcausingthevariationof soil
mass strength. Inviewof thecharacteristic unload-
ingcausedby excavations, they deducedthestrength
ratioof theunloadedsoil tosoil subjectedtocompres-
sionloadings. Laboratorytestssimulatingexcavation
werecarriedout basedonHvorslevsstrengththeory.
By comparingtheoretical results withthelaboratory
data, theyconcludedthatthesoil massisoverconsoli-
dated.Asaconsequence,theauthorsstatedthatthesoil
microstructureisdamaged, andthesoil massstrength
90
isreducedintheunloadingprocess. Theauthorscon-
cluded that analyses of theresults arehelpful to the
understanding of the effect of excavation unloading
onthevariationof thesoil massstrength.
3.7 Earth pressure
LinandLeecontributedthepaper A simplifiedspa-
tial methodology of earth pressure against retaining
piles of pile-row retaining structure. When using a
pile-rowretainingstructuretosupport excavation, the
authorsstressedtheimportanceof obtainingthemag-
nitude and distribution of the earth pressure against
the retaining piles. Based on the mode of failure, a
new methodology is proposed to evaluate the earth
pressure against the retaining piles of such a struc-
ture. In theproposed method, both thespatial effect
andintermediateprincipal stresseffectareconsidered.
Theauthors provideanexampleof themethodology
applied to engineering practice. They demonstrated
that thestrength theory has moreinfluenceon earth
pressure.
4 COMMENTSREGARDINGCOMPUTED
ANDOBSERVEDRESULTS
Many factors affect ground movements caused by
excavations, including stratigraphy, soil properties,
support systemdetails, construction activities, con-
tractual arrangementsandworkmanship.Inthistheme,
most papers described numerical simulations that
analyzed ground response arising from excavation.
Finiteelement predictions always contain uncertain-
ties related to soil properties, support systemdetails
andconstructionprocedures. Furthermore, whilesup-
portedexcavations commonly aresimulatednumeri-
cally by modeling cycles of excavation and support
installation, it generally is necessary to simulate all
aspectsof theconstructionprocessthataffectthestress
conditionsaroundthecuttoobtainanaccuratepredic-
tionof behavior.Thismayinvolvesimulatingprevious
construction activities at the site, installation of the
supportingwall andanydeepfoundationelements, as
well astheremoval of cross-lotsupportsor detension-
ingof tiedbackgroundanchors. Issuesof timeeffects
causedbyhydrodynamiceffectsor material responses
may be important. The following sections summa-
rize some of the factors that may impact computed
responsesof groundmovementsassociatedwithexca-
vations. Proper consideration must be given to such
factors whenmakingsuchanalyses, as well as when
criticallyevaluatingpublishedresultsof thesame.
4.1 Drainage conditions
Animportantpreliminarydecisioninanyanalysisisto
match theexpected field drainageconditions, which
impacts theformulation required. Clough and Mana
(1976) and field data have shown that for excava-
tions throughsaturatedclays withtypical excavation
periods of several months, the clays remain essen-
tially undrainedwithlittledissipationof excess pore
pressures.
For undrainedconditions, onecanemploy either a
coupled finite element formulation where both dis-
placements and pore water pressures are solved for
explicitly(e.g. Carter et al. 1979) or apenaltyformu-
lation(e.g. Hughes1980) whereinthebulkmodulusof
water or asufficientlylargenumber thatdependson
theprecisionof themachinemakingthecomputation-
isaddedtothediagonal termsintheelement stiffness
matrixduringglobal matrixassembly. Thisadditional
termconstrains thevolumetric strain to nearly zero,
i.e., undrained. Inboththeseapproaches, theconstitu-
tiveresponseof thesoil isdefinedintermsof effective
stress parameters. A simpler, alternateapproachis to
defineundrainedconstitutiveresponseintermsof total
stress parameters, withcarebeingtakento makethe
diagonal terms of theelement stiffness matrix large,
typicallybyusingaPoissonsratiocloseto0.5. Inthis
case, aYoungsmoduluscorrespondstoanundrained
valueandfailureisexpressedintermsof anundrained
shear strength, S
u
(e.g., =0andc=S
u
).
However, theremay becases (e.g., ORourkeand
ODonnell 1997) wheresubstantial delaysduringcon-
struction occur and excess pore pressures partially
dissipate, andinthesecasesonemustuseamixedfor-
mulationtoaccount for theporewater effects. When
usingtop-downtechniquestoexcavate,itcantakeupto
several yearstoreachfinal gradeforlargeexcavations,
and hence partially drained conditions would apply
therein, requiringacoupledfiniteelementsimulation.
4.2 Initial conditions
A reasonable prediction of the ground response to
construction of adeepexcavation starts with agood
estimate of the initial stress conditions, in terms of
botheffectivestresses andporewater pressures. The
effective stress conditions for excavations in well-
developed urban areas rarely correspond to at-rest
conditionsbecauseof themyriadpastusesof theland.
Existenceof deepfoundationsand/or basementsfrom
abandonedbuildings andnearby tunnels changes the
effectivestresses fromat-rest conditions prior to the
start of excavation. For example, Calvello andFinno
(2003) showedthatanaccuratecomputationof move-
ments associated with an excavation could only be
achievedwhenall thepre-excavationactivitiesaffect-
ing the site were modeled explicitly. They used the
caseof theexcavationfor theChicago-Statesubway
renovation project (Finno et al. 2002), wherein con-
structionof bothatunnel andaschool impactedthe
groundstressespriortothesubwayrenovationproject.
91
Ignoringtheseeffectsmadeadifferenceof afactor of
3inthecomputedlateral movements.
One also must take care when defining the ini-
tial ground water conditions. Even in cases where
thegroundwater level isnot affectedby near surface
constructionactivities, non-hydrostaticconditionscan
existfor avarietyof reasons. For example, Finnoetal.
(1989) presentedpneumaticpiezometerdatathatindi-
cated thepresenceof adownward gradient within a
20mthick sequence of saturated clays. This down-
ward flow arose froma gradual lowering since the
1950s of thewater level intheupper rock aquifer in
theChicago area. A non-hydrostatic water condition
affects themagnitudeof theeffectivestresses at the
start of anexcavationproject.
An engineer has two choices to define such con-
ditions tomeasurethein situ conditionsdirectly or
tosimulateall thepast constructionactivitiesat asite
startingfromappropriateat-rest conditions. Because
bothapproachespresentchallenges, itisadvantageous
todobothtoprovidesomeredundancyintheinput. In
any case, careful evaluation of the initial conditions
is required when numerically simulating supported
excavationprojects, especiallyinurbanareas.
4.3 Wall installation
Manytimestheeffectsof installingawall areignored
inafiniteelementsimulationandthewall iswished-
into-place with no change in the stress conditions
in the ground or any attendant ground movements.
However, there are abundant data that show ground
movementsmaydevelopasawall isinstalled.
ORourke and Clough (1990) presented data that
summarized observed settlements that arose during
installationof fivediaphragmwalls. They notedset-
tlementsaslargeas0.12%of thedepthof thetrench.
Theseeffectscanbeevaluatedby3-dimensional mod-
elingof theconstructionprocess(e.g., Gourvenecand
Powrie 1999), but not without several caveats. The
specificgravityof thesupportingfluidusuallyvaries
duringexcavationof apanel as aresult of excavated
solids becomingsuspended increasingthespecific
gravityabovethevalueof thewaterandbentonitemix-
tureand subsequently decreasing when theslurry is
cleanedprior totheconcretebeingtremiedintoplace.
Consequently,itisdifficulttoselectonevaluethatrep-
resentsanaveragecondition. Furthermoretheeffects
of thefluidconcreteonthestressesinthesurrounding
soil dependuponhowquicklytheconcretehardensrel-
ativetoitsplacementrate. Someguidanceinselecting
thefreshconcretepressureisprovidedbyLingset al.
(1994).
Itislessstraightforwardwhenmodelingdiaphragm
wall installation effects in a plane strain analysis
becausethearchingcausedbytheexcavationof indi-
vidual panels cannot be taken directly into account.
Toapproximatetheeffectsof thisarchingwhenmak-
ing such an analysis, an equivalent fluid pressure,
generallyhigher thanthelevel of thefluidduringcon-
struction, canbeappliedtothewallsof thetrenchto
maintainstability.Thus, somedegreeof empiricismis
requiredtoconsidertheseinstallationeffectsinaplane
strainanalysis. Onecanback-calculateanequivalent
fluidpressurecorrespondingto theobservedground
responseif good records of lateral movements close
to the wall are recorded during construction. More
dataof thistypeareneededbeforeanyrecommenda-
tionscanbemaderegardingmagnitudesof appropriate
equivalent pressures.
Theeffects of installing asheet pilewall aredif-
ferent thanthoseof adiaphragmwall, yet theeffects
on observed responses also can be significant. In
this case, ground movements may arise fromtran-
sient vibrationsdevelopedasthesheetingisdrivenor
vibratedintoplaceandfromthephysical displacement
of the ground by the sheeting. The former mecha-
nismis of practical importance when installing the
sheeting through loose to mediumdense sands, and
canbeestimatedby procedures proposedby Clough
et al. (1989). However, theseeffects arenot included
in finite element simulations. The latter mechanism
inclayswasillustratedby Finnoet al. (1988). Inthis
case, thesoil wasdisplacedawayfromthesheetingas
it wasinstalled. Thismovement wasaccompaniedby
anincreaseinporewaterpressureandagroundsurface
heave. As theexcess porewater pressures dissipated,
the ground settled. The maximumlateral movement
andsurfaceheavewasequal toone-half theequivalent
width of the sheet pile wall, defined as the cross-
sectional areaof thesheetpilesectionperunitlengthof
wall. Sheet-pileinstallationcanbesimulatedinplane
strainby usingprocedures summarizedinFinno and
Tu(2006).
Inadditionto themovements that occur as awall
isinstalled, installingthewallscanhavealargeinflu-
enceonsubsequentmovements, especiallyif thewalls
areinstalledrelatively closetoeachother, as may be
the case in a cut-and-cover excavation for a tunnel.
Sabatini (1991) conducted a parametric study as a
function of thedepth, H, to width, B, of an excava-
tion, whereintheeffectsof sheet-pilewall installation
in clays were compared with simulations where the
wallswerewishedintoplace. Theresultsof thestudy
areshowninFigure1wherethecomputednormalized
maximumlateral movements, H
(max)
/H, are plotted
versusH/B.
It is apparent for wide excavations (H/B0.25)
that the decision to include installation effects in
a simulation is not critical. However, these effects
becomepronouncedfor narrowexcavations(H/B1)
andshouldbeexplicitly considered. Theresults also
show that for the wished-in-place case when the
sheet-pile installation effects are ignored, the lateral
92
0
1
2
3
0 0.25 0.5 0.75 1
H/B
H
(
M
A
X
)
/
H

[
%
]
Sheet-pile effects
No sheet-pile effects
Figure 1. Effects of sheet-pile installation on computed
lateral movements.
-5
0
5
10
15
20
25
-5 0 5 10 15 20 25 30 35 40
Distance From South End of Excavation (m)
D
e
p
t
h

(
m
)
Day 28
Day 57
Day 58
Day 60
Day 75
Day 80
Day 88
Day 89
Day 95
Day 110
Day 113
Day 163
Tiebacks
Struts
Inclinometer 2
Location
Inclinometer 1
Location
Strut 1
(56)
Strut 2
(58)
Strut 3
(59)
Strut 4
(60)
Strut 5
(74)
Strut 6
(79)
North
South Scale (m)
0 5
Note:
Upper level tiebacks stressed between Day 81 and Day 105
Lower level tiebacks stressed between Day 98 and Day 113
Initial
Grade
East
Escalator Pit
Figure2. Construction progress at excavation in Chicago
(Finnoet al. 2002).
movementsarelarger for wider excavations, asimilar
trendreportedbyManaandClough(1981).
Sheet-pileinstallationhastwomaineffects: thesoil
adjacent to theexcavationis preloadedandtheshear
strength on the passive side is (partially) mobilized
prior to the beginning of the cycles of excavation.
Wall installationtendstopreloadthesoil ontheactive
sideof theexcavation as aresult of thereduction in
shear stress at approximately constant mean normal
effectivestress.Thismechanismprovidesthesoil out-
sidethewallswithmoreavailableshearingresistance
whenthecyclesof excavationstart. However, thesoil
betweenthewallshaslessavailablepassiveresistance
as a result of the preloading and this promotes the
larger movements during excavation as compared to
thecaseof ignoringthesheet-pileeffects (Finnoand
Nerby1989).
4.4 Plane strain versus 3-dimensional analyses
Figure 2 illustrates some of the challenges of using
field observations to calibrate numerical models of
anykind, evenwhendetailedrecordsexist.Thisfigure
summarizestheconstructionprogressattheChicago-
State excavation in terms of excavation surface and
support installationononeof thewallsof theexcava-
tionfor selecteddays after constructionstarted. Also
shown arethelocations of two inclinometers placed
several meters behind the wall. If one is making a
computation assuming plane strain conditions, then
it is clear that onemust judiciously choosedatasets
sothat planar conditionswouldbeapplicabletothose
selected.
Evenwhenasufficientlyextensivehorizontal exca-
vatedsurfaceisidentified, 3-dimensional effectsmay
still arisefromthehigher stiffnessatthecornersof an
excavation.Theseboundaryconditionsleadtosmaller
groundmovementsnear thecornersandlarger ground
movementstowardsthemiddleof theexcavationwall.
Another, andlessrecognized, consequenceof thecor-
ner stiffeningeffectsisthemaximummovement near
thecenter of anexcavationwall may not correspond
to that found froma conventional plane strain sim-
ulation of the excavation, i.e., 3-D and plane strain
simulations of the excavation do not yield the same
movementatthecenterportionof theexcavation, even
if themovementsinthecenterareperpendiculartothe
wall (Ouet al. 1996). Thisaffect canbequantifiedby
theplanestrainratio, PSR, definedhereinasthemax-
imummovement in thecenter of an excavation wall
computedby 3-D analyses dividedby that computed
byaplanestrainsimulation. Finnoetal. (2007) devel-
opedthefollowingexpressionforPSRfromtheresults
of a finite element parametric study of excavations
throughclay:
whereL istheexcavationlengthalongthesidewhere
themovement occurs, B istheother areal dimension,
andH
e
istheexcavationdepth.Thevalueof C depends
onthefactor of safetyagainst basal heave, FS
BH
, and
istakenas:
Thevalueof k dependsonthesupportsystemstiffness
andistakenas:
whereEI isthebendingstiffnessof thewall, isthe
total unitweightof thesoil andh istheaveragevertical
spacingbetweensupports. WhenL,H
e
isgreater than
6, thePSRisequal to1andresultsof planestrainsim-
ulationsyieldthesamedisplacementsinthecenter of
anexcavationasthosecomputedbya3-Dsimulation.
WhenL,H
e
islessthan6, thedisplacementcomputed
fromtheresultsof aplanestrainanalysiswill belarger
than that froma 3-D analysis. When conducting an
inverseanalysis of an excavation with aplanestrain
93
simulationwhenL,H
e
is relatively small, theeffects
of thiscorner stiffeningisthat anoptimizedstiffness
parameter will be larger than it really is because of
the lack of the corner stiffening in the plane strain
analysis. Thiseffectbecomesgreater asanexcavation
is deepenedbecausetheL,H
e
valueincreases as the
excavatedgradeis lowered. This trendwas observed
intheoptimizedparametersforthedeeperstrataatthe
Chicago-State subway renovation excavation (Finno
andCalvello2005).
5 CONSTITUTIVE MODEL
CONSIDERATIONS
Whenoneundertakesanumerical simulationof adeep
supportedexcavation, oneof thekey decisions made
earlyintheprocessistheselectionof theconstitutive
model. Ingeneral, thisselectionshouldbecompatible
withtheobjectivesof theanalysis. If theresultsform
thebasisof apredictionthat will beupdatedbasedon
fieldperformancedata, thenthetypesof fielddatathat
formthebasisof thecomparisonwill impacttheappli-
cability of a particular model. Possibilities include
lateral movements based on inclinometers, vertical
movements at various depths and distances froman
excavationwall, forcesinstructural supportelements,
porewaterpressuresoranycombinationsof thesedata.
Whenusedfor acasewherecontrol of groundmove-
ments is akey design consideration, theconstitutive
model must beabletoreproducethesoil responseat
appropriatestrainlevelstotheimposedloadings.
5.1 Incremental non-linearity
It is useful to recognizethat soil is an incrementally
nonlinear material, i.e., itsstiffnessdependsonload-
ing direction and strain level. Real soils are neither
linear elastic nor elastic-plastic, but exhibit complex
behavior characterized by zones of high stiffness at
very small strains, followed by decreasing stiffness
withincreasingstrain.Thisbehaviorunderstaticload-
ingwasidentifiedthroughback-analysisof foundation
and excavation movements in the United Kingdom
(Burland, 1989). The recognition of zones of high
initial stiffnessunder typical fieldconditionswasfol-
lowedby efforts to measurethis ubiquitous behavior
inthelaboratoryfor varioustypesof soil (e.g., J ardine
et al. 1984; Clayton and Heymann 2001; Santagata
et al. 2005; CalistoandCalebresi 1998, Cho2007).
To illustrate this behavior, Figure 3 shows the
resultsof drained, triaxial stressprobesconductedon
specimens cut fromblock samples of lightly over-
consolidated glacial clays obtained at an excavation
in Evanston, IL. Each specimen was reconsolidated
under K
0
conditions to the in-situ vertical effective
stress
v0
, subjected to a 36 hour K
0
creep cycle,
followed by directional stress probing under drained
0.0001 0.001 0.01 0.1 1
Local Triaxial Shear Strain,
sl
(%)
0
10
20
30
40
50
60
70
80
90
S
e
c
a
n
t

S
h
e
a
r

M
o
d
u
l
u
s
,

G
s
e
c

(
M
P
a
)
FB3AL1
FB1TC2
FB1CMS2
FB1RTC2
FB1RTE1
FB1TE1
FB2CMSE1
unloading
loading
G
BE
Range
Figure3. Secant shear modulus as afunctionof direction
of loading.
axisymmetric conditions. Bender element (BE) tests
were conducted during all phases of the tests. The
secant shear moduli areplotted versus triaxial shear
straininFigure3for natural specimens whosestress
probes involved changes in the shear stress q. The
overconsolidation ratio of these specimens was 1.7,
soif oneassumestheresponseisisotropicandelasto-
plastic, thenGshouldbethesameforatleasttheinitial
portionof all curves. Thestressprobeswhereinqand
the stress ratio, =q,p
, is increased ( loading)
are clearly softer than those where q and initially
decrease( unloading). Therearenoobviouszones
of constant G
sec
at shear strainsgreater than0.002%,
andthusnoelastic zoneisobservedinthesedatafor
strainlevels.Completedetailsandresultsof thetesting
programarepresentedbyCho(2007).
Burland(1989) suggestedthatworkingstrainlevels
insoil aroundwell-designedtunnelsandfoundations
areontheorder of 0.1%. If oneuses datacollected
with conventional triaxial equipments to discern the
soil responses, onecanreliably measurestrains0.1%
or higher. Thus inmany practical situations, it is not
possibleto accurately incorporatesite-specific small
strainnon-linearityintoaconstitutivemodel basedon
conventionally-derivedlaboratorydata.
5.2 Model selection
There are a number of models reported in litera-
turewhereinthevariationof small strainnonlinearity
can be represented, e.g., a three-surface kinematic
model developfor stiff Londonclay (Stallebrass and
Taylor 1997), MIT-E3(WhittleandKavvadas 1994),
hypoplasticity models (e.g. Viggiani and Tamagnini
1999), andadirectional stiffnessmodel (Tu2007). To
derivethenecessaryparameters, thesemodelsrequire
eitherdetailedexperimental resultsorexperiencewith
themodel in agiven geology. Whilethemodels can
beimplementedinmaterial librariesinsomecommer-
cial finiteelementcodes, theseroutinesarenotreadily
94
Figure 4. Shear strains behind excavation: 57mmmaxi-
mumlateral movement (contoursin%).
availabletomost practitioners. Thusfor most current
practical applications, oneusessimpler, elasto-plastic
models containedinmaterial libraries incommercial
codes.
For these models, a key decision is to select the
elastic parameters that are representative of the
secantvaluesthatcorrespondtothepredominantstrain
levels inthesoil mass. Examples of thestrainlevels
behindawall for anexcavationwithamaximumlat-
eral wall movement of 57mmareshowninFigure5.
Thesestrainlevelswerecomputedbasedontheresults
of displacement-controlledsimulationswherethelat-
eral wall movements and surface settlements were
incrementallyappliedtotheboundariesof afiniteele-
ment mesh. Thepatterns of movements weretypical
of excavationsthroughclays, andwerebasedonthose
observedatanexcavationmadethroughChicagoclays
(FinnoandBlackburn2005). Becausethesimulations
weredisplacement-controlled,thecomputedstrainsdo
not dependontheassumedconstitutivebehavior.
AscanbeseeninFigure4, maximumshear strains
as high as 0.7% occur when 57mm of maximum
wall movement develop. The maximum strains are
proportional maximum lateral wall movement; for
example, when 26mmmaximumlateral wall move-
ment develops, the maximumshear strain is about
0.35%. These latter strain levels can be accurately
measuredinconventional triaxial testing, andthus if
onecanobtainspecimensof sufficientlyhighquality,
thensecantmoduli correspondingtothesestrainlevels
canbedeterminedviaconventional laboratorytesting.
Becausethemaximumhorizontal wall displacement
can be thought of as a summation of the horizontal
strains behindawall, themaximumwall movements
canbeaccurately calculatedwithaselectionof elas-
tic parameters that corresponds to these expected
0
20
40
60
80
100
120
0 0.2 0.4 0.6 0.8 1.2
Maximum Shear Strain (%)
M
a
x
i
m
u
m

L
a
t
e
r
a
l

M
o
v
e
m
e
n
t
(
m
m
)
1
Figure5. Relationbetweenmaximumwall movement and
shear strain.
strain levels. In this case, the fact that small strain
non-linearityisnotexplicitlyconsideredwill nothave
alargeimpact on thecomputed horizontal wall dis-
placements because the maximumhorizontal move-
ment at thewall is dominatedby thelarger strains in
thesoil mass. Thesecomputedmovements wouldbe
compatible with those measured by an inclinometer
locatedclosetothewall.
However, if oneneedstoobtainanaccuraterepre-
sentationof thedistributionof groundmovementswith
distancefromthewall, then this approach of select-
ing strain-level appropriate elastic parameters is not
applicable. Small strainnon-linearity of soil must be
explicitly considered to find theextent of thesettle-
ment becausethestrains in theareaof interest vary
fromthemaximumvalueto zero. As aconsequence,
many cases reported in literature indicate computed
wall movementsagreereasonablywell withobserved
values, buttheresultsfromthesamecomputationsdo
not accurately reflect thedistribution of settlements.
Indeed, thiswasthecaseinseveral paperspresentedas
part of thistheme. Goodagreement at distancesaway
froma wall can be obtained only if the small stain
non-linearity of thesoil is adequately represented in
theconstitutivemodel.
Therelationbetweenlateral wall displacementsand
shear strain levels in the soil behind the wall can
beevaluatedfromresults of displacement-controlled
finiteelementsimulations.Similartotheresultsshown
inFigure4, differentdisplacementprofileswerestud-
iedbyimposinglateral wall displacementsandsettle-
mentprofiles, representingconditionswithmaximum
lateral movementsattheexcavatedsurface, cantilever
movements, deep-seatedmovementsandcombination
of thelatter two(Andrianis2006). Thestratigraphies
used in the models were based on typical Chicago
soils. Theresults in Figure5 show that therelation-
ship between maximumshear strain behind thewall
andmaximumdisplacementof thewall isalmostlinear
forlateral wall displacementsbetween10and110mm.
Figure 5 also shows that the results forma narrow
95
band, suggestingthat therelationbetweenstrainand
wall displacement is not greatly affectedby thetype
of movement.
Figure5canbeusedtoestimateshear strainsfor a
specifiedmaximumwall movement.Withthisvalueof
shearstrain, thesecantshearmoduli foruseinconven-
tional elasto-plasticmodelscanbeestimatedbasedon
strain-stressdatafromhighqualitylaboratoryexperi-
ments. Thevaluesof maximumshear strains, evenin
the cases with the relatively low values of displace-
ments, areabout 0.2%and increaseas thespecified
displacement becomeslarger. Thisisimportant when
onedetermines soil stiffness in thelaboratory. Con-
ventional soil testingwithoutinternal instrumentation
allows one to accurately measure strains as low as
0.1%. Thus for many cases, the secant shear mod-
uli can be determined fromconventional laboratory
testsonhighqualitysamples. However, if strainlevels
are0.1%or less, then onemust select thesemoduli
fromtest resultsbasedoninternally-measuredstrains
in equipment not normally available in commercial
laboratories.
In summary, using a simulation based on con-
ventional elasto-plastic models limits the type and
location of the data that can be used as observa-
tions inaninverseanalysis. Bothvertical andlateral
movements measured at some distance froma wall
cannot becalculated accurately in this casebecause
thevariationof stiffnesswithstrainlevelsmustbeade-
quatelyrepresentedinthesoil model. Onlythelateral
movementsclosetoasupport wall canbereasonably
computedwithconventional modelssincethatresultis
dominatedbythezonesof highstrainsbehindthewall.
6 CONCLUDINGREMARKS
The papers presented at the symposium included
widely variable levels of information regarding the
details of the finite element analyses. As such, the
author tentatively proposes that the following infor-
mationbeincludedinanypaper describingtheresults
of any finite element simulation of geotechnically-
relatedconstruction.
1 Thefiniteelement codeused.
2 The assumed drainage conditions, e.g., drained,
undrainedor partiallydrained.
3 The dimensionality of the problem, e.g., plain
strain, axisymmetricor three-dimensional.
4 Theconstitutivemodel(s) employedfor bothsoils
andstructural elements.
5 Theparametersfor eachmaterial andadiscussion
of thebasisof their selection.
6 A descriptionof themesh, includingboundarycon-
ditionsandtypeof elementsusedforsoil, structural
componentsandinterfaces.
7 Constructionrecords, simulationsteps anddetails
of howeachconstructionactivity wasidealizedin
thefiniteelement simulation.
Finally, comparisons between computed and
observedresults, as well as adiscussionof thecom-
parisons, shouldbeincluded.
REFERENCES
Andrianis, G.E. 2006. Excavation-induced Strains and Can-
tilever Deflections in Compressible Clays. MS thesis,
NorthwesternUniversity, Evanston, IL.
Burland, J.B. 1989. Small isbeautiful thestiffnessof soils
atsmall strains: NinthLauritsBjerrumMemorial Lecture.
Canadian Geotechnical Journal 26: 499516.
Calisto, L. & Calebresi, G. 1998. Mechanical behavior of a
natural soft clay. Geotechnique 48(4): 495513.
Calvello, M. & Finno, R.J. 2003. Modeling excavations in
urbanareas: effectsof pastactivities. Italian Geotechnical
Journal 37(4): 923.
Carter, J.P., Booker, J.R. & Small, J.C. 1979. Theanalysisof
finiteelasto-plastic consolidation. International Journal
for Numerical and Analytical Methods in Geomechanics
3: 107129.
Cho, W.J. 2007. Recent stress history effects on compress-
ible Chicago glacial clay. PhD thesis, Northwestern
University, Evanston, IL.
Clayton, C.R.I. & Heymann, G. 2001. Stiffnessof geomate-
rialsat verysmall strains. Geotechnique 51(3): 245255.
Clough, G.W. & Mana, A.I. 1976. Lessons learnedinfinite
element analysisof temporaryexcavations. Proceedings,
2nd International Conference on Numerical Methods in
Geomechanics, ASCE, Vol. I: 496510.
Clough, G.W., Smith, E.M. & Sweeney, B.P. 1989. Move-
ment control of excavation support systems by iterative
design. Current Principles and Practices, Foundation
Engineering Congress, Vol. 2, ASCE: 869884.
Finno, R.J., Atmatzidis, D.K. & Nerby, S.M. 1988. Ground
response to sheet-pile installation in clay. Proceedings,
Second International Conference on Case Histories in
Geotechnical Engineering, St. Louis, MO.
Finno, R.J.,Atmatzidis, D.K. &Perkins, S.B. 1989. Observed
Performance of a Deep Excavation in Clay. Journal of
Geotechnical Engineering, ASCE, 115(8): 10451064.
Finno, R.J. & Blackburn, J.T. 2005. Automated monitor-
ing of supported excavations. Proceedings, 13th Great
Lakes Geotechnical and Geoenvironmental Conference,
Geotechnical Applications for Transportation Infrastruc-
ture, GPP 3, ASCE, Milwaukee, WI.: 112.
Finno, R.J., Blackburn, J.T. & Roboski, J.F. 2007. Three-
dimensional Effects for Supported Excavations in Clay.
Journal of Geotechnical andGeoenvironmental Engineer-
ing, ASCE, 133(1): 3036.
Finno, R.J., Bryson, L.S. & Calvello, M. 2002. Performance
of astiff support systeminsoft clay. Journal of Geotech-
nical and Geoenvironmental Engineering,ASCE, 128(8):
660671.
Finno, R.J. &Calvello, M. 2005. Supportedexcavations: the
observational method and inverse modeling. Journal of
Geotechnical and Geoenvironmental Engineering.ASCE,
131(7).
96
Finno, R.J. & Nerby, S.M. 1989. Saturated Clay Response
DuringBracedCutConstruction. Journal of Geotechnical
Engineering, ASCE, 115(8): 10651085.
Finno, R.J. &Tu, X. 2006. SelectedTopicsinNumerical Sim-
ulation of Supported Excavations. Numerical Modeling
of Construction Processes in Geotechnical Engineer-
ing for Urban Environment, International Conferenceof
ConstructionProcesses inGeotechnical Engineeringfor
Urban Environment, Th. Triantafyllidis, ed., Bochum,
Germany, Taylor & Francis, London: 320.
Gourvenec, S.M. & Powrie, W. 1999. Three-dimensional
finite element analysis of diaphragm wall installation.
Geotechnique, 49(6): 801823.
Hughes, T.J.R. 1980. Generalizationof selectiveintegration
procedures to anisotropic and nonlinear media. Interna-
tional Journal of Numerical Methods in Engineering 15
(9): 14131418.
J ardine, R.J., Symes, M.J. & Burland, J.B. 1984. Themea-
surementof small strainstiffnessinthetriaxial apparatus.
Lings, M.L., Ng, C.W.W. & Nash, D.F.T. 1994. Thelateral
pressure of wet concrete in diaphragmwall panels cast
under bentonite. Geotechnical Engineering, Proceedings
of theInstitutionof Civil Engineers, 107: 163172.
Mana,A.I.&Cough,G.W.1981.Predictionof movementsfor
bracedcut inclay. Journal of Geotechnical Engineering,
ASCE, NewYork, 107(8): 759777.
ORourke, T.D. &Clough, G.W. 1990. Constructioninduced
movementsof insituwalls. Proceedings, Design and Per-
formance of Earth Retaining Structures, Lambe, P.C. and
HansenL.A. (eds). ASCE: 439470.
ORourke, T.D. & ODonnell, C.J. 1997. Deep rotational
stability of tiedback excavations in clay. Journal of
Ou, C.Y., Chiou, D.C. & Wu, T.S. 1996. Three-dimensional
finite element analysis of deep excavations. Journal of
Sabatini,P.J.1991.Sheet-pile installationeffects oncomputed
ground response for braced excavations in soft to medium
clays. MSthesis, NorthwesternUniversity, Evanston, IL.
Santagata, M., Germaine, J.T. & Ladd, C.C. 2005. Factors
AffectingtheInitial Stiffnessof CohesiveSoils.Journal of
131(4): 430441.
Stallebrass, S.E. &Taylor, R.N. 1997. Thedevelopment and
evaluation of a constitutive model for the prediction of
ground movements in overconsolidated clay. Geotech-
nique 47(2): 235253.
Tu, X. 2007.Tangent stiffness model for clays including small
strain non-linearity. PhDthesis, NorthwesternUniversity,
Evanston, IL.
Viggiani, G. & Tamagnini, C. 1999. Hypoplasticity for
modelingsoil non-linearity inexcavationproblems. Pre-
failure Deformation Characteristics of Geomaterials,
M. J amiolkowski, M, Lancellotta, R. and Lo Presti, D.
(eds.), Balkema, Rotterdam: 581588.
Whittle,A.J. &Kavvadas, M.J. 1994. Formulationof MIT-E3
constitutivemodel for overconsolidatedclays. Journal of
120(1): 173198.
97
Constructionmethod, groundtreatment, andconditioningfor tunneling
T. Hashimoto& B.Ye
Geo-Research Institute, Osaka, Japan
G.L.Ye
Department of Civil Engineering, Shanghai Jiaotong University, Shanghai, P.R. China
ABSTRACT: This general report reviews a selected group of papers of Session 2 which is related to
ConstructionMethod, GroundTreatment, andConditioningforTunneling.Thepapersaredividedinto5groups
basedontheir topics: (1) constructionmethodswithcasestudies, (2) groundtreatment, (3) loadandpressure,
(4)conditioningadditivesforEPB,(5)others.Beforereviewing,thegeotechnical aspectsinthesefieldsarefirstly
summarized, andthentheessencesof thesepapersarepresented. Thedeficienciesandfuturedevelopmentsare
alsodiscussed.
1 INTRODUCTION
Session2includes19papersfromChina, J apan, UK,
theNetherlands, Germany, Iran, Slovakia, Argentina
and Brazil. Especially, Tongji University, Shanghai
and GeoDelft, the Netherlands contribute to present
somepapers respectively. It is becausethat Shanghai
andtheNetherlands havebeenperformedmany tun-
nelsduringthepast decadeinthesoft ground. These
papers are divided into 5 groups and 8 subgroups
based on their topics, which are shown in Table 1.
Althoughall thesepaperhascontributedtosomespec-
ifiedaspectof constructionmethod, groundtreatment,
andconditioningfor tunneling, somepaperswithsig-
nificantimportanceareselectedtobereviewedinthis
General Report.Thereviewwill becarriedoutaccord-
ing to thegrouping of thepaper. Beforethereview,
the geotechnical aspects in these fields are firstly
summarized.
2 CONSTRUCTIONMETHODWITH
CASE STUDIES
2.1 Bored tunnel by TBM
More and more practices of bored tunnels by TBM
bringforwardmoreandmorerequirementsfor shield
tunnel. Table2displaysthecurrent trendof develop-
ment of shieldtunnel basedontherequirementsfrom
theworldmarket of tunneling.
To meet theserequirements, technologies of TBM
are also developed at the same time. The recent
Table1. Groupingof thepapersinSession2.
Num. of
Topics papers Authors
1. Constructionmethods 6papers
withcasestudies
1.1Boredtunnel by (3) Bakker & Bezuijen
TBM (shield (A, B) Heet al.
tunneling)
1.2Shotcretemethod (3) SfrisoGuatteri et al.
(mountaintunneling Fillibeck&Vogt
method, NATM)
2. GroundTreatment 5papers
2.1Groundfreezing (2) Hu& Pi
Fillibeck&Vogt
2.2Grouting (4) Guatteri et al.
Bezuijen& vanTol
Gafar et al
Fillibeck&Vogt
3. Loadandpressure 7papers
3.1Liningpressure (5) Hashimotoet al.
Talmon& Bezuijen
Talmonet al.
Bakker & Bezuijen
(A, B)
3.2PressureonTBM (4) Bezuijen& Bakker
Song& Zhou
Bakker & Bezuijen
(A, B)
4. Conditioning 2papers Hajialilue-Bonab
additivesfor EPB et al. (A, B)
5. Ohters 3papers Deng& Zhang
Kuzme& Hrustinec
Li et al.
99
Table2. Current trendinshieldtunneling.
Longdistance 3km10km
Highspeedexcavation 300m1000m/month
Deepexcavation 40m100m
Largecrosssection 10m15mof diameter
Deformedcrosssection 2faces4faces, non-circular
Highdurabilityof tunnel 100years
Cost performance Not cheapbut highquality
withreasonablecost
Table3. Geotechnical aspects for bored tunneling (shield
tunneling).
TBM type Bothof slurryandEPB typeinthe
soft soil withgroundwater
Applicableground Soft tostiff clay, loosetodensesand,
gravel
Groundloss Possibletobecontrolledlessthan
0.11%innormal condition
Facestability Needsomecontrollingtechnologies
for eachslurrytypeor EPB type
Fillingtail void Simultaneousgroutingcanreduce
groundlossandgiveanuniform
distributionof liningpressure
Segmental lining Manytypesof segmental lininghave
beendeveloped
development of TBM andits technologies areshown
asfollowing:
Durabilityof TBM
Durabilityof cutter bits
Exchangeablecutter bits
Installationof linings, newsegmental linings
Drivingcontrol system
Dockingmethod
Backfill grouting
In the practice of bored tunneling by TBM, the
geotechnical aspectsshowninTable3areof themost
importanceandshouldbewell considered.
Bakker & Bezuijen (A, B) shared their invaluable
experiences and findings on shield tunneling in soft
groundobtainedinlasttenyears. Duringtheconstruc-
tion of the 2nd Heinenoord Tunnel that is approxi-
matelyinthemiddleunderneaththeriver OudeMaas
intheNetherlands.Theyfoundoutthatbecauseblow-
out occurred during TBM driving under the river,
facesupportpressuredroppedwithin15secondsafter
thecutter faceworking, shownasFigure1.According
totheirinvestigation,theypointedoutthatfacesupport
pressureshouldbecontrolledbetweenlowerandupper
limits for situations withlittleoverburdenor thesoil
cover itself isrelativelylight. Wealsoareinterestedin
the15seconds, whichindictedthat thefront insta-
bility occuredwithout any omen, acareful control of
front pressureis necessary. Someanalysis results of
Figure 1. Support pressures before, during and after the
Blow out at the 2nd Heinenoord tunnel (by Bakker &
Bezuijen (A)).
Figure 2. Surface settlements; measured and back-
calculated with different material models (by Bakker &
Bezuijen (A)).
surfacesettlement werealsodisplayed, showninFig-
ure2. Itwasconcludedthatfor anadequateprediction
of deformationsit isimportant tomodel thegrouting
pressureasaboundarycondition, incombinationwith
theuseof small strainmaterial model.
As to the structure issues of the 2nd Heinenoord
Tunnel, Bakker & Bezuijen (B) investigatedthecrack-
ingandpallingthatoccurredduetoconstructionload,
seeFigure3. Thenalargescaletunnel ringtestswas
carried out, shown as Figure 4. By combining the
model tests as well as numerical tests, it was found
that theusageof kaubit intheringjoint wasthemain
reason.Thecompressionof theflexiblekaubitstripsby
jackingforceresultedinaslippingof differentsegment
piece, leadingtolocal stressconcentrationandirreg-
ular deformation. Byreplacingitwithstiffer plywood
plates, thedamagewasprevented.Theinfluenceof the
durationof plywoodtothelong-termbehavior of tun-
nel, however, isstill questionable. Duringconstruction
of thefirst tubefor theWesternscheldt Tunnel, they
foundout that highgrout pressuresandinabsenceof
beddingmaycausethebucklingof theTBM.Certainly,
100
Figure3. Damageto thedowel and notch sockets during
thefirst 150mof constructionof the2ndHeinenoordtunnel
(byBakker & Bezuijen (B)).
Figure 4. Large-scale tunnel ring testing (by Bakker &
Bezuijen).
someotherfactorsthatwerenotdiscussedinthepaper
mayalsocauseTBM deformation.
He et al. studiedthefirst applicationof DOT tun-
neling in Shanghai. They conducted an in-situ test
to investigatethedistribution of stress and displace-
ment around thetunnel. Figure5 shows thevertical
soil stress increment ahead of cutter face. Beauti-
ful distribution of vertical earth pressure increment
andsettlement troughs wereobserved. It is expected
that more detailed information about the measuring
methodscanbegivenout. They alsoreportedaDOT
shield passed under afive-floor building with adis-
tanceof 1msuccessfullybycareful operation, shown
as Figure6. Themaincountermeasures wererelative
lowadvancingspeedandextrabackfill grouting.
Figure5. Thevertical stressincrementin1.5maheadof the
openingface(byHe et al.).
Figure 6. Dot shield tunnel run across the buildings
(byHe et al.).
2.2 Shotcrete method (Mountain tunnel
method, NATM)
Thegeotechnical aspectsof shotcretemethod(Moun-
taintunnel method,NATM)aresummarizedinTable4.
Thedesign and construction procedures of Metro
tunnels in Buenos Aires from 19982007 were
reported by Sfriso. The characterization of Buenos
Airessoilsfortunnelingisoverconsolidatedcemented
101
Table4. Geotechnical aspectsof shotcretemethod(moun-
taintunnel method, NATM).
Facestability Prelining, facebolt, horizontal
grouting, piperoof, shotcrete,
et al.
Settlement mitigation Facestabilization, shotcrete,
foot pile, rockbolt, minibench,
groundimprovement (jet
grouting, chemical grouting,
compensationgrouting, et al.)
Geological survey Geophysical survey(elastic
wave, sonicwave, electric
resistivity, et al.), pilot boring
Monitoringtechnology Extensometer, 3Dlaser
scanning, optical fiber sensing,
digital photogrammetric
system, et al.
Predictionof groundwater Loweringof groundwater
inflowandpreservation table, subsidence, dryingwell
of groundwater
soil withN
spt
>20, whichisvery favorablefor exca-
vation. As shown in Figure 7, shotcrete tunneling
methods evolved from German method to Belgian
method, and reached an optimal full face excava-
tion. Cut & cover methodandundergroundexcavated
methodwereusedfor stations. Accordingtothefiled
measurement, the surface settlement is in the range
28mmingeneral, 415mmat stations.
Guatteri et al. described the state-of-the-art of
application of ground improvement with all round
(360
) horizontal jet grouting in Sao Paulo and

Barcelona, shown as Figure 8. Horizontal jet grout-
ing columns were executed around the excavated
section, including the invert, and at the far end of
the conical treatment, to create a watertight cham-
ber. This ground improvement achieves good results
of pre-consolidation, settlement mitigation, reduction
of water flow, and keep of face stability. Accord-
ing to field measurement, ground movements were
controlledwithin2030mm.
Shotcrete excavations with ground freezing, jet
grouting, pipescreenandcompressedair supporting
methods wereappliedintheconstructionof Munich
Subway. Fillibech & Vogt made a comparison of
different methods of facesupport in settlement sen-
sitiveurbanareasbasedonthegrounddeformations.
In the case of heading with ground freezing under
important structure, measurementsfor reducingfrost
heaves were taken, namely reducing operation time
and careful temperaturecontrol. Therecorded verti-
cal displacements inFigure9showthat amaximum
heaveof 35mmwasachieved.Thereportof jetgrout-
ing displayed alargeheavedueto installation of jet
groutingcover, as shown in Figure10. Although the
facestabilityincreases,thesettlementisnotreducedso
Figure7. Various shotcretetunnelingmethods inthecon-
structionof MetrotunnelsinBuenosAires(bySfriso).
muchasexpected.Inthepaper,itispointedoutthatthe
installation of crown supporting measures must lead
to higher safety potential, but it is difficult to judge
whether thesespecial measuresarenecessaryor not.
3 GROUNDTREATMENT
The geotechnical aspects of the two sub-subjects of
ground treatment, ground freezing and grouting, are
summarizedinTable5andTable6.
102
Section Tunnel
Septum
ColumnCCPh
ColumnsCCPh
Figure 8. Ground improvement with all round horizontal
jetgroutingappliedinSaoPauloandBarcelona(byGuatteri
et al.).
Figure 9. Vertical ground surface displacement along a
tunnel protectedbygroundfreezing(byFillibech &Vogt).
TheworkbyBezuijen &vanTol aimedtomakeclear
againstthequestionwhyfracturescanoccurmoreeas-
ilyinthefieldthaninmodel testswiththesameW/C
ratio. Startingfromaconceptual model that shownin
Figure11, theydemonstratedinananalytical waythat
heterogeneityof soil inthefieldandthestressreduc-
tionby theinstallationof pipes (so-calledTAM) and
other causesbeforeinjectionaremainreasons.
Garfa et al. performedasereiesof laboratoryscale
grout injection tests in which various factors affect-
ingfracturingof sandwerestudied. Figure12shows
theschematicdiagramof theexperimental setup. The
experimental resultsconfirmedthatfractureinitiation
Figure10. Largeheavedueto horizontal jet grouting (by
Fillibech &Vogt).
Table5. Geotechnical aspectsongroundfreezing.
Mechanismof freezing Segregationpotential, icelens,
processandevaluationof structureof soil, et al.
laboratoryfreezingtest
Propertyof frozensoils Strength, stiffness, freezing
point, temperature, thermal
conductivity, salinity
consistency, et al.
Frost heaveandthaw Laboratorytesting, prediction,
settlement countermeasure
Applicationof freezing Crosspassage, dockingof
methodonunderground TBMs, launchandarrival of
construction TBM
Table6. Geotechnical aspectsongrouting.
J et grouting Uniformityof improvedsoil, ground
deformationduringjet grouting,
applicablegroundcondition
(boulder, obstacle, et al.)
Groundinjection
Material Chemical grout, micro-cement,
CB, LW, polyurethane, et al.
Grouting Penetrationgrouting,
method compactiongrouting, double
packer, et al.
Evaluationof Fracturing, compactioneffects,
improvement uniformity, strengthening, reduction
of permeability
Settlement Compensationgrouting
control
insandrequiressomelocal inhomogeneityaroundthe
injectionpoint, rapiddevelopmentof afiltercakewith
alimitedthicknessandagroutwithlowviscosityand
a limited yield stress. Grouts with high w/c (water-
cement) ratiowill exhibitfractureswiththeformation
of filter cake. If thew/c is low, no fractures will be
formed.
103
Figure11. Sketchwithpossibledeformationmodesof the
injectionhole(byBezuijen & van Tol ).
Figure12. Schematicdiagramof theexperimental setup(by
Safar et al.).
4 LOADANDPRESSURE
Geotechnical aspects onliningpressureandpressure
actingonTBM aresummarizedinTable7.
Hashimoto et al. analyzedaseriesof observedearth
pressuremeasuredbyPADtypeearthpressurecell in
soft clay, stiff clayandsandrespectively. It wasfound
thatthedistributionsof earthpressureareveryuniform
ineachkindof soils. Intheverysoft clay, alargepor-
tionof theoverburdenwill act uponthelining, while
inthestiff clayandthedensesand, themagnitudeand
distribution of earth pressurealso depend largely on
thebackfill grouting, shownasFigures13& 14.
Talmo &Bezuijen presentedaveryinterestingpaper
onthepredictionof thegroundpressure(liningpres-
sure) based on the flow theory of backfill grout in
combinationwiththetimedependentconsolidationof
groutmaterial.Themeasuredresultsandthepredicted
resultsareshowninFigure15. Intheirpaper, theydis-
playedtheresultsthattheliningpressuredropslargely
withtime, shownasFigure16. However, accordingto
theresearch by Hashimoto et al. thelining pressure
Table7. Geotechnical aspectsonliningpressureandpres-
sureactingonTBM.
Liningpressure Liningpressureduringconstruction
for design andat longterm, magnitudeand
distributionof pressure, effectsof
backfill grouting(groutingpressure
andmaterials), soil typesandground
condition
Longitudinal Backfill grouting, liveloadanddead
deformationand loaddistribution, subsoil reaction, et al
bendingof tube
PressureonTBM Backfill groutingpressure, jackforce,
slurrypressure, greasepressureat tail
seal, earthpressureat aface, driving
control of TBM, shapeandrigidityof
shield, subsoil reaction, tround
deformationbypressureandload
fromTBM
Figure 13. p
v
,p
v0
vs. 2C,p
v0
in clayey ground (by
Hashimoto et al.).
dropisverysmall.Themaindifferencesareconsidered
tobethegrout material andinjectionmethods.
Talmon et al. studied the longitudinal tube bend-
ing due to grout pressure. They carried out beam
actioncalculationusinginputparametersof thebend-
ingmoment byTBM jacks, transverseforcebyTBM,
vertical grout pressure gradient behind TBM, load-
ingdiagramandunsupportedlengthof tunnel lining
inTBM, et al. Thecalculatedresult fitstheobserved
onestosomeextent, shownasFigure17.
Bezuijen & Bakker described the interaction
betweentheslurry fromthefaceandthegrout from
thetail, seeFigure18. Thepressuredistributionalong
thelongitudinal directionof TBM iscalculatedtheo-
retically based on thepressureloss (LP) dueto the
flow.ThecalculatedresultisshowninFigure19.They
104
Figure14. p
v
,p
v0
vs.SPT-Ninsandyground(byHashimoto
et al.).
Figure 15. Meausred and calculated grout pressures (by
Talmo & Bezuijen).
foundthatLP dependsontheshear stressof thegrout
along theTBM (
r
), gap betweenTBM and ground,
andlengthincrement alongtheTBM.
Song & Zhou did a research work on the earth
pressure distribution of excavation chamber in EPB
tunneling. Accordingtotheir work, thetotal support-
ingpressurecanbecomposedbytwoparts: (1) Earth
supportingpressureP
E
inworkingchamber; (2) Cut-
ter head plane supporting pressure P
P
. The authors
proposedanestimationmethodof earthpressureratio
based on the empirical relation among cutter head
torque, trapezoidal shape of pressure distribution,
Figure16. Dropsof liningpressurewithtime(byTalmo&
Bezuijen).
Figure 17. Measured and calculated bending moments
compared (Groene Hart Tunnel, the Netherlands) (by
Talmon et al.).
opiningratioof cutter face. They foundthat EPSR in
clayislarger thanthat incobbleandsand.
5 CONDITIONINGADDITIVESFOR EPB
Required properties of conditioned soil for EPB are
summarized in Table 8. Until now, there are many
types of conditioningadditives havebeenutilizedin
practice, including slurry, foam, polymer, water (for
clayeyground), cellulose, sodiumalginate, et al.
TherearemanydataintheworldespeciallyinJ apan
for this subject, but thesedatahas not been summa-
rized in general. In Session 2, there are two papers
by Hajialiue-Bonal et al. concerningconditionaddi-
tivesfor EPB. Thetwopapersdescribedthefollowing
resultsfromlaboratorytest for foamandconditioned
sandybyfoam:
1 Polymer typefoamshowsagoodstability;
2 Withsomecombination, foam/sandmixtureshave
highcompressibility;
105
Figure18. Possibleflowdirections andsketchedpressure
distributionsalongtheTBM.
Figure19. Pressuresandjoint widthalongaTBM.
3 The soil conditioning by foamcause decrease of
shear strength(c, +);
4 WhenFoamExpansionRatiois10-FER-18,the
changeof strengtharenegligible.
6 CONCLUSIONS
Thepapersinthissessionprovideabundantcasestud-
iestoadvanceour knowledgeof tunnel constructions,
Table8. Requiredpropertiesof conditionedsoil for EPB.
Highflowability Lowshear strength, reductionof
cuttingtorqueandwear of cutter bits,
stableandprecisemonitoringof earth
pressureinworkingchamber
Impermeability Preventionof water inflowandpiping
High Reducingearthpressurefluctuation
compressibility duringexcavation
Uniformity Preventionof water inflowandpiping,
precisemonitoringof earthpressure
and many new technologies developed in recent
yearswereintroduced. Thisisparticularly important,
because through case study, we can understand the
advantage,disadvantage,applicablefield,andfeasibil-
ityof thenewconstructionmethodsandtechnologies.
Somepapersintroducednewmonitoringandmea-
suringtechnologies. This is also animportant aspect
for tunnel construction. On-time and accurate mon-
itoring and measuring can maketunnel construction
workquicker, safer, andmoreeconomical. Of course,
abundance of measuring data help us analyze and
understand the mechanism and the essence of the
interactionbetweentunnel structureandground.
Predictionandtheoretical analysis was concerned
in some papers. In general, prediction and analysis
results were compared with observed results to ver-
ify their validity. But weshould pay attention to the
limitationandapplicablefieldof thesemethods.
REFERENCES
Bakker K.J. & BezuijenA. 10 years of bored tunnelling in
the Netherlands: Part I geotechnical issures.
Bakker K.J. & BezuijenA. 10 years of bored tunnelling in
the Netherlands: Part II structureal issures.
BezuijenA. & Bakker K.J. The influence of flow around a
TBM machine.
Bezuijen A. & van Tol A.F. Mechanisms that determine
betweenfractureandcompactiongroutinginsand.
ChongH., Li T. &YanJ. The double-o-tube shield tunnel in
Shanghai soil.
DengZ.G. &ZhangQ.H. Research of non-motor vehicle -rail
transit-tube interchanging transport system pattern.
FillibeckJ. &Vogt. N. Shotcrete excavations for the Munich
subway Comparison of different methods of face support
in settlement sensitive areas.
Gafar K., Soga K., Bezuijen A., Sanders M.P.M. &
vanTol A.F. Fracturing of sand in compensation grouting.
Guatteri G., Koshima A., Lopes R., Ravaglia A. &
Pieroni M.R. Historical cases and use of horizontal
jet grouting solutions with 360 distribution and frontal
septum to consolidate very weak and saturated soils.
Hajialilue-Bonab M., Ahmadi-adli M., Sabetamal H. &
Katebi H. The effects of sample dimension and grada-
tion on shear strength parameters of conditioned soils in
EPBM.
106
Hajialilue-Bonab M., Sabetamal H., Katebi H. & Ahmadi-
adli. M. Experimental study on compressibility behavior
of foamed sandy soil.
Hashimoto T., Ye G.L., Nagaya J., Konda T. & Ma X.F.
Study on earth pressure acting upon shield tunnel lining
in clayey and sandy ground based on field monitoring.
Hu X.D. & Pi A. Frozen soil properties for cross passage
construction in Shanghai Yangtze River Tunnel.
Kuzma K. The influence of engineering-geological condi-
tions on the construction of radioactive waste dump.
Li Z.X., HanX. &WangK.S. Critical ventilation velocity in
large cross-section road tunnel fire.
Sfriso A. O. Metro tunnels in Buenos Aires: Design and
onstructionprocedures19982007.
SongT.T. & ZhouS.H. Study on the Earth Pressure Distri-
bution of Excavation Chamber in EPB tunneling.
TalmonA.M. & BezuijenA. Backfill grouting research at
Groene Hart Tunnel.
TalmonA.M., BezuijenA. &HoefslootF.J. Longitudinal tube
bending due to grout pressures.
107
Physical andnumerical modelling
P.L.R. Pang
Geotechnical Engineering Office, Civil Engineering and Development Department,
Government of the Hong Kong Special Administrative Region, HKSAR
ABSTRACT: This General Report has been prepared based on areviewof twenty papers submitted to the
session on Physical and numerical modelling related to geotechnical aspects of underground construction.
The papers cover a wide range of model feature types in different materials. The problems studied include
ground/tunnel facestability, ground/tunnel deformationandearthpressures, ground/tunnel-structureinteraction,
seismicbehaviour, andvehiclefiresinaroadtunnel. Thisreport highlightsanddiscussestheapproachesused
inmodellingandpresentsthekey findings. Someremarksaregivenat theendontheobjectivesof modelling
andtheworkof TC28.
1 INTRODUCTION
20papershavebeensubmittedtothissession(Table1).
Threeof thepapersarejointcontributionsfromauthors
of twocountries.
The papers cover a wide range of model feature
types (Table 2). These include tunnels in clay, sand,
aluminumrods (modellingagranular mass), layered
soils, aswell astunnelsinsoftorweakrock.Thereisa
paper onmodellingof deepexcavationswithstepped-
twinretainingwalls, andapaper onvehiclefiresina
roadtunnel.
2 PHYSICAL MODELLING
Eleven papers present results of physical modelling
(Table3). Theseincludesix models at 1g, aphotoe-
lasticmodel andfour centrifugemodels.
Table1. Geographicdistributionof thepapers.
Country Papers
China 5
China/France 1
Denmark 1
France 1
Italy 1
Italy/UK 1
J apan 2
J apan/UK 1
Korea 4
TheNetherlands 1
UK 2
20
For thethreemodels that usealuminumrods, the
testswerecarriedoutat1g. Numerical modellingwas
also carried out to compare with the results of the
1gtests.
Thesandmodel at1gwaspreparedbycompaction
of the sand using a plate vibrator. The compaction
could have created locked-in compaction stresses
on the model braced wall and the adjacent tunnel
thus influencing themodel test results. This was not
discussedinthepaper.
Twopaperspresentresultsof modellingof rocktun-
nel problemsusing1gtests. Oneusedbaritepowder,
sandandplaster mixedwithwater, andtheother used
concretebrickstomodel thesoftrock.Therewassome
discussiononthemodellinglawsinthepapers. While
theconclusionsonthequalitativebehaviour seemrea-
sonable, and arenot unexpected, it is not sureif the
quantitativeresultsarevalidat prototypescalewhere
discontinuitiesintherockandthehigher stresslevels
couldinfluencethemagnitudeof thedeformations.
Table2. Featuretypescoveredinthepapers.
Featuretype Papers
Tunnelsin:
(a) Clay 5
(b) Sand 3
(c) Aluminumrods/crushedglass 4
(d) Layeredsoils 3
(e) Soft/weakrock 3
Deepexcavation(aluminumrods) 1
Vehiclefiresinroadtunnel 1
20
109
Table3. Papersonphysical modelling.
Techniqueandscale 2D/3D Materialsused Papers
Laboratory1gmodel 2D Aluminumrods 3
(scales: 1/10, Sand 1
1/19, 1/80) Baritepowder/ 1
sand/ plaster
Concretebricks 1
Photoelasticmodels 2D Crushedglass 1
Centrifugemodels 3D Clay 1
(75g, 100g, 160g) Sand 2
Sandoverlying 1
clay
There are four papers on centrifuge tests. One of
thepapersistostudytheeffectsof pileloadingonan
existingtunnel inanoverconsolidatedclay, twoareon
centrifugetestswheredrysandwasusedtoconstruct
themodels for studyingtheinteractionmechanisms,
and one on tunnelling in an overconsolidated clay
overlainbysandunder thewater table.
3 NUMERICAL MODELLING
18 papers present numerical modelling results
(Table4). Different numerical modelling techniques
wereused.
14 out of these 18 papers used either 2D or 3D
codes based on the finite difference method (FDM)
or the finite element method (FEM). Some of the
codes, e.g. CRISP, FLACandPLAXIS, arewell estab-
lishedcodesandthe2Dversionsarecommonlyusedin
current engineeringpractice. Intheanalyses, thesoil
was modelled either as alinear elastic or an elastic-
perfectly plastic material withtheMohr-Coulombor
Drucker Prager failurecriterion. Whereacomparison
was made, the elastoplastic model performed better
thantheelasticmodel.
In one paper, a slope stability analysis program
SLOPE/W based on the limit equilibrium method
was used to compute the factors of safety of a clay
slope. Theresults werecomparedwiththeresults of
FLACandPLAXISwhichusedthestrengthreduction
method. However, noinformationisgivenonthethe-
oretical method used (afew options areavailablein
SLOPE/WsuchasJ anbu, BishopandMorgenstern&
Price) and the choice of slip surfaces, which could
affect thecomputedsafety factors. Also, noinforma-
tionisgivenonwhattheslopedeformationandthesoil
shear strainwere, whenthesoil strengthisreducedfor
thefactor of safetytoapproachunity.
A visco-elastic model adopting a nonlinear rela-
tionshipbetweenthenormalizedshear modulus (and
dampingratio)andtheshearstrainamplitudewasused
for a 1D ground dynamic shear response analysis.
The code EERA was used for the analysis, the
objective of this study was to calibrate a linear
visco-elastic, effectivestressbased, constitutivemodel
for use in coupled 2D dynamic analyses using the
finiteelement programPLAXIS. Theviscous damp-
ingwasaccountedfor usingtheRayleighformulation
(Woodward& Griffiths, 1996).
Thesubloadingtij finiteelementmodel (developed
by Nakai & Hinokio (2004)) was used in two cases
to provideresults for comparing with physical mod-
ellingat 1gwhichusedaluminumrods inthemodel
tests. Thetij model takes into account theinfluence
of theintermediateprincipal stress by introducing a
modifiedstresstij. Also, thesubloadingconcept(pro-
posedbyHashiguchi (1980)) isadoptedtomodel the
influence of soil density. Five of the seven param-
eters in the tij model are the same as those in the
Cam-clay model, with onemoreparameter added to
describe the influence of soil density and confining
pressure, andanother parameter addedtocharacterize
theshapeof theyieldsurface. Laboratorybiaxial tests
were carried out to compare the stress-strain curves
obtained from the finite element program FEMtij-
2D. In the biaxial tests, shearing of the aluminum
rods, whichhadlowfrictionangles, induceddilatant
behaviour.Thematchbetweenthebiaxial testsandthe
finiteelement analysisresultsappearsreasonablebut
this is up to ashear strain level of about 12%only
(Figure1).
TheDistinct Element codeUDEC wasusedinone
casetocomparewiththeresultsof large-scalemodel
testscarriedout usingconcretebrickstomodel rock.
However, thepaper doesnot indicatehowthediscon-
tinuitiesintherock weremodelled. For theother two
papers ontunnels inrock, thenumerical simulations
werecarriedout usingfiniteelement codes adopting
an elastoplastic rock model with theDrucker-Prager
failurecriterion. It seemsthat theneedfor modelling
the discontinuities that may be present in the rock
was not considered. It is not too clear fromthetwo
papers howtherock parameters weredeterminedfor
thecontinuummodelsandthefieldprototypes.
Results obtained from closed form solutions
derived using upper bound limit analysis were pre-
sented in two of the papers, for comparison with
the results of centrifuge modelling and numerical
modellingrespectively.
The Fire Dynamics Simulator code incorporating
alargeeddy simulationmodel was usedto carry out
computational fluiddynamics modelling. Theobjec-
tive of this work was to study the heat release rates
fromvehiclefiresinaroadtunnel of 15mindiameter.
Thecomputedresultswerecomparedwithanempiri-
cal equation.Thisindicatesthattheempirical equation
requiresimprovementforthecaseof small firesinroad
tunnelswithalargecrosssection.
110
Table4. Papersonnumerical modelling.
Constitutivelaw Modelling Program Papers
Linear elastic 2DFEM PLAXIS 2
Nonlinear visco-elastic 1Dshear EERA 1
Elasto-plastic 2DFEM CRISP, Msc.MARC 2
(Mohr Coulomb) 2DFDM FLAC 1
3DFEM MIDAS-GTS 1
3DFDM FLAC3D 3
Elasto-plastic 3DFEM MARC 2
(Drucker Prager)
Elasto-plastic(Cam 2DFEM FEMij-2D 2
clay+2parameters)
Distinct element 2DDEM UDEC 1
Rigid-plastic Limit analysis Closedformsolution 2
Largeeddysimulation CFD FireDynamicsSimulator 1
Figure1. Stress-strain-dilatancyrelation.
4 PROBLEMSSTUDIED
Theproblemsstudiedasreportedinthepapersinclude:
1. ground/tunnel facestability(5papers),
2. ground/tunnel deformationandearthpressures (8
papers),
3. ground/tunnel-structureinteraction(5papers),
4. seismicbehaviour (1paper), and
5. vehiclefiresinaroadtunnel (1paper).
A brief reviewof selectedpapersisgivenbelow.
4.1 Ground/tunnel face stability
Thesubject of facestability is avery important one.
If thefacepressureappliedistoolow, therecouldbe
acollapseor excessivegroundsettlement, andif the
facepressureis too high, therecould bea blow-out
failureor excessivegroundheave.
A number of researchers have studied this prob-
lem(e.g. Anagnostou&Kovri, 1994). Thefollowing
papershaveaddedtotheknowledgebase.
Li et al investigated the failure of a large slurry
shield-driven tunnel using upper bound limit anal-
ysis and numerical modelling. The study is for the
15.43mdiameterShanghai YangtzeRiverTunnel con-
structedinsoft clay. A shallowgroundcover section,
with a ground cover to tunnel diameter (C/D) ratio
of 0.7, was selected for thestudy. Undrained condi-
tions wereassumed in themodelling. A multi-block
failuremechanismwithauniformfacepressure(sug-
gestedbySoubra,2002)wasusedforthelimitanalysis.
FLAC3Dwasusedforthenumerical modelling(which
adoptedanelastic-perfectlyplasticconstitutivemodel
withaMohr-Coulombfailurecriterion).Theresultsof
theupper boundlimit analysis andthe3D numerical
modellingshowedthat partial blow-out failureof the
upper part of thetunnel faceoccurs whentheslurry
pressureislarge, whereasglobal collapseof thewhole
tunnel faceoccurs when theslurry pressureis small
(Figure2).
Theauthors noted that thedifferencebetween the
slurry pressure and earth pressure at the crown and
invert for alargediameter slurryTBM tunnel canbe
large and this could have a significant effect on the
failuremechanismandthecritical slurrypressure.The
failuremechanismsandthecritical slurrypressuresat
the tunnel axis level obtained fromthe limit analy-
sisandthenumerical modellingagreewell witheach
other (Figure3).
Caporaletti et al reviewedthepast researchontun-
nel stability in undrained conditions (Davis, et al,
1980; Kimura& Mair, 1981; Sloan&Assadi, 1992),
in drained conditions (Atkinson & Potts, 1977) and
inlayeredground(Grant & Taylor, 2000). They con-
ductedcentrifugeteststoinvestigatethestabilityof a
circular tunnel in layered ground, with clay overlain
byamediumdensesandylayer, belowthewater table.
TheC/Dratioof thetunnel was2.38.Theclaywascon-
solidated fromaslurry, to givean overconsolidation
ratiorangingbetween1.4and2.8withdepth.All tests
111
Figure2. Twokindsof partial failuremechanisms.
werecarriedoutat160g. Theconditionof tunnel col-
lapsewas takenas volumeloss greater than20%. In
thecentrifugetests themechanismof failurefor the
layered ground involved a wide area of soil both in
sand and in clay, with pseudo-vertical settlements at
thesand-clayinterface(Figure4).
Itwasfoundthatthecontributiontostabilitydueto
frictionactingwithintheupper sandlayer represented
asignificant contribution. A significant overestimate
of the tunnel support pressures to prevent collapse
might result if the theoretical solutions obtained for
homogenousclaysareusedwiththesandlayer treated
as a surcharge. The authors proposed a new failure
mechanismwhich provided an upper bound to the
experimental data obtained (Figure 5). It would be
interesting to examine whether the proposed mech-
anismisapplicablefor thecaseof aloosesandlayer.
Date et al carried out a series of centrifuge tests
at 75g to investigate the ground deformation pat-
terns during excavation of tunnels in dry sand. The
C/Dratioof themodel tunnelswasone, andsomeof
themodels incorporatedreinforcements. Theground
deformation was found to be small even when the
facepressurewas reducedto half theinitial pressure
of 100kPa, but oncemovement started upon further
reductionof thefacepressureitincreasedsharplylead-
ingto instantaneous collapse(i.e. abrittlefailure).
Figure 3. Comparison of failure mechanisms of Case 2
(velocitycontour for FLAC
3D
analysis).
Figure 4. Mechanism of failure from centrifuge tests
(VL
=20%).
The model tests without reinforcement collapsed at
a support pressure which agrees with the centrifuge
test results of Chambon & Cort (1994). The study
found that introduction of face bolts and forepoling
yieldeddifferent tunnel collapsemechanisms, which
dependedonthedensityof thefaceboltsandforepol-
ingbolts. Surprisingly, thereinforcementscontributed
112
Figure5. Mechanismof failurefor layeredground.
Figure6. Tunnel failurepatternsonthelongitudinal section.
to only a slight reduction in the support pressure
requiredtokeepthetunnel facestable,comparedtothe
case without reinforcement. The face bolts installed
stiffenedthegroundaheadof thefaceandwerefound
to beableto reducethefaceextrusion. Theforepol-
ingdividedthegroundaroundthetunnel faceintotwo
zones, withtheouterzoneforminganarchcomprising
the forepoling bolts. The geometries of the collapse
mechanisms are similar to those observed by other
researchers for tunnels in sands, e.g. as reported by
Chambon& Cort(1994) andMair &Taylor (1997).
They all involve a narrow chimney, propagating
almost vertically fromthe tunnel up to the ground
surface(Figure6).
FLAC3Danalyseswerealsocarriedout.TheMohr-
Coulomb soil model with strain softening/hardening
wasfoundtogiveabetter matchtothecentrifugedata
thantheMohr-Coulombmodel without strainsoften-
ing/hardening.Thedeformationpatternobtainedfrom
theanalysisforamodel reinforcedwithfaceboltswas
similar tothatof thecentrifugetestbutthemagnitude
was smaller. Theauthors recommendedto study fur-
ther theeffectof meshshapeandtheeffectof changes
Figure7. Measuredexcessporepressureinfrontof aslurry
shieldandapproximation.
insoil-bolt interactionproperties uponexcavationin
thenumerical analyses.
Theinformationonfailuremechanisms presented
intheabovepapersisinterestinganduseful. Thereis
recent improvement inunderstandingof theground-
tunnellinginteractionprocessesassociatedwithinflu-
ence of grouting pressures, removal of the filter
cake and the pore pressures generated during the
advanceof aslurryTBM(Figure7).Thiswasachieved
throughfieldmeasurementsobtainedduringconstruc-
tion (Bezuijen & Talmon, 2008). Further data and
study in this area will no doubt augment theresults
of existinglaboratoryandanalytical modelling, which
havenotaccountedfor suchprocesses. Further under-
standing of theprocesses could help to evaluatethe
needtorefinethecalculationmodelsanddesignmeth-
odsforestimationof facepressuresrequiredtoprevent
collapseandblow-out.
4.2 Ground/tunnel deformation and earth
pressures
A number of papers inthis sessionpresent results of
modellingtostudy thegrounddeformationandearth
pressuresaroundatunnel.
Shahinet al developedanewcircular tunnel appa-
ratus and conducted 1g model tests to examine the
groundmovementsinducedbytunnellingandtheearth
pressures around the tunnels. Aluminumrods were
usedtomodel agranular soil mass.Thesurfacesettle-
ment was measured using a laser typedisplacement
transducer with an accuracy of 0.01mm, and pho-
tographs were taken during the experiments which
were later used as input for the assessment of the
groundmovementsusingtheParticleImageVelocime-
try technique (White et al, 2003). To compare with
the model test results, numerical simulations were
carried out using 2D finite element analyses under
plane strain and drained conditions. The computer
program FEMtij-2D was used. The initial stresses
113
Figure8. Distributionof shear strain: tunnel invertisfixed.
appliedcorrespondto theself-weight condition. Two
C/Dratios, viz. 1and2, wereexamined.Theeffectsof
full faceexcavation(withthecentreof theexcavation
keptfixed) andtopdriftexcavation(withtheinvertof
thetunnel kept fixed) werealso studied. Thesurface
settlement andearthpressuresaroundthetunnel were
foundto besignificantly influencedby thedisplace-
mentatthetunnel crownfor thesameoverburdenand
samevolumeloss.Thevolumelosswaslesssignificant
comparedtothecrowndrift inthecaseof theshallow
tunnel.Thefull faceexcavationcaseproducedawider
shear deformationregionthanthat for thecaseof top
driftexcavation(Figure8). Theuseof anelastoplastic
soil model producedbetter matchwiththemodel test
surfacesettlement profilethananelastic soil model.
Thedistributionof earthpressures aroundthetunnel
dependedontheexcavationpattern. Theauthorsindi-
catedthatthenumerical simulationsweregenerallyin
goodagreement withthemodel test results. However,
it is no clear whether thetij finiteelement model is
capable of describing the behaviour of tunnels con-
structedinreal soilsespecially insoilswhichexhibit
contractilebehaviour.
Lianget al studiedtheeffectsof soil stratification
on tunnelling-induced ground movements. 3D anal-
yses were carried out using the computer program
FLAC3D.Thebehaviourof the2.47mdiameter Thun-
der Bay sewer tunnel inCanada, constructedusinga
TBM withsegmental concretelining, insoft to firm
clayswithsiltandsandseams, wassimulated.TheC/D
ratioof thetunnel was3.8.Thesoil strataweredivided
into four sub-layers for thepurposeof theanalyses.
Figure9. Lateral displacement15mbehindthetunnel face.
The ground surface settlement, lateral displacement
profileat 15mbehindthetunnel faceandthesubsur-
facesettlement withdepthabovethetunnel axisfrom
theanalyses werecomparedwiththeanalysis results
obtained by Lee & Rowe (1991) using the FEM3D
program(also based on an elastoplastic soil model).
Theywerealsocomparedwiththefielddatareported
byBelshaw&Palmer(1978).Additional comparisons
werecarriedout withtheanalytical solutiongivenby
Loganathan& Poulos (1998). Thestudy showedthat
theelastoplastic soil model couldsimulatethedefor-
mationprofilesbetter thanthosebasedontheelastic
model.Theresultsof theelastoplasticsoil model indi-
cated that soil stratification had little effects on the
groundsurfacesettlementbutsignificantlyinfluenced
the lateral displacement and subsurface settlement
profiles (Figure9). This was different fromtheelas-
tic soil model whichpredictedthat soil stratification
had significant effects in all cases. This is an inter-
esting casehistory of benchmarking a3D computer
programusing data froma past project, illustrating
the value of documenting good data and making it
availablefor research.
Song et al studied the time-dependent behaviour
of soft ground tunnels constructed using steel rein-
forcementsgroutedintothegroundaheadof atunnel
114
Figure10. Time-dependent characteristics of elastic wave
velocitiesof asand-cement mixture.
(atechniquewhichtheauthorscalledthereinforced
protectiveumbrellamethod). Laboratorydirectshear
testsandPandSwavevelocitytests(usingpiezoelec-
tric bender elements) were carried out to determine
thestrengthandstiffnessof thesand-cement mixture
at different curingtimes. Thetest resultsshowedthat
thesand-cement mixturegainedsignificant increases
in stiffness after about 6 hours whereas the appar-
ent cohesion increased to about 2MPa after 7 days
(Figure10).
3D finiteelement analyses werecarriedout using
acomputer programMIDAS-GTS(2005) tosimulate
thebehaviour of such atunnel. Thetunnel is 18.8m
wideand10.4mhigh, at 15mbelowground. It was
constructedinweatheredrock, using12mlongsteel
pipesasreinforcement. Thewater tablewasatground
surface.Theanalysesincorporatedthetime-dependent
material properties of thesand-cement mixture. The
excavation rate was taken as 0.75m per day. The
studyconcludedthat useof the23daysstrengthand
stiffness parameters was adequatefor predicting the
time-dependent deformation behaviour, for practical
designpurposes, providedthatthereissufficientover-
lapbetweenthereinforcements. Nocomparisonwith
anyfieldperformancemonitoringresultswashowever
presented.
Leeet al studiedthebehavior of a2-archrocktun-
nel usingalarge-scaletestmachine(6mwide6.5m
high). Themodel tests(1/19scale) wereconductedat
1g.Therockwasmodelledusingconcretebricks.The
tests showed that the ground displacements induced
by tunnelling were mainly within a zone of 0.25D
fromthetunnel, whereD is thetunnel width. Hori-
zontal displacements of morethan 40%and vertical
displacements more than 20% of the total displace-
mentsoccurredduringexcavationof thepilot tunnel.
Theauthors suggestedthat thestability of the2-arch
tunnel couldbedominatedbythestabilityof thepilot
tunnel excavationandthat therockbolt lengthshould
be longer than 0.25D. Displacements obtained from
UDEC analyses werepresented. Whiletheseshowed
the same pattern, details of the analyses were not
given. Based on thelimited measurements obtained,
theauthorssuggestedthat therock loadactingonthe
centrepillar of the2-archtunnel may betakento be
0.15Wfor preliminarydesign, whereWisthecentre-
to-centredistancebetweentheleft andright tunnels,
whentheRMR of therock massismorethan60. No
numerical analyses were carried out. More research
wasrecommendedtoconfirmtheproposedempirical
relationship. It wouldbeuseful toexaminetheinflu-
enceof rockdiscontinuitiesandtheeffectof rockblock
sizerelativetothetunnel diameter.
4.3 Ground/tunnel-structure interaction
Broere&Dijkstrainvestigatedtheinfluenceof tunnel
volumelossonpilesusingthephotoelastictechnique.
2Dplanestrainmodel testswereconductedtoexamine
the tunnel-pile interaction. Crushed glass (a photo-
elasticmaterial)wasusedtomodel thesoil.Theeffects
of volume loss were simulated by making the tun-
nel diameter contract vertically. Fromthetests, it was
found that significant stress changes occurred close
tothepiletips. Thetests withavolumeloss of 0.6%
showed a clear influence of the volume loss on the
stresses near thepiletips up to onetunnel diameter
away. Thestudysuggestedthat theinfluencezonefor
displacementpileswithbothendbearingandskinfric-
tion, might beslightlylarger thanfor boredpileswith
endbearingalone. Theauthors indicatedthat further
fieldobservations, model testingandnumerical mod-
ellingarerequiredtodeterminetheinfluencezone.
Lee & Yoo studied the ground shear strain pat-
ternsdevelopedaroundatunnel andtheexistingpiles
nearby dueto tunnel construction. Small-scalelabo-
ratory model tests at 1gwereconducted. Aluminum
rods wereused to model thesoil mass and thepiles
embedded in it. A tunnel diameter reduction system
capable of achieving a tunnel volume loss of up to
20%wasspeciallydeveloped. Thestrainedcontrolled
testscarriedout usingthissystemresultedinground
shear strainswhichwerecapturedbycloserangepho-
togrammetry. 3Dnumerical analyseswerealsocarried
outusingthefiniteelementprogramCRISP. Compar-
ison between thephysical model tests and thefinite
element analysesshowedgoodagreement intermsof
shear strain patterns. Based on the maximumshear
115
Figure11. Schematicillustrationof shear strainmodesfor
pile-soil-tunnellinginteraction.
straincontours, twodistinctshear strainpatternswere
observed, viz. withandwithouttunnel-pileinteraction
(Figure11). Theboundary betweenthesetwo modes
of behaviour dependedonthelocationof thepiletip
fromthetunnel andthemagnitudeof thetunnel vol-
ume loss. The authors suggested that this boundary
might serveasauseful guideintheplanningthetun-
nel alignment inareaswherepilesarepresent. It may
beworthwhileto comparetheresults given in these
paperswiththefindingsof J acobszetal (2004; 2005)
fromcentrifugetestsandSelemetaset al (2005) from
fieldtests.
Yao et al studied the effects of loading of bored
pilesonexistingtunnels. Centrifugemodel testswere
carried out at 100g. The model tunnel was formed
in firmto stiff clay consolidated froma slurry. The
tunnel liningdeformation, porepressures intheclay,
pileloadapplied, pilesettlementsandtunnel facepres-
suresweremonitoredwhilethepileloadingwasbeing
applied.TwoC/Dratios, viz. 2and3, werestudied.The
tests examined thebehaviour after pileconstruction.
Theinfluenceof pileexcavationwasnotconsidered.In
theteststhepilebasewassetattwodifferentpositions:
tunnel crownandinvert level. Therateof loadingwas
designedtocreateundrainedconditions. Preliminary
analysisof theresultsindicatedthatthepilesettlement
hadalinear relationshipwithincreaseinappliedload
whentheloadexceedshalf thedesignedultimateload.
Thetunnel centrealwaysmoveddownwardsandaway
fromthe pile. Increasing the pile-tunnel clear spac-
ingreducedthedeformationof thetunnel lining. The
longpilehadmoreeffectonthetunnel liningthanthe
shortpileregardlessof theC/Dratio.Thetunnel crown
wasalwayssubjecttosignificantmovementduetopile
loading.
Marshall & Mair investigated the soil-structure
interaction mechanisms resulting from tunnel con-
struction beneath buried pipelines using centrifuge
modelling. The study aimed to validate visually the
interaction mechanisms that account for pipeline
behaviour. Particle Image Velocimetry was used to
measure displacements for characterising the soil-
structure interaction. The model tests were carried
out at 75g, using sand prepared to a relative den-
sityof 90%. TheC/Dratioof thetunnel was2.4. The
studyshowedthatestimationof thetunnel volumeloss
(defined as change in tunnel volume divided by the
original total tunnel volume) usingsoil displacement
data was not simple for sands. This was due to the
uncertainty ontheextent of thedilationandcontrac-
tilebehaviour of thesandaroundthetunnel. Thesoil
volumeloss(definedasthevolumecalculatedbyinte-
gratingthesoil settlement profileanddividingbythe
original total tunnel volume) wasnot alwaysthesame
asthetunnel volumeloss.Themagnitudeof theformer
calculatedat thegroundsurfacecanbegreater or less
thanthelatter. Thecentrifugepipelinetest illustrated
that agapformedbelowthepipelineat atunnel vol-
umeloss of between1and2%. Thegapgrewas the
tunnel volumeloss increased. Thebendingmoments
inducedinthepipeincreasedfromtheonsetof tunnel
volumeloss but didnot appear to besensitiveto the
growthof thegapheight (Figure12).
Lee&Kimstudiedthebehaviour of abracedexca-
vation in sand adjacent to atunnel using large-scale
(1/10 scale) model tests at 1g. Thebraced wall was
subjected to preloading to limit the wall deflections
during the ground excavation. The tunnel was at a
distanceof half thetunnel diameter fromthebraced
wall. Thesand was prepared to arelativedensity of
56%.2Dnumerical analyseswerecarriedoutusingthe
finiteelement programPLAXIS. It is not clear what
constitutivemodel was used for thesand. Thestudy
found that if the wall deflections were significantly
reduced by preloading, the stability of the adjacent
tunnel wouldgreatlyincrease. Themaximumbending
momentandshear forceinthetunnel liningdecreased
duetothepreloading. Thegroundsurfacesettlement
also decreased as a result of preloading. The wall
deflection profiles fromthemodel tests agreed well
withthenumerical analysisresults. Itisnotedthatthe
116
Figure12. Derivationof bendingmomentsfromdeformed
shapeof pipeline(fromPIV data).
sandwascompactedtoconstructthemodels.However,
itisnotclear whatinitial soil stresseswereusedinthe
numerical analyses. Also, noinformationis givenon
whether thewall installationandexcavationsequence
intheanalysesmatchedthoseinthemodel tests.
5 CONCLUDINGREMARKS
In the papers submitted to this session, the mod-
ellingobjectivesaregenerallynotexplicitlystatedbut
they probably include one or more of the following
objectives:
Toobserve/understandcollapsemechanisms
To observe/understand deformation patterns and
interactivebehaviour
Toassess/verifytheusefulnessoraccuracyof theo-
retical solutions,softwareorempirical rulesagainst
laboratory(1gor Ng) model test data
Same, but against fieldmeasurement data
Tobenchmarktheoretical solutionsor software
Topredict fieldperformance
Other than to gain knowledge and to understand
the problem, an important goal of the modelling
researchshouldbetoprovideuseful andreliabletools
or to enhance the existing tools for prediction of
fieldperformance, for useinengineeringpractice. In
this regard, some of the papers have contributed to
thisgoal.
TC28hasrecently set uptwoworkinggroups, one
on databases on underground works and another on
preparingguidelines for comparingfieldor physical
modelling with numerical simulations. Thefirst ini-
tiativewill beuseful formodellersinthatgoodquality
datawill bearchivedsystematicallyfor easyreference
andretrieval, physical modellers couldusetheinfor-
mationtoplantheirresearchandchecktheirmodel test
resultsagainstothersworkforbenchmarkingpurpose,
andnumerical modellers couldusethedatato check
thereliability andlimitations of theexistingtheoret-
ical closed formsolutions and numerical codes. The
secondinitiativewill beuseful for thosewhoarecar-
ryingoutmodellingtounderstandreal behaviour or to
validatenumerical codes.
Oneof thedatabasesonundergroundworkscould
includefailures observedinmodel tests andfailures
inactual projects. Datainthelatter categoryaremore
difficult toobtainunlessthereisaforensicinvestiga-
tionandtheinformationissubsequentlymadepublicly
available. Information on failures would be invalu-
abletoprovidelessonslearnt, forcalibrationof design
methods andfor providinginsights for risk manage-
ment. If thisistoproceed, thentheremaybemerit to
collect dataonthesizeof thefailureinfluencezones
for different groundandgeometrical conditions, and
alsothetimefor anycavitycreatedatdepthtomigrate
tothegroundsurface. Suchinformationispotentially
useful forriskmanagement,inparticular,forpreparing
monitoring plans, planning of risk mitigation mea-
sures, andpreparationof emergencypreparednessand
contingencyplans.
Another database will be on monitoring results.
For such databases, the monitoring data should be
accompanied by the necessary data on ground and
groundwater conditions, the way the soil and rock
parameterswasmeasuredandinterpreted, themethod
of wall installation, informationongroundtreatment
and the sequence of construction. Such data would
allow numerical modellers to check the capabilities
and limitations of the existing computer programs.
Fromapractisingengineerspoint of view, it isoften
not practical to use overly sophisticated software
requiringmultipleparameterstocharacterizethesoils
for design. Thisisbecauseof thecost, timeanddiffi-
cultiesinobtaininghighquality groundinvestigation
data, theuncertainties associatedwithmodellingthe
ground and the hydrogeological conditions (includ-
ing the boundary conditions), the effort needed to
model therangeof designsituationsandtocarry out
sensitivityanalyses, theneedforhavingrelativelysim-
pletools for undertaking design reviews in atimely
manner duringconstruction, thedifficulties inincor-
porating effectively the wall installation and ground
treatment effectsintheanalyses, andthelackof com-
petent personnel intheuseof sophisticatedcodesand
checking of the computed results fromsuch codes.
117
Thereisalackof systematiccomparisonontheresults
obtainedfromsophisticatedsoftwarewiththosefrom
lesssophisticatedones. Theavailabilityof goodqual-
itymonitoringdataandbenchmarkingof theexisting
numerical codes using good quality monitoring data
couldhelptoaddressessomeof theseissues.
TC2onphysical modellingingeotechnicshassim-
ilar initiatives ondatabases (seehttp://www.tc2.civil.
uwa.edu.au). Cross committee communication will
createsynergy.
ACKNOWLEDGEMENTS
Theauthor isgrateful toPaul Wufor hisassistancein
preparingthisGeneral Report.
LIST OF PAPERSREVIEWED
Boldini, D. & Amorosi, A. Tunnel behaviour under seis-
mic loads: analysis by means of uncoupledandcoupled
approaches.
Broere, W. &Dijkstra, J. Investigatingtheinfluenceof tunnel
volumelossonpilesusingphotoelastictechniques.
Caporaletti, P., Burghignoli, A., Scarpelli, G. &Taylor, R.N.
Assessment of tunnel stabilityinlayeredground.
Date, K., Mair, R.J. & Soga, K. Reinforcing effects of
forepolingandfaceboltsintunneling.
Du, J.H. & Huang, H.W. Mechanical behavior of closely
spacedtunnelslaboratorymodel testsandFEManalyses.
Idris, J.,Verdel,T. &Alhieb, M. Stabilityanalysisof masonry
of anoldtunnel bynumerical modellingandexperimental
design.
Iwata, N. Shahin, H.M., Zhang, F., Nakai, T., Niinomi, M. &
Geraldni, Y.D.S. Excavationwithstepped-twinretaining
wall: model testsandnumerical simulations.
Kasper, T. & J ackson, P.G. Stabilityof anunderwater trench
inmarineclayunder oceanwaveimpact.
Lee, S.D., J eong, K.H., Yang, J.W. & Choi, J.H. A study on
behavior of 2-archtunnel byalargemodel experiment.
Lee,S.D.&Kim,I.Behaviorof tunnel duetoadjacentground
excavationunder theinfluenceof pre-loadingonbraced
wall.
Lee, Y.J. &Yoo, C.S. Twodistinctiveshear strainmodesfor
pile-soil-tunnellinginteractioninagranular mass.
Li, Y., Zhang, Z.X., Emeriault, F. & Kastner, R. Stability
analysisof largeslurryshield-driventunnel insoft clay.
Liang, F.Y., Yao, G.S. & Li, J.P. Effectsof soil stratification
onthetunneling-inducedgroundmovements.
Marshall, A.M. & Mair, R.J. Centrifugemodelingtoinvesti-
gatesoil-structureinteractionmechanismsresultingfrom
tunnel constructionbeneathburiedpipelines.
Shahin, H.M., Nakai, T., Zhang, F., Kikumoto, M.,
Tabata, Y. & Nakahara, E. Ground movement and earth
pressureduetocirculartunneling: model testsandnumer-
ical simulations.
Song, K.I., Kim, J. & Cho, G.C. Analysis of pre-reinforced
zone in tunnel considering the time-dependent perfor-
mance.
Wang, K.S., Han, X. &Li, Z.X. Vaulttemperatureof vehicle
firesinlargecross-sectionroadtunnel.
Wang, X.M., Huang, H.W. & Xie, X.Y. Effects of different
benchlengthonthedeformationof surroundingrock by
FEM.
Yao, J., Taylor, R.N. & McNamara, A. Theeffectsof loaded
boredpilesonexistingtunnels.
You, G.M. 3DFEManalysisongrounddisplacementinduced
bycurvedpipe-jackingconstruction.
REFERENCES
Anagnostou, G. & Kovri, K. 1994. The face stability of
slurry-shielddriventunnels.Tunnelling and Underground
Space Technology. 9(2), 165174.
Atkinson, J.H. & Potts, D.M. 1977. Stability of a shallow
circular tunnel incohesionlesssoil. Gotechnique, 27(2),
203215.
Belshaw, D.J. &Palmer, J.H.L. 1978. Resultsof aprogramof
instrumentationinvolvingaprecast segmentedconcrete-
linedtunnel inclay. Canadian Geotechnical Journal, 15,
573583.
Bezuijen,A. &Talmon,A.M. 2008. ProcessesaroundaTBM.
KeynoteLecture. Pre-printVolumeof theProceedingsof
the6thInternational SymposiumonGeotechnical Aspects
of UndergroundConstructioninSoft Ground, Shanghai.
Chambon, P. & Cort, J.F. 1994. Shallow tunnels in cohe-
sionlesssoil: Stabilityof tunnel face. Journal of Geotech-
nical and Geoenvironmental Engineering, ASCE, 120(7),
11481165.
Davis, E.H., Gunn, M.J., Mair, R.J. & Seneviratne, N.
1980. Thestability of shallow tunnels and underground
openings in cohesive material. Gotechnique, 30(4),
397416.
Grant, R.J. & Taylor, R.N. 2000. Stability of tunnelsinclay
withoverlyinglayersof coarsegrainedsoil. Proceedings
of GeoEng2000. Melbourne, Australia.
Hashiguchi, K. 1980. Constitutiveequationof elastoplastic
materialswithelasto-plastictransition. Journal of Applied
Mechanics, ASME, 102(2), 266272.
J acobsz, S.W. Standing, J.R., Mair, Soga, K., Hagiwara, T. &
Sugiyama, T. 2004. Centrifuge modelling of tunnelling
near drivenpiles. Soils and Foundations, 44(1), 5158.
J acobsz, S.W., Bowers, K.H. and Moss, N.A. 2005. The
effectsof tunnellingonpilestructuresontheCTRL. Pre-
print Volumeof theProceedings of the5thInternational
Symposium on Geotechnical Aspects of Underground
ConstructioninSoft Ground, Amsterdam.
Kimura, T. & Mair, R.J : 1981. Centrifugal testingof model
tunnels in soft clay. Proceedings of 10th International
Conference on Soil Mechanics and Foundation Engineer-
ing. Stockholm, 1, 319322.
Lee, K.M. & Rowe, R.K. 1991. An analysis of three-
dimensional groundmovements: theThunder Baytunnel.
Canadian Geotechnical Journal, 28, 2541.
Lee, K.M., Rowe, R.K. & Lo, K.Y. 1992. Subsidenceowing
totunnelling. I. Estimatingthegapparameter. Canadian
Geotechnical Journal, 29, 929940.
Loganathan, N. & Poulos, H.G. 1998. Analytical prediction
for tunnelling-inducedgroundmovementsinclays. Jour-
nal of Geotechnical and Geoenvironmental Engineeing,
ASCE, 124(9), 846856.
Mair, R.J. & Taylor, R.N. 1997. Themelecture: Bored tun-
nelling in the urban environment. Proceedings of 19th
118
International Conference on Soil Mechanics and Foun-
dation Engineering, Hamburg, 23532384.
MIDAS-GTS. 2005. Geotechnical &Tunnel Analysis System,
MIDASInformationTechnologyCo., Ltd.
Nakai, T. & Hinokio, M. 2004. A simpleelastoplasticmodel
for normally and over consolidated soils with unified
material parameters. Soils and Foundations, 44(2), 5370.
Selematas, D., Standing, J.R., & Mair, R.J. 2005. The
responseof full-scalepilestotunnelling.Pre-printVolume
of the Proceedings of the 5th International Symposium
onGeotechnical Aspectsof UndergroundConstructionin
Soft Ground, Amsterdam.
Sloan, S.W. &Assadi,A. 1992. Stabilityof shallowtunnelsin
soft ground. Predictive Soil Mechanics, ThomasTelford,
London, 1993, 644662.
Soubra,A.H.2002.Kinematical approachtothefacestability
analysis of shallow circular tunnels. Proceedings of 8th
International Symposiumon Plasticity, BritishColumbia,
Canada, 443445.
Wang, J.N. 1993. Seismic Design of Tunnels: A State-of-
the-art Approach. Monograph 7, Parsons, Brinckerhoff,
Quade& DiuglasInc., NewYork.
White, D., Take, A. Bolton, M.D. 2003. Soil deformation
measurementusingparticleimagevelocimetry(PIV) and
photogrammetry. Gotechnique, 53(7), 619631.
Woodward, P.K. &Griffiths, D.V. 1996. Influenceof viscous
damping in thedynamic analysis of an earth damusing
simpleconstitutivemodels. Computers and Geotechnics
19(3), 245263.
119
Casehistories
A. Sfriso
Department of Estabilidad, University of Buenos Aires, Argentina
ABSTRACT: Twentypapersacceptedfor publicationunder IS-Shanghai 2008Session3onCaseHistoriesare
classifiedandsummarized. Papers deal with open pit excavations, groundimprovement, tunnels, monitoring
systemsandimpact onsurroundings, most of themrelatedtoprojectsperformedinchallengingurbanenviron-
ments. Ithasbeenfoundthatdifferentauthorsfollowdifferentapproacheswhenreportingcasehistories, mainly
withrespect to thequantitativedescriptionof groundconditions andbehavior. Whilethis canbeattributedto
different scientific andprofessional schools, it isjudgedthat ahigher degreeof consistency andcompleteness
of thebasicinformationisrequiredfor abetter usabilityof theinformeddata. Tocontributetothisgoal, ashort
guidelineonreportingcasehistoriesisproposed.
1 INTRODUCTION
IS-Shanghai 2008becameamajoropportunitytoshare
recentexperiencerelatedtoundergroundconstruction
in soft ground. In Session 3, devoted to case histo-
ries,manyprojectsreflectingadvancesingeotechnical
engineering related to challenging urban conditions
werediscussed.
Eightcountriescontributedatotal of twentypapers
to this session: eleven fromChina, one fromJ apan,
two from Korea, one from Singapore, one from
Taiwan, onefromItaly, twofromFranceandonefrom
Germany.
In the following sections, the twenty papers are
classifiedandsummarized. Thepurposeof this clas-
sificationis toguidetheinterestedreader tospecific
informationthat might beuseful for his/her research.
Papers aregrouped as follows: i) seven papers deal-
ingwithopenpit excavations; ii) four papersdealing
with NATM and drill&blast tunnels; iii) fivepapers
dealing withTBMs and shield tunnels; and iv) four
papersdealingwithmonitoringsystemsandtheeval-
uation of the impact of under-ground projects on
surroundings.
Almost all kindof difficult groundconditionsdue
to existing infrastructureand spaceconstraints were
described. For instance, papers dealing with TBM
tunnelling describe crossing beneath a shield tunnel
and arailway line, across thefoundations of ahigh-
waybridgeandaboveexistingtunnelsof anoperating
metroline. Openpit excavationprojects arenot sim-
pler, showing thechallenges that urban construction
posestogeoengineering.
Twofacts becameevident duringtherevisionpro-
cess, asfollows: i) groundconditionsaredescribedin
widelydifferent waysandwithhighlyvaryingdegree
of completeness; andii) groundandstructurebehavior
are characterized by some representative numbers
selectedwithampleliberty. Whilethereporteddatais
veryvaluable,someeffortmustbedonetofullyexploit
itsusability.
Itisremarkablethatsevenoutof twentypapersdeal
with either recent or on-going underground projects
in Shanghai, thus reflecting the impresive rate of
infrastructuredevelopmentof thecity.A largeamount
of information is provided with respect to Shanghai
soils, including laboratory and field tests and exten-
sivereportingof groundbehaviorduringconstruction.
It is desirable that this valuable information be fur-
ther analyzed by researchers to produceaconsistent
andcompleteset of material parametersfor Shanghai
soils, astherawdataprovidedbythepapersdoesnot
allowfor acompleteunderstandingof soil conditions
andsoil behavior.
In each section, the list of papers is listed in a
table and a brief description is presented for each
paper.Thedescriptionmerelystatesthetypeof project,
geology conditions whereknownandthedescription
of a few selected contributions. These contributions
canbeof anytype, fromanoverall descriptionof con-
structionprocessestoaquantitativemeasureof ground
behavior or detailed monitoring data. The writer
recommends the reading of all papers, as the valu-
ableinformationcontainedthereisonlysuperficially
grasped by the short descriptions that follow in this
report.
121
Table1. Papersonopenpit excavations.
Author Project
Hsiung& Threeexcav. 20mdeepinKaohsiung,
Chuay Taiwan
Kondaet al Elevenbracedexcavationsat Osaka, J apan
Liu, D. et al 18mdeepproppedexcavationinShanghai
Liu, G. et al 21mdeepproppedexcavationinShanghai
Liu, T. et al 40mdeepproppedexcavationinShanghai
Mei et al 6mdeepproppedexcavationinShanghai
Osborneet al J MM groundtreatment inSingapore
2 OPENPIT EXCAVATIONS
Thelist of papersdealingwithopenpit excavationsis
giveninTable1.
2.1 Hsiung and Chuay. Observed behaviour of
deep excavations in sand
The behavior of three excavations in Kaohsiung,
Taiwan, isdescribed.Theexcavationsareapprox. 20m
deep, supported by propped diaphragmwalls 1.0m
thick and36mlong, andexcavatedinmediumdense
silty sand with clay layers (N
SPT
: 530). The water
tableisreportedat 3mto6mbelowgroundsurface.
Themaximumlateral wall displacement
hmax
and
surfacesettlement
vmax
arereported. Valuesarenor-
malizedbytheeffectiveheightof theexcavationH
e
for
comparison among thethreeprojects. It is observed
that
hmax
/H
e
falls in the range 0.03% to 0.3% and
that
vmax
isabout onehalf of
hmax
. Theeffect of the
constructionsequenceandremedial effectstoreduce
surfacesettlementsarediscussed.
2.2 Konda et al. Measurement of ground
deformations behind braced excavations
The paper reports surface and wall deformations of
braced support systems used at eleven sites of the
OsakaSubway L8 project in J apan. Whiledetails of
the support systems are not informed, the geotech-
nical conditions are reported to vary widely among
the sites analyzed, fromgravels to soft clays. Wall
deformationandsurfacesettlementsaredescribedby
areaindices as shown in Fig. 1. It is concluded that
theground settlement areaA
s
is about 20%30%of
the wall deflection area A
for excavations approx.

21mdeep, but can be much higher if consolidation
settlementsoccur.
2.3 Liu, D. et al. Research on the effect of buried
channels to the differential settlement of
building
The paper deals with the impact of a deep excava-
tion on adyacent structures in Shanghai, China. The
Figure1. Symbol definition(Kondaet al 2008).
Table2. Descriptionof thesoils(Liu, D. et al 2008).
Shear pars
Bottom e
Name level [m] % kN/m
3
- c[kPa] [
]
Fill 2.93
Clay 0.33 34.6 18.2 0.99 21 17.5
Siltyclay 3.87 43.0 17.3 1.21 13 17.0
Siltyclay 11.87 49.1 16.8 1.39 14 11.0
Clay 14.87 38.9 17.6 1.12 16 14.0
Siltyclay 21.37 34.9 17.9 1.02 15 18.5
Sandysilt 35.87 32.2 18.0 0.94 4 29.0
Siltysand 26.3 18.8 0.77 1 32.0
excavation is 18mdeep, supported by a diaphragm
wall 0.8mthick and 26mlong with steel supports,
excavatedinShanghai soft clays. Groundconditions
aredescribedas showninTable2. Thewater tableis
assumedtobe1mbelowgroundlevel.
The maximumlateral wall displacement
hmax
is
reported to be 60mm, or 0.33% of the excavation
height. Extensive analysis of the settlement behav-
ior of an adyacent building is reported in thepaper,
withemphasisonthenon-uniformsettlementratedur-
ingtheexcavationstage. Whilethecomplexityof the
geological conditionsisassessed, nodataonthecom-
pressionandpermeability parametersof thesoft clay
layersisgivenandtheconsolidationprocessisnotdis-
cussed, despitethefact that thereportedsettlement of
thebuildingwasupto125mm.
2.4 Liu, G. et al. Performance of a deep excavation
in soft clay
Thedeformationbehavior a21mdeepexcavationin
Shanghai, China, isdescribed. Thegeologicprofileis
the quaternary soft alluvial and marine clay deposit
typical of Shanghai City. The authors brief on the
geologyandpresentFig. 2thatallowsforafirstunder-
standingof thesiteconditions. GWL isreportedtobe
1mbelowgroundsurface.
122
Figure2. Soil profileandparametersinShanghai (Liu, G.
et al 2008).
Figure3. Thetwo-drill-onegrabmethodfortheconstruc-
tionof thediaphragmwall (Lui, T. et al 2008).
The support system is a 0.8m40m propped
diaphragmwall withbasecompensationgroutingand
prestressedstruts.Themaximumlateral wall displace-
ment
hmax
is reportedto be55mm, or 0.32%of the
excavationheight. Thisresult iscomparedwithother
measured values in Shanghai and other sites having
rather similar soil conditions.
Theeffectof thestiffnessof multi-proppedsupport
systemsisanalysedandthethreedimensional behavior
of theexcavationisassessed. Itisconcludedthatacor-
ner effect existsthat reducesthelateral wall displace-
mentcorner-to-center ratio
hmax(corner)
/
hmax(center)
to
about 0.390.74.
2.5 Liu, T. et al. The construction and field
monitoring of a deep excavation in soft soils
Thepaper describes theconstruction procedureof a
verylargeanddeepexcavationperformedinShanghai
clays. Theexcavationwas263mlong, 23mwideand
38m41mdeep, supportedbya1.2mthickand65m
longmulti-proppeddiaphragmwall.
Thedeepdiaphragmwall constructionprocedureis
describedindetail, includingthesocalledtwo-drill-
onegrabconstructionmethodshowninFig. 3andthe
employmentof acounterweighttobetter cleanthelast
panels lateral surfacebeforepouringthenext panel,
asshowninFig. 4.
J et groutingwas extensively employedto improve
soil conditions in the passive zone. Reported incli-
nomenter data shows wall behavior along the
Figure4. Procedureusedtocleanthelateral surfaceof the
last panel (Lui, T. et al 2008).
construction stages. The max lateral wall deflection
was
hmax
=50mm, or0.12%of thetotal heightof the
excavation. Surface settlements are reported but not
associatedby theauthors to aconsolidationprocess.
Moreover, neitherasetof compressionparametersnor
ananalysisof compressionbehaviorof Shanghai clays
isreported.
2.6 Mei et al. Excavation entirely on subway
tunnels in the central area of the Peoples
Square
Thedesignandconstructionof ashallowexcavation
6mdeep in Shanghai, China, is described. Thepar-
ticular challengeof this project was that thebottom
of theexcavationwas placed3maboveexistingtun-
nels. A support systemconsisting in a soil-cement
pilewall 3.2mthickwithdirectional jet groutingwas
designed and passivetension piles wereprovided to
control uplift.
2.7 Osborne et al. The benefits of hybrid ground
treatment in significantly reducing wall
movement: a Singapore case history
The paper reports the first major use in Singapore
of ahybridgroundimprovement procedurecalledJ et
Mechanical Mixing (J MM). J MM is a large diame-
ter deep mixing method that forms acentral coreof
mixedsoil combinedwithajet-groutedouter annulus.
A schematic diagramof thedrillingtool is shownin
Fig. 5.
Thesystemwas employed in theNicoll Highway
Station Project. Theexcavation was 27mdeep, sup-
portedbya1.5mthickand51mlongdiaphragmwall.
TheJ MM wasusedtomakeabaseplugof improved
soil 7mthickbelowtheexcavation.Groundconditions
includefill,fluvial sand,fluvial clayandnormallycon-
solidatedmarineclay. Theauthorsconcludedthat the
groundimprovementtechniqueemployedreducedthe
lateral wall displacements
hmax
byafactor of three.
123
Figure5. Schematic diagramof thedrilling tool showing
themixingarmof theJ MM machine(Osborneet al 2008).
Table3. Comparisonof wall deformationdata.
H
e

hmax
/H
e
Author Soil m %
Hsiung& Chuay Sand 20 0.030.30
Kondaet al Varies 21 0.100.24
Liu, D. et al Clay 18 0.33
Liu, G. et al Clay 20 0.32
Liu, T. et al Clay 40 0.12
Osborneet al Clay 27 0.09
2.8 Comparison between wall deformation data
Table3lists thelateral wall displacement
hmax
as a
fractionof theexcavationheight H
e
for thedifferent
projects and construction procedures described. No
correlation can be observed between H
e
and
hmax
,
confirming the well-known fact that wall deflection
heavily depends on the particular construction pro-
cedure, to the extent that it might be concluded
that diaphragmwallsandconstructionproceduresare
designed to accomplish lateral wall deflections that
balance the performance requirements of engineers
andclients.
3 NATM/DRILL&BLAST TUNNELS
Thelist of papersdealingwithNATM anddrill&blast
tunnelsisgiveninTable4.
3.1 Eclaircy-Caudron. Displacements and stresses
induced by a et al tunnel excavation: case of
Bois de Peu (France)
Thepaper describes theground responseduring the
constructionof thetwotwindrill&blasttunnelsinBois
dePeu, France.Thetunnelshaveacrosssectionareaof
130m
2
, alengthof 520m, andwereexcavatedthrough
Table4. Papersondrill&blast andNATM tunnels.
Author Description
Eclaircy- Drill&blast tunnel inBoisdePeu, Fr.
Caudronet al
Guilouxet al Drill&blast tunnel inMorocco
Quicket al NATM tunnel inMainz, Germany
Yooet al Subsidenceduetowater drawdown
claysandsoft rocksunder 8mto140mof overburen.
Thesupportsystemwasformedbyshotcrete, steel ribs
and radial bolts. Unfavourable ground conditions in
theclaysoilsdemandedtheuseof astructural invert,
forepolingandfacebolting.
An interactive design and construction procedure
wasemployed, wheremonitoringdatawasusedinan
adaptive design process. The paper reports the con-
structionsequence, theuseof monitoringinformation
toadjust design, andextensivedataonfacedisplace-
ments measured at four instrumented sections. It is
shownthat extrusionextendedonediameter aheadof
thetunnel faceandthat highfaceextrusionprovedto
beagoodindicator of poor groundperformanceand
riskof facefailure.
3.2 Guiloux et al. Case history on a railway tunnel
in soft rock (Morocco)
The construction process of the Ras RMel tunnel
in Morocco is described. The tunnel is 2.6kmlong
and has a cross section of 60m
2
. It was excavated
throughweak flyschunder 50mto150moverburden
bydrill&blastmethod. Thesupportsystemconsistsin
23cmof shotcreteandsteel ribs. A particular feature
of theconstructionprocedureistheuseof aformwork
toreduceshotcreteloss, asshowninFig. 6. Stress-to-
UCSratiosupto3.5andconvergencesupto300mm
werereported, valueshigher thanusual for drill&blast
tunnelsinrock.
3.3 Quick et al. Challenging urban tunnelling
projects in soft soil conditions
Thedesignandconstructionof theNewMainz Tun-
nel inGermanyispresented. NewMainzTunnel runs
parallel to Old Mainz Tunnel, built in 1884, with a
clearanceof 4mto 50 meters. Thetunnel is 1250m
long, with across section of 140m
2
andruns below
buildings with 10m to 23m overburden. Soils are
marly clays, silts and sand, and the support system
isacomplex combinationof bolting, umbrellas, face
boltingandreinforcedshotcrete.Groundimprovement
techniques employedat somesections to reducesur-
face settlements are described. It is reported that a
reductionof settlementsfrom11cmto1.5cm2.5cm
wasachievedbygroundimprovement.
124
Figure6. Constructionof RasRMel tunnel (Guilouxet al
2008).
Figure7. Constructionprocedurefor OldandNewMainz
Tunnels(Quicket al 2008).
It is very interesting to note the differences and
similitudes in construction procedures used in two
similar tunnels separated in time by one century, as
showninFig. 7.
3.4 Yoo et al. Characteristics of tunneling-induced
ground settlement in groundwater drawdown
environment
The paper studies the effect on surface settlements
of groundwater drawdownduetotunnel construction.
The case analyzed consists in a 70m
2
tunnel exca-
vated through weathered granite with 20mto 30m
overburden formed by fill, alluviumand weathered
rock. The support systemconsisted in pre-grouting,
pipeumbrellas, rockboltsandshotcrete.
Water drawdownproducedsurfacesettlementsthat
started approximately six diameters ahead of tunnel
faceand that stabilized six diameters behind it. The
problemwas analyzedby aparametric study usinga
2Dfiniteelement model withMohr-Coulombconsti-
tutivemodel. Itwasconcludedthatsurfacesettlements
duetotunnel constructionhavedifferentpatternswhen
ground-waterdrawdownispresent, whencomparedto
thenormal case.
It must be noted that surface settlements due to
groundwater drawdown are a well-known problem
Table5. PapersonshieldtunnelsandTBMs.
Author Description
Antiga& EPB tunnelsinMilano, Italy
Chiorboli
Gong&Zhou Tunnel beneathrailwaylineinShanghai
Wanget al Crossingbelowexistingtunnel inShanghai
Wonget al Crossingaboveexistingtunnel inHKSAR
Xuet al Crossingbridgefoundations, Shanghai
of geotechnical engineering that is accounted for by
consolidationtheoryandthatissimulatedwithconsti-
tutiveequationsthataccountforinelasticcompression.
TheMohr-Coulombconstitutivemodel reportedtobe
usedinthemodel, however, neithersimulatesinelastic
compressionnor includescompressionparameters.
4 SHIELDTUNNELSANDTBMS
The list of papers dealing with shield tunnels and
TBMsisgiveninTable5.
4.1 Antiga and Chiorboli. Tunnel face stability and
settlement control using earth pressure balance
shield in cohesionless soil
Thepaperanalyzesandcomparestwocasehistoriesof
EPBtunnelsdriveninMilano, Italy. Bothtunnelswere
excavatedthrough50mto 60mof medium-denseto
densealluvial sandsandgravels.
Theauthors provideacomprehensivesummary of
factorsaffectingsubsidenceof shieldtunnelsinsands.
They concludethat ahighadvancerateproducesless
volume loss and reduces subsidence and show that
EPB face pressure is poorly correlated to surface
settlements but depends on technological aspects of
backfillingoperations.
4.2 Gong and Zhou. Shield tunneling beneath
existing railway line in soft ground
The design and construction of the Metro Line 11
tunnel running below the Hu-Ning railway line in
Shanghai, Chinais described. Thetunnel was driven
throughShanghai soft clays below11moverburden.
Waterlevel isreportedtobe1mbelowgroundsurface.
Thetunnel hasacrosssectionof 30m
2
andissup-
ported by asegmental lining 35cmthick. Extensive
soil improvement, including jet-piles and grouting,
was performed to control surface settlements. It is
reportedthatsettlementsintheimprovedsectorswere
85%lowerthanthoseof theunimprovedsectors. Fig. 8
shows thelongitudinal irregularity of thetracks after
thetunnel wasdrivenbelowtherailway.
125
Figure 8. Longitudinal irregularity of the tracks by
tunneling-induceddeformation(GongandZhou2008).
4.3 Wang et al. Supervision and protection of
Shanghai Mass Rapid Line 4 shield tunneling
across the adjacent operating metro line
Thepaper reportsthedesign, constructionandmoni-
toringprocedureof thecrossingof MassRapidLine4
(MRL4) shieldtunnelsbelowexistingL2Metrotun-
nels inShanghai, China. MRL4tunnels haveacross
sectionof 32m
2
andweredriveninShanghai softclays
onlyonemeter beneathL2tunnelsat asmall crossing
angle. Noinformationisprovidedwithrespect tosoil
parameters and support systems of both theexisting
andnewtunnels.
Groundcontrol measurestakentoreducedisplace-
ments in the existing tunnels are described. It is
remarkable that, despite the short distance between
thenewandexistingtunnels, control measuresdidnot
include ground improvement due to lack of surface
space.
Extensivemonitoringdatawasgeneratedduringthe
operation, andsomeof it issummarizedinthepaper.
It was foundthat astrict control of shielddeviation,
careful tail groutingandaslowadvancerateaidedin
controllingL2tunnel displacements. Shieldtail grout-
ingisreportedtohaveinfluencedsettlementssometen
to fifteenmeters aboveandbehindthegroutingsec-
tion, dependingongroutpressureandgroutingstages.
Asageneral conclusion,authorsrecommendveryslow
advanceratestominimizetunnel inducedsubsidence.
4.4 Wong et al. Kowloon Southern Link TBM
crossing over MTR Tsuen Wan Line tunnels in
HKSAR
Thepaper describestheconstructionof MassTransit
Railway(MTR) Crossing. MTR Crossingisthepoint
where the new Kowloon S. Link twin tunnels cross
(2mabove) theexistingMTR tunnelsinHongKong.
KowloonS. Linktunnelshaveacrosssectionof 51m
2
andweredriventhroughdecomposedandsoundgran-
ite under 6.8moverburden by a shield-slurry TBM
withanhorizontal clearanceof 900mm. Water table
isreportedtobe2.5mbelowgroundlevel.
No restrictions to service of MTR tunnels were
allowed, andthereforeaseriesof groundimprovement
Figure9. Horizontal umbrellaplacedbetweenthenewand
existingtunnels(Wonget al 2008).
and ground control measures had to be undertaken.
Thesemeasuresincludedextensivejet-groutingof the
whole area and the installation of a physical bar-
rier made by an umbrella of horizontal pipe piles
placed below the new tunnel as shown in Fig. 9.
Whileasophisticatedmonitoringsystemisreportedto
havebeeninstalled, thereisnoinformationof ground
or tunnel behavior during the construction of MTR
Crossing.
4.5 Xu et al. Application of pile underpinning
technology on shield machine crossing through
pile foundations of road bridge
Thepaper deals with thedesign problemof a39m
2
EPB tunnel hiting 14 piles of a bridge foundation.
The tunnel belongs to Metro Line 10 in Shanghai,
China. Soil conditions are described as fill, organic
soil,clayleysiltsandclays.Twounderpinningschemes
areproposedinthepaper: i) thebridgesfoundations
be reinforced before eliminating the existing piles;
and ii) the existing piles be replaced after founda-
tion reinforcement. It is unclear whether the project
iscompletedor not.
5 MONITORINGANDIMPACT TO
SURROUNDINGS
Thelistof papersdealingwithmonitoringsystemsand
impact tosurroundingsisgiveninTable6.
5.1 Kim et al. Environmental problems of
groundwater around the longest expressway
tunnel in Korea
A 3D hydrogeologic model that simulates theimpact
of Injetunnel on grounwater level is presented. Inje
tunnel is in fact two 14.5m wide by 11km long
drill&blast twin tunnels, claimed by the authors to
formthelongest expressway in Korea. Thegeologic
126
Table 6. Papers on monitoring systems and impact to
sourroundings.
Author Description
Kimet al Hydrogeological model for InjeTunnel, Korea
Liu&Wang Descriptionof deformationmonitoringsystems
Qiuet al Monitoringsystemappliedat Beijing
SubwayL1
Zhaoet al Math. model of settlement inducedliningstress
Figure 10. 3D model of fracture network and tunnel for
steadystatesimulationof groundwaterflow(Kimetal 2008).
profiles is composed by methamorphic rocks with
somesuperficial debris.
AMODFLOWcontinuousmodel wasimplemented
for thefar fieldsimulationof groundwater flow, while
a MAFIC discontinuous model was developed for
thenear fieldmodel includingtheeffect of grouting
on water flow, as shown in Fig. 10. The conclusion
is that grouting might reduce groundwater inflow
to the tunnel by 53% to 3.6m
3
/h/km, and that the
expecteddrawdownmightbereducedby65%toabout
0.61.1meters.
5.2 Liu and Wang. Deformation monitoring during
construction of subway tunnels in soft ground
The objectives, methods and required precision for
open field and urban tunnel ground monitoring are
discussed in this paper. Tunnel Profile Scanners are
introduced.Thesearetwodigital camerasmountedon
arigidframethatproduceastereoscopicdigital image
of thetunnel surface.
A highamount of low-precisiondisplacement data
isrecorded,withanestimatederrorof 5mm.Statistical
analysisof thisdata, however, isreportedtobeuseful
asamonitoringtool.Anautomaticdeformationdevice
using an advanced Geodetic Monitoring Software,
capableof managinghighprecisionmonitoringdata,
isalsodescribed.
5.3 Qiu et al. 3D deformation monitoring of
subway tunnel
Thepaper describes theapplication of LIDAR tech-
nology to Beijing Subway L1 tunnel. LIDAR tech-
nology allows for a rate of 3D data acquisition of
100.000 points/sec by 3D laser scanning. With this
technology, expensivereflectingprismsdonotneedto
beusedandcanbereplacedbyreflectionsheetsplaced
onthetunnel walls. NURBS(nonuniformrational B
splines) technologyisusedtointerpolatetheobtained
data, andamathematical model is developedfor the
analysisof theinformation.Anexampleisgivenwhere
adifferential settlementof 0.29mmcouldbemeasured
usingthistechnology.
5.4 Zhao et al. Effect of long-term settlement on
longitudinal mechanical performance of tunnel
in soft soil
Thepaper presentsastructural model for theinduced
longitudinal stress developed in a segmental tunnel
duetononuniformsettlementsandapplythetheoryto
acasehistory.
Asreportedbytheauthors, anunspecifiedhighway
shieldtunnel 30yr old, settled/heavedupto30mmin
thelast10years.Themathematical model wasusedto
evaluatethestructural performanceof thetunnel based
onlongitudinal curvatureradiusR. Itisconcludedthat
R-27300mmayinduceleakage; stressesinducedby
R-18800mmayfail segments; bolt yieldingshould
beexpectedfor R-15000m; andthat tensilefailure
wouldprobablyoccur for R-302m.
6 SUMMARY OF CASE HISTORIESRELATED
TOPROJ ECTSINSHANGHAI
Seven out of the twenty case histories presented at
Session 3 are related to challenging underground
projectsinShanghai,China.Thisisanuniqueopportu-
nity toadvanceintechnology andtocalibratedesign
procedures for soft soils with valuableexperimental
evidence.
However,itmustbenotedthatnocompleteandcon-
sistent description of Shanghai soils has been found
amongall papers. Thereis almost no informationon
basic index parameterslikeliquidlimit, compression
or recompression indexes and apparent OCR due to
ageing.
The shear strength parameters as reported by the
different authorsarelistedinTable7andcanbeused
as an exampleto further illustratetheobserved lack
of information. WhileparameterslistedinTable7are
127
Table7. Strengthparametersreportedfor Shanghai clays.
Depth: 1015m Depth: 2025m
c c
Author kPa

kPa

Gong&Zhou 8 24.0 45 15.0
Liu, D. et al 14 11.0 15 18.5
Liu, G. et al Fig. 2.
Liu, T. et al 7 32.0 43 15.5
Mei et al nodata
Wanget al nodata
Xuet al nodata
not and cannot be either undrained or effective
stressparameters, noneof theauthorsexplainedwhat
theseparametersactuallymean.
Shanghai clays are normally consolidated clays
withsomedegreeof ageing. Strengthof thesesoilsis
universallycharacterizedbyundrainedshear strength
s
u
, either determined by in situ testing or lab test-
ing. While it might be argued that there are many
undrainedshearstrengthsforagivenclayduetostress-
pathdependencyof shearstrength, itmustbeaccepted
thatthegeotechnical communitywouldappreciateany
reportedvalueof shear strength. Theoverall lackof a
completedescriptionof Shanghai clayshighlightsthe
valueof Fig. 2.
Drained shear strength parameters are probably
fewer inquantity, thoughalsopresumedtobewidely
available, given the impressive pace of city growth
andtheexcellent degreeof geoengineeringinvolved.
Unfortunately, noclear discussiononthecritical state
frictionangleof Shanghai soilswasfoundamongall
papers.
7 SHORT, DRAFT GUIDELINE ON
REPORTINGCASE HISTORIES
After the experience of reviewing papers submitted
to Session 3 to write this report on Case Histories,
thewriter believesthat ashort anddraft guidelineon
reportingcasehistoriesfor projectsdealingwithsoft
soils might be useful. The guideline neither intends
to becompletenor definitive, as it is only based on
theinformationsearchedbut not foundinthepapers
duringtheprocessof writingthisreport.
1. Nameof theproject andlocation.
2. State of the project by the time of submittal of
the paper: design phase, under construction, or
completed.
3. Basic information on geometry: i) for tunnels,
length,areaandapictureshowingthecrosssection
with main dimensions; ii) for excavations, type,
dimensions, structural descriptionof thesupport
systemandapictureshowingthesupport system
andsoil profile.
4. A description of geological/geotechnical ground
conditionsandwater table.
5. Acomprehensivesetof clearlydefinedsoil param-
eters. Ideally, SPT and/or CPT profiles should
beincluded. Bothtotal stress andeffectivestress
shear parameters should be indicated, either as
measured or estimated values. If other strength
parametersarealsoreported, theirmeaningshould
befully explained. For problems involving large
subsidenceor other compression-drivenphenom-
ena, compressionparametersandmaterial perme-
abilityshouldalsobeindicated.
6. A brief descriptionof theconstructionprocess.
7. Description of ground behavior and unexpected
changesingroundconditionsduringconstruction
activities.
8. Monitoringinformationwhenavailable, orastate-
ment otherwise. Someamount of basic rawdata
should beincluded to better understand and use
somederivedparameterslikeA
, seeFig. 1. Fig8
is agoodexampleof informationrelevant to the
subject beingdiscussed.
9. Forunusal equipmentsorconstructionprocedures,
somefigures/picturesthat better explaintheidea,
seeFigs. 3to7and9.
10. For nonconventional calculationsandmodels, an
illustrativepicture, seeFig. 10.
Inall cases, thesourceanddegreeof confidenceof
theprovidedinformationshouldbeassessed.
A goodexampleof reportingacasehistorycanbe
foundinapaperbyShaoandMacari (ShaoandMacari
2008), selectedbecauseitdealswithadeepexcavation
inShanghai clays. Twentythreekeyparametersiden-
tify each of the six main layers that formShanghai
soils profile, including water content and void ratio,
classificationdata, shearandcompressionparameters,
permeabilityandSPT values(ShaoandMacari 2008).
8 CONCLUSIONS
Session 3 of IS-Shanghai 2008 becamean excellent
opportunity to share experience related to under-
ground construction in soft ground in challenging
urbanconditions.
Twenty papers fromeight countries, dealing with
openpit excavations, NATM anddrill&blast tunnels,
TBMs and shield tunnels, and monitoring systems
wereclassifiedandsummarizedinthisreport. While
thereporteddataisveryvaluable, someeffortmustbe
donetofullyexploititsusabilitybecausenoconsistent
procedurewasfollowedbytheauthorstoreportground
conditions and ground behavior during construction
activities.
128
A largeamount of informationwas providedwith
respect to Shanghai soils, including laboratory, field
testsandgroundbehaviorduringconstruction. Lackof
definitionof thereportedparametersisjudgedtomaje
theinterpretationof thereportedinformationnoteasy.
Toallowfor abetter consistencyandcompletenessof
reporteddata, ashortdraftguidelineonreportingcase
historiesisproposed.
ACKNOWLEDGEMENTS
Thewriter wishes to acknowedgetheauthors of the
summarized papers for sharing valuableinformation
withthegeo-community andtheorganizingcomittee
for invitinghimtodeliver thisgeneral report.
REFERENCES
Antiga, A. andChiorboli, M. 2008. Tunnel facestabilityand
settlement control usingearthpressurebalanceshieldin
cohesionlesssoil. IS-039.
Eclaircy-Caudron, S., Dias, D. and Kastner, R. 2008. Dis-
placements andstresses inducedby atunnel excavation:
caseof BoisdePeu(France). IS-107.
Gong, Q. and Zhou, S. 2008. Shield tunneling beneath
existingrailwaylineinsoft ground. IS-013.
Guiloux, A., LeBissonnais, H., Marlinge, J., Thiebault, H.,
Ryckaert, J., Viel, G., Lanquette, F., Erridaoui, A. and
Hu, M. 2008. Case history on a railway tunnel in soft
rock(Morocco). IS-367.
Hsiung, B. andChuay, H. 2008. Observedbehaviour of deep
excavationsinsand. IS-005.
Kim,S.,Yang,H.andYoon,S.2008.Environmental problems
of groundwater aroundthelongest expressway tunnel in
Korea. IS-087.
Konda, T., Ota, H., Yanagawa, T. and Hashimoto, A. 2008.
Measurements of ground deformations behind braced
excavations. IS-337.
Liu, D., Wang, R. andLiu, G. 2008. Researchontheeffectof
buriedchannelstothedifferential settlement of building.
IS-118.
Liu, G., J iang, J. and Ng, C. 2008. Performanceof adeep
excavationinsoft clay. IS-082.
Liu, S. andWang, Z. 2008. Deformationmonitoringduring
constructionof subwaytunnelsinsoft ground. IS-120.
Liu, T., Liu, G. andNg, C. 2008. Theconstructionandfield
monitoringof adeepexcavationinsoft soils. IS-029.
Mei, Y., J iang, X., Zhu, Y. and Qiao, H. 2008. Excavation
entirely on subway tunnels in the central area of the
PeoplesSquare. IS-140.
Osborne, N., Ng, C. and Cheah, C. Thebenefits of hybrid
ground treatment in significantly reducing wall move-
ment: aSingaporecasehistory. IS-378.
Qiu, D., Zhou, K., Ding,Y., Liang, Q. andYang, S. 2008. 3D
deformationmonitoringof subwaytunnel. IS-151.
Quick, H., Michael, J., Meissner, S. and Arslan, U. 2008.
Challengingurbantunnellingprojects insoft soil condi-
tions. IS-358.
Shao, Y. andMacari, E. 2008. Informationfeedback analy-
sisindeepexcavations. ASCE Int. J ou. Geom. Vol. 8, 1,
91103.
Wang, R., Cai,Y. andLiu, J. 2008. Supervisionandprotection
of Shanghai MassRapidLine4shieldtunnelingacrossthe
adjacent operatingmetroline. IS-033.
Wong, K., Ng, N., Leung, L. and Chan, Y. 2008. Kowloon
Southern Link TBM crossing over MTR Tsuen Wan
LinetunnelsinHKSAR. IS-370.
Xu, Q., Ma, X. andMa, Z. 2008. Applicationof pileunder-
pinning technology on shield machinecrossing through
pilefoundationsof roadbridge. IS-326.
Yoo, C., Kim, S. and Lee, Y. 2008. Characteristics of
tunneling-induced ground settlement in groundwater
drawdownenvironment. IS-329.
Zhao, H., Liu, X., Yuan, Y. andChi, Y. 2008. Effect of long-
termsettlement onlongitudinal mechanical performance
of tunnel insoft soil. IS-199.
129
Theme 1: Analysis and numerical modeling of
deep excavations
Optimizationdesignof compositesoil-nailinginloessexcavation
G.M. Chang
Urban Construction and Environment Engineering Department, West Anhui University, LuAn, Anhui, P.R. China
School of Civil Engineering, ChangAn University, XiAn, Shanxi, P.R. China
ABSTRACT: Theloess excavation has its uniquecharacteristics comparedwith theothers dueto its struc-
tural property and collapsibility. In order to acquire the work mechanismand design methods of composite
soil-nailinginloessexcavation, areasonablefiniteelement analysismodel isselected. Theoptimizationdesign
methods areintroduced based on theresults of finiteelement analysis conducted to determinetheregularity
of deformation, thesafety factor andtheendogenforceof thestructurealongwiththechangeof designvari-
able. Finally, the optimization design methods are validated contrasted with the data measured in an actual
project.
1 INTRODUCTION
Composite soil-nailing combined soil nails with
other forms of supporting measures has avoided the
soil-nailing technology from excessive dependence
onthesoil andexpandeditsapplicationfield. Among
the different kinds of composite soil-nailing forms,
theanchor compositesoil-nailing support method is
widely appliedfor its powerful location adaptability,
easyconstruction, lowcostandreducingthepitdefor-
mationremarkably. However, itsworkmechanismand
designmethod, especially theLoess Pit anchor com-
posite soil-nailing research, fall behind the project
practicebyfar. Inthefirst instance, thispaper aimsat
studyingnail designparameter selectioninplainsoil-
nailingonthepremiseof maintainingsoil-nailingtotal
length, andthenreplacingaanchor for asoil nail to
researchtheparametervalueof anchorcompositesoil-
nailing structure and optimization design under the
circumstancesof maintainingplainsoil-nailingdesign
parametersamoreoptimal value.
2 PARAMETERANALYSIS
Theoverall stabilityandworkingperformanceof exca-
vation supporting is closely related to the design
parameters. Understandingandgraspingtherelations
of theoverall stabilitysafetyfactor withthechangeof
thesedesignvariables, particularlythiskindof sensi-
tivity degreethat variety, havespecial andimportant
meaningfor guidingengineeringpractice.
2.1 Hypothesis
To simplify thecalculations, wemakethefollowing
assumptionswhencarry onthenumerical analysisto
thecompositesoil-nailingnumerical analysis:
1 Composite soil-nailing problems are plane strain
problems;
2 Soil-nailingandassistancereinforcementmaterials
areelasticmaterials;
3 Thesoil ispresumedastheelastic-plasticmaterial.
2.2 Computation diagram and parameter
of material
2.2.1 Computation diagram
Engineering experience shows that the influence of
excavationwidthisabout 3to4timesof theexcava-
tion depth, influence depth is about 2 to 3 times of
excavation depth. Thecaseassumes that theexcava-
tiondepthis9.5m,thetotal lengthof thefiniteelement
model is 45m, thetotal height is 25mandtheslope
gradient is1:0.1(Fig. 1).
2.2.2 Boundary conditions and loads
Ontheleftandrightboundaryof themodel, wesetthe
X-directiondisplacementtozeroandallowtheY direc-
tiondeformation; theX andY directiondisplacement
of thebottomboundaryarezero; thetopisafreesur-
face. Initial stressfieldisgravitystressfield; thevalue
of Gravitational Accelerationis 9.8m/s
2
. Sincecom-
positesoil-nailingisusuallyconstructedafter precip-
itation, theimpact of groundwater isnot considered.
133
Figure1. Finiteelementanalysismodel (Anchoratmiddle).
2.2.3 Material parameters
This research was completed against theloess exca-
vations and the soil parameters were provided from
a engineering investigation report in Xian City.
Because the soil distributes in certain scope are
uneven, it isdiscommodioustocarry ontheNumeri-
cal Calculationandtakethesoil strengthaveragevalue
of eachlevel. Soil nailsandanchorstaketheformof
thecommonly used procedurein Xian: Soil-nailing
110mmdiameter bored, steel bar 1 22; theanchor
holediameter 150mm, steel bar 2 18, surface100
for C25 thick concrete, reinforced with distribution
steel bar network. Soil nail andanchor weremadeof
steel bars that wrappedwithcement slurry composi-
tion. Slurry tightly wrappedtheexternal part of steel
bars, andoccludedwiththesoil indogtooth. Inorderto
simulatethemechanical behavior of soil-nailingand
anchors correctly and simplify finiteelement analy-
sis process, we regard the steel bar and the cement
pastebodyasakindof compoundmaterial. Materials
geometricandmechanical parameters, asfollows:
Soil: c =30kpa, =18
, gravity =18KN/m
3
,
deformation modulus E
0
=1.810
7
Pa, Poissons
ratio j=0.3; Soil-nailing: diameter 0.11m, sec-
tional area is 0.0094985m
2
, moment of inertia
1.832410
6
m
4
, equivalent modulus of elasticity
E
eq
=210
10
Pa, Poissonsratioj=0.3;
Anchor: the sectional area of free segment is
5.086810
4
m
2
, moment of inertia is 1.030077
10
8
m
4
, elastic modulus E
s
=210
11
Pa, thesec-
tional area of anchorage segment is 0.0176625
10
4
m
2
, moment of inertia is 2.48378910
5
m
4
,
equivalent elastic modulus E
eq
=2.0310
10
Pa.
Poissons ratio j=0.3; Surface (unit length) :
The sectional area is 0.1m
2
, moment of inertia
is 8.3333310
5
m
4
. Equivalent elastic modulus
E
eq
=2.110
10
Pa, Poissonsratioj=0.3.
Contact surfacefriction: Accordingtotheresearch
of reference(Chen, 2000, Wang, 1997), thesoil-nail
contact surface and the soil-anchor contact surface
frictionvalueis60kPa.
0
1
2
3
4
5
6
7
8
9
10
28 24 20 16 12 8 4
lateral deformation/mm
d
i
s
t
a
n
c
e

t
o

g
r
o
u
n
d
/
m
0 5 10
15 20 25
Figure 2. Relations of soil-nailing angle and pit
displacement.
Table1. Relationsof soil-nailingangleandsafetyfactor.
Soil-nailing 0 5 10 15 20 25
angle
Safety 1.593 1.623 1.654 1.615 1.568 1.53
factor
2.3 Soil-nailing support
2.3.1 The angle of soil-nailing
Concerning with theconstruction method, theangle
of soil-nailing has great influence on the pit
displacement, the safety factor and the surface sub-
sidence.Takingthetotal lengthof soil-nailingis40m,
establishingfiverowsof soil-nailing, takingthesoil-
nailinglevel andthevertical spacingtakes1.8m, the
first row of soil-nailing depth of burying is 1.8m.
Dividing fivesteps excavates, thefirst step of exca-
vation depth is 2.3m, and the other step of cutting
depthis1.8meach. Soil-nailingobliquitiesarecalcu-
latedby inclinationof 0
, 5
, 10
, 15
, 20
and25
respectively.
Figure2andTable1showthat thehorizontal dis-
placement isgradually increasingandchangingat an
increasingly rapid pace as soil-nailing angle from0
degreesto25changesgradually. Whenthepit design
requiresstrict control of thehorizontal displacement,
they should use a smaller angle. Safety factor in
soil-nailing angle reduces 10 degrees at the largest
and declines rapidly with the angle increases after
10degrees.
On the other hand, Soil-nailing angle is related
withconstructionmethods andsoil-nailingconstruc-
tion usually adopt the self grouting methods, in the
hopeof soil-nailinghasmoreinclinationtomakethe
cement grout fill soil-nailingholes under theweight
easily.
134
Figure3. Longat upper rowandshort at lower row.
Figure 4. Long at middle row and short at upper and
lower row.
Figure5. Short at upper rowandlongat lower row.
Soafterconsideringthepitdisplacement, thesafety
factor and construction factors, the angle should be
about 1015degrees.
2.3.2 Scheme of soil-nailing layout
Other researchersdomoreabout theschemesof soil-
nailinglayout (Hu& Song, 1997, Li & Zhang, 1999).
But their studies focus moreoncomparisonbetween
longatupperrowandshortatlowerrowscheme(long-
shortscheme) andshortatupperrowandlongatlower
rowscheme(short-longscheme). But inpractice, we
oftenuselongatmidrowandshortatupper andlower
rowscheme(mid-longscheme), especially whenthe
soil-nailingisusedwithanchor together (Figs. 35).
0
1
2
3
4
5
6
7
8
9
10
25 20 15 10 5
d
i
s
t
a
n
c
e

t
o

g
r
o
u
n
d
/
m
long-short
mid long
short-long
Figure6. Relationsof soil-nailinglayoutanddisplacement.
Table 2. Relations of scheme of soil-nailing layout and
safetyfactor.
Schemeof layout Long-short Short-long Mid-long
Safetyfactor 1.478 1.447 1.376
Selecting nail angleis 10 degrees then calculates
and analyzes on three different layouts, we get the
resultsFigure6andTable2below.
FromFigure6andTable2wecanfindthatthereare
smallest displacement andlargest safety factor when
using long-short scheme. On the contrary, there are
largest displacement andsmallest safety factor when
usingshort-longscheme.Asthesameconclusionwith
our forerunners, displacement and safety factor that
use short-long short scheme are between the other
two modes and we can see that when using plain
soil-nailing support a long-short scheme should be
adopted.
2.4 Anchored soil-nailing support
At present, the application of prestressed anchor in
compositesoil-nailingdesignisoftenusedempirically
and there has not a determinate calculative method
to set the anchor position, the length of anchorage,
prestressedvalue.
Basedonpreviousstudies, wechosethenailsangle
of 10degreesandlong-shortlayoutschemetoresearch
theanchoredsoil-nailingsupport.
2.4.1 Location of prestressed anchor
After replacingthe1st and3rdrowsandthefifthrow
of soil nailswithanchor, let usstudytheinfluenceof
blotlocationonthepitslevel displacementandsafety
factors. Diagrams shows in Figures 78, 1, and the
resultsareshowninFigure11andTable2below.
135
Figure7. Anchor at upper row.
Figure8. Anchor at lower row.
0
1
2
3
4
5
6
7
8
9
10
25 20 15 10 5 0
lateral deformation /mm
d
i
s
t
a
n
c
e

t
o

g
r
o
u
n
d
/
m
soil
upper anchor
mid anchor
lower anchor
Figure9. Relationsof anchor locationsanddisplacement.
Table3. Relationsof anchor locationsandsafetyfactor.
Anchor locations Upper Middle Lower
Safetyfactor 1.496 1.546 1.511
Figure 9 and Table 3 show that add prestressed
anchor into soil-nailing can significantly reduce the
maximumhorizontal displacement, especially inand
near theanchor location. Inaddition, comparedwith
0
1
2
3
4
5
6
7
8
9
10
25 20 15 10 5 0
d
i
s
t
a
n
c
e

t
o

g
r
o
u
n
d
/
m
0kN
100kN
150kN
200kN
250kN
Figure10. Relationsof prestressinganddisplacement.
Table4. Relationsof prestressingandsafetyfactor.
Prestressing
value/kN 0 100 150 200 250
Safetyfactor 1.546 1.597 1.617 1.624 1.628
plain soil-nailing, it can significantly reduce the
level of surface displacement through adding pre-
stressed anchor, particularly the top-anchor scheme
and bottom-anchor schemehavethemost obviously
effect on the surface of the horizontal displacement
control. Anchor locations also affect the safety fac-
tor. It hasthebiggest safetyfactor whenanchor at the
central pit.
Therefore, tocontrol thepit deformation, theangle
of anchor wouldfavor theupper-anchor scheme; the
mid-anchor schemeisthemost beneficial toimprove
the safety factor. However, in the engineering prac-
tice, becauseexcavationsconcentratemoreandmore
onurbanareas andtheanchor may into thepit slope
outsidemoredistance, theremay affect anchor con-
structionfor theadjacent buildingswhenusedthetop
anchor scheme.
2.4.2 Level of prestressing value
Used prestressed anchor replace with the 3rd soil-
nail, taking the anchor free segment length is 10m,
anchoragesegment length is 8m, prestressing value
is 0kN, 150kN, 200kN and 250kN respectively for
calculating. Theresults calculated fromthehorizon-
tal displacement and thesafety factors areshown in
Figure10andTable4.
Figure 10 and Table 4 show that the impact of
prestressing value on the horizontal displacement
is greater. When the magnitude of prestressing is
100kN, horizontal displacement decreasesmorethan
plain soil-nailing from the maximum displacement
of 23.1mmto 17.6mmlower. If prestressing value
136
0
1
2
3
4
5
6
7
8
9
10
20 16 12 8 4 0
d
i
s
t
a
n
c
e

t
o

g
r
o
u
n
d
/
m
6m
8m
10m
12m
14m
Figure 11. Relations of anchorage segment length and
displacement.
Table5. Relationsof anchoragesegment lengthandsafety
factor.
Anchorage
length/m 6 8 10 12 14
Safety 1.586 1.617 1.629 1.637 1.641
factor
increases to 150kN, the maximum horizontal dis-
placementwill decreaseto15.3mm.If theprestressing
valueis morethan 150kN, such as 200kN, 250kN,
horizontal displacement will continueto decline, but
the reduction range is insignificant. As the value of
the prestressing increases, the Pit safety factor will
gradually increase in a limited extent, which means
thevalueof prestressinghasnosignificant impact on
safetyfactor.
2.4.3 The length of anchorage segment
Selecting a middle-anchor scheme, the prestressing
value is 150kN, taking the length of anchorage for
6mand8m, 10m, 12m, 14mtocalculate, theresults
areshowninFigure11andTable5.
Figure11andTable5showthat thehorizontal dis-
placementof Pitgraduallydiminishesastheanchorage
lengthincreases, butthereductionismodest. Pitsafety
factorwouldincreaseasthelengthof anchorincreased
either, but not markedly.
3 ANCHORANDSOIL-NAILINGWORKING
TOGETHER MECHANISM
Plain soil-nailing support is a passive support sys-
temandsoil-nailingwouldhavearoleonlywhenthe
soil generatessufficientdeformation. Anchor belongs
to the initiative support system and through pre-
stresstocontrol soil deformation.Anchoredcomposite
soil-nailingisaspecial kindof support,whichbetween
plain soil-nailing and prestressed anchoring sup-
port. It has the advantages of both the plain soil-
nailing support and the prestressed anchor support
simultaneously.
3.1 Anchor and soil-nailing working together
It isat theinitiativestressful conditionasanchor sup-
port constructioncompletedbecauseof theexistence
of prestressing. As a result of the anchor prestress-
ing reaction, the soil is caused to be at the pressed
condition, reduced soil lateral deformation. On the
other hand, anchor is wrapped in the cement paste,
andadhibitedwithcementpaste. Becauseof theholes,
pores and crannies existed in soil; the cement and
thesoil assumethezigzaglinking. After anchor ten-
sion deformation, there will have a shear stress due
to elastic deformation and retraction in the anchor-
soil interface, whichdirectiononthesoil deformation
under shear stressisthecontrary. Itreducessoil inter-
nal tensilestress, andwill alsolimitthedeformationof
soil. Theaxial forceof soil-nailingisrelatedwiththe
deformation of theearth. Becausestress reduces the
soil deformation, soil-nailinginternal forceisreduced
more remarkably than plain soil-nailing .The closer
the anchor approaches the soil nails, the more the
axial forcedecreases.Therefore,theroleof restrictions
pit deformationisthebaseof anchor andsoil-nailing
workingtogether.
3.2 Anchor contribution to resistance moment
When thePit Slopein theevent of damage, theslip
surface have too much plastic deformation to make
slidemassalongfor destructionunder sliding. Gener-
ally, theslidingmomententirelydependsonthedepth
of excavationandthesoil gravity. For acertainpit, the
soil depthandits weight usually areconstant andits
slidingmoment canbeseenasaconstant. Meanwhile
resistancemomentisprovidedbytheundisturbedsoil,
shear strength, soil nailsandanchor.
Thecontribution of soil nails performancelies in
three main aspects: soil-nailing presence gives the
slip surfaceplaceto thepost-transfer slip, improves
theslidingareaandincreases thefrictionof slipsur-
face. Upliftroleof thesoil-nailingthatoutsidetheslip
surface, andthebendingresistanceroleof soil-nailing.
Thecontributionof anchor mainfeatures (Chang,
2007):
1 The anchors anchorage is long in general and
extendsintothesteadysoil massincentral slipaway
fromtheexcavation surfaceto provideastronger
uplift capacity.
2 Theanchorsprestressingmakesslideandstability
soil mass tightly squeezed each other to improve
the friction resistance to sliding and increase the
resistancemoment.
137
3 When the slip surface crosses the anchorage
segment, the anchor resistance to bending has
somecontribution, but thecontributionisweak in
general.
3.3 The impact of prestressing to soil-nailing
axis force
Therearemanystudiesabout theimpact of prestress-
ingto soil-nailingaxis forceandtheconclusions are
thesame. Generally, theprestressingwill significantly
reducesoil-nailingforcesandthegreater thevalueof
prestressing, thesmallerthesoil-nailinginternal force.
And moreover, thecloser theanchor, thegreater the
internal force reduction (Zhang & Liu, 2002, Zhen
et al. 2005).
4 DESIGNOPTIMIZATION
1. Soil-nail design. Recommending taking the soil-
nailings long-short layout and it is advisable to
select 1015degrees.
2. Loess has strong structure, which should beused
asmuchaspossible.
3. Anchorshouldbeinstalledclosetothecentral verti-
cal partof thepit, whichcanachieveahighersafety
factor andrestrict thepit deformation.
4. Itdoesnotprovideagreatersafetyfactorandbetter
control deformationeventheanchor lengthis too
long. Soit hadbetter beabout 812m.
5. Since the anchor prestressed reaction limits soil
lateral deformation, reduces the lateral displace-
ment of compositesoil-nailingretainingandaxial
force of the-soil-nailing near the anchor. On the
premise of meeting deformation control request,
wecanshortenthelengththat several soil-nailing
topof thepit appropriately inlong-short scheme,
long-short-long scheme while forming to reduce
projectcost. However, inorder nottoreducesafety
factor, it isnot recommendedtoshortenthelength
of soil nailsinthelower side.
6. Thefixingontheprestressedvaluemustbeaccord-
ingtothesoil shearstrengthvalues, itwill beabout
100kN to 200kN as well. Dueto theprestressed
valuehasnoobviousinfluenceonproject cost, we
caninclinesafetytochoosethevaluesgreater. But
toomuchprestressedhasnosignificant impact on
pit retainingperformance.
5 ENGINEERINGANALYSIS
5.1 Project overview
AprojectinXian,theexcavationdepthis11.0m,both
thePitslengthandwidthareabout 100m. Intheeast
Table 6. Site layer structure and geotechnical
characteristics.
Unit Angleof
Soil Thickness/ weight Cohesion/ internal
class (m) (kN/m3) kPa friction/
Miscellaneous 0.90 18.20 25.00 18.00

LoessQ3
2EOL
6.60 16.20 28.00 18.00
LoessQ3
2EO
L 1.70 18.00 26.80 18.10
Ancient soilQ3lal 3.40 18.90 32.20 17.60
LoessQ3
al+PL
8.00 19.50 20.00 18.00
bynorthfromthepit 6.3misaseven-storeymasonry
structureresidential buildings, thelimesoil founda-
tion depth is about 3m.On the south-east there is a
18-storeyhigh-risebuildingwithone-storeybasement
whichdepthis 6m, reinforcedconcretepilefounda-
tionsare36mlong, fromPit 9.10m, andtheadjacent
side of the project pit used soil-nailing in the con-
struction;northof theseven-storeyresidential building
masonry structure, androughly parallel to pit edges,
buildingslengthis42mandthewidth13m, thenear-
est to excavationis 5.3m; thewest sideof theSouth
andanadjacent hotel podiumwhichis atwo storeys
buildingwithanundergroundlayer fromtheedgeof
pit 4m, framework and infrastructure end elevation
is 7.13m; the south side is close to a main road,
therearewater andgaspipelinesunder thesidewalks.
5.2 Engineering geological conditions
According to geotechnical engineering investigation
reportthattheprojectsitegeomorphicunitsbelongto
theLoessbeam-swamplandscape. Proposedsitelayer
structureandgeotechnical characteristicsareinrange
of 30.0mdeepinTable5.
5.3 Retaining design
Thedesignof anchoredcompositesoil-nailingadopts
themethodsproposedinpart 4of thispaper, whichis
the excavation depth was 11.0mand the slope was
1:0.1. The basic design parameters were shown in
Table6.Therearesixlayersof soil nailsandthelayout
is cinquefoil. Wealso set aprestressedanchor at the
depthof 6.0mtoreducethelateral displacementand
ensurethehigh-risebuildings insafety andstability.
Using two 18mmdiameter grade 60 bars in anchor
(10mfree, 8manchorage); Usingone22mmdiame-
ter grade60bar insoil-nailingwith1.5mspacingand
inclinationof 15degrees, prestressingvalueis150kN.
(Fig. 12).
5.4 Monitoring results
ThisprojectwasconstructedsinceMarch20, 2006and
lasted75days. Monitoringresultsshowedthegreatest
138
Figure12. Compositesoil-nailingsupport diagram.
lateral displacementoccurredintheeastcentral pitand
thelargest displacement was16mm. It wasnoexces-
sivelateral deformationandearthsurfacesubsidence
andsurroundingbuildingswerenogreater settlement.
6 CONCLUSIONS
Inthis paper, theauthor analyzedtheparameter sen-
sitivity of compositesoil-nailing in loess excavation
using the finite element method. The optimization
designmethodsareintroducedbasedontheresultsof
finiteelementanalysis.Thewriterbelievesthatanchor
should be installed in the pit in the central vertical,
whichcanachieveahigher safety factor andrestrict
thepit deformationobviously.
Wecanshortenthelengththat several soil-nailing
topof thepitappropriatelyinlong-shortscheme, mid-
long scheme while forming to reduce project cost.
Anchor lengthshouldnotbetoolongandprestressing
alsoshouldnot betoolarge. Practical project proved
thatthisoptimizationmethodinloesspitisapplicable.
ACKNOWLEDGMENT
This paper was supported by Youth Foundation of
Anhui Education Committee (No. 2007jql181), and
supportedbyYouthFoundationof WestAnhui Univer-
sity(No. wxzq2006018).
REFERENCES
Chang, G.M. 2007. Study on theApplication of Compos-
itesoil-nailing in Loess Excavations. XiAn: ChangAn
University
Chen, Z.Y. 2000. The application of soil-nailing in excava-
tions. Beijing: ChinaArchitecture& BuildingPress
Hu, K.G. & Song, Q.G. 1997. Nonlinear analysis of action
mechanismof soil-nailingwall . Industrial Construction
27(11): 1013
Li, S.H. & Zhang, Y.J. 1999. Numerical simulations by 2D
FEM in process of excavation and supporting of deep
foundationditch. Chinese Journal of Rock Mechanics and
Engineering 18(3):342345
Wang, B.Y. 1997. Designof soil-nailing. Geotechnical Engi-
neering Technique (4): 3041
Zhang, F., Liu, Z.C. & Chen, G.G. 2002. The mechanical
workingmechanismresearchontheunitedsupportingof
prestressed soil anchor and soil-nailing. Rock and Soil
Mechanics 23(3): 292296
Zheng, Z.H., et al. 2005. In-situ testing study on retaining
miscellaneousfill slopebyusingcompoundsoil-nailing.
Chinese Journal of Rock Mechanics and Engineering
24(5): 898904
139
Three-dimensional finiteelement analysisof diaphragmwallsfor
top-downconstruction
J. Hsi, H. Zhang&T. Kokubun
SMECAustralia Pty Ltd, Sydney, NSW, Australia
ABSTRACT: TheTugunBypassTunnel inGoldCoast, Australiawasconstructedusingdiaphragmwallswith
thetop-downcut-and-cover methodto allowsimultaneous constructionof anairport runway extensionabove
thetunnel, whilst excavation of thetunnel continued underneath. Thetunnel was built in an environment of
high groundwater table and deep deposits of alluvial and estuarine soils with the toes of the walls founded
in soil deposits. There was a potential risk for differential settlements between the diaphragmwall panels,
causedbytherunwayfill placedoverthetunnel roof duringexcavation.Three-dimensional numerical modelling
was undertakento predict thedifferential settlements of thetunnel withconsiderations of varyingsubsurface
profile, stagedexcavation anddewatering, non-uniformloadingandcomplex soil-structureinteraction. Field
instrumentationandmonitoringwasimplementedtoconfirmnumerical predictions.
1 INTRODUCTION
The 7km long Tugun Bypass forms part of the
PacificHighway, andconnectssouth-eastQueensland
to northern NewSouthWales, Australia. Oneof the
keyfeaturesof theprojectwasatunnel of about334m
in length (Ch5588 to Ch5922.4), constructed below
theproposedrunwayextensionof theGoldCoastAir-
port. Figure1presentstheproject routeplanshowing
thelocalityof theproject.
Figure1. Project routeplan.
Asthetunnel wastobeconstructedintheproximity
of theairport runway, therewasastrict height restric-
tion for the construction activities. Low headroom
plant and equipment were chosen to construct the
diaphragmwalls for thecut andcover tunnel. As the
constructionof therunway extensioncoincidedwith
the tunnel construction, the top-down construction
methodwasadopted.
Thesubsurfaceof thetunnel sitecomprisedmainly
alluvial and estuarine soils up to depths of about
35munderlainby weatheredrock of NeranleighFer-
nvaleformation. To minimizeconstructioncosts, the
diaphragmwallswerefoundedinsoil depositswhich
weresubjected to settlement under theloading from
therunwayextension.
Excessivedifferential settlement of thediaphragm
walls couldoverstress thetunnel structureandaffect
the tunnel serviceability. Detailed numerical mod-
ellingwascarriedoutusingthefiniteelementpackage
PLAXIS3DFoundation(Version1.6) wherethespa-
tial subsurface variation and non-uniform loading
patternscouldbetakenintoconsideration.
Instrumentation and monitoring were undertaken
todemonstratethefieldperformance, whichwasthen
comparedwiththenumerical predictions.
2 SITE GEOLOGY
The tunnel was situated in a flood plain which was
subjected to periodical flooding. Thegeology of the
site comprised Neranleigh Fernvale Beds overlain
by Cenozoic estuarine and coastal deposits. These
141
Figure2. Siteinvestigationplan.
depositswereupto35minthickness, comprisingriver
gravels, sandsandclays, andfloodplainandtidal delta
mudsandsilts. At thetunnel location, thesubsurface
horizons consisted of dune sands, Coffee Rock
(local termgiven to cemented silty sands), estuarine
interbeddedclaysandsands, andresidual soilsderived
fromtheweatheredbedrock.Groundwaterwasslightly
saline due to the close proximity to the ocean. The
water tablewas influenced by both tidal movements
and rainfall events recharging Cobaki Broadwater.
Dueto low-lying ground surfaces, potentials existed
for acidsulphatesoils.
3 GEOTECHNICAL MODEL
Asthesubsurfaceconditionsvariedspatiallyalongthe
lengthandwidthof thetunnel, extensivesiteinvesti-
gationsusingboreholes(BH) andpiezocones(CPTU)
were undertaken at the wall and barrette locations.
Withinthefootprintof therunwayextension,theinves-
tigationsweredoneataspacingof approximately20m
intervals.Theplanof thesiteinvestigationisshownin
Figure2.
The geotechnical model of the site included sub-
surface stratigraphy and geotechnical parameters.
The subsurface was divided into discrete soil units,
classifiedaccordingtomaterial typeandconsistency
or densityandissummarizedasfollows(topdown):
Top soil thin skinned (-1m) comprising peaty
sandy organic topsoil, having loose consistency.
The ground surface was marshy and generally
untrafficable;
Dunesands asequenceof generallyloosetovery
loosesands of upto about 8to 10minthickness,
finetomediumgrainedsands;
Coffee Rock (CR) a sequence of medium
densetovery densecementedsilty sandsof about
7 to 10m in thickness with occasional loose
consistency;
Estuarine a sequence of about 15mthickness
comprising shell fragments, sand and silty sand,
clayandsandyclay, siltandclayeysilt, clayeysilty
sandandgravels. Relativedensityvariedfromvery
loosetodense, andconsistencyvariedfromfirmto
verystiff;
Figure3. Subsurfaceprofile.
Table 1. Typical soil profile and key geotechnical
parameters.
E
ref
50
&
RL
k E
ref
oed
E
ref
ur
(m) Soil (deg) (m/day) (MPa) (MPa)
0.5 Sand(VL) 30 1.0 10 30
4.0 CR (MD) 32 0.1 50 150
11.2 CR (D) 34 0.1 80 240
13.5 Sand(L) 32 1.0 30 90
17.5 Clay(St) 28 110
4
10 30
21.1 Sand(L) 32 1.0 30 90
23.0 Clay(F) 24 110
4
7 21
28.6 Clay(VSt) 29 110
4
25 75
30.8 Bedrock
Note: RL (reducedlevel) is at topof eachlayer; VL is very
loose; L islose; MDismediumdense; Disdense; F isfirm;
St isstiff; VSt isvery stiff;
isdrainedfrictionangle; k is
permeability; E
ref
50
is secantYoungs modulus at areference
pressureof 100kPa; E
ref
oed
istangentYoungsmodulusfor pri-
mary odometer loading at areferencepressureof 100kPa;
and E
ref
ur
is unloading/reloadingYoungs modulus at a ref-
erence pressure of 100kPa. Refer to PLAXIS manual for
HardeningSoil (HS) model.
Residual soil comprising clay and silty clay
with some sands, and with residual fragments of
extremely weathered and extremely low strength
interbeddedargilliteandgreywackeof theNeran-
leighFernvaleBeds.Thethicknessrangedbetween
about 1mand6m;
Bedrock comprising extremely weathered to
moderately weathered and extremely low to low
strengthinterbeddedargilliteandgreywacke, hav-
inganirregular contact withtheoverlainresidual
material at adepthof approximately30to35m.
The subsurface profile based on the boreholes
along the centre line of the tunnel is presented in
Figure3. Thegeotechnical parametersfor eachof the
unitsweredeterminedfrominterpretationof thefield
and laboratory test results, and based on local expe-
rience. Thetypical soil profileand key geotechnical
parametersassumedareshowninTable1.Theground
142
surfacelevel was approximately at RL 0.5mandthe
groundwater tablewasat thesurface.
4 ISSUESANDCONSTRAINTS
Construction of a tunnel in soft ground at shallow
depthsisconventionallyundertakenusingthecutand
cover method. However, to allow for construction
of the runway extension that occurred concurrently
withthetunnel excavation, thetop-downconstruction
methodhadtobeadopted. Diaphragmwallsandcast
insitutunnel roof slabshadbeenchosentofacilitate
the construction requirements and time constraints.
Figure2showsthefootprint of therunway extension
obliquetothetunnel alignment.
Following the handover of ground surface, up to
23mof fill for the airport runway extension was
placed above the tunnel roof. Loads acting over the
entirewidthof theroof slabsweretransferreddirectly
to the diaphragm walls and the barrettes. The site
investigationsrevealedpresenceof estuarinedeposits
consisting of loose materials below the toe of the
walls. Therefore, therewas apotential for thetunnel
to settleduringexcavation. Oneof thecritical issues
wasthedifferential settlementsbetweenthewallsand
thecentral barrettes, andalongthewalls.Thesediffer-
ential settlementscouldpotentiallyinducesignificant
stressesintheroof structuresandinthewalls.
Other issues in relation to thetunnel construction
arelistedbelow:
ObstacleLimitationSurface(OLS) appliedat
bothendsof therunwaytoprovidesafeairspacefor
approaching aircrafts. This required all construc-
tionactivitiestobeundertakenwithinaheadroom
of as low as 8m. Use of cranes or heavy-lifting
equipment was only allowed outside the airport
operatinghours;
Highgroundwaterlevel duetoitscloseproximity
totheseaandCobaki Broadwater.Thegroundwater
waspracticallyatthegroundsurfacelevel. Reliable
dewateringsystemwasessential duringexcavation;
Environmental requirements strict environmen-
tal controls wereenforcedsuchthat drawdownof
thegroundwater tableoutsidethediaphragmwalls
was insignificant. Also, all acidic sulphate soils
excavated from the tunnel had to be dried and
neutralizedwithlimeprior to placement as fill in
embankments.
5 CONSTRUCTIONMETHOD
Suitableconstructionmethodswerechosentoaddress
theissuesandconstraintsmentionedabove. Inorderto
adheretotheOLSrequirements,special lowheadroom
hydraulic grab(Leibherr HS852HD) and2.8mwide
trenchcutter(CBC25)wereused.Theguidewallswere
Figure4. Diaphragmwall constructionsequence(courtesy
of Bauer/PilingContractors).
built first followed by construction of the 6mwide
primary panels (Steps 1 to 5 of Figure4) and 2.8m
widesecondarypanels(Steps6to8). Theopentrench
wassupportedbymixtureof bentoniteslurry,whenthe
cutterundertookfull excavation(Steps2to3and6).A
steel reinforcement cagewasloweredwhenthepanel
wasexcavatedtofull depth(Steps4and7). Concreting
of thepanelswasthenachievedbythetremiemethod
(Steps 5 and 8). Figure 4 presents the construction
sequenceof thediaphragmwall.
Following completion of the diaphragm walls
andbarrettes, dewateringandexcavationcommenced
insidethewalls. Excavationwas initially undertaken
to depths of up to about RL 2mto allowfor con-
structionof theroof slab. Water-tight membranewas
installedas part of thewater-proofingsystem. When
theroof slabwascompleted, it wasbackfilledandthe
sitewas clearedfor handover to theGoldCoast Air-
port. Theseactivities commencedinApril 2006after
environmental approvalsweregranted, andwerecom-
pleted by November 2006 which was the scheduled
dateof handover of thesitesurface. Excavationbelow
the runway extension continued through to J anuary
2007, and the remaining construction of the tunnel
continued.
6 STRUCTURAL DETAILS
The tunnel structure consisted of diaphragm walls
andbarrettes locatedat thecentreof thetunnel. The
diaphragmwallswere1minthickness, andextended
from the Northern Portal (Ch5588) to the South-
ern Portal (Ch5922.4). The walls were installed to
the depth of RL 17m, fromthe top of the guide
wall at RL 2m. The internal width between the
diaphragm walls ranged from about 25.7m at the
northern portal to 28mat the southern portal. Bar-
rettes were 0.8mthick and 2.8mwide with a clear
spacingof 2.8mthroughoutthecentral axisof thetun-
nel, extendingtoRL 17mindepth.Thesestructures
hada100year designlife, usingN-gradereinforcing
143
Figure5. Typical tunnel crosssection.
steelsand50MPahighstrengthconcrete. Therewere
nomechanical joints at theinterfaceof theprimary
and secondary panels in the longitudinal direction.
However, thebarrettesandthediaphragmwallswere
rigidlyconnectedtothe1mthickroof slab.Therewere
threejetfannicheswheretheroof slabwasslightlyele-
vated.Thebaseslabwasalso1mthickwithafounding
level rangingfromRL 5.5mtoRL 9.5m. Figure5
showsthetypical crosssectionof thetunnel.
7 DESIGNCONSIDERATIONS
Geotechnical design of the tunnel was required to
satisfythefollowingthreekeyissues:
Excavation support during construction the
diaphragmwall structuresweredesignedtoensure
stability of the excavation. Issues included struc-
tural design of the walls, base heave, hydraulic
uplift, piping, andliquefaction;
Longtermstabilityof thetunnel buoyancyof the
tunnel whenthegroundwater tablewasclosetothe
surface;
Serviceability assessment duetosettlement of the
tunnel during construction the tunnel was sub-
jected to loading fromairport runway fill which
resulted in settlements. The influences of dif-
ferential settlements on structural capacity were
assessed.
8.1 Two-dimensional numerical modelling
Design of the tunnel was initially undertaken using
thefiniteelement softwarePLAXIS (Version8.4) at
selected sections. This numerical package was used
to analyze two-dimensional plane-strain problems
involving complex soil-structure interaction for the
design of the structural members. Structural beam
elements wereusedtosimulatethediaphragmwalls.
Global factorof safetyduringeachof theconstruction
stages was calculated based on the c
reduction
methodtoensuretheminimumFoSwasachieved.The
softwareallowedmodellingof constructionsequence,
changinggroundwater levels, andvaryingsubsurface
conditionsacrossthewidthof thetunnel.
8.2 Three-dimensional numerical modelling
A three-dimensional numerical modelling package,
PLAXIS3DFoundation(Version1.6), wasemployed
topredict thesettlementsof thetunnel causedbyrun-
wayfill loadingandexcavation. Duetothelimitation
of theprogram, settlement analyses wereundertaken
in sections, each of approximately 40 to 60m in
length. The major advantages of the 3D modelling
wereasfollows:
Abilitytomodel thephysical dimensionsof thewall
andbarrettestructures.Thisimprovedtheaccuracy
of settlement prediction, asit accountedfor longi-
tudinal stiffnessof thetunnel whichassistedinload
redistributionandtoeresistanceof thestructures;
Ability tosimulate3D loaddistributionwherethe
runway fill wasplacedobliquetothelongitudinal
axisof thetunnel;
Ability to model 3D subsurface profile based on
probeholesat discretelocations;
Ability to simulate dewatering within the tunnel
excavationarea.
The Hardening Soil (HS) model was considered
mostappropriatetosimulatesoil behaviourinanopen
excavation.TheHSmodel tookintoaccountunloading
andreloadingbehavior andirreversibleplasticstrains
of soil.TheHSstiffnessparametersweredefinedwith
respect to a reference pressure of 100kPa. The key
parameters included E
ref
50
, E
ref
oed
, and E
ref
ur
as shown in
Table1. Thepublisheddataindicatetheratio of E
ref
oed
toE
ref
50
isabout 0.7to1.4andtheratioof E
ref
ur
toE
ref
50
varies from2to 4. Theanalysis adoptedE
ref
50
=E
ref
oed
,
andE
ref
ur
=3E
ref
50
.
Presentedhereisa41.2mlongsectionof thetun-
nel between Ch5728.8 and Ch5770. This section of
the tunnel was of the deepest location of the tun-
nel, beneaththethickest layer of therunway fill, and
underlainbyslopingbedrock level andchangingclay
thickness. A jet fannicheof approximately12mlong
alsoliedwithinthecentreof thissectionwhichhadalso
beenincorporatedinthemodel. Withinthischainage
range, thereweresevenboreholes. Duetothecapacity
of theprogram, four representativeboreholes, which
wereevenlydistributedspatially, wereselectedfor the
analysis. Theassumedsubsurfaceprofiles areshown
inTable2.
8.3 Assumptions of analysis
Theconstructionsequencewasconsideredintheanal-
ysis to simulate the load transfer fromthe runway
fill tothediaphragmwalls. Theassumedconstruction
sequenceisdescribedbelow:
1. Applicationof loads exertedonthevirginground
fromthe working platformbuilt to RL 2m(for
construction of theguidewalls) and construction
loadof 10kPa;
144
Table2. Subsurfaceprofiles.
Borehole #1 #2 #3 #4
Location LHS RHS LHS Centre
Chainage 5730 5737 5757 5768 Soil type
(density/
Unit RL at topof eachlayer consistency)
1 0.5 0.5 0.5 0.5 Sand(VL)
2 4.8 3.4 5.0 2.8 CR (MD)
3 10.8 14.4 9.8 9.7 CR (D)
4 13.8 15.9 12.6 11.7 Sand(L)
5 17.8 17.1 17.5 17.5 Clay(St)
6 24.3 21.8 20.8 17.5 Sand(L)
7 27.3 25.1 21.8 18.0 Clay(F)
8 33.3 30.8 25.0 25.2 Clay(VSt)
9 36.3 32.0 27.4 27.4 Bedrock
Note: LHS is left hand side of tunnel facing increasing
chainage direction; RHS is right hand side of tunnel; and
Centreiscentrelineof tunnel.
2. Installationof diaphragmwallsandbarrettestoRL
17m;
3. Removal of theworkingplatform, andapplication
of 10kPaconstructionloadonsurface;
4. Dewateringandexcavationtoundersideof theroof
slab;
5. Installationof theroof slab(andjetfanniche), and
backfill toexistinggroundsurface;
6. Placement of runway fill todesignheights (simu-
latedaspressures) with10kPaliveloadabovethe
runway;
7. Staged dewatering and excavation within the
diaphragmwallstoundersideof thebaseslab;
8. Casting of the base slab and completion of the
tunnel structure;
9. Return of the groundwater table to the ground
surfaceandremoval of 10kPasurfaceloads.
The settlement assessment was undertaken at
stage 7, which was considered most critical with
maximumexcavationunder full runwayloading.
The assumed levels within the modeled chainage
rangearesummarizedinTable3.
8.4 Results of analysis
Thedeformedmeshof the3Dfiniteelement analysis
under thefull runway loading and at thefinal stage
of theexcavationisshowninFigure6. Thepredicted
settlementprofilesatthetopof theroof slabalongthe
diaphragmwalls andbarrettes prior to castingof the
baseslabarepresentedinFigure7.
Thepredictedsettlementof thetunnel duringexca-
vation was about 45mmon theLHS, 43mmon the
RHS, and 35mm along the central barrettes. The
maximumdifferential settlement was predictedto be
12mmbetweenthewalls andthebarrettes. To allow
Table3. Assumedgeometryduringconstruction.
5728.8 5737.6 5743.6 5755.2 5761.2
to to to to to
5737.6 5743.6 5755.2 5761.2 5770.0
Chainagerange
Feature RL (m)
Natural Ground 0.5 0.5 0.5 0.5 0.5
Level
Topof Roof Slab 0.8 0.25 +0.4 0.25 0.8
Bottomof Roof 1.8 1.25 0.6 1.25 1.8
Slab
Initial Excavation 2.8 2.8 2.8 2.8 2.8
Initial Dewatering 3.8 3.8 3.8 3.8 3.8
Intermediate 6.0 6.0 6.0 6.0 6.0
Excavation
Intermediate 7.0 7.0 7.0 7.0 7.0
Dewatering
Topof BaseSlab 8.4 8.4 8.4 8.4 8.4
Bottomof Base 9.4 9.4 9.4 9.4 9.4
Slab
Final Excavation 9.7 9.7 9.7 9.7 9.7
Final Dewatering 11.7 11.7 11.7 11.7 11.7
Toeof Diaphragm 17.0 17.0 17.0 17.0 17.0
Wall
Figure6. Deformed3Dfiniteelement mesh.
for uncertainties, thetunnel wasdesignedfor amaxi-
mumdifferential settlement of 25mm. Thestructural
analysisshowedthatthelongitudinal in-planestiffness
of thetunnel wouldsmoothoutdifferential settlements
alongthetunnel alignment, withthepresenceof thejet
fannicheandvariabilityof thesubsurfaceconditions.
9 FIELDPERFORMANCE
The performance of tunnel during construction was
assessed based on the field monitoring results. This
wasameanstoconfirmthatthestructural integrityof
thediaphragmwallsandbarretteswerenot adversely
affectedby differential settlements. Threeinstrumen-
tation arrays were set up at Ch5655, Ch5718, and
Ch5770correspondingtolocationsof therunwayfill
(seeFigure8).
Each array consisted of three settlement plates
placed above the LHS and RHS diaphragm walls
145
55
50
45
40
35
30
45 40 35 30 25 20 15 10 5 0
Distance along Centre Line (m)
S
e
t
t
l
e
m
e
n
t

(
m
m
)
Central Barrettes
RHS Diaphragm Wall
LHS Diaphragm Wall
CH5728.8 CH5770
Figure7. Predictedsettlement profilesat topof roof.
Figure8. Planof instrumentationarrays.
Figure9. Typical instrumentationsection.
and the central barrettes (see Figure 9). These were
installed prior to runway fill placement and excava-
tionof thetunnel inorder tocaptureall construction
induced movements. In addition to the settlement
plates, surveytargetswerealsoinstalledatinner walls
tothetunnel torecordtunnel movement duringexca-
vation. This informationhadto becalibratedagainst
thesettlementplatemeasurementsastheinitial tunnel
movement recordwasnot available.
Figure 10 shows a summary of construction
activities, recorded settlements, and the predicted
settlements at diaphragmwall and barrettelocations
at Ch5718. Thesettlement predictionadoptedhereis
60
50
40
30
20
10
0
10
26/11/2006 16/12/2006 5/01/2007 25/01/2007 14/02/2007 6/03/2007 26/03/2007 15/04/2007
Date
S
e
t
t
l
e
m
e
n
t

(
m
m
)
10
8
6
4
2
0
2
4
R
L

o
f

E
x
c
a
v
a
t
i
o
n

a
n
d

T
o
p

o
f

F
i
l
l

(
m
)
Measured at LHS
Measured at Centre
Masured at RHS
RL of Excavation
RL of Top of Fill
Predicted 35mm (Center)
Predicted 43mm (RHS)
Predicted 45mm (LHS)
Figure10. Settlement monitoringresultsat Ch5718.
theresult of analysisbetweenCh5728.8andCh5770.
Monitoring commenced at thebeginning of Novem-
ber 2006. Excavationof thetunnel commencedinmid
December 2006fromtheNorthernPortal at Ch5588.
TheexcavationprocessreachedCh5718inearlyJ an-
uary.Placementof runwayfill aboveCH5718followed
in mid J anuary, which had resulted in visiblesettle-
mentsof thetunnel. Thesettlementsappearedtohave
ceasedafter theexcavationreachedfinal depthinmid
February. Themonitoring datashowed that thefield
performanceof thetunnel wasconsistentwiththepre-
dictions obtained fromthe PLAXIS 3D Foundation
modelling.Maximumdifferential settlementsbetween
thebarrettes andthediaphragmwalls wereless than
25mmat all stagesof construction.
10 CONCLUSIONS
TheTugunBypasstunnel hadtobeconstructedunder
many strict constraints in a challenging geotechni-
cal environment. Extensive site investigations were
undertaken to better characterize ground conditions
and reduce risks of geotechnical uncertainties. The
top-downconstructionmethodwas adoptedto allow
extension of the airport runway to occur simultane-
ouslyduringtunnel construction.Theadditional loads
fromtherunwayfill inducedsettlementsof thetunnel
duringconstruction. Settlement analysisof thetunnel
using 3D numerical modelling techniques had been
undertaken. Thedifferential settlementsof thetunnel
weresuccessfully predicted. Theperformanceof the
tunnel wasmonitoredduringconstructionandthefield
measurements were consistent with the numerical
predictions.
REFERENCES
PLAXIS2D, Version8.4, PLAXISBV Netherlands, 2006.
PLAXIS 3D Foundation, Version 1.6, PLAXIS BV
Netherlands, 2006.
146
Numerical evaluationof dewateringeffect ondeepexcavationinsoft clay
L. Li
Tianjin Institute of Ubran Construction, Tianjin, P.R. China
Tianjin Key Laboratory of Soft Soil Characteristics and Engineering Environment, Tianjin, P.R. China
M.Yang
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University,
Shanghai, P.R. China
Tongji University, Shanghai city, P.R. China
ABSTRACT: Thispaper describestheapplicationof FLAC3Danalysisfor modelingatop-downconstruction
of a four-story (33.7mdepth), underground transformer substation in downtown of Shanghai city. There is
unconfinedaquifer andconfinedaquifer onthesiteof thisprojectanddrainagebydesiccationinthefoundation
pit isadopted. Theeffectivestressmethodsof analysisincorporateexcavationanddewateringof thefoundation
pitfor real-timesimulationof constructionactivities. Theresultsnotconsideringdewateringarecomparedwith
theresult consideringdewatering, includingwall deflections, basal heave, surfacesettlement. Theanalysis of
considering leakage of the wall and leakage between confined aquifer is provided also. The analysis shows
that although thedifferenceis small in soft clay dueto thelowpermeability of thesoil, dewatering enhance
thedeformationof thefoundationpit andthefoundationpit is inclinedto benot security if dewateringis not
considered, theeffect of leakageof thewall canbeobviousonthesurfacesettlement.
1 INTRODUCTION
Fordeepexcavationincongestedurbanenvironments,
designers are particularly interested in making reli-
able prediction of the magnitudes of movements in
thesurroundingsoil (Peck 1969, Cloughet al. 1989,
ORourke 1981) and then estimating the effects of
thesemovementsonadjacent structuresandfacilities
(Burland&Wroth1974, Boscardin&Cording1989).
In principle these prediction can be achieved using
powerful numerical methods such as finite element
analyses, but there is difficulties in achieving reli-
ableanalytical predictionsof soil deformationswhich
can be attributed to a variety of factors including
dewatering.
Dewateringisnecessaryintheexcavationunderthe
ground water level in soft clay, which provideadry
environment for theexcavationandis benefit for the
slopestability andreducingtheharminducedby the
groundwater, but dewateringhaveimportant effect on
thebehaviour of thefoundationpit andthesurround-
ingsoils, seepageinducedbydewateringfromoutside
to theinsideof thefoundationpit haveeffect onthe
stabilityandthedeformationof thefoundationpitand
surfacesettlementforthevertical consolidationbythe
underground-water drawdown out of the foundation
pit, so effective stress methods of analysis distinct
fromvast majority of analyses relies on total stress
methods of analysis andreal-timesimulationof cou-
plingbetweenground-water flow(porepressure)and
soil deformations is adoptedinthis numerical analy-
sisconsideringdewatering. Inthispaper, considering
dewateringmeansconsideringthedifferencebetween
inandout of thefoundationpit inthesimulating, and
notconsideringdewateringmeansnotconsideringthe
difference.
This paper describes the application of FLAC3D
programfor predictingsoil deformationsandground
water flowassociatedwiththetop-downconstruction
of acolumniformundergroundtransformersubstation
indowntownof Shanghai city.Themodel incorporates
anumber of advancedfeaturesof analysisincluding:
(1) A FLAC3Dmodel whichcanconsider interaction
betweensupport systemof thefoundationpit andthe
surroundingsoil;(2)Dewateringandexcavationareall
consideredinthenumerical analysisatthesametime;
(3) The drawdown of the ground water and the sur-
facesettlementduetodewateringinthefoundationpit;
(4) Thesurfacesettlement inducedby theleakageof
thediaphragmwall. Sitestratigraphy, material proper-
ties, andinitial ground-waterconditionareall selected
usinginformationprovidedprior toconstruction, the
simulation of theconstruction sequenceis based on
thescheme.
147
82.3m
77.4m
60.3m
45.3m
73.4m
37.5m
31.6m
26.5m
21.3m
16.5m
10.5m
3.2m
2.0m
100.0m
11
1
1
2.0m
7.0m
Roof slab
Temporary bracing
Floor slab
Floor slab
Floor slab
Temporary bracing
Temporary bracing
11.5m
16.5m
22.0m
26.5m
31.0m
33.7m Floor slab
1
7
1
2
3
4
5
12
5
6
7
1
8
2
8
3
8
1
9
2
9
2
7 2
7
Figure 1. Soil profile and the location of the wall and
bracing.
2 GENERAL SITUATIONOF THE PROJ ECT
2.1 Project description
Thecolumniformundergroundtransformersubstation
of afour floor and33.7mhighundergroundstructure
occupies aplanareaof 13000squaremeter (interior
diameter is130m) inthedowntownof Shanghai city
and is bounded by buildings and viaduct and lots of
undergroundpipeline.
The underground transformer substation design
uses a cast in situ, reinforced concrete, diaphragm
wall (1.2mthick) extendingdown into theelevation
57.5m (With respect to the Shanghai City Base
datum), the circular wall is braced internally by the
floorslabsand3temporaryannularbracing(Figure1),
which arein turn supported by theinterior columns
(steel and reinforced concrete & angel iron lattice)
foundedonthebearingpile(boredfillingpile, thepile
tipat depths8090mbelowgroundlevel. Both
thediaphragmwall andinterior columnsareinstalled
prior to excavationusingslurry trenchmethods. The
roof andfourfloorlevelsarecastinsequencefromthe
top-downbyexcavatingthesoil frombeneaththemost
recentlyconstructedslab. Duringexcavation, dewater-
ingisaccomplishedusingdrainagebydesiccation.
2.2 Engineering geological and groundwater
conditions
Table 1 shows an averaged profile of subsurface
stratigraphyinterpretedfromboringsconductedatthe
site. It shouldbenotedthat theborings logs actually
Table1. Input parametersinMohr-coulombmodel.
E j C D
Stratum MPa kPa (
) (
)
1.2 0.36 0.32 22.5 0
9.3 0.36 25.9 17.4 0
8.9 0.37 5.1 21.2 0
8.2 0.39 8.0 19.7 0
1
11.4 0.38 13.4 15.7 0
2
14.9 0.37 27.1 15.8 0
26.8 0.34 42.7 13.7 0
1
44.9 0.32 5.0 31.2 0
2
80.2 0.31 0.0 33.0 0
1
16.4 0.34 21.2 23.1 0
2
28.3 0.33 16.9 24.1 0
3
72.2 0.33 22.1 20.3 0
1
99.2 0.31 0.0 35.0 5
2
133.4 0.29 0.0 37.0 7
Table2. Porosityandcoefficient of permeability.
Coefficient of
permeability
Vertical Horizontal
Stratum Soil name Porosity m/s m/s
Artifical soil 0.56
Siltyclay 0.49 2.510
9
5.510
7
Muckysilt 0.57 1.710
8
3.510
6
Muckysilt 0.58 7.210
9
8.110
8
11
Clay 0.52 3.910
9
4.110
8
12
8
3.210
8
1
9
4.110
8
1
Siltysilt 0.46 2.610
6
2.610
5
2
Silt sand 0.44 5.410
6
3.810
5
1
Silt clay 0.51 8.210
9
4.010
6
2
8
2.810
6
3
6
3.010
4
1
Mediumsand 0.37 5.710
6
3.010
4
2
Coarsesand 0.35 7.910
6
3.010
4
showsignificantvariationsinthethicknessof theindi-
vidual strata across the site, The assumption of an
averageprofileisconsistent withthelimitedareaand
uncertainties in engineering properties of individual
strata.Thematerial parameterof thelayersisintable1.
Thereis unconfined aquifer and confined aquifer
inthesite. Thegroundwater level of theunconfined
aquifer is 12munder thegroundandthe6thlayer
is relative impervious layer. The confined aquifer is
dividedintothe1st confinedaquifer andthe2ndcon-
fined aquifer by the
1

2
layer. The 1st confined
aquiferliesinthe
1
and
2
layerandthe2ndconfined
aquifer lieinthe
3
andlayer, theremay besome
relationshipbetweenthe1stand2ndconfinedaquifer.
Theporosityandthecoefficientof permeabilityof the
layersisintable2.
148
2.3 FLAC3D model description
The three dimensional numerical analysis program
FLAC3D is developed by Itasca Consulting Group,
Inc. Thegroundwater flowmodel may becoupledto
the stress model. The finite element model extends
far beneaththeexcavation(to100mdepth) andlater-
ally a distance of 200mbeyond the perimeter wall
where soil displacements, due to the simulation of
undergroundtransformer substationconstruction, are
negligible. Constitutivemodelingof soil behavior and
selectionof inputparametersrepresentamajor source
of uncertaintyinfiniteelementanalysis.Thesoil con-
stitutive model use Mohr-coulomb failure criterion
andinorder tomodel realisticallythedepthvariations
in properties theelastic shear and bulk modulus are
assumedtobeproportional tothemeaneffectivecon-
finingstress. Theunloadingmodulusof thesoil inthe
foundationpitistrinal loadingsandthemodulusof the
soil undertheultimatebaseof excavationadoptmixed
modulus because of tension pile. The parameters of
thesoil areshownintable1.
The soil and diaphragmwall adopt 8 nodes solid
element and the permanent floor slab adopted shell
element and the temporary annular bracing adopted
beamelement, and interface is adopted in the joint
of the diaphragm and soil. Elastic model is used
for the diaphragmwall and the youngs modulus is
2.310
4
MPa, andthePoissonsratiois0.167. Elastic
model isalsousedforfloorslabandtemporaryannular
bracingandtheparameter of themis consistent with
thescheme. TheFLAC3D model of this project is in
Figure2.
Boundariesconditionissummarizedasfollows: (1)
Theundersidedisplacement of themodel iszero, the
horizontal displacement of the side of the model is
zero, theuppersurfaceisfree; (2)Theundersideof the
model isimperviousboundary, thesideof themodel is
perviousboundaryandtheporepressureisfixed, the
porepressureof theupper surfaceis fixedas zeroin
theformationof theinitial stressfieldandhydrostatic
pressureandisfreeduringdewateringandexcavation.
Dewateringinthefoundationpit issimulatedbycon-
trolling the saturation and pore pressure at specific
locationsintheelement model.
2.4 Construction sequence
Based on the actual record of site activities and the
sequenceof events occurs in thefiniteelement sim-
ulation, this process is simulated by 17 stages. Each
stage in the analysis represents a distinct change
in either the geometry, boundary conditions or time
elapsedbetweenevents.Thefirststageistheformation
of the initial stress field and hydrostatic pressure
andafter every dewateringandexcavation(including
addingbracing) is astage. Thenumerical simulation
assumes that theconstruction of thediaphragmwall
Figure2. FLAC3Dmodel of theproject.
has no effect on the surrounding soil (i.e., the wall
iswished-inplace)anddoesnot consider theinstal-
lation of load-bearing elements used to support the
internal column.
3 RESULT ANDANALYSIS
3.1 Analysis of the seepage field
Duetounconfinedaquifer andconfinedaquifer exist-
inginthesite, thereisdifficultyinsimulatingthemat
thesametime. Theconfinedaquifer isnot takeninto
149
Figure3. Theneural pressurecontour andflowvector.
Figure4(a). Lateral deflectionof wall.
Figure4(b). Contrast of thewall deflection.
accountwhenanalyzingtheseepageof theunconfined
aquifer, whichhavenoinfluenceonthephreaticline.
Theneutral pressurecontourandtheflowvectorwhen
excavatingto33.7misasFigure3. Itshowsthatthe
drawdownof thephreaticwaterissmall andthedepres-
sionconeis not obvious dueto thelowpermeability
of thesoft clay.
3.2 Analysis of the lateral deflection of wall
Figure4(a)isthecomputingresultof thelateral deflec-
tions of thediaphragmwall alongthedepthat every
Figure4(c). Contour of thelateral deflection.
excavationstage. Themaximumlateral deflectionof
thewall is23.3mm, theratioof themaximumlateral
deflection to theend excavation depth is 0.07%, the
top-downandthehighrigidityof theslabisthereason
forthesmall ratio.Thelocationof thelateral deflection
is alittlehigher thantheexcavationfaceandfalling
withtheexcavatingandthelocationof themaximum
lateral deflectionisat28.1m. Figure4(b) isthecon-
trast of thelateral deflection considering dewatering
andnot inthelast 3excavationstage. Themaximum
lateral deflection considering dewatering is 27.7mm
andhigher thantheresultnotconsideringdewatering.
Thedifferencemostlyhappeninthemiddleandunder-
sidewall andis increasingwiththeexcavationdepth
fortheseepageforceisconcentratedinthemiddleand
underside wall and the hydraulic head is increasing
withtheexcavationdepth. Figure4(c) is thecontour
of lateral displacement of thediaphragmwall.
3.3 Analysis of the basal heave formation
The basal heave is including the elastic rebounding
andthelocal plasticfailureandthedeep-seatedplastic
failure. Theelastic reboundisbecauseof theunload-
ing, the local plastic failure is that the soil near the
wall yield, and thedeep-seated plastic failureis that
thesoil inthebottomisshort of bearingpower. When
the excavation depth is little, the elastic rebound is
themost, themaximumheaveliesinthecenter, with
theexcavation depth increasing theheavein thecir-
cumferencepreponderateover theheaveinthecenter
becauseof thefailureinthecircumference, Figure5(a)
isthecurveof thebasal heave, Figure5(b) isthecon-
tour of thebasal heavewhenexcavatingto 33.7m.
Figure5(c)isthecontrastof thebasal heaveresultcon-
sideringdewateringandnot inthelast twoexcavation
stages, it showsthebasal heaveconsideringdewater-
ingishighthanthenotandthedifferenceisincreasing
withtheexcavationdepth.
3.4 Analysis of the surface settlement
Figure 6(a) shows the surface settlement after the
every excavation stage. The surface settlement and
150
Figure5(a). Thedeformationof thebasal heave.
Figure5(b). Contour of thebasal heave.
Figure5(c). Contrastof thedeformationof thebasal heave.
thedistancebetweenthemaximumsurfacesettlement
andthefoundationpit is increasingwiththeexcava-
tiondepth. Themaximumsettlement is 15.7mmand
the maximumdistance is 44mwhen excavating to
33.7mdepth.Thediaphragmwall moveupandraise
thesoil near thefoundationpitbecauseof theunload-
ing, whichhasaeffectonthelocationof themaximum
surfacesettlement .
Figure6(a). Surfacesettlement at stages.
Figure6(b). Contour of thesurfacesettlement.
Figure6(c). Contrast of thesurfacesettlement.
Figure6(b) isthecontour of thesurfacesettlement.
Figure6(c)showsthecontrastof thesurfacesettlement
consideringdewateringandnot. Itshowsthatthemag-
nitudeandrangeof thesurfacesettlement inducedby
thedewateringissmall forthereasonof thelowperme-
abilityof thesoil andthesmall descentof groundwater
level induced by dewatering. The location of max-
imum surface settlement induced by dewatering is
closer tothefoundationpitthanbyexcavationandthe
151
Figure7. Leakageinthe6thlayer.
maximumsurfacesettlement consideringdewatering
is15.9mm.
3.5 Analysis of leakage
By the excavation experience in Shanghai city, the
leakageof diaphragmwall oftenoccur whenthedepth
beyond 28m and the probability increase with the
excavationdepth. Themaincauseisthat thebadjoint
of the wall and the dimension error in construction
andthedistortionof thediaphragmwall. Theleakage
will inducethedeclineof thegroundwater level and
theadditional surfacesettlement, so theeffect of the
leakagehavetobetakenintoaccount.
If the leakage occurs in the sixth layer, Figure. 7
showstheneutral pressureandtheflowvelocity vec-
tor, by the contour of the neutral pressure, only the
groundwater level near thefoundationpit descendlit-
tle, sotheleakageinthe6thlayer havelittleeffect on
thesettlementaroundthefoundationpitbecauseof the
lowpermeabilityandsmall discharge.
If the leakage occurs in the 1st confined aquifer
which of the water pressure is 56m under the
ground, thewater pressureof the1stconfinedaquifer
descendclearly andthesurfacesettlement is distinct
for thepermeabilityof the1stconfinedaquifer ishigh
andthedischargeismuch.
The2ndconfinedaquifer canleaktointerior of the
1st confined aquifer for the fall of the neutral pres-
sureinthe1stconfinedaquifer duetodewateringand
excavatingin
1
layer. Thenumerical analysisshows
leakageoccurs, but thedischargeislittle, sotheleak-
agehavelittleeffectonthesettlementforpermeability
of theof the
1
grayclayislow(8.2110
9
m/s) and
thethicknessof
1
layer reachto15m.
4 CONCLUSIONS
This paper introduce the FLAC3D model for simu-
lating the top-down construction of an underground
transformer substationat Shanghai city. Results con-
sideringdewateringarecomparedwiththeresult not
considering including wall deflection, basal heave,
and surface settlement, the effect of leakage is also
analyzed. Themain conclusions of this study areas
follows:
1. Dewatering by desiccation in the foundation pit
haveeffectonthebehaviorof thefoundationpit,the
seepagecouldenhancethewall deflectionandthe
deformation of thebasal heave, thedrawdown of
thegroundwater level outsideof thefoundationpit
couldresultinthevertical consolidationandaddthe
surfacesettlement.Thefoundationpitisinclinedto
benot securityif dewateringisnot considered.
2. Thewall deflectionandbasal heaveduetotheseep-
ageislittlebecauseof thelowpermeabilityof the
soil insoftclay. Thedrawdownof thegroundwater
level duetodewateringislittle, whichresultinlittle
surfacesettlement for thesamereasonabove.
3. Theleakageof thewall haveimportant effect on
thesurfacesettlement, whichisaproblemneeded
tosolve.
REFERENCES
Andrew J. Whittle, Youssef M. A. Hashash & Robert V.
Whitman. 1993. Analysis of deep excavation in boston,
Journal of Geotechnical Engineering, 119(1): 6990.
Chang-Yu, Ou, Tzong-Shiann Wu & Hsii-Sheng Hsieh,
1996. Analysis of deep excavation with column tye of
groundimprovementinsoftclay. Journal of Geotechnical
Itasca Consulting Group, Inc. 2002.6. FLAC3D (Fast
LagrangianAnalysis of Continua in 3DDimensions) User
Manuals, Version2.1. Minneapolis, Minnesota.
ItascaConsultingGroup, Inc.2005.5. FLAC(Fast Lagrangian
Analysis of Continua )User Manuals, Version 5.0. Min-
neapolis, Minnesota.
J acob Bear, 1983. Dynamics of Fluids in Porous Media.
Beijing: ChinaArchitecture& BuildingPress.
Lin Li, 2007. Studies on the behavior of deep excavation
and surroundings due to dewatering effect, Ph.D, Thesis,
Universityof Tongji, Shanghai, China.
LinLi & MinYang, 2007. Theanalysisof deformationchar-
acteristicsof thedeepexcavationinsoft clay. China Civil
Engineering Journal, 40(4): 6672.
Sunil S.Kishnani &RonaldoI.Borja,1993.Seepageandsoil-
structureinteractioneffectsinbracedexcavtion. Journal
of Geotechnical Engineering, 119(5): 912927.
Youssef M.A. Hashash & AndrewJ Whittle, 1996. Ground
movement prediction for deep excavations in soft clay.
Journal of Geotechnical Engineering 122(6): 474486.
Yuqi Li, 2005. Studies on the behavior of foundation pit with
excavation considering seepage, Ph.D,Thesis,University
of Zhejiang, Hangzhou, China.
152
Analysisof thefactorsinfluencingfoundationpit deformations
Y.Q. Li
Department of Civil Engineering, Shanghai University, Shanghai, P.R. China
K.H. Xie
Institute of Geotechnical Engineering, Zhejiang University, Hangzhou, P.R. China
J. Zhou& X.L. Kong
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education,
Tongji University, Shanghai, P.R. China
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R. China
ABSTRACT: Due to the complexity of excavation and groundwater seepage, the behavior of foundation
pitsisnot yet well understood. Inthispaper, basedonthree-dimensional (3D) Biotsconsolidationtheory and
nonlinearDuncan-Changsmodel, finiteelementequationsconsideringthecouplingof groundwaterseepageand
soil skeletondeformationduringexcavationarededucedandacorrespondingthree-dimensional finiteelement
programis developed. Using the program, the influence of soil permeability, rigidity and tiers of supports,
rigidity of retainingwall andconstructionperiodof excavationongroundsurfacesettlement, wall horizontal
displacement and pit baseheaveareanalyzed in detail. Someuseful conclusions aredrawn by analyzing the
influenceof thesefactors on theexcavation deformations, which arevery significant for guiding design and
constructionof excavations.
1 INTRODUCTION
Inurbanareas, moreandmoreundergroundspaceis
utilized with the fast development of city construc-
tion, andthusalotof excavationengineeringappears.
However, the pit deformations induced by excava-
tion greatly influence the safety of not only the pit
itself but also the buildings and municipal facilities
around it. Therefore, study of the behavior of foun-
dationpitshasreceivedmuchattention. Whittleet al.
(1993) described the application of a finite element
analysis for modelling thetop-down construction of
a seven-storey, underground parking garage at Post
OfficeSquareinBoston.Theresultsdemonstratedthat
reliableandconsistentpredictionsof soil deformations
and groundwater flow can beachieved by advanced
methods of analysis without recourse to parametric
iteration, but emphasizedtheneedfor adequatechar-
acterizationof engineeringpropertiesfortheentiresoil
profile.Vaziri (1996) describedasimple, efficientand
practical numerical model for analysisof cantilevered
and strutted flexible retaining walls. The model had
incorporated a variety of features that affected the
performance of the retaining walls in the field such
as installation and removal of struts, application of
surcharge, changes in groundwater table, changes in
soil properties andsimulationof stagedexcavations.
Themodel canbeusedeffectivelytoperformabroad
suite of parametric studies in the design stage and
also as a reliable tool for predicting performance.
Ou et al. (1996) further proposed a nonlinear, 3D
finite element technique for deep excavation analy-
sis. Thetechniqueaswell astheanalytical procedures
formodelingtheexcavationprocesseswerecodedinto
acomputer program, andtheaccuracyof theprogram
wasassessed. Thecaseof anirregularly-shapedexca-
vationwithfieldmeasurementsof wall deflectionwas
studiedandtheresults showedcloseagreement with
field measurements. Zdravkovic et al. (2005) stud-
iedtheeffect of excavationonthesurroundingareas
andprovidedadetailedassessmentof wall andground
movements.
Therehavebeen afew studies on theinfluencing
factors of foundationpit deformations. Inthis paper,
3Dconsolidationfiniteelementequationsarederived,
andthecorrespondingfiniteelementprogramisdevel-
oped. Someuseful conclusionsaredrawnbyanalyzing
the influence of factors such as soil permeability,
rigidity and tiers of supports, rigidity of retaining
wall andconstructionperiodof excavationonthepit
153
deformations, whicharebeneficial tooptimisationof
excavationdesign.
2 FINITE ELEMENT EQUATIONS
BasedonBiots3Dconsolidationfiniteelementequa-
tions(Xie&Zhou2002),andconsideringgroundwater
seepageinducedbythewaterheaddifferencebetween
the inside and outside of a pit, the finite element
equationsof excavationareasfollows:
where is an integral constant; Lt is the time
increment; [K
eij
] and[K
cij
] arerespectively thesub-
matrices of the stiffness matrix and the coupling
matrix; K
sij
is an element of seepage matrix; Lu
i
,
Lv
i
andLw
i
arethedisplacement incrementsof ele-
ment node i; P
i(n+1)
is the soil water potential of
element nodei at t =t
n+1
; LR
xi
=LR
xi
+[K
cij
]P
i(n)
,
LR
yi
=LR
yi
+[K
cij
]P
i(n)
, LR
zi
=LR
zi
+[K
cij
]P
i(n)
,
andLR
pi
=LR
pi
LtK
sij
P
i(n)
, LR
xi
, LR
yi
andLR
zi
are the equivalent load increments of element node
i, and LR
pi
is theequivalent water runoff increment
of element nodei, P
i(n)
is thesoil water potential of
element nodei at t =t
n
.
The soil water potential of a saturated soil can
beexpressed using thefollowing equation when the
solutepotential of thesoil isneglected:
wherethespatial coordinatez isupwardspositive; P is
soil waterpotential of saturatedsoil; p isthesumof the
pressurepotential andtheloadpotential, i.e. thetotal
porewater pressure; and
w
z isthegravitypotential.
3 ANALYSISOF THE INFLUENCING
FACTORSOF PIT DEFORMATIONS
Inorder toanalyzetheparametricinfluenceonthepit
deformations, a3D consolidationfiniteelement pro-
gramis developedon thebasis of thefiniteelement
equations derived. Usinganumerical examplegiven
below, themain factors influencing thepit deforma-
tions such as soil permeability, rigidity and tiers of
supports, rigidity of retaining wall and construction
periodof excavationareanalyzedrespectively.
Figure1. Meshof finiteelements.
Table 1. Duncan-Chang model parameters
of soil.
Parameters Values
K 150
n 0.7
R
f
0.85
c
15kPa
35
F 0.15
G 0.35
D 3.5
K
ur
300
3.1 Reference case numerical example
Theexcavatedlength, widthanddepthof thefounda-
tionpitinacertainhomogenousandisotropicstratum
of soft soil are60m, 50mand8mrespectively. The
soilsvertical andhorizontal permeabilitycoefficients
areboth2.010
6
cm/sandtheeffectiveunit weight
of the soil is 9.0kN/m
3
. The retaining wall is 0.6m
thickandembedded16mdeepinsoftsoil. Reinforced
concretesupportsareinstalledat different excavation
stages and the horizontal spacing between supports
along the pits long side (i.e. y-direction) and short
side (i.e. x-direction) is 6mand 5mrespectively in
everytier.
In order to minimize the boundary effects and
improvethecomputational efficiency, thecalculation
domainsinx-, y- andz-directionare100m, 100mand
40mrespectively in consideration of the symmetry
about the pit centerline. The finite element mesh of
thesoil massandretainingwall areshowninFigure1.
All soil units are discretized using eight-node
hexahedral isoparametric elements, modelled using
thenonlinear Duncan-Changmodel with parameters
listed in Table 1, where c
and
are the effective

cohesion and the effective friction angle of the soil
respectively, R
f
is the failure ratio, and K, n, F,
G, D and K
ur
are some parameters determined by
tests. Theretainingwall adoptsWilsonnon-harmony
154
elements, modelled as alinear elastic model, whose
modulus of elasticity and Poissons ratio are25GPa
and0.167respectively.A rowof 0.1mthickinterfaces,
connectingthesoil massandtheretainingwall isatthe
twosidesof retainingwall, adopting3Dthininterface
elements derivedfromYins rigidplastic model (Yin
et al. 1995) with the outer friction angle=1.0
and
cohesion=0.5kPa,anditsothermodel parametersare
thesameasthoseof thesoil masselements. Thesup-
ports are modelled using a linear elastic model and
spatial bar elements, with0.6m0.6mcrosssection,
whoseelasticitymodulusis23GPa.
Theexcavationinvolves threestages. Thedetailed
description of the staged excavation of the pit is as
follows:
1. Stage1: 2.0mexcavationdepthwithout supports
for four days, andfour days excavationintermis-
sionfor installingsupports at thenext excavation
stage.Thezvalueis1.5mforthefirsttierof sup-
portsand2.0mfor thecorrespondingexcavation
level belowthesupports.
2. Stage 2: 3.0m excavation depth (excavation to
5.0mdeep) with a tier of supports in six days,
andsixdays excavationintermissionfor installing
the next tier of supports. The z value is 4.5m
for thesecondtier of supportsand5.0mfor the
correspondingexcavationlevel belowthesupports.
3. Stage 3: 3.0mexcavation depth (full excavation
to 8.0mdeep) with two tiers of supports in eight
days, andtwentydays excavationintermissionfor
castingthepit baseconcrete.
3.2 Influencing factors
3.2.1 Soil permeability
In this section, theinfluenceof soil permeability on
the pit deformations at the y =0 section after the
thirdexcavationstageis studied. Thesoil permeabil-
ity for the reference case is 2.010
6
cm/s. Four
more analyses are carried out for soil permeability
of 2.010
5
cm/s, 5.010
6
cm/s, 5.010
7
cm/s
and2.010
7
cm/s. Figures 24showtheinfluence
of soil permeabilityrepresentedbypermeabilitycoef-
ficient k onthewall horizontal displacement, ground
settlementandpitbaseheave.Withthesoil permeabil-
ity increasing, the vertical effective stresses outside
thefoundationpit alsoincrease, but thosebeneaththe
pit basedecrease, so ground settlement and pit base
heaveincrease, whichareshowninFigures 34. For
the wall horizontal displacement, with the soil per-
meability increasing, thehorizontal effectivestresses
insideandoutsidethepit bothincrease, andthewall
horizontal displacement decreases as a result of the
greaterinfluenceof lateral pressuresactingonthewall
insidethepit, whichcanbeseeninFigure2.
-16
-12
-8
-4
0
5 6 7 8 9
Wall horizontal displacement/cm
D
e
p
t
h
/
m
k=2e-7cm/s
k=5e-7cm/s
k=2e-6cm/s
k=5e-6cm/s
k=2e-5cm/s
Figure 2. Influence of soil permeability on wall
displacement.
-6
-5
-4
-3
-2
-1
0
0 20 40 60 80
Distance from the wall/m
G
r
o
u
n
d

s
e
t
t
l
e
m
e
n
t
/
c
m
k=2e-7cm/s
k=5e-7cm/s
k=2e-6cm/s
k=5e-6cm/s
k=2e-5cm/s
Figure 3. Influence of soil permeability on ground
settlement.
0
3
6
9
12
0 5 10 15 20 25
Distance from the pit center/m
P
i
t

b
a
s
e

h
e
a
v
e
/
c
m
k=2e-7cm/s k=5e-7cm/s
k=2e-6cm/s k=5e-6cm/s
k=2e-5cm/s
Figure4. Influenceof soil permeabilityonpit baseheave.
3.2.2 Rigidity of supports
The influence of support rigidity on wall horizontal
displacement, ground settlement and pit base heave
at the y =0 section after the third excavation stage
are shown in Figures 57. When the support rigid-
ity becomes larger, the retaining wall movement is
more restricted, so the wall horizontal displacement
is smaller. However, theinfluenceof support rigidity
ongroundsettlement andpit baseheaveis relatively
insignificant.
3.2.3 Tiers of supports
The influence of support tiers on the pit deforma-
tions at the y =0 section after the third excavation
155
-16
-12
-8
-4
0
5 6 7 8 9 10
wall horizontal displacement/cm
D
e
p
t
h
/
m
5GPa
10GPa
23GPa
50GPa
100GPa
Figure 5. Influence of support rigidity on wall
displacement.
-6
-4
-2
0
0 20 40 60 80
G
r
o
u
n
d

s
e
t
t
l
e
m
e
n
t
/
c
m
5GPa 10GPa
23GPa 50GPa
100GPa
Figure 6. Influence of support rigidity on ground
settlement.
0
3
6
9
12
0 5 10 15 20 25
P
i
t

b
a
s
e

h
e
a
v
e
/
c
m
5GPa 10GPa
23GPa 50GPa
100GPa
Figure7. Influenceof support rigidityonpit baseheave.
stageis studied. Thereferencecasehas two tiers of
support. Threemoreanalysesarecarriedout: nosup-
port, onetier at 2.0mexcavationdepthandonetier at
5.0mexcavationdepth. Figures810showtheinflu-
enceof support tiersonwall horizontal displacement,
groundsettlementandpitbaseheaverespectively.The
deformationsof thefoundationpit duringexcavation
withnosupport arethelargest, andtheyevidentlyare
smallerwithaddingsupporttiers.Tiersof supportalso
influencethepitdeformations, whichfortwo-tiersup-
ports are less than those with one-tier. In addition,
-16
-12
-8
-4
0
0 10 20 30 40
D
e
p
t
h
/
m
two tiers
one tier at
upper depth
one tier at
lower depth
no support
Figure8. Influenceof support tiersonwall displacement.
-15
-12
-9
-6
-3
0
0 20 40 60 80
G
r
o
u
n
d

s
e
t
t
l
e
m
e
n
t
/
c
m
two tiers
one tier at upper depth
one tier at lower depth
no support
Figure9. Influenceof support tiersongroundsettlement.
0
5
10
15
20
0 5 10 15 20 25
P
i
t

b
a
s
e

h
e
a
v
e
/
c
m
two tiers
one tier at upper depth
one tier at lower depth
no support
Figure10. Influenceof support tiersonpit baseheave.
the position of the supports also greatly influences
thepit deformations. Thedeformationswithsupports
installed at a higher level are less than those with
supportsinstalledatalower level under thesamecon-
ditions, so theformer approachis moreeffectivefor
controllingthepit deformations.
3.2.4 Rigidity of retaining wall
Figures 1113 show the influence of rigidity of the
retainingwall onwall horizontal displacement,ground
settlementandpitbaseheaveatthey =0sectionafter
the third excavation stage. The wall horizontal dis-
placement will obviously decrease with an increase
in the rigidity of the retaining wall. However, the
156
-16
-12
-8
-4
0
0 3 6 9 12
D
e
p
t
h
/
m
5GPa
10GPa
25GPa
50GPa
100GPa
Figure 11. Influence of rigidity of retaining wall on wall
displacement.
-6
-5
-4
-3
-2
-1
0
0 20 40 60 80
Distance from thewall/m
G
r
o
u
n
d

s
e
t
t
l
e
m
e
n
t
/
c
m
5GPa 10GPa
25GPa 50GPa
100GPa
Figure12. Influenceof rigidityof retainingwall onground
settlement.
0
2
4
6
8
10
12
0 5 10 15 20 25
Distance fromthe pit center/m
P
i
t

b
a
s
e

h
e
a
v
e
/
c
m
5GPa 10GPa 25GPa
50GPa 100GPa
Figure13. Influenceof rigidityof retainingwall onpitbase
heave.
influenceof rigidityof retainingwall ongroundsettle-
ment andpit baseheaveisnot significant. Therefore,
increasingtherigidityof theretainingwall caneffec-
tively reducethewall horizontal displacement andis
beneficial tothesafetyof excavations.
3.2.5 Construction period
Theconstructionperiodincludestheexcavationperiod
andintermissionsatall excavationstages,whichis48d
inthereferencecase. Four moreanalyseswerecarried
outforconstructionperiodof 24d, 36d, 60dand72d.
-16
-12
-8
-4
0
0 2 4 6 8 10
D
e
p
t
h
/
m
t=24d
t=36d
t=48d
t=60d
t=72d
Figure14. Influenceof construction period of excavation
onwall displacement.
-5
-4
-3
-2
-1
0
0 20 40 60 80
G
r
o
u
n
d
s
e
t
t
l
e
m
e
n
t
/
c
m
t=24d t=36d t=48d
t=60d t=72d
ongroundsettlement.
4
6
8
10
0 5 10 15 20 25
P
i
t

b
a
s
e

h
e
a
v
e
/
c
m
t=24d t=36d
t=48d t=60d
t=72d
onpit baseheave.
Theinfluenceof constructionperiodonwall horizon-
tal displacement, groundsettlementandpitbaseheave
at they =0sectionafter thethirdexcavationstageis
shown in Figures 1416. On the one hand, with the
constructionperiodincreasing, theexcessporewater
pressureshavealonger timetodissipate, andthesoil
strata can achieve a higher degree of consolidation,
gaining higher strength and stiffness, thus the wall
horizontal displacement decreasestoacertainextent.
On theother hand, thepit baseheaveincreases with
an increase in construction period. The influence of
constructionperiodongroundsettlement isrelatively
insignificant.
157
4 CONCLUSION
Based on Biots consolidation theory, finiteelement
equationswerededucedandacomputer programwas
developed.Theinfluenceof thekeyparameterssuchas
soil permeability,rigidityandtiersof supports,rigidity
of theretainingwall, andtheconstruction periodon
pit deformations is studied using the finite element
program. The study and the results reported in this
paper arehelpful toguideexcavationengineering.
ACKNOWLEDGEMENTS
ThisresearchprojectwassupportedbytheChinaPost-
doctoral ScienceFoundation(No. 20060400672) and
InnovationFundof Shanghai University, China.
REFERENCES
Ou, C.Y., Chiou, D.C. & Wu, T.S. 1996. Three-dimensional
finite element analysis of deep excavation. Journal of
Geotechnical Engineering, ASCE 122(5):337345.
Vaziri, H.H. 1996. A simplenumerical model for analysisof
proppedembeddedretainingwalls. International Journal
of Solids Structures 33(16):23572376.
Whittle,A.J.,Hashash,Y.M.A.&Whitman,R.V.1993.Analy-
sisof deepexcavationinBoston. Journal of Geotechnical
Engineering, ASCE 119(1):6990.
Xie, K.H. & Zhou, J. 2002. Theory and Application of Finite
Element Analysis in Geotechnical Engineering. Science
Press, Beijing.
Yin, Z.Z., Zhu, H. &Xu, G.H. 1995. A studyof deformation
intheinterfacebetweensoil andconcrete. Computers and
Geotechnics 17:7592.
Zdravkovic, L., Potts, D.M. & St J ohn, H.D. 2005. Mod-
elling of a 3D excavation in finite element analysis.
Geotechnique 55(7):497513.
158
Constructionmonitoringandnumerical simulationof anexcavation
withSMWretainingstructure
Z.H. Li & H.W. Huang
ABSTRACT: The soil mixing wall (SMW) retaining structure is applied in two long strip excavations in
Shanghai. Firstly, thebearinganddeformingmechanismof SMWisanalyzedinbrief. Thestructural analysis
methodof SMWisdiscussed. Secondly, basedonthein-situexcavatingconstructionprocedures, theconstruction
stepsof excavatingandsupportingaresimulatedinthenumerical calculationwiththemethodof FastLagrangian
Analysisof Continua3D. Therearetwocasessimulatedinnumerical calculation, case1isthenormal casein
whichthesupportsareinstalledtimely, andcase2isacaseinwhichthesupportsarenotinstalledtimelybecause
of somereasons.Then, thedeformationof theretainingstructure, thehorizontal displacementatthetopof SMW
andtheaxial forces of steel pipesupports areanalyzedandcomparedwiththeactual observationdataintwo
cases. A goodagreement canbefoundbetweenthecalculationresultsandobservationdata. It canbeseenthat
incase1theexcavationisstableandsafe; theaxial forcesarelower thanthealarmvaluesandthedisplacement
duetoexcavatingisinthepermissiblerange. Incase2, however, theexcavationisindanger of instability and
somemeasuresshouldbetakentoprotect theSMWretainingstructurefromfailure.
1 INTRODUCTION
The composite structure with H-shaped steel and
deepcemented-soil pilesiscalledSMWmethod. This
methodcanbeappliedincohesivesoil, sandysoil and
sandy gravel layers. It has been widely accepted in
China, whichismainlyappliedtodeepexcavationsin
soft soilsof easternandsouthernChina.
Thispaperstudiesthebearinganddeformingmech-
anismof SMW. Basedonthein-situexcavationcon-
structionprocedures of theengineeringexample, the
construction steps of excavating and supporting are
simulated in the numerical simulation method with
FLAC3D. Becausetheexcavations aretoo long, the
methodof excavatingissimilar totunnel excavating,
whichis fromonesidetoanother. Andtherearetwo
cases simulated, oneis thenormal caseinwhichthe
supports areinstalled timely. However, in middleof
March 2007 during excavating, because of the bad
weatherandsomeotherreasons, thesupportswerenot
installedtimely so that theexcavationwas indanger
of collapse. Therefore, another caseisacaseinwhich
thesupportsarenot installedtimely.
Through comparing the actual observation data
in-situ and the calculation results, some useful con-
clusionsareachieved.
2 THE PROJ ECT GENERAL SITUATION
ANDMONITORINGSCHEME
2.1 The project general situation
Theundergroundchannels project contains two long
stripexcavations, theeasternandwesternexcavations,
which are similar to each other. The length is both
820.5m, andthewidthis16.5m.Theexcavationdepth
is between 6.645m8.039m. It is clear around the
constructionsite.
The foundation soil layers belong to Quater-
nary Pleistocene-Holocene deposit, including cohe-
sive soils, silt and sandy soils which distribute in
planes. The physico-mechanical parameters of soils
insiteareshownonTable1.
The650 SMW method is applied and steel pipe
supports are installed. The SMW retaining struc-
ture is 17m long in depth, which is inserted with
5002001016H-shapedsteels.Inordertofacil-
itatetheearthwork excavating, twosteel pipesupport
tiers including theupper and thelower supports are
installedconsideringthecharacteristic of theexcava-
tions. Theupper steel supportslocate1.0mbelowthe
groundsurface. Thelower steel supportslocate2.5m
above the excavation bottom. The sizes of the steel
159
Table1. Physico-mechanical parametersof soilsinsite.
Thesoil Depth C E
layers (m) (kN/m
3
) (KPa) (
) (MPa) j
Siltyclay 3 17.8 10.0 23.0 4.38 0.3
andclayey
silt
Sandysilt 2 18.2 4.0 29.0 4.78 0.33
Siltyclay 2 17.3 10.0 21.5 3.84 0.35
Sandysilt 2 17.6 3.0 25.5 5.08 0.3
andsilty
clay
Siltyclay 10 16.2 11.0 11.0 13.62 0.3
Clay 11 17.3 13.0 12.0 17.0 0.3
Figure1. General layoutof monitoringpoints: (a)Thesouth
section of the excavations; (b) The strain gauges welded
aroundthesupport.
supportsareall 60912. Thedistancebetweenthe
adjacent supportsis5m.
2.2 The monitoring scheme
Aroundthetwoexcavations, themonitoringpointsfor
thehorizontal displacement andsettlement at thetop
of theretainingstructurearelocatedaboutevery10m.
And they are numbered using E and W, in which E
denotes the eastern excavation, W denotes the west
excavation. There are all 324 monitoring points for
thetophorizontal displacement andsettlement of the
SMW in two excavations, as shown in Figure 1(a).
Because the excavations are so long that the south
sectionisgivenonly.
14 inclinometer tubes for lateral deformation of
SMW are set in every excavation. They are located
in the same distance about 100m symmetrically.
The location of every monitoring point is shown in
Figure1(a), in which ECX denotes theinclinometer
tubeintheeasternexcavationandWCX denotes the
inclinometer tubeinthewesternone.
In every excavation there are 6 pairs of monitor-
ing points for axial forces of supports, and every
pair hastwopointsincludingtheupper andthelower
supports.AsshowninFigure1(a), for example, EZC1
includestheupperEZC1andthelowerEZC1supports.
Therefore, there are 12 monitoring points for every
excavation, EZC denotes the axial force monitoring
pointsintheeasternexcavation, andWZCdenotesthe
western one. The distance between each pair moni-
toringpoints is about 120m. Thesteel straingauges
areweldedaroundthesteel pipesupportsasshownin
Figure1(b).
3 BEARINGANDDEFORMING
MECHANISM OF SMW
The cemented soil material that is produced gener-
ally has a higher strength, lower permeability, and
lower compressibility thanthenativesoil. Therefore,
theSMWmethodcanmakeit possibletoformwater-
preventingandearth-retainingwallsquicklybymixing
earth collected at a construction site with cement
slurry.Therigidityof theearthretainingwallswasfur-
ther enhancedbyformingacompoundearth-retaining
wall withH-shapedsteel materialsweldedwithstuds
thatactasstressmaterial arrangedwithintheimproved
soil walls. And under the suitable conditions, the
H-shapedsteelscanberecycled.
Stress-straincharacteristicsof SMWareextremely
complex duringthecourseof thepit excavation. The
curves of H-shaped steel strain are under the linear
elasticscope, butcemented-soil isnonlinearresponse,
andtherigiditychangesof compositestructuremainly
by thecemented-soil. It iscommonly consideredthat
theH-shapedsteelsbear all thelateral water andearth
pressureandthecementdeepmixingpilesareusedto
prevent water. However, it istestifiedthroughexperi-
mentsthatcementsoil canenhancetheH-shapedsteels
to reduce the deformation. In addition, the cement
soil can also haveconfinement effect to prevent the
H-shaped steels instability. The composite flexural
stiffnessis20%greaterthanonlyH-shapedsteels.The
stiffnessenhancingcoefficient candenotethedegree
of stiffnessenhancingasfollows:
whereE
cs
and E
s
aretheelastic modulus of cement
deepmixingpilewithH-shapedsteel andtheelastic
modulusof H-shapedsteel, respectively; I
cs
andI
s
are
theinertiamoment of cement deepmixingpilewith
H-shaped steel and the inertia moment of H-shaped
steel.
In this numerical calculation, the cement deep
mixing pile with H-shaped steel is equivalent to
diaphragmwall andtheinfluenceof stiffnessenhanc-
ing coefficient is considered. According to the
principlethat thestiffness is equal to eachother, the
equationisgivenby
160
Figure 2. The stiffness equivalence between SMW and
diaphragmwall.
Figure3. Elementsof model incase1.
Basedonequation(1), thus
Here, is considered as 1.2. The equivalent thick-
ness of diaphragm wall in this numerical calcu-
lation is h =0.65m and the Young modulus is
E =12.6GPa.Theequivalent figurefromSMWwith
H-shaped steels to diaphragmwall in this project is
shownasFigure2shown.
The interfaces are installed to simulate the inter-
face characteristics between the retaining structure
and thesoils. In FLAC3D, Interfaces havetheprop-
ertiesof friction, cohesion, dilation, normal andshear
stiffness, andtensileandshear bondstrength, which
are characterized by Coulomb sliding and/or tensile
andshearbonding. Inthiscomputation, theequivalent
diaphragmwall isconsideredaselasticity.
4 CALCULATIONCASE
4.1 Case 1
Becausetherearetwolongstripexcavationsandthey
aresimilar, onepartof theeasternexcavationischosen
tobesimulated.Themodel sizeis60minextent, 60m
inbreadthand30minheight. Themodel isshownin
Figure3.
Theearthworkssoilsforexcavatingaredividedinto
3layers. Thefirst layer is from0.0mto 2.0m, the
second layer is from2.0mto 6.5m, the third is
from6.5mto 8.0m. There are upper and lower
twosupportsinstalled, theupper supportsarelocated
at 1.0mandthelower supportsareat 6.5m. And
thedistancebetweentwoadjacentsupportsinydirec-
tionis5m. Therefore, thelengthof soilsexcavatedin
every layer is 5min y direction in every excavating
step.Becausetheexcavationistoolong,theexcavating
methodissimilar totunnel excavatingmethodwhich
is fromonesideto another. Theconstruction proce-
dureof excavatingandsupportingisdividedintolots
of steps, asfollows:
Theconstructionof SMW.
Thefirst layer is excavated5miny directionand
thefirst upper steel support isinstalled.
Thefirst andsecondlayersareexcavated5miny
directionandthesecondupper andthefirst lower
supportsareinstalled.
All thethreelayersareexcavated5minydirection,
andthethirdupper andthesecondlower supports
areinstalled.
Do this until the earthworks excavation is
completed.
The whole procedure of excavating steps and
installing supports is simulated by 3D numerical
method. Thereare11excavatingandsupportingsteps
except theconstructionof SMW.
In this numerical simulation calculation, the
mechanical soil behavior is modeled with Mohr-
Columnmodel andthesupportingstructuresarecon-
sidered as elastic model. Theinterfaces areinstalled
between SMW and soils. The top of the model, at
z=30m, is afreesurface. Thebaseof themodel, at
z=0m, is fixedinthez-direction, androller bound-
ariesareimposedonthesidesof themodel, atx=0m,
x=60m, andy=0m, y=60m.
4.2 Case 2
Inmiddleof March2007, becauseof thebadweather
and other reasons, the steel pipe supports were not
installedintime.Therewereabout30minlengthwith-
outsupportsfromtheexcavatingfacetothelattermost
supportsfor alongtime. Meanwhile, accordingtothe
in-situmeasurements, therewasasharpincrement in
horizontal displacement of thesoil mixingwall. This
caseissimulatedtoanalyzetheinfluence.
Inthis case, thematerial properties andboundary
conditions are same to case 1. The excavating and
supportingproceduresaresame, too.
5 RESULTS
5.1 Case 1
Inordertoanalyzethecalculationresultsconveniently,
some key points are set in the model, as shown in
161
Figure4. Thehorizontal displacement at pointA incase1.
Figure3. Firstly, thehorizontal displacementatthetop
of SMWisanalyzed.Thecurveof horizontal displace-
mentsatpoint A isshowninFigure4(a). Itcanbeseen
that thehorizontal displacement increasesbeforestep
6butthendecreasesinthefollowingsteps. Themaxi-
mumvalueis6.76mmatstep6andtheultimatevalue
is4.63mm. Accordingtotheexcavatingstepsincalcu-
lation,whentheexcavatingfaceexceedspoint Aabout
15m, thevalueof thehorizontal displacement begins
todecline. Theactual observationdatafor point A is
showninFigure4(b). Itcanbeseenthattheactual val-
uesarebigger thanthecalculationresults. Thecurve
ismonotoneincreasingby stepsandtendstobecon-
stant after havingreachedacertainlevel. Itsultimate
valueis 10.4mm. Thecalculation result of thehori-
zontal displacement at point A is muchless thanthe
observationvalue.
The curves of calculation results with excavating
stepsandtheobservationdatawithdatefor thehori-
zontal displacement of theretainingstructureinline
B areshowninFigure5. Accordingtothecalculation
results, its maximumhorizontal displacement of the
SMWoccursatthepointof 6.5mdepth,anditsvalueis
22.35mm.AsshowninFigure5(a), whentheexcavat-
ingfacereachesthelineBatstep3, thedeformationof
SMWincreasesdramatically.Whenithaspassedaway
fromline B about 20m, the deformation increases
Figure5. Thehorizontal displacement of theSMWinline
B incase1.
slowly.Figure5(b)showstheobservationcurveof hor-
izontal displacement of SMWinlineB withthedate.
A rathergoodagreementcanbefoundbetween(a) and
(b). Accordingtotheobservationdata, themaximum
horizontal displacement inlineB occurredat point of
5.0mdepthanditsvalueis26.8mm, whichisgreater
thanthemaximumcalculationvalueby4.45mm.With
theexcavatingfaceadvancing, thedeformationincre-
mentisbecomingsmaller. Fromtheactual observation
dataandcalculationresults, itcanbeseenthatthehori-
zontal displacementof theSMWmainlyoccurreddur-
ingtheperiodof excavatingsurfacepassingthisline.
The curves of axial forces of the steel pipe sup-
ports C and D with excavating steps are shown in
Figure6(a). Thefinal axial forceof theupper support
C is657.60kN, andthelower support Dis1467.7kN
in calculation results. As shown in Figure 6(b), the
observationdataisgreaterthanthecalculationresults.
Themaximumaxial forceof theupper support C is
995.83kN, andthelowersupportDis1575.7kN. Both
C andDhaveanascendingfirstlyandthendeclining
processwithlapseof timeinactual observation. This
isbecausethefoundationmat boreapart of soil pres-
surewithitspouringandstrengthening. However, the
procedureof pouringandstrengtheningof foundation
matisnotsimulatedinnumerical calculation. Sothere
162
Figure6. Theaxial forceof supportsC andDincase1.
is no decliningtrendof axial forces. But intheearly
stage, thetrendandshapeof calculationresultscurve
withexcavatingstepsandobservationdatacurvewith
dateisagreedgenerally.AsshowninFigure6(b),on28
April thelower support Dwasremoved, therefore, the
axial forceof theuppersupportChadasignificantrise
by102.5kN. However, incalculationthisprocedureis
not simulated.
Through analysis, it can be seen the values of
observation data are greater than the calculation
results universally. The main discrepancy between
calculation and measure can be explained that the
physico-mechanical parametersof soilsarenot accu-
rateenough. However, inreality, thisareawaslessstiff
thaninitiallyplanned.Accordingtothenumerical sim-
ulationandtheactual observationdata, theexcavation
is stableif thesteel pipesupports canbeinstalledin
time. The numerical results and actual data of axial
forces arelower thanthealarmvalues. Thedisplace-
ment duetoexcavatingisinthepermissiblerange.
5.2 Case 2
In case2, thenumerical calculation model is shown
in Figure7. Figure8 shows thehorizontal displace-
ment of themodel and theaxial forces of steel pipe
supportsatlaststep. Itcanbeseenthemaximumhori-
zontal displacementatthetopof SMWoccursatpoint
Figure7. Elementsof model incase2.
Figure8. Thehorizontal displacement andtheaxial forces
of steel pipesupportsincase2.
Figure9. Thehorizontal displacement of point E incase2.
E with 48.96mm, as shown in Figure 7. The curve
of horizontal displacementsat point E whenexcavat-
ing fromstep 1 to step 11 is shown in Figure 9. It
canbeseenthedisplacement at point E isnearlyzero
before step 6 until the excavating face passes point
E. With the excavating face advancing after step 6,
thedisplacementincreasessignificantly.Accordingto
calculationresults, themaximumhorizontal displace-
ment of SMWoccursat thepoint of 3mdepth, which
is in lineF, and its valueis 50.44mm. As shown in
Figure10, thehorizontal displacementof SMWinline
Fdevelopsslightlyuntil theexcavatingfacepassesthis
lineat step6. Becausethesupports arenot installed
nearthisline, thedeformationof SMWdevelopsmore
rapidly. Inactual observation, therearethreepointsof
which horizontal displacements in lineF aregreater
than the alarmvalue with 50mm. The axial forces
163
Figure10. Thehorizontal displacementof theSMWinline
F incase2.
Figure11. Theaxial forcesof supportsGandHincase2.
of lastly installedsupports G andH aremuch larger
becauseof withoutinstallationof subsequentsupports.
The axial force of support G is 1924.50kN, and H
is 2095.60kN. Thecurves of theaxial forces of sup-
portsGandH withtheexcavatingstepsareshownin
Figure11. Accordingto calculationresults, theaxial
forces of G and H are much greater than the adja-
cent supports by 903.75kN for upper support and
546.00kN for lower support. Compared with Figure
6(a), themagnitudeis much greater. Theaxial force
of Gexceedsthealarmvaluewiththeupper supports
for 1500kN and H exceeds thealarmvalueof lower
supportsfor 2000kN.
TherearetwoinclinometertubesECX2andECX13
for lateral deformation of SMW around this site.
According to the observation data of 19 March
and 20 March, the velocity of horizontal displace-
ment exceeded the alarmvalue of 3mm/d for two
days between the depth of 5m8m at ECX13.
The velocity values were 4.18mm/d, 3.97mm/d,
3.64mm/d and 3.20mm/d when the depths are 5m,
6m, 7mand 8mrespectively on 19 March. More-
over, thevelocity valueswere5.18mm/d, 5.66mm/d,
4.98mm/dand3.44mm/dinthenextday. Meanwhile,
thevelocitiesof horizontal displacement weregreater
than3mm/dinthedepthfrom1mto12matECX2on
19March, whichwereover 10mm/dwithin3mfrom
thetopof SMW.
According to the results of numerical simulation
andactual observationdata, itwaspossibletocollapse
for this excavation because the supports were not
installed in time. The risk was existent so the cor-
responding measures should be taken. After being
alarmed, the construction team stopped excavating
and installed the supports speedily. Its turned out
that themeasures arevery effectiveaccordingto the
subsequent observeddata.
6 CONCLUSIONS
Inthispaper, firstly, thebearinganddeformingmech-
anism of SMW is analyzed in brief; secondly, the
constructionmonitoringschemeisintroduced; thirdly,
a 3D numerical simulation of this long stripe exca-
vation is described, including all the components of
theproject(theSMWconstruction,steppedexcavation
andsupports installation); thenthenumerical results
arecomparedwiththeactual datain-situobservation.
The3D numerical methodcansimulatethewhole
excavation construction very well. A good agree-
ment can befound between thenumerical results
and the actual observation data except for some
small deviations.
Theexcavationisstableandthedisplacement due
toexcavatingisinthepermissiblerangeif thesteel
pipesupportsarestalledtimely. However, because
of badweatherandotherreasonsthesteel pipesup-
portsarenotinstalledintimeandwithoutsupports
for alongtime, suchascase2, theexcavationisin
danger of collapse.
Accordingtothein-situobservationdata, thecon-
structionteamcantakecorrespondingmeasuresto
protecttheexcavationawayfromsomeundesirable
eventsandrisks.
REFERENCES
Commend, S. Geiser, F. & Crisinel, J. 2004. Numerical sim-
ulation of earthworks and retaining systemfor a large
excavation. Advances in Engineering Software Vol. 35:
669678.
Li, J.C. Zhang, Z.Y. & Xu, Q 2005. Study on three-
dimension numerical simulation of deformation of the
deep-foundationpit withexcavation. Journal of Nanjing
University of Technology 27(3): 17.
Liu, H.Y. et al 2006. Numerical analysis on excavation
safety of deep foundation engineering. Chinese Journal
of Geotechnical Engineering 28(Supp): 14411444.
Liu, J.G. & Zeng, Y.W. 2006. Application of FLAC3D to
simulation of foundation excavation and support. Rock
and Soil Mechanics 27(3): 505508.
Liu, J.H. & Hou, X.Y. 1997. The Handbook of Founda-
tion Engineering, Beijing: ChinaArchitecture&Building
Press. 569572.
Zhang,P&Liu,R.H.2000.TheApplicationof SMWMethod
inFoundationPit.Chinese Journal of Rock Mechanics and
Engineering 19(Supp): 11041107.
164
A simplifiedspatial methodologyof earthpressureagainst
retainingpilesof pile-rowretainingstructure
Y.L. Lin
Geotechnical Research Institute, Hohai University, Nanjing, P.R. China
Key Laboratory for Geotechnical Engineering of Ministry of Water Resource, Hohai University, Nanjing, P.R. China
X.X. Li
ABSTRACT: Pile-rowretainingstructureiswidelyadoptedintheexcavationof deepfoundationpit. Inthesup-
portingsystem, retainingpilesarethemainbearingmembers. Itisextremelyimportanttoobtainthemagnitude
anddistributionof earthpressureagainstretainingpiles. Basedonthemodeof failure, anewmethodologyispro-
posedtoevaluatetheearthpressureagainstretainingpilesof pile-rowretainingstructure.Intheproposedmethod,
bothspatial effect andintermediateprincipal stresseffect areconsidered. Finally, themethodologyisappliedto
practiceengineering. It is demonstratedthat thestrengththeory has moreinfluenceonearthpressureandthe
potential strengthof fillingmaterialsissufficientlydevelopedunder theguidanceof theunitedstrengththeory.
1 INTRODUCTION
Thepressureagainst theback of aretainingstructure
causedbybackfill andsurchargeonthegroundsurface
is a classical problemof soil mechanics. It is influ-
enced by retaining structuretypes, movement mode,
stiffnessandcontactconditionsbetweensoil andstruc-
ture(Fang& Ishibashi, 1986; Harrop-Willrams,1989;
Zhou,1990; Fang et al., 1994; Wang, 2000; Pal &
Salgado, 2003). Inaddition, thedistortionof soil mass
has acertaineffect ontheearthpressure. Beforethe
soil achievesbreakage,themagnitudeof earthpressure
cannotbedetermined. Evenif itreacheslimitstate, the
earthpressurecannotalsobecalculatedbecauseinner
soil masscannotsynchronouslyarriveatlimitequilib-
riumstate. Sothereliableparameter of soil cannot be
acquired. Thus, to apply inpracticeexpediently, it is
usuallyassumedthesoil isonideal failurestate.
In theexcavation of deep foundation pit pile-row
retainingstructureiswidelyadopted.Inthesupporting
system, retaining piles are the main bearing mem-
bers. It isveryimportant toobtainthemagnitudeand
distribution of earth pressureagainst retaining piles.
Because of failure mechanismof soil behind piles,
theinfluenceof interactionbetweenretainingstructure
andsoil onearthpressurecantbeachievedaccurately
accordingtoclassical earthpressuretheory. Soearth
pressureshouldbetakenasthespatial problemrather
thanplaneproblem. Mohr-Coulombstrengththeoryis
usually introducedinto thecomputationanalysis and
theinfluenceof intermediateprincipal stressisomit-
ted. However, plenty of experiments reveal the soil
strength varies with theintermediateprincipal stress
(Yu,2004),whichisquitedifferentfromwhathasbeen
depicted in the conventional Mohr-Coulomb theory.
Theunified strength theory is asystemof yield and
failurecriteriaof material sunder complexstresses. It
hasaclear physicsandmechanicsbackground, auni-
fied mathematical model, and a simple and explicit
criterion which includes all independent stress com-
ponentsandsimplematerial parameters(Yu, 2002).
In this study, a new methodology based on the
plasticslimitanalysisisproposedtoevaluatetheearth
pressureagainst retaining piles of pile-rowretaining
structurebasedonthemodeof failure. Intheproposed
method, bothspatial effect andintermediateprincipal
stresseffect areconsidered. Thesolutionof theequa-
tionisobtained, givingatheoretical resultfortheearth
pressureonretainingpiles.
2 UNIFIEDSTRENGTHTHEORY
Based on atwin-shear element and themultipleslip
mechanism,YuandHe(1992) establishedtheunified
strength theory. It has a unified model and simple
165
Figure1. Varietiesof theunifiedstrengththeoryondevia-
toricplane(Yu, 2004).
unified mathematical expression that is suitable for
various materials (Yu, 1994). The unified strength
theory covers all the regions fromthe lower bound
to upper bound, as shown in Figure 1. The unified
strength theory considers the different contributions
of all stress components actingonthestress element
totheyieldor failureof materials.
Themathematical modelingisexpressedasfollows
(Yuet al., 2002):
If it is prescribed that press stress is positive,
Equation(1) andEquation(2) canberewrittenas
wherec
0
=cohesion,
0
=frictionangle, b =unified
strength parameter that reflects theinfluences of the
intermediate principal stress on the yielding of the
material (0b1),
t
and
c
are uniaxial tensile
strength and compressive strength, respectively, and
istensile-compressivestrengthratio.
IntroducingLodeparameter j
, thus
SubstitutingEquation(5) intoEquation(3) and(4),
letting
The unified cohesion c
t
and the unified friction
angle
t
canbedefinedas
According to Mohr circularity of stress state at a
point, theunifiedexpressionof shear strengthcanbe
obtained
3 FAILURE MODE OF SOIL HEHINDPILES
For pile-rowretainingstructure, thearchingeffect in
the retaining soil mass occurs (Hu et al., 2000). It
is a stress redistribution process by which stress is
transferred around a region of the soil mass, which
then becomes subject to lower stresses. So theearth
pressureactingon piles is enhanced, whiletheearth
pressureonsoil aroundpilesisdepressed.Thesmaller
ispilespace, thestronger issoil archingeffect. Andit
ismorepropitioustothestabilityof foundationpit.
3.1 Simple shear failure mode
Figure2showsthefailuremodeA homogeneousfoun-
dationpitof depthH andthenetspaceBisconsidered.
166
Figure2. Simpleshear failuremodeof soil mass between
piles.
Figure3. Rip-shearfailuremodeof soil massbetweenpiles.
Thesoil wedgeisassumedtoberigidandslidealonea
planar surface. Thecritical inclinationof failureplane
isexpressedas.
WhenB2b 0,simpleshearfailuremodeoccurs,
letting
AccordingtoEquation(10), thefailureconditionof
simpleshear failuremodecanbeobtainedas
whereH
cr
=critical height of failuremode.
3.2 Rip-shear failure mode
Rip-shear failure mode will arise when H >H
cr
, as
showninFigure3.Thesoil massbehindpilesisdivided
into two portions fromthecritical height H
cr
. Above
theheight H
cr
, thesoil is rip failure, and it is shear
failurebelowtheheight H
cr
.
4 CALCULATIONOF ACTIVE EARTH
PRESSURE
4.1 Earth pressure of simple shear failure mode
Figure4(a) showsthemechanismof simpleshear fail-
ure. Thesinglepileendures theearthpressureof the
Figure 4. Schematic for earth pressure of simple shear
failure.
regionsoil of blockBCDB
. Itisassumedthatthe
soil is perfectly plastic andtheir deformationis gov-
ernedbytheassociativeflowrule. Then, Kinematical
admissibility requires thevelocity jumpvector, v, be
inclined to the velocity discontinuity at angle
t
, as
presentedinFigure4(b).
The Cartesian coordinates systemestablished for
the present analysis is shown in Figure 4(a). The
point C istheoriginof theCartesiancoordinatessys-
tem, andplaneXY islevel plane. SectionAC (or A
)
isslippagetangentof fracturesurfaceBCC
, anditis
inclinedtothevelocitydiscontinuityatangle
t
. Slim-
ily, thevelocityof fracturesurfaceBCDandB
is
atanangle
t
totheslippagetangentDC andD
. In
theplaneBCD, accordingtothedirectional derivative
of thevelocityv, theangle canbeobtainedas
Therateof workof soil weightcanbecalculatedas
theworkrateof blockABC-A
plustheworkrates
for blocks C-BAD andC
-B
. Consequently, this
workratetakestheform
where =the soil unit weight, d =the diameter of
pile, f
1
isafunction, it canbedeterminedas
Duetothehomogeneoussoil massesbeingrigid,the
internal energyisonlydissipatedalongtheslidingsur-
face.Theworkdissipationratecanbecalculatedasthe
167
work dissipationrateof block CC
DD
plusthework
dissipationratesfor blocks DBC andD
. Conse-
quently, thisdissipationworkratecanbecalculatedas
where
Sincesoil-pileinterfacecanbeconsideredasveloc-
itydiscontinuityrather thanstresscharacteristic, ideal
plastic model is not applicableto theinterfacemate-
rial. The relative movement between soil and pile,
whichdependsontheinterfacecharacteristicsandthe
property of theadjacent soil, is not always of purely
frictional sliding. If it is assumedthat thetotal hori-
zontal earthpressureinducedbysoil massisP
au
. The
frictionangleof soil-pileinterfaceis.Twoconditions
areconsideredinthefollowing.
1 Smoothpile( <
t
)
Theexternal work ratecontributedbytheresultant
horizontal earthpressureP
au
is
Theworkdissipationratealongpilesurfaceis
Equatingtherateof internal energy dissipationto
therateof theexternal work, wecanobtain
2 Roughpile(
t
)
Theexternal work ratecontributedbytheresultant
horizontal earthpressureP
au
is
Theworkdissipationratealongpilesurfaceis
Similarly, equatingtherateof internal energy dis-
sipationtotherateof theexternal work, theresultant
horizontal earthpressurecanbeobtained
Figure5. Schematicfor earthpressureof rip-shear failure.
4.2 Earth pressure of rip-shear failure mode
Therip-shear failuremechanismfor thepresent anal-
ysisisshowninFigure5.Theearthpressureactingon
singlepileisinducedbytheregionsoil massFCDG-
G
.Similartotheearthpressureof simpleshear
failure mode, the rate of work due to the soil mass
weight canbeexpressedas
where
whereS =thecenter distanceof twoadjacent piles.
For the rigid material considered, the internal
energy is only dissipated along the sliding surface
and the interface surface of soil-pile. The rate of
energy dissipation along the sliding surface can be
expressedas
168
Figure6. Relationsbetweentheunifiedstrengthparametersandthesoil strength.
where
Theratedissipationalongtheinterfacesurfaceof
soil-pile is similar to the Equation (16) and (19).
For Smooth pile ( -
t
), according to the energy
conversationlaw, wecanobtain
Similarly,forroughpile(
t
),thetotal horizontal
earthpressureP
au
canbederivedas
Accordingly, thecorrespondingresultant P
a
acting
onthepileis
The magnitude of active earth pressure can be
obtained fromEquation (25). Obviously, for agiven
example, theresultant P
a
is only determinedby fail-
ureangle. Bytakingthefirstderivativesof Equation
(25) with respect to , and equating it to zero and
solving it, wecan obtain thecritical angle
cr
. Sub-
stituting
cr
intoEquation(25), wehavethemaximal
upper-boundfor theactiveearthpressure.
5 RESULTSANDDISCUSSION
Toevaluatethevalidityof theproposedmethod, apile-
row retaining structure without anchor is analyzed.
Numerical resultsarepresentedandcompared.
5.1 Effects of the unified strength parameters on
soil parameters
Theinfluences of theunifiedstrengthparameters on
soil mass strength are represented in Figure 6 for
0
=15
. It is observed that the soil mass strength

varies with the variety in the value of the unified
strengthb. Thesoil mass parameters of c
t
and
t
are
piece-wisefunctionsandtheyachieveextremumwhen
j
=sin
0
.
failure mode
Figure 7 presents the effects of the unified strength
parameters on critical height H
cr
for =17kN/m
3
,
0
=15
, H =10m. As a whole, the critical height

H
cr
decreases with the increase in the value of b.
Figure7(a) and Figure7(b) also show theinfluence
of S andpilediameter d. For c
0
=0, thecritical height
increases with the increase in the value of S, while
it decreases with the increase of d. It is also clear
169
Figure7. Influenceof theunifiedstrengthparametersoncritical height H
cr
.
that under thesameconditionsH
cr
decreaseswiththe
increaseof c
0
or inFigure7(c) andFigure7(d).
active earth pressure
Based on the unified strength theory, the values
of active earth pressure is shown in Table 1 for
=17kN/m
3
,
0
=15
, H =10m, d =1.0m, c
0
=0.
The proposed formula can be degenerated into the
expressions inducedby Mohr-Coulombstrengththe-
ory. From the table, it is found that the strength
Table1. Effectsof theunifiedstrengthparametersonactive
earthpressure.
b =0 b =0.5 b =1
Deg (1) (2) (1) (2) (1) (2)

0 483.6 483.6 414.9 396.3 378.8 350.0
15 364.1 364.1 305.4 378.8 275.5 252.0
b =0issimplifiedtoMohr-Coulombtheory; b =1istwin
shear theory. Situation (1) represents j
=0, and situation

(2) isj
=sin
0.
170
theory has moreprodigious influences on theactive
pressure. The resultant earth pressure P
a
decreases
withtheincreaseinunifiedstrengthparameter b. For
j
=0, when b changes from0 to 1, the resultant

earth pressure P
a
decreases by about 21.7%. Simi-
larly, theearth pressureP
a
is also influenced by the
Lodeparameter j
.
6 CONCLUSIONS
The estimation of active earth pressures acting on
retaining piles is very important in geotechnical
design. However, unlike the assumption used in the
analysisof Coulomb, whichgenerallycalculatesearth
pressureaccordingto planestrain, theearthpressure
behindthepiles shouldbetakenas thespatial prob-
lem.Thisisduetoarchingeffectsintheretainingsoil,
whichresultfromthefrictional resistancebetweenthe
pilesandthesoil.
In this paper, a new methodology is proposed to
evaluatetheearth pressureagainst retaining piles of
pile-rowretainingstructure.Theadvantageof thepro-
posedmethodliesinthefactthatbothspatial effectand
intermediateprincipal stresseffectareconsidered. Itis
indicatedthat thestrengththeory has moreinfluence
onearthpressureandthepotential strengthof filling
materialsissufficientlydevelopedunder theguidance
of the united strength theory. But the methodology
requiresexperiment or fieldverification.
REFERENCES
Fang, Y. & Ishibashi, I. 1986. Static earth pressures with
various wall movements. Journal of Geotechnical Engi-
neering, ASCE, 112(3):317333.
Fang,Y.S.,Chen,T.J.&Wu,B.F.1994.Passiveearthpressures
with various walls movements. Journal of Geotechnical
Engineering, ASCE, 128(8):651659.
Harrop-Willrams, K.O. 1989. Geostaticwall pressures. Jour-
nal of Geotechnical Engineering, ASCE, 115(9):1321
1325.
Hu, M.Y., Xia, Y.C. & Gao, Q.Q. 2000. Calculationprinci-
ple of earth pressure against retaining piles of pile-row
retaining structure. Chinese Journal of Rock Mechanics
and Engineering, 19(3):376379.
Palk, K.H. & Salgado, R. 2003. Estimation of activeearth
pressureagainst rigidretainingwallsconsideringarching
effects. Geotechnique, 53(7):643653.
Wang,Y.Z. 2000. Distributionof earthpressureonaretaining
wall. Geotechnique, 50(1):8388.
Yu, M.H. 2004. Unifiedstrengththeoryanditsapplications.
Berlin: Springer.
Yu, M.H. 2002.Advanceinstrengththeoryof materialsunder
complex state in the 20th Century. Applied Mechanics
Reviews, 53(3):159218.
Yu, M.H. 1994. Unified strength theory for geomaterials
and its applications. Chinese Journal of Geotechnical
Engineering, 14(2):110.
Yu, M.H., He, L.N. & Liu, C.Y. 1992. Generalized twin
shear stressyieldcriterionanditsgeneralization. Chinese
Science Bulletin, 37(24):20852089.
Yu, M.H., Zan, Y.W. & Zhao, J. 2002. A unified strength
criterionfor rock material. International Journal of Rock
Mechanics and Mining Sciences, 39:975989.
Zhou, Y.Y. & Ren, M.L. 1990. Experimental study of the
active earth pressure on rigid retaining wall. Chinese
Journal of Geotechnical Engineering, 12(2):1926.
171
Considerationof designmethodfor bracedexcavationbasedon
monitoringresults
H. Ota, H. Ito&T.Yanagawa
Osaka Municipal Transportation Bureau, Osaka, Japan
A. Hashimoto
Kotsu Service Co., Ltd., Osaka, Japan
T. Hashimoto&T. Konda
ABSTRACT: A comparisonbetweenobserveddataanddesignvalueof earthretainingwall deflectiondueto
bracedexcavationwascarriedoutinsoftandsensitiveclaygroundof someconstructionsitesof OsakaSubway
LineNo.8.Thebeam-springmodel wasemployedinthebraceddesignmethod, anditwastakenintoaccountthe
characteristicsof theOsakasoft ground, andtherewasgoodagreement betweentheobserveddataanddesign
valueinpast results. Accordingtothis comparison, theobservedwall deflectionwas larger thanthedesigned
oneinsomeconstructionsites consistedof thesoft andsensitiveclay layer with10to 20mthickness. Inthis
paper, themeasuringprocessof thehorizontal coefficient of subgradereactionk
h
intheexcavationsideof soft
claylayer isdiscussed.Asthevalueof k
h
becamesmall dependedonthewall deflection, thenewknowledgehas
beenemployedonthedesignmethod. It isfoundthat thecalculationwiththereviseddesignmethodagreewell
withthemonitoringdata.
1 INTRODUCTION
In densely populated city, it is necessary to use the
undergroundspacehighlyandeffectivelyforthedevel-
opment of city. It is believed that the demand for
much deeper underground excavation will increase
gradually. Therefore an applicable design method is
demandedfor deep, safeandeconomical excavation.
OsakaMunicipal TransportationBureau(OMTB)sug-
gested an original design method for braced exca-
vation based on the characteristics of the Osaka
groundandsubway constructions. At eachconstruc-
tion site (elevens stations and railway depot) where
opencut methodwasadoptedinOsakaSubway Line
No.8, bracedexcavationdesignbasedonthisoriginal
designguidelinewascarriedout,andtheobservational
methodwasalsoutilizedeffectively.
Inthispaper, somecomparisonsbetweenobserved
dataanddesign valueof earth retainingwall deflec-
tionduetobracedexcavationhavebeencarriedouton
softandsensitiveclaygroundof twoconstructionsites
of OsakaSubway LineNo.8. Theevaluationmethod
for design has been described based on the ground
properties.
2 CHARACTERISTICSOF THE MODIFIED
BEAM SPRINGMODEL
East andWest sides of Osaka ground areconsistent
withtheHolocenelayers(softclayandloosesand),but
Pleistocenelayersexist aroundthegroundsurfaceof
Uemachi plateau. Thewater levelsarehighinuncon-
finedandconfinedaquifers, alsothepermeability of
theseaquifersarelarge.
Thebeam-springmodel for thebracedexcavation,
whichisindicatedinStandardSpecificationsforTun-
nel [OpenCutTunnel]publishedbytheJ apanSociety
of Civil Engineers(J SCE, 2006), isfrequentlyimple-
mentedasawidelyusablemethodinJ apan. However,
sincetheresultof thepredictionof wall deflectionand
strut forcearenot alwaysconsistent withtheobserva-
tiondata, OMTB proposedthemodifiedbeamspring
design method (OMTB, 1993) (Kishio et al., 1997)
173
which can consider the characteristics of Osaka
groundandtheconditionsof subwayconstruction.
Theoutlineof theOMTB model isshowninFigure1.
2.1 Active lateral pressure above the excavation
bottom
Becausetherearesomepossibilities of gap between
the braced wall and back ground, the active earth
pressureisestimatedbyRankine-Resalsequationwith
thewater pressure. Insandy layer, thewater pressure
isassumedashydrostaticpressure. Inclayeylayer, the
water headof theupper sandylayer isextendedinthis
clayeylayer.
2.2 Active lateral pressure below the excavation
bottom
In thecaseof sufficient penetration depth of braced
wall, the wall deflection near the tip is small and
the lateral pressure is kept as the initial condition.
So, if theactivelateral pressureis defined basically
only by thelimit equilibriumtheory, therearesome
cases which the wall deflection is overestimated by
givingmuchlateral pressure.Therefore, theactivelat-
eral pressurewhichisdeeper thanthebottomlayer of
excavationisgraduallydecreasingintheareaof trian-
gleformedfromthelateral pressureat thebottomof
theexcavationtothetipof wall.
2.3 Resisting lateral pressure of the excavated
ground
Resisting lateral pressure of the excavated ground
is the multiplication of the coefficient of horizontal
Figure1. Theconcept of themodifiedbeamspringmodel
(presentedbyOMTB, 1993).
subgradereaction and thewall deflection. However,
this value should not exceed the coefficient of the
passivelateral pressurewhichisthesubtractionfrom
limit passive lateral pressure defined by Coulombs
equationtothelateral pressureat rest.
2.4 Water pressure in clayey ground
Sinceit is difficult to distinguish thewater pressure
fromthelateral pressureinclayeyground, lateral pres-
sureis often identified as theintegration of soil and
water. Ontheother hand, it is consideredthat soil is
separatedfromwater inmodifiedbeam-springmodel.
Because the pore water pressure acts on the braced
wall hydrostatically, thewater pathispossiblyformed
betweenthewall andtheback grounddueto braced
excavation.
For these reasons, the effective stress method is
adopted in both sandy layer and clayey layer. The
groundwater table in clayey ground is taken on the
higher gravitational pressuredistributionbycompari-
sonbetweentheupper water-bearinglayer anddown
sidewater-bearinglayer.
2.5 Supported effect of covering plates
Becausetheeffectof depressingthewall deflectionis
recognized when thecovering plates areconstructed
inthesamedirectionasstruts, thesupportedeffect of
coveringplatesisconsideredby reducing10%of the
spring-beamcoefficient.
2.6 Horizontal coefficient of subgrade reaction of
excavated side
The coefficient of subgrade reaction k
h
used in the
J SCE model is taken into consideration the geo-
metrical effect related to the difference of loaded
width based on some plate loading test results per-
formednear thegroundsurface. However, thelateral
pressure acts on the horizontal direction against the
earth retaining wall, because the wall is installed
to thevertical direction in subway construction site.
Therefore, it is not always appropriate to apply the
coefficient of subgrade reaction used in the J SCE
model tobracedexcavationdesigndirectly. Sointhe
OMTB model, the coefficient of subgrade reaction
is expressedas equation(1) and(2) (Yanagidaet al.,
1981) empirically.
174
3 COMPARISONSBETWEENOBSERVATION
ANDDESIGNOF BRACEDWALL
DEFLECTION
The comparison between observed data and design
valueof earthretainingwall deflectionduetobraced
excavation was carried out in soft and sensitiveclay
groundof No.8Lineconstructionsites.
In general, the design value can estimate the
observeddataappropriatelyinmostconstructionsites.
Butinsomesensitiveandsoftalluvial claylayer accu-
mulate from10mto 20mthick, observed data are
larger than the design value due to braced excava-
tion. The causes for these phenomena are described
asfollows.
Figure2. Crosssection(A site).
Table1. Earthretainingwall andeachstrut (A site).
Soil mixingwall (H-steel) condition
Size Pitch Length EI Area
(mm) S(m) L(m) (kNm
2
/m) A(m
2
)
H-5883001220 0.60 27.25 399000 0.01925
Excavationcondition Strut condition
Depth Depth
Size Span Pitch Area
Step (GL-m) Stage (GL-m) (mm) L(m) S(m) A(m
2
)
0th 1.51 Cover beam 0.51 H-5883001220 16.25 2.00 0.01925
1st 4.50 1st 3.50 H-3003001015 14.76 2.50 0.01048
2nd 7.00 2nd 6.00 H-3003001015 14.86 2.50 0.01048
3rd 8.70 3rd 7.70 H-3003001015 14.66 2.50 0.01048
4th 11.20 4th 10.20 H-3003001015 14.76 2.50 0.01048
5th 13.90 5th 12.90 H-3503501219 14.46 2.50 0.01549
6th 16.45 6th 15.45 H-3503501219 14.46 2.50 0.01549
7th 19.05 7th 18.05 H-3503501219 14.46 2.50 0.01549
8th 21.55
3.1 A-site
Thelayer of thisA-sitegroundconsistsof thealluvial
layer,upperlowerPleistocenefromthegroundsurface.
The uniformity coefficient of this fine sand Aus
is small, and is called the first water-bearing layer.
Thealluvial clayeylayerAucissensitive(N-value =0
to 3, liquid limit I
L
=0.4 to 1.0, cohesion c =20 to
100kN/m
2
),andistypical softgroundinthisconstruc-
tionsite. Ontheother hand, under thealluvial layer,
theupper Pleistocenesandyandgravel layer Tsg(the
secondwater-bearinglayer),thelowerPleistoceneclay
layer Oc3(OsakaGroup, MarineClayMa3, c =about
400kN/m
2
) and the lower Pleistocene sandy layer
Os3 (OsakaGroup, thethird water-bearing layer, N-
value >60) aredeposited continuously to horizontal
direction.
Thecrosssectionof A-siteisshowninFigure2, the
wall andstrutsconditionsareshowninTable1andthe
soil parametersareshowninTable2. Inthisconstruc-
tionsite, theseepagecontrol methodwasadoptedby
extendingtheearthretainingwall tothelowpermeable
layer Oc3, excavationwidthis16.2m, thefinal exca-
vationdepthis GL-21.5mandthepenetrationdepth
is4.8m.
Figure3shows thecomparisonbetweenobserved
dataanddesignvalueof theearthretainingwall deflec-
tion. Theobserved wall deflections in east and west
sidesaresymmetrictill the4thstep.Itisconfirmedthat
thedesignvaluecanestimatestheactual phenomenon
adequately. However, sincethe5thstep, theobserved
wall deflection of west sidewas larger than theeast
side, whichcanbeseenfromtheresultsof the8thstep
175
in Figure3. It can beassumed that thecauseof this
phenomenonwas thedifferenceof constructionpro-
cessanddevelopmental patternof creepdeformation.
Moreover,theobserveddataexceededthedesignvalue
at thewest side. It was consideredthat this disparity
occurredfor thereasonthat theplasticzoneunder the
excavationbottomexpandedintheAmcandTsglayer
fromthe5thstepdrastically. Inaddition, it was con-
firmedthatthestressinthewall wascontrolledwithin
theallowablestress.
In theexcavation stageat theAmc layer, thecal-
culation result considering the 75%stress reduction
under the5mfromtheexcavationbottomwasshown
together inFigure3. Duringanexcavationincohesive
Table2. Soil parameters(A site).
Bottom Internal
Soil depth Cohision friction E
50
layer (GL-m) N-value c (kN/m
2
) angle (
) (MN/m
2
)
B 1.8 2 0 19.3
Auc 4.9 4 42 0 4.1
Aus 8.3 2 0 19.3
Amc1 13.8 0 29 0 4.7
Amc2 16.8 1 60 0 6.9
Amc3 19.4 4 91 0 15.9
Alc 21.8 6 108 0 15.9
Tsg 25.4 26 0 32.7
Oc3 31.6 14 360 0 83.6
Figure3. Comparisonbetweenobservationanddesignvalueof bracedwall deflection(B site, (a) : the4thstep, (b) : the8th
step).
soil, if anexcavationstagetakesalongtime, thesuc-
tionof thesubgradesoil will disappearduetoswelling
causedbythewater infiltrationfromcontinuousrain-
fall, whichleadstoreductioninstrength(Hashimoto
et al., 1997). In conjunction with this arrangement,
the ultimate passive lateral pressure and coefficient
of subgradereactionwerereduced. Thisphenomenon
wasverifiedbytheconsolidationwithun-drainedtri-
axial compressiontest,inwhichshearstrengthreduced
to 70% after the suction was disappear completely
and after measuring the water pressure and suction.
Inshort, it isprovedthat thereisapossibilitythat this
phenomenonmayoccur (Katoet al., 2006).
Inthe8thstep, thecalculationresultconsideringthe
stressreductionexceededthedesignvaluewhichcould
explaintheobserveddataappropriatelytosomeextent.
However, under the bottomof the excavation, espe-
cially inTsglayer, thetendency that thedesignvalue
andcalculationresultexceededtheobserveddata.The
wall deflectiondistributionwasdifferent betweenthe
observeddataanddesignvalueandcalculationresult.
It wouldappear that oneof thereasonsfor theseten-
dencies is thedeformation at thebottomof thewall
towardstheexcavationside.
3.2 B-site
The layer of this B-site ground constitutes the
alluvial layer, upper lower Pleistocene from the
176
groundsurface. Especially, thisconstructionsitewas
located in the Neyagawa lowland, and it is peculiar
that the very soft and sensitive alluvial clay layer
(N-value
=0, liquid limit I

L
=0.6 to 1.5, cohesion
c =30 to 100kN/m
2
), which is specific for theEast
side of Osaka Plain, deposited with 15 to 20m
thickness. The upper Pleistocene sandy and gravel
layer Ts & Tgandthelower Pleistocenesandy layer
Os3 (Osaka Group) constitute the second water-
bearinglayer under thealluvial layer.
Thecrosssectionof B-siteisshowninFigure4, the
wall andstrutsconditionsareshowninTable3andthe
soil parametersareshowninTable4.
Inthisconstructionsite,theseepagecontrol method
was adoptedby extendingtheearthretainingwall to
the low permeable layer Oc7 (about GL-42m), too.
Thebottomdepthof Soil MixingWall (H-steel) was
extendedtotheOs8. Theexcavationwidthis17.2m,
Figure4. Crosssection(B site).
Table3. Earthretainingwall andeachstrut (B site).
Soil mixingwall (H-steel) condition
Size Pitch Length EI Area
(mm) S(m) L(m) (kNm
2
/m) A(m
2
)
H-5883001220 0.60 27.52 399000 0.01925
Excavationcondition Strut condition
Depth Depth
Size Span Pitch Area
Step (GL-m) Stage (GL-m) (mm) L(m) S(m) A(m
2
)
0th 1.42 Cover beam 0.42 H-4883001118 17.15 2.00 0.01592
1st 2.81 1st 1.81 H-3003001015 16.55 2.59 0.01048
2nd 5.96 2nd 4.96 H-3503501219 16.45 2.59 0.01549
3rd 8.26 3rd 7.26 H-3503501219 16.45 2.59 0.01549
4th 11.51 4th 10.51 H-3503501219 16.45 2.59 0.01549
5th 14.51 5th 13.51 H-3503501219 16.45 2.59 0.01549
6th 17.21 6th 16.21 H-4004001321 16.35 2.59 0.01977
7th 19.61 7th 18.61 H-4004001321 16.35 2.59 0.01977
8th 21.70
thefinal excavationdepthis about GL-22mandthe
penetrationdepthisabout 5m.
Figure 5 shows the comparison between the
observeddataanddesignvalueof theearthretaining
wall deflection in the 4th and 8th excavation steps
(Oota et al., 2007). The wall deflection distribution
modewassimilar bothsideinthe4thexcavationstep.
However theamount of theobservedwall deflection
was two times of thedesign value. Moreover, in the
8th excavation step, the wall deflection distribution
modewasdifferent inbothandobserveddataexceed
thedesignvalue. Inaddition, itwasconfirmedthatthe
stress inthewall was controlledunder theallowable
stress.
Table4. Soil parameters(B site).
Bottom Internal
Soil depth Cohision friction E
50
layer (GL-m) N-value c (kN/m
2
) angle (
) (MN/m
2
)
B 0.8 2 0 19.9
Auc 2.0 0 27 0 2.2
Aus 4.0 2 0 19.9
Amc1 8.0 0 42 0 2.2
Amc2 13.0 0 63 0 5.5
Amc3 16.0 0 76 0 7.4
Alc 19.0 3 73 0 5.6
Tc 20.8 7 129 0
Ts 23.3 42 0 37.5
Tg 26.0 45 0 38.2
Os8 39.1 60 0 41.8
177
Figure5. Comparisonbetweenobservationanddesignvalueof bracedwall deflection(B site, (a) : 4thstep, (b) : 8thstep).
As the ground condition under the bottom of
the excavation is considered as the plastic zone in
thecalculationusingJ SCE model, thebottomof the
wall deformedtowardstheexcavationsideinalarger
valueandthewall deflection distribution haddiffer-
ent phenomenoncomparedtotheobservation. Inthe
excavationstageattheAmclayer, thecalculationcon-
sideringthe75%stressreductionunder the5mfrom
theexcavationbottomwasshowntogetherinFigure5.
Unlike the comparison result inA-site, this calcula-
tionwassimilar tothedesignbecausewall deflections
around thebottomof theexcavation arein theplas-
tic zone. It was impossible to explain the observed
phenomenon adequately used by some calculation
model.
Thehorizontal coefficientof subgradereactionk
h
of
clayeygroundforexcavationsideintheOMTBmodel
isdeterminedbyequation(2). Thissettingmethodof
k
h
wastheempirical equationbasedonmanyobserved
datainthecasethatthewall deflectionwasabout3cm
(Yanagida et al., 1981). This reference bring up the
problemthat k
h
istendtodecreaseduetotheincrease
of thewall deflection.
In the actual construction site, as k
h
is depended
on the ground mechanical characteristics and some
boundary conditions and so on, it is known that
k
h
changes every second due to braced excavation.
For example, k
h
is determined as solid line by the
wall deflectionfunctiontakingintoconsiderationthe
nonlinearity(J apanRoadAssociation, 1986).
Theinverseanalysisbasedonsomeearthretaining
monitoringresultswascarriedouttoestimatethevalue
of k
h
. ModifiedPawell Methodwasemployedfor the
inverseanalysis. It ispossibletoobtaintheoptimized
solutionstablyonthemanyunknownparameter prob-
lem(Kishio, et al., 1995). Input valuesfor theinverse
analysis areearth retaining wall deflection (angleof
inclination) andaxial forceof struts, andoutputvalues
arelateral pressureontheearthretainingwall andk
h
.
Figure 6 shows the inverse analysis results based
on themonitoring datain OsakaSubway LineNo.8
touchedtotheKishio, et al., 1997.Thevertical scaleis
theratioof theestimationvaluek
h
bytheinverseanal-
ysistodesignvalueof k
h0
. Inother words, k
h
,k
h0
=1
meanstheinverseanalysisresultsanddesignvalueare
thesame.
Inthecasethat thewall deflectionwasabout 1cm,
the relation between k
h
and k
h0
was about the same
in both past actual results and Line No.8 results. In
short,theapplicabilityof k
h
inthedesignisreasonable.
178
Figure 6. Dependence for brace wall deformation of k
h
(touchinKishio, et al., 1997).
However inthecasethatthewall deflectionwasabout
2to4cm, inverseanalysisresultsk
h
aresmaller than
the design value k
h0
, the ration k
h
,k
h0
decreased to
about 0.5.
The relation of =1/40 was presumed on the
assumption that k
h
decrease due to the increase of
thewall deflection. Figure5shows therecalculation
resultsunder thisrelation.
Inthecaseof the4thexcavationstep, therelation
between observation and recalculation was in good
agreement. Inthecaseof the8thexcavationstep, the
wall deflectiondistributionmodebetweenobservation
andrecalculationwasstill different, butthemaximum
amount of wall deflection was similar. It is believed
that the cause of differences in the wall deflection
distributionmodeis thelimit explainedby thebeam
springmodel indesign.
In accordancewith theseestimations results, it is
preferabletoconsider thek
h
settingmethodcarefully
asequation(3) withconsideringthetraditional design
ideainthecaseof earthretainingdesigninthesoftand
sensitiveclayey layer, whichN-value is about 0to 2,
withthicklayer (about 10to20m).
4 CONCLUSIONS
Theresultsareshownasfollows;
1. At theA-site, theobserved wall deflection in the
east andwest sidearesymmetric till the4thstep,
and it is confirmed that the design value esti-
matestheactual phenomenonadequately. However
since the 5th step, the observed data exceeded
the design value. It was assumed that the plastic
zoneexpanded drastically to thepenetration area
indesign.
2. It was possiblethat thecalculation result consid-
eringthe75%stressreductionunder the5mfrom
thebottomof theexcavationexplainedtheobserved
datatosomeextent. However, under thebottomof
theexcavation, thetendency that thedesignvalue
andcalculationresult exceededtheobserveddata.
It wasconsideredthat oneof thereasonsfor these
tendenciesisthedeformationatthebottomof wall
towardstheexcavationside.
3. At theB-site, thewall deflectiondistributionwas
similar betweentheobservationanddesigninthe
4th excavation step. However the observed wall
deflectionistwotimesof thedesign. Inthe8thstep,
thewall deflectiondistributionmodewasdifferent
inboth, andobservedwall deflectionexceededthe
designvalue.
4. Owingthat=1/40waspresumedontheassump-
tionthatk
h
decreaseduetotheincreaseof thewall
deflection, the relation between observation and
recalculation was in good agreement in the case
of the 4th excavation step. In the 8th excavation
step, thewall deflectiondistributionmodeinboth
wasstill different, butthemaximumwall deflection
wasclose.
5. It isrecommendedthat thek
h
settingmethodcare-
fullyasequation(3)withconsideringthetraditional
design idea in the case of earth retaining design
in the soft and sensitive clayey layer, which N-
value is about 0 to 2, with thick layer (about 10
to20m).
REFERENCES
Hashimoto,T.,J.Nagaya,J.KishioandT.Shiotani : Investiga-
tionof StrengthDegradingduetoSwellingof theGround
in Excavation, Proc. of the Int. Conf. on Foundation
Failures, pp.393398, 1997.
J apanRoadAssociation: DesignGuidelinefor Underground
Multipurposeduct, 1986(inJ apanese).
J SCE : StandardSpecificationsforTunneling-2006, Cutand
CoverTunnels, pp.142181, 2006(inJ apanese).
Kato, S., T. KondaandH. Shinkai: Effect of SuctionReduc-
tionCausedbyWettingProcessonShear StrengthChar-
acteristicsUnderLowConfiningPressure, JSCEJournals
C, Vol.62, No.2, pp.471487, 2006(inJ apanese).
Kishio, T., N. Nakai, H. Arimoto, T. KondaandK. Takami :
Inverseanalysisexampleof BracedWall usedbyModified
Pawell Method, Proc. of the50thJ SCE Annual Meeting,
III-520, pp.10401041, 1995(inJ apanese).
Kishio, T., H. Oota, T. Hashimoto, T. Konda, E. Saito
and N. Kobayashi : Estimation of Lateral Pressure and
Coefficientof SubgradeReactionduringExcavationWork
in Osaka, JSCE Journals, No.560, VI-34, pp.107116,
1997(inJ apanese).
179
Kishio, T., H. Oota, T. Hashimoto and T. Konda : Some
Aspects of Designing Earth Pressures for Braced Wall
under the Bottom of Excavation, Tsuchi-to-Kiso JGS,
Vol.45, No.10, pp.2022, 1997(inJ apanese).
Oota, H., H. Ito, T. Yanagawa, T. KondaandT. Hashimoto:
Considerationof DesignMethodfor BracingExcavation
BasedonMonitoringResults, Proc. of the 42nd JNCSFE,
2007(inJ apanese).
OsakaMunicipal TransportationBureau: DesignGuideline
for temporarystructure(draft), 1993(inJ apanese).
Yanagida, S., T. Watanabe, I. Yamaguchi, H. Nakamuraand
S.Mizutani :TheStudyof Lateral EarthPressureforEarth
Retaining Design (Part II), Proc. of the 16th JNCSFE,
14491452, 1981(inJ apanese).
180
Groundmovementsinstationexcavationsof Bangkokfirst MRT
N. Phienwej
Asian Institute of Technology, Bangkok, Thailand
ABSTRACT: Thecharacteristicsof movementsof diaphragmwall andgroundintheexcavationof 18stations
of thefirst BangkokundergroundMRT linewereevaluated. Threemodesof deflectedshapesof thewallswere
observedatdifferentexcavationdepths, namelycantilever mode, bracedmodeswithbulgeinsoftclayandbulge
in stiff clay. Theratio of maximumlateral wall deflection developing with excavation depth and theratio of
groundsurfacesettlement toexcavationandthenormalizedzoneof groundsurfacesettlement variedwiththe
modeof wall deflection. Back-calculationof undrainedsoil moduli for differentsoil layersweremadefromwall
movementdataof threeselectedstationsusingthe2-Dlinear elasto-plasticFEM analyses. Themodulusvalues,
whichwerehigher thanthosecommonlyobtainedfromconventional triaxial tests, canbeusedasguidelinefor
futureworksinBangkok.
1 INTRODUCTION
Deep excavation by means of strutted concrete
diaphragm walls is often used for construction of
multi-level buildingbasements inBangkok soft soil.
It is superior to theexcavation with steel sheet piles
for control of groundmovement toavoiddamagesto
adjacentstructures.Priortotheconstructionof thefirst
Bangkok MRT project, fewstudiesweremadeonthe
characteristics of the ground movement and its pre-
diction (e.g. Phienwej and Gan, 2003 andTeparaksa
et al, 1999. etc.). However, it was the implemen-
tation of the first Bangkok MRT subway line that
provided systematic and comprehensive monitoring
dataontheexcavationof stationboxesthatallowedin-
depthevaluationonthecharacteristicsof thewall and
groundmovementsassociatedwithdeepexcavationin
Bangkok subsoil conditionsusingstrutteddiaphragm
walls.Theprojectinvolvedthedeepesteverexcavation
madeinBangkoktodate.Theexcavationsof all station
boxeswerefully instrumented. Monitoringdatafrom
18stationexcavationswerecompiled,summarizedand
interpreted.
The construction of the first MRT underground
projectinBangkok,theMassRapidTransitInitial Sys-
temProject(MRT ISPBlueLine) wasstartedin1998.
Prior tothattherewerepublicdoubtsontechnical via-
bilityof theconstructionandoperationsafetyof under-
ground MRT in Bangkok soft soil. That pessimistic
outlookledtoacall foranin-depthinvestigationonthe
applicationandperformanceof theexcavationmethod
andsupportsystemstobeintegratedintheexcavation
insoft andsubsidingBangkok ground. Thecontracts
madeitscompulsorythatfull instrumentationprogram
beimplemented during excavation for design verifi-
cationandsafety assurance. Evaluationontheactual
performanceatsiteswasperformedtoconfirmthesuf-
ficiency inthedesignof thesupport systems for the
MRT stationexcavations.A comprehensivestudywas
madeontheaspectof thewall andgroundmovements
(Hooi, 2003) andthesalient pointsfromthestudyare
reportedherein.
2 PROJ ECT DESCRIPTION
TheISP BlueLineisthefirst undergroundMRT line
constructedinBangkok.Itcomprises22-km-longtwin
single-tracktunnels, 18stationsandadepot.Thehori-
zontal alignmentmainlyfollowstherightof wayof city
roads. Constructionof theundergroundstructureswas
implemented under two fast track design-built con-
tracts, each having approximately thesameamounts
of work. TheSouthContract involvedconstructionof
atwinboredtunnelsfromtheinter-cityrailwaytermi-
nal at HuaLamphongeastwardsfor 5kmbeneaththe
busyRamaIV roadtotheQueenSirikitNational Con-
ventionCenter, then4.5kmnorthbeneaththenarrow
businessAsokeroad, andRatchadaphisekroadending
onthesurfacenearthedepot.Worksof theNorthCon-
tractcontinuedfor4.5kmnorthalongRatchadaphisek
roadtoLadPhraoroadthenturnedwesttoChatuchak
Park and finally terminated beneath the Bang Sue
181
railway station. The18 stations of theproject areas
follows:
South Contract North Contract
1. HuaLamphongStation 1. ThiamRuamMit Station
2. SamYanStation 2. Pracharat BumphenStation
3. SilomStation 3. SutthisanStation
4. Lumphini Station 4. RatchadaStation
5. BonKai Station 5. LadPhraoStation
6. Sirikit CentreStation 6. PhahonyothinStation
7. Sukhumvit Station 7. MoChit Station
8. Phetchaburi Station 8. KamphaengPhet Station
9. RamaIX Station 9. BangSuStation
3 GROUNDCONDITIONS
Bangkok is situated on thesouthern part of thelow
lyingChaoPhrayaplain, whichextendsnorthfromthe
coast lineat theGulf of Thailanduptoapproximately
350kmand spans east-westward up to 150km. The
flat topography plain is covered with athick marine
clay layer, whichoverlies avery thick series of allu-
vial depositsof alternatingstiff tohardclayanddense
toverydensesandtogravel. Thethick soft claylayer
generallyextendsfromthegroundsurfacetoadepthof
12to15m.ThesoftclaywhichisknownasBangkok
soft clay has high water content (70120%), high
plasticity, lowstrengthandhighcompressibility.
Theshallowsubsoil of theupper 35mzoneisrela-
tivelyuniformandgenerallyconsistsof layersof soft
tomediumclay, stiff tohardclayandsand. Belowthis
shallowzone,alternatinglayersof stiff tohardclayand
densesandexist toagreat depth. Bedrock isfoundat
depths morethan 450m. Thetypical subsoil profile
for thefirst 50mdepthislistedasfollows:
MadeGround
BangkokSoft Clay
First Stiff Clay
MediumDenseClayeySand, SandyClayandSilty
Clay
VeryStiff SandyClay/First BangkokSand
SecondHardClay
SecondBangkokSand
Maconochie (2001) summarized the general soil
profileandpropertiesat theBangkok MRT ISP Blue
Line. The variation in soil profile along the align-
ment was observedprimarily intheVery Stiff Sandy
Clay/First Bangkok Sandlayer andthesoilsimmedi-
atelybelowit. Figure1showthesoil profilealongthe
MRT alignment andat thestations.
Deep well pumping in Bangkok and its environs
hasreducedtheporewaterpressuresinthesandlayers
by approximately 23mfromtheoriginal hydrostatic
profile. The groundwater pumping has also created
regional subsidencethroughout Bangkok metropolis.
At thelocations of theproject, aperchedwater table
istypically encounteredinMadeGround. Belowthis
horizon, hydrostaticconditionsaregenerallyfoundto
adepth of approximately 8mto 10mdepending on
thelocation and thickness of First and Second Sand
layers. Typically, theupper fewmetres of sandlayers
underlying First Stiff Clay andVery Stiff Clay have
beendewatered.
4 STATIONCONSTRUCTION
TheMRT stationshadfollowingfeatures:
Typically,threelevelsof structure,withacentreplat-
formthatisfedbystairsandescalatorsbetweentwo
linesof columnsdownthemiddleof thestation.
Up to 230mlong and approximately 25 mwide,
excavateduptoadepthof 25mto30mbelowthe
groundsurface.
Theperimeterswereof diaphragmwalls,1.0mthick
and30mto35mdeepandsolidin-situreinforced
concrete slabs, typically 0.9mthick for the roof
slabs, 0.7mthick for the intermediate slabs and
1.75mthick for the base slabs, which were used
as the excavation support systemand permanent
structureslater on.
Therewerethreestackedstations, atSamyan, Silom
and Lumphini Stations, due to space constraints
causedbytheexistenceof thefoundationpilesof the
longroadflyover andawater transmissiontunnel at
thesebusyintersections.
Therewas asideplatformstationat BangSuSta-
tion, withtwolevelsonly, toaccommodatethetrack
alignment for futureelevatedextensionof theline
tothenorth.
TheremainingstationswereconstructedasCentre
IslandPlatformstations.
PracharatBamphenandSutthisanStationsincorpo-
ratedintersectionroadunderpassesontheroofsof
thestations.
Silom, LadPhraoandPhahonyothinStationswere
excavated underneath foundation piles of existing
road flyovers and thus, the station structures and
foundation were designed to support the flyovers
viacross-beamandunderpinningboredpiles.
Thetopdownconstructiontechniquewasadopted
for all stationbox excavations withdiaphragmwalls
andconcreteslabs as theexcavationsupport system,
which was later utilized as thepermanent structures
of the stations. The designs were made with an aid
of FEM analyses. Theexcavationdepths andthetoe
depthof thediaphragmwallsof all stationexcavations
182
Figure1. Soil profilealongMRT alignment andat thestations.
aresummarizedinFig. 2. It shouldbenotedthat the
ratioof thedepthof thewall embedment tothedepth
of excavationwas significantly different betweenthe
two contracts, primarily due to the difference in the
designcriteriaadoptedbythetwodifferent designers.
5 INSTRUMENTATIONDATA
The measurement data frominclinometers and sur-
facesettlement pointswerecompiledandinterpreted
toevaluatetheoverall performanceof thestationexca-
vations in Bangkok subsoil using diaphragmwalls.
The data were screened to preclude movements not
expresslyrelatedtotheexcavationandsupport instal-
lation, such as diaphragmwall construction, tempo-
rarydeckingworksandtheinitial 2.3mexcavationthat
involved driving sheet piles, backfilling and extract-
ingthesheetpilessubsequently.Adetailedstudyof the
instrumentationdataobtainedwasundertakenbycom-
paringobservedgroundmovement amongthestation
excavations. Factors that may result in such patterns
of datawasexaminedanddeduced, suchas:
Stationboxconfigurationanddimension
Constructionsequences
183
Figure2. Depthof excavationandtoedepthof D-wall.
Figure3. Modesof wall deflection.
Variationinsoil profileandproperties
Temporaryworksor presenceof structuresadjacent
toexcavationbox
5.1 Lateral wall movement
Ingeneral, most inclinometersall stationexcavations
showedthatthecantilevermodewasthemostpredom-
inant of wall deflectionshapeat theinitial excavation
stage, whilethebracedexcavationmodedevelopedin
thesubsequent stages as theexcavations weredeep-
ened. Figure 3 shows the three modes of deflected
shape of wall movement, which occurred at differ-
ent excavationdepths. Thecantilever modewasmost
commonduringthefirstexcavationstage. Thebraced
excavationmodewithbulgeinsoftsoil prevailedatthe
second and third excavation stages. This mode con-
tinued to dominatethepattern of lateral movements
for NorthContract, butdataof SouthContractexhibit
that the braced mode with bulge in the underlying
NORTH CONTRACT
0
5
10
15
20
25
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Maximum Lateral Wall Movement/Excavation Depth, d
Hmax
/H
E
x
c
a
v
a
t
i
o
n

D
e
p
t
h
,

H

(
m
)
Cantilever mode (H=1.6m-4.0m) Braced mode-soft soil (H=6.5m-10.8m)
Braced mode-stiff soil (H=12.4m-16.7m) Braced mode-stiff soil (H=20.0m-32.6m)
SOUTH CONTRACT
0
5
10
15
20
25
30
35
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Maximum Lateral Wall Movement/Excavation Depth, d
Hmax
/H
E
x
c
a
v
a
t
i
o
n

D
e
p
t
h
,

H

(
m
)
Cantilever mode (H=1.6m-4.0m) Braced mode-soft soil (H=6.5m-10.8m)
Braced mode-stiff soil (H=12.4m-16.7m) Braced mode-stiff soil (H=20.0m-32.6m)
Figure 4. Normalized maximum wall deflection versus
excavationdepth.
stiff soil layers was more predominant. This differ-
entbehavior maybeattributedtothedifferenceinsoil
profiles. In addition, South Contract had three deep
stationexcavationsof verticallystackedplatforms.
Themaximumlateral wall deflectionof the18sta-
tionexcavationswereintherangeof 1047mm. The
monitoring data also showed that there were signif-
icant variations in the shape and magnitude of the
lateral wall movementwithinsomeof thestationexca-
vations. Thevariationmay beattributedto anumber
of factors including the variation in soil profile and
properties over theplan areaof thestation, adjacent
temporarysurfaceworks, andconfinement fromroad
pavements and buried utility structures. The area of
theexcavationwas quitelarge(about 25mwideand
200mlong). For the three stations adopted for the
detailed analysis in this study, the variations in the
maximumlateral wall movementswereasfollow: 28
38mmfor the deepest SilomStation, 15 to 40mm
for Sirikit Station, and 1827 for ThiamRuamMit
Station. Figure4summarizestherangesof thenormal-
izedmaximumlateral wall movementwithexcavation
depth(
Hmax
/H) versusexcavationdepthsrecordedby
all inclinometers at the 18 stations. The plots show
that inthefirst stagecantilever modeexcavationthe
value of
Hmax
/H was as high as 1.60. The maxi-
mumvalue of
Hmax
/H decreased with the increase
in the excavation depth in the braced modes of the
184
Figure5. Normalizedsettlementversusdistancefromwall.
walls. Thedecreaseinthevaluewithdepthwasdueto
thechangeinsoil conditionat theexcavationbottom
fromsoft clay to stiff clays as the excavation depth
increased. Whentheexcavationdepthreachedstiffer
soils,
Hmax
/H decreased and the range of the value
becamenarrow. Whentheexcavationbottomwasstill
inthesoft clay layer, themaximumvalueof
Hmax
/H
wassmaller than0.5. Thevaluewassmaller than0.2
whentheexcavationsweredeeper, instiff clays.
5.2 Ground surface settlement
The maximum ground surface settlement was
observedat 58mmat BonKai Stationand75mmat
Pracharat Bumphen, in South and North Contracts,
respectively. Figure5showsplot of theratioof maxi-
mumsettlement toexcavationdepth(
vmax
/H) versus
thedistancefromexcavationnormalizedbytheexca-
vation depth (D/H). For shallow excavations under
cantilever modeof wall deflection, thezoneof ground
settlement may extend up to D/H of 7 to 10. As for
the excavation depths while the wall deflection was
under thebracedmodewithbulgeinsoft clay layer,
the
vmax
/H may extend up to D/H of 7. For deeper
excavations while the wall was deflected under the
modeof bulgeinstiff clay layer, thezoneof ground
settlement may extend up to adistanceof D/H of 4
and
vmax
/H valuemay beas highas 0.35Under the
bracedmodeinsoftclay,
vmax
/Hvaluemayreach0.55.
For shallowexcavationunder thecantilever modethe
value of
vmax
/H can be higher. The characteristics
of ground surface settlement behind the excavation
withdiaphragmwallscanbecategorizedaccordingto
themodeof wall deflectionof excavationdepth. The
zonesof groundsettlementforthethreemodesof wall
deflection in Bangkok soil are marked in figure. In
addition, thethreecategoriesof groundmovement in
bracedwall excavationforflexiblewalls(sheetpilesor
soldierpiles) suggestedbyPeck(1969) arealsoshown
intheplot. Thelevel of groundsettlement inexcava-
tionwithdiaphragmwall inBangkoksoftsoil ismuch
smaller thanthatpredictedbyPeckschartfor flexible
wall. However, theinfluencezones of ground settle-
ment werewider thanthosesuggestedbyPecks. This
findingcanbeusedasageneral guidelinefor predic-
tiongroundsurfacesettlement fromdeepexcavation
withdiaphragmwallsinBangkoksubsoil condition.
6 PREDICTIONOF MOVEMENTS
For the design of the MRT station excavations of
both contracts, FEM analysis were made to deter-
minegroundmovement andforcesonthediaphragm
walls andbracings. Theanalyses mainly utilizedlin-
earelasto-plasticMohr-Coulombsoil parameters.The
parameters wereobtainedfromthesoil investigation
andtestingprogrammadefor eachstationexcavation.
Triaxial compression tests as well as pressuremeter
tests wereconducted to determinethevalues of soil
modulusfordesignanalysis.Theinstrumentationdata
providedvaluableinformationto evaluatetheappro-
priateness of thesoil model and thesoil parameters
usedinthedesigncalculation. Inthisstudy, monitor-
ingdatafromthreerepresentativestationexcavations,
i.e. Silom, Sirikit and Thiam Ruam Mit Stations,
were examined in details and suitable soil parame-
ters were back-calculated using a continuum FEM
analysis. Computer codePLAXIS2Dwasadoptedin
the study. Effective stress strength parameters were
adoptedintheundrainedanalysis.Thedrawdownphe-
nomenonof thepiezometric levelswasconsideredin
thesimulation.
Silom Station was the deepest excavation in
Bangkok to date (32.6m deep), with a vertically
stackedplatformsthusithadfour levelsof slabbelow
the roof. The station was designed to underpin the
existing flyover roadway running over the station
length. A densesandlayer of thefirst Bangkok Sand
was encountered fromdepth 8.5mabove the final
excavationlevel.Hence,theexcavationrequireddewa-
tering. Thediaphragmwall wastoedintotheSecond
Sandlayer.
Sirikit Stationhadthetypical configurationof the
centreislandplatformwiththreelevelsbelowtheroof
slab. The first stage and final excavation depths of
185
3.65mand23.6mrespectively, whichweresimilar to
majorityof other stations.
ThiamRuamMit Stationwasselectedbecausethe
soil profilewasslightlydifferent fromthosefoundin
thefirsttwostations. Theareahadathicker FirstStiff
Claylayer withlensesof clayeysands. Inaddition, the
first stageexcavationwasveryshallowwithroof slab
wasonlyat 1.8mdepth.
Theback-calculationusingthelateral groundmove-
ment datafromtheexcavations of thethreestations
suggestedthesuitableundrainedsoil modulusparam-
etersasfollows.
Soft andMediumClay : E
u
=500C
u
kN/m
2
First Stiff Clay : E
u
=700N
60
kN/m
2
ClayeySandandSilty/ : E
u
=900N
60
kN/m
2
SandyClay
SecondHardClay : E
u
=1600N
60
kN/m
2
ThirdHardClay : E
u
=2500N
60
kN/m
2
whereC
u
isthecorrectedfieldvaneshearstrengthand
N
60
isthecorrectedSPT N valueaccordingtoLiao
andWhitman(1986).
These back-calculated values of soil modulus are
higher than those commonly obtained fromconven-
tional laboratorytriaxial tests. It reflectsthemodulus
valuesatlowstrainlevel whichwouldbethedominat-
ingresponseof soil intheexcavationproblem(Mair,
1993). The soils would be mainly under unloading
conditionof stresses.
Insimilar early studies, Phienwej andGan(2003)
andTeparaksa(1999) both proposed thesamemod-
ulus parameter of thesoft clay as E
u
=500c
u
. While
for stiff clay, thevalueof E
u
=1200C
u
and 2000C
u
were suggested, respectively. Based on the relation-
shipof C
u
=0.6N
60
kN/m
2
typicallyusedforBangkok
subsoil, theparametersareequivalent toE
u
=720N
60
and1200N
60
kN/m
2
,respectively.Theback-calculated
values from this study were comparable to those
suggestedbyPhienwej andGan(2003).
7 CONCLUSIONS
Thefollowingconclusionscanbedrawnfromthestudy
of thewall andgroundmovementsintheexcavationof
thestationsof thefirst BangkokMRT underground.
Threemodesof deflectedshapesof thewall move-
ment were observed at different ranges of exca-
vation depth. Mode 1: Cantilever mode (H=1.6
m4.0m), Mode2: Bracedmodewithbulgeinsoft
clay layer (H=6.5m11m), and Mode3: Braced
modewithbulgeinstiff soil (H=12.4m32.6 m).
Themaximumlateral wall movement (
Hmax
) was
smaller than47mminbothcontracts. Thenormal-
izedwall deflection,
Hmax
/Hinthecantilevermode
of movement was as high as 1.60, while it was
reducedtonomorethan0.60and0.40inthelatter
stagesof excavationwhenthewall deflectiondevel-
opedinthebracedmodewithbulgeinsoftclayand
bracedmodewithbulgeinstiff soil, respectively.
The maximumground surface settlement (
Vmax
)
was58mmfor SouthContractand75mminNorth
Contract. Thenormalizedmaximumgroundsettle-
ment with excavation depth,
vmax
/H, was smaller
than 0.55 and 0.35 for Modes 2 and Mode3 wall
deflection, respectively. The normalized distance
fromexcavationof thezoneof groundsettlement,
D/H, varied from7.0 and 4.0 for the two modes
of wall deflection. In the initial excavation stage
of cantilever mode, thevalues of bothnormalized
settlement anddistanceof groundmovement were
higher that thoseinthebracedmodes.
Back-calculation of soil moduli of different soil
layers using monitoring data fromthree selected
stations showed higher values than those com-
monly obtained for conventional laboratory tests.
Thevaluesare: SoftandMediumClay: E
u
=500c
u
,
First Stiff Clay E
u
=700N
60
kN/m
2
, Clayey Sand
andSilty/SandyClayE
u
=900N
60
kN/m
2
, Second
HardClay E
u
=1600N
60
kN/m
2
, ThirdHardClay
E
u
=2500N
60
kN/m
2
REFERENCES
Hooi, K.Y. 2003. Ground Movements Associated with Station
Excavations of the First Bangkok MRT Subway. Master
Thesis, AsianInstituteof Technology, Bangkok.
Liao, S. &Whitman, R.V. 1986. OverburdenCorrectionFac-
torforSPT inSand. J ournal of Geotechnical Engineering.
American Society of Civil Engineers 112(3): 373377.
Mcconochie, D. 2001. Geotechnical Completion Report
MRTA Chaloem Ratchamongkhon Line. A CSC Report
submittedtotheMRTA.
Mair, R. 1993. Developments in geotechnical engineering
research: application to tunnels and deep excavations.
UnwinMemorial Lecture1992,Proceedings of Institution
of Civil Engineers, Civil Engineering, 93, Feb: 2741
Phienwej, N. & Gan, C.H. 2003. Characteristics of
GroundMovements inDeepExcavations withConcrete
Diaphragm Walls in Bangkok Soils and their Predic-
tion. Journal of The Southeast Asian Geotechnical Society
34(3): 167175.
Teparaksa, W.,Thasnanipan, N. &Tanseng, P. 1999.Analysis
of Lateral Wall Movementfor DeepBracedExcavationin
Bangkok. Proc. of AIT 40thAnniversary Conference,AIT,
Bangkok, Thailand.
186
Numerical modellingandexperimental measurementsfor aretainingwall
of adeepexcavationinBucharest, Romania
H. Popa, A. Marcu& L. Batali
Technical University of Civil Engineering, Geotechnical and Foundations Department, Romania
ABSTRACT: Althoughcivil engineers disposeof various calculationmethods for retainingstructures, none
of themhavedefinitelyimposeditself, eachonebringingitsownbenefitsor limitations. FiniteElementMethod
(FEM) offers the benefits of complex models allowing taking into account the majority of soil structure
characteristic parameters. However, theexperienceshows that thedifferences between theexperimental and
thecalculationresults areoftenquiteimportant. Thepaper presents thecasehistory of adiaphragmwall for
adeepbasement of anewbuildingincentreof Bucharest. Nearby thenewbuildingthereis anancient cathe-
dral historic monument. Thedeepexcavationisalsoneighboringat different distanceswithanother existing
buildingsandaheavy traffickedroad. All theseconditionsledtochoosethetop-down technology inexecu-
tionof thebasement. Thenumerical results obtainedby FEM arecomparedwiththemeasurements recorded
duringtheconstruction. Thedifferences betweentheobtainedvalues (displacements) arecomprisedbetween
15%and 75%, depending on the enclosure sides. The main factors leading to these differences are the soil
parameters.
1 INTRODUCTION
Theeffect of deepexcavations onneighboringstruc-
tures can become important and therefore special
measureshavetobetakeninthedesignandmonitoring
of retainingwalls.
Calculation of such structures must be based on
methodstakingintoaccount thesoil structureinter-
action and, with this respect good soil knowledgeis
indispensable.
Paper presents the case of a diaphragmwall for
a deep basement in the very centre of Bucharest.
The basement is developed on 4 underground lev-
els andneeds anexcavationof about 15mdeep. The
groundwater level is at about 6mdepth. Theground
iscomposedof alluviumslayerscomprisingmedium
soft silty clays, as well as fine to coarse medium
densesand. Near thenewconstruction, at about 6m
distance, there is an ancient cathedral, classified as
historical monument. As well, thepit has onanother
sidesomebuildings, whileon theother two sides it
is delimitedby aheavy traffic road. All thesecondi-
tions led to choosethetop down technology for
buildingtheinfrastructure, inorder tohaveminimum
deformations and displacements of the wall, so that
the integrity of the neighboring buildings not being
affected.
2 WORK ANDSITE DESCRIPTION
2.1 Site and geometrical characteristics
Thenewbuildingislocatedatthecentreof Bucharest,
next to theRomano Catholic Cathedral St. J oseph.
Figure1presentsaphotoof thesite.
Asit canbeseenonthephoto, ontheWesternside
thepitisveryclosetotheSt. J osephCathedral (about
Figure1. Locationof thedeepexcavation.
187
Figure2. Crosssectionthroughthediaphragmwall.
6mdistance), ontheSouthernsidethereisabuilding
complexatabout10mdistance, whileonthetwoother
sidesthereisapublicroadwithheavytraffic.
Figure 2 shows a cross section through the
buildings infrastructure, presenting also the ground
lithology.
Thediaphragmwall ismadeof panels80cmthick
and21mdeep. Theembedmentdepthhasbeenestab-
lished considering the wall stability and excavation
bottom imperviousness. Dewatering has been per-
formed only inside the enclosure, the groundwater
level outsidebeing left unchanged to avoid undesir-
ablesettlements of theground around thepit dueto
dewatering, whichwouldbeinadditionto theinher-
ent settlementsduetoexcavationinsidetheenclosure
andtotheerectionof thenewstructure, whichcould
affect especiallySt. J osephCathedral.
2.2 Geotechnical characteristics of the ground
Thegeotechnical parametersarespecificforBucharest
area, whichischaracterizedbyalluviumsoils.Thetri-
axial testsperformedwithimposedstresspathallowa
direct determinationof theshear strengthparameters
(, c), of thesecantmodulus(E) andof theearthcoef-
ficient at rest (k
0
) for theclayey layers; for thesands
theseparametershavebeenestablishedbasedonSPT
tests.
Table1. Designvaluesfor thegeotechnical parameters.
Thickness, E,
K
0
Stratum m kN/m
3
MPa (
) kPa
Clay(1) 4.00 19 25 25 30 0.7
Sand& gravel 5.50 20 40 38 0 0.4
Clay(2) 7.50 20 50 22 50 0.7
Siltyclayey 2.00 20 75 28 20 0.5
sand
Clay(3) 4.00 20 75 22 50 0.4
Finesand 7.00 20 75 36 0 0.4
Thegroundlithology andthedesignvalues of the
geotechnical parameters are shown in table 1 (level
0.00mrepresentsgroundlevel).
Thegroundwater level is +74.0mand, according
to the site investigations, it can vary with 1.00m.
A secondaquifer, confined, has beenfoundbetween
+63.0+61.0m(withinthesandylayer).
3.1 Numerical model
Numerical modeling has been performed using 2-D
FEM, the model having 1933 elements and 5866
nodes.
For thesoil, aperfectelasto-plasticconstitutivelaw
hasbeenused,withMohr Coulombcriteria,usingthe
geotechnical parameters issued fromlaboratory and
insitutests.
3.2 Calculation stages
Calculationshavebeenorganizedin6stages, follow-
ingthetechnological phases:
phase 0 initialization of the stress state in the
ground;
phase 1 excavationdowntothelower level of the
first floor slab; executionof thefirst floor slab;
phase 2 excavationbelowthefirstfloorslabdown
tothelowerlevel of thesecondfloorslabanddewa-
teringinsidetheenclosure; executionof thesecond
floor slab;
phase 3 excavation between floor slabs no. 2
and3; executionof floor slabno. 3;
phase 4excavationbelowthethirdfloorslabdown
tothefinal level (15.00m).
3.3 Results
Figure3presents theevolutionof thehorizontal dis-
placement of thewall as afunction of theexecution
stages.
188
Figure3. Horizontal displacementsfunctionof calculation
phases.
Figure4. Bendingmoment functionof calculationphases.
Figure5. Shear forcefunctionof calculationphases.
Table2. Floor slabreactionforces.
Force, kN/ml
Floor 1 Floor 2 Floor 3
Phase1
Phase2 142.2
Phase3 2.3 456.5
Phase4 6.7 249.2 729.2
Accordingto thecalculations, themaximumhori-
zontal displacementsof thewall areof about 15mm,
at 15mdepth. Duetothetop-down technology, the
shapeof thedisplacementcurvesshowsgreatervalues
inthelower part of thewall. Thedisplacementsof the
upper part arepracticallyblockedbythealreadybuilt
floor slabs.
Theupper maximumhorizontal displacements are
estimatedtobeof 56mm.
Bendingmomentandshear forcegraphsareshown
figures4and5, respectively.
Thereaction forces on thebasements floor slabs
areshowntable2.
4 MEASUREMENTS
In order to record theinfluencethedeep excavation
and, moreover the whole new building, has on the
189
Figure6. Experimental measurements.
neighbouring structures, a monitor of the displace-
mentshasbeenperformed.
Theretainingwall wasequippedoneachsidewith
inclinometers for measuring thehorizontal displace-
mentsand, thus, thewalls deformation.
Markswereinstalledonthesurroundingbuildings
to monitor their settlements. As well, extensometers
were installed, just behind the wall for measuring
grounds settlements.
Another monitoring concerned the groundwater
level outside the enclosure. For this purpose, wells
weredrilledandequippedaspiezometers, locatedon
eachsideof theenclosure. Themeasurementsshowed
that thedewateringworks insidetheenclosuredidnt
affect thegroundwater level outside.
Figure6showsthemonitoringequipment usedfor
theSt. J osephCathedral side.
It canbeseenthat, inorder toreducefurthermore
therisksof anegativeinfluenceof theretainingwall on
theCathedral, betweenthesetwo first astabilization
wall wasbuilt usingcement-basedinjections.
Figure6showsalsothemeasuredlateral displace-
mentsof thewall duringthelastexcavationstage(stage
4). It canbeseenthat theshapeof thedisplacement
190
curvecorresponds to theoneobtained by numerical
calculation. But the values are much less than the
estimatedones.
So, at theupper part of thewall, themaximumdis-
placementsareof about2mm,representingabout50%
of theestimatedones, whileatthelowerpartthediffer-
encesaremoreimportant, themeasuredvaluesbeing
only25%of thecalculatedones.
Thisimportantdifferencebetweencalculationsand
measurementscanbeduealsototheprotectioninjec-
tion screen locatedon this sideof theenclosure. On
the other sides, where no such protection has been
installed, themaximumwalls displacementswereof
about13mm, beingquiteclosedtotheestimatedvalue
(85%).
Anyway, estimations arestill higher thanthemea-
suredvaluesfor all excavationstages.
Concerning the ground settlement behind the
retainingwall (at about 1mdistance), extensometers
showed a maximumvalue at ground level of about
7mm. Thesettlement evolutionversus thedepthcan
alsobeseeninfigure6andonecannotethatitbecomes
negligibleat about 23mbelowwalls toe.
MarksfixedontheSt. J osephCathedral indicateda
maximumsettlementof 0.7mm, itsintegritynotbeing
endangered.
Fromthis point of view it can also be noted the
beneficial roleof theinjectionscreen, thedifference
of settlementsononesideandtheother of thescreen
beingsubstantial.
Concerningtheotherneighboringbuildingslocated
atabout10mdistance, maximumsettlementsof about
3mmwererecorded, insignificant for their stability.
5 CONCLUSIONS
Retaining structures imply complex soil structure
interactionphenomena. A correct estimationof their
behavior is possible only by using numerical mod-
els, allowing a complex modeling of the system
formedby theretainingwall, foundationgroundand
neighbouringbuildings.
Evenwhensuchmethods areused, theresults can
present significant differences fromthe real behav-
ior.Thereasonsforthesedifferencesaremany, among
them:
incertituderegarding thegeotechnical parameters
usedfor thecalculations, especiallywhencomplex
constitutivelawsareusedfor theground;
difficulty inestimationof theinitial stressstatein
the ground, taking into account its lithology, the
presenceof neighbouringstructures, theexecution
of theretainingwall itself etc.;
complexity of thenumerical model itself, consid-
eringall impliedparameters;
three-dimensional behaviour of the retaining
structure.
In order to obtain reliable results using numeri-
cal modelingit is important to calibrateandvalidate
themodel basedonexperimental measurements per-
formedonsimilar structures andinsimilar sitecon-
ditions. The experience in such modeling is also an
important aspect.
REFERENCES
Marcu, A., Popa, H., Marcu, D., Coman, M., Vasilescu, A. &
Manole, D. 2007. Impactdeepexcavationsonneighboring
buildings, National ConferenceAICPS, 1J une2007(in
Romanian).
Marcu, A. &Popa, H. 2004. Calculationsandmeasurements
of deformations and displacements of a retaining wall
for adeepexcavationandof theneighboringstructures.
10th National Conferenceof Soil Mechanics and Foun-
dation Engineering, 1618 September 2004, Bucharest,
Romania, pp. 311322(inRomanian).
Popa, H. 2002. Contributions to the study of the soil
structure interaction in case of underground structures,
PhD thesis, Technical University of Civil Engineering
Bucharest, p. 311.
191
3Dfiniteelement analysisof adeepexcavationandcomparison
withinsitumeasurements
H.F. Schweiger
Computational Geotechnics Group, Institute for Soil Mechanics and Foundation Engineering, Graz
University of Technology, Austria
F. Scharinger & R. Lftenegger
GDP Consulting Engineers, Graz, Austria
ABSTRACT: Thepaper describes theanalysis of adeep excavation project in clayey silt in Salzburg. The
excavationwassupportedbyadiaphragmwall, ajetgroutpanel andthreelevelsof struts. Becauseof insufficient
informationonthematerial propertiesof thejetgroutpanel thestiffnessof itwasvariedinaparametricstudy.The
effectof takingintoaccountthestiffnessof acrackeddiaphragmwall onthedeformationswasalsoinvestigated.
Insomeof the3Dcalculationsanon-perfectcontactbetweendiaphragmwall andstrutwassimulatedbymeans
of anon-linear behaviour of thestrut. Theevaluationof theresultsandcomparisonwithinsitumeasurements
showedthatanalyseswhichtookintoaccountthereducedstiffnessof thediaphragmwall duetocrackingachieved
thebest agreement withthemeasurements. Furthermoresettlementsof buildingscouldbebest reproducedby
thethree-dimensional model.
1 INTRODUCTION
Softsubsoil depositsinAustriaaremainlyfreshwater
deposits, sedimentedinthepost-glacial lakesafter the
boulder periods. Thesedeposits areknown as lacus-
trineclays onthefoothills of theAlps. Oneexample
for awidespread lacustrineclay deposit is thebasin
of Salzburg, where the city of Salzburg is situated
on subsoil sediments, which partly showathickness
upto70m, calledSalzburger Seeton, whichcanbe
classifiedasclayeysilt.
Inthedesignstageof deepexcavationsinsuchprob-
lematic soils finiteelement calculations areauseful
tool toobtainreasonablyrealisticpredictionsof defor-
mationsexpected. Inpractical engineering2D-models
arestill prevailing, but3D-model becomeincreasingly
attractive. It will beshown, andthis is themainpur-
poseof thispaper, that thebest overall matchwithin
situmeasurements,inparticularwithrespecttosurface
displacements behindthewall, is achievedwith 3D-
models. If only wall deflectionisconsideredalso2D
analysesshowreasonableagreement. Themechanical
behaviourof thesoil ismodelledwithanelasto-plastic
constitutivemodel, namelytheHardeningSoil model
as implemented in the finite element code Plaxis
(Brinkgreve2002). For theproject theclass A 2D
analysis predictedtheoverall deformationbehaviour
withsufficientaccuracyfromapractical pointof view,
but amoredetailedcomparisonwithinsitumeasure-
mentshasbeenmadeafter constructioninvolving3D
finiteelementanalyses.Furthermoresomedetailswith
respecttothestruttinghavebeenchangedduringcon-
struction which havenot been taken into account in
theoriginal analysis.
The input parameters for the constitutive model
have been determined not solely from site investi-
gations but also fromprevious experience of finite
element analyses under similar conditions (see e.g.
Schweiger & Breymann2005).
Inthefollowingabrief descriptionof theproblem
will be provided together with the material parame-
ter used. The different assumptions with respect to
modellingthediaphragmwall andthejet grout panel
arediscussed. Finallyresultsfromvarious2Dand3D
analyses arecomparedwithinsitumeasurements of
wall deflectionandsurfacedisplacements.
2 PROBLEM DESCRIPTIONANDMATERIAL
PARAMETERS
A cross section of the excavation with strut levels
and final excavation depth is shown in Figure 1. In
plantheexcavationis roughly square, approximately
193
Figure1. Crosssectionof excavationandstrut levels.
1920m, whichof coursemustraisedoubtswhether
a2Danalysisisatall appropriateinthiscase.Attention
ispaidtothefact that ajet grout panel just belowthe
final excavation level has been constructed to act as
lateral support. This has beenconstructedbeforethe
startof theexcavationandallowedexcavationwithout
installing afourth strut level. Groundwater lowering
insidetheexcavationwas achievedby vacuumwells
(commonly usedinSalzburg) whichextendedbelow
theexcavationlevel inorder toreduceuplift.
The construction sequence is closely reflected in
the analysis. Starting from the initial stress state
(K
0
=0.55 for all layers) and theloads of thefoun-
dations of theneighbouring buildings (80kN/m
2
for
the Novotel, 200 and 250kN/m
2
for the strip foot-
ings of Object 24) the wall and jet grout panel
havebeenintroducedwish-in-place. Thenexcavation
steps, groundwater changes andinstallationof struts
have been modelled in a step by step analysis. Soil
behaviour below20mis assumedto beundrained,
above20m, duetothepresenceof thinsandylayers,
asdrained.
2.2 Material parameters
The soil parameters used in the analysis for the top
soil layer (04mbelow surface) and the clayey silt
aresummarizedinTables1and2. Asmentionedpre-
viously, parameter determinationisnot onlybasedon
siteinvestigationsandlaboratoryexperimentsbutalso
fromexperienceof back analysesof other deepexca-
vations in Salzburg. Therfeore soil parameters have
not been varied in this study. In Table 1 E
50
, E
oed
and E
ur
are the reference stiffness in primary load-
ing (for deviatoric and oedometric stress paths) and
unloading/reloadingrespectively.
Theaxial stiffnessof thestruts(Table3) differsfor
thethreelevels, thematerial behaviour isassumedto
Table1. Stiffnessparametersfor soil layers.
E
50
E
oed
E
ur
m p
ref

ur
(MPa) (MPa) (MPa) (kPa)
Soil layer 3 3 12 0.0 40 0.2
(04m)
Clayeysilt 37.6 37.6 150.4 0.30 100 0.2
Table2. Strengthparametersfor soil layers.
c
(kPa) (
) (
)
Soil layer (04m) 5 28 0
Clayeysilt 1 30 26 0
Table3. Axial stiffnessof struts.
EA spacing
(kN) (m)
Strut level 1 3.234E6 3
Table4. Parameters for wall, jet grout panel and founda-
tions.
E R
inter
UCS
(kN/m
2
) (N/mm
2
)
Diaphragmwall 2.9E7 0.2 0.7 18.8
J et grout panel 5.0E5 0.2 0.7 2.25
Foundations 3.0E7 0.2 0.7
belinear elastic. Table4lists thebasic set of param-
eters used for diaphragmwall, jet grout panel and
the foundation structures of Novotel and Object 24.
Inthe2D analysesaMohr-Coulombfailurecriterion
hasbeenusedfor wall andjetgroutpanel whereasthe
cohesionwaschoseninsuchaway toobtaintheuni-
axial compressivestrength(UCS) aslistedinTable4,
assuming
=45
.Tensioncut-off wassettoUCS/10.
R
inter
denotes thereductionof soil strengthto model
wall friction. In the 3D analyses the wall was elas-
tic andstiffnesswaseither assumedtocorrespondto
uncrackedconditions or crackedconditions. The
stiffnesspropertiesof thejetgroutpanel havebeenvar-
iedbecauseof thesignificantuncertaintyinobtaining
reliablevaluesfor theinsitustiffnessof suchpanels.
194
Figure2. Locationof pointsusedfor comparison.
3 INSITU MEASUREMENT PROGRAMME
Theexistenceof structuresintheclosevicinityof the
excavationrequiredacareful observationof deforma-
tions duringconstruction. Therefore, about 30settle-
mentgaugeswereinstalledtomonitor thesettlements
outside the excavation, in particular of the adja-
cent buildings. In addition, four inclinometers were
installedin thediaphragmwalls in order to measure
thehorizontal deflectionof thewall inall construction
stages.Twoof themwerelocatedapproximatelyinthe
cross section chosen for the 2D analysis, i.e. along
thecentrelineof theexcavation. Figure2depictsthe
pointschosenfor thecomparisonof measurementand
analysisfor settlements.
4 NUMERICAL MODELS
As mentioned previously 2D and 3D analyses have
beenperformedusingPlaxis2DandPlaxis3DFoun-
dations.The2Dmodel consistsof approximately2,300
15-noded elements (Figure 3) and the 3D model of
approximately11,00015-nodedwedgeelements(Fig-
ure 4). Lateral boundaries are fixed in horizontal
direction and the bottom boundary in vertical and
horizontal direction in both models. It can be seen
that the 3D mesh is much coarser as compared to
the2D meshbut studiesperformedonthe3D model
showedthat ameshwithmorethan20,000elements
resultedinonlymarginal differencesindisplacements.
However, bendingmomentsaremoresensitivetodis-
cretisationandastabilityanalysiswouldcertainlynot
yieldcorrectresultswiththemeshadoptedfor the3D
analyses.
5 RESULTSOF 2DMODEL
Four different analyseshavebeenperformedwiththe
2Dmodel:
Variation 1 (V1): Wall and jet grout panel elastic
withelasticpropertiesaccordingtoTable4.
Figure3. 2Dfiniteelement mesh.
Figure4. 3Dfiniteelement mesh.
Variation 2 (V2): Diaphragm wall modelled as
elastic-perfectly plastic material with UCS as given
inTable4.
Variation3(V3): V2andincreaseof stiffnessof jet
grout panel byafactor of 3.
Variation4(V4):V3andincreaseof tensioncut-off
indiaphragmwall byafactor of 2.
Figures 5 and 6 compare the deflection of the
wall for the final construction stage for all four
analyses with the measurements obtained fromthe
inclinometers.
Itfollowsthatthedifferentassumptionsmadehave
littleinfluenceontheresults intheupper part of the
wall becauseinthispartthedeformationsaregoverned
by thestruts. Results for theright wall comparewell
withmeasurementsintheupper part, for theleft wall
thisisnot thecase. For thelower part onlyV3andV4
produceareasonably match and it turned out that it
isdifficulttoobtainthewall curvatureasmeasuredat
thelocationof thejet grout panel.
Figures7to9showacomparisonof calculatedand
measuredvertical displacements at various points on
thegroundsurface.Thetwosetsof squaresineachdia-
gramrepresentpairsof settlementgaugeswhicharein
closedistancetothepointspickedfromthenumerical
analysisatvariousstagesof construction(thedatesare
givenwithinthediagram, theaxisrepresentscalcula-
tion steps, representing the progress of construction
195
Figure5. 2Danalysis: wall deflection left wall.
with time). Only for point H a reasonable agree-
mentbetweencalculationandmeasurementscouldbe
achieved, although onehas to mention that absolute
values are very small, with about 10 mmas maxi-
mumsettlement. InpointI theanalysispredictsheave
whereassettlementshavebeenmeasured, butforpoint
E settlementsareoverpredicted.
6 RESULTSOF 3DMODEL
Inthissectionresultsfrom3Danalysesarepresented.
These analyses have been performed because the
geometry of theexcavation(approximately quadratic
inplanview)andalsopartof thebracingsystem(struts
across thecorners of theexcavation) cannot beade-
quately representedinplanestrainconditions. Inthe
first series of analyses emphasis has beenput onthe
stiffnessof thediaphragmwall and3different calcu-
lationshavebeenperformed: thefirst assumedlinear
elasticbehaviour for thewall withastiffnessassigned
representingstiffness I (uncrackedconditions), the
secondoneassumedstiffnessII(crackedconditions)
andthethirdoneintroducedanon-linearbehaviourby
Figure6. 2Danalysis: wall deflection right wall.
Figure7. 2Danalysis: surfacedisplacements Point I.
meansof apre-definedcurverelatingallowablebend-
ingmomentstothecurvatureof thewall. InFigures10
and11thesearedenotedwithZ1, Z2andnon-linear
respectively. It has been observed already in the2D
analysesthattheassumptionfor thestiffnessof thejet
groutedpanel has asexpected asignificant influ-
ence on the curvature of the wall. The inclinometer
measurementsindicatethatthelower value obtained
fromlaboratoryexperiments seemstounderestimate
thesupportinsitu. Thishasbeenconfirmedalsofrom
196
Figure8. 2Danalysis: surfacedisplacements Point H.
Figure9. 2Danalysis: surfacedisplacements Point E.
Figure10. 3Danalysis: wall deflection left wall.
Figure11. 3Danalysis: wall deflection right wall.
3D analyses and thereforeonly results assuming the
highstiffness(1,500MPa, asusedinV3andV4of the
2Dcalculations) arepresentedinthefollowing.
Thecomparisonof horizontal displacements (Fig-
ures10and11) clearlyshowtheeffect of varyingthe
stiffness of the diaphragmwall in the unsupported
zone whereas the assumption of cracked stiffness
is closer to the measured curvature than the analy-
siswithhighwall stiffness, at least for theright wall.
In theupper part theinfluenceof varying wall stiff-
nessismuchlesspronouncedbecausethebehaviouris
dominatedbythestruts, however predictedhorizontal
displacementsarelessthanmeasured. Thenon-linear
model is, not surprisingly, between the two extreme
cases.
Finally, after some discussion with the designer,
an additional analysis was performed assuming a
non-perfect connection of struts and wall, i.e. it was
assumedthat thereis animperfectionbeforethefull
support of thestrut can bemobilised. This has been
achieved by a nonlinear model for the strut which
197
Figure12. Comparison2D-3Danalysis left wall.
results ina0.25mm/mgap tobeclosedbeforethe
full support is activated (this variation is denoted as
V7 in the following diagrams). The consequence of
thisfollowsformFigures12and13, inwhichresults
fromthe2Danalysis(Variation4) arealsoplottedfor
comparison. For theleftwall thecurvatureattheposi-
tion of the grout panel is still not in full agreement
with measurements but the upper part corresponds
much better than in previous analyses. For the right
wall thecurvatureandtheupper part arenowinrea-
sonablegoodagreement withmeasurements (for the
right wall the2Danalysisisalsoingoodagreement).
Figures14to17plot settlementsat variousobserved
points. It is immediately noticedthat incontrast to
the2D model the3D analysis predicts settlements
alsoforPointI, althoughtheyarestill slightlyloweras
comparedtomeasuredvalues. PointHcorrespondsin
thesensethatmeasuredandcalculatedsettlementsare
almostzero. PointE showsslightlyhigher settlements
for later stagesof constructionthanmeasuredandthe
sameholdsfor point G.
Figure13. Comparison2D-3Danalysis right wall.
Figure14. Comparison2D-3Danalysis: Point I.
Figure15. Comparison2D-3Danalysis: Point H.
198
Figure16. Comparison2D-3Danalysis: Point E.
Figure17. Comparison2D-3Danalysis: Point G.
7 CONCLUSIONS
Results form2D and 3D finite element analyses of
a deep excavation have been compared to in situ
measurements. The excavation is supported by a
diaphragmwall, 3rowsof strutsandajet grout panel
locatedjustbelowthefinal excavationdepth.Inapara-
metricstudythestiffnessof thediaphragmwall andthe
jetgroutpanel havebeenvaried.Thestudyshowedthat
a 2D analysis would reasonably predict wall deflec-
tions(inparticular for theright wall) but if bothwalls
and vertical displacements of all surface points are
consideredthe3Danalysisproducesasomewhatbetter
overall agreement withthemeasurements.
REFERENCES
Brinkgreve,R.B.J.2002.PLAXIS, Finite element code for soil
and rock analyses, usersmanual. Rotterdam: Balkema.
Schweiger, H.F. & Breymann, H. 2005. FE-analysis of
fivedeepexcavations inlacustrineclay andcomparison
within-situmeasurements. In(K.J. Bakker, A. Bezuijen,
W. Broere, E.A. Kwast, eds.), Proceedings 5th Int. Symp.
Geotechnical Aspects of Underground Construction in
Soft Ground, Amsterdam, June 1517, 2005, Taylor &
Francis/Balkema, Leiden, 887892.
199
Theeffect of deepexcavationonsurroundinggroundandnearbystructures
A. Siemi nska-Lewandowska& M. Mitew-Czajewska
Warsaw University of Technology, Warsaw, Poland
ABSTRACT: Inthepaper problemsrelatedwiththeexecutionof 29mdeepexcavationof NowySwiatStation
(S11) of 2ndmetrolineinWarsawarediscussed. Inthecentral section, Warsaw2ndmetrolinerunsbelowthe
center of thecity(officeandhousingbuildingsandhightrafficroads) aswell asbelowVistulariver.Thiscentral
sectionconsistsof 7stationsand6runningtunnels 6kmlengthintotal. Runningtunnelswill beconstructed
usingTBM, stations cut andcover method. Deepexcavationwill beexecutedwithindiaphragmwalls. The
stabilityof thewallswill beprovidedbyseveral levelsof slabsandstruts. Theanalysisof settlementsof ground
surface, surroundingfoundations anddisplacements of walls of theexcavationhavebeenmade. Additionally,
settlementsof thesurfacewerecalculatedabovetheTBM (runningtunnels). Resultingvaluesof settlementsin
bothcaseswerecomparedanddiscussed.
1 INTRODUCTION
Constructionof 2ndMetrolineinWarsawisscheduled
tobegininJ anuary2008, announcementof designand
buildtender hasbeenalreadypublished. Inthecentral
section,Warsaw2ndMetrolinecrossesbelowthecen-
ter of thecity (officeandhousingbuildingsandhigh
trafficroads)aswell asbelowVistulariver.Thiscentral
sectionconsistsof 7stationsand6runningtunnels
6kmlengthintotal.
Figure1. Longitudinal sectionof thecentral part of the2ndMetrolineinWarsaw.
Running tunnels will be constructed usingTBM,
stations cutandcovermethod. Inthepaperproblems
relatedwiththeexecutionof 29mdeepexcavationin
Quarternarysoils(siltysands, sands, clayeysandsand
Plioceneclays) arediscussed. Within theexcavation
Nowy Swiat Station(S11) of the2ndmetrolinewill
bebuilt.TheS11stationwill befoundedatthedepthof
29mbelowgroundsurface(b.g.s.) inthevicinityof so
calledWarsawSlope, wherethedenivelation(differ-
enceingroundsurfacelevels) reaches30m(Figure1).
201
Figure2. Longitudinal sectionof thecentral part of the2ndMetrolineinWarsaw.
Thedepth of thestation is aconsequenceof sud-
denloweringof thetunnel fromtheupper slopelevel
to the level below the bottomof the river with the
considerationof appropriatesoil cover resultingfrom
TBM technology. Deep excavation will be executed
within100cmthickdiaphragmwalls. Thestabilityof
thewalls will beprovided by several levels of slabs
andstruts. Intheclosevicinity of theexcavationand
abovethetunnel therearemanyoldbuildings, suchas:
historic buildingsbuilt inXIXthcentury, partially
destroyed during the 2nd World War and rebuilt
after thewar. PolishAcademyof Sciences, Warsaw
University andaHospital arelocatedthere. These
buildings arefounded on spread foundations at a
depth of 4,80mb.g.s. Shortest distance between
theexcavationwall andfoundationof thebuilding
amountsto3m;
residential buildings constructed in 30. of XXth
century,probablyfoundedonpiles.Thesebuildings
arelocatedabovethetunnel drilledusingTBM;
residential andofficebuildings constructedin50.
and 60. of XXth century on old pre-war foun-
dations. These buildings are founded on spread
foundations at adepth of 4,00mb.g.s., 6mapart
fromtheexcavationwall;
masonryandconcreteresidential andofficebuild-
ings constructedin 60. of XXth century, founded
at a depth of 6,00m b.g.s., 5m apart from the
excavationwall.
Polish Central State Bank and the Ministry of
Financearelocatedthere.
Thesesbuildingsaremostlymasonryor reinforced
concretestructuresingoodtechnical state.Themajor-
ity of themis protected by the heritage conservator
law. The location of the excavation of S11 Station,
running tunnels and surrounding buildings is shown
onFigure2.
Theanalysisof settlementsof groundsurface, sur-
roundingfoundationsanddisplacementsof excavation
wallshavebeenmade.Additionally, settlementsof the
surfacewerecalculatedabovetheTBM, T11running
tunnel (cross-sectionmarkedby greenline). Figure2
showsthelocationof all calculationcross-sectionsin
thevicinityof S11StationandT11runningtunnel.
2 GEOLOGY
ThereareQuaternaryandTertiarysoilsintheareaof
thedeepexcavationof thestationandrunningtunnels.
According to the geotechnical investigations report,
followinggeotechnical layersaredistinguished:
layer I uncontrolledfills 1,52mthick, insome
placesupto3m;
layer II morainedeposits reachingdepthof 4m
b.g.s., consistingof mediumandstiff sandy clays
andclayeysandsof Wartaglaciation;
202
Figure3. CalculationsectionN
3.
layer III mediumsands and silty sands of Odra
glaciation, tothedepthof 10mb.g.s.;
layer IV morainedepositsreachingdepthof 13m
b.g.s. consistingof mediumandstiff sandyclaysof
Odraglaciation;
layerV pliocenclaystill thedepthof 50mb.g.s.
Therearethreelevels of groundwater table. Con-
sidering temporary stability of the bottom of the
excavation, it wasassumedthat thewater tablewould
beloweredduringconstructionof thestation.Geotech-
nical conditions,distributionof soil layersandlocation
of foundationsareshownonFigure3andFigure4.S11
stationandT11runningtunnel arebothlocatedwithin
thelayer of stiff andverystiff Plioceneclays.
3 DESCRIPTIONOF THE DEEP EXCAVATION
OF THE S11STATION
It was designedthat thedeepexcavationof S11Sta-
tionwill beexecutedwithin100cmthick diaphragm
walls, founded10mbelowthebottomof theexcava-
tion(thatmeanstheheightof wallsis39m). Duetothe
greatdepthof theexcavation, amountingto29m, slab
methodof theexecutionof theexcavationwaschosen
inordertoprovidemaximumsafetyof theconstruction
works. Thestability of diaphragmwalls will bepro-
vided by 8 levels of 35cmthick underground slabs.
Vertical spacingof slabsis3m, whichgivesanoppor-
tunitytoadopt undergroundsurfacefor car parksand
retail. Constructionstagesareconsideredasfollows:
execution of guide-walls, 1m thick diaphragm
walls and 1mhigh reinforced conretegirt on the
entireperimeter of theexcavation,
excavation till the depth of 2mb.g.s., i.e. below
the slab at level 1, execution of barrettes and
temporaryslabsupports,
constructionof theslabatlevel 1, backfillingthe
excavationandallowtrafficback,
excavationtill thedepthof 5mb.g.s., i.e. belowthe
slabat level 2,
constructionof theslabat level 2,
excavationtill thedepthof 8mb.g.s., i.e. belowthe
slabat level 3,
excavationtill thedepthof 11mb.g.s., i.e. below
theslabat level 4,
theslabat level 5,
theslabat level 6,
theslabat level 7,
theslabat level 8,
excavationtill thedepthof 26,5mb.g.s.,
installationof temporarystrutsatthedepthof 26m
b.g.s.,
final excavationtill thedepthof 29mb.g.s.,
constructionof 1,5mthickfoundationslab,
deinstallationof thetemporarystruts.
Calculations were made in 3 sections, chosen
becauseof thevicinityof significant buildings.
Figure 3 presents example cross-section N
3,
located close to the beginning of the running tun-
nel (for thelocationof thesectionrefer toFigure2),
showinggeotechnical conditionsandsurcharges.
4 DESCRIPTIONOF THET11TUNNEL
Two versions of the tunnel structure has been
considered: 1tubeincluding2tracksand2tubes, sin-
gletrackeach.
The lining of the tunnel was assumed to be con-
structedof 40cmthicksegments. Followingstagesof
theexecutionof thetunnel weremodeled:
initial stress includingoverburdenandsurcharges
(buildingsandtraffic),
203
excavation of the tunnel and construction of the
liningof thetunnel.
Figure4showscalculationcross-sectionN
I-I (for
thelocationof thesectionrefer toFigure2) including
geotechnical conditions, tunnels(2tubes)andlocation
of existingbuildings.
5 CALCULATIONS
5.1 Calculations of the excavation of S11 station
Finiteelement plain strain analysis werecarried out
usingPLAXIS v. 8software, Coulomb-Mohr consti-
tutive soil model was chosen for modeling the soil
body,diaphragmwallsaswell asslabsweremodeledas
3-nodes, linearbeamelements. Non-associatedplastic
flowlawwasconsidered. For modelingwall frictions
Coulomb-Mohr lowwasused. Model dimensionsare:
65m (vertical), 100m (horizontal), they were esti-
matedtakingintoaccountpolishregulationsaccording
totherangeof influencezoneof theexcavation.
FEM model mesh, generated automatically, was
built of 807 15-nodes triangle elements and 9773
nodes. For thepurposeof thepaper 3rdcross-section
waschosentobepresentedanddiscussedbecauseof
Figure4. CalculationsectionNoI-I (T11).
its vicinity to theT11 running tunnel. Geotechnical
conditionsandlocationof existingbuildingshasbeen
presentedonFigure3, FEM model is shownonFig-
ure5. Figure6presentsmaximumdeformationsof the
model inthefinal constructionstage. Maximumcal-
culatedlateral displacement of thediaphragmwall in
section3amountsto49,3mm; maximumfoundation
displacement 30,6mm.
Table 1 presents maximum calculated values of
horizontal and vertical displacements of the wall as
well as settlements of thesurroundingbuildings in3
cross-sectionschosenfor calculation.
Figure5. Numerical model section3(PLAXIS).
204
5.2 Calculations of the T11 running tunnel
PlaceFiniteelementplainstrainanalysiswerecarried
out usingGEO4TUNNEL software, Coulomb-Mohr
constitutivesoil model was chosen for modeling the
soil body, tunnel lining was modeled using 3-nodes,
linearbeamelements. Non-associatedplasticflowlaw
wasconsidered. Formodelingwall frictionsCoulomb-
Mohr lowwasused.
SectionI-I, 1tube, 2tracks:
model dimensions: 60m (vertical) and 240m
(horizontal);
Figure6. Final displacements section3(PLAXIS).
Table1. Resultsof calculationsof deepexcavationS11.
Maximum Maximum
displacementsof settlements
diaphragmwall of buildings
Ux Uy U
Section [mm] [mm] [mm]
1-1 46,1 32,8 24,5
2-2 61,0 52,8 35,2
3-3 49,3 53,3 30,6
Figure7. FEM model, T11tunnel 2tubes, (GEO4TUNNEL).
built of 70606-nodestriangleelementsand15011
nodes.
SectionI-I, 2tubes, singletrackeach:
model dimensions: 60m (vertical) and 240m
(horizontal);
built of 81786-nodestriangleelementsand17284
nodes.
Geotechnical conditions, tunnelslocation(2tubes,
singletrackeachcase) andlocationof existingbuild-
ings has been presented on Figure 4, corresponding
FEM model is shownonFigure7. Figure8presents
maximumdeformationsof that model.
Table 2 presents maximum calculated values of
bending moments, and displacements of the lining
aswell assettlementsof thesurfaceandsurrounding
buildings.
6 CONCLUSIONS
Taking into consideration results of analysis of the
excavationof S11Stationaswell astheresultsof T11
runningtunnel calculationsfollowingconclusionsare
formed:
1. Inthevicinity of the29mdeepexcavation, which
will be executed during the construction of S11
MetroStationestimatedsettlementsof thesurface
andsurroundingbuildingsamountto24,535mm.
2. Calculated settlements of theground surfaceand
surroundingbuildingsabovetheT11runningtun-
nel constructedby themeansof TBM, takinginto
consideration both cases 1 two track tunnel and
2 single track tunnels are similar and amount to
37,537,8mm.
3. Theoretical values of settlements as well as dis-
placements and forces in the structures were
205
Figure8. Maximumdeformationsof themodel, T11tunnel.
Table2. Resultsof calculationsof runningtunnel T11.
Finiteelementsmethod
Maximum
bendingmoments, Maximum Maximum
displacements surface settlements
of tunnel lining settlements of buildings
Typeof Mmax Umax Umax Umax
tunnel [kNm/m] [mm] [mm] [mm]
1-1 306,6 24,2 37,7 37,8
(1tunnel)
1-1 290,2 8,1 37,5 37,6
(2tunnels)
calculatedconsideringthat thevalueof themodu-
lusof deformationof Plioceneclays, withinwhich
thestructuresarelocated,isE=50MPa.Thisvalue
must beverifiedbymeansof in-situtestsandthen
thecalculationswill beadjusted.
4. Due to the expected differences in the values of
settlementsof thegroundsurfaceclosetothedeep
excavation and above the tunnel further analysis
of thecaseincluding 3D modeling of thecontact
of 2types of tunnel structure(runningtunnel and
station) will beperformed.
5. During the construction, the results of analysis
describedinthepaperwill becarefullyverifiedand
discussed.
REFERENCES
FINE Ltd. 2007. GEO4 Users manual. Prague: FINE Ltd.
Geoteko Sp. z o.o. 2004. Evaluation of the technical state
of buildings in the influence zone of 2nd metro line in
Warsaw, section Rondo Daszy nskiego Station Powi sle
Station. Warsaw: GeotekoSp. z o.o.
Geoteko Sp. z o.o. 2004. Geotechnical and hydrological
report for theconstructionof 2ndmetro lineinWarsaw,
Nowy
SwiatStation. Warsaw: GeotekoSp. zo.o., SGGW,

Geoprojekt Sp. z o.o.
Grodecki, W., Siemi nska-Lewandowska, A. & Lejk, J. 2007.
Second metro line in Warsaw possibility and meth-
ods of realization, In zynieria i Budownictwo 7-8/2007:
365368.
Kotlicki, W. &Wysoki nski L. 2002. Protection of structures
in the vicinity of deep excavations (376/2002). Warsaw:
BuildingResearchInstitute.
PLAXISBV. 2005. PLAXIS Users manual. Roterdam: A. A.
Balkema.
Polish Committee of Standardisation. 2002. PN-EN 1538-
2002 Execution of special geotechnical works. Diaphragm
walls. Warsaw: PolishCommitteeof Standardisation.
206
Multi-criteriaprocedurefor theback-analysisof multi-supported
retainingwalls
J. Zghondi
Arcadis, Lyon, France and LGCIE, INSA-Lyon, France
LGCIE, INSA-Lyon, France
ABSTRACT: A numerical back-analysis procedurefor multi-supporteddeepexcavations is proposedbased
ontheoptimizationof several indicators, takinginaccount theforcesinthestrutsandthedifferential pressures
derivedfromthewall displacement. Theevaluationof theprocedureisperformedon1gsmall scalelaboratory
experiments(Masrouri 1986) onsemi-flexibleretainingwallsembeddedinSchneebelli material. Theproposed
numerical procedurewasappliedonanexcavationwith2passiveslowstiffnessstruts. TheresultingHardening
Soil Model parametersarefurther usedtoback-calculatethe14different testedconfigurations. Theresultsare
comparedwiththeclassical methods, SubGradeReactionMethod, FiniteElementanalysiswithMohr Coulomb
model withparameters proposedby Masrouri (1986) andwiththeback-analysis usingHardeningSoil Model
parametersbasedontriaxial testsresults.
1 INTRODUCTION
Numerical back-analysisof insitumonitoringresults
of multi-supported deep excavations is generally
extremelycomplex(Hashash&Whittle1996,Finno&
Calvello2005, Delattre1999): soil characteristicscan
beheterogeneousor determinedwithalowdegreeof
confidence, thedifferent stagesof theexcavationcan
bedifficult toreproduceina2Dnumerical approach,
keymechanical parameterscanbeunknown(forexam-
ple the actual stiffness of the strut-to-wall contact)
and thenumber of measured quantities such as wall
displacements, settlements and strut forces is gener-
allytoosmall toperformacomprehensivecomparison
betweentheactual behavior andthenumerical results.
Thus thefull validationof aback-analysis numer-
ical procedure (including in particular the choice of
theconstitutivelaw and thedetermination of all the
requiredparameters) israrelydirectlypossibleonreal
casehistories.Therefore,thenumerical procedurepro-
posedinthispaperisvalidatedon1gsmall scalelabo-
ratoryexperimentsperformedbyMasrouri (1986) on
semi-flexibleretainingwallsembeddedinSchneebeli
material (mixtureof steel rods of different diameters
representingin2Dthebehaviorof acohesionlesssoil).
The14consideredexperimentscorrespondtoaretain-
ingwall, whoselengthandmechanical propertiesare
keptconstant, supportedbyoneor twolevelsof active
or passivesteel strutswithvariousaxial stiffnessand
prestressing.
Even for such simple comprehensive laboratory
experiments, usual design method like SubGrade
Reaction Method(SGRM) or classical limit equilib-
riummethodsdonotcaptureall theobservedbehaviors
andtest results. It isthusnecessary toproposeauni-
fied numerical procedure to back calculate with an
acceptabledegreeof confidence, all theresultsof the
excavationstestsof Masrouri (1986).
2 EXPERIMENTSANDCOMPARISONWITH
CLASSICAL/SGRM CALCULATIONS
Theexperimentscorrespondtosmall scale2Dmodels
of flexibleretainingwalls(Figure1).
Figure1. Masrourisexperimental set up.
207
Table1. Summaryof Masrouri (1986) experiments.
First strut Secondstrut
Prestressed Stiffness Prestressed Stiffness
Experiments (kN/m.ml) (kN/m.ml) (kN/m.ml) (kN/m.ml)
B1 2.133 83333 (not used)
B2 3.025 83333 (not used)
B3 0.208 816 (not used)
B4 2.1 816 (not used)
B5 3 816 (not used)
B6 0.191 404 (not used)
B7 2.016 404 (not used)
B8 2.916 404 (not used)
B10 0.383 83333 0.333 83333
B11 1.330 83333 2.691 83333
B12 0.366 816 0.400 816
B13 2.075 816 2.700 816
B14 1.733 816 4.041 816
Schneebelli 2D analogic soil was used: this mate-
rial offer a good repeatability and enables to build
a homogeneous 2D soil model with quick handling
for the experiments. However Schneebeli materials
havesomeinconvenients: theunit weight is closeto
6.5kN/m
3
, theangleof frictionissmaller thanthat of
most of thesoils (21
) andit only presents adilatant

behavior.
While maintaining the same geometrical and
mechanical characteristics for the wall (EA =1.2
10
6
kN/mand EI =14.4kN/m
2
/m), a wide range of
configurations(1or 2struts, withdifferent prestress-
ing and stiffness) was considered, see Table 1. The
phases used in all the excavations were planned as
follows;
1. strut cases: 10cmof excavation, installationof the
first strut at 5cmfromthe top and prestress-
ing, then3excavationsof 10cmeachtill 40cm,
then several excavations of 5cmuntil failure is
obtained.
2. struts cases: the same procedure as for the 1
strut casesisfolloweduntil theexcavationreaches
40cmfromthetop, thenthe2ndstrutisinstalled
at thelevel 25cmandprestressed. Thestep-wise
excavation(byincrementof 5cm) isresumeduntil
failureoccurs.
Thestruttowall contactishingedinordertoprevent
bending moment to betransmitted to thestruts. For
eachexcavationphase,thehorizontal displacementsof
thetopor bottomof thewall aremeasured.Thecurva-
tureisalsomeasuredin26locations(bothsidesonthe
wall)allowingtodeterminewithareasonableaccuracy
thedifferential pressureactingonthewall fromthetop
to thebottom(polynomial approximation performed
bythePalpanprogramcreatedbyBoissieret al. 1978).
Photographswerealsotakentodeterminethedisplace-
ment fieldswithastereophotogrammetric technique.
0
10
20
30
40
50
60
70
80
-15 -5 5 15
Differential pressure (kN/m2)
exc -45cm
Experiment
SGRM
B3
Passive strut
low stiffness
Figure2. Differential pressurecalculatedwithSGRMcom-
pared to experimental values: Case B3 low stiffness
passivestrut (45cmexcavationlevel).
Load cells determine the forces on the cylindrical
struts.
2.1 Classical and SGRM calculations
Thewholeseriesof experimentswerefirst compared
with classical (modified Blummethod) and SGRM
calculations (Terzaghi 1955). The subgrade reaction
modulus used in RIDO calculations (Fages 1996) is
increasingwithdepthinthefollowingmanner:
K
h
= K +K
v
. withK = 0, K
h
= K
v
Classical methods calculation totally neglects the
influence of the stiffness of the wall, construction
steps, stiffnessandprestressingonthestruts, arching
effect, etc. OntheoppositeSGRMexplicitlyconsiders
thestiffnessof thewall, theconstructionstepsandthe
stiffnessof thestrut.
In thecaseof onepassivestrut with lowstiffness
(tests B3, B6), the SGRM method reproduces in an
acceptable manner the differential pressures results
(Figure2).
208
0
10
20
30
40
50
60
70
80
-20 -10 0 10 20
Differential pressure (kN/m2)
exc -45cm
Experiment
SGRM
B10
Passive stiff
strut
0
10
20
30
40
50
60
70
80
-20 -10 0 10 20
Differential Pressure (kN/m2)
exc -45cm
Experiment
SGRM
B14
Active low
stiffness
Figure3. Calculated (SGRM Method) and measured dif-
ferential pressurefor B10andB14at 45cmof excavation.
Figure4. Meshusedfor thePlaxiscalculationsof 2struts
supportedwalls.
Withonestiff or activestrut or inthecaseof exca-
vations with 2 struts, both the SGRM and the limit
equilibriummethodsfail toreproducetheexperiment
differential pressures(Figure3a Test B10). Insome
cases, a good description of the pressure diagrams
seemstobeobtained(Figure3bTestB14). Itactually
results fromtwoerrors compensatingeachother: the
overestimationof thepressureinducedbyprestressing
andthelackof abilitytoreproducethearchingeffect.
2.2 Finite element back calculation with Mohr
Coulomb model
Itappearsthattheclassical andSGRMmethodsdonot,
eveninsimplecasesof excavation, accuratelydescribe
all theobservedresults. A finiteelement approachis
thereforeproposed.
The finite element calculations were performed
withPlaxisV8.2, themodel representsavertical slice
of Masrouris experiment. Themesh is composed of
triangular 15nodeselements(Figure4).
For thesecalculations, thesamecharacteristics of
the Schneebelli material were used (c=0, =21
,
=65kN/m
3
), taking into account for the soil-wall
interfaceaninterfacefactor R
int
=0.55(relatedtothe
frictionanglevaluenotedbyMasrouri).
Masrouri (1986) estimated theelastic modulus at
4500kPa with an increment of 26830kPa/m after
Figure 5. Strut forces for cases B1 to B8 with 1 strut
variationduringthelast 4excavationphasesbeforefailure.
0
100
200
300
400
500
600
0,00 1,00 2,00 3,00 4,00 5,00
200 kPa
300 kP a
400 kPa
D
e
v
i
a
t
o
r
i
c

S
t
r
e
s
s

k
P
a
200 kPa
300 kPa
400 kPa
Axial Strain %
Experiments
HSM1
-4
-2
0
2
4
6
8
10
12
0 6
500Kpa
200Kpa
400Kpa
200Kpa
1 2 3 4 5
Axial Strain%
500Kpa
200Kpa
400Kpa
200Kpa
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
/
1
0
0
0
Experiments
HSM1
Figure6. SimulationwithHSM1parametersof thebiaxial
testswith200, 300and400kPa.
0.6mof depth(valuesbackcalculatedontestB4with
1activestrut of lowstiffness).
Inall thepresentedcalculationsthedifferencewith
themeasuredstrut forces andwall displacements do
not exceed20%(seeFigure5for the1-strut casesB1
toB8).
A sensibility analysis is performed considering
all the parameters of the Mohr Coulomb model. It
appears that the differences between calculated and
experimental results can not besatisfyingly reduced
(especiallyinthecaseswith2struts).Thereforeamore
sophisticated constitutivelaw is required (Figure6):
theHardeningSoil model.
209
Figure7. Differential pressurefor test B12andexcavation
at 50cm: comparisonof theexperimental values withthe
resultsof SGRM, FEwithMohrCoulomblawandHardening
Soil Model (HSM1andHSM2).
2.3 FE back-calculation with Hardening soil model
parameters based on triaxial tests
Hardeningsoil model (Shanz et al. 1999) cancapture
soil behavior inaverytractablemanner. Thevaluesof
thedifferentparameterswerefirstfittedonthebiaxial
testsresults(under a200Kpaconfiningpressure) per-
formedon2010cmsamplesbyKastner(1982).The
200kPaconfiningstressisgreaterthanthemeanstress
generally observedinMasrouri experiments, but ina
firstapproachitwasconsideredasmorerepresentative
of thesoil behavior thanthebiaxial testsusinglower
confining pressures: the size of the sample and the
highvalueof theunit weight of Schneebelli material
inducesanonhomogeneousstressstateinthesample
that could greatly affect the accuracy of the results.
The obtained set of parameters noted HSM1 in the
sequel providesasatisfyingdescriptionof thebiaxial
test with200kPaor moreconfiningstress(Figure6).
Thissetof parameterswasfurther usedtobackcal-
culatethewholeseriesof Masrourisexperiments(14
cases). It appearsthat thesecalculationsdonot repro-
duce well the test results in termof strut forces or
differential pressuresonthewall (Figure7represents
onlytheresultsof differential pressures).
3 PROPOSEDBACK ANALYSISPROCEDURE
The aimof this procedure is to find the proper set
of parameters for the constitutive soil model (Mohr
CoulomborHardeningSoil models) consideredinthe
Finite Element simulations. Considering one partic-
ular test configuration, the parameters will be first
obtained fromthe minimization of indicators based
on differential pressures and struts forces errors.
The resulting set of parameters will then be con-
frontedwiththeresults of thebiaxial tests andof 14
Figure 8. Differential pressure curves for indicator
explication.
configurationsconsideredexperimentallybyMasrouri
(1986).
3.1 Definition of indicators
Inorder toconsider themainfeaturesof theretaining
wall behaviour, two indicators were defined: E
sm
is
relatedtothestrutforcesandE
pd
relatedtothediffer-
ential pressureonthewall (linkedtothedisplacement
profile).
whereE
b1
=f
1
f

1
andE
b2
=f
2
f

2
andwithf
1(2)
and
f

1(2)
thestrut forcecalculatedor measuredinthe1st
(2nd) strut level.
TheE
sm
indicator isbasedontheerror of thesum
of strutsforces: thereforeanerror onthestrutforcef
1
canbecompensatedbyf
2
.
TheE
pd
indicator (Figure8) takesintoaccount the
absolutevalueof thedifferencebetweenthemeasured
andcalculateddifferential pressures (respectively P
1
andP
2
), dividedby theintegral of themeasureddif-
ferential pressureP
1
. Integralsarecalculatedfromthe
top of the wall to 10cmbelow the final excavation
level.
E
pd
isthemainindicator whilethepossibleinaccu-
racy of thestrut forcemeasurement especially at the
beginningof theexcavationmakestheE
sm
indicator a
secondvalidationindicator.
3.2 Parameter optimization for Mohr Coulomb
model based on a 2 strut excavation test
Thedetermination of asecond set of parameters for
Mohr Coulomb model is performed by fitting the
final resultsof a2strutsexcavationcase(excavation
level 50cm). Theparticular caseof test B12 (cor-
responding to passive struts with low stiffness) was
selectedbecauseitclearlyappearsthattheSGRMfail
210
Figure9. Variationsof indicatorsE
sm
andE
pd
withE
ref
50
.
toreproducethedifferential pressurediagramobtained
insuchconfiguration.
Only oneindependent parameter was usedfor fit-
ting the results, the elastic modulus E. The other
parameters arekept constant (c, , . . .). Thefinal
valueof E is determinedthroughtheoptimizationof
theindicatorsE
sm
andE
pd
. Actually, alinear variation
of E withthedepthisconsidered:
Both the initial value of the elasticity E
0
and its
variation with depth were considered in the opti-
mization procedure. It appeared that none of these
parameterscouldbemodifiedtoimprovethedescrip-
tion of the experimental results, indicating that the
Mohr Coulombmodel is unableto reproduceimpor-
tantfeaturesof thesoil behaviourinvolvedintheglobal
behaviour of theretainingwall.
3.3 HSM model optimized by fitting on a 2 strut
excavation test
Thedetermination of asecond set of parameters for
HardeningSoil Model (notedHSM2inthesequel) is
performed by fitting the final results of the same 2
strutsexcavationcase(B12excavationlevel 50cm)
asinsection3.2.
Onlyoneparameter wasusedfor fittingtheresults,
thereferencestiffnessmodulusE
ref
50
.Theotherparam-
eters areeither constant (c, , ) or keepadirect
relationshipwithE
ref
50
. For example:
Thefinal valueof E
ref
50
isdeterminedthroughtheopti-
mization of theindicators E
sm
and E
pd
. Considering
theaccuracy of theexperimental results (in particu-
lar theprocedureleading to thedifferential pressure
diagrams), E
ref
50
is determined with a precision of
1000kPa(Figure9).
Thevalueof E
ref
50
=6000kPais chosenbecauseit
appearstominimizebothindicators. HSM2isfurther
usedto simulatethebiaxial tests withlowconfining
stress(50and100kPa). Figure10showsagoodagree-
ment with theexperimental results, even though the
latter canbeaffectedby thenon-homogeneity of the
initial stressstatein20cm10cmsamples.
Figure 10. q-
1
biaxial curves: experimental results and
calculatedvalueswithHSM2.
The proposed back analysis procedure shows its
ability to verify or justify the HSM parameters that
accuratelydescribethebiaxial test results.
3.4 Back calculation of the 14 configurations tested
by Masrouri (1986)
TheHSM2set of parametersisnowusedtobackcal-
culatethe14different tests(Table1). Figures11and
12presentthedifferential pressurediagramsobtained
withHSM2andother methodsandtheratios f ,f

of
thecalculatedtothemeasuredstrutforcesrespectively
for tests B1 to B8 (1 strut) and tests B10 to B13 (2
struts).
The strut forces and the differential pressure are
well represented compared to theclassical methods,
the SGRM method or Finite element analysis with
a simpler constitutive model (Mohr Coulomb). The
effectof theprestressingof thestrutiswell reproduced
bythat procedure(comparingB3andB5inFigure11
or B10andB11inFigure12), as well as thearching
effect (caseof B10inFigure12) andtheinfluenceof
thestrutstiffness(comparingB1andB7inFigure11).
The proposed back analysis procedure and the
resulting set of parameters HSM2 show their effi-
ciency in all the configurations (unlike the SGRM
wherethearching effect and theprestressing on the
strutsarenot well reproduced).
3.5 Summary
Figure13presentsacomparisonof thevaluesobtained
for theselectiveindicator E
pd
in3casesof excavation
andfortheSGRM, MohrCoulomb, HSM1andHSM2
models:
B3correspondstoasinglepassiveandlowstiffness
strut
B12uses2passivelowstiffnessstruts
andB102passiverigidstruts.
In all of these 3 cases, it appears that the HSM2
set of parametersclearlyminimizestheerror between
experimental andnumerical results.
211
Figure11. Differential pressurediagrams andstrut forces
obtained with HSM2 (excavations with 1 strut at 50cm)
compared with experiments, Mohr Coulomb, and classical
methodsresults.
Figure12. Differential pressurediagrams andstrut forces
obtainedwithHSM2(excavations with2struts at 50cm)
comparedwithexperimentsandclassical methodsresults.
Figure13. E
pd
indicator calculatedfor testsB3, B12, B10
(excavation 50cm), for the SGRM, HSM1, HSM2 and
Mohr Coulombcalculation.
With the same constitutive model (HSM1 and
HSM2) andwiththesameparametersexceptE
ref
50
, the
error isdividedby3to5.
4 CONCLUSION
A comprehensive series of 14 small scale experi-
ments on flexibleretaining walls with different strut
212
stiffnessandprestressingisusedtovalidateanumeri-
cal back calculationusingtheHardeningSoil Model.
The final set of parameters HSM2 is fitted on one
single test (B12 at the final excavation level). The
proposed model is based on the simultaneous mini-
mization of two indicators E
sm
and E
pd
respectively
relatedtothestrutforcesanddifferential pressuredia-
gram. Theverificationof theproposedback analysis
procedureshowedthattheHSM2model givesthemost
acceptabledescriptionof thedifferential pressuresand
forcesonthestrutsinall of the14testedconfigurations
compared to theSGRM method or aFiniteElement
approachwitheitherMohrCoulombmodel orHarden-
ingSoil model withparametersbasedonbiaxial tests
(HSM1).
Further developmentswill includetheverification
of theability of proposedback-analysis procedureto
determinedtheHSM parametersnot onlyonE
sm
and
E
pd
at thefinal excavationlevel but alsooninterme-
diatelevels, for examplethefirstexcavationstepafter
theinstallationandprestressingof thelower strut.
Despite the already mentioned difficulties, fur-
ther validationof theprocedureonwell-instrumented
excavationsiteswill alsobetested.
REFERENCES
Boissier, et al. 1978. Dtermination des moments et des
pressions exercs sur uncranpartir demesures incli-
nomtrrique.Revue Canadienne de Gotechnique, 15,(4),
522536.
Brinkgreve, R.B.J. & Broere, W. 2004. PlaxisV8manuel.
Delattre, L. 1999. Comportementdescransdesoutnement-
Exprimentations et calculs. PhD dissertation, ENPC,
Paris, ENPC, 498p.
Fages, J. 1996Rido Users manual, RFL, Miribel, France.
Finno, R. & Calvello, M. 2005. Supported excavations:
observational methodandinversemodeling. J. Geotech.
Geoenv. Eng. ASCE, 131, (7): 826836.
Hashash,Y. M. A. &Whittle, A. J. 1996. GroundMovement
Predictionfor DeepExcavations insoft Clay. Journal of
Geotechnical Engineering, 122(6): 474486.
Kastner, R. 1982. ExcavationprofondeensiteurbainProb-
lmes lis la mise hors deau- Dimensionnement des
soutnements butonns. PhD dissertation, INSA Lyon &
Universit Claude Bernard-Lyon 1, 409p.
Masrouri, F. 1986. Comportement DesRideaux desoutne-
ment Semi- Flexibles. PhD dissertation, INSA Lyon,
France.
Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. The
Hardening-Soil Model: Formulation and verification.
In: R.B.J. Brinkgreve, Beyond 2000 in Computational
Geotechnics. Balkema, Rotterdam: 281290.
Terzaghi, K. 1955. Evaluation of coefficients of subgrade
reaction, Gotechnique, 5: 297326.
213
Monitoringandmodellingof riversidelargedeep
excavation-inducedgroundmovementsinclays
D.M. Zhang& H.W. Huang
W.Y. Bao
China State Construction Engineering Corporation (SH), Shanghai, P.R. China
ABSTRACT: TheRiversidelargedeepexcavationof Shanghai international passenger center was800mlong
and100150mwidewiththedepthof 13m. Thesouthlongsideof thedeepexcavationwas at adistanceof
4.6mfromtheparallel floodwall of HuangpuRiver. Thenorthlongsidewas5mawayfromahistoricbuilding.
Problemsresultedfromthelargedeepexcavationwastheasymmetricgroundmovementsalongthelongsides
duetothecomplexsurroundingconditionandsurfacesurcharge.Themonitoringduringtheexcavationprovided
numerousdatatostudythecharacteristicsof thegroundmovementandearthpressure.Thenumerical modelling
wasalsoadoptedaimtopredict thegroundmovements.
1 INTRODUCTION
Thedevelopmentof undergroundspacealongthebund
of HuangpuRiver inShanghai, Chinahasresultedin
excavationsbecomingprogressivelylarger andcloser
totheRiver, wherethegroundwatertablewasjustnear
thegroundsurfaceandagreatnumberof underground
works are within a few meters of the surface. The
riversideexcavationswereall locatedclosetotheexist-
ing buildings, network and the city lifeline of flood
wall. It hasbecomeagreat challengetoprotect these
neighboringbuildings andpublic utilities fromdam-
age during the deep excavation due to the complex
geotechnical constraints and thesmall opening from
theHuangpuRiver. Thesoilsnear theHuangpuRiver
wasusuallyweakwithaverylowstrengthandhigher
water content, which were a potential causes of the
largergroundmovement.Meanwhile,thecomplexand
denseenvironments put forward astrict requirement
onthegroundmovement controlling. It was difficult
todeterminetheearthpressureactedontheretaining
wall withanyconventional earthpressuretheorycon-
sideringthesmall soil bodyleftbetweenretainingwall
andfloodwall. Besides, theretainingwall of riverside
deep excavation was usually asymmetrically loaded
with much higher earth pressureon oneside, which
wascausedbygreatsurfacesurchargeduetotheexist-
ingbuildingsandthepileof theconstructionmaterial.
Thestability of thedeep excavation as a wholewas
worth considering to avoid any kinds of failure of
the deep excavation and consequent damage on the
environments.
However, there were few references for the con-
struction of the large deep excavation because of
thegeotechnical conditionandcomplex environment
alongthebundof HuangpuRiver. Thedeepexcava-
tionof Shanghai international passengercenter(SIPC)
wasthelargestandclosestonetoHuangpuRiversofar.
Theconstructionandtheanalysismethodof thedeep
excavationof SIPCandtheinducedgroundmovement
aswell will beauseful andpractical referencefor the
subsequent riversidelargedeepexcavation.
2 PROJ ECT OUTLINEANDSOIL CONDITIONS
2.1 Project outline
Thedeepexcavationof SIPCwas800mlongwiththe
widthof 100150mandthedepthof 13m. Thelarge
deepexcavationwasdividedintotwosub-excavations
with thelengths of 480mand 218mrespectively to
reducetheriskof damagefortheexistingstructureand
thefailureof deepexcavation. Thestudypresentedin
thispaperwascarriedoutbasedonthedeepexcavation
withthelengthof 480m.Thesketchviewof theproject
wasillustratedinFigure1.Thespacebetweenthedeep
excavationandfloodwall of HuangpuRiver wasonly
215
Figure1. Sketchviewof theproject.
Figure2. Theplaneviewof strut arrangement.
4.6mat south side. A number of existing buildings,
including ahistoric building, werelocated along the
north sideof thedeep excavation with adistanceof
about 5m.
BoredpilessupplementedbySMWpileswall were
used as the retaining structure. The bored pile was
950mmindiameter withacenter-to-center spaceof
1150mm. Theeffectivelengthof theboredpilewas
26mandtheembedmentwasadequatetoprovidesuf-
ficient passiveearthpressureto keepthestability of
the retaining wall. The SMW piles were 850mmin
diameter withaneffectivelengthof 20.8m. Thedis-
tancebetweenSWM piles was 600mmto guarantee
thewaterproof performance. Themixratioof cement
wasashighas20%for SMWpiles. Threereinforced
concretestrutsweresetatthedepthof 0.9m, 5.7m
and9.6mwiththecrosssectionof 1250800mm
forthefirststrutandof 1200800mmforthesecond
andthirdstrut. Theplanespaceof thestrut wasabout
1.2mand illustrated in Figure2. Thecross sections
of the deep excavation were presented in Figure 3a
and3b.
The jet grouting belt of 4mwide and about 4m
high was employed closely above the bottomof the
deepexcavationalongtheretainingwall.Thegrouting
could significantly increase the capacity of the soil
resistancefor theretainingwall duringtheexcavation.
Theboredpilewas extendedfrom26mto27mnear
Figure3a. A-A crosssectionof deepexcavation.
Figure3b. B-B crosssectionof deepexcavation.
thehistoricbuildingtoprotectthebuildingfromcrack
andtilting. Besides, theisolationpileswerespecially
designedtoreducetodeepexcavation-inducedeffect
onthehistoricbuilding.
2.2 Soil conditions
Thesoil profilethroughout thedeepexcavationcom-
prises the mixed filling to a depth of 6.4m, which
containsmanyobstaclesandmadealot of troublefor
thedeep excavation, underlain by silt, silty clay and
mucky clay. The retaining wall including the water-
proof wall wasembeddedinthesiltyclay.Thedetailed
characteristicsof thesoilswerepresentedinTable1.
216
Table1. Soilscharacteristicsthroughout thedeepexcavation.
Water Bulk Compression Cohesion Friction
Depth content density modulus kPa angle
Soil m % kN/m
3
kPa

Mixedfilling 6.38
Silt 4.89 31.7 18.2 8170 8 28.5
Muckyclaywithsilt 4.96 40.5 17.5 4130 11 22
Muckyclay 7.17 49.7 16.6 2580 14 13
Siltyclay1 7.31 34.0 18.0 4570 17 17
Siltyclay2 4.05 33.3 17.8 8130 8 29
Siltyclay3 17.23 33.2 17.9 5240 17 23.5
Figure4. Monitoringlayout of lateral displacements.
Table2. Progressof excavation.
corresponding
excavation
Date depth(m)
11/5/200511/11/2005 3
11/11/200512/2/2005 6.2
12/2/20051/3/2006 13
1/13/2006 completionof
bottomplate
3 MONITORINGOF EXCAVATION
As showninFigure4, 14inclinometers, whichwere
denoted by CX1 to CX14, were set into the retain-
ingwall aroundthedeepexcavation. 2earthpressure
gaugesdenotedasTY1andTY2inFigure4werealso
installedclosetotheretainingwall tostudytheevolu-
tionof earthpressureof thesmall soil body between
retainingwall andfloodwall duringexcavation. The
earthpressuregaugeswereinstalledevery 5minthe
vertical overall 25m.
Theprogressof excavationwaspresentedinTable2.
3.1 Monitored lateral displacements
Figures 58 showed the lateral displacements at
monitoring points of CX1, CX3, CX12 and CX10
correspondingtothestudiedexcavationstages. These
inclinometerswereclosetothecenter of thelongside
0
5
10
15
20
25
30
0 10 20 30 40 50
11/ 11/ 2005
12/ 2/ 2005
1/ 3/ 2005
1/ 13/ 2006
d
e
p
t
h
/
m
lateral displacement/mm
Figure5. Lateral displacement at inclinometer CX1.
0
5
10
15
20
25
30
0 10 20 30 40
12/2/2005
1/3/2005
1/13/2006
d
e
p
t
h
/
m
of the deep excavation and thus the readings were
representative of the maximumdisplacement of the
retaining wall. It could be found fromfigures 58
that themaximumlateral displacement was less than
60mmduringthewholeexcavationstage. Meanwhile,
217
0
5
10
15
20
25
30
0 10 20 30 40 50 60
11/11/2005
12/2/2005
1/3/2005
1/13/2006
d
e
p
t
h
/
m
0
5
10
15
20
25
30
0 10 20 30 40 50
11/11/2005
12/2/2005
1/3/2005
1/13/2006
d
e
p
t
h
/
m
thelateral displacement exhibitedasymmetricbehav-
ior along the two long sides of the deep excavation
becauseof thefollowingtworeasons. Firstly, theearth
pressure acted on the retaining piles was asymmet-
ric becauseof thesmall soil body betweenHuangpu
River anddeepexcavation. Secondly, thesurfacesur-
chargewasasymmetric duetotheexistingbuildings.
Comparingtherecordsof CX1withCX12, CX3with
CX10, it couldbefoundthat lateral displacement of
retainingwall was 1529%smaller at thesouthside
than north side. Unfortunately, thelarger lateral dis-
placement at north side would result in a potential
damageto theneighboring historic building. Conse-
quently, thejetgroutingshouldbeimmediatelycarried
out toimprovethefoundationof thehistoricbuilding
andit was provedtobeaneffectiveway toavoidthe
damageof crackandtilt of thebuilding.
Therecordedlateral movementsatthetopof retain-
ingwall werepresentedinFigure9.Thefivemovement
Figure 9. Distribution of lateral movement at the top of
retainingwall.
0
5
10
15
20
25
30
0 50 100 150 200 250 300 350 400
11/11/2005
12/2/2005
1/3/2005
1/13/2006
active pressure
d
e
p
t
h
/
m
earth pressure/kPa
Figure 10a. Comparison of recorded earth pressure with
calculatedactiveearthpressureatTY1.
curves fromoutsideto insidewerecorresponding to
theexcavationdepthof 2.3m, 6.2m, 13m, completion
of bottomplateandcompletionof undergroundstruc-
turerespectively. Themaximumlateral movement at
thetopof retainingwall reached106.5mmatthenorth
side, whileit wasonly50mmat thesouthside, when
theundergroundstructurewascompleted. Thesephe-
nomenaalso confirmedtheinfluenceof asymmetric
earthpressureonthemovement of theretainingwall.
3.2 Evolution of earth pressure
The monitored earth pressure was illustrated in
Figure 10a and 10b with calculated one. The calcu-
latedactiveearthpressurewasobtainedusingRankine
earthpressuretheory.
Fromfigure10aandfigure10b, it couldbefound
thatthemonitoredearthpressureattop15mwasvery
218
0
5
10
15
20
25
30
0 50 100 150 200 250 300 350 400
11/11/2005
12/2/2005
1/3/2005
1/13/2006
active pressure
d
e
p
t
h
/
m
earth pressure/kPa
Figure 10b. Comparison of recorded earth pressure with
calculatedactiveearthpressureatTY1.
closetotheactivepressurefor bothmonitoredpoints.
Itbecamemuchlargerthanactivepressurebelow15m.
Two factors could be contributed to the distribution
of theearthpressure. Onewas that themagnitudeof
lateral displacementof theretainingwall waslarger in
thetop15mthanthat of below15m, andit couldbe
verifiedfromFigures58.Theothercausewasthatthe
small bulkof thesoil bodyagainsttheretainingwall at
top15m. It couldbefoundfromFigure3that thesoil
bodyinthetop15mwasmuchsmaller thaninbelow
15m. Bothfigure10aandfigure10bimpliedthat the
soil bodyhadasignificanteffectonthedistributionof
earthpressureagainst theretainingwall.
However, theearthpressureatnorthsideof thedeep
excavationwas not monitored. No comparisoncould
beperformedbetweenthetwosides.
4 MODELLINGOF DEEP EXCAVATION
2-Dnumerical modellingwascarriedout usingFEM
codeof Plaxisv8consideringthenarrowplanechar-
acteristic of the deep excavation. The cross section
11showninFigure2was adoptedinFEM analysis
becauseitwasalmostthecenterof thedeepexcavation
andnear thehistoricbuildingaswell.
4.1 Numerical model
Theoverall widthof deepexactionatcrosssection11
was100mwithexcavationdepthof 13m.Thewidthof
thenumerical model was240m, whichwas18times
as wide as the depth of the excavation. The vertical
dimensionwas50m, whichwasmorethan3.5times
thedepthof theexcavation.Themodel dimensionwas
Figure11. TheFEM mesh.
Table3. Parametersof retainingpilesandgroundimprove-
ment.
Elasticmodulus
kPa Poisonsratio
Retainingpiles 3.310
7
0.15
Strut 3.010
7
0.15
Groundimprovement 1.410
5
0.20
large enough to lower the boundary effect. A linear
elastic model was adopted for the ground improve-
ment. The retaining wall as well as the strut was
simplified as elastic beamin numerical modelling.
ThesoilsweresimulatedwithMohr-coulombmodel.
Thenumerical simulationwasperformedwith15-node
isoparametricfiniteelementsunder theassumptionof
planestrainconditions.TheFEMmodel waspresented
inFigure11.
The boundary conditions in the numerical simu-
lation contain the following two types, one was the
displacement boundary condition, and theother was
thedrainagecondition. A freedisplacementboundary
condition was adopted at theground surface. It was
assumedthat nohorizontal nor vertical displacement
takenplaceat thelower boundary, for it was beyond
theinfluenceof deepexcavation. Thelateral displace-
ments at left andright handboundary wereboth
fixed as zero. The drainage condition at the ground
surfacewasassumedtobefree, hencetheexcesspore
pressurewas kept as zero along theground surface;
meanwhile the lower boundary as well as the left
andright handboundaryconditionwereconsidered
tobekept ashydrostaticporepressureduringexcava-
tion. Theinitial effectivestressesandhydrostaticpore
pressurewerecalculated based on theweight of the
soil andtheundergroundwater condition.
4.2 Parameters used in numerical modelling
The parameters of retaining structure and ground
improvement used in the numerical analysis were
listedinTable3. Thesoil parameters couldberefer-
encedasTable1.Theinterfacebetweenretainingpiles
andsoil wasadoptedandtheinterfaceparameterswere
determinedaccordingtoPlaxismanual. Themodulus
219
0
5
10
15
20
25
30
0 10 20 30 40 50
1.4 m
6.2 m
13 m
lateral displacement/ mm
d
e
p
t
h

/
m
Figure12. Evolutionof lateral displacementsat southside
of deepexcavation.
Figure13. Evolutionof lateral displacementsat northside
of deepexcavation.
of resilience was adopted for soils. The modulus of
resilience was obtained by back-analyzing the mon-
itored lateral displacements of the first excavation
progressshowninTable2. Itwasfoundthatthemodu-
lusof resiliencewasas5timeshighasthecompression
modulusfor soils.
4.3 Numerical modelling procedure
Theexcavationwasmodeledwith7consecutivesteps
shownasfollowing: STEP1wastodeterminetheini-
tial stressstateduetothegravityof soils. STEP2was
usedtoexert theloadingof existingbuildingsonthe
surface with the magnitude of 6070kPa according
tothetypeof thebuildings. Themovements induced
inSTEP 1and2werereset tozerointhemodelling.
STEP 3representedtheconstructionof retainingwall
andgroundimprovement, thesurchargeof 20kPawas
alsoloadedat thisstep. STEP 4meant thefirst exca-
vationto1.4mdeepandtheconstructionof firststrut.
InSTEP5, thesecondstrutwassetafter excavatingto
6.2mdeep. Excavatingto10mdeepandthethirdstrut
wasfinishedinSTEP6.Theexcavationwascompleted
andbottomplatewasconstructedinSTEP7. Dewater-
ingwasconsideredduringtheexcavationbychanging
water table.
4.4 Calculated lateral displacements
Figures 1213 presented the evolution of lateral
displacements with excavation progress. The lateral
displacements of south side of the retaining piles
reached 28.8mm, 38.4mmrespectively when exca-
vatedtothedepthof 6.2mand13m.Theyweresmaller
than thoseof north sideof theretaining wall, which
were 32mmand 60mm. It could be found that the
maximumlateral displacement at northsidewas 1.5
timeslargerthanthatof southsideof theretainingwall
bycomparingfigure12andfigure13.
Thecomparisonbetweencalculatedandmonitored
lateral displacementcouldbecarriedoutbecausecross
section 11 was coordinate with inclinometer CX3
andCX10. Themonitoredlateral displacementswere
57.8mmand 35mmat monitored points CX3 and
CX10correspondingtotheexcavationdepthof 13m,
whiletheaccordinglycalculatedoneswere60mmand
38.4mmrespectively. Theagreement betweencalcu-
lation and monitoring implied the validation of the
simulation procedure of FEM modelling with back-
analysisonthemodulusof resilienceof soils. Besides,
theconsiderationof maininfluential facts, suchassur-
facesurchargedueto existing loading and piling of
construction material, theprocess of excavation, the
supplemented techniques of dewatering and ground
improvement, wasessential intheFEM modellingto
reasonablypredict thebehavior of theretainingpiles.
5 CONCLUSIONS
Thelateral displacementof Riversidedeepexcavation
with complex surrounding environment was studied
withmonitoringdataandFEM modelling. Thelateral
displacements of theretainingwall wereasymmetric
becauseof theasymmetric earthpressure. Themaxi-
mumlateral displacementatnorthsidewasalmost1.5
220
timesaslargeasthatof southsideof thedeepexcava-
tion. Theearthpressurewas closeto theactiveearth
pressureintop15mduetothelargelateral displace-
ment and small soil body against theretaining wall.
Theearthpressurewas muchlarger thanactivepres-
surebelow15m. Itwasfoundthesoil bodybulkhada
noticeableeffect onthedistributionof earthpressure
against retainingwall.
REFERENCES
Chang, C. T. & Sun, C. W. et al. 2001. Responseof aTaipei
rapidtransit systemTRTS tunnel toadjacent excavation.
Tunnelling and Underground Space Technology 16:
151158.
Yamagushi, I. &Yamazaki, I. et al. 1998. Studyof ground
tunnel interactions of four shieldtunnels driveninclose
proximity, in relation to design and construction of par-
allel shield tunnels. Tunnelling and Underground Space
Technology 13(3): 289304.
Zhang, D. M. & Huang, H. W. 2007. Ground movements
and controlling measurements in deep excavation under
asymmetricloading. Proceeding of 10th national confer-
ence on soil mechanics. Chongqing, 14 Novermber 2007
(in press). (inChinese).
221
GPSheight applicationandgrosserror detectioninfoundationpit
monitoring
H. Zhang
School of Safety and Resource Engineering, China University of Mining &Technology, Beijing, P.R. China
S.F. Xu
College of Architecture & Civil Engineering, Zhejiang University of Technology, Hangzhou, P.R. China
T.D. Lu
Department of Survey, East China University of Technology, Fuzhou, P.R. China
ABSTRACT: Theauthorintroducesadeformationmonitoringmodel combinedbytraditional measuringtech-
nology and modern GPS measuring technology based on technical attribute of foundation pit deformation
monitoringandauthorsexperienceof deepfoundationpit constructionproject of undergroundtunnel inLishui
Road, Hangzhoucity. WhenanalyzingGPSheightconversion, inorder toimproveGPSdatummarksreliability,
onecan useDixons test in GPS datummark reliability test to find out height anomaly, thus provideconve-
nienceto searchanddeletemarks withgross error. This test also improves deformationmonitoringprocesss
efficiency.
1 BACKGROUNDPROJ ECT INTRODUCTION
Lishui Road(fromHuzhouRoadto QingfangRoad)
projectisoneof HangzhouCitygovernments33929
engineeringproject. Thetunnel of theproject iscom-
posedbyU-tanksandboxculverts.1+5681+638,
1+7941+864 are U-tanks. Each tank is 70m
long and the width of banks is summed to 22m.
1+6381+794areboxculverts.Thesumof lengths
of all box culverts reaches 156m. Equally dividedit
into4parts, eachboxculvert is38.25mlongandthe
widthof all box culvertsissummedto21.4m. Rein-
forced concrete piles with diameter of + 100 steel
pipeweretakenas support. They are21mlongwith
concrete outside. The concrete piles with +60@30
wereusedtokeepdry fromwater. Thedepthof con-
cretepileis 10m. They areconnected to each other
side by side. The steel in the shape of I is used as
the inside supports. The distance between two sup-
ports is 6mwide. Thedepthof thefoundationpit is
8m. Thisfoundationpit islevel 2foundationpit. The
constructions 0.000mlevel is equal to Huanghai
height+4.125m.Thesituationaroundsiteareaisquiet
complex,especiallyJ inghangCanal onthewestsideof
siteandancientmunicipal heritageGongchengBridge
whichisclosetothenadirof undergroundlot, smallest
distanceisabout 2m.
2 GPSHEIGHT APPLICATION
GPS positioning technology has advantages such as
no need of keeping vision between measuring sta-
tions, not restrained by weather conditions, able to
measuring the targets 3D displacement and highly
automated. Theaccuracy of short distancedeforma-
tion monitoring can reach minor millimeter level
[1]
,
thusprovidesanewmethodfor high-accuracy defor-
mationmonitoringof largeconstructionandfounda-
tion pit. In Lishui Road projects case the visibility
condition in foundation pit construction site is bad
and most datummarks cant share vision, monitor-
ing marks and datummarks are in different height,
and also thereis a across-river benchmark problem.
To solve these problems above, this project take a
monitoring plan using both modern and traditional
measuring technology: using GPS technology to set
up a 3D datummark network, and using traditional
measuring methods to monitor after the network is
established
[2]
.
AfteradjustingGPSmeasuringresults,theoutcome
height is geodetic height H
GPS
relevant to WGS-84
ellipsoid. Sincethebenchmarkheight(normal height)
is using in foundation pit engineering application,
the geodetic height H
GPS
should be transferred into
normal height H
0
in this project. The difference
223
between normal and geodetic height is called height
anomaly
[3]
:
In solving GPS height anomaly, known marks
height anomalyvaluesreliabilityiscrucial tosolving
result accuracy. Becauseof restraint fromsitecondi-
tion, it isimpossibletohaveenoughGPSmarksmeet
benchmarks or taking benchmark co-measuring. So
every singlemarks height anomaly valuewill make
considerable affect to calculating result accuracy, a
markwithgrosserrorheightanomalyvaluecouldeven
lead to a totally useless result and complete failure.
Thustheinitial datashouldtakeagrosserrortest. Dur-
ingthetest, thedataisnormallycheckedbygeometric
conditional closure, like triangle closure in triangle
network or pole condition closure, which monitor-
ing valuemust meet or by residual fromadjustment
error. Since gross error is hard to distinguish from
limited error, this method is hard to discover small
grosserror. Alsoit ishardtofindapplicablegeomet-
ric conditionclosureduringGPS height transferring.
To solvethis problem, onecanpick upsometrustful
geometricbenchmarkspot height andgeodeticheight
in the GPS network to fit other benchmark height,
or pick somespot separately to processingrepetitive
trail calculation, thenobtainother measuredgeomet-
ric benchmarks trail height with mathematic model
fromfit and using the equation below to obtain fit
residual:
H
i
,
i
is trail height and trail height anomaly, H
i
,
i
is measured benchmark height and measured height
anomaly. Then, one can use residual to process rel-
evant spots measured benchmark height gross error
test, after carefully analysis of measured value with
gross error, select enoughreliablemeasuredvalueto
runfit again.
3 HEIGHT ANOMALY GROSSERRORTEST
METHOD
Accordingto DixonTest
[4]
, assumetherereaset of
residual V
1
, V
2
, V
n
, sortthemfromlowtohigh, and
get asequencelikebelow:
Thenwehave:
If onefromr
10
, r
11
, r
21
, r
22
andr
10
, r
11
, r
21
, r
22
islarger
than critical value, then wecan consider V
(n)
or V
(1)
as anomaly value. After analyzes the sensitivity of
anomaly inr statistics test, Dixonclaimedthat when
3n 7, itisbettertouser
10
orr
10
; when8n 10,
user
11
or r
11
; when11n 13, user
21
or r
21
; when
14n 25, user
22
or r
22
.
It isnatural tousedifferent statisticsdependingon
different n. When n is small, rangeestimation has a
betterefficiency, butwhilen becomelarger, rangeesti-
mations efficiency decreaseaccordingly. So when n
is relevant large, userangeV
(n)
V
2
or V
(n)
V
(3)
to
estimate. Statistics r
ij
or r
ij
s critical value is given
inr(n, ) inreference
[4]
. isType1probability, also
calledsignificance. Itsvalueusuallyis0.05or 0.01.
Whenrunningthetest, onecancalculateanddis-
criminatefrombothends of residual sequencesepa-
rately, until thereis no gross error suspicion in both
endsof thetest.
4 GROSSERRORTEST EXAMPLE
InLishui Roadundergroundtunnel foundationpitcon-
structionproject, thedatummarksarethedeformation
monitoringdatumcontrol system. Sotheyareusually
built in the area outside and far fromthe construc-
tionsitetomaintaintheir stability. Theyshouldnotbe
too far though for theconsideration of havingbetter
monitoringaccuracy andalsofor our convenienceof
work. Our monitoringnetworkisdividedintotwolev-
els.Thefirstlevel of monitoringnetworkiscomposed
bythedatummarksandworkingspots, measuredonce
a week to maintain its stability. Thesecond level of
thenetwork is set upby workingspots andmonitor-
ingspots, usingstabledatummarkstoverifyworking
spots. Six datummarks were set up: four are at the
eastbankandother twoareatwestbankof theancient
J inghangCanal.UsingGPStointroducethetwodatum
marks at west bank of thecanal tothecanals east in
favor of monitoring network. The datumnetwork is
surveyedfour times; followingtheofficial construc-
tionstandardsentitledGlobal PositioningSystemfor
UrbanSurveyTechniquestandards CJ J 73-97. Three
224
Trimble4600LSGPSsingle-frequencyreceiverswere
setuptoreceivethesignal atthesametime.Theobser-
vationtimelastedmorethan90min. Informationfrom
510satelliteswereefficientlyreceived.Theelevation
angleof satellitesis15degreeandabreakof 20sec
wassetfor everytwoobservations. 12baselineswere
observedandfourof themaretherepeatones. Specific
softwareprovidedbyAmericasupplier wasemployed
toprocessthedataandtocarryout theeffectivesolu-
tions. Themaximumof error is about 5mmwhile
theminimumis2mm.Theobservationresultswere
further checkedbytimesynchronizedandunsynchro-
nizedcircle. Datumheightnetworkmonitoringdatais
fitfrom4spotsand16setsof dataof GPSbenchmarks
geometricbenchmarkheight residual.
Running Dixon test, first discriminate the largest
residual V
(16)
, since n =16, so take r
22
and r
22
as
statistics.
Using n =16, =0.05 as argument, according to
table
[4]
, r
0
(16, 0.05)=0.507, sincer
22
-r
0
(16, 0.05),
the conclusion is the geometric benchmark height
whichV
(16)
referstodoesnthavegrosserror. Discrim-
inationof smallest residual V
(1)
:
Asr
22
>r
0
(16, 0.05), theconclusionisthegeometric
benchmarkheightwhichV
(1)
referstohasgrosserror,
shouldbeeliminated.
After theeliminationof residual V
(1)
, thebothends
test shouldberunagain.
First discriminatethelargest residual V
(15)
Using n =16, =0.05 as argument, according to
table
[4]
, r
0
(15, 0.05)=0.525, sincer
22
-r
0
(15, 0.05),
the conclusion is the geometric benchmark height
which V
(15)
refers to doesnt have gross error. Then
test thesmallest residualV
(1)
:
Since r
22
>r
0
(15, 0.05)=0.525, and r
22
-r
0
(15, 0.01)=0.616 it can be concluded that the geo-
metric benchmark height whichV
(1)
referstodoesnt
havegrosserror.
5 CONCLUSION
WithGPStechnologyswidelyapplication, peoplecan
simply and efficiently obtain horizontal accuracy of
certainspot onminor millimeter level, but still cant
obtain the spots height on same accuracy level. So
in order to extend GPSs superior ability in survey-
ing 3D displacement, we should put our efforts on
researchinghowto improveGPS survey accuracy of
vertical displacement, thus it can match with survey
accuracyof horizontal displacement. Thereasonwhy
GPS has alowsurvey accuracy of vertical displace-
mentisthatthoughGPScouldprovideahighaccuracy
geodetic height, thelack of ageodic model withrel-
evant accuracy lead to a serious accuracy decrease
duringtransferringfromGPS geodetic height tonor-
mal height. To seek theGPS height anomalys value,
thereliabilityof knownspots heightanomalyvalueis
critical toresultsaccuracy
[5]
,isthekeytoimprovever-
tical deformation accuracy. To apply theGPS height
survey inour projectsfoundationpit monitoring, the
questionsbelowshouldbeconsidered:
1. Height anomaly is unstable, it maybe smooth in
small range or flat-contour region, where height
datumnetworkof foundationpitmonitoringisoften
established, thusiseasytoseekanomalyvalue; but
it isveryvariant inwiderangeor complexcontour
region, possible to occur several value with high
residual. Soinorder toimprovereliabilityof gross
error detection, when discriminated an anomaly
value, oneshouldanalyzecarefullybeforedeleteit.
2. Incalculationof GPSheightanomaly, thesourceof
error isvarious; itcouldbesurveyingerror of GPS
geodeticheight or GPSgeodeticheight difference,
orerrorfromgeometricbenchmarksurveying.This
problemdirectly leads to adifficulty of deciding
error distributionpatternfor heightanomaly. Since
test methodusually runinacertainerror distribu-
tion pattern, (e.g. Dixon test, requires residual is
randomsamplefromnormal distribution) thecred-
ibilityof usingthistesttorungrosserror detection
isdecreasedinreal application.
3. Many factors couldaffect GPS height component
accuracy.Variousmeasuresshouldbetakentoguar-
anteetheaccuracy in specific projects. Minimize
themultipath effect when surveying with GPS in
urbanarea, choosinggeodetictypeof GPStomon-
itoring datumduring foundation pit construction
period. Experienceprovedthatusingthesemethods
notonlyavoidtherestrictionsbroughttositecondi-
tionsfromconventional methods, butalsoimprove
workingefficiencyandassureconstructionquality.
GPS static relativepositioningsurvey hastremen-
douspractical significancetopreciseengineeringsur-
vey. Withthereasonablemonitoringplanaccordingto
engineeringconditionandpurpose, itsaccuracycould
225
meetalmosteveryrequirementsof preciseengineering
survey. It also has multiple advantages such as low
cost, highefficiencyandahighdegreeof automation.
TheapplicationinLishui Roadfoundationpit project
isauseful experience.
REFERENCES
Li, Z.H. & Huang, J.S. GPS measurement and date process.
Wuhan: WuhanUniversityPress, 2005
Lu, T.D. Zhou, S.J. GuanY.L. Height anomaly Gross Error
Test andAnalysisinGPSHeight Conversion. Geotechni-
cal Investigation & Surveying, 2004(4) 5154
Yang, J.T. J iang Y.X. Zhou J. Analysis on Reliability and
Accuracy of Subsidence Measurement with GPS Tech-
nique. Journal of Geodesy and Geodynamics, 2006(1)
7075
Zhang, H. &Gu, J.S. Deformationmonitoringanddataanal-
ysis of foundation in municipal engineering. Journal of
Zhejiang University of Technology, 2003(5) 571574
Zhang, F.G. &ZhangJ.Y. Statistical Distribution and Test of
Survey Error. Beijing: ChinaMeasurement Press, 1991
226
Studyondeformationlawsunder theconstructionof semi-reversemethod
J. Zhang, G.B. Liu&T. Liu
Department of Geotechnical Engineering Tongji University, Shanghai, P.R. China
ABSTRACT: Takinga24.09-m-deepfoundationpit of Shanghai Metro Line1whichuses thesemi-reverse
constructionprocessof threeopenexcavating-onetunneling asanexample, throughgatheringandanalyzing
fieldmonitoringdataandmakinguseof forwardandbackanalysismethods, wefoundout deformationlawsof
foundationpitundertheconstructionof semi-reversemethod.Theimplementationresultsof thisprojectindicated
thatthesemi-reversemethodisaneffectivewaytoimproverigidityof theexteriorsupport,control thedeformation
of excavation, and ensuresafety of thesurrounding buildings and pipelines. Meanwhile, theresults coincide
essentiallywithtime-spaceeffect.Thedeformationof theexcavationiscloselycorrelativewithexcavationspeed
andexposuretime. It providedsomeuseful referencefor thedesignof deepexcavationinsoft soil.
1 INTRODUCTION
With the development of urban construction, more
attentions havebeenpaidto theutilizationof under-
ground space, and theconstruction technology level
of foundationengineeringhasbeenimprovedcontin-
uously. At the existing construction process of deep
foundation engineering, open excavation method is
the most common construction method at present.
Because it boasts many advantages, such as more
constructionoperationsurface, short periodandless
cost. However intheapplicationof somedeepexcava-
tions which havecomplicated adjacent environment,
narrow operation space, and complicated geological
conditions, openexcavationmethodwouldcausegreat
influence on the traffic flow. At the same time, the
pollution of mud fluid, dust particle, acoustic noise,
andvibrationwhichcausedintheconstructionwould
inducediscommoditytotheresidents life. Especially,
openexcavationmethodwouldgoagainst withdefor-
mationcontrol of pits, whichwouldcauseperimeter
buildings and structures cracking, and bring great
economic loss or unfavorable social influence. The
completereversemethodhas littleeffect onadjacent
environment, but itsspeedof excavationisslow, con-
structiontechnologic process is complicated, andthe
cost of pillar pilesishigh.
Combined with advantages of open excavation
method and complete reverse method, semi-reverse
constructionmethodemergesasthetimerequire, and
hasbeenusedmoreandmorewidelyinShanghai deep
foundation constructions. Taking a pit of Shanghai
metro line 1, which uses semi-reverse construction
method, asanexample, throughgettingfieldmonitor-
ingdataandsettingupthefiniteelement model, this
paper has given an evaluation for thecharacteristics
of semi-reversemethodsuchasconstructiontechnol-
ogyanddeformationcontrol laws, kindlyexpectedto
provide with a beneficial reference to those similar
projectsinfuture.
2 ENGINEERINGCASE
2.1 General engineering situation
A railwaystationof Shanghai Rail TransitLineNo. 10
(metroline1) issituatedattheintersectionof SouthXi
ZangRoadandFuXingRoad, andcrosstransferred
with metro line8. Thegeographical position of this
stationisshowninFigure1.
Thestationof metroline1isbelowthatof line8.The
structureformof thissubwaystationwiththreefloors
is two pillars andthreespans, theoutsidedimension
are 179.2m(length) 23.8m(width). And the size
of east and west end well is 27.8m16.1m, which
bottomfloor burieddepthare24.06m, 24.09m.
Accordingtotherequirementsof waterproof design
andconstructionplan, thewholerailwaystationmain
bodystructureisdividedintotwoconstructionregion
witheight parts. Thesubsectionconstructiondrawing
of thisstationisshowninFigure2. Thewest endwell
is the first construction part, which requires higher
environment protection. Thisendwell approachesthe
227
Figure1. Geographical positionof thestation.
Figure2. Subsectionconstructionof thestation.
Figure3. Distributionof monitoringpointsinwestendwell.
International Squire(28floors) andShenNengInter-
national Building (26 floors), and surrounding with
lotsof pipelines. Accordingtotherequirementsof the
firstclassenvironmentprotectionspecifiedforShang-
hai subway station, the horizontal deformation of
diaphragmwall shouldbe1.4H(Histhedepthof
excavation), andthemaximumsettlement of perime-
ter groundsurfaceshouldbe1H. Theexcavation
depthof thewest endwell isabout 24.09m. It adopts
theundergrounddiaphragmwall withwidth1000mm
anddepth44m.Thebracesystemapplies1piececon-
cretebraceof 900800and7piecessteel tubebrace
of +60916. Thedistributionof monitoringpointsis
showninFigure3.
2.2 Geological condition
Basing on the geological prospecting data, the soils
of engineeringsitearedividedinto 9layers fromup
to bottom. They arefill soil layer, silt clay layer,
mud-silt clay layer, muddy clay layer,
11
clay
layer,
12
silt clay layer,
3
silt clay layer,
4
silt
clay layer,
2
finesandlayer. Table1shows physical
andmechanical characteristicsof differentsoil layers.
Figure4showsGeotechnical sectionof excavation.
Mainhydrology conditionof this stationis as fol-
lows:theshallowgroundwaterfieldisphreaticaquifer,
whichmainlycomesfrominfiltrationof precipitation
andseepageof surfacewater.Theannual averagewater
stage of Shanghai ranges from0.50m0.70m, and
generally0.5mischosenasdesignvalue.
Inthereportof geological prospectingdata, thesoil
of
2
finesandlayer is distributedinthesite, whose
buried depth is 4446mand confined water head is
10.511.0m. Consideringtheworst factors, whenthe
pit excavated to 24m, thecoefficient of upheaval in
thebottomof thepit wouldnot meet therequirement
of safetyfactor, soit shouldbeadoptedmeasurement
for decreasingconfinedwater head.
2.3 Construction procedure
Considering the actual factors such as construction
period,trafficorganization,undergroundpipelinesand
environmentprotection, thesemi-reverseconstruction
processof threeopenexcavating-onetunnelingwas
adoptedinthisproject.
Thedetail of theprocessisasfollows: Firstlyexca-
vatethesoil tothefifthbrace, andthenconstruct the
secondmedianplatebetweentheforthandfifthbrace.
With thetop reinforced concretebrace, thereversed
medianplateandundergrounddiaphragmwall formed
aframesystem. Whilethepavement maintenanceof
median plate has been finished, utilize two shield
structureholes of theend well to dig thesoil below
theplateuntil thebottomplatefinished.
Thesemi-reversedconstructionhasbrought lotsof
inconveniencetoexcavatingandsupportingof thepit
under the second median plate. This inconvenience
generallyreflectsatthenarrowperpendicular channel
andthecomplicatedsupportsinstallation. Commonly,
installationprocedureof supportsunder theconstruc-
tionof semi-reversedmethodisthat, dividethebrace
into several pieces, bring thesepieces to thebottom
one after another, and then assemble themtogether
to thedesignelevation. For theprocess as it is men-
tioned, installing one straight brace generally needs
7hours, and installing one diagonal brace needs 10
hours, whichfar fromtherequirement of time-space
effect. Time-space effect requires that the exca-
vation width should be no more than 6m, and the
excavation plus supporting time should be no more
than24h(excavationtime-16h; supportingtime-8h).
Thedeformationshouldbecontrol ineffectively, if the
pit was not supportedwithinsuchtime. So basedon
theengineeringtraits, theproject excavatethesoil as
soonaspossiblewhileintheopencut period. It uses
opencutmethodtoexcavatethesoil until 0.5mbelow
228
Table1. Physical andmechanical characteristicsof different soil layers.
peakvalueof
Permeability
Buried consolidated Compression
coefficients(m/s)
depth gravity quickshear moduleEs(MPa)
No. (m) r (kN/m
3
) C (kPa) (
) Kv Kh
2.31
3.41 18.6 18 14 4.80
9.21 17.5 11 16 3.51 1.97E-92.18E-9 2.08E-09
19.81 16.7 13 12 2.27 1.17E-91.48E-9 1.33E-09
11
22.91 17.5 16 14 3.93 1.77E-9 1.77E-9
12
27.81 18.1 15 17.5 4.98 1.52E-9 1.52E-9
3
42.11 18.2 15 19.5 5.08 1.60E-92.13E-9 1.87E-09
4
44.71 19.7 39 15.0 8.00 9.3E-101.46E-9 1.20E-09
2
19.3 0 31 15.11 2.48E-82.57E-8 2.53E-08
Figure4. Geotechnical sectionof excavation.
thefifthbrace(3mbelowthesecondmedianplate).
Whentheseventhsoil wasexcavated, disassemblethe
supportsformthelocationof thefifthbraceandinstall
themtothelocationof theseventhbrace. Sotheinstal-
lationof everybracejustneeds0.5hours,whichshould
acceleratetheconstructionspeedaswell ascontrol the
deformationof thepit effectively.
3 FINITE ELEMENT MODELING
Combinedwithfieldmonitoringdata, thefiniteele-
ment softwarePLAXIS 8.2(Brink Greve& Vermeer
1998) was used to compute the response of the soil
aroundtheexcavationforeffectiveanalysis.Theprob-
lemwas simulated assuming plane-strain conditions
andchosenhalf of thepit astheresearchsubject. The
sideboundaries of themesh(total size90m75m)
wereestablishedbeyondthezoneof influenceof the
settlements induced by the excavation (Caspe 1966;
Hsieh & Ou 1998). Thefiniteelement mesh bound-
aryconditionswereset usinghorizontal restraintsfor
theleftandrightboundariesandtotal restraintsforthe
bottomboundary. Thesoil stratigraphy was assumed
tobeuniformacrossthesite. Thosesoilswithsimilar
propertieswouldbecombinedbyweightedsimilarity
method. Sosixsoil layerswerecompartmentalizedfor
thecalculationsimplify.
Thesoil model usedtocharacterizetheclaysinthe
PLAXISsimulationof theexcavationisthehardening-
soil (H-S) model (Schanz et al. 1999).
Thiseffectivestressmodel isformulatedwithinthe
framework of elastoplasticity. Plastic strains arecal-
culatedassumingmultisurfaceyieldcriteria. Isotropic
hardening is assumed for both shear and volumetric
strains. Theflowruleis nonassociativefor frictional
shearhardeningandassociativeforthevolumetriccap.
Theinitial valuesof thebasicH-Sinputparametersfor
thesoil layersarereferencedasTable1andcalibrated
byinverseanalysis.
Thelinear spring-layer model is adopted to simu-
late the braces; the plate element model is adopted
tosimulatetheundergrounddiaphragmwall, reversed
medianplateandbottomplate. Consideringthebuild-
ingsaroundthepit, 50kN/m
2
overloadisappliedfor
calculation.
Figure5showsthecalculationmodel.Table2shows
11calculationphasesandtheconstructionstagesused
in the finite element simulations. PLAXIS employs
a penalty formulation so that undrained conditions
canbeexplicitly modeled. Becausetherewas along
timeinterval between Phase6 and Phase7, thedis-
placementsareduetopartiallydrainedconditions. So
consolidationshouldbeconsideredinthisstage. Other
stages whichnot notedas consolidation inTable2
weremodeledasundrainedandtheexcessporewater
pressureswerecomputedrelativetosomesteady-state
value(1m) that changeswithdredgelinelevel.
229
Table2. Calculationphasesandtheconstructionstagesusedinthefiniteelement simulations.
Identification Phaseno. Calculation Stages
Initial equilibrium 0 Plastic construction
Set updiaphragmwall andapplyoverload 1 Plastic construction
Excavatethefirst soil andsupport thefirst brace 2 Plastic construction
Excavatethesecondsoil andsupport thesecondbrace 3 Plastic construction
Excavatethethirdsoil andsupport thethirdbrace 4 Plastic construction
Excavatetheforthsoil andsupport theforthbrace 5 Plastic construction
Excavatethefifthsoil andsupport thefifthbrace 6 Plastic construction
Construct medianplate 7 Plastic constructionand
consolidation
Excavatethesixthsoil andsupport thesixthbrace 8 Plastic construction
Excavatethelast soil 9 Plastic construction
Construct bottomplate 10 Plastic construction
Figure5. Calculationmodel.
4 COMPARISONOF FIELDDATA WITH
CALCULATIONRESULTSOF FINITE
ELEMENT SOFTWARE
Inorder tostudythedeformationlawsunder thecon-
structionof semi-reversedmethod, lotsof fieldmoni-
toringdataof west endwell havebeenfinishedfrom
thesecondbracehasbeensupportedtotheroof plate
hasbeenfinished. Choosetheinclinationsurveypoint
CX3andthesettlement points J 6-1, J 6-2, J 6-3, J 6-4,
J 6-5 which have the same cross section with CX3
as representative points. Combined with calculation
results of finite element software, it could be got
detailedanalysis.
4.1 Inclination deformation of underground
diaphragm wall
Intheconstructionprocessof foundationpit, thedis-
placement curveof inclinationpoint CX3at different
depth which changed with the working condition is
showninFigure6.
FromFigure 6, it could be found that the maxi-
muminclination displacement of diaphragmwall is
only33.69mmwhenthebottomplatehasbeenpoured,
which is satisfied with the requirement of Class 1
accumulative displacements (mm)
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36
d
e
p
t
h

(
m
)
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
support second brace
support third brace
support fourth brace
support fifth brace
construct median plate
support sixth plate
support seventh brace
construct bottom plate
Figure 6. Displacement curve of CX3 at different depth
whichchangedwiththeworkingcondition.
environment protection. Basedontheexperiences of
Shanghai underground works these years, with the
similar excavation depth, excavation size, geological
conditionsandperipheral circumstance, if thefounda-
tionpitadoptsopencutmethod, thedeformationvalue
couldnot becontrolledat sosmall range.
Calculation results of finiteelement softwareand
field data were compared from the time that fifth
braceshadbeeninstalled, whichisshowninFigure7.
In the figure, dashed line represents the calculation
value, andsolidlinewithcirclerepresents measured
value. Table3showsthespecificcomparativevalue.
230
Theshapeof calculationcurvewasingoodagree-
mentwiththemeasuredcurve,andthemaximumvalue
of calculation deformation was in accordance with
fielddatawhilethebottomplatehasbeenconstructed.
Theresults showthat finiteelement methodcancor-
rectly reflect excavation deformation regularity. It
shows that thediaphragmwall engenderedcompara-
tivelargerdeformationwithintheperiodfromthefifth
bracesupportedtothemedianplateconstructed. This
isbecausethediscrepancyof thetwoworkconditions
lasts as long as 20 days. Though soils werent exca-
vated, exposuretimeforthefoundationpitwithbraces
wascomparativelong. Theexcavationfaceissituated
inmuddyclaylayerwhichhasverystrongflowprop-
erty, andthepermeabilityof thesoil isrelativelylarge
(
=1.7710
9
m/s). For abovereasonsthediaphragm
wall engenderedlarger deformation. Inthefiniteele-
ment calculation, consolidationhas beenconsidered,
soit couldcorrectlyreflect theactual deformation.
-2 0 2 4 6 8 10 12 14
d
e
p
t
h
(
m
)
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
measured value
calculation value
accumulative displacement(mm)
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
d
e
p
t
h
(
m
)
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
measured value
calculation value
Support the fifth brace Construct the median plate
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
d
e
p
t
h
(
m
)
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
measured value
calculation value
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
d
e
p
t
h
(
m
)
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
measured value
calculation value
Support the sixth brace Support the bottom plate
Figure 7. Measured versus computed horizontal
displacements.
Table3. Specificvaluesof measuredversuscomputedhorizontal displacements.
Project database Fifthbrace Medianplate Sixthbrace Seventhbrace Bottomplate
Completiontime 06-11-16 06-12-6 06-12-29 07-1-1 07-1-15
MeasuredValue Maximumvalue 12.33 27.94 30.25 31.85 33.69
Depth 14 19 19 21 21.5
Calculationvalue Maximumvalue 13.19 27.32 31.05 35.35 37.68
Depth 15 20 21 21 22
Inorder toanalyzetherelationshipbetweendefor-
mation of diaphragm wall and time, we chose the
department of the maximum deformation for filed
data CX3 (CX3-43 point whose depth is 21.5m) as
akey point. Thevariations of deformationwithtime
for thispointinthewholeexcavationconstructionare
inspectedasFigure8shows.
From Figure 8, we can find that though the
diaphragmwall engendered large deformation from
the fifth brace supported to the median plate sup-
ported, in the process of concrete maintenance, the
deformation stopped to growand it even had alittle
falling,andalsowhenthesoil underthereversedmedia
plate was excavated, the deformation rate is smaller
thanthatof previous. Withthetopreinforcedconcrete
supports, thereversedmedianplateandunderground
diaphragmwall could beformed as aframesystem,
whichconstrainedthespreadingof soil deformation.
The reduction of deformation rate in this phase has
releasedthecomparatively largedeformation, which
engenderedasaresult of soil creepinforwardphase.
This is beneficial for the reduction of foundation
deformationandassuranceof pit stability.
4.2 Ground settlement
Groundsettlement points J 6-1, J 6-2, J 6-3, J 6-4, J 6-5
areinthesamesectionwiththeinclinationpointCX3,
whichisdistributedwiththedistanceof 3mfor every
point fromtheedgeof pit. Figure9 shows thefield
settlement curvesintheprocessof construction.
From Figure 9, it shows that ground settlement
increased with excavation depth. When the bottom
plate has been finished, the maximumsettlement is
only 7.1mm. Thesoil presented alittleuplift at 6m
fromthe edge of excavation. With reference to the
actual engineeringproject, highpressurejet grouting
was used in the end well for the stability of shield
accesstotunnel. Itmightbethereasonthatcausedthe
soil uplifting.
4.3 Building settlement and pipeline settlement
ToMarch2007, whiletheroof platehasbeenfinished,
themaximumbuildingsettlementwasonly3.3mm,
whichis at F05point. Thevariations of deformation
231
-5.00
0.00
5.00
10.00
15.00
20.00
25.00
30.00
35.00
40.00
10-20-06 10-30-06 11-9-06 11-19-06 11-29-06 12-9-06 12-19-06 12-29-06 1-8-07 1-18-07 1-28-07
d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
second
brace
third
brace
fourth
brace
fifth
brace
median
plate
sixth
brace
seventh
brace
bottom
plate
Figure8. Variationsof displacement withtimefor CX3-43inthewholeexcavationconstruction.
Figure9. Fieldsettlementsintheprocessof construction.
Figure10. Variationsof deformationwithtimefor F05.
withtimefor F05areshowninFigure10. Asaresult
of highpressurejet grouting, intheformer phasesof
constructionthevertical deformationof F05presented
anupliftingtrend. It was not until J anuary 2007that
the soil deformation fell back. This was one of the
main reasons for theso small accumulativebuilding
settlement.
The conditions of pipeline settlement were as
follows: the maximum deformation point of gas
pipe was M01, whose accumulative settlement was
8.6mm; themaximumdeformation point of water
supply pipe was S02, whose accumulative settle-
ment was8.4mm; themaximumdeformationpoint
of rain pipe is Y01, whose accumulative settlement
Figure11. Time-historycurvesof M01, S02andY01.
was8.2mm. ChoosingpressurepipepointM01, S02
andnon-pressurepipepointY01askey point, whose
time-historycurvesisshowninFigure11.
Thesettlement trends of thesepipelines wereuni-
form,andthesetime-historycurveshapesweresimilar
tothat of inclinationpoint inFigure8. FromDecem-
ber 6thtoDecember 29th, whenthemedianplatehas
been maintained, thedeformation valueof pipelines
alsopresentedastableperiod. In1969, Peck put for-
ward stratumcompensation theory, which indicated
that theshapesandtheenclosedareaof lateral defor-
mationcurvescausedbyfoundationpitexcavationare
similar tothatof groundsettlementcurves. FromFig-
ure 11, it can be found that this similarity changed
uniformlywithtime, thatitistosaythegroundsettle-
mentchangedwiththelateral deformationatanytime,
whichisfavorablefor theenvironment protection.
5 CONCLUSION
1. Adopting semi-reverse construction method in
metro foundation pit could control the deforma-
tion of pit effectively, and decrease the influence
of excavationconstructiononitssurroundingenvi-
ronment. Semi-reverseconstructionmethodowns
a deep foundation support technology with prac-
tical value and brilliant prospects, which would
232
be further developed and applied in rail transit
construction.
2. The reversed median plate and underground
diaphragmwall formedaframesystem. Inthepro-
cessof medianplatemaintenance, thedeformation
of soil behindretainingwall wasstable. Whenthe
platemaintenancefinishedandthesoil excavated,
thedeformationratewas smaller thanthoseengi-
neeringworks whichadoptedopen-cut methodof
thesameconditions.
3. Intheprocessof reversedmedianplatesupporting,
alongperiodwasneededfor reinforcementassem-
bleand scaffold erection. As aresult of soil flow
property, larger deformation may begenerated at
thisperiod.
4. As theexcavation is in clay, longer times of con-
struction may result in partial drainage as well.
Consolidation of thesoil should beconsidered in
finiteelement calculation.
5. The shapes of lateral deformation curves caused
by foundation pit excavation are similar to that
of groundsettlement. Thissimilaritychangeduni-
formly together with time pass. It is conjectured
thatthegroundsettlementchangedwiththelateral
deformationshapeat anymoment.
REFERENCES
Brinkgreve, R.B.J. &Vermeer, P.A. 1998. Finite element code
for soil and rock Analysis. PLAXIS7.0manual, Balkema,
Rotterdam, TheNetherlands.
Liu, J.H. &Hou, X.Y. 1997. Foundation engineering manual.
Beijing: ChinaConstructionIndustryPress.
Richard, J., Finno, M. &Michele, C. 2005. Supportedexcava-
tions: observational methodandinversemodeling. Jour-
nal of geotechnical and geoenvironmental engineering
10.1061:826836.
Zhao, G.W. &Guo, H.B. 2006. Applicationof semi-reversed
constructionmethodtorail transit construction. Building
Construction 28(10):815818.
233
Comparisonof theoryandtest onexcavationcausingthevariation
of soilmassstrength
J. Zhou& J.Q. Wang
Department of Geotechnical Engineering Tongji University, Shanghai, P.R. China
L. Cong
ABSTRACT: Inviewof theexcavationunloadingcharacteristic, thevariationof soilmassstrengthisstudied
throughthetheoretical deductionandthetestanalysis. BaseedontheHvorslevsreal strengththeory, thestrength
ratioof theunloadingsoil andthenormal compressedsoil consideringthepore-water pressureisdeducedand
thetest simulating excavation is carried out. Through comparing dataof thetheory and test, thesoilmass is
causedtobeat theoverconsolidatedstate, andthesoil microstructureisdamaged, thenthesoilmassstrengthis
reducedintheunloadingprocess. Theanalysisresult of theoryandtest arehelpful tothefurther understanding
of theeffect of unloadinginexcavationonthevariationof thesoilmassstrength, whichareverysignificant for
avoidingproject accidents.
1 INTRODUCTION
With the rapid and remarkable development of city
construction, an increasingly large number of the
exploitationof theundergroundspaceshaveemerged,
suchas high-story building, theundergroundmarket
andundergroundgarageetc., whichneedtoexcavate
for buildingfoundation.Theexcavation, includingthe
influenceof thesoilmasss engineering property and
thevariation of theenvironment characteristic, have
been systematically studied by numerous scholars.
Rutledge(1944) summarized thesoil sampledistur-
bancetotheinfluenceof theunconfinedcompression
strength and the initial tangential modulus in stress
and strain curve, and the result showed that the ini-
tial tangential modulus of theremouldedsoil sample
is smaller than the one of the undisturbed value by
about 20%, some were only even 3%4%. On res-
onant column test, Drnevich & Massarsch (1979)
discoveredthat evenif thesoil samplesufferedfrom
the small disturbance, its initial tangential modulus
also obviously reduced. Broms (1980) pointed out
that the soil sample disturbance in the brittle soil
to the stress and strain curves influence was much
bigger than in the plastic soil. Zeng (1995) studied
the subway double lines shield tunnel construction
to the influence of the surface, the buildings and
the underground pipelines, and analyzed the tunnel
interval to the influence of stress and the displace-
ment of surrounding soilmass. Zeng & Pan (1988)
studied stress path to theinfluenceof theundrained
strength in excavation. Wei (1987) has studied the
relationof theexcavationunloadingandpassivesoil
pressure.
Several examplesof thecollapseof foundationpits
inthepasthadveryseriousconsequences,whichurged
thepeopletostudythedesignandconstructionof foun-
dationpitdeeply.Atpresentthemaintenancestructure
of foundationpit isdesignedandcalculatedby using
theelastic foundation beamlawor theelastoplastic-
ityfiniteelementmethod.Theroutine-testparameters
generally were adopted as the computation parame-
ters, which had not really considered theexcavating
and unloading to theinfluenceof soilmass strength.
Thevariationof soilmassstrengthafterexcavatingand
unloadingisstudiedthroughtheanalysisof theoryand
testinthispaper. Theresultof thestudyindicatesthat
theunloadinginexcavationhasinfluenceonthevari-
ationof soilmassstrength, whichcanbeof somehelp
toavoidproject accidents.
235
2 THEORETICAL DEDUCTIONOF SOILMASS
STRENGTHUNDER UNLOADING
After excavation, the surrounding soil can be seen
as the overconsolidated soil layer, and Wei (1987)
deduced the undrained strength of the excavation
unloadingsoft clay accordingtotheHvorslev (1960)
real strengththeory, whichwasthesameformulathat
Mayne(1980)obtainedtheundrainedstrengthof over-
consolidatedclay soil accordingto thestatistics of a
largenumber of testdata.After theexcavation, infact,
theeffectivestressof bottomsoil layer of foundation
pitisinunceasinglydevelopingandchangingprocess,
rather thanthestaticoverconsolidatedstatethatabove
formula derives. In this process, becauseof unload-
ing, thenegativeporewaterpressuredissipatesslowly,
andeffectivestressdecreasesgradually,andeventually
stops at theoverconsolidated state. According to the
Hvorslev real strengththeory, theundrainedstrength
of thesoilsafter excavationandunloadingisdeduced.
TheHvorslevstrengthformulaisasfollows:
where c
e
= p
e
; In normal consolidated soil, p
e
is
equal to the current effective stress (drained shear
strength) or the consolidation pressure (undrained
shear strength); In overconsolidated soil, p
e
is the
consolidation pressure that the test specimen fail-
uresporosityratiocorrespondsinthenormal pressure
dense curve. tg
e
is the increment ratio which the
shearing strength increases along with the effective
stresschangewhenthewater content isconstant; P is
theeffectivestress.
Accordingtothereal strengththeoryandthecritical
states concept, when soilmass has withstood a sim-
ple loadingunloading cycle in stress history, it can
beassumedthat thefailurepoint of overconsolidated
soil is coincidencewith thefailurepoint of thenor-
mal consolidatedsoil at thesamewater content when
stress path reaches at critical stateline, as shown in
Figure1andFigure2. Thentheeffectivestressof the
overconsolidatedsoil isasfollows:
Based on the confirmation of a large number of
tests, it can be considered approximately that the
shapes of undrained stress path of the normal com-
pacting soil sample is geometrical similarity under
thedifferent consolidationpressure, andthevariation
valueof effectivestressandconsolidationpressureare
inproportion(Wei 1987). Thereforetheir relationcan
beproposedfromthefollowingequation:
p
e c
A
E
C
B
p'
a
p'
b
p'
e
p' p p
over-compacting soil
normal compacting soil
normal compacting soil
c
r
i
t
i
c
a
l

s
t
a
t
e

l
i
n
e
initial compressed line
p
Figure 1. The constant consolidation pressure and the
undrained stress path for the normal consolidated soil and
theoverconsolidatedsoil (Wei 1987).
p
c
p
e
p log p
A
E
C
B
Figure2. Therelationcurves for thenormal consolidated
soil andtheoverconsolidatedsoil.
The undrained shear strength of overconsolidated
soil sampleB isasfollows:
Takingtheequation(2),(3),(5) intotheequation(4),
then:
236
The undrained shear strength of normal consoli-
datedsoilmassA ininitial consolidationpressureP
c
isasfollows:
InFigure2:
Ontheinitial compressionlineAEC,
The equation (12) is the undrained strength ratio
of the excavation unloading soil and the normal
consolidationsoil.
Figure3. Thediagramof thestressvariationforexcavation.
Intheabovedeductionprocess:
S
ub
The undrained strength of the excavation
unloadingsoils;
S
ua
Theundrainedstrengthof thenormal consol-
idatedsoils;
u
t
Thenegativepore-waterpressureof theunload-
ingsoilmass, withtimedissipation;
P
t
Thesoilmassconsolidationpressureof current
state;
p
c
The soilmass consolidation pressure under
natural state;
C
c
, C
s
Thecompressionindexandswellingindex
of soilmass.
Thelawis reflectedintheequation(12) that soil-
mass strength reduces gradually with the negative
pore-water pressuredissipationafter unloading. when
u
t
=0, soil is at thecompleteover-compactingstate.
The key that estimates the soilmass strength after
unloading is to determine the parameter i. Because
of theerrors of thesamplingdisturbanceandinstru-
mentation equipment, the computed result of the
parameteri whichisdeterminedbyconsolidationtests
result C
c
, C
s
, isbigger thanthereal value, needingto
calculatetheparameter i bythestrengthtest.
After the synthetical comparison, Mayne (1980)
propose that 1C
s
,C
c
takes the statistical average
value 0.64 to be quite reasonable according to the
statistics of a large number of experimental data
andinsitumeasurement; Furthermore, Zhang& Wei
(1987) haveconfirmedthis viewpoint. Thereforethe
parameter i istaken0.64inthispaper.
3 EXPERIMENT ANALYSISOF SOILMASS
STRENGTHAFTER UNLOADING
3.1 Test plans
In theexcavation process, as shown in Figure3, the
soilmassunitA of aroundfoundationpitwall islateral
237
Table1. Thetriaxial testplansof constantpressureconsol-
idation.
Consolidationpressure
v
=
H
(kPa)
Variationmodeof
thestress 100 200 300 400
Unloading
v
unloading I01 I02 I03 I04
failure
Unloading
v
loading J 01 J 02 J 03 J 04
failure
v
unloadingfailure K03
v
loadingfailure L03
Table2. Thetriaxial test plansof K
0
consolidation.
Consolidationpressure
v
=
H
/K
0
(kPa)
Variationmode
of thestress 180 240 400
H
unloadingfailure M01 M02 M03
unloading, and vertical pressureis nearly invariable;
ThesoilmassunitBof thelateral andvertical inthebot-
tomof foundationpitsimultaneouslyunloads, andthe
vertical stress drops morequickly thanlateral stress,
but still retains quiteapart of incompleteunloading
stress, therefore the influence of the unloading only
possiblyexistsinacertainscopebelowfoundationpit
bottomsurface.Thestageexcavationmethodisgener-
allyselectedforthefoundationpit.Thefirstsupporting
structureimmediatelybetakenwhenthefirstlayersoil
isexcavatedtoreachat thedesignelevation; Thenthe
secondlayer soil isexcavated. Duringtheexcavation,
the soilmass units are in the process that the soil is
unloadingandexpandingandthenegativepore-water
pressureisdissipatingslowly.
Accordingtotheunloadingcharacteristicof above
soilmassunitA, B, testplansaredesignedasshownin
Table1andTable2.
I, J grouptests simulatethestress pathof theunit
B, andthesoil samplesareconsolidatedfor 24hours
under constant pressure; thenaccordingto thestress
pathL
v
=
v
,3, L
v
,L
H
=2, thetestspecimenare
unloaded simultaneously on thevertical and thelat-
eral, and the value was recorded when the negative
pore-water pressure are stable after unloading; This
processneedsfor 23hour fromstartingunloadingto
stabilityof readingvalue, thenturnsonthedrainvalve,
and makes thenegativepore-water pressuredissipa-
tion, the soil sample completes consolidation under
thenewlowstresscondition, theconsolidationtimeis
8hours.ThentheI groupisloadedtothetestspecimen
compression failureon thevertical, and theJ group
is unloadedto thetest specimenextrusionfailureon
Table3. Thebasicindexof mechanical propertyof Siltclay.
Parameters Values
w 52.6%
16.9KN/m
3
e 1.487
Sr 97.3%
I
p
23.2
I
L
1.231
1-2
1.20MPa
1
c 11kPa
9.3
Table 4. The negative pore-water pressure data of soil

sampleafter unloading(kPa).
Serial number I (J ) 01 I (J ) 02 I (J ) 03 I (J ) 04
Negative 30.9 34.1
pore-water
pressure(kPa)
Note: Becauseof instrument failure, thedataof 01 and 02
cannotbedetected, thelatter twodataareaveragevalueof I,
J group.
thevertical. Asareferencetest, K groupandL group
specimen areconsolidated in theconfining pressure
of setting, withoutunloadingdisturbance, andrespec-
tively are loaded and unloaded to the test specimen
failureonthevertical.
Theunloadingstressvariationprocessof unitA is
simulatedbyMgrouptest. K
0
=0.6, theconsolidation
pressureisexertedbystagedloading; Foravoidingthe
accidental failureof thetestspecimen, thestagedload-
ingis divided10levels to exert; Theaxial stress L
isexertedineachlevel loading, at thesametime, the
confiningpressureL
H
=K
0
L
v
isexerting, consol-
idationtime24hours. After consolidationcompletes,
L
v
ismaintainedinvariable, andthetestspecimenis
compressedtofailurebythelateral unloading.
3.2 Test results
Thesoil samplesof testsarethetypical softsoil of the
Shanghai area, itsbasicindexof mechanical property
asshowninTable3.
After unloading, thenegativepore-water pressure
isshowninTable4.
4 THE COMPARISONANALYSISBETWEEN
THEORY RESULT ANDTEST RESULT
OF THE SOILMASSSTRENGTHAFTER
EXCAVATINGANDUNLOADING
The theory deduction and the test simulation about
soil strength of the excavation has been discussed.
238
Figure4. Thenormalizationstressof theunloadingsoil of
I group.
Figure5. Thenormalizationstressof theunloadingsoil of
J group.
Nowwewill discuss thecomparisonreuslt of theory
andtest.
Thestressstrainrelationcurveof differentconfin-
ing pressureof I, J two groups tests arenormalized,
and
m
=(
1
+2
2
),3=233kPa, thenthecurvescan
bedrawnasshowninFigure4andFigure5, inwhich
K Line, L linearethesoil stress-straincurvesof natu-
ral compactionafter normalization, andI02, I03, J 01,
J 02, J 04 are the normalizing stress-strain curve of
over-compactingsoil after theunloading.
After excavationunloading, thestress valueof the
soil samplestress-strainrelationscurveapproachesor
surpassesthestressvalueof thenatural normal com-
pacting soil as shown in Figure4 and Figure5; For
eliminating the test error, after unloading, soil sam-
plepeak value(
1
3
)
max
canbetakentheaverage
valueof eachnormalizedcurvepeak value, asshown
Table 5. The soil strength contrast between theory result
andtest result after unloading.
Project
Peakvalue
(
1
3
)
max
(kPa)
Natural Soil sample Test Theory Difference
Test soil of unloading result result value
number sample disturbance S
ub
/S
ua
S
ub
/S
ua
(%)
I 208.2 219.0 1.052 1.175 12.3
J 205.1 212.5 1.040 1.175 13.5
inTable5; For comparingwithtest result, theoretical
calculationis takenby theformula(12), theparame-
ter P
c
=300kPa, P
t
=233kPa, thetestresultsandthe
theoretical formularesults, areshowninTable5.
In the Table 5, S
ub
is the undrained strength of
excavation unloading soil in I, J series tests by
normalization; S
ua
is the undrained strength of the
normal compressed soils in K, L series tests by
normalization.
In above tests, the influence of negative pore-
water pressure(Table4) isconsideredinsoil sample.
According to the computation of the formula (12),
thesoilmassstrengthratioS
ub
/S
ua
is1.229and1.235,
which is higher about 5%6%than theratio of the
pore-water pressuredissipatingcompletely.
ThecomparisonfromTable5canbefoundthatthe-
oretical calculation result is bigger about 10% than
testresult. NotonlyThesoilmassiscausedtobeatthe
overconsolidatedstate, butalsothesoil microstructure
isdamaged, andthesoilmassstrengthisreducedinthe
unloadingprocess. Inthistriaxial test, soil samplesis
unloadedaccordingtothestress pathof thetest, soil
stress is redistributed, andconsolidated, theovercon-
solidatedsoil isformated, soil structureof theoriginal
systemisalsodamaged.
5 THEVARIATIONOF SOILMASSSTRENGTH
PARAMETERSAFTER EXCAVATINGAND
UNLOADING
Based on the stress-strain relation curves of tests,
according to the Mohr-Coulomb criterion, the soil-
mass strength parameters c$$ of the simulating
excavation unloading and the parameters c$$ of
routine-test arelistedinTable6.
The cohesion, the angle of internal friction that
unloadingfailureof I andM groupobtainedarequite
close, their c value is bigger than J group, but the
value is smaller, furthermore, the values of c$$
are obviously different from the result of conven-
tional consolidatedquick shear test. Becauseof lack
239
Table6. Thestrengthparameter valuec$$ of thesoilmass
failure.
Test number
A unit of B unit of B unit of Theresult of
M group I group J group conventional
(unloading (unloading (loading consolidated
Parameter failure) failure) failure) quickshear
Cohesion 26.5 24.3 7.36 11
c (kPa)
Angleof 15 13 16 9.3
internal
friction
(
)
of thesufficienttestdata, therelationshipbetweenthe
unloadingstressandthestrengthparametersc$$ can
not beobtained, whichneedsfurther test todetermine
whether theseparametershavetheinevitablerelation.
6 CONCLUSION
This paper chooses thetypical excavationas thetest
study object, designs and carries out different stress
pathsindoor triaxial testsinI, J, K, L, M grouptests.
Someuseful conclusions aredrawnby analyzingthe
influence of excavation on the result of theory and
test, which are very significant for avoiding project
accidents:
1. Bytheassumptionthatthesoil asoverconsolidated
soil with dissipation of the negative pore-water
pressure after unloading, the undrained strength
ratiobetweenthesoilsof excavationunloadingand
thenormal consolidated soils is deduced, namely
theformula(12). Withthedissipationof soil neg-
ativepore-water pressure, thesoilmass strengthis
reduced.Accordingtotheanalysisresult, thescope
of reductionisnotlarge. Undertheaboveteststress
condition, the strength ratio range of variation is
about 5%6%.
2. TheundrainedstrengthratioS
ub
/S
ua
fromthetests
issmaller about 10%thanundrainedstrengthratio
fromthe above theoretical formula computation.
Thedifferencevalueinthetestcanbeidentifiedas
theresult that of unloadingdisturbance.
3. Intheunloadingprocess, thesoilmassiscausedto
beattheoverconsolidatedstate,thesoil microstruc-
ture is damaged, and the soilmass strength is
reduced.
4. Thetotal stressstrengthparametersc$$ obtained
fromthe different stress path tests are much dif-
ferent fromthe parameters fromthe routine-test.
Due to the insufficiency of data, the relation-
shipbetweentheunloadingstressandthestrength
parametersc$$ cannot beobtained, whichneeds
further test todeterminewhether theseparameters
havetheinevitablerelation.
REFERENCES
Broms, B.B. 1980. Soil Sampling in Europe: State-of-the-
Art. Journal of the Geotechnical Engineering Div. 106:
6598.
Drnevich, V.P. & Massarsch, K.R. 1979. Sample Distur-
bance and Stress Strain Behaviour. ASCE Journal
of the Geotechnical Engineering Division 105(GT 9):
10011016.
Hvorslev, M.J. 1960. Physical component of the shear
strengthof saturatedclays. Research Conference on Shear
Strength of Cohesive Soils, ASCE: 169274.
Mayne, P.W. 1980. Cam-clay prediction of undrained
strength. Geotech Engrg Div ASCE 106(GT11):
12191242.
Rutledge, P.C. 1944. Relation of undisturbed sampling to
laboratorytest. Transactions ASCE, (109): 11551183.
Wei, R.L. 1987. Soft clay strength and deformation. Beijing:
Chinacommunicationspress.
Zeng, X.Q. 1995. Subway project double thread tunnel
parallel advancement interaction and construction
mechanics research. Shanghai:Tongji Universitydoctoral
dissertation.
Zeng, G.X., Pan, Q.Y. & Hu, Y.F. 1988. The Behavior of
Excavation in Soft Clay Ground. Chinese Journal of
Geotechnical Engineering 10: 1322.
240
Theme 2: Construction method, ground treatment,
and conditioning for tunnelling
Tenyearsof boredtunnelsinTheNetherlands: Part I, geotechnical issues
K.J. Bakker
COB, Delft University of Technology, Delft, The Netherlands
A. Bezuijen
Deltares, Delft University of Technology, Delft, The Netherlands
ABSTRACT: Ten years have passed since in 1997 for the first time construction of bored tunnels in the
Netherlands soft soil was undertaken. Before that date essentially only immersed tunnels and cut-and-cover
tunnelswereconstructedintheNetherlands.ThefirsttwoboredtunnelswerePilotProjects, the2ndHeinenoord
tunnel and theBotlek Rail tunnel. Sincethen aseries of other bored tunnels has been constructed and some
arestill under constructiontoday. At thebeginningof thisperiod, amongst othersBakker et al (1997), gavean
overviewof therisksrelatedtoboredtunnelsinsoft groundandaplanfor researchrelatedtothepilot projects
was developed. After that in1999the2ndHeinenoordtunnel openedfor thepublic, theJ ointedplatformfor
Boredtunnelling, inshortGPB, wasorganized, tocoordinatefurther researchandmonitoringof boredtunnels.
This platformis supervisedby theCenter for UndergroundConstruction. Inthis paper asummary is givenof
someof themost characteristicobservationsonthese10yearsof undergroundconstructionintheNetherlands.
Inthefirst part of thispaper thefocusisongeotechnical interactions, andstability, whereaspart twowill focus
moreonstructural relatedissues.
1 INTRODUCTION
In 1992 afact-finding mission was sent to J apan by
theDutchgovernment, whichreportedthat it should
be possible to construct bored tunnels in the Dutch
soft soil conditions. Up to that timeessentially only
immersedandcut-and-cover tunnelswereconstructed
intheNetherlands,asboringof tunnelsinsoftsoil con-
ditions, atthattime, wasconsideredtobetooriskfull.
Afterthisconclusionthingswentquitefast; in1993
the Dutch minister of Transport and Public works
orderedtheundertakingof twopilot projects, the2nd
HeinenoordTunnel and theBotlek Rail Tunnel. The
projects were primarily aimed at constructing new
infrastructure and besides that for monitoring and
researchinorder to advancethedevelopment of this
new construction method for the Netherlands. The
projects started in 1997 and 10 years have passed
sincethen.
Atthestartof thepilotprojects, thedifficultieswith
respect toconstructingboredtunnelsinsoft soil con-
ditionswereevaluatedandaplanfor monitoringand
research was put forward, see Bakker et al (1997).
Sincethen, the2ndHeinenoordtunnel, seeFig. 1, and
aseriesof other boredtunnelswereconstructed.
After the completion of the pilot projects a J oint
PlatformforBoredtunnelswasestablished(GPB) that
coordinatesthemonitoringandresearchatthevarious
Figure1. Geological profileat the2ndHeinenoordtunnel.
otherDutchtunnellingprojects.TheGPB, aninitiative
of thelarger clientsfor undergroundinfrastructureon
thegovernmentside, wasorganisedunder supervision
of theNetherlandsCentrefor UndergroundConstruc-
tion; COB. Theresearchwasorganisedinsuchaway
that resultsof aproject wouldbebeneficial for anext
project startingalittlelater.
Unquestionablyalothasbeenlearnedfromtheper-
formed monitoring and research. The results of this
processhavebeennoticedabroad. In2005theNether-
lands hosted the fifth International symposium of
TC28onUndergroundConstructioninSoftGround.
Researchersandexpertsfromall over theworldcame
toAmsterdam, tolearnabout theDutchobservations
ontunnellingandto visit theconstructionworks for
thenewNorth-SouthcitymetrosysteminAmsterdam.
243
Table1. Boredtunnelscompletedafter 1997intheNetherlands.
Completion Boredlength Depth OutwardDiameter
(Year) (m) (m) (m)
2ndHeinenoordtunnel Road 1999 945(dual) 30 8.3
WesternScheldt tunnel Road 2003 6700(dual) 60 11.30
BotlekRail tunnel Rail 2004 1835(dual) 26 9.60
SophiaRail tunnel Rail 2005 4000(dual) 27 9.60
PannerdenschCanal Rail tunnel Rail 2005 1615(dual) 25 9.60
GreenHart tunnel Rail 2006 8.620(single) 30 14.90
The above event was also the occasion for the
presentationof abook; A decadeof progressintun-
nellingintheNetherlands by BezuijenandvanLot-
tum(2006), wherethis researchis describedinmore
detail.
This paper(s) gives some highlights of the main
researchresult of thepast decade.
2 REVIEWOF THE 1997SITUATIONAND
WHAT CAMEAFTER
Inthedesignphasefor the2ndHeinenoordtunnel a
main concern were the soft soil conditions in com-
binationwithhighwater pressures. Ingeneral inthe
Netherlandsthewater tableisjust belowthesoil sur-
face. Furthermorethe8.3moutwarddiameter for this
first largediameter tunnel was amajor stepforward,
comparedtopastexperienceintheNetherlands; expe-
rience that was mainly based on constructing bored
tunnels, pipesor conduitsuptoabout 4.0mdiameter.
Thisgavesomeconcernwithrespecttotheamountof
extrapolationof empiricknowledge.
Withrespecttothesoft-soil conditions,thelowstiff-
nessof theHoloceneclayandpeatlayersandthehigh
groundwater table; nearlyuptothesoil surface, were
consideredapotential hazardandachallengeforbored
tunnels.Thesoil profileatthe2ndHeinenoordtunnel,
seeFig. 1, isindicativefor theheterogeneouscharac-
ter and on occasion sudden changes in underground
soil layeringthat onemight encounter. Inadditionto
theheterogeneityandthegroundwater, deformations
dueto tunnellingmight influencethebearingcapac-
ity of any existing piled foundations in the vicinity.
And as the common saying is that the Amsterdam
Forest is underground, one might realize the poten-
tial risksinvolvedfor theNorth/SouthMetroworksin
Amsterdam.
Characteristic for ahigh water tableis buoyancy;
theeffect that thetunnel might befloatingupintothe
softupperlayersabovethetunnel duetothegradientin
thegroundwaterpressure. Besidestheriskof breaking
upof thesesoil layers, therather flexiblebeddingof
thetunnel and thedeformations that this may cause
needtobeanalysed.
Therefore research was aimed at clarifying the
effects of thesoft underground, groundwater effects,
andtheeffect of tunnellingonpiledfoundations.
After thesuccessful construction of thetwo Pilot
projects, anumber of other boredtunnellingprojects
followed, seeTable1. MentionworthisthattheGreen
Hart Tunnel holds until recently the record as the
largest diameter boredtunnel intheworld.
Still under construction arethetunnels for Rand-
stadRail inRotterdam, theHubertusTunnel for aroad
in The Hague and the North/South metro works in
Amsterdam.
Withrespecttotheconstructionof theNorth/South
metro works in Amsterdam, the station works have
madequitesomeprogress andtheboredtunnel is in
a preparation phase. The elements of the immersed
tunnel; theextension toAmsterdamNorth under the
river IJ, arewaitingfor thecompletionof theimmer-
siontrenchunder theAmsterdamCentral Station. For
theboredtunnellingpart, theTBMisexpectedtostart
excavationat theendof 2008.
Tenyearsafterthepilotprojects, thequestionarises
whether the observations and related research have
confirmedtheaboveissues to bethecritical ones or
that advancinginsight mayhaveremovedtheseissues
fromthestage and swapped thesefor other topics
givingmoreconcern.
Inthispaper someof thecharacteristic eventsand
resultsof thispastdecadewill bedescribed.Thechoice
for the topics being discussed is influenced by the
projects that both authors wereinvolved with, with-
outintenttominimizetheimportanceof otherresearch
thatisnotdiscussedinthispaper.Furtherissuesrelated
to groundwater effects andgroutingaredescribedin
moredetail inaseparatepaper inthissymposiumby
Bezuijen&Talmon(2008).
3 EXPERIENCESWITHBOREDTUNNELSIN
THE NETHERLANDSINTHE PAST DECADE
3.1 An instability of the bore front
Duringtheconstructionof the2ndHeinenoordTunnel,
approximatelyinthemiddleunderneaththeriverOude
Maasaninstabilityat theexcavationfront developed,
244
Figure 2. Support pressures before, during and after the
Blowout at the2ndHeinenoordtunnel.
see Fig. 2; afterward commonly referred to as The
Blow-out (seealsoBezuijen& Brassinga, 2001).
Backtrackingthesituationlearnedthat after that a
pressuredrop was observed, in his efforts to restore
frontal support, themachinedriver first pumpedben-
tonitetotheexcavationchamber; consideringadefi-
ciencyinthebentonitesystem.Whenthisdidnothelp,
air waspumpedtotheborefront; notrealizingthatthe
front itself already had collapsed. This collapsecre-
ated ashortcut between theexcavation chamber and
the river. The action of pumping air was noticed by
shipmasters on theriver, which reported air bubbles
risingtothewater surface, whichcausedthefailureto
beknownastheblow-out. Inthiscasethepumping
of air hadnotbeenbeneficial totherestorationof sta-
bilitybecausepressurelosswasnot thecausebut one
of theresultsof theevent.
This frontal stability at the 2nd Heinenoord tun-
nel has attracted some public attention. Presumably
it is less knownthat loss of frontal stability has also
occurredsincethenwithsomeregularity at theother
tunnelsunder constructionintheyearsafter, e.g. dur-
ing construction of the Sophia Rail Tunnel and the
Green Hart Tunnel, however without much delay-
ingtheconstructionprocess. At the2ndHeinenoord
Tunnel, construction work was delayed for several
weeks beforethecrewsucceededinrestoringfrontal
stability, fillingupthecrater intheriver bottomwith
clay and bringing in swelling clay particles in the
excavationroom.
From the evaluation of the monitored pressures
in the excavation room, it appeared that before the
developmentof theinstability, thefrontal pressurewas
raisedabovetheadvisedpressurefor frontal support;
i.e. at about 470kPa instead of about 310kPa. see
Fig.2(pressuregaugeP15isintheexcavationchamber
at tunnel axislevel).
In retrospect it was understood that during stand-
still,thepressureswereraisedtogetalargergradientin
thepipesinordertoimprovethetransportof excavated
Figure 3. Pore water pressure distribution in front of
theTBM.
material;i.e.Kedichemclaythatwasfoundinthelower
partof theexcavationfrontandappearedtobedifficult
topumpthroughthehydraulicmucktransportsystem.
The measurements indicate that excavation had
startedwithoutreleasingpressuretothestandardsup-
portlevel duringexcavation. Inthatconditioninstabil-
ity developedwithin15seconds after that thewheel
startedcutting. Atstandstill, whensufficienttimehas
passedfor aproper vertical cakesealingof bentonite
to build up at the front, a high support pressure is
not muchof aproblem, asthepressuresusedareway
belowthosethatmightoverridethepassiveresistance
at the front. However, as the pressure itself is fluid
pressure, whenthecake-sealingistakenaway during
excavation, andwater canpenetratethefront, accord-
ingto Pascals lawfor afluidwithout shear stresses,
thepressurealso works inthevertical direction, and
if this pressureexceeds thevertical soil pressurethis
will triggeranupliftandpossiblyabreakingoutof soil
layers, andapparentlythat iswhat hashappenedhere.
In their paper on face support J ancsecz and Steiner
(1994), for thisreasongaveawarningaboutthelimits
to thefacesupport pressure, for situations withlittle
overburden.
Researchlearnsthat for thefinesandthat wehave
in the Pleistocene sands layers in the Netherlands,
penetration of bentonite in the pores is negligible.
Excavationthereforemeansremoval of thecakeseal-
ing; Research by Bezuijen and Brassinga (2001),
indicatesthatitnormallytakesabout4to5minutesto
buildupanewcakesealingafter theexcavationwheel
has removedthesealingduringexcavation. Thetime
betweenpassingsof chisels, intheorderof 20seconds
is too short for that. It must beemphasized that this
effect isnot onlyimportant for theupper limit toface
support pressures, but may also give a limitation to
thelower limit of thesupport pressure. A methodto
discount for this effect was given by Broere(2001),
seealsoFig. 4.
245
Figure 4. The effect of removal of the cake sealing dur-
ingexcavationonpore-pressuresinthefront. Theinfluence
zonefor excess pore-pressures may belarger that thezone
normallyconsideredinstabilityanalysis.
Thesituationof alowsoil coverunderneaththeriver
bottomisnottheonlysituationthatmightbecritical to
theaboveeffect, alsoif thesoil coveritself isrelatively
light, suchas inthecaseof thethicker layers of peat
overlayingthesandwheretheGreenHartTunnel was
excavated, thismightleadtoacritical situation.Alocal
failuremight betriggeredwherethegeneratedexcess
porepressureinfrontof thetunnel facecanliftthesoft
soil layers.
Theknowledgegainedwiththemonitoringof the
2ndHeinenoordtunnel wasappliedfortheGreenHart
tunnel, and may have prevented instabilities at the
borefront at larger scales; seeBezuijenet al. 2001&
Autuori & Minec(2005).
3.2 Tail void grouting and grouting pressures
Tomeasurethesoil pressuresonatunnel liningisdiffi-
cult. Inthestart-upphaseforthemonitoringof the2nd
HeinenoordTunnel, anumber of international experts
on tunnel engineering advised not to put too much
effort onthistopic, astheresultswouldprobablybe
disappointing. Duetothehardeningof thegrout, the
periodfor meaningful pressuremeasurements would
beshort andtoprevent bridgingeffectsthesizeof the
pressure cells would have to be large and therefore
costly.
Still, against this advice, the measurement of
grouting pressures was undertaken, and repeated for
a number of tunnel projects. It appeared that the
interpretationwas difficult whenthegrout has hard-
ened,butforthefreshgroutuntil 17hourafterinjection
Figure5. Under circumstancestheGroutmaterial fromthe
tail voidmight flowintothegapbehindthetail of theTBM,
givingcausetoincreasedloads.
itwaspossibletogiveanacceptedinterpretationof the
measurement results (Bezuijen & Brassinga, 2004),
and a lot of experience has been gained that has
contributedto abetter understandingof thegrouting
processandthepressuresactingonthetunnel lining.
Withtheseresults it was possibleto predict grouting
pressuresandtunnel loading, seeTalmon& Bezuijen
(2005).
Basedonvariousevaluationsof theforcedistribu-
tioninthetunnel lining, seeamongst others, Bakker
(2000), it came forward that the initial in-situ soil
stresses around the tunnel do not have a dominant
influence on the compressive loading of the tunnel.
Duetothetaperingof theTBM andinspiteof thetail
voidgroutingthereisasignificantreleaseof theradial
stressesaroundthetunnel, seeFig. 5.
Thefinal loadingontheliningrelates moretothe
effectivenessof thegroutingprocess, thedistribution
of thegroutopeningsandtheconsolidationof thegrout
than to the initial in-site soil stresses, see Bezuijen
et al. (2004). Whether this reduction of the in-situ
radial stresses is alasting effect that will remain for
thefull lifespanof thetunnel maydependonthecreep
sensitivityof thesoil, seealsoBrinkgreveandBakker
(2001), andHashimotoet al. (2008).
3.3 Surface settlements
Hoefsloot et al. (2005), haveshownthat theapplica-
tion of a stress boundary condition between tunnel
and soil in 3D tunnel analysis has a better corrobo-
ration between measurement and calculation of soil
deformations around thetunnel and subsequently of
theloadingonthetunnel, thantheapplicationof the
socalledcontractionmethod.
Althoughthiseffectwasknownintheliterature, see
for exampleMair andTaylor (1997), for theresearch
teamthat worked at the 2nd Heinenoord tunnel the
observationthat thenumerical predictions of surface
settlementslackedaccuracywasdisappointing.Atthe
start theexpectationsonnumerical analysishadbeen
quite high. Shortly after the first observations were
evaluatedit wasrealizedwithintheteam, that it were
only the empirical predictions by Peck (1969) for a
246
Figure 6. Surface settlements; measured and
back-calculatedwithdifferent material models.
volumeloss of about 1%that gaveareasonablecor-
roborationwiththemeasurements. Thefiniteelement
calculations, at that time mainly based on an appli-
cationof theelastic-plastic Mohr-Coulombmodel in
combinationwithacontractiontypeof modellingfor
thetail voidloss, predictedatoowideandtooshallow
surfacesettlement.
Thisdisappointingresult createdaproblemfor the
intentionstoapply3Dnumerical analysisindeforma-
tionpredictionsfortunnel projectsinurbanareas, such
asfor theAmsterdamNorth-Southlinemetroworks.
Since then, however, a lot of effort has been put
in the improvement of the numerical prediction of
soil deformations. To begin with it was the project
teamfor theAmsterdamMetroworks, seeVanDijk&
Kaalberg (1998), that gave a first indication for an
improvement, withtheproposal tomodel thestresses
atthetunnel soil interfaceinsteadof thedeformations.
Withtheintroductionof thisgroutpressuremodel the
resultsimproved.Lateron,whenthephysicsinthepro-
cessbecamebetter understood, i.e. theimportanceto
account for thehighstiffnessof thesoil inunloading,
double hardening was introduced with the develop-
mentof HardeningSoil, asamaterial model; withthis
development, thecalculationresultslargelyimproved
comparedtothemeasurements, seeFig. 6. Thelatest
development is theintroduction of small strain stiff-
ness in theHardening Soil Model, seeBenz (2006),
which up to now gives the best results, see Mller
(2005).
Theoretically theresult might further beimproved
introducing anisotropy in the model; such models
are being developed nowadays, e.g. in the frame-
work of European Research; AMGISS, e.g. see
www.ce.strath.ac.uk/amgiss.
4 EVALUATIONOF THE LEARNINGISSUES
Nowadays its not the soil deformation during nor-
mal excavation process that makes us worry about
surfacesettlements. Withanaveragetail voidloss of
about 0.5%of thediameter or less, thedeformation
might only be a problem for situations of under-
excavationof buildingsor if thestructuresarelocated
veryclosetotheexcavationtrack. Fortunnelsinurban
area, thereismoreconcernwithrespect tobore-front
stability; especiallywhentheupper stratumof thesoil
above the Pleistocene layers, where the tunnels are
usually positioned in, consists of soil with arelative
lowdensity, asintheNetherlands. For thesesituations
withrelativelylight upper layersof peat or clayswith
organic parts, onehas to bevery careful controlling
thesupportpressuresduringexcavation, asontheone
hand there is a lower bound value of the support to
prevent cavein, but ontheother hand, theupper limit
triggeredbyanupliftof lightupper layersmayalsobe
notfar.Thiswill limitthepressurewindowtoworkin.
Front instabilityhasoccurredat varioustunnelling
projects in the Netherlands. If the tunnel is outside
anyurbanareathismightnotgivetoomuchproblems;
however if the tunnel is underneath a city road sys-
tem, or closetopilefoundationsthismaycausesevere
problems, asinstability might causeasinkholeinthe
pavement andfoundationsettlements.
With respect to the accuracy in the prediction of
soil deformations:Apartfromthewell knownempiric
model of Peck (1969) that predicts the shape of the
settlementtroughbutnotthevolumeloss, thenumeri-
cal models have become quite reliable in predicting
surface and subsurface deformations, both vertical
and horizontal. The improvement, mainly achieved
in 2D analysis has opened up the possibility for a
reliable deformation analysis in 3D of tunnelling in
urban areas. For an adequateprediction of deforma-
tionsitisimportanttomodel thegroutingpressuresas
aboundaryconditiontotheexcavation,incombination
withtheapplicationof ahigher order material model,
that takes into account the small strain deformation
behaviour of sand, seeBenz (2006).
Furtheritisrecommended, andplannedfor, tointe-
grate the Delft Cluster Grout pressure model in the
Plaxis3DTunnel software.Thelatterwouldcontribute
totheapplicabilityof thenumerical modelsasamore
general tool forundergroundconstruction.Thiswould
enableabetter analysis for theloadingonthetail of
theTBM andof thetunnel lining.
Within certain limits some cost saving structural
improvements areexpected to bepossibleand, even
moreimportant, insight isobtainedinthemechanism
involved.
5 CONCLUDINGREMARKS
TenYears have passed since the first large diameter
boredtunnellingprojectintheNetherlandsinSoftsoil
was undertaken. Sincethensomeworldrecords with
respect totunnellinghavebeenbrokenintheNether-
lands; i.e. the largest diameter (for the Green Hart
247
Tunnel), thehighest outsidepressureonasegmental
tunnel (fortheWesternscheldtTunnel), theapplication
of an Earth Pressure Balance shield in coarse sand,
andthelargest lengthof constructedtubeinoneday,
(PannerdenschCanal Tunnel).
Before the underground construction works were
started,andthetunnellingprojectswereinapre-design
stage, the softness of the Netherlands underground
attracted a large part of the attention, see Bakker
(1997). In retrospect theinfluenceof alowstiffness
asasourceof riskandinfluenceonundergroundcon-
structionwasconfirmed, but sometimesinadifferent
perspective, orrelatedtootherphysical processesthan
foreseen.
Withrespect thesenewinsightsthefollowingcon-
clusionsweredrawn:
1 The low stiffness of the soil may also lead to
increasedflexibility of thetunnel tube. Thedefor-
mationof thetubeduringhardeningof thegrout,
andtheadditional Eigenstressesthatthismaycause
isstill aresearchtopic.
2 For aproper predictionof surfacesettlementsand
soil deformations, it is important to model the
groutingpressuresattheinterfacebetweensoil and
tunnel (or grouting zone). Further to improvethe
prediction of the width of the settlement trough,
theuseof small strainanalysisisadvised.
3 Duringexcavationinfinesand, suchas thePleis-
tocenesandlayersintheNetherlands, thesupport-
ing cake fluid will be removed by the chisels on
theexcavationwheel.Therefore, incasesof limited
overburdentheupperboundtothesupportpressure
must becarefullydeterminedtoprevent instability
of theoverlayingsoil.
4 Inaddition; for thedeterminationof thelower limit
tothesupportpressure,theincreasedporepressures
inthefront alsoneedstobetakenintoaccount.
With acknowledgement to the Netherlands Cen-
tre for Underground Construction for their consent
topublishabout theresearchthey commissionedand
coordinated.
REFERENCES
Autuori,P.&Minec,S.2005.Largediametertunnellingunder
polders, Proceedings 5th Int. Symposiumon Underground
Construction in soft Ground, IS-Amsterdam2005.
Bakker, K.J., v.d. Berg, P. & Rots, J. 1997. Monitoringsoft
soil tunnellingintheNetherlands; aninventoryof design
aspects, Proc. ISSMFE, Hamburg.
Bakker, K.J. 2000. Soil Retainingstructures, developmentof
modelsforstructural analysis,Balkema,2000,Rotterdam.
Benz, T. 2006. Small strain stiffness of soils and its conse-
quences, DoctorThesis, IGSUniversitt Stuttgart.
Bezuijen,A.&Brassinga,H.E.2001.Blow-outpressuremea-
sured in a centrifuge model and in the field. Proc. Int.
Symp. on Modern Tunnelling Science and Techn. Kyoto.
Fieldmeasurementsof grout pressuresduringtunnelling
of theSophiaRail Tunnel. Soils and Foundations vol, 44,
No1, 4150, Feb.
Bezuijen, A. & Talmon, A.M. 2005. Grout properties and
their influenceonback fill grouting. Proceeding 5th Int.
Symposiumon Underground Construction in soft Ground,
IS-Amsterdam2005.
Bezuijen, A., Pruiksma, J.P. &vanMeerten, H.H. 2001. Pore
pressures in front of tunnel, measurements, calculations
and consequences for stability of tunnel face. Proc. Int.
Symp. on Modern Tunnelling Science and Techn. Kyoto.
Decade of Progress. GeoDelft 19952005.
Bezuijen, A. & Bakker, K.J. 2008. The influence of flow
aroundaTBM machine. Proceeding 6th Int. Symposium
Bezuijen, A. & Talmon, A.M. 2008. Processes around a
TBM. Proceeding 6th Int. Symposium on Underground
Brinkgreve, R.B.J. & Bakker, K.J. 2001. Time-dependent
behaviour of bore tunnels in soft soil conditions; a
numerical study, Proceedings, ICSMGE Istanbul, Turkey.
Broere, W. 2001. Tunnel facestability & newCPT applica-
tions, Delft UniversityPress.
Hashimoto, T., Ye, G.L., Nagaya, J., Konda, T. & Ma, X.F.
2008. Study onearthpressureactinguponshieldtunnel
lininginclayey andsandy grounds basedonfieldmon-
itoring. Proceeding 6th Int. Symposium on Underground
Hoefsloot, F.J.M. & Verweij, A. 2005. 4D grouting pres-
suremodel PLAXIS. Proceeding5thInt. Symposiumon
UndergroundConstructioninsoftGround, IS-Amsterdam
2005.
J ancsecz,S.&Steiner,W.1994.FaceSupportforalargeMix-
Shieldinheterogeneousgroundconditions.Tunnelling 94,
BritishTunnellingAssociation, 57J uly1994.
Mair, R.J. & Taylor, R.N. 1997. ThemeLecture: Boredtun-
nelling in the urban environment. Proc. 14th ICSMFE,
Balkema, Rotterdam.
Mller, S.C. &Vermeer, P.A. 2005. Predictionof settlements
andstructural forcesinliningsduetotunnelling. Proceed-
ing 5th Int. Symposium on Underground Construction in
soft Ground, IS-Amsterdam2005.
Pachen, H.M.A., Brassinga, H. & Bezuijen, A. 2005.
Geotechnical centrifuge tests to verify the long-term
behaviour of aboredtunnel, Proc. 5th Int. Symposium on
UndergroundConstructioninsoft Ground,IS-Amsterdam
2005.
Peck, R.B. 1969. Deep excavations and Tunnelling in soft
Ground. Proceedings 7th ICSMFE Mexico.
Talmon, A.M. &Bezuijen, A. 2005. Groutingthetail voidof
boredtunnels: theroleof hardeningandconsolidationof
grouts. Proceeding 5th Int. Symposium on Underground
Construction in soft Ground, IS-Amsterdam.
vanDijk, B. & Kaalberg, F.J. 1998. 3-Dgeotechnical model
for the North/Southline in Amsterdam. In A. Cividini
(Ed.), Applicationof numerical methods togeotechnical
problems, Wien, 739{750. Springer-Verlag.
248
Tenyearsof boredtunnelsinTheNetherlands: Part II structural issues
K.J. Bakker
COB, Delft University of Technology, Delft, The Netherlands
A. Bezuijen
Deltares, Delft University of Technology, Delft, The Netherlands
ABSTRACT: In1997for thefirst timeconstructionof boredtunnelsintheNetherlandssoft soil wasunder-
taken. Beforethat dateessentially only immersed tunnels and cut-and-cover tunnels wereconstructed in the
Netherlands. Thefirst two boredtunnels werePilot Projects, the2ndHeinenoordtunnel andtheBotlek Rail
tunnel. Sincethenaseries of other boredtunnels has beenconstructedandsomearestill under construction
today. At thebeginningof this period, amongst others Bakker (1997), gaveanoverviewof therisks relatedto
boredtunnelsinsoft groundandexplainedabout aplanfor researchrelatedtothePilot projects. Tenyearshave
passed, alot of monitoring and research has been done. In this paper that is split in two parts asummary is
givenof someof themost characteristicobservationsof thesepast 10yearsof undergroundconstructioninthe
Netherlands. Inthissecondpart, theemphasiswill beonstructural relatedissuesdiscussedwhereasinpartone,
frontal stability, groutingandsoil deformationsarediscussed.
1 INTRODUCTION
In1992theDutchgovernmentsentafact-findingmis-
siontoJ apan, toreport onthepossibilitytoconstruct
boredtunnelsintheDutchsoft soil conditions. Upto
thattimeessentiallyonlyimmersedandcut-and-cover
tunnels wereconstructed in theNetherlands, as bor-
ingof tunnelsinsoft soil conditions, at that time, was
consideredtobetooriskfull.
After thereport, that advisedpositive, thingswent
quite fast; in 1993 the Dutch minister of Transport
and Public works ordered the undertaking of two
pilot projects, the 2nd Heinenoord Tunnel and the
BotlekRail Tunnel.Theprojectswereprimarilyaimed
at constructing new infrastructure and besides that
for monitoring and research in order to advancethe
development of thisnewconstructionmethodfor the
Netherlands.Theprojectsstartedin1997and10years
havepassedsincethen.
Atthestartof thepilotprojects, thedifficultieswith
respecttotheconstructionof boredtunnelsinsoftsoil
conditions wereevaluatedandaplanfor monitoring
andresearchwasputforward,seeBakker(1997).Since
then, the2ndHeinenoordtunnel, andaseriesof other
boredtunnelswereconstructed. Unquestionablyalot
hasbeenlearnedfromall themonitoringandresearch
that wasperformed.
Figure1. Trumpet effect intunnel ringconstruction.
Theresultsof thisprocesshavebeennoticedabroad.
In2005theNetherlandshostedthefifthInternational
symposiumofTC28onUndergroundConstructionin
Soft Ground. Theaboveevent wasalsotheoccasion
for thepresentationof abook; A decadeof progress
intunnellingintheNetherlands byBezuijenandvan
Lottum(2006), where this research is described in
moredetail.
In the present paper some highlights of the main
researchresult of thepast decadewill begiven. The
paper is split in two parts, where part one includes
somegeneral observationsanddiscussesfacesupport,
249
groutingandsurfacesettlements, whereaspart twois
moreabout structural issues.
2 REVIEWOF THE 1997SITUATIONAND
WHAT CAMEAFTER
A mainconcernwithrespect toboringtunnels inthe
Netherlandswerethesoftsoil conditions; thelowstiff-
nessof theHoloceneclayandpeatlayersandthehigh
groundwater table; nearly upto thesoil surfacewere
consideredapotential hazardandachallengeforbored
tunnels.
Furthermore the 8.3moutward diameter for the
first largediameter tunnel was amajor stepforward,
comparedtopastexperience; experiencethatuptothat
timewasmainlybasedonconstructingboredtunnels,
pipesor conduitsuptoabout 4.0mdiameter.
Inadditiontothat, ingeneral deformations dueto
tunnellingmightinfluencethebearingcapacityof any
existingpiledfoundations inthevicinity. Andas the
commonsayingisthattheAmsterdamForestisunder-
ground, onemight realizethepotential risksinvolved
for theNorth/SouthMetroworksinAmsterdam.
Characteristic for ahighwater tablearebuoyancy
effects.Besidestheriskof breakingupof thesoftupper
soil layers, theratherflexiblebeddingof thetunnel and
thedeformationsthatthismaycauseneedtobeanaly-
sed. Therefore research was aimed at clarifying the
effects of thesoft underground, groundwater effects,
andtheeffect of tunnellingonpiledfoundations.
Ten years later, the question arises whether the
observationshaveconfirmedtheaboveissuestobethe
critical ones. Inthis paper someof thecharacteristic
eventsandresultsof thispastdecadewill bedescribed.
Thechoiceforthetopicsbeingdiscussedisinfluenced
by theprojects that bothauthors wereinvolvedwith,
without intent to minimize the importance of other
researchthat isnot discussedinthispaper.
3 EXPERIENCESWITHBOREDTUNNELS
INTHE NETHERLANDSINTHE PAST
DECADE
3.1 Structural damage
Anearlyexperiencewiththedifficultiesforboredtun-
nels insoft groundwas thedamagetotheliningthat
occurred during thefirst 150 metres of construction
of the2ndHeinenoordTunnel. Onaveragethedam-
agewastoohighcomparedtoexperiencesfromabroad
andwasconsideredtobeunacceptable. Although, the
integrityof thetunnel wasnotatstake, therewasworry
aboutthedurabilityof thetunnel andthelevel of future
maintenance.
Characteristic to the damage was cracking and
spalling of concrete near the dowel and notches see
Fig. 2. Quite often the damage was combined with
Figure2. DamagetotheDowel andnotchsockets.
Figure3. Large-scaletunnel ringtestingintheStevinLab-
oratoriesatDelftUniversity(thediameter of the(gray) inner
ringis8.3m).
differential displacements between subsequent rings
and with leakage. Theevaluation report, seeBakker
(2000), attributed thedamageto irregularities in the
constructionof theliningat therear of theTBM and
subsequent loading duringTBM progress. Further a
correlation of thedamagewith high jack forces was
observed; theseappearedtobenecessarytoovercome
thefrictioninthis part of thetrack, whichprevented
smoothprogress.
With respect to the tunnel ring construction, it is
difficulttoerectastressfreeperfectcircular ring. The
250
Figure4. Test sitefor thePile-tunnel interactiontest.
ring needs to bebuilt onto theend of aformer ring
that already has undergonesomeloading and defor-
mation fromthe tail void grouting while it partially
has left the tail of theTBM, see Fig. 3. The further
deformationis characterisedby thetrumpet shapeof
thetubingthatdevelops, seeFig. 1, withtheinevitable
relatedstress development inthelining. Thetrumpet
shapeandthehighjackingforces leadtolocal stress
concentrations andirregular deformations inthelin-
ing and occasional to slipping between thedifferent
tunnel elements.Theslippingof elementswasblamed
to theuseof abituminous material called Kaubit in
theringjoint.
Originally Kaubit strips hadbeenusedinthering
joint. TheseKaubit strips, of flexiblebituminouslike
material, wereusedtopreventtheoccurrenceof stress
concentrations; sosomeslippingwasmeant tooccur,
but thedynamic character of theslippingthat actu-
ally occurred that influenced the final geometry of
theliningandhadtriggeredcrackingwasunexpected.
Especially thecrackingandoverloadingof thedowel
andnotchsystemwasunforeseen.
Failureof thedowel andnotchsystem, seeFig. 2.
led to spalling and in some cases to leakage. In the
cases that leakagewas observedthis must havebeen
correlated to damage to the notch at the outer side
of thelining, creatingashortcut towater penetrating
behindtherubber sealingthere.
After the main conclusions were drawn, it was
decidedtoexchangetheKaubitstripsforthinplywood
plates. Dueto thelarger stiffness andshearingresis-
tance, shearingof theconcreteelements at largewas
further preventedandthedamagelimited.
Besidesthistechnical measure, theevaluationwas
thetriggerfortheundertakingof fundamental research
into lining design that included large scale physical
testingof tunnel tubingat Delft UniversityseeFig. 4.
Inthisproject that wasacombinedeffort of physical
andnumerical testing, thedetails of assemblingtun-
nel segments into subsequent tunnel rings and these
intoatubewereinvestigated.Amongstothersthemain
results of theproject werereportedby Blom(2002),
30
45
zone A zone B zone C
Figure 5. Zones that indicate different effects on piles
foundations.
and Uijl et al (2003). Based on this research it was
decidedto omit thedowel andnotches for theGreen
Hart tunnel; whichledtoanearlydamagefreetunnel
lining.
A different issue, not settled yet, is the durabil-
ity of plywoodandtheconsequences of woodrot on
thelong-termtunnel behavior.Anunwantedlossof the
longitudinal pre-stressof atunnel mightinfluencethe
tunnel flexibility and deformations, possibly leading
toleakages. Ontheother hand, experiencelearnsthat
compressionlargelyincreasesthedurabilityof wood.
Theplywoodmaterial iscompressedtoastrainof more
than50%duringtunnel construction. At suchahigh
level of strainingthewoodcellsmighthavecollapsed.
3.2 Deformations of the TBM machine during
construction of the Westernscheldt tunnel
Duringconstructionof thefirst tubefor theWestern-
scheldt tunnel, unexpecteddeformationsof thetail of
theTBM were observed; i.e. the air space between
tubing and tail of the TBM narrowed at a certain
stageinanunexpectedway.Theshapeof theobserved
deformations didnot coincidewiththeassumedsoil
loading and gave the impression that it was a large
deformationseffect; i.e. buckling.
At first buckling was not accepted as a cause
because the tapering of the TBM was assumed to
givesufficient stressreleasetoguaranteeasufficient
decreaseinisotropicstress. Further acertainbedding
effectwasassumedtobealwayspresentandthecombi-
nationwouldmakebucklingunlikely. Bucklingwould
onlybeplausiblefor amuchhigher loadingof thetail
of theTBM in combination with theabsenceof any
beddingreaction.
However, theinsights havechangedsincethen. In
general theremay beno overall contact between the
soil andthetail of theTBM; whengrout isinjectedin
thetail void, theincreasedpressureonthesoil, com-
paredtotheoriginal stresswill pushthesoil fromthe
TBM and grout will flowbetween theTBM tail and
thesoil, seeFig. 5inpart I of thispaper. Thismeans
that the pressures on the TBM tail are higher than
anticipatedinthepast andtheremight benobedding
251
reaction. This could well explain the occurrence of
bucklingandthedeformationsof theTBM tail.
A 1-Dcalculationmodel hasbeendevelopedandis
verified with FEM simulations (Bezuijen & Bakker,
2008). This model shows that also thehigh stiffness
of soil duringunloading, whichledtotheHSandthe
HS
small
material models, madeit likely that thecom-
mon tapering, approximately equal to an equivalent
volumelossof 0.4%, issufficienttolosethelargerpart
of theeffectiveradial stresses, whichhelpstodevelop
agapbetweenthetail of TBM andthesoil.
The grout pressures exerted on the tail of TBM
might be much higher than the soil stresses, and in
absenceof bedding, buckling could well explain for
thedeformations.
3.3 The influence of tunnel boring to piled
foundations
Largescaletestingof pilefoundationswasperformed
during construction of the 2nd Heinenoord Tunnel.
This was donein order of theProject Bureau of the
AmsterdamNorth/South metro works to get abetter
understandingof theprocesses.
A trial fieldwithloadedpiles andpileconfigura-
tionswasinstalledintheareanear andabovethetrack
of theTBM, seeFig. 4. Oneof themainconcernswas
that dueto anincreaseinporepressuretheeffective
stressesaroundthepiletipmight beaffectedandthat
areleaseinisotropic stresses might trigger adropin
pilebearingcapacity.
However, againstthisreasoningthereisalsonumer-
ical and analytic evidence, (assuming cylinder sym-
metric analysis), that indicates that the release in
stresses dueto tunnellingis limitedto arather small
plastic zoneintheclosevicinity of thetunnel lining,
seealsoVerruijt (1993). Theanalytical model reveals
thatstrainasafunctionof thedistancedropsasafunc-
tionof 1,r
2
, whichwouldindicatethat theinfluence
zonewouldbelimitedinsize.Thisreasoningincombi-
nationwiththefactthatthestrainsduetotunnellingin
general arequitesmall; thelargest strainsoftenbeing
lessthan0.5or1.0%, makesplasticzonesfurtheraway
than D/2, measured fromthe tubing, unlikely. Only
abovethetunnel thiszonecanbelarger.
However, reasoningandanalysisisonething; mea-
suring and validation is another; based on the field
measurements and physical model research in Delft
andCambridgeKaalbergetal. (2005),proposedazon-
ingasshowninFig. 5, withthefollowingindicators;
azoneA abovethetunnel wherethesettlement of a
pileisexpectedtobelargerthanthesoil displacements.
A zone B adjacent to the tunnel, with an inclined
influenceline, wherethepilewill followthesoil defor-
mationat thetipof thepile, andfurther azoneC,
outsideZoneB, whereat soil surfacelevel thesettle-
mentof thepilewill belessthanthatof thesoil surface.
Thiszoningproposal moreor lesscoincideswiththe
main results as published by Selemetas (2005) that
weremainlybasedof physical testinginageotechnical
centrifuge.
The results published by Kaalberg et al. and oth-
ers are valid for the average volume loss that can
be expected during tunnelling (0.5 to 1%) Earlier
centrifuge testing by GeoDelft indicated that larger
deformation effects are possible for higher volume
losses(upto7%wastested). Suchvolumelossesare
well abovenowadayspractice, butitmeansthatduring
acalamity, pilesover alarger areamaybeaffected.
3.4 Longitudinal deformations of the tunnel tube
In the paper by Bakker et al. (1997), the develop-
ment of longitudinal stresses in a tunnel lining due
to irregular bedding in soft soil was mentioned as
anitemfor research. Irregular beddingthat couldbe
the result of zones with different elasticity or else
due to the stiff foundation of a shaft or bedding
in the deeper Pleistocene layers; especially near the
transition between Holocene and Pleistocene layers.
Themeasurement of longitudinal stressesinitself has
turnedout tobecumbersome. Withinthemonitoring
schemefor the2nd Heinenoord atrial measurement
was undertaken. In addition to that measurements
fromtheSophiaRail Tunnel wereback-analysedwith
4D finite element analysis, i.e. (time dependant 3D
analysis), andafter that thelongitudinal stresseswere
also measured during the construction of the Green
HartTunnel.
To begin with the latter situation: measurements
were taken with a tubular liquid level devise of the
longitudinal deformations of the tunnel during the
groutingprocess. Fromthesemeasurementstheobser-
vation came forward that the tubing exhibited large
vertical movements, up and down, between 20 to
30mmduring excavation and tail void grouting was
measured, andatotal vertical shift of thetubingverti-
cal of about60mmatonelocation(SeealsoTalmon&
Bezuijen, 2008).
This amplitude was surely unexpected and is not
fullyacceptedyet. Neverthelessitisclear thatvertical
deformations do occur in the zone where the grout
material is still fluid, andduringexcavationandmay
leadtoanalternatingdeformation; upwardswhenthe
TBMisexcavatingandgroutinganddownwardsif the
TBM isat standstill.
Withrespect to the3D stagedconstructionanaly-
sisof tunnel constructionfor theSophiaRail Tunnel,
that was undertakenfor theCOB F220committee, a
combinedDIANA andPLAXIS3Danalysiswasper-
formed, seeHoefsloot et al. (2005). Theoutcomeof
thesevarious analyses moreor less coincided; which
might have been expected as the mathematical base
of bothmodelsisquitesimilar, andingeneral defor-
mations remainsmall, sothesoil reactions will most
probablymainlyhavebeenelastic.
252
Figure6. Conceptual model fortheanalysisof beameffects
inthetubeof aboredtunnel byBoogaards(1999).
Themainconclusionwithrespecttothiseffectwas
that this issuecan beproperly analysed with arela-
tively simplemodel basedontheconcept of abeam
withanelasticbeddingandaseriessummation, such
as developed by Boogaards & Bakker (1999), see
Fig. 6andlater onappliedby Hoefsloot (2002). See
Fig. 7for acomparisonbetweenmodel outcomeand
measurementsfromthe2ndHeinenoordtunnel.
However, usinggenerallyacceptedparameters, the
measureddeformationsaremuchhigher thanaccord-
ing to these models. Recently, Talmon et al. (2008)
havepresentedresultsthatmayexplainthelowerstiff-
ness that are found in the measurements (the lining
stiffnesscanbelowerduetoonlylocal contactbetween
the elements and the soil stiffness reduces due to
unloadingof thesoil aroundthetunnel), but theseare
not yet generallyaccepted.
4 CROSSPASSAGES
The design for the Westernscheldt tunnel in the
Netherlandsdidtriggeradebateontunnel safety.Some
major accidentswithtunnel fires, suchasoccurredin
theChannel tunnel andattheMontBlanctunnel inthe
Alpsdidreveal thevulnerabilityandrelativeunsafesit-
uationsintunnelswithoncomingtrafficor inasingle
tubeingeneral.
For theWesternscheldt tunnel, atwin tunnel with
oneway traffic per tube, thediscussion focussed on
whatdistancebetweencrosspassageswouldbeaccept-
abletoguaranteethat escapingpeoplewouldbeable
tofindasafehavenbyenteringtheother tube; assum-
ingthat thetrafficisstopped, byanautomaticcontrol
system.Theoutcomeof thesesafetystudieswasacross
Figure 7. Longitudinal bending moment due to the con-
struction forces measured and calculated acc. to Bogaards
model.
connection at least every 250m, which is nowadays
moreorlessthereferencesituationintheNetherlands.
Thetask toconstruct thesecrosspassagesisafur-
ther technical effort. During the construction of the
Botlek Rail Tunnel avertical shaft andfreezingwere
themainconstructiontechniquesasthecrosspassages
couldbepositionedoutsidetheareaunder theOude
MaasRiver. Thepositiveexperiencewithfreezingfor
the Botlek Rail Tunnel was helpful in the decision
making for theWesternscheldt Tunnel, but therethe
freezing was donefromthetunnel tubeas thetrack
underneaththeestuary is toolongandtoodeepwith
respect of the water table to enable the shaft type
method.
Althoughthemethodinitself iscostly, itsreliability
isanimportantadvantageandthereforeitisalsoused
for thecross passages of theHubertusTunnel andis
expectedto beusedinfutureprojects. For thesingle
tubeGreen Hart Tunnel tunnel safety is achieved by
constructionof aseparationwall withdoors.
5 EVALUATIONOF THE LEARNING
ISSUES
Theresearchongrout pressures, incombinationwith
the structural research on lining design has gained
us the insight that the lining thickness and the nec-
essary reinforcement are mainly determined by the
loading in the construction phase and to a lesser
degree to the soil pressures. In engineering prac-
ticethethickness and reinforcement of thetubing is
mainly determined by the most unfavourable jack-
forces during TBM excavation in combination with
anunfavourabletail voidgroutingscenario. Difficulty
with these is that its the contractors prerogative to
decide on the necessary jack-forces that will enable
himtoconstruct thetunnel andalsowhat scenariohe
will usefor thetail voidgrouting. Thismight leadto
253
conservativeassumptionsinthedesignofficeinorder
to avoid liabilities if a problemwould occur during
construction.
Withrespect tothegenerality of thisconclusionit
has to beconsidered that themain observations that
werediscussedrelatetotunnelsthataresafelylocated
instiff Pleistocenesandlayers. Wemusthowever con-
sider the possibility of tunnels in softer soil layers
that aremoresusceptibleto consolidationandcreep.
Consolidationandcreepmight counteract thegeneral
tendency of stressreleaseandarchinginthesoil and
leadtoamuchhigher radial loading. Onemight think
of a soil pressure on the lining that may be on the
level of theinitial soil stressesbeforetunnel construc-
tion; theK
o
stresssituationor evenhigher thanthese
initial stresses. Suchasituationwasaccountedfor in
thedesignforRandstadRail inRotterdam, whereafull
steel liningwaschosenforapartof thetrackwherethe
tubingmainlyrestsintheuppermuchsofterHolocene
clay and peat layers, that foreseeablewould havean
extraloading on thelining dueto consolidation and
creep(Pachenet al. 2005).
However, withrespect toliningdesign, withincer-
tainlimits somecost savingstructural improvements
areexpectedtobepossibleand, evenmoreimportant,
insight isobtainedinthemechanismsinvolved.
6 CONCLUDINGREMARKS
TenYears have passed since the first large diameter
boredtunnellingprojectintheNetherlandsinSoftsoil
wasundertaken. Beforetheundergroundconstruction
works werestarted, and thetunnelling projects were
inapre-designstage, thesoftnessof theNetherlands
undergroundattractedalargepartof theattention, see
Bakker (1997). In retrospect the influence of a low
stiffness as asourceof risk and influenceon under-
ground construction was confirmed, but sometimes
inadifferent perspective, or relatedtoother physical
processesthanforeseen.
Withrespect tothenewinsightsgainedthefollow-
ingconclusionsweredrawn:
1 Thelowstiffness of thegroundsupport may give
risetoincreasedvulnerabilityof theliningforjack-
ingforcesbytheTBMduringexcavation.Caremust
betakentopreciseshapeof theelementsandjoints
toprevent toohighstressesduringassembly.
2 The low stiffness of the soil may also lead to
increasedflexibility of thetunnel tube. Thedefor-
mationof thetubeduringhardeningof thegrout,
andtheadditional Eigenstressesthatthismaycause
isstill aresearchtopic.
3 ThestifferPleistocenesandlayersmightnotalways
be able to follow the tapering of theTBM. It is
expectedthat thismaygiverisetogappingbehind
the tail of theTBM. If grout penetrates this gap,
this may cause higher loads on theTBM than is
normallyassumed.
4 No proof was found that tunnel driving, in nor-
mal operation, might givecausetolossof bearing
capacityof piles. Settlementsingeneral arerelated
tothesettlement of thegroundandthepositionof
thepiletoewithrespect to thezones indicatedin
fig. 5.
With acknowledgement to the Netherlands Cen-
tre for Underground Construction for their consent
topublishabout theresearchthey commissionedand
coordinated, andwiththankstoCeesBlomfor theuse
of someof thefigures.
REFERENCES
Bakker, K.J., v.d. Berg, P. & Rots, J. 1997. Monitoringsoft
soil tunnellingintheNetherlands; aninventoryof design
aspects, Proc. ISSMFE, Hamburg.
Bakker, K.J. 2000. Soil Retainingstructures, developmentof
modelsforstructural analysis, Balkema, 2000, Rotterdam
aroundaTBM machine. Proceeding 6th Int. Symposium
Decade of Progress. GeoDelft 19952005.
Blom, C.B.M. 2002. Designphilosophy of concretelinings
fortunnelsinsoftSoil,DelftUniv.Press,TheNetherlands.
Bogaards, P.J. & Bakker K.J. 1999. Longitudinal bending
inGeomechanicsProc. NUMOGVII: p. 317321.
Hoefsloot, F.J.M. & Bakker, K.J. 2002. Longitudinal effects
boredHubertustunnel inTheHague. Proceeding 4th Int.
Symposiumon Underground Construction in soft Ground,
IS-Toulouse.
Hoefsloot, F.J.M. & Verweij, A. 2005. 4D grouting pres-
suremodel PLAXIS. Proceeding5thInt. Symposiumon
UndergroundConstructioninsoftGround, IS-Amsterdam
2005.
Kaalberg, F.J.,Teunissen, E.A.H., vanTol,A.F. &Bosch, J.W.
2005. Dutchresearchontheimpactof shieldtunnellingon
pilefoundations. Proceedings of 16th ICSMGE, Osaka.
Pachen, H.M.A., Brassinga, H. & Bezuijen, A. 2005.
Geotechnical centrifuge testst to verify the long-term
behavior of aboredtunnel. Proc. 5th Int. Symposium on
UndergroundConstructioninsoft Ground, IS-Amsterdam
2005.
Selemetas, D., Standing, J.R. &Mair, R.J. 2005.Theresponse
of full-scalepilestotunnelling. Proceeding 5th Int. Sym-
posium on Underground Construction in soft Ground,
IS-Amsterdam2005.
Talmon, A.M. & Bezuijen, A. 2008. Backfill grouting
researchatGroeneHartTunnel. Proceeding 6th Int. Sym-
posium on Underground Construction in soft ground,
IS-Shanghai 2008.
Uijl, J.A., den, A.H., Vervuurt, J.M., Gijsbers, F.B.J. & van
der Veen, C. 2003. Full scaletests on asegmented tun-
nel lining. InProc. ITA World Tunnelling Congress 2003,
Amsterdam, TheNetherlands, 1217April 2003.
Verruijt, A. 1993. Soil Dynamics, Delft University of
Technology.
254
Theinfluenceof flowaroundaTBM machine
A. Bezuijen
GeoDelft, Delft University of Technology, Delft, The Netherlands
K.J. Bakker
Delft University of Technology, Delft, The Netherlands
ABSTRACT: Theflowof groutandbentonitearoundaslurryshieldTBMisinvestigated,usinga1-dimensional
calculationmodel.Theresultsshowthattheshieldof theTBMisonlypartlyincontactwiththesurroundingsoil
andthatfor alargeextentitisincontactwiththeliquidgroutand, dependingontheamountof over cutting, also
incontact withthebentonitethat is injectedat thetunnel face. Theseliquids alsodeterminetheforces onthe
TBMshield.Thecalculationmodel presentsthepressuredistributionontheshieldandshowstheinfluenceof the
soil properties, theovercuttingandthepropertiesof thegrout andthebentonite. Assumptionsinthemodel are
checkedwith2-Dfiniteelement calculations, planestrainandaxi-symmetric, that showqualitativeagreement,
andareasonablequantitativeagreement. Themodel isuseful toexplainthephenomenaandasafirstestimation
of theinfluenceof flowaroundtheTBM.
1 INTRODUCTION
Tunnelling through urban areas asks for a minimal
groundloss. ModernTBMscanoperateatanaverage
groundlossof 0.5%orless(Bowers&Mos, 2005).To
beabletocalculatethegroundlossthatcanbeexpected
inthisrangeit isnecessarytoimprovethecalculation
methods. Onedecadeagoitwassufficienttocalculate
whether thefront pressurewassufficient toprevent a
collapseof theborefront andnot highenoughtoget
ablow-out. Nowadays 4-D finiteelement modelling
isusedtoestimatethegroundlossandthesettlement
troughthat istheresult of it.
In most simulations on TBM tunnelling it is
assumedthat thenon-excavatedsoil aroundtheTBM
slides over theshieldskinof theTBM (Hoefsloot &
Verweij, 2005; Kasper & Meschke, 2006). As acon-
sequencethetaperingof theTBM resultsdirectlyinto
avolumelossandasettlement trough.
However, inrealitythecalculatedsettlementtrough
is too high, theremay beaspacebetweentheshield
skinof theTBMandthesoil.Thisisbecausethecutting
wheel of aTBM isinmost casesabit larger thanthe
TBM itself. There is only a small difference (a few
centimetresonaTBMwithadiameter of 10metresor
more), butthisissignificant. Itmeansthatfor aslurry
shieldbentonitecanflowfromthetunnel facebackto
thegrout used to fill up thetail void. However, it is
alsopossiblethat grout canflowfromthetail voidto
thetunnel face.
A 1-dimensional model is developed to get infor-
mation on the order of magnitudes of these flows,
thepressuredistributionandthesoil deformation.The
model assumesagivenpressuredistributionatthetun-
nel face(thefacepressure) andat thetail of theTBM
(thegroutpressure)andincorporatestheyieldstrength
of boththegroutandthebentonite.Ittakesintoaccount
the overcutting of the cutting wheel and the conical
constructionof theTBM (withabit smaller diameter
at thetail compared to thefront). Linear elastic soil
behaviour isassumed.
The paper will briefly describe the model, more
detailedinformationispresentedbyBezuijen(2007),
someexamplecalculations will bepresentedandthe
assumptions will be checked by using 2-D finite
element calculations.
2 GEOMETRY OF A TBM
A TBM shieldseemstobeatubewithaconstant dia-
meter, sketched in a way as in Figure 1. However,
looking more in detail the shield is part of a cone.
Thediameter at thefront islarger thanat thetail. The
differenceisonlysmall, normallyaround0.4%of the
diameter of theTBM. For aTBMshieldwithadiame-
terof 10m, thismeansthatthediameterof theshieldis
4cmsmaller at thetail comparedtothediameter just
after thecutter head.Thissmall differenceindiameter
allowstheTBM tomanoeuvreinthesoil.
255
Figure1. Sketchof aslurryshieldTBM.
3 GROUT FLOWAROUNDA TBM
3.1 Theory
Assume aTBM with a small change in diameter (a
diameter decrease fromfront to back). TheTBM is
boringinsoil thatisassumedlinearelasticwithashear
modulusG.
Assumethat thesoil in contact with theshield at
everylocationof theshield. Thetaperingof theshield
will lead to adecreasein stresses in thesoil around
the TBM going fromthe tunnel face to the tail. A
simpleapproach is to neglect theinfluenceof grav-
ity and assume a tunnel that is positioned perfectly
symmetric in the bore hole. In such a situation the
relationbetweendeformationandstressreductioncan
bewrittenas(Verruijt, 1993):
WhereL isthechangeinpressure, Lr thechangein
radius, r theradiusof thetunnel andthegrout andG
theshear modulusof thesoil aroundthetunnel.
Calculating the pressure fromfront to tail of the
TBM, withouttheinfluenceof groutor bentonite, will
leadtoanongoingpressurereduction.
However, bentonite is injected at the front of the
TBMandgroutinthetail.Normallythebentonitepres-
sureislowerthantheinitial soil pressureandthegrout
pressureishigher. Usingeq. (1) thismeansthat there
will beadiameter decreaseatthefrontandadiameter
increaseatthetail.However,whenovercuttersareused
it isstill possiblethat thereisanopeningat thefront
of theTBMbetweenthetunnel faceandthesoil where
bentonitecanflowover theshieldto thetail. Dueto
thehighgroutpressureitisalsolikelythatgroutflows
fromthetail over theshieldtothefront of theTBM.
Somewhere in the middle around theTBM these
flowswill meet. Calculatingthepressuredistribution
isnowcomplicatedbythefactthatthedirectionof the
Figure 2. Possible flow directions and sketched pressure
distributionsalongtheTBM. The3rdsituationisworkedout
quantitatively, seeFigure3.
flow is of importance. Both grout and bentoniteare
usuallydescribedasBinghamliquids.Thismeansthat
acertain pressuregradient is needed to start aflow.
Thereforethereare3possibleflowsituations, seealso
Figure2.
1 Grout flows fromthe tail to the tunnel face and
bentoniteflowsfromthetunnel facetothetail (this
situationcanonlyoccurwhenthereissomevolume
loss inthegapbetweentheTBM andthesoil for
exampledueto bleeding of thegrout or penetra-
tionof thebentoniteintothesoil. For thissituation
thelowest pressurewill bepresent therewherethe
bentoniteandthegrout meet.
2 Bentoniteisflowingbackwardstothetail andpush-
ingthegrout out of thegapbetweentheTBM and
thesoil. Thepressurewill bethehighestatthetun-
nel faceandwill decreasewhengoingto thetail.
Thiscannotbeacontinuoussituation, butcanoccur
temporally.
3 Grout isflowingtothetunnel faceandpushesthe
bentonitetothetunnel face.Thepressureishighest
at thetail closetotheinjectionpointsof thegrout
andwill decreasegoingtotunnel face.
Themodel developedcaninprinciplecopewithall
these3options. However, for simplicity inthispaper
256
wewill only deal withthe3rdoptionbecausethis is
themostrealisticsituationformostof thedrillingpro-
cess.Asthemachinemovesforward, aforwardflowof
material isneededtofill theannulusthatiscreatedby
theTBM. It isthereforeassumedthat thegrout flows
alongtheTBM shieldandpushesthebentoniteaway.
ConsideringtheadvancingTBM duringdrilling, this
can beastablesituation. Both thebentoniteand the
grout can be described as Binghamliquids. For the
lowflowvelocities duringtunnellingtheyieldstress
of theliquidwill bedeterminingthepressuredropand
theviscousforcescanbeneglected.Assumingthatthe
flowinducedfrictionwill developbetweenthesoil and
thegrout(inthe3rdoptionthegroutandbentonitewill
notmovewithrespecttotheTBMinastablesituation),
thepressuredropcanbewrittenas:
WhereLP is thechangeinpressuredueto theflow,
Lx alengthincrementalongtheTBM, s thegapwidth
betweenthetunnel andthesoil and
theshear stress
of thegrout aroundtheTBM.
The following calculation procedure is used. The
soil aroundthetunnel isassumedtobehaveasindepen-
dentsliceswithathicknessLx. Knowingthegeometry
of thetunnel, thegroutpressureatthetail andtheben-
tonitepressureatthetunnel face, thesoil pressureand
theelastic properties of thesoil, thegapwidthat the
faceandtail of theTBMcanbecalculatedusingEqua-
tion(1). That gapwidthcanbeusedto calculatethe
pressuredropduetotheflowof thegrout tothefront
of theTBM over thedistanceLx with Equation (2)
and thepressureincreasethat will occur in theben-
tonitewhenitispushedtothefront.Theresultinggrout
pressureandbentonitepressureiscalculatedindepen-
dently. As longas at acertainlocationthecalculated
grout pressureishigher thanthecalculatedbentonite
pressure, it isassumedthat thegrout pushestheben-
toniteinthedirectionof thetunnel faceandtherewill
begrout at that location. Inthecasethat thebentonite
pressureishigher, therewill bebentonite.
Theresultof suchacalculationisshowninFigure3.
Theupper plot of thisfigureshowsthepressuresand
gapwidths that wouldoccur if therewereonly grout
(G) or onlybentonite(B) indotsandinlinesthecom-
binedgapwidthandpressure. Inthis calculationthe
grout penetrates between the shield and the soil up
to 1.8mfromthetail. Theremainingpart of thegap
betweentheTBMandthesoil isfilledwithbentonite.
Theamount of penetration varies with thepressures
and shear strength that are chosen. The parameters
usedinthiscalculationareshowninTable1.Although
thereisanopenconnectionbetweenthefaceandtail
of theTBM (Seethelower plot of Figure3, thegap
widthof thecombinedcalculationisneverzero), there
Figure3. PressuresandgapwidthalongaTBM.Groutpres-
suresandbentonitepressures. ParametersseeTable1. Plots
show pressures and gap width for the bentonite and grout
pressureseparatelyandthecombinedresult.
Table1. Inputparametersusedincalculationwithbentonite
andover cutting.
lengthTBM shield 5 m
Diameter 10 m
diameter reduction 0.2 %
over cutting 0.015 m
asymmetric(1) or 2
symmetric(2)
grainstress 150 kPa
grout pressure 400 kPa
porepressure 200 kPa
pressureontunnel face 250 kPa
shear modulus(G) 90 MPa
shear strengthgrout, 1sided 1.6 kPa
frictionassumed
shear strengthbentonite2 0.08 kPa
sidedfrictionassumed
isstill astableboundarybetweenthegroutentheben-
tonite. SincebothliquidsbehaveasBinghamliquids,
therecanbesomedecreaseinthegrout pressureand
still theposition of thegrout-bentonitefront will be
thesame.
257
Accordingtotheresults, thetail of theTBMisnever
incontact withthesoil. Inreality therecanbesome
contact due to forces not taken into account in this
calculation(vertical loadingfromthelining, moments
induced by thehydraulic jacks). However, it is clear
that taperingof theshielddoesnot necessarilyleadto
acertain volumeloss, as was assumed by Kaspar &
Meske(2006). Thecalculationresultspresentedhere
arealso in linewith theTBM data, which normally
showthat moregrout havetobepumpedintothetail
voidthancorrespondingtotheoriginal tail void. The
grout volumethat flowsintothegapbetweenthesoil
and the TBM is not an extra volume, because it is
a constant volume during tunnelling, but the grout
volume increases because the grout pressure makes
awider gap.
IntheWesternScheldtTunnel projectasoil sample
wastakentroughthetail of theshield. Inthat caseno
grout was found between theshield and thesubsoil,
onlysomebentoniteinthesand(Thewes, 2007). This
indicates that inreality thesoil deformationis not as
uniformassuggestedhere. Atsomelocationsthetun-
nel will remainincontactwiththeTBM.Thiswill also
betheresult of thenumerical calculations presented
inthenext session.
4 NUMERICAL SIMULATIONS
Someof theassumptionsinthecalculationmodel have
beentestedbymeansof numerical calculations, using
thePlaxisprogramme.
The assumptions tested are the axial-symmetric
deformationof thesoil that isassumedandtheinflu-
enceof thenonlinearbehaviourof thesand.Therefore
a2-dimensional planestrainsimulationisperformed
usingthemeshshowninFigure4. Thesimulationis
performedfor soil conditionsthat arecommoninthe
westernpartof theNetherlands: A Holocenetoplayer
or10mandatunnel withadiameterof 10membedded
in Pleistocene sand. The small strain hardening soil
model was used in the simulation (Brinkgeve et al.,
2006). TheparametersusedarepresentedinTable2.
In the simulation it is assumed that the pressure
on top of thetunnel is 350kPaat thebeginning and
the pressure is increased up to 400kPa to simulate
the increase in pressure due to grouting behind the
TBM. Inasecondcalculationthestressisfirstreduced
to 320kPa and increased to 400kPa after that. This
simulatestheunloadinginthesoil that canoccur due
to the overcutting and/or lower bentonite pressures.
Theresults will show to what extent a linear elastic
soil behaviour canbeassumed.
The pressure increase with depth in the grout is
assumedtobe18kPa/m.Thecalculateddisplacements
aroundthetunnel openingduetothepressureincrease
areshowninFigure5forbothloadingsituations.With
Figure4. Meshusedin2-Dsimulation.Thediameterof the
tunnel is10m.
Table2. Input parameters for theFiniteElement analysis.
Parameters give (from top to bottom) the wet volumet-
ric weight, elasticity parameters (Brinkgeve et al., 2006),
cohesion, frictionanddilatancyangle.
Holocene PleistoceneSand Grout
sat
kN/m
3
18 20 pm
ur
0.35 0.2 0.499
E
ref
oed
MPa 10 180 pm
E
ref
50
MPa 10 180 0.45
E
ref
ur
Mpa 720
G
ref
0
Mpa 675
0.7
1.00E-04
c
kPa 5 1 15

20 32 0
+

2
Figure5. Deformationpatternaroundtunnel duetoapres-
sureincreaseof 50kPastartingatapressureof 350kPa(left),
resulting in amaximumdisplacement of 22mm. Right the
samecalculationsbut first astressreductionto320kPawas
applied, resultinginamaxdeformationof 25mm.
theparameters usedinthesimulation, themaximum
displacements arein thesameorder as thedisplace-
ments calculated with the analytic model (27mm),
but this depends on the choice of the parameters.
258
It is clear that the displacement is not symmetrical
around the tunnel axis. The soil will predominantly
moveupwards. For thesituationwheretherewasfirst
unloadingthedeformationpatternis abit moreegg
shaped (seeFigure5).
Thedeformationpatternisthereforenotcompletely
axial-symmetricasisassumedintheanalytical calcu-
lations. Furthermore, in reality theTBM can rotate
with a horizontal rotation axis perpendicular to the
axisof theTBM inthejust createdopening. Thiscan
alsoleadtoanunevendistributionof thegaparound
the TBM. Therefore fromthis calculation it cannot
beconcludedwhat will bethegapwidthdistribution
betweentheTBMandthesoil aroundthetunnel atdif-
ferentlocations, whatcanbeconcludedisthatthegap
widthshall not beevenly distributedalongtheTBM
circumference. Simulationsof thedeformationof the
soil movementsfor atunnel at alarger depthresulted
inamoresymmetricdeformationpattern.
5 DISCUSSION
Theresultsof thenumerical calculationsshowqualita-
tiveagreementwiththeresultsof theanalytical model.
However, thesymmetricdeformationthat isassumed
inthenumerical model isnot foundinthenumerical
simulations, but lookingat theresultsit isreasonable
to assume such a symmetric deformation as a first
assumption. Theorder of magnitudeof thedeforma-
tionisthesame, but thisistoalargeextendamatter
of tuning theparameters. Theparametersusedinthe
small strain model lead to a stiffer behaviour of the
sandthanisnormallyassumed. Wheninacasestudy
reliablesoil dataareavailable, boththeelasticparam-
eters andtheparameters fromthesmall strainmodel
canbedetermined.
Theinfluenceof unloadingbeforeloadingseemsto
belimitedandthereforeanelasticmodel canbeused.
Basedonanalytical calculationsof Wang& Dussault
(1994) itcanbeexpectedthattheinfluenceof unload-
ing becomes moreimportant when thepressures are
decreasedfurther.
Themodel,asitisnow,issuitabletogetaqualitative
indication on what is theinfluenceof bentoniteand
groutflow. Inthiswayitisastepforwardcomparedto
modelsthatsimplyassumethatthesoil flowsover the
TBM andthat predict asettlement troughthat is too
large. For quantitativeresultsandtoseewhat partsof
theTBMareincontactwithsoil andwhatareincontact
withthegroutorbentonite, moresophisticatedmodels
areneededthentheonedescribedhere.
6 CONCLUSIONS
A onedimensional analytical model has been devel-
opedtodescribetheinfluenceof bentoniteandgrout
flowaroundaTBM onthepressuredistributioninthe
soil aroundtheTBM andtheinfluenceonthesettle-
ment trough. Calculations withthis model showthat
theinfluenceof theflowalaongtheTBM canbesig-
nificant, leadingto calculatedvolumelosses that are
moreinagreement withthemeasurementscompared
toresultsof calculationsthat donot takeintoaccount
theflowaroundtheTBM. A consequenceis that the
TBM is (partly) not incontact withthesoil. At high
soil pressuresandthereforealsohighgrout pressures
this may increasetherisk of deformation of thetail
shield.
Comparingtheresultsof theanalytical model with
theresultsof numerical calculationsshowedthatsome
of the assumptions in the model (symmetric defor-
mation, linear elastic soil behaviour) areasomewhat
crude representation of reality. This means that it is
worthwhileto includetheprinciples describedinthe
paper in more sophisticated numerical models. The
model as described herecan beused for afirst esti-
mateonwhat is theinfluenceof bentoniteandgrout
flowaroundaTBM onthesettlement trough.
REFERENCES
BezuijenA. 2007. Bentoniteandgrout flowaroundaTBM.
Proc. ITA 2007, Prague.
aboredtunnel. Proc ICOF 2003 Dundee.
Fieldmeasurements of grout pressures duringtunneling
of theSophiaRail tunnel. Soils and Foundations 44(1):
4150.
Bowers, K.H. & Moss, N.A. 2005. Setllement due to tun-
nellingonhteCTRL Londontunnels. Proc. 5th. Int. sym-
posium on Geotch. Aspects of Underground Construction
in Soft Ground, Amsterdam.
Brinkgreve, R.B.J., Bakker, K.J. & Bonnier, P.G. 2006. The
relevanceof small-strainsoil stiffnessinnumerical simu-
lationof excavationandtunnellingprojects. Proceedings
Numog XII, Graz, Austria.
Hoefsloot, F.J.M. &Verweij, A. 2005. 4DGroutingpressure
model PLAXIS. Proc. 5th. Int. symposium on Geotch.
Aspects of Underground Construction in Soft Ground,
Amsterdam, pp 529534.
Kasper, T. & Meschke, G. 2006. On the influence of face
pressure, grouting pressure and TBM design in soft
ground tunnelling. Tunnelling and Underground Space
Technology 21 160171.
Thewes, M. 2007. Privatecommunication.
Verruijt, A. 1993. Soil Dynamics. Delft University of Tech-
nology, b28.
Wang,Y. &Dusseault, M.B. 1994, Stressesaroundacircular
openinginanelastoplastic porous MediumSubjectedto
repeatedhydraulicloading. Int. J. Rock Mech. Min. Sci. &
Geomech abstr. 31(6): 597616.
259
Mechanismsthat determinebetweenfractureandcompaction
groutinginsand
A. Bezuijen&A.F. vanTol
Delft University of Technology/Deltares, Delft, The Netherlands
M.P.M. Sanders
Present affiliation: Royal Haskoning, The Netherlands
ABSTRACT: Laboratory tests haveshownthat injectionof grout insandcanleadtodifferent shapes of the
grout in the sand. At low water cement ratios (12) there is usually compaction grouting, which leads to a
spherical shapeof thegrout. At higher water cement ratios(520) fracturingof thesandwill occur, leadingto
rather thingrout structures. Infieldobservationsthisdifferenceislessobvious. It isassumedthat inmost cases
fractures occur, although it is hardly ever possibleto examinewhat is madein thesoil. This paper describes
conceptual modelsandcalculationmodelsthat explain, at least qualitatively thedifferencesinthebehavior of
thegrout anddiscussesapossiblereasonfor thedifferencebetweenthemodel testsandthefieldtests.
1 INTRODUCTION
Compensationgroutinghasbeensuccessfullyapplied
inseveral projectstopreventor tocompensatefor sur-
face settlements induced by for example tunnelling
(Mair &Hight, 1994; Chiriotti etal. 2005; Christiaens
et al. 2005). Compensation grouting uses hydraulic
fracturingtoget aheavethat cancompensatetheset-
tlementor compactiontodensifythesoil withlimited
heave. In the latter situation the compaction lead to
improvement of thesoil characteristics.
Laboratorytestshavebeenperformedtoinvestigate
themechanismsthatareof importancefor compensa-
tion grouting in clay (J aworski et al. 1981, Mori &
Tamura 1987, Andersen et al. 1994, Chin & Bolton
1999, Au, 2001). Comparable tests have been per-
formed for compensation grouting on sand (Chang,
2004, Kleinlugtenbeltetal. 2006, Bezuijenetal. 2007
and Gafar & Soga, 2006). These tests showed that
compensationgroutingbyfracturingisnot sostraight
forwardinsand. After quiteanumber of tests witha
water cement ratioof 12it appearedthat therewere
hardly any fractures in the sand, but a more or less
spherical shape of grout is formed in the sand, see
Figure1. Fracturescouldbeobtainedwhengroutwith
ahigher water cement ratio (520) is used(Bezuijen
et al. 2007).
Similar results were found in at least 3 different
laboratories. Different resultswere, however reported
fromthefield(Grotenhuis, 2004) basedonmeasured
Figure 1. Results from Cambridge University (left) and
GeoDelft that showmoreaspherical shapethanfractures.
surfacedisplacements. Normally it is not possibleto
investigatethefracturescreated, but thiswaspossible
duringtheconstructionof thestationforthehighspeed
train inAntwerp and during the construction of the
cross-passagesat theHubertusTunnel inTheHague,
theNetherlands.
InAntwerpcompensationgroutingwas appliedto
preventsettlementsontheexistingstation(Christiaens
et al. 2005). Theresulting fractures werefound in a
soil layer with a lot of shells. A clear fracture was
foundbetweentheshells. Hereit is possiblethat the
shellshaveinfluencedthefracture.TheHubertusTun-
nel casewasanunintendedfracturethatoccurredwhen
thelancesfor freezingwerebrought inandthegrout
pressureusedtoinstall theselanceswaslargeenough
tocreatefractures.
261
Thepaper will describetheHubertusTunnel case
moreindetail, sincethisisoneof thevery fewcases
where fractures in relatively homogeneous sand at
11m depth and 5m below the waterline could be
studiedanddescribestheprincipleandresultsof cal-
culationmodelsthataredevelopedtodescribevarious
parts of thegrouting process. Possiblecauses of the
differencesbetweenthefieldtest andthemodel tests
are: 1) the preconditioning has an influence. 2) the
in-homogeneities in real soil serve as a trigger for
thestartof theprocess. 3)Theinstallationof theTAMs
(TAM standfor Tube-a-Manchette, atubewithinjec-
tionopeningscoveredbyarubber Manchettethrough
whichthegroutcanbeinjected) whichcausesunload-
ing of the soil has an influence. In this paper we
will elaborate the 3rd possibility and mention some
aspects of the 2nd. The first possibility is not dealt
withinthispaper.
Results of themodel tests itself will bedescribed
and analyzed in another paper on this conference
(Gafar at al., 2008).
2 THE HUBERTUSTUNNEL
TheHubertusTunnel isadoubletrackroadtunnel that
is constructed in dunesand. Thetunnels haveadia-
meter of 10m. Fiveconnections weremadebetween
thetwotunnels. Soil freezingwasusedtomakethese
connections. Lanceswereconstructedfromthetunnel
lining into the sand to freeze the soil. A connection
betweenthetwotunnelswasmadethroughthelinings
andtheconcreteconnectionstructurewas made. For
across-sectionseeFigure2.
Thetunnel crownisplaced11mbelowthesoil sur-
faceandthephreaticlevel is5mbelowthesurface.The
vertical pressureat thetopof thetunnel is estimated
tobearound140kPa.
Grout was used to install the lances for freezing.
Duringtheinstallationof theselancesforthefirstcon-
nection,thepressurewasincreasedsofarthatthegrout
created hydraulic fractures. These fractures became
visiblewhenthefrozensoil wasremoved.
Figure3showsanexampleof fracturesinthesand
aroundthetunnel andbetweenthetail voidgrout and
thesand.
Thepictureis takenstandinginthenewly created
opening, fromleft to right you see the opening, the
tunnel lining, thetail void grout, grout fromalance
thatpenetratedbetweenthetail voidgroutandthesand
andthefrozensandwithafractureintheupper part
of thesand. Therewerealso fractures that wereonly
visibleinthesand.
The normal pressure to inject the grout is 3 bar.
However, accordingto theemployees of thecontrac-
tor on the site the pressure could have been higher
duringthisinjectionbecausethereweredoubtsonthe
Figure2. Sketchof aconnectionbetweenthetunnelsmade
byfreezing.
pressure gauge. The grout had a water-cement ratio
(W/C) of approximately 1 and 3%of bentonite was
usedinthemixture.
3 COMPARISONWITHLABORATORY TESTS
The laboratory tests performed at GeoDelft, as
described by Bezuijen et al. (2007) and Gafar et al.
(2008) havebeenperformedinacylindrical tankwith
adiameter of 0.9mand aheight of thesand sample
of 0.85m. Thesandwas pressurizedto get avertical
effective stress of 100kPa. Different grout mixtures
wereinjectedanddifferent injectionprocedureswere
used. Most mixtures injected had 5% or 7% ben-
tonite. A fracturelikebehaviour only occurredwhen
theW/C ratio inthemixturewas 2or higher. Lower
W/C ratioswouldnot createfractures, but leadtothe
shapes shown in Figure1 (right picture). Themaxi-
muminjectionpressuresvariedinmostcasesbetween
20and30bar.
Comparingthemainresultsof themodel testswith
that of thefieldobservations, thereseemstobeadis-
crepancy. Thegrout that does not leadtofractures in
the model, does result in fractures in the field. The
injection pressure in the field situation described is
not known, but it is unlikely that where the normal
262
Figure3. Fracturesfoundduringtheconstructionof thefirst
connectionbetweenthetunnelsof theHubertustunnel.
injection pressure is 3 bar, it was 28 bar during the
constructionof thefirst cross-section(this wouldbe
the injection pressure based on the laboratory tests
andextrapolatingtheresultsof theseteststo140kPa
vertical pressure).
A possiblereasonfor thediscrepancy betweenthe
model tests and the field behaviour can be the soil
boundary condition, as will beexplainedin thenext
sections.
4 RELEVANT GROUT PROPERTIES
4.1 Plastering
Bezuijen&VanTol (2007) havedescribedhowplas-
teringof thegroutinfluencesthefracturingbehaviour.
Fracturing occurs because the radial stress in the
sand around a fracture is higher than the angular
stress. UsingtheMohr-Coulombcriterion, therelation
betweenradial stressandangular stressperpendicular
ontheradial stresscanbewrittenas:
Where
r
is theradial stress,
thetangential stress
and thefrictionangleof thesand.Takingforexample
Figure4. Sketch with possibledeformation modes of the
injectionhole. Inrealitytherewill bemoregrainsaroundthe
injectionhole. Thefigurejust showstheprinciple.
africtionangleof 35degrees, thetangential stressis
only0.27timestheradial stress.
Looking at the scale of the grains, sand is not
homogeneousandisotropic, asassumedincontinuum
mechanics. At theboundaryof theopeningtherewill
be some grains that are in closer contact and some
betweenwhichthereissomespace, seeFigure4.
Suchspacecanbesufficient tohavethefluidpres-
surenot onlyinradial but alsointangential direction.
Thefluidpressureis withperfect plasteringequal to
theradial stress andmuchhigher thanthetangential
stress and therefore this will lead to an opening of
thespacebetweenthegrainsandafracturecanoccur.
Some in-homogeneity in the soil as is often present
inafieldsituation(morethaninthemodel test) will
facilitate the fracture formation, because weak sec-
tionsinthesoil will deform. If thereisabeginningof
afracturethenthefluidpressurewill penetratefurther
inthesandandthefracturewill easily growfurther.
A fracturewill stoptopropagatewhenthepressurein
thefracturetipdropsduetofrictionlosses, leak-off or
increaseof thevolumewithinjectionfluid.
Plastering and the formation of a filter cake in
theinjectionholewill hamper fracturingof thesand,
becausenowpossibleirregularities at theboundaries
of the injection hole will be filled with plaster see
Figure 5. This plaster also has certain strength and
thereforepreventsthefluidpressurefrompenetrating
intothespacebetweenthegrains.
AsisalsomentionedinBezuijenenVanTol (2007),
plasteringiscausedbothbyconsolidationof thegrout
andby leak-off of grout inthesandwhilelarger par-
ticles inthegrout arefilteredby thesand, remainat
intheinjectionholeandcauseaplasteringlayer atthe
boundarybetweentheinjectionliquidandthesand.
With aconstant injection pressure, theconsolida-
tion of the grout mixture, without leak-off, can be
approximatedbyformula(Bezuijenet al. 2007):
263
Figure5. Influenceof plastering.
Wheres
1
isthethicknessof theconsolidatedlayer(the
plasteringlayer), n
i
theinitial porosity, n
e
theporosity
after consolidation, k thepermeabilityof theconsoli-
datedlayer,L thedifferenceinpiezometricheadover
thecolumnandt thetime.
Thisformulaimpliesthat thethicknessof theplas-
tering layer increases with the square root of the
appliedpressure, alowerpressurewill leadtoathinner
plasteringlayer.
4.2 Leak-off
Assuming that the flow during leak-off can be
described as a Bingham liquid flow, the relation
between the difference in piezometric head and the
thickness of theplasteringlayer of courser materials
canbewrittenas(for v >0, flowintothesand):
Where L is the difference in piezometric head, s
1
thethickness of thegrout cake(theplasteringlayer),
s
2
thedistancetheliquidhaspenetratedintothesand,
k
1
and k
2
the permeability of the cake and the sand
respectively for thepenetratingliquid, v thevelocity,
1
and
2
factorsforthecakeandthesandrespectively
thatdeterminethedropinpiezometricheadcausedby
the yield stress in a Binghamliquid and R the flow
resistance fromthe soil around the fracture. When
this equation is dominant, thethickness of theplas-
teringwill beproportional withtheapplieddifference
inpiezometrichead.
Experimental research by Sanders (2007) has
shownthat consolidationdeterminesthethicknessof
theplasteringlayerforgroutwithalowW/Cratio, leak
off isthedominantmechanismforthethicknessof this
plasteringlayeratW/Cratiosof 5andmore. Normally
afiltercakecausedbyconsolidationof leakoff ismade
of finer particlesthanthesandof thesubsoil. Dueto
thisleakoff will stopwhenthereissomethicknessof
thefilter cake.
Figure6. Sketchof casingandTAM duringinstallation.
5 INFLUENCE OF SOIL DEFORMATION
IntheexperimentspreformedinWaterloo, Cambridge
andDelft, thesandaroundtheinjectionpoint is rea-
sonableuniformandthevertical confinementpressure
isconstantoverthesoil sample. Injectionof groutwill
leadtocavityexpansionuntil afractureoccurs.
A different pressuredistributionis present around
theinjectionpointsattheHubertusTunnel andingen-
eral when compensation grouting is applied in the
field. At theHubertustunnel thefreezinglanceswere
appliedbetweenthetwotunnels. Consolidationof the
tail voidgrout leadstoastressreductioninthevicin-
ity of thetunnel (Bezuijen&Talmon2003). Arching
betweenthe2tunnels canleadtoafurther reduction
of thestresses.Alsotheplacementof lances, orincase
of compensationgroutingof TAMs, leadstoareduc-
tionof stresses. TheTAMsareplacedusingacasing;
seeFigure6for asketch. Figure6showsthesituation
wheretheTAM isplacedbydisplacement of thesoil.
Another optionisthat thebentoniteisusedtoremove
thesoil. Thelatter optionleadstomoreunloadingof
thesoil. Whatever optionis chosen, whenthecasing
is removed therewill besomeunloading of thesoil
althoughgroutisinjectedtostabilizetheborehole, as
appearsfrom(minimal) settlementsmeasuredduring
theinstallationof theTAMs. (Kleinlugtenbelt, 2006).
The situation that occurred during compensation
groutingcanbecomparedwiththesituationof acavity
underrepeatedloading. Suchasituationisanalyzedby
Wang&Dusseault(1994).Theyusedcavityexpansion
theory to calculate the stress distribution in the soil
aroundaboreholeduringunloadingandre-loading. In
thecaseof compensationgrouting, theremoval of the
casingwill result inunloadingandthereloadingwill
occur duringcavityexpansion. Cavitycontractionand
expansiontheorycannotdescribewhenafracturewill
start, butitcanbeusedtocalculatewhatpressureswill
leadtoaplastic zoneinthesoil. Someplastic defor-
mationinthesandis necessary beforeafracturecan
occur. Plasticdeformationwill leadtodilatancyinthe
sand, whichresultsinmorespacebetweenthegrains.
Thismakesiteasierforthegrouttoseparatethegrains
further accordingtothemechanismdescribedinSec-
tion4.1leadingtoafracture, becausetherearelocally
zoneswithahigher permeability.
264
Parts of Wang & Dusseaults analyses will be
presented here in a shortened and slightly adapted
versiontoallowcalculationof theinfluenceof previ-
ousunloadingontheloadingpressureatwhichplastic
deformation occurs. It is assumed that after installa-
tion of aTAM thereis apressurereleasedueto the
removal of the casing that leads to an active failure
of thesoil aroundthehole. Thismeansthat theradial
stress,
r
, issmallerthanthetangential stress,
.After
installation, theradial pressureisincreasedduetothe
fracturingprocessandpassiveyieldwill occur. Insuch
asituation
r
,islargerthanthetangential stress,
and
a fracture can occur. Here we assume that the frac-
tureinitiationpressureisrelatedwiththepressurethat
resultstopassivefailureof thecavity.
Assuming a linear Mohr-Coloumb criterion, the
plasticstressesmust fulfill thecriterion:
Wheretheprimeindicatestheeffectivestress.
Thematerial parameters N andS aredifferent for
activeandpassiveyieldandcanbewrittenas:
and
(Thepublicationof WangandDusseaultdiscriminates
betweenpeak valuesandresidual values, hereweuse
onlyonevalue.)
Here isthefrictionangleandc
0
thecohesionof
thesoil material. Activeyieldwill occur when:
Wherep
a
isthepressureattheboundaryof theopening
intheactivestate(assumingaperfectplasteringonthis
boundary) and
0
theinitial effectivestressaroundthe
opening.
When the effective pressure around the TAM
remainshigherthanthecriterionmentionedinEq. (7),
therewill benoplasticdeformationandthelimitpres-
sure for passive plastic deformation will remain the
sameasif therewasnounloadingonthesoil.
Incasethecriterionof Eq. (7) isfulfilled, therewill
beanactiveplasticzone. Increasingthepressureinthe
cavity afterwards will lead to, thesituation sketched
inFigure7. Therewill beanactiveplastic zoneand
withinthat apassiveplastic zone. Whenthepressure
isincreasedfurther theactivezonewill disappear, but
Figure7. Definitionsketch, forthesituationwithactiveand
passivesoil aroundthecavity, seealsotext.
Table1. Input parametersusedincalculation.
0
original stressinsand 100 kPa
frictionanglesand 35 degr.
c
0
cohesion 0 kPa
R
w
(radiusof tube, seeFigure7) 0.035 m
b (radiuswithconstant press) 5 m
our interest is thepressureat whichthepassivezone
started.
Duringunloading, whentheactivezoneisformed
P
r
, Based on the equations presented by Wang and
Dusseault, thepressureinthecavity, canbeexpressed
as afunction of thethickness of theplastic zoneR
a
andother parameters(seealsoFigure7):
Where:
Whenafter activeyieldthepressureis increasedthe
first soil will comeinthepassiveplasticstatewhen:
Without theactivefailurethisrelationwouldread:
Table1presentstheinput parametersusedfor cal-
culationsusingtheformulaspresentedabove. Results
265
Figure8. Pressurethat isnecessary tocreatepassiveplas-
tic deformation after that first activeyield is imposed as a
functionof theoriginal activeplasticzone.
are shown in Figure 8. Without active plasticity the
pressure where passive plasticity starts is 157kPa.
Fromtheseresultsitisclearthatplasticdeformationin
theactivestatecanconsiderably reducetheborehole
pressureat whichpassiveplasticdeformationoccurs.
Thepressureneededtoachievepassiveplastic defor-
mation decreases as theactiveplastic radius created
by the low pressure before the pressure increase is
larger. This presents quantitativeinformation for the
statement already mentioned by Wang & Dusseault:
Our studysuggeststhat initial activeformationdam-
age reduces the pressure required to initiate such a
fracture.
6 DISCUSSION
In Section 4 it was shown qualitatively that only
with alimited filter cakebetween thegrout and the
subsoil a fracture can occur in homogeneous soil.
Inin-homogeneoussoil theformationof afractureis
easier, because also the initial deformation will not
be symmetric. For this reason there was a fracture
betweenthetail voidgroutandthesandattheHubertus
Tunnel.
Thecalculationmodel by WangandDusseault, as
describedinSection5, showsthat thepressuretocre-
ateafractureisconsiderablylower incaseof acavity
thathasbeensubjectedtoplasticunloading, compared
toacavitythatisstill inanundisturbedstate. Thislast
phenomenonis themost reasonableexplanationthat
the injection pressure in the field is lower than the
pressuremeasuredinthelaboratory.A lower injection
pressurealso means that thefilter cakedueto plas-
tering and leak-off is thinner and thus that thegrout
mixtureislesscritical.Thismaybeanexplanationthat
fractures are created in the field for conditions that
does not lead to fractures in the laboratory. Further
experimental workhastoproveandquantifythisidea.
Consequencemaybethat thecompensationgrout-
ingissandisinfluencednotonlybythegrout, butalso
toalargeextendby thestressdistributionaroundthe
boreholewheretheTAM isinstalled. Theinstallation
procedurewill influencetheshapeof thefractureand
thepressureneededtocreateafracture.
7 CONCLUSIONS
Fromourstudywecametothefollowingconclusions:
1 Fractures can occur in afield situation wherefor
comparablesoil andusingthesamegroutonlycav-
ityexpansionismeasuredinamodel test. Thiscan
becausedby theheterogeneity inthefieldandby
thewaytheTAMsareinstalledinthefield.
2 Unloadingof thesoil beforeitisloadedbyinjection
of grout will lead to a reduction of the injection
pressure.
3 A reducedinjectionpressurewill leadtoareduced
cakethicknessbecausebothbleedingandleak-off
will bereduced.
4 In further experimental research it will beneces-
sary toinvestigatetheinfluenceof theinstallation
procedureandin-homogeneityof thesoil.
ACKNOWLEDGEMENT
ThisworkwassupportedbyDelftCluster.Theauthors
wishto thank TheHubertusTunnel Combinationfor
theinformationpresentedonthefirstcross-connection
of thistunnel.
REFERENCES
Andersen, K.H., Rawlings, C.G., Lunne, T.A. & Trond, H.
1994. Estimation of hydraulic fracture pressure in clay.
Journal of Canadian Geotechnical 31: 817828.
Au, S.K.A. 2001. Fundamental study of compensation grout-
ing in clay. PhDthesis. Universityof Cambridge.
Bezuijen, A., Sanders, M.P.M., Hamer, D. & Tol, A.F. van
2007. Laboratory tests on compensation grouting, the
influenceof groutbleeding. Proc. WorldTunnel Congress,
Prague.
aboredtunnel. Proc ICOF. Dundee: ThomasTelford.
Bezuijen,A. &Tol,A.F. van2007. Compensationgroutingin
sand, fracturesandcompaction. Proc. XIVEuropean Con-
ference on Soil Mechanics & Geotechnical Engineering,
Madrid.
Chang, H. 2004. Hydraulic fractures in particulate materials,
Phd. Thesis, GeorgiaInstituteof Technology, November.
Chin, C.Y. & Bolton, M.D. 1999. Factors influencing
hydrofracture in clay. Proc. 13th ASCE Engineering
Mechanics Conference, Baltimore.
Chiriotti, E., AvgninaN. & Grasso P. 2005. Compensation
groutingforTBM tunnelingbeneathshallowcover. Proc.
5th Int. Symposium Geotechnical aspects of underground
construction in soft ground, Amsterdam2005.
266
Christiaens M., Hemerijckx E. & Vereerstraeten J.C. 2005.
Tunnelling under the city centre of Antwerp. A new
underground Railway link for the HSL Paris-Brussels-
Amsteram. Proc. 5th Int. Symposium Geotechnical
aspects of underground construction in soft ground,
Amsterdam2005.
Gafar, K. & Soga, K. 2006. Fundamental investigation of
soil-grout interaction in sandy soils. Report. University
of Cambridge.
Gafar, K., Soga, K., BezuijenA., SandersM.P.M. &Tol A.F.
van. 2008. Fracturing of sand in compensation grout-
ing. Proc. 6st Int. Symposium on Geotch. Aspects of
Underground Construction in Soft Ground, Shanghai.
Grotenhuis, R. 2004. Fracture Grouting inTheory, Modelling
of fracture grouting in sand. MScthesis, Delft University
of Technology.
J aworski,G.W.,Seed,H.B.&Duncan,J.M..1981.Laboratory
studyof hydraulicfracturing. Journal of the Geotechnical
Engineering Division 107(6): 713732.
Kleinlugtenbelt,R.2005.Compensationgrouting, laboratory
tests in sand. MScthesis. DelftUniversityof Technology.
Kleinlugtenbelt, R. 2006. PrivateCommunications.
Kleinlugtenbelt, R., Bezuijen, A. & Tol A.F. van. 2006.
Model testsoncompensationgrouting.Proc. WorldTunnel
Congress, Seoul.
Mair,R.J.&Hight,D.W.1994.Compensationgrouting.World
Tunnelling November: 361367.
Mori, A. & Tamura, M. 1987. Hydro-fracturing pressure
of cohesive soil. Journal of the soil and foundations,
Japanese Society of Soil Mechanics & Foundation Engi-
neering 27(1): 1422.
Sanders, M.P.M. 2007. Hydraulic fracture grouting, lab-
oratory tests in sand. MSc thesis. Delft University of
Technology.
Wang,Y. &Dusseault, M.B. 1994. Stressesaroundacircular
openinginanelastoplastic porous MediumSubjectedto
repeatedhydraulicloading. Int. J. Rock Mech. Min. Sci. &
Geomech abstr 31(6): 597616.
267
Researchof non-motor vehicle-rail transit-tubeinterchanging
transport systempattern
A.Z.G. Deng& Q.H. Zhang
ABSTRACT: Inthelightof therelationshipbetweenlanduseandrail transit, itconsidersthatthedevelopment
of thepublictransportisoneof theeffectivewaystosolvetheproblemsintraffic. Urbanrail transitthatcanlead
thedevelopmentof thecitiesshouldbethebackboneinthepublictraffic. Thispaper presentstheresearchabout
non-motor vehicle-rail transit interchangingtransport systempatterns. Theinterchangepatternisprovedtobe
feasiblebyinvestigationandcalculationresearch. Thepurposeistomakepublictransportationmoreeffective,
andelevatetheleadingroleof rail transit. Thepaper alsoindicatestheeffectivewaystosolvetheproblemsin
trafficinorder toexalt theproportionof theurbanrail transit.
1 GENERAL INTRODUCTION
Nowadays, therearemanyresearchesabout theinter-
changeof rail transit andpublic transportation(such
asairport, railway, subway) alongwiththeincreasing
urban rail transit lines in Beijing and Shanghai etc.
However, these researches hardly mention the inter-
change of non-motor vehicle traffic to rail transit.
Throughtheinvestigationongeneral distributionsta-
tion (refers to the station around the inhabited area,
public transportation, non-motor vehicle and walk-
ingsystemprimarily), thispaper presentstheresearch
about non-motor vehicle-rail transit interchanging
transport systempatterns. Thepurposeistofacilitate
thepublictransportation, andelevatetheleadingrole
of rail transit. Withthedevelopmentof rail transitnet-
work, it isvery important toinvestigatethemodel of
theaffectedfactor instationandthelandusearound
toprovideacomfortable, safeandconvenientenviron-
ment intransit interchanging. Thepaper proposesthe
effectivewaystosolvetheproblemsintraffic, inorder
toexalt theproportionof theurbanrail transit.
2 THEORY OF THE LANDUSEAROUND
THE RAIL TRANSPORT STATIONS
Thefundamental characteristic of therailway station
can be summarized as the following: forming the
buildings around therailway station, which generate
the core-axis pattern, developing the traffic organi-
zationandfunctionlayout withthecenter of railway
station. Landisimportanttoacity, andreasonableuse
of landwill dobenefittotheeconomy, societyaswell
as the ecosystem. The aimof land use shows com-
plex, multidimensional andintegrativeinspaceform,
which request closecooperation between infrastruc-
turedevelopment and civil development to optimize
landuse. Intherangeof railway station, it shouldbe
emphasized of the relation and merge between dif-
ferent city functions, which can make people enjoy
theuseof railway andembody thetheory of people
oriented.
In developed countries, somenewurban planners
break throughthelimit of thetraditional urbanplan-
ningtheory under thecitys sustainabledevelopment
background, andproposedTransit-OrientedDevelop-
ment Model, i.e. TOD Model. It is expected theuse
of public transport to leadthedevelopment of acity,
andback to themodel of usingbicycleandwalking.
Themaincontent of TODis: Usingpublictransporta-
tionstationas acenter, walkingdistanceas aradius.
In this area, high density of the land use is empha-
sizedandpublictransportationfacilitiesarearranged
aroundstation. Bywalking, bicycleandpublicbusetc,
higheffectiveinterchangeto replacetheleadingrole
of car incity.
Becauseof thediversity between Chinaand west
developed countries on population density and city
developmentstage, therearesomedifferencesof using
TODtheory, explainedasfollows:
1. Different background: After theWorldWar II, the
cars become popular in cities of US, the energy
269
wastebecomesserious, suburbanizationandurban
blightof thecitydowntown. TheTODconceptwas
right proposed under this background. But cities
inChinaarestill inhighlydevelopment. Themain
problemis how to develop public transportation,
to solve the traffic jamcarried up by high den-
sity of citys population and deterioration of the
environment.
2. Population density: In US, low population den-
sity, andthemaintransportationhome-based-trips
are cars used. But in China, population density
is higher. Even the residential area in suburb,
thepopulation density is still higher than foreign
country.
3. Willingness to usepublic transportation: Because
of thedistributedsuburb, highpercentageof peo-
ple owing cars and more highways, the public
transportationhas less attractionto publicationin
westerncountries. ButinChina, itisbecauseof the
large population, short of effective facilities, and
undeveloped construction techniqueand manage-
ment techniquethat therealways exits thetraffic
jam, lowerspeedof transportationandlackof com-
fortableconditions. Therewill beavicious circle
alongwiththemorecarsusing.
4. The difference of influence radius of station: In
foreignTOD model, there are a 600 meter influ-
ence radius with the center of the station. And
theinterchangewith rail transit is most based on
walksystem. But inChina, thedensityof rail tran-
sit network is lower, instead of higher population
density, andtheinfluenceradiusof railwaystation
is larger, peopleuseinterchangemostly by public
bus andbicycleetc withrail transit, whichmakes
theinterchangepattern of therailway station and
surroundingareadifferent.
It can be seen fromthe operation of foreign big
cities transportation that the percentage of railway
passengeris45%to60%tothetotal one. ButinChina,
thepercentageof busmakesover75%, andthatof rail-
wayismuchlower. It isbecausethat thedevelopment
of rail transit transportationis onthebeginning, and
hasnt formed atransport net. Thedensity of station
and network is over 1500kmin London, NewYork,
ParisandTokyo, andthereisahighrailwaydensityin
down town also. Thedensity of railway and stations
in town is shown inTable1. It will generally take5
to 10 minutes for peoplewalking fromhometo the
railwaystation. TakeBeijingfor example, thedensity
of subwayindowntownisonly0.32km/km
2
, whichis
onetenthof that comparedtodevelopedcountries. It
ispointedout inliteraturethat thedowntowndistrict,
middledistrict and outskirt district of abig city can
betakenas1.6, 0.8and0.4km/km
2
basedonChinas
situation and it is found acceptable. It is still a gap
of thedensity of therailway systemwith developed
countries.
Table1. Thedensityof stationandnetworkinLondon, New
York, ParisandTokyo.
London NewYork Paris Tokyo
Densityof network 3.49 3.47 3.76 2.6
km/km
2
Densityof station 2.52 3.92 4.58 1.59
unit/km
2
Inconclusion, duetolownetworkdensityinchina,
the application of TOD model in Chinas cities will
bemodifiedas: it shouldbebasedonthestationasa
center, rail transit, bus andnon-motor vehicles as an
interchangefor development. Themaintask isnot to
increase the residential density but how to combine
landuseandpublictransporttogether, whichcanpro-
videacomfort, convenientandattractivepublictrans-
port system. Bydoingsocanwemakeuseof thenet-
workeffectivelyasthebackboneincitysdevelopment.
3 RAIL TRANSPORT SYSTEM PATTERNSIN
DIFFERENT CHARACTERISTIC STATION
To different railway transportation types of stations
(It is classified into large scale interchange station,
general interchange station and general distribution
station) area, accordingly there are different major
functions (seen in Table 2). Generally, for station
in downtown area(large-scaleinterchangestation or
general interchangestation), thecommercial business
development areaconsist upthemajority of this dis-
trict, and the residential development area takes up
theminority of this district. For stationinoutskirt or
newly developedarea(general interchangestationor
general station), themostsuitabledevelopmentinthis
areashouldberesidential development.
In planning the station and surrounding area, the
main concern point should be the land use around
thestation. For example, incityCBDarea, thestation
shouldbebuiltasthecommercial center of, especially
ininterchangestationof multi-triprailway lines. The
underground space exploitation should also be con-
sidered to develop a district, which is incorporating
transportationandbusinessor other functionstoform
anundergroundurbancomplex.
For interchangestationinoutskirt of thecity or in
development zone, it should be considered together
with other transport method e.g. privatecars, buses,
non-enginecars.
For stations at suburbs of the city, the majority
method of interchangeis using Cars Peripheral Park
andRide(P&R). TheP&Rreferstotheinterchanging
facilitiesforpeopleinoutskirtparkingcarsaroundthe
railwaystationor publictransportationsendstop. By
usingpublic transportationsystemtodowntown, the
270
Table2. Classificationindifferent rail transit station.
Classificationindifferent
rail transit station Landusearoundstation Different interchangeforms
Rail transit Large-scaleinterchange Towncenter commercial Withinrail transits(3or
station station circleandCBDarea over 3lines)
Citysuburb Rail transit withair plane
Citydistributioncenter Rail transit withrailway
Rail transit withRoadpassenger
transportation
commercial areaand Rail transit withpublictransport
residential district
General interchange Citysub-center Withinrail transits(2lines)
station Rail transit withpublictransport
Development zone Peripheral ParkandRide(P&R)
General distribution Residential district Rail transit withpublictransport
station Rail transit withnon-motor vehicle
Rail transit withwalksystem
use of private cars in down town area is decreased.
Theoutskirt interchangefacilitiesarethelink of cars
andpublictransportationsystem. It major functionis
toprovidetheeffective, safe, convenient, comfortable
linkagebetween outskirt cars systemand downtown
public transportation system, so as to increase the
attraction of the public transportation system and
decreasetheuseof carsindowntownarea. Bydoing
so,thetransportationpressureindowntownareacould
bediminishedandrealizethesustainabledevelopment
of thecitystransportationsystem.Thismethodissuit-
ablefor outskirt of thecity. Therailway stationmust
providethesufficient parking facilities, parking lots
to meet the interchange need. The parking facilities
must benear tothestation, andhaveconnectionpath
tothestation.Theenoughroadsshouldbeplannedand
constructed.
Regardinginurbangeneral distributionstation,pre-
cededby text analysis, themain consideration is the
walk system, the non-motor vehicle transportation
systemandthebus, thetaxi convenient interchanges.
The walking systemis the most primary connec-
tionwiththerail transit, thecontent mainly includes
sidewalks systeminstation, thefacility andthepas-
sengers separates fromvehicle facility plan design,
theguidanceinformational signdesign,walkingroutes
organizationdesignandsoon. Thereforethewalking
systemshouldbegivenpriority inrail transit design.
Consideredtransportationstationaroundthelanduse
intensity is high, passengers activity is alsofrequent,
and in order to increase commercial stores suitable
sidewalksshouldbeprovided.Trafficislandandcross-
walk must beconsiderableanddesignedinthis area.
The walk systemin the station which connect plat-
formmeets the convenient needs in the station, but
also achieve the evacuated request, simultaneously
alsomust havedistinct guidancesymbol.
Thebicycleisoneof effectivewaysconnectedwith
rail transit also. Enoughbicycleparksshouldbepro-
videdinjointsof stationdesign, appropriativeparking
zonebesettingwhichisconnectedwithunderground
tunnel, accommodationroadshouldbeconstructedin
order to lower theeffects fromnon-motor vehiclein
traffic.
The interchange between bus transportation and
rail transit mainly includes theroadsideparkingpat-
tern, parallel pattern, vertical patternandcentralized
pattern. Bus accommodation road can be presented
andthebusstopshouldbebuilt near subway station
entrance-exitwhichcanprovidecomfortable, safeand
convenient environment intransit interchanging.
4 RESEARCHOF NON-MOTOR
VEHICLE-RAIL TRANSIT-TUBE
INTERCHANGINGTRANSPORT
SYSTEM PATTERN
Bicycleisoneof theeffectivemannersinpublictraffic.
Especiallyinchinaat present, inferior serviceinpub-
lictraffic, undevelopedmechanization. Bicycleshows
cheap, convenient and unpolluted characteristics, it
takesimportantstatusinourcountrypublictraffic.The
interchanging content mainly includes bicycle park
systeminstation, thefacility of thepassengerssepa-
ratesfromvehiclefacilityplannedandsuitableroutes
organizationdesignednear thestation.
Bicycle-rail transit-tubetransferringtransport sys-
tempatternsdesignshouldfollowprinciples:
1. Theinterchangepatternissuitableinthecitysuburb
or in residential district around rail transit station
(General indistributionstation), andit canbenot
appliedincitycenter.
271
2. In order to avoiding occupying of the space,
decreasetheeffectstothetransport, it shouldpro-
videenoughquantityif thebicyclespecial parking
spots.
3. Appropriative bicycle parking lots on the ground
shouldbeprovidednearthestationwhichpassenger
flowislarge, andpassengersmaytakeinterchange
throughtunnel;dispersedbicycleparkingzonemay
bedesignedaroundthestationwhichhasfewpas-
senger flow, yet is not so closed to the station
entrance-exitinordertoavoideffectingpassengers.
4. In park essential facilities must be supplied and
shouldarrangethespecialist tomanagewithinex-
pensivecharge.
5. To display bicycle superiority in short-distance
home-based-trips, and limits its proportion in
long-distancehome-based-trips. Reasonablebicy-
cle routes and accommodation road system can
reducethebicycle effects in traffic and provides
comfortable, safe and convenient environment in
transit interchanging.
In general distribution station, which interchange
with rail transit almost are public traffic (bus) and
non-motor vehicle (such as automobile, bicycle and
walksystem).Atpresentinmanystationsusual show:
bicyclesarelaidonthestreetorsometemporaryopen-
air non-motor vehicle parks spot all around, it has
occupied the limited station space. Simultaneously
thebicyclearound thestation may on thepassenger
way, hinder thepedestriantraffic, andpotential secu-
rityriskexistsalso.Althoughmanystationsperipheral
non-motor vehicleparkshavetakespecialists, accom-
modation tunnel which may connect rail transit and
bicycles are not been designed. It enlarges inter-
changingdistance, takes inconvenient andlowers the
interchangingefficiency.
Based on the interchange principle between non-
motor vehicleand rail transit introduced above, it is
proposed in this paper the Bicycle-rail transit-tube
transferringtransportsystempatternsconceptbyusing
ageneral undergroundsubwaystationasanexample.
For a general station, the length is about 200m,
thewidth is about 20m(3035mif including exit).
Four exits aredesigned on both sides of thestation,
andareconstructedinconjunctionwiththeventilation
shaft etcaffiliatesasshowninFigure1. Thestationis
situatedunder thecross, andfour exitsaresituatedat
four cornersof theroad. Inconstructionof thestation,
themainstructurewill beconstructedfirst, andthen
theexit. It canbeseenfromtheplanelayout that the
twoexits inthesamedirectionwouldbeconstructed
separately, andno connectionbetweenthetwo exits.
Theshadow areain themiddlehas no use. If it can
beconstructedat thesametimewiththeexits of the
station, thetwo exits andthemiddlepart couldform
theundergroundparkingbicycleetc. Bydoingso, the
land around the station can be used intensively, and
Figure1. Bicycle-rail transit-tubeinterchanging transport
systempatternsinGeneral distributionstationonthe1stflood
underground.
Figure2. Sketchmapof Bicycleparksunderground.
decreasethestationgroundareaoccupationareaand
passenger flow on the ground. The phenomenon of
scramblebetween bicycleand car will bedecreased,
and theinterchangeefficiency of non-motor vehicle
andrailwaywill behighlyincreased.
The underground non-motor vehicle is shown as
shadow area in fig. 1, the four corners of the road
couldbeset asbicyclesparkingexit (showninFig. 2)
whichmakesthepassengersfrombothdirectionscan
parktheirbicycleseffectivelyandinterchangewithrail
transit(interchangingmayusethemutual wall between
stationhall andbicycleparkingareaorusetunnel). For
considerationof parkingareaof bicycle, thisshadow
areaisabout onethirdof thestationmainarea(about
60m) withthewidthof 20m, areaof 1200m
2
(total
area 2400m
2
). The major design reference can be
planedbyBicycleParkingDesignStandard.
Theprinciplefactors consideredfor thescaleof a
normal park for bicyclesincludethenumber of bicy-
cles reaching to station, the area occupied by one
bicycle, and the piece of the bicycle park. Accord-
ingto thepiece, thescaleof thebicyclepark canbe
calculatedasfollowing:
272
Table3. Themaindesignindexinbicyclepark.
Parkingwidth Space Aislewidth
between
single double bicycles single double
Parkingtype (m) (m) (m) (m) (m)
Diagonal 30
1.00 1.60 0.50 1.20 2.0

45
1.40 2.26 0.50 1.20 2.0

60
1.70 2.77 0.50 1.50 2.0

Vertical 2.00 3.20 0.60 1.50 2.0
Unit area
Single Single Double Double
one-side two-sides one-side two-sides
Parkingtype m
2
m
2
m
2
m
2
Diagonal 30
2.20 2.00 2.00 1.80

45
1.84 1.70 1.65 1.51

60
1.85 1.73 1.67 1.55

Vertical 2.10 1.98 1.86 1.74
where: S
bic
denotes the needed scale for bicycles in
the station (m
2
); V
bic
is the number of bicycles dur-
ing rush hour piece/hour);
bic
is thevelocity of the
bicyclepark; s
bic
is theaveraged areaoccupied by a
bicycle(m
2
).
Thevelocitycanbesetas1duetolittlechangeof the
numberof bicyclesbecausetheusersof thebicyclesin
thestationarealmost onlytoworkor backfromwork
bybicycles. Theareaoccupiedbyabicycle, obtained
fromTable3,is1.7m
2
.Throughcalculating,thecapac-
ity of the underground bicycle park, excluding the
publicarea, is1200units.
Throughthesamplingsurvey of numbers of bicy-
clesnear theavailablestationsinShanghai rail transit
carried out by the author, the following point are
given: thenumber of personsusingbicyclesfor inter-
changing transportation, including the persons who
arbitrarily lay their bicycles at the edges of road or
near someshops, is about 800to1000. Accordingto
theaboveresults, itcanbeseenthatthecapacityof the
underground park areain anormal station can meet
therequirement of paringbicycles for transportation
interchanging.
For thestationof mainbodylayonthesideof road,
theundergroundpark areafor bicycles canbeset up
asshowninFigure3. For thistypeof parkingarea, the
ceiling can be mounted by transparent materials for
accessingof sunlight.Thevegetableprocessingonthe
topof thepark will improvetheinner environment in
an extent through theholebetween theparking area
andstationhall.
Figure 3. Sketch map of Bicycle-rail transit-tube inter-
changing transport systempatterns (station on one side of
road).
Takingthevarietyof theplanelayoutof therailway
transportation stations into accounting, the under-
ground parking area for bicycles combined railway
station, proposedinthispaper, isaconceptual pattern.
Theexact arrangement will beadjustedaccordingto
theactual planelayout of stations. Theprincipleidea
of themodel is toefficiently utilizetheareanear the
entrancesandexits, realizetheinterchangingbetween
bicyclesandrailwaytransportationwithoutextracost,
andincreasetheefficiencyof interchanging.
For theundergroundparkingareaof bicycles ina
station, theconstruction is feasiblein techniques. In
general, thestations whichneedthis typeof parking
areaareinthesub-center, outskirt or near thesatellite
city. Themainbuildingsnear thestationsareresiden-
tial districts also, theconstructionof thestationwill
almost not affected by theenvironment around. The
parkingareacanbeconstructedwiththemainstructure
of thestationby openexcavationmethodinfounda-
tion. Also, its feasible to firstly construct the main
structure of the station and then construct the park-
ing area combined the entrances and exits by open
excavationmethodor top-downmethod.
5 CONCLUSIONS
Throughtheanalysis above, it canbeconcludedthat
theundergroundbicycleparkingareafor transitinter-
changing is feasible in reality. The advantages as
below:
1. The users of bicycles can interchange rail transit
through the entrance between the under ground
parkingareaandstationconveniently, whichreal-
izetheinterchangewithout extracost, increasing
theefficiencyinagreatextent, providingthecom-
fortable environment in transit interchanging at
sametimeandexpressingtheconceptionof people
oriented.
273
2. Theundergroundinterchangebetweenbicycleand
railway transit decreases theoccupied areaof the
station, facilitatethetravelersusingother typesof
transportation which transfigure the environment
aroundthestation.
3. This patternof transit interchangingcandecrease
the flow of bicycles on the ground, increases the
efficiency of motor passing, eliminates the pos-
sibility of traffic accident and realizes thePedes-
trian SystemSeparated fromVehicle Systemby
passenger interchangingthroughtunnels.
Thecost of constructionfor theundergroundbicy-
cles parking area is higher than normal station.
Nevertheless, synthetically considering the Traffic
Diversionontheground, theeffect onthepassenger
flowbyother patternsof transportation, andtheenvi-
ronment near the station, the society and economic
benefit resulting fromwhat it brings will exceed far
fromwhat it costs.
REFERENCES
Ge, L. 2005. A Study on Planning & Design Methods for
UrbanPassengerTransit Hubs. Doctoral Dissertations In
Southeast University. Nanjing.
Gu, B.N. & Rao, X.P. 2001. On the Reduction of Subway
StationScale. Urban mass transit. 2001(3):1417.
Huang, J.Z. 2006. Modelsfor SuburbanizationLanduseand
Corresponding Transportation Development in Metro-
polis. Beijing: ChinaArchitecture& BuildingPress.
Yang, D.Y. & Han, H. 2000. A Study on Metropolitan Rail
TransitandTransportationStructure. Urban mass transit.
2000(4):1015.
Zheng,M.Y.2006.UrbanDevelopinLight-Rail Age.Beijing:
ChinaRailwayPublishingHouse.
274
Shotcreteexcavationsfor theMunichsubway Comparisonof
different methodsof facesupport insettlement sensitiveareas
J. Fillibeck& N. Vogt
Zentrum Geotechnik, Technische Universitt Mnchen, Mnchen, Germany
ABSTRACT: For theconstructionof shallowtunnelsinsettlement-sensitiveurbanareasit isvery important
to reducethesettlements andto increasethestability of thetunnel faceduringtheexcavation. Inthecaseof
shotcreteexcavation, theuseof different methodsof facesupport hasbecomemoreandmorecommon. These
methods are: groundfreezing, piperoofs, jet groutingandinjectionsupport. Thepaper shows theexperience
madetoduetheinstallationof theabovementionedfacesupports, especiallysincethespecificfocusisrelatedto
thearisingsettlements. If onlysmall deformationsareallowedtooccur, astheexamplesshow, deformationsthat
havetobeconsideredduringtheconstructionprocessaswell asthosetoestablishthebearingload; theycould
besignificant dependingontheprocess. Suggestionshavebeenmadeastowaysinwhichthedeformationscan
bereducedbymakingadditional measurements.
1 INTRODUCTION
For the construction of safe shallow tunnels in
settlement-sensitiveurbanareas, itisveryimportantto
reducethesettlementsandtoincreasethestability of
thetunnel faceduringtheexcavation. Theuseof dif-
ferentmethodsof facesupportinthecaseof shotcrete
excavationsisbecomingincreasinglycommon. These
methodsaregroundfreezing, piperoofs, jet grouting
andinjectionsupport.
The report presents the experience gained in the
installationof theabovementionedworkingfacesup-
ports, with particular focus on the induced settle-
ments. Four different projects of Munichs subway
are described. After a short project description the
results of the measurements are illustrated (geode-
tic andboreholemeasurements) andevaluated. With
this background, the different methods of face sup-
port are compared and the different advantages and
disadvantagesarediscussed. Finally, special attention
is focused on the installation process. Proposals are
madefor thereductionof settlementsinfuturetunnel
projects.
2 GEOLOGICAL ANDHYDROGEOLOGICAL
CONDITIONS
In theMunich subsoil thequaternary gravels follow
underfillingsof small thickness.Thequaternarygrav-
els can reach a thickness of more than 20m. They
predominantly consist of mediumdensity layers, are
laminatedandhave, dependingonthedeposit condi-
tionsandtheirage, adifferingamountof sandandfine
grain. The average permeability amounts to approx.
k=5 10
3
m/s. Tertiary layers liebelowthequater-
narygravels. Theyconsist of changinglayersof fine-
tomedium-grainedsandswithhighdensityandclays
orsiltsinstiff tofirmconsistency.Thethicknessof the
layerscanchangeexcessivelywithinasmall distance.
Theaveragepermeabilityof thesandamountsapprox-
imately fromk=1 10
4
to 1 10
5
m/s, thetertiary
clayandsiltcanfor all practical purposesbeassumed
impermeable.
The quaternary gravels possess a mostly free
phreatic water level, which can reach ground level.
Therearestill confinedaquiferswithinthesandlayers
withfine-grainedcover. Thepressureof theground-
water approximately corresponds to that of the free
phreaticsurfaceinthequaternarygravels.
3 HEADINGWITHGROUNDFREEZING
UNDERTHE CITY HALL OF MUNICH
3.1 Construction process
Theextentionof thestationMarienplatzof thesubway
linesU3/ U6under theMunichCityHall wasbuilt by
the company Fa. Max Bgl GmbH & Co KG. The
project wasfinishedin2006.
Parallel tothetwoexistingplatforms, twodirectly
joiningtunnels werebuilt inshotcretemethodunder
atmospheric conditions with a vertical distance of
275
Figure1. CrosssectionwithCityHall, tunnelsandgeologic
situation.
about 10mto the city hall. In order to avoid dam-
age to the landmark city hall, the deformations had
to bestrictly limited. Theconstruction company Fa.
Bgl,plannedfreezingarchesinthecontextof analter-
natebid, inorder tosupportthecrownandtokeepthe
retainingwater awayfromthetunnel face. Thefreez-
ing arches were provided for through pilot galleries
abovethecrown(Figure1).
The tunnels are embedded in the tertiary layers
(figure 1). The water bearing sand layers had to be
dewateredwiththehelpof filter wells.
3.2 Measurements to reduce frost heave
For thesuccessful realizationof thespecificproposal
it wascrucial toreducefrost heaveinsuchawaythat
no damageoccur to thecity hall. Frost heavecanbe
essentiallyattributedtotworeasons:
homogenous frost heave(Lh
vol
) becauseof a9%
increaseinvolumecausedbythechangeover from
water toice.
growing of ice lenses with corresponding frost
heave(Lh
icel.
) becauseof thetendency of thesoil
todrawwater near theinterfaceof thefrozentothe
unfrozensoil (zero-degree-front). Thisfrost heave
increaseswithtime.
Thefrost heavetests, whichwereperformedinthe
laboratory of theCentrefor Geotechnicsat theTech-
nical Universityof Munich(ZentrumGeotechnik, TU
Mnchen), showed, that in the tertiary fine-grained
soils frost heave Lh
icel.
still occurs at load-levels
of more than 400kN/m
2
if water can be drawn at
theinterfaceto thepermeablesandlayers. Therefore
the alternating layers of permeable sands and frost-
sensitiveclayspresentacritical risksourcewhenfrost
heaveisconsidered.
-40
-30
-20
-10
0
10
20
30
0 20 40 60 80 100
days
C
permanent
operating
permanent
operating
intermittent
operating
intermittent
operating
core freezing area
border freezing area
Figure 2. Operating Control of the artificial ground
freezing.
In order to reduce frost heave to a minimumthe
followingmeasurementshadtobetaken:
measuring and controlling thetemperaturein the
soil with the help of 5 measuring cross sections
per tunnel. Everycrosssectionincludes18thermo
couples.
reducingtheoperationtimeof thefrozenarchesby
dividingthetunnelsinto3differentsections: north,
middleandsouth.
further partitioningwithinthefreezingsectionsby
the installation of groups of freezing tubes with
separatecontrol.
In figure 2 the temperature development of a
sectionisshownschematically.Twentydaysafterstart-
ingthefreezingprocessinthecoreof thefreezingbody
thefreezingof theborderareastarted.Afterthefrozen
bodyreachedapprox. 22
Cinthecoreareaand0
C
in the border area, the freezing tubes were operated
intermittently, anaverageof 8to24hoursinthecore
areaand12to24hoursintheborderarea. Duetointer-
mittenthandling, thezero-degree-frontdoesnotmove
outside (enlargement of the frozen body), but stays
inanarrowzone, whichagainandagaingets frozen
anddefrosted. Thusfrost heavereducessignificantly.
Theoperation of onesection could bestopped after
approx. 90days. Theadjacentdefrostingprocesstook
about threemonths.
3.3 Measuring of the settlements
Thedeformationswhichoccurredduringtheconstruc-
tion process were measured by a geodetic precise
levelling systemon the surface and a closed water
levellingsysteminthe2ndbasement of thecityhall.
Theclosedwater levellingsystemconsistedof 10
measuring points with aresolution of 1/10mm. The
measuringresultscouldbecheckedonlineatall times.
Figure3 shows thelocation of themeasuring points
S03, S06andS09of theclosedwater levellingsystem
aswell asthedevelopmentof thesettlementandheave
276
Figure 3. Vertical displacements of the measuring points
S03, S06andS09.
dependingonthetime. Thethreemeasurement points
weresituatedinthesectionsnorth, middleandsouth.
Thesettlementsatthebeginningof thefreezingpro-
cessresult fromgroundwater drawdown. At theonset
of freezingtheexpectedfrostheavestarted. Itreached
amaximumvalueof 3to5mm.Thesettlementsdueto
thetunnellingprocessoccurredafter theheadinghad
passedthemeasuringpoints andthey still continued
afterthefreezingprocesswasstopped.Thesettlements
sloweddowncontinuouslyandstopped3monthslater
withamaximumsettlement of about 10to12mm.
Figure 3 clearly shows the temporary displace-
ment of the settlements according to the heading.
Thedrivereachedthemeasuringpointsindescending
order, resultinginthemreachingthemaximumheave
successively.
Themeasureddeformationswereapproximatelythe
sameas thecalculatedones. Thereby half of theset-
tlements couldbeattributedto dewateringmeasures,
whichleadto largeareasettlements andcorrespond-
ingly low differential settlements. Furthermore, no
settlement damages weredeterminedat thecity hall,
soitcanbeassumedthattheheadingwasverysuccess-
ful. It wasessential that for thesuccessof theproject,
larger frost heaveby icelenses could beavoided by
applyingtheabovementionedmeasures. Frost lenses
wouldotherwisehaveledtoasofteningof larger soil
areasandthereforetolargersettlementsandsettlement
differences.
4 J ET GROUTINGCOVER FORA LARGE
CROSSSECTIONFOR U3NORTHLOT 1
TheconsortiumEd. ZblinAG/Max Bgl GmbH &
CoKGcarriedout theconstructionof thesubwaylot
U3North 1inthenorthof Munich. Theworkswere
completedin2006.Theshotcreteheadingswithatotal
lengthof around1950mweredrivenwithandwithout
Figure4. HeadingW3andW4withjet groutingcover.
Figure5. Deformations in cross section MQ 8 after con-
structionof thejet groutingcover.
compressed air support and with several methods of
crownsupport whichwill beintroducedhereafter.
TheheadingsW3andW4withacrosssectionarea
of upto200m
2
beganfromastartingshaft withatop
heading. First watertight pits with thin slurry walls
wereproduced(figure4) inorder tolower theground-
water table. The safety of the excavation face was
increased by 13 jet grouting covers (total length of
about15.5meach, overlap4.3m) aswell asfurtherjet
groutingpilesinthefaceof thecrown.
The quaternary gravels were cut with suspension
(simplex method) atapressureof upto400baratthe
cone. At anytimeduringthemakingof ajet grouting
pipeacontrolledoutflowof thesuspensionisrequired,
assuring that the pressure does not lift the soil. The
top heading followed after the installation of the jet
grouting took place. The heading of the bench and
invert beganonfinishingall topheadings.
4.2 Heave during the installation of the
jet grouting cover
Figure5showsthedeformationinthecrosssectionat
adistanceof 50mawayfromthestartingshaftdirectly
after theinstallationof thejet groutingcover.
277
Figure 6. Required suspension backflow during jet
grouting.
TheheaveabovethecrownreachedinMQ8about
140mmandintotal amaximumof 250mm. At first,
theheavewas deemed uncritical becausetherewere
no buildings closeto thetunnel, however they result
particularly in soil strains in a narrow band directly
abovetheinterfacetothetertiarysoils. Becauseof the
proximity of the thin slurry wall to the jet grouting
cover, theheaveledtoacrack inthethinslurry wall,
makingthewall permeable.
Theheaveresultsfromthefact, that theoutflowof
thesuspensionintheannular spaceof thejetgrouting
piles,whicharefacedupwards,cannotbecontrolledin
asuitableway(figure6).Owingtothelackof backflow
inthelayeredsoil withstronglydifferingconductivity
theoverpressurespreadover agreater area, resulting
inarisingof thesoil abovethejet groutingcover.
The large heave could only be limited by reduc-
ingtheoverpressurethroughtheinstallationof further
boreholes from the ground level, resulting in high
costs.
With increasing soil cover the heave reduced
becauseof theincreasingload. However, evenunder
more than 12mof soil cover and the installation of
thejet groutingcover intertiary clays, theheavestill
amountedtoapproximately20mm.
For further projects, where only very few defor-
mations areallowedduringtheinstallationof thejet
groutingcover, sufficient attentionshouldbepaidto
thecontrol of thesuspensionbackflowduringgrout-
ing. This problemcould for exampleberesolved by
improving thetechniqueof thegrouting machineor
withthehelpof adoubletube, whichispulledalittle
ahead during jet grouting or likewise with the help
of special valves which control the pressure of the
backflow.
If thepressuregetstohighduringgrouting, heave
can be avoided or at least reduced by additional
horizontal or vertical arrangedboreholes.
Figure7. Comparisonof settlementsof different shotcrete
tunnels, drivenunder atmosphericconditions.
4.3 Settlements during the heading
After theinstallationof thejet groutingcover, thetop
headingwithtemporaryshotcreteinvertfollowedstep
bystep, overthewholeheadingdistance.Afterthisthe
headingof thebenchandinvert followed. Thesettle-
mentswhichoccurredduringtheheadingsamounted
toamaximumof 26mmincrosssectionMQ8and30
to40mmintheareawithlarger soil cover.
Inorder tobeabletojudgetheresults, infigure7
the above mentioned measurements are compared
withthoseof atmospheric shotcreteheadings having
nearly thesamesoil cover but weredriven in partial
faceadvancewithout jet groutingcover.
Overall maximumsettlements were measured as
almostthesamesize, whichmeansthatthejetgrouting
cover doesnot reducethesettlements, incomparison
totunnelsdriveninpartial faceadvance. Astheslid-
ingmicrometer measurementsshow, theforceswhich
weretaken fromthejet grouting cover, lead to con-
centrated high stresses in the small bedding area of
thejet groutingcover. Thisstressconcentrationleads
to comparatively highcompressions andsettlements.
On theother hand partial faceadvanceleads to less
stress concentration, however the different delayed
headings lead to multiple load rearrangements and
therefore the surface experiences approximately the
samesettlements.
Finallyitcanbeconcludedthatwiththejetgrouting
coverthesettlementsarenotreducedincomparisonto
thosecaused by partial faceadvances. However, the
facestabilityclearlyincreasesbyusingajet grouting
cover.
5 PIPE SCREENCOVER FORTHE
UNDERPINNINGOF A BUILDINGINU3
NORTHLOT 1
Thetwoshotcreteheadingsof sectionW1intheabove
described subway Lot U3 North-1 in Munich had a
cross sectional areaof A=41m
2
andweredrivenin
the tertiary soils under atmospheric conditions with
thehelpof wells, dewateringthetertiarysandlayers.
278
Figure8. CrossingthebuildingWerner-Friedmann-Bogen,
longitudinal section.
Figure9. Settlements of Werner-Friedmann-Bogenduring
tunnelling(crosssection).
In this section the underpinning of the Werner-
Friedmann-Bogen, abuildingcomplexwith12floors,
is of special interest. Thefoundation pressureof the
3mwidestripfoundation, whichliesinthecentreof
thebuildingandcarriesthemainloads, amountingto
nearly300kN/m
2
.
At a vertical distance of approx. 12m between
the foundation and the crown, a pipe screen cover
was planned as an additional measure of protection
(figure 8), because the tertiary soil cover amounted
only 4mandfull water pressurewas actingfromthe
quaternary to thesurfaceof thetertiary soils. At the
southwest side of the Werner-Friedmann-Bogen an
undergroundgaragefollows.
For every pipescreen38pipeswereinstalled. The
lengthof thepipesamountedto12mwithanoverlap
of 4m. Theborediameter amountedto146mmwith
a6mmannular space.
Infigure9thesettlement troughalongtheWerner-
Friedmann-Bogen is shown as dependent on the
development of theheading.
Figure10. Longitudinal sectionof thegroutingsectionO2.
Due to the dewatering of the tertiary sand layers
settlementsof 5mmto7mmwererecorded.Theinstal-
lation of the pipe screen and the forward directed
settlements of the heading of track 2 increased the
maximumsettlements to approx. 10mm. Thelargest
settlements resultedfromthe2headings. Finally the
maximumsettlementsamountedto25mm.
As a comparison with measuring of further cross
sectionswithoutpipescreenshows, themaximumset-
tlementsweremeasuredundertheWerner-Friedmann-
Bogen. It is clear that the foundation loads lead to
higher settlements and because of the smaller soil
cover only limited arching develops. Furthermore,
installationalsocausesettlements. However, it iscru-
cial to settlements, that the pipe screens as well as
thesurroundingsoil layersexperiencesomedeforma-
tion, before the systemcan carry the expected load
inboththelongitudinal andlateral directions. That is
why the predominant settlements respectively occur
shortlybeforeanddirectlyduringtheheading.
It canthereforebeconcluded, that thepipescreen
primarilyincreasesthesafetyof thetunnel face.Forthe
installationandformationof thebearingeffects,defor-
mationsarehowevernecessary, whichleadinthiscase
tosettlementsof 25mm. Pipescreensareonly appli-
cablefor thereductionof settlements, if substantially
higher settlementsareexpectedwithout them.
6 HEADINGWITHCOMPRESSEDAIR
SUPPORT ANDGROUTINGINU3NORTH
LOT 1
Thegeological conditionsinthesectionO2of thesub-
way lot U3 north-1 are shown in figure 10. In this
sectionashotcreteheadingwithcompressedair sup-
port was providedfor. If thethickness of thetertiary
soilsabovethecrownreachedlessthan1.5m,theover-
layingquaternarygravelsweregrouted.Inthegrouting
section1withalengthof approx.40mthegravelswere
groutedfromthesurface. Insection2thesurfacewas
not accessible. Thegravel wasthereforegroutedfrom
279
Figure11. Surfacesettlements dueto shotcretetunnelling
usingcompressedairinMunichwithdependenceonthepillar
ratioA/D.
thetunnel. At theendof theheadingthegroutingwas
doneagainfromthesurface.
The aimof the grouting was to reduce the per-
meability of thegravel to k5 10
5
m/s. This was
controlledbypermeabilitytestsintheborehole.
After thegroutingactivity theheadingfollowedin
shotcretemethodwithcompressedair support witha
maximumoverpressure of 0.7bar. Settlements were
measuredbetween3mmand11mm(without consid-
eratingof settlementsduetowater drainage). Inorder
toassessthisresult, theresultsof settlementmeasure-
ments of headings in Munich, with compressed air
support and without grouting (shield tunnelling and
shotcretemethod) depending on thepillar ratioA/D
arecompared with theabovementioned result of the
U3N1measurement infigure11.
It can be seen that the settlements measured if
groutingwasapplieddonot differ fromthosewithout
grouting. It appears that thegrouting did not reduce
thesettlements.
Overall, theresultsconfirm, thatshotcreteheadings
with compressed air support lead only to very small
settlements withsmall tangential inclinations, which
causenodamagetoconventional buildings.
Consideringthatthemeasuredcompressedair con-
sumption was almost just as small as the calculated
value, thevery extensivegroutingmeasure(21500m
grouting boreholes with more than 43000 grouting
sleeveshavebeeninstalled) canbedeterminedasvery
successful.
7 CONCLUSIONSANDFINAL REMARKS
In order to construct shallow tunnels in settlement-
sensitiveurbanareaswiththeshotcretemethod, mea-
sureshavetobetakentoincreasethefacestabilityand
toreducethesettlements.
Besidesthecommonmeasures(for examplereduc-
ingthelengthof theadvancestep, etc.) special crown
support measures areoften usedfor this purpose. In
thispaper, theexperiencesof theauthorsdemonstrated
thepurposeof usingafrozencover,apipescreencover,
ajet groutingcover andagroutingcover. It hasbeen
shown, that theinstallationof crownsupportingmea-
surescanhaveanextensiveinfluenceontheappearing
deformations
As the settlements caused by shotcrete headings
withconventional crosssections(approx. 40m
2
)inthe
Munichundergroundarecomparativelysmall (smaller
than 20mmto 25mmfor atmospheric headings and
smaller than 10 to 20mm for headings with com-
pressed air support), thecrown supporting measures
donothave,astheexamplesshow,decisiveadvantages
regarding the deformations. However, if the crown
supportingmeasuresareusedinaadequateway, acon-
siderablyhigher safetypotential occurs.Thishastobe
consideredif adecisionhastobemade, astowhether
or not crown supporting measures are necessary in
difficult sections.
280
Fracturingof sandincompensationgrouting
K. Gafar & K. Soga
University of Cambridge, Cambridge, UK
A. Bezuijen, M.P.M. Sanders&A.F. vanTol
TU-Delft/GeoDelft, Delft, The Netherlands
ABSTRACT: Thephenomenonof fracturinginsandasaresultof compensationgroutingwasstudied.Processes
of fractureinitiationandpropagationwereexplainedandaparametricstudywasconductedinordertoinvestigate
thefactorsthat causesandfracturingtooccur. Experimental resultsindicatethat fractureinitiationrequiresthe
existenceof alocal inhomogeneityaroundtheinjectionposition. Grout mixtureintermsof water-cement ratio
and fines content had major roles in sand fracturing, whereas injection ratehad aminor influenceunder the
testedconditions.
1 INTRODUCTION
Compensationgroutinghasbeenwidelyinusetocon-
trol ground settlements during tunneling processes.
Nevertheless, its use is still hindered in many cases
bytheuncertaintiesinthegrout mechanical behavior.
As thecurrent grouting practiceis highly dependant
on field experiencerather than scientific knowledge
of soil-grout interactionbehavior, issuessuchassuit-
able injection pressure, soil fracturing and bleeding
(amountof water forcedoutof thegroutmixture) still
needfurther investigation.
In particular, soil fracturing stands out as one of
the major challenges that could affect the results
of a grouting project. Accidentally-created fractures
could cause considerable damage to near-by struc-
tures, whereasfailingtocreatefractureswhentheyare
requiredcouldresultintunneling-inducedsettlements
(for example) not beingfullycompensatedfor.
Whilefracturingof cohesivesoils was extensively
studiedbymanyresearchers(e.g. J aworski etal. 1981,
Mori & Tamura 1987, Andersen et al. 1994, Chin
& Bolton 1999, Soga et al. 2005 & 2006), there
has been limited work for compensation grouting in
cohesionless materials (e.g. Chang2004). Fracturing
of sand was studied in relation to the oil industry
(Khodaverdian&McElfresh2000, Bohloli &dePater
2006) and horizontal directional drilling (Bezuijen
et al. 2002). Recently, the prospected use of frac-
ture grouting in Amsterdam to tackle settlements
encountered during the construction of the North-
South Metro line triggered a thorough research into
thephenomenonof sandfracturing.
This paper follows on fromthework reported by
Sanders (2007) andBezuijen& vanTol (2007). Uti-
lizing the reported optimumgrout mixture for soil
fracturing, aseriesof laboratoryscalegrout injection
testswasperformedinwhichvariousfactorsaffecting
fracturingof sandwerestudied.
2 THEORY & BACKGROUND
Hydraulicfracturingisdefinedastheconditionlead-
ingtothecreationandpropagationof athinphysical
separation in a soil or rock mass due to high fluid
pressures. Thefracturingprocessismainlycharacter-
izedbyfractureinitiation,propagationandorientation.
Understanding the factors controlling these parame-
tersisthefirststepinunderstandingandpredictingthe
grout fracturingbehaviour. Thefactorsaffectingfrac-
turingphenomenoncouldbedividedintotwogroups:
factorsrelatedtosoil, suchassoil properties(particle
size, shapeanddistribution, relativedensity, cohesion,
friction angleetc), stress stateand themagnitudeof
confining pressureand factors related to grout itself
andthegroutingprocess,suchasgroutrheology(com-
ponents and viscosity), injection rate and injection
pressure.
2.1 Fracture initiation
The two main theories explaining fracture initiation
are tensile failure and shear failure. J aworski et al.
(1981)suggestedthat,forahydraulicfracturetooccur,
theeffectivestresshastobecometensileandequal in
281
Figure 1. Realistic particle arrangement around an injec-
tionholeandpossibledeformationmodes. P
f
istheinjection
pressureto causeplastic deformations insoil and
inthe
effectivestress around theinjection hole. (Bezuijen & van
Tol 2007).
magnitudetothetensilestrengthof thesoil. Thissitu-
ationisclearlyunconceivableincaseof cohesionless
soilslikesand. Theother theoryonfractureinitiation
proposes ashear failureas themainreasonfor frac-
turinginclays. Mori &Tamura(1987) suggestedthat
shear failure occurs within a short duration under a
highinjectionrate, withthedurationbeingtoo short
for thegrout topenetrateintomicrofissurestocreate
wedgeactionand/or enter thesoil porestoweakenthe
soil strength.
For fracturing in sand, Bezuijen et al. (2007) and
Bezuijen & van Tol (2007) suggested that the local
contact forces between sand grains haveto beelim-
inated. As in case of cohesionless soils there is no
tensile strength between the grains, this means that
thefracturingpressurehas to overcometheeffective
stress in the direction perpendicular to the fracture.
In reality, sand is never perfectly homogeneous and
therefore, thearrangement of particles aroundacer-
tainboundary(injectionhole, for example) will bein
such away that someparticles arein closer contact
thanothers, asshowninFigure1.
When injection is conducted, grout will fill the
spacebetweenparticlesandstart topushthemapart.
Whetherthisinitiatedfracturewill propagatefurtheror
theresultwill bearoughlysymmetriccavityexpansion
dependsmainlyonthepropertiesof theinjectedgrout.
This explanationof fractureinitiationinsandagrees
with the findings of Thallak (1991), who, based on
themicromechanicsof granular media, suggestedthat
hydraulicfractureinitiationrequiresthelocal contact
forcesbetweenparticlestobecomezeroor tensile. As
this requirement is complicatedby local forcedistri-
butionat microscopicscale(whichcanleadtoawide
variationinthefracturingpressure), it wasconcluded
thathydraulicfractureinitiationdependsmainlyonthe
local microscopicinhomogeneitiesinthesoil.
2.2 Fracture propagation
Starting from the condition shown in Figure 1, a
roughlysymmetriccavityexpansionwill alwaysbethe
Figure2. Leak-off andfiltercakeformationassociatedwith
fracturing.
resultif theinjectionpressureishighenoughandclose
toperfect cavityexpansionpressure. Factorsthat dic-
tatehowhightheinjectionpressurewill bearemainly
thegrout materials and rheology, confining pressure
and stress stateand soil density. Results reported by
Kleinlugtenbelt(2005), Gafar&Soga(2006), Sanders
(2007) and Bezuijen & vanTol (2007) confirmthat
highinjectionpressuresareassociatedwithhighgrout
viscosities(loww/c ratioor morecement content for
cement-basedgrouts). Gafar & Soga(2006) reported
that injectionpressureinthecaseof no confinement
was increased by 15 times when a100kPaconfine-
ment was introduced. Bezuijen & van Tol (2008)
highlightedtheinfluenceof thestress state. Unload-
ingof thesoil aroundacavityasaresultof equipment
installation leads to plastic deformations in thesoil,
whichmayresult inlower injectionpressures.
Bezuijen&vanTol (2007) explainedthat, for frac-
tures to propagate, grout mix has to contain enough
content of fine bentonite particles and has enough
water to ensure good flowability at the same time.
Inthiscase, pressureapplicationwill causethewater
inadjacent sandpores to bereplacedwithamixture
of water and finer particles leaking fromthe grout
mix, as showninFigure2. This permeationactionis
termed as leak-off. With introduction of fines in the
soil matrix, theleak-off will reducethepermeabilityof
sandaroundtheinjectionhole, causinggroutbleeding
tobesloweddown.
On the sides of the propagating fracture, a filter
cakeisformedasaresultof theaccumulationof larger
cement particles filteredat thesand-grout boundary.
Filter cake formation is crucial for the propagating
fractureto beableto keep itself open and to sustain
forcesinducedbygroutpenetration.Atthesametime,
bleeding is restricted, keeping agood workability of
the grout. A suitably high initial water-cement ratio
(w/c ratio) will ensure that the grout mix will have
282
Figure 3. Schematic diagramof the experimental setup,
showingthedimensionsinmm(Bezuijen& vanTol 2007).
sufficientwatercontenttokeepafracturepropagating
without the need for very high pressures. Eventu-
ally, the pressure at fracture tip will not be enough
toovercometheeffectivepressureinthedirectionper-
pendiculartofracture. Nonetheless, thepressureatthe
tipwill still behighenoughfor leak-off andbleeding
tocontinue. Thiswill causeafilter caketobeformed
at thefracturetip, completely blockingfurther grout
propagation.
3 EXPERIMENTAL SETUP
Grout injectiontests wereconductedinacylindrical
steel container of a900mmdiameter andchangeable
height. Two sample heights were used: 840mm(for
Tests1and2)and600mm(fortherestof thetests).Fig-
ure3showsaschematicdiagramof theexperimental
setup.Theinjectiontubepositionwasfixedat360mm
abovethebottomof thecontainer. A PVC platerests
onthetopof thesaturatedsandsample, tightlysealing
it off froman upper water chamber. Confinement is
appliedby means of pressurizingthis water chamber
usingcompressedair.Airpressureisappliedthrougha
glasscylinderthatalsoshowsthechangeinwaterlevel
resultingfromsoil heave. Thischangeiscontinuously
measured during the test by means of a differential
pressure gauge. Two tubes connect the top (through
thewater chamber) andthebottomof thesoil sample
toanother graduatedglasscylinder whichrestsonthe
topof thesetup, providingadoubledrainagesystem.
A simplifiedmodel of thetubeManchette(TAM),
asshowninFigure4, runsacrossthediameter of the
cylindrical container.Thetubehasaninternal diameter
of 22mm,with4equallyspaced7mmholesatthecen-
tre. A rubber sleevecoverstheholesand2rings, one
oneither side, preventtheinjectedgroutfromflowing
alongthetube.
Figure 4. Simplified model of Tube Manchette (TAM)
used for injection. Actual space between the two rings is
40mmduringinjection(Bezuijenet al. 2007).
Figure5. Locationof instrumentswithrespecttotheinjec-
tiontube(inj).Paretheporepressuretransducers,Vmeasures
the vertical pressure and H the horizontal (Bezuijen et al.
2007).
Changes in pore water pressure during injection
were monitored by four pore pressure transducers,
distributedaroundtheinjectionpointasshowninFig-
ure5. Thereadings fromthesetransducers werenot
usedinthis paper. Twototal stress cells wereusedto
recordthechangeinhorizontal andvertical pressures.
Two types of sand were used: Baskarp sand
(d
50
=130m) and Leighton Buzzard type D sand
(d
50
=234m). Inbothcases, sandwaswet-pluviated
intowaterinthemodel container. Loosesandwasthen
densifiedtorequiredrelativedensity (70%) by drop-
pingthewholecontainer over 25mmas many times
asrequired.
In order to raise K
0
to a closer value to 1, sand
was pre-stressed by applying aconfining pressure
of 300kPaatthebeginning. Confinementwasreduced
to100kPaprior togrout injection. Thispre-stressing
283
Table1. Summaryof conductedexperiments.
W/C Rate Bentonite Materials/
No ratio (l/m) (%) sand Remarks
1 5.0 10.0 7.0 GD/Ba Repeatability
test
2 5.0 2.0 7.0 GD/Ba Slower inj.
rate
4 5.0 10.0 7.0 GD/Ba Wall friction
effect
5 5.0 10.0 7.0 Ca/Ba Effect of
materials
6 1.0 2.0 4.0 Ca/LB Lower w/c
ratio
7 1.0 10.0 4.0 Ca/LB Faster inj. rate
Notes: Rate=injection rate, GD=GeoDelft cement
and bentonite, Ca=Cambridge cement and bentonite,
Ba=Baskarpsand, LB=LeightonBuzzardsand. Bentonite
percentageisbyweight of mixingwater.
wasonlypartiallysuccessful andvaluesof startingK
0
werestill lessthan, but closeto, 1.
Grout injection was conducted using a plunger
pump. A bladder (not shown in Fig. 3) was used as
aninterfacebetweenpumpedwater andinjectedgrout
inordertoavoiddamagingthepumpbythegrout.The
injectionpumpwas capableof reachingamaximum
pressureof 4MPa. Injectedgrout was allowedto set
for24hoursbeforethesandwasdugoutandtheshape
of hardenedgrout wasphotographed.
4 RESULTS& DISCUSSION
Using the same injection setup, Sanders (2007) and
Bezuijen & van Tol (2007) reported that the best
grouting efficiency was attained by using acement-
bentonitegroutof aw/cratioof 5.0tofractureBaskarp
sand. OrdinaryPortlandcementandsodium-activated
bentonite(7%by weight of mixing water) wereuti-
lized. Injectionwasmadeunder aninjectionrateof 10
litersper minute(l/m).
Thecurrent seriesof testsadoptedtheabovemen-
tioned test as a reference and Table 1 summarizes
some of the experiments carried out. Two sets of
grout materials wereused: (a) rapidhardening, ordi-
naryPortlandcement andsodium-activatedbentonite
(GeoDelft, theNetherlands), and(b) normal harden-
ing ordinary Portland cement and sodiumbentonite
(University of Cambridge, UK). Injection pressures
resultingfromall thetestsareshowninFigure6.
4.1 Repeatability check
Duetoaproblemwithsamplepreparationthatyielded
asamplewhich was not perfectly homogeneous, the
repeatabilitytest(Test1, usinggroutwithw/cratioof
5.0) resultedinasinglefracturewhichpropagatedto
Figure6. Changeof injectionpressurewithinjectedvolume
for different tests.
Figure7. Fracturesfrom(a) referencetest(left) and(b)Test
2withslower injectionrate(right).
near thecontainer wall. Nevertheless, theinitial pres-
surewassimilar inmagnitudetothestartingpressure
of the reference test (2.7 and 2.8MPa respectively).
Thepressureduringinjectionwasabout40%lessthan
theinjectionpressureduringthereferencetest, which
highlightstheeffect of soil inhomogeneity.
4.2 Effect of injection rate
ReducingtheinjectionrateinTest2byafactorof 5still
yielded fracturing of thesand model. Thegrout mix
containedenoughwater andfineparticles. Recorded
injectionpressurewasabout10%lower thanthevalue
for faster injectionrate. Sectioningof hardenedgrout
revealedthat thicker fractureswereformed, withnar-
rower leak-off zoneandthicker filter cake, as shown
inFigure7b.
Under theslower injectionrate, thereismoretime
for thefilter caketo develop. Accordingto Bezuijen
etal. (2007), thethicknessof filtercakeincreaseswith
thesquarerootof thetimethatthegroutispressurized.
For a given injection pressure and injected volume,
reducingtheinjectionratebyafactor of 5will leadto
approximately 2times thicker filter cake. Formation
of athicker filter cakewill hamper further leak-off, as
thefiner particlesareblockedbythefilter cakethatis
alreadyformed.
284
Figure 8. Comparison of the change in vertical pressure
withinjectedvolumebetweenthereferencetestandthetests
withreducedsampleheight.
4.3 Side wall friction
Thesampleheightfortherestof thetestswas600mm.
Test 4wasarepetitionof thereferencetest under the
newsampleheight. Resultsshowedthat theinjection
pressure under the new testing condition was about
1MPalower, eventhoughtheinjectedgrout managed
to fracturethesand in asimilar way. Thechangein
overburdenpressurethat correspondstochangingthe
height of sandabovetheinjectionpoint isonly afew
kiloPascals.Therefore, itcouldnothavebeentherea-
sonfor thisreductionininjectionpressure. However,
measurements of total vertical stress showedthat the
frictionbetweenthesandandthewallsof thecontainer
havereducedthevertical stress at theinjectionlevel
fromthatappliedatthetopof thesample(Fig. 8).This
explainsthereductioninrecordedinjectionpressures
forall thetestsconductedunderthenewsampleheight.
4.4 Grout materials
TheordinaryPortlandcementusedintheexperiments
was CEM I cement and either sodiumbentonite or
sodium-activated bentonite was mixed. Bruce et al.
(1997) reportedthat sodiumbentoniteisthebest type
of bentonites to be added to cement grouts. This is
mainlyattributedtoitsswellingpotential, asit swells
upto 18times its original volume. Sodium-activated
bentonite, ontheother hand, couldswell upto 10to
15times.
Results showed that there was no significant dif-
ferenceresultingfromusingsodiumbentoniteinstead
of sodium-activated bentonite in terms of fracturing
sand. Comparing the injection pressures of tests 4
and5(Fig. 6), theonlydifferenceistheslightlyslower
build-upof pressureincaseof sodiumbentonite. This
shouldhaveresultedfromaproblemwiththeinjection
systematthebeginningof injection,asalmostnoheave
or drainagewas recorded over thedelay period. The
hardening speed of the used Portland cement (rapid
Figure9. Dehydrated layer around theboundary of hard-
enedgrout for (a) Test 6(left) and(b) Test 7(right).
forGeoDelftandnormal forCambridge) didnotaffect
theresults, asmostof theprocessesthatinfluencesand
fracturinghappenwell beforehardening. Interms of
groutingefficiency(definedasheavedvolumedivided
byinjectedvolume),bothtypesof bentonitegavemore
or thelessthesameefficiency.
4.5 No-fracture tests
Two injection tests were conducted using a low w/c
ratio grout (w/c ratio of 1.0, 4% sodiumbentonite
added; seeTests 6 and 7 inTable1). Thetests were
carriedoutinsandmodelsof typeDLeightonBuzzard
sand(d
50
=234m).
Injection under both slow (Test 6) and fast injec-
tionrates(Test 7) yieldednofracturing. Theinjection
pressures werevery closeto each other, but slightly
higher thanthevaluesfor fracturingexperiments. The
slowerinjectionrategavemoreuniformshapeof hard-
ened grout (Fig. 9a), whereas some fingering was
observedfor thefaster injectionrate(Fig. 9b). Inboth
cases,sectioningthehardenedgroutrevealedalayerof
dehydratedmaterial aroundthegrout-soil boundary.
With higher cement content and less water in the
grout mix, no leak-off occurred and there was lim-
itedamount of freebentonitetodevelopafilter cake.
Bleedingdidoccur andthisinturnreducedthegrout
mobility. It ispossiblethat thecalciuminthecement
changesthecoagulationstructureof thebentoniteby
cation exchange, increasing the permeability of the
consolidatedgroutandacceleratingbleeding(Sanders
2007). Most of thegrout stayedaroundtheinjection
point.Theexistenceof adehydratedlayeratthegrout-
soil boundariessuggeststhatmorebleedinghappened
aroundtheboundaries, whichisinagreementwiththe
theorysuggestedbyMcKinley& Bolton(1999). The
faster injectionrateallowedlesstimefor bleedingand
hence, the grout managed to create some fingering
beforeit becametooviscoustoflow.
Consolidationof thegroutleadstothepossiblelocal
irregularitiesat theboundariesof aninjectionholeto
befilledup,orplastered,inthewayshowninFigure10.
285
Figure 10. Influence of plastering on fracture initiation
(Bezuijen& vanTol 2007).
Such plastering will hamper fractureinitiation, as it
prevents thefluid pressurefrompenetrating into the
space between sand particles. As the created plaster
will haveacertainstrength, part of theinjectionpres-
sure will be acting on the plaster rather than being
usedtopushparticlesawayfromeachother toinitiate
afracture.
5 CONCLUSIONS
The experimental work conducted confirmed that
fractureinitiationinsandrequiressomelocal inhomo-
geneityaroundtheinjectionpoint, rapiddevelopment
of afilter cakewith alimited thickness and agrout
withlowviscosityandalimitedyieldstress. Whether
theinitiated fractures will propagateor not depends
mainlyonthegroutmixture. Water-cement(w/c) ratio
andfineparticlescontent playamajor roleinfractur-
ingof sand. Grouts withhighw/c ratios andenough
fineswill exhibitaleak-off of fineparticles, accompa-
niedwiththeformationof afilter cake, whichresults
infracturing. For groutswithloww/cratiosandlarge
grout permeability, bleedingisthedominant process,
leadingtonon-fracturingof sand.
For asuitablegrout mixture, faster injectionrates
will result in thinner fractures, whereas slower rates
give thicker fractures with less leak-off and thicker
filter cake. If thew/cratioistoolow, nofractureswill
beformed, regardlessof theinjectionrate.
The longer fractures experienced under relatively
lowinjectionpressuresinthefieldaremainly dueto
thenatural inhomogeneityof sandlayers. Injectionin
almost perfectly homogeneoussandinthelaboratory
gives shorter fractures and requires higher injection
pressures.
REFERENCES
Andersen, K.H., Rawlings, C.G., Lunne, T.A. & Trond, H.
1994. Estimation of hydraulic fracture pressure in clay.
Journal of Canadian Geotechnical 31: 817828.
Bezuijen, A., PruiksmaJ.P. & Pater C.J.de2002. Maximum
pressures in tunnelling limited by hydraulic fractures.
Proc. International Tunnelling Association Conference,
Amsterdam.
Bezuijen, A., Sanders, M.P.M., Hamer, D. & Tol, A.F.van
2007. Laboratory tests on compensation grouting, the
influenceof groutbleeding. Proc. WorldTunnel Congress,
Prague.
Bezuijen, A. &Tol, A.F.van2007. Compensationgroutingin
sand, fracturesandcompaction. Proc. XIVEuropean Con-
ference on Soil Mechanics & Geotechnical Engineering,
Madrid.
Bezuijen, A. & Tol, A.F.van 2008. Mechanisms that deter-
minebetweenfracturegroutingandcompactiongrouting
insand. Proc. 6th International Symposiumon Geotechni-
cal. Aspects of Underground Construction in Soft Ground,
Shanghai.
Bohloli B. & Pater C.J.de 2006, Experimental study on
hydraulic fracturingof soft rocks: Influenceof fluidrhe-
ologyandconfiningstress. Journal of Petroleum Science
and Engineering 53(12): 112.
Bruce, D.A., Littlejohn, G.S. &Naudts, A.M.C. 1997. Mate-
rials for ground treatment- A practitioners guide. Proc.
Grouting: compaction, remediation & testing, Logan.
Chang, H. 2004. Hydraulic fractures in particulate materials.
PhD. thesis. GeorgiaInstituteof Technology.
Chin, C.Y. & Bolton, M.D. 1999. Factors influencing
hydrofracture in clay. Proc. 13th ASCE Engineering
Mechanics Conference, Baltimore.
Gafar, K. & Soga, K. 2006. Fundamental investigation of
soil-grout interaction in sandy soils. Report. University
of Cambridge.
J aworski, G.W., Seed, H.B&Duncan, J.M. 1981. Laboratory
studyof hydraulicfracturing. Journal of the Geotechnical
Engineering Division 107(6): 713732.
Khodaverdian, M.M. & McElfresh, P.M. 2000. Hydraulic
fracturestimulation in poorly consolidated sand: mech-
anismandconsequences. SPE Annual Technical Confer-
ence, Dallas.
Kleinlugtenbelt,R.2005.Compensationgrouting, laboratory
tests in sand. MScthesis. DelftUniversityof Technology.
McKinley J.D. and Bolton M.D. 1999. A geotechnical
description of fresh cement grout Filtration and consoli-
dation behaviour. Magazineof ConcreteResearch51(5):
295307.
Mori, A. & Tamura, M. 1987. Hydro-fracturing pressure
of cohesive soil. Journal of the soil and foundations,
Japanese Society of Soil Mechanics & Foundation Engi-
neering 27(1): 1422.
Sanders, M.P.M 2007. Hydraulic fracture grouting, labo-
ratory tests in sand. MSc thesis. Delft University of
Technology.
Soga, K., Gafar, K.O., Ng, M.Y.A. &Au, S.K.A. 2006. Macro
and micro behaviour of soil fracturing. Proc. Interna-
tional Symposium on Geomechanics and Geotechnics of
Particulate Media, Yamaguchi.
Soga, K., Ng, M.Y.A. & Gafar, K. 2005. Soil fractur-
ing in grouting. Proc. 11th International Conference of
the International Association of Computer Methods and
Advances in Geomechanics, Tornio.
Thallak, S. 1991. Numerical simulation of hydraulic fractur-
ing in granular media. Phdthesis. Universityof Waterloo.
286
Historical casesanduseof horizontal jet groutingsolutionswith360
distributionandfrontal septumtoconsolidateveryweakandsaturatedsoils
G. Guatteri
Novatecna Consolidaes e Construes S.A So Paulo, Brazil Terrajato Lisbon, Portugal
A. Koshima, R. Lopes&A. Ravaglia
Novatecna Consolidaes e Construes S.A So Paulo, Brazil
M.R. Pieroni
Novatecna Consolidaes e Construes S.A So Paulo, Brazil Terrajato Lisbon, Portugal
ABSTRACT: Horizontal jet groutingtechnology has provedto bequiteaversatiletool for dealingwiththe
particular geological conditionsencounteredwhenexcavatingtunnelsinsoil mass, especiallyinthepresenceof
water flowandhighhydraulicgradient. Wedescribethestate-of-the-art applicationknownas360distribution,
inwhichhorizontal jet groutingcolumnsareexecutedaroundtheexcavatedsection, includingtheinvert, andat
thefar endof theconical treatment, tocreateawatertight chamber. Thisresults. inaheadingthat isconstantly
protectedbypre-consolidatedsoil andminimizestheeffectsof excavationonnearbystructures.Wealsodescribe
aspecial tool (calledthepreventer) designedtocontrol drillingandinjectionfluidoutflow, whichprovedtobe
absolutely essential tosuccessful outcomes. Wealsopresent theexperiencegainedby theauthorsinanumber
of projectsinBrazil andVenezuelaandtheresultsof afull scaletest, executedfor thefirst timeinEurope.
1 INTRODUCTION
Horizontal jet groutinghas provedto beanefficient
and versatile technical solution and is increasingly
beingusedfortunnellingindifficultsoils,usuallywith
highgranulometry, presenceof intensewaterflowand
highhydraulicgradient.
Further developments have facilitated more non-
TBM excavations in locations with low overburden
and superstructure presence, such as urban areas
wherestableandsafeconditionsduringexcavationare
requiredatall times, withminimumimpactonnearby
structures.
Onetypeof applicationthat is not yet incommon
useistheoneinwhichhorizontal jetgroutingcolumns
are executed all around the tunnel cavity in what is
calleda360
distribution(roof, sidewallsandinvert)
andinwhich, dependingonthecircumstances, consol-
idation is extended to formahorizontal soil-cement
frontal wall (septum) at the far end of the conical
treatment.
Therefore, for theheadingof thetunnel, asequence
of watertight chambersarecreatedsothat theexcava-
tion activity is always protected by pre-consolidated
soil.
We provide a brief description of this innovative
technical solution, itsmaingeometrical characteristics
anduseconditions.
Fromtheexecutivepoint of view, wedescribethe
useof aspecial tool namedthePreventer, whichwas
developedtocontrol theoutflowof drillingandinjec-
tionfluids,andisabsolutelyessential inordertoobtain
goodperformance.
Finally practical cases, executed by the authors,
aredescribed, along with theresults of thefirst full
scaletest of a360
jet groutingtreatment executedin

Europe.
2 360
DESIGN
Inrecentyears, horizontal jetgroutingtechnologyhas
frequently been used for tunnels wheresafety was a
priorityandmoretraditional techniqueswerenot fea-
sible or might not guarantee good performance. As
readers will know, this applies mainly to consolidat-
ingsoil masswithpoorgeo-mechanical characteristics
which do not allow excavation in safe conditions,
mainlyinsandy, siltyor claylysoilsor acombination
of these, withor without presenceof water.
287
Figure1. Typical distributionforroof andlateral protection
andforepolednucleus.
Figure2. Typical distributionof afull face360
distribu-
tion.
This technology is normally associated with the
NATM . Inatypical configuration, columns areexe-
cuted side by side to forma pre-consolidated arch
which follows the external profile of the excavated
section.
Dependingontheactual soil conditions, theconsol-
idationeffectmaybeimprovedbyjetgroutedcolumns
executed inside the section to reinforce the tunnels
nucleus, or by means of horizontal drains drilled at
thebottomor aroundthesection. Inmoreseveresitu-
ations, thejet groutedcolumnsmay bereinforcedby
meansof steel pipesor fiberglass, anddrainsmay be
equippedwithvacuumsystems. Fig. 1showsatypical
traditional distribution.
Theencouragingresultsobtainedfromthehorizon-
tal jet grouting techniqueand thegrowing needs for
executingtunnelsinincreasinglydifficult conditions,
atshallowerdepths, generallyinalluvial orevenresid-
ual depositsbutwithhighgranulometry, intensewater
flow, hydraulicgradientandthepresenceof important
superstructures, asinmany urbanprojects, prompted
thedevelopment of the360
distribution.
Themaincharacteristicsof the360
solution, com-
paredtothetraditional, are: a) soil consolidation, pre-
viouslylimitedtoroof andsidewalls, isnowextended
totheinvert, andb) theexecutionof ajetgroutedplug
at thebottomof theconical treatment.
Thusasealedchamber canbecreated, andinterms
of stability, soil-massloadsaroundthecavityareredis-
tributedto prevent possibleinflows of material from
Figure3. Exampleof 360
distributionfor half-sectionand
benchheading.
Figure4. 3DAnalysisof a360
distribution.
the excavation face and the invert, which are quite
commoninsandysoils.
Fromthehydraulicpointof view, thedrasticreduc-
tion of water inflow during the excavation phase
reducessoil-massdewateringandthereforekeepsset-
tlementundercontrol andminimizeseffectsonnearby
structureswhiletheheadingisexecuted. Fig. 2shows
atypical exampleof a360degrees distributionwith
full-faceexcavation.
Dependingonthetunnel dimensions, thesolution
mayalsobeadjustedfor usewithahalf-sectionhead-
ing. In this case, the bench or final invert may be
executed, if required by the geological conditions,
with the help of vertical or inclined jet grouting to
ensuresaferconditionsandoptimizeworkingschedule
(fig.3).
As far as theoperational aspect is concerned, the
360
solution became feasible due to advances in

drilling equipment, involving redesign and improve-
mentof itsmaneuverability,setupandmastalignment.
Special homothetic templates with the projection of
theconsolidation elements for theperfect alignment
of thedrillingstringhadto bedevelopedandimple-
mented.
Also, 3D graphic tools were introduced in order
toverify theintegrity andcontinuity of thetreatment
designed(fig.4).
Special considerationmust begivento thedevice
used to control the outflow of drilling and injection
fluids,whichevolvedfromasimilaroneusedintheoil
industry,hencethenamepreventer.Properuseof this
device proved to be essential to achieve appropriate
288
Figure5. Preventer typical configuration.
monitoringandcontrol of movementof excavatedsoil
mass.
3 THE PREVENTER CONTROL VALVE
Thepreventervalve(orblow-uppreventer)isused
to avoid gas leaks and explosions when drilling oil
wells.
Preventer valves andretainers at themouthof the
holecontrol outflows of solidandliquidspoils from
drillingwhengroundissandyandsusceptibletopip-
ing. They also facilitate control of slurry (excess
soil-cement mixture) during jet grouting jobs, thus
avoidingpiping alongtheholeor thecolumn.
Notethat adrillingoperationinnon-cohesivesoils
belowthewater tableproduceswithdrawal of material
duetotheintenseflowestablishedby thewater table
hydraulic gradient and water needed for the drilling
operationitself.
Insuchcase, itquitecommontolosecontrol of the
volumeof material flowingoutof theholeif thereisnot
proper perceptionof thephenomenon, thustriggering
apipingeffect.
This piping occurs when pressure inside the hole
duringthedrillingoperationislessthantheeffective
hydrostatic pressureactinginto theground. Andthis
canoccur alsoduringthejet groutingphase.
In this situation, with the preventer installed at
the hole mouth, we may control pressure inside the
hole, thusavoidingdecompressionof thesurrounding
material leadingtoanincreaseof permeability.
Thisdevice(fig.5) hasasystemof valvesandseals
tomaintainstrictcontrol of volumesof materialswith-
drawn during drilling or injection. At theend of the
jetting operation, it can also be sealed for the time
neededfor thesoil-cement mixturetoset.
Figure6. Tunnel facewithstartingdrillingpoints.
Figure7. Septumandinstallationof yieldingarch.
4 HISTORICAL CASES
4.1 Copacabana subway tunnel (Rio de Janeiro
Brazil)
Rio de J aneiros subway company planned a 750m
extensiontoline1, fromCardinal ArcoverdeStationto
thecenterof Copacabana, alsobuildingSiqueiraCam-
pos Station and turn-offs or maneuvering areas. The
undergroundstructurecomprisedtwoindependentbut
juxtaposedeyeglass type tunnels crossinggneissic
rock, saproliteandmicaciousresidual silty-sandysoil
andsandyclayishsoil of marineorigin, of highhydro-
geological complexity;theroutealsocrossedadensely
populatedbuilt-uparea.
Oneof thegreatest challengeswastheNATM sec-
tion of the tunnels under the direct foundations of
a 50-year-old 7-storey building, standing on sandy
sediment (beachsand), witharound6moverburden.
Thesetunnelswerepreviouslytreatedwithhorizon-
tal jetgroutingformingaclosed360
conical chamber
with treatment throughout the excavation cross sec-
tionandfrontseptumprotectingeachadvancemodule.
(figs. 6, 7).
289
Figure8. PlazaItalia/CapuchinosTwinTunnels.
Figure9. Sectiondetails.
Thework wassuccessfully concludedontimeand
the preventer valve was a crucial factor in ground
treatment. The maximum settlement recorded was
30mm.
4.2 Plaza Italia Tunnel Line 4 subway (Caracas
Venezuela)
Part of the new Line 4, the North and South Tun-
nelsInitial Sections, headingfromPlazaItliatoward
Capuchinos Station, involves the expansion of an
existing station where two tunnels of approximately
6.5mdiameter wereto beexcavatedoneachsideof
anexistingtunnel, whichhadtobemaintainedopera-
tional (fig. 8and9). Inthiscasetheexistingstructure,
actedasabarrier tothegroundwater flow, raisingits
level andpressure.
Onstartingtheexcavation, insteadof theresidual
soil predictedbytheoriginal geological investigation,
tunnelers foundhighly permeablealluvial sandy soil
with gravel at the bottomof the excavation section,
toppedbyaclaylylayer of residual origin(micaceous
schist), and a sedimentary fine sandy layer, at the
upper part of thetunnel. Theoverburdenwas around
10meters.
Figure10. 360
horizontal jet groutingscheme.

Figure11. Geological profileandtunnel pathdetails.
In order to cope with these difficulties, various
preventivemeasuresweretakeninthedesignandexe-
cutionphases. First of all, thedesignwasadjustedto
the in situ conditions, extending the vault treatment
to the invert or locally, whenever it was necessary,
soas topre-consolidateandreducewater flow(360
Horizontal J et Grouting, fig. 10).

Due to the presence of water under pressure, the
drillingandjetgroutingoperationsattheinverthadto
beexecutedusingthepreventer valve.
Sincethecentral portion of thesection proved to
be in clay material, there was no need to execute a
jet-groutedfrontal septum.
However, specificmeasurestocontrol pressureand
waterflowusingdeephorizontal drains,equippedwith
checkvalveswhennecessary, hadtobetaken.
4.3 Tamanduatei river service tunnel (So Paulo
Brazil)
To our knowledge, this is thefirst caseworldwidein
whichthe360
distributionhasbeenadoptedasapre
consolidationshapeinaNATM tunnel.
Inthiscasethechallengewasthetunnel runningat
about 10munder street level, but deepeningto 25m
when crossing the river (Fig. 11) due to the project
requirementforaminimumof 5munderthebottomof
thedeepest diaphragm-wallscontainingthecanalized
river;thisstructurewouldposeaseriousriskof becom-
ingapreferential water-pathcommunicatingwiththe
river bed.
The geological profiles (Fig. 11) showed that the
tunnel is embedded in an alluvial mass deposited
during the more recent Tertiary period. The matrix
comprises coarse, mediumandfinelight-gray sands,
with N(SPT) 10 to 40 and is very pervious at the
tunnelslevels.
290
Figure12. Treatment schemewithfull sectionseptum.
Insidethismatrix, twohighlyplasticsiltyclaylay-
ers, 2to5mthick, areintercalatedbut onlyrandomly
intersectingthetunnel excavation.
Water tableobservations showedaquick response
during rainy periods. Thesameeffect was produced
whenever the level of the river rose due to rainfall
anywherealongitsupper course.
Before the excavation, as additional precaution,
watertightness was checked by installing threehori-
zontal drains inside the chamber so as to verify the
presence of water under pressure and/or continuous
flow.
4.4 Full scale test for a railroad tunnel (Barcelona
Spain)
TheMadrid-Zaragoza-Barcelona-FrenchFrontierhigh
speedrail linkssectionconnectingtoSantsStationin
Barcelona, requiredatunnel witha120m
2
crosssec-
tionlocatedunderneaththepathof theexistingground
level railway.
Thegeological investigationdetectedthepresence
of alluvial sandy material with high water tableand
high permeability, which was potentially dangerous
for thetunnelsstabilityandtheexistingnearbystruc-
tures, without special soil consolidationtreatment.
In order to verify theefficiency of thehorizontal
or inclined jet grouting consolidation techniques in
submergedconditions,itwasdecidedtocarryoutafull
scaletesttoconfirmthefeasibilityof somealternative
technical solutionsproposed, verifytheconsolidation
designparametersandcheck theeffect of theactivity
at groundlevel, recordingsettlementsandwater table
variations.
Figure13. Full ScaleTest General lay-out.
Figure14. Geological Profilerelatedtothechamber posi-
tion.
On a slightly smaller scale (80% of the origi-
nal dimensions), the test reproduced the executive
conditions of a consolidated horizontal half-section
heading with all the structural and waterproof ele-
mentsexecutedinaccordancewiththeproceduresand
specificationsstipulatedfor theactual structure.
Fig. 13 shows the general lay-out and geometry
of the proposed treatment. The heading has been
dimensioned to simulate the excavation of a stretch
of tunnel 5meterslongbetweentwojet groutedsep-
tums.Thesoil coverabovethetunnelscrownisaround
6meters. Alsothelogisticandoperational conditions
facingtheexcavationcrewduringtheactual workwere
accuratelyreproduced.
Thegeological profileat thetest areais shownin
fig.14. Basicallyit consistsof asequenceof sandyto
siltysoilswithSPT varyingfrom5to15, intercalated
withverythinlayersof siltyclay.Thepresenceof stiff
grayclaywasdetectedonlywell belowtheinvert.
Between the tunnels crown and invert levels, the
soil is a very fine silt that is under hydraulic gradi-
ent andflowingout of theexcavationalmost without
control. Sinceabovethecrownlevel thereisnomore
291
Figure15. Instrumentationlay-out.
Figure16. BenchMarkHN40 Readings.
consistent strata, evenasmall outflowisimmediately
reflectedat thesurface.
Toverifytheeffectsof thejetgroutinginjectiononto
nearby structuresandtheefficiency of thechambers
watertightness, piezometers and bench marks were
installed, properly distributed above and around the
test area(seefig. 15).
A number of horizontal drainswithvacuumsystem
wereinstalled insidetheexcavation section in order
to facilitate the chamber dewatering. A connection
betweenthechamber andthesoil massoutsidewould
bedetectedbytheexternal piezometers. Thereadings
of settlementsandwater tablelevelsweretakendaily.
Fig. 16and17showasummaryof thedatarecorded
throughoutthetestperiod. Despiteverypoorsoil char-
acteristics, it was possible to keep soil movements
aroundthechamber andatgroundlevel under control,
withanaveragevariationof about +/20mm.
The preventer (control valve system) was perma-
nently installedexcept inaparticular situationwhen,
duetoatechnical problemwiththisdevice, amassive
loss of material was experienced. Once the preven-
ter was repaired, theoriginal groundconditions were
reinstatedby injectingandrecompressingtheground
area affected. A pressure gauge was installed at the
mouthof thedevicetoverifythebuild-upof pressure
insidethesoil mass.A maximumvalueof 0.4Mpawas
recorded.
Piezometers readings were stable during the jet
grouting consolidation phase. During the chamber
excavation, onepiezometer showedananomaly indi-
catingapossiblegapintheconsolidatedarea.
Theproblemwassolvedbyreinforcingtheconsol-
idationwithasmall number of additional columnson
Figure17. PiezometersReadings.
onesideof theexternal alignment limitedtothearea
affectedbythephenomenon. Thechamber excavation
was successfully concluded confirming the stability
andwatertightnessof theproposedtreatment.
The test results were fully adopted in the design
and subsequent execution of the main tunnel under
theexistingrailway.
5 CONCLUSIONS
The360
consolidation treatment is thestate-of-the-

art intermsof pre-consolidationtechniqueassociated
withNATM mainlyinalluvial or residual formations,
withhighgranulometryandrelevantpresenceof water
pressure, and in general in all types of soil werejet
groutingtechniquecanformthecolumnelement.
To ensure the level of efficiency required, the
instrumentstopursuetreatmentsimperviousnessand
geometry arethepreventer valvethat is essential in
ordertoavoiduncontrolledlossof material duetopip-
ingtogether withthespecial homothetictemplatesfor
perfect alignment of thedrillingstring.
The 360
consolidation scheme proved to be

extremelyflexible, asitcanbeadjustedtoactual insitu
geological conditions,andwhenexecutedintheappro-
priate manner and with proper equipment, achieves
extremelygoodresultsintermsof groundmovements,
witharecordedaverageof 20to30mmthusallowing
aproper control of sideeffectsonexistingstructures.
So far this technique has be implemented in
urbantunnelswithaminimumoverburdenof around
6 meters. The encouraging experience gained since
1998whenwasexecutedforthefirsttime, isatpresent
leading the authors to study its application in even
shallower tunnelswithasoil cover of 3to4meters.
292
REFERENCES
Guatteri, G. Mosiici, P. Doro Altan, V. Koshima, A. 1998.
BrazilianExperienceinJ et GroutingTreatments inDif-
ficult Tunnels. World Tunnel Congress 98, ITA CBT,
TunnelsandMetropolises, SoPaulo.
Guatteri, G. Koshima, A. Lopes, J. Doro Altan, V. 2000.
360
J etGroutedConical ChamberAllowSafeTunneling
UnderRiverWithinaHighlyPerviousEnvironment. Geo-
enge2000AnInternational ConferenceonGeotechnical
& Geological Engineering; Melbourne, Australia.
Guatteri, G. Koshima,A. Lopes, J. Pieroni, M. 2004. Resume
of BrazilianExperienceaboutUseof J etGroutinginTun-
nelsandUndergroundExcavations.1
BrazilianCongress
of Tunnels SoPaulo.
Guatteri, G. Koshima, A. Ravaglia, A. Duarte, M. 2007.
J et Grouting Solutions for the Caracas Subway Line 4.
ChallengesinUrbanProjects XIII PanamericanConfer-
enceonSoil MechanicsandGeotechnical Engineering
Venezuela
Szio, L. 2002. J et Setting in Rio. Tunnels & Tunneling
International, I-1,15.
293
Theeffectsof sampledimensionandgradationonshear strengthparameters
of conditionedsoilsinEPBM
M. Hajialilue-Bonab, M. Ahmadi-adli, H. Sabetamal & H. Katebi
University of Tabriz, Tabriz, Iran
ABSTRACT: Mechanical properties of conditioned soils in EPBM tunneling consist of lots of unknowns.
Inthis research, thetests has beenarrangedto fulfill of four goals. Firstly theeffects of conditioningonthe
shear strengthvariationhavebeeninvestigated. Secondly aninvestigationoneffects of conditioningonshear
strengthparameters(C,) hasbeenperformed. Inthirdsteptheresultsexertedfromtwopreviousstageshave
beencomparedfor twoshearingapparatusesof conventional andlargeshear boxes. Thelastgoal of theresearch
isexplorationof effectsof changesinconditioningparametersonshear strength. It isfoundthat theC& for
testedsoils obtainedfromlargeshear box areusually greater thantheresults of thesamesoil inconventional
shear test. Thisresultislesssignificantfor conditionedsoil withcomparedtounconditionedsoil anditisfound
tobeafunctionof injectedfoamcontent.
1 INTRODUCTION
Determiningtheshear strengthparametersof thecon-
ditionedsoils usually leadto theuseof conventional
shearboxtestsbecauseof noavailabilitytolargeshear
boxes. In order to use this apparatus, soil gradation
must bemodifiedbasedonthecodeswhichdemands
to eliminationof great mass of coarseparticles. This
eliminationaffectstheshearstrengthparametersespe-
ciallyinconditionedsoil. Inthisresearchtheeffectsof
sampledimensionandsoil gradationonpropertiesof
foamconditionedsoilshavebeeninvestigated.A mod-
ificationonsoil gradationhas beendoneinsamples
of largeandconventional shearboxes.Thevariablesin
this researcharenormal stress andtheinjectedfoam
quantityintoeachof soils.
2 MATERIALSANDPROCEDURESINSOIL
CONDITIONING
2.1 Conditioning parameters
Concentration(C
f
)
This parameter is thecontent of foamingagent in
water unit weight.
FoamExpansionRatio(FER)
This is thevolumetric expansionof aunit volume
of foamingsolution.
FoamInjectionRatio(FIR)
It consists of theratioof injectedfoamvolumeto
volumeof theconditionedsoil.
The last parameter is of great importance in suc-
cessful soil conditioning.
2.2 Laboratory foam generation system
To produce a conditioner (foam) with specific and
controlled parameters, a foamgenerating systemin
laboratory scalewas designed and constructed. This
apparatushastheabilityof controllingfoamproperties
alongtheproductionandcanproduceuniformfoam.
The schematic plan of this systemwhich manufac-
turedinsoil mechaniclaboratoryof TabrizUniversity
isshowninfigure1.
Foamgeneration is performed by mixing process
of air andliquid(consist of foamagent) under pres-
sure. In thefirst step thereservoir must befilled by
foamSolution. Foamgenerationisperformedbymix-
ingprocess of air andliquid(consist of foamagent)
under pressure. In the first step the reservoir must
be filled by foamsolution. A regulator controls the
air pressuresuppliedtothereservoir. Liquidvalveis
opened and flow control valve is used to adjust the
flowthrough theliquid flowmeter to thefoamgen-
erator unit inlet. The pressure in the air flow line is
controlled with a control valve of liquid flow meter
and measured by a pressure gauge. Air pipe valve
is opened to allow the air to flow through the con-
trol valveandtheair flowmeter tothegenerator unit
inlet. Thepressurizedair andfoamsolutionthenflow
through thegenerator unit to producethefoam. The
design of the foamgenerator allows the liquid and
295
Figure1. Schematicof laboratoryfoamgenerationsystem.
0
10
20
30
40
50
60
70
80
90
100
0.01 0.1 1 10 100
mm
%
pA
A
A'
C'
Figure2. Particledistributionof soilsusedinthisresearch.
air flowratesandpressurestobeadjustedandmoni-
toredseparately tocontrol thepropertiesof thefoam
produced.
3 SHEARTESTSPATTERNS
In order to perform large shear box (300300
150mm) testsandconventional shear box(6060
20mm) teststhesoil excavatedfromTabrizmetroline
has beenused. This soil labeledpA. Consideringthe
maximumdimensionof thesoil particleswithrespect
totheshear boxdimension(large, conventional), soil
gradationhasbeenmodifiedtoA, A (by elimination
of the oversized particles) and C (same as A with
30%moreinfinecontent).Particlesizedistributionfor
thesesoilshasbeenillustratedinfigure2. Directshear
testsonsoilsA, A andC wereperformedusingcon-
ventional andlargeshear boxes. Thenormal stresses
onsamplesalsowere37.52, 64.75, 119.29and228.28
(kPa).
a
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp(mm)
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp(mm)
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
Unconditioned
Conditioned
b
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
V
e
r
t
i
c
a
l

d
i
s
p
(
m
m
)
Unconditioned
Conditioned
Figure 3. (a) Variation in ratios of shear to normal stress
and(b) vertical displacementof samplecapversushorizontal
displacement in Conventional shear boxes on soil A under
normal stress37.52kPa.
4 RESULTSANDDISCUSSIONS
4.1 Conditioning effects on shear strength
variation trends
Thebehavior of soil A in conditioned and uncondi-
tionedstateunderdifferentnormal stress(37.52, 64.75
and119.29kPa) indirect shear testshavebeenshown
inFigure3to5. Thevariationof shear/normal stress
ratio versus horizontal displacement has been inves-
tigated. The variation of vertical displacement with
respect to horizontal displacement during shear test
has beenalso considered. Figure3shows theresults
forsoil A inbothconditionedandunconditionedstate.
A reductionof almost0.4inmaximumvaluesof stress
ratioscanbeobserved. Thecurvaturesindicatesome
peak point inlower stressratiosinunconditionedsoil
test. This peak is going to disappear by increasing
the stress level. But in conditioned samples there is
no peak point in any stress level. It can beobserved
that increasinginstresslevel canconvert thedilative
behavior tocontractivebehavior. However thecondi-
tionedsamplesgiveacompletelycontractivebehavior
regardlessof initial specificvolume.
4.2 Conditioning Effects on shear strength
parameters
Table1showscalculatedC&forperformedtests.The
friction anglefor peak and residual states aregiven.
296
a
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp(mm)
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp(mm)
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
Unconditioned
Conditioned
b
-0.4
-0.2
0
0.2
0.4
0.6
0.8
v
e
r
t
i
c
a
l

d
i
s
p
(
m
m
)
Unconditioned
Conditioned
Theresidual frictionanglerepresents thestateof the
soilsinexcavationprocess.
Comparison of corresponding shear tests per-
formed on both conventional and large shear boxes
indicate that the large boxes give internal friction
angle6to 7degrees greater than conventional shear
tests. Howeverinthetestsonconditionedsamplesthis
reducesto3to4. Largeapparatusesalsoresultinlower
cohesionincomparisontothestandardsamples. Also
inconditionedsamplesthistrendistracedbutitisless
intensive.
4.3 Effects of variation in conditioning parameters
on shear strength parameters
Theneededconditioner for soilsusedintestshasbeen
predictedusingtheKusakabe1999formula.
WhereX =thepercentageof soil passing0.074mm;
Y =thepercentageof soil passing 0.25mm; Z =the
percentageof soil passing2.0mm;a =1.0forC
u
>15,
a =1.2for 15>C
u
>4, a =1.6for 4>C
u
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp(mm)
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp(mm)
a
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
Unconditioned
Conditioned
b
v
e
r
t
i
c
a
l

d
i
s
p
(
m
m
)

Unconditioned
Conditioned
Usingthisformulaforsoil A, theFIRis50%. Soil A
has been tested with FIR equal 50%, 30%and 70%
and results are shown in figure 6. For example in
treatment with FIR=30% an increasing of 0.11 in
normal/shearstressratioscanbeobserved. Contrarily,
withFIR=70%no remarkableeffect was observed.
It is clear that increasingFIR do not changethenor-
mal/shear stressratioswhilealittlereductiononFIR
hasconsiderableeffect onthisratio.
Figure7givestheresultsfor another groupof sim-
ilar testsonsoil A withvariousFIR. If oneapplythe
soil gradationA to theKusakabeformula, FIR will
result in40%. Byincreasingfoamconsumption.
In soil samples, a negligible reduction in stress
ratios canbeobserved, but alittlereductioninfoam
consumptionwill resultasignificantincreaseinthese
ratios.Thisrelationinconventional sheartestsisabout
0.17. This means that consuming more foam indi-
catesslighteffectonshearstrengthof samples. During
shearing, porepressurewill generatebut ingranular
soil this will dissipate very quickly. In the presence
of foamthedissipationof porepressurewill postpone
andtheshear strengthwill decreaseinfoamedgranu-
larsoil. Comparisonof changesinstressratiosinlarge
andconventional testsshowsthat conventional direct
shear testsresultsaremoresensitivetothechangesin
297
Table1. Shear parameters.
F
i
n
e
c
o
n
t
e
n
t
A
m
o
u
n
t
o
f
c
o
n
d
i
t
i
o
n
e
r
N
o
r
m
a
l
s
t
r
e
s
s
kPa peak kPa residual

L
a
r
g
e
s
h
e
a
r
b
o
x
5 0 37.52 45.322 50.38 9.221 44.2
64.75 72.888 48.39
119.29 125.08 46.36
5 50 37.52 26.095 34.82 4.42 31.1
64.75 44.93 34.76
119.29 75.929 32.48
5 0 37.52 42.281 48.41 14.62 37.4
64.75 63.667 44.51
119.29 105.95 41.61
228.28 186.68 39.28
5 40 37.52 27.076 35.81 7.259 28.4
64.75 43.066 33.63
C
o
n
v
e
n
t
i
o
n
a
l
s
h
e
a
r
b
o
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
3.5 3 2.5 2 1.5 1 0.5 0
Horizontal disp (cm)
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
FIR=50%
FIR=30%
FIR=70%
Figure6. Variationinratiosof sheartonormal stressversus
horizontal displacement inlargeshear boxesonsoil A under
FIRthantheresultsinthesamesamplesinlargeshear
test. Theresults for other tests givethesameconclu-
sion.ThiscanbeobservedinFigure8.Thisfiguregive
somemoreinformationabout soil C whichissimilar
tosoil A withsomemorefinecontent. Itsobviousthat
theincreasingFIRinsamplesthathaveexcessivefine
causesmoresignificant lossininternal frictionangle.
But in samples which have more fines this phe-
nomenafirstlyoccurssteeplybut inFIRshigher than
optimumthistrendisnot observedat all.
It canbeconcludedthat thesoil whichhavemore
finecontent needlessFIR toget aspecific reduction
of shear strength.
Inorder tostudythepossibleeffectsof variationin
foamexpansionratio(FER), afoaminthreeFERsof
5, 10and22werepreparedfor similar shear tests on
a
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 1 2 3 4 5 6 7 8 9 10
Horizontal disp (mm)
0 1 2 3 4 5 6 7 8 9 10
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
FIR=40%
FIR=20%
FIR=60%
b
v
e
r
t
i
c
a
l

d
i
s
p
(
m
m
)
FIR=40%
FIR=20%
FIR=60%
0
10
20
30
40
50
60
0 10 20 30 40 50 60 70 80
FIR %
A
A'
C'
Figure8. Internal frictionangleversusvariationsinFIR.
soil C. Figure9showsthatareductiononFERcauses
noeffect onshear parametersandstressratios.
Thereasonwhichcanbestatedisthat thelowFER
indicatesverywetfoamsothatitisunabletoreducethe
relativedensityandconsequentlyincreasethespecific
volume. Ontheother handwhentheFERincreasesan
unstablestateoccursandconsequently nochangesin
stressratioshappen. It isresultedthat aFER closer to
15givesastablestatefor foam.
Figure 10 gives a view of obtained variations in
stress ratios of conditioned samples during shearing
298
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
a
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
FER=5%
FER=10%
FER=22%
b
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.9
0.8
v
e
r
t
i
c
a
l

d
i
s
p
(
m
m
)
FER=5%
FER=10%
FER=22%
displacement inConventional shear boxes onsoil C under
becauseof changesinCf. ThreetestsperformedonC
withthreedifferentinitial Cf of 3%, 4%and5%. If we
considerthepossiblechangesinstressratiosthereisno
significant differencebetweenthesesamples. But the
volumeof samplestendstoreduceduringthetest.The
authors consideredthis logical as well as declaration
of producersof conditioners.
5 CONCLUSIONS
1. Thesoil conditioningcauseadecreaseof 7to 11
degreesininternal frictionangleof soils. A lossof
49.1kPafor thecohesionof conditionedsoil was
observed.
2. Decreasing FIR from its optimum given by
Kusakabeformula, will greatly affect thetrendof
reductioninshear strength. But increasinginthat
quantity does not show remarkable effect. FER
lower than 10 and more than 18 causes no con-
ditioningandconsequentlythechangesinstrength
arenegligible.
3. Largesheartestgivesgreaterinternal frictionangle
about 6to7degreesinunconditionedsamplesand
a
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
s
h
e
a
r
/
n
o
r
m
a
l

s
t
r
e
s
s
Cf=3%
Cf=4%
Cf=5%
b
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
V
e
r
t
i
c
a
l

d
i
s
p
(
m
m
)
Cf=3%
Cf=4%
Cf=5%
10b
-
Figure10. (a) Variationinratios of shear to normal stress
displacement inConventional shear boxes onsoil C under
3 to 4 angles in conditioned samples comparing
conventional shear test.
REFERENCES
Ahmadi adli, M, Investigation on mechanical properties
of conditioned soils for EPB mechanized tunneling in
Tabriz, 2006, M.Sc. thesis, Facultyof Civil engineering,
Universityof Tabriz. Tabriz: Iran.
Defence Standard 4240. 2002. FoamLiquids, Fire Extin-
guishing (Concentrates, Foam, Fire Extinguishing).
Ministryof Defence. Issue2. USA.
EFNARC. 2005. Specifications andGuidelines for theUse
of SpecialistProductsforMechanizedTunnelling(TBM).
InSoft GroundandHardRock. EFNARC, UK.
Merritt, A. 2004. Conditioning of clay soils for tunneling
machinescrewconveyors.PH.D.thesis,St.J ohnscollege,
Cambridgeuniversity. London: England.
Milligan, G.W.E. 2000. Soil conditioning and lubrication
in tunneling, pipe jacking and micro tunnelling. A
stateof artreview. http/www.civils.eng.ox.ac.uk/research/
pipejacking.htm
Psomas, S. 2001. Propertiesof foam/sandMixturesfor tun-
nelling applications. M.Sc. thesis, St. Hughs college,
Oxforduniversity. London: England.
299
Experimental studyoncompressibilitybehavior of foamedsandysoil
M. Hajialilue-Bonab, H. Sabetamal, H. Katebi & M. Ahmadi-adli
University of Tabriz, Tabriz, Iran
ABSTRACT: Inordertoassesstheinfluenceof differentfoamtypesoncompressibilitybehaviorof conditioned
sand, aset of testswereperformedonthreegradationof sandysoil. Someindextestswerealsoundertakenfor
verifyingfoamagentscharacteristicsandfoamgeneratorquality.Detailsanddiscussionaboutdifferentaspectsof
mentionedcaseshavebeenpresentedinthispaper. Compressibilitytestswereperformedbya151mmdiameter
RoweCell andfoamgenerationwascarriedout byfoamgenerator whichwasconstructedbytheauthors.
1 INTRODUCTION
Ideal groundconditionsfor EPB machinesconsist of
soilswithrelativelyhighfinescontentssuchasclayey
silts or silty sands, withaconsistency to formalow
permeability, softplasticpastewhenexcavated.These
propertiesallowthesupportpressuretobetransferred
uniformly to the tunnel face and controlled flow of
thesoil throughthemachine. Thecompressibility of
soil has an important function on machine perfor-
mance. However, natural soilsrarely havetheseideal
properties, andconditioningof thesoil isusuallynec-
essary to change its properties to suit the machine.
Soil conditioning for EPB machines involves inject-
ing conditioning agents into the excavated soil. The
objectiveistomodifythesoil propertiestoformasoft
plasticpaste, leadingtoimprovementsinthemachine
performanceandcontrol of theexcavationprocessin
awiderangeof soils.Thespecifictreatmentsrequired
to effectively conditiondifferent types of soil. Many
factorsinfluencethespecificationandperformanceof
soil conditioningtreatments.
Soil conditioningisclearlyperformedbymeansof
somechemical &physical materialssuchasfoamand
polymerwhichleadtodecreasepermeability, increase
plasticity, etc. TheSCA (Soil ConditioningAgent) is
also responsiblefor controllingtherheological prop-
erties of theextractedsoil, andminimizingwear and
abrasionof thecutter headduringtunneling. Inorder
to achieve all of these objectives, the SCA compo-
sition must be tailored to the properties of the soil.
Foams, mainly consisting of surfactants, areused in
fine-grainedmaterialstoreducematerial adhesionto
exposed EPBM surfaces, and to fluidize the muck
(reduceballingandenhanceflowability). Foams are
alsousedingranulardepositstoimprovematerial rhe-
ology andreducesoil permeability. Whiletheuseof
SCAs in EPBM tunneling is on the rise, the use of
SCAsingeneral isverymuchablack box technol-
ogy. AccordingtoPsomas(2001) foam/sandmixtures
exhibit low shear strength and they have very com-
pressible, so inorder to assess theinfluenceof foam
and polymer on shear strength and Compressibility
behavior and influence of foamparameter in those,
supplementary tests performed with variety of foam
agentsandfoamparameterssuchasFER, FIR andCf
onthreegradationof sandysoil (seeSabetamal 2006).
This paper focused on theresults of compressibility
testswhichhavebeenundertakenbytwotypesof foam
agents.
2 EQUIPMENT ANDMATERIALS
2.1 Foam generator
Lab scaleTBM foamgenerator with full controlling
possibilityonvariouspartof systemwasdesignedand
constructed.Theoperationof thelaboratoryfoamgen-
eratorisdescribedherewithreferencetotheschematic
diagramshown in figure1. Foamgeneration is per-
formedbymixingprocessof air andliquid(consistof
foamagent) under pressure. Inthefirst stepthereser-
voir must befilledbyfoamsolution. Thenit flowson
route2viaairpressure.A regulator(b) controlstheair
pressuresuppliedtothereservoir andthecompressed
air isflowedinroute(1). Theentranceair pressureis
measuredbygaugea. Valve(2) isopenedandflow
control valveisusedtoadjusttheflowthroughtheliq-
uidflowmeter tothefoamgenerator unit inlet. Valve
(1) isopenedtoallowtheair toflowthroughthecon-
trol valveandtheair flowmeter tothegenerator unit
inlet.Theairpressureismeasuredbyapressuregauge
(e). Thepressurized air and foamsolution then flow
301
Figure 1. a: Photo of laboratory foam generator.
b: Schematicof laboratoryfoamgenerator.
through thegenerator unit to producethefoam. The
designof thefoamgenerator allowstheliquidandair
flowratesandpressurestobeadjustedandmonitored
separatelytocontrol thepropertiesof thefoam.
2.2 Rowe cell
Compressibility tests were carried out in a standard
151mmdiameter RoweCell. Stiff porous discs were
placed on the top and bottomof the sample so that
the boundary conditions are fixed strain rather than
freestrain. All tests wereundertaken at seven stress
level andeachlevel composedof twostages:undrained
stage(immediatesettlement is obtained) anddrained
stage (settlement due consolidation is obtained). At
each stress level, themeasurement of drained liquid
volumes in second stage leads to calculate the air
andliquidvoidratio. Vertical settlement andvertical
Figure2. Compressibilitytestsequipment set up.
Figure3. Sandsgradations.
Table1. Sandsproperties.
Soil GS emin emax D50mm
A 2.65 0.452 0.695 1.0
B 2.65 0.660 0.940 0.25
C 2.65 0.580 0.766 0.60
pressure are measured using LVDT and pressure
transducer respectivelyaccordingtofigure2.
3 STUDIEDMATERIALS
3.1 Sand
The gradation curves of different sandy soils which
wereusedinthisresearchareshowninfigure3. The
soil typeA andBarequiteuniform. Thesoil typeCis
amix of soil A andB. Theproperties of thesesandy
soilssuchasG
s
, e
min
, e
max
areshownintable1
3.2 Foam agents
Four different types of foamagent and one type of
polymer wereused in foamindex tests. Only two of
302
themwere used in compressibility tests. Agent type
DisproducedinIranandvastlyusedinShiraz metro
project. Other agents (namedA, B, C) andpolymers
(namedP) areproducedbyDegussa. Thefoamgener-
ator wasabletooperatewithall of themthat produce
acceptablequalitymicro-foam.
4 FOAM INDEX TESTS
The foam properties depend on its different com-
pounds like air, water, surfactant and sometimes
polymer.Theparameterswhichcharacterizefoamare:
Surfactant Dosage=C
f
[%]
Polymer Dosage=C
P
[%]
Air Ratio(FoamExpansionRatio)=FER
FoamInjectionRatio=FIR
25%or 50%drainagetimeaccordingtoMinistry
of defenseStandard4240
The amount of air introduced to the soil can be
changedwiththeairratioFERwhichcharacterizesthe
ratiobetweenair andliquidvolume. Thefoaminjec-
tionratioFIR indicates thevolumeof foamusedper
1m
3
excavatedsoil.Inordertoverifyingtheproperties
of producedfoamandtoinvestigatethecapability of
foamgenerator toproducequalifiedfoamwithdiffer-
entagents,50%drainagetimeweretestedforproduced
foams.
Accordingtotheindextests, it isfoundthat topro-
duceof foamwithequal FER, theamount of air and
liquid flows must be adapted with each foamagent
type. Theamount of FER withrespect toCf for foam
D at the same condition is shown in figure 4a. It
canbeobservedthat, withincreasingof Cf, FER will
increase.
Thestabilitytimesfor different foamareshownin
figure4b.Thestabilitytimevariesfrom6to15minutes
for different foams. Byadding2.5%polymer tofoam
agent B the stability time will be doubled. This can
beobservedfor all foamagents. Thisisanimportant
result asthestability timeisvery important for some
kindof excavations.
5 COMPRESSIBILITY TESTS
Thecompressibilitytestswerenormallyperformedat
four stress level in which thestress bedoubled with
respecttopreviouslevel.Becauseof samplesensitivity
tostresslevel, thetestswerecarriedoutatsevenstress
levels. For thisreasontheincremental ratiodecreased
from2 to 1.414 fromtertiary step (i.e. doubling in
two loadsteps). Duringeachloadstagethedrainage
valve was closed at the beginning, in order to mea-
sureanundrainedcompression. Suchacompression
wouldbenegligiblefor asaturatedspecimen, but for
Figure4. a: Effect of Cf onFER for foamDb: Comparing
of foamstability.
Figure 5. Drained and undrained settlement at end stage
(@450kPa).
thesand/foammixturesthereisasignificantcompres-
sionof theair bubblesinthefoam. Thenthedrainage
valvewasopenedandfluidallowedtodrainfromsam-
ple. A typical plotof displacementagainsttimewhich
showsboththeundrainedanddrainedphasesisgiven
infigure5.
Figures 6, 7 illustrate the variation of void ratio
against vertical stress in semi logarithmic scale for
303
Figure6. Voidratioagainst loadlogarithmfor sandB.
sands B, C. Only thelast points for eachloadincre-
mentareplotted, i.e. thevoidratioachievedattheend
of thedrainedstage.Thesoil typeAwascoarserthanB
andCandinorder toachieveahomogeneousmedium
for this soil, other conditioning agent such as poly-
mer andbentonitewasneeded. Concentration(Cf) of
foamagent for all samples is 3%. Index f at curves
indicates foamed sands. The lowermost curve is for
unconditionedsand, initiallypreparedinaloosestate.
At low stress level the void ratio lies just below
the maximumvoid ratio for the sand. As the stress
increases, thevoidratioreducesbyasmall increment,
but the sand remains in a loose to mediumdensity.
(Demonstrating the static stress is not sufficient for
sandcompaction). Theupper curves arefor different
testsonfoam/sandmixtures. Thedifferencesbetween
theinitial voidratios at lowstress levels areentirely
duetotheamountsof addedfoam.Atlaststagethevoid
ratio of thesand/foammixtureis still well abovethe
voidratio at 0%relativedensity. It was unsurprising
that sand/foammixtures couldbemadeat highvoid
ratios, but it wasquiteunexpectedthat suchhighvoid
ratios couldbesustainedat aremarkably highstress
level. Notethatthesand/foammixturehasatrulycom-
posite action rather than the sumof the component
parts. Sand would have been compacted to a much
lower density, andthefoambyitself wouldhavebeen
crushed at such a stress but the sand/foammixture
is stablein a remarkably loosestate. This may have
fundamental implicationsfor tunnelingoperations.
Accordingto performedtests, thefoam/sandmix-
tures voidratio aregreater thanmaximumvoidratio
of drysandandthischaracteristicsof foam/sandmix-
tures is independent with respect to thefoamtypes.
Withincreasingof FIR thevoidratioincreasesinten-
sively but increasing of FER would have negligible
effectsonvoidratio. Infigures6, 7twocurvesshow
different behavior withrespect toother curves. These
twocurvesareBf22andCf10teststhatarepreparedby
foamC.TheothersampleswerepreparedwithfoamD.
Figure7. Voidratioagainst loadlogarithmfor sandC.
Voidratioinfoamedsands has beencomposedby
twocomponents: air andliquidvoidratio, majorityof
voidratioinfoam/sandmixtureconsistsair voidratio.
If thefoambubblesarestable, thecompressibility of
bubblescouldleadtohighcompressibilityof mixture.
Thisamountof compressibilityviawaterdrainagewill
benotpossibleandalsoitneedmoretimetodrain.The
role of consolidation settlement is much lesser than
immediatesettlement (Fig. 5).
Accordingtofigures6, 7, thesamplesthat arepre-
paredbyfoamDshowlowercompressibilityandlesser
sensitivitywithrespecttosamplesthatarepreparedby
foamC.
Thisphenomenonpresumablyisrelatedtolowsta-
bilityof foamDagainstloading.Butdifferentdrainage
timesfor twotypesof agentsarenot considerable. To
investigate this phenomenon, the volume change of
thismixturewasverifiedbyeachvolumechangecom-
ponent (air and water). In figure 8, the void ratio
changesforsoil anditscomponent, airandliquid, have
beendrawnwithrespecttopressurefortwosand(B,C)
which mixed by foamtypeC and D. Theinitial and
final valuesareusedfor thisplot. General tendencyof
voidratiochangesisdistinct. It isclear that thewater
voidratiovariationsfor sampleswhichpreparedwith
foamD(Figs8a, 8c) aregrater thanair voidratiovari-
ations. Theincreasingof air voidratiohasbeingseen
insomecases(Bf20) becauseof excesssampledrying
indrainageprocess. AlthoughinBf22andCf10tests
(Figs 8b, 8d) air voidratiodecreasemorethanwater
voidratio. Itmeansthattheamountof drainedwateris
lesser thanpreviouscases, andthemajorityof volume
changesweretakenplaceviacompressibilityof foam
bubbles.
Theairvoidratiochangesarecomparedinfigures9,
10. Theair voidratiofor samples withfoamD gives
either anegligiblechangeor asmall increase(except
Cf7). It can be concluded that the foamD increase
thewater content and consequently thevoid ratio of
mixture preserving the homogeneity. But the com-
pressibilitysislesser becauseof lowstabilityagainst
loadingandduringthetestprocessfoambubbleswash
304
Figure8. Voidratiosvariationversusvertical stressforsand
B,C withtwodifferent foamagents.
Figure9. Air voidratioversusvertical stressfor sandB.
Figure10. Air voidratioversusvertical stressfor sandC.
out viawater drainageinsecondstagesof tests. Foam
C also cause the increasing of water content in the
mixture, moreover the sample is high compressible
consequently shows awell sensitivity withrespect to
loading(Figs6, 7). Thisdifferencebetweentwotypes
of foam agents are related to their chemical com-
positions. Note that the similar shear strength was
obtainedfor samplespreparedbytwodifferentagents
(Sabetamal 2006). Butdependsontheir compressibil-
itydifferentsettlementswereobtained.Thus, thefoam
D must beimprovedfor thats lowstability timeand
lowcompressibility inorder to useinTBM machine
insandysoil.
6 CONCLUSION
1. Index tests on various foamagents show that the
maindifferences betweentheseagents areonsta-
bility or drainagetime. Thepolymer increasethe
stabilitytimesignificantly.
2. One of the foam/sand mixtures characteristics is
highcompressibility. Existenceof foambubblesin
thesoil skeletonincreasevoidratios,thisfeaturefor
all foamagentsiscommonbutthebehaviorof sam-
pleduringloadingisaffectedby foamtype. Thus,
useof samefoamagent for different condition is
not suitable.
305
3. Largepart of voidratioinfoam/sandmixturecon-
sistsof air voidratio. If thefoambubblesarestable
against vertical stress and during thetest process
foambubblesdonot washout viawater drainage,
thecompressibility of bubbles could lead to high
compressibilityof mixture.
4. SomefoamagentssuchasfoamDincreasethevoid
ratio at first but because of their unfitted chem-
ical compositions they can not sustained against
loading.
5. Withincreasingof soil particlesizeatuniformgra-
dation, useof foamaloneas aconditioningagent
isnotsufficientasthefoambubbleseasilyescaped
fromthe soil voids and assembled on soil mass.
Consequently the composite action of soil/foam
fails. So in coarser soils other conditioning agent
suchas bentoniteas afiller of voids must also be
used.
REFERENCES
Defence Standard 4240. 2002. FoamLiquids,Fire Extin-
guishing(Concentrates, Foam, FireExtinguishing). Min-
istryof Defence. Issue2. USA.
EFNARC, 2005. Specificationandguidelinesfor theuseof
specialist productsfor soft groundtunneling.
Head, K.H. 1994. Manual of soil laboratory testing: shear
strength, Compressibility and Permeability. 2nd edition,
Vol.2, Pentechpress: London.
Milligan, G.W.E. 2000. Soil conditioningandlubricationin
tunneling, pipejackingandmicrotunneling.A stateof art
review. web site: http/www.civils.eng.ox.ac.uk/research/
pipejacking.htm
Psomas, S. 2001. Propertiesof foam/sandmixturesfor tun-
neling applications: M.Sc. thesis, St. Hughs college,
OxfordUniversity. London: England.
Sabetamal, H. 2006. Study of soil conditioningongranular
soils for EPB Tunneling: M.Sc. thesis, Faculty of Civl
engineering, Universityof Tabriz. Tabriz: Iran.
306
Studyonearthpressureactinguponshieldtunnel lininginclayeyandsandy
groundsbasedonfieldmonitoring
T. Hashimoto, G.L.Ye, J. Nagaya&T. Konda
X.F. Ma
ABSTRACT: Inthis study, firstly, aseries of fieldmonitoringdataonearthpressureinsoft clay, hardclay
andsandgroundareanalyzedrespectively. Inthesoft clay ground, theearthpressurefluctuatedwithbackfill
groutingpressureatfirst, itsettleddowntowardasteadyvaluebetween(staticvertical pressurePv0cohesion
C) anddistributeduniformlyover theringfinally. Inthehardclayground, theearthpressurewasmoregreatly
influencedby thebackfill grouting, especially for thelateral earthpressureinhardclay with2C,p
v0
0.5. In
thesandground, althoughtheearthpressurewasalsoinfluencedby thebackfill grouting, thedistributionwas
relatively uniformthaninhardclay groundsincethehydraulic pressureaccounts for alargeportioninearth
pressure.Theinsightsobtainedfromthisstudycancontributetoanimprovementof loadconsiderationsinshield
liningdesign.
1 INTRODUCTION
In the design code of shield tunnel in J apan (J SCE,
1996), the earth pressure acting upon the segment
lining is calculated by the overburden pressure or
Terzaghis loosening earth pressure according to the
stratum condition and the overburden height only.
However, it is known that the earth pressure is also
influencedby theconstructionconditions (e.g. back-
fill grouting, position adjusting of shield machine),
andtheinteractionbetweenthegroundandman-made
structures(e.g. tunnel lining, pilefoundation). Inmost
cases, thesefactors work together andundistinguish-
able. Therefore, nowadaysthemechanical behavior of
the earth pressure upon shield tunnel lining has not
beenclearlyclarifiedyet.
Someresearcheshavebeendoneonthisproblemin
thelast decades. After theTerzaghis theory onloos-
ening earth pressure, Murayama (1968) studied the
vertical earthpressureinsandylayersbytrapdoortests.
Accordingto thetest findingthat theslidingsurface
is similar to logarithmspirals heproposedaformula
to calculatethevertical earthpressure. However, the
up-to-data shield technology equipped with precise
pressurecontrol systemat cutter faceandsimultane-
ousbackfill groutingsystemmakesitpossibletobuild
a tunnel without loosening the surrounding ground.
Therefore, the actual earth pressure cannot be cor-
rectlypredictedbyconventional methods(Ohtaet al.,
1997, Hashimotoetal., 1997). Moreover, Suzuki etal.
(1996) reportedthatthemaximumloadsoccurreddur-
ing backfill grouting in a shield tunnel with large
overburden.
In this study, firstly, threecases of field monitor-
ingonearthpressureinsoft clay, hardclay andsand
grounds are analyzed carefully. And focus is set on
thelong-termbehavior. Theresults distributions of
earth pressure, axial force and bending moment of
lining are compared with the calculated value by
conventional design method respectively. And more
than twenty measurement data are summarized and
organizedaccordingto thestrengthof thegroundto
findout someempirical rules of earth pressure. The
insights obtained fromthis study can contribute to
animprovement of loadconsiderationinshieldlining
design.
2 BRIEF DESCRIPTIONOF MONITORING
SITES
Inordertoclarifythecharacteristicsof earthpressures
actinguponthelinings inclayey andsandy grounds,
three typical monitoring jobs are chosen fromthree
307
Table1. Descriptionsof shieldsandgeologyconditions.
Kadoma Osakajo-A Osakajo-B
Sitename (Soft clay) (Hardclay) (Sand)
Shieldtype Mud-soil Earthpressure Earthpressure
pressure balanced balanced
balanced
Segment type Ductile RC RC
Shield 5300mm 5300mm 5300mm
diameter
Backfill Simultaneous Simultaneous Simultaneous
groutingtype
Overburden 14.09m 28.2m 16.8m
height
Aroundsoil Alluvial clay Diluviumclay Sand
type
SPT-Nvalue 35 89 >50
Unconfined 170200kPa 540kPa
compressive
strength
types of ground respectively, namely, soft clay, hard
clayandsandgrounds. Somebasicinformationonthe
shieldsandgeological conditionsarelistedinTable1,
andthesoil profilesof monitoringsitesareshownin
Figure1. Moreover, two comparing monitoring sec-
tions weresetupat Osakajo-A andOsakjo-B sites to
check theinfluenceof backfill grouting. In order to
obtainareliableearthpressuredata, apadtypeearth
pressure cell (Hashimoto et al., 1993), as shown in
Figure 2, was adopted in all monitoring jobs. The
water pressures were recorded by piezometers from
thegroutingholes.
3 FIELDMONITORINGSOF EARTH
PRESSURESACTINGUPONLINING
Thetimehistoriesof observedearthpressuresaround
thelining in 3 types of ground areshown in Figure
35. Theleftpartof thefiguresrepresentstheshort-
termdata, andtheright part isthelong-termone. The
earthpressureandwater pressurearenotedasEP and
WP hereinafter. Without mentions, theearthpressure
meanstotal earthpressure.
3.1 In soft clay ground
Figure3showsthechangesof observedEPsandWPs
at crown, left spring-line, right spring-lineandinvert
of thelining in soft clay ground. After thetail pass-
ing, the EPs fluctuated greatly by backfill grouting
pressure, and the fluctuation almost disappeared at
7thringsafter tail passing, whichindicatedtheextent
of influencefromgrouting holes is about 7 rings in
suchagroundcondition. After that, theEPsandWPs
decreasedgraduallyinthefirst12months, andthen
Figure1. Soil profilesat monitoringsites.
Figure2. Padtypeearthpressurecell.
theWPs remained unchanged while the EPs turned
toincreasedlittleby little, finally reachedaconstant
status (except R-spring). Thesefinal constant values
liebetween(static pressureP
0
cohesionC). Above
phenomenacanbeexplainedasfollowing.Theexcess
porewater pressureadjacent tolining, whichwasgen-
eratedduringshieldadvancingandbackfill grouting,
dissipatedinthefirst12monthsresultingdropdowns
of EPsandWPs, andthentheexcessporewater pres-
sureinthefarersurroundingsoil dissipatedalongwith
thedecorationof soil skeleton, resultedinaveryslow
buildupof effectiveEPsuponlining.
308
Figure3. Observedearthpressureactinguponlininginsoft
clayground(KadomaShield).
Figure4. Observedearthpressureactinguponlininginhard
clayground(Osakajo-A Shield).
3.2 In hard clay ground
Figure4showsthechangesof observedEPsandWPs
attwocomparedcasesinhardclayground. InCase1, a
regular backfill groutingwasimplemented. InCase2,
in order to eliminate the influences (the squeezing
force to the ground) of backfill grouting, a special
groutingwithsmaller injectionpressureandinjection
ratio was used, as shown inTable2. Sincethelong-
termchanges in both cases were influenced by the
parallel shieldpassing, earthquakeandconcretecast-
ingof invertandarcade, andshowedincreasingtrends,
theattentionwas paidto thechanges beforethe2nd
shieldpassing. InCase1, after all theEPs primarily
climbingupto 300400kPaunder theinfluenceof
backfill grouting, theEPsat crownandinvert turned
to decrease. On the other hand, the EPs in Case 2
was smaller (about 200kPa) at first and then turned
to be increased by backfill grouting of next several
Table2. Backfill groutingof Osaka-A shield
Case1 Case2
Injectionpressure 150kPa 50kPa
Injectionratio 139% 100%
Groutingmaterial Standardstrength* Lowstrength
*Twocomponentsgroutingmaterial withagel timelessthan
10seconds.
Table3. Backfill groutingof Osaka-B shield
Case1 Case2
Injectionpressure 300kPa 170kPa
Injectionratio 135% 100%
Groutingmaterial Standardstrength Lowstrength
Figure 5. Observed earth pressure acting upon lining in
sandyground(Osakajo-B Shield).
rings. Comparisons betweentwo cases showthat the
final differences were very small. The final EPs at
thecrowninbothcaseswere370380kPa. Thefig-
ureshows that themain part of each EP at different
location was determined at thefirst 10 rings, which
indicates that theEP largely depends onthebackfill
groutingin thehardclay ground. Andit also can be
seenthat thesettlingdownof EP onlyneededseveral
days, muchquicker thaninsoft clay.
3.3 In sand ground
The same as that in Osaka-A shield, here also two
cases of comparing measurements were carried out,
as shown in Table 3. Figure 5 shows the changes
309
Figure6. Conventional design method in J apan (effective
stressmethod).
of observed EPs and WPs in both cases. In Case 1,
right after the backfill grouting of current ring the
EPswas100150kPa, thenroseupto200kPaafter
thegroutingof next23rings, andfurther increased
to 200250kPa4 months later. On theother hand,
in Case 2, right after the backfill grouting of cur-
rent ringtheEPswereonly80110kPa, anddidnot
increased significantly after next several rings. The
final valueswere120160kPaeven4monthslater.
Thefigurealsoshowsthat althoughthelong-termEP
inbothcasesinfluencedbytheparallel shieldpassing
and earthquake, and therefore showed an increasing
trend, however, themagnitudes did not changea lot
fromthoseafter thefirst several rings backfill grout-
ing. AndcomparingtotheEPsinclayeygrounds, the
settlingdownof EPs inthesandy groundwas much
faster.
4 COMPARISONBETWEENDESIGNAND
OBSERVATION(HASHIMOTOET AL. 2002)
InJ apan,theconventional model andthebeddedframe
model are main design method for shield lining. In
most commonsituation, suchas theoriginal designs
of the three tunnels in previous session, the engi-
neer prefers the conventional model for the sake of
convenience. We also used conventional model for
comparisoninthisstudy. Figure6isaconceptual fig-
ureof conventional model, whichconsiderstheEPand
WP aswell assoil reactionasliningloads.
The comparisons between the observed EP and
member forces as well as the calculated values are
Figure7. p
v
,p
v0
vs. 2C,p
v0
andp
h
,p
v
vs. 2C,p
v0
(clayey
ground).
shown in APPENDIX. The EPs over the ring in all
groundconditionswererelativeuniformover thering
by comparing with the design values. Consequently
theaxial forceandthebendingmoment weresmaller
than thedesign ones. As mentioned in previous ses-
sion, the EP was greatly influenced by the backfill
groundinrelativehardground, anditisnotdifficultto
imagethat thebackfill groutingcanflowintothetail
voidquicklyandcover thewholeringwithauniform
pressure.Therefore, theuniformdistributionof EPsis
directlyrelatedwiththebackfill grouting.
5 RELATIONBETWEENEARTHPRESSURE
ANDGROUNDCONDITIONANDITS
INTERPRETATION
Besides aforementionedthreemeasurement data, we
alsocollectedmorethantwentydata(long-termdata)
fromother monitoring jobs in J apan, including the
shield tunnels for subway, sewer, power-supply and
communication, and summarized themaccording to
thestrengthof theground. Theauthorswouldliketo
emphasizethat based on theselimited datait is still
difficult to executeaquantitativeanalysis. Therefore
inthissessionaqualitativeanalysisonthebehavior of
EPsinvariousgroundconditionsisgivenout.
5.1 In clayey ground
The relation between p
v
,p
v0
and the normalized
strength2C,p
v0
isplottedinFigure7(a), wherep
v
is
theeffectivevertical EP at thecrown, p
v0
istheeffec-
tiveoverburden pressure, p
v0
is thetotal overburden
pressure, andC isthecohesion. It iswell knownthat
innatural clayey ground, therehas suchanempirical
relationasq
u
(=2C)=0.30.4p
v0
. Whentheground
isunconsolidatedwithaverysmall 2C,p
v0
(-0.3), the
tail voidiseasytocollapseimmediatelyaftertail pass-
ing, andcannotbeeasilyfilledbythebackfill grouting
intime.Therefore, thesurroundingsoilswill yieldand
mostof theoverburdenweightwill actuponthetunnel
(p
v
,p
v0
1).Andwhenthegroundisconsolidatedor
310
alittleover-consolidated(2C,p
v0
0.5), thegrouting
material canfill thetail voidintimebeforecollapse.
Ontheother hand, whenthestiffnessof thegrouting
material andsurroundinggroundareclose, theirinter-
actionbecomesactive. Consequently, p
v
,p
v0
depends
onthegroutingpressureandvarieslargely according
totheconstructionconditions. Furthermore, whenthe
ground is stiff enough (2C,p
v0
becomes large), the
shrinkageof groutingmaterial duringhardeningwill
belarger thanthedeformationof surroundingground.
Thebackfill groutinghasminoreffectontheEP,result-
inginasmall p
v
,p
v0
. ThedatainFigure7(a) indicate
thosekinds of phenomenon. For abetter understand-
ing, two boundary lines aredrawninthefigure, and
for theareaslackingof databrokenlinesaredrawn.
The relation between p
h
,p
v
and the normalized
strength2C,p
v0
is showninFigure7(b), wherep
h
is
theeffectivehorizontal EP at thespring line, andp
v
is the effective vertical EP at the crown. It is found
thatwhen2C,p
v0
issmall (-0.3), p
h
,p
v
=0.450.8.
Consideringthecoefficientof lateral earthpressureat
rest (K
0
) of clayey groundis about 0.5andtheloca-
tionof springlineisdeeperthanthecrown,suchvalues
of p
h
,p
v
arerational. When 2C,p
v0
0.5, thevalue
of p
h
,p
v
scatters in a wide range, implying that the
circumferential distributionof earthpressureinsuch
groundalsodependsonthebackfill groutingandother
constructionconditions.
5.2 In sandy ground
Therelation between p
v
,p
v0
and theequivalent SPT
(Standard Penetration Test) N value which is
regardedasastrengthindex isplottedinFigure8(a).
By applying theTerzaghis loosening earth pressure
toanassumptiveshieldtunnel, wecanget acurveof
looseningearthpressureagainst SPT-N value, as the
dotlineinFigure8(a). However, duetotheinfluences
in construction process, the EP acting on the lining
will varyaboveorbelowthetheoretical line.Whenthe
groundconditionisrelativelygoodandtheloosening
earthpressureissmall, aproper backfill groutingcan
reducetheEP dramatically, otherwisealargegrout-
ing pressure may remain on the lining. The former
phenomenonoftenoccursinthesandgroundmingled
withcohesivesilt/clay. ThedatawithSPT-N between
4090 in Figure 8(a) are just some vivid instances.
Furthermore, when the ground becomes very dense
sandor gravel, similar totheclayeygroundwithlarge
2C,p
v0
, theshrinkageof groutingmaterial maylarger
thanthedeformationof surroundingground, andthe
EPconsequentlybecomesverysmall, suchasthedate
withSPT-N>100inthefigure.
Although thereis no enough observed datain the
SPT-N-40ground, wecanmakeanestimationfrom
theknowledgeof soil mechanics.Whenthesoil isvery
loose, thedependencyonconstructionshall beweak-
ened(thesameasthesoft clay ground), andthenthe
Figure 8. p
v
,p
v0
vs. SPT-N and p
h
,p
v
vs. SPT-N (sandy
ground).
distributionof EPwill getclosetothetheoretical loos-
eningearthpressure, asthebrokenlineinFigure8(a).
For abetter understanding, theauthorsalsodrawtwo
boundarylinesinthefigure.
Therelationbetweenp
h
,p
v
andtheequivalentSPT-
N value is shown in Figure 8(b). It is found that
p
h
,p
v
scattersbetween00.6, indicatingthatthecir-
cumferential distribution of earth pressure in sandy
groundalsodependsonthebackfill groutingandother
constructionconditions.
Theshapesof theboundarylinesinFigure7(a) and
8(a)havethesamecharacteristics: twoendsarenarrow
whilethemiddleis wide. This kind of shapeclearly
tellsusthatwhencalculatingtheEPinthegroundwith
mediumstiffness, more attention should be paid to
thebackfill groutingandotherconstructionconditions
than thosein soft or very hard ground. It should be
pointedout that, however, thebackfill groutingmust
beconsideredwhendealingwiththesettlementinsoft
clayground(Hashimotoet al. 1999).
6 CONCLUSIONS
In order to clarify the mechanical behavior of earth
pressure acting upon shield tunnel lining in various
groundconditions, sayingsoftclay, hardclayandsand
grounds, aseriesof fieldmonitoringhasbeencarried
out.Byanalyzingtheobservationscarefully,following
conclusionswereobtained.
1. Theearthpressureisinfluencedbytheinjectionof
backfill groutingat 78ringsawayinthecaseof
simultaneousbackfill grouting.
2. Inthesoftclayground, theearthpressurefluctuates
withthebackfill groutingintheearlyphase. How-
ever,itsettlesdowntoasteadyvaluebetween(static
pressure P
v0
cohesion C) finally, regardless of
thebackfill groutingpressure.
3. Inthehardclay andsandground, theinitial earth
pressurethatbuildsupgraduallybybackfill grout-
ingremainsinthelong-termearthpressure.Inother
words, the earth pressure depends on the back-
fill groutingto suchanextent that sometimes the
311
earthpressurewill belarger thanthepredictionby
Terzaghislooseningearthpressure.
4. The settling down of earth pressure in soft clay
groundneeds several months, whilethoseinhard
clay and sand grounds only need several days or
evenseveral hours.
5. The distributions of earth pressure are more uni-
form than predictions by conventional design
method in all ground conditions. In other word,
the bending moment is apt to be overestimated,
especially in thesandy ground wherehydrostatic
pressure plays a dominant role in earth pressure.
Therefore, it is suggestedthat whendesigningthe
lining, theinfluenceof backfill groutingshouldbe
takenintoconsideration.
6. Byanalyzingmorethantwentymonitoringdata, it
isfoundthat intheclayey ground, if 2C,q
u
-0.3,
a large portion of the overburden will act upon
thelining. If 2C,p
v0
0.5, themagnitudeanddis-
tribution of earth pressure depend largely on the
backfill grouting. Andinthesandy ground, if the
equivalent SPT-N valuelies between 4080, the
magnitudeand distribution of earth pressurealso
dependlargelyonthebackfill grouting.
REFERENCES
Hashimoto, T., Nagaya, J. & Konda, T. 1999. Prediction of
ground deformation due to shield excavation in clayey
soils. Soils and Foundations, Vol.39, No.3, pp.5361.
Hashimoto, T., Nagaya, J., Konda, T. & Tamura, T. 2002.
Observation of lining pressure due to shield tunneling.
Geotechnical aspects of underground construction in soft
ground, IS-Toulouse, Kanster et al. (eds), Specifique,
pp.119124.
Hashimoto, T., Yabe, K., Yamane, S. & Ito, H. 1993. Devel-
opmentof Padtypeearthpressurecell forshieldsegment.
Proc. of 28th annul report of JGS, pp.20552058. (in
J apanese)
J apaneseSocietyof Civil Engineers(J SCE). 1996. Japanese
Standard for Shield Tunneling. 3rdedition.
Murayama, S. 1968. Earth pressure on vertically yielding
section in sand layer. DPRI Annuals, No.11(B), Kyoto
University, pp.549565.
Ohta, H., Shiotani, T., Sugihara, K., Hashimoto, T. &
Nagaya, J. 1997. Considerationof DesignEarthPressure
forShieldTunnel BasedonMeasurement. Proc. of Tunnel
Engineering, JSCE. Vol.9. pp.3742. (inJ apanese)
Suzuki, M., Kamada,T., Nakagawa, H., Hashimoto,T. &Sat-
sukawa,Y.1996.Measurementof earthandwaterpressure
actingonthegreatdepthshieldtunnel segments. Geotech-
nical aspects of underground construction in soft ground,
IS-London, Mair &Taylor (eds), Balkema, pp. 613619.
APPENDIX
FigureA1. Comparisonof observedanddesignedvaluesof
earthpressureandmember forces(Hashimotoet al, 2002).
312
Thedouble-o-tubeshieldtunnel inShanghai soil
C. He
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, P.R. China
L. Teng
Shanghai Tunnel Engineering CO. LTD, Shanghai, P.R. China
J.Y.Yan
Shanghai Metro Operation CO. LTD, Shanghai, P.R. China
ABSTRACT: Thehighbuildingdensityof shanghai andthewishtoconstructbiggercitytransportsystemlead
toanobviousconflict. Thedouble-o-tubeshieldtunnelingwasintroducedtosaveundergroundspace. In2002,
Shanghai has built its first double-o-tubeshieldtunnel inNo.8line. Thestress anddisplacement distribution
aroundthedouble-o-tubeshieldtunnelingwasinvestigatedfromanin-situtest.
1 GENERAL INSTRUCTIONS
Shanghai, as one of the biggest city in China, the
density of building is quite high. But with the big
development of urban city, more and more building
and transportation constructions are needed. So the
questionsthat uselimitedspaceof shanghai tosatisfy
all kinds of needs come out these years. In subway
area, the double-o-tube shield tunneling was intro-
ducedtosaveundergroundspace. In2002, Shanghai
has built its first double-o-tubeshieldtunnel inNo.8
line. Figure1showsthedoublecircular shieldwhich
Figure1. Breakthroughof adoublecircular shield.
just finishedonepart of tunnel andbrokethroughthe
workingshaft.
Theoriginof thedoublecircular shieldtunnel may
be traced to 1981, when a basic patent was applied
for inJ apan. Thepatent wasregisteredin1987, anda
horizontal doublecircular shieldtunnel fieldtrial was
performedinthesameyear. In1988, avertical double
circular shieldtunnel trial wasconductedinthefield.
In theconstruction of thenational road No. 54 tun-
nel inHiromiosa, adoublecircular shield, whichwas
designedandmanufacturedby Ishikawajima-Harima
Heavy Industries Co. Ltd, was usedto buildthefirst
doublecircular shieldtunnel intheworldwithlength
of 853.8m(Moriya,Y., 2000).
TheshieldusedinShanghai subway was madein
J apanby Ishikawajima-HarimaHeavy Industries Co.
Ltd and assembled in Shanghai by Shanghai TBM
Company.Thewidthof shieldis11.12m; thediameter
of acircular is6.52m. Thediameter of mostly single
circular shieldsusedinShanghai is6.4m. Thelength
of shieldis 12.76m. Thedistancebetweenthelining
andshieldis11cm.Figure2showsthemaindimension
of double-o-tubeshieldtunnel shieldmachine.
The double-o-tube shield tunneling Method is
appliedfor anearthpressurebalancedshieldmachine
withinterlockingspoke-equippedmultiplecuttersthat
are positioned in the same plane. Adjacent cutters
rotateintheoppositedirections to avoidtouchingor
smashing one another and are thus controlled syn-
chronously. The double circular shield machine is
313
Figure2. Thedimensionof thedoublecircular shield.
equipped with cantilever-arm-type erector to erect
joint andpanel segments, soit provideswideworking
space. The double-o-tube shield tunnel is composed
of 11 prefabricate concrete segments (including the
columnsinthemiddleof thetunnel) per section.
2 COMPARISONBETWEENSINGLEAND
DOUBLE-O-TUBE SHIELDTUNNEL
Compared with the single circular shield, the most
important advantageof thedouble-o-tubeshieldtun-
nel issavingspace. Most of Shanghai metrowasbuilt
under the road, and the space is limited when the
road is relatively narrow. Apparently, thedoublecir-
cular shield tunnel may pass narrower underground
corridors, andtheimpact onnearbystructuresismin-
imized (Moriya, Y., 2000 and R.C. Sterling, 1992).
Meanwhile, thedoublecircular shield tunnel has an
optimizedcross-sectionwithaminimizedsectionarea,
enablingthemost efficient useof undergroundspace.
Moreover, the cross-passage between two circular
shield tunnels is unnecessary for a double circular
shield, and construction risk is thereforeeliminated.
Figure3 shows you theconcept that doublecircular
shieldtunnel cansavespacecomparingwithtwosingle
tunnels.
On the other hand, the double-o-tube shield tun-
nel has its own disadvantages. Thefirst oneis cost.
Chinacanproducethesinglecircular shieldnow, but
thedoublecircular shieldismadeinJ apanandassem-
bledinChina. Moreover, theconstructioncost of the
double-o-tubeshieldtunnelingismoreexpensivethan
twosinglecirculartunnels. Secondly, thedoublecircu-
larshieldisnoteasytobecontrolledinsmall curvature
route. Sometimes it needs extra load to balance the
shield. Thirdly, thespeedof double-o-tubeshieldtun-
neling is slower than two single circular tunneling
whichsharedthesameshield. Themainreasonisthat
theliningsfixspendsmoretime.
Figure 3. Comparison between single shield tunnel and
doublecircular shieldtunnel.
Duringthetunnelingprocess, therearethreeimpor-
tant construction parameters: face pressure, speed
and grout volume. The face pressure which the
double-o-tube shield tunneling used is little higher
than which used in the single circular tunneling.
Higher facepressurecanreducethesettlement while
tunnel isfinished, however it wouldinducethehigher
heavebeforetheopeningface.
Thespeedof thedouble-o-tubeshieldtunnelingis
much lower than the single circular tunneling. The
main reason is the lining fix cost more time in the
double-o-tube shield tunneling. And maintaining a
certain lowspeed is efficient way for thesettlement
control. According to theconstruction experienceof
Shanghai, lower speed correspond smaller ground
settlement especiallyfor shallowcover depthcase.
Although grout pressure is more frequently men-
tionedintheoretical andnumerical calculationof the
tunnelingas themechanical parameter, inthepracti-
cal tunnelingworkof Shanghai, grout volumeusedto
present theeffect of grout instead of grout pressure.
Themainreasonisthat thegrout pressureisunstable
and hard to control for real work. Thegrout volume
whichthedouble-o-tubeshieldtunnelingneedismuch
morethanthevolumewhichsinglecircular tunneling
314
Figure4. Typical profileof groundconditioninShanghai.
need. But thegrout volumeis not afixedvalue. It is
dependondifferent soil types, cover depth, grouting
point, and grout material. Even whileall thefactors
aresame, thegroutvolumewouldstill bedifferentfor
different company whichinchargeof tunneling. It is
morebasedontheexperiencebutnottheoretical result.
3 GEOLOGY CONDITION
Since Shanghai is located near the confluent of
Huangpu River and Chang jiang (Yangzi River),
the soil has been progressively constituted by sed-
imentation and ended in a superposition of layers.
The Shanghai soil can be assumed as homogeneous
in most Shanghai land except partial place where
some soil layers are missed. In Shanghai area, the
mucky soil stratum is down to about 30m deep
fromground surface. It is basically saturated fluid-
plastic or soft-plastic clay with low shear strength
(0.0050.01Mpa), highwater content (above40%),
high compressibility (0.51.0MPa1), sensitivity
varying in 45, and evident rheological behavior
(Wang Zhen Xin & Bai Yun, 2004). Thewater table
is rangedfrom0.3m1m. Inthis very soft ground,
deepexcavationsfor constructingundergroundmetro
station and high building basement encounter many
difficulties in environment protection. The general
ground condition of Shanghai Metro Line is shown
inFigure4.
There are mainly two soil types which
double-o-tubeshieldtunnel crossthrough. Oneisthe
Grey muddy clay and Grey clay which arethetypi-
cal soft clay layers. Another situation is in theplace
wheretheGrey muddy clay is missedandGrey clay
is very thin. It means that layer Grey muddy silty
clayalmostconnectswithlayer Greysiltyclay. Sothe
tunnel mainlycrossesthesiltyclay.
Figure5. Gaugeslayout incrosssectionA.
Alongthetunnelingroute, most of thecover depth
varies from6mto 16m. Therearetwo typical cases
accordingtothedifferentcoverdepth.Oneis8mcover
depthrepresentingshallowcover.Another oneis16m
for deepcover. Surfacesettlement duetotunnelingis
significant different betweenthesetwocases.
4 IN-SITUTEST
4.1 Stress distribution in the soil due to the
double-o-tube shield tunneling
Geotechnical engineersdiscussedonatheoretical base
about thesoil stressdistributionduetodouble-o-tube
shield tunneling at the beginning of double circular
methodappliedinShanghai. Theytrytocomparethe
soil stressdistributionduetodouble-o-tubeshieldtun-
nelingwiththeresultsfromsinglecircular tunneling.
Two theories came out. One opinion is that the soil
stress distributionfromthedouble-o-tubeshieldtun-
neling could be assumed as the combination of soil
stressfromtwoseparatesinglecircularitieswhichtheir
locationarecoincidentwithdouble-o-tubeshieldtun-
nel. Other engineers prefer to believe that it should
bemodeled fromabigger equivalent singlecircular
whichsharesamecenter point withthedouble-o-tube
shieldtunnel.
An in-situ test has been done by Shanghai Tun-
nel EngineeringCo. Ltdto investigatethestress and
displacement distribution around the double-o-tube
shield tunneling. There are 4 sections in which the
displacement inspector andstress gaugewereplaced
alongthelineof tunneling.
CrosssectionA wasdesignedforobservingthesoil
stressincrement andwater stressincrement. Figure5
showsthelayout of gaugesoncrosssectionA.
Figure6showsthevertical soil stressincrement in
theplanewhich in 1.5mahead of theopening face.
315
Figure6. Thevertical stressincrementin1.5maheadof the
openingface.
Figure7. Thevertical stressincrementdistributionin1.5m
aheadof theopeningface.
Therearethreesoil stress increment distributionline
in figure 7 with different depth. It is drawn from4
inspectorsat3.02mdeep, 4at6.52mand6at10.78m.
It is different tendency for thesedistributionlines. A
triangleshapewhichthelargest soil stress increment
occurredonthemiddlevertical lineof double-o-tube
shieldtunnel isgot for thetopdistributionlinewhich
neartogroundsurface. Forthebottomdistributionline
whichlies onthesamedepthlevel withtunnel hori-
zontal central line, thestressincrement inthemiddle
point of tunnel is smaller than theonein thecenter
point of twocircles.
Figure 7 shows the soil stress distribution due to
thedouble-o-tubeshieldtunnelingof onesection. The
black line in figure represents the isoline soil stress
aroundthetunnel atthetimewhiletheopeningfaceis
1.5mbehind.
Fromtheresult, thesoil stressdistributioninduced
bytunnelingismoreclosedtoanellipsethanthecom-
binationof twosinglecircularitiesor abigcircularity.
Figure8. Thesettlement troughswithcover depthof 8m.
Figure9. Thesettlement troughswithcover depthof 16m.
Andtheeffectiveinfluencewidthfromdouble-o-tube
shieldtunnelingismostly2.5timestowidthof tunnel.
4.2 Surface settlement due to the double-o-tube
shield tunneling
Therequirement of surfacedisplacement dueto tun-
neling is +13cmin Shanghai. In thebeginning
of the double-o-tube shield tunneling, the settle-
ment was surprisingly high. Many people suspected
this tunneling method at the beginning of double-
o-tube shield tunnel applying, however after some
projects; theengineers acquiretheexperienceof the
double-o-tube shield tunneling, the settlement has
beenunder control inmost conditions.
Figure8 and Figure9 givethesurfacesettlement
troughduetotunnelingintwocases. Andeachfigure
showstwosoil typeswhichtunnel crossthrough.
316
Figure 10. The plan view and transverse section of the
double-o-tubeshieldtunnel crossingbuildingA.
Itishardtoobtainequal volumelossforeachcircle
intheadvanceof tunneling, especiallyinthecurve. So
fromthesetwofigures, it isquitenormal phenomena
that thesettlement troughis asymmetry tothecenter
of shield.
Concerningtheinfluenceof soil type,themaximum
settlementof siltyclayishigher thanthevalueof clay.
Meanwhilethesettlementtroughof clayiswider. The
mainreasonisthat theinfluenceareainthesoft clay
is lager thantheoneinsilty clay withthesamecon-
structionconditions, not only inthecrosssectionbut
alsointhelongitudinal direction.
Comparing the data in different cover depth, the
maximumsettlement of 16mcover depthislessthan
theoneof 8mcover depth. Inthesametime, theset-
tlement trough of 16mis wider than theoneof 8m
cover depth. Thisphenomenoniscoincidingwiththe
Peckformula.
5 DOUBLE-O-TUBE SHIELDTUNNEL
CROSSINGBUILDING
Here is an interesting project during the
double-o-tube shield tunneling. There are three key
constructionphaseof thisdouble-o-tubeshieldtunnel
line. The shield need across three key section- pipe
channel, five-floor buildingsandshallowcover depth
Figure11. Themeasurement resultsof buildingsettlement
after tunnel crossing.
area. Thecurrent paper will represent thesecondkey
sectionwhichtunnel crossingafivestoriesbuilding.
Figure10shows theplaneviewof threekey con-
structionphasesandthecrosssectionof secondone.
The distance of the building and the tunnel is very
close; it ranged from1.2mto 1.0m. It is great risk
that thetunnel crossesabuildingwithsuchclosedis-
tance. Especiallythisbuildingwithraftfoundationhas
already had 0.3%inclination. According to Chinese
foundation code, themaximuminclination of build-
ing should be controlled under 0.4%. Meanwhile it
has beenmentionedbeforethat therearenot enough
experiences of the double-o-tube shield tunneling
comparingwiththesinglecircular tunneling.
Shanghai Tunnel Engineering Co. Ltd take series
actions to control the ground settlement and reduce
theinfluenceof tunnelingtobuildingraft foundation.
Intheprocessof tunneling, thefacepressureremains
high level. Thegrout volumehappened in thetail is
larger than thevaluewhich used normally. Thetun-
nelingspeedwasmaintainedat verylowinthewhole
process. Moreover, extragroutwasusedwhilethetail
passed.Withall theseconstructionmethod, itresultsin
that thesettlement andtheinclinationof thebuilding
wascontrolledwell.
Therewere6 displacement inspectors installed in
thefoundationof thebuilding.Thefinal measurement
dataweregiveninFigure11.
The blue number shows the location of displace-
ment inspectors. The red number is the settlement
of raft foundationrespectively correspondingtoeach
inspectors and the unit is centimeter. The measure-
mentof settlementrangedfrom0.32.2mm. Itmeans
thatthetunnelingissuccessful.Thebuildingisalmost
intact andnochangecanbedetectedbyeyes.
6 CONCLUSIONS
Thedouble-o-tubeshieldtunnelinghas beenapplied
in Shanghai successfully. The settlement can be
317
controlled in a very low value. The construction is
morebasedontheexperiencebutnottheoretical result.
Furtheringresearchneedtobedonetogetatheoretical
guaranteefor futurework.
ACKNOWLEDGEMENTS
The work described in this paper was part of the
project funded by Shanghai Tunnel Engineering Co.
Ltd. Theauthors wouldliketo acknowledgethecol-
laboration throughout the project of Shanghai TBM
Company. Particular thanksareduetoTangxuanand
Wuhuiming.
REFERENCES
Moriya,Y., 2000. Special shieldtunnelingmethodsinJ apan.
In: Proceedings of the International Conference on Tun-
nels and Underground Structures, Singapore, pp. 24925.
Shanghai Tunnel EngineeringCo. Ltd., 2006. The report of
double circular tunneling (1), Sept.
Sterling, R.C. 1992. Developmentsinexcavationtechnology,
acomparisonof J apan, theUSandEurope,Tunneling and
Underground Space Technology. 7(3), pp. 221235.
Wang, Z.X. & Bai, Y. 2004. Urbansoft groundtunnelingin
China experiences fromShanghai and other cities,
Proceedings of World Tunnel Congress and 13th ITA,
Singapore, pp. 2227.
318
Frozensoil propertiesfor crosspassageconstructionin
Shanghai YangtzeRiverTunnel
X.D. Hu
A.R. Pi
Department of Civil Engineering, Anhui Institute of Architecture & Industry, Hefei, P.R. China
ABSTRACT: Eightcrosspassagesbetweenthetwotubesof Shanghai YangtzeRiverTunnel areconstructedby
Artificial GroundFreezingMethod. Theformationsaroundthetunnel arecharacterizedwithsalinesoil, which
raisessuchdisadvantagesasfreezing-point depression, freezingphaselengthening, thegrowthratedecreaseof
frozenbodiesandthestrengthlossof thefrozenground. Inorder for successful construction, thepropertiesof
theartificial frozensoils aremadeout by test, suchas salinity, freezingpoint, uniaxial compressivestrength,
thermal conductivity, frost heaveratio, thawconsolidationratioandother relatedparameters. Basedonthetest
results somesuggestions aregiven for freezing schemedesign, thecross passages construction and freezing
processmonitoring.
1 INTRODUCTION
Asinmany other tunnel projects(Crippaet al. 2006,
Huetal.2006&Zhaoetal.2005),eightcrosspassages
betweenthetwotubesof Shanghai YangtzeRiverTun-
nel (seeFigure1) areconstructedbyartificial ground
freezing method. Shanghai Yangtze River Tunnel is
situatedattheestuaryof theYangtzeRiver, wheresalt-
water intrusionoccurs frequently (Shenet al. 2003).
It is supposedthat thesalineconcentrationof soil at
theriverbedishigher thanusual, thusthehigher risk
inconstructingthecrosspassagesbyartificial ground
Figure1. Layout of thecrosspassagesof Shanghai YangtzeRiverTunnel.
freezingmethod. Theproperties of frozensalinesoil
havearemarkablechange(Romanet al. 2004) which
isdisadvantageoustotheconstruction.Asaresult, itis
of great necessitytoproceedaresearchof theproper-
tiesof thesalinesoil suchasthesalinity, thefreezing
pointof thesalinesoil andthefrozensoil strength, etc.
Theresearch concentrated on thebasic geotechnical
parameters, thephysical mechanicparametersandthe
thermodynamicparametersof thedesignedsoil layers,
includinggraymuckyclay([4]), grayclay([5]1), gray
clayeysilt([5]2), graysiltyclay([5]3) andgrayclayey
silt ([7]1-1).
319
Table1. Resultsof geotechnical tests.
Water Permeability
Soil layer Content Density coefficient Voidratio Plasticlimit Liquidlimit Salinity Chlorinity
(undisturbed) w (%) (g/cm
3
) K (cm/s) e
0
w
P
(%) w
I
(%) () ()
[4] 44.02 1.69 3.0510
7
1.40 40.78 23.72 18.370 10.169
[5]
1
38.21 1.77 2.1510
7
1.07 33.97 21.68 17.392 9.627
[5]
2
29.96 1.80 8.5510
5
0.94 30.42 19.92 13.228 7.322
[5]
3
31.44 1.78 1.9910
6
1.01 36.73 21.17 8.039 4.450
[7]
11
26.83 1.86 7.9010
4
0.85 36.02 20.35 14.744 8.161
2 TESTS
2.1 Requirement and standard
Thetestsareproceededaccordingtothestandardsof
PeoplesRepublicof Chinaasfollows:
Standardfor soil test method(GB/T50123-1999)
Coal Industrial Standard(MT/T593-1996)
2.2 Content
2.2.1 Basic geotechnical tests
The basic geotechnical tests include test of density,
water content, permeability coefficient, void ratio,
plasticlimit, liquidlimit, salinityandchlorinity.
2.2.2 Frozen soil tests
Frozensoil testsinclude:
1. Uniaxial compressivestrengthtest
Uniaxial compressivestrengthtestsareconducted
at the temperature of 8
C, 15
C, 20
C and
25
C respectively to achieve the uniaxial com-

pressivestrength, modulus of elasticity and Pois-
sonsratio.
2. Frost heavetest
Frost heave tests are conducted to achieve frost
heaveratioandfrost heaveforce.
3. Thawconsolidationtest
4. Thermal conductivitytest
5. Freezingpoint test
6. Specificheat capacitytest
3 RESULTS
3.1 Basic geotechnical tests
Theresultsof basicgeotechnical testsarepresentedin
Table1.
3.2 Frozen soil tests
3.2.1 Uniaxial compressive strength test
(1) Uniaxial compressivestrength
Table 2 shows the result of uniaxial compres-
sion test, which indicates the fact that the uniaxial
compressive strength of the frozen soil rises when
Table2. Resultsof uniaxial compressivestrength.
Soil layer 8
C 15
C 20
C 25
C
[4] 3.14 4.07 5.29 5.94
[5]1 3.19 4.82 5.47 6.58
[5]2 4.18 5.26 6.17 7.07
[5]3 3.07 3.95 5.16 6.60
[7]
11
4.29 5.44 6.55 7.61
Table 3. The relationship between uniaxial compressive
strength (MPa) andtemperatureT (
C).
Soil layer Linear fittedrelationship
[4] =0.1708T +1.706 R
2
=0.9844
[5]
1
=0.1943T +1.7118 R
2
=0.9901
[5]
2
=0.1704T +2.7725 R
2
=0.9982
[5]
3
=0.2073T +1.1713 R
2
=0.9641
[7]
11
=0.1965T +2.6328 R
2
=0.9949
temperature falls. Furthermore, these two variables
obey a favorable linear relationship. The linear fit-
tedcorrelationbetweenuniaxial compressivestrength
and temperature T according to the test data is
presentedinTable3andFigure2.
(2) Modulusof elasticity
The result of the test indicates that, generally,
modulus of elasticity of the frozen soil rises when
temperaturefalls. Dataacquiredfromthetestisshown
inTable4. Theexponential correlationbetweenmod-
ulusE of elasticityandtemperatureT couldbefitted
informof
as presentedinFigure3andthecoefficients a andb
areshowninTable5.
(3) Poissonsratio
The result of Poissons ratio is shown inTable 6,
which demonstrates that when temperaturefalls, the
Poissons ratio of the frozen soil tends to decline,
thoughthistrendisnot remarkable.
320
2
3
4
5
6
7
8
-30 -25 -20 -15 -10 -5 0
Temperature/C
S
t
r
e
n
g
t
h

/
M
P
a
[4]
[5]1
[5]2
[5]3
[7]1-1
Figure 2. The relationship between uniaxial compressive
strengthandtemperature.
Table4. Result of modulusof elasticity.
Modulusof elasticity/MPa
Soil layer 8
C 15
C 20
C 25
C
[4] 53.1 78.7 146.5 246.8
[5]1 58.2 97.9 165.2 269.0
[5]2 50.7 82.4 155.7 273.5
[5]3 57.1 93.2 161.1 306.1
[7]
11
79.2 155.7 238.1 401.0
Table5. Thecoefficientsa andb intherelationshipbetween
modulusE (MPa) of elasticityandtemperatureT (
C).
Exponential fittedrelationship
Soil layer a b R
2
[4] 23.173 0.0921 0.9726
[5]
1
26.975 0.0907 0.9939
[5]
2
20.923 0.1005 0.9824
[5]
3
23.848 0.0985 0.9796
[7]
11
37.176 0.0945 0.9989
3.2.2 Frost heave test
Frost heave ratio test: frost heave test without axial
confinement so that thespecimencanexpandfreely.
Therelationshipbetweenaxial displacementandtime
is measured at the time according to code require-
ment. Themaximumof frost heave
max
shouldalso
berecorded. Frost heaveratioof thespecimenmeans
theratiobetweenthemaximumof frostheave
max
and
initial lengthof thespecimen.
0
50
100
150
200
250
300
350
400
450
-30 -25 -20 -15 -10 -5 0
Temperature/C
M
o
d
u
l
u
s

o
f

e
l
a
s
t
i
c
i
t
y

/
M
P
a
[4]
[5]1
[5]2
[5]3
[7]1-1
Figure 3. The relationship between modulus of elasticity
andtemperature.
Table6. Result of Poissonsratio.
Poissonsratio
Soil layer 8
C 15
C 20
C 25
C
[4] 0.281 0.263 0.230 0.214
[5]1 0.277 0.251 0.229 0.204
[5]2 0.298 0.269 0.248 0.221
[5]3 0.259 0.242 0.217 0.206
[7]
11
0.266 0.244 0.218 0.180
Table7. Theresultsof frostheaveforceandfrostheaveratio
tests.
Frost heaveforce Frost heaveratio
Soil layer (MPa) (%)
[4] 0.66 7.93
[5]1 0.78 6.84
[5]2 0.84 7.26
[5]3 0.71 6.35
[7]
11
0.69 6.48
Frost heave force test: frost heave test proceeded
withdisplacementconfinement. Lengthwiseconfine-
ment isappliedtotheupper endof thespecimenand
then frost heave force is measured by load sensor.
Therelationshipbetweenfrostheaveforceandtimeis
recordedduringthewholeprocesstogetthemaximum
of frost heaveforce
max
.
Themaximumof frost heaveratioandfrost heave
forceof eachsoil layer ispresentedinTable7.
321
Table8. Theresult of thethawconsolidationratio.
Soil layer Thawconsolidationratio(%)
[4] 6.83
[5]1 7.09
[5]2 7.76
[5]3 7.25
[7]
11
6.05
Table9. Theresult of thethermal conductivity.
Thethermal conductivity(W/(mK))
Soil layer positive negative
[4] 1.39 1.62
[5]1 1.43 1.75
[5]2 1.47 1.68
[5]3 1.42 1.74
[7]
11
1.54 1.81
3.2.3 Thaw consolidation test
The temperature of the specimen is 8
C, the hot
end temperature is 402
C. The dimension of the

specimenis +79.840mm. Thethawconsolidation
ratioisdeterminedasfollows:
where a0=thaw consolidation ratio (%); Lh0=
quantityof thawconsolidation(mm); andh0=initial
height of thespecimen(mm).
The result of thaw consolidation test of each soil
layer isshowninTable8.
3.2.4 Thermal conductivity test
Theresult of thethermal conductivity ispresentedin
Table9. Inthistable, positiveisfor theunfrozensoil
andnegativefor thefrozensoil.
3.2.5 Freezing point test
Generally speaking, thecrystallizationfromliquidto
solidof thewater inthesoil undergoes threestages:
First a small group of molecule is formed which is
calledthecrystallizationcenter or growthpoint, then
it growup to abigger crumb called crystal nucleus,
whicheventuallydevelopstoicecrystal.Thetempera-
tureof theformationof theicecrystal iscalledfreezing
pointor icepoint. Thecrystallizationcenter isformed
at acertaintemperaturebelowthefreezingpoint. As
a result, theformation of frozen soil is consisted of
fourstages,i.e.supercooling,jump,invariablenessand
degradation.Atthestageof jump, theelectricpotential
will suddenlyreduce, andthenbecomestableatacer-
tainnumber at thetemperaturewhenfreezingbegins.
Table10. Theresult of freezingpoint.
Soil layer Freezingpoint (
C)
[4] 2.5
[5]1 2.3
[5]2 1.7
[5]3 2.1
[7]
11
1.7
Table11. Result of undisturbedsoil specificheat capacity.
Soil layer Specificheat capacity/(gK)
[4] 1.56
[5]1 1.62
[5]2 1.69
[5]3 1.65
[7]
11
1.51
Accordingtothisprinciple, thefreezingpoint canbe
calculatedasfollows:
whereT=freezingpoint(
C);V=thestabilizedvalue
after the jump of the hot electric potential (j);
andK =thedemarcationcoefficientof thermoelectric
couple(
C/j).
Theresult of thefreezingpoint of eachsoil layer is
presentedintable10.
3.2.6 Specific heat capacity test
Thespecific heat capacity test is performedwiththe
specificheat capacitytestingdeviceDTBR-01. Intro-
ducecoolantof thesametemperatureandvolumeinto
bothsidesof theequipment. Temperaturesensorsare
previouslyputintocoolant.Thespecimenwithtemper-
aturesensor embeddedisputintooneside.According
tothelawof conservationof energy, theenergyforone
degreechangingcanbedeterminedbytherelationship
betweenthethreevaluesof temperature. Theresult is
presentedintable11.
4 DISCUSSIONANDSUGGESTIONS
4.1 Salinity and its influence
Thesalinityof porewaterintheundisturbedsoil ranges
from8.03918.370%(seeTable1). To findout the
salinity differencebetweenthesoils near theseaand
far fromthesea, anadditional testwasmadefor asoil
layer thegray clayey silt ([7]
11
). Intheresult, the
salinityof thespecimenof thelayer intheurbanarea
(at thesiteof therestorationproject of thecollapsed
322
tunnels of Shanghai Metro Line4) is 6.320%, while
that of thesamelayer at theestuary is 14.744%(see
Table1). Therefore, aconclusioncouldbedrawnthat
thesalinity of porewater inthestratanear theseais
higher thanthat distant formthesea.
Influenceof salinityonfrozensoil propertiescould
befreezing-pointdepression, freezingphaselengthen-
ing, thegrowthratedecreaseof frozenbodiesandthe
strength loss of thefrozen ground, on which agreat
attentionmust bepaid.
Given lack of enough test data, no clear relation-
shipshavebeenfoundbetweenthesalinityandfrozen
soil strength, frost heaveforce, frost heaveratio, thaw
consolidation ratio and other parameters, as well as
freezing point, but further tests arebeing performed
bytheauthors.
4.2 Freezing-point depression
The freezing point ranges from1.7 to 2.5
C (see
Table10). Itisfoundthatthefreezing-pointdepression
of thesoil attheestuaryisgreaterthanthatdistantfrom
thesea. Froexample, thefreezing-point temperatures
of thegray clayey silt [7]
11
near theseaanddistant
fromtheseaare1.7
C (seeTable10) and1.4
C
(Xiao, etal. 2003), respectively. Ontheotherhand, soil
layersaresortedaccordingtothedescendingorder of
themagnitudeof freezing-pointdepression,i.e.mucky
clay, clay, siltyclayandclayeysilt.
Freezing-point depression causes asmaller thick-
nessof thesoil frozenwall thanthethicknesswhenthe
freezingpointisconsideredas0
C.Attentionmustbe
paiduponthethicknesslossof thefrozensoil wall due
tofreezing-point depression.
4.3 Frozen soil strength
Theuniaxial compressivestrengthof thefrozensoils
and the temperature of the specimen bear a favor-
ablelinear relationship, withregressioncoefficientall
higher than0.96. Meanwhile, themoduli of elasticity
of thefrozensoilshaveanexponential correlationwith
thetemperature,withanaverageregressioncoefficient
of 0.985. Theregressionformulascanbeusedtocal-
culate the strength values at any temperature within
therangeof thetest temperature.
A suggestion hereis that theregression formulas
shouldbeusedto calculatethebearingcapacity and
deformationof thefrozenwall.
4.4 Frost heave
Thefrost heaveforceranges from0.66to 0.84MPa.
The frost heave ratio ranges from 6.35 to 7.93%.
Accordingto Codefor Designof Soil Foundationof
BuildinginFrozenSoil Region(J GJ 118-98), whenthe
frostheaveratioof theclayishigher than6%, thesoil
is categorizedas highly frost-heavingsoil. Therefore
greatemphasisshouldbelaidonthedesignandappli-
cation of the construction. It is suggested to install
asteel reinforcement framearound theopening as a
protection device to prevent the tunnel lining from
excessivedeformationduetothefrost heave.
4.5 Thaw consolidation
Theratioof thethawconsolidationrangesfrom6.05%
to 7.76%. Thethawsettlement of thegroundaround
thecrosspassagescouldberemarkable. Therefore, it
isadvisedtoperformenforcedthawingasaneffective
measuretoreducethawsettlement.
4.6 Thermal properties
Thethermal conductivityforunfrozensoil rangesfrom
1.39 to 1.54W/(mK), and 1.62 to 1.81W/(mK) for
frozen soil. The soils are hard to be frozen, i.e. the
growthrateof thefrozenbody coulddecreaseor the
freezingphasecouldbelengthened, becausethecoef-
ficient of thermal conductivity is unfavorably low. It
is necessary to enhancefreezing process monitoring
andtousethemonitoreddatatojudgethesizeof the
frozen wall instead of rude estimate of the freezing
phase.
5 CONCLUSIONS
Althoughthestrengthof thefrozensoils at theestu-
ary of the Yangtze River is fairly high, attention
must bepaid upon thedisadvantageous influenceof
freezing-point depressiondueto thehighsalinity on
thecalculationof thethicknessof thefrozensoil walls
for constructingthecrosspassages.
Further study should be carried out to find out
the influence of salinity of pore water in the saline
soils uponthegeotechnical andphysical-mechanical
propertiesdiscussedbefore.
ACKNOWLEDGEMENTS
The work were supported by the National Natu-
ral Science Foundation of China (No. 50578120),
the National High Technology Research and Devel-
opment Program of China (863 Program) (No.
2006AA11Z118) and Shanghai Leading Academic
DisciplineProject(Project Number: B308).
REFERENCES
CrippaC.&ManasseroV.2006.Artificial GroundFreezingat
Sophiaspoortunnel (TheNetherlands) Freezingparam-
eters: Data acquisition and processing. GeoCongress
2006: Geotechnical Engineering in the Information Tech-
nology Age, Atlanta, Feb 26 Mar 1, 2006. Reston, VA:
ASCE.
323
Heilongjiang Province Academy of Cold Area Building
Research. 1999. Code for Design of Soil Foundation of
Building in Frozen Soil Region (JGJ118-98). Beijing:
ChinaArchitecture& BuildingPress.
Hu X.D. & Chen R. 2006. Construction technology of
freezingmethodappliedtocross-passageof double-deck
cross-river road tunnel. Low Temperature architecture
technology. (5). 6466.
RomanL.T., AksenovV.I. &VeretehinaE.G. 2004. Charac-
teristics of influence of the content salts on the frozen
soils strength. Journal of glaciology and geocryology
26(Suppl): 3538.
ShenH.T., Mao Z.C. & ZhuJ.R. 2003. Saltwater Intrusion
in the Changjiang Estuary. Beijing: ChinaOceanPress.
TheMinistryof Coal Industrial P.R.C. 1996. Coal industrial
standard: Physical-mechanical properties test of artificial
frozen soil (MT/T593-1996). Beijing.
The Ministry of Water Resources P.R.C. 1999. Standard
for soil test method (GB/T50123-1999). Beijing: China
PlanningPress.
Xiao Z.H., Hu X.D., Pi A.R. & Liu R.F. 2003. Study on
uniaxial compressivestrengthof Shanghai soilsundersec-
ondary freeze-thawaction. Rock and Soil Mechanics 27
(Suppl): 497500.
ZhaoY.H., Hu X.D., Zhao G.Q. 2005. Experimental study
onartificial frozensoil boundarydetectingduringcross-
passageconstructionintunnel. Progress in Electromag-
netics Research Symposium, Hangzhou, China, 2226
August 2005.Cambridge:TheElectromagneticsAcademy.
324
Theinfluenceof engineering-geological conditionsonconstructionof the
radioactivewastedump
J. Kuzma& L. Hrustinec
Slovak University of Technology, Faculty of Civil Engineering, Bratislava, Slovakia
ABSTRACT: Securestabilityandreliableserviceabilityof radioactivedumpisadifficultengineeringproblem.
Duetodifficult geological formationsdeterminedmainlybygreat compressibility, lowshear strengthof soils,
and high ground water level, or great upward hydrostatic pressurewill thesedemands increase. Influenceof
required reliability and lifespan on the structure of these specific objects is considerable. We are trying to
contributetoaproblemsolvingof thesedifficult andcomplicatedproblemsinsubmittedcontribution.
1 INSTRUCTION
The Republic Radioactive Waste Dump (RRWD)
located about 2kmfromthe area of the Mochovce
Nuclear Plant in south-western Slovakia, Central
Europe, wasbuilttostorewastewithlowandmedium
activityarisingfromoperationanddisposal of nuclear
powerfacilitiesaswell asfromresearchinstitutes, lab-
oratories, hospitals and other institutions engaged in
activities generating nuclear waste. Ensuring stabil-
ity and reliableoperation of RRWD in Mochovceis
achallengingengineeringtask. Thedumpis built in
demanding engineering-geological conditions deter-
minedinparticularbyhighcompressibility, lowshear-
ingstrengthof soil andhighundergroundwater level
and/or high lifting forces exerted in the basement.
Also required reliability and lifetime have a signifi-
cant influenceon layout of thesespecific structures.
This paper discusses a solution for this complex
engineeringtask.
Thetopicof howtosecureareliablestorageof final
processed low and mediumradioactive waste in the
long termis a complex interdisciplinary issue. The
timeaspect isasignificant factor affectingacomplex
approachtothisproblemandspecificperiodsof active
operationandplannedlifetimeof therepositorymust
betakenintoaccount. Therepositoryisbuilt infairly
difficult geological and hydro-geological conditions.
Itconsistsof acomplexof buildingsandtechnological
facilities. In thefirst construction stage, two double
lines of storage boxes with the total surface area of
approx. 112.000m
2
werebuilt. Thegeneral situation
aroundthedumpanditsadjacentterritoryisshownon
Figure1. Sectionof theRRWDisshownonFigure2.
Figure1. Viewof placement of theRepublic Radioactive
WasteDump(RRWD) inMochovce.
Figure2. Sectionof theRRWDinMochovce.
Suchlargestructural work challengingintermsof
operational safetywill haveasignificantimpactonthe
surrounding natural environment during its lifetime.
Extensiveconstruction activities, activeoperation of
therepository andthefinal coveragewill haveasig-
nificant influenceon a changein thebalanced state
325
of therock environment. Thischangeinthebalanced
statewill causedeformationsymptomsattheaffected
territory with a number of structural buildings and
functional (technological) units. Knowing theextent
of themodifiedbasement andconstructionat theter-
ritoryconcernedwill bedecisivetodeterminestability
of theterritoryandfunctionalityof thealreadybuiltor
planneddrainagesystems, engineeringandtechnolog-
ical distributionlines. Thispaper will discussindetail
theissueof repository reliability in terms of assess-
ingthebasement accordingtorequirementsof group
I andII limitstates, i.e. accordingtosafetyandusabil-
ityof thestructural work. A set of geotechnical issues
isrelatedtothedraft constructionproject of thefinal
repository coveragewithits primary functionto pre-
ventwater infiltrationtopremiseshousingthestorage
boxesafter theactiveoperationof thestructural work
until itslifetime.
2 MORPHOLOGICAL, GEOLOGICAL AND
HYDROGEOLOGICAL CONDITIONSAT THE
REPUBLIC DUMP
RepublicRadioactiveWasteDumpislocatedinsouth-
western Slovakia, Central Europe. The terrain is
slightlyfoldedwithanunstablesuperficial waterriver.
Thebasement is madeof sandy limestoneclay with
spotsof dustyfine-grainsandandclaysilt.Thesespots
formlenses separatedfromeachother. Sediments of
Ivanian strata (lower Panonium) with thickness of
11.523.0mwereidentifiedatthenorthernpartof the
repository. They are represented by sandy clay with
thinspotsof dustysandsandisolatedlensesof gravel
sands. Quaternary deposits arerepresentedprimarily
by deluvial rain-washsandclay withthickness upto
3.0 to 4.0m. The basement directly under the first
double-line of boxes is made of residual quaternary
deposits (clays), especially in themiddleof thebox
structureandunder thebottomdoubleline. Panonian
deposits(clays, sands) arefoundunderneath.
Hydrogeological conditionsintheRRWDbasement
andinitsdirect vicinityareoneof thecrucial factors
withregardto operational safety of thedump. It can
besaidthat thehydrogeological conditionsat thesite
are fairly complicated due to the complex nature of
thegeological environmentwithflowsof underground
water. Surveys had done show existence of several
continuous water-saturated aquifers of underground
water. Thehighest mainaquifer (H) isanaquifer with
a free water level, fairly sensitive to precipitation.
Other twoaquifersaremorepermeable, hydraulically
fairlyisolatedsandyspotsindeeper zonesof theSar-
matian basement. Theunderground water has lifting
characteristics and its piezometric level reaches 1.0
to 2.0mabove the free level of the H aquifer. The
original hydrogeological conditionsweresignificantly
Figure 3. Typical cross section of the 1st and 2nd dou-
ble-lineof storageboxesinRRWDinMochovce.
modified by constructional activities. In terms of
assessingthefoundationconditions, theengineering-
geological andhydrogeological conditionsat thearea
of interest canbeassessedasdifficult.
3 CONSTRUCTIONAL ANDTECHNICAL
CHARACTERISTICSOF THE REPOSITORY
At present, the Republic Radioactive Waste Dump
consists of two double-lines of storage boxes made
from reinforced concrete. Four boxes are grouped
into one dilatation block with layout dimensions of
18.66.0mandheightof 5.5m, placedonacommon
foundationslab0.6mthick. Therearefivedilatation
blocks in one line. An expansion gap between the
blocks is planned to be 50.0mmthick. The storage
boxes are covered by 0.5mthick panels fromrein-
forced concrete. A systemof two drainages (control
drainageandmonitoreddrainage) is built at thebot-
tomof the storage boxes and under the foundation.
The main goal of the dump drainage system is to
ensureinspectionandsaferemoval of potential leaks
of underground and surface water contaminated by
radioactivesubstancesfromtheinnerspaceof thestor-
ageboxesintothebasementand/or fromthebasement
intothestorageboxes. Sidewallsandthebottomof the
monolithicstorageboxesfromreinforcedconcreteare
protectedbyalayer of clayishsoil.A 1.0mthicklayer
of compacted clayish soil is located under thetanks
toensurewaterproof basement for thestorageboxes.
Placed on this layer is a 0.6mthick layer of sandy
gravel. Sidewalls areprotectedby a3.5mthick ver-
tical claysealinglinkedwithahorizontal sealingsoil
layer, formingacompactunit theso-calledclayseal-
ingtub.Thedouble-linesareprotectedagainstclimatic
effects by amobilesteel hall whenthestorageboxes
arefilledwithwaste. Consideringthegeneral reliabil-
ity of RRWD, thestructures of thestorageboxes are
themost important and decisivebuilding structures.
At thesametime, thesestructureswill facethehigh-
est demandsduetoeffectsof forces, deformationand
radiation. A typical cross andlongitudinal sectionof
the1st and2nddouble-lineisshownonFigure3and
Figure4.
326
Figure4. Typical longitudinal sectionof the1stdouble-line
of storageboxesinRRWDinMochovce.
4 ASSESSINGSTABILITY INRRWD
ACCORDINGTOLIMIT STATES
In order to make a general assessment of RRWD
safety and usability, it is necessary to examineaset
of geotechnical issuespredeterminedbythechalleng-
ingnatureof buildingstructures at thedumpandby
complex engineering-geological conditions. Interms
of compliancewithrequirements accordingto group
I andII of limit states, solutionsfor theproblemscan
bedividedintotwomainsetsasfollows:
Stability calculations to assess overall stability of
RRWDandtheadjacentrockenvironmentaffected
by theconstructioninspecific stagesof construc-
tion, operationandexpectedlifetimeof therepos-
itory, includingbasement for thestorageboxes.
Deformation calculations to assess final and
uneven settlement of decisive building structures
of thedump, takingintoaccount specificstagesof
construction,operationandexpectedlifetimeof the
dump.
Tosolvethefirstsetof problemsrelatedtosecurity
andoperativecapacityof RRWDinMochovceinevery
construction stage and after the construction, it was
necessary tomakeaseriesof stability calculationsto
determine:
Stabilityof the1st and2nddouble-line,
Stabilityof adjacent slopes,
Bearingcapacityof reinforcedconcreteboxes.
Thesecondsetof problemsdealswithhowtoensure
usability andsmoothoperationof thebuildingstruc-
turesandtechnological unitsat thedumpintermsof
size of the final and uneven settlement of the stor-
ageboxes.A requirementtoassesshorizontal shiftsof
storageboxwallsatplacesof expansiongaps, planned
to be 50.0mmwide, stems out fromthe structural-
technical solution of decisive bearing structures,
includingbearingwallsof thestorageboxes.Consider-
ingtheheightof storageboxwalls,evenasmall uneven
settlement or inclinationof toughreinforcedconcrete
structuresof thestorageboxescancausetheirdamage,
resultinginalossof theirplannedfunction.Theextent
of settlement andunevensettlement of therepository
basement must be known to assess also function of
thedrainageand sealing systems located in revision
andinspectionshafts. Incaseof drainagesystems, this
involvesespeciallytherequiredgradient neededfor a
gravitational drainageof leakedwater. Thestructural
designof thedumpshowsaconcentratedintensityof
loadinthegeometric centreof thedouble-lines with
storageboxes, so themethod and systemof placing
fiber-concretecontainersintothestorageboxesof the
dumpisveryimportant.
Theeffects of uneven settlement of thebasement
resultinformationof local depressionsinthedrainage
system, so their thickness must follow the extent of
theuneven settlement of thebasement to ensurethe
required gradient for gravitational water drainagein
thelongterm. Occurrenceof local depressionswould
result inhigher andpermanent hydraulicloadof spe-
cific sealing layers in the dump, affecting physical
propertiesof thesoil inthesealinglayer, inparticular
thestateof consistencyandthenstrengthanddeforma-
tion properties. Such changes wouldresult in higher
permeability of thesealingtub, higher compressibil-
ity andlower bearingcapacity of theRRWD subsoil.
Localizationof maximumtensionsanddeformations
given by effects of external and deforming load can
indicatepotential defect areasthat must betakeninto
account inthedesignof structures of specific layers
for thefinal coverage.
Themost demandingandthemost important task
in the process of addressing these problems is to
transform actual material properties of the natural
rock toanidealizedcalculationmodel that must real-
istically depict the actual behavior of the assessed
structures. Itisthereforenecessarytopayproperatten-
tiontothistask. Thisissueiscomplicatedevenmore
by significant non-homogeneity of the natural rock
environment, but primarily by the fact that material
properties of fine-grain soil in the subsoil are very
variable in time and tension. It is very difficult to
definecomplex constitutional relationships incalcu-
lation models that would truly describe the actual
processes inthesoil massif resultingfromeffects of
theload.Thistaskcanbesimplifiedtoacertaindegree
byareasonableidealization(simplification) of actual
soil properties, respectingtheinevitablerequirement
to preservethephysical essence. It is thereforenec-
essary todeterminephysical, strengthanddeforming
propertiesof thebasementsoil andof thesoil incorpo-
ratedinthesealingtubbasedonstandardlaboratory
testonintegral andundamagedsamplesinaccordance
withapplicablestandards.
To definecalculationmodels andlimit conditions
of specifictasks, itisnecessarytodefinethefollowing
input data:
Spatial arrangement and geometrical shape of
building structures and construction at the area
concerned;
Extent and distribution of permanent, incidental,
extraordinaryloadsandtheir combinations;
327
Figure5. Calculationmodel dividedintofiniteelements.
Physical, strengthanddeformingpropertiesof arti-
ficial buildingmaterials(concreteandreinforced-
concretestructures);
Physical, strengthanddeformingpropertiesof the
natural rockenvironmentattheareaof concernand
of thesoil incorporatedinRRWD.
Inordertoassessgeneral reliabilityof thedump,itis
not permissibletoexceedanylimit stateandtoallow
deformation of any structural unit during the whole
period of active operation of the repository (organi-
zation of filling the storage boxes by fiber-concrete
containers)andduringtheperiodof theplannedand/or
actual lifetime of the dump once it is finally cov-
ered. Specific calculations must therefore depict all
the crucial load states that occur or might occur in
the future. The extent of load on specific building
structures and subsoil in RRWD and effects of the
load on the environment is a very variable quantity
in terms of time. It will depend mostly on the con-
structionprocessappliedfor themaindumpbuilding
structures, schedule of organized filling of the stor-
ageboxes by fiber-concretecontainers during active
operation of the dump and constructional stage of
the final dump coverage. The intensiveness of load
will therefore depend on a progressive increase in
permanent, long-termandshort-term, incidental and
extraordinaryloadandtheir combinations.
Mathematical modelingusingthemethodof finite
elements was used to calculate tensions and defor-
mations intheRRWD subsoil. Thespatial model for
the model calculations was developed according to
assumptions of thelinearly elastic half-spacetheory
(Figure5).
Basic parameters of thegeotechnical model arise
fromanengineering-geological survey andstructural
designof thedump. Geotechnical calculationsdefine
Figure6. Isoareas of subsoil settlement for storageboxes
calculatedfor theLoadState2.
four loadstates(LS) describingthefollowingcrucial
stagesof dumpconstructionandoperation:
LS 1 Loadonthesubsoil dueto ownweight of
thereinforced-concretestorageboxes,
LS 2 Loadafter fillingthe1st doublelinewith
wastecontainers,
LS 3 Loadafter fillingthe2nddoublelinewith
wastecontainers,
LS 4 Loadafter closingthe1st and2nddouble
linesandthefinal coverageof thedump.
Resultant settlement (vertical displacement) of the
subsoil underthestorageboxescalculatedfortheLoad
State2isshownonFigure6.
Thefactorof timeisveryimportanttoaddressthese
problems, because the expected lifetime for RRWD
was determined by the Slovak Nuclear Supervision
Office to be 300 years based on the period of the
institutional inspection of the dump. Consolidation
processesinthedumpbasementandprogressiveredis-
tributionof forcesoccurringindependenceontherate
of loadingthefoundationgap(fillingthestorageboxes
by containers) are closely related to the time factor.
Calculationsonforecastsmakeit possibletodesigna
suitablemethodof containerplacement,withfavorable
effects on theoverall averagesettlement and uneven
deformations that will havebeen developed depend-
ing on time in couples of decades. Stability of the
Republic Radioactive Waste dump must be ensured
withrequiredreliability duringthewholetermof its
lifetime.
5 CONCLUSION
Construction of specific buildings intended to store
low and mediumradioactive waste is an extremely
328
challenging task for engineers. Given the demands
and required reliability of storing radioactive waste
inRRWD Mochovce, theissuemust bediscussedin
itscomplexity. Thetimeaspect isasignificant factor
affectingacomplexapproachtothisproblemandspe-
cificperiodsof activeoperationandplannedlifetime
of thedumpmust betakenintoaccount. Toformulate
specificgeotechnical problems,itisnecessarytofocus
primarily onassessment of structural works interms
of safetyandusabilityof buildingstructuresandtech-
nological units. Much emphasis should be given on
definitionsof limit conditionsfor specific taskswith
focusontheir variabilityintime. A correct definition
of interactions between thesubsoil and thebuilding
structuretoensuretherequiredlevel of environmental
protectionisanessential requirement toaddresssuch
acomplexinterdisciplinaryissue.
ACKNOWLEDGEMENT
Thispaperwasdevelopedintermsof agranttaskof the
SlovakRepublicMinistryof EducationNo1/2135/05.
REFERENCES
Kuzma,J.etal.1998.Experimental Assessmentof theRepub-
licRadioactiveWasteDumpinMochovce. Bratislava.
329
Critical ventilationvelocityinlargecross-sectionroadtunnel fire
Z.X. Li, X. Han& K.S. Wang
Shanghai Institute of Disaster Prevention and Relief, Tongji University, Shanghai, P.R. China
ABSTRACT: BasedonCFDmethod, thispaper conductedsimulationanalysisof lorryandcar firesfor large
cross-sectionroadtunnel. Thesmokedistributionfeaturewasdiscussedandthecorrespondingresultsof critical
ventilationvelocity werealso comparedwiththevalues calculatedby relatedempirical formula. It wouldbe
helpful toworkout aventilationsystemstrategyfor theroadtunnel.
1 INTRODUCTION
Recent incidents however have drawn widespread
attention to therisks of fires in road, rail and mass-
transit tunnels (e.g. Kings Cross, Channel Tunnel,
Mont Blanc, Tauern, Gotthard). The incidents have
resultedincasualtiesaswell aslargedirectandindirect
economic damage. Thecontrol of smokeflowduring
a road tunnel fire is often an important part of fire
safety measure. On theother hand, CFD (Computa-
tional Fluid Dynamics) simulations are increasingly
used to estimatetheeffects of fires in road tunnels.
However, littleworkhasbeendonetoanalyzecompre-
hensivefirecharacteristics of thelargecross-section
road tunnel. On the basis of CFD method and with
thehelpof FDS simulationsoftware, this paper con-
ducted simulation analysis of lorry and car fires for
large cross-section road tunnel. Two fire scenarios
were constructed and five different kinds of longi-
tudinal ventilation conditions were considered. The
smokedistributionfeaturewasdiscussedandthecor-
respondingresultsof critical ventilationvelocitywere
also compared with the values calculated by related
empirical formula. It wouldbehelpful to work out a
ventilationsystemstrategyfor theroadtunnel.
2 CFDSIMULATIONSCENARIOS
2.1 Brief introduction of FDS
Inthesesimulations, FDS(FireDynamicsSimulator)
4.06 which was released by NIST (National Insti-
tute of Standards and Technology, USA) was used.
FDSisaComputational FluidDynamics(CFD)model
withLES(LargeEddySimulation) of fire-drivenfluid
flow. The model solves numerically a form of the
Navier-Stokes equations appropriate for low-speed,
Figure1. Cross-sectionof roadtunnel.
thermally-drivenflowwithanemphasisonsmokeand
heat transport fromfires. The partial derivatives of
theconservation equations of mass, momentumand
energyareapproximatedasfinitedifferences, andthe
solution is updated in time on a three-dimensional,
rectilinear grid. Thermal radiationiscomputedusing
afinitevolumetechniqueonthesamegridastheflow
solver.
2.2 The design of the fire scenario
Asfor largecross-sectionroadtunnel, thefiresmoke
distribution feature under different ventilation sit-
uation was simulated. One road tunnel which has
100mlength and 15mdiameter was constructed as
331
Figure2. CFDmodel inscenario1.
Figure3. CFDmodel inscenario2.
simulationmodel, asshowninFigure1, Figure2and
Figure3. Onelorrywhichdesignedasfiresourcewas
located at 37mfromtheupstreamventilation cross-
section and 63m from the downstream ventilation
cross-section, asshowninFigure2.Thelorrywassup-
posedtobeonfirefirst, andtheheatamountreleased
by thelorry wouldemblazeacar near thelorry. Two
firescenarios wereset up, including: (1) scenario 1:
thecarattherightof thelorry; (2)scenario2: thecarin
front of thelorry. Bothscenariosinvolvedfivediffer-
ent kinds of longitudinal ventilation conditions. The
ventilationvelocitiesweresetas2m/s, 2.5m/s, 3m/s,
3.5m/s, and4m/srespectively.
Duringthesimulation, thelorry wasburningfirst.
Thenat acertaintime, whentheheat amount thecar
receivedexceeding16kW/m
2
, thecar wouldbeginto
burn.Theheatreleaserateof thelorryandthecarwere
20MWand5MWrespectively. Thetotal heat release
ratewasshowninFigure4andFigure5separately.
3 ANALYSISOF CFDSIMULATIONRESULTS
3.1 The brief analysis of the simulation results
Since the heat release rates of two scenarios were
almost of samevalue, thesimulation results demon-
strated that therelativeposition of thelorry and the
car didnt distinctively affect the smoke distribution
feature. As for the same environmental conditions,
there was not much difference between these two
Figure4. Heat releaserateof firesourceinscenario1.
Figure5. Heat releaserateof firesourceinscenario2.
Figure6. Thesmokedistributionfeatureunder ventilation
velocitiesof 2m/s, 2.5m/s, 3m/sand3.5m/srespectivelyin
scenario2.
fire scenarios, as shown in Figure 6, Figure 7 and
Figure 8 respectively. For example, under the same
longitudinal ventilationcondition, thesmokedensity
atthelocationof 10mupstreamthefiresourcewasof
similar distributiontype.
332
Figure7. Smokedensity at thelocationof 10mupstream
thefiresourceinscenario1.
Thesimulationresultsillustratedthat theroadtun-
nel fireyieldedhighproductionof smoke. Asaresult
of firetemperature, thesmokewouldhavealower rel-
ative density than the surrounding air in the tunnel
andthereforerosetotheceilingof thetunnel. Conse-
quentlythesmokewouldremoveitself awayfromthe
firesourceinalayer alongtheceiling. Alongthebot-
tomthecoldairwouldbesuckedtowardsthefire.This
createdtwoflowinglayers,i.e.,thehotsmokelayerand
thecoldsmoke-freelayer of air.Thephenomenonwas
calledstratification.
The hot smoke layer gradually formed as the hot
smoke curled upwards. With continuous yielding of
smoke from the fire source, the hot smoke layer
swelledrapidly andspreadtowards bothsides of the
tunnel. Meanwhile, thecoldair fromthelower part of
thetunnel flewtothefiresource. Inthisway, thesym-
metry re-circulated air was formed on both sides of
thefirefield. When thelongitudinal ventilation was
started, the smoke on both sides of the fire source
emergedasymmetry. If theventilationvelocityisquite
small, it couldnt prevent thesmokefromdiffusingin
thedirection oppositeto theventilation stream. This
situation, namelysmokeflowsagainsttheventilation,
isadversetothepreventionof thesmokespread(the
hotsmokemaybeemblazethetrafficintheupper side
of thefire) andwouldbeharmful tothesafety of the
firemen.Inordertopreventthebackflowof thesmoke,
thelongitudinal ventilationvelocityshouldbegreater
thanthecritical one.
As ventilation velocities were set as 2m/s and
3m/s, and at the location of 10m upstream the
fire source, the maximum smoke density would
approached 250mg/m
3
and 280mg/m
3
separately.
While the ventilation velocity was set as 4m/s, the
correspondingsmokedensityalmostkeptzero.Itobvi-
ously indicated that the critical ventilation velocity
went between3m/sand4m/s.
Figure8. Smokedensity at thelocationof 10mupstream
thefiresourceinscenario2.
3.2 Obtain of the critical ventilation velocity
Oka andAtkinson had proposed a method to calcu-
latethecritical velocity, whichwas to get thesmoke
backflowdistanceindifferent ventilationthroughthe
detectors first. Then the value of the distance was
placed in the reference frame for backflow distance
and ventilation velocity. To joint these points in the
reference frame and extend the line to the y-axis,
they-axis valueof thepoint of intersection was just
thecritical velocity. Thispaper wouldalsoobtainthe
critical ventilationvelocityinthisway.
Thespreadof thesmokeundertheventilationveloc-
itiesof 2m/s,2.5m/s,3m/sand3.5m/swerepresented
inabovementionedfigures. Thesefigurescouldgive
outtherelateddistancevalueswhichthesmokespread
upstreamthefiresource. Basedonthedetectors, we
could get thebackflowdistancein four situations of
120m, 80m, 48mand 10m. These numerical val-
ues wereput upin thereferenceframefor backflow
distance-ventilation velocity. J ointing thepoints and
extending the line in accordance with the method
referredpreviously, thenthecritical ventilationveloc-
itycouldbeobtainedas3.65m/s.
3.3 Comparative analysis of the critical ventilation
velocity
OkaandAtkinsonhadsetupaformulafor calculation
of tunnel critical ventilationvelocityasfollows:
333
Figure9. Critical ventilationvelocityforlargecross-section
roadtunnel.
whereQ=theheatreleaserate, kw; V
critical
=thecrit-
ical velocity, m/s; Q
=dimensionless heat release

rate, and V
=the dimensionless critical velocity,
0
=ambient air density, kg/m
3
; CP=specific heat
capacity, kJ kg
1
K
1
; T
0
=ambient temperature, K;
g=accelerationduetogravity, m/s
2
; H=heightfrom
thesurfaceof firesourceto tunnel ceiling, m. Using
these values we could obtain the critical ventilation
velocity of the tunnel as 3.208m/s. Comparing the
simulation value with the one given out by formula
(1), therelativeerror was12.12%.
Wu and Bakar further amended the formula
(1). They considered the influence of the breadth
of the tunnel and replaced the height of the tunnel
withthehydraulictunnel heightH, whichwasdefined
as theratio of 4times thecross-sectional areato the
tunnel wetted perimeter. Thus, thenew formulawas
illustratedasfollows:
where Q
=dimensionless heat release; V
=
dimensionless critical ventilation velocity, other
parameter were of the same meaning as formula
(1). Through formula (5), we could obtain the criti-
cal ventilationvelocity as 2.893m/s. Comparingthis
valuewith thesimulation one, therelativeerror was
20.7%.
4 CONCLUSION
Thesimulationresultsdemonstratedthat specifyinga
critical ventilationvelocitywashelpful topreventback
layeringof smokeduringatunnel fire. Hence, it was
veryimportantfor smokemanagementof largecross-
section road tunnel. In order to find cost-effective
methods to upgradefiresafety inexistingtunnels or
tunnelstobebuilt, thepractical testsaswell assimula-
tionanalysisshouldbefurther carriedout. Hence, the
correspondingperformance-basedcalculationmethod
couldbeput upto improvethecomputationof criti-
cal ventilationvelocity. Meanwhile, moreinfluencing
factors need to be considered, such as the impact
of roadtunnel physical characteristics andthetraffic
capacity, etc.
ACKNOWLEDGMENTS
The support of the Natural Science Foundation of
China(GrantNo. 50678124) isgratefullyappreciated.
REFERENCES
Bettis, RJ., J agger, SF., Wu, Y. Interimvalidation of tunnel
fireconsequencemodels; summary of phase2tests.The
Health and Safety Laboratory Report IR/L/FR/93/11, The
Health and Safety Executive. UK, 1993.
Danziger, NH., Kennedy, WD. Longitudinal ventilationanal-
ysis for the Glenwood canyon tunnels. Proceedings of
the Fourth International Symposium Aerodynamics and
Ventilation of Vehicle Tunnels.York, UK, 1982: 169186.
Heselden, AJ M. Studies of fire and smoke behavior rele-
vant totunnels. Proceedings of the Second International
Symposium of Aerodynamics and Ventilation of Vehicle
Tunnels. Paper J 1, 1976.
Kennedy,WD.,Parsons,B.Critical velocity:past,presentand
future. Paper presented in the One Day Seminar of Smoke
and Critical Velocity in Tunnels. London, 2April 1996.
Oka,Y.,Atkinson, GT. Control of smokeflowintunnel fires.
Fire Safety Journal. 1995; 25(4): 305322.
Wu, Y., Bakar, MZA. Control of smokeflowintunnel fires
usinglongitudinal ventilationsystems-astudyof critical.
Fire Safety Journal. 2000; 35(4): 363390.
334
MetrotunnelsinBuenosAires: Designandconstructionprocedures
19982007
A.O. Sfriso
Department of Estabilidad, University of Buenos Aires, Argentina
ABSTRACT: TheCityof BuenosAires,Argentina,isexpandingitsmetronetwork.Some13kmof newtunnels
havebeenexcavatedsince1998andsome20kmarescheduledfor constructioninthenear future. Manymajor
improvementshavebeenimplemmentedduringtheseyearsinthefieldsof designandconstructionprocedures.
Someof theachievementsandlessonslearnedaredescribedinthispaper, including: characterizationof Buenos
Airessoilsforthenumerical modelingof NATMtunneling, descriptionof thedesignandconstructionprocedures
inuseandsomecommentsontheobservedgroundbehavior duringconstruction.
1 INTRODUCTION
The City of Buenos Aires is extending its metro
network as shown in Figure 1. On-going projects
are: LineA, extended 4km, LineB, extended 4km;
LineE, extended 2km; and newLineH, 5kmlong.
Some20kmof new Lines F, G, I arescheduled for
constructioninthenear future(SBASE 2006).
Landmarks of newconstructionprocedures are: i)
introduction of shotcrete, Line B, 1998 (Fig. 2); ii)
Figure1. Metro network inBuenosAires. Existing(A, B,
C, D, E, H) andnewprojects(F, G, I).
belgiantunnelingmethod, LineH, 2000(Fig. 3); iii)
full faceexcavation, LineB, 2004(Fig. 4).
Geotechnical and structural analysis techniques
evolvedconcurrently, fromearth-loadtheory to state
Figure2. First useof shotcrete, LineB.
Figure3. Belgiantunnelingmethod, LineH, 2000.
335
Figure4. Full faceexcavation, LineB, 2004.
of theart computer simulationof constructionproce-
duresandcalibrationof constitutivemodelsviaback
analysisof monitoringdata(Nez1996, Sfriso1996,
1999, 2006).
2 CHARACTERIZATIONOF BUENOSAIRES
SOILSFORTUNNELING
BuenosAiresCity soilshavebeendescribedinother
contributions (Bolognesi 1975, Fidalgo 1975, Nez
1986a, 1986b). Briefly, the Pampeano formation
underlyingBuenosAiresisamodifiedLoess,overcon-
solidated by dessication and cemented with calcium
carbonatein noduleand matrix impregnation forms.
Except for theheavedupper threetosix meters, pen-
etration resistance is systematically N
SPT
>20 with
someheavily cementedzones that exhibit very weak
rockbehavior withN
SPT
>50(Nuez 1986b).
The most used site investigation technique in
BuenosAires is SPT penetration using a2 sam-
pler along with standard lab testing and CTUC test-
ing on recovered samples. Some plate load testing
andMenardpressuremeter testinghavebeenrecently
included as part of thefield investigation specifica-
tionsfor metroprojects(Sfriso2006).
2.1 Underground construction in the Pampeano
Formation
ThePampeanoformationisveryfavourableforunder-
ground construction due to its high stiffness, reli-
able compressive strength, rapid drainage and good
frictional behavior whendrained.
Twoparticularcharacteristicsof theformationmust
beaccountedforinthedesignof undergroundprojects:
i) thePampeanoformationisfissuredandhaslenses
of quasi-granular behavior, forcingtheinstallationof
aprimary support closeto thefaceinorder to avoid
Table1. Assumedin-depthvariationof K
0
.
Depthm K
0
0 to8/12 0.550.70
8/12 to20/24 0.651.00
20/24 to30/32 0.550.80
Table2. PLT Modulusof subgradereaction.
Depth Primaryloading Un-Reloading
m MN/m
3
MN/m
3
0 to8/12 200300 500800
8/12 to12/14 400600 8001200
12/14 to20/24 600800 12001800
20/24 to30/32 250500 6001400
crownoverexcavation;andii)materialsdrainataspeed
comparabletothat of theconstruction.
Duetothesefactors,themaxallowabledriftwithout
support isabout 2.5meters. Uptothismaximum, the
unsupporteddrifthasverylittleinfluenceontheresult-
ingsettlements, assoil behavior remainsquasi-elastic
(Sfriso2006, Nez 2007).
2.2 In situ stresses
It isaccepted(Bolognesi 1991, Nez 1986a, 1986b,
Sfriso 1999, 2006) that upper Pampeano soils are
overconsolidatedbydessicationtoanequivalentpres-
sure0.81.2MPa. Table1lists theassumedin-depth
variation of K
0
used for the design of underground
structures(Sfriso2006). Thesefigureshavenot been
actually measured directly but estimated after back-
analysisof monitoringdata.
2.3 Modulus of subgrade reaction
The most reliable information of in-situ stiffness is
retrievedviaplateloadingtestsperformedinvertical
shaftsorpilottunnels.Typical in-depthvariationof the
modulusof subgradereaction, asdeterminedbyPLT,
islistedinTable2.
2.4 Parameters for numerical modelling
Hyperbolic model (Duncan 1970, Vermeer 1998)
has been extensively used for the numerical analy-
sisof undergroundconstructioninBuenosAiressoils
(Sfriso 1999, 2006). After eight years of continuous
usage and calibration, a set of input parameters for
thePlaxis implemmentationof thehyperbolic model
(Vermeer 1998) has been found to best represent
336
Table3. Material parametersusedfor numerical simulations.
Fill 08/12 8/1220/24 >20/24
min max min max min max min max
c
u
(KPa) 20 50 50 100 110 220 40 120
u
(
) 8 15 10 20 5 20 0 5
c (KPa) 0 5 10 25 25 50 15 30
(
) 28 30 28 31 30 34 28 31
(
) 0 0 0 3 0 6 0 3
E
r
50
(MPa) 10 20 60 100 75 150 60 100
E
r
ur
(MPa) 25 50 150 250 180 300 140 220
m() 0 0 0 0 0 0 0 0
() 0.20 0.20 0.20 0.30 0.20 0.30 0.25 0.35
R
f
() 0.85 0.90 0.80 0.90 0.80 0.90 0.80 0.90
the observed behavior of tunnels, caverns and open
pit excavations. This set is listed inTable 3. Stress-
strainrelationshipof theHSMmodel isreproducedin
Equations1ato1d.
InEquations 1ato 1dandTable3,
1
and
3
arethe
major and minor principal stresses,
1
is the major
principal strain, c is either undrained cohesion c
u
or
drainedcohesionc
,iseitherundrainedfrictionangle
u
or drainedfrictionangle
, is dilatancy angle,
E
r
50
andE
r
ur
arereferenceloading/unloadingYoungs
modulus, m is stiffness exponent, is Poissons ratio
andR
f
isthefailureratio.
3 CONSTRUCTIONPROCEDURES
3.1 Tunnels
Construction procedures evolved from german
method(Fig. 2) to belgianmethod(Fig. 3) andhave
probablyreachedanoptimal stagewithfull faceexca-
vation(Fig. 4). Figure5shows thecross sectionof a
typical two lane, full facetunnel, as usedinLines B
andH.A 15cmunreinforcedshotcretelayerand1.0m
spacedlightweight latticegirdersaccount for thepri-
mary support of thetunnel, later supplemented with
3040cmof cast-in-placeunreinforcedconcrete.
Figure 5. Cross section of a typical two lane, full face
tunnel, linesB andH.
The Metro authority requires that tunnels remain
dryduringoperation, thusrenderingcast-inplacesec-
ondaryliningasthecheapest option, whencompared
tomembranebarriersandsecondaryshotcretelining.
Somefull face, all-shotcretesections withimperme-
abilization barriers have been successfully built in
linesHandB.
Noclosureof thestructural ringisneededfor sta-
bility, andthereforeadvanceratesof 2.5m3.5mper
12hr shift areconsistently achieved. After thetunnel
isexcavated, acastinplaceinvertisplacedin5m6m
segments, allowingfortheplacementof thesecondary
lininginsinglepoured5msegments. Figure6shows
atunnel after placement of theinvert, whileFigure7
showstheformworkbeingdrivenintothetunnel.
3.2 Underground stations
Underground caverns for metro stations have been
built usingmanytechniquesincluding: i) cut & cover
slab-on-piles; ii) undergroundexcavatedmaincavern
&openpitexcavatedupper hall; and, iii) underground
excavatedmaincavern& upper hall.
337
Figure6. Tunnel after placement of theinvert, LineB.
Figure 7. Formwork used to cast the secondary lining,
LineB.
Asper 2007, threestationsareinexcavationstage.
Thefirst twoareEcheverraStation(Fig. 8) andVilla
UrquizaStation, LineB, wherethegermantunnelling
methodwasused. ThethirdstationisCorrientesSta-
tion, LineH, wherefull faceexcavationisbeingused
(Fig. 9).
CorrientesStationisthelatest andmorechalleng-
ing improvement to construction procedures used in
BuenosAires metro tunnellingso far. It is anunder-
ground cavern 14.1mhigh, 18.9mwide and 135m
long (Fig. 10). On top of the main cavern, a 6m
high access hall shall be excavated full-face after
completionof thesecondaryliningof themaincavern.
Theprimaryliningof CorrientesStationisformed
by2040cmmeshreinforcedshotcreteplacedintwo
layers, and 1.0mspaced lightweight lattice girders.
The construction procedure is shown in Figure 11.
Full-face excavation is accomplished via a series of
four benches, eachone5mlong. Two excavators are
permanentlysetatthetwotopandtwobottombenches,
respectively.Thebottombenchexcavatoralternatively
lies on soil or on top of the cast-in-place invert,
included into the primary support lining to reduce
Figure 8. Echeverra Station, Line B. German tunnelling
method.
Figure9. CorrientesStation, LineH. Full faceexcavation.
Figure 10. Longitudinal sketch of Corrientes Station,
LineH.
costsandtimeschedule.Whileexcavationof Echever-
raStationtook some14months, it shall took some6
monthstocompleteCorrientesStation, withadvance
rates1.0mper day.
338
Figure11. Constructionprocedurefor Corrientes Station,
LineH.
4 DESIGNPROCEDURES
4.1 Primary lining
Thepreliminarydesignof theprimaryliningislargely
basedonexperience. Bythetimethetunnel shownin
Fig. 2wasbeinganalyzed, asimplifieddesignmethod
wasdevelopedtoestimateforcesactinginthecrownof
theprimarysupportof circularsections(Nez1996).
Theexpressionsare
whereN isthenormal forceatcrown, M istheflexure
moment at crown, p
v
is the vertical pressure on the
crown, D isthetunnel diameter, A istheunsupported
drift,E, aretheelasticparametersof thesoil massand
E
r
,
r
aretheelasticparametersof thesupportsystem.
Structural forcesobtainedwithequations2a,2band2c
comparewithin10%15%withthosecomputedusing
themoreinvolvedprocedureby Einstein& Schwarz
(Einstein1979).
4.2 Simulation of construction procedures
Construction procedures are simulated using 3D
elastoplastic models that allow for theestimation of
surfacesettlements, thecomputationof facestability
and the determination of structural forces acting on
the primary lining. Structural forces computed with
3DFEMaresome20%lowerthanthoseobtainedwith
Eqns. 2a, 2b, 2c. Theseequations, whenappliedtothe
tunnel showninFig. 5, resultedinM=0.55KNm/m
and N=416KN/m. 3D numerical models yielded
M=0.47KNm/mand N=415KN/m. (Sfriso 2006,
Nez 2007).
4.3 Secondary lining
Metroauthority requiresthat thesecondary liningbe
designed using earth-load procedures and beamon
springsanalyses. Bothprimaryliningandtheeffectof
constructionproceduresaredisregardedinthedesign
of thesecondarylining.
5 GROUNDBEHAVIOR
Groundbehavior has beenlargely elastic for all con-
structionproceduresandundergroundstructuresbuilt
sofar. Disturbancetosurroundingstructuresandfacil-
itieshasalwaysbeenminimal, andsurfacesettlements
intherange2mm8mmfortunnelsand4mm15mm
for underground caverns have been observed for all
constructionproceduresandsoil covers. Whilethisis
a desirable behavior fromthe point of view of con-
structionandsafety, it alsomeansthat uncertainty of
the predictions remain high, because it is unknown
howsafetheconstructionproceduresreallyare.
A numerical excercisehasbeenperformedtocom-
paretheconstructionproceduresforsafetyandimpact
tosurroundings.A tunnel section10mwide, 8mhigh
with a soil cover of 5mwas used, and the low side
parameterslistedinTable3wereadopted. Theresults
arelisted inTable4 (Sfriso 2006). It can benoticed
that the german method proved to be the least safe
construction method, due to the low safety of the
unsupportedaccesstunnelsexcavatedtobuildtheside
walls(Fig. 12).
Echeverra Station and Corrientes Station have
proven enlightening experiences for the purpose of
checkingpredictions. Forsafetyconsiderations, itwas
decided that Echeverra Station be excavated using
thegermanmethod, andamax. surfacesettlement of
15mmwas predicted. After aseries of small access
339
Table4. Numerical comparisonbetweenconstructionpro-
ceduresfor tunnels.
Full
German Belgian face
Max. surfacesettlement,
undrainedparameters(mm) 4.9 4.3 4.6
Max. surfacesettlement,
drainedparameters(mm) 7.4 5.3 6.7
Max. angular distorsion,
undrainedparameters(10
3
) 0.26 0.22 0.18
Max. angular distorsion,
drainedparameters(10
3
) 0.30 0.26 0.26
Factor of safety,
undrainedparameters() 2.6 >7 4.7
Figure12. Unsupportedpilottunnel for sidewalls, german
methodof tunnelling.
tunnels were excavated to improve the construction
schedule, andbeforethemaincavernexcavationwas
started, a surprisingly high 10mm settlement was
observedat surface.
When the numerical model was re-run with the
access tunnels included but without any change in
material parameters, theobservedsurfacesettlement
could be reproduced. It turned out that the uncon-
finementproducedbythetoomanyintersectingsmall
tunnelswasresponsiblefortheundesiredbehaviorand
yieldedatemporarilyunsafecondition. Itwasdediced
that the abandoned access tunnels be supported by
struts(Fig. 13).
At Corrientes Station, theobservedsurfacesettle-
ment 5mm8mmis much lower than the predicted
value of 20mm. After interpretation of the moni-
toring data, it has been concluded that the unload
Youngsmodulusof Pampeanosoilsislowerthanorig-
inallyestimated, andthatthedepositreboundispartly
Figure13. Strut supportedaccesstunnels, EcheverraSta-
tion, LineB.
responsiblefor thesmall settlementsobserved. Being
thefirstlargeclosedringstructureeverbuiltinBuenos
Aires, Corrientes Station is the first opportunity to
properlycalibratetheunloadingYoungsmodulusand
theeffect of soil rebound.
6 REMAININGCHALLENGES
The advancement in design and construction pro-
cedures is an endless activity. Despite the efficient
methods actually in use, someremaining challenges
needto beaddressedinthenear future. Theseare: i)
theimplemmentationof areliableproceduretomea-
sureK
0
; ii) theabandonmentof cast-in-placeconcrete
and dry tunnels; iii) theuseof robot-placed, fiber
reinforcedshotcrete; iv) theimplemmentationof more
advanced topographic guiding systems; v) optimiza-
tions in the usage of lattice girders; and vi) better
control of groundwater duringconstruction.
7 CONCLUSIONS
13kmof metrotunnelshavebeenexcavatedinBuenos
Aires in the period 19982007. Construction proce-
dures in 2007 include shotcrete and full face exca-
vation both in tunnels and caverns, while design
procedures includestateof theart numerical simula-
tionof constructionprocesses. Best fit parametersfor
theconstitutivemodelsusedwereintroducedandsome
observedfeaturesof soil behaviorhavebeendescribed.
ThePampeanoformationunderlyingBuenosAires
City is very favourable for underground excavation
duetoitshighstiffness, reliablecompressivestrength,
rapid drainage and good frictional behavior when
drained.
340
Groundbehaviorhasbeenlargelyelasticforall con-
structionproceduresandundergroundstructuresbuilt
sofar. Disturbancetosurroundingstructuresandfacil-
itieshasalwaysbeenminimal, andsurfacesettlements
intherange2mm8mmfortunnelsand4mm15mm
for underground caverns have been observed for all
constructionproceduresandsoil covers.
CorrientesStationisthelatestimprovementtocon-
structionproceduresusedinBuenosAiresmetrotun-
nellingsofar.Anundergroundcavern135mlongshall
becompletely excavated in six months with surface
settlementslessthan10mmandminimal disturbance
tosurroundings.
ACKNOWLEDGEMENTS
The working teams fromthe companies involved in
theseprojects, Techint, Dycasa, Roggio andSBASE,
sharedtheirimpressions,opinionsandexperiencewith
the writer, allowing himto learn many things from
them. EduardoNeztaughtthewriter thebasics, the
involvedtheoriesandthetricksof tunnelling, carefully
supervisedthewritersearlyworks, andremainsasthe
day-by-daywriterssourcefor expert advice.
REFERENCES
Bolognesi,A. 1975. CompresibilidaddelossuelosdelaFor-
macinPampeano. V PCSMFE, Argentina, V: 255302.
Bolognesi, A. andVard, O. 1991. Subterrneos enBuenos
Aires. IX PCSMFE, Chile, III:13291350.
Duncan, J. andChang, C. 1970. Nonlinear analysisof stress
andstraininsoils. JSMFD, ASCE, 96, SM5, 16291653.
Einstein, H. andSchwartz, C. 1979. Simplifiedanalysisfor
tunnel supports. JGED, ASCE, 105, 4, 499518.
Fidalgo, F., DeFrancesco, F. andPascual, R. 1975. Geologa
superficial delallanuraBonaerense.VI Arg. Geol. Conf,
110147.
Nez, E. 1986a. Panel Report: Geotechnical conditions in
BuenosAiresCity. V Conf. IAEG, 26232630.
Nez,E.andMucucci,C.1986b.Cementedpreconsolidated
soilsasveryweakrocks. V Conf. IAEG, 403410.
Nez, E. 1996. Tnelesdeseccincircular enlaformacin
pampeano. bull. SAMS, 29, 115.
Nez, E. 2007. Casagrande lecture: Uncertainties and
ApproximationsinGeotechnics.XIII PCSMGE,Venezuela.
Inpress.
SBASE, 2006. Institutional web page. www.sbase.com.ar.
Sfriso, A., 1996. Tnelesdeseccincircular enlaformacin
Pampeano comentario. bull. SAMS, No. 29, 1619.
Sfriso, A. 1999. Tunnels in Buenos Aires: Application of
numerical methodstothestructural designof linings. XI
PCSMGE, Brasil, 637642.
Sfriso, A. 2006, Algunosprocedimientosconstructivospara
laejecucindetnelesurbanos. XIII CAMSIG,Argentina,
117.
Vermeer,P.1998.Plaxis Users Manual.Balkema,Rotterdam.
Ne, 577p.
341
Studyontheearthpressuredistributionof excavationchamber inEPB
tunneling
T.T. Song& S.H. Zhou
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R. Chnia
ABSTRACT: Inshieldtunneling, theexcavationstabilityisveryimportant. Thebalancepressurewhichsup-
ports the work face is offered by the the mucks in the earth chamber. The ideal supporting pressure on the
excavationfaceistrapezoidal. Infact, thepressureisirregular. So, theconceptionsof earthpressuresupporting
ratio(EPSR), regular modulusof earthpressureandbuffer abilityof chamber areputforward. Earthpressuresin
twodifferentsituationsof shieldtunnelinginclaygroundandcobblesandgroundarestudied. Thestudiesshow
that EPSR andregular modulus of earthpressureinsoft groundarebothbetter thanthoseincobbleground.
EPSR andregular modulusof earthpressurearetwoappropriateassessingindexesfor EPB tunneling.
1 INTRODUCTION
Withtheadvantagesof surroundinginfluence, excava-
tionspeed,structurequality,workingenvironmentetc.,
theshieldtunnelingmethodispopularlyappliedinthe
tunnelsof metro, railway, road, municipal engineering
andsoon.
Earth pressurebalancemachine(EPB) andslurry
shieldtunnelingarethetwomostlycommonmethodin
metrotunnels. Comparingwithslurry shieldmethod,
theEPBhasthemeritsof small constructionyard, less
cost, simpletechnology. SotheEPB is morepopular
inmetrotunnels.
EPB machineapplies pressureto workingfaceby
themuckswhicharecutfromthefacesometimeswith
injectionof soil conditioningadditives. Withthepres-
suretheworkingfacecanmaintainstable.Thestability
of workingfaceisakey factor intheEPB tunneling.
Theaccidents brought by destabilization of working
facearethemainaccidentsaccordingtostatistics(Qin
J ianshe, 2005).
Present researchesabout theshieldtunnelingpres-
suremainly focus onthetheoretical pressureneeded
fortheworkingface(AnagnostouG, Kovari K., 1996;
Abdul-Hamid Soubral, 2000 & Qin J ianshe, 2005).
As for how to apply the pressure and the pressure
properties are seldommentioned and researched. In
this research, features of workingfacepressure, dis-
tribution of pressure in the excavation chamber and
the pressure in the clay and cobble sand strata are
researched. Some conceptions are put forward, and
principlesaresummarized.
2 CLASSICAL EPBTHEORY
Theexcavatingsystemof EPB machineis composed
by shield, cutter head embedded with cutters, work-
ingchamber, pressurewall, screwconveyor, andthrust
cylinders (seeFig1). Whenexcavation, thrust cylin-
dersapplyforcetopressurewall, andthepressurewall
appliesthepressuretotheworkingfacebythemucks
intheearthchamber. Thepressureisusedtobalance
thewater andsoil pressureinground.
EPBmachineadjustspressuretobalancewater and
soil pressureinworkingfacebythemannersof chang-
ing the pressure of thrust cylinders and the rotation
speed of screw conveyor (see Figure 1). When it is
necessary to increasethepressurein earth chamber,
therotationspeedshouldbedecreasedandthethrust
forceshouldbeincreaseviceversa.
Classical earthpressureassumesthese:
1. Thepressureappliedbyshieldmachinetothework-
ingfaceistrapezoidal.Andthetrapezoidal pressure
balanceswiththewaterandsoil pressureinworking
faceinorder tomaintainfacingbalance.
2. Thepressureappliedtothegroundis equal tothe
pressureinthepressurewall (M. Herrenknecht &
U.Rehm,Aug,2003).So,theworkingfacepressure
isacquiredthroughthepressurevaluesshowninthe
operatingpanel.
Obviously, theconditions for classical earth pres-
surearesatisfiedwiththefollowings:
1. The mucks in earth chamber should have the
propertiesof ideal plasticfluidity.
343
Figure1. Soil pressure+water pressure=earthpressureinworkingchamber (Wassmer, Treceno&ANdreossi, 2001).
2. Theopeningrateof cutter headisenough.
3. Thereisnopressurelossinthescrewconveyor.
In fact, EPB cant reach this ideal state in the
practiceof excavation.
3 DEFINITIONSRELATEDEARTHPRESSURE
Thepressureapplied by shield machineis related to
thestability of workingfacedirectly. Inpractice, its
hardtoobserveandgaugethepressureof theworking
face. So, someindirect waystoestimatethepressure
of workingface, asbelow:
1 Thepressuresensor inthepressurewall. Thisway
is thought as the most efficient way. But it has
differencebetweenthetwopressures.
2 Theforceof thrust cylindersandthetorqueof cut-
ter head. They increases with the earth pressure
increase;
3 Thevolumeof mucks discharge. Moredischarge,
lessearthpressure.
For further research, thefollowingdefinitions are
given:
3.1 Earth pressure supporting ratio (EPSR)
Theclassical earth pressuretheory assumes that the
supporting pressureon theworking faceis supplied
by earthintheearthchamber (seeFig1). Infact, the
supportingpressureiscomposedbytwoparts: earthin
workingchamber andtheplaneof cutter head. As to
theFigure2, theearthpressureis mainly transferred
though the open of cutter head. But the plane also
contributessupportingpressure. Sotheearthpressure
supportingratio(EPSR) isdefinedastheratioof earth
pressureinthetotal supportingpressure.
Figure2. Schematicplanof openandplane.
whereEPSR=EarthPressureSupportingRatio;
earthsupportingpressure: thetotal pressureof earth
pressureontheworkingface;
total supporting pressure: the total supporting pres-
sure on working face including earth pressure and
cutterheadplanesupportingpressure;
Thetotal supporting pressureis assumed thebal-
ancepressureontheworkingface.
Theratioisrelatedtotheshapeof cutterhead, open-
ingrateandthecontrol of earthpressure. Thevalueof
theratioisbetween0and1whichindicatesthestateof
earthcontrol whenexcavating. For example, theratio
of cutter head of spoke shape is bigger than that of
planeshape.
3.2 Regular modulus of earth pressure (RMEP)
Fivepressuresensorsarelocatedonthepressurewall
(seeFig1). Thepositionof thesensors areshownin
Figure3, andtheheight isshowninTable1.
344
Figure3. TheDistributionof earthpressuresensors.
Table1. Positionof thesensors.
Sensor No. 1 2 3 4 5
Position(m) 1.2 1.8 2.7 3.4 5.4
The regular modulus of earth pressure RMEP is
usedfor indicatingtheregularityof theearthpressure
inworkingchamber. Thevalueof RMEP isdefinedas
relatedcoefficients: R
2
betweenvaluesof earthpres-
sureandthecorrespondingvertical position.Thevalue
of themodulusisbetween0and1whichindicatesthe
fitting extent of the fitted regression line. The ideal
valueis1.Thevalueiscloserto1, betterregularof the
earthpressureis.
where
Y
i
: actual datapointsof earthpressure
Y
i
: predictedvalueof theregressionmodel
Thedetail definitionof R
2
canreferredRichardA.
J ohnson(1992).
3.3 Buffer ability of working chamber
It cant keep the balance pressure when the mucks
in the working chamber are over discharged due to
improperoperation.Thevolumeof theworkingcham-
ber shouldhavetheabilityof maintainingthepressure
duetoimproper dischargingoperation. Sothecham-
ber shouldbuffer theearthpressurewhenthemucks
aresuddenloss.
Figure 4. Total pressure makeup: additive pressure of
facing+lateral pressureof mucks.
Here, thebufferabilityof chamberisdefinedasthe
proportion of discharged mucks in thewholecham-
ber at theconditionof screwconveyor workingfor 5
minuteswithmaximumrotationspeed.
It is deemed that the buffer ability is good if the
proportionof dischargemucksislessthan30%.
4 PROPERTIESOF EARTHPRESSURE
OF EPB EXCAVATION
4.1 Analysis of pressure makeup
Thepressurewall isinthebehindof workingchamber.
Thewall undertakesthemainfacepressure. Herethe
seriesStypeEPBshieldmachineof Herrenknechtare
take an example to research the wall pressure. The
earthpressuresensorsinthewall aredistributedasthe
Figure3. Therearefivesensorsinthewall. Theserial
number andthepositionareshowninFigure3.
Themeasuredpressureshapeistrapezoidal inmost
commonsituation(seeFig1&Fig4).Thetrapezoidal
pressureiscomposedof twoparts: oneisrectangular
andanotheristriangular.Thetriangularcanbedeemed
asthelateral pressureof themucks,andtherectangular
canbedeemedastheadditivepressureappliedbythe
water andsoil pressureof theworkingface.
Analysishereinbefore, therewontbeadditivepres-
sureif thepressureshapeistriangular.Atthissituation,
if thethrustforceisbigandthefaceisstable, thecutter
headwill undertakelargeportionof facepressure. So,
the earth pressure support ratio EPSR is low which
lead to the large contact stress between cutter head
andfaceandlargetorqueof cutter wheel. Usually, in
this situation, thewear of cutters andcutter headare
serious.
345
Figure5. Earthpressuredistributiondiagram(Herrenknecht & Rehm, 2003).
Table2. Thegroundparametersof Guangzhou.
(g/cm
3
) w(%) e c(kPa) +(
) Es1-2(MPa)
1.95 31 0.8 22.8 18.1 4.7
4.2 Pressure properties in clay
Usually, themucks of clay stratahaveagoodplastic
fluiditywhichcanapplyaregularpressuretotheexca-
vationface. So, theearthpressuresupportingratiois
high; andthemodulusof theearthpressureishightoo.
Thestabilityof workingfacecanbewell controlledif
theworkingchamber satisfiesthemucksbuffer.
Nevertheless,thereispressurelossinthescrewcon-
veyorinthiskindof strata.Thepressuregradientcurve
isshowninFig4. It isinevitablethat pressurefallsin
thescrewconveyorbecausethepressureis0inthedis-
chargingoutlet. Thepressuregradient dependsonthe
fluidity and impermeability of themucks. Thepres-
sureintheinlet of screwconveyor has fallento80%
(seeFig5). It indicatesthat thepressureisnt regular
duetheexistenceof screwconveyorthoughthemucks
isof plasticfluidity. (seeleft inFig5)
Theexcavationdataof ringNo.219inonesectionof
line5of Guangzhoumetroaretakenfor anexample.
The cutter head of the EPB machine is plane type,
and opening rateis about 28%. Therearetwo types
of cutters: disccutter andscrapecutter. Thelengthof
workingchamberis1m.Thepositionof earthpressure
sensors are shown in Figure 3. The tunnel is in the
stratumof <42>whosemainparametersareshown
inTable2.
The Figure 6 is the figure of earth pressure data
which are acquired fromthe earth pressure sensors
in pressure wall. The number order of the pressure
sensorsisfromthetopdown. Theyare1#, 2#, 3#, 4#
and 5#. Fromthe data it can obtain that the regular
modulus of earthpressureRMEP is about 0.98. This
valueshowstheearthpressureof pressurewall ismore
regular.
Earth pressure of the sensors
R
2
= 0.9761
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13
MPa
m
Figure6. Meanearthpressureinthepressurewall.
Table3. Assessment of earthpressuresupportingratio.
Additivestress/ Max. stress
Actual torque/
Max. torque 030% 3050% >50%
-0.8 -0.5 0.50.8 >0.8
>0.8 -0.3 0.30.5 >0.5
-1.1
>1.1 -0.1 0.10.3 0.305
The proportion of earth pressure and cutter head
is hard to bedivided. Theearth pressuresupporting
ratiocantbecalculatedprecisely. So, itcanbejudged
by the earth pressure in the pressure wall and the
torqueof cutterheadbythefollowingway.Theadditive
346
Table4. Mainindexof theEPB (S365).
Machiner Excavation Rating Max. Max. Opening Cutter
type diameter(m) torque(KNm) thrust(kN) speed(mm/min) ratio head
EPBTBM 6.28 5980 34210 80 28% diskandscrape
Table5. Theindexesof thestratum.
(g/cm
3
) w(%) c(kPa) +(
) E
0
(MPa) K(m/d)
2.2 31% 01 38 45 27
pressurecanbeknownfromthepressurewall.Because
theadditivepressureis appliedby face, higher addi-
tivepressure, higher earth pressuresupporting ratio.
If thecutter headplateundertakes morefacingpres-
surethetorqueof thecutter headwill becomebigger.
According to empirical summary and modification,
themethodusingadditivepressureandtorquetoesti-
mate earth pressure supporting ratio is put forward.
(showninTable3)
The actual torque of this ring is 3650kNmwhile
the rating torque is 4500kNm. The value of actual
torque/ratingtorqueis0.81; andthevalueof additive
stress/max. stress is 0.32. So according to the earth
pressureandthetorqueandassessingby theTable3,
theEPSR isabout 0.65.
4.3 Pressure properties in cobble sand strata
Theline1of Chengdumetroisinstrataof cobbleand
sand.TheEPBmachineis+6.28mof Herrenknt.The
mainparametersof themachineareshowninTable4.
Fromtheanalysishereinbefore, theEPSR isabout
0.60.7inthesoftclaystrata.Andinmostof thesitu-
ationstheEPBtunnelingcantreachtotal earthbalance
inworkingface.Thevaluesof EPSRsareusuallycon-
trolledbigger than0.5. Themainbalancepressureis
earthinthisstratum. Theregularityof theearthpres-
sureis good. Theregular modulus of earth pressure
reaches0.98.
Thetunnel lineisinthestrataof cobbleandsand.
Thecobblecontent ishighandthepermeabilitycoef-
ficientishigh. Themainindexesof strata<37>are
listedinTable5.
ThepressuredataintheFigure7arecollectedinthe
ringof 112whichis themeanvalueinthering. And
thetorqueof cutter headis3020kNminthisring. The
EPSR isabout 0.4fromtheTable3andFigure7.
Theregularmodulusof earthpressureRMEP inthe
chamberis0.93whichislessthanthevalueinclay.The
pressuregradientcanreflectthemuckunitweight.The
gradientvalueistheproductof unitweightandlateral
Earth pressure of the sensors
R
2
=0.93
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0 0.05 0.1 0.15
MPa
m
Figure7. Meanearthpressureinthepressurewall.
pressurecoefficient. Dueto theirregularity of earth
press in cobbleandsandstrata, its hardto calculate
theunit weight. However, it shows theunit weight is
bigger accordingtoFigure6andFigure7.
Thedatashowthat theEPSR is low. So thecutter
headplatewill undertakebigger pressure. LowEPSR
isunfavorablefor wear of cuttersandcutter head. So
theEPSR should beincreased by taking somemea-
sures. Such as soil conditioning should be taken to
increase the mucks plastic fluidity. Simultaneously,
theregularmodulusof earthpressurecanbeincreased.
4.4 Buffer ability of working chamber
The discharging ability of the screw conveyor is as
below:
347
where:- discharging efficiency, if there is no loss,
=1. Incommon is about 0.9; D
1
- inner diameter
of screwconveyor; D
2
- outer diameter of center shaft;
P- Distanceof paddle; n- Rotationspeed.
Accordingtoformula(6) themaximal discharging
abilityof screwconveyoris1.2m
3
/min.Thevolumeof
dischargingis6m
3
in5min. Thecubageof thecham-
berisabout30m
3
.Accordingtoformula(5),thebuffer
abilityis20%(6/30=20%) of thecubageof chamber
(showninformula7).
Accordingtocalculation, 30%lossof workingcham-
ber volume can cause big ground settlement. The
bigger buffer proportioncancausemoregroundloss
and instability of working face. If the proportion is
over 30%, it canbedeemedthat thebuffer ability of
thechamber is not enough. If it is below30%it can
bedeemed that thechamber has theability of earth
pressurebuffer for improper operation.
5 CONCLUSIONS
Fromtheanalysis and study of thetheory and mea-
sureddata, themainconclusionsareasbelow:
1 Theclassical earthpressuretheoryhaslimitedand
premises.
2 Thepressureinthepressurewall isdividedintotwo
parts: additivepressureof workingfaceandlateral
pressureof mucks.
3 Theconceptionsof earthpressuresupportingratio,
regular modulus of working chamber and buffer
ability of chamber areput forwardtoresearchthe
pressureproperties.
4 Claystratumandcobblesandstratumaretakenfor
examplestostudy.
5 Theresearch shows that theactual pressurecant
reachthesituationof classical pressure.
6 EPSR andregular modulusinclayarebothhigher
than in cobble and sand. So the invalid wear is
higher incobbleandsandstrata.
REFERENCES
Abdul-HamidSoubral 2000.Threedimensional facestability
analysisof shallowcirculartunnels. International Confer-
ence on Geotechnical and Geological Engineering, 2000.
Australia: Melbourne.
Anagnostou G & Kovari K. 1996. Face stability condi-
tions earth pressure balanced shields. Tunneling and
Underground Space Technology11(2):165173.
Herrenknecht,M.&Rehm,U.2003.EarthPressureBalanced
ShieldTechnology. Internal Lecture in Colorado School
of Mine, USA.
HisatakeM. &EtoT Murakami T. 1995. Stabilityandfailure
mechanisms of atunnel facewithashallowdepth. Pro-
ceedings of the 8th Congress of the International Society
for Rock Mechanics. J apan: Fujii.
J ohnson, R.A. &Bhattacharyya, G.K. 1992. Statistics: prin-
ciples and methods. NewYork: Wiley.
Qin, J.S. 2005. studyonmechanismof thefacedeformation
andfailureinshieldtunneling. P.H.D thesis, University of
Hohai.China: Nanjing.
Shikibu, N. 1995. Stabilityof faceduringshieldtunneling
A survey of J apanese shield tunneling 1995. In: Fujita,
Kusakabe. Underground Construction in Soft Ground.
Rotterdam: Balkema.
Stallebrass, S. E., Grant, R.J. & Taylor, R.N. 1996. A Finite
element study of ground movements measured in cen-
trifuge model tests of tunnels. Geotechnical Aspects of
Underground Construction in Soft Ground, 1996. Rotter-
dam: Balkema.
Wassmer, L., Treceno, O. &ANdreossi, E. 2001. Tunnel bor-
ingmachine(TBM)ApplicationinSoftGroundCondition.
I.M.I.A. Meeting 2001, Australia: Sydney.
348
Backfill groutingresearchat GroeneHartTunnel
A.M. Talmon
Deltares (Delft Hydraulics) & Delft University of Technology, Delft, The Netherlands
A. Bezuijen
Deltares (GeoDelft) & Delft University of Technology, Delft, The Netherlands
ABSTRACT: Measurementsonback-fill groutingat GroeneHartTunnel (GHT) aredescribedandanalysed.
TheresearchutilisesTBM data, monitoredgrout pressures andmeasuredrheological properties of tunnelling
fluids. Grout injection strategy and vertical grout pressuredistribution behind theTBM werefound to agree
withestablishedtheory, likefoundinother tunnellingprojects. However, onsomeoccasions different smaller
vertical groutpressuregradientswerefound, whichmighthavebeencausedbyanincidental upwardmovement
of thetunnel. Itisshownthattheconsolidationof groutdeterminesthedecayof groutpressures. Theassociated
net hydraulic resistanceof surrounding soil is substantially higher than according to radial flowtheory. This
higher resistanceisattributedtobentoniteoriginatingfromthefaceof theTBM, havinginvadedthegrout-soil
interface.
1 INTRODUCTION
TheGroeneHartTunnel (GHT) intheNetherlandsis
part of ahighspeedrailwaylinebetweenAmsterdam
and Brussels. Thelocation of thetunnel is about 20
kilometers South-West of Amsterdam. Thelength of
thetunnel is7.16km. Thetunnel issituatedat adepth
of about30meters. Hydro-geotechnical conditionsare
describedbyAimeetal. 2004.Thisdoubletracktunnel
has adiameter of 14.5meters. Aristaghes et al. 2002
reportedonthedesignof thetunnel.
Grouting research at GHT was conducted as a
cooperation between Delft Institutes, HSL project
organisationandCOB-research.Thestudyfocusedon
theinteraction of soil, grout andtunnel construction
Figure 1. Geotechnical profile Groene Hart Tunnel and measurement locations: Instrumented plot no 1 at ring 344
(29/03/2002), COB-passageat ring2117(03/06/2003) andNoordplaspolder at ring3241(05/11/2003).
at the injection of grout. Three specific aims were:
1) Analysis of grout pressures measured at GHT.
2) Quantificationof theconsolidationof grout andits
influenceongroutpressures. 3)Modellingof thegrout
pressuredistribution; specificallyfromgroutinjection
at theTBM tothebuoyancydominatedregionfurther
behind theTBM. Novel conditions at the time were
theimpressivediameter of thetunnel: 14.5m, incon-
trasttothe10mdiametertunnelsmonitoreduptothen,
thecritical hydro-geotechnical softsoil conditionsand
adifferent typeof grout. Thegeotechnical profileis
showninFig.1.
Laboratory characterisations were conducted on
the grout: vane testing for rheological properties,
consolidation experiments, and tests to determineto
349
0
1
2
3
4
5
6
7
0 10 20 30 40 50
hours since production [hh]
s
h
e
a
r

s
t
r
e
s
s

[
k
P
a
]
peak strength
remoulded strength
peak strength (Dec 03)
Figure 2. Peak shear strength and remoulded strength of
GHT tail voidgrout, determinedbyvanetesting.
what degree of grout consolidation the grout pres-
suretransducers used in thelining will givereliable
results. Also therheological andconsolidation prop-
ertiesof bentonitethat couldsurroundthegrout were
measured.
Theconstructionof GroeneHartTunnel wasmon-
itoredonanumber of locations. Threelocationswere
availablefor Delft Institutes research: Instrumented
plot no1, aso-calledCOB-passage andapassage
inNoordplaspolder, seeFigure1.
In each of these locations grout pressures on the
tunnel lining weremeasured. TheCOB-passagehad
additional instrumentation: porewater pressuremea-
surements, axial andtangential strainmeasurementin
the tunnel lining, convergence measurements, verti-
cal positionmeasurementsof thetunnel liningandtilt
measurementsof thetunnel lining. Thispaper focuses
ongrout pressures.
2 PROPERTIESOF TUNNELINGGROUT
2.1 Rheology of grout
Yield stresses are governing grout pressures behind
a TBM. Shirlaw et al. (2004) describe the ingredi-
entsof thecement-lessgroutemployedatGroeneHart
Tunnel (GHT): sand, flyash, limeandchemical addi-
tives.Theyalsoaddresssomeof itsphysical properties.
Freshgrout samples weretaken(27/03/2002) froma
supply container locatedintheTBM. Thegrout was
testedinDelft.Thedensityof thegroutis1850kg/m
3
,
thewater content is0.201.
Peak shear strengthandremouldedshear strength
were measured with a vane apparatus as a function
of time. Figure 2 shows the results. The peak shear
strengthof freshgrout isabout 1to1.5kPa. Thetests
also showthat peak shear stresses increaseslowly in
time. Thisiscausedbypuzzolanicreactionof thefly-
ashpresentinthegrout.Theremouldedshearstrength,
onthecontrary, is nearly constant intime. Thegrout
wasfinegrained, thoughnosievetestswereconducted
for quantification. Similar tests were conducted on
groutcollectedinDecember2003.Alsotheshearmod-
ulusof groutwasdetermined. Thispropertyisneeded
tocalculatethetransitiontoelasticbehaviour (=input
toDCLong: Talmon& Bezuijen2005).
Thepeak shear strength, insteadof theremoulded
strength, isheretherelevant parameter, sincethrough
continuedconsolidationthefailuresurfacesareclosed
continuously, andit isherethat thepeak strengthhas
tobesurmountedcontinuously.
2.2 Consolidation of grout
In porous soil conditions elevated grout pressures,
comparedtotheporewater pressureat thesameloca-
tion, leadtoconsolidationof grout, or groutbleeding.
The consolidation of grout, in turn, influences the
effectivestress distributionsomedistancebehindthe
TBM. Consolidationstartsat thegrout-soil interface,
andproceedsslowly intothegrout layer that fillsthe
tail void (this layer is called agrout cake). Consoli-
dation properties of GHT grout were determined in
Delft by meansof element testssuchasdescribedby
Bezuijen & Talmon (2003). Permeability and poros-
ity changeof thegrout arethegoverningparameters.
Relevant results to the calculation of grout consoli-
dation in the tail void are: permeability grout cake
k =2.4*10
8
[m/s], porosityfreshgroutn
i
=0.31and
porosityof theconsolidatedgrout caken
e
=0.24.
3 LABORATORY TESTINGOF GROUT
PRESSURE SENSORS
Pressuresensors(SenSym/ICT 19C100P) havebeen
mounted in the tunnel lining segments. A cavity
(D=82mm, h=23mm) filledwithtail brushgrease
separated the sensor membrane fromthe grout (this
construction is comparable to that at Sophia Rail
Tunnel, whichisdescribedbyBezuijenet al. 2004).
Sinceat theGHT beamactionof thetunnel lining
wastheprimarysubjectof theCOB-research, ithadto
besurethatthegroutpressuredatawasreliable,specif-
ically under consolidated/hardened conditions of the
grout. Thefunctioningof thesegroutpressuresensors
wastestedinthelaboratory, Bezuijen(2004).A sketch
of thetest cell is showninFigure3. A pressuresen-
sor wasplacedinagreasefilledcavityposition, asat
GHT.
Thegrout inthetest cell wasloadedby increasing
the air pressure in the top of the cell. The resulting
grout pressurewasthesameasinthefield.
Uptoabout onehour, thissensor adequately mea-
suredexternally imposedvariationsof total pressure,
seeFigure4. After that thepressurereadingswerein
350
valve
water
collection
geotextile
0.2m
pore pressure gauge
total press.
gauge
grease
grout press.
gauge
plate
grout
pressurized
air
Figure3. Laboratorytestingof grout pressuresensor.
-50
0
50
100
150
200
250
300
350
P
r
e
s
s
u
r
e

(
k
P
a
)
-20
-15
-10
-5
0
5
10
15
20
d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
0 1 2 3 4 5
time (hours)
displacement
air press.
grout press.
GD tot. press
pore press.
Figure4. Exampleof pressuresensor response.
betweentheimposedtotal pressureandthepore-water
pressure.
Duringthisprocessanotherflushmountedpressure
sensorwithverystiff membrane,continuedtomeasure
thetotal stresses. It is concluded that after onehour
consolidatedgrout reachedthepressuresensor. Sub-
sequently grout-archingmight haveoccurredinfront
of theGHT pressuresensor whenloadeddynamically.
As aconsequence, pressurereadings at GHT haveto
beinterpreted with care. It should however bemen-
tionedthat thesetests might havebeentoo stringent;
theimposedpressurevariationsarenotlikelytooccur
in the field. Instead there is a slow decay of grout
pressurewithtime, givenarchinglesschancetooccur.
Furthermorethesensor islocatedat theimpermeable
lining. InSection6itwill beshownthatinatunnelling
processitcantakemorethan25hoursbeforethegrout
closetotheliningisconsolidated
-9
-6
-3
0
3
6
9
0 50 100 150 200 250 300 350 400
pressure [kPa]
v
e
r
t
i
c
a
l

p
o
s
i
t
i
o
n
[
m
]
28 m behind TBM
pore water
slurry front
Figure 5. Pressures measured on tunnel lining at 28m
behindTBM.
4 GROUT PRESSURESMEASUREDAT GHT
4.1 Measured grout pressures at Instrumented
plot no.1
TwelvegroutpressuresensorswereutilisedinInstru-
mented plot no 1. The time-series were relatively
short: thedatacoveredatmax3ringsbehindtheTBM.
ThewirelessFM-datatransfer of thesensorssuffered
interruptionsandmalfunctioning.
Thegrout wasinjectedthroughsixinjectionpipes.
The pressure drop over the pipes was about 3 bar.
Thispressuredropisequivalent toawall shear stress
of 1kPa. This is significantly higher than found for
cementiousgroutssuchasat SophiaRail Tunnel and
Botlek Rail Tunnel, where about 0.2kPa was found
(Talmonet al. 2001).
Pump-strokecountsshowedthat about 60%of the
grout wasinjectedthroughthethreeinjectionportsin
theupper half of theTBM. Thusanet downstreamof
grout iscreatedat thebackof theTBM.
Despite the sparse set of tail void pressure data,
a picture emerged that the vertical grout pressure
gradient behind the TBM (10kPa/m) could have
beenrather small incomparisonto earlier tunnelling
projects such as SophiaRail Tunnel (Bezuijen et al.
2004).
4.2 Measured grout pressures at COB passage
At COB passage, tail voidgrout pressuresweremea-
sured by a total of 32 pressure sensors. Ring 2117
and2118wereeachcircumferentially equippedwith
10 pressuresensors. Theremaining 16 sensors were
equallydistributedover ring2119, 2120and2121.
Thegrout pressureprofile, 28mbehindtheTBM,
is shown in Figure5. At this point it approaches the
hydrostaticpressuredistribution. Immediatelybehind
theTBM, measuredpressuresareabout100kPaabove
porewater pressure. Only two sensors locatedinthe
351
Figure6. Meangroutpressureof ring2117, 2118and2120.
Figure7. Vertical groutpressuregradientsmeasuredatring
2117, 2118and2120. Between16and25metreselectronics
didnot functioncorrectly.
lower half of thetunnel, strangely, showedpressures
equal toporewater pressure.
Average grout pressures measured at ring 2117,
2118 and 2120 areshown in Figure6. It shows that
grout pressures decay with distance fromtheTBM,
in correspondence with the fundamental processes
describedinTalmon& Bezuijen(2005).
The pressure distribution over the circumference
of thetunnel liningis determinedby specific weight
of thegroutandflowresistancewhenthegroutisdis-
tributedaroundthelining.Thevariationof thevertical
groutpressuregradientwithdistancebehindtheTBM
is showninFigure7. This figureshows that at upon
exitingof ring2117this vertical pressuregradient is
rather small: 6kPa/m. Thisimpliesthattherehasbeen
adown-flowof grout withahighflowresistance.
Measuredvertical displacementsof thetunnel lin-
ing(measuredbyawater levellingsystemmeasuring
vertical displacements between ring 2117, 2118 and
ring2057) indicatethat thetunnel lininghadmoved
to a high position before the pressure sensors came
intothegrout. Beforeexitingof Ring2117, thelining
hadalready movedupwardsome6cm, seeFigure8.
Ring2057, located120mbehindtheTBM servedas
areference.
Figure8. Vertical positionof ring2117measuredbyaliquid
levelingsystemandtiltof ring2117measuredwhenthegrout
sensorsfirst cameintothegrout (3-6-200306:00).
Ring2117and2118werealsoequipedwithtiltsen-
sorsthatwereattachedtotunnel ringsegmentssituated
at tunnel axisheight. Thetimetraceof thetilt meters
variedinconcertwithvertical positionsmeasuredwith
theliquidlevellingsystem. Neartheendof drillingfor
ring2120(3-6-200312:00), theliquidlevellingsys-
temhad to bedismantled, thetilt measuring system
continuedmeasuringthough. Thetilt-measuringsys-
temshowedthat after thedrillingof ring2120, thetilt
haddropped0.04degree, whichaccordingto Figure
8 corresponds to a vertical downward movement of
about 6 centimetres sincering 2117 and 2118 came
intothegrout.
Overviewing the data, it is concluded that when
ring 2117 and 2118 came into the grout the tunnel
lining was at atemporally high position. Thecourse
of tilt meter data, vertical pressuregradient andverti-
cal positioncorrespondqualitatively. A representative
vertical groutpressuregradientatthebackof theTBM
is12kPa/m.
4.3 Measured grout pressures at Noordplaspolder
The data from Noordplaspolder produced a more
familiar picture: vertical grout pressuregradients of
about 15kPa/mimmediately behind theTBM and a
vertical gradientof about8kPa/mat17metersbehind
theTBM, seeFigure9.
5 ANALYSISOF GROUT PRESSURESBEHIND
THETBM
A finitedifferencemodel for thecalculationof grout
pressures immediately behind aTBM was first pre-
sentedbyTalmonet al. (2001). Themodel calculates
the distribution of grout issued from the injection
ports, andconsequentlypredictsthepressuredistribu-
tiondirectly behindtheTBM. A tail voidof constant
thickness is assumed and thecomputational domain
stretchesfor about 5metersbehindtheTBM.
352
-8
-6
-4
-2
0
2
4
6
8
0 100 200 300 400 500 600
grout pressure [kPa]
h
e
i
g
h
t

[
m
]
behind TBM (drilling)
8 rings behindTBM
water
grout static
Figure9. Vertical grout pressureprofilesmeasuredat ring
3241of Noordplaspolder.
Figure10. Grout pressuresmeasuredupto1mbehindthe
TBM at Instrumented plot no 1 and calculated pressures
(model DCgrout). s=axial coordinate. Two-sided friction
grout layer:
y
=1.5kPa, 60%of grout is injected through
injectionportsintheupper half of theTBM.
When theresults of theInstrumented plot no 1
cameavailableacalculationof thegroutpressuredis-
tributionimmediatelybehindtheTBM wasmade, see
Figure10. Inputswerethevanetest resultsof March
2002andthemeasureddistributionof flowratesover
thesixgrout-injectionpipes.
Only if two-sided friction of the grout layer is
assumed, likeintheoriginal model, afair correspon-
dencewiththemeasurementsisachieved. A pressure
difference of nearly 150kPa is calculated between
thebottomandthecrest of thetunnel. This pressure
difference is equivalent to a vertical grout pressure
gradientof 10kPa/m. For reference, self-weightof the
grout would produce a vertical pressure gradient of
18.1kPa/m. For 300kPaat thecrest, this wouldlead
toapressureof 562kPaat thebottomof thetunnel.
When first data of the COB-passage came avail-
able it was found that the distribution of flow rates
over the injection ports was quite comparable: 55%
of the grout was injected through the three upper
ports. Thegrout pressuredata(ring 2117) showed a
small vertical grout pressure gradient of about 4 a
6kPa/m, seeFigure7. Thiscouldnot beexplainedby
means of theDCgrout model. Themeasuredvertical
pressure gradient more is compatible with a situa-
tion where all of the grout is forced downwards. In
that case the vertical pressure gradient is given by:
18.1-2*
y
/h=3.1kPa/m.
Atring2118and2120avertical groutpressuregra-
dientof about12kPa/mismeasuredbehindtheTBM.
Here the results correspond more with a situation
wheregrout distributes inboththeup- as downward
directionfromtheinjectionports.
Thepaceof theconstructioncycleatCOB-Passage
wassloweddownbyinstallationof instrumentation.
The measured pressure distribution at Noord-
plaspolder is morein linewith earlier experienceat
other tunnels: f.i. SophiaRail Tunnel (Bezuijenet al
2004).
Giventherheological propertiesof grout, themea-
sured vertical grout pressure gradients can only be
achieved when there is friction on both sides of the
grout layer: friction at the soil-grout interface and
friction at thetunnel-grout interface. Given thehigh
frictional resistancemeasuredover thesupplylines, it
seems justifiedto concludethat this grout is provid-
ingsignificantfrictionwiththeconcretetunnel lining
(contrary to cementious grout employed in previous
projects).
It has been found that the pressure distribution
immediatelybehindtheTBM variesstronglybetween
measurement locations. A likelycandidatethat might
have affected the pressure distribution is vertical
movementof thetunnel lining. Thehigher yieldstress
of thegrout used, compared to thegrout in theear-
lier tunnels, hasasaconsequencethat themovements
of thelining havealarger influenceon thepressure
distributionaroundthelining.
6 CALCULATIONOF THE CONSOLIDATION
OF GROUT
Pressurizedgroutsituatedagainstapermeablesurface
issubject tofluidloss. Twodifferent situationscanbe
discerned: conditions during stand still of theTBM
andconditionsduringdrillingof theTBM.
Under stand-still conditions no freshgrout is sup-
pliedandgroutpressuresdropslowly.Thepressurised
soil will unload. Filtration properties of grout were
measured in the laboratory, and a model was made
for grout consolidationduringstandstill (seeBezui-
jenandTalmon2003). This situationis verifiableon
measured pressure decay during standstill. Involved
propertiesarepermeabilityof grout, porositychanges
of thegrout, hydraulic resistanceof surroundingsoil
andelasticityof surroundingsoil.
353
Figure 11. Skematisation of consolidation of grout. Ele-
vated grout pressures are represented by the force F, the
hydraulicresistanceof porewater flowisrepresentedbyR.
Figure 12. Numerical simulation of grout pressure decay
behind theTBM at COB-passagecompared with measure-
ment: calculatedpressuredecay andcalculatedthicknessof
thegrout cake, at R=210
7
[s].
Under drilling conditions fresh grout is supplied
and the grout pressure is constant. The surround-
ing soil will be under a constant pressure, and only
permeabilityandhydraulicresistancearerelevant.
TheconsolidationprocessisschematisedinFigure
11. The hydraulic resistance R was back-calculated
fromobservedgrout pressuredecay. Measuredpres-
suredecayandcalculatedgrout pressuresmatchwell,
see Figure 12. This back-calculated hydraulic resis-
tance is however substantially higher than that of
surrounding soil. A typical value for the hydraulic
resistance of radial pore water flow into the sur-
roundingsoil isR=65000[s], Bezuijen(2005). It is
hypothesisedthatbentoniteslurryfromthefaceof the
TBM hasinvadedthegrout-soil interface.
Theconsolidationtheory allows thecalculationof
thedevelopmentof thegroutcakealongthegrout-soil
interface, see Figure 12. The thickness of this filter
cakeisimportanttothemodellingof thegroutlayeras
aninterfacebetweentunnel liningandsoil: itincreases
theintegral stiffnessof thegroutlayer.Thefoundation
of thetunnel occursbyconsolidationof thegrout.This
conclusionwasalsoreachedfortheSophiaRail Tunnel
(Bezuijen et al. 2004), where cementious grout was
used.
7 CONCLUSIONS
Thefinegrained grout at GHT has ayield stress of
about 1a1.5kPaand, dueto absenceof cement, the
yield stress of this grout increases only slowly with
time. It was shownthat consolidationof grout deter-
mines thedecay of grout pressures over at least five
tunnel rings behind theTBM. Thedeveloping grout
cakeincreasestheintegral stiffnessof thegrout layer
andhencethetunnel isfounded.
Theassociatednethydraulicresistanceof surround-
ingsoil issubstantiallyhigher thanaccordingtoradial
flowtheory.Thishigherresistanceisattributedtoben-
toniteoriginating fromthefaceof theTBM, having
invadedthegrout-soil interface.
Giventhehighfrictional resistancemeasuredover
thesupply linesandthesmall vertical grout pressure
gradients behind theTBM, it is concluded that this
groutisprovidingsignificantfrictionwiththeconcrete
tunnel lining(contrarytocementiousgrout employed
inpreviousprojects).
For a number of reasons the pressure readings at
GHT werelessreliablethanatothertunnels.TheGHT
dataisnonethelessvaluablebecauseitshowsthegen-
eral applicabilityof theresultsfromearlier researchat
SophiaRail Tunnel, it isprovidingdataontheperfor-
manceof cement-lessgroutwithhighyieldstressand
it showsamarkedinfluenceof vertical movementsof
thetunnel liningonthegrout pressuredistribution.
ACKNOWLEDGEMENT
The work was conducted as a co-operation between
Delft Cluster, HSL-South Organisation and Centre
for Underground Construction. The authors grate-
fullyacknowledgeHSL SouthOrganisation, andtheir
treasurer the Dutch Public Works Department, and
Centrefor UndergroundConstructionfor permission
topublish.
REFERENCES
Aime, R., Aristaghes, P., Autuori, P. & Minec, S. 2004. 15m
Diameter Tunneling under Netherlands Polders, Proc.
Underground Space for Sustainable Urban Development
(ITA Singapore), Elsevier.
Bezuijen, A. 2004. Calibration grout pressure sensor
GHT/Results of measurements (in Dutch: Ijking grout-
drukopnemers GHT/Resultaten metingen), COB report
F512-O-04-127.
354
Bezuijen, A. 2004. Vintestsgrout GHT/Resultsof measure-
ments (in Dutch: Vinproeven op grout GHT/Resultaten
metingen), GeoDelft 403050/0008.
Bezuijen, A. 2005. Analysis of pore pressures close to
the tunnel; influence of grout flow, (in Dutch: Analyse
waterspanningennaast tunnel; invloedgroutstromingop
waterspanning), COB report F512-05-04.
Bezuijen, A. & Talmon, A.M. 2003. Grout the foundation
of aboredtunnel, 2003, Proc. int. conf. on Foundations
Innovations, Observations, Design & Practice (ICOF
2003 Dundee), ThomasTelford, London.
Fieldmeasurementsongrout pressuresduringtunnelling
of theSophiaRail Tunnel, J. of Soils and Foundations of
the Japanese Geotechnical Society, Vol.44,no.1,pp3948.
Shirlaw, J.N., Richards, D.P., Ramond, P. & Longchamp, P.
2004. Recent experienceinautomatic tail voidgrouting
withsoftgroundtunnel boringmachines, inproc. Under-
ground Space for Sustainable Urban Development (ITA
Singapore), Elsevier.
Talmon, A.M. & Bezuijen, A. 2005. Grouting thetail void
of bored tunnels: the role of hardening and consolida-
tion of grouts, 5th int. Symp. Geotechnical Aspects of
Underground Construction in Soft Ground, 1517 June,
ISSMGE-TC28, Amsterdam, TheNetherlands.
Talmon, A.M., Bezuijen, A., Aanen, L. & vander Zon, W.H.
2001. Grout pressures around a tunnel lining, proc.
IS-Kyoto conference on Modern tunneling Science and
Technology, pp817822.
355
Longitudinal tubebendingduetogrout pressures
A.M. Talmon
Delft Hydraulics & Delft University of Technology, Delft, The Netherlands
A. Bezuijen
GeoDelft & Delft University of Technology, Delft, The Netherlands
F.J.M. Hoefsloot
Fugro Ingenieursbureaux B.V., Leidschendam, The Netherlands
ABSTRACT: A mathematical relation is described between the distribution of grout over injection ports,
rheological properties and vertical grout pressure gradient immediately behind aTBM. This vertical grout
pressuregradient isanimportant loadingparameter inbeamactioncalculationof tunnel linings. Longitudinal
momentsintheliningareof importanceintunnelingespeciallyfor thedesignof thelining. Thesemomentsare
determinedby theunevendistributionof jack forces fromtheTBM but also by thebuoyancy forces that are
exertedbythegrout that isstill intheliquidphaseclosebehindtheTBM.
1 INTRODUCTION
Earlier work (Bezuijen & Talmon 2005) has shown
that the grout properties have a large influence on
the longitudinal moments and that it is even possi-
blethatthesemomentspreventanadequatetunnelling
process.
This earlier work did not take into account that
the tunnel is stage constructed and that a specific
beam calculation method has to be used to take
that into account. An analytical staged beamcalcu-
lation will be used in combination with measured
andcalculatedgrout pressuredistributiontocalculate
longitudinal bendingmomentsinthelining.Thecalcu-
lationmethoditself will bedescribedinanother paper
tothisconference(Hoefsloot, 2008).
The vertical grout pressure distribution immedi-
ately behind a TBM is governing to a large extent
the stresses in the tunnel lining. This vertical grout
pressure distribution is determined by the injection
strategy and rheological properties of the grout. A
finitedifferencemodel waspublishedbyTalmonetal.
(2001) for thecalculationof thegrout pressuredistri-
bution up to adistanceof afewtunnel rings behind
a TBM. Using this model it was learned that there
might beasimplerelationbetweenthedistributionof
grout injection rates over theinjection ports and the
vertical grout pressure gradient immediately behind
aTBM.
2 SIMPLE MODEL FORVERTICAL
GROUT PRESSURE DISTRIBUTION
AT TBM
2.1 Injection conditions
Many TBMs are equipped with six injection ports.
Typicallytheflowratethroughtheportslocatedinthe
upper half of theTBM islarger thaninthelower half.
Thisway, thegrout isforcedtoflowdownwards, par-
tially. As a consequence, the vertical grout pressure
gradient will belower than thestatic grout pressure
gradient. Such a grout injection pushes the tunnel
downwards. A definitionsketchof grout distribution
isgiveninFigure1. Dataontypical grout injectionat
GroeneHartTunnel isgiveninTable1.
Theouterdiameterof theGroeneHartTunnel lining
is14.50m.Thetheoretical thicknessof thegroutlayer
is0.18m.
2.2 Mathematical modeling of tangential
grout flow
Eachinjectionporthasitsownsupplyarea. Groutflow
velocitiesinthegroutingannulusaresmall. Giventhe
Bingham-likerheological character of grout, only its
yield-stress is relevant. As a consequence, for both
clock and anti-clock wise direction, grout pressures
decay linearly frominjection ports. Grout pressure
357
Figure 1. Definition sketch tangential grout distribution
behind aTBM, grout supply area and the z-coordinate of
thecentreof gravityof grout injection.
Table1. Measureddistributionof grout flowratesover the
injectionportsat GroeneHartTunnel.
vertical contributionto
Injection position position total grout
point [degrees] [m] injection
1 8.4 7.32 0.2
2 65.2 3.10 0.178
3 122 3.92 0.156
4 180.9 7.40 0.133
5 237.7 3.95 0.156
6 294.5 3.07 0.177
total 1
decayhaltsuponconvergencewithgroutfromadjacent
injectionports.Thewidthof supplyareaisdictatedby
therelativeinjectionratesof associatedinjectionports.
Considering the width of supply area, and the fact
that no discontinuities ingrout pressureareallowed,
thepreciselocationsof thestagnationpointsbetween
supply area are calculated. Next, with yield stresses
on both sides of the grout layer, the grout pressure
distributioniscalculatedmathematically.Thelocation
of injectionportsandthedefinitionof supply areais
giveninFigure1.
An example of the calculation results for Groene
HartTunnel isgiveninFigure2.Inthiscalculationboth
sides of the grout layer experience tangential shear
stressesequal totheyieldstressof freshgrout.Physical
propertiesarelistedinTable2.
In reality thereareno sharp discontinuities at the
confluence of the supply area. These discontinuities
are an artifact inherent to the employed schemati-
zation. The calculated discontinuities are acceptable
becauselocally clock-wiseand anti-clock-wiseflow
compensate.
Figure2. Calculatedgroutpressuredistributionfromgrout
injection. Injectionportsareindicated. A linear approxima-
tion of calculated grout pressure distribution is indicated.
Thetheoretical staticgroutpressuredistributionisshownfor
comparison. Groutingparameters areaccordingtoTable1,
physical parametersareaccordingtoTable2.
Table2. Input-parametersfor thecalculationof
groutpressuredistributionatGroeneHartTunnel.
D
tunnel
[m] 14.5
h
annulus
[m] 0.18
grout
[kg/m
3
] 1850
y
[Pa] 1500
P
crest
[Pa] 200000
2.3 Simplification of vertical pressure gradient
calculation
Theloadingforce(F) of grout onatunnel lining, per
unit length, isgivenby:
The average vertical grout pressure gradient might
bedetermined fromlinear approximations as shown
in Figure 2. This is however not the most straight-
forward method. Themathematical model presented
in Section 3 gives grout pressures at each circum-
ferential position. By integration of calculated grout
pressurestheloadingforce(F) iscalculated. Byvirtue
of theaboveequationthecorrespondingvertical grout
pressuregradient isdeterminednext.
A number of calculationswasconductedfor differ-
ent injection strategies, where the grout distribution
between the upper and lower part of theTBM var-
ied. Upon inspection of theresults it was found that
the centre of gravity of grout injection is a decisive
358
Figure 3. Calculated dimensionless contribution of
grout-injectiontovertical grout pressuregradient. Location
of injectionportsaccordingtoTable1.
parameter. Thez-coordinateof thecentreof gravityis
calculatedby:
with: Q
i
=grout-flow-rateof injectionport (i),
Q
total
=total grout-flow-rateof all injectionports,
z
i
=vertical coordinateinjectionport (i),
z
g
=z-coordinatecentreof gravity
Definitions for thecentreof gravity areshown in
Figure1.
Calculationresultsforvertical groutpressuregradi-
entfromgroutinjection(withoutstaticgroutpressure)
areshown, inFigure3. Quasi independent of thepre-
cise distribution, a linear relation is found between
the centre of gravity of grout injection and vertical
groutpressuregradient.Withsix-injectionports, with-
out grout overlapping adjacent injection ports, the
maximumvertical positionof thecentreof gravity is
z
g
(max)=R/3, and the influence of grout injection
ismaximal: dp/dz=8/
y
/h. Inthatcasetheinjection
portatthecrestof thetunnel supplies1/3of total grout
injection.
Adding the static grout pressure to the results of
Figure3leads to thefollowingdesign-formula for
vertical pressuregradientimmediatelybehindaTBM:
For the grouting conditions at Groene Hart Tunnel,
given in Table 1, the vertical position of the cen-
tre of gravity is at z
g
=0.047R. According to the
simplified theory presented above, a vertical grout
pressuregradient of 15.6kPa/mwouldbeexpectedat
GroeneHartTunnel.Measuredvertical pressuregradi-
entsmeasuredatGroeneHartTunnel were: (5kPa/m
at ring2117, 12kPa/mat ring2118and17kPa/mat
Figure 4. Back-side of the TBM showing tail-grease
brushesandplay(=jeu) betweentunnel liningandback-side
of theTBM-shield.
ring 3241, seeTalmon & Bezuijen 2008). With the
methoddescribedinSection2.2and2.3thevertical
grout pressure gradient behind aTBM is calculated
as afunction of injection strategy and grout proper-
ties.ThismethodisagoodalternativefortheDCgrout
model. Theadvantageisthat nonumerical simulation
is needed. The disadvantage however, is that it only
presents thegradient directly after thetunnel. In the
calculationsthatfollowweassumethatin8mthisgra-
dientdecreasedownto9.3kPa/maswill beexplained
later.
2.4 Grout pressures in eccentric tail void
Thegrout pressurecalculation model of Section 2.2
assumes aconcentric positionof thetunnel liningin
thesoil cavity, asdoestheDCgrout model of Talmon
et. al. (2001). At GroeneHart Tunnel however some
7cmupwardmovementwasmeasuredbyawater lev-
elingsystemjustbeforethepressuresensorscameinto
thegrout, Talmon& Bezuijen(2008). It is unknown
if, or how, surrounding soil moved. A movement of
around2cmwasfoundin2-DcalculationsbyBezuijen
& Bakker (2008) for atunnel of 10mdiameter. How-
ever, thisresultwill dependtoalargeextendonthesoil
parameters andthey werenot obtainedfor this loca-
tion. Theresultsof these2-Dcalculationsdoindicate
that asoil movement of somecentimetersistheorder
that canbeexpected. Fromageometricpoint of view,
themaximumrelativemovement betweenthetunnel
liningandtherear of theTBM isat max 9cmbefore
thetunnel liningtouchesthetail-endof theTBM, see
Figure4.
It is concluded fromFigure 4 that the maximum
eccentricity between the tunnel lining and soil cav-
ity will beabout 45mm. Theexistingmodels do not
necessarily have to be modified. An adaptation of
inputted grout flow rates is considered a workable
alternative. For 45mmupwardeccentricityof thetun-
nel liningat GHT, theapparent centreof gravity has
beencalculatedat z
g
=0.22R.
359
Figure 5. Loading diagram tunnel lining Groene Hart
Tunnel: loadingper unit length.
Figure 3 shows that for a relative centre of grav-
ityof z
g
,R=0.22, grout injectionproducesavertical
pressure gradient of dp/dz =1.6
y
,h =13.6kPa/m
for a yield stress of
y
=1.5kPa and a tail void
thickness of h=0.18m. Addition of thestatic grout
pressure gradient (at =1850kg/m
3
) gives for the
calculatedvertical pressuregradientbehindtheTBM:
dp/dz =18.113.6=4.5kPa/m. This is closeto the
measured initial vertical gradient of 4 a 6kPa/mat
ring2117.
2.5 Different positions of injection ports
The present application applies to the layout of the
GroeneHartTunnel intheNetherlands. Similar calcu-
lationsshouldbeconductedfor other positionsof the
grout injectionports, for instancefor alayout suchat
Botlek Rail tunnel, wherethereisnoinjectionport at
thecrest. Expectedisthattheresult(Figure3) will not
differ much.
3 BEAMACTIONCALCULATION
3.1 Tunnelling conditions
The(local) uniformloadsinthetunnel aresketchedin
Figure5.
Theforces exerted by themain jacks of theTBM
arecalculatedfrommeasuredhydraulic oil pressures
andpiston/barrel constructionof thehydraulic cylin-
ders. Duringexitof thefirstringsof theCOB-passage,
ring2117, thepoint of applicationof theaxial force
was1.5mbelowtheTBM-axis: thebendingmoment
exertedbytheTBM jackswas79MNm.
ThebendingmomentexertedbytheTBMjackswas
foundtobevirtually identical for eachring, withthe
exceptionof thosefirstringsof theCOB-passage, see
Figure6. A representativevalueis65MNm. Assum-
ingthattheother external influences(transverseforce
fromTBM, vertical groutpressuregradientbehindthe
TBMandlongitudinal uniformityof soil) remaincon-
stant, atime-distancetransformation of axial strains
measured in tunnel ring 2117 produces the bending
moment-curveof thetunnel liningshowninFigure7.
Figure6. Bending moment exerted by TBM jacks during
drillingphase.
Figure7. Measuredbendingmomentintunnel lining(com-
putedfromstrain-gaugesprecastedintunnel segments).
Figure8. Averageinjectionpressureof six grout injection
linesandthevertical positionof thecentreof gravityof grout
injection.
At thebeginningof theCOB-passagethevariation
inbendingmomentexertedbyTBMjacksisprecisely
synchronous with the bending moment calculated
fromthestraingaugesintunnel ring2117.
Figure8shows thecourseof grout injectionpres-
sure(averageof all six injectionlines) measured4m
beforetheoutletof theinjectionpipes.Alsothecentre
of gravityof groutinjection, calculatedbyEquation2,
360
Figure9. Averageinjectionpressureandvertical pressure
gradient of grout injectionlines, measuredat theinlet of the
injectionpipes.
isshown. Grout injectionpressuresvary between4.5
bar and7.5bar. Thevertical positionof thecentreof
gravity of grout injectionvaries between0and0.2R
abovetunnel-axis.
For aconcentricconfigurationof tunnel liningand
soil cavity, at agrout yieldstressof 1.5kPaandtwo-
sided friction of the grout layer, this corresponds to
a variation of vertical grout pressuregradient at the
TBM between7.1-dp/dz-18.5kPa/m.
Figure9showsthevertical pressuregradientcalcu-
latedfrominjectionpressures. Basicallythisgradient
variesbetween0and10kPa/movertheCOB-passage.
Thisgradientislowerthanmeasuredwiththepressure
sensorsonthelining. Itcouldbethatthepressuredrop
over thefour meterspipelengthishigher intheupper
part of theTBM, becauseof higher grout flowrates
(typical grout flowvelocityis0.3[m/s]).
TBM jacksmaytransfer transverseforcesbetween
the tunnel lining and TBM. The magnitude of such
forces should be limited, because jacks are not
designed for that. This unknown transverse force is
howeverimportanttobeamactionof thetunnel lining,
andhastobedetermined.
Inthetail voidthesegmentsaresupportedbyforces
fromthegrout layer. InsidetheTBM tunnel ringsare
unsupported. At the Groene Hart Tunnel the typical
length of unsupportedtunnel liningin theTBM is 4
meters.
Vertical support by fluidgrout isdeterminedfrom
grout pressures measured around the tunnel lining.
Uponexit vertical grout pressuregradients as lowas
6kPa/moccurred,thatafter3mprogressstabilizedtoa
gradientof 12kPa/m(Talmon&Bezuijen2008). This
latter valueisconsideredarealisticandrepresentative
valuefor grout loadingjust behindtheTBM.
3.2 Result beam action calculation
The staged beamaction model described by Hoefs-
loot (2008) is used. Input-parameters are the above
Figure 10. Measured and calculated bending moments
compared(GroeneHartTunnel, theNetherlands).
mentionedbendingmomentbyTBMjacks, transverse
forcebyTBM, vertical groutpressuregradientbehind
theTBM, loadingdiagramandunsupportedlengthof
tunnel liningintheTBM.
Model-specific is the so called end of the fluid
groutzone. Inthestagedbeammodel, thisistheloca-
tionwherethereaction-forcefromthesoil iszero: new
elementsaremathematicallyaddedinastress-lesscon-
dition. Heretheassociatedvertical pressuregradient
acrossthetunnel liningis9.3kPa/m(weight of grout
layer hasbeensubtractedfroma10kPa/mhydrostatic
pressure).Themeasuredpositionof 9.3kPa/misposi-
tionedabout tenmetersbehindtheTBM. Thevertical
grout pressuregradient is assumed to decay linearly
from12kPa/mat theTBM to9.3kPa/mat theendof
thefluidgrout zone. Consequently theexternal load-
ing by fluid-grout connects smoothly to thereaction
forcefromthesoil. A further secondorder condition
isthattheaxial gradientof vertical groutpressuregra-
dientandreactionforcefromthesoil areidentical.An
endof fluidgroutzonesituated8mbehindtheTBM
satisfies thesecriterion. However, it is not clear how
general applicable the implicit boundary conditions
are. Thiswill besubject tofurther research.
Themodel requires also informationonstructural
properties: bending stiffness of the lining and mod-
ulus of subgradereactionof surroundingsoil. These
haveto comply with fundamental properties of seg-
mentedliningandsurroundingsoil, but also haveto
complywiththevertical displacementsthathavebeen
measured(Talmon&Bezuijen2008).Thereforelower
valuesareusedthenonewouldapplyonpuretheoret-
ical grounds. The following values have been used:
EI =3.2 10
9
[kNm
2
] andk=7.3 10
4
[N/m
2
].
A modest downward force of 1.5 [MN] fromthe
TBM is needed to reach perfect agreement between
measuredandcalculatedbendingmomentcurveinthe
tunnel lining, seeFigure10.
At ring 2117 and ring 2118 a total of four incli-
nometersmetershavebeenattachedtoliningsegments
361
Figure 11. Measured and calculated inclination of the
tunnel (GroeneHartTunnel, theNetherlands).
situated at tunnel-axis-level. These four instruments
producedexactlythesameresults. Figure11showsthe
resultsof oneof these. Itshowsthatafter theTBMthe
tunnel inclinationisof theorder 1:1000(comparedto
itsfinal orientation)andthatinclinationdecreaseswith
distance. The graph shows that there remains a dis-
crepancy betweeninclinationcalculatedby thebeam
actionmodel andmeasuredinclination.
4 CONCLUSIONS
Grout injection:
Thecentreof gravity of grout injection, incombina-
tionwithgrout yieldstress, isshowntodeterminethe
vertical grout pressuregradient immediatelybehinda
TBM.
Incaseof aneccentricpositionof thetunnel lining,
theinfluenceof eccentricityongroutpressuresisalso
quantifiable.
Infutureapplicationof beamactionmodelsfortun-
nel linings, the model of Section 2 can be used to
calculate beamloading caused by the vertical grout
pressuregradient behindtheTBM.
Beam action tunnel lining
The measured bending moment curve in the tunnel
liningis well reproducedby thebeamaction model.
Theway howthetransition fromliquidgrout with a
constant force on the tunnel to elastic behaviour is
incorporatedinthemodel needsfurther study. Simu-
lationshaveshownthatthebuoyancyforcesthatoccur
at the location where the tunnel lining is floating
inliquidgrout areessential tosimulatethemeasured
bendingmoment.
Lower values for thebendingstiffness of thetun-
nel lining and the modulus of subgrade reaction of
surroundingseemtoapply thanareexpectedonpure
theoretical grounds.
The construction phase of a bored tunnel leads
to axial forces that remain in the tunnel lining after
completion(Blom2002). Theconstructionphasealso
producesaremainingbendingmoment, aswasfound
at the Groene Hart Tunnel. The associated point of
application of this axial force is situated about 1.5
meter abovethetunnel axis.
ACKNOWLEDGEMENT
The work was conducted as a co-operation between
HSL-SouthOrganisationandCentrefor Underground
Construction. The authors gratefully acknowledge
HSL SouthOrganisation, andtheirtreasurertheDutch
Public Works Department, and Centre for Under-
groundConstructionfor permissiontopublish.
REFERENCES
around a TBM machine. Proc. 6st Int. Symposium on
Geotch. Aspects of Underground Construction in Soft
Ground, Shanghai
Bezuijen, A. & Talmon, A.M. 2005. Grout properties and
their influenceonbackfill grouting. Proc. 5th Int. symp.
Geotechnical aspects of underground construction in soft
ground, Amsterdam1517J une.
Blom, C.B.M. 2002. Design philosophy of concrete linings
for tunnels in soft soil. PhD-thesis, DelftUniversityPress.
Hoefsloot, F.J.M. & Bakker, K.J. 2002, Longitudinal effects
bored Hubertus tunnel in The Haque. 3rdInternational
Construction in Soft Ground ISToulouse, 2325october
2002.
Hoefsloot, F.J.M. 2008. Analytical solution longitudinal
behaviour tunnel lining, TC28Shanghai.
Koek, A.J., Bakker, K.J. & Blom, C.B.M. 2006. Axial pre-
stresses in the lining of a bored tunnel. 5thint. Symp.
Soft Ground, 1517 June, ISSMGE-TC28, Amsterdam,
TheNetherlands
Talmon, A.M. & Bezuijen, A. 2005. Grouting thetail void
of bored tunnels: the role of hardening and consolida-
tion of grouts. 5thint. Symp. Geotechnical Aspects of
Underground Construction in Soft Ground, 1517 June,
ISSMGE-TC28, Amsterdam, TheNetherlands.
Talmon, A.M. & Bezuijen, A. 2008. Backfill grouting
researchatGroeneHartTunnel. Proc. 6st Int. Symposium
on Geotch. Aspects of Underground Construction in Soft
Ground, Shanghai.
Talmon, A.M., Bezuijen, A., Aanen, L. & van der Zon,
W.H. 2001. Grout pressuresaroundatunnel lining. Proc.
IS-Kyoto conference on Modern Tunneling Science and
Technology, pp. 817822.
Talmon, A.M., Bezuijen, A. & Hoefsloot, F.J.M. 2007. Eval-
uatietweedefaseGroeneHart tunnel: Ontwerpparame-
ters bij krachtswerking, groutdruk en invloed beweging
tunnelbuis. COB-report, COB-F512-07-02.
362
Theme 3: Case histories
Tunnel facestabilityandsettlement control usingearthpressurebalance
shieldincohesionlesssoil
A. Antiga
Soil S.r.l., Milano, Italy
M. Chiorboli
Metropolitana Milanese S.p.A., Milano, Italy
ABSTRACT: Softgroundtunneling, inurbanareasaregenerallyboredusingpressurizedTBM.Tunnel design
involves theneedto revaluetheclassical concepts of deformationresponseto theexcavationandthecontrol
priorities, focalizedonfaceandcavity, loosetheir importance; theevaluationof thesettlements becomes the
main aspects to consider. The definition and the control of the face pressure are fundamental design steps.
It is important to giveaproper importanceto thefacepressurerelating to settlements control. Thepressure
increaseleadstohigher safetyfactorsof thefacestabilitybut it isnot abletoguaranteecomparablesettlements
decrease. Weexaminestheessential parametersinrelationtotheuseof theEPBSintwocasehistoriesinMilan:
line1extensiontothenewFair andPassanterail infrastructure. Thereferencevaluesof thecitedparametersare
identifiedfor cohesionlesssoil i.e. fluvio-glacial sanddeposit andgravel.
1 INTRODUCTION
Inrecent years, thegrowingneeds for transportation
infrastructurehas leadto asharpincreaseintheuti-
lization of underground. During 1993 and 1994, an
EPBSwithadiameter of 8.03mwasusedinthearea
of thePassante(undergroundrailwayline) inMilan.
Thismethod, withadiameter of 6.56m, wasthenused
in2004toextendtheMilansubway tothenewtrade
fair of Rho. Thisdocumentconsidersthemainproject
aspects relating to the use of the EPBS, with refer-
ence to the two cited projects in Milan and so to a
geotechnical context of cohesionless soil under high
water head. The reference values for the basic tun-
nel designparameters aredefined; areferencevalue
for theparameter i (point of inflectioninthetrans-
versesettlement trough) is identified. Thedocument
also describes how, given the values necessary to
guaranteethefacestability, thetunnel facepressure
wouldnotinfluencethefinal settlements, andhowthis
parametercouldbebestdefinedusingonlycalibration
duringthework.Thetheoretical pressurevaluestouse
canbedeterminedusingsimpleequilibriumratiosin
K0conditionsasinitial startingvaluesfor tunnelling
operations.
2 MECHANIZEDTUNNELLINGDESIGNIN
URBANENVIRONMENT
Peck (Peck 1969) highlightedthebasicpointstocon-
sider whenplanningundergroundworksinthecaseof
soft ground:
1 thestabilityof thetunnel withparticular reference
to thestability of thetunnel faceand of thezone
immediatelybehind;
2 theevaluationof thedeformations at thefaceand
aroundthetunnel and, inthecaseof atunnel with
alowdepth, theresultingsuperficial subsidence;
3 thedefinitionof stabilizationactionstoguarantee
compliancewithconditions(1) and(2) andof the
liningstructuretoensurethelong-termstabilityof
thetunnel.
Withtheconstructionof tunnelsinurbanenviron-
ment, the control of superficial subsidence and its
effects assumes primary importance with regard to
thoseitems listedabove. Theconditions for stability
of thefaceandof thecavity andtheacceptability of
respectivedeformations, with regard to thebreaking
values, areconditionsthatarenecessarybutnotsuffi-
cient toguaranteethevalidityof agivenconstruction
365
Figure1. Buildingsubsidence: mainparameters.
systembecausetheyarenotabletoguarantee, apriori,
theacceptability of thesubsidence. This statement is
particularlyvalidinthecaseof tunnellingsystemswith
shieldmachinesandfull-facepressurizationforwhich
thereisnodirect control of thestrainstateof theface
andof thecavityduringtheconstructionphase, andfor
whichthemostsignificantquantitativeparametersare:
stabilizationpressureat theface
volumeof extractedsoil
superficial subsidence
3 SUBSIDENCE EVALUATION
Theprocessof designfor tunnelsinurbanareasmust
bebasedonthesettlement valueparameter.
Theobjectivesof thesettlement analysisshouldbe
asfollows:
definitionof acceptablesettlement values
study of theprocessthat generatesthesubsidence
andthecontrol methods
technical and economical optimization of the
projectbydefiningtheconstructionmethodstouse
andanyrequiredmitigationactions.
BurlandandWrothidentifiedtheparameterstocon-
sider when valuating subsidence(figure1) (Burland
and Wroth 1974). Rankin subsequently focussed on
thedeflectionratioDR=L
max
/L andonthehorizontal
deformation, hemoreover showedthat inengineering
practice, youcanrarelyrelyonthecompletedataiden-
tifiedby BurlandandWroth(Rankin1988). Similar
toanormal Gaussianprobabilityfunction, theprofile
of thetransversesettlement troughisconsolidatedby
a mass of literature on the subject and is proven by
experience.
The curve that defines the transverse settlement
troughischaracterizedbytwoparameters:(a)themax-
imumsettlement S
max
(corresponding to the tunnel
Figure2. Transversesettlement trough.
axis), (b) thedistancei betweenthetunnel axisand
thepoint of inflection in thesubsidenceprofilethat
definesthesizeof thetransversesettlement trough.
Thevertical settlement at distancey onthetunnel
axisisgivenby:
Thetotal volumeof thesubsidencetroughV
s
(for
theunitlengthforthetunnel) canbeobtainedfromthe
integrationof (1) andtheresult is:
Thevolumeof thesubsidencetroughV
s
isdirectly
dependent onthevolumelossV
p
(volumeof soil that
isexcavatedinexcessof thetheoretical volumeof the
tunnel).
In agreement with the proposal by OReilly and
New, for z >D (D=diameter of the tunnel), the
parameteridependsonthetypeof soil andthedepth
of thetunnel andappearstobeindependentonthetun-
nel diameter and thetunnelling mode(OReilly and
New1991); giving:
wherez isthedepthof thetunnel andK isacoefficient
that dependsonthetypeof soil.
For cohesive soil, the value of K is generally
between0.4and0.6. For sandandgravel, thevalues
for K aremoredispersed, however theusual rangeis
between0.25and0.45(Mair 1997).
Fromthefunction(1) youcanderivetheslopeand
curveof thetransversesettlement trough, factorsthat
arerelevant onthepre-existent structures.
Tosummarize, thevalueof thesettlement isthere-
foredependent oni andonV
p
throughV
s
.
366
Figure3. EPB Shield_ Facestabilizationpressure.
In the case of non-complex situations that can
be well planned following the hypothesis described
above, there are no significant improvements in the
reliabilityof theresultsafterswitchingfromanempiri-
cal Gaussianmethodtoamoresophisticatednumerical
model.
4 FACE STABILIZATIONPRESSURE
The definition and the control of the face stabiliza-
tionpressurerepresentsoneof thebasicstagesinthe
constructivephase. Thedeterminationof this param-
eter intheplanningphase, however, giverisetosome
uncertainties: despite there being a large amount of
literature, thereis not yet agenerally agreedanalysis
methodandat thesametime, therearenonormative
references. Also, ignoring theuncertainties resulting
from the geotechnical characterization of the soil,
usingthedifferent approachesdiscussedinliterature,
you obtain frequently significant differences in the
theoretical valuestoapply.
However, it isnecessary tolay proper emphasisto
the face stabilization pressure in terms of the final
subsidenceresult.
Ononehand, theincreaseinvalueof thepressure
applied to the face guarantees higher safety coeffi-
cientsforthefacestability; however, ontheotherhand,
acorrelatedreductioninsettlementsatthesurfaceare
not absolutelyguaranteed.
Thedefinitionof thestabilizationpressuretoapply
mustachievetheobjectiveof guaranteeingthestability
of thefacewhilechangingtheinsitustressaslittleas
possible. In afully-operational situation, you should
trytoexcavateinK0condition, thatiswithaconstant
advancerate.
A study conducted by AFTES identified how the
ideal advance condition, for minimizing the defor-
mations and ensuring thefacestability, is a balance
betweenthevolumeof material extractedandthethe-
oretical tunnel volume and how with this condition
thepressureremains constant. Inthis condition, you
could advancewith aconstant volumeand thepres-
sureapplied to thefacewould beequal to theearth
pressureat rest (i.e. K0conditions); by checkingthe
twoidentifiedparameters(volumeandpressure) you
would obtain the definition of the correct face sta-
bilizationpressure(AFTES2001). Asaresult, during
constructionphase, theresearchshouldconcentrateon
theachievement of thebalancebetweenthevolumes,
boththeoretical andextracted, basedonstabilization
pressurethat issufficientlycautious.
Themachines usedinrecent years, mainly dueto
thecorrectdefinitionof thesoil conditioningtechnolo-
gies, enablegoodcontrol of pressuresand, therefore,
of the stability of the tunnel face, as we will see in
theMilanexamples. Withthecontrol of thepressure,
you can influencethesituation ahead of thefaceof
thetunnel and this represents acondition necessary,
but not sufficient, toguaranteesubsidencecontrol. In
fact, to greatly increasethepressurewouldbringthe
soil aheadof thecuttingheadinconditions nearer to
that of passivepressurewith an increasein thetotal
pressuretoapplyand,asaresult,withaseriesof effects
that bringabout areductionintheadvancespeedand
operatingconditionsthatarelessfavourableandmore
risky(for example, anincreaseintherequiredtorque,
an increasein energy consumed, an increasein tool
wear, anincreaseof temperatureinthetunnel working
chamber asaresult of increasedfrictionwiththepos-
sibilityof creatingmaterial blocks).Thesoil infrontof
thefacedoesnotrecordanyexcesspressureappliedto
thefaceinthesuccessivephasesinwhichit issubject
tothephenomenaof looseningthatresultsfromthedif-
ferentannular cavitiesinthepassageof theshield, i.e.
over-cut, duetothetoolsattheperipheryof thecutting
wheel (thestatement appliestoincoherent earth).
To summarize, thepressureintheworkingcham-
ber of theTBM must bemaintainedat alevel capable
of ensuring stable working conditions with a safety
coefficient that allows theabsorption of thefluctua-
tionsresultingfromadynamic situation. Thisshould
besufficient toavoiduncontrolledcollapsesof mate-
rial within the working chamber but not so that it
causes soil deformations and blow-up with aconse-
quent lossof thestabilizationconditionedmixturefar
fromtheface(AFTES 2001). Thedefinitionapriori
of thefacestabilizationpressurewould, infact, bea
secondary element only guided in choosing the ini-
tial value; it is optimizedduringtheprocess of work
onthebasis of thebalanceof theexcavatedsoil vol-
ume. Theproblemsrelatingtothesubsidencerequire
maximumattention also to other parameters, in par-
ticular advancerateandbackfillingoperationsbehind
theface. Research, moretechnological thantheoreti-
cal, must alsoanalysethemechanismsthat riseahead
of the face: grouting systems above the shield have
been recently introduced, mainly using polymers or
bentonitemuds.Thesemuds, injectedatpressure, hold
the soil above the shield for the time period from
the passage of the head to the unthreading of the
shield and the related final backfilling with cement
mortar.
367
Figure4. EPB Shield _ Individual factors of volumeloss
VL.
5 VOLUME LOSSVL
The volume loss VL represents the fundamental
index parameter for thevaluation of thesurfaceset-
tlement and, consequently, of thedisturbancecaused
by thetunnelling. Thetotal volumeloss is derived
fromthesumof various components whoserelative
contributiondependsonthemechanical andphysical
characteristics of thesoil, onthetechnical character-
istics of the tunnelling equipment, on the execution
method, with particular attention to the pressures
appliedatthefaceandaround, andontheadvancerate.
Thesefactorsareessentially(figure4) (Mair 1997):
volume loss at the face (face loss): owing to the
deformationof theportionof soil aheadof theface
of thetunnel andto theloss resultingfromlocal-
izedmicro-stabilityproblems. Thislossof volume
iscounteractedbythefacestabilizationpressure.
volume loss around the shield essentially due to
theconvergenceof thetunnel profiletowards the
extrados of the shield machine. This factor can
besubdividedintothefollowingterms: (2a) over-
cutting due to greater diameter of the tunnelling
cutterheadcomparedtotheshieldmachine, neces-
sary to reduce the friction of the shield machine
onthesoil andtofacilitatesteering; (2b) planoal-
timetriccharacteristicsof thetrail; thepresenceof
narrowcurveranges results inover-cuttingvalues
higher than thosepresent in thestraight sections.
This aspect is accentuated by possible misalign-
ments caused by the operator who may let the
machinezigzag; (2c) over-cuttingcausedbypossi-
bleconeshapedof theshield; (2d) roughnessof the
shieldwhichcanleadtomovementsandsubsidence
duetothefrictionagainst thesoil.
volumelossaroundthelinedtunnel sectiondueto
thegapbetweenthetunnel profileandtheextrados
of thepre-cast segments; thisresultsinatendency
toconvergetowardstheextradosof thelining.
volumelossduetodeformationsof thelining.
long-termeffects/consolidation.
Withregardtothefivepointslistedabove:
thecomponentrelatingtothestressrelief of thesoil
atthefacewiththeuseof pressurizedfacetunneling
machines becomes unimportant if thepressureat
thefaceisadequatelycontrolled.
the volume loss during the passage of the shield
is difficult to counteract and can be the weakest
element; thelimitationof thiscomponent isabove
all toreducethetechnologicalover-cuttingtothe
minimumpossible.Inrecentyears,shieldmachines
that allow annular grouting for filling the region
around the shield have been used. It can also be
added that holding an high advanceratebrings a
significantreductioninthiscomponent, preventing
that thegapclosescompletely.
the volume loss around the lined section of tun-
nel isminimizedbyusingcement mortar fromtail
of theshield; however problemsmay still occur in
the execution which render this component criti-
cal (insufficient groutingpressures, washingaway
of the mixture etc.). In soil that has a self-stable
capacity adequate to the advance rate, the injec-
tionof cement mortar cancreatearecompression
of thegroundandallowstherecovery of apart of
thevolumedefinedinpoint 2.
isingeneral unimportant comparedtoothers.
issignificant insoil subject toconsolidation.
Mair showedthat, inthecaseof Slurry Shields or
EPB-S, possiblereferences areVL 0.5%for sandy
soil andVL =12%in soft clay (Mair 1997). The
definition of areliablevalueis, however, very diffi-
cult indesignphasesbecausethisparameter depends
on factors that are highly dependent on the specific
contextandcannotbegeneralized(soil, technological
factors of tunnelling equipment, workmanship skill,
hydrogeologyetc).
6 EPB-STUNNELLINGINMILAN
Between2004and2005, parallel to theconstruction
of thenewtradefair inRho/Milan, line1of thesub-
waywasextended. Thetrack, withalengthof around
2.1km, ischaracterizedbythepresenceof highwater
tableconditions(upto12m) andvariabletunnel depth
between10and20m; twotunnelswithonesingleplat-
formwereexcavated using an EPB-S machine. This
methodof tunnellingwaspreviouslyusedinMilanin
1993/1994 during the construction of the Passante
Ferroviario (undergroundrailway line) for asection
of around4kmwhendiggingtwotunnelswithvariable
depthsbetween4and16m.
Some details of the two works are reported by
Chiorboli and Marcheselli (Chiorboli, Marcheselli
1996) and by Cavagna and Chiorboli (Cavagna,
368
Chiorboli 2004); the next table reports the main
data:
PASSANTE
LINE 1 FERROVIARIO
CLIENT MUNICIPALITY MUNICIPALITY
OF MILAN OF MILAN
REGIONOF
LOMBARDY
BUILDING TORNO ICLET TORNO CMB
CONTRACTORS FIAT COGEFAR
ENGINEERING IMPRESIT
LODIGIANI
COLLINI P&C
TETTAMANTI
PERIOD 20042005 19941995
The Milan area is characterized by fluvio-glacial
depositsconsistingof aheavy deposit (5060m) of
gravel and sand with amediumto high density. The
typical, grainsizedistribution, whichcanbedefined
astheaveragegranulometriccurveis:
D-0.074mm=1020%
D>2mm=70%
D>10mm=1020%
7 FACE STABILIZATIONPRESSURE
Themaximumsettlement valuesarealwaysshownin
associationwiththelowestpossiblepressurevalues, in
termsof minimumpressurevaluepeaks. Thesemini-
mumpressurepeaks, if just momentary, reachvalues
significantlylower thatthevaluesrequiredfor thesta-
bilityandtheyare, almostalways, appearedwherethe
conditions around (exit or entry of the machine in
theconstruction pit,) or operating conditions (initial
start phase, mechanical rests) cause a loss of pres-
sure that is difficult to control. If you consider this
intermsof averagerecordedpressurevalue, youcan-
not define an effective link between the increase in
pressure and the reduction in final settlement. The
two following figures show a summary of the aver-
agerecordedpressurevalues, thetheoretical pressure
valuesinK0conditionsandthemeasuredsettlement
for thetwocitedworks.Youcanseethat byassuming
apressurevaluecloseto thetheoretical valuein K0
conditionsasthestart valueof themachine, thepres-
sureissubsequentlyreduced: themeasuredsettlement
doesnot showrelativeassociatedincreases.
8 TRANSVERSEANDLONGITUDINAL
TROUGHS
The transverse and longitudinal settlement troughs
havebeenanalysedinbothcases.Thetroughsinareas
Figure5. Line1: settlements facepressure.
Figure6. PassanteFerroviario: settlements facepressure.
disturbed by thepresenceof buildings or grouting
treatmentshavebeenignored. Withregardtothecase
of line1andtheundergroundrailway line, n. 18and
n. 5transversesettlement troughstogether withn. 78
and n. 28 longitudinal settlement troughs havebeen
analysed. Thegraphic infigure7plots thevalues of
the position of the transverse point of inflection i
relating to the tunnel depth and the corresponding
interpolating lines; thesamegraphic also shows the
valuescomparedwiththerangeof literatureindicated
byMair for sandandgravel (Mair 1997). Notethatfor
bothworks considered, thereis agoodconcentration
of dataaroundanidentifiablevaluesuchas:
It is possible to conclude that for the soil of the
MilanareathereferencerangeK forthecalculationof
settlementswiththeGaussian formulais:
Figure8shows, forbothcases, thelongitudinal sub-
sidence profiles with settlements normalised to the
369
Figure7. Transversepoint of inflection tunnel depth.
Figure8. Normalizedlongitudinal settlement troughs.
maximumsettlement(i.e. ratioof settlementstomax-
imumsettlement ispresented) theabscissashowsthe
position of the face, normalized to the diameter. In
terms of absolutesettlements, they showsimilar val-
ues inthetwocases whilethetwopercentagecurves
showdifferent values.
For theline1 it is observedthat around40%of
thesubsidenceisduetothepassageof theshield(9m
at theback of thehead) and, consideringthat around
25%isaheadof theface, theremaining35%isbehind
themachineingroutingphasesfor theannular space
intheextradosof thesegments. Thiswouldappear to
showgoodmanagement of thepressureinthetunnel
workingchamber althoughthesubsidencepercentage
behind the shield appears rather high. For the Pas-
santeFerroviario, youcanseethat at thetunnel face
anaveragevalueof 70%, of thesubsidence, isreached;
thepercentagebecomes94%withthepassageof the
shield. After the passage of the face, the behaviour,
apart from rare cases, was similar for all points
irrespectiveof thecoverageandother parameters.
9 CONCLUSIONS
Theanalysis of thedatacollected and of thequoted
referencesenablestostressthat:
1 for the Milan area soil the reference range of
the parameter K for settlement calculation is
0.430.46.
2 Thenet differenceinbehaviours betweenthetwo
casesessentially referstotechnological aspectsof
backfillingoperationsforthegapabovethepre-cast
segments. In line 1 a simple systemof grout-
ing from4 nozzles placed around the shield was
used. In Passante a more complex systemwas
adopted (Hochtief system) which uses injections
of acement groutingfromnozzles placedaround
theshield; however, thisbentonitefluidwasableto
exertapressureonthesoil, becauseitwasconfined
to amobileannular mouldwithtransducers. This
equipment enablesthevariationof pressureof the
groutingandworks inasimilar way to compen-
sationgrouting, however, it isextremelytaxingin
termsof maintenanceandtherequirementof highly
specializedpersonnel.
3 Theanalysisof thedevelopmentof thelongitudinal
subsidencecurveover timeinrelationtotheposi-
tionof thefaceandtheadvancerateshowedthat,
in the Milan soil, the strain effects of the tunnel
advancespreadvery quickly, almost immediately,
over thesurface.
4 Given the values necessary to guarantee the face
stability, thetunnel facepressurewouldnot influ-
ence the final settlements The pressure increase
leads probably to higher safety factors of theface
stability but it is not able to guarantee compara-
ble settlements decrease. Moreover, as explained
by Anagnostou and Kovari (Anagnostou, Kovari
1996), if the support pressure exceeds a certain
upper limit someoperational problems(highwear
of the cutter, excessive torque increase, difficul-
ties in muck discharge) may occur; it may also
cause high fluctuations of the distribution of the
effectivepressureactedonthefacethat may lead
alsotolocal instability. Sosupportpressureshould
bedefinedusingonly calibrationduringthework
startingfromaninitial valuedeterminedusingsim-
ple equilibriumratios in K0 conditions (AFTES
2001).
REFERENCES
A.F.T.E.S., 2001. Synthese Eupalinos 2000 EPB Shield
ThemeB1: Laboratory studies on reduce models.
370
Anagnostou G., Kovari K., 1996. Face stability condi-
tionswithearth-pressure-balancedshields.Tunnellingand
Underground Space Technology, 11(2):165173. Elsevier
Science.
Attewell P. B., 1978. Ground movements caused by tun-
nelling in soil. Large Ground Movements and Struc-
tures, 1978London, 812948. Ed. Geddes, PentechPress
London.
BurlandJ. B., WrothC. P., 1974. Settlementof buildingsand
associateddamage. Proceedings of a Conference on Set-
tlement of Structures, 1974Cambridge, 611654. Pentech
PressLondon.
BurlandJ. B., 1997. Assessment of riskdamagetobuildings
duetotunnellingandexcavation. Earthquake Geotechni-
cal Engineering, 1997Ishihara, 11891201. Balkema.
ChambonP., CorteJ. F., 1994. Shallowtunnelsincohesion-
lesssoil: Stabilityof tunnel face. Journal of Geotechnical
Engineering, 120, 11481165. ASCE.
Chiorboli M., Marcheselli P., 1996. Analysis andcontrol of
subsidenceduetoEarthPressureShieldtunnellinginPas-
santeFerroviario of Milano. North American Tunneling,
1996WashingtonDC, 97106. Balkema.
CavagnaB., Chiorboli M., 2004.Theuseof EPBintheMilan
subsoil: line1extension.International Congress on mech-
anized tunnelling. 2004 Torino, 107118. Politecnico
Torino.
Mair R. J., Taylor R. N., BurlandJ. B., 1996. Prediction of
ground movements and assessment of risk of building
damageduetoboredtunnelling. International Symposium
on Geotechnical Aspects of Underground Construction in
Soft Ground, 1996London, 713718. Balkema.
Mair R. J., Taylor R. N., 1997. Theme lecture: bored tun-
nelling in the urban environment. Conference on soil
mechanics and foundation engineering, 1997 Hamburg,
vol. 4, 23532385. Balkema.
RankinW. J., 1988. Groundmovementsresultingfromurban
tunnelling: predictions andeffects. Engineeringgeology
of underground movements, Geological Society Engi-
neering Special Publication, 7992. Geological Society,
London.
371
Displacementsandstressesinducedbyatunnel excavation: Caseof
BoisdePeu(France)
S. Eclaircy-Caudron, D. Dias& R. Kastner
INSA Lyon, Civil and Environmental Engineering Laboratory (LGCIE), Villeurbanne cdex, France
ABSTRACT: Thetunnel of BoisdePeuispart of aproject of thesouth-easternBesanon(France) by-pass.
Anexplorationgallery permittedtohighlight thepresenceof eighteengeological unitsandtodistinguishfour
sorts of materials: limestone, marls, clays and interbedings of marls and limestone. Despiteof theimportant
number of laboratory andinsitutestscarriedout, many uncertaintiesremainedonthemechanical parameters
valueandonthepositionof thedifferentgeological units.So,itwasdecidedtoapplytheinteractivedesignmethod
duringtheconstructionto adapt theexcavationandsupport to theactual conditions found. Intheframework
of thismethod, animportant monitoringprogramwasforeseen. Thisarticleshowsthebehavior observedinthe
differentkindof soils. Intheclayeyzonewherethesupportisthemostcomplex, strainsinthegroundandinsteel
ribsaremonitored. Finally, theavailablemeasurementspermittobetter understandthebehavior of soil-structure
interactionproblems.
1 INTRODUCTION
Thediggingof atunnel inducesamodificationof the
initial stressfieldinthegroundwhichcreatesanunbal-
ancestate.Thisunbalanceresultsinmovementsof soil
likeconvergenceof thecavity, pre-convergenceahead
of theface, extrusionof thefaceandsettlements.
Duringtheexcavationanarchingeffectiscreated.If
itisnotsufficienttostabilizethecavity,liningsupports
aresetupinorder tolimitsoil movementsandthus, to
avoidthefailureof thestructure.
Theliningsupportcanincludeshotcrete,ribsand/or
radial bolting. Incertaincases, whenthegroundhas
lowmechanical characteristics, additional supportlike
groundreinforcementorgroundimprovementatsome
stageof theexcavationisnecessary. Several reinforce-
ment systems have been developed. Pelizza & Peila
(1999)presentedthedifferentmethodsof soil androck
improvement used to permit safe tunneling in diffi-
cultgeological conditions. Lunardi (2000) dividedthe
support methodsintothreegroups: pre-confinement,
confinement and pre-support. Each group exerts a
different kindof effect onthecavity.
Whenthetraditional methodof diggingis usedin
difficult geological conditions, the main problemis
thecontrol of movements. Withoutsupportor adapted
treatment, theground tends to sink into theopening
(tunnel face failure, tunnel face extrusion): it is the
phenomenonof decompression. Inordertoreducethis
phenomenon, an action of pre-confinement may be
required. A pre-confinement actionisdefinedasany
activeaction that increases theformation of an arch
effect in the ground ahead of the tunnel face. The
pre-confinement can be achieved by reinforcement
or protectiveinterventionaheadthetunnel face. The
umbrellaarchmethodandthefaceboltingareincluded
inprotectiveinterventionsor pre-support methods.
Thesupport and forepoling introducemany three
dimensional soilstructureinteractions.Consequently,
it is difficult to understandthesephenomenaanalyt-
ically. Moreover, althoughtheumbrellaarchmethod
is widely used, there are no simple approximations
to simulate this method in numerical analyses. The
design of an umbrella arch is still based today on
empirical considerations or on simplified schemes
(Oreste& Peila1997). Several numerical studiescar-
riedout in2Dandin3D, werefocusedonthemanner
of takinginto account umbrellaarch(Tan& Ranjith
2003, Baeetal. 2005). Inthesameway, several numer-
ical modelingwerecarriedoutin2Dandin3Donthe
manner of taking into account thefacebolting (Yoo
2002, Dias1999).
For somelargegeotechnical engineeringprojects,
amonitoringprogramis generally definedinproject
phasetorecordthesoil movementswhichreallyoccur
during construction and to evaluatetheperformance
of theconstruction design. In most of thecases, the
recordeddataarejust usedtocontrol theconstruction
process. But, these data can be also used to update
predictions by using inverse analysis processes on
373
Figure1. A planviewof theBoisdePeutunnel project.
these measurements in order to decide of a possi-
bleadaptationof theconstructionprocessinthecase
where unsafe values would be predicted. This later
practiceis apart of theobservational method (Peck
1969, Powderham& Nicholson 1996). TheAFTES
guidelines (2005) related to themonitoring methods
of undergroundworkspresent themajor methodsand
giveadvicesonthemeasurementsfrequency.
In this article, monitoring results obtained in the
different geological unitsfoundduringtheexcavation
of thedownwardtubeof theBoisdePeutunnel arepre-
sented.First,thiscommunicationintroducesthetunnel
project. Then, inasecondpart, themonitoringresults
arepresented.Theypermittobetterunderstandthesoil
and the umbrella arch behavior set up in the clayey
zone.
2 PRESENTATIONOF THETUNNEL
2.1 General presentation
The tunnel of Bois de Peu (cf. Fig. 1) is part of
the project of the south-eastern Besanon (France)
by-pass entitled La Voie des Mercureaux. This
project includes several engineering structures: two
tunnels and one bridge, and several retaining walls.
Thetunnel iscomposedof twotubesof 520mlength.
Thecover height varies between8mand140m. The
excavation, achievedinSeptember 2006, wascarried
outfull facebydrill andblastfor themajor partof the
tunnel.
2.2 Geology and geotechnics
An exploration gallery was dug in 1995 in order to
assessthemechanical propertiesof theground. It has
awidthof 3mandaheight of 3.5m. Variouslabora-
toryandinsitutestswerecarriedout. Theresultslead
to concludethat thetunnel is situated in adisturbed
area. Eighteengeological unitsareidentified. A geo-
logical cross section is showed in Figure 2. Among
these eighteen units, four sorts of materials can be
distinguished: clays, marls, limestone and interbed-
ings of marls andlimestone. Geotechnical properties
of these materials defined at the end of site inves-
tigations are summarized in Table 1. Two types of
characteristicsweredefinedforthemarl (probableand
Figure2. Thegeological crosssection.
Table1. Geotechnical propertiesof material.
E
Material kN/m
3
MPa C MPa
Limestone 26.3 329010000
Marl 24.8 1600 0.7 40
(probable)
Marl 24.0 750 0.21 36
(exceptional)
Clay 23 80 0.250.4 1317
exceptional) becausethein situ and laboratory tests
leadedtovariablevalues. ThePoissonratio, thedila-
tancy angleandtheearthpressureratio arethesame
for thematerialsandarerespectively equal to0.3, 0
and0.7.
Theexactpositionof thedifferentunitsisdifficultto
knowbeforethedigging. Finally, followingtheobser-
vationmadeinthis explorationgallery, many uncer-
taintiesareremaining. So, it wasdecidedtoapplythe
observational method during thedigging in order to
adapttheliningsupporttothereal groundconditions.
In this framework, an important number of experi-
mental measurementswereforeseenandfour sortsof
support were defined in project phase. During con-
struction, thechoiceof theadaptedsupport depended
on the monitoring results and on the quality of the
soils assessedby geological surveys. Inthis context,
two other supports wererealized during theexcava-
tion. Thedigging step for each support was defined
variable.
3 MONITORINGRESULTS
Four monitored sections arepresented: onein inter-
bedings of marl and limestone (entitled D1), one in
marls(D2) andtwoinclay(D3& D4). Eachof them
arenotcircular. Fortwoof them, onlyconvergenceand
leveling measurements are available. For the others,
morespecificmeasurementswerecarriedout suchas
strainsmeasurementsaheadthefacebyextrusometers.
374
Table2. Maincharacteristicsof eachstudiedsection.
PM H D
face
Section m Material m m
D1 9.75 marl/limestone 22 3.1
D2 457 marl 40 13
D3 510.3 clay 15 1.3
D4 493.8 clay 22 3.84
Eachsectionispresentedandanalyzed.Table2reports
thelocalization in thetube(PM, referred to Fig. 1),
thetypeof material, theoverburden fromthetunnel
axis(H) andthedistancefromthefaceattheoriginof
theconvergenceandlevelingmeasurements(D
face
)for
eachstudiedsection. D1, D3andD4aresituatednear
oneof thetwotunnel portals(seeFig. 1, lowoverbur-
den).ForthesectionsD1,thewall supportiscomposed
of shotcreteandsteel ribsset upevery0.75to1.75m.
ForD2, itismadeupbyshotcreteandradial bolts.The
digingstepvariesbetween3.5and4.5m. For D3and
D4, thesupport is morecomplex. It includes awall
support andanarchinvert byshotcreteandsteel ribs,
set upat thetunnel advanceevery 1.5m. Andafore-
polingby umbrellaarchandfaceboltingarerealized
every 9min thegeneral case. In D3, theexcavation
wasmadeinpartial faceotherwiseitwasmadeinfull
faceinD4.
3.1 Convergence and leveling
Tunnel wall convergences between reference points
arerealizedbyoptical sights. Fiveoptical reflectortar-
getsareinstalledinthemonitoredsectionswherethe
excavation is madein full face(at thecrown, at 45
andatthespringline).Whentheexcavationisrealized
inpartial face, thefivetargetsset upafter theexcava-
tion of thehalf higher section arecompleted by two
other targetsinstalledafter theexcavationof thelower
part at thesidewalls. Levelingmeasurementsareless
accuratethanconvergences(5mmagainst 1mm)
becauseof theuseof several referencestations. But,
for all studiedsectionsinthedownwardtube, leveling
curvesareexploitable.
Themaximal convergenceandlevelingvalues are
obtainedintheclayeyzone. Inthehalf higher section
D3, levelingmeasurementsreached25mmfor targets
1and2(Fig. 3) whileconvergencesremainlower than
7mm.
The maximal convergence value is recorded for
wire4. Althoughthesectionisclosedbyatemporary
buton, vertical movementsareimportant.Astargets1,
2and3presenthigherdisplacements, thedeformation
of thesectionis dissymmetric. Thesidewall located
neartheothertubepresentslargervertical movements.
Aftertheexcavationof thelowerpart, convergences
measurementsarehigher andreach12mmfor wire4
Figure 3. Leveling measurements in the half higher
sectionD3.
Figure 4. Convergences measurements in the half lower
sectionD3.
(Fig. 4). For the other wires, convergences remain
lower than7mm. Thelevelingmeasurementsarenot
so high than in thehalf higher section. They remain
lower than 6 mm. All targets show similar displace-
mentssothedeformationof thesectionissymmetric
after theexcavationof thelower part. Measurements
in section D3 seemto be stabilized after the return
towardsanexcavationinfull face.
InsectionD4, 25mmof convergenceisregistered
for wire 4 (Fig. 5) and the speed of convergence is
highdespiteof thedistancefromthefaceat theori-
ginwhichisimportant (4m). Vertical movementsare
important but remainlower thanthoseregistereddur-
ing the digging of the half higher section D3. They
reach14mm. Thesectionis closedby anarchinvert
setupatthetunnel advanceevery3m. So, thesection
isnot immediatelyclosed. Thiscanexplainthelevel-
ingvalues. Measurementsarenotstabilizedattheend
of theexcavation.
Displacements recorded in theclayey zonein the
downward tubearemoreimportant than thosemea-
suredintheother tube(Eclaircy-Caudronetal. 2007).
Consequently, theclayey zoneseems to beof better
qualityinthedownwardtube. Thisconclusionisalso
375
Figure5. ConvergencesmeasurementsinsectionD4.
Figure6. LevelingmeasurementsinsectionD1.
verified by the strains measurements performed by
extrusometersaheadtheface.
ThesectionD1issituatedininterbedingsof marls
andlimestone. Theconvergencemeasurements show
animportant dissymmetry confirmedby theleveling
measurements(Fig. 6). Infact, vertical displacements
aremoreimportantfortargets4and5andreach4mm.
Maximal convergences values areobtainedfor wires
3and4andreach8mm.
This section presents lower measured values than
sections situated in a similar soil in the upper tube
(Eclaircy-Caudronet al. 2007). Thiscanbeexplained
by the dip of the interbedings which is more favor-
ablein thedownward tube(Fig. 7). In fact, theface
surveyshowsalmost horizontal interbedingsof marls
and limestone. Moreover, the face survey presents
lessfracturesandfaultsinthissectionthaninsimilar
sectionsintheother tube. So, thesoil seemstobeof
betterqualityinthedownwardtube.Measurementsare
stabilizedat adistancefromthefaceequal to60mso
9R against 150mintheother tube.
InsectionD2, locatedinmarls, themaximal con-
vergencevalueisobtainedfor wire4asinthesection
D1andinmarlsintheupwardtube. It reaches12mm
beforetheendof theexcavationfromtheDoubsportal
at 03/17/06. Levelingmeasurements showadissym-
metric deformation of the section as in the upward
Figure7. FacesurveyinsectionD1.
Figure8. FacesurveyinsectionD2.
tube. Maximal vertical movements are recorded for
targets 3 and 2 and reach 15mmbefore the end of
the excavation from the Doubs portal and almost
20mmafter. The side wall situated on the side of
theother tubepresents moredisplacements than the
other whileinsectionD1theoppositeisnoticed. This
dissymmetry canbeexplainedby thepresenceof the
exploration gallery and not by thegeology which is
perfectly symmetric(Fig. 8). Convergencesmeasure-
ments arestabilizedbeforetheendof theexcavation
fromtheDoubsportal (at 25m). After theexcavation
fromtheDoubsportal andbeforethebeginningof the
excavationfromtheVallonportal, themeasuredcon-
vergences increase. These displacements increments
canbeduetotheconstructionphases (drillingof the
firstfaceboltsandof thefirstumbrellaarch).After
thebeginningof theexcavationfromtheVallonportal,
theconvergencemeasurementsstopincreasing.So,the
digging fromtheother portal does not influencethe
convergences measurements of this section while in
the other tube the opposite was noticed in this kind
of soil. Concerningtheleveling, noconclusioncanbe
drawn becauseafter thebeginning of theexcavation
fromtheVallon portal, these measurements present
someunexplainedvariations.
376
Figure9. Extrusionevolutionversusthedistancefromthe
faceat different facepositions(extrusometer 1).
3.2 Radial displacements in marl at PM 325
These measurements are realized by three borehole
extensometers installedradially fromthetunnel wall
(at the crown and at 45
). Each extensometer has a

lengthof 12mandincluded7measurementspointsat
0, 1, 2, 4, 6, 8and10mfromthetunnel wall. These
measurementscanbeusedtoassesstheextent of the
zone of influence around a tunnel. Movements are
measured automatically with an accuracy of 0.02m.
Thefacewas at adistanceof 4mat theoriginof the
measurements. Thepoint locatedat 10mis assumed
to beoutsideof theinfluencezone. So, thedisplace-
ment of each point is computed by considering that
this point is fixed. Thecomputed displacements can
becomparedtotheconvergencesmeasurementsreal-
izedinmarls. Displacements arevery lowcompared
toconvergencesandremainlowerthan2mm.Thedis-
tancefromthefacebeingequal to4mat theoriginof
displacements measurements, lowdisplacements can
beexplainedbythefact that apart of themwaslost.
3.3 Displacements measured ahead the face in the
clayey zone
Two extrusometers of 20 meters length were set up
at PM 521 and 501. The feature consists to mea-
sure relative displacements between two successive
points spaced by one meter. It is destroyed at the
tunnel advance. If the anchor point can be consid-
ered outside the zone influenced by the excavation
thenabsolutedisplacement of eachpoint canbecom-
puted. Generally, thezoneof influenceextends until
one radius ahead the face. So, to consider that the
last point is fixed, theextrusometer has to belonger
thanoneradius. Themeasurementsobtainedfor each
extrusometer arepresented.
3.3.1 Extrusometer 1 (PM 521)
Figures9and10presenttheextrusionevolutionversus
thedistancetothefaceandthePM.
Figure10. Extrusionevolutionversus thePM at different
facepositions(extrusometer 1).
Thelengthof theextrusometer isindicatedineach
case. Four measurements were performed when the
facewasstoppedatPM510aftercollapseswhichtook
placeinMay2006.Aftertheresumptionof theexcava-
tioninpartial faceonlyonemeasurementwasrealized
andit isnot exploitableduetotheshort lengthof the
extrusometer.Themaximal valueof extrusionreaches
10cm(soastrainU/Requal to1.5%withR=6.8m).
Thisvalueisfour timesgreater thanintheother tube.
80%of thisextrusionisobtainedinthefirst4msoata
distancelowerthanoneradiusasintheothertube. Due
toalack of measurements, it isnot possibletodeter-
minetheradiusof influenceof theface. Measurements
shouldberealizedat eachdiggingphases.
3.3.2 Extrusometer 2 (PM 501)
Figure11 shows themost important extrusion mea-
surements performed with the second extrusometer.
Fourteenmeasurementswererecorded. Onewasmade
duringtheexcavationof thehalf higher section. Four
wereperformedbetweentheendof thediggingof the
half highersectionandthebeginningof theexcavation
of thelower part. Four wererealizedafter thebegin-
ningof thelower partexcavationandthreeweremade
afterthecompleteexcavationof thepartinpartial face.
Finally, onlytwomeasurementswererealizedafterthe
returntowards anexcavationinfull face. Thecurves
evolutionversusthedistancefromthefaceisdifferent.
Thiscanbeexplainedbyadifferentbehaviorof soil or
of thebolts. Themaximal extrusionreaches4cm(so
U/R=0.7%).Thisvalueissimilar totheoneobtained
intheother tube.
3.4 Strains in the steel rib at PM 493
Thesemeasurementsarerealizedby14vibratingwires
extensometers installed on steel ribs as shown in
Figure 12. They permit to obtain the strain of the
steel rib. Movements aremeasuredwithanaccuracy
of 1m/m.Temperaturesarealsomonitored.Theface
377
Figure11. Extrusionevolutionversusthedistancefromthe
faceat different facepositions(extrusometer 2).
Figure12. Locationof theinstrumentationat PM 495.
wasat adistanceof 1mat theoriginof themeasure-
ments. Maximal valuesarerecordedinextensometers
S2-CV1 and S2-CV2 and reach 400m/m. So, the
deformation of the rib is dissymetric. However, the
geology is the same in both sides. Fromthe strains
measuredby apair of extensometers it is possibleto
computethemaximumstressinducedinthesteel rib.
For example, thestress S
12
is determined fromthe
measurementsof S2-CV1andS2-CV2.Thesestresses
are important and the maximal value is obtained
in extensometers 1 and 2 and reaches 100MPa.
However, this value remained lower than the limit
stress(180MPa). Thecomputedstressesareshownin
Figure13. Thesamebehavior than in theother tube
is observed. Stresses induced in thesteel rib permit
also to compute the normal force and the bending
moment. It appearsthat thesteel ribworksessentially
in compression as in the upward tube. As regard to
themeasurements, structural elementsarelessloaded
in the downward tube than in the other tube but
displacementsarehigher.
Figure 13. Evolution of the stresses in the steel rib at
PM 493.
4 CONCLUSIONS
Thispaper presentsthemost representativemeasure-
ments which occurred during the excavation of the
downwardtubeof theBoisdePeutunnel inBesanon
(France). The measurements permit to better under-
stand thebehavior of thecomplex support set up in
theclayey zoneandtoadapt thesupport andexcava-
tionto thefoundconditions duringtheconstruction.
Theconvergences measurements vary between6and
25mmandthelevelingbetween4and25mmfollow-
ing the soil. It appeared that structural elements are
less loadedintheclayey zoneof thedownwardtube
thanintheupwardtubebut displacementsarehigher.
In fact, vibrating wires extensometers set up on the
steel rib permit to conclude that this structural ele-
ment isloadedat 55%of itsadmissiblestressagainst
70%intheupwardtubewheretheclayeyzoneseems
tobeof better quality.
In the upward tube, higher displacements are
observed in the geological unit including interbed-
ings of marl and limestone where a heterogeneous
geology is observed with many faults. In thedown-
wardtube, thisunitislessfracturedanddisplacements
arereduced. Thedisplacements monitored by extru-
sometersshowthat theperturbedzoneaheadtheface
extendstoonediameter. And, 80%of theextrusionat
thefaceisobtainedatadistancelowerthanoneradius.
Comparisons betweenpredictions andmeasurements
permitted to evaluate the actual mechanical param-
eters of the soils found by the tunnel. Comparisons
in the clayey zone showed that this layer had sim-
ilar properties than those defined at the end of site
investigations.
REFERENCES
AFTES. 2005. Recommandations relatives aux mthodes
dauscultation des ouvrages souterrains. Tunnels et
Ouvrages Souterrains 187: 1047.
378
Bae, G. J., Shin, H. S., Sicilia, C., Choi, Y. G. & Lim, J. J.
2005. Homogenizationframework for threedimensional
elastoplastic finite element analysis of a grouted pipe-
roofing reinforcement method for tunnelling. Interna-
tional Journal for Numerical and Analytical Methods in
Geomechanics 29: 124.
CETU. 2003. Dossier de consultation aux entreprises. Tunnel
de Bois de Peu.
Dias, D. 1999. Renforcementdufrontdetailledestunnelspar
boulonnage Etudenumriqueetapplicationuncasrel
ensiteurbain.Thesis, InstitutdesSciencesAppliquesde
Lyon, Lyon, France.
Eclaircy-Caudron, S., Dias, D. & Kastner, R. 2007. Move-
mentsinducedbytheexcavationof theBoisdePeutunnel
inBesancon(France).ECCOMASthematic Conference on
Computational Methods in Tunneling (EURO:TUN 2007
Vienna, Austria, 2729 August 2007).
Lunardi, P. 2000. The design and construction of tunnels
using the approach based on the analysis of controlled
deformation in rocks and soils. Tunnels and Tunnelling
International: 330.
Oreste, P. P. &Peila, D. 1997. Laprogettazionedegli infilaggi
inavanzamentonellacostruzionedellegallerie. Convegno
di Ingegneria Geotecnica; Proc. Intern. Symp., Perugia.
Peck, R. B. 1969. Advantages andlimitations of theObser-
vational Methodinappliedsoil mechanics. Gotechnique
19: 171187.
Pelizza, S. & Peila, D. 1999. Soil androck reinforcement in
tunneling.Tunnelling and Underground SpaceTechnology
8: 357372.
Powderham,A. J. &Nicholson, D. P. 1996.TheObservational
MethodinGeotechnical Engineering. Institution of Civil
Engineers: 195204. London: ThomasTelford.
Tan, W. L. & Ranjith, P. G. 2003. Numerical Analysis of
Pipe Roof Reinforcement in Soft Ground Tunnelling.
16th International Conference on Engineering Mechan-
ics, ASCE, Seattle, USA.
Yoo,C.2002.Finiteelementanalysisof tunnel facereinforced
by longitudinal pipes. Computers and Geotechnics 29:
7394.
379
Shieldtunnelingbeneathexistingrailwaylineinsoft ground
Q.M. Gong& S.H. Zhou
The Key Laboratory of Road and Traffic Engineering at Tongji University, Shanghai, P.R. China
ABSTRACT: The paper discusses the geodynamic challenges and technical countermeasures in the shield
tunnel projectspassingunder anexistingrailwayline, especiallyfor soft groundconditions. Threemainissues
neededtobesolvedbytheengineeringcommunity: trainsafetyduringtheperiodof tunnel construction, accel-
eratedtrackdegradationof trackstructureinthelongtermandtunnel structuredurability. Thispaper describes
andpredictsthenatureof thethreeissues, thecountermeasuresarealsogiven.
1 INTRODUCTION
Cities and urban population steadily grow, and put
stringent requirementsfor metrotransportation. With
thedevelopment of themetrosystem, moreandmore
metro tunnels need to pass under an existing rail-
way line, suchasmetroline9, 11and7inShanghai,
China. At thesametime, therailway lines havebeen
speeded up and are planned for increasingly higher
trainspeedsandhigherloads, whileexistinglinesneed
tobeupgradedtoallowfaster, heavier andmorefre-
quent trains. Inthiscontext, threemainissuesneedto
beconsidered:
1 TheTrain Safety issue: excessivesettlement of
the track due to the tunnel excavation must be
avoided to ensure the permissible deviations in
trackgeometry.
2 The Track Structure Durability issue: increas-
ing traffic density, loads and metro tunnel struc-
turebringabout higher dynamic actionsandmore
intensivetrackmaintenance.
3 TheTunnel StructureDurability issue: dynamic
actionsaretobesustainedbytunnel structure.
This paper discusses these issues in Shanghai,
China. The train safety issue and countermeasures
duringthemetrotunnel constructionareconcentrated.
2 SITE DESCRIPTION
Thesitechosenfor theproject wastheJ iadingareain
Shanghai, wheremetroline11passesunder Hu-Ning
railway Line. Figure1shows their planeandvertical
position.Theanglebetweenthemis85degree, almost
perpendicular.Theoverburdensoil depthof shieldtun-
nel is 11.08m, its outsidediameter is 6.2mand the
thickness of tunnel liningis 0.35m, thegroundwater
level isabout 1mbelowthegroundsurface.
Table1. Soil parameters.
Soil layer
Parameters
1

2

1

Gravity() 18.4 18.1 17.4 17.7 19.3
(kN/m
3
)
CohesionC(kPa) 13.0 8.0 9.0 13.0 45
Angleof friction 18.5 24.0 17.0 13.0 15
+(degrees)
Lateral ratioof 0.46 0.41 0.50 0.50 0.40
earthpressureK
0
Compressive 5.25 7.22 5.50 6.77 10.5
Modulus(MPa)
* InTable1,
1
isTansiltyclay,
2
isSallowsiltyclay,
1
is
greysiltyclay, isgreyclay, isSapgreenClay.
The 4.5mhigh embankment of Hu-Ning railway
line has been constructed in 1908. The train speed
have increased up to 250km/h at 18th, April, 2007
and demands for shortening travel times are rising.
Figure 1(b) also shows the geological profile of the
chosensite,thesoil propertiesareintable1.According
tothegeotechnical investigation, thesitecanbechar-
acterizedbya1.5mweatheredcrustoveralayerof soft
claywithathicknessof about13.0m. Undertheselay-
ersliessiltyclaywhosestiffnessismoregreater than
theoverlyingsoils.
Track-sidebuildingsarealmost not existing, sothe
influencesonenvironment canbeignored.
3 THE SAFETY OF RUNNINGOF THETRAINS
DURINGTUNNEL EXCAVATION
Tunnel excavation may induce adverse effects on
nearbyexistingstructuresandservices(e.g., deforma-
tionontracks, derail of trains).Theaccurateprediction
381
(a) Plain position
(b) Geological profile
Figure1. Layout of railwaylineandshieldtunnel.
of tunneling effects poses a major challenge during
design.
The effects of tunnel advancement during con-
struction on the ground and railway track are three
dimensional andtransient. Inthispaper, a3D, elasto-
plasticanalysiswasperformed. Figure2showsthe3D
finiteelement mesh. Themeshwas 50mlong, 45m
wideand 40mhigh. Eight-nodebrick elements and
four-nodeshell elementswereusedtomodel thesoil
andtheconcretelining, respectively. Roller supports
wereappliedonall vertical sidesof themesh, whereas
fixedsupportswereassignedtothebaseof themesh.
Therefore, themovement normal to all vertical sides
of the mesh and the movement in all directions at
thebaseof themeshwererestrained. Thewater table
waslocatedatthegroundsurface.Anelasto-perfectly-
plastic soil model using the Mohr-Coulumb failure
criterionwithanon-associatedflowrulewasadopted
inthisstudy.Thetunnel liningwasmodeledasalinear
elastic material. TheYoungs modulus and Poissons
ratio for thetunnel lining weretaken as 30GPaand
0.3, respectively. Theunit weight of thetunnel lining
was24kN/m
3
.
Figure 3 shows the progressive changes in the
tunneling-inducedsurfacesettlement under thetrack
as the tunnel advances. With further excavation, the
surfacesettlementscontinuetoincrease. Considering
thetunneling-induceddeformation, theirregularityof
thetrackisshowninFigure4.
Figure2. Finiteelement mesh.
Thederail coefficientandrateof wheel loadreduc-
tion were shown as table 2 using locomotive-track
dynamic coupling model. According to the protocol
published by the Ministry of Railway in China, the
stability and safety of thetrain will bethreatened if
therunningspeedof trainisabove100km/h(passen-
ger car) and60km/h(freight), sothetrainspeedmust
belimitedduringthetunnel construction.
3.1 Influence on tunnel lining by running trains
After the construction of the tunnel, the dynamic
stressesinducedbyarunningtrainwill actonthetun-
nel liningfor alongtime, evenuntil therailway line
or themetrolinewill beabandoned. Sotheinfluence
of thedynamic stresses must beconsidered. In gen-
eral, thedynamic stresses inducedby atrainrunning
dissipate quickly because of the nonlinear material
andtheviscousdamping. Theinfluencedepthisonly
about 3munder thesub-gradebed. But thevalues of
dynamicstresseswill begreater withthetunnel under
382
-25
-20
-15
-10
-5
0
5
0 5 10 15 20 25 30 35 40 45 50
m
S
u
r
f
a
c
e

s
e
t
t
l
e
m
e
n
t

(
m
m
)
Uplink tunnel excavates 10.6m
Figure3. Tunneling-inducedsurfacedeformation.
Figure 4. Longitudinal irregularity of the tracks by
tunneling-induceddeformation.
Table 2. Maximumvalue of derail coefficient and wheel
loadreduction.
Vehicle Passenger car Freight
Speed
(km/h) 90 100 50 60
Item
Pair of I 0.63 0.13 0.75 0.20 0.69 0.21 0.98 0.25
wheels II 0.51 0.10 0.59 0.13 0.56 0.15 0.78 0.22
III 0.48 0.08 0.56 0.11 0.53 0.15 0.75 0.20
IV 0.60 0.11 0.73 0.21 0.67 0.19 0.93 0.23
*InTable 2, represents Derail coefficient, represents
wheel loadreduction.
therailwayline, becausethestiffnessof tunnel ismuch
greater thenthesoft soil at thesamedepth.
Thedynamic stresses actingontunnel lininghave
been established using a dynamic numerical model.
Laboratory tests have shown that soil stiffness and
damping change with cyclic strain amplitude under
dynamic cyclic loading conditions. To simplify, the
soil is modeledusingequivalent Linear analysis, the
dampingratioalsochangeswithstrain. Figure5shows
the loads acting on the tunnel lining including the
p
12 p
11
q
2
q
1
P
1
P
2
q
1
2
q
O
11
q
21
q
22
q
12
q
12
p 11
p
p
13
p
13
Figure5. Load-structuremodel includingdynamicstress.
dynamic stresses. It is quite different fromthe con-
ditionswithonlystaticstresses.Thelong-termactions
of dynamicstresseswill influencetheservicelifetime
of tunnel structure.
4 SAFETY COUNTERMEASURES
4.1 Countermeasures to ensure the train safety
during tunnel excavation
Therearebasically threeways to decreasetheinflu-
enceonsafetyof thetrainsduringtunnel construction:
(a) by decreasingthetunnelinginduceddeformation;
(b) by increasing the longitudinal bending stiffness
of track structure itself; (c) limit the running speed
of thetrain.
Decreasingthetunnelinginduceddeformationmay
for instancebedonebymeansof improvingthestiff-
nessandstrengthof thesoil aroundthetunnel.Thiscan
be done by for example lime-cement piles, jet-piles
or other deep-mixing methods. These countermea-
suresareeasier toimplementaspartof thefoundation
workof newlines, thanasaretrofittingmethodunder
existinglines.
Figure6shows thereinforcingareas andmethods
around the tunnel, including jet-piles and grouting.
These countermeasures decrease the track deforma-
tionduringthetunneling. Thevalueof thesettlements
isonly15%of theoriginal deformation, sotheremain-
ing track irregularity is small. At thesametime, the
bearingcapacityof thefoundationof therailwaytrack
after thetunnel constructionwill increase.
Theconstruction parameters should becontrolled
duringthemetrotunnel excavationtodecreasethedis-
turbing of thesoil. Table3 shows therecommended
parameters. Thecontrolling and monitoring systems
shouldalsobeperfecttofollowthetrackdeformations.
Figure7showsthemeasureddeformationsat thesite
duringtunnel construction. is thepreliminary set-
tlement, isthesettlement inducedbyshieldarrival,
isthesettlementduringtunnel construction, isthe
383
Tunnel structure jet-pile area
Secondary reinforcing area Main reinforcing area
Figure6. Reinforcedsoil aroundthetunnel.
Table3. Constructionparameters.
Constructionsite
Hu-Ning
Constructionparameters Railway
Support pressure (MPa) 0.200.22
Groutingpressure (MPa) 0.190.21
Groutingamount (m
3
) 3.34.1
Groutingspeed (L/s) 1.11.4
Constructionspeed (cm/min) 2.02.5
Time (h)
-2 0
-1 5
-1 0
-5
0
5
10
0 30 60 90 120 150 180
C
u
m
u
l
a
t
i
v
e

S
e
t
t
l
e
m
e
n
t
(
m
m
)
Subsequent settlement
Figure7. Measured deformation at thesiteduring tunnel
construction.
settlement induced by sub-optimal filling of shields
rear, isthesubsequent settlement. Theresultsshow
that thedeformationof thetrack has beencontrolled
effectively. Theseapproachescanbeused.
4.2 Countermeasures to protect the tunnel lining
Considering the long life time and the exhausted
strength, the lining stiffness and strength should be
increased. So the reinforcement ratio of the tunnel
liningunder therailwaylinehasbeenincreased.
5 CONCLUSIONS
Tunnels passing under an existing railway line must
ensure the train safety, the track structure durability
and the tunnel structure durability. These problems
areespeciallyimportant insoft groundareas. Inorder
todecreasethedynamic stressesactingonthetunnel
lining, to decrease the tunneling induced deforma-
tion, andtoevenincreasethebearingcapacity of the
ground, thesoft soil andtunnel liningstiffness must
bereinforced.
ACKNOWLEDGEMENTS
This research was sponsored by National Natu-
ralScience Foundation of China (No.50678131) and
Science and Technology Commission of shanghai
Municipality(No.06ZR14083). Grateful appreciation
isexpressedfor thesesupports.
REFERENCES
Esveld, C. 1989. ModernRailwayTrack. MRT-Productions,
ISBN90-800324-1-7.
Esveld, C. 1997. Innovationfor thecontrol of infrastructure
maintenance. Rail International/Schienen der Welt.
Suiker, A.S.J. 1996. Dynamic behaviour of homogeneous
and stratified media under pulses and moving loads.
TU Delft, Report 7-96-119-1.
384
Casehistoryonarailwaytunnel insoft rock(Morocco)
A. Guiloux, H. LeBissonnais& J. Marlinge
TERRASOL, Geotechnical consulting engineers, France
H. Thiebault, J. Ryckaert, G. Viel & F. Lanquette
SETECTPI, France
A. Erridaoui
MAROC SETEC, Morocco
M.Q.S. Hu
TEC Engineering, P.R. China
ABSTRACT: Ras RMel Moroccan tunnel is a 2.6kmlong and 60m
2
section singletrack railway tunnel.
It was excavated in a complex geological context of highly heterogeneous and highly deformablesoft rock.
Thepurposeof this paper is first to describethemethodology usedfor specific designof support andlining.
Thedeformationmonitoringprocess appliedduringconstructionwill thenbedetailed, andananalysis of the
deformations measured will begiven, making acomparison with design calculations results. Main concerns
werepotential faceinstabilityandhightunnel deformations.
TheRasRMel tunnel presentedinthiscase-historyis
asingletrack railway tunnel, part of thenewrailway
linebeingconstructedinMorocco betweenthetown
of TangieranditsnewMediterraneanharbour, located
on thecoast about 30kmnorth-east of thecity. The
tunnel itself islocatedsome15kmeastof Tangier,near
the location of Ras RMel. Western portal is called
Tangier portal, eastern one is named Ras RMel
portal. Kilometricpointsontherailwaylineincrease
fromTangierportal (PK 26+841) toRasRMel portal
(PK 29+445).
Chinesecompany TEC Engineering won thecon-
tractforconstructionof thetunnel,whichalsoincluded
constructionof asecondtunnel, shorter (600mlong)
butinasimilar geotechnical context, andconstruction
of the railway platformin between the two tunnels.
For both tunnels, following aprevious collaboration
ontheMeknestunnel project (another Moroccanrail-
way tunnel, the study of which was described in a
previouspaper forAFTESOctober 2005international
congress),TECEngineeringchoseFrenchengineering
companiesSETECTPI andTERRASOL, subsidiaries
of FrenchgroupSETEC,asitsconsultingengineersfor
specificdesignof supportandlining, andfordeforma-
tionanalysis duringconstruction. This paper focuses
ondesignandconstructionissuesregardingRasRMel
tunnel.
Thetunnel, of a2600mlengthanda60m
2
section,
wasexcavatedinhighlyheterogeneousflyschs,withup
to150mdepthinitscentral part(Cf. Fig. 2).Theexca-
vationwasgivenahorse-shoeshapeof a8.5mheight
and a7.6mto 7.8mwidth, depending on thelining
thicknessinorder topreservetheultimatecirculation
clearance (Cf. Fig. 1). Four zones with provisional
enlarged sections were excavated, in order to allow
enteringandexitingtrucks to cross duringconstruc-
tion. Theenlargedsectionsareof a10.2mwidthand
a9.4mheight. At final stage, theover-excavationof
thosezoneswill befilledwithliningconcrete.
Theentranceportalswereat adepthof only 10m.
Asthoseportalswerealmostvertical,andasthequality
of thepelites, classifiedassoftrocks, israther poor, it
wasdecidedtostabilisetheportalsbuildingtwo22m
long falsetunnels (oneat each portal) using acut
andcover technique.
Indicativetunnel profileisgivenonFigure2.
Fromeachportal, andonatotal lengthof around
1600m, theexcavationisat adepthlower than50m:
great surfacesettlements wereexpectedinthis zone.
Inthecentral partof thetunnel, onalengthof 1000m,
excavationisatadepthof about150m:lowsurfaceset-
tlementsbut highconvergenceandtunnel settlements
385
Figure1. Typeprofiles(basicandenlargedcross-sections).
Figure2. Tunnel profileanddeformationsmeasuredbefore
liningconstruction.
wereexpectedinthiszone. Asthetunnel isinacoun-
trysideenvironment with no construction at surface,
surfacesettlementswerenotexpectedtobeof amajor
concern, but tunnel convergenceswere.
2 GEOTECHNICAL CONTEXT
RasRMel tunnel islocatedinageologicallycomplex
and highly tectonised zone. Two flyschs formations
canbefound, theolder one(theTisirennappe, alter-
natinglevelsof sandstoneandpelite) beingontopof
theyounger one(theBeni Ider Nappe, softer clayey-
calcareous flysch) being athrust-nappe. Thecontact
planebetweenthoseformationsslightly dipstowards
east. It was expected to beintersected by thetunnel
but could not be firmly identified during construc-
tion.Thiscontactisknowntobeintersectedbyseveral
sub-vertical faultsdisturbinglocallythelithology.
Table1. Geomechanical characteristics.
Mainly Mainly Fully
Formation pelite sandstone sandstone
(kN/m
3
) 24 24 24
UCS(MPa) 0.66 46
C
u
(kPa) 200 500
u
(
) 0 0
C
(kPa) 20 100 1000
) 28 30 35
E
0
(MPa)* 500 1000 1500
E
(MPa)* 250
0,3 0,3 0,3
g
(kPa) 200 0 0
* E
0
and E
are respectively short and long termYoungs

moduli.
Furthermore, the geological formations are com-
plex themselves: both flyschs formations consist of
alternatinglevelsof pelites(softrockwithUnconfined
Compressive Strength (UCS-1MPa) and highly
resistant sandstones (UCS=30to 40MPa). Four (4)
kindsof formationswerethenexpected:
non-altered, highly fractured pelites, with occa-
sional breccia zones, Rock Quality Designation
(RQD) rangingfrom50%to75%, andsandstones
levels RQD, expressed as a percentage, is the
summedlength of corepieces greater than 10cm
measuredfor a1mlongcorepass; onceexcavated,
thepelitesget quicklyaltered;
blockysandstones, withoccasional pelitelevels;
alternatinglevelsof pelitesandsandstonesof low
thickness, that can be found intricated in blocky
sandstones;
claystonebreccia.
Asblockysandstoneslayerswherenotpersistent,no
continuityforthoselevelscouldbedrawnbetweenthe
boreholelogs,andnoaccurategeological profilecould
bebuiltfortheproject. Ithasthenbeendecided, topro-
poseanindicativegeological zoningalongtheaxisof
thetunnel, forsupportandliningdesignof theproject,
pointingat threemaincategoriesof formations:
formationsmainlypelitic;
formationsmainlycomprisingsandstones;
formationsfullycomprisingsandstones;
Table1sumsuptheaveragemechanical character-
isticsgiventotheseequivalentformations. Shortterm
cohesion values for mainly pelitic formation arethe
updatedvaluesfollowinganalysisof firstmeasuresof
tunnel deformations.
Potential swellingpressure
g
inpelitesisevaluated
to200kPa, applyingunder lininginvert.
Because of the thrust nappe context, high val-
ues for horizontal stresses wereexpected, whichwas
386
taken into account in thedesign calculations setting
parameter K
0
to 1(horizontal stress equal to vertical
one).
During construction, sandstone levels were con-
firmed non-persistent, which meant that the Fully
Sandstoneformationwasnonrelevant.Thetwoother
formationsdescribesoft rockswithrather weak char-
acteristics: under expected geostatic stress state, the
stabilityratio
0
/UCSrangesfrom2to3,5(>1) mean-
ingthat stress level is highenoughfor excavationto
beunstablewithout support.
Fromahydrogeological pointof view, captivewater
of limitedextent hadbeenidentifiedintheboreholes,
shelteredby sandstonelayers. Thesurroundingpelite
matrix is littlepermeable. However, thehighly frac-
turedpelitesandsandstonesmightallowlimitedwater
flowstoreachthetunnel. Itwasthendecidedtoprotect
tunnel lining with a watertight membrane on crown
andshoulders.
3 PROJ ECT DESIGN
TERRASOL and SETEC TPI performed specific
design studies for support and lining of the tunnel,
aswell asstabilityanalysisof tunnel face. Inorder to
control tunnel and surface deformations, a full face
excavationmethodwasconsidered.
Based on the indicative geotechnical zoning pro-
posed for the tunnel, six calculation profiles were
defined, named P1 to P6. Thoseprofiles differed by
the formation and depth taken into account for cal-
culation(eachprofileissupposedtobecomposedof
one single formation). Five additional profiles were
studied, corresponding to potential enlarged section
zones. Thepreciselocation of thosezones would be
defined precisely during construction, depending on
real geological conditions.
3.1 Tunnel face stability analysis
Tunnel face stability analysis was performed using
TERRASOL convergence-confinement code TUN-
REN. ThecodeusesaC-Phi reductionmethod, based
on theanalytical model EXTRUSION developed by
Wong et al. (1999). Hypothesis are the ones of
convergence-confinement method (circular section,
isotropic, homogeneousandinfinitemedium, uniform
stressfield), withtwoadditional ones:
tunnel faceissupposedspherical, andstress-strain
fieldfollowsaspherical symmetry;
tunnel face heading is modelled by taking into
accountadecreasingradial pressureattunnel face.
Material behaviour is supposedelasto-plastic, fol-
lowingMohr-CoulomborTrescacriterion.
Frominitial (c,) short-termvalues, mechanical
characteristicsof thegroundareprogressivelyreduced
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
55%
60%
1 1.2 1.4 1.6 1.8 2
Safety factor F
f
= u
f
/R (%)
Figure 3. Tunnel face stability analysis for P1 profile
(mainlypeliticformationat 50mdepth).
byasecurityfactor F. For eachsetof mechanical char-
acteristics, averageradial strainiscalculatedattunnel
face, andresultsarepresentedviaachartgivingtunnel
facerelativedeformation(
f
=u
f
/R,whereu
f
isextru-
sion value and R excavation radius) versus security
factorF.Stabilityisevaluatedconsideringsecurityfac-
torvaluesF(
f
=2,5%)andF(
f
=5%)givingrelative
deformationsof respectively
f
=2,5%and
f
=5%.
Figure 3 shows an example of result chart, where
F(
f
=2,5%)=1.2andF(
f
=5%)1.4.
Calculation showed that tunnel face stability
depends on tunnel depth. Whatever the formation
considered, both security factors F(
f
=2,5%) and
F(
f
=5%) weregreater than 1for depth lower than
50m. For maximal depth (150m), security factors
wherelower than 1. Sensitivity analysis showed that
tunnel facebecameunstableunder 60 to 70mover-
burden. Groundbehaviour appearstobelimit-plastic
at 50mdepth, andtotallyplasticat 150mdepth.
In order to perform construction using full-face
excavationmethod, TERRASOL-SETECTPI advised
facereinforcement wouldthenbenecessary.
3.2 Support and lining design
Theconstruction method chosen was ahalf-sections
method. Supportandlining, asdefinedinthecontract,
wereasfollows:
support: all over thesection, includinginvert, 5cm
shotcrete confinement layer, plus HEB 180 steel
ribs settled in a18cmthick shotcretelayer; steel
ribs spacinghadto bedefinedby designcalcula-
tions, and was expected to range from0.75mto
1.5mdependingongeological conditions;
lining: watertightlayer oncrownof thetunnel, and
40to60cmthickconcretereinforcedatlateral walls
andinvert junction; reinforcement sections hadto
bedefinedbydesigncalculations.
In order to estimate tunnel deformations and
solicitations of lining, a staged calculation was
run with finite elements model (FEM). Calculation
387
Table2. FEM support calculations (after calculation pro-
filename, P andS arerespectively for Mainly Pelite and
Mainlysandstones, figureindicatesdepth).
Crown Horizontal
Surface ground convergences(mm)
Calculation settlement settlement
profile (mm) (mm) Ground Support
P1(P, 50) 3.2 13 44 9
P2(S, 50) 1.1 4.5 11 4
P3(S, 150) 25.8 60 16
P4(P, 150) 97.6 190 12
P5(P, 10) 0.4 1.8 6.5 3
P6(P, 30) 0.7 3 8 3
procedurewasleadthroughTERRASOL convergence-
confinement code TUNREN, and through French
LCPCfiniteelementscodeCESAR. For eachprofile,
calculationstepswereasfollows:
Convergence-confinementcalculations:estimation
of the deconfined rate
0
after tunnel full-face
excavation, usingTUNRENcode;
FEMcalculations: full-faceexcavation(deconfine-
ment up to
0
value previously estimated, load
appliedtoexcavationwallssetto(
N
=(1
0
)
0
);
completion of support and final deconfinement
(
N
=0); completion of lining (mechanical char-
acteristics are set to long-term values, swelling
pressureisappliedunder invert andgroundcreep-
ing is taken into account by option EFD of the
calculationcode).
Resultswereexpressedasdeformationsandstress
fieldsfor support, andonlyasstressfieldsfor lining.
Table 2 sums up displacements results after support
calculations.
Results showed that surface settlements were
expectedtobelower than5mm.
For a given formation, tunnel deformations were
expectedto growwithtunnel depths. For formations
withmainlysandstones,maximal convergencesinsup-
port and crown settlements are respectively 16mm
and 25.8mm, which is rather limited. For mainly
peliticformations, theyareof respectively12mmand
97.6mm, which shows that at maximal depth, tun-
nel behaviour iscompletely plastic anddeformations
can become important. Numerical model indicates
crowngroundsettlementsabouthalf ashighasground
horizontal convergencesfor depthgreater than10m.
Support calculationsindicatedthat great deforma-
tions (of about 100mm) could occur in support at
depth greater than 50m. Support completion would
thenbeof amajor concernregardingcontrol of those
deformations.
Lining calculations showed that only minimum
reinforcementwouldbenecessaryatjunctionbetween
lateral wallsandinvert.
Figure4. Upperhalf-sectionexcavationwithcentral partof
thefaceleft inplace(Tangier face, aroundPK 27+960).
4 CONSTRUCTION
WorksbeganinFebruary2006, withaccesscut exca-
vation. Tunnel excavation began in March 2006 at
bothportals. J unctionof thetwofacesoccurredatPK
28+098onJ ulythe5thof 2007, after15months.Aver-
ageheadingwas of about 80mper monthat Tangier
face, andof about 90mper monthat RasRMel face.
Ras RMel face appeared to be in better geological
conditions thantheother face, whichallowedhigher
heading rateand lighter support: steel ribs could be
spacedfrom1,2to1,5minaverage, against0,75min
averageatTangier face.
TEC Engineering decided not to apply face rein-
forcement treatment. Astunnel-facewasfearedtobe
unstable, it was thendecidedto apply thesamehalf-
sectionsexcavationmethodasusedforMeknestunnel.
Inorder to control tunnel facestability, duringexca-
vation of section upper-half, and in zones with poor
qualityground, central partof thefacewasleftinplace
until upper-half support completion(Cf. Fig. 4).
Furthermore, aslower-half excavationandsupport
completionoccurredinaverage4daysafterupper-half
excavation,behaviourof tunnel facewasalmostequiv-
alent to afull faceexcavation, which helped control
deformations.
Openingrangedfrom0,75minzoneswithmainly
pelites, to1,5minzoneswithmainlysandstones, and
was adaptedto thedeformations measured. A provi-
sional formwork of woodenboardsholdbysteel bars
prevented shotcreteloss during shotcreteprojection,
ascanbeseenonFigure3.
A total of 4 crossing zones with enlarged section
wasexcavated, 2at eachface.
As expected, main concern during construction
was tunnel deformations due to excavation of soft
deformablerocksunder highstresslevel.
388
4.1 Surface deformations monitoring and analysis
Fromeachportal,andonatotal lengthof 1500m,exca-
vationis at adepthlower than50m, Ras RMel face
evenhavingto cross azoneat adepthof only 15m.
Surfacesettlementswherethenmonitored.Every20m
alongtunnel axis, amonitoringcross-sectionwasset
using5surveyorsrodsset perpendicularlytothetun-
nel axis(numberedfrom1tofive, with10mto15m
spacing,central rodnumber3beingabovetunnel axis).
For agivenmonitoringsection(namedfollowingthe
corresponding kilometric point of the tunnel), mea-
surements started 2 to 4 weeks before tunnel face
reachedthegivenkilometricpoint.Thefrequencywas
onemeasureper dayuptostabilisation, andonemea-
sureaweekoncestabilisationreached. Settlementsin
thecentrepartof thetunnel, atgreatestdepth(150m),
were not monitored.A chart showing final surface
settlements measured before lining construction is
presentedonFigure2.
4.1.1 Ras RMel face
At RasRMel face, settlementsfor agivenmonitored
section began when tunnel face was approximately
20m ahead of the corresponding PK. Stabilisation
wasreachedastunnel facewas80mfurther (approxi-
matelyafter1month).Attunnel axis, final settlements
are of about 20mmfor the first 100mafter portal.
Further on, they are inferior to 5mm. No relation
couldbeestablishedbetweentunnel depthandsurface
settlements. Thelowsettlementsdidnot allowobser-
vationof relevant transversesettlement throughs, and
thecrossingof thevery lowdepthzone, betweenPK
28+924andPK 28+870, didnot induceanyincrease
of surfacesettlements.
Surface settlements measured were as expected
after FEM calculation.
4.1.2 Tangier face
AtTangierface, settlementsforagivenmonitoredsec-
tionbeganastunnel facewasabout 20to40mahead
of the corresponding PK. Stabilisation was reached
when tunnel face was 80 to 120mfurther (1 to 1.5
monthlater). Abovetunnel axis, final settlementsare
ratherhighinthefirst100maftertheportal (from20to
50mm).Thesevalues, higherthantheonescalculated,
couldbeexplainedbytheproximityof theaccesscut
totheportal, andbythelowdepthof thetunnel inthis
zone. Nevertheless, asa45moverburdenwasreached,
andsettlementsseemedtodecrease, tunnel facehit a
rather weak zone(fromPK 26+978toPK 27+140),
in which surfacesettlements weremuch higher than
expected(greater than20mmat tunnel axis, reaching
upto190mm).
Figure 5 shows settlements evolution at PK
27+000. Tunnel face upper-half reached the corre-
sponding section, when already 20mm settlements
had occurred at surface. The effect on surface
-200
-100
0
07/05/06 21/05/06 4/06/06 18/06/06 02/07/06
T
o
t
a
l

s
e
t
t
l
e
m
e
n
t

(
m
m
)
Point 1
Point 2
Point 3 (Tunnel axis)
Point 4
Point 5
Section upper-half
excavation (04/06/2006)
Support completion
(10/06/2006)
Figure5. SettlementsatPK 27+000(WeakzoneatTangier
face).
settlements of tunnel face lower-half reaching the
section, can hardly be seen as no stabilisation had
occurred at that moment. Stabilisation was reached,
two weeks after completion of tunnel support. The
final settlementvalueabovetunnel axisisof 175mm.
Thosesettlementscanberelatedtohightunnel defor-
mationsaswill bediscussedfurther on.
Final settlements for points 5 and 4, respectively
compared with those for points 1 and 2, are lower.
Almost nosettlement occursat point 5. Thisdissym-
metry in transversesettlement through is relevant of
what could beobserved at Tangier face: surfaceter-
rainisnaturallydippingsouth,sothatthecross-section
monitoredisindirectionof thedip. Points1and2are
ontheuppersideof thedip, whereaspoints4and5are
onthelower side. Crossingaweak zoneexaggerated
thisphenomenon.
As no building was constructed abovethetunnel,
suchhighsettlementswerenot of amajor concern.
After this zone of high deformations, settlements
arelowerthan20mm. Frompk27+140topk27+340,
surfacesettlementsat tunnel axisincreasefrom7mm
to 17mm, following depth increase from 40m to
90m.ForgreaterPK,settlementsdecreasewhiledepth
increase. As thecentrepart of thetunnel is reached
(withdepthof 150m), final settlementsareof a5mm
average, which fits thesettlements calculated by the
FEM simulations, and is compatible with measures
relativetoRasRMel face.
Surface settlements were then partly dependent
on tunnel depth, and partly on lithology (greater
settlementsinhighlydeformablesoft rock).
4.2 Tunnel deformations monitoring and analysis
Assaidbefore, tunnel deformationswereexpectedto
beof amajor concern. Every 20malongtunnel axis,
amonitored cross section was established after sup-
portcompletion, equippedwith5measuringtargets. It
allowedmeasurementsinthesupportof crownandside
wall settlements, andof sectionupper andlower-half
convergence.
389
Duetoconstructionconstraints, convergencemea-
surements areof questionablereliability. However, it
clearly appears that final measuredconvergences are
highlyvariablefromonesectiontoanother, depending
onlithology rather thanontunnel depth. Theconver-
gencevaluesreachedrangefrom20mmtomorethan
300mm, which is far greater than estimated by cal-
culation. Furthermore, differencesappearbetweenthe
two tunnel faces, Ras RMel faceonceagain show-
inglower deformationsthanTangiers. Figure2shows
tunnel deformations.
Atbothfaces, tunnel deformationsappearedtostop
as soon as support was completed, closing theexca-
vated sections. Later creeping could beidentified as
liningwasbeingconstructed.
4.2.1 Ras RMel face
FromRas RMel portal, up to junction PK, tunnel
deformations werelimited, section upper-half defor-
mations ranging from10mmdivergence to 30mm
convergence, crown settlements being of a compa-
rablesize. Thosevalues arelimited, but higher than
calculatedfor correspondingprofiles(P1, P2, P3, P5
andP6).
FromPK 28+620uptoPK 28+400, tunnel defor-
mations are higher, reaching 130mmconvergences
and 78mmcrown settlements. In this zone, crown
settlement is about half of convergences. This local
increaseindeformationsshouldberelatedtodecreas-
ingqualityof excavatedrockratherthantoanincrease
of overburden. Thecontactplanebetweenthetwofly-
schs formations might havebeencrossedalongwith
thischangeof rockquality, butcouldnotbeidentified
firmly.
ApproachingPK of facesjunction,thedeformations
increasedandreachedvaluesconsistent withtheones
measuredat theother face.
4.2.2 Tangier face
As shown for surface settlements, after a zone with
rather highdeformationsduetoperturbationsinduced
by thevicinity of Tangier portal, tunnel suddenly hit
a zone of very high deformations at limited depth
(around50m), locatedbetweenPK 26+978andPK
27+140.
High convergences in the section upper-half
occurred, greater than100mmandreaching330mm.
Crown settlements wereof values half as high. Try-
ingto securethezone, it appearedthat deformations
stopped as soon as sections support was completed.
Final stabilisationwasreachedwithinonemonthafter
section opening, although later creeping could be
identifiedlocally causingsupport deformations. The
values measured overcome by far the values calcu-
lated for support deformations. They are thought to
belinkedwithlocal mechanical characteristicsfor the
rocklower thanexpected.
-340
-240
-140
-40
22/05/06 05/06/06 19/06/06 03/07/06
S
e
t
t
l
e
m
e
n
t
/
C
o
n
v
e
r
g
e
n
c
e

v
a
l
u
e

(

m
m
)
Crown Settlement (mm)
Section Upper-half convergence (mm)
Section upper-half excavation (23/05/2006)
Support completion ( 30/05/2006)
Figure 6. Greatest tunnel deformations, measured at PK
26+978.
Itwasfearedthatdeformationswouldbecomeunac-
ceptable when reaching maximumdepth and exca-
vation would become highly unstable. As soon as
it became obvious that support section completion
blockedfurtherdeformations,ithasbeendecidedtogo
onconstructionwiththesamemethod, operatingexca-
vation carefully, and keeping opening and steel-ribs
spacingat only0.75maslongasnecessary.
Figure 6 shows the evolution of tunnel greatest
deformations, whichoccurredat PK 26+978(Stabil-
isationwasreachedbymid-J une).
ThehighdeformationszonegoesuptoPK 27+780,
where greatest depth is reached. Fromthis PK, and
up to the junction of tunnel faces, deformations are
still important (about 90to100mminsectionupper-
half convergences, and50mmincrownsettlements)
but lower thanthepreviouszone. Thisissupposedto
be linked with depth increase, and a better stiffness
of soft rock mass. Thedeformationsmeasuredinthis
zone, arehigher thanconvergencescalculatedfor the
correspondingprofile(P4).
Deformations values increasedas junctionsection
approached, reachingvaluescompatiblewiththeones
observedat RasRMel face.
Froma general point of view, it is thought that
part of thedeformationscalculatedasgrounddefor-
mations before support completion (theoretically
non-measurable), aremeasuredasupper-half support
convergences, becauseof thehalf-sectionsexcavation
method used, (measures begin as section excavation
isnot completed, andupper-half support isnot totally
blockedasnoinvert isconstructedat that stage). This
couldexplainwhy measuredtunnel deformationsare
greater that theones calculated. Sectionbehaviour is
thenintermediatebetweenfull-face(as calculatedin
specificdesignstudies) andhalf-sectionexcavation.
Face potential instability had been pointed at by
calculation. The cautious excavation method chosen
allowed no collapse to occur on Ras Rmel tunnel
works.Theothertunnel fortherailwayproject(Sidi Ali
tunnel), operatedbythesameteamof companyinsim-
ilar context but withslightly better geological condi-
tionsdidnotreceivesuchcautioustreatmentattunnel
390
face. Tunnel face collapsed on 31/07/2006, which
showsthatTEC Engineeringmethodwasappropriate
regardingRasRMel tunnel face.
4.2.3 Zones with enlarged section
Thefour zones with enlarged section did not induce
greater deformation compared to surrounding basic
sections.
5 CONCLUSION
RasRMel tunnel hadtobeexcavatedinhighly vari-
able geological formations, highly deformable soft
rocks, and at a high stress level. To control tunnel
deformations and tunnel face potential instability, it
was then necessary to use half sections excavation
and, when necessary, divided section excavation for
upper-half. Thismethodderivedfromamethodprevi-
ously usedat Meknes tunnel (Marocco), operatedby
thesameteamof companies.Duringconstruction,sec-
tionsbehaviour wasintermediatebetweenfull-section
andhalf-sectionsexcavations.
Design studies gave a rough size range for sur-
face and tunnel deformations, but several zones
with tunnel high deformations were crossed, which
requiredcautiousexcavation. Nevertheless, theexca-
vationmethodusedallowedto control deformations,
that wereblockedassoonassupport wascompleted.
At timethis paper was written, tunnel completion
wasprogrammedfor mid-September 2007.
REFERENCES
Guilloux, A., LeBissonnais, H. & Pre, M. 2005. Tunnel de
Meknes(Maroc): conceptionetralisationduntunnel en
terrain meuble sous faible couverture, AFTES, Les tun-
nels, cl dune Europe durable; Proc. Intern. Congress.,
Chambry, 1012 October 2005. Lyon: Spcifique.
Terrasol, Tunren. Tool for tunnel conception(liningandtun-
nel face stability), convergence-confinement and extru-
sion calculation code.
Wong, H., Trompille, V. & Dias, D. 1999. Dplacements
du front dun tunnel renforc par boulonnage prenant
en compteleglissement boulon-terrain: approches ana-
lytique, numriqueetdonnesinsitu, Revue Franaise de
Gotechnique n
89: 1328.
391
Observedbehavioursof deepexcavationsinsand
B.C.B. Hsiung
Department of Civil Engineering, National Kaohsiung University of Applied Sciences, Kaohsiung, Taiwan, R.O.C.
H.Y. Chuay
Mott MacDonald, Kaohsiung, Taiwan, R.O.C.
ABSTRACT: In this paper, structural and ground behaviours of several excavations in Kaohsiung, Taiwan
were described and examined. Based on behaviours observed fromthese excavations, it was found that the
maximumlateral wall displacement (
hmax
) inrelationtothemaximumexcavationdepth(H
e
) isapproximately
0.034to0.3%.Theratioof maximumsurfacesettlement(
vmax
) to
hmax
variesfrom0.5to0.7for theexcavation
constructedbybottom-upmethodandfrom1.3to1.8fortheexcavationusingasemi-top-downmethod.Thezone
atgroundlevel affectedbytheexcavationisupto3timesthemaximumexcavationdepthbehindthediaphragm
wall whichretains theearthduringconstruction. This findingis different fromprevious conclusions reported
byClough& ORourke. It isapparent that chemical churningpiledoesnot effectivelyreduceinfluencesof the
excavationonadjacent buildings.
1 INTRODUCTION
Grounddeformationsinducedbydeepexcavationsin
clays havebeenexploredwidely (Wonget al., 1997,
Hsieh&Ou, 1998, Hsiung, 2002, Liuetal., 2005) but
studiesregardingobservedbehavioursof deepexcava-
tionsinsandarecomparativelylimited(Burchell,2000
and El-Nahhas, 2006). In this paper, case histories
from excavations in Kaohsiung City, Taiwan pro-
videanopportunity to explorestructural andground
behaviours induced by excavations in sand. Empiri-
cal approachesfor evaluatinglateral wall movements
and surface settlements were studied and discussed.
Further, the effectiveness of chemical churning pile
for houseprotectionandpredictionof proploadwere
alsodiscussedinthisstudy.
2 THE SITES
Several deep excavations for the Orange Line of
the Kaohsiung Rapid Transit Systems (KRTS) have
recentlybeencarriedout inanareawheretheground
conditionsarerelativelyuniform. Amongthem, exca-
vationsof threeundergroundstations, O6, O7andO8
wereselected for this study. Thesites arelocated at
thecentreof Kaohsiung City, Taiwan and themaxi-
mumexcavationdepthvaries from19.6mto 20.9m.
Lengths and widths of these excavations are in the
rangeof 178to215mand22to24m, respectively.All
excavationsareretainedby1.0mthickreinforcement
concretediaphragmwalls. Constructionsequencesof
Table1. Constructionsequenceat O6.
Period
Constructionactivity (day/month/year)
Excavateto3.4mbelowgroundlevel 24/10/0327/10/03
Install 1st level propat 2.5mbelow 02/11/0306/11/03
groundlevel
Install 2ndlevel propat 5.9mbelow 13/11/0316/11/03
groundlevel
Install 3rdlevel propat 9.1mbelow 04/12/0307/12/03
groundlevel
Install 4thlevel propat 12.6mbelow 28/12/0331/12/03
groundlevel
groundlevel
O6, O7 and O8 are listed in Tables 1 to 3. H-type
steel sections wereselectedfor horizontal props and
thelevelsof propsat thesesitesarealsodescribedin
Tables1to3. Thehorizontal spacingof propsissim-
ilar at O6. O7andO8andisapproximately 4.5m. In
addition, anglebracing and waling wereconstructed
usingH-typesteel sectionsinorder tostrengthenthe
strut system.
Themainsoil stratumof thesesites is similar and
consists of mediumto dense silty sand with several
393
Period
groundlevel
groundlevel
groundlevel
groundlevel
groundlevel
Period
groundlevel
groundlevel
groundlevel
Construct concourse-level slab(0.8m 11/04/0415/05/04
thick) at 14.2mbelowgroundlevel
groundlevel
groundlevel
thin layers of clay. The SPT-N value of ground is
from5to30. Theinitial groundwater level was36m
belowgroundlevel anditremainedhydrostaticbefore
commencement of excavation. Most excavations are
surroundedby4-storeyto7-storeybuildingsbutsome
high-risebuildings(upto12-stories)arealsoobserved
onsite.
3 OBSERVATIONS
Monitoring instruments installed on site included
inclinometers inside the diaphragm wall and soils,
0
5
10
15
20
25
30
35
40
-5 15 35 55
Lateral wall displacement(mm)
D
e
p
t
h
(
m
)
1m thick ,
36m deep
diaphragm
wall
Excavation
depth 19.57m
Silty Sand
Silty Sand
Silty Clay

Silty Sand
Silty Clay

Silty Sand
Silty Clay
Figure 1. Lateral wall deformations and ground profile
at O6.
0
5
10
15
20
25
30
35
40
-8 12 32 52 72
D
e
p
t
h
(
m
)
1m thick ,
38m deep
diaphragm
wall
Excavation
depth 21.7m
Silty Sand
Silty Clay
Silty Sand
Silty Clay

Silty Sand
at O7.
observation wells, standpipe/electrical piezometers,
vibrationwiregaugeonstruts, tiltmeter andprcision
levellingonbuildings.
Figures 1 to 3 showthelateral wall deformations
observedfromO6, O7andO8at thefinal excavation
stagetogether withtheir groundprofiles, respectively.
It isevident that themaximumlateral wall movement
varies from32 to 64mm. The shape of lateral wall
deformationis acantilever at theshallowexcavation
394
0
5
10
15
20
25
30
35
40
-5 5 15 25 35
D
e
p
t
h
(
m
)
1m thick,
37m deep
diaphragm
wall
Excavation
depth 20.9m
Silty Sand
Silty Clay
Silty Sand
Silty Clay
Silty Sand
at O8.
0
5
10
15
20
25
30
0 10 20 30 40 50 60 70
Distance to the wall(m)
S
e
t
t
l
e
m
e
n
t
(
m
m
)
Shallow excavation
End of excavation
Figure4. Surfacesettlement at O6.
0
5
10
15
20
25
30
0 10 20 30 40 50
Distance tothe wall(m)
S
e
t
t
l
e
m
e
n
t
(
m
m
)
Shallow excavation
End of excavation
level but tends to becomeprop-modethedeeper the
excavation.Thedepthwherethemaximumlateral wall
movement occurs is generally at the same depth of
excavation.
Figures 4 to 6 present induced surfacesettlement
andtheobservedmaximumsurfacesettlementisupto
20to28mmatthecompletionof excavation.However,
0
5
10
15
20
25
30
0 10 20 30 40 50 60 70
Distance to the wall(m)
S
e
t
t
l
e
m
e
n
t
(
m
m
)
Shallow excavation
End of excavation
0
5
10
15
20
25
30
10/6/03 1/4/04 4/3/04 7/2/04
Measured date
P
i
e
z
o
m
e
t
r
i
c

l
e
v
e
l
(
m
,
b
g
l
)
ELP001 31mbgl
ELP002 36mbgl
ELP003 31mbgl
Start of excavation
End of excavation
Figure7. Piezometriclevelsinsidetheexcavationat O7.
0
1
2
3
4
5
6
7
8
05/25/03 08/23/03 11/21/03 02/19/04 05/19/04 08/17/04
Measured date
P
i
e
z
o
m
e
t
r
i
c

l
e
v
e
l
(
m
,
b
g
l
PS005 27m bgl
PS006 34m bgl
PS007 27m bgl
PS008 34m bgl
Start of ex-
cavation
End of excavation (

Figure8. Piezometriclevelsoutsidetheexcavationat O6.
asshowninFigures4to6, limiteddatawasavailable
for theareaclosetotheexcavation.
Since piezometers inside the excavation at O8
were all broken, only piezometric levels inside the
excavation at O6 and O7 were measured, as indi-
catedinFigure7. Thepiezometric levels continueto
decreaseasexcavationprogresses. Figure8indicates
the piezometeric level outside the excavation. It is
395
Figure9. Therelationshipof
vmax
,
hmax
andH
e
apparentthatitisaffectedbyachangeof porepressure
insidetheexcavationthereforeit shows aslight drop
asexcavationprogresses.
Theloadsonstrutswerealsomeasuredandthemax-
imumobservedmeasuredloadis intherangeof 220
to320tons.
4 DISCUSSIONS
4.1 Maximum movements
Figure9presentstherelationshipof maximumlateral
wall movements andexcavationdepthat O6, O7and
O8. Itisfoundthatthemaximumlateral wall displace-
ment (
hmax
) in relation to themaximumexcavation
depth(H
e
) isapproximately 0.034to0.3%. Theratio
of
hmax
/H
e
iscomparativelyhighatO7andthiscould
beaccredited to less stiffness of thesupporting sys-
tem. Mana & Clough (1981) described the ratio of
hmax
/H
e
mainlyintherangeof 0.4to2.0%. Ouet al.
(1993) reportedthatthesameratiowasapproximately
from0.2to 0.5%for excavations inTaipei. Theratio
obtained fromthis study tends to be lower and it is
considered that varying ground conditions could be
thereasonfor thedifference.
AsshowninFigure10, thedepthwheremaximum
lateral wall movement occurs is inarangeof 0.86to
1.26H
e
. Themaximumlateral wall movementisclose
totheexcavationlevel.
Thelateral wall movementatO8issmaller thanthe
othersanditisassumedasemi-top-downconstruction
method might provide a greater strut stiffness and
reducelateral wall movement intheexcavation.
Figure10. Therelationship of thedepth of themaximum
lateral wall displacement (D
m
) andexcavationdepth(H
e
).
4.2 Surface settlement troughs
AsshowninFigure9,themaximumsurfacesettlement
(
vmax
) is approximately 0.050.13% of H
e
. Wang
(2003) suggested that
vmax
induced by the excava-
tioninKaohsiungmightreach0.04%to0.25%of H
e
.
Observationsinthisstudyappeartobeconsistentwith
Wangsconclusion.
Further, the relationship between
vmax
and
hmax
wasexploredandFigure9presentsthedistributionof
thecalculated ratio of
vmax
to
hmax
. It appears that
theratio of
vmax
/
hmax
falls inarangeof 0.3to 0.75
at O6andO7but tends to begreater at O8(approx-
imately from1.3to1.8). Sincedifferent construction
methods (bottom-up for O6 and O7 but semi- top-
downfor O8) wereemployedthisisconsideredtobe
themainreasonfor thedifference. As additional lat-
eral wall movements at O6 and O7 were generated
by thedelayedinstallationof struts at several stages,
this might reduce the ratio of
vmax
to
hmax.
Some
datafromO6 and O7 werethereforeignored, hence
it is suggested that the ratio of
vmax
/
hmax
varies
from0.5 to 0.7 at O6 and O7. However, only lim-
iteddatawerecollectedfor areasnear theexcavation
(0 to 5m from the excavation). This might affect
the accuracy of the measurement of ground surface
settlement.
Clough & ORourke (1990) suggested that the
influencezoneat surfacelevel affected by theexca-
vation in sand was two times the excavation depth
for the ground behind the wall retaining the soil,
but it tends to be greater (up to three times the
excavation depth) at O6, O7 and O8. Pumping from
deeper levelsunder groundisthoughttobethereason
for that.
396
0
10
20
30
40
50
60
6/18/03 9/16/03 12/15/03 3/14/04 6/12/04 9/10/04
Measured date
S
e
t
t
l
e
m
e
n
t
(
m
m
)
SB69
SB70
SB71
SB81
SB87
Start of ex-
cavation
End of ex-
cavation
Installation of CCPs
Figure11. Buildingsettlement at O6.
4.3 Use of chemical churning pile
Chemical churningpiles(CCP) arecommonlyusedin
Taiwaninorder to reducetheinfluenceonbuildings
resultingfromadjacentexcavation. Materialsusedfor
CCPgenerallyincludecementandwater. Woo&Moh
(1990) reported that the movements of the building
could beprevented if thetip of thechemical churn-
ingpilewasinstalleddeeper thanthepotential failure
surface.CCPswereusedatO6inordertoprotectstruc-
tures nearby. For the design of CCPs, piles have to
penetrateat least 2meters belowthe45
activefail-
uresurfaceof theexcavation (Maeda- LongdaJ oint
Venture, 2003). CCPswereinstalledbetweentheexca-
vation and buildings, approximately 7.59.0mfrom
theexcavation. Thediameter of thepileis0.35mand
thelengthof thepilevariesfrom12.0to14.5m. Dur-
ingtheconstruction,pressureusedforjet-groutingwas
keptat19.6MPa.Therateof rotationandraisingof the
rodwere20rpmand4.0min/m,respectively.Although
CCPswereinstalledaroundO6, thebuildingsadjacent
to O6 continued to settle after installation of CCPs,
as showninFigure11. CCPs may only providevery
limitedinfluenceonreducingcontinual settlement of
buildingsandthisisconsistent withobservationsthat
Hsiung (2002) reported. Insufficient CCPs stiffness
couldbeapossiblereason.
4.4 Prop load
Peck (1969) suggested that the prop load could be
estimatedassociatedwiththeapparent earthpressure
method. Considerationof casehistoriesfromdifferent
projects, Twine & Roscoe (1997) revised the appar-
ent earthpressuremethod. Figures12and13present
thecomparisonof strut loadfromfieldmeasurement
0
5
10
15
20
0 50 100 150 200 250
Strut load(kPa)
D
e
p
t
h
(
m
)
Peck
Field measurement
0
5
10
15
20
0 50 100 150 200 250
Strut load(kPa)
D
e
p
t
h
(
m
)
Twine and Roscoe
Field measurement
Figure12. Comparison of prop load fromfield measure-
ment andapparent earthpressureat O6.
0
5
10
15
20
25
0 50 100 150 200 250
Strut load(kPa)
D
e
p
t
h
(
m
)
Peck
Field measurement
0
5
10
15
20
25
0 50 100 150 200 250
Strut load(kPa)
D
e
p
t
h
(
m
)
Twine and Roscoe
Field measurement
Figure13. Comparison of prop load fromfield measure-
ment andapparent earthpressureat O7.
and apparent earth pressure suggested by Peck and
TwineandRoscoe.ItsuggeststhatPecksmethodobvi-
ously underestimates theactual loadonthepropbut
themethodsuggestedbyTwineandRoscoeprovidesa
397
moreaccuratepredictioninpractice. Suchconclusion
is similar to those by Ou et al. (1998) and Hsiung
(2002).
5 CONCLUSIONS
This study provides an opportunity for exploration
of groundbehaviours inducedby excavationinsand.
Based on this study, conclusions can be made, as
follows:
1. Thelateral wall movement inducedbyexcavations
selected in this study is up to 0.3% of the exca-
vationdepth. Suchdisplacement islessthanthose
foundfromseveral relatedpreviousliteratures. Itis
suspectedthevaryinggroundconditions couldbe
afactor inthedifference.
2. Themaximumlateral wall displacement is0.86to
1.26of theexcavationdepth. Thedepthwherethe
maximumlateral wall movementoccurstendstobe
closetotheexcavationlevel
3. Associatedwithobservationsinthisstudy, theratio
of
vamx
to
hmax
varies from0.5 to 1.8 and the
useof constructionmethod(bottom-upandsemi-
top-down) mightinducetheobviouschangeinthis
ratio.
4. Theinfluencezonecausedbyexcavationcanbeup
tothreetimestheexcavationdepth.
5. The installation of CCPs does not provide an
effective solution for reducing the settlement of
buildings near the excavation. Insufficient CCPs
stiffnessmight beareason.
6. The apparent earth method suggested by Peck
underestimates the prop load. The method sug-
gestedbyTwineandRoscoemight provideamore
acceptableresult inpractice.
ACKNOWLEDGEMENTS
Theauthorswouldliketothankfor thefinancial sup-
port fromNational ScienceCouncil, Taiwan for this
project (Project Number: NSC92-2218-E-151-003).
Data for this research provided by Maeda- Longda
J oint Venture and comments on paper content and
EnglishfromMr. TonyShieldarealsoappreciated.
REFERENCES
Burchell, A. J. 2000, Cairometroline2- constructionprob-
lemsandtheirsolutions,TunnellingAsia2000,NewDelhi
Clough, G. W. and ORourke, T. D. 1990, Construction-
induced movement of insitu walls, Design and per-
formance of earth retaining structure, ASCE, special
publication, No. 25, pp. 439470
El-Nahhas, F. M. 2006, Tunnelingandsupporteddeepexca-
vationsintheGreater Cairo, International symposium on
utilization of underground space in urban area, Sharm
El-Sheikh, Egypt
Hsieh, P. G. and Ou, C. Y. 1998, Shape of ground sur-
facesettlement profilescausedby excavation, Canadian
Geotechnical Journal, Volume35, No. 6, pp. 10041017
Hsiung, B. C. 2002, Engineering performance of deep
excavationsinTaipei, PhD thesis, Universityof Bristol
Liu,G.B.,Ng,C.W.W.andWang,Z.W.,2005,Observedper-
formanceof adeepmultistruttedexcavationinShanghai
SoftClays, Journal of Geotechnical and Geoenvironmen-
tal Engineering., volume131, issue8, pp. 10041013
Maeda- LongdaJ oint Venture, 2003, Working proposal for
houseprotectionat O6station(inChinese)
Mana, A. I. and Clough, G. W. 1981, Prediction of move-
mentsfor bracedcut inclay, Journal of the Geotechnical
Division, Proceeding of theAmerican Society of Civil
EngineersVolume107, No. GT6, pp. 759777
Ou, C.Y., Hsieh, P. G. andChiou, D. C. 1993, Characteristics
of groundsurfacesettlementduringexcavation, Canadian
Geotechnical Journal, Volume30, pp. 758767
Ou, C. Y., Liao, L. T. andLin, H. D. 1998, Performanceof
DiaphragmWall ConstructedUsingTop-DownMethod,
Journal of Geotechnical andGeoenvironmental Engineer-
ing, volume124, issue9, pp. 798808
Peck, R. B. 1969, Deep excavation and tunneling in soft
ground, proceeding of the 7th international conference
on soil mechanics and foundation engineering. Mexico
City, stateof art volume, pp. 266277
Twine, D. and Roscoe, H., 1997, Prop loads guidance on
design, Funders Report FR/CP/48, 224 pages, CIRIA,
London
Wang, C. C. 2003, thestudy of engineering characteristics
of deepexcavations inKaohsiung, NSC research report,
Taiwan(inChinese)
Wong, I. H., Poh, T. Y. and Chuah, H. L. 1997, Perfor-
manceof Excavationsfor DepressedExpressway inSin-
gapore, Journal of Geotechnical and Geoenvironmental
Engineering, Volume123, Issue7, pp. 617625
Woo,S.M.andMoh,Z.C.1990,Geotechnical characteristics
of soil in Taipei basin, Proceedings of the 10th South-
east Asian Geotechnical Conference, Volume 3, Taipei,
pp. 5165
398
Environmental problemsof groundwater aroundthelongest expressway
tunnel inKorea
S.M. Kim, H.Y.Yang& S.G.Yoon
Bau Consultant CO. LTD., Seoul, Korea
ABSTRACT: Long tunnels areusually located in mountainous areas with limited infrastructure. Therefore
verylittleisknownaboutthegeological, hydrogeological andgeotechnical conditions: thelonger thetunnel, the
higher theprobability of encounteringadverseconditions for tunneling; thegreater thecost anddurationfor
tunnel construction. Injetunnel will bethelongest expressway tunnel inKorea. Incaseof Injetunnel, Special
problemsrelatedtothelengthof thetunnel arefor exampletypicallythelogistics, ventilationandenvironmental
impacts. Eventhoughmodernexcavationmethods of tunnels havebeendeveloped, various types of problems
suchas changeingroundwater distributionandtransformationof geographical features still remain. It is not
uncommonthatprivatewellsandsmall streamsareusedfor dailylifeintheregionswheremountaintunnelsare
located.Thenserioussocial problemssuchaswell waterlevel fall, beingattributabletotunnel excavationoccurs.
In thedesign stage, weevaluated that thequantity of leakagewater into tunnels and groundwater drawdown
areawassimulatedusingnumerical modelingsuchasMODFLOWandMAFICtoreduceadverseeffectsonlife
environment aroundtunnels.
1 INTRODUCTION
Nowadays road construction projects in Korea have
shown a tendency toward linear route designs, for
the purpose of increasing running speed and reduc-
ingtransportationtime. Moreover, agrowingnumber
of projects aredesigned with tunnels and bridges in
ordertominimizedamagetothenatural environments,
whichcanbecausedbyslopecutting.
However, someroad construction projects, which
werestarted after theyear 2000, provoked thedelay
of projectsandcreatingalawsuit battle. Suchaprob-
lemprecipitatedbecausetheimpacts of groundwater
outflow during tunnel excavation were dealt with
superficial experiences rather than scientific verifi-
cation. Since 1998, there have been four important
environmental conflicts related to tunnel groundwa-
ter outflow in Korea. Every tunnel construction that
was mentioned above has passed legal environmen-
tal impactassessment(EIA)procedures.Nevertheless,
theseconstruction projects weredelayed or stopped,
within a span of 2 or 3 years on average, due to
strongobjectionsfromcivil environmental NGOsand
local residentsneartheprojectsites.TheNGOinsisted
thatthegroundwateroutflowduringtunnel excavation
accompanied moreserious secondary environmental
impactssuchasabrupt changes4MitigationPlanfor
Environmental Impactsof Groundwater inTunnelsin
thefaunaandfloraecologynear theupper part of the
tunnel sites. (Lee, J. et al 2005).
Generallyspeaking, themovement of groundwater
is based on meteorology conditions, surface vegeta-
tion, hydrogeological conditionsandsoon. Becausea
tunnel isalinear structure, thepossibilityof changing
hydrogeology conditions is highandit is difficult to
avoid adverse effects to life environment in regions
where small-scale water usage remains. Therefore,
manystudiesonthequantityof leakagewaterintotun-
nelsandgroundwater drawdownareaareexecutedto
reduceadverseeffectsonlifeenvironmentaroundtun-
nelsaswell asretainsafetyandworkabilityintunnel
constructions.
In this study, weperformed thefollowing investi-
gationandnumerical analysistoevaluatetheenviron-
mental influence of groundwater with excavation of
tunnel inmountainousarea;
GroundInvestigation: Boreholelogging, geophys-
ical survey, BIPS, Lugeontest, groundwater level
monitoring, water well surveyandsoforth
Numerical Analysisfor groundwater flow: Contin-
uousandfracturemediamodel
Evaluationof influence: drawdownof groundwater
level, groundwater inflowrateintotunnel Etc.
399
2 OVERVIEWOF THETUNNEL
The longest tunnel in Korea will be the Inje tunnel
withthelengthof 11km. Thetunnel consists of two
parallel double-lanetubeeachwithawidthof 14.5m.
Thetubesareconnectedbycrosstunnel asemergency
facility every 250mand 750m. The gradient of the
tunnel fromChuncheontoYangyangis1.95%down-
ward. AnInclinetunnel withthelengthof 1.5kmwas
designedonthepurposeof accessfor excavationface
andescapetunnel inemergency.Table1showsgeneral
overviewof thetunnel.
3 GROUNDINVESTIGATION
3.1 General overview on the topology and
geological conditions
The dominant topographic features of the area are
ruggedmountainsandafewof streams. Thewestern
part of theareaislower altitudeandmorerelief than
theeasternpart. Thisareashowsearlymaturestagein
geomorphologiccycle.
This areaconsists of Pre-Cambrian porphyroblas-
ticgneiss, bandedgneiss, andJ urassicbiotitegranite,
two-micagranite, Cretaceous basic andacidic dikes,
and Quaternary alluviumand diluviumas shown in
Figure3.
3.2 Ground investigation
Wecarriedout geological andgeotechnical surveyas
showninFigure2.
Figure1. Locationof Injetunnel anditsvertical sectionprofile.
4 NUMERICAL ANALYSIS
4.1 General aspects of groundwater flow model
The environmental assessment of tunnel groundwa-
ter aims to forecast the drawdown and variation of
domestic groundwater near planned routes due to
tunnel excavation. As usual, numerical methods of
groundwater flowmodelingareusedfor theenviron-
mental impact assessment of tunnel groundwater.
The programpackages used for the modeling of
tunnel groundwater outflow are different fromeach
other, depending on the regional groundwater level
variation in wholetunnels or small-scalegroundwa-
ter level variationinnarrowfracturezones. Normally,
continuousnumerical model packagesarecommonly
usedfor modelinglarge-scalegroundwater flowvari-
ations, while fracture media models are applied for
Table1. Layoutsof InjeTunnel.
Length 10,965m
Alignment R=2000R=4991.3
Gradient 1.95%
Shapeof Portal Archwall type
TrafficType 2tube(2- lane)
Ventilation J et Fan: 80
VentilationShaft : 4
ExcavationMethod Drill andBlast
EmergencyFacilities Escapeconnectingtunnel : 44
Interval : 750mfor vehicle
: 250mfor human
400
groundwater flow modeling in smaller zones com-
posedof jointedrockaquifers.
Similarly, when we carry out the modeling of
groundwater outflow, the MODFLOW package is
commonly used for numerical analysis of regional
groundwater flows in whole areas of tunnel and
drainage, andMAFIC isappliedfor detailedground-
water modeling in small discrete zones in tunnel
areas.
In this study, two representative models are used
toforecast thetunnel groundwater outflowpatternas
followingTable2.
The ground conditions of modeling section are
Very-good(rocktype1)Moderate(rocktype3)grade
Figure2. Flowchart of GroundInvestigation.
Figure3. Geological featuresof Injetunnel.
in rock classification and the rock covers of tunnel
understreamandvalleyareaareabout100morhigher.
4.2 Continuous media modeling
4.2.1 Setup of the model
MODFLOW was usedto simulategroundwater flow
as continuous media model in the studied area. The
wholeareais4km3.5km.
The boundary of the model is set at the summit
of northernmountainsandsouthernmountainsinthe
upper andlower end, respectively.
Hydraulic conductivities are classified into fif-
teen groups according to the permeability of the
regions from Lugeon test. The range of value is
4.5E-5cm/sec7E-6cm/sec.
4.2.2 Steady state simulation
Wecalibratedthevalueof groundwater headbetween
the calculated and observed one in order to modify
Table2. Layout of groundwater flowmodeling.
Program MODFLOW MAFIC
Model 3D-Continuous 3D-Discontinuous
Model Model
Section Sta. Sta.
2+5006+500 4+0004+500
Figure4. Groundconditionandmodelingsection.
401
Figure5. Boundaryconditionof themodel.
Figure 6. Distribution of equivalent head in steady state
condition.
the hydraulic model. Figure 6 shows distribution of
equivalent headinsteadystatecondition.
We can see that groundwater inflow was concen-
tratedat tunnel alignment andflowfromthesummit
areas of northern and southern mountain to Bangtae
stream.
4.2.3 Transient state simulation
Resultof Transientflowsimulationisusedtoevaluate
variationof groundwater level withtunnel excavation.
We can know that the change of groundwater level
withtimeisverysmall amount of 0.008m0.015m
beforetunnel excavation.
4.2.4 Change of groundwater level due to tunnel
excavation
Wecanseethedistributionof groundwater level and
drawdownatthecheckpointduetothetunnel excava-
tionasfollowingFigure8andTable3.Thecheckpoints
inmodel arelocatedevery500mfromSta. 3+000to
Sta. 6+500.
Thedrawdownof groundwater duetotunnel exca-
vationis1.92m3.29masyoucanseetheFigure8.
It is small amount in general, but weworried about
theinfluenceongroundwater systemsaroundtunnel.
Therefore, we carried out simulation for waterproof
Figure7. Distributionof equivalent headintransient state
condition.
Figure8. Distributionof groundwater level.
grouting with theaimof reducing water inflow into
tunnel inrockmass.
Groundwater level at Sta. 5+000 will fall down
to 2.03mafter tunnel excavation and water inflow
rateper kmis 0.127m
3
/min. Incaseof usingwater-
proof grouting to theground near tunnel, drawdown
of groundwater level is 0.79mandwater inflowrate
402
Table3. Groundwater level andwater inflowrateinto the
tunnel.
Drawdownof Drawdownof
groundwater level groundwater level
(m) (m)
Checkpoint (without grouting) (withgrouting)
Sta. 3+000 3.20 0.94
Sta. 3+500 2.94 0.87
Sta. 4+000 2.32 0.58
Sta. 4+500 1.92 0.73
Sta. 5+000 2.03 0.79
Sta. 5+500 2.98 1.04
Sta. 6+000 3.20 0.77
Sta. 6+500 3.29 1.05
Water inflow 0.127m
3
/min 0.060m
3
/min
rateper km (730.75m
3
/day) (343.27m
3
/day)
Figure9. Drawdowngroundwater level incaseof applying
waterproof groutingat Sta. 5+000.
per km is 0.060m
3
/min. It can be notified that if
waterproof groutinginto thegroundis performedin
advance, theamount of groundwater inflowinto the
tunnel will decreaseasshowninFigure9andTable3.
4.3 Fracture media modeling
MAFIC(MatrixAndFractureInteractionCode) deter-
mines solutions of flow and pressure by conjugate
gradient finite element methods, and solves solute
transportbyparticletrackingdiscretefracturenetwork
models. Finiteelement meshes arebasedonfracture
networkanalysisandgeometricmodeling.
4.3.1 Set up of the fracture model
Usingthepackageof MAFIC, wesimulatedground-
water flowthroughfracturemodel inthestudiedarea
(Sta. 4+0004+500). Input data was estimated
based on joint data fromdetailed surface and bore-
holesurvey. Infacturemodel, threeof major joint set
wasestimatedasfollowingTable4.
Figure10showsthreedimensional fracturenetwork
andtunnel.
4.3.2 Steady state condition
As thestageof settingboundary condition in steady
statesimulation, weestimatedgroundwater level and
Table4. Major joint set andfault inthisarea.
S.Dof
J oint Length Length Distribution
set Orientation (m) (m) of length
Set 1 80/194 15.36 10.38 Lognormal
Set 2 78/264 12.93 9.63 Lognormal
Set 3 30/124 34.47 13.18 Lognormal
Fault 85/105
Figure10. threedimensional model of fracturenetworkand
tunnel for steadystatesimulation.
Figure11. Distributionof water headinsteady-statecon-
ditionbeforetunnel excavation.
transmissivity(T) basedonresults of MODFLOW in
steady-state flow condition, water well survey and
boreholetest andsoon.
Especially, average transmissivity evaluated from
theresult of Legeontest andBIPSis1.24m
2
/sec.
403
Figure 12. Distribution of water head in transient-state
conditionbeforetunnel excavation.
Figure 11 shows distribution of water head in
steady-stateconditionbeforetunnel excavation. Itwas
set asthevalueof initial water headintransient flow
condition.
4.3.3 Transient state condition
We simulated the groundwater flow due to tunnel
excavationintheconditionof transient state. Astun-
nel excavation was performed, groundwater flowed
into tunnel and water head fell down suddenly and
afterwardgradually.
Inthecaseof 500mtunnel advance, water inflow
rateperkmintotunnel is0.10m
3
/min.Thatwassimilar
totheresult of MODFLOW(0.127m
3
/min).
5 CONCLUSIONS
Inspiteof thedifficulty for estimationof water flow
characteristicsintotunnel beforeexcavation, itisvery
important topredict thevariationof hydraulicsystem
duetothetunnel.
Inthisstudy, tworepresentativemodelsareusedto
forecast thetunnel groundwater outflowpatternsuch
asMODFLOWandMAFIC.
Basedontheresult of numerical modeling, it can
benotifiedthatif waterproof groutingintotheground
is performedinadvance, theamount of groundwater
inflowintothetunnel will decreasecomparedtonon-
groutingcase.
REFERENCES
Kim, S. & Yoon, S. 2006a. Environmental Problems of
Groundwateraroundthemountaintunnel (2),Tunnel Tech-
nology Journal of KoreanTunnel Association, Vol. 8, No. 2,
8298.
Kim, S. & Yoon, S. 2006b. Environmental Problems of
Groundwateraroundthemountaintunnel (1),Tunnel Tech-
nology Journal of KoreanTunnel Association, Vol. 8, No. 1,
6478.
Kim, S.,Yoon, S. &Yoon, J. 2005. Influenceof Groundwater
intheRailwayTunnel ConstructionKorea Society for Rock
Mechanics 2005 Fall National Conference, October 27,
2005, Chuncheon, Korea.
Lee, J. & Lee, Y. N. 2005. A Study on Mitigation Plan for
Environmental Impactsof GroundwaterinTunnels. Korea
Environment Institute, 536.
404
Measurementsof grounddeformationsbehindbracedexcavations
T. Konda
H. Ota&T.Yanagawa
Osaka Municipal Transportation Bureau, Osaka, Japan
A. Hashimoto
Kotsu Service Co., Ltd., Osaka, Japan
ABSTRACT: Thedeformationbehavior of earthretainingwall andbacksidegroundduetobracedexcavation
weremeasured at theconstruction sites of OsakaSubway LineNo. 8. It was observed that backsideground
movedto excavation& downsidealongacircular slipfollowingthewall deflection, andits influencespread
tothegroundsurface. Thebacksidegroundsurfacesettlement near thewall wassuppressedduetothefriction
between wall and surrounding ground, and maximumsettlement occurred at somedistancefromwall in the
4560degreeareafromthebottomof thewall. Thebacksidegrounddeformationareadecreasedexponentially
withthedistancefromthewall. Asaresult eachobserveddatafrommonitoringsiteswasanalyzedstatistically
accordingtothesoil characteristic, wall deflectionandsurfacesettlementinprominentsoftclaylayerwerelarger
thanthoseinprominent sandandgravel layer. Therelationbetweenthewall deflectionareaA
andbackside
groundsettlementareaA
s
wasthatA
s
=(0.20.3)A
inprominentsoftandsensitiveclayeylayer.Thebackside
groundsettlement of thisobserveddataconvergedsmaller thanpast data.
1 INTRODUCTION
There are many neighboring constructions in recent
subway works. Inorder toguaranteetheconstruction
safety and control the influence on the surrounding
environment to the minimum, it is necessary to use
theobservational methodandpredictionresultseffec-
tively.Forexample,inthecaseof cutandcovermethod,
whichisusuallyusedtoconstruct thesubwaystation,
it isnecessarytoevaluatethebacksidegrounddefor-
mationof earthretainingwall andgroundwater level
fluctuation, tocarefullymonitor someparticular con-
cern points, and of course to estimate the safety of
timberingof acut andtheexcavationbottom.
In this paper, earth retaining wall deflection and
backsidegrounddeformationduetobracedexcavation
wereevaluatedcircumstantiallyinopencut construc-
tionsites, whichmainexcavationsoil wasthealluvial
soft clay, inOsakaSubway LineNo. 8. All observed
datain11monitoringsitesweregottogether andana-
lyzedwiththesameidea. Theevaluatedresults were
describedfor eachsoil characteristic. Somebehaviors
of grounddeformationduetobracedexcavationbased
onthesemonitoringresultswereconsidered.
2 CONSTRUCTIONSUMMARY
OsakaSubwayLineNo. 8wasconstructedat theeast
sideof Osakacity asnorthsouthrail routeinunder-
ground. Every stationwas built by opencut method.
Inordertocontrol thesettlementatthegroundsurface
andtheinfluenceontheneighboringgroundenviron-
ment, seepagecontrol methodwasadoptedinall of the
stationconstructionsiteswithextendingearthretain-
ingwalls, wherethemost wereSoil MixingWalls, to
thePleistoceneclayeylayer under theartesianaquifer.
Itispossibletoclassifythreeareasroughlybasedon
thedifferenceof soil characteristicasFigure1shows.
Thisnewrail routelocatedattheEastsideof Uemachi-
plateau, and this area was an inland sea in ancient
times.
(1) NorthArea
This area interleaved theYodo-river, and constitutes
thealluvial layer, upper andlower Pleistocene. Allu-
vial layer is composed of a fine sandy layer having
lowuniformity coefficient Uc, andsoft andsensitive
clayey layer havingunconfinedcompressivestrength
q
u
=40100kN/m
2
. Tenma-gravel layer accumulated
405
Figure1. Plainviewandsoil crossingviewof OsakaSubwayLineNo. 8(Unei et al., 2002).
continuously with just about uniform thickness as
upper Pleistocenelayer withhighpermeability. Lower
Pleistocene layer Osaka Group is laminated ground
withstiff clayeylayerof cohesionc=300600kN/m
2
,
and dense sandy layer of Standard Penetration Test
N-valueis morethan 60. Each OsakaGroup dips in
thesouthandeast directionneighbor theYodo-river.
(2) MiddleArea
Thisareaisconstitutedof thethinalluvial layer, upper
andlower Pleistocene. Upper Pleistocenelayer iscon-
stituted with stiff clayey layer (N-value=510) and
verydensesandandgravel layer withhighpermeabil-
ity. Lower Pleistocenelayer OsakaGroupislaminated
ground with stiff clayey layer (c=200400kN/m
2
)
anddensesandy layer (N-value>60). Thegroundof
thisareaisharder thanother areascomparatively.
(3) SouthArea
ThisarealocatedinNeyagawa-lowlandanditisdiffer-
entfromtherestareaforthattheverysoftandsensitive
alluvial clayey layer (N-value=about zero, q
u
=40
100kN/m
2
, liquid limit I
L
=about 1.0) deposited
withinabout15meters.UndertheAlluvial layer,upper
Pleistocene layer (N-value=2260) with high per-
meability and lower Pleistocenelayer OsakaGroup,
withstiff clayeylayer(c=200400kN/m
2
) anddense
andslightlycohesivesandylayer (N-value=30 and
above60). Each OsakaGroup dips in thesouth and
east directionparallel toUemachi-plateau.
3 BEHAVIOR OF BACK GROUND
DEFORMATIONDUETOBRACED
EXCAVATION
Here, somemonitoring results of theearth retaining
wall deflectionandbacksidegrounddeformationdue
tobracedexcavationat theA-sitearerepresentedand
thedispersedbehavior inbacksidegroundfromwall
togroundsurfaceisdescribedasFigure2shows(Ito
et al., 2006).
Alluvial layer andupper & lower Pleistocenelayer
areformattedfromthegroundsurfacenear theA-site.
Alluvial layer is composedof afinesandy layer (1st
aquifer)with2mthicknesshavinglowUc andN-value
equalstoabout2, andasoftandsensitiveclayeylayer
withN-value=03, I
L
=0.41.0, c=20100kN/m
2
.
ThisAlluvial clay layer is atypical soft layer in this
constructionsite. Ontheotherhand, upperPleistocene
sandy and gravel layer Tsg (2nd aquifer) with par-
tially scatteredgravel, lower Pleistoceneclayey layer
Oc3 (c=about 400kN/m
2
) and lower Pleistocene
sandy layer Os3 (3rd aquifer, N-value >60) existed
continuouslyunder theAlluvial layer.
Seepage control method was adopted in this
construction site with extending the wall to the low
406
Figure2. Earthretainingwall deflection, backsidegrounddeformationvector anddistributionof backsidegroundsurface
settlement at theA-site.
permeable layer Oc3 (GL-26.3m). The excavation
widthwas16.2m, andthefinal excavationdepthwas
GL-21.5m, in short, the wall penetration depth is
4.8m. Becausethesoil improvementworkwascarried
out near the monitoring site after GL-1.5mexcava-
tiondepthtoprotect thedeficient part of wall, initial
valueof eachobserveddatawassetthevalueafter 1st
excavationtoGL-4.5m.
Themaximumwall deflectionoccurredataslightly
deeper depth fromtheexcavation bottomdueto the
bracedexcavationtill the6thexcavationstep, andthis
influences reached thebacksideground and surface.
Forexample, themaximumwall deflectionwas38mm
attheexcavationbottom,whichwasthemiddleofAmc
layer (GL-12.5m), in the5th step. Accordingto this
influences, themaximumhorizontal grounddeforma-
tionof theNo. 1wasabout29mmattheslightlyabove
themaximumdeformation depth of wall. Themaxi-
mumhorizontal grounddeformationof theNo. 2was
about12mminAuclayer andAuslayer. Itwasknown
that thewall deflection caused theback sideground
deformation like the circular slip and this influence
reachedthegroundsurface.
On the other hands, the maximum surface
settlement at theNo. 2was6mm, andthesettlement
distribution was convex downward with 0mm set-
tlement at the point, 20m outside from the wall.
It was thought that the friction between the wall
and surrounding ground supported the surface set-
tlement. These ground movements happened in
the 4560 degree area from the bottom of the
wall.
Themaximumwall deflectionincreasedto51mm
at just above fromthe excavation bottom, which is
the lower of Amc layer (GL-14m), in the 6th step.
Sincethe6thstep, wall deflectiondidnotincreaseany
more. Itwasbelievedthatthecauseof thisinhabitation
phenomenonwasfromthesubgradereactionintheTsg
andOc3layer waslargeandtheeffectsof pre-loadon
thewall.
Figure3 shows thedefinition of symbols used in
thispaper. Therelationbetweentheexcavationdepth
tothemaximumwall deflectionratio
max
/Z
e
andthe
maximumdeformation ratio S
max
/
max
are shown in
Figure4.
East andwest wall deflectionbecamelargedueto
thebracedexcavationinthemiddleof Amc layer till
the4thstep, andS
max
/
max
wastendtodecrease. How-
ever, thewall deflectionwasdecreasedgradually, and
S
max
/
max
wastendtoincreaseadversely.
407
Figure3. Definitionof eachsymbol.
Figure4. Relationbetween
max
/Z
e
andS
max
/
max
.
Therelationbetweentheexcavationdepthtowidth
ratioZ
e
/B, themaximumdeformationratioS
max
/
max
andthedeformationarearatioareshowninFigure5.
Z
e
/B isthenon-dimensional valuefor excavationsize
andincreaseduetothebracedexcavation.
As the increase ratio of the wall deflection and
surface settlement is small till the 4th step due to
the braced excavation in the middle of Amc layer,
S
max
/
max
was tend to decrease. However, S
max
/
max
wasincreasedconverselyduetothebracedexcavation
after the4thstep. Thistendencyoccurredinthesame
waywiththerelationof A
s
/A
.
Figure6showstherelationbetweenthewall deflec-
tion areaA
and thebacksideground surfacesettle-

ment areaA
s
at thebeginning of theexcavation and
after removing all struts. More or less A
and A
s
tend to increasefromthestart to theend. Both side
wall deflectionshaveA
s
=0.2A
relationshipinthe
8thstep, andshiftedtoA
s
=0.3A
relationshipdue
to removal of struts. Unalterably, the west side wall
deflectionwaslarger thantheeast one.
Figure5. RelationbetweenZ
e
/B, S
max
/
max
andA
s
/A
.
Figure6. RelationbetweenA
andA
s
.
Therelationbetweenthedistancefromthewall to
backside, wall deflectionareaA
andbacksideground
deformationA
1
&A
2
areshowninFigure7. Broken
lineistheapproximatecurvebasedonthecalculation
resultbyleast-squaremethodwiththreemeasurement
dataineachexcavationstage.
Thegrounddeformationareaattenuatedgradually
asgettingawayfromthewall positiontill the6thstep.
Ontheotherhand,astheincreaseof thewall deflection
becamesmall, each deformation areachanged little.
Theattenuationrateof thebacksidegrounddeforma-
tionareaonwall deflectionareawasjustaboutsteady,
about 70%at theNo. 1andabout 25%at theNo. 2.
408
Figure7. Influencepropagationaspect tobacksideground
duetobracedexcavation.
andA
s
inthecaseof soft andsensitiveclayeylayer.
4 STATISTICAL COMPARISONBETWEEN
WALL DEFLECTIONANDBACK GROUND
SURFACE SETTLEMENT
Figures 8 and 9 show the relation between the wall
deflection areaA
and backside ground surface set-

tlementareaA
s
atthebeginningof theexcavationand
afterremovingall strutsof all monitoringsiteforexca-
vation soil characteristic. Thebroken lineshows the
observeddatafromfinal excavationstagetoall struts
removal stage.
Thecaseof softandsensitiveclayeylayer isshown
inFigure8. For examplethereweredifferent tenden-
cies between east and west wall deflection behavior,
however, A
s
=(0.20.3)A
relationshipwas given.
Additionally, thewall deformationandsurfacesettle-
ment increased dueto theremoval of struts. A
and
A
s
werelarger thanoneintheother differentgrounds.
Thecaseof alluvial sandy layer is struts. A
andA
s
were larger than one in the other different grounds.
Thecaseof alluvial sandylayer isshowninFigure9.
Therearegreat variances inthis reference, however,
409
andA
s
inthecaseof alluvial sandylayer.
A
s
=(0.10.4)A
relationshipwasgiven.Thedefor-
mation area of wall and surface increased due to
removal of strutsthesameasthebehavior of thesoft
and sensitive clayey layer case. But the amount of
deformationwasrelativelysmall.
Figure 10 shows the predictive method of back-
sideground surfacesettlement based on therelation
between A
and A
s
due to the braced excavation.
A
=A
s
means that theinfluenceof thewall deflec-
tionpasses to thesurfacesettlement directly without
volumechange(volumeloss=0%).
In the case of including the consolidation settle-
ment,A
s
waslargerthanA
accordingtotheliterature.
On the other hand, A
s
was smaller thanA
in other
cases (J SCE, 1993). Someobserveddatainthecase
of alluvial soft claygroundwereaddedtopast exam-
plesshownasFigure10. Becausetheseepagecontrol
methodwas adoptedinNo. 8Lineconstruction, and
actually thechangeof groundwater level was small,
it is possibleto estimatethesmall amount of degra-
dation for the consolidation settlement caused by
the groundwater level. However past observed data
were distributed along the A
=A
s
relation, some
monitoring data in No. 8 Line construction were
located along the A
s
=(0.20.3)A
relation and
undertheA
s
=0.4A
relation.Thebacksideground
settlement of this observed data converged smaller
thanthepast dataandthevolumelossfromA
toA
s
was7080%. Oneof thereasonsfor thisphenomenon
isthoughttobethesupporttotheneighboringground
based on the arching effect developed in backside
ground.
5 CONCLUSIONS
Theconclusionsaredrawnasfollows:
1. Itwasfoundthattheearthretainingwall deflection
causedtheback sidegrounddeformationlikethe
circular slip and spread influences to the ground
surfaceinA-site.
2. The distribution of backside ground surface set-
tlement occurred as convex downward with the
maximumvalueat somedistancefromthewall. It
wasthought that thefrictionbetweenthewall and
around ground supported the surface settlement.
410
andA
s
.
Theseground movements happened in the4560
degreeareafromthebottomof thewall.
3. Theattenuationrateof thebacksidegrounddefor-
mationareaonwall deflectionareawasjust about
steady, about 70%at 2mdistanceandabout 25%
at 9mdistancefromthewall.
4. The relation between the wall deflection areaA
and backside ground settlement areaA

s
was that
A
s
=(0.20.3)A
inprominentsoftandsensitive
clayeylayer, andA
s
=(0.10.4)A
inprominent
Alluvial sandylayer dependedontheconstruction
condition.
5. SomemonitoringdatainNo. 8Lineconstruction
werelocatedunder theA
s
=0.4A
relation. The
backsideground settlement of this observed data
convergedsmaller thanpast data, andthevolume
lossfromA
toA
s
was7080%. Oneof thereasons
for thisphenomenonisthoughttobethesupportto
theneighboringgroundbasedonthearchingeffect
developedinbacksideground.
It is necessary to carry out theevaluation prelim-
inary in detail and use the observational method to
voidsomekinds of risk. It is believedthat this study
will become useful for predicting the influences of
neighboringconstructioninfuture.
REFERENCES
Ito, H., Yanagawa, T., Konda, T. & Hayakawa, K. 2006.
Relationbetweenbracedwall deflectionanddeformation
behavior of backside ground due to braced excavation,
Proc. of Tunnel Engineering, JSCE, vol. 16, 439446(in
J apanese).
J SCE. 1993. Design Calculation Example of Makeshift
Structure based on the J apanese Standard for Cut and
CoverTunneling, Tunnel Library 4, 107(inJ apanese).
Unei, N., Sakaguchi, Y. & Yamaguchi, H. 2002. Challenge
to CloseApproachto or CrossingwithShinkansenLine
and Other Existing Important Structures Construction
planfor subway stationonOsakaMunicipal LineNo. 8,
Tunnels and Underground, Vol. 33, No. 12, 3140 (in
J apanese).
411
Researchontheeffect of buriedchannelstothedifferential settlement of
building
D.P. Liu, R. Wang& G.B. Liu
ABSTRACT: Greatanddifferential settlementoccursunder thebuildingduringanadjacentmetrodeepfoun-
dationpit construction. Factorsinfluencingthebuildingsettlement indifferent constructionstagesareanalyzed
basedonthemonitoringdataandgeologydistributioncondition.Theresultshowsthatthelargelateral deforma-
tionof retainingwall andtime-dependeddeformationof soft clayunder thebuildingarethereasonsthat make
themajorsettlementoccur.Themainfactorinducingdifferential settlementof thebuildingislocal distributionof
adversegeological condition. Somefeasiblemeansof improvementsinpracticearepresented, whichcanreduce
effectivelytheinfluenceof adversegeological conditionsuchasburiedchannel onexcavationandsurroundings.
Thoughthepapertheauthorhopesthatengineerscanattachmoreimportancetothenegativeinfluenceof adverse
geological condition, andtheresultscouldbevaluablereferencetoother engineering.
Shanghai hastherepresentativesoft groundinChina.
Thestratumof Shanghai mainlymadeupof saturated
clay, siltyclayandsand. Most metrostationsandtun-
nelsof Shanghai lieabout20mbeneaththegroundsur-
face. Soil of suchdepthmostlyissoftclaywhichchar-
acteristicssuchaswater content, degreeof sensitivity,
compressibilityandrheologyarenotable, andtheunit
mass, strengthandpermeabilityarebad. Exceptthose,
the adverse geological factors such as buried chan-
nel, shallow-buriedmethane, undergroundbarrier etc.
maybeexistwhichwill doharmtotheconstructingof
deepexcavationandadjacentbuildings.Sotheadverse
geological factorsshouldbepaidenoughattention.
During the construction of a metro excavation in
Shanghai, great anddifferential settlement occurs on
an adjacent building. Factors influencing the build-
ing settlement in different construction stages are
analyzed based on themonitoring dataand geologi-
cal distribution condition. Reasons which effect the
major settlement arediscussed and theresults could
bevaluablereferencetoother engineering.
2 GENERAL SITUATION
Themetro station of this study is designed as atwo
layers underground framestructure. Thetotal length
of thestationis364.7mandthewidthof itsstandard
Figure1. Strut profileof standardsegment.
segmentis24.5m.Theburieddepthof thebottomplate
is17m, andthethicknessof retainingwall is800mm.
As for theedgewells of thestation,theburieddepth
is about 18.6m, thethickness of its retainingwall is
800mm, Pre-stressedsteel pipesof 609mmindiam-
eter (external) and16mminthicknesswereinstalled
at each levels (standard segment: level 1 to level 5;
edgewells: level 1tolevel 6) tosupport theretaining
wall. Theprotectivegradeof theexcavationis grade
1. Thecrosssectionof thestandardsegment isshown
inFigure1. Soil profileandgeological descriptionof
thesoilsunder theadjacentbuildingof theexcavation
aregiveninTable1.
413
Table1. Profileandgeological descriptionof thesoils.
Shear strength(peak)
Serial Bottom Water Unit Void
number Name Thickness/m level /m content /% weight /kNm
3
ratio C / KPa +,
1
filledsoil 1.2 2.93
1
clay 2.6 0.33 34.6 18.2 0.99 21 17.5
siltyclay 4.2 3.87 43.0 17.3 1.21 13 17.0
1
siltyclay 8.0 11.87 49.1 16.8 1.39 14 11.0
11
clay 3.0 14.87 38.9 17.6 1.12 16 14.0
12
siltyclay 6.5 21.37 34.9 17.9 1.02 15 18.5
22
sandysilt 14.5 35.87 32.2 18.0 0.94 4 29.0
2
siltysand Nopenetrated Nopenetrated 26.3 18.8 0.77 1 32.0
Figure2. Planof monitoringpointsandburiedchannel.
2.1 Adjacent buildings and surrounding condition
The buildings around the excavation are relatively
dense.Theprimeprotectionobjectisthenearer build-
ingsonthesouthsideof theexcavation. Great settle-
ment occurredunder ateachingbuildingbelongedto
amiddleschool whichis sited1416maway from
theexcavation on thesouth side. Total length of the
buildingparallel totheexcavationis87.882mandits
widthis 15.188montheeast, 20.883monthewest.
Thebuildingwasconstructedattheendof 1970s, and
it isthereinforcedmasonrystructurewithringbeam,
shallowstripfoundation.
According to thegeological documents, thereare
two buried channels under the building (Figure 2).
Influencedbytheadversegeological factor, therehas
beenabout 400mmtotal settlement happenedbefore
theexcavationconstruct.
2.2 Building settlement
Thecurvesof buildingsettlementdevelopedwithtime
areshowninFigure3after theexcavationconstructed
August, 2005. Figure4shows theaccumulativetotal
Figure3. Buildingsettlement withtime.
Figure4. Accumulativetotal settlement of buildingduring
different stage.
buildingsettlementof several mainmonitoringpoints
duringdifferent constructionstages.
From Figure 4, monitoring points of F13 and
F23 have the large accumulative and differential
414
Figure5. Buildingaccumulativetotal settlementindifferent
stages.
settlement. Until March, 17, the differential settle-
ment in longitudinal direction between F13 and F23
has reachedto 45mmandincliningslopeis 0.0005;
andwhichis 60.7mmand0.0041respectively trans-
versely. Inclining slope of the two directions have
exceeded0.003whichistheallowablevalueaccording
to national criterion. Many cracks appeared in walls
of thebuilding, somelocal width of cracks is about
10mm.sodidtheparallel crakesonthegroundsurface
betweenthebuildingandtheexcavation.
3 BUILDINGSETTLEMENT ANALYSISIN
DIFFERENT STAGES
Greatandobviousdifferential settlementof thebuild-
ingistheresultaffectedbyseveral factors, whichhave
dissimilar effect indifferent constructionstages. Fig-
ure5and6respectivelyshowstheaccumulativetotal
settlementandsettlementvelocitywithdifferentparts
of thebuildingduringthemainconstructionstages.
Basedonthemainconstructionstages of thepart
of excavationnearthebuilding, settlementof different
parts of thebuilding in different construction stages
wasanalyzed.
3.1 Stage of retaining wall construction
Generally, deformationof thesurroundinginducedby
retainingwall constructionisrelatively small andthe
effectedareais limited. Thebuildinglies outsidethe
influencingareaof theretainingwall constructionso
that the effect can be neglected when the building
settlement isanalyzed.
Field data of the building in retaining wall con-
structionstageisanalyzed. Constructionof thepartof
retainingwall beganonOct.4, 2005andfinishedon
Oct.17. Duringtheperiodbuildingsubsidedlittleand
Figure6. Buildingsettlement velocityindifferent stages.
theaveragesettlementis2.7mm. Settlementof F21to
F23arelessthan1mmandthatof F15toF16are1.2to
1.6mm, asshowninFigure4. Takethenormal fluctu-
ationof monitoringdataintoaccount, theconstruction
of retainingwall haslittleinfluenceonbuildingsettle-
ment, whichtests theconclusionof other correlative
studies.
InFigure5, settlement indifferent locationof the
buildingisof variationfromtheminimum0.5mmof
F23 to the maximum5mmof F19. Because of the
existingof theburiedchannel, buildingsettlement of
the location under which is the buried channel are
greater thanthatof normal subgradesoil.Atthattime,
thereisnootherconstruction, sothemainreasonlead-
ingtothesettlement canbeensuredastheexistingof
thehighrheologyof thesoft soil.
3.2 Stage of excavation
Sequence of the excavation is fromthe east to the
west. Differential settlementbetweenthetwoendswas
45mmtill March, 17. InFigure7, thelateral deflection
of theretainingwall at Q17is gatheredaccordingto
eachstageof theexcavation. Sodothedeformationof
excavationandthesurroundingsasshowninTable2.
415
InFigure7, largedeformationproducedattheearly
stageof theexcavation. Therehasbeen32mmlateral
deformation after the 3rd level strut installed which
is mainly relatedto hightime-dependeddeformation
of the3rd level soft soil. In addition, long exposure
timewithout struts andlonglayinguptimeafter the
installingof strutsalsomadefor thelateral deflection
of theretainingwall increasing.
Cooperating with theexcavation, basal heaveand
thegroundsurfacesettlementcausedbylateral deflec-
tionof thewall areincreasing, whichalso accelerate
thesettlementof thebuilding. (Figure5andFigure6)
Settlement velocitiesof pointsF6F8intheloca-
tion of buried channel aredoublegreater than those
Figure7. Lateral deflectionof thewall at Q17.
Table2. Deformationof excavationandthesurroundings.
Accumulativetotal ondifferent Stageof structure
moment / mm Stageof excavation construction
Structure
Beforethe Theexcavation construction Increment / Velocity/ Increment/ Velocity/
Item excavation finished finished mm (mm/d) mm ( mm/d)
Lateral Q13 31.8 59.1 76.4 27.3 1.05 17.9 0.15
deflection Q17 32.7 47.2 60.3 14.5 0.60 13.1 0.13
of thewall Q19 49.7 61.6 68.6 11.9 0.70 7.0 0.07
Basal L6 16.5 45.9 51.1 29.4 1.13 5.2 0.18
heave L7 7.7 34.2 35.2 26.5 1.02 1.0 0.03
L8 7.7 49.6 41.0 41.8 1.67 8.6 0.28
L9 24.4 63.0 63.3 38.6 2.27 0.3 0.01
L10 32.1 50.4 60.3 18.3 1.66 9.9 0.23
Ground D65 28.2 52.3 83.2 24.1 0.93 30.9 0.30
surface D66 35.3 76.9 118.8 41.6 1.60 41.9 0.41
settlement D67 49.6 93.6 130.5 44.0 1.76 36.9 0.35
D68 52.3 86.3 128 34.0 2.0 41.7 0.37
D69 70.6 82.2 130.0 11.6 1.05 47.8 0.41
of points F1F4. Settlement velocities of points
F13F15, F17F20 arealso greater than theoth-
ers. (Figure8) Themaximumlocal incliningslopeof
the building in longitudinal direction is 0.0006, and
whichis 0.0052intransverse. They all exceededthe
allowablevalue0.003. Sothereasonof differential set-
tlement of thebuildingcouldbemainly attributedto
thelocal distributionof buriedchannel.
3.3 Stage of structural construction
Comparedwiththeexcavationstage, lateral deflection
of theretainingwall, basal heaveandbuildingsettle-
ment inthisstagearesteadyrelativelybecauseof the
strengtheningof thestructural stiff. But thebuilding
andgroundsurfacesurroundtheexcavationstill sanka
littleattherate0.30.4mm/d. Thiscanbeexplained
Figure 8. Comparison of settlement velocity between the
partsinburiedchannel andtheothers.
416
asthereasonof time-dependeddeformationof thesoil
aroundtheexcavation.
4 REASONSANALYSIS
1. Relatively great lateral deflection of theretaining
wall causesthewholesettlementof groundsurface.
Time-depended deformation of the soft soil dis-
turbedfromtheconstruction, overloadof thebuild-
ing, long exposure time without struts and long
layinguptimeafter installingthestruts madethe
lateral deflectionof theretainingwall increasing.
2. Time-depended deformation of the soft soil
under the building increased the whole building
settlement.
Soft soil has obvious rheology and sensitivity.
Time-dependedcurveof typical softsoil inShang-
hai is shown in Figure 9. High time-depended
deformation becauseof thedisturbing of excava-
tionconstructioncauses thesustainingsettlement
of thebuilding.
3. Differential settlement of the building mainly
attributes to thelocal distributionof adversegeo-
logical action.
Settlementsindifferentlocationof thebuildingare
obviously of variation becauseof thelocal distri-
butionof buriedchannel, thebuildingsettlements
of thepart under which is theburied channel are
greater thanthoselocatedonnormal subgradesoil.
5 CONCLUSION
1. Optimizingthedesignof theretainingsystemof the
excavationincomplexgeological condition, short-
ening the exposure time of retaining wall during
excavation should be done to control the lateral
deflection of retaining wall, which in turn could
reducetheeffect onthesurroundings.
Figure9. Dynamictriaxial rheologytest of soft clay.
2. The adverse geological condition such as buried
channel should be paid enough attention. The
buriedchannel shouldbestrengthenedorseparated
from the building to decreasing the disadvanta-
geouseffect onthebuildinginpractice.
3. Monitoringdatacan givefeedback to thedesign-
ers about theconstruction so that they can adjust
designparametersintimetoinstruct theconstruc-
tionandensurethesafetyof theexcavationandthe
surroundings.
REFERENCES
Gao,Y.B., Wu, X.F. &Ye, G.B. 2003. StudyontheInfluence
of DiaphragmWall Construction on adjacent Building
Settlement, Underground Space. 23(2):115118.
Liu, G.B. & Lu, H.X. 2004. Study on the Influence
of Building Settlement Diaphragm Wall Trench Con-
struction, Chinese Journal of Geotechnical Engineering.
26(2):287289.
Liu, J.H. &Hou, X.Y. 1997. Manual of Excavationengineer-
ing, ChineseConstructionIndustryPress, Beijing.
Tang, H.F. & Li, B. 2004. Analysis of measured environ-
mental influenceandprotectionof speciallybiganddeep
excavation engineering in soft soil areas, Rock and Soil
Mechanics. 25(s1):553558.
Xiong, J.H. & Lou, X.M. 2004. Monitoring and Analy-
sis of displacement of a Foundation Pit in Soft Soil,
Underground Space. 24(3):354358.
Zhang, L 2004. Analysis of the Settlement of Buildings
Caused by the Construction of Metro Pit under Spe-
cial Geological Condition, Urban Railway Traffic Study.
23(5):3841.
Zhu, R.J. , Gao, Q & Qi, G. 2006. Settlement Analysis/n
Building near Reining Pile of Deep Foundation Pit,
Journal of Chongqing Jianzhu University. 28(2):5255.
417
Performanceof adeepexcavationinsoft clay
G.B. Liu& J. J iang
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R. China.
C.W.W. Ng
Department of Civil Engineering, the Hong Kong University of Science and Technology, HKSAR
ABSTRACT: Theobservedresponseof diaphragmwall andthesurfacesettlement of adeepexcavationfor a
metrostationinShanghai softclayarepresented, comparedwiththesimilarexcavations(retainedbyadiaphragm
wall) insoftclayareasinAsiaandfivemetrostationexcavations(similar depth) inShanghai. Resultsshowthat
themaximumlateral wall deflection(
hm
) andgroundsettlement (
vm
) are0.32%and0.1%of final excavation
depth(H
e
), respectively. Lateral wall deflectionsarenear theaveragemagnitudevalueinShanghai, smaller than
thoseinSingapore.Thelargestlateral wall deflectionisneartheexcavationcenterandtheratioof
hmcor
/
hmcen
of
themaximumwall deflectionis0.390.74. Thegroundsettlementisrelativelysmall andfallsintheZoneI limit
describedbyPeck(1969). Theadoptionof prestressedsteel struts, short excavationsections, fast workmanship
sequences andcompactiongroutingmay contributetothedeformationcharacteristics of deepexcavations for
metrostationsinShanghai.
Excavations for metro stations are in general deep,
longandnarrow,andlocatedinurbanenvironments.In
Shanghai, China, somemetrolineshavebeenbuiltand
several linesareunderconstructionorplanned,tocross
the large and congested city. Thick soft-to-medium
alluvial and marine deposit strata exist in this area.
Under suchcircumstances, movement control, rather
thanstability at thedeepexcavationsitebecomesthe
governing factor for design and construction (Liu &
Hou1997).
Research shows that field monitoring and perfor-
mance analysis are essential to deep excavations in
soft soils. Monitoring the field performance of the
excavation and the magnitude of movements in the
surroundingsoil (Peck1969; ORourke1976; Clough
et al.1989) along with estimating theeffects of such
movements andfieldperformanceonadjacent struc-
tures (Burland& Wroth1974, Boscardin& Cording
1979) areimportanttodesignengineers.Theobserved
performanceof deepexcavationsandcasehistoriesare
very useful to verify design assumptions and reduce
theconstructionriskintheexcavationprocess(Whittle
et al. 1993; Ng1998, Ou1998, Long2001, Finno &
Bryson2005, 2007, Leung& Ng2007).
Theobservedresponsesof diaphragmwallsandthe
ground surfaceof adeep multistrutted metro station
excavation in Shanghai soft clay arepresented here.
Diaphragmwall deformationsandgroundmovements
arecomparedwithsomecasehistoriesinShanghai and
excavationsinsoftclayinotherlocationsinAsia.From
themeasureddatainthisstudy,somerecognitionof the
characteristicsof deepexcavationscanbeobtainedand
canofferreferencesforbackanalysisandcasestudies.
2 SITE CONDITION
2.1 Introduction
Shanghai islocatedinthefront fringeof theYangtze
River Delta in China. The Da Muqiao excavation is
situatedinthesouthwest of Shanghai andthereis an
interchangefor two lines, Metro 4andMetro 9. The
siteplanisshowninFigure1. Notethattheexcavation
widthisvariable.
Thesitecrosses two roads, DaMuqiao Road and
LingLingRoad.TheshadowedareainFigure1shows
the interchange excavation (35m25m20.7m),
in the shape of cross. The excavation depth is
3.8mdeeper thantheexcavationdepthinother parts
(16.9m). There are several concrete, three-story to
six-storybuildingsalongthesidesof theexcavation.
Figure2showsacrosssectionof section1-1of the
interchangepart (INP); includingthesupport system,
419
Figure1. Siteplanof themetrostationandexcavationdirection.
Figure2. Crosssectionof theINP (Section1-1).
the strut level and excavation depth. The concrete
diaphragmwall for the INP is 40mdeep and 0.8m
thick; the16.9mdeepexcavationisretainedbya34m
deep and 0.8mthick perimeter concrete diaphragm
wall. Thegroundlevel of thesiteis3.75mabovethe
Shanghai City Datum(SCD). Thereis 3meter com-
pactiongroutingbelowthefinal excavationtoincrease
thatsoil strengthinthispart. Sixsteel strutswereused
at theINP invertical planewhiletherewerefivesteel
strutlevelsinotherexcavationparts.All strutsaresteel
tubesandthefirstonehasa580mmdiameterandother
ones are 609mmin diameter and are 16mmthick.
Twosteeltubes(12.45m) wereinter-joinedtoformthe
struts (24.6m). Thestruts havea3mlateral spacing.
The specific installment of struts was described by
Wang(2005).
2.2 Geology
Thestrataof Shanghai arethick soft soilscomprising
Quaternaryalluvial andmarinedeposits.A highwater
Figure3. A typical soil profileandsoil parameters.
content, lowshear strength, highcompressibility and
lowgroundbearingcapacityaretypical characteristics
of thesoft soil inShanghai.
Thegroundisunderlainbyrelativesofttomedium
soft marinedeposits, whichgenerally consist of uni-
formbedding planes. A typical soil profilefromup
to down obtained during site investigation is shown
in Figure 3. The top layer is artificial fill, 1.6min
thickness.About45mthickclayunderliesthefill.The
insideof theexcavation is among theclay, which is
aquaternary alluvial andmarinedeposit. Ingeneral,
groundwaterconditionsareapproximatelyhydrostatic
1.0mbelowgroundlevel.
3 CONSTRUCTIONPROCEDURE OF THE
EXCAVATION
Frominstalling the diaphragmwall to the comple-
tion of casing the bottomslab, the working period
is divided into eight main construction stages, listed
inTable1. Theexcavationdepthof theINP is 3.8m
deeper thanother parts (seeFigure2) sothereis one
moreexcavationstageindicatedbyitalicsTable1.
420
Table1. Summaryof themainconstruction.
Stage Constructionoperation Day
(1) (2) (3)
1 Constructionof diaphragmwall 81
2 Constructionof boredpile 116
3 Reducelevel digtoL1& installation 122
of L1strut
4 ExcavationtoL2& installationof L2strut 129
8 ExcavationtoL6& castingbottomslab 160
9 ExcavationtoL7& castingthebottomslab 168
The excavation of this long metro station was
dividedintoseveral shortsectionsinthehorizonplane.
Notethat theexcavationwasfromthetwoshort sides
tothemiddlepart.TheINPandtheeastpartweresyn-
chronously excavated, indicatedasPart 1. Part 2was
at the west end and was excavated one month after
Part 1was started. After finishingthesetwo excava-
tions, Part 3 was excavated fromboth ends. A short
sectionof soil excavationwasadoptedinPart 3also.
Eachexcavationdepthwasdugto0.3mbelowthestrut
level tooffer spacefor installingthestruts.
4 INSTRUMENTATION
Inclinometer tubes werefixed to thesteel reinforce-
ment cages andconcretedintheretainingwalls. The
casingwasinstalledwithapair of groovesorientedin
theexpecteddirectionof movement.Toavoidtwisting
of thecasing, thegrooveswerecheckedbeforeinstall-
menttoensurethattheir twistanglewasnomorethan
0.1%.Thecasingwasfixedtightlytothesteel cagesof
theretainingwall. Thegroovedirectionwas checked
againafter loweringthecagesintothebentonite. The
twist angles of the grooves were measured after the
concreting. The allowable value of each twist angle
was0.2%. Probeswitharesolutionof 0.02mm/50mm
and temperature of 20
to +50
were used in the

inclinationmonitoring.
Ground surface settlement monitoring markers
werelocatedonthesidesof theexcavation.Themark-
erswerelocatedabout1.535mperpendicularlyaway
fromwall at each rectangular box station and were
secured0.5mbelowthegroundsurface. Thesection
settlementsweremonitoredwithalevelinginstrument
withastatedaccuracyof 0.5mm/km.
5 OBSERVEDWALL DEFLECTION
5.1 Lateral diaphragm wall deflection
Along the south side of this deep and long excava-
tion, seventeeninclinometerswereplacedtomonitor
0
10
20
30
40
0 20 40 60
Lateral wall deflection (mm)
E
x
c
a
v
a
t
i
o
n

d
e
p
t
h

(
m
)
I1 I2
I3 I4
I5 I6
I7 I8
I9
Figure 4. Maximum lateral wall deflection at the final
excavationdepth.
the lateral wall deflection. Only nine of these incli-
nometers arepictured, as I1toI9, inFigure1, repre-
sentingthreetypical locationsof thediaphragmwalls.
(1)InclinometersI1,I2,I5,I6andI7wereatornearthe
corners; (2) Inclinometers I3andI4wereat thecen-
tersof theexcavation; (3) InclinometersI8andI9were
setfortheINP.Thesenineinclinometersmeasuredthe
datathat areanalyzedherefor lateral wall deflection.
Figure4shows thelateral wall deflection at I1to
I9 at the end of the final excavation (Stage 8). The
upper final excavationdepthisthatof thegeneral part
while the lower one is that of the INP. As indicated
in this figure, the profiles of lateral wall deflection
at the inclinometers are similar but with different
deformationmagnitudes.
Thelateral wall deflectionof type(1)issmallerthan
type(2).Thisfindingshowsthatwall deflectionswere
affectedbythestiffnessof thecorners. Themaximum
wall deflectionof thisdeepexcavationwasatthecenter
of excavationI3, withavalueof 54.5mm; whilethe
minimumwall deflection was at I5, with a value of
20.2mm. I5is at Corner 3andtheexcavationwidth
onthewest side(seeFigure1) issignificantlyshorter
than that on east side. This varied excavation width
maycontributetothesmallest wall deflectionat I5.
The lateral wall deflection of type (3) (INP) is
betweentype(1) andtype(2). Themaximumvalues
forI8andI9were38.2mmand51.0mm, respectively.
Notethat thelocations of all lateral wall deflections
wereabovethefinal excavationlevel.
5.2 Relationship between maximum lateral wall
deflection and excavation depth
Lecturesonlateral retainingwall deflectioncausedby
deepexcavations inAsiansoft clay werereportedby
someresearchers (Ou et al. 1993, 1998, Wong et al.
1996, Tamanoet al. 1996, Leeet al. 1998, Wanget al.
421
Table2. Summaryof thesixmetroexcavations.
H
e
Strut
hm
/
Name (m) No. K
s
FOS
base
H
e
(%)
Pudian 16.3 4 626.4 2.58 0.42
Yangshupu 16.5 4 626.4 2.59 0.57
Pudongdao 16.5 4 626.4 2.61 0.15
Luban 18.2 5 777.6 2.35 0.40
Pudong 17.3 5 1101.6 2.53 0.28
Damuqiao 16.9 5 1414.3 2.43 0.32
0
3
6
9
12
15
18
0 10 20 30 40
Excavation depth (m)
M
a
x
i
m
u
m

w
a
l
l

d
e
f
l
e
c
t
i
o
n

(
c
m
)
UOB, Singapore
Bugis, Singapore
Singapore Parking
CTE, Singapore
Newton, Singapore
Formosa, Taiwan
Taiwan Chi Ching
TNEC, Taipei
Osaka A, Japan
5 cases inShanghai
This study
hm
/H=0.3%
hm
/H=0.6%
hm
/H=0.9%
hm
/H=0.14%
Figure5. Relationshipof themaximumlateral wall deflec-
tiontotheexcavationdepth.
2005, Wallace, 1992, Hulme et al. 1989,) Figure 5
showsthemeasuredlateral wall deflectioncompared
totheexcavationdepthof thisdeepexcavation. Data
of fiveother deepexcavations inShanghai (listedin
Table2.) andseveral excavationsinTaiwan, J apanand
Singaporesoft clay areplotted in Figure5 for com-
parison. Notethatall theexcavationswereinsoftclay
andtheretainingwallswereall diaphragmwalls.
Figure5showsthatthemagnitudeof themaximum
wall deflection in Shanghai soft clay is smaller than
that in Singaporesoft clay whileit is similar to that
in Taiwan and J apan soft clay. The ratio of
hm
/H
e
inmetroexcavations inShanghai is between0.14%
0.6%, whilethatinSingaporeisbetween0.3%0.9%,
and that in Taipei and J apan is in 0.3%0.6%. The
ratio of
hm
/H
e
in this study is 0.32%. A compari-
sonof geotechnical parametersof thesoil inShanghai
clay andTaiwan clay shows that undrained strength
and water content are similar, but the average mea-
sured deflections in the Shanghai soil were smaller
than in the Taiwan soft soil. The use of the multi-
strutted support system, the prestressed steel struts,
theshort excavation sections and thefast workman-
shipsequencesinShanghai probablyaccountforthese
differences.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
10 100 1000 10000
K
s
h
m
/
H
e

(
%
)
This study
FOS
base
=2.0
FOS
base
=1.4
FOS
base
=3.0
(2.59)
(2.58)
(2.35)
(2.53)
(2.61)
(2.43)
Design curves (Clough et al. 1989)
Figure6. Normalizedlateral wall deflection(
hm
/H
e
) ver-
sussystemstiffness(K
s
).
5.3 Effect of the stiffness of the support system on
the maximum wall deflection
Theeffectof thestiffness(K
S
)of thesupportsystemon
lateral wall movement wasproposedby Cloughet al.
(1989). K
S
isexpressedasfollows:
where E
W
is theYoungs modulus of the wall; I is
the second moment of the area of the wall section,
I =t
3
,12, t isthewall thickness; h istheaveragever-
tical propspacingof themultistruttedsupport system
and
W
istheunit weight of water.
Figure6showsthenormalizedlateral wall deflec-
tionof thelongexcavationinthechartassuggestedby
Cloughet al. (1990). Another fivemetroexcavations
with the same retaining wall type and support sys-
teminShanghai withexcavationdepths from16.3m
to18.2m, arechosenfor comparison. A summary of
thesemetroexcavationsislistedinTable2.Theretain-
ingstructuresof all theseexcavationswere0.8mwidth
diaphragmwallsandthevalueof E
W
was30kN/mm
2
.
ThisformbyClough(1990) isfor excavationinsoft-
to-mediumclay to account for systemstiffness and
FOS
base
. The value of FOS
base
is calculated by the
definitionproposedbyCloughet al. (1989).
TheK
s
valuerangesfrom626.4to1414.3andthe
ratioof
hm
/H
e
variesfrom0.15to0.57withthevalue
of FOS
base
from2.35 to 2.61. However, for a given
K
s
, thereisarelativelyscatter in
hm
/H
e
values, there
may benovery strongcorrelationbetweenmeasured
hm
/H
e
andK
s
value. It canbeseenthat measuredlat-
eral wall deflectiondecreaseswithincreasingFOS
base
.
With similar system stiffness, the larger value of
FOS
base
thesmaller theratioof
hm
/H
e
. Andthemea-
sureddatado not seemto correspondvery well with
thedesigncurves.Thesuggestedchartonlycouldgive
422
0
10
20
30
40
50
60
0 110 220 330
Distance from east corner (m)
h
m

(
m
m
)
I1
I2
I8
I9
I3
I5
I6
I7
I8
I4
Figure7. Maximumlateral wall deflectionversus.
preliminarypredictionsof wall movementinsixmetro
excavationsinShanghai.
5.4 Three-dimensional deformation of excavation
Researches on three-dimensional responses of exca-
vations by measured data or finite analysis (Bono
et al.1992, Ou et al. 1993, Wong 1993, 1996, Lee
et al. 1998, Finno et al. 2007) showed that ratios of
corner
/
center
indeeper excavationsaresmaller ingen-
eral. Figure7 shows themeasured maximumlateral
wall deflection (
hm
) at the end of the final excava-
tion in this study. The lateral wall deflection at and
near thecenter of theexcavationislarger thanthewall
deflection near or at the corners, which means that
thestiffeningeffect of thecornersaffectedthedefor-
mationof thediaphragmwalls. Notethat thereduced
lateral wall deflectionat I5issignificant. Theratioof
eachmaximumwall deflectiontothemaximumwall
defectionat thecenter,
hmcor
/
hmcen
, of thewall was
0.390.74.
6 SURFACE SETTLEMENTS
In some cases, settlement was measured after the
startingof constructionof thediaphragmwall; some
weremeasuredat different times. Makingconsistent
comparisons is therefore difficult. In this study, the
settlement of thegroundsurfaceduringandafter the
final excavationwasmeasured.
Eight groupsof groundsettlementsweremeasured
alongoneof thelongsides(southside) andtwoends
of theexcavation(seeFigure1).All datawereselected
fromonesideof thesite. Figure8showstherelation
of the maximumsurface settlements to the distance
fromthediaphragmwallsatthefinal excavationdepth
of thisstudy.
In comparison with surface settlements in Singa-
poreandTaiwan soft clay, themeasured datain this
deep excavation fell into a relatively small range,
despite the weak ground conditions in Shanghai.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0 1 2 3
Normalized distance from wall to
excavation depth
N
o
r
m
a
l
i
z
e
d

s
e
t
t
l
e
m
e
n
t

t
o
e
x
c
a
v
a
t
i
o
n

d
e
p
t
h

(
%
)
TEN, Taipei
IMM Building, Singapore
Fareast, Taipei
Settlement in Shanghai
0
0.05
0.1
0 1.5 3
I
Peck (1969)
Figure 8. Distribution of the ground surface settlement
normalizedbytheexcavationdepth.
Wang (2005) summarized thesurfacesettlements of
some metro excavations in Shanghai and also indi-
catedthat themagnitudewasrelativesmall (
hm
/H
e
is
lessthan0.1%), plottedinFigure8also.
The3mthickcompactiongroutingbelowthefinal
excavation level, which improved the ground con-
dition, and the high-stiffness diaphragm wall and
prestressed multistrutted steel struts and fine work-
manshipmaycontributethesmall surfacesettlements.
7 DISCUSSIONOF THE MOVEMENT
MECHANISM
Comparedwiththelateral wall deflectionandground
settlement causedbyexcavationinsoft clayinSinga-
pore, J apan andTaiwan, the measured deformations
inthis deepexcavationfor ametro stationinShang-
hai is relatively smaller (see Figure 7). The 3m
thickcompactiongroutingbelowthefinal excavation
level, usageof prestressedmulti struts andastiffing
diaphragmwall, theshortsoil excavationsectionsand
quickandfineconstructionaremainfactorsassociated
withthis measuredresult. Amongtheabovereasons,
onethatoughttobenotedisthatthevertical spacingof
thestrutsintheShanghai excavationwassmaller than
thatinSingaporeandTaiwan, whichcontributedtothe
small wall deflection. Generallyin12mto18mdeep
Shanghai metroexcavations, four tofivestrutsareset
tosupportthesystem, whileinthe9.9mdeepexcava-
tionof CTEinSingapore, forexample, onlytwostruts
wereset inthevertical plane. Thesystemstiffnessin
theShanghai excavationwaslarger; itisnotsurprising
toobservethesmaller deformationinShanghai.
8 CONCLUSIONS
Observed data froma deep and long metro excava-
tion, with an interchangemetro station (3525m
2
)
423
crossing it, in Shanghai soft clay is presented. On
the basis of the interpreted observed field data, the
followingconclusionsaredrawn:
1. Thewall deflectionsalongthelongsideandthree
cornersaredifferent, butwithsimilar profiles. The
ratioof
hm
/H
e
inthisdeepexcavationislessthan
0.32%. The measured maximumwall deflection
is near the center. Three-dimensional analysis of
thisdeepexcavationshowsthat thecorner effect is
found. The
hmcor
/
hmcen
of thewall was0.390.74,
thoughtheaspect ratiowaslarge.
2. Comparedwiththeexcavationsinsoft clayinTai-
wan, J apan and Singaporeretained by diaphragm
walls and another five similar metro excavations
inShanghai, thismaximumlateral wall deflection
fallsneartheaveragemagnitudeforShanghai andis
at thelowlimit linefor Singapore. Theundrained
strength and water content of Shanghai clay and
Taiwan clay is similar, but the average measured
deflectionsinShanghai aresmaller thaninTaiwan.
Theuseof prestressed steel struts, asupport sys-
tem, shortexcavationsections, afastworkmanship
sequence, andcompactiongroutingbelowthefinal
excavationlevel inShanghai maycontributetothis
smaller deflection.
3. The influence of support stiffness on wall move-
mentwasanalyzed.Asmallerlateral wall deflection
of excavationwithlargeKs values was observed.
Though scatter exists in the chart proposed by
Cloughet al. (1990), thesuggestedchart seemsto
allow preliminary predictions of wall movements
inexcavations for metro stations inShanghai soft
soil.
4. Thesettlement alongthecornersandthelongside
isrelativelysmaller (
hm
/H
e
islessthan0.1%) than
that in other Asian soft clay, though thesoft clay
stratais thick inthis deepexcavation. This settle-
ment is locatedinthezoneI as proposedby Peck
(1969), whichis consistent withthecasestudy in
Shanghai (Wanget al. 2005).
ACKNOWLEDGMENTS
Theauthorswouldliketothank colleagueswhocon-
tributed to the field monitoring in Shanghai and
acknowledge the earmarked research grant 618006
provided by the Research Grants Council of the
HKSAR.
REFERENCES
Bono, N.A., Liu, T.K. & Soydemir, C. 1992. Performanceof
aninternallybracedslurry-diaphragmwall forexcavation
support. Slurry walls: Design, construction, and quality
control, ASTM STP 1129, D.B.Paul, R.G.Davidson, and
N. J. Cavalli, eds., ASTM, Philadelphia. pp.169190.
Boscardin, M.D. & Cording, E.J. 1979. Case studies of
buildingbehaviorinresponsetoadjacentexcavation. Uni-
versityof IllinoisRep. for theU.S. Department of Trans-
portation, Rep. No. UMTAIL-06-0043-78-2, Washington
D. C.
Burland, J.B. & Worth, C.P. 1974. Settlement of buildings
and associated damage. Proceeding of conference on
settlement of structures, Cambridge. pp. 611654.
Clough, G.W., Smith, E.M. & Sweeney, B.P. 1989. Move-
ment control of excavation support systems by iterative
design. Inproceedingof current principlesandPractices
onfoundationandengineering, ASCE, NewYork. Vol. 2,
pp.869884.
Clough, G.W. &ORourke, T.D. 1990. Constructioninduced
movement of in-situ walls. In proceeding of design and
performanceof earthretainingstructures. ASCE special
conference, Ithaca, N.Y. pp. 439470.
Finno, R.J., Bryson, L.S. & Calvello, M. 2005. Performance
of astiff support systeminsoft clay. J ournal of Geotech-
nical andGeoenvironmental Engineering, ASCE, 128(8).
pp. 660671.
Finno, R.J., Blackbum, J.T. & Roboski, J.F. 2007. Three-
dimensional effectsforsupportedexcavationinclay. J our-
nal of Geotechnical and Geoenviromental Engineering,
ASCE, 133(1). pp. 3036.
Hulme, T.W., Potter, J. & Shirlaw, N. 1989. SingaporeMRT
system: Construction. Proc., Instn. of Civ. Engrs.,Vol. 86,
London, 709770.
Lee, F.H.,Yong, K.Y., Quan, K.C. &Chee, K.T. 1998. Effect
of corners in strutted excavations: field monitoring and
case histories. J ournal of Geotechnical and Geoenviro-
mental Engineering, ASCE, 124(4). pp. 339349.
Leung, E.H.Y. & Ng, C.W.W. 2007. Wall andgroundmove-
ments associated with deep excavations supported by
cast in situ wall in mixed ground conditions. J ournal of
Geotechnical andGeoenvironmental Engineering,ASCE,
133(2). pp. 129143.
Liu, J.H. & Hou, X.Y. 1997. Excavation engineering hand
book. ChineseConstructionIndustry Press, Beijing, P.R.
China.
Long, M. 2001. Database for retaining wall and ground
movements dueto deepexcavation. J ournal of Geotech-
nical and Geoenviromental Engineering, ASCE, 127(3).
pp. 203224.
Ng, C.W.W. 1998. Observedperformanceof multi-propped
excavation in stiff clay. J ournal of Geotechnical and
Geoenviromental Engineering, ASCE, 124(9). pp. 889
905.
ORourke, T.D., Cording, E.J. & Boscardin, M.D. 1976. The
groundmovementsrelatedtobracedexcavationandtheir
influenceonadjacentstructures, Univ. of IllinoisRep. for
theU.S. Dep. of Transportation, Rep. No. DOT-TST-76T-
22, WashingtonD.C.
Ou, C.Y., Liao, Hsied. P.G. & Chiou, D.C. 1993. Charac-
teristics of groundsurfacesettlement duringexcavation.
CanadianGeotechnical J ournal, 30(5). pp. 758767.
Ou, C.Y. Liao, Hsied. & Lin, H.D. 1998. Performance of
diaphragmwall constructedusingtop-downmethod.J our-
nal of Geotechnical and Geoenviromental Engineering,
ASCE, 124(9). pp. 798808.
ground. Inproceedingof the7thinternational conference
on soil mechanics and foundation engineering, Mexico
City. Vol. 1, pp. 225281.
424
Tamano,T.,Fukui,S.,Mizutani,S.,Tsuboi,H.&Hisatake,M.
1996. Earthandwater pressuresactingonabracedexca-
vation in soft ground. Proc., Int. Symp. GeoAspects of
Underground Constr. in Soft Ground, City University,
London, 207212.
Wallace, J.C., Ho, C.E. & Long, M.M. 1992. Retaining
wall behavior for adeep basement in Singaporemarine
clay. Proc., Int. Conf. Retaining Struct., ThomasTelford,
London, 195204.
Wang, Z.W., Ng, C.W.W. & Liu, G.B. 2005. Characteris-
tics of wall deflections and ground surface settlements
in Shanghai. Canadian Geotechnical J ournal, 42(10).
pp. 12431254.
Whittle,A.J., Hashaha,Y.M. &Whitman, R.V. 1993.Analysis
of deep excavation in Boston. J ournal of Geotechnical
Engineering, ASCE, 119(1). pp. 6990.
Wong, L.W. & Patron, B.C. 1993. Settlements induced by
deepexcavationsinTaipei. 11thSoutheastAsianGeotech-
nical Conf., Singapore. pp. 787791.
Wong, L.W., Poh, T.Y. & Chuash, H.L. 1996. Analysis of
case histories fromconstruction of the central express-
wayinSingapore. CanadianGeotechnical J ournal, 33(1).
pp. 732746.
SYMBOLS
E
W
=Youngsmodulusof thewall;
FOS
base
=factor of safetyagainst thebasal heave;
h =average vertical prop spacing of multi-strutted
support system;
H
e
=final excavationdepth;
I =secondmoment of inertiaof thewall section;
INP=interchangepart excavation;
Ks =systemstiffness;
t =thewall thickness;
W
=unit weight of water;
corner
=deformationat corner of excavation
center
=deformationat center of excavation
hm
=magnitudeof maximumhorizontal diaphragm
wall deflection;
hmcor
=maximumlateral wall deflectionat corner;
hmcen
=maximumlateral wall deflectionat center;
vm
=maximumgroundsurfacesettlement;
425
Deformationmonitoringduringconstructionof subwaytunnelsin
soft ground
S.T. Liu
Department of Surveying and Geo-Informatics, School of Civil Engineering, Tongji University,
Department of Civil Engineering, Henan Institute of Engineering, Zhengzhou, P.R. China
Z.W. Wang
Fourth Team of Henan Province Coal Field Geology Bureau, Xinzheng, P.R. China
ABSTRACT: Monitoringthedeformationsof subwaytunnelsinsoftgroundisaprincipal meansfor selecting
theappropriateexcavationandsupportmethodsinthedesign.Inthispaper,brieflydescribesthebasicdeformation
monitoringrequirements,analyzesthetunnel deformationmonitoringaccuracydifferencebetweenmountainarea
andurbanarea, discussesthemeasurementsof deformationsat groundsurface, measurementsof deformations
onthegroundandmeasurementsinthetunnel, especiallygivesanexampleof vertical convergencemonitoring,
introducesthedataprocessingmethodandthequality evaluatingmethodbasedonthemeasureddata. Mainly
explores the technique of tunnels crown settlement monitoring, convergence monitoring and theAutomatic
DeformationMonitoringmethods. It isput forwardthat themonitoringjobduringtheconstructionof subway
tunnelscanbedonewithbothtraditional andmodernsurveyinginstrumentsandthejobwill besignificant to
thesamekindsof projects.
1 INSTRUCTIONS
Determining shape and position changes occurred
in engineering is one of an application area of
geodeticsurveys.Temporaryandpermanentdeforma-
tions occur in engineering structures such as dams,
bridges, tunnels, viaducts and towers due to natural
and artificial forces. Causes of these deformations
are usually physical properties of ground, weight of
structure, active external forces etc. (Clough 1960).
Theneedto upgradeandfurther developtransporta-
tioninfrastructurehas leadtotheon-goingconstruc-
tion of large-diameter, long tunnels under difficult
conditions.
Suchconditions usually arisefromacombination
of adversegroundandgroundwaterregimes, veryhigh
overburdenpressuresor, inthecaseof urbantunnels,
the existence of sensitive structures within the zone
of influenceof thetunnel. It is imperativeto provide
accurateandfrequent monitoringof tunnel liningsto
detectanymovementsthatcouldposeasafetyhazard.
Deformation monitoring in tunneling usually
includes some of the following measurements
(Dunnicliff 1993):
1. Convergenceof thetunnel wall, andusually crest
settlement andspring-lineclosure.
2. Deformations at ground surface including settle-
mentsandtiltsof surfacestructures.
3. Deformationsinthegroundaroundthetunnel.
Tunnel deformationscanbemonitoredwithgeode-
tic or geotechnical methods. Geodetic measurements
provideabsolutecoordinatesof thetarget locationsin
time, while geotechnical measurements usually pro-
viderelativedisplacementsof thetargetlocationswith
respect to an initial condition (at the time when the
initial measurement isrecorded) (Hisatake1999).
Geotechnical measurements can provide absolute
coordinates of the target locations in time, if the
initial positions of the targets are obtained using
geodetic means. Depending on the tunneling appli-
cation, deformation measurements can be recorded,
processed and evaluated in real-time using digital
recording and telecommunication systems, or can
be recorded manually and processed later in batch
mode.
Real-timeprocessingof grounddeformationsoffers
the possibility of rapid response to upcoming sit-
uations but requires advanced technology and an
appreciably higher cost. Thus, real-time monitoring
islimitedtocaseswhererapidresponseisabsolutely
necessary, i.e., mainly inurbantunnelsnear sensitive
structures.
427
In this paper, briefly describes the deformation
monitoringrequirements,andthendiscussesthedefor-
mation monitoring at ground surface, in the ground
and in the tunnel separately, furthermore gives an
exampleof traditional vertical convergencemonitor-
ing, atlastmainlyintroducestheAutomaticDeforma-
tionMonitoringmethod.
2 BASIC REQUIREMENTS
Thebasicrequirementsare: layoutmonitoringbench-
marks, monitoringtargets timely, takemeasurements
regularly, analyzesthesurveydataandfeedbackinfor-
mation to relative departments in time (Kavvadas
1999).
Measurementaccuracydependsonthepurposesof
thedeformationobservation. Inorder toensuremoni-
toringaccuracy, Operatingpersonnel mustbefamiliar
tothepurposesandequipmentoperatingrules, survey
teammembersshouldcorporateandworkcarefully.
Eachmonitoringstepshouldabidethefour fixed
principle(Liu2006). Theso-calledfour fixed prin-
ciplesnamely: instrumentsandrodsfixed, observing
staff fixed, monitoringmethodfixedandmonitoring
environment fixed.
Thefirststepindeformationmonitoringistocarry
out a basic survey, consisting of individual surveys
insideanarrowtimeframe.All followingsurveyswill
becomparedto theresults of this survey. It is there-
foreimportantthatthissurveyiscarriedoutbeforethe
deformationstobeobservedcanbeexpectedtooccur
anditisalsoimportantthatthebasicsurveyiscarried
out withahighdegreeof reliabilityregardingresults,
asit isobviouslyimpossibletofollowupwithcheck-
surveysafter deformationshavetakenplaceor canbe
expectedtohavetakenplace. It isrecommendedthat
thefirst elevationresult shouldbefiguredout inthe
averageof atleasttwotimesof monitoring(Liu2007).
Theequipment beingusedshouldbetestedbefore
the first time monitoring, after continuously use for
36 months the equipments should be tested again
(Kaiser 1993).
Benchmarks and stations and Monitoring points
insidethetunnels should beset in areawhererocks
arestable. Benchmarksandstationsinsidethetunnel
areinstalledinspecial brackets, whichforcedcentered
(Kontogianni & Stiros2002, 2003).
3 MONITORINGACCURACY DIFFERENCE
BETWEENMOUNTAINAREA AND
URBANAREA
3.1 Objectives difference
Theobjectivesof grounddeformationmonitoringare
different inmountainandurbantunnels. Inmountain
tunnels, themain objectiveof deformation measure-
ments during construction is to ensure that ground
pressuresareadequatelycontrolled, i.e., thereexistsan
adequatemarginof safetyagainst collapse, including
roof collapse, bottomheave, failureof theexcavation
face, yieldingof thesupport system, etc. (Mihalis &
Kavvadas1999).
Adequatecontrol of groundpressuresensuresasafe
andeconomical structure, well adaptedtotheinherent
heterogeneityof groundconditions. Thisprocedureis
compatiblewithmoderntunnel designmethodswhich
includearangeof excavationandsupport systemsto
cover theanticipatedspectrumof conditionsalongthe
tunnel, withselectionof theapplicablesystemineach
case relying on the encountered geology at the tun-
nel face, experienceontunnel behavior at previously
excavated sections under similar conditions and, on
accuratedeformationmeasurements, i.e., byapplying
theso-calledobservational method.
This method of construction can ensure adequate
safety and, at the same time, an economical con-
struction. Onthecontrary, inurbantunnels, themain
objectiveof grounddeformationmonitoringistolimit
ground displacements to values sufficiently low to
prevent damage to structures and utilities at ground
surface. Thus, thefundamental differenceindeforma-
tionmonitoringstems fromthefact that inmountain
tunnels theobjectiveis to guard against an ultimate
limit state(i.e., collapse) whilein urban tunnels the
objectiveistoguardagainst serviceabilitylimit states
(i.e., crack initiation) for structures and utilities at
groundsurface.
3.2 Monitoring accuracy difference
As aresult of thesedifferences in objectives, design
philosophies, and construction techniques, thetypes
andrequiredaccuracyof themeasuredgrounddefor-
mations vary between thetwo classes of tunnels, as
follows.
1. In mountain tunnels, considerable ground defor-
mationsaredeliberately permitted(andoftenpro-
voked) in order to reduce the initially very large
geostatic loads on the temporary support by
increasing ground de-confinement. Such reduc-
tion of ground loads on the tunnel support can
beappreciableand, thus, extremelybeneficial pro-
videdthat excessiveloosening of therock mass
isprevented(suchloosening cancauseroof fail-
uresandaneventual increaseof thegroundloads).
De-confinement is achievedby controlledinward
ground deformation at the excavation face (face-
take), controlled delay in the completion of the
temporarysupportmeasures(byincreasingthedis-
tance from the face where the tunnel invert is
closed), arelativelyflexibletemporarysupportsys-
tem(e.g. longpassiverock-bolts andthinsprayed
428
concreteliners) and, finally, by installingtheper-
manent lining at a later time when evolution of
the long-term (creep) ground deformations has
practically stopped. In extreme cases of strongly
squeezinggroundconditions, slidingsupportsmay
beinstalledtopermit tunnel wall convergencesof
several tensof centimeters. Inall thesecases, con-
trol of ground deformations depends strongly on
efficient and timely deformation measurements.
However, due to the large ground deformations
(several centimeters andevenseveral tens of cen-
timeters), therequired level of precision of these
measurements needs not be excessive; typically,
accuracyof theorderof onecentimeterissufficient
inmountaintunnel applications.
2. In urban tunnels, the main objective is limit-
ing ground deformations around the tunnel and
thuscausingtheminimumpossiblemovement and
disturbance at ground surface and the structures
founded there. This is achieved by (a) limiting
inwardgrounddeformationat theexcavationface
(face-take), e.g. by face pre-reinforcement using
fiber-glass nails, stiff steel beams (fore-poles),
cement- orjet-groutingtechniques,(b)byinstalling
a stiff temporary lining, usually including invert
closure, as early as possibleand (c) by installing
the final lining as quickly as possible, especially
whentunnel wall convergencescontinuetoevolve
withtime. Theabovestiff constructionmethods
tendtoreducegroundde-confinementandthusthe
groundloadsonthetunnel liningareasignificant
fraction of the initial geostatic loads are much
smallerthanthoseindeepmountaintunnels. Dueto
thesmall grounddeformationsinducedbytunnel-
ing(usuallylessthan10mmatgroundsurfaceand
occasionally less than 5mm), measurement pre-
cision and theearly installation of themeasuring
devicesisof utmost importance.
4 DEFORMATIONMONITORINGAT
GROUNDSURFACE
Measurementsof deformationsat groundsurfaceare
crucial in urban tunneling projects wheredamageto
surface structures and utilities should be prevented.
These measurements typically include settlements
(andheaves) of structuresaswell astilting. Suchmea-
surements areperformedwithsurveyinginstruments
(PreciseGeodeticLevel, total stationsandGNSS), or
withgeotechnical instruments likeElectronic Liquid
Level Gauges,ElectrolyticTiltSensors(electro-levels)
surface clinometers/tilt meters, precise taping, and
crack-meters.
Preciselevelingandfaademonitoringarethemost
common methods for monitoring displacements at
groundsurface. Theaccuracy of thesemeasurements
istypically0.2mm(overabout100mlengths) forpre-
cise leveling and 1
for angles and (1mm+2mm

10
6
D) for distances in the case of faade moni-
toring with total stations. Faade monitoring can be
automated and measurements can be obtained and
transmittedinpracticallyreal-time.
Insidebuildingsandinareaswithlimitedvisibility
for the application of the above geodetic measure-
ments, geotechnical measurementscanbeperformed
using the following precision instruments (Moaveni
2003):
1. Electronic Liquid Level gauges, for the mea-
surement of settlements at several locations. The
method consists of installing a number of liquid
filledpots, hydraulically connectedto areference
pot located in a stable area. The elevation of the
liquid in thereferencepot is maintained constant
by means of amini-pump, reservoir and an over-
flowunit. LVDT float sensors monitor theheight
of theliquidineachpot. Whensettlementor heave
occurs, thesensor detects theapparent changein
theheightof theliquidandtransmitsthesignal toa
dataloggerforcontinuousmonitoringandreal-time
processing. Theaccuracyof thesystemis0.3mm.
2. Electrolytictilt sensorsareprecisionbubblelevels
that areelectrically sensed as aresistancebridge.
The bridge circuit outputs a voltage that is pro-
portional to the tilt of the sensor. The sensors
areusually attachedonmetal beams, oneto three
meterslong, withtheir endsmountedonthestruc-
tural elementstobemonitored. Chainsof suchtilt
sensorsareofteninstalledinsequenceinthehori-
zontal directiontomonitor differential settlements
along long walls or beams. The precision of the
instrument istypically1
.
3. Surface clinometers (tilt meters), precise taping
using invar tapebetween fixed anchor points and
various types of crack-meters are also used in
measuringdeformationsat groundsurface.
5 DEFORMATIONMONITORINGIN
THE GROUND
Thesemeasurementsrecordthedeformationof target
positions in the ground, either around the tunnel or
deepbelowthegroundsurface. Theyareoftenusedto
calculatestrainsor control thedeformationof charac-
teristicpointsintheground(e.g. belowthefoundation
of buildings, belowutilitylines, etc.) (Sakurai 1981).
Such measurements are performed with geotech-
nical instruments including single- or multi-point
borehole rod extensometers, magnetic extensome-
ters, slidingmicrometers, inclinometers, probedeflect
meters (often called sliding curve meters) and deep
settlement plates. Theseinstruments canbeinstalled
429
either fromthegroundsurface(beforethetunnel face
reaches the area of the instrument) or from inside
the tunnel (radically fromthe tunnel wall or along
thetunnel axisaheadof theexcavationface).
6 DEFORMATIONMONITORINGIN
THETUNNEL
Tunnel wall convergence(closure) betweenreferences
points (hooks) bolted on the tunnel walls is usually
measured with standard metal tape extensometers.
For distances up to about 1015 meters, the accu-
racy of such measurements is typically 0.2mm. The
methodis easy touseandmaintainbut it only offers
the magnitude of the deformation along the line of
measurement.
Becauseof this disadvantage, inmost present-day
tunneling applications, deformations of the tunnel
walls are obtained in three dimensions, by routine
geodeticsurveyingusingtotal stationswithintegrated
distance measurement. In such applications, optical
reflectortargetsareinstalledatregulardistancesalong
thetunnel axis (e.g. at sections every 1520m) and,
on each section, at selected locations of the tunnel
wall (e.g. fivereflectorsper section: at thecrest, at 45
degreesandat thespring-line). Astunnelsareusually
long,thefixed(stable)referencepositionsaretypically
locatedoutsidethetunnel, oftenat distances exceed-
ingonekilometer andusuallyout of sight frominside
thetunnel.
Thus, measurementsof thetargetsinsidethetunnel
areobtainedbyplacingthetotal stationatpre-defined
ruggedstations1(boltedonthetunnel wall) andsuc-
cessivelymovingtheinstrument forward(towardsthe
tunnel excavation face) whilemeasuring thecoordi-
nates of the visible targets from each station. The
theoretical accuracyof thesemeasurementsoveradis-
tance of about 100m) is about 23mmfor lengths
and 2
for angles. For long tunnels, the accuracy

of the measurements is usually reduced by unclean
atmosphereand dueto themultiplepositions of the
instrument (especiallyif thesepositionsarenot stable
duetocreepdeformationsof thetunnel).
A recent development of measuring the geome-
try of tunnel walls in cross section (and thus assess
thedeformationintheinterval betweentwomeasure-
ment epochs) is theTunnel ProfileScanners (profile
meters). Inadditiontomeasuringtunnel wall conver-
gences intime, profilemeters arealso employedfor
avariety of other purposes likecomparing theactu-
ally excavated tunnel cross section with the design
requirementandformeasuringthevolumeof shotcrete
placed on theexcavated rock surface(by measuring
theprofilebeforeandaftershotcreting).Tunnel profile
metersarefullydigitizedphotogrammetricmeasuring
devices.
Figure1. Anexampleof settlementmonitoringresult(Villy
Kontogianni et al. 1999).
A typical suchsystemconsistsof twoCCDcameras
which are mounted on a portable frame. The cam-
erasproducestereoscopicdigital imagesof thetunnel
surface. Thepositionof thecameraframeisautomat-
ically determined by a total station with automatic
target recognition placed up to amaximumdistance
of 100meters.
For this purpose, three reflector targets are per-
manently mounted on the frame. Digital images are
automatically storedinalaptopcomputer andcanbe
processed to providethe3-D coordinates of thesur-
veyedtunnel wall surfacewithanaccuracyof 5mm
for eachcoordinate.
Althoughtheaccuracy of thismethodislowcom-
paredtoroutinegeodeticsurveying, theadvantageof
recordingaverylargenumber of pointsonthetunnel
wall outweighsthelowaccuracyinmanyapplications.
6.1 Traditional geodetic measurements
Threedimensional coordinatesof objectpoints(defor-
mationpoints) whichwill beconstructedonceiling,
sole, andsidewallsof cross- sectiondeterminedatdif-
ferentintervalsontunnelsweremeasuredbyelectronic
instruments. The coordinates of object points which
wereconstructedonceiling, sole, andsidewallswere
measured at determined period intervals. By using
coordinates measured at different periods, forming
movementsweredetermined. (Figure1isanexample
of settlement monitoringresult).
6.2 Automatic deformation monitoring
6.2.1 Automatic deformation monitoring
requirement
High-quality precision optical monitoring targets
mounted on support of excavation and tunnel lin-
ings, with fully automated motorized total stations
under computer control tomonitor remotelythethree
components of movement. Also requiredwas amea-
surement precisionof 1mmfor sight distances upto
100m, withwirelessdatalinkstothecontrol site. The
430
robotic total stations would have to be totally auto-
matic andoperateunattended24/7under all weather
conditions, andalsobeinsensitivetorefractioneffects
caused by temperature or pressure variations. Each
targetwouldhavetobehitatleastevery30minutes.
Multiply TCA2003s, located where they could
monitor hundredsof targetprismstobeaffixedatpre-
defined intervals to the structures to be monitored.
Themeasurement datawouldbetransmittedat speci-
fiedintervalsviaradiodatalinkstoacentral location,
situated miles from the actual instrument monitor-
ingstations. Thetotal network wouldbetiedtogether
and controlled by GeoMoS (Geodetic Monitoring
Software) system.
6.2.2 Installation challenges
ThereweretwoTCA2003 total station sites on each
side, for a total of four. The total stations were per-
manently mounted on fixed location with vented
heavy-dutyglassenclosurestoprotectthesystemfrom
theelements. Eachinstallationincludedanintercom
radioandmodemwithdirectional antennatotransmit
datafromthesiteto thecontroller. Bandpass filters
wereaddedtoovercomethehighlevelsof RF activity
intheairportenvironment.Thepedestalswereisolated
toeliminatevibrationandmovement.
The GeoMoS software was installed on the com-
puter network, andwasconfiguredsothatdatacanbe
accessedviaasecureIP link byauthorizedpersonnel
andtheresident engineer. Protectingtheremotesites
fromwindandweather presentedspecial challenges.
Theglassenclosureshadtoberobustandrugged,capa-
bleof withstandingheavy winds, rain, snowandice,
andtheyhadtobenon-reflectivesoasnottodistortthe
EDMsignal. Necessitybeingthemother of invention,
theyexperimentedwithanumber of differenttypesof
enclosuresincludingclear rounddogigloosuntil they
foundtheright solution.
6.2.3 Monitoring software
The GeoMoS software is a powerful tool for con-
trolling the network of the remote sites, as well as
collecting data, providing alarms, post-processing,
reportingandvisualizingdata.Thesoftwarerepresents
thedataandresultsingraphical or numerical format.
Youcanselectatime-linegraphshowingthetrendsof
movementover selectedtimeperiods(Burland2001).
Multiplepoints canbeviewedsimultaneously inthe
samegraph.Alternatively, youcanselectavectorview
that showsdisplacement for aselectedarea, toeasily
seewherethegreatest movement hasoccurred.
Senior-level projectengineersandother authorized
personnel canaccesstheGeoMoSdatathroughareal-
timeweb-based portal fromany PC or laptop. They
can log onto thesecureGeoMoS siteand download
systemstatusandreports. Measurementtolerancesare
establishedandloadedintotheGeoMoSsystem,If any
Figure2. Anexampleof tabularreportof AutomaticDefor-
mationMonitoringfor HongKongKSL Railway(Tangetal.
2007).
of thesetolerances areexceeded, anautomatic alarm
isactivated(Figure2isanexampleof tabular report).
7 CONCLUSION
Inorder todeterminedeformationsof thesubwaytun-
nel in soft ground, deformation monitoring must be
takenonthegroundsurface, inthegroundandinthe
tunnel. Practiceshowsthatthetraditional deformation
measurement can be achieved with 0.51mm dis-
placement accuracy, but deformation measurements
can also be recorded, processed and evaluated in
real-timebydigital recordingandtelecommunication
systems.Real-timeprocessingof grounddeformations
offers thepossibility of rapid responseto upcoming
situations but requires advanced technology and an
appreciably higher cost. So real-time monitoring is
limited to cases where rapid response is absolutely
necessary, i.e., mainly inurbantunnelsnear sensitive
structures.
REFERENCES
Burland, J. B. & Standing, J. R. 2001. Building Response to
Tunneling Case Studies from Construction of the Jubilee
Line Extension. London: ThomasTelfordpublishers.
Clough, R.W. 1960. The Finite Element Method in Plane
StressAnalysis, Proceedings of American Society of Civil
Engineers, 2nd Conference on Electronic Computations.
23: 345378.
Dunnicliff, J. 1993. Geotechnical Instrumentation for
Monitoring Field Performance. U.S.A: J ohn Wiley &
SonsInc.
Hisatake, M. 1999. DirectEstimationof Initial Stressesof the
GroundAroundaTunnel. Proceedings of the Numerical
Methods in Geomechanics: 373377.
Kaiser, P. 1993. Deformation Monitoring for Stability
Assessmentof UndergroundOpenings.Compressive Rock
Engineering: 607630.
Kavvadas, M. 1999. Experiencesfromtheconstructionof the
AthensMetroproject,Proceedings of 12th European Con-
ference of Soil Mechanics and Geotechnical Engineering:
16651676.
431
Kontogianni, V. & Stiros, S. 2002. Shallow Tunnel Con-
vergence Predictions and Observations. Engineering
Geology63(34):333345.
Kontogianni, V. & Stiros, S. 2003. Tunnel MonitorigDuring
theExcavationPhase:3-D KinematicAnalysis Basedon
Geodetic Data. Proceedings of 11th FIG Symposium on
Deformation Measurements. Santorini, Greece.
Kontogianni, V., Tesseris, D. & StirosS. 1999. Efficiencyof
geodeticdatatocontrol tunnel deformation. Proceedings
of The 9th FIG International Symposium on Deformation
Measurements. Olsztyn, Poland:206214.
Liu, S. 2006. Deformation Measurements During theCon-
struction of Large Dam Projects. Chinese Journal of
Underground Space and Engineering 06(Z2):13461348.
Liu, S. & Zhao, Z. 2007. DeformationMonitoringof 70m
SpanBoxGirdersof Hang-ZhouBaySea-CrossBridgeat
ConstructionStage. World Bridge 07(2): 5860.
Mihalis,I.&Kavvadas,M.1999.GroundMovementsCaused
by TBM Tunnelling in theAthens Metro Project. Proc.
Int. Symp. on the Geotechnical Aspects of Underground
Construction in Soft Ground, Tokyo, J apan, J une 1999:
269274.
Moaveni, S. 2003. Finite Element Analysis. New J ersey:
PearsonEducation.
Sakurai, S. 1981. Interpretation of Displacement Measure-
ments. Proceeding of the International Symposium on
Weak Rock, Tokyo: 751756.
Tang, E., Lui, V. & Wong, A. 2007. Application of Auto-
matic Deformation Monitoring Systemfor Hong Kong
KSL Railway. Monitoring Strategic Integration of Sur-
veying Services. Proceedings of FIGWorking Week 2007,
HongKongSAR, China.
432
Theconstructionandfieldmonitoringof adeepexcavationinsoft soils
T. Liu& G.B Liu
Tongji University, Shanghai , P.R. China
C.W.W. Ng
Department of Civil Engineering, the Hong Kong University of Science and Technology, HKSAR
ABSTRACT: Thispaper describestheconstructionof a40mmulti-proppeddeepexcavationinthedowntown
areaof Shanghai, Chinaandtheinterpretationof themonitoringdata.Theentireexcavationwas263minlength,
23mwideand3841mdeep. Theexcavationwasdividedinthreesections, i.e., theeastern, themiddle, andthe
westernpits. Noticeablecharacteristicof thisprojectaretheexcavationof anearly60mwidesectionembedded
in theHuangpu River and thepresenceof many sensitivebuildings nearby. In this paper, someconstruction
detailsof theexcavationandthemonitoringresultsarepresentedanddiscussed.
1 INTRODUCTION
1.1 Excavations for underground railway networks
in Shanghai
With the rapid development of Shanghai and the
citysundergroundrailwaynetworks, metroconstruc-
tionhasbecomeamultidisciplinaryandmultifaceted
project. Duringconstruction, regularserviceof neigh-
boringundergroundrailway lines must remainopen.
Figure1showsthedistributionof completedandon-
goingexcavations for undergroundrailway networks
inShanghai asof 2007. 92%of themetrostationsare
deeper than 15mand 31% are over 20min depth.
Zhao & Yang (2004) have suggested that any exca-
vation, which is over 20m in depth, will impose
severegeotechnical challengesinsoftsoils.Thispaper
describestheconstructionof a40mdeepexcavationin
thedowntownareaof Shanghai,Chinaandtheanalysis
of itsconstructionimpactstotheenvironment.
1.2 Engineering background
Thetunnel between theSouth Pudong Road Station
andtheNanpuBridgeStationisacross-riversectionon
Line4of Shanghaisundergroundrailwaysystem(see
Figure2). On1J uly2003, afailureoccurredresulting
infloodingof thetunnel fromasub-artesianaquifer.
Thisledtothecollapseof somesectionsof thetunnel,
excessiveground settlements and theclosurenearby
buildingsbecauseof thepotential dangertothepublic.
Figure 1. Statistics of depth of completed and on-going
excavationsfor undergroundrailwaynetworksinShanghai.
Anin-siturestorationprogramwasimplementedin
whichanopencut wasmadeinthedamagedsections
of thetunnel toexposethecollapsedsections, clearthe
debris, andthentoconstruct anewopenstructure.
Astheextent of thecollapsewaslarge, therestora-
tion work was constructed both on shore and in the
river. Using a cofferdam, a piled steel platformwas
constructedabovetheriver to excavatethecollapsed
tunnel after it had been backfilled. Two undamaged
tunnel sections, whichwereflooded, werede-watered
by high-pressure pumps and then reinforced. The
entirerestorationproject includedfivemajor compo-
nents: threedeep excavations at theeastern, middle,
and western parts of the tunnel, clearance work for
433
Figure2. Line4of Shanghaisundergroundrailway.
Figure3. General layout of therestorationproject.
theundamagedtunnel inPudong(1003m) andinPuxi
(760m),andtheconnectionworkatthetwoendsof the
damagedandundamagedtunnel sections.Thegeneral
layout of therestorationproject isshowninFigure3.
A 65mdeep1.2mthick concretediaphragmwall
wasusedtoretainthethreeexcavations(or so-called
pits). The total length of the excavations was 263m
in length and 23mwide (see Figure 3). The depth
of theexcavationsvariedfrom38mto41m. Tenand
ninelevels of concreteproppingslabs wereinstalled
to support thediaphragmwall at theeasternandthe
westernpits, respectively(seeFigure4). Inthemiddle
pit, 35mlong, 1.2min diameter cast-in-place piles
wereconstructedtosupport thesteel columns whose
cross-sectional areais 650650mm
2
. Inturn, these
columns wereused to carry thehorizontal propping
slabs.
2 GEOLOGICAL CONDITIONS
Therestorationproject waslocatedadjacent toOuter
River Road, Dongjiadu Road and Zhongshan South
Road, wheretheterrainwasquiteflat. Figure5com-
pares thecurrent geological profileobtained froma
shift put downafter thetunnel hadcollapsedandthe
previously surveyed data obtained before the tunnel
Figure4. Groundprofileat therestorationproject site.
Figure5. Comparisonof strataonsite.
collapse. The physical and mechanical properties of
thesoil layersaresummarizedinTable1.
Based on the ground investigations, groundwater
wasexpectedalongandabovethetunnel.Theground-
watertableis0.51.0mbelowthegroundsurface.The
groundcomprisesclayandsiltyclaylayersat shallow
depths and more sandy materials at greater depths.
Layer 7isthefirst aquifer inShanghai, whilelayer 9
thesecondaquifer. They areconnectedhydraulically
on site but apparently they are not connected to the
Huangpu River. The groundwater and the Huangpu
River havenoobvioushydrauliclinkonsite.
3 ENGINEERINGDIFFICULTIESAND
CONSTRUCTIONTECHNOLOGY
The scale of the restoration work was large and the
depth of excavations was thedeepest in Shanghai at
that time. Aroundtheexcavationsite, thereweresen-
sitive buildings nearby such as the Linjiang Garden
Building and theNanpu Bridgewhich required pro-
tections. Groundconditionswereverycomplexowing
to thetunnel collapseandthematerials left fromthe
434
Table1. Physical andmechanical propertiesof thesoilsat therestorationproject site.
Water Unit Average
Layer Soil Thickness Depthbelow content, weight, Void Cohesion, Friction SPT N
No. Type (m) ground(m) w(%) (kN/m
3
) ratio, e (kPa) angle(
) values
Fill 7.70 4.29 33.5 18.7 0.98 14 27.5
2
Silt clay 1.8 15.09 30.9 18.7 0.90 7 32.0 10
1&2
Grayclay 7.10 22.19 43.3 18.2 1.24 14 12.5
Greensiltyclay 3.5 25.69 24.4 20.2 0.70 43 15.5
1
Siltysand 12.30 37.99 35.1 20.3 1.04 0 33.0 40
2
Siltyfinesand 22.86 60.85 28.2 19.8 0.78 0 37.0 50
Siltyfinesand 25.2 0.71 0 35.5 50
rescue of the tunnel including slurry, polyurethane,
freezing agents, rails, and reinforced concretepipes.
Sincethedepth of excavations reached theconfined
water aquifer at layer 7 (see Figure 4), dewatering
wasthereforerequired. Asthepumpingrateof asin-
glewell fromdewateringtestswasfoundtobeabout
90m
3
/h, thelossof groundwater wasexpectedtohigh
duringpumping. Special measuresweretakentomin-
imize ground settlements and hence to protect the
surrounding buildings. Several important construc-
tionprocessesof thisrestorationprojectaredescribed
below.
3.1 Treatment of underground materials
Therestorationwork coveredthemainareasaffected
by the tunnel collapse. Within the collapsed areas,
there were materials frombuilding foundations and
abandonedundergroundpipelines, emergency rescue
materials, construction waste, sand backfill, previ-
ously abandonedbuildings(e.g. theWenmiaoPump-
ing Station), and the septic tanks of the Linjiang
Garden building. In the deep strata, there were two
refrigerationunits, tracks, sleepers, tunnel reinforced
steel supports, alargenumber of water pipesandthe
collapsedreinforcedconcretetunnel debris.
Materials at 1015m depth were excavated and
removed using a heavy plant. The materials further
below were removed fromexcavation pits. Cutting
equipmentwasusedtoremoveanyobstructions,which
were taken away with the earthworks. When deep
undergroundobstructionsaffectedtheconstructionof
thediaphragmwall,aboringmachinewasusedtoclear
theobstructions(seeFigures6and7). Thediaphragm
wall wasthenconstructedafter backfilling.
3.2 Construction of the diaphragm wall
Toretainthedeepexcavationpits, whichvariedfrom
38mto41m, arecorddepthof 65mand1.2mthick
concrete diaphragmwall in Shanghai was required.
Theconstructionof suchadeepwall wasexpectedto
encounter manydifficulties. Theavailableequipment
Figure6. Cuttingof theformer tunnel byboringmachine.
Figure7. Barrier removal at thepositionof thediaphragm
wall.
andtechnology hadto beimprovedfor theconstruc-
tion. As shown in Figure8, thetwo-drill-onegrab
method was used. The method involves drilling two
orientedholesfirst witharotarydrillingmachineand
then grabbing thesoil between theholes effectively.
Using this method, the verticality of the holes was
better than1/300andtheslurry trenchexcavationin
435
Figure8. Thetwo-drill-onegrabmethodfor theconstruc-
tionof thediaphragmwall.
Figure9. Brushingworkfor thediaphragmwall.
thestratawithobstructionscouldbecarriedoutmuch
faster.
To achievebetter efficiency incleaningthesoil at
thepreviouslyinstalleddiaphragmwall panel, acoun-
terweight was placedat thebottomof thediaphragm
wall trench (see Figure 9). This generated horizon-
tal forces that made the brushing machine cling to
thejoints by its directional bearings. Inaddition, the
method enhanced thewaterproofing performanceof
thewall.
3.3 Cofferdam and steel platform
Thecofferdamandthesteel platformintheHuangpu
Riverwereimportanttemporarystructures,whichcon-
tributedto thesuccess of therestorationwork inthe
river (Figure10). Thefoundationof theplatformwas
supportedby0.8mdiameter of steel pipeswithawall
thicknessof 12mmand1.2mindiametercast-in-place
piles. Thesuperstructureconsistedof H700steel pile
cap beams, H700 steel stringers, 18#I beambridge
plandistributionbeams, 10mmbridgedecksteel, and
+45steel pipegrates.Theplatformwas+3.5mabove
groundanditswidthwas89m.
3.4 Strengthening of the ground by jet grouting
Inordertoreducethedeflectionof thediaphragmwall
during excavation and construction of the concrete
proppingslabs, rotaryjetgrouting(seeFigure11) was
carriedout beloweachlevel of slab, startingfromthe
fourthproppinglevel andaroundtheexcavation.
Figure10. Thecofferdamandsteel platformintheHuangpu
River.
Figure11. J et groutingconstruction.
The grout formed a frame-like earth beam to
enhancetheshear strength of thesoil in thepassive
zone. J et grouting was also carried out at the bot-
tomof the excavation. The grouting was combined
with dewatering to protect the excavation fromany
adverse effects such as inflow of water or sand and
baseheave.Twojet-groutedpileswereinstalledateach
joint betweendiaphragmwall panelstominimizethe
ingressof groundwaterintotheexcavation(seeFig.7).
3.5 Dewatering and monitoring
The excavation depth of the two open cut pits was
morethan38m, whichreachedtheaquifer of layer 7.
Dewateringbecameoneof themost vital activitieson
site. Sincethedewateringwasat great depth, ground
settlement hadto bestrictly controlledto protect the
neighboring buildings such as the Linjiang Garden
and the ramp of the Nanpu Bridge fromany dam-
age. Basedonanoptimizeddesignusingresultsfrom
in-situ dewatering tests, 56 wells of 61mdeep were
sunk andrechargewellswereinstalledat appropriate
positions outside the pits. In addition to monitor-
ing the ground settlements, water level observation
436
Figure12. Dewateringmonitoringpointsof theeasternpit.
wells, sub-surfacesettlement monitoring points, and
pore pressure observation holes were installed (see
Figure12).
3.6 Three excavation pits
Theexcavation was 41.2mdeep at thetwo end sec-
tions and 38mat themiddlesection (seeFigure4).
The pit was 22.3mwide at the middle section and
23.7mwideatthetwoendsections. Duetoconstraints
imposed by thepresenceof thedeep obstructions, a
partitionwall fordividingtheentireexcavation, which
was 174.1mlong in the eastern pit, and 62.5min
the western pit, could not be built. Because it was
a very long and narrow excavation, the stability of
theundergroundsupportingsystemwas very impor-
tant. A three-dimensional structural supportingframe
with high rigidity, formed by upright column piles,
reinforced concrete supports and purlins, was used
toprovidetherequiredstiffness tocontrol thelateral
deformations.
4 INTERPRETATIONOF MONITORINGDATA
Excavationfor theeasternpit beganon1March2006
whereasthatforthewesternandmiddlepitson6Octo-
ber2006.Thelateral deflectionsof thediaphragmwall,
settlement of thesurrounding ground and buildings,
andbottomheaveateachstageof theexcavationwere
measured at the monitoring points installed before
excavation(seeFigure13). Duetothepagelimit, only
selecteddataarepresented.
4.1 Lateral deformations of diaphragm wall
Thirty-nineinclinometerswereplacedinthediaphragm
wall tomeasureits lateral deflections duringexcava-
tion.A significantincreaseinthemeasureddeflection
of thediaphragmwall was usedas awarningsignal.
Figure 14 shows the measured wall profiles by the
inclinometer (I12) atsomekeystagesof excavationin
themiddlezoneof theeasternpit.
Figure 13. Plan showing the instrumentation installed on
site.
Figure 14. Deflecion profile of the diaphragm wall (at
inclinometerI12)ateachstageof excavationattheeasternpit.
As shown in Figure 14, the measured deflection
profiles of the wall are typical for a multi-propped
excavation in Shanghai (Liu et al. 2005; Wanget al.
2005). Astheexcavationwent deeper, themagnitude
of inwardwall deflectionincreasedas expected. The
largest deflection of thewall was 49.6mmrecorded
at adepthof 28m. Figure15comparesthemeasured
437
Figure 15. Comparisons of the largest wall deflections
measuredbyall inclinometers.
G
u
t
a
i

R
e
s
t
a
u
r
a
n
t
Figure16. Thelargestsettlementvaluesmeasuredatall the
monitoringpointsinstalledat Gutai Restaurant.
maximumdeflections at all inclinometers. Themea-
sured maximumdeflections ranged from20mmto
50mm, whicharefairly small for a40mdeepexca-
vationas comparedwithother shallower excavations
insoft claysinShanghai (Wanget al. 2005). Thiswas
becauseof thelargerigidity of theretaining system
provided and the early installation of props prior to
eachstageof excavation.
4.2 Settlement monitoring
Theadjacentbuildingsandstructuresaroundtheexca-
vationsuchastheGutai Restaurant, LinjiangGarden
and the ramp of the Nanpu Bridge (see Figure 3)
required major protective measures and their settle-
menthadtobemonitored. Figure16showsthelargest
settlement valuesmeasuredat theGutai restaurant. It
can be seen that largest recorded settlement of the
restaurant was 72.1mmat thecorner J 5-2. This was
becausedewatering was carried out near this corner
region. Figure 17 shows the measured settlement at
J 5-2 versus time. As expected, the measured settle-
mentincreasedastheexcavationprogressed.However,
upwardmovement (heave) wasrecordedafter the9th
stage of excavation and the rate of upward move-
ment seemed to accelerate after the construction of
-75
-70
-65
-60
0 50 100 150 200 250 300
Elapsedtimefrom1-1-2006 (days)
S
e
t
t
l
e
m
e
n
t

(
m
m
)
4th excavation
5th excavation
6th excavation
7th excavation
8th excavation
9th excavation
Final excavation
Construction of
bottom slab
3th excavation
Figure 17. The measured settlement at J 5-2 of the Gutai
Restaurant versustime.
Linjiang Garden

Figure18. Thelargest settlement values measuredat Lin-
jiangGarden.
-40
-35
-30
-25
-20
-15
0 50 100 150 200 250 300
Elapsed time from 1-1-2006 (days)
S
e
t
t
l
e
m
e
n
t

(
m
m
)
4th excavation
5th excavation
6th excavation
7th excavation
8th excavation
9th excavation
Figure19. Themeasured settlement at J 6 of theLinjiang
Gardenversustime.
thebottomslab. Thiswasprobably of theincreasein
pore water pressure and hence a reduction in effec-
tivestressduetotherechargingof groundwater after
dewatering.
Figure 18 shows the measured maximumsettle-
ment valuesat theLinjiangGarden. Asexpected, the
measuredmaximumsettlementof 38.9mmwasatthe
corner closet to theexcavation(i.e. at J 6). Figure19
shows the measured settlement at J 5-2 versus time.
Similar tothat showninFigure17, soil swellingwas
recordedbyJ 6afterthe8thstageof excavationbutthe
ratewassmaller thanthat recordedbyJ 5-2.
438
Figure 20. The largest settlement values measured at the
ramp.
Elapsed time from 1-1-2006 (days)
S
e
t
t
l
e
m
e
n
t

(
m
m
)
The diaphragm wall was being
installed at western pit and the
eastern pit was being excavated.
The western pit was
being excavated.
Figure 21. The measured settlement at the middle of the
ramp(S4) against time.
Figures 20 and 21 show the measured maximum
settlement at all monitoringpointsat therampof the
NanpuBridgeandthesettlementhistoryof S4, respec-
tively. The measured maximum settlement ranged
from2.2mmatS9to29.7mmatS4,whichwaslocated
at themiddleof theramp. Asexpected, themeasured
settlementalongtherampclosertothewestexcavation
pitwasmuchlargerthanthevaluesrecordedawayfrom
thepit. However, nosoil swellingwasrecordedat S4.
Thiswasprobablybecausethedurationof monitoring
atthispointwasnotlongenoughafter thecompletion
of thewesternpit.
5 CONCLUSIONS
This paper provides a case history which illustrates
thatacomplexmulti-proppeddeepexcavationinsatu-
ratedsoftsoilscanbeeffectivelyengineeredbyproper
designandconstruction.Themaximumlateral deflec-
tion (49.6mm) of the1.2mthick diaphragmwall is
relativelysmall fora40mdeepexcavationinsoftclays
inShanghai.
ACKNOWLEDEMENTS
The authors would like to thank many colleagues
who have contributed to the construction and field
monitoring of the project and to acknowledge the
earmarked research grant 618006 provided by the
Research Grants Council of theHong Kong Special
AdministrativeRegion.
REFERENCES
Liu, G.B., Ng, C.W.W. & Wang, Z.W. 2005. Observedper-
formanceof adeepmulti-struttedexcavationinShanghai
softclays.Journal of Geotechnical and Geoenvironmental
Engineering, ASCE 131(8): 10041013.
Wang, Z.W., Ng, C.W.W. &Liu, G.B. 2005. Characteristicsof
wall deflectionsandgroundsurfacesettlementsinShang-
hai. Canadian Geotechnical Journal 42(5): 12431254.
Zhao, X.H. & Yang, G.X. 2004. Practice and theory for
specially big and deep excavation engineering. China
Communication Press: 152156.
439
Excavationentirelyonsubwaytunnelsinthecentral areaof thePeoples
Square
Y.B. Mei
Shanghai No. 7 Construction Co. Ltd., P.R. China
X.H. J iang,Y.M. Zhu& H.C. Qiao
Shanghai No. 1 Construction Co. Ltd., P.R. China
ABSTRACT: The Open and Go-down square is located at the center of the city where the constructional
surroundingsareverycomplex. Itsagreat challengetotheexcavationbecausemetroline1and2tunnelsjust
underliethepit formingashapeof doublecrossing#. Thedetailedexcavationprocess andmany technical
measuresarestatedinthepaper whichmaybereferredbycomingsimilar projects.
1 INTRODUCTION
Withtherapiddevelopment of thestateeconomyand
accelerationof urbanization, manydomesticbigcities
arefacing problems such as land limitation, popula-
tionexpansion, traffic jam, environment pollutionet
al. whichrestrict thecontinuabledevelopment of the
cities. Toexploit theundergroundspacecanfindnew
spacefor thecontinuabledevelopmentof thecity, and
todeveloptherail trafficisaneffectivewayforsolving
theproblemof trafficjam.
With the large-scale exploitation of the under-
ground space and formation of the city rail traffic
network, more and more deep excavations will be
veryclosetotheexistingrail trafficfacilities, includ-
ing 1) sharing thesameretaining wall with existing
subway stations, 2) excavationat thesideof running
subway tunnels, 3) excavationaboverunningsubway
tunnels, 4) excavationbetweenthecolumnsof theele-
vated light-rail, and all thesecases arechallenges to
thenewexcavationengineering.
2 GENERAL CONDITIONOF THE CASE
PROJ ECT
TheOpenandGo-downsquareislocatedatthecross-
ing point of West Nanjing Road and MiddleXizang
Road, neighboring New World Mansion and World
TradeBuilding,anditsonepartof thekeyproject the
pivotal rail traffic project of thePeoplesSquare. The
areaof theexcavationisabout3300squaremeterwith
depthabout4meter.Thetunnelsof runningmetroline
1andline2underliethepitformingashapeof double
crossing#, andtheminimumdistancefromthebot-
tomof thepit to thetopof existingtunnels of metro
line1isonly3.3meter.Thefeaturesof thisprojectare
mainlyasfollows:
1. Thesiteisat thecenter areaof thecity, sothesur-
roundingtrafficisverybusy, andtheorganization
of constructional vehiclesisdifficult.
2. Environmentprotectionisverystrictforthecontrol
of thenoise, vibration, dust, andwastewatercaused
byconstructionactivities.
3. Timelimitationiscritical becausemany construc-
tion activities can not be carried out before the
underground rail traffic stopped running at mid-
night.
4. For thesafetyof therunningmetrolines, thedefor-
mation of the underlying subway tunnels should
becontrolledwithincertainscope, whichmakesit
verydifficult for theconstruction.
5. The deformation of the nearby roads and under-
groundpipelinesshouldbestrictlycontrolled.
6 Manyconcreteobstructionsunderlieandshouldbe
demolishedfirst.
3 GEOLOGICAL CONDITION
Aswell-known, Shanghai isoneof theclassical areas
that have deep soft soil which has poor mechanical
characteristicssuchashighcompression, largedefor-
mationandobviousrheology. Therefore, thesoft soil
iseasilydisturbedandexcavationsshall displaysharp
TimeandSpaceEffect.
According to the geotechnical report, within the
scopethatexcavationaffects, thesoil layersaremainly
441
Figure1. Locationof OpenandGo-downsquare.
Figure2. Planof retainingstructures.
Table1. Soil layers.
Thickness
Soil layer Description m
1
Fill 1.2
2
Fill 0.6
Siltyclay 1.1
Muddyandsiltyclay 3.2
Muddyclay 7.4
1 Clay 8.5
3 Siltyclay 10.0
4 Siltyclay 1.9
2 Finesand 12.3
composedby muddy silty clay andmuddy clay with
thinlayerof finesandbetweenthem. Becauseof some
certaingeological reasons, thesoil layersand
1
are
absent. Thesoil layers fromtoptodownarelistedin
table1.
Thegroundwaterbelongstophreaticwatertypeand
is suppliedmainly by rainfall. Thewater tablevaries
from0.5mto1.2mbelowthegroundsurface.
4 RETAININGSYSTEMANDREQUIREMENT
OF DEFORMATIONCONTROL
Theareaof OpenandGo-downsquareisabout 3300
squaremeterwithlengthof 66m, 55minlongitudinal
andlatitudinal directionrespectively.Theslabof Open
and Go-down square ramps fromground surface to
3.8mbelowinwest-east direction.
Fortheexcavationisnotdeep,adamformedbysoil-
cementmixingpilesisadoptedastemporaryretaining
structure which has a perimeter of 137m, width of
3.2m. Thelength of soil-cement mixing piles varies
from6mto 11m, and its strength should beno less
than 1.2MPa. To reduce the impact on the subway
tunnels, high pressure jet grouting piles with same
parametersareusedtosubstitutefor soil-cementpiles
aboveandbetweenmetroline1tunnels.Thejetgrout-
ing piles have diameter of 800mm, overlapped by
150mm.
The life of subway facilities is 100 years for its
essentiality, anditssafetyduringexcavationisthekey
pointof thisproject. Thedeformationrequirementsof
station structures andtunnels areregulatedas below
bymetro-managingunit.
1. Differential deformation of subway rails in
transversal directionshouldbelessthan2mm,
2. Differential deformation of subway rails in lon-
gitudinal direction should beless than 2mmper
10m,
3. The radius of deformation curvature of subway
structuresshouldbegreater than15000m,
4. The relative deformation curvature of subway
structuresshouldbelessthan1/2500,
5. Thefinal absolutesettlementanddisplacementof
subwaystructuresshouldbelessthan10mm,
6. The rate of structures deformation caused by
excavationshouldbelessthan0.5mmper day,
7. Thewidthof newlygeneratedcrackshouldbeless
than0.2mm,
8. The final accumulated settlement of station and
tunnelsshouldbelessthan10mm,
9. The final accumulated displacement of station
retainingwallsshouldbelessthan2.5mm,
10. Theaccumulated settlement of structures in lon-
gitudinal directionshouldbelessthan4mmper
10m,
11. Thedistancechangebetweensubwayrailsshould
belimitedwithinthescopeof 2mmto+6mm.
For the particularity of this project, although the
excavation depth is not large, the pit engineering is
still designatedtothehighest grade, sothecontrol of
442
Figure3. Sectionof retainingstructure.
deformationindicesareratherstrictthanbeforewhich
arelistedbelow,
1 Theaccumulatedhorizontal displacement of tem-
porary retaining structure should be less than
0.1%H(Hisexcavationdepth),
2 Thesettlement of groundnearby thepit shouldbe
lessthan0.14%H.
5 TECHNICAL MEASURESOF EXCAVATION
5.1 Treatment of soft soil
Because of its poor mechanical characteristics,
Shanghai soft soil makes it rather more difficult in
undergroundengineeringthaninother areas. For the
softsoil iseasilydisturbedandeasilydeforms, usually
certainwaysareadoptedtoimproveitspoor mechan-
ical characteristics. Toreducethelateral deformation
of retainingstructureduringexcavation, jet grouting
pileswithdiameter of 1200mmatspacingof 800mm
areusedforsoil treatment.Nearlythesoftsoil inthepit
areall treatedfromthegroundsurfaceto3mbelowthe
bottom, andthetotal areareaches 2880m
2
. Over the
metro1tunnels, thejetgroutingpilesextendfromthe
groundto500mmtothetopof tunnels. Forthetreated
soil ishardtoremove, sotherequirementsbelowand
above the pit bottomare different. The strength of
treatedsoil belowandabovethebottomshouldreach
1.2MPaand0.6MParespectively.
Usually, when thejet grouting piles is being pro-
cessed, there is fairly high pressure acting on the
veryclosesoils, thusthosesoilswouldbecompressed
and disturbed. To avoid the high pressure acting on
the subway tunnels, semicircular other than circu-
lar jet grouting piles are used adjoining the metro
line1tunnels (seefig. 4). Thesemicircular piles are
formedbyjetgroutingonlyatoppositesideof tunnels
whichiscalleddirectional jetgroutingtechnology, and
this technology is developedduringthepast years in
Shanghai.
Figure4. Planof directional jet groutingpiles.
Figure5. Special anti-uplift structure.
5.2 Special anti-uplift structure
Theexistingrunningsubwaytunnelsareinbalanceby
thesoil belowandover them. If part of thesoil over
themisremoved, thenthebalancestatewill bebroken,
andthereseems to haveaforcedraggingthetunnels
upward, sothetunnelsmayuplift. Toprevent thetun-
nelsfromuplifting, newbalancemust beestablished.
In this project, anew kind of anti-uplift structureis
inventedwhichisindictedinfigure5. At eachsideof
tunnel, reinforcedconcrete(abbreviatedasR.C. here-
after) boredpileswithdiameter of 600mmareset up
closetothetunnel by500mm.Theendof drillingpile
is only 1mfromtheunderlyingmetro line2tunnels
at crossingpoint. Whenasmall areaof soil over the
tunnel isremoved, theR.C. slabinsameareawill be
constructedquickly. Atthesametime, rebarsof bored
pilesareanchoredintoslabandweldedtogether with
itsrebars. Thus, theuplift capacityof boredpilescan
443
Figure6. Excavationstep1.
balance the uplifting force acting on tunnels which
equalstheweight of removedsoil.
5.3 Dividing excavation into pieces
It is not difficult to beunderstoodthat moresoil are
removedfromthetopof thetunnels at onetime, the
uplifting displacement will be larger. To control the
upliftingdisplacement withincertainscope, thepit is
dividedintothreepartswhicharedividedintopieces
either,sotheprocedureof excavationisfairlycomplex.
Thewholeexcavationincludestreestepscorrespond-
ingtothetreeparts. Thenumbersinfigure6indicate
thesequenceof excavationinstep1. Excavationover
the metro 1 tunnels is relatively shallower which is
arranged to be carried out first. Usually, the width
of eachexcavationpieceis 3mwhichis determined
by experience. After thesoil is excavatedto thebot-
tom, thebeddingcushionlayer andR.C. slabwill be
constructed on time. The main rebars in the slab of
different pieces would be connected by mechanical
connector pre-embeddedintheconcrete. Figure7and
figure8demonstrateexcavationinstep2and3, and
thenumbersdonot indicatetheexcavationsequence.
5.4 Loading on the finished slab
Another measureistakentoassurethesafety of sub-
way tunnels during excavation. That is, onepieceof
Figure9. Loadingonslab.
slab will be loaded with heavy materials soon after
beingfinishedandreachingcertainstrength.Theload-
ingmust beequivalent totheweight of soil removed
fromthetunnels top.
6 INFORMATION-BASEDMONITORING
It is well-known that information-based monitoring
is necessary during theconstruction in underground
engineering. Due to vagueness and variability of
mechanical characteristics of soft soil, it is impossi-
ble to predict the deformation of retaining structure
accuratelyduringexcavation, somonitoringistheonly
measure that can see the state of safety. Figure 10
shows thelayout of monitoring points of metro line
1tunnels.
Figure 11 demonstrates the deformation curve of
metroline1tunnel (upper one) duringthefirst exca-
vationstep. Theaccumulatedmaximumupliftingdis-
placement of underlying tunnel is 3.05mmafter the
complement of excavationinstep1. It isobviousthat
444
Figure10. Monitoringpointsof subwaytunnels.
monitoring point
Oct.
27
th
Nov.
2
nd
Oct.
18
th
4.0
3.0
2.0
1.0
0.0
-1.0
S
C
J
0
1
S
C
J
0
3
S
C
J
0
5
S
C
J
0
7
S
C
J
0
9
S
C
J
1
1
S
C
J
1
3
S
C
J
1
5
S
C
J
1
7
S
C
J
1
9
S
C
J
2
1
displacement
Figure11. Upliftingdisplacementof subwaytunnelsduring
excavationstep1.
monitoring point
Jul.
26
th
Aug.
3
rd
Jul.
18
th
7.0
5.0
3.0
1.0
-1.0
-3.0
X
C
J
0
1
X
C
J
0
3
X
C
J
0
5
X
C
J
0
7
X
C
J
0
9
X
C
J
1
1
X
C
J
1
3
X
C
J
1
5
X
C
J
1
7
X
C
J
1
9
X
C
J
2
1
displacement
Figure12. Upliftingdisplacementof subwaytunnelsduring
excavationstep2.
thedeformationis fairly small andthesafety of run-
ningtunnel isguaranteed. Figure12demonstratesthe
deformationcurveof metroline1tunnel (lower one)
during thesecond excavation step. Theaccumulated
maximumupliftingdisplacementof underlyingtunnel
is6.47mminstep2whichincludesdisplacementgen-
eratedinstep1. Generallyspeaking, theaccumulated
deformationisfairlysmall duringtheconstructionof
undergroundengineering.
7 CONCLUSION
With therapid development of theurban rail traffic,
moreandmoreexcavationsaretoberestrictedbyexist-
ingundergroundsubwayfacilitiessuchasexcavation
under,aboveorbetweensubwaytunnelsetal.Onecase
excavationcompletelyontherunningsubwaytunnels
isintroducedindetail inthispaper.Thetechnical mea-
sures employedduringtheprocess of excavationand
thesuccessful experiencecan bereferred by similar
projectsinfuture.
ACKNOWLEDGMENTS
This research was supported by Science and
Technology Commission of Shanghai Municipality
(No. 062012002and07QB14019)
REFERENCES
GarskeE., Kauer H., & Von Soos P. 1989. Excavation lin-
ing and foundation for thenewKreissparkassebuilding
aboveexistingsubway tunnels inMunich, Bauingenieur
(German), 64(11): 505511.
LoK.Y &RamsayJ.A. 1991. Effectof constructiononexist-
ingsubway tunnels, Tunnelling and Underground Space
Technology (6)1: 287297.
Eisenstein Z. Dan, Martin Geoffrey R., & Parker Harvey.
1997.Challengesof tunnelingfortheLosAngelessubway,
ASCE Construction Congress Proceedings, Managing
Engineered Construction in Expanding Global Markets:
281289.
Ookado, & Nobuyuki. 1998. Large-scale excavation in
Tokyofor theconstructionof theNambokusubway line,
Geotechnical Special Publication 86:124143.
ZhuZ. F., Tao X. M., & XieX. S. 2006. Theinfluenceand
control of deep excavation on deformation of operating
subway tunnels, Chinese Journal of Underground Space
and Engineering (1):128131.
ZhangJ. L., LiuG. B., &LiuH. 2006. Studyonconstruction
technologyof excavationonsubwaytunnels,Construction
Technology (4): 8590.
445
Thebenefitsof hybridgroundtreatment insignificantlyreducingwall
movement: A Singaporecasehistory
N.H. Osborne, C.C. Ng& C.K. Cheah
Land Transport Authority, Singapore
ABSTRACT: Thefirst major useof hybridgroundtreatment, J et Mechanical Mixing(J MM), inSingapore,
was in theconstruction of thenewNicoll Highway Station (NCH) for theCircleLineProject (CCLP). This
systemplayedamajor roleinthesuccessof theproject. Theretainingwall designwasa1.5mthickdiaphragm
wall to providesupport of theground during thestation construction and also as thepermanent walls of the
station. Theproposedconstructionschemeisthat of top-downwiththeuseof a7mthickJ MM belowthebase
slablevel, whichcomprisesalargediameterdeepsoil mixingmethodcomprisingacentral corecombinedwitha
jet-groutedouter circumferencetoformalargediameter of improvedsoil mass. Thispaper aimstodemonstrate
theeffectiveness of theexcavation support systemand benefits of J MM in reducing wall movement of deep
excavationinsoft groundwith35mthickof MarineClay.
1 INTRODUCTION
1.1 Background
Theconstructionworks for theoriginal Nicoll High-
wayStation(NCH) ontheCircleLineProject(CCLP)
washaltedwhenacollapseof thecutandcovertunnels
leadingtothestationoccurredinApril 2004. Follow-
ingthecollapse, several options werestudiedfor the
recommencement of theworks. Theoptiontorealign
part of the project to avoid the collapsed site was
eventuallyadopted. Asaconsequenceof thisrealign-
ment, NCHwasrelocatedapproximately100mtothe
south, as shown in Figure1, with thestation design
andexcavationrestartingafresh.
Figure1. Tunnel Alignment Drawing.
1.2 Ground condition
AtNCH, thegroundconsistsof man-madefill, fluvial
sands, fluvial clayandtheMarineClayof theKallang
formation, underlainby theOldAlluvium, as shown
inFigure2. Thethickness of thefill is typically 3to
6meters. Underlyingthefill isalayer of fluvial sand.
Beneaththis, itistheverysofttosoftMarineClay.The
thicknessof thesandlayeris3to7m.Thedepthof the
MarineClay varies from30mto 40mbelowground
level.
Locally, the Marine Clay is separated by a layer
of laterallydiscontinuousfluvial deposits. Thefluvial
sands foundat NCH aretypically describedas loose
tomediumdensegraysandsor siltysands. Theprop-
ertiesof thefluvial sandsaredescribedbyChu, et al.
(2000). Theproperties of theSingaporeMarineClay
andproblemsassociatedwithit fromatunnellingand
deep excavations perspective have been well estab-
lishedinSingapore; seeTan(1972), Shirlaw&Copsey
(1987), Chang (1991) and Tanaka et al (2001), and
generally relate to its softness. The Marine Clay is
normally consolidated or slightly over-consolidated,
withaundrainedshear strength(Cu) startingat about
20kPa and increasing slowly with depth. The com-
pressionindex is typically intherangeof 0.6to 1.0.
Thepermeabilityislowandisintheorder of 10
9
to
10
10
m/s. TheOldAlluviumistypicallydescribedas
sandysiltor clayeysilt.Atdepththematerial isgener-
ally foundtohavesomecementation. However much
of thecementationhasbeenlost duetoweatheringat
shallowdepth. Thepermeabilityof theOldAlluvium
447
Figure2. Groundconditionsat thenewNCHStation.
Figure3. Typical crosssectionof NCHStation.
depends onweatheringandgrainsizedistribution. It
typicallyrangesbetween10
6
to10
9
m/s.
1.3 Section details
ThenewNCH was designedas atopdownstructure
witha1.5mthickpermanentdiaphragmwall anchored
intotheOldAlluviumstrata.Toelevel isat60mbelow
groundlevel.Thisalsoreducestheneedfortemporary
struts as thepermanent concreteroof and concourse
slabs areconstructedduringexcavation. Theexcava-
tiondepthis 20mbelowgroundlevel andthewidth
of excavationis24m. Thecrosssectionof thestation
boxisshowninFigure3.
1.4 Hybrid type of ground treatment
Groundtreatmentunderneaththebaseof thestationis
oftenusedtolimit thewall deflection, act asawork-
ingplatformandpreventupliftingof thesesoftclayey
Figure4(a). J MM Machine.
Figure4(b). Schematicdiagramof thedrillingrodshowing
themixingarmof theJ MM machine.
soils. Theapplicationof groundtreatment suchasjet
groutpiles(J GP) for deepexcavationsinMarineClay
inSingaporehasbeenpresentedbyPageetal. (2006).
For this project, thegroundtreatment optionwas J et
Mechanical Mixing (J MM), ahybrid of jet grouting
and deep soil mixing. A proprietary name, RASJ ET
is givento it by thespecialist contractor fromJ apan,
RaitoKogyo.Thiswasthefirsttimesuchasystemhad
beenusedinlargescaleinSingapore.
J MMisacombinationof soil mixingandjetgrout-
ingthatproducesoverlappingcolumnswithaninternal
column of mixed soil by the auger and an external
column created by a slurry jet into the in-situ soil.
Theprocess of formingthecolumns is similar to the
methodof formingJ GP columnswiththeadditionof
dual and counter rotation mixing blades on thedrill
rodtoensureintensivesoil mixing. Figures4(a) &(b)
showtheJ MMmachines, thedrillingrodandthemix-
ing armof the machine. The rod/auger had a large
diameter of 457mmas compared to the traditional
J GP rod of 200mm. The high stiffness of the drill
448
Figure5. Typical layout of theJ MM columns.
Figure6. Columnscreatedduringwithdrawal.
rod contributes to a more accurate drilling vertical-
ity. Combined with the rod were the mixing blades
whichcreatedaninnermechanical soil mixingcolumn
of 1.6mdiameter. A jet grout nozzleon themixing
bladeintroduces cement slurry mix withpressurized
air intothesoil andaddsafurther 0.6mof jet grout-
ing around thesoil mixing column, creating a2.8m
column within the ground. These columns are then
designedwithappropriateoverlaptoprovideafull cov-
erageof thetreatedareas. Figure5shows thelayout
of themechanical mixingpartandjetgroutingpartof
theJ MM column.
Therearenumerousadvantagestothissystem(Page
et al. 2006& Uedaet al. 2007), principally theben-
efits of mixing and grouting are experienced. From
the mechanical soil mixing, a known treated area is
assured and fromthe jet grouting, a sizeable over-
lapandpenetrationintoanyshadowareasclosetothe
retainingsystemisachieved.
Toinstall aJ MMcolumn, theaugerisfirstdrilledto
thebaselevel of theJ MMcolumnwithwaterinjection,
and withdrawn to the top level of the J MM column
withmechanical mixingwithoutanyinjection. Itthen
descendswithslurryinjectionandmechanical mixing
to formthe internal soil mixing column up to base
level. After which, it ascendswithjettingtoformthe
external jet groutingperimeter. Thewholeprocess is
automatedandmonitoredreal timebydataloggersto
ensurethat ahigh level of quality control. It should
be noted that a further benefit is gained during the
withdrawal theauger after completingtheJ MM slab.
Astheauger iswithdrawn, lower quantitiesof cement
areaddedandthegroundismixed.Thiscreatesa1.6m
diameter treatedcolumnall thewaytothesurface, as
shown in Figure6. Consequently thestrength of the
soil abovethetreatedgroundissignificantlyenhanced.
2 INITIAL DESIGNOF NCHSTATION
TEMPORARY WORKS
2.1 Initial design assumption
Theinitial designof NCHwascarriedoutusingmod-
eratelyconservativeparametersfortheoriginal ground
andJ MM. A conservativeapproach was adoptedfor
theinitial designandthereforenoadditional strength
attributed to the soil cement mix above the J MM
layer was considered. TheUndrainedShear Strength
(Cu) andYoungsModulus(E) of theJ MM layer was
assumedtobe300kPaand90MParespectively, based
upontraditional designparametersusedfor jet grout-
inginSingapore. Table1showstheparametersof the
groundandJ MM usedintheinitial design.
2.2 Geotechnical analysis
Thegeotechnical analysisof theexcavationsequence
was doneusingatwo-dimensional analysis of defor-
mation and stability with PLAXIS (Version 8), a
geotechnical finiteelement program. Thenon-linear
and stress-dependent stress-strain properties of the
soilsaremodeledaselasticperfectlyplasticusingthe
Mohr Coulomb model. The undrained behaviour of
claysandcohesivematerialsissimulatedinPLAXIS
usingMohr-Coulombsoil model intermsof effective
stresses with undrained strength parameters (com-
monly known as Method B in Singapore). This was
donebyspecifyingtheeffectiveYoungsModulus(E
)
and Poisson ratio (
) with Cu under undrained set-

ting. In this method of analysis, theCu of theclays
andcohesivematerials werecappedat theinput val-
uesgivenby theuser. Theinput soil parameterswere
basedonthemoderatelyconservativevaluesshownin
Table1.
Figure7showsthecrosssectionof temporaryworks
for theexcavation. The1mand1.8mdiameter bored
piles weremodeledas afixed-endanchor withhori-
zontal spacing of 12.325mcentreto centreto avoid
possible ill effect of soil-beaminteraction in a 2-D
plane strain analysis. The equivalent length of the
anchor is taken as half the pile penetration length,
which is 16m. Temporary steel plunge-in column is
modeled using the node-to-node anchor to simulate
thebehaviorof anaxiallyloadedmembertoavoidpos-
sibleill effect of soil-beaminteraction. Thoughlarge
449
Table1. Soil parametersandstrengthof treatedsoil usedintheinitial design.
Standard Undrained Drained
penetration Bulk shear shear Friction Elastic
test density, strength, strain, angle, modulus,
Stratum SPTN (kN/m
3
) Cu(kN/m
2
) C
(kN/m
2
)

(
) E (kN/m
2
)
Fill 64 19.0 25 0 30 10000
M 01 16.0 Design: 15(0to10m) 0 22 Design: 300Cu
Worst Credible: 13(0to10) Worst Credible:
Design: 15+1.2(z10) 200Cu
Worst Credible: 13+1.1
(z10)
F2 115 19.0 20(0to10m) 0 22 Design: 300Cu
20+1.14(z10) Worst Credible: 200Cu
F1 117 20.0 0 0 30 1500N
E 04 15.0 15(0to10m) 0 18 Design: 300Cu
15+1.2(z10) Worst Credible: 200Cu
OA(W) N-30 25 20.0 Sandy: 0 Sandy: 32 Design: 2000N
Clayey: 10 Clayey: 25 Worst Credible: 1000N
OA(SW-2) 356 20.0 Design: 5N(max. 500kPa) Sandy: 5 Sandy: 32
30-N-50 Worst Credible: 3N Clayey: 20 Clayey: 25
OA(SW-1) 7013 20.0 10 30
30-N-100
OA(CZ) N>100 >100 20.0 25 32
J MM 16 300 90000
Figure7. Crosssectionof temporaryworksdesign.
voidsarepresentatconcourselevel atreducedlevel of
90.92m, thediaphragmwall panelsaresupportedby
theconcourseslababuttingthediaphragmwall acting
in-planeasawaler beam. Thiswaler beamsupport is
modeledasafixed-endanchor inlateral direction.
AsshowninFigure7, therewere2layersof tempo-
rarysteel strutsabovethepermanent roof slabduring
the excavation. The first and second layers of struts
(S1 & S2) wereH-beamof size414405232 at
spacingof 5.7m. Thefirst andsecondlayersof struts
wereinstalledat reducedlevels of 102m(1mbelow
thegroundlevel) and99.87mrespectively. Thestruts
were modeled in the programas fixed end anchors
witheffectivelengthof 12m, whichishalf of thestrut
lengthbetweenthetwodiaphragmwallssupport. The
pre-stressing on the struts was typically 50%of the
designstrut load.
On comparingthepredictedfiniteelement model
and the actual maximumwall deflections measured
by in place inclinometers, a very significant differ-
enceof 60mmof movementisobserved.A maximum
measured deflection of 25mm compared to a pre-
dicted 85mmwall deflection is observed, as shown
in Figure8. This is avery notablediscrepancy. This
difference rapidly became apparent as the excava-
tion progressed. Although it was evident that much
of this disparity could be attributed to the J MM, as
suggested by the strength and stiffness results from
thesitetestingof theJ MM, back analysiswasunder-
takentoquantifythesedifferences.Althoughachange
in deflection from 0mm to 20mm was observed
over 5m length of the diaphragm wall, no visible
crack was observed. The deflection of diaphragm
wall could be less. This could be due to an inher-
ent problem of the inclinometer installed in steel
pipes in the diaphragmwall with cement bentonite
backfill.
450
Figure8. Predicted wall deflection of theoriginal design
andthemeasuredall deflection.
3 ACTUAL STRENGTHPARAMETERSFOR
J MMANDSOIL CEMENT MIX ABOVE
3.1 JMM strength parameters
Post installation and prior to the commencement of
excavation, anextensivequalitycheckonthestrength
parametersof theJ MMlayeraswell asthesoil cement
mix abovewas carriedout. Figure9shows thesum-
mary of theanalysis of thetest results fromthe7m
thick layer directly beneath the base slab. It shows
the comparison between the average strength of the
J MM layer and theparameters assumed in theorig-
inal design. Also included is a strength factored by
amass correctionfactor to account for any variation
caused during theconstruction process, for uselater
in further analysis (refer to Section 4.1 for details
of correction factor). The average strength, Cu, of
theJ MM layer is about 1845kPa, with very consis-
tent strengthsachievedinthesamplestested, ranging
fromalowest of 1150kPatothehighest of 2370kPa.
Theaveragestrength is morethan fivetimes higher
than that originally assumed valueof 300kPain the
design.
3.2 Soil cement mix strength parameters
Similar findings are experienced on analysing the
resultsfromthestiffnesstesting, Figure10showsthe
comparisonof thestiffnessof J MM fromtest results,
theoriginal designvalueandthefactoredvaluetaking
into account themass correction factor. Theaverage
stiffness of the J MM fromtest results is 572MPa,
Figure 9. Comparison of J MM strength used for initial
designandactual valuesfromtest data.
Figure 10. Comparison of J MM stiffness used for initial
designandactual valuesfromtest data.
Figure 11. Average parameters of J MM and Soil Mixing
above.
withalower boundof 400MPaandanupper boundof
700MPa. Againthetestedaveragestiffnessissignifi-
cantlyhigher thanthe90MPaassumedintheoriginal
design. The various strength of the treated ground
with J MM and soil cement mix above the J MM is
summarizedinFigure11.
451
4 BACK ANALYSIS
4.1 Accounting for better strength parameters
As it was evident in the early stages of the excava-
tion that theretaining wall performance, in terms of
movement, was appreciably better than that design
prediction. The review of the J MM test results sug-
gestedthattheywerethemostplausiblereasonfor the
reducedmovement. Backanalysisof theperformance
wasundertakenfor twomainobjectives: (i) tomodel
theperformanceof thecofferdamwiththeJ MMmod-
eledmoreaccurately; (ii) tojustifytheomissionof the
thirdtemporary strut, whichwas originally proposed
betweentheconcourseandbaseslab. Inprincipal, the
analysis was first donethroughmodifyingtheinitial
finiteelement model usingmost probableparameters
tomatchtheactual wall deflectionandstrutloadof the
secondstageof excavation. Itwasrefinedthereafteras
theexcavationproceeds. At thetimeof thebackanal-
ysis, thesecond stageof excavation was thecurrent
excavationprogress.
The most probable parameters include the most
probable soil parameters, J MM parameters and the
improvedparametersof thesoil mixingcolumnsabove
the J MM. The most likely parameters for 7mthick
J MMslabandthesoil mixingcolumnabovetheJ MM
slabwerebasedontheactual coretestsresultsmulti-
pliedby afactor (0.725), whichcoincides withmost
probableor meanmasscorrectionfactor for theJ MM.
The correction factor is applied to account for the
potential non-uniformityof thegroundtreatment.This
correctionfactor is derivedfromtheestimatedcoef-
ficient based on the square of the ratio of seismic
wavevelocity, obtainedfromcross holeseismic test-
ing(Vsf), totheseismicwavevelocityobtainedfrom
laboratoryUltrasonictest (Vsc), i.e.
CorrectionFactor=(Vsf/Vsc)2
Theresultsof thewavevelocities(Vsf andVsc) and
therational of derivingthecorrectionfactor fromthe
wavevelocities is outlined in apaper by Uedaet al.
2007. Based on theseparameters, back analysis was
carriedouttomatchtheactual wall deflectionandstrut
forces, ascloseaspossible, uptosecondstageexcava-
tion. Table2showsthemost probableparametersfor
theJ MM andsoil abovetheJ MM.
Table2. Most probableJ MM parametersfor backanalysis.
Most probableparametersfor soil cement
Most probableJ MM parameters mixabovetheJ MM (CorrectionFactor=0.725)
(CorrectionFactor=0.725) 94.6m-RL -102.5m(GL) 81.3m-RL -94.6m
Cu
(kPa) 0.7251845=1337.6 0.725178=129 0.725339=245.8

E
50
(MPa) 0.725572.1=414.8 0.72524=18 0.72544=31.9
4.2 Interpretation of the back analysis results
Theresults fromthevarious back analysis aresum-
marized in Figure 12. The wall deflection fromthe
back analysis (Analysis 1) using the most probable
parametersat secondstageexcavation, withthemass
correction, isabletoreasonablymatchthereadingof
the in wall inclinometers at the same stage. This is
supported by the fact that the reading of strut force
at first level strut is quite similar to the strut force
obtainedfrombackanalysis. However, thewall deflec-
tion obtained fromtheAnalysis 2 using unfactored
parametersgivesamuchclosermatchtotheactual wall
movements.Thissuggeststhatthequality,strengthand
stiffness of the J MM are consistent throughout the
entire J MM slab and no correction factor is needed
tobeapplied.
5 CONCLUSIONS
Thiscasehistory demonstratesthat J MM, if properly
installed, hasmajor benefitsincontrollingthestabil-
ity and movements induced by deep excavations in
soft ground. Thereasonscanbeattributedtothefact
that theinner soil columnis comprehensively mixed,
combinedwiththeattributes of theouter jet grouted
columnwithsufficientoverlapping.Thewholeprocess
undergoestightqualitycontrol andrigoroustestingto
ensureacontinuousandcomprehensiveslab. Inaddi-
tion to the J MM slab, there is the major benefit of
Figure12. Wall deflectionfor thebackanalyses
452
the discrete soil mixing columns formed during the
withdrawal of theauger.
Based on the limited usage to date it is difficult
tosuggest what parametersshouldbeusedfor future
design. However it is clear that the key is in the
quality control of theprocess inensuringatotal and
uniformtreatment. With todays engineering sophis-
tication, this canbeachieved. Thestrength, stiffness
andqualityof theJ MMissignificantlyhigher thanjet
grouting, butthechoiceof actual designparametersto
beusediscomplex.Itisrecommendedthattheyshould
bedeterminedonacasebycasebasiswithlocal trials
specifictothegroundconditions, but consideringthe
strengthsandstiffnessalreadyachieved. Thebenefits
of J MM shouldnotbeignoredandthistechniquewill
beafuturebenefittotheindustryincontrollingground
movements.
REFERENCES
Chang, M.F. 1991. TheStress History of Singaporemarine
clay. J. Geotechnical Engineering, Vol. 22, 1991.
Chu, J., Wen, D., Kay, R.E. & Tay, T.H. 2000. Engineering
Propertiesof fluvial sandsat RaceCourseRoad. Proc. of
the International Conference on Tunnels & Underground
Structures, Singapore, 2000.
Page, R.J., Ong, J.C.W, Osborne, N. & Shirlaw, J.N. 2006.
J et Grouting for Excavations in Soft Clay Design and
ConstructionIssues. Proc. of International Conference on
Deep Excavations, 2006.
Shirlaw, J.N. & Copsey, J.P. 1987. Settlement over Tunnels
inSingaporeMarineClay. Proc. of the5thInternational
Geotechnical Seminar Case Histories in Soft Clay,
NTI, Singapore, 1987.
Tan, S.B. 1972. Foundation Problems in SingaporeMarine
Clay, Asian Building and Construction 1972.
Tan, S.B., Tan, S.L. & Chin, Y.K. 1985. A Braced Sheet-
pile Excavation in Soft Singapore Marine Clay. Proc.
of 11th Conference on Soil Mechanics and Foundation
Engineering, San Francisco, 1985.
Tanaka, H., Locat, J., Shibuya, S., Tan, T.S. and
Shiwakoti, D.R. 2001. Characterization of Singapore,
BangkokandAriakeClays, Can. Geot. J. 38:378400.
Ueda,Y., Furusone, T., Sato, Y. & Imamura, S. 2007.
Design, ConstructionandQualityControl of theGround
Improvement Method Adopted for Singapore Marine
Clay Mechanical Soil Mixing with Cement Slurry J et
Grouting. Proc. of 4th Civil Engineering Conference in
the Asian Region 2007.
453
3Ddeformationmonitoringof subwaytunnel
D.W. Qiu, K.Q. Zhou&Y.H. Ding
Beijing University of Civil Engineering and Architecture, Beijing, P.R. China
Q.H. Liang& S.L.Yang
Beijing Jiaotong University, Beijing, P.R. China
ABSTRACT: As akindof modernvehicle, subway has shownus theadvantages of safety, speediness, low
power consumption, lowpollutionetc. Ithasbeenthemainpartof theurbanhigh-capacitypublictrafficsystem.
Moreandmoresubwaylinesappear inorder tomeet theneedof thesociety. Thecivil engineeringconstruction
especiallyinSoft Groundmust leadtothedeformationof theadjacent subwaytunnel, andit causesthesevere
influences to the stabilization and safety of the tunnel. This paper puts forward a kind of three-dimension
deformationmonitoringmethod. It canprecisely monitor thedeformationof thetunnel liner real-timewhich
donot interrupt thesubwaytransport, andexpedientlyprovidemechanicsanalysisonthedeformationof tunnel
construction. Thepaper discusseshowtobuildthe3Dmathematical model for thesubwaytunnel bytheground
lidar surveyingtechnology. UsetheGeorobot tosurveythedeformationof thetunnel crosssectionandtherail
automatically. According to dataprocessing and analysis using amodel interpolation method, the3D digital
model of thedeformation and displacement for thewholemetro is fitted finally. Thus wecan obtain the3D
imageof thesubwaydeformationinaprecision, real-time, stereoandvisual way.
1 INTRODUCTION
Inorder toalleviatethepressureof thegroundtrans-
portation, urbansubwayhasbeenprogrammedorbuilt
inbigcities. Asakindof moderntransportationfacil-
ities, subway has shown us theadvantages of safety,
speediness, lowpowerconsumption, lowpollutionetc.
Ithasbeenthemainpartof theBeijingshigh-capacity
public traffic system. Alongwiththedevelopment of
cityinBeijing, moreandmorecivil engineeringcon-
structionmust leadtothedeformationof theadjacent
subway tunnel. It causes thesevereinfluences to the
stabilizationandsafetyof thetunnel,thusendangerthe
whole urban transportation system. During the con-
struction of periphery foundation ditch engineering
and tunnel engineering, how to guaranteethesafety
of the subway tunnel has been the severe difficulty
neededtobesolved.
Thetraditional monitoringmethodistosetmonitor-
ingcrosssectionsonthedeformationdistrictof tunnel.
By surveying these monitoring points with conver-
gencerule, total stationandleveling, wecanmonitor
thedeformation of thetunnel structure. Somedraw-
backs exist here. First, thenumber of themonitoring
points is limited, which can not reflect thedeforma-
tion tendency. Load analysis of the deformation of
tunnel structureis constrictedandthecorresponding
reinforce measure is difficult to carry out. Second,
the traditional one has not used the remote monitor
method, so it disturbs the transportation of the sub-
waytosomeextent.Third, thedimlight, narrowspace
andthecomplicateenvironment dodisturbthesafety
monitoring.
A method of three-dimension safety deformation
monitoring for urban subway is put forward here. It
can obtain the 3D digital data of the subway tunnel
deformation, whichdonotinterruptthesubwaytrans-
portation. It cannot onlypreciselymonitor thedefor-
mation of the tunnel structure and get the tendency
of subway deformation, but also provide mechanics
analysisontunnel structureandtherail. Thismethod
hasbeenappliedtothesafetymonitoringof theBeijing
subwayline1(BabaoshanStation Bajiaoamusement
parkStation).
2 THREE-DIMENSIONSAFETY MONITORING
METHOD
2.1 Model building
Lidar technology isalsocalledthree-dimensionlaser
scanningtechnology, whichisanewkindof non-touch
surveying method. It can obtain thearray geometric
455
Figure1. Reflectiontarget setting.
imageof surveyobjectfromlaser pointclouds, which
wereemittedbyscanprismandthequicklaser ambu-
lator. So the three-dimension space model can be
made.Thistechnologycanobtainthethree-dimension
coordinates of one point without reflecting prism,
andthespeedcanreach100,000points/second. This
technologyiswell suitedformanyapplications: indus-
trial, architectural, civil surveying, urbantopography,
reverseengineering, archaeology.
Thedetailedmonitoringproject isintroducedhere:
1. Within the deformation district, set the annular
closedsurveycontrol networkalongthemiddleline
of theup-downrail.
2. Put aset of cross sectionevery 3mfollowingthe
middlelineof tunnel.Setthereflectiontargetonthe
arch, vertical wall, subgradeof therail (As Fig. 1
shows), andcollectthepointcloudsinformationby
three-dimensionlidar scanner.
3. BasedontheNURBS Curvedfacefunction(refer
with: Eq. 1), thethree-dimensionmodel of tunnel
isestablishedbythedataprocesssuchasdatajoint
andregistering.
Let the order of polynomial is pq, NURBS
Curvedfacefunctioncanbewrittenas:
where d
i, j
(i =0,1, , m j =0,1, , n) represent
thecontrol vertex, andw
i,j
denotes theweight factor
of vertex, andN
i, p(u)
, N
j, q(v)
is gageB splineprimary
function.
2.2 Real-time safety monitoring system
The24 hour safety monitoring is necessary to guar-
antee the safety of both the tunnel structure and
transportation of the subway. Considering the high
density of thesubways transportation, weadopt the
remoteautomatic monitor systemto real-timemoni-
tor thetunnel structure, vertical wall, andsubgradeof
therail bytheGeorobot.
2.2.1 Measurement principle of Georobot
Georobot is often called automatic electronic total
station(ETS). Itisakindof intelligentelectronictotal
station, which is able to search target automatically,
recognize, trace, collimatepreciselyandobtainthe3D
coordinates.
Target points observation of tunnel deformation
adopts freestationingprincipleof Georobot. Reflec-
tion sheets are set on the target points. To achieve
higher resolutionandimprovereliability of observa-
tiondata, Georobot canfinishredundant observation
automaticallyunder thecontrol of on-boardsoftware.
Thenadjustment of observationdataof different peri-
ods, 3D coordinates values in different periods will
bedonebypost-softwareof computer, finallywecan
get3Dcoordinatedisplacementof targetpoints: (LX,
LY, LZ). For observationnetworksof freestationing,
wechooseindirect adjustment model toprocessdata.
Let t benecessary observationnumber andnbetotal
observationnumber(n>t).Thenadjustmentmodel is:
Thecorrespondingerror equationbythematrixis:
Theabove-mentionedmethodisthat getsdisplace-
mentbycomparingcoordinatesof observationpoints.
Inthecourseof subway tunnel deformationanalysis,
wealsoconsider lateral vector after observationvalue
adjustment of different periods:
where, x
i
, y
i
, z
i
areadjustment value. k isobservation
times of tunnel deformation, L
k1,k
is the conver-
genceof measurementlines, theydonotincludeerror
frompossibledisplacementof basepoints, sotheycan
accurately reflect tunnel deformation. By regressive
analysisof theseadjustmentdeformationvaluesof dif-
ferent periods, wecanconcludeforecastingresultsof
tunnel structuredeformation.
2.2.2 The setting of Georobot
Thespecially madeinstrument pier isput outsidethe
right of thefirst rail. Georobot is forced to befixed
456
on the instrument pier through the pedestal and is
protectedwiththeglasscover. 616reflectingsheets
areinstalledfor eachmonitoringstation, whichisdis-
tributedinthearch, vertical wall, orbitdrainageditch,
rail fastener and so on. With the monitor program,
Georobot collects the coordinates automatically and
transportsthedatatothecontrol server viadatawire.
3 ENGINEERINGAPPLICATIONS
Thethermal pipelineengineeringof Babaoshansouth
road crosses over the structure of Beijing subway
line1 (Babaoshan Station Bajiao amusement park
Station). Engineeringconstructioncausesthesubway
tunnel deformation. Thekilometer post of thetunnel
deformationdistrict isfromK3+770toK3+810.
Thesafety monitoring result is shown as follows:
theaccumulativedeformationvalueof themaxdefor-
mation point on the tunnel structure is +1.90mm;
the max accumulative value of the rail deformation
is +1.86mm; the max differential settlement of the
rail subgrade is 0.29mm. So such conclusion can
bemadethat theaccumulativedisplacementsof both
tunnel structure and rail caused by the engineering
construction areless than 2mm, which is within the
allowed deformation rangeand put no influences on
thesubwaytransportation.
4 CONCLUSIONS
This method of three-dimension safety monitor-
ing has the advantages of high automation and
three-dimensional measurement. It canbeappliedto
thesafetymonitoringof high-risebuilding, sideslope,
deepfoundationditchengineeringandsoon.Of course
this method is not so mature. Its theory needs to be
further researchedthroughengineeringexperience.
ACKNOWLEDGEMENTS
The authors wish to acknowledge the support of
the Key Technologies R&D Programme of China
(No. 2006BAJ 15B01) and the Beijing University
of Civil Engineering and Architecture Science
Research Foundation of China (No. 100701205,
No. 100601905).
REFERENCES
Bassett, R.H., Kimmance, J.P. & Rasmussen, C. 1999. Auto-
mated electrolevel deformation monitoring system for
tunnels. Proceedings of the Institution of Civil Engineers,
Geotechnical Engineering: 117125.
Wu,Z.A.1989.Processing Deformation Data of Engineering
Construction. Beijing: SurveyingPublishingHouse.
Xu, W.P. 2000. ExploringonFeasibilityof NewWaytoMon-
itor Tunnel Displacement. Railway Engineering Transac-
tion 66(2).
Yu, L.F. &Duan, D.Q. 1996. Real-time Theodolite Industrial
Surveying System. Beijing: SurveyingPublishingHouse.
457
Challengingurbantunnellingprojectsinsoft soil conditions
H. Quick, J. Michael & S. Meissner
Prof. Dipl.-Ing. H. Quick, Ingenieure und Geologen GmbH, Darmstadt, Germany
U. Arslan
Univ.-Prof. Dr.-Ing. U. Arslan, Technische Universitt Darmstadt, Germany
ABSTRACT: Different challenging tunnel projects in thedowntown areaof thecity of Mainz in Germany
arepresented. Thesetunnels runparallel inadistanceof 4mtomax. 50m. Thetunnels werebuilt insoft soil
conditionsconsistingof filling, clayandmarl layersof theTertiary.Thepaper presentsthedifferentconstruction
techniques, thecalculation methods for thetwo tunnels as well as theresults of measurements for theNew
Tunnel Mainz. Theexperiencefor theconstructionof thistunnel andtheresultsof themeasurementswerethe
basisfor thechosenconstructionandcalculationmethodfor therehabilitationof theOldTunnel Mainz, which
iscurrentlyunder construction.
1 INTRODUCTION
TheNewTunnel Mainz had been constructed in the
years1998to2001directlyadjacenttotheexistingOld
Tunnel Mainz.Thisoldtunnel builtin1884iscurrently
beingrehabilitated, convertedandenlargedduringthe
nextyears.Duetothesmall overburdenof bothtunnels
and sensitive structures on the ground surface chal-
lenginganduniquetunnellingtechniqueswerechosen
toguaranteethestabilityandserviceabilityof thetun-
nelsandof sensitivestructures. Inadditioncalculation
methodsandresultsfromgeotechnical measurements
are presented in the following. The situation of the
tunnelsisshowninfigure1.
Figure1. Planviewtunnel situation(Quicket al. 2001).
2 GROUNDCONDITIONS
The geological condition is mainly characterized
bythetertiarystrata(Miocene) of theMainzer Basin.
TheTertiary strata sequence consists of an alternat-
ing sequenceof marly clays, chalk marl, sandy silts
(hydrobiasilts, hydrobiaoyster shells) andsandsinan
alternating sequence with chalkstone banks (fig. 2).
Thechalkstonebanks arepartly compact/massiveto
weathered. The consistency of the in-situ ground is
stiff tosemi-solid, turningintosoft/paste-likeif water
intrudes.Thegroundwatercanbefounduptothelevel
of thefloor/upperedgeof track; otherwisethereisonly
local stratumwater of littleimportance.
459
Figure2. Geotechnical longitudinal section OldTunnel Mainz.
Figure3. GermancorecentretechniqueOldTunnel Mainz
(Maidl 2004).
3 OLDTUNNEL MAINZ
3.1 Construction
TheOldTunnel Mainz waserectedintheyears1881
until 1884asonecontinuousdoubletracktunnel. The
tunnel has anhorseshoeshapeandtheliningconsist
of sandstonewith a thickness of approx. 0.9m. The
Germancorecenter tunnelingtechniquewas usedto
built theOldTunnel Mainz withanoverall lengthof
1200m. The German core center technique (fig. 3)
usespartial drivings, mostlysidewall drivings. Dueto
thebadconditionof thetunnel, especiallythemasonry
and to improve the smoke venting system, a 300m
long and up to 26mdeep open cut was built in the
early30iesof thelastcentury, whichdividesthetunnel
nowadays into theTunnel Mainz Central Stationand
theTunnel Mainz south(fig. 1/2).
The Old Tunnel Mainz consists of the following
structures:
Tunnel Central Station: 663m
Opencut: 300m
Tunnel Mainz South: 246m
Figure4. Geological crosssection.
4 CONSTRUCTIONOF THE NEWTUNNEL
4.1 Construction techniques
Parallel to the existing Old Tunnel Mainz, the New
Tunnel Mainz was built in the years 1998 to 2000.
The new 1250m long double track railway tunnel
with a low overburden of 10mto 23mruns under
buildings including a hotel with basements up to
10m under ground level. Moreover, there are old
(Roman) underground hollow spaces (gallery sys-
tems) tobeundercrossed.
TheclearancebetweentheOldTunnel Mainz and
thenewonevariesbetween4mand50m(fig. 4).
Regardinggroundconditions, existingsettlement-
sensitivestructures andthepossibleinfluenceonthe
OldTunnel Mainz theexcavationof theNewTunnel
Mainz hadtobecarriedout only withlittledeforma-
tion. Hence, anuniversal shotcretetunnellingmethod
with side wall drifts followed by the excavation of
thecalotteandcore/benchwaschosenasconstruction
method(fig. 5).
The distance between the side wall faces and the
final lining (ring closure) was limited to less than
100m and in particularly sensitive parts to 50m.
The distance between the calotte face and the ring
closure of the preliminary support was restricted to
30m. Apart from the usual measurements in tun-
nellingadditional securingmeasures wereappliedin
460
Figure 5. Undercrossing of a hotel; driving concept and
securingmeasures(Steiger et al. 1999).
areas of settlement-sensitive structures. They are as
follows:
Horizontally injected steel pipe roof shelter
(strengthening of the longitudinal rock bearing
arch) (fig. 5).Theroof shelterisplacedintheupper
areaof theface.Thelengthof thedrillingis20.5m.
Theminimumoverlappingtothenextroof shelteris
3m. Theadvantageof thistechniqueisquitevery
obvious; awidening of theroof areato placethe
drillingsisnot necessary.
Inorder topreventanydilatational effectabovethe
tunnel roof, 45degreeinclined, groutedPVC-fans
areinstalledincontinuousdistancesof about 5m.
Injections fromthe ground surface, pre-installed
injectionsystemsunderthefoundationof buildings
aswell assystematicfaceboltingswithinthetunnel
areappliedadditionallyinordertominimizedefor-
mations.All thesemeasurestogether inconnection
with prior determined combinations of available
measures were the basis of a successful driving
withlittledeformationsunder settlement-sensitive
structures.
4.2 Calculation method
For proof of thestability and serviceability 2D- and
3D-numerical calculationwerecarriedoutbymeansof
theFinite-Element-MethodwiththeprogramAbaqus
(figs7, 8). Continuumelementswereusedforthesoil,
whereasbeamelementsforthelining. Fortherealistic
simulation of thesoil an elastoplastic soil behaviour
was chosen (Quick et al. 2001). The modified
Figure6. Principleof thealpha-method(Quicketal. 2001).
Figure7. 2D-finiteelementmeshof theNewTunnel Mainz.
Drucker-Prager material law with cap was imple-
mented.Theyieldsurfaceof thiselastoplasticmodel is
notconstantintheprincipal stressspace. Itcanexpand
dueto plastic straining. Furthermoreit distinguishes
betweendifferentstiffnessforloading, unloadingresp.
reloading.
Forthecalculationof theprecedingdeformationsas
well astoaccountforthreedimensional archingeffects
aroundtheunsupportedtunnel thealpha-methodwas
applied (fig. 6). The principle of this method is to
reducethestiffnessof thefiniteelementswhichareto
beremovedinthenextcalculationstep. Thereduction
causes changes in the initial stress field and there-
fore leads to preceding deformation. In case of the
sidewall and calottedrivings thefactor is set to 0.5.
Thisassumptionwhichcontrolsmainlythepreceding
deformationswasverifiedbymeasurements(fig. 11).
The3D-finiteelementmodel wascreatedbyextrud-
ing the 2D-mesh. Under respect of the construction
procedurethelengthwaschosento100m(fig. 8).
4.3 Monitoring
Regarding the extraordinary situation to undercross
several settlement-sensitivestructures with only low
461
Figure 8. 3D-finite elements mesh of the New Tunnel
Mainz.
Figure 9. Measured surface settlements due to drivings
(TM 117).
overburden in soft ground an extensivegeotechnical
monitoringprogramhadbeencarriedout. At thesur-
facethedeformationsduetotunnellingaremeasured
in close distances by levelling as well as deforma-
tionmonitoringsystems, workingontheprincipleof
correspondingtubes.Figure9showsthemeasuredsur-
facesettlementsduetothedrivingat stationTM 117.
The surface settlement adds up to 5cm. The settle-
mentsmeasuredatthegroundsurfacealongthetunnel
(fig. 10)wereinmostareasbetween1.5cmand2.5cm
inaverage; at thevery beginningof thedrivingaddi-
tional securingmeasures as describedprior have
notbeenapplied; thesettlementsatsurfacereachedup
tounacceptable11cm.
Figure 10 shows the surface settlements under
respect of thedifferent drivings (sidewall drift, exca-
vationof crownetc.) at stationTM 117. This station
is closetotheportal north(fig. 1). Most of themea-
suredsurfacesettlementsarerelatedtoexcavationof
thesidewall driftsandthecrown, whileonlyasmaller
amount of settlements arise from the bench/invert
excavation.
Figure10. Surfacesettlement at surfaceduetotunneling.
Figure11. Comparisonof measurements andcalculations
(roof displacements).
Figure11showsthecomparisonbetweenthemea-
suredroof displacementsof thetunnel andthecalcu-
latedroof displacementsatstationTM117.Theresults
of the 2D-calculation show a good correspondence
withmeasurementsregardingtheprecedingdisplace-
mentaswell asthedisplacementsduetotheexcavation
of thesidewall driftsandthecrown. However thecal-
culatedheaveof theroof duetotheexcavationof the
bench/invert doesneither matchthemeasurement nor
theexpectedgroundbehaviour (fig. 11). For suchsoft
soil conditionsitisthereforerecommendedtoincrease
thestiffnessof thefiniteelementsbelowthetunnel.
5 REHABILITATIONOF THE OLDTUNNEL
5.1 Construction technique
The rehabilitation and enlargement of the Old Tun-
nel Mainz is going to be done under respect of the
462
experiences of the New Tunnel Mainz. The partial
looseground (old back filling) around theOldTun-
nel isimprovedby injections. Inthefirst steptheold
filling around the tunnel is injected. Voids will be
filled, improved by injections (bulk filling). For this
10drillings upto alengthof 2.5minalongitudinal
distanceof 1.5marecarriedout (fig. 12).
In the second step up to 8.5m long injection
drillings aroundthetunnel to activateasupport ring
aredrilled. Thesetwo groundimprovement steps are
doneunderprotectionof theexistingOldTunnel. Sub-
sequentlythecalottedrivinginshotcretemethodwith
an additional forepiles takes places. For thestability
of the preliminary lining of shotcrete with a thick-
ness of 0.30mawideningof thecalottefootingwas
established. Inthefinal constructionsteptheinvert is
excavatedandthepreliminaryringclosureisachieved.
Thedistancebetween thefaceof thecalottedriv-
ingandthepreliminaryringclosurewithshotcreteis
limitedtolessthan12m.
5.2 Calculation method
For the design and the proof of the serviceability
of the tunnel and the mentioned structures 2D- and
3D-numerical calculationarecarriedoutwiththepro-
gram Plaxis. The numerical calculations regard all
constructionphases as well as theformer excavation
of theOldTunnel Mainz andtheNewTunnel Mainz.
Inorder to createrealistic results thematerial lawof
HardeningSoil withayieldsurface, whichisnotfixed
intheprincipal stressspaceisused. Theyieldsurface
expandsduetoplasticstraining. Inadditionthemate-
rial lawcandistinguishbetweendifferent stiffnessfor
loadingandunloadingresp. re-loading.
Toaccountfor thethreedimensional archingeffect
of the unsupported enlargement of the Old Tunnel
Mainzthebeta-methodisappliedunder respectof the
usedcalculationprogram.Theprincipleof thismethod
isdescribedin3steps(fig. 13):
1. Generationof theinitial stressfield
.
2. De-activation (excavation) of the tunnel clusters
without activation of the tunnel lining and gen-
eration of (1-beta) reduced forces, which can be
donebyareductionof theultimatelevel of thefull
calculationstep.
3. Activationof thetunnel lining.
The beta-value was obtained by an iterative back
analysisof theNewTunnel Mainz. Thepredictedsur-
facesettlementsof the2D calculationfor theground
surfaceamount toapprox. 2cm(fig. 14).
5.3 Monitoring
Therehabilitationof theOldTunnel Mainz isaccom-
paniedbyanextensivemonitoringprogramwithinthe
Figure12. Tunnellingtechnique OldTunnel Mainz.
Figure13. Principleof thebeta-method.
Figure14. Calculationresults OldTunnel Mainz.
463
tunnel and on ground surface. For a quick interpre-
tationof thedatathreelevels of settlement limits for
different areas under respect of the overburden and
the existing structure were defined. On the basis of
pre-defined measures such as an additional tempo-
rary shotcreteinvert settlements canbesloweddown
andreducedto guaranteethestability of theexisting
structures.
6 CONCLUSION
The presented tunnels in the inner city of Mainz in
soft soil conditions showavariety of different mea-
sures inthegroundfor thedrivings inorder to meet
therequirementsof theserviceabilityof existingbuild-
ings. 2Dor 3Dcalculationswerecarriedouttopredict
thedisplacements. It is shown that theevaluation of
the input parameters and the calculation methods is
complicated but decisive for the calculated results.
Hence, everytunnellingprojectmustbeaccompanied
by an intelligent monitoring programto observethe
impact on the environment due to the drivings and
to ensure the stability and serviceability of the tun-
nel andneighboringstructures as well as to generate
datafor possibleback-analysisinorder toimprovethe
calculatedresults andto verify theassumptions. The
paper showsalsonumerical approacheswithdifferent
material lawsandcalculationmethods.Thesedifferent
approaches can all lead to tolerableresults, if meth-
ods are used properly and the input parameters are
evaluatedappropriately.
REFERENCES
Emeriault, F., Bonnet-Eymard, T. &Kastner, R. 2005. Move-
ments induced on existing masonry buildings by the
excavation of a station of Toulouse subway line B
5TH international symposium, Amsterdam, Geotechni-
cal Aspectsof UndergroundConstructioninSoftGround.
J une2005
Kovacevic, N., Edmonds, H.E., Mair, R.J. & Higgins, K.G.
1996. Numerical modelling of the NATM and compen-
sation grouting trials at Redcross Way Geotechnical
Aspects of Underground Construction in Soft Ground,
Rotterdam, 1996
Maidl, B. 2004. Handbuch des Tunnel- und Stollenbaus. 3.
Auflage. VerlagGlckauf GmbH, Essen
Quick, H., Michael, J., Arslan, U. 1999. Tunnelling for
German High Speed Railway Lines Proc. Civil andEnvi-
ronmental EngineeringConference NewFrontiers and
Challenges, Bangkok, Vol. 2, part I, pp. II117128
Quick, H., Michael, J., Arslan, U. 2001. About the effect
of preliminary measures on ground movements due to
tunnelling Responseof BuildingtoExcavation-Induced
GroundMovements, London, 1718. J uly2001
Quick, H., Meiner, S., Michael, J., Arslan, U. 2001.
Vergleich von Ergebnissen numerischer Berechnungen
mit in-situ Messungen am Beispiel eines Tunnelvor-
triebes STUVA-Tagung, Mnchen, 1922. November
2001, Studiengesellschaft
Quick,H.,Michael,J.,SchttnerV.,Arslan,U.,Katzenbach,R.
2000. Tunnelling and Deep Excavation in Soft Ground
GeoEng2000, Melbourne, 1924. November 2000
Steiger, H., Theissen-Wenzel, C., Quick, H. 1999. Neuer
Mainzer Tunnel Behandlung geotechnischer Grenzflle
in der Planung und Ausfhrung Vortrge zum 6.
Darmstdter Geotechnik-Kolloquium, Darmstadt, 11.
Mrz 1999, Mitteilungen des Institutes und der Ver-
suchsanstalt fr Geotechnik der TechnischenUniversitt
Darmstadt, Heft Nr. 44
464
Supervisionandprotectionof Shanghai MassRapidLine4shieldtunneling
acrosstheadjacent operatingmetroline
R.L. Wang&Y.M. Cai
Shanghai Metro Operation Co. Ltd., P.R. China
J.H. Liu
Shanghai Municipal Engineering Administration Bureau, P.R. China
ABSTRACT: Shanghai Mass Rapid Line4 shield tunnels cross theoperating Line2 with 1.03mdistance
beneaththeoperatingline2, small angleandsmall turningradius. Thetunneling-acrosswascarriedoutwithout
groundimprovement.Thisprojecttookgreatchallengetotheoperatingline.Theinfluenceof thetunnel crossing
waspredictedandascientificconstructionschemeandmeasureswerearrangedtoguaranteesafeoperationof
Metro Line2. Theconstructionwas performedbasedonthefollowingprinciplestrictly to pushstepby step
slowly, to turnequably withshort step, to maintainstablepressure, andto improvethefoundationwithlower
pressureandsmall amount.Thankstothesemeasures, theshieldtunnel crossedMetroLine2successfully.
Zhangyang RoadPudian Road tunnel of Metro
Line4wasconstructedwithshieldmethod.Theshield
tunnel crossedunder theoperatingLine2theoperat-
inglinewithasmall angleandsmall turningradius.
Figure 1 shows you the # shape of the relation
between Metro Line 2 and 4. Following difficulties
wereencounteredinthisproject.
1. Distance. Thenewtunnel cross blowtheexisting
tunnel withminimumdistance1.03m.
2. Thetunnel atthecrossingareaiscurvedwithsmall
turningradius. Theradiusof Line4tunnel isonly
380m. For the shield, to advance with a small
radius curve would influence the surroundings
moreseriouslythanadvancingalongstraight line.
3. Influencearea. For Line2, theshieldadvancewill
directlyinfluence60mat theupdirectionlineand
94m at the down direction line. Length of the
influenced range, in which the distance between
theupper andlower tunnel is less thanthetunnel
diameter, isabove300m.
4. No ground improvement. The construction point
waslocatedinaplacewhereseveral roadscross, so
thatthegroundtrafficisveryheavy. Itisimpossible
toconduct anygroundimprovement.
5. Badgeological condition. Line2tunnel islocated
in the fourth layer stratum, which exhibits high
compressibilityandobviousrheological behavior.
6. No relevant experience. Prior to this project, the
tunnel of Line 2 had ever crossed below Line 1
Figure1. Planeandprofileof thecrosssection.
when constructing, but in that case the ground
improvement was carried out. To sumup, tunnel
crossingwithout soil improvement ishighlyrisky.
2 PROCESSANALYSISAND
COUNTERMEASURE OF THE SHIELD
TUNNELING
2.1 Process analysis
Theinfluenceof constructiontometrooperationwas
predictedandthechangeof constructioncharacterwas
465
Figure2. Deformationof tunnel Line2.
analyzed. Thetheoretical analysis shows: beforethe
shield-advancereachthecentrelineof Line2tunnel,
it would disturb the equilibrating stratumstress and
changethestressdistribution. Itwouldinduceupward
andforwarddeformationto Line2. Whentheshield
facepassedthrough, theLine2tunnel woulddeform
continuouslyduetothereasonsincludingexcesspore
water pressure dispersing, consolidation settlement
andsecondaryconsolidationsettlementandoperation
trains vibration. Inordertocontrol thedeformationof
Line2tunnel, it is necessary to conduct synchroniz-
inggroutingandsecondgroutingtoLine4tunnel.The
groutingcouldgiveapre-heavingtoLine2andthefur-
thersettlementcouldbesmaller.If theoperatingtunnel
settlement approached the limitation, a re-grouting
would be needed. Meanwhile ground improvement
could be carried out between two tunnels to reduce
settlement of Line2. Consequently, when theshield
crossedtheLine2tunnel,longitudinal deflectioninthe
horizontal and vertical direction might be occurred.
It is composed with tunnel horizontal displacement,
vertical heave or settlement and radial convergence
deformation. It isshowninfigure2.
2.2 Main disturbing factors in the shield advance
process and its control measure
Indeed, various factors would cause Line 2 tunnel
deformation, thefactors, however, could bereduced
tothefollowingthreecategories:
1. Duringtheshieldadvanceprocess, theshieldshell
and some projecting parts on the shield shell
would cause stratumloss. The stratumloss vol-
ume, marked as V
e1,
is closely relevant with the
shieldadvancevelocity. Inorder toreducethestra-
tumloss, wemustlimittheshieldadvancevelocity
strictly and try our best to make it suitably slow
and constant. According to the theoretical analy-
sisandtheexistingexperience, theshieldadvance
velocity was set as v=510mm/min. Further-
more, we request the shield advance must pause
for1030minevery10minadvance.Accordingto
themetroprotectionstandardmentionedaboveand
themonitoringdatafromautomaticequipments,we
wouldadjustandre-determinethepausetimeatany
moment.
2. Volume loss caused by curved shield advance is
marked as V
e2
. When the shield is advancing in
curve, in every 10cm advance, an angle devia-
tion, markedas, shouldoccur at theshieldaxes.
Volumeof V
e2
varies directly as theshieldlength
squareandinverselyasthecurveradius. Toreduce
the volume loss, articulated equipment of shield
machine must be used correctly, to reduce the
shield effective length. The deviation-correcting
degreeper timemust belimitedstrictly to ensure
V
e2
as small as possible. Theoretically speaking,
deviation-correcting at every ring advance with
1200mmringwidthis17.6mm. Whenitisdivided
into 12 times, however, the deviation-correcting
volumeper timeis only less than1.5mm. That is
tosay thesmall degreeandmulti-timesdeviation-
correctingwouldinfluenceonly1/12areaasinflu-
enced by the original way. The disturbed area is
thereforeconsiderablyreducedandthevolumeloss
volume, V
e2,
isalsoreducedconsumedly.
3. Volumelosscausedby unbalancedpressureat the
shieldface, markedasV
e3
. V
e3
isthevolumeloss
causedby differencebetweenshieldfacepressure
and stratum original static pressure LP. During
thecoursethat theshield advanceis approaching
Line 2 tunnel, in order to reduce the metro tun-
nel settlement caused by shield tail grouting lack
andconsolidationsettlement, LPshouldbeaposi-
tivevolume, namelytheshieldfacepressureshould
be05%larger than stratumoriginal static pres-
sure. BecausetheLine4tunnel isinasmall radius
curve in this section and it crosses the operating
Line2 with asmall angle, pressureon theshield
cutter head is unbalanced at left and right sides
during the shield crossing through process. And
suddenpressurechangemightoccurduringthepro-
cess. Tomaintainthepressureat shieldfacestable
(especially whenthesegment is erecting, thejack
wouldbegettingloose) andto makethesoil bea
littleuplifting is very beneficial to Line2 tunnel
settlementcontrol. Consequently, whether thecon-
structionissuccessful or not isdependonwhether
thesoil pressureat theshield facecould becon-
trolled and maintained at a suitable level. What
should be adjusted in the construction process is
theparametermentionedabove: theshieldadvance
velocity, v; pressureat shieldface; andaxesdevia-
tionoccurredinevery10cmadvance, , (generally,
thisindexisdefinedasthepredeterminedvolume).
General effectivenesscausedbythesethreefactors,
whenreflectedintheLine2tunnel horizontal dis-
placement, vertical uplift or settlement andtunnel
radial convergence, is
v
h
and the tunnel radial
convergencedeformation.
Accordingto themetro tunnel deformationspeci-
fication and metro safety operation requirement, for
theoperating Line2 tunnel, themaximumtolerance
of the longitudinal deflection in the horizontal and
vertical level planes are determined as 5mm; and
466
the tunnel radial convergence deformation tolerance
isstrictlycontrolledwithin20mm.
2.3 Real time information construction
Duringtheshieldadvanceprocess, thevertical deflec-
tion of the Line 2 tunnel, both in vertical level
horizontal level, marked as
v,
and in the vertical
plane, markedas
h
mustbecontrolledstrictly. When-
ever theshieldadvancesastep, thechangerateof
v
and
h
must beanalyzedandtheconstructionparam-
eters must be adjusted within 5mmtolerance, so
that to control the tunnel deformation. In the shield
advance process, high-automatic and high accurate
monitoring systemwas adopted to supervise Line 2
tunnel status lively. High-precise automatic electri-
cal level supervisionsystemwas adoptedtomeasure
thetunnel longitudinal settlement andtransversesub-
sidence difference between two tracks; and Bassett
convergencesystemadopted to supervisethetunnel
convergencestatus.Afterthetrainservicefinished, we
adopt additional manual supervision to monitor set-
tlement, displacement, andconvergencestatus of the
Line2 tunnel (to set 3 settlement monitoring points
at eachprofile; 1onthetrack bedandtheother 2on
thesegment themonitoringprofilesareset every2m
alongthetunnel.)
2.4 Analysis on grouting and foundation
improvement
After the shield cross over the Line 2 tunnel centre
line, whether theinterspaceat theshieldtail couldbe
refilledtimelyandsuitablyisakeyfactorsthat cause
tunnel longitudinal deformation. In this process, the
shieldjackstressonthetunnel segmentwouldcausean
oppositelongitudinal deflection,butcontrol volumeof
h
must belessthan5mm. After theshieldpassed,
theconsolidation settlement of thelayer beneath the
tunnel isalsoanunneglectablefactor. Therefore, after
the shield passed below the Line 2 tunnel, several
timely groutingwithlowpressureandsmall amount
isveryimportantanddirectlyrelatedtothesuccessof
Line2tunnel settlement control.
3 TECHNOLOGY KEYSOF THE SHIELD
ADVANCEANDPRE-CONTROL OF MAIN
CONSTRUCTIONPARAMETER
Basedonthedeepanalysesandclearunderstandingon
theconstructiondifficulties, theoperator conducteda
45kph speed limit in the relevant section of Metro
Line2.
Wecomply with thefollowing principlecarefully
andstrictlyintheconstruction topushstepbystep
slowly, toturnequablywithshortstep, tomaintainsta-
blepressure, andtoimprovethefoundationwithlower
pressureandsmall amount.Thetechnologykeysare
decomposedintothefollowingpoints:
1. Settlement or uplift control volumeof theLine2
tunnel is5mm. Whentheuplift volumereaches
50%andsettlement, 30%, of thecontrol volume,
themonitoringsystemwill alarm. That makesthe
tunnel alwaysinaslight deformedcondition.
2. The shield advance slowly and constantly;
the shield advance speed is maintained at
510mm/min.Andtheshieldadvancemustpause
for 1030min every 10min advance. The pause
timelengthdependsonwhether theLine2tunnel
hasuplift fall back totheprotectionstandard.
3. Theshieldfacesoil pressureshouldbeadjustedto
acorrect volumeso that it wouldcauseminimum
influenceontotheLine2tunnel. Tomaintainasta-
blefacesoil pressureis extremely important. The
shieldfacepressureshouldbe100105%of stra-
tumoriginal static pressure, however, thevolume
isflexibleandshouldbeadjustedbasedonthelive
monitoringdata. Whenthesegments areerecting,
thereshouldbeeffectiveprecautiontoprevent the
facesoil pressurefromdescending;andtheerecting
timeshouldbeasshort aspossible.
4. Shield Advance Deviation-correcting. The
deviation-correctingdegreeper timeisthesmaller
the better. The deviation-correcting is required
to taken every 10cm advance. And the articu-
lated equipment of shield machine must be used
correctly, toreducetheshieldeffectivelength.
5. Synchronizing grouting pressure should be less
than 0.4mpa or even lesser. Grouting amount is
1.11.8 times of theinterspacecubage. Detailed
groutingamount dependsonthemonitoredLine2
tunnel data.
6. Second liquid grouting would be adopted in the
secondgrouting.Thegroutingconductprincipleis:
lowpressure,multi-times,moderateandtimely.The
pressureshouldbemaintainedat0.2mpa. Grouting
amountpertimeperringis50100liter.Werequest
thegroutingshouldbeconductedeveryother ring,
toreduceitsinfluencetoLine2above. Generally,
we request the grouting only be conducted when
thetrainservicefinishes.
7. Foundation improvement is necessary to prevent
theLine2tunnel fromsettlingcontinuously. Inthis
process, all relevant parameters such as, grouting
pressure, groutingamount, groutingtime, grouting
pipewithdrawingvelocity, andgrout mix propor-
tionandconsistency, shouldbelimitedcarefullyto
reducetheharmful influencetothetunnel aboveas
muchaspossible.
8. Setautomaticandhigh-preciseelectrical level sys-
temalongtheLine2tunnel. Theelectrical level is
set toacquiredataevery5minutesanddeliver the
datatoshieldoperationcontrol centre. Theopera-
tioncentrewouldadjustconstructionparameterand
467
Live Automatic Monitoring
Construction Goes On
Within the Tolerance
Compare the Collected Data with the
Tolerance
Construction Parameter Adjustment
Tunnel Deformation Control
B
e
y
o
n
d

t
h
e

T
o
l
e
r
a
n
c
e

Y
N
Figure3. Pre-control systemsketch.
guideshieldconstructionbasedonthesedata. The
constructiontechnology keys andquantifiedcon-
structionparameteraretogetherformedaneffective
andreliablepre-control system, whichisshownin
figure3.
4 SUPERVISIONONTHE SHIELD
CONSTRUCTIONANDTHE OPERATING
TUNNEL DEFORMATIONANALYSES
Theshieldmachineis8.625mlongandtheouterdiam-
eter +=6.34m. Thesegmentsare350mmthick and
singlesegmentlongitudinal lengthis1.2m.Thewhole
constructionprocessisasfollows:
1. The first tunneling across process of the shield.
Shield at Line4 down linetunnels across Line2
up linefirst, and then tunnels cross Line2 down
line.
2. After theshieldmachinearrivedZhangyangRoad
Station, soil between the tunnels at the crossed
section was improved to ensurethemetro Line2
operationsafety.
3. Thesecond tunneling cross process of theshield.
The shield turned back to advance in the Line 4
downline. It crossLine2downlinefirst, andthen
tunnelscrossLine2upline.
4. Afterthetunnelingfinished, soil of thissectionwas
improved. Four tunnelsatthecrossthroughsection
formeda#shapewithsmall angleonaprojection
plane. Theproject thereforeinfluencedvery large
rangeof Line2. Lengthof therange, inwhichthe
distancebetweentheupperandlowertunnelsisless
thanthetunnel diameter, isabove300m.
Figure4. TheLine2 up linetunnel settlement/uplift data
duringthefirst tunnelingcrossprocess.
Figure5. TheLine2downlinetunnel settlement/upliftdata
duringthefirst tunnelingcrossprocess.
4.2 Process description and site monitor of the first
tunneling cross
The first shield tunneling cross process was started
fromPudian Road Station. Thefirst tunneling cross
processlastedfor 15days.
Attheverybeginningof theshieldadvance, thepre-
set constructionparameter makestheLine2tunnel to
maintainaslight upliftingtrend.
Whentheshieldcutterheadwasapproachingbelow
theLine2uplinetunnel projection, theshieldadvance
velocity and face soil pressure were adjusted, but
theLine2tunnel still maintainedupliftingtrendand
theuplift volumewas upto0.20.5mm. Duringthe
shieldadvanceprocess, soil atdifferentpointshaddif-
ferentstress,thuscauseddifferentinfluenceintheLine
2tunnel.Whendeviation-correctingconducted, soil at
headandleft of theshieldmachinewas stressedbut
stress of soil at head and right was released; mean-
while, at the shield tail, soil at the right side was
stressedbut stressof soil at left sidewasreleased.
When the shield advanced just below the axis of
Line2tunnel (theshieldwasonly1.44mawaybelow
Line2tunnel then),thefacepressureontosoil hadbeen
adjustedspeedilyfromtheprevious0.230.27mpato
0.166mpa, muchsmaller thantheprevious pressure.
468
Figure6. Settlement/upliftdataof theclosestpointfromthe
Line2uplinetunnel duringthefirsttunnelingcrossprocess.
Figure 7. Settlement/uplift data of the closest point from
theLine2downlinetunnel duringthefirst tunnelingcross
process.
At that time, theuplineof Line2tunnel haduplifted
for 2.53.1mm.
When the shield cutterhead was escaping from
theprojectionof Line2tunnel, becauseof incorrect
andunsuitabledeviation-correctingandsynchronized
grouting, instantaneous uplift volumeof Line2 tun-
nel was up to 3.3mm. Figure 4 and figure 5 show
thesettlementandupliftconditionduringshieldcross
through Line 2 up line tunnel and down line tunnel
respectively.
Figure6andFigure7showthesettlementanduplift
conditionof theclosestsettlementpointduringshield
crossthroughLine2uplinetunnel.
4.3 Process description and site monitor of the
second tunneling cross
Thesecondshieldtunnelingcrossprocesswasstarted
from Zhangyang Road Station. By J une 24, 2003,
when the shield of Line 4 construction arrived 2m
awayfromtheprojectionlineof theLine2downline
tunnel, theshieldadvancehadcausedsettlement and
uplift at Line 2. On J une 30, the shield cutterhead
entered theprojection lineof Line2 down linetun-
nel, thenit crossedover belowtheLine2downline
anduplineoneafter another. By J uly 22, theshield
Figure8. TheLine2downlinetunnel settlement/upliftdata
duringthesecondtunnelingcrossprocess.
Figure9. TheLine2 up linetunnel settlement/uplift data
duringthesecondtunnelingcrossprocess.
escapedfromLine2projectioncompletely. Thesec-
ond shield tunneling cross process totally lasted for
23days.
TheshieldcrossesbelowtheLine2downlinetun-
nel first with theminimumvertical distance1.03m.
ItsinfluenceontheLine2tunnel isobvious, although
we have made very strict limit on the construction
parameter. ByJ uly1, theLine2downlinetunnel has
upliftedupto 2.5mm. Becausetheuplinetunnel of
Line4isveryclosedtoLine2tunnel, theLine2tun-
nel assumedobviousdynamicsettlementduringcross
throughprocess. It canhardlybecontrolledexcept by
synchronized grouting and second grouting method.
Figure8toFigure11showsdetailedcondition.
4.4 Grouting and foundation improvement issues
In the shield advance process, in order to control
the Line 2 tunnel settlement after shield advance,
unreactive-grout-synchronized grouting was con-
ducted, so that the grout would refill interspace at
theshieldtail. Meanwhile, synchronizedgroutingand
second grouting was conducted at the cross section
where dynamic settlement occurred. Double liquid
grout was used in second grouting to improve the
stratumsmechanical properties. 16groutingholesare
469
Figure 10. The Line 2 down line tunnel settlement/uplift
dataateverytimeduringthesecondtunnelingcrossprocess.
Figure11. TheLine2uplinetunnel settlement/uplift data
at everytimeduringthesecondtunnelingcrossprocess.
Figure12. Schemaof theFoundationImprovement.
reservedat everysegment inthecrosssection, sothat
itwouldbeeasytoconductstratumgroutingimprove-
ment after the shield escaped fromthe section. The
arrangement of thefoundationimprovement grouting
pointisshowninfigure12.After theshieldpassed, by
repeatingfoundationimprovement intherecent year,
theLine2tunnel settlement situation is nowwell in
hand.
5 CONCLUSION
After theelaborateconstructionandstrict sitesuper-
vision and protection, the Line 2 tunnel settlement
situationisnowwell inhandandthesafemetrooper-
ation during the project are also realized. Valuable
experience, which could be reference for the future
construction, wasaccumulatedinthiscase.
1. When the shield advancing in curve with small
radius, becauseof different stressconditionat dif-
ferent point of the shield, it would cause various
influencetotheupper metrotunnel.
2. Theshieldadvancingvelocityisanimportant fac-
tor that causes upper tunnel uplift. Theinfluence,
however, couldbereducedbysuitableconstruction
parameter adjustment.
3. Shield deviation-correction is another important
factor that causes upper tunnel uplift. The uplift
or settlement condition is dependent on relative
positionof thetwotunnels.
4. Shieldtail groutingwouldinfluencequitealarge
range, upto 1015m. It is necessary to maintain
low pressure, small amount and correct grout in
thegroutingprocess, toreducetheinfluencetothe
surroundings.
5. Dynamicsettlementcausedbytrainrunningisvery
dangerous. Synchronizing grouting and second
grouting could be adopted to correct that. Foun-
dationimprovementanddoubleliquidgroutingare
important methodtocontrol tunnel settlement.
6. Theprinciplestopushstepbystepslowly, toturn
equablywithshortstep, tomaintainstablepressure,
andtoimprovethefoundationwithlower pressure
andsmall mountareprovedcorrecttechnical keys.
7. Stricthigh-preciseautomaticmonitoringandinfor-
mation based construction is one of the most
important technical support.
ACKNOWLEDGEMENTS
I would like to extend grateful appreciation to
Mr.Bai Tinghui,Mr.GeShiping,Mr.ShenChengming,
Mr. Sun Lianyuan and Mr. Ni Chengyu for their
assistanceduringcomposingthisthesis.
470
KowloonSouthernLink TBM crossingover MTRTsuenWanLine
tunnelsinHKSAR
K.K.W. Wong, N.W.H. Ng& L.P.P. Leung
Kowloon-Canton Railway Corporation, HKSAR
Y. Chan
MTR Corporation Limited, HKSAR
ABSTRACT: TheKowloonSouthernLink(KSL) isa3.8kmundergroundrailwayprojectinabusyurbanarea
withcomplexgeological andalignmentconstraints.Akeychallengefortheprojectistoconstructacritical section
of thetunnelsabovetheexistingMTR tunnelsandbelowaheavilytraffickedroadjunction, whichistermedas
theMTR Crossing. Cut andcover methodwasoriginallyenvisagedtoconstruct theMTR Crossing. Inviewof
thecomplexgroundconditionsandother physical siteconstraints, thecontractor proposedtoextendthebored
tunnelsusingatunnel boringmachine(TBM) acrosstheexistingMTR tunnelsasanalternative. Adoptingthe
TBMmethodfor theMTRcrossingintroducedspecificchallengeswhichcalledfor comprehensiveengineering
studies, extensivegroundtreatment, tailoredprotectivemeasurefor theMTR tunnels, aswell assophisticated
instrumentationinorder toensurethat therailway operationsandthepublic safety arenot compromised. This
paper istodiscussthedesigndevelopment of theschemeandthemethodologyof theMTR Crossing.
1 INTRODUCTION
TheKowloon Southern Link (KSL), running 3.8km
underground, will link West Rail from its current
terminusatNamCheongStationtotheEastRailster-
minusat EastTsimShaTsui Station, withanewWest
KowloonStationinbetween. ByjoiningEastRail with
WestRail, itwill providecommuterswithaconvenient
interchangebetweenthesetworailwaycorridors.
TheKSL constructionworks arebasically divided
into two sections, northernsectionandsouthernsec-
tion. Thenorthern section covers atwin 2.1kmrun-
ningtunnels betweenNamCheongStationandWest
KowloonStation, mainlyonreclaimedlandwherethe
conventional cutandcover constructionistheprevail-
ing method. The 1.7kmsouthern section comprises
a new station and twin 1.1kmTBM bored tunnels
locatedinabusytouristandcommercial areabetween
West Kowloon Station and East TsimShaTsui Sta-
tion.Theprojectincludesthecrossingperpendicularly
abovetheexistingMassTransitRailway(MTR)Tsuen
WanLine(TWL) tunnelswhichruninthenorth-south
direction underneath the heavily trafficked Nathan
Road(Figure1).
This paper focuses on theTBM crossing over the
MTRTWL tunnels.
Figure1. TheKSL Alignment.
2 SITE CONDITIONSANDREGIONAL
GEOLOGY
ThesolidgeologyintheKowloonpeninsulaandalong
theKSL tunnel routeandat theMTR crossingcom-
prises intrusive igneous bedrock belonging to the
Kowloon Granite Formation. Salisbury Road is 4m
abovesealevel andconstructedoveroldreclaimedfill
underlainbysuperficial deposits.Thedepositsoverlie
471
Figure2. Geological sections.
arelativelythinlayerof completelydecomposedgran-
ite(CDG) abovebedrock. Despitecloseproximity of
theharbour thegroundwater tableinSalisburyRoad
was typically 2.5mbelow ground level with minor
influencefromtidal fluctuations.
The geological interpretation indicated that the
MTR down track tunnel lies within the granite
bedrock. For theMTRuptracktunnel, as-builtrecord
indicates the bedrock and completely decomposed
granite interface to be at the axis of the MTR up
tracktunnel. However, baseduponadditional geotech-
nical informationandboreholeresults, this interface
was interpreted to be higher above the tunnel axis
(Figure2).
3 TUNNEL DESIGN
3.1 Original design based on cut and cover
The original scheme design of the KSL tunnels
along Salisbury Road was based on the conven-
tional cut-and-cover constructionmethod. KCRC, the
project proponent, submitted its original proposal
basedupontheoriginal schemedesigntotheHKSAR
Government.
However, MTRraisedanobjectiontotheproposed
schemefor thereasonthattheproposedcut-and-cover
method at the MTR crossing would give rise to the
risk of floatation due to removal of the overburden
abovetheMTR tunnelsat thecrossing. Subsequently,
analternativemicro-tunnellingtechniquewasfurther
proposedfor theMTR crossing.
The proposed micro-tunnelling method required
handmininginsoftgroundwithinapre-installedhori-
zontal pipepiledenclosurearoundtheperimeterof the
tunnel structure. It wascontendedthat thisalternative
methodwouldamelioratetherisk of floatationof the
existingMTR tunnelscomparedtothecut-and-cover
method. Assuch, MTR agreedtoallowauthorization
of theschemeby theHKSAR Government basedon
themicro-tunnellingmethod.
DespiteMTRsnoobjectiontotheuseof themicro-
tunnelling method fromrailway protection point of
view, thereweredrawbacksof thismethod. Itisrecog-
nizedthat handmininginsoft ground, inparticular in
marinedepositsandbelowgroundwatertable, isahigh
riskactivity.Therisksof handminingincludetheindi-
rectsupportsystem,instabilityof minedface,potential
for seepageandpipingthroughthegapsbetweenhor-
izontal pipepiles, andthelikelihoodof largeground
movements resulting in a risk of damaging existing
utilities. To mitigate these risks, substantial ground
treatment wouldberequiredfor themicro-tunnelling
constructionattheMTRcrossing. Furthermore, exca-
vation by hand mining with staged installation of
temporarysupportsandconstructionof thepermanent
tunnel structurewouldbetimeconsuming.
DuringtheKSL tenderingstage, someof theten-
derersproposedtheuseof TBM methodfor theMTR
crossing. Due to the limited time for the tendering,
noneof thetendererswasabletoobtainMTRsagree-
mentonadoptingtheTBMmethod.Assuch, theTBM
crossingasapotential alternativeremaineduncertain
at thetimeof thetendering.
3.2 Selling the TBM crossing alternative
Immediatelyafterthecontractaward,avalueengineer-
ingworkshopwasheldamongKCRC, thecontractor
anditsdesigner. ConsensuswasreachedthattheTBM
crossing alternative should be seriously considered.
Thecontractors designer undertook adetailed engi-
neering study of the TBM crossing over the MTR
TWL tunnels. Thestudy includedatwo-dimensional
finiteelement numerical modelinganalysisusingthe
softwarePLAXISVersion8.2andathree-dimensional
analysisusingthesoftwareFLAC 3DVersion3.0.
Inboththe2-Dand3-Danalyses, thegeotechnical
parametersadoptedwerebasedonthegeologydeter-
minedfromexistingsiteinvestigationinformationand
theMTR constructionrecords. Duetotheclosesepa-
rationof thetwoKSL tunnel bores, groundtreatment
using jet grouting was recommended by KCRC and
its contractor (Figure 3). The analyses showed that
the movements and distortions of the MTR tunnel
lining due to the KCRCs TBM crossing would be
withintherequiredstructural tolerance. Furthermore,
a parametric study was carried out with varying jet
grout strengths between 0.5MPa and 1.5MPa and
it indicated that the impact on the MTR tunnel lin-
ingmovement wouldbesatisfactoryandthejet grout
strengthwasnot acritical factor.
Despitethefavorableresults of thedesignassess-
ment, MTR expressed a specific concern regarding
thepresenceof bouldersabovethecrownof theMTR
tunnels and below the invert level of the KSL tun-
nels. MTRconsideredthattherewasariskof theTBM
472
Figure3. Layout of J et Grout Block.
dislodgingaboulder whichmay beconnectedto the
MTR tunnel lining, resultingindamageto theMTR
tunnel lining. Inorder tomitigatethisrisk, it waspro-
posedto drill closely spacedhorizontal pipepiles to
physically isolatetheMTR tunnelsfromthematerial
throughwhichtheTBM wouldbore. Furthermore, jet
grouting theentireareaabovetheMTR tunnels was
proposedtoformaconsolidatedblockof soil andgrout
matrix. Onthisbasis, MTR agreedinprincipletothe
TBM methodof crossingtheMTR tunnels.
However, thealternativeTBMproposal still needed
toundergothestatutoryprocessof theamendment to
theauthorizedscheme, whichalsoinvolvedconsulta-
tionwithrelevant Government authorities, stakehold-
ers, as well as local communities. Themerits of the
TBM methodcomparedtotheoriginal schemeof cut
and cover tunnels wereobvious fromthepublic and
social benefit pointsof view, apart fromtheconstruc-
tionbenefits.TheTBMmethodminimizesdisturbance
tothetrafficandexistingutilitieswithmuchlessenvi-
ronmental impacts interms of noise, dust andvisual
impact, in particular in such abusy commercial and
tourist areawheretherearenearbyprestigioushotels,
shoppingmallsanddistinguishedcultural facilities.As
such, thealternativeproposal for adopting theTBM
method was well received by all therelevant parties
without anyobjection.
4 TECHNICAL CHALLENGESFORTBM
CROSSING
KCRC andits contractor neededto revisethedesign
details of both the permanent works and temporary
workstocater for therevisedTBM crossingmethod.
This includedaddressingthevertical alignment con-
straints imposed by the existing MTR tunnels, the
connectionintotheexistingKCRC overruntracksof
theETS Station, aswell asthegroundconditionsfor
TBM tunnelling(Figure4).
Figure4. KSL Alignment constraints.
Inorder tomaintaintheminimum1.5mclearance
of any KSL works fromthe extrados of the MTR
tunnels, the depth of the ground cover to theTBM
at the MTR crossing was as shallow as only 6.8m
which was less than the diameter of theTBM (viz.
8.05m). Giventhegeologyat thecrossingwithboul-
ders present at thetunnel invert andmarinedeposits
withinthetunnel horizonandabovethetunnel crown,
theshallowcoverisconsideredachallengetotheTBM
tunnellingincontrollingthegroundsettlementsonone
handandminimisingtheriskof groundheaveorslurry
leakageontheother. Therewereexistingcritical utili-
tiesabovethecrossing.Theconsequenceof damaging
these utilities would be very serious. Also, any sig-
nificant disturbanceto thesurfacesuch as excessive
groundmovementsor slurryleakagewouldinevitably
causedisruptionto thetraffic at this busy roadjunc-
tion.Thesekindsof incidentswouldhavemajorimpact
on theroad users and KCRCs corporateimageand
thereforemeasuresweretakentomitigatetherisksof
their occurrence.
Duetotheabovevertical alignment constraint, the
structural clearancebetweentheKSL tunnelsandthe
MTR tunnelswasonly2m. Thepresenceof boulders
at the KSL tunnel invert had been a cause for con-
cern on potential damageto theMTR tunnel lining.
TheMTR TWL carries morethan amillion passen-
gers daily. Apart fromthe public safety which is of
paramount importance, any disruptionto therailway
services would havesubstantial impact to thepublic
andisunacceptabletoMTR andGovernment author-
ities. Careful planning, control andmonitoringof the
TBM drives crossing above the MTR tunnels were
deemedtobecritical andessential inorder toensure
that there would be no adverse impact on the MTR
tunnels. Thesecalledfor theneedof thedevelopment
of widerangeof additional protectivemeasures and
geotechnical instrumentation.
Theexistingoverruntracksof theETSstationtothe
eastof theMTRcrossinghadbeenoperational priorto
theKSL construction. Thealignments of theseover-
runtrackshadbeenbasedontheoriginal cutandcover
473
Figure5. ThemixedgroundTBM.
tunnel schemeof theKSL withaverynarrowsepara-
tionbetweentheuptrackanddowntrack. hestructural
clearance between the two KSL tunnel bores at the
MTRcrossingwasonly900mm.Thisextremelyclose
separationof thetunnelsgivesrisetoatechnical chal-
lengefor theTBM tunnelling, inparticular onhowto
avoidcausingexcessivemovements tothefirst com-
pleted KSL tunnel during theadjacent secondTBM
drive.
5 TUNNEL BORINGMACHINE
A purpose-builtmixedgroundTBM, usingmixshield/
slurry technology, was used to excavate the tunnels
through both rock and soil deposits (Figure 5). The
8.05mdiameter Herrenknecht TBM, named Little
Dragon Girl (Xiaolognu), was designed to suit the
prevailinggroundandgroundwater conditionsandso
limited the potential for face instability and ground
settlement whenexcavatinginsoft andmixedground
conditions. TheTBM cutter head was also equipped
withheavy-duty cuttingtools for excavationincom-
petent rock. A hydraulically powered jaw type rock
crusherwasinstalledinfrontof thesuctiongrill behind
the cutter head and was capable of crushing blocks
suchasbouldersupto500mminsize.
Theoperatingprincipleof themixshieldTBM isto
utilisethepressurizedbentoniteslurryattheconfined
tunnel faceto withstandthehydrostatic pressureand
groundpressureinall instances. Thissystemenables
thetunnel facestabilitytobemaintainedat all times,
eveninamixedsoil/rockinterface.
Thebentonitesuspensionforthetunnel facesupport
ispressurisedbyanaircushionorairbubble. Little
DragonGirlwasequippedwithacompressedairreg-
ulationunit whichaccurately controlledtherequired
air bubble pressurewithasensitivity of +0.05bar.
Pressurelevelswerecalculatedfor varioussectionsof
theTBM tunnel drivesaccordingtothedifferent geo-
logical andgroundwaterconditions. Extensivetestson
themarinedeposits intheareaof theMTR crossing
werecarriedout inorder toobtainsufficient datafor
calculatingtheair bubble pressuretobeadoptedby
theTBM whencrossingover theMTR tunnels.
6 ADDITIONAL PROTECTIONTOMTR
TUNNELS
6.1 Ground treatment
Therearemultipleobjectivesof thegroundtreatment
abovetheMTR tunnelsasfollows:
To stabilize the ground whilst the TBM passes
abovetheMTRTWL tunnelsandbeneathsensitive
utilities,
To provide water cut-off and prevent lowering of
groundwaterandlossof soil whentheTBMbreaks
intothereceptionshaft, and
To cater for TBM low ground cover and narrow
separation
Ground treatment in the formof 1.5mdiameter
jet grout columns wereadopted using cement based
grout(includingmicrofinecements) andsilicates. For
quality control purpose the unconfined compressive
strengthof post-drilledcores was set at 1MPamini-
mumat28days. J etgroutcolumnsgenerallyterminate
at1.5mfromthecrownof theMTRtunnelsand2.5m
abovethecrownof thefutureKSL tunnels. Thetotal
number of jet grout columnsdesignedis690.
Thejet groutingemployedadoublefluidjet grout-
ing systemutilizing a swivel head that directs two
different fluids of grout and air through concentric
double tube rods. Grout and air exit horizontally
throughthetwoconcentricnozzlesattachedtothebot-
tomof thegrout string. Theair andgrout areinjected
at highpressuresof 6barsand200barsrespectively.
Theair andgrout injectionbreaksupandmixeswith
thesoil matrixresultinginagroutandsoil slurry. The
diameter of thegrout columniscontrolledbytherate
of rotation and withdrawal of the string. In general,
a rotation rate of 6 to 7rpmand withdrawal rate of
1820min/mwereadoptedtosuit thesoil stratum. In
the cohesive clay layer, precutting by water jet and
grouting was carried out at a slower drill rod with-
drawal rateof 2022min/m. Excesssludgeandslurry
isremovedthroughtheannulusbetweentheboreholes
andthedrill rods by air liftingeffect. Thecompleted
column is formed as ahardened cylindrical body of
soil cement mixture.
Workingspaceandaccesswereseverely restricted
duetotherequirementtomaintaintrafficflowthrough
abusyjunctionaswell asnumerousundergroundutil-
ities (Figure6). It was necessary to divert thetraffic
in multiplestages to facilitatejet grout column con-
struction within thefootprint of theavailabletraffic
management schemes. A combinationof vertical and
474
Figure6. J etgroutinginprogress.
rakingholeswereconstructed,dependingonthetraffic
andutilityconstraints.
Qualitycontrol testingwascarriedout toprovethe
effectiveness of theground treatment. Coresamples
weredrilledandcollectedfrom5%of thecolumnsfor
unconfinedcompressivestrengthtesting.Permeability
testswerealsocarriedout invertical drill holesfrom
groundsurfaceandwater inflowmeasurements were
madeinhorizontal probeholesfromtheTBMretrieval
shafttoverifythewater-tightnessof thecompletedjet
grout block. Theresultsprovedtobesatisfactory.
6.2 Horizontal pipe piles
Horizontal pipe piles were installed through the jet
groutedzoneabovetheMTRTWL tunnelstoprovide
physical separationbetweentheTBMandMTRTWL
tunnels, andtoisolateany corestoneboulderswhich
might bedisturbedbytheTBM bore.
Conventional Down-the-Hole (DTH) techniques
were adopted for the drilling operation using an
ODEX bit. Thehorizontal pipeswere273mmdiame-
ter 6.3mmthickpipepilesinstalledat 500mmcentre
to centrespacing. Thetotal length of each pipepile
is about 23m in order to extend over both MTR
tunnels and 5mbeyond the centre line of MTR up
track tunnel. Structural clearance of the horizontal
pipe piles to the MTR tunnels varies froma max-
imum of 2.0mm to a minimum of 1.397m. Tight
structural clearance is a combined result of the ver-
tical alignment constraints as described in section 4
aboveand thedownward deviation of thehorizontal
pipepiles.Drillingof isolatedpileswhichinfringedthe
railway protection zonewerepermitted however this
wasconditional onadditional requirementsfor work-
ing during non-traffic hours, detailed survey checks
andcomprehensivemonitoring.
Controlling the alignment of the horizontal pipe
piles is critical to ensurethepipepiles do not devi-
ate downwards within 1.2mof the MTR tunnels or
upwards into theTBM bore. Alignment control was
Figure7. Horizontal pipepilesinstalled.
achieved using a down-the-hole survey method. A
surveyprismisinsertedintothefarendof thepiles,fol-
lowedbysurveychecksevery3masthesurveyprism
iswithdrawnfromthepipe.Thesurveycheckiscarried
out at 5intermediatestages for eachpileas follows:
(1) beforedrillingandafter theset-upof thedrilling
mast; (2)at4mdepthbeforearrivingattheMTRdown
track tunnel; (3) at 10mdepthbetweentheMTR tun-
nels; (4) at 16mdepthbeforearrivingat theMTR up
track tunnel; and (5) after reaching the final length
of 23m.
Various contingency measures were developed in
theevent of recordingunacceptableupwardor down-
warddeviation.Theseincludedreplacingthepipepiles
withbundledrebar withinabentonitecementmixand
down-sizingthepipepilesto219mmindiameter.
7 INSTRUMENTATIONANDMONITORING
The instrumentation and monitoring for the MTR
tunnels was required to meet statutory and spe-
cificcontract requirements. Inaddition, Alert, Action
andAlarmResponselevels (AAA) for building and
structuremovement, utilities, ground settlement and
groundwater changeweredeveloped.
Itisapre-requisiterequirementfortheworkswithin
theMTRrailwayprotectionzonethattheMTRrailway
anditscontinuousoperationmust not bejeopardized
andbesafelymaintainedduringtheTBMcrossing.As
revealed by thenumerical modelling analyses, some
movementsof theMTRtunnel liningmaybecausedby
theconstructionof theTBMtunnelsdirectlyabove.To
ensurethesafeandcontinuousoperationof theMTR
railway at all times, an extensive regime of instru-
ments and monitoring programme was established
bothinsideandoutsideof thetunnels.Themonitoring
programmewasalargepartof themitigationmeasures
forthecrossingtominimizetherisksof damagetothe
MTR tunnel liningandconsequential interruptionto
475
therailwayservicesandincludethefollowingtypesof
instruments and method of monitoring. Instruments
andmonitoringwithintheMTR tunnels:
AutomaticDeformationMonitoringSystem(ADMS):
real-timedeformationmonitoringof thetunnel lin-
ingbyautomatictotal stationswithintheinfluence
zoneat thecrossing
VibratingWireStrainGauges:real-timemonitoring
of changesinthestressof tunnel lining
Seismograph:real timemonitoringof groundborne
vibrationsinducedbytheworks
Tapeextensometers: manual convergencesurveyof
thetunnels
Manual tracklevel offset surveys
Instruments and monitoring of the ground sur-
roundingtheMTR tunnels:
VibratingWirepiezometers: real-timemonitoring
of changesinporewater pressure
Manual andautomatedin-placevertical inclinome-
ters: monitoring deformation of the ground and
cofferdamwalls
In-placehorizontal inclinometers: installedinside
selectedhorizontal pipepiles aboveMTR tunnels
monitoringreal timegroundmovement
Magnetic probeextensometer: manually monitor-
ingof deepgroundsettlement.
All monitoring data was uploaded directly to an
instrumentationdatabasesystemcalledGEOMON
whichisaproprietorproductfromthespecialistinstru-
mentationsubcontactor. GEOMONalsomanagedthe
data by automatically alerting designated recipients
of AAA responsevalueexceedances viae-mails and
SMS, and could also be viewed and downloaded
remotelyviatheInternet.
Upon exceeding theAAA levels, procedures out-
lined in the Exceedance Plans were implemented
accordingly. During excavation for the first of the
two KSL tunnels theTBM successfully crossedover
theMTRtunnels. Insignificantmovementandground
bornevibrationwasrecordedintheMTR tunnels.
8 CONCLUSION
Thechangefromtraditional cut-and-cover construc-
tiontoTBM methodenabledtwonewrailwaytunnels
to be constructed directly above one of the worlds
busiest operating railways with minimumclearance,
but no trainspeedrestrictions or interruptions to the
servicesandwithanexcellent safetyrecord.
Thesiteconditions and theexisting railways pre-
sentedanunusually onerous rangeof technical chal-
lenges. Thesuccess of theMTR Crossinghas estab-
lished a new benchmark for the TBM tunnelling
techniqueand with such recognition received would
no doubt leadto its further potential useincomplex
urbanenvironment similar tothosefoundinHKSAR.
Theprojecthasstronglydemonstratedthat, for par-
ticular spatial and operational constraints, theTBM
tunnellingmethodcanprovidethebest overall value
and performance in constructing new underground
infrastructures.
ACKNOWLEDGEMENT
TheMTRCrossingrequiredalsothedetailedplanning
andsubstantial cooperationandpartneringinteraction
between thetwo railway corporations. Theproactive
involvement fromMTR intheselectionof anappro-
priate day for the crossing minimizing all possible
adverseimpacts, theprovision of stand-by teamand
engineeringtrainsduringthecrossingnightsreacting
to incidents, if any, instantaneously, andthecarrying
outadvancedimprovementworksinsideTWL tunnels
beforethecrossing, areall contributingfactorstothe
successof MTR crossingwhichwasdeliveredsafely
andwithout interruptiontothetrainservices.
Theauthorswouldalsoliketothankthecontractor
Link200J V andtheirdesignteamMottMeinhardtJ V
for theinvaluableinput in developing and finalizing
theMTR Crossingscheme.
REFERENCES
HKSARWorksBureau, Environmental TransportandWorks
Bureau, Technical Circular No. 19/2002 Mass transit
RailwayProtection, May2002.
476
Applicationof pileunderpinningtechnologyonshieldmachinecrossing
throughpilefoundationsof roadbridge
Q.W. Xu
Shanghai University of Science and Technology, Shanghai, P.R. China
Shanghai Jiaotong University, Shanghai, P.R. China
X.F. Ma
Z.Z. Ma
Shanghai Shentong Metro Co. Ltd., Shanghai, P.R. China
ABSTRACT: Shanghais subway construction has entered a new period of rapid development. A notable
characteristic of subway network construction is that the construction environment is increasingly complex.
TakeShanghai subwayline10asanexample, inthepositionof ShajinggangBridgeonSipingRoad, theshield
machinewill havetopassthroughthebridgespilefoundation. Sincetheembeddeddepthof pileshasalreadyrun
throughtheentiretunnel section, thesepileshavetobepulledout or cut off. But, asoneof themainarteriesof
Shanghai city, Si PingRoadhasbigtrafficflow, anditdoesnotallowthetrafficvolumetobeaffectedduringthe
constructionperiod. Whatsmore, thereareintensiveresidential quartersaroundthebridgeandtheconstruction
spaceislimited. Therefore, thispaper will mainlyintroducehowthepileunderpinningtechnologycanbeused
inthesubwayconstructionof Shanghai areafor thefirst time.
1 INTRODUCTION
Alongwiththecontinuousconstructionof metrolines
incitiesandunceasingimprovementof metronetwork,
there are more and more cases of new tunnel lines
passingthrougholdnearbybuildings
[1][3]
.
Inthepast20years,Shanghai hasbeenexperiencing
theunprecedented climax of city infrastructurecon-
struction. Therearetotal 17linesplanedintheurban
rail transitnetwork, of which11linesaresubwaywith
length of 385km, presently, 5 lines havebeen com-
pleted, 5linesarebeingconstructed. By2010, total 10
lineswill beput intousewithlengthof 250km.
A notable feature of the new period of metro
construction is that the construction environment is
increasinglycomplex, that is, examplesof intercross-
ing between newlineand old line, crossing through
various existing structures etc., are becoming more
andmore.TakeShanghai subwayline10asexample, it
spansfromHongqiaoAirport toNewJ iangwanTown
withtotal lengthof 28.8km, whichformsaconvenient
channel betweenthenorthandthewestpartof Shang-
hai city. Duringitsconstructionprocess, themainpile
foundationsthatimpedetheadvanceof shieldmachine
arelocatedatShajinggangBridge, ZoumatangBridge,
R4 Line
M7 Line
R2 Line
M4 Line
R1 Line
M7 Line
M3 Line
Zou Matang
Sha Jinggang
Huangxing road
Chengjia Bridge
Suzhou River
Huangpujiang River
M10 Line
Figure1. Thegeneral pictureof subwayLine10.
Chenjiaqiao Bridge and Huangxin Road Bridge, as
showninFigure1.
Regarding to the situation of new line crossing
through existing structures, generally speaking, the
line should be chosen to avoid existing buildings in
thestageof planning. Infact, therearevarious prac-
tical situations often make it difficult to avoid such
477
structures. As to the examples of tunnel crossing
thebridgefoundationmentionedinthis paper, when
the line can not avoid existing pile foundation, the
general construction method is: construct temporary
alternativebridgechangetrafficcourse, removethe
old bridge, extract piledrive the shield machine
forwardconstructanewbridgerestoretheorig-
inal trafficdemolishthetemporarybridge.
If follow this method, it results in long construc-
tionperiod, highconstructioncost, andgreat impact
on the society. But if we adopt the pile underpin-
ning technology, that is, based on the premise of
keeping the existing structure can be worked nor-
mally, while a series of construction technology are
used to extract or truncate piles, and finally reach
thepurposeof driving shield machineforward. This
cannot onlyshortenconstructionperiod, reducecon-
structioncost, minimizeitsimpactonthecommunity,
but alsocanpromoteShanghaissubwayconstruction
level.
2 PROJ ECT OVERVIEW
2.1 Status of Shajinggang Bridge
According to theplan of Shanghai subway Line10,
theinterval tunnel fromLiyangRoadtoQuyangRoad
will cross through the piles of Shajinggang Bridge
on Siping Road. Thebridgeis asimplebeamstruc-
turewiththreespans; itsspansare6m, 12mand6m
respectively. This bridgeincludes two piers and two
abutments. Each pier uses 23 quadratic reinforced
concrete piles as its foundation, with dimension of
40040026000mm; whileeachabutmentuses14
quadratic reinforcedconcretepiles as its foundation,
withdimensionof 40040027000mm. Sincethe
elevation of tunnel vault beneath thebridgeis about
67m, in theprocess of shield drive, thereare
about 4pilesat eachpier and34pilesat eachabut-
ment will affect tunneling excavation, as shown in
Figure2and3.
Around Shajinggang Bridge, there are intensive
buildings and underground pipelines, which should
betakenintoaccountduringtheconstructionprocess.
Theimportantbuildingsnearthebridgeareahigh-rise
buildinginthesouthwest, asinglestorypumpstation
in thesouth, a24-storey building in thesoutheast, a
reinforcedconcreteandbrickbuildingwith4floorsin
thenortheast, thepit of Quyang Road Station in the
northandtheHuaxi Stock ExchangeBuildinginthe
northwest. Theimportantundergroundpipelinesarea
groupof power cable, awater supplypipewithdiame-
ter of 1500mmandagroupof telephonecablestothe
eastof thebridge; whiletwogaspipeswithdiameterof
300mmand700mmrespectively, awater supplypipe
withdiameter of 300mmandtwogroupsof telephone
cablestothewest of thebridge.
3
0
0
0
1
9
0
0

2
1
5
0
3
2
1
0
0
1
5
5
3
0
1
5
5
3
0
2
7
3
0
3
0
0
0
1
9
0
0
1
1
2
2
3
0
=
2
4
5
3
0
1
1
2
2
3
0
=
2
4
5
3
0
2
1
5
0
3
2
1
0
0
15
1
5
8
4
0
1
5
5
9
0
1
5
8
4
0
1
5
5
9
0
3
1
0
3
1
0
Figure2. Planviewof tunnel crossingbridgefoundation.
26900
410
20
6000
40
6980 6980
40
6000
20
410
14-400

400 27000
quadraticprecastpile
23-400 400 26000
quadraticprecastpile
13688
EPB shieldmachine
olive~greysilt
olivedrab~grey sandysilt
sapgreen~olivedrabgreysiltclay
greysiltclay
greycalysoil
greysiltsoil
greysandysilt
greyclay silt
backfilling
-32.69
-25.24
-21.04
-17.24
-13.24
-10.04
-7.69
+1.16
-22.180
-22.700
4.020
2.500
5.940
5.000
6.000
toChifengRoad toTianshui Road
7
7
6
5
5
4
2
2
1
3-1
3-2
1-1
1-2
1
2
Figure 3. Elevation view of tunnel crossing bridge
foundation.
2.2 The geological conditions of the project
Accordingtoengineeringgeological surveyreport, the
geological conditions at Shajinggang Bridge can be
divided into the following layers. The first layer is
miscellaneous backfillingwiththickness from1.9to
3.7m, containingcinder, stone, etc. Thesecondlayer
is grey and yellow clayey mingled with 1.0mthick
blackhumusinit. Thethirdlayer isgreysilt soil con-
tainingclay init. It belongs to mediumcompression
soil, about 10mthick. The fourth layer is grey clay
with sand body trapped in it. It is high compression
soil, including whitestiff block and black humus in
it, about 512mthick. Thefifthlayer isdark green
hard clayey. It belongs to mediumcompressed soil,
about 2.42.8mthick. Thesixthlayer is olivedrab
hardclay, whichismediumcompressionsoil also.The
seventhlayer isolivedrabsilt soil.
2.3 Existing problems and solutions
In this project, theuniqueness and difficulties is not
only reflected on complicated site topography and
moreintensivesurroundingbuildings, but alsocanbe
seenfromthefollowingaspects.
478
1. SinceQuyangRoadsubwaystationisverynear to
thebridgepilefoundation, coupledwiththerestric-
tionof lineslopeinlongitudinal section, thetunnel
cannot crossbeneaththepilefoundationdirectly.
2. Since the outer diameter of shield machine is
approximately7m, inthecrossingprocess, 4piles
ateachpier and3pilesateachabutmentshouldbe
removed.
3. SinceSipingRoadisoneof mainroadsof Shanghai
city, itstrafficflowisbigat dayandnight, it isnot
allowedthatthetrafficvolumetobeaffectedduring
theconstructionperiod.
In view of the above considerations, under the
premiseof existingtraffic not tobeaffected, thepro-
posedconstructionstrategiesinclude: adoptingpile
foundationunderpinningtechnology at eachpier and
abutment; removing piles affecting shield tunnel-
ing. SinceShanghai has not yet hadtheprecedent of
pileunder-pinninginsubwayconstruction, itisneces-
sarytoconductseriescomprehensivetechnical studies
onthistechnologyanditsaccessorymethod, thuspro-
videacommontechnical guidancefor futuresimilar
projects.
3 ACCESSORY CONSTRUCTIONMETHODS
INPILE UNDERPINNING
3.1 Methods of foundation reinforcement
Thepurposeof foundationreinforcement istoensure
the stability of existing structures, such as pier and
abutment, andtoprovidestablefoundation.
Themainmethodsof foundationreinforcementcan
beconcludedas:
1. Grouting method: themain roleof grouting is to
seal water, but it isnot applicabletocohesivesoil.
2. Deep mixing method: it can ensure founda-
tion intensity and its sealing characteristics. This
method includes two categories, that is, vertical
construction method and transverse construction
method.
3. Freezing method: it can guarantee foundation
strength and sealing effect; but it is difficult to
resolvethegrounddeformationwhendefrosting.
3.2 Method of removing existing piles
Therearetwo methods, that is, extractingpiles from
groundsurfaceanddemolishingpileswithinlayers.
(1) Extractingpilesfromgroundsurface
Deep excavation method: it is very simple, cheap
and the required operation space is small also. The
mainprocedureisshowninFigure4.
Extractingpileswithcasingpipe:accordingtosome
overseas related experiences, this method includes
processes of extracting piles with casing pipe and
Casing pipe Pile
Cutted pile sector
Portal-type frame
Figure4. Sketchmapof deepexcavation.
(a) Uplifting the pile (b) Cutting the pile (c) Removing the pile
casing pipe
pile
steel bar
cutting position
chain
truncated pile
Figure5. Sketchmapof extractingpilewithcasingpipe.
cuttingpiles, onlyneeding56mconstructionspace
inheight, asshowninFigure5.
(2) Demolishingpileswithinlayers
This method can also be divided into two types:
demolishing piles manually inside the chamber and
cuttingpiles directly by shieldmachine. For thefirst
type,itshouldreinforcethefoundationfirstly,andthen
removepilesmanuallyinthechamber. For thesakeof
safety, compressedair canbeusedas supplementary
measuretostabilizetheexcavationface.Forthesecond
type, piles can be cut directly by the special device
installedonthecutter disc of shieldmachinewithout
humanoperation. But, theshieldmachineusedhereis
relativelymoreexpensive.
4 PROPOSAL FOR SHAJ INGGANGBRIDGE
Based on overseas experience and some successful
instances[4], twofeasibleschemesareprovidedhere.
A scheme is reinforcing foundation and eliminating
pilesinsidetheshieldchamber,whileBschemeisrein-
forcing foundation, adding new piles and extracting
pileswithcasingpipe.
4.1 A scheme
(1) Foundation reinforcement and construction of
retainingwall
Foundation reinforcement and construction of
retainingwall istoformoperationspacewhichisused
toreinforcebridgestructure, asshowninFigure6and
7. Sincetheriver flows perennially under thebridge,
cofferdamshouldbebuilt firstly.
(2) Pit excavationbehindtheabutment
Sincethefoundationof pier andabutment cannot
be reinforced on road surface, so it is necessary to
479
foundation reinforcement behind abutment
retaining wall
foundation reinforcement behind retaining wall
Figure6. Elevationviewof foundationreinforcement.
foundation reinforcement behind abutment
foundation reinforcementbehind abutment
retaining wall
Figure7. Planviewof foundationreinforcement.
4m
excavation
behind
abutment
steel plate
Figure8. Elevationviewof pitexcavationbehindabutment.
digpit betweentheabutment andretainingwall. It is
noteworthy that the excavated depth below the road
surfaceshouldbenolessthan4meterssoastorein-
forcethefoundation smoothly. After excavation, the
pit top should be covered with steel plate to ensure
that thegroundtraffic will not beaffected, as shown
inFigure8.
(3) Foundationreinforcement of pier andabutment
Oncethefoundationof pier andabutmenthadbeen
reinforced, as shown in Figure 9 to 11, it can not
foundation reinforcement outside the tunnel
foundation reinforcement
of pier and abutment
backfilling
Figure 9. Elevation view of structure foundation
reinforcement.
foundation reinforcement of pier
backfilling
Figure10. Foundationreinforcement of pier.
backfilling
foundation reinforcenment of abutment
Figure11. Foundationreinforcement of abutment.
only transfer theload borneby theremoved piles to
thenewaddedpilesandtheremainingpiles, but also
integrate the foundation and piles as a whole body.
In addition, thetunnel lining located hereshould be
redesignedashighstiffnesssegment towithstandthe
loadtransferredfromtheremovedpiles. Considering
480
foundation reinforcement outside the tunnel
Structure reinforcement of pier and abutment
Figure12. Elevationviewof reinforcingpierandabutment.
foundation reinforcement
outside thetunnel
concrete layer
high stiffness segment
compressured air method
Figure 13. Demolishing piles and propelling shield
machine.
thesecurityof manuallyoperationandstabilityof tun-
nelingface, thefoundationoutsidethetunnel should
also bereinforcedfirstly, andthereinforcedscopeis
indicatedbythebluelineinfollowingfigures.
(4) Structurereinforcement of pier andabutment
The purposes of reinforcing pier and abutment
structureistotransfer theloadborneby theremoved
pilestothereinforcedfoundationandremainderpiles,
and improve the capacity to resist shear force and
moment perpendicular to theaxial of pier and abut-
ment. Duringreinforcement, steel beamcanbeused
tosupport thestructure, asshowninFigure12.
(5) Demolishing piles and propelling forward the
shieldmachine
After all thereinforcement work hadbeendone, it
will beableto demolishpiles, propel forwardshield
machineandassemblesegments subsequently. Since
theworkof demolishingpilesisdoneinsidethecham-
ber,theexcavatedsoil initshouldbedischargedfirstly.
Thereafter, tunnelingfaceisintheunsupportedstatus,
socompressedaircanbeusedtostabilizetheface.The
wholesketchmapisshowninFigure13. If necessary,
highstiffnesssegment may beusedwithinthewhole
add new piles
Figure14. Elevationviewof foundationreinforcement.
lengthbelowthebridge, whichisindicatedbythepur-
plelineinthefigures. Meanwhile, thedeformationof
pier andabutment shouldbemonitoredstrictly.
(6) Maintainceof steel beamandrestorationof road
surface
To prevent the corrosion of steel beam, it should
becoatedwithalayer of concrete. After theobstacle
piles had been removed and theshield machinehad
passedby, inorder toguaranteeregular transportation
running, itisnecessarytodemolishtheretainingwall,
backfill thepit andrestoretheroadsurface. But, dur-
ingtheseprocesses, moreattentionshouldbepaidto
monitor thedeformationof structure.
4.2 B scheme
(1) Foundationreinforcementandtheconstructionof
retainingwall
Foundation reinforcement and theconstruction of
retaining wall are the same as that inA scheme, as
showninFigure6and7.
(2) Pit excavationbehindtheabutment
Thisprocessisalsoidentical withthatinA scheme,
asshowninFigure8.
(3) Foundation reinforcement of pier and abutment
andtheconstructionof addingnewpiles
ComparewiththatinAscheme,therearetwodiffer-
encesinthisprocessaltogether, first isthereinforced
scope, second is adding newpiles, as shown in Fig-
ure14to17. After thefoundationreinforcement, the
foundation at thetopsidearound thepiles should be
excavatedtoformworkingspaceforaddingnewpiles.
Therefore, therequired vertical clearanceunder pier
andabutment shouldbenot lessthan3.5meters.
(4) Reinforcement of pier andabutment structure
Reinforcement of pier and abutment can better
improveits capability to resist exterior loads. There-
fore, steel beamcanbeusedtosupport thestructure;
481
new piles
Figure15. Planviewof foundationreinforcement.
backfilling
Figure16. Foundationreinforcement of pier.
backfilling
foundation reinforcement ofabutment
Figure17. Foundationreinforcement of abutment.
themainprocess is identical to that inA scheme, as
showninFigure18.
(5) Foundationexcavationandpileextraction
The purpose of further excavation of foundation
belowthebridgeistoensurethattheworkingspacefor
extractingpilecanbesatisfied. But, onlythoseplaces,
wherethepilesneedtobeextracted, shouldbeexca-
vateddown.Sincecasingpipemethodisusedtoextract
steel beam
pit excavation behind abutment
steel cover ploate
new piles
existing piles
foundation
reinforcement
retaining wall
foundation
reinforcement
Figure 18. Reinforcing bridge structure and adding new
piles.
5m pit excavation, pile extraction
Figure19. Further excavationof foundation.
removed piles
backfilling
Figure20. Extractingpilewithcasingpipeat pier.
piles, therequiredvertical clearanceunderthebearing
platformshouldbenolessthan5m, asshowninFig-
ure19to21.Duringtheconstructionof extractingpile,
moreattentionshouldbepaidonthemonitoringof pier
andabutment, especiallyfor structuredeformation.
(6) Propellingshieldmachineandinstallingsegments
After the impeded piles had been removed, the
shield machine can then move forward and pass
482
backfilling
removed piles foundation reinforcement of abutment
Figure21. Extractingpilewithcasingpipeat abutment.
backfilling
demolishing the retaining wall, restoring
the ground surfacement
Figure22. Backfillingandrestorationof roadsurface.
throughbridgefoundation, followedbyinstallationof
segments. During the crossing process, deformation
of pier and abutment structure should be monitored
primarily.
(7) Backfillingandrestorationof roadsurface
Once the shield machine had passed through the
pilefoundation, itisnecessarytodemolishtheretain-
ingwall, backfill thepit andrestoretheroadsurface
to guaranteetheregular traffic running, as shown in
Figure 22. During such process, it is important to
put emphasis on the deformation monitor of bridge
structure.
5 CONCLUSIONS
Combined with thepractical situations of in Shang-
hai areaandtheexampleof shieldmachinecrossing
through thepilefoundations of Shajinggang Bridge,
this paper mainly study how theunderpinning tech-
nology can be used in subway construction. Based
onconcreteanalytical studies, twomainconstruction
schemesarepresentedhere. Oneisreinforcingfoun-
dationandeliminatingpiles insidethechamber; the
other isreinforcingfoundation, addingnewpilesand
extractingpileswithcasingpipe.Sincethepileunder-
pinningtechnologyhasnever beenusedinthesubway
constructionof Shanghai area, thetwo schemes pro-
videdinthispapercannotonlybeusedasreferenceof
designandconstruction, also as precious experience
for futuresimilar projects.
REFERENCES
Ishimura, T., Metoki, M. & Shimizu, M. 2006. Develop-
mentof removedpilemethodwithcutting.Tunnelling and
Underground Space Technology 21: 411412.
Iwasaki,Y., Watanabe, H., Fukuda, M., Hirata, A. & Hori,Y.
1994. Construction control for underpinning piles and
their behavior during excavation. Geotechnique 44(4):
681689.
Sharma, J.S., Hefny, A.M., Zhao, J. & Chan, C.W. 2001.
Effect of large excavation on deformation of adjacent
MRT tunnels. Tunnelling and Underground Space Tech-
nology 16: 9398.
Yamaguchi, I., Yamazaki, I. & Kiritani, Y. 1998. Study of
ground-tunnel interactionsof four shieldtunnelsdrivein
close proximity, in relation to design and construction
of parallel tunnels. Tunnelling and Underground Space
Technology Vol.13, No.3: 289304.
483
Characteristicsof tunneling-inducedgroundsettlement ingroundwater
drawdownenvironment
C.Yoo& S.B. Kim
Sungkyunkwan University, Suwon, Korea
Y.J. Lee
RIST, Yongin, Korea
ABSTRACT: This paper presents theresults of aninvestigationonthecharacteristics of tunnelling-induced
ground settlement in groundwater drawdown environment. Thedynamics of theeffect of groundwater draw-
downonthegroundsettlementsarefirst investigatedusingacasehistoryconcerningaconventional tunnelling
situationinwhichtheinteractionbetweenthetunnellingandthegroundwater inducedexcessivegroundsurface
settlements. A 2D stress-porepressurecoupledfiniteelement analysis is thenconductedonatunnellingcase
withgroundwater drawdown, aimingatinvestigatinggroundsurfacesettlementcharacteristics. Theresultsindi-
catedamongother thingsthatsignificantportionof groundsettlementcanoccur beforetunnel facereaches, and
that theerror functionapproachdoesnot provideagoodfit tothesettlement troughsfor tunnellingcaseswith
groundwater drawdown.
1 INTRODUCTION
Tunnelling beneath the groundwater table causes
changesinthestateof stressandtheporewater pres-
sure distribution. In such tunnelling problems, the
tunnelling work inevitably causes water inflows into
excavated area, thus causing the change in the pore
water pressure distribution. The direct environmen-
tal consequenceof water inflowsduringtunnellingis
thedrawdown of groundwater level in thesurround-
ingaquifer(Yoo2005).Therelatedgroundsubsidence
occurringasaresultof thereductioninwaterpressures
inthesoil layerscandamagenearbystructures/utilities
(Figure 1). Oneof themajor casehistories illustrat-
ingdamageduetogroundsettlement associatedwith
tunneling-inducedgroundwater drawnis perhaps the
Romeriksporten tunnel in which the highspeed rail-
way tunnel construction caused more1mof ground
subsidenceduetogroundwaterdrawdown, raisingsig-
nificant technical and political issues pertaining to
theeffect of tunnelling on surrounding environment
(NSREA 1995).
Surprisinglystudiesconcernedwiththetunneling-
inducedgroundmovementsingroundwaterdrawdown
are scarce as indicated by Yoo (2005). Although a
number of studies on the ground subsidence caused
by groundwater pumping froman aquifer havebeen
conducted (Shen et al. 2006; Qiao & Liu 2006;
U
GWT
Reduction in
pore pressure
Groundwater
lowering
Ground
surface
Figure1. Illustrationof groundsettlement associatedwith
tunnelling-inducedgroundwater drawdown.
Xu et al. 2006). The results of their studies can-
not be directly applied to the tunnel excavation
problems as they focused only on the groundwater
drawdown due to groundwater pumping. As urban
tunnelling projects tend to involve potential prob-
lemsrelatedtogroundwater drawdowngroundmove-
ments during tunnelling, thereis an urgent need for
better understanding on the mechanisms involved
in tunnelling-induced ground movements associated
withgroundwater drawdown.
This paper presents theresults of investigationon
thecharacteristics of tunnelling-induced ground set-
tlement ingroundwater drawdownenvironment. The
485
Figure2. Typical groundprofile.
Table1. Geotechnical propertiesof soil/rocklayers.
c
E k
Type (kN/m
3
) (kPa) (deg) (MPa) (cm/sec)
fill 18 0 27 5 0.40 3.810
4
alluvial 20 15 30 10 0.40 3.810
4
weathered 25 15 30 50 0.33 2.410
4
soil
weathered 25 60 35 120 0.30 8.810
5
rock
hardrock 26 100 35 200 0.25 5.010
5
Note: =unit weight, c
=cohesion, =internal fric-

tion angle, E =youngs modulus, =poissons ratio,
K =coefficient of permeability
dynamics of theeffect of groundwater drawdownon
the ground settlements are first investigated using a
casehistoryconcerningaconventional tunnellingsit-
uationinwhichtheinteractionbetweenthetunnelling
and the groundwater induced excessive ground sur-
facesettlements.A parametricstudyusingacalibrated
2Dstress-porepressurecoupledfiniteelement model
is then conducted on a number of factors influenc-
ing the tunnelling-induced ground settlements with
groundwaterdrawdown.Basedontheresults,theinter-
actionmechanismbetweenthetunneling,groundwater
lowering, andgroundsettlement isidentified.
2 GROUNDSURFACE SETTLEMENT
CHARACTERISTICS FIELDMONITORING
DATA
2.1 Tunnelling condition
A casehistoryconcerningtheconventional tunnelling,
i.e.,NATM,wasconsidered.Thetunnel hasexcavation
widthandheight of approximately 10.5mand8.7m,
respectively, withacoverdepthrangingapproximately
2030m, andconstructedinamulti-layeredground
including a fill, alluvium, and a weathered zone as
illustratedinFigure2. Thegeotechnical propertiesof
thegroundaregiveninTable1.
Figure3. Support pattern(typical).
2.2 Tunnel design
Figure3shows atypical tunnel support patternused
for a100mlongsectionfor thegroundprofilegiven
in Figure2. On account of thedifficult ground con-
ditiontheringcut excavationmethodwasadoptedto
promote the tunnel face stability during excavation.
Theprimarysupportsystemconsistedof a0.2mthick
steel fibrereinforcedshotcrete(SFRS) layer with4m
longsystemrock boltsat 1.0and1.2m, respectively,
longitudinal andtransversespacing.Thepipeumbrella
techniqueusing800mmdiameter groutinjected12m
longsteel pipeswasadditionallyimplementedtopro-
mote the face stability through improving the load
carryingcapacityof thegroundaheadof theface.Also
adopted was atrumphet shaped micro cement injec-
tion(MSG) pre-groutingaroundthetunnel periphery
to createa 5mthick watertight shell for sections in
which the weathered soil layer extended to the tun-
nel crown level. The pre-grouting scheme was later
extendedtocoverthefaceafterthesettlementproblem
hadbecomeanissue.
2.3 Measured ground surface settlements
Figure 4 shows the progressive development of set-
tlements during the tunnel advancement at various
monitoringstations. Inthisfigurethemeasuredsettle-
mentsareplottedagainsttherelativedistancebetween
the tunnel face and the monitoring stations normal-
izedbythetunnel diameter (D). Thesedata, measured
usingtheconventional levelingtechnique, thusindeed
represent thesettlement history duringthetunneling
processfor themonitoringstations.
A total of fivesettlement curves arepresented in
thisfigure.Ascanbeseeninthisfigurethefivecurves
aresimilar bothinqualitativeandquantitativeterms,
despitethevertical extent of thedecomposedsoil rel-
ative to the tunnel varies along the route, showing
themaximumconverged settlements in therangeof
486
Figure 4. Progressive development surface settlements at
variousmonitoringstations.
Figure5. Normalizedsettlement historycurves.
1.6%1.8%D. Of salientfeaturethatcanbeobserved
in this figure is the tendency of settlement increase
duringwhichthetunnel advancementwashalted, sug-
gesting time dependent (tunneling activity indepen-
dent) settlement development (to bediscussedlater).
Another of interest trend is the resemblance of the
settlementcurveswithatypical logt curvefromacon-
solidationtest, beingcharacterizedby threezones as
aninitial compression, primary, andasecondaryzone.
Suchatrendstronglysuggestsapossiblecausebeing
the volume change effect due to tunneling-induced
groundwater drawdown.
Thedatain Figure4 arefurther analyzed by nor-
malizing the settlement values with their respective
maximumvalues (S
v,max
) inFigure5. As seeninthis
figure, thenormalizedcurvestendtocollapseintoone
curve. A further inspection of thenormalized settle-
menthistorycurvesshowsthatthesettlementsstarted
todevelopwhenthetunnel facewasapproximately6D
away fromthe monitoring stations. The settlements
tend to accelerate when the tunnel face reached 3D
awayfromthemonitoringstations,anddecelerateafter
the tunnel advanced 56D beyond the monitoring
stations.Alsoshownarethatapproximately6070%
of thefinal settlement (S
v,max
) was completedbefore
thefacepassedamonitoringstationwiththeremain-
ing 3040% occurred after the full passage of the
tunnel face. Suchapercentageof settlement aheadof
thefaceisconsiderablylargerthanthetypical valueof
4050%,suggestingalargerportionof theconverged
settlement occurredprior tothearrival of tunnel face
inthistunnelingconditionthanatunnelingcondition
withoutthegroundwaterdrawdown.Moreover,theset-
tlementstendtoconvergetoaconstant valueafter the
tunnel faceadvancedtoadistanceof 67D beyond
themonitoringstations, suggestingslower settlement
convergencethananormal condition.
Theseresults in fact aresomewhat different from
typical trends that canbeobservedintunnelingcon-
ditionswithoutsignificantgroundwater lowering, and
ledtoaconclusionthatfactorsotherthantheunloading
effect duetothetunnel excavationmayhaveplayeda
role. Suchatendencyisdirectlylinkedtotheground-
water drawdown as will be shown in a subsequent
chapter.
3 PARAMETRIC STUDY
3.1 Stress-pore pressure coupled analysis
A commercial finite element package ABAQUS
(Abaqus, Inc. 2002) wasusedfortheparametricstudy.
A 2Dstress-porepressurecoupledeffectiveformula-
tion was adopted in order to realistically capturethe
interactionmechanismbetweenthetunnellingandthe
groundwater.
In ABAQUS a porous mediumis approximately
modelled by attaching the finite element mesh to
thesolid phase. Equilibriumis expressed by writing
the principle of virtual work for the volume under
considerationinitscurrent configurationat timet:
wheredv is avirtual velocity field, d is thevirtual
rateof deformation, isthetrue(Cauchy) stress, t are
surfacetractions per unit area, andf body forces per
unitvolume. f includestheweightof thewettingliquid
f
w
definedasEq. (2)
inwhichs isthedegreeof saturation, n istheporosity,
andn
t
isthevolumeof trappedwettingliquidper unit
of current volume. Eq. (1) canthenberewrittenas
487
Table2. Geotechnical properties.
E c k
(kN/m
3
) (kPa) (kPa) (
) (cm/sec)
soil 25 50,000 30 30 5.810
3
Weathered 25 100,000 50 30 1.310
4
rock
wheref areall body forces except theweight of the
wettingliquid.
Thecontinuity equationissatisfiedapproximately
in thefiniteelement model by using excess wetting
liquidpressureas thenodal variable(degreeof free-
dom8), interpolatedover theelements. Thebackward
Euler approximation is used to integrate the equa-
tion over timeandtheNewton iterations areusedto
solvethenonlinear, coupled, equilibriumandcontinu-
ityequations.Fundamentalsof thestress-porepressure
coupled formulation adopted in ABAQUS can be
found in theABAQUS users manual (Abaqus, Inc.
2005).
3.2 Condition analyzed
A tunnelling condition frequently encountered in
urban situations was considered in theanalysis. The
tunnel consideredisa10mdiameterhorseshoeshaped
tunnel with acover depth of 3.0D, excavated by the
benchcut method. Theprimary support systemcon-
sists of a 20cmthick shotcrete lining with system
rock boltsinstalledat 1.5mcenter-to-center spacing.
A 1.5D thick soil layer was assumedtoexist abovea
weatheredrocklayerthroughwhichthetunnel isexca-
vated.Tables2summarizesgeotechnical propertiesof
theground.
3.3 Finite element model
Figure6 shows thefiniteelement model adopted in
thisstudy. Thefinite-elementmeshextendstoadepth
of two times thetunnel diameter (D) belowthetun-
nel springlineandlaterallytoadistanceof 15D from
thetunnel centerdependingonthecoverdepthH.The
lateral locationwasselectedbasedonaseriesof pre-
liminary analysis as it has asignificant influenceon
theresultsof astress-porepressurecoupledanalysis.
Atthelateral boundarydisplacementsperpendicularto
theboundaries wererestrainedwhereas pinsupports
wereappliedtothebottomboundary.
With regard to thehydraulic boundary conditions
and with reference to Figure 6, a no-flow condition
wasassignedtothevertical boundariesperpendicular
tothetunnel drive.Atthelateral vertical boundarythe
groundwater table was assumed to be at the ground
surfaceandconstant throughout theanalysis.
Figure6. Finite-element model usedintheanalysis.
Thegroundandtheshotcreteliningwerediscretized
using8-nodedisplacementandporepressureelements
with reduced integration (CPE8RP). The rock bolts
weremodeledusingthe2-nodetruss elements. With
regard to the material modelling, the soil and rock
layers wereassumed to bean elasto-plastic material
conforming to the Mohr-Coulomb failure criterion
together withthenonassociatedflowruleproposedby
Davis (1968), whiletheshotcreteliningandtherock
boltswereassumedtobehaveinalinear elasticman-
ner. Thegeotechnical propertiesfor thegroundgiven
inTable1wereusedfor analysis. Theyoungsmoduli
of theshotcreteandrockboltswerechosenas15GPa
and21GPa, respectively.
Theactual tunnellingprocessconsistingof aseries
of excavation and support installation stages was
closelysimulatedintheanalysisbyaddingandremov-
ingcorrespondingelementsat designatedsteps. After
establishingtheinitial stressandporepressurecondi-
tionswithappropriateboundary conditions, thestep-
by-steptunnellingprocess pertinent to thebenchcut
excavationmethod,wasthensimulated.The3Deffects
of advancingatunnel headingwastakenintoconsider-
ationusingthestressrelaxationmethodinwhichthe
boundary stresses arising fromtheremoval of exca-
vated elements areprogressively applied to simulate
theprogressivereleaseof theexcavationforcesasthe
tunnel headingadvances.
3.4 Ground settlement characteristics
Figure7 presents therelationship between themax-
imumsurface settlement (S
v,max
), directly above the
tunnel crown, obtainedduringvarious stages of tun-
neling. As seen the maximum surface settlement
S
v,max
tends to linearly increasewith theincreasein
thegroundwater drawdownlevel H
D
. Thesettlement
occurred after the completion of tunnel is in fact
twicethat during thetunnel excavation. It should be
notedthat theplot giveninthisfigurerepresent those
causedbythegroundwater inflowintothetunnel after
the completion of tunnel excavation, until a steady
stateconditionisachieved. Thissuggestsadirect link
between the ground settlement and the groundwater
drawdown, thusdemonstratingtheimportanceof cre-
atingawatertight shell for tunnellingcaseswherethe
controllinggroundsurfacesettlement isof concern.
488
Figure7. Variation of S
v,max
with H
D
after completion of
tunnel excavation.
Figure8. Normalizedsurfacesettlement troughs.
In urban tunnelling situations, characteristics of
a ground surface settlement trough, such as slope
and width of inflection point, areimportant as they,
together withlateral displacements, determinepoten-
tial for damageof adjacent structures. Figure8shows
normalizedgroundsurfacesettlementtroughsfor dif-
ferent groundwater drawdown levels together with
error functions constructed using different values of
inflectionpoint. Of interest trendsaretwofold. First,
theextentof groundsettlementtroughissignificantly
greater than typical tunnelling conditions without
groundwater drawdown. Infact, for theparticular tun-
nelling condition considered, the ground settlement
zoneextendsmorethan10timesthetunnel diameter
fromthetunnel centerline. Second, theerror function
approach(Attewell et al. 1986; Peck 1969) knownto
well describe the surface settlement trough for tun-
nelling cases without ground water drawdown does
not provide a good fit to the settlement troughs for
caseswithgroundwaterdrawdown.Anotherimportant
observation is that the computed settlement troughs
Figure9. Normalizedsurfacehorizontal displacementpro-
files.
tendtocollapseintoonecurvedespitesomediscrep-
anciesintheregionfather away, i.e., 4D, fromthe
tunnel center.
Normalized horizontal displacement (S
h
) profiles
areshowninFigure9for different levels of ground-
water drawdown. As seen, the maximumhorizontal
displacements (S
h,max
) tend to develop at locations
3D away from the tunnel center with decreases in
magnitudes thereafter. Again the S
h
profiles tend to
collapseinto onecurvealthoughsomediscrepancies
areobservedintheregionawayfromthetunnel center.
Theresultsobtainedinthisstudysuggest that theset-
tlement troughaswell asthehorizontal displacement
profilemay beconstructed using normalized curves
when relationships between the surface movements
andother factorscanbeestablished.
4 CONCLUSIONS
This paper presents theresults of aninvestigationon
thecharacteristics of tunnelling-induced ground set-
tlement ingroundwater drawdownenvironment using
the measured surface settlement for a site where
thetunnelling-inducedgroundwaterdrawdowncaused
significant surfacesettlement. A stress-porepressure
coupled finite element model was additionally con-
ducted aiming at identifying the ground movement
characteristics when tunnelling induces asignificant
level of groundwater drawdown. Basedontheresults
thefollowingconclusionscanbedrawn.
1. For tunnelling cases in which tunnel excavation
causessignificantgroundwaterdrawdown, theper-
centage of settlement that develop prior to the
tunnel facearrival tothefinal settlement issignif-
icantly larger than for cases without groundwater
drawdown.
489
2. Continued groundwater drawdown after thecom-
pletionof tunnel excavationmay cancausesettle-
ment larger thanthat occur duringexcavation.
3. Theerror function does not provideagood fit to
thesettlement troughsfor caseswithgroundwater
drawdown.
4. Normalizationcanholdfor thesurfacesettlement
andhorizontal displacementprofilesfortunnelling
caseswithgroundwater drawdown.
ACKNOWLEDGEMENT
ThisresearchissupportedbyKoreaMinistryof Con-
structionandTransportationunder Grant No. C4-01.
Thefinancial support isgratefullyacknowledged.
REFERENCES
Abaqus users manual, Version6.5. 2005. Hibbitt, Karlsson,
andSorensen, Inc., Pawtucket, Providence, R.I.
Attewell, P.B., Yeates, J. & Selby, A.R. 1986. Grounddefor-
mationandstrainequations. Soil movements induced by
tunneling and their effects on pipeline and structures,
Blackie, Glasgow: 5366.
NorwegianSoil andRockEngineeringAssociation(NSREA).
1995. Norwegian urban tunnelling. PublicationNo. 10.,
Norway.
ground Proc., 7th Int. Conf. on Soil Mech. And Found.
Engrg. : 225290.
Qiao, S. & Liu, B. 2006. Predictionof grounddisplacement
and deformation induced by dewatering of groundwa-
ter. Underground Construction and Ground Movement:
7379.
Shen, S.L., TangC.P., Bai, Y. & Xu, Y.S. 2006. Analysis of
settlement dueto withdrawal of groundwater around an
unexcavated foundation pit. Underground Construction
and Ground Movement: 377384.
Xu, Y.S., Shen, S.L. & Bai, Y. 2006. State-of-art of land
subsidence prediction due to Groundwater Withdrawal
inChina. Underground Construction and Ground Move-
ment: 5865.
Yoo, C. 2005. Interaction between Tunnelling and
Groundwater-Numerical Investigation Using Three
Dimensional Stress-Pore Pressure Coupled Analysis.
Journal of Geotechnical and Geoenvironmental
490
Effect of long-termsettlement onlongitudinal mechanical performanceof
tunnel insoft soil
H.L. Zhao, X. Liu,Y.Yuan&Y. Chi
ABSTRACT: Thispaperintroducesthemethodforanalyzingsettlementdataof tunnel insoftsoil.Themethod
is appliedto thesettlement monitoringanalysis of sometunnels across theriver. In thecasestudy, thelong-
termmonitoringdatafor shieldtunnel settlement is discussedconcretely. Firstly, unitary longitudesettlement
curvesaregained, andthecorrespondinglongitudinal curvatureiscalculatedtoinvestigatetherulesof absolute
settlement andunevensettlement varyingfromtime. Secondly, basedontheunevensettlement magnitude, the
mechanical propertiesof shieldtunnel arediscussedbymeansof theequivalentaxisstiffnesscontinuousmodel
of shieldtunnel. Inthisway, therelationshipsbetweenlongitudinal curvatureandcrackof shieldtunnel segment
joint, thestress of tunnel concretesegment, and thestress of bolts for segment joint arefound generally. So
thecrack andleakageof shieldtunnel canbeinferredconsequently. Finally, theresultsgottenfromanalytical
solutionarecomparedwiththefieldinspectioninformation, andthefavorableagreement isshown. Analysisof
effect of long-termsettlement onlongitudinal mechanical performanceof tunnel insoft soil offersguidanceto
evaluatethesafetyandserviceability, structural protectionandwaterproof designof shieldtunnel insoft soil.
1 INTRODUCTION
Thedramaticlongitudinal settlementsoccurinsoftsoil
aftermanyyearsoperationbecauseof anumberof fac-
tors, including vibration induced by trains, leakage,
geotechnical conditions, and depth of embedment,
construction methods, surface loads and mutative
environment actions.
Duetothegreat spanof timeandhugeinspecting
cost, theinvestigationfortunnel long-timesettlements
hasbeenlimited. Byfar, therearefewspecial reports
for long-termsettlement research of the tunnel. All
of these factors restrict the development of safety
andserviceabilityevaluationfor tunnelsinlong-term
operatingprocess.
Thesettlement reflects thelongitudedeformation
of tunnels. Generally, thetunnel in soft soil is weak
relatively to resist deformationcausedby extraneous
factors, andthecrackof circumferential segmentseam
ispronetoappear,whichleadstoleakageandsegments
damage under longitudinal tension or compression.
Therefore, its crucial and significant to investigate
andanalyzesomerules for shieldtunnel suchas the
relationship of the segment inner force caused by
unevendeformation, crackof segmentjoint, andstress
of bolts.
Inthispaper, thetunnel longitudedeformationcur-
vatureisobtainedfrommonitoringdataof long-term
settlement. Then, therelationship between longitude
deformation curvature with the segment inner force
andcrack of segment joint areconcluded. Thispaper
aims at evaluating the performance of tunnel after
many years service according to the corresponding
longitudedeformationcurvature. Theanalysisresults
offer important foundational materials for structural
safetyanddurabilityevaluationof shieldtunnel.
2 LONG-TERM SETTLEMENT DATA
ANALYSISFOR SHIELDTUNNEL
Thispaperfocusesonahighwaytunnel acrosstheriver.
Thistunnel locatedinsoft soil hasservicedfor more
than30years.Tunnel settlementmonitoringhasbeing
executedintheservicestage. Inthesettlement moni-
toring, therearetwoelevationbenchmarksset inboth
endsof tunnel. Settlement monitoringpointsbetween
thetwobasepointsarearrangedinbothsidesof way
respectivelyalongtunnel length,soitisaclosedcircuit
stylefor inspection.
The investigation of this paper is based on the
settlement dataof thistunnel for recent tenyears.
491
Figure1. Longitudinal settlement curveof tunnel.
Figure2. Longitudinal curvatureof tunnel fromShaft No. 4toNo. 2
2.1 Longitudinal settlement curve
There is certain rigidity for the structure between
two adjacent monitoring points, and local structural
deformation will affect adjacent structural deforma-
tion and internal force. In order to reflect integral
tunnel longitudesettlement comprehensively, theset-
tlementanalysisshouldadoptunitaryanalysismethod.
A unitary longitude settlement curve can be drawn
based on monitoring results, shown as Figure 1. It
shouldbenotedthatmonitoringpointslocatedinboth
sidesof thistunnel. Sothesettlementsof thesemoni-
toringpointsaretermedassettlementsof leftlineand
right linerespectivelyinthispaper.
The varying rules of longitudinal settlement
obtained fromlongitudinal settlement curve can be
concludedasfollows.
1. Mostsettlementmagnitudeof observationpointsis
between30.0mmand30.0mm. Theannual set-
tlement curve is parallel approximately. It means
that the increase of settlements of observation
points is uniformand tunnel longitude deforma-
tion shows favorable integrity. The deformation
differenceislittlebetweenadjacentpoints,andthey
almost riseandfall together.
2. Thesettlement magnitudesarevariablefromtime.
Most settlement pointsoscillateupanddown. For
instance, theaccumulativelongitudinal settlement
in2004islargerthanthatin2002,butthesettlement
in 2006 is less than that in 2004. Some of these
settlementsincreasecontinuouslyintheoscillating
andsomeinclinetowardssteady.
2.2 The curvature of longitudinal settlement curve
Thecurvaturechangeis theimportant characteristic
of longitudinal settlementfor shieldtunnel, becauseit
demonstratestheunevensettlement of tunnel.
Thesettlementdatafromtunnel longitudeobserva-
tionpointsismadetosplineinterpolationfitting, and
then, basedontheinterpolationresult, thecurvature
is elicitedby numerical differential. Therefore, the
curvaturevarying fromlongitudinal distancecan be
drawn, shownasFigure2. Thecurvaturevariesfrom
tunnel longitudinal distanceandtheincreasedrangeof
curvatureislarge.Thisisinlinewiththeruleof tunnel
492
Figure3. Longitudinal curvatureof right andleft linefromShaft No. 2toNo. 1
longitude settlement development. But it should be
notedthatthecurvaturechangeof leftcurveisdifferent
fromright curve, soit will leadtolateral deformation
andlateral additional inner force.
1. The curvature of settlement curve for the points
fromNo. 4toNo. 2shaftissmall. Butthecurvature
where closed to shaft is larger than that between
shafts.
2. The curvature is very large fromNo. 2 to No. 1
shaft, soanother figureisdrawntodescribeit. The
averagevalueof curvatureis0.01mm-1, evenupto
0.021mm-1, shownas Figure3. So thestructural
safety anddurability of this part tunnel shouldbe
calculatedandevaluatedindetail.
3 SHIELDTUNNELINGLONGITUDINAL
STRUCTURE MODEL
At present, there are two main kind of theoretical
researchfortunnel longitudestructure. Oneisnumeri-
cal solvingmodel basedonthefinite-elementmethod.
Andtheother istheoretical solvingmodel.
J apanese scholars proposed two kinds of tunnel
longitudestructuretheorybasedonthedifferent sim-
plification for tunnel segments joints andbolts. One
isbeam-springmodel (Zhu, 1998); theother isequiv-
alent continual rigidity model (Shiba& Kawashima,
1988). Thestudyhereinadoptstheequivalent contin-
ual rigiditymodel proposedbyShibaandKawashima
(1988) toanalyzetunnel longitudeperformance. Con-
sidering the tunnel analyzed in this case study has
structural lateral deformation caused by long-term
operation, so a relative correction factor is intro-
duced to reduce the affection on tunnel longitude
deformationandinner force.
3.1 Equivalent rigidity continuous model
Equivalent rigidity continuous model is supposed to
be continuous lining rings and homogeneous in the
cross sectionof tunnel. Therings composeatubular
structureconnectedbylongitudinal steel boltstermed
as girth joint of tunnel segment. Thegirth joint bolt
is simulated to spring with different characteristic
in compression and tension. The simulate spring is
Figure4. Relationshipfor P
Figure 5. Stress distribution for segments in tension and
compression.
bilinear material intension, whileis rigidabsolutely
incompression. Therelationshipof P isshownas
Figure4.
Herein, k
j1
is elastic stiffness for individual bolt,
andk
j2
isplasticstiffnessfor individual bolt, P
y
isthe
elasticultimatepressurefor bolt.
The assumptions in equivalent continual rigidity
model areasfollows.
1. Circumferential inhomogeneityof liningringisnot
taken into consideration. Additionally, the shear
stress and deformation in longitudinal seamare
ignored.
2. Compressivepressureis resistedby concreteseg-
ment singly, and tensile pressure is resisted by
concretesegment insharewithsteel bolts.
3. Plane sections remain planeandsmall deformation
assumptionsareapplicableinthismodel.
Themajor resultsof equivalent rigiditycontinuous
model referencedwiththeanalysisof thispaper areas
follows.
493
Figure6. Stressdistributionfor segment inelasticstate.
3.2 Equivalent compressive and tensile rigidity
A segment with length l
s
is taken as a calculate
unit. Its stress distributionintensionor compression
can beseen in Figure5. Theequivalent axial tensile
andcompressiverigidity is gainedfromdeformation
compatibility andforcebalanceconditions, shownin
Equation1.
(EA)
C
eq
, (EA)
T1
eq
, (EA)
T2
eq
is equivalent longitude com-
pression rigidity, the elastic and plastic equivalent
tensilerigidityrespectively. K
j1
, K
j2
iselasticandplas-
ticspringcoefficient. K
j1
=nk
j1
; K
j2
=nk
j2
. N, N
0
and
N
y
is longitudeaxial forcefor unit, pre-pressurefor
jointsandtheelasticultimatetensionfor unit respec-
tively. E
c
, A
c
isYoungsmodulusof concreteandarea
of segment.
3.3 Equivalent bending rigidity
When thesegment is in absoluteelasticity stateand
thetensilestress of all thebolts is less than Py, the
unit hasthestressandstrainshownintheFigure6, in
which, x, indicatethepositionof neutral axis, where,
x =r sin.
Intermsof thedeformationcompatibilitycondition
and the force balance equation, where, meets the
Equation2:
Figure7. Stressdistributionfor segment inplasticstate.
Thentheequivalent elastic bendingstiffness for tun-
nelscanbeexpressedasfollow:
Elastic ultimateflexural moment for priestesses seg-
ments, definedas:
Therefore, the tunnel radius of curvature corre-
spondingtoelasticultimateflexural moment is:
Wheny=D,2, themaximumcompressionstressfor
concretesegment canbeexpressedas:
When y= D,2, the maximum tensile stress for
concretesegmentscanbeexpressedas:
Wheny= r, themaximumtensionanddeformation
for joint boltsrespectivelyexpressedas:
When the flexural moment of tunnel segment is
larger than elastic ultimate flexural moment M
y
,
the outermost joint bolts would demonstrate plastic
494
Table1. Mainstructural parametersfor shieldtunnel.
Bolt Ratioof
Youngs Compression diameter Youngs Yield Ultimate elastic
Inner Outer Ring modulusof strengthof /length/ modulus strength strength stiffness
diameter diameter width concrete concrete quantity of bolt of bolt of bolt andplastic
(m) (m) (m) (kPa) (kPa) (mm/mm) (kPa) (kPa) (kPa) stiffness()
8.8 10.0 0.9 3.3510
7
2.9610
4
30/580/64 2.0610
8
5.410
5
8.3510
5
0.01
characteristic. The stress and strain of segment in
which some bolts reach the plastic state is shown
in Figure 7. and demonstrate the critical posi-
tionwherethedeformationof boltsiselasticultimate
deformation
y
, and =r sin.
Intermsof thedeformationcompatibilitycondition
andtheforcebalanceequation, anequationconcerned
with and canbeobtainedasfollows.
Wheny=r,themaximumdeformationof jointbolts
max
expressedas:
Where,
y
is elastic ultimate deformation for joint
bolts,
(Zheng, 2005)
Where
0
ispre-pressureof bolts,
y
isyieldstress,
u
is ultimate stress and E
j
isYoungs modulus. R
1
,
R
2
are coefficients in equation, R
1
=
1
1+
EcAc
lsK
j1
, and
R
2
=
1
1+
EcAc
lsK
j2
.
Themaximumcompressionstressof concreteseg-
ments
The tunnel longitude deformation curvature can be
definedasfollows:
Therelationships betweentunnel longitudecurva-
tureandstressof concretesegment, jointscrack, bolts
inner forceand tunnel ultimateflexural moment are
establishedaccordingto theconclusions drawnfrom
analyticsolutionabove.Itcanbeusedtoevaluatestruc-
tural performanceof tunnels if thelongitudeuneven
settlement hasbeenknown.
4 LONGITUDINAL STRUCTURAL
PERFORMANCE OF TUNNEL
4.1 Critical state for evaluating performance
There are several critical conditions in the develop-
mentof longitudinal stressanddeformationof tunnels.
Inorder toevaluatethesafetystatusof tunnels, some
primarycritical stateswill bestudiedinthefollowing
contents.
1. Crack of joints. The joints crack is limited for
meetingthedemandsof waterproof andstructure.
2. Stress of bolts. There are two key critical states,
namely, the stress of outermost bolts reaches to
yieldstressandultimatestress.
3. Compressionstrengthof concrete. Concretestress
in compressive area reaches flexure compression
strengthof concretesegments.
4.2 The evaluation for tunnel longitudinal
structural performance
The elastic and plastic spring stiffness coefficient
may be calculated by parameters listed in Table 1.
k
j1
=E
j
A
j
,l
j
=250929.0kN/m, k
j2
=2509.3kN/m.
When k
j1
is introduced in Equation 2, will
be 0.9895. The elastic equivalent flexural stiffness
(EI )
1
eq
=2.0951 10
9
kNm
2
. Then the value of R
1
and R
2
are gained based on the upper parame-
ters, as R
1
=2.3810
2
, R
2
=2.4410
4
. When
y
=1.82mm, and arecalculatedas =1.36368,
=1.20615. So thecorresponding curvaturecan be
calculated. The inner force and deformation corre-
sponding to thecurvatureof longitudinal settlement
curvefor several importantcritical statesareacquired
bytheanalytical methodstatedabove, listedinTable2
indetail.
495
Table2. Inner forceanddeformationof tunnel incritical states.
Crackof
Radiusof segment Concrete
Curvature curvature Bolt stress joint stress
Critical state (m
1
) (m) (kPa) (mm) (kPa)
Waterproof requirement 1.1010
4
27300 2.9710
5
1.00 2.4710
4
(crackof segment joint
1mm)
Concretestressreaching 1.5910
4
18818 4.2910
5
1.21 2.9610
4
Compressionstrength
Outermost bolt stress 2.0010
4
15000 5.4010
5
1.82 3.7210
4
reachingyieldstrength
Outermost bolt stress 9.9410
3
302 8.3510
5
83.06 8.3510
5
reachingultimatestrength
When the radius of curvature is greater than
27300m, thestressof segments, tensionof boltsand
meet thedemands, sothetunnel issaferelatively, but
theleakagemay occur. When theradius is less than
18818m, thecompressionstress of concreteincom-
pressionareawill exceedthecompressionstrengthof
segments. Whentheradius is less than15000m, the
outermostboltsincompressionstatemaybeintoplas-
tic state. Whentheradius is less than302m, theend
outboard bolts have reached at ultimate stress state,
andmaybetensilefailure.
4.3 Evaluation on structural performance
As for the part of tunnel between Shaft No. 3 and
No. 4, settlements absolutevalueof segmentbetween
issmall relativelyandfloatsintherangeof 20.0mm.
The radius of curve is large. So the structural inner
forceand deformation arein alow level , whilethe
longitudinal deformationcurvatureislargeinthearea
near Shaft No.3and thecrack of segment joints can
not meet thewaterproof requirements.
As for thepart of this tunnel betweenShaft No. 3
andShaft No. 2, theradiusof curvatureisless, about
20000m.Thefieldinspectionresultsshowthatthereis
severeleakageof inner liningof thetunnel, especially
near Shaft No. 3.
AsforthepartbetweenShaftNo. 2andShaftNo. 1,
there is great tunnel longitude settlement, and the
uneven settlement develops dramatically, especially
nearbyShaft No. 2, whichbecomestheweakposition
inthewholetunnel.
5 CONCLUSIONS
Fromthetunnel settlement curveandcurvaturedata,
thesettlement of shield segment trends to besteady
aftermanyyears operation. Buttheunevensettlement
of tunnel developsgradually.
Themagnitudesof longitudedeformationcurvature
correspondingtoseveral significant crucial statesare
obtained. Thesemagnitudes can beused to evaluate
the structural performance according to the tunnels
longitudesettlement.
Whentheradiusof curvatureislessthan27300m,
thewaterproof requirementfor crackof segmentjoint
doesnotmeetthedemands,andtheleakagemayoccur.
Whentheradiusislessthan18818m, thecompression
stressof concretewill exceedthecompressionstrength
of segments. When the radius is less than 15000m,
the outermost bolts in compression state may yield.
Whentheradius of curvatureis less than302m, the
outboardbolts wouldreachultimatestress state, and
tensilefailuremayoccur.
ACKNOWLEDGMENTS
This work was supportedby Shanghai LeadingAca-
demicDisciplineProject(Project Number: B308).
REFERENCES
Liu, J. H. & Hou, X.Y. 1991. Shield tunnel. Chinarailway
press:369395
Shiba, Y. & Kawashima, T. 1988. The evaluation of the
ductslongitudinal rigidityintheseismicanalysisof shield
tunnel. In: Proceedings of the Civil Academy :319327.
Zheng,Y.L. 2005. Studyonlongitudinal crackof shieldtun-
nel segment joint due to asymmetric settlement in soft
soil. Chinese journal of rock mechanics and engineering,
24(24):45524558.
Zhu, H.H. &Tao, L.B. 1998. Studyontwobeam-springmod-
elsforthenumerical analysisof segmentsinshieldtunnel.
Rock and Soil mechanics, 19(2):2632.
496
Theme 4: Safety issues, risk analysis hazard
management and control
Researchonstochasticseismicanalysisof undergroundpipelinebasedon
physical earthquakemodel
X.Q. Ai
J. Li
Department of Building Engineering, Tongji University, Shanghai, P.R. China
ABSTRACT: Inthispaper, anewapproachfor thestochasticanalysisof finiteelementmodeledunderground
pipelinesystemunder earthquakeexcitationsisproposed, whereanewdevelopedphysical randomearthquake
model andtheprobabilitydensityevolutionmethodareadopted. Basedonthephysical processof seismicspread,
therandomearthquakemodel adoptednotonlyindicatesthephysical relationshipbetweentherandomearthquake
groundmotionandseveral key randomparameters, but alsopresents theprobability configurationof random
seismicmotions. Associatingwiththedevelopedprobabilitydensityevolutionmethodfor stochasticstructures,
theinstantaneousprobabilitydensityandtheevolutionprocessfor theseismicresponseof undergroundpipeline
canbeanalyzednumerically, andthenthestochastic responsecanbeobtainedeasily. Withthefiniteelement
method, thestochastic responseof underground pipelineunder seismic excitation can bestudied effectively.
Usingtheproposedmethod, numerical exampleunder different randomconditionsisinvestigated, showingthat
theproposedmethodisof highapplicationinconsideringrandomseismicexcitationwithrandomsoil parameter
at thesametime.
Asabasic andimportant probleminseismic reliabil-
ity researchfor theundergroundlifelinessystem, the
stochasticseismicresponseof undergroundpipelines
has for a long time been extensively studied, com-
ing up with a variety of theoretical and numerical
methodsapplicabletoengineeringpractice(Machida
& Yoshimura 2002; Nedjara et al. 2007). However,
on account of the calculation limitation in previous
studies, the stochastic response analysis considering
the randomness simultaneously caused by the seis-
micinputandthestructural parametersof soil-pipeline
systemisstill anunsolvedproblem.
For thenon-linear stochastic structures, it is diffi-
cult to capturetheaccurateprobabilistic information
of dynamic performance. The dynamic response of
non-linear stochastic structures was analyzedusually
by the randomsimulation, the randomperturbation
method or by theequivalent linearization technique.
In 1986, Liu et al.(1986) started to investigate the
problemwith extension of the randomperturbation
technique. In 1980s, the Monte Carlo simulation is
also employed (Deodatis & Shinozuka 1988), and
someresearchersstudiedthetechniquestoreducethe
considerablecomputational work (Spanos & Zeldin,
1998). As the alternative approaches, the equivalent
linearizationtechnique(Klosner et al. 1992) andthe
extensionmethodsusedforlinearstochasticstructures
such as the orthogonal polynomial expansion (Iwan
& Huang 1996), are also investigated. However, for
thedynamic responseanalysis of non-linear stochas-
tic structures, the randomperturbation method had
great difficulty becauseof thesecular termsproblem
andtherequirement of small coefficient of variation
of therandomparameters. Therearesomearguments
for theequivalentlinearizationmethodbecauseof the
misleading results in some occasions. The applica-
tionof orthogonal polynomialsexpansionseemsalso
unfeasibleformultiple-degree-of-freedomsystemand
non-polynomial-formnon-linearity.
Evidently, more investigations are necessary for
the dynamic response analysis of the underground
pipelines. Inresentyears, anewlydevelopedprobabil-
ity density evolutionmethodfor stochastic structures
hasbeenproposed.Theinstantaneousprobabilityden-
sityfunctionandtheevolutionagainstthetimecanbe
obtainedprecisely(Li &Chen2005). Forthedynamic
responseof stochastic structures, thesolutioncanbe
derived through solving theprobability density evo-
lution equation with an initial value condition. On
the other hand, different fromthe models fromthe
499
power spectral density function, anewphysics-based
stochasticearthquakemodel toreflect thefundamen-
tal correlation between the critical factors and the
random earthquake motions has been proposed by
the authors (Li & Ai 2006). Several key random
parameters are based on to realize the purpose, and
arelational expressionwithphysical backgroundcan
beconstructedconsideringthepropagationprocessof
earthquakemotioninanengineeringsite. Associated
withtheprobabilitydensityevolutionmethod, thepro-
posedrandomearthquakemodel canbeof comparative
advancetoprovidethebasisfor thestochasticseismic
analysis.
In this paper, based on the probability density
evolution method and the physics-based stochastic
earthquake model, a new approach for the dynamic
responseanalysisof thefiniteelementmodeledunder-
groundpipelinestructures under randomearthquake
excitationsanduncertainsoil parametersisproposed
andnumerical exampleisinvestigated.
2 PROBABILITY DENSITY EVOLUTION
METHODFOR STOCHASTIC STRUCTURES
2.1 Principle of preservation of probability
The principle of probability conservation can be
describedas, duringaconservativeprobability trans-
formation process, within the state space, the incre-
mentof theprobabilitywithinaunitvolumeequalsto
theinflowprobabilitythat getsacrossthisunit.
If theordinarydifferential of astatevector Ycanbe
expressedas,
whereY =(y
1
, y
2
, . . . , y
n
)
T
, G=(g
1
, g
2
, . . . , g
n
)
T
.
If p
Y
(y, t) issupposedtobetheprobabilitydensity
of Y(t), basedontheprincipleof probability conser-
vation, theprobabilitydensityevolutionequationcan
beexpressedasfollowed,
2.2 Probability density evolution equation of the
seismic response of stochastic structures
Thedynamic equationof thenonlinear structurecan
bewrittenas,
where istherandomparameter vector whichrepre-
sents thephysical character of astochastic structure,
anditsjointprobabilitydensityfunctionisp
(x); M,C
arestochasticmassanddampingmatricesrespectively,
which include randomparameters with the rank of
n n, nisthedynamicfreedomdegree; U,

U,

U are
the displacement, the velocity, and the acceleration
vectors, respectively; f (, U) is thenonlinear restor-
ingforcevector; and

U
g
is theaccelerationvector of
input earthquake.
Let aresponsevector X =(U
T
,

U
T
)
T
, thedynamic
equationisthenchangedintoaformatof stateequation
includingrandomparameters,
If

X =

t
X(, t)=G(, t), basedontheaboveproba-
bilitydensityevolutionequation, thejoint probability
densityfunctionof X and will satisfythefollowing
equation,
Sincethevelocity component

X
i
is only thefunction
of variable, theaboveequationcanbere-writtenas
Having integral at both sides with x
1
, . . . , x
l1
, x
l+1
,
. . . , x
2n
,theprobabilitydensityevolutionequationwill
bedecoupled,
where p
X
l
(x
l
, x
, t)=
p
Y
(x, t)dy
1
. . . dy
l1
dy
l+1
. . .
dy
2n
, is the joint probability density function of
(X
l
,
T
)
T
.
Whentheinitial displacement andthevelocity are
independent of thephysical parameters of thestruc-
ture, thecorrespondinginitial condition is described
as,
whereX
l,0
isthedeterminateinitial valueof X
l
, for a
initial staticstructure, thereisX
l,0
=0; ()istheDirac
function; and p
(x
) is the joint probability density

functionof therandomvector .
Solvingtheabovepartial differential equationwith
initial-boundary-values, whichincludethedecoupled
probability density evolution equation and the ini-
tial condition, the joint probability density function
500
p
X
l
(x
l
, x
, t) canbecalculated. After theintegral with

x
, theprobabilitydensityfunctionof X
l
(t) canalsobe
solved,
The probability density evolution method of the
stochasticstructuresisaneffectiveapproachtoobtain
theinstantaneousprobability density functionandits
evolutionfor thestochasticresponse.
2.3 Computational algorithm of probability density
evolution equation
Step1: Dispersethevalueregionthat correspondsto
, anddispersetheinitial conditionsimultaneously;
Step2:Afterdispersing, foreachdeterminatevalue
of variable, adeterminatedynamicanalysisiscarried
out andthecorrespondingdifferential itemwithtime
of thetarget response

X
l
(x
, t) canbededuced;
Step 3: Using the method of finite difference to
solvetheprobabilitydensityevolutionequation, then
thejointprobabilitydensityfunctionp
X
l
(x
l
, x
, t) and
theprobabilitydensityfunctionof thetarget response
canbecalculated.
3 RANDOM MODEL OF EARTHQUAKE
GROUNDMOTIONFOR ENGINEERING
SITE
UsingarandomFourierfunctionwithadherentproba-
bilityasthemodellingform, theproposedmodel isto
reflect theintrinsic relationship between therandom
seismicmotionandthecritical parameters.
3.1 The physical relation and modeling
If searching from the mechanism of the seismic
wave propagating through engineering site, earth-
quake ground motion can be regarded as a physical
process, whichincludesbeinginput fromthebedrock
baseandfilteredbythesite.Accordingly, itisbelieved
that because of the uncontrollability of the random
factors including the afferent wave energy and the
soil mediumcharacteristicsthat theactual earthquake
groundmotionsareobservedwithsignificantrandom-
ness as aresult. Thementionedfactors aresupposed
tobetheenergy factor of afferent wave, theperiodic
factor andthedissipativefactor of seismicsite. Corre-
spondingly, thesefactors canbepractically indicated
bythefollowingstochasticvariableswithoperational
physical meaning: thebasal spectrumparameter, the
freeangular frequencyandthedampingratioof engi-
neeringsite.Consequently,aphysical relationbetween
randomearthquakemotionandcritical factorscanbe
constructedandmodeled.
Without loss of generality, the soil layer can be
simulated as an equivalent linear single-degree-of-
freedomsystemwhich is input by a seismic motion
from the base, and then the absolute response of
this system can be supposed to present the seis-
mic ground motion process. Therefore, if this linear
single-degree-of-freedomsystemis input with one-
dimensionseismic motion, infrequency domain, the
absoluteaccelerationresponseisexpressedas
where is independent parameter, read as angular-
frequency;

Y() istheabsoluteoutput accelerationin
frequency domain;

U() is the absolute basal input
acceleration in frequency domain;
0
is the natural
angular-frequency of thesite;
0
isthedampingratio
of thesite.
Accordingtothepreviousstatement, thebasal input
motion and themediumcharacteristics of soil layer,
whichleadto therandomness of earthquakeacceler-
ation motion, areboth of uncertainty. Therefore, the
parametersof soil mediumthefreefrequency
0
and
the damping ratio
0
are apparently stochastic vari-
ables,notedasX
andX
respectively.Assumerandom
variable vector X
H
=(X
, X
)
T
to relate to the soil
medium.
If thebasal inputFourierspectrum
U()isthefunc-
tionof stochasticvariablesX
1
, . . . , X
n
,

U()isdefined
with
Assumerandomvariablevector presenting thebasal
input Fourier spectrumtobeX
G
=(X
1
, . . . , X
n
)
T
, and
assign randomvariable vector X=(X
T
H
, X
T
G
)
T
with
the joint probability p
X
(x). Then the randomearth-
quakegroundprocessdescribedwithaccelerationcan
beexpressedby arandomFourier spectrumfunction
F(X, ), whichisexpressedas
3.2 The basal input spectrum
If presented by power spectral density function, the
bedrock seismic responses aregenerally assumed to
beaband-limited whitenoisespectrum. However, a
large amount of observed acceleration records have
shown that the bedrock spectrumpresents filtering
and limited band character. Moreover, relevant anal-
ysis indicates that the frequency range greater than
501
Figure1. Basal Input Spectrum.
Figure 2. Contrast of the proposed random earthquake
model andreal recordsfor Soil TypeI.
94Rad/s (15Hz) can beneglected becauseof corre-
spondingminuteamplitude.
Accordingtotheprecedingthoughts, abasal input
spectrum is constructed (Li & Ai 2006), which is
shown in Figure 1, where,
1
,
2
are the con-
trol angular-frequency;
e
is the truncated angular-
frequency; X
u
is the random variable defining the
amplitudeof thebasal input spectrum.
3.3 Random earthquake model basing on physical
process
In order to determine the physical variables in the
proposed random earthquake model, according to
model andreal recordsfor Soil TypeII.
stochastic modeling theory, actual strong seismic
recordsareregardedasatargetsampleaggregationand
numerical methods are adopted to identify different
distribution parameters of the variables correspond-
ingtotheproposedmodel (Li &Ai 2006). According
to different site class, a collection of acceleration
records, whichchiefly comefromwesternAmerican
strongseismic records, arecollectedandreorganized
to establish different equivalent randomearthquake
model.
For differentsiteclass, thecontrastsbetweenactual
seismic records and the proposed model are shown
in Figures 25. The contrasts imply that the pro-
posedmodel isshownof definitephysical conceptand
propriety to reflect the variation of randomseismic
motion.
4 STOCHASTIC SEISMICANALYSISOF
UNDERGROUNDPIPELINE
4.1 Computational example
Thesiteis a5010muniformsaturated sandy soil
spaceandthegroundwaterissetatthegroundsurface.
Andthepipemadeof cast ironisburiedat thedepth
of 1mwiththediameter is0.4m.
502
model andreal recordsfor Soil TypeIII.
model andreal recordsfor Soil TypeIV.
Soil Pipeline
Boundary condition
Contact surface
1
0
M
50M
A
Figure6. TheStructural Mode.
Figure 6 shows the finite element mode of the
computational example. Thestructureis subjectedto
earthquakeexcitationintheshapeof theconditional
randomearthquake function which is defined sub-
sequently. The earthquake is input fromthe bottom
horizontallywiththepropagationvelocity250m/sand
thedurationis20s.Thejointelementsadoptthesame
formwiththepipelineelement,whiledifferentYoungs
moduleanddensityaredefined.
For the deterministic analysis of the seismic
responseof buriedpipelines,thefiniteelementmethod
is adopted to study the seismic response of buried
pipelines and the surrounding soil. The soil that
surrounds the pipeline is regarded as a solid-liquid
two-phase medium. The effective stress method and
nonlinear constitutivemodel of soil areusedtostudy
the increase and the dissipation of pore water pres-
sure during the seismic process. At the same time,
thecontactinterfacebetweenthepipelineandthesur-
roundingsoil isalsoincluded. Detailedtechniquesand
thedynamicparametersof thesandysoil andthecon-
tact surfacecanbereferredtotherelatedpaper of the
authors(Ai & Li 2004).
4.2 Stochastic seismic analysis of underground
pipeline
Inthispaper, therandominputsandtheuncertainsoil
parameter are both considered and defined accord-
ingtothereference(Li andAi 2006): associatedwith
the seismic risk analysis, in the future definite time
range, theconditional randomearthquakefunctionis
defined as thecaseof thetranscendental probability
of theearthquakemotionbeingp =3%andthepeak
valueof theseismicaccelerationis0.2g; Theinternal
frictionangle, whichis definedas therandomsoil
parameter, hasalogarithmicnormal distributionwith
themean15
0
andthestandarddeviation0.30.
Undertheproposedstochasticconditions, theprob-
abilitydensityfunction(PDF) of theresponseforjoint
E ispresentedinFigure7, includingthemeanandthe
standarddeviation, typical instantaneousPDFsatcer-
tain instants of time, evolution of PDF against time
and contour to the PDF surface. These indicate the
503
Figure7. ThePDF of theresponsefor joint E.
stochasticfluctuationcharacter of anonlinear random
response.
5 CONCLUSIONS
A newapproachisproposedfor thestochasticseismic
response analysis of finite element modeled under-
ground pipeline structures under earthquake excita-
tions and uncertain soil parameters. Theapproach is
establishedbasedonthethoughtsof thenewlydevel-
opedprobabilitydensityevolutionmethod.Associated
with the physics-based stochastic earthquake model
andthefiniteelementmethod, theinstantaneousprob-
ability density andtheevolutionprocessfor theseis-
mic responseof undergroundpipelinecanbestudied
effectively. A computational example is investigated
with stochastic input and random soil parameters.
Somefeaturesof theresponsesareobservedanddis-
cussed. It is found that the proposed method is of
highapplicationinanalyzingthestochastic response
of underground pipeline and the stochastic seismic
response for different structural forms requires fur-
ther investigation. Furthermore, theseismicreliability
problemcan be investigated easily by imposing the
failurecriterionof thefirstpassageproblemreliability
theory.
ACKNOWLEDGEMENTS
The support of the Natural Science Foundation
of China for Innovative Research Groups (Grant
No.50321803) isgratefullyappreciated.
REFERENCES
Ai, X.Q. & Li, J. 2004. Seismic responseanalysisof under-
groundpipelinesusingeffectivestressmethod. Proceed-
ing of 13thWorld Conference on Earthquake Engineering.
Vancouver, Canada
Deodatis, G. & Shinozuka, M. 1988. Stochastic FEM anal-
ysis of non-linear dynamic problems. In: Shinozuka M
(ed). Stochastic Mechanics III, Department of Civil Engi-
neering and Operations Research. PrincetonUniversity.
Iwan, W.D. & Huang, C.T. 1996. Onthedynamic response
of non-linear systems withparameter uncertainty. Inter-
national Journal of Non-Linear Mechanics 31(5):
63145.
Klosner, J.M., Haber, S.F. &Voltz, P. 1992. Responseof non-
linearsystemswithparameteruncertainties. International
Journal of Non-Linear Mechanics 27(4): 54763.
Li, J. & Ai, X.Q. 2006. Study on Random Model of
EarthquakeGround Motion Based on Physical Process.
Earthquake Engineering and Engineering Vibration (in
Chinese) 26(5): 2126.
504
Li, J. & Chen, J.B. 2005. Dynamic responseand reliability
analysis of structures with uncertain parameters. Inter-
national Journal for Numerical Methods in Engineering
62(2): 289315
Liu, W.K., Belytschko,T. &Mani,A. 1986. Probabilityfinite
elements for nonlinear structural dynamics. Computer
Methods in Applied Mechanics and Engineering 56:
6181.
Machida, H. & Yoshimura, S. 2002. Probabilistic frac-
ture mechanics analysis of nuclear piping considering
variation of seismic loading. International Journal of
Pressure Vessels and Piping 79:193202
Nedjara, D., Hamanea, M., Bensafia, M., Elachachia, S.M. &
Breysse, D. 2007. Seismic responseanalysis of pipes by
aprobabilistic approach. Soil Dynamics and Earthquake
Engineering 27:111115
Spanos, P.D. & Zeldin, B.A. 1998. Monte Carlo treatment
of randomfields: abroadperspective.Applied Mechanics
Review 51(3): 21937.
505
Riskassessment for thesafegradeof deepexcavation
X.H. Bao& H.W. Huang
ABSTRACT: This paper embarks fromrisk idea, obtains safety factors of excavationandtheir percentages
frominvestigationresultsof excavationaccidents, thesafegradeof excavationisdesignedandanalyzed. First, the
presentdevelopingsituation,possiblyexistedrisksinexcavationprojectsandthestateof presentexcavationgrade
divisioninshanghai areaisbrieflyintroduced. Reasonsof accidentsareobtainedfromanalyzingalargenumber
of excavation accidents. Later, fuzzy synthetic evaluation method is introduced and two levels of judgment
on excavation safety arecarried on. Finally, thepaper takes theexampleof shanghai international passenger
transport centre, theresultsconformtotheactual situation.
1 INTRODUCTION
1.1 The present risk study of deep excavation
engineering
In recent years, excavation engineering develops
quickly, and excavations are becoming larger and
larger. Theemergenceof theselargeexcavations has
brought new opportunities and challenges to engi-
neeringdesign, constructionandmanagement. Onthe
other hand, a great many of engineering practices
helpedtechnical development, improvedmanagement
level,alsomadetheriskanalysisapplyinmanyaspects
of engineering. Currently, most of theexcavationsare
mainlydesignedaccordingtothespecifications, rules
and accepted practice. However, excavations can be
designed according to the risk idea that is, to study
latentriskof excavationduringtheperiodof construc-
tion and operation, and reduce the unavoidable risk
to the insignificant level by increasing construction
cost. This method bases on the principles of relia-
bility; it can make the probability of reaching limit
state small enough, and reduce the excavation risk
to acceptablelevel according to risk acceptancecri-
teria.Theresearchinthisaspectathomeandabroadis
still at abeginningstage, specializedresearchof risk
concerning deep excavation engineering is seldom,
the results obtained are almost qualitative, quantifi-
cational research is seldom, applicable risk analysis
andevaluation methods that areboth qualitativeand
quantificational areneeded(HuangandBian, 2005),
soitisdifficulttoconstructprobabilitymodelsof risk
factors.
1.2 The present grade division and risk
factors in excavation
Therearestill someproblemsinpresentgradedivision
bothat homeandabroad. For example, uncertaintyof
various factors which can influencetheresult, small
applicability scopeof calculationtheories andso on.
Thesewill makeconstructioncost beincreased. Cur-
rently, the grade of excavation is divided according
to deformation values (Specification for Excavation
inShanghai MetroConstruction, SZ-08-2000). Every
gradehasitscontrol values.Whenthegradedivisionis
not precise, if higher whichmeansthesafetywarning
valueis smaller, it will forecast thesafeconditionas
dangerouscondition,thentheexcavationrequestshigh
rigidity retaining structures, also more bracings and
bottomstrengthening may beneeded to resist defor-
mation.Actuallyinthiscondition,investmenthasbeen
increased. If lower which means the safety warning
valueislarger, theconditionof excavationhasalready
beendangerousbeforeforecasting, it maybetoolater
toremedy, soitcancausegreatlossandtheworkmay
bedelayed.
Thecharacteristicsof excavationengineeringmean
that all owners have to face huge risk during con-
struction.Asthesoil layerconditionsandgroundwater
circumstances are uncertain, technologies are com-
plex, man may make mistakes in techniques and
managements, accidentsmayalwayshappenindisad-
vantageous soil layers, so, therewouldnot only exist
environment risk, but also other risks such as over-
spendingandextensionfor completiondate, etc. The
seriouscollapsesappearedrecentlyhaverevealedthat
507
Table1. Statisticsof reasonsfordeepexcavationaccidents.
Reason frequency Percentage(%)
Investigation 16 4.65120
Design 125 36.3372
Construction 176 51.1628
Supervision 5 1.45350
Monitor 10 2.90700
Owner 12 3.48840
Total 344 100.000
Table 2. Retaining structures for the failed deep excava-
tions.
Retainingmethod frequency Percentage(%)
Rowpiles 150 66.3717
Soil-nailingsupport 30 13.2743
Deepmixingpile 13 5.75220
Diaphragm 21 9.29200
Excavatedslope 5 2.21240
Soil nailedwall 5 2.21240
Others 2 0.88500
Total 226 100.000
accidentscancausemany bigtroublesintheexcava-
tionengineering. Theexcavationconstructedincities
cancauselifeandpropertylosstothethird.Thesocial
problemsandpublicprotest causedbyexcavationcan
extend construction period (International Tunneling
Association, Second Unit, Guiding Rules of Tunnel
RiskProcessing, 2002), sospecial attentionshouldbe
paid.
2 RISK ASSESSMENT FORTHE SAFE GRADE
OF DEEP EXCAVATION
2.1 Reasons for deep excavation accidents
Therearenumerousreasonsfor deepexcavationacci-
dents, andalot of scholarshavepapersinthisaspect.
Thefollowingstatisticsaregottenfromabout300deep
excavationaccidents(h>6m).
Fromtable1 it can beseen that construction and
Designarethemainreasons. Ontheother hand, con-
sider 226accidentsof thesefailedexcavations; divide
themaccordingto retainingmethodintable2. Most
of themareretainedbyrowpiles. Althoughtheremay
beproblemsinthisdivision, anditcannotbesaidthat
rowpileismoredamageablethanotherkindsof retain-
ingstructures, but accidents happenedfrequently, so
moreattentionshouldbepaidonexcavationsretained
byrowpiles.
2.2 Fuzzy synthetic evaluation method
Thefuzzysyntheticevaluationprocess(LiangandBi,
2001) isasfollows:
U is influence-factor set, U ={U
1
, U
2
, . . . U
m
},
U
m
(i =1, 2, . . . m) is thei factor, inthesecondary
synthetic evaluation, U
i
={U
i1
, U
i2
, . . . U
in
}, U
in
is
thesub-factor of U
i
.
V is comment set, V ={V
1
, V
2
, . . . V
p
}, V
j
( j =1,
2, . . . p) isthej grade.
Construct evaluationmatrixR
Where R is evaluation matrix of single factor,
r
ij
is the relative membership of U
i
to V
j
, R
i
(i =1, 2, . . . m) issubordinationvector.
A is weight set of mainfactors, A=(a
1
, a
2
. . . a
m
),
a
i
is theimportancedegreeof U
i
comparedto the
other factors, 0a
i
1.
Choose composite operators and get composite
evaluationresult B bymultiplyingA andR:
WhereB is subordinationvector of mainfactor to
comment congregation, R is evaluation matrix of
singlefactor, b
i
subordinationvector.
Analyzeresults
2.3 Influence factors
Considering investigations and analysis of other
researches(YangandDing, 1998), thispaper consid-
ersfivemainfactorsthathaveinfluenceonexcavation
safety: size of foundation pit U
1
, hydrogeology U
2
,
designU
3
, constructionU
4
andsurroundingenviron-
ment U
5
. Size of foundation pit includes area U
11
,
shape U
12
, depth U
12
. Hydrogeology includes layer
conditionU
21
, confinedwater U
22
, driftingsandU
23
.
Design includes calculation method U
31
, value of
parameter U
32
, valueof load U
33
, material selection
U
34
andretainingstructureU
35
. Constructionincludes
if according to design and rules U
41
, construction
method U
42
, construction experience U
43
, dewater-
ing U
44
, supporting U
45
. Surrounding environment
includesdynamicloadU
51
, surchargeloadU
52
, adja-
cent construction U
53
. In fuzzy synthetic evaluation,
thesefactorsarealsocalledindexes; themembership
betweenfactorsandtheirsub-factors(indexessystem)
isshowninfigure1.
508
S
a
f
e
t
y

f
a
c
t
o
r
s

o
f

e
x
c
a
v
a
t
i
o
n
Surrounding
Environment
Shape
Area
Depth
Layer condition
Confined water
Drifting sand
Calculation method
Parameter value
Load value
Material selection
Retaining structure
Method
According to
design and rules
Dewatering
Surcharge load
Supporting
Experience
Adjacent construction
Dynamic load
Construction
Design
Hydrogeology
Size
Figure1. Factors influencingthesafegradeof deepexca-
vation.
Table3. Weightsof U
i
.
Number Influencefactors Weights
1 Size 0.038
2 Hydrogeology 0.046
3 Design 0.360
4 Construction 0.510
5 Surroundingenvironment 0.046
2.4 Weights of factors
2.4.1 Weights of the factors U
i
Fromthe investigation of 344 excavation accidents
(h>6m), thefollowingresultscanbegottenasshown
intable3.
UsetheAnalyticHierarchyProcess. First, construct
a judgment matrix to calculate the maximumEigen
valueof thematrix and vector feature, and then test
theconsistency of judgment matrix to determinethe
weightsof sub-factors(WangandHuang, 2005).
Table4. Themeaningof 19scale.
Scale Meaning(comparetwofactors)
1 Theimportant degreeisequal
2 Theformer isslightlyimportant
3 Theformer isobviouslyimportant
4 Theformer ismightilyimportant
5 Theformer isextremelyimportant
* 2, 4, 6, 8meanthemiddlevalueof twoscalesclosetogether.
Table5. J udgment matrixU
1
.
U
1
U
11
U
12
U
13
U
11
1 3 1/3
U
12
1/3 1 1/5
U
13
3 5 1
U
2
, U
3
, U
4
, U
5
canbegotteninthesameway
2.4.2 Construct judgment matrix
J udgment matrices can be constructed using Expert
Gradesmethod, accordingtoT.L. Sattys19scale.
U
ij
is subordinated to factor U
i
, compare mutu-
ally the importance degree of every sub-factor U
ij
,
ratio scale u
ij
can be gotten, it reflects the relative
importanceof two sub-factors, if thefirst sub-factor
comparestothesecondsub-factor andtheresultisu
ij
,
then the second compares to the first and the result
isu
ij
=1,u
ij
, sojudgmentmatrixU
1
=(u
ij
)
nn
canbe
gottenasfollows:
2.4.3 Weights of sub-factors
As it does not request high accuracy, to be simple,
calculatethemaximumEigenvalue
max
of thematrix
U
i
andthefeaturevectorA, andA isalsotheweightset
of sub-factors. Thecalculationprocessof A and
max
isasfollows:
Elementsinthematrixmultiplyeachother inrows
andopenan (n =3) power toget A
1i
So, A
11
=1
NormalizeA
1i
toget A
1i
:
So, A
11
=0.296
Usingthesamemethod, A
12
, A
13
canbegotten, and
A
1
=(0.2960.1200.584), A
2
=(0.1200.5840.296),
509
Table6. AveragerandomconsistencyindexR.I .
Order of matrix 1 2 3 4 5
R.I . 0 0 0.52 0.89 1.12
A
3
=(0.104 0.246 0.104 0.045 0.501), A
4
=(0.203
0.0860.4660.2030.042), A
5
=(0.1430.7140.143).
Thelargest Eigenvaluecalculation:
So,
1max
=3.072,
2max
=3.072,
3max
=5.125,
4max
=5.125,
5max
=3.001
2.4.4 Test of consistency
Onaccount of fuzzinessof manyfactorsandpeoples
understandingisdifferent becauseof subjectivity, so,
to overcomethesesubjectiveerrors, theconsistency
of judgment matrixshouldbetested. Whenstochastic
CR satisfies: C.R. 0.1, throughSaatysCR standard,
thedirectionandextent of theimprovement wouldbe
controlled,andtheoptimal matrixwithacceptablecon-
sistencycouldbeavailable, thathadpreservedfurthest
thedecisionmakersprimitivejudgment information.
ConsistencyindexC.I .
So, C.I .
1
=0.036, C.I .
2
=0.036, C.I .
3
=0.031,
C.I .
4
=0.031, C.I .
5
=0.0005
Check table 6 (Gong and Xu, 1986) to get the
correspondedaveragerandomconsistencyindexR.I .
The average randomconsistency indexes R.I . of
everymatrixare: 0.52 0.52 1.12 1.12 0.52
ConsistencyratioC.R.
So, C.R.
1
=0.069-0.1, C.R.
2
=0.069-0.1,
C.R.
3
=0.028-0.1, C.R.
4
=0.028-0.1, C.R.
5
=
0.001-0.1, theconsistency of judgment matrices is
acceptable.
2.5 Comprehensive evaluation and results analysis
Fuzzy synthetic evaluation method is akind of very
validmethodthat cancompletely evaluatethethings
which are influenced by various factors, so, it can
beused largely in many projects and other systems.
This paper carries on two levels of judgment to the
excavationsafety,firstprimaryfuzzysyntheticevalua-
tionof sub-factors, andthensecondaryfuzzysynthetic
evaluationof mainfactors.
Commentsset
Generally,thedeepexcavationsafetycanbedivided
into4grades: very safe, safe, less safe, unsafe, these
can reflect the safety condition of deep excavation,
so thecomments set V ={very safe, safe, less safe,
unsafe}. To calculatesimply, quantify thecomments
set, thenV ={4 3 2 1}.
Construct evaluationmatrixR
Obtaintherelativemembershipdegreeof 19factors
to comment set V and evaluation matrix R by using
Delphi Method.Invite10expertswhoknowwell about
the excavation as an evaluation group, and then get
marks of all indexes fromexperts. The steps are as
follows:
Invite 10 experts to evaluate, the factor sets
areU
1
={U
11
, U
12
, U
13
}; U
2
={U
21
, U
22
, U
23
}; U
3
=
{U
31
, U
32
, U
33
, U
34
, U
35
}; U
4
={U
41
, U
42
, U
43
, U
44
,
U
45
}; U
5
={U
51
, U
52
, U
53
}. Therelativemembership
degreeisr
ij
: 0r
ij
1, thesizeof r
ij
showstheinflu-
ence degree of the factor on safe grade of deep
excavation, and if bigger, more close to the corre-
spondinggrade, conversely, haslittleinfluenceonthe
corresponding grade. For the index U
ij
, if from10
experts, m
i
experts consider the comment is v
1
, n
i
expertsconsiderthecommentisv
2
,p
i
expertsconsider
thecommentisv
3
, q
i
expertsconsider thecommentis
v
4
, thenthesubordinationvectorR
ij
of theindexU
ij
is:
R
ij
=(r
i1
r
i2
r
i3
r
i4
)=(m
i1
,10 n
i2
,10 p
i3
,10
q
i4
,10), i =1, 2, 3, 4, 5.
Evaluation matrices for the five main fac-
tors can be constructed by using subordina-
tion vectors as rows. So, R
1
=(R
11
R
12
R
13
)
T
,
R
2
=(R
21
R
22
R
23
)
T
, R
3
=(R
31
R
32
R
33
R
34
R
35
)
T
,
R
4
=(R
41
R
42
R
43
R
44
R
45
)
T
, R
5
=(R
51
R
52
R
53
)
T
.
Primaryandsecondaryfuzzysyntheticevaluation
Thepaper usesweightedaveragingfuzzysynthetic
evaluation model which can consider all factors and
single-factor evaluationresults.
For the primary fuzzy synthetic evaluation, the
weightedaveragingfuzzysyntheticmodel is:
Where A
is the weight vectors of sub-factors, R
is
single-factor evaluation matrix and R subordination
vector of sub-factor tocomment congregation.
So, for everysingle-factor: R
i
=A
i
R
i
(
i=1,2,3,4,5
)
For thesecondaryfuzzysyntheticevaluation:
Where A is weight set of main factors, R is the
secondary evaluation matrix that is made up of R
j
,
B is subordinationvector of mainfactor to comment
congregation.
So, for eachsingle-factor: B
i
=A
i
R
i
(
i=1,2,3,4,5
)
510
Thefinal valueof comprehensiveevaluation
For thegradeof deepexcavation, weusethecom-
mentssetV ={verysafe, safe, lesssafe, unsafe}, after
qualification, V ={4 3 2 1}. W is the evaluation
valueof deepexcavationgrade.
So, for everysingle-factor: W
i
=B
i
V(
i=1,2,3,4,5
)
If,
3.5<W<4.0, verysafe;
2.5<W<3.5, safe;
1.5<W<2.5, lesssafe;
0.0<W<1.5, unsafe.
3 THE EXAMPLE
3.1 The general situation of the project
Shanghai International Passenger Transport Centre
(west area) locates in south of east Daming Road,
west of Liyang Road, east of Gaoyang Road, north
of Huangpu River. The project covers an area of
128400squaremeters, theshapeof foundationpit is
rectangular andthedepthis13.10meters.
3.2 Primary fuzzy synthetic evaluation
Themembershipof thefactorsandtheir sub-factorsis
knowninfig. 1.
3.2.1 Construct judgment matrix
UseDelphi Method. From10 experts, for theindex
U
11
: the area of this foundation pit, no expert con-
siders it very safefor theproject, 3experts consider
it safe, 6 experts consider it less safe, and 1 expert
considers it unsafe. So, the comment score to dif-
ferent safe grades are: 0, 3, 6, 1, so the subordina-
tion vector R
11
=(0,10 3,10 6,10 1,10)=(0 0.3
0.6 0.1).Aftergettingtheothersubordinationvectors,
theevaluationmatrixfor U
1
isasfollows:
3.2.2 Primary evaluation
As is known, the weights vectors of the sub-factors
to main factor U
1
is: A
1
=(0.296 0.120 0.584),
B
1
=A
1
R
1
=(0 0.475 0.425 0.1)
So, the evaluation value for single-factor U
1
is:
W
1
=B
1
V =2.375, in thesameway, W
2
=1.978,
W
3
=2.893, W
4
=2.533, W
5
=3.158.
Table7. Thesafetygradeof eachmainfactor.
Factor Value Grade
Size 1.5-W
1
=2.375-2.5 lesssafe
Hydrogeology 1.5-W
2
=1.978-2.5 lesssafe
Design 2.5-W
3
=2.893-3.5 safe
Construction 2.5-W
4
=2.533-3.5 safe
Surrounding 2.5-W
5
=3.158-3.5 safe
environment
3.2.3 Secondary evaluation
Thecomprehensiveevaluationmatrixis:
Astheweightsfor fivemainfactorsareknown,
A=(0.038 0.046 0.360 0.510 0.046),so,thesub-
ordination vector B with assessment target U to
comment set V is:
3.2.4 Evaluation results analysis
For thefivemaininfluencefactors of thedeepfoun-
dationpit, thesafetygradedivisionisintable7:
Final comprehensiveevaluationvalue:
So on the whole, the situation of this deep exca-
vation is safe; the system safety can be basically
accepted. But according to theevaluation values for
single-factors, the size and hydrogeology are disad-
vantageous for the whole stability. To prevent this
deep excavation fromaccident, some measurements
should be taken during construction, for example
proper retaining method, good dewatering measure,
timelymonitoring, etc.
Thesizeandhydrogeologyevaluationvaluescanbe
goodadvicetomanagerswhentheymakedecision.As
thesizeandhydrogeologyaredisadvantageousforthis
excavationsafety, sotheyareconsideredtobetworisk
factors, thenduringtheproject management process,
risk management can be carried on to decrease the
probabilityof failurethat maybecausedbythesetwo
risk factors. Through the master and supervision of
risk factors, managerscanadopt activemeasuresand
scientificmanagement.
511
4 CONCLUSION
Indeepexcavationengineering, usingfuzzysynthetic
evaluation method can quantify qualitative analysis
and make inaccurate expression become numeral,
so the evaluation process is more scientific. The
safety influencefactors indeepexcavationengineer-
ingarevarious; thepaper obtainsthesafefactorsand
its weights fromthe investigation results of excava-
tion accidents, so the safe grade evaluation is more
scientificandrational.
Intheprimary evaluation, theresults areadvanta-
geous for discovering thetroubleand weakness that
lurked in engineering; they can provide decision to
managersinordertodoprevention,sothewholesafety
level of deepexcavationisraised.
Theexampledemonstratesthattheresultsconform
to the projects actual situation; and they can help
raising theoverall safety level of theexcavation and
providesomereferencevalues for theprojects. Thus
thispaper hasprovedthat theFuzzySyntheticEvalu-
ation, AnalyticHierarchyProcessandDelphi Method
arefavorableindeepexcavationengineering.
ACKNOWLEDGEMENT
Theauthor isverygrateful of DoctorY.H. Bianfor the
investigationresultsof deepexcavationaccidents.
REFERENCES
Guo, Z.W. 1986. Risk Analysis and Decision. Beijing: China
MachinePress.
Huang, H.W. & Bian, Y.H. 2005. Risk management indeep
excavationconstruction. Chinese Journal of Underground
Space and Engineering 1(4): 611614.
Liang, S. &Bi, J.H. 2001. Fuzzysyntheticevaluationmethod
of construction engineering quality grade. Journal of
Tianjin University 34(5): 664669.
Shanghai Metro Construction Corporation LTD, SZ-08-
2000. Specification for Excavation in Shanghai Metro
Construction. Shanghai.
Soren, D.E., Per, T. & J orgen, K. 2004. Guidelines for tun-
neling risk management: International TunnelingAsso-
ciation, WorkingGroupNo. 2. Tunnel and Underground
Wang, Y. & Huang, H.W. 2005. Hierarchyfuzzy synthetic
safetyevaluationmethodof tunnel insubway. Journal of
Underground Space 24(3): 301305.
Yang, Z.M. & Ding, X.G. 1998. Safety accident andreason
analysisof excavationengineering. Safety Regulations for
Geological Prospecting Operation 5(1): 1013.
512
Multi-factorsdurabilityevaluationinsubwayconcretestructure
C. Chen& L.Yang
C. Han
School of Economics & Management, Tongji University, Shanghai, P.R. China
ABSTRACT: Inthispaper, anerosionmodel of subway concreteliningstructurewasproposed, thenfactors
whichtheinfluencedurabilityof subwaystructureweredetermined, andtheframework of evaluatingmeasure
index was established. Finally, Fuzzy MultipleAttribute Decision Making was used to carry out durability
predictionmodel onthebasis of thesinglefactor evaluationinthesubway structure. Throughcasestudy, the
durabilityof subwaystructureisevaluatedbythemethodsinthisarticleandget goodor badschemes.
1 INTRODUCTION
Shanghai SubwaybelongstotheFastOrbitTransporta-
tionSysteminthecity. Becauseof suchcharacteristics
as being freefromweather influenceand being safe
andfast, itbecomesthekeyvehiclewithinnetworkof
thepassenger transportation (Shen et al. 1998). The
subwaystructureisapermanentconcretestructure, so
theinvestment is usually huge, andit always crosses
key districts of city. Therequirement of durability of
subway is higher than that of other engineering (the
baseperiodfor designis100years), andtheproblem
of durabilityisof most urgency.
The factors influencing concrete structural dura-
bility are numerous and very complicated, and they
are interrelated and influence on each other. There-
fore, it has two kinds of manifestations: one is the
uncertainty of the occurrence of an affair, and the
other is theuncertainty of thebehavior of theaffair.
Theexistingevaluationmethods of concretedurabil-
itymostlyevaluatetheinfluenceof asinglefactor by
setting up a model. However, actually, the concrete
structuredurability isacomprehensivefunctioncon-
tainingmanyfactors, anditsevaluationisacompound
decision process of multipleattributes, multiplefac-
torsandmultipleindexes.Thatistosay, theevaluation
of concretedurability is aFMADM (Fuzzy Multiple
AttributeDecision Making) (Li 2002, Leung & Cao
2000) problem. Inviewof thisproblem, FMADMisto
beusedtocarryoutthemultiplefactorsevaluationon
thebasisof thesinglefactor evaluationinthesubway
structure(Heilpern1997, Zimmermann1987).
2 ANALYSISOF SUBWAY STRUCTURE
DURABILITY
2.1 Some concepts
Thedurability of concretestructureis thecapability
of thestructurewhich keeps its safety, function and
requirementof appearanceundersuitablymaintaining
conditionduringthebaseperiodof designwherethe
effect of theenvironment must beconsidered. Here,
theenvironment isageneral concept, whichcontains
anyfactor resultingindegradationanddestructionof
theperformanceof thestructure.
Therefore, thedurability of concretestructureis a
integrated functional index which is concerned with
not only rawmaterialsandproportioningof concrete
butalsomechanicsenvironment, natural environment,
service environment and production technology of
concrete (Tan 2003). When studying the durability
of concrete structure, two states, which include the
limit of bearing capacity and the limit of function
and appearance requirement, should be considered.
Namely, it is better to obtain economical and ratio-
nal life-spanonthepremiseof ensuringthestructure
safety.
2.2 Influencing-factors of subway structure
durability
Thereareafewfactorswhichcanmodifythedurabil-
ityof concretestructure.Whenproportioninghasbeen
confirmed, themaximal modifyingfactor to durabil-
ityisitsenvironment. Here, thefactorsof environment
513
mainlyincludecarbonation, chlorideionsingress, sul-
fateattack andcycleof freezingandthawing(Du&
Zhang2003).
The erosion model of concrete lining structure is
shownasFigure1.
The concrete lining of subway structure is sub-
jectedtointeractiveactionof multi-destroyfactorsthat
maybecomefrominteriorenvironmentof atmosphere
or formexterior corrosion of geotechnical medium.
Obviously, durability of subway structure should be
analyzed on the basis of its circumstance as it is.
Compared with others concrete structure, the sub-
way structure has two prominent characters. Firstly,
thetemperatureof subwaystructurechangesverylit-
tlebecauseit is long-termlocated under thesurface
of earth. Therefore, theprobability of destruction of
subway structureis very littleowingto theactionof
cycleof freezingandthawing. Secondly, thesubway
Figure1. Erosionmodel of concreteliningstructure.
Table1. Evaluationmeasureindexframeworkof durabilityof subwaystructure.
Secondary
Objectivelevel Primarycriterialevel criterialevel Operational level
Durabilityof StrayCurrent Coltagefactor Loadcurrent of train; Contact potential difference;
Subway Corrosion Transformer substationspacing
Structure Resistance Electrical resistivityof concrete; Transitionresistance;
factor Longitudinal resistanceof rail
Chlorideions Concrete Dosageandvarietyof cement; Varietyandgradedof mixture;
Ingress characteristic Admixturevariety; Water cement radio; Air content;
Curingtime; Hydrationdegree
Diffusivity Environmental temperatureandhumidity; Open-assembly
Coefficient time; Chlorideionconcentration; Stressstate; Crack
SulfateAttack Exterior factors Sulfateionconcentration; Magnesiaion; Chlorideion;
PHvalue; Alternationof wettinganddrying; Freezing
andthawingcycle; Stressstate; Crack
Interior factors Dosageandvarietyof cement; Varietyandgradedof
mixture; Water cement radio; Admixturevariety; Pore
content anddistribution
Carbonation Exterior factors Light andtemperature; Relativehumidity; co2concentration;
Stressstate; Crack
Interior factors Dosageandvarietyof cement; Varietyandgradedof
mixture; Water cement radio; Compressivestrength
of concrete; Constructionquality; Curingmethod
is operated by the traction of electric power. As a
resultof this, itproducesstrayelectrical currentwhich
plays avery important rolein causing thecorrosive
destructioninsubwaystructure.
Therefore, theinfluencing-factors of durability of
subway structure mainly include stray current cor-
rosion, chloride ions ingress, sulfate attack and car-
bonation. Thelife-spanof durability of structureis a
functionof theseinfluencing-factors.
2.3 Evaluating measure index framework of
subway structure durability
As described previously, the influencing-factors of
durability of subway structure mainly include stray
currentcorrosion, chlorideionsingress, sulfateattack
and carbonation. Actually, each factor interact with
many others relative factors. Due to lack of space,
this paper analyzes only thesefour factors described
previously.
Table1showstheevaluationmeasureindexframe-
workof durabilityof subwaystructure.
2.4 Evaluation of subway structure durability
Durability of subway structure is a comprehensive
function containing many factors, such as stray cur-
rentcorrosion, chlorideionsingress, sulfateattackand
carbonation (Figure1). At thesametime, thesingle
factorisalwaysthefunctionof otherinfluential factors
(Table1).
514
whereL
cr
is thedurability of subway structure,
1
is
theinfluential factor of stray current corrosion,
2
is
theinfluential factor of chlorideionsingress,
3
isthe
influential factor of sulfateattack,
4
istheinfluential
factor of carbonation, i is theinfluential factor index
of thesinglefactor.
Thedurabilityevaluationof asinglefactorisrelated
to the same subject that influences the durability of
subway structure, and it is the basis of the durabil-
ity evaluation containing multiple factors. However,
becauseof thedifferent damagemechanismof each
singlefactor, so, each factors evaluation model and
evaluation method are different, and the influential
degreetodurabilityof subwaystructureof eachfactor
isof difference.
For example: thesulfateattack is theentranceof
SO
2
4
fromexternal corrosion media into the inner
concrete through the surface of the concrete, and if
chemical reaction happens between SO
2
4
and sub-
stance of the cement, such as Ca(OH)
2
, then the
chemical reactionwill generateAft, whichisof disten-
sibilityandof hazardtoconcretebycausingexpansive
cracks, andtheconcreteis infragileandloosestate.
Therefore, theprocedureof sulfateattack isfromthe
external totheinterior, andthedamagedegreedeep-
ens withthecorrosiondepth. However, chlorideions
ingress through theconcretesurfaceand theperme-
ationisof littledamagetoconcrete. But, theresult of
thepermeationchangesthealkali environment of the
concrete, which makes the passivating filmof con-
crete structure disappear to induce the rust-eaten of
bars, thus, theconcretestructureis damaged. There-
fore, thedamageprocedureof chlorideionsingressis
fromtheinteriortotheexternal.Thedamagedegreeis
slight ininitial state, anddevelopsingeometricseries
withthetime.
Durabilityof subwaystructureisthecomprehensive
interactionof multiplefactors, andit isinsufficient to
analyzetheissuefromtheactionof thesinglefactoror
thesimpleperspectiveof singlefactors. Becauseof the
fuzzinessof manyfactors, itisnotsuitabletoevaluate
it with an accuratemathematic model. Therefore, to
guaranteetheobjectivismof theevaluationof subway
structure, it isof ultimatecautiontochooseasuitable
evaluationmethod.
On the basis of the Fuzzy Multiple Attribute
Decision Making theory, the article sets up fuzzy
multi-factors evaluationmodel for durability of sub-
way structure applying the methodAHP and expert
investigationtodeterminetheweight of eachfactor.
3 FUZZY MULTIPLEATTRIBUTE DECISION
MODEL (FMADM)
Fuzzy MultipleAttributeDecisionModel (FAMDM)
is acomprehensiveanalysis methodwhichcansolve
multi-factors and indefinite problems. In this paper,
theauthor usefuzzymultipleattributedecisionmodel
to evaluate the durability of subway structure. The
stepsaredescribedasfollows:
1. Suppose the collection of evaluating scheme
X ={X
1
, X
2
, X
3
, , X
m
}, which X are m dimen-
sions pending optimumschemes. Determine the
index vector G
m
={G
1m
, G
2m
, , G
nm
}
T
if each
schemehasndimensionsevaluatingindex.
2. Determine attribute value matrix of scheme col-
lection G
ij
(i =1, 2, ...n, j =1, 2, , m), which is
abbreviatedasr =(r
ij
)
nm
.
3. Transformattributevaluematrix to relativegrade
of membershipmatrix.
Multi-targetsdecision-makingislack of commen-
surability. Namely, eachtargethasntunifiedstandard
of measurement andso they cant becomparedeach
other. For compared easily, the value of attribution
shouldbequantifiedandchangedthemall into[0, 1].
Basedthetypeof target, different quantifiedmethod
should bechosen. Thetarget can bedivided thefol-
lowingtypes: thecost type(thesmaller oneis better
one), the benefit type (the larger one is better one),
the moderation type (appropriate one is better) and
interval type(belocatedinstablezoneisbetter).
Herethebenefit typeis used as quality matrix of
targetattributeof durabilityof subwaystructure. (The
life-spanof durabilityisolder, theeffect isbetter).
where m
k
=minimum eigenvalue; M
k
=maximum
eigenvalue; andd
k
=[m
k
, M
k
]
The base period for subway design is 100 years.
Hence,hereonegetsm
k
=100.Whatsmore,consider-
ingeconomical efficiencyanddemandof sustainable
development inurbanconstructionsynthetically, here
onegetsM
k
=150.
4. Ascertain the weight of each factor, the weight
vector of attribution is (w
1
, w
2
, , w
n
)
T
, here
n
j =1
w
j
=1, andw
j
0. Theweight vector isascer-
tained by the method of arrangement analysis,
whichcanbeseenin4.4indetails.
5. Calculatetheequation:
A
i
is the collection of evaluating result. a
j
(j =
1, 2, m) is the fuzzy evaluating value of each
515
Table2. Alternativeschemeof concreteproportioningandattributeof influencing-factorsof durability.
Attributeof influencing-factors
Primarymaterial content of concrete(kg/m
3
) of durability(years)
Stray Chloride
Scheme Cobble Fly Mineral Current ions Sulfate
Number Cement Water Sand (525mm) ash Powder HLC Corrosion ingress attack Carbonation
D300 339 165 784 1148 0 0 45 114 110 163 112
D310 254 161 713 1138 108 0 45 126 159 144 100
D320 138 159 718 1051 0 228 45 134 134 119 150
D331 163 150 703 1080 110 110 45 143 150 122 113
D332 110 157 712 1062 110 148 45 180 173 135 118
D300: PrimaryStandardConcreteD310: ConcreteContainingFlyAshD320: ConcreteContainingMineral Powder

D331D332: ConcreteContainingFlyAshandMineral Powder
HLC: crack-resistanceandseepage-proofingagent
scheme. The superior or inferior rank of schemes
is estimated by the evaluating value (01), and the
evaluatingvalueislarger, theschemeisbetter.
4 ILLUSTRATIVE EXAMPLE
Hereare5Alternativeschemes for researching high
performance concrete. Through the researching to a
seriesof experimentsandcorrelativedurabilityworks,
thelife-spanof durabilityof theseschemesisobtained
under theaction of stray current corrosion, chloride
ionsingress, sulfateattackandcarbonation.
TheAlternativeschemesof concreteproportioning
andtheresearchingattributeof thelife-spanof dura-
bilityunder theactionof eachfactor individuallycan
beshowninTable2.
4.1 The collection of decision-making evaluation
schemes and collection of attributions
As analyzed previously, the collection of decision-
making evaluation schemes can be described as:
X ={A, B, C, D, E}; the collection of attribution can
bedescribedas:
G={G
1
, G
2
, G
3
, G
4
}
T
={Stray Current Corrosion,
ChlorideIonsIngress, SulfateAttack, Carbonate}
T
4.2 The evaluating attribution matrix of subway
durability
Thevalueof attributioncanbedescribedasfollowing
matrix:
4.3 The relative quality matrix of evaluation
system of subway durability
Accordingtoformula(2), therelativequalitymatrixof
evaluationsystemof subwaydurabilityiswrittenas:
4.4 The weight of each factor in the evaluation
scheme
The ascertainment of weight adopts to the method
AHP. Accordingtothestraycurrent corrosionandthe
concentrationof CO
2
, analyzetheinfluenceof single
factor to thelife-span of durability of concrete. And
useDelphi methodtoanalyzeandget thecomparing
estimationmatrixbetweenanytwoonesinthesefac-
tors. Thestandardof ascertainedindex is: when one
compareswithanother one, theindex get 1if bothis
thesameimportant; theindexget3if theformeroneis
onlyabitmoreimportantthanthelatterone; theindex
get 5 if the former one is markedly more important
thanthelatter one; theindex get 7if theformer one
ismuchmoreimportant thanthelatter one; theindex
get 9if theformer oneis absolutely moreimportant
thanthelatter one. Thecalculationprocedureindetail
can beshown in reference(Wei 2002). Hereget the
solutionof comparison:
516
Use the method of square root to make certain the
weight of eachindexandget:
Theseweight valuesareavailablethroughproving.
4.5 Fuzzy evaluation of the system of
subway durability
Usingformula(3), thefuzzyvaluesare:
Therefore, the alternatives A
i
(i =1, 2, 3, 4, 5) can
be ranked as A
5
>A
3
>A
4
>A
2
>A
1
, the relevant
order of proportioningfromgoodto badis sortedas
followed: E >C >D>B>A
Obviously, theschemeE is best one. Therecom-
mendation uses thebest proportioning for theserial
number E.
5 CONCLUSION
Durabilityof subwaystructureisaproblemof compre-
hensivefunctionundermanyfactors,anditsevaluation
isacompounddecisionprocessof multipleattributes,
multiple factors and multiple indexes. In view of
thisproblem,thepaperdeterminedinfluencing-factors
of durability of subway structure for stray current
corrosion, chloride ions ingress, sulfate attack and
carbonation, establishedtheevaluatingmeasureindex
framework of durabilityof subwaystructure. Further,
FMADMistobeusedtocarryoutthemultiplefactors
evaluationonthebasisof singlefactor evaluationsin
thesubway structure. Then, thedurability fuzzy val-
ues of alternative scheme of concrete proportioning
areobtained. Finally, thesevaluesareorderedbytheir
sizeandmakethemclearwhichschemeisgoodorbad.
Thepresent resultsshowthat schemeE isthebest
proportioningschemeof subwayconcretewhichobvi-
ouslyenhancesdurablelife-span(thedurablelife-span
islonger than100year) bymixingflyashandmineral
powderreasonable.Therecommendationusesthebest
proportioningfor theserial number E.
It should be mentioned that superimposed effect
between various factors is not to be considered in
the paper. For instance, carbonation affect chloride
ionsingress, chlorideionsingressaffectsulfateattack.
Therefore, durability of subway structure consider-
ingsuperimposedeffectbetweenvariousfactorsneeds
further research.
ACKNOWLEDGEMENT
ThisworkwassupportedbytheNational Natural Sci-
enceFoundationof ChinaunderGrantNos.50678135.
The author would like to acknowledge Prof. Yang
Lindefor thisresearchproject.
REFERENCES
Du, Y.J. & Zhang, H.H. 2003. Elementary study in evalua-
tion of enduring span for HPC of subway engineering.
Northwest Water Resources & Water Engineering (3):
4953.
Heilpern, S. 1997. Representationandapplicationof fuzzy
numbers. Fuzzy Sets and System: 259268.
Leung, L.C. & Cao, D. 2000. On consistency and rank-
ing of alternatives in fuzzy AHP. European Journal of
Operational Research 124:102113.
Li, T.J. 2002. Theory and application of Fuzzy Multiple
Attribute Decision Making. Beijing: SciencePress.
Shen, G. 1998.The NO. 1 subway project. The series of design
and construction of shanghai municipal works. Shanghai:
Shanghai ScienceandTechnologyPress: 13.
Tan, W.Z. 2003. Holistic view of durability of concrete
structure. Architecture Technology (1): 1922.
Wei, R.Q. 2002. Operational research. Beijing: Tsinghua
UniversityPress,: 461466.
Zimmermann, H.J. 1987. Fuzzy Set Theory and its Applica-
tion. Kluwer2Nijhoff Publication, USA,: 5152.
517
Theuseof artificial neural networkstopredict groundmovementscausedby
tunneling
I. Chissolucombe, A.P. Assis& M.M. Farias
University of Brasilia, Brazil
ABSTRACT: Duringthedesignphaseof atunnel, oneof theconcernsof tunnel engineersisestimatingground
movementsinducedbytunneling, inorder totakestepstopreventor minimizepossibledamagecausedbythese
movements, to adjacent structures. Inthis work, anestimateis madeof displacements inthegroundbrought
onbytunnel excavationbyusinganartificial intelligencetechniquecalledArtificial Neural Networks(ANN).
Thecomputer tool utilizedwas theMatLabprogramandthenetworks wereof thefeedforwardtype, withthe
Resilient Backpropagation(trainrp) learningalgorithm.
1 INTRODUCTION
Tunnelling provokes changes in the stress state of
a ground and consequently, some displacements are
generatedthat spreadacross theexcavations zoneof
influence, which may induce damage in the struc-
tureslocatedtherein.Several methodologieshavebeen
utilized to estimate the displacements generated by
theexcavation of atunnel and thedamagethat such
displacements cancauseto thestructures lyingadja-
cent to theexcavationarea. Thesemethodologies are
distributed by the empirical, analytical and numeri-
cal approaches. The empirical approaches are easy
to use, but have the disadvantage of not consider-
ing the ground resistance and deformability param-
eters (Chissolucombe et al., 2005a). The analytical
approachesarealsoeasytouse, but havetheprimary
disadvantageof determiningthevarious coefficients
necessary for the use thereof. In using numerical
approaches, onemust haveaccess to atest program
that can supply the necessary parameters in a way
as to be able to use constitutive models that faith-
fullyrepresent theconditionsof thegroundandoften
haveasinthecaseof three-dimensional analysesa
highcomputational cost.
Theaimof this study is to estimatethedisplace-
ments of the ground induced by tunneling and the
inflectionpointof thesurfacesettlementsbasinutiliz-
ingoneof theartificial intelligence(AI) techniques,
called artificial neural networks. To verify which of
the input variables had the greatest impact on the
output variable of the neural network a sensitivity
analysis was carried out utilizing the values of the
weightsbetweenandamongthenetwork connections
asproposedbyGarson(1991).
Inthetrainingof theneural network, twodatasets
wereused, oneof whichwasobtainedbyinstrumenta-
tionduringtheexcavationof theMetr-DF(subway)in
thecityof Brasilia, andtheotherwasobtainedthrough
numerical simulationsbytheFiniteElement Method.
Inall of thesituations, thetrainingalgorithmutilized
wasResilientBackpropagation(trainrp),withsigmoid
transfer functions.
Theuseof two sets of dataobtaineddifferently is
duetothefollowingfactors: (a) theinput variablesof
theset of dataof Metr-DF weremostly geometrical
parametersof thetunnel or of theground, havingthe
SPT valueasasingleparameterof resistance; (b) need
to knowthebehavior of theneural network uponthe
presentationof theinputvariablesascohesionandthe
angleof friction, values obtained through numerical
simulations.
2 ARTIFICIAL NEURAL NETWORKS
A neural network is a massively parallel distributed
processor that has a natural propensity for storing
experiential knowledgeandmakesitavailablefor use.
It resembles the human brain in two respects: (a)
knowledgeisacquiredbythenetworkfromitsenviron-
mentthroughaprocessof learning, and(b) connection
strengths amongneurons knownas synaptic weights
areusedfor storingtheacquiredknowledge. (Haykin,
2001).
The most important property of Artificial Neural
Networks is their ability to learnfromtheir environ-
ment and to improve their performance with such
learning.Thislearningoccurswhentheneural network
reaches a generalized solution for a particular class
519
Figure1. Typesof partializationof thecrosssection.
of problems. In1970, Mendel & McClaren(citedby
Haykin, 2001) definedlearningasbeinganinteractive
process whereby thefreeparameters of aneural net-
work areadaptedthroughaprocessof stimulationby
theenvironmentinwhichthenetworkisinserted.This
definitionimpliesthefollowingsequenceof steps: (a)
theneural network is stimulatedby theenvironment;
(b) theneural network undergoesmodificationsinits
freeparametersasaresult of thisstimulation; and(c)
theneural network responds to theenvironment in a
newway, duetothemodificationsthat occurredinits
internal structure.
Theset of well-definedrules for thesolutionof a
learningproblemiscalledalearningalgorithm. There
arevariouskindsof learningalgorithms, andthemain
differenceamongthemisthewayinwhichtheweights
areadjusted.
3 MODELINGWITHARTIFICIAL NEURAL
NETWORKS
3.1 Modeling 1 (surface marks)
Inthismodeling, thedatabaseutilizedwascomprised
of dataobtainedfromtheinstrumentedsectionsduring
theexcavationof thetunnel for Metr-DF (subway) in
thecityof Brasilia.
TheBrasiliaMetrohasanextensionof 42km, being
12kmunderground. TheSouthWingsectionistotally
underground, with7.2kmof length, andencompasses
nine stations. These stations were built by the cut-
and-cover method and the tunnel excavated by the
sequential excavation method, following the NATM
(New Austrian Tunnelling Method) principles. The
geometry shape of the tunnel cross-section was an
ovoid, withanequivalentdiameter of 9.6m.Thetypes
of partializationof thecrosssectionof theMetrDF
tunnel are represented in Figure 1. The South Wing
tunnel of the Brasilia Metro was mostly excavated
inside a geological domain, called Paranoa group,
whichencompasseslayersof porousclayandresidual
Figure2. Thegeological profileof thetunnel fromtheSouth
Wing.
Table1. Input statisticsfor datasets.
Variables Mean Minimum Maximum
H(m) Training 10.24 6.70 14.30
Testing 10.03 7.70 12.90
D(m) Training 8.49 7.26 9.56
Testing 8.61 7.26 9.46
SPT roof Training 22.92 0.40 50.00
Testing 22.70 0.90 50.00
SPT side Training 3.75 1.50 8.50
Testing 3.75 1.50 8.50
SPT floor Training 8.72 3.50 18.00
Testing 8.79 3.50 17.00
soil of slate(Chissolucombeet al., 2005b). Thegeo-
logical profileof thetunnel fromtheSouthWing is
representedinFigure2.
Theinput variablesinthenetworkwereasfollows:
ground cover abovethetunnel roof (H), thetypeof
partializationof thesection(TS), theequivalentdiam-
eter of thesection(D), thelengthof thebalance(x),
thelevel of thewater table(N.A.), theSPT valueof
thetunnel roof, sides andfloor. Theoutput variables
werethevaluesof thestabilizedsurfacesettlementon
thetunnel axisobtainedfromadjustmentsafter taking
readingsof thesurfacemarksandtheinflectionpoint
(i). Themedium, minimumand maximumvalues of
someinput variablesarepresentedinTable1.
Theset of datawascomprisedof 74pairsof exam-
ples, 62 being utilized in the training phase and 12
inthetestingphase. Thetrainingwasperformedcon-
sidering each output separately; this procedure was
adoptedbecausetheperformanceof thenetwork was
better thanwhenbothoutputswereconsideredsimul-
taneously. Thearchitecturethat offeredbest general-
izationwas8-40-1for theoutput variablemaximum
surfacesettlement and8-10-1for theoutput variable
inflectionpointof thesettlementbasin. Figures3, 4,
5and6showtheresults obtainedduringthetraining
andtestingphasesof theneural network, andanaver-
ageof 3,000interactionswasnecessary toobtainthe
performancepresented.
520
Figure 3. Comparison between the predicted and mea-
suredresultsfor themaximumsurfacesettlement duringthe
trainingphase.
Figure4. Comparisonbetweenthepredictedandmeasured
resultsforthemaximumsurfacesettlementduringthetesting
phase.
3.2 Modeling 2 (numerical simulation)
In this modeling, the data obtained from two-
dimensional numerical simulations under conditions
of flat deformationwasutilized, withthePlaxisfinite
elementprogram(Brinkgreve&Vermeer,1998)devel-
opedatDelftUniversityof Technology. Thesectionto
beexcavatedwaspartializedandthenumerical simu-
lationobeyedthefollowingsteps: (a) generationof the
initial stressbasedoncoefficientof lateral earthpres-
sure(Ko); (b)headexcavationwithapplicationof 50%
of thenodal forcesintheareaaroundtheexcavation;
and (c) activation of thesupport with application of
theremainder of theload. Theexcavationof thesides,
nucleusandinvertedarchfollowedthesamesequence
described above. Thefiniteelement mesh was com-
prisedof 275triangular elements with6nodes each,
totaling636nodesand825Gausspoints.
resultsfor theinflectionpoint of thesettlement basinduring
thetrainingphase.
resultsfor theinflectionpoint of thesettlement basinduring
thetestingphase.
The model constituent adopted was the Mohr-
Coulomb, themainparametersof whicharecohesion
(c), angleof friction(),Youngmodule(E), andPois-
soncoefficient (). For themodeling, 432numerical
simulationswereperformed, varyingthecohesionval-
ues(10; 20; 30; 40; 60; and80kPa), angleof friction
(17
; 22
; 26
and30
),Youngmodule(3; 6; 10; 15; 20

and30MPa) andcoefficient of lateral earthpressure
(Ko 0.35; 0.55 and 0.60). Thevalueof thediame-
ter (D) of thesoil cover abovethetunnel roof andof
thePoissoncoefficientwereconstant(4.5m, 10mand
0.33respectively).
521
Figure7. Performanceof theneural networkinthetraining
phasefor theoutput variablemaximumsurfacesettlement.
Figure8. Performanceof theneural network inthetesting
phasefor theoutput variablemaximumsurfacesettlement.
The following variables: cohesion, angle of fric-
tion, Young module, and coefficient of lateral earth
pressure, constituted the input variables in the neu-
ral network, while the maximumsurface settlement
andthesettlement at tunnel roof constitutedtheout-
put variables in theneural network. Out of thetotal
of 432pairsof examplesavailable, 352wereutilized
in the training phase and 80 were used in the test-
ingphase. Thebest network architecturewas4-40-1,
and each output variable was calculated separately.
An average of 1,500 interactions was necessary for
trainingtheneural network inorder to obtainagood
generalization.
Figure9. Performanceof theneural networkinthetraining
phasefor theoutput variablesettlement at tunnel roof.
Figure10. Performanceof theneural networkinthetesting
phasefor theoutput variablesettlement at tunnel roof.
Figures7, 8, 9and10showtheresultsobtainedin
thetrainingandtestingphasesfor theoutputvariables
maximumsurfacesettlementandsettlementattun-
nel roof withrespectivecorrelationcoefficients(R)
of thebest linear adjustment. AccordingtoFigures7,
8, 9and10, excellent performancewas obtainedfor
bothvariablesinthetrainingandtestingphases.
4 SENSITIVITY ANALYSIS
Toverifytherelativeimportanceof eachinputvariable
in the estimate of the output variables, a sensitivity
522
Figure 11. Relative importance of the input variables in
the network to estimate maximumsurface settlement and
inflectionpoint (Modeling1).
analysiswasconductedaccordingtothethatproposed
by Garson (1991). Figure 11 shows the sensitivity
analysis for Modeling 1, i.e., for thecasewherethe
network was trained using data obtained by instru-
mentationduringtheexcavationof thetunnel for the
Metr-DF(subway)inBrasilia. Fortheoutputvariable
maximum surface settlement, the SPT resistance
parameter wastheonethat presentedthegreatest rel-
ativeimportanceanditsmaximumvaluewas32.26%
occurringfortheSPT inthetunnel side; theinputvari-
ablebalancewastheleastsignificant, witharelative
importanceof 3.84%. In caseof theoutput variable
inflectionpointof thesettlementbasin,themostsig-
nificantinputwasalsotheSPT inthetunnel side, with
avalueof 17.52%, andtheleast significant input was
balance, 7.13%.
In Modeling 2, where the data were obtained
through the numerical simulations with the Plaxis
finiteelement programandtheoutput variableswere
maximumsurfacesettlementandsettlementattun-
nel roof,therelativeimportanceof theinputvariables
wasnearlyidentical tothetwooutputvariables. Inthat
case, theresistanceparametersweretheonesthatpre-
sentedgreater relativeimportance, havingacohesion
valueof 48%, followed by aYoung moduleof 28%,
frictionangleof 15%, andcoefficient of lateral earth
pressurewithanaveragevalueof 9%(Figure12).
5 CONCLUSIONS
Thefollowingconclusionswerereachedafterthetrain-
ingof thetwo sets of datautilizingtheAI technique
calledartificial neural networks: (a)thecomputational
tool utilized proved highly effective in the estimate
of thedisplacementsinducedby tunnelingandof the
inflectionpoint of thesurfacepressurebasin, having
obtainedhighcorrelationcoefficient valuesandvery
minor errors when comparing the values measured
withthevaluesestimatedbythenetwork; (b) whenthe
Figure12. Relativeimportanceof theinputvariablesinthe
networktoestimatethemaximumsurfacesettlementandthe
settlement at tunnel roof (Modeling2).
objectiveistoexcavateatunnel inaregionwhereaset
of dataonpreviousexcavationsisavailable, theuseof
neural networkswill provokeadrasticreductioninthe
programsof geotechnical geological investigationand
measurements of theconstruction work; (c) thenet-
worktrainedwiththedataobtainedbyinstrumentation
duringtheexcavationof theMetr-DF tunnel exhib-
ited an inferior performance to the network trained
withthedataobtainedthroughnumerical simulations.
Thisisduetothenoisethatthedataobtainedinthecon-
structionworkpresents. Neverthelesstheperformance
of this network was reasonablewithaveragecorrela-
tion coefficients of 0.99 for training and 0.95 in the
test phase; (d) theneural networks present anadvan-
tage in relation to traditional empirical methods as
theypermittheconsiderationof variablessuchas:con-
structivemethod, water level, geometryof thetunnel,
and geometrical parameters and parameters of resis-
tanceof theground; (e) comparedwiththenumerical
methodsit hastheadvantageof estimatingwithhigh
precision the displacements induced by the excava-
tion, precisionthatisnotalwaysachievedinnumerical
analysesduetofactorslike: lack of anadequatecon-
stitutivemodel; difficultyinrepresentationof thereal
geometry; consideration of the tridimensional effect
inbidimensional analyses; (f) thesensitivity analysis
showed in thetwo modelings that theparameters of
resistance and deformability of the ground have the
greatestimpactontheestimateof displacements, with
theSPTvaluesonthesides,roof andfloorof thetunnel
inmodeling1andsoil cohesionandYoungs module
in modeling 2; (g) the main disadvantage of neural
networksisthelackof agoodset of datafor training.
REFERENCES
Brinkgreve, R.B.J. & Vermeer, P.A. 1998. Finite Element
Code for Soil and Rock Analyses. Plaxis Manual. Rot-
terdam, Netherlands.
523
Chissolucombe, I., Assis, A.P. & Farias, M.M. 2005a. Mto-
dos para previso dos deslocamentos do macio induzidos
por escavaes subterrneas. 5
Simpsio Brasileiro de
Aplicaes de Informtica EmGeotecnia (INFOGEO).
MinasGerais.
Chissolucombe, I., Assis, A.P. & Farias, M.M. 2005b. Soil-
structure interaction and its influence on displacements
induced by tunnel excavations.FifthInternational Sympo-
sium Geotechnical Aspects of UndergroundConstruc-
tioninSoft Ground. Netherlands.
Garson, G.D. 1991. Interpreting neural-network connection
weigts. AI Expert, 6(7): 4751.
Haykin, S. 2001. Redes Neurais: Princpios e Prtica.
Bookman, 2oedio, RS, 900p.
524
Researchandapplicationof roadtunnel structural optimization
W.Q. Ding&Y. Xu
Department of Geotecnical Engineering, Tongji University, Shanghai, P.R. China
ABSTRACT: Theoptimizationprocessof roadtunnel isestablishedinthispaperwithanengineeringexample.
Through numerical computation, we can find a serious stress concentration occurring at the spring of arch.
To solvethis problem, atunnel section shapeoptimization method is established at first. In this method, the
dimensionsof thetunnel sectionaretakenasthedesignvariables, thestressof springistakenastheobjective
function. No augmentation of theexcavated areaand thesatisfaction of construct boundary aretaken as the
constraint condition. A Complex Method is used as theoptimization method and C++ languageis used to
programand implement theoptimization. Using this optimization method, thestress condition of thetunnel
liningcanbegreatlyimproved. Then, theoptimizationmethodfor thethicknessandreinforcementof thetunnel
liningisalsoestablishedandtheconstructioncost isreduced.
1 INTRODUCTION
Withtherapiddevelopment of highway inmountain
area, thequantity of road tunnel becomes moreand
more. However, thecost of roadtunnel ishigher than
that of roador bridge, so, howtoreducethetotal cost
of theroadtunnel, andhowtoconstructroadtunnel in
themost economic way aretheimperativeproblems
whichneedtobesolvedbydesigners.
Optimizationdesignof theroadtunnel isaneffec-
tivemethodtoreducetheconstructioncost. But now,
intheoptimizationdesignof theroadtunnel, thetra-
ditional passive analysis method is used most. The
designers often define several feasible cases, then,
chooseonefromthemthroughcomputationandanaly-
sis.Thedisadvantageof thetraditional passiveanalysis
methodliesinthatitcannotcontainall conditions, and
it istimeconsuming. Andthenumerical optimization
method is not used widely, moreresearch is needed
andimperative.
Thenumerical optimizationmethodisanewbranch
of applied mathematics. In recent twenty years, it
develops very quickly with the widespread applica-
tionof computer.Applyingthenumerical optimization
method in civil construction design makes thetradi-
tional passive analysis develop to the active search.
Thisisagrateprogressinthecivil engineeringdesign,
andthenumerical optimizationmethodisamoresci-
entific, moreeffectiveandmoreeconomicmethodin
structural optimizationdesign.
But in the domain of optimization of tunnel and
undergroundstructure, therearemanyproblems. Due
to the complexity of tunnel and underground struc-
ture, agreat manyof parametersshouldbeoptimized
(suchasparametersof surroundingrock,parametersof
supportstructureanddecisionoptionsinconstructing
andsoon). Astotheoptimization, thereareproblems
such as multi-variable and highly non-linear, so the
researchanditsapplicationarestill inthestart stage,
thereferencesandexamplesarefew.
With an engineering example, an optimization
design method of automatic search is expounded in
this paper. Thesectionshapeandtheliningstructure
of thetunnel canbeoptimized. A ComplexMethodis
usedas thenumerical optimizationmethod, theC++
languageis usedto implement it. Withtheoptimiza-
tion analysis and design, the stress condition of the
structureisgreatly improved, andthecost of thetun-
nel is effectively reduced. This is of great practical
significancetotheconstructionof roadtunnel.
2 THETHEORY OF COMPLEX METHOD
2.1 The mathematic model of complex method
Themathematicmodel of complexmethodusuallycan
beexpressedasfollows:
Solve: designvariable X
Min f (x)
525
S.T (Subject toconstraint condition):
Inequationconstraint condition:
Boundaryconstraint condition:
where a
i
is the upper limit of the variable; b
i
is the
lower limit of thevariable.
2.2 The iterative process of complex method
2.2.1 The building of initial complex
The complex is composed of k(k >n +1) vertexes.
Therearetwomethodstodefinethesevertexes.
a. Definitivemethod
Designerscandefinethesevertexesbythemselves
accordingtothepropertiesof theproblem.
b. Randommethod
Thevertexesaredefinedbythefollowingformula:
wherei is thenumber of variable; j is thenumber of
vertex;
ij
isarandomvaluebetween0and1.
Thevertexes definedby randommethodconform
toall boundaryconditions, butdonotalwaysconform
toall inequationof constraint conditions.
Supposingthereares(1sk) vertexes conform
toall constraint conditions, thecenter X
s
of thevalid
vertexescanbedefinedbyfollowingformula:
Then-svertexesthatcannotconformtoall constraint
conditionscanbedealt withbyfollowingformula:
If thenewvertexes still can not conformto all con-
straintconditions,theyaredealtwithformula(5)again
until theycanconformtoall constraint conditions.
2.2.2 Search for reflection point
Figureout all thevalues of objectivefunction f (X
j
),
andfindouttheworstvertexX
h
whichmakesthevalue
of f (X
j
) maximum, X
c
isthecentreof thevertexesthat
donot includetheworst vertexX
h
.
Defineareflectioncoefficient (1), thereflec-
tionpoint X
a
canbedefinedbyfollowingformula:
Checkingwhether X
a
is valid, if not, reducingthe
valueof to its half, then deal X
a
with formula(7)
until it becomesvalid.
2.2.3 Compare the value of f (X
j
) at X
a
with that at
X
h
Therearetwopossibleresults:
If f (X
a
)-f (X
h
), i.e. X
a
isbetterthanX
h
, replaceX
h
withX
a
, thenewcomplexisformed, then, turntostep
2andgoon.
If f (X
a
)f (X
h
), i.e. X
a
isnotbetterthanX
h
, reduce
thevalueof to its half againuntil X
a
is better than
X
h
, thenturntostep2andgoon.
2.2.4 Criterion of convergence
There are many criteria of convergence, but the cri-
terion used most widely is that the values of f (X
j
)
at all vertexescanconformtoconstraint conditionas
follows:
wheref (X
s
) is thevalueof objectivefunctionat the
centrepoint; isalittlepositivenumber.
Find the vertex at which the value of objective
functionisleast, theoptimumvertexisfound.
3 OPTIMIZATIONOF SECTIONSHAPEAND
LININGOF ROADTUNNEL
3.1 General situation of the road tunnel
TheBai YangChongNo. 1tunnel isatwintunnel con-
nectedwitharc mid-wall whichis locatedinJ iangxi
province, China. It is 215mlong. The longitudinal
gradeof this tunnel is 2.499%. Thethickness of the
overburden ranges from10mto 20m. The stability
of thesurroundingrockisnotgood. Theunderground
water conditionis well. Thewidthof tunnel is about
11m, andtheheight isabout 5m.
3.2 The optimization problem of road tunnel
Lining is the most important component of tunnel
structure.Thestressconditionof liningisthekeyinflu-
encefactor of thestabilityandthelife-spanof tunnel,
it usually becomestheoptimizationobject whentun-
nel is designed. Whenoptimizingthetunnel section,
the stress condition of lining should be analyzed at
first.Themostimportantinfluencefactoronthestress
conditionof liningbecomestheoptimizationobject.
With theTongji GeoFBA numerical analysis soft
(a type of numerical analysis soft which can solve
two-dimensional problemwithfiniteelementmethod,
andit isdevelopedbytheDepartment of Geotecnical
EngineeringofTongji University),wecangettheinner
forceof thetunnel liningasfollows:
According to Figure 1 and quantities of calcula-
tionresults, it canbefoundthat themaximuminner
forces (including bending moment, shear force and
526
Figure1. Inner forcediagramof tunnel lining.
axial force) occur at thetunnel archspring. Thesec-
tion shape of this tunnel is monocentric circle, and
theinvert arch is joined by arc, i.e. thespring is the
linkagebetween theinvert arch and thearch ring. It
isthemost critical part of theliningring. Duetothis,
thesafety factor of theliningis reduced. Thereason
isthat thespringisthelink of twoarcswithdifferent
radius, anditsradiusissmall, thestressconcentration
occurseasily.Whenoptimizingthesectionformof the
tunnel, thedimensionparameter of this part is taken
as thedesignvariable, andthemaximuminner force
of liningbecomesthecontrol objective.
Tunnel lining takes about 50% of the total cost.
Afterimprovingthestressconditionbyoptimizingthe
sectionshape, thethicknessandreinforcement of the
tunnel liningcanthenbeoptimizedtoreducethelining
cost.
Generally, with an engineering example, theopti-
mization of section shape and lining design of road
tunnel isstudiedinthispaper. Byoptimizationdesign,
Figure2. Flowchart of theoptimizationdesign.
thestress conditionis greatly improved, andthecost
iseffectivelyreduced.
3.3 General process of the optimization design
Therearetwosteps of theoptimizationdesign. First,
optimizethesectionshapeof thetunnel,then,optimize
thicknessandreinforcement of thetunnel lining.
3.4 Optimization of the tunnel section shape
3.4.1 The building of optimization model
3.4.1.1 Selectionof thedesignvariable
AsshowninFigure3,todefinethetunnel section,there
are4independent variables: R
1
, R
2
,
1
,
2
. According
tothegeometrical relationship, R
3
,
3
canbedefined
asfollows:
Thestressconcentrationoccursat thelinkagepart,
thedimensionparameters of this part R
2
,
2
is taken
asthedesignvariables, R
1
,
1
istakenasconstant.
527
Figure3. Sketchof tunnel section.
whereR
1
istheradiusof roof; R
2
istheradiusof the
linkagepart; R
3
is theradius of invert arch;
1
is the
anglebetweenR
1
andspringingline;
2
isthecentral
angleof thelinkagepart;
3
is theanglebetweenR
3
andthecenter lineof tunnel section.
3.4.1.2 Thebuildingof theobjectivefunction
Theradiusof thelinkagepartissmall, thecurvatureis
big, this is thekey factor of thestress concentration.
So, the curvature of the linkage part is taken as the
objectivefunction, theoptimizationproblemisgetting
itsminimumvalue.
Objectivefunctionisshownasfollows:
3.4.1.3 Thebuildingof theconstraint conditions
R
1
,
1
areconstants, theinnercontourlineof thetunnel
sectionwill notenterintotheconstructionboundary.In
ordernottoincreasetheexcavationcost, theminimum
excavation areashould not belarger than that of the
initial design.
The minimumexcavation area can be defined as
follows:
Thentheconstraintconditioncanbegainedasfollows:
whereR
1
,
1
areconstants; R
2
,
2
aredesignvariables;
R
3
,
3
canbedefinedbyformulas(9) and(10).
Theupper limit andthelower limit canbedefined
accordingtothepractical situation.
3.4.2 Implement of the optimization process
Complex method is used as thenumerical optimiza-
tion method, programmed with C++ language, and
theoptimumsolution of thedesign variables can be
gainedbyrunningtheC++ program.
3.4.3 Optimization design and comparative
analysis
Definethevalues of every designvariableaccording
totheoptimumsolution.
Theinitial design:
Theoptimizationdesign:
Theoptimizationof roadtunnel is realizedby the
Tongji GeoFBA numerical analysis soft, the inner
forces of the tunnel lining can be gained after the
optimizationdesignasfollows:
AsshowninFigure4, themaximuminnerforcealso
occursat spring, but itsvaluehasbeenreducedalot.
Thestressconditionhasbeenimprovedgreatlybythe
optimizationdesign.
FromTable 1, we can find that the stress condi-
tionof tunnel lininghasbeenimproved, especiallythe
bendingmomentwhichisthedeterminingfactorof the
liningsafetyhasbeenreducedgreatly.Itillustratedthat
decreasingcurvatureof thelinkagepartisaneffective
measuretoimprovethestresscondition.
3.5 Optimization of the tunnel lining
3.5.1 The building of optimization model
3.5.1.1 Definingof objectivefunctionanddesign
variable
After the optimization design of the tunnel section
shape, thestressconditionhasbeenimprovedgreatly,
then, wecanoptimizethetunnel lining, andtakethe
liningcost astheobjectivefunction.
Costof thesteel concreteliningiscomposedof the
cost of steel andtheconcrete. Thecost of liningcan
bedefinedasfollows:
whereC
c
isthecostof concrete; C
s
isthecostof steel;
Z istheobjectivefunction.
528
Figure 4. Inner force diagram of tunnel lining of the
optimizedsectionshape.
Table1. Innerforceof tunnel liningcomparingbetweenthe
optimizationdesignandtheinitial design.
M
max
N
max
Q
max
(kN.m) (kN) (kN)
Initial design 311.1100 715.7400 285.8700
Optimizationdesign 191.6200 578.9500 221.4300
Reductionratio 38.41% 19.11% 22.54%
Multiplyingthequantityof consumptionbytheunit
pricecanget thecost of thematerials. Asthesection
shape is already determined, C
c
is only related to h
(thickness of the tunnel lining), as the thickness of
theprotectiveconcretelayer is aconstant, C
c
is only
related to h
o
(effective section height). The longitu-
dinal reinforcement is arrangedas theconstructional
reinforcement, the hoop reinforcement is the main
workingreinforcement. Soh
o
andA
g
(areaof thehoop
reinforcement of lining per meter) are taken as the
designvariablesof theobjectivefunction.
3.5.1.2 Buildingof theconstraint conditions
1) Constraint conditionof shear resistance
Checkthesafetyof thesectionwheretheshearforce
reachesthemaximumvalue.
whereK isthesafetyfactor, it canbedefinedaccord-
ing to the standard code; Q is the maximumshear
forceof thetunnel lining; R
a
iscompressivestrength
of concrete; bis thesectionwidth, whichis equal to
1000mm; h
o
isthedesignvariable.
2)Constraintconditionof compressiveandbending
resistance
Check the safety of the sections where the axial
force reaches the maximum value, or the bending
momentreachesthemaximumvalue.Thenwecanget
thefollowingequations.
Inbigeccentriccompressioncondition:
Insmall eccentriccompressioncondition:
or
3) Constraint conditionof boundary
The upper limit and the lower limit of the lining
thickness are determined according to the criterion.
Theupper limit andthelower limit of thereinforce-
ment areaaredeterminedaccordingtothemaximum
reinforcement ratio and theminimumreinforcement
ratio.
3.5.2 Implement of the optimization process
Complex Methodis usedas thenumerical optimiza-
tion method, C++ languageis used to programand
gaintheoptimumsolutionof thedesignvariables.
3.5.3 Optimization design and comparative
analysis
The values of every design variable can be deter-
mined according to the optimumsolution, then the
comparisoncanbedonebetweeninitial andoptimized
design.
Initial design: h =550cm, A
g
=1571mm
2
Optimizeddesign: h =450cm, A
g
=1451mm
2
where h is the thickness of the tunnel lining; for
symmetricarrangement, A
g
isthehoopreinforcement
areaof onesideof thetunnel lining.
For this tunnel, the lining cost of the optimized
design can be reduced by 12.1%, and 630,000RMB
canbesaved. Itcanconcludethatoptimizationdesign
is an effective measure to reduce the engineering
cost, whilethesafety of thestructurecanbeensured
simultaneously.
529
4 CONCLUSION
Fromoptimizationdesignof theroadtunnel andthe
comparative analysis, conclusions can be gotten as
follows:
1. When defining the objective function and the
design variables, many factors can be taken into
account, one step optimizing not only makes the
problemcomplex, but also get no effect results.
By dividingtheoptimizationproblemintoseveral
steps, the optimization model and operation pro-
cesscanbesimplifiedgreatlywhensolvedstepby
step, it canget goodandpractical results.
2. By numerical computationandanalysis, it canbe
foundthat themaximuminner forces occur at the
tunnel spring, thereasonisthatthecurvatureof this
part is small, and the stress concentration occurs
easily,reducingthecurvatureof thispartisaneffec-
tivemeasuretoimprovetheliningstresscondition,
andoptimizedsectionshapecanthenbegained.
3. Introducing optimization method into structure
designcanreducetheconstructioncost effectively
without loweringthesaftygrade, it shouldbeused
widespread.
4. Tosolvethenon-linear optimizationproblemwith
inequationconstraint condition, Complex Method
is an effective optimization method, its theory is
simple, andtheoperationiseasyandeffective.
ACKNOWLEDGEMENTS
Thispaper aresponsoredbyTheNational HighTech-
nology Research and Development Program (863
Program) of China, No. 2006AA11Z118, Popular-
ization Project of Key ResearchTechnology of Twin
Tunnel ConstructionandShanghai LeadingAcademic
DisciplineProject, Project Number: B308.
REFERENCES
Ding, W.Q., Yue, Z.Q., Tham, L.G., Zhu, H.H., LeeC.F. &
Hashimoto, T. 2004. Analysis of ShieldTunnel. Interna-
tional Journal for Numerical & Analytical Methods in
Geomechanics. 28: 5791.
J TGD70-2004. Road tunnel design code.
Liu, Y.J. & Gao, G.F. & Feng, W.X. 2004. Study on opti-
mizationdesignof crosssectionof thelarge-spanhighway
tunnel. Liao Ning communication science and techlonogy.
(2) : 4850.
Qian, N. 1999. C++ program design course. Bei J ing: Qing
HuaUniversityPress.
Zhang, B.H. 1998. Civil structure optimization design. Shang
Hai: Tongji UniversityPress.
530
Floor heavebehavior andcontrol of roadwayintersectionindeepmine
B.H. Guo&T.K. Lu
School of Energy Science & Engineering, Henan Polytechnic University, Jiaozuo, Henan, P.R.China
ABSTRACT: For investigatingfloor heavebehavior andcontrollingtechniqueof roadwayintersectionindeep
mine, creep deformation characteristics of floor around roadway intersection was studied by Flac
3D
and the
effectsof shear plasticcritical value, reinforcementmeasurementoncreepdeformationof floor aroundroadway
intersectionwerediscussed. Asaresult, thecreepdeformationcurvecanbedividedintotwostagesincluding
initial creepdeformationstageandsofteningdeformationstage, whichcanexplainthecasethat deformationof
roadwayintersectionislittlebeforeacertaintime, butthenincreasesatahighspeed; onlywhentotal shearplastic
valueexceedsthecritical valuecansecondcreepstagetakeplace; reinforcement measurement withboltingin
roof andribshaslittleeffectonfloor heave, yetexertingpressureagainstfloor canreducefloor heaveobviously.
1 INTRODUCTION
Rocksmassof roadwayindeepmineshowsoftbehav-
iorbecausetheyarelocatedinhighstressenvironment.
Duetoitslargercrosssectionandcomplexgeometries
deformation of floor around roadway intersection is
usually greater thanroadway doesnt intersect. Floor
heavehasagreat influenceonventilation, transporta-
tion, andso on, thus alot of literatures already exist
about this issue, but there are only a few in which
wasbasedonalternatingeffectsbetweensofteningand
creeping(Yanget al. 2006).
Strataof Northventilationroadwayat990mlevel
inTangkoucollieryissiltymudrock inmainlygreen
andcinereous, mingledwithalittlefinesandrock at
some locality. Protodrakonov scale of hardness f is
2.2, pressivestrengthis 19MPa, andtensilestrength
ranges from1.67MPato 4.45MPa. Theburial depth
of roadway is nearly from1029mto 1035m, maxi-
mummainstress is about 31.6MPa, andthevertical
stressisbetween25MPaand26MPa(Liuetal.2005),
whichislarger thanitspressivestrength, thussoften-
ingdeformationandcreepingdeformationwill occur
simultaneously (Wang et al. 1994). A casehas been
observed that the deformation of roadway intersec-
tion was less before about the 45th day after it was
formed, butthenacceleratedandconvergencebetween
roof andfloor reached565mmatthe90thday, among
which floor heave accounted for more than 70 per-
cent.Therefore, behavior andcontrollingtechniqueof
roadway intersection in deep mine was investigated
according to conditions of aroadway intersection at
990m level in Tangkou colliery, with effects of
softening/creeping of rocks mass interaction being
consideredbymeansof numerical simulationmethod.
2 NUMERICAL SIMULATIONSPROCEDURE
2.1 Failure criteria
Pwippmodel (avisco-plasticmodel combiningWIPP
model (the rock creep visco-elastic model) and the
Drucker-Prager model) is used in this study. Total
strainincludesdeviatestrainandaveragestrain. Devi-
atedstraincontainsthreecomponentsof elasticstrain,
plasticstrainandviscositystrain, whileaveragestrain
containstwocomponentsmentionedaboveexcept for
viscositystrain(Liuet al. 2005).
Strength of rocks mass decreases gradually along
with development of deformation after rocks mass
begin to failure(Wang et al. 2006), and thebearing
capacityof rocksmassinplasticregionsislower than
inelasticregions. Cohesionandfrictionanglelowerat
differentdegrees(Xiaoetal. 2005, Zhangetal. 2005)
alongwithdevelopmentof plasticdeformation, resid-
ual cohesion will lost totally andbearingcapacity is
providedbyonlyfrictionforce(You2005).Therefore,
strain-softeningmodel canbeusedtostudydeforma-
tionof roadwayintersectionundoubtedly(Yanget al.
2002).
Changeof strainespecially plastic strainexternal-
izestheloadingpathandhistory, andreflectssoftening
processof material frominitial conditiontofinal fail-
ure(Zheng2007). Referringto relationshipbetween
softeningprocessandplastic deformation(seeFig.1)
fromYou (2000), Diagrammatic sketch of full shear
plastic strain-softening curve (see Fig. 2a) was got.
Figure 2a shows that the softening coefficients of
strength k (theratio of stress corresponding to plas-
tic straintostrengthpeak value) increaseswithshear
plastic strain when shear plastic strain is below
p1
,
531
10
confining
pressure
0
10
20
30MPa
p
/mm
1.6 1.2 0.8 0.4
0
50
100
150
200
(
3
)

M
P
a
Figure1. Relationshipbetweensofteningprocessandplas-
ticdeformation(afterYou, 2000).
k
1
0
p1 p2 p
k0
1
1
0
k
pc p
(a)Full curve (b)Simplified curve
Figure 2. Diagrammatic sketch of softening coefficient-
shear plasticstraincurve.
decreasewhentheshear plastic strainis between
p1
and
p2
, and keep unchanged as a constant valuek
0
whenshear plasticstrainisabove
p2
. Because
p1
and
k
0
areusually very little, they arebothassumedas 0
here, thusthesimplifiedshear plasticstrain-softening
curve (see Fig. 2b) was obtained. Besides, when
total strainof rocks mass exceeds thestrainvalueat
peak strength, deformationmoduluslessengradually
withstrain, butresidual deformationmodulusisnever
reach0(seeFig.3).Otherresearchers(Zengetal.2005,
Liangetal. 2005, Chengetal. 2005, Chenetal. 2005,
Li et al. 2006, Qiang et al. 2006, Wang
et al. 2007) alsoreportedthesimilar opinion.
If wedefinek
1
asstrengthsofteningcoefficientand
k
2
as deformation softening coefficient, they can be
definedbyfollowingformulas:
0
2 4 6
1
(10
3
)
1
(
M
P
a
)
8 10
10
20
30
40
Figure 3. Stress-strain curves under cycle loading (after
Zhu, 1985).
Figure4. Simulatedmodel.
where
p
=shear plasticvalue; and
pc
=shear plastic
critical value, 2e-3here.
2.2 Model development
Assumption of rocks mass being homogeneous and
in hydrostatic-pressurestatewas madein this work.
Simulatedmodel containingaTshapedroadwayinter-
section and bolt supporting sketch were plotted in
Figure4 and Figure5 respectively. Thesizes of cal-
culatedmodel were40mlength, 40mwidthand40m
height.Thebottomof themodel wasfixedinall direc-
tions, foursideswerefixedinhorizontal direction, and
overburdenweightwasexertedonthetopof themodel.
Thesection of roadway was rectanglewith width of
4mandheight of 3m. Thelengthanddiameter of 2
anchorsboltedintheroof were6300mmand17.8mm
respectively, theinterval alongtheaxial andcircum-
ferential directionof theroadway were3000mmand
2400mmrespectively; for cables, thecorresponding
valueswere2300mm, 18mm, 1000mmand1200mm
532
Figure5. Bolt supportingsketch.
Table1. Mechanical parametersof rocksmass.
Bulk Shear Shear Tensile
modulus modulus Density strength strength
Parameters K G D
f

t
Units GPa GPa kg/m
3
MPa MPa
Values 8 4.8 2640 9.5 3.06
Table2. Creepparametersof rocksmass.
Activation Zone Gas WIPP-
energy, temperature, constant, Model
Parameters Q T R constant, D
Units J mol
1
K J mol
1
K
1
Pa
n
s
1
Values 50160 300 1.987 28.95e-36
WIPP-model WIPP-model WIPP-model
Parameters constant, A constant, B exponent, n
Values 22.8 25.4 4.9
Material Material Critical
parameter, parameter, steady-state
Parameters q
q
k
creeprate
Values 0.55 0.5 1.078e-8
respectively.Thecablesnear corner deviated20
from
normal directionof wall to corner. Accordingto rel-
ativereferences (Liu, J.H et al. 2005, Liu, T.S et al.
2005), mechanical parameters and creep parameters
of rocks mass werechosen and listed in table1 and
table2; mechanical parametersof cablesandanchors
werelistedintable3. Following4reinforcementmea-
surement methods were investigated by numerical
simulation: methodA was for naked roadway inter-
section, method B and C were for bolted roadway
intersectionwithoutandwithcablesappliedinfloorat
twocorners, andthelast methodwasmethodC com-
paniedwithexertedpressureagainstfloor,thepressure
valuesrangedfrom0.1MPato1.0MPawithinterval
of 0.1MPa.
Table3. Mechanical parametersof cablesandanchors.
Grout Grout
Elastic cohesion exposed
modulus strength perimeter
Parameters MPa MPa m
Cable 45 0.2 1
Anchor 195 0.42 1
Tensile Grout Cross-section
strength stiffness area
Parameters MPa MPa m
2
10
6
Cable 0.25 17.5 254
Anchor 1.85e3 5.35e3 249
3 RESULTS
3.1 Analysis of creep stages
XU et al. (2007) discussed that deformation pro-
cess of theroadway can bedividedinto threestages
includingadjustment deformationstage, stabledefor
mation stage and accelerated deformation stage by
investigationinthefield.
Figure 6 shows deformations of naked roadway
intersectionversus creeptimecurves under different
critical values. When
pt
>
pc
, Figure 6a illustrates
that creepcurvescanbedividedintotwocreepstages
named initial creep stage and softening creep stage
respectively, and each stage include a decelerated
creep stageand astablecreep stage. At initial creep
stage where the shear plastic value is lower than
thecritical value, deformation increases rapidly and
then comes to a nearly constant value until shear
plastic valueexceeds thecritical value. At softening
creepstage, deformationof theroadway intersection
increasesrapidly withcreeptimefor alonger period,
and comes to stable state finally. Therefore, we can
explainthecasementionedinsection1thatthedefor-
mationislittlebeforeacertaintime, butincreasesata
highspeedlater.
For naked roadway intersection, when
pt
-
pc
,
only initial creep stage occurs (see Fig. 6b). Roof
deformationraises rapidly withagradual decreasing
velocity, then approaches a relatively constant value
whilethedeformationvelocitycometoabout 0; floor
heaveraisesrapidly also, but it reachesapeak value,
then decreases a little, finally come to a lower con-
stant value. There is a deformation rebound in the
deformationprocessof floor heave.
When
pt
>
pc
, and roadway intersection is rein-
forcedby boltingandexertingpressureagainst floor,
therelationshipsbetweendeformationandcreeptime
areshowninFigure6c. Figure6c illustrates that the
533
Figure6. Deformation-creeptimecurvesunder threecon-
ditions,
pt
representstotal shear plastictrainvalue.
floor heaveisverylittlecomparedwithroof deforma-
tion;theprofileof roof deformationissimilarwiththat
showninFigure6a,yetfloorheavecurveisverydiffer-
ent. Thebeginningtimeof softeningcreepstagetakes
place late compared with that of roof deformation,
anditstotal deformationvalueislessgreatlythanthat
of roadwayintersectionwithout reinforcement shown
inFigure6a. Therefore, Reinforcement for floor can
reducefloor heaveeffectively.
Figure7. Therelationshipsbetweendeformations, vertical
stresspeak valueandreinforcement measurement: A, B and
CrepresentMethodA, methodB andmethodCdescribedin
section2.2; 0.11.0MParefertothepressurevaluesexerted
against floor basedonmethodC.
3.2 Controlling technique
Figure 7 illustrates that roof deformation decreases
abruptly when naked roadway intersection are rein-
forcedbyboltinginroof andribs, yet other reinforce-
ment measurements have little affection on it; floor
heavedecreasesalittlewhennakedroadwayintersec-
tion arereinforced by bolting in roof and ribs, even
when cables were applied in floor near two corners
(see Fig.7a), but floor heave lessens obviously with
increment of pressurevalueexertedonthefloor, and
vertical stress peak value around roadway intersec-
tion also decreases (seeFig.7b); vertical stress peak
values around roadway intersection with reinforce-
ment measurement areall lower thannakedroadway
intersection. Only when the pressure value against
floor is0.1Mpaandwithreinforcementmeasurement
method C, the vertical stress peak value has a little
rebound.Therefore, wecanreducefloorheaveamount
by exertingpressureagainst floor together withbolt-
ingsupport. If it is not enoughyet, measurements of
vertical cuttinginfloor (Guo, unpubl.) andsooncan
beusedadditionally.
534
Figure8. Relationshipbetweencreeptime, total deforma-
tionandshear plasticcritical value.
4 DISCUSSION
4.1 Affection of shear plastic critical value
Itisveryimportanttochoosereasonableshear plastic
critical valuewhennumerical simulationisperformed.
Figure8illustratesthatwhenthecritical valueisabove
2e-3, the second creep stage doesnt occur, and the
deformationof theroadwayintersectionisverylittle,
andthetotal creeptimethatcalculationlastisshorter.
Otherwise, thesecondcreepstagetakesplace. At the
second creep stage, the deformation of the roadway
intersectionislargerandthedeformationvalueof each
critical valueseemstobesimilar. Butalongwithincre-
mentof thecritical value,thebeginningandendtimeof
thesofteningcreepstagedelayalittle, andtotal defor-
mationloweralittle. Insummary, theshearplasticcrit-
ical valuedeterminesif thesofteningcreepstagetakes
place, andhasakeyinfluenceonthebeginningtime,
end time and total deformation of the second creep
stage. So creep deformation of roadway intersection
canbereducedbyimprovingtheshear plasticvalue.
4.2 Generation mechanics of creep stages
Researchworkers(Zhuet al. 2002, Wanget al. 2004,
Li et al. 2006, Fan2007) consideredthat rocks sam-
ples will failurefinally if shear plastic valueis large
Figure9. Distributionof highervertical stressandsoftening
zonedistributionsunder threeconditions.
enough. Whycanroadwayintersectioncometostable
statefinally?
Distribution of shear plastic zones (i.e. softening
zone(Wanget al. 2004)) andvertical stress filedare
plottedinFigure9. AstheresultsshowninFigure9,
thefinal shearplasticzoneissmall when
pt
-
pc
(see
Fig.9a),greaterwhen
pt
>
pc
(seeFig.9b),andcanbe
reducedbyreinforcementmeasurements(seeFig. 9c).
Whenrocksmassinthesofteningzonelosttheir bear-
ingcapacitytosomeextent, adjacentrocksmassmust
providemorebearingcapacity, if theycant, theywill
be softened also, and overburden pressure will con-
tinueseekinganother rocksmassuntil therocksmass
can bear it sufficiently and keep stable finally (see
Fig. 9d). Alongwithsofteningzones generationand
enlargement, peak stress valuein surrounding rocks
mass increases andits locationdiverts to rocks mass
thathasntbeensoftened. All inall, althoughsoftened
rocks mass adjacent to excavated roomlost bearing
capacityatadifferentdegree, butrocksmassadjacent
tosoftenedrocksmassprovidehigher bearingcapac-
ity, sothattheconstructionof roadwayintersectiondo
not lost itsstabilitycompletely.
5 CONCLUSIONS
The conclusions of this research are summarized
below:
1. Pwippmodel of flac
3D
softwarecanbeusedeffec-
tively to simulate creep deformation of roadway
intersection in deep mine. Strain-softening used
535
in this study includes softening of strength and
deformationmodulus.
2. Creep deformation-creep time curves can be
dividedintotwostagesincludinginitial creepstage
andsofteningcreepstage,andonlywhentotal shear
plasticvalueexceedsshearplasticcritical valuecan
softeningcreepstageoccur. Determinationof the
critical valueisvery important incalculation, and
creepdeformationof roadway intersectioncanbe
reduced by means of improving the shear plastic
value.
3. Bolting in surrounding rocks mass and exerting
pressure against floor can reduce deformation of
roadway intersection in deep mine. Method of
exerting pressure against floor has more obvious
effects to reducefloor heavethan other measure-
mentsusedinthisresearch, if itisnotenough, other
measurement canbeusedadditionally.
ACKNOWLEDGEMENTS
This work was carried out under the Outstand-
ingTalent Innovation Fund of Henan province (No.
0621000400).
REFERENCES
Chen, Z.J. &Yang, J.W. 2005. Measuresfor supportingdeep
HighStresscrack-expansioncreeprockmassinJ inchuan
miningdistrict. Metal Mine(1): 1822.
Cheng, R.H., Qian, M.G. & Miao, X.X. 2005. Numeri-
cal simulation in mining pressure control of thick and
strongstratumcavingbywater-infusionsofteningmethod.
Chinese J ournal of Rock Mechanics and Engineering
24(13):22662271.
Fan,Q.Z.2007.Experimental studyoncreepanditsdisturbed
effect of rocks. ChineseJ ournal of Rock Mechanics and
Engineering26(1):216216.
Li, Y.J., Pan, Y.S. & Zhang, M.T. 2006. Timeeffect onzonal
disintegration process of deep rock mass. The Chinese
J ournal of Geological Hazardandcontrol 17(4): 119122.
Liang, Z.Z., Yang, T.H. & Tang, C.A. et al. 2005. Three-
dimensional damagesoftenmodel for failureprocess of
heterogeneousrocksandassociatednumerical simulation.
Chinese J ournal of Geotechnical Engineering 27(12):
14471452.
Liu, J.H., Zhu, W.S. & Li, S.C. et al. 2005. Analysisof rhe-
ological characteristicsandstability of surroundingrock
massesof Xiaolangdi hydrojunctionundergroundcaverns
byusingFlac3D. ChineseJ ournal of RockMechanicsand
Engineering24(14): 24842489.
Liu,T.S., Zhao, J.Z. &Yang, J.E. etal. 2005. Supporttechnol-
ogy of bolt andanchor trussappliedinlargechamber of
Tangkoumine.Coal ScienceandTechnology33(5):1215.
Qiang, H., Zhou, H.Q. &Chang, Q.L. 2006. Regressanalysis
of mechanicscharacteristicof therockstrengthatthepre-
peak andpost-peak. J iangXi Coal Science&Technology
(2): 4850.
Wang, H.P., Gao, Y.F. & Li, S.C. 2007. Uniaxial experiment
studyonmechanical propertiesof reinforcedbrokenrocks
pre-and-post grouting. ChineseJ ournal of Underground
SpaceandEngineering3(1): 2731, 39.
Wang, J.A., J iao, S.H. & Xie, G.X. 2006. Study on
influence of mining rate on stress environment in sur-
roundingrockof mechanizedtopcavingminingface. Chi-
neseJ ournal of Rock MechanicsandEngineering25(6):
11181126.
Wang, J.J., Lu, Z.Y. & Liu, X.F. 2005. Discussion on floor
upheaval mechanismof mine soft rock roadway. Coal
Engineering(9): 6768.
Wang, L.G., WangY.J. &Liu, X. 1994.Therheological insta-
bility theory for rock sampleand its criteria. J ournal of
FuxinMiningInstitute(Natural Science) 13(3):9397.
Wang, X.Q., Yang, L.D. & Gao, W.H. 2004. Creep damage
mechanismand back analysis of optimumsupport time
for softenrockmass. ChineseJ ournal of RockMechanics
andEngineering23(5): 793796.
Xiao, H.F., He, X.Q. & Feng, T. et al. 2005. Research on
coupling laws between EME and stress fields during
deformation and fracture of mine tunnel excavation by
Flac3D simulation. ChineseJ ournal of Rock Mechanics
andEngineering24(5):812817.
Xu,X.L.,Zhang,N.&Xu,J.G.etal.2007.Principleandprac-
ticeof processcontrol oversoftbrokenroadwaywithhigh
groundstress. J ournal of Mining&SafetyEngineering24
(1):5155.
Yang, C., Chui, X.M. & Xu, S.P. 2002. Establishment and
studyof strain-softeningnumerical constitutivemodel for
soft rock. RockandSoil Mechanics23(6):695697.
Yang, C.H. &Li, J.G. 2006.Analysisof creepingmechanisms
of non-uniformitysoftrocks.J ournal of MiningandSafety
Engineering23(4):476479.
You, M.Q. 2005. Study of deformation and failure of rock
based on properties of cohesion and friction. J ournal of
Geomechanics11(3):287291.
You, M.Q. 2000. Strengthof rock specimensandprocessof
deformationandfailure. Beijing: GeologyPress.
Zeng, S., Yang, S.J. & Man, C. et al. 2005. Statistical
constitutivemodel for limestonerock damageunder uni-
axial compression and its experimental study. J ournal
of Nanhua University (Science and Technology) 19(1):
6972, 95.
Zhang, F., Dong, Z.H. & Ding, X.L. 2005. Numerical analy-
sisof excavationprocessfor tailracetunnelsof Pengshui
project. J ournal of Yangtze River Scientific Research
Institute22(6):5962.
Zheng,Y.R. 2007. Discussiononyieldandfailureof geoma-
terialsandstabilityanalysismethodsof slope/landslide
communion and discussion summary of special topic
forumon geologic disasters in the three gorges project
region. Chinese J ournal of Rock Mechanics and Engi-
neering26(4):649661.
Zhu, H.H. &Ye, B. 2002. Experimental studyonmechanical
propertiesof rockcreepinsaturation. ChineseJ ournal of
RockMechanicsandEngineering21(12):17911796.
Zhu, Z. F. 1985. Stiff test machine. Beijing: coal industry
press.
536
Squeezingpotential of tunnelsinclaysandclayshalesfromnormalized
undrainedshear strength, unconfinedcompressivestrengthand
seismicvelocity
M. Gutierrez
Colorado School of Mines, Golden, CO, USA
C.C. Xia
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University
ABSTRACT: Thetunnel squeezingphenomenonwasfirstdescribedbyTerzaghi (1946)whoassociatedsqueez-
ingmainly withclay-richrocks. Consistent withTerzaghis original description, thefocus of this paper is on
tunnelsinclaysandclay-richrockssuchasclayshales. Thepaper developsasimpleproceduretopredict tunnel
squeezingpotential usingnormalizedundrainedshear strength, unconfinedcompressivestrengthandP-wave
velocity. A collectionof alargeamount of undrainedtriaxial test dataisusedtoshowthat theundrainedshear
strengthof clayshalescanbenormalizedwithrespect totheeffectivevertical stress. Thenormalizedundrained
shear strengthcanthenbeusedtopredictthesqueezingpotential sinceitcanbedirectlyrelatedtoPecks(1969)
stability factor N. Values of N provideestimates of thedegreeof squeezing potential, with values of N >1
indicatingpotential for squeezing. Thenormalizedundrainedshear strengthof clay andshales is showntobe
relatedtotheapparent overconsolidationratioOCR, whichaccountsfor bothmechanical andother diagenetic
pre-consolidation, followingtheSHANSEP procedurefor clays (LaddandFoott, 1977). To facilitatetheuse
of theproposednormalizedundrainedshear strengthvs. OCR relationship, empirical equationsareestablished
to predict theapparent pre-consolidationstress fromunconfinedcompressivestrength
c
andseismic P-wave
velocity. Together withtheinsitueffectivevertical stressprior toexcavationat thetunnel location, theapparent
pre-consolidationfrom
c
orP-wavevelocitycanbeusedtoestimatetheapparentOCRandthencethesqueezing
potential.Theproposedapproachiscomparedwithavailablefielddataandtoexistingmethodstopredicttunnel
squeezingpotential.
1 INTRODUCTION
Barla (2001) defined squeezing as the large time-
dependent convergence during tunnel excavation. It
takesplacewhenaparticular combinationof induced
stresses and material properties pushes some zones
aroundthetunnel beyondthelimitingshear stress at
which creep starts. Deformation may terminatedur-
ing construction or continue over a long period of
time. Several other definitions of tunnel squeezing
havebeenproposedincludingthosebyGioda(1982),
ORourke (1984), Kovari (1988), Singh (1988), and
Aydan et al. (1993). A full review of the different
definitionsisgiveninBarla(2001).Ingeneral,squeez-
ing cannot always be distinguished from swelling
conditions(Steiner, 1993).
The tunnel squeezing phenomenon was first
described by Terzaghi (1946) who associated
squeezingmainlywithclay-richrocks.Consistentwith
Terzaghisoriginal description, thefocusof thispaper
is on clay-rich rocks such as clayshales, although
squeezing can also occur in other rock types. Sev-
eral procedures have been developed to predict the
squeezing potential of tunnels. One of the first sta-
bility criteriato predict squeezingwas developedby
Peck (1969) for tunnelsinclaysbasedonBromsand
Bennermarks(1967) stabilitycriterionfor openexca-
vations. Heproposed astability number N which is
expressedastheratiobetweenthetotal vertical stress
v
atthetunnel locationandtheundrainedshearof the
clayS
u
:
Valuesof N havebeencorrelatedwithobservations
of tunnel stabilityresponse, andthesecorrelationsare
537
Table1. Stabilitycriteriafortunnelsincohesivesoils(Peck
1969).
N Problemsencountered
15 Tunnelingwithout unusual difficulties
56 Claymaysqueezerapidlyintoshieldvoid
67 Shear failureaheadof tunnel causesground
movementsintothefaceeveninshieldtunneling
>7 General shear failuresandgroundmovements
aroundtunnel headingcauseshieldcontact to
becomedifficult; shieldtendstodive
summarized inTable1. Cases whereN >5 arecon-
sideredto havehighdegrees of rapidsqueezing, and
higher valuesof N indicatepotential for general shear
failure and large tunnel displacements close to the
tunnel heading.
For tunnels inrocks, several tunnel squeezingcri-
teria, which are mostly empirical, have also been
proposed. Singh et al. (1992) developed a criterion
based on the Q-systemof rock mass classification
(Barton et al. 1974) and the overburden height H
(inm) whichseparates thesqueezingcases fromthe
non-squeezingcases:
Tunnels with overburden height greater than that
giveninEq. (2) will experiencesqueezing.
Semi-empirical approaches have also been pro-
posed by Goel et al. (1995, 2000), J ethwa et al.
(1984), Aydan et al. (1993), and Hoek and Marinos
(2000). Thelast threecriteriauseastability number,
which is a reciprocal of Pecks stability. Almost all
these squeezing criteria summarized are empirically
or semi-empirically based on direct observations of
tunnel response. Inthesemi-empirical approachesof
J ethwaet al., Aydanet al., andHoekandMarinos, the
ratio of capacity and load (or strength and stress) is
analogous to thefactor of safety FS (or therecipro-
cal if theloadtocapacityratioisused). Inadditionto
linkingthedegreeof squeezingtothefactor of safety,
theadvantageof usingstressandstrengthisthatthese
canberelatedtostrainlevels, providedthefull stress-
straincurveareknown, whichinturncanberelatedto
degreeof squeezinginthetunnel.
The main challenge in the use of these semi-
empirical approaches is in the determination of the
rockmassstrength. InPecksapproach, thestrengthis
expressed in terms of the undrained shear strength,
while in J ethwa et al., Aydan et al., and Hoek and
Marinos, the strength is expressed in terms of the
unconfined compressivestrength of either theintact
rockor therockmass. It shouldalsobenotedthat the
proposedmethodologieshavebeendevelopedmainly
forclaysorhardrocks. Fewerstudieshaveinvestigated
the applicability of the above mentioned criteria to
intermediatematerialssuchashardsoilsor soft rocks
(e.g. HoekandMarinos, 2000). Basedontheseobser-
vations, themainobjectiveof thispaper istopropose
simplemethodstoestimatetheshear strengthof inter-
mediate soil-rock materials, particularly cemented
claysandshales, andtousethesimplemethodstoesti-
matesqueezingpotential intunnels. Estimates of the
shearstrengthof shalesareobtainedfromadatabaseof
triaxial testdataonseveral clayshales.Theseestimates
of shear strengthsarethenlinkedtofieldobservations
of squeezingtodevelopsimpleproceduresforthepre-
liminaryinvestigationof squeezingintunnelsinclays
andshales.
2 NORMALIZEDUNDRAINEDSHEAR
BEHAVIOR OF CLAY ANDSHALES
A procedure that is widely used to characterize the
undrained shear strength of clays is the SHANSEP
(Stress History and Normalized Soil Engineering
Properties) procedure developed by Ladd and Foott
(1977). According to this procedure, the undrained
shear strengthS
u
of normallyconsolidated(NC) clays
normalizedwithrespect tothecurrent effectiveverti-
cal stress
v
isunique, andfor overconsolidated(OC)
clays, thefollowingrelationshipadequatelyrepresents
thenormalizedundrainedshear strength:
whereOCR is theoverconsolidationratio definedas
theratiobetweenthemaximumpast effectivevertical
stress
p
and thecurrent effectivevertical stress
v
,
that is:
The parameter b is an empirical exponent, and
a =(S
u
,
v
)
NC
is the normalized undrained shear
strength of NC clay, i.e., the value of S
u
,
v
for
OCR=1.
Numerous studies in the literature have shown
the applicability of SHANSEP in representing the
undrained shear strength of many types of clays,
includingmarine, residual andglacial soils. Although
some studies have shown the successful use of
SHANSEPfor clayswithsomedegreeof cementation
(e.g. Bo et al. 2003), theapplicability of SHANSEP
to lithified materials like shales has not been fully
established.
SHANSEP has beensuccessfully usedinpractice
in geotechnical analysis and design, and it would
538
Figure1. Normalizedundrainedshear strengthasfunction
of apparent OCR for 25different typesof shales.
be valuable to extend the procedure to shales and
highlycementedcohesivematerials. Themaindiffer-
ence with clays is that overconsolidation and shear
strength in shales areeffected not only by mechani-
cal loading-unloading, butalsobytheotherdiagenetic
processes particularly cementation at the clay parti-
clecontacts. Theincreaseinoverconsolidationdueto
non-mechanical processesiscalledapparent or quasi
preconsolidation(e.g. BjerrumandWu, 1960).
To investigate the applicability of SHANSEP to
shales, adatabaseof triaxial test results on 25 types
of clayshalesfromdifferent locationswasassembled.
Includedinthis databaseareconsolidatedundrained
(CU) triaxial test resultsonshaleswithdifferent con-
solidation stresses. The normalized undrained shear
strength of 25 types of materials are plotted against
apparent OCR in Fig. 1. Only clayshales, which are
shales containing more than 50% clay particles by
weight, and stiff cemented clays are included in the
study. The apparent OCR is defined as ratio of the
current effectivevertical stress andtheapparent pre-
consolidation stress, i.e. OCR=
y
,
v
, where
y
is
simply theeffectivevertical stress at which yielding
canbeobservedfromtheexperimental consolidation
stress-strain curve. In caseof uncementedmaterials,
y
=
p
.
Theresults showalinear relationshipbetweenthe
log(S
u
,
v
)andlog(OCR),whichagreeswiththepower
functiongiveninEq. (3).Thereasonablygoodcorrela-
tionbetweenS
u
,
v
andOCR for 25typesof materials
is very promising, andindicates that SHANSEP can
provideareliableapproachtopredictingtheundrained
shear strengthof shales.Theaveragevaluesof a andb
forthe25differentmaterialsareequal to0.37and0.87,
respectively. In comparison, Ladd and Foott (1977)
Figure 2. Correlation between apparent preconsolidation
andunconfinedcompressivestrengthfor different shales.
obtained values of a =0.20 and b =0.77 for several
claysfromCUtriaxial tests. Itisimportanttonotethat
thedatashowninFig. 1arefor clayshaleswithdiffer-
ent degrees of diagenesis andcementation. Although
thenormalizedundrainedshear strengthof shalesam-
ples are similar, more lithified samples are actually
stronger(i.e., havehigherundrainedshearstrengthS
u
)
thanyounger uncementedsamples becauseof higher
valuesof theapparent preconsolidationstress.
3 DETERMINATIONOF THEAPPARENT
PRECONSOLIDATIONSTRESSINSHALES
Themost widely used laboratory approach to deter-
mine
y
experimentally is Casagrandes (1936) pro-
cedurewhere
y
corresponds to thesharpest bendin
consolidation plot. In addition to this approach, it is
useful to develop procedures to estimate the appar-
ent preconsolidationstressfromsimpleindexteststo
facilitatetheapplicationof SHANSEP toshales. One
parameter that can beobtained with relativeeasein
thefieldor inthelaboratory is theunconfinedcom-
pressive strength
c
. In the following, it is assumed
that theintact rock androck mass
c
arethesame, as
wasdonebyAydanet al. (1993), andHoekandMari-
nos(2000). Figure2presentsaplot of
y
vs.
c
(both
inMPa) for different clayshales, whichshows area-
sonablecorrelationbetweenthetwoparametersof the
followingform:
Although there is some scatter in the data, they
are in the range of typical
c
values of about 1 to
539
Figure 3. Correlation between apparent preconsolidation
andcompressional wavevelocityfor different shales.
70MPafor shales. It shouldbereasonableto expect
that a correlation exists between
y
and
c
since
bothcanbeconsideredasmaterial parameters, which
dependonthedegreeof mechanical consolidationand
cementation.
A moreconvenient procedureis to relate
c
with
datathatcanbedirectlymeasuredinthefieldusingin
situtestssuchasthecompressional wavevelocityV
p
.
Onesuchcorrelationobtainedbycurve-fittingthrough
experimental data, showninFig. 3interms of
y
(in
MPa) and compressional wave velocity V
p
(in m/s),
alsoassumedthesamefor theintactrockandtherock
mass, isobtainedas
Again, data are limited but are within the typical
range of V
p
of 2 to 5km/s for shales. To support
the reliability of Eq. (6), an additional relationship
isformulatedviathefollowingcorrelationfor shales
developedbyUrai (1995) between
c
(inMPa) andV
p
(inm/s):
SubstitutingEq.(7)in(5)providesthesecond
y
vs.
c
curveshowninFig. 3, whichisincloseagreement
withEq. (6). Thesimilarity of thetwo curves shown
inFig. 3appearstosupportthecorrelationbetween
y
andV
p
.Obviouslymoredataareneeded,butEqs.(5)to
(7) canprovidepreliminaryestimatesof theapparent
preconsolidationstressinthefield.
4 SQUEEZINGPOTENTIAL OF TUNNELSIN
CLAYSANDSHALES
Substituting Eq. (3) in Eq. (1) yields the following
expressionfor thestabilitynumber:
Thetotal andeffectivevertical stresses arerelated
tothedepthtothetunnel H andthetotal unitweight
andtheeffectiveunit weight
of thematerial above
thetunnel, i.e.
v
=H and
v
=
H. Boththeseunit
weights depend on thespecific gravity of thesolids
G
s
, theporosity n of thematerial, and thedegreeof
saturation. Assuming fully saturated conditions, the
ratio
v
,
v
canbeexpressedas:
Re-writingEq. (8)
andsolvingfortheeffectivevertical stressgivesarela-
tionshipbetween theeffectivevertical stress andthe
apparent yieldstress:
In turn, substituting Eqs. (5) and (6) in theabove
equationresultsinthefollowingrelationshipsbetween
theeffectivevertical stress, andtheunconfinedcom-
pressivestrengthandtheP-wavevelocity:
Equations (12) and (13) may beviewed as stabil-
ity criteria which relate the effective vertical stress
correspondingtothetunnel depth
v
, theunconfined
compressive strength
c
or P-wave velocity V
p
, the
stability number N, the porosity function F(n), and
the empirical constants a and b. Tunnels with
v
,
larger thanthosegiveninEqs. (12) and(13) havehigh
potential for squeezing.
540
Figure 4. Comparison of the stability criterion given in
Eq. (14)withobservedcasesof squeezingandnon-squeezing
intunnelsinclays(datafromBromsandBennermark, 1967).
5 COMPARISONWITHFIELDDATA AND
OTHER EMPIRICAL TUNNEL SQUEEZING
CRITERIA
Toshowtheirvalidity,thestabilitycriteriagiveninEqs.
(12)and(13)arecomparedwithfielddataonresponse
of tunnelsinclaysandclayshales. Thevalidityof Eq.
(12) for tunnels in clays is shown in Fig. 4, which
showsdatatakenfromBromsandBennermark(1967).
Thedatahavebeenre-plottedintermsof theeffective
vertical stress and unconfined compressivestrength.
A boundary curve between the non-squeezing and
squeezingcases is establishedusingEq. (12) incon-
junctionwithavalueof F(n)2.5(basedontypical
valuesof n 60%andG
s
=2.7forclays), andastabil-
ityof number of N =5.3.Also, valuesof a =0.20and
b =0.77 were used, which are representative values
for clays given by Ladd and Foott (1977). Substi-
tuting these values in Eq. (12) gives the following
criterion which demarcates cases of squeezing and
non-squeezingfor tunnelsinclays:
It canbeseenthat Eq. (14) providesaclear bound-
arybetweencasesof squeezingandnon-squeezingin
tunnelsinclays.
For clayshales, thevalidityof Eq. (12) isvalidated
by comparison with dataon Himalayan tunnels col-
lected by Bhasin (1991). Thedata of Bhasin (1991)
havebeen re-plotted as for thedatafor clay and are
showninFig. 5. A boundary curvebetweenthenon-
squeezing and squeezing cases is established using
Eq. (12) with values of F(n)1.8 (for typical val-
uesof n 30%andG
s
=2.7for shales), N =2.5, and
intunnelsinclayshales(datafromBhasin, 1991).
a =0.37andb =0.87takenfromFig. 1. Substituting
thesevaluesinEq. (12) givesthefollowingequation:
It canbeseenthat Eq. (15) providesaclear bound-
arybetweencasesof squeezingandnon-squeezingin
tunnelsinshales.
Figure6shows thecombineddatafromclays and
clayshales. A boundary curve demarcating the case
of squeezingfromnon-squeezingfor tunnels inboth
claysandshalesisestablishedfromtheaverageof the
stabilitycriteriagiveninEqs. (14) and(15).Thiscurve
is shown in Fig. (6), which shows that Eq. (15) can
adequatelyseparatethecasehistoriesof squeezingand
non-squeezingfor tunnel inbothclaysandclayshales.
Figure6alsoshowsthestabilitycriterionof Singh
etal. (1992)combiningEq. (2)withtheempirical rela-
tionshipbetween
c
andQ-valuealsofromSinghetal.
(1992). It can be seen fromFig. 6 that the Singhs
criterion significantly underestimates the boundary
between squeezing and no-squeezing for both clays
andshales.
Thefinal comparison with field datais shown in
Fig. 7 to show the validity of the criterion given in
Eq. (13). The unconfined compressive strength data
of Bhasin(1991) wereconvertedtoV
p
usingEq. (7).
Using similar parameters used for Eq. (15), thefol-
lowing stability criterion is obtained in terms of
v
andV
p
:
541
Eqs. (16) and (18) with observed cases of squeezing and
non-squeezing in tunnels in clays and shales (data from
BromsandBennermark, 1967; andBhasin, 1991).
intunnelsinshales(datafromBhasin, 1991).
Ascanbeseen, Eq. (17) providesanadequatecri-
terion for separating casehistories of squeezing and
non-squeezing in tunnels in clay and shales. Also
shownisthecriterionof Singhetal. (1992) re-plotted
intermsof V
p
byusingtherelationshipbetweenV
p
and
Q-valuedeveloped by Barton (2002). It can beseen
that Singhscriterionexpressedintermsof V
p
under-
estimates the boundary between cases of squeezing
andnon-squeezing.
6 CONCLUSIONS
Squeezingcriteriafor materialsthat rangefromclays
to clayshales wereproposed. Thecriteriawerebased
on normalized undrained shear strength as embod-
ied in the SHANSEP procedure for clays. It was
shownthat theundrainedshear strengthof clayshales
alsoexhibitnormalizedbehaviorsimilartounlithified
clays. In addition to relating stability to overconsol-
idation ratio, squeezing criteria in terms of uncon-
finedcompressivestrengthandP-wavevelocitywere
proposed.
The validity of the proposed criteria was demon-
stratedby comparisonwithfieldobservedsqueezing
in tunnels in clays and shales. It was shown that the
proposed criteriaarecapableof delineating cases of
squeezingandnon-squeezingintunnels inclays and
clay shales. Theproposedcriteriaprovidebetter pre-
dictions of squeezing in clays and clayshales than
other empirical criteriafor hardrockssuchastheone
proposedbySinghet al. (1992).
Theproposedcriteriacanbeusedfor preliminary
analysisof squeezingintunnelsinclaysandclayshales
wheretherearelimiteddata.Fortunnelswithsufficient
shear strength data, the paper has demonstrated the
possibilityof usingSHANSEP procedureinconjunc-
tionwithproject specific soil parameters to estimate
tunnel stability.
ACKNOWLEDGEMENT
This paper are based upon works supported by
the National Science Foundation under Grant No.
0324889andShanghai LeadingAcademicDiscipline
Project, Project Number: B308. Any opinions, find-
ings, andconclusionsor recommendationsexpressed
in this material are those of the authors and do not
necessarily reflect theviews of theNational Science
Foundation.
REFERENCES
Aydan, O., Akagi, T. & Kawamoto, T. 1993. Thesqueezing
potential of rock around tunnels: theory and prediction.
Rock Mech. Rock Eng. 2: l37163.
Barla, G. 2001. Tunneling under squeezing rock conditions.
LectureNotes,Eurosummer-School inTunnel Mechanics,
Innsbruck.
Barton, N., Lien, R. & Lunde, I. 1974. Engineeringclassifi-
cationof rock masses for thedesignof tunnel supports.
Rock Mech. 6(4): 189239.
Barton, N. 2002. SomenewQ-valuecorrelationstoassist in
sitecharacterisationandtunnel design. Intl. J. Rock Mech.
Mining Sci. 39: 185216.
Bhasin, R. 1991. Evaluation of soft rock conditions in tunnels
through the Lower Himalayan regions; A Contribution for
updating of the Q-system. MScThesis, Universityof Oslo,
1991.
Bjerrrum, L. & Wu, T.H. 1960. Fundamental shear strength
properties of the Lilla Edet clay. Gotechnique, 10(3):
101109.
542
Bo, M.W., Choa, V. & Hong, K.H. 2003. Material charac-
terization of Singapore Clay at Changi. J. Eng. Geol.
Hydrogeol. 36: 305319.
Broms, B. & Bennermark, H. 1967. Stability of clays at
vertical openings. Swedish Geotechnical Institute Publ.
No. 16.
Casagrande, A. 1936. The determination of the pre-
consolidationloadanditspractical significance. Proc. 1st
Intl. Conf. Soil Mech. Fnd. Eng., Cambridge, Mass. 60.
Gioda, G. & Cividini, A. 1996. Numerical methods for the
analysisof tunnel perform-anceinsqueezingrocks. Rock
Mech. Rock Eng. 29(4): 171193.
Goel, R.K., J ethwa, J.L. &Paithakan, A.G. 1995. Tunnelling
through the young Himalayas A case history of the
Maneri-Uttarkashi power tunnel. Engrg. Geol. 39: 3144.
Hoek, E. & Marinos, P. 2000. Predicting tunnel squeezing
problems in weak heterogeneous rock masses. Tunnels
Tunnel. Intl.: pp. 4551(part one), 3336(part two).
J ethwa, J.L., Singh, B. & Singh, B. 1984. Estimation of
ultimate rock pressure for tunnel linings under squeez-
ing rock conditions a new approach. In E.T. Brown,
J.A. Hudson (eds.), Design and Performance of Under-
ground Excavations, ISRM Symposium, Cambridge:
231238.
Kovari, K. 1998. Tunnelbau in druckhaftem Gebirge
Tunnellinginsqueezingrock. Tunnel 5: 1231.
Ladd,C.C.&Foott,R.1977.Newdesignprocedureforstabil-
ityof soft clays. J. Geotech. Eng. Div., ASCE, 100(GT4):
763779.
ORourke, T.D. 1984. Guidelines for tunnel lining design.
ASCE.
ground. State of the art volume, 7th Intl. Conf. Soil Mech.
Fnd. Eng., Mexico: 225282.
Singh, B., J ethwa, J.L., Dube, A.K. & Singh, B. 1992. Cor-
relationbetweenobservedsupportpressureandrockmass
quality. Tunnel. Undergr. Space Tech. 7: 5974.
Steiner, W. 1993. Swelling rocks in tunnels: Rock charac-
erization, effect of horizontal stress and construction
procedures. Intl. J. Rock Mech. Min. Sci. & Geomech.
Abstr. 30(4): 361380.
Terzaghi, K. 1946. Rockdefectsandloadsintunnel supports.
Rock tunneling with steelsupports. In R.V. Proctor &
T.L.White(eds.),The Commercial ShearingandStamping
Co., Youngstown, Ohio: 1799.
Urai, J.L. 1995. Brittleandductiledeformationof mudrocks.
EOS Transactions, American Geophysical Union, Nov. 7,
1995, F656.
543
Frameworkof performance-basedfireprotectiondesignmethodfor
roadtunnel
X. Han
Shanghai Institute of Disaster Prevention and Relief, Tongji University, Shanghai, P. R. China
G.Y. Ding
Shanghai Tunnel Engineering Co. Ltd, Shanghai, P. R. China
ABSTRACT: Under thebackgroundof performance-basedcodefor fireprotectiondesignof building, andon
thebasisof thestate-of-the-art of firesafety engineeringaswell asprescriptive-basedcodefor fireprotection
designof roadtunnel, thispaperoutlinestheframeworkof performance-basedfireprotectionanalysisanddesign
methodfor largecross-sectionroadtunnel. Theframeworkprovideapreliminaryguidanceontheapplicationof
scientificandengineeringprinciplestotheprotectionof suchroadtunnelsfromtheunwantedeffectsof fire.
1 INTRODUCTION
With rapid development of tunnel construction in
China, the number of road tunnels has increased
sharply to servetheperpetual growthof bothfreight
and transport for the populace. The consequence of
thegrowthinthenumber of roadtunnelsandthevol-
umeof traffic is an increasing occurrenceof severe
fire incidents in tunnels. Also, as tunnel construc-
tion technology has evolved, so the feasible length
andcross-sectionof roadtunnelshaveexpanded. Sub-
sequently, the control of such incidents has become
more difficult for the emergency services of large
cross-sectionroadtunnel.
Currently, the fire protection design of most
road tunnels are in accordance with prescriptive-
basedcodes.Prescriptive-basedcodesprovidespecific
requirementsfor broadclassificationsof tunnelsthat
establish acceptable or tolerable levels of risk for a
variety of health, safety, and public welfare issues.
These requirements are generally stated in terms of
fixedvalues, suchas maximumtravel distance, min-
imumfire resistance ratings, and minimumfeatures
of required systems (e.g. detection, alarm, suppres-
sion, and ventilation), and not in terms of overall
tunnel performance. However, the construction of
largecross-sectionroadtunnel wouldchallengemore
complicated fire protection requirement and should
be provided scientific and more cost-effective fire
protectionsolution.
Thispaperestablishedaframeworkof performance-
based fire protection analysis and design method
for large cross-section road tunnel. The framework
provided preliminary guidanceon theapplication of
scientific and engineering principles to the protec-
tion of such road tunnels fromtheunwanted effects
of fire. It alsoconstructedaprocessfor undertakinga
performance-based fire safety engineering approach
to road tunnel fire safety analysis and design. The
methodrequirestheuseof avarietyof toolsintheanal-
ysis, bringing increased engineering rigor and often
resulting in innovative design options. It allows the
safety levels provided by alternative design options
to be compared. In this way, it could result in a
comprehensivefireprotectionstrategyforlargecross-
sectionroadtunnel inwhichall firesafetysystemsare
integrated, rather thandesignedinisolation.
2 MAINFRAMEWORK
Theestablishment of performance-basedfireprotec-
tionanalysisanddesignmethodforlargecross-section
roadtunnel isunder thebackgroundof performance-
basedcodefor fireprotectiondesignof buildingand
theprescriptive-basedcodefor fireprotectiondesign
of tunnel. Thewholeperformanceof fireprotection
for road tunnel should be comprehensively consid-
ered. By means of developing design firescenarios,
thefireprotection goals, such as providelifesafety,
protectpropertyandfireproof of tunnel structure, have
tobeanticipated. Thereby, thefireprotectionlevel of
roadtunnel couldbesyntheticallyassessedaccording
todesigncriteriaandthecorrespondingfireprotection
measurescouldbedetermined.
545
The construction process of the main framework
involvesthreesteps: (1)ConceptEstablish; (2)System
Analysis; (3) SystemDesign. Amongthesesteps, step
of Concept Establish is to definite core need of the
method,stepof SystemAnalysisistosetuptotal model
of theframeworkandstepof SystemDesignistocreate
theimplement system.
2.1 Concept establish
Inorder toenablethemethodbeappliedintheprac-
tical fireprotectionengineeringof largecross-section
road tunnel, the fundamental concept of the frame-
workshouldbeestablishedonthebasisof currentfire
protectioncodes.
During the evaluation process of fire protection
feature for large cross-section road tunnel, the per-
formance of corresponding fire protection facilities
could be confirmed. The traffic volume in the road
tunnel could beanticipated and therelated firepro-
tectionmanagement couldbeascertained. Ingeneral
situation, the assumed conditions tally with practi-
cal project. As for somespecial situation, thefurther
researchworkshavetobecarriedout.
2.2 System analysis
Based on fundamental concept of framework estab-
lishment, facingqualifieddesigner, authority having
jurisdiction,ownerandmanagementpersonnel involv-
ingfireprotectionof roadtunnel, thebasicframework
of fireprotectionmethodfor largecross-sectionroad
tunnel isconstructed.Theframeworkconsistsof three
modules, includingComprehensiveEvaluation, Infor-
mationIntegrationaswell asPerformancePrediction
respectively.
Theframework of performance-basedfireprotec-
tionanalysisanddesignmethodforlargecross-section
isshowninFigure1.
Themainfunctionof modulesisasfollows.
Module1: ComprehensiveEvaluation
This module is the key part of the framework,
includingmainfunctionssuchas:
1. Defineobject of fireprotectiondesign;
2. Establishdesigncriteriaof fireprotection;
3. Evaluate comprehensively the fire protection
features.
Module 1 is appropriate for stakeholders of
road tunnel, including road tunnel owner, qualified
designer, authorityhavingjurisdictionandinsurer, etc.
Module2: InformationIntegration
This module is the basic part of the framework,
includingmainfunctionssuchas:
1. Collect correspondingdesignmaterials according
tocurrent prescriptive-basedcode. Determinedif-
ferentkindsof parametersforfireprotectiondesign
of roadtunnel;
Module 1:
Comprehensive
Evaluation
Fire Safety
Engineering
Prescriptive-based Fire
Protection Design Code
Module 2:
Information
Integration
Module 3:
Performance
Prediction
Figure 1. Framework of performance-based fire protec-
tionanalysisanddesignmethodfor largecross-sectionroad
tunnel.
2. Collect basic data of fire protection design. The
detailed design parameters included road tunnel
feature, vehiclevariety, firecharacteristic andfire
protectionfacility, etc.
3. Definiteothercomprehensiveperformance,includ-
ingbackgroundinformationrelatedwithfirepro-
tectiondesign.
Module2 is applicablefor qualified designers of
roadtunnel.
Module3: PerformancePrediction
Thismoduleisthequantitativeanalysispart of the
framework, includingmainfunctionssuchas:
1. Definedesignfirescenario.Accordingtostatistical
dataof tunnel fireanddesigngoal of fireprotection,
thedesignfirescenariosshouldbechosen.
2. Definite performance anticipation approach. The
analyses may be deterministic or probabilistic,
deterministicmethodwouldnormallybeused.
3. Predictsystemperformance. Takingthewholefire
processasobject of study, all componentsor sub-
systems which possibly affect the fire protection
feature of the road tunnel should be thoroughly
analyzed.Asforperformanceevaluations, thecom-
ponents or subsystems of the fire safety system
aswell astheinteractionsbetweenthesubsystems
wouldneedtobeconsidered. Thesubsystemsthat
may beconcerned include: (1) fireinitiation and
development; (2) spread, control andmanagement
of smoke; (3) firedetection; (4) firesuppression;
(5) personnel behavior andegress; (6) passivefire
protection.
546
Module 3 is applied for professional assessment
personnel andresearcher.
2.3 System design
In accordance with the function specificities deter-
minedbythethreemodulesforfireprotectionanalysis
anddesignmethodof largecross-sectionroadtunnel,
thedetailedcontentof correspondingimplementation
structureconstructedby SystemDesignis presented
inFigure2.
2.3.1 Define scope of large cross-section road
tunnel
The scope of large cross-section road tunnel is an
identification of the boundaries related with the
performance-basedanalysisordesign, including: road
tunnel constraints, designandconstructionteamorga-
nization, projectschedules, applicableregulationsand
other useful information to assist in comprehending
thescopedefinition.
2.3.2 Determine design goal and objective of
fire protection
Fireprotectionof largecross-sectionroadtunnel gen-
erally has four interrelated fundamental fire safety
goals: (a) Provide life safety for the public and fire
fighters. Minimize fire-related injuries and prevent
undueloss of life. (b) Protect roadtunnel. Minimize
damageto road tunnel fromfireand fireprotection
measures. (c) Provide for continuity of road tunnel
operations(i.e., protectthepassageof vehicles). Min-
imize undue loss of operations due to fire-related
damage. (d) Limit the environmental impact of fire
andfireprotectionmeasures.
Oncethefireprotectiongoalshavebeenestablished
and agreed to, thedesign objectives to meet thefire
protection goals must bedefined. Thedesign objec-
tiveprovides moredetail thanafireprotectiongoal,
andisoftenstatedintermsmeetingtherequirements
of aspecificcodeor standardprovision(prescriptive-
orperformance-based), of aspecificinsurance-related
requirement,orinadditiontoaspecificcode,standard,
or insuranceprovisionor requirement.
2.3.3 Collect related performance information
There are many types of information that may
affect road tunnel design performance that should
be considered and collected, including characteris-
tics such as: vehicle and personnel, location of the
tunnel, fire service, utilities, environmental consid-
erations, tunnel management and security, economic
and social value of the tunnel, the tunnel delivery
process, applicableregulations, etc: It isquiteimpor-
tant to identify the appropriate codes, regulations
andinsurancerequirementsfortheperformance-based
analysis.
Evaluate
performance
of fire safety
system
Assess
evacuation of
personnel &
vehicles
start
Define scope of large
cross-section road tunnel
Determine design
goal and objective
of fire protection
Collect related
performance
information
Select fire protection
strategies for tunnel
Meets
criteria ?
Develop trial design of
fire protection for tunnel
Yes
No
end
Determine design fire scenario
Establish design criteria
of fire protection
Figure2. Fundamental content of performance-basedfire
protectionanalysisanddesignmethodforlargecross-section
roadtunnel.
547
2.3.4 Establish design criteria of fire protection
The design criteria of fire protection for road tun-
nel in this paper are mainly concerned performance
criteria. Performance criteria are usually thresh-
old values, ranges of threshold values, or distribu-
tions that are used to develop and evaluate trial
designs for a given design of large cross-section
road tunnel. Performance criteria may include tem-
peratures of materials, gas temperatures, smokecon-
centration or obscuration levels, radiant flux levels,
and human response, decision, reaction, and move-
ment times. For example, performance criteria may
includevaluesforthermal radiationexposure(kW/m
2
)
or gas (air) temperature. Other types of perfor-
mance criteria include concentration of toxic gases
(ppm), distance of the smoke layer above the floor
(m), visibility(m), or other measurableor calculable
parameters.
2.3.5 Determine design fire scenario
Incasetheperformancecriteriahavebeenestablished,
the qualified designer needs to focus on the devel-
opment and analysis of design alternatives to meet
these criteria. The possible fire scenarios should be
considered arethen filtered into selected design fire
scenarios. As design firescenarios constructed, then
trial designs of road tunnel could be developed and
evaluatedto determinewhether theperformancecri-
teriawill besuccessfullymet bythetrial designfor a
givendesignfirescenario.
A firescenariorepresentsoneof aset of firecon-
ditions that arethought tobethreateningtoacertain
roadtunnel. For agivenfirescenario of roadtunnel,
there are many factors that may affect fire develop-
ment. These different factors may include: (a) form
of ignition source; (b) different items first ignited;
(c) ignition in different locations of a road tunnel;
(d) effectsof tunnel geometry; (e) ventilation, whether
longitudinal ventilationor transverseventilation, etc;
(f) formof intervention(i.e. personnel, sprinklers, the
firedepartment, etc.).
2.3.6 Develop trial design of fire protection
for tunnel
After theperformancecriteriahavebeen established
and thedesign firescenarios havebeen determined,
thetrial designof theroadtunnel shouldbedeveloped.
Developing trial design may simply require select-
ing features similar to that of theprescriptivebased
designoption, but withenhancedcapabilities or fea-
tures. The design features being developed should
giveconsiderationtothecapabilities, reliability, costs
and maintenance. However, it should be considered
that thetrial designwouldbeevaluated. Trial design
canbeevaluatedonasystemperformancethat relies
on an evaluation relativeto established performance
criteria.
2.3.7 Evaluate trial design of fire protection
for tunnel
Evaluation is the process of determining if a trial
design meets the performance criteria during the
postulated design fires. Theintent is to demonstrate
that in thedesign firescenario, performancecriteria
will not beexceeded.
If thetrial designof theroadtunnel is foundsuc-
cessful, anyremainingtrial designsmaybeevaluated
asnecessary. If oneof trial designsisnot foundtobe
successful, thenit may bemodifiedandre-tested, or
it may be dismissed. After the selected trial designs
have been tested, a final design should be selected
fromamong those found successful. If there are no
trial designs that arefound successful, thequalified
designershouldensurethatthetrial designsconsidered
all possiblemitigationstrategies. If after considering
all possible mitigation strategies, there still are not
anytrial designsthat arefoundsuccessful, thedesign
goal and objective of fire protection as well as the
performancecriteriashouldbereexamined.
Many techniques canbeusedto evaluatetheade-
quacyof atrial designfor roadtunnel incaseof fires.
Thesetechniquestypicallyfit intotwoprincipal cate-
gories,probabilisticanddeterministic.Adeterministic
analysis examines thehazard posed by thepotential
design fire scenarios independently. A probabilistic
evaluationusesriskanalysistoidentifyconsequences
of specificeventsandtheir respectivelikelihood.
Theevaluationof aperformance-basedfireprotec-
tion design for a large cross-section road tunnel is
a function of several factors. These factors include:
(a) complexity of road tunnel geometry; (b) level of
subsysteminteraction; (c) type of performance cri-
teria; (d) sensitivity of subsystemoutput to design
objectives; (e) absolute or comparative evaluation;
(f) knowledgelevel; (g) benefitversuscost; (h) expert
judgment andexperience. Theevaluationshouldalso
account for knownvariationsanduncertainties.
Themainevaluationcontent of thetrial designfor
road tunnel may concern evaluating performance of
correspondingfiresafety systemaswell asassessing
evacuation of personnel and vehicles, etc. It is nec-
essary to point out that timelines could be valuable
tools in evaluation. Therefore, it may beessential to
determinethetimeof keyeventssuchas: (a) ignition;
(b) fire detected; (c) evacuation begins; (d) unten-
ableconditionsreachedinroadtunnel; (e) firespreads
scope; (f) suppressionbegins; (g) failureof structural
elements; (h) fireextinguished.
2.3.8 Select fire protection strategies for tunnel
Intermof performance-basedfireprotectiondesignof
alargecross-sectionroadtunnel, whiletheacceptable
trial designsareidentifiedbytheevaluation, theycan
beconsideredfortheselectionof fireprotectionstrate-
gies. The choice of which acceptable trial design is
548
selectedfor thefinal designmaybebasedonavariety
of differentfactors,includingfinancial considerations,
timeliness of installation, systemand material avail-
ability, easeof installation, maintenanceanduse, and
other factors.
Oncethefireprotectionstrategiesareidentified,the
relevantdesigndocumentsneedtobeprepared. Proper
documentationwill ensurethatall stakeholdersunder-
standwhatisnecessaryfor thestrategiesimplementa-
tion, maintenanceandcontinuityof thefireprotection
design. Thedocumentationshouldincludethedesign
brief, aperformancedesignreport, detailedspecifica-
tionsanddrawings, andaroadtunnel operationsand
maintenancemanual.
3 CONCLUSION
It is evident that the framework of performance-
based fire protection analysis and design method
for a large cross-section road tunnel offers a num-
ber of advantages over prescriptive-based design. It
provides a basis for development and selection of
alternativefireprotection options based on theroad
tunnels needs and results in a composite fire pro-
tection strategy in which all fire safety systems are
integrated, rather than designed in isolation. Hence,
such a comprehensive engineering approach often
provides more cost-effective fire protection solution
for the road tunnel. Apparently the intensive study
about theframeworkshouldbefurther conducted.
ACKNOWLEDGMENTS
REFERENCES
Eberl, G. 2001. TheTauernTunnel Incident: What happened
andWhathastobeLearned, Proc. 4th Int. Conf. on Safety
in Road and Rail Tunnels, Madrid, Spain, 26th April
2001: 1730.
Han, X., Shen, Z.Y. et al. 2000. Study on a framework of
performance-basedassessment methodfor firesafety of
large public building, Proceeding of the Sixth Interna-
tional Symposium on Structural Engineering for Young
Experts,YunnanScienceandTechnologyPress,Kunming,
China, August 2000: 1820.
Lacroix, H.D. 2001.TheMontBlancTunnel Fire: WhatHap-
pened andWhat Has Been Learned, Proc.4th Int. Conf.
on Safety in Road and Rail Tunnels, Madrid, Spain, 26th
April 2001: 316.
Turner,S.2001.St.GotthardTunnel Fire,NewCivil Engineer,
1st Nov. 2001: 57.
549
Predictionof surfacesettlementsinducedbyshieldtunneling: AnANFIS
model
J. Hou, M.X. Zhang& M. Tu
Department of Civil Engineering, Shanghai University, Shanghai, P.R. China
ABSTRACT: A newmethod fuzzysystemcombinationneural networkwasusedtoestimategroundsurface
settlements.Accordingtothemeasureddataof Shanghai No.2Subway, andconsideringvariouskindsof factors
synthetically, anANFISfuzzyneural networkpredictionmodel wasbuilt. Comparingwiththepredictionresults
byotherthreekindsof methods, thevalidityof theANFISfuzzyneural networkmodel wasappraised. Confirmed
by theinstance, ANFIS fuzzy neural network is valuablein predicting ground settlements induced by shield
tunneling.
1 INTRODUCTION
Shield tunneling has becomeoneof themost popu-
lar methodsusedintheconstructionof urbantunnels,
suchasrapidtransitsystemsandlargediameterunder-
groundpipelines. However, shieldtunnelingconstruc-
tion inevitably disturbs the ground and the original
stress fieldof soil, which in turn causes surfaceset-
tlement that may yield damage of existing adjacent
structures andundergroundfacilities. Therefore, it is
of significant importancefor engineers to accurately
predict thesurfacesettlement during thedesign and
constructionstages.
Considering the complexity of the problem that
involves intricate geological makeup of the ground
composedof differentmaterialswithvaryinglayerpat-
ternsplusdifferentconstructionmethods,itisapparent
thatusingclassical methodmakesitdifficulttoprovide
accuratepredictionsof groundsettlement. Inaddition,
it isvery difficult todetermineinput parametersrep-
resentative of the mixed geological compositions in
apredictionmodel. Therefore, theanalytical methods
relyingonobserveddataarewidelyusedininvestiga-
tionof surfacesettlement. A hybridintelligentsystem
calledANFIS(J ang 1993) (Adaptive-Network-Based
Fuzzy Inference System) combining the ability of a
neural network tofuzzy logic havetheadvantages of
bothneural networks(e.g. learningabilities, optimiza-
tion abilities, and connectionist structure) and fuzzy
systems (e.g. humanlike if-then rules, and ease of
incorporating expert knowledgeand judgment avail-
able in linguistic terms). Such a hybrid intelligent
system holds much potential in prediction. In this
paper, amodel basedonANFISisproposedtopredict
thegroundsettlement inducedbyshieldtunneling.
Figure1. Architectureof ANFIS.
2 DEFINITIONOF THE MODEL
2.1 Brief description on the
Adaptive-Network-Based Fuzzy Inference
System (ANFIS)
Bothartificial neural networkandfuzzylogicareused
in ANFISs architecture (Chang & Lee 2003). The
ANFIS is one of the methods to organize the fuzzy
inferencesystemwith given input-output datapairs.
TheANFIS optimizes the parameters of consequent
part usingleast squaremethodandthoseof premise
part usingsteepest descent method.
For simplicity, weassumethefuzzyinferencesys-
temunder considerationhastwoinputs(x andy) and
oneoutput(z) inthepresentstudy. Thecorresponding
ANFISarchitectureisshowninFigure1.
Layer 1: In this layer x and y represent different
typesof input parametersof theadaptivenodei of an
adaptivefunction, andA
i
isthelinguisticlabel associ-
atedwiththis nodefunction. Parameters inthis layer
arereferredtoaspremiseparameters.
Layer 2: Every nodein this layer is a fixed node
labeledH whichmultipliestheincomingsignalsand
551
sendstheproductout. Eachnodeoutputrepresentsthe
firingstrengthof arule.
Layer 3: Every nodein this layer is a fixed node
labeledN. Theithnodecalculatestheratioof theith
rules firing strength to the sumof all rules firing
strengths.
Layer 4: Every nodei in this layer is an adaptive
node. Parameters in this layer will be referred to as
consequent parameters.
Layer 5: The single node in this layer is a fixed
nodelabeledYthatcomputestheoverall outputasthe
summationof all incomingsignals.
Thesteepestdescentmethodcanbeappliedtofind
thepremiseparametersandleast squareestimatecan
heappliedtooptimizetheconsequent parameters.
2.2 Input and output parameters
A relatively too largesurfacesettlement may induce
damageorevencrashof existingstructures. Itisthere-
foreverycrucial toaccuratelypredictthetotal surface
settlement(themaximumsurfacesettlementaftercon-
struction) of apoint beforetheshield arrives at this
point. It isstraightforwardtoset thetotal groundsur-
facesettlementtobeanoutputvariableinaprediction
model. Inthispaper, similartomanyotherresearchers,
wechoosethetotal surfacesettlement 5metersahead
of the shield face as an output parameter. However,
howtochoosetheinput variablesdependsonthecol-
lectionof monitoreddata. Systematic analysesof the
measureddatashowthat thedisturbedareaby shield
tunnelingof theoriginal stressfieldof soil isapproxi-
matelylimitedinazonewithin15mfromtheworking
point, andalmost hasnoeffect ontheareaout of this
zone. M. Karakuss measured data showed also that
themaximumsettlement isover thetunnel centerline.
Consideringthesetwoaspects, wechoosesix param-
eters, i.e., fivetotal groundsurfacesettlementsat the
pointsover thetunnel centerline(i.e., 0, 5, 10, 20and
30metersbehindthetunnel face) andoneof theshield
workingparameters, thetotal number of theworking
cyclesof theday, astheinputs.
Theinvestigationinthispaper isbasedonthemea-
sureddataintheshieldingtunnel fromZhongshanPark
stationtoLongdongRoadstationintheShanghai No.2
SubwayTunnel Project. Thetotal lengthof thetunnel
is 1624m. The earth pressure balance (EPB) shield
wasusedintheproject.Theouterdiameterof theEPB
machineis6340mmandthelengthis6540mm. The
tunnel shield tunneling was started in 18 J uly, 1997
and completed by 9 November, 1997, and the sur-
facesettlementsweremeasuredfrom21August.Large
numbersof surfacesettlement markerswereinstalled
tomeasuresurfacesettlementsduringexcavation.The
installationof surfacesettlementmarkersisdescribed
in detail by Sun andYuan. The soil profile around
the tunnel can also be found in the paper. The col-
lecteddatahaveformedadatabase. Thedatabasenot
Table1. Samplesusedfor networktraining.
Serial
number X
1
X
2
X
3
X
4
X
5
X
6
y
1 11.65 24.70 31.60 34.35 34.85 5 21.20
2 1.50 7.65 24.00 36.40 42.29 5 55.25
3 6.80 52.10 72.57 89.67 98.75 8 70.30
. . . . . .
81 3.75 14.85 38.20 36.85 39.70 6 57.05
Table2. Samplesusedfor networkvalidationtesting.
Serial
number X
1
X
2
X
3
X
4
X
5
X
6
y
1 5.20 20.30 26.75 23.20 9.95 8 40.95
2 0.05 1.07 0.35 2.45 4.95 12 62.45
3 13.07 54.42 82.62 96.60 25.45 11 89.87
4 10.00 27.25 63.00 81.10 91.60 11 53.50
5 14.45 51.30 57.85 74.85 90.52 10 54.50
6 4.75 12.65 28.55 38.11 46.20 4 57.00
7 6.80 52.10 72.57 89.67 98.75 8 83.15
8 3.20 9.75 2.60 1.42 2.27 11 99.57
9 4.17 7.32 7.32 17.00 44.90 7 41.10
10 14.05 16.15 14.39 22.1 34.50 8 47.22
11 0.95 2.85 16.25 54.45 47.65 12 76.42
12 3.30 15.55 33.30 45.30 52.00 6 53.90
13 1.70 15.60 30.25 47.05 55.40 11 47.30
14 2.95 4.95 13.75 44.15 86.40 10 61.30
15 13.25 43.75 76.30 74.85 68.52 10 56.73
16 3.75 22.02 47.77 66.90 72.75 13 88.15
17 4.57 18.75 46.40 68.75 47.15 7 79.02
18 4.00 11.65 34.05 47.65 77.40 8 61.30
19 0.72 6.92 18.37 43.37 51.65 10 79.60
20 1.90 12.20 30.40 49.65 39.30 11 71.95
onlyallowsonetostudythebehaviorof groundmove-
mentsoccurringduringexcavation, but alsobecomes
auseful sourcefor developingpredictivemodels for
thegroundsettlement. Werandomly chose81inthe
databaseexperimental data(fromAugust to October
in1997, seeTable1, detailedlist canbefoundinTu
(2005)) asatrainingsetand20dataasavalidatingset
(datainNovember, seeTable2).
Here, X
1
, X
2
, X
3
, X
4
, X
5
is total ground surface
settlements (unit: mm) at the points over the tunnel
centerline(i.e., 0, 5, 10, 20and30metersbehindthe
tunnel face) separately. X
6
is thetotal number of the
workingcyclesof theday.
2.3 Data pretreatment
To speeding calculation, we standardize the original
datatohaveaminimumof 0andamaximumof 1by
552
0
6 12 18 24 30 36 42 48 54 60 66 72 78
20
40
60
80
100
120
Prediction
Experiment
s
u
r
f
a
c
e

s
e
t
t
l
e
m
e
n
t
(
m
m
)
serial number of measured point
Figure 2. The fitting map of the ANFIS output and the
measureddata.
linear transformation. It isshownthat suchatransfor-
mation will makethecalculation very efficient. The
algorithmof thelinear transformationisasfollows.
The original data of six input variables can be
expressedas:
Wenotethat i =1, 2, . . . P is thenumber of samples
andtotally 81samples areusedinthis paper to train
themodel, i.e., P =81.
Wenormalizeeachinput datumas
Thenthenormalizedinput datacanbeexpressedas
3.1 Text and indenting
The present model is realized via a Matlab pack-
age of ANFIS. The fitting map of the ANFIS out-
put and the measured data was obtained (Fig. 2).
Table 3. The comparison of predicted surface settlement
andmeasuredsettlement.
Measured Predicted Absolute Relative
Serial settlement Settlement error error
number (mm) (mm) (mm) %
1 40.95 39.24 1.71 0.042
2 62.45 63.3 1.15 0.018
3 89.87 92.34 2.47 0.027
4 53.5 52.46 1.04 0.019
5 54.5 52.59 1.91 0.035
6 57 58.38 1.38 0.027
7 83.15 81.55 1.6 0.019
8 99.57 103.28 3.71 0.037
9 41.1 38.25 2.85 0.069
10 47.22 45.39 1.83 0.039
11 76.42 74.23 2.19 0.029
12 53.9 54.96 1.06 0.02
13 47.3 48.71 1.41 0.03
14 61.3 62.37 1.07 0.017
15 56.73 54.42 2.31 0.041
16 88.15 86.29 1.86 0.021
17 79.02 80.16 1.14 0.014
18 61.3 63.91 2.61 0.043
19 79.6 78.22 1.36 0.017
20 71.95 73.28 1.33 0.018
The comparison of predicted surface settlement and
measuredsettlement wasalsoshowninTable3.
Theyall showthatall thepredictedresultsfromthe
ANFIS model areingoodagreement withmeasured
data.
Peck (1969) presented the first available method
for estimatingthegroundsettlement duetotunneling
andexcavation. Inhismethod, chartsweredeveloped
basedonthefielddataobtainedfromsubwayconstruc-
tions at different places worldwide. Thedatapoints,
though scattered, revealed the shape of a Gaussian
distributioncurve. Thecharts havebeenwidely used
forestimatingthetransversegroundsettlementprofile
causedbytunnelingandexcavation. Theoretical mod-
elsbasedonthecombinedpi-sigmaapproach(Gupta
& Rao1994) andBP (Back Propagation) neural net-
work arealso used to predict thesurfacesettlement.
Dueto thelimitation of thepaper length, wedo not
discussthemhere.
Comparison between the above methods and the
ANFISmethodisshowninFig. 3.
It is found that the current prediction is in good
agreement with measurement. The maximum error
betweenthemodel andthetestingdataisnotmorethan
7%. For comparison, weproposedalso thepredicted
results fromtwo theoretical models basedontheBP
(BackPropagation) neural networkandthecombined
pi-sigmaapproach(whicharefrequently usedinpre-
dictingthesurfacesettlement), andthosefromPecks
equation (Fig. 4). We can see fromthe figure that
553
0 5 10 15 20
30
40
50
60
70
80
90
100
110
120
measured data
ANFIS
Combined pi-sigma neural network
BP package
Peck formula
s
u
r
f
a
c
e

s
e
t
t
l
e
m
e
n
t
(
m
m
)
Figure3. Comparison of predicted results with measured
ones.
0 5 10 15 20
0.10
0.05
0.00
0.05
0.10
0.15
0.20
0.25
r
e
l
a
t
i
v
e

e
r
r
o
r
ANFIS
Combined pi-sigma neural network
BP package
Peck formula
Figure4. Comparisonrelativeerror fromANFIS toother
methods.
theBP neural networkhasthemaximumerror among
thesemethods; that thecombinedpi-sigmaapproach
underestimates but thePecks formulaoverestimates
thesurfacesettlement.
4 CONCLUSION
Thispaper hasshownthepotential for applyingfuzzy
neural networks to ground surfacesettlement analy-
sis. fivetotal groundsurfacesettlementsat thepoints
over the tunnel centerline (i.e., 0, 5, 10, 20 and 30
meters behindthetunnel face) andoneof theshield
workingparameters, thetotal number of theworking
cycles of theday, as input parameters to predict the
surfacesettlementinducedbytunnel shieldtunneling.
Thepredictedresultsfromtheproposedmodel arein
goodagreement withfieldobservationsandthemax-
imumerror betweenthemodel andthetestingdatais
not morethan7%inour example. Comparisonwith
someotherpredictingapproachshowsthatthepresent
model isaccurate, steadyandefficient.
REFERENCES
Chang,K.H.&Lee,J.J.2003.AdaptiveNetwork-basedFuzzy
InferenceSystemwithPruning. SICEAnnual Conference
in Fukui, 46 August 2003. Japan: Fukui University.
Gupta, M.M. & Rao, D.H. 1994. Ontheprinciplesof fuzzy
neural networks. Fuzzy Sets and Systems, 61(1):18.
J ang, J.S. 1993. ANFIS: Adaptive-Network-based Fuzzy
InferenceSystems. IEEE Transactions on Systems, Man
and Cybernetics, 23(3): 665685.
Karakus, M. & Fowell, R.J. Effects of different tunnel face
advance excavation on the settlement by FEM. Tun-
nelling and Underground Space Technology, 2003, 18(5):
513523.
ground. Proceedings of 7th International Conference on
Soil Mechanics and Foundation Engineering, Mexico.
Sun, J &Yuan, J.R. Soil disturbanceandgroundmovement
under shield tunnelling and its intelligent prediction by
usingANNtechnology. Chinese Journal of Geotechnical
Engineering, 2001, 23(3):261267. (inChinese)
TuM. Thepredictionof thegroundsettlements inducedby
basedontheFuzzyNeural Network. Shanghai university
masters thesis. 2005. (inChinese)
554
Experimental studiesof ageological measuringsystemfor tunnel with
ultrasonictransducer
D.H. Kim, U.Y. Kim, S.P. Lee& H.Y. Lee
GS Engineering & Construction, Seoul, Korea
J.S. Lee
Korea University, Seoul, Korea
ABSTRACT: Predicting ground conditions ahead of the tunnel face has been one of the most important
requirementsof tunnel construction.Thisstudyinvestigatedthedevelopmentandapplicationof ahighresolution
ultrasonicwaveimagingsystem, whichcapturesthemultiplereflectionsof ultrasonicwavesat theinterface, to
detect discontinuitiesat laboratoryscalerockmassmodel. UltrasonicwavereflectionimagingbasedonA- and
B-modes was obtained through stacking, signal compensation, demodulation, and display. Experiments were
carriedoutbyusinghorizontal scanningandrotational scanning. Experimental studiesshowedthattherotational
scanningmethodwasabletoidentifyhorizontal andinclineddiscontinuitiesandthecavityontheplaster block
atafixedlocation. Furthermore, twodiscontinuitiesincludinghorizontal andinclineddiscontinuityplaneswere
detected. Therotatingscanningtechniqueproducedimagessimilar tothoseobtainedby thetypical horizontal
scanningtechnique. Thispaper containsbasictheoriesabout theultrasonictransducer andseveral experimental
applicationresults. Thefull-scalefieldapplicationandother applicationwill bescheduledinthefuture.
1 INTRODUCTION
With therapid growth of theworlds population and
economics, increasing number of tunnels have been
constructed to provide better transport links. Safe
andeconomic tunnelinghas always beenachalleng-
ing topic in tunnel construction becauseof complex
groundconditions likefaults, fractures, caverns, and
wet layers. Therefore, theprediction of ground con-
ditions aheadof tunnel faceis consideredoneof the
important requirements.
Many techniques to predict the ground condi-
tion ahead of the advancing tunnel face have been
developed, improved and widely applied in tunnel
construction projects. Three types of techniques are
representives for tunnel groundconditionpredicting:
coring, displacement monitoring and analysis, and
wave-basednon-destructiveevaluations(Figure1(a)).
In this study, anew high resolution seismic tech-
niquebasedonrotatingscanningequipment is intro-
duced to detect geological discontinuities, as shown
inFigure1(b). This techniqueuses thereceivedtime
series which captures multiple reflections due to
the impedance mismatch at a geological interface.
Thesourceandreceiver transducersarelocatedonthe
tunnel faceat onefixedpoint. Asapreliminarystudy
of this technique, this paper describes experimental
studies carried out on lab-scaled models. This paper
Figure 1. Seismic method for predicting tunnel face:
(a) TSP method; (b) Rotatingscanningmethod.
presentsbasictheoriesabouttheultrasonictransducer
includingtransducer frequency characteristics, trans-
ducer beamcharacteristics, and coupling layer char-
acteristics, experimental setupandsignal processing,
applicationexamples, discussion, andconclusions.
2 ULTRASONICTRANSDUCER
Ultrasonicwavereflectionimagingsystemmaybean
economical and effectivetool for theforecast of the
ground condition ahead of the tunnel face. A trans-
ducerreferstoanydevicethatcanconvertanelectrical
signal into a mechanical energy and vice versa. In
ultrasonic reflection imaging, a piezoelectric lead-
zirconate-titanate(PZT) typetransducer is generally
used to as a source and detector (Wells, 1977). The
selectionof thetransduceristhemostimportantfactor
555
inultrasonicreflectionimagingbecausethetransducer
determinestheimageresolutionandskindepth.
2.1 Transducer frequency characteristics
The choice of transducer frequency in ultrasonic
reflection imaging is the result of a compromise
betweentheresolution(lateral andaxial) requirement
andtheacquirement of satisfactorybeampenetration
for theimaging of thepart of interest. Lateral reso-
lution refers to theability to distinguish two closely
spacedreflectors, whicharepositionedperpendicular
totheaxisof theultrasoundbeam. Lateral resolution
is most closely related to thetransducer beamwidth.
Axial resolutionreferstotheminimumreflector spac-
ingalongtheaxisof anultrasonicbeamthatresultsin
separate, distinguishableechoesonthedisplay.
For agiven frequency, theshorter thepulsedura-
tion, thewider thefrequency bandwidth. If ashorter
pulsedurationisused, highresolutioncanbeattained,
but it lowersthesensitivityandskindepth. But sensi-
tivitycanbeimprovedbyeither increasingtheenergy
of thetransmitter or by amplifyingthecapturedsig-
nals at thereceiver. Note, skindepthis theability to
detect ananomaly at agivendepthandit dependson
the amplitude of the reflected signal. When the res-
olutionincreases withincreasingfrequency, theskin
depthdecreases(LeeandSantamarina2005).
2.2 Transducer beam characteristics
Near FieldandFar Field. Thetransducerbeamischar-
acterizedasnearfieldandfarfield, whicharesketched
inFigure2.Inthenearfield,whichiscalledtheFresnel
zone, waveamplitudefluctuates. Notethebeams are
almost parallel rather thandivergent inthenear field.
Thenear field length NFL or Fresnel zonelength is
dependent onthetransducer radius r andwavelength
(Krautkramer andKrautkramer, 1990; Rose, 1999).
wheref isthefrequency, andV istheultrasonicwave
velocity of the medium. Therefore, the near field
length NFL increases with theincreaseof thetrans-
ducer radius r and/or theincreaseof thefrequency f.
Thezonebeyondthenearfieldisthefarfield, whichis
alsocalledtheFraunhofer zone. Thedivergenceangle
dependsonthedirectivity.
Directivity. Thefar fielddivergenceangledepends
onthewavelengthandradiusr (or diameter d) of the
transducer (Zagzebski, 1996), asshowninFigure2.
The divergence angle should be made small to
increase lateral resolution. Higher frequency and
larger diameter render higher lateral resolution (far
fieldisassumed).
Figure2. Beamcharacteristicsof transducer.
Figure3. Transducer typeandfocal length: (a) Schematic
drawingof transducer; (b) Focal length.
Thedirectivity of thetransducer is affectedby the
velocityof thetestmedium, andthefrequencyandsize
of thetransducer. Highdirectivity yieldshigher reso-
lution. Thedirectivity of an ultrasonic beammay be
alteredbyfocusingthebeam: 1) acurvedPZTelement
insteadof aplanePZT elementisusedand2) alensis
attachedinfront of theplanePZT element (Schmerr,
1998), asshowninFigure3. Becausethewaveveloc-
ityinlensisgreater thanthatinwater, theshapeof the
lenswill beconcave. Thefocal lengthF isdefinedas
thedistancefromthepoint onthecurvedsurfaceon
thecentral axistothemidpointof theregionof conver-
gence. Note, if thelensisused, thefocal zoneinstead
of thefocal pointappears, asshowninFigure3(b).The
focal lengthisdependentontheradiusof curvatureR,
theradiusof thelensa, andtheapertureangle2 (see
Figure3).
For the investigation of the directivity, two types
of transducers areused: focal typetransducer (Pana-
metricsA3441)andnon-focal transducer(Panametrics
V318). Botharehighdampingimmersiontransducers
having500kHz inresonant frequency and19mmin
diameter. Directivities were investigated by measur-
ingtheamplitudeof thesignal atfixedaxial distances
(50mm, 100mmand150mm) andthelateral offsets
of thetransducers, as shown in Figure4(a). A focal
556
Figure 4. Directivity: (a) Test procedures; (b) Focal type
transducer; (c) Non-focal transducer; (d) Lens focal type
transducer.
Figure 5. Coupling layer effects: (a) Vacuumgrease-not
adheredtightly;(b)Vacuumgrease-adheredtightly;(c)Water.
typetransducer hasgooddirectivity, asshowninFig-
ure4(b).Whereasanon-focal typetransducerdiverges
much of theenergy, as shown in Figure4(c), anon-
focal typetransducer withaconcavelens (lens focal
typetransducer) improvesthedirectivitydramatically,
asshowninFigure4(d).
Coupling layer characteristics. Thecouplinglayer,
which is an agent between transducer and medium,
shouldminimizethereflectionfromthesurfaceof the
mediumtestedtomaximizetheenergy transferredto
themedium.Vacuumgreaseandwaterwereexamined
as agents in this study. Thetest mediumwas aplas-
ter block of 300mminheight, 300mminwidthand
150mminthickness. Theminor energywasreflected
withvacuumgreasewhenthetransducerswerelightly
placed on the top of the vacuum grease (it is not
adhered tightly), as shown in Figure 5(a). However,
whenthetransducersweretightlyadhered, theampli-
tude of the reflected signal increased, as shown in
Figure5(b). Note, theamplitudeof thedirectlytrans-
mittedwave, whichpropagatesfromthesourcetothe
receiver through the tested medium, also increases.
Furthermore, it is very difficult to maintainthecon-
stant contact betweenthetransducer andthevacuum
greaseduringthescanning. Whenwater was usedas
acouplinglayer, it producedarelatively highampli-
tudereflectionasshowninFigure5(c). Inaddition, the
directly transmittedwavecanbeeffectively shielded
by using several layers of aluminumfoil. Therefore,
water wasselectedasthecouplinglayer for scanning
inthisstudy.
Figure 6. Signal processing: (a) Measured signal after
stacking; (b) logsigfunction(windowfunction) for timegain
compensation; (c) Compensated signal; (d) Rectification;
(e) Movingaveraging; (f) Rejection.
3 EXPERIMENTAL SETUPANDSIGNAL
PROCESSING
3.1 Experimental setup
A Pulser (J SR, DPR300) was used to generate the
ultrasonicwavesthroughthetransducer (Panametrics
A3441). Twotransducerswereusedasthesourceand
receiver, respectively. Thereflectedsignalsduetothe
impedance mismatch at the interface were detected
and converted into electrical signal by the receiver
transducer. The electrical signal was fed through an
amplifier (Krohn-Hite 3945: frequency range from
170Hzto26.5MHz) becausetheamplitudeof thesig-
nalsmeasuredbythereceiverwasgenerallytoolowto
identifythemeaningful reflection. Theamplifiedsig-
nal wasdigitizedintheoscilloscope(Agilent54624A
or National InstrumentsPXI-5112).
3.2 Signal processing
Signal processing for ultrasonic imaging consists of
stacking, signal compensation, demodulationinclud-
ingrectificationandsmoothing, rejection, anddisplay
(Zagzebski 1996).
Stacking. Signals were captured after stacking,
whichisthemosteffectivesignal processingtechnique
for removingthehighfrequency, uncorrelatednoise.
The stacking means averaging multiple signals. The
1024signalswereaveragedtoobtainthesignal trace,
asshownFigure6(a).
557
Signal compensation. The amplitude of the
reflected signals at the receiver generally decreases
as the distance of the interfaces, which the waves
are reflected from, increases due to the geometrical
spreadingunder identical impedancemismatch. This
attenuation may becompensated by using timegain
compensation(TGC), whichincreases theamplitude
withtime. Thus, theamplificationfactor is higher at
alonger distancethan at ashorter distancefromthe
receiver.Inthisstudy,theadoptedTGCisalogsigmoid
(logsig) transferfunction, asshowninFigure6(b).The
compensatedsignal isobtainedthroughpoint-by-point
multiplicationbetweentheoriginal signal andlogsig
function. Figure 6(c) shows the compensated signal
fromthe original signal. The amplitude of the first
andsecondreflectionsdecreases.
Demodulation. Demodulation includes rectifica-
tion and smoothing. Rectification is an inversion of
negativecomponents. Thus, thesignal has only pos-
itive values, as shown in Figure 6(d). Smoothing
(moving average) was carried out by using the ker-
nel of [1/20, 2/20, 4/20, 6/20, 4/20, 2/20, 1/20] inthis
study.Thesignal afterthesmoothingprocessisshown
inFigure6(e).
Rejection. Rejection is an elimination process,
whichremovesthesignal whoseamplitudeislessthan
threshold value. Thus, rejection removes noises and
lowamplitudesignals.Inthisstudy,thethresholdvalue
wasabout20%of themaximumamplitude.Thesignal
after rejectionisshowninFigure6(f).
Display. Twomodeswereusedtodisplaytheultra-
sonicreflectionimaginginthisstudy. First, theampli-
tudemode(A-mode), whichrepresentstheamplitude
of thesignal after rejectionversustime, wasused. In
theA-mode,thefirstarrival timeof thereflectedsignal
canbeeasilydetermined. Notetheheight of thetrace
in theA-mode is the amplitude. Second, the ampli-
tudeof eachsignal wasconvertedtobrightness, which
is proportional to theamplitudeof thereflected sig-
nal. Thebrightnessversustimeisthebrightnessmode
(B-mode).
4 APPLICATIONEXAMPLES
Several uniqueapplications of ultrasonic wavemon-
itoring and imaging were explored by using several
small scaleplaster blocks andonelargescaleplaster
block.
4.1 Rotating scanning test for the small scale
plaster blocks
Threekinds of specimens wereused for therotating
scanningtests: anintact plaster block, aplaster block
withaninclinedcrack,andaplasterblockwithacavity.
Thedimensions of theplaster block were300mmin
height, 300mminwidthand150mminthickness.The
middleof thetopsurfaceof theplaster blockwasdug
Figure7. Rotatingscanningtest for anintact plaster block:
(a) Test setup; (b) A-modeimage; (c) B-modeimage.
Figure8. Rotating scanning test for a plaster with crack:
(a) Test setup; (b) A-modeimage; (c) B-modeimage.
toformtheconcaveshapefor therotational scanning,
as shown in Figures 79. Water, which was filled in
theconcavepart of theplaster block, wasusedasthe
couplinglayer.
Intact Plaster Block. Therotational scanningtests
ontheintact block werecarriedout, asshowninFig-
ure8(a).Therotational scanningtestonanintactplas-
ter block clearly showstheinterfaceat thebottomof
theplaster blockasshowninFigures7(b) and(c).The
brightestpartintheB-modeoccurredinthemiddleof
thebottom. Becausethereflectedsignal atthebottom
of theplaster blockwastrappedinthewater coupling
layer, thesignalswiththeintermediateamplitudewere
alsodetectedbelowthebottomof theplaster.
Plaster Block with an Inclined Crack. After the
intact plaster block was cut into two parts in the
inclined direction as shown in Figure 8(a), the two
parts were filled with thin vacuumgrease to maxi-
mize the energy transmitted through the crack. The
result of therotational scanningtest is showninFig-
ures 8(b) and (c). Two strong reflected signals were
detectedfromtheinclinedcrackandfromthebottom
of theblock, respectively. Thefirst strong reflection
occurredat theinclinedcrack. Notetheinclinationof
thecrackwasobservedintheB-mode. Theamplitude
or thebrightness of thereflectedsignals increases as
theangleof incidenceof thetransducerbeambecomes
zerodegreewiththenormal tothecrack (right side).
Notethebrightest part intheB-modecorrespondsto
thestrongest reflectionfromthediscontinuity, which
is perpendicular or normal to thetransducer. There-
fore, the brightest section can be used to determine
the angle of the inclined discontinuity. In addition,
becausethereflectedsignal at theinclinedcrack was
alsotrappedinthewater couplinglayer, themultiple
reflectionswereobserved. Notetheinclinationinthe
558
Figure9. Rotatingscanningtestforaplastercavity: (a)Test
setup; (b) A-modeimage; (c) B-modeimage.
Figure10. Rotatingscanningtest for alargescaleplaster
block: (a) Test setup; (b) B-modeimage.
multiplereflectionswasidentical withthatof thefirst
reflection, which camefromtheinclined crack. The
secondstrongreflectionwas measuredat themiddle
of thebottomof theplaster blockbecausetheangleof
theincident wavebecamezerowiththenormal tothe
bottomof theplaster block.
Plaster Block with a Cavity. A hole (Diameter=
30mm) was drilled on theintact block, as shown in
Figure9(a). Theresultsof therotational scanningtest
areplottedinFigures9(b) and(c). Strongsignalswere
alsoreflectedfromthebottom.Theestimateddiameter
of thecavityfromtheB-modewasabout43mm. Note,
thedifferencebetweenthereal size(30mm) andthe
estimatedsize(43mm) resultsfromthedivergenceof
thebeam, Fresnells ellipse, and thetransducer size.
Although the inclined crack produced a continuous
reflectionlineasshowninFigure8(c), thecavityonly
yieldedthereflectionsinalimitedrange.Furthermore,
theamplitudeof theimagefromthecavitydecreased
whentheangleof theincidentwavesdivertedfromthe
normal tothecavity.
4.2 Rotating scanning test for a large scale plaster
block
Thedimensions of thelargescaleplaster block were
1000mmin height, 1200mmin width and 150mm
inthickness. For therotational scanningtest, theleft
part of the top surface of the plaster block was dug
to form the concave shape, and water was filled,
as shown in Figure 10. The rotational interval was
0.5degree, whichcorrespondedto8.7mmhorizontal
displacement at thebottomof thelargescaleplaster
Figure11. Effect of inclinationangle: (a) Real discontinu-
ity; (b) Ultrasonicreflectionimage.
block. To simulate the multiple discontinuities, two
discontinuities wereprefabricated: onewas horizon-
tal, and the other was inclined. The first horizontal
crack whose gap and length were about 10mmand
250mmwas filled with very weak and low viscous
plaster paste for partial transmission and reflection
of theultrasonic wavesat thisdiscontinuity. Thesec-
ond long and inclined crack was filled with vacuum
grease. B-mode is represented in Figure 10(b), that
showsclear reflectionsfromthefirst horizontal crack
andthesecondinclinedcrack. Inaddition, thebottom
of theplaster blockcouldbeidentified, eventhoughit
wasnot clear.
5 DISCUSSION
5.1 Inclination angle of images
The inclination angle of the reflected image from
the crack was flatter than that froma real crack, as
shown in Figure 8. The relationship of the inclina-
tionanglebetweenthereflectedimage andthereal
discontinuity is
where, p
1
andp
2
arethedepthsof thediscontinuityat
two points separatedby thehorizontal distances (see
Figure11), andnote, p
1
inthereflectedimageis the
travel distance based on the travel time and veloc-
ity of the ultrasonic wave. The inclination angle of
the discontinuity estimated by Equation (4) may be
confirmedby thestrongest brightnessintheB-mode
(seeFigure8).
5.2 Horizontal scanning versus rotating scanning
Afteraparaffinblockwasinstalledunderwater,typical
horizontal scanning(seedetailsinLeeandSantama-
rina2005)androtatingscanningtestswerecarriedout.
The experimental setups are shown in Figure 12(a).
The thickness of the paraffin was 35mm. For the
typical horizontal scanningtest, theultrasonic waves
559
Figure12. Comparisonbetweenrotatingscanningandhor-
izontal scanningtests: (a) Test setup; (b) A-modeimage; (c)
B-modeimage.
reachedperpendicularly to theinterface. That is, the
angleof incidenceof thetransducer beamwasalways
zero degree with the normal to the interface. Thus,
onlyreflectionandtransmissionoccurredat theinter-
face. Theamplitudes of thereflectedandtransmitted
wavesdependontheimpedancemismatch. Thescan-
ninginterval wasset to1mmtoavoidspatial aliasing
(wavelengthinwater is =3mmbecausetheveloc-
ityof theultrasonicwavesinwater isabout 1500m/s
andthefrequencyof thetransducer is500kHz). Note
no migration processing was required because high
directivity transducers were used (Figure 4). Hori-
zontal scanning could clearly detect the horizontal
paraffinwax layer inwater. Furthermore, thebottom
of the water box was also clearly seen, as shown in
Figures12(b) and(c).
For therotatingscanningtest, theultrasonic wave
may reach non-perpendicularly to the interface, and
therefore, amodeconversion may occur (seedetails
inRichart et al. 1970; Aki andRichards1980). Asthe
angleof incidenceof thetransducer beamincreases
fromzerodegreeasshowninFigure12(a), thearrival
timeof thereflectedwaveincreases, theamplitudeof
thereflectedwavesdecreasesandfinally diminishes,
asshowninFigures12(b)and(c),fortherotatingscan-
ningtest. However, B-modeobtainedbythehorizontal
scanningandby rotationscanningisalmost identical
duetothesizeof thetransducer (19mmindiameter),
divergence(seeFigure4), andFresnellsellipse. Note
images obtainedby horizontal scanningandrotating
scanning techniques arerelated to thelateral resolu-
tion. Thus, thereflectedwavesarestill measuredeven
whentheincident wavesdivergesfromthenormal to
theinterface.
6 CONCLUSIONS
Thedesignandapplicationof ultrasonic wavereflec-
tion imaging were documented in this study for the
detection of the discontinuity planes or cavities in
laboratory scalerock models. Thesignal processing,
whichincludesstacking,signal compensation,demod-
ulation, rectification, smoothing, andrejection, were
carried out to producetheamplitudeand brightness
modes (As andB-modes). Themainobservations of
thisstudyfollow:
Althoughvacuumgreasetransmitsmoreenergyto
themediumtested, water isrecommendedasthecou-
pling layer for the horizontal and rotating scanning
techniquesbecauseconstantcontactareaismaintained
in water and the directly transmitted wave between
transducersmaybeeffectivelyremoved.
The discontinuities of the plaster block, includ-
ing horizontal and inclined cracks, and the cavity
wereclearly detectedby usingthenewrotatingscan-
ning technique. While the horizontal and inclined
cracksyieldedcontinuousreflections, thecavity pro-
duced reflections at alimited zone. B-modeis more
appropriatefor thedetectionof discontinuities.
Theparaffinwax underwater was clearly detected
by thetypical horizontal scanningandrotatingscan-
ning techniques. Furthermore, the two techniques
producedalmost identical images. However, cautions
arerequiredfor theanalysisof resultsobtainedbythe
rotatingscanningtechnique.
Theangleof theinclinationobtainedbytherotating
scanningtechniquemaygivetheangleof theoriginal
inclinedcrack throughthecomparisonsof brightness
andthroughasimplecalculationbasedongeometry.
REFERENCES
Aki, K. & Richards, P.G. 1980. Quantitative Seismology
Theory and MethodsVol. 1and2: 932. FreemanCompany,
SanFrancisco.
Gomm, T.J. & Mauseth, J.A. 1999. State of the Technol-
ogy: Ultrasonic Tomography, Materials Evaluation Vol.
57: 737755.
Krautkramer, J. & Krautkramer, H. 1990. Ultrasonic Testing
of Materials: 677. Springer-verlag, London.
Lee, J.S. & Santanmarina, J.C. 2005. P-Wave Reflection
Imaging. Geotechnical Testing Journal Vol. 28: 197206.
Richart, F.E., Hall, J.R. & Woods, R.D. 1970. Vibrations of
Soils and Foundations: 414. Prentice-Hall, USA.
Schmmer, L.W. J r. 1998. Fundamentals of Ultrasonic Nonde-
structive Evaluation A ModelingApproach: 559Plenum
Press.
Wells, P.N.T. 1977. Biomedical Ultrasonics: 635. Academic
Press, London.
Zagzebski, J.A. 1996. Essentials of Ultrasound Physics: 220.
Mosby, Inc., Missouri.
560
Performancereviewof apipejackingproject inHongKong
T.S.K. Lam
Geotechnical Engineering Office, Civil Engineering and Development Department, Government of
the Hong Kong Special Administrative Region
ABSTRACT: The pipe jacking method is commonly used in Hong Kong for construction of underground
cableduct crossingsandstormwater drains. Themethodminimizesthedisturbancetoor interferencewiththe
activitiesandfacilitiesonthegroundsurface. Inthispaper, detailsof apipejackingprojectcompletedrecentlyin
HongKong, involvinguseof apressurizedslurrytunnel boringmachinetoforma222mlong, 1.95mdiameter
cabletunnel, aredescribed. Resultsof theperformancereviewcarriedout oncompletionof theproject arealso
presented.
1 INTRODUCTION
A cable duct crossing was constructed in an urban
areaat KowloonWest of HongKong. Thecableduct
crossing had to traverse a highway and some rail-
waytracks. Conventional opencutexcavationmethod
was not used because of the disturbance that could
becausedtothefacilitiesonthegroundsurface. Pipe
jackingmethodwasusedinstead.Thecableductcross-
ingwasconstructedinfill comprisingloosetomedium
dense,siltycoarsesand,andapressurizedslurrytunnel
boringmachine(TBM) wasselectedfor theproject.
Inthispaper, detailsof theproject aredescribed. A
performancereviewwascarriedout oncompletionof
theproject. Thekey constructionaspects, monitoring
dataobtained at thesiteduring construction, impact
onsensitivefeatures inandsurroundingthesiteand
a summary of the observations and decisions made
during construction are presented. The information
giveninthis paper is obtainedduringauditingof the
projectbytheGeotechnical EngineeringOffice(GEO)
of theCivil EngineeringandDevelopmentDepartment
(CEDD) of theGovernment of theHong Kong Spe-
cial AdministrativeRegion. Theauditingwas carried
out to exercisegeotechnical control intheinterest of
publicsafety.
2.1 Details
The site is located near West Kowloon Highway at
KowloonWest (seeFigure1), Kowloon.
Theproject involves constructionof a222mlong
1.95mdiameter tunnel toserveascableductcrossing
Figure1. Locationplan.
acrossahighway, theMTRTungChungline, theAir-
port Express Line and a public park. Pipe jacking
methodwasusedtoformthe2mdiameter pipeopen-
ingat8to9mdepthinfill, thepropertiesof whichare
shown in Figure8. Thegroundwater level was mea-
suredatabout2.0mbelowground.Twoworkingpitsof
14mlong4mwide10mdeepwereconstructed
atbothendsof thecableductcrossingforthetunneling
operation.
2.2 Geotechnical aspects of the tunnel works
At thedesign stage, plans and supporting documen-
tation of thegeotechnical design of thepipejacking
561
Figure 2. Locations of monitoring prisms on the railway
tracks.
works, a geotechnical risk assessment and a risk
mitigationplanwerepreparedbythedesigner.
Themost sensitivefeatures affected by theworks
were the railway tracks and the underground utili-
ties, includingdrainagepipesandsewers, closetothe
alignment of thepipejacking.
Prior to thecommencement of works, acondition
survey of theexistingroadandthestructures within
25mof thealignmentof thepipejackingandaCCTV
surveyof theexistingdrainagepipesandsewerswithin
20mof theworks werecarried out. Another CCTV
survey was also carried out on completion of the
works.
Aninstrumentationschemeconsistingof 168set-
tlement markers for theroadsurface, 123settlement
monitoring prisms for the railway tracks and four
piezometersforgroundwaterlevel wasadopted. Loca-
tionsof thesettlementmarkers, settlementmonitoring
prismsandpiezometersareshowninFigures1and2.
Thesettlement markers wereinstalled adjacent to
thejackingpit andreceivingpit andontheroadsur-
facealongthealignmentof thepipejacking. Outof the
123 settlement monitoring prisms installed, 32 were
installedontheTungChungLine(up) (TCL up) (on
bothrails), 30ontheAirport ExpressLine(up) (AEL
up), 31ontheTungChungLine(down) (TCL down)
and 30 on the Airport Express Line (down) (AEL
down) (seeFigure2).
Onepiezometereachwasinstalledclosetothejack-
ing pit (VBH1) and receiving pit (VBH2) and two
Figure3. Close-upviewof settlement monitoringprism.
piezometerswereinstalledalongthealignment of the
pipejacking(PZ1andPZ2) (seeFigure1).
A monitoring plan including monitoring of set-
tlement, groundwater level and vibration, and a site
supervision plan for the works were prepared. The
methodstatementfortheworksanddetailsof thepres-
surizedslurrymachineselectedwerealsosubmittedto
therelevantdepartmentsincludingtheGEOforreview.
Thestakeholders affectedby theproject werealso
notifiedandconsultedbythedesigner. For therailway
tracks, trigger levelsof alert, action andalarm
(or AAA) of 12mm, 16mmand20mmrespectively
for settlement and1in1500, 1in1250and1in1000
respectivelyforangulardistortionweresetandagreed
by the MTR Corporation Limited. At alert level,
readings would be reviewed and plans for remedial
measuresandcontingencyactionswouldbeprepared.
At actionlevel, theplannedremedial measureswould
beimplementedandtheworkscouldonlycontinueif
theremedial measurestakenwereeffective. Thealert
andaction levels wouldalso berevisedif necessary.
At alarmlevel, all works would be stopped or con-
tingency actions taken and the design, construction
methodandtheplannedremedial measureswouldbe
reviewed. A maximumtolerable track settlement of
25mmwasallowed.
AnAutomaticDeformationMonitoringSystemwas
usedtotakereadingsfor monitoringtherailwaytrack
settlement. Three CYCLOPS theodolites were posi-
tionedonoppositesidesalongthetracks, andthetrack
settlementwasmonitoredonadailybasisinreal time.
Thedatatakenweretransferredto acomputer inthe
siteofficewhichwerethenprocessedandpostedonto
theinternetforreal timemonitoringbyrelevantparties
(Figures3to5).
2.3 Construction
A RASA DHL1650 pressurized slurry TBM was
used (Figure 6). The diameter of the shield body
562
Figure4. Positionsof settlement monitoringprismsonthe
track.
Figure5. A CYCLOPStheodolite.
and the equipment tube of this TBM is 2,000mm
and1,990mmrespectively. Theexcavateddiameter is
2,040mm. The pipe installed has an outside diame-
ter of 1,950mmandaninternal diameter of 1,650mm
(Figure7).
The construction method involved use of a pres-
surizedslurry system. Theslurry support pressureat
theexcavationfacewasset tobalancethegroundand
groundwater pressureasindicatedinthecontrol panel
of theTBM. TheTBM operator checkedthepressure
gaugetocontrol thepressureattheexcavationfaceand
theslurry-dischargepressuregaugetocontrol thecir-
culationpressure.Thetypical groundwaterpressureat
theexcavationfacewas50to70kPawithanaverage
value of 60kPa at the tunnel axis level. To balance
theground and groundwater pressure, aslurry pres-
sureof 10to20kPaabovethewaterpressure, whichis
60to90kPa, wasappliedattheexcavationfaceforthe
operation(Figure8).Theslurrypressureappliedtothe
Figure6. PressurizedslurryTBM.
Figure7. Dimensionsof TBM andconcretepipe.
Figure8. Typical groundwater pressure.
excavationfacewascontrolledbythepressurecontrol
valveof aby-passunit placedinthejackingpit.
During operation, alubricant consisting of water,
bentoniteandmineral oil was injectedintotheannu-
lusaroundthepipestofill thevoidsaroundthepipes
to reducesoil movement. Oncompletionof thepipe
jacking, thecavitiesaroundthepipesweregrouted. A
cement grout with aminimumcompressivestrength
of 20MPawasused.
563
3 PERFORMANCE REVIEW
3.1 Control of amount of materials excavated
In a pressurized slurry system, the ground was
mechanicallyexcavatedwhiletheexcavatedfacewas
stabilized and supported by slurry. The excavated
materialswereremovedbyslurrytransport. Tocheck
if voidshadbeenformedinthesoil abovethetunnel
duringexcavation, theamount of materialsexcavated
was monitoredduringthetunnelingoperation. A de-
sander wasusedtoseparatethesolidsfromtheslurry.
Thesolidsweremainlyfinesandwhichwaslater used
for backfilling thepits. Theweight of thesoil exca-
vatedfromthejackingpit, receivingpitandthetunnel
was workedout to be2,262tonnes beforetheworks
commenced. Duringtheworks, 593tonnesof sandand
1,461tonnesof finer soil weredisposedof. Thetotal
amount of soil disposed was therefore 2,054 tonnes
which was less than the calculated value. Although
limitedbytheaccuracyof measurement, thisprovided
aroughindicationthat nosignificant voidshadbeen
formed in theground abovethetunnel. No boulders
wereencounteredandnocutter hadtobereplacedfor
thisproject.
3.2 Settlement of railway tracks
In this project, the most sensitive features are the
railwaytracks.
Theexcavationsequenceof thepipejackingworks
in terms of the distance fromthe railway tracks is
shown in Figure 9. Pipe jacking commenced on 29
March 2007. Fourteen days after commencement of
thepipejacking(12April 2007), theTBMreachedthe
TCL Uptrack and5moredayslater (17April 2007),
itwentpastthroughall thetracks. Includedinthefig-
ureisalsothetracksettlement inresponsetothepipe
jacking.
Thetracks settled as soon as thepipewas jacked
past andawayfromthesection. At theTCL Uptrack,
thetracksettledby12mmwhenthepipewasinstalled
at adistanceof 23.5mawayfromthesection. Similar
rateof settlement was notedat tracksAEL Up, AEL
DownandTCL Downexcept inthosecases, thetrack
settlementwasless.Oncompletionof thepipejacking,
maximumsettlementof tracksTCL Up,AEL Up,AEL
DownandAEL Downwere22mm,15mm,10mmand
3mmrespectively whichareless thanthemaximum
tolerabletracksettlement of 25mm.
In pipe jacking, the whole length of the pipe
is jacked and moved forward, resulting in distur-
bance to the soil around the whole length of the
pipe. With the cutter head about 90mmlarger than
the pipe to be jacked (see Figure 7), it is possi-
ble that voids would form around the pipe which
would result in ground movement, and hencesettle-
mentalongthelengthof thepipeevenafterexcavation
Figure9. Excavationsequenceof pipejacking(a) andset-
tlement profilesat tracksTCL Up, AEL Up, AEL Downand
TCL Down(b).
Figure10. Settlement troughat trackTCL Up.
has beencarriedout to somedistanceaway fromthe
tracks.
ThesettlementtroughsattracksTCL Up, AEL Up,
AEL DownandTCL DownareshowninFigures 10
to 13respectively. Thesettlement troughs arealmost
symmetrical inthesecases.
The maximum settlement on the pavement was
15mmwhichwasrecordedbysettlement point SC19
adjacent toTCL Uptrack(Chainage29362).
Giventhegroundconditionsandthemethodof con-
structionasdescribedabove, amaximumvolumeloss
of 3.6%wascalculated(seeFigure14).
Themaximumgroundwaterdrawdownrecordedby
the piezometers at the jacking pit and receiving pit
(VBH2andVBH1) was1.1m, whichwasonlyslightly
abovethepermissiblevalueof 1m. At PZ1location,
564
Figure11. Settlement troughat trackAEL Up.
Figure12. Settlement troughat trackAEL Down.
Figure13. Settlement troughat trackTCL Down.
themaximumdrawdown recorded was 0.7m, and at
PZ2 location, no significant amount of groundwater
movement wasobserved.
Nounduesettlement was recordedat thedrainage
pipes, sewersandthestructuresnearby.
3.3 Actions taken at action level during
construction
Twelve one days after commencement of the pipe
jacking (19April 2007), asettlement of 15mmwas
recordedat theTCL Uptrack whichwasclosetothe
action level. An urgent meeting was held among the
client, designer and contractor on that night. After
Figure 14. Properties of error function to represent the
settlement troughaboveatunnel (after Peck, 1969).
the meeting, it was decided that for the remaining
works, analternativetypeof lubricant, inlieuof amix-
tureof water, bentoniteandmineral oil, wasusedtofill
thegapbetweenthepipeandtheTBMover-break.This
lubricant formedasolidsubstanceonceinjectedinto
thegapandit helpedtoreducesettlement of theover-
burden fill caused by closureof theoverbreak voids
around thepipe. Thedensity of thebentoniteslurry
wasalsoincreased.Theadvancespeedof theTBMwas
reducedfrom9mpernightshifttoabout6mpernight
shift. With the actions taken, the rate of increase of
track settlement wasreducedandthefinal settlement
wascontainedtowithinthepermissiblevalue.
4 CONCLUSIONSANDRECOMMENDATIONS
Thefollowingconclusionsandrecommendationscan
bemadefromtheproject:
1. The use of pipe jacking method was effective in
forminga222mlong, 1.95mdiametercabletunnel
565
at8to9mdepth.ThepressurizedslurryTBMused
is appropriate for the type of ground conditions
encountered.
2. It is necessary to apply appropriate geotechnical
control measures to this typeof project affecting
sensitive features in the interest of public safety.
Withthegeotechnical control measuresapplied, the
geotechnical riskisassessedandthepotential haz-
ards areidentifiedearly. Any unduesettlement or
irregularities observedduringconstructioncanbe
detected promptly and appropriateremedial mea-
sures, suchasuseof adifferentannulusfillertosuit
theactual groundconditions encountered, change
of theadvancespeedof TBM, etc. canbetakento
prevent catastrophicfailurefromhappening.
3. It is important to control the amount of materi-
als excavated fromthetunnel opening to prevent
significant groundloss. Thiscouldbeachievedby
comparingtheexcavatedvolumewiththetheoreti-
cal excavationvolumetocheckif significantvoids
havebeenformedinthesoil abovethetunnel and
checkingtheamount of filler/grout usedtofill the
voids around the pipes. In this project, only the
total amount of materials excavatedfromthetun-
nel openingisobtained. A better methodshouldbe
devisedfor futuresimilar operationmeasuringthe
amount of materials excavatedfor eachsectionof
thetunnel excavationandcheckingtheamount of
filler/grout usedtofill thevoidsaroundthepipes.
ACKNOWLEDGEMENT
The author would like to thank the Director of the
Civil EngineeringandDevelopment Department and
theHeadof theGeotechnical EngineeringOfficefor
thepermissiontopublishthispaper. Commentsonthe
paper fromMr J oeB.N. LeungandDr RichardPang
aregratefullyacknowledged.
Theauthor wouldalso liketo thank theEmployer
of theProject CLP Power HongKongLimited, the
Designer Black & VeatchHongKongLimitedand
theMainContractor KumShing(K.F.) Construction
Co. Limitedfor theco-operationtoexercisegeotech-
nical control in the interest of public safety in the
project and the permission to publish the technical
data/detailsgiveninthepaper. Theviewsgiveninthe
paper reflect only thoseof theauthor and not of the
abovecompanies.
REFERENCES
ground. Proceedings of the 7th International Confer-
ence on Soil Mechanics and Foundation Engineering,
State-of-the-ArtVolume: 225290.
566
Geotechnical control of amajor railwayproject involvingtunnel works
inHongKong
W. Lee, S.S. Chung, K.J. Roberts& P.L.R. Pang
Geotechnical Engineering Office, Civil Engineering and Development Department, Hong Kong Special
Administration Region
ABSTRACT: This paper describes theroleof theGeotechnical Engineering Office, Civil Engineering and
DevelopmentDepartmentof theHongKongSpecial AdministrationRegion(HKSAR)Governmentinexercising
geotechnical control foramajorrailwayprojectinHongKong. Ithighlightsthegeotechnical aspectsof thetunnel
worksandhowthegeotechnical control processprotectspublicsafetyandaddsvaluebyensuringanadequate
standard of design, sitesupervision and risk management is applied. Thesuccessful implementation of best
geotechnical risk management practiceintheproject is strongly influencedby thecommitment of theproject
clienttofollowthecoreelementof theJ ointCodeof Practicefor RiskManagementof TunnellingWorksandto
followupontheresultsof theindependent auditingunder thegeotechnical control process.
1 INTRODUCTION
TheKowloonCantonRailwayCorporations(KCRC)
Kowloon Southern Link (KSL) Project links West
RailsNamCheongStationandEast RailsEastTsim
ShaTsui (ETST) StationintheHKSAR. The3.8km
longrailwaywill haveonenewstation, WestKowloon
(WKN) Station.
The civil construction works of the project
were packaged into three Design and Build con-
tracts. The total project cost, including railway sys-
tems, is expected to be about HK$8.3 billion. The
project is scheduled for completion in late 2009.
Figure1. Aerial photographshowingthealignment of Contract KDB200.
The details of the three contracts are described
below.
1.2 Contract KDB200
ContractKDB200involvesthesectionbetweenETST
StationandJ ordanRoadviatheformer MarinePolice
Headquarters. WKN Station, tworailwaytunnelsand
two emergency vertical shafts, namely Canton Road
EmergencyAccess Point (CREAP) andPekingRoad
Emergency Egress Point (PREEP), were included
in this contract (see Figure 1 below). The contract
was awardedto Link 200J V (formally knownas the
Leighton Balfour Beatty Kumagai J ohnHolland
J ointVenture).
567
Figure2. Slurry-typetunnel boringmachine.
Thegeological sequencealongthealignment gen-
erally comprisesthefollowingprincipal geotechnical
units:
fill (generally a few metres thick but up to 20m
thickat theWest KowloonReclamationsite);
marine deposits and alluvium (described as
clayey/silty sand and sandy silt/clay with some
gravel, mainlyfoundalongSalisburyRoadbutalso
presentlocallynearHaiphongRoadandundredged
pocketsat theWKNStation);
residual soil and saprolite, mainly completely to
highlydecomposedgranite; and
moderately decomposedto freshmedium-grained
granite.
At the preliminary design stage, a cut-and-cover
optionwas plannedfor thetunnels. However, KCRC
decidedto employ aTunnel BoringMachine(TBM)
for the twin railway tunnels between the launching
shaft at the southern tip of WKN Station and the
retrieval shaftatSalisburyRoadtominimizepotential
disruption to road users, pedestrians, business oper-
ations and residents in the area. The twin railway
tunnelsareapproximately8mindiameter and1.1km
inlength. Thedepthtotunnel crownvariesfrom8m
to24m.
The slurry-type TBM (see Figure 2) uses 3800
kilowatts of electricity, which is equivalent to over
5000 horsepower. It was design in Germany and
its components were manufactured in Germany and
China and finally assembled in Guangzhou. The 50
steel cuttingdiscsenabletheTBMtocutthrough1.5m
of rockinabout 40minutes.
Notwithstandingtheuseof TBM, someworkssuch
as ground investigation and grouting were required
under thecontract onthealready very busy roadsur-
face. Such works are to provide additional data for
geotechnical designandtomitigatetheimpactsbefore
constructionof therelevant tunnel workssections.
The excavation for WKN Station was supported
by diaphragmwalls, whilethat for theremainingcut
andcover tunnels was supportedby temporary walls
comprisingsheet piles, pipepilesandstruts.
The most challenging part of these works froma
geotechnical perspectiveistheprovisionof aretrieval
shaft outside the Sheraton Hotel in Salisbury Road
serving as a works portal for the retrieval opera-
tionof theTBM. Trafficdiversionswereunavoidable.
Thispart of theworkswasconstructedusingthecut-
and-cover method. Temporary roaddeckinghasbeen
erected to facilitate the underground works and to
shortentheconstructionperiod.
1.3 Contracts KDB300 and KDB400
Contracts KDB300 and KDB400 involve the sec-
tion between J ordan Road and NamCheong Station
of the KCRCs existing West Rail (WR). The two
contracts are divided at theYau Ma Tei Ventilation
Building(YMTVB) (seeFigure3).Thetunnel lengths
are0.85kmand1.06kmrespectively. Thesecontracts
wereawardedtoChinaStateConstructionEngineering
(HongKong) Limited.
In view of the ground conditions (mainly fill,
marine deposits, alluvium and completely decom-
posed granite), the depth of the tunnel (about 20m
to soffit) and the lack of constraints on the ground
surface, the cut-and-cover tunnel method has been
adopted as the most suitable method in terms of
time and cost for the majority of the tunnel length.
The exception is the tunnel beneath Cherry Street,
whichwouldbeconstructedusingaminedtunnelling
method. Thecut andcover excavationwassupported
bytemporarywallscomprisingsheetpiles, pipepiles,
diaphragmwalls, boredpilewallsandstruts.
Thetunnel alignment is very closeto many exist-
ing buildings, structures and utility services. Some
are very sensitive to construction-induced ground
movements suchas theoperatingMassTransit Rail-
way tracks andbuildings, highway bridgestructures,
Drainage Services Department structures (e.g. box
culverts) and nearby buildings (e.g. theHSBC Cen-
tre). Inorder tomonitor theeffectsof theconstruction
on the surrounding buildings, structures and util-
ity services, extensive geotechnical instrumentation
568
Figure3. Aerial photographshowingthealignment of ContractsKDB300andKDB400.
has beeninstalled, includingground/ structure/utility
settlementcheckpoints, inclinometers, extensometers,
tiltmeters, vibration monitoring checkpoints, stand-
pipesandpiezometers. Instrumentreadingsaremoni-
toredagainstalert, actionandalarmvaluesasdefined
individually for particular instruments and sensitive
receivers.
2 THE ROLE OF THE GEOTECHNICAL
ENGINEERINGOFFICE
2.1 Buildings ordinance and regulations
In the HKSAR, the regulatory control of building
works, in the interest of protecting public safety, is
by application of theBuildings Ordinanceand Reg-
ulations (BOR). Under the BOR the definition of a
building includes, inter alia, any underground
space adapted or constructed for occupation or use
for any purpose including its associated access tun-
nelsandshafts.Therefore, thedevelopmentof private
undergroundspacesuchastunnelsor cavernsinclud-
ing their planning, design and construction all fall
under theBOR.
A KCRC project is considered a private project
in which the tunnel works may be exempted from
theadministrativeproceduresof approval andconsent
under the Buildings Ordinance. Under certain con-
ditions, an Instrument of Exemption (IoE) would be
prepared and issued by theBuildingAuthority (BA)
of the HKSAR Government. For the KSL Project,
because of the significant risk to life and property,
the design, risk management, construction and site
supervisionwouldneedtobeimplementedtoagood
standard to protect public safety. Auditing of the
standardof geotechnical design, sitesupervisionand
risk management is carried out by the Geotechnical
Engineering Office (GEO) of the Civil Engineering
andDevelopment Department, as atechnical adviser
to the Buildings Department (BD) of the HKSAR
Government.
2.2 Technical standards
GEO has issued a technical guidance note TGN24
(GEO 2005a), on specific aspects of site investiga-
tionfor tunnel works intheHKSAR. It supplements
guidance on site investigation given in Geoguide 2
(GEO1987) andGeoguide4(GEO1992).Thetechni-
cal guidancenotewaspreparedwiththebenefitof the
experiencegainedfromtheHarbour AreaTreatment
SchemeStageI (Panget al 2006, Masseyet al 2007).
GEOhasalsoissuedTGN25(GEO2005b), onthe
implementation of geotechnical risk management in
relationtotunnel works. Tunnel worksaredefinedas
tunnels, shafts, caverns and associated underground
facilities, however constructed. Construction of tun-
nel works may involve use of drill and blast meth-
ods, tunnel-boringmachines, cut andcover methods,
techniques that incorporate insitu ground treatment,
groundwater control, installation of temporary and
permanent supports, etc.
TherelevantGovernmentdepartmentsandthepro-
fession were consulted in the preparation of these
guidancenotes.
3 GEOTECHNICAL CONTROL PROCESS
3.1 Instrument of exemption
WithinGEOareviewpanel hasbeensetuptoagreeon
thegeotechnical auditingrequirementstobeincluded
intotheIoE, examinethekeygeotechnical aspectsof
major submissions related to tunnel construction for
the project, and to oversee the standard of auditing.
Theinput startedright fromtheplanningstageof the
projectinwhichKCRCdemonstratedtoGEOthatthey
had adequately identified and assessed the geotech-
nical risks, and had taken suitable risk mitigation
and control actions to managetherisks. In addition,
KCRC committedtoadopt theJ oint Codeof Practice
(ABI & BTS2004). Therisksidentifiedfor theTBM
tunnels relatedto plant procurement, manufactureof
569
segmental linings, delay in launch shaft availability,
break-intoretrieval shaft,TBMassembly/removal and
operationincludingencounteringadversegroundcon-
ditions,unforeseenundergroundobstructions,causing
excessivesettlementsorevendamagetonearbybuild-
ings/structures/utilities during TBM operation and
interventions, compressedair blow-out alongareasof
lowsoil cover and crossing over theoperating MTR
tunnels.
Inearly 2005, KCRCs consultants submittedpre-
liminary scheme designs for auditing by BD/GEO,
including a Geotechnical Basis of Design Report,
Ground Movement Prediction Report, Geotechnical
InstrumentationReportandanExistingBuildingsand
StructuresReport, supportedby aGeotechnical Data
Report.
The BA issued the IoE to the KSL construction
workson30J uly2005pursuant totheKowloonCan-
tonRailwayCorporationOrdinance.Theexemptionis
confined to only thoseprocedures and requirements
relating to approval of plans, consent to commence-
ment and resumption of works and occupation of
buildingsprovidedunder theBOR, suchthat theBAs
duties andsanctioningpowers to ensurestandards of
healthandsafetyarenot undermined.
GuidelineswhichKCRCwererequiredtofollowfor
makingsubmissions toGEOfor geotechnical works,
including tunnel works, under theKSL project were
agreedandincludedintotheIoE andsummarizedina
Management Plan(KCRC 2005).
Under the agreement, KCRC had submitted the
designstatements andvarious methodstatements for
theTBM tunnels, thecut andcover tunnels, CREAP,
PREEP, thelaunchingshaft andtheretrieval shaft to
BD/GEOfor consultation.
KCRC hadalso submittedtheexcavationandlat-
eral support (ELS) plans for theTBM tunnels where
thereis excavation, loading/unloading of theground
or changes to the groundwater regime. The critical
parametersshownintheTBM ELS plansincludethe
operatingslurry pressures, theplannedinterventions
(useof freeairorcompressedairdependsontheantic-
ipatedgroundconditions) andtheair pressureswhere
compressedair interventionsareplanned.
3.2 Other requirements for KCRC under the IoE
Under theIoE, KCRC is required to appoint Autho-
rized Persons (AP), Registered Structural Engineers
(RSE) andRegisteredGeotechnical Engineers(RGE)
to co-ordinatetheworks andto certify theplans and
documents as well as completion of the works; and
to appoint Registered General Building Contractors
(RGBC), andRegisteredSpecialistContractors(RSC)
inthecaseof specializedrailwayconstructionworks,
tosuperviseandcarryout therelevant works.
KCRCisalsorequiredtoinstigateanassurancesys-
temand control scheme to ensure that management
of theconstructionworksisat astandardnot inferior
tothatrequiredunder theBOR. KCRCalsoemployed
ateamof ResidentSiteStaff (RSS), ledbyaConstruc-
tionManager toact astheEngineersRepresentative,
tosupervisetheworksandtheimplementationof the
ProjectRiskManagementPlan(KCRC2006). Inaddi-
tion, AP, RSE, RGE, RGBC and RSC have jointly
preparedSiteSupervisionPlans(SSP) inaccordance
withtheCodes of Practicefor SiteSupervision(BD
2005a, b).
The Contractors were also required to appoint
Independent Checking Engineers (ICE) to provide
certification of consultation documents and verify
geotechnical designsubmissions for bothpermanent
andtemporaryworkspriortoforwardingtoKCRCfor
reviewunder thecontracts.
KCRC ensure that submissions are made at all
stages to BD, and to GEO where geotechnical
aspects are involved, in a timely manner prior to
the commencement of elements of the construction
works through the AP/RSE/RGE. KCRC and the
AP/RSE/RGEarerequiredtoensurethatall comments
given by BD and other relevant parties, in connec-
tionwiththeirsubmittedconsultationdocuments,have
beenresolvedtothesatisfactionof BD andtheparty
concernedprior tocommencement of constructionof
therelevant part of theworks.
3.3 During construction
During the course of works, the AP/RSE/RGE are
required to keep on site copies of certified working
plans, inspection and test records and other rele-
vant reports for regular audit inspections by BD
and GEO. When significant changes in design
or method of working are necessary, then KCRC
throughtheAP/RSE/RGE arerequiredto report this
to BD andGEO andensurethat all comments given
byBDandGEOareresolvedtothesatisfactionof BD.
KCRC through theAP/RSE/RGE are required to
report the following to BD and GEO immediately
whenthefollowingcircumstancesarise:
Construction accidents causing nuisance to the
public.
Irregularities causing inconvenienceto thepublic
and/or damagetonearbyproperty.
Constructionnon-conformities.
Sinceawardof thefirstcontractfortheKSL Project
inAugust 2005, monthly Buildings OrdinanceMan-
agement Committee (BOMC) Meetings have been
heldbyKCRCwithrepresentativesfromKCRCMan-
agement, the AP/RSE/RGE and from Government
includingBDandGEO. Oneof theitemsof discussion
is the scope, standard and timing of the geotechni-
cal submissionsfromtheRGE toGEO. Geotechnical
auditingof theKSL Project iscurrentlyinprogress.
570
ACKNOWLEGEMENTS
Thispaperispublishedwiththeapproval of theHeadof
theGeotechnical EngineeringOfficeandtheDirector
of Civil EngineeringandDevelopment.
Theauthorsacknowledgewithgratitudetheinfor-
mationandmaterialsprovidedbytheKowloon-Canton
Railway CorporationandLink 200J V for production
of thispaper.
REFERENCES
ABI/BTS 2004. A Joint Code of Practice for the Procure-
ment, Design and Construction of Tunnels and Associ-
ated Underground Structures [In the United Kingdom].
London: TheAssociationof BritishInsurers/TheBritish
TunnellingSociety.
BD2005a. Code of Practice for Site Supervision 2005. Hong
Kong: BuildingsDepartment.
BD 2005b. Technical Memorandum for Supervision Plans
2005. HongKong: BuildingsDepartment.
GEO1987. Guide to Site Investigation (Geoguide 2). Hong
Kong: Geotechnical EngineeringOffice.
GEO 1992. Guide to Cavern Engineering (Geoguide 4).
HongKong: Geotechnical EngineeringOffice.
GEO 2005a. Site Investigation for Tunnel Works (TGN24).
HongKong: Geotechnical EngineeringOffice.
GEO 2005b. Geotechnical Risk Management for Tunnel
Works (TGN25). Hong Kong: Geotechnical Engineering
Office.
KCRC 2005. Kowloon Southern Link Management Plan for
Compliance under the Buildings Ordinance Revision D.
HongKong: Kowloon-CantonRailwayCorporation.
KCRC 2006. Kowloon Southern Link Project Risk Man-
agement Plan. Hong Kong: Kowloon-Canton Railway
Corporation.
Massey, J.B., Pang, P.L.R., Lo, J.Y.C. & Salisbury, D. 2007.
Developments in Tunnel Engineering in Hong Kong.
Proceedings of the HKIE Geotechnical Division Annual
Seminar 2007. HongKong: 137155.
Pang, P.L.R., Woodrow, L.K.R. & Massey, J.B. 2006. Devel-
opment of geotechnical control andrisk management for
tunnel works in Hong Kong. Proceedings of the HKIE
Geotechnical Division 26thAnnual Seminar. HongKong:
7588.
571
Researchonstructural statusof operatingtunnel of metroinShanghai and
treatment ideas
J.P. Li, R.L. Wang& J.Y.Yan
Shanghai Metro Operation Co., Ltd., Shanghai, P.R.China
ABSTRACT: Recently, with network operatingandservicingtimegoingon, thesafety of operatingtunnel
in Shanghai is becoming afocus in thecircleof civil engineering. It is well known that safety of operating
tunnel isgreatlyinfluencedbythestructural statusof thetunnel. Thisarticlefirstlygiveanintroductionof main
diseases of operatingtunnel basedonthelargeamount of informationanddatecollectedinthepast decades
about Shanghai Metro. Thencausesfor thediseaseareanalyzedandrelatedsuggestionstoprevent thediseases
deterioratearegiven. It isshownthat themainproblemof operatingtunnel includeleakage, crack, longitudinal
settlement andconstringency.
1 INTRODUCTION
Recently, with network operatingandservicingtime
goingon, thesafety of operatingtunnel in Shanghai
is becoming a focus in the circle of civil engineer-
ing. However, only decades after shield tunnel was
introduced to China, fewsuccessful experiences can
be used for reference from domestic and abroad,
especially for Shanghai metrowheretunnel wascon-
structedinthicksoftclaysoil.Thisarticlefirstlygives
an introduction of main disease of operating tunnel
basedonthelargeamountof informationanddatacol-
lectedinthepastdecadesaboutShanghai Metro.Then
causesforthediseaseareanalyzedandtherelatedsug-
gestions for prevention aregiven. It is expected that
thisarticlecanbeuseful for thosewhoisinterestedin
thehealthof tunnel of Shanghai metro.
2 MAINDISEASES
2.1 Leakage
Based on the large amount of information collected
recently about Shanghai tunnel of metro, more than
tenseepagespotscanbefoundateachsectionbetween
adjacent stations. Seepage spots in several sections
coulduptofiftyplaces. Figure1indicatesthatleakage
mainlyoccursat bothsidesof tunnel. jointsof station
andtunnel andtheby-passbetweenupanddownline.
Minority leakageplaces werefoundedat theholeof
bolt andtheholeof grouting. Accordingtotechnical
specification for water-proof of shield-driven tunnel
(DBJ 08-50-96, 1996), Leakageof theoperatingtunnel
of Shanghai metrobelongs tothesecondor thethird
level.
After decades of operating, thefunctionof water-
proof of sealingrodhasalreadyreducedtoalowgrade
for the non-uniformsettlement of tunnel caused by
adjacent construction, pumpingof ground-water, etc.
For thespecial stressstate, mostsegmentringsexhibit
an ellipseshapewith horizontal radius enlarged and
vertical radius reduced. Correspondingly, the com-
pressive stress at the hance of tunnel segment ring
decrease near outer surface and increase near inte-
rior surface. Gapsat thejoint of segment ringsmaybe
formed. Sincethesealingrodwasset at theouter side
of segment ring, sothecapability of water-proof was
reducedatthehanceof thetunnel, andthisleadtothat
most leakagetakeplaceat thesideof tunnel. Inaddi-
tion, for thegreat differenceof structural styleamong
theshieldtunnel, thestationandtheby-pass, differen-
tial settlement occurs at thesepositions andthis also
leadstoseriousseepage.
2.2 Segment crack
Fewcracks areformed in thesegment itself, but the
phenomenonthatjointfiller haspull-outcanbefound
for non-uniformsettlement of operatingtunnel which
makesthejointsuncoupled. Mostcrackslocatedatthe
corner of segmentor unfilledcorner takeplaceduring
the construction phase, such as production, mainte-
nance, handlingandconsolidation. Additional, cracks
can be found between segment and track bed at the
positionwherelargedifferential settlement takeplace
andwhereturningradiusissmall.
573
Figure1. Leakageof thesegment ring.
2.3 Longitudinal settlement
Largesettlementtakeplacetothetunnel after decades
of operation, and the longitudinal settlement repre-
sents with regional characteristic. Figure2 indicates
that the longitudinal settlement curve of Line 1 of
Shanghai metrocontainstwohugesettlepits. Oneof
themis about 1400mlong, located at theHengshan
road station and the maximum settlement is about
20cm. The other one is located at the interval from
theSouthHuangpi roadstationto Shanghai Railway
stationwiththemaximumsettlement isnearly30cm.
2.4 Rate of settlement
Figure 3 is duration curves of settlement of tunnel
nearthePeopleSquare.Althoughthetunnel havebeen
constructedformorethantenyearsandtherateof set-
tlementhasbeingsloweddown, itcannotleadstothat
the tunnel has already been stabilized for the creep
properties of Shanghai soft clay. Howto control the
settlement is still a difficult problemfaced by civil
engineering.
2.5 Convergence
Horizontal diameterof mostsegmentringenlargefrom
2cm to 4cm, few of them even up to 7cm, This
already exceed the design safety limit (15D).
According to statistics, segment ring with horizon-
tal convergencedeformationgreaterthan3cmoccupy
69.98%andthoseof greater than6cmoccupy6.63%.
Themaximumconvergenceisabout15cmandthegap
alongthelongitudinal jointis11mmwhichmeansthat
circumferential bolt havealreadyreachyieldlimit.
3 GENETICANALYSISOF DISEASE
Accordingtodecadessubwaymonitoringresults, the
main factors related to the deformation of metro
structurearelistedasfollowing:
(1) Local GroundSubsidence
Basedonthegroundsubsidencedatabaseof Shang-
hai city, itshowsthatthesettlementsof subwaystation
and tunnels are large if they located at the center
574
Figure2. Longitudinal Settlement Curveof Line1of Shanghai Metro(1995.52007.12).
Figure3. Durationcurveof tunnel adjacent toPeopleSquare.
of subsidenceregion. Comparisonof time-settlement
curves between ground surface and tunnel structure
measurement in People Square region gives great
agreement. Figure4showsthetime-settlementcurves
from1999 to 2008. But the exacted ratio of tunnel
settlement to ground subsidence is not able to be
obtained.
(2) GeologyConditions
InShanghai area, thesoft soil stratumis about 30m
deepfromgroundsurface.Thesubwaystationandtun-
nel mainly located in soft soil. The soil is basically
saturatedflowing-plasticor soft-plasticclaywithlow
shear strength(0.0050.01MPa), highwater content
(above 40%), high compressibility (0.51.0MPa-
1), sensitivity varying from4 to 5, and rheological
behavior. In this very soft ground, the influence of
excavation and tunnel drive to environment could
not be ignored either for construction period or for
long-termoperatingperiod.
(3) Qualityof Construction
It wouldleadtolargedeformationif thereweresome
accidents occurrring during the construction of sta-
tion and tunnel driving. The differential settlement
woulddevelopalongtheoperationlifeandoverstepthe
operatingsafety standardeventually. Therearesome
typical cases presentingthis phenomena, suchas the
tunnel deformationnear toWest Ninghai Rd. pump-
ingstationof metroLine.1, leakageof water andsand
inpumpingstationof metroLine.2crossingrivepart,
ShilongRd. stationof metroLine3, havebeenfound
andtreatedagainst totheexcessivesettlement.
(4) MaintenanceandOperationWork
Vibrationsof runningtrainswouldleadtotunnel set-
tlement, andthenwater leakagewill takes placedue
to great tunnel settlement. In casethewater leakage
becomes serious, it will inducemorewater/soil loss
andbehaveas larger settlement. Thevicious circleis
formed in this way. The in-time and frequent main-
tenance work is a good way to prevent this vicious
circle.Theresearchonlong-termsoil mechanicsunder
vibrationof highconsistercyandlowfrequencyisstill
undergoing. Butitcouldbeensuredthatdisadvantage
of thecomplicatedsoil behavior isthemainfactor of
tunnel operationsafe.
(5) LoadingandUnloadingactivitiesnear totunnel
Duetotheshortageof landingresourceinurbanenvi-
ronmentlikeShanghai city, theprojectsof foundation
pitlocatedintheareaof subwayprotectionareatendto
575
Figure4. Time-settlement curveof groundsurfaceandtunnel.
bedeeper, bigger, closer, moredifficultandmorerisk.
Asaconsequence, itisnecessarytopaymoreattention
totheinfluenceof deepfoundationpitexcavationand
high-risebuildingtothesubwaystationandtunnel.
(6) Groundwater exploitation, dewatering and pore
pressurereduction
Soil consolidation induced by the groundwater
exploitation would also influence the subway struc-
tures.
4 CONTROL MEASURES
The safety guarantee system for subway structure
should take many aspects into account, including
theregulation, standard, monitoringmanagement and
pre-alarmmanagement.
(1) RegulationandStandardfor SubwayControl and
Protection
Supervisedregulationshouldbeestablishedfor sub-
waylinesurveys, inspectionandprojectsconstruction.
Process management and responsibility definition
should be enhanced to make sure that the whole
subwaylinesareunder control.
(2) Standardof theMonitoringProcedure
Itistheresponsibilityof monitoringengineer tomon-
itor, inspectandanalyzesubwaystructure, whichaims
to findtheproblemof structureintimeandto guar-
antee the safety of subway. There are at least twice
settlement measurements and oneconvergencemea-
surement for operating line; 34times settlement
measurementsandonceconvergencemeasurementfor
new line; several times inspection for key positions.
Moreattentionshouldbepaidto themonitoringand
inspectionof projectslocatedinthesubwayprotection
area. Upto now, 30000settlement monitoringpoints
and1convergencemonitoringpoint every 5segment
ringswereset for metroLine1toLine4whichareall
operatingmetrolinesinShanghai.
(3) DiseaseRecordCardSystemfor Subway Struc-
ture
Basedondecades observationof longitudinal settle-
ment curves of subway tunnel, positions with large
longitudinal settlement curvaturearefound. Disease
record card of subway structure were established to
makesurethat inspectionandmonitoringcouldfocus
on thesepositions. Then quick responseand actions
couldbetakenaccordingto theresults of inspection
andmonitoringincaseof anyabnormal situation.
(4) Digital InformationSystem
Anoverall digital scanfor runningsubway andabig
GISsystemareplanningandconstructingrespectively.
TheGIS systemwould control thesafety of subway
structure and operation risk using information tech-
nology including inspects of geological conditions,
tunnel structure, waterproof system, settlement and
convergence.
(5) NewtechnologyandEquipment
The monitoring system with long distance, large
range,highprecisionandautomatizationcharacteristic
shouldbedevelopedcontinuously.Remotemonitoring
systemandequipment configurationwouldbeset on
thekeyposition. Thedata, graphsandinformationare
abletobetransferredandanalyzedinstantly.
(6) Project locatedintheSubwayProtectionArea
Seting protection area aims to reduce the influence
of loading and unloading effect frombuilding and
excavation to tunnel by enhancing control standard
576
to the projects located in the subway protection
area.
(7) ScientificResearchProjectlocatedintheSubway
ProtectionArea
The investigation to structural problem of metro,
including durability of tunnel, seepage and leakage,
settlementandconvergence,hasbeenpreformedcoop-
eratingwithresearchinstitute. Evaluationsystemand
indexwill beestablishedtoprotect subwaystructure.
REFERENCES
Cui Z.D., TangY.Q., LuC., et al. 2007. Predictionof ground
settlement induced due to changes in engineering envi-
ronment in Shanghai. Journal of Engineering Geology
15(2):233236.
DuanJ.L.,TanY.L., WangY., etal. 2006. Casestudyof foun-
dation pits crossing subway tunnel. Chinese Journal of
Geotechnical Engineering 28(Z):18771879.
Li X.Z., HuangM.S. &WangL.M. 2007. Experimental study
oncreepandcycliccreepcharacteristicsof saturatedsoft
clay. Journal of Chongqing Jianzhu University 29(2):56
59.
YeY.D., ZhuH.H. &WangR.L. 2007.Analysisonthecurrent
status of metro operating tunnel damagein soft ground
and its causes. Chinese Journal of Underground Space
and Engineering 3(1):157161.
Zhu Z.F., Tao X.M. & Xie H.S. 2006. The influence and
control of deep excavation on deformation of operating
metrotunnel. Chinese Journal of Underground Space and
Engineering 2(1):128131.
577
Maximisingthepotential of straingauges: A Singaporeperspective
N.H. Osborne, C.C. Ng& D.C. Chen
Deputy Project Manager, Land Transport Authority, Singapore
G.H. Tan
Managing Director, SysEng Pte Ltd, Singapore
J. Rudi
Instrumentation Engineer, Kiso Jiban Consultants Co. Ltd, Singapore Branch
K.M. Latt
Instrumentation Engineer, Soil & Foundation Pte Ltd, Singapore
ABSTRACT: Loadmonitoringof support strutsintemporary excavationsplaysacrucial roleinconfirming
the stability and safety of the excavation. Data gained provides valuable feedback for the design engineer
to facilitaterefinement of futuredesigns. Much of this monitoring is undertaken by strain gauges, sensitive
instruments attachedto thetemporary supports, whicharethenlinkedto automatedalarms throughreal time
systems. The success of the monitoring is directly linked to the performance and interpretation of the data
derivedfromthestraingauges, andthereliabilityof thereal timesystem. Thereareanumber of factorswhich
caninterferewiththeperformanceof themonitoringsystem, tosuccessfullyandusefullyinterpret thedata; the
influenceof thesefactorsneedstobeunderstood. Theemphasismust beontheproductionof highqualitydata
that can bereliably processedandrapidly given to theend-user, such that erroneous readings areminimized
andgenuineloadchangesareidentifiedfor interpretation. Throughanumber of casestudiesof deepexcavation
projectsinSingapore, influencesonthemonitoringsystemarereviewed, their potential impactsdiscussed, and
recommendationsgiventoproduceahighqualityandreliablemonitoringsystem, thusmaximizingthepotential
of straingaugestobeusedfor monitoringof performance.
1 INTRODUCTION
1.1 Importance of instrumentation
Instrumentation and monitoring has always played a
crucial rolewithintheconstructionindustryandrecent
worldwidehighprofileconstructionfailureshavefur-
ther raiseditsprofileandimportance. Thisringsvery
truefor Singapore. Technological advancementshave
allowedtheinstrumentationindustrytobecomemore
sophisticatedinhowdataismonitored, collectedand
presented to the end-users. However the emphasis
remains on theproduction of quality datafor useful
interpretation. Strain gauges are one very important
component of a fully integrated and comprehensive
monitoring system used to control the movements
andloadsgeneratedduringexcavations. InSingapore
theyareusedextensivelyfor strut monitoringindeep
excavationsandprovideinvaluabledataensuringthat
constructioncontrol ismaintainedduringexcavation.
Thebuild-upof loadinthestrut is monitoredinreal
timeduringconstructionandcomparedwiththedesign
predictionsatthevariousexcavationstages. If signifi-
cantdiscrepanciesareobserved, reanalysisisrequired,
and areview of thedesign assumptions undertaken.
Howeverstraingaugesarenotoriouslysensitiveinstru-
ments and their readings can be influenced by a
number of factors, leadingtomisinterpretationwhen
thedataisreviewed.
Thereareanumberof wide-rangingpotential influ-
ences impactingtheaccuracy of straingaugeresults,
encompassinganumber of relatedconstructionareas:
Theseincludeinstallation, environmental effects, con-
structionactivitiesandtheir positionrelativetostruc-
tural members, all of which can result in erratic
changesinthestraingaugereadings,someof whichare
genuineloadchanges,somenot.Withthedevelopment
of real time systems, the results can be automati-
cally transferred to the internet for interpretation or
sent to amobilephone, but without any filtering of
theerroneous readings. If receivedby inexperienced
579
Figure1. Comparisonof straingaugeandloadcell results.
personnel, commonly, all readingsaretreatedasgen-
uine, with the potential to cause unnecessary false
panic andevacuationof theexcavation, or erroneous
andignoringgenuinechangesinload. Either scenario
results inaloss of confidenceinthemonitoringsys-
tem. Itisimportanttherefore, thattheseverysensitive
instruments and their performance be fully under-
stood during interpretation and every effort taken to
maximizethequalityof thedata.
1.2 Reliability of strain gauge a comparison with
load cell
Generally, loadcell isknowntoprovidereliableload
measurementasitmeasuresthefull loadacrossthefull
section of thestrut. Strain gauge, on theother hand,
measurelessthen1%of thestrut crosssectional area
andthenattributethatloadtothewholestrut. Figure1
showsthedatathat givesanindicationof therelative
accuraciesof straingaugesandloadcells. Thegraph
showsageneral variancebetween5%and10%. Much
of thisdifferencecanbeaccountedfor bytheir differ-
ent responses to thevarious construction influences,
external factorssuchastemperatureandEMI, andthe
different ways they monitor load. A detailed discus-
sionof thistopicisbeyondthescopeof thispaper and
needs to bepart of aseparatestudy. Thecorrelation
betweenthetwoinstruments, however, issufficient to
indicatethatstraingauge, iscapableof monitoringthe
strutsaccurately.
With this fact established, strain gauge has other
advantages over load cell. One fundamental benefit
isthecost effectivenessof straingaugeinSingapore;
withastraingaugecostingapproximately U.S. $200
andaloadcell considerablymoreatU.S. $4000. With
cost,normallytotheindustrysdetriment,significantly
influencing the selection process, strain gauges are
the instrument of choice. Cost aside froma techni-
cal perspective, introduction of load cell in thestrut
can createnon typical loading conditions, but strain
gaugedoes not changetheloading condition on the
strut. In addition, strain gauge can also be replaced
easilyif damaged.
1.3 Choice of strain gauge type
Once a strain gauge is chosen to monitor a strut,
therearetwocommonlyavailabletypesfor strutmon-
itoring; spot weldable strain gauges (SWSG) or arc
weldable(AWSG), also referredto as surfacemount
strain gauges. Both strain gauges arewidely used in
Singapore, withslightlydifferentattributesandadvan-
tages. SWSGhavetheadvantageof beingattachedlow
to thestrut, thereforeminimizing theimpact caused
bybendingerror. However theAWSGismorerobust,
its greater areafixedto thestrut resultinginabetter
areatoareaconnection, makingitintheorylesssensi-
tivetofluctuationfromvibrationcausedbyaccidental
impacttothestrut. Byfar themostimportantfactor in
ensuringtheperformanceof thestraingauges is that
they are installed correctly and not damaged during
installation.
For SWSG the gauge flanges are spot welded to
the strut, this requires exposing the thin 1.45mm
gauge, increasingthepossibilityof damage. After the
welding, thevibrating wireplucking coil housing is
mountedontothegaugeformeasurementandasaform
of protectivecover. It iscrucial that thecoil doesnot
touchthegauge, whichcanoccur withsomebrands,
otherwisesubsequentreadingsareaffected.AWSGare
arc welded to the flange, however a dummy gauge
shouldalways beusedduringweldingto avoiddam-
age, thenreplacedwiththeactual gauge, followedby
mountingthevibratingwirepluckingcoil andfurther
protectionasrequired.Theonlypotential problemwith
this installation is theintroduction of residual stress
intothestructural member bytheweldingprocess. On
overall comparisonof therespectivemeritsof thetwo
gaugetypes theauthors believeAWSG will provide
more reliable results a view shared by Broone &
Crawford(2000).
2 FACTORSINFLUENCINGTHE
PERFORMANCE OF STRAINGAUGE
2.1 Location of strain gauge
When bending within the prop is likely to be sig-
nificant, strain gaugereading will beaffected by its
locationintheprop. Straingaugesplacedat thesame
section may give different reading as they experi-
encedifferent stressesinthepropsubject tobending.
Connections between the prop and waling result in
non-uniformstresses, asdoconnectionstokingposts,
cross-bracing or runner beams. Thus strain gauges
installed near to these locations are likely to be
affectedbythenon-uniformstressesandwill not give
arepresentationof theloadsintheprops.
2.2 Impact of electromagnetic interference
As vibrating wire strain gauges operate at a fre-
quency between 600 to 1500 Hz, they are subject
580
Figure2. Impact of EMI noiseonstraingaugereadings.
to electromagnetic interference(EMI). This compro-
misestheaccuracyof thereadingsbyintroducingnoise
intotherawdata, whichcanbeverydifficult tosepa-
ratefromgenuinedata, andthereforecanbeprocessed
andcalculatedas load. Therearenumerous potential
sourcesof EMI noiseonconstructionsitesincluding:
arcwelding, machineryignition, powergeneratorsand
power cables onthesite. Thenoisetakes oneof two
forms, either asageneral underlyingtrendimpacting
theoverall accuracyof databyincreasingitsspread.Or
asahighvoltagesurge,causingaspikeintheloadread-
ingswhen, for example, amachineignitionisstarted,
electronicnoisetendstolowerreadings, whereasmag-
neticnoiseincreasesthem.Withtheadventof real time
monitoringanddataprocessingat10minuteintervals,
theimpact of thisinterferencebecomesmoresignifi-
cant. Thefirst caseleadstoageneral questioningthe
accuracy of thestraingaugereadingsastheaccuracy
rangeappears wider. Thedataspikes canresult inan
alarmbeingbreached, withthepotential for work to
stopunnecessarily.
Theimpactof EMI noisecanbeclearlyseeninFig-
ure2. Duringtheworkingdayon9th& 10thSeptem-
ber, electronicnoiseinterferencefromagenerator and
power cablecauseda200kN fluctuatingreductionin
load.Thelunchhour canalsobeclearlyseenwhenthe
generator was turnedoff. By the11thSeptember the
noisehadbeenidentifiedandthegenerator removed,
hencemorestablereadings.
2.3 Temperature effect
Theimpactof temperatureontheperformanceof strain
gaugesinstalledonstrutshaslongbeenrecognizedas
asignificant factor affectingtheir performance, with
papers first publishedontheissueintheearly 1960s
inNorwayandJ apan(NGI 1962&Endo&Kawasaki
1963). Despitetheadventof thermallymatchedstrain
gauges, thermal influencepersists. IntheUnitedKing-
domforanun-deckedexcavationsubjecttothethermal
effectsduetosunlight, anannual temperaturerangeof
50
C wasmeasuredleadingtoavariationof 2750kN
instrutloads, after thebaseof excavationwasreached
(BattenMetal. 1996).Thisequatestoachangeof load
of 65kN per 1
C. Similarly intheUnitedStates sig-

nificantvariationsweremeasuredandreported.Boone
andCrawford2000, recordedan18.75kNchangeper
1
C, withanannual temperaturevariationof 45
Con
their siteresultinginanannual fluctuationof 844kN.
Thedifferencein thesevariations is dueto thestrut
area, incombinationwiththestiffnessof theretaining
system, theground it supports and theend restraint,
with agreater stiffness resulting in agreater impact
duetotemperaturevariation.
Singaporelying1.5
North of theequator experi-

ences minimal seasonal variationintemperature, but
asignificant diurnal range, withtemperaturesfluctu-
atingfromalowof 20
Ctoahighof 36
C.Thisposes
adifferent set of problems. Theproblemof tempera-
turevariationwas first publishedby Niuet al (2005)
discussing fluctuations due to a deep excavation in
Singapore for the North East Line metro, in 1999.
A loadchangeof 37kNper 1
Cwasrecorded, against
a theoretical change of 48kN per 1
C. This being
calculatedby:
Where LP is change in load due to temperature
change(LT), A
s
is cross sectional area of thestrut,
E
s
isYoungs modulus of steel and is the thermal
coefficient of expansionfor steel.
For thisexcavation, 23%of thetheoretical increase
inloadwas not observedintheinstrumentation. The
absence of this monitored load was explained and
demonstrated by an outward movement of 2mmof
theretaining systemduring thehigher temperatures.
ThisphenomenonhasbeenobservedintheU.S. where
a potential 13mm movement into the ground was
recordedinglacial till, (D. Druss 2000) andonother
projectsinSingapore, witha2mmmovementintothe
groundof a1.5mthickdiaphragmwall insoftMarine
Clay.
Thesametemperaturephenomenonisseenconsis-
tentlyacrossdeepexcavationsinSingapore.Fora25m
deepexcavationinsoftMarineClayontheCircleLine
project an increaseof 30kN per 1
C was measured,
Figure3, across threedifferent struts, over four days
of nonexcavation.Thisequatestoonly56%of thethe-
oretical increasebeingtransferredtothestrut load. In
thiscasethecritical factor inmobilizingthefull effect
of thetemperatureliesintheground.Theretainingsys-
temwasverystiff, with1.5mthick diaphragmwalls,
comparedto theNELP examplewhereasoldier pile
systeminconsiderably stiffer groundwas used. Fur-
ther to thegeneral trend in thefigure, thereis some
scatter in the data, this is due to readings taken at
between 8:00and9:00amandattributedto thevery
localizedeffect of plant start-upupcausingEMI.
581
Figure3. Impact of temperatureonstrut load.
Figure4. Lossof preloadafter strut installation.
2.4 Preloading
Preloading has an important bearing on the strain
gaugeperformanceinearly stages. Thefundamental
actionistoinstall thestraingauge, takebasereadings
and check it is working correctly prior to the onset
of preloading. Basereadings shouldideally betaken
prior tothestrut beingconnectedtoitsendsupports.
Duringpreloadingtheloadregisteredinthestrain
gaugeandloadcell shouldbecheckedagainstthejack
load; however loadscannot beexpectedtomatchper-
fectly. Thejackregistersthehighest load, followedby
thestraingaugeandthentheloadcell; thissequence
tendstobethecaseobservedinSingapore.
Thiscanbeattributedtoanumberof factors: therel-
ativepositionsof theinstrument, relatively lowloads
being used; construction difficulties in placing the
strut trulyperpendicular totheretainingsystem; tem-
perature effects; the introduction of load cell in the
strutsaffectingtheoverall loadtransfer pattern.
2.5 Disturbance by construction activities
As with all excavations, the accuracy of the instru-
mentationreadingscanbeaccidentallyinfluencedby
anumber of different constructionactivities, leading
topotential misinterpretationof thereadings. Oneof
theobvious is theaccidental damageof straingauge
Figure5. Impact of welding on adjacent strain gauges on
thesamestrut.
by worker, whichcaneither permanently damagethe
gaugeor partiallydamagethegaugeor connection.
2.6 Welding
One of the construction effects resulting in erratic
and significantly high loads registered by the strain
gaugeiswelding. Highheatfromweldingof horizon-
tal ties or lacings to thestrut member, whichusually
commencesafter preloading, canresult inahighand
sudden increasein thestrut loads. As shown in Fig-
ure5, theweldingof lacingto astrut has causedthe
strut loadmeasuredbyapair of straingaugeslocated
closetothelacing, torisesuddenly.
Theimpactof thisweldingonthestrutloadisclearly
evident.
As shown in Figure5, thereadings on both sides
of thestrut webshowasuddendrop, probablyassoci-
atedwithEMI noise, followedby asharpincreaseof
loadontheweldingside. Anincreasefrom1300kN
to 1900kN was recorded. Onthenon-weldingsidea
minimal rise in load was registered. On completion
of weldingtheimpactedgaugedidnot recover to its
original loadbut remainedat itselevatedlevel, which
is not representative of the overall load in the strut.
Thisresidual stress, recognizedsince1964, isnotrep-
resentativeof theactual loadof thestrut, andif clearly
identifiedfromthereadingsandconstructionactivity,
thereadingcanbeadjustedtoaccount for thiseffect.
2.7 Casting of permanent slab
Another construction impact on temporary supports,
and their strain gauges, is the effect due to casting
of permanent components of atop-downexcavation.
Duringthecastingof a1.5mthickroof slab,theimpact
of thecuringandexpansionof theslabcanclearlybe
seen in the two layers of struts above the roof slab,
Figure6. A significant dropof 500kN was observed
acrossthefull excavation, followedbyanincreasesev-
eral dayslater andareturntotheongoingtrendof the
582
Figure6. Impact of castingof permanent roof slabonstrut
loadsabove.
load. Onceclearlyidentified, thisphenomenoncanbe
easily linkedtoconstructionactivity, andnot usedto
cast doubt ontheaccuracyof thestraingaugeresults.
2.8 Negative values
Another phenomenon seen in strain gauge in deep
excavation, particularly in soft clay, is that of nega-
tive loads in the top strut layers, indicating that the
strutsareintension. Thisisfrequentlyblamedonthe
instrumentsthemselvesandregardedaserrorreadings.
However investigations into anumber of thesecases
haveidentifiedthat thestraingauges arefunctioning
well andrecordinggenuineloads. Independentchecks
throughcut off testsandinsertingjackshavedemon-
stratedthat thesestrutsareintension. Thisisduetoa
combinationof factors:theloadsinthestrutstendtobe
originallylowathigher levels; lossof preload; andthe
deflectionprofileof theretainingwall.Withsoftclays,
significant retaining wall movements are recorded,
withdeepseatedmovementsoccurringbelowexcava-
tionlevels. Astheexcavationprogresses, stiffer struts
withgreater preloads areused. Combinedwiththese
largemovements belowthestruts, theretainingwall
can rotateabout thestrut, resulting in asmall back-
ward movement into thesoil at higher level. This is
alsoreflectedintheinclinometer readings.
2.9 Problem related to real time system
Tomakethemosteffectiveuseof straingaugedata, for
deepexcavations, it isprudent tolinktheinstruments
via data logger to the office computer and mobile
phonesinareal timesystem. However byimplement-
ingsuchasystemtwonoteworthyproblemsneedtobe
considered.
Firstthepotential highnumber of alertalarmsgen-
erated. The erroneous alerts can lead to a loss of
confidence in the systemand potentially a genuine
alert beinglost amongst thefalsealarms.
Thesecond and potentially moreserious problem
lieswiththerobustnessof thereal timesystemitself.
The simile, a chain is only as strong as its weakest
link, ringsverytruewhenappliedtoanyreal timesys-
tem. Anyfailureof anycomponent withinthesystem
compromises the whole monitoring scheme leading
to an absenceof results to theend user. Apart from
the strain gauges, the potential numbers of points
that can fail within thesystemarenumerous. These
includethecabling, thedatalogger itself, thephone
system, the power and the server. Failures of all of
thesecomponentshavebeenexperienced.
3 HOWTOMAXIMIZETHE POTENTIAL OF
STRAINGAUGE
3.1 Right location of strain gauge on strut
There are two important considerations when locat-
ingastraingaugeonastrut. First itslocationonthe
strut, andthenits positionrelativeto other structural
members. If bending within the prop is likely to be
significant, straingaugesshouldbelocatedtoaccount
for it, four gauges for acircular prop or proportion-
atelyspacedalongthewebforanI-beam. Connections
between the prop and waling result in non-uniform
stresses, asdoconnectionstokingposts, cross-bracing
orrunnerbeams.Thereforethegaugelocationsshould
beat maximumdistancefromtheseareas, tobefully
representativeof theloadspassingthroughthestrut.
3.2 Cross referencing with load cell
British CIRIA C517s guide on temporary design
(Twine & Roscoe 1999) recommends strain gauges
over loadcells. However as cross referencingof data
fromdifferent instrument typesiscritical togaininga
full pictureof theexcavationinducedmovement, it is
recommendedthatatleastsomeloadcellsbeincluded
in the instrumentation design. This concept is fully
recognizedinSingapore.
The Building Control Authority of Singapore
(BCA) statesthatdeepexcavationsmustbemonitored
by a combination of strain gauges and loads cells.
Land Transport Authority of Singapore (LTA), the
clientforthemajorityof deepexcavationsinSingapore
which are for the continually growing underground
rail network, specifies that as a minimumrequire-
ment, 25%of all struts for deepexcavations shall be
monitoredforloadbystraingaugeand/orloadcell,and
of thosemonitoredbystraingauge, 15%shall alsobe
monitoredbyloadcell.
3.3 Minimizing EMI effect
EMI noisecanbeavoidedby afewsimplemeasures
onsite. Itisrecommendedthatcablinglengthsbekept
to aminimum; thecables and datalogger belocated
at least 5maway fromany potential sourceof EMI.
Figure7 shows atypical layout of datalogger in an
excavationsite. Regular checksshouldbeundertaken
583
Figure7. Typical layoutof dataloggeronanexcavationsite.
byaportableEMI meter toensurethat cableanddat-
alogger zones remainEMI free. Althoughcables can
be protected fromelectronic noise, cable joints and
theinevitableonsitecuttingandre-splicingof cables
areweakpointsthatcanbeeasilycorrupted, therefore
particular attentionshouldbepaidtotheselocations.
3.4 Account for temperature effect
ThetemperaturerangeislessinSingapore, andthere-
fore the impact of load less than elsewhere in the
world. However strut loads are very closely linked
withmonitoringcontrol of excavations.Themaximum
designloadof astrut, usingmoderately conservative
soil parameters, isusedasthework suspensionlevel,
withanalert set at 70%of thiscapacity. Therecanbe
significant implications onthework, includingstop-
pageof worksif thetemperatureeffectisnotproperly
accountedfor inthestrutloads. A number of different
solutionsincludepaintingstrutswhiteanddailyspray-
ingtoreducetemperatureimpact. It issuggestedthat
themostappropriatesolutionistoaccountfor thethe-
oretical temperatureeffects duringdesignandtoadd
themto themonitoring control values to ensurethat
workisnot impactedunnecessarily.
3.5 Protection against disturbance by construction
activities
Damageto strain gaugeand disturbanceon thedata
duetoconstructionactivitiescanbeavoidedthrougha
combinationof adequateprotection, educationof site
staff through tools box talks and clear well marked
signs on site. If damage occurs on a frequent basis,
the more extreme measure of fining individuals or
companiescanbeconsidered.
3.6 Maximizes strain gauge reading with real
time system
To maximizestrain gaugereading for monitoring of
deepexcavations, it isprudent tolinktheinstruments
via data logger to the office computer and mobile
phones in a seamless fully automated machine to
machine(M2M) systemwherenohumanintervention
isrequiredfor onwardtransmissionof results, result-
inginafully real timesystem(Tanet al 2004). It is
stronglyrecommendedthatthecapacityfordatatrans-
ferof anysuchsystemisinminutesandthatawireless
systembeutilized.
An understanding of the potential problems can
reducefalsealerts, combined with alerts going only
to knowledgeablepersonnel who arefully cognizant
of theconstructionworksbeingundertaken.
Potential failureof thecabling, thedatalogger, the
phonesystem, thepowerandtheserverwithinthesys-
temcanleadtolossof dataor delayintransmissionof
data. Toensurethat thesystemisfullyautomatedand
seamless, all theseareasneedtoberigorouslychecked
and fail safes written into thesystems to informthe
systemmanagerif anyof thesecomponentsfail, rather
thanassumingthat all arefunctioningsmoothly.
4 CONCLUSIONS
It isclearly evident that straingaugesareessential in
monitoringandcontrollingdeepexcavations.Withthe
increasingsophisticationof thereal timesystemspro-
ducing vast quantities of data, combined with M2M
capabilitiesallowingautomatedalerts, andstrictalarm
limits on the monitored loads within the temporary
structural systems, theresults fromstraingauges are
under very closescrutiny. Thereforequality dataand
aclear understandingof bothstraingauges andhow
theconstructionactivitiesimpact that dataarecrucial
to theinterpretation of strain gaugeresults. Without
this, confidence in the performance of the system
is lost, resulting in results being ignored as errors,
very dangerous in the monitoring environment, or
numerousunnecessaryalarmsreceivedimpactingthe
constructionprogress.
To maximize the potential of strain gauges, their
locations and that of the cabling must be planned,
installationcarriedout by skilledpersonnel awareof
the problems and the potential to compromise data.
Datainterpretationshouldalsobebyskilledpersonnel
fully awareof thedesign predictions for theexcava-
tion, theexcavationprogressandthepotential impact
of theexcavationonthestraingaugeresults.Finallythe
processingsystemtakingdatafromthestraingaugeto
the end user must be seamless and robust such that
this component does not fail leading to a complete
breakdownof thewholesystem.
REFERENCES
Batten, M., Powrie, W., Boorman, R. &Yu, H. 1996. Mea-
surementof PropLoadsinalargeexcavationduringcon-
structionof theJ LEstationatCanadaWater, EastLondon.
Proceedings of Geotechnical Aspects of Underground
Construction in Soft Ground, London 1996 pp 712
584
Boone, S.J. & Crawford, A.M. 2000. Theeffectsof tempera-
tureandUseof VibratingWireStrainGaugesfor Braced
Excavations.Geotechnical News September 2000.
Druss, D.L. 2000. Discussion The effects of temperature
and Use of Vibrating Wire Strain Gauges for Braced
Excavations, Geotechnical News Vol. 18 No. 4 pp 24.
Endo, M. & Kawasaki, T. 1963. Study of Thermal Stresses
ActingonStruts. Transactions of the Architectural Insti-
tute of Japan No. 63 pp 689692
NGI. 1962.VibratingWiremeasuringDevicesUsedatStrut-
ted Excavations, Technical Report No. 9. Norwegian
Geotechnical Institute, Oslo.
Niu, J.X., Wong, I.H. & Makino, M. 2005. Temperature
Effects on Strut Loads and Ground Movements for a
31mdeepexcavationinSingapore. Proceedings of Under-
ground Singapore 2005.
Tan, G.H., Ng, T.G. & Brownjohn, J. 2004. Real Time
monitoringandAlertSystemsforCivil engineeringAppli-
cationsusingMachinetomachineTechnologies. Int. Con-
ference on Structural and Foundation Failures, Singapore
2004.
Twine, D. & Roscoe, H. 1999. TemporaryProppingof Deep
Excavations, CIRIA Guide C517.
585
Discussionondesignmethodfor retainingstructuresof metro
stationdeepexcavationsinShanghai
R. Wang, G.B. Liu& D.P. Liu
Department of Geotechnical Engineering of Tongji University, Shanghai, P.R. China
Z.Z. Ma
Shanghai Shentong Holding Co., Ltd, Shanghai, P.R. China
ABSTRACT: By investigatingmany projects inShanghai metro construction, notabledifferences arefound
to exist in present design schemes, especially in steel bar content of diaphragmwall and struts load. In this
paper, somedesignschemes of retainingstructures for metro stationdeepexcavations whichhavethesimilar
geological and structural situations are compared. Their rationality is evaluated and tested according to the
resultsof fieldmonitoring(e.g. strutload, bendingmomentof diaphragmwall deducedfromlateral deformation
curves). Byadoptingthesemeasurements, aseriesof software, calculatingmodels, methodsandparameterscan
besummarized. Somekeyfactorswhichaffectthecorrectnessof thedesignareemphasized. Finally, suggestions
that satisfy demandsof safety andeconomy aregiven. Thesesuggestionshelpimprovethedesignof retaining
structuresfor deepexcavationsothat thedesignstandardinShanghai canbeunifiedgradually.
In order to prepare the 2010 EXPO, construction
activities of civil infrastructures andbuildings inthe
municipalityof Shanghai havebeenincreasedsignif-
icantly. To meet thedemands of high efficiency city
traffic, extensiveundergroundmetrosysteminShang-
hai hasbeenrapidly constructed. Howtoaccomplish
thetargetof completing400kmnetworkinbothquick
andsatisfactorywayunder thesituationof largescale
andhighrisk becomes aquestionto theconstructors
inShanghai. Manyproblemsduringconstructionhave
been researched by many researchers and engineers
over theyears(Liu& Hou1997; Liuet al. 1999; Liu
etal. 2000, 2001). Butfewreportsaboutthedesignof
retainingstructuresof excavationsareanalyzed. Some
retaining structure design schemes of normal metro
station excavations which have the similar geologi-
cal andstructural situationsareselected.Accordingto
theresultsof fieldmonitoring(e.g. strutload, bending
momentindiaphragmwall deducedfromlateral defor-
mationcurves), therationalityof designsareanalyzed.
A seriesof software, calculatingmodels, methodsand
parameterscanbesummarized, andkeyfactorsaffect-
ingthecorrectnessof thedesignareemphasized. The
intention of this paper is optimizing design methods
satisfy demands of safety and economy. These sug-
gestions help improve design of retaining structures
for deep excavations so that the design standard in
Shanghai canbeunifiedgradually.
2 DESIGNOF RETAININGSTRUCTURES
OF DEEP EXCAVATION
2.1 Background of the design
2.1.1 Shape of the excavation
Comparedwiththediversificationof commonexca-
vations, almost all metrostationdeepexcavationsare
similar. The main excavation of metro station looks
likeadumbbell ontheplane, whichdividedintothree
partsstandardsegment inthemiddleandendwells
at twoedgesof theexcavation.
2.1.2 Geological condition
Under thegroundof Shanghai, thereissoft soil about
dozens meter deep, whose water content, degree of
sensitivity, compressibilityandrheologyarehigh, and
theunit weight, strengthandpermeabilityarelow.
587
Table1. Thecontrol criterionof environment protectioninexcavation.
Protection Therequirement of maximumgroundsettlement and
grade maximumretainingwall horizontal displacement Therequirement of surroundingprotection
1 Maximumsettlement of ground0.1%H*; Intheareanot morethan0.7Hawayfromexcavation,
Maximumhorizontal displacement of retaining therearemetro, municipal pipes, gaspipes,
wall 0.14%H; important water pipesandsoon. Thoseimportant
Ks** 1.8 buildingandutilitiesmust besecured.
2 Maximumsettlement of ground0.2%H; Inthearea12Hawayfromexcavation, thereare
Maximumhorizontal displacement of retaining important mainlines, water pipes, important in-used
wall 0.3%H; structuresandbuildings.
Ks1.6
3 Maximumsettlement of ground0.5%H; Intheareanot morethan2Hawayfromexcavation,
Maximumhorizontal displacement of retaining therearelessimportant branchlinesandgeneral
wall 0.7%H; buildingsandinstallations.
Ks1.4
* H isthedepthof excavation; ** Ks isthesafetycoefficient against basal heave.
2.2 Design method of the retaining
structure in Shanghai
2.2.1 Excavation protective grades
Accordingtotheengineeringexperienceof Shanghai
metro deep excavation engineering and the require-
ment of the surrounding environmental protection
(Shanghai standard foundation design code DGJ 08-
11-1999), thedeformationcontrol criterionof excava-
tionisclassifiedintothreeprotectivegrades(Table1).
2.2.2 Model
Finiteelementmethod(FEM)of two-dimensional ver-
tical elasticsubgradebeamwasadoptedfor analyzing
thedeflection of diaphragmwall. In theFEM mod-
eling, lateral earth pressure was taken as triangular
and rectangular. The distribution of horizontal load
is slope gradient on the excavation plane and rect-
angular under theexcavation plane. Thedistribution
of horizontal spring stiffness coefficient of the pas-
sivezonesoil wasalsoconsideredastwodistributions.
They havetriangular distributionwithslopegradient
withineffectivedepthunder theexcavationlevel and
rectangular distributionbelowthedepth(Figure1).
3 DIVERSITY OF THE DESIGNRESULTS
Consideringthesimilarityof excavationformandgeo-
logical condition of metro stations in Shanghai, the
retainingstructuresdesignof metrostationdeepexca-
vations should be unified. But it is to find notable
differencesexistinginthepresent designschemesby
investigating many projects in Shanghai metro con-
struction, especiallyinsteel bar content of diaphragm
wall andstrutsparameters. Resultsof retainingsystem
design of fifteen deep excavations of metro stations
in Shanghai arelisted inTable2. Regard two layers
Figure1. Vertical elasticsubgradebeammodel.
underground stations as the objects of analysis, the
differencesof themareasfollows:
3.1 Steel bar content of diaphragm wall
Theexcavationdepthisattherangeof about1517m
instandardsegmentof twolayersundergroundexcava-
tion,buttheunitweightof longitudinal barsplushoops
of diaphragmwall isquitedistinction, fromthemini-
mum127.05kg/m
3
tothemaximum232.63kg/m
3
, the
valuediscrepancyisabout 83%.
Station 1 and 4 are similar in excavation depth,
but about 22% discrepancy in steel bar content of
diaphragmwall.
Thestrutsnumberof crosssectionof station8isone
morethanthat of station2, but still 44%higher steel
bar content of diaphragmwall thanthat of station2.
588
Table2. Designresultsof bracingstructural systemof somestationsof metrolineM1.
Standardsegment of theexcavation
Serial Lengthof the Thicknessof Embedded Steel bar content Vertical
number of Structure Excavation diaphragm thediaphragm ratioof the of thediaphragm struts
station form depth/m wall/m wall/m diaphragmwall wall/kg m
3
number
1 Towlayers 17.28 34.0 0.8 0.97 127.05 5
2 underground 16.80 30.0 0.8 0.79 128.86 4
3 16.60 32.0 0.8 0.93 145.82 5
4 17.29 31.0 0.8 0.79 155.66 5
5 16.30 30.0 0.8 0.84 155.85 5
6 16.30 28.5 0.8 0.75 156.43 5
7 15.66 30.0 0.8 0.92 169.47 4
8 16.44 30.0 0.8 0.82 185.58 5
9 15.57 28.5 0.6 0.83 186.44 4
10 15.59 28.0 0.7 0.80 192.56 4
11 14.56 28.6 0.6 0.96 196.97 4
12 14.92 28.3 0.6 0.90 232.63 4
13 Threelayers 18.37 32.0 0.8 0.74 139.80 5
14 underground 24.24 43.0 1.0 0.77 145.21 7
15 21.56 37.0 1.0 0.72 153.10 6
*Note: Steel bar content of thediaphragmwall intable1referstheunit weight of longitudinal barsplushoops.
Figure2. Variationof steel bar content of diaphragmwall
withexcavationdepth.
The thickness of diaphragmwall of station 10 is
10cmthicker than that of station 9, but almost the
sameinsteel bar content.
All above shows that the regularity of steel bar
content of diaphragmwall isnot obviousnow.
In addition, the decreasing tendency of steel bar
content of diaphragmwall alongwiththeexcavation
depthincreasingisshowninFigure2.
Thesteel bar content of diaphragmwall equals to
AgdividedbyV, whichAgisthetotal weightof longi-
tudinal barsplushoopsof adiaphragmwall; V isthe
volumeof adiaphragmwall. Accordingtothedesign
schemes, thediameter andspacingof thelongitudinal
barsandhoopsaresimilar. Butthespacingof bothbars
increasesattherangeof approachingwall toe;itcauses
theincrement of weight of steel barslessthanthat of
thevolumewhentheexcavationdepthandlengthof
thediaphragmwall increase. Sothesteel bar content
of diaphragmwall decreased.
3.2 Design parameters of steel struts
3.2.1 Comparison of design parameters of different
stations
Strut designparameters of 3stations whichhavethe
closely excavationdepthandstratumdistributionare
compared(Table3).
Spacingof thestrutsoftenascertainedaccordingto
experience, sotherearelessdifferential init. But the
designstrutsloadcalculatingbymodel isdiscrepancy
obviously.
3.2.2 Comparison of struts load by designing and
by field monitoring
Fieldmonitoringis animportant means of providing
immediate feedback to designers during excavation
andof documentingtheactual performanceasacase
historyfor futurereferences.
Testingby themonitoringdataof station9and13
inTable2after excavation, real strutsloadarefar less
thanthedesigning.
Station9: Thepercentages occupiedby real struts
loadtodesignvaluesare67%(level 2); 58%(level 3)
and67%(level 4).
Station13: Thepercentagesoccupiedbyreal struts
loadtodesignvaluesare72%(level 1); 69%(level 2);
46%(level 3); 37%(level 4) and36%(level 5).
Considerablebearingcapacityiswastedespecially
in the lower struts. So the struts design should be
optimized.
589
Table3. Comparisonof strutsparametersof threestationexcavations(Standardsegment).
Station I II III
Excavationdepth/m 15.57 14.92 15.59
Spacingonplane/m about 3.0 about 3.0 about 3.0
Vertical spacing/m Level 1groundlevel 0.5 0.4 1.0
Level 2Level 1 3.9 4.34 4.24
Level 3Level 2 3.9 3.7 4.1
Level 4Level 3 3.1 3.42 3.4
Final levelLevel 4 3.07 2.6 2.85
Strutsloadof design/kN Level 1 / 800 800
Level 2 1500 1660 1850
Level 3 2000 2300 2000
Level 4 1300 1975 1350
Figure 3. Struts load comparison of designing with field
monitoring.
4 FACTORSANALYSIS
Byinvestigatingeight designinstituteswhichtakeon
mostdesignworkof metroprojectsinShanghai, there
aremanyfactorscausingthedisperseof designresults.
Themainreasonsareasfollows:
4.1 Diversification of calculating tools
Thesoftwarecalculatingtheinternal forceandlateral
deflectionof diaphragmwall isnot unified. SUPER-
SAP, ANSYS and MIDAS are used in calculating
themainbody of theexcavation; FRWS, LIZHENG,
YAO1, YAO2etc. areusedincheckingtheresults or
designingtheaffiliatingstructures.
4.2 Values of the parameters
Effectedbythedesignerssubjectiveunderstandingof
the code, parameters like overload, lateral pressure,
restrictionsandsoonareadopteddifferently. Thedif-
ference of these initial parameters will lead to the
different calculatingresultsdirectly.
In addition, thephenomenaof thedesigners con-
centratingonsafewhileignoringtheeconomy made
thedesignschemesarequiteconservative.
Figure4. Usagepercentof softwareincalculatingthemain
bodyexcavations.
4.3 Relationship between design and construction
Construction is the means to realize the design
schemes. It can test the rationality of the design.
Therefore, mutual cooperationcanhelpbothof them
gettingbetter. But littlecontact betweenthemmakes
the design theoretical; say nothing of optimizing
the design according to the feedback of the field
construction.
5 DESIGNSUGGESTION
According to the design code and practical experi-
enceof Shanghai, somekey points about thedesign
of theretainingstructureof normal metrostationdeep
excavationaresuggestedor emphasized:
5.1 Stricter design demands
Theloadcoefficienthasalreadyincreasedfrom1.25to
1.35;Thestructuresignificancecoefficient1.1should
beconsidered;Theconcreteprotectionlayer isthicker
thanbefore.
5.2 Value of kv
The restriction under the toe of the diaphragmwall
is simulatedby vertical springwhich kv often adopt
10000kN/m
3
inShanghai.
590
Figure5. Empirical valueof k inShanghai.
5.3 Calculation of lateral loads
Resultsof lateral loadsdonthavemuchdifferenceno
mattertheearthandwaterpressurecalculatedtogether
or not inShanghai. But it hadbetter tocalculatethem
separately.
5.4 Earth pressure
Earthpressureisakeyparameterof excavationdesign.
Inthepractices of soft clays excavationengineering
inShanghai, bymeansof alotof fieldmeasurements,
indoorstestsandcentrifugal modetests, therelationof
earthpressurewithdifferent factorssuchasretaining
walls displacement, time, and construction parame-
tersiserected.A practical formulaof calculatingearth
pressurehasalreadybeenputforward.Theactiveearth
pressurecoefficient k canbeascertainedby Figure5
after excavatingtothefinal level.
5.5 Equivalent subgrade coefficient K
h
K
h
is amaincalculationparameter inthecalculation
methodof displacement andinternal forceof retain-
ingwall usedbydesigninstitutes. Anequivalent level
resistancecoefficient K
h
isusedtoreflect thecapac-
ity that soil resists deformation. It is a coefficient
that involves viscidity, elasticity, plasticity and other
diversifiedconstructionfactors.
For purpose of facilitating design and engineer-
ing application, according to practices in Shanghai,
usuallythefollowingsimplifiedformulacanbeadopt:
whenthesubgradewithout improvement:
whenthesubgradewithimprovement:
Figure6. Maximumstrutsloadbyfieldmonitoring.
Where
i
, c
i
,
i
isspecificgravity, cohesion, angleof
internal frictionof layer i respectively; h
i
andh
i1
are
bottomdepthof layeri andlayeri 1respectively; B
j
,
T
j
, H
j
isexcavationwidth, excavationtime, excavation
depthof progressj respectively.
5.6 Struts load
By survey of 26 station excavations, the maximum
monitoring loads of each level struts are suggested.
According to Figure6, most values of struts load of
excavation in Shanghai are 5001000kN (level 1),
591
Figure7. Deflectionratioof thewall indifferent casesand
wall thickness.
Figure 8. Comparison of deflection, bending moment in
different wall thickness(excavatetofinal level).
12001800kN (level 2), 12002200kN (level 3),
10002500kN(level 4and5) respectively. Themax-
imumvalues are about 1300kN (level 1), 2800kN
(level 2), 3000kN (level 3), 3000kN (level 4and5)
bydeletingsomespecial factors.
AsshowninFigure7, althoughthedeflectionof the
diaphragmwall gradually increasing along with the
excavation deeper, the increment of the deflection
become lower with the wall getting thicker. But the
methodof confiningthedeflectionbysimplyincreas-
ing the wall thickness is neither economical nor
effective. It isnot onlyplayingtheinvalidfunctionto
limit thedeflectionbut causingthebendingmoment
of the wall increasing obviously when the thickness
increase to specific value. (Figure 8) So the thinner
wall shouldbebetter whenthedemandof safety can
besatisfied. Thethickness of thediaphragmwall in
Shanghai often adopted 600mm, 800mm, 1000mm
and1200mmwhichincrementis200mm.Asthecon-
structionmachinedeveloped, thickness increment by
100mmas modulus can be realized. Nowadays the
attempt to apply the 700mm diaphragm walls has
succeededinmetrolinem1inShanghai.
Figure9. Video, datacollection and net pageplatformof
theremotemonitoringsystem.
5.7 Remote monitoring, management and
consulting system by use of internet technology
Remotemonitoringandmanagement systemisintel-
ligentized monitor, measurement and management
systembased on network technology and it is built
onlocal networkandInternet. Suchasystemcandeal
withthedisperseprojects, maketheinformation, man-
powerandotherresourceof relativecompaniesshared
eachother,makeitconvenienttofollowtheprojectand
knowwhat isgoingon.
People taking on monitoring of the construction
field inputs the monitoring data into the system
592
immediately after monitoring. Theproject managers
anddesignerscanconsult thedatathroughaspecific
net pageplatform; if thereis any risk existinginthe
project, someconstructionmethodsor designadjust-
ment can bedoneto prevent theaccident happen. It
will bring utmost profit to companies by providing
intelligent andeffectiveway of project management.
At thesametime, it can bealso convenient to build
andmanagedatabaseandprovideremotetechnology
consultation.
6 CONCLUSION
Based on thesurvey of largequantity of monitoring
data of theretaining structureof metro station deep
excavations in Shanghai, the following conclusions
canbedrawn:
Differencesinthedesignschemesof theretaining
structureof normal metro station deep excavation is
depictedandanalyzed. It mainly focuses onthesteel
bar content andsteel strut load, whichdisaccordance
withthecharacterof similarityandregularityof metro
stationexcavation.
A lotof factorsaffectthedesignresults. Difference
of calculating software, initial inputting parameters,
designer objectiveunderstandingandso onbecomes
themainreasons.
Suggestions are given according to the code and
practical experiencetohelpunifythedesign.
REFERENCES
Liu, G.B., Huang, Y.X. & Hou, X.Y. 2001. Study onEquiv-
alent Lateral Resistance Factor of Soils in Excavation,
China Civil Engineering Journal.
Liu, J.H., Liu, G.B. & Fan, Y.Q. 1999. Theory of Timeand
Space Effect and its Practice in Soft Soil, Journal of
Underground Engineering and Tunneling.
Liu, J.H. &Hou, X.Y. 1997. Manual of Excavationengineer-
ing, Chinese Construction Industry Press, Beijing.
Series of New Technologies of Environment Protection in
the Construction of Subway; from Shanghai Subway
Corporation (General).
Shanghai Construction Commission. 1999. Shanghai stan-
dardfoundationdesigncode(DGJ 08-11-1999), Shanghai
Construction Commission (SCC), Shanghai, China. [In
Chinese].
Sun, G.S. & Zheng, D.T. 1987. Soft foundationsandunder-
groundengineering,ChinaArchitecture &Building Press,
Beijing, China.
593
Riskanalysisfor cutterheadfailureof compositeEPB shieldbasedon
fuzzyfault tree
Y.R.Yan& H.W. Huang
Q.F. Hu
Shanghai Institute of Disaster Prevention and Relief, Shanghai, P.R. China
ABSTRACT: All factors andrelatedhazards inthecutterheadof compositeEPB shieldweresystematically
analyzed, andthenfuzzyFaultTreeAnalysis(FTA) model of cutterheadfailurewasproposed. Mainbasicevents
affectingtheoccurrenceprobability of thetopevent wereverifiedby aquantitativeanalysis, whichcouldbe
appliedintheriskanalysisof EPBshieldmachinescutterhead. Comparedwiththetraditional faulttreeanalysis,
thefuzzy fault treemethodcanget thecutterheadfailurepossibility distributionof compositeEPB shield. At
last, themeasuresthat helpreducingthecutterheadfailureoccurrencewerepresented.
1.1 The actuality of composite EPB shield
cutterhead failure
Withthemanufacturetechniquedevelopmentof shield
tunnelling, thescopeof stratainwhichshieldsworks
becomes more and more widely. Shield technology
is not only used in relatively uniformity or single
ground, but also used in mixed ground that changes
fromhardrocktomixedfaceandsoftground(andvice
versa) at thetunnel level. Duringtheexcavation, the
highly abrasive and frequently changing mixed face
ground causes high cutter wear, especially flat cut-
ter wear; theaccident rateof cutter disc to thetotal
shieldis highly to onehalf. Themainforms of acci-
dents arecutter abrasion, cutter disc abrasion, cutter
discdistortionandsoon(Lei Guo2006). Theseacci-
dentsmadetheadvancerateandcutterheadservicelife
largelydepressed. Accordingtostatistics, indomestic
compositeEPB (EarthPressureBalance) shieldcon-
struction, thefailureof cutter abrasioncameforthat
differentlydegree(WeibinZhuet al. 2006).
1.2 The introduction of fuzzy FTA
Inthetraditional faulttree, afailureeventsystem(top
event)isdividedintomanysub-eventswithacombina-
tionof seriesandparallel. Itsfailureprobabilitycanbe
back-calculatedaccordingtothelogical relationshipof
thefaulttreewhenthefailureprobabilityof eachbasic
eventisknown. FaulttreeanalysisisbasedonBoolean
algebra. A quantitativeanalysisneedsprobabilitiesof
all basiceventsortheminimal cutset,whicharemostly
obtained by statistical data or subjectivejudgmental
databased on experts experiences. Thesedatahave
uncertaintybecauseof variousinfluencefactorsduring
statistical proceduresandlimitationof experts expe-
riences. It is necessary todefineafuzzy valueinthe
probabilisticspacetorepresent asingleprobability.
Basicconceptsandmethodsof thefuzzyfault tree
were proposed in 1980s (Tanaka et al.1983, Furuta
1984). At present, thestudy of thefuzzy fault treeis
almost focused on algorithmand theintegrated the-
orysystemshavenot beenestablishedandverifiedin
practice (Singer 1999). A fuzzy set is introduced in
this paper andthefailureprobability of basic events
isreplacedbythefuzzyfailureprobability. A triangu-
lar fuzzynumber isintroducedtorepresentthefailure
probabilityof abasiceventandthefuzzyfailureprob-
ability of a top event is obtained by fuzzy number
operation. Measures to reducethecutterhead failure
possibilityof theEPBshieldareverifiedbyanalyzing
theimportanceof basicevents.
2 FUZZY SET THEORY ANDITSOPERATIONS
2.1 Fuzzy set
Fuzzy set theory was introducedby Zadeh(1965) to
deal with the problemin which the phenomena are
595
impreciseandvague. LetX beacollectionof objects,
calledtheuniverse, whoseelementsaredenotedbyx.
AfuzzysubsetAinX ischaracterizedbyamembership
function f
A
(X)which associates with each element x
inX areal number intheinterval [0, 1]. Thefunction
valuef
A
(X) represents thegradeof membershipof x
inA.Thelarger thef
A
(X) is, thestronger thedegreeof
belongingnessfor x inA.
2.2 Fuzzy numbers and its operations
Switchbetweenthefuzzynumbersareusedtohandle
impreciseinformationsuchascloseto5, highreli-
ability, lowfailurerate, etc. Therearemany forms
of fuzzy numbers to represent the linguistic values.
Inhere, triangular fuzzy numbers areapplied. Let x,
a, m, b R (real line). A triangular fuzzynumber isa
fuzzy number A in R, if its membership function f
A
:
R[0, 1] is
In this study, triangular fuzzy numbers are
employed. A triangular fuzzy number canbedefined
byatriplet =(a, m, b). Themembershipfunctionis
with a mb. Thetriangular fuzzy number can be
denoted by a triplet =(a, m, b). The parameter m
givesthemaximal gradeof f
A
(X), i.e. f
A
(m)=1, it is
themost probablevalueof theevaluationdata. a and
marethelower andupper boundsof theavailablearea
for theevaluationdata.
For a given in thetheinterval [0, 1], thearith-
meticoperationsof fuzzynumberscanbedefinedby
meansof -cut operationsaccordingtotheextension
principle(Zadeh1965):
3 FTA OF CUTTERHEADFAILURE
3.1 Fault tree model for the risk of composite EPB
shield cutterhead failure
Themainpurposeof thefault treeanalysis is tofind
out all failuremodesof thesystemandtheevent with
arather largefailureprobability. After weak sections
havebeenenhanced, occurrenceprobabilitiesof these
accidentsarereducedsothat thesystemreliability is
improved. For researchingthecutterheadfailurerisk
of compositeEPBshield, thereweresomesupposition
intheanalysisasfollows:
1. Itdidnttakeaccountof thedisadvantageousinflu-
encethat theengineering construction brought to
the surroundings building, road surface, under-
groundpipelineetc.
2. Dragbitsandscraperswouldoftenproducenormal
wear during theshield advance, so their abrasion
wasnot considered.
Accordingtotheaboveinvestigationaccidentsdata,
thecutterheadfailureriskof compositeEPB shieldin
mixedfacegroundwassetbymeansof thefuzzyFTA
method, showninFigure1.
Thesystemic risk probability of cutterheadisana-
lyzedbasedontheestablishedfaulttreemodel. Firstly,
theminimumbasic event sets which causethemain
event occurrenceis solved, i.e. theminimumcut set
(MCSfor short) of fault tree. EachMCScorresponds
tooneaccidenttype, andthereareseveral MCSswith
different occurrenceprobabilityof onefault tree. The
MCSwiththemaximumoccurrenceprobabilityisthe
most probablypotential factor whichmaycauseacci-
dent. Booleanalgorithmisrelativelysimpleinsolving
theMCS. Thefault treeinFigure1is obtainedfrom
theBooleanalgorithmasit showninEq. (6).
Fromtheresultobtainedabove,thetopeventTisthe
unionof 28setswhicharetheMCSof thefaulttree,i.e.
{X
1
X
7
}, {X
2
X
7
}, {X
3
X
7
}, {X
3
X
11
}, {X
3
X
16
}, {X
3
X
18
},
{X
3
X
24
}, {X
3
X
25
}, {X
4
X
11
}, {X
4
X
12
}, {X
4
X
13
},
{X
4
X
14
},{X
4
X
18
},{X
4
X
24
},{X
4
X
26
},{X
22
X
23
},{X
5
},
{X
6
}, {X
8
}, {X
9
}, {X
10
}, {X
15
}, {X
17
}, {X
19
}, {X
20
},
{X
21
}, {X
27
}, {X
28
}, whichcorrespondto28accident
modes.
3.2 Quantitative analysis
Quantitative analysis includes evaluation of failure
probabilityof thetopevent andimportant analysisof
thebasicevents. Inpractice, theoccurrenceprobabil-
ity of atopevent (PT) isobtainedusingapproximate
probabilityformulaof independenteventsasshownin
Eq.(7)
Where P(M
i
) is the occurrence probability of the
ith MCS. For example, P(M
i
) is the occurrence
596
Figure1. Fault treeof thecutterheadfailureriskof compositeEPB shield.
597
probability of thefirst MCS {X
1
X
7
}, whichdepends
on the probability multiplication of basic events X
1
andX
7
.
Probabilities of the basic events must be known
in advance, in order to evaluate failure probability
of the top event and important analysis of the basic
events. Expert elicitation and fuzzy set theory will
be used to get the probabilities of the basic events
in this paper. Because the experts cannot exactly
evaluate the probability of events, and sometimes
some of the events are vague, the experts tend to
apply natural linguistic expression, such as Impos-
sible, Infrequent, Occasional, PossibleandFrequent,
todescribetheprobabilityof events. Accordingtothe
Guidelinesof RiskManagementforMetroTunnelling
andUndergroundEngineeringWorks (2007), itranks
theoccurrenceprobabilityof riskinto5class,shownin
Table1. Conventional mathematical wayscannothan-
dlenatural linguisticexpressionefficientlybecauseof
its vagueness. Therefore, fuzzy set theory is used to
copewithit.Accordingtothetriangularfuzzynumber
discussedabove, it isassumedherein
Theoccurrenceprobabilitiesof randombasicevents
usingEq.(8) andEq.(9), andtheir fuzzy probabilities
areshowninTable2.
Figure2showsthatthefuzzyprobabilityof thetop
eventcanbeexpressedastriangularfuzzynumbersand
theparametersare(0.9925, 0.9960, 0.9982). Thecor-
respondingoccurrenceprobabilityis0.99250.9982,
however, the most possible probability is equal to
0.9960withamembershipgradeequal to1.
3.3 Sensitivity analysis
Themainbasic eventsaffectingtheoccurrenceprob-
ability of thetopevent can bedeterminedandsome
Figure2. Possibilitydistributionof cutterheadfailureprob-
abilityof compositeEPB shield.
Table1. Theclassificationof riskoccurringprobability.
Rank Accident description Interval probability
One Impossible P-0.01%
Two Infrequent 0.01%P-0.1%
Three Occasional 0.1%P-1%
Four Possible 1%P-10%
Five Frequent P10%
effectivemeasuresareverifiedbysensitivityanalysis
to reduceoccurrenceprobability of thebasic events
and the top event. According to the fuzzy number
model defined in this paper, sensitivity evaluation
indexV
i
issimplydefinedinEq.(10) (ChenandZhang
2002)
Where j
g
is the occurrence probability of the top
event, j
xi
istheaverageoccurrenceprobabilityof the
basicevent x.
If V
i
V
j
,itismoreeffectivetominimizetheoccur-
rence probability of the top event by reducing the
occurrenceprobability of theevent i rather than the
event j.
598
Table 2. Fuzzy probability of the basic event in the fault
tree.
Fuzzyprobabilityvalue
Basic Sensitivity
event a m b index
X
1
0.000095 0.01% 0.000105 1.0043e-6
X
2
0.000475 0.05% 0.000525 5.0216e-6
X
3
0.095 10% 0.105 0.0083
X
4
0.8075 85% 0.8925 0.9532
X
5
0.0855 9% 0.0945 0.0803
X
6
0.0475 5% 0.0525 0.0502
X
7
0.0095 1% 0.0105 0.0010
X
8
0.76 80% 0.84 0.8035
X
9
0.000665 0.07% 0.000735 7.0302e-4
X
10
0.000475 0.05% 0.000525 5.0216e-4
X
11
0.0057 0.6% 0.0063 0.0057
X
12
0.00076 0.08% 0.00084 6.8294e-4
X
13
0.095 10% 0.105 0.0854
X
14
0.0038 0.4% 0.0042 0.0034
X
15
0.076 8% 0.084 0.0803
X
16
0.057 6% 0.063 0.006
X
17
0.00855 0.9% 0.00945 0.009
X
18
0.00076 0.08% 0.00084 8.0345e-5
X
19
0.000665 0.07% 0.000735 7.0302e-4
X
20
0.000855 0.09% 0.000945 9.0389e-4
X
21
0.76 80% 0.84 0.8035
X
22
0.0855 9% 0.0945 6.3272e-4
X
23
0.00665 0.7% 0.00735 6.3272e-4
X
24
0.0057 0.6% 0.0063 0.0051
X
25
0.0038 0.4% 0.0042 4.0173e-4
X
26
0.00057 0.06% 0.00063 5.1e-4
X
27
0.00038 0.04% 0.00042 4.0173e-4
X
28
0.057 6% 0.063 0.0603
Sensitivity indices of all basic events areobtained
bysensitivityanalysisof thefaulttreeof thecutterhead
failureriskof compositeEPBshieldasshowninFig.1.
The basic event in which the sensitivity index
is greater than 5% is chosen and arranged as fol-
lows: V
4
=95.32%, V
8
=V
21
=80.35%, V
13
=8.54%,
V
5
=8.03%, V
28
=6.03%, V
24
=5.1%, V
13
=5.02%.
4 MAINLY INFLUENCE FACTORAND
IMPROVENMENT MEASURE
Accordingtotheorder result, singleminimumcut set
X
5
, X
6
, X
5
, X
8
, X
21
and so on easily cause the fail-
ureof cutterhead; Secondly, thebasic events X
4
, X
7
,
X
24
whichappear moretimesalsoeasilycausethetop
event failure. They aretheweaknesspartsof thesys-
temandthemainriskfactorsarousingthefailureof the
shieldcutterheadinthemixedfaceground.Therefore,
during the tunnel construction, it aims at surveying
andmanagingthebasiceventswhichgreatlyinfluents
thetopeventoccurrencetolower theriskaccidentsof
cutterheadfailure.
1. For lowering the risks X
5
(badness geology), X
8
(soft andstickygeology), X
21
(alternatedwithsoft
and rigidity rock terrane) influences on the cut-
terhead failure, it should strengthen to run the
geologyforecast, accuratelycertainthepositionof
thebadness geology and its distribute, and adopt
correspondingmeasuresinadvance.
2. ForloweringtheriskX
6
(irrationalitytypeof cutter-
head) influenceonthecutterheadfailure, it should
accord to the geology and hydrology condition,
structural design, constructionadvancerequestetc.
factor, choose much adaptability cutterhead, and
makeanadequacyadjustmentof shieldinstallation
under concreteconditionsduringconstruction.
3. For loweringtherisksX
4
(exceedingallowedabra-
sion), X
7
(misgovern construction) influences on
the cutterhead failure, it should choose reason-
ableadvancemodel andparameters, continuously
accumulatetheexperience, andreduceman-made
breakage; installing the cutter wear monitor sys-
tem, itcanaccuratelyobtaintheinformationof the
cutter wear, then adopt corresponding measures;
enhancingcuttersreplacingrateandquality.
5 CONCLUSIONS
Fromriskanalysisof compositeEPBshieldcutterhead
failureinthemixedfacegroundbasedonfuzzyFTA,
theconclusionsareasfollows:
1. Having a directly view and simple character, the
FTA isavalidmethodfor analyzingthefailurerisk
of compositeEPB shieldcutterheadinmixedface
ground.
2. It totallyconsidered28basiceventsfor thefailure
risk fault tree of composite EPB shield cutter-
head. Through the fuzzy fault tree calculation, it
candefinitetheweaknesspartsof cutterheadfail-
ure, confirmthe key factors of risk occurrence,
and makean order for theimportanceof various
influencefactors.
3. Through fuzzy FTA of the cutterhead failure, it
made sure the mostly reason and mechanismof
leadingtothefailureriskof compositeEPB shield
cutterhead,broughtforwardtheimprovedmeasures
andsuggestion. Soit gavesomeuseful conference
for preventingor reducingthefailurerisk of com-
positeEPB shieldcutterheadduringconstruction.
REFERENCES
Cao,H.L.,Lv,C.T.&Li,J.X.2004.Failureanalysisforcutters
andits preventionmeasures. Tunnel Construction 24(6):
910.
Evans,J.R.2001.Itroduction to Simulation and RiskAnalysis.
2/e, PrenticeHall BusinessPublishing.
599
Furuta, H. & Shiraishi, N. 1984. Fuzzy importanceinFault
treeanalysis. J.Fuzzy sets and systems 12: 205213.
Guo, L. 2006. Fault elimination of mixed face shield dur-
ingconstruction. Modern Tunneling Technology (supple-
ment): 436439.
Meng, X.J. 2004. Causation analysis and resolvemeasures
for theordinary fault (damage) of themixedEPB shield
cutterhead. Tunnel Construction 24(2): 6166, 73.
Singer, D. 1999. Fuzzy set approach to fault tree and
reliabilityanalysis. J.Fuzzy sets and systems 34: 145155.
Tanaka, H., Fan, L.T. & Taguchi, T. 1983. Fault treeanaly-
sisby fuzzy probability. J. IEEE Trans on Reliability 32:
453457.
Tongji University2007. Guidelinesof RiskManagement for
Metro. Tunnelling and Underground Engineering Works.
2007.
Zadeh, L.A. 1965. Fuzzy sets Information and Control 8:
338353.
Zhang, G.D. & Lu, Y.X. 1990. Analysis and design
of reliability and maintenance of system. Beijing:
Beijing Aeronautics and Astronautics University Press:
120125.
Zhao, J., Gong, Q.M. & Eisensten, Z. 2007. Tunnelling
throughafrequentlychangingandmixedground: A case
historyinSingapore. Tunnelling and Underground Space
Technology 22: 388400.
Zhu, W.B., J u, S.J. etc. 2006. Shield tunnelling technology in
mixed face ground conditions. Beijing: ChinaScienceand
TechnologyPress.
600
Riskassessment onenvironmental impact inXizangRoadTunnel
C.P.Yao, H.W. Huang& Q.F. Hu
ABSTRACT: Nowadays, with theincreasein thedemands for environmental protection, theassessment of
constructionontheenvironmental impact has becomeanewtopic. BasedontheRisk Management Software
(TRM 1.0) whichisdevelopedbyTongji University, theauthorsinvestigatethesurroundingenvironment con-
ditionandconstructionprocess, calculatetheprobability of hazardrisksanditsloss, andprovidetheranksof
hazardrisksandthecorrespondingtreatmentonrisks. Itmaygivesomeadviceandsuggestionsonunderground
construction.
1 INTRODUCTION
Becauseof itsuniqueunseencharacter, isolationfrom
nature and complexity, the construction of under-
ground works is more difficult than the process of
general buildings. In the age of high demands of
environmental protection, the environmental impact
causedby undergroundworksconstructionisattract-
ing increasing attention. The paper is for the risk
assessment on environmental impact caused by the
construction of Shanghai Xizang South Road Tun-
nel fromthe point view of risk. The authors make
quantitative analysis on the impact of construction
onsurroundingbuildings, theimpact onsurrounding
roadsandtheimpactonsurroundingpipelines.Atlast,
theranks of hazard risks areprovides as well as the
Figure1. Xizangsouthroadtunnel plan.
riskreductionmeasures. It maygivesomeadviceand
suggestionsonundergroundconstruction.
2 PROJ ECT PROFILE
XizangSouthRoadTunnel isaplanningriver-crossing
tunnel inShanghai Expo.Themainlineof thetunnel is
fromtheintersectionof Puxi XizangSouthRoadand
ZhongshanSouthRoadtotheintersectionof Pudong
BinzhouRoadandYunlianRoad, includingtheinter-
sectionof BinzhouRoadandPudongSouthRoad.The
total project is 2673meters long. Figure1shows the
roadtunnel plan.
The tunnel is advanced by Slurry Balance Shield
machine of 11.36-meter diameter. Project consists
601
Risk assessment on
environmental impact
Impact on
surrounding
buildings
Impact on
surrounding
roads
Impact on
surrounding
pipelines
Figure2. Riskidentificationmapof environmental impact.
of the approach of Puxi, Puxi rectangular tunnel
(includingwells), thecircular tunnel, Pudongrectan-
gular tunnel (includingwells) andPudongapproach.
Alongthesettinglocation, therearegreat population,
manybuildings, complexpipelinesandcomplextraf-
ficsystems. Also, thetunnel will up-crossLine7and
down-crossLine8, bothof whicharecomponentsof
Shanghai RailwayTransit System. Moreover, thetun-
nel will intersect withmagnetic trainlinewhichwill
beconstructedsoon. Inthelater part, theauthorswill
assesstheriskof environmental impactcausedbycon-
struction, includingimpact onsurroundingbuildings,
onsurroundingroadsandonsurroundingpipelines. It
isshowninFigure2.
3 RISK ASSESSMENT ONENVIRONMENTAL
IMPACT
3.1 Risk assessment process
Risk assessment is usually divided into three steps
(Guo, 1986& Chen, 2004):
1 Risk identification: Analyzeall potential risk fac-
tors whichmay influencetheenvironment during
Table1. Riskinvestigationform(part of it).
Environmental assessment Occurrenceprobability Loss
Riskhazard Risk-inducingfactor Probabilityrank Confidenceindex Lossrank Confidenceindex
Table2. Riskoccurrenceprobabilityranks.
Ranks A B C D E
Occurrence Impossible Seldom Occasional Possible Frequent
probability P-0.01% 0.01%P-0.1% 0.1%P-1% 1%P-10% P10%
Table3. Lossranksof riskhazards
.
Ranks 1 2 3 4 5
Details Ignored Considered Serious Veryserious Disastrous
Here, thecriteriononrankof lossdiffersfromcountrytocountry. InChina, thelossisdisastrouswhen5liveslost.

theconstructionprocess, andthenclassify, collat-
ingthoseparametersaffectinggreatly.
2 Riskanalysis: Calculatetheprobabilityandconse-
quenceof riskfactors.
3 Riskevaluation:Accordingtosomecertaincriteria,
evaluatetheriskfactors.
Where, risk evaluation is to use Experts Investi-
gationMethod(EIM) andConfidenceIndex Method
(CIM), to analyze the probability and consequence
of risk factors, and to gain the ranks of risks. EIM
is akind of gathering information method. Wesend
investigation forms concerning with the risk factors
in engineering to experts, professors, senior consul-
tersandsoon. TheCIM isjust anindextoshowtheir
confidence when doing the judgment. The data can
beput intothedatabaseof TRM 1.0, whichis devel-
opedbyTongji University (Huang, et al 2006b). The
detailed data-input procedure can refer to the refer-
enceHuang, etal 2006b.Table1showspartof therisk
investigationform.
In this paper, the applied risk evaluation crite-
riais Guidelines for TunnellingRisk Management
enacted by the International Tunnel Association in
2002andGuidelinesof RiskManagementfor Metro
Tunnelling and Underground Engineering Works
enacted by Tongji University (International Tunnel
Association, 2002 & Huang, et al 2006a). They are
showninTables2, 3, 4and5.
3.2 Risk assessment on environmental impact
on surrounding buildings
3.2.1 Risk identification
The possible risk hazards of surrounding buildings
causedby constructioncanbeillustratedinFigure3
(Yao, et al 2006).
602
3.2.2 Risk analysis
The overall sinking of buildings is caused by the
even settlement of ground. Uneven settlement can
inducethetiltingandcrackingof buildings. Duringthe
constructionprocess, many factors cancauseground
settlement (Yao & Huang, 2007). For example, the
seepageof shield and pit, pit collapse, theimproper
earth pressure of shield and support damage. Such
factors may also bethecauseof buildingdamage. If
this happens, it will causegreat social influenceand
economicloss.
Table4. Riskevaluationmatrix.
Loss
Risk 1. Ignored 2. Considered 3. Serious 4. Veryserious 5. Disastrous
Occurrence A: P-0.01% 1A 2A 3A 4A 5A
probability B: 0.01%P-0.1% 1B 2B 3B 4B 5B
C: 0.1%P-1% 1C 2C 3C 4C 5C
D: 1%P-10% 1D 2D 3D 4D 5D
E: P10% 1E 2E 3E 4E 5E
Table5. Riskacceptancecriteria.
Ranks Risk Acceptancecriteria Measures
I 1A,2A,1B,1C Ignored Noneedfor management andsupervision
II 3A,2B,3B,2C,1D,1E Allowable Call for attention, needregular management andsupervision
III 4A,5A,4B,3C,2D,2E Accepted Call for great attention, needprevention, andmonitoringmeasures
IV 5B,4C,5C,3D,4D,3E Unaccepted Needthedecisionof policy-maker, needcontrollingmeasures
V 5D,4E,5E Cannot beaccepted Immediatelystop, needalternativeplans
Table6. SurroundingbuildingsinPuxi andPudongDistrict.
District Buildings
Puxi Manufacturing 6buildingswith56floors, 2buildingswith4floors,
district bureau 2buildingswith1floor, 2buildingswith3floors, 1buildingwith10floors, 1building
with7floors
Xizangsouthroad Manybuildingswith37floors, 2buildingswith15floors, 1buildingwith12floors,
1buildingwith10floors, several buildingswith1floor, 1buildingwith19floors,
1buildingwith28floors(reservedfor Shanghai Expo)
Southstationroad 1buildingwith16floors, 1buildingwith5floors, 2buildingswith6floors, 1building
with7floors
J iangnanshipyard Three-floor defenseproject, 1buildingwith6floors, distributionsubstationwith
5floors, 3docks
Pudong Shanghai hangbiao Floatingpier, floodwall, 3steel pipepiles, several buildingswith36floors
district factory
Nanchuanfactory Plant with13floors, 1buildingwith4floors, 2buildingswith6floors
Shangnanroad Several buildingswith13floors, 1buildingwith15floors, 1buildingwith13floors,
alibrarywith4floor
Pudongsouth Several buildingswithhighfloors, buildingswith6floors
road-Yaohuaroad
Binzhouroad 10buildingswith46floors
Thepassingbuildings of thetunnel canbeshown
inTable6, Appendix1& 2, respectively.
Wemaysee, thebuildingsareverydense, therefore
therisk is high. It is very easy totilt or crack for the
buildingsduringtheconstruction.
3.2.3 Risk evaluation
Bysendingoutinvestigationforms,andusingtheTRM
1.0, theauthors get therisk ranks of theserisk haz-
ards. TheresultsareshowninTable7, fromwhichwe
may seebuildings crack is morerisky and is highly
probabletohappen.
603
Impact on surrounding buildings
Overall
sinking of
buildings
Building tilt
Building
crack
Figure 3. Risk identification of impact on surrounding
buildings.
Table 7. Risk assessment of impact on surrounding
buildings.
No. Riskhazard Probability Loss Ranks
1 Overall sinking B 2 II
of buildings
2 Buildingtilt B 3 II
3 Buildingcrack C 3 III
Road risk
Road damage Traffic jam
Surface
subsidence
Road uplift
Road
fracture
roads.
3.3 Risk assessment on environmental impact on
surrounding roads
Thepossibleriskhazardsof surroundingroadscaused
byconstructioncanbeillustratedinFigure4.
3.3.2 Risk analysis
Thetrafficsituationalongtheworksisextremelycom-
plex. Thetunnel isthelink betweenPuxi andPudong
district, along with Nanpu Bridge, Lupu Bridgeand
Da-puroadtunnel. Oncetheroadisdamaged, causing
trafficjam, theconsequenceswill beserious.
Therearemanyroadsalong, butbecauseof thelimit
of thepages,thefigureof roadsisomitted.Theanalysis
isjust thesameasabove.
1.0, theauthorsgettheriskranksof theseriskhazards.
Table8. Riskassessment of impact onsurroundingroads.
No. Riskhazard Probability Loss Rank
1 Road Surface C 2 II
damage subsidence
Roaduplift C 2 II
Road C 3 III
fracture
2 Trafficjam D 2 III
Pipeline damage
Gas
pipeline
damage
Water
pipeline
damage
Sewage
pipeline
damage
Telecom
pipeline
damage
Electricity
pipeline
damage
pipelines.
Table 9. Risk assessment of impact on surrounding
pipelines.
No. Riskhazard Probability Loss Rank
1 Gaspipelinedamage B 3 II
2 Water pipelinedamage B 4 III
3 Sewagepipelinedamage C 2 II
4 Telecompipelinedamage B 3 II
5 Electricitypipeline C 3 III
damage
TheresultsareshowninTable8, fromwhichwemay
seetraffic jamand road fracturearemorerisky and
arehighlyprobabletohappen.
3.4 Risk assessment on environmental impact on
surrounding pipelines
Thepossibleriskhazardsof surroundingpipescaused
byconstructioncanbeillustratedinFigure5.
3.4.2 Risk analysis
There are many pipelines along, but because of the
limit of thepages, thefigureof pipelines is omitted.
Theanalysisisjust thesameasabove.
1.0, theauthorsgettheriskranksof theseriskhazards.
TheresultsareshowninTable9, fromwhichwemay
seewater pipelinedamageandelectricitylinedamage
aremorerisky.
604
Table10. Riskassessment of environmental impact.
No. Riskhazard Rank
1 Impact onsurroundingbuildings III
2 Impact onsurroundingroad III
3 Impact onsurroundingpipelines III
3.5 Summary
Thewholerisk rank for environmental impact is III.
TheresultisshowninTable10. Itshouldbepaidmore
attention.
Roadtunnel constructiononsurroundingbuildings
impact was mainly due to surface subsidence dur-
ing theprocess, causing buildings sinking, tilting or
cracking. Themaintreatmentsare:
1 Grout timely, and make sure the shield is well
sealed.
2 Control the earth pressure during the digging to
prevent the face instability, a surface subsidence
anduplift.
3 Protect theimportant buildings. If it is necessary,
consolidatethefoundation.
4 All kinds of anti-leakage measures are needed.
Pay attentiontofoundationstrengtheningprocess,
otherwiseit will most pronetolandslides.
5 Focus on the steps of construction during the
foundationengineering. Control every stepof the
excavationdepthandslope, includinggoodsupport
andthetimelyinstallation;
Themeasures about road and traffic involvewith
followinginstructions.
1 Strictlycontrol thequalityof temporaryroadcon-
struction, toensureroadsafety.
2 Control theloadof vehiclestravellingontheroad
toprevent overweight andcrushingroad.
3 Detailedly understand the traffic flow near the
project before starting the construction. Disperse
trafficflowtoavoidjam.
There are many large diameter pipelines near the
project, and oncethepipelines aredestroyed, it will
not onlyhavehugeeconomiclosses, but alsoresulted
in a very bad social influence. Therefore we should
takeprotectivemeasuresagainst dangerouspipelines.
1 Since there may be some errors occurring when
planting the pipelines, which may induce incor-
rect positionof pipelines. So beforeconstruction,
positionsof important pipelinesmust beverified.
2 Protect important pipelinesby consolidatingearth
or isolationmethod.
3 Removesomeloadpushedonthepipelines.
4 Take informational construction. Monitor the
pipelinesfrequently.
4 CONCLUSION
Fromthestudy andresearchonthis project, wemay
get someuseful conclusions:
1 Thewholerisk rank for XizangSouth RoadTun-
nelsenvironmental impact isIII. It shouldbepaid
more attention. When the tunnel is constructed,
theworkersmaystrictlycontrol groundsettlement.
Informationconstructionisrecommended.
2 There are plenty of methods to evaluate risk. It
needs todiscuss deeply whichmethodis adopted.
Inthispaper, ExpertsInvestigationMethod(EIM)
andConfidenceIndex Method(CIM) areapplied,
andtheyareapplicable.
3 Theriskrankof environmental impact isobtained,
whichmaygiveadviceonprojectdecision, bidding
andinsurance.
4 Risk management isadynamicprogress. Withthe
development of theproject, therisk rank may be
changed.Theauthorsshouldfollowtheproject, and
realizethedynamicriskmanagement.
REFERENCES
Chen, L. 2004. RiskAnalysis andAssessment DuringCon-
structionof Soft Soil ShieldTunnel inUrbanArea. PhD
Thesisof Tongji University.
Guo, Z. W. 1986. Risk analysis and decision. Beijing:
MachineryIndustryPress.
Huang,H.W.etal 2006a.Guidelinesof RiskManagementfor
MetroTunnellingandUndergroundEngineeringWorks.
Tongji University.
Huang, H. W. et al. 2006b. Risk Management Software
(TRM1.0) BasedonRisk Databasfor ShieldTunnelling.
Chinese Journal of Underground Space and Engineering.
Vol. 2: 3641.
International Tunnel Association. 2002. Working Group
No. 2. Guidelines for tunnelling risk management.
Balkema.
Yao, C. P., Huang, H. W. & Hu, Q. F. 2006. The Study on
ChoosingConstructionMethods of UndergroundWorks
BasedonTheRiskAnalysis. Modern Tunnelling Technol-
ogy. Vol. 43: 581585.
Yao, C. P. & Huang, H. W. 2007. Risk analysis on the
construction of connection tunnels in Shanghai Yangtze
River Tunnel. Underground Construction and Risk
Prevention Proceedings of the 3rd International Sym-
posium on Tunnelling-Shanghai 2007: 99104. Tongji
UniversityPress.
605
Appendix1. SurroundingbuildingsinPuxi district.
Appendix2. SurroundingbuildingsinPudongdistrict.
606
Riskanalysisandfuzzycomprehensiveassessment onconstruction
of shieldtunnel inShanghai metroline
H.B. Zhou, H.Yao&W.J. Gao
Shanghai Jianke Project Management Co., Ltd., Shanghai, P.R. China
ABSTRACT: Basedonriskmanagementof metrolineinShanghai theriskanalysisandfuzzycomprehensive
assessmentmodel wasintroducedtotheverylongmetrolineprojectriskassessmentof shieldtunnel construction.
Firstly, Work BreakdownStructuremethodwas brought forwardtobreak downtheshieldtunnel construction
works accordingtothefeatures of theengineeringgeology andsurroundingbuildingandundergroundpublic
pipes, thefault treemethodwasusedtoidentifytheshieldtunnel constructionsriskeventsandriskfactorsand
therisklistcanbeestablished. Secondly, thefuzzycomprehensiveassessmentmodel waspresentedtoassessthe
constructionrisk inviewof thecomplexity of undergroundspaceanddifficulty of quantifyingtheevaluation
parameters, andthesub-projectsrisk level andtheproject overall risk level canbecalculated. Finally, acase
about verylongmetrolineproject inShanghai wasstudied, theassessment resultsisconsistent withreality.
1 INTRODUCTION
Eleven metro lines projects with the total length of
more than 400 kilometers will be constructed in
Shanghai before2010, theconstructionof metrosys-
tems inShanghai is inLarge-scaleconstructionnow.
Metro systems can effectively decrease traffic con-
gestion, at thesametime, dueto theconstruction of
metromainlytakenundergroundwithmanyuncertain-
ties as well as limited time, many serious accidents
has happenedsuchas theconstructionaccident hap-
pened in Shanghai Rail Transit Line 4 in J uly 2003
and thecollapsein Guangzhou Metro Linein 2004.
Theseaccidentsnotonlybroughtenormouslosstothe
country, but also exertedanegativeinfluenceonthe
public. So, theriskassessmentontheshieldconstruc-
tionhasvital significancetoinsuretheprojectproceed
smoothly(Mao2004, Su2004).
In this paper, a new risk analysis and assessment
method during very long shield tunnel construction
wasput forward. Firstlytherisk analysismethodwas
introduced, which analyzed metro construction risk
combining theWBS method with fault tree method
because that the very long metro constructions risk
wasverycomplex,intheriskanalysisprocess,thevery
longmetrolineprojectwasclassifiedaccordingtothe
features of theengineeringgeology andsurrounding
building. ThentheWorkBreakdownStructure(WBS)
methodwasbroughtforwardtobreakdowntheshield
tunnel construction works. The fault tree method
was used to identify theshield tunnel constructions
risk events and risk factors, so the risk list can be
established. Secondly, a new fuzzy comprehensive
assessment model was presented to assess the con-
structionriskinviewof thecomplexityof underground
space and difficulty of quantifying the evaluation
parameters, which was proposed based on thefuzzy
mathematical theoryandriskmatrixmethodproposed
byInternational TunnelingAssociationin2004.Atlast
a case about one very long metro line construction
riskassessmentwasintroducedtodemonstratetherisk
analysisandrisk assessment method. Theassessment
resultswereconsistent withreality. All theriskanaly-
sisandassessmentideasandmethodgivenaboutvery
longmetrolineproject areof great practical valuefor
similarmegaengineeringconstructionriskassessment
inthefuture.
2 RISK ANALYSISMETHODRESEARCH
Now, thereweremany risk analysis methods suchas
WBS-RBSmethod, faulttreemethodandsoon. If the
WBS-RBS methodwas usedto analyzetherisk fac-
torsof constructionof verylongmetrolineproject,one
veryhugematrixwouldbeestablishedtoidentifythe
riskandthemethodcouldnotdistinguishrelationship
between the risk factors. The fault tree method also
couldnot solvetheproblems. So anewrisk analysis
methodthatcombinedtheWBSmethodwithfaulttree
methodwasproposed. Firstly, theverylongmetroline
project wasclassifiedaccordingtothefeaturesof the
607
engineeringgeology andsurroundingbuilding. Then
theWork Breakdown Structuremethod was brought
forwardtobreakdowntheshieldtunnel construction
works. The fault tree method was used to identify
the shield tunnel constructions risk events and risk
factors, sotherisklist canbeestablished.
2.1 Work breakdown structure method
Inordertoanalyzetherisk,Theworkbreakdownstruc-
ture (WBS) method was brought forward to break
down the metro construction works and to analysis
theriskeventsandriskfactors.
2.1.1 The first grade structure breakdown
The metro line was divided into several sections to
analyze risk according to the features of the engi-
neering geology, surrounding building, underground
publicpipesandtrafficconditions.
2.1.2 The second grade structure breakdown
In order to analyze the risk of every section com-
prehensivelyduringtheconstruction, theactivitiesof
eachsectionaredividedfurther intoportal-out, shield
driving, segment works, groutingconstruction, shield
machine, portal inandcontact channel.
2.2 Risk identification
The fault-tree method was used to identify the risk
events and risk factors based on breaking down the
works(Loosemore, Raftery, Reillyetal 2006), thenthe
riskevent list canbeestablished.
3 FUZZY ASSESSMENT METHOD
RESEARCH
Therehavebeenlimitedattemptstoexploitfuzzylogic
withintheconstructionriskmanagementdomain. Pre-
vious approaches to the use of fuzzy logic within
constructionriskmanagementhaveprovedtobeeither
too simplistic for useinthereal world, or havebeen
very specific intheir approach, targetingaparticular
areaof constructiononwhichtoact or concentrating
onspecifictypesof risks.
Inthispaper,anewfuzzycomprehensiveevaluation
model usedtotheriskassessmentwasproposed,which
attemptedtosolvepreviousproblem. Thefuzzycom-
prehensiveevaluationmodel wasestablishedbasedon
thefuzzymathematical theoryandriskmatrixmethod
proposed by International Tunneling Association in
2004. Itcancalculatetheevaluatingindicators degree
of membership and the weights, and the model can
makesurethecomputedresult ismoreobjective.
3.1 Calculating the indicators weights
Theindicators weightsweredeterminedby theAna-
lytic Hierarchy Process(AHP) methodby comparing
Table1. Estimationmethodof theriskeffect consequence.
Estimate
Degree value Explanation
Slight 1 Thelossisnot obvious
Medium 2 Thelossisfew(minimumin
100,000RMB)
Serious 3 Thelossiscompensable
(minimumin1,000,000RMB)
Significant 4 Thelossisgreat but compensable
(minimumin10,000,000RMB)
Disastrous 5 Thelossistooenormousto
compensate(above10,000,000
RMB)
Table2. Estimationof theriskprobability.
Probability Estimatevalue Frequency
Rare 1 -0.0003
occasional 2 0.00030.003
Possible 3 0.0030.03
anticipated 4 0.030.3
frequent 5 >0.3
theindicators importance(Fang&Wang1997, Shao
2004).
3.2 Calculating the risk events degree of
membership
Firstly, theexpertscoringmethodwasusedtoestimate
eachrisk events values of effect consequenceC and
possibilityP, thentheproduct of P multiplyCwasled
to the risk ranks membership function and the risk
level degreeof membershiptotheeveryriskeventare
determined.
3.2.1 Determination of the risk event effect
consequence C
Theeffect consequenceC of risk event isdetermined
accordingtotheshieldtunnelingconstructionsactual
situation(Mao 2004). Theestimatemethodis shown
ontable1.
3.2.2 Determination of the risk events possibility
Therisk eventsoccurrenceprobability isrankedinto
fivelevels; estimatemethodisshownontable2.
3.2.3 Calculating the degree of membership
3.2.3.1 Determinationof membershipfunction
In2004, theInternational Tunnel Associationput for-
ward therisk matrix theory used to assessment risk,
in this theory; the risk was classified four rank. In
608
level 1 level 2 level 3
level 4
0
0.2
0.4
0.6
0.8
1
1.2
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
PC
d
e
g
r
e
e

o
f

m
e
m
b
e
r
s
h
i
p
Figure1. Membershipfunctionfor risk.
Table3. Themembershipfunction.
Risklevel Membershipfunction
level 1
level 2
level 3
level 4
this paper, therisk rank also was classifiedinto four
levels. Themembership function was theL-R fuzzy
functionwhichusedinproject verywidely, asshown
on Equation 1, to be calculated easily, and accord-
ing to the feature of shield tunnel construction risk
assessment, thefunctionwas designedto trapezoidal
function, as shown on figure1 and table3. (Carr &
Tah2001, Wang& Li 1996,Yao& zhou2007).
3.2.3.2 Determinationof membershipdegreevalue
of riskevents
Taketheproduct of occurrenceprobability estimated
value P and multiply consequence effect estimated
Figure2. Thechart of themetroroutemap.
valueC intotherisk rank membershipfunction, then
the corresponding membership degree of the risk
eventsaredetermined.
4 CASE STUDY OF RISK ANALYSISAND
ASSESSMENT
Onelengthof themetrolineisabout34.243kilometers
inall inShanghai area, whichpassesthroughfivedis-
trict of Shanghai including Baoshan district, Putuo
district, J inandistrict, Xuhui districtandPudongnew
districtfromnorth-westtosouth-east.Themetrolines
routemapisshownonfigure2.
4.1 Geological conditions
4.1.1 Engineering geological conditions
Theon-sitesoilswithin65mdeptharetheQuaternary
sediments, which are mainly composed of saturated
soft clay, silty clay and sand. Themain soil profiles
of theconstructionof shieldtunnel aresaturatedsoft
clayandsiltyclay.
4.1.2 The hydrogeological conditions
There are three kinds of groundwater in the soils,
including the upper phreatic water mean embedded
depthis0.50.7munderthegroundsurface,themid-
dle feeble confined water mean embedded depth is
36munder thegroundsurfaceinthe2-2sub-layerof
fifthlayersoilsandlowerconfinedwatermeanembed-
deddepthis414mintheseventhlayersoil.Thewater
elevationof groundwaterwaschangedwiththechange
of season, climateandtide.
4.2 Risk analysis
The metro line was divided into 12 sections to ana-
lyzerisk accordingtothefeatures of theengineering
geology, surrounding building, underground public
pipes and traffic condition, as shown on table 4
then the shield tunneling constructions work struc-
turewas brokedown, so therisk list was established
609
Table4. Thelist of dividedevaluatingsectionsof metroline.
Category surroundingenvironments Mainlytraversingsoilsof shiled
The1st section sub-center of thecityandthecomplexof surrounding soft clayandsiltyclay
environmentsiscommon
the2ndsection sub-center of thecityandsurroundingenvironment is soft clayandsiltyclay
relativelycomplex
the3rdsection center of thecityandtraverseunder theNorthZhongshanroad soft clayandsiltyclay
the4thsection center of thecityandsurroundingenvironment iscomplex soft clayandsiltyclay
the5thSection center of thecityandtraversesthroughSuzhouRiver soft clayandsiltyclay
the6thSection center of thecityandtraversesmetroNo. 1lineandmetro soft clayandsiltyclay
No. 2lineclosely
the7thSection center of thecityandtherearepilesfoundationsintrudingin soft clayandsiltyclay
themetrotunnel
the8thSection crosstheHuangpuriver andconstructionenvironment isvery soft clayandsiltyclay
complex
the9thSection Traverseunder or besidepilesfoundationandsurrounding graysoft clay, clayandsandysilt
environment iscomplex
The10thsection thecomplexof surroundingenvironment iscommon graysoft clay, clayandsandysilt
The11thsection traversestheriver belowandthecomplexof graysoft clay, clayandsandysilt
surroundingenvironmentsiscommon
The12thsection traversesunder metroNo. 2line graysoft clayandclay
M
e
e
t
i
n
g

o
b
s
t
r
u
c
t
i
o
n
s
S
i
n
k
i
n
g
S
o
i
l
s

p
o
u
r
i
n
g

o
u
t
,
q
u
i
c
k
s
a
n
d
C
o
l
l
a
p
s
e
S
e
a
l
e
d

e
q
u
i
p
m
e
n
t
l
e
a
k
i
n
g
A

d
e
p
a
r
t
u
r
e

f
r
o
m

t
h
e
d
e
s
i
g
n
e
d

a
x
e
s

o
f
t
u
n
n
e
l
i
n
g
C
o
n
s
t
r
u
c
t
i
o
n

r
i
s
k
t
h
r
o
u
g
h

t
h
e

m
i
d
d
l
e
c
e
s
s
i
o
n
T
o
o

m
u
c
h

d
i
s
t
o
r
t
i
o
n
o
f

t
h
e

f
l
o
o
d

p
r
e
v
e
n
t
i
o
n
w
a
l
l
p
o
s
i
t
i
o
n

d
e
v
i
a
t
i
o
n
o
v
e
r
s
i
z
e
d
H
a
s

n
o
t

b
e
e
n

a
b
l
e

t
o
s
i
n
k

p
r
o
m
p
t
l
y

p
u
t
s

t
h
e
a
r
r
i
v
i
n
g
S
u
r
v
e
y

u
n
c
l
e
a
r
f
a
i
l
u
r
e

o
f
w
o
r
k
i
n
g

f
a
c
e
g
r
o
u
t
i
n
g

i
s

n
o
t
t
i
m
e
l
y
s
i
n
k
h
o
l
e
f
a
i
l
u
r
e

o
f
w
o
r
k
i
n
g
f
a
c
e
s
e
a
l

o
i
l

i
s
u
n
q
u
a
l
i
f
i
e
d
c
h
a
n
g
e

o
f
g
e
o
t
e
c
h
n
i
c
a
l
c
o
n
d
i
t
i
o
n
p
u
s
h
i
n
g

f
o
r
c
e

i
s
u
n
b
a
l
a
n
c
e
s
h
i
e
l
d

s
t
a
n
d
s
t
i
l
l
s
o
f
t

g
r
o
u
n
d
s
e
a
l

o
i
l
s
q
u
a
n
t
i
t
y

i
s

s
h
o
r
t
d
r
i
v
i
n
g

a
x
i
s

n
o
t
t
i
m
e
l
y

c
o
r
r
e
c
t
i
v
e
o
v
e
r

d
r
i
v
i
n
g

o
r
l
e
s
s

d
r
i
v
i
n
g
e
x
c
e
s
s
i
v
e
m
e
a
s
u
r
e
m
e
n
t

e
r
r
o
r

s
e
a
l
e
d

e
q
u
i
p
m
e
n
t
l
o
s
s

f
l
e
x
i
b
i
l
i
t
y
m
e
c
h
a
n
i
c
a
l

f
a
i
l
u
r
e
c
o
n
s
t
r
u
c
t
i
o
n
p
a
r
a
m
e
t
e
r
s

o
f
i
m
p
r
o
p
e
r
c
h
a
n
g
e

o
f
g
e
o
t
e
c
h
n
i
c
a
l
c
o
n
d
i
t
i
o
n
f
a
i
l
u
r
e

o
f
w
o
r
k
i
n
g

f
a
c
e
Shield tunneling (the 8th section)
Figure3. Shieldtunnelingconstructionsriskeventsandriskfactors.
by using fault treemethod to analyzeshield tunnel-
ing construction risk. As an example, the result of
risk identification about shield driving construction
of the8thkindof sector project (under Huangpuriver
section) wasshownonfigure3.
4.2.1 Weights calculation
TheAHPmethodwasusedtodeterminetheweightof
eachrisk event. Theestimationmatrix wasformatted
basedonthemutual comparisonof important degree
amongtheriskevents.
Operating on the estimation matrix, the vector
corresponding valued W of shield tunneling was as
follows:
W = [0.0460.2070.2700.0260.0460.0780.2070.12]
4.2.2 Membership function determining
Takingtheproductof possibilityPof theriskeventand
consequences C into themembership function, each
risk events degree of membership was determined.
ThejudgingmatrixRwasalsoformattedandthevalue
wasasfollows:
4.2.3 Fuzzy comprehensive assessments
Taking weight vector W and judging matrix R
into Fuzzy Comprehensive Assessment, the whole
610
Table5. Thepoint of eachevaluatedindex.
riskevents P C
Shield meetingobstructions 2 3
tunneling collapse 1 5
construction soilspouringout, quicksand 3 3
sealedequipment leaking 3 2
sinkingof theshield 3 3
adeparturefromthedesigned 4 2
axesof tunneling
constructionriskthroughthe 4 4
middlecaisson
toomuchdistortionof theflood 2 4
preventionwall
Table6. theshieldtunnelingconstructionrisk events risk
level of the8thsection.
construction
workof the risk risk
8thsection level riskevents level
shield level 3 meetingobstructions level 2
tunneling collapse level 3
soilspouringout, level 3
quicksand
sealedequipment leaking level 2
sinking level 3
constructionriskthrough level 3
themiddlecession
adeparturefromthe level 4
designedaxesof tunneling
toomuchdistortionof the level 3
floodpreventionwall
evaluationvector B about thetunnelingconstruction
wasdetermined. Thevaluewasasfollows:
Basedonthemaximal membershipdegreeprinci-
ple, thetunnelingconstructionrisk rank about shield
tunneling of the 8th section is level 3, in which the
highest risk is the construction through the middle
caisson and the lowest risk is construction of meet-
ing underground obstructions and sealed equipment
Table7. Risk level calculatedof eachsectionof themetro
line.
category risklevel
the1st section Level 2
the2ndsection Level 3
the3rdsection Level 4
the4thsection Level 4
total risklevel Level 3
leaking. Inthemeanwhile, risksof other construction
activityduringshieldtunnel constructionin8thsection
werecalculated; theresultof riskassessmentaboutthe
8thsectionisasshownontable6. Intheend, risklevel
of the8thconstructionsectioncouldbedeterminedby
thecalculatingupwards.
Therisklevel of other11constructionsectioncould
bedetermined by thesamemethod and theresult is
shown on table 7. Fromthe table 7, it can be con-
cludedthewholerisk level of this ultra-lengthmetro
lineprojectis3. Socertainmeasuresmustbetakento
reducetheriskbasedontheriskacceptanceprinciple.
5 CONCLUSIONS
In accordance with the study on the risk analysis
andassessment methodof shieldtunnelingconstruc-
tion of very long metro line project and case study,
conclusionsweregivenasfollows:
1 AnewriskanalysismethodthatcombinedtheWBS
method with fault tree method was proposed, in
which classified thevery long metro lineproject
accordingto thefeatures of theengineeringgeol-
ogy and surrounding building, then broke down
the shield tunnel construction works, and used
fault treemethodtoidentifytheshieldtunnel con-
structionsriskeventsandriskfactors. Themethod
not only canidentify therisk events andrisk fac-
tors more comprehensively, but also distinguish
the relationship between each risk events or risk
factors.
2 The fuzzy comprehensive risk assessment model
couldbeusedtosolvetheproblemof non-accuracy,
quantify thesubjectiveconcept, convenient math-
ematical processing and reduce the difference
causedbysubjectivejudgment.
611
3 The fuzzy comprehensive risk assessment model
proposed in this paper could beoperated upward
from the bottom layers of risk structure up to
obtain theoverall evaluation results of thewhole
shieldtunnel constructionproject.Theresultof risk
assessmentisveryconsistentwiththepractical con-
structionconditions. Therefore, themethodabove
mentionedisverypractical andcredible.
REFERENCES
Carr, V. &Tah, J.H.M. 2001. A fuzzy approachtoconstruc-
tion project risk assessment and analysis: construction
projectriskmanagementsystem.Advances inEngineering
Software (32): 847857.
Fang, S.C. &Wang, D.W. 1997. Fuzzy mathematics and fuzzy
optimization. Beijing: SciencePublishingCompany.
Loosemore,M.,Raftery,J.,Reilly,C.&Higgon,D.2006.Risk
Management in Projects. London:Tayor andFrancis.
Mao, R. 2004. Risk management onundergroundengineer-
ing. International Proseminar Memoir about Engineering
Managemengt(11): 9498.
Shao, R.Q. 2004. A multi-level fuzzy synthetic evalua-
tiononinvestment programsinshipping. Transportation
Engineering 144(93): 497502.
Su,Y. 2004. Seismicriskmanagementandinsuranceanalysis
of underground structures in soft soil. Doctor Thesis of
TongJi University: 85105.
Wang, Z.P. & Li, H.X. 1996. Fuzzy system theory and fuzzy
computer. Beijing: SciencePublishingCompany.
Yao, H. & Zhou, H.B. 2007. Fuzzy synthetic evaluation on
theconstructingriskof EPBSduringtunnellinginsoftsoil
area. Rock and Soil Mechanics. Vol. 28(8):17531756.
612
Theme 5: Physical and numerical modelling
Tunnel behaviour under seismicloads: Analysisbymeansof
uncoupledandcoupledapproaches
D. Boldini
Department of Chemical, Mining and Environmental Engineering, University of Bologna, Bologna, Italy
A. Amorosi
Department of Civil and Environmental Engineering, Technical University of Bari, Bari, Italy
ABSTRACT: In this paper different approaches to investigatethebehaviour of tunnel under seismic loads
arepresented. Theyincludeone-dimensional (1D) numerical analysesperformedmodellingthesoil asasingle
phasenon-linear visco-elasticmedium, theresultsof whicharethenusedtoevaluatetheinput datafor selected
analytical solutionsproposedintheliterature(uncoupledapproach), and2DfullycoupledFiniteElement(FEM)
simulationsadoptingavisco-elasticeffectivestressmodel for thesoil (coupledapproach).
1 INTRODUCTION
The dynamic response of tunnels to seismic actions
can be assessed by means of uncoupled or coupled
approaches, depending on whether theevaluation of
theseismicwavepropagationandof thecorresponding
actionsonthestructureisundertakenintwoseparated
stepsor inonesingleanalysis, respectively.
In this paper the transverse seismic behaviour of
an idealised shallow tunnel in soft clay is analysed
bymeansof uncoupledandcoupledapproaches.They
includeone-dimensional (1D) numerical analysesper-
formedmodellingthesoil asasinglephasenon-linear
visco-elastic medium and 2D fully coupled Finite
Element (FEM) simulations adopting avisco-elastic
effectivestressmodel for thesoil.
In theuncoupled approach 1D visco-elastic anal-
yses, performed using the equivalent linear scheme
implementedinthecodeEERA (Bardet et al. 2000),
are aimed at establishing the role of stiffness and
dampingnon-linearityonthefree-fieldsiteresponse.
The results of the analyses at the tunnel depth are
thenusedto evaluatetheinput datafor selectedana-
lytical solutions proposed in theliteratureto predict
thetransverseresponseof thestructurebothfor full-
slip and no-slip conditions (e.g., Wang 1993). This
approach, widely used in the engineering practice,
is based on the assumptions that the free-field site
responseisrepresentativeof theproblemunder study
andthat theseismic event at thetunnel depthcanbe
satisfactorily representedby means of themaximum
shear straininducedinthesoil andthecorresponding
mobilisedstiffness.
Toovercomesomeof thelimitationsof theapproach
describedaboveafullycoupledFiniteElementanaly-
siscanbeadopted, simulatinginthetimedomainthe
soil-structuredynamic interactionduringtheseismic
event. Thislatter isinthiscaserealisticallydescribed
by an accelerogramand the soil by an appropriate
effectivestressformulation.
Theconstitutiveassumptionforthesoil isakeyele-
ment of this class of analyses. A linear visco-elastic
model is often adopted at the scope, for its appar-
ent simplicity and limited number of parameters. In
fact, anumber of commercial codes nowadays allow
theuser to performcoupleddynamic analyses based
on linear elasticity and viscous damping, this latter
accounted for by the Rayleigh formulation. In this
casealimitationof thecoupledapproachisinthenon
obviousselectionof adequateelasticandviscoussoil
parameters, whichsensiblyinfluencetheresultsof the
analysis.
Inthisworkastrategytocalibratetheparametersfor
thevisco-elasticmodel adoptedintheFEManalysesis
proposedbasedonthefreefieldsoil responseresults
obtainedinthecontext of theuncoupledapproach.
A critical comparisonbetweencoupledanduncou-
pledapproachesisoutlinedinthepaperwithreference
to a shallow tunnel excavated by a TBM. A realis-
tic geotechnical characterisation is assumed for the
idealisedsoft claydeposit under study.
2 CASE STUDY
In the present study the acceleration time history
recordedat Kalamata(Greece) duringthe13.XI.1986
615
earthquakeisconsidered.Theoriginal seismicsignal is
characterisedbyadurationof 29.74sandamaximum
accelerationof 0.24g.
The input signal adopted in this work was scaled
at 0.35gandwasfilteredtoprevent frequency levels
higher than 7Hz. A picture of the selected accel-
eration time history after manipulation is given in
Figure1whiletheresultingFourierspectrumisshown
inFigure2.
In all the analyses the seismic signal is assumed
tobeappliedat theoutcropof thedeposit. Thecorre-
spondingbedrockmotionwascalculatedbymeansof a
deconvolutionanalysisperformedbythecodeEERA,
asdescribedinthenext paragraph.
A 60-mthick ideal deposit of soft clay isassumed
asthereferencesoil profile, characterisedby thefol-
lowingphysical andmechanical parameters: plasticity
indexIP=44%, unitweightof volume =17kN/m
3
,
overconsolidation ratio in terms of mean effective
stress R=1.5, small strain shear stiffness G
O
=
variablewithdepth, Poissons ratio
=0.25, damp-
ing ratio D=variablewith depth, coefficient at rest
Figure 1. Modified acceleration time history scaled at
0.35g.
Figure2. Frequency-filteredFourier spectrum.
K
0
=0.6. The water table is assumed at the ground
surface.
A circular tunnel, located at a depth of 15mand
characterised by a 10.10m diameter is selected as
the reference underground structure for the present
study.Theliningisassumedtobecomposedby0.50m
thick precast concretesegments characterisedby the
followingparameters: Youngs modulus E
l
=38GPa,
Poissonsratio
l
=0.25, dampingratioD
l
=5%.
3 UNCOUPLEDAPPROACH
The 1D ground response analysis was performed
withthecodeEERA (Bardetetal. 2000) thatanalyses
the vertical propagation of shear waves in a one-
dimensional layered systembased on an equivalent-
linear visco-elastic scheme. It assumes that theshear
modulusG anddampingratioD arefunctionof shear
strainamplitude.
Theadoptedprofileof thesmall-strainshear stiff-
ness G
O
with depth (Figure3) was evaluated by the
relationshipproposedbyViggiani (1992) asafunction
of theinsitumeanstress, R andIP.
Figure4showsthecurvesof thevariationof thenor-
malisedshear stiffness anddampingratio withshear
strain,definedaccordingtoVucetic&Dobry(1991).
Atotal numberof 31layerswasadoptedtodiscretise
thesoil stratum.
Figure5showstheresultsof theanalysisintermsof
maximumshearstrain
max
, normalisedshearstiffness
G/G
O
, damping ratio D and maximumacceleration
a
max
withdepth.
Values of
max
and G obtained at the depth of
15m, i.e. at thetunnel depth, weresubsequentlyused
Figure3. Profileof thesmall-strainshear stiffnessG
O
.
616
to evaluate the increment of hoop forces and bend-
ing moments acting on the tunnel lining during the
earthquake, accordingtoselectedanalytical solutions
proposedintheliterature.
Here, the solutions proposed by Wang (1993) to
predict the response of the structure are taken into
considerationforbothfull-slipandno-slipconditions.
Themaximumincrementof hoopforceandbending
moment in thetransversedirection of thetunnel are
givenby:
for full-slipconditionsandby:
for no-slip conditions, the bending moment being
the same for the two cases. E
u
and
u
indicate the
Figure4. ModulusreductioncurveG/G
O
andvariationof
dampingratioD withshear strain.
Figure5. Resultsof the1DgroundresponseanalysisperformedwithEERA.
mobilised soil Youngs modulus (evaluated with ref-
erence to the previously calculated mobilised shear
modulus G) and the Poissons ratio (assumed equal
to 0.5) inundrainedconditions, respectively. r is the
tunnel radius. K
1
and K
2
aregiven by thefollowing
expressions:
where:
arethecompressibilityandflexibilityratiosandt and
I thethicknessandthemomentof inertiaof thetunnel
lining, respectively.
Table 1 summarises the computed increments of
hoopforceandbendingmomentfor bothfull-slipand
no-slipconditions.
4 COUPLEDAPPROACH
The coupled analyses were performed by the Finite
Element commercial codePLAXIS, whichdescribes
617
Table 1. Increments of hoop force and bending moment
accordingtoWang(1993).
Full-slipconditions No-slipconditions
LN
max
LM
max
LN
max
LM
max
(kN/m) (kNm/m) (kN/m) (kNm/m)
159 802 473 802
Figure6. Calibrationof theG andD profiles assumedin
theFEM analysesonthebasisof EERA results.
the mechanical behaviour of the soil in terms of
effectivestress.
Inorder toperformacomparativeanalysiswiththe
EERA resultsalinear visco-elasticconstitutivemodel
wasselected. Intheconstitutivemodel viscousdamp-
ingwasaccountedfor bytheRayleighformulation.
4.1 Calibration of the visco-elastic parameters
Theconstitutivemodel employedintheFEManalyses
makesuseof constant-valueelasticandviscousparam-
eters for eachsub-layer of thediscretiseddeposit. In
this respect, it is of paramount importanceto define
appropriatevaluesof theseparametersasafunctionof
thestrainlevel attainedinthesoil deposit duringthe
earthquake.
Inthis paper arecently developedcalibrationpro-
cedureof thevisco-elastic parameters tobeassumed
indynamic FEM analyses is adopted(Amorosi et al.
2007). G andD profilesaresetinsuchawaytomatch
thecorrespondingprofiles obtainedby thefree-field
EERA analysis. To this end the numerical model is
dividedintoarelativelylargenumber of sub-layersin
order toobtainanascloseaspossiblecorrespondence
(Figure6).
Figure 7. Fourier spectra computer by EERA at different
depthsandhigh-energyfrequencyinterval.
Figure8.
R
e
R
profilesassumedintheFEManalyseson
thebasisof EERA results.
Theprofilesof theRayleighcoefficients
R
e
R
are
obtainedcorrespondingly, accordingtothefollowing
relationshipwiththedampingratioD (e.g., Clough&
Penzien2003):
where
n
e
m
arethecircular frequenciesrelatedto
the cyclic frequency interval f
n
f
m
. The frequency
rangewasdeterminedbyselectingtheinterval where
the Fourier spectra computed by EERA at different
depthswerecharacterisedby thehighest energy con-
tent (indicatedinthebox of Figure7). Theresulting
R
e
R
profilesareshowninFigure8.
618
Figure9. MeshemployedintheFEM analyses.
Figure10. Detail of themesharoundthetunnel section.
4.2 Numerical model
The dimension and boundary conditions of the 2D
FEM numerical model were set up after a consid-
erablenumber of preliminary analyses performed to
minimisetheinfluenceof boundaryconditionsonthe
computedresults.
Themeshemployedinthepresentstudyisreported
inFigure9: it is characterisedby awidthequal to 8
times its height. Thebaseis assumedto berigidand
atthelateral sidestheviscousboundariesproposedby
Lysmer & Kuhlmeyer (1969) wereused.
Thedomainwasdiscretisedin243115-nodeplane
strain triangular elements. In the central part of the
mesh, wherethetunnel islocated, theelement dimen-
sionh alwayssatisfiesthecondition:
where V
S
is the shear wave velocity and f
max
is the
maximumfrequencyof theseismicsignal.
A detail of themesharoundthetunnel isshownin
Figure10.
Thefollowingstagesweresimulatedinthenumer-
ical analysis:
first, thetunnel wasexcavatedinundrainedcondi-
tionsimposingavolumelossof 0.4%;
next,theliningwasinstalledandthepost-excavation
consolidationphasewasstudied;
the select seismic signal was then applied at the
bottomof themodel inundrainedconditions.
Inall thestatic stagesof theanalysisthesoil stiff-
nessfor eachsub-layer wasselectedscalingdownthe
valueof thevery small strainshear stiffnessG
O
by a
factor of 0.45 to account for an averageshear strain
level involvedintheexcavationstagesof =0.1%.
Aninterfacewasactivatedbetweenthesoil andthe
lining, characterisedbyanormal andtangentstiffness
Figure 11. Comparison between EERA and PLAXIS
free-fieldsoil responseanalysesat 15mdepth.
correspondingtothatof theadjacentsoil:suchacondi-
tioncanbeconsideredsimilar totheso-calledno-slip
conditionsof theWangssolutions.
4.3 Free-field soil response results
In this paragraph apreliminary comparison between
free-field soil responseresults at thetunnel depth is
providedtochecktheconsistencybetweenthe1Dand
2D proposed approaches. In this casethe2D model
doesnotincorporatethetunnel and, assuch, theresult
canbedirectlycomparedtothat of thecorresponding
1Dfree-fieldanalysis.
Figure11showstheaccelerationtimehistoriesand
theFourier spectracomputedbyEERA andPLAXIS.
A satisfactory agreement is obtained between the
two solutions, demonstratingtheeffectiveness of the
proposedcalibrationstrategy.
619
Figure12. Incrementsof hoopforceandbendingmoment
duringtheseismicevent.
4.4 Results of the coupled analyses
Themaximumincrementof theseismic-inducedabso-
lute hoop forces and positive and negative bending
moments in thetunnel lining werecomputed by the
2D FEM coupled analyses and are represented in
Figure 12. The results are reported as a function of
theangle indicatedinFigure12anddefinedpositive
incounter-wisedirection.
Results indicateasatisfactory match between the
maximum increments of hoop force and bending
moments predicted by the visco-elastic FEM solu-
tion(LN
max
=423kN/meLM
max
=724kNm/m)
andthoseresultingfromtheWangssolutionsfor the
no-slipcase.
5 CONCLUSIONS
Inthispaperthetransversedynamicresponseof ashal-
lowtunnel subjectedtoseismicactionsisinvestigated
bymeansof uncoupledandcoupledapproaches.
Theformerapproach, traditionallyusedintheengi-
neering practice, combines afree-field siteresponse
analysis withaset of analytical solutions to evaluate
the maximumincrement of hoop force and bending
moment in thetunnel lining. This approach is based
on a limited numbers of parameters to describe the
earthquake, thesoil behaviour andthawt of thetunnel
lining. The reliability of the uncoupled approach is
supportedbythesatisfactoryperformanceof anumber
of undergroundstructuresdesignedbasedonit.
Nowadays the availability of relatively sophisti-
cated commercial FEM codes and the continuous
increasein computational power allows to approach
thesameproblembymeansof asingleanalysiswhich,
atthesametime,accountsfortheseismicsiteresponse,
thesoil-structureinteractionandtheactionsinthetun-
nel lining. This coupledapproachis characterisedby
afar moredetaileddescriptionof theseismic actions
at thetunnel depthascomparedtotheuncoupledone
andallowstodescribethesoil behaviour bymeansof
effective stress based constitutive models. Concern-
ingthislast featureinthispaper asimpleassumption
was made: linear visco-elasticity was assumed for
theclayey deposit under study, whereviscous damp-
ing was accounted for by the Rayleigh formulation.
Although very simple this constitutive assumption
already poses some problems in the calibration of
the stiffness and Rayleigh damping parameters. In
fact, it is a well established fact that the profiles of
these parameters with depth crucially influence the
results of any FEM dynamic analysis (e.g.: Wood-
ward& Griffiths1996). Relatedtothis, astrategyfor
thecalibrationof thestiffnessandRayleighdamping
parametersisproposedinthepaper. Theresultsof the
coupledanalysisbasedonthiscalibrationsatisfactory
match those obtained by the uncoupled approach in
termsof hoopforceandbendingmomentsinthelining.
IntheAuthors opinionthisencouragingresultrep-
resents a necessary condition to extend the use of
theproposedcalibrationstrategy to thecaseof more
complexconstitutiveassumptionsincludingplasticity.
REFERENCES
Amorosi, A., Boldini, D., Elia, G., Lollino P. & Sasso M.
2007. Sullanalisi dellarispostasismicalocalemediante
codici di calcolonumerici. IARG 2007, Salerno.
Bardet, J.P., Ichii, K. & Lin C.H. 2000. EERA-A computer
program for Equivalent-linear Earthquake site Response
Analyses of layered soils deposits. User Manual.
Clough, R. & Penzien, J. 2003. Dynamics of Structures.
ComputersandStructuresInc.
Lysmer, J.& Kuhlemeyer, R.L. 1969. Finitedynamic model
for infinitemedia. ASCE EM 90: 859877.
PLAXIS. 2003. Reference Manual, version8.
Viggiani, G.M.B. 1992. Small strain stiffness of fine grained
soils. PhDthesis. CityUniversity. London.
Vucetic, M. & Dobry, R. 1991. Effects of thesoil plasticity
oncyclicresponse. Journal of Geotech. Eng. Div., ASCE
117(1): 89107.
Wang, J.N. 1993. Seismic design of tunnels: a state-of-the-art
approach. Monograph7, Parsons, Brinckerhoff, Quade&
DiuglasInc., NewYork.
Woodward, P.K. &Griffiths, D.V. 1996. Influenceof viscous
damping in thedynamic analysis of an earth damusing
simpleconstitutivemodels. Computers and Geotechnics
19(3): 245263.
620
Investigatingtheinfluenceof tunnel volumelossonpilesusing
photoelastictechniques
W. Broere
A. Broere BV, Amsterdam, The Netherlands
J. Dijkstra
ABSTRACT: This paper presents theresults of planestrainmodel tests ontunnelpileinteractioninapho-
toelastic material. The effects of volume loss are simulated by contracting a tunnel. The soil in this test is
represented by crushed glass. This allows for thedetermination of stresses in themodel by thephotoelastic
method. Theinfluenceof thestresschangeinthesoil duetovolumelossisshown, aswell astheeffect onthree
rowsof pilesat varyingdistancefromthetunnel. Fromthetestsit isclear that significant stresschangesoccur
closetothepiletips.
1 INTRODUCTION
Theupcomingboringof theNorth-Southmetrolinein
Amsterdam(Netzel & Kaalberg2000, Kaalberget al.
2005) will involveTBM excavation closeto thepile
tips of the wooden piles on which much of the his-
toricinner cityisfounded. Theinfluenceof avolume
lossduetotunnellingonthebearingcapacityof these
pilesisrelativelyuncertainandhardtoquantifyeven
by numerical means (Broere & van Tol 2006). This
promptedadifferent approach, tovisualisetheinflu-
ence of volume loss on the stresses around the pile
tips using the photo-elastic method. Photoelasticity
hasbeenusedextensivelytoquantifystressesinhomo-
geneousmaterials(e.g. Frocht1948).Andisextended
to particle assemblies in index matching liquids by
Drescher (1976).
A small research effort was started in 2002 to
attempt to quantify the influence of a volume loss
on the stresses close to pile tips in a photoelastic
scalemodel.Astheinterpretationof themeasurements
provedmorecomplicatedthanoriginally anticipated,
it took considerably moreeffort tofinalisetheanaly-
sis andderivethecompletestress state. Anoverview
of thesecomplexities is given in Petrucci & Restivo
(2007). Standing& Leung(2005) hadsimilar consid-
erationsasoutlinedabovewhenthey didcomparable
photoelastic scale model tests, but they decided to
invert theproblemand only qualify theinfluenceof
pileinstallationonthestressinexistingtunnel linings.
2 STRESSMEASUREMENTS
It is possible to obtain detailed information on the
stressdistributionof agranularmaterial usingthepho-
toelastic measuring method. The soil is replaced by
grainsof aphotoelasticmaterial, inthiscaseglasspar-
ticles (Wakabayashi 1957, Drescher 1976). Crushed
glass behaves similar to sandparticles, although, the
grains are more angular. Therefore it is a reason-
able substitute to investigate sand behaviour. How-
ever, other authors, most noticably Lesniewska &
Sklodowski (2005) argue that when uniform glass
beadsareusedinsteadof crushedglassthestresstra-
jectoriesalongfailureplanescanbebetter visualized.
When compared to thephotoelastic measurement
methods for continuous materials the quantification
of thestresspathsfor agranular material ishampered.
First, inorder toeliminatelight scatter theporeshave
tobefilledwithaliquidwhichhasasimilar refraction
index as theglass particles. Secondly theanalysis of
thefringepatternsisimpossiblebecauseof thestacked
nature of the sample. Each layer of grains produces
its own fringe pattern resulting in a mix of fringes
whichcannot beinterpreted. Thefirst conditionleads
tothequalitativevisualizationof thesestresspathsbya
circular polariscope, seei.e. Wakabayashi (1957). For
thesecondproblem, quantificationof thestresspaths,
Allersma(1982) developedanautomatedpolariscope.
In general, granular material does not behave
elastically, as is often assumed when formulating a
621
Figure1. Front viewof model setup, all measuresinmm.
stressstrain relationship. Therefore it is not possible
to derivethestress increments fromstrain measure-
ments.Thestresseshavetobemeasuredindependently
of thestrains.
The method developed by Allersma (1987) and
Allersma & Broere (2002) for measurement of the
stress in a photoelastic material is followed. The
method already published by Allersma in 1982, is
most comparabletophase-steppingphotoelasticityas
described in e.g. Ajovalasit et al. (1998). The direc-
tion of the principal stress and the principal stress
differencealongthelight patharemeasuredwithan
automated circular polariscope. Therefore averaging
of the stress in the direction normal to the plane is
takenintoaccount.
In a photoelastic material the refraction in each
direction is dependent on the normal stress in that
direction. Thestressdifferencebetweentheprincipal
stressdirections, i.e. twicethemaximumshear stress,
2*
max
, leadstoarelativelightvelocitydifferenceand
subsequentlyachangeinpolarisation. Thischangein
retardationcanbemeasuredtoobtaintheshear stress
inapoint.
Themaximumshearstressinapointcanbederived
fromtheelliptical polarisationof thelight. Thelight
changes fromcircular polarised light into elliptical
polarised light when travelling through the sample,
causedbythestressinthematerial.
If at oneor morepointsthecompletestressstateis
known, theabsolutevaluesof thestresstensor canbe
determinedby integratingtheequilibriumequations.
Unfortunate, thisschemeisvery sensitivefor experi-
mental noise, whichhamperstheresults. Inthepresent
teststhevertical stressiscalculatedfromthemeasured
surchargeat thetopof thesample.
3 TEST SETUPANDTEST PROCEDURE
3.1 Test setup
Thetestsetupconsistsof amodel container withinner
dimensions of heightwidth of 690mm530mm
andadepthof 70mm. InFigure1afront viewof the
testsetupisgiven.Thefrontandrearsidesaremadeof
glass. Therectangular pilesaremadeof stainlesssteel
withacrosssectional areaof 1010mm. Tworows,
with a c.t.c. distanceof about 25mm, of threepiles
are installed at respectively 0.5D
tunnel
, 1D
tunnel
and
1.5D
tunnel
, simulatingapiledfoundation. Also atun-
nel isembeddedat345mmfromthesideand232mm
fromthebottomof thecontainer.
The tunnel consists of an upper and a lower part
milledof stainlesssteel whichcanbepushedoutverti-
callybyaninner stepper motor.Therefore, thevolume
changeof thetunnel is primarely caused by vertical
contraction. Themechanical realisationof thistunnel
622
Figure2. Mechanical realisationof thetunnel.
can be seen in Figure 2. An additional surcharge is
appliedonasteel beamtoincreasethestresslevel in
thesample. Thissurchargeisappliedby springload-
ingthebeam, i.e. between afixedboundary andthe
beamamechancial springis applied, suchthat small
displacementsinthesampledonot totallyunloadthe
boundary. Betweenthespringandthefixedboundary
aloadcell ismonitoringtheappliedload.
Thebox is filled with crushed glass particles and
theindexmatchingliquid.Theparticlesizeof theglass
particlesisbetween2and3mm.Thematerial behaves
similar to sand of the same particle size as shown
intriaxial testsperformedbyAllersma. Accordingto
Allersma, theangleof internal frictionatconstantvol-
ume,
cv
, for angular silicasandof thisparticlesizeis
ca. 33
andfor crushedglassca. 39
.
3.2 Test procedure
Mediumdenseconditionsareobtainedbyfirstpouring
theglassintothestrongbox, followedbydensification
of thesample, andfinallypumpingtheliquidthrough
thesamplefrombottomto top. Thetunnel is undis-
turbedlyembeddedduringpreparation, whilekeeping
thediameter at its maximumposition. Thepiles are
pushed in afterwards before applying the surcharge.
A surchargeof 75kPaisappliedatthesteel beamafter
preparationof thesampleisfinished. Beforestarting
the test first an equilibriumstage has to be reached
inwhichnoadditional creepis apparent. Duringthis
phasethesurchargeiskept constant at 75kPa.
Theobjectiveof thistest istoinvestigatetheeffect
of volume loss on existing pile foundations, there-
fore at the beginning of the test the tunnel is in its
openedposition.Alsothepilesarealreadyembedded.
Subsequentlyaphotoelasticmeasurement ismadeby
meansof amechanical polariscope. Inthenextphase,
Figure3. Principal stress direction and magnitudebefore
tunnel contraction.
withoutalteringthesurcharge, thetunnel iscontracted
bydecreasingthedistancebetweentheupperandlower
tunnel partsby2mmor1.2%.Thisresultsinavolume
loss of only 0.6%, as the tunnel contracts vertically
onlyandnot uniformly. Thiswill result inanunload-
ingof thesoil aroundthetunnel. Againaphotoelastic
measurement ismade.
Themechanical polariscopeismountedonacom-
puter controlledx-yscanner andscansapproximately
1200stresspointsspacedat10mminterval inanarea
of ca. 340mm340mmcoveringhalf of thetunnel
and thelower part of theembedded piles. Themea-
surement zoneis also showninFig. 1(dottedlines).
4 TEST RESULTS
Theresults of theperformedtests aresummarizedin
three figures. The first Figure, Fig. 3, is showing
the principal stress directions and the magnitude of
the principal stresses measured before tunnel con-
traction, the second Figure, Fig. 4, is showing the
samedata, but nowfor thesituationafter tunnel con-
traction. The principal stress crosses are all plotted
with the same scale, i.e. the length determines the
magnitude. Finally, the last figure, Fig. 5, plots the
contour plot of thedifferenceinthemeasuredmaxi-
mumshear stress. Thebeforemeasurementistaken
as reference, resultinginapositivevaluefor thedif-
ference when the maximumshear stress in a point
before contraction was higher than after contraction
and logically a negative value when the value after-
wards is higher thanbefore. Themeasuredsurcharge
623
Figure 4. Principal stress direction and magnitude after
tunnel contraction.
Figure5. Maximumshear stressdifference(kPa) before-
after.
after thefirstmeasurementis74.8kPa, whileafter the
secondmeasurement thereadingfor thesurchargeis
74.3kPa. Therefore, the changes found between the
two measurements are mainly caused by the tunnel
contraction and not by additional creep in the sam-
ple. As already elaborated, theangleof theprincipal
stress andthemaximumshear stress aredirect mea-
surements, while the principal stress magnitude is
processed fromthe stress integration algorithm. All
geometries, likepilesandtunnelsshowninthefigures
areplottedtoscaleat theproper position.
4.1 Before tunnel contraction
Whenstudyingtheresultsof themeasurementstaken
before tunnel extraction, given in Figure 3 several
thingsarenoticed, for senseof scalethemeasurment
pointsarehorizontallyandverticallyspacedat10mm.
Firstly theprincipal stress rotations right on top and
belowthetunnel areall negligible,thetunnel isloading
thesoil incompression, whiletothesidelargestress
rotations occur. The behaviour is therefore ver-tical
symmetrical.
Anotherobservationisthatthespacingbetweenthe
piles is such that arching seemto occur, spots with-
out any rotation and with a lower stress magnitude
can be observed. Clearly, a spacing of about seven
timesthepilediameter isstill sufficient tocausethis
effect, howeverscaleeffectscannotbeexcludedasthe
grainsarerelativelylargecompared, d
grains
=23mm,
to thepilediameter which is 10mm. Themost right
pairof piles(remembertworowsof pilesareintstalled
behindeachotherwhenseenfromthefrontside),orthe
pilerownext totheboundary of themodel container
is distributing its load immediately in the side wall.
As directly below a large undisturbed zone is seen.
Thisisclearlyaboundaryeffect. Lastlythepileclos-
esttothetunnel isinfluencedbythetunnel, asbetween
thepileandthetunnel adirect pathof parallel stress
directionscanbeseen. Thispilerowisat atiptunnel
distanceof about 12cmor 12D
pile
or 0.75D
tunnel
.
4.2 After tunnel contraction
Againtheprincipal stressrotationandmagnitudeare
plotted in Figure 4, this time for the situation after
tunnel contraction.All former observationsstill apply,
pilearchingseemstostill occuring,theclosestpilerow
is still influenced by the tunnel and is unexpectedly
inahigher stress regime. However, closer inspection
revealsthat thebandof stressbetweentheclosest pile
rowandthetunnel seemssmallerinwidthi.e.thestress
intensityislower. Also, thestressesdirectlybelowthe
tunnel seemchangedalso, theundisturbedzonebelow
is smaller. Froman experimental point of view this
canbeexpected, becauseasthetunnel isdecreasedin
height, thelargestchangesareexpectedinthevertical
stressdistribution. Not onlyunloadingonthetopside
of thetunnel occursbutalsoanelasticreboundonthe
bottom.
4.3 Shear stress difference
It is rather difficult so observe differences in mag-
nitude in principal stress from the already shown
figures, thereforeoneadditional figureispresented. In
Figure5thedifferenceinmeasuredmaximumshear
stress between the situation before and after tunnel
contraction is plotted as thedifferenceis moreclear
fromthisdatathanfromtheprocesseddata. Thenum-
bers plotted next to the x-axis and y-axis are the
distancefromtheoriginof themeasurement inmm.
The values given in the colourbar of the contour-
plot areinkPa, positivevaluesindicatealower stress
thanbeforetunnel contraction, negativevalues point
towardsahigher stressthanbeforecontraction.
624
It is immediately clear from this data that an
increaseinstressaroundthepiletipsof thethreepile
rowsisfound.Anotherobservationisthatthefirsttwo
pile rows are most effected by the contraction. The
zones with the highest stress concentrations, in this
casearoundthepiletipsandat thepoint at whichthe
stressarchreachesthetunnel aretakingall thestress
change, whilethegeneral fielddoeshardlychangeat
all. Thelast findingfurther supports theobservation
that creep in thesampleduring thetests is kept at a
minimum. Theincreaseinmaximumshear stresscan
beexplained by thefact that in themodel setup due
to volumeloss thesoil immediately begins to create
movements of soil below the pile rows resulting in
negativefrictiononthepiles.
5 DISCUSSION
It isour opinionthat for thecaseconsidered, ascaled
planestrainsituationwithrather largegrainsizeand
low stress situation, the results are still convincing.
Theeffects of nearby tunnellingareseeninthemea-
suredstress field. Increaseof maximumshear stress
canbeexplainedbynegativefrictiononthepiles.How-
ever, the change of the horizontal or vertical stress
is more difficult to acquire, both due to difficulties
fromnoiseinexperimental data, as measuringprop-
ertiesof light isalwaysverysensitivetoexperimental
errors,andduetotheintegrationof themeasuredstress
field. Therefore, noconclusiveanswer canbegivenat
thisstage.
All pilerowsinthepresentedsetupsuffer fromthe
tunnel contractioneffects, but thefirst rowis clearly
most influenced, not onlybythetunnelingbut alsoin
thereferencesituationinteractionisseenbetweenthe
twostructural elements.Theinfluenceof thetunneling
onthelastpilerowisminor, butthispilerowisalready
locatedat 1.5D
tunnel
tothecenter lineof thetunnel.
Whencomparedtotheresultsof Standing&Leung
(2005) a bit more information is derived fromthe
measurements. In thecurrent setup themodel scale,
especially when considering the dimensions of the
pile,themodel containerandtunnel,aremorerealistic.
Similar toStanding& Leung(2005) inour measure-
ment nearby piles attract more stress lines, which
reducein intensity after thetunnel contraction. This
indicates astress releasenear thepile, whichis con-
sistent withtheincreaseinshear alongthepileasthe
pilestartstosettle.
6 CONCLUSIONS
Theinfluenceof acontractingtunnel onthestresses
near thetips of displacement piles has been studied
using photoelastic model tests. Fromthesetests, the
changes in principal stresses and shear stresses near
thetunnel andpiles havebeenquantified. Giventhe
limitedsizeof thescalemodel, andtherelativelylarge
grainsusedtoimprovethevisibilityof thephotoelastic
effect,somelimitationsof themodel havecometolight
andthesehavebeendiscussed. Especially as thereis
someevidenceof stress arches betweenthedifferent
piles, the results of these tests cannot be translated
directly to afieldsituationbut shouldbeinterpreted
withcare.
Nonetheless, the tests show a clear influence of
a volume loss of the tunnel of 0.6%on the stresses
near the pile tips up to 1D
tunnel
. Here stress bands
aredecreasinginwidth, indicatingaloweringof the
principal stresses, and at the same time an increase
of theshear stresses alongthepileshafts is seenfor
these piles. These observation corroborate those of
Standing & Leung (2005), but also indicatethat for
displacement pilesclosetofailuretheinfluencezone
of atunnel volumelossiswider thansuggestedthere.
Contrarytotheirfindingsnotonlypileswiththeirtoes
inside the influence zone defined by J acobsz et al.
(2003) areaffected, but alsothosejust outsideof that
zone.
Thissuggeststhattheinfluencezonetobetakeninto
accountfor displacementpiles, whichdependonboth
endbearingandskinfriction, might beslightlylarger
thanfor boredpiles, whicharemostlyfoundedonend
bearingalone.Todeterminethiswouldrequireacom-
bination of further field observations, model testing
andnumerical modellingof theproblem.
REFERENCES
Ajovalasit, A., Barone, S., & Petrucci, G. 1998. A review
of automatedmethods for thecollectionandanalysis of
photoelasticdata. Journal of StrainAnalysis 33(2), 7591.
Allersma, H. 1982. Determination of the stress distribu-
tioninassembliesof photoelasticparticles. Experimental
Mechanics 22(9), 336341.
Allersma, H. 1987. Optical analysis of stress and strain
in photoelastic particle assemblies. Ph. D. thesis, Delft
Universityof Technology.
Allersma, H. & Broere, W. 2002. Optical analysis of stress
aroundapenetratingprobeingranular material. InPhysi-
cal Modelling in Geotechnics: Proceedings of the interna-
tional Conference on Physical Modelling in Geotechnics,
pp. 149154.
Broere, W. & van Tol, A. 2006. Modelling the bearing
capacity of displacement piles in sand. Proceedings of
the Institution of Civil Engineers, Geotechnical Engineer-
ing 159(3), 195206.
Drescher, A. 1976. An experimental investigation of flow
rulesfor granular materialsusingopticallysensitiveglass
particles. Gotechnique 26(4), 591601.
Frocht, M. 1941and1948. Photoelasticity volumes 1 and 2.
NewYork: J ohnWiley& Sons.
J acobsz, S., Standing, J., Mair, R., Soga, K., Hagiwara, T., &
Sugiyama, T. 2003. Tunnelling effects on driven
piles. proc. int. conf. on response of buildings to
excavation-induced ground movements. In F. J ardine
625
(Ed.), CIRIA Special Publn 201, pp. 337348. CIRIA,
London.
Kaalberg, F., Teunissen, E., VanTol, A., & Bosch, J. 2005.
Dutch research on the impact of shield tunnelling on
pilefoundations. InProceedings of the 5th international
Construction in Soft Ground, pp. 123131.
Lesniewska, D. & Sklodowski, M. 2005. Photoelasticinves-
tigationof localizationphenomenaingranular materials.
InPowders and Grains, pp. 6972.
Netzel, H. & Kaalberg, F. 2000. Numerical damage risk
assessment studies on masonry structures due to TBM-
tunneling in Amsterdam. In O. Kusakabe, K. Fujita,
andY. Miyazaki (Eds.), Geotechnical Aspects of Under-
ground Construction in Soft Ground. Balkema. preprints
pp. 235244.
Petrucci, G. & Restivo, G. 2007. Automated stress sepa-
rationalongstress trajectories. Experimental Mechanics
published online.
Standing, J. &Leung, W. 2005. Investigatingstressesaround
tunnels and piles using photo-elasticity techniques. In
Proceedings of the 5th international Symposium on
Soft Ground, pp. 171177.
Wakabayashi, T. 1957. Photoelastic methodfor determining
of stressinpowderedmass. InProceedings of the seventh
Japanese National Conference on Applied Mechanics,
pp. 153158.
626
Assessment of tunnel stabilityinlayeredground
P. Caporaletti
Atkins Geotechnics, London, UK (formerly University of Rome La Sapienza, Rome, Italy)
A. Burghignoli
University of Rome La Sapienza, Rome, Italy
G. Scarpelli
Universit Politecnica delle Marche, Ancona, Italy
R.N. Taylor
City University, London, UK
ABSTRACT: Thestability of acircular tunnel inlayeredground, withbothfine-grainedandcoarsegrained
soilsbelowthewater table, isinvestigatedexperimentally andtheoretically. Centrifugetestshavebeencarried
out at CityUniversity, London, investigatingindetail tunnellingeffectsinlayeredground, intermsof bothsoil
displacementsandstrainsandintermsof thekinematicsat failure. Also, analytical upper boundsolutionsfor
layeredgroundwhichcloselyreflect theobservedfailuremechanismof thetunnel havebeenderivedindepen-
dently. Theresultsfromall differentapproacheshavebeencomparedbyemphasisingtheeffectsof thedifferent
hypothesesontheassessment of tunnel stability.
1 INTRODUCTION
The assessment of tunnel stability is an important
issueconsideringthecatastrophic effects inducedby
thetunnel collapse, especiallywhenitconcernsurban
areas. Surfacesurchargeloadingsmay occur inprac-
tice where tunnels are excavated below pre-existing
structuresthroughfine-grainedsoil whichisoverlain
bygranular material: typical conditionsof excavation
inurbanareas. Althoughtheweightsof theoverlying
coarse-grainedmaterialsareeasilytakenintoaccount
whenassessingthetunnel stability, thecontributionof
their strengthandstiffnessontunnel stabilityisoften
ignoredor toodifficult tobequantified.
Theproblemwill beapproachedthroughtheupper
bound plasticity theory together with results of cen-
trifuge tests on model tunnels. The case of a tunnel
excavated within an over consolidated clay deposit
overlainbyasandylayerisconsidered.Testsincludeda
masonrywall settinginthecoarsegrainedupper layer
of soil andthusalsogaveadditional informationonthe
interactionof thetunnel withpre-existingbuildings.
Thisstudybeganfromliteraturereferences, follow-
ingtwodifferentapproaches.Inthefirstcasethesandy
layer has been taken into consideration as surcharge
acting on the top of the clay layer where the tunnel
is excavated. In the second case, the sandy layer is
properly consideredby assigningits appropriateunit
weight, thickness, and strength in order to simulate
thereal initial stressstateof soil aroundthecavityand
toemphasisethedifferent behaviour intermsof both
undrainedanddrainedconditions.
2 STABILITY OF A SHALLOWTUNNEL
Theexcavationof ashallowtunnel isclearly athree-
dimensional problem. Neglecting the effect due to
the volume loss at the front face of excavation, this
problemmight be analyzed under plane conditions.
Figure 1 shows an idealization of shield tunnelling,
whereacircular tunnel of diameter D is constructed
withadepthof cover C. Thetunnel liningisregarded
Figure1. Anidealizationof shieldtunnelling.
627
asrigidandinfront of it thetunnel headingisrepre-
sentedby acylindrical cavity of lengthP inwhicha
uniformfluidpressure
t
acts, andauniformpressure
s
acts onthesoil surface(Figure1a). For relatively
bigvalueof thedistancefromthetunnel liningtothe
tunnel face, theproblemcanbeidealizedunder plane
conditions(Figure1b).
The collapse of tunnel heading is usually a sud-
den event caused, for example, by a sudden loss of
tunnel support pressure. Then, thestabilityof tunnels
infine-grainedsoils canbeevaluatedby referringto
undrainedconditions, whilethestabilityof tunnelsin
coarse-grained soils can be analyzed under drained
conditions.
2.1 Stability of a tunnel in undrained conditions
Thestability of atunnel excavatedinundrainedsoils
maybeassessedusingthestabilityratio, N:
where is thesoil unit weight; H is thetunnel axis
depthfromthegroundsurface;
s
and
t
arerespec-
tively the surface surcharge pressure and the tunnel
supportpressure(accordingtoFigure1); andS
u
isthe
soil undrainedshear strengthat thetunnel axislevel.
Davisetal. (1980) estimatedthesafetyof ashallow
tunnel excavatedwithout internal support by consid-
ering three different shapes of shallow underground
opening, for which upper and lower bound stabil-
ity solutions were derived. The soil strength S
u
was
assumed to be constant with depth. Sloan & Assadi
(1992) presentedarigorousstudytakingintoaccount
for the variation of the undrained shear strength
with depth. The theoretical approach of upper and
lowerboundsolutionswerecomparedtoexperimental
resultsfromacomprehensivestudyinclaysconducted
at CambridgeUniversity over thelast decades e.g.
Kimura & Mair (1981) who gave a range of design
linestoestimatethestabilityratiofor different tunnel
geometries, mainlyfocusingonundrainedconditions.
Generally speaking, the bounding solutions gave
goodestimatesof thecollapseloadandsupportedthe
use of classical limit analyses for undrained condi-
tions. It should benoted that thestability of aplane
strainunlinedcirculartunnel (L
R
/D) ismorecrit-
ical thanthatforthelinedcirculartunnel headingwith
L
R
/D=0andtheanalysisisthereforeconservative.
Previousliteratureproposeddifferent chartswhere
the tunnel pressure at collapse is related to soil
mechanical propertiesandtunnel geometry: theterm
(
s
t
)/S
u
orthestabilityratioNisgivenasafunction
of C/D for different ratios of D/S
u
-4. If S
u
is con-
stantwithdepth, thenfor valuesof C/Dgreater than3
theupper andlower bounds of N do not changesig-
nificantlywithD/S
u
. BelowC/Dequal to3, thereis
alarger spreadbutthelower boundfor D/S
u
=0can
beadoptedasasafecriteriontocalculatetherequired
tunnel support pressure: seeFigure4, by referringto
the upper and lower bound curves after Davis et al.
(1980). If theratioD/S
u
issufficientlylargethenthe
collapsewill takeplacefor anyvalueof uniformtun-
nel pressure. Sloan & Assadi (1992) also concluded
that for tunnels with C/D>3, thesesolutions arenot
fullyreliablesincethestabilityof deeptunnelsisusu-
allyrelatedtoacomplicatedlocal collapse, involving
bothelastic andplastic deformation, withonly small
settlementstakingplaceat thegroundsurface.
2.2 Stability of a tunnel in drained conditions
Atkinson and Potts (1977) derived kinematic upper
bound, andstatically admissiblelower boundplastic-
ity solutions for the two-dimensional idealization in
Figure1. Only dry sandwas considered, but thethe-
oretical solutions, based on effective stress, may be
appliedtosaturatedsandsprovidedthattheporewater
is stationary andtheporewater pressures aroundthe
tunnel areknown.Thecollapseof atunnel insaturated
sandisthensimplythesumof thepressurespredicted
by either theupper or lower boundsolutions andthe
porewater pressure. Thesituationisrather morecom-
plicatedif thereisasteadystate, or transient, seepage.
Theauthorsgaveequationstoobtainthedimensionless
ratio
t
/Dasfunctionof C/D: theupper boundsolu-
tion(beinginherently unsafe) gives lower valuethan
thelower boundsolutionwhichisinherentlysafe.The
authors concluded that the support pressure is inde-
pendent of the ratio C/D. They also obtained good
comparisons between thesetheoretical solutions and
experimental datafromcentrifugetests.
2.3 Stability of a tunnel in layered ground
Aspreviouslydiscussed, asfarastheauthorsareaware
there is currently no standard procedure to account
for thecontributiontothetunnel stabilityof overlying
layersof different materials.
Grant &Taylor (2000) studiedthestability of tun-
nelsexcavatedinclaywithoverlyinglayersof coarse
grainedsandsandgravels, referringtothedesignline
for clay only given by Kimura & Mair (1981) for
L
R
/D. Experimental data fromcentrifuge tests
withanupper layer of loosesandfittedthedesignline
for clay only, indicating that it may act as surcharge
loading and not contribute to the stability of tunnel
except in terms of weight. In contrast, the presence
of asignificant thick layer (at least 1D) of relatively
dense coarse grained material, combined with little
cover above the tunnel crown (C-2D), will aid the
628
Figure2. Pictureof theexperimental model: wall-ground
interactionconfiguration.
stability of a tunnel, and the reasonable value of N
equal to4might beassumed.
3 CENTRIFUGE MODEL TESTS
Centrifugemodel testinghasprovedaveryuseful tool
toexaminethebehaviourof shallowtunnelsinlayered
ground,intermsof soil displacementsandstrains,such
asintermsof kinematicfailure.Anumberof testshave
beenperformedat CityUniversity, bymodellingboth
thegreenfieldconditionsandtheinteractionproblem
betweentunnelsandpre-existingmasonrystructures.
Details of theapparatus andprocedures of tests have
been given elsewhere (Caporaletti 2005; Caporaletti
et al. 2006) and only the essentials features will be
describedinthispaper.
All centrifuge tests were performed under plane
strain conditions: Figure2 shows aschematic of the
centrifugemodel. A layer of pre-consolidatedkaolin
is overlain by a layer of medium dense sand. The
kaolinslurrywaspre-consolidatedbyapplyingaone-
dimensional loadinapresstoavertical effectivestress
of 500kPa, beforeallowingtheclay to swell back to
250kPa. Theoverconsolidationratiorangedbetween
1.4to 2.8with depth. Thestratumof mediumdense
sandwasmadebymanual pluviationandthenfrozen.
A 50mmcavitywascutthroughtheclaylayer usinga
thinwalled-cutter andlinedwithathinrubber mem-
braneof negligiblestiffnessandstrength.All testswere
carriedoutatthesamescalefactorN
g
=160inorderto
model areal tunnel of diameterequal to8m, excavated
inaclay layer of 22mdepthoverlainby asandlayer
of 12.8mdepth, with the tunnel axis at about 23m
fromthegroundsurface. Theeffectsof tunnellingon
pre-existing structures were studied by modelling a
Figure3. Frontviewof themodel duringthetestfromimage
processing: a) before reduction of
t
;
t0
=386kPa; b) at
tunnel collapsewith
t
=25kPa.
completelyburiedthinmasonrywall perpendicular to
thetunnel axis, withfoundationsjust at thesand-clay
interface. It represents astiffer andheavier inclusion
inthesandlayer.
Duringthecentrifugespin-upthetunnel airpressure
t
appliedauniformradial total stressintothecavity
tobalancethetotal stressat tunnel axislevel:
where
S
istheunitweightof sand;
C
istheunitweight
of clay; C
s
isthethicknessof thesandlayer; C
c
isthe
cover abovethetunnel intheclay layer; andR is the
radiusof tunnel asshowninFigure2.Thistunnel pres-
surerepresentsthecompressedair, bentoniteslurryor
ashieldusedinpracticetoachievethetunnel stability
duringtheexcavationprocess. Water wassuppliedto
themodel through astand-pipeto maintain aprede-
terminedwater-tablethroughout themodel: thewater
level was set up at different depths fromtheground
surfaceinorder toevaluatetheinfluenceof thediffer-
ent effectivestressdistributiononthetunnel stability.
Equilibriumporepressurewasmeasuredbyminiature
porepressuretransducersaroundthecavity. Thetun-
nel excavationhasbeenperformedbyreducingtheair
tunnel pressureuntil thecavitycollapse, inaperiodof
34minutesat arateof approximately85kPa/min.
Digital imagesweretakeneverysecondbyusinga
videocamerafixedontheswingingplatformtoview
thefront faceof thestrongboxduringtestsandtofol-
low the movements of black targets pushed both in
clayandinsand. Figure3showsimagesof themodel
inflight.
In spite of the relatively high value of the kaolin
permeability, the stratum of clay performs general
629
Figure 4. Stability ratio, N: experimental data against
theoretical solutionsfor claysunder planeconditions.
undrained conditions during the reduction of tunnel
pressure. However, considering that the upper layer
of sandrepresentsasignificant boundaryof drainage
for thelower kaolin layer, andconsideringthesmall
thickness of the clay cover above the tunnel cavity,
local drainedconditionsinthisupperpartof thekaolin
layer werelikelyinthosetests.
3.1 Centrifuge test results
Thevalueof thetunnel pressure,
t
, hasbeenevaluated
for eachtest, inorder tocompareexperimental results
totheguidelinesfromliteraturereferences.
Thecavitywasexcavatedinclay,andsodesignlines
forhomogenousundrainedsoilshavebeenconsidered
first. Figure4andFigure5present thecomparisonin
termsof thestabilityratio, N, andof
t
/S
u
depending
onthesoil mechanical propertyandtunnel geometry.
Theprofileof theundrainedshear strengthwithdepth
hasbeencalculatedfollowingtheequation(Koutsoftas
& Ladd1985):
were
v
is the vertical effective stress, and OCR is
theoverconsolidationratio. Consideringtherelatively
small thickness of theclay cover, aconstant valueof
S
u
has been chosen in order to compare test results
with theoretical solutions by Davis et al. (1980) for
the corresponding value of the term
c
D/S
u
=2. In
the interpretations analyzed afterwards, the constant
valueof S
u
referstoacharacteristicdepthrespectively
equal to 5mand1.8mfromthesand/layer interface
(Ribacchi et al. 1993).
Experimental data represented by open squares
havebeenanalyzedbysimplyassumingthesandlayer
assurchargeactingonthetopof theclaylayer,
s
, and
it is equal to its own weight. Then, thetunnel cover,
C, is only equal to C
c
. Onthecontrary, full symbols
Figure 5. Tunnel pressure,
t
: experimental data against
theoretical solutionsfor claysunder planeconditions.
represent theexperimental results analyzed by prop-
erly includingthesandlayer intheevaluationof the
tunnel cover. Thesandlayer isallowedtomakeacon-
tributionto thetunnel stability as asoil withsimilar
strength to theclay. Thesurfacesurchargepressure,
s
, is therefore equal to zero, and the tunnel cover
C=C
s
+C
c
. Thecomparisonbetweenthetwodiffer-
ent data interpretations emphasizes that the ratio of
thecover tothediameter of thetunnel isclearlyasig-
nificant aspect on the assessment of the pressure at
collapse. A goodfittingof experimental datatoprevi-
ousplasticitysolutionsfromliteraturereferenceswas
obtainedonlywhenthetunnel coverisproperlyconsid-
ered(full symbols).While,theopensquaresarealways
out of therangegivenby theoretical lower andupper
bounds. However, this agreement definitely worsens
if thereal valueof S
u
attunnel axisisassumedinstead
of thevalueat thecharacteristic depth as previously
explained.
Afterwards, experimental datahavebeencompared
totheoretical solutionsfor cavity excavatedinsands.
DesignlinesafterAtkinsonandPotts(1977)areshown
inFigure6, bydistinguishtheoriginal lowerandupper
boundapproachesfordrysands, andthelinesobtained
for saturated soils. Once again, the physical results
areclearly out of thetworanges givenby theoretical
solutions.
Theloadfactor, whichistheratiobetweenthesta-
bility ratioat workingconditionsandat collapse(the
subscript
0
refers to thebeginning of thetest before
thereductionof thetunnel support pressure):
has beencalculatedat different values of thevolume
loss, V
L
, definedas thevolumeof settlement trough
at thetunnel axis, andplottedinFigure7.
In order to have a clear representation of exper-
imental data, only results fromfew tests have been
shown. Test PC2andtest PC8hadgreenfieldcondi-
tionsbut withtwodifferent hydraulicboundaries: the
630
Figure 6. Tunnel pressure,
t
: experimental data against
theoretical solutionsfor sandsunder planeconditions.
Figure7. Curvesof loadfactorsagainst volumeloss.
water level isveryclosetothegroundsurface(PC2),
andvery closetothesand/clay interface(PC8). Soil-
structureinteraction conditions existed inTests PC6
and PC7, in which were modelled respectively the
weakest and the strongest masonry structure, with
thesamehydraulic conditions fixedduringtest PC8.
As expected, focusingat first onthegreenfieldtests,
highervaluesof thetunnel supportpressureweremea-
suredduringtest PC2, whichmeans amoreunstable
configurationduetothedifferenteffectivestressdistri-
butionexperiencedastheconsequenceof differences
inthehydraulicboundaryconditions.Atthesameway,
analyzingtheresultsof soil-structureinteractiontests,
themasonry wall built in thesand layer represented
astiff inclusion within thesoil and madethetunnel
less stable. The tunnel collapse occurred for higher
values of the cavity pressure, and the worst condi-
tionwasreachedwhenthestrongwall wasmodelled.
Lookingat thepoint of intersectionbetweentheload
factor curves related to thesoil-structureinteraction
tests, andthereferencegreenfieldcurve, it isevident
that the weakest wall always stayed in contact with
theupper faceof kaolin, whichdidnot occur withthe
strongwall thatlostthiscontactatvalueof V
L
15%.
Inspiteof this, thefailuremechanismobservedinthe
soil aroundthetunnel seemedtobeverysimilar. It is
particularly interestingto emphasisethat for all tests
theexperimental curves approximately show ahori-
zontal asymptoteafter volumelossesof around20%.
Therefore, thecondition correspondingto thetunnel
collapsehasalwaysbeentakenasthat corresponding
tothestagewhenV
L
>20%.
3.2 Tunnel mechanism of failure
Experimental datahavebeenanalyzedintermsof the
effect duetothelayeredconfigurationonthefieldof
soil stress and strain (Caporaletti, 2005; Caporaletti
etal., 2006). Thetwofine-grainedandcoarse-grained
materials, sandandclay, haveclearlyshowndifferent
mechanical behaviours. In particular, the sand ten-
dency to dilate was constrained by the lateral walls
of the strongbox, and resulted in vertical settlement
decreasingwithdepth.Atthebaseof thesandlayer,the
pattern of soil strains indicates expansion near the
tunnel axis. In contrast, vertical settlements within
theclay layer always increasedwithdepth, andnon-
zero volumetric compressive strains were measured
abovethecavityclosetothetunnel axis, maybedueto
therenot beingperfectlyundrainedconditionslocally
maintained during the tests. According to literature
references, the mechanism of failure for tunnel in
clayspropagatesupwardsandoutwardsfromthecav-
ityinvertbecomingsignificantlywiderthanthetunnel
diameter. In sands, failuretends to involveanarrow
chimney, propagating almost vertically from the
cavityuptothegroundsurface(Mair&Taylor(1997)).
However, in thesecentrifugetests themechanismat
failureforalayeredconfigurationinvolvesawidearea
of soil bothinsandandinclay, withpseudo-vertical
settlements at thesand-clay interface. Thekinematic
mechanismis clearly characterized by soil displace-
ments pointing towards the cavity on the top of the
sandlayer andeverywherewithintheclay layer, and
by arigidvertical block translationat thebaseof the
sandlayer, for athicknessof aboutD/2(seeFigure8).
Thefieldof grounddisplacements insandis strictly
relatedtosoil movementsinducedatthetopof theclay
layer morethantothereductionof thetunnel support
pressureitself.
4 THEORETICAL ANALYSIS
Thecomparisonof experimental datatopublishedlit-
eratureguidelineshasclearly showndiscrepanciesin
terms of evaluationof thesoil behaviour at collapse,
whenbothfineandcoarsegrainedmaterial strataare
involvedintheproblem. Theeffect duetothelayered
631
Figure 8. Mechanism of failure from centrifuge tests
(VL
=20%).
Figure9. Mechanismfor layeredground.
configurationneedtobebetter understood: kinematic
analysescouldexplaintheobservedbehaviour.
A mechanismrelated to the problemof stability
of aplanestrainunlinedcircular tunnel excavatedin
layered ground has been evaluated. The calculation
reliesupontheassumptionthat failurehappensunder
drainedconditionsinsandandundrainedconditionsin
clay.Thesoil isidealizedasaperfectlyplasticmaterial
withanassociatedflowrule: unit weight,
s
for sand,
and
c
forclay,undrainedshearstrengthof clay,S
u
,and
angleof dilancyof sandequal tothemaximumangleof
shear resistance, +=
. Themechanismisshownin
Figure9. ItstartsfromthatgivenbyDavisetal. (1980)
for clays, but it assigns the appropriate unit weight,
thickness, andstrengthtothecoarse-grainedstratum.
Duetotheassumptionof associatedflow, themecha-
nismshowsvertical movementswithinthesandlayer,
accordingtotheexperimental data(seeFigure8).Ver-
tical movements arealso assumedto begeneratedin
clay above the cavity, by simplifying the real kine-
matic observed fromcentrifugetests (seeFigure8).
The work calculation is performed by changing the
angle in order to minimizethevalueof thetunnel
pressureandtoachieveanupper boundsolution, with
anunsafeestimationof thepressureat collapse.
Figure10. Theoretical solutionsandexperimental data.
Thetheoretical solutionfor thismechanismisplot-
ted in Figure10together with centrifugeresults and
literaturesolutions for homogenous undrained soils:
the stability ratio, N, refers to the actual undrained
strengthof clay at thetunnel axis. Thecover insand
has been assumed equal to D/2, following the kine-
matic mechanism observed during centrifuge tests,
whilepartof thethicknessof thesandlayerisevaluated
asanexternal surcharge. Thestabilizingcontribution
due to the friction acting in sand is clearly evident,
anditsinfluenceontheevaluationof thestabilityratio
shouldnotbeneglected. Experimental dataareclearly
out of therangegivenbylower andupper boundsfor
clays, whereasasatisfactory upper boundisassessed
by adoptingthetheoretical solutionby followingthe
mechanismherewithproposed.
5 CONCLUSIONS
Physical modellingandtheoretical solutionshavebeen
studiedinthisresearchandcomparedtooneanother,to
highlighttheeffectonthetunnel stabilityduetoalay-
eredconfigurationthatinvolvesbothfine-grainedand
coarse-grained materials, since previous works only
refer tohomogenoussoils.
In the case of tunnels excavated in clay overlain
by sand, the contribution to the stability due to the
frictionactingwithintheupper layer representsasig-
nificant contribution, whichshouldnot beneglected.
Coarse-grainedmaterial cannotsatisfactorybeconsid-
eredasanexternal surchargeequal toitsownweight
only. A significantover-estimateof thetunnel support
pressuretopreventcollapsemightresultif simpleref-
erenceto theliteraturetheoretical solutions obtained
for homogenous clays is made and the sand layer
treatedonlyasasurcharge. Theproposednewmecha-
nismprovidesagoodupperboundtotheexperimental
datathoughsomesimplificationsof theproposedkine-
matic mechanismcould be removed, and though it
632
is recognised that a new lower bound is needed to
completetheanalysis.
REFERENCES
Atkinson, J.H. & Potts, D.M. 1977. Stability of a shallow
circular tunnel incohesionlesssoil. Gotechnique 27(2):
203215.
Caporaletti, P. 2005. Tunnelling in Layered Ground and its
Effects on Pre-existing Masonry Structures. PhDThesis,
Universityof RomeLaSapienza, Italy.
Caporaletti, P., Burghignoli, A. & Taylor, R.N. 2006. Cen-
trifuge Study of Tunnel Movements and their Interac-
tionwithStructures. Proceedings 5th Int. Conference on
Soft Ground. Amsterdam, TheNetherlands: 99105.
Davis,E.H.,Gunn,M.J.,Mair,R.J.&Seneviratne,H.N.1980.
Thestabilityof shallowtunnelsandundergroundopenings
incohesivematerial. Gotechnique 30(4): 397416.
Grant, R.J. & Taylor, R.N. 2000. Stability of tunnelsinclay
withoverlyinglayersof coarsegrainedsoil. Proceedings
GeoEng2000. Melbourne, Australia.
Kimura, T. & Mair, R.J : 1981. Centrifugal Testingof Model
TunnelsinSoftClay. Proceedings 10th International Con-
ference on Soil Mechanics and Foundations Engineering.
Stockholm, Vol. 1: 319322.
Koutsoftas, D.C. & Ladd, C.C. 1985. Designstrengthof an
offshore clay. Journal of the Geotechnical Engineering
Division, ASCE 111(3): 337355.
Mair, R.J. & Taylor, R.N. 1997. Themelecture: Bored tun-
nellingintheurbanenvironment. Proceedings 19th Inter-
national Conference on Soil Mechanics and Foundations
Engineering. Hamburg: 23532384.
Ribacchi, R. 1993. Recenti orientamenti nellaprogettazione
statica delle gallerie. XVIII Convegno Nazionale di
Geotecnica. Proceedings 18th National Conference in
Geotechnics. Rimini, pp. 37.
Sloan, S.W. &Assadi,A. 1992. Stabilityof shallowtunnelsin
soft ground. Predictive soil mechanics, ThomasTelford,
London, 1993: 644662.
633
Reinforcingeffectsof forepolingandfaceboltsintunnelling
K. Date
Kajima Technical Research Institute, Kajima Corporation, Tokyo, Japan
R.J. Mair & K. Soga
Engineering Department, University of Cambridge, Cambridge, UK
ABSTRACT: Grounddeformationinducedbytunnellinginshallowsandygroundcanbereducedbyplacing
somereinforcementssuchasfaceboltsandforepolingboltsfromthetunnel.A seriesof centrifugetestshavebeen
carriedoutinorder toinvestigatethegrounddeformationpatternduringtunnel excavationwithreinforcements.
Threedimensional numerical analysis of theproblemwas also performedusingFLAC3D andthesimulation
resultsshowgoodagreement withthecentrifugedata.
1 INTRODUCTION
Tunnel reinforcementhasbeenappliedtoboredtunnel
excavationinorder tokeepthecuttingfacestable, to
reduceground (sub)surfacesettlements and to avoid
any adverse influence on adjacent structures. Fore-
polingandfacebolts arethetwomost popular tunnel
reinforcements; theformer isoftenusedinEuropean
countries, whereas thelatter is frequently applied in
Asiancountries. However, thespecificationsof using
themarebasedonlocal andempirical designs or on
experience of past constructions in similar ground
conditions.
In order to excavate a larger tunnel under poor
ground conditions safely, it is necessary to establish
anewdesignmethodfor tunnel reinforcements such
as forepolingandfacebolts. Inparticular, inorder to
evaluatetherelativemerit of thesetwotechniquesfor
grounddeformationcontrol,itisimportanttocompare
themusingthesamemodelingtechniques(centrifuge
tests or numerical analysis) under the same ground
conditions.
2 CENTRIFUGETESTS
In this study, the effect of tunnel reinforcements on
ground deformation in shallow tunnels was investi-
gated. Chambon and Cort (1994) performed cen-
trifuge modeling of tunneling in sandy ground and
showed the minimumpressure to support a cutting
facewasindependentof cover diameter ratio(C/D, C:
cover, D: tunnel diameter). Theyalsoshowed, whena
model tunnel wasinstalledwithC/D=4andP=0.1D
(P: unsupportedlength), thefailurelines extendedto
theheight of about 2.5D fromthecrownanddidnot
extendto thegroundsurface. Thecentrifugetests by
LecaandDormieux(1990) showthat thefailurelines
reachedthegroundsurfacewhenC/D=1. Thesetwo
studiesindicatethatgrounddeformationpattern,when
C/Dislessthan1.0, isdifferentfromthatC/Dexceeds
1.0. Hence, it was decided to perform centrifuge
experimentswithatunnel model of C/D=1.0.
2.1 Models
A schematic viewof themodel isshowninFigure1.
Thestrongboxshowsthehalf of theprototypesothat
the ground deformation changes could be observed
through a perspex window that is installed on the
longitudinal sectionof themodel. Thegrounddefor-
mationwasanalyzedusingPIV programdevelopedby
White&Take(1996).
Figure1. Schematicviewof strongboxandmodel tunnel.
635
Table1. Matrixof reinforcement bolts.
Test Type Arrangement* Number
KDC10 Noreinforcement 0
KDC04 Facebolts FB01 14
KDC05 Facebolts FB02 14
KDC06 Forepoling FP01 14
KDC07 Forepoling FP02 28
* EachpatternisdisplayedinFigure2.
Themodel box was filledwithdry LeightonBuz-
zard Fraction E Silicasand with therelativedensity
of 87% (2%), corresponding to a unit weight of
15.9kN/m
3
.
Themodel tunnel, of diameter D=100mm(7.5m
inprototype), issemicircular andthedepthtothetun-
nel crownCwasequal tothediameter D(C/D=1.0).
Theexcavationof thetunnel wassimulatedbydecreas-
ingtheinternal pressureof arubber bagplacedat the
tunnel face. Thebag was covered with an alminium
rigid lining, which was installed at a distance of P
(=0.2D) behindthetunnel heading.Theinternal pres-
sure was reduced from 100kPa to tunnel collapse
pressure. Thecentrifugetestswereperformedat 75g.
A series of fivetests werecarried out as listed in
Table1. Themodel reinforcement bolts, which were
madeof alminium, wereinstalledperpendicularly to
the tunnelling head during the sand-pouring. They
werecoatedwith thesamesandas usedin thetests,
theouter diameter of themwere2.4mm(180mmin
prototype). Theforepolingmodel boltswereattached
to themodel liner withglue. Thenumber, thelength
andthearrangementof thereinforcementswerevaried
asshowninTable1andFigure2.
2.2 Tunnel collapse
The model tunnel without tunnel reinforcement
(KDC10) collapsedatthesupportpressureof 3.1kPa,
which agrees with the past centrifuge results by
Chambon& Cort(1994). Ontheother hand, for the
tunnelswithreinforcements(KDC04-07), tunnelscol-
lapsed at lower pressures (2.22.7kPa). This shows
that not onlyfaceboltsbut alsoforepolingcontributes
to reducing theminimumsupport pressure, but both
techniquesdidnot dramaticallydecreasethepressure
requiredtokeepthefacestability.
The tunnel collapse mechanismon the longitudi-
nal section is shown in Figure 3. When there is no
reinforcement(KDC10), thefrontslippagelinestarted
at the bottomof the tunnel, extended upwards with
a quasi-logarithmic curve, and then reached to the
ground surface vertically. The line behind the tun-
nel wasnearlyvertical butalittleinclinedbackwards.
Thefailuremechanismsusingfaceboltsareshownby
KDC04 and 05 in Figure 3. The distinct difference
fromthenon-reinforcementcasewasfoundatthefront
failureline, whichstartedfromsomepointattheupper
face and then did not extend ahead of the face but
extendedupwardsalmost vertically. Thisshowedthat
facebolts wereeffectivetoimprovethefacestability.
Inother words, this changeinfailuremechanismled
to thereduction of thecollapsevolume. When face-
bolts wereinstalledonly at theupper face(KDC05),
a chimney-like collapse was observed. This may be
becausethetunnel collapsewasmoredominatedbyP,
that is, thestress relaxation at thecrown rather than
theheight of themodel tunnel, H.
Whenforepolingwasintroducedinasparsemanner
(KDC06), thefrontslippagelinewassimilar tothatof
non-reinforcement case(KDC10), but theback slip-
pagelinedidnot developoutwardbut inwardtoward
thetunnelingdirection.Thisbacklinepatternwassim-
ilar to that when a denser pattern of forepoling was
introducedinKDC07. However, thegeometry of the
front line was totally different. In KDC07, the slip-
pagelinedevelopedfromthemiddleof theheading,
andextendeduptothehorizontal linewhereforepoling
boltsareembedded.Thenthelinesuddenlydeveloped
verticallyupwardstothegroundsurface.
When forepoling bolts are densely installed, they
dividethesurroundinggroundintotwozones; (a) the
outsidezoneof invisiblearchconsistingof forepoling
bolts, and (b) the inside zone of the forepoles. The
mechanisms of collapseof thetwo zones haveto be
consideredseparately.Itappearsthatthecollapsedarea
shiftedforwardinthelongitudinal direction.
2.3 Changes in displacement vectors with the
decrease in tunnel support pressure
Thegrounddeformationinsandygroundisquitesmall
evenwhentunnel support pressuredecreasestoabout
half of theinitial pressure. However, onceit starts to
develop, it abruptly increases and then reaches the
collapserather instantaneously. Therefore, inthepast,
it has been difficult to obtain thedisplacement vec-
tors in sandy ground as the tunnel support pressure
decreases.ThePIV analysis, developedbyWhiteetal.
(2003), wasusedtomonitor subtlechangesinground
deformationinsandyground.
Figure4showsthat thechangesinthedistribution
of displacement vectors on the longitudinal section
in KDC10 as thetunnel support pressuredecreases.
Thedisplacementvectorsweredifficulttobedetected
evenbyusingthePIV analysisuntil theinternal stress
was unloaded to around 25kPa. Initially, the region
deformed by the stress release was widespread. As
shown in Figure 4, with decrease in the tunnel sup-
portpressure, thedeformationbecamemorelocalized
around the face rubber bag. The first large ground
movement wasobservedat 5.4kPaandit wasanear-
likeshapeon thelongitudinal section. Theobserved
636
Figure2. Thearrangement patternsof bolts.
shapeissimilartothethreedimensional tunnel failure
mechanismproposedbyLeca& Dormieux(1990).
2.4 Displacement vectors near collapse in
centrifuge tests
Figure5 shows thedisplacement vectors at pressure
closetothetunnel collapse, whichvariesfrom3.5to
5.0kPa.
The distributions are comparable to the failure
shapesshowninFigure3. ItwasobviousfromKDC04
andKDC05thatfaceboltswereeffectivetoreducethe
front areaaffectedby tunnel excavationandalso the
ground movements, in particular, for the horizontal
one. Intheforepolingcases(KDC06andKDC07), it
appearsthat thegroundmovedunder different mech-
anisms at the inside and outside of the half arches
createdby forepolingbolts as showninFigure1and
Figure2. Somecontinuitybetweentheinsideandthe
outsideisobservedinKDC06, buttheyarediscontinu-
ousinKDC07.Hence,itwasconcludedthatforepoling
boltscanbeeffectivetoreducethedisplacement out-
sidetheforepolingarchas longas they areinstalled
denselyenoughtodividethesurroundinggroundinto
twozones.
Figure6 shows thedistribution of horizontal dis-
placement alongthecuttingheadat different internal
pressures for KDC10, KDC04 and KDC07. Results
showthatfacebolts(KDC04) contributedtothereduc-
tionof faceextrusion. Forthecaseof denselyinstalled
forepoling bolts (KDC07), the effect to reduce the
faceextrusionwas not so apparent as that withface-
bolts, but the bulging pattern changed. That is, the
maximumextrusionwasfoundatthelocationbeneath
the crown, although it was found at about 1/4 the
height of the face from the crown in KDC10 and
KDC04. Grounddeformationstartedat2025kPafor
all cases, butthedeviationfromthenon-reinforcement
case(KDC10) tothereinforcementcases(KDC04and
KDC07) becameevident whenthefacepressurewas
1015kPa.
637
Inactual construction, great caremust betakenin
order to avoid fatal ground (sub)surface settlement
aboveatunnel, andhenceit isessential tostudy how
thesettlementtroughdevelopswiththedecreaseinthe
tunnel pressureespecially at locations just abovethe
tunnel crown.
Figure 7 shows the subsurface settlement at just
above the tunnel crown in KDC10, KDC04 and
KDC07. Thesetroughsappearedwhenthefacepres-
surewas 2025kPafor all cases, but thedifferences
amongthembecameapparent whenthefacepressure
was1015kPa. At around5.56.5kPa, themaximum
settlements in KDC04 and KDC07 were about half
Figure3. Tunnel failurepatternsonthelongitudinal section.
Figure4. Displacementvectorchangeswiththedecreasein
tunnel support pressure.
Figure5. Displacement vector distributiononthelongitu-
dinal section.
Figure6. Distributionof faceextrusionat theface.
as large as those in KDC10. Hence, both facebolts
638
Figure7. Subsurfacesettlement just abovecrown.
and forepoling were effective to reduce ground set-
tlements. The shapes of the troughs are sharper in
KDC10andKDC04thaninKDC07. Themaximum
settlement in KDC4 was positioned behind that in
KDC10. The trough in KDC07 was wider than the
troughs observedinKDC04and10, and, as aresult,
thepositionof maximumsettlement shiftedaheadof
themaximumsettlementoccurredinKDC10. Densely
installed forepoling bolts were capable of reducing
theinfluencesfromthestressreleaseat boththeface
and the crown, while facebolts only contributed to
counteract theeffect of thestressreleaseat theface.
3 SIMULATIONMODELINGANDANALYSIS
3.1 Modeling
Inordertosimulatethecentrifugetestresults, 3Danal-
yses wereperformed using FLAC3D. Thegeometry
wasidentical totheinternal sizeof thestrongboxused
inthecentrifugemodelingtests; 400200300hin
mm, as shown in Figure8. For theboundary condi-
tionsatsidewalls, roller conditionswereapplied. The
analysisbasicallyfollowedtheorder of thecentrifuge
Figure8. Geometryof 3Dnumerical analysismodel.
Table2. Soil propertiesfor test simulationwithFLAC3D.
Model YoungsmodulusE c(Pa) (
) (
)
Mohr-Coulomb-1
E=p
* 1400 0.1 40 0, 15
p
=(
v
+
h
+
h
),3
Mohr-Coulomb-2

h
=
v
,(1) 0.1 SeeFigure9
* Mohr-Coulombmodel without strainsoftening/hardening
model
** Mohr-Coulomb model with strain softening/hardening
model
tests; swing-upto75g, andthendecreaseof theinter-
nal pressureinsidetherubber bag. Thetunnel lining
wasassumedtoberigid.
3.2 Soil property
Mohr-Coulomb without strain softening/hardening
model (MC), oneof thestandard and basic models,
wasused. Twodilationangleswereusedasshownin
Table2.
In addition, in order to simulate more accurately
theresults fromthecentrifugetests, Mohr-Coulomb
withstrainsoftening/hardeningmodel (SSH) wasalso
used. The changes in parameters with plastic shear
strain werederived fromthetriaxial tests by Coelho
et al.(2007). As shown in Figure 9, the simulation
of triaxial test results using FLAC3D showed good
agreement with the test performed under the initial
confiningstressof 120kPa.
3.3 Simulation results for the non-reinforcement
case
Figure 10 shows the development of the maximum
faceextrusionswiththedecreaseinthetunnel support
pressure. Thecentrifugeexhibitedtheentirecollapse
at 3.1kPa. ThepredictionwiththeSSHmodel wasin
goodagreement withtheexperimental results. When
the MC models were used, the deviation from the
experimental results starts at around 3040kPa and
639
Figure9. Simulationresultsof triaxial testat
3
=120kPa.
Figure 10. Simulation of the maximumface extrusion in
KDC10with3Dnumerical analysiswiththeMohr-Coulomb
modelswith/without strainsoftening/hardeningmodel.
then abrupt increase in face extrusion was found at
around15kPa.
ThisisbecausetheSSHmodel isabletosimulatethe
plasticbehaviouratandaftersmall strainlevels, corre-
spondingtothefaceextrusionlargerthan1.5mm.Asa
result, theSSHmodel canfollowthegradual increase
infaceextrusionwiththedecreaseintunnel support
pressure.
Figure11shows thesimulationresults of thedis-
tributionof faceextrusioninKDC10whenthetunnel
support pressure was 8.09.0kPa. Both constitutive
models were successful in estimating the extrusion
curvejust beforetunnel collapse, but, inorder tosim-
ulatethedeformationpatternobservedintheexperi-
mentsmoreprecisely, theSSHmodel wasfoundtobe
moreappropriatethanMohr-Coulombmodel.
However, therewas adifferenceinpositionwhere
themaximumextrusionoccurredinFigure11. Inthe
numerical analyses, the peak was found at the mid-
dle and this location is lower than the location of
the peak observed in the centrifuge test. This may
be due to the difference in the shape of the excava-
tionrangesinbetweencentrifugetestsandnumerical
analyses. The model ground in centrifuge tests can
movesmoothly at thecorner near thecrownandthe
facebecausetherubber bagis flexibleenoughto be
smoothly deformed. However, as for the numerical
Figure 11. Simulation of the face extrusion bulges in
KDC10whenthesupport pressurewas8.09.0kPa.
Table3. Input parametersfor facebolts.
Model Density Youngsmodulus Poissonsratio
Pile 2700(kg/m
3
) 7*10
4
(MPa) 0.2
Figure12. Changesinshear stiffnessandyieldingstrength
of soil-bolt interfacewithoverburdenpressure.
analyses, rectangular meshesmightpreventthemodel
groundfromextrudinginwardssmoothly.
Furtherimprovementinthenumerical analysis(e.g.
meshmaking) isrequiredinordertoillustratenotonly
thedevelopment of themaximumextrusionbut also
itsdistributionalongtheface.
3.4 Reinforcing effects of facebolts on reducing
face extrusion
Thereductionof faceextrusionbyfaceboltswasalso
simulated. TheSSHmodel wasadoptedfor bothsim-
ulations. Theinterfaceproperty betweengroundand
boltsurface, whichisshowninTable3andFigure12,
was derived fromthepull-out tests reported in Date
et al. (2007). Figure13. shows thesimulationresults
of KDC04. Results fromKDC10 arealso presented
for comparisonpurpose.
Thedeformationpatternsimulatedbythenumerical
analysiswassimilar tothat of thecentrifugetest, but
themagnitudepredictedbythenumerical analysiswas
640
Figure13. Simulationof faceextrusioninKDC4with3D
numerical analysiswiththeSSHmodel.
smaller thanthecentrifugedata. This may bedueto
themeshproblemmentionedabove, thedifferencein
thesoil-boltinteractionpropertiesandthedifferencein
soil propertiesduringbetweenloadingandexcavating.
Further investigationisneeded.
4 CONCLUSIONS
A series of thecentrifugetests showedthat introduc-
tionof faceboltsandforepolingboltsfor tunnelingin
shallowsandygroundyieldeddifferentshapesof tun-
nel collapseandcontributedtosomereductioninthe
tunnel supportpressuretokeepthecuttingheadstable.
Thesetechniques werealso effectivein reducingthe
vertical settlement at locations just abovethetunnel
crown. Facebolts wereabletoreducetherisk of face
extrusion and hence can make the ground ahead of
thefacestiffer. Forepolingboltscandividetheground
aroundthetunnel faceintotwozones; theinsideand
theoutsideof thearchof forepolingbolts.
Numerical analysis results show that the SSH
model (Mohr-Coulomb model with strain soften-
ing/hardening) gave better match to the centrifuge
datathantheMCmodel (Mohr-Coulombmodel with-
out strainsoftening/hardening). However, inorder to
simulatemorerealisticbehaviour observedinthecen-
trifuge tests, further investigation of the settings of
numerical analysissuchasmesh-makingandsoil-bolt
interactionpropertiesneedstobeconducted.
REFERENCES
sionlesssoil: stabilityof tunnel face. Journal of Geotech-
nical and Geoenvironmental Engineering, ASCE, 120(7):
11501163 1994.
Coelho, P.A.L.F. 2007. Insitudensificationasaliquefaction
resistance measure for bridge foundations. PhD Thesis,
Cambridge University, UK., 2007.
Date, K., Mair, R.J., Soga, K. Centrifugetestsfortunnel rein-
forcement. Soils and Foundations, 2008. (inpreparation)
Leca, E., Dormieux, L. 1990. Upper andlower boundsolu-
frictional material. Gotechnique 40(4), 581606 1990.
White, D., Take, A. Bolton, M.D. 2003. Soil deformation
measurementusingparticleimagevelocimetry(PIV) and
photogrammetry. Gotechnique 53(7), 619631 2003.
641
Mechanical behavior of closelyspacedtunnels laboratory
model testsandFEM analyses
J.H. Du& H.W. Huang
ABSTRACT: Reduced-scalemodel testsandcorrespondingfinite-elementanalyseswereperformedtoinves-
tigatetheeffect of different excavationsequences for twintunnels under theunsymmetrical pressureinweak
rock. Inthesetests, thealternativesequencesof theexcavationof theleft tunnel first andthentheright one, or
theright first andthentheleft, weresimulated. Thedisplacement of groundandtheearthpressureintherock
weremeasured. Particular attention was paid to thedifferent behaviors of thedisplacements and thestresses
causedbythedifferent excavationsequences. Theresult of thepresent studyindicatesthat thestateof stresses,
thedisplacementsaroundthetunnelsandthegroundsurfacesettlementsareall different whentheexcavation
sequenceisalternated. Whentherightoneisexcavatedfirstly, thestratumconditionismoredeteriorated. Itcan
beconcludedthat thesequenceof theleft first andthentheright oneisbetter for suchtwintunnels.
1 INTRODUCTION
Withthedevelopment andtheupgradeof infrastruc-
turessuchashighway,subway,railway,andmanyother
facilities, tunnel constructions aregradually increas-
ingintherecent years. ThepreferenceinChina, and
aswell asinother countries, istousetwintunnelsfor
thenewtransportationlinesrather thanasinglelarger
diameter tunnels, whenthespaceislimited.
Model test and numerical simulation aretwo key
methodstoinvestigatethetunnelingproblems. Inthe
pasttwodecades,therapidadvanceshavebeenmadein
tunnel model tests,manyof whichareinvestigationson
thetwintunnels. Dharetal. (1981)performedthefrac-
turepattern around twin openings in weak materials
of sand wax and sand wax mixtures with the plas-
ter of Parisunder controlledloadingconditions. Twin
openingsfor different orientationsinrespect toload-
ingdirectionswerestudied. Adachi et al. (1993) also
used model tests to analysis theinteraction between
twin tunnels. The tunnel excavation was simulated
by tailor-madediameter reducibledevice. Chu et al.
(2006) performedmodel testsof twincircular tunnels
inhomogenousmaterial, two-layeredformations, and
three-layered formations to understand themechani-
cal behavior of atwin-tunnel of circular crosssection
in multi-layered formations, and a two-dimensional
numerical simulationwasalsodeveloped.
Using2Dand3Dnumerical simulations,Ghaboussi
(1977), Soliman (1993), Addenbrooke (1997), Ng
Figure1. Typical crosssectionof PingnianTunnel.
(2004) et al. investigatedsomeaspects of interaction
mechanisms.
PingnianTunnel isatypical twin-tunnel, whichwas
constructed in accordancewith theprinciples of the
NewAustrianTunnelingMethodinweak rock, being
part of LuofuexpresswayinYunnanProvince. Tunnel
widthis12.5m, andthelengthisabout360m. Spacing
betweentwotunnelschangesfrom18m(at Luochun-
kousite)to25m(atFuningsite).Thetunnelsareunder
theinclinedgroundsurfaceandthetwotunnelsarenot
constructedinthesamelevel, andthevertical distance
between themis from3mto 4m. The typical cross
sectionisshowninfigure1.
643
Figure 2. Sketch of steel frame with model tunnel
(Unit: mm).
So far, the behavior of surrounding ground dur-
ingthesingletunnel excavationhasbeenextensively
investigated, however, thecasethat atunnel isdriven
paralleling to another adjacent tunnel is of interest
totunnel engineers. DuringthePingnianTunnel con-
struction, acollapsehappened at theleft lineportal,
andtheslipalso affectedtheadjacent portal of right
linebecauseof thesmall spacebetweenthetwotun-
nels. Inorder to study theactual interactionbetween
two tunnels, two-dimensional laboratory model tests
wereperformedtocapturetheearthpressuretransfer
andthedisplacementof surroundingrockduringcon-
struction. Theparticular attentionwaspaidtotheevo-
lutionof thedisplacementsandstressescausedbythe
different excavationsequence. Thenumerical simula-
tionsbytheFEMwerealsoconductedfortheproblem
tocomparetheanalytical resultsbymodel tests.
2 SIMULATIONMETHODS
2.1 Two-dimensional plane strain test
Althoughactual tunnelingisathreedimensional prob-
lemand actual ground is heterogeneous as well as
anisotropy in nature, in this study, plane strain tests
were conducted in a homogeneous ground by sim-
plification. The box used in tests was made of a
steel frame, asshowninfigure2, withdimensionsof
1.6m1.2m0.4m. Thefour sidewalls of thetest
Table 1. Properties of the ground material for the model
tests.
Unit weight (KN/m
3
) 26.9
YoungsModulusE (MPa) 38
Cohesionc (KPa) 25
Frictionangle (deg) 39
boxwereassembledusingsteel sheets. For thecaseof
theobservationof thegroundmovementpatternsdur-
ingtesting, twotransparent Perspex plates wereused
at thefront andbacksides.
Thetests wereconductedaccordingto thetypical
cross sectioninthesteel frameas showninfigure2.
Thetunnel width is 156mm, thehorizontal distance
betweentwotunnels is 225mm, thevertical distance
is500mmunder inclinedgroundwithangle18
, and
thecoverdepthsof leftandrighttunnel are212mmand
138mmrespectively.All dimensionsarecontrolledby
thegeometrical similarityratioof 1/80.
Thegroundmaterial usedinthemodel testwascom-
posedof baritepowder, sandandplaster mixedwith
water,andthequalityproportionwasachievedthrough
aseriesof material tests. Themainparametersof the
groundmaterial areshownintable1, determinedby
thefollowingexpressions accordingto thesimilarity
theory.
whereC
, C
, C
, C
E
, C
c
, C
l
arethesimilarity ratios
for friction angle , stress , youngs modulus E,
cohesion c, unit weight and geometrical dimen-
sionl, whichrepresentrelationshipfor theparameters
betweenthemodel test andthereal situation.
The excavation was simulated by removing the
groundinsidetheexcavationzonesmanually, withthe
twodifferentexcavationsequencesconducted: theleft
tunnel excavatedfirst andthentheright one(named
test A), and the right first and then the left (named
test B). A monitoringprogramwas utilizedtorecord
the process of the ground movements and the earth
pressurechangesaroundthetunnels.Monitoringitems
includedearthpressurecells, deeprodsandmicrome-
tergauges, andthelocationsof itemsshowninfigure3
andfigure4.All monitoringitemswereattachedtothe
high-speedinstrumentof staticstraingauge(YE2539),
collectingdataat anyinterval operator set.
644
Figure3. Locationsfor displacement gauges.
Figure4. Locationsfor earthpressuregauges.
2.2 Numerical simulation
Corresponding numerical simulation was performed
to compare the above experimental tests. The sim-
ulations were based on the model test conditions
and thematerial properties. Thefiniteelement soft-
ware Msc.Marc was adopted for the simulations. In
theanalyses, thegroundmaterial was assumedto be
elastic-plastic confirming to the Druker-Prager fail-
ure criterion together with the associated flow rule.
Theanalysisconsistedof twophases: thefirst phrase
istocreateaninitial geostaticconditionandthesecond
phaseto simulatetheexcavationprocess. Duringthe
firstphase, theinitial geostaticconditionwasachieved
by applyinggravity forces toall groundelements. In
the second phase, the excavation process was simu-
latedby removingtheelementsinsidetheexcavation
zone. The simulation of the test A and the test B
performedin laboratory model tests wereconducted
accordingly.
3 RESULTS
3.1 Displacement
Figure 5 is a surface displacement-time diagram,
whichisobtainedinthemodel test. Fromthefigurewe
Figure 5. Surface settlement curves (obtained in model
test).
canfindthatthesettlementof everytestpointincreases
step by step along with theadvancement of excava-
tion.A substantial increaseof thedisplacementoccurs
when excavation section passes thetest section. The
settlementdevelopscontinuouslyafterexcavationfin-
ishedandreachesthesteady-stateinaperiodof time.
Thecreep displacement is themajor part during the
period of theexcavation termination, accounting for
about 10%30%of thetotal displacement.
There is still some ground surface settlement (at
G2, G4) increasing during theexcavation process of
adjacenttunnel,whichrevealstheexistenceof aninter-
actionbetweenthetwintunnels. G1locatedaboveand
lefttoG2, whosehorizontal distancetotheleftwall of
thelefttunnel isonetimeof widthof thetunnel,doesnt
undergoadistinctdisplacementduringtheexcavation
of the right tunnel while it does during the excava-
tionof theleftone. Thedisplacementof thelower G5,
locatedintherighttotheG4, istotallydifferent, which
ismainlyaffectedbytheexcavationof righttunnel.For
thespotG3, locatedatthesurfaceabovethepillarcen-
terline, thesettlement isstill inthedevelopment state
andtheexcavationof boththerightandthelefttunnels
will produceeffectonit; thetwoabruptchangesof the
displacementhappenedintheexcavationsectionpass-
ingthetest one. Thedisplacement changeof thespot
G3appearssmaller thanthedisplacement of thespot
G2andG4.
Wecanmakeaconclusionfromtheabove:theinflu-
enceof tunnel excavationdecreaseswiththeincrease
of thedistancetotheexcavatedtunnel.Thezoneinflu-
encedmostly is thoseabovethetunnels. Under such
leaning stratum, the influence of the excavation to
rocks in one time of the width of the tunnel region
still exists, but it is small whenthedistancebetween
the tunnel and the rock is more than three times of
thewidthof thetunnel. Excavations of left andright
tunnel all haveeffect onsurfaceabovethepillar.
The comparison between deep settlement (settle-
mentof theinnerstratum, indicatedbyD1, D2andD3,
645
Figure 6. Surface and deep ground displacement curves
(obtainedinmodel test).
Figure 7. Surface and deep ground displacement curves
(obtainedinFEM analyses).
locatedabovethelefttunnel crown, inthepillarcenter
lineparallel to theright tunnel springlineandabove
therighttunnel crownrespectively)andcorresponding
surfacesettlement isshowninfigure6. Thedevelop-
ment trendsarecoincident, however, deepsettlement
occurs prior to that of the surface, and its displace-
mentrateisalsorelativelybigger.Timetosteadystate
neededby thedeepdisplacement is shorter thanthat
of thesurfacesettlement. Whenit comes tocompare
twotunnels, thedisplacement rateof theleft tunnel is
bigger thantheright one.
InthetestA, settlementsof all measurepointsare
morethan thosegot in thetest B. Themaximal dis-
placement is35.9mmwhentheleft tunnel excavated
first and then theright one, and it is 44.4mmwhen
the right one first and then the left. So it indicates
that the left tunnel excavated first can control sur-
face displacement more effectively. The comparison
of surfacesettlementsbetweentwooppositeconstruc-
tionsequencesachievedinnumerical simulationsare
Figure 8. Transverse settlement trough corresponding to
different excavationsequences(obtainedinFEM analyses).
shown in figure 7. It reaches the same conclusion
that the ultimate surface settlement occurred by the
righttunnel excavatedfirstisbiggerthanthatoccurred
inthereverseexcavationsequence. Theevolutionlaw
of thedisplacement rateisthesameof thesettlement.
So it is not beneficial for the tunnel stabilization to
adopt theconstructionsequencewiththeright tunnel
excavatedbeforetheleft tunnel under suchlocations
asthePingniantwintunnel undertheinclinedstratum.
The whole surface settlement curves can be
obtained through FEM analyses, shown in figure 8.
Basedonthesecurvesitcanbeconcludedthat,withthe
right tunnel finished, theexcavationof theleft tunnel
(thecurveB infigure8) will generatemoredisplace-
mentthanthatcausedbytheexcavationingreenfield
(thecurveA infigure8). Thedisturbancedueto the
right tunnel construction lowers the stiffness of the
rock aroundtheleft tunnel andwhentheleft is con-
structed, therock inthepreviously disturbedzoneis
disturbed again, and it will movemorethan what is
expected. So, theexistingof theright tunnel isunde-
sirable for the left one. However, when there is the
adjacenttunnel existing,additional groundsurfaceset-
tlementcausedbyrighttunnel excavated(thecurveD
infigure8) islessthanthat causedingreenfield(the
curve C in figure 8). The existing of the left tunnel
whichishigher thantheright oneisanadvantagefor
theright tunnel toacertainextent.
3.2 Earth pressure
Earth pressure near the crown and the invert (R1 is
abovethecrown and R2 is under theinvert, refer to
figure4) of therighttunnel aregiveninfigure9.There
is the same development rule for the earth pressure
of R1 and R2; thepressureincreaseabit dueto the
previous tunnel excavation, and decrease rapidly as
excavationsectionpasses thetest sectionbecauseof
theloadreleasecausedbyremovingrockof thetunnel
space.
646
Figure 9. Earth pressure near crown and invert of right
tunnel (obtainedinmodel test).
Figure10. Earthpressureat crownandinvert of theright
tunnel.
Thepressurechangeat R2 near thecrown is big-
ger than that at R1 near the invert. The phenomena
should beassociated with thelocations of two mea-
suringpoints, theinvert isdeeper thanthecrownand
initial stressnear theinvertisbigger thanthatnear the
crown. Thesameconclusioncanbedrawnbynumer-
ical simulationandtheresultsareshowninfigure10.
The stress changes caused by the excavation of the
right tunnel ingreenfieldaremorethanthat caused
withtheexistingleft tunnel. Andit comestoopposite
conclusion for theleft tunnel. So, depending on the
stresschanges, thesameconclusioncanbedrawn: the
existingof theright tunnel is undesirablefor theleft
one; and the existing left tunnel is beneficial to the
right one.
4 CONCLUSION
The reduced-scale laboratory tests and their cor-
responding numerical analyses were performed to
investigatethetransfer of theearth pressureand the
displacement of thesurroundingrockduringthecon-
struction. At the same time the effect of different
excavationsequenceswasanalyzed.
Thedegreeof theinteractionbetweenthetwotun-
nelsisproportional tothedistancebetweenthem. The
changes of stresses and displacements of the rocks
along the tunnel are mainly affected by its excava-
tionwhiletheeffect causedby theadjacent tunnel is
small. Theinteraction effects on thetwo tunnels are
different: theexistingof theright tunnel deteriorates
theproperties of rocks aroundtheleft tunnel, which
isharmful totheleft tunnel; whereastheexistingleft
tunnel reduces theasymmetric degree, andtherefore
benefitstotheright tunnel.
Stressdecreasesrapidly withtheexcavationof the
tunnel, which causes settlements of the surrounding
rocksandconsequentlyexpandstosurface.
Displacementsof therockincreasegraduallyalong
with theexcavation advancement, andthemain por-
tionoccurswhentheexcavationsectionpassesthetest
section.Aftertheexcavationfinished,acreepdisplace-
menttakesplace, whichaccountsforabout10%30%
of thetotal displacement.
Thedifferenceof theexcavation sequencecauses
thedifferent development ruleof thestress fieldand
the displacement field. The values of displacements
and stresses increase on adopting right tunnel exca-
vatedfirstly. For thetunnelslikethePingniantunnel,
whichisashallowtunnel under unsymmetrical pres-
surewiththevertical distancebetweenthetunnels, it
isprofitabletoexcavatetheleft tunnel firstly.
REFERENCES
Adachi, T., Kimura, M. & Osada, H. 1993. Interac-
tionbetweenmulti-tunnels under construction. Eleventh
Southeast AsianGeotechnical Conference, Singapore, 48
May 1993: 5160.
Addenbrooke, T.I. 1996. Numerical analysis of tunnelling
in stiff clay. Ph. Degree thesis, London: University of
London.
Chu, B.L., Hsu S.C., ChangY.L. et al. 2006. Mechanical
behavior of a twin-tunnel in multi-layered formations.
Tunnelling and Underground Space Technology.
Dhar, B.B., Ratan, S., Sharma, D.K. etal. 1981. Model study
of fracturearoundundergroundexcavations. Proceedings
of the International Symposium on Weak Roc: 267271.
Ghaboussi, J. &Ranken, R. E. 1977. Interactionbetweentwo
parallel tunnels. International Journal for Numerical and
Analytical Methods in Geomechanics. 1:75103.
Ng, C.W.W., Lee, K.M. & Tang, D.K.W. 2004. Three-
dimensional numerical investigations of new Austrian
tunnelling method(NATM)twin tunnel interactions.
Geotechnique. 41:523539.
Soliman, E., Duddeck, H. & Ahrens, H. 1993. Two and
threedimensional analysisof closelyspaceddouble-tube
tunnels. Tunnelling and Underground Space Technology.
8(1):1318.
647
Stabilityanalysisof masonryof anoldtunnel bynumerical modellingand
experimental design
J. Idris&T. Verdel
Laego Ecole des Mines de Nancy, Parc de Saurupt, Nancy Cedex, France
M. Alhieb
Ineris Ecole des Mines de Nancy, Parc de Saurupt, Nancy Cedex, France
ABSTRACT: Thepresentpaper proposestwonumerical modelsof anoldtunnel supportedbymasonry; these
modelsweredevelopedby thewell-knownUniversal Distinct Element Code(UDEC). A masonry mechanical
behaviour analysis and numerical simulation of the masonry ageing phenomenon will also be addressed by
meansof anexperimental designtostudytheinfluenceof masonryblockphysical propertiesonthemechanical
behaviour of tunnel support structure.
1 INTRODUCTION
Masonryvaultsareoneof themostfamiliar structural
shapes present in architectural heritage worldwide.
Historical domedbuildings archedstonebridges and
vaultedtunnelsareamongthemostfamousexamples.
Over thelastfewyears, thedevelopmentof numer-
ical toolsinthefieldof structural analysishasenabled
researchers to establish approaches for numerically
modelling masonry structures, yet an analysis of
the mechanical behaviour of such blocks and joints
structures remains challenging due to the extent of
correlated factors. This paper seeks to study tunnels
masonry behaviour usedinoldtunnels by numerical
modellingandexperiencedesign.
2 MASONRY STRUCTURESNUMERICAL
MODELLINGAPPROACHES
Several modelling approaches to masonry structures
(continuous and discontinuous modelling) are cur-
rentlyunderdevelopmentbyseveral researchteams.A
number of bridgearchmodelshavebeenproposedto
study thebehaviour andcertaininstability problems,
suchascollapse(Fordetal., 2003; Sumonetal., 1995;
Hughes et al., 1998; Bicanic et al., 1995; Brookes
et al., 2004). Many historical masonry construction
simulations and numerical analyses were presented
by (Valluzzi et al., 2004; Loureno, 2001; Giuriani
et al., 2001; Bicanicet al., 2002). Yet nosinglepubli-
cationactuallyfocusesonor addressestheanalysisof
oldtunnel masonrystructures. Threebasicmodelling
strategies for masonry structures can at present be
identified(seefigure1):
1 Detailedmicro-modelling: blocksandmortarinthe
jointsarerepresentedasacontinuum, whereasthe
unit/mortar interfacesaremodelledbydiscontinu-
ouselements;
2 Simplified micro-modelling: geometrically-
expanded continuum units, with discontinuum
elementsincorporatingthebehaviour of bothmor-
tar jointsandinterfaces;
3 Macro-modelling: all three principal features of
structural masonryarerepresentedbyanequivalent
continuum.
Figure1. Basicapproachestomasonrystructures.
649
Figure2. Frenchtunnel classification(Idriset al., 2007).
3 TUNNELSAGEINGPHENOMENA
Old underground structures, especially tunnels, dis-
play specific characteristics regarding behavioural
evolution over time. Theinfrastructureenvironment,
surroundingsoils andconstructionmaterials usedall
contribute to this evolution. Consequently, several
types of disorders may appear and develop in these
underground structures due to specific ageing pro-
cesses that, in reality, arecomplex phenomena. One
impact is the alteration in mechanical properties of
thevariousconstructionmaterials.
A majorityof theworldstunnelsarecurrentlymore
than100years old; thesewouldall beconsideredas
ancient infrastructure. Figure2 illustrates theclassi-
fication of tunnels in France, where the oldest one
exceeds180years.
The majority of old tunnels are supported by
masonry elements. The type of masonry support or
liningdepends uponutilisationof thehighcompres-
sivestrengthinthestones, whichexplainsthevaulted
sectionshapeof oldtunnelssupportedbymasonry.
Apart fromtheenvironment and evolution in sur-
rounding soil and in the absence of an effective
isolationsystemforsuchundergroundstructures, sub-
soil water caneasilypenetratethemasonryjointsand
circulatewithin.
Over timeandinthepresenceof other aggressive
ambientfactors,several physical,chemical andbiolog-
ical processesmaydevelopinsidethemasonrystruc-
ture; thisphenomenonanditsimpactsarecollectively
calledthetunnel-ageingphenomenon.
As a result, various types of disorders appear
insideoldtunnels; thesewouldinclude: longitudinal
or transversestructural cracks, convergenceandpar-
tial masonrycollapse. Figure3providessomeimages
of suchdisorders.
4 NUMERICAL MODELLINGOF ANCIENT
TUNNELS
A tunnel masonrystructureisadiscontinuousmedium
consistingof blocks bondedtoeachother by mortar;
Figure3. Sampledisorder typeswithinoldtunnels(CETU
Tunnel StudyCentre, 2004).
Figure 4. A typical old tunnel cross-section (extracted
fromKerisel, 1975).
in addition, such astructureforms an interfacewith
the surrounding soil. The Distinct Element Method
(DEM) is a suitable technique for modelling these
structures. By means of the Universal Distinct Ele-
ment Code(UDEC), twosimplifiedmicro-modelsof
an ancient tunnel have been derived. The geometric
and geomechanical properties of both models were
inspiredfrompreviouscasesof ancient tunnelsavail-
ableintheoldtunnelsub-database(Idrisetal.2004,
2007) as well as fromother bibliographical sources,
including Kerisel, 1975 (see Figure 7, in which the
thickness of the masonry support structure equals
80cm).
Thetwo representativemodels arepositionedat a
shallowdepthof 20m; theybothdisplayavaultedsec-
tionshape. Inthefirstmodel, themasonry-supporting
section consists of a regular rectangular and square
limestoneblocks(Fig. 5a). Inthesecond, themasonry
consistsof twolayersof limestoneblocks: regular and
irregular (Fig. 5b). Masonry blocks arebonded by a
limemortar. Themasonry support thicknessis80cm
andthesidewall heightineachmodel amountsto3m.
All other geometricdetailsareshownonFigure5.
By taking into account model section symme-
try, only half of each set-up needed to be mod-
elled. Figure 6 shows the tunnel position within the
650
Figure5. Twoproposedmodels(aandb) for oldtunnels.
surrounding soil, along with the selected boundary
conditions, the chosen dummy joints and the mesh.
Dummy joints were introduced in order to obtain a
suitablegridaroundtheexcavatedtunnel.
Thesoil surroundingthetunnel consistsof ahomo-
geneous mix of clay andsand. Table1lists thebasic
mechanical properties assigned to the surrounding
soil,masonryandmasonryjoints,basedontheworkby
(Verdel etal.,1999),(Hoek,2000)and(J anssen,1997).
Calculations were carried out in plane strain: the
soil andmasonryfollowaperfectelasto-plasticMohr-
Coulomb plasticity criterion. The calculation step
proceededbytwomainstages: model consolidationin
theinitial stress conditionprior totunnel excavation;
and tunnel excavation and simultaneous installation
of masonry support. The calculation could then be
continueduntil reachingmodel equilibrium.
5 MECHANICAL BEHAVIOURANALYSIS
OF MASONRY BLOCKS
Figure8presentsMohr circlesfor everygridzoneof
themasonryblockusedintheinitial model (according
to which, every masonry block is divided into 2 or
Figure6. Tunnel position, boundaryconditionsandmesh.
4 grid zones). In the initial model, no block stress
exceedstheMohr-Coulombfailurecriterion(asshown
onFigure7), andall masonryblocksdisplayperfectly
elastic behaviour. Based on thestress distribution of
masonryblocks, weareabletoapproximatethecriti-
cal valuesformasonryblockcohesion, tensilestrength
andfrictionangle,all of whichcanmodifythemasonry
behaviour fromelastictoplastic.
This analytical approach has enabled us to define
the variation range for these three parameters over
651
Table 1. Selected mechanical properties of the surround-
ingsoil, masonry andmasonry joints(M: Volumic mass; E:
Youngs modulus; : Poissons ratio; C: Cohesion; : Fric-
tionangle; Tr: TensilestrengthJ kn, J ks: Normal, Tangential
joint stiffness; J C: J oint cohesion; J : J oint friction angle;
J Tr: J oint tensilestrength).
Surroundingsoil Masonry
Parameter Unit Value Parameter Unit Value
M Kg/m
3
1900 M Kg/m
3
2100
E MPa 200 E MPa 6000
0.3 0.2
C MPa 0.50 C MPa 3

20

30
Tr MPa 0.10 Tr MPa 1
Masonryjoints
Parameter Unit Value Parameter Unit Value
J kn GPa/m 150 J

25
J ks GPa/m 69.7 J Tr MPa 0.4
J C MPa 1.2
Figure7. Mohrcirclesformasonryblockgridzonestresses
intheinitial model.
which it is appropriate to study how the three cho-
sen parameters influence the overall behaviour of
masonry structures. As an example, the decrease in
bothcohesionandfrictionanglecanservetosignifi-
cantlyincreasethenumber of plasticzoneswithinthe
masonrystructure.
6 PROPOSEDEXPERIMENTAL DESIGN
FOR MASONRY SUPPORT AGEING
SIMULATION
Constructionmaterialsaresubmittedtovariousmod-
ifications and degradations under several physical,
Figure8. A sampleof numerical simulationresultsfor the
first model (test, cohesion, tensionandfriction).
chemical andbiological combinedprocesses. A num-
ber of physical phenomena, such as dilation, con-
traction, freezing, hydration and desiccation cycles,
potentiallyleadtoaconsiderableevolutioninmaterial
mechanical properties. Constructionmaterialsageing
processesdetail wereexplainedby(Idriset al. 2007).
Thebehaviourof masonryisdictatedbythephysical
andmechanical properties of bothblocks andjoints.
Our survey focused on evolution in the mechanical
behaviour of masonry blocks due to ageing; special
attentionwasthereforedevotedtotheparametersthat
defineblockfailure, suchascohesion, tensilestrength
andfrictionangle.
Toevaluatetheinfluenceof eachchosenfactor on
masonry structure behaviour, it proved necessary to
observesignificant changesinmodel behaviour once
factor values hadbeenchanged. Theresponsefactor
selected herein is the number of blocks that change
statusfromelastictoplasticbehaviour.
For this purpose, acompletefactorial design was
proposed; thisthree-level designiswrittenasK
n
fac-
torial design(withK =3: thestudiedfactor number,
n: level number). This nomenclature means that 3
factors are considered, each one at 3 distinct levels
(Barrentine, 1999).
Consequently, acompletefactorial designwith27
experimentswasproposed. Table2containsall of the
experimental results (i.e. changed experimental fac-
tors and observed responses), and Figure8 provides
someof theresultsfor bothmodels.
Wehaveto indicatethat theobtainedresults from
thetwodevelopedmodelsarequitesimilar.
7 RESULTSANALYSISBY MULTIPLE
LINEAR REGRESSION
The general purpose of a multiple regression analy-
sis is to establish arelationship (criterion or model)
652
Table 2. Numerical experimental design results for
masonryblockparametersinthefirst developedmodel.
Plastic
Tensile Friction block
Test Cohesion strength angle (model 1)
(Number) (MPa) (MPa) (
) (Number)
1 1 1 30 1
2 1 1 15 9
3 1 1 0 24
4 1 0.5 30 1
5 1 0.5 15 9
6 1 0.5 0 24
7 1 0 30 16
8 1 0 15 22
9 1 0 0 55
10 0.5 1 30 29
11 0.5 1 15 76
12 0.5 1 0 112
13 0.5 0.5 30 29
14 0.5 0.5 15 77
15 0.5 0.5 0 112
16 0.5 0 30 46
17 0.5 0 15 96
18 0.5 0 0 116
19 0.2 1 30 116
20 0.2 1 15 116
21 0.2 1 0 116
22 0.2 0.5 30 116
23 0.2 0.5 15 116
24 0.2 0.5 0 116
25 0.2 0 30 116
26 0.2 0 15 116
27 0.2 0 0 116
betweenadependent variableandoneor moreinde-
pendent or predictor factors (Pedhazur, 1997). Inour
case, thedependent variableis thenumber of plastic
masonry blocks PN while the exploratory variables
are: masonryblockcohesion(C), masonryblockten-
silestrength(Tr),andmasonryblockfrictionangle(),
respectively. The general formof the adopted linear
model includinginteractiontermsisasfollows:
Where PN is the dependent variable,
0
the con-
stant and
1
,
2
,
3
,
13
,
23
and
123
theregression
coefficients of thevarious terms to besolvedby the
regressiontechnique.
The diagramin Figure 9 provides all of the cal-
culated multiple regression factor coefficients, with
which amultipleregression was applied to thestan-
dardisedexperimental designfactors.Thestandardisa-
tionprocesschangesall factor valuesover theinterval
[1, +1]. This multiple regression was performed
usingthewellknownMathematicasoftware.
Figure 9. Calculated regression factor coefficients and
their interactions(model 1).
Theproposedlinear model for theinitial masonry
blockshapethusbecomes:
For thesecond shapeof masonry blocks (model),
thelinear model becomes:
Thesetwolinearmodelsreveal thatamongthethree
studiedfactors, only cohesion(
1
) andfrictionangle
(
3
)significantlyinfluencemasonryblockmechanical
behaviour, though the model does show that cohe-
sionremains thesinglemost significant factor of all
thosestudied.Theinteractioncoefficientsindicatethat
only cohesionandfrictionangleexhibit asignificant
interactioncapableof influencingtheresponse.
8 CONCLUSION
Inthispaper, anexperimental designwasproposedto
simulateageingeffectsonoldtunnel behaviour;acom-
pletefactorial experimental design, which expresses
the evolution of three selected masonry mechanical
properties, was then forwarded. Thefactors selected
for thepresent study were: masonry block cohesion,
tensilestrengthandfrictionangle.
All experimental design tests were modelled by
means of thedistinct element method, executedwith
theUDEC code.
Results analysis indicated that among the three
studiedmasonryblockmechanical factors, onlycohe-
sion and friction angle along with their interaction
mayexertasignificantinfluenceonmasonrymechan-
ical behaviour; furthermore, cohesion proved to be
653
the most significant of all factors studied. The non-
influenceof tensilestrengthfactor isexplainedbythe
fact that the masonry structure is mainly loaded in
compressionduetoitsvaultedsectionshape.
For a typical model of an old tunnel, only three
mechanical properties (factors) of masonry support
in old tunnels were studied. It is important to point
out however that several other mechanical properties
of masonry must be taken into account in order to
studytheir influenceonthebehaviour of older tunnel
supports. Our subsequent researchgoal will focuson
involvingothermasonryblockandjointparametersin
studiesonoldtunnel behaviour.
REFERENCES
Barrentine, L.B. 1999. AnIntroductiontoDesignof Experi-
ments:ASimplifiedApproach.Publishedby theAmerican
Society for Quality, Chapter 3: experiments with three
factors: 2735.
Brookes, C.L. & Mullett, P.J. 2004. Services load testing,
numerical simulationandstrengtheningof masonry arch
bridges, CIMNE, Barcelona: 10pages.
Bicanic, N., Stirling, C. & Pearce, C.J. 1995. Discontinuous
modelling of masonry bridges, Computational Mechan-
ics, V. 31: 293314.
Bicanic, N., Stirling, C. & Pearce, C.J. 2002. Discontinuous
Modelling of Structural Masonry. Fifth World Congress
on Computational Mechanics, Vienna, Austria: 18pages.
CETu (Centre dEtude des Tunnels). 2004. Guide de
linspectiondugniecivil destunnelsroutiers, Ministre
de lquipement, des transports se du logement direction
des routes France. ISBN: 2-11-084749-2: 95pages.
Kerisel, J. 1975. Oldstructuresinrelationtosoil conditions.
Geotechniques 25, No. 3, p. 433484.
Loureno, P.B. 2001. Analysis of historical constructions:
From thrust-lines to advanced simulations,Historical
Constructions P. Roca (Eds.) Guimare : 91116.
Ford, T.E., Augarde C.E. & Tuxford, S.S. 2003. Mod-
ellingmasonryarchbridgesusingcommercial finiteele-
ment software, the 9th International Conference on Civil
and Structural Engineering Computing, Netherlands: 20
pages.
Giuriani, E., Gubana, A. & Arenghi, A. 2001. Structural
rehabilitation of masonry vault, Proceedings from the
UNESCO-ICOMOSMillenniumCongress, Paris: 6pages.
Hoek, E. 2000. Practical Rockengineering, Chapter11, Rock
massproperties. rocscience.comonline: 161203.
Hughes,T.G. &Davies,A.W. 1998.Theinfluenceof soil and
masonry type on the strength of masonry arch bridges,
proc, sec, int . arch bridge conf., A. sinopoli Ed: 321330.
Idris, J., Verdel, T. & Al Heib, M. 2004 1. Feedback from
adatabasecreatedfor reportingaccidentsintunnelsand
galleries. Urban Transport X. Urban Transport and the
Environment in the 21st century. Dresden, Germany, WIT
press, ISBN1-85312-716-7: 4150.
Idris, J., Al Heib, M. &Verdel, T. 2004 2. Basededonnes
desaccidentset desincidentssurvenusdanslesouvrages
souterrains. Tunnels et ouvrages souterrains, No. 182:
363368.
Idris, J., Verdel, T. & Al Heib, M. 2007. Numerical mod-
elling and mechanical behaviour analysis of ancient
tunnel masonry structures. Tunnelling and Underground
SpaceTechnology.(Articleinpress) :13pages.
J anssen, H.J.M. 1997. Structural Masonry. Structural
Masonry, Numerical studieswithUDEC. Centrefor civil
engineeringresearchandcodes.A.A.Balkema,Roterdam,
Netherlands, ISBN9054106808: 96106.
Olofsson, T. 1985. A non-linear model for the mechanical
behaviour of continuous rock joints. In Proc. of Inter-
national Symposium on Fundamental of Rock Joints, Ed
StephanssonO. Bjrkliden, Sweden: 395404
Pedhazur, E.J. 1997. Multiple regression in behavioral
research, third edition, Chapter 5, Multiple regression,
New York: Harcourt Brace College Publishers. ISBN
0-03-072831-2: 95135.
Sumon, S. K. &Ricketts, N. 1995. Strengtheningof Masonry
Arch Bridges. Chapter in Arch Bridges. Publ Thomas
Tellord, London.
Valluzzi, M.R., Binda, L. & Modena, C. 2004. Mechani-
cal behaviour of historicmasonrystructuresstrengthened
by bed joints structural repointing. Construction and
Building Materials V.19: 6373.
Verdel, T. & Bigarre, P. 1999. Modlisation de tunnels
anciensaveclelogiciel UDEC, Rapport INERIS, Socit
SIMECSOL : 112.
654
Excavationwithstepped-twinretainingwall: Model testsandnumerical
simulations
N. Iwata, H.M. Shahin, F. Zhang, T. Nakai, M. Niinomi &Y.D.S. Geraldni
Department of Civil Engineering, Nagoya Institute of Technology, Nagoya, Japan
ABSTRACT: Bracedexcavationusingstepped-twinretainingwall isbecomingpopularinJ apan.Asitisanew
techniqueusedtoprevent movementsof double-elevatedground, mechanismof deformationduetoexcavation
andchangeof stresses havenot beenfully understood. Thedesignmethodology of this techniqueis also not
properlyestablished. Inthisresearch, two-dimensional model testsareconductedtoinvestigatethedeformation
mechanismof thegroundandtheearthpressureof thestepped-twinretainingwall. Numerical simulationswith
finiteelementmethodarealsocarriedoutforthesamescaleof themodel tests.Theaimof theresearchistomake
clear themechanismof thebracedexcavationusingstepped-twinretainingwall andtoestablishaneffectiveway
toevaluatingthemechanical behaviorsof theretainingwall andthesurroundingground.
1 INTRODUCTION
Inurbanarea, openexcavationoftencauseproblemsto
surroundinggroundandadjacent structures. Inprac-
tical daily design works, however, earth pressure is
usually predicted by conventional methods such as
a frame model together with Rankines earth pres-
sure theory. There is also no appropriate method to
predict surfacesettlements of ground, which is usu-
ally predicted by empirical method and/or elastic
finite element method. For braced excavation using
stepped-twinretainingwall, conventional methodcan-
not takenintoconsiderationproperlytheinfluenceof
nearbystructuresaswell astheconstructionsequence
in evaluating the ground movements and the earth
pressure.
Inthisresearch, two-dimensional (2D) model tests
are conducted to investigate the deformation mech-
anismof the ground and the earth pressure of the
stepped-twin retaining wall. Numerical simulations
with finite element method are also carried out for
the same scale of the model tests. In the finite ele-
mentanalyses, subloadingt
ij
model (Nakai &Hinokio
2004), is used in the analysis to model the ground
material. Thismodel candescribetypical stressdefor-
mationandstrengthcharacteristicsof soilssuchasthe
influence of intermediate principal stress, the influ-
enceof stresspathdependencyof plasticflowandthe
influenceof density and/or confiningpressure. Mass
of aluminumrodsisusedinthemodel ground. Several
patternsof themodel testsareperformedvaryingthe
lengthof theretainingwall andchangingthedistance
Figure1. 2DModel test device.
between the two walls. The results of stepped-twin
retainingwall arecomparedwiththesingleretaining
wall.
2 OUTLINE OF MODEL TESTSAND
NUMRICAL SIMULATIONS
2D Model tests and the corresponding numerical
simulations of braced excavation using stepped-twin
retaining wall, were carried out to make clear the
mechanical behavior theproblem. Figure1showsthe
outline of the 2D Model test device. Four cases of
model testswithdifferentlengthof outerretainingwall
anddifferent spacingof stepped-twinretainingwalls
wereconsidered, asshowninFigure3.
655
Figure2. Outlineof 2DModel test device.
Figure3. Casesof study.
Figure2shows theschematic diagramof thetwo-
dimensional apparatus.Thesizeof themodel groundis
68cminwidthand45cminheight.Aluminumrodsof
5cminlength,havingdiametersof 1.6mmand3.0mm
andmixedintheratioof 3:2inweight, areusedasthe
model ground(unitweightof themassis20.4kN/m
3
).
In the experiment, the model ground was excavated
with athickness of 1.5cmeach timeand two struts,
locatedat thelevelsof 1.5cmand7.5cmrespec-
tively, wereset intoplaceat thetimewhenexcavating
level reached1.5cmbelowitsposition. Theretain-
ingwalls, withdifferent lengthandspacing, wereset
beforetheground was excavated. Table1 shows the
material parameters of themodel ground, theretain-
ingwall (Aluminumplate) andthestruts. A laser type
Table1. Material Parameters.
Ground Aluminumrods
Unit weight =20.4(kN/m
3
)
Retainwall Aluminumplate
EI =0.88(kN*m
2
/cm)
EA=4.2210
4
(kN/cm)
Strut Upper: k
1
=3.64(kN/m/cm)
Lower: k
1
=4.13(kN/m/cm)
Table2. Parametersof groundmadeof aluminumrodmass.
0.008
0.004
N e
NC
at 0.3 sameparametersas
p 98kPa& Cam-claymodel
q =0kPa
R
cs
=(
1
,
3
)cs 1.8
e
0.2
1.2 shapeof yieldsurface(same
asoriginal Cam-clayat =1)
a 1300 influenceof densityand
confiningpressure
displacement transducer is used to measure surface
settlementof theground.Bytakingphotographswitha
digital cameraandusingimageprocessingof thepho-
tos, thedistributionof movementandconsequentlythe
strainof thegroundcanbemeasured.
Numerical analyses arecarried out with thesame
scaleof themodel test consideringplanestraincon-
dition using isoparametric element. An elastoplastic
constitutive model, named as subloading t
ij
model
(Nakai & Hinokio, 2004) is used in the finite ele-
mentanalysestosimulatethemechanical behaviorsof
themodel ground. Themodel cantakeintoconsider-
ationautomatically theinfluenceof theintermediate
principal stress, by introducing a modified stress t
ij
(Nakai andMihara, 1984; Nakai andMatsuoka, 1986).
Subloadingsurfaceconcept proposedby Hashiguchi
(1980) was also adopted in the model to consider
theinfluenceof densityof groundmaterials. Detailed
description about the performance and the reason-
ing of the model can be referred to aforementioned
references.
Table2liststheparametersof model groundmade
of aluminumrod. Figure4showstheperformanceof
themodel. Figure5showsthefiniteelement meshof
Case1-b. Smoothboundary conditionis assumedfor
sideboundaries, andthebottomof themeshesiskept
fixed. Theinitial stressesof thegroundarecalculated
bysimulatingtheself-weight consolidationbyapply-
ingbody forces, startingfromanegligibleconfining
pressure(p
0
=9.810
6
kPa)andaninitial voidratio
656
Figure4. Stress-strain-dilatancy curves for aluminumrod
mass.
Figure5. Finiteelement mesh(Case1-b).
e=0.35. Theretainingwall wassimulatedwithbeam
elementandthestrutissimulatedwithspringelement.
Betweenthegroundandtheretainingwall, joint ele-
ment whose mechanical behavior is simulated by a
perfect-plasticjointelements(Nakai,1985),wasintro-
ducedtoconsider possibleslidingbetweentheground
andthewall.
Figure 6 shows the observed surface settlements at
different excavation stages in four cases. Thelength
of outer retaining wall does not affect too much the
settlement, whilethespacingof twinwallshasagreat
influenceonthesettlement.Theshorterthespacingis,
thelarger thesettlement will be. Thesametendency
canbeobtainedinthecorrespondent calculations, as
showninFigure7.
Figures 8and9showtheobservedandcalculated
deflections of theretaining walls during theexcava-
tion. Similar tothesurfacesettlement, themainfactor
affectingthedeflectionisthespacingof twinwalls.
From Figures 6 to 9, it is also known that the
numerical calculationconductedinthispaper cannot
only well describe the deformation patterns of the
ground and the retaining wall qualitatively but also
quantitativelytosomeextent.
Figure6. Surfacesettlements(Observed).
Inthecalculation, thefrictional angleof joint ele-
mentswhichareusedtosimulatethefrictionbetween
theground and thewall is determined with constant
normal stress frictional test and is found to be 17
degree. Inthecalculation, however, thedisplacement
of thegroundalongthewall doesnot fit theobserved
onewell. This is duetothefact that aperfect-plastic
model is used for the joint elements which do not
allowanyelasticdeformationbeforethejointelement
reachesyieldingstate. Thereasonwhy wedonot use
elasto-plasticmodel isthat it isdifficult todetermine
657
Figure7. Surfacesettlements(Computed).
theshear stiffness of joint element. Further research
needstobedowninfuture.
Figures 10 and 11 show thedisplacements of the
groundsurroundingtheexcavatedarea, obtainedboth
frommodel testandnumerical calculation. Thedefor-
mation of theground is limited to thearea near the
excavation. Thecalculatedresultsagreewell withthe
observedones.
Figures 12 and 13 show the distribution of shear
strain(Thesecondinvariantof straintensor), obtained
bothfrommodel test andnumerical calculation. It is
foundthattheshearstrainof thegroundisalsolimited
Figure8. Displacementsof retainingwall (Experiment).
658
Figure9. Displacementsof retainingwall (Computed).
Figure 10. Displacement of ground (Observed,
d=180mm).
Figure 11. Displacement of ground (Computed,
d=180mm).
totheareanear theexcavation. Thecalculatedresults
agreewell withtheobservedones.
Figure14givesacomparisonbetweentheobserved
and the calculated results of axial forces within the
struts at different excavating stages. Similar to the
surfacesettlement, themainfactor affectingtheaxial
forceisthespacingof twinwallsinsteadof thelength
of theouter wall. Thecloser thespacingis, thehigher
the axial force of the second strut will be. The first
659
Figure 12. Distribution of shear strain (Observed,
d=180mm).
Figure 13. Distribution of shear strain (Computed,
d=180mm).
strut, however, is not affectedtoomuchby thesetwo
factors, thatis, thelengthof wall andthespacing. The
numerical calculationcanwell describethetestresults,
bothqualitativelyandquantitatively.
Figure15givesacomparisonof calculatedchange
of earth pressures on retaining walls during excava-
tionfor thecases of 1-aand1-b. It is foundthat the
earth pressures on inner retaining wall do not show
muchevidentdifference.Theouterwall,however,does
behavequitedifferently, thatis, thepassiveearthpres-
sureonthebottomincase1-a(narrowspacing)ismuch
smallerthanthosein1-b(widespacing), implyingthat
thegroundbetweenthetwinwall cannot beexpected
to resist thedeflection of theouter wall as what we
Figure14. Comparisonof axial forcesinstruts.
Figure 15. Change of earth pressures on retaining walls
duringexcavation.
usually expectedto beapassiveearthpressurewhen
thespacingof thetwowall isenoughnarrow.
Figure16showsacomparisonof calculatedresults
of axial forces instruts indifferent excavationmeth-
ods. Insingle-wall excavation, theaxial forceinupper
strut increases firstly and then decreases after the
lower strut comes into action. In twin-wall excava-
tion, however, boththeaxial forcesinupper andlower
strutsincreaseduringtheexcavation. Meanwhile, the
660
Figure16. Comparisonof different excavationmethods.
axial forceof lower strut increasesmuchfast intwin-
wall excavationthanthoseinsingle-wall excavation.
Thereforethemechanismof twin-wall excavationand
single-wall excavationismuchdifferentandshouldbe
consideredcarefullyindailydesign.
4 CONCLUSIONS
Laboratorymodel testsandthecorrespondingnumer-
ical simulations are conducted for investigating the
deformation mechanism of the ground and earth
pressureof thestepped-twinretainingwall. Fromthis
researchthefollowingpointscanbeconcluded:
1. Thedisplacements of thewalls areinversely pro-
portional tothedistancebetweenthewalls.
2. Thesurfacesettlement follows thesametendency
of the wall displacements, and it is very much
dependent onthedistancebetweenthetwowalls.
3. Thedistancebetweenthewalls is moreimportant
factor thantheembeddedlengthof thewall.
4. Unlike the single retaining wall the struts of the
stepped-twinretainingwall shareaxial loadamore
efficient way.
Finite element analysis conducted in this paper,
whichisbasedonsubloadingt
ij
model and, iscapable
todescribethemechanical behaviorsof thetwin-wall
excavationqualitativelyandquantitatively.
REFERENCES
Hashiguchi, K. 1980. Constitutive equation of elastoplas-
tic materialswithelasto-plastic transition. Jour. of Appli.
Mech. ASME 102(2): 266272.
Nakai, T. 1985. Finiteelement computations for activeand
passive earth pressure problems of retaining problems.
Soils and Foundations 25(3): 98112.
Nakai, T. & Hinokio, M. 2004. A simpleelastoplasticmodel
material parameters. Soils and Foundations 44(2): 5370.
Nakai, T. & Matsuoka, H. 1986. A generalizedelastoplastic
constitutivemodel for clayinthree-dimensional stresses.
Soils and Foundations 26(3): 8198.
Nakai, T. & Mihara, Y. 1984. A new mechanical quantity
for soils and its application to elastoplastic constitutive
models. Soils and Foundations 24(2): 8294.
661
Stabilityof anunderwater trenchinmarineclayunder oceanwaveimpact
T. Kasper & P.G. J ackson
COWI A/S, Kongens Lyngby, Denmark
ABSTRACT: A longsectionof theimmersedtunnel of theBusan-GeojeFixedLink inSouthKoreaistobe
constructedinatrenchinsoft marineclay. Thesiteis exposedto largetyphoonwaves andthetrenchis to be
left openfor approximately oneyear. Thetrenchprofilewas chosenas abalancebetweenexcavatedvolume
(costs) and risk of slope instability under large waves. To evaluate this risk, far-shore wave data have been
transformedintonear-shorewaveconditionsbymeansof numerical modellingandwaveflumetestshavebeen
carried. Potentiallycritical waveshavebeenidentifiedandtheir impact hasbeenanalysedbymeansof coupled
hydro-mechanical numerical simulations. Basedonthestrengthreductionmethod, safety factorsfor theslope
stabilityduringwaveimpact havebeendetermined. Theresultsandtheconsequencesfor theeconomicdesign
of suchatrenchsubject tolargewavesarediscussed.
1 INTRODUCTION
TheBusan-GeojeFixedLink betweenSouthKoreas
secondlargestcityBusanandtheislandof Geojecon-
sistsof a3.2kmlongimmersedtunnel (Odgaardetal.
2006), tworock tunnelsandtwo1.7and1.9kmlong
cable-stayed bridges. About 2.2kmof theimmersed
tunnel is constructed in a12 to 15mdeep trench in
softmarineclayatwater depthsbetween20and50m.
Theareaispronetofrequentraidsof typhoonsandthe
trench is dredged approximately 1 year before con-
struction of the tunnel. When the tunnel is placed
in the trench, the lower half of the trench is filled
withtunnel protectionmaterial (backfill). Itwasthere-
forenecessary to design thetrench profileas abal-
ancebetweentheexcavatedvolume, i.e. construction
costs on the one hand and the risk of slope failure
under wave impact within a 1 years period on the
other hand.
This paper discusses in Section 2, how the wave
characteristics at the site have been determined and
howthemostcritical waveswithregardtotrenchslope
stabilityhavebeenidentified. Basedonthegeotechni-
cal characterizationof thesoftmarineclaypresentedin
Section3, theimpactof thecritical wavesonthetrench
isinvestigatedbymeansof coupledhydro-mechanical
numerical simulations(Section4). Strengthreduction
analysesaremadeinthesimulationsinregular inter-
valsof onesecondtodeterminethefailuremodesand
factors of safety (Dawson et al. 1999). Thelast part
of thepaper contains adiscussion of theresults and
general conclusions.
Figure 1. Numerical wave modelling: Calculated signifi-
cantwaveheightsfora10,000yearsreturnperiod.Thetunnel
isindicatedbythedashedline.
2 DERIVATIONOF DESIGNWAVES
2.1 Numerical wave modelling
Basedonastatistical analysisof waveandwinddata
inthearea, theextremefar-shoreboundarywavesand
extremewind speeds havebeen derived. Fromthese
input data, thewaveconditions at thelocationof the
tunnel have been determined by means of numeri-
cal wavemodellingwiththeprogramMIKE 21(DHI
Water & Environment).
The obtained maximumsignificant wave heights
alongthetunnel alignmentare6.2mforareturnperiod
of 10years, 8.0mfor areturnperiodof 100yearsand
9.2mfor the10,000yearswaveevent (Figure1). The
largewavescomefromsoutherlydirectionwithawave
velocity of approximately 13m/s andpass thetunnel
663
Figure2. Effectof typhoonwavesonabreakwaterinBusan
(typhoon Maemi, 12.9.2003): (a) before and (b) after the
storm.
Wave pressure profiles at seabed
-4
-2
0
2
4
-100 -50 0 50 100
Length (m)
E
x
c
e
s
s

p
r
e
s
s
u
r
e

(
m
)
10 years wave
100 years wave
10,000 years wave
Figure 3. Most critical seabed pressure profiles for the
trenchstabilitydeterminedfromthewaveflumetests.
trench almost perpendicularly. The potential devas-
tating effect of such large waves is illustrated for a
breakwater inFigure2.
2.2 Wave flume model tests
Basedontheresultsof thenumerical wavemodelling,
waveflumetestshavebeencarriedout for tunnel ele-
ment TE7, whichis subject to thehighest waves at a
relativelyshallowwater depthof 20m. Eachtest con-
sistedof aseriesof several hundredsof waves. From
the pressure measurements in the tests, the seabed
pressureprofiles withthelargest pressuredifference
across the 32mwide trench slopes have been iden-
tified. Thesepressureprofiles showninFigure3are
consideredtobemostcritical for theslopestability. It
shouldbenotedthatduetohydrodynamics, theseabed
pressuresunderwavesdonotsimplycorrespondtothe
physical waveheight above.
3 GEOTECHNICAL PROPERTIES
Extensiveground investigations and laboratory tests
havebeencarriedout for theproject (Steenfelt et al.
Figure4. Represenationof undrainedstrengthinthenumer-
ical model withtheMohr-Coulombmodel basedoneffective
strengthparameters.
Marine clay
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
0.0 10.0 20.0 30.0 40.0
c
u
(kPa) in situ
D
e
p
t
h
(
m
)
Measured lower
bound (SHANSEP,
prec. stress 10 kPa
Measured upper
bound (SHANSEP,
prec. stress 65 kPa)
Numerical model
(Mohr-Coulomb)
Figure 5. Comparison between measured and modelled
undrainedstrengthof themarineclayfor insituconditions.
2008). The undrained strength of the clay has been
assessed from 30 cone penetration tests along the
alignment and can be described according to the
SHANSEP approach(Steenfelt & Foged1992) as
withthepreconsolidationstressL of themarineclay
typically ranging between 10 and 30, max. 65kPa.
Theeffectivestrengthparameters
=25
, c
=3kPa
have been derived from19 consolidated undrained
triaxial tests. In the numerical model, the material
strengthisdescribedwiththeMohr-Coulombmodel.
Duetothefact that theclay is only slightly overcon-
solidated, theMohr-Coulombmodel canalsobeused
withoutmodificationtocorrectlymodel theundrained
strengthincaseof undrainedconditions(Figure4and
Figure5).
The stiffness properties and permeabilities have
been derived from oedometer tests. The material
parametersaresummarisedinTable1.
664
Table1. Material parametersforthenumerical simulations.
Marineclay Alluvium
Material model Mohr-Coulomb Mohr-Coulomb
sat
(kN/m
3
) 14.7 20
K
0
(-) 0.6 0.426
E (kPa) 8000 50000
(-) 0.2 0.25
) 25 35
c
(kPa) 3 0
k(m/s) 1 10
9
1 10
5
FLAC (Version 5.00)
LEGEND
12-Mar-07 23:19
step 149210
Flow Time 0.0000E+00
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
User-defined Groups
Alluvium
Marine clay
Grid plot
0 5E+01
-60
-20
20
60
100
20 60 100 140 180
JOB TITLE : Trench stability analysis
COWI A/S
Denmark
Figure6. FLAC model of thetrench.
4 NUMERICAL SIMULATIONOF WAVE
IMPACT
4.1 Model description
Theeffectof wavesonthetrenchismodelledbymeans
of numerical simulations using the finite difference
programFLAC (Itasca2005). Thetwo-dimensional,
plane-strainmodel isshowninFigure6.
Thevertical boundaries of themodel arefixed in
horizontal direction, while the bottomof the model
isfixedbothinhorizontal andvertical direction. The
vertical boundariesandthebottomareassumedtobe
impermeable.
Soil stressesandporewaterpressuresareinitialised
according to the self-weight of the soils, K
0
=1
sin
andastill watertableat+63mmodel coordinate

(still water pressuresp
still
).
In a first step, the trench excavation is modelled
either asadrainedor undrainedprocess. Thisisdone
to allow for modelling of wave impact shortly after
trench excavation (undrained excavation modelling)
andwaveimpact1year after trenchexcavation, where
the excess pore pressures in the clay in the relevant
areanext tothetrenchslopes will almost completely
bedissipated(drainedexcavationmodelling).
Afterwards, wave impact is modelled by means
of coupledhydro-mechanical analyses withtransient
Wave pressure profile at seabed
-4
-2
0
2
4
-100 -50 0 50 100
Length (m)
E
x
c
e
s
s

p
r
e
s
s
u
r
e

(
m
)
10 years wave
Approximation
Figure7. Approximationof the10yearswavepressurepro-
fileintheFLACmodel (A
1
=3.4m, L
1
=70m, A
2
=3.4m,
L
2
=80m).
hydraulic andmechanical boundary conditionsalong
theseabedandthetrenchsurfaces. Thewavepressure
profileisdescribedbymeansof twosinecurves, one
for thewavecrest andonefor thewavetrough. This
allows to consider different shapes of thewavecrest
andwavetroughaccordingto
The described approach is able to model the wave
pressureprofilesquiteaccurately, asillustratedfor the
10yearswaveinFigure7.
BasedonEqs. (2) and(3), thewavepressures are
evaluated for each point in time at each gridpoint
alongtheseabedandthetrenchsurfacesandthecor-
respondingtotal normal stressandporewaterpressure
boundaryconditions
are applied. It should be noted that water pressures
are defined positive, while compressive stresses are
definednegative. Accordingtotheprincipleof effec-
tivestresses, theseboundaryconditionsimplythatthe
effectivenormal stressesat theseabedandthetrench
surfaces remain0. Basedonalinear variationof the
boundaryconditionsalongeachzoneedge, thewhole
wavepressureprofileisapproximatedasapiecewise
linear function with the correct values at each grid-
point. The model is defined such that a wave train
withanarbitrary number of waves canbesimulated.
As a compromise between computational effort and
concernaboutapossiblechangeinsafetiesfromwave
towave, awavetrainwith2wavesisconsideredinthe
basicsimulations.Duetothefactthateachof theinves-
tigatedlargewaves passes thetrenchwithinaperiod
of approximately6seconds,thesimplifiedassumption
of quasi-staticconditionsinthesimulationappearsto
beadequate. Thecompressibility of theporewater is
665
FLAC (Version 5.00)
LEGEND
12-Mar-07 23:19
step 149210
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
Boundary plot
0 5E+01
Factor of Safety 1.86
Max. shear strain-rate
0.00E+00
1.50E-07
3.00E-07
4.50E-07
6.00E-07
7.50E-07
9.00E-07
Contour interval= 1.50E-07
COWI A/S
Denmark
-60
-20
20
60
100
20 60 100 140 180
Figure8. Predictedfailuremechanismwithoutwaveimpact
after full consolidation of the trench excavation (drained
conditions): Safetyfactor 1.86.
taken into account by a bulk modulus K
w
=2 10
9
Pa(purewater), whilethecompressibility of thesoil
grainsisneglected.Failuremechanismsandsafetyfac-
torsaredeterminedintimeintervalsof 1secondduring
waveimpact bymeansof strengthreductionanalyses
(Dawsonet al. 1999).
4.2 Results
Figure8showstheresult of asafety analysiswithout
waveimpactafterfull consolidationof thetrenchexca-
vation. Thefailuremechanismisillustratedbymeans
of maximumshearstrainratesattheendof thestrength
reductionanalysis.
The FLAC result can be verified by comparison
with other solutions (Table 2 and Figure 9). The
drained results can be compared with a traditional
limit equilibriumsolutionandafiniteelement stress
basedsolutionobtainedwiththeprogramSLOPE/W
(GEO-SLOPE) as well as with a strength reduction
analysis with the finite element program PLAXIS
(Brinkgreve&Bakker 1991). Inthecaseof undrained
conditions, a FE calculation is needed with all pro-
gramsinordertocorrectlypredicttheeffectivestresses
and excess pore pressures. The shapes of the slip
surfacesandthesafetyfactorsshowsatisfyingagree-
ment inall cases. It is obvious that thesafety factors
decrease with consolidation. The reason is that the
excavation represents an unloading process leading
to a time-dependent reduction in effective vertical
stresses,whicharecrucial forthesoil strength.Accord-
ing to theMohr-Coulomb failurecriterion theshear
strengthislinearlydependent ontheeffectivenormal
stress. Itisestimatedthatafteroneyear, theconditions
intheclayinthecritical areaclosetotheslopesurfaces
maybenearlydrained.
Figure 10 shows the evolution of the computed
safetyfactorsduringwaveimpactafter full consolida-
tionof thetrenchexcavation. It canbeobservedthat
Table2. Comparisonof safetyfactorswithoutwaveimpact.
FLAC GEO-SLOPE PLAXIS
Undrainedexcavation 2.67 2.62
2
2.64
Drainedexcavation 1.86 1.83
1
, 1.83
2
1.87
1
General limit equilibriumsolution
2
Solutionbasedonfiniteelement stresses
Figure 9. Illustration of results for drained conditions:
(a) Traditional limit equilibrium solution (SLOPE/W),
(b) Finiteelement strengthreductionsolution(PLAXIS).
Safety factors during wave impact after full
consolidation of the excavation
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
0 10 20 30 40 50
Time (s)
S
a
f
e
t
y

f
a
c
t
o
r

(
-
)
10 years wave
100 years wave
10,000 years wave
Figure 10. Computed safety factors during wave impact
after full consolidationof thetrenchexcavation.
thesafety factors showalargevariationintherange
between0.91and4.12duringpassageof thewaves.
The safety factors exceed 1.86 obtained for still
water when a wave crest is above the trench. Then,
increasedpressuresactalongthebottomof thetrench
anddecreasealongthetrenchslopestowardsthecrown
pointsof thetrench(Figure11).Thispressuredistribu-
tioninthetrenchcounteractsthedestabilizinggravity
forces, thus leadingto increasedsafety factors of up
to4.12.
Ontheother hand, awavetroughabovethetrench
causeshighpressuresontheseabedbesidethetrench
which decrease down along the trench slopes (Fig-
ure12). Thispressuredistributionaddstothegravity
forces,thusleadingtodecreasedsafetyfactorsof down
to 0.91. Thepredictionof safety factors smaller than
1for the100and10,000years wavehas beenfacili-
tatedbyscalingtheclaystrength(tan
andc
) upand
scalingtheobtainedsafetyfactorsdown.
According to Table 3, the difference in minimum
safety factors between the10, 100 and 10,000 years
666
FLAC (Version 5.00)
LEGEND
13-Mar-07 5:25
step 109770
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
Boundary plot
0 5E+01
0.00E+00
2.00E-07
4.00E-07
6.00E-07
8.00E-07
1.00E-06
1.20E-06
COWI A/S
Denmark
-60
-20
20
60
100
20 60 100 140 180
Figure11. 10yearswave, t=16s: Illustrationof wavepres-
sureprofileandpredictedfailuremechanismwiththelargest
safetyfactor of 4.12for waveimpact after full consolidation
of thetrenchexcavation.
FLAC (Version 5.00)
LEGEND
13-Mar-07 23:11
step 203328
-1.089E+01 <x< 2.069E+02
-9.239E+01 <y< 1.254E+02
Boundary plot
0 5E+01
0.00E+00
1.50E-08
3.00E-08
4.50E-08
6.00E-08
7.50E-08
9.00E-08
COWI A/S
Denmark
-60
-20
20
60
100
20 60 100 140 180
Figure 12. 10 years wave, t=22s: Illustration of wave
pressure profile and predicted failure mechanismwith the
smallest safety factor of 1.01 for wave impact after full
consolidationof thetrenchexcavation.
Table3. Summaryof minimumsafetyfactors.
Waveimpact Waveimpact after
shortly full consolidation
after trench of thetrench
excavation excavation
10yearswave 1.38 1.01
100yearswave 1.29 0.95
10,000yearswave 1.24 0.91
wavesisrelativelysmall andthesafetyfactorsforwave
impactafterfull consolidationof thetrenchexcavation
are approximately 36% lower than the correspond-
ingsafetyfactorsfor waveimpact shortlyafter trench
excavation.
Figure10showsthesafetyfactorsfor waveimpact
after full consolidation of the excavation (drained
modelling of theexcavation). Graphs with thesame
general pattern, but with larger safety factors are
obtainedfor waveimpact shortly after excavation of
thetrench(undrainedmodellingof theexcavation).
It may be concluded from the calculated safety
factors that failure of the slopes may occur if these
extreme waves would pass the trench after 1 year
consolidation. However, it can be argued that slid-
ingof slopesisadynamic processwhichtakesplace
in the range of at least several seconds. Figure 11
and Figure12 illustratethat themaximumand min-
imumsafety factors are only a few seconds apart.
Even if theexceptional 100 and 10,000 years waves
would pass the open trench after 1 year consolida-
tion, larger sliding movements may beprevented by
thefact that critical safety factors occur only within
about 1or 2seconds. Althoughwater waveimpact is
aslower phenomenon, thesituationmaybecompared
withearthquakeimpactwheretheexceedanceof yield
accelerations in shorter periods induces some irre-
versibleplasticdeformationswhichdonotnecessarily
leadtocompletefailure.Anestimateof expectedslid-
ingmovementsduringwaveimpactcouldbeobtained
e.g. by means of dynamic coupledhydro-mechanical
simulationswiththedynamicversionof FLAC.
A simulation of the 10 years wave event with
K
w
=2 10
7
Pa(representingporewater withahigh
content of dissolved/entrappedair) yieldsaminimum
safety factor of 0.99 for waveimpact after full con-
solidationof theexcavationand1.35for waveimpact
shortly after excavation of thetrench. Thesesafeties
forhighlycompressibleporewaterareslightlysmaller
thanthecorrespondingsafetiesof 1.01and1.38forthe
basiccase. Theinfluenceof porewater compressibil-
ity onthebehaviour of different soil structuresunder
drawdownandwaveimpact hasbeeninvestigatedin
detail e.g. byKhler (2000).
No perceptible change in safety factors can be
observedbetweenthefirstandsecondwaveinthesim-
ulations(Figure10). Thishasfurther beenconfirmed
byasimulationwithawavetrainof 6waves.
5 DISCUSSIONANDCONCLUSIONS
Dueto frequent raids of typhoons inthearea, it was
necessarytoconsiderthepossibleeffectof largewaves
inthedesignof the2.2kmlongand12to15mdeep
trenchfor theimmersedtunnel of theBusan Geoje
Fixed Link in South Korea. Safety factors for the
trenchstabilityunder extremewaveswithreturnperi-
odsof 10, 100and10,000yearshavebeendetermined
by means of numerical simulations. The minimum
values have been found to be between 1.24 (10,000
yearsreturnperiod) and1.38(10yearsreturnperiod)
for wave impact shortly after trench excavation and
between 0.91 (10,000 years return period) and 1.01
667
(10 years return period) for wave impact at the end
of theconstructionperiod(after consolidationof the
trench excavation). These safety factors have been
consideredtobeacceptablefor thefollowingreasons:
Theinvestigatedsituationisatemporaryconstruc-
tionphase, wherethetrenchisopenfor about 1year.
Under thesecircumstances, the10yearswaveeventis
consideredtobetherelevant designcase.
Although theminimumsafety factors for the100
and10,000yearswavesareslightlybelow1(Table3),
failureisunlikelyevenundertheseextremewavesdue
tothefactthattheminimumsafetyfactorsoccurwithin
only1or 2seconds(Figure10). Thisperiodisconsid-
eredtobetooshortforcompletefailuretooccur. High
strain rates would berequired for failureto develop
and it is a well-established fact that the strength of
clayisactuallyhigher for highstrainrates. Minimum
and maximumsafety factors areonly afewseconds
apart (Figure10).
Failure of the trench slopes represents a finan-
cial risk andwouldrequireaclean-upafterwards, but
wouldneither affect lifesnor thetechnical successof
theproject. Thecostsfor excavationof atrenchinoff-
shoreconditionsatwater depthsbetween20and50m
arehighandshouldbelimitedtoaminimum.
6 PROJ ECT PROGRESS
Theexcavationof thetrenchbeganinJ uly 2006and
was finished in September 2006. Busan only expe-
riencedaminor tropical storminsummer 2006. The
productionof thesoil improvement(cementdeepmix-
ing) of themarineclayinthetrenchfor thefoundation
of thetunnel beganinNovember 2006andhas been
finished in May 2007. The offshore construction of
thetunnel isscheduledtostart at theendof 2007and
finishin2010.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the permission
fromDaewooEngineering& Constructiontopublish
thispaper.
REFERENCES
Brinkgreve, R.B.J. & Bakker, H.L. 1991. Non-linear
finite element analysis of safety factors. In G. Beer,
J.R. Booker & J.P. Carter (eds), Computer Methods
and Advances in Geomechanics: 11171122. Rotterdam:
Balkema.
Dawson, E.M., Roth, W.H. & Drescher, A. 1999. Slope
stability analysis by strength reduction. Gotechnique
49: 835840.
DHI Water & Environment, Denmark. MIKE 21. www.
dhisoftware.com/mike21/
GEO-SLOPE International, Ltd. GeoStudio 2004, version
6.19. www.geo-slope.com.
ItascaConsultingGroup, Inc. 2005. FLAC FastLagrangian
Analysisof Continua, version5.0. www.itascacg.com.
Khler, H.J. 2000. Pressure spreading at soil water inter-
faces and its influence on soil structure design. In
A. Cancelli et al. (eds), EUROGEO2000; Proc. Sec-
ond European Geosynthetics Conference, Bologna, 1518
October 2000: Vol. 2, 687694. Bologna: PatronEditore.
Odgaard, S.S., J ensen, O.P., Kasper, T., Yoon, Y.H.,
Chang, Y., & Park, R.Y. 2006. Designof longimmersed
tunnel for highway in offshore conditions Busan
Geoje Fixed Link. Tunnelling and Underground Space
Technology 21(34),Special Issue:SafetyinUnderground
Space (CD-ROM Proceedings of the ITA-AITES 2006
WorldTunnel Congress, Seoul, Korea).
PLAXISb.v. PLAXISversion8.5. www.plaxis.nl.
Steenfelt, J.S. & Foged, N. 1992. Clay till strength
SHANSEP andCSSM. NGM-92, 8186.
Steenfelt, J.S., J ackson, P.G., Christensen, C.T., Lee, J -S., &
Ha, Y.-B. 2008. Ground investigations for the Busan
GeojeImmersedTunnel. Submittedto 3rd International
Conference on Site Characterisation, ISC3,Taipei, 2008.
Zienkiewicz, O.C., Humpheson, C. & Lewis, R.W. 1975.
Associatedandnon-associatedvisco-plasticity andplas-
ticityinsoil mechanics. Gotechnique 25(4): 671689.
668
A studyonbehavior of 2-archtunnel byalargemodel experiment
S.D. Lee
Department of Civil Engineering, Ajou University, Korea
K.H. J eong, J.W.Yang& J.H. Choi
Dodam Engineering and Construction Co., Korea
ABSTRACT: It is tendency that theparallel tunnels areconstructedclosetoeachother inorder todiminish
civil complaints and environmental damage. The 2-Arch tunnel is similar to two parallel tunnels with very
short centre-to-centredistance. Recently, constructionof 2-Archtunnel isincreasing. However, it isexecuting
without enough studies for behavior of the 2-Arch tunnel. In this research, a study for behavior of 2-Arch
tunnel isexaminedusinglargemodel testmachine.Atfirstwemakethemodel groundwithhorizontal joint-set.
Thenembody in-situstress by applyingpressureto boundary of model ground. Thenexcavatemodel ground
accordingto constructionsteps of the2-Archtunnel. As aresult, duringexcavationof pilot tunnel, measured
grounddisplacementsareabout 4050%of wholedisplacement, whichisconcentratedin0.25D(D: diameter
of tunnel) regionaroundtunnel. Height of loosenareabyconstructionof the2-Archtunnel is0.15w(w:centre-
to-centredistancebetweenleftandrighttunnel).TheseresultsarecomparedwithDEMtoconformthereliability
of results.
1 INTRODUCTION
The 2-Arch tunnel is similar to two parallel tunnels
withveryshort centre-to-centredistance. The2-Arch
tunnel has beenincreasinginorder todiminishenvi-
ronmental damageandensureslinkingwithstructure
closebytunnel.Therehavebeenaseriesof researches
onthebehaviorof 2-Archtunnel insoil andweathered
rock. However, theuseof 2-Archtunnel inkoreahas
been rapidly increasing in hard rock. So, researches
of 2-Arch tunnel in hard rock are urgently needed.
The stability of center upper part of 2-Arch tun-
nel is weak comparing with other parallel tunnels;
therefore it is necessary to install pillar and secure
structural stability. In2-Archtunnel, mutual effect of
precedenceand afterward tunnel is bigger than gen-
eral parallel tunnel. Thestability of pillar andmutual
effect are very important in research of the 2-Arch
tunnel.
Therefore, thisstudyisconductedtomodel charac-
teristic of sedimentary rock that similar withground
conditionof prototypetunnel. Stepindifferent stage
displacement tendency, precedencetunnel stresstran-
sition by excavateafterwardtunnel, pillar andlining
behavior examinebyachievinganexperimentaccord-
ingtocarryingoutconstructionstepsof 2-Archtunnel.
Withthis, comparewithnumerical analysis(DEM) in
sameconditionandconfirmthereliabilityof results.
2 SUMMARY OF EXPERIMENT
2.1 The model package
Thedimensionsof themodel groundwere3m3m
0.27m(BHT).Themaincharacteristicof thetest
machineisapplicationof boththevertical andlateral
loadingsystemtoembodyin-situstress.
Figure1. Largemodel test machine.
669
Table1. Propertyof model ground.
Block J oint surface
E C C
(MPa) (KN/m
3
) (MPa) (
) (MPa) (
)
980 0.25 19.8 2.55 35 0 32
Theinsideof themodel container wascoatedwith
greasetoreducesidefrictionsothatshearstresstrans-
fer at thegroundandthecontainer interfacecouldbe
minimized.
2.2 The model ground
It has been well known that rock mass is not homo-
geneous and exists discontinuities. Thebehaviour of
tunnelsandsurroundinggroundareheavilydependent
onthecharacteristicof therockmass. Inthisstudyto
model characteristic of sedimentary rock, themodel
groundwasmadeof anumber of concretebricks. The
model ground was classified by Rock Mass Rating
(RMR) used in tunnel design in Korea widely. The
model ground used in the model tests had slightly
roughsurfacewithaperturethicknesslessthan1mm.
Overall themodel groundhadtotal RMR ratingof 67
and hence the model ground was classified as class
number II, i.e., Goodrock.
2.3 Tunnel cross section and stiffness
reduction rate
There is a problemin problemin passing of equip-
ment andconstruct of pillar, becausecenter tunnel is
small and narrow(Fig. 2). Therefore, an experiment
achievedintoimproved2-Archtunnel sectionthatcon-
siders equipment exit and entrance center tunnel of
approximatelyenlarge, andimprovewaterproof sheet
establishmentlocationandcarryingoutorder (Fig. 3).
Usually, theproperty of thematerial and sizeare
decidedbythroughthesuitablestiffnessreductionrate
when achieve a model experiment. The property of
model tunnel wasdecidedbythemethodof Duddeck
andErdmann(1985) that usestiffnessratio() of lin-
ing. Actuality stiffnessratioof researchobject tunnel
andmodel tunnel dependedonstiffnessof eachground
andpropertyof tunnel lining.Thoseareexpressedwith
belowway. Thereduction rateapplications aresame
withtable2.
Stiffnessratioof prototypetunnel
: =(E
k
R
3
)/(E
b
I
b
)
Stiffnessratioof model tunnel
: =(E
km
R
3
m
)/(E
bm
I
bm
)
E
k
(E
km
): Modulusof elasticityof ground(MPa)
R(R
m
): Diameter of prototype(model) tunnel (cm)
Figure2. Theexist 2-Archtunnel cross-section.
Figure3. Improved2-Archtunnel cross-section.
Table2. Applicationof reductionratio.
E(ground) E(lining) Diameter Thickness
Tunnel (MPa) (MPa) (m) lining(cm)
Prototype 9.8*10
3
1.96*10
5
23 40
tunnel
Model 9.8*10
3
1.96*10
5
1.2 0.6
tunnel
Figure4. Measurepoint of grounddisplacement.
E
b
(E
bm
): Modulusof elasticityof lining(MPa)
I
b
(I
bm
): Secondmoment of arealining
t(t
m
): Thickness of prototype(model) tunnel lining
(cm)
2.4 Instrumentation
Inthemodel test thebehaviour of thetunnel andthe
groundduetoexcavationwasmonitoredusingLVDT
(Linear Variable Displacement Transducer, ground
displacements) and load cell (pillar load). Theloca-
tionsof LVDTsareshowninFigure4.
670
Figure5. Thetest procedures.
2.5 The center pillar
Inorder tomeasuretherockloadactingonthecentre
pillar, load cell was installed prior to the formation
of the model ground. In the real tunnel excavation,
thecentrepillar is tobeinstalledafter completionof
thepilot tunnel. However, tomeasurelooseningload
duringthepilottunnel excavationthecentrepillarwas
installed prior to theassembly of themodel ground.
Thereafter the pillar load has been arranged to zero
tomeasurerock load, exclusively associatedwiththe
maintunnel excavation.
2.6 Experiment condition and method
Themodel testwasconductedunderauniformsurface
loadingof 980kPa.Anachievedexperimentconsiders
actual constructionstepasfollows.
1. Madeuphorizontality beingstratiformrock mass
bytheuseconcreteblockof fixedsize.
2. The model test was conducted under a uniform
surfaceloadingof 980kPawithlateral earthpres-
sure of 1470kPa and hence the earth pressure
coefficient was1.5.
3. Excavationof upper part of thepilot tunnel.
4. Excavationof lower part of thepilot tunnel.
5. Arrangement of loadcell readingtozero.
6. Excavationof precedencetunnel.
7. Excavationof afterwardtunnel.
3 THE RESULTSOF THE EXPERIMENT
This research is that investigate behavior of 2-Arch
tunnel that is conducted through a large model test
andDEM analysisfromthestratifiedrock. Themain
examinationcontentsaresameasfollows.
1. Displacement tendencyat different steps
2. Precedence tunnel stress transition by afterward
tunnel excavation
3. Estimateof behaviorof centerpillarandloosenarea
rockload.
3.1 Displacement tendency of different steps
Thegrounddisplacementisthemainfactor for tunnel
and adjacent structures stability judgment by tunnel
excavation. The displacements that appear in exper-
iment results are compared with numerical analysis
(DEM) in equal condition and analyzed. Boundary
conditionof thisnumerical analysisset limitstohor-
izontality displacement in side wall and lower part
set limits to perpendicular displacement. Input data
of numerical analysisisdecidedbyaxial compression
test, directsheartestandRMRvalueof model ground.
3.1.1 Displacement concentration region
In the model test, most of the ground displacement
wasexaminedbylargemodel testandnumerical anal-
ysisoccurredwithin0.25D. Experimentconsequences
andnumerical analysisconsequencesof displacement
werevariancemoreor less.
This is judgedby joint actionthat is happenverti-
cally joint betweenblock andblock that canproduce
cause and experiment special quality upper that did
671
Figure6. Displacement concentrationregion.
not accordwithcorrectlypropertiesvalueof actuality
model groundinnumerical analysis.
But, a model experiment and numerical analysis
result of displacement concentration extent and ten-
dency aresimilar, thereforeinidentical branchresult
wasjudgedthatitismeaningthatcomparemutual and
analyzeddisplacement todifferent steps.
3.1.2 Horizontal displacement of different steps
Inordertoevaluatestabilityof thesidewall, thelateral
ground displacements at the right part of the tunnel
havebeen monitoredusingLVDTs. Thedirection of
displacementispositivetotunnel anddisplacementof
oppositedirectionisnegative.
a. Excavationupper part of thepilot tunnel.
b. Excavationlower part of thepilot tunnel.
c. Installationof pillar.
d. Excavationof theprecedencetunnel.
e. Excavationof theafterwardtunnel.
Duringexcavateto pilot tunnel, most of displace-
ment occurred. Ingeneral thestress conditionof the
groundafter tunnel excavationmay bedifferent from
thein-situcondition. Inparticular, directionandmag-
nitudeof theprinciplestresseswill bechangeddueto
stressre-distributionthroughthearchingeffect, hence
themeasuredresults areindicatethat majority of the
Table3. H-displacement of different steps(mm).
Horizontal displacement (mm)
Measure
point a b c D e
Large 2 1.49 1.83 1.85 3.21 3.38
model 4 1.36 1.62 1.69 2.18 2.31
test 29 1.4 1.54 1.55 2.02 2.17
30 1.14 1.33 1.33 1.87 1.94
UDEC 2 1.943 3.578 3.577 7.127 7.349
4 1.423 2.254 2.254 5.203 5.382
29 1.617 2.344 2.344 4.984 5.141
30 1.231 1.618 1.618 3.280 3.387
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
initial a b c d e
H
o
r
i
z
o
n
t
a
l
l
y
d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
(a) H-displacement of different steps
point 2 point 4 point 29 point 30
H
o
r
i
z
o
n
t
a
l
l
y
d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
(b) Completion of tunnel excavation
modeltest
UDEC
M-T No.2 M-T No.4 M-T No.29 M-T No.30
UDEC No.2 UDEC No.4 UDEC No.29 UDEC No.30
Figure7. H-displacement.
stress changeoccurredduringstagebafter thestress
conditionmay not bechangedmuch. Also, Themea-
sured results show that thelonger thedistancefrom
thetunnel basethesmallerthelateral grounddisplace-
ment.(measurepoint 2,4s horizontal displacement >
measurepoint29,30shorizontal displacement). Over-
all themeasurements suggest that thestability of the
2-arch tunnel depends mainly on the excavation of
the pilot tunnel. Therefore, the pilot tunnel should
bestabilized prior to excavation of theother part of
thetunnel. A largemodel experiment and numerical
analysisresult issamewithtable3.
3.1.3 Vertical displacement of different steps
Inordertomeasurevertical grounddisplacementsdur-
ing the tunnel excavation at different steps, ground
672
Table4. V-displacement of different steps(mm) M.T.
Measure
point a b c d e
7 0.25D 1.05 1.53 1.55 1.84 2.33
13 0.50D 0.82 1.21 1.16 1.38 1.89
8 0.25D 1.24 1.66 1.67 2.01 2.99
14 0.50D 0.99 1.25 1.35 1.59 2.38
9 0.25D 1.32 1.72 1.74 2.52 2.74
15 0.50D 0.99 1.31 1.39 1.94 2.11
10 0.25D 1.2 1.59 1.59 1.94 2.21
16 0.50D 0.86 1.24 1.3 1.51 1.81
Table5. V-displacement of different steps(mm) UDEC.
Measure
point a b c d e
7 0.25D 0.95 1.44 1.45 1.75 4.39
13 0.50D 0.95 1.63 1.63 1.95 4.26
8 0.25D 1.43 2.01 2.01 2.60 5.89
14 0.50D 1.29 2.05 2.05 2.67 5.41
9 0.25D 1.8 2.43 2.43 3.99 6.14
15 0.50D 1.43 2.22 2.22 3.66 5.63
10 0.25D 0.95 1.44 1.44 3.26 4.09
16 0.50D 0.97 1.65 1.66 3.09 4.05
-7
-6
-5
-4
-3
-2
-1
0
point 7,13 point 8,14 point 9,15 point 10,16
v
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
modeltest 0.25D
modeltest 0.50D
UDEC 0.25D
UDEC 0.50D
Figure8. V-displacement (completionof excavation).
displacements at 0.25D and 0.5D above the tunnel
crownhavebeenmonitored(0.25D: points7100.5D:
points 1316), where D is the width of the tunnel
(Table4).
Themajorityof thedisplacementsoccurredatexca-
vation of pilot tunnel after insignificant changes of
the ground displacement had developed. The Verti-
cal displacementin0.25Dwasobservedgreatly120%
when compare with displacement in 0.5D and most
displacement wasconcentratedon0.25Dextent.
Horizontal displacementsmorethan40%of whole
displacements and vertical displacements more than
20% are occurred during excavation of pilot tun-
nel. The displacement of model test is larger than
numerical analysisresult(Table5)atpilottunnel exca-
vation steps. The displacement was concentrated on
-7
-6
-5
-4
-3
-2
-1
0
initial a b c d e
v
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
(a) 0.25D from upper part of tunnel
-6
-5
-4
-3
-2
-1
0
initial a b c d e
v
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
(b) 0.5D from upper part of tunnel
Figure9. V-displacement of different stages.
0.25Dextent. Suchresultmeansthatstabilityof whole
tunnel is dominatedby stability of pilot tunnel exca-
vationin2-Archtunnel. This suggests that rock bolt
lengthshouldbelongerthan0.25Dtopreventtherock
loosening.
3.2 Precedence tunnel stress transition by
afterward tunnel excavation
Precedencetunnel stability by afterwardtunnel exca-
vationismainconcernstoestablishment2-archtunnel
that parallel tunnel isvery near. Inthecaseof prece-
dence tunnel, displacement is converged at comple-
tionof excavationandestablishment of support. But
ground displacement and additional loading to sup-
port is increased by excavation of afterward tunnel.
Additional load that happen to precedence tunnel at
afterward tunnel excavateis 1030%of wholeload
to the tunnel. (Lee and kim 2001, The considera-
tionof improve2-Archtunnel Designandconstruction
method).
At the large model test result, displacement of
precedencetunnel is occurred812%of wholedis-
placement by excavation of afterward tunnel. And
2035% of UDEC analysis result increased at the
samepoint. It isconsideredsimilar result if takeinto
673
2.74
2.21
3.26
3.99
1.94
2.52
4.09
6.14
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
M-T No.9 M-T No.10 UDEC No.9 UDEC No.10
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
excavation completion of
precedence tunnel
excavation completion of
afterward tunnel
Figure10. Displacement of precedencetunnel by excava-
tionof afterwardtunnel.
account difference in behavior of joint that happen
in experiment and joint properties value that apply
in analysis. This is similar with suggested result by
Lee, kim. This result means that effect to support of
precedencetunnel byexcavationof afterwardtunnel.It
canbecomeprobleminstabilitysecureof precedence
tunnel or support amount excess of afterwardtunnel
that suchconduct has precedencetunnel support and
afterwardtunnel supportsequally. Thereforehereafter
designsneedtoestablishsuitablesupportpatternplan
inconduct of 2-Archtunnel.
3.3 Behavior of the center pillar
Matsuda (1998) proposed an empirical equation of
rock load for a tunnel in soils and weathered rocks
by considering soil depth (H) and tunnel centre-to-
centredistance(B). However, to datestudies of rock
loadthat canbeusedinpreliminarytunnel designfor
2-archtunnel inrock is rather limited. Inthecurrent
study, anempirical relationfor rockloadbasedonthe
measurement is proposedfor therock whenRMR is
greater than60.
Theloadwidthact onpillar isdistanceapart two
tunnel. (W)
Whendistanceof tunnel togroundsurface(H) is
longer thantunnel width(D)
P= DW
: unit weight of ground
Whendistanceof tunnel width(D) islonger than
tunnel togroundsurface(H)
P= HW
1. Initial stage
2. Excavationcompletionof upperpartof pilottunnel.
3. Excavationcompletionof lowerpartof pilottunnel
4. Loadconvergence
5. Excavationcompletionof precedencetunnel
6. Excavationcompletionof afterwardtunnel
Figure11. Theloadactingonpillar of 2-Archtunnel insoil
andweatheredrocks(Matsuda, 1998).
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
a b c d e f
T
h
e

l
o
a
d

o
f

a
c
t

o
n

p
i
l
l
a
r
(
K
N
)
Installation of Pillar
Figure12. Changeof theloadactingonpillar at different
steps.
Table6. Theloadactingonpillar (KN).
a b c d e f
28.6 33.0 26.4 0 13.2 31.2
Themaximumloadactingonpillar of 2-Archtun-
nel is33.0KN(Table6). InFigure12, theloadof until
converge step that act on pillar of 2-Arch tunnel is
supportedbyrockboltandshotcrete.Thefinal loadof
about 31.2KN istheactual rock loadtobesupported
bythecenter pillar. Inthecurrent study, anempirical
relationforrockloadbasedonthemeasurementispro-
posedfor therock whenRMR isgreater than60. The
empirical equationproposedfromthecurrentresearch
isH=0.15W,whereWisthecentre-to-centredistance
between left and right tunnels. Although morestudy
isobviouslyrequired, thisrelationmaybeusedfor the
preliminary designof thecentrepillar for the2-arch
tunnel.
4 CONCLUSIONS
The 2-Arch tunnel is similar to two parallel tunnels
withveryshort centre-to-centredistance. However, it
674
isexecutingwithoutenoughstudiesforbehaviorof the
2-Arch tunnel. In this research, astudy for behavior
of the2-Arch tunnel is examined using largemodel
test machine. Thenembodyin-situstressbyapplying
pressuretoboundaryof model ground.
1. Thegrounddisplacementsmainlyoccurredwithin
0.25D, whereDisthetunnel width. Horizontal dis-
placementsmorethan40%of wholedisplacements
andvertical displacementsmorethan20%occurred
duringexcavationof pilottunnel.Suchresultmeans
that stability of wholetunnel dominatedby stabil-
ityof pilot tunnel excavationinthe2-Archtunnel.
Thissuggeststhatrockboltlengthshouldbelonger
than0.25Dtoprevent therockloosening.
2. Basedonthelimitedinformationobtainedfromthe
measurement, the rock load acting on the centre
pillarmaybeH=0.15W, whereWisthecentre-to-
centredistancebetweenleftandrighttunnelswhen
RMR is more than 60. However, obviously more
research is required to generalize the proposed
relation.
REFERENCES
Amadei, B. & Stephanson, O. 1997. Rock stress and its
measurement, KluwerAcademicPublishers.
Brady, B.H.G. & Brown, E.T. 1985. Rock mechanics for
underground mining, KluwerAcademicPublishers.
Goodman, R.E. 1988. Introduction to Rock Mechanics, J ohn
WileyandSons.
Hoek, E. & Brown, E.T. 1980. Underground excavation in
Rock, Instn. of Mining& Metallurgy.
Hoek, E. 2000. Practical RockEngineering, Coursenote.
Lee, S.D., Choi, S.I. & Gu, J.G. 1994. Design and construc-
tion of stability underground structure.
Lee & Kim. 2001. The consideration of improve 2-Arch
tunnel Designandconstructionmethod.
Lee, I.M. 2001. Principle of rock mechanics.
Matsuda, T. 1998. Groundbehavior andSettlement control
of twintunnels insoil ground. Tunnels and Metropolies:
pp11931198
Matsuda,T.1998.Discussionbehaviorandsettlementcontrol
of twintunnel insoil ground, Tunnels and Metropolises.
Priest, S.D. 1993. Discontinuity Analysis for Rock Engineer-
ing, KluwerAcademicPublishers.
Yun, J.S. 1991. Investigationandtest of rockmass.
675
Behavior of tunnel duetoadjacent groundexcavationunder theinfluence
of pre-loadingonbracedwall
S.D. Lee
Department of Environmental, Civil and Transportation Engineering, Ajou Univ, Suwon, Gyeongi-Do, Korea
I. Kim
Sambon Eng. Anyang, Gyeonggi-Do, Korea
ABSTRACT: Pre-loads couldbeimposedonthebracedwall to prevent its horizontal displacements during
the ground excavation even though a tunnel exists adjacent to the braced wall. New pre-loading systemfor
largeloadswasdevelopedandappliedtothelargescalemodel tests. Model testswereperformedinthesandy
ground, whichwashomogeneouslyandisotropicallyconstructedinthetestpitandnumericallyanalyzedbyFinite
Element Method. It was foundthat thestability of existingtunnel was greatly enhancedwhenthehorizontal
displacementsof abracedwall wasreducedbyapplyingpre-load, whichwaslarger thanthedesignload.
1 INTRODUCTION
Horizontal displacement of abraced wall during the
groundexcavationcouldsoftentherearground, which
couldcausetheexistingstructureintherear groundto
beunstable(Lee, 1999). Pre-loadingmethodiswidely
usedtoreducethelateral displacementof bracedwall.
ORourke (1976) found that lateral displacement
of abraced wall could bereduced by pre-loading in
halfsizeof thedesignforce. But toolargepre-loading
couldjeopardizethestrut. ORourke(1981) alsopro-
posedthat theeffectiverigiditycouldbeincreasedby
imposingpre-load.
To establish the optimum range of pre-loading
in clayey soil, Mana and Clough (1981) performed
numerical analysesandfoundthatdisplacementcould
bereducedby pre-loadingthedesignforce. But they
pointed out that too large pre-loading could cause
local deformationof thesteel joint of thesupporting
structureandcoulddamageit.
Canadian Geotechnical Society (1997) found that
pre-loading which amounted to the design force
should be imposed to reduce the displacement of a
bracedwall.
Moststudiesonthepre-loadinghavefocusedonthe
groundbehavior, theeffectiverigidityof strut, andthe
displacement of abracedwall.
To keep the tunnel in the rear ground to be sta-
ble during the adjacent ground excavation, horizon-
tal displacement of braced wall should be reduced.
Horizontal displacement, however, couldnotbesuffi-
cientlyreducedbypre-loadingthedesignforceandit
isdifficult tobeconvincedof thetunnel stability.
In this paper, behavior of thebraced wall and the
tunnel in rear ground was studied, when pre-loading
wasimposedonabracedwall duringthegroundexca-
vation. For thispurpose, largescalemodel testswere
conductedandnumericallyanalyzed.
2 LARGE SCALE MODEL TESTS
2.1 Summary
Pre-loadingmethodwas verifiedthroughlargescale
model tests. They wereperformed by imposing pre-
loadonastrut duringthegroundexcavationbyanew
pre-loadingsystem,throughwhichlargepreloadcould
impose.
Model tests were conducted in a large test
box (2.0m wide, 4.0m long, 6.0m tall) under 2-
dimensional boundary conditions. Lateral wall of
test box was specially treated to minimize the fric-
tion. Existing tunnel was detached 0.5D from the
bracedwall.
Until excavationwascompleted, wall displacement
waskepttobezero(18stages)bypre-loading.Model
testswithout pre-loadingwerealsoconducted.
2.2 Test ground
For model tests, homogeneous and isotropic model
ground was constructed. Sand was put into test pit
in 0.3mdepth at a time and compacted by a plate
vibrator. Relativedensity of test groundwaschecked
by extracting asampleusing DIN 4021 CoreCutter
(Lee, 1998).
677
Figure1. Model test box.
Physical characteristicsof testgroundforthemodel
tests wereconfirmed through grain sizedistribution
test, specific gravity test, field unit-weight test, and
water content test.
According to theUnified Soil Classification Sys-
tem(USCS), test groundwaspoorlygradedSandSP.
Mechanical properties of test ground were obtained
throughadirect shear test. Internal frictionangle()
was 38
and cohesion (c) was 6.0kPa. Modulus of

elasticity(E) was28,000kPa.
Test resultsareshowninFigure2andTable1:
2.3 Model tunnel and model braced wall
Specifications of themodel tunnel weredetermined
from the rigidity ratio by Duddeck and Erdmann
(1985). Rigidity ratio between the ground and the
tunnel lining for actual and model tunnel was as
follows.
RigidityRatioof Actual Tunnel;
Figure2. Grainsizedistributioncurve.
Table1. Physical propertiesof test ground.
Parameter Value
Max. DryUnitWeight (
dmax
) 16.86(kN/m
3
)
Min. DryUnitWeight (
dmin
) 13.82(kN/m
3
)
DryUnitWeight of Test Ground(
d
) 15.39(kN/m
3
)
RelativeDensity(D
r
) 56(%)
Water Content () 6.8(%)
SpecificGravity(Gs) 2.63
RigidityRatioof Model Tunnel;
Here, subscript m meansmodel tunnel.
whereE
k
(E
km
)=YoungsModulusof Ground(kPa);
R (R
m
)=Radiusof Tunnel (m); E
b
(E
bm
)=Youngs
ModulusLining(kPa); I
b
(I
bm
)=Momentof Inertiaof
Tunnel Lining; and
Thicknessof model tunnel t
m
wasestimatedunder
theassumptionthattherigidityratioof real andmodel
tunnel wasequal (Solimanet al. 1993).
wheret (t
m
)=Thicknessof tunnel lining(m)
Modulusof elasticityof theactual ground(E
k
) was
400,000500,000kPa. But modulus of elasticity of
model ground(E
km
) was20,00030,000kPa. For the
model tunnel liningwiththediameter of 0.6m, 6mm-
thick steel was used, whichwas equivalent to 0.3m-
thick concretelining. For abracedwall, 16mm-thick
steel wasused, whichwasequivalenttoan0.8m-thick
concreteslurrywall.
Hobbs (1966) estimated the time reduction rate
under theassumptionthat theaccelerationof gravity
wasconstant. Reductionrateof lengthwasestablished
first andthenreductionratefor time, density, weight,
678
Table2. Reductionrateof model test.
Reductiontate
(Tunnel lining
Property Dimensions andbracedwall)
Length [L] 1/10
Time [T] 1/3.16
Weight [M] 1/3.120
Density [ML
3
] 1/3.12
Stress [ML
1
T
2
] 1/31.24
Acceleration [LT
2
] 1.0
of Gravity
Figure3. Measuringpointsfor model tests.
andstresswasestimated. Thelawof similarity could
bebasedonweight, time, andacceleration. Steel with
unit weight 78kN/m
3
was usedfor model tunnel lin-
ingandthebracedwall (Yanget al. 2007, Shimet al.
2007).
Table3. Propertiesof ground.
Parameter
Modulusof Elasticity(E) 20,000kPa
PoissonsRatio() 0.25
UnitWeight () 16.39kN/m
3
Internal FrictionAngle() 38
Cohesion(C) 6.0kPa
Figure4. Finiteelement meshandboundaryconditions.
2.4 Measurement
Moment of braced wall (40 points), lateral dis-
placement (8points), rear ground surfacesettlement
(9points), liningmoment (32points), andtunnel dis-
tortion (16 points) was automatically measured and
savedatevery30minutes.Measuringpointsareshown
inFigure3andTable3.
3.1 Analysis program
Tunnel was detached 0.5D from the braced wall.
TestswerenumericallyanalyzedusingFEM program
PLAXISVer. 8.2.
Triangular planestrainelementswithfifteennodal
pointswereusedfortheground. Interfacewasadapted
ontheboundaryat thebracedwall andtheground.
Finiteelement mesh and boundary conditions for
thenumerical analysisareshowninFigure4.
Vertical displacement at the lateral boundary and
horizontal displacement at the lower boundary were
possible.
679
Table 4. Material properties for braced wall and tunnel
lining.
Parameter Bracedwall Tunnel lining
Axial Rigidity(EA) 3,293,000kN/m 1,929,000kN/m
Flexural Rigidity(EI) 70.0kNm
2
/m 5.788kNm
2
/m
Thickness(d) 0.016m 0.006m
PoissonsRatio() 0.3 0.3
Figure5. Horizontal displacement of bracedwall.
3.2 Input data
Properties of ground and structuremembers used in
numerical analysisareshowninTables3and4:
4 RESULTSOF LARGE SCALE MODEL
TESTSANDNUMERICAL ANALYSES
4.1 Horizontal displacement of braced wall
Horizontal displacement of abracedwall inthelarge
scale model tests and numerical analyses is shown
inFigure5. At thefinal excavationstage, maximum
lateral displacement (max) of bracedwall decreased
from2.290mmto 0.03mmby preloading, whilethe
resultof numerical analysisdecreasedfrom2.914mm
to0.193mm(Fig. 5).
4.2 Tunnel distortion
Distortion of tunnel lining is shown in Figure 6.
Without preloading, tunnel distorted gradually due
to ground excavation. Tunnel was pressed vertically
andexpandedhorizontally. Bypre-loadingtunnel dis-
tortiongreatly decreased. Maximumdisplacement of
thetunnel liningdecreasedfrom2.22mmto0.21mm
in the tests and from2.670mmto 0.337mmin the
numerical analysis(Fig. 6) bypre-loading.
4.3 Member stress of tunnel lining
Bypre-loading, maximumflexural momentdecreased
from 0.15kNm/m to 0.068kNm/m and maximum
Figure6. Distortionof tunnel lining.
shearforcedecreasedfrom1.160kN/mto0.436kN/m.
Axial force, however, increased from 17.436 to
22.837kN/minthenumerical analysis(Fig.6).Max-
imumflexural moment at thefinal excavation stage
decreased from 0.195kNm/m to 0.021kNm/m
by pre-loading while the shear increased from
0.331kN/mto0.178kN/m. Maximumaxial force
increasedfrom34.377kN/mto124.863kN/m, in
themodel tests.
In the numerical analyses, on the contrary, by
preloading, maximum flexural moment decreased
from0.150kNm/mto0.068kNm/m, maximumshear
decreased from 1.160kN/m to 0.436kN/m. Maxi-
mum axial force, however, slightly increased from
17.463kN/mto 22.837kN/m, which was in an
allowablerange. Momentof thetunnel liningisshown
inFigure7.
4.4 Lateral displacement of braced wall and
tunnel distortion
Withoutpre-loading, thelargestdisplacementandtun-
nel distortion were observed in the 6th excavation
stage, inthetunnel level.
Significant horizontal displacement of bracedwall
and tunnel distortion would not occur throughout
theentireprocess, whenpre-loadingwasimposed. If
bracedwall was not deformedlaterally, tunnel shape
wouldhardlychange(Fig. 8).
680
Figure7. Bendingmoment andstressof tunnel lining.
Figure8. Lateral displacement of bracedwall.
Figure9. Rear groundsurfacesettlement.
4.5 Ground surface settlement
By pre-loading maximumground surfacesettlement
at final excavation stage decreased from2.1mmto
0.12mmin the model tests and from2.506mmto
0.726mm in the numerical analysis. Rear ground
surfacesettlement isshowninFigure9.
5 CONCLUSION
It was investigated how to ensure the stability of
the adjacent tunnel during the ground excavation
by reducing thehorizontal displacement of abraced
wall by gradually imposing pre-loading. A new pre-
loading system, through which large pre-loading
could impose was developed and applied to large
scale model tests. Performed tests were numerically
analyzed.
Theresultsareasfollows:
1 If lateral displacement of the braced wall was
greatly reduced by imposing pre-loading, which
was larger than design force, the stability of the
adjacent tunnel at the rear ground would greatly
increase.
2 Maximumflexural momentandmaximumshearof
tunnel liningweredecreasedbuttheaxial forcewas
increased by imposing pre-loading in both large
scalemodel testsandnumerical analyses.
3 Largest horizontal displacement of a braced wall
occurred in the excavation stage, which was the
tunnel level, whenpre-loadingwasnot imposed.
4 Tunnel was pressed vertically and expanded hor-
izontally during the ground excavation without
preloading.Butbyimposingpre-loadingnosignifi-
cantdistortioninitsshapewasobservedthroughout
theentireexcavationprocess.
5 Most of the ground surface settlement was
decreasedbyimposingpre-loading.
681
REFERENCES
CanadianGeotechnical Society. 1997. Foundation Engineer-
ing Manual. 3rd. ed.
Duddeck, H. &Erdmann, J. 1985. Onstructural designmod-
els for tunnels in soft soil. Underground Space, Vol. 9,
PergamonPress; 246259.
Hobbs, D.W. 1966. Scale model study of strata movement
around mine roadways. Apparatus, technique and some
preliminaryresults, Int. J. of Rock Mech. Min. Sci., 3.
Lee, S.D. 1998. Soil Mechanics(2nd. edition). Saeron Press;
353354.
Lee, S.D. 1999. Foundation Engineering. Saeron Press;
252254.
Lee, S.D. 2003. Understanding of the Latest Tunneling
Technology. Ajou University Geotechnical Engineering;
122.
Mana,A.I. &Clough, G.H. 1981. Predictionof movementfor
bracedcutsinclay. J. Geotech. Engineering. Div., ASCE,
vol. 107, No. 6; 759778.
ORourke, T.D., Cording, E.J. & Boscardin, M. 1976. The
Ground movements related to braced excavation and
their influenceonadjacent buildings. U.S Department of
Transportation, Report no. DOT-TST 76, T-23.
O Rourke,T.D. 1981. GroundMovementsCausedbyBraced
Excavation. Journal of Geotechnical Engineering Divi-
sion. ASCE. Vol. 107. NO. GT9; 11591178.
Shim, H.J. and others. 2007. Model Test andTest Blasting
toDesignAdjacentTunnel. KTA 2007 Annual Conference
April 2021 Seoul Korea; 267278.
Soliman, E., Duddeck, H. & Ahrens, H. 1993. two-and
three-dimensional analysisof closelyspaceddouble-tube
tunnels. Tunnelling and Underground Space Technology,
Vol. 8, No. 1; 1318.
Yang, H.S. andothers. 2007. A Study onCharacteristics of
Materials for Scaled Model Test. Ministry of Construc-
tion &Transportation R&D Performance Forum Journal;
3336.
682
Twodistinctiveshear strainmodesfor pile-soil-tunnellinginteraction
inagranular mass
Y.J. Lee
Soil-Structure Interaction Group, Steel Structure Research Laboratory, Research Institute of Industrial Science &
Technology (RIST), South Korea
C.S.Yoo
Department of Civil & Environmental Engineering, Sungkyunkwan University, South Korea
ABSTRACT: Therearemanyhigh-risebuildingsinurbanareas, whicharenormallysupportedbypiledfoun-
dations. Consequently, amajor concernfor geotechnical engineersistheconstructionof atunnel adjacenttothe
piledfoundations, sincegroundbehaviour betweenanexistingloadedpileandtunnellinghasnot beenunder-
stoodwell sofar, particularly for granular soils rather thanclay soils. Inorder tofigureout suchcomplicated
groundbehaviour, theinteractiveshear strainpatternsaregeneratedbytwo-dimensional laboratorymodel tests
andfiniteelement analyses (FEA). For themodel testing, amulti-sizedaluminiumrodmixtureconsideredas
acontinuumgranular massandcloserangephotogrammetrictechniquefor obtainingdisplacement datawithin
therods areintroduced. Twodistinctiveshear strainmodes, viz. connectiveandisolatedmodes, arepresented
throughthecomparisonof themodel testsandFEA accordingtothepiletiplocations.
1 INTRODUCTION
Tunnel excavation work in thesoft ground results in
significant reductionintotal stressinthevicinityof a
tunnel boundary.Thereductioninstresscausesground
movementswhichaffectadjacentbuildingfoundations
consistedof arowof loadedpiles.
In order to identify the pile-soil-tunnelling inter-
action behaviour, small-scale physical model tests
andnumerical methods wereemployedinthis study.
Careful assessment of the pile-soil-tunnelling inter-
action problems is relatively new and only limited
informationiscurrentlyavailable.
This study focuses on two distinctiveshear strain
modes of thegroundbetweentheexistingpileanda
newtunnel constructionasshowninFigure1.Thelab-
oratory model tests wereconductedby straincontrol
ratherthanstresscontrol,andthencomparisonwiththe
finiteelement analyseswascarriedout toidentifythe
shear strainmodes. This researchincorporates strain
controlled tests with idealised two-dimensional alu-
miniumrods consideredasagranular mass taken
to very highvolumeloss (upto about 20%) to high-
light the full shear failure formation. Digital image
processingtechniquehasallowedoverall deformation
patterns of ground movements to be obtained from
the analysis of digital images. Detailed shear strain
patterns of thegroundcan beobtainedwhich givea
clear insight into the pile-soil-tunnelling interaction
events.
Themodel pilewasprincipallyanend-bearingpile
wheremostof thepileloadisconcentratedonthepile
tip rather than pileshaft. Thepileworking load was
maintainedconstantduringthetest.Thisworkingload
waschosenbyreferencetoadisplacement-controlled
pile-loadtest(thepileworkingload, 3.6kN, is77%of
theultimatepileload). Inthisstudy, inorder toavoid
complexity of pile loading the influence of lateral
loadingwasnot considered.
2 LABORATORY MODEL TEST
2.1 Test equipment
Thedisplacement-controlledmodel tunnel consistsof
6segments, thetwo ends of asegment beingcarved
ontaperedcones. Eachsegment movesinwardasthe
tapersarewithdrawn, simulatingthetwo-dimensional
volumelossduringtunnellingoperations. Thetunnel
diameterisreducedbyrotatingthetwoknobsasshown
inFigure2.Theouterdiameterof thetunnel isinitially
100mm. Thereduction of thetunnel diameter gives
directly a2Dvolumeloss(V
L
). This2Dvolumeloss
perrevolutionisdeterminedfromthecalibrationresult
(Lee, 2004).
683
Figure1. Schematic illustration of shear strain modes for
thepile-soil-tunnellinginteraction.
Figure2. Diameter reductionsystemof model tunnel.
Figure3. Multi-sizedaluminiumrodmaterial.
Figure4. Identificationof piletiplocation.
Thealuminiumrodmixtureconsists of six differ-
ent diameters(viz. 2mm, 3mm, 6mm, 9mm, 12mm
and 20mm), which havethesamelength of 75mm.
Itrepresentsawell graded, idealisedtwo-dimensional
granularmaterial, asshowninFigure3.Thetestcham-
berisrigid, arectangularsteel frame(width: 1058mm,
height: 930mm). Themodel pile(25mm75mmin
crosssection,embeddedlength,L:370mm)ismadeof
aluminiumalloy. Nosignificant effectsof thebound-
aryconditionswereobservedduringthetest, sincethe
smallestrods(2mmand3mmdiameters) weremainly
usedintheinteractiveregionsandthelargestrodswere
usedinthevicinityof thesteel frameboundaries.
2.2 Pile-soil-tunnelling interaction test
Thepiletipidentificationsystemadoptedwastolabel
thedistancesof thepiletipawayfromthetunnel centre
line(O) asA (1d
0
: model tunnel diameter), B (1.5d
0
),
C(2d
0
),D(3d
0
)andthepiletiplevel as0(onthecentre
line), +1(onthecrownlevel), +2(at 1d
0
abovethe
centrelinelevel), 1and2beingsimilarlybelowthe
centrelinelevel. Onefinal location(O+2) waswith
thepiledirectlyabovethecentrelineof thetunnel on
the+2level (Figure4).
684
Figure 5. Set-up stage with test equipment and reflective
target points.
Thepilewasinplaceandloadedbeforeanyof the
pile-soil-tunnelling interaction occurred. The model
tests, therefore, simulatednewtunnellingadjacent to
arowof existingloadedpiles.
Eachmodel test normallyconsistedof threestages
as follows: (1) set-up stage, material compacted as
before, withreflectivenodesshowninFigure5(both
pileandtunnel installedinpositionsbut withnopile
loading and tunnel at initial diameter, d
0
); (2) pile
loadedto workingloadby deadweight (P
w
fromthe
pile-load test); (3) tunnelling stage(reduction of the
tunnel diametertoamaximumvalueof 12revolutions,
i.e. V
L
=18.65%).
The tests using aluminiumrods in this study are
much quicker to carry out in terms of testing time
than any conventional test using a real soil such as
sand or clay. In addition, detailed observation of the
sidesectioncaneasilybecarriedout, enablingdetails
of thefailuremechanisms, whichareassociatedwith
strainfieldsrather thanstressfields, tobeexamined.
3 CLOSE RANGE PHOTOGRAMMETRY
Thecloserangephotogrammetrictechniquefor deter-
mining strains used in this study has recently been
appliedtoanumberof geotechnical engineeringprob-
lems. A Kodak DC 290 digital camera was used to
captureboththeframe(or chamber) referencepoints
and the target points fixed to the centre part at the
ends of the smaller size rods within the multi-sized
rodmatrix. Thedigital cameracanprovideacceptably
highresolution.Tenindependentimagesweretakenat
eachepoch(or stage) usingtheKodakDC290digital
cameraasshowninFigure6.
Figure6. Typical digital camerapositionsfor capturing10
images.
A pixel resolution of 17921200 was available.
Eachretro-reflectivetargetwasidentifiedandtheposi-
tionof itsimagewithineachof thetenphotographswas
measuredbytheVMS(visionmetrologysystem) pro-
gram. Themeasured2Dx-y coordinatesof thetarget
points fromtheVMS are arranged into a triangula-
tionmeshbymeansof theEngVisprogram. Figure7
shows an exampleof theimaging processing by the
VMSandthesubsequent triangulationbytheEngVis
respectively.
Thesystemand its application weredescribed by
Woodhouse et al. (1999), Woodhouse (2000) and
Kwok and Swajani (2001). Fromthe measured dis-
placements, strains were calculated based on the
assumptionof linear strainineachtriangular element
(LeeandBassett, 2006). Moredetailedmodel testpro-
ceduresmatchedwiththeimagecapturingstagesare
showninFigure8.
Inordertocarryoutthefiniteelementanalyses(FEA),
the 3D pile-soil-tunnelling interaction situation was
idealised to two-dimensional plane-strain conditions
andtocomparewiththemodel pile-soil-tunnellingtest
result, thesametunnel geometry, pilesize, andloca-
tionof thepilerelativetothetunnel wereadoptedfor
theFEA. TheFE analyseswerecarriedout usingthe
continuumfiniteelement programCRISP (Brittoand
Gunn, 1987; WoodsandRahim, 2001).
Groundbehaviour wasassumedtobegovernedby
an elastic-perfectly-plastic constitutive model based
on the Mohr-Coulomb failure criterion with a non-
associatedflowrule.Thecritical stateangleof friction
(
cs
) andtheangleof dilation() weredeterminedto
be23
and15
forthemodel granularmaterial, respec-

tively (based on a best FEA fit of load-settlement
relationship for the model pile loading tests, Lee,
2004). It shouldbenotedthat later theseparameters
were also obtained fromshear box tests (area void
685
Figure 7. Digital image analysis by VMS and EngVis
programs.
ratio, e=0.340.38,
cs
and evalues arefound to
besimilar to thosegiven byYamamoto and Kusuda,
2001). Parameter valuesfromtheshear boxtestswere
comparabletothebest fit values. Theeffectivecohe-
sion(c
) was assumedtobe0.1kPa. Thevariationof

Youngsmodulus(E) wasassumedtoincreaselinearly
withdepth. Thegroundparametersaresummarisedin
Table1.
Thetunnel supportandpilewasmodelledasatwo-
node bar element and linear elastic material respec-
tively. TheYoungs modulus and Poissons ratio for
boththetunnel support andpilewereassumedto be
15.5GPaand0.2, respectively. Theunit weight of the
pileand thecross section areaof thetunnel support
ring were 23kN/m
3
and 0.003m
2
, respectively. The
model parameters for thetunnel support andthepile
aresummarisedinTable2.
Figure8. Relationshipbetweenmodel teststepsandimage
capturingstages.
Table1. GroundparametersusedintheFEA.
Groundsurfacelevel, Y
0
(m) 0.72
Youngsmodulusat Y
0
, E
0
(kPa) 1600
Gradient of Youngsmodulus, m
E
(kPa/m) 10,000
Poissonsratio, 0.35
Unit weight of soil, (kN/m
3
) 24
Effectivecohesion, c
(kPa) 0.1
Critical stateangleof friction,
cs
(degrees) 23
Angleof dilation, (degrees) 15
Note:
frombottomtotopof mesh;
varyingwithdepth
Table2. Parametersfor tunnel support andpile.
Parameters Tunnel support Pile
E (GPa) 15.5 15.5
0.2 0.2
Unit weight, (kN/m
3
) 23
Crosssectionarea, A (m
2
) 0.003
FNR (Full Newton-Raphson) iterative solution
schemewasadoptedtogether withatoleranceof 0.05
and a maximumiteration number of 100. K
0
(0.66)
wasappliedastheinitial in-situstressconditions.Dou-
bleconvergencecheckbasedonboththedisplacement
andforcenormsandatotal of 1055incrementswere
used. Itshouldbenotedthatthelargest2Dvolumeloss
values(7.63%to18.65%) weregeneratedinorder to
capturetheinteractionfailurepatternsbetweenthepile
andthetunnel.
686
Figure9. Comparisonof maximumshear straincontoursat
V
L
=3.94%.
5 RESULTS
Figures 9 and 10 show the maximum shear strain
contours of the model tests at V
L
=3.94% and
V
L
=10.94%respectively. Thelocationandintensity
of shear strain clearly identify the developing shear
failureformation between thepilebaseand thetun-
nel. It is notedthat thecaseof O+2was omittedin
thiscomparison(thiscaseisnot anormal practicefor
consideringaproper tunnel positionadjacenttoarow
of loadedpiles).
A clear neutral or deadblock X anareawithlow
or no strain was observed clearly in caseC+1 at
10.94%of volumelossasshowninFigure10.
Figure10. Comparisonof maximumshear straincontours
atV
L
=10.94%.
The shear failure formation appeared to com-
prisetwo distinctiveshear strainmodes: (1) onethat
includesaneutral soil block X, separatingaforma-
tionrunningfromthepilebasetothetunnel invertarea
andasecondmechanismrunningfromthepileshaftto
thetunnel crownarea, and(2) theother onethat isan
independent shear behaviour mode, i.e. nointeractive
shear strainmodebetweenthepileandthetunnel. In
summary, thetwodifferent shear strainmodesshown
in Figure 1 can be identified according to pile tip
locationsasshowninFigure11.
687
Figure 11. Boundary of two different shear strain modes
accordingtopiletiplocations.
Similar shear strainmodebehaviour wasobserved
atbothsmall andlargemagnitudesof thevolumeloss.
While the model tests and FEA are two-
dimensional, it is expected that similar behaviour
wouldbefoundinreal projects wherethetunnelling
presents a three-dimensional problem. However, the
trueboundary of two different shear strainmodes in
thelatter may bedifferent fromtheformer tests and
analyses.
6 CONCLUSIONS
Comparisonbetweenthephysical model testsandthe
finiteelementanalysesshowedmanysuccessful points
of agreement interms of shear straindata. Basedon
themaximumshear strain data, it was observed that
theshear strainmodesdevelopedarestrongly depen-
dent onthepiletiplocationandthemagnitudeof the
volumeloss. Throughthisstudy, it isrecognizedthat
theboundaryof twodifferent shear strainmodesmay
be a useful guide for the tunnel planners who need
tomakeadecisionontheproper positioningof tunnel
constructionadjacenttoarowof loadedpilesinurban
areas.
REFERENCES
Britto, A.M. & Gunn, M.J. 1987. Critical state soil mechan-
ics via finite elements. Chichester, UK: Ellis Horwood
Limited.
Kwok, H.Y. & Swajani, C. 2001. Precise measurement tech-
nique. 3rd year project report, Department of Civil and
Environmental Engineering, UniversityCollegeLondon,
Universityof London.
Lee, Y.J. 2004. Tunnelling adjacent to a row of loaded piles.
PhD Thesis, University College London, University of
London.
Lee, Y.J. & Bassett, R.H. 2006. Application of a pho-
togrammetrictechniquetoamodel tunnel.Tunnelling and
Underground Space Technology 21(1): 7996.
Woodhouse, N.G. 2000. Geometric models appropriate for
engineering analysis from vision metrology data. PhD
thesis, UniversityCollegeLondon, Universityof London.
Woodhouse, N.G., Robson, S. & Eyre, J.R. 1999. Vision
metrology and three dimensional visualisation in struc-
tural testing and monitoring. Photogrammetric Record
16(94): 625641.
Woods, R. & Rahim, A. 2001. SAGE-CRISP Techni-
cal Manual, Version 4. The CRISP Consortium Ltd.
http://www.mycrisp.com/demo/TECHMAN.pdf
Yamamoto, K. & Kusuda, K. 2001. Failuremechanismsand
bearingcapacitiesof reinforcedfoundations. Geotextiles
and Geomembranes 19(3): 127162.
688
Stabilityanalysisof largeslurryshield-driventunnel insoft clay
Y. Li & Z.X. Zhang
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, P.R. China
INSA-Lyon, LGCIE, France
ABSTRACT: Thepossibilityof partial failureof largeslurryshield-driventunnelsisinvestigatedbyanupper-
boundapproachinlimit analysisandathree-dimensional numerical modelingfor theShanghai YangtzeRiver
Tunnel. Theresultsof theupper-boundlimit analysisfailuremechanismsandthe3Dnumerical modelinghave
shownthat thepartial blow-out of theupper part of thetunnel faceoccurswhentheslurrypressureistoolarge
whiletheglobal collapseof thewholetunnel faceoccurs when theslurry pressureis too small. Thefailure
mechanismsandcritical slurrypressuresobtainedfrombothapproachesarepresentedanddiscussed.
1 INTRODUCTION
In recent years, the rapid growth in urban develop-
ment has resulted in an increased demand for the
construction of tunnels for electric and communica-
tion cables, and transportation systems. For obvious
practical reasons such as accessibility, serviceability
andeconomy, thesetunnelsareconstructedby shield
machinesof largediameter andatshallowdepths.The
GroeneHart Tunnel, constructed in 2005 in Nether-
lands, wascarriedoutbyaslurry-shieldmachinewith
anoutsidediameter of 14.87m. TheM-30Tunnel in
Madridexcavatedby EPB shieldmachine, 15.2min
diameter, was until recently thebiggest shieldtunnel
completedintheworld. InSeptember 2006, twomas-
sive 15.43mdiameter slurry shield machines began
work on Shanghai Yangtze River Tunnel. With the
increaseof theshieldtunnel diameter, theexcavated
volumeis increaseddramatically andtheprobability
of excavation in complicated stratumwith different
typesof soil layersincreasesgreatlytoo. Thestability
of thesoil itself decreases at thesametime. Thus, in
recent years, moreandmoreattentionwaspaidtothe
facestabilityof largeshield-driventunnels.
Theanalysisof thefacestabilityof shallowcircular
tunnelsdrivenbythepressurizedslurryshieldrequires
the determination of the pressure to be applied by
theshield. Thispressuremust avoidboththecollapse
(activefailure) and theblow-out (passivefailure) of
thesoil massnearthetunnel face.A numberof studies
haveconcernedtunnel facestability. Most resultsare
analytical andarebasedonlimit equilibriummethod
(Horn, 1961; Anagnostou & Kovri 1994; Broere,
2001) and limit analysis method (Davis et al. 1980;
Leca & Dormieux, 1990; Chambon & Cort, 1994;
Soubra2000, 2002; Subrin& Wong, 2002). A ratio-
nal and well-defined approach for the computation
of thesupportingpressureistranslational multiblock
failuremechanismof theupper-boundmethod, which
is presentedby Soubra(2002) basedonthe3D limit
analysis model of Leca and Dormieux (1990). This
mechanismallows the slip surface to develop more
freely in comparison with theavailablemechanisms
givenby LecaandDormieux, andthus, improvesthe
bestupper-boundsolutionsgivenbytheseauthors.The
multiblock mechanismis convenient for a constant
tunnel pressurewhichisanacceptableassumptionfor
small tomediumtunnel diameters(10m). However,
studies on face stability of very large slurry shield-
driventunnels, for whichthehypothesisof aconstant
slurrypressureisnot applicable, arefairlyfew.
The non-constant supporting pressure of slurry
shield-driven tunnel is caused by the density of the
slurry. Sincethedensity of theslurry shouldremain
within a certain range to obtain high-quality filter
cakes, anditisalwayssmaller thanthedensityof soil,
therewill beapressuredifferencebetweentheslurry
pressure and the earth pressure at tunnel crown and
invert.Thispressuredifferenceincreaseswiththetun-
nel diameter increase, shown in Figure1, where, S
u
andS
b
aretheslurrypressuresat thelevel of thetun-
nel crownandinvert; P
u
andP
b
arethecorresponding
689
Figure1. Pressuredifferenceindifferent sizetunnels.
Figure2. Geological conditionwithC/D=0.7.
earth pressures; D
0
is thediameter of thesmall tun-
nel and D is the diameter of the larger one. This
pressure difference may induce a different failure
mechanismfromtheonecorrespondingtoaconstant
slurrypressure, especiallyinlargetunnels.
Inthispaper, asimplifiedcomputationschemecon-
sideringthenon-constantslurrypressureisadopted, in
whichthemultiblockfailuremechanismsuggestedby
Soubra(2002) isemployedtoinvestigatethepossibil-
ity of partial failureinlargesizeslurry shield-driven
tunnel. Also, a more rigorous 3D numerical model-
ingiscarriedouttocomparewiththeobtainedresults
of critical slurrypressuresandthecorrespondingsoil
massat failure.
2 CASE STUDY
The main part of Shanghai Yangtze River Tunnel is
7.5kmriver-crossing tunnel connecting the Pudong
andChangxingIslandinShanghai, China. Excavation
of the tunnels is carried out by a pressurized slurry
shieldmachinewithanoutsidediameter of 15.43m,
whichistheworldslargest until now.
Figure3. Multiblockfailuremechanisms.
Thetunnel ismainlyexcavatedbelowtheriver bed
of YangtzeRiver, which is composed of muddy clay
and soft clay, with some local lenses of silty fine
sand. Thetunnel will bedrivenat adepthupto65m.
A hydrostatic pressure up to 650kPa is anticipated.
Theshallowest groundsectionisunder theriver with
acover-to-depthratioof C/D=0.7, whereC andDis
thecover depth and diameter of thetunnel. Because
of theunfavorablegeological conditionandthelarge
dimensionof excavatingface, thefacestabilityof the
tunnel isoneof thekeytechnical aspectsinthisproject.
Themost dangerousprofilewithC/D=0.7ischosen
for thestudy, as showninFigure2, whereH
w
is the
height of thewater table.
3 PARTIAL FAILURE MECHANISM
3.1 Multiblock failure mechanism
Theproblemof thefacestabilityanalysisrelevanttoa
circularrigidtunnel of diameterDdrivenunderadepth
of cover C couldbeidealized, as showninFigure3.
A surcharge
s
isappliedonthegroundsurface, and
t
istheuniformsupportingpressureonthetunnel face.
Themultiblock failuremechanismconsideredinthis
690
paper is described in Soubra(2002). It is composed
of several truncated rigid cones with circular cross-
sections andwithopeningangles equal to 2, where
isthefrictionangleof thesoil. Thedifferent blocks
of this mechanismmoveas rigidbodies. Theserigid
conestranslatewithvelocitiesof different directions,
whicharecollinear withthecones axesandmakean
angle with the discontinuity surface. The velocity
of eachconeis determinedby theconditionthat the
relativevelocity betweentheconesincontact hasthe
directionthat makes anangle withthecontact sur-
face.Thepresentmechanismiscompletelydefinedby
n angularparameters and
i
(i =1. . .n-1), wheren is
thenumber of rigidblocks. Thegeometrical construc-
tionof this mechanismis similar to that of Lecaand
Dormieux (1990). Thepresent mechanismis charac-
terizedbymorefreedomanglesthanthat of Lecaand
Dormieux (1990) andthus improves thesolutions of
thecritical supportingpressure.
Theexternal forcescontributingtotherateof exter-
nal workconsistof (i) theself-weightof thetruncated
rigid cones; (ii) thesurchargeloading
s
(in caseof
outcropof theupper rigidblock) and(iii) thepressure
t
atthefaceof thetunnel. Therateof energydissipa-
tionoccursalongthelateral surfacesandradial planes
of thefailuremechanism. By equating thetotal rate
of external work to the total rate of internal energy
dissipation, thepressure
t
at thefaceof thetunnel is
obtainedasfollow:
where, N
s
andN
aresurchargeandsoil weight coef-

ficient; c, arethecohesionandfrictionangleof the
soil and isthesoil unit weight.
3.2 Critical slurry pressures
In order to study the possibility of partial failure,
two partial failuremechanisms areusedas shownin
Figure4, correspondingto theblow-out of theupper
partof tunnel andthecollapseof thelowerpartof tun-
nel. Thepartial failureisassumedtooccur inanarea
with avertical axis D
L
, whereD
L
(0, D). In upper
part blow-out mechanism, as shown in Figure 4(a),
thetop of thefailureareapasses through thetunnel
crown. Inthelowerpartcollapsemechanism, asshown
in Figure4(b), thebottomof thefailureareapasses
throughthetunnel invert. Theupper part collapseand
lower part blow-out arenot considered becausethey
arelessdangerous.
Thecritical slurry pressurecorresponding to par-
tial upper part blow-out and lower part collapse are
computed in a simplified approach using the multi-
blockfailuremechanismasfollow: thetunnel pressure
obtainedfromthemultiblockmechanismfor different
Figure4. Twokindsof partial failuremechanisms.
prescribedvaluesof thetunnel diameters(correspond-
ingto different values of D
L
in thepresent analysis)
are computed. The value of D
L
giving the minimal
(respectively maximal) blow-out (respectively col-
lapse) pressureis considered as thecritical pressure
causing a partial failure. A five-block model (i.e.
n =5)isemployedforthisstudy(asshowninFigure3)
since it was shown in Soubra (2002) that n greater
than5will notsignificantlyimprovetheresults. Water
abovethegroundsurfaceisconsideredasasurcharge.
Becausetheexcavationisexecutedquicklycompared
tothesoil consolidation, atypical undrainedanalysis
isemployed.Table1summarizesthecharacteristicsof
thesoil. Theunit weight of slurry is
F
=12kN/m
3
.
Incase1, thecohesionisconsideredtobeaconstant.
In case 2, cohesion of the soil increases with depth
asc
u
=0.95z+0.4, wherez isthedepthof thesoil
layer. Themeanvaluefromthelevel of tunnel invert
to theground surfacein case2 is equal to thec
u
in
case1. Case2ischosentotakeintoaccount thevari-
ation of the undrained cohesion between the tunnel
691
Table1. Soil parametersusedinnumerical modeling.
unit friction Youngs
weight cohesion angle modulus Poissons
0
c
u

u
E ratio
(kN/m
3
) (kPa) (
) (MPa)
case1 18.2 24.5 0.01 3.21 0.495
case2 18.2 0.95z+0.4 0.01 3.21 0.495
crown and invert in extra large tunnel, especially in
normallyconsolidatedclay.
Accordingtothebasicassumptionof themultiblock
failuremechanism, theintersectionof thefailureblock
andthetunnel facewill beanellipsewithalongaxisof
D
L
invertical direction.Thecritical slurrypressure
t0
obtainedbymultiblockmodel isauniformpressure(as
showninFigure3). Consideringthetotal forcebalance
onthefailurefaceof diameterD
L
, anequivalentslurry
pressureis obtained. Theslurry pressure
t
at tunnel
crowncanbecomputedasfollow:
t
=
t0
s
D
L
,2.
Actually, sincetheslurrypressureincreasewithdepth,
thetotal forceduetotheslurryisactingonthetunnel
face with eccentricity e. When using uniformpres-
sure
t0
, amoment shouldbetakenintoconsideration
at the same time. However, the moment is not con-
sideredheresincethenormalizedeccentricity e,D is
verysmall inthisstudy(about 2%4%).
Thecritical slurrypressure
t
iscalculatedfor par-
tial diameter D
L
varyingin1msteps, whichleads to
themaximum(or minimum) valueof
t
, asshownin
Figure5. Incase2, thecohesionfor differentD
L
isthe
meanvaluefromthetopto thebottomof thefailure
block. Inupperpartblow-out, cohesiondecreaseswith
thedecreaseof D
L
, sincethefailureblock becomes
moreandmoreshallow. Whileinlower part collapse,
the failure block is always starting fromthe tunnel
invert tothegroundsurface, themeanvalueof cohe-
sion for different D
L
is equal to the one in case 1.
Therefore, thereisnodifferencebetweenthevalueof
t
incase1and2incollapse.
Figure 5 shows that in case 1, the maximumand
minimumcritical slurrypressureisobtainedwhenpar-
tial failurediameter D
L
=D=15.43m, whichmeans
thattheglobal failurewill happenbothincollapseand
blow-out cases.
In case 2, the minimumslurry pressure in blow-
out isobtainedwhenpartial failurediameter is12m,
whichmeansthatinblow-outfailure,thepartial failure
of upper part tunnel facewill bemoredangerousthan
theglobal failureof thewholetunnel face. However,
there are no great differences among critical slurry
pressures when D
L
(10, 15.43). Therefore, in prac-
tice, global failureof theentirefaceandpartial failure
withD
L
>10mwill havethesameprobability.Forcol-
lapse, themaximumslurrypressureisobtainedwhen
Figure5. Critical slurry pressureof partial failuremecha-
nism.
D
L
=15.43m. Withthedecreaseof thepartial failure
diameter, thecritical slurry pressurewill decreaseat
thesametime:global failureof thewholetunnel faceis
moredangerousthanpartial failure.Thecritical slurry
pressure at the tunnel crown level is 619.9kPa and
265.9kPa corresponding to blow-out (partially) and
collapse(globally), whichis712.5kPaand358.5kPa
at thelevel of tunnel center.
4 NUMERICAL ANALYSISWITHFLAC
3D
4.1 FLAC
3D
numerical modeling
In order to investigate the behavior of the tunnel
faceduringfailure, numerical analysis is carriedout
withthecommerciallyavailablefinite-differencecode
FLAC
3D
, whichisaneffectiveprogramfor situations
wherephysical instabilitymayoccur.
In the model, due to symmetry, only one half is
included. The model is sufficiently large to allow
for any possible failure mechanismto develop and
to avoid any influence from the model boundaries
(as shown in Figure6). Thewater tableis 2D above
the tunnel crown. In order to focus the analysis on
692
Figure6. Studyingprofileand3Dnumerical model.
the face failure in front of the shield machine, the
excavationprocess was simulatedusingasimplified
single-stepexcavationscheme, assumingthat thetun-
nel isexcavated13m(thelengthof theshieldmachine)
instantaneously. Such asimplified modeling scheme
had been successfully adopted in previous studies
(Gioda& Swoboda, 1999).
Thedifferent soil layers areassumedto beelastic
perfectly-plastic materials conforming to the Mohr-
Coulomb failure criterion. An undrained analysis is
carriedout by usingtheundrainedparameters of the
soil asshowninTable1.
Theinitial slurry pressure
s0
is equal totheearth
pressureat rest at center of tunnel face(inthis case,
s0
=537.6kPa). The slurry pressure is increasing
withthedepthaccordingto thedensity of theslurry.
Slurrypressureatthecenter of thetunnel face
si
will
be increased (or decreased) by multiplying
s0
by a
pressurefactor M ineveryconstructionphasei, until
blow-out (or collapse) occurs. Thefailurecriterionis
definedas follows: theconstructionphaseis consid-
ered as the beginning of failure, where for the first
timeconsiderablevalueof unbalanceforce(not equal
to0) andvelocity (greater thane
11
e
12
m/stepin
this case) is observed. Then the critical slurry pres-
surecorresponding to collapseand blow-out will be
obtainedbythefollowingequation:
4.2 Numerical modeling results
For case1, collapseoccurs at theconstructionphase
when M =0.65 and blow-out occurs at M =1.30;
M =0.70andM =1.30for case2respectively. Max-
imumandminimumcritical slurry pressurecouldbe
obtained by equation (2). Displacement contours at
thefailureof case2 is plotted in Figure7. A global
collapsefailureof thewholetunnel faceisobserved.
Thismechanismiswell coincidentwiththecentrifugal
experimentresultof ChambonandCort(1994).How-
ever failuremechanismof blow-outisapartial failure
Figure7. Displacement contours whenfailureis observed
incase2.
Figure 8. Velocity contours of tunnel face when failure
occurs.
in the upper part of the tunnel face. Partial failure
mechanismismoreobviousincase2thanincase1.
Two cases are compared in Figure 8. In collapse,
asshowninFigure8(a) and(b), thetotal tunnel face
will haveconsiderablevalueof velocity; andthemost
dangerous point with maximumvelocity is near the
tunnel invert. Thefailureareaonthetunnel faceisthe
whole face, which is circular in shape. In blow-out,
as weseein Figure8 (c) and (d), thefailureis con-
strainedinupper3/4Dand1/2Dpartof thetunnel face,
corresponding to thecase1 and case2 respectively.
The failure mechanismin both cases has an elliptic
shape, withalongaxisinthehorizontal direction.The
most dangerous point with themaximumvelocity is
thepointnearthetunnel crown.Thefailuremechanism
of case2couldbeobviously observedby theplot of
693
Figure9. Velocitydevelopmentof blow-outalongthecenter
vertical axisof thetunnel faceof case2(unit: m/step).
velocity evolution of themonitored points along the
center vertical axis of the tunnel face, as shown in
Figure9.
5 COMPARISONOF FAILURE MECHANISMS
Thefailuremechanisms of five-block model and3D
numerical modelingarecomparedincase2(asshown
inFigure10). Infive-block model, twosets of and
i
are obtained by optimization of the coefficients
N
and N
s
. As shown in Figure 10, there are only
small differencesbetweenthetwosoil massat failure
correspondingtothetwosetsof and
i
.
Basicallyspeaking, thefailuremechanismsof five-
block model and 3D numerical modeling well agree
witheachother. Bothof themwell predict thepartial
failureontheupperpartof thetunnel faceinblow-out,
andglobal failureof wholetunnel faceincollapse.
6 CONCLUSIONS
1. Boththeresultsof partial failuremechanismbased
on multiblock model and 3D numerical analysis
show that thepartial failuremechanismwill hap-
peninblow-outbutglobal failurewithentiretunnel
facewill dominatethecollapsefailure, especially
incasewherecohesionischangingwithdepth.
Figure 10. Comparison of failure mechanisms of Case 2
(velocitycontour for FLAC
3D
analysis).
Table 2. Comparison of critical slurry pressure at tunnel
center level incase2.
Critical slurry Pressurefactor
pressure/kPa M
Collapse Blow-out Collapse Blow-out
5-BlockModel 358.5 712.5 0.67 1.33
3DNumerical 376.3 698.9 0.70 1.30
2. Thecritical slurrypressuresof bothmultiblockpar-
tial failuremechanismand 3D numerical analysis
areasshowninTable2. Theresultswell agreewith
each other. It should benoted that themultiblock
model is the upper-bound solution; it is possible
that asmaller valueof thecritical slurry pressure
for blow-out andgreater onefor collapsecouldbe
found.
694
3. The results of 3D numerical analysis show that,
in blow-out, thefailurearea on thetunnel faceis
described by an elliptic shape, with a long axis
inhorizontal direction, andashort axis invertical
direction. However, thisshapeof failuremechanism
is different fromtheassumptionof themultiblock
model of upper-boundtheorem, whichisanelliptic
shapewithlongaxisinvertical direction.
4. In multiblock model analysis of case 2, a mean
value of cohesion c
u
fromthe top to the bottom
of thefailureblock isemployed. However, accord-
ingto theshapeof thefailureblock, thechanging
cohesionwill influencetheenergydissipationalong
thelateral surfaceof failureblock. A moreprecise
calculation of energy dissipation along thefailure
surface with c
u
changing with depth is necessary
to take into account. Also, the rotation of failure
block causedbytheeccentricityof theslurrypres-
sureshould beconsidered in multiblock model to
obtainbetter results.
ACKNOWLEDGEMENTS
Theresearchwereconductedwithfundingprovidedby
theNational HighTechnologyResearchandDevelop-
ment Program(863Program) of ChinaandShanghai
LeadingAcademic DisciplineProject, Project Num-
ber: B308. The first author is grateful to EGIDE to
providetheFrenchfundingscholarshipof herdoctoral
stay inINSA Lyon. Particular thanksareduetoProf.
Soubra. A. H. for important discussionandtheuseof
multiblocksoftware.
REFERENCES
Anagnostou, G. & Kovri, K. 1994. The Face Stability
of Slurry-Shield-DrivenTunnels. Tunnelling and Under-
ground Space Technology. Vol.9, No. 2:165174.
Broere, W. 2001. Tunnel Face Stability & New CPT Applica-
tions. PhD thesis, Delft University of Technology. Delft
UniversityPress, theNetherlands.
sionless soil: Stability of tunnel face. ASCE Journal of
Geotechnical Engineering. 120: 11481165.
Davis, E.H., Gunn, M.J., Mair, R.J. & Seneviratne, H.N.
1980. TheStabilityof ShallowTunnelsandUnderground
Openings in Cohesive Material. Geotechnique. 30 (4):
397416.
Gioda, G. & Swoboda, G. 1999. Developments andAppli-
cations of the Numerical Analysis of Tunnels in Con-
tinuous Media. International Journal for Numerical and
Analytical Methods in Geomechanics. 23: 13931405.
Horn, M. 1961. Horizontal earth pressure on perpendicu-
lar tunnel face. Hungarian National Conference of the
FoundationEngineer Industry, Budapest. (InHungarian)
frictional material. Gotechnique 40, 4: 581606.
Soubra, A.H. 2000. Three-dimensional facestability analy-
sis of shallowcircular tunnels. International Conference
on Geotechnical and Geological Engineering, 1924
November 2000. Melbourne, Australia.
Soubra, A.H. 2002. Kinematical approach to the face sta-
bility analysis of shallow circular tunnels. 8th Interna-
tional Symposiumon Plasticity, 443445. Canada, British
Columbia.
Subrin,D.&Wong,H.2002.Tunnel facestabilityinfrictional
material: anew3Dfailuremechanism. C. R. Mecanique,
330: 513519. (InFrench)
695
Effectsof soil stratificationonthetunneling-inducedgroundmovements
F.Y. Liang, G.S.Yao& J.P. Li
ABSTRACT: Theavailableanalytical predictionsfor tunneling-inducedgroundmovementsareusuallybased
ontheassumptionsthat thegroundishomogeneousandelastic. Actually, effectsof soil stratificationshouldbe
takenintoaccount. Inordertoclarifythisproblem, aFLAC3Dcomputational model wasconductedtoinvestigate
theeffect of vertical stratificationinsoil onthetunneling-inducedgroundmovements. Theapplicabilityof the
presented model is verified with other availablepublished results as well as thefield casehistories. Surface
settlementsandlateral displacementsinducedby tunnelingwithlayeredsoil model werecomparedwiththose
basedonhomogeneoussoil.Theresultsshowthatthestratificationof soil shouldnotbeneglectedintheanalysis
of thetunneling-inducedgroundmovements, especiallywhenthesoil isassumedtobeanelasticmedium.
1 INTRODUCTION
Tunneling process will inevitably result in ground
movements, and as a result tunnel constructions in
urbanareamay causeseriousdamagetotheadjacent
preexisting buildings or public facilities. Therefore
one of the important issues of tunneling in urban
areas is the estimation of potential ground move-
ments. Methods for estimating theground deforma-
tions due to tunneling may be classified into three
categories: empirical methods, numerical methods,
andanalytical methods.
Intheengineeringpractices, grounddeformations
areoftendescribedasanormal distributioncurvewith
empirical formulas basedonfieldobservations (e.g.,
Peck 1969; New & OReilly 1991). These methods
arelimited in their applicability to copewith differ-
ent ground conditions and construction techniques,
andtheinformationtheycanprovideislimitedinthe
horizontal movementsandsubsurfacesettlements.
Someattempts havebeenmadeto developsimple
closed-form analytical solutions for tunneling-
inducedgroundmovementsinclayswiththeassump-
tion of a uniform radial or oval-shaped ground
deformation pattern around the tunnel section (e.g.,
Sagaseta1987;Verruijt&Booker1996;Loganathan&
Poulos 1998; Park 2004). However, theeffect of soil
stratification was not considered in these analytical
methods.
Basedonthethree-dimensional (3-D) natureof the
stress changes anddeformations, numerical methods
such as finite element method and boundary ele-
ment method could be used to analyze the effect of
soil stratification on the tunneling-induced ground
movements(Mair et al. 1996) and3-Dfiniteelement
simulationmodelsforshield-driventunnel excavation,
taking into account relevant components of thecon-
struction process, havebeen developed (e.g., Lee&
Rowe1991; Rowe& Lee1992).
In order to analyze the effects of soil stratifica-
tionontunneling-inducedgroundmovements, a3-D
computational model wasusedtosimulatethetunnel
excavationinsoftsoil inthispaper. Firstly, thefactors
of thecomputational model takingintoaccountthesoil
stratificationweredescribed. Secondly, thesimulation
of theThunder Bay Tunnel inlayeredsoft soil using
FLAC3D with Mohr-Coulomb material model was
carried out. The predicted ground movements were
investigatedindetail. Inorder tochecktheapplicabil-
ity of theproposedcomputational model, theresults
werecomparedwiththosefromLee& Rowe(1991)
and the field data observed by Belshaw & Palmer
(1978). In theend, another simulation for thetunnel
usingtheelasticmodel wascarriedout. Theresultsof
thegroundmovements werecomparedwiththeana-
lytical resultsof Loganathan& Poulos(1998) andthe
fielddataobservedbyBelshaw& Palmer (1978) and
Leeetal. (1992).Theeffectsof soil stratificationwere
discussed and some suggestions for future research
werealsoput forward.
2 METHODOF ANALYSIS
2.1 Simulation of the ground loss
The settlements caused by tunneling are often char-
acterizedby thetermgroundloss. Themaincause
tothegroundlossisthat thereisthegapbetweenthe
697
Figure1. Definitionof GAP (Leeet al. 1992).
constructionsectionandthedesignsection. Thenthe
actual groundmovementatthetunnel openingsection
isoval shaped.Thesemovementscanbeapproximately
incorporatedtothevolumeof groundlossthroughthe
heading and over the shield (as shown in Figure 1).
Thus thegapparameter (GAP) canbeconsideredas
themaximumsettlement at thetunnel crown, and it
maybeexpressedasLeeet al. (1992).
where G
P
=physical gap (G
P
=2L+) that repre-
sents thegeometric clearancebetweentheouter skin
of the shield and lining; L=thickness of the tail-
piece; =clearancerequiredforerectionof thelining;
u
3D
=equivalent3-Delasto-plasticdeformationatthe
tunnel face; and=valuethat takesintoaccount the
qualityof workmanship.
Applyingtail voidgroutwill leadtoamuchsmaller
valueof GAP thanindicatedhere, but this valueis a
possiblemaximum.
2.2 Principal assumptions
Themediumaround thetunnel almost always exists
intheformof stratificationfeatures, andthetunnel-
ing analysis is a matter of three-dimensions rather
thantwo-dimension. FLAC3Disusedtosimulatethe
tunneling-induced ground movements in this paper.
FLAC3Disafinitedifferencethree-dimensional soft-
ware package based on the Lagrangian difference
method. FLAC3D canoffer anideal analysistool for
solution of three-dimensional problems in geotech-
nical engineering. Theassumptions of theanalytical
methodarepresentedinbrief asfollows.
1. Aplanecontactontheinterfaceof differentsoil lay-
ersisassumed.Thetransversemovementsbetween
thesoil layers areneglected, andthenthefriction
forceof thesoil layerscanbeignored.
2. The inner material is homogeneous in the same
soil layer. The influence of the ground water is
neglected.
Figure2. 3-Dmodel geometryusedinanalysis.
3. Uniformpressureisappliedtothegroundsurface
as the external loads. The ground stress field is
generatedbythegravityforce.
4. Themediumis simplifiedas thecontinuum, ideal
elasto-plastic material, and the Mohr-Coulomb
material model isadopted.
5. The oval-shaped ground deformation pattern is
imposed as the boundary condition at the tunnel
opening.
3 EFFECTSOF SOIL STRATIFICATION
USINGELASTO-PLASTIC MODEL
3.1 Calculation model and boundary conditions
A 3.3-km-long, 2.47-m-diameter section of the san-
itary trunk sewer tunnel in thecity of Thunder Bay,
Ontario, Canada, wasconstructedin1976. Thetunnel
was constructed through soft clay using a tunnel-
boring machine together with a segmented precast
concretetunnel lining. Thedetailsof tunnel construc-
tion, soil condition, and thetunnel dimensions were
publishedelsewherebyBelshaw&Palmer (1978) and
Lee& Rowe(1991).
Becauseof thesymmetry of thestructureandthe
influencerangeof thetunnel excavation, onlyhalf the
tunnel is considered. Thedepthfromthegroundsur-
facetothetunnel axisis10.7m. Themodel geometry
is30mlong(ydirection), 25mdeep(zdirection) and
20mwide (x direction). Figure 2 illustrates the 3-D
meshusedinthisanalysis. Thereare21240elements
and 23464 nodes. Thelateral and bottomsurfaceof
themodel aredisplacement boundaries. Themodel is
698
fixedinthehorizontal directionatlateral surface, and
thevertical displacement islimitedat thebottomsur-
face.Thetopsurfaceof themodel isthegroundsurface
andisfreeinall directions.
For theThunder Bay tunnel, theestimationof the
gapparameter proposedbyLeeet al. (1992) hasbeen
usedandthephysical gapG
P
is90mm.Thecalculation
hastwostages. Firstly, theradial oval displacement as
aboundary conditionisappliedonthetunnel crown.
The3Dmovementsaheadof thefaceintothezoneto
beexcavatedcanthenbedetermined, andthenu
3D
can
beevaluatedandusedtocalculatethetotal gapGAP.
Secondly, the model based on the tunnel diameter
2R+GAP, is established. The radial oval displace-
ment asaboundaryconditionisappliedonthetunnel
crown. Hencetheground settlements and thelateral
displacement aroundthetunnel canbesimulated.
Inthis study, thestructural behavior of thetunnel
shieldandliningareneglected. Inorder toensurethe
accuracyof thesimulation, threecasesareassumed.
1. Thetunnel is assumedto beexcavatedunder per-
fect alignment, andthetunnel machineis pressed
hardagainstthefacesothat3-Dmovementwill be
minimized.
2. Full release of axial stress at the tunnel face is
assumed, which would result in the full develop-
ment of 3-Dgroundlossaheadandover thetunnel
shield.
3. Theexcavation of thetunnel is theinstantaneous
excavation, and the interrelationship between the
supporting and lining of the tunnel is neglected.
Results of the ground movements (surface and
sub-surfacesettlements, lateral displacement)from
thenumerical simulation aredefined as thefinal
groundmovementsafter thetunnel excavation.
3.2 Comparison with observed results and other
numerical methods
Inthissection, thesoil isassumedtohaveanelastic-
perfectlyplasticconstitutiverelationshipandaMohr-
Coulombmaterial model.Thesiltandsiltysandstrata
for thetunnel aredividedintofour sublayers accord-
ingtothedatainLee&Rowe(1991). Thelayerswere
assigned soil parameters on the basis of soil densi-
ties,laboratorytesting,andempirical correlationswith
SPT data. Theparametersadoptedfor thesematerials
in the present analysis are based on the research of
Lee& Rowe(1991) andsummarizedinTable1. The
material parametersfor theuniformsoil model arethe
averagevaluesof thefour layersrespectively.
Resultsfromthepresentedanalysiswerecompared
withthosefromLee& Rowe(1991) as well as field
measurementsrecordedbyBelshaw&Palmar (1978).
Theanalysis model usedby Lee& Rowe(1991) was
alsobasedonelasto-plasticconstitutiverelationship.
Figure 3. Surface settlement troughs at various distances
fromtunnel face.
3.2.1 Surface settlement along transverse section
The development of transverse surface settlement
troughs(i.e., settlement profilealongthegroundsur-
faceperpendicular to thedirection of tunneling) for
variouscalculationmethodsat2.2mbehindthetunnel
faceareshowninFigure3.
Theshapeof surfacesettlement troughscalculated
withlayeredsoil model is consistently slightly wider
thantheobservedtroughs, anditsvaluesareapprox-
imateto theresults fromLee& Rowe(1991). From
thecurves, thesettlementvaluesbasedontheuniform
soil model slightlyunderestimatethevaluesbasedon
thelayeredsoil. Figure3shows that considerationin
lamination of thesoil has only asmall effect on the
calculatedsurfacesettlement trough.
3.2.2 Variation of lateral displacement
Thelateral displacement valueswhenthetunnel face
was15mawayfromthemonitoringpointsareshown
inFigure4. Themonitoringdataare2.2mawayfrom
thevertical centerlineof thetunnel.Thepresentresults
based on the layered soil are very similar to those
obtainedby Lee& Rowe(1991) usingthe3-D finite
elementanalysis.Theresultsfromthetwomethodsare
ingoodagreementwiththefielddata. Fromthecurves
in Figure 4, the horizontal movement values based
on theuniformsoil model greatly underestimatethe
fielddataabovethetunnel horizontal centerline, and
thelatter arealmost 1.58timeslarger thanthefor-
mer. Underthetunnel horizontal centerline, thelateral
displacements basedonuniformsoil model decrease
rapidly and are close to those based on layered soil
and the field data. As well known, the lateral dis-
placementabovethetunnel centerlinehasanimportant
influenceonthemovementof structureslocatedclose
to the tunnel. Figure 4 shows that consideration of
soil laminationhas asignificant effect onthelateral
displacement.
699
3.2.3 Variation of subsurface settlement with depth
above the tunnel axis
The observed and calculated subsurface settlement
distributions withdepthat 15mfromthetunnel face
areshown in Figure5. Results based on thelayered
soil arevery closeto theresults obtained by Lee&
Rowe (1991), and they are also in good agreement
with the field data. The settlement values based on
theuniformsoil model underestimatethefielddatain
thewholemeasurementfield, andthelatter arealmost
1.061.30timeslarger thantheformer.
Figure5. Subsurfacesettlements with depth abovetunnel
axis.
Table1. Mechanical parametersof materials.
No. Descriptionof soil layers Thickness(m) (kNm
3
) E
u
(MPa) E
b
(MPa) c (kPa) (
o
) K
0
1 Peat 1.0 14.00 12.3 213.0 39 30 0.5
2 Loosesiltysand 7.0 19.77 15.4 119.3 40 32 0.52
3 Siltyclay 5.2 17.45 9.2 110.5 29 26 0.88
4 Varvedclay 11.8 18.70 32.0 240.0 59 34 0.8
Notes: unit weight of soil; E
u
shear modulus; E
b
bulkmodulus; ccohesivestrength;
internal frictionangle; K
0
coefficient of staticearthpressure.
Fromtheaboveanalyses, results fromthis model
takingintoaccount thestratificationof thesoil arein
goodagreement withthoseobtainedby Lee& Rowe
(1991)aswell asthefielddata, andmoreaccuratethan
thosebasedontheuniformsoil model. Thereforethe
FLAC3Dmodel establishedinthispaper provestobe
reasonableandreliableforthesimulationof tunneling-
inducedgroundmovement.
4 EFFECTSOF SOIL STRATIFICATION
USINGELASTIC MODEL
Loganathan & Poulos (1998) redefined the ground
loss parameter with respect to the gap parameter
and incorporated this into the closed formsolution
derived by Verruijt & Booker (1996). The proposed
approachtendedtooverpredictthegroundlossparam-
eter for tunnelsinsoft clay, andthepredictedsurface
settlement troughs were wider than the field obser-
vations. Nevertheless, in general, the observed and
calculatedsettlement andhorizontal movement arein
goodagreementfor tunnelsinuniformclay. However,
theapplicabilityof thesolutionbasedonuniformsoil
model should be discussed when it was applied to
layeredsoils.
Inthissection, thesoil isassumedtohavealinear
elastic constitutive relationship. The ground around
the tunnel is also divided into four sublayers. The
parametersadoptedfor thesematerialsinthepresent
analysis are summarized in Table 1. The material
parameters of uniform soil model are the average
valuesof thefour layersrespectively.
Resultsfromthepresentanalysisarecomparedwith
those fromLoganathan & Poulos (1998) as well as
field measurements recorded by Belshaw & Palmar
(1978) andLeeet al. (1992). Theeffectsof thestrat-
ification of the soil on the ground movements are
analyzedinthefollowing.
4.1 Surface settlement along transverse section
Settlementresultsfromthiscalculation,Loganathan&
Poulos(1998) andthefielddata(Leeet al. 1992) are
shown in Figure 6. Apparently, the values predicted
byLoganathan&Poulos(1998) overestimatethefield
700
Figure 6. Surface settlement troughs at various distances
fromtunnel face.
data. The values based on the layered soil underes-
timatethevalues predictedby Loganathan& Poulos
(1998) andareclosetothefielddatadespitethefact
that theyareslightlygreater thanthefielddata. From
thecurves, thesettlementvaluesbasedontheuniform
soil model are greater than the values based on the
layered soil in most areas and thetrough of thefor-
mer isobviously wider thanthat of thelatter. But the
maximumsettlements obtainedby this model under-
estimatethevalueobtainedby Loganathan& Poulos
(1998) andthefielddataat thetunnel crown.
4.2 Variation of lateral displacement
The values of the lateral displacements monitoring
pointsareshowninFigure7. Thevaluesobtainedby
Loganathan & Poulos (1998) overestimate the field
data(Belshaw&Palmer 1978) withinthedepthrange
from5mto15m.Thevaluesbasedonthelayeredsoil
arein good agreement with thefield data. Fromthe
curves, whenadoptedtheuniformsoil model, theval-
ues areless than thefielddataat therangeof depth
from0to7m, andclosetothefielddataat therange
of depthfrom7mto 11mandgreater thanthefield
dataattherangeof depthfrom11mto25m. Figure6
shows that the prediction of the ground lateral dis-
placements basedonthelayeredsoil model is better
thantheuniformmodel.
4.3 Variation of subsurface settlement with depth
above the tunnel axis
The results of the subsurface settlement above the
tunnel centerline are shown in Figure 8. The values
obtained by Loganathan & Poulos (1998) underesti-
matethefield data(Leeet al. 1992) greatly in spite
thatthetrendsof bothprofilesaresimilar.Thepresent
results based on layered soil show larger values for
thesurfacesettlement thanthosefromLoganathan&
Poulos (1998) andcloseto thefielddataexcept that
Figure 8. Subsurface settlements with depth above the
tunnel axis.
thereisanerrorof 711percentattherangeof depth
from0to2m. At other depthstheerror islessthan5
percent.Thesubsurfacesettlementvaluesbasedonthe
uniformsoil model obviouslyunderestimatethefield
dataandthereisanerror of 2030per cent between
them.
5 CONCLUSIONS
A FLAC3D computational model for simulating the
displacementcausedbyexcavationof theThunderBay
tunnel hasbeenperformed, andtheeffectsof thestrat-
ificationof thesoil onthetunneling-inducedground
movementsareanalyzedsufficiently.
Resultsfromthemodel basedonlayeredsoilsarein
goodagreement withthoseobtainedby Lee& Rowe
(1991) aswell asthefielddata. Comparatively, results
disregardingthestratificationof thesoil will havedis-
tincterrors. Consequently, inengineeringpractice, the
stratification of soils should not be neglected in the
analysisof thetunneling-inducedgroundmovements.
701
Furthermore, the results based on the elasto-plastic
model cansimulatethedeformationcurvesbetterthan
thosebasedontheelasticmodel.
Based on the elasto-plastic model, the analysis
resultsshowedthat thesoil stratificationhasveryfew
effects on the ground surface settlement, but it has
greater effectsonthegroundlateral displacementand
sub-surfacesettlement.
While based on the elastic model, the analysis
resultsshowedthatthesoil stratificationhassomedis-
tinct effects in theground surfacesettlement, lateral
displacement andsub-surfacesettlement respectively.
Inorder toobtainmorereasonableanalytical solu-
tion to the tunneling-induced ground movements, it
is necessary to incorporatethestratification of soils
intotheanalytical solutionproposedbyLoganathan&
Poulos(1998).
ACKNOWLEDGEMENTS
ThisworkreportedhereispartlysupportedbyNational
Natural Science Foundation of China (Grant No.:
50708078) andProgramfor YoungExcellent Talents
in Tongji University (Grant No.: 2006KJ 043). The
authorswishtoexpresstheirgratitudeforthefinancial
assistances.
REFERENCES
Belshaw, D.J. &Palmer, J.H.L. 1978. Resultsof aprogramof
instrumentationinvolvingaprecast segmentedconcrete-
linedtunnel inclay. Canadian Geotechnical Journal 15:
573583.
Lee, K.M. & Rowe, R.K. 1991. An analysis of three-
dimensional groundmovements: theThunder Baytunnel.
Canadian Geotechnical Journal 28: 2541.
Lee, K.M., Rowe, R.K. & Lo, K.Y. 1992. Subsidenceowing
totunnelling. I. Estimatingthegapparameter. Canadian
Geotechnical Journal 29: 929940.
Loganathan, N. & Poulos, H.G. 1998. Analytical predic-
tion for tunnelling-induced ground movements in clays.
J. Geotech. Geoenviron. Eng. 124(9): 846856.
Mair, R.J. 1996. Predictionof groundmovementsandassess-
ment of risk of building damage due to bored tunnel-
ing. Proceedings of geotechnical aspect of underground
construction in soft ground, Rotterdam, BalkemaPress:
713718.
New,B.M.&OReilly,M.P.1991.Tunnellinginducedground
movements, predictingtheirmagnitudeandeffects. Proc.,
4th Conf. on Ground Movements and Structures, Cardiff,
Wales, Pentech Press: 671697.
Park, K.H. 2004. Elastic solution for tunnelling-induced
ground movements in clays. Int. J. Geomech., 4(4):
310318.
Peck, R.B. 1969. Deep excavations and tunnelling in soft
ground. Proc., 7th Int. Conf. on Soil Mech. and Found.
Engrg.: 225290.
Rowe, R.K. & Lee, K.M. 1992. Subsidence owing to tun-
nelling.II.Evaluationof apredictiontechnique.Canadian
Sagaseta, C. 1987. Analysis of undrained soil deformation
duetogroundloss. Geotechnique 37: 753756.
Verruijt, A. & Booker, J.R. 1996. Surface Settlements
Due to Deformation of a tunnel in an elastic half
plane.Geotechnique 46(4): 753756.
702
Centrifugemodellingtoinvestigatesoil-structureinteractionmechanisms
resultingfromtunnel constructionbeneathburiedpipelines
A.M. Marshall & R.J. Mair
Department of Engineering, University of Cambridge, Cambridge, UK
ABSTRACT: Newundergroundconstruction is undertaken increasingly closeto existingburiedstructures.
Theresulting effects on thesestructures must beproperly evaluated. This paper examines thecaseof tunnel
constructionbuilt transverselytoexistingburiedcontinuouspipelinesusingdataobtainedfromtestswithinthe
University of CambridgeGeotechnical BeamCentrifuge. Thisresearchaimstovisually validatehypothesized
soil-structureinteractionmechanismsthat account for pipelinebehaviour. Thisisaccomplishedby placingthe
tunnel-soil-pipelinesystemdirectlyagainst aPerspexwall withinthecentrifugepackagesothat digital images
canbetakenof thesoil andburiedstructures. ParticleImageVelocimetry(PIV) isusedtomeasuredisplacements
andprovideacompletedescriptionof thesoil-structureinteractions.
1 INTRODUCTION
Beneath thesurfaceof any major city is an intricate
andincreasinglycongestedseriesof tunnels,pipelines,
and buried structures. New tunnels are constructed
for various purposes and forman important part of
urbaninfrastructure. Tunnel constructionhasvarying
effectsonsurroundinggrounddependingonsoil type
andconstructionprocess(Mair andTaylor, 1997).The
designof newtunnelsmust account for likely effects
onnearbyburiedstructures.
Analytical methods for this problemtypically fall
within three categories: [1] finite element (FE) or
finitedifference(FD)methods, [2]Winkler-typemod-
els, and[3] continuumsolutions. Winkler modelsand
continuumsolutionsgenerallyprovidequicksolutions
(see for example Attewell et al., 1986; Klar et al.,
2005; Vorster et al., 2005; Klar et al., 2007) however
theyall havesimilarsimplifyingassumptionsthatlimit
their applicability. Klar & Marshall (2007) validated
someof thesimplifiedassumptionsof thecontinuum
methodbycomparisonwithmorerigorousshell struc-
turesolutions. FE andFDmethodsarebest suitedfor
theproblembecausetheycanincorporatethecomplex
tunnel-soil-pipelineinteractions. For thesemodels to
beapplicable, theymust bevalidatedagainst real soil
andsoil-structurebehaviour.
Vorster et al. (2005) proposed a series of global
and local soil deformation mechanisms to account
for observed pipelinebehaviour aboveatunnel dur-
ingcentrifugetesting. Theproposedmechanisms are
basedondeformationandstressmeasurementsmade
withinthesoil mass.
Figure1. Centrifugepackage.
The current research aims to elucidate the soil-
structureinteractionmechanismsthat account for the
behaviour of existing tunnels and buried pipelines
located above new tunnel construction by providing
visual evidence of soil and structure deformations.
This is accomplished by performing similar tests to
thosedescribedinVorster etal. (2005) butwithanew
centrifugepackagethatplacesthetunnel-soil-pipeline
systemadjacent to a Perspex wall such that digital
images of thesystemcanbeanalyzedusingParticle
ImageVelocimetry(PIV).
2 CENTRIFUGE PACKAGE
A centrifuge package was developed that allowed
visual observations to be made of the soil, model
703
Figure2. Schematicof centrifugepackage.
tunnel, andburiedpipeline. Indoingso, digital images
couldbetakenduringthetest andPIV (Whiteet al.,
2003) could be used to accurately measure soil and
structuredisplacements.
Figures 1 and 2 present the developed centrifuge
package. The relevant components are labelled as
follows:
A: The model tunnel: composed of a stiff inner
cylinder, fixedwithintheboxwalls, andsealedwithin
aflexiblerubber membrane. Fluid is extracted from
within thesealed rubber membraneto replicatetun-
nellingvolumeloss;
B: Three 8 megapixel digital cameras. These are
usedto photographthesoil andburiedstructures for
PIV analysis;
C: Five lasers used to measure surface vertical
displacements within the middle third of the box
where boundary effects are negligible (see Figure 2
for locations);
D: Five linear variable differential transformers
(LVDTs)usedtomeasuresubsurfacevertical displace-
ments withinthemiddleof thebox andat adepthof
70mm,correspondingtotheaxisof themodel pipeline
(seeFigure2for locations);
E: The model pipeline: consists of an aluminium
half-cylinder withalengthof 700mm, anouter radius
of 9.5mm, andawall thicknessof 1.6mm.
All testswereperformedat75gusingdryLeighton
Buzzard fraction E sand at a relative density of
approximately90%.
3 BASELINETESTS NOPIPELINE
Twotestswerecompletedinwhichnomodel pipeline
wasincluded.Thesetestsservedtwopurposes: first, to
investigatetheeffectsof placingalayer of glassatthe
Perspexinterface(thoughttoreduceboundaryeffects)
andsecond, to obtainbaselinedisplacement datafor
thesimulationof tunnellinginsand.
Figure3. Vertical soil displacementdatafora) surfacemea-
surementswithglassinterface, b) sub-surfacemeasurements
withglass interface, c) surfacemeasurements withPerspex
interface, and d) sub-surface measurements with Perspex
interface.
Volumeloss (VL) is used as areferencethrough-
out the following text. Two methods of determining
volumelossarereferredtoandaredefinedasfollows:
1 Tunnel volume loss: the change in tunnel vol-
ume (or volume of fluid extracted from the
model tunnel) divided by the original total tun-
nel volume, expressed as a percent (VL
tunnel
=
LV
tunnel
/V
tunnel
100)
2 Soil volumeloss: calculatedbyintegratingthesoil
settlement profileat agivendepthanddividingby
theoriginal tunnel volume, expressedasapercent
(VL
soil
=V
trough
/V
tunnel
100).
3.1 Assessment of boundary friction using glass
and Perspex
BoundaryeffectsareanissuewhenobtainingPIV data
fromcentrifugetests. Anassessment of theboundary
effectsontheobtainedPIV resultswascarriedout by
comparingthePIV displacementdatatothatobtained
fromthelasersandLDVTswhichwereplacedwithin
the middle third of the thickness of the box (where
boundaryeffectsarenegligible). Glassisoftenusedas
aninterfacebetweenthesoil andthePerspexwall when
performingPIV testsbecauseitisharder thanPerspex
andthereforepreventssandgrainsfromscrapinginto
thePerspexathighstresslevels(aphenomenonwhich
causesincreasedboundaryfriction).
Twotestswerecarriedout: onewithglassandone
with Perspex alone. Figure 3 presents a comparison
between the PIV centreline data (directly above the
tunnel)andthesurface(laser)andsub-surface(LVDT)
data, for boththeglass andaPerspex interfacetests.
Figure 3 shows very good correlation between the
704
PIV dataandthat obtainedfrommeasurementsmade
within themiddleof thebox for both glass andPer-
spex interfaces, indicating very little friction at the
interface.
Theuseof glasscausedoptical problems(coloured
zonescausedbyrefractionof light) whichaffectedthe
quality of thePIV analysis. It was thereforedecided
that glass wouldnot beusedsincethePIV displace-
ment data was effectively the same for glass and
Perspex.
Thereasonforthedepartureof thePIVdatafromthe
general trendinFigure3b) at atunnel volumelossof
about3.7%isnotknown. Itmaybearesultof soil loss
betweentheglass andthePerspex aroundthemodel
tunnel. This potential problemwas addressedinsub-
sequent tests by sealing circumferenceof thetunnel
endswithgrease.
3.2 Tunnelling in sand
Thebaselinetests provideduseful dataregardingthe
effectsof tunnellinginsand. Figure4presentsthevari-
ationof volumelossat variousdepthswithinthesoil
for a) thebaselinetest with aglass interface, andb)
thebaselinetestwithaPerspexinterface. The1:1line
representsthevolumelossof themodel tunnel. When
theslopeof alineisgreater than1, thesoil isexperi-
encingavolumelossthat islarger thanthat provided
bythevolumelossinthetunnel.
Figure 4 illustrates that the value of volume loss
for sands is not unique. In clays, it is correctly
assumedthat volumelossdoesnot changewithdepth
(undrainedcase,constantvolume)andthereforecalcu-
lationof volumelossderivedfromsurfacesettlements
shouldgiveanaccurateestimateof volumelosswithin
thetunnel. However, estimationof tunnel volumeloss
usingsoil displacementdataisnotsosimpleforsands.
As Figure4 illustrates, it depends very much on the
magnitudeof tunnel volumelossachieved. Thetrend
of soil volume loss presented in Figure 4 was also
notedbyVorster(2005) andJ acobsz(2002) forsimilar
centrifugetestsinsands.
Examining Figure 4, we observe that the surface
andsub-surfacevolumeloss arevery similar upto a
tunnel volumelossof about0.5%, after whichthevol-
umelossescalculatedatdifferentdepthswithinthesoil
diverge. Themaximumratioof surfacevolumelossto
tunnel volumeloss is 1.38for theglass interfacetest
and1.67for thePerspexinterfacetest. Thismaximum
ratio occurs at a tunnel volume loss of between 1.0
and1.2%. Astunnel volumelossisincreasedfurther,
therateof changeof soil volumelossrelativetotun-
nel volumelossbeginstoreduce. Surfacevolumeloss
fallsbelowtunnel volumelossatatunnel volumeloss
of between2.3and2.6%.Theslopefor thesubsurface
soil nearesttothetunnel isshowntoincreasetoabout
1after atunnel volumelossof approximately3%.
Figure4. Soil volumeloss at various depths compared to
tunnel volumefor testwitha) glassinterface, andb) Perspex
interface(VL
soil
calculatedusingPIV displacement data).
Figure5presentsaprobableexplanationfor some
of theresults shown in Figure4 by examining volu-
metric strains within the soil (derived fromthe PIV
analysisof thePerspexinterfacetest). Figure5a) illus-
trates that at alowvolumeloss of 0.54%, thesoil is
generally inacontractivemodeinafan-shapedzone
abovethetunnel. This explains thetendency towards
largersoil volumelossesatlowertunnel volumelosses
inFigure4. Astunnel volumelossisincreasedto1%
andhigher, alocal zoneof dilationdevelopsinthesoil
abovethetunnel. As this dilatant zonegrows insize
andmagnitude, itcompensatesfor thecontractivesoil
aboveit andresultsinlower calculatedvaluesof soil
volume loss, thus reducing the slope of the lines in
Figure4.
4 PIPELINEABOVETUNNEL
Centrifuge testing of a model pipeline placed trans-
versely above the tunnel was performed in order to
investigatesoil displacementsandstructural behaviour
of thepipe.
705
Figure5. Development of volumetric strainas tunnel vol-
ume loss is increased (fromPIV data of Perspex interface
test).
The relevant dimensions of the test are given in
Figure1.Inprototypescale,thetestrepresentsa4.65m
diameter tunnel constructedatadepthof 13.65mand
runningtransverselybeneatha1.43mpipelineburied
atadepthof 5.25m.Thetunnel covertodiameterratio
(C
T
/D
T
) is 2.4 while the pipeline cover to diameter
ratio(C
p
/D
p
) is3.1.
Bendingmoment datawas obtainedusingthedis-
placement data of the pipeline provided by the PIV
analysis (Fig. 6). The vertical deflection data of the
pipelinewasfoundtofit well toamodifiedGaussian
curve(Equation 1) in theformpresented by Vorster
et al. (2005).
where S
v
=vertical displacement of pipe lining;
x =offset from tunnel centreline; S
max
=maximum
vertical displacement of pipelining; i = distanceto
inflexion point of thedisplacement curve; n =shape
functionparametertocontrol thewidthof thedisplace-
mentcurve;and= parametertoensurei isconsistent
withtheGaussiancurvepresentedbyPeck(1969).The
optimumfit was obtained using i =55mmand =
0.01 (resulting in n =0.03). Note that when =0.5
(n =1), Eq. 1becomestheGaussiancurve.
Bendingmomentswerederivedusingelasticbeam
theory whereby momentsarerelatedtothedeformed
shapeof abeamby:
Figure6. Derivation of bending moments fromdeformed
shapeof pipeline(fromPIV data).
where M =bending moment; E
p
=Youngs modulus
of the pipe material; and I
p
=second moment of
the cross-sectional area of the pipe. Note that these
curvesrepresentanapproximationof thetruebending
momentssincethefittedcurvecannot matchexactly
theactual deformedshapeof thepipe.
Thereis aconcernthat tunnellingmay causegaps
belowcertainoverlyingstructures. Figure7presents
PIV displacement data to illustrate that a gap does
indeed formbelow pipelines affected by tunnelling.
Soil displacements directly below the pipeline are
comparedtopipelinedisplacementsatthetunnel cen-
terline. The upper and lower pipe linings are shown
to displace the same amount (i.e. no cross-sectional
distortionof thepipeoccurs). Gapformationisshown
to commence at a tunnel volume loss of between 1
and2%. Thelengthof thegap(alongthepipe) varied
from2.9pipediametersatlowvolumelossto3.4pipe
diametersat 10%tunnel volumeloss.
Figure7alsoshowsaplotof themaximumsagging
moment of the pipe versus tunnel volume loss. The
datasuggeststhatbendingmomentsincreasesubstan-
tially with theonset of tunnel volumeloss, however
pipe bending behaviour does not appear to be very
sensitiveto theformationandgrowthof thegapthat
forms belowthepipe. Theheight of thegap(i.e. the
separationbetweenthesoil andthelower pipelining)
is shown to increase significantly at a volume loss
of about 6%withno equivalent responseinthepipe
706
PIV patch on soil
just below pipe
PIV patches on top and
bottom lining of pipe
maximum sagging
bending moment
(from PIV data)
represents vertical
separation between
soil and lower pipe
lining
0.0
0.4
0.8
1.2
1.6
2.0
2.4
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
tunnel volume loss
0
1
2
3
4
5
6
7
8
(bracketed values indicate
number of pipe diameters)
(2.9)
(3.1)
(3.2)
(3.4)
52
56
60
64
68
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
pipe - upper lining
pipe - lower lining
soil directly below pipeline
maximum sagging bending moment
g
a
p

l
e
n
g
t
h

(
m
m
)
v
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
b
e
n
d
i
n
g

m
o
m
e
n
t

(
N
m
)
Figure 7. Gap formation below pipeline and increase
in maximum sagging moment as tunnel volume loss is
increased.
bendingmoment. Thewidthof thegapdoesnot grow
substantiallyandisthelikelyreasonwhythebending
moments do not respond to sudden increases in gap
height at thetunnel centreline.
5 CONCLUSIONS
Soil-structure interaction mechanisms are important
when evaluating the response of buried structures
to ground movements. The research described in
this paper has elucidated the soil-structure interac-
tionmechanisms that occur whentunnellingbeneath
buriedpipelines.
The results presented illustrate some important
aspectsof tunnellingwithinsandyground(i.e. drained
material). Theseinclude:
soil volume loss will not always be the same as
tunnel volumeloss;
the value of soil volume loss depends on the
magnitudeof thetunnel volumeloss;
soil volume loss calculated at the surface can be
greater or lower thantunnel volumeloss;
themaximumratioof surfacevolumelosstotunnel
volumelosswasashighas1.67andoccurredat a
tunnel volumelossof between1and1.2%;
alocal zoneof dilationformedjustabovethetunnel
at atunnel volumelossof about 1%.
The zone of dilation effectively reduces the soil
volumelossesaboveit.
Thepipelinetest illustratedthat agapformsbelow
thepipeat avolumeloss of between1and2%. The
length of thegap grewfrom2.9 tunnel diameters at
lowvolumelossto3.4tunnel diametersat10%tunnel
volume loss. Bending moments induced in the pipe
increasefromtheonset of tunnel volumelossanddo
not appear to be sensitive to the growth of the gap
height.
REFERENCES
Attewell, P.B.,Yeates, J. &Selby,A. R. 1986. Soil movements
inducedby tunnellingandtheir effects onpipelines and
structures. BlackieandSonLtd, UK.
J acobsz, S.W. 2002. The effects of tunnelling on piled
foundations. PhDThesis, Universityof Cambridge.
Klar, A., Vorster, T. E. B., Soga, K. &Mair, R. J. 2005. Soil
PipeInteractiondueto tunnelling: Comparisonbetween
Winkler andElasticContinuumSolutions. Geotechnique,
55(6), 461466.
Klar, A. & Marshall, A.M. 2007. Shell versus beamrepre-
sentationof pipesintheevaluationof tunnelingeffectson
pipelines. accepted, Tunnelling and Underground Space
Technology.
Klar, A., Marshall, A.M., Soga, K. & Mair, R.J. 2007. Tun-
neling effects on jointed pipelines. accepted, Canadian
Geotechnical Journal.
Mair,R.J.&Taylor,R.N.1997.BoredTunnellingintheurban
environment. Proceed. 14th International conference on
soil mechanics and foundation engineering. Hamburg:
Balkema. 4:23532385.
Peck, R.B. 1969. Deep excavation and tunnelling in soft
ground. Proceed. 7th International conference on soil
mechanics and foundation engineering. Mexico City.
266290.
Vorster, T. E. B. 2005. The effects of tunnelling on buried
pipes. PhDThesis. CambridgeUniversity.
Vorster,T.E.B.,Klar,A.,Soga,K.&Mair,R.J.2005.Estimat-
ingtheEffectsofTunnelingonExistingPipelines.Journal
of Geotechnical and Geoenvironmental Engineering, 131
(11), 13991410.
White, D. J.,Take,W.A. &Bolton, M. D. 2003. Soil deforma-
tionmeasurementusingparticleimagevelocimetry(PIV)
andphotogrammetry. Geotechnique, 53(7), 619631.
707
Groundmovement andearthpressureduetocircular tunneling: Model tests
andnumerical simulations
H.M. Shahin, T. Nakai, F. Zhang, M. Kikumoto,Y. Tabata& E. Nakahara
Nagoya Institute of Technology, Nagoya, Japan
ABSTRACT: Two-dimensional (2D) model testsontunnel excavationusinganewlydevelopedcircular tunnel
apparatusarecarriedout. Numerical simulationsarealsoconductedusingfiniteelement methodunder plane-
strain and drained conditions. In the finite element analyses, elastoplastic subloading tij model is used as a
constitutivemodel of thegroundmaterial. Fromthemodel testsitisrevealedthatinthecaseof thesamevolume
lossduetotunnel excavation, surfacesettlement andearthpressurearoundtunnel aresignificantly influenced
by thedisplacement appliedat thetunnel crownfor thesameoverburden. Thevolumeloss is less significant
comparetothecrowndrift inthecaseof shallowtunneling. Thenumerical resultsshowvery goodagreement
withtheresultsof themodel tests.
1 INTRODUCTION
Shallow tunneling is one of the essential methods
to make underground space in urban area. With the
ongoing demand of tunneling technology research
worksongroundmovementsandearthpressuresdue
to tunnel excavationhavebeenpotentially increased.
The trap door apparatus has been used to investi-
gatethemechanismof tunneling problems by many
researchers (MurayamaandMatsuoka, 1971; Adachi
et al. 1994, Nakai et al., 1997; Shahin et al., 2004).
Inour previouswork, wehavebeencarriedout labo-
ratorymodel testsusingtrap-door tunnel apparatusto
investigatethedeformation mechanismand redistri-
butionof stresssurroundingthetunnel. Toinvestigate
thedeformationmechanismandearthpressureof the
groundmoreprecisely, anewandmorerealistictunnel
apparatushasbeendevelopedwherethecrosssection
of thetunnel is circular. Theapparatus can simulate
variousexcavationmethodssuchasfull faceexcava-
tion, topandsidedrift andbenchcut excavation. This
paper reports 2D model tests usingthenewly devel-
opedcirculartunnel apparatusandnumerical analyses
using thesubloading t
ij
model. This model can con-
sider influenceof intermediateprincipal stressonthe
deformationandstrengthof soils, Dependenceof the
directionof plasticflowonthestresspaths, Influence
of density and/or confiningpressureonthedeforma-
tionandstrengthof soils. Model tests areperformed
with different soil covers and for different excava-
tionpatterns.Thedeformationmechanismisdescribed
focusing the ground movements and shear strain
patterns.
Figure1. Schematicdiagramof 2Dtunnel apparatus.
2 DESCRIPTIONOF 2DMODEL TEST
2.1 Apparatus of model test
Figure 1 shows a schematic diameter of 2D tun-
nel apparatus. Figure2representsanewly developed
model tunnel withcircularcrosssection.Itconsistsof a
shimatthecenterof thetunnel surroundedwith12seg-
ments. Thesegmentsarestronglytightenedall around
theshimwithrubberband. Onemotorisattachedwith
theshimtopull it out inthehorizontal direction. The
model tunnel iskeptinspacewithavertical shaft, and
can be moved in the vertical direction with another
motor. Therefore, thedeviceconsistsof twomotors-
oneisfor shrinkingthetunnel andtheother for mov-
ing thetunnel vertically to fix it at achosen ground
709
Figure2. Circular tunnel device.
depth. Itispossibletomakethesemotorsworksimul-
taneously and individually together with controlling
thespeed of themotors. With themotor theshimis
pulledoutgraduallywhichchangesthediameterof the
shim, consequentlythesegmentsmoveinwardandthe
diameter of thetunnel isreduced. Changingtheshape
of theshimdifferentkindsof excavationprocess, such
as full faceexcavation, top and sidedrift and bench
cut excavationcanbereproducedwiththisapparatus.
Thereduction of tunnel diameter and theamount of
radial shrinkageareobtainedfromadial gaugeread-
ing which is determined fromthe calibration result.
Thevertical movement (if requires to impose) of the
tunnel isalsomeasuredwithanotherdial gauge.There-
fore, theshrinkageof thetunnel canbeattainedina
controlledmanner, whichcansimulatethecondition
of areal tunnel construction.
In the apparatus 12 load cells are used to mea-
sure earth pressure acting on the tunnel. The load
cellsareattachedwiththeblock whichisplacedsur-
rounding thesegments of thetunnel. Each load cell
block is 2.35cmin width and 5.0cmin length. The
blocks aretightly fastened with rubber band. There-
fore, earth pressurecan beobtained at 12 points on
theperiphery of thetunnel at atime. Earth pressure
can also be obtained at other positions by rotating
the tunnel. However, in this case it will be required
to makethemodel groundonceagain. Includingthe
loadcell bocksthetotal diameter of themodel tunnel
is10.0cm. Thecircular tunnel deviceisplacedonan
irontablethat wasusedfor thetrapdoor tunnel appa-
ratus (Nakai et al., 1997; Shahinet al., 2004). It has
10moveableblocksabovewhichthegroundismade.
Thereasonof usingthis typeof baseis to adjust the
initial stress conditionof thegroundsuchaway that
thestress distributionbecomes similar to theground
without tunnel. Thesurfacesettlement of theground
is measured using a laser type displacement trans-
ducer withanaccuracyof 0.01mmanditspositionin
thehorizontal directionisincurredwithasupersonic
wavetransducer. Photographsaretakenduringexper-
iments which arelater on used as input datafor the
simulationof groundmovementswithParticleImage
Velocimetry.
Figure3. Schematicexplanationof excavationpatterns.
2.2 Model ground and excavation patterns
Firstly, thetunnel deviceissetataheightof 10cm; the
height is measuredfromthebottomboundary to the
tunnel invert. Varyingthedistancebetweenthetunnel
invert andthebottomboundary, several experiments
wereconducted.Itwasfoundthatthisheight(10cm)is
freefromtheinfluenceof thebottomboundary. After
settingthetunnel device, massof aluminumrods, hav-
ingdiametersof 1.6and3.0mmandmixedinaratioof
3:2inweight, isstackeduptoaprescribeddepth. The
unit weight of thealuminumrodmassis20.4kN/m
3
,
and thelength is 5.0cm. Theinitial ground is made
insuchawaysothat theearthpressurebecomessim-
ilar to theearthpressureat rest adjustingthebottom
moveableblocksof theapparatus.Greatcareistakento
makeauniformgroundandnottoapplyanyundesired
loadintheground.
In this study two types of excavation patterns are
considered. Pattern 1 (full face excavation) corre-
spondstotheexcavationwherethecenterof thetunnel
iskept fixedandthediameter of thetunnel isreduced
applyingshrinkageof 4mmall aroundthetunnel as
shown in Figure 3. Pattern 2 represents the excava-
tion pattern where the invert is kept fixed (top drift
excavation).Thisisobtainedbydescendingthetunnel
during the application of shrinkage. Here, the same
amount of shrinkage(4mm) is applied. However, as
the center is moved downward by 4mmthe amount
of imposeddisplacementatthetunnel crownis8mm.
Inthebothexcavationpatternsthevolumelossof the
ground is the same, which is equal to 15.36%. The
model testshavebeenconductedforfourkindsof over-
burdenratio, D,B equals0.5, 1.0, 2.0and3.0, where
D is thedepthfromthegroundsurfaceto thetopof
thetunnel andB (10cm) isthewidthof thetunnel. In
Figure3, d
r
representstheamount of shrinkageinthe
radial directiontowardsthecenterof thetunnel, andd
c
indicatestheamountof descendedof thetunnel center.
3 NUMERICAL ANALYSES
Figure 4 shows the mesh used in the finite element
analyses. Isoparametric4-nodedelementsareusedin
themesh. Bothvertical sidesof themesharefreeinthe
710
Figure4. FEM mesh(D/B=2.0).
Table1. Material parametersfor aluminumrods.
0.008
0.004
N (e
NC
at p =98kPa 0.3 Sameparametersas
&q =0kPa) Cam-claymodel
R
CS
=(
1
/
3
)
CS(comp.)
1.8
e
0.2
1.2 Shapeof yieldsurface
(sameasCam-clay
at =1)
a 1300 Influenceof densityand
confiningpressure)
vertical direction, andthebottomfaceiskeptfixed.To
simulatethetunnel excavation, horizontal andvertical
displacementsareappliedtothenodesthatcorrespond
tothetunnel block. Analysesarecarriedout withthe
sameconditionsof themodel tests. Inthecaseof the
fixed invert excavation, both horizontal and vertical
displacementsareappliedtothenodescorrespondto
theblocks of thetunnel intheexperiment except the
bottommost blocks, whereonly horizontal displace-
ments are applied to the corresponding nodes. The
displacements patterns have been decided fromthe
actual movements of theground in theexperiments.
Two-dimensional finiteelementanlysesarecarriedout
withFEMtij-2Dusingthesubloadingt
ij
model (Nakai
and Hinokio, 2004). Model parameters for the alu-
minumrodmassareshowninTable1.Theparameters
arefundamentally thesameasthoseof theCamclay
model except the parameter a, which is responsible
for the influence of density and confining pressure.
Theparameter representstheshapeof yieldsurface.
Theparameterscaneasilybeobtainedfromtraditional
laboratorytests. Figure5showstheresultsof thebiax-
ial tests for the mass of aluminumrods used in the
model tests. Thefigureshows thepositiveandnega-
tivedilatancy of aluminumrod mass; and it is clear
that the strength and deformation behavior is very
similar to those of dense sand. The dotted lines rep
resent thenumerical results for aconfining pressure
of 1/100timestheconfiningpressureof experiments.
Fromthe stress-strain behavior of the element tests
simulatedwithsubloadingt
ij
model, it isnoticedthat
Figure5. Stress-strain-dilatancyrelation.
Figure6. Observedsurfacesettlement profiles.
this model can express the dependency of stiffness,
strengthanddilatancyonthedensityaswell asonthe
confiningpressure.Theinitial stresses, correspondent
to thegeostatic (self-weight) condition, areassigned
tothegroundinall numerical analyses.
4 RESULT ANDDISCUSSION
4.1 Surface settlement
Figure6 shows theobserved troughs of surfaceset-
tlement in the full surface excavation and top drift
excavationfor theamount of shrinkaged
r
=4mmin
thecaseof D,B=1.0and2.0. Figure7representsthe
computedresultscorrespondingtotheobservedones.
Theabscissarepresentsdistancefromthecenterof the
tunnel, while the vertical axis shows the amount of
surfacesettlement. For bothpatternsof excavationthe
maximumsurface settlement occur vertically above
thetunnel crown. Surfacesettlementsbecomesmaller
withtheincreaseof thetunnel depth, but theyextend
over a wider region similar to theprevious research
conducted with trap door tunnel apparatus (Shahin
et al., 2004). The shape of the surface settlement
711
Figure7. Computedsurfacesettlement profiles.
profiles is thesamefor all thesoil covers inthecase
of full faceexcavation. For thesamevolumelossthe
maximumsurfacesettlementislargerinthecasewhere
theinvertisfixedthanthatof thefull faceexcavation.
Thisisbecausetheapplieddisplacement at thecrown
is8mmfor thefixedinvert and4mmfor thefull face
excavationthoughthevolumelossisthesame.Surface
settlement occurs locally for 8mmapplied displace-
mentinthecaseof fixedinvert,consequentlytheshape
of surfacesettlementprofilevarieswiththesoil cover
inthiscase. Thetendencyof larger surfacesettlement
forthefixedinvertismoresignificantuptoD/B=2.0.
However, inthecaseof D/B=3.0(notshownhere) the
differenceof thesurfacesettlement significant. From
theseresultsitisrevealedthatforthesamevolumeloss
surface settlement profiles vary with the excavation
patternsinthecaseof shallower tunneling. Therefore,
thesurfacesettlement may not beproperly estimated
using themethod of volumeloss (Mair et al., 1993)
for shallowtunneling. Theresultsof numerical analy-
sesshowthesametendencyof model testsnotonlyin
shapebut alsoinquantity.
4.2 Shear strains
The distribution of shear strain of the model tests
are obtained fromthe simulation of Particle Image
Velocimetry (PIV) technique. The PIV is originally
developed in the field of fluid mechanics (Adrain,
1991). In this paper, two images are divided into a
finitearea;theaveragemovementrateof thealuminum
rods of each area is being output as nodal displace-
ment. Thestrainfor onegridiscalculatedfromthese
displacements by using the shape function and the
J acobianmatrixthatareusedinfiniteelementmethod
fordisplacementandstrainrelationship. Figures8and
9 show the distribution of shear strain for full face
excavation and fixed invert excavation, respectively,
inthecaseof D/B=1.0and2.0for d
r
=4mm. It is
seen in Figure 8 that the shear band of the ground
Figure8. Distributionof shear strain: center isfixed.
is developed fromthe tunnel invert and covered the
entiretunnel duringtunnel excavationfor thefull sur-
face excavation. FromFigure 9 it is seen that shear
banddevelopsfromthesideof thetunnel notfromthe
tunnel invert. Inthiscasethelengthof theshearbandis
longer thanthat of thefull faceexcavation. Therange
of thedeformedregionfor thefixedinvertisnarrower
comparetothefull faceexcavation. Thedifferentpat-
terns of shearing strain dueto thedifferent types of
thetunnel excavation lead thechangeof theground
behavior. Moreover, theshear strainof thenumerical
analysesshowsvery goodagreement withtheresults
of themodel tests.
4.3 Earth pressure
Figures 10and 11showtheobserved and computed
earth pressuredistributions for D/B=1.0 and 2.0 in
thecaseof full faceexcavationandfixedinvert exca-
vation, respectively.Theplotsaredrawninthe12axes
corresponding to the radial direction of the 12 load
cellstowardsthecenter of themodel tunnel. Thefig-
ures represent the value of earth pressure in Pascal
correspondingtotheamount of applieddisplacement
(amount of shrinkage). It is seen in Figure 10 that
earthpressuredecreasesall aroundthetunnel for the
full face excavation due to the arching effect. The
results appear to bein agreement with theresults of
tunnel experimentsperformedbyMurayamaandMat-
suoka, 1971; Adachi et al., 1994; Shahinet al. 2004.
As shear band develops surrounding the entire tun-
nel (Fig.8) thesurrounding ground undergoes to the
712
Figure9. Distributionof shear strain: invert isfixed.
Figure10. Distributionof earthpressure: center isfixed.
loosen state which reduces stresses in that place. It
is also noticedthat theearthpressuredecreases sud-
denly after applying shrinkage of the tunnel within
0.00 to 0.20mm. Further shrinking thetunnel, earth
pressure decreases gradually at a lower rate up to a
certainextent after whichtheearthpressurebecomes
almost constant. Sudden changein earth pressureis
dueto soil arching, immediately after disturbing the
ground. For thefixedinvert excavationearthpressure
Figure11. Distributionof earthpressure: invert isfixed.
distributionsaredifferentfromthefull surfaceexcava-
tion.Inthiscaseearthpressuredecreasesall aroundthe
tunnel till d
r
=1mm, for further shrinkageof thetun-
nel itincreasesinthebottompartof thetunnel whileit
remainsalmostsameintheupper partof tunnel. Itcan
beexplainedwiththeshear straindistributionshown
inFigure9. Asthegroundbecomesloosenonlyinthe
upper part of thetunnel after d
r
=1mm, theconfin-
ingpressureinthebottompartincreases, therefore, the
increaseof earthpressureinthebottompart of tunnel
canbespeculated. Fromtheabovediscussions it can
besaidthatthedistributionof earthpressureishighly
dependent ontheexcavationpatterns.
Figure 12 and 13 illustrates the change of earth
pressureat load cells 4 and 9 against theamount of
shrinkageof thetunnel for different soil covers. Load
cell 4 is located in thevicinity of thetunnel crown,
andloadcell 9isinthepart of thetunnel invert. This
figure confirms the sharp change of earth pressure
during tunnel excavation. In thecaseof fixed invert
excavation(Fig. 13), atthepositionof loadcell 9earth
pressures increases after aroundd
r
=1mm, whichis
different fromtheresults of thefull faceexcavation
(Fig. 12) whereearth pressureremain constant after
thatamountof shrinkage.Thephenomenonof theearth
pressure increase after the reduction to some extent
canbedescribedasthechangeof archingeffectdueto
thenon-linearandelastoplasticbehaviorof theground
materials. It cannot bedescribedwithausual linear
elastic model. The results of numerical analyses are
in good agreement with the results of model tests.
Therefore, it can be said that a proper elastoplastic
713
Figure12. Earthpressurehistory: center isfixed.
Figure13. Earthpressurehistory: invert isfixed.
constitutivemodel isrequiredtopredictearthpressure
aroundtunnel for liningdesign.
5 COMPARISONSWITHTHE ELASTIC
ANALYSES
In this study numerical analyses with a linear elas-
tictheoryhasbeencarriedout tocomparetheresults
withtheelastoplasticanalyses. Thissectiondescribes
sometypical resultsof theanalyses.Youngsmodulus
for theelastic analyses is calculatedfromthestress-
strain relation (Fig. 14) of biaxial test performed in
laboratory for themass of aluminumrods. Thevalue
of E =5500kPa is chosen fromthe figure, and the
assumedvalueof Poissonsratiois0.33fortheground
of aluminumrodsmass.
Figure15showsthesurfacesettlement profilesof
themodel test, elastoplasticanalysisandelasticanaly-
sisforsoil coverD/B=1.0. Itisseeninthisfigurethat
elastoplasticanalysiscanpreciselyexpresstheresults
Figure14. Stress-strainrelationof aluminumrodsmass.
Figure15. Comparisonsof surfacesettlement profiles.
of themodel test.Theelasticanalysisproducesawider
surface settlement profiles compare to the observed
one, the maximumsurface settlement is smaller as
well. Asthereisnoyieldpointinaliner elasticmodel
itcannotexpressthedeformationoccurredlocally. In
this analysis displacement is appliedtosimulatetun-
nel excavation, therefore, there is no relation of the
magnitudeof theYoungsmodulusintheshapeof the
settlement troughexcept thevalueof Poissonsratio.
6 CONCLUSIONS
To investigatethedeformationmechanismandearth
pressure of the ground, a new tunnel apparatus has
beendeveloped.Withthisapparatus2Dmodel testand
elastoplasticfiniteelementanalyseshavebeencarried
out. Fromthemodel testsandnumerical analyses, the
followingpointscanbeconcluded:
1. Surfacesettlement andearthpressurearoundtun-
nel aresignificantlyinfluencedbythedisplacement
appliedatthetunnel crownforthesameoverburden
andthesamevolumeloss.
2. Thevolumelossislesssignificant comparetothe
crowndrift inthecaseof shallowtunneling.
3. Thefull faceexcavationproduceswiderrangeof the
deformationregioncomparetothetopdrift (fixed
invert) excavation.
714
4. Thegrounddeformationmechanismsaredifferent
for thesamevolumelosswithdifferent excavation
patterns.
5. Thedistributionof earthpressureishighly depen-
dent ontheexcavationpatterns.
The finite element analysis with subloading t
ij
model is a useful tool to predict earth pressure and
groundbehavior duringtunnel excavation.
REFERENCES
Adachi, T., Tamura, T., Kimura, M. & Aramaki, S. 1994.
Earthpressuredistributionintrapdoortests: Proc. of 29th
Japan National Conference of SMFE, 3, 19891992 (in
J apanese).
Adrain, R. J. 1991. Particle imaging techniques for exper-
imental fluid mechanics: Ann. Rev. Fluid Mech. 23,
261304.
Murayama, S. and Matsuoka, H. 1971. Earth pressure on
tunnelsinsandyground: Proc. of JSCE, 187: 95108(in
J apanese)
Mair, R.J., Taylor, R.N. & Bracegirdle, A. 1993. Subsurface
settlement profilesabovetunnelsinclays. Geotechnique
43, No.2, 315320.
Nakai, T., Xu, L. &Yamazaki, H. (1997): 3Dand2Dmodel
testsandnumerical analysesof settlementsandearthpres-
sure due to tunnel excavation, Soils and Foundations,
37(3), 3142.
Nakai,T.,andHinokio,M.2004.Asimpleelastoplasticmodel
material parameters. Soils and Foundation. 44(2): 5370.
Shahin, H. M., Nakai, T., Hinokio, M., Kurimoto, T., and
Sada, T. (2004): Influenceof surfaceloadsandconstruc-
tionsequenceongroundresponseduetotunneling, Soils
and Foundation, 44(2), 7184.
715
Analysisof pre-reinforcedzoneintunnel consideringthe
time-dependent performance
K.I. Song, J. Kim& G.C. Cho
Department of Civil & Environmental Engineering, Korea Advanced Institute of Science and Technology,
Daejeon, South Korea
ABSTRACT: Auxiliarysupportsystemssuchasthereinforcedprotectiveumbrellamethodareappliedbefore
tunnel excavationtoincreasegroundstiffnessandtoprevent thelargedeformationinsoft groundandshallow
depthtunnelling. This study suggests amethodto characterizethetime-dependent behavior of pre-reinforced
zonesaroundthetunnel usingelasticwavesanddirect shear test. Theresultsobtainedfromthelaboratorytests
areappliedtonumerical simulationsof atunnel consideringitsconstructionsequences. Accordingtonumerical
analyses, thetime-dependenttunnel stabilityismostcritical intheinitial installationpartof pre-reinforcedzone
andtheportal of tunnel. However, time-dependenteffectontunnel behaviorisnotsignificantduringconstruction
aslongasaproper overlaplengthisapplied. Finally, thesuggestedanalysismethodof combiningexperimental
and numerical procedures that consider the time-dependent effect on the pre-reinforced zone on the tunnel
behavior will providereliableandpractical designandanalysisfor tunnelsinsoft ground.
1 INTRODUCTION
Recently, instancesof theconstructionof largeunder-
ground structures in soil and soft rock layers are
increasing.Auxiliarysupportsystemssuchastherein-
forcedprotectiveumbrellamethodareappliedbefore
or duringtunnel excavationto ensurethestability of
the tunnel face, to increase ground stiffness and to
preventlargedeformationsinsoftgroundduringshal-
lowdepthtunneling. Theseauxiliarysupportschange
thestateof stressandderivethearchingeffectaround
thetunnel. However, theeffectsof pre-reinforcement
on tunnel stability and waterproofing of large sec-
tiontunnelsinsoft groundhaveyet tobeclearly and
quantitativelydefined.
In conventional tunneling, there are typically 1
to 2 days between one face and the next face after
thepre-reinforcement step. Duringthistimeinterval,
it is known that changes of the material properties
caused by theeffect of curing of thegrouting mate-
rial exist. However, 28days of stiffness andstrength
afterconstructionaregenerallyappliedtothematerial
properties of thepre-reinforcedzoneindesignstage
without consideringtheeffect of thetime-dependent
behavior of theinjectedgrout material.
The present study suggests a method to charac-
terizethetime-dependent behavior of pre-reinforced
zonesaroundalargesectionof tunnel insoft ground
usingelasticwaves.Anexperimental analysiswasper-
formed to characterize the time-dependent behavior
of the pre-reinforced zone. Direct shear tests were
performedat different timestagestoobtainthetime-
dependent strength parameters. In addition, elastic
wave velocities (i.e., Vp and Vs) were continu-
ouslymeasuredusingpiezoelectricbenderelementsto
obtainthetime-dependent stiffness parameter. Time-
dependent strengthandstiffnessparametersobtained
fromlaboratory tests were applied in the numerical
simulation of a large section tunnel in soft ground,
takingintoaccountitsconstructionsequence.Thepro-
posedanalysismethod, whichcombinesexperimental
andnumerical procedureswhileconsideringthetime-
dependent effect on the pre-reinforced zone on the
tunnel behavior, will provideareliableand practical
design basis and means of analysis for largesection
tunnelsinsoft ground.
2 TIME-DEPENDENT EXPERIMENTAL
ANALYSIS
2.1 Measurement of elastic wave velocity and shear
strength
AsshowninFigure1(a), S-waveandP-wavevelocity
were measured in this study using bender elements.
Thepropagationdirectionwasvariedbyarrangingthe
installationdirectionof thebenderelementsuchthatit
wasparallel for thep-waveandperpendicular for the
s-wave. Wavevelocities weremeasuredcontinuously
717
Figure1. Schematicdrawingof test setup.
Table1. Material propertiesfosamplesusedintest.
J oomunjin
Material properties StandardSand PortlandCement
Frictionangle(
) 39
Min. voidratio 0.62
Max. voidratio 1.13
Density(kN/m
2
) 25.9 30.8
Realtivedensity(%) 60W/C ratio(%) 170
Vertical Stress 160Dimension 10(W)10(D)
(kPa) (cm) 7(H)
for 28 days under 160kPa vertical stress. From
the wave propagation transition, arrival time was
calculated. The wave velocities were calculated by
dividingtheheight of thespecimenby themeasured
arrival time(Dano, 2004; Khan, 2006).
A schematicdrawingof themotorizeddirect shear
apparatus is shown in Figure 1(b). The shear speed
of the direct shear apparatus was maintained at
0.5mm/min.A loadcell of 2toncapacitywasinstalled
to calibratetheload during shearing. Measured ver-
tical and horizontal displacements were saved in a
computer automatically using a LVDT. All material
properties of samples used in the tests are listed in
Table1.
0
200
400
600
800
1000
1200
1400
0 5 10 15 20 25 30
Curing time (Days)
V
e
l
o
c
i
t
y

(
m
/
s
)
P-wave
S-wave
(a) Elastic wave velocity (Normal scale)
(b) Elastic wave velocity (Log scale)
0
200
400
600
800
1000
1200
1400
0.001 0.01 0.1 1 10 100
Curing time (Days)
V
e
l
o
c
i
t
y

(
m
/
s
)
P-wave
S-wave
6 hours
Figure 2. Time-dependent characteristics of elastic wave
velocities.
2.2 Results and analyses of experimental tests
2.2.1 Time-dependent characteristics of wave
velocities
The results show that the wave velocity increases
exponentially accordingto curingtimeandbecomes
almostconstantafter7days. Figure2(a) showstypical
velocity datafor two types of measured waves. It is
found that the strength of the mixture becomes an
importantfactor fromabout6hours.AsshowninFig-
ure2(a), P-wavevelocityisfasterthanS-wavevelocity
andrepresentsthewavethattravelsthroughtheskele-
tonof sandandcementmixture. FromFigure2(b), itis
seenthatthewavevelocitystartstoincreasewhenthe
curingtimeis46hours. Thismeansthatthemixture
takes46hourstoreveal acementationeffect.
2.2.2 Time-dependent characteristics of shear
strength and strength parameters (c and )
The direct shear tests performed for specimens less
than 7 days of curing time correspond to the wave
velocity measurement. Theresultsof thedirect shear
test were very similar to those of the wave veloc-
itymeasurements. However, thetestswereperformed
withintervals of aday or more, andthus theprocess
of thestrengthrevelationthroughcementationcould
not beobserved.
Figure3showsthetimedependentcharacteristicsof
shear strengthparameters obtainedfromdirect shear
test. InFigure3(a), thefrictionangledoesnot change
in accordance with curing time. On the other hand,
718
Figure3. Time-dependent characteristicsof shear strength
parameters.
Figure 3(b) shows time-dependent characteristics of
cohesion. Through Figure 3(b), it is noticeablethat
thecohesionincreaseswithcuringtimeandafteracer-
tainamount of curingtimethecohesionconverges. It
canbededucedthatthebondingof cementcausesthis
increaseincohesion. Therefore, strengthiscontrolled
bythenormal stress(i.e., frictional) attheearlystages
and then by cohesion (Schnaid et al., 2001). In the
numerical analysis, itisassumedthatthefrictionangle
doesnot changeandthecohesionincreasesascuring
timeincreases.
Ultimately, thecohesioncontrolstheincreaseof the
shearstrength, andcohesiondependsontimeinasim-
ilar manner as elastic wavevelocity. Hence, Youngs
modulus and cohesion are selected as improvement
propertiestoreflect time-dependent characteristicsin
order tofacilitateaccuratesimulationviathenumeri-
cal analysisprograminthisstudy.
2.3 Verification of testing results and acquirement
of time-dependent design properties
Anempirical determinationmethodof pre-reinforced
zonestoinvestigatethestabilityinatunnel ispresented
and the results are compared with characteristics of
elasticwavevelocityfromlaboratorytest results.
As shown in Eq. (1), cohesion is related to the
unconfinedstrengthandfrictionangle. Herein, uncon-
finedstrengthof groutbulbat7daysof curingtimeis
about 8MPafromtheuniaxial compressivestrength
tests. Whenthefrictionangleof asoil layer innatural
conditionisabout39degrees, thecohesionof thepre-
reinforcedzoneisabout1.91MPaaccordingtoEq.(1).
Figure 4. Normalized shear modulus of sandy soil (Kim
et al. 2004).
Sandysoil innatural conditionhasnocohesioninthe
direct shear test. However, after 7days, it hasacohe-
sionvalueof about 2.17MPa, whichissimilar tothe
result obtainedbyEq. (1).
Youngsmodulusof thegroutbulbcanbepresented
asafunctionof theweight densityanddesigncriteria
strengthintheconcretedesignspecification, asgiven
byEq. (2).
Here, when the weight density is 18kN/m
2
and
theunconfinedstrengthat acuringtimeof 7days is
8MPa, Youngsmodulusisabout 8300MPa. Theini-
tial Youngs modulus (i.e., 45Mpa) canbecalculated
fromequation(3)(6), this represents anincreaseof
184times.
Regardingthecharacteristicsof elasticwavepropa-
gation, themaximumshearmoduluscanbecalculated
byEq. (3) fromthemeasuredshear wavevelocity.
Thenormalizedshear modulus intherangeof the
tunnel strainlevel canbeobtainedby aresonant col-
umnandtorsional sheartestandtheRamberg-Osgood
model (1943), as shown in Figure 4. Typically, the
strainlevel of atunnel is about 0.11%. Hence, the
normalizedshear modulusof sandy soil isabout 0.3,
as shown in Figure4. Accordingly, theshear modu-
lusof thepre-reinforcedzonecanbecalculatedbythe
maximumshear modulus fromEq. (3) and the nor-
malized shear modulus in Eq. (4). Furthermore, the
Youngs modulus of model canbederivedby Eq. (5)
719
andEq. (6), whichdelineatestherelationshipamong
Youngsmodulus, shear modulus, andPoissonsratio.
Although it is clear that themeasured shear wave
velocity has characteristics of infinitesimal deforma-
tion, it is reasonable that the increase in Youngs
modulus can vary according to any increase in the
elasticwavevelocitydependingonthetime.
With these equations and theory, Youngs mod-
ulus of sandy soil in natural condition is 45MPa
and Youngs modulus for a curing time of 7 days
is 7400MPa. Therefore, Youngs modulus of a pre-
reinforcedzoneincreases165timescomparedtothat
beforepre-reinforcement. Itisapparentfromtheequa-
tions(3)(6)presentedherethatstiffnessincreasesat
asimilar ratetotheresultcalculatedbyEq. (2). More-
over, the Youngs modulus of pre-reinforced sandy
soil after 28 days of curing time increases by about
200timesincomparisonwithnormal conditionsandy
soil. Basedontheseresultsandequations, numerical
analysesshouldbeaccompaniedtoevaluatethetime-
dependent characteristicsof pre-reinforcedzonesina
tunnel.
Youngs modulus andcohesionof apre-reinforced
zone should be expected to increase by 200 times
increaseafter acuringtimeof 28days.
3 NUMERICAL SIMULATIONOF LARGE
SECTIONTUNNEL INSOFT GROUND
3.1 Large section tunnel model and boundary
conditions
Comparative analyses were performed to analyze
thetime-dependent effect of apre-reinforcedzoneon
thelargesectiontunnel behavior whiletunneling. The
material properties of the pre-reinforced zone have
fivedifferentconditionsthatareeithertime-dependent
conditions or constant time conditions (i.e., 1d, 2d,
3d, and 28d). A commercial 3D FEM analysis pro-
gram(i.e., MIDAS-GTS) was used as a numerical
simulationtool. Elasto-plasticgroundmaterial, which
is linear elastic and perfectly plastic with following
Mohr-Coulombyieldcriterionandnonassociatedflow
rule, wasusedfor analyses.
For the application of the stiffness and strength
parameters dependingonthetimeobtainedfromthe
experimental study, atunnel model shouldbelocated
at thesamestress level of theexperiment condition,
as propagation characteristics of elastic waves are
affectedbythestateof thestress.
Figure 5. 3D tunnel model and time-dependent material
propertiesof thepre-reinforcedzoneafter 12mexcavation.
Figure5shows thesimulated3D four-lanetunnel
model, whichwasconstructed15munder thesurface
of theground. It hasaradiusof 9.4mandaheight of
10.4m. Theboundaryof 3DFE model ishorizontally
fixedat the4endsections andvertically fixedat the
bottomsection. Water tableislocatedat thetopof the
surface. Thissimulatedtunnel isanactual sectionof a
four-lanetunnel that wasdesignedandconstructedin
weatheredrock inKorea. Four sub-sectional excava-
tionswithalengthof 0.75mper stageweremodeled.
Steel pipes 114.3mmindiameter weremodeledas a
beamelement: a 12msteel pipe was installed at an
inclination of 11
with a 6moverlap, with a 0.7m

transversal interval betweenthesteel pipes that were
installed along the tunnel crown 60
fromthe cen-
ter. Thegrout expansionradiusis0.3mfromthepipe
center.
3.2 Numerical analysis of a tunnel considering the
time-dependent behavior
Thetime-dependentbehavior of apre-reinforcedzone
canbemodeledusingtheproceduresdescribedbelow.
Thematerial properties(i.e., stiffnessandstrength) of
thepre-reinforced zoneareregistered as thebound-
ary conditions fromDay 1to Day 28. Theregistered
initial boundary conditions were applied to a pre-
assignedmeshinthepre-reinforcement construction.
The boundary conditions were applied and updated
accordingto thefieldconstructionprocess. Figure5
shows the conceptual time-dependent stiffness and
strengthcharacteristicsof thepre-reinforcedzoneafter
thethirdpre-reinforcement and12mof excavationat
anexcavationspeedof 0.75m/day.
Thetunnel behaviorswereanalyzedconsideringthe
time-dependent effect of the pre-reinforced zone in
terms of the vertical displacement and the horizon-
tal displacement. Thematerial propertiesusedfor the
analysis aretabulatedinTable2. Figure6shows the
time-dependent elastic modulus andcohesionvalues
obtainedfromtheexperimental studyaswell asthose
usedwiththenumerical analysis.
720
Table2. Material propertiesusedforthenumerical analysis.
Ground (kN/m
2
) E (MPa) c(kPa) (
)
Weatheredsoil 18 45 0.32 0 39
Weatheredrock 21 200 0.3 250 35
Steel Pipe 33 210000 0.3
where: unit weight, E: elastic modulus, : Poissonsratio,
c: cohesion, : frictionangle.
Figure6. Time-dependentstiffnessandstrengthfor numer-
ical analysisandobtainedexperimental result.
Figure7. Normalizedvertical displacement variation.
4 COMPARATIVEANALYSISOF THE
NUMERICAL SIMULATIONRESULT
For quantitativeanalysis, displacements of eachcase
are normalized with the result of a pipe-only case.
Withthenormalizeddisplacements, theeffect of the
time-dependent behavior of the pre-reinforced zone
wasexamined.
4.1 Time-dependent effect on vertical displacement
Figure7shows thenormalizedvertical displacement
at the portal according to the time-dependent con-
dition and constant timeconditions. Thenormalized
vertical displacement behavior of thetime-dependent
condition is similar to the results of one day of the
constant timeconditionsat theinitial excavationsec-
tion. Astheexcavationlengthincreases, theresultsof
time-dependent conditionbecomenearly identical to
Figure8. Normalizedhorizontal displacement variationat
tunnel face.
theanalysisresult of 23daysof constant timecon-
dition, and thevertical displacement converges. The
stiffness and strength of the pre-reinforced zone of
12 days of theconstant timecondition correspond
to3050%of 28daysof theconstant timecondition.
Inotherwords,areductionof thematerial propertiesof
thepre-reinforcedzonemakesitpossibletomodel the
time-dependenteffectof thepre-reinforcedzoneonthe
global tunnel behavior attheinitial tunnel excavation.
4.2 Time-dependent effect on horizontal
displacement at the tunnel face
Figure 8 shows the normalized horizontal displace-
ment at the tunnel face according to the time-
dependent condition and constant time conditions.
Fromtheanalysis result, variationof thenormalized
displacement at thetime-dependent conditionranges
from0.940.98, whichisvery similar toother cases
during excavation. Therefore, pre-reinforcement can
beconsideredas preventionof acollapserather than
adisplacement reductioncontrol at thetunnel face. It
canbeconcludedthat groutingreducesthehorizontal
displacement by approximately 26% at the tunnel
facewiththepre-reinforcement method.
4.3 Time-dependent effect on horizontal
displacement at the tunnel side wall
Figure9showsthehorizontal displacement variation
atthetunnel sidewall accordingtothetime-dependent
conditionandconstanttimeconditions. Fromtheanal-
ysis result, the normalized horizontal displacement
variationof thetime-dependentconditionissimilar to
theresultsof oneday of theconstant timeconditions
at the initial excavation section. The horizontal dis-
placement variation of thetime-dependent condition
is locatedbetween2and3days of theconstant time
condition as the excavation length increases. There-
fore, between2and3daysof aconstant stiffnessand
strength condition can beused for atime-dependent
analysis of a pre-reinforced zone for the horizontal
721
Figure 9. Normalized horizontal displacement at tunnel
sidewall.
Figure 10. Normalized displacement at 10.5mfromthe
portal.
displacement of a tunnel, which is identical to the
variationof thevertical displacement.
Figure10 shows thenormalized vertical and hor-
izontal displacements 10.5m from the portal. Fig-
ures 10(a) and10(b) areingoodagreement withthe
result of 28dayof theconstant timeconditionaswell
as the result of the time-dependent condition of the
normalized vertical and normalized horizontal dis-
placement. Whentheexcavationlengthis10.5m, the
pre-reinforced zoneinitially has 14 days of stiffness
andstrength;secondly,thepre-reinforcementzonehad
6daysof stiffnessandstrength.
Thestiffness andstrengthlevels of 6and14days
of the constant time condition correspond to 87.8%
and 99.3%, respectively, of 28 days of the con-
stant time condition. Therefore, the time-dependent
characteristicsof thepre-reinforcedzonedonotaffect
tunnel displacementduetothesufficientlyoverlapped
pre-reinforced zone. However, the time-dependent
characteristics of the pre-reinforced zone should be
consideredinthecaseof aportal andaninitial support
sectionof aweak layer, whichcancausealargedis-
placement andunsafeconstructionconditions. There-
fore, itcanbeconcludedthat23daysforthestiffness
and strength of pre-reinforced zones is appropriate
to model thetime-dependent behavior. Inaddition, a
properlyoverlappedsectionof pre-reinforcedzonecan
be assumed as 3 or 7 days of constant stiffness and
strengthwithout atime-dependent effect duetosuffi-
cient exposureof thestiffnessandstrengthduringthe
excavationstagesinconservativedesigns.
5 CONCLUSIONS
Based on an experimental study of a pre-reinforced
zone, the time-dependent characteristics were ana-
lyzed in terms of the strength and stiffness. Results
show that the shear strength and the elastic wave
velocities increaseas thetimestageincreases. Shear
strength and strength parameters (i.e., the cohesion
and friction angle) can be uniquely correlated to
elastic wave velocities. The results obtained from
laboratory tests wereapplied to anumerical simula-
tion of atunnel, taking into account its construction
sequence. According to the results of the numeri-
cal simulation, vertical displacement and horizontal
displacement resultsfor fewer than23daysof con-
stant time boundary conditions are nearly identical
to the analysis results of the time-dependent condi-
tion. Therefore, it can be concluded that 23 days
for thestiffness andstrengthof pre-reinforcedzones
is appropriateto model thetime-dependent behavior
of alargesectiontunnel. Finally, thesuggestedanal-
ysis method combining experimental and numerical
proceduresthatconsider thetime-dependenteffecton
thepre-reinforced zoneon tunnel behavior will pro-
videareliableandpractical designbasisandmeansof
analysisfor tunnelsinsoft ground.
ACKNOWLEDGEMENTS
This paper was supported by the Korea Institute of
ConstructionandTransportationTechnology Evalua-
tionandPlanningunder theMinistryof Construction
andTransportationinKorea(Grant No. 04-C01) and
BrainKorea21Project in2006.
722
REFERENCES
Dano, C., Hicher, P.Y. & Tailliez, S. 2004. Engineering
Properties of Grouted Sands. Journal of Geotech. and
Geoenviron. Engineering, 130(13): 328338.
Khan, Z., Majid, G., Cascante, G., Hutchinson, D.J. &
Pezeshkpour, P. 2006. Characterizationof acementedsan
withthepulse-velocity method. Canadian Geotechnical
Journal, 43: 294309.
Kim, D.S., Kwon, K.C. & Oh, S.B. 2004. Development of
Experiment-Analysis Integrated System for the Evalu-
ation of Deformation Behavior of Geotechnical Struc-
tures on Weathered Residual Soils at Whole Strain
Range. Research Report Korea Science and Engineering
Foundation.
MIDAS-GTS. 2005. Geotechnical & Tunnel Analysis
System, MIDASInformationTechnologyCo., Ltd.
Ramberg, W. & Osgood, W.R. 1943. Description of stress-
strain curves by three parameters. Technical Note 902,
National Advisory Committee for Aeronautics, Washing-
ton, D.C.
Schnaid, F., Prietto, P.D.M. &Consoli, N.C. 2001. Character-
izationof CementedSandinTriaxial Compression. Jour-
nal of Geotech. and Geoenviron. Engineering, 127(17):
857868.
723
Vault temperatureof vehiclefiresinlargecross-sectionroadtunnel
K.S. Wang, X. Han& Z.X. Li
ABSTRACT: Onthebasisof CFDmodels, thesimulationanalysisof alorryandcar fireinlargecross-section
roadtunnel was carriedout in this paper. Thevault temperaturewas simulatedandcomparedwith thevalue
calculatedbycorrelativeempirical equation.Thecorrespondingresultswouldbecontributedtoprovideeffective
techniquefor thefireprotectionmeasuresof themajor structureof roadtunnel.
1 INTRODUCTION
Large scale tunnel fires often caused considerable
material damage, not only tovehicles, but alsotothe
tunnel facilitiesaswell.Thedamagewasbroughtabout
by the massive development of heat and aggressive
combustiongases, whichledtoenormousdifficulties
toreconstructionof tunnel. Inordertoprovidetheroad
tunnel with effective fire protection measures, it is
important topredict thevault temperatureindifferent
road tunnel fire scenarios. There were some experi-
ments on testing the structure characteristics under
fire. But most of themweretested by following the
standardtemperature-timecurve(Yasudaetal., 2004).
Veryfewdataareavailableonmaximumsmoketem-
peratureunder thetunnel ceilingwithdifferentceiling
heights, longitudinal ventilation velocities and fire
intensities.BasedonCFDmodels,thesimulationanal-
ysisof alorryandcar fireinlargecross-sectionroad
tunnel was carried out in this paper. The vault tem-
peraturewas simulatedandcomparedwiththevalue
calculatedbycorrelativeempirical equation. Thecor-
responding results would be contributed to provide
effective technique for the fire protection measures
of themajor structureof roadtunnel.
2 DESIGNOF FIRE SCENARIOFOR CFD
2.1 Brief introduction of FDS
Inthesesimulations, FDS(FireDynamicsSimulator)
4.06 which was released by NIST (National Insti-
tute of Standards and Technology, USA) was used.
FDSisaComputational FluidDynamics(CFD)model
withLES(LargeEddySimulation) of fire-drivenfluid
flow. The model solves numerically a form of the
Navier-Stokes equations appropriate for low-speed,
thermally-drivenflowwithanemphasisonsmokeand
heat transport fromfires. The partial derivatives of
theconservation equations of mass, momentumand
energyareapproximatedasfinitedifferences, andthe
solution is updated in time on a three-dimensional,
rectilinear grid. Thermal radiationiscomputedusing
afinitevolumetechniqueonthesamegridastheflow
solver.
2.2 Simulation scenarios of tunnel fire
Inthis paper, CFD simulationof themaximumvault
temperaturewasconductedbyFDS(4.06) indifferent
fire scenarios with constant longitudinal ventilation
velocity set as 3m/s. The simulation length of the
model which was three-driveway largecross-section
roadtunnel withdiameterof 15mwas100m,asshown
inFigure1. TheHeat ReleaseRate(HRR) of thefire
Figure1. Cross-sectionof roadtunnel.
725
Figure2. CFDanalysismodel inscenario1.
Figure3. CFDanalysismodel inscenario2.
sourcewasbasedontheEUREKA projectconclusion,
whichwascompletedbynineEuropecountriesinanti-
quated road tunnel. Thefireposition was located at
37mfromtheupstreamventilationcross-sectionand
63mfromthedownstreamventilation cross-section.
Inthissimulation, thecar wasignitedbythelorryand
thedesignscenariosinclude: scenario1: thecar inthe
flank of the lorry; scenario 2: the car downstream
thelorry. Inthesetwoscenarios, thespacingbetween
thecar andthelorry was 2m, as showninFigures 2
and3.
Inthesimulationprocess, thelorry firstly burned
and for about 220s the car was ignited by the lorry
whichreceivedmoreheat radiant fluxthanitscritical
valueof 16kw/m
2
. ThemaximumHRR of thelorry
and the car was 20MW and 5MW separately. The
total HHRgeneratedbythesetwovehicleswasshown
inFigure4.
3 DESIGNOF FIRE SCENARIOFOR CFD
3.1 HRR curve
As showninFigure4, HRR curves inthesetwo sce-
narios wereconstructed. Therelativeposition of the
lorry and the car had less influence on total HHR.
During the simulation, the HHR was approximately
23MWandlessthanthealgebrasumof 25MWwhich
was calculated by the HHR of these two vehicles
respectively.
0
5000
10000
15000
20000
25000
0 500 1000 1500 2000
TIME (S)
H
R
R

(
K
W
)
scenaio 1
scenaio 2
Figure4. Heat releaserateof firesourceinscenario1, 2.
0.00
50.00
100.00
150.00
200.00
250.00
300.00
350.00
0 200 400 600 800 1000 1200 1400 1600 1800 2000
TIME (S)
T
E
M
P
E
R
A
T
U
R
E

(
C
)
scenaio 1
scenaio 2
Figure5. Vault temperature-timecurvesinscenario1, 2.
3.2 Vault temperature-time curve
As showninFigure5, thevault temperaturereached
themaximumvalueof 321
C at theposition of 4m
downstreamfromthefiresourceafter941sinscenario
1 and achieved 331
C at the position of 5mdown-

streamfromthefiresourceafter 980s inscenario 2.
Thevault temperaturetimecurves werebasically of
superposition in these two scenarios with the same
ventilationvelocityof 3m/sandgeometryproperties.
It wasindicatedthat therelativepositionof thesetwo
vehicleshadlessinfluenceonvaulttemperaturecurve.
AsshowninFigures6and7, becauseof largecross-
sectionof theroadtunnel andrelativelysmaller HRR
valueof thefiresource, thesimulated vault temper-
ature was not much high. On the other hand, under
thelongitudinal ventilationvelocity of 3m/s, thefire
flameswereslopedtowardsdownstreamfar fromthe
vault, so it attenuatedtheenvironmental temperature
andmadelesscontributiontovault temperature.
3.3 Comparative analysis of vault temperature
An empirical equation which was derived to predict
the maximumsmoke temperature under the tunnel
ceilingispresentedasfollows(Kuriokaet al., 2003):
726
Figure6. Fireflamesinscenario1after 940s.
where Q=heat release rate of the fire, kW; Q
=
dimensionless heat release rate; Fr =dimensionless
Froude number; , =experimental constant; C
p
=
specific heat capacity of air at constant pres-
sure, kJ kg
1
K
1
;
a
=ambient air density, kgm
3
;
T
a
=ambient air temperature in tunnel, K; LT
max
=
maximum excess temperature of smoke under the
tunnel ceiling, K; u =representativelongitudinal ven-
tilationvelocity,ms
1
;g =accelerationduetogravity,
ms
2
; H
d
=height fromthesurfaceof firesourceto
tunnel ceiling, m.
For validatingtheavailabilityof equation(1) tothe
vault temperature in large cross-section road tunnel
fires, threeotherscenarioswereintroducedtothesim-
ulation analysis, including: (1) scenario 3: car fire;
(2) scenario 4: passenger car/lorry; (3) scenario 5:
heavygoodsvehicle. Theheat releaseratesof thefire
source were also in accordance with the EUREKA
project conclusion, as shown in Figure 8. The fire
flamesinscenario3,4,5wereshowninFigure9,10,11
respectively.
Figure8. Heatreleaserateof firesourceinscenario3, 4, 5.
Thevault temperaturereachedthemaximumvalue
atthepositionof 4mdownstreamfromthefiresource
in scenario 3, 4, 5 with the same longitudinal ven-
tilation velocity of 3m/s. Thesimulation results and
empirical equation valueof thevault temperaturein
thesescenarioswereshowninFigure12andTable1.
As shown in Figure 13 and Table 1, there were
rather differences betweenthesimulationresults and
727
0
100
200
300
400
500
600
700
800
0 10 20 30 40 50 60
HRR (KW)
T
E
M
P
E
R
A
T
U
R
E

(
C
)
simulation value equation value
Figure12. Thesimulationvalueandtheequationvalueof
vault temperatureindifferent scenarios.
Table 1. The simulation value and the equation value of
vault temperatureindifferent scenarios.
3 4 1, 2 5
Scenario (5MW) (20MW) (23MW) (50MW)
Sv 54 275 (321+331)/2=326 753
Ev 116 312 369 628
Re 53.4% 11.9% 11.7% 16.6%
Sv=simulation value; Ev=equation value; Re=relative
error.
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
0 10 20 30 40 50 60 70 80 90
HRR (KW)
r
e
l
a
t
i
v
e

e
r
r
o
r
relative error
17 68
Figure13. Therelativeerror among thesimulation value
and the equation value of vault temperature in different
scenarios.
the empirical equation values of the vault tempera-
ture. The relative error exceeded 20% for the HRR
whichwerelessthan17MWandgreater than68MW.
Itshouldbementionedthatforsmall scalefire, suchas
HRR of 5MW, therelativeerror approached53.4%.
Therefore, much moresimulation analysis and prac-
tical tests need to becarried out. On theother hand,
it wassuggestedthat theempirical equationcouldbe
further improvedtobeapplicablefor small scalefire
inlargecross-sectionroadtunnel. Perhapssomeother
factorshadtobeconsideredduringthecalculationof
vaulttemperature, involvingdifferentkindsof ventila-
tionconditionsandgeometrical propertiesof theroad
tunnel.
4 CONCLUSION
1. Asforlorryandcarfiresinlargecross-sectionroad
tunnel andunder longitudinal ventilationvelocity
of 3m/s, therelativepositionof vehicles hadless
influenceontotal HHR whichwas approximately
23MWandlessthanthealgebrasumof 25MW.
2. Owing to large cross-section of the road tunnel,
small scalevehiclefireandlongitudinal ventilation
velocity of 3m/s, thesimulatedvault temperature
wasnot veryhigh.
3. In connection with the calculation of the vault
temperature, there was limitation to the related
empirical equation. The relative error which was
between the simulation analysis results and the
empirical equation value exceeded 20% for the
HRR less than 17MW and greater than 68MW.
Obviouslyitshouldbefurther improvedespecially
for small-scalefireof theroadtunnel.
ACKNOWLEDGMENTS
REFERENCES
Gu, Z.P. & Cheng, Y.P. 2003. Researchontemperaturefield
distribution in transportation tunnel fires. Paper collec-
tions of first session international meeting on city and
industrial safety in China (Nan jing): 265268 (in Chi-
nese).
Haack, A. 1998. Fire protection in traffic tunnels: general
aspects and results of theEUREKA project. Tunnelling
and Underground Space Technology 4(13): 377.
Kurioka, H., Oka,Y., Satoh, H. &Sugawa, S. 2003. Fireprop-
ertiesinnear fieldof squarefiresourcewithlongitudinal
ventilationintunnels. Fire Safety 38(4): 319340.
Yasuda, F., Ono, K. & Otsuka, T. 2004. Fireprotection for
TBM shield tunnel lining. Tunnelling and Underground
728
Effectsof different benchlengthonthedeformationof surrounding
rockbyFEM
X.M. Wang, H.W. Huang& X.Y. Xie
Department of Geotechnical Engineering, Tongji University, Shanghai, P.R.China
ABSTRACT: Bench cut method has been extensively used in mountain tunneling. This is mainly due to
its flexibility to adaptingto different groundconditions. Induceddisplacements areempirically controlledby
adjustingthespeedof excavation, thebenchlength, partial-faceexcavationandclosureof invert. Inthispaper,
a series of three-dimensional, numerical, elastoplastic analyses were conducted to investigate the effects of
different bench length on the deformation of surrounding mass in soft rock. The closure of invert was also
investigatedtotheir roleincontrollingthefinal displacement. Whenbenchcutmethodadoptedinsoftrock, the
benchlengthshouldnot betoo longor too short and0.5times of tunnel diameter aroundfor benchlengthis
appropriate.
1 INTRODUCTION
Inmountaintunneling, thebenchcut methodis used
extensively due to its simplicity and flexibility to
adapting to different ground conditions. And the
methodprovidesanadvantageof simultaneousexca-
vationof theupper andlower sections. Thekey issue
of adopting this method is selecting a length and a
shapeof benchtoassurethestabilityof theface, espe-
ciallyfor tunnelinginsoft rock. Inaddition, auxiliary
methodsareusedasrequired.Whenthegroundisgood
enough, havingenoughself-supportingproperties, the
benchlengthcanbereducedto0m(full facecut). In
thecaseof aweakground, thebenchlengthisempiri-
callydecided. Itisthereforeimportanttoevaluateand
compare the effect of different bench length on the
deformationof surroundingrock andonthestability
of excavationfront.
Usually,numerical analysesworkasakindof model
test in which many relevant design variables can be
investigated in parametric studies (Ng & Lee, 2005;
Karkus&Fowell, 2003; Galli etal., 2004). Inthisway
it is possible to quantify the relative importance of
eachpossibleinterventioninorder tochoosethemost
effectivemeasuresfromtheeconomicandsafetypoint
of view.
Seki et al. (1994) conducted a series of three-
dimensional finite-elementelasticanalysesof unlined
tunnelingtodeterminetheeffect of benchlengthand
shape. Theinitial stress was givenby external force,
ignoring thedead weight effect. They found that the
longer the bench, the smaller the displacement due
to squeezing at the face. Moreover, they found that
thebenchlengthscarcely exertedinfluenceuponset-
tlement of thecrown. Finally they concluded longer
benchtendedtooffer agreatsafetyfactor andleaving
thecorewaseffectivefor increasingthefacestability.
Farias et al. (2004) conducted a series of three-
dimensional elasticfiniteelement analysesof tunnel-
ingby theNewAustrianTunnelingMethod(NATM)
to investigate relative importance of relevant tech-
niquesforsettlementcontrol.Thetechniquesincluded
partial-face excavation, free span distance and sup-
port activation. Thetunnel has9.6mof diameter and
the soil cover is 10m. They found that tunnel sup-
port lining including freespan and closureof invert
wasthemost relevant singlefactor analyzedinreduc-
inginducedsettlements. Thecloser tofacethelining
was concreted, thesmaller thedisplacements. More-
over, they foundthebenchhelpedto keephorizontal
pressureintheexcavationface.
Eberhardt (2001) conducted a series of compre-
hensive, three-dimensional, elasticandelasto-plastic,
numerical analyses of tunneling in thecentral Swiss
Alps with different assumed initial stress states to
demonstratethree-dimensionstress rotationaheadof
anadvancingtunnel face. Thediameter of tunnel was
10m and the bench length was 10m. They found
that high stress concentrations in association with
largerotation of themaximumprinciplestress were
729
Figure1. Threedimensional finiteelement mesh.
observed when the initial maximumprinciple stress
alignment was horizontal and parallel to the tunnel
axis.
Inthispaper, aseriesof three-dimensional, elasto-
plastic, numerical analyseswerecarriedout toinves-
tigate the effects of different bench lengths on the
deformation of surrounding mass in soft rock. The
objectivesof theseanalysesaretofindoptimumbench
lengthsfor different rockmass.
2 THREE-DIMENSIONAL NUMERICAL
MODELLING
2.1 Numerical approximations
Ahypothetical tunnel excavationinsoftrockwasmod-
eled in this three-dimensional numerical study. The
diameter of thetunnel (D) was takenas 12m, witha
constantcover depth30m.Thetunnel wasassumedto
beabenchexcavationandlinedwithspray shotcrete
and bolt. The finite element program, MARC, was
adoptedtomodel thetunnel excavation.
Thethree-dimensional finiteelement meshusedin
thepresentanalysesisshowninFigure1.Thebedrock
wasset 37.2mbelowthebottomof thetunnel andthe
domain expands laterally 80mfromthe tunnel cen-
terline. Themodel tookadvantageof symmetryof the
problem. Boundaryconditionsaretotallyfixedat the
bottomof themodel andonly vertical displacements
arefreeinthevertical sides. Themodel consistedof
10114elements and10373nodes. Eight-nodedbrick
elementsandfour-nodedshell elementswereusedto
model therock andconcretelining, respectively. And
two-noded truss elements were used to model bolt.
Themeshwas dividedinto 22longitudinal blocks of
Table 1. Rock parameters used in the finite element
analyses.
Rock E/GPa /KN*m
3
C/MPa ,(
)
IV
upper
6 0.3 23 0.7 39
IV
lower
2 0.35 20 0.2 27
V
lower
1 0.35 17 0.05 20
Figure2. Schematicrepresentationof L
1
andL
2
.
variablesizes. Thefirst 20blockslengthcorresponds
1/4of thetunnel diameter (D). Thelast2blocklength
is1/2D.A monitoredsection, locatedinthemiddleof
themesh(i.e., atz=30m), wasstudiedduringevery
stageof excavation and construction. To account for
therelativelylargestressandstraingradientsnear the
tunnel opening, small finiteelementswereused.
An elastic perfectly plastic rock model, using
Drucker Prager failurecriterion with anonassoci-
atedflowrule, was adoptedinthis study. Thetunnel
lining and bolt were modeled as linear elastic. The
YoungsmodulusandPoissonsratiofor thetunnel lin-
ingweretakentobe25GPaand0.2, respectively. The
unitweightof thetunnel liningwas22kN/m
3
.Forbolt,
theYoungsmodulusof 210GPawithPoissonsratioof
0.3, wereadopted.Table1providestherockparameter
usedinthis study. Thesevalues werebasedonCode
for designof RoadTunnel (2004).
2.2 Numerical modeling procedures
A given cross-section was divided into three parts:
upper section, lower section and invert section. For
simplicity, upper benchlengthequals to lower bench
length(L
b
).Thefreedistancebetweenexcavationface
andthesupport headingwill bereferredasfreespan
(L
1
). The distance between the excavation face and
thefirst wholelining section will bereferred as full
supportdistance(L
2
).Theschematicrepresentationof
thesewasshowninFigure2.
730
0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30
4
2
0
2
4
6
H
e
i
g
h
t

f
r
o
m

t
h
e

s
p
r
i
n
g
l
i
n
e

(
m
)
Squeezing (mm)
0.25D
0.50D
0.75D
1.0D
1.5D
2.0D
top heading
full face
Figure 3. Distribution of squeezing at the face (IV
upper
rock).
Tunnel excavationandconstructionweresimulated
bydeactivatingtherockelementswithintheproposed
tunnel excavationzoneandby activatingthesupport.
The tunnel excavation rate was modeled 3.0m(i.e.,
D/4) per day, which was used as a step size in the
numerical analyses. No support was applied to the
tunnel face. Theunsupportedlengthequals 3m(i.e.,
L
1
=3m). Theexcavationsequencesare:
1 Excavateupper sectionrock until theupper bench
lengthequal L
b
, andinstall tunnel liningtothepre-
viouslyexcavationspansimultaneously. Leavefree
spanof 3m.
2 Excavateupper sectionrockandlower sectionrock
until the lower bench length equal L
b
, and install
tunnel lining to the previously excavation span
simultaneously. Leavefreespanof 3m.
3 Excavateuppersectionrock, lowersectionrockand
invert sectionrock, andapplylining.
4 Advance the tunnel by repeating step 3 until the
upper tunnel facehas passed3.0D fromthemoni-
toringsection.
Theinitial stress was given by gravity becauseof
shallow tunnel. The different bench lengths (0-2D),
topheadingcut andfull facecut werestudiedinthe
analyses.
3 ANALYTICAL RESULTS
3.1 Squeezing at the tunnel face
When a large displacement is created because of
squeezingat thetunnel face, topplingor collapseby
slippingof thetunnel faceisverylikelytotakeplace.
Figures35show, for eachbenchlength, thedistribu-
tionof squeezing(intheZdirection,i.e.,tunnel driving
direction) alongthetunnel centerlineat themonitor-
ingsection. Figure6shows therelationshipbetween
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
4
2
0
2
4
6
H
e
i
g
h
t

f
r
o
m

t
h
e

s
p
r
i
n
g
l
i
n
e

(
m
)
Squeezing (mm)
0.25D
0.50D
0.75D
1.0D
1.5D
2.0D
top heading
full face
Figure 4. Distribution of squeezing at the face (IV
lower
rock).
0 1 2 3 4 5 6 7 8 9 10 11
4
2
0
2
4
6
H
e
i
g
h
t

f
r
o
m

t
h
e

s
p
r
i
n
g
l
i
n
e

(
m
)
Squeezing (mm)
0.25D
0.50D
0.75D
1.0D
1.5D
2.0D
top heading
full face
Figure5. Distributionof squeezingattheface(V
lower
rock).
0.0 0.5 1.0 1.5 2.0
1.0
1.5
2.0
2.5
3.0
3.5
R
a
t
i
o

o
f

m
a
x
i
m
u
m

s
q
u
e
e
z
i
n
g

t
o

t
h
a
t

o
f

t
o
p

h
e
a
d
i
n
g

m
e
t
h
o
d
Bench length (D)
IV lower
V lower
Figure6. Relationshipbetweenbenchlengthandmaximum
squeezingdisplacement.
benchlength(IV
lower
andV
lower
rock) andmaximum
squeezingdisplacement.
For different rock, as the bench length increases,
thesqueezingdistributiontendstothedistributionof
731
Figure7. Crownsettlement versusfacedistancefor differ-
ent benchlength(IV
upper
rock).
topheadingcut. Anincreaseinbenchlengthgreatly
decreases the maximum squeezing. The longer the
bench, thesmaller thesqueezing: forV
lower
typerock,
comparedwiththeresultsfor thetopheadingmethod,
thesqueezing of thefull facemethod is 3.43 times,
with0.25Dbenchlength1.49times, with0.50Dbench
length1.13times, andwith0.75D benchlength1.03
times. For another two types of rock, as the bench
lengthsincrease, thechangingtrendsof squeezingare
the same as the trend of V
lower
type rock. However,
withthedeteriorationof rock, thestablevalueof bench
length(whenthebenchlonger thanstablevalue, there
isnosignificantchangeof squeezing) isdifferent. For
theIV
upper
, IV
lower
andV
lower
rock, thestablevaluesof
benchlengthare0.50D,0.50Dand0.75D,respectively.
It is very interesting to note that for IV
upper
rock
thecomputedsqueezingdisplacementsat thetopand
bottomof upper section arenegativevalues. That is
duetoloadtransfer inlongitudinal directionresulted
from arcing in the unsupported zone. That has an
effect similar to the classical Terzaghis trap door
experiment.
3.2 Crown settlement
When alargedisplacement of thecrown at theface
is produced, the tunnel face or its vicinity is prone
to failure. Figures 79 show the crown settlements
at themonitoring section versus thenormalized dis-
tance of the excavation face to monitoring section,
for different bench lengths. A negativevalueof dis-
tanceindicatesthattheexcavationfacehasnotreached
themonitoringsectionyet.A significantpercentageof
thefinal stabilizedsettlement is inducedbeforeface
passage. This can only be adequately reproduced in
three-dimensional analyses.
The crown settlement before the face passage
doesnot significantly dependuponthebenchlength.
However, comparedwithfull facemethod, thecrown
ent benchlength(IV
lower
rock).
ent benchlength(V
lower
rock).
settlement of bench method before the face pas-
sagehas aremarkabledecrease. As thebenchlength
increases, thestabilizedsettlementbecomeslargerand
closertothevalueof settlementof topheadingcut. For
V
lower
rock, the stabilized settlement is 5.85mmfor
0.25Dbenchlength, 6.41mmfor 0.50Dbenchlength,
and 6.83mmfor 0.75D. When the bench length is
greater than 1.0D, the stabilized settlement value is
larger than7.0mm, buttheexactvaluecouldnotbe
obtainedinthisstudyfor thelimitationof themodel.
Thereasonforsettlementdecreasingmainlyattributes
to full activation with invert closure. As the bench
lengthdecreases, thevalueof L
2
becomessmallertoo.
Fromthispoint of view, tokeeptunnel facestability,
thebenchlengthshouldnot betoolongfor soft rock.
On theother hand, to prevent tunnel squeezing, it is
necessarytomakebenchlengthlongenough. Conse-
quently, thebenchlengthtoo longor too short is not
helpful to stability of excavation face, and thereis a
reasonablevalueof benchlengthfor somerock. For
732
good rock (i.e., IV
upper
rock), becausethestabilized
settlementissmall, itisnotnecessarytoreducebench
lengthtodecreasethesettlementandthebenchlength
canbesettoonelargervaluewhichbenefitsthesimul-
taneous excavation of the upper and lower sections.
For badrock (i.e., IV
lower
andV
lower
rock), thebench
lengthshouldnot betoolong, and0.5timesof tunnel
diameter aroundfor benchlengthisappropriate.
Asmentionedintheintroduction, Seki etal. (1994)
conductedthree-dimensional finite-element analyses
to investigate the effect of bench length and shape
onthetunnel facestability. Intheir studies, theycon-
cludedthatlongerbenchesaremadeinpoorergrounds.
However, in this study, the contrary result can be
obtained. ThisisbecauseSekismodelingwaselastic
andthetunnel wasunlined. Thestabilizedsettlement
isthesamefor different benchlength.
4 CONCLUSIONS
Thebenchcut methodby whichtheupper andlower
sectionsareexcavatedat thesametimeisusedexten-
sively inmountaintunneling. Results werepresented
fromadetailedthree-dimensional finiteelementanal-
yses directed towards the effects of different bench
lengths on the deformation of surrounding mass in
soft rock.
Thelonger thebench, thesmaller thedisplacement
duetosqueezingattheface. Whenthebenchlengthis
longerthanstablevalue, thereisnosignificantchange
of squeezing, compared with the squeezing of top
heading method. Thestablevalues of bench lengths
aredifferent for different rock.
Comparedwithfull facemethod, thecrownsettle-
ment of bench method before the face passage has
a remarkable decrease, and it does not significantly
dependuponthebenchlength.
Thereisareasonablevalueof benchlengthforsome
rock. Thebenchlengthshouldnot betoolongor too
shortand0.5timesof tunnel diameteraroundforbench
lengthisappropriatefor soft rock.
ACKNOWLEDGMENTS
demicDisciplineProject, Project Number: B308.
REFERENCES
Eberhardt, E. 2001. Numerical modelingof three-dimension
stress rotationaheadof anadvancingtunnel face. Inter-
national Journal of Rock Mechanics & Mining Sciences
38(4): 499518.
Farias, M.M., J unior, A.H., Assis, A.P. 2004. Displacement
control intunnelsexcavatedbytheNATM: 3-Dnumerical
simulations. Tunneling and Underground Space Technol-
ogy 19(3): 283293.
Galli, G., Grimaldi, A., Leonardi, A. 2004. Three-
dimensional modeling of tunnel excavation and lining.
Computers and Geotechnics 31(3): 171183.
Karakus, M. & Fowell, R.J. 2003. Effects of different tun-
nel faceadvanceexcavation on thesettlement by FEM.
Tunneling and Underground Space Technology 18(5):
513523.
Ministry of Communications of the Peoples Republic of
China. 2004. Code for design of Road Tunnel. Beijing:
ChinaComunicationsPress.
Ng, C.W.W. & Lee, G.T.K. 2005. Three-dimensional ground
settlements andstress-transfer mechanisms duetoopen-
face tunneling. Canadian Geotechnical J ournal 42(4):
10151029.
Seki, J., Noda, K., Washizawa, E. etal. 1994. Effectof bench
lengthonthestabilityof tunnel face. InAbdel Salam(ed.),
Tunnelling and Ground Conditions: 531542. Rotterdam:
Balkema.
733
Theeffectsof loadedboredpilesonexistingtunnels
J.Yao, R.N. Taylor &A.M. McNamara
City University, London, UK
ABSTRACT: Thispaper presentsthedevelopment of aseriesof centrifugetestscarriedout toinvestigatethe
effects of loadingof boredpiles onexistingtunnels. Theapparatus was designedto monitor thetunnel lining
deformationwhilealiveloadingwasapplied. Four factswereconsidered: rateof loadingthepile, pilebaselevel
relativetotheexistingtunnel, theratioof thedepthof clay cover totunnel diameter, andthedistancebetween
thepileandtunnel.
1 INTRODUCTION
High-rise buildings with deep pile foundations are
more and more used in the fast developing urban
environment. Inevitabledisturbancetothegroundand
surrounding underground structures caused by their
constructionandsubsequentloadingmayhavesignif-
icant impact intermsof settlement anddeformation.
Over the last thirty to forty years, tunnel owners
have become concerned about this possibility, and
anexclusionzonewas introducedto protect thetun-
nels. However, these guidelines, are mainly based
ontheir empirical correlations fromsimilar projects,
and generally apply limits on theminimumdistance
betweentheexistingtunnelsandnewpilefoundations.
(Chudleighet al. 1999).
Manyresearchershavesummarisedthepile-tunnel
interactionproblems(Schroeder,2002;Yaoetal 2006),
whichcanbecategorisedinto two groups: effects of
tunnelling on piles and effects of piling on tunnels,
wherethesecondgroupreceivedlessattention. Some
fieldcasestudieshavebeenpresented, whichinvesti-
gatedtheeffectof pilingonexistingtunnels(Chapman
etal, 2001, Higginsetal, 1999, andBenton&Phillips,
1991). Schroeder (2002a, 2002b) alsoconductedaset
of FE analyses to investigatetheinteractionbetween
pilefoundationsandexistingtunnels.Yaoetal (2006)
describedaseriesof centrifugetestsdesignedtoinves-
tigatetheeffect of bored pileexcavation on existing
tunnels.
Theeffectof boredpilefoundationsonexistingtun-
nelscanbecategorisedintotwoparts: pileinstallation
and the post piling period. In this project, the main
objectiveis focusedonthepost pilingperiod, which
is the investigation on the effect of pile loading on
existingtunnels.
Thepaper presentsthecentrifugeapparatusdesign,
thecentrifugetestresultsandsomepreliminaryanaly-
sisof theresults.All thetestspresentedaresolelywith
regardstopostpiling; theinfluenceof pileexcavation
isnot considered.
2 CENTRIFUGE MODELLING
Theapparatuswasdesignedtomeetfiverequirements:
thepilecouldbeloadedatanytime, arangeof loading
rates, theappliedloadcouldberecorded, pilesettle-
ment couldberecordedandtherateof loadingcould
bechangedduringthetest.
2.1 Model overview
Figure 1 shows a picture of the model on the cen-
trifugeswing.Thecentrifugestrongboxisconstructed
fromaluminiumwith atransparent Perspex window
in the front to enable a view of the experiment in
progress. A tunnel lining with an internal deflection
gaugesysteminstalledinthecentreof thesample, two
rubbagswereusedtoseal thetunnel at thetwoends
of thetunnel lining(Yaoetal, 2006) andapileloading
apparatuswasfastenedontopof themodel container.
Druck PDCR81porepressuretransducers wereused
to monitor theporewater pressurein themodel and
airpressureinthetunnel liningsystemduringthecen-
trifugetesting. Digital imageprocessingwas usedto
tracethedeformationatthefrontsurfaceof themodel.
All the tests were conducted in the Geotechni-
cal Engineering Research Centre at City University,
London. TheAcutronic 661 centrifugewith amaxi-
mumpayloadof 200kgat200gandradiusof 1.8mis
described in detail by Schofield andTaylor, (1988).
All tests presented in this paper were conducted at
735
Figure1. Model oncentrifugeswingprier totest.
an acceleration level of 100g, according to scaling
law(Taylor, 1995), 1cminthemodel equalsto1mat
prototypescale. Kaolin clay was used for modelling
(Al Tabba, 1987).
2.2 Pile loading unit
A pileloading systemwas designed to load thepre-
installed pile during the centrifuge test, and also
providethefacilitiestomeasurethesettlement of the
pileandappliedload. Figure2shows aschematic of
thepileloadingunit, whichconsist amodel pile(pre-
installedinthemodel), anactuator, twoLVDTs, aload
cell, andsupport units.
For ease of centrifuge operation, the pile was
pre-installed into the model during model prepara-
tion. Having a pile pre-installed has two important
requirements:
Nosettlementwouldoccur duetotheself weightof
pile
The pile must be strong enough to withstand the
appliedload
A 22mmouter diameter, 10mminternal diameter
aluminiumtubewithtwoendssealedwasselectedas
themodel pile. Two different lengths weremadefor
different pilebaselevelsandcover totunnel diameter
(C/D) ratios.Themodel pilewasmadetohavealower
density than the clay sample, so it would not cause
Figure 2. Schematic of the loading unit. C: depth of the
cover abovetunnel, D: diameter of thetunnel.
anysettlementsduringcentrifugetestingwithout any
appliedload. AscanbeseenfromFigure2, aloading
pinsittingontopof thepilecap, whichhadtwoLVDTs
(Linearly variabledifferential transformers) installed
oneitherof ittomonitorthesettlementof thepile.The
entireloadingunit will stopthepilebeingpushedout
duetothelighter density.
The actuator contains a 35mm 90 watt motor,
RE35-118783, a planetary Gearhead GP 42 C gear
box, and an aluminiumscrew jack, MSZ-Alu. The
actuator fitted directly abovethepilewas supported
by a channel unit sitting on top of the strongbox as
canbeseenfromFigures1and2.Theloadingpinwas
attachedtoit, toapplytheloadtothepile. A loadcell
wasusedtomonitor theloadappliedtothepile. Itwas
mountedbetweentheloadingpinandthescrewjack.
2.3 Deflection gauge tunnel system
Figure3showsthecrosssectionof themodel inplan,
which has similar layout to thepileexcavation tests
carriedoutbyYaoetal (2006).Thetunnel diameterwas
50mmwiththethicknessof 0.15mm, andwasmade
of carbonfibre. A 16mmdiameter titaniumtubewas
usedtosupportthetunnel unitandprovidethepathway
for all the wires and cables. The tunnel was sealed
by two air-pressurised rubber bags at the two ends.
Tunnel liningdeformationdetector wasattachedtoit.
736
Figure3. Crosssectionof themodel,showingthedeflection
gaugesandpileborehole(s). D
p
isappliedpile-tunnel spacing
duringtest (22or 44mm).
Figure4. Assembleddeflectiongaugessystem.
Figure4showsthedeflectiondetector unit developed
byYaoet al (2006). That wasmadeof four deflection
gauges, which were calibrated strain gauged carbon
fibre cantilevers, and their output was recorded via
anonboardPC. Inorder tomeasurebothinwardand
outwarddeformation,thedeflectiongaugesweregiven
a4mmpre-deformation.
2.4 Other equipment
Strongbox: Thecentrifugemodel wassetupwithinan
aluminiumalloy strongbox, whichhadaninner plan
area of 550mm200mm, and can contain a solid
model up to 300mmhigh, a 80mmwidth Perspex
windowat thefront toenablethemodel tobeviewed
duringthetest.
Water wassuppliedfromastand-pipesittingonthe
centrifuge swing, and the pore pressure in the clay
model wasmonitoredusingporepressuretransducers
pre-installedintothesamplebeforethetest. Detailsof
thisequipment havebeenwell discussedbyanumber
of researchersatCityUniversity,London.(Tayloretal,
1998, McNamara, 2001).
A pilecutter set wasusedtoexcavatetheborehole
for the model pile. The cutter was made of 22mm
stainlesssteel thinwalledtube.
2.5 Sample preparation and test procedure
Kaolinslurrywithawatercontentof 120%waspoured
into the strongbox, which had a 3mmthick porous
plastic sheet with a 0.5mmthick filter paper in the
bottom. Thestrongboxwastransferredintoaconsoli-
dationpress, andloadeduptoavertical effectivestress
of 500kPa. Uponcompletionof normal consolidation
it was swelledback to250kPa. Thebulk unit weight
of thekaolinwasabout 17.44kN/m
3
. A de-airedand
calibratedporepressuretransducer wasinstalledinto
thesampleafter theswellingperiod.
Atypical sampleset-upandtestprocedureconsisted
of thefollowingsteps:
Freewateratthetopof themodel wasremovedafter
closingthedrainagetaps closedat thebaseof the
strongbox; thiswastoavoidclayswellingback.
The applied vertical stress was reduced to zero
andthestrongbox removedfromtheconsolidation
press.
The front wall was removed, so the front surface
of theKaolinsamplecouldbecleanedto ensurea
better imageprocess, andthetopof thesamplewas
trimmedtotherequiredheight.
Tunnel was cut andpre-assembledtunnel unit was
installedintothemodel.
Marker beads for image processing were pushed
intothesamplefront surfaceona10mmgrid.
The Perspex window was then bolted onto the
strongbox.
Thepileshaft holewasexcavatedandpilewaspre-
installedintothemodel.
The pile loading unit was mounted on top of the
strongbox; loadingpinwasdrivendowntothepile
cap.
The strongbox was weighed and placed on cen-
trifugeswing.
The model was accelerated to 100g on the cen-
trifuge, andleftovernighttoreachtheporepressure
equilibrium.
Air pressurein therubber bags at theends of the
tunnel wasreduced
Loadappliedtothepileusingtheactuator.
Thetunnel liningdeformations, porepressureinthe
model, loadapplied, pilesettlementsandair pressure
intherubber bags, weremonitoredanddatastoredon
thecomputer inthecontrol room. Global movement
aroundthetunnel wasmeasuredusingtheimagepro-
cessingsystem. However, theresultsof digital image
analysiswill not bediscussedinthispaper.
3 TEST RESULTSANDDISCUSSION
Eighttestswerecarriedout, of which2weretrial tests
usedtocommissionandimprovetheapparatus.Table1
summarisesall thetests:
Pilewasinstalledat twodifferent tunnel-pileclear
spacing: 22mmand44mm.
Twodifferent C/Dratioswereused: 2and3.
Pilebasewasinstalledattwodifferentlevelsrelative
tothetunnel position: tunnel crownlevel andinvert
level.
Onlyasinglepilewastested.
737
Table1. Tableof tests.
Test ID C/D PileLength Pilebase Spacing
J Y13 2 100mm Crown 2.2
J Y14 2 100mm Crown 2.2
J Y15 2 150mm Invert 4.4
J Y16 2 100mm Crown 2.2
J Y17 2 150mm Invert 2.2
J Y18 2 100mm Crown 4.4
J Y19 3 150mm Crown 2.2
J Y20 3 150mm Crown 4.4
*Spacingis theminimumdistancebetween theoutsides of
thetunnel andpile(minprototypescale).
All thedatapresentedinthepaper areintendedto
demonstratethedesignof thisresearchandthebasic
dataobtainedfromthecentrifugetest.
3.1 Rate of loading the pile
Themainpurposeof twoinitial trial testswastotestify
theloadingunitandtodeterminethebestrateatwhich
toloadthepile.
Burland et al (1966) produced a simple method
to predict the load/settlement curve. It is assumed
that the curve is linear up to full mobilisation with
takesplaceat asettlement of about 0.5%of theshaft
diameter. Brownetal (2002) presentedaseriesof Stat-
namicteststoinvestigatetheinfluenceof loadingrate
on pile behaviour in clay. The pile bearing capacity
was found to be particularly sensitive to pile defor-
mationrate. Dayal & Allen(1975) foundthesimilar
response.Frischmann&Fleming(1962)statedthatthe
recordedsettlementwaslargelyelastic. All settlement
was assumedto beas adirect result of shear strains.
Skempton(1951) presentedthesimilar result:
where:
s
=shaft settlement, p
s
=loadappliedtopile
shaft-soil interface, L =effectivelengthof pileshaft,
G
ave
=mean shear modulus of soil along pile shaft,
r
m
=radius frompileat whichstrainbecomes negli-
gible(Randolph&Wroth, 1978) andr
o
=pileradius.
Whitaker & Cooke(1966) statedthat whenthesettle-
mentisabout0.5percentof theshaftdiameter, thepile
shaftfrictional resistancedevelopsrapidlyandlinearly
withthesettlement.
The speed of pile displacement also affects the
pore pressure at the base of the pile, which can be
understoodasonincreaseinmagnitudeof excesspore
pressurewithincreasingpenetrationrate(Brownetal,
0
20
40
60
80
100
120
140
160
20 25 30 35 40 45 50
Depth (mm)
S
h
e
a
r

s
t
r
e
n
g
t
h
JY14 JY15 JY17 JY19
Figure5. Undrainedshear strengthmeasuredagainstdepth
inmodel.
2002). Therateof loadingusedinthecentrifugetests
were designed to create undrained conditions. The
loading speed of 2mmper minute was selected on
thebasisof thetrial tests.
Theultimateloadof apilecanbedefinedaseither
the load at which the pile settlement continues to
increasewithout further increaseof resistance, or the
load at which pile settlement reaches 10% of the
pile base diameter (Fleming et al, 1992). For most
soil conditions, thesecondcategory ismorelikely to
be the controlling factor for end bearing resistance
(Burland et al, 1966). In this research, the ultimate
loadisdefinedastheloadwhichcausesasettlement
of 10% of pile foundation base diameter. The ulti-
mateloadcapacity of pilescanbeestimatedinterms
of undrainedstrength(S
u
), inthis researchmeasured
fromquick undrainedtestsdirect takenafter thetest,
and undrained pileshaft adhesion factor, which was
chosenas 0.6inthis paper. Figure5shows themea-
sured undrained shear strength again depth directly
takenafter four tests.
Figure6showstherecordedloadappliedinthetest
J Y16. Half of themaximumcalculatedloadreachedas
soonastheloadapplied,thenincreasedmoregradually
towards themaximum. Theload only achieved 85%
of thedesignedultimateload. Applyingthefactor of
safetyof 1.5to2, theachievedloadwasacceptablefor
our researchpurposes.
3.2 Tunnel lining deformations
Tables2and3summarisethemaximumdeformations
measured by thedeflection gauges for tests J Y13 to
J Y20, andpositivevaluesindicatemovement towards
thetunnel centre. Duringthepileexcavationtest car-
ried out by Yao et al 2006, the tunnel crown was
subjected to the greatest deformations for all cases,
and moved towards the tunnel centre, but there was
lesseffectattunnel invert. Asitcanbeseen, whilethe
pilewas under load, thecrown was still affected the
738
0
50
100
150
200
250
300
350
400
450
0 10 20 30 40 50 60 70
Time (s)
A
p
p
l
i
e
d

L
o
a
d

(
/
)
Figure6. Therecordedincreasingof loadagainst time.
Table2. Summaryof maximumtunnel liningdeformations
recordedduringpileloadingperiod(mmatprototypescale).
ID Crown Left Right Invert
J Y15 11 1 2 8
J Y16 18 1 2 12
J Y17 22 2 3 19
J Y18 10 1 1 9
J Y19 24 3 4 18
J Y20 16 2 3 14
Table3. Summary of maximumdeformations in percent-
age: deformationover tunnel diameter (,D).
ID Crown Left Right Invert
J Y15 0.22 0.02 0.04 0.16
J Y16 0.36 0.02 0.04 0.24
J Y17 0.44 0.04 0.06 0.38
J Y18 0.2 0.02 0.02 0.18
J Y19 0.48 0.06 0.08 0.36
J Y20 0.32 0.04 0.06 0.28
most, followedinturnbytheinvert, theright sideand
theleft sideof thetunnel. (Theright sideof thelining
isthenearestsitetothepile). Bothrightandleftsides
werealwaysmovedawayfromthepile.
Figure7showstherecordedtunnel liningdeforma-
tionfor test J Y16. It canbeseenthat thetunnel lining
didnot start tomoveuntil theloadreachedhalf of its
maximum. Together withFigure6, it shows therela-
tionshipbetweentheappliedload, pilesettlement and
tunnel liningdeformation. Similar resultswerefound
for most of thetests.
Figure 8 shows the relationship between applied
loadandpilesettlements.Thepiledidnotstarttomove
until theloadachievedhalf of itsmaximum.
Table4liststhechangeof diameterof thetunnel lin-
inginthevertical andhorizontal directions. Figure9
showstherelationshipbetweenthetunnel liningdefor-
mation and pile-tunnel clear spacing and Figure 10
Figure7. Thetunnel liningdeformationsagainsttheapplied
load. Inall cases, positivevaluesindicatemovementtowards
tunnel centre, deformationsat model scale, for test J Y16.
Figure 8. Applied loads against pile settlements, for test
J Y16.
Table 4. Change of tunnel diameter and tunnel centre
position(inprototype).
ID v v,D h h,D Cv Ch
J Y15 3 0.06 1 0.02 9.5 1.5
J Y16 6 0.12 1 0.02 15 1.5
J Y17 3 0.06 1 0.02 20.5 2.5
J Y18 1 0.02 0 0 9.5 1
J Y19 6 0.12 1 0.02 21 3.5
J Y20 2 0.04 1 0.02 15 2.5
v
and
h
: Changeof tunnel diameter at vertical/ horizontal
direction(mm); C
v
andC
h
: Changeof tunnel centreposition
at vertical/ horizontal direction(mm), C
v
movetowardspile
openingandC
h
movedownward.
presentstherelationshipbetweentheliningdeforma-
tionandpilebaseposition. Itcanbeseenfromchanges
of thediameter, that deformationof thetunnel lining
isnon-uniform.Thechangesof thetunnel centreposi-
tion in the vertical and horizontal direction are also
summarisedinTable4. Figure11showsthechangein
positionsof thetunnel centreforeachcase.Bothcrown
andinvertweresubjectedtosignificantimpactfortest
J Y17andJ Y19wherelongpilewithlarger C/Dratio
739
Figure 9. The change of movement due to change of the
pile-tunnel clear spacing(C/D=2).
Figure10. Thechangeof movement duetochangeof pile
baseposition(C/D=2).
was used. Movements up to 0.48% of tunnel diam-
eter wererecorded on thecrown when thelong pile
wasinstalledtogether withdeeper cover. For thesame
C/D ratio andtunnel-pileclear spacing, intest J Y16
andJ Y17amovement of 0.08%of thetunnel diame-
ter wasrecordedfor thelonger pile. Summarizingthe
tables and the figures, it can be see that in general,
asclear spaceincreased, thedeformationreduced; the
deeperthepilebaseslevel orthelongerthepilelength,
themoretheeffect of thepileat thetunnel.
4 CONCLUSIONS
This paper presents a study of the influence of pile
loading on an existing tunnel, which is a part of a
research on the effect of bored pile installation and
subsequent loadingonanexistingtunnel. Thedesign
of themodel and preliminary analysis of theresults
werepresented.
Based on the literature review and the centrifuge
model tests, thefollowingobservationsweremade.
By reviewing previous research and based on two
trial tests, the rate of the loading was chosen as
Figure 11. The positions of tunnel centre for all cases.
Movement onX axisistowardspileopening, movement on
Y axisistowardsthebottomof thestrongbox.
2mmper minute.Thetrial testsalsoconfirmedthat
increasing in therateof loading will increasethe
loadachievedonthepile.
Pile settlement has a linear relationship with
increasingof theappliedloadoncetheloadexceeds
half of themaximumdesignedultimateload.
Tunnel lining response to the pile movement as
observed from the results shows that the tunnel
centrealwaysmovedownwardsandaway fromthe
pile.
Increasingthepile-tunnel clear spacingwill reduce
thedeformationof thetunnel lining.
Usingthelonglengthpilewill havemoreeffect on
thetunnel liningregardlessof theC/Dratio.
Tunnel crownisalwayssubjecttosignificantmove-
ment duetoeither pileexcavation(Yaoet al, 2006)
or pile loading. Tunnel invert affected more from
pileloadingthanpileexcavation.
REFERENCES
Al-Tabbaa, A. 1987, Permeability andstress-strainresponse
of Speswhitekaolin,PhDthesis,Universityof Cambridge.
Benton, L. J. & Phillips, A., 1991, The Behaviour of two
tunnelsbeneathabuildingonpiledfoundationsProc. 10th
European Conference on Soil Mechanics and Foundation
Engineering, Florence, 665668.
Chapman, T., Nicholson, D. & Luby, D. 2001, Use of the
observational methodfor theconstructionof pilesnextto
tunnels. Proc. Int. Conf. Response of Buildings to Exca-
vation Induced Ground Movements, (ed F. M. Jardine)
London:CIRIA.
Chudleigh, I., Higgins, K. G., St J ohn, H. D., Potts, D. M. &
Schroeder, F. C. 1999. Piletunnel interaction problems.
Proc. Tunnel Construction & Piling 99, London. The
HemmingGroupLtd, 172185.
Higgins, K. G., Chudleigh, I., St J ohn, H. D. & Potts
D. M. 1999, An example of pile tunnel interac-
tion problems. Proc. Int. Symp. Geotech. Aspects of
Underground Construction in Soft Ground, IS-Tokyo
740
99(eds O.Kusakabe, K. Fuita &Y. Miyazake) Rotterdam:
Balkema, 99103.
Schroeder, F. C. 2002a,Theinfluenceof boredpilesonexist-
ingtunnels: acasestudy. Ground Engineering 35, No7,
3234.
Schroeder, F. C. 2002b,Theinfluenceof boredpilesonexist-
ing tunnels. PhD thesis, Imperial college, University of
London.
Schofield, A. N. &Taylor, R. N. 1988, Development of stan-
dard geotechnical centrifuge operations. Centrifuge 88
(ed Corte), Balkema, 2932.
Taylor, R. N. 1995, Geotechnical Centrifuge Technology,
BlackieAcademicandProfessional, Glasgow.
Taylor, R. N. Grant R. J., Robson, S. & Kuwano, J.
1998, An image analysis systemfor determining plane
and 3-D displacements in soil models. Centrifuge
98, (eds, Kimura, Kusakabe & Takemura), Balkema,
7378.
Yao, J,Taylor, R. N. &McNamara, M.A. 2006,Theeffectsof
boredpileinstallationonexistingtunnels.The proceeding
of 6th International Conference on Physical Modelling in
Geotechnics, HongKong.
741
3DFEM analysisongrounddisplacement inducedbycurved
pipe-jackingconstruction
G.M.You
Shanghai Municipal Engineering Design General Institute, Shanghai, P.R. China
ABSTRACT: Thepower tunnel of Tibet roadinShanghai isthemost difficult pipe-jackingproject inChina
withlongdistanceandlargediameter. Thetunnel crossesthroughSuzhouriver, undergroundpipelinesandarea
of densebuildings, especially theundergroundpassageof subway andtheexistingsubway line2, so its vital
important to protect thesurrounding environment when thepipe-jacking is constructed. Thedeformation of
groundsurfaceandtheexistingtunnelsduringtheconstructionof curvedpipe-jackingisstudiedwith3Dfinite
elementmethods. First, thecalculatedresultsareanalyzedandcomparedwiththemeasureddatainsitetoverify
thecorrectnessof the3DFEMmodel. Second, therearemanyfactorswhichcanaffectgrounddisplacementwhen
curvedpipe-jackingis constructed. Amongthesefactors, slurry sleeves qualities, soil pressureonthefaceof
pipe-jacking, slurryinjectionpressureandearthresistancearediscussedonthebasisof finiteelementsimulation.
Then, thedeformationof existingtunnelsof subwayline2isstudied. Finally, thegrounddisplacement formula
of curvedpipe-jackingisdiscussed.Theresultsshowthatthecontinuityof slurrysleevesandthepressureof the
facearevery important factorstogroundsurfacedeformationduringtheconstructionof curvedpipe-jacking.
Thegrounddisplacementsinducedbycurvedpipe-jackingarenot larger thanthoseof linear pipe-jackingif the
slurrysleevesaroundpipearegoodandcontinual. Becauseof additional earthresistance, themaximumof the
groundsurfacedeformationperpendicular topipeaxisisat thesideof thecenter of thepipe-jackingcurve, and
thedistanceof deviationdependsontheradiusof pipe-jackingcurve.
1 INTRODUCTION
Curvedpipe-jackingtechnologywasappliedinJ apan,
EuropeandAmericamany years ago, andsomesuc-
cessful experienceswerealsoacquired. Nomuraet al.
(1985) developed a pipe jacking method (D301) to
facilitatelong-span, curvedandhigh-speedcapabili-
ties in the construction of small diameter (300mm)
tunnels.Nanno(1996)proposedanewcurvingmethod
calledtheunit curvingmethod inwhichfour joint-
adjustersareinstalledbetweenpipes. Thejoint angle
is controlled by the adjusters and the thrust is also
distributed uniformly in thefour adjusters. Thenew
method solved most of the technical problems in
curveddrives andperformedmany jobs successfully
in actual construction sites. Vogler & Georg (2002)
studied the stresses on the curved pipe. The predic-
tionequationfor acurvedjackingareawas analyzed
inorder toexplainthecharacteristicsof thrustandthe
frictionresistance(Shimadaet al. 2004). Thethrusts
inslurry pipe-jackingcanbepredictedaccurately by
using theresistancebetween themud slurry and the
concretepipesandtheresistancebetweenthesoil and
thepipesinthecurvedjackingarea.
InChina, pipe-jackingtechnology hasbeendevel-
oped rapidly in recent years and many successful
constructionswereobtained, suchasthesewerageout-
fall project inShenzhen(Mao, 2001), thesecondary
project of improving combined sewerage systemin
Shanghai (Ge, 2002), theprojectof Hefangstreetwith
sewagepipeandrainwaterpipeinHangzhou(J inetal.
2002), theproject of main trunk sewer pipeof Yan-
gli wastewater treatment plant inFuzhou(Liu, 2003).
However, only singlecurve(horizontal curveor ver-
tical curve) was used in these projects. Ding et al.
(2001) calculatedandanalyzedthejackingforce, joint
stretching value, pipeinternal force, stability of soil
andearthresistanceof curvedpipe-jackingbymeans
of pipe-joint mechanical model and beamon elastic
foundationmethod.
Fang & Wang (1998) analyzed the ground settle-
ment dueto pipejackinginsoft soils anddeveloped
amethodto predict thesettlement profileof straight
pipe-jacking. Wei et al. (2003) analyzed the mecha-
nismand reason of ground deformation caused by
pipejackingconstruction.Wei etal. (2005)derivedthe
computing formulas of ground deformation induced
bybulkheadadditivethrust, forceof frictionbetween
shieldandsoil, andforceof frictionbetweenfollow-up
pipesandsoil byusingtheMindlinsolutioninelastic
mechanics. Furthermore, theformulaof total ground
deformationinducedbypipejackingconstructionwas
743
obtainedby combiningtheformulaof grounddefor-
mationinducedbygroundlosswiththepreviousone.
However, thereis nostudy ontheformulaof ground
deformationof curvedpipe-jacking.
Thepower tunnel of Tibet roadinShanghai is the
most difficult pipe-jackingproject inChinawhichis
3.033kilometerslong. Thispaper discussesthedefor-
mation of ground surface and the existing tunnels
by means of thenumerical analysis. Someimportant
parameters during the construction and the ground
displacement formulaof curvedpipe-jackingarealso
discussed.
2 3DNUMERICAL ANALYSISOF CURVED
PIPE-JACKING
2.1 Mechanical model
Theradiusof pipe-jackingplanecurveis600m. The
distance along the pipe axis range form0 to 70m.
Thedistanceperpendiculartopipeaxisis66mandthe
height of themodel is40m. Thetunnel islocatedat a
depthof 7.5m(fromthetopof pipetogroundsurface).
Theouter- andinner-diameter of theconcretepipeare
3,200mmand2,700mm, respectively. Inthisanalysis,
weusethefollowingassumptions:
1. Thepressureon theexcavation faceis uniformly
distributedwithcircular shape. Thevalueisequal
to the actual measurement pressure of soil and
watercabin(facepressure).Accordingtotheactual
observationrecord, thefacepressureis0.18MPa.
2. During the construction of pipe-jacking, time
dependent behavior of soil isnot considered.
3. Thefrictional resistancebetweensoil andpipe-ring
is uniformly distributed along pipe axis. Accord-
ingto theactual observationrecord, thefrictional
resistanceisabout 2.0kPa(Youet al. 2006).
4. To distinguish with straight pipe-jacking, earth
resistance should be accounted which induced
by curved pipe-jacking. Therelationship between
earthresistanceandradius of pipe-jackingcurva-
tureisapproximatelyinversely-proportional linear
(Dinget al. 2001). Theearthresistanceis12.0kPa
at aradiusof 600m.
2.2 Boundary conditions
Displacement boundaryconditionsareappliedtothis
model.Thetopsideof themodel isfreeboundary.Ver-
tical displacementsof thebottomsideandnormal dis-
placementsof thevertical sidesarefixed, respectively.
2.3 Compute parameters
Accordingtothegeological report, thematerials and
their parameters that areusedfor this simulationare
listedintable1.
Table 1. Physical and mechanical parameters of the
materials.
Unit Youngs Friction
weight modulus Poissons Cohesion angle
Materials (kN/m
3
) (MPa) ratio (kPa) (deg)
Brownclay 18.6 26.7 0.35 21.6 14.5
Siltyclay 17.3 14.2 0.37 14.4 13.0
Muddyclay 16.9 11.8 0.39 14.0 10.5
Grayclay 17.3 16.1 0.36 15.7 11.5
Siltyclay 17.8 27.2 0.35 18.5 17.5
Figure1. 3D-FEM meshmodel.
Thebreakingcriterionusedinthemodel isDrucker-
Prager criterion.
2.4 Initial stress state
Initial stress state is obtained by the FEM software
directly. Only self-weight stress of soil is considered
without tectonicstress.
2.5 Finite element mesh
The mesh, consisting of 11120 nodes and of 10696
elements, issubdividedinto6regions, havingdifferent
material properties. Eightnode, solidelementandfour
node, shell elementareusedtosimulatesoil andpipe-
jackingring, respectively. Inordertosimulatethesup-
port effect of slurry sleeves, contact surfaceelement
is engagedbetweenpipe-jackingringandouter soil.
Figure1showsthe3DFEM model usedinthisstudy.
2.6 Analysis of numerical results
Figures2and3showthecomparisonbetweenground
surfacedeformationobtainedfromnumerical analysis
and formmeasured data. The predictions fromthe
FEM compare reasonably well with the observed
results.
744
Figure2. Displacementsof groundsurfacealongpipeaxis.
Figure3. Displacements of ground surfaceperpendicular
topipeaxis.
Becauseof thecomplexfactorsaffectingthedefor-
mationof groundsurface, it isimpossibletoconsider
every factor during the simulation of construction.
Only several main factors areinvolved in this study.
These factors are slurry sleeves, the pressure of the
faceandtheearthresistance.
2.6.1 Contribution of slurry sleeves
The effect of slurry sleeves to reduce resistance
betweenpipeandsoil relatesnot only tothematerial
and mixture ratio of slurry but also to the injec-
tion parameters such as the location of injection
holes, injectionpressureandslurryinjectionquantity.
The location of injection holes and slurry injection
quantity can be simulated by different locations of
slurry sleeves aroundpipe. In this study, 5cases are
considered, asshowninFigure4.
Figures5and6showthesurfacedisplacementswith
different locations of slurry sleeves. Figure 5 indi-
cates that the continuity of slurry sleeves has vital
importanteffectonthesurfacedisplacements.Thesur-
face displacements will be reduced sharply if good
slurry sleeves can be formed around the pipe. This
Figure4. Different locationsof slurrysleevesaroundpipe.
topipeaxiswithdifferent locationsof slurrysleeves.
Figure6. Displacementsof groundsurfacealongpipeaxis
withdifferent locationsof slurrysleeves.
phenomenonismoreobviousatthetopcenter of pipe.
Therangeof deformationiswider if theslurrysleeves
are less continuous. Figure 6 indicates that the less
continuoustheslurry sleeves, thelager istheground
surfaceupliftmovementaheadthetunnel faceandthe
lessarethegroundsurfacesettlementsabovethetun-
nel face. The less continuous the slurry sleeves, the
lessarethegroundsurfacesettlementsbehindthetun-
nel face(intherange0 15m). Thesettlementsare
invariantafter15mbehindthetunnel faceif theslurry
sleevesarecontinuous. Moreover, thesettlementswill
developfast if theslurrysleevesarenot continuous.
2.6.2 Contribution of face pressure
Figures7and8showthesurfacedisplacementswith
different facepressures. Figure7 indicates that face
745
topipeaxiswithdifferent facepressures.
withdifferent facepressures.
pressurehasimportant effect onthesurfacedisplace-
ments at the top center of pipe. Figure 8 indicates
that the larger the face pressure, the more obvious
is theuplift movement of surfaceground. Thepoint
withmaximumupliftmovementiscloser tothetunnel
facewithincreaseof facepressure. Thedeformations
ahead and behind the tunnel face will be increased
with increaseof facepressure. Thesettlement at the
topof thetunnel faceis invariant withdifferent face
pressure.
2.6.3 Contribution of earth resistance
Figures 9 and 10 show the surface displacements
withdifferentearthresistances.Thealterationof earth
resistance does not change much the displacements
perpendicular to pipe axis. The displacements are
almost thesamewithdifferent earthresistance. From
this result, it means that if goodandcontinual slurry
sleevescanbeobtained, thedisplacementsof ground
surface of curved pipe-jacking are almost the same
asstraight one. Moreover, becauseof additional earth
resistance, the deformation profile perpendicular to
pipeaxisisnot symmetric. Thelarger theearthresis-
tance,themoreobviousisthedifference.Furthermore,
topipeaxiswithdifferent earthresistances.
withdifferent earthresistance.
thepointwithmaximumgroundsurfacedisplacement
perpendicular to pipe axis is not on the projection
line of pipe axis but at the side of the center of the
pipe-jackingcurve.
3 NUMERICAL ANALYSISOF SUBWAY
LINE 2SDEFORMATION
The power tunnel passed over the existing subway
line2, andtheminimumcleardistancebetweenpower
tunnel andexistingsubway line2is 1.5m. Thetotal
projectedlengthof crossover zoneisabout25mlong.
The acute angle between power tunnel and subway
line2isabout 75degree.
Figure12shows therelationshipbetweenvertical
displacement of subway line2 and jacking distance.
Figure13 shows therelationship between horizontal
displacement of subway line2 and jacking distance.
In Figures 12 and 13 origin of jacking distance is
40mawayfromthecrosspoint. Forthetotal projected
lengthof crossoverzoneisabout25mlong,thejacking
746
Figure11. Meshof power tunnel andsubwayline2.
Figure12. Relationship between vertical displacement of
subwayline2andjackingdistance.
distancerangedfrom40to65mwhenthepowertunnel
passedthroughsubwayline2.
Figure 11 shows the 3D finite element mesh of
the power tunnel and subway line 2. The region of
the model is: 100m(along pipe axis)60m(per-
pendicular to pipeaxis)40m(depth). Eight node,
solid element and four node, shell element areused
to simulate soil and pipe-jacking ring, respectively.
Displacement boundaryconditionsareappliedtothis
model. The top side of the model is free boundary.
Vertical displacementsof thebottomsideandnormal
displacements of thevertical sides arefixed, respec-
tively.Theparametersthatareusedfor thissimulation
arelistedintable1.
InFigure12positivedisplacement indicatesheav-
ingmovement.Itcanbeseenthatvertical displacement
of subway line2 increased sharply after thejacking
facepassedthroughthesubway. Therewasalmost no
displacement of subway line2beforethepipebegan
totraversethesubway. Becausethepipelinestraversed
above the subway line 2, only uplift displacements
Figure 13. Relationship between horizontal displacement
of subwayline2andjackingdistance.
Figure 14. Regression curve of ground surface displace-
mentswith10kPaearthresistance.
occurredduringtheconstruction. Figure13indicates
that thedirectionof horizontal displacement pointed
tothetunnel face.
TheFEM results show that themaximumof ver-
tical and horizontal displacements were3.3mmand
0.39mm, respectively.Thevaluesmeetthedemandof
thenormal operationof existingsubwayline2.
4 GROUNDDISPLACEMENT EQUATIONOF
CURVEDPIPE-JACKING
Becauseof earth resistance, thedeformation profile
perpendicular to pipeaxis of curved pipe-jacking is
notsymmetrictothenormal lineof pipeaxis.Accord-
ing to numerical results as well as measured data,
the ground displacement formula of curved pipe-
jacking is obtained by means of regression analysis
(Fig. 14). Thevertical displacementsperpendicular to
pipeaxisare:
747
Figure 15. Relationship between curvature radius r and
regressivecoefficient x
c
.
Where x
c
is the regression coefficient related to
theradiusof pipe-jacking, S
max
andi arethemaximum
settlement of groundsurface(whenx =x
c
) andstan-
darddeviationof thesettlement curve, respectively.
Based on the research consequence (Ding et al.
2001), we can get the curvature radiuses corre-
spondence to the earth resistances mentioned above
(10kPa, 15kPa and 20kPa). Then, according to the
previousfittingresult, therelationshipbetweencurva-
tureradiusr andregressivecoefficient x
c
isobtained
(Fig. 15). The fitted linear regression line can be
expressedasfollows:
It should be noted that the equations (1) and (2)
are obtained with the special geological conditions
of this project. The equations suitability for other
areas and geological conditions need more practical
verifications.
5 CONCLUSIONS
1 Because of additional earth resistance, the max-
imumvalue of the ground surface displacement
perpendiculartopipeaxisisatthesideof thecenter
of thepipe-jackingcurve, andthedistanceof devi-
ationdependsontheradiusof pipe-jackingcurve.
Thelarger theradius of pipe-jackingplanecurve,
thelessisthedeviation.
2 Thedeformationprofileperpendicular topipeaxis
of curvedpipe-jackingisnot symmetric about the
normal lineof pipeaxis.
3 Slurrysleeveshavevital effectongrounddisplace-
ments. If goodandcontinual slurrysleevescanbe
obtained, the displacements of ground surface of
curvedpipe-jackingarealmostthesameasstraight
one.Thekeytocontrol thedisplacementsof ground
surfaceistocontrol thequalityof slurrysleeves.
4 Thefacepressureisanimportant factor toground
surfacedisplacements. Duringtheconstructionof
curvedpipe-jacking, thepressureof soil andwater
cabinshouldbecontrolledstrictly.
5 Only uplift displacements of subway line 2
occurred during theconstruction of pipe-jacking.
Thedirection of horizontal displacements of sub-
wayline2pointedtothetunnel face.
REFERENCES
Ding, W.Q. & Zhu, H.H. et al. 2001. Beamonelastic foun-
dationmethodconsideringpipe-jointandanalysisof pipe
packingconstruction. Journal of Tongji University 29(5):
616620.(inChinese)
Fang, C.Q. &Wang, C.D. 1998.Ananalysisandpredictionof
groundsettlementduetopipejacking. Journal of Jiangsu
University of Science and Technology 19(4): 106110.(in
Chinese)
Ge, J.K. &Zhang,Y. 2002.Applicationof jackingtechnology
to sharply curvedpile. Chinese Journal of Geotechnical
Engineering 24(2): 247250.(inChinese)
J in, W.H. & Gong, F.X. et al. 2002. Applicationof longdis-
tance curve pipe-jacking to Hangzhou project. Special
Structures 19(4): 6265.(inChinese)
Liu, P.R. 2003. The Application of long-distance curved
pipejackingtechnology. Fujian Construction Science and
Technology (2): 3839.(inChinese)
Mao, B.Q. 2001. Technology of curved pipe-jacking with
long-distance and large-diameter in sea area. Special
Structures 18(3): 4853.(inChinese)
Nanno, T. 1996. A method for driving curved pipe-jacked
tunnels. Tunnelling and Underground Space Technology
11(2): 325.
Nomura,Y. &Hoshina, H. etal. 1985a. Pipejackingmethod
for long curve construction. Journal of Construction
Engineering and Management 111(2): 138148.
Nomura, Y. & Hoshina, H. et al. 1985b. Design and char-
acteristicsof D301pipejackingmachine. Denki Tsushin
Kenkyusho Kenkyu Jitsuyoka Hokoku 34(12): 17891799.
Shimada, H. &Khazaei, S. etal. 2004. Small diametertunnel
excavationmethodusingslurrypipe-jacking. Geotechni-
cal and Geological Engineering 22(2): 161186.
Vogler & Georg. 2002. Stresses on jacking pipes when
driving curved alignments. Betonwerk und Fertigteil
Technik/Concrete Precasting Plant andTechnology 68(7):
5061.
Wei, G. & Huang, Z.Y. et al. 2005. Study on calculation
methods of ground deformation induced by pipe jack-
ingconstruction. Chinese Journal of Rock Mechanics and
Engineering 24(Suppl. 2): 58085815.(inChinese)
Wei, G. & Xu, R.Q. et al. 2003. Analysis of ground defor-
mation caused by pipe jacking construction. Trenchless
Technology 20(6): 2427.(inChinese)
You, G.M. & Ge, J.K. et al. 2006. Study on environmental
protection and control techniques of 3-D curved
pipe-jacking construction. Rock and Soil Mechanics
27(Suppl.): 398401.(inChinese)
748
Theme 6: Calculation and design methods,
and predictive tools
Calculationof thethreedimensional seismicstressedstateof
MetroStationEscalatorOpenLineTunnels system, which
islocatedininclinedstratifiedsoft ground
R.B. Baimakhan, N.T. Danaev, A.R. Baimakhan, G.I. Salgaraeva, G.P. Rysbaeva,
Zh.K. Kulmaganbetova, S. Avdarsolkyzy&A.A. Makhanova
Scientific Center of Fundamental Research, Almaty, Kazakstan
S. Dashdorj
Hokkaido University, Sapporo, Japan
ABSTRACT: This includes examples of a destructive effect of earthquakes on the underground structures
andthenewmechanicmathematical model of theinclined-stratifiedmassif, whichdiffersfromthelikemodel
in such away that enables to research stressed stateof an underground structurefor arbitrary orientation of
theextendedaxis. Besides, thisincludestheresultsof calculatingathree-dimensional seismicstressedstateof
theundergroundsystem, whichincludesametrostationescalator openlinetunnelsinthesoft groundthat has
inclinedandstratifiedstructure.
1 INTRODUCTION
A lot of cases of damage and destruction of under-
ground structures during a strong earthquake are
known. They include, for example, undergroundpip-
ing, mountaintunnels, stationaryandmetroopenline
tunnelsof bothshallowanddeepshaft.Therearemany
examplesof considerabledamageandstrongdestruc-
tionsinundergroundstructures, frompipingtomines,
all aroundtheworld. Undergroundstructuresseemto
bestable. But this is not quitetrue. For example, as
shownontheFigure1, inDaikai, J apan, duringearth-
quakes in Kobein 1995 (Magnitude 7.5) not only
the roofs and sides of the metro station but also its
columns werebeingdestroyedlaidat aepthof 50m
fromthesurface(Iidaetal.1996,Ishihara1998).There
occurred mass destruction in underground mines at
a depth of 326m near the town of Shurab during
theIsfara-Batkent earthquakein 1977 (MiddleAsia,
magnitude 6.5). (Rashidovet al. 1975).
Based on the analysis of damage of Based
on the analysis of damage of 71 underground
structures caused by earthquakes, Ch. Daudin,
D.V. Monakhenko, S.G. Shulmanandothers propose
toassessthembyfivecategories: 1shiftsalongcon-
tacts; 2general distortion; 3local cracks; 4rock
fracturing and failure and 5partial failure (Mon-
akhenko& Shulman1980).
A general analysis shows that during intensive
earthquakesundergroundstructuresmaybedestroyed
Figure 1. Characteristic features of destructions of the
metrostationduringearthquakesinKobe, J apanin1995.
bothintheshallowanddeepshaftandalsointheearth
stratumandrock mass. Factors, whicheffect onsuch
destructionsarediverse. Thestrengthof structureand
fasteningelementsdependsnotonlyonmanufactured
materials, but alsoonphysical & mechanical proper-
tiesof thesurroundingmassif. However, onemaynote
acharacteristicfeaturethat undergroundstructuresof
deep shaft are destroyed owing to dynamic stresses
surpassingthebreakingpoint of fasteningmaterials.
751
The ground stratumsurrounding an underground
structureoftenischaracterizedby natural anisotropy.
Theexistingmethods allowus to consider static and
seismic stressedstatemainly inaflat state. It is con-
nected,firstly,withrestrictedpossibilitiesof analytical
methods, and, secondly, with an undeveloped model
of undergroundstructuresinathree-dimensional state
subjecttothearbitraryorientationof theextendedaxis
inspaceinarbitrarydirections.
Thirdly, tofullyprovideseismicstabilityof acom-
plexsystemof undergroundstructures, it isnecessary
to consider their interference (for instance, that of
station, escalator andmetro pairedopenlinetunnels
duringseismicvibrations)
2 MECHANICALMATHEMATIC MODELSOF
THE INCLINEDSTRATIFIED
TRANSTROPIC MASSIF
The Figure 2 includes different directions of wave
propagationinrelationtotheelementsof theinclined
layered massif subject to the modeling by means of
atransversal isotropicbodywithaninclinedplaneof
isotropy. ThevariantI correspondstothewavepropa-
gationalongthelineof layerspreading(intersectionof
theinclinedplaneof isotropyandthehorizontal plane
at theangle), thevariant IIacross thelineof layer
spreadingattheangle.Theywerepreviouslyconsid-
eredbyJ.S.Yerzhanov,Sh.M.Aitaliev,J.K.Massanovin
theabove-mentionedwork.
Thereby underground openings are on the hori-
zontal plane, though they are oriented in a different
way related to the line of layer spreading (plane of
isotropy): drifts and crosscuts, if they are oriented
towards the line of layer spreading across and per-
pendicular, respectively; diagonal openingsif theyare
orientedatanangle.ThevariantIII correspondstothe
common case, wherethedirection of seismic waves
constitutes anarbitrary angle withthelineof layer
Figure2. Different directionsof seismic waivesspreading
ininclined-stratifiedtranstropicmassif relativelythelengthy
axisof undergroundstructure.
spreading.Thelongitudinal axisof anextendedunder-
groundopeningmay havethesamedirection, i.e. the
openingmay beinclinedto thehorizon. This variant
was first analyzed by R.B. Baimakhan (Baimakhan
2002).
Healsohasobtainedthemost general expressions
of velocities of quasi-longitudinal and two quasi-
transversal wavesthat spreadinarbitrary directionin
the inclined layered transtropic massif with normal
n = n{cos , cos, cjs}.
Forsuchmassif, theequationof thegeneralizedlaw
inamatrixformshouldbewrittenasfollows
Thefollowingvaluesof elasticityfactorsc
ij
for ahor-
izontal stratifiedstateareestablished. (Erzhanovetal.
1980).
where E
K
,
K
, (k =1, 2) J ungs modules and Pois-
sonsratios. G
2
shear modulus. (Lekhnitskiy1965)
whereq
p
im
q
p
jn
, (p =, , , i, j = 1, 2, , 6) matrixes
of cosinesof turningangles.Asseen, saidcalculations
areinterconnected. Basedontheexpression(5) inthe
workof R.B. Baimakhan, expressionsarereceivedfor
speeds of elastic wavepropagationinthetranstropic
752
massif withanormal of n = n{cos, cos, cos} in
anarbitrarydirectionasfollows(Baimakhan2002)
where =arccos (0.5q(p,3)
3,2
), environ-
mental density, , , anglesbetweenthenormal of
thewavefrontandtheaxesof theCartesiancoordinates
OXYZ.
3 DESTRUCTIVE EARTHQUAKE
ACCELEROGRAM CONSTRUCTION
SUBJ ECT TOTHE GROUNDANISOTROPY
Themethodsof receptionof elasticequivalentcharac-
teristicsfor alternateisotropicstrataarealsoavailable
in the work of J.S. Erzhanov, Sh.M. Aitaliev and
J. K. Masanov(Erzhanovet al. 1980).
For transtropic(transversal-isotropic) strata, elastic
constantsshouldbecalculatedbyusingthefollowing
formulas:
whereh
k
thicknessof thekstratum; n thenumber
of strata; E
k
,
k
, G
k
elastic characteristics of thek
isotropicstratum; 2(1+
k
).
With provision for data of J SC Almaty
metrokurylys andreferencedataongrounds (Buly-
chev 1989) someaveragevalues of elastic character-
isticsaresystematisedandadducedat intheTable1.
Calculatedbyformulas(7) valuesof elasticconstants
andfor differentlocationsof layersareadducedatthe
Table2.
Thesedataby valueof deformation modulus per-
pendicular tothelayer E
1
conditionally ispossibleto
refer to the three ground conditions of a city. Their
minimumvaluescorrespondtoalluvial groundsatthe
debrisconesof riversBigandSmall Almatynkanorth-
wardfromtheRaimbekavenue.Averagemeanings to
thegroundsalongtheAbayavenuebetweentworivers
AlmatinkaandVesnovka.TheMostvaluescorrespond
to grounds of flood-plain areas of rivers. Values of
elasticpropertiesmaybespecifiedbywayof detailed
determinationof lithologic thicknessesof groundsin
regionsof all stationsandroutesof drivingtunnels.
If threecomponentsof theaccelerogramonthesur-
faceareknown, thentoreceivetheir modifiedvalues
withadepthalongthevertical linedown, according
totheworkof R.B. Baimakhan, wehavethefollowing
formulas(Baimakhan2002)
Table1. Elasticpropertiesof groundsE
k
,
k
andthickness
h
k
of their layersat theundergroundrote.
E
k
h
k
Layersgroundpaper (m) (m)
1 Soft pouredsoil 7.0 0.40 2.8
2 Loamysoil 30.0 0.36 2.2
3 Gravel pebbles 25.0 0.28 3.3
4 Loamwithpebbles 8.5 0.21 2.7
5 Boulder groundwithgravel 120.0 0.27 3.2
6 Boulder groundwithpebbles 80.0 0.35 4.8
7 TickLoam 50.0 0.25 6.5
8 Sandof averagesize 22.0 0.36 4.5
9 Boulder withpebbles 200.0 0.32 5.1
10 TickClay 300.0 0.31 5.9
Table 2. Calculated equivalenttranstropic properties of
soft stratifiedgroundE
1
, E
2
, G
2
,
1
,
2
.
E
1
E
2
G
2

1

2
Layer (Mpa) (Mpa) (Mpa)
1 71.19 14.47 5.42 0.30 0.06
2 73.61 15.80 5.95 0.30 0.07
3 73.45 15.61 5.88 0.30 0.06
4 76.51 16.07 6.04 0.30 0.06
5 80.88 16.40 6.14 0.31 0.06
6 83.16 15.87 5.93 0.31 0.06
7 83.02 14.94 5.56 0.31 0.06
8 83.85 14.58 5.42 0.31 0.05
9 83.73 14.23 5.29 0.31 0.05
10 81.55 13.73 5.10 0.31 0.05
753
Figure 3. Synthetic accelerogram of a 10 Intensite
earthquake.
whereq=0.009;
Havingdividedtheknownvalues of theaccelero-
gramon the surface of the ground by modification
factors, well receive their values a depth along the
vertical linedown.
Using the formulas (2) (11) by the methods
mentioned in the work of R.B. Baimakhan, design
accelerograms of intensive9 10gradeearthquakes
areconstructedfor aparticular buildingsiteasrelated
bothtothesurfaceandtheinner strataof aheteroge-
neousmassif (Figure3, Baimakhan2003).
4 CALCULATIONOF A SEISMIC STRESSED
STATE OF THE UNDERGROUND
STRUCTURE SPACE SYSTEM
Byamethodof finiteelements, aseismicstressedstate
of acomplexsystemof themetrotunnelsisresearched.
TheFigure4shows adesignarea. For clearness, the
metrostationisshowninadoubleenlargedform.
Paired open line tunnels come to the metro sta-
tion and go away fromthe opposite side. Boundary
conditions of the task are as follows: in the verti-
cal planesABDC andEFHG thereareno horizontal
displacements.
Figure4. Designareaof acomplexsystemof underground
structures.
=0; inthevertical planesBFHDandAEGCthere
arenohorizontal displacementsu =0; inthehorizon-
tal planeCGHD thereareno vertical displacements
w=0. The upper horizontal planeAEFB is free of
stresses.Theroofingof themetrostationisatadepthof
17mfromthesurface. Geometricdimensions: height,
widthandlengthof saiddesign 125, 45and160m
respectively; thoseof thestation 9, 22 and 100m;
thoseof openlinetunnels theheightandwidth 2m;
diameter of the escalator tunnel 9, length 54m;
angleof inclinationfor theescalator against thehori-
zontal axis0X for thelyingside30
. Length, width
andheightof thegallery, whichconnectsthelowerend
of theescalator tothestationare13, 8and5m.
Said area is divided by eight-nods isoparametric
prismaticelementswitheightadditional internal inte-
grationpointsfor 3508spaceelementswithtotal nods
of 4687. Physical & mathematical properties of the
massif areasfollows: E
1
=1.028 10
4
Mna,
1
=0.31,
2
=0.10, E
2
=0.292 10
4
Mna, G
2
=0.11 10
4
Mna
=2.2m,m
3
. Liningmaterialsfor thestationandthe
openlineescalator tunnelsare:
E
obdeku
=2, 5 10
4
Mna,
obdeku
=0, 25,
obdeku
=2, 5m/m
3
(timberlining)
Valuesof theangles, andvarywidely.Accord-
ing to the metro project, the angle of inclination
cannot exceed4
for openlinetunnelsand30
for
escalator andother special tunnels. Theangles and
canchangefrom0to90
.
TheFigure5showsseismicstressesdiagramsatthe
tunnel contoursif =0, =0, =4
andtheangleof
wavefall =60
, =30
, =0.
5 CONCLUSION
Under thestaticload, theareasof theroofingandthe
lyingsideof thestationareexposedtothemainstress
concentration. Seismic stress diagrams, though they
754
Figure5. Seismic stressesdiagramsat thetunnel contours
in different times of a non-stationary seismic load of 10
gradeearthquake: astationsectionatadistanceof 22mfrom
the front end; bstation section at a level of the escalator;
csectionof openlinetunnelsat adistanceof 35mfromthe
station; dcentral sectionof thegallerybetweentheescalator
andthestation. Thecurvescorrespond: 1tothestatiload-
ingof themass of themassif upper strata; 2and3 to the
seismicloadsat thetimesof 18.244sec. and35.132sec.
arecomplex,however,showsomeregularities.Inaddi-
tiontotheangular areas, sideareasareexposedtothe
most intensive seismic load. Such case is typical of
station, escalator andopenlinetunnelsandgalleries.
Themiddlepart of theescalator tunnel is in amore
seismicstressedstate.
The results of said research make it possible to
strengthenelements of anundergroundstructureand
toreducedestructiveeffectscasedbyactsof Godsuch
asearthquakes.
REFERENCES
Baimakhan, R.B. 2002.Analysis of seismic stability of under-
ground structures in a heterogeneous thickness by a
method of finite elements. Almaty: Daur.
Baimakhan, R.B. 2003. Development of the analysis of seis-
mic stressed state of underground structures subject to the
peculiarities of the geodynamics of the region. Thesisfor
competitionof anacademicdegreeof Doctor of technical
sciences.Almaty: InstituteMathematical andmechanical.
Erzhanov, J.S. et. al. 1980. Seismic stability of underground
structures in the stratified anisotropic massif. Almaty:
Nauka(Science).
Iida, H. etal. 1996. DamadgetoDaikai SubwayStation. Soil
and Foundation. Special Issue: 283300.
Ishihara, K. 1998. Performanceof tunnelsandunderground
structures during earthquakes. International Conference
on SoilStructure Interaction in Urban Civil Engineering.
89 October: 1931
Lekhnitskiy, S.G. 1965.Theory of elasticity of an anisotropic
body. Moscow: Gostechizdat.
Monakhenko, D.V. & Shulman, S.G. 1980. Issuesof seismic
stability of undergroundstructures. News from institutes
of higher education. Series Construction & Architecture
No. 8: 315.
Rashidov, T.R. et al. 1975. Seismic stability of metro tunnel
structures. Moscow: Transport.
755
A complexvariablesolutionfor tunneling-inducedground
movementsinclays
H.L. Bao, D.M. Zhang& H.W. Huang
ABSTRACT: A closed-formplanestrainsolutionispresentedfor anelastichalf-planewithacircular tunnel,
whereanoval- shapedgrounddeformationpatternis imposedas theboundary conditionof thedisplacement
aroundthetunnel opening. Thesolutiontranslatescomplex variablesfromtheresearchphysical regionontoa
circularringinthemirrorimageregionbyconformal mapping.ThecoefficientsintheLaurentseriesexpansions
of thestressfunctionscanbeobtainedfromtheboundaryconditionsandtheconvergenceof theseries. Finally,
acasestudyisalsoperformedbasedonthemetrolineNo. 2inShanghai andasatisfactoryagreement between
thepredictedsettlement andtheobservedoneisobtained.
1 INTRODUCTION
Due to recent city developments within the limited
land of urban areas, more and more complex facil-
ities aredeveloped under theground surface, which
maycauseseriouspotential damagetoadjacent over-
lying services and structures. Engineers responsible
for thedesignandconstructionof tunnelsshouldpre-
dict tunneling-inducedgroundmovementsinorder to
protect theexisting structures and tunnels. Therewe
three different approaches for estimating the poten-
tial tunneling-induced ground movements: empirical
methods, numerical methodsandanalytical methods.
Empirical procedures have been widely used to
assess potential ground movement owing to tunnel-
ing. In practice the ground deformations are often
describedbaseduponfieldobservations, for instance,
anormal GaussiandistributioncurveproposedbyPeck
(1969), whichhas no theoretical basis. It is assumed
that the surface settlement through can be approxi-
matedby thenormal probability curveor error func-
tion. However in reality, ground movements depend
onanumber of factors suchas tunnel geometry and
depth, tunnel constructionmethod, thequality of the
workmanship and management, behavior of the soil
around tunnel. Therefore the empirical methods are
alsosubject totheseimportant limitations.
Recentlysomeattemptshavebeenmadetodevelop
closed-formanalytical solutionsfortunneling-induced
groundmovementsinsoftground.VerruijtandBooker
(1996) presentedasimpleanalytical solutionforatun-
nel inahomogeneouselastichalf spacebythevirtual
imagetechnique. LoganathanandPoulos(1998) intro-
ducedanequivalentundrainedgroundlossparameter,
whichcanbeestimatedusingthegapparameter pro-
posedby Leeet al. (1992). Verruijt (1997) proposed
thecomplexvariablesolutionfor circular tunnel inan
elastichalf planewiththeboundaryconditionof apre-
scribeduniformradial displacementatthecavityopen-
ing.Bobet(2001)presentedanotherelasticsolutionfor
grounddeformationsof ashallowtunnel inasaturated
groundwhichwasmadewiththeboundarycondition
of uniformradial displacement at thetunnel opening.
However in practice the radial ground movement is
not uniformbut oval-shaped. Park (2004) proposed
the elastic solution for the tunneling-induce ground
deformation in clay by imposingtheprescribedfour
different typesof oval-shapeddisplacement at tunnel
opening.Wang(2007)expandedVerruijtssolutionsby
incorporatingfour different oval-shapeddeformation
pattern at tunnel opening proposed by Park (2004).
In this paper, the complex variable method by Ver-
ruijt (1997) is also usedfor thesolutionof elasticity
problems for a half-plane with a tunnel of the third
oval-shaped deformation pattern at tunnel opening.
The metro line No. 2 in Shanghai is used to check
theapplicabilityof theanalytical solution.
2 STATEMENT OF THE PROBLEM
The problemdeals with an elastic half-plane with a
circular tunnel, seeFigure1. Theradiusof thetunnel
is denoted by r, thedepth of its centreby h, and the
757
Figure1. Half-planewithacircular tunnel.
Figure 2. Oval-shaped deformation pattern of tunnel
boundary.
coverbyd.Theupperboundaryof thehalf-planeisfree
of stress. As inpracticetheradial groundmovement
aroundthetunnel isnotuniformbutoval-shaped. Park
(2004) proposed four different oval-shaped ground
deformationpatternastheinner boundaryconditions.
Wang(2007) suggestedthethirdboundaryconditions
as thegood agreement with thefield measurements
had been obtained. Therefore in this paper, loading
takes place along the inner boundary of the tunnel,
intheformof thethirdoval-shapeddeformationpat-
tern(Figure2), proposedby Park (2004), whichcan
bedenotedby,
3 BASIC EQUATIONS
The complex variable method for the solution of
two-dimensional linearlyelasticprobleminvolvestwo
analyticfunctionsof complexvariable, +(z) and+(z).
Figure3. Theregionafter conformal mapping.
Thehorizontal andvertical displacements, u
x
andu
y
,
intheelasticregionR canbeexpressedas
where j is the shear modulus of the clay, and is
relatedtoPoissonsratio by
As planestrain condition areassumed in this paper,
thestressescanbeexpressedas
It is convenient to express theboundary conditionin
termsof theintegral of thesurfacetraction, integrated
alongtheboundary,
whereC isanintegrationconstant.
4 CONFORMAL MAPPING
WeconformallymaptheregionR intheZ-planeonto
a ring region in the mirror image region, which
is referredas -plane, boundedby thecircles || =1
and|| =, seeFigure3. Theconformal mappingis
givenby
758
Where
If 0theradiusof thecircular cavingisprac-
tically zero, which indicates a very deep tunnel. If
1 the covering depth is very small. For every
value of d/h the corresponding value of can be
determinedfromequation(8).
By virtueof thesubstitutionz=() thefunction
+(z) and+(z) canbewrittenintermof ,
Because the conformal transformation function
() is analytic intheringregion, thefunction+(z)
and+(z) canberepresentedbyLaurent seriesexpan-
sions,
The coefficients of the series can be determined
fromtheboundaryconditions.
5 BOUNDARY CONDITIONS
Thefirst boundaryconditionisthat theupper bound-
aryy=0must beentirelyfreeof stress. Accordingto
equation(6), Verruijt (1997) gavetherelationshipof
thecoefficients,
Therefore, one half of the unknown coefficients
havebeenexpressedintotheother half.
In this paper, thesecondboundary valueproblem
isconsidered, inwhichthethirdoval-shapeddisplace-
ment pattern proposed by Park (2004) is prescribed
along the tunnel boundary. According to Verruijts
solution, if the function G
(), which defines the

boundary condition at the tunnel boundary, can be
writtenasaFourier series,
thecoefficient must satisfytheequations
and
Wang(2007) hadgiventheexpressionof thecoef-
ficientsof theFourier seriesfor thethirdoval-shaped
displacement pattern,
Thereforeall thecoefficientsof theLaurent series
havebeendetermined, exceptfora
0
.Thisconstantcan
bedeterminedfromtherequirementof convergenceof
theLaurent series by thelinear interpolationmethod
givenbyVerruijt (1997).
6 CASE STUDY
A case study is performed with the background of
Shanghai metroline2. Thetunnel liningfor Shanghai
metroline2is6.2minexternal diameter and5.5min
internal diameter. Theaveragedepthtothecenter-line
of thetunnel isabout 11m. Thetunnel wasexcavated
by EPB shield machine. The shield body is 6.24m
longwith6.34mindiameter. Thereforetheclearance
between the external diameters of shield body and
the tunnel, which is usually named physical gap G
p
759
-60
-50
-40
-30
-20
-10
0
0 10 5 15 20
distance from tunnel center(m)
g
r
o
u
n
d

s
e
t
t
l
e
m
e
n
t
(
m
m
)
Observed
Analytical
Figure4. Surfacesettlement.
according to the definition of Lee et al. 1992), will
reach140mm(Zhang,2004). TheYoungsmodulusof
theclayistakenas25MpaandPoissonsratioas0.3.
Figure4is aplot of comparisonbetweenpredictions
of surfacesettlement fromthecomplexvariablesolu-
tionandobservedmovements.Thecomparisonisquite
good,whichprovesthattheanalytical solutioncangive
reasonablepredictionsof groundmovementfor shield
driventunnelsinsoft clays.
7 CONCLUSIONS
Current design practiceto predict tunneling-induced
ground movements is based primarily on empirical
methods, which aresubject to someimportant limi-
tations. In practice, the geometry of the tunnel, soil
properties and construction methods may affect the
settlements of the ground. In this paper, complex
variablemethod is used for thesolution of elasticity
problemsfor ahalf-planewithatunnel of oval-shaped
deformation.ThemetrolineNo. 2inShanghai isused
to check theapplicability of theproposed analytical
solution. Itgivesgoodpredictionof tunneling-induced
groundmovements. However theplasticityof softsoil
is not consideredinthis paper, whichmeans that the
methodinthispaper canonlybeapreliminarydesign
of tunnelsinsoft clays. Morefactorsshouldbetaken
intoaccount toimprovethereliabilityof themethod.
ACKNOWLEDGMENTS
The authors wish to acknowledge the finan-
cial support from National Science Foundation of
China (No. 50608058) and Hi-tech Research and
Development Program (863 Program) of China
(No. 2006AA11Z118).
REFERENCES
Bobet, A. 2001. Analytical solutions for shallow tunnels
in saturated ground. J ournal of Engineering Mechanics
127(12): 12581266.
Lee, K.M., Rowe, R.K. & Lo, K.Y. 1992. Subsidence due
to tunnelling: Estimating the gap parameter. Canadian
Geotechnical J ournal 29(6): 929940.
for tunneling-inducedgroundmovements inclays. J our-
nal of Geotechnical and Geoenvironmental Engineering
124(9): 846856.
Park, K.H. 2004. Elastic solution for tunneling-induced
ground movements in clays. International J ournal of
Geomechanics4(4): 310318.
ground. 7thInt. Conf. onSoil Mech. andFound. Engrg.,
Mexico, 1969.
Verruijt, A. 1997. A complexvariablesolutionfor adeform-
ingcircular tunnel inanelastic half-plane. International
J ournal forNumerical andAnalytical MethodsinGeome-
chanics21: 7789.
Geotechnique46(4): 753756.
Wang, L.Z. &Lv, X.J. 2007. A complexvariablesolutionfor
differentkindsof oval deformationaroundcircular tunnel
inanelastic half plane. ChineseJ ournal of Geotechnical
Engineering29(3): 319327.
Zhang, D.M., Huang, H.W. & Hicher, P. 2004. Numerical
prediction of long-termsettlements of tunnels in clays.
ITA-AITES 2004WorldTunnel Congress and 30th ITA
General Assembly., Singapore, May2004.
760
Simulationof articulatedshieldbehavior at sharpcurveby
kinematicshieldmodel
J. Chen, A. Matsumoto& M. Sugimoto
Nagaoka University of Technology, Nagaoka, Niigata, Japan
ABSTRACT: Tunnelling cases at sharp curve have increased by using articulated shield with copy cutter.
At sharpcurve, theshieldbehavior, i.e., theposition, therotationangle, andtheadvancedirection, shouldbe
precisely controlledto followtheplanedalignment. However, it is sometimes difficult to control andpredict
theshieldbehavior without caserecords. Tosimulatetheshieldbehavior, thekinematicshieldmodel hasbeen
proposedbytheauthors, takingintoaccounttheexcavatedarea, thetail clearance, therotationdirectionof cutter
face, theshieldslide, andthedynamic equilibriumcondition. This paper reports thesimulationresults of the
articulatedshieldbehavior duringexcavationat asharpcurve. Thisstudyyieldsthefollowingfindings: 1) The
kinematicshieldmodel simulatesthemeasuredpathof theshieldreasonably; 2) Thearticulatedangle, thecopy
cutter areaandlengtharethepredominant factorsaffectingtheshieldbehavior.
1 INTRODUCTION
Recently, aiming at the completeness of city func-
tions in urban areas, many infrastructures such as
roadways, water supply systems, sewerageand elec-
tric power lines, etc. havebeen constructed densely.
Dueto spatial limitation, sometimes it is difficult to
choosesimplyanideal routefor anewtunnel. Onthe
otherhand, withnewtechniquesadoptedcontinuously,
shieldtunnelingmethodhas achievedgreatly. Under
these circumstances, many shield tunneling cases at
sharpcurvehavebeenreported. Inthecaseof exca-
vation at a sharp curve, articulated shield becomes
popular since conventional single shield met some
difficultiesinoperational control andgeneratedwide
rangeof grounddisturbance.
The direction control systems have been applied.
However, thesesystems arebasedonempirical rela-
tionships and are lacking in the precise theoretical
background (Shimizu & Suzuki 1994). Therefore it
issometimesdifficult tocontrol theshieldat asharp
curveandtopredict theshieldbehavior without case
records.
To clarify the shield tunneling behavior and the
behavior of thesurroundingground, numerical meth-
ods suchas FEM, DEM havebeenadopted(Komiya
etal. 1999; Kasper&Meschke2004; Melis&Medina
2005). The enforced displacement by means of the
gapbetweentheexcavatedsurfaceandthetunnel lin-
inghasbeenappliedinFEM andthistechniquegave
a very good prediction of theground movement for
shieldtunneling(Rowe& Lee1992). DEM isproved
tobeeffectivefor soil stabilityproblemattunnel face.
However, thesenumerical methodsrequiretheshield
movementasoneof knownconditions.Theimmediate
ground movement during excavation stage is diffi-
cult to be simulated, since it is related to the shield
movement andtheexcavatedarea.
Bytakingintoaccountgrounddisplacementaround
theshield, thekinematic shieldmodel hadbeenpro-
posedforsinglecircularshield(Sugimoto&Sramoon
2002). This model was validated by the simulation
of an earth pressurebalanced (EPB) shield behavior
along astraight alignment in asinglelayer of sandy
gravel (Sramoon et al. 2002) and along acurvein a
multilayeredground(Sugimotoetal.2007).Extending
thismodel toarticulatedshield, thearticulatedshield
model wasdevelopedandvalidatedbysimulatingthe
steady behavior of curve-only excavation (Sugimoto
et al. 2002).
Thetransient behavior of anarticulatedshieldina
way of sharpcurvaturemovement is of interest tobe
cleared.Inthisstudy,thesimulationresultsof theartic-
ulatedshieldbehavioratasharpcurvewithradius20m
arereported.Thevalidityof themodel isexaminedby
comparing the simulation results with the measure-
ment data. Furthermore, the factors affecting shield
behavior arealsodiscussed.
2 KINEMATIC SHIELDMODEL
The articulated shield model is composed of five
forces: forcedueto self-weight of machine, f
1
; force
761
ontheshieldtail, f
2
; forceduetojackthrust, f
3
; force
onthecutter disc, f
4
; andforceontheshieldperiph-
ery, f
5
, as illustratedinFig. 1(Sugimoto& Sramoon
2002). Amongthem, f
1
andf
5
act onbothsectionsof
theshield. f
3
comesfromshieldjacksandarticulated
jacks. f
5
iscomposedof thegroundreactionforceand
thedynamicfrictional forceontheshieldskin, which
areduetotheearthpressureactingontheshieldskin
plate. Sinceearthpressureisreliant ongrounddefor-
mation, andexcavatedcross sectionareais usually a
littlebit larger thantheshieldcross sectionarea, the
ground reaction forcefromtheground to theshield
canbeobtainedbyconsideringthecoefficientof earth
pressureK, which is given by afunction of thedis-
tancebetweentheoriginal excavatedsurfaceandthe
shieldskinplateU
n
, as showninFig. 2. InFig. 2, K
at U
n
=0 means the coefficient of earth pressure at
rest K
0
, and the gradient of K at U
n
=0 represents
thecoefficient of subgradereactionk. Here, notethat
the subscripts h and v mean horizontal and vertical
directions, respectively; thesubscriptsmin, o, andmax
defineminimum, initial, andmaximum, respectively;
and
v0
is overburden pressure. K in any direction
K
n
can be calculated by interpolation between K
h
andK
v
.
Theshieldbehaviorisrepresentedbythemovement
of theshieldinx, y, andz directions(Lx, Ly, Lz), and
theshield postures (yawing angle
y
, pitching angle
p
, and rolling angle
r
). Since the change of
r
is
limitedinpractice, thefactor of shear resistancedue
to the cutter torque
sg
was adopted as a parameter
insteadof
r
.
Figure1. Model of loadsactingonarticulatedshield.
Thearticulatedshieldbehaviourcanbeobtainedby
solvingthefollowingequilibriumconditionsof forces
andmoments:
where F and M are the force and moment vectors
andthesubscriptsF andR denotefront andrear sec-
tions of theshield respectively. Themoment vectors
are generated by the cross product of the position
Figure2. Groundreactioncurve.
762
vector and theforcevector. Here, thesuperscript M
indicates themachinecoordinatesystem, which can
be transformed to total coordinate systemby using
transformationmatrices.
3 SHIELDTUNNELLINGSITE
3.1 Test site description
The test site was established at a cable tunnel. The
total lengthof this cabletunnel is 264mandthetest
siteisabout60mlongataleftwardsharpcurvesection
withradiusof 20m.Theinclinedgradientof thetunnel
alignment isupwardof 0.2%.
Figure 3 shows the geological profile at the test
section, wheretheoverburdendepthisabout34mand
thegroundwater level is2.8mbelowthegroundsur-
face. Table1showsgroundproperties. Thegroundis
composed mainly of alluvial layers (As1, Ac1, Acs,
As2, andAc2) anddiluvial layers(Ds1, Dc2, Dc3, and
Figure3. Geological profileat test site.
Table1. Soil parametersat test site.
Unit weight Cohesion Frictionangle Coef. of subgrade Coef. of earth Youngsmodulus
Soil layer (kN/m
3
) c (kN/m
2
) (
) reactionk (kN/m
3
) pressureat rest K
0
E (kN/m
2
)
Bs 19.0 0 26 4737 0.562 8400
As1 19.0 0 38 9474 0.384 16800
Ac1 15.3 59 0 789 0.844 1400
Acs 18.1 0 24 3158 0.593 5600
As2 19.2 0 29 7895 0.515 14000
Ac2 15.4 50 0 2368 0.832 4200
Ds1 19.2 0 37 19737 0.398 35000
Dc2 15.8 91 0 8684 0.848 15400
Dc3 15.0 160 0 7895 0.847 14000
Ds3 19.7 0 45 67895 0.293 120400
Ds3).Thetunnel isexcavatedinthediluvial sandlayer
Ds3 with N-values over 50 by standard penetration
test.
Thearticulatedslurryshieldwithouter diameter of
3.95mand 5.765min length was used. After exca-
vation, thereinforcedconcretesegmentswithoutside
diameter of 3.8mand1.2minwidthwereinstalledat
thestraightsections. Thesteel segmentswith0.3min
widthwereadoptedat thecurvesection. Thedimen-
sions of thetunnel andtheshieldaresummarizedin
Table2.
Table2. Dimensionof tunnel andmachine.
Item Component Value
Tunnel Horizontal curveradius 20m
(leftward)
Vertical slopegradient 0.2%
(ascending)
Overburdendepth 34m
Groundwater level G.L. 2.8m
Outer radiusof segment 1.9m
Widthof segments 1.2m, 0.3m
Shield Outer radius 1.975m
Total length 5.765m
Lengthof front section 2.24m
Lengthof rear section 3.525m
Self-weight 810kN
Openratioof cutter face 20.0%
Thicknessof cutter face 0.35m
Radiusof chamber 1.943m
Lengthof chamber 0.80m
Radiusof cutter face 1.98m
Shieldjack Number of jacks 12
Cross-sectional area 346.361cm
2
Radiusof jack 1.67m
Articulatedjack Number of jacks 12
Cross-sectional area 433.736cm
2
Radiusof jack 1.47m
763
3.2 In situ data
Figure4showsthedataof tunnelingoperation, shield
behavior, andexcavationconditionmeasuredcontinu-
ouslybyautomaticmeasurement system. Theparam-
eters of tunneling operation are articulated angle in
horizontal direction
CH
(+: leftward), areaof applied
copy cutter CC range (measured fromthe invert of
shieldinclockwisedirection, viewedfromshieldtail),
jackthrustF
3r
, horizontal jackmomentM
3p
(+: right-
ward), vertical jackmoment M
3q
(+: downward), and
cutter torqueCT (+: anticlockwisedirection, viewed
fromshield tail). Theparameters of observed shield
behavior are defined as yawing angle
y
(+: right-
ward), pitchingangle
p
(+: downward), androlling
angle
r
(+: clockwisedirection, viewedfromshield
tail).Theparametersof excavationconditionareshield
velocity v
s
, slurry pressure
m
, andslurry density
m
inthechamber, whichareusuallycontrolledtostabi-
lizethetunnel face.ThemuckingratioR
v
istheratioof
themeasureddischargedsoil volumetothetheoretical
excavatedsoil volume.
Thearticulationof theshieldwasappliedtonego-
tiate the sharp curve. To follow the planed align-
ment,theappliedarticulatedangleincreasedgradually
fromthe beginning point of the curve, then it kept
steady valueof 500minutes, whichsuitsfor theleft-
ward curve of 20mradius. CC was approximately
0.1m in length and was applied mainly in range
of 30
180
(measured from the shield invert in

clockwisedirection, viewedfromtail) toincreasethe
excavatedareaaroundthecutter disc, whichreduces
theactingearthpressureontheshieldskinplateand
makesashieldadvanceeasily. F
3r
wasappliedtodrive
the shield forward against the earth pressure at the
cutter faceandthedynamicfrictionaroundtheshield
skin plate. After the transient section fromcurve to
straight line(fromthedistance75mto thedistance
80m), F
3r
increased obviously. M
3p
was applied to
turntheshieldtofollowthehorizontal leftwardcurve.
M
3q
was mainly appliedagainst thevertical moment
due to the earth pressure on the cutter disc and the
self-weight of theshield. CT wasgeneratedduetothe
shearingresistanceonthecutter disc.
Theobserved
y
revealsthedetailedyawingbehav-
ior of theshield. Theobserved
p
indicates that the
shield negotiates the inclination of the tunnel align-
ment. Theobserved
r
fluctuates at thesharp curve
alignmentandissomewhatrelativetoCT andtherota-
tiondirectionof thecutterdisc.Thispointsoutthatthe
shieldrollsarounditslongitudinal axisintheopposite
rotationdirectionof thecutter disc.
Theshieldvelocityv
s
decreasedatthesharpcurve.
To stabilize the face,
m
was applied based on the
lateral earth pressureat thetunnel face, and
m
was
kept approximately 12kN/m
3
. R
v
was closeto unity
throughout thetest site, whichindicatesthat excellent
excavationcontrol hasbeenachieved.
Figure4. Shieldtunnelingmeasurement data.
764
Figure5. Simulatedandobservedshieldtraces.
Figure6. Simulatedandobservedshieldbehavior.
4 SIMULATIONRESULTS
4.1 Shield behavior
Figure 5 shows the simulated and observed shield
traces. Fromthis figure, themaximumdifferenceof
1cmfor vertical position and the maximumdiffer-
enceof 3cmfor horizontal position can beverified.
Thesimulatedandobservedtimedependent parame-
ters
y
,
p
, v
s
arecomparedinFig. 6. Thesimulated
y
indicates that the shield performs good negotia-
tiontothesharpcurve. Asfor thesimulated
p
, there
is about 15 minutes uplift at the first straight sec-
tion (fromthe distance 30mto the distance 42m),
comparedwiththeobserveddata. Fromthedistance
60mto the end point of the test site, maximum20
minutes uplift can be found. Considering the pos-
sibility of change of geological conditions at some
locationsandthe5minutesprecisionof theinclinome-
ter, these differences are acceptable. Except for the
transient section fromcurveto straight line, thecal-
culated v
s
is generally consistent with theobserved.
At the transient section, the maximumdifference of
0.01m/min for v
s
is revealed. According to thecon-
structionreport, intermittent excavationis appliedto
confirmthetail clearanceatthissection, thereissome
possibilitytorecordslowervelocitythanthatobtained
inthesimulation.
4.2 Ground-shield interaction
Ground-shield interaction is discussed by using the
calculateddistancebetweentheoriginal excavatedsur-
faceand theshield skin plateU
n
and thecalculated
normal effectiveearthpressureactingontheshield
ns
atstraightlineandsharpcurve. Figs. 7and8showthe
distributionof U
n
and
ns
aroundtheshieldperiphery
at thestraight linewithdistanceof 39.3mandat the
sharpcurvewithdistanceof 58.4mrespectively. Here,
notethattheshieldperipheryisunfoldedasaflatplate,
i.e., thevertical axisshowsthelengthof theshieldand
thehorizontal axisrepresentsthecircumferenceof the
shield.
In the case of excavation at straight line, the fol-
lowingarefoundfromFig. 7: (1) thecontour linesof
U
n
becomedensearound45
, 80
, 260
, 335
andU
n
is about 30mmaround 60
and from270
to 300
,
becausethecopy cutter is applied from300
to 90
showninFig. 4andtheeffectiverateof over cutting

decreased between 30
and 330
by the assumption
that themuckingat theinvert isnot sufficient; (2) the
distributionof U
n
at thecrosssectionof theshieldis
almost samealong thelongitudinal direction, which
is natural at straight line; (3) The contour lines of
ns
appear aroundtheinvert andthecrownof shield,
sinceU
n
aroundbothspringlinesissmaller thanU
n
at
otherareaandthehorizontal effectiveearthpressureis
smaller thanthevertical oneduetoK
0
of D
s3
=0.293
showninTable1.
Whenshieldexcavatedat sharpcurve, thefollow-
ing are found fromFig. 8: (1) U
n
becomes positive
at theendof thefront bodyandtherear bodyaround
theright spring lineof shield. At thesametime, U
n
becomespositiveatthemiddlelengthof thebothbod-
ies alongtheleft springline. Theshieldskinplateat
theselocations pushes theground, whereas U
n
at the
oppositesidebecomesnegative, wheretheearthpres-
sureis inextensionstate. Thesecharacteristics result
fromtheequilibriumconditionandcorrespondtothe
leftwardcurveof thetunnel alignment; (2) thecontour
765
Figure7. U
n
and
ns
aroundshieldat thestraight linewithdistanceof 39.3m.
Figure8. U
n
and
ns
aroundshieldat thesharpcurvewithdistanceof 58.4m.
linesof U
n
becomedenseattherightbottomof therear
part andat theleft bottomof thefront part, sincethe
applying rangeof copy cutter shifts at theboundary
positionduetothechangeof cutterfacerotationdirec-
tionshowninFig. 4; (3) U
n
alongbothspringlineshas
afluctuationdueto thewrigglemotionof theshield
duringexcavation; (4) theintensity of
ns
appears at
theareawhereU
n
ispositive, whichisreasonablefrom
theviewpoint of ground-skinplateinteraction.
5 CONCLUSIONS
The articulated shield behavior at sharp curve was
simulated and the calculated shield behaviour was
compared with the observed one. Furthermore,
ground-shieldinteractionwasdiscussedusingthedis-
tributionof U
n
and
ns
aroundtheshield. Asaresult,
thefollowingconclusionscanbemade:
1. Thekinematic shield model for articulated shield
simulatedtheshieldpositionwithin3cmdifference
and the shield pitching angle within 20 minutes
difference at sharp curve. This indicates that the
proposed model can simulate shield behavior at
sharpcurvereasonably.
2. Thearticulatedangleandthecopy cutter areaand
length are the predominant factors affecting the
shieldbehavior especiallyat sharpcurve.
766
REFERENCES
Kasper, T. & Meschke, G. 2004. A 3D finiteelement simu-
lationmodel for TBM tunnelinginsoft ground. Interna-
tional Journal for Numerical and Analytical Methods in
Geomechanics 28(14): 14411460.
Komiya, K., Soga, K., Akagi, H., Hagiwara, T. &
Bolton, M. D. 1999. Finite element modelling of exca-
vationandadvancement processes of ashieldtunnelling
machine. Soils and Foundations 39(3): 3752.
Melis, M. J. & Medina, L. E. 2005. Discrete numerical
model for analysisof earthpressurebalancetunnel exca-
vation. Journal of Geotechnical and Geoenvironmental
Engineering 131(10): 12341242.
Rowe, R. K. & Lee, K. M. 1992. Subsidenceowingto tun-
nelling: II. Evaluationof predictiontechnique. Canadian
Shimizu, Y. & Suzuki, M. 1994. Movement characteris-
tics and control method of shield tunnelling machine
of articulate type. Transactions of the Japan Society of
Mechanical Engineers (Series C) 60(571): 141148. (in
J apanese)
Sramoon, A., Sugimoto, M. &Kayukawa, K. 2002. Theoret-
ical model of shieldbehavior duringexcavationII: Appli-
cation. Journal of Geotechnical and Geoenvironmental
Sugimoto, M. & Sramoon, A. 2002. Theoretical model of
shield behavior during excavation I: Theory. Journal of
138155.
Sugimoto,M.,Sramoon,A.,Konishi,S.&Sato,Y.2007.Sim-
ulationof shieldtunnellingbehavioralongacurvedalign-
ment inamultilayeredground. Journal of Geotechnical
and Geoenvironmental Engineering 133(6): 684694.
Sugimoto, M., Sramoon, A., Shimizu, T., Dan, A. &
Kobayashi, T. 2002. Simulation of articulated shield
behavior byin-situdatabasedonkinematicshieldmodel.
Journal of Tunnel Engineering 12: 471476. (inJ apanese)
767
Deformationandporepressuremodel of thesaturatedsiltyclay
aroundasubwaytunnel
Z.D. Cui &Y.Q. Tang
X. Zhang
ABSTRACT: Shanghai subwayLineNo. 2passesthroughthecenterof Shanghai fromSonghongRoadstation
inthewesttoZhangjianggaokestationintheeast.Thetotal lengthof Shanghai subwayLineNo. 2is25km. Con-
tinuousdynamicmonitoringisconductedbymeansof embeddedearthpressurepiezometersandporepiezome-
tersaroundthetunnel. BasedoncontinuousfieldmonitoreddataandusingthelaboratoryGDS(Global Digital
System)testapparatus,thedevelopinglawof theporewaterpressureof thesaturatedsiltyclayaroundthetunnel is
exploredwiththedistance. BytheGDStest, themodel of theincreasingporewaterpressureof thesaturatedsilty
clayisputforwardunderthevibrationloading.Itisalsoamendedbyfieldmonitoreddata.Theporepressuremodel
andcharacteristicsof deformationof thesaturatedsilty clay of Shanghai under thesubway trainareanalyzed.
Theresult offersavaluablereferencetothedesign, constructionandthesafeoperationof thesubwaytunnel.
1 INTRODUCTION
Thelawof increasingporewaterpressureundervibra-
tionloadingisanimportantfactor for thedeformation
andstrengthof soil andisalsothekeytouseeffective
stresstheoryfor thedynamicanalysis. Itisof primary
importance to predict correctly the changing law of
porewater pressureof soil. Researchershavestudied
the increase and dissipation law of pore water pres-
sureinsandysoil under vibrationloading. Zenget al.
(2005) andGuanet al. (2004) studiedthecharacteris-
tic of porewater pressureinsilt andsilty sandunder
thecyclicloadingandputforwardthechanginglawof
porewater pressure. Zhou et al. (2002) andPradhan
etal. (1998) triedtofindtherelationbetweenthepore
water pressureand thepath of strain and stress, and
summeduptheeffect onthedevelopment of thepore
water pressure in the liquefaction of the sandy soil.
Lee et al. (1975) and Shao et al. (2006) studied the
test parametersof deformationcapabilityof thesatu-
ratedsandysoil under thecyclicporewater pressure.
Li et al. (2005) analyzedthemechanismof liquefac-
tionof silt, influencefactors andthedevelopinglaw
of porewater pressureof silt in thecourseof vibra-
tion. Guo et al. (2005) studiedtheeffect of different
directionsof theprincipal stressonthecharacteristic
of undrainedcirculationof saturatedloosesandinthe
course of vibration. Zeng et al. (2001) studied the
increase and dissipation law of saturated clay under
shockloading. Menget al. (2004) studiedthecharac-
teristicof dynamicresponseof porewater pressurein
saturatedsiltyclayunder shockloading. Butthereare
few researches on the developing law of pore water
pressureof saturatedclayunder thelongtermsubway
vibrationloading. Thegraincomposition, mechanical
characteristicsandmoisturemigration,etc.of thesatu-
ratedsiltyclayaredifferentfromsandysoil.Thespeed
of theincreaseanddissipationof porewater pressure
underlong-termsubwayvibrationloadingisrelatively
slowandthedevelopinglawof porewater pressureis
also different fromsandy soil. Based on continuous
fielddataandwiththeuseof GDS inthelaboratory,
thispaper studiesthelawof theincreasingporewater
pressure and deformation of the saturated silty clay
aroundthesubwaytunnel at different depths.
2 DYNAMIC MONITORING
Inorder tostudy theinfluencecausedby thesubway
vibration loading on the saturated silty clay around
thetunnel, thefieldtestandmonitoringareconducted
inthisresearch. ThesiteisselectedbetweenJ ingansi
StationandJ iangsuRoadStation. Thedynamicmon-
itoringsystemisadoptedfor fieldmonitoringandits
769
samplingfrequencycanreach200Hzanditsprecision
is0.1kPa. It canfullyreflect thesoil responsearound
thetunnel dueto thesubway vibration loading. The
dynamic monitoring systemconsists of a resistance
sensor, adynamicstrainamplifier, adataselector and
acomputer. Thesystemcan record all thesampling
datacollectedbythecomputer inreal time.
Figure1shows thelayout of boreholes at thesite.
Intheplane, therearefiveboreholes, each110mmin
diameter, parallel and vertical to thesubway tunnel,
respectively. The distance between the site and J in-
gansi Station is 210m. Boreholes No. 3, No. 4 and
No. 5areparallel to thetunnel axis only 1.8maway
fromtheoutsideof thesegment of thesubwaytunnel
and thedistancebetween themis 15.0m. Boreholes
No. 1, No. 2, andNo. 3arevertical tothetunnel axis.
Inordertostudytheattenuationof theeffectonthesoil
aroundthetunnel withtheincreasingdistanceunder
the subway vibration loading, the distance between
No. 1 and No. 2 is 3mand that between No. 2 and
No. 3is2m, sothatthereisastep-upcourse. Figure2
showsthedistributionof strataandinstruments. Inthe
section, the subway tunnel lies in gray silty clay of
layer No. 4. Theearthpressurepiezometersandpore
piezometersarelocatedat thedepthsof 8.5m, 11.5m
and13.5m, respectively, inlayer No. 4, tomonitor the
responsecharacteristicof thevibrationfor thesubway
running.
3 GDSTEST
Under thenatural stress, thestressstateof theundis-
turbed saturated silty clay is in thek
0
consolidation.
The response amplitudes and response frequency of
soil have not been studied under the subway vibra-
tion loading. In order to study the characteristics of
soil under the subway vibration loading, this paper
usesfieldmonitoreddataandfieldinvestigationdata
3000
2000
15000 15000
1
8
0
0
Zhong Shan Park
Monitoring
Room
People Square
Subway
No.5 Hole No.3 Hole
No.4 Hole
No.2 Hole
No.1 Hole
Figure1. Layout of boreholes(unit: mm).
to design the laboratory test. GDS (Global Digital
Systems) apparatus is used in the test, as shown in
Figure 3. It can monitor the test process at the real
timeandcollect dataat highspeedandstorethem. It
has merit of high precision, easy operation, reliable
results, etc. ItsworkingprincipleisshowninFigure4.
In the cyclic test, the soil samples are first satu-
ratedunder back pressure. In order to simulatefield
conditions to the utmost, the samples are consoli-
dated under k
0
condition. Theconfining pressure
h
isobtainedbycalculationaccordingtothenatural soil
strata, that is,
h
=k
0
v
(
v
=

i
h
i
). After consol-
idation, the cyclic triaxial test begins and the cyclic
stress
d
shouldsimulatethebearingdynamicloading
totheutmost.
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~ ~
2.70m
7.80m
20.0m
18.0m
~
~
~
~
~
~
~ ~
~ ~
~
~
~
~
~ ~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~ ~
20.0m
Strain gauge
1.8m
8.5m
11.5m
13.5m
R3m
Tunnel
1.30m 1
2
3
4
5
~
~
~
~
~
~
~
~ ~
~
~
~ ~
~ ~
~
~
~
~
~
~
~
~
~
~
~ ~
~ ~
~
~
~
~
Figure2. Soil profileandembeddedinstruments.
Note: No. 1layer ismixedsoils; No. 2layer isbrownyellow
siltyclay; No. 3layer isgraymuckysiltyclay; No. 4layer is
graymuckyclay; No. 5layer isgraysiltyclay.
Figure 3. GDS system for dynamic multi-function
triaxial test.
770
Bythecontinuousdynamicfieldmonitoring,plenty
of dataof boreholesareobtained. Bycollecting, ana-
lyzing and arranging them, two kinds of response
frequency of soil areobtainedwhenthesubway train
runsacrossthesite.Thehighfrequencyis2.42.6Hz
andthelowfrequency is 0.4 0.6Hz. Thetest uses
thefrequencyof 2.5Hz.
Moreover, thestressresponseamplitudesof soil at
differentdepthsunderthesubwayloadingareobtained
by fieldmonitoring. Themaximumchangingampli-
tudes of soil stress response are 0.23kPa at 8.5m,
0.70kPaat11.5mand1.15kPaat13.5m.Thechange
of soil stress amplitudeis approximately linear with
depth. Moreover, thestressamplitudeintherushhour
inthemorningislarger thanthatintheeveningwhich
is also larger than that at noon. The test adopts the
maximumstressamplitudesasreferencefor theworst
caseinconstruction.
The field monitored data indicate that the soil
dynamicresponsearoundthetunnel under therailway
vibrationloadingisthecyclicresponse.Therefore, the
test usesthestress-controlled, cyclicloadingmodule.
TheschemeisshowninTable1.
4 DEVELOPINGLAWOF PORE WATER
PRESSURE
Throughthetest dataof GDS, thecurveof porewater
pressurewithincreasingtimeisobtained, asshownin
Figure5.
Figure4. Workingprincipleof GDS.
Table1. Schemeof undraineddynamiccyclictriaxial test.
Axial Confining Back Dynamicload Circular
Test Sampling Simulating pressure pressure pressure amplitude stressratio Frequency
number depth(m) depth(m) (kPa) (kPa) (kPa) L(kPa) CSR
(Hz)
D1 8.09.0 8.5 155 130 80 0.23 0.001 2.5
D2 11.012.0 11.5 220 180 110 0.7 0.003 2.5
D3 13.014.0 13.5 250 200 130 1.15 0.006 2.5
Theformulaof circularsstressratio: CSR=

L
2
(Tang, 2003), with
=

1
+
2
+
3
3
=

1
+2
3
3
(
2
=
3
).
Underthedifferentconsolidatedpressures, thepore
water pressures increases quickly along the straight
lineunder theoriginal loading, thenthecurvebends.
The increase becomes slow. After a long-time, the
increase of pore water pressure becomes smooth.
Finally, theporewaterpressurereachesthelimitvalue.
The limit values of the three groups of pore water
pressure are 115kPa, 160kPa and 185kPa, respec-
tively. They are 80% of their effective confining
pressures.
Fromtheaboveanalysis, theincreaseof porewater
pressurecanbedividedintothreeobviousstages: the
sharply increasing stage, theslowly increasing stage
andthesmoothstage, asillustratedinFigure5(c).
1 Sharpincreasingstage(AB)
After loading, the excess pore water pressure
increases rapidly in astraight line, and in ashort
time (about 1300s, 3250 times of vibration), it
reaches50%of thelimitvalue.Thentheincreasing
speedattenuatesrapidlywiththeincreasingnumber
of vibrationandattheendof thestage, itreachesa
stablevalue.
2 Slowlyincreasingstage(BC)
Asatransitionstage, theexcessporewater pres-
sure comes to a smooth stage fromthe sharply
increasing stage. The pore water pressure still
increases, but the increasing speed obviously
becomesslow.
3 Smoothstage(CD)
In this stage, the pore water pressure hardly
increaseswiththeincreaseof thenumber of vibra-
tion. It reaches a stable value called the limit
value, whichisabout80%of itseffectiveconfining
pressure.
Whenthecharacteristicsof siltyclayarecombined
with engineering conditions in Shanghai, Logistic
model of pore water pressure is put forward (the
correlationcoefficient reachingover 0.99):
whereN is thenumber of vibration; u
0
is theannual
valueof hydrostatic pressure; u is thelimit valueof
excessporewater pressure; N
0
and p are regression
771
Figure5. Theincreasingporewater pressurewithtime.
parameters, whichareobtainedbyregressionanalysis
for test data, asshowninTable2.
5 FITTINGOF FIELDMONITORINGDATA
The depth of 11.5mis chosen as an example. The
changing curve of the field monitored pore water
Table2. Parameter tableof themodel.
Corresponding
Number u
0
u
t
N
0
p coefficient
D1 78.96 128.22 8125.4 0.43 0.9933
D2 106.44 173.00 6383.2 0.59 0.9979
D3 132.94 191.73 2130.0 0.65 0.9978
pressure under the subway loading is shown in Fig-
ure6. y Axis(unit: kPa) istheresponsevalueof pore
waterpressureandx axis(unit: s) istime.Accordingto
thecalibrationcoefficients, it is transformedintothe
curveof excess porewater pressurewiththenumber
of vibration, asshowninFigure7.
ComparedFigure7withFigure5(b), theincreas-
ingspeedof thefieldmonitoredporewater pressure
is slower than that of the test, and the increasing
amplitudeis also slower. Themain causeis thesize
effectandthedifferencebetweenthefieldtestandthe
laboratoryone. Thecoefficient c isusedtoamendthe
developingmodel of porewater pressureinFormula
(1) andtheamendingmodel isobtained:
whereC istheamendingcoefficient.
Substituting the field monitored pore water pres-
sures at 8.5m, 11.5mand 13.5mand their number
of vibrationintoFormula(2), wehavethevalueof c,
0.0350for D1, 0.2112for D2and0.2975for D3.
Similarly, thedepthof 11.5mischosenasanexam-
ple.TheamendingparameterissubstitutedinFormula
(2) andthefittingcurveisobtained. Thefittingcurve
is comparedwiththemonitoredcurveandtheresult
is shown in Figure 8. Fromthe figure, the value of
excess porewater pressureinFormula(2) is closeto
themonitoredvalueandtheyarefittingwell.
The field monitored data indicate that the excess
porewater pressureis not bigwhenthesubway train
isrunningacross.Thenitdissipatesquickly.Theinter-
val of the subway train is 35 minutes, when the
excess porewater pressurecanalmost bedissipated.
Thatis, basically, theexcessporewater pressurespro-
ducedbythetwoadjacenttrainscannotbesuperposed.
Therefore, thedevelopinglawisat theinitial stageof
Formula(2) only, that is, thesharplyincreasingstage.
6 DEFORMATIONOF THE SATURATED
SILTY CLAY
The model can be used to predict the excess pore
water pressurewhenthesubwayisrunningacrossand
thedeformationandthechangeof stress of soft clay
aroundthetunnel areanalyzedby theeffectivestress
772
Figure6. Theoscillogramof fieldporewater pressureat 11.5m.
Figure 7. The curve of excess pore water pressure and
number of vibration.
theory.Theyofferatheoretical referencetothesubway
operation.
The energy of the subway loading is delivered to
thesoil by thesidewall of thetunnel andthelining.
Thesensitivity of porewater is morethanthat of the
grains skeleton. At the start of the subway loading,
porewater absorbsall theenergyof thesubwayload-
ing, whichresultsintherapiddeclineof theeffective
principal stressof thesaturatedsiltyclayandtheelas-
tic dischargeof thetexturecell of thesoil. With the
increasing number of vibration of the running way,
theeffectiveprincipal stressof thesaturatedsiltyclay
Figure8. Thecomparedcurveof fieldmonitoringdataand
fittingdata.
reaches asteady valuegradually andthetexturecell
of soil startstobear theenergyof thesubwayloading.
Theweak connectionof thesoil cell startstobecome
looseandbreaksandslight cracksoccur, but thesoil
cell is still integrated. Withaccumulatingenergy, the
crack of the soil cell expands gradually and the big
soil cell splitsintosmall soil cell andcrumb.Theshear
zoneoccursinaplaceof seriousbreaking. Thesmall
soil cell andcrumbintheshear zonearecrushedand
deformation occurs. The accumulation of microcos-
micdeformationresultsinthedeformationof theaxis
of thesubwayandthegroundsettlement.
773
Thedatamonitoredshowthat theaxial settlement
of thesubwayLineNo. 1exceeds20cminsometun-
nel sections, whichgreatlyaffectstherunningof Line
No. 1andcausessomeoldhousescracking. Withthe
lapseof time, plasticdeformationandlargeasymmet-
rical settlement occur onthebottomof thesubgrade,
whichwill affectthesubwayrunningandwill possibly
result ingroundsettlement, instabilityof thebuilding
foundationnear thesubway, crackingandincliningof
the building, etc. Although no large deformation of
soil occurs atethesidewall of thetunnel inaperiod
of time, butaperceptibledeformationoccurspossibly
at thesidewall of tunnel withthepassingof timeof
thesubwayworking.
7 CONCLUSIONS
This paper uses field monitored data as a parame-
ter for the GDS test. This guarantees the reality of
test conditions. Logistic model of pore water pres-
sureat different depths is put forward. Theincrease
of porewater pressurecanbedividedintothreeobvi-
ous stages: the sharply increasing stage, the slowly
increasingstageandthesmoothstage. Themodel is
comparedwithfieldmonitoreddata, findingthat the
increasing speed of the field monitored pore water
pressure is slower than that of the test and that the
increasing amplitudeis also slower. Themain cause
isthesizeeffect andthedifferencebetweenthefield
test andthelaboratory one. Then, theparameter c is
used to amend the developing model of pore water
pressureandtheamendingmodel is obtained. It can
be used to predict the increase of pore water pres-
sure when the subway train is running across. The
effectivestresstheorycanbeusedtoanalysethedefor-
mation and thechangeof stress of silty clay around
thetunnel. Theresult offersatheoretical referenceto
thefailuremechanismof thesaturatedsilty clay, the
axial deformationof thesubwaytunnel andtheground
settlement.
ACKNOWLEDGEMENTS
This work are supported by the research grant
(40372124) fromNational Natural Science Founda-
tion of China, Shanghai Key Subject (Geotechni-
cal Engineering) Foundation and Shanghai Leading
AcademicDisciplineProject(ProjectNumber: B308).
REFERENCES
Guan, Q.M., Zhou, S.H. & Wang, B.L. 2004. Variation of
pore pressure and liquefaction of soil in metro.Chinese
Journal of Geotechnical Engineering 26(2): 290292(in
Chinese).
Guo,Y., Luan, M.T., He,Y. &Xu, C.X. 2005. Effectof varia-
tionof principal stressorientationduringcyclicloadingon
undraineddynamicbehaviorof saturatedloosesands.Chi-
nese Journal of Geotechnical Engineering 27(4):403409
(inChinese).
Lee, K.L. & Focht, J.A. 1975. Strengthof clay subjectedto
cyclicloading. Marine Geotechonolgy 3(2):165168.
Li, L.Y., Cui, J., J ing, L.P. & Du, X.L. 2005. Studyonlique-
factionof saturatedsilty soil under cyclic loading. Rock
and Soil Mechanics 26(10): 16631666(inChinese).
Meng, Q.S. &Wang, R. etal. 2004. Porewaterpressuremode
of oozy silty clay under impact loading. Rock and Soil
Mechanics 25(7): 10171022(inChinese).
Pradhan,T.B.S.,Tatuoka, F. &Sato,Y. 1989. Stressdilatation
of sandsubjectedtocyclic loading. Soil and Foundation
29(1):3546.
Shao, L.T., Hong, S. &Zheng,W.F. 2006. Experimental study
ondeformationof saturatedsandunder cyclicporewater
pressure. Chinese Journal of Geotechnical Engineering
28(4): 428431(inChinese).
Zeng, C.N., Liu, H.L., Feng,T.G. &Gao,Y.F. 2005.Teststudy
onporewater pressuremodeof saturatedsilt. Rock and
Soil Mechanics 26(12): 19631966(inChinese).
Zeng, Q.J., Zhou, B., Gong, X.N. & Bai, N.F. 2001. Growth
and dissipation of porewater pressurein saturated silty
clayunderimpactload. Chinese Journal of Rock Mechan-
ics and Engineering 20(1): 11371141(inChinese).
Zhou, H.L. & Wang, X.H. 2002. Study on the pore water
pressureof saturatedsandindynamictriaxial test. Journal
of the China railway Society 24(6): 9398(inChinese).
774
Analytical solutionof longitudinal behaviour of tunnel lining
F.J.M. Hoefsloot
Fugro Ingenieursbureau, Leidschendam, The Netherlands
ABSTRACT: Stagedconstructionof segmentedtunnelsresult inpermanent andconstant bendingmoment in
thelongitudinal direction.Thisfacthasbeenconfirmedempirically, andanalytical solutionsfor thelongitudinal
behaviourof aboredtunnel lininghavebeenpresented.Thispapersummarizespublishedanalytical solutionsfor
simpleloadingconditions, andincludescorrectionsfor thesolutionswherenecessary. Solutionsfor additional
loadingconditionsrelevant toTBM tunnel constructionarealsopresented. Theanalytical solutionshavebeen
builtinapowerful Excel spreadsheetforrapidanalyses.Theresultshavebeenvalidatedwithstagedconstruction
FEMcalculationsinPLAXIS2D. For final validationof thestagedconstructionbehaviour, tunnellingdatafrom
theGroeneHarttunnel intheNetherlandswereanalysed. Segmentswereinstrumentedwithaxial straingauges,
andtheresults havebeenanalysedandconvertedto longitudinal bendingmoments. It has beendemonstrated
thatthemeasuredbehaviour isreproduciblewiththeanalytical model, althoughselectionof inputparametersis
complex.Thepresentedmodel iswell suitedforquickanalysesof TBMback-uptrainlayout, groutingconditions
andmomentsfromjackingforceswithrespect tolongitudinal liningbehaviour.
1 INTRODUCTION
TBM tunnellingis characterisedby phasedconstruc-
tionof thesegmentedlining. Excavationbythetunnel
boring machine is followed by erection of a single
ring of lining segments. The complete structure can
beregardedas abeamonelastic foundationaccord-
ing to classic structural engineering theory. In each
phase of construction, a load free member is intro-
duced, which contributes to the structural system.
Inthemeanwhile, loadingprogressessimultaneously
withtheTBM. Examplesarejackingforcesonthefirst
ring,buoyancyforceswithinthegroutingzoneandself
weight of theback-uptrain.
Previoustheoretical analysesof thestructural sys-
tem have shown that the distribution of bending
moment and shear forces is fundamentally different
fromawished-in-placebeamonanelasticfoundation.
Measurements fromthe Groene Hart tunnel clearly
confirmthisdifference.
2 ANALYTICAL SOLUTIONS
2.1 General procedure
Theprogressiveloadingconditionsare:
1 uniformloading
2 shear forceat front endof thebeam
3 bendingmoment at front endof thebeam
4 local uniformlydistributedload
Bogaards & Bakker (1999) have given analytical
solutions for the section forces of a beam on an
elastic foundation with progressive extension of the
beamunder auniformload. Bakker (2000) published
solutionsfor items1to3.
Section forces can be derived in several ways.
Bakkers (2000) derivation is explained and thecor-
rectedresultsaregivenbelow.
2.2 Uniform load
A stagedextensionof auniformlyloadedbeamcanbe
solvedasthesumof theanalytical solutionsof apar-
tiallyloadedbeam(Figure1,leftside).Themechanical
Figure1. Stagedconstructionuniformload.
775
schemeisequivalent totheschemegivenontheright
sideof Figure1.
ThesectionforcesM
q
(x) andD
q
(x) (Equation1to
4) caneasily besummedfromtheanalytical solution
of abeamloadedwithashear forceandthebending
moment at thebeginning of thebeam. Thesesimple
solutionshavebeengivenby Hetnyi (1946), Bouma
(1993) andYounget al. (2002).
with x=coordinate starting on the left side;
dx=lengthof beamincrement; k=modulus of sub-
gradereaction; E=Youngs modulus; I =moment of
inertia.
2.3 Shear force at front end of the beam
Figure2showsthemechanical schemeof ashearforce
at the front end of a stage-constructed beam. As in
Figure 1, the analytical solution of a staged (beam)
extension with shear force at the front end can be
foundasthesumof theanalytical solutionof apartially
loadedbeamasexplainedontheleft sideof Figure2.
The mechanical scheme is equivalent to the scheme
givenontheright sideof Figure2. Theresultingsec-
tion forces M
Q
(x) and D
Q
(x) (Equation 6to 8) have
beenderivedfromtheright sideof Figure2.
Figure2. Stagedconstructionshear force.
2.4 Bending moment at front end of the beam
Progressiveextension of thebending moment at the
front endof thebeamresults inthesimplesolutions
M
M
(x) andD
M
(x) inEquations9and10.
2.5 Local uniformly distributed load
Theanalytical solutionforaprogressivelocal uniform
loadcanbederivedasshowninthemechanical scheme
inFigure3. Theresultingsectionforcesarecomplex
andarenot presentedhere.
3 SPREADSHEET-MODEL
The mechanical scheme for the design of a tunnel
lining as a beamon an elastic foundation is given
in Figure4. An unsupported part of tunnel lining is
presentwithintheTBM(l
i
) andalsobetweentheTBM
andstablegrout (l
u
). Thefollowingloadsarepresent:
Bendingmoment fromjackforces(M
jack
)
Shear forcefromjackingforces(D
jack
)
Shear forcefromsteel brushes(D
br
)
Weight of liningsegments(q
w
)
Uniformly distributed load (q) starting behind the
TBM at distance(l) withalength(l
q
).
An Excel-spreadsheet model was built for this
mechanical schemewiththebasic solutions givenin
paragraph 2. The model is equipped with 6 local,
uniformly distributed loads to account for buoyancy
forces and load configurations of the back-up train,
backfill and permanent structure within the tunnel.
Therefore the model is a powerful tool to calculate
section forces for a wide variety of load conditions
simplybyenteringthenecessaryparameters.
776
Figure3. Stagedconstructionlocal uniformload.
Figure4. Loadschemeandsubgradereactiontunnel beam.
Determiningtheangular andvertical deformation
of the tunnel is not straight-forward. According to
classical structural engineeringtheory, theangularand
vertical deformationcanbefoundbyintegrationof the
bendingmoment.
Thesolution of astage-constructed tunnel results
inaconstant bendingmoment at great distancefrom
the TBM. Direct integration of this constant value
resultsinalinearincreaseof angulardeformation, and
aquadratic increaseof vertical displacement. Indeed,
finiteelementcalculationsof astage-constructedtun-
nel with strain-less extension of thebeamshowthis
type of deformation. Since the final tunnel must be
Figure5. Angular offset installationnewring.
positioned according to its design alignment, tunnel
rings must be installed with an inclination onto the
previous ring. Figure5 shows thebasic ideaof ring
installation.
Thespreadsheet-model isequippedwiththecalcu-
lation of angular and vertical deformation based on
thisprinciple. Part of theresult istherequiredinstal-
lationoffset anglerelativetothevertical andrelative
tothepreviousring.
Themodel has been verified using finiteelement
modeling with PLAXIS 2D version 8.4. For each of
thebasicloadingconditionsgiveninparagraph2.1, a
verificationhasbeenperformed. InPLAXIS, astruc-
tural beamonanelasticsoil masswasmodeledunder
theapplicableloadingconditions. Phasedcalculations
havebeen performed with each phaseextending the
beamandsoil support, andmovingtheloadonestep
ahead. Resultsformtheanalytical spreadsheet-model
and finite element model show good to excellent
agreement.
4 GROENE HART TUNNEL
For the High Speed Railway link between Amster-
damand Brussels, a 7.2kmlong TBM tunnel was
recently constructed. The tunnels outer diameter is
14.5mand the lining thickness is 0.6m. The Dutch
research committee COB (Center for Underground
Construction) organized an extensive measuring
campaign to study the longitudinal behavior of the
lining. Straingauges, tilt sensors andtwo systems to
measurevertical deformationwereplacedonsomeof
thetunnel rings.
4.1 Strain gauges
Straingaugeswereusedtomeasureaxial andtangen-
tial strain during construction. Pairs of gauges were
placedinaxial direction, bothontheinsideandonthe
outsideof thesegments.
After 4 days, the signals reach general equilib-
rium, with minor fluctuations dueto jacking forces.
Completeresults fromaxial gauges for ring number
2117areshowninFigure6. Theaverageof thestrain
measurements fromtheinsideandtheoutsideof the
segment aregiveninFigure7.
777
Figure6. Resultsof all straingaugesring2117.
Figure7. Averageresultsstraingaugesring2117.
In Figure 8 typical results of the strain distribu-
tion have been plotted against the vertical position,
shortlyafterhavingpassedthroughtheTBM.Thesame
has beendoneat theendof themeasuringcampaign
(Figure9). Theslopeof thetrend linethrough these
measurementschangesof signbetweenthebeginning
andendof themeasurementprogram.Thisaxial gradi-
ent isdeterminedbythebendingmoment inthecross
section.
4.2 Bending moment and normal force
Usingtheaxial strainattunnel axisandthegradientof
thetrendline, theaveragenormal forceandbending
momentcanbedeterminedwithEquations13and14.
with N=average normal force;
av
=average axial
stress;
av
=averageaxial strain; =axial straingradi-
ent; E=Youngs modulus=38500MPa; A=section
area=26.2m
2
; I =moment of inertia=634m
4
.
In Figure 10, the derived axial normal force and
bendingmoment fromstraingaugemeasurementsfor
Figure 8. Axial strain ring 2117, just after having passed
throughtheTBM, J une32003, 1:00hrs.
Figure9. Axial strain ring 2117, at theend of measuring
campaign, J une112003, 0:00hrs.
Figure10. Normal forceandbendingmoment fromaxial
strainmeasurement ring2117.
theentireloggingperiodareshown. Resultsof normal
forceandbendingmomentattimeof leavingtheTBM
havebeencomparedwithTBMdataof jackingforces.
A reductionfactor hasbeenincludedintheresultsof
Figure10. Whenareductionfactor of 0.9isappliedto
theaxial strain, theback-calculatedjackforcesmatch
thedirectlymeasuredjackforces.Thereductionfactor
takesintoaccounttheinfluenceof non-uniformstrain
distributionwithinthesegments.
Alsothesectionforcesasfunctionsof distancefrom
thetail end of theTBM arepresented in Figure11.
Theselinescanonlyberegardedasnormal forceand
778
Figure11. Normal forceandbendingmoment ring2117.
Table1. Input parameter.
Identification
Description Fig. 4 Valueunit
Outsidediameter 14.50m
Wall thickness 0.60m
Youngsmodulus 3.85E+07kN/m
2
Reductionfactor 0.650
stiffness
Modulusof subgrade 367,000kN/m
2
reaction
Lengthsegmentsin l
i
6m
tunnel
Unsupportedlength l
u
2m
Moment jackforces M
jack
79,000kNm
Shear forcejack D
jack
0kN
Shear forcebrushes D
br
0kN
bendingmomentdistributionalongthetunnel axisdur-
ingconstant tunnelling. FromFigure11it is evident
that thebendingmoment reaches aconstant valueat
approximately 60mbehind the TBM. This remain-
ing bending moment is aconsequenceof thestaged
constructionof tunnelling.
4.3 Back analyses
Theanalytical model hasbeenappliedtocalculatethe
distributionof thebendingmoment withinthelining
behindtheTBM.Theinputparametershavebeengiven
inTables 1 and 2. They consist of simplegeometric
andmaterial parameters andloadingconditions. Two
parametersneedfurtherexplanation. First, areduction
factor for the bending stiffness has been applied to
account for thestructural behaviour of jointsbetween
segments. Secondly, theexternal grout load requires
further analyses. Refer tothepaper onthissubject by
Talmonet al. (2008).
The calculated and measured bending moments
are in Figure 12. There is good agreement between
measurementsandcalculationresults. It isnotedthat
Table2. Input parameter; uniformlydistributedload.
Loadidentification Load PositionbehindTBM
Fig. 4 [kN/m] [m]
q
w
629 6to1000
q
1
2064 0to1000
q
2
437.5 2to26
q
3
70 30to1000
q
4
300 52to1000
q
5
187 82to108
Figure 12. Analytical model and measurement results
ring2117.
alternatecombinations of input parameters may also
resultinfairagreement. However, theanalytical model
clearlyshowsthesensitivityof different loadingcon-
ditions.Thereforethemodel iswell suitedtoanalysing
loading conditions like position of back-up train,
backfill, jackingforcesandgroutingconditions.
5 CONCLUSIONS
The mechanical behaviour of a stage-constructed
TBMtunnel isfundamentallydifferentfromawished-
in-place tunnel. Analytical solutions for bending
moment, shear force, angular deflection andvertical
displacement havebeenusedtocreateaspreadsheet-
model. Measurements fromthe Groene Hart Tunnel
clearlyshowaresidual bendingmomentfarbehindthe
TBM, aresultwhichisingoodagreementwithresults
of theanalytical model. Thelongitudinal behaviour of
thetunnel asaconsequenceof thestagedconstruction
needstobeaccountedfor intunnel design.
ACKNOWLEDGEMENT
Thework describedinthispaper hasbeenperformed
for the research committee F512 of the Centre for
UndergroundConstruction(COB).Theauthor wishes
779
tothank thecommitteefor theopportunitytopublish
thiswork.
REFERENCES
Bakker, K.J. 2000. Soil Retaining Structures. Rotterdam:
Balkema.
Bogaards, P.J. & Bakker, K.J. 1999. Longitudinal bending
in Geomechanics Proc. NUMOGVII: 317321.
Bouma, A.L. 1993. Mechanica van constructies. Delft:
DelftseUitgeversMaatschappij (inDutch).
Hetnyi, M. 1946. Beams on Elastic Foundations. Michigan:
TheUniversityof MichiganPress.
Talmon, A.M., Bezuijen, A. & Hoefsloot, F.J.M. 2008.
Longitudinal tube bending due to grout pressures.
Shanghai: TC28.
Young, W.C., Budynas, R.G. & Roarke, R.J. 2002. Roarkes
Formulas for Stress and Strain (seventhed.). NewYork:
McGrawHill.
780
Designof tunnel supportingsystemusinggeostatistical methods
S. J eon& C. Hong
School of Civil, Urban & Geosystem Engineering, Seoul National University, Seoul, Korea
K.You
Department of Civil Engineering, University of Suwon, Gyungki-do, Korea
ABSTRACT: Rockmassclassificationprovidesaguidelinefor atunnel excavationandreinforcementdesign.
Theboreholedataandgeophysical siteinvestigationresultshavebeenpopularlyusedforrockmassclassification,
butthelocalityandlimitedinformationfromtheboreholedataandqualitativecharacteristicsof geophysical data
havebeenproblematic. A geostatistical methodsuchas krigingcanbeanalternativetosolvetheseproblems.
Thispaperdescribesadesignof tunnel supportingsystembasedongeostatistical tools. Koreantunnel supporting
systemistypically composedof six different typesof combinationof shotcrete, rockbolts, andconcretelining
based on rock mass rating (RMR). Ordinary kriging (OK), indicated kriging (IK), and sequential indicator
simulation (SIS) wereused to estimateRMR around thetunnel. Kriging methods could estimateRMR with
the best linear unbiased estimator. Using SIS, RMR was presented in the probabilistic termsuch as mean,
variation, andconfidenceinterval. Reliability of theestimatedvalues was verifiedby split-samplevalidation
andcomparedwiththereal RMR obtainedfromthesidewall of thetunnel whileexcavatingcarriedout. Based
on100equallyprobablesimulations, RMRcouldbepresentedintheformof aprobabilitydistributionfunction
andtheuncertaintyof estimationcouldbesuccessfullyquantified.
1 INTRODUCTION
Tunnel excavationandreinforcementdesignaremade
according to therock mass classification. Engineers
havebeenusingtheboreholedataof rockmassclassi-
ficationandgeophysical siteinvestigationresults. Due
tothelocalityandlimitedinformationfromthebore-
holedataandqualitativecharacteristicsof geophysical
data, geostatistical methodsuchas krigingshouldbe
consideredfor therockmassclassification.
Krigingisoneof themostwidelyusedinterpolation
methodsingeostatistics. Therehasbeenconsiderable
researches conducted using this technique (Taboada
et al., 1997; Facchinelli et al., 2001; Marinoni, 2003;
Pardo-Igzquiza and Dowd, 2005). Despite its wide
use, thekriging map flattens out thelocal details of
thespatial variationwiththeoverestimationof small
valuesandunderestimationof largevalues. Thistype
of selectivebias is aserious shortcomingbecauseof
thelossof thedistributionfeaturesof theoriginal data.
Kriging is focused on theestimation of unknown
pointsbyonedeterministicvalue, whereassequential
simulation is on thestochastic simulation by proba-
bilistic form. J uang et al. (2003) showed the spatial
distribution of soil contamination by the sequential
indicator simulation, andFenget al. (2006) proposed
animprovedsequential indicator simulation.
Thispaper describesadesignof tunnel supporting
systembased on geostatistical tools. Typical Korean
tunnel supportingsystemwas composedof six types
of combination of shotcrete, rockbolts, and concrete
liningbasedontherockmassrating(RMR). Ordinary
kriging (OK), indicated kriging (IK), and sequential
indicatorsimulation(SIS) wereusedtoestimateRMR
around the tunnel. For IK, we estimated the three-
dimensional distributionof RMR withthefielddata
of borehole logging and geophysical data. And this
result was compared with the results using OK and
SIS. Using SIS, an equally probablesimulation was
performed 100 times to quantify the uncertainty of
estimation. Theaccuracy of estimation was checked
bysplit-samplevalidation.
2 ESTIMATIONPROCESS
2.1 Ordinary kriging
1 Construct avariogramfromthescatter point set to
beinterpolated.
whereh =lagdistance; z(x)=valueof positionx;
andn =number of total data.
781
2 Define a theoretical variogram. Spherical model
wasusedinthisstudy.
3 Calculatetheweightsforeachpointandestimation
valueisthelinear combinationof weightedknown
values.
withaconstraint
where
2
ab
isthecovariancebetweenaandb.
2.2 Indicator kriging
1 Determine thresholds of borehole data and seis-
micdata. Becauseboreholedataisquantitativeand
seismicdataisqualitativeregardingtoRMR, both
boreholeandseismicdatachangedintotheindica-
tors between 0 and 1. The thresholds were RMR
20, RMR 40, RMR 60, and RMR 80 for bore-
holedata, and800km/s, 1500km/s, 2400km/s, and
3600km/sfor seismicvelocity.
wherev
t
=indicator value; V(x)=datafunction.
2 Calculateanindicator of unknownnodeswiththe
sameprocessof ordinarykriging.
3 Convert an indicator into RMR using cumulative
probability distributionfunctionof estimatedfour
indicatorsfor four thresholds.
2.3 Sequential indicator simulation
1 Determine the thresholds (1st quartile, medium,
and3rdquartile)anddividethedataintoindicators.
Theindicator functionisgivenbyequation(5).
2 Calculate the experimental variogramand deter-
minethetheoretical variogramfor eachthreshold.
3 Selectanunsamplednodeusingarandompathand
calculatetheindicator values at theselectednode
byordinarykriging.
4 Calculate the CDF (Cumulative probability Dis-
tribution Function) using three thresholds and
samplingfromtheCDF.
5 Includethecalculatedvalueasconditioningpoints.
6 Gobacktorandompathselectionuntil all unknown
nodesarecalculated.
Figure1. Boreholeandseismicvelocitydataintheresearch
area.
3 EXAMPLESANDRESULTS
3.1 Estimation of three-dimensional RMR
Estimation of three-dimensional RMR distribution
wasperformedinthehighway project fromSosa to
Noksan. Itwas1700minlengthfromSTA 3k600to
STA 5k300, withadepthfrom40to200m.Thegrid
was86zonesinlength(x-direction), 16zonesinwidth
(y-direction) and25zonesinheight(z-direction). The
dimensionsof oneelement were20minlength, 20m
inwidthand10minheight. RMRestimationwasper-
formed by borehole logging data and seismic data.
Boreholelocationandseismicdatawerepresentedin
Figure1(a) andFigure1(b), respectively.
Theresults of ordinary kriging, indicator kriging,
andSISareshowninFigure2. Ordinarykrigingused
a borehole logging data as input data, and indicator
krigingusedbothboreholeloggingdataandseismic
survey data. Figure2(c) shows thefirst result of 100
SISresults. TheRMRdistributionaroundtheplanned
tunnel ispresentedinFigure2(d). Koreantunnel sup-
portingsystemwascomposedof sixtypesbasedonthe
RMR. Fivegradesof RMR arematchedwiththefive
typesof supportsystemandsixthsupportsystemisfor
theportalsof atunnel. Therefore, themost important
issueinthedesignof tunnel support systemcanbea
determinationof reliableRMR values.
3.2 Reliability analysis of estimated RMR
Split-sample validation was performed to verify the
accuracy of theGA (GeneticAlgorithm) simulation.
A subset that was composedof 100datapoints from
782
Figure2. EstimationresultsandRMR distributionabound
thetunnel.
theoriginal borehole-loggingdatawassetasideastest
data; thereminderwastrainingdata.Andtheresultsof
split-samplevalidationarepresentedinFigure3. The
perfectlyestimatedresultispresentedasastraightline
inclined at 45
. Theresult of split-samplevalidation
showsdotslocatedaroundtheperfect estimate line.
Thedots intheupper andlower parts of thelineare
approximately random, andtheir numbersarealmost
identical.
Figure3. Theresultsof split-samplevalidation.
The coefficient of variation was 0.482 and 0.342
for ordinaryandindicator kriging, respectively. These
valuesarevery sensitivetolocal area. Withthesame
boreholeinputdata, SISwasperformedto100equally
probable times. Through these analyses of results,
RMR couldbepresentedintheformof aprobability
distributionfunction,andtheuncertaintyof estimation
couldbesuccessfullyquantified.
AsshowninFigure3(a) andFigure3(b), theorig-
inal input RMR ranged from15 to 95, whereas the
output RMR ranged from45 to 80 by both ordinary
and indicator kriging. Distribution features of orig-
inal geological data were disappeared by kriging in
theprocessof minimizingtheerror variation, andthis
phenomenoniscalledassmoothingeffect.Thecoef-
ficient of variationfor averagedSIS result was0.656
asshowninFigure3(c).
783
Figure4. Boxchartandprobabilitydistributionfunctionof
RMR distributionintheplannedtunnel areausingSIS.
Table1. Variationof RMR intheplannedtunnel area.
Station Mean Standarddeviation
3k700 44 16.3
3k940 62 18.7
4k210 72 13.4
4k450 76 12.5
4k690 75 6.5
4k960 62 17.6
5k300 52 21.1
Inorder toinvestigatethereliabilityof theestima-
tion, an equally probable simulation was performed
100 times in order to quantify theestimation uncer-
tainty. The RMR distribution around the planned
tunnel ispresentedinFigure4.Theleftdotsanddistri-
butioncurverepresent theRMR realizations, andthe
right box chart presents their normal distribution. In
thefigure, X, thevertical line, thediamondshapebox
chart, andthecenter dot represent themaximumand
minimumvaluesof 1%and99%respectively, thebox
rangefrom25%and75%of theCDF, andthemean
value, respectively. Themeanandstandarddeviation
of the 100 simulations are presented in Table 1. As
thereliabilityof theestimationincreases, thevariance
decreases.
4 CONCLUSIONS
The objective of this study was to estimate reliable
RMR values and correctly design a tunnel support
system. Theresultsmaybesummarizedasfollows:
1 Kriging and sequential indicator simulation have
its special characteristics andsequential indicator
simulationcouldestimateRMR effectively.
2 Estimation values could be shown in the form
of a probability using the 100 stochastic simu-
lations that were simulated on the condition of
equi-probability. Theestimationuncertainty could
bequantifiedbyavarianceof RMR.
3 Reliabilityanalysiswasperformedbysplit-sample
validation.Thedifferencesbetweentruevaluesand
estimationvaluescouldcheck theprecisionof the
estimation.
REFERENCES
Facchinelli, A. Sacchi, E. & Mallen, L. 2001. Multivari-
atestatistical andGIS-basedapproachto identify heavy
metal sources in soils. Environmental Pollution 114(3):
313324.
Feng, Y. Tang, S. & Li, Z. 2006. Application of improved
sequential indicator simulation to spatial distribution
of forest type. Forest Ecology and Management 222:
391398.
J uang, K. Chen,Y. &Lee, D. 2003. Usingsequential indicator
simulationtoassesstheuncertaintyof delineatingheavy-
metal contaminated soils. Environmental Pollution 127:
229238.
Marinoni, O. 2003. Improving geological models using a
combinedordinaryindicatorkrigingapproach.Engineer-
ing Geology 69(12): 3745.
Pardo-Igzquiza, E. & Dowd, P.A. 2005. Multipleindicator
cokrigingwithapplicationtooptimal samplingfor envi-
ronmental monitoring. Computers & Geosciences 31(1):
113.
Taboada, J. Vaamonde, A. Saavedra, A. &Alejano, L. 1997.
Application of geostatistical techniques to exploitation
planning in slate quarries. Engineering Geology 47(3):
269277.
784
Comparativestudyof softwaretoolsontheeffectsof surfaceloads
ontunnels
D.K. Koungelis& C.E. Augarde
School of Engineering, Durham University, Durham, UK
ABSTRACT: Inthispaper resultsarepresent fromparametricstudiesof twin-tunnellingschemescarriedout
using various finiteelement packages. Theaimis not to makecomparisons with field databut to show the
differences obtained using different FE models, as might be used in a design office. The effects of varying
surfaceloadingonthetunnelsthemselvesareinvestigatedandthechangingeffectsarestudiedastunnel layout
isaltered.
1 INTRODUCTION
Theongoingdevelopment intheworlds urbanareas
inevitably leads to the construction of structures in
closeproximitytoalreadydriventunnels. Careshould
be taken to ensure that construction is carried out
without damaging the tunnels or any other adjacent
or overlyinginfrastructure. Considerableresearchhas
beenundertakenfor thecaseof asingletunnel where
empirical methodsforpredictingtunnel induceddefor-
mationsareapplicable. Formorecomplexgeometries,
however, empirical methodsfail tomakeaccuratepre-
dictionssincetheydonot account for thesoil-tunnel-
structure interaction mechanism. The finite element
(FE) methodappearstobeasolutiontothisprediction
problem, however manydifficultiesinitsuseremain.
Theaimof this paper is to validatetools for numeri-
cal modellingof tunnellingrelatedinteractionsinsoft
ground.
Plenty of publications exist whichstudy theinter-
action mechanismof soil, lining and a pre-existing
structure(i.enon-greenfieldsite) responseduringtun-
nellingoperations. Howevertheliteratureontheeffect
of surfaceloading on an existing tunnel is sparseto
theauthors knowledge. Most refer tothecaseof sur-
faceloadingabovepipes, or pileconstructionandpile
loading, andtheir effectsontunnels. Themainreason
preventingengineersfromdealingwiththesubject of
surfaceloadingisthedifficultytheyfaceinaccurately
measuringthechangeof stressesactingonthelining
duetotheappliedload. Thelack of fielddataresults
inperformingmerelytheoretical analysis.
Peck (1969) states three conditions for success-
ful tunnelling. The first refers to safe operation of
tunnellingworks. Thesecondrequirement isthepro-
tection of adjacent structures. The final condition
refers to thetunnels ability to withstandall external
loads whichact uponit duringits servicelife. These
loadsandtheir influenceontunnel liningwill becon-
sidered here. According to ORourke (1984) linings
do not carry thetotal overburdenweight of theover-
lying ground. The vertical (
v
) and horizontal (
h
)
stressesinstead, arere-distributedaroundthefacedue
tomobilisationof thesoil shear strengthandcontinu-
ity. This effect is oftentermedarching. Thetunnel
thereforehastowithstandonlythestresseswhichare
not arched. Mair andTaylor (1997) presentedfield
datafrom12different tunnel casesdriveninLondon
clay. The lining load is expressed as a percentage
of full overburdenweight at tunnel centreline(CL).
Thedatacollection refers to points at oneweek and
oneyear after lining installation. Theseindicatethat
the measured lining load even after a year is below
70%of full overburden.Inmostcasesitvariesbetween
40%and60%.
Moore(1987) describedasemi-analytical solution
that makes use of the Boussinesq method and other
closedformsolutionstoestimatethedeformationof a
buriedpipe(rather thanatunnel) inaninfiniteelastic
mediumduetosurfaceloading. 2-D FE analysiswas
also employed. Provided that realistic elastic ground
properties are selected the semi-analytical method
compares well with the numerical results. This pro-
cedure can be used for estimations of hoop forces,
bendingmomentsandringdeformations.Clearlythere
areproblems however withtheassumptionof elastic
ground.
Thispaper investigatestheeffectof surfaceloading
onpre-existingtunnelsinsoftgroundassumingplane
strainconditions(2-DFEanalysis).Thefollowingtwo
commercial FE packages are used for this purpose,
Strand7andPlaxis. Thepurposeof thiscomparisonis
totryandidentifythedifferencesintheFEpredictions
byusingvariouscodes, whichmight besignificant to
785
industrial usersof theseprograms.Thereasonsfor not
usinganyanalytical method(e.g. Boussinesqmethod)
to estimate the tunnel deformations due to surface
loading werethat thesemethods areonly applicable
to elastic medium. They do not takeinto account the
propertiesof themediumortheinteractionbetweenthe
mediumandanypre-existingstructureinit.According
toMoore(1987) theBoussinesqmethodfailstotake
into account the effect of shear stresses and strains
developingintheoverlyingstrata.
2 ANALYSISGEOMETRY
Inthecurrentplanestrainanalysesthreedifferenttun-
nel geometric configurations are considered. In the
first case a single tunnel analysis is carried out (ST
case). Inthesecondatwintunnel analysis is carried
out, wherebothtunnels arehorizontally aligned(TH
case). Finally inthethirdcasetwintunnelsareverti-
callyanddiagonallyaligned(TVDcase).A parametric
study was performedfor theabovethreecases vary-
ing the position of the tunnel axis (z
o
), pillar width
(P), pillar depth (P
D
) as well as the position of the
surfaceloadedarea(W). Figure1showstheparame-
ters variedinthis study. For theST case(shownas a
solidcircle) z
o
varies. For thetwintunnel configura-
tion(wherethesecondtunnel ispresentedasadotted
circle)P andP
D
vary. Inall casestheloadedareashifts
fromW1toW6.Throughoutthisparametricstudythe
dimensionsof thedomain(x, y), tunnel diameter (D),
magnitude (400kN/m) and area (W) of the applied
loadwereconstant. Surfaceloadwasapplieddirectly
tothesurfaceof thefiniteelementshencemodelling
a flexible footing. No interface elements were used
to model theexistenceof any typeof foundations or
treatment of thegroundprior toitsloading.
Tunnel diameter (D) was chosen to be4mwhich
is comparableto thediameter of runningtunnels for
the Underground in London (Attewell, 1978). The
dimensions of the modelled domain were chosen to
be70mlong (or 17.5D) in thex direction and 50m
deep(12.5D)inthey direction.Forz
o
=15mand20m
thechosenvalues liewithinlimits proposedby Potts
et al. (2002). They suggestedthat for tunnels inclay
the depth of the mesh should be approximately 2D
to 3D below tunnel invert. As for theoptimal width
of thedomaintwofactorshavetobeconsidered. The
meshhastobesufficientlywidetoensureminimal dis-
placements along the vertical boundaries. However,
the larger the domain the larger the number of the
degrees of freedom(d.o.f). This immediately affects
thesolutionintermsof computational time. Hencea
compromisehastobemadebetweenthesetwocrucial
factors. Itwasdecidedthattheabovedimensionswere
appropriatefor thisstudy.
Thesurfaceloadisconstantat400kN/m.Thismag-
nitudewaschosentoresembletheuniformstressfrom
Figure1. Geometricparametersof thedomain.
a10-storey building, assuming astress of 10kN/m
2
per storey for a 4m wide loaded area. The latter
valueoverafull buildingwidthisprobablyunrealistic.
However, it was chosen as a worst case value (per-
hapsincludingtheeffectof anaccidental concentrated
load) to accentuatethedifferences in theparametric
study. Thevalueof 10kN/m
2
per storey was chosen
after BS 8002 (British Standards Institution, 1994)
recommendations.
ThemeshinStrand7iscreatedandimportedfrom
Gmsh(afreewareFE meshgenerator) sinceStrand7
can neither producean unstructuredmesh nor can it
beasflexibleasGmshinthepre-processingstageof
theanalyses. Six-nodedtriangular elements areused
to model thesoil, andtwo-nodedbeamelements are
used to model the tunnel lining. In Plaxis the auto-
generated mesh consists of fifteen-noded triangular
elements. Curvedbeamelementsareusedtomodel the
tunnel lining.Thereasonforusingdifferenttypeof ele-
ments (fifteen-noded triangles rather than six-noded
triangles) compared to Strand7 is to achievegreater
accuracy. InPlaxistheusercannotimportameshfrom
another softwarepackage.Thusitwasentirelycreated
inPlaxisspre-processingstage.Asaresultdifferences
intherefinement areevident (Fig. 2).
Therealisticdeterminationof theinitial stresscon-
ditions is of great importance in FE modelling in
geotechnics. Several approaches exist for this pur-
pose. Themost commonof whichistheK
0
procedure
wherestressespriortoanyconstructionareinitialised.
This methodis only applicablefor horizontal ground
surfaces and greenfield sites. This is not the case
in this study. Consequently a different approach is
adoptedtosimulateinitial groundconditions. Tunnel
excavationisnot modelled. Insteadtunnelswiththeir
permanent lining appear in themesh as if wished in
place.Then,gravityloadingisuniformlyappliedtothe
wholedomain(gravity loading method).Theresulting
786
Figure 2. Generated mesh for the TVD case when
z
o,1
=15mfor the upper tunnel, z
o,2
=20mfor the lower
tunnel and P =1D. Theupper figurerefers to theStrand7
meshwhilethelower tothePlaxismesh.
displacements arethen set as thezero datumfor the
subsequent steps of theanalysis. Thestratigraphy is
thesamethroughouttheanalysis. Thisconsistsof one
claylayer, thecharacteristicsof whicharepresentedin
Table1. Undrainedanalysesareperformedthroughout
usingeffectivestiffnessparameters.Asfor thebound-
ary conditions, inplanestrainanalysis no horizontal
or vertical movements are permitted along the hori-
zontal boundary at thebaseof themesh. Onthetwo
vertical meshboundaries, onlyvertical movementsare
allowed. Thetopmeshboundaryisfreetomove.
The beamelements used to model the lining are
assumed to behave in a simple linear elastic way.
Thus two parameters (Youngs modulus E and Pois-
sons ratio v) are required for this model. Table 2
shows the full characteristics of the lining, includ-
ing thegeometrical properties. Soil does not behave
in a linear nor an elastic way. Thus a more realistic
and advanced constitutivemodel should beadopted.
A simpleelasto-plasticconstitutivemodel istherefore
used. For theplastic regionMohr-Coulombyieldcri-
terion with associated flow is used in Strand7 while
non-associated flow is used in Plaxis. A valid criti-
cismhereis that this elasto-plastic model is still too
Table1. Material propertiesof thesoil.
Typeof behaviour: ElasticRegion, Linear Elastic
Typeof behaviour: YieldSurface, Mohr-Coulomb
Parameters Name Value Unit
Youngsmodulus E 6.20710
3
kPa
Poissonsratio v 0.33
Unit weight 20 kN/m
3
Cohesion c 5 kPa
Angleof friction 25

Table2. Material propertiesof thetunnel lining.
Typeof behaviour: Linear Elastic
Parameters Name Value Unit
Youngsmodulus E 10
8
kPa
CrossSectional Area A 0.168 m
2
/m
SecondMoment of Area I 3.9513610
4
m
4
/m
Unit weight
s
24 kN/m
3
Poissonsratio v
s
0.3
crudefor accurateanalysis of this problem, whichis
true. However theuseof thismodel matchesmuchof
the routine analysis carried out for tunnelling prob-
lems in design offices. The purpose of the paper is
not to providealink to fielddatabut to demonstrate
thedifferent predictionswhichfollowfromtheuseof
different FE models.
In all calculations carried out the analysis proce-
durebeganwiththetunnelsdrivenandthepermanent
lininginstalled. Displacementsfromthisstagearenot
measured. Two load stages arethen defined. During
thefirstgravity load isappliedtothemesh. Inthefol-
lowingstagethesurfaceload(400kN/m) isvertically
appliedtothepre-definedsurfaceareas(W1toW6).
Thedisplacementsduetothefirst loadstagearecon-
sideredas thezero datum. Thus only thosepredicted
by the FE analysis due to the second stage (surface
loading) areexamined.
3 RESULTS
3.1 Strand7 FE predictions
Inthis sectionfindings arepresentedfor thecaseof
surfaceloadingabovepre-existingtwintunnelswhich
arehorizontally aligned(TH case). Figure3presents
the lining deformations (scaled up) of the first tun-
nel (Fig. 1) when z
o,1
=15mandP =1D dueto the
effect of surface loading only (dotted circles). This
is then compared to the original tunnel shape prior
to any loading stage(thick circle). Thefirst obvious
outcomeis that thewholetunnel seems to squat. In
other words there is an elongation of the horizontal
787
Figure3. Deformedshapeof thefirst tunnel (TH case) due
to the surface loading for six different loaded areas. The
tunnel axisisdrivenat z
o,1
=15mandP =1D.
Figure4. Changesof horizontal andvertical tunnel diame-
tersof thefirsttunnel (THcase) expressedasapercentageof
theinitial tunnel diameter duetothesurfaceloadingagainst
thepositionof theappliedloadwhenz
o,1
=15m.
diameter withasimultaneousdecreaseof thevertical.
Furthertothisobviousvertical translationasecondary
typeof movement seemstooccur coincidentally. The
deformedliningseemstoslightlyrotateanti-clockwise
opposingthepositionof theappliedsurfaceloadasthis
shiftstowardsW4. Similarpredictionswereidentified
for theST andtheTVD cases.
Figure4showsplotsof thechangeof tunnel diam-
eter (as a percentage of the initial tunnel diameter)
againsttherelativepositionof thesurfaceloadedarea
for the first tunnel (TH case). The thin lines refer
tothehorizontal tunnel diameter (alongspringlines).
The dotted lines on the other hand refer to the ver-
tical (crown to invert). An increaseof thehorizontal
diameter withasimultaneousdecreaseof thevertical
isobserved. Themaximumincreaseof thehorizontal
diameter aswell asthemaximumdecreaseof thever-
tical (0.8%of thetunnel diameter) occurs when the
surfaceloadisdirectly appliedabovefor z
o,1
=15m.
Thus, acleartrendforthemagnitudesof thesechanges
can be identified. These changes appear to fade as
the load is applied further away from the tunnels
centreline.
Figure5. Plots of thecrownsettlements duetoloadingof
thefirst tunnel intheTH caseasapercentageof theST case
for various surface loading areas and two different tunnel
depths.
thefirst tunnel (TVD case) asapercentageof theST casefor
varioussurfaceloadingareaswhenz
o,2
=20m.
For theTH casecrownsettlement predictionsfrom
thefirst tunnel areplotted(Fig. 5) asapercentageof
thesingletunnel caseagainstthepositionof thesurface
loadedarea(W) for twodifferent depths(z
o,1
=15m
and20m)andforthreedifferentpillarwidths(P =1D,
2D and3D). Thinlinesrefer totheshallowcasewhile
dottedtothedeeper. It canbeseenthat greater inter-
actionoccurs for theshallowtunnel casesincethere
is adifferenceof 5%to 15%comparedto thesingle
case. For thedeeper caseresults arealmost identical
tothesingle. Thisimplieslessor evennointeraction
betweenthetwoparallel driventunnels.
FortheTVDcasecrownsettlementpredictionsfrom
the first tunnel are plotted (Fig. 6) as percentage of
theST caseagainst thepositionof thesurfaceloaded
area(W) for different pillar widths (P =piggy-back,
0and1D) whenz
o,1
=15mandz
o,2
=20m. It canbe
observedthatasmall amount(lessthantheTHshallow
case) of interaction exist (none to 6%) compared to
theST results. Greaterinteractionappearsbetweenthe
tunnelswhentheyarecloselyspaced(P =piggy-back
and0) of approximately5%. Whenthelower tunnel is
788
Figure7. Changesof horizontal andvertical tunnel diame-
tersduetothesurfaceloadingexpressedasapercentageof
thetunnel diameter against thepositionof theappliedload
whenz
o,1
=15m.
drivenfurther away(P =1D) lessinteractionbetween
thetwoverticallyalignedtunnelsoccurs(noneto3%).
3.1.1 Comparison of the different cases in Strand7
In this section theFE predictions regarding thefirst
tunnel (in theTH and theTVD cases) arecompared
with the predictions regarding the single tunnel (ST
case) inordertoinvestigatetheinteractionmechanism
of soil-tunnel-structurein2-D. Theabovementioned
comparisonis madeinterms of liningdeformations.
TheFE predictionsregardingthefirst tunnel (TH and
TVD cases) are smaller than those regarding the ST
case.Thisisafirstindicationof theexistenceof inter-
action. In both of the compared cases the shape of
thetunnels seems tosquat whilethereductionof the
vertical tunnel diameter equals to theincreaseof the
horizontal. Themaximumliningdeformation occurs
when the surface load is applied at W2. No lining
deformation is predicted when theload is applied at
W6(Fig. 4) which indicates that at that distancethe
interactionceases.
Intotal it seems that interactionoccurs withinthe
regionof W1toW4(P 2D) andthenfor (P 3D) it
startstoreduce(Fig.4).Greaterinteractionispredicted
intheTHcasecomparedtotheTVDcase.Thisimplies
that theexistenceof thelower tunnel (TVD case) does
not contributetothecomplex interactionmechanism
inthesamewayasitdoesthesecondtunnel intheTH
case(Fig. 1).
3.2 Plaxis FE predictions
Figure7isakintoFigure4. Thisgraphindicatesthat
themaximumchangeof D (Roughly0.8%of thetun-
nel diameter. Thesepredictionsareinagreement with
Strand7) regarding thefirst tunnel (TH case) occurs
when theload is applied directly above. As theload
shiftsfurther awaythesechangesreducetowardszero
(no change of D). However when the surface load
is applied at its furthest possibledistance, theverti-
cal diameter is seen to increasewhilethehorizontal
reduces.
Figure8. Plotsof thecrownsettlementsduetoloadingof the
firsttunnel intheTHcaseasapercentageof theSTpredictions
for various surface loading areas and two different tunnel
depths.
thefirsttunnel intheTVDcaseasapercentageof theST case
for varioussurfaceloadingareaswhenz
o,2
=20m.
Figure 8 shows the interaction between the first
tunnel (TH case) and the surface load compared
to the single tunnel case (ST) for two different
depths(z
o,1
=15mand20m) andthreedifferent pil-
lar widths (P =1D, 2D and 3D) in terms of crown
settlements (similar to Fig. 5). Thinlines refer to the
shallowcasewhiledottedrefer to thedeeper case. It
canbeseenthatPlaxispredictionsfor theTH caseare
almost identical totheST caseresults (i.e. almost no
interactionbetweenthetwoparallel tunnels).Thispic-
tureisdifferenttotheStrand7resultsespeciallyforthe
shallowcase.
Figure9showstheinteractionbetweenthetwotun-
nelsandthesurfaceloadintheTVD casecomparedto
the(ST) caseintermsof crownsettlementsfor differ-
ent pillar widths (P =piggy-back, 0 and 1D) when
z
o,1
=15mand z
o,2
=20m(akin to Fig. 6). It can
beobservedthat asmall amount of interactionexists
(none to 5%) compared to the ST case predictions.
Resultsaresimilar comparedtotheTH casefromW1
789
toW3.AsthesurfaceloadshiftstowardsW4thougha
differenceinthepredictionof interactionisidentified.
3.2.1 Comparison of the different cases in Plaxis
FEpredictionsregardingthefirsttunnel (intheTHand
theTVDcases) arecomparedwiththoseintheST case.
Themaximumdeformationof thetunnel liningoccurs
whentheloadis appliedat W2forcingthetunnel to
squat. These deformations reduce as the load shifts
towardsW5.AtW6thoughtheloadseemstoproduce
an ovalisation of the lining with the vertical tunnel
diameter greater thanthehorizontal (incontrasttothe
previousloadcases). Ingeneral Plaxispredictssimilar
amountof interactionbetweenthetwodifferenttunnel
geometricconfigurations(TH andTVD).
4 DISCUSSION
Inthecurrentpaper2-DFEpredictionswerepresented
investigatingtheeffect of surfaceloadingabovepre-
existing tunnels driven in soft ground. Two different
FE packages wereused for this purposeto compare
andvalidatetheproducedresults. Several parametric
studieswerecarriedoutvaryingtheexcavationdepth,
thepillarwidth, thepillardepthandthepositionof the
surfaceloadedarea.Thegeneral trendbetweenthetwo
FE packages regardingtheliningdistortions andthe
crownsettlementswassimilar. Bothpackagespredict
thatwhenthesurfaceloadisappliedwithintheregion
of W1toW4[i.e. ahorizontal distanceof P 2Dfrom
thefirsttunnel (THcaseandTVDcases)] theexistence
of theinteraction mechanismwas evident regardless
of the tunnel geometric configuration. Further from
that distancenointeractionoccurred.
Small differences in the predictions between
Strand7andPlaxisoccurred. Strand7inparticularpre-
dictedtheexistenceof strongerinteractionmechanism
fortheTHcasecomparedtotheTVDcase.Plaxisonthe
other hand predicted asimilar amount of interaction
between thetwo different tunnel geometric configu-
rationsandsmaller comparedtoStrand7predictions.
Thesedifferencesareattributedtothefollowingthree
factors:
Different types of finite elements were used to
model soil inthedomain. Six-nodedtriangleswere
used in Strand7. Even though this type of finite
element was available in Plaxis as well, it was
decided that the fifteen-noded triangle should be
usedinsteadfor greater accuracy.
Different meshes weregeneratedbetweenthetwo
FEpackages.ThereasonwasthatinPlaxistheuser
cannot import ameshasinStrand7.
Finally, even though the same elasto-plastic soil
model with the Mohr-Coulomb failure criterion
was usedbetweenthetwo FE packages, theplas-
tic potential function was different. Associated
flow was used in Strand7 whilePlaxis used non-
associatedflow.AccordingtoPottsandZdravkovic
(1999) the latter way of modelling real soil
behaviour ismorerealistic thanthefirst, although
giventheloadinginthisproblemit isperhapsnot
significant.
Fromtheaboveit isclear that differingpredictions
can befound fromroutineuseof two FE modelling
packages. Greater detail of thisandanextendedstudy
of this problemcan be found in Koungelis (2007).
Suchdifferencesinpredictionsmaynot beimportant
but without undertakingsuchstudiesthat fact will be
difficult toassert inmost cases.
ACKNOWLEDGEMENTS
This work was fundedby astudentshipfromtheUK
Engineering and Physical Sciences Research Coun-
cil (EPSRC), the School of Engineering at Durham
UniversityandHalcrowLtd.
REFERENCES
Attewell, P.B. 1978. Groundmovementscausedbytunnelling
insoil.In: Procs. Of Conf. on large ground movements and
structures, UWIST/July 1977: 812948.
BritishStandardsInstitution1994.BS8002:Code of Practice
for Earth retaining Structures.
Koungelis,D.K.andAugarde,C.E.2004.Interactionbetween
multiple tunnels in soft ground. In: Developments in
Mechanics of Structures and Materials: procs. 18th Aus-
tralasian Conference, Perth, Australia. Deeks, and Hao,
(eds), London: Taylor andFrancis. Vol. 2: 10311036.
Koungelis, D.K. 2007. Tools for numerical modelling of
tunnelinginteractions. PhDThesis, DurhamUniversity.
Mair, R.J. and Taylor, R.N. 1997. Bored tunnelling in the
urbanenvironment. In: Procs 14th Int. Conf. Soil Mech.
And Found. Engng., Balkema, 1997: 23532385.
Moore, I.D. 1987. Responseof buried cylinders to surface
loads. J. Geotech. Engng. 113(7): 758773.
Moore, I.D. and Brachman, R.W. 1994. Three-dimensional
analysis of flexiblecircular culverts. J. Geotech. Engng.
120(10): 18291844.
ORourke, T.D. 1984. Guidelines for tunnel lining design.
PublishedbyASCE,Technical committeeontunnel lining
designof theUndergroundTechnologyResearchCouncil.
ground. In: 7th Int. Conf. Soil Mech. and Found. Engng,
Mexico, 1969: 225290.
Potts, D.M., Axelsson, K., Grande, L., Schweiger, H. and
Long,M.eds.2002.Guidelines for the use of the advanced
numerical analysis. London: ThomasTelford.
Potts, D.M. andZdravkovic, L. 1999. Finiteelement analy-
sisingeotechnical engineering: Theory. London: Thomas
Telford.
Spasojevic, A.D., Mair, R.J. and Gumbel J.E. 2007. Cen-
trifugemodellingof theeffectsof soil loadingonflexible
sewer liners. Gotechnique 57(4): 331341.
790
GeologicModel TransformingMethod(GMTM) for numerical analysis
modelingingeotechnical engineering
X.X. Li & H.H. Zhu
Shanghai, P. R. China
Department of Geotechnical Engineering, School of Civil Engineering, Tongji University, Shanghai, P. R. China
Y.L. Lin
Geotechnical Research Institute, Hohai University, Nanjing, P. R. China
ABSTRACT: Duringnumerical simulationof geotechnical engineering, duetothecomplexengineeringgeo-
logical condition, thesimplified geologic model is usually adopted. However, theaccuracy and reliability of
calculatingresultaredirectlyinfluencedbycharacteristicof model.The3Dgeologicmodel canbetterreflectthe
actual geological condition. Thepaper presentsanewandpractical modelingmethod(Geologic Model Trans-
forming Method) of numerical analysis by integrating geologic model and numerical model. This method is
performedwiththefollowingprocedures: (1) cuttingthe3Dgeologicmodel accordingtonumerical calculation
region; (2) extractingcontrol datafromcuttingmodel toreconstructthesurfacemodel; (3) meshingandforming
thenumerical model automatically by thestratumattribute; (4) importing themodel into numerical analysis
system. An example is given to illustrate the application of the method. The implementation of the method
resultsinhighefficiencyandautomaticityof modeling.
1 INTRODUCTION
With the development of the computer technology,
finite element method (FEM) is widely used in
geotechnical engineeringandother researchdomains.
During the solution of FEM, preprocessing, solving
andthepost-processingareincluded. Thepreproces-
sor, which involves collecting data, inputting infor-
mation, meshinganddefiningmaterial property, etc.,
is thefussiest process and spends 4050 percent of
thetotal timeintheanalysis courseof FEM. That is,
about half effort of FEM analysisisexpendedonthe
preprocessor (Liuet al. 2002). Therefore, preprocess-
ingsystemisoneof thecorepartsof thefiniteelement
analysis andagoodpreprocessor plays animportant
roleinthesuccessof themethod.
For apractical problemof geotechnical engineer-
ing, numerousdataisdemandedduringthemodeling
process becauseof complex geological environment,
structure, construction procedure and so on. If the
fussydataisinputtedbymanual handling, it isineffi-
cient, andisdifficult tocheckandmodifydata. Thus,
the simplified geologic model is generally adopted.
However, a poor model may give unreliable results
that areusually misleadingandmay leadtoincorrect
conclusions. Therefore, it isnecessary todevelopthe
researchonasimplifiedgoodFEM preprocessor, and
theproblemhas received alot of interests (Yu et al.
1999, Zhouet al. 2002, Lianget al. 2004).
Theinformationvisualizationof engineeringgeol-
ogyisanimportantresearchsubjectingeosciences.At
present, thegeologysimulationhasthepowerful func-
tionof 3D geological modeling. 3D stratamodel can
beestablishedbyboreholedata, andpartspatial analy-
siscanalsobeoperated,suchassectioningandcutting,
etc. Therearethecertainsimilaritiesinmodelingand
spatial analysisbetweengeologic model andnumeri-
cal model. If geologicmodel canbetransformedinto
numerical analysismodel, itsdatacanalsobedirectly
introduced. Thus, thepreprocessor of FEM islargely
simplified(Houet al. 2002, Xiaet al. 2005).
In full consideration of the characteristics of the
geologic model and thenumerical model, this paper
presentsanewmodelingmethodof numerical analy-
sis calledtheGeologic Model TransformingMethod
(GMTM).ThismethodissuitableforFEMpreproces-
sor ingeotechnical engineeringandsimpler andmore
efficient than the traditional operations approach.
The GMTM can be used to form finite element
meshdirectlyfrom3Dgeologicmodel. Theproposed
technology will give some useful references to the
continuousresearchonthissubject.
791
Theremainderof thispaperisorganizedasfollows:
inthenextsectionthe3Dgeologicmodel isdescribed.
Section3introducestheimplementationandthealgo-
rithmof thenewmethod. Section4givesanexample
toillustratetheresultsof thismethod.
2 3DGEOLOGIC MODEL
2.1 Choosing data model
At present, there are more than twenty data models
presented for 3D geosciences modeling. According
to theformof datastructure, thesegeologic models
presented or discussed could beclassified into three
typesasfacial models, volumetric modelsandmixed
models. Thefacial model emphasizes onthesurface
representationfor terrain, stratainterface, outlinesof
the constructions, buildings and underground engi-
neeering. Thetriangulatedirregular network (TIN) is
widelyappliedfor thesurfacemodeling(Baker 1989,
Huang et al. 2002). The volumetric model is based
on the spatial partition and the real 3D object con-
struction.Thethreedimensionspacecanbefilledwith
regularorirregular3Dvolumes, andtherepresentative
volumeisirregular tetrahedronnetwork(TEN) (Chen
etal. 1994, Pilouketal. 1994).Themixedmodel isthe
combinationof facial model andvolumetricmodel.
Thesmall storagespaceof dataanditsfast visual-
izationarethecharacteristic of thefacial model, and
theinformationof stratainterfaceandfault couldbe
representedintheinterior of geologicsurfacemodel.
Thevolumetric model is convenient for thedescrip-
tionof theattributesof eachvolumeandthestorageof
itsrelatedspatial location. However, duetotheenor-
mous dataamount of thesolidmodel, theefficiency
of Booleanoperationto stratais lower. Althoughthe
mixedmodel takes theadvantageof facial model for
fast visualization and volumetric model for spatial
analysis in theory, the modeling technology is quite
complicated. By contrast, the paper chooses the 3D
geologicsurfacemodel asbasicmodel.
2.2 3D strata surface modeling
The paper is concerned with the 3D geologic sur-
facemodel basedonTIN representation(Zhanget al.
2006).Tothegeological modeling,thepractical imple-
mentationjust requires:
1. Transaction of borehole data. The information of
drill hole is extracted fromdatabase. Then every
borehole is numbered with the dichotomic topo-
logicdatastructure, andisblockedwiththeblock-
ingrulesof data. Thus, thestratumintegritycanbe
openedout.
2. Constructionof thehollowsurfacemodel. Theref-
erenceTIN isconstructedwiththesurfacedataof
theborehole. Intermof thetopological relationsof
theformedTIN, eachstratuminterfaceisobtained
bymappingthealtitudeandattributeof underlayer
borehole. For the phantom, the modeling can be
established by the sliced interface rules and the
interpolation of thespaceattribute. So thewhole
hollowsurfacemodel isfinished.
3. Construction of the surrounding surface model.
With the nearest neighbor first algorithm and
Delaunay methodto thescatteredpoints inexter-
nal outlineof everyinterface,thestratasurrounding
surfacemodel isconstructed. Finally, the3Dstrata
surfacemodel iscomposedof thehollowinterface
model andthesurroundingmodel.
3 TRANSFORM GEOLOGIC MODEL INTO
NUMERICAL ANALYSISMODEL
The finite element mesh needs to coincide with the
consistency principle and the geometrical property
principle. In consideration of the difference in data
structure, thegeologicmodel cannotbedirectlytaken
asthenumerical model (Wanget al. 2004). Tosatisfy
theregulationof thefiniteelementmesh,thenewmod-
elingmethodpresentedinthispaper emphasisonthe
processof transforminggeologicmodel intonumerical
model.
3.1 Flow process
The realized procedures from geologic model to
numerical model arelistedinthefollowing:
1. Calculationregioncutting(section3.2). Basedon
the3Dstratasurfacemodel established, theregion
cutting is performed along the partial research
region in conformity to the request of numerical
analysis.
2. Reconstructionof surfacemodel (section3.3). By
virtue of the intersection points and the strata
control points extracted fromthe cutting model,
the strata surface model of the calculate region
is reconstructed. Thus, theinitial surfacemesh is
generated.
3. FEM surface meshing (section 3.4.1). Evaluate
whetherthequalityof theinitial surfacemeshissat-
isfiedwiththerequestof thefiniteelementmeshor
not. If thejudgment istrue, theinitial meshcanbe
actedas thecalculationmesh. Otherwise, thesur-
facemeshneedstoberegeneratedautomaticallyby
anewapproach.
4. FEM solid meshing (section 3.4.2). Based on the
generated surface mesh which satisfied with the
FEM request, therelevant solidmeshisgenerated
automaticallybyanefficient approach.
5. Importingintonumerical analysissystem(section
3.5). Thetopological dataof theFEM meshmodel
is imported into thenumerical analysis systemin
accordancewiththerequesteddataformat. Inthe
end, themodelingworkisfinishedgradually.
792
Figure1. Flowchart of model transformation.
In theGeologic Model Transforming Method, the
key technologies includethecutting, thereconstruc-
tionandthemeshing.Figure1showsaflowchartof the
transformingprocess.Thefollowingsectionsdescribe
thekeystepsindetail.
3.2 Calculation region cutting
Accordingtotheengineeringgeological conditionand
thesurvey data, the3Dstratasurfacemodel isestab-
lishedwiththegeological modelingtechnology. Ithas
theadvantages of better accuracy for strataattribute
andlarger overlyingregion. Toapractical calculation
model, the partial research region just is the region
of numerical model. Theconditionswhichdetermine
the calculation region usually include removing the
boundary effect and satisfying the calculation accu-
racy. Soit isunnecessarytotaketheoverlyingregion
of the geologic model as the research region of the
calculationmodel. Toreducethecomputingtimeand
scaleof thenumerical analysis, theregion cutting is
performedinthemethod.Bythemeansthatchoosethe
calculationregionof thenumerical analysis, therea-
sonablecuttingboundarycouldalsobedetermined, as
showninFigure2.
To benefit thecalculation, thecuttingregiongen-
erally is a rectangular solid. Of course, it can also
be composed with many planes or polygons. The
algorithmof the region cutting is a procedure that
Figure2. Calculationregioncutting.
the intersection lines and the intersection points are
obtainedwiththeintersectioncalculationof thespace
planes each other. By the region cutting, the partial
researchregioncanbearbitrarily acquiredonthe3D
geologicmodel.Thus,itisunnecessarytobuildthecal-
culationmodel againandagain. At thesametime, the
constructionprocessingeotechnical engineeringcan
besimulated, suchastheexcavationof thefoundation
andthetunnel.
3.3 Reconstruction of surface model
Thetopological relationsarechangedwhentheregion
cutting is performed on the original strata surface
model. In order to keep the same topological struc-
ture, thesurfacecuttingmodel isreconstructedbythe
intersectionlines, theintersectionpointsandtheorigi-
nal stratacontrol points. Thereconstructionalgorithm
includes two sections: the reconstruction of strata
interfaceandthereconstructionof stratasurrounding
surface.
Reconstruction of strata interface:
1. Theintersection lines, theintersection points and
the original strata control points in the cutting
region are classified and restored by the strata
attribute.
2. To each stratum, the intersection lines are taken
as the constraint boundary, and then the con-
straint Delaunaytriangularizationisappliedtothe
scatteredpointsintheplaneregion.
3. Thetopological structureof theformedmeshiskept
invariability. Thealtitudeof eachpoint ismapped.
Thus, thestratuminterfacemodel isobtainedagain.
4. Repeat step (2) and (3). Each stratum interface
mesh could be reconstructed. Finally, the whole
interfacemodelingisfinished.
Theillustration of thereconstruction to thestrata
interfaceisshowninFigure3.
Reconstruction of strata surrounding surface:
5. Thesurroundingoutlineloopof thecuttingmodel
is constructed by the intersection points in the
interfaceandisorientedcounterclockwise.
6. Toeachcuttingplane, theintersectionlinesof the
planearetakenastheinterior constraint lines, and
793
Figure3. Stratainterfacereconstruction.
Figure4. Surroundingsurfacestratareconstruction.
then the constraint Delaunay train-gularization is
applied to the intersection points of the cutting
plane.
7. The attribute of the formed mesh in the cutting
planes is determined with the attribute of the
intersectionpoints.
8. Repeat step (6) and (7). The surface mesh of
the closed surrounding cutting planes could be
generated. Finally, thewholesurroundingsurface
modelingisfinished.
Theillustration of thereconstruction to thestrata
surroundingsurfaceisshowninFigure4.
The surface model of the cutting region is the
combinationof thestratainterfacemodel andthesur-
rounding surface model. The cutting model has the
samegeometricrelationandthetopological datastruc-
ture. Itsmeshisalsotakenastheinitial surfacemesh
withthestrataattribute.
3.4 Finite element meshing
At present, thealgorithmsof thefiniteelement mesh
generationarerather sophisticatedandefficient. Most
relevantresearchesarefocusedonthealgorithmitself
(George1991; Shephardet al. 1991; Lauet al. 1996).
However, there are many preconditions of the mesh
Figure5. MultipleTINRegions(MTR) method.
generation need to be judged by the man-machine
interactionwayduringtheprocessof numerical mod-
eling, suchasthepreconditionswhether themeshing
regionisclosedor notandwhether thetotal redundant
edge length is zero or not, etc. Thus, the automatic
meshing is often not ableto beperformed normally.
Thereasonof theproblemdoesnotlieinthealgorithms
butinthepreconditions. Thesefactorscausethemod-
elingdifficultytobeenlargedandtheautomaticityto
bereduced. Therefore, it isnecessarytoresearchhow
toimprovethewholemodelingefficiencyentirely.
3.4.1 Surface meshing
Any single irregular triangle is closed. Such the tri-
angleregion is called theSingleTIN Region (STR)
inthepaper. Theset of many STR whichdonot pass
througheachother iscalledtheMultipleTIN Region
(MTR). Thesurfacemodel basedonTIN representa-
tionistakenastheMTR model. Sotheinitial surface
mesh of the cutting model also is the MTR model.
Byperformingcircularlythealgorithmof thesurface
meshing in 3D space plane to each STR, the finite
element surfacemesh of MTR model could begen-
erated automatically, as shown in Figure 5. In order
tocoincidewithtopological consistencyof thenodes
insharededge, theregionboundaries arediscretized
with fixed length partition or integer division way.
Themethodof thesurfacemeshingiscalledtheMTR
method.
Based on the MTR method, some man-machine
interactionoperationscanbeavoided, suchasseeking
manuallytheclosedregionanddiscretizingthebound-
ariesoneby one. TheMTR methodisrealizedeasily
by programming, and has the advantages of better
element qualityandhighefficiencyandautomaticity.
3.4.2 Solid meshing
A singleclosed3D spacecomposedof theFEM sur-
facemesh is called theSingleSpaceRegion (SSR).
Theset of manySSR whichdonot passthrougheach
other is called the Multiple Space Region (MSR).
Thealgorithmof thesolid meshing needs to beper-
formed in SSR. Generally, there are many different
spaceregionsinacalculationmodel.Thus,almosteach
SSRneedtobeformedwithman-machineinteraction.
It isveryinconvenient tothenumerical modeling.
794
Figure6. MultipleSpaceRegions(MSR) method.
So, under the direction of the block thought that
the surface mesh having the same stratumattribute
can constitute the closed space region, a method is
presentedtobuildautomaticallytheSSRof thediffer-
entstratum. Byperformingcircularlythealgorithmof
solid meshing to each SSR, the finite element solid
mesh could begenerated automatically, as shown in
Figure6. Inconsiderationof thedual attributesof the
interfacemesh, it needstobeusedtwotimesinform-
ingstrataspaceregion. ThemethodiscalledtheMSR
method.
The thought of the blocking by strata attribute is
applied in the MSR method. The construction for
theSSR andthegeometry inspectioncanbeavoided
beforethesolidmeshing. Theefficiencyof thewhole
modelingis improvedthoroughly. It is aninnovation
totraditional modelingmethod.
3.5 Importing into numerical analysis system
Thetopological dataof thetakenFEMsurfacemeshis
importedintothenumerical analysissysteminaccor-
dancewiththerequesteddataformat. Thenumerical
analysis model is formed gradually and the whole
modelingisfinished.Thefiniteelementmodel gener-
atedwiththeGMTMisnotdependedonthenumerical
analysis system. Only changing into the demanded
data format, it can be used in the different systems
of theuniversal FEM software, such asANSYS and
Marc.
4 EXAMPLE
Toillustratetheapplicationof theGMTM, anexam-
pleispresentedinthissection. Thedetailedmodeling
proceduresarethefollowing.
1. The information of boreholes is extracted from
database.Thenumberof boreholesandstrataare16
and3, respectively. Thetotal number of thestrata
control pointsis64. Thegeologicsurfacemodel is
constructedbythealgorithmof thegeologicmod-
eling (seeFig.7). Thetotal number of theTIN is
Figure7. 3Dgeologicsurfacemodel.
Figure8. Cuttingregion.
Figure9. Topological relations.
144, andthedifferentstratumattributeisindicated
withdifferent colors.
2. The calculation region is determined by import-
ing the information of the cutting surface. The
paper assumes thecuttingsurfaceis composedof
four vertical planes. Theplanegraphof thecutting
regionisshowninFigure8,andthenewtopological
relationsareshowninFigure9.
3. Thecuttingsurfacemodel isreconstructedaccord-
ingtothestratacontrol pointsincuttingregion.The
stratainterfacemodel andtheinitial surfacemesh
areshowninFigure10.
4. Basedontheinitial surfacemesh,thesurfacemesh-
ing and the solid meshing are automatically per-
formedbytheMTR methodandtheMSR method
respectively.ThestratainterfaceFEMmeshandthe
solidmeshareshowninFigure11.
795
Figure 10. Surface model: (a) strata interface model;
(b) initial surfacemesh.
Figure11. FEM mesh: (a) stratainterfacemesh; (b) solid
mesh.
Figure12. Meshof theexcavationmodel: (a) wholemesh;
(b) part mesh.
5. Theexcavationprocessalsocanbesimulatedbythe
regioncutting. Inthis example, thecuttingregion
is takenas thefoundationpit. Thewholeandpart
FEM meshareshowninFigure12. Inthesurface
model, thetotal numberof thepointandthetriangle
element generatedare1140and3646respectively.
In the solid model, the total number of the point
andthetetrahedronelementgeneratedare1821and
8838, respectively.Thetotal timespentinthewhole
modelingprocessisabout 10minutes.
Figure 13. Stress distribution: (a)before excavation;
(b) after excavation.
6. The topological data of the formed mesh is
importedintotheFEM softwareMarc. Inorder to
check theelement meshquality, asimplenumeri-
cal calculationisimplemented. Figure13showsthe
self-weightstressesof themodel beforeexcavation
andafter excavationintheactionof gravity.
5 CONCLUSIONS
Inorder tosimplifythepreprocessor of thenumerical
analysis, thepaper presents anewmodeling method
basedonthegeologicmodel. It differsfromthetradi-
tional modelingmethod. Inour method, thenumerical
model isgeneratedby thefollowingprocedures: cut-
tingthecalculationregion, reconstructingthesurface
model, finiteelement automatic meshing, importing
theanalysissystem. Thecharacteristic of themethod
isgeneralizedas:
1. The element mesh is marked by the stratum
attribute, so the model further nears to the real
geological condition.Thepartial regioncanberan-
domlyextractedfromthe3Dstratamodel,andmust
not beestablishedrepeatedly.
2. Taking the advantage of the characteristic of the
geologicmodel, thecomplexmeshingalgorithmis
avoided. Theprogrammingof themeshingissim-
ple and efficient. The whole modeling is highly
automaticandrapid.
3. The excavation of the foundation and the tunnel
can be conveniently simulated. The boundary of
theexcavationmust not beset inadvance.
4. Thefiniteelement meshcanbeimportedinto the
differentanalysissystemtocalculate.Themodel is
applicableandflexible.
5. Inaword, theGMTM realizesthedreamtoestab-
lishnumerical model directlybasedon3Dgeologic
model. It provides anicepossibility andprospect
for FEM numerical simulation. Thetransforming
method based on other geologic model is consis-
tentwiththemethodpresentedinthepaper, butthe
material algorithmneedstobestudiedfurther.
796
REFERENCES
Baker, T.J. 1989. Developments and trends in threedimen-
sional meshgeneration.Applied Numerical Mathematics.
Chen, X.Y. & Kozo, I. 1994. Three dimensional modeling
of GIS based on delaunay tetrahedral tessellations. In:
International Archives of Photogrammetry and Remote
Sensing, Munich, Germany, 30: 132139.
George, P.L. 1991. Automatic mesh Generation: Application
to Finite Element Methods. Witey, NewYork.
Hou, E.K, Wu, L.X. & Li, J.M. 2002. Studyonthecoupling
of 3D geoscience modeling with numerical simulation.
Journal of China coal Society, 27(4): 388392.
Huang, Y.Y. & Chen, S.Q. 2002. Geologic modeling by
marchingcube. Journal of Engineering Graphics, 23(2):
6569.
Lau, T.S. & Lo, S.H. 1996. Finiteelement meshgeneration
over analytical curvedsurface. Computers & Structures,
59(2): 301309.
Liang, S.W, YangX.H. &Yang, W.B. 2004. Systemintegra-
tionof ANYSIS preprocessor andAutoCAD. Journal of
Huazhong University of Science and Technology (Urban
Science Edition), 21(1): 8184.
Liu, L.M, Liu, H.L. & Zhu, Z.D. 2002. Integrationmethod
of geoengineeringfiniteelementanalysissystembasedon
GIS. Chinese Journal of Rock Mechanics and Engineer-
ing, 21(Supp.1): 19951998.
Pilouk, M, Klaus, T. & Martien, M. 1994. A tetrahedron-
based 3D vector data model for geoinformation.
In: Advanced Geographic Data Modeling, Nether-
lands Geodetic Commission, Publications on Geodesy,
1994(40): 129140.
Shephard, M.S. & Georges, M.K. 1991. Automatic three
dimensional mesh generation by the finite octree tech-
nique. International Journal for Numerical methods in
Engineering, 32: 709749.
Wang, C.X. & Bai, S.W. 2004. Study on integration of 3D
stratainformationsystemandFEM. Chinese Journal of
Rock Mechanics and Engineering, 23(21): 36953699.
Xia, Y.H, Bai, S.W. & Ni, C.S. 2005. Study on coupling
of 3D visualization with numerical simulation for pow-
erhouseexcavationof acertainhydrojunction. Rock and
Soil Mechanics, 26(6): 969972.
Yu, C, Zhou, X.H. & Zhang, Y.Q. 1999. Theintegration of
the feature-based modeling and finite element method.
Journal of computer-aided design and computer graphics.
Zhang, F, Zhu, H.H. & Ning, M.X. 2006. Modelingmethod
of 3D strata suitable for massive data. Chinese Jour-
nal of Rock Mechanics and Engineering, 25(Supp.1):
33063310.
Zhou, E.H, Zhu, Y.W. & Wang, T. 2002. Transitional-
zone method in hexahedral finite element meshing in
geotechnical engineering. Engineering Journal of Wuhan
University, 35(3): 2429.
797
Reviewandinterpretationof intersectionstabilityindeepunderground
basedonnumerical analysis
T.K. Lu, B.H. Guo, & L.C. Cheng
School of Energy Science and Engineering, Henan Polytechnic University, Jiaozu, Henan, P.R. China
J. Wang
Shenhua Ningxia Coal Group,Yinchuan, Ningxia, P.R. China
ABSTRACT: Thispaper conductedapreliminarystudyonthestabilityof roadwayintersectionindeepunder-
groundconditionsonthebasisof theundergroundobservationsandnumerical modeling. Firstly, accordingto
theintersectionsusedincurrent coal mines, fivedifferent geometrical shapesof two-dimensional intersections
areselectedandmodeledunder deepgroundcondition. Itisnotedthatthecross-intersectionisthemostunstable
twodimensional intersectioncurrently usedincoal mines. Also, thedepthof cover isoneof important factors
for thisinvestigation, whichclearly indicatedthat thestability of theintersectionisthreatenedseriously when
thedepth of cover reached at 1000mor morebelowthesurface. In addition, theconstruction sequenceand
direction for thecross-intersection is studied. It is believed that theconstruction sequenceand direction are
sensitivefactorsfor thestabilityof twodimensional cross-intersection. It issuggestedthat theformationof the
cross-intersectionshouldbeconductedonesidebyanother individually, eventhiskindof constructionsequence
will causetwicestress concentrationandredistribution, it is still less influenceonthestability of intersection
comparingwithother constructionmethodsselected.
1 INTRODUCTIOIN
Fromthestatistical data, itisnotedthattheroof failure
mainlyoccurredaroundthelargespanof underground
opening, andthesituationis gettingworth, whenthe
complexgeometrical structuresanddeepminingcon-
ditions are encountered. This paper is describes the
stability andfailurecharacteristics of intersectionon
the basis of literature review and study of the sta-
bility and failure behaviour of different geometrical
structure of intersection, which commonly used in
underground of coal minein China, associated with
variousdepthsof cover usingnumerical simulation.
Itisnotedthatthegeometrical shapesusedinChina
coal mining industry is much complex, comparing
with major coal mining countries in overseas, such
as Australia and USA. With increasing the depth of
cover, itissusceptibletogroundcontrol problemsdue
tothecomplexgeometriesof undergroundstructure.
2 PREVIOUSINVESTIGATIONON
UNDERGROUNDINTERSECTIONINCOAL
MINES
Stability of underground intersection had been paid
attention previously, even in the shallow condition.
The investigations had been conducted to study
the structural characteristics, stability and failure
bahaviour in the shallow condition by former US
Bureauof Mine(Hanna, 1991).
In1976, Balachandrastudiedthestabilityof inter-
sectionusingundergroundmonitoringandcomputer
simulation.
Thedepartment of miningengineeringfromWest
VirginiaUniversity studiedthestability of three-way
intersection with different geometrical parameters,
anditfoundthatthetensilestressisthemainreasonto
causethefailureof theintersection.Comparatively,the
three-wayintersectionismorelikelyhavingshearfail-
urethanthecrossintersection. Itisconcludedthatthe
tensileand shear failureon theroof strataincreased
with the reducing of angle between two roadways
(Peng, 1978).
Thedepartment of mining engineering fromUni-
versityof Wollongongalsostudiedthestabilityof the
T-intersection(Singh, 2001). Basedonthenumerical
modeling and underground monitoring, it indicated
that the depth of cover, horizontal stress and geo-
metrical parameter are three major factors which
significantly influencethestability of T-intersection.
Italsoindicatedthatthestressesinducedduringinter-
sectionformationmayresultinhighincidenceof roof
andribfailures.
799
InChina, thedepthof cover incoal mineisincreas-
ing 812mannually, and there are many collieries
are going to face deep mining situation, with depth
of cover around10001500m. Withincreasingof the
depth of cover, the stress regime and deformation
characteristics of roadway is changed significantly,
it resulted in failureand difficult to ground control.
Also, construction condition is getting worse than
shallow conditions (J in and Sun, 2001). Under such
situation, thestability of intersection is facing more
seriousproblemduringthedeepmining.
The depth of cover is a sensitive factor to the
groundcontrol. For example, thegroundproblemwas
occurredwhenthedepthof cover was changedfrom
550mtoover 600mbelowthesurfaceinXiezhuang
colliery, Shandong. Previously, thestability of inter-
sectioncanbemaintainedby usingsteel framewhen
thedepthof cover lessthan600m, butwhenthedepth
of cover reached over 800m, ground control around
intersection becameamajor problem, dueto: 1) the
softrockcharacteristicsof surroundingrockmasswas
encountered and themechanical properties was get-
tingworse; 2) thecapacityof reinforcingelementwas
reducedsignificantly; 3) theplasticzonearoundinter-
sectionincreasedsignificantly, 4) roof separationwas
formed significantly and it is considered as amajor
factor to cause the stability of intersection (Wang,
2001).
Thefurther studyontwowayintersectionindicated
that, indeepunderground(700900m), theangleof
intersectionisnotamajor factor toinfluencetheplas-
tic zone around intersection, but when the angle is
less than35degree, thereinforcingareais increased
significantlywithdecreasingof theangle. It is, there-
fore, suggestedthat theanglebetweentwo roadways
shouldnot belessthan35degreefor thedesignation
of intersection(ZhuandCao, 2005).
On theother hand, many ground control problem
relatedtointersectionarenotaffectedbysinglefactor,
but combination phenomenon, which involved roof
stress, physical propertiesof mudstone, reinforcement
techniqueandparametersusedarenotsuitableforsuch
undergroundenvironment (He, et al, 2005).
3 MODELINGOF INTERSECTIONSTABILITY
Themodelingwas conductedby usingFLAC
3D
with
three dimensional and non-linear simulations. The
numerical model used for the analysis is presented
in Figure 1, hard rock to soft rock, the mechanical
propertiesaregiveninTable1.
3.1 Effect of geometries on stability of intersection
Therearelargenumber of roadway intersections are
constructedannuallyinChinacoal minesindustrywith
Figure1. Model usedinthesimulation.
variousgeometries, includingthree-way intersection,
four-wayintersection, twodimensional roadwayinter-
section(Figure2a,b,c) andthreedimensional roadway
intersection (Figure 2d,e) and geological conditions,
e.g. strongrockmassandsoftrockmass.Withincreas-
ingof theminingdepth, thestabilityof theseroadway
intersectionsissusceptibletogroundcontrol problems
due to inherently wide roof spans and complicated
intersectiongeometryused.
The study conducted in this section is to clarify
theeffect of thegeometrical shapesonthestabilityof
intersection.Todoso, thedepthof coverisdetermined
as800m, andthetwodimensional intersectionswith
five different geometrical shapes, plus one normal
singleroadwayarecomparedas:
Cross intersection, with 90
angle between two

roadways;
X intersection, similar with cross, but with 45
anglebetweentworoadways;
T intersection;
L intersection;
Y intersection;
Singleroadway
Themodelingwasconductedtoevaluatethestabil-
ity behaviour of theintersectioninterms of theroof
andribdeformationandthefailuredepthintotherock
masswithdifferentshapes, onthebasisof mechanical
propertiesgiveninTable1. Thedeformationandfail-
ureof theintersection havebeen compared with the
normal singleroadway, it isindicatedthat thesimpler
thegeometry, themorestabletheundergroundstruc-
tures. The results also shown that the most unstable
two-dimensional intersectionisthecross-intersection
(Figures3and4).
800
Table1. Mechanical parametersof rockmass.
Modular Modular Angle Tensile Residual Residual
E of volume of share Density Cohesion of inter- Strength Cohesion angleof inter-
No (GPa) j (GPa) (GPa) (kg/m
3
) (Mpa) friction(o) (MPa) (MPa) friction(o)
1 12 0.25 8 4.8 2500 3.7 26 1.4 0 22
Figure2. MainintersectionsusedinChinesecoal industry.
Figure3. Comparisonof roof/ribdeformationwithdifferent
geometrical shapes.
Figure 4. Comparison of roof/rib failure with different
geometrical shapes.
3.2 Effect of excavation sequence on the stability of
intersection
Stresses induced during the intersection formation
may result in high incidence of roof and rib failure
(Singh, 2001), thustheconstructionsequence(proce-
dure) may beanother sensitivefactor that influences
Figure5. Sequenceof intersectionconstruction.
the stability of intersection. As the indicated above,
thecross-intersectionisthemost unstabletwodimen-
sional intersection, thus, theeffect of theexcavation
sequenceon stability of cross-intersection is studied
accordingly.
Figure5 shows theconstruction sequences of the
cross-intersection, thesequencesaredesignedasfive
different typesof a, b, c, d, ande, whichrepresent the
sequencesanddirectionsof theconstruction. Accord-
ing to the modeling results, the minimum vertical
stressanddeformationhavebeenfoundfromtheexca-
vation sequenceNos. 1 and 2, and both represented
similar construction sequence, that is, the roadways
formedtheintersectionaredevelopedatdifferenttime.
It is indicatedthat eventhestress statearoundopen-
ing are interrupted twice by using these developing
sequences,buttheeachinterruptionisconsideredtobe
smaller comparing with other construction sequence
proposed.
The developing direction is also sensitive to the
stability of intersection. If the excavation is toward
the existing opening (Figure 5e and Figures 6a&b),
the significant affect on the stability of intersection
isdetached, andif theexcavationdirectionismoving
awayfromintersection(Figure5bandFigures6a&b),
theaffect iscomparativelysmall (Figure6).
3.3 Effect of depth of cover on stability of
intersection
Thedepthof cover isalwaysasensitivefactor, which
shouldbetakenintoconsiderationduringthestability
analysis. Based on the previous work, the cross-
intersection is analyzed with various levels of depth
of cover from200mto1100mrespectively. Figure7
shows the stability of the intersection various with
depth of cover, and it indicated that the stability is
not affected directly with increasing of thedepth of
cover. Before the depth of cover reached a certain
level, such as 1000m, thereis not significant defor-
mation and failureoccurred around theintersection,
thus, the stability can be properly maintained. After
801
Figure6. Effect of construction sequences on stability of
intersection.
Figure7. Effectof depthof coveronstabilityof intersection.
thedepth of cover reached 1000m, thedeformation
and failureis developed significantly, it implies that
reinforcement strategies used for the intersection in
shallow conditions may not be suitably used in the
deepcondition.
4 SUMMARIESANDCONCLUSIONS
Dueto inherently wideroof spans used, thestability
of theroadwayintersectionwasalwayspaidattention
bytheresearchersandengineers. Now, withcontinual
increasingof theminingdepthof cover,moreandmore
serious stability problems, particularly related to the
roadway intersection, will be faced in Chinese coal
mines.
Thestudy confirmedthat thestability of theinter-
sectioncouldbemaintainedintheshallowcondition,
but when thedepth of cover reached acertain level,
suchasover 1000mbelowthesurface, thestabilityof
intersectionisunder thethreat.
Thegeometry of theintersection is concerned, as
comparatively thegeometrical shapes used in China
coal miningindustryarerelativelycomplicated, which
results in much more wide roof span than normal
heading roof. Comparing with six different types of
theunderground opening, it is noted that, under the
similar conditions, thenormal roadwayhasminimum
displacement and failure than other intersections. It
implies that thelargeroof span will causemoresta-
bility problemthansmall roof span. Also, comparing
different geometrical shapes of the two-dimensional
intersections, the most unstable one is the cross-
intersection, whichis themost comment intersection
usedincoal mines.
Theconstructionsequencesalsoinfluencethesta-
bility of the intersection during the excavation. To
minimizethestress interruption to existing roadway
during theintersection development, particularly for
thecross-intersection, it issuggestedthat all theaddi-
tional roadwayconstructionshouldalwaysmoveaway
fromrather thanforwardto theintersection. Inaddi-
tion, theconstruction for additional roadway around
the intersection should be conducted one side by
another individually (Figure 5b), even this kind of
construction sequence will cause twice stress con-
centration and redistribution, but it still is the less
influence on the stability of intersection comparing
other methodsof theconstruction.
Thedepthof coverisrecognizedasthemostimpor-
tant factor for thestability of undergroundopenings.
Accordingtothemodelingresults, thestabilityof the
intersectionisnotlinearlyrelatedtothedepthof cover,
the significant effect only occurred when the depth
of cover reachedto1000munder typical surrounding
rockmassconditionsselected.
ACKNOWLEDGMENTS
Authors wish to acknowledge The Innovation Fund
for Outstanding Scholar of Henan Province(Project
No.:0621000400) for providing the financial assis-
tance.
802
REFERENCES
Balachandra, M.B. 1976. The three-dimensional structural
analysis of double-entry and single-entry coal mines.
Volum I: three-dimensional finite element analysis of
crosscut and entry intersection of a double-entry coal
mine, research report, report number(s), PB-80-150345.
Hanna, K. etal. 1991. Coal-mine-entryintersectionbehavior
study, Report of Investigations.
He, M.C. et al. 2005. Research on Stability and Support
Measuresof CrossingPoint inMineSoft Rock Roadway
of Deep Third System, Mine Construction Technology,
Vol.26, No.34, pp. 3235(inChinese).
J in,H.H.&Sun,Q.G.2001.Highstressandsoftrockroadway
reinforcement problems and strategy, Coal Technology,
Vol.3, No.3, pp. 3435(inChinese).
Peng, S.S. 1978. Roof bolting patterns at three-way
entry intersections, research report, report number(s)
NP-23983.
Singh, R.N., Porter, I. & Hematian, J. 2001. Finiteelement
analysisof three-wayroadwayjunctionsinlongwall min-
ing, International Journal of Coal Geology,Vol.45, Issues
23, J anuary, Pages115125.
Wang, P.L. 2001. Discuss of supporting types in roadway
cross areain deep seam, Coal mining, Vol.4, pp. 5759
(inChinese).
Zhu,Y.S. &Cao, S.H. 2005. ResearchonRoof LoosedLoop
andSupport ReinforcedExtent inRoadway CrossPoint,
MineConstructionTechnology, Vol.26, No.2, pp. 2832
(inChinese).
803
Analysisof surfacesettlement duetotheconstructionof ashield
tunnel insoft clayinShanghai
Z.P. Lu& G.B. Liu
Department of Geotechnical Engineering, Tongji University, P.R.China
ABSTRACT: Bycomparisonof methodsof settlement calculationswhichhavebeenappliedintocomputing
surfacesettlementduetoshieldtunneling,thepaperanalyzedtherelativitybetweenPecksequationandstochastic
mediummethod, andresearchedtheapplicationprobleminpractical engineering.Associatedwiththemeasured
dataof aconstructingtunnel, somecharactersof surfacedeformationhavebeenanalyzedtoappraisecurrentlevel
of constructiontechnologyinsoftgroundinShanghai.Atthebasisof establishednormal groundloss, thequality
of constructioncanbedeterminedandjobpracticecanbeadjustedtomeet therequirement of environmental
protection.
1 INTRODUCTION
Ground movement, which is caused by ashield tun-
nelinginsoftclay, will givebirthtosurfacesettlement
of different level. When surface settlement arrives
at certain level, the regular service of building and
underground installation around will be influenced.
Therefore, wemust findout theregularity of ground
movement,andmoreexactlypredictthevalue,rangeof
settlementandmaximumgradientof settlementcurve.
Besides, in order to adopt some measures reducing
groundmovementduringconstructionanddesign, we
needtoanalyzeall kindsof factorsinfluencingsurface
settlement.
Presently, theresearchabout predictionof surface
settlementduetotheconstructionof shieldtunnel can
begeneralizedasfollows:
Empirical method;
Theoretical method;
Model experiment method.
This paper presents it most important to predict
the settlement value in allowable error variation by
the most economical and convenient method, which
is used to direct construction process and control
peripheral environmental safety. Thefinal settlement
is mainly caused by ground loss and soil consoli-
dation. And in the period of construction, surface
settlement is mainly causedby soil pilinginto shield
gapinundrainedcondition. Sobasedonthecompar-
ison of several computing method, the paper firstly
discusses the relationship and applicability of them,
andsecondlyanalyzestheissueof parameterselecting.
Intheend, somecharactersof groundmovement and
thecalculationof groundloss arepresented. To sum
up,lesssurfacesettlementindicateslessgrounddistur-
bance, andthegoal istopresentreasonableprediction
methodandcontrol measuresof surfacesettlement.
2 REASONANDREGULARITY OF SURFACE
SETTLEMENT
2.1 Reason of surface settlement
The reason of surface settlement caused by shield
tunnelingcanbegeneralizedasfollows:
1. Groundloss
Groundmovement of working;
Shieldsetback;
Soil pilingintoshieldgap;
Changingdirectionof shielddriving;
Ground friction and shearing because of shell
movement of shieldmachine;
Deformationandsettlement of tunnel lining.
2. Consolidationof disturbedsoil.
2.2 Universal regularity of surface settlement
When shield method is adopted in saturated soft
clay, the universal regularity of ground deformation
alonglongitudinal axisof shieldtunnel ispresentedin
805
Figure1. Universal regularityof surfacedeformation.
Figure2. Figureof surfacesettler abovetunnel.
Figure1.Usuallygroundsurfacewill givebirthtolittle
humpingbecauseof uprisingandheadingof squeezed
soil body in front of shield machine. When shield
machinepasses, bilateral soil bodywill moveoutward.
When tunnel lining disengages fromtheend part of
shieldmachine, groundsurfacewill occur rather large
settlement and thespeed rateof settlement is rather
large.
3 ANALYSISOF SETTLEMENT EQUATION
3.1 Pecks equation
Peck(1969) adopted the following expression to
describethesettlementcomponentduetotheconstruc-
tionof ashieldtunneling. Pecksequationisexpressed
inFigure2.
whereS(x)=valueof settlement;V
l
=valueof ground
loss in unit length of shield tunnel; x =distance to
center lineof tunnel; i =spreadfactor of settler.
Pecks equationexpresses theconceptionthat sur-
face settler is approximate normal distribution. And
this estimated equation considers that ground move-
ment is caused by ground loss. Because it is based
on engineering experience, Pecks equation exists
Figure3. Schematicplanof tunnel driving.
some shortages (e.g. considering fewer factor and
lackof theoretical foundation). If valueof parameters
existsmoredifferenceincommoncondition, result of
calculationmaybeinfluencedobviously.
3.2 Stochastic medium model
Schematicplanof tunnel drivingof stochasticmedium
model isshowninFigure3. Inthismethod, afteralong
time(t ), thefinal valueof element subsidence
ispresentedinplanestrainstateasfollows:
whereS
e
(x)=valueof element subsidenceinZ level
plane; r(z)=mainrangeof influencedueto element
drivinginZlevel plane; =mainangleof influence.
Applying superposition principle, value of cross-
sectionsurfacesettlement(z =0)duetotunnel driving
maybecalculatedby
where S(x)=value of cross-section surface settle-
ment; O, = area of tunnel driving and tunnel in
unit length; O=areaof groundloss.
Thesolutionof equation(4)dependsondoubleinte-
gral. However, theintegrandsof theseintegralscannot
beintegrated, sothemethodof numerical integration
mustbeapplied. Here, valueof integral canbegained
byLegendre-Gaussmethod,
806
Figure4. Model of uniformradial displacement.
where n =number of integral point; A
i
, A
j
=weight
number of Gauss;
1i
,
2j
=interpolation points of
integratedvariable.
3.3 Analysis of computing settlement
Basedontheory of grounddeformationpresentedby
Sagaseta (1987), Verruijt & Booker (1996) extend
the theory and present new analytical equation of
ground deformation due to tunnel driving in elastic
half-space, whichis alsobasedonequivalent ground
loss model (Lee et al. 1992). Verruijt & Booker
adopted the model of uniform radial displacement
(Fig. 4).
Parameter of equivalent ground loss model is
definedas
whered =external diameter of tunnel; g =parameter
of gap.
Verruijt & Bookers equation is expressed as fol-
lows(not consideringdeformationof tunnel liningin
long-timeeffectsandconsideringundrainedstate):
whereH =embedded depth of tunnel; =Poissons
ratio;z
1
=z H;z
2
=z +H;r
2
1
=x
2
+z
2
1
;r
2
2
=x
2
+z
2
2
;
m=1/(12).
Whenz =0, surfacesettlement iscalculatedby
4 COMPARISONOF METHODSOF
SETTLEMENT COMPUTATION
Infact, thegapof tunnel driving(groundloss) must
be defined for several universally adopted method
of surface settlement above. And it corresponds to
separately:
V
l
inPecksequation;
O instochasticmediummodel;
inVerruijt andBookersequation
Assumingthat groundlossof tunnel isainfinitesi-
mal element dd.
CalculatingbyPecksequation:
Calculatingbystochasticmediummethod:
Contrastingequations(10) and(11), if
thetwoequationsarecompletelyidentical.
Ontheonehand, theapproximatesolutionof sur-
facesettlement correspondingto different valueof x
canbeobtained, bythenumerical integrationof equa-
tion(4). Thenthecurveof surfacesettlement dueto
tunnel driving will begained by curvefitting of the
resultabove.Ontheotherhand,regressionanalysiscan
beappliedintoequation(4) accordingtotheformatof
Gausscurve:
The model of ground loss is usually obtained by
assumption, and heresurfacesettlement will becal-
culated separately by threemethods above, adopting
model of uniformradial displacement.
Assuming R=d/2=3m, g =0.1m, =50
, and
surfacesettlementwill becalculatedaccordingtotwo
conditions (z
0
/R=1.3 or 6.7, z
0
=distance between
groundsurfaceandcenter of tunnel). Byconversion:
Byregressionanalysis, equation(12) isexpressed:
If z
0
,R=1.3,
807
Figure5. Comparisonof computingmethods.
If z
0
,R=6.7,
Figure56indicatethatthecomputingresultsexist
less differencebetweenPecks equationandstochas-
ticmediummodel if embeddeddepthof tunnel ishigh
(e.g.z
0
/R>6).If embeddeddepthof tunnel islow(e.g.
z
0
/R-2), calculationerror isrelativelylargebetween
twomethods. Besides, thecurveof surfacesettlement
by stochastic medium model corresponds with the
superimpositionof Gausscurvesof surfacesettlement
duetoseriesof infinitesimal elementsdriving. There-
fore, it isrelatedwithgeometriccharacter of working
face, convergencestyleafter driving, not correspond-
ingtoGaussdistributioncompletelyinstrictmeaning
andespecially generatingrelatively largeerror being
close to maximum settlement. On the other hand,
Pecksequationdirectly agreestoGaussdistribution.
In a word, stochastic mediumtheory should be the
theoretical basisof Pecksequation, andPecksequa-
tion is the simplified method of stochastic medium
method, but adoptingdifferent methodsfor parameter
selecting.
Figure78indicatethat Poissons ratio of Verruijt
and Bookers method locates between 0.1 and 0.2,
adoptingrelativemethodof parameters. Anditsresult
isclosetoPecksequationwhencomputingmaximum
settlement. If embedded depth is relatively high, the
maximumvalueof settlement isveryclose.
5 PRINCIPLE OF APPLICATIONOF
COMPUTINGMETHODS
At present, methodsof settlement computationabove
havebeenappliedintopractical engineering. Accord-
ingtoanalysisabove, itisthekeytocorrectlyselecting
parameters for different methods. In other words,
whether method is applied, the separate method of
parameter selecting should beestablished by experi-
enceor theory.
Pecks equationhas beenappliedinthedistrict of
softsoil inShanghai, becauseof itscharactersof sim-
plicityandpracticality.Andtheprincipleof parameter
808
SZ181-6
SZ181-5
SZ181-4
SZ181
SZ181-3
SZ181-1
SZ176-6
SZ176-5
SZ176-4
SZ176
SZ176-3
SZ176-1
SZ171-6
SZ171-5
SZ171-4
SZ171
SZ171-1
SZ171-3
SZ15-6
SZ15-5
SZ15-4
SZ15
SZ15-3
SZ15-1
SZ10-6
SZ10-5
SZ10-4
SZ10-1
SZ10-3
SZ10 Direction of
shield driving
SZ185
SZ05-6
SZ05-5
SZ05-4
SZ05-3
SZ05-1
SZ05 SZ01
SZ05-2 SZ10-2 SZ15-2 SZ171-2 SZ176-2 SZ181-2
Figure9. Theplacement of monitoringpoints.
selectinghasbeengeneralized. Spreadfactorof settler
isexpressedasfollows(Clough& Schmidt 1981):
Stochastic mediumtheory canconsider multiform
factors (e.g. different job practices of tunnel and
sectional styles, etc.), and can calculate multiform
conditions of ground movement (e.g. horizontal dis-
placementof groundsurfaceandsurfacesloping,etc.).
About parameter selecting, integral domain (O)
andangel of influence() needtobeobtainedbyback
analysisof measuredmonitoringdata.
Verruijt & Bookers method was modified later
(Loganathan& Poulos1998, Bobet 2001). Theresult
relativelycorrespondswithpractice,butthesemethods
still existsomeproblem(e.g.uncertaintyof parameters
of groundloss, curveshape, etc.).
In aword, surfacesettlement can bepredicted by
Pecks equation or analytical equation when solving
problemof parameter selecting according amass of
statistic of practical engineering. To important and
complicated conditions, stochastic medium method
associated with back analysis is able to gain more
informationandmorepreciseresolution.
6 ANALYSISBASEDONMEASURED
MOINTORINGDATA
Thepaper chooses aconstructingtunnel at Shanghai
Metroline7assubjectinvestigated.Thetunnel adopts
shieldmethodof earth-pressurebalancethroughtwo
mainlayers:
Muddyclaybelongingtotypical soft soil;
Siltyclaybelongingtohardsoil.
Thelongitudinal directionof tunnel isVinshape,
and thetunnel passes through thecrossover facet of
two soil layers. The placement of monitoring points
ongroundsurfaceisshowninFigure9, andthespac-
inginterval of monitoringpoints is 5m. Thesurface
settlement will be separately researched by mainly
considering two factors (ground movement of work-
ingfaceandsoil pilingintogapof endpart of shield
machine).
Maximum head deformation (mm)
-8 -6 -4 -2 0 2 4 6 8 10
C
o
n
s
t
r
u
c
t
i
o
n

s
e
t
t
e
l
m
e
n
t

(
m
m
)
0
5
10
15
20
25
30
35
40
Figure 10. The relationship between maximum surface
deformation before shield arriving and maximumsurface
settlement after shieldheadingat thesamepoints.
6.1 Surface settlement caused by ground movement
of working face
Figure10generalizestherelationshipbetweenmaxi-
mumsurfacedeformationbeforeshieldarrivingand
maximumsurface settlement after shield heading at
thesamepoints. Infact, theyappear todispersivedis-
tribution: surfacedeformationbeforeshieldarrivingis
either humpingor settlement andcorrespondingsur-
facesettlement after shieldheadingis either largeor
small. Thereforethewholeprocess of surfacedefor-
mation at certain point is influenced by multiform
factors (e.g. soil layer of tunnel driving, embedded
depth, constructionprocedure, etc.).
6.2 Final settlement during construction
V
l
inequation(1) iscalculatedby
whereD
y
=external diameter of tunnel; V
l
%=ratio
of groundloss; V
h
=groundlossduetogroundmove-
ment of workingface.
Table12andFigure11indicate:
1. Inorder tocontrol surfacesettlementinthedistrict
of complicated surrounding, the method of low-
speeddrivingandsynchronal injectingas soonas
possible can be adopted to gain good result (e.g.
S77,S117).
2. Theratioof groundlosscausedbygroundhumping
beforeshieldarrivingingroundlosscausedbysoil
pilingintogapof endpartof shieldmachineisvery
little. Sogrounddisturbancebeforeshieldarriving
canbecontrolledat reasonablerangeat present in
Shanghai.
3. Theamount of humpinghas aneffect onthecon-
striction of maximum surface settlement during
construction, but larger humpingwill givebirthto
largergrounddisturbanceandinfluencesubsequent
valueof consolidationsettlement.
809
Table1. Parameterscalculationof monitoringpointsbased
onPecksequation.
Monitoring Z
0
i S
hmax
V
h
S
max
V
l
points m m mm mm
2
m mm
2
SZ47 14.89 6.25 1.0 0.02 17.1 0.27
SZ77 16.80 6.88 1.2 0.02 8.50 0.15
SZ107 16.90 6.91 1.5 0.03 16.7 0.29
SZ117 16.61 6.82 2.7 0.05 5.60 0.10
SZ127 15.85 6.57 1.8 0.03 14.5 0.24
SZ137 14.70 6.18 3.6 0.06 14.4 0.22
SZ157 13.13 5.65 1.4 0.02 13.4 0.19
SZ171 11.95 5.24 1.7 0.02 24.3 0.32
SZ176 11.95 5.24 3.4 0.04 17.6 0.23
*S
hmax
is the maximum value of humping at monitoring
points, and S
max
is the maximum value of settlement at
monitoringpoints.
Table 2. Controlled maximum settlement and measured
maximumsettlement.
Monitoring S
cmax
S
max
points mm mm
SZ47 14.0 17.1
SZ107 12.5 16.7
SZ127 12.0 14.5
SZ137 14.0 14.4
SZ157 16.5 13.4
SZ171 15.3 24.3
SZ176 15.3 17.6
*S
cmax
is the maximum allowable value of settlement at
monitoringpoints.
Distance to the center of tunnel (m)
-20 -15 -10 -5 0 5 10 15 20
S
u
r
f
a
c
e

s
e
t
t
e
l
m
e
n
t

(
m
m
)
-5
0
5
10
15
20
25
30
35
SZ10
SZ27
SZ37
SZ47
SZ57
SZ77
SZ87
SZ107
SZ117
SZ127
SZ137
SZ147
SZ157
SZ171
SZ176
Figure11. Final settlement duringconstructionperiod.
4. Constructing shield-driven tunnel in Shanghai,
30mmisregardedascontrollingstandardfor max-
imum settlement (considering that the depth of
earth covering is 4m). With many metro lines
constructed and larger embedded depth, the new
standardof controllingsurfacesettlement isestab-
lishedaccordingto V
l
%1%inPecks equation.
By analyzing monitoring result, the construction
technologyissatisfiedwiththerequirementof con-
trollingsurfacesettlement whenthedepthof earth
coveringis15mor so.
7 CONCLUSIONS
Bycomparisonof methodsof settlement calculations
whichhavebeenappliedinto computingsurfaceset-
tlement due to shield tunneling, the paper analyzed
the relativity between Pecks equation and stochas-
tic mediummethod, and researched the application
probleminpractical engineering. Associatedwiththe
measured data of a constructing tunnel, some char-
acters of surfacedeformation havebeen analyzed to
appraise current level of construction technology in
soft ground in Shanghai. At thebasis of established
normal groundloss, thequalityof constructioncanbe
determinedandjobpracticecanbeadjustedto meet
therequirement of environmental protection.
REFERENCES
Bobet, A. 2001. Analytical solutions for shallow tunnels
in saturated ground. Journal of Engineering Mechanics
127(12):12581266.
Clough, G.W. &Schmidt, B. 1981. Design and performance
of excavations and tunnels in soft clay. NewYork: Elsevier
SciencePublishingCompany.
Lee,K.M.&RoweR.K.1992.Subsidenceowingtotunneling
I:Estimating thegap parameter. Canadian Geotechnical
Journal 29(6): 929940.
Liu, J.H. & Hou X.Y. 1991. Shield-driven tunnel. Beijing:
ChinaRailwayPublishingHouse.
for tunneling-induced ground movement in clays. Jour-
nal of Geotechnical and Geoenvironmental Eengineering
124(9): 846856.
ground. Proceeding of 7th international conference on
soil mechanics and foundation engineering. MexicoCity:
Stateof theArt Report.
Sagaseta, C. 1987. Analysis of undrained soil deformation
duetogroundloss. Geotechnique 37(3):301320.
Geotechnique 46(4):753756.
Zhu, Z.L. &Zhang, Q.H. 2001. Stochastictheoryforpredict-
inglongitudinal settlement insoft-soil tunnel. Rock and
Soil Mechanics 22(1):5659.
810
Urbantunnelsinsoil: Reviewof current designpracticeinBrazil
A. Negro
Bureau de Projetos, So Paulo, Brazil
ABSTRACT: The current design practice of urban tunnels in soil as perceived in Brazil is reviewed and
discussed. Thereviewisbasedonanswerstoaquestionnairesentin2006topractitionersinvolvedinthedesign
orinthedesignsupervisionof tunnel projects.Theresultsof thisinvestigationarecarefullyconsideredtoidentify
trendsandneedsfor development.
1 INTRODUCTION
Thefirst assessment of thedesign practiceof urban
tunnels driven in soil performed in Brazil was car-
riedout in1993throughasurvey whoseresultswere
published in the 1st. International Symposium on
Underground Constructions in Soft Ground in New
Delhi (NegroandLeite, 1995).
Thistimecut-and-cover structuresarenotincluded
and thereview covers underground openings of any
shape with more than 0.5m diameter, directional
drillings also not included. The survey is not lim-
ited to tunnels built in Brazil and covers any tun-
nel project conducted with criteria and procedures
presently adoptedinBrazil. Inall cases tunnels were
designedandbuilt inareaswithbuildingsonthesur-
face and utilities on the subsurface, both liable to
damagesinducedbygroundmovementsassociatedto
tunnel excavation.
Multiple choice questions to which one or more
answers couldbeselectedweresent to professionals
involved in the design or in the design supervision
of tunnel projects over thelast few years. Consider-
ingthemultidisciplinary natureof tunnel design, the
questionnaire was sent to practitioners froma vari-
ety of expertise and included topics beyond strict
geotechnical context. Accordingly, questions were
sent to geotechnical engineers, geologists, structural
engineers, expertsinnumerical modelling.
These professionals could opt out questions tran-
scending their specialties. Most questions asked for
answersexpressingthepractitionercurrentpreference
on each technical aspect of tunnel design, in such a
waythattheytranslatedtherespondentsroutinedesign
practice. Thequestionsweresent to30tunnel experts
and20repliedthesurvey.
The results presented below are grouped within-
related topics. One should note that respondents
offered sometimes morethan oneanswer to aques-
tion.Therefore, frequenciesshownrefer topercentage
of total answersofferedortopercentageof respondents
asapplicable.
2 THE SCENARIOOF THE PRACTICE
Tomakesimplertheanalysisof theanswers, fewques-
tions were formulated to define the scenario of the
designpracticetowhichit refers.
Morethantwothirdsof theanswersrefer totunnels
builtfor trains, metrosystemsandvehiclesingeneral.
Accordingly, morethan four fifths of theanswers
(from26given)refertotunnelswithmorethan4to6m
diameter (almost twothirds withdiameters inexcess
of 6m).
The ground condition more frequently encoun-
tered, isthatof tunnellingwithmixedface.Tunnelling
through cohesionless ground is not frequent. On the
other hand, morefrequentlythannot, tunnellingtakes
placebelowwater table.
3 LININGDESIGNANDWATER PROOFING
The type of secondary lining more frequently used
over the recent years is the wet mix sprayed con-
crete(45%), followedinequal proportionsbythedry
mix and by the cast in place concrete (23% each).
Theanswersareconsistentwiththelarger tunnel sizes
focused. Steel liner plates as final lining are not in
use, possibly for being liableto electric or chemical
corrosioninadverseurbanscenarios.
Thesurveyrevealedthat wiremeshesarethemost
popular typeof reinforcement for secondary linings.
It is somewhat surprising to see that lattice girders
and steel sets are still in use, though they are not
usuallytakenintoaccount inthedesignof secondary
811
linings. Steel fibersarenotmuchusedanditisalmost
non existent the old practice of unreinforced con-
crete secondary linings: it looks like that there is a
general perception that it is necessary to ensure an
adequateductility renderedby steel reinforcement to
final tunnel linings, assuringthat brittlefailureof the
liningdonottakeplaceunderunexpectedcatastrophic
conditions. This is in line with the requirement of
minimumsteel reinforcement acceptedby85%of the
respondents(theremainingdonot adopt it).
Forty percent of the respondents explained that
they useonly theminimumsteel sectionfor bending
whereas60%adopttheminimumbothforbendingand
for shrinkagecontrol.
60%of therespondentseither never designtunnels
to bewaterproof (15%) or they didit only fewtimes
(45%). The most favorite technique to water proof
tunnel liningsareimperviousgeomembranes, system
(39% of the answers) that has been just introduced
inthecountry. Ontheother handthisissecondedby
agroupof moretraditional procedures inBrazil that
aimsthereductionof concretepermeability, byspecial
mix design, special admixturesandcontrolledcuring
of theconcrete(summing32,3%of theanswers). This
groupcombinedwiththedesignof concretewithlim-
itedcrackwidthstotal 45%of theanswers, while5to
10yearsagotheywouldcountfor100%of theanswers.
4 GEOTECHNICAL DESIGN
All respondents declared they normally performsta-
bility analysis of thetunnel heading. Figure1shows
that almost onethirdof theanswers indicatetheuse
of thesafeLower BoundsolutionsfromtheTheoryof
Plasticity. Some prefer the use of limit equilibrium
methods or the sole use of Upper Bound solutions
fromPlasticity,whicharenormallyunsafeapproaches.
Empirical methods(that caneither besafeor unsafe)
and finiteelement analysis with elastic plastic mod-
els ranked equally, with almost 14%of the answers
both. It should be noted that within the latter, none
of themwerereallyNumerical LimitAnalysis(NLA),
suchasthat presentedbyDurandel al (2000). Infact
they refer to 2D FE analysis in which afull ground
stressunloadingisappliedtocheckif the2Dopening
withstandsit, without anysupport.
It is worthnotingthat respondents adopt, inequal
proportion (40%), distinct approaches for stability
analysisof headingsof tunnelsbuilt withashieldand
without it (minedtunnel, NATM). No formal reason
justifiessuchpractice.
All respondent confirmthat they normally predict
thesettlementsinducedbytunnel constructionintheir
routinepracticeandthatthisisdonethroughnumerical
analysis(bothfiniteelementsandfinitedifferences)
seeFigure2. Comparingtheseresults, withthoseof
the1993survey, aconsiderabledecreaseintheuseof
Figure 1. Methods of analysis for assessment of tunnel
headingstability.
Figure2. Methodsfor settlementsestimates.
empirical andsemi-empirical methodsisnoted(from
60% to 20% now). With respect to the assessment
of induceddamagesonbuildings(Figure3), noclear
preferenceto asinglemethodor procedureis noted:
the methods by Burland and Wroth (1974) and by
BoscardinandCording(1989) werethosewithmore
indications(21%of theanswersboth). Themost used
plane static systems to calculate normal forces and
bendingmomentsinthetunnel lining, accountingthe
interactionbetweenitandtheground(seeFigure4)are
numerical solutionswith74%of theanswers(48%for
FE and26%for FD). A considerableincreaseinthe
useof thesetypes of solutions was seeninthecoun-
trysince1993, whenonly43%of theanswersfavored
them.
It was noted also that in the design of the pri-
marylining, using2Dstaticsystems, 85%of thecases
consider areductionof thegeostatic stresses andthe
remaining15%of answersconsider or not thisreduc-
tion, dependingof thespecific conditionsof thecase
beingstudied. Theproceduremorefrequentlyusedto
estimatethereductiononthegroundstresses(Figure5)
812
Figure3. Methods for assessment of damages inducedby
tunnellingonbuildingsat groundsurface.
Figure4. 2Dstaticsystemstoassessliningloads.
arethosederivednumerically (NegroandEisenstein,
1997,forinstance),followedbyTerzaghisarchingthe-
ory, notwithstandingthefact that this corresponds to
largegroundstressreleasesassociatedtolargeinduced
displacements, whicharenot compatibleto anurban
scenario.
Whendesigningthesecondarylining, only59%of
the practitioners assume a reduction on the ground
stresses, possibly reflecting an old and debatable
believe of long termcreep of geological materials,
leadingtofullydecayedshear strength.
Practitionerswereaskedhowtoaccountforground-
water loads on the sprayed concrete primary lining
of a tunnel driven below water table (see Figure 6).
50%of theanswers indicatethat no account is taken
of water pressures, ontheassumptionthat thesoil is
Figure 5. Procedures to estimate reduction of ground
stresses.
Figure6. Account of groundwater loadingover aprimary
liningof atunnel belowwater table.
fullydrained(45%of answers) bycompulsorydewa-
tering; few (5%) take into account the body forces
resultingfromdewateringtheground. Theother half
of responses indicatethat they takeinto account the
actingporewater pressures andtheeffectivestresses
ontothelining(either reducedor not).
Itisworthnoting(seeFigure7) thatmorethanhalf
of answersofferedindicatethatpractitionerseither do
not takeinto account theeffects of porewater pres-
sures inthesoil behaviour (4%) or limit theaccount
to thewater pressureonto thelining at most (54%).
J ust over 20% performnumerical analysis coupling
soil stressesandporewater pressuresusingfiniteele-
ments. Alsoworthnoting(Figure8) that almost 80%
of theanswersindicatetheuseof elastic-plasticmod-
elsassociatedtotheMohr-Coulombfailurecriterion.
Theproblemwiththispopular optionisthatconsider-
ingthatwearedealingwithanurbantunnel, inwhich
considerableefforts arespent inreducingthedegree
of groundstressrelaxationandsoil deformation, plas-
ticstrainingisusuallypreventedandgroundresponse
813
Figure7. Accountof porewater pressureonsoil behaviour.
Figure 8. Type of constitutive models used for soil in
numerical analysis.
isreducedtoalinear elasticbehaviour, anoversimpli-
fiedportrayof amaterial knowntoexhibit nonlinear
behaviour evenat small strainrange.
Despitethe3D stress redistribution around atun-
nel heading, rarelyor never (55%of respondents) 3D
numerical analysis are performed in routine design.
Whenthey areperformed, they refer totunnelsinter-
sectionsorstartupshaftsof tunnels.When3Danalysis
areperformed, themost frequentlyusedtypeof anal-
ysis (in 32%of the answers) involve shell elements
for theliningandbarsor springstorepresent thesoil.
3D finiteelement programs normally used areSAP,
Ansys, Plaxis and FLAC 3D. The more intense use
of 3D analysisalready notedintheacademy (seethe
5thProceedingsof TC28SymposiuminAmsterdam,
2005), resultingfromtheavailabilityof morepowerful
PCsandmoreefficient numerical codes, will alsobe
noteinindustryinthenext fewyears.
Figure9. Choicesfor geotechnical investigations.
5 GEOTECHNICAL INVESTIGATIONSAND
MONITORING
Oneof thequestions preparedreferredto theprefer-
enceof practitionerswithrespecttodifferentgeotech-
nical investigations performedinthefieldandinthe
laboratory.
They were asked to indicate, in a scale from 0
(never use) to 3 (always use), their choices regard-
ingeachclassof investigation. Resultsof thissurvey
are shown in Figure 9, fromwhich it can be noted
that:a)standardpenetrationtestsarealwaysrequested;
b) conepenetrationtestingandundisturbedsampling
followed by laboratory special testing are never or
rarely requested; c) disturbed sampling followed by
characterizationtests inthelaboratory arefrequently
used; d) dilatometers, DMT, pressuremetersandplate
bearingtestsarerarelyused.
With respect to field monitoring, 81% of the
answersconfirmedthat thisisalwaysusedinpractice
or that it is frequently requested(19%). Theanswers
indicatedalsothefrequencyinusingdifferent instru-
ments(Figure10). Levelingsurfacesettlementpoints,
buildinglevelingandconvergencemeasurements are
performed in 100% of the cases. Deep settlement
points and lining levelings and the use of piezome-
tersorof waterlevel indicatorsareobservedin85%to
814
Figure10. Frequencyinusingfieldmonitoring.
90%of thecases. Inclinometers, loadcellsandstrain
or stressmetersarelessfrequentlyused.
6 SATISFACTIONWITHTHE PRACTICE
Finally, thesurvey investigatedtheareaswhereprac-
titioners believe that there are needs for technical
developments to be introduced in their practice and
areas they aresatisfiedwiththecurrent practiceand
new developments are less needed (Figure 11). The
highest levels of satisfaction refer to available con-
structionmethodsandliningtypes.Theareawithleast
satisfactionreferstothepredictionof porewater pres-
sure and water flow. Practitioners are also unhappy
withtheavailablewater proofingtechniquesandwith
theavailablemethodsfor stabilityanalysis.
7 CONCLUSIONS
Theconclusionsfoundrefertothescenarioof theprac-
ticeinBrazil, definedinbroadterms by therespon-
dents: large size tunnels, with equivalent diameter
larger than6m, designedfor rails, metros, highways,
drivenunder mixedfacecondition, incohesivesoils,
below water table using sprayed concrete as lining
(NATM).
Steel liner plates are not much in use, and are
being replaced by wet mix sprayed concrete or by
cast in placeconcreteas secondary lining. This may
beexplained by thehigher risks of corrosion of the
former inurbanenvironments, butattentionshouldbe
Figure11. Satisfactionwiththecurrent practice.
paidtoproblemsrelatedtoalkali-silicareaction(ASR)
frequentlyoverlookedinconcreteliningdesign.
It was also noted that the old practice of using
unreinforced concrete secondary lining is declining,
possiblyrevealingageneral perceptionthatthedesign
shouldensureanadequateductilitytotheconcretelin-
ing, renderedbythesteel reinforcement. Ontheother
hand, itshouldbearguedwhatistheimpactof thesteel
reinforcement in thelining safety when subjected to
thehightemperaturesof afire.
A considerableincreasein theuseof water proof
membraneshasbeenobservedwhereasuntil recently,
the control of water ingress through the lining was
exerted only by using low permeability concrete.
Onewonders if this changemay not befollowed by
lesseningof theconcretequalityinthefuture.
815
Itisstill highthefrequencyintheuseof limitequi-
libriummethods and of empirical methods to assess
thestability of thetunnel headingandface, thoughit
isreckonedthat theseapproachesmayprovideunsafe
estimates.
Further, anunjustifiedpostureandcriterionregard-
ingtheassessmentof stabilityof aNATM andaTBM
driventunnelswerenoted. For thelatter, it lookslike
that thetunnel designer delegatestothecontractor or
to themachinery manufacturer, theresponsibility to
ensurethetunnel facestability and, in general, they
accept it, inasmuch they may not havethetechnical
abilitiestojudgewhatisthebestoperational procedure
of theequipment at hand, for acertain geotechnical
scene. Thissubtleomissionof thegeotechnical engi-
neer is technically and professionally noxious. The
studyof casesof frequentinstabilitiesobservedinEPB
shieldsmayput somelight onthisissue.
The use of numerical methods (FE or FD) is
widespread in thelocal practicefor settlements esti-
matesor for liningdesign. Equally widespreadisthe
assumption of reduced ground stresses in 2D tunnel
models. However, it isunjustifiedthepersistingprac-
ticeof usingTerzaghisarchingtheorytoestimatethat
reductionongeostaticstresses, whichisdoomedtobe
put asideinthepracticeof urbantunnels.
It is intriguing to observethat practitioners either
donot takeintoconsiderationtheeffect of porewater
changes in the soil behaviour or account only the
water pressure over the lining. FE numerical analy-
sis couplinggenerationof porepressureandstresses
areinfrequent.
The most used constitutive models in numerical
analysis are those with the elastic plastic behaviour
with Mohr-Coulomb failure criterion. These models
whenappliedto tunnels built withlimitedrelaxation
of groundstresses, resultingfromrestrictiveconstruc-
tionmethods, usedto inhibit groundmovements and
associateddamages inaurbanscene, result inlinear
elasticsoil responses, withinhibitedor limitedplastic
zones. Thisisnot inlinewiththereckonednon-linear
behaviour of soils even at small straining: thereis a
clear need of using more adequate soil modeling in
thelocal practice.
Theuseof three-dimensional modelinginpractice
isyet limited, but it isbelievedthat theincreaseduse
of thistypeof modeling, alreadynotedintheacademy,
will beseeninlocal practicesoon.
A poor practice in geotechnical investigation for
tunnel design was noted. It tends to be limited to
soundingswithSPT blowcountsandsimplelabtest-
ingindeformedsoil samples. Thismaybesobecause
wearefocusingurbantunnelsinenvironmentswhich
are, in general, well known both in geological and
geotechnical terms, for which sizeable data bank is
available. Perhaps a way to compensate such defi-
ciency is to stimulatetheuseof in situ testing such
asthepressuremeters, thedilatometersandothers.
A favourable situation is seen in the practice, in
terms of field monitoring. Field instrumentation is
always present. The noted deficiency is the lack of
measurementsof liningloads. Thereareanumber of
robust andreliableprocedurestoassessloadsincon-
cretelinings, usingstressreleasetechniques(see, for
instance, Negro, 1994) thatcouldbeusedatlowcosts.
The respondents believe that the area in need for
further technical developments is that of pore water
pressureestimates andgroundwater flow. Moreover,
they are not satisfied with available techniques for
water proofing tunnels and for assessment of stabil-
ity analysis of tunnel heading and face. The survey
indicatesareasinthepracticewheretheacademymay
contributesignificantlyfor technologyadvancement.
ACKNOWLEDGEMENTS
Thisinvestigationhasbeenpossibleonlythankstothe
kindattentionof thosethat repliedthequestionnaire
sent.Theauthorisgrateful toall of themfortheefforts
andtimespent inreplying.
REFERENCES
Bjerrum, L. 1963. Discussion to European Conf. On Soil
Mech. Found. Eng. (Wiesbaden), Vol. II, p. 135.
Boscardin, M.D. &Cording, E.J. 1989. Buildingresponseto
excavation induced settlement. Journal of Geotechnical
Burland, J.B. & Wroth, C.P. 1974. Settlement of buildings
andassociateddamage. Conf. on Settlement of Structures,
London: 611654.
Durand, A.F., Vargas, E.A. & Vaz, L.E. 2000. SomeExper-
imentsin3D Numerical LimitAnalysis(NLA). 3rd Intl.
Conf. On Advances of Computer Methods in Geotech and
Geoenvironmental Engineering Geoecology and Comput-
ers, Vol. 1: 223227.
Mair, R.J., Taylor, R.N. & Burland, J.B. 1996. Predictionof
Ground Movements and Assessment of Risk of Build-
ingDamageduetoBoredTunneling. Proceed. Intl. Symp.
Geotechnical Aspects of Undeground Construction in Soft
Ground, London, UK: 713718.
Namba, M., Ruiz, A.P.T., Queiroz, P.I.B., Negro, A. &
Vasconcellos, C.A. 1999. Assessment of BuildingDam-
ages Dueto UrbanTunnelling. Proc. 11th Panam. Conf.
Soil Mech. And Geotech. Eng., Vol. 2: 549555.
Negro, A. 1994. Soil Tunnels and Their Supports. Special
Conferencetothe10th Brazilian Congress of Soil Mech.
Found. Engineering, Iguau Falls, Brazil, ABMS, Proc.
Vol. 5: 3360.
Negro, A. & Eisenstein, Z. 1997. Delayed Lining Activa-
tion and Ground Stress Relaxation in ShallowTunnels.
Proc. of The 4th Intl. Conference on Soil Mechanics and
Foundation Eng, Hamburg, Germany: 23912396.
Negro, A. & Leite. R.L.L. 1995. Design of Underground
StructuresinBrazil National Report onTunnellingand
BracedWall ExcavationinSoft Ground. Proc. Intl. Sym-
posium on Underground Construction in Soft Ground,
NewDelhi, India: 4956.
816
A studyonloadsfromcomplexsupport systemusingsimple2Dmodels
Z. Shi, W. Bao, J. Li, W. Guo& J. Zhu
China State Construction Engineering Corporation (SH), Shanghai, P.R. China
ABSTRACT: Indeeppit engineering, spatial displacementsarereadilyavailable, commonlyalonghorizontal
or vertical lines. Itstraditionally difficult toget strut loadsfromcomplex concretestrut systemduetolimited
scaleof instrumentation and uncertainties in field measurements. This results in problemin predicting wall
deformation where not instrumented. This paper proposes to utilize spatial wall displacements measured to
back-analyzesupportingloadsonthewall fromconcretestrutsthroughmodelingwall-soil system. Theanalysis
canbeappliedto eachstrut layer to obtainloads betweenwailingandwall inhorizontal plane. Theobtained
support loads at different levels arethen used as input in avertical section model, fromwhich deformation
profileof wall canbepredicted. Theapplicationisverifiedinproject caseandshowsclosecorrelationtofield
measurements.
1 INTRODUCTION
Reinforced concrete strut systemis widely adopted
with diaphragm wall for support of deep founda-
tion pits in soft soil ground. During excavation, the
diaphragmwall is supportedby concretestruts from
insideandsubject tosoil pressureontheother side. It
is typical for deepfoundationpits to haveafewlay-
ers of concrete struts at different excavation depths.
The interaction between diaphragm wall and con-
crete struts (through wailing), which is one of the
most important factors inanalyzingdeepfoundation
pits, presentsaverycomplicatedscenariounder loads
fromsoil.
Becausewailingisincontinuousdirectcontactwith
diaphragmwall, it isdifficult tomeasuretheinternal
loads between them. By direct method, the internal
forceof struts is commonly measured by embedded
loadcells. Thismeasurementisusuallyinsufficientto
estimatesupport loadsondiaphragmwall because:
1 The scale of field measurements of strut force
is limited; and the geometry of strut system is
complex. Therefore, it is difficult to derive reli-
ablesupport loads on diaphragmwall or through
complexstrut systembymechanical analysiswith
limitedstrut forcemeasurements.
2 Field measurement of strut force could be devi-
ated by other factors such as creep, contraction
andthermal stressof concretematerial (Xia& Li,
1999). Zhao (1996) reported significant thermal
stressmeasuredinconcretestruts.
Therefore, indirect methods are widely used for
analyzingthesupport loads. Back-analysis fromdis-
placement is oneof thecommonly used. By indirect
method, therehavebeennumerous2Dand3Dstudies
ontheanalysisof suchsupportsystem(Li &Hu, 1995;
Zhao, et. al., 1996).
Incontrarytothedifficultyof direct measurement
of support force, spatial displacement informationof
diaphragmwall is more readily available and more
reliable by field measurement. Therefore, wall dis-
placementmeasurements, for exampleinplansection
at astrut level, canbeusedtoback-analyzeloads on
diaphragmwall byasimplewall-soil model.
During excavation, the diaphragm wall deforms
subject to the soil pressure on one side and support
loadsontheother side. Thesoil pressurecanbeesti-
mated by empirical method. Therefore, for asystem
consisting of supports, diaphragmwall and soil, the
measureddisplacement patternof diaphragmwall on
theplancanbeusedtoback-analyzesupportingloads
on thewall fromconcretestruts. It is recommended
tocarryouttheback-analysisinplansectionsincethe
vertical sections normally involve much more com-
plexity associatedwithconstructionsequenceetc. In
planof eachstrut layer, asimple2Dmodel caneasily
represent suchcasescenario. Thiscanberepeatedfor
eachlayer of strut. Therefore, support loads at asec-
tioncanbeback-analyzedfromspatial displacement
informationfromfield.
Basedonacasehistory, thispaper presentsasim-
plemethodof estimatingsupport loadsondiaphragm
817
wall bywailings.Theestimatedsupportloadsarethen
verifiedinmodelsfor vertical crosssection.
2 ANALYSISOF SUPPORT LOADSINPLAN
2.1 Assumptions and simplifications
Ontheplansectionof eachstrut layer, thesystemof
diaphragmwall andsoil resemblesaplainstrainsce-
narioatthecenterelevationof wailing.Thediaphragm
wall deforms under thesupport loads and soil pres-
sure after excavation. The sidelines of pit, which in
mostcasesarestraight, canbesimplifiedasanelastic
continuouswall, i.e. acontinuousbeamonplan. The
deflection pattern of thebeamcan beeasily defined
by displacement measurement along the side and is
takenastarget deformationinback-analysis. Thetwo
endsof thesidelinecannormallybedeemedaspinned
ends, asshowninFigure1.
For simplicity, followingassumptions aremadein
theanalysis:
Onplansection, thediaphragmwall ismodeledas
elasticcontinuousbeam.
Theloadsbetweenthewailinganddiaphragmwall
areevenly distributedalongvertical direction but
varyalonghorizontal directionfor eachstrutlayer.
Onplanof eachstrut layer, thesupport loadsfrom
wailingaresimplifiedaspoint loadsalongthepit
side.
Invertical plan, thesupportloadsof eachstrutlayer
areinvariableduringconstruction.
Theinitial stresslevel of soil inthemodel iscalcu-
latedfromitsdepthfollowingRankinsactiveearth
pressureformula.
Intheplanmodel, pointloadsequivalenttotheini-
tial stress level prior to excavation are applied onto
thebeam(diaphragmwall) sothat themodel isinini-
tial balancewithzerodisplacementindiaphragmwall.
Inback-analysis, theappliedpoint loadsareadjusted
frominitial valuesinacontrolledmanner till that the
diaphragmwall deforms approximately to thetarget
displacementpattern,whichisdefinedfromfieldmea-
surements. Theresult point loadscanbeconvertedto
internal stress betweenthediaphragmwall andwail-
ing(Fig. 1(d)) alongthelengthof sideline.Thiscanbe
repeatedfor eachstrutlayer withoutmajor changesto
themodel. Inthisway, thesupport loadsof eachstrut
layer canbefound.
Inthemodel foravertical crosssection,analysiscan
becarriedouttosimulatetheconstructioncasescorre-
spondingto planmodels. Thepredicteddeformation
pattern of diaphragmwall can checked against field
measurements and the quality of estimated support
loadscanbeverified.
A historycaseof adeepfoundationpit insoft clay
inShanghai areaisselectedfor theapplicationof the
Figure1. Plansection:(a)Planof strutlayout;(b)Lineloads
ondiaphragmwall fromwailing; (c) Convertedpoint loads;
(d) Calculatedsupport stressondiaphragmwall.
method. Theprogramusedinthestudy is theFLAC
byItascaCG(Itasca, 1997).
2.2 Model conditions and material properties
Inthecasehistory, onesidelineof thepitwithalength
of about 100misselectedfor application(Fig. 1(a)).
The initial stress level is calculated according to
Rankins activepressurewith asurchargeof 20kPa,
as shown in Figure 2 (b). The excavation process is
dividedintofour stepsequalingtothenumber of strut
layers(Fig. 2(b)).
818
Figure2. Vertical sectionandsimulationconditions.
On each plan section, the loads from wailing
on diaphragm wall are simplified as five point
loads (F
h1
F
h5
) at the elevation of each strut layer
(Fig. 1(c)). Thetargetdisplacementsfor eachsupport
layer (S
1
S
i
) are taken at the end of excava-
tion at time t
k
. The back-analyzed support forces
(F
h1
F
h5
) represent the site condition at the end
of excavation.
The geometry of the plan model is shown in
Figure3.Thesoil layersarelistedinTable1withmate-
rial properties. The soil is modeled as elasto-plastic
material withMohr-Coulombstrengthcriteria.
The reinforced concrete diaphragm wall reaches
41mfromgroundsurface, modeledaselasticcontin-
uousbeam.
2.3 Analysis of support loads at strut layers
There are five monitoring points available for each
layer withfieldmeasurementexceptfor thefirstlayer
as listedinTable2. Dueto availability of fielddata,
Figure3. Model meshandset upfor plansections.
therearemeasurementsforthreemonitoringpointsfor
thefirst strut layer.
Accordingtoexcavationhistory, thecorresponding
excavation steps for each strut layer are defined as:
up to 2.8mfor thefirst layer; 6.3m(2.8m8.3m)
for the second layer; 3.7m(8.3m12.0m) for the
thirdlayer; and4.0m(12.0m16.0m) for thefourth
layer. Theinitial stressconditionfor eachstrutlayer is
determinedinthesamewayasabove(Fig. 2(b)).
Toreducenoiseduringback-analysis, onlythefield
measurements of thecenter threepoints (Q10Q12)
are set as target deformation because the other two
monitoring points (Q9 and Q13) could be easily
affectedbyfixedendsof diaphragmwall. Thematch-
ing criteriafor model prediction is set as within 5%
of fieldmeasurement, i.e. themodel reachesitstarget
conditionwhenthemodel predictionsateachselected
points arewithin5%differencefromfieldmeasure-
ments. Under this condition, thepoint loads on wall
aremeasuredfromthemodel.
Theresultdeformationsof back-analysisareshown
inTable3withcorrespondingcalculatedpoint loads
of support inTable4.
3 VERIFICATIONOF CALCULATEDSUPPORT
LOADSINVERTICAL SECTION
3.1 Models of vertical section
Twovertical crosssectionsintersectingthediaphragm
wall at Q10 and Q12 are selected to verify the
calculatedsupport loads.
Thevertical sectionmodel hasatotal widthof 120m
with40monthesideof thepit asshowninFigure4.
Theheight of model is80mfromthegroundsurface.
Thestrut levels andexcavationsteps areshownin
Figure2. Themodel is prescribedwithinitial lateral
stress conditions as calculated above. The two sides
arefixed only in horizontal direction. Thebottomis
fixedinbothdirections.
Consideringthehighstiffnessandlimitedheightof
wailing, theloadsof wailingareassumedevenly dis-
tributedalongitsheight at eachstrut layer. Therefore
in plan model, thesupport loads can berepresented
819
Table1. Material propertiesof soil anddiaphragmwall.
Thickness UnitWeight Modulus Poissons Cohesion Friction
No Soil Layer (m) (kN/m3) (MPa) ratio (kPa) (
)
12 TopSoil 3.5 18.3 11.25 0.35 24 15.5
3 MudSiltyClay 3 17.4 5 0.35 15 20
4 MudClay 9.9 16.6 2.75 0.4 12 11.0
5
1a
GreyClay 4.2 17.5 5 0.35 18 11.5
5
1b
GreySiltyClay 7.2 17.9 12.5 0.35 17 20.5
5
1c
SandedSiltyClay 10.4 17.9 25 0.3 16 23.5
6 SiltyClay 2.4 19.9 36 0.35 51 24.5
DiaphragmWall 1.0 20.0 30000 0.17
Table2. Fieldmeasurementsof deformationat strut layers
(mm).
Q9 Q10 Q11 Q12 Q13
Layer 1 0.1 0.3 0.2
Layer 2 15.9 24.3 23.7 19.6 13.2
Layer 3 32.3 52.4 49.0 43.5 23.3
Layer 4 42.0 62.7 59.9 48.0 33.2
Table3. Calculateddeformationsat strut layers(mm).
Q9 Q10 Q11 Q12 Q13
Layer 1 0.1 0.3 0.1
Layer 2 24.3 24.0 19.6
Layer 3 52.7 50.1 44.1
Layer 4 63.0 59.1 48.8
Table 4. Calculated support point loads at strut layers
(kN/m).
Q9 Q10 Q11 Q12 Q13
Layer 1 150 106 120 136 108
Layer 2 89 34 80 39 170
Layer 3 220 50 220 70 382
Layer 4 310 150 320 200 540
by point loads at thecenter elevation of wailings as
illustratedinFigure2(c).
The calculated point loads fromplan models are
converted to lineloads along horizontal direction at
all strut layers, wherethevaluesat Q10andQ12can
beobtained(Fig. 1(d)) for useinvertical model. The
valuesof point loadarelistedinTable5.
Incorrespondencetothecalculatedsupport loads,
thefollowingconstructioncases aresimulatedinthe
planmodel:
Case1: Initial conditionbeforeexcavation; instal-
lationof diaphragmwall;
Figure4. Overviewof model for vertical sections.
Table5. Convertedsupportloadsforvertical model (kN/m).
Layer 1 Layer 2 Layer 3 Layer 4
Section(Q10) 318 136 200 300
Section(Q12) 408 156 280 400
Case2: Excavateto2.8m; apply support loadat
first strut layer;
Case3: Excavateto8.3m; apply support loadat
secondstrut layer;
Case4: Excavateto12.0m; applysupportloadat
thirdstrut layer;
Case5: Excavateto16.0m; applysupportloadat
fourthstrut layer.
3.2 Results of vertical section models
Thedeformationpatternsof diaphragmwall at target
pointsareverysimilar asQ12showninareshownin
Figure5.
820
Figure 5. Horizontal deformation pattern (mm) of
diaphragm wall along depth (m) in vertical section
throughQ12.
Thecalculateddeformationsof diaphragmwall are
compared with field measurements in Figure5. The
maximumcalculateddisplacement is 40.5mmwhile
51.1mm in field measurement on vertical section
throughQ12. It isnotedthat all casesshowgoodcor-
relationof deformationpatternbetweenmodel calcu-
lationandfieldmeasurementasshowninFigure5(b).
4 DISCUSSIONSANDCONCLUSIONS
Byusingspatial displacementmeasurementsinsimple
2D models, thedifficulties associated with complex
supportsystemanduncertaintiesinloadmeasurement
areavoidedwith reasonablecorrelation to fielddata
invertical sections.
However,thedisplacementinvertical sectionshows
notable discrepancies on magnitude, which is partly
a result of simplifications made in the analysis. In
the system of diaphragm wall and soil, the calcu-
lated support loads are considerably affected by the
stress conditions in the model. The influence from
stress can be minimized if soil stress measurements
areincorporatedinthebackanalysis.
The discrepancies in prediction of vertical model
could also result fromtheassumption that thestress
releaseduetoanexcavationstepisundertakenentirely
by aspecific strut layer. Thereforethesupport loads
can only becalculated for limited cases in thesame
number of excavationstepsasstrut layers.
However, themethodproducesreasonableestima-
tionof supportloadsthatleadtoacceptableprediction
of deformationpatternof diaphragmwall. Thiswould
behelpful inmanyengineeringcases.
ACKNOWLEDGEMENT
TheAuthor would like to acknowledge great assis-
tanceandsuggestioninpreparationof thispaper from
Dr Dongmei ZhangfromTongji University.
REFERENCES
Bolton, M. &Powrie,W. 1988. Behaviorof DiaphragmWalls
in Clay Prior to Collapse. Geotechnique, 1988, 38(2):
167189.
Li,Y. & Hu, Z. 1995, DeformationandLoadsof Diaphragm
Wall for Deep Foundation Pits. Architectural Construc-
tion, Vol 17, 1995(3).
Itasca, C.G. 1993. Fast Language Analysis of Continua.
ItascaConsultingGroupInc.
SIDA, 1997. Shanghai Standard Codefor Design of Exca-
vationEngineering(DBJ 08 6197).
Xia, C. &Li,Y. 1999. MonitoringTheoryandTechnologyin
UndergroundEngineering. Tongji UniversityPress.
Zhaoet al. 1996. PracticeandStudy of Support Systemfor
DeepFoundationPit. Tongji UniversityPress.
821
Groundreactionduetotunnellingbelowgroundwater table
Y.J. Shin
Purdue University, West Lafayette, USA
J.H. Shin
Kon-Kuk University, Seoul, Korea
I.M. Lee
Korea University, Seoul, Korea
ABSTRACT: Tunnelling below the groundwater table influences the hydraulic regime in the surrounding
ground.Thiswill, inturn, causeseepageintothetunnel throughtheporesanddiscontinuitiesandoftenproduces
a long-terminteraction between thetunnel and theground. In this paper an attempt is madeto identify the
behavior of surroundinggroundduetoseepage, andgroundreactioncurves(GRC) consideringseepageforces
arepresented for atunnel under drainageconditions using an analytical method. It is found that theflowof
groundwater hasasignificant effect ontheradial displacement of atunnel wall. Whiletheeffectiveoverburden
pressureis reducedby thearchingeffect duringtunnel excavation, seepageforces still remain. Therefore, the
presenceof groundwater induceslargeradial displacement of thetunnel wall comparedtodryconditions.
1 INTRODUCTION
Tunnelling below the groundwater table affects the
hydraulicequilibrium.Thiswill,inturn,causeseepage
into thetunnel through thepores and discontinuities
in the ground and often produce long-terminterac-
tion between the tunnel and the ground. The effect
of seepage on the tunnel is initially reflected in the
ground loading, transmitted through theground and
then, eventuallyappliedtothetunnel resultinginaddi-
tional stressesinthelinings. It wouldbeappropriate,
therefore, toincludetheinfluenceof seepageforcein
estimatinggroundbehavior duetotunnelling.
Thegroundbehavior duetotunnellingcanbeindi-
catedtheoreticallybythegroundreactioncurvewhich
showstheincreasingtrendsof radial displacement as
theinternal pressureof thetunnel decreases.
Inthispaper,thetheoretical solutionof groundreac-
tioncurveconsideringseepageforcesduetoground-
water flowunder steady-stateflowwas derived. The
studieswereperformedfor anon-supportedcondition
aswell asasupportedconditionwithshotcretelining.
2 GROUNDREACTIONCURVE WITH
SEEPAGE FORCES
2.1 Theoretical solutions
It isassumedthat asoil-massbehavesasanisotropic,
homogeneous and permeable medium. Also, an
Figure 1. Elasto-plastic model based on Mohr-Coulomb
yieldcriterion.
elasto-plasticmodel basedonalinear Mohr-Coulomb
yieldcriterionisadoptedinthisstudy, asindicatedin
Figure1:
where,
1
indicatesthemajorprincipal stress,
3
isthe
minor principal stress,
k and a are the Mohr-Coulomb constants, c is the
cohesion, and isthefrictionangle.
Figure2 shows acircular opening of radius r
0
in
an infinite soil-mass subject to a hydrostatic in situ
stress,
0
.Theopeningsurfaceissubjecttotheoutward
radial pressureto thetunnel surface, p
i
. Considering
823
Figure2. Circular openinginaninfinitemedium.
Figure3. Bodyforcesunder thegroundwater table.
all the stresses on an infinitesimal element abcd of
unitthicknessduringexcavationof acirculartunnel in
Figure3, equilibriumequationcanbeexpressedas:
If the tunnel is excavated under the groundwater
table, thenit actsasadraininunsupportedcondition.
Thebody forceis theseepagestress, as illustratedin
Figure3:
Inthis state, i
r
andi
arethehydraulic gradient in r
and directionsrespectively, and
w
istheunitweight
of groundwater.
If thestressdistributionissymmetrical withrespect
totheorigin(O) inFigure3, thenthestresscompo-
nents arenot varied with angular orientation, , and
therefore they are functions of the radial distance r
only. By putting Equation (4) into Equation (2) the
equilibriumequationreducestothesingleequationof
equilibriumasfollows:
By substituting
1
and
r
=
3
in Equation (1)
andbyputtingEquation(1) into(6), Equation(6) can
begivenasfollows:
Theabovepartial differential equationcanbesolved
by using the boundary conditions
r
=p
i
at r =r
0
.
Then, theradial andcircumferential effectivestresses
intheplasticregionareasfollows:
Inthisequation,p
i
isall thesupportpressuredeveloped
by in situ stress and seepage. Subscripts rp and p
indicatetheradial andtangential effectivestresses in
theplasticregionrespectively.
Inordertoestimatetheeffectivestressintheelastic
region, thesuperpositionconceptisused.Asshownin
Figure4, theeffectivestress consideringtheseepage
forcecan beassumed as a combination of thesolu-
tionof equilibriumequationindry conditionandthe
effectivestressonly consideringseepage. TheKirsch
solutionsareappliedtosolvetheeffectivestressesin
the elastic region under dry condition (Timoshenko
and Goodier, 1969). And the solution proposed by
Stern(1969) isadoptedtoobtaineffectivestressesin
elasticregionwithconsiderationof seepageforces.
ThefollowingEquation(10) isderivedbycombin-
ingtheKirshsolutionswiththeSternssolutioninthe
elasticregionasfollows:
824
Figure4. Concept of superpositioninelasticregion.
where
Finally, attheinterfacebetweentheplasticandelas-
ticregions,r =r
e
asshowninFigure2,theradial stress
calculatedintheplasticregionmustbeidentical tothat
intheelastic region. ThefollowingEquation(11) of
the radius of the plastic zone, r
e
can be derived as
follows:
The radial displacement for a circular tunnel can
beworkedout basedontheelasto-plastic theory. By
following the same procedure proposed by Sharon
(2003), theexpressionfor theradial displacement in
theplasticregioncanbeobtainedasfollows:
where,
Theradial displacementu
r(r=r
0
)
attheopeningsurface
r =r
0
isgivenbyEquation(13).
Figure5. Theexampletunnel for seepageanalysis.
Figure 6. The ground reaction curve (C,D=10,
H,D=10).
Thetheoretical solutionshownabovewasverifiedby
performingnumerical analysisandcomparingthetwo
analysisresults(Shin, 2007).
2.2 Example problem
As defined in previous section, the ground reaction
curve, GRCistherelationshipbetweenthedecreasing
internal pressure, p
i
andtheincreasingradial displace-
ment, u
r
. Givenavalueof p
i
, r
e
canbecalculatedusing
Equation(11), then, theu
r
valueusingEquation(13).
As shown in Figure 5, the ground reaction curve
is calculatedfor asamplecircular tunnel withdiam-
eter of D=5m, thecover depth of C =50m, under
the ground surface, and the groundwater table of
H =50m, above the tunnel crown. As shown in
Figure 6, the ground reaction curve with considera-
tionof seepageforceisbiggerthanthegroundreaction
curveindry condition, whichmeans that thereis no
ground water while the cover depth of tunnel, C is
10 times diameter of tunnel, D. This is due to the
fact that even if the effective overburden pressure
825
Figure 7. The distribution of water pressure in fully and
restricteddrainedcases.
canbedecreasedby thearchingeffect duringtunnel
excavation, seepageforcesstill remain.
2.3 Theoretical solution for lined tunnel when
including consideration into seepage forces
For areal tunnel, thefirst lining(suchas shotcrete),
isinstalledduringtunnelling, andthepermeabilityof
this material is relatively lower than that of ground.
Thetheoretical solutionwhenconsideringseepageis
obtainedbyassumingthat thetunnel actslikeadrain.
However, theinstallationof liningmakestheground-
water flowchangefromafullydrainedconditiontoa
partlydrainedcondition.
Whentheinstallationof theliningisconsidered, the
distributionof water pressureinthegroundisshown
inFigure7. Thewater pressurecausedby tunnelling
can bedifferent according to theratio of permeabil-
ities of thesoil and thetunnel lining. In caseof the
fully drained condition, thewater pressureincreases
withdepthtoacertaindepthbutitconvergessmoothly
to zero, u
w
=0 at tunnel wall. On theother hand, if
thedrainedconditionischangedtoapartiallydrained
condition due to installation of the lining, then the
water pressure increases with depth like the hydro-
static pressure, but it convergesrapidly tozeroat the
insideof thetunnel wall. It developsaresidual stress
onthesurfaceof lining, thereforeeventually induces
a bigger seepage force adjacent to tunnel wall than
thatof fullydrainedcondition. Forthedevelopmentof
residual waterpressure, alargetotal headlossisdevel-
opedwithinrelativelyshortdistance(i.e., thicknessof
lining).
Whenconsideringtheinstallationof aliningwitha
relativelylower permeabilitythanground, theground
water flow is changed as shown in Figure 7. This
inducesthedevelopment of aresidual water pressure
onthesurfaceof liningandmakesthehydraulicgradi-
entvary. Thismeansthati
r
inEquation(4) ischanged
toi
r
consideringthelininginstallation. Figure8shows
thevariationof hydraulicgradient astheresult of the
installationof thelining.Therefore, Equation(11) and
Figure8. Thevariationof hydraulic gradient duetolining
set up.
(13) canberewrittenwiththeconsiderationintolining
installation as follows, wherer
el
is theelasto-plastic
interfaceconsideringlininginstallationunderground-
water table, assumingthestiffness of theshotcreteis
thesameasthat of theground.
The ground reaction curve is estimated through
theoretical solutionwhenconsideringalininginwhich
the permeability is one thousandth of the soil per-
meability, k
l
,k
s
=0.001, under thegroundwater table.
Thegeometryissameaspreviousone. Beforethelin-
ingisinstalled, thegroundreactioncurveisthesame
as thefully drainedcurve. Then, after installationof
thelining, aresidual water pressuredevelops on the
surface of lining. It induces an increase of seepage
forces near thetunnel wall, thus thegroundreaction
curve increases as shown in Figure 9. Accordingly,
it is knownthat thepermeability andtheinstallation
timeof theliningis themost important factor inthe
behavior of theground.
3 THE SIMPLE METHODOF GROUND
REACTIONCURVE
3.1 Estimation of equation of hydraulic gradient
In this study, to evaluatethetermrelated to thesee-
pageinthetheoretical solution, thehydraulicgradient
shouldbeestimatedinadvanceunder conditionscor-
responding to the existing geometry. The seepage
analysis is conducted for a sample circular tunnel
826
Figure9. Thegroundreactioncurveconsideringliningset
upunder groundwater table(C,D=10, H,D=10).
Figure 10. Generalized hydraulic gradient distribution in
exampletunnel.
showninFigure5. Theradial componentof hydraulic
gradient, i
r
as estimated by the seepage analysis, is
normalized by squareof thedepth fromtheground-
watertabletothetunnel, z
2
asthesampletunnel shows
inFigure10, wherez isunit of [m], thusunit of i
r
,z
2
becomes[m
2
].If thedistributionof thehydraulicgra-
dientisexpressedasahyperbolicfunction, thenitcan
beeasily integratedinto thetheoretical solution. The
equation of distribution of hydraulic gradient shown
inFigure10isasfollows:
Figure 11. The variation of parameter A for water height
andtunnel diameter.
where, i
r
=theradial componentof thehydraulicgra-
dient, z =thedepthfromthegroundwater tabletothe
tunnel (ex. crown, shoulder, invert), r =thedistance
fromthe center of the tunnel, and D=the diameter
of thetunnel. In general form, Equation (16) can be
expressedas
where, A andB areparametersdependingonC,D and
H,C. Shin(2007) performedparametricandsensitiv-
ity analysis to assess thetwo parameters. Theresults
of hisanalysisareasfollows. Firstly, theparameter B
canbeassumedasaconstant of onethroughthesen-
sitivityanalysis. Inother words, theestimationof the
hydraulicgradientcanbereducedtotheproblemof the
onlyparameterA.Thirdly, whenthegroundwaterlevel
ishigherthancoverdepth(H,C >1), theparameterA
decreases as thegroundwater level increases. Onthe
contrary, whenthegroundwater level islower thanthe
cover depth, theparameter A increasesaswater level
increases.
Also, when the diameter of tunnel increases, the
parameter A tends to decrease. Thus, if thediameter
of tunnel, groundwaterlevel andcoverdeptharegiven
inany geometric condition, thenormalizedequation
of hydraulic gradient can be estimated by using the
tendency establishedabove. This canbeusedtoesti-
matethehydraulicgradient necessarytocalculatethe
ground reaction curvewithout aseepageanalysis in
apreliminary design. Figure11isthediagramof the
parameter A including various conditions; it can be
usedtoestimatetheequationof hydraulicgradient in
anygeometriccondition.
827
3.2 Simplified theoretical solution with
consideration of seepage forces
As mentioned before, the equation of normalized
radial hydraulic gradient canbepredicted. Theequa-
tion of normalized radial hydraulic gradient can be
expressedasfollows:
where, C =A
z
2
d
Therefore, theelasto-plastic interfaceand radial dis-
placement canbewrittenasfollows, wherer
ecs
isthe
simplifiedelasto-plasticinterfaceconsideringseepage
forces.
4 CONCLUSIONS
Thegroundwaterhasasignificanteffectonthebehav-
ior of tunnel. This is dueto thefact that even if the
effectiveoverburdenpressurecanbedecreasedbythe
archingeffectduringtunnel excavation,seepageforces
still remain. Theresultsobtainedfromthisstudy can
besummarizedasfollows:
1 The flow of groundwater has a significant effect
ontheradial displacement of atunnel wall. While
the effective overburden pressure is reduced by
the arching effect during tunnel excavation, see-
pageforcesstill remain. Therefore, thepresenceof
groundwater induces thelargeradial displacement
of tunnel wall comparedtodrycondition.
2 When theshotcretelining (with arelatively lower
permeabilitythanthat of ground) wasinstalled, the
residual water pressureoccursonthesurfaceof the
lining.Theseepageforcesnear tunnel wall increase
and consequently the radial displacement of tun-
nel increases, too under the assumption that the
shotcretehasthesamemechanical propertiesasthe
surroundingsoil.
3 The hyperbolic curve (i
r
=Cr
1
) that can esti-
matethedistribution of hydraulic gradient dueto
tunnelling under groundwater table is suggested
throughparametricstudiesandsensitivityanalysis.
Usingit, thesimplifiedgroundreactioncurvecan
beachievedwithout performingseepageanalysis.
ACKNOWLEDGEMENTS
This paper was supported by the Korea Institute of
ConstructionandTransportationTechnology Evalua-
tionandPlanningunder theMinistryof Construction
andTransportationinKorea(Grant C04-01).
REFERENCES
Sharan, S.K. 2003. Elastic-brittle-plasticanalysisof circular
openingsinHoek-Brownmedia, International Journal of
Rock Mechanics and Mining Sciences 40: 817824.
Shin,Y.J. 2007. Elasto-plasticgroundresponseof underwater
tunnels consideringseepageforces, Ph.D. Thesis, Korea
University, Seoul
Stern, M. 1965. Rotationally symmetric plane stress dis-
tribution, Zeitschrift fur Angewandte Mathematik und
Mechanik 45(No. 6): 446447.
Timoshenko, S.P. & Goodier, J.N. 1969. Theory of elasticity,
McGraw-Hill, NewYork.
828
Basal stabilityof bracedexcavationsinK
0
-consolidatedsoft clayby
upper boundmethod
X.Y. Song& M.S. Huang
ABSTRACT: Design of braced excavations in soft clays is usually controlled by theshort-termundrained
stability.Thepapertakesintoconsiderationtheprincipal stressaxial rotationinducedbyexcavationandundrained
anisotropicsoil strength.AssumingthePrandtl soil slipfailuremodesandconsideringtheanisotropicsoil strength
recommendedbyCasagrande&Carilloandthenon-homogeneousfeatureof softsoils, amethodfor evaluating
thebasal stability isproposedbasedontheupper boundanalysis. Resultsobtainedfromtheproposedmethod
indicatethat thebasal stability is significantly influencedby theanisotropy ratio of soil, as well as theplane
geometry of theexcavationandthethicknessof soft soil layer betweenexcavationbaseandhardstratum. For
fieldcasestudies, theproposedmethodistestifiedbyfieldobservationsaswell asfiniteelement methods.
1 INTRODUCTION
For deepexcavationsinsoft clay, designof thelateral
earth support systemis often controlled by stability
requirements. In current practice, basically thereare
threemethods availablefor performing stability cal-
culationsof bracedexcavations: (1) limit equilibrium
methods; (2) displacement-basedelastoplastic finite-
element methods; (3) upper and lower bound limit
analysis. The limit equilibriummethods are widely
usedindesignpracticeandincludeseparatecalcula-
tions of basal stability (basedonfailuremechanisms
proposedbyTerzaghi 1948; BjerrumandEide1956)
or overall slopestability(usingcircular or noncircular
arc mechanisms) based on well established methods
(MorgensternandPrice1965; Bishop1966; Spencer
1967). It is often difficult to assess the accuracy
of these solutions due to ad hoc assumptions: (1)
in selecting the shape of the failure surface; (2) in
thesearch procedures used to locatethecritical sur-
face; and(3) intheapproximations usedtosolvethe
equilibriumcalculations(Ukritchonet al., 2003). Fur-
ther complications arise in analyzing soil structure
interactions for embedded supported walls, tieback
anchors, etc.
Displacement-based elastoplastic finite-element
methodsprovidesacomprehensiveframeworkthatcan
evaluate multiple facets of excavation performance
rangingfromthedesignof thewall andsupport sys-
tem, to thepredictionof groundmovements, andthe
effects of construction activities such as dewatering,
groundimprovement, etc. They areindispensablefor
predictingthedistributionof groundmovementcaused
byexcavations, andforsimulatingprocesswherethere
ispartial drainagewithinthesoil. Excavationstability
isusuallyassessedbyfactoringthestrengthparameters
of thesoil.
A powerful methodof calculatingstabilityisbased
on theupper and lower bound theory. Chang (2000)
analyzedthesafety factor of basal stability basedon
thePrandtl slidefailuremodeusingupper boundthe-
ory. Ukritchonet al.(2003) formulatedthenumerical
limit analysis by assuming theanisotropic yield cri-
terion proposed by Davis and Christian (1971). Zou
(2004) assumedakindof complex velocity fieldand
analyzed the basal stability without considering the
anisotropicfeatureof K
0
consolidatedclay.
However, the undrained shear strength of natural
K
0
-consolidated clay is anisotropic and has a close
relationwiththevertical effectivestress.Theclassical
rotationof principal stress directionof clays induced
by excavation is illustrated in Figure 1 (Clough and
Hansen, 1981) and it is well-known that the undri-
anedshear strengthof clayvarieswiththeangleof the
majorprincipal stressreorientationduringtheloading.
Someresearchers(J ianget al., 1997; Suet al., 1998;
Ukritchonet al.,2003)pointedoutthatwithoutconsid-
eringtheanisotropicandnon-homogeneousfeatureof
claystheresultof basal stabilityisnotrealisticandthe
safetyfactor maybemuchsmaller thanthereal one.
Inthis paper, anupper boundanalysis methodfor
evaluating the basal stability of deep excavations in
829
Figure1. Typical rotationof principal stress.
soft clayisintroducedbytakingintoconsiderationof
anisotropicfeatureof K
0
consolidatedsoftclaysinthe
world, especiallyinShanghai, aswell astheundrained
shear strengthvariationwithvertical effectiveconsol-
idation stress. Practical application of the proposed
method is demonstrated through case studies using
datapublishedinliterature.
2 UNDRAINEDANISOTROPIC SHEAR
STRENGTHOF K
0
-CONSOLIDATED
SOFT CLAYS
2.1 Undrained shear strength obtained from
triaxial tests
Wei andHuang(2006) presentedthedetailedformu-
lationof aconstitutivemodel for soft clayswhichcan
consider theanisotropicfeatureof claysbasedonthe
bounding surfaceplasticity. In thecaseof theshape
parameter R=2, thevolumetricstrainratecanbefor-
mulated fromthe yield function and hardening rule
whichcanbeexpressedasfollows.
where =q,p =stressratio; e
0
=initial voidratio; ,
=theslopesof virginconsolidationlineandswelling
line respectively in the e ln p space; M=critical
stress ratio; and =slope of yield surface in p q
space.
Under thetriaxial undrainedcondition, i.e.
v
=0,
equation(1) canberewrittenas
By integrating both sides of equation (2), the
undrainedstresspathcanbeexpressedas
Table 1. Comparison between authors and measured
results.
Test Undrainedshear Test value Theoretical
method strengthexpression (Ladd, 1973) value
K
0
UC eq. (7) 0.330 0.367
K
0
UE eq. (8) 0.155 0.140
Theundrainedlimit shear strengthis theintersec-
tion of undrained stress path and critical statestress
line. If let =M, theundrained limit shear strength
obtainedfromcompressivetest canbeexpressedas
Similarly, if let = M, theundrainedlimitshear
strength obtained fromtriaxial extensivetest can be
expressedas
where p
c
=mean effective stress; =the second
invariantof anisotropictensor:
ij
. Intheanalysis, the
initial valueof
ij
isdeterminedbytheinitial consol-
idationstateof thesoil. For theK
0
-consolidatedclay,
p
c
canbeexpressedas
where
v0
=vertical effectiveconsolidationstress.
Substituting equation (6) and S
u
=q
ult
,2 to equa-
tions (4) and (5) respectively leads to the following
expressionsas
and
where S
uv
=undrained shear strength obtained from
K
0
UEtriaxial test; andS
uh
=undrainedshear strength
obtainedfromK
0
UC triaxial test.
Table1 shows thecomparison between theafore-
mentioned formulas and experimental data from
Ladd(1973), which testifies that the current expres-
sion of undrained strength for soft clays is suitable.
AccordingtoJ ianget al. (1997), 1, =0.76. Other
input parametersareasfollows:
K
0
=0.5,
=33
, =3(1K
0
),(1+2K
0
)=0.75,
M =6sin
,(3sin
)=1.331.
830
Figure2. Definitionof geometricparameters.
2.2 Undrained shear strength considering the
rotation of principal stress direction caused by
excavations
In order to consider the rotation of principal stress
directioncausedby excavationinclays, theclassical
anisotropic strength formula for soft clays recom-
mended by Casagrande& Carillo (1944) is adopted
here. Inthevertical plane, theundrainedshearstrength
under anyprincipal stressdirectionisexpressedas
whereS
ui
=undrainedshearstrengthof clayswhenthe
anglebetween principal stress direction and vertical
directionisi. S
uh
andS
uv
canbeobtainedfromgeneral
undrainedtriaxial tests.
AccordingtothegeometricrelationshipinFigure2,
i canbeexpressas
where =anglebetweenthedirectionof soil failure
surfaceandvertical direction; =anglebetweenthe
directionof soil failuresurfaceandtheprincipal stress
direction.
According to the test result fromLo (1965) is
a constant which does not vary with the rotation of
theprincipal stressdirection. For anundrainedanaly-
sis, =,4is preferred. Inaddition, theanisotropic
strength ratio is defined as k =S
uh
,S
uv
. For an
isotropicclayS
uv
=S
uh
, k =1. Substitutingtheexpres-
sionsof k and resultsinthefollowingexpression,
By substituting equation (4) to equation (11), the
anisotropic strength considering both the vertical
effective stress and the rotation of principal stress
directioncausedby excavationcanthenbedescribed
asfollows:
Takingadvantageof thedefinitionof safety factor
by strength reduction finiteelement method, wecan
definethebasal stabilityfactor as
whereS
u
(z)=real undrainedshear strengthof clays;
andS
u
(z)
critical
=critical undrainedshear strength.
3 UPPER BOUNDANALYSISOF BASAL
STABILITY OF EXCAVATION
CONSIDERINGUNDRAINED
ANISOTROPIC SHEAR STRENGTH
OF CLAYS
3.1 Soil slide failure mechanism
Chen(1975) strengthenedthat theassumedsoil slide
failuremodeis crucially important tothefinal result
of thelimit analysis. Chang(2000) andFaheemet al.
(2003) suggestedthat thePrandtl slidefailureisclose
to thereal basal failuremodeof excavations. In the
paper, thePrandtlsslidefailuremodeisadopted.
According to the distance between the hard stra-
tumunder thebaseof excavationandtheexcavation
base, theslidefailuremodecanbeclassifiedintotwo
typical modes as shown in Figure 3. Based on the
strengthreductionfiniteelement method, Goh(1990)
andFaheemet al. (2003) foundthattheshear strength
of clay-wall interfacehasslight influenceonthebase
stability. Here, wesimplyassumethat theadhesionof
clay-wall interfaceis neglectable, whichleads to the
simplicityintheupperboundanalysis.Theslidingsur-
faceconsistsof a90
circulararc sandwichedbetween
two 45
isosceles wedges, an elastic wedge gjh and

plasticjik. Thesoil columnefji actsasasurcharge.
According to the upper bound limit theorem, the
rate of external work should be equal to the rate of
internal energydissipationinthesysteminastability
problem.If arealistic,kinematicallyadmissiblesliding
mechanismisassumed, areasonablecollapseloadcan
beevaluated.
3.2 Calculation of basal stability
Assume the clay has a unit weight and a uniform
surchargeq ispresent adjacent totheexcavation.
Case One: T T
c
=B,
2
FromFigure3(a), therateof external workdoneby
(1) theweightof thesoil columnsefji andmnjg, (2) the
831
Figure3. Velocity fieldbasedonassumedPrandtl sliding
mechanism.
weight of soil inthetwoisosceleswedgesjik andgjh,
and(3) theweight of soil intheradial zonejhk is
The total rate of internal energy dissipation from
(1) sliding along fi (2) sliding along ik and gh, and
(3) theradial shear combinedwitharc slidinginthe
radial shear zonejhk is
where
By equatingdW to dE, thebasal stability of exca-
vationis
Whentheclay is homogeneous andtheundrained
shear doesnt varywiththesoil profile.
Case Two: T - T
c
Similarly,
Whentheclay is homogeneous andtheundrained
shear doesnt varywiththesoil profile.
4 PARAMETRIC STUDIES
Thecurrentanalysisof basal stabilitycanconsider the
rotation of principal stress direction and anisotropic
832
Figure4. InfluenceD/Honthefactor of safety.
feature of clays. To study the effect of the strength
anisotropy, D/H and T/T
c
to the basal stability of
deep excavation in clays, an infinitely long verti-
cal deep excavation in soft clays is analyzed using
the aforementioned method. The unit weight of the
clay is =18kN/m
3
, the undrained shear strength
is S
uv
(z)=0.33
v
, the width of the excavation is
B=15m, the depth of the excavation is H =12m,
thepenetrationof thediaphragmis6m, 8m, 10mand
12m, respectively.
4.1 Effects of strength anisotropy
Figure4showsthefactor of safetyof thebasal stabil-
ity of deepexcavationvarieslinearly withthedegree
of strengthanisotropy of clay. Inaddition, thelarger
degreeof anisotropy, thesmaller of safetyfactor. With
thesameanisotropic ratio, increasingthepenetration
of the diaphragmcan not increase the basal safety
factor greatly.
4.2 Effects of D/H
Figure5 indicates thesafety factor of basal stability
increases with the increasing of D/H, which shows
thecontributionof D/H toF
s
. However, withthesame
D/H, theanisotropicratioismoreinfluential thanD/H
tothesafetyfactor.
4.3 Effects of T/T
c
Figure6indicates that presenceof thebedrock close
tothebaseof theexcavationincreasesF
s
. Thiscanbe
explainedthat thesizeof theyieldingzoneisaffected
sincethedisplacement of thesoil beneathandaround
theexcavationis restrainedwhentherigidstratumis
closetothebaseof excavation.However,withthesame
T/T
c
thefactorof basal safetyisdecreasedclearlywith
thedecreasingof anisotropicratio.
Figure5. InfluenceD/H onfactor of safety.
Figure6. InfluenceT/T
c
onthefactor of safety.
5 FIELDCASE STUDIES
5.1 Boston case
Hashash andWhittle(1996) analyzed an deep exca-
vation in Boston Blue Clays through finite element
methodincorporatinganadvancedeffectivestresssoil
model, MIT-E3 (Whittle and Kavvadas, 1994). The
study focus on an idealized (symmetric) plain strain
excavationgeometrywithhalf-width, B/2=20m; the
depth of excavation is 10m, 15m, 22.5mand 30m
respectively; thepenetrationof thediaphragmis2.5m,
5m, 17.5mand30mrespectively. Thevertical effec-
tive stress is expressed as
v0
=8.19z + 24.5, the
gravity of the clay is =18.0kPa. When the OCR
is1.0, 2.0and4.0, thestrengthanisotropyratiois0.5,
0.48and0.43respectively. Inthecurrent analysisthe
author assumes thestrengthanisotropy ratio is about
k =0.5for BBC. Accordingtotheproposedmethod,
thefactor of basal stability is1.16whenthedepthof
833
excavationis10mandpenetrationof diaphragmis2.5
whichquiteagreeswiththeresult(Hashashet al.1996)
basedonthefiniteelementmethodincorporatingwith
thecomplicatedMIT-E3model.
5.2 Shanghai case
One failure case of excavation for a railway station
inShanghai was reportedby J ianget al. (1997). The
width of the excavation is 23m, the depth and the
length is 16.7mand 600m. The penetration of the
diaphragmis13.3mand0.8mthick. Soil parameters
concernedareK
0
=0.65, 1k, =0.91, M =0.818.
According to the formula for calculating the basal
stability recommended by the Shanghai Foundation
Design Code and the formula recommended by the
Shanghai Tunnel Engineering Design and Research
Institute, the safety factors are 2.4 and 1.4, respec-
tively. However, thesafety factor will reduceto 0.97
when the undrained shear strength is considered as
anisotropicandnon-homogeneous.
5.3 Taipei case
Suet al. (1998) analyzedthebasefailureof anexca-
vationinTaipei usingananisotropicstrengthformula
introducedbythemselves.Thedimensionof thisexca-
vationisabout100mlong, 17.5mto25.8mwide, and
13.45mdeep. Thediaphragmwall retainingtheexca-
vationis24.0mdeepand0.7mthick. Failureoccurred
abouttwoandahalf hoursafter completionof thelast
stageexcavation and only two minutes wereneeded
before the entire internal bracing systemcollapsed.
The excavation site was located in a reclaimed land
alongtheKeelungRiver, whichmeandersthroughthe
Taipei Basin. Thetop8.7mof thesubsoil profilewas
backfill and hydraulic fill materials. Underlying the
fillsarea2.0mthicksiltysandlayer, athicksoft clay
layer rangingfromGroundLevel 10.7to 44.7m,
and a silty sand layer ranging fromGL 44.7mto
the bed rock located at GL 55.0m. The distribu-
tionof S
uc
andS
ue
withdepthareS
uc
=0.271
vc
and
S
ue
=0.189
vc
, respectively, i.e. k =0.7. The calcu-
latedsafetyfactor is1.10whichshowsapossibilityof
basal failure.
6 CONCLUSIONS
A method for analyzing the base stability of deep
excavations in anisotropic soft clay is presented in
this paper. It is derivedfromtheupper boundtheory
andstrengthreductiontheory assumingthePrandtls
soil slide failure mode suitable for a critical basal
stability whichcanconsider theanisotropy andnon-
linear featureof clays. Theundrained shear strength
and anisotropic ratio required for the method can
be determined from the conventional CK
0
UC and
CK
0
UE triaxial tests. Comparisonsbetweentheresult
of thecurrent methodandthenumerical results pre-
sented by Hashash and Whittle (1996) for an exca-
vation in normally consolidated Boston blue clay
demonstratetheabilityof theproposedmethodtosim-
ulatethestrengthanisotropy andnonhomogeneity of
soils on thebasal stability of deep excavations. Fur-
thermore, this method also shows good accuracy in
back-analyzingthesafety factor for twobasal failure
casesinShanghai andTaipei respectively. Conclusions
of thepaper canbesummarizedasfollows.
1 Thenecessitiesof consideringtheanisotropicfea-
ture of clays for the calculation of basal stabil-
ity of deep excavations in soft clays have been
demonstrated.
2 Without taking into consideration of strength
anisotropy of clays, thesafety factor against base
heavefromthetraditional methods is higher than
thereal one.
3 The parameters used in the current method are
easily obtained by conventional lab tests (e.g. tri-
axial compressiontest andextensiontest or plain
compressionandextensiontest). Additionally, the
methodcanconsider therelationshipbetweensoil
strengthandvertical effectiveconsolidationstress.
REFERENCES
Bishop, A.W. 1966. The strength of soils as engineering
materials. Geotechnique 16(2): 89130.
Bjerrum, L. &Eide, O. 1956. Stabilityof struttedexcavations
inclay. Geotechnique 6: 3247.
Casarande,A. &Carillo, N. 1944. Shearfailureof anisotropic
soil. J. of the Boston Society of Civil Engineers 31(4).
Chang, M.F. 2000. Basal stability analysis of braced cuts
in clay. Journal of Geotechnical and Geoenvironmental
Enigineering, ASCE, 126(3): 276279.
Chen, W.F. 1975. Limit analysis and soil plasticity. Elsevier
Scientific, Amsterdam.
Clough, G.W. & Hansen, L.A. 1981. Clay anisotropy and
bracedwall behavior.J. Geotech. Eng. Div.,ASCE,107(7):
893913.
Davis, E.H. & Christian, J.T. 1971. Bearing capacity of
anisotropic cohesive soil. J. Soil Mech. Found. Div.,
ASCE, 97(5): 753769.
Duncan, J.M. & Seed, B.H. 1966. Strength variation along
failuresurfacesinclay. J. Soil Mech. Found. Div., ASCE,
92(9): 81104.
Faheem, H., Cai, F., Ugai, K. & Hagiwara, T. 2003. Two-
dimensional base stability of excavations in soft soils
usingFEM. Comp. and Geotechnics 30(2): 141163.
Goh, A.T.C. 1990. Assessment of basal stability for braced
excavation systems using the finite element method.
Comp. and Geotechnics 10(4): 325338.
prediction for deep excavations in soft clay. Journal of
122(6): 474486.
834
Hu, Z.F., Zhou, J. &Yang, L.D. 2001. Study of subsoil sta-
bility under deep excavation. Chinese Journal of Civil
Engineering, 34(2): 8495.(inChinese)
J iang, H.W., Zhao, X.H. & Hang, B.L. 1997. Analysis of
heave-resistant stability for deep braced excavation in
soft clayunder anisotropiccondition. Chinese Journal of
Geotechnical Engineering, 19(1): 17.(inChinese)
Ladd, C.C. 1973. Discussion. MainSession4. In: Proc. 8th
ICSMFE, eds. Committee of 8th ICSMFE, Moscow, 4(2):
108115.
Lo,K.Y.1965.Stabilityof slopesinanisotropicsoils.Journal
of the soil mechanics and foundations division, ASCE,
91(SM): 85106.
Morgenstern, N.R. & Price, V.E. 1965. Theanalysis of the
stabilityof general slipsurface. Geotechnique 15:7993.
Spencer, E. 1967. A method of analysis of the stabil-
ity of embankments assuming parallel interslice forces.
Geotechnique 17:1126.
Su, S.F., Liao, H.J. & Lin,Y.H. 1998. Basal stabilityof deep
excavation in anisotropic soft clay. J. Geotechnical and
Geoenvironmental Eng., ASCE, 124(9): 809819.
Terzaghi, K. &Peck, R.B. 1948. Soil mechanics in engineer-
ing practice. Wiley, NewYork.
Ukritchon, B., Whittle, A.J. & Sloan, S.W. 2003. Undrained
stabilityof bracedexcavationsinclay. J. Geotechnical and
Geoenvironmental Eng., ASCE, 129(8): 738755.
Wei, X. & Huang, M.S. 2006. Anisotropic bounding sur-
face model for clays. Chinese Journal of Hydraulic
Engineering 37(7): 831837.(inChinese)
Whittle, A.J., DeGroot, D.J., Ladd, C.C. & Seah, T.H. 1994.
Model prediction of theanisotropic behavior of Boston
Blueclay. Journal of Geotechnical Engineering, ASCE,
120(1): 199224.
Zou, G.D. 2004. Analysis of stability against upheaval of
deepexcavationbyanupper limit method. Rock and Soil
Mechanics, 25(12):18731878.(inChinese)
835
Analytical twoandthreedimensionmodelstoassessstabilityand
deformationmagnitudeof undergroundexcavationsinsoil
L. E. Sozio
Promon Engenharia, Sao Paulo, Brazil
ABSTRACT: Twoandthreedimensional elasticandplasticanalytical solutionsarepresented, envisagingthe
evaluationof stability anddeformationof atunnel or cavern. Thesesolutionsarebasedonthick wall cylinder
or thick wall spheremodels, whereinner andouter radiusaredefinedaccordingtotunnel or caverngeometry
andcorrespondingsoil cover togroundsurface. Radial bodyforcesareintroducedintotheequationstoemulate
gravityforces. Cohesiveandfrictional materialscanbeconsidered. Theresultingmodel permitsthecalculation
of theplastic zoneradiusasafunctionof thesupport pressureactingat thetunnel or caverninner surfaceand
obtainthecritical supportpressure. Byadoptinglinearelasticbehaviourattheelasticzoneandnovolumechange
at theplasticzoneit ispossibletoderiveacompleteconvergenceconfinement curve. Radial displacement and
stressresultingfromthesemodelsarecomparedwithpublishedcasehistories. Comparisonisalsomadewitha
knownpublishedelasticperfectlyplasticplanestrainmodel.
1 INTRODUCTION
A fundamental parameterintunnel designistheallow-
abledistancebetweentunnel faceandthepointwhere
supportiseffective.Supportisgenerallyeffectiveupon
closureof theinvertarchinNATM(NewAustrianTun-
nellingMethod) tunnels(Fig. 1), or wheresegmental
liningiserectedandgrout isinjectedinnonpressur-
ized shield tunnels. In any casethetunnel Engineer
mustanticipateif theopeningwill remainstablewithin
suchlength. If not, heshouldspecify constructionof
theinvert archcloser tothefaceof theNATM tunnel,
oremployanearthpressurebalancedtypeof shield. In
bothcaseshewouldberequiredtoestimatethemag-
nitudeof soil deformationupto theactivationof the
Figure 1. Tunnel length L, where support is not fully
effective.
support. Theloadonthesupport will dependonthis
deformation.
2 SOMEASPECTSOF CURRENT DESIGN
PRACTICE
Notwithstanding the increasing availability of three
dimensional (3D) numerical computerprogramcodes,
thepredominant designpracticeof atunnel liningis
still basedonatwodimensional (2D)model.Therelief
of groundstressesprior totheactivationof thelining,
if accountedfor, isgenerally simulatedby aprogres-
sive softening of the soil within the excavated area.
The design engineer must rely either on his experi-
enceor onsomepublished3Danalytical or numerical
analyses, which generally do not match the particu-
lar characteristicsof hisproblem, todefinehowmuch
softeningwill beallowedpriortotheliningactivation.
If he allows little or no softening, the lining design
maybeultraconservative, particularlyif thesoil cover
abovethetunnel islarge. If heallowstoomuchsoften-
ing, thestressesontheliningmaybeunderestimated,
and, if he defines a construction method allowing
for an excessive softening, his tunnel is in risk of a
collapse.
A comprehensivestudy into3D effectsat atunnel
headingwasperformedbyNegro(1988).
The analytical models presented herein are pro-
posedtobeusedasapreliminaryestimateof atunnel
837
stress relief and corresponding soil displacements.
Moresophisticated(andtimeconsuming) analysiscan
be performed by using 3D numerical models, at a
detaileddesignstage.
3 ELASTIC-PLASTICANALYTICAL
MODELS
Thesemodelsarebasedonathick wall sphere (3D
model) andonathick wall cylinder (2D model) in
which part of thesphereor thecylinder is in elastic
state and the remainder in plastic state. The area of
the tunnel where support is considered either to be
not effective, or where a supporting pressure p can
beapplied, is simulated as asphere(Fig. 1) with an
inner radiusRi. Theouter radiusRo(Fig. 2) simulates
theground surface, whereasurchargepressureload
s can be considered. This concept has already been
used by Muhlhaus (1985) for alimit analysis where
thewholesphereisassumedtobeinplasticstate, but
no informationis obtainedondisplacements prior to
collapse.
The analytical model was derived in a way that
either a3D or a2D analysis can beperformed. The
3D analysis is the objective of this concept model.
Althoughless interesting, the2D model canbeused
to roughly evaluate average lining load vs. soil dis-
placement, after lining installation, bearing in mind
thattheaxi-symmetrymodel conditiondoesnotallow
for bendingof thelining.
Intheplasticzoneof thesphereorcylinderthestress
stateis foundby combiningthedifferential equation
of equilibrium(in polar coordinates) and the Mohr
Coulombcriterion.
Intheelasticzoneof thesphereorcylinderthestress
stateis foundby combiningthedifferential equation
of equilibrium, thestress strainrelationships andthe
compatibilitystraindisplacementrelationships. Refer
toFigure3for notationandequations.
Radial body forces are included in the equilib-
riumequations of both the plastic and elastic zones
toemulategravityforces.
Figure2. ThickWall Cylinder (2D) or Sphere(3D).
Theboundarybetweentheelasticandplasticzones
(RpinFigure2) isdeterminedbyequatingtheelastic
stressestotheMohr Coulombcriterion. Notethat for
unstableconfigurations no solution is found for that
equation. On theother extreme, if RpRi, then the
wholesphereor cylinder isinelasticstate.
Theradial displacementattheelastic-plasticbound-
aryisobtainedfromtheelasticequation(9)of Figure3,
whichisderivedfromthedifferential equationof equi-
librium, the stress strain relationships and the com-
patibility strain displacement relationships, applied
to the elastic zone, and considering the previously
determinedradial stressat elastic-plasticboundary.
Fordisplacementsintheplasticzoneitisneccessary
toconsider theplasticdeformationbehaviour. Novol-
umechangewasadoptedfor thisanalysis, asimposed
onequation(10). Thisequationshouldbemodifiedif
dilatingor contractingbehaviour istobemodelled.
If Ro , equation(9) will reproducetheclassi-
cal elasticexpressionfor radial displacement duetoa
cylindrical or spherical openinginaninfiniteelastic
medium.
Frictionless materials (undrained analysis) havea
particular set of equations to define plastic radius,
sinceforfrictionangleequal tozero, parameterM =0
(equation2of Fig. 3) makethegeneral equationsinde-
terminate. A checkhasbeenmadebyadoptingavery
small valueof M, for instanceM =1e
5
, andthegen-
eral andundrainedanalysesgivepracticallythesame
results.
The equations of Figure 3 can be inserted into a
programmablepocket calculator (suchas theHP32s,
whichcontainsanalgorithmthat solvesequation(7a)
or (7b)=equation(8)) to enablerapidassessment of
agivencase.
4 LIMITATIONS
These models were developed to give a preliminary
insight into stability and deformation behaviour of
underground excavations, with a particular interest
in tunnel headings. Themain limitations that should
alwaysbeconsideredarelistedbelow:
4.1 Radial simmetry
Themodelsimply inradial simmetry, thusany shape
of tunnel or sphere, and ground surfacewill haveto
be approximated to a cylinder or sphere. One type
of approximation required when considering a tun-
nel headingisthechangeof shapeof theunsupported
lengthof tunnel into asphere, as showninFigure1.
Some geometrical situations may arise for which a
greatdistortionoccurs, andoneof particularinterestis
alongheadingof ashallowtunnel. If suchacondition
istobeanalysed, isadvisabletocomplementtheeval-
uationby conservatively checkingtheresultsof a2D
838
Figure3. 2Dand3Delastic-plasticmodel equations.
839
analysis, for aninfinitecylinder of inner radiusequal
tothelongheadingradius.
4.2 Constitutive model
Soil isassumedtobehomogeneous, isotropicandlin-
ear elastic with Mohr Coulomb plastic criterion and
no volumechange. Soils stress strainbehaviour is in
general not linear, and this would affect mainly the
displacementsestimate.
4.3 In situ stresses
Radial simmetry implies inradial body forces, other
thanvertical gravity.Theeffectof ahorizontal toverti-
cal insitustressratiodifferentthanoneisnotpossible
tobeconsideredinthesemodels.
Also, theeffectof aflownetresultingfromground
water tableabovethetunnel axiscannot beaccounted
for, exceptforthesimplifiedassumptionof adescend-
ingvertical flownet, withhydraulicgradient equal to
one,thusimplyinginasaturated(inlieuof submerged)
unit weight of thesoil. Ingeneral aflownet towards
thetunnel headingrepresentsriskof ahydraulicpiping
failure, andshouldbeavoidedbyvertical orhorizontal
drains, compressedair, or aslurrytypepressure.
4.4 Consolidation
Settlements due to consolidation are not considered
in this model, analyses areeither fully undrained or
drained.
4.5 Local collapse
Local Collapseas defined by Davis et ali (1980), is
not includedinthesemodels.
5 APPLICATIONSANDCOMPARISONS
WITHOBSERVEDDATA
Figure4showsanexampleapplicationof the2Dand
3D modelsfor atunnel, consideringagivenlengthL
betweentunnel faceandeffectivesupport activation.
Thesettlement of thetunnel roof onactivationof the
liningis estimatedas thedisplacement of thesphere
innersurface,whichistheintersectionof the3Dmodel
convergence curve with the abcissa (Fig. 4). At this
point theliningisactivated.
Thefinal pressureon theliningandcrown settle-
ment areestimatedfromtheintersectionof thetunnel
liningstressdisplacementcharacteristiccurvewiththe
2Dmodel convergencecurve.
An analysis can bemadeon thesensitivity of the
unsupported length; the crown settlement increases
withincreaseof thislengthuptothepoint wherethe
wholesphereis in plastic state, resulting in collapse
conditions.
Figure4. Exampleof convergence confinement curve.
For the 3D analysis the inner radius was defined
toequatetheunsupportedlengthtunnel volume, addi-
tionalyconsideringthetunnel faceasahalf sphereof
diameter equal tothetunnel diameter,
whereD isthetunnel Diameter.
Eighttunnel casehistorieswereanalysedintermsof
themodelspresentedherein, seeTable1forreferences
andTable2for parametersandresults.
Thetunnel geometryandsoil strengthanddeforma-
bility parameters were inferred fromthe respective
casehistorypapers. Drainedparameterswereusedfor
sandysoil casesandfullyundrainedparameterswere
used for clayey soil cases. In the absence of better
information, the undrained deformability moduli of
clays weretaken as 600 times their undrained shear
strength. Themeasuredsettlement figuresweretaken
onactivationof thelining, althoughthiswasnot sig-
nificantly different fromthe final settlements on all
cases. For theWashingtonD.C. shieldcaseit appears
that soil movementswerepartiallyrestrainedtomove
freelyattherear of theshieldsteel hood. For thiscase
the unsupported length in the analysis was adopted
slightly smaller thantheactual distancefromfaceto
thesegmental lininginstallation.
The stresses on the lining were measured in the
Edmontoncases (case7andcase8, seeTable1) and
their authorspublishedtheidealizedsoil convergence
confinement curve plus the carachteristic curves of
twotypesof linings, concretesegmentsandsteel ribs
withwoodlagging. The2Dmodel wasappliedtothis
caseandtheresultingcurveisshowninFigure5. The
convergenceconfinementcurveobtainedfromthe3D
model tosimulatetheunsupportedlengthof thetunnel
isalsoshown, for theconcreteliningcondition.
840
A reasonableapproximationtothemeasuredloads
and/ or settlementsfor thisandtheother casescanbe
observed.
A comparisonwas also madewiththemodel pre-
sentedby Mair &Taylor (1993), for theunloadingof
planestrain cylindrical cavity, for linear elastic, per-
fectly plastic undrained soil behaviour. Curves from
bothmodels andfor aspecific set of parameters are
shown in the graph of Figure 6. It was found that
the conformity between these models is affected by
thesoil covertotunnel radiusratio. Ingeneral theratio
betweenradial displacementsof eachmodel increases
asthetunnel becomesshallower. Thiscouldbedueto
thefact that theMair &Taylor model isapplicableto
aninfinitemediumcondition.
6 CONCLUSIONS
Despitethesignificant evolution on numerical com-
puterprogramcodes, twodimensional tunnel analyses
are still predominant in the current design practice.
Consequentlythe3Dstressstateinthezoneof thetun-
nel headingisofteneither ignoredor unsatisfactorily
simulated.
Table1. Casehistoriesandreferences.
Case Reference
1 Frankfurt, NATM, Baulos25 Cording(1976)
2 Frankfurt, Shield, Fahrgasse Cording(1976)
3 WashingtonDC, F2A Line1 Cording(1976)
4 HeathrowCargo Cording(1976)
5 GreenPark J ubileeLine Attewell (1974)
6 SaoPauloE-WLine Sozio(1978)
7 Edmonton, concrete Eisenstein(1979)
8 Edmonton, steel ribs Eisenstein(1979)
Table2. Applicationof 3Dmodel vs. observeddata.
Case 1 2 3 4 5 6 7 8
Diameter m 6.5 6.5 5.5 10.9 4.1 6.1 2.6 2.6
L m 3.0 8.0 5.0 5.0 4.0 7.0 6.0 11.0
Ri m 3.44 4.32 3.38 5.77 2.59 3.98 2.03 2.44
Ro m 14.6 12.4 20.1 13.3 29.3 20.0 26.9 26.9
kN/m
3
19.0 18.5 20.0 19.0 19.0 19.0 20.0 20.0
c kPa 55 15 15 100 150 25 110 110

0 30 35 0 0 33 0 0
E MPa 33 35 80 60 90 60 66 66
0.5 0.25 0.25 0.50 0.50 0.25 0.50 0.50
Rp m 6.68 9.07 5.20 6.11 4.33 5.20 4.52 5.33
w mm 42 69 23 11 20 22 38 43
calculated
w mm 45 70 21 14 17 25 33 38
observed
The elastic plastic 3D analytical model presented
hereinismeanttobeusedasanauxiliarytool toenable
assessment of tunnel headingstabilityconditionsand
deformationmagnitudepriortoinstallationof support.
Figure 5. Convergence confinement curves, Edmonton
cases.
Figure 6. Comparison between Mair & Taylor and 2D
model.
841
The2Dmodel permitsanestimateonthestresslevel
tobeexpectedonthesupport.
Comparisons with some published case histories
and with a known elastic perfect plastic analyti-
cal model arepresented, nosignificant discrepancies
beingobserved.
A seriesof limitationshavetobeconsideredwhen
employingthesemodels, as outlinedabove(Heading
4). Therefore these models are not to be used as a
sole design tool, but as complementary elements to
theacceptedtunnel designproceduresandstandards.
REFERENCES
Attewell, P. & Farmer, I. 1974. Grounddeformationsresult-
ing from shield tunnelling in London clay. Canadian
Geotechnical Journal 11.
Cording, E. & Hansmire, W. 1976. Displacements around
tunnelsinsoil. U.S. Department of Transportation Report
TST-76T-22.
Davis, E., Gunn, M., Mair, R. & Seneviratne, H. 1980. The
stabilityof shallowtunnelsandundergroundopeningsin
cohesivematerial. Geotechnique, 30.
Eisenstein, A., El-Nahas, F. & Thomson, S. 1979. Pres-
sure displacement relations in two systems of tunnel
linings. 6th Soil Mechanics and Foundation Engineering
Conference, Lima.
Mair, R.J. &Taylor, R.N. 1993. Predictionof claybehaviour
aroundtunnels usingplasticity solutions. Predictive Soil
Mechanics, Proc. Wroth Memorial Symp., Oxford, U.K.
Muhlhaus, H.B. 1985. Lower bound solutions for circular
tunnelsintwoandthreedimensions. Rock Mechanics &
Rock Engineering, 18.
Negro, A. 1988. Design of shallow tunnels in soft ground.
PhD. Thesis, Universityof Alberta, Canada.
Sozio, L.E. 1978. Settlements inaSao Paulo shieldtunnel.
Tunnels &Tunnelling, vol. 10n
7.
842
Dynamicresponseof saturatedsiltyclayaroundatunnel under subway
vibrationloadinginShanghai
Y.Q. Tang& Z.D. Cui
X. Zhang
ABSTRACT: Thispaper studiesthesaturatedsoftclayaroundthetunnel betweenJ ingansi StationandJ iangsu
Stationof Shanghai subwayLineNo.2.Thecontinuousdynamicmonitoringisconductedbymeansof embedded
earthpressurepiezometers andporepiezometers aroundthetunnel at different locations anddifferent depths.
Theresponsefrequency andstress amplitudeof thesaturatedsoft clay arestudiedwiththedistancefromthe
tunnel duetothesubwayvibrationloading. A formulaisproposedfor theattenuationof thedynamicresponse
of thesoil. Thedistanceof influenceandtheamplitudeof thedynamicresponsearecalculatedandtheinfluence
onthesurroundingbuildingsunder thesubwayvibrationloadispredicted, whichoffersvaluablereferencesfor
thedesign, theconstructionandthesafeoperationof thesubway.
1 INTRODUCTION
Thesubwayisindispensableasasafe, comfortableand
highspeedtransport vehicle, inmoderncities, but in
thecourseof itsrunning, thevibrationproblemshould
notbeignored. Somemodels(Alabi, 1992; Lipenand
Chigarev, 1998; Sheng, etal., 1999; J ones, etal., 2000)
wereused to analyzetheground vibration under the
train loading on the railway. Pan (1995), Hirokazu
(2001) andXie(2002) studiedthedynamic response
of the railway system, but they focused only on the
responseof theupper structureof thefoundationand
didnot study thedynamicresponseof thesoil. Upto
now, onlyafewresearchershavestudiedthedynamic
response of the saturated silty clay around the sub-
waytunnel.Yet, thelong-termvibrationloadingof the
subway hasresultedinlargedeformationof thesatu-
ratedsiltyclay(Chen, etal., 2002; Wang, etal., 2003).
Accordingto monitoreddata, alargedeformationof
theaxisof thesubwaytunnel andthegroundsettlement
hadoccurredinsomesectionsof thetunnel of Shang-
hai subway LineNo.1(Lin, et al., 2000) andaffected
thenormal operationof thesubway.Thesettlementand
deformationoriginallybeganwiththechangeof pore
waterpressureandsoil stress(Tang, etal., 2003;Tang,
et al., 2005). So it is important to study thedynamic
responseof thesoil aroundthesubway tunnel for the
safeoperationof thesubway.
As thesoil aroundthesubway tunnel inShanghai
ismainlythesaturatedsiltyclay, thispaper focuseson
it under thesubway vibration loading. Theresponse
frequencyof thesaturatedsoft clayandthelawof the
amplitudeof stressresponsechangingwiththedepth
andthedistanceaway fromthetunnel areall studied
under the subway vibration loading. The result can
offer avaluablereferenceto thedesign, construction
andthesafeoperationof thesubway.
2 DYNAMIC MONITORING
Inorder tostudy theinfluencecausedby thesubway
vibrationloadingonthesaturatedsoftclayaroundthe
tunnel, fieldtest andmonitoringareconductedinthis
research. ThesiteisselectedbetweenJ ingansi Station
and J iangsu Road Station. The dynamic monitoring
systemis adopted for field monitoring and its sam-
plingfrequency canreach200Hz anditsprecisionis
0.1kPa. It can fully reflect thesoil responsearound
thetunnel dueto thesubway vibration loading. The
dynamic monitoring systemconsists of a resistance
sensor, a dynamic strain amplifier, a data selector
843
resistance
sensor
electric signal
strain
amplifier
electric signal
data
selector
digital signal
computer
Figure1. Schemeof dynamicmonitoringsystem.
Monitoring
Room
3000
2000
1
8
0
0
15000 15000
BH5
BH2
BH4 BH3
BH1
People Square
Subway
Zhong Shan Park
Figure2. Layout of boreholes(unit: mm).
and acomputer. Thesystemcan record all thesam-
plingdatacollectedbythecomputer inreal time. The
dynamicmonitoringsystemisshowninFigure1.
Figure2shows thelayout of boreholes at thesite.
In the plane, there are five boreholes, each 110mm
in diameter, parallel and vertical to thesubway tun-
nel, respectively. The distance between the site and
J ingansi Station is 210m. Boreholes BH3, BH4 and
BH5 areparallel to thetunnel axis only 1.8maway
fromtheoutsideof thesegment of thesubwaytunnel
and thedistancebetween themis 15.0m. Boreholes
BH1, BH2, andBH3arevertical tothetunnel axis. In
order tostudytheattenuationof theeffect onthesoil
aroundthetunnel withtheincreasingdistanceunder
the subway vibration loading, the distance between
boreholesBH1andBH2is3mandthatbetweenbore-
holes BH2andBH3is 2m, sothat thereis astep-up
course.
Figure3showsthedistributionof strataandinstru-
ments. In thesection, thesubway tunnel lies in gray
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~ ~
2.70m
7.80m
20.0m
18.0m
~
~
~
~
~
~
~ ~
~ ~
~
~
~
~
~ ~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~ ~
20.0m
Strain gauge
1.8m
8.5m
11.5m
13.5m
R3m
Tunnel
1.30m 1
2
3
4
5
~
~
~
~
~
~
~
~ ~
~
~
~ ~
~ ~
~
~
~
~
~
~
~
~
~
~
~ ~
~ ~
~
~
~
~
Figure3. Soil profileandembeddedinstruments.
Note: layer No.1ismixedsoils; layer No.2isbrownyellow
silty clay; layer No.3isgray muddy silty clay; layer No.4is
graymuckyclay; layer No.5r isgraysiltyclay.
siltyclayof layerNo.4andtheearthpressurepiezome-
tersandporepiezometersarelocatedat thedepthsof
8.5m, 11.5mand13.5m, respectively, inlayer No.4,
tomonitor theresponsecharacteristicof thevibration
for thesubwayrunning.
3 RESPONSE FREQUENCY OF SOIL
The train of Shanghai subway has six carriages and
itswholelengthis139.46m. Itsnormal runningspeed
is 60km/h and thebreak at each station is 3040s.
The monitoring site is near the subway station. The
trainwill decreaseitsspeedwhenrunningintothesta-
tion. Accordingtostatistics, thespeedis generally at
3040km/hwhenthetrainpasses throughthemon-
itoring site, and thetimeis 1216s in general. The
interval of thesubway train crossing themonitoring
siteisunequal,generally36minand34minduring
rushhours.
Therearetwo groups of wheels in every carriage
of thesubwaytrain, andthesubwayvibrationloading
is transferredto thesoil outsidethetunnel segments
through the system of wheels-segments. When the
subwaytraincomesnearthemonitoringsite, thevibra-
tion loading is produced and spread in waves, and
the soil can be induced to respond. In Figure 4, the
horizontal axis is timeand thevertical axis is stress
response, andthewavecanbeseenclearly whenthe
subway trainruns across themonitoringsite. Owing
to the different loading of each carriage (the num-
berof passengersbeingalwaysdifferent), theresponse
amplitudeisalsodifferent. Theresponsefrequencyof
soil isalsodifferentbecauseof thedifferentinterval of
thegroupsof wheels. Therearetwotypesof response
844
Figure4. Waveof thesoil response.
0
0.5
1
1.5
2
2.5
3
No.1 No.2 No.3 No.4 No.5
Hole number
F
r
e
q
u
e
n
c
y

(
H
z
)
High frequency
Low frequency
Figure 5. The response frequency of soil for the five
boreholes.
frequency: the high frequency f
h
and low frequency
f
l
. Thehighfrequencyisproducedmainlybytherear
wheelsof thefirstcarriageandthefrontwheelsof the
secondcarriage. Thedistancebetweenthemis short,
so theresponsetimeof soil is also short. When the
two groups of wheels runacross themonitoringsite,
thehighfrequencyisproduced. Thelowfrequencyis
producedmainly by thefront andrear wheels of the
samecarriageandthedistanceislonger, soitislower.
Thevibrationresponseof soil canbeseenclearlywhen
everygroupof wheelsrunsacrossthemonitoringsite,
asshowninthefigure.
Bycontinuousdynamicfieldmonitoring, plentyof
dataareobtained.All thewavesareanalyzedandcom-
paredandthetypical responsewaveof thesaturated
soft clay is obtained. Theresponsefrequency of soil
(siltyclayof layer No.4) of everyboreholeisobtained
by collecting, analyzing and arranging its data. The
statistic values are 2.42.6Hz and 0.40.6Hz for
thehighandlowfrequency, respectively, asshownin
Figure5.
When thesetwo kinds of responsefrequency and
thenatural frequencyof softclayinShanghai areaare
Figure6. Attenuationof horizontal stressvertical totunnel
axis.
taken into consideration, thecorresponding measure
should be adopted to avoid the equality of monitor-
ingfrequenciesandnatural frequencyinthedesignof
subway tunnel. This canresult inresonance, causing
different kinds of calamity, such as theinstability of
thesubwaytunnel, thefissureof thetunnel segments,
andsoon.
4 DYNAMIC RESPONSE OF SOIL
4.1 Attenuation of dynamic response of soil vertical
to the axis of the subway tunnel
Thetest andresearchontheattenuationof vibration
of the train on the ground with the increasing dis-
tance fromorigin of vibration were mainly focused
on thetransversedirection. FromChens test (Chen,
et al., 1998), thevibration almost attenuated to zero
at 25maway fromtherail of therailway, andhadno
effectbeyond25m.Thesameideawasprovedbyother
tests, andonly thedistanceof attenuationwasdiffer-
entfordifferentvibrationloadingsandconditions.The
subway tunnel is embedded in thesemi-infinitesoil
body, andthevalueof attenuationis larger thanthat
of theground. BoreholesBH1, BH2andBH3arehor-
izontally ardistancedat different locations vertical to
thetunnel axis and are6.8m, 3.8mand 1.8maway
fromtheedgeof thesubway tunnel, respectively. In
order tofindtheattenuationlawmoreclearlyandtry
to avoid theinterferenceby other factors, thesoil at
13.5mdepthischosenfor study. Themonitoreddata
areaveragedandtheresult isgiveninFigure6.
According to the statistic result of large amounts
of monitored data, the relationship between the soil
responseandthedistancevertical tothesubwaytunnel
axisisgiveninFormula(1):
845
where Lp (kPa) is the value of dynamic response
of soil when thesubway is running; K
0
(kPa) is the
valueof dynamic responseof soil at theedgeof tun-
nel; K
1
(kPa/m) is the first order coefficient of the
attenuationof dynamical responsewiththedistance;
K
2
(kPa/m) isthesecondordercoefficientof theatten-
uationof dynamical responsewiththedistance;x isthe
distanceawayfromtheedgeof thesubwaytunnel.
According to field monitored data by fitting For-
mula(1), thecoefficients areobtained: K
0
=1.3526,
K
1
=0.0321andK
2
=0.0077.
Substitutionof coefficientsintoFormula(1) yields:
Formula(2) is givingthedynamic responseatten-
uation law of horizontal soil vertical to the tunnel
axiswithdistanceinthisfieldmonitoring, wherethe
distanceof x is011.33m.
Using Formula (2), the distance of influence and
theamplitudeof dynamicresponsecanbecalculated.
Theinfluenceonthesurroundingbuildingsunder the
subwayvibrationloadingcanbepredicted. Itcanoffer
atheoretical referencetothedesignandconstruction
of thebuildingnear thesubwaytunnel.
It canbeseenfromFigure6that theattenuationof
thesubway vibrating loading takes on acertain rule
withtheincreasingdistance. Theinfluencedistanceis
11.33mobtainedfromFormula(2).Theresultisquite
different fromthat of thedynamicrailwayloadingon
theground. Inadditiontothefactor of soil, themain
reasonis that theshear modulus andthedampingof
soil increasewiththeincreasingdepth.Thismakesthe
dynamicloadingattenuaterapidlytovanish.Alongthe
subway, therailsarejoinedwithtunnel segmentsand
therigidityismuchlarger thanthatof thesoil vertical
tothesubway.
4.2 Dynamic response law of soil at different
depths
Sensorsareembeddedatthetop, themiddleandatthe
bottomof thetunnel. Theearlystaticmonitoringdata
indicatethat thesoil pressureapproximately takeson
thelinear increasewiththeincreasingdepth. Inorder
to findtherelationshipbetweenthestress amplitude
anddepth,thestatisticdataattherushhourareselected
for analysis, asshowninFigure7.
It canbeseenfromFigure7that under thesubway
vibrationloading, themaximumchangingamplitudes
of the stress response of soil are 0.23kPa at 8.5m,
0.70kPaat11.5mand1.15kPaat13.5m.Thechange
of stressamplitudeisapproximatelylinearwithdepth.
Moreover, thestressamplitudeat therushhour inthe
morning is larger than that at the rush hour in the
evening, which is also larger than that at noon. This
indicatesthat thenumber of passengersisthelargest
Figure7. Amplitudeof soil pressureresponsewithdepth.
inthemorning.Thisshouldbetreatedasareferenceto
designandcalculatetheworstcaseintheconstruction.
5 CONCLUSIONS
1 Theattenuationof thesubwayvibrationloadingver-
tical totheaxisof subwaytunnel takesonacertain
rulewiththeincreasingdistance. Accordingtothe
statisticresultsof largeamountsof monitoreddata,
theformulafortheattenuationof dynamicresponse
of thehorizontal soil vertical totheaxisof thesub-
waytunnel isobtainedwiththeincreasingdistance.
Usingthis formula, theinfluencedistanceandthe
amplitudeof dynamic responsecanbecalculated.
The influence on the surrounding building under
the subway vibration loading can be predicted. It
offers a theoretical reference to the design and
constructionof buildingsnear thesubwaytunnel.
2 The change of stress amplitude of soil is approx-
imately linear with depth when subway trains are
running across. Moreover, the stress amplitude is
thelargest duringtherushhour inthemorning. It
indicates that theflowof passengers is thelargest
duringtherushhour inthemorningwhenthesoil is
intheworstcondition.Thisvalueshouldbeadopted
as referencefor thedesignandconstructionof the
subwaytunnel.
3 Accordingto thetwo kinds of responsefrequency
by thefieldmonitoringof thesoil aroundthetun-
nel, which are produced by the subway vibration
loadingandmatchedwiththenatural frequency of
soft clayinShanghai area, thecorrespondingmea-
surement should be adopted in the design of the
subway tunnel in order to avoid the response fre-
quencyequal tothenatural frequency. Itcouldresult
inresonancecausingkindsof calamity, suchasthe
instability of thesubway tunnel, thefissureof the
tunnel segments, etc.
846
ACKNOWLEDGEMENTS
This work are supported by the research grant
(40372124) fromNational Natural Science Founda-
tion of China, Shanghai Key Subject (Geotechni-
cal Engineering) Foundation and Shanghai Leading
AcademicDisciplineProject(ProjectNumber: B308).
REFERENCES
Alabi, B. 1992. A parametric study on some aspects of
ground-borne vibrations due to rail traffic. Journal of
Sound and Vibration 153(1): 7787.
Chen,Y.M., Chen, R.P. &Lu, S. 2002. Several soil mechanics
problemsinmetroconstructionandoperationonsoftclay
foundation. In:Proceedings of Seminar on Metro Con-
struction and Environmenta1 Geotechnical Engineering.
Hangzhou: (inChinese).
Hirokazu,T., Shuhei, S. &Xie,W.P. 2001.Traintrack-ground
dynamics dueto high speed moving sourceand ground
vibration transmission. Journal of Structure Mechanics
and Earthquake Engineering 682(7): 299309.
J ones, C.J., Sheng, X. & Petyt, M. 2000. Simulations of
ground vibration from a moving harmonic load on a
railway track. Journal of Sound and Vibration 231(3):
739751.
Ling, Y.G., Liao, S.M. & Liu, G.B. Discussion of Influ-
encing factors on axial deformation of subway tunnel.
Underground Space 20(4): 264267(inChinese).
Lipen, A.B. & Chigarev, A.V. 1998. Thedisplacementsinan
elastichalf-spacewhenaloadmovesalongabeamlying
onitssurface. Journal of Applied Maths Mechanics 62(5):
791796.
Pan, S.C., Li, D.W. & Xie, Z.G. 1995. The discussion of
theinfluencingonenvironmentbysubwaytrainvibration.
Journal of Vibration and Shock 14(4): 2934(inChinese).
Sheng, X., J ones, C. & Peryt, M. 1999. Ground vibration
generatedby aharmonic loadactingonarailway track.
Journal of Sound and Vibration 225(1): 328.
Tang,Y.Q., Hang,Y. &Ye,W.M. 2003. Critical dynamicstress
ratioanddynamicstrainanalysisof soil aroundthetunnel
under subwaytrafficloading. Journal of Rock Mechanics
and Engineering 22(9): 15661570(inChinese).
Tang,Y.Q., Zhang, X., Zhou, N.Q. &Huang,Y. 2005. Micro-
scopicstudyof saturatedsoftclaysbehavior under cyclic
loading. Journal of Tongji University 33(5): 626630(in
Chinese).
Wang, C.J., J i, M.X. & Chen, Y.M. 2003. Subsequent settle-
ment of saturated soft clay ground induced by train. In:
Proceedings of the 9th Conference on Soil Mechanics and
Geotechnical Engineering. Beijing: TsinghuaUniversity
Press(inChinese).
Xie,W.P., Hu, J.W. &Xu, J. 2002. Dynamicresponseof track-
ground systems under high speed moving load. Journal
of Rock Mechanics and Engineering 21(7):10751078(in
Chinese).
847
Lateral responsesof pilesduetoexcavation-inducedsoil movements
C.R. Zhang, M.S. Huang& F.Y. Liang
ABSTRACT: Lateral soil movements induced by excavation of adeep foundation pit may adversely affect
nearby pilefoundations. In this paper, asimpleanalysis method is proposed for computing lateral responses
of passivepilegroups subject toexcavationinducedlateral soil movement. Basedonatwo-stagemethod, the
Winkler model isadoptedfor simulatingthepile-soil interaction, combinedwithfinitedifferencemethodinthe
caseof multi-layeredsoils. A specifiedfree-fieldsoil movement profileis usedas input. Then, thegoverning
equationfor apilegroupisobtainedconsideringtheshieldingeffect inpilegroupsbythesimplifiedMindlins
equation. Comparisonsaremadebetweentheobservedbehavior of centrifugemodel testsandthosecomputed
bytheproposedmethod. Thepresentmethodcaningeneral giveasatisfactorypredictionof thelateral response
of passive pile groups. However, the major limitation is the assumption of linear elastic soil springs, which
providesonlyanupper boundestimate.
1 INTRODUCTION
Deep excavations for basements and other under-
ground facilities are unavoidable in big cities. The
lateral soil movementsresultingfromexcavationswill
imposeadditional bending moments and deflections
onnearbypiles,whichmayleadtostructural distressor
failure(Panet al. 2002, Gohet al. 2003). Developing
reliable and simple methods to estimate the behav-
ior of piles next to excavations is urgent in practical
engineering.
Availablemethodsof analysiscanbebroadlyclas-
sified into complete three-dimensional method and
two-stage method. The former, which is carried out
by a finite element analysis, can consider complex
pile-soil interaction and thewholeconstruction pro-
cess(Gohet al. 2002, Miaoet al. 2005). However, itis
computationally expensiveandtheaccuracy depends
ontheaccuracyof theconstitutivesoil models, which
arecurrently under astageof calibrationwithresults
fromphysical tests (J uirnarongrit & Ashford 2006).
Furthermore, it ismoresuitedtoobtainabenchmark
solutionor toobtainsolutionsof detailedanalysisfor
final design,ratherthanasapreliminaryroutinedesign
tool (Kitiyodom& Matsumoto2002).
Comparatively, the simplified two-stage method
appearstobeamoreattractivechoice.Theknownfree-
fieldsoil movementisaprerequisite(Gohet al. 1997).
As awealth of experienceaccumulations on estima-
tionof thesoil deformationresultingfromexcavation
havebeenobtainedbyengineers,thetwo-stagemethod
cangiveasatisfactoryresulttoguideconstructionand
design. Intheworksof Poulos(Poulos& Chen1997,
Chen & Poulos 1997, 1999), a combined finite ele-
mentmethodandboundaryelementmethodwasused
to analysepiles adjacent to an excavation. However,
mostof theirworksarefocusedonsinglepassivepiles
andthesurroundingsoil ismodelledasanelasticcon-
tinuum.Issuessuchasgroupeffectduetopile-soil-pile
interaction and theeffect of non-homogeneous soils
are not well-understood. Further research is clearly
required.
Thispaper describesasimplifiedtwo-stagenumer-
ical procedure for analyzing response of piles in
group subjected to excavation-induced lateral soil
movements. A numerical model, consideringof non-
homogeneous soils, is achieved using the finite dif-
ferencemethod and theconcept of shielding effects
is introducedto analysethepile-soil-pileinteraction.
Theassessmentisperformedbycomparingtheresults
fromthepresentedmethodwiththosefromcentrifuge
model testsandthepredictedresultsof centrifugetests
byLeung(Leunget al. 2000, Leunget al. 2003).
2 ANALYSISMETHOD
According to thetwo-stagemethod, theanalysis can
bedecomposedintotwocomponents(Poulos&Chen
1997). First, thefree-fieldsoil movement(withoutthe
presence of piles in the substratum) is obtained by
849
measurementor calculation. Second, theacquiredsoil
movement isimposedonanearbypiletocalculateits
response. Theflexural bendingof thepileismodelled
by an elastic beamwhile the complex phenomenon
of pile-soil interactionismodeledbylinearelasticsoil
springsbasedontheWinklermodel.Thelateral deflec-
tion equation for a single pile is formulated. Then,
considering the restriction of soil movement due to
pile-soil-pileinteraction, theshieldingeffect between
twopilesisimposedusingsimplifiedMindlinselastic
solutionforalateral pointloadinanelastichalf-space.
At last the response of group piles is obtained with
superpositiontheorem.
2.1 Free-field soil movement induced by the
construction of deep foundation pit
Analysismethodsfor estimatingfree-fieldsoil move-
ment include an empirical method, a finite element
methodandananalytical method.Asameanstoverify
thevalidity of thetwo-stagemethod, supposed free-
fieldsoil movement or that fromin-situmeasurement
canbeadopted. Bigotet al. (1982) clearlyshowedthat
the displacement-based method of analysis provides
very good predictions of bending moment profiles
andpiledeflectionsif themeasuredfree-fieldsoil dis-
placementisusedasinput, or if anaccurateprediction
of soil movement canbemade.
2.2 Analysis of a single pile
As an approximate method, the nonlinear springs
whichrepresenttheactual pile-soil interactionandten-
sioncracksdevelopedaroundpilesarenot takeninto
account. Inother words, thefollowinghypothesesare
adopted:
1 thepileisrepresentedbyanelasticbeambasedon
aWinkler subgradereactionmodel;
2 the complex phenomenon of pile-soil interaction
ismodeledusinglinear elastic soil springsandno
crackappearsbetweenpileandsurroundingsoils;
3 theeffect of axial loadonthepileisignored.
Thelinear elasticsoil springisrepresentedthrough
amodulusof subgradereaction, whichisdefinedas
wherek
z
hastheunitsof force/length
2
, p=soil reac-
tion per unit-length of pile, in unit of force/length,
y= the relative displacement between pile and sur-
rounding soils. Vesic (1961) analyzed an infinite
horizontal beamin an elastic foundation and com-
pared the results with those obtained by the use of
subgrade-reactiontheory, whichrelatedthemodulus
of subgradereactionk
z
totheelasticparameter E
s
and
s
of thesoil mass, asfollowin:
whereE
p
I
p
ispilerigidity.
Then, thegoverningdifferential equationof single
pileisgivenby:
inwhichU
t
(z) isthelateral deflectionof pilecaused
by excavation; h
x
(z) is thefree-fieldsoil movements
duetoexcavation; and iswrittenas
Theequation(3) canbesolvedeither byfinitedif-
ferencemethodorbyanalytical method.Theanalytical
methodhasbeenundertakenbytheauthorsbefore,and
canonlybeappliedforhomogeneoussoils. Inorderto
consider theinfluenceof non-homogeneoussoils, the
numerical finitedifferencemethodis adoptedinthis
paper.
Thetotal pilelengthL isdividedinton cellsandthe
lengthof eachcell is, withanodeineachcell, which
is0,1,. . . n 1, n. Inadditionton real cells, thereare
four additional imaginarycells(twoateachendof the
pileand additional nodes of 2, 1, n +1, n +2) to
implement boundaryconditions.
Thebasicdifferential equation(3) canbewrittenin
thefinitedifferenceformfor anyreal cell i, as
Thedifferentboundaryconditionsatthetopandtip
of thepileprovidefour different additional equations
(Poulos& Davis1990).
With free-head pile exerted load H and bending
moment M, equationsareasfollow:
Withfixedpilehead, theequationsare
Whendeflectionandrotationof pileheadispossible,
theequationischangedcorrespondingtotherelevant
boundarycondition.
850
Figure1. Computationmodel for twopassivepiles.
Similarly, theequationsfor cell n 1, n canalsobe
acquired, basedontheir boundaryconditions.
Substitutingtheequations of boundary conditions
into equation for nodes 0, 1, n 1 and n, then +1
unknowns, which represent the deflection of pile
in n +1 nodes, can be solved by the whole sys-
temof n +1 simultaneous equations. The rotation,
shear force and bending moment of the pile can
be obtained fromdeflection based on the theory of
material mechanics.
2.3 Interaction between two piles
In general, piles do not followexactly thefree-field
soil movement at the pile location and the soil pro-
fileisalsoalteredbypile. Thehinderedfree-fieldsoil
movement affects nearby piles, theamount of which
depending on therelativestiffness between pileand
surroundingsoils. Thisiscalledshieldingeffect.
The problemof interaction between two piles is
depictedinFigure1, wheretwopileswithpilespacing
s arerepresented. Bydiscretizingeachpileinton cells
withanodeineachcell, thefree-fieldsoil movement
inarbitrarynodei atthepositionof pile1ish
s1,i
andthe
deflectionof pile1causedisU
p1,i
. Thecorresponding
lateral soil deflectionisU
s1,i
, whereU
p1,i
isidentical
toU
s1,i
basedoncompatibilityof lateral displacement
betweenpileandsoil. Thelateral soil displacement in
nodei duetoshieldingeffectof pile1isthenexpressed
as
Thecorresponding lateral soil shielding displace-
ment inanynodej at thepositionof pile2duetosoil
displacement innodei at thepositionof pile1, is
where
ij
is the attenuation function of the lateral
shieldingmovementbasedonthesimplifiedMindlins
equationas follows (Poulos & Davis 1990). Suppose
theloadinarbitrary nodei at thepositionof pile1is
k
zi
U
s1,i
fromWinkler subgrademodel inthecaseof
activepiles. Basedon Mindlins equation, thecorre-
spondingsoil displacement inarbitrary nodej at the
positionof pile2(without thepresenceof pile2) due
toloadat nodei is
whereR
2
1
=s
2
+(z
j
z
i
)
2
, R
2
2
=s
2
+(z
j
+z
i
)
2
, z
i
and
z
j
is the depth of cell i, j, x is the distance between
twopilesinthedirectionof lateral soil deflection.The
coefficient
ij
canbethereforewrittenas
Withthesuperpositiontheorem, thelateral soil shield-
ing displacement in any node j at the position of
pile2is
The lateral equilibrium equation of pile 2 due to
shieldingeffect of pile1iswrittenas
whereU
t21
(z) is thecorresponding lateral deflection
of pile 2. The equation can also be solved by finite
differencemethodinthesameway.
2.4 Pile group analysis
For the usual case of a group of n piles, the total
responseof asinglepileinthegroup(i.e. thetotal dis-
placementandthetotal internal force) iswrittenasthe
sumof twocomponents:
1 displacement duetoexcavation-inducedfree-field
soil movement.
2 shieldingdisplacement resultingfrompile-to-pile
interaction (shielding effect) which decreases the
responseof singlepilecausedbyexcavation.
Considering the arbitrary pile i in the group, the
total displacement of thepileheadcouldbeobtained
bysuperposition
851
Figure2. Lateral soil movement inducedbyexcavation.
Figure3. Comparisonof free-headpileresponsefor pilelocated1mbehindwall.
Figure4. Comparisonof free-headpileresponsefor pilelocated3mbehindwall.
where U
tii
represents the response of pile i due
to excavation-induced free-field soil movement, and
U
tij
(i =j)representstheshieldingdisplacementof pile
i duetotheexistenceof pilej.
3 VERIFICATIONBY COMPARISONWITH
CENTRIFUGE MODEL TESTSANDFEM
ANAYSIS
3.1 Experiments
Only very limited field data are available so far,
especially in the case of pile groups. Leung et al.
(2000, 2003) have published a series of centrifuge
model tests on unstrutted deep excavation in dense
sand and its influence on an adjacent single pile
andpilegroupfoundationbehindtheretainingwall.
852
Due to approximate assumption about linear elastic
soil springs in the present simplified method, cases
in which soil experienced failure are not taken into
account. ThepredictedresultsbyLeunginthecaseof
singlepilewithfiniteelement methodarealsoshown
herein.
3.2 Analysis of single piles
Twotestsfor singlefree-headpilelocatedat1and3m
behind the retaining wall and two tests for restraint
head (fixed deflection-free rotation head and fixed
deflection-fixed rotation head) at 3m behind the
retainingwall arepresentedhere.Theprototypesquare
pilehasawidthof 0.63mandalengthof 12.5mwith
E
p
I
p
=22010
3
kNm. The free field soil displace-
ment profiles with depth at the location of the pile
Figure5. Comparisonof fixed(deflection)-free(rotation) headpileresponse.
Figure6. Comparisonof fixed(deflection)-fixed(rotation) headpileresponse.
areinstrumentedandshown in Figure2. Es =6z (in
MPa, z is the depth below ground surface, in m) is
appliedfor theanalysis as suggestedby Leunget al.
(2000). Figures 34 show comparisons between the
measuredandpredictedbendingmomentsanddeflec-
tion profiles alongafree-headpilelocatedat 1, 3m
fromthewall. Figures56givetheresultsfor thecase
of headrestraint. FE analysis by Leunget al. (2000)
gives a relatively better estimation than the present
method, especially at the position of the maximum
bending moment. Thereason maybetheassumption
of linear elastic soil springs along thepile. Thesoil
aroundthelongpileuptoadepthlessthan4timesthe
pilediameter mayhavereachedlimitedsoil pressures,
evenexperiencingverysmall displacement(Panet al.
2002). In Leungs works, theultimatesoil pressures
actingonpilewereintroducedtoconsider thiseffect,
853
Figure7. Configurationof pilegroup.
Figure8. Comparisonof freeheadpilegroup.
Figure9. Comparisonof front pileincappedheadpilegroup.
while they are not taken into account in the pre-
sentedmethod. However, thesimplifiedmethodstill
gives reasonableestimationwithlittlecomputational
effort andit canbeusedwithsomeconfidenceinthe
preliminarydesign.
3.3 Analysis of pile group
17group-piletestswithdifferent number of pilesand
differentconfiguresareshowninLeunget al. (2003).
Duetothelimitedlengthof thispaper, only2typical
pilegroups aredescribedhere. Theprototypesquare
pilehasawidthof 0.48mandalengthof 12.5mwith
E
p
I
p
=240 10
3
KNm
2
. Thepilecapwithathick-
nessof 0.55misplacedabovetheground, whichcan
be treated as rigid cap. Pile groups with four piles
infree-headandcapped-headaredescribedhere, the
configuration of which are shown in Figure 7. The
predictedresultsbythepresentsimplifiedmethodare
comparedwiththedatafromcentrifugetests inFig-
ures 810. Thecalculated results providereasonable
approximation to the centrifuge tests data for free-
head and capped-head pile groups. The discrepancy
betweenthepredictedandmeasuredbendingmoment
profiles is seemto besmall along theupper portion
of thepile, whileit isrelatively largealongthelower
portionof piles. However, deflectionsof capped-head
pilesshowthetendency that thefront pileisdragged
backbytherear filethroughconnectionof pilecap.
4 CONCLUSIONS
This paper presents a numerical analysis with finite
differencemethod for studying thebehavior of piles
subjectedtoexcavation-inducedlateral soil movement
innon-homogeneous soils. Responseof asinglepile
is determined by imposing theknown free-field soil
movement profileto thepassivepile. TheMindlins
854
Figure10. Comparisonof rear pileincappedheadpilegroup.
equationisemployedtocalculatetheshieldingeffect
of passivepilegroupsduetopile-soil-pileinteraction.
Comparisonswithcentrifugemodel testsconfirmthat
themethodprovidesreliableestimatesinsimpleway.
However, themajor limitationof themethodisthe
assumption of linear elastic soil springs, which pro-
vides only anupper boundestimateof themaximum
bending moments and pile deflections. Hence, the
analysisconsideringthenonlineareffectisstill needed
inorder todecisivelyassesstheresponsesof pilesdue
toexcavation-inducedlateral soil movements.
ACKNOWLEDGMENTS
demicDisciplineProject, Project Number: B308.
REFERENCES
Bigot, G., Bourges, F. &Frank, R. 1982. EtudeExperimental
DunPieuSoumisAuxPousseesLateralesDuSol. Revue
Francaise de Geotechnique 18: 2947.
Chen, L.T. & Poulos, H.G. 1997. Piles subjected to lateral
soil movements. Journal of Geotechnical and Geoenvi-
ronmental Engineering, ASCE, 123(9): 802811.
Goh, A.T., Teh, C.I. & Wong, K S. 1997. Analysis of
pilessubjectedtoembankmentinducedlateral soil move-
ments. Journal of Geotechnical and Geoenvironmental
Engineering, ASCE, 123(9): 792801.
Goh, A.T.C., Wong, K.S., Teh, C.I. & Wen, D. 2003.
Pile response adjacent to braced excavation. Journal of
129(4):
J uirnarongrit, T. & Ashford, S.A. 2006. Soil-pile response
toblast-inducedlateral spreading. II:analysisandassess-
ment of the py method. Journal of Geotechnical and
Geoenvironmental Engineering 132(2):163172.
Kitiyodom, P. & Matsumoto, T. 2003. A simplifiedanalysis
method for piled raft foundations in non-homogeneous
soils. Int. J. Numer. Anal. Meth. Geomech. 27:
85109.
Leung, C.F., Chow, Y.K. & Shen, R.F. 2000. Behavior of
pilesubjectedtoexcavation-inducedsoil movement.Jour-
nal of Geotechnical and Geoenvironmental Engineering
126(11): 947954.
Leung, C.F., Lim, J.K., Shen, R.F. &Chow,Y.K. 2003. Behav-
ior of Pile Groups Subject to Excavation-Induced Soil
Movement. Journal of Geotechnical and Geoenvironmen-
tal Engineering 129(1):5865.
Miao, L.F., Goh, A.T.C., Wong, K.S. & Teh, C.I. 2006.
Three-dimensional finite element analyses of passive
pile behavior. Int. J. Numer. Anal. Meth. Geomech. 30:
599613.
Pan, J.L., Goh,A.T.C.,Wong, K.S. &Teh, C.I. 2002. Ultimate
soil pressure for piles subjected to lateral soil move-
ments. Journal of Geotechnical and Geoenvironmental
Engineering 128(6):530535.
Pan,J.L.,Goh,A.T.C.,Wong,K.S.&Selby,A.R.2002.Three-
dimensional analysisof singlepileresponsetolateral soil
movements. Int. J. Numer. Anal. Meth. Geomech. 26:747
758.
Poulos, H.G. & Chen, L.T. 1997. Pile response due to
excavation-induced lateral soil movement. Journal of
Geotechnical and geoenvironmental engineering 123(2):
9499.
Poulos, H.G. & Davis, E.H. 1990. Pile foundation anal-
ysis and design. Florida: Pobert E. Krieger Publishing
Company.
Stewart, D.P. 1999. Analysis of piles subjected to embank-
mentinducedlateral soil movements. Journal of Geotech-
nical and Geoenvironmental Engineering, ASCE, May:
425426.
Vesic, A.S. 1961. Bendingof beamrestingonisotropicelas-
tic solid. Journal of. Engineering Mechanics Division,
ASCE, 87(2): 3553.
855
Elastic-plasticanalysisfor surroundingrockof pressuretunnel withlining
basedonmaterial nonlinear softening
L.M. Zhang& Z.Q. Wang
College of Science, Qingdao Technological University, Qingdao, P.R. China
ABSTRACT: Theelastic-nonlinear softening-residual plasticsurroundingrockmode1isanalyzed.According
tothetotal straintheory, therelationshipbetweenequivalentstressandequivalentstrainisdeducedfromuniaxial
compression of practical rock. Therelational expressions arerelated with triaxial stress (
,
z
, and
r
) and
triaxial strain(
,
z
, and
r
) of surroundingrock. Themechanismof loadbearingandactingrelationbetween
surroundingrockandsupportarestudied. Stressdistributionanddisplacementof brokenzone, plasticzoneand
elasticzoneof tunnel arepresentedout. Theultimatebearingcapacityof surroundingrockisgiven. Thecritical
pressureleadingto yieldfirstly of suroundingrock that causedby inner pressureis introduced. Theobtained
resultsaremorecloselytoactual valuesof surroundingrock. It ispointedout that thereareobviouslimitations
inKastener formula, whichisbasedonideal elastic-plasticmodel. Analysisshowsthat theideal plasticmodel
andthebrittlemodel arespecial casesof theproposedsolution.
1 INTRODUCTION
Kastnerssolutionisoftenusedinelastic-plasticanal-
ysis for surrounding rock of circular tunnel. There
are obvious limitations in Kastners formula, which
is based on ideal elastic-plastic constitutive model.
This leads to the Kastners solution is far away to
corespondingactual valuesinsuroundingrockof tun-
nel. Followingalongthepathof pioneeredbyKastner,
researcherssuchasMa(1995, 1996), J iangandZheng
(1996, 1997), Ma(1998, 1999),Yu(2002), Fan(2004),
Ren and Zhang (2001) and Pan and Wang (2004)
publisheddifferent solutionsfor surroundingrocksof
circular tunnel. However, thesesolutionsareresticted
toverysimplematerial models, suchassimplelinear
relationship between stress-strain. They are of lim-
ited practical value. This study successfully gets the
stress distribution laws of surrounding rock plastic
andbrokenzoneof tunnel accordingtothetotal strain
theory. Therelationshipbetweenequivalentstressand
equivalent strainis deducedfrompractical rock. The
relational expressions are related with triaxial stress
(
,
z
, and
r
) and triaxial strain (
,
z
, and
r
) of
surroundingrock.
2 ELASTIC-PLASTICANALYSISFOR
SURROUNDINGROCK
Fig. 1showsthegeometricmodel conditionof atunnel
inaplanestrainstatesubjectedtoapressuredifference
between its internal and external pressures. Wherea
and b are respectively the inner and outer radius of
tunnel, P
0
andPaarepressuresactingontheinner and
outersurfacesof tunnel,andRistheradiusof theinter-
faceof elasticandplasticzones.Surroundingrockmay
be generally divided into broken, plastic and elastic
zonesonthebasisof theirstates,asshowninFig.1.The
surroundingrockwithintheelasticzoneisinanelas-
tic state, withintheplastic zone, inastrainsoftening
state,andwithinthebrokenzone,inaresidual-strength
state. Sothesurroundingrockwithinthebrokenzone
isthedirect object of tunnel support.
2.1 Constitutive model of plastic zone
The tunnel may be simplified as an axisymmetrical
planestrain problem. Substituting r =r
3
,
z
=0 and
i
=
c
into theinterfaceof elastic andplastic zones,
theequivalent strainisdefinedby(Zheng, 1988)
If the volumetric strain of softening zone equals to
zero, wecanobtain
857
Figure1. Model of tunnel.
Figure2. Equivalent stress-equivalent strain.
Theequivalent stressisgivenby
The constitutive model of uniaxial compression is
calculatedfrom(Guo, 2004)
where
c
is a constant,
i
refers to the stress cor-
responded to peak strength,
i
refers to the strain
corresponded to peak strength. Theconstant
c
may
bedeterminedbytheresultsof aset of uniaxial tests.
Thesofteningsectiononequivalent
i
i
curvecan
beplottedasFig. 2shown. Wefindthat theassumed
rockmode1agreeswell withpractical rocks.
Theultimatebearingcapacity of surroundingrock
in complex stress state is analyzed in the follow-
ing parts. We consider that the strain component
of surrounding rock keep constant propotion, ie
z
:
:
r
=0:1: (1). Soit maybesimplifiedassim-
ple loading condition. According to the total strain
theory(Zheng, 1988), therelationshipbetweenequiv-
alent stress and equivalent strain can be deduced
fromEq. (4)
Therelational expressions arerelatedwithtriaxial
stress(
,
z
, and
r
) andtriaxial strain(
,
z
, and
r
)
of surroundingrock.
2.2 Stresses in the plastic zone
Itscorrespondingmechanical equilibriumequationis
Substituting(
r
)
r =r
2
=p
2
intoEq. (3) andEq. (5)
andusingtheEq. (6), thestresses intheplastic zone
canbeobtainedasfollows
2.3 Stresses in the elastic zone
Thestressesintheelasticzonemaybeexpressedby
Substituting thebounding condition on theinterface
of elastic and plastic zones (
e
)
r =r
3
+(
e
r
)
r =r
3
=
2p
0
=(
p
)
r =r
3
+(
p
r
)
r =r
3
into Eq. (7) and Eq. (8),
weget
858
Theplasticzonehasthesupporteffecttosurround-
ingrock. Eq. (9)showsthatforagivenp
0
, surrounding
rock can bebalanced by itself through adjusting the
plasticzone. Soitisalsonamedtheequilibriumequa-
tion of surrounding rock. Surrounding rock without
support has ultimate bearing capacity. If r
3
,
weget theultimatebearing capacity of surrounding
rock. Inpractical, thesurroundingrock iscollapedas
r
3
. However, it gives us the theoretical result.
Theultimatebearingcapacity of surroundingrock in
practical cannot belarger thanthetheoretical result
of surroundingrock.
2.4 Deformation in the plastic zone
Accordingtotheelastic-plastictheory, thetoatal strain
of plastic zone can be calculated by the following
formula(Zheng, 1988)
Theelasticstrainof plasticzoneisdefinedby
Theplasticstrainof plasticzoneisdefinedby
Thetotal strainof plasticzonecanbeexpressedby
whereE
c
referstoelasticmodulusof surroundingrock,
G
c
refers to shear modulus, j
c
refers to Poissions
radio, refers to plastic function. The plastic func-
tio is zero in elastic deformation. Using Eq. (13),
thedeformationontheinterfaceof elasticandplastic
zonescanbecalculatedfrom
SubstitutingEq. (7) intoEq. (13), wehave
CombiningEq. (14), wecansolveEq. (15) andgetthe
followingformula
where
859
Substitutingr =r
2
intoEq. (16), wehave
2.5 Stresses and deformation in the broken zone
The broken zone cannot bear the tangential stress,
so the tangential stress is zero. Its corresponding
mechanical equilibriumequationis
Thedeformationinthebrokenzonemaybeexpressed
by
CombiningEq. (17), wecansolveEq. (21) andgetthe
deformationformula
wheref =
(1j
2
0
)r
1
E
0
.
CombiningEq. (22) andEq. (20), weget
3 SUBMISSIONOF MATERIAL TOTHE
EDITOR
The lining can be considered as thick-wall cylin-
der in inner pressure p
a
and outer pressure p
1
. The
deformationof liningmaybeexpressedby
CombiningEq. (23) andEq. (24), weget
CombiningEq. (25), Eq. (20) andEq. (9), weget
860
Figure3. Ideal plasticmodel of rock.
Figure4. Brittlemodel of rock.
Using Eq. (26), we may get r
3
. Substituting r
3
into
Eq. (25), wefindthestressformulaof liningasfollows
4 DISCUSSIONOF RESULTS
(1) AccordingtoEq. (5), weknow
If
c
=0, weget =
c
.Thecurveof strengthafter
yield changes into horizon line, as shown in Fig. 3.
Which means the equation stands for ideal plastic
material. This suggests that the ideal plastic model,
such as Fenners solution and Kastners solution is
special casesof theproposedsolution.
If
c
, weget =0.Thecurveof strengthafter
yield changes into zero, as shown in Fig. 4. Which
means the equation stands for brittle material. This
shows that the brittle model is special case of the
proposedsolution.
If 0-
c
-, theequationstandsfor thebehavior
of post-failurestrength. Differentintegrationconstant
c
means different behavior of strength after yield,
as shown in Fig. 2. The equation betterly describes
thecharacteristic of thestrengthdroppingafter soft-
ening of surrounding rocks plasticity. It is suitable
for analyzing the stress of plastic zone of wall rock
openings.
(2) If the rock does not enter plastic state, but has
brokenzone, substitutingr
3
=r
2
intoEq. (25), wefind
If the rock does not enter broken state, but has
plasticzone, substitutingr
2
=r
1
intoEq. (25), wefind
If therock does not haveplastic zoneandbroken
zone,substitutingr
3
=r
2
,r
2
=r
1
intoEq.(25),wefind
(3) Substitutingr
3
=r
2
into Eq. (26), weget thefor-
mula of critical pressure leading to yield firstly for
suroundingrockthat causedbyinner pressure
5 CALCULATIONEXAMPLEAND
EXTENSIONOF ITSAPPLICATION
Typical cross-sectionof apressuretunnel isshownin
Fig. 1.Thedesignlengthof tunnel is150mwithinner
diameter of a =3.0m, outer diameter of r
1
=3.5m.
The mechanics parameters of rock and lining can
be gotten by test, E
0
=E
c
=30MPa, E
d
=60MPa,
j
c
=u
0
=u
d
=0.25,
d
=50MPa, r
2
=3.5m, p
0
=
1.3MPa, p
a
= 2MPa.
AccordingtoEq. (31), weget thecritical pressure
p
cr
a
=1.67MPa. As the inner pressure p
a
is greater
thanthecritical pressurep
cr
a
, plasticzoneoccurs.Then
weget r
3
=5.01mfromEq. (26). Theloosen range
of surrounding rock is obtained by Ultrasonic tests
to be 5.15m. It is very close to the theory result to
be 5.29mfromEq. (26). Table 1 shows stresses of
different positonof liningfromEq. (29).
861
Table1. Stressesof different positionof lining.
r/m 3.00 3.25 3.50
rc
(kPa) 1.96 1.74 1.47
L
c
(kPa) 1.92 1.62 1.37
6 CONCLUSIONS
Herewemaydrawthefollowingconclusions.
1 Theelastic-nonlinearsoftening-residual plasticsur-
roundingrockmodel isanalyzed. Accordingtothe
total straintheory, therelationshipbetweenequiv-
alent stress andequivalent strainis deducedfrom
uniaxial compression of practical rock, which is
relatedwithtriaxial stress(
,
z
, and
r
) andtri-
axial strain (
,
z
, and
r
) of surrounding rock.
Stress distribution laws of different position of
surroundingrock, themechanismof loadbearing
and acting relation between surounding rock and
support arestudied. Analysis shows that theideal
plasticmodel andthebrittlemodel arespecial cases
of theproposedsolution.
2 Different radial stressesof theinterfaceunder dif-
ferent conditions, such as elstic-plastic condition,
elastic-broken conditions are obtained. The ulti-
matebearingcapacityof surroundingrockisgiven.
The critical pressure leading to yield firstly for
surounding rock caused by inner pressureis also
obtained.
REFERENCES
Fan, H. & Yu, M.H. 2004. An analytic solution of elasto-
plasticpressuretunnel consideringmaterial softeningand
dilatancy. Engineering Mechanics 21(5):1624
Guo, Z.H. 2004. The strength and constitutive model of
conctrete. ChinaArchitectureandbuildingPress, Beijing
J iang, M.J. & Sheng, Z.J. 1996. Onexpansionof cylindrical
cavitywithlinearsofteningandsheardilatationbehaviour.
Journal of Rock Mechanics and Engineering 16(6):
550557
Ma, N.J. 1996. A newanalysis ongroundpressures around
openings. Journal of Rock Mechanics and Engineering
15(1): 8489
Ma, N.J. 1999. Multilinearstrengthattenuationmodel of rock
bodyandplasticareaof wall rockof openings. Journal of
Metal Mine 9: 1012
Ma, G.W., Iwasaki, S. & Miyamoto, Y. 1998. Plastic limit
analysis of circular plates with respect to unified yield
criterion. Int. J. Solids & Structure 43: 1137115
Ma, G.W., Iwasaki, S. &Miyamoto,Y. 1999. Dynamicplastic
behaviorof circularplatsusingtheunifiedyieldcriterion.
Int. J. Solids & Structure 36: 32573275
Pan, Y. & Wang, Z.Q. 2004. Research on relationshp of
load-displacementforcavernsurroundingrockwithstrain
nonlinear softeningng. Journal of Rock Mechanics and
Engineering 25(10):15151521
Ren, Q.W. & Zhang, H.C. 2001. A modification of fenner
formula. Journal of Hohal University 29(6): 109111
Yu, M.H. 2002. Advances in strength theory of materials
under complex stress state in the 20th century. Applied
Mechanics Reviews 55(3): 169218
Zheng, Y.T. 1998. Fundamentals of elastic-plastic-sticky
theory of rockmechanics. Coal industryPress, Beijing
862
Modificationof keyparametersof longitudinal equivalent model for
shieldtunnel
W. Zhu& X.Q. Kou
Geotechnical Research Institute, Hohai University, Nanjing, Jiangsu, P.R. China
X.C. Zhong
College of Civil Engineering, Hohai University, Nanjing, Jiangsu, P.R. China
Z.G. Huang
Suzhou Traffic Design and Research Institute, Suzhou, Jiangsu, P.R. China
ABSTRACT: Theshieldtunnel actuallyisaslenderconstructionwhichhasmanyweakspotsinthelongitudinal
direction. Becauseof thecharacter, thelongitudinal non-uniformdeformationof thetunnel wouldhappeneasily.
Withincreasingof shieldtunnel inChina, theproblemof thelongitudinal non-uniformdeformationhasbecome
attracted.Theequivalentcontinuousmodel whichhasclearconceptandcouldbecalculatedsimplyhasbeenused
widelyinprojectstoanalyzethelongitudinal performanceof thetunnel. Theshell-springmodel whichbaseson
thethree-dimensional FEM wasintroducedtoanalyzeandmodifytheerror of theequivalent continuousmodel
whichwasduetotheincreasingof thewidthof liningssegment. Thesuggestedrangeof theeffectiveratioof
thelongitudinal rigidityincommonsituationsof theshieldtunnel wasinduced.
1 INTRODUCTION
With the development of municipal works in china,
theshieldtunnelswhicharemainlyusedinsoftground
havebeeninawidespreadavailability.Thelayerwhich
has been traveled by the shield tunnel has become
more and more sophisticated. So the problem that
longitudinal non-homogeneousdeformationoccurson
thetunnel liningbecauseof itstravelingsophisticated
layershasbecomeattracted.
Now therearemainly two kinds of exiting longi-
tudinal calculationmodels whichareequivalent con-
tinuousmodel andlongitudinal beam-springmodel to
analysis thelongitudinal mechanical property of the
shieldtunnel linings.Theequivalentcontinuousmodel
simplifies thetunnel acontinuous beamwhichis on
theelasticfoundationinthelongitudinal directionand
recognizesthatthecrosssectionof thetunnel liningis
uniform. At thesametime, it considersthat thedefor-
mation of the tunnel in the longitudinal direction is
continuousat thelongitudinal joints. Theinfluenceof
the longitudinal joints of rings is simplified by the
way of equivalent rigidity. Theequivalent continuous
model has the clear conception and it is easy to be
calculated, so themodel has beenusedwidely inthe
actual projects.
Withtheadvancementinthedesignandconstruction
experience, thereisatrendtoincreasethewidthof the
segmental linings. For example, in J apan, the width
of theliningsegment was 0.80.9minpast, it was
developedto1.0min1975, andthewidthof thetun-
nel linings used in Gulf of Tokyo was indeed up to
1.5m; inChina, thewidthwasdevelopedfrom1.0m
for shanghai subway to 1.5mfor Guangzhou No.2
subway, and thewidth of 1.2mhas been commonly
adoptedinGuangzhousubwayNo.1line,Nanjingsub-
way, ShenzhensubwayandBeijingsubwayNo.5line.
Thesegment ringwidthof thetunnel whichtraverses
theYangzi River in Nanjing was 2.0m. Thenumber
of joints can be reduced by 20% by increasing the
widthof theliningsegment. Theproductioncost can
bereduced, and thewaterproof quality of thetunnel
canbeimprovedbecauseof reductionof thenumber
of joints.
Theequivalent continuous model discussedabove
assumes that thejoints of ringhavetheinfluenceon
mechanical performanceof rings segment along the
wholewidth. This model considers that thedelivery
of thedeformationandinternal forceof liningsalong
thelongitudinal directioniscontinuousandhomoge-
neous.Actuallywiththeincreasingof thewidthof the
segment ringof linings, theshear forceproducedby
863
theboltsof thejointsleadstotheincreasesinbending
stresswhichareconcentratedontheedgeof thering.
Thedelivery of thedeformationandinternal forceof
linings alongthelongitudinal directionis inhomoge-
neous especially at thejoints of rings becauseof the
concentrationof thestress. Theequivalentcontinuous
model whichisthe2D model isnt abletoreflect the
complexity of thedeformationof theliningsincreas-
inglywhilethewidthof segment liningsincreasing.
Thusitisaneedtomodifytheequivalentcontinuous
model andmakeit suitabletothedevelopment of the
shieldtunnel.
Huang& Zang(2003) consideredtheinfluenceof
theprestress of thelongitudinal bolts onthelongitu-
dinal rigidity of linings, andmodifiedtheequivalent
continuousmodel.
Liao (2002) improved the equivalent continuous
model, consideredthat theinfluenceareaof jointsof
ringsisfiniteandnot thewholeringlengthandsug-
gestedthat theeffectiveratioof longitudinal bending
rigidity of theShanghai subway was in therangeof
1/71/5.
Huang & Xu (2005) suggested the values of the
effective ratio of longitudinal and transversal rigid-
ity of shield linings in Shanghai areathough indoor
modelingtests.
In this study, the 3D numerical simulation called
theshell-spring model which could reflect theactu-
allymechanical performanceof thesegmental linings
properlywasintroducedtomodifytheequivalentcon-
tinuous model. In this model, the segments of the
shield tunnel lining aresimulated by theassociation
of shell elements, thejoints of thesegments aresim-
ulatedby thespring-elements, andeffects of thesoil
support arerealizedbythesoil-springelements.
2 THREE-DIMENSIONAL MODELLINGOF
SHIELDTUNNEL LINING
Figure1shows atypical sectionof shieldtunnel lin-
ing. In theshield tunnel construction, thesegmental
linings of each ring arejointed together by thebolts
in transversedirection of tunnel. Similarly, two con-
secutiverings areconnected together by thebolts in
thelongitudinal directionof tunnel. Inthenumerical
modeling, two coordinatesystems wereused: global
coordinates (x
, y
, z
) as showninFigure1andlocal
coordinates (x, y, z) which wereused to describethe
stressesanddeformationsfor asingleelement.
2.1 Shell element
Inthisstudyeachsegmental liningwasmodeledbya
meshof three-dimensional shell elements. Eachshell
element hasfour nodes. Eachnodehassixdegreesof
freedom, which areaxial and angular displacements
inx, yandz direction.
Figure1. Typical sectionof tunnel lining.
Figure2. Local coordinatesandglobal coordinates.
2.2 Segmental joint element
Each ring of tunnel lining is composed of several
segmental linings, which are connected together by
theboltsinthetransversedirectionof tunnel.Theseg-
mental joint element was used to simulatethejoints
between the segmental linings. Zhu and Tao (1998)
have put forward the beam-joint discontinuous ele-
ment to simulatethemechanical behavior of linings
andjointsinbothtransverseandlongitudinal direction
of tunnel.Theyadoptedthenon-linearnumerical anal-
ysisthroughtheconceptof Goodmanelement(1968).
Intheir model thejoint element exhibitstensileresis-
tancetosimulatetheboltconnection. Their numerical
result showsthat thebeam-joint discontinuousmodel
was suitablefor simulatingthenon-linearity of joint
anddiscontinuousdeformationof linings.
Inthisstudy, theconceptof Goodmanelementwas
used together with the three-dimensional joint ele-
ment. Theaxial, shear and rotation of thesegmental
joint element in the local coordinates system are
described by axial stiffness (K
sx
), shear stiffness in
864
Figure3. Segmental joint element.
Figure4. Longitudinal joint element.
radial direction (K
sy
), shear stiffness in tangential
direction (K
sz
) and rotational stiffness (K
), respec-
tively. Each segmental joint element has two coinci-
dence nodes. The discontinuity of the joint between
two segmental linings is denoted by therelativedis-
placementof twonodesinthelocal coordinatessystem
x, yandz direction. Eachnodeof thesegmental joint
elementhasfourelementhasfourdegreesof freedom,
whichareaxial displacement inx, y, z directionand
angular displacement inz direction.
2.3 Longitudinal joint element
Theshear model wasusedtosimulatethereinforcing
effectsinthelongitudinal directionof tunnel induced
by thestaggeredarrangement of liningsinthebeam-
joint discontinuous model. The shear model include
radial andtangential shear of thelinings. Inthisstudy,
thebeam-jointdiscontinuouselementwasextendedto
thethree-dimensional jointelement. Inadditiontothe
radial andtangential shear inthelinings, thetension
and compression along the longitudinal direction of
tunnel werealsoconsidered.
Theaxial, shearinradial andtangential directionof
thelongitudinal joint element inthelocal coordinates
system(x, y and z) are described by axial stiffness
(K
rz
), shear stiffness in radial direction (K
ry
), shear
stiffness in tangential direction (K
rx
), respectively.
Eachlongitudinal joint element has two coincidence
nodes. Thediscontinuityof joint betweentworingsis
denotedby therelativedisplacement of twonodes in
thelocal coordinates systems. Eachnodeof thelon-
gitudinal joint element has threedegrees of freedom
which areaxial displacement in x y and z direction.
Theradial andtangential shear, andaxial forceinthe
longitudinal directionof tunnel areconsideredineach
node. Compressivestiffnessisassumedtobeinfinite.
Tensile stiffness is the same as the stiffness of bolt,
k=EA/L, whereE isYoungs modulus of bolt, A is
cross-sectionarea, L islength.
2.4 Soil-spring element
Theinteractionbetweenthesegmental tunnel linings
andthesurroundinggroundwas modeledby thesoil
springinthis study. Inthis paper, theradial andtan-
gential spring was used to simulate the radial and
tangential forceactingonthesegmental liningsfrom
theground. Thespringswerearrangedover thewhole
circular sectionof thetunnel lining.
3 NUMERICAL SIMULATIONAND
ANALYSIS
The numerical simulation was used to calculate the
longitudinal rigidity of the shield tunnel linings. In
thisstudy, theeffect of thesoil didnt needtobecon-
sidered. Sothesoil-springelementsinthesimulation
systemwereneglected.Thesimulatedmodel isshown
in figure 5. There are four rings in the model and
boundary condition is selected to be one side fixed
and one side free. The outside and inside diameter
is 6.2mand 5.5m. the linings were constructed by
C50 concrete. The youngs modulus of concrete is
3.4510
4
N/mm
2
(seeTable1).
The arrangement of segmental linings is stagger
joint. Figure5showsthat therearesixsegmental lin-
ingsineachsection. SegmentA BandK haveacentral
angleof 67.5
, 68.0
and21.5
, respectively. Thelin-
ings are arranged in sequence and in stagger in the
longitudinal directionof tunnel, asshowninFigure5.
Thereare18bolts which connectedtwo consecutive
ringsof liningsinthelongitudinal directionof tunnel.
For analyzing thelongitudinal banding rigidity of
thelinings, themodel of numerical simulation(shown
in Figure 6) was adopted. In this study, the linings
is one end fix and one end free for loading. The
865
Figure 5. Arrangement of segmental linings in first and
secondsections.
Table 1. Influence of the lining segment width to the
longitudinal equivalent bendingrigidity.
Widthof liningsegment/m 1 1.4 1.8
Analytical solution/GPa 72.43 100.64 128.48
Numerical simulation/GPa 100.8 159.7 205.9
Figure6. Simulationmodel.
displacement of the free end of lining would be
observed.
3.1 Model parameters
Themodel parametersusedinthesegmental andlongi-
tudinal jointelementwerechosenbasedonthenational
andinternational shieldtunnelingexperienceandare
summarized. For thestiffness of segmental joint, the
valuesarethetotal stiffnessof onering. Inthenumer-
ical simulation, eachringis further dividedinto four
portions in the longitudinal direction of tunnel as
showninFigure4.
Figure7. Equivalent continuousmodel for shieldlining.
Themainlymodel parametersinnumerical simula-
tionareshownbelow.
Outsidediameter of thetunnel: D=6.2m; inside
diameter: d =5.5m;
Thethicknessof thelining: t =350mm;
Thediameter of thebolts: d =30mm;
Thelengthof thebolts: l =400mm;
Thenumber of thebolts: n =17;
Youngs modulus of the bolts: E
s
=2.1
10
5
N/mm
2
;
Poissonsratioof thebolts: =0.3;
Yieldstressof thebolts: [
y
]=6.410
5
N/mm
2
;
Breaking stress of the bolts: [
f
]=6.4
10
5
N/mm
2
Thegradeof theconcrete: C50
Youngsmodulus: E
c
=3.45104N/mm
2
;
Poissonsratio: =0.3.
3.2 The influence of the lining width to longitudinal
rigidity of tunnel lining
Theequivalentcontinuousmodel simplifiesthetunnel
acontinuousbeamwhichisontheelasticfoundationin
thelongitudinal directionandrecognizesthatthecross
section of the tunnel lining is uniform. At the same
time, it considers that the deformation of the tun-
nel in thelongitudinal direction is continuous at the
longitudinal joints. Theinfluenceof thelongitudinal
joints is simplified by theway of equivalent rigidity
(Figure7).
Theequivalent bandingrigidityof thetunnel:
Theeffectiverateof equivalent bandingrigidity:
couldbesolvedwiththisformulation:
866
WhereK
j
theelasticstiffnessrigidityof all longitu-
dinal bolts, K
j
=nk
j
k
j
theelasticstiffnessrigidityof onelongitudinal
bolt
I
c
inertiamoment of thesection
I
c
=

64
[D
4
(Dt)
4
]
E
c
Youngsmodulusof thelinings
A
c
cross-sectionareaof thelinings
thepositionangleof theneutral axis.
Figure8shows therelationshipbetweentheaver-
agevertical displacements of theobserved points at
thefreeendwhichcalculatedbythesimulationmodel
based on shell-spring model and theloads when the
width of linings is 1m, 1.4mand 1.8m. It could be
concludedthat thedisplacementswhichiscalculated
by thesimulationmodel aregenerally lower thanthe
analysissolutionwhichiscalculatedbytheequivalent
continuousmodel.
3.3 Analysis of simulation results
As the Table 1 shown, the equivalent rigidity of
the equivalent continuous model is lower than the
numerical values. Actually, theinfluenceareaof the
longitudinal joints to the whole linings deforma-
tion is finite. And the simplified assumption of the
model which considers theinfluenceis at thewhole
ring of lining is not correct. Theequivalent banding
rigidity of the equivalent continuance model should
be lower than the actual values. Besides that, the
influence of the arrangement of segmental linings
to the longitudinal performance of the tunnel isnt
considered by the equivalent continuous model and
the influence of that should not be ignored when
theinternal forcedelivers alongthelongitudinal lin-
ings. Thelocal deformationat thejointsof ringsalso
could influence the delivery of the endogen force.
This phenomenonis just so-calledlocal arrangement
effect.
Withtheliningwidthdeveloped, thedifferenceof
longitudinal equivalentrigiditybetweenthesimulation
model and theequivalent continuous model became
bigger (Figure8&Table2).
It was also approved that with thelining segment
width increasing the influence of the joints of rings
tothewholerings performancebecamereduced.And
thischangeabovecannotbereflectedbytheequivalent
continuancemodel whichgenerallyaddstheinfluence
of thejointsof ringstothewholeringsperformance
asthewidthof liningdeveloped. And, theerror of the
equivalentcontinuousmodel wouldincrease(Table2)
whilethewidthof liningsincreases.
AstheTable2shown,theeffectiveratioof thelongi-
tudinal rigidityhasbeenincreasedwhenthewidthof
Figure 8. Relationship between the load and the average
vertical displacement of theobservedpointsat thefreeend.
Table 2. Influence of the lining segment width to the
effectiveratioof thelongitudinal rigidity.
Widthof liningsegment/m 1 1.4 1.8
Equivalent continuous 1/12.14 1/8.74 1/6.85
model
Numerical simulation 1/8.73 1/5.51 1/4.27
Differencebetweenthe 39.1 58.6 60.4
abovetwomodels(%)
lining increased. Actually, with thenumber of joints
of rings and its influence decreased the longitudi-
nal performance of the segmental lining tunnel has
becomemoresimilar tothehomogeneoustunnel.
867
4 CONCLUSIONS
1 The equivalent bending rigidity which is used in
theanalytical solutionof theequivalent continuous
model is lower than theactual values. Becauseof
theassumptionof themodel, thecalculatedmethod
of theequivalent bendingrigidityintheequivalent
continuous model couldnt reflect theinfluenceof
thesegmental arrangement effect which generally
causes the delivering discontinuity of the internal
forceof theliningsat thejointsof rings.
2 With the width of lining increasing, the error of
the equivalent bending rigidity in the classical
equivalent continuousmodel wouldincrease.
3 When thewidth of linings is thegeneral rangeof
1.0m1.8musedintheshieldtunnel engineering
inChina,theeffectiveratioof thelongitudinal bend-
ingrigidityof thecommonsituationsof thesubway
couldbeintherangeof 1/8.731/4.27.
REFERENCES
Goodman, R.E., Taylor, R.L., & Brekke, T.L. 1968. A model
forthemechanicsof jointedrock. Journal of Soil Mechan-
ics and Foundation Division, ASCE, 94: 11211129.
Hashimoto,T., Zhu, H. H., &Nagaya, J. 1994.A NewModel
forSimulatingtheBehaviorof SegmentsinShieldTunnel.
Proceeding of the 49th Annual Conference of the JSCE,
12421243.
Huang, H.W., & Xu, L. 2006. Study on Transverse Effec-
tiveRigidityRatioof ShieldTunnels. Chinese Journal of
Geotechnical Engineering, 28(1):1118. [inChinese]
Huang, H.W., & Zang, X.L. 2002. Research and Analysis
on Longitudinal Deformation Characteristics of Shield
Tunnel. UndergroundSpace, 22(3):244250. [inChinese]
ITA WorkingGroup2, 2000, International tunnelingassoci-
ation, guidelines for the design of shield tunnel lining.
Tunneling and Underground Space Technology, 15(3),
pp.303331.
Lee, K.M., Hou, X.Y., Ge, X.W., & Tang Y. 2001. An
analytical solutionfor ajointedshield-driventunnel lin-
ing. International Journal for Numerical and Analytical
Methods in Geomechanics, 25(4):365390.
Zhu, H.H., Yang, L.D., & Chen, Q.J. 1996. Two design
models for segment joint of lining system in shield
tunnel. Engineering Mechanics, Supplement, 395399.
[inChinese]
868
Author Index
Ahmadi-adli, M. 295, 301
Ai, X.Q. 499
Alhieb, M. 649
Amorosi, A. 615
Antiga, A. 365
Arslan, U. 459
Assis, A.P. 519
Augarde, C.E. 785
Avdarsolkyzy, S. 751
Baimakhan, A.R. 751
Baimakhan, R.B. 751
Bakker, K.J. 243, 249, 255
Bao, H.L. 757
Bao, W. 817
Bao, W.Y. 215
Bao, X.H. 507
Batali, L. 187
Bezuijen, A. 3, 243, 249, 255,
261, 281, 349, 357
Boldini, D. 615
Bolton, M.D. 15
Broere, W. 621
Burghignoli, A. 627
Cai,Y.M. 465
Caporaletti, P. 627
Chan,Y. 471
Chang, G.M. 133
Chao, H.C. 67
Cheah, C.K. 447
Chen, C. 513
Chen, D.C. 579
Chen, J. 761
Cheng, L.C. 799
Chi,Y. 491
Chin, C.T. 67
Chiorboli, M. 365
Chissolucombe, I. 519
Cho, G.C. 45, 717
Choi, J.H. 669
Chuay, H.Y. 393
Chung, S.S. 567
Cong, L. 235
Cui, Z.D. 769, 843
Danaev, N.T. 751
Dashdorj, S. 751
Date, K. 635
Deng, A.Z.G. 269
Dias, D. 373
Dijkstra, J. 621
Ding, G.Y. 545
Ding, W.Q. 525
Ding,Y.H. 455
Du, J.H. 643
Eclaircy-Caudron, S. 373
Emeriault, F. 77, 207, 689
Erridaoui, A. 385
Farias, M.M. 519
Fillibeck, J. 275
Finno, R.J. 87
Gafar, K. 281
Gao, W.J. 607
Geraldni,Y.D.S. 655
Gong, Q.M. 381
Guatteri, G. 287
Guiloux, A. 385
Guo, B.H. 531, 799
Guo, W. 817
Gutierrez, M. 537
Hajialilue-Bonab, M. 295, 301
Han, C. 513
Han, X. 331, 545, 725
Hashimoto, A. 173, 405
Hashimoto, T. 99, 173, 307
He, C. 313
Hoefsloot, F.J.M. 357, 775
Hong, C. 781
Hou, J. 551
Hrustinec, L. 325
Hsi, J. 141
Hsiung, B.C.B. 393
Hu, M.Q.S. 385
Hu, Q.F. 595, 601
Hu, X.D. 319
Huang, H.W. 159, 215, 507, 595,
601, 643, 729, 757
Huang, M.S. 829, 849
Huang, R. 29
Huang, Z.G. 863
Idris, J. 649
Ito, H. 173
Iwata, N. 655
J ackson, P.G. 663
J eon, S. 781
J eong, K.H. 669
J iang, J. 419
J iang, X.H. 441
Kasper, T. 663
Kastner, R. 77, 207, 373, 689
Katebi, H. 295, 301
Kikumoto, M. 709
Kim, D.H. 555
Kim, I. 677
Kim, J. 717
Kim, S.B. 485
Kim, S.M. 399
Kim, U.Y. 555
Kokubun, T. 141
Konda, T. 173, 307, 405
Kong, X.L. 153
Koshima, A. 287
Kou, X.Q. 863
Koungelis, D.K. 785
Kulmaganbetova, Zh.K. 751
Kuzma, J. 325
Lam, S.Y. 15
Lam, T.S.K. 561
Lanquette, F. 385
Latt, K.M. 579
LeBissonnais, H. 385
Lee, H.Y. 555
Lee, I.M. 45, 823
Lee, J.S. 555
Lee, S.D. 669, 677
Lee, S.P. 555
Lee, W. 567
Lee,Y.J. 485, 683
Leung, L.P.P. 471
Li, J. 499, 817
Li, J.P. 573, 697
Li, L. 147
Li, X.X. 165, 791
Li,Y. 689
Li,Y.Q. 153
Li, Z.H. 159
Li, Z.X. 331, 725
Liang, F.Y. 697, 849
Liang, Q.H. 455
Lin,Y.L. 165, 791
Liu, D.P. 413, 587
Liu, G.B. 227, 413, 419, 433,
587, 805
Liu, J.H. 465
869
Liu, S.T. 427
Liu, T. 227, 433
Liu, X. 491
Lopes, R. 287
Lu, T.D. 223
Lu, T.K. 531, 799
Lu, Z.P. 805
Lftenegger, R. 193
Ma, X.F. 307, 477
Ma, Z.Z. 477, 587
Mair, R.J. 635, 703
Makhanova, A.A. 751
Marcu, A. 187
Marlinge, J. 385
Marshall, A.M. 703
Matsumoto, A. 761
McNamara, A.M. 735
Mei,Y.B. 441
Meissner, S. 459
Michael, J. 459
Mitew-Czajewska, M. 201
Nagaya, J. 307
Nakahara, E. 709
Nakai, T. 655, 709
Negro, A. 811
Ng, C.C. 447, 579
Ng, C.W.W. 419, 433
Ng, N.W.H. 471
Niinomi, M. 655
Osborne, N.H. 447, 579
Osman, A.S. 15
Ota, H. 173, 405
Pang, P.L.R. 109, 567
Phienwej, N. 181
Pi, A.R. 319
Pieroni, M.R. 287
Popa, H. 187
Qiao, H.C. 441
Qiu, D.W. 455
Quick, H. 459
Ravaglia, A. 287
Roberts, K.J. 567
Rudi, J. 579
Ryckaert, J. 385
Rysbaeva, G.P. 751
Sabetamal, H. 295, 301
Salgaraeva, G.I. 751
Sanders, M.P.M. 261, 281
Scarpelli, G. 627
Scharinger, F. 193
Schweiger, H.F. 193
Sfriso, A. 121
Sfriso, A.O. 335
Shahin, H.M. 655, 709
Shi, Z. 817
Shin, J.H. 823
Shin,Y.J. 823
Siemi nska-Lewandowska, A. 201
Soga, K. 281, 635
Song, K.I. 717
Song, T.T. 343
Song, X.Y. 829
Sozio, L.E. 837
Sugimoto, M. 761
Tabata,Y. 709
Talmon, A.M. 3, 349, 357
Tan, G.H. 579
Tang,Y.Q. 769, 843
Taylor, R.N. 627, 735
Teng, L. 313
Thiebault, H. 385
Tu, M. 551
vanTol, A.F. 261, 281
Verdel, T. 649
Viel, G. 385
Vogt, N. 275
Wang, J. 799
Wang, J.Q. 235
Wang, K.S. 331, 725
Wang, R. 413, 587
Wang, R.L. 465, 573
Wang, X.M. 729
Wang, Z.Q. 857
Wang, Z.W. 427
Wong, K.K.W. 471
Xia, C.C. 537
Xie, K.H. 153
Xie, X.Y. 729
Xu, Q.W. 477
Xu, S.F. 223
Xu,Y. 525
Yan, J.Y. 313, 573
Yan,Y.R. 595
Yanagawa, T. 173, 405
Yang, H.Y. 399
Yang, J.W. 669
Yang, L. 513
Yang, M. 147
Yang, S.L. 455
Yao, C.P. 601
Yao, G.S. 697
Yao, H. 607
Yao, J. 735
Ye, B. 99
Ye, G.L. 99, 307
Yoo, C. 485
Yoo, C.S. 683
Yoon, S.G. 399
You, G.M. 743
You, K. 781
Yuan,Y. 491
Zghondi, J. 207
Zhang, C.R. 849
Zhang, D.M. 215, 757
Zhang, F. 655, 709
Zhang, H. 141, 223
Zhang, J. 227
Zhang, L.M. 857
Zhang, M.X. 551
Zhang, Q.H. 269
Zhang, X. 769, 843
Zhang, Z.X. 689
Zhao, H.L. 491
Zhong, X.C. 863
Zhou, H.B. 607
Zhou, J. 153, 235
Zhou, K.Q. 455
Zhou, S.H. 343, 381
Zhu, H.H. 791
Zhu, J. 817
Zhu, W. 863
Zhu,Y.M. 441
870
GEOTECHNICAL ASPECTS OF
UNDERGROUND CONSTRUCTION IN SOFT GROUND
C.W.W. Ng, H.W. Huang & G.B. Liu, editors
This volume comprises a collection of four special lectures, six general
reports and 112 papers presented at the Sixth International Symposium of
Geotechnical Aspects of Underground Construction in Soft Ground (IS-Shanghai) held
between 10 and 12 April 2008 in Shanghai, China.
The Symposium was organised by Tongji University and supported by China
Civil Engineering Society, Chinese Society for Chinese Society for Rock
Mechanics and Engineering, Geotechnical Division, the Hong Kong
Institution of Engineers, Hong Kong Geotechnical Society, Hong Kong
University of Science and Technology, Science and Technology Commission
of Shanghai Municipality, Shanghai Changjiang Tunnels and Bridge
Development Co. Ltd and Shanghai Society of Civil Engineering. This was
the most recent symposium in a series of symposia starting in New Delhi,
India (1994), followed by symposia in London, UK (1996), Tokyo, Japan
(1999), Toulouse, France (2002) and Amsterdam, the Netherlands (2005).
The four invited special lectures from A. Bezuijen, Huang Rong, M.D. Bolton
and I.K. Lee are about Processes around a TBM, Overview of Shanghai
Yangtze river tunnel project, Supporting excavations in clay - from analysis
to decision-making and Underground construction in decomposed residual
soils, respectively.
The six general reports from Richard Finno, Tadashi Hashimoto, Alejo Sfriso,
C.T. Chin, Richard Pang and Richard Kastner, cover the symposium themes:
1. Analysis and numerical modelling of deep excavations
2. Construction method, ground treatment, and conditioning for tunnelling
3. Case histories
4. Safety issues, risk analysis, hazard management and control
5. Physical and numerical modelling
6. Calculation, design methods, and predictive tools
This volume provides a valuable source of information on the the state-of-
the-art in geotechnical engineering associated with the design, construction
and monitoring of tunnels and excavations in soft ground, and will be of
interest to academics and professionals involved in these areas.
GEOTECHNICAL
ASPECTS OF
UNDERGROUND
CONSTRUCTION IN
SOFT GROUND
Ng
Huang
Liu
editors
an informa business
cover_Xie_Ng_D.qxd 14-10-2008 15:06 Pagina 1

Geotechnical Aspects of Underground Construction in Soft Ground - C. NG., Et Al., (CRC, 2009) BBS PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Geotechnical Aspects of Underground Construction in Soft Ground - C. NG., Et Al., (CRC, 2009) BBS PDF

Hochgeladen von

Copyright:

Verfügbare Formate

GEOTECHNICAL ASPECTSOF UNDERGROUNDCONSTRUCTIONINSOFT GROUND

PROCEEDINGS OF THE 6TH INTERNATIONAL SYMPOSIUM (IS-SHANGHAI 2008),

theyieldstrengthof thegrout, ands

Figure3. Incremental displacement fields.

C within 120 minutes is selected to protect

=9.0kPaas thecohesiondecreases from120kPa

) horizontal jet grouting in Sao Paulo and

for excavations approx.

Miscellaneous 0.90 18.20 25.00 18.00

are the effective

. It is observed that the soil mass strength

, H =10m. As a whole, the critical height

Deg (1) (2) (1) (2) (1) (2)

=0, and situation

=0, when b changes from0 to 1, the resultant

=0, liquid limit I

SwiatStation. Warsaw: GeotekoSp. zo.o., SGGW,

) andit only presents adilatant

Table 4. The negative pore-water pressure data of soil

1.00 1.60 0.50 1.20 2.0

1.40 2.26 0.50 1.20 2.0

1.70 2.77 0.50 1.50 2.0

2.20 2.00 2.00 1.80

1.84 1.70 1.65 1.51

1.85 1.73 1.67 1.55

jet groutingtreatment executedin

solution became feasible due to advances in

horizontal jet groutingscheme.

Horizontal J et Grouting, fig. 10).

consolidation treatment is thestate-of-the-

consolidation scheme proved to be

kPa peak kPa residual

C respectively to achieve the uniaxial com-

C. The dimension of the

=dimensionless heat release

=the dimensionless critical velocity,

=dimensionless heat release; V

). Each extensometer has a

(kPa) 20 100 1000

are respectively short and long termYoungs

and thebacksideground surfacesettle-

and backside ground surface set-

and backside ground settlement areaA

were used in the

for angles and (1mm+2mm

for angles. For long tunnels, the accuracy

) with Cu under undrained set-

(kPa) 0.7251845=1337.6 0.725178=129 0.725339=245.8

=cohesion, =internal fric-

) is the joint probability density

, t) canbecalculated. After theintegral with

is the weight vectors of sub-factors, R

D300: PrimaryStandardConcreteD310: ConcreteContainingFlyAshD320: ConcreteContainingMineral Powder

),Youngmodule(3; 6; 10; 15; 20

C. Similarly intheUnitedStates sig-

North of theequator experi-

Here, thecriteriononrankof lossdiffersfromcountrytocountry. InChina, thelossisdisastrouswhen5liveslost.

andastill watertableat+63mmodel coordinate

and cohesion (c) was 6.0kPa. Modulus of

forthemodel granularmaterial, respec-

) was assumedtobe0.1kPa. Thevariationof

aresurchargeandsoil weight coef-

with a 6moverlap, with a 0.7m