You are on page 1of 306

National TechnicalUniversityofAthens

SchoolofCivilEngineering

GeotechnicalDivision

.
...

BuriedPipelineSubjectedtoNormal
andReverseTectonicFaultRupture
DiplomaThesis

Vasileiadis Michail
Supervisedby:

ProfessorG.Gazetas
Ast.Prof.I.Anastasopoulos
2012

,
,

.
. .

.
,
.
, , . .
,
.
,
,
.
..

.
, , .
,

.

ii


,
,.

.

, ,

.,
.
,
Winkler,
.

.

(D/t).
.
,D/t,
,
.

iii

iv

Synopsis

The present diploma thesis investigates the mechanical behavior of a buried steel
pipeline, commonly used for oil and gas transportation, subjected to DipSlip
tectonic fault rupture. The investigation is based on numerical simulation with the
finiteelementcodeABAQUS2011.Themainobjectiveofthisstudyistodevelopa
proper model that accounts for large displacements and strains, inelastic material
behaviorofthepipeandthesoil,aswellaspropercontactandfrictionatthepipe
soil interface. Different approaches are examined concerning the pipe boundaries,
finally leading to the development of a hybrid boundary capable of simulating the
naturalcontinuationofthepipe.Themodelisusedtostudythenormalandreverse
fault ruptures. The main results are presented in terms of critical fault
displacement for different performance criteria, which are monitored throughout
theanalyses,withrespecttotheD/tratio.Otherparametersthatareexaminedare
the soil layers depth, the stressstrain relation of the pipe steel, and the internal
pipepressure.

vi


TABLEOFCONTENTS

CHAPTER1
INTRODUCTIONANDSTUDYMOTIVATION.................................................................1
1.1Introduction...................................................................................................3
1.2Faults..............................................................................................................4
1.3Pipelines.........................................................................................................6
1.4LiteratureReview...........................................................................................9
1.5MotivationoftheStudy...............................................................................13
FiguresofChapter1....................................................................................................21

CHAPTER2
PROBLEMANDMODELDEFINITION...........................................................................55
2.1ProblemDefinition.......................................................................................57
2.2ModelandMethodofAnalysis....................................................................59
2.2.1FiniteElementModel..........................................................................59
2.2.2SoilConstitutiveModel........................................................................61
2.2.3PipelineProperties..............................................................................69
2.2.4ModesofPipelineFailure....................................................................71
FiguresofChapter2....................................................................................................77

vii

CHAPTER3
PIPEBOUNDARIES......................................................................................................89
3.1FreePipeEndsModel..................................................................................92
3.2FixedPipeEndsModel.................................................................................95
3.3HybridBeamModel.....................................................................................98
3.4EquivalentSingleSpringModel..................................................................106
FiguresofChapter3..................................................................................................109

CHAPTER4
REVERSEFAULTRUPTURE........................................................................................153
4.1BriefReviewoftheFixedandFreeEndsModel........................................156
4.2HybridBeamModel...................................................................................157
FiguresofChapter4..................................................................................................163

CHAPTER5
SOILLAYERDEPTHEFFECT........................................................................................185
5.1FreeFieldModel........................................................................................188
5.2NormalFaultRupture................................................................................190
5.3ReverseFaultRupture................................................................................192
FiguresofChapter5..................................................................................................197

viii

CHAPTER6
PIPESTEELWITHHARDENINGBEHAVIOR................................................................225
6.1NormalFaultRupture................................................................................227
6.2ReverseFaultRupture................................................................................228
FiguresofChapter6..................................................................................................231

CHAPTER7
INTERNALPIPEPRESSURE........................................................................................247
7.1NormalFaultRupture................................................................................250
7.2ReverseFaultRupture................................................................................251
7.3InternalPipePressurewithHardeningPipeSteelBehavior......................252
FiguresofChapter7..................................................................................................257

CHAPTER8
CONCLUSIONSANDPROPOSALSFORFURTHERRESEARCH....................................275
8.1Conclusions................................................................................................277
8.2ProposalsforfurtherResearch..................................................................279

REFERENCES.......................................................................................................281

ix





Chapter 1
Introduction and study motivation

Introductionandstudymotivation

1.1 Introduction

Buried pipelines
used to transport

Buried pipelines are commonly used to support electric power distribution,


telecommunications, and to transmit water, sewage, liquid and gas. Pipelines, also
mentioned as lifelines, often cross tectonically active areas and may cross active
faultscapableofproducinglargeearthquakesandlargegrounddeformations.There
arevariousexamplesofearthquakesthatseverelydamagedburiedpipelinessuchas
the earthquakes of San Fernando 1971, Managua 1972, Haicheng 1975, Tangshan
1976,MiyagikenOki1978,Northridge1994,Kobe1995,ChiChi1999,Kocaeli1999
and the more recent ones, those of Chile 2010, Christchurch 20102011 and Japan
2011. Based on the observed damage mechanisms of buried pipelines, seismic
effectscanbeeithercausedbytransientstrainandcurvatureinthegrounddueto
traveling wave effects,or caused by permanent ground deformations such as fault
deformation,landslide,andliquefactioninducedsoilmovements.Amongthem,the
groundmovementsofactivefaults,eventhoughrelativelylimitedtosmallregions,
canhavethemostsevereeffectsonburiedpipelines.Therearefewcasesinwhich
pipelines were damaged only by wave propagation (Mexico City 1985Michoacan
earthquake).Faultmovement,though,isappliedonthepipelinequasistaticallyand
mightcausesignificantdamagetothepiperegardlesstheseismicintensity.

Introductionandstudymotivation

1.2 Faults

Althoughthisstudyfocusesonfaultsthatpropagatethroughtheuppersoillayers,it
isconsideredreasonabletopresentabrieftheoreticalbackgroundconcerningfaults
andtheirassociationwiththegenesismechanismofearthquakes.Anactivefaultisa
discontinuity between two portions of the earth crust along which relative
movements can occur. The theory that is currently considered to be the most
accurate for the description of this mechanism is the Elastic Rebound Theory.
Because of friction andthe rigidityof the rock, the rocks cannot glide or flow past
eachother.Rather,stressbuildsupinrocksandwhenitreachesalevelthatexceeds
the rocks strength, the accumulatedpotential energyis dissipated by the release
ofstrain, which is focused into a plane along which relative motion is
accommodated,thefault.Strainisbothaccumulativeandinstantaneousdepending
on therheologyof the rock. The ductile lower crust and mantle accumulates
deformation gradually viashearing, whereas the brittle upper crust reacts by
fracture instantaneous stress release to cause motion along the fault. A fault in
ductilerockscanalsoreleaseinstantaneouslywhenthestrainrateistoogreat.The
energyreleasedbyinstantaneousstrainreleasecausesearthquakes.Therearetwo
ways to simulate the faults surface: the Asperities model, a model of
inhomogeneousmacroscopicroughnessandtheBarriersmodel,wherethesurface
isconsidereddiscontinuouswithareasofvariableshearstrength.


Basic types of faults

The two sides of a fault are known as thehanging wallandfootwall. By definition,


the hanging wall occurs above the fault plane and the footwall occurs below the
fault.Faultsaredistinguishedonthebasisofthemovementofthefootwallrelative
tothehangingwall.

Introductionandstudymotivation

Strike slip faults

Strikesliparethefaultsthatprimarilyexhibithorizontalmovement(Figure1.2.1).
Thefaultsurfaceisusuallynearverticalandthefootwallmoveseitherleftorrightor
laterallywithverylittlevertical motion.


Dipslip faults

Dip slip faultsare those in which vertical displacement primarily occurs. They are
divided in to 2 categories: normal and reverse faults. Normal (Figure1.2.2) are the
faults in which the hanging wallmoves downrelative to the foot wall due to
tensionalstress.Thetermnormaldoesnothavethesenseofcommonandwhat
isnormalaboutthemisthattheirmovementtendstofollowthegravitationalpull
onthefaultblocksinvolved.Areversefault(Figure1.2.3)isoneinwhichthehanging
wallmoves uprelative to the foot wall due to compression. If the hanging wall is
pushedupandthenoverthefootwallatalowangleitiscalledathrustfault.


Obliqueslip faults

Afaultwhichhasacomponentofdipslipandacomponentofstrikeslipistermed
anobliqueslip fault (Figure1.2.4). Nearly all faults have some component of both
dipslip and strikeslip, so defining a fault as oblique requires both dip and strike
componentstobemeasurableandsignificant.Finally,ithastobeclarifiedthatnot
all of the faults reach the surface. Whether the rupture will propagate up to the
surface depends on various parameters such as the magnitude of bedrock
movement, the depth of the soil layer, the faults angle and the soil properties.

Introductionandstudymotivation

1.3 Pipelines

Overview

Oil and gas pipelines are usually made fromsteelof various strength and
characteristics(Table1.1)orfromplastic(HDPE),withinnerdiametertypicallyfrom
4to48inches(100to1200mm).Anothermaterialthathasbeenusedistheglass
reinforced plastic (GRP) a composite material made of plastic reinforced by glass
fibers.Mostpipelinesaretypicallyburiedatadepthofabout0.9to2.5m.Theoilis
keptinmotionbypumpstationsalongthepipelinesitedattheoriginatedstationof
thelineandat30to160kilometerintervalsalongthelengthofthepipelinewhile
the gas is kept in motion by compression stations sited at 70 to 500 kilometer
intervals,dependingonpipelinedesign,topographyandcapacityrequirements.The
usualflowspeedisofabout1to6meterspersecondfortheoiland20to60m/sfor
thegas.
Asarulepipelinesforallusesarelaidinmostcasesunderground.Howeverinsome
cases it is necessary to cross a valley or a river on apipeline bridge. Pipelines for
petroleum running through permafrost areas often run overhead in order to avoid
melting the frozen ground by hot petroleum which would result in sinking the
pipeline in the ground. Laying the pipeline on the ground is also preferred
sometimesincaseswherethepipelinecrossesafaultofimportantoffsetpotential
suchastheTransAlaskaoilpipeline(Figure1.3.1).
Abriefcomparisonofburiedtosurfacepipelinesleadstothefollowingobservations:
Buriedpipelines

are more protected against temperature fluctuations and weather


phenomena in general, as well as against accidents like crashes and
terroristattacks.

Haveasmallimpacttothesurroundingenvironment.

Introductionandstudymotivation

Can extend for thousands of kilometers without interfering and


causingproblemstootherinfrastructuresystemssuchasroads.

Have the same, or even smaller, cost compared to the on ground


pipelines,sincetheexcavationcostiscounterbalancedifnotexceeded
bythecostofthespecialfootingsneededforthesurfacepipelines.

Surfacepipelines

Canbeeasilyinspected

Canaccommodate,generally,biggerfaultoffsets

Pipelinescanbecategorizedas:
Gatheringpipelines
Group of smaller interconnected pipelines forming complex networks with the
purposeofbringingcrudeoilornaturalgasfromseveralnearbywellstoatreatment
plantorprocessingfacility.

Distributionpipelines
Composed of several interconnected pipelines with small diameters, used to take
theproductstothefinalconsumer.

Transportationortransmissionpipelines
Mainlylongpipeswithlargediameters,movingproducts(oil,gas,refinedproducts)
betweencities,countriesandevencontinents.Transmissionpipelinesarethemain
arteriesoftheoilandgasbusiness.

Introductionandstudymotivation

Construction Sequence

Before any other step of the construction, the construction corridor has to be
clearedofbrush,trees,largerocksandanyotherobstructions.Then,thetrenchin
whichthepipewillbelaidisexcavated(Figure1.3.2).Machineexcavatorsareused
for the excavation along with any necessary hand digging and general field labor
activities. The pipeline parts are placed alongside the trench. Next step is the
bendingofthepipes.Since,therouteofthetrenchisnotaconstantlinethepipeline
shouldgetbendedatcertainpoints,inordertoconformtothetrench.Afterwards,
thejointsofthedifferentpartsareweldedintobiggerpipeparts.Althoughthepipe
would arrive at the rightofway with a corrosionresistant coating, crews apply
additionalcoatingtotheweldedareasandrepairanydamagetothefactoryapplied
coating to prevent corrosion. The pipe coating is a special material that prevents
waterfromcontactingthesteelofthepipeandcausingcorrosion.Whenthewelding
and coating are complete, the pipe is suspended over the ditch by sideboom
tractors, which are crawler tractors with a special hoisting frame attached to one
side(Figures1.3.4,1.3.5).Then,thepipelineisgraduallyloweredtothebottomof
theditch.Inrockysoilorsolidrock,itissometimesnecessarytoputabedoffinesoil
inthebottomoftheditchbeforeloweringthepipeline.Thefinefillmaterialprotects
thepipecoatingfromdamage. Thefinalcutsonthepipearemadetoconnectthe
entire pipeline and coat the final tiein welds (Figure 1.3.6). The pipeline is then
testedtoproveitsstructuralsoundnessandabilitytofulfillitsdesignfunction.The
line is filled with water and then pressurized to check for leaks. Finally, the trench
getscoveredbythebackfillandthesurfacegetscleanedupandrestoredifpossible
toitspreviousstate.

Introductionandstudymotivation

1.4 Literature Review

Theproblemofpipelinescrossingactivefaultshasbeenapproachedusinganalytical,
numerical,aswellasexperimentalmethods.Theinitiatorsoftheanalyticalstudies
wereNewmarkandHall(1975)(Figure1.4.1).Intheirpioneerworktheystudiedthe
abilityofacontinuouspipetoresistlargedisplacementscausedbystrikeslipfaults
Focusing upon cases where the fault rupture provokes severe elongation of the
pipeline,soastensionistheprevailingmodeofdeformation.Intheirmodelthepipe
isassumedtobefirmlyattachedtothesurroundingsoilattwoanchorpointslocated
at distance L from fault trace. These anchor points correspond to elbows, thrust
blocks,tieinsandotherfeatures,whichcandevelopsubstantialresistancetoaxial
movement. The authors neglected the bending stiffness of the pipe as well as the
lateral interactions at the pipesoil interface. They discovered that the bearing
capacityofaburiedpipelinetofaultmovementdependsonthesoilconditions,pipe
andfaultcharacteristicsandthatminimizingthelongitudinalresistanceofthesoilto
the pipe motion maximizes the pipe resistance. They have also suggested that the
pipe should be placed in a trench with shallow sloping sides so that it can
accommodateitselftothetransverseaswellasthelongitudinalcomponentsofthe
fault movement. Subsequently, Kennedy et al (1977) (Figure 1.4.2) extended
Newmark and Halls procedure to determine the pipe capacity to resist fault
movement by taking uniform passive soil pressure and the large deflection theory
intoconsideration.Itisassumedthatthepipelineisaflexiblecabledeformedintoa
single constant curve approaching asymptotically to the undistorted portion of the
pipeline.Inthismodel,onlytheaxialtensileforceatthepointofinflectionisused
forequilibrium,noflexuralresistancewasconsidered.Thereforewiththeomission
oftheflexuralrigidityofthepipe,thismodelcannotsatisfytheequilibriumcondition
for a pipeline crossing a strikeslip fault or obliqueslip fault that will cause
compressioninthepipeline.VougioukasandCarydis(1979)(Figure1.4.3)proposed
a model applicable both for strikeslip and for dipslip faults, either normal or
reverse, and calculated the ductility demands for the pipelines to resist large fault
displacements. Wang and Yeh (1985) (Figure 1.4.4) proposed a localized large

Introductionandstudymotivation

deflection beam model to analyze pipeline behavior crossing the strikeslip fault.
They modeled a large deflection pipe as a constant curvature curved segment and
the remaining small deflection pipe as a semiinfinite beam on elastic foundation.
This model includes the bending rigidity of the pipe, a shear force at the point of
inflectionofthecurvepipecrossingthefaultzone,andaboundaryconditionrelated
to semiinfinite beam on elastic foundation at some distance away from the fault
zone.Theirmethodologyrefersonlytostrikeslipfaultsandreliesonpartitioningof
thepipelineintofourdistinctsegmentstwointhehighcurvaturezoneonbothsides
of the fault trace, and another two outside this zone. This model yielded more
realistic results provided that the appropriate parameters are chosen. Chiou et al.
(1994)cametotheconclusionthatthecurvatureofthedeformedpipeisunknown
and can hardly be predicted in advance. Moreover, the constant curvature model
overestimatesthestiffnessoftheburiedpipeandconsequentlyunderestimatesits
strain. Karamitros et al. (2006) (Figure 1.4.5), trying to improve the models of
KennedyandWang,proposedamoresophisticatedanalyticalmethodforstrikeslip
faults, which they later extended to be applicable for normal faults as well (2011).
They presented an analytical methodology, refining existing methodologies using a
combination of beamonelasticfoundation and beam theory, and computed axial
force, bending moment and maximum strain on the pipeline. They also compared
their analytical predictions with results from a threedimensional finite element
model, which employed a relatively coarse shell element mesh for describing the
pipeandnonlinearspringstosimulatethesoil.Asinalltheaforementionedstudies,
theiranalyticalmodeldoesnottakeintoaccountsecondordereffectssuchaslocal
bucklingandsectionaldeformationandappliesonlyundercertainstrainlimitsand
onlywhenthefaultresultsinelongationofthepipe,thus,whentensionandbending
aretheprevailingmodesofdeformation.

There is also a number of researchers that studied the problem using numerical
models. The majority of them used the Finite Element Method and studied mostly
thestrikeslipcase.Takadaetal.(2001)studiedelastoplasticshellmodebucklingof
a pipe subjected to normal and reverse fault movement using shell finite element

10

Introductionandstudymotivation

method. Based on analytical results and on Kennedys work they proposed a


simplified method for the maximum axial strain in steel pipes considering the
deformation of the pipe crosssection (Figure 1.4.6). O'Rourke and Liu (1997)
developed a numerical model to analyze pipe response to fault offsets and
comparedtheirnumericalresultswiththeresultsfromotheranalyticalapproaches.
They concluded that Kennedy's analytical solution is the best available for pipe
responsetostrikeslipfaultsprimarilyintension.Theseismicanalysisofburiedand
unburied pipelines, under both transient and permanent ground movements have
been examined by Arifin et al. (2010), using beam finite elements for the pipeline
and nonlinear springs to model the effects of the surrounding soil. Odina and Tan
(2009)investigatedburiedpipelineresponseunderseismicfaultdisplacement,using
abeamtypefiniteelementmodelwithelasticplasticspringsforthesoileffects.Ina
subsequentpublication,OdinaandConder(2010),extendedtheworkbyexamining
the effects of Lders plateau of the stressstrain material curve on the pipeline
responsecrossingactivefaults.

ItisworthmentioningtheworkofVazourasetal.(2010,2012)onwhichwebased
our approach. They simulated a buried pipeline crossing a strikeslip fault using
ABAQUS(Figures1.4.7,1.4.8).Theyconductedvariousparametricstudiesregarding
the type of soil, the fault angle and the diameter to thickness ratio of the pipe
(Figure 1.4.9). They also examined the effect of internal pipe pressure (Figure
1.4.10).Althoughtheirmodelislimitedtotheupper5mofthegroundanddoesnot
take into consideration the actual fault propagation, it has the advantage that the
soil is simulated by continuum brick elements and not by unconnected soil springs
which have been the most common option by the other researchers. That has
certainsignificantadvantagesasitwillbedescribedinthefollowingchapter.

Significant experimental research has been conducted, under the NEESR project
(Network for Earthquake Engineering Simulation) by Cornell University and
Rensselaer Polytechnic Institute supervised by Professor T.D.ORourke (2008). The
research,elaboratedattheCornell(CU)andRensselaer(RPI)NEESequipmentsite,

11

Introductionandstudymotivation

utilizestheequipmentforlargescalesoilstructureinteractionandcentrifugescale
split box testing. The equipment at Cornell consists of largedisplacement servo
hydraulicactuatorsandancillaryhydraulicsystems,soilstoragefacilitiesandframe
support system for largescale lifeline soilstructure interaction, a variety of
instrumentation,anddataacquisitionsystems(Figure1.4.11).TheRPIfacilitiesuse
advanced splitboxcentrifuge containers for simulation of lifeline systems. These
containers are used in the upgraded RPI 150 gton centrifuge (Figure 1.4.14). It is
worth noticing that the largescale ground rupture experiments involve the largest
laboratory tests ever performed on pipeline response to ground deformation. The
test not only demonstrates the ability of such pipelines to accommodate severe
movement, but provides important data about strain concentrations, changes in
shape, and soilstructure interaction Combined testing at CU and RPI provides
informationessentialfordesignandconstructioninresponsetoearthquakes,floods,
landslides, large deformation induced by tunneling and deep excavations, and
subsidencecausedbyseveredewateringorwithdrawalofmineralsandfluidsduring
miningandoilproduction.ThesamelargescaleequipmentatCornellhasbeenused
to test the behavior of a different construction method according to which the
pipeline is placed in a segmental, reinforced concrete vault with special joints that
canaccommodatelateraloffsetandcompressivedeformationduringfaultrupture,
therebyallowingforrotationandcompressionofthepipelineinsidethevaultatball
andslipjoints,respectively(Figure1.4.16).Finally,Simetal.(2011)(ImperialCollege
of London and University of Tokyo) performed a series of shaking table tests
modelingsmalldiameterpipescrossingaverticalfault(Figure1.4.18).Theirresults
indicated that the magnitude of bending moment is directly affected by the
magnitudeoffaultdisplacementirrespectiveofanyotherfactors.Inparticular,the
intensity of shaking does not play a significant role in the magnitude of bending
moment caused by fault action. In their experiments they also used tyre derived
aggregate (TDA) backfill and proved that it is capable of reducing the bending
moments incurred due to simultaneous faulting and shaking for maximum
accelerations less than or equal to 0.5 g and vertical fault displacements less than
thepipediameterbyupto74%.

12

Introductionandstudymotivation

1.5 Motivation of the Study

Energyisoneofthemostimportantissuesofmodernworld.Forthetimebeingand
probably for lots of future decades, energy production will be based primarily on
hydrocarbons. The transportation of energy is quite as vital as its very production
andtoday,themostcommonwayoftransportforoilornaturalgasarethepipeline
networks.Everyhourtheworldconsumesmillionsofcubicmetersofgasandoiland
almostallofitmovesthroughapipeline.

Past and Present

ThefirstrecordeduseofapipetotransportahydrocarbonwasinChinaabout2,500
years ago. The Chinese used bamboo pipe to transmit natural gas from shallow
wells:theyburneditunderpanstoboilseawatertoseparatethesalt,andmakethe
waterdrinkable.LaterrecordsindicatethattheChineseusedbamboopipe,wrapped
inwax,tolighttheircapital,Peking,asearlyas400BC(Britannica).Todaysoiland
gaspipelineindustryhasitsoriginsintheoilbusiness.OilhadbeendrilledinBaku,
Azerbaijan in 1848, and Poland in 1854, but the first major exploitation and
commercialization started in 1859 in the USA. In the early 1860s, the oil was
transported in wooden barrels on rivers by horsedrawn barges. Later, it was the
railwaythattooktheprimaryroleinoiltransportation.In1865a152mmdiameter
gravity oil line was built in Pennsylvania, USA, transporting 7000 barrels/day.
However,therealchangeinpipelineengineeringwasthebuildingoflongdistance,
largediameterpipelines:thesewerepioneeredintheUSAinthe1940sduetothe
energydemandsoftheSecondWorldWar.

13

Introductionandstudymotivation

Today,oilandgasprovidesmostoftheworldwithitsenergy.Thefuelsprovidingthe
worldwithitsprimaryenergyneedsare:
Oil=34%
Coal=24%
Gas=21%
Nuclear=7%
Hydro=2%
Other=12%
The oil and gas business is big, and it is going to become bigger. The US Energy
Information Administrations World Energy Outlook has predicted that fossil fuels
willremaintheprimarysourcesofenergy,meetingmorethan90%oftheincreasein
futureenergydemand.Thisexpanding,secureindustryisalsohighlyprofitable.Big
oil companies have announced annual profits even of $US36 billion, the largest
company profit ever accomplished. These profits are expected to continue in the
foreseeable future, as the price of fuels continues to rise. The same augmenting
trend is observed in pipeline construction market. The market size for oil and gas
pipelineconstructionexperiencedtremendousgrowthandgrewfrom$23billionsin
2006 to over $45billions today. To support the growth in energy demand, pipeline
infrastructure has grown by a factor of 100 in approximately 50 years. It has been
estimatedthatworldpipelineexpansioncouldbeupto7%peryearoverthenext15
years.Internationally,32,000kmofnewonshorepipelinesareconstructedeachyear
corresponding to a $US28billion business. The total length of high pressure
transmission pipelines around the world has been estimated at 3.500.000km
(Figures1.5.21.5.6).

14

Introductionandstudymotivation

Ofthose:
64%carrynaturalgas
19%carrypetroleumproducts
17%carrycrudeoil
Pipelines are a very safe form of transporting energy, both for human and the
environment.Relativereportsdemonstratedthatpipelinesare40timessaferthan
railtanks,and100timessaferthanroadtanksfortransportingenergy.Accordingto
theUSAAssociationofOilPipelines,oilpipelinespillsamounttoabout1gallonper
millionbarrelmiles.Thisislessthanoneteaspoonofoilspilledperthousandbarrel
miles.Pipelinesare,ingeneral,themosteconomicalwayforgasandoildistribution.
Itisalongterminvestmentandprobablythemostenergyefficientsolutionmeaning
that they consume less energy than the other options having at the same time
highercapacity.
Pipeline systems are critical transportation infrastructures in most nations and
essentialtobothstandardsoflivingandeconomies.Thefutureforpipelinesisboth
bright and challenging. They will continue to carry the bulk of our primary energy
sourcesandisuptoengineerstoensuretheyperformbothsafelyandefficiently.

Performance in past earthquakes



The vulnerability of buried pipelines to seismic hazards has been demonstrated by
theextensivedamageobservedduringpreviousearthquakes(Figures1.5.71.5.10).
Examplesofdocumentedpipelinedamage,regardlessitsuseandmaterial,include:
the 1905 San Francisco, 1933 Long Beach, 1952 Kern Country, 1964 Alaska,1964
Niigata,1971SanFernando,1979ImperialValley,1987Ecuador,1989 LomaPrieta,
1990ManjilEarthquake,1991CostaRica,1994Northridge,1995Kobe,1999ChiChi,
1999Kocaeli, 2010 Chile, 20102011 Christchurch and 2011 Japan. Three

15

Introductionandstudymotivation

characteristic cases are those of Alaska, San Fernando and Ecuador. In 1964, the
Anchorage, Alaska earthquake caused over 200 breaks in gas pipelines and 100
breaksinwaterdistributionpipelinesatAnchorage.Gaslineswithinfaultzoneswere
ruptured.Mostofthepipelinedamagewasduetolandslidesandgroundcracking.
The 1971 San Fernando earthquake resulted in 1,400 breaks in various piping
systems. The city of San Fernando temporally lost water, gas and sewage services.
Mostofthedamagewascausedduetoliquefactioninducedlateralspreading.The
1987 Ecuador earthquake destroyed the transEcuadorian pipeline, which
representedthelargestsinglepipelinelossinhistory.Itcostover$850millioninlost
salesandreconstruction.

The situation in Greece

Currently, in Greece, the major pipeline networks operating are the 300km
ThessalonikiAthensgaspipelineandtheITG(InterconnectorTurkeyGreece)(Figure
1.5.11). However, Greece is about to play a significant role for Europes energy
supply. There are 4 proposed plans for transporting gas from the Caspian and
Middle East areas to Europe: TAP (TransAdriatic Pipeline), ITGI(Interconnector
TurkeyGreeceItaly),NabuccoandSEEP(SouthEastEuropePipeline).Inthefirsttwo
ofthosescenariosGreecewillplaytheroleoftheenergyhubforEurope.TAPwould
run800kmfromKomotinithroughGreeceandAlbania,toendnearSanFoca,Italy
via a 110km offshore part. ITGI includes 600km onshore pipeline running all along
the northern part of Greece and 200km offshore pipeline across the Ionian Sea. It
also includes the connection of Bulgaria with Greece via the 170km
IBG(Interconnector GreeceBulgaria) pipeline (Figure 1.5.13). It is obvious how
important this is for Greece especially if we consider that, according to Energy
InformationAdministrationEIA,Officeofoilandgas2008 Report,thecostperKm
for the construction of a pipeline varies from 0.8 to over 3 million euros, with the
partofthemanagementandengineeringcorrespondingtothe2025%ofthetotal
cost.LookingatthetectonicmapofGreece(Figure1.5.12)itiseasytoobservethat

16

Introductionandstudymotivation

itisalmostinevitableforthepipelinenottocrossanyfaultzonealongitsway.Itis
quite interesting, as well, that the type of the great majority of faults of northern
Greeceisdipslip,somethingthatmakestheinvestigationofthisproblemevenmore
importantandontime.

Scope of the study

As already mentioned, little analytical work has been done for a pipe crossing a
normalorreversefault.Forapipesubjecttoanormalfault,thepipesoilsystemis
nolongersymmetric,andthetransverseinteractionforceatthepipesoilinterface
for downward movement of the pipeline is much larger than that for upward
movement.Similarly,forapipesubjecttoareversefault,itappearsthatnoclosed
formsimplifiedanalyticalapproachiscurrentlyavailable.Thebehaviorinsuchcases
is difficult to generalize, in part because there are two angles of intersection in
addition to the aforementioned asymmetric nature of the soil resistance in the
verticalplane.Theinternationalregulationsforthiskindofproblemarethuslimited.
Apartfromtherelativelyinadequateanalyticalresearchtherearealsofewnumerical
modelssimulatingdipslipfaultrupture.naddition,mostofthemsubstitutethesoil
with soilsprings attached to the pipe having as a result to neglect the interaction
betweenthesprings(asitwouldhappenintherealsoil)aswellastheoriginalfault
propagation. Finally, none of the previous researchers has given a documented
answertotheproblemofboundaryconditionsattheedgesofthepiperatherthan
assuming fixed conditions at the edges of the pipe. What is pioneering about our
work is the model we used to simulate the problem, which is probably the most
realistic model ever used for this kind of problems. We tried to overcome the
aforementioned problems using a hybridbeam model, in order to take into
considerationtheverylargedimensionsandaffectedareaoftherealproblem.We
also adopted and incorporated in the finiteelement code, an elastoplastic Mohr
coulomb constitutive model with isotropic strain softening that has been proved
capable of capturing the predominant mode of deformation caused by faults

17

Introductionandstudymotivation

comparedwithlaboratoryexperimentresults(Anastasopoulos2007).Apartfromthe
significance that pipeline networks are going to have for Greece and the various
problems of the previous models that we want to solve, it has been a great
motivationforourstudythefactthattheverysameproblemwillbeexaminedinour
laboratory. Using a splitbox device the behavior of pipes under dipslip faults is
goingtobeexamined,having,thus,theopportunitytoinvestigatecloselytheactual
mechanismsthatoccurandtocompareanalyticalwithexperimentalresults.
Tosumup,throughourresearchweaimto:
Proposeofapropermodelfortheproblem
Investigate the mechanical behavior of the pipe and the effect of various
parameters

18

19

20






Figures of Chapter 1

21

22

BasicTypesofFaults

Figure1.2.1.StrikeSlipFault.

Figure1.2.2.NormalFault.

Figure1.2.3.ReverseFault.

Figure1.2.4.ObliqueFault.

23


Figure1.2.5.TheSanAndreasFault,arightlateralstrikeslipfault,thatcausedthe
massive1906SanFranciscoearthquake(www.SanAndreasFault.org).

Figure1.2.6.TheDenaliFault,Alaska(www.arcticgas.gov).

24

Figure1.2.7NormalfaultatTuckiMountain,DeathValley,California(www.geotripper.org).

Figure1.2.8.ReversefaultatColorado(USGSgallery).

25

Table1.1.Gradesofsteelpipesandtheirminimumstrength.(AmericanPetroleumInstitute)

Figure1.3.1.ThetransAlaskaoilpipelinewhereitcrossestheDenaliFault.Thespecial
footingsallowthepipetomove6mlaterallyand1.5mvertically(AlaskaNaturalGas
TransportationProjects).

26


Figure1.3.2.Excavationofthetrench(www.constructionphotography.com).

Figure1.3.3.Excavatingpipetrenchusingtrenchingmachine(Lincolnshiregaspipeline
works,UnitedKingdom).

27


Figure1.3.4.LayingdownthegaspipelinethatrunsfromEdmontontoVancouver,Canada
(CanadianEnergypipelineAssociation).

Figure1.3.5.Pipelayingprocessofthe370kmFujairahoilpipelineintheUnitedArab
Emirates(www.pipelinesinternational.com).

28


Figure1.3.6.Weldingprocess.SouthWalesgaspipeline. UKs largest high-pressure gas
pipeline (317 km) (www.nationalgrid.com).

29

LiteratureReview

Figure1.4.1.NewmarkandHallmodel(1975).

30

Pipedeformationnearthefault,afterKennedyetal.
Figure1.4.2.Kennedysmodel(1977).

31

Modelofpipelinesubjectedtoverticalfaultmovement(sideview).

Detailoflargedeformationareaofthemodel.

Figure1.4.3.VougioukasandCarydismodel(1979).

32


Figure1.4.4.PipelineanalysismodelproposedbyWangandYeh(1985).

Figure1.4.5.Partitioningofthepipelineintofoursegments.WangandYehmethodon
whichKaramitrosapproachwasbased.

33

Figure1.4.6.FlowofthesimplifiedmethodproposedbyTakadaetal(2001).

34

Figure1.4.7.TheFEmodelusedbyVazourasetal.(2011):(a)soilformationwithhorizontaltectonicfault,(b)
crosssectionand(c)steelpipeline.

35

Figure1.4.8.ResultsfromVazourasetal.model(2011).(a)Planviewofdeformedshapeofapipelineforfault
offsetd=14mand(b)distributionoflongitudinalnormalstrainforseismicfaultdisplacementequalto1and
4m(X65pipe,D/t=72).

36


Figure1.4.9.Normalizedultimatefaultdisplacementforvariousperformancelimitsat
differentanglesof(65pipe,Clay,D/t=72,zeropressure),Vazourasetal.

Figure1.4.10.Effectofinternalpressureontheevolutionofcrosssectionaldeformationin
pressurizedpipelinesforthesamefaultdisplacementd=1m(X65,Clay,D/t=96,=35),
Vazourasetal(2011).

37

Table1.2.TheCornellandRPItestparameters.

Table1.3.Thewholeofthefullscale,centrifugeandnumericalNEEStests.

38

Figure1.4.11.Cornellsplitboxtestbasin.

Figure1.4.12.PlanViewofFullScalePipelineTest.

39

Figure1.4.13.MeasuredAxialandBendingStrainforallfullscaletests(NEESRFinalReport
2008).

Figure1.4.14.RPI150gTonCentrifuge.

40


Figure1.4.15.DistributionofnormalPressureMeasuredbyTactileForceSensors(NEESR
FinalReport2008).

41

Figure1.4.16.TestBasinatCornellforthepipelineembeddedinconcreteblockstest
(NEESRFinalReport2008).

42

Figure1.4.17.Photoanddrawingofthebasinat2moffaultdisplacement
(NEESRFinalReport2008).

43

Figure 1.4.18. The shaking table of the University of Tokyo, capable of simulating at the
sameexperimenttheseismicvibrationaswellasthefaultrupture(Simetal.2012).

44

Figure1.4.19.Bendingmomentdistributionforapipecrossingthefaultat90subjectedto
noshakingandAmax=0.2g,10Hzatdifferentfaultdisplacements(Simetal.2012).

Figure 1.4.20. Comparison of pure sand and TDA trench backfills through normalized
bendingmomentdistributionwithdistancealongthepipeaxisfromtheversesubjectedto
Amax=0.5g,10Hzat4differentfaultdisplacements(Simetal.2012).

45


Figure1.5.1.Pennsylvania,USA,1872.Oneofthefirstgaspipelinesrunning5kilometers
fromNewtonWellstoTitusville.(AmericanOilandGasHistoricalSociety,aoghs.org)

Figure1.5.2.ThegaspipelinenetworkattheUSA(API).

46


Figure1.5.3.TheoilpipelinenetworkofUSA(API).

Figure1.5.4.MajorlifelinesatCentralAsia(USCC.gov).

47


Figure1.5.5.PipelinenetworkofCentralandWesternEurope(InternationalGasUnion).

Figure1.5.6.MajorgaspipelinesofRussiaandEasternEurope(InternationalGasUnion).

48

Figure1.5.7.FaultinduceddamageofembeddedpipelineduringtheManjilearthquake
1990,Iran(Simetal.2012).

Figure1.5.8.Largedeformationofsteelpipesectioncrossinganactivefault
(1999ChiChiearthquake, Takadaetal.2001).

49

Figure1.5.9.Failureofsteelpipescrossingfaults.(a)1999Kocaeliearthquake,Turkeu,
(b)1999ChiChiearthquake,Taiwan.(Takadaetal.2001)

Figure1.5.10.Wrinkledpipelinefailedincompression(Takadaetal.2001).

50


Figure1.5.11.ThemaingaspipelinesinGreece
(..).

Figure1.5.12.MajortectonicfaultsandlocationmapofGreece(Papazachos,1997).

51

Figure1.5.13.TheSoutherngascorridor.Thedifferentproposedplansandtheirroutes
(EuropeanCommissionProject,2009).

52

53

54





Chapter 2
Problem and model definition

55

56

Problemandmodeldefinition

2.1 Problem Definition

As already mentioned, relatively limited amount of research has been done


regarding the behavior of a pipeline crossing a dipslip type of fault. The analytical
approachesextendtoacertainlimit,havingtroubletodescribetheproblemusing
formulas because of its complexity. In the strikeslip rupture the soils passive
resistance can be assumed the same along the pipeline. However, in dipslip cases
thisisnotrealisticfortheupwardmovementofthepipefoundslessresistancethan
the downward movement since the soil above the pipeline is of significantly less
height (thus less weight, easier to get lifted up by the pipe) and less strength
comparedtotheinfinitesoillayerunderthepipe.Furthermore,anotherdifficulty
that occurs when we examine a pipeline subjected to dipslip fault rupture is an
extra added parameter. Unlike, strikeslip fault where there is only one angle of
intersection to be examined in dipslip faults there are two: the angle in plan
betweenthefaultandthepipeline,aswellasthedipangleofthefault.Inaddition,
analytical studies have difficulties in including the nonlinearity aspects of the
problem(eithergeometricalormaterialnonlinearity)aswellassecondordereffects
such as local buckling and ovalization of the pipe that are crucial modes of failure
andcannotbeneglected.Thus,arigoroussolutionoftheproblemshouldinvolvean
advancednumericalanalysiswhichcanaccountconsistentlyforthenonlinearstress
strain response of the pipeline steel, the possible forming of gap between the
pipelineandthesoil,thelongitudinalandtransversesoilresistanceaswellassecond
order effects induced by relatively large displacements. Since such analyses are
definitely possible with the available commercial computer codes, many of the
researchers approached the problem using numerical methods. The great majority
ofthemusedthefiniteelementmethodsimulatingthepipewithshellelementsand
thesoilwithsoilsprings.Thesubstitutionofsoilwithsoilspringshastheadvantage
of limiting the size of the model and as a result reducing significantly the
computationaleffortandtimeneededfortheanalysis.However,soilspringsmodel,
in general, does not take into account either the interaction between the soil

57

Problemandmodeldefinition

elements or the real propagation of the fault. This is probably the reason why
severalresearcherspreferredtouseintheirmodelscontinuumbrickelementsfor
thesoil.Similarly,themodelthatweusedinthisstudyisafullmodelthatconsistsof
the pipe and the soil represented by shell and continuum brick elements
respectively.
Another significant drawback of lots of studies is the boundary conditions of the
modelandspecificallythesupportsofthepipeatthe2edgesofthemodel.Itisa
fact that the deformation caused by the fault rupture to the pipe, as well as the
bending stress and strain that occur to the pipe, are concentrated to a relative
narrowarea(dependingonthemagnitudeofthefaultsoffsetandthepipeandsoil
properties).Howevertheaxialstressesthatdevelopaffectasubstantiallybiggerpart
of the pipe that cannot be fully modeled for computational reasons. Thus, the
problemoftheboundaryconditionsarises,whichisthesubjectofChapter3.
Investigating the problem of pipe boundaries we try different models and
approaches. However, all of those models simulated a normal fault rupture. In
Chapter 4 we design a reverse fault model to examine the behavior of the pipe
comparedtothenormalfaultcase.
In Chapter 5 we focus on the aforementioned problem of faults propagation. In
most of the relative studies where a full model is used, despite the fact that
continuum soil elements can depicture the faults propagation, researchers could
designandusealimitedmodelthatconsistedofthepipeandonlyasmalllayerof
soil.Thelimitedcomputationalpowerandthetimelimitsforthecompletionofthe
study, commonly, do not allow the use of a bigger and more realistic model that
would include a soil layer of 20, 40 meters or more. To overpass this problem we
extractresultsfroma2Dfreefieldmodel,wherefaultspropagatefromthebedrock
tothesurfacethroughasoillayerofrealisticwidth,andthenweintegratetheminto
thehybridfull3Dmodel.Thismethodcombinedthe3Daccuracywiththerealistic
fault propagation of the fault without the highcomputational force and time a big
3Dmodelwouldneed.

58

Problemandmodeldefinition

InChapter6,weexaminetheeffectofanelasticplasticwithhardeningstressstrain
relation for the pipe steel, whereas in Chapter 7, we investigate the effect of the
internal pipe pressure both for elasticperfectly plastic and for hardening steel
behavior.
Finally, in Chapter 8, we gather the most significant conclusions of our study and
propose additional factors and parameters that could be examined in a future
research.

2.2 Model and Method of Analysis

Asalreadystated,inordertoconcludetoafinalmodelwetestedseveraldifferent
approaches.Thefollowingparagraphsincludethegeneralpropertiesandconditions
thatarecommonforallthemodelsweused(unlessstatedotherwise).

2.2.1 Finite Element Model

Geometry of the model

The analyses for the investigation of the problem were conducted in three
dimensionalspaceusingthefiniteelementcodeABAQUSv.6.11.
Thedimensionsofourbasicmodelare60x10x5m.Thepipelineisplacedatthe
center of our model and its diameter is 0.9144m (3feet). In all of our tests, we
maintained the same angle of intersection between the fault and the pipeline
investigating only the effect of the dip angle of the fault. For the model definition
proceduretheangleofthefaultwasselectedequalto60.Thefaultmovementwas
integrated in the model by moving a part of the base and the corresponding side
(Figure 2.2.1), according to the wanted fault angle, while the rest of the base

59

Problemandmodeldefinition

remained immovable. Specifically, the boundary conditions of our model are


depicted in Figure 2.2.2. Concerning the conditions at the pipe edges we tried
variousapproacheswhichwillbediscussedinthenextchapter.

Type of elements and material properties


SOIL

Itisnecessarytomakeclearthatthevaluesmentionedinthisparagraphwereused
in order to find a proper model for the problem. More realistic values, as well as
parametricanalysesareissuesthatweexamineinthenextchapters.
Thesoilbodywasmodeledusing8nodehexahedralcontinuumelementsC3D8,with
acorrespondingconstitutivemodel,asdescribedinparagraph2.2.3.Themeshingof
thesoilbodyisdividedinthreeareas:a20mareaofrelativelydensediscretization,
inthemiddleofthemodelwhereweexpectthefaulttocrossthepipeline,thusthe
main soil (and pipe) deformations to occur and two other areas of 20m of coarser
meshing(Figure2.2.2).
Thesoilweusedwasdensedrysandwiththefollowingproperties.
ElasticModulus
Poisson'sRatio

E
v

25MPa
0.3

AngleofInternalFriction

40o

DilationAngle

10o

Density

2.1kg/m3

SpecificWeight

20.6kN/m3

Ko

0.29

NeutralEarthPressureCoefficient

For computational reasons we also implemented to our sand soil model a small
cohesionparameterc=2Kpa.

60

Problemandmodeldefinition

2.2.2 Soil Constitutive Model

Severalexperimentalandnumericalstudieshaveshownthatpostpeaksoilbehavior
is a decisive factor in fault rupture propagation and its possible emergence on the
groundsurface.ScottandSchoustra(1974)utilizingtheFEmethodandanelastic
perfectly plastic constitutive soil model with MohrCoulomb failure criterion,
produced results contradicting both reality and experiments. Walters and Thomas
(1982) employed a more elaborate nonlinear incremental constitutive model with
DruckerPrager failure criterion, nonassociated flow rule, and strain softening to
simulate reverse fault rupture propagation through cohesionless soil. Comparing
their analysis results with benchmark smallscale tests, they proved that laborator
reality could only be reproduced with a nonassociated flow rule and strain
softening. Bray (1990) and Bray et al. (1994), utilizing a FE code with a hyperbolic
nonlinear elastic constitutive law, also achieved satisfactory agreement with
experiments. Also successful were the analyses performed by Roth et al. (1982),
Whiteetal.(1994),Nakaietal.(1995),Loukidis(1999),andEricksonetal.(2001),all
of which made use of the finitedifference (FD) method with an elastoplastic
constitutive model, MohrCoulomb failure criterion, and strain softening. Similar
constitutive models have been employed successfully to model the failure of
embankment dams and the delayed collapse of cut slopes in stiff clay (Potts et al.
1997). Following a thorough review of the literature (Anastasopoulos 2005,
Anastasopoulos et al. 2007) we adopted a similar elastoplastic MohrCoulomb
constitutivemodelwithisotropicstrainsoftening.

61

Problemandmodeldefinition

Elastic behavior
The elasticity form that we implemented in our model is linear isotropic elasticity
wherethestressstrainrelationshipisgivenby:

TheelasticpropertiesarecompletelydefinedbygivingtheYoung'smodulus,E,and
thePoisson'sratio, .(Theshearmodulus,G,canbeexpressedintermsofEand as

Plastic behavior: yield criteria


MohrCoulomb
TheMohrCoulombcriterionassumesthatyieldoccurswhentheshearstressonany
pointinamaterialreachesavaluethatdependslinearlyonthenormalstressinthe
sameplane.TheMohrCoulombmodelisbasedonplottingMohr'scircleforstates
ofstressatyieldintheplaneofthemaximumandminimumprincipalstresses.The
yieldlineisthebeststraightlinethattouchestheseMohr'scircles.

62

Problemandmodeldefinition

Therefore,theMohrCoulombmodelisdefinedby

tan c
whereisnegativeincompression.FromMohr'scircle,

s cos
s sin
m

Substituting forand, multiplying both sides bycos, and reducing, the Mohr
Coulombmodelcanbewrittenas

s m sin c cos 0
where

1
s ( 1 3)
2
ishalfofthedifferencebetweenthemaximumprincipalstress,1,andtheminimum
principalstress,3(andis,therefore,themaximumshearstress),

1
2

m ( 1 3)
istheaverageofthemaximumandminimumprincipalstresses,andisthefriction
angle.

63

Problemandmodeldefinition

The MohrCoulomb criterion in three dimensions is often expressed as

Forgeneralstatesofstressthemodelismoreconvenientlywrittenintermsofthree
stressinvariantsas

F Rmcq p tan c 0
where

Rmc (, )

1
1

sin cos tan


3 3
3
cos

istheslopeoftheMohrCoulombyieldsurfaceinthep
frictionangleofthematerial
Cisthecohesionofthematerial;and
isthedeviatoricpolarangledefinedas

r
cos 3
q

1
p trace istheequivalentpressurestress,
3

64

stressplane,the

Problemandmodeldefinition

3
S : S istheMisesequivalentstress,
2

S
:
S

S pI

1
3

isthethirdinvariantofdeviatoricstress,and

isthedeviatoricstress.
Thefrictionangle,,controlstheshapeoftheyieldsurfaceinthedeviatoricplane
asshowninFigure2.2.4.Thetensioncutoffsurfaceisshownforameridionalangle
of=0.Thefrictionanglerangeis090.Inthecaseof=0theMohrCoulomb
modelreducestothepressureindependentTrescamodelwithaperfectlyhexagonal
deviatoric section. In the case of=90 the MohrCoulomb model reduces to the
tensioncutoffRankinemodelwithatriangulardeviatoricsectionand
R mc

(this limiting case is not permitted within the MohrCoulomb model

describedhere).
Isotropic cohesion hardening is assumed for the hardening behavior of the Mohr
Coulombyieldsurface.Thehardeningcurvedescribesthecohesionyieldstressasa
functionofplasticstrain.

Plastic flow on the MohrCoulomb yield surface


Theflowpotential,G,fortheMohrCoulombyieldsurfaceischosenasahyperbolic
functioninthemeridionalstressplaneandthesmoothellipticfunctionproposedby
MentreyandWillam(1995)inthedeviatoricstressplane:

c |

tan Rmwq p tan


2

65

Problemandmodeldefinition

where

Rmw(, e)

4 1 e 2 cos 2 2e 1

2 1 e 2 cos 2e 1


Rmc ,
3
4 1 e 2 cos 2 5e 2 4e

and

3 sin
Rmc ,
3 6 cos

isthedilationanglemeasuredinthepRmwqplaneathighconfiningpressureand
candependontemperatureandpredefinedfieldvariables.
c|0istheinitialcohesionyieldstress.
isthedeviatoricpolarangledefinedpreviously.
isaparameter,referredtoasthemeridionaleccentricity,thatdefinestherateat
which the hyperbolic function approaches the asymptote (the flow potential tends
toastraightlineinthemeridionalstressplaneasthemeridionaleccentricitytends
tozero);and
eisaparameter,referredtoasthedeviatoriceccentricity,thatdescribestheout
ofroundedness of the deviatoric section in terms of the ratio between the shear
stress along the extension meridian ( 0 ) and the shear stress along the
compressionmeridian(

Adefaultvalueof

).

isprovidedforthemeridionaleccentricity,.

Bydefault,thedeviatoriceccentricity,e,iscalculatedas

66

Problemandmodeldefinition

3 sin
3 sin

whereis the MohrCoulomb friction angle; this calculation corresponds to


matching the flow potential to the yield surface in both triaxial tension and
compression in the deviatoric plane. This flow potential, which is continuous and
smooth,ensuresthattheflowdirectionisalwaysuniquelydefined.Inourmodelswe
used nonassociated flow rule. A family of hyperbolic potentials in the meridional
stressplaneandtheflowpotentialinthedeviatoricstressplaneareshowninFigure
2.2.5.

Strain Softening

As depicted by the following equations, strain softening is introduced by reducing


the mobilized friction angle mob and the mobilized dilation angle mob with the
increase of octahedral plastic shear strain where p and res=ultimate mobilized
frictionangleanditsresidual(orcriticalstate)value;p=ultimatedilationangle;and

fp =plastic octahedral shear strain at which softening has been completed. The
model is incorporated in the finiteelement code ABAQUS 2011 through a user
subroutine.

p res p

oct
p
p

f
mob

res

p
fp
for 0 oct

p
fp
for oct

67

Problemandmodeldefinition

p
p 1 octp
mob f

res

p
p
for 0 oct f

p
p
for oct f

Figure 2.2.6 shows a typical variation of the stress ratio, /v, and volume change
(expressed through vertical displacement y of the top platen) with respect to the
imposed horizontal displacement x in a direct shear test of dense Toyoura sand
[based on data of Shibuya et al. (1997)]. Soil response can be divided in four
characteristicphases.

1.Quasielasticbehavior(OA):UptoPointAthesoildeformsquasielastically(Jewell
andRoth1987).Somenonlinearityisclearlyobserved,butwithoutanydilation.xy
isdefinedasthehorizontaldisplacementforwhichy/x>0.

2.Plasticbehavior(AB):FromPointAtoBthesoilyields,enterstheplasticregion,
anddilates.PeakconditionsarereachedatPointB.xp=horizontaldisplacementfor
which /v=max.

3. Softening behavior (BC): From Point B to C the soil experiences softening. Right
after the peak, a single horizontal shear band develops at mid height of the
specimen(JewellandRoth1987;Gerolymosetal.2007).Softeningiscompletedat
PointC,andxf=horizontaldisplacementforwhichy/x0.

4.Residualbehavior(CD):Shearingisaccumulatedalongthedevelopedshearband.

68

Problemandmodeldefinition

2.2.3 Pipeline properties

The pipeline was modeled using shell elements S4R, a robust, generalpurpose
quadrilateral element that is suitable for a wide range of applications (Figure
2.2.7). Theseelementsprovideaccuratesolutionsinallloadingconditionsforthin
andthickshellproblems.Thicknesschangeasafunctionofinplanedeformationis
allowedintheirformulation.Theydonotsufferfromtransverseshearlocking,nor
do they have any unconstrained hourglass modes. These elements consider finite
membrane strains. In geometrically nonlinear analyses in Abaqus/Standard the
crosssection thickness of finitestrain shell elements changes as a function of the
membranestrainbasedonauserdefinedeffectivesectionPoisson'sratio, .The
thickness change based on the effective section Poisson's ratio is calculated as
follows.
Inplanestress33=0;linearelasticitygives

33

( 11 22)

Treatingtheseaslogarithmicstrains,

t
ln
1
t0

l1
l2

A
ln
ln 0 ln 0
1 A0
l2
l1

whereAistheareaontheshell'sreferencesurface.Thisnonlinearanalogywith
linearelasticityleadstothethicknesschangerelationship:

t A

t 0 A0

For = 0.5the material is incompressible; for = 0the section thickness does not
change.
Likebefore,themiddlepartofthepipelinehasamorerefinedmeshwithasizeof
shellelements3.2cmx9cminsteadof20cmx9cmwhichistheelementsizeforthe
restofthepipe(Figure2.2.7).

69

Problemandmodeldefinition

Although the diameter of the pipe D remained the same in all of our tests, the
thicknessofthepipetrangesbetween0.00635mand0.0127mgiving4differentD/t
ratios.
Diameter

0.9144m/36''

Thikness

0.00635m/0.25''

0.00762m/0.3''

0.009525m/0.375''

0.0127m/0.5''

D/t

144

120

96

72

ThepipeweusedisanAPI5LX65steelpipe,whichiswidelyusedintheoilandgas
industry.
Yield Strength

Tensile Strength

API 5L
Grade

min.

min.

Yield to Tensile
Ratio
max.

ksi / Mpa

X65

ksi / Mpa
65 / 450

77 / 530

0.93

18

Elongation
min.

Elasticmodulus
Poisson'sRatio

Es
v

210Gpa
0.3

Density

7.8kg/m3

Asafirstapproachweassumedelasticperfectlyplasticbaviourofthesteelwhereas
in a later chapter we integrated the original hardening behavior of the material as
givenbytheAmericanPetroleumInstitute.
Thefrictioncoefficientofthepipesoilinterfacewaschosenequalto0.42.

70

Problemandmodeldefinition

2.2.4 Modes of pipeline failure

Themainconcernofthepipelinedesignistoensurethattherewouldbenolossof
containment.Itisalsocrucialtomaintainthepipelinefunctional.Inordertoquantify
the amount of damage a buried pipeline can endure it is necessary to adopt
appropriate criteria which will pose certain limits and make clear whether the
pipeline is safe and operational or not. Traditionally, the pipeline design has been
basedonthecriteriaoftheallowablestress.However,steelisaductilematerial
capable of sustaining significant amount of inelastic deformation. Thus, under the
severe deformations caused by the fault rupture the pipeline exhibits important
inelasticdeformation.Consideringthat,itisreasonabletodeterminethelimitstates
accordingtothestrainanddeformationlevelsratherthantostressvalues.Basedon
the paper of Vazouras et.al(2012) we categorize the performance limit states as
follows(Figure2.2.9):
a) tensile strain in the longitudinal direction of the pipeline that causes pipe wall
rupture
b) localbuckling
c) excessivedistortionofthepipelinecrosssection(ovalization)

A) Maximum tensile strain



Determination of tensile strain capacity of a pipeline is of primary importance
for establishing an efficient strainbased pipeline design procedure. In the
absence of serious defects and damage of the pipeline, the tensile capacity is
controlled mainly by the strength of the girth welds, which are usually the
weakest locations due to weld defects and stress/strain concentrations. Tensile
strain limits are experimentally determined through appropriate tension tests on
strip specimens and in wide plates. For design purposes, a simple and
straightforward approach for determining tensile strain limit of girth welds in

71

Problemandmodeldefinition

pipeline is provided by the Canadian CSA Z662 pipeline design standard, through
the following equation that considers surfacebreaking defects:

u= (2.36-1.583-0.101) (1+16.1-4.45) ( -0.157 + 0.239-0.241 -0.315 )


where:
u:theultimatetensilestraincapacityin%
: the CrackTip Opening Displacement. CTOD is a measurement of the stretching
across the crack tip just before rapid growth starts at the crack and is used as a
standardmeasurementoftoughnessoftheweld.(0.11)
:theyieldtotensilestrenghtratio.(0.70.95)
:theratioofdefectlengthoverthepipewallthicknessequalto(2c/t).(110)
:theratioofdefectheightoverthepipewallthickness(/t).(0.5)
Considering a slightly defected pipe with =0.3, =0.85, =1 and =0.1 the critical
tensilestrainisequaltou=3.5%.
ThisisveryclosetothevalueproposedbyEurocode8,EN19984andbyASCEMOP
119wheretheproposedtensilestrainlimitis3%forsteelweldedpipes.

Aplanarsurfacebreakingdefectinthepipewall.

72

Problemandmodeldefinition

CSAZ662pipelinedesignstandardalsoprovidesanequationforburieddefects:

Aplanarburieddefectinthepipewall.
where:
,,,:aspreviouslydefined
:ratioofdefectdepthtopipewallthickness(d/t).
Using the previous values and assuming =0.1 for a 12.7mm thickness pipe one
obtains tcrit=4.5%. Although we do not take into consideration effects of
temperature, strain rate, prior strain history and strain aging these equations give
quite conservative values as stated in the CSA regulations. Based on the above
equationsandontheproposedvaluesoftheinternationalregulationsweinvestigate
theproblemusingtwovaluesofcriticaltensilestrain3%and5%.

73

Problemandmodeldefinition

B) Local Buckling

Undergroundinducedactions,togetherwiththedevelopmentoftensilestrainson
the pipeline wall, compressive strains due to pipe bending deformation may also
occur. When compressive strains exceed a certain limit, pipeline wall exhibits
structural instability in the form of local buckling. In the presence of buckling the
pipelinemaystillfulfillitstransportationfunction,providedthatthesteelmaterialis
adequately ductile. However, the buckled area is associated with significant strain
concentrationsand,inthecaseofrepeatedloading(e.g.duetovariationsofinternal
pressureortemperature),fatiguecracksmaydevelop,imposingaseriousthreatfor
thestructuralintegrityofthepipeline.Therefore,theformationofalocalbuckleor
wrinklecanbeconsideredasalimitstate.Extensiveresearchonaxialcompression
andbendingofsteelpipesegmentshasdemonstratedthatcompressivestrainlimits
for steel pipes depend on the diametertothickness ratio (D/t), the yield stress of
steelmaterialy,thepresenceofinternalorexternalpressure,aswellastheinitial
imperfections and the residual stresses.(karamanos et al). The local buckling strain
Cucanbeestimatedusingthefollowingdesignequation,adoptedbytheCSAZ662
specification:

with

Wherehisthehoopstressandpistheinternalpipepressure.

74

Problemandmodeldefinition

C) Distortion of pipeline crosssection

Significant distortions of the pipeline crosssection even though may not harm the
integrityofthepipeshouldbeexaminedasamodeoffailuresinceaboveacertain
limit the pipeline is no longer functional. A simple and efficient measure of cross
sectiondistortionisovalizationwhichwequantifieusingtheflatteningparameterf
andconsiderasalimitvaluef=0.15asalsoadoptedbytheDutchspecificationNEN
3650.

Do D '
f=
Do

Do:theinitialdiameterofthepipe
D:theminimumdiameterofthefinaldistortedcrosssection

75

76







Figures of Chapter 2

77

78

(a)

(b)

60

60

(c)

z
x
y
Figure2.2.1.(a)Thebasicmodelofourstudy.(b)Thefaultwasinducedbymovingapartof
thebaseandthe correspondingsideattheleftedgeofthemodel.(c)Thedistortedshape
ofthepipe.

79

5m

0.5m

2m

0.625m
5m

60m

10m

y
x

0.625m

25Mpa

5m

Figure2.2.2.Thesoilpartofthemodel,itsvariouscrosssectionswiththeirproperties,
dimensionsandboundaryconditions.

80

(a)

(b)

Figure2.2.3.(a)ViewofMohrCoulombfailuresurfacein3Dspaceofprincipalstresses.(b)
TraceoftheMohrCoulombyieldsurfaceinthe 12 plane.

81

(a)

(b)

Figure2.2.4. MohrCoulombandtensioncutoffsurfacesin (a)meridional and (b)deviatoric


planes..

82

(a)

(b)

Figure2.2.5.(a)Hyperbolicpotentialsinthemeridional stressplaneand (b)theflowpotential


inthedeviatoric stressplane.

83

Figure2.2.6.Atypicalvariationofthestressratio,/v,andvolumechange(expressed
throughverticaldisplacementyofthetopplaten)withrespecttotheimposedhorizontal
displacementxinadirectsheartestofdenseToyoura sand[basedondataofShibuyaetal.
(1997)].

84

(a)

0.02m

0.032m

0.07m

(b)

0.9144m

500
450
400
350
300
Stress(Mpa) 250
200
150
100
50
0
0%

5%

10%

Strain(%)

Figure2.2.7.(a)Thepipeasdesignedinourmodel.dividedinthreeparts,accordingtothe
neededdensity ofmeshing,(b)Thepipecrosssectionandthediagramthatdescribedthe
stressstrainrelationofthepipesteel.

85

(a)

(b)

Figure2.2.8.(a)Theelementsusedinourbasicmodel.Shellelementsforthepipeand
continuumbrickelementsforthesoil.(b)Comparisonofshelltocontinuumelements
(Abaqus Documentation).

86

Modesofpipelinefailure
(a)

(b)

(c)

h=0

h=0,2m

h=0,5m

h=1m

h=1,45m

Figure2.2.9.(a)Excessivetensilestrain,(b)Localbuckling,(c)Thedevelopmentofthe
ovalization ofthepipecrosssectionwithrespecttotheverticalfaultdisplacementh.The
criticalstatebeyondwhichthepipeisconsideredasnonoperationalcorresponds,forthis
exampletoh=1m,[D/t=144,steelwithhardeningbehavior,HybridBeamModel(Chapter6)].

87

88





Chapter 3
Pipe boundaries

89

90

Pipeboundaries

Introduction

Asalreadymentioned,theboundaryconditionsofthiskindofproblemsisyettobe
answered. A very large model would be needed if we wanted to capture the real
pipesoil interaction and pipe behavior. Such a big model would be, at least, time
consuming, hence, all of the researchers had made an assumption about the
boundaryconditionsthattheyshouldapplytothemodel.Usually,thepipeedgesare
consideredasfixedtothemarginalsidesofthemodel.Nonetheless,wedecidednot
totakethatassumptionasgrantedandtoexaminedifferentboundaryconditionsfor
the pipe edges. Apart from the free and fixed conditions that we examined, we
designedahybridmodelthataddressestotheboundaryproblembycombiningthe
previousfullmodelwithbeamelementsthatrepresentthecontinuationofthepipe
and extend for a kilometer, starting from the edges of the pipe, and soil springs
attachedtothosebeamsasthesurroundingsoil.Inthefollowingparagraphsofthe
chapter,wepresentallthestepsandallthedifferentmodelswetried,beforeending
upwiththemostrealisticandreliablesimulation.
We examine four different cases, testing a normal fault and a pipeline of variable
thickness.

91

Pipeboundaries

3.1 Free Pipe Ends Model

Intheseanalyses,wereleaseallthedegreesoffreedomofthepipeedges.Initially,
duringthefirstfewcentimetersoffaultmovement,thepipeisundertensilestress
thatiscausedprimarilybythefrictionbetweenthepipeandthesoilbecauseoftheir
trend for relative movement. Very soon, the bending mechanism starts to prevail.
Thedeformedshapeofthepipeindicatestwoinflectionpoints(Figure3.1.1)thatis
inabsoluteaccordancewiththeanalysisstressandstrainresults(Figure3.1.3).In
those two bending areas, tensile and mostly compressive strains start to
concentrate, until the point where the compression causes local buckling to occur,
firstlyinthefootwallandlaterinthehangingwall(Figure3.1.2).
At this point, it is important to describe the way we define and observe the
commencement of the local buckling phenomenon which is our failure criterion
apartfromthevisualobservationofconcentratedwrinklingofthepipe.Wecanspot
the start of buckling both at the strainposition diagram, which demonstrates the
strain distribution along the pipe for various fault offsets (Figure 3.1.4) and at the
strainfaultoffsetdiagramatthepointofthepipewherethebucklingoccurs(Figure
3.1.5). Concerning the first diagram, as the fault offset augments the compressive
axial strains increase as well, having a substantially bigger ratio of increase at the
area where buckling will occur, forming a bell shaped strain distribution. When
buckling begins, additional fault movement causes not only a general increase of
compressivestrainsbutalsoadecreaseofaxialstrainsatcertainpointsofthatarea
which in later steps can even turn to tensile strains. The start of buckling can be
definedevenmoresimply,bythelaterdiagramwementioned,justbypointingout
themoment(point)whentheincreaseratioofcompressivestrainsaugmentsgreatly
andstrainsbegintoincreasealmostverticallycomparedtotheirpreviousmildtrend
ofincrease.
TheresultsobtainedarepresentedinFigures3.1.43.1.8.

92

Pipeboundaries

As we can observe in Figure 3.1.6, smaller D/t ratio pipes (thicker pipes since we
maintainthesamediameter)canendurebiggerfaultoffsetsandbiggercompressive
strainsbeforetheappearanceofbuckling.Itisalsodemonstratedthat,eventhough
bucklingdidnothavetheexactsameforminallD/tcases,itoccursinsidearelatively
short area of 1m very close to the pipe fault intersection point. This indicates that
hypothetically if we knew the faults plane we could presume the buckling area
despite D/t ratio (about 12 diameters length from the faultpipe intersection).
However,bedrockat5mdepthisnotveryrealistic,sotheaboveconclusioncouldbe
notofgreatinterestandwillbereviewedinChapter5.
Even though, our approach, as defined in Chapter 2, concentrates on strain based
criteriaweexamined,aswell,theaxialforce(N)andmoment(My)distributionalong
the pipe. Since the pipeline is simulated using shell elements, different values of
forcesandmomentscanbeexpected,notonlyalongthepipeslengthbutalongthe
crosssections of the pipe as well. For this reason, we examined and edited the
output data for different points of the pipes crosssection. The axial forces and
moments along the top and bottom pipe fibers are demonstrated in Figures 3.1.7
and3.1.8respectively.
Concerning the axial forces diagram, the distribution of forces is in absolute
agreementwiththepreviousdiagramsofaxialstrains,makingthetwobendingareas
easy to observe. Apart from that, the axial force at the pipe edges (and at the
adjacent area) is practically zero, something we clearly expected since we had
designedthepipeedgestobefree.Anotherinterestingobservationwecanmakeby
lookingthesegraphsisthatasthefaultdisplacementincreasessodothevaluesof
the axial force. At some characteristic areas, after a certain value the axial forces
remainstableandtheseareasstarttoexpand.Theseareascorrespondtothepipe
parts where sliding between the pipe and the soil has occurred and the
aforementioned maximum value that the axial forces can obtain is the marginal
friction force. The exact value of this maximum force is not easy to be calculated,
since,itdependsnotonlyonthefrictioncoefficientofthepipesoilinterfaceandthe
vertical(tothepipe)soilpressurebutontheexactsurface(thustheexactlength)of

93

Pipeboundaries

thepipethatisactivatedandtendstomoverelativelytothesurroundingsoil.Aswe
alsoexpected,themaximumfrictionvalueforthelowerpartofthepipeisslightly
biggerbecauseofthebiggerdepthandtheaddedweightofthepipe.Finally,wecan
noticethat,afterbucklingappears,theaxialforcesofthatareaaredestabilizedand
decreased.Thisdecreaseiscausedbythewrinklingthatoccursduringbucklingand
causesatsomepointsthecompressiontoincreaseandatotherpointstodecrease
(eventoturnintotension).
Regardingthemomentdiagramstherearefewpointstobemade.Unliketheaxial
forcedistribution,momentispracticallythesamealonganypipecrosssection.That
doesnotapply,obviously,tothephasewhenbucklingoccurs,whichhasasaresult
verydifferentvaluesofmomentbetweenthetopandbottomfiber,butagain,only
atthelimitedareaofbuckling.
Finally,weexaminedthedistributionofsoilpressuresalongthepipeanditscross
sections.UsingthedataoutputfromABAQUSandanalgorithmwritteninMATLAB
we obtained the graphs of external pressure distribution along the pipe cross
section.Figure3.1.9presentsthepressuredistributionforthreecrosssections,A,B
andC,locatedatthehangingwall,atthemiddleofthemodelandatthefootwall
respectively. As it is demonstrated, at the area where the fault intersects the
pipeline(B),gapingoccursattheveryfirststepsoffaultdisplacement(indicatedby
thezerosoilpressuresatthelowerpartofthepipecrosssection)whereas,atthe
areasAandCittakessignificantlymorefaultmovementforpipesoilcontactlossto
occur. Specifically, during the first steps of small fault movement, the pressures
around these areas are almost geostatic. Further movement causes the loss of
contactfirstlyinthehangingwallandafterwardsinthefootwallpart.

94

Pipeboundaries

3.2 Fixed Pipe Ends Model

Apart from integrating all the aforementioned properties and geometries in our
model,werestrained,inaddition,allthedegreesoffreedomofthepipeendcross
sections.Thiswouldbeindeedavalidapproachifwecouldprovethatthestressesof
thepipehavealreadybeennullified,beforereachingthelimitsofourmodel,thus,
that there would be not significant changes between our model and a bigger one
thatwouldincludealongerpipeline.Itisafactthatwedidntexpectthistohappen
ina60mmodelmainlybecauseoftheaxialstressesthatoccurandhave,ingeneral,
awideareaofaction.However,wefoundreasonabletoruntheseanalyses,bothto
besureaboutourspeculationandforfurthercomparisonwiththeothermodelslet
alonethatthisisthemostcommonoptionbetweentheresearchers.Theresultsthat
cameoutofthismodelaredemonstratedinFigures3.2.13.2.9.
The pipe behavior is defined as follows. The pipe, at first, experiences almost only
axialtensilestress.Thetensilestressesandstrainsareinitiallymoreintenseattwo
pipeareas,rightandleftfromthepointwherethefaultplanecrossesthepipeline.
Loss of contact between the pipe and the surrounding soil occurs by the time the
fault reaches the pipeline. During almost all of the fault evolution, because of the
bending mechanism that occurs, the fibers with the greatest tensile strain are
located anti diametrically to those with the smaller tensile strain. In the free ends
scenario,thoselatterareasexperiencecompressivestressandarethosewherelocal
buckling occurs. Afterwards, the tensile strains continue to increase and reach the
marginoftheelasticstrain.Atthatarea,(thetopfibersofthepipenearthefault,in
the footwall part) the concentration and the rate of increase of plastic strains
augmentsleadingfinallytoexcessivetensilestrainsbeyond3%and5%thatwehave
setaslimits.Fromthere,eventhoughitisbeyondtheinterestofthisstudysincewe
considerthatthepipehasalreadyoperationallyfailed,thetensilestrainscontinueto
increase and a change of loading mechanism is spotted. This specific feature, the
transitionfromprevailingbendingtoprevailingtension,isdiscussedinthefollowing
caseofHybridBeammodel.

95

Pipeboundaries

AswenoticeinFigure3.2.1,regardlessthepipethickness,allofthepipesreachthe
failurecriteriaof3%and5%attheexactsamefaultoffsetandfailureoccursalmost
atthesamepointofthepipe.Theexplanationoftheabovecoincidenceliesinthe
modeldesign.Thesideofthehangingwall,wheretheoneendofthepipeisfixed,
movestheexactsameway(direction,valueandrateofdisplacement)asthebaseof
thehangingwallthatwemoveinordertosimulatethefault.Thedisplacementsof
thesidewerethesameforalltheanalysesofsamefaultangle.Hence,sincethepipe
isfixedbothtothemovablesideofthehangingwallandtotheimmovablesideof
the footwall we subject it to elongation, causing mainly (at the first steps at least)
thesameaxialdeformation(valuesandrate)despitethethicknessofthepipe.Soit
isnormalthatthestrainlimitsarereachedforalloftheD/tratiosatthesamefault
offset. This happens of course because, as shown by the analyses of this case, the
loading mode is very close to pure axial loading. It is possible that under different
faultangleswewouldnothavethisbehaviorsincethedeformationwouldbeeven
more concentrated and would cause more severe bending and shearing near the
faultplane.Asaconclusion,incaseswheresomeonewouldexpectsignificantaxial
loadingitwouldbeclearlywrongtousethefixedendsmodel.
It is evident, that under an ideal axial stress or strain loading the behavior of the
pipe, keeping the elastic modulus E and the length L the same, would be
independentofthethicknessofthepipeandthegeometricareaAingeneral.

F K x

E A
L

E A
x
L

96

Pipeboundaries

E A
L

F
x

(Fr

F A

E
x
L

y )

Comparing fixed to free edges scenario, in the latter case the pipe at the hanging
wall follows only a percentage of the movement magnitude of the adjacent soil
whereas the pipe on the footwall is dragged towards the direction of the hanging
wall,incontrasttothesurroundingsoilofthefootwallthatremainsrelativelystable.
Hence,nowthatwedesignedthepipeendstobefirmlyattachedtothesurrounding
soil we provoke axial tensile stress to the pipe having as a result bigger tensile
stressesandmoretensilebehavioringeneral.Thisisthereasonwhyinthefreeends
case the bending mechanism prevails leading to local buckling failure while in the
fixedendsscenariothetensilestressdoesnotallowthecompressivestrainsandasa
consequence the buckling to occur, leading to excessive tensile strain failure.
However,wecannotbesurewhichofthosetwoassumptionsisclosesttothereality
just by comparing them. For this reason we designed a more sophisticated model
thataddressestothisproblemasdescribedinthefollowingparagraph.
Regarding the axial forces (Figures 3.2.53.2.7.) and moment (Figure 3.2.8),
schematically, things do not differ a lot from the previous case of free edges.
However,asmentionedbefore,fixededgescausetensilebehaviorhavingasaresult
large values of tensile axial force at the pipe ends (instead of zero in the previous
case) and along the pipe in general, which remain tensile during all of the fault
movementsteps.Soilpresses,aswell,arenotpracticallyaffectedbytheboundary
conditionsofthepipe(Figure3.2.9.)Theonlydifferencespottedisthatgapingisless
intense,comparedtothefreeendsscenario,somethingthatisattributedprobably
tothefixedconditionsofthepipeendthatobligethepipe,atthatpoint,tofollow
the exact same movement of the soil, hence the loss of contact is slightly more
limited.

97

Pipeboundaries

3.3 HybridBeam Model

General geometry

Apartfromtheprevioustwoextremecases,wetriedtoapproachtherealbehavior
of the soilpipe system by making the least simplifications possible. In order to
achievethatweneitherfixedthepipeedgesnorletthemfree.Instead,wekeptthe
same 60m model and then we expanded it by 1 kilometer from each side (Figure
3.3.1). The key point is that, for the expansion parts, we now use beam elements
PIPE31forthepipeandnonlinearspringelementsSPIRNG2forthesurroundingsoil,
elements that are commented in the following paragraphs. Specifically, we added
beamelementsof1mwhosenodeswereconjoinedwithfive,nonlinear,1msprings
alongthevertical,horizontalandaxialdirectionrespectively(Figure3.3.1).

Beam element overview

A beam is an element in which assumptions are made so that the problem is


reduced to one dimension mathematically: the primary solution variables are
functions of position along the beam axis only. For such assumptions to be
reasonable,itisintuitivelyclearthatabeammustbeacontinuuminwhichwecan
define an axis such that the shortest distance from the axis to any point in the
continuum is small compared to typical lengths along the axis. The simplest
approach to beam theory is the classical EulerBernoulli assumption, that plane
crosssectionsinitiallynormaltothebeam'saxisremainplane,normaltothebeam
axis,andundistorted.Thisapproximationcanalsobeusedtoformulatebeamsfor
largeaxialstrainsaswellaslargerotations.ThebeamelementsPIPE31arebasedon
such a formulation, with the addition that these elements also allow transverse
shear strain; that is, the crosssection may not necessarily remain normal to the
beamaxis.ThisextensionleadstoTimoshenkobeamtheory(Timoshenko,1956)and

98

Pipeboundaries

is generally considered useful for thicker beams, whose shear flexibility may be
important.Weassumethat,throughoutthemotion,theradiusofcurvatureofthe
beamislargecomparedtodistancesinthecrosssection:thebeamcannotfoldinto
a tight hinge. A further assumption is that thestrain in the beam's crosssection is
thesameinanydirectioninthecrosssectionandthroughoutthesection.
In some piping applications thinwalled, circular, relatively straight pipes are
subjectedtorelativelylargemagnitudesofinternalpressure(weexaminetheeffect
ofinternalpressureinChapter6).Thishastheeffectofcreatinghighlevelsofhoop
stressaroundthewallofthepipesectionsothat,ifthesectionyieldsplastically,the
axial yield stress will be different in tension and compression because of the
interactionwiththishoopstress.ThePIPEelementsallowforthiseffectbyproviding
uniform radial expansion of the crosssection caused by internal pressure. Since
consideration of planar deformation only provides considerable simplification in
formulatingbeamelements,beamelementsonlymoveinthe(X,Y)plane.

Spring element overview

Spring elements couple a force with a relative displacement. SPRING2 is between


two nodes, acting in a fixed direction. The relative displacement across a SPRING2
elementisthedifferencebetweentheithcomponentofdisplacementofthespring's
firstnodeandthejthcomponentofdisplacementofthespring'ssecondnode:

99

Pipeboundaries

Springs Calibration

The forcedisplacement relation was obtained by the following procedure. We


designed a 3D F.E. model that included the pipeline and the soil of the
aforementioned characteristics and we subjected the buried pipeline to vertical,
horizontal and axial displacement. Based on the computed reaction forces and
displacements that came out of those pushover analyses, we obtained the spring
properties for the hybridbeam model. The spring forcedisplacement relation
diagramsareshowninFigures3.3.2and3.3.3.Asweobserve,theaxialspring(KA)is
activatedbothfornegativeandpositiveaxialdisplacement,whereas,theotherfour
springs are activated only for one direction of displacement and develop no force
during displacement towards the opposite one. The horizontal spring (KH) is the
sameforbothsides,sincethereisnodifferenceneitherregardingthesoilproperties
norregardingtheavailablesoillengthateachside,thus,thepassivereactionofthe
soil at the sides is the exact same. On the contrary, the vertical springs differ
significantly. The spring that is activated during downward displacement (KVdown)
developsmuchbiggerforcesthantheupperspring(KVup)forthepassivesoilreaction
that the limited surface soil layer can develop is significantly smaller than the
resistancethatisdemonstratedbytheinfinitesoillayerunderthepipe.

Results of HybridBeam model

TheresultsoftheanalysesconductedwiththebeammodelareillustratedinFigures
3.3.43.3.17. As it is depicted, this more realistic approach gives substantially
differentresultscomparedtotheprevioustwomethods.

100

Pipeboundaries

Comparison with the fixed and free ends models.

Firstofall,itisnowprovedthatthemodeoffailureforthiskindoffaultruptureis
the excessive tensile strain. This is in absolute disagreement with the results that
were taken from the free ends analysis. Free ends model does not take into any
considerationtheeffectoftherealeffectivelengthofthepipeandofitssurrounding
soil,whichexceedsbyfarthedimensionsofthebasicmodel.Thus,itisclarifiedthat
tochoosefreeboundaryconditionsforthepipe(atleastforthiskindofproblem)is
notinaccordancewithreality.
Compared to the fixed ends scenario, once again great differences are observed.
Apart from the mode of failure and the area where failure occurred, that are
practicallythesameinbothcasesandapartfromthegeneralshapeofdistributionof
axialstrainsattheareaoffailurethatissimilar,alltheotheraspectsofthesetwo
models differ substantially. Now that the axial forces and deformations have
adequate length to be developed and are not limited by the boundaries of the
model,thepipecanendurealotmorefaultdisplacement.Significantlyforthecase
ofD/t=72thepipecanendurealmost6timesbiggerfaultmovement(Figure3.3.3.).
The pipe now demonstrated bigger resistance, with respect to the fault
displacement,simplybecauseitisnolongersubjectedtosuchhightensilestressand
strainslikethosethatwerecausedbythefixedpipeboundaries.Comparingtheaxial
forcesthatdevelopinthismodelandinthefixedone,wenowobservesignificantly
smallertensileforces(Figure3.3.9.).Furthermore,unlikethefixedendsresults,the
3%and5%strainlimitarenowreachedfordifferentvaluesoffaultoffset,unlikethe
fixed ends case where the critical fault displacement was almost the same for 3%
and5%axialstrain(Figure3.3.3.).Theonlyexceptionconcerningthis,isthecaseof
D/t=72wherewenoticeoncemoreaverysmalldistancebetweenthecriticalfault
movementforthe3%axialstrainandthecriticalfaultmovementfor5%axialstrain.

101

Pipeboundaries

A change of loading mechanism

Apossiblereasonfortheaboveobservationisthataboveacertainlimitofrelative
displacementbetweenthefootwallandthehangingwallthedeformationsbecome
moreintenseandaccumulatealongthefaultsplanehavingasaresultthechangeof
thecriticalmodeofloadingfrombendingtoshearingthatresemblestodirectshear
loading. However, in the fixed ends case, the 3% and 5% limits are reached for
almostthesamehandforverysmallvaluesofit.Thus,thischangeofmechanismis
spotted even for limited relative displacement between the hanging wall and the
footwall.Inaddition,asalreadymentioned,fixedendscaseresultsinanalmostideal
axialtensiledeformation,somethingthatsuggeststhatthechangeofmechanismis
actually a transition from prevailing bending to prevailing tension. This mechanism
provokes a rapid development of strains which has as a result that one or two
additional centimeters of vertical fault displacement are enough for the strains to
increasefrom3%to5%.Themechanismcanbeidentifiedfirstlybythefactthatthe
criticalareamovestothepointofpipefaultintersection,whichisatthemiddleof
themodeledpipewherethereis,practically,nobendingdeformationandsecondly
by the strain distribution that change form and shape. Specifically, as long as the
bending prevails and causes the main tensile strain accumulation, the strains are
distributedoverarelativelylargeareaattheupperpartofthepipe,whereas,when
theprevailingfailuremechanismistension,strainsaccumulatealonganarrowarea.
Additionally,tensioncausestensilestrainallalongthecriticalpipecrosssectionsin
contrastto thebendingmechanismthattendstocompress(oratleastreducethe
tension) half of the pipe crosssection. All of the aforementioned observations are
depictedin(Figures3.3.6).Becauseofthischangeofloadingmechanismweobserve
twocriticalpoints:theonewhere3%strainisreachedbecauseofbendingandthe
otherwhere5%strainlimitisreachedbecauseofshearing(Figure3.3.8).

102

Pipeboundaries

Effective pipe length, Leff.

Veryinterestingresultscameupfromthismodelconcerningtheaffectedareaofthe
pipe.Thisisreallyimportanttobeexamined,sinceitisofgreatinteresthowmany
meters away from the pipefault intersection the effect of the rupture stops,
somethingthatwouldgiveapictureofhowlongaF.E.simulationshouldbeorinan
actual field case, something that could help make a speculation about how long
would be the affected length of the pipe from a potential fault rupture. For these
reasons westudied theeffective (or affected) length of the pipe Leff related to the
verticalfaultoffsetFIGUREandtotheD/tratio.Weconsideraseffectivelengththe
length of the pipe that expands from the pipefault intersection point to the point
where the axial force N is practically zero. Another way to calculate the effective
lengthisbyusingthepointwheretheaxialstrainsarealmostzero.Thismakesno
difference since at that area the pipe behaves totally elastically, hence, stresses
(forces) and strains are absolutely proportional. We also calculated the effective
lengthbothforthehangingwallpartandthefootwallpartandobservednegligible
difference (Figure 3.3.11). As we notice in Figure 3.3.12 at the very first steps of
fault movement the Leff is augmented almost vertically. After that it follows a
parabolictrenduntilitisfinallystabilized.Thisstabilizationoccursbecause,aftera
certainfaultdisplacement,thedeformationiscentralizedtothepointoffaultpipe
intersection and continues to augment in that failure area, until probably the
breakingofthepipe,causingnofurthereffecttothepipeawayfromthatarea.The
faultdisplacement,atwhichtheaxialstrainsstarttoincreasealmostverticallyatthe
pointoffailure,istheonebeyondwhichtheeffectivepipelengthremainspractically
unchanged.Figure3.3.12alsodemonstratesthatLeff isbiggerforsmallerD/tratios
thusforthickerpipes.Thickerpipeshavebiggermomentofinertia,hence,justlikein
the case of beam on elastic foundation, bigger inertia moment means, logically,
largerLeff.

103

Pipeboundaries

Detachment

It is demonstrated that the boundaries of the pipe do not affect that much the
gaping between the soil and pipe, since, once again, the contact between them is
lost from the first steps of fault displacement. The occurrence of gaping is
demonstrated in Figures 3.3.13 and 3.3.14. The best way to locate the loss of
contactistospottheareaswherethesoilpressuresarezero.Thisisthemethodwe
used in Figure 3.3.14 which shows that detachment occurs and expands mainly in
thehangingwallpartandthataboveacertainfaultdisplacementthelengthofthe
area of gaping remains stable. The only advantage of Figure 3.3.13, where we
demonstrategapingbyplottingthedifferentverticaldisplacementsofthepipeand
the soil, is that it gives us the exact height of the gaping. However, this is not
something of interest since it does not affect the pipe and it is only the area of
detachmentandthetimewhenitoccursthataffectthepipesbehavior.Lookingat
thecrosssectionalsoilpressuredistributionofFigure3.3.15wenoticethatgaping
mayoccurnotonlyatthelowerpartofthepipebutalsoattheupperpartofit.It
wouldbereasonablenottoexpectgapingtooccuratthatarea,sincenormally,sand
abovethepipe,justbecauseofthegravity,shouldfollowpipesmovementandfill
anygapthatisabouttooccur.However,asalreadymentionedinChapter2,because
ofcomputationalreasons,weintegratedasmallcohesioncoefficient(c=2kPa)toour
sandwhich,incombinationwithpossiblesoilarchingeffect,restrainthesoilabove
thepipefromcollapsingandallowsthegapingtooccur.

Effect of the pipe to the fault propagation

Regardingtheeffecttothepipetothefaultpropagationandthesurroundingsoilin
general,wedonotexpectsignificantdifferencescomparedtofreefieldfaultrupture
case. As demonstrated in Figures 3.3.16 and 3.3.17, the faults surface outcrop is
slightlyaffectedbytheintersectionwiththepipe.Themajoreffectofthepipeisthe
plastification that induces at the surrounding soil because of its resistance to the

104

Pipeboundaries

downwardmovementofthehangingwallandthebulgingthatoccursbecausethe
piperestrainsthesoilabovefromfollowingthedisplacementsoftherestofthesoil
ofthehangingwall.

Effect of the fault angle

Theeffectofthefaultangleisalsoexaminedbyrunningtheanalysisfora45fault
angle.Thecomparisonbetweenthe60and45faultanglecasesisshowninFigures
3.3.18 and 3.3.19. Now that we change the fault angle we cannot make pipe
resistance comparisons just in terms of vertical fault displacement. For this reason
wecomparetheaxialstraindevelopmentatthecriticalpointforbothcasesinterms
of vertical and horizontal fault displacement and in terms of total magnitude of
displacement.Asitdemonstrated,forthenormalfaultcase,failureoccursearlierfor
the45angleintermsofverticalandtotaldisplacementbutforbothanglesfailure
occurs almost at the same value of horizontal displacement. This observation
signifies that the horizontal component of the fault, thus the horizontal
displacement of the pipe is the critical one for the failure to occur. This is quite
logical,ifweconsiderthatthemodeofpipefailure,subjectedtonormalfaulting,is
theexcessivetensileaxialstrain.

105

Pipeboundaries

3.4 Equivalent Single Spring model

Basedonthepreviousmodel,wesubstitutedthebeampartsandsoilspringswith
justtwosprings,oneateverypipeend(Figures3.4.1.).Thespringcalibrationwas
basedontheHybridBeammodel.Specifically,theaxialreactionforcesattheedges
ofthepipewererecorded,throughoutthefaultmovement,withrespecttotheaxial
displacements.ThesedatawaslaterintegratedintotheForceDisplacementrelation
diagramoftheseaxialsprings,implying,thisway,theeffectsofanonlimitedmodel
withoutthelongbeampartsofthepreviousmodelthatcostintimeand
computationalpower.
AsdepicturedinFigures3.4.2and3.4.3.,thereisnotanysubstantialdifference
comparedtothehybridbeammodel,neitherregardingthecriticalfault
displacementwithrespecttotheD/tratio,norregardingthestraindevelopment
anddistribution.
Thisensuresthatthespringedgemodelis,indeed,abletocapturethebehaviorof
thesoilpipesystem,yieldingreliableresults.However,wedecidedtoproceedour
researchusingthehybridbeammodel,insteadofthelatterone,partlybecauseof
thesmall(howevernegligible)differencesthatoccurredandmainly,becauseby
usingspringedgeswewouldnothadbeenabletocomputeusefulandimportant
parameterssuchastheeffectivepipelength.

106

Pipeboundaries

Main conclusions of Chapter 3

Themorerealisticapproach,achievedwiththehybridbeammodel,provesthehigh
importance of the proper choice of the pipe boundaries. Comparing fixed to free
endscase,fixedmodeldoescapturetheactualmodeoffailurebut,nonetheless,the
resultsdiffersubstantiallyfromthoseobtainedfromthehybridbeammodel,which
weconsiderandproposeasthemostreliablemodelandtheonethatistheclosest
to the real conditions. Hence, a model, such as the proposed one, that combines
reducedtimeconsumptionwithrealismandreliability,isconsideredasacorrectand
practicalchoiceofboundariesinordertosimulatepipefaultsystemsusingafinite
elementprogram.
Pipes of smaller D/t ratios (thicker pipes of the same diameter) are able to
withstandlargerfaultmovements.
According to the results of the hybrid beam model a change of the prevailing
loadingmechanismoccurs.Asthefaultdisplacementincreases,sodoestherelative
movement between the hanging wall and the footwall, which leads to transition
fromprevailingbendingtoprevailingshearinglikeloadingmechanism.
Useful results came up by the examination of the effective length of the pipe.
Firstly, the effective length of the pipe increases with the increase of fault
displacement.Howeverthetrendofincreasereducesasthefaultdisplacementgoes
on,until,finally,intenseplastificationandfailureoccuratthemiddlepartofthepipe
hence the effective length is no longer affected. It is also demonstrated that Leff
dependsontheD/tratioandaugmentsasthepipethicknessincreases.
The critical displacement component for the normal fault case, that defines the
pointwhenthepipefails,isthehorizontalone.

107

108







Figures of Chapter 3

109

110

(a)

Freepipeends

(b)

Hoggingdeformation
Shaggingdeformation

Figure 3.1.1. (a) The deformed model shape . The deformed pipeline is highlighted in red.
The pipeline edges are free and the pipe is pulled towards the fault. (b) An example of
deformed pipeline and axial strain distribution. The correspondence of the two inflection
points of the deformed pipe shape with the strain distribution is depicted.
111

h=0.30m

h=0.90m

h=1.20m

Figure 3.1.2. Axial strain distribution for three vertical fault displacement h. Buckling occurs
at two areas making its first appearance in the footwall pipe part (D/t=96, Free ends).

112

h=0.30m
0.25%
0.20%
0.15%
0.10%
0.05%
x 0.00%
0.05% 20

25

30

35

40

35

40

35

40

0.10%
0.15%
0.20%
0.25%

X(m)

h=0.90m

6.00%
4.00%
2.00%
0.00%
2.00%
x

20

25

30

4.00%
6.00%
8.00%
10.00%
12.00%

X(m)

10.00%

h=1.20m

5.00%
0.00%
20

25

30

x 5.00%
10.00%

Topfiber

15.00%
20.00%

Bottomfiber
X(m)

Figure 3.1.3. The axial strain diagrams for the top and bottom pipe fiber along the critical
area, that correspond to the deformed shapes of the previous figure.

113

(a)

X(m)
33.2

33.4

33.6

33.8

34

34.2

34.4

34.6

0.10%

0.15%

0.20%

0.36m
0.37m

0.38m

0.25%

0.39m
0.40m

0.30%

0.41m

0.42m

0.35%

(b)

32

32.5

33

X(m)
33.5

34

34.5

35

0.10%
0.30%
0.50%
0.70%

D/t=144,h=0.42m

x 0.90%

D/t=120,h=0.48m
D/t=96,h=0.60m

1.10%

D/t=72,h=0.83m

1.30%
1.50%
1.70%

Figure 3.1.4. (a) The evolution of axial compressive strain at the buckling area, as the vertical
fault displacement increases.. h=0.42m corresponds to buckling initiation . (D/t=144). (b)
Comparison of the axial strains when buckling occurs for four different D/t ratios. (Free Ends)

114

h/D

(a)

0.1

0.2

0.3

0.4

0.5

0.6

0.0%

0.5%

X=33.93m

x 1.0%

1.5%

2.0%

(b)

h/D
0

0.2

0.4

0.6

0.8

0.0%
D/t=144
1.0%

D/t=120
D/t=96

2.0%

D/t=72

x 3.0%
4.0%

5.0%

6.0%

Figure 3.1.5. (a) Axial strain development at the point where buckling occurs (D/t=144). The
red circle indicates the start of buckling. (b) Comparison of axial strain developments at the
buckling areas for four D/t ratios.(Free ends model)

115

(a)
1
0.9
0.8
0.7
0.6
hcr/D 0.5
0.4
0.3
0.2
0.1
0
0

24

48

72

96

120

144

96

120

144

D/t

(b)
1.2%
1.0%
0.8%

xcritical 0.6%
0.4%
0.2%
0.0%
0

24

48

72

D/t

Figure 3.1.6. (a) Vertical critical displacement to pipe diameter hcr/D related to the pipe
diameter to thickness ratio for the free ends case. (b) The axial, compressive strain x when
buckling occurs for different D/t ratios for the free ends case.

116

(a)

Topfiber

8000
6000
4000

h=0.05m

2000

h=0.20m
N(kN)

h=0.50m
0

10

20

30

40

50

60

2000

h=0.80m
h=1.0m

4000
6000
8000

X(m)

(b)

Bottomfiber
8000
6000

N(kN)

4000

h=0.05m

2000

h=0.20m
h=0.50m

0
2000

10

20

30

40

50

60

h=0.80m
h=1.0m

4000
6000
8000

X(m)

Figure3.1.7.Axialforcedistributionanddevelopmentforfiveverticalfaultdisplacementsfor
(a)thetopand(b)thebottompipefiber. (D/t=72,Freeends)

117

Topfiber

(a)
2

0
0

10

20

30

40

50

60

2
h=0.05m

M(kNm) 4

h=0.50m
h=0.85m

10

(b)

X(m)

Bottomfiber

4
2
0
0

10

20

30

40

50

60

M(kNm) 2
h=0.05m
4

h=0.50m
h=0.85m

6
8
10
X(m)

Figure3.1.8.Momentdistributionanddevelopmentforthreeverticalfaultdisplacementsfor
(a)thetopand(b)thebottompipefiber.(D/t=72,Freeends)

118

h =0.05m

h=0.50m

h=1.00m

Figure 3.1.9. The soil pressures along three pipe crosssections (X=20, 40 and 60 m for A,B and
C respectively) for three different vertical fault displacements h. (D/t=72)

119

Fixedpipeends
(a)

(b)
0.40

0.30

hcr/D

3%

0.20

5%
0.10

0.00
0

24

48

72

96

120

144

168

D/t

Figure 3.2.1. (a) The undeformed and deformed shape of the Fixed ends model for a normal
fault of a vertical movement up to 2m. For this fault offset the pipe is deformed way above the
operational strain limits. (b) Critical fault displacement to pipe diameter hcr /D ratio versus the
diameter to thickness ratio D/t for D/t=72114.
120

(a)
h=0.10m

h=0.20m

h=0.30m

h=0.50m

(b)

Figure 3.2.2. (a) The deformation of the pipe for four vertical fault. (b) Axial strain at h=0.3m
exceeding the upper limit of 5%.(D/t=72, Fixed Ends)

121

(a)

6%

0.2m

5%

0.22m
4%
0.24m

x 3%

0.26m

2%

0.28m

1%

0.29m

0.30m

0%
31.9

32.1

32.3

32.5

32.7

32.9

33.1

X(m)

(b)
7%
6%
5%
4%

x
3%
2%
1%
0%
31.9

32.1

32.3

32.5

32.7

32.9

33.1

X(m)

Figure 3.2.3. (a) The evolution of axial tensile strain for vertical fault displacement up to 0.30m
with 0.28m and 0.30m corresponding to hcr for 3% and 5% respectively. (D/t=144). (b) Comparison
of the axial strains between four different D/t ratios for the critical state of strain values above 5%.
(Fixed Ends)
122

(a)

6%
5%
4%

3%

X=32.38m

2%
1%
0%
0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

h/D

(b)
6%
5%
4%

D/t=72
D/t=144

x 3%

D/t=120
2%

D/t=96

1%
0%
0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

h/D

Figure 3.2.4. (a) The axial strain of the most deformed element of the pipe at position X=32.38
versus the h/D ratio. (b) Comparison of the axial strain of the most deformed point of the pipe
between four different D/t ratios (Free Ends)

123

h=0.05m

h=0.30m

h=0.50m

Figure3.2.5.Thedistributionoftheaxialforcealongthepipelineforthreedifferentvaluesof
verticalfaultmovementh.(D/t=72,Fixedends)

124

Bottomfiber
6000
5000
4000
h=0.05m
h=0.10m

3000

N(kN)

h=0.15m
2000

h=0.20m
h=0.30m

1000
0
0

10

20

30

40

50

60

X(m)

Topfiber
6000

5000

4000
h=0.05m
3000

h=0.10m

N(kN)

h=0.15m
h=0.20m

2000

h=0.30m
1000

0
0

10

20

30

40

50

60

X(m)

Figure 3.2.6. The distribution of the axial force along the pipeline for five different values of
vertical fault movement h. (D/t=72, Fixed ends)

125

Topfiber
6000

5000

4000

N(kN)
3000
h=0.30m
h=0.40m

2000

h=0.50m
1000

0
0

10

20

30

40

50

60

X(m)

7000

Bottomfiber

6000

5000

N(kN)

4000

3000
h=0.30m
2000

h=0.40m
h=0.50m

1000

0
0

10

20

30
X(m)

40

50

60

Figure 3.2.7. The distribution of the axial force along the pipeline for three different values of
vertical fault movement h. The h=0.30m is the critical displacement above which we observe a
general decrease and distortion of the axial force distribution. (D/t=72, Fixed ends)
126

(a)

h=0.05m

h=0.30m

h=0.50m

(b)

2
h=0.05m
M(kNm) 0
0

10

20

30

40

50

60

h=0.15m
h=0.30m

X(m)

Figure 3.2.8. (a) Schematic distribution of moment along the pipeline and (b) diagrammatic
distribution of moment along the middle pipe fiber., for three fault displacements.
(D/t=72, Fixed ends)
127

h=0.00m

h=0.05m
A

h=0.25m
A

h=0.50m
C
A

Figure 3.2.9. The distribution of soil pressures along the pipeline and along the pipe cross
section a three points A,B and C (X=20, 40 and 60m respectively) for four vertical fault
displacements h. (Dense sand, D/t=72, Fixed ends)
128

HybridBeammodel
(a)

(b)

Figure 3.3.1. (a) The hybridbeam model, initial and deformed shape . (b) Detail of the beam
and spring elements connected to the pipe end.

129

KVup

KH
KA

KH
KVdown
F (kN)

KA

40
30
20
10
0

0.006

0.004

0.002

10

0.002

0.004

0.1

0.2

0.006
Ux (m)

20
30
40

KH

F(kN)

700
600
500
400
300
200
100
0

0.3

0.2

0.1

0.3
Uy (m)

Figure 3.3.2. Spring force with respect to the axial displacement (KA) and to the horizontal
displacement (KH).
130

KVup
KH
KA

KH

KVdown
F(kN)

160
140
120
100

KVup

80
60
40
20
0
0.2

0.1

0.1

0.2
Uz (m)

F(kN) 600
500
400
300

KVdown

200
100
0
0.4

0.2

0.2

0.4
Uz (m)

Figure 3.3.3.. Spring force with respect to the vertical displacement Uz for the two vertical
springs that are activated by upward (KVup)and downward (KVdown) movement respectively.
The forces of the downward spring are greater because of the bigger passive soil resistance
that develops against downward movement in comparison with the smaller soil resistance
that the limited surface soil layer activates against the upward movement of the pipe.
131

(a)

HybridBeamModel
2
3%

1.8

5%

1.6
1.4
1.2
hcr/D

1
0.8
0.6
0.4
0.2
0
0

24

48

72

96

120

144

168

D/t

(b)

ComparisonwithFixedModel
2
1.8

3%FixedModel

1.6

5%FixedModel

1.4

3%HybridBeamModel
5%hybridBeamModel

1.2
hcr/D

1
0.8
0.6
0.4
0.2
0
0

24

48

72

96

120

144

168

D/t

Figure 3.3.4. (a) Vertical critical displacement to pipe diameter hcr/D related to the pipe
diameter to thickness ratio for the Hybrid Beam model. (b) Comparison of the previous
diagram with the relative one of Fixed ends model.
132

(a)
4.0%
3.5%
3.0%
2.5%

FixedEnds

x 2.0%

HybridBeamModel

1.5%
1.0%
0.5%
0.0%
31

32

33

X(m)

34

35

(b)
7%
6%
5%
4%

FixedEnds

3%

HybridBeamModel

2%
1%
0%
31.9

32.1

32.3

32.5

32.7

32.9

33.1

X(m)

Figure 3.3.5. Comparison of the axial strain distribution between the Fixed ends model and
the Hybrid Beam model, at the two critical states of (a) 3% tensile strain and (b) 5% tensile
strain. (D/t=96)
133

Figure 3.3.6. Bending mechanism prevails from the beginning of fault displacement until the
point where the big relative displacement of the hanging wall and the footwall causes the
change of the loading mechanism.

134

(a)

7%

6%

0.50m

0.57m

5%

0.65m

4%

0.70m

3%

0.75m
0.80m

2%

0.85m
1%

0.93m

0%

0.90m
31.5

32

32.5

33

33.5

34

34.5

X(m)

(b)

3.5%
3.0%
2.5%

2.0%

D/t=144

1.5%

D/t=120

1.0%

D/t=96

0.5%

D/t=72

0.0%
32

32.5

33

33.5

34

34.5

35

35.5

X(m)
7%

(c)

6%
5%

4%

D/t=144

3%

D/t=120

2%

D/t=96

1%

D/t=72

0%
31.5

32

32.5

33

33.5

34

X(m)

Figure 3.3.7. (a) Axial tensile strain distribution and development along the critical area
(D/t=144). Comparison of axial tensile strain distribution between the four different D/t
ratios for the limit states of (b) 3% tensile strain and (c) 5% tensile strain.
135

10%
9%

(a)

8%
7%
6%

x 5%

X=33.78m

4%

X=32.34m

3%
2%
1%
0%
0

0.2

0.4

0.6

0.8

h/D

5%

(b)

4%
3%

D/t=72,X=33.78m

D/t=96,X=34.10m

2%

D/t=120,X=34.38m
D/t=144,X=34.83m

1%
0%
0

(c)

0.2

0.4

0.6

0.8

h/D
6%
5%
4%
D/t=72,X=32.59m

x 3%

D/t=144,X=32.34m

2%

D/t=120,X=32.35m
D/t=96,X=32.34m

1%
0%
0

0.5

1.5

h/D

Figure 3.3.8. (a) Axial tensile strain development with respect to the vertical fault
displacement to diameter ratio at two points X=33.78 and X=32.34m where the limits of 3%
and 5% tensile strain are reached , caused by bending and shearing respectively (D/t=144).
(b) The axial strain development at the points where the 3% tensile strain is reached for four
D/t ratios. (c) The axial strain development at the points where the 5% tensile strain limit is
reached for four D/t ratios.
136

(a)
Bottomfiber

7000
6000
5000
4000

N(kN)

3000

0.10m

2000

0.50m

1000

1.00m
1.70m

0
1000

10

20

30

40

50

60

2000
3000
4000

(b)

X(m)

7000

Topfiber

6000
5000
4000
3000

0.10m

N(kN)

0.50m

2000

1.00m
1000

1.70m

0
0

10

20

30

40

50

60

1000
2000

X(m)

Figure 3.3.9. Axial forces along (a) the top and (b) the bottom pipe fiber, for four values of
vertical fault displacement.

137

(a)
6

2
h=0.10m

0
0

10

20

30

M(kNm)

40

50

60

h=0.50m
h=1.00m

h=1.75m

X(m)

(b)
h=0.10m

h=0.50m

h=1.00m

h=1.75m

Figure 3.3.10. (a) Diagrammatic distribution of moment along the top pipe fiber and (b)
schematic display of moment values along the pipe.(D/t=72)
138

(a)

(b)

Leff
16000
14000
12000

Footwall

10000
N(kN) 8000

Hangingwall

6000
4000
2000
0
30

230

430

630

830

1030

|X|(m)

X=30

X=0

X=30

Figure 3.3.11. (a) The axial force distribution along our model. We consider as Leff the length
of the segment that extends from the faultpipe intersection point (practically at the middle
of the model) to the point where the axial force is almost 0. (b) Comparison of the axial
forces along the two beam parts of the model at the hanging wall and at the footwall
respectively., no significant difference is observed.
139

(a)

(b)
700
0.25

600

0.20

500
400

Leff/D

300

0.15
0.10

200
0.05

100
0
0

0.5

1.5

0.00

0.5

h/D

(c)

1.5

h/D

16000
14000
12000
10000

h=0.10m

N(kN) 8000

h=0.50m
h=1.00m

6000

h=1.50m

4000

h=1.75m

2000
0
30

(d)

230

430

630

830

1030

X/D
600
500
0.25m

400

0.5m

Leff/D 300

0.75m
0.9m

200

1m
100
0
0

24

48

72

96

120

144

168

D/t

Figure 3.3.12. (a) The effective length (hanging wall side) with respect to the vertical fault
displacement . (b) The axial tensile strain at the point of failure. (c) The axial forces along the
beam elements of the hanging wall for five values of h. The increase rate of Leff remains
stable for a few centimeters of fault displacement and then decreases as the fault
displacement augments. When the tensile strain starts to increase rapidly (point of failure
because of shear mechanism) the Leff practically stops increasing. (D/t=72) (d) The effective
length Leff with respect to the diameter to thickness ratio D/t.
140

h=0.02m
0.005
0
0

10

20

30

40

50

60

0.005
Soil

Uz(m) 0.01

Pipe
0.015
0.02
0.025

X(m)

h=1.0m
0

10

20

30

40

50

60

0
0.2
0.4
Soil

Uz(m) 0.6

Pipe

0.8
1
1.2

X(m)

Figure 3.3.13. Detachment of the pipe from the surrounding soil depicted for two values of
vertical fault displacement h, by comparing the vertical displacement of the bottom pipe
fiber with the vertical displacement of the soil beneath. Gaping occurs almost instantly
within the very first centimeters of fault displacement. (D/t=72)
141

h=0.01m

h=0.01m

h=0.10m

h=0.10m

h=1.00m

h=1.00m

h=1.75m

h=1.75m

Figure 3.3.14. Pipe detachment from the surrounding soil can be clearly depicted based on
the soil pressure on the pipe. When the soil pressure becomes zero (blue areas) ,the loss of
contact between soil and pipe is indicated. The gaping is formed from the very beginning of
fault displacement, its length increases as the fault displacement augments until a maximum
point above which further fault displacement does not affect the length of the gaping area.

142

h=0.00m

h=0.05m

h=0.50m

h=1.00m

h=1.75m

Figure 3.3.15. Soil pressure along three pipe crosssections A,B and C (X=20,30 and 40m
respectively) for several vertical fault displacement h. (D/t=72)
143

Figure 3.3.16. Magnitude of plastic strains for h=1.75 . Planes of various distances from
the pipe are depicted. (y=0 corresponds to the vertical plane of the central pipe axis)
(D/t=72).
144

Figure3.3.17.Verticaldisplacementdistributionforverticalfaultdisplacementh=1.75m.
Planesofvariousdistancesfromthepipe.(y=0correspondstotheverticalplaneofthe
centralpipeaxis)(D/t=72).
145

(a)

(b)
0.1

0.1

0.09

0.09

0.08

0.08

0.07

0.07

0.06

0.06

x 0.05

x 0.05

0.04

0.04

0.03

0.03

0.02

0.02

0.01

0.01

0
0

0.5

1.5

0.5

1.5

l(m)

h(m)
0.1
0.09
0.08
0.07
0.06

60

x 0.05

45

0.04
0.03
0.02
0.01
0
0

0.5

1.5

2.5

(m)

Figure 3.3.18. (a) The deformed soil and pipe for a normal fault of 45. (b) Comparison
between the 60 and 45 normal fault in terms of axial deformation with respect to the
vertical fault displacement h, the horizontal fault displacement l and the total magnitude of
displacement , (D/t = 72).
146

(a)

700
600
500
400

Leff/D
300
200
100
0
0

0.5

1.5

h(m)

(b)
16000
14000

45

12000

60

10000
N(kN) 8000
6000
4000
2000
0
30

130

230

330

430

X(m)

Figure 3.3.19. Comparison of a) the effective pipe length and b) the beam force
distribution between the 45 and 60 analyses , for normal faulting (D/t = 72).

147

SpringEdgeModel
(a)

KA
(b)

20000

F (kN)

15000
10000
5000
0
0.6

0.4

0.2

5000

0.2

0.4

0.6

Ux (m)

10000
15000
20000

Figure 3.4.1. (a) The Spring edge model where we substituted the beams and soil springs of
the Hybrid Beam model with a single axial spring at the edge of the pipe.(b) The axial spring
force F with respect to the axial displacement Ux.
148

Springedgemodel
(a)

2
1.8
1.6
1.4
1.2
hcr/D

3%

0.8

5%

0.6
0.4
0.2
0
0

24

48

72

96

120

144

168

D/t

ComparisonSpringEnd Hybridbeam

(b)

HybridBeammodel

2
1.8

Springendsmodel

1.6
1.4
1.2
hcr/D

1
0.8
0.6
0.4
0.2
0
0

24

48

72

D/t

96

120

144

168

Figure 3.4.2. (a) Critical fault displacement to pipe diameter hcr /D ratio versus the
diameter to thickness ratio D/t for D/t=72114. (b) Comparison of the spring edge model to
the hybrid beam model with respect to the critical vertical fault displacement for D/t=72
114.
149

(a)
9%
8%
7%

Springedgesmodel

6%

HybridBeammodel

5%

4%
3%
2%
1%
0%
31.5

32

32.5

33

33.5

34

34.5

35

35.5

X(m)

(b)
10%
9%
8%
7%
6%
HybridBeammodel

x 5%

Springedgemodel

4%
3%
2%
1%
0%
0

0.5

1.5

h/D

Figure 3.4.3. (a) Comparison of axial tensile strain distribution for vertical fault movement
h=1.63 and 1.65m and (b) comparison of the axial tensile strain with respect to the
vertical fault movement, between the Spring edges and Hybrid Beam model . (D/t=72)

150

151

152





Chapter 4
Reverse fault rupture

153

154

Reversefaultrupture

Introduction

In this chapter, we design and examine the behavior of a pipe when subjected to
reverse fault rupture. For our analysis we use the hybrid beam model whose
reliability has been discussed and proven in the previous chapter. However, we
conductabriefcomparisonbetweentheresultsofthismodelandtheresultsofthe
modelswithfreeandfixedpipeends,togetaclearviewabouthowclosetoreality
thelattermodelsare.

Model Properties

ThemodelsweuseinthischapterarepracticallythesamewiththoseofChapter2:
geometry, elements, constitutive model, and pipe and soil properties are kept the
same as in Paragraph 2.2. Two are the major differences implemented in these
analyses.
The first is the obvious change of applied movement. Instead of a downward we
applyanupwardmovementkeeping,however,thesameangleof60,inordertobe
abletodirectlycomparethetwocases,normalandreverse.
Secondly,weadjustedtheneutralearthpressurecoefficient(Ko)changingitsvalue
from0.29to1.0.Thereasonforthischangeisthat,normallyinthecaseofreverse
fault, before the occurrence of rupture, horizontal stresses have already increased
becauseofthecompressivenatureofreversefault.Hence,inordertocapturethis
initial state we increase the value of Ko to be closer to the passive earth pressure
coefficient Kp. Respectively, the normal fault case is of tensile nature thus, the
horizontalstresseshavealreadybeendecreasedbeforetheruptureandthevalueof
KoischosentobeclosertotheactiveearthpressurecoefficientKa.

155

Reversefaultrupture

4.1 Brief Review of Fixed and Free


Ends Model

Free ends model

Wepresenttheresults(Figures4.1.24.1.4)onlyforacertainpipethickness(D/t=
72).Sincewehavealreadyconcludedthatthehybridbeammodelisthemost
reliable,wesimplywanttobrieflydemonstratethedivergenceofthetwomodels.
AsdepictedinFigure4.1.2thepipelinefailsbecauseoftheoccurrenceofbucklingat
verticalfaultoffseth=0.49m(h/D=0.54).Figure4.1.3demonstratestheaxialstrain
distributionalongthecriticalareaofthepipe.Weobservethatthemajorityofthe
pipeissubjectedtocompressionwiththeexceptionofthetwoinflectionareas
wheretensilestressesandstrainsappearandwhereatlaterhvalueslocalbuckling
occurs.First,bucklingappearsinthehangingwallside(wherehoggingdeformation
appears)andthisiswhenweconsiderthefailureofthepipe.Asthefaultmovement
continuesasecondlocalbucklingoccursatthepipepartinthefootwall(where
shaggingdeformationhasbeenformed).Therelativeaxialstraindiagrams,fortop
andbottomfiber,areshowninFigure4.1.4.


Fixed ends model

Onceagain,wepresenttheresultsonlyforapipethicknesst=0.0127m(D/t=72).
Bucklingisthemodeoffailureinthiscase,aswell.Bucklingappearsfirstlyinthe
hangingwallandtheninthefootwall,likeinthepreviouscase.AsshowninFigure
4.1.5,thecriticalverticalfaultdisplacementismerely0.07m(h/D=0.08).Thehuge
differencebetweenfreeendsandfixedendssolutionsisexplainedbythefactthat
reversefaultsubjectsthepipemainlytocompressionandsincetheboundariesare
fixed,thepipeisunabletomoveandthusthestressesincreasemorerapidly.

156

Reversefaultrupture

4.2 HybridBeam Model

Incontrastwiththepreviousextremeandunrealisticcasesoffreeandfixedends,
thisapproachaddressestotheproblemofboundariesinrationally.Withouteven
examiningtheresultsoftheanalysis,wecouldpredictthatthefailurewilloccur
somewherebetweenthepreviousextremeapproaches.Asmentionedbefore,
fixededgesrestrainthepipefrommoving,thus,compressivestressesincreaseand
bucklingoccursearly,whilefreeedgesallowthepipetoaxiallymovealmostfreely
(affectedonlybythefrictionforce).Sinceinbothcasesfailurewasduetobuckling,
weexpectbucklingtobethefailuremodeinthiscaseaswell.Sincethehybridbeam
modeltakesintoaccountthecontinuationofthepipe,weexpectbuckling(forD/t
=72)tooccurforaverticalfaultmovementbetween0.07m(fixedends)and0.49m
(freeends).
Indeed,asshowninFigures4.2.1and4.2.3localbuckling(forD/t=72)occursat
0.25mofverticalfaultdisplacementh.Similartothenormalfaultcase,thickerpipes
canendurebiggerfaultdisplacements(Figure4.2.1)andlargeraxialstrains(Figure
4.2.4)beforetheoccurrenceofbuckling.Again,bucklingoccursalonganarrowarea
ofabout0.2mlengthforeveryD/tratio.

Effective pipe length, Leff

WedefineLeffasinParagraph3basedonthepointwhereaxialforcesinthebeam
partsbecomezero.InFigure4.2.6.a,acomparisonofbeamaxialforcesinthe
hangingwallandthefootwalldoesnotindicategreatdifferencesbutmakesclear
thatinthefootwallthepipeismorestressedcomparedthaninthehangingwall.The
effectivelengthincreasesrapidlyduringthefirststepsofverticaldisplacement(until
h/D=0.06)andthentherateofincreasediminishesuntiltheLeffremainspractically
stable(h/D=0.38)(Figure4.2.6).Inthesamefigure,theeffectofD/tratiois
demonstrated,showingthatgreaterlengthofpipeisactivatedforthethickerpipes.

157

Reversefaultrupture

Furthermore,wecomparethemagnitudeandareaofeffectoftheaxialbeamforces
inthefootwallandasitisdepictedinFigure4.2.6.b,thereversefaultactivates
greaterlengthofpipeforthesamevalueofdislocation.

Detachment and pipe effects to the soil

ThedetachmentofthepipefromthesoilisdemonstratedinFigures4.2.8and4.2.9
intermsofrelativepipeandsoilverticaldisplacementsandsoilpressures,
respectively.Inthefirstcase,weusetheverticaldisplacementsofthelowerpartof
thepipeandofthesoilsurfacebeneaththepipeandinthesecondfigure,lossof
contactisdepicturedbytheareaswheresoilpressureiszero.Thelatterwayof
displayismoreefficientsinceitcandemonstratethedetachmentareasalongthe
entirepipesurfaceandnotonlythelowerpipepart.Unlikethenormalfaultcase,
reverserupturecausesdetachmenttooccurmainlyatthefootwallpartofthepipe.
Detachmentcanalsobedemonstratedbyplottingthesoilpressurearounddifferent
pipecrosssections(Figure4.2.10).
Finally,regardingthegeneraleffectofthepipetothesoil,weobserveinFigure
4.2.11,thatthepipehindersthefaultfrompropagatinguptothesurfaceandin
Figure4.2.12,wenoticethatcomparedtothefreefieldrupture(y=5m)thepipe
provokesadditionaldisplacementtothesoilsinceitpushesitupwardsand
rightwards,accordingtothefaultmovementandthepipedeformation.

158

Reversefaultrupture

Effect of the fault angle

Theeffectofthefaultangleisalsoexaminedbyrunningtheanalysisfora45fault
angle.Thecomparisonbetweenthe60and45faultanglecasesisshowninFigures
4.2.13and4.2.14.SimilarlytotherelativeparagraphofChapter3,weexaminethe
pipe resistance in terms of vertical, horizontal and total fault displacement. We
observe that, unlike the normal fault case, the critical parameter now is the total
fault displacement since in both fault angle cases the pipe fails for the same total
fault displacement. This is also quite logical, since the mode of pipe failure when
subjected to reverse faulting, is the local buckling. For local buckling to occur it is
importanttheincreaseofcompressiveaxialstrains(causedmainlybythehorizontal
movementcomponent)aswellastheoccurrenceofbending(causedbythevertical
movementcomponent).

159

Reversefaultrupture

Main conclusions of Chapter 4

Onceagain,thelimitedmodelsoffixedandfreepipeendsareproveninadequate
tocaptureanddescribecorrectlythebehaviorofthesoilpipesysteminthecaseof
reversefaultrupture.
Themodeoffailureforthiskindoffaultislocalbuckling,duetothecompressive
natureofreverserupture.
Subjected to reverse fault rupture, pipe can endure significantly less fault
movementbeforetheoccurrenceoffailure,comparedtonormalfaultrupture.
Comparedtothenormalfaultcase,reversefaultactivatesalongerpartofthepipe.
Ascriticalfaultmovementcomponent,itappearsthatthecriticalparameteristhe
totalfaultdisplacement.

160

161

162







Figures of Chapter 4

163

164

FreePipeEnds

60

60

z
x
y

Figure 4.1.1. Simulation of reverse fault rupture, using the basic model of Chapter 3 with free
pipe ends.

165

(a)

x
(b)

28

28.2

28.4

28.6

28.8

29

29.2

29.4

29.6

29.8

30

0.00%
1.00%
2.00%
3.00%
h=0.40m

x 4.00%

h=0.45m
h=0.49m

5.00%
6.00%
7.00%
8.00%

X(m)

Figure 4.1.2. (a) The deformed pipeline shape is highlighted in red. The relative movement
between the pipe and the soil indicates that , if not for the free boundaries of the pipe, a
compression would have occurred, caused by the continuation of the pipe , which is not taken
into consideration in the free edges model. (b) Axial strain distribution in the critical area.
Buckling appears at 0.49 m of vertical fault displacement.

166

h=0.10m

h=0.30m

h=0.50m

h=0.90m

x
Figure 4.1.3. Axial strain distribution along the pipe for four vertical fault displacements.
(D/t = 72,Free ends)

167

h=0.10m

0.10%
0.05%
0.00%

22

24

26

28

30

32

34

36

38

40

34

36

38

40

34

36

38

40

0.05%
0.10%
0.15%

h=0.30m
0.20%
0.10%
0.00%

0.10%

22

24

26

28

30

32

0.20%
0.30%
0.40%

h=0.50m

1.00%
0.00%
1.00%

22

24

26

28

30

32

2.00%
3.00%

Topfiber

4.00%
5.00%

Bottomfiber

6.00%
10.00%

h=0.90m

5.00%
0.00%
5.00%

22

24

26

28

30

32

34

36

38

40

x 10.00%
15.00%
20.00%
25.00%
30.00%

X(m)

Figure 4.1.4. Diagrammatic axial strain distribution along the critical area for four vertical
fault displacements. (D/t = 72,Free ends)

168

FixedEnds
(a)
h=0.02m

h=0.07m

h=0.10m

(b)
28.8

28.9

29

29.1

29.2

29.3

29.4

29.5

29.6

0.00%
1.00%
2.00%
3.00%

h=0.06m

x 4.00%

h=0.07m

5.00%
6.00%
7.00%
8.00%

X(m)

Figure 4.1.5. (a) Axial strain distribution along the critical area. The pipe is under compression
from the very beginning of fault movement. (b) Axial strain distribution along the critical area
(bottom fiber) when first buckling occurs. (D/t = 72, Fixed ends)

169

Reverse HybridBeammodel
(a)

(b)
0.3
0.25
0.2

hcr/D 0.15
0.1
0.05
0
0

24

48

72

96

120

144

168

D/t

Figure 4.2.1. (a) The aforementioned HybridBeam model used for the simulation of reverse
fault motion. (b) The summarized dimensionless results of the analysis in terms of critical
vertical fault displacement with respect to the diameter to thickness ratio D/t.

170

h=0.10m

h=0.20m

h=0.25m

buckling

h=0.40m

1st

2nd

Z
X
Figure 4.2.2. The axial strain distribution along the critical pipe area (X=2540 m) for four
vertical fault displacements h. (D/t = 72). First buckling occurs in the hanging wall side.

171

(a)

28

28.2

28.4

28.6

28.8

29

29.2

29.4

29.6

29.8

30

0.00%

1.00%

2.00%
h=0.10m

x3.00%

h=0.20m
h=0.25m

4.00%

5.00%

X(m)

6.00%

(b)

29

29.1

29.2

29.3

29.4

29.5

29.6

0.00%

0.50%

D/t=144

1.00%

D/t=120
D/t=96

x 1.50%

D/t=72
2.00%

2.50%

X(m)

3.00%

Figure 4.2.3. (a) Axial stain development for three vertical fault displacements, for D/t = 72.
(b) Axial strain distribution at the moment of buckling initiation for four D/t ratios..

172

(a)
0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.00%
0.50%
1.00%
D/t=144
D/t=120

x 1.50%

D/t=96
D/t=72

2.00%
2.50%
3.00%

h/D

(b)

24

48

72

96

120

144

168

0
0.005
0.01

xcritical

0.015
0.02
0.025
0.03

D/t

Figure 4.2.4. (a) Axial strain development at the point where buckling occurs for four D/t
ratios. (b) The critical axial strain at which buckling occurs , for four D/t ratios.

173

h=0.30m

h=0.10m
3000

2000

2000
1000
N(kN)

1000

0
0

20

40

60

N(kN)

1000

1000

20

40

60

2000
2000

3000
4000

3000

5000

4000

6000
5000

3000

3000

h=0.35m

1000

1000
0

1000 0

h=0.40m

2000

2000
N(kN)

X(m)

7000

X(m)

20

40

60

N(kN)

0
0

20

40

1000

2000

2000

3000
4000

3000

5000
4000

6000
7000

X(m)

X(m)

Topfiber
Bottomfiber
Figure 4.2.5. Axial forces of the upper and lower fiber, for four vertical fault displacements.
The selected steps indicate the initial condition, the initiation of the first buckling , the
initiation of the second buckling and the axial forces distribution several steps after both of
buckling failures have occurred. (D/t = 72)

174

60

EffectivepipelengthLeff

(a)

30

80

130

180

230

280

330

0
1000
2000

Hangingwall

3000

N(kN)

Footwall

4000
5000
6000

|X|(m)

7000

(b)

8000
6000

Normal

4000

Reverse

2000

N(kN)

0
30

230

430

630

830

1030

2000
4000
6000
8000

|X|(m)

X=30

X=0

X=30

Figure 4.2.6. (a) Comparison of axial forces along the beam in the hanging wall and the
footwall. The differences are not large but demonstrate that footwall is more stressed
compared to hanging wall. (b) Comparison between normal and reverse cases regarding the
effective pipe length . Reverse activates a longer part of the pipe for the same fault
displacement.

175

400

(a)

350
300
250

Leff/D

200
150
100
50
0
0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

h/D
0

(b)

30

230

430

630

830

1030

1000
2000

h=0.05

3000

h=0.10m

N(kN)

h=0.20m

4000

h=0.30m
5000

h=0.40m

6000
7000

(c)

X(m)

300
250
200

Leff/D

h=0.05m
h=0.10m
h=0.13m

150
100
50
0
0

24

48

72

D/t 96

120

144

168

Figure 4.2.7. (a) Effective pipe length as fault displacement increases. Just like with normal
fault case, during the first steps the increase is large while, subsequently, the rate of increase
diminishes until the Leff gets a maximum value. (b) The development of axial forces, as the fault
movement increases ,indicates the value of Leff (D/t = 72). (c) The axial forces along the beam
elements in the footwall for three values of h.

176

h=0.03m

0.035
0.03

Soil

0.025
0.02

Pipe

Uz (m) 0.015
0.01
0.005
0
0.005

10

20

30

40

50

60

40

50

60

40

50

60

X(m)

h=0.20m

0.25
0.2
0.15

Uz (m)

0.1
0.05
0
0

10

20

30

0.05

X(m)

h=0.40m
0.45
0.4
0.35
0.3
0.25
Uz(m) 0.2
0.15
0.1
0.05
0
0.05 0

10

20

30

X(m)

Figure 4.2.8. Detachment of the pipe from the soil beneath it , depicted by plotting and
comparing the vertical displacements off pipe and soil (D/t = 72).

177

h=0.03m

h=0.10m

h=0.25m

h=0.40m

Figure 4.2.9. Detachment of the pipe, indicated by the pipe areas where the soil pressure is
zero (blue areas).

178

h=0.10m

h=0.20m

h=0.40m

Figure 4.2.10. Soil pressure around the pipe crosssection in three positions A,B and C of
X=20,30 and 40m respectively. (D/t = 72)

179

Figure 4.2.11. Magnitude of plastic strains for h = 0.40 m . Planes of various distances from
the pipe are depicted. (y=0 corresponds to the vertical plane of the central pipe axis).
No significant lateral effects are spotted.(D/t = 72)

180

Figure 4.2.12. Vertical displacement distribution for vertical fault displacement h=0.40 m.
Planes of various distances from the pipe. (y = 0 corresponds to the vertical plane of the
central pipe axis) (D/t=72).

181

(a)

45

(b)
0
0.02
0.04
0.06
0.08
x 0.1
0.12
0.14
0.16
0.18
0.2

0.1

0.2

0.3

h(m)
0
0
0.02
0.04
0.06
0.08
x 0.1
0.12
0.14
0.16
0.18
0.2

0.1

0
0.02
0.04
0.06
0.08
x 0.1
0.12
0.14
0.16
0.18
0.2
0.1

0.2

0.2

l(m)
0.3

0.4

60
45

(m)

Figure 4.2.13. (a) The deformed soil and pipe for a reverse fault of 45. (b) Comparison
between the 60 and 45 reverse fault in terms of axial deformation with respect to the
vertical fault displacement h, the horizontal fault displacement l and the total magnitude of
displacement , (D/t = 72).

182

0.3

(a)

400
350
300
250

Leff/D 200
150
100
50
0
0

0.1

0.2

0.3

0.4

0.5

h/D

(b)

30

130

230

330

430

0
1000
2000
3000

N(kN) 4000
5000
6000

45

7000

60

8000
X(m)

Figure 4.2.14. Comparison of a) the effective pipe length and b) the beam force
distribution between the 45 and 60 analyses , for reverse faulting (D/t = 72).

183

184





Chapter 5
Soil layer depth effect

185

186

Thesoillayerdeptheffect

Introduction

Havingtestedandinvestigatedseveralmodelsforthesimulationoftheproblemwe
came to the conclusion that the most realistic and reliableone is the hybridbeam
model. The reason is that it does not need any arbitrary assumption about the
boundaryconditionsofthepipe,sinceitispracticallynotlimited(inthelongitudinal
direction)anddepicturesaccuratelyboththeaffectedareaandtheactualbehavior
ofthepipe.Nonetheless,therearestillsomeaspectsofthismodelthatarenotin
accordancewiththecommon,realconditions.Theseremaininginaccuracieshaveto
do once again with the size of the model and specifically with the depth of the
simulatedsoillayer.Imposingthebedrockmovementjust2.5metersbelowthepipe
doesnotallowthefaultpropagationtoevolvenaturally.Onthecontrary,thisway
the soil and the pipe are subjected to a more centralized and steep deformation
comparedtothenaturalone.Inaddition,a5metersurfacesoillayerwithanelastic
modulus of 25Mpa cannot be considered realistic. In order to overcome these
inaccuracies,toexaminetheeffectofthesoillayerdepthandatthesametimeto
limit the needed time for the analysis, we designed a two dimensional free field
model, subjected to normal and reverse fault rupture and then implemented its
resultstothe3Dbeammodel.

187

Thesoillayerdeptheffect

5.1 FreeField model

Model properties and geometry

The model we used is a simple 40m x 160m model consisted of square 1m x 1m


continuumCPE4elementsthatsubstitutethesoil.Onceagain,thehangingwallside
and one part of the base were moved to simulate a 60 fault rupture (normal and
reverse). The purpose of these analyses was to obtain the vertical and horizontal
nodedisplacementsat5mdepthandthentoimposethemtothebaseofthe3D,
5mdeephybridbeammodel(Figure5.1.2).
Regardingthesoilproperties,inthischapter,everythingbuttheYoungsmodulusE
distribution,remainedthesameasinthepreviouschapters.Fortheelasticmodulus
distribution along the depth, it was decided to examine various proposed
distribution models and compare them with a linear distribution model which has
been proved trustworthy by various researches conducted in our department.
Specifically, we compared the distributions of the linear model and of the sand
models of Yu and Richart (1984), Iwasaki et al. (1978), Zhou and Chen(2005) and
Hardin (1978). As depicted in Figure 5.1.3, the parabolic model that approaches
betterthelineardistributionisthatofYuandRichart.
Hence, the comparison was focused on these two models, in order to examine
whetherornottheparabolic(andmorerealistic)distributionyieldssimilarresultsto
thelinearone,whichhasalreadybeenprovedreliable.

The linear distribution is, approximately, a simple relation between the depth and
theelasticmodulus:

4

188

Thesoillayerdeptheffect

TheYuandRichartmodelforpuresandisbasedonthefollowingrelation:

G max AG F (e) ( ' 0) nG


where

AG 7000
nG 0.5

dependingonthesoiltype

'1 ' 2 ' 3

'1 z

' 2 ' 3 K 0 '1

'0

and

2.17 e
F (e)
1 e

Whereeistheporeindexequalto

s
1 0.286

It must be clarified that the Yu and Richart model was proposed for dynamic
problems, thus, the values we get from the above equations are very large to be
implied into our quasistatic problem. Hence, we derived the Gmax values with an
integer in order to obtain smaller values and to achieve relatively similar values to
those of the hybridbeam model of the previous chapters, in order to be able to
makevalidcomparisons.

189

Thesoillayerdeptheffect

The results of the comparison, both for normal and reverse fault rupture are
presented in Figures 5.1.4 5.1.7, in terms of surface displacements and surface
angulardistortion,foranimposedbedrockmovementupto3m.Itisquiteobvious
that the two different soil profiles yielded quite similar results. The final surface
displacementsarealmostidentical,whereasthesurfaceshearstrainslightlydiffers
betweenthetwoprofiles.Themaindifferenceisthattheparabolicprofileseemsto
resulttoquickerfaultpropagationtothesurfaceandtoaslightlymorecentralized
deformation.
However,thecomparisonsarecloseenoughtoallowtheuseoftheparabolicprofile
inour3Dmodel,withoutexpectinganypracticaldifferencescomparedtothelinear
one.

5.2 Normal Fault Rupture

The 3D hybrid model is used to examine the pipe behavior, when subjected to a
fault rupture that initiates 40m below the surface. The results we obtained are
demonstrated in Figures 5.2.1 5.2.8. It must be clarified, that 40 m model
correspondstothe3DHybridbeammodelof5mdepthwhichtakesintoaccounta
fault that initiates from the bedrock layer, 40 m below the surface, whereas 5 m
modelcorrespondstothepreviouslydescribedHybridbeammodelofChapter3.In
addition, we may refer to vertical fault displacement meaning the vertical
componentofthebedrockmovement.
Despitethedifferencesofthe40mandthe5mmodels,theresultsarequitesimilar.
Theaxialstraindevelopmentanddistributionisverymuchalikeandtheupperlimit
of 5% axial tensile strain is reached for the exact same values of bedrock vertical
displacement. As described in Chapter 3, a transition from prevailing bending to
prevailingtensionisagainspotted(Figure5.2.2).Thecritical,failureareasofthe

190

Thesoillayerdeptheffect

40mmodelareslightlydifferentandlocatedabout3maway,comparedtothe5m
model.Asexpectedfromtheaforementionedsimilarities,substantialdifferencesare
notspottedeitherintermsofaxialforces(Figure5.2.4).
Regardingtheeffectivepipelength,itappearsthatthelengthoftheactivatedpipeis
the same towards the hanging wall and the footwall side. Compared to the 5 m
modelofChapter3,theLeffispracticallythesameanddiffersonlyduringthefirst0.2
m of vertical bedrock movement, where the 40 m model demonstrates a much
smootherincreaseoftheLeff(Figure5.2.5).
Pipedetachmentfromthesoilwasalsoexaminedandispresentedbothintermsof
areasofcontactlossandintermsofrelativesoilpipeverticaldisplacement(Figure
5.2.6).Itisclearthat,detachmentoccurslaterinthecaseof40mmodel,butagain,
ittakesonlyasmallbedrockmovement(h=0.08m)fordetachmenttooccur.For
completenesssake,soilpressuresaroundthreepipecrosssectionsarepresentedin
Figure5.2.7.
Regardingthesubstantialdifferencesthatoccurasaresultofthesoillayersdepth
effect, the different soil displacement distribution is one of the most important
factors. As demonstrated in Figure 5.2.8, unlike the 5 m model, where the
displacements (thus the deformations) are concentrated along a relatively narrow
area,inthecaseof40mmodel,notonlythevaluesofdisplacementarelowerfor
thesamebedrockmovement,butalsothedistributionissignificantlylesssteep.This
is also depictured by the displacement diagrams of the same figure, where the
displacementsofthesoilrightbeneaththepipearecompared.
However,themostimportantdifference,fromapracticalpointofview,isthatthe
twomodelsreachthelimitof3%axialstrainfortotallydissimilarvaluesofvertical
bedrock movement h, with the exception of the case of D/t=72. The 40 m model
provokesalessintensebendingloading,thus,theincreaseofaxialstrainsbecauseof
bendingislessrapid.Hence,inthe5mmodel,the3%strainlimit,whichiscausedby
bending, is reached quicker, before the transition to prevailing tension. On the
contrary, in the 40m model, the axial strains caused by bending demonstrate a

191

Thesoillayerdeptheffect

milder increase, as a result of the wider displacement distribution, so that the


prevailingtensionappearsbeforethe3%strainlimitisreached.Afterthechangeof
loadingmechanismthereisarapidincreaseoftheaxialstrains,reachingthe3%and
5%limit,practicallyforthesameh.
This relatively limited difference, in terms of critical bedrock movement, has to do
with the nature of normal fault propagation. Because of its tensile nature, (thus
active conditions) the rupture propagates relatively rapidly up till the surface.
Anastasopoulosetal.(2007)haveshownthatfornormalfault,propagatingthrough
dense sand, it takes about (0.75% Z) of vertical bedrock movement for fault
outcropping,whereZ=thesoillayerdepth.Onthecontrary,ittakes3timesbigger
bedrock movement for a reverse fault to outcrop (2.5% Z). Indeed we observe
limited differences for the normal fault case whereas for the reverse fault, the
differencesaremuchgreaterasshowninthefollowingparagraph.

5.3 Reverse Fault Rupture

The effect of soil layers depth was also examined for the case of reverse fault
rupture.Thedifferencescomparedtothe5mmodelareobviousineveryaspectof
theobtainedresults.
Eventhoughlocalbucklingisonceagainthefailuremode,thestraindistributionand
developmentarequitedifferentinthiscase.InFigure5.3.2thestraindistribution,as
the fault movement increases, is demonstrated and it is obvious that the two
inflectionpointswehadinthe5mmodeldonotappearasquicklyandasclearly.
Furthermore,thesequenceoftheoccurrenceoflocalbucklingisoppositetothatof
5mmodelandnowthefirstbucklingappearsinthefootwallandthesecondonein
the hanging wall. This is due to the different displacement distribution the pipe is
subjectedtoasshowninFigure5.3.3.

192

Thesoillayerdeptheffect

Themostsignificantalterationthatoccurredconcernsthecriticalfaultdisplacement.
AsdemonstratedinFigure5.3.11,pipecanendureupto5times(dependingonthe
D/t ratio) bigger vertical fault movement. The critical axial compressive strains are
alsogreatlylarger(Figure5.3.5).Theaboveobservation,aswellasalltheobserved
differences, is the result of the totally different distribution of displacements, as
demonstrated in Figure 5.3.10, and of the passive conditions that occur during a
reverse fault, which cause slower rupture propagation. In simple words, since the
fault movement needed for fault outcropping is about 2.5% of Z (instead of only
0.75% as in normal fault) a deeper soil layer affects much more intensively the
results of the reverse fault case. The fault rupture needs a lot more of bedrock
displacementinordertoreachthepipeandingeneral,inordertoreachthesame
deformationstatewiththe5mmodel.
Concerningthepipedetachment,contrarytoallthepreviouscases,lossofcontact
between the pipe and the soil appears only after 0.60 m of vertical fault
displacement(Figure5.3.7).Again,forthesakeofcompleteness,theLeffandthesoil
pressures along certain pipe crosssection are demonstrated in Figures 5.3.6 and
5.3.8respectively.

193

Thesoillayerdeptheffect

Main conclusions of Chapter 5

We examined various parabolic E distributions in order to add realism to our


simulation. We concluded that the Yu and Richart distribution is the best among
thoseweexamined,sinceityieldsonlynegligibledifferencescomparedtothelinear
distribution,whichhasbeenusedvarioustimesfromourdepartmentandhasbeen
testedforitsreliability.
Asexpected,thegreatestalteration,betweenthismodelandthemodelofChapter
3, is the distribution of soil displacement. This is the cause of every observed
difference.
Thelowerdisplacementvaluesandthemoresmoothsoildisplacementdistribution
at the level of the pipe lead to larger pipe endurance in terms of vertical bedrock
displacement.

194

195

196







Figures of Chapter 5

197

198

5m

40m

Sand
Bedrock

Figure 5.1.1. Two normal fault ruptures, propagating through a surface soil layer of 5 m and
40 m respectively. In Chapter 5, the differences of those two models regarding the pipe
behavior are examined, both for normal and reverse faults.

199

(a)

40m

160m

1m

(b)

CPE4
Continuumplainstrain4nodeelement

1m

5m

(c)

Figure 5.1.2. (a) The 2D model consisted of square finite elements 1m x 1m.b) The deformed
2D free field model, used for obtaining the displacements during the development of the fault,
at level 5m below the surface. Those values where applied to the base of the 3D model (c).

200

E(MPa)
0

50

100

150

0
5
10

Linear
YuandRichart

15

Iwasakietal.
Z(m) 20

ZhouandChen
Hardin

25
30
35
40

YuandRichart
20

40

80

100

120

10

10

15

15

20

Z(m)

Z(m)

E(MPa)
60

Linear
20

40

E(MPa)
60
80 100

120

140

160

20

25

25

30

30

35

35

40

40

Figure 5.1.3. The distribution of Youngs modulus along the depth of our model, according to
various proposed models. Yu and Richarts relation was selected to be compared with the
linear distribution.

201

Normalfault

Linear

YuandRichart

Z
X
0

20

40

60

80

100

120

140

h=0.40m
0.5

h=0.80m
1

h=1.50m

Linear

h=2.00m

YuandRichart

Uz (m) 1.5
2
2.5

h=3.00m
3

X(m)

Figure 5.1.4. The deformed models are almost identical for Linear and Yu and Richart
distribution. This is also depicted by the comparison of the vertical displacements of the
surface of the two models (h = 3 m).

202

160

YuandRichart

(a)

70%
60%

50%

0.5m

40%

1.0m

30%

1.5m

20%

2.0m

10%

2.5m

0%

3.0m
60

65

70

75

80

85

90

(m)

Linear

70%
60%

50%

0.5m

40%

1.0m

30%

1.5m

20%

2.0m

10%

2.5m

0%
60

65

70

75

80

85

3.0m

90

X(m)

(b)

60%
50%
40%
YuandRichart

30%

Linear

20%
10%
0%
65

70

75

80

85

90

X(m)

Figure 5.1.5. (a) The development of angular distortion along the surface area where the
fault outcrop appears, for linear and Yu and Richart E distribution. (b) Comparison of the
angular distortion for vertical fault displacement h = 2 m.

203

Reversefault

Linear

YuandRichart

3.5

h=3.00m
3
2.5
2

Uz (m)

1.5

h=2.00m

Linear

h=1.50m

YuandRichart

1 h=0.80m
0.5 h=0.40m
0
40

60

80

100

120

140

160

X(m)

Figure 5.1.6. The deformed models of reverse fault for h = 3 m. The vertical displacements
demonstrate no significant differences between linear and parabolic distribution model.

204

Linear
(a)

35.00%
30.00%

25.00%

h=0.50m

20.00%

h=1.00m

15.00%

h=1.50m

10.00%

h=2.00m

5.00%

h=2.50m

0.00%

h=3.00m
60

70

80

90

100

X(m)

YuandRichart
35.00%
30.00%

25.00%

h=0.50m

20.00%

h=1.00m

15.00%

h=1.50m

10.00%

h=2.00m

5.00%

h=2.50m

0.00%

h=3.00m
60

70

80

90

100

X(m)

(b)

25.00%
20.00%
15.00%
Linear

10.00%

YuandRichart

5.00%
0.00%
60

70

80

90

100

X(m)

Figure 5.1.7. a) The development of angular distortion along the surface area where the fault
outcrop appears ,for linear and Yu and Richart E distribution. b) Comparison of the angular
distortion for vertical fault displacement h = 2 m.

205

Normalfaultrupture

(a)

(b)
2
1.8
1.6
1.4
1.2

hcr/D

3%

0.8

5%

0.6
0.4
0.2
0
0

24

48

72

96

120

144

168

D/t

Figure 5.2.1.. (a) The deformed soil and pipe. (b) The critical vertical fault displacement to
diameter ratio (hcr/D) with respect to the diameter to thickness ratio (D/t).

206

h=0.50m

h=1.00m

h=1.50m

h=1.65m

h=1.67m

X
Figure 5.2.2. The axial strain distribution along the critical pipe area (X=2737 m) for five fault
displacements h. (D/t = 72)

207

(a)

7.00%
6.00%
5.00%
h=0.50m
4.00%

h=1.00m
h=1.50m

3.00%

h=1.65m
h=1.67m

2.00%
1.00%
0.00%
27

(b)

29

31

33

35

37

X(m)

7.00%
6.00%
5.00%

D/t=72

4.00%

D/t=144
3.00%

D/t=120

2.00%

D/t=96

1.00%
0.00%

(c)

27

29

31

6.00%

X(m)

33

35

5.00%
4.00%

D/t=144

x 3.00%

D/t=120
D/t=96

2.00%

D/t=72

1.00%
0.00%
0

0.5

h/D

1.5

Figure 5.2.3. (a) The axial strain distribution along the critical area for D/t = 72. (b)
Comparison of axial strain distribution, when critical state is reached (x=5%), for four D/t
ratios. (c) Comparison of axial strain development at the points where the limit strain is
exceeded, with respect to the vertical fault offset for D/t=72,96,120 and 144.

208

(a)

Topfiber

7000
6000
5000

h=0.10m
4000

h=0.25m
h=0.50m

N(kN) 3000

h=1.00m

2000

h=1.50m
h=1.75m

1000
0
0

10

20

1000

(b)

30

40

50

60

X(m)

Bottomfiber

7000
6000
5000

h=0.10m

4000

h=0.25m

3000

h=0.50m

N(kN)
2000

h=1.00m

1000

h=1.50m
h=1.75m

0
0

10

20

30

40

50

60

1000
2000

X(m)

Figure 5.2.4. Axial force distribution and development for six vertical fault displacements for
(a) the top pipe fiber (b) the bottom fiber (D/t = 72)

209

16000

(a)

14000
12000
10000

Footwall

N(kN) 8000

Hangingwall

6000
4000
2000
0
30

130

230

330

430

530

630

730

830

930

1030

|X|(m)

(b)
350
300
250
200

Leff/D

150

40mmodel
100

5mmodel
50
0
0

0.2

0.4

0.6

0.8

1.2

h/D

Figure 5.2.5. (a) Axial beam force in the hanging wall and the footwall. No significant
difference is spotted. In this diagram, X=0 is considered at the middle of the model, where the
fault crosses the pipe, approximately. (b) The activated pipe length with respect to the vertical
fault offset (D/t = 144).

210

0.02
0
0.02

20

40

60

Uz (m) 0.04

h=0.10m

0.06
0.08
0.1

X(m)

0.1
0
0.1

20

40

60

0.2

Uz (m)

h=0.50m

0.3
0.4
0.5
0.6

X(m)

0.2
0
0.2

Uz (m)

20

40

60

0.4

h=1.00m

0.6
0.8
1
1.2

X(m)

Figure 5.2.6. Pipe detachment areas from the surrounding soil are defined by the blue colored
areas. The detachment of the pipe bottom from the soil beneath it, is demonstrated by the
diagrams of vertical displacements of the pipe bottom and the soil for three values of vertical
bedrock displacement.(D/t = 72)

211

C
h=0.10m
C

h=0.50m
A

h=1.00m
A

Figure 5.2.7. Soil pressures along the pipeline and soil pressures around three pipe cross
sections A,B and C (X=20,30 and 40 m respectively) for three vertical fault offsets h.

212

(a)
5mmodel

40mmodel

(b)
h=0.10m

0.02
0
0.02

Uz(m)

0
0

20

40

60

0.2

0.04

20

40

60

0.4

Uz(m)

0.06
0.08

0.6
0.8

0.1
0.12

h=1.00m

0.2

X(m)

40mmodel

1.2

X(m)

5mmodel

Figure 5.2.8. (a) Displacement distribution for the limited 5 m depth model and the model that
takes into account a soil layer of 40 m depth. (b) Comparison of soil vertical displacements
beneath the pipe bottom between the 5 m and 40 m model.

213

Reversefaultrupture
(a)

(b)

0.9
0.8
0.7
0.6
0.5

hcr/D

0.4
0.3
0.2
0.1
0
0

24

48

72

96

120

144

168

D/t

Figure 5.3.1. (a) Plastic strain distribution along the deformed soil and pipe for the reverse
fault case. (b) The critical vertical fault displacement to diameter ratio (hcr/D) with respect to
the diameter to thickness ratio (D/t).

214

h=0.72m

h=0.60m

h=0.50m

h=0.90m

2nd

1st

Z
X
Figure 5.3.2. The axial strain distribution along the critical pipe area (X=1340 m) for four fault
displacements h. (D/t = 72). First buckling occurs in the footwall .

215

(a)

30

31

32

33

34

35

0.00%
0.20%
0.40%
h=0.50m

0.60%

h=0.60m
0.80%

h=0.68m

1.00%

h=0.72m

1.20%
1.40%
1.60%
1.80%
X(m)

(b)

21

21.5

22

22.5

23

23.5

24

0.00%
0.50%
1.00%

h=0.70m

1.50%

h=0.75m

x 2.00%

h=0.80m

2.50%
3.00%
3.50%
4.00%

X(m)

Figure 5.3.3. The distribution of axial strains until (a) the 1st appearance (h = 0.72 m) and (b)
the 2nd appearance of buckling (h = 0.80 m). (D/t = 72)

216

(a)

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.00%
0.10%
0.20%
D/t=144
0.30%

D/t=120
D/t=96

0.40%

D/t=72
0.50%
0.60%
0.70%
0.80%

h/D

(b)

24

48

72

96

120

144

168

0.00%
0.10%
0.20%
0.30%

xcritical

0.40%
0.50%
0.60%
0.70%
0.80%
0.90%

D/t

Figure 5.3.4. (a) Axial strain development at the buckling points with respect to the h/D ratio,
for D/t=72144. (b) The critical axial strain values , beyond which buckling occurs., for all the
four values of D/t.

217

Topfiber
1000

500

1000

N(kN)

20

40

500 0

60

20

40

60

1000

2000

1500

3000

N(kN) 2000
2500

4000

h=0.05m

5000

h=0.20m

6000

h=0.50m

7000

h=0.60m

X(m)

1st buckling

3000
3500
4000

h=0.75m

4500

h=0.85m

X(m)

h=0.70m

Bottomfiber
1000

1000

0
0

20

40

60

1000

1000

20
2nd

40

60

buckling

2000

2000

N(kN) 3000

N(kN)
3000

4000
h=0.05m

4000

h=0.20m
5000

h=0.50m
h=0.60m

6000
X(m)

h=0.70m

h=0.75m

5000

h=0.85m

6000
7000
X(m)

Figure 5.3.5. The axial stain along the bottom and top pipe fiber before and after the
occurrence of buckling.

218

(a)

30

130

230

330

430

530

630

730

830

930

1030

0
1000
2000
h=0.05m
3000

h=0.10m
h=0.40m

N(kN) 4000

h=0.50m
h=0.75m

5000

h=0.85m

6000
7000
X(m)

8000

(b)

400
350
300
250

Leff/D 200
150
100
50
0
0

0.2

0.4

0.6

0.8

1.2

h/D

Figure 5.3.6. (a) The axial beam forces in the footwall, demonstrating the activated pipe
length with respect to the vertical bedrock displacement. (b) The effective pipe length to
diameter ratio with respect to the h/D ratio (D/t = 72).

219

0.12
0.1
0.08

Uz (m)

h=0.60m

0.06
0.04
0.02
0
29

31

33
X(m)

Uz (m)

0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0

h=0.80m

20

25

30

35

X(m)

0.6
0.5

Soil

0.4

Uz (m)

Pipebottom

h=0.90m

0.3
0.2
0.1
0
20

25
X(m)

30

Figure5.3.7.Detachmentareasaredefinedbythebluecoloredareas.Thedetachmentofthe
pipebottomfromthesoilbeneathit,isdemonstratedbythediagramsofvertical
displacementsofthepipebottomandthesoilforthreevaluesofverticalbedrock
displacement.(D/t=72)

220

h=0.60m

h=0.80m

h=0.90m

Figure 5.3.8. Soil pressures along the pipeline and soil pressures around three pipe cross
sections A,B and C (X=20,30 and 40 m respectively) for three vertical fault offsets h.

221

(a)
5mmodel

40mmodel

(b)
h=0.10m

0.12

h=0.40m

0.1
0.08
0.06

Uz (m)

0.04
0.02
0
0
0.02

20

40
X(m)

0.45
0.4
0.35
0.3
0.25
Uz (m) 0.2
0.15
0.1
0.05
0
60
0.05 0

40mmodel

20

40

60

X(m)

5mmodel
Figure 5.3.9. (a) Displacement distribution for the limited 5 m depth model and the model that
takes into account a soil layer of 40 m depth. (b) Comparison of soil vertical displacements
beneath the pipe bottom between the 5 m and 40 m model.

222

(a)

2
1.8
1.6
1.4
3%5mmodel

1.2

5%5mmodel

hcr/D

3%40mmodel

0.8

5%40mmodel

0.6
0.4
0.2
0
0

(b)

24

48

72

h/D

96

120

144

168

0.9
0.8
0.7
0.6
0.5

hcr/D

40mmodel

0.4

5mmodel
0.3
0.2
0.1
0
48

72

96

120

144

168

D/t

Figure 5.3.10. (a) Comparison of the failure points in terms of hcr/D between the 5 m and 40 m
model, for the normal fault case. (b) Comparison of the failure points in terms of hcr/D
between the 5 m and 40 m model, for the reverse fault case.

223

224





Chapter 6
Pipe steel with hardening behavior

225

226

Pipesteelwithhardeningbehavior

Introduction

In all the previous chapters, the models we examined were based on an elastic
perfectly plastic behavior of the steel pipe material, as described in Chapter 2.
However,theoriginalstressstrainbehavior,asofficiallygivenbymanufacturersand
institutes (e.g. API), is consisted of a limited elastic brunch, followed by a short
plateauwhichissucceededbyanascendingpart,untiltheultimatestressof550MPa
is reached, corresponding to approximately 20% of axial strain (Figure 6.1.1). This
chapteraimsattheinvestigationoftheeffectthattheaforementionedchangecould
have,concerningthepipeenduranceandbehavior.

6.1 Normal Fault Rupture

The model we use is the HybridBeam model, keeping the same properties as of those
describedinChapters2and3,withtheexceptionofthehardeningpipebehavior.
The results of the normal fault rupture model can be summarized by the graph given in
Figure6.1.1.Asitisdemonstrated,applyinghardeningbehaviortothepipematerialleads
to a quite different outcome. As we would have expected, hardening grants the pipe the
capability of developing bigger stresses, caused by the strain increase, leading to bigger
capacityintermsofcriticalfaultdisplacement.
Whatisevenmoreinterestingisthat,aswecannoticeatthesamefigure,another,different
modeoffailureappearsbetweenthestrainlimitstatesof3%and5%ofaxialtensilestrain.
This operational failure occurs due to excessive crosssectional distortion, also known as
crosssectionflatteningorovalization, asdescribed inChapter2.Figure6.1.3presents the
imageofsuchafailureandthedevelopmentofthecrosssectionalflattening.
Hardeningresultsinadifferentredistributionofstressesandstrains,aspresentedinFigures
6.1.2and6.1.4,andthisisthecauseforallthenoticeddifferences.Similarlytotheprevious
model of Chapter 3, above a certain fault displacement, the prevailing loading mode
switchesfrombendingtoamoreshearlikeloading,accompaniedbyasmalltranspositionof
thecriticalarea.However,theformoftheaxialstraindistributionissignificantlydifferent.

227

Pipesteelwithhardeningbehavior

Regardingthelengthofthepipethatgetsactivated,resultsdonotdemonstrateanygreat
differences. Figure 6.1.6 proves the aforementioned observation since, the applied
hardeningbehaviorhasonlyaslighteffect,causingsmallerLeff.

6.2 Reverse Fault Rupture

Once again, the HybridBeam model of Chapter 3 is used, applying the hardening stress
strainrelationofthesteel.
As depictured in Figure 6.2.1 not only the mode of failure remains the same but also the
criticalfaultdisplacementisidenticaltothatofthecaseofelasticperfectlyplasticmaterial.
Thisisquitelogical,sincelocalbucklingisrelatedtothestabilityandcapacityofawiderarea
andnotonlytothestrengthofacriticalcrosssection.
Thesequenceofbucklingappearancesisnotalteredsinceitisonlyaffectedbythewaywe
impose the soil displacements. The shape of strain distribution is also slightly altered,
especially as regards the 2nd buckling area (Figure 6.2.3) where the strain distribution is
morewrinkled,declaringthedifferentstressandstrainredistributionmode.
Regarding the Leff, hardening does not seem to affect the activated pipe length at all.
Similarly to Chapter 4, the footwall beam part of the model gets more stressed and is the
one on which we are based in order to compute the Leff. Figure 6.2.5 shows that the
evolutionoftheLeffforthereversefaultruptureisalmostidenticaltothatofChapter4.

228

Pipesteelwithhardeningbehavior

Conclusions of Chapter 6

Probably the most important conclusion made in this chapter is that the
application of the original hardening behavior of the pipe material is necessary in
ordertocapturethecrosssectionalfailuremodeofovalization.Ovalizationappears
onlyinthenormalfaultcasewheretensilestressprevails.
It has been proved that in the case of normal fault rupture, hardening results in
bigger capacity in terms of fault displacement and thus we can claim that is more
conservative to ignore it and to apply an elasticperfectly plastic stressstrain
relation.However,theaxialstrainlimitsof3%and5%arealreadyconservativeand
taking also in consideration that the elasticplastic model fails to capture the
ovalizationfailuremode,weproposethathardeningshouldbeintegratedinmodel.
Concerning the affected pipe length the importance of hardening is not of great
importanceespeciallyinthereversefaultcase.
Hardening, obviously, does not impose any direct alteration to the pipesoil
interactionapartfromthelimitedchangesinsoilpressuresanddetachmentthatare
causedbythedifferentdeformedshapeofthepipe.
Inthecaseofreversefaultrupture,wherelocalbucklingisthemodeoffailure,no
actual differences are spotted and the critical fault displacements are identical to
thoseoftheelasticperfectlyplasticcase.

229

230







Figures of Chapter 6

231

232

NormalFaultRupture
(a)

600
550
500
450
400
350
Stress
300
(MPa)_
250
200
150
100
50
0

600
550
500
450
400
350
Stress
300
(MPa)
250
200
150
100
50
0
0%

5%

0%

10%

5%

Strain

(b)

10%
Strain

15%

2
1.8
1.6
1.4
1.2
hcr/D

3%

0.8

5%

0.6

Ovalization

0.4
0.2
0
0

24

48

72

96

120

144

168

D/t

Figure 6.1.1. (a) The deformed model due to normal fault rupture. In this chapter, instead of
ElasticPerfectly Plastic stressstrain relation, we applied the original ElasticHardening Plastic
behavior of the pipe steel. (b) The summarized dimensionless results demonstrate the critical
vertical fault displacement with respect to the D/t ratio, for the axial strain limits of 3% and
5%, as well as, for the failure due to ovalization.

233

20%

h=0.10m

h=0.50m

h=1.10m

h=1.40m

Figure 6.1.2. The axial strain distribution along the critical pipe area (X=2040 m) for four
vertical fault displacements h. (D/t = 72).

234

(a)

z
x

(b)

h=0

h=0.2m

h=0.5m

f=

h=1m

h=1.45m

Do D '
0.15
Do

Figure 6.1.3. (a) Excessive crosssectional pipe distortion (ovalization, flattening). b) The
evolution of the ovalization with respect to the vertical fault displacement. (D/t = 120)

235

(a)

6%
5%
h=0.25m

4%

h=0.50m

x 3%

h=1.00m

2%

h=1.10m
h=1.20m

1%

h=1.40m
0%
31.2

(b)

32.2

X1

6%

33.2
X (m)

X234.2

35.2

5%
4%

x 3%

X2

2%

X1

1%
0%
0

(c)

0.5

1.5

h/D
3.5%
3.0%
2.5%
D/t=72

2.0%

D/t=96
1.5%

D/t=120

1.0%

D/t=144

0.5%
0.0%
31.2

32.2

33.2

34.2

35.2

X (m)

Figure 6.1.4. (a) The axial strain distribution along the critical area. We observe that the point
of critical strain is transposed from the X1 position to the X2 (D/t = 120). (b) The axial strain
development at points X1 and X2. (c) The axial strain distribution at the point when the 3%
strain limit is reached, for four D/t ratios.

236

Topfiber
4000

3000

2000
h=0.10m
h=0.50m

N(kN) 1000

h=1.00m
h=1.40m

0
0

10

20

30

40

50

60

1000

2000

X(m)

Bottomfiber
4000

3000
h=0.10m

2000

h=0.50m
h=1.00m

N(kN) 1000

h=1.40m
0
0

10

20

30

40

50

60

1000

2000

X(m)

Figure 6.1.5. The axial force distribution along the top and bottom fiber of the pipe.
(D/t = 120)

237

16000

(a)

14000
12000
10000

h=0.10m

8000

h=0.50m

6000

h=1.00m

4000

h=1.50m

2000

h=1.95m

N(kN)

0
30

230

630

830

1030

|X|(m)

700

(b)

430

600
500
400

Leff/D

Hardening

300

Elasticplastic
200
100
0

(c)

0.5

600

1.5

h/D

500
400

h=0.25m

Leff/D 300

h=0.50m
h=1.00m

200

h=1.40m
100
0
0

50

100

150

200

D/t

Figure 6.1.6. (a) The axial beam force development in the footwall. (D/t = 72). (b) The effective
length of the pipe with respect to the h/D ratio, compared to the ElasticPerfectly Plastic case.
(c) The effective length to diameter ratio with respect to the D/t ratio for four vertical fault
displacements.

238

ReverseFaultRupture
(a)

(b)
0.3
0.25
0.2

hcr/D 0.15
0.1
0.05
0
0

24

48

72

96

120

144

168

D/t

Figure 6.2.1. (a) The deformed model due to reverse fault rupture, with hardening pipe steel.
(b) The summarized dimensionless results that demonstrate the critical vertical fault
displacement, at which buckling occurs, with respect to the D/t ratio.

239

h =0.10m

h =0.20m

h =0.30m

1st

2nd

h =0.50m

Figure 6.2.2. The axial strains along the critical area for several fault displacement values. The
buckling occurs firstly in the hanging wall and afterwards in the footwall.(D/t = 72)

240

(a)

26

27

28

29

30

31

32

0.00%

0.50%

1.00%
h=0.10m

x 1.50%

h=0.20m
h=0.24m

2.00%

2.50%

3.00%

(b)

X(m)

34

35

36

37

38

39

40

0.00%
0.20%
0.40%
0.60%

h=0.10m
h=0.20m

x 0.80%

h=0.30m

1.00%

h=0.34m

1.20%
1.40%
1.60%

X(m)

Figure 6.2.3. The axial strains until the occurrence of (a) the first buckling, that occurs at the
bottom pipe fiber, in the hanging wall and (b) the second buckling at the top pipe fiber,
located in the footwall, (D/t = 72).

241

(a)

0.00%
0

0.05

0.1

0.15

0.2

0.05%

0.10%
D/t=72
D/t=96

x 0.15%

D/t=120
D/t=144

0.20%

0.25%

0.30%

(b)

h(m)

24

48

72

96

120

144

168

0.00%
0.20%
0.40%
0.60%

xcritical 0.80%
1.00%
1.20%
1.40%
1.60%

D/t

Figure 6.2.4. (a) The axial strain development at the critical pipe where buckling initiates, with
respect to the vertical fault displacement for four D/t ratios. (b) The critical axial strain right
before the occurrence of buckling with respect to the D/t ratio.

242

(a)

30

130

230

330

430

0
1000
2000
3000

Hangingwall

N(kN)
4000

Footwall

5000
6000
7000

(b)

|X|(m)

400
350
300
250

Hardening

Leff/D 200

Elasticplastic

150
100
50
0
0

0.1

0.2

0.3

0.4

0.5

h/D

Figure 6.2.5. (a) The axial strain development at the critical pipe where buckling initiates, with
respect to the vertical fault displacement for the four D/t ratios. (b) The critical axial strain
right before the occurrence of buckling with respect to the D/t ratio.

243

Summary
NormalFault

(a)
2
1.8
1.6
1.4

3%HardeningModel

1.2

hcr/D

5%Hardeningmodel

OvalizationHardeningModel

0.8

3%ElasticPlasticModel

0.6

5%elasticPlasticModel

0.4
0.2
0
0

24

48

72

(b)

96

120

144

168

D/t

ReverseFault
0.3
0.25
0.2

ElasticPlasticModel
HardeningModel

hcr/D 0.15
0.1
0.05
0
0

24

48

72

96

120

144

168

D/t

Figure 6.2.6. a) Comparison of the critical fault displacements between the ElasticPerfectly
Plastic model of the previous chapters and the Hardening model of this chapter, concerning
the normal fault rupture. Significant differences are spotted, both in terms of hcr and in terms
of failure mode. b) The same comparison for the reverse fault case. As demonstrated, the
effect of the hardening behavior of the pipe material has no significant effects.

244

245

246





Chapter 7
Internal pipe pressure

247

248

Internalpipepressure

Introduction

In the following chapter, the effect of internal pipe pressure is examined. As


mentionedinthefirstchapters,thepipelinesweexaminearecommonlyusedinoil
andgastransportation.Obviously,gasandoilhavetobeunderpressureinorderto
move along vast distances and variable topographies. This means that the pipe,
duringitsoperationalperiod,issubjectedtoacontinuousinternalpressure.Theaim
ofthischapteristoinvestigatewhataretheeffectscausedbytheinternalpressure
applicationandespecially,toclarify,whetherornot,theinternalpressureincreases
thepipecapacity(intermsoffaultdisplacementmagnitude)andasresult,whether
or not, it should be taken into consideration when studying the pipe resistance to
dipslipfaultruptures.

Internal Pressure

The maximum internal pressure value is given by the following equation, which,
considering a factor of safety equal to 0.72, as proposed by the ASME code for
pressurepiping,isgivenas:

p max 0.72 2 y
D

However,similarlytothecommon,realoperationalpressurevalues,weappliedonly
a percentage of the maximum pressure allowed, equal to approximately 40% Pmax.
Since,themaximumpressuredependsontheD/tratiothepressureappliedforeach
oneoftheD/tcasesisgivenbythefollowingtable:

D/t

144

120

96

72

Papplied(kPa)

1700

2040

2550

3400

249

Internalpipepressure

7.1 Normal Fault Rupture

Once again, we make use of the HybridBeam model. The material of the pipe is,
initially, chosen to be elasticperfectly plastic, whereas in one of the following
paragraphs,thecombinationofinternalpressureandhardeningmaterialbehavioris
examined.
Theresultsobtainedarequitedifferentcomparedtothecaseofabsenceofinternal
pressure. Figure 7.1.1 demonstrates the pipe capacity in terms of critical fault
displacement and indicates that the presence of internal pressure has an adverse
effectconcerningthepiperesistancetofaultmovement.Specifically,forthe5%axial
tensile strain limit, the pipe resistance is reduced approximately by 35% (Figure
7.2.6). This can be simply explained by the fact that the internal pressure causes
radial hoop dilation which then provokes additional axial tensile stress and strain,
thus,thetensilestrainlimitisreachedearlier.
In Figure 7.1.2, the axial strain distribution along the critical area is displayed for
severalfault movementmagnitudes.Combiningthisimagewiththeinformationof
Figure 7.1.3, we observe that the concentration of axial tensile strains is now
spotted in several points along the critical area. Especially as the pipe thickness
increases,hencethepipeissubjectedtobiggerfaultmovementbeforefailure,the
critical point is displaced to another, adjacentcrosssection. However, dissimilar to
the model of Chapter 3, we do not observe the aforementioned transition from
prevailingbendingtoprevailingshearingloadingmode.
Regarding the effective pipe length, similarly to the previous chapters, we firstly
examinewhichsideismoreintensivelystressed,theoneinthefootwalloreinthe
hanging wall. Once again, no significant difference is observed, as shown in Figure
7.1.5. In the same figure, the effect of D/t ratio is also demonstrated and proves,
thathighermomentofinertiasignifieslongeractivepipelength.Finally,itisproved
thattheevolutionofLeffisabsolutelysimilartothecasewithoutinternalpressure,
with the exception that now the Leff does not seem to reach a maximum value. In

250

Internalpipepressure

general,mostoftheanalysesstopaftertheoccurrenceoffailure,duetoexcessive
distortion and inability of numerical convergence. Thus, it is logical for the model
withinternalpressuretostopearlier.However,wewouldexpectthatsincefailureis
reached,theLeffshouldstabilizeafterafewadditionalstepsoffault displacement.
Thereasonthatthisdoesnothappenistheaforementioneddifference,that,now,
bending remains the prevailing loading mechanism and there is not such a great
strain concentration and strain increase according to a shearlike mode of
deformation,asinthecaseofChapter3.

7.2 Reverse Fault Rupture

Theeffectofinternalpressureisalsoinvestigatedforareversefaultrupture.
The differences are not as significant as in the normal fault case. The summarized
outcomeoftheanalysisisdemonstratedinFigure7.2.6,wherethepiperesistance
tofaultmovementisdepictedintermsofcriticalfaultdisplacementwithrespectto
the D/t ratio, both with and without the internal pressure application. It is proven
that internal pipe pressure leads to an earlier failure, even though the capacity
decrease is only about 17%. This is attributed to the development of additional
stressesandstrainsinthepipelinewallthatcauseearlyyieldingandprematurelocal
buckling formation. Figures 7.2.2 and 7.2.3 describe the axial strain development
and distribution and show no significant alterations with the case of absence of
pressure.
Finally, concerning the Leff, things are almost identical. The pipe in the footwall is,
again, more stressed than in the hanging wall and compared to the case without
pipepressure,nothingbutaslightincreaseofLeffisdepicted(Figure7.2.5).

251

Internalpipepressure

7.3 Internal Pipe Pressure with


Hardening Pipe Steel Behavior

In this paragraph, we investigate the results of a model that combines the


application of internal pressure with elastichardening plastic stressstrain relation
forthepipematerial.

Normal fault rupture

The results are presented in Figures 7.3.1 and 7.3.2. As demonstrated, the use of
hardening plasticity intensifies the strain distribution behavior that was observed
withtheapplicationofinternalpressure.Thisstraindistributionbehaviorisdefined,
in the previous paragraphs, as the redistribution of stress and strains in several
adjacentcrosssectionsandthus,thedecreaseofexcessivestrainconcentrationata
singlepoint.ItisclearlydemonstratedinFigure7.3.2that,asthefaultdisplacement
augments and the strains increase, strains concentrate in several positions and
hinderthemaximumstrainvaluefromreachingbiggermagnitudes.
In the same figure, the (b) diagram compares the axial strain development at the
criticalpoints,forallthefourexaminedcombinations:
1) ElasticPerfectlyPlastic/AbsenceofInternalPipePressure
2) ElasticHardeningPlasticity/AbsenceofInternalPipePressure
3) ElasticPerfectlyPlasticwithInternalPipePressure
4) ElasticHardeningPlasticitywithInternalPipePressure
Ingeneraltheeffectsofhardeningappear,logically,aftertheoccurrenceofcertain
axial strain. Hardening prevents the almost vertical increase of axial strains that is
observedinthecasesofelasticperfectlyplasticmaterialaftertheplastification.

252

Internalpipepressure

The effect of internal pressure is clearly amplified when combined with hardening
material and leads to a constant redistribution of strains that retain the maximum
strainmagnitudetorelativelowvalues.

Reverse fault rupture

The results are demonstrated in Figures 7.3.3 and 7.3.4. The differences between
the four cases are not great. However, it is clear that, in general, the presence of
internal pressure reduces the pipe strength, whereas the effect of hardening is
slightly beneficial but can be practically spotted only when combined with the
applicationofinternalpressure.

253

Internalpipepressure

Conclusions of Chapter 7

Theapplicationofinternalpressurehasasignificanteffect,inthecaseofaburied
steel pipeline subjected to normal fault rupture, since it causes premature failure
(about 35% less capacity in terms of fault movement when 0.40% Pmax is applied)
compared to the case without pipe pressure. The strain distribution is also altered
andisdescribedbystrainredistributionsandstrainconcentrationsinseveralpoints.
For the reverse fault rupture case, the presence of pipe pressure reduces once
againthepipecapacity,althoughbyasmallerpercentageofapproximately17%.The
straindistributionisnotsignificantlychanged.
TheLeff doesnotseemtobepracticallyaffectedbytheactionofinternalpressure
bothinthenormalandinthereversecase.
The combination of hardening with internal pressure has as a result the
intensificationofinternalpressureeffects,concerningtheaxialstraindistributionin
thecaseofnormalrupture,whereas,inthereversecase,theapplicationofinternal
pipepressureisnecessaryfortheeffectofhardeningtoappear.
Basedontheaboveconclusions,amodelthattakesintoconsiderationtheinternal
pipepressureisproposed.

254

255

256







Figures of Chapter 7

257

258

(a)

p=0.4pMAX

(b)
1.2

0.8

h/D 0.6

3%
5%

0.4

0.2

0
0

20

40

60

80

100

120

140

160

D/t

Figure 7.1.1. (a) Internal pipe pressure was applied equal to 40% of the value of pmax. Pmax is
different for every D/t ratio. (b) The capacity of the pipes for the normal fault case, in terms of
maximum fault displacement, with respect to the D/t ratio.

259

h =0.10m

h =0.40m

h =0.80m

h =1.10m

Figure 7.1.2. The axial strain distribution along the critical area (X=1638 m) for different fault
displacements.(D/t = 72)

260

8.00%

(a)

7.00%
6.00%
h=0.10m

5.00%

h=0.40m

x 4.00%

h=0.80m

3.00%

h=1.03m
h=1.10m

2.00%
1.00%
0.00%

(b)

8.00% 33

34

35

X(m)

36

37

38

7.00%
6.00%
5.00%

D/t=72

x 4.00%

D/t=96
D/t=120

3.00%

D/t=144

2.00%
1.00%
0.00%
33

(c)

34

35

X(m)

36

37

38

20%
18%
16%
14%
12%
10%
8%
6%
4%
2%
0%

D/t=72
D/t=96
D/t=120
D/t=144

0.5

1.5

h/D

Figure 7.1.3. (a) The distribution development of the axial strains along the critical area,
(D/t = 72). (b) The shape of axial strains distribution, at the point where the 5% limit is
reached, for four D/t ratios. (c) The evolution of axial strain at the critical pipe points with
respect to the vertical fault displacement to diameter ratio, for four D/r ratios.

261

Topfiber
7000
6000
5000
4000
h=0.10m

3000
N(kN)

h=0.40m

2000

h=0.80m

1000

h=1.10m

0
1000

10

20

2000

30

40

50

60

X(m)

Bottomfiber
8000
6000
4000
h=0.10m
h=0.40m

N(kN) 2000

h=0.80m
0

h=1.10m
0

10

20

30

40

50

60

2000
4000

X(m)

Figure 7.1.4. The axial force distribution along the top and bottom pipe fiber, for different
vertical fault displacements.(D/t = 72)

262

14000

(a)

12000
10000
8000

Footwall

6000

Hangingwall

N(kN)

4000
2000
0
700

(b)

30

230

430

630

830

1030

|X|(m)

600
500
400
300

WithoutInternal
Pressure

200

WithInternal
Pressure

Leff/D

100
0
0

(c)

0.5

1.5

2.5

h/D

450
400
350
300
Leff/D

250

h=0.10m

200

h=0.20m

150

h=0.60m

100
50
0
0

24

48

72

96

120

144

168

D/t

Figure 7.1.5. (a) Comparison of the axial forces, for three fault displacements, along the beam
parts of the model, in the footwall and the hanging wall. (b) Comparison of the Leff between
the case without internal pressure and the case where internal pipe pressure is applied. (D/t =
72) (c) Leff to Diameter ratio with respect to the D/t ratio for three fault displacements.

263

ReverseFaultRupture
(a)

p=0.4pMAX

(b)
0.3
0.25
0.2

hcr/D 0.15
0.1
0.05
0
0

24

48

72

96

120

144

168

D/t

Figure 7.2.1. (a) The internal pipe pressure effect was examined for the reverse fault case as
well. (b) The capacity of pipes of different thicknesses (D=1m) in terms of critical fault
displacement.

264

h =0.10m

h =0.20m

h =0.30m

1st

2nd

h =0.40m

Figure 7.2.2. The axial strain distribution along the critical area (X=2342 m) for different fault
displacements. The buckling areas and the sequence of buckling appearance. (D/t = 72)

265

(a)
28

28.5

29

29.5

30

0.00%
0.50%
1.00%

h=0.17m
h=0.21m

1.50%

h=0.22m

h=0.23m

2.00%

h=0.24m
2.50%
3.00%
X(m)

3.50%

(b)

0.1

0.2

0.3

0.4

0.5

0.00%
0.50%
1.00%
D/t=72

1.50%

D/t=96

x 2.00%

D/t=120
2.50%

D/t=144

3.00%
3.50%
4.00%
h(m)

Figure 7.2.3. (a) The distribution of axial compressive strains along the critical buckling area
until the appearance of buckling.(D/t = 72). (b) The evolution of axial strains at the critical
buckling points for the four D/t ratios.

266

BottomFiber

(a)

3000
2000
1000
0
0

10

20

30

40

50

60

1000

h=0m
h=0.10m

N(kN)

h=0.20m

2000

h=0.30m
3000

h=0.40m

4000
5000
6000

X(m)

TopFiber

(b)

3000
2000
1000
0
0

10

20

30

1000

40

50

60

h=0m
h=0.10m

N(kN)

h=0.20m

2000

h=0.30m
3000

h=0.40m

4000
5000
6000

X(m)

Figure 7.2.4. The axial force distribution along (a) the bottom pipe fiber and (b) the top pipe
fiber. Before the initiation of fault movement (h = 0m), the application of internal pressure
subjects the pipe to tensile stress.

267

(a)

30

230

430

630

830

1030

0
1000
2000

Footwall
N(kN) 3000

Hangingwall

4000
5000
6000

(b)

|X|(m)

400
350
300
250
Leff/D 200

WithoutInternallPressure
WthInternalPressure

150
100
50
0
0

0.1

0.2

0.3

0.4

0.5

h/D

Figure 7.2.5. (a) Comparison of the axial beam forces, for two fault displacements, in the
footwall and in the hanging wall . (b) Comparison of the Leff with respect to the fault
displacement for the case with and without internal pipe pressure.

268

(a)

Normalfaultrupture

2
1.8

WithInternalPressure

1.6

WithoutInternalPressure

1.4
1.2

hcr/D 1
0.8
0.6
0.4
0.2
0
48

72

96

120

144

168

D/t

Reversefaultrupture
(b)

0.3

WithInternalPressure
0.25

WithoutInternalPressure
0.2

hcr/D 0.15
0.1
0.05
0
48

72

96

120

144

168

D/t

Figure 7.2.6. (a) Comparison of the pipe capacity, in terms of critical fault displacement, with
respect to the Diameter to thickness ratio, between the case with and without internal pipe
pressure, for normal fault rupture. (b) Comparison of the pipe capacity, in terms of critical
fault displacement, with respect to the Diameter to thickness ratio, between the case with and
without internal pipe pressure, for reverse fault rupture.

269

HardeningwithInternalPressure NormalFault

h =0.50m

h =1.00m

h =1.36m

h =1.80m

Figure 7.3.1. The axial strain distribution along the central model area (X=1040 m) for
different fault displacements.(D/t = 72)

270

(a)

3.50%
3.00%
2.50%
h=0.50m

2.00%

h=0.75m

h=1.00m

1.50%

h=1.50m
1.00%

h=1.80m

0.50%
0.00%
31

33

35

37

39

X(m)

(b)
5%

4%
ElasticPlastic+Internal
Pressure

3%

Hardening+InternalPressure

x
2%

ElasticPlastic
Hardening

1%

0%
0

0.5

1.5

2.5

h(m)

Figure 7.3.2. (a) The distribution of axial compressive strains along the critical buckling area,
until the 3% limit of tensile strain is reached.(D/t = 72). (b) The evolution of axial strains at the
critical points for the 4 models.(D/t = 72)

271

HardeningwithInternalPressure ReverseFault

h =0.10m

h =0.15m

h =0.30m

h =0.50m

Figure 7.3.3. The axial strain distribution along the central model area (X=2045 m) for
different fault displacements.(D/t = 72)

272

28

(a)

28.5

29

29.5

30

30.5

31

0.00%

0.50%

1.00%
h=0.15m
h=0.20m
x 1.50%

h=0.24m

2.00%

2.50%

(m)

3.00%

0.3

(b)
0.25

ElasticPlastic
Hardening

0.2

HardeningwithInternal
pressure

hcr/D 0.15

ElasticPlasticwithInternal
Pressure

0.1

0.05

0
48

72

96

120

144

168

D/t

Figure 7.3.4. (a) The distribution of axial compressive strains along the critical buckling area
until the appearance of buckling.(D/t = 72). (b) Comparison of the pipe capacity, in terms of
critical fault displacement, with respect to to the D/r ratio, between the four models.

273

274





Chapter 8

Conclusions and proposals


for further research

275

276

ConclusionsandProposalsforfurtherresearch

8.1 Conclusions

The studys purpose was, mainly, to propose a finite element model that properly
addressestotheproblemofapipecrossinganactiveDipSlipfaultandbyusingthis
model,toexaminetheeffectofvariousparameterssuchasthetypeoffault(Normal
or Reverse), the diameter to thickness pipe ratio, the depth of the soil layer, the
stressstrainrelationofthepipematerialandtheinternalpipepressure.
Themodelweusedisa3DF.E.modelthatsimulatesthepipewithshellelements
andthesurroundingsoilwithcontinuumbrickelements.Inordertoconfrontthe
problemthatarisesfromthecontradictionofthecontinuityofthepipelineandthe
limiteddimensionsofthemodel,weexpandedthepipebyonekilometerfromeach
pipe end using linear beam elements. The soil around the expansion parts was
substitutedbyaxial,verticalandhorizontalsoilsprings,whosepropertieshadbeen
takenbytheresultsofanumericalpushoveranalysis.Theuseofthebeamelements
and soil springs allows the increase of our models dimensions without significant
increaseofcomputationaleffortandtime.Comparingtheresultsofthismodelwith
models of free and fixed pipe ends, both for normal and reverse fault case, we
concludethatthebehaviorofthepipeisheavilyaffectedbythepipeboundariesand
we recommend the proposed model as the most reliable and realistic one. Its
expanded dimensions also allow the computation of the effective pipe length that
denotesthelengthofthepipethatisaffectedbytherupture.
Theexaminationofbothnormalandreversefaultruptureprovedthat,becauseof
theirtensileandcompressivenaturerespectively,themodeoffailureforthenormal
caseistheexcessivetensilestrainwhereasforthereversecaseisthelocalbuckling.
Inallofthecasesexaminedinthisessay,itisdemonstratedthatthedecreaseofD/t
ratio (or the increase of pipe thickness, since we keep the diameter unchanged)
increasesthepipeenduranceintermsoffaultdisplacement.

277

ConclusionsandProposalsforfurtherresearch

The investigation of the soil layers depth effect concludes that the pipe can resist
significantlymorebedrockdisplacementbeforeitreachesthefailurestate,because
ofthewiderdistributionofsoildisplacements.
All the aforementioned investigations were initially conducted for elasticperfectly
plastic pipe steel and were repeated for an elastichardening plastic pipe material.
Theresultsoftheanalysesindicatethatforthenormalfaultcase,hardeningplaysa
significantrolesincefirstly,itaugmentsthepiperesistancetofaultmovementand
secondlyitisprovennecessaryforthemodeltobeabletocaptureanothermodeof
failure knows as ovalization or flattening, related to excessive crosssectional
distortion and operational failure. However, the role of hardening is not that
significantforthereversefaultcasewherelocalbucklingisthefailuremode.
Regardingtheinternalpipepressure,ithasbeenproventhatithasanegativeresult
to the pipe resistance since it causes additional stresses and strains, leading to
prematureyieldingandfailure.Finally,itprovokesadifferentstraindistributionwith
multiple critical areas in the case of normal fault and is proven necessary for the
beneficialeffectsofthehardeningtoappearinthereversefaultcase.

278

ConclusionsandProposalsforfurtherresearch

8.2 Proposals for further Research

ThisessayisapreliminaryinvestigationofthebehaviorofapipesubjectedtoDip
Slipfaultrupture.Apartfromtheaspectsthatweexaminedtherearelotsofother
parametersthatcouldbeinvestigatedinafutureresearchsuchas:
The type of soil. We have conducted analyses for different types of soil
(Dense/LooseSandandClay).However,wefocusedonlyintheDenseSandcasein
ordertobeabletocomparetheeffectsoftheotherparameters.Furthermore,itis
proved that the behavior of the system depends on the relative stiffness of the
system,inotherwords,ontheratioofsoilstiffness topipestiffness,whichwedo
changebyadjustingthethicknessofthepipe.
Thepresenceofaquiferofalevelclosetothesurface.
Thebackfillmaterial.Wehavemadetheassumptionthatthebackfillmaterialhas
thesamepropertiesastherestofthesoil,somethingthatisincontrasttothereality
since the soil that is used to cover the pipe is of less density because of the
excavation procedure. An interesting idea would be to examine different backfill
materials,apartfromtheexcavatedsoil.
The type of fault. Apart from our study, which concentrates on DipSlip fault
rupturesandapartfromtheStrikeSliprupture,whichiswidelyexaminedbyvarious
researchers, the case of an Oblique fault rupture could be investigated. Different
pipe behavior is expected in that case, because both of the Dip and Strike
components of the fault are measurable and significant. The oblique effect can be
alsoexaminedbychangingthepipefaultplaneintersectionangle.

279

280





References

281

282

References

ABAQUS(2011):UsersManual,Simulia,Providence,RI,USA.
AbdounT.H.,Ha,D.,ORourke,M.J.,Symans,M.D.,ORourke,T.D.,Palmer,M.C.,
and Stewart, H. E. (2009), Factors influencing the behavior of buried
pipelines subjected to earthquake faulting., Soil Dynamics and Earthquake
Engineering,Vol.29,pp.415427.
American Society of Civil Engineers (2009). Buried flexible steel pipe; design and
structuralanalysis.In:WhiddenWR,editor.ASCEManualofPractice,MOP.p.
119.
American Society of Mechanical Engineers (2006), Pipeline Transportation Systems
forLiquidHydrocarbonsandOtherLiquids,ANSI/ASMEB31.4.
AmericanSocietyofMechanicalEngineers(2007),GasTransmissionandDistribution
PipingSystems,ANSI/ASMEB31.8.
AnastasopoulosI.,GazetasG.,M.ASCE,BransbyM.F.,DaviesM.C.R.,NahasA.El,
(2007), Fault Rupture Propagation through Sand: FiniteElement Analysis
andVaidationthroughCentrifugeExperiments,JournalofGeotechnicaland
GeoenvironmentalEngineering,ASCE,p.943958
Anastasopoulos I., Callerio A., Bransby M. F., Davies M. C., Nahas A. El, Faccioli E.,
GazetasG.,MasellaA.,PaolucciR.,PeckerA.,RossigniolE.(2008),Numerical
analysesoffaultfoundationinteraction.,BulletinofEarthquakeEngineering,
Sprin\ger,6(4),645675.
ArifinR.B.,ShafrizalW.M.,WanB.,YusofM.,ZhaoP.,andBaiY.(2010),Seismic
analysisforthesubseapipelinesystem.,ProceedingsoftheASME201029th
International Conference on Ocean, Offshore and Arctic Engineering,
OMAE201020671,Shanghai,China.
ANSI/API Spec 5L, Specification for Line Pipe, 44th edition, American Petroleum
Institute,October2007.
Bray,J.D.(1990).Theeffectsoftectonicmovementsonstressesanddeformations
in earth embankments. Ph.D. dissertation, Univ. of California at Berkeley,
Berkeley,Calif.
Bray,J.D.,Seed,R.B.,Cluff,L.S.,andSeed,H.B.(1994a).Earthquakefaultrupture
propagationthroughsoil.J.Geotech.Engrg.,120(3),543561.

283

References

Canadian Standard Association (2007), Oil and Gas Pipeline Systems, CSAZ662,
Mississauga,Ontario,Canada.
Chiou YJ, Chi SY, Chang HY. A study on buried pipeline response to fault
movement.PressureVesselTechnology,ASME1994;Vol:116,3641.
Comit Europen de Normalisation (2006), Eurocode 8, Part 4: Silos, tanks and
pipelines,CENEN19984,Brussels,Belgium.
Dama E., Karamanos, S. A., and Gresnigt, A. M. (2007), "Failure of Locally Buckled
Pipelines.",JournalofPressureVesselTechnology,ASME,Vol.129,No.2,pp.
272279.
DouglasG.Honeggeretal.(2004),TransAlaskaPipelineSystemPerformanceinthe
2002DenaliFault,Alaska,Earthquake.,EarthquakeSpectra,Volume20,No.
3,pages707738,August2004.
Erickson, S. G., Staryer, L. M., and Suppe, J. (2001). Initiation and reactivation of
faults during movement over a thrustfault ramp: Numerical mechanical
models.J.Struct.Geol.,23,1123.
Gazetas. G., Anastasopoulos I. and Apostolou. M. (2007), Shallow and deep
foundation under fault rapture or strong seismic shaking., K. Pitilakis (ed.),
EarthquakeGeotechnicalEngineering,Springer,185215.
ICFInternational,InterstateNaturalgasAssociationofAmericaReport:Naturalgas
pipelineandstorageinfrastructure,Projectionsthrough2030,October2009
Karamitros D.K., Bouckovalas G.D., Kouretzis G.P., (2006),Stress analysis of buried
steel pipelines at strikeslip fault corssings, Soil Dynamics and Earthquake
Engineering27(2007),p.200211.
KaramitrosD.K.,BouckovalasG.D.,KouretzisG.P.,GkesouliV.,Ananalyticalmethod
for strength verification of buried steel pipeline at normal fault crossings,
SoilDynamicsandEarthquakeEngineering31(2011),p.14521464.
Kennedy,R.P.,Chow,A.W.andWilliamson,R.A.(1977),Faultmovementeffectson
buried oil pipeline, Journal of Transportation Engineering, ASCE, Vol. 103,
pp.617633.
Kennedy R. P., Kincaid R. H. (1983). Fault crossing design for buried gas oil
pipelines.ASME,PVPconference;77:19.

284

References

Liu, X. and O'Rourke, M. J. (1997), Behaviour of continuous pipeline subject to


transversePGD.EarthquakeEngng.Struct.Dyn.,26:9891003.
Loukidis, D. (1999). Active fault propagation through soil. Thesis, School of Civil
Engineering,NationalTechnicalUniv.,Athens,Greece.
MaCaffrey MA, ORourke TD (1983). Buried pipeline response to reverse faulting
during the 1971 San Fernando Earthquake. ASME, PVP conference 1983;
77:151159.
Nakai, T., Muir Wood, D., and Stone, K. J. L. (1995). Numerical calculations of soil
responseoveradisplacingbasement.SoilsFound.,35(2),2535.
NEESRSG Final Report, Cornell University, Rensselaer Polytechnic Institute, The
SciencenterDiscoveryCenter,June25,2008.
Newmark,N.M.,HallW.J.(1975),Pipelinedesigntoresistlargefaultdisplacement.
Proceedings of U.S. National Conference on Earthquake Engineering; 416
425.
Odina,L.andTan,R.(2009),SeismicFaultDisplacementofBuriedPipelinesUsing
Continuum Finite Element Methods., Proceedings of the ASME 2009 28th
International Conference on Ocean, Offshore and Arctic Engineering,
OMAE200979739,Honolulu,Hawaii.
Odina,L.andConder,R.J.(2010),SignificanceofLder'sPlateauonPipelineFault
Crossing Assessment, Proceedings of the ASME 2010 29th International
Conference on Ocean, Offshore and Arctic Engineering, OMAE201020715,
Shanghai,China.
ORourkeT.D.,Geohazardsandlarge,geographicallydistributedsystems.,(2010),
Geptechnique60,No.7,507543
Papazachos B., Papaioannou Ch., Papazachos C., Savvaidis A., (1997), Atlas of
Isoseismal Maps for Strong Shallow Earthquakes in Greece and Surrounding
Area(426BC1995).ZitiPublicationCo,Thessaloniki(1997).
PeterJ.Haeussleretal.,(2004),SurfaceRuptureofthe2002DenaliFault,Alaska,
Earthquake and Comparison with other StrikeSlip Ruptures., Earthquake
Spectra,Volume20,No.3,pages565578,August2004.

285

References

Potts, D. M., Kovacevic, N., and Vaughan, P. R. (1997). Delayed collapse of cut
slopesinstiffclay.Geotechnique,47(5),953982.
Roth,W.H.,Sweet,J.,andGoodman,R.E.(1982).Numericalandphysicalmodeling
of flexural slip phenomena and potential for fault movement. Rock Mech.,
12,2746.
Scott, R. F., and Schoustra, J. J. (1974). Nuclear power plant sitting on deep
alluvium.J.Geotech.Engrg.Div.,100(4),449459.
Sim W.W., Towhata I., Yamada S., Moinet G.J.M., (2012), Shaking table tests
modeling small diameter pipes crossing a vertical fault., Soil Dynamics and
EarthquakeEngineering,Volume35,p.5971,April2012,Elsevier.
Takada,S.,Hassani,N.andFukuda,K.(2001),Anewproposalforsimplifieddesign
of buried steel pipes crossing active faults, Earthquake Engineering and
StructuralDynamics,2001;Vol.30:pp.12431257.
Trifonov,O.V.andCherniy,V.P.(2010),Asemianalyticalapproachtoanonlinear
stressstrain analysis of buried steel pipelines crossing active faults., Soil
Dynamics&EarthquakeEngineering,Vol.30,pp.12981308.
Trifonov,O.V.andCherniy,V.P.(2012),Elastoplasticstressstrainanalysisofburied
steel pipelines subjected to fault displacements with account for service
loads.,SoilDynamics&EarthquakeEngineering,Vol.33,No.1,pp.5462.
U.S.EnergyInformationAdministration,OfficeofOilandGas,Report:Naturalgas
compressorstationsontheInterstatepipelinenetwork:Developmentssince
1996,November2007
Vougioukas E. A., Theodossis, C., Carydis P. G. (1979), Seismic analysis of buried
pipelines subjected to vertical fault movement., Journal of Technical
Councils,ASCE,Vol.105(TCI),pp.432441.
VazourasP.,KaramanosS.A.,DakoulasP.,(2010),Finiteelementanalysisofburied
steel pipelines under strikeslip fault displacements., Soil Dynamics and
EarthquakeEngineering(2010),doi:10.1016/j.soildyn.2010.06.011
Vazouras P., Karamanos S. A., Dakoulas P., (2012), Mechanical behavior of buried
steel pipes crossing active strikeslip faults, Soil Dynamics and Earthquake
Engineering,Volume41,October2012,Pages164180

286

References

Walters, J. V., and Thomas, J. N. (1982). Shear zone development in granular


materials.Proc.,4thInt.Conf.onNumericalMethodsinGeomechanics,Vol.
I,Edmonton,Canada,263274.
Wang,L.R.L.andYeh,Y.A.(1985),Arefinedseismicanalysisanddesignofburied
pipelineforfaultmovement,EarthquakeEngineering&StructuralDynamics,
Vol.13,pp.7596.
Wang,Y.Y.,Horsley,D.andLiu,M.,(2007),StrainBasedDesignofPipelines.,16th
JointTechnicalMeeting,AustralianPipelineAssociation,Canberra,Australia.
White, R. J., Stone, K. J. L., and Jewel, R. J. (1994). Effect of particle size on
localization development in model tests on sand. Proc., Inst. Centrifuge
Conf., C. F. Leung, F. H. Lee, and T. S. Tan, eds., Balkema, Rotterdam, The
Netherlands,817822.
YoshizakiK.,ORourkeT.D.,HamadaM.,(2003),Largescaleexperimentsofburied
steel pipelines with elbows subjected to permanent ground deformation,
StructuralEng/EarthquakeEng.,JSCE,Vol.20,No.1,111.
.K.,,

287

References

InternetReferences

AmericanOilandGasHistoricalSociety,www.aoghs.com

AmericanPetroleumInstitute,www.api.org

UnitedStatesGeologicalSurvey,www.USGS.org

..,www.desfa.gr

InternationalGasUnion,www.inu.org

InterconnectorTurkeyGreeceItalyPipeline,
www.edison.it/en/company/gasinfrastractures/itgi.shtml

TransAdriaticPipeline,www.transadriaticpipeline.com

288