9/21/2015

JohnS.HorvathConsultingEngineer:onlineinformationpageSSI#1

TheCoefficientofSubgradeReaction:
IsIttheMostMisunderstoodConceptinFoundationEngineering?
....AndHowDidItGetThatWay?

"Thedesignof...raftfoundationsundoubtedlyrepresentsoneofthemoredifficulttechnicalaspectsofcivil
engineeringpractice...Untilfairlyrecently,therewaslittlealternativebuttoproceedonthebasisofgreatlysimplifying
assumptionscombinedwithrudimentaryanalysis.Butalthoughmanysuchdesignsweredevelopedwithremarkable
success,thelimitationsofthistraditionalapproachcannotbedisregardedandoftenareunacceptableinmodern
practice."
J.A.Hemsley,editorofDesignApplicationsofRaftFoundations,2000
Callitwhateveryoulike...coefficientofsubgradereaction,modulusofsubgradereaction,subgrademodulus,soilspring
constant,Winkler'sspringconstant...thereisprobablynootherparameterinalloffoundationengineeringthatissooftensought
afterforuseinpracticeyetsomisunderstoodastowhatitreallyisandhowitshouldbeproperlyevaluatedinaparticular
application.Inmanywaysitistofoundationengineeringwhattheunicornistomythologyfoundintheliteratureofmanyancient
civilizations.Statedanotherway,thewidelyheldperceptionthatthecoefficientofsubgradereactionisafundamentalsoilproperty
thatcansomehowberationallyestimatedbeforehandinagivenproblemapplicationiseverybitasmythicalastheunicornoflore
andyore.
ItwasthisrecurrentambiguityinpracticethatattractedDr.Horvathtothesubjectin1973whenhebeganhisdoctoralstudies.
Aftermorethanfourdecadesofresearchintothistopic,whichincludedpublicationin1979ofhisdoctoralthesisonthesubject,
hisconclusioncanbestatedsimply:

Thecoefficientofsubgradereactionisaproblemspecificobservedresultandnotafundamentalsoilproperty.

Thereforeanyefforttomeasureorcalculateitsvaluebeforehandbasedonsomesoiltestisinherentlyflawedanddoomedto
failure.Howeverthishasnotstoppedgenerationsofcivilengineersoverdecadesoftimefromchasingafterthisnonexistentgoal.
Sohowdidthingsgettowheretheyare?Simple,conciseexplanationsaredifficulttomakebutoneisattemptedhere.
Imagineaclassicsoilstructureinteraction(SSI)problemsuchasamat(raft)foundationsupportingabuildingsuperstructureand
restingonasoilsubgrade.Reactionsfromthesuperstructureresultinloadsareappliedtothetopofthematandthemat
displacesverticallybysettlingintotheunderlyingsoilsubgrade.Inanidealworld,therewouldbeasingleanalyticalmodelthat
wouldallowalltherelevantproblemparameters...thegeometryandfundamentalmaterialpropertiesofthesuperstructure,mat,
andunderlyingsoil...tobeinputintoaclosedformsolutionor,morelikely,besolvednumericallysothattherelationshipbetween
superstructureloadsandresultsofinterest(typicallythetotalsettlementofandmomentswithinthemat)couldbecalculated
directlyandcorrectly.
Whilesuchasingle,unifiedanalysisisnowadayspossibleinconcept,problemsinvolvingSSItendtobecomplexbecauseofthe
nonlinear,stressdependentnatureofsoilbehaviorandtheinherentthreedimensionalnatureofmostpracticalproblems.Even
withtheavailabilityofpowerfulcomputerhardwareandsoftware,manySSIproblemsarestilltoocomplextoanalyzeroutinelyas
theyshouldbebecauseoftheirthreedimensional(3D)nature.Certainlyinthedaysbeforecomputers,mostSSIproblemswere
simplyunsolvableintheirtrueform.SohistoricallyandeventothisdaymanySSIproblemshavetobesimplifiedinroutine
practicetorenderthemsolvable.
Formatfoundations,thetraditionalapproachthatwasusedinprecomputerdayswastobreaktheproblemintoitsthreedistinct
componentsandanalyzeeachseparately(butreasonablyaccurately)althoughcertainsimplifyingassumptionshadtobemadeas
tohowthecomponentsinteractedwitheachother.Thustherewouldbeaseparatestructuralanalysisofthesuperstructuretobe
supportedonthemataseparatestructuralanalysisofthematitselfandasettlementanalysisofthesoilsubgrade.Theanalysis
ofthesuperstructurecouldverywellneglectanydifferentialsettlementsatthesupports,i.e.thetopofthemat,whichmeans
thatsuperstructurematinteractionwasneglected.Theinteractionbetweenmatandsubgradewasassumedtobethroughthe
normalstressattheircontact.Thisstresshasbeenreferredtovariouslyasthebearingstress,bearingpressure,contactstress,
orcontactpressurebutthetermsubgradereactionstressorsimplysubgradereactionistheoneusedinthisdiscussion.
Therearetwothingstonoteaboutthesubgradereaction:
itisagenerictermandnotdependentonorindicativeofaparticularwayinwhichthesoilismodeledmathematicallyand
itis,inreality,aninternalstresswithintheoverallsuperstructurematsubgradesystemthatwouldneverappearexplicitlyin
thecalculatedproblemresultsexceptforthefactthattheoverallproblemisartificiallycutatthematsubgradeinterface.
Thelatterpointisanalogoustoabeamthatisanalyticallycutforthepurposesofcalculatinginternalstressesatsomepoint

http://www.jshce.com/online_info_ssi01.html

1/4

9/21/2015

JohnS.HorvathConsultingEngineer:onlineinformationpageSSI#1

withinthebeam.Theseinternalstressesarenotpartofthefundamentalproblemresults(beamdeflectionandmoments,and
supportreactions)thatwouldnormallybeseen.
Aswillbeseensubsequently,italsoprovesusefultodefineagenericparametercalledthecoefficientofsubgradereactionthatis
theratioofthesubgradereactionatapointalongmatsubgradeinterfacetodisplacement(settlementinthematproblem)ata
givenpoint.Again,notethatthecoefficientofsubgradereaction:
isagenerictermandnotdependentonorindicativeofaparticularwayinwhichthesoilismodeledmathematically
will,ingeneral,varyfrompointtopointacrossthematsubgradeinterfaceand,
mostimportantly,isacalculatedresultandnotaprobleminputparameter.
Nowadaysitisnolongernecessary,andthusnolongerjustifiableordefensibleassupportedbythequoteatthetopofthis
page,toanalyzematfoundationsusingthistraditionaltripartiteapproach.Thecurrentstateofpracticeandcomputersoftware
allowsthesuperstructureandmattoberoutinelymodeledandanalyzedasasingle,combinedstructuralsystemwithgreat
accuracy.However,thecurrentcapabilitiesofcommerciallyavailablestructuralanalysissoftwarestillrequiresignificant
simplificationofthesoilsubgradeeffects.Insimpleterms,thedepthdimensionofthesoilsubgrademustalwaysbeeliminated
fromexplicitmodelingasdiscussedsubsequently.
Thisnecessarysimplificationinroutinepracticeunfortunatelyperpetuatestheperceptionthatthesubgradereactionandthe
coefficientofsubgradereactionarefundamentalproblemparametersandsoilpropertiesrequiringinputasopposedtobeing
problemresultsoroutcomesastheyactuallyareinan'ideal'analysis.Thereasonisthatinsteadofmodelingthesoilunderlyinga
matinitstruenonlinear,3Dnatureandasaninherentpartoftheoverallproblem,theentireeffectofthesoilsubgradeis
artificiallyseparatedfromthestructuralpartsoftheproblem(superstructure+mat)andreplacedbyamuchsimplersystem
called
asubgrademodel.Subgrademodelsaresometimescalledsurfaceelementorboundaryelementmodelsbecauseatwo
dimensional(2D)mathematicalexpressionisappliedoverthesubgradesurfacetorepresentwhatisreallyhappeninginthree
dimensions.Effectively,thethirddimension,depth,ismathematicallycollapsedtoazerothickness.Similarly,a2D(planestrain)
problemcollapsesthesecond(depth)dimensionofthesoilsubgradesothataonedimensional(1D)subgrademodelisapplied
alongthematsubgradeinterface.
Itisimportanttonotethatthegeneralapproachofusinga2Dsubgrademodelin3Dproblemsora1Dsubgrademodelin2D
problemsisnotinherentlyflawed.Indeedresearchandexperiencetodateindicatethatsatisfactoryresultscanbeobtainedusing
subgrademodels.Theproblem,ofcourse,istofindthesubgrademodelforagivenapplicationthatstrikesabalancebetween
accuracyofcalculatedresultsandeaseofuse,and,perhapsmostimportantly,canbeaccommodatedwithinthecapabilitiesof
commerciallyavailablestructuralanalysissoftware.
WithoutadoubtthemostcommonlyusedsubgrademodelingeotechnicalengineeringpracticetodateisWinkler'sHypothesis,a
mathematicalabstractionthatdatesbacktothe19thCentury.Winklerstatedmathematically(apparentlywithoutanyjustification
otherthansimplicity)thatthereisadirectcauseeffectrelationshipbetweenanormalstressappliedtoaunitarea(squaremetre,
squarefoot,etc.)ofasubgradesurfaceandthenormaldisplacementofthatunitarea.Statedanotherway,astressappliedto
thesubgradesurfacehasaneffectonlyunderthatstress.
SubsequenttoWinkler'sstatinghishypothesis,othershavephysicallyinterpretedhismathematicalabstractionasmeaningthat
thesubgrade'sreactiontoanappliedstresscanbevisualizedasaseriesofindependent(butnotnecessarilylinearorelastic)axial
springs(averyappealingconcept)oraliquid(muchlessso).Althoughaspringandaliquid(curiouslymostcallita'denseliquid'
buttheadjective'dense'ismeaningless,irrelevant,andthusunnecessary)aretwoverydifferentphysicalobjects/materials,they
areequivalentinthiscontextasaphysical/visualinterpretationofWinkler'sabstractmathematicalhypothesis.Thereasonisthat
theparameterinWinkler'sHypothesisthatrelatesthedirectrelationshipbetweenappliedstressanddisplacementatsomepoint
onthesubgradesurface(whichwasdefinedpreviouslyinamoregenericcontextasthecoefficientofsubgradereaction)has
dimensionsofforceperlengthcubedwhichismathematicallythesameaseitheraspringconstant(forceperlength)supportinga
unitarea(lengthsquared)orfluidunitweight(forceperlengthcubed).Notethatthepreferrednameofthisparameterthatisthe
heartofWinkler'sHypothesisisWinkler'sCoefficientofSubgradeReaction(todistinguishitfromthegenericcoefficientof
subgradereaction)anditsusualnotationisk(althoughkWispreferredtoeliminateambiguity).
ArguablythesinglemostimportantaspectofWinkler'sHypothesisanditsuseasasoilmodelisthatittransformsthegeneric
coefficientofsubgradereaction(whichisaratiooftwocalculatedresultsinan'ideal'SSIanalysis)intotheWinklerCoefficientof
SubgradeReactionwhichalwayshastobeaknowninputparameter.Thisrolereversalissomethingthatdoesnotappeartobe
widelyappreciated.
ItisofinterestandrelevancetonotethatWinklerneverstatedthathisCoefficientofSubgradeReactionhadtobeaconstantfor
agivenproblem.Itcouldvaryfrompointtopointalongtheinterfacebetweenstructure(matinthisdiscussion)andsubgrade.
TheuseofaconstantvalueoftheWinklerCoefficientofSubgradeReactionissomethingthatevolvedinpracticefromearly(i.e.
precomputer)closedformsolutionsdevelopedduringthe20thCenturythatrequiredasingle,constantvalueofkW.
Furthermore,thereisnothinginherentinWinkler'sHypothesisthatagiven'spring'beeitherlinearorelasticalthoughhistorically
linearelastic'springs'havebeenusedinmatfoundationproblemsbutnotforlaterallyloadeddeepfoundationswheretheso
calledpycurvesarenothingmorethannonlinear,inelasticWinkler'springs'.
WhateveronewantstocallthisparameterkW(inanyevent'coefficient'ismostdefinitelypreferredover'modulus'asmodulusin
acivilengineeringcontexttypicallyhasdimensionsofforceperlengthsquared,notforceperlengthcubedasdoeskW),itis
fundamentally,inherently,andfatallyflawedbythefactthatitistheparameterofasubgrademodelthatpoorlyreplicateshow
realsoilsbehave.Simplystated,astressappliedatapointonanactualsubgradesurfacecausesdisplacementatmanypoints,
notjustbeneathit.Conversely,thedisplacementofsomepointonthesubgradesurfaceistheresultofstressesappliedatmany
points,notjustdirectlyaboveit.Thisactualsoilbehavior(whichissometimescalledloadspreadingandisthephysicalresultof
thefactthatsoilcanreadilytransmitshearstresses)isreferredtointhecontextofdiscussingsubgrademodelsasspring
couplingbecause'real'soilspringscanbevisualizedasbeinglinkedtogether,notindependentasWinkler'sHypothesisimplies.It

http://www.jshce.com/online_info_ssi01.html

2/4

9/21/2015

JohnS.HorvathConsultingEngineer:onlineinformationpageSSI#1

isthissignificantflaw(i.e.thelackofspringcoupling)inWinkler'sHypothesisthatsingularlymakesitsuchapoorsubgrade
model.
Thepracticalimplicationsofthisflawcanbeseenifwereturntooursimplematfoundationproblem.Whatonewouldfind,for
example,isthatforagivencombinationofmatandsubgradetheactualsubgradereaction,andthustheactualgenericcoefficient
ofsubgradereaction,wouldnotonlyvaryalongthematsubgradecontactbutwouldvarydependingontheexactloading.For
example,ifonetookthesametotalforceandinonecaseapplieditinarelativelyconcentratedfashion(suchasfromabuilding
column)andinanothercaseinamoredistributedfashion(suchasfromadistributedfloorload)thematsettlementwouldbe
different.Thusnotonlywouldthecoefficientofsubgradereactionnotbeaconstantforthismat/subgradecombinationthere
wouldbetwodifferentnonuniformdistributionsofcoefficientsofsubgradereactionthatwouldlegitimatelybecalculatedforthe
samemat/subgradecombinationreactingtothesametotalforcebutappliedintwodifferentways.
ItisalsousefultocitetheclassicalresearchofVesicandSaxena(1970)whostudiedthesimple,basicproblemofaslabwitha
singlepointload.TheydemonstratedthatitwouldtaketwodifferentvaluesoftheWinklercoefficientofsubgradereactionto
correctlypredicttheslabsettlementsandslabbendingmoments.Eventhen,thecalculatedandactualresultsforboth
settlementsandmomentswouldmatchonlyatthepointdirectlyundertheappliedload.Notethatsincebothsettlementand
momentsareimportantinslabormatdesignnosinglevalueoftheWinklerCoefficientofSubgradeReactioncouldreliablyserve
theoneproblem.Ofcourseifoneweretochangetheflexuralstiffnessoftheslabormateverythingchangesaswell.
Insummaryandconclusion,subgradereaction,andthethusthegenericcoefficientofsubgradereactionthatisderivedfromit,is
actuallyanobserved,calculatedresultinanySSIprobleminthesamewaythatforcesandmomentswithinstructuralmembers
areobservedresultsfromastructuralanalysis.ItisonlywhenarelativelycrudesubgrademodelsuchasWinkler'sHypothesisis
used('crude'inthesensethatitdoesnotproperlyreplicateeventhemostbasicaspectsofactualsoilbehavior)thatthe
coefficientofsubgradereactionmustnowbecomeaninputparameterifthecorrectresultsaretobecalculatedusingamodel
suchasWinkler's.
Furthermore,itisnowwidelyrecognizedthattheWinklerCoefficientofSubgradeReactioncanneverbeaconstantinagiven
problembutliterallymustvaryfrompointtopointbecausetheactualcontactstress(subgradereaction)betweenafoundation
elementandunderlyingsubgradevariesfrompointtopoint.However,themostastoundingaspectofthisentirediscussionisthat
onemustbasicallyknowtheanswers(i.e.subgradereactionandsettlementmagnitudesanddistributions)toagivenproblem
beforehandsothatthe'correct'valuesofWinkler'sCoefficientofSubgradeReactioncanbecalculatedandinputtoyieldthe
correctresultsforsubgradereactionandsettlement.Wasthereeveramoreabsurdexampleofcircularlogicingeotechnical
engineering?Yetthishasbeenthecaseforover100yearsnoweverytimeanengineerusesWinkler'sHypothesis.Thisisthe
pricetopayforusingsuchasimplisticsubgrademodelthatdoesnotinherentlymodeltheimportantbehavioralphenomenonof
soilshear.Theburdenisonthegeotechnicalengineertoestimatetheeffectofsoilshearinagivenproblemandreflectitinthe
inputvalueofWinkler'sCoefficientofSubgradeReaction.
NotethatthisconclusionthatknowingthecorrectanswerbeforehandisnecessarytouseaWinklersubgradetoproducethe
correctanswerisnotillogicalinallSSIapplications.Forexample,thisispreciselywhatisdonewithlaterallyloadeddeep
foundationswherethepycurvesareessentiallythenonlinear,inelasticspringstiffnessesofaWinklersubgradeunderhorizontal
stressapplication.Inthiscaseithasbecomepossibleafterdecadesofresearchandcountlessinstrumentedloadteststo
essentiallybeabletoknowthecorrectanswersbeforehandforawelldefinedrangeofproblemparameters(typesofdeep
foundationelementsandsubgrades).Unfortunately,thereisnocorrespondingbodyofknowledgefrominstrumentedmat
foundationsorotherSSIproblemsthatallowsonetoknowthecorrectanswerbeforehandsothatWinklerspringvaluescanbe
accuratelychosentoproducethecorrectanswer.
Hopefully,itisnowobviouswhytryingtoestimateormeasurethe'correct'valueoftheWinklerCoefficientofSubgradeReaction
isalwaysanimpossibletask.Evenmoreimpossibleisthinkingthatasinglevalueforthisparameterexistsinanygivenproblem.
Thepointsmadeinthisdiscoursehavebeenknownfordecades.Researchpapersthatsoughttodevelopsubgrademodelsthat
improveonWinkler'sHypothesiswerepublishedintheEnglishlanguageatleastasfarbackas1940.Subgrademodelsthat
inherentlyincorporatesoilshear(springcoupling),atleastinsomerudimentarybuteffectiveway,haveexistedfordecades.Yet
civilengineerscontinuetouseWinkler'sHypothesisandpursueafruitlessgoaltoquantifyhisCoefficientofSubgradeReaction,
kW.TheonlyrationalexplanationtothisirrationalsituationisthepowerfulappealofthesimplicityofusingaWinklersubgrade(it
isveryeasytomodelindependentaxialspringsinstructuralanalyses).
However,allisnothopelessinSSIanalysesforproblemsinvolvingmatfoundationsandsimilarstructures.InrecentyearsDr.
HorvathincollaborationwithMr.RegisJ.Colasanti,P.E.havenotonlycollaboratedtodevelopanadvancedsubgrademodel,the
HorvathColasantiSubgradeModel,thatinherentlyincorporatesspringcouplinginitstheoreticalderivationtheyhaveshown:
thatthemodelparameterscanbedefinedintermsoffundamentalsoilparameters(elasticmoduli)andtheproblem
geometry(HorvathandColasanti2011a)
howthesemodelparameterscanbeevaluatedinpracticeusingcasehistoriesofactualmatfoundations(Horvathand
Colasanti2011b)and
howthemodelcanreadilybeimplementedincommerciallyavailablestructuralanalysissoftware(ColasantiandHorvath
2010).
Asaresult,thereisnoneedforpractitionerstohavetoresorttousingWinkler'sHypothesisasasubgrademodelanylonger.

References
Vesic,A.S.andSaxena,S.K.(1970)."AnalysisofstructuralbehaviorofAASHOroadtestrigidpavements",NationalCooperative
HighwayResearchProgramReport97,HighwayResearchBoard,Washington,D.C.,U.S.A.

http://www.jshce.com/online_info_ssi01.html

3/4

9/21/2015

JohnS.HorvathConsultingEngineer:onlineinformationpageSSI#2

BeamColumns:
AnInterestingbutMisunderstoodConceptinStructuralMechanics

Oneofthemostmisunderstoodconceptsinstructuralmechanicsinvolvesastructuralmembercalledabeamcolumn.Itisused
routinelyinsoilstructureinteraction(SSI)modelingsoadiscussiontoclarifythemisconceptionssurroundingthisuniqueand
intriguingstructuralmember(manyassumethatanybeamwithbothtransverseandaxialloadingisabeamcolumnwhichis
simplyincorrect)isusefulsothatbeamcolumnsarereferredtoandusedproperlyinSSIanalyses.
Thekeytounderstandingthe'true'beamcolumninvolvesoneoftheassumptionsandapproximationsinherentindeveloping
EulerBernoullior'simple'beamtheory.ConsiderthebeamshowninFigure1below(notethattheappliedtransverseload,q(x),
isnotnecessarilyuniforminmagnitudeorevencontinuousalongtheentirelengthofthebeam):

Assumingaconstantflexuralstiffness,EI,theclassicaldifferentialequationdefiningthebehaviorofthisbeamis
EIw''''=q(x)(1)
wherew''''isthefourthderivativeofthebeamdeflection,w,withrespecttothexaxis.Themostimportantthingtonoteforthis
discussionisthatEquation1isbasedontheinitial,undeformedgeometryofthebeam.Thisisoneoftheclassicelementsof
simplebeamtheory.Theimplicationsofthisassumptioncanbeillustratedbyaqualitativeexample.
ConsiderthebeamshowninFigure2belowwhichisidenticaltotheoneshowninFigure1butwiththeadditionofanaxialforce,
P,thatisdefinedasbeingpositiveincompression.Asbefore,theappliedtransverseload,q(x),isnotnecessarilyconstantor
continuous.

Undersimplebeamtheory,theaxialforcehasnoeffectwhatsoeverontheflexuralbehaviorofthisbeamandonlycausesaxial
stresswithin,andconcomitantaxialstrainof,thebeam.Axialonlyeffectsaresometimescalledtrusseffectsorbehavior.Thisis
becauseaconceptanalogoustosimplebeamtheory(i.e.astructuralmemberthathasflexuraleffectsonly)thatisoftenusedin
structuralmodelingistohavestructuralmembers(elements)thatrespondonlytoaxialforceswithnoflexuraleffects.Thisisthe
http://www.jshce.com/online_info_ssi02.html

1/3

9/21/2015

JohnS.HorvathConsultingEngineer:onlineinformationpageSSI#2

assumptionmadeinbasicorsimpletrussanalysisforexample.Flexuralstructuralelementsandtrussstructuralelementsarethe
basicelementsofwhatisnowcalledlinearanalysisinstructuralengineering.Infact,forasimplebeamwithelasticmaterial
behaviortheloadPcanbeincreasedwithouttheoreticallimitandwillnevercauseaxialbucklingofthebeam.
Inreality,anactualstructuralmemberundercombinedtransverseandaxialcompressiveloads,q(x)andPrespectively,will
developadditionaldisplacements,forces,andmomentsnotpredictedbysimplebeamtheoryandeventuallybuckle.Thisfacthas
longbeenrecognizedbystructuralengineers.Inthepast,thehigherorderbehaviorbeyondthatpredictedbysimplebeam
theorywascalledthePdeltaeffectbecausetheseadditionaldisplacements,etc.werecausedbytheaxialcompressiveforce,P,
beingdisplacedtransversely(verticallyinFigure2)throughsomedistancedelta(paralleltothezaxisinFigure2).Nowadays,a
higherorderstructuralanalysisperformedconsideringthePdeltaeffectisreferredtoasanonlinearanalysis.
Restatedsimplythen,linearstructuraltheoryandanalysisisbasedontheinitial,undeformedgeometryofastructureand
nonlinearstructuraltheoryandanalysistakesintoaccountthedeformedgeometrythatdevelopsunderloadapplication.
Beforeproceeding,itisofinteresttodigressandcommentoncommerciallyavailablestructuralanalysissoftwarethatisused
routinelyinpractice.Whilemostsoftwarenowadayscanperformanonlinearanalysis(sometimesthisisanextracostmoduleor
option),thesimplerlinearanalysisistypicallythedefaultoperatingmodeofthesoftware.Furthermore,itisimportantto
understandthatmost,ifnotall,softwareaccomplishesanonlinearanalysisbyiterativeapplicationoflinearanalysisasopposedto
adirectanalysis.Thisisanimportantpointtonote.
Returningnowtothediscussionathand,asnotedatthebeginningofthisdiscussionmanycivilengineers,especiallygeotechnical
engineers,callanystructuralmemberwithbothtransverseandaxialloadsasshowninFigure2a'beamcolumn'.However,this
broad,genericterminologyisnotstrictlycorrect.Thisisbecause,strictlyspeaking,theterm'beamcolumn'onlyappliestoa
particularformulationoftheproblemshowninFigure2.Specifically,whatconstitutesa'true'beamcolumnandmakesitunique
in
structuralmechanicsisnotthecombinedtransverseandaxialloadingperse.Rather,itisthefactthattheproblemformulation
andderivationofthedifferentialequationdefiningthebehaviorofthetruebeamcolumnisbasedonthedeformedgeometryof
thebeam.Thisisasignificantdeviationfromsimplebeamtheorywhichisbasedontheundeformedgeometryasnoted
previously.Notethatusingthedeformedgeometrymakesthetruebeamcolumnadefactononlinearprobleminthecontextof
theprecedingdiscussion.However,whatisintriguingisthatifcertainboundaryconditionsexist(specifically,ifaxialdisplacements
causedbytheaxialload,P,areignored),itispossibletogetthefollowingclosedformsolutiondefiningtheflexuralbehaviorofa
truebeamcolumn:
EIw''''+Pw''=q(x)(2)
wherew''isthesecondderivativeofthebeamdeflection,w,withrespecttothexaxis.
ComparingEquation2toEquation1forthelinear,simplebeamsolution,itcanbeseenthattheflexuraleffectsoftheaxialforce,
P,onthedeformedshapeofthebeamarereflectedinthesecondtermonthelefthandsideofEquation2.Also,ifP=0the
beamcolumnEquation(2)revertsbacktothesimplebeamEquation(1)asexpected.
However,thetrulyintriguingandusefulaspectofthetruebeamcolumnsolutionisthatitturnswhatisfundamentallyanonlinear
problemintoaquasilinearproblembecauseitispossibletogetarelativelysimple,exact,closedformsolution(Equation(2))
definingthenonlinearflexuralbehaviorofthebeam.Thustheinherentlyapproximate,iterativesolutionsnormallyrequiredfor
nonlinearproblemswhenlineartheoryisusedasthebasisforanalysisarenotrequiredinthiscase.
Althoughnotillustratedhere,itisofinteresttonotethat,fromapracticalperspective,theeffectoftheaxialforce,P,onthe
flexuralbehaviorofabeamistomodifytheflexuralstiffnessofthebeam(thisismuchmoreapparentusingmatrixanalysisand
examiningthedifferencesinstiffnessmatrices).Acompressiveforce(positiveP)'softens'abeam,i.e.makesitmoreflexible.On
theotherhand,atensileforce(negativeP)'stiffens'abeam,i.e.makesitmorerigid.Alsonotillustratedhereisthatatrue
beamcolumnwillalwaysundergoaxialbucklingiftheaxialcompressiveforce,P,isincreasedtotheEulerbucklingload,Pcr.
Again,
thisisbestvisualizedandunderstoodusingstiffnessmatrices.
TherearesomeinterestingpracticalissuessurroundingthebeamcolumnEquation(2).Althoughithashistoricallyreceived
treatmentinsolidmechanicsandstructuralengineeringliterature(e.g.TimoshenkoandGere1961),itappearsthatthe
beamcolumnequationhasseenmuchmoreextensiveuseinpractice(atleastintheU.S.A.)notbystructuralengineersbutby
geotechnicalengineerswhohaveuseditsinceatleastthe1950stosolveawidevarietyoffoundationproblemsthatinvolveSSI.
Examplesofapplicationsincludedeepfoundationssubjectedtolateral/momentloadingflexibleearthretainingstructures
(primarilyanchoredandcantileveredsheetpilebulkheads)mat(raft)foundationsandsubaqueousconduits(Haliburton1971,
Dawkins1982,Horvath1988).Morerecently(ca.early1990s),thebeamcolumnequationwasusedasawaytovisualizethe
Pasternak/LoofHypothesisandFilonenkoBorodichmultipleparametersubgrademodels(Horvath1993).Thusthereisan
interestingparadoxthatwhatisessentiallyastructuralmechanicssolutionismuchmorefamiliarto,andapparentlyusedby,
geotechnicalengineerscomparedtostructuralengineers.
Perhapsasaresultofitsrelativelyunknownstatusamongstructuralengineers,itisofinteresttonotethatcommerciallyavailable
structuralanalysissoftwaregenerallydoesnothavethecapabilitytomodeltruebeamcolumnbehaviordirectly.Statedanother
way,atruebeamcolumnisnotinthecatalogofavailablestructuralmembers.Truebeamcolumnbehaviorcanonlybeanalyzed
bysubjectinganaxiallyloaded'normal'(simple)beamelementtoanonlinearanalysis.Whilethisessentiallyaccomplishesthe
sameendresultasusingthetruebeamcolumnequationdirectly,thestructuralanalystmustclearlyunderstandthesubtle
differencesinvolvedsothatanonlinearcomputeranalysisisinvokedwhentruebeamcolumnbehaviorisdesiredorrequired.
Asafinalcomment,giventheimportanceofthetruebeamcolumntoSSIanalysesitisusefultonotethatthebeamcolumn
equationcanbeextendedtoincludesubgradesupportwhichisalwaysakeyelementofSSIanalyses.Figure3belowillustrates
thegeneralproblemwherep(x)isthecontactstress(subgradereactionstressorsimplysubgradereaction)betweenthe
beamcolumnandsubgrade.Notethatp(x)isnotnecessarilyconstantalthoughinmostapplicationsitwillbecontinuous.As
http://www.jshce.com/online_info_ssi02.html

2/3

9/21/2015

JohnS.HorvathConsultingEngineer:onlineinformationpageSSI#2

before,theappliedtransverseload,q(x),isnotnecessarilyconstantorcontinuous.

Thedifferentialequationforatruebeamcolumnrestingonasubgradeis:
EIw''''+Pw''+p(x)=q(x)(3)
Howtomodelp(x)simplyyetaccuratelyusingwhatiscalledasubgrademodelistheperpetualchallengeofsubgrademodeling.

References
Dawkins,W.P.(1982)."User'sGuide:ComputerProgramforAnalysisofBeamColumnStructureswithNonlinearSupports
(CBEAMC)".InstructionReportK826,U.S.ArmyEngineerWaterwaysExperimentStation,Vicksburg,MS,90pp.
Haliburton,T.A.(1971)."SoilStructureInteraction:NumericalAnalysisofBeamsandBeamColumns".TechnicalPublicationNo.
14,OklahomaStateUniversity,SchoolofCivilEngineering,Stillwater,OK,179pp.
Horvath,J.S.(1988)."NumericalAnalysisofBeamsandBeamColumnswithLinearandNonLinearSpringSupportsUsingFinite
Differences".ResearchReportNo.CE/GE881,ManhattanCollege,CivilEngineeringDepartment,Bronx,NY.
Horvath,J.S.(1993)."BeamColumnAnalogyModelforSoilStructureInteractionAnalysis".JournalofGeotechnicalEngineering,
AmericanSocietyofCivilEngineers,Vol.119,No.2,pp.358364witherratainVol.119,No.7(July1993),p.1183.
Timoshenko,S.P.andJ.M.Gere(1961).TheoryofElasticStability.McGrawHillBookCompany,NewYork,NY,2ndedition,541
pp.

Copyright20082015byJohnS.Horvath.Allrightsreserved.Themoralrightsoftheauthorhavebeenasserted.Thispagewaslastrevisedon1January2015.

http://www.jshce.com/online_info_ssi02.html

3/3