Dual boundary element method in modelling of fatigue crack propagation

$UAL"OUNDARY%LEMENT-ETHODIN-ODELINGOF&ATIGUE#RACK0ROPAGATION

!HMAD+AMAL!RIFlN-OHD)HSANAND.IK!BDULLAH.IK-OHAMED

&ADHLUR2AHMAN-OHD2OMLAY

#OMPUTATIONAL-ECHANICS'ROUP

&ACULTYOF%NGINEERING"UILT%NVIRONMENT 5NIVERSITY+EBANGSAAN-ALAYSIA

5+-"ANGI3ELANGOR -ALAYSIA

&ACULTYOF-ECHANICAL%NGINEERING

+OLEJ5NIVERSITI+EJURUTERAAN4EKNOLOGI-ALAYSIA+5+4%- 'AMBANG+UANTAN0AHANG

-ALAYSIA

2ECEIVED$ATE^TH!UGUST !CCEPTED$ATE^TH-ARCH

!"342!#4

4HISPAPERDEALSWITHTHEMODELLINGOFFATIGUECRACKPROPAGATIONONACENTREMEMBERBARUSINGADUAL BOUNDARYELEMENTMETHOD4HEEFFECTSOFLIFECYCLETOTHEMULTIPLESITEFATIGUECRACKPROPAGATIONWERE STUDIED!NALYSISOFSTRESSINTENSITYFACTORWASPERFORMEDBYTHEDETERMINISTICAPPROACHUSINGADUAL BOUNDARYELEMENTMETHOD4HEDUALBOUNDARYELEMENTMETHODWASUSEDTOSIMPLIFYTHECRACKMODEL THROUGHTHENUMERICALAPPROACH4HECOMPLEXPROBLEMSHAVEBEENSOLVEDUSINGTHEINFORMATIONFROMA BOUNDARYCONDITIONONLY.EXTTHEINITIALCRACKANDLIFECYCLEOFTHESTRUCTUREHAVEBEENPREDICTEDUSING PROBABILISTICMETHODWHICHIS-ONTE#ARLO4HECRACKSIZEANDFATIGUELIFEWERECOMPUTEDUNTILFAILUREOF THESTRUCTURE4HEFAILUREANALYSISWASPERFORMEDBYALINEARELASTICFRACTUREMECHANICS4HESCENARIOSOF THEFATIGUECRACKPROPAGATIONWEREGIVENBYANINTEGRATIONOFBOTHDUALBOUNDARYELEMENTAND-ONTE

#ARLOMETHOD4HEREFOREFATIGUELIFEOFMULTIPLESITECRACKSTRUCTURECANBEPREDICTED +EYWORDS#RACKPROPAGATIONFATIGUE-ONTE#ARLOBOUNDARYELEMENTMETHOD

!"342!+

+ERTASKERJAINIMEMPERSEMBAHKANPERMODELANPERAMBATANRETAKLESUTERHADAPKOMPONENRASUKTENGAH MENGGUNAKANKAEDAHDWIUNSURSEMPADAN+ESANKITARANHAYATBAGIPERAMBATANRETAKLESUBERBILANG TEMPATDIKAJI!NALISISFAKTORPENUMPUANTEGASANDILAKUKANBERDASARKANPENDEKATANKETENTUANIAITU MENGGUNAKANKAEDAHDWIUNSURSEMPADAN+AEDAHDWIUNSURSEMPADANBERPERANANUNTUKMEMODELKAN RETAKSECARAPENGIRAANBERANGKADENGANPERMASALAHANKOMPLEKSDISELESAIKANSECARAPENGIRAANDISEMPADAN MODEL3ETERUSNYAPENJANAANRETAKAWALDANKITARHAYATSESUATUSTRUKTURDIANGGARKANMELALUIKAEDAH KEBARANGKALIANIAITU-ONTE#ARLO3AIZRETAKDANTEMPOHHAYATLESUSTRUKTURDIKIRASEHINGGAKEGAGALAN STRUKTURBERLAKU!NALISISKEGAGALANDILAKUKANBERDASARKANSIFATMEKANIKPATAHELASTIKLINEAR3ENARIO

PERAMBATANRETAKLESUDIPEROLEHIHASILINTEGRASIANTARAKAEDAHDWIELEMENSEMPADANDENGAN-ONTE

#ARLO/LEHITUHAYATLESURETAKBERBILANGTEMPATBAGISTRUKTURYANGDIKAJIDAPATDIANGGARKAN +ATA+UNCI0ERAMBATANRETAKLESU-ONTE#ARLOKAEDAHUNSURSEMPADAN

).42/$5#4)/.

)NTHEAREAOFFATIGUERELIABILITYANESTIMATIONOF PROBABILITYOFFAILUREISREQUIRED4HISISCAUSEDBY UNCERTAINTIESININITIALCRACKSURFACEROUGHNESS MATERIALPROPERTYAPPLIEDLOADmAWDEFECTSSUCH ASSCRATCHESORWELDDEFECTSFROMMANUFACTURING PROCESS 9ANG ET AL )N OTHER WORDS AS THE CRACK GROWS THE CRACK SIZE HAS A VARIATION ACCORDINGTOTHOSEUNCERTAINTIESANDTHERESIDUAL LIFE OF THE STRUCTURE IS NOT DETERMINISTIC BUT STOCHASTIC&ATIGUECRACKPROPAGATIONISINHERENTLY ARANDOMPROCESSBECAUSEOFTHEINHOMOGENEOUS

OFMATERIALCONNECTEDWITHITSCRYSTALSTRUCTURE ANDWITHVARIATIONSOFCONVECTIVElLMCOEFlCIENT ATTHESTRUCTURESSURFACEDUETOITNONSMOOTHNESS AND OTHER SIMILAR REASONS #HERNIAVSKY 4HE EXPERIMENTAL RESULTS FOR THE FATIGUE CRACK GROWTH UNDER CONSTANT AMPLITUDE LOADING SHOWTHATTHEMATERIALRESISTANCEAGAINSTCRACK PROPAGATION HAS THE INTERSPECIMEN AS WELL AS THE INTRASPECIMEN VARIABILITY ! STOCHASTIC MODEL CONSIDERING BOTH TYPES OF VARIABILITY IS THUS NEEDED FOR THE RATIONAL ASSESSMENT OF FATIGUE CRACK PROPAGATION4HEREFORE THE ANALYSISOFFATIGUECRACKPROPAGATIONSHOULDBE BASED ON THE PROBABILISTIC APPROACH AND THE INSPECTION INTERVAL OR THE REPAIR METHOD MUST BE DETERMINED ACCORDING TO THE POSSIBILITY OF STRUCTURAL FAILURE CONSIDERING THE UNCERTAINTIES MENTIONEDABOVE

4HIS PAPER PRESENTS THE DEVELOPMENT OF AN INSPECTION PROGRAMME FOR THE FATIGUE CRACK PROPAGATION WHICH IS AN ENHANCEMENT OF AN EARLIER PROGRAMME +EBIR ET AL AND THE MAJOR DIFFERENCES BETWEEN THESE TWO PROGRAMMESARESUMMARIZEDBELOW

4HE CRACK PROPAGATION HAS BEEN MODELLED USING "%- PRINCIPAL OF "EASY SOFTWARE WITH THE COMBINATION OF RANDOM FUNCTION OF -ATLABSOFTWARE

4HE LIFE CYCLE OF A CENTRE MEMBER BAR WITH MORETHANONENOTCHHASBEENANALYZEDUSING DUALBOUNDARYELEMENTAND-ONTE#ARLO

"OUNDARY%LEMENT-ETHOD"%-

4WODIMENSIONALNUMERICALSTRESSANALYSISWAS CARRIEDOUTUSINGTHEBOUNDARYELEMENTMETHOD

"%-ISWELLSUITEDFORTHECOMPLEXPROBLEMLIKE CRACK BY MODELING ONLY THE BOUNDARIES OF THE PART)NORDERTODERIVEELEMENTSTIFFNESSMATRIX FORACRACKEDDOMAINAMETHODBASEDONDUAL BOUNDARYELEMENTHASBEENADOPTEDINWHICH REQUIRESTWOEQUATION"OUNDARYDISPLACEMENT AND TRACTION ARE THE FUNDAMENTAL OF CRACK CHARACTERISTIC AND THE RELATION BETWEEN THE CHARACTERISTICSARESHOWNBELOW

WHERE4_IJXX AND5_IJXX IN%QUATION ISA+ELVIN TRACTIONANDDISPLACEMENTFORXPOINTlELDINΓ DOMAIN ! DISTANCE BETWEEN THE SOURCE POINT ANDlELDPOINTISPRESENTEDBY\XnX|\

4HE INTERNAL OR EDGE SURFACES WHICH ARE INCLUDED NO AREA OR VOLUME AND ACROSS WITH THE DISPLACEMENT FIELD HAVE BEEN DEFINED AS MATHEMATICAL DISCONTINUOUS CRACKS &OR SYMMETRIC CRACK PROBLEMS ONLY ONE SIDE OF THE CRACK NEED TO BE MODELLED AND A SINGLE REGIONBOUNDARYELEMENTANALYSISMAYBEUSED (OWEVERASOLUTIONOFGENERALCRACKPROBLEMS CANNOT BE ACHIEVED IN A SINGLEREGION ANALYSIS WITH THE DIRECT APPLICATION OF THE BOUNDARY ELEMENT METHOD BECAUSE COINCIDENCE OF THE CRACK BOUNDARIES THAT GIVE RISE TO A SINGULAR SYSTEM OF ALGEBRAIC EQUATIONS4HE EQUATION FOR A POINT LOCATED AT ONE OF THE BOUNDARIES OF THE CRACK IS IDENTICAL TO THOSE EQUATIONS FOR THE POINT WITH THE SAME COORDINATES BUT ON THEOPPOSITESURFACE4HISISBECAUSETHESAME INTEGRALEQUATIONWASCOLLOCATEDWITHTHESAME INTEGRATIONPATHATBOTHCOINCIDENTPOINTS 4HE,ANGRARIANCONTINUOUSORDISCONTINUOUS BOUNDARY ELEMENTS ARE USED TO SATISFY #AUCHY PRINCIPLE VALUE INTEGRAL WHICH IS DElNED AS A DISPLACEMENTEQUATION4HE(ADAMARDPRINCIPLE VALUE INTEGRAL TRANSFORMS THE DISCONTINUOUS ELEMENT TO THE CONTINUITY REQUIREMENT FOR THE lNITEPART INTEGRAL4HE DISCONTINUOUS ELEMENT IS DElNED FROM ALL NODES WHICH IS AN INTERNAL u x_i

_°

_°

' '

' '

POINT4RACTION EQUATION IS THEN DElNED FROM THE (ADAMARD PRINCIPAL VALUE INTEGRAL 4HE PRINCIPALVALUEINTEGRALHASBEENPERFORMEDTO IMPOSETHEDUALBOUNDARYINTEGRALEQUATION"Y THE CHANGING OF THE DISCONTINUOUS QUADRATIC ELEMENTS CRACK MODELLING IS PRESENTED IN THE

*INTEGRALFUNCTIONASGIVENBY J

WHERESISANARBITRARYCONTOURSURROUNDINGTHE CRACKTIP7ISTHESTRAINENERGYDENSITYGIVENBY σ_IJε_IJWHEREσ_IJANDε_IJARETHESTRESSANDSTRAIN TENSORSRESPECTIVELYT_JARETRACTIONCOMPONENTS GIVENBYσ_IJN_IWHERENIARETHECOMPONENTSOF THE UNIT OUTWARD NORMAL TO THE CONTOUR PATH 4HERELATIONSHIPBETWEENTHE*INTEGRALANDTHE lRST AND SECOND LEVEL STRESS INTENSITY FACTOR+₎ AND+₎₎ISGIVENBY

J K K

E

I II

^{2 2} '

WHERE % IS THE ELASTICITY MODULUS FOR PLANE STRESS CONDITIONS AND %@ %νV FOR PLANE STRAINCONDITIONS4HEALGORITHMOFTHEBOUNDARY ELEMENTMETHODWASSHOWNBELOW

s #ARRYOUTADUALBOUNDARYELEMENTMETHOD STRESS ANALYSIS OF THE STRUCTURE HAS BEEN DONE

O 3TRESS INTENSITY FACTORS HAVE BEEN COMPUTEDBYTHE*INTEGRALTECHNIQUE #OMPUTE THE DIRECTION OF THE CRACK

EXTENSIONINCREMENT

O %XTENDTHECRACKONEINCREMENTALONGTHE DIRECTIONCOMPUTEDINTHEPREVIOUSSTEP s 2EPEAT ALL THE ABOVE STEPS SEQUENTIALLY

UNTILASPECIlEDNUMBEROFCRACKEXTENSION INCREMENTSWEREACHIEVED

,!7/&&!4)'5%#2!#+02/0!'!4)/.

)N 0ARIS AND %RDOGAN CREATED A 0ARIS LAW EQUATIONFORCALCULATINGFATIGUECRACKPROPAGATION RATEda/dNASGIVENIN%QUATION

∆++_MAXn+_MINISTHERANGEOFSTRESSCONCENTRATION FACTORAND#ANDMARETHEMATERIALPROPERTIES 4HESTRESSCONCENTRATIONFACTORISONEOFTHE PARAMETERSTHATARECONSIDEREDINLINEARELASTIC FRACTURE MECHANIC4HE THEORY ONLY ACCEPTABLE

FORTHESITUATIONWHENTHEREISNOYIELDOCCURSAT THECRACKTIP4HEREFORE%QUATION CANBEUSED FORHIGHCYCLICFATIGUECASES&ORMANETAL TRIEDTOMODIFYTHE%QUATION ASITISNOTINCLUDE THESTRESSCONCENTRATIONRATIO2+_MIN+_MAXANDTHE FRACTURESTRENGTHK_C&ROMTHEDElNITIONOF∆+

+_MAXn2 AND+_MAX+_CTHEBOUNDARYCONDITION FORTHECRACKPROPAGATIONRATEIS

lim

$K R K_c

d dN

m d

1

3UBSTITUTING%QUATION IN%QUATION GIVES

da dN

C K

R K K

m

c

( )

( )

$

$

1

&ROM%QUATION &ORMAN FOUNDTHAT THE M VALUE FOR ALUMINIUM ALLOY 4 AND 4WAS%QUATION ISKNOWNAS&ORMAN EQUATION 3TARTING FROM THE &ORMAN EQUATION ANDCONSIDERINGTHECRACKWILLNOTPROPAGATEIF THE∆+VALUEBELOW∆+_TH&IGURE AGROWTHLAW WASINTRODUCEDASIN%QUATION TOCALCULATETHE FATIGUECRACKPROPAGATIONRATE

da

dN C K K

K K ^th C

c

¥

§¦

´

¶µ

$ $

max

'

2

4HISMODELISVALIDFORASOFTMETALTHATUNDER BOTH THE FIX AND RANDOM LOADING AMPLITUDE WHICH#ISAPPROACHINGTOXMMCYCLIC #RACK BEHAVIOUR IS DIAGNOSED BY THE CRACK GROWTH PROCESS WHICH ARE THE FUNCTION OF AN APPLIEDLOADANDGEOMETRYOFTHECRACKSTRUCTURE 4HE CRACK GROWTH PROCESS IS FORMED BY CRACK EXTENSIONBEHAVIOUR

#RACKBEHAVIOURISPREDICTEDBASEDONSTRESS INTENSITY FACTOR4HIS ANALYSIS ALSO CALL DAMAGE TOLERANCEANALYSISISDEVELOPEDBASEDONLINEAR ELASTICFRACTUREMECHANICSTHEORY

,).%!2%,!34)#&2!#452%-%#(!.)#3

&RACTURE MECHANICS SEEK TO ESTABLISH THE LOCAL STRESSANDTHESTRAINlELDSAROUNDACRACKTIPIN TERMOFGLOBALPARAMETERSSUCHASTHELOADING ANDTHEGEOMETRYOFTHESTRUCTURE,INEARELASTIC CHARACTERISTIC IS USED IN CRACK PATH MODEL4HE LONG CRACKS HAVE BEEN MODELLED SUBJECTED TO CONSTANT AMPLITUDE CYCLIC LOADING &OR LINEAR ELASTICSOLUTIONTHESTRESSINTHEVICINITYOFTHE CRACKISEVALUATEDBYSTRESSINTENSITYFACTORS

Fatigue crack grwth rate, da/dN

Stress intensity range, ∆K Stage I

Region

Stage II Region

Failure

Stage III Region

m K_th

&)'52%3CHEMEDIAGRAMOFSHORTANDLONGFATIGUECRACKPROPAGATION$HARANI 3.CURVEASSUMEDTHEFATIGUELIFEAVERAGE

._IATCERTAINPOINTFOR4ALUMINIUMALLOY ISILLUSTRATEDASBELOW

Ni S S

IQF S

m

p

¥

§¦

´

¶µ 10^{5 lim}

lim

WHERE 0)1&-0A3_LIM-0AAND 3_M =THEAVERAGESTRESS

)NLINEARELASTICFRACTUREMECHANICSTHEREARE SEVERAL MIXEDMODE PROPAGATION CRITERIA /NE OF THEM IS THE STRESS INTENSITY FACTOR +_I WHICH CONTROLSTHENEARTIPSTRESSlELD-AGNITUDEOF THECRACKTIPSTRESSESISGOVERNEDBYTHESTRESS INTENSITYFACTORS+₎AND+₎₎ASSHOWNIN%QUATION )TISALSOOBSERVEDTHATTHEDISPLACEMENTSARE CONTROLLEDBYTHESTRESSINTENSITYFACTORSASSHOWN IN%QUATION 4HEDISTRIBUTIONOFTHESTRESSES ISGOVERNEDBYTHEPOSITIONRELATIVETOTHECRACK TIPGIVENBYRANDθ(EREμISTHESHEARMODULUS

OFMETALALLOYMATERIAL

4HETWODIMENSIONALNUMERICALSTRESSANALYSIS HASBEENCARRIEDOUTUSINGTHEBOUNDARYELEMENT METHOD"%-ISWELLSUITEDFORCOMPLEXMODELLING CASESUCHASCRACKPROBLEMSBYMODELLINGONLY THEBOUNDARIES

#2!#+-/$%,).'342!4%'9

4HE DOMAIN REGION HAS BEEN TREATED AS DUAL BOUNDARYELEMENTBY"OUNDARY%LEMENT3YSTEM

"%!39 SOFTWARE )T IS NECESSARY TO CALCULATE THERELATEDSTIFFNESSMATRIXANDEFFECTIVESTRESS INTENSITYFACTOR+EFFBYMEANSOF$UAL"OUNDARY

%LEMENTS-ETHOD$"%- 4HECRACKMODELLING STRATEGYSHOWNBYALGORITHMBELOW

s #ARRYOUTADUALBOUNDARYELEMENTMETHOD FORSTRESSANALYSISOFTHESTRUCTURE

s #OMPUTE THE EFFECTIVE STRESS INTENSITY FACTORS+_EFFWITHTHE*INTEGRALTECHNIQUE s #OMPUTE THE DIRECTION OF THE CRACK

S P

Q Q Q

R Q

11 2 2 1 2

2 3

2 2 2 2

¥

§¦

´

¶µ K

r

K r

I cos sin sin II sin ¥

§¦

´

¶µ

cos cosQ Q ( ^/ ) 2

3 2

O r1 2

M P

Q Q Q

₄¹_{u 2}^{r K}I^¥_§¦

3 2

¥

§¦

´

¶µ

¨

ª© ·

¹¸ sin Q

EXTENSIONINCREMENT

s %XTENTTHECRACKONEINCREMENTALONGTHE DIRECTIONCOMPUTEDINTHEPREVIOUSSTEP s 2EPEAT ALL THE ABOVE STEPS SEQUENTIALLY

UNTILASPECIlEDNUMBEROFCRACKEXTENSION INCREMENTSHAVEBEENACHIEVED

4HE BOUNDARY STIFFNESS MATRIX AND+_EFF AFTER CONDENSATION HAVE BEEN INSERTED INTO CRACK INITIALANDPROPAGATIONROUTINEUSING-ONTE#ARLO METHOD WHICH IS PROVIDED BY -!4,!" SOURCE CODE 5SING THE 3. CURVE AT INCLUDES THE DETERMINISTICAPPROACH

"Y RUNNING A -ONTE #ARLO METHOD USING -!4,!" PROGRAM IT POSSIBLE TO CALCULATE THE CYCLENUMBERFOREACHOFTHEPROPAGATIONAND THECRACKLENGTH4HEINITIALPOINTALSOINDICATED BY RANDOM PROCESS AS SHOWN IN &IGURE 4HE

Monte Carlo method

Sampling Crack Length Sampling Number of Cycle Dedicate Initial Point

&)'52%2ANDOMPARAMETERFORFATIGUECRACKPROPAGATION

&)'52%3CHEMATICDIAGRAMOFPLATEHOLES 900

375

Notch 1 ...Notch 24

Constraint Load

25 25

Holdhole Notch No

2 4 6 8 10 12 14 16 18 20 22 24

MODIlED DATA lLES IN "%!39 HAS BEEN RUN TO UPDATETHECRACKPARAMETER

.5-%2)#!,2%35,43/&0,!4%(/,%3 )N ORDER TO VALIDATE THE GLOBAL PROBABILISTIC APPROACHTHERESULTSHAVEBEENCOMPAREDWITH THEFATIGUETESTONAPLANEPLATEWITHFREEHOLES THATWASCONDUCTEDBY+EBIR(ETAL AT

!EROSPATIALE-ATRALABORATORYIN3URESNES&RANCE ASSHOWNIN&IGURE

4HETESTINGWASCONDUCTEDUSINGALUMINUM ALLOY 4 SHEETS WITH A THICKNESS OF MM4HE LOAD HAS BEEN APPLIED ON TRANSVERSAL DIRECTION4HE9OUNGMODULUSOFTHESAMPLEWAS '0A4HEINITIALSTRUCTUREHASBEENDISCREET WITH ELEMENTS IN ONE ZONE WITH DEGREES OF FREEDOM )T HAS INTERNAL POINTS

Cycle of propagation, Np

Cycle of initiation, Ni

×10³

Monte Carlo-Beasy Test (Kebir H. et al., 2001) Deterministic

100 80 60 40 20 0

0 100 200 300 400

&)'52%&ATIGUEPREDICTIONLIFE PATCHINTHEMODEL4HENUMERICALRESULTSHAVEA

GOODCOMPROMISEBETWEENTHETESTRESULTS4HE TOTAL NUMBERS OF CYCLES WITH THE PROBABILISTIC APPROACHARECLOSELYSIMILARTOTHETESTEXPRESSED IN&IGURE

4HE SYNTHESIS OF THE PROBABILISTIC RESULTS IS EXPRESSED IN &IGURE )N THE DETERMINISTIC APPROACH THE PROPAGATION PHASE WAS SHORT WITHIS˜CYCLE)TWASBECAUSEALLTHECRACKS ASSUMED BEGIN AT THE SAME TIME SINCE ALL THE SITESWEREUNDERGOINGTHESAMESTRESSLEVEL3O THEPROBABILISTICAPPROACHHASANADVANTAGEOF GIVING THE VIEW OF INITIAL CRACK PROPAGATION ! LARGECRACKSIZEHASBEENDOMINATEDTHEFAILURE PROBABILITY AT THE BEGINNING OF THE FAILURE

&)'52%,IFECYCLEOFFATIGUECRACKPROPAGATIONBYITERATIONS Cycle Crack propagation at notch 1 to notch 24

Notch number

1 24

0.29 × 10⁵ 0.50 × 10⁵ 0.73 × 10⁵ 1.25 × 10⁵ 1.38 × 10⁵ 1.51 × 10⁵ 1.72 × 10⁵ 1.98 × 10⁵

PROCESS)NALONGDURATIONTIMETHESMALLCRACKS SIZEMAYHAVETHEMOSTDOMINANTEFFECTONTHE FAILUREPROBABILITY

4ABLESHOWSAMAXIMUMCRACKLENGTH MM BEFORE THE COMPONENT IS FAILURE !T THIS

MOMENTTHELIFECYCLEONLYXCYCLESAS PRESENTSIN^THITERATION4HISHAPPENEDBECAUSEA CRACKNOTCHHADENOUGHENERGYTOPROPAGATE ANINITIALCRACK4HECRACKGROWTHVERYFASTAND VERIlESTHEHIGHPROPAGATIONRATE(OWEVERTHE FAILUREOFTHESAMPLEHASBEENNOTHAPPENEDYET UNTILTHELIFECYCLEREACHEDATXCYCLES 4HE NOTCH IS RANDOMLY PROPAGATED THE INITIAL CRACK APPROPRIATE WITH THE INCREASE OF STRESS INTENSITYFACTORVALUE4HESTRESSINTENSITYFACTOR

VALUEISCONSTANTLYINCREASEDFORAFEWITERATIONS UNTILITISSLOWLYTRENDEDTOACHIEVEAMAXIMUM VALUE&ORTHEFOURTEENHOLESPLATENOTCHES ANDHAVEBEENCHOSENFORPROPAGATESAN INITIALCRACKASSHOWNIN&IGURE4HEINCREASING WASCONTINUINGFORCERTAINITERATIONS!FTERTHAT THECRACKHASBEENRANDOMLYPROPAGATEDATOTHER NOTCH WHICH HAD LOWER STRESS INTENSITY FACTOR )NTHISSCENARIOTHENOTCHHAVE BEENRANDOMLYPROPAGATEDTHECRACK4HECRACK HAS BEEN CONTINUING PROPAGATE FOR A CERTAIN ITERATIONSTOGETCLOSEWITHTHEMAXIMUMSTRESS INTENSITYFACTORATTHATTIME4HEINCREASEDOFTHE LIFECYCLEWASCONTINUINGTHECRACKPROPAGATION AT HIGH PROBABILITY LOCATION RANDOMLY !T THIS

MOMENT THE CRACK PROPAGATION CAN MAKE THE SAMPLEFAIL.OTCHWASHAVINGACATASTROPHIC FAILUREWHENITSEFFECTIVESTRESSINTENSITYFACTOR REACHED THE VALUE OF -0A M4HE HIGHPOTENTIALENERGYHASBEENASSEMBLESATA LOWSTRESSINTENSITYFACTORNOTCHANDCATASTROPHIC FAILUREWASOCCURREDBECAUSEOFTHEHIGHGROW CRACKPROPAGATIONRATE

4HERESULTOFTHECRACKPROPAGATIONBASEON THE COMPONENT LIFE CYCLE WAS SHOWN THAT THE CRACKPROPAGATESINTHREEPHASES4HElRSTPHASE STARTEDATCYCLESANDlNISHEDAT CYCLES!TTHISPHASEEACHOFTHEITERATIONCAUSES OFCMCRACKLENGTHWHICHISCONSIDEREDAS AMINORCRACK

4!",%2ESULTSOFFATIGUECRACKPROPAGATION

)TERATION #RACK,ENGTH #YCLE._4OTAL 0OINT.O

&AILURE

&)'52%'RAPHOFEFFECTIVESTRESSINTENSITYFACTORVERSUSNOTCHESNUMBERFOR^STTHAND^THITERATION 1 st Iteration

7 th Iteration 13th Iteration

0 5 10 15 20 25 1000000

100000 10000 1000 100 10

Notch number Effective stress intensity factors, Keff (MPa m1/2)

Number of sample, n

Standard deviation of life cycle, Ó

0 100 200 300 400 500

× 10² 50

40

30

20

10

0.0

&)'52%3TANDARDDEVIATIONOFLIFECYCLEVERSUSNUMBEROFSAMPLEFOR^THITERATION

&)'52%-EANLIFECYCLEVERSUSNUMBEROFSAMPLEFOR^THITERATION

×10⁵ 3.0

2.5

2.0

1.5

1.0

0.5

0.0

0 100 200 300 400 500 Number of sample, n

Mean of life cycle, NTotal

4HE SECOND PHASE WAS STARTED AT CYCLES"ETWEENANDCYCLESA MAJORCRACKWASPROPAGATED4HESIZEISCM FOREACHLIFECYCLESITERATION

&ORTHETHIRDPHASEMINORCRACKPROPAGATION ISOCCURREDASINlRSTPHASE(OWEVERTHETHIRD PHASEWASTOOSHORTWHICHONLYCYCLES 4HE COMPONENT TOTALLY FAILED AFTER CYCLES

&IGUREAND&IGURESHOWAMEANLIFEANDA STANDARDDEVIATIONPREDICTIONFORA^THITERATION BYTHEEFFECTOFNUMBEROFSAMPLES)TISSEENTHAT THENUMBEROFSAMPLESINmUENCESTHEFATIGUELIFE CYCLE4HERESULTSARECONSTANTWHENTHENUMBER

&)'52%!PHOTOGRAPHOFACENTREMEMBERBAR

The notches locate at each engine holder.

&)'52%4OPVIEWOFTHECENTREMEMBERBAR

&)'52%"OUNDARYELEMENTSOFTHECARCENTREMEMBERBAR

Load Load

Notch 4 Notch 3 Notch 2 Notch 1

OF SAMPLES IS OVER 3O THE -ONTE #ARLO

"%- STATISTICAL TEST IS ONLY VALID WITH THE HIGH NUMBEROFSAMPLES4HEMEANLIFEANDASTANDARD DEVIATIONPREDICTIONHAVEGIVENTHESAMETREND RESULTLIKETHE^THITERATION

.5-%2)#!, 2%35,43 /& #%.4%2 -%-"%2

"!2

4HEBOUNDARYELEMENTMETHODHASBEENAPPLIED TOACENTREMEMBERBAR&OURHOLESWITHNOTCHES HAVEBEENMODELLEDASSHOWNIN&IGUREAND 4HELOCATIONOFTHENOTCHESBASEDONTHEHIGH FORCECONCENTRATIONANDHIGHPROBABILITYOFTHE

&)'52%#OMPONENTDEFORMATIONATITERATIONS

&)'52%#RACKATA NOTCHB NOTCHC NOTCHANDD NOTCHE #OMPONENTDEFORMATIONAT ITERATIONS

E

D C

B A

&)'52%&AILUREOFTHECOMPONENTAFTERITERATIONS CRACKPROPAGATION4HECENTREMEMBERBARWAS

MADEBYSTEELWITH9OUNG-ODULUSOF'0A 4HE FATIGUE LOAD IS -0A BASED ON THE ENGINELOAD

4HE ELEMENTS HAVE BEEN GENERATED ONLY AT THE BOUNDARY AS SHOWN IN &IGURE 7ITH THE LOADGIVENTHEELEMENTSOFTHECOMPONENTHAVE DISPLACEDASSHOWNIN&IGURE4HECRACKSHAVE

BEEN PROPAGATED AT ALL NOTCHES AS SHOWN IN

&IGURE&INALLYTHEFAILUREOCCURREDAT CYCLESAFTERITERATIONSASSHOWNIN&IGURE

#/.#,53)/.

!NOVERALLASSESSMENTMETHODPROPOSEDINTHIS PAPER WAS DEVELOPED IN ORDER TO VALIDATE THE FATIGUECRACKPROPAGATIONWITHTHEPROBABILISTIC

2%&%2%.#%3

!NDERSON - 2 &ATIGUE DAMAGE ANALYSIS BY USE OF CYCLIC STRAIN APPROACH *OURNALOF3HIP 4ECHNOLOGY2ESEARCH

!PICELLA ! #ITARELLA 2 %SPOSITO 2 -3$

RESIDUAL STRENGTH ASSESSMENT FOR A CRACKED JOINT5+ #OMPUTATIONAL -ECHANICS "%!39 0UBLICATIONS

"REBBIA#!$OMINGUEZ"OUNDARY%LEMENTS n AN )NTRODUCTORY #OURSE #OMPUTATIONAL -ECHANICS0UBLICATIONS3OUTHAMPTON5+

#ALÖ # #ITARELLA 2 3OPRANO ! &%-"%-

#OUPLED -ETHODOLOGY FOR #RACKED 3TIFFENED 0ANELS.APOLI)TALY

#HERNIAVSKY !/ 0ROBABILISTIC !PPROACH TO #ALCULATION OF +INETICS OF #RACK -ESHES

$YNAMICS 3TRENGTH 7EAR2ESISTANCE OF -ACHINES

$HARANI &ATIGUE CRACK GROWTH 5NIVERSITY OF -ISSOURI2OLLA

METHODTHROUGHTHE-ONTE#ARLO4HEMODELING PROCESSWASUSINGTHEDUALBOUNDARYELEMENT METHOD4HE RESULTS OF THE BOUNDARY ELEMENT METHOD AND -ONTE #ARLO ANALYSIS SHOW THAT THE LIFE CYCLE OF STRUCTURE CAN BE PREDICTED AND OBTAINED IN GOOD AGREEMENT WITH THE EXPERIMENT RESULTS4HE RESULTS OBTAINED PROVE

&ORMAN 2' 3HIVAKUMAR6 .EWMAN *#

&ATIGUECRACKGROWTHCOMPUTERPROGRAMv.!3!

&,!'2/v 3ECOND %DITION 4EXAS .ATIONAL

!ERONAUTICSAND3PACE!DMINISTRATION0RESS +EBIR(2OELANDT*-'AUDIN*-ONTE#ARLO

SIMULATIONS OF LIFE EXPECTANCY USING THE DUAL BOUNDARYELEMENTMETHOD%NGINEERING&RACTURE -ECHANICS

0ARIS 0 # %RDOGAN & #RITICAL !NALYSIS OF 0ROPAGATION,AWS*OURNALOF"ASIC%NGINEERING

9ANG*.3ALIVAR'#3CHWARTZ"*3TATISTICAL MODEL FOR PREDICTION OF FATIGUE CRACK GROWTH UNDERABLOCKTYPESPECTRUMLOADING%NGINEERING

&RACTURE-ECHANICS

4ONG 9 # ,ITERATURE REVIEW ON AIRCRAFT STRUCTURALRISKANDRELIABILITYANALYSIS!USTRALIA

$34/ !ERONAUTICAL AND -ARITIME 2ESEARCH ,ABORATORY

THAT THE COMPUTER SIMULATION CAN BE USED TO PREDICTFATIGUECRACKPROPAGATION4HEPROPOSED ALGORITHMCANBEUSEDFORAGUIDELINETOHAVEA RISK AND RELIABILITY ANALYSIS AND LIFE EXPECTANCY OFTHESTRUCTURE