NASA-TM-107147

IqGO lalO

NASA
Technical Memorandum 107147

Army Research Laboratory

Technical Report ARL-TR-957

Gear Crack Propagation Investigations

David G. Lewicki
VehiclePropulsion Directorate
U.S.Army Research Laboratory
Lewis Research Center
Cleveland, Ohio
and
Roberto Ballarini
Case WesternReserve University
Cleveland, Ohio

: : -'

- ,.. --,_-_
r:.,..

:
_

_i " , ,'_
, eli

'1./'_ !('5
r._'-,- .... p,_ bCi'_
,-r_ -.'r_
.... [_' l']_2_.iC./;.',,,,_,l-i
t LI_
........

-_-._.

;"

_,;_--_,_

Prepared for
An Integrated Monitoring, Diagnostics, & Failure Prevention
Technology Showcase
sponsored by the Society for Machinery Failure Prevention Technology
Mobile, Alabama, April 22-26, 1996
,

U.S. ARMY

NationalAeronauticsand
SpaceAdministration

RESEARCHLABORATORY

3
r

GEAR CRACKPROPAGATIONINVESTIGATIONS

DavidG.Lewicki
VehiclePropulsionDirectorate
U.S.Army ResearchLaboratory
NASALewisResearchCenter
Cleveland,Ohio44135
RobertoBallarini
Departmentof CivilEngineeringand
Mechanical& AerospaceEngineering
CaseWesternReserveUniversity
Cleveland,Ohio44106

Abstract: Analyticaland experimentalstudieswere performedto investigatethe effectof gearrim thicknesson

crack propagationlife. The FRANC (FRactureANalysisCode) computerprogramwas used to simulatecrack
propagation.The FRANCprogramusedprinciplesof linearelasticfracturemechanics,finiteelement modeling,
and a uniquere-meshingschemeto determinecrack tip stressdistributions,estimatestress intensityfactors, and
modelcrackpropagation.Variousfatiguecrackgrowthmodelswereusedto estimatecrackpropagationlife based
on the calculated stress intensity factors. Experimentaltests were performedin a gear fatigue rig to validate
predictedcrack propagationresults. Testgears were installedwith specialcrack propagationgages in the tooth
fillet region to measure bendingfatiguecrack growth.Goodcorrelationbetweenpredictedand measuredcrack
growth was achievedwhen the fatigue crack closureconceptwas introducedinto the analysis.As the gear rim
thicknessdecreased,the compressivecyclic stressin the geartooth fillet regionincreased.This retardedcrack
growthand increasedthe numberof crack propagationcyclesto failure.

KeyWords: Crack propagation;finiteelementmodeling;fracturemechanics;gears;rims.

Introduction: A commondesigngoalforgearsin helicopteror turboproppowertransmissionsis reducedweight.

To help meet this goal, somegear designsuse thin rims.Rims that are too thin, however,may lead to bending
fatigue problemsand cracks. The most common methodsof gear design and analysis are based on standards
published by the AmericanGear ManufacturersAssociation.Includedin the standardsare rating formulas for
geartooth bendingto preventcrack initiation[1]. Thesestandardscanincludetheeffectofrim thicknesson tooth
bending fatigue [2]. The standards,however,do not give any indicationof the crack propagationpath or the
remaininglife once a crack has started.Fracturemechanicshas developedintoa useful disciplinefor predicting
strength and life of crackedstructures.

Ahmad and Loo [3] applied fracture mechanics to gear teeth to illustrate the procedure and estimate crack
propagationdirection.HondaandConway[4]also appliedfracturemechanicsto simulatetoothcrackpropagation,
compute thresholdloads, and calculatetooth life. Flaskerand Jezernik [5] applied fracture mechanicsto gear
teeth to estimate stress intensity factorsand gear life. Researchersat TohokuUniversityin Japan performed a
seriesofanalysesand experimentsto determinetheeffectofresidualstressoncrackinitiationandpropagation[6,7].
Also, Daniewicz,et al. [8] developeda comprehensive,self-containedanalysispackageto refinethe spur gear
1

bending fatigue theory using fracture mechanics.Lastly, Flasker and Pehan [9] described their method for
calculatingcrack propagationin gear teeth usingfracturemechanics.Much of the workof the abovereferences
considered only an initial crack and propagationpaths were not considered. Many of the references that did
considercrackpropagationassumedthepropagationoccurredin a straightpath.Inaddition,experimentalvalidation
of the cited analyseswassparse.Finally,no workusing fracturemechanicswas performedfor thin-rimgears.
The objectiveof this studywas to determinethe effectof gearrim thicknesson crack propagationlife.From an
extensivestudy [10],linearelasticfracturemechanicswasusedto analyzegeartooth bendingfatiguein standard
and thin-rimgears.Finite elementcomputerprogramswere usedto determinestressdistributions,estimatestress
intensityfactors, and modelcrack propagation.Variousfatiguecrack growthmodelswere used to estimatecrack
propagationlife. Experimentaltests were performedto validatepredictedcrack propagationresults.
Fatigue Crack Growth: Many machine elements,such as gear teeth,are cyclicly loadedin application.The
overall fatiguelife of suchcomponentsmaybe representedby threedistinctphases: 1)crack initiation,2) crack
propagation,and 3) finalfailure.Oncecrack initiationhas occurred,fracturemechanicsmaybe used to estimate
crack propagationfatiguegrowthrate and time to finalfailure.
The most universallyusedmethodto calculatecrackpropagationfatiguecrack growthwas postulatedby Paris
and Erdogan[11]. ConsideredwerepurelymodeI loadedspecimenssubjectedto cyclicload. Furtherconsidered
was unstablecrack growthsuch that the stressintensityfactor grewwith increasingcrack size. Parispostulated
that therate of crack growthwith respectto numberof stresscycleswasa logarithmicrelationshipwith the stress
intensityfactor range as
da
m = C(aK)n
(i)
dN
wheredo is the changein cracklengthfor dNnumberof stresscycles,AX is the rangeof themodeI stress
intensity factor at a given time, and C and n are material constants. The material constants, C and n, must be

determined
by someexperimental
means.
Furtherresearch
of fatigue crack growth hasshownthatthereexiststhreeimportant factors not considered in the
Parismodel.Firstwastheeffectof loadratio,R,oncrackgrowth(R = minimumcyclicload/ maximumcyclic
load).Secondwastheinstabilityof crackgrowthobservedwhenthestressintensityfactorrangeapproached
the
material'sfracturetoughness index, Kto Thirdwasthepresence
of a stress intensity threshold factor, Z_th. The
stress intensity threshold factor isthe highest stress intensity factor in whichnocrackgrowth wouldoccur.The
Collipriestcrackgrowthmodel[12]accounts
for theseeffectswhere

. tanh_l/.

aN

_,(1- R)KIcAKt_h

_,AKth )

L
)]
I ln(- R)K'c

C(KlcZth).exp
[

da

(2)

In addressingapplicationsto gears,Inoue, et al. [7] describesfatiguecrack growthof gearbending fatiguetests.

Here, crack growth equationswere derivedas a function of crack depth through a gear tooth. The expression
derivedfor crack growthrate damn, as a functionof stressintensityrange,zlK,was

da

[ -_n
(I

l'_ _--'_

) (!_r(''q- AKthrl)

_/c_

frLur_th<_r_<ZkKC

frM(c<AK<AKIc

- _-_)

where the parameters Kjc, o:, Z_c, Z_th, 71,and _.were all estimated as a function of tooth hardness [7].

(3)

CrackPropagation Simulation: The analysisof the currentstudyusedthe FRANC(FRactureANalysisCode)

computerprogramdescribedbyWawrzynek[13].FRANCis a generalpurposefiniteelementcode for the static
analysisof cracked structures.FRANC is designedfor two-dimensionalproblemsand is capable of analyzing
plane strain,plane stress,or axi-symmetricproblems.
Among the varietyof capabilities, a unique feature of FRANC is the ability to model a crack in a structure.
FRANC uses a methodcalled "delete and fill" to accomplishthis. To illustrate,the user would first define an
initial crack byidentifyingthe nodeof thecrack mouthand coordinatesof thecrack tip.FRANC willthendelete
the elementsin the vicinityof thecracktip. FRANCwillnextinserta rosetteof quarter-point,six-nodetriangular
elementsaroundthe cracktip to modeltheinversesquare-rootstresssingularity[14,15].Finally,FRANCwillfill
the remaining area betweenthe rosette and original mesh with conventionalsix-nodetriangularelements. The
user canthenrunthe finiteelementequationsolver to determinenodaldisplacements,forces,stresses,and strains.
Afurtheruniquefeatureof FRANCis the automaticcrack propagationcapability.Afteran initialcrackis inserted
in a mesh,FRANCmodelsa propagatedcrack asa numberof straightline segments.Foreach segment,FRANC
models the crack tip using a rosetteof quarter-pointelements.FRANCthensolvesthe finiteelement equations,
calculatesthe stress intensity factors, and calculatesthe crack propagationangle. After the crack propagation
angle is determined,FRANC thenplaces the new crack tip at the calculatedangle and at a user-definedcrack
incrementlength.Themodelis thenre-meshedusing the"deleteand fill"methoddescribedabove.Theprocedure
is repeateda specificnumberof times as specifiedby the user.In the current study,the stress intensityfactors
were determinedfromthe calculatednodal displacementsusing the displacementcorrelationmethod [16].The
methodof Erdoganand Sih [17]was usedin the currentstudy to determinethe crack propagationangle.
Once the stress intensity factorsare determinedfor each segment,the predictednumber of crack propagation
cycles can be estimatedusing the fatiguecrack growthmodels.Regardlessof the modelused,the crack growth
rates,daldN, were of the form
da
= g(M()
(4)
where g(AK) is givenby Eq. (1) for the Parisrelationship,Eq. (2) for the Collipriestrelationship,or Eq. (3) for
Inoue's method.Thepredictednumberof crackpropagationcyclesfor the ithcracksegment,Ni,wasestimatedby
Ni = ai - ai'_l+

g(tai)

Ni. 1

(5)

where ai was the crack length of the ithsegment,ai.1was the crack length of the (i-1)thsegment,Ni-! was the
number of cycles of the (i-1)thsegment,and g(z5_i) was the average crack growth rate of the ith and (i-1)th
segments.Notethat al was the initialcrack length,N1=0,and i variedfrom 2 to the total numberof segments.

"

Gear Finite Element Modeling: Basic gear tooth geometrydata was input to a tooth coordinategeneration
computerprogram.The toothcoordinategeneratorprogramusedthe methodofHefeng,et al. [18] to determine
the tooth coordinates.The output was tooth coordinateand rim coordinatedata which defineda single-tooth
sectorof a gear.This outputwasusedby a commercialavailablepre-and post-processingfiniteelement analysis
softwarepackage[19].This packagecreatedthe finiteelementmeshof the completegear.FRANCthenusedthis
meshand performedcrack propagationsimulations.
Figure 1showsa samplefiniteelementmesh of an uncrackedgear.The toothgeometryusedmodeled thatof the
test gears of the NASALewis Spur Gear FatigueRig (describedin the following section).The analysisused
8-node, plane stress,quadrilateralfiniteelements. The mesh was refinedin the region of the loaded tooth for
improvedaccuracy.Themodelof Fig. 1 had2353elementsand 7295nodes.Materialpropertiesusedwerethat of

AISI 9310steel.The tooth load wasplaced at the highestpoint of singletooth contact.For boundaryconditions,
fourhub nodeswerefixed.In addition,gearswithvariousrimthicknessesweremodeled.Theparameterdescribing
the rim thicknesswas the backup ratio,mB,where
b
mB = -(6)
h
where b was the rim thickness, and h wasthe tooth wholedepth. Gearswithvariousbackup ratios weremodeled
by incorporatingslots in the model.All casesused the samefinite elementmeshfor the loadedtooth.

TestFacility: Crackpropagationexperimentswereperformedin theNASALewisSpurGearFatigueRig (Fig.2).

The test standoperated on a torque-regenerativeprinciplein which torquewascirculatedin a loop of test gears
and slave gears.Oil pressure was suppliedto loadvanesin one slavegear whichdisplacedthe gearwith respect
to its shaft.This produceda torqueon the testgears,slavegears,andconnectingshaftsproportionalto the amount
of appliedoil pressure.A 19kW (25-hp),variable-speedmotorprovidedspeedto the drive shaftusing a belt and
pulley.The lubricantused for the gears,bearings,and loadingsystemwas a syntheticparaffinicoil. The testgear
lubricantwas filtered througha 5-micronfiberglassfilter.
Test Gears: The test gears were 28-tooth,8-pitch,20 pressureangle externalspur gears with a face widthof
6.35 mm (0.25in.). Theteeth had involuteprofileswithlineartip reliefstartingatthe highestpointof singletooth
contact and endingat the tooth tip at an amountof 0.013mm (0.0005in.).All test gearsused in the experiments
were fabricatedand machined from a singlebatch of material.The testgear materialwas consumable-electrode
vacuum-meltedAISI 9310 steel.The gears werecase-carburizedand ground.The teeth were hardenedto a case
hardness of Re61 and a core hardness of Re38. The effectivecase depth (depth at a hardnessof Re 50) was
0.81 mm (0.032 in.). Two differenttest gear designswereconsidered.The first was a thick-rimmedgearwith a
backup ratioof mB=3.3(Fig.3(a)).The secondwasa thin-rimmedgearwhich incorporatedslots(Fig. 3(b)).The
backup ratio of the thin-rimmedgear wasmn=0.3.
It wasbelievedthat tooth bendingfatiguecrackswouldbedifficultto initiatebasedonthe loadcapacityof thetest
rig.Due to this, notches were fabricatedin the fillet region(loadedside)on one tooth of each of the test gearsto
promotecrack initiation.The notcheswerefabricatedusingelectrodischargemachining(EDM) with a 0.10-mm
(0.004-in.)diameterwire electrode.Thenominalnotchdimensionswere0.20mm(0.008in.)inlengthand0.13 mm
(0.005in.) in width along the full face widthof the tooth.Thenotcheswere locatedat the same locationfor both
test gears. This location was at a radius of 40.49 mm (1.594in.) on the fillet, which was the position of the
greatest tensile stress for the solid gear (mB=3.3).The notches produced a stress concentration factor of
approximatelythree as determinedusing a finiteelementanalysis.
Instrumentation: The standard test rig instrumentationmonitoredtest gear speed, oil load pressure,test gear
and slave gear oil pressure, and oil temperatures.Also, overall test stand vibration was monitored using an
accelerometermounted on the top housing. In additionto the standardfacility vibration sensor, an advanced
vibration processing diagnostic system was installedin the test stand to help assist in crack detection. Crack
propagationgageswere usedin the experimentsto determinefatiguecrackgrowth.Specialgageswere fabricated
for installationin the toothfillet regionofthe test gears.The gageshad ten circularstrandswithan inner radiusof
1.52mm (0.060in.)and anouter radiusof 3.05 mm(0.120in.)(Fig.4).Thestrandswere designedto breakasthe
crack propagatedthoughthem,which inturn,increasedtheelectricalresistanceof thegage (Fig.4(a)).Figure 4(b)
shows the installation of a gage in the fillet regionof a notched tooth.A gage was installedon each side of the
toothflankfor eachgearinstrumentedwithcrackgages.Theelectricalresistanceof thecrackgagesweremonitored
along withthe loadcycle countto estimatecyclesas a functionof cracklength.The informationfromthe rotating
crack gages was transferred throughbrush-typeslip rings. Also, an infrared tach sensor was used to measure
number of loadcycles.
Measured Gear Fatigue Crack Growth: The thin-rimmedgear wasusedin test 1.The test was run at 89 N.m
(786 in..lb) torque and 10,000rpm speed for 6.5 hr, at which time rim fractureoccurred. Figure 5 plots the

number of load cycles as a function of the measuredcrack length.The crack gage results indicatedthe crack
growthwasnon-uniformthroughoutthe toothface width.A crack startedon the rear flankof the tooth at the tip
ofthe notchandreachedan initialsizeof0.46mm(0.018in.)at 1,060,000cycles.Thecrackcontinuedtopropagate
throughthe rearflankbut didnotreach the frontflank untilapproximately2,680,000cycles.At 2,910,000cycles,
the crackreacheda sizeof 0.64 mm(0.025in.)onthe frontflank,butcompletedpropagatedthroughthe reargage
by this time. Even though the crack initiationtime wasnot uniformthroughoutthe tooth face width, the crack
propagationrate was uniform.This wasindicatedby the similarityin slopes of the curves in Fig. 5 for gages 1
and 2.
The thick-rimmedgear was used in test 2. This gear was run at 136N-m(1200 in..lb)torque and 10,000rpm
speed for 15 minutes, at which time tooth fracture occurred.Figure6 gives the processedcrack propagation
results for test 2.Note that the crackinitiationand crackpropagationwasfairlyuniformthroughoutthe toothface
widthfor this test.
Comparisonof Predictedand MeasuredCrackGrowth: The FRANCcomputerprogramwasusedto simulate
crack propagationand calculatemodeI stressintensityfactorsas a functionof cracklength.Thepredictedstress
intensityfactorswere thenusedwiththreedifferentfatiguecrack growthmodels(Paris,Collipriest,and Inoue)to
estimatecrack propagationlife.
A comparisonof predictedcrack propagationcyclesusingthe Paris,Collipriest,and Inouemethodsis shownin
Fig. 7. For this, the thin-rimmedmodel(mB=0.3)was usedto simulatethe testgearof Fig. 3(b).An initial crack
of0.64 mm(0.025in.)wasplacedinthetoothfilletatthelocationofthemaximumtensilestress.Crackpropagation
was thensimulatedand the modeI stressintensityfactor as a functionof cracklengthis givenin Fig. 7(a). From
this, six differentfatiguegrowthcaseswereconsidered.Thefirst fourcasesusedthe Parisequation and material
constantsofAIS19310specimensfromexperimentsof Auand Ke [20].ThefifthcaseusedtheCollipriestequation
and AISI 9310materialconstantsfrom Formanand Hu [21].The loadratiousedwasR=-2.6as determinedfrom
the finiteelement analysis.The sixthcaseusedInoue'smethodandthe materialconstantsofthe SCM415material
(SCM415is a high-strengthJapanesesteel,similar in propertiesto AISI9310).Thepredictednumberof cycles
per crack lengthvaried significantlyamongthecases studied(Fig.7(b)).Note thatthe cyclesweredefinedas the
number of crack propagationcyclesafter an initial crackof 0.64 mm (0.025in.).
Predicted crack growth for the m8---0.3and 3.3 gears were comparedto the measuredcrack growth from the
experiments.Again, the six differentpredictionschemesas mentionedabovewereused.Thepredictednumberof
crackpropagationcyclesusingthe sixthschemeswere,onthemostpart,extremelylowcomparedto the measured
numberof cyclesfromthe experiments.To accountfor this,theconceptof fatiguecrackclosure wasinvestigated.
Elber [22] performed crack experimentson aluminumalloys and deduced that residual compressivestresses
existednear the crack tip regiondue to plastic deformation.These residualstressesreduced the effectivestress
intensity factor range (and thus, increasedcrack propagationlife) and provideda betterfit to experimentaldata
than other empiricalexpressions.Elber proposedan effectivestressintensityrangeratio, U, suchthat

zXKeg
= U(aK)

(7)

where zlKeffwas the effective stress intensity factor range. Elber then used the effective stress intensity factor
range in the Paris fatigue crack growth model. In addition, Elber defined U through experimental studies as a
linear function of the load ratio, R.

The conceptof fatiguecrack closure wasappliedto the currentgearcrackexperimentsand predictions.A study

was thenconductedto estimatetheeffectivestressintensityfactorrangeratiofor thegearcrack experiments.The
predictednumber of crack propagationcycles using the previouslymentionssix schemeswere plotted versus
crack lengthat a varietyof arbitrarilychosen Uratios.Forthe Parisequationand materialconstantsn=2.954and

C=8.433x 10-9mm/cyc/(MPa_/m)n,
goodcorrelationbetweenpredictedcrackcyclesand theexperimentsoccurred
when: 1) U---0.4for R=-2.6, and 2) U---0.8for R=-0.1.Assuminga linear relationbetween Uand R produced
U= 0.82+0.16(R)

(8)

Figure 8 shows a sample comparisonof predictedand measuredcrack growth when the fatigue crack closure
concept was used. The cycles were definedas the numberof crack propagationcycles after an initial crack of
0.64 mm(0.025 in).It shouldbe notedthat goodcorrelationwasalso achievedwhen the Collipriestequationwas
used with certain Uvalues.This produceda relationshipsimilarto Eq. 8 but with differentcoefficients[10].
Figure 9 displaysthe effect of rim thicknesson predictedmode I stress intensity factors and predicted crack
propagationcycles.The stressintensityfactorsweredeterminedfromFRANCusing theappropriatefiniteelement
models. The Paris equation was used alongwith the effectivestress intensity range ratios of Eq. 8. The initial
cracks of the various models were placed at the locationof the maximumtensile stress in the tooth fillet. The
stressintensityfactorswerelowestfor the mB---0.5
case.This gavethe highestpredictednumberof cyclesfor the
cases studies. The cycles all were defined as the number of crack propagationcycles after an initial crack of
0.28mm (0.011 in).Thestress intensityfactorswerehighestfor themB=0.3case. However,the predictedlife for
this was somewherebetween the case of mB--0.5and 1.0 due to the fatiguecrack closure effect. The cases of
mB=3.3and 1.0 gave nearlythe samepredictedlife.
Conclusions: Analyticaland experimentalstudieswereperformedto investigatetheeffectof gear rimthickness
on crack propagationlife. The followingconclusionswere made: 1)Good correlation between predictedand
measuredgear crack growth was achievedwhen the predictionsused the Paris crack growth equation and the
conceptof fatiguecrackclosure.2)For thinrims,a decreasein rimthicknesscausedan increaseinboth thestress
intensity factor and the compressivecyclic stress in the gear tooth fillet region.The increasein stress intensity
factor promotedcrack growthwhile the increasein cycliccompressivestresstended to retard crack growth and
increasethe number of propagationcyclesto failure.
Acknowledgment: The authorswishto thankDr.PaulA.Wawrzynekof FractureAnalysisConsultants,Inc., for
fruitful discussionsand for providingthe FRANC program.
References:
1. "FundamentalRating Factors and CalculationMethodsfor InvoluteSpur and Helical Gear Teeth",ANSI/
AGMA2001-B88,AmericanGearManufacturersAssociation,Alexandria,VA,1990.
2. Drago, R.J., and Lutthans,R.V., "CombinedEffectsof RimThickness and Pitch Diameter on Spur Gear
ToothStresses",Journalof theAmericanHelicopterSociety,Vol.28, Jul., pp 13-19,1983.
3. Ahmad,J., and Loo, ET., "On the Use of Strain EnergyDensityFractureCriterion in the Design of Gears
UsingFinite ElementMethod",ASMEPaperNo.77-DET-158,presentedattheDesignTechnicalConference,
Chicago,IL, Jun., 1977.
4. Honda, H., and Conway,J.C., "AnAnalysisby Finite ElementTechniquesof the Effects of a Crack in the
GearToothFillet and its Applicabilityto EvaluatingStrengthof the Flawed Gears", Bulletin of the JSME,
Vol.22, No. 174,Dec.,pp. 1848-1855,1979.
5. Flasker,J., and Jezernik,A.,"The ComparativeAnalysisof CrackPropagationinthe GearTooth",Proceedings
of the InternationalConferenceof Applicationof FractureMechanicsto Materialsand Structures,Freiburg,
WestGermany,Jun., pp. 971-982, 1983.
6. Kato, M., Inoue, K., Deng, G., and Jeong, B.S., "Strength Evaluationof Carburized Gear Teeth Based on
FractureMechanics",Proceedingsofthe KSME/JSMEJoint Conference"Fractureand Strength'90", Seoul,
Korea,pp. 248-253,1990.

7. Inoue, K., Kato, M., Deng, G., and Takatsu,N., "FractureMechanics Based Evaluation of Strength of
Carburized Gear Teeth", Proceedings of the JSME International Conference on Motion and Power
Transmissions,Hiroshima,Japan,Nov.,pp. 801-806,1991.
8. Daniewicz,S.R., Collins,J.A., and Houser,D.R., "The StressIntensityFactor and Stiffnessfor a Cracked
Spur Gear Tooth",Journalof MechanicalDesign,Vol.116,No. 3, Sep., 1994.
9. Flasker,J., and Pehan, S., "CrackPropagationin ToothRoot WithVariableLoading",Communicationsin
NumericalMethodsin Engineering,Vol.9, No. 2, Feb.,pp. 103-110,1993.
10. Lewicki,D.G.,"CrackPropagationStudiesto DetermineBenignorCatastrophicFailureModesforAerospace
Thin-RimGears", Ph.D.Dissertation,CaseWesternReserveUniversity,May 1995.
11. Pads, P.C.,and Erdogen,F.,"A CriticalAnalysisofCrackPropagationLaws",Journalof BasicEngineering,
Vol.85, pp. 528-534, 1963.
12. Collipriest,J.E., "An Experimentalist'sView of the Surface FlawProblem",The Surface Crack:Physical
Problemsand ComputationalSolutions,AmericanSocietyof MechanicalEngineers,pp. 43-61, 1972.
13. Wawrzynek,P.A.,"DiscreteModelingof CrackPropagation:TheoreticalAspectsandImplementationIssues
in Two and Three Dimensions",Ph.D.Dissertation,CornellUniversity,1991.
14. Henshell, R.D., and Shaw, K.G.,"Crack Tip Finite ElementsAre Unnecessary",InternationalJournal for
NumericalMethodsin Engineering,Vol.9, pp. 495-507, 1975.
15. Barsoum,R.S., "Onthe Use of IsoparametricFiniteElementsin Linear FractureMechanics",International
Journal for NumericalMethodsin Engineering,Vol.10,No. 1,pp. 25-37, 1976.
16. Tracey,D.M.,"Discussionof 'On the Useof IsoparametricFiniteElementsin LinearFractureMechanics'by
R.S. Barsoum",InternationalJournalfor NumericalMethodsin Engineering,Vol.11, pp. 401-402, 1977.
17. Erdogan,E, and Sih, G.C., 1963,"On the CrackExtensionin PlatesUnder Plane Loadingand Transverse
Shear",Journal of Basic Engineering,Vol.85, pp. 519-527, 1963.
18. Hefeng,B., Savage,M., and Knorr,R.J., "ComputerModelingof Rack-GeneratedSpurGears", Mechanism
and Machine Theory,Vol.20, No. 4, pp. 351-360,1985.
19. P3/PATRANUser Manual, PDAEngineering,CostaMesa, CA, 1993.
20. Au, J.J., and Ke, J.S., "CorrelationBetween FatigueCrackGrowthRate and FatigueStriation Spacing in
AISI 9310 Steel",Fractographyand MaterialsScience,ASTMSTP733,pp. 202-221, 1981.
21. Forman,R.G.,andHu,T., 1984,"ApplicationofFractureMechanicsonthe SpaceShuttle",DamageTolerance
of MetallicStructures,ASTM STP842,pp. 108-133,1984.
22. Elber,W., "The Significanceof Fatigue CrackClosure",DamageTolerancein Aircraft Structures,ASTM
STP486, pp. 230-242, 1971.

Figure 1.--Finite element model of gears used in

crack propagation studies, solid model, mB = 3.3.
(a)

C-94-02304

r Test
Oil-seal
gas

flow7

/i

gears

Slave-system

.-"

....

oil inlet
Drive

Viewing port -7/_/

Test-gear
cover

....

shaft -7

-//_

)/

::
r

-_

Shaft

oil

seal

Test-

lubricant:."

Load _
pressure -_

inlet _

_.

_"-

Loading vane
Slave

z. Test-lubricant
tern pe rature
measurement

outlet

gear

._._
/
_'_/

location
Figure 2.--NASA

Lewis Spur Gear Fatigue Rig.

(Ls)

......

C-94-02303

Figure 3.--Test gears used to determine effect of rim

thickness on crack propagation.
(a) m B = 3.3. (b)
m B =0.3.

4
Gage 1, front tooth flank
r-r-.--

30

Crack growth direction

J-

o
(-.

Gage 2, rear tooth flank

20_

.=_
2

I---

r---r-0

-----,-----'_
[ I
L
0 1 2

{
3

I
4

I
5

I
6

I
7

1
0.0
!
8

I
9

Number of broken strands

Figure 5._rack
1, mB = 0.3.

0.5

1.0
1.5
Crack length, mm

2.0

propagation fatigue growth for test

0.20
Gage 1, front tooth flank

o. o

r-

._o0.15

0.05
0.00
0.0
_bll

_IL

Figure 4.--Specialized crack propagation gages for

gear tooth crack growth measurements. (a) Increase in gage electrical resistance as the number
of broken strands increase. (b) Installation of crack
propagation gage on test gear.

0.5
1.0
Crack length, mm

1.5

Figure 6.---Crack propagation fatigue growth for test

2, mB = 3.3.

1.0

"6
>,
o
t- 0.5

ictions

Experi
....

11: 50

T_
0.5

0.0
0.0

a.
=-o

40

.__ 30
,-_

t
'
1.0
1.5
Crack length, mm

'
2.0

2.5

Figure8.--Comparison of predictedcrack propagation cyclesto experiments. Paris fatigue crack

growth model, n = 2.954, C = 8.433 x 10-9 mm/cyc/
(MPaq-m)
n, R = -2.6, U = 0.4, usedfor predictions.
Test 1, gage 1, front flank,for experiments.

20

off)

if)

10

25(a)

-o
o

D_

:_

0 [a)
0.0

,
0.5

,
,
,
1.0
1.5
2.0
Crack length,mm

,
2.5

3;- 20
_
m
15

mB = 0.3
F

mB = 1.0
mB = 3.3

u_

1 Parisequat_0n,n = 2.264, C = 1.460x10"7rnm/cyc/(MPa_-_)

n

0.2

PadsequatJon,n=2.954,

Pads equation, n = 2.420, C = 3.988x10al mm/cycJtl_Pa_'-_)

n

r-

C=8.433x10"grr_n/cyc/(MPa_-m)n

3 PadsequatJon,
n=2.555, C=2.S12x10"Smm/cyc/(MPa_r-m)
n
5 Collip_stequation

6 Inoue
equation

10

//_m

B 05
= "

u}

cn

"_

o
>"
o

:_ 0

= 0.1

1
1

.o_

,
2

,
3

l
4

,
5

Crack length, mm

,_

3
4

6 ib)
5

0.0
0.0

0.5

1.0
1.5
2.0
2.5
Crack length, mm
Figure 7.--Comparison of predicted crack propagation cycles using Paris equation, Collipriest equation, and Inoue equation. Model for mB = 0.3.
(a) Mode I stress intensityfactor. (b) Life
comparison.

_mB

= 0.5

4
_>, 3
o
_" 2
.9

_=-

1..__.0

:_ 1
0

Crack length,mm
Figure9.--Effect of backupratioon stressintensity
factorsand crackpropagationcycles. (a) Mode I
stressintensityfactors. (b)Crackpropagation
cycles,Parisfatiguecrackgrowthmodel,n = 2.954,
C = 8.433x 10-9 mm/cyc/(MPa_/'m)
n, U = 0.82 +
0.16 R.

]0

REPORT DOCUMENTATION

Form Approved

PAGE

OMBNo.
0704-0188

Pub c reportng burden for this collectionof informS=onis estimated to average 1 hour per response, includingthe time for ravi..ewinginstructions,.se,;u_h;r,_existing data sources..,
gatheringand maintainingthe data needed.,and cornplet.ing
.a.n.
d rev.iewing!he .collection.
d inforrna_io_ Send co.rnmentst .agard,ngthis burden e s't,mateor,.,arly=o=ther^_5
jl_effol _ns
collectionof information,including$uggestzonsfor reducingthis ouroen, to washington Heaoquarters_e_rvlces,ucecto.ratetot Inlorm_.jonuj:)er=atjons.ano_epon_s,i_o_ene_
Davis Highway, Suite 1204, Atlington,VA 22202..4302, and to the Office of Management and Budget, paperwork HOOUC110n
vrolect (g/o4-1Ol_}, wasnmglon, t._ z 0_rj.

1. AGENCY USE ONLY (Leave blank)

2. REPORT DATE

3. REPORT TYPE AND DATES COVERED

January 1996

Technical Memorandum

4. TITLE AND SUBTITLE

5. FUNDING NUMBERS

Gear CrackPropagation Investigations

WU-505-62-36
1L162211A47A

6. AUTHOR(S)

David G. Lewicki and Roberto Ballarini

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)

8. PERFORMING ORGANIZATION

NASA LewisResearch
Crater
Cleveland, Ohio 44135-3191
and
Vehicle Propulsion Directorate
U.S. Army Research Laboratory
Cleveland,Ohio 44135-3191
9. SPONSORING/MONITORING

REPORTNUMBER

E-10080

AGENCY NAME(S) AND ADDRESS(ES)

10. SPONSORING/MONITORING
AGENCY REPORT NUMBER

National Aeronauticsand Space Administration

Washington, D.C. 20546-43001

and

NASA TM-107147

U.S. ArmyResearchLaboratory
Ad,'lphi,Maryland 20783-I 145

ARJ_,,-TR-957

11. SUPPLEMENTARY NOTES

Prepared for An Integrated Monitoring, Diagnostics, & Failure Prevention Technology Showcase, Mobile, Alabama, April
22-26, 1996. David G. Lewicki, Vehicle Propulsion Directorate, U.S. Army Research Laboratory, NASA Lewis Research
Center; and Roberto Ballarini, Case Western Reserve, Department of Civil Engineering and Mechanical & Aerospace
Engineering, Cleveland, Ohio 44106. Responsible person, David G. Lewicki, organization code 2730, (216) 433-3970.
12a. DISTRIBUTION/AVAILABILITY

STATEMENT

12b. DISTRIBUTION CODE

Unclassified -Unlimited
Subject Category 37
This publication is available from the NASA Center for Aerospace Information, (301) 621-0390.
13. ABSTRACT (Maxlmum 200 words)

Analytical and experimental studies were performed to investigate the effect of gear rirn thickness on crack propagation
life. The FRANC (FRacture ANalysis Code) computer program was used to simulate crack propagation. The FRANC
program used principles of linear elastic fracturemechanics, finite element modeling, and a unique re-meshing scheme to
determine crack tip stress distributions, estimate stress intensity factors, and model crack propagation. Various fatigue
crack growth models were used to estimate crack propagation life based on the calculated stress intensity factors. Experimental tests were performed in a gear fatigue rig to validate predicted crack propagation results. Test gears were installed
with special crack propagation gages in the tooth fillet region to measure bending fatigue crack growth. Good correlation
between predicted and measured crack growth was achieved when the fatigue crack closure concept was introduced into
the analysis. As the gear rim thickness decreased, the compressive cyclic stress in the gear tooth fillet region increased.
This retarded crack growth and increased the number of crack propagation cycles to failure.

14. SUBJECT TERMS

15. NUMBER OF PAGES

12

Crack propagation; Finite element modeling; Fracture mechanics; Gears; Rims

17. SECURITY CLASSIFICATION
OF REPORT

Unclassified
qSN 7540-01-280-5500

18. SECURITY CLASSIFICATION

OF THIS PAGE

Unclassified

19. SECURITYCLASSIFICATION
OF ABSTRACT

_6. PRICE
CODE
A03
20. LIMITATION OF ABSTRACT

Unclassified
Standard

Form 298 (Rev. 2-89)

Prescribed
298-102

by ANSI Std. Z39-18

National Aeronautics and

Space Administration
Lewis Research Center
21 000 Brookpark Rd.
Cleveland, OH 44135-3191
Official Business
Penalty for Private Use $300

POSTMASTER: If Undeliverable - Do Not Retu