1、08FTM05AGMA Technical PaperGear Failure AnalysisInvolving Grinding BurnBy G. Blake, M. Margetts, andW. Silverthorne, Rolls-RoyceGear Failure Analysis Involving Grinding BurnGregoryBlake,andMichaelMargetts,Rolls-Royce - TransmissionsandStructuresand Wilson Silverthorne, Rolls-Royce - Materials and Te
2、chnologyThe statements and opinions contained herein are those of the author and should not be construed as anofficial action or opinion of the American Gear Manufacturers Association.AbstractWhen gears are case-hardened it is known that some growth and redistribution of stresses will occur whichres
3、ult in geometric distortion. Aerospace gears require post case-hardening grinding of the gear teeth toachieve necessary accuracy. Tempering of the case hardened surface, commonly known as grinding burn,occursinthemanufacturingprocesswhencontroloftheheatgenerationatthesurfaceislost. Excessiveheatgene
4、ratedatthesurfacecaninducesurfacetemperingand/orre-austenitizethesurfaceinalocalizedarea.The localized area will have reduced or altered mechanical properties in addition to an unfavorable residualstress state 1.A gearbox with minimal service time was removed from an aircraft, disassembled, and visu
5、ally inspected.Linearcracksalongthededendumoftheworkinggeartoothfacewerefoundduringvisualinspectioninthreeadjacent teeth. No teeth had been liberated. A detailed inspection of the incident gearbox found no othercomponents with distress.Metallurgicalevaluationdeterminedthatthecracksinitiatedinservice
6、attheboundaryofalocalizedgrindingburn, which had re-astenitized. The cracks propagated inward from the tooth surface in fatigue to a depthgreater than the depth of case.The metallurgical evaluation could not conclude if the crack trajectory would propagate across the toothcross-sectionorradialintoth
7、egearrim. Thecross-sectiontrajectoryresultsintheliberationofteeth. Linearelastic fracture mechanics (LEFM) was then used to predict the cracks future propagation path based onassumptions from the one incident gear.AGMA2007-C002providesdetailsofthetemperetchprocessandspecificallyrequirestheusesanitri
8、cacidetch process, which is typically used in production quality inspections. The incident gear was processed forgrindingburnusinganammoniumpersulfateetchsolution. Qualityrecordsdocumentedvariationinchemicalconcentration levels during the time the failed gear was manufactured. A design of experiment
9、s wasconductedtounderstandtheeffectsofthefactorsandinteractionsthatimpactthecapabilityoftheammoniumpersulfate process used in production to detect grinding burn.Presented are the metallurgical findings, load distribution analysis of actual geometry, crack propagationanalysis, and design of experimen
10、t results of the ammonium persulfate etch process.Copyright 2008American Gear Manufacturers Association500 Montgomery Street, Suite 350Alexandria, Virginia, 22314October, 2008ISBN: 978-1-55589-935-63Gear Failure Analysis Involving Grinding BurnGregory Blake and Michael Margetts, Rolls-Royce Transmis
11、sions and Structures andWilson Silverthorne, Rolls-Royce Materials and TechnologyOverview of gear systemTheincidentgearboxhadaccumulated1,650hoursofuseatthetimeofevent. Thedesignlifeoftheen-gine is 20,000 hours with field demonstrated life inexcess of 20,000hours. Aneventthis earlyinlifeisrare for t
12、his gearbox.The failed gear was one of many gears in a reduc-tion gearbox that is integrated with a Rolls-Roycegasturbineengine. Theengineisusedinamultien-gine fixed wing application.Apartialillustrationofthegeartrainandtheincidentgear is showninFigure1. The failedgear andmat-ing pinion are carburiz
13、ed, ground, and shot peenedAMS6265. TheenginetorquepathisalsoshowninFigure 1.Figure 1. Partial illustration of gear trainshowing torque pathThe incident gear is part of the first stage reductionin a legacy aerospace gearbox that has demon-strated high field reliability.Background of failureThegearbo
14、x has achipdetector inthescavengeoilpassage. Duringflight,achiplightwarningwasacti-vatedbythesystemforoneofthefourengines. Theaircraft safely landed without issue and groundcrews began investigation to determine the sourceof the chip light indication.The ground crew first removed the chip detectoran
15、d found a large sliver of material clinging to theend. ThesizeofthesliverisshowninFigure2. Typ-ically, wearing or pitting components will generatevery fineparticles, that whenmixedwithoilcreateapaste. Thepasteiswhatmosttechniciansexpecttofind when investigating a chip light indication. Thesize of th
16、e debris found on the plug was cause forremoval of the gearbox.Figure 2. Debris found on chip detectorTear down inspectionThe gearbox was removed and sent to a mainte-nance facility for tear down inspection. The teardownwas overseenby Rolls RoyceEngineeringaspart of standard practice.Cracks were fou
17、nd in the first stage spur gear asshowninFigure3.ToothnumbereightshowninFig-ure 3 was found to have a divot close in size to themetallic sliver found on the chip detector. Thecracks were located in three adjacent teeth on thedrivenside. Thecracksstretchedacrossthecentral75%of thefacewidthwithin thea
18、ctive toothprofile.The cracks were arc shaped, higher on the activetooth profile at the ends than in the center.4Figure 3. Three adjacent teeth cracked -observable without MPI or magnificationThe visual cracks and distress of the incident gearwas contained tothe threeteeth shownin Figure3.The remain
19、der of the gear teeth showed no evi-dence of premature failure.A detailed inspection was performed on all otherparts of the gearbox. No other components showsany signs of degradation or indications of high loadexperience. The mating pinion gear showed nosigns of surface distress or mal-distributed l
20、oad asshown in Figure 4.Magnetic particle inspection (MPI) was performedontheincidentgearatthemanufacturer. Nocracks,in addition to those found visually, were found. Noetch inspection was performed at this time to pre-ventalteringthecracksurfaces. Thecracksweretobeevaluatedindetailaspart ofthedestru
21、ctivemet-allurgical investigation.Inspection of the pinion and incident gear tooth ge-ometry was performed. The geometry of bothmembers was within specification and was of highquality.Metallurgical evaluationAphotographicmontagethroughthecrack ontooth6(seeFigure3) ispresentedinFigure5. Thecrackwasap
22、proximately0.103inchlongand0.061inchindepth. Several smaller cracks were observedbranching from the main crack. The crack inter-sectedthesurfaceapproximately0.23inchfromthetooth tip.Figure 5 shows the crack trajectory to be inwardfromthetoothsurface. Analysiswillbepresentedinthe later portion of thi
23、s paper to bound the crackpropagation path.Figure 4. Mating pinion gear - no surfacedistress or evidence of mal-distributed loadFigure 5. Photo montage of crack from toothsurfaceThe cross sectionwas etchedas shownin Figure6and grinding abuse was observed on both sides of5the tooth. Detailed views of
24、 the grinding abuse areshown in Figures 7 and 8. The grinding abuse pro-ducedarehardenedlayeronthesurfacemeasuringup to 0.007 inch deep. The crack followed theheataffected zone as illustrated in the upper image inFigure 7.Similar grinding abuse was observed on the coastside of the tooth as shown in
25、Figure 8. The dottedlines in Figure 9 show the approximate location ofthe five hardness surveys. The hardness was re-duced in the over tempered areas (dark coloredzone).Figure 6. Etch cross section showingtempered and rehardened (burned) zone indedendum area of the driven and coast sideof the gear t
26、oothFigure 7. Detailed views of the grindingabuse and crack on the driven side. Etchant5% NitalFigure 8. Coast side showing rehardened andtempered zone and locations of hardnesstraversesFigure 9. Hardness vs depth at locations withand outside of the rehardened zoneThe No. 6 gear tooth section was la
27、boratory frac-turedtoexposethe crack surface. The crack couldnotbecompletelyopenedduetothecracksshallowdepth and orientation. A SEM photograph of thefractureisshowninFigure10. Thecrackmorpholo-gy was indicative of fatigue progression from themid face towards end face as illustrated in Figure10. Thea
28、rrowsindicatethedirectionoflocalfatiguecrack progression.AnalysisAnalysis wasperformedtounderstandthelocation,shape, and expected crack trajectory. A crack tra-jectorythatresultsintheejectionofasingletoothormultiple teeth has a different end results than onethat propagates into the gear blank.6Figur
29、e 10. SEM photo showing fatiguedirection from mid face towards end faceThe flank form inspection of the incident gear lo-cated the approximate radial position of the crack.Theprofiletracesweremadeatthemidfaceofeachtooth.TheprofilechartshowninFigure11isfourad-jacent teeth, three of which contain the
30、crackedteeth. The approximate location of the crack isshownasarapidchange(bump)inthelower25%ofthe profile chart.Load distribution vs. crack locationLoadDistributionProgramversion10.9wasusedtopredicttheincidentgearloaddistribution. Theactu-al measured geometry of the pinion and incidentgear was input
31、.The contact stress distribution of the incident gearrelative to the approximate crack location is shownin Figure 11. The crack location is near the start ofsingletoothcontactofthegearbutnot intheareaofhighest contact stress. However, the crack seemstofollowthecontourofthesurfacestressintheaxialdire
32、ction.Figure 11. Approximate crack location vscontact stress distributionThemarginofsafety(MOS),usingtheAGMAindexmethodof theincident gear was greaterthan1.0asshown in Table 1. None of the AGMA index MOSwould suggest premature crack initiation.Table 1. AGMA index MOSMOSBending 1.30Contact 1.10Flash
33、temperature 1.08Crack trajectory analysisAnalysis was performed to predict the crack trajec-tory. A 2D finite element model (FEA) with actualtoothgeometry was createdas shownin Figure12.Figure 12. 2D FEA model of incident gearwithtoothgeometryThecrackasmeasuredinFigure5wasaddedtothemodel. Linear Ela
34、stic Fracture Mechanics (LEFM)was then used to predict the crack trajectory fromthis initialpoint. Thesolutionis shownin Figure13.The crack tip was modeled as a quarter point ro-sette as shown in Figure 14.The solution shown represents maximum continu-ousspeedandmaximumcontinuoustorqueappliedat the
35、highest point of single tooth contact.The depth of the initial crack depth is slightly belowthecase-coretransitionpoint. Themodelassumesno residual stress in this area of the gear.7Figure 13. Crack trajectory solution startingfrom initial crack using maximum continuoustorque and speed first principl
36、e stresscontours shownFigure 14. Quarter point rosette used tomodel crack tipThe model was then used to predict the effects ofspeed on crack trajectory. Reference 3 highlightsthe effects of rotational speed on crack trajectory.Atmaximumcontinuousspeed,thecracktrajectorywas across the tooth. The traj
37、ectory changes to-wardtheblank center as speedincreases giventhesame applied torque (Figure 15).Published analytical and test results in references3, 4, and 5 were used to validate the model.Reference6 highlights theeffects of rim thicknesson the alternating stress range experienced in thegear tooth
38、 root. The supporting geometry is there-fore expected to have a strong influence on cracktrajectory.Quality investigation & etch inspectiondesign of experimentsA quality investigation focused on two areas, thecause of the burn and cause of the escape.Cause of the grinding burnThegrindingburnwas isol
39、atedtoafinitepopulationof gears that were gear ground on an alternatema-chine during a short period of potentially affectedgears.Cause of the escapeAn investigation team identified two causal vari-ablesforthequalityescapei)humanelements,andii) chemical concentration variation in the etch pro-cess.Th
40、e manufacturing process and inspection stepswere typical of those for carburized aerospacegears. A post grind etch inspection was performedon all gears to detect grinding temper and burn.AGMA 2007-C00 provides details of the temperetchprocess andspecifies theuseanitric acidetchprocess,whichistypical
41、lyusedinproductionqualityinspections. The incident gear was processed forgrinding burn using an ammonium persulfate etchsolution.Chemical concentration levels were recorded atregular intervals. The run charts of this data (Fig-ures 16 and 17) indicate variation in chemical con-centrationlevels durin
42、gthetimethefailedgearwasmanufacturedFigure 15. Crack trajectory vs. rotational speed8Figure 16. Run chart of ammonium persulfateconcentration levels pre and post time ofquality escapeFigure 17. Run chart of HCL concentrationlevels pre and post quality escape time periodThe period between the two ver
43、tical lines repre-sents the production period of the suspect popula-tion.Both the ammonium persulfate concentration andthe HCL concentration are outside of the statisticalcontrol limits at the time the incident gear wasmanufactured.A design of experiments (DOE) was conducted tounderstand the effects
44、 of the factors and interac-tions that impact the capability of the ammoniumpersulfate process used in production to detectgrinding burn.The DOE was a full factorial withreplication, 2Lev-els per factor with center point. The variables were%ammoniumpersulfateand%HCL. Theresponsevariable was burn ind
45、ication (faint, light, dark).TheetchprocesswasreplicatedintheFailureAnal-ysis Laboratory as shown in Figure 18.Figure 18. Laboratory replication ofproduction etch processThe process steps for the DOE are shown in Table2.Table 2. DOE etch processStep 1 Heatspecimento170Ffor3min.Step 2 Blow dryStep 3
46、Immerse in enchant chemical10.0 (min spec) secStep 4 Rinse cold H2OStep 5 Heatspecimento170Ffor1min.Step 6 Blow dryStep 7 Immerse in bleach solution 10.0(max spec) secStep 8 Rinse in cold H2OStep 9 Blow dryThe maximum observed concentration levels wereused as upper and low test points. Table 3 detai
47、lsthe concentration levels for each test. Further, themax and min time for exposure to the ammoniumpersulfate and HCL were tested.The specimens were standard four point bendingspecimens made from carburized AMS6265 mate-rial. The specimens were ground with aggressiveparameters to induce surface temp
48、er.Figure 19 shows the specimens post processing.Each end of the specimen was etched separately,thus allowing for more test points.Theuseof bothends creates anoverlapareainthemiddle that must be excluded during finalevaluation.9Figure 19. Etch DOE specimens post testA qualitative scale was then crea
49、ted and the speci-mens evaluated by engineering. Table 3 containsthe test matrix with the qualitative results.An upward arrow indicates dark etch of temperedareas. Downward arrow indicates faint indicationsof temperedarea. Therelativeangleofthearrowisproportionaltothedegreeofdarknessoftheetchedindications.Figure 20, is a scatter plot of the manufacturersconcentration levels. The DOE results listed inTable3definethepreferredareain theupper left orrightquadrant. Thehorizontalandverticallinesrep-resent the manufacturers target values.The finite po
