warwick.ac.uk/lib-publications
Original citation:
Hooper, James Michael, Marco, James, Chouchelamane, Gael Henri, Chevalier, Julie Sylvie
and Williams, Darren. (2018) Multi-axis vibration durability testing of lithium ion 18650 NCA
cylindrical cells. Journal of Energy Storage, 15. pp. 103-123.
Permanent WRAP URL:
http://wrap.warwick.ac.uk/94735
Copyright and reuse:
The Warwick Research Archive Portal (WRAP) makes this work of researchers of the
University of Warwick available open access under the following conditions.
This article is made available under the Creative Commons Attribution 4.0 International
license (CC BY 4.0) and may be reused according to the conditions of the license. For more
details see:
http://creativecommons.org/licenses/by/4.0/
A note on versions:
The version presented in WRAP is the published version, or, version of record, and may be
cited as it appears here.
Multi-axis
vibration
durability
testing
of
lithium
ion
18650
NCA
cylindrical
cells
James
Michael
Hooper
a,*
,
James
Marco
a,
Gael
Henri
Chouchelamane
b,
Julie
Sylvie
Chevalier
b,
Darren
Williams
caWMG,UniversityofWarwick,Coventry,CV47AL,UK b
JaguarLandRover,BanburyRoad,Warwick,CV350XJ,UK
c
MillbrookProvingGroundLtd,Millbrook,Bedfordshire,MK452JQ,UK
ARTICLE INFO
Articlehistory:
Received5June2017
Receivedinrevisedform7November2017 Accepted8November2017
Availableonlinexxx
Keywords:
Sixdegreesoffreedom(6DOF) Multi-axisshakertable(MAST) Electricvehicle(EV)
Li-ionbatteryageing Vibration
Durability
ABSTRACT
This paperpresentsnewresearchtodetermineiftheelectromechanicalattributesofNickelCobalt Aluminium Oxide (NCA) 18650 battery cells are adversely affected by exposure to vibration commensuratewiththatexperiencedbyelectricvehicles(EVs)throughroadinducedexcitation.This investigationappliedvibrationtoasetofcommerciallyavailablecellsinsixdegreesoffreedom(6-DOF) usingamulti-axisshakertable.Thismethodofmechanicaltestingisknowntobemorerepresentativeof thevibrationexperiencedbyautomotivecomponents,as6motionsofvibration(X,Y,Z,roll,pitchand yaw) areapplied simultaneously.Withinthecontextofthis study,cellcharacterisationwithinthe electricaldomainisperformedviaquantificationofthecell’simpedance,theopen-circuitpotentialand the cell’s energy capacity. Conversely,the mechanical properties of the cell are inferred through measurementofthecell’snaturalfrequency.Experimentalresultsarepresentedthathighlightthatthe electromechanical performances of the 18650NCA cellsdo not, in the main, display statistically significantdegradationwhensubjecttovibrationrepresentativeofatypical10-yearEuropeanvehicle life.However,astatisticallysignificantincreaseinDCresistanceofthecellswasobserved.
©2017PublishedbyElsevierLtd.
1.Introduction
To help reduce the risk of excessive warranty costs and to ensurein-marketreliability,automotivemanufacturersperforma varietyofliferepresentativedurabilitytestsoncomponentsand sub-systemsatthephysicalprototypestageofthevehicledesign anddevelopmentlife-cycle[1–5].Vibrationdurabilitytestingvia theuseofelectromagneticorhydraulicshakersisonesuchtest programme that is strategicallyemployedwhen evaluating the suitabilityof differentconcept designs. The test programme is conductedtodeterminethedurabilityofkeyvehiclecomponents whensubjecttovibrationenergythatisrepresentativeofthe in-service,real-world,environment.Theimportanceofsuchtestsis widelyacceptedwithinindustry,aspoorlyintegratedcomponents, assemblies or structures subject to vibration can result in a
significantly reduced service life [2,4,6–8]. In the worst case, catastrophic structural failure, through mechanisms such as fatiguecrackingorwork hardeningofmaterials,mayalsooccur [4,8].
Manyelectricvehicle(EV) manufacturersareadopting cylin-dricalformatcellswithintheconstructionoftheirbatteryelectric vehicle(BEV)andhybridelectricvehicle(HEV)batteryassemblies
[9,10].Cylindricalcellsareoftenchosenforintegrationwithinthe
rechargeableenergystoragesystems(RESS)foracombinationof commercial and technical reasons, that include: built in safety features such as a current interrupt device (CID), excellent economies of scale due to their high manufacturing volumes andadimensionallystablesteeloutercasemakingthemeasierto packagewithinanEVmoduleassembly[9–12].However,thereis currently limited data and evidence within industry or the academicliteraturethatdefinestheimpactthatvibrationenergy mayhaveontheperformanceorservice-lifeofEVbatterysystems
[13].Muchofthepublishedresearchintothemechanicaltestingof
EVcellsisunderpinnedbytheneedformanufacturerstocomply withmandatorytransportlegislation, suchas UN38.3[14] and vehiclecrashhomologation[15].Subsequentlytherehasbeena
* Correspondingauthor.
E-mailaddresses:[email protected](J.M.Hooper),
[email protected](J.Marco),[email protected]
(G.H.Chouchelamane),[email protected](J.S.Chevalier),
[email protected](D.Williams).
https://doi.org/10.1016/j.est.2017.11.006
2352-152X/©2017PublishedbyElsevierLtd.
ContentslistsavailableatScienceDirect
Journal
of
Energy
Storage
clearfocus onstudies investigating mechanical robustness and crashworthinessviamechanicalcrush[16–18],penetration[19,20] andmechanicalshock[16,21].Asdiscussedwithin[13],alimited bodyofresearchexists,forexamplepresentedin[21–26],which investigatestheeffectof vibrationonli-ioncellchemistriesvia single-axistestmethods.However,ashighlightedwithin[13],the academicliteraturepresentsconflictingevidencewithregardto thesusceptibilityofli-ioncellstovibrationenergyappliedalonga single-axis.Anumberofpotentiallimitationsexiststhatfurther compoundthechallengeofunderstandingthecausalitybetween vibrationloadingandthecellperformanceandageing.Firstly,the
single-axis vibration profiles employed within [21–26] are not representativeofreal-worldEVuse.Theswept-sinewavesutilised withinthedurabilityassessmentofthecells[21–24]areequally deemedtobeanunrealisticrepresentationofthevibrationloading thatoccurswithinroadvehicles[27].Finally,thevibrationprofiles employeddo notrepresent a knownservice-life condition, e.g. 100,000milesofvehicleuse,resultinginthedatageneratedbeing unsuitableforthelifeestimationofthecells[13].
Thisstudyis anextensionofprevious publishedresearchto definetheimpactofvibrationthat isrepresentativeof100,000 miles of vehicle durability on 18650 Nickel Manganese Cobalt (NMC) battery cells and 18650 NickleCobalt AluminiumOxide (NCA)batterycells[13,28].Withinthesestudies,degradationin cellelectricalandmechanicalperformancewasobserved, regard-less of the cells’ stateof charge (SOC) or physical orientation. However,unlikecomplementaryEVcellvibrationstudies,which havefocusedonconductingrobustnessanddurabilityevaluation using single-axis vibration test methods [13,21–26,28], this researchinvestigatestheeffectof vibrationwhen appliedin 6-degreesoffreedom(6DOF)simultaneously.Thiswasperformedby subjecting cells tovibration signals measuredfromthebattery assembly of a commercially available Smart ED vehicle when drivenoverindustrystandardprovinggrounddurabilitysurfaces. Unlikewithin[13,28]whichapplyvibrationuni-axiallywithinthe frequency domain via the use of acceleration spectral density (ASD)profiles,thisstudyappliesthemeasuredvibrationwithinthe time domain.Theadvantageof this methodologyis thaterrors resultingfromtesttimecompressionareavoided[2].Also,because thevibration motionis appliedin thetime domain, it is more representativeofreal-worldin-vehicleloading.Asdiscussedwithin [29–32] within the context of traditional vehicle testing and componentevaluation,theapplicationofcombinedaxialmotions willoftenhighlightadditionalfailuremodesthatwouldotherwise notbeobservedthroughsingle-axistesting.
The primary objective of this research is to quantify the underpinning causality between the application of vibration energyand thecell’selectrical performanceand itsmechanical attributes.Thisstudyalsoaimstoidentifyifthein-packorientation ofthecellscaninfluencethedegreeofmechanicalandelectrical ageingthatmaybeobserved.Withinthecontextofthisstudy,cell performance is quantified through themeasurement ofnatural frequency, pulse power capability, electrochemical impedance spectroscopy(EIS)anddischargeenergycapacity.Full characteri-sationofthecellshasbeenundertakenbeforeandaftervibration durability testing. Anadditional motivationfor this study is to provide a comprehensive framework for multi-axis vibration testing within the field of EV battery test and evaluation that encompassesboththeexperimentalset-upandanappreciationof keysafetyrequirementsandconstraints.Thispaperisstructured asfollows:Section2 providesa detailed descriptionof thetest methodologyemployedandthedesignofkeytestfixturesthatare pre-requisitetotheresearch.Section3presentstheexperimental resultsand quantification ofthecellelectricalperformanceand mechanicalattributesbeforeandaftervibrationdurabilitytesting. Statisticalanalysisisundertakentoidentifyifanychangeincell performanceisstatisticallysignificant.Discussion,FurtherWork andConclusionsarepresentedinSections4,5and6,respectively.
2.Experimentalmethod–multi-axisvibrationdurability testingin6DOF
Thetestmethodemployedwithinthisstudyissummarisedin Fig.1.Thissectiondefines,ingreaterdetail,thetestmethodology employedtobetterunderstandthedurabilityandageing behav-iour18650NCAcellswhensubjecttovibrationenergyin6DOFthat iscomensuratewith10-yearsofEVservicelife.
Nomenclature
6DOF Sixdegreesoffreedom
Ah Amperehour
ANOVA Analysisofvariance
ASD Accelerationspectraldensity BEV Batteryelectricvehicle CAE Computeraidedengineering
CC Constantcurrent
CID Currentinterruptiondevice
CT Computertomography
CV Constantvoltage
DC Directcurrent
E Young’smodulus
EIS Electrochemicalimpedancespectroscopy EMS Electromagneticshaker
EOT Endoftest
EV Electricvehicle
g Grams
gn The force of acceleration equivalent to gravity
where1gn=9.81m/s
HEV Hybridelectricvehicle
Imax Maximumappliedcurrentpulse
Li-ion Lithiumion
MAST Multiaxisshakertable
mm Millimetre
MPG Millbrookprovingground MSD Millbrookstructuraldurability NCA Nicklecobaltaluminiumoxide NMC Nickelmanganesecobalt PLC Programmablelogiccontrolled
PPT Powerpulsetest
RAC Remoteairconditioning RCT Chargetransferresistance
RDC DCresistanceofcellsasmeasuredbyPPT
RESS Rechargeableenergystoragesystem
RMS Routemeansquare
Ro Ohmicresistance
S Seconds
SmartED Smartelectricdrive SOC Stateofcharge SOT Startoftest
V10s Voltageafter10scurrentpulseatimax
VOCV Voltagepriortotheapplicationofcurrentpulseat
imax
Wh/kg Watthourperkilogram WMG Warwickmanufacturinggroup
X Xaxis
Y Yaxis
Z Zaxis
2.1.Testsamples
Twelve186503.1AhNCAcellswereevaluated.Thecellshavea nominalvoltageof3.7V.Duringthisinvestigation,the18650NCA cellswereassessedatanenergylevelof75%SOC.75%SOCwas identifiedwithina previousvibration durabilitystudy[13]asa chargestatewhereincreasedcelldegradationmaybeobserved. Thecellsweredividedintotwobatches.Onebatchcomprisedof9 cells and was subject to the vibration profile in 6DOF. The remaining3 cells weredefined as controlsamples.Thecontrol sampleswereplacedintostorageat10Cforthedurationofthe test programme. Control samples that were not subject to vibration,wereincludedtofacilitateamorecomplete understand-ing of vibration related degradation relative to other ageing mechanisms,forexamplecalendarageing[33,34].
Cellsareoftenpackagedwithindifferentphysicalorientations, either within a single RESS or by different EV manufacturers exploringdifferentdesignoptions.Anexampleofthisvariationin cellpackagingistheTeslaModelSthatpackagesthe18650cellsin theZorientation,whilsttheTeslaRoadsterpackagesthe18650 cellsintheYorientation.Asaresult,thisstudyevaluatestheeffect ofthethreedifferentX,YandZaxiscellorientations,definedin Fig.2.Detailsof eachcellsample, includingitsorientationand uniqueidentifieraredefinedinTable1.
2.2.Experimentalfixturesandcommissioning
In order to provide a comprehensive understanding of the experimental method employed, a detailed description of the
differentcellmountingfixturesisprovided.Particular consider-ationisgiventothoseaspectsofthemethodologythatarekeyfor ensuringthesafetyofthetestandformaintainingahighdegreeof testaccuracyandrepeatability.
2.2.1.Cellfixtures
Fig.3apresentsoneofthreecell-mountingfixturesthatwere designedandfabricatedtosupportvibrationtesting.Eachfixture holdsuptothreecellsandisintendedtorecreatearepresentative 18650 EV RESS mounting condition [13]. Three cell-mounting fixturesweremanufacturedtoallowfortheconcurrentevaluation ofmultiplecellsindifferentorientationsduringasinglevibration
testontheMAST[13].Thedifferentcellorientations(X,YandZ)
were achieved by mounting the three cell fixtures onto the differentsurfacesofthecubeshapeddurabilityfixtureillustrated
inFig.3b[13].Toundertakeresonanceevaluationstomechanically
characterisethecellsatthestartoftest(SOT)andendoftest(EOT), as discussed within Section 2.5.5, an additional cell resonance searchplatewasfabricated.ThisfixtureisshowninFig.3candwas requiredtoprovideaninterfacewiththesingleaxis electromag-netic shaker table.The single-axisshakerwas employedin the naturalfrequencymeasurementofthecellsduetoitsfrequency capabilitythanthehydraulicMAST.
Adetaileddiscussionintothedesign andmanufactureofthe different cell mounting fixtures is provided in [13] and will therefore not be repeated here. All vibration durability and assessment fixtures employed were constructed from 6082-T6 gradealuminium.ThisisduetothehighPoisson’sratioassociated withthismaterial(circa:0.33).CalculationofthePoisson’sratiois definedinEq.(1),wherethematerialsYoung’smodulusisdefined asEandthedensityas
r
[35].Ahighmaterialsratio(e.g.greater than0.3)indicatesagreaterpotentialforahighnaturalfrequency[35]. This is important to ensure that the fixtureresonance is
beyondthatfrequencyrangeofthetest.Itmustbenotedhowever thattheresonancebehaviourofthefixtureis
Poisson0sRatio¼
r
E ð1Þ2.2.2.Testfacilityandsetup
ThecompletetestfacilityisshowninFig.4.Thetestrigemploys a TEAMcube MAST. The MAST was installed within a climatic
Fig.1.SchematicofTestProcessforCells.
Table1
TestSampleInformation.
SampleNo TestProfile OrientationforTest
1 DatafromSmartEDsequencedto10-yearsEuropeancustomerstructuraldurability Z
2 Z
3 Z
4 Y
5 Y
6 Y
7 X
8 X
9 X
10 Storageatatemperatureof10C Control
11 Control
12 Control
Fig.3. a)SingleThree18650CellHoldingFixture,b)18650BlockFixture,c)SingleThree18650CellHoldingFixtureonResonanceSearchPlate.
chamber which, in turn, was connected toan L.Kel remote air conditioning(RAC)unit.
Duringtesting, theambienttemperaturewas conditionedto 21C3C. Withinthetestenvironment,K-typethermocouples andvibrationsensors(accelerometers)wereemployedtoprovide testcontrolaccuracyandsafety.Tofacilitateclosed-loopvibration control,six blocksofthreeDJBInstrumentsA/130/V accelerom-eterswereinstalledatkeylocationsontheshakertable(discussed further in Section 2.4). A Labview PXie-1075 chassis with an integrated Ni-PXIe-8133 controller and input modules for 32 thermocouples (NI PXIe-4353), 4 channels for accelerometer measurements(NIPXI-4462)andamultifunctionaldatamodule (NiPXIe-6363)wereusedforadditionaldatarecordingandsafety monitoring.TheMASTwascontrolledbyaMoogCC02310-301test controller(SerialNo0069).
Severalsafetymeasureswereincludedwithinthedesignofthe testset-up.Tosafeguardagainstcatastrophiccellfailure,thetest fixturewasinstalledwithinastainless-steelfireproofenclosure. Theenclosurewasfabricatedfrom3mmthick304gradestainless steel.AgasNitrogeninjectionsystemwasintegratedwithinthe enclosure so that the test items could beplaced into an inert environmentifasignificantincreaseintemperaturewasdetected. Integrated with the enclosure was a programmable logic controlled (PLC) monitoring system that would automatically activateanaudiblealarmifthesurfacetemperatureofthecellsis greater than 60C or a rate of increase greater than 4C/s is observed.The thermocouples wereset torecordat a rate of 1 sampleasecond.
Aremotelyactivatedgasextractionsystemwasinstalledinto theclimaticchamberandthetestenclosuretoallowfortheforced extraction of potentially harmful gases. Gas monitoring was performedremotelythroughairsamplingviaaRKIInstruments Eagle gas analyser Serial No (04-021-118) at a measurement sampleintervalof30s.Thissystemwouldalsooutputanaudible alarm if hazardous quantities of hydrofluoric gas (>1 partper million),unsafeoxygenlevels(<19%)oranyhydrocarbonswere detected.
Thebaseplateofthefixturewasfabricatedfrom25mmthick 6082-T6aluminiumandcontainedcastnyloninserts.Castnylon wasincorporatedintothefixturedesigntoreducetheriskofan electricalearthpathbeingcreatedfromthetestitemtotheMAST. 10mmthickG10(fibreglassepoxylaminate)sheetingwasplaced betweenthetestfixtureandMASTtoreduceheattransfertothe castmagnesiumalloycomponentsoftheMAST.TheG10sheeting also provided additional electrical insulation between the NCA cellsandthefacility.Fig.5illustratesthetestsetupontheMAST.
2.3.Rigandfixturepre-testingcharacterisationandcommissioning
Theprimaryrequirementofdurabilitytestingistoensurethat the vibration profile demanded by the electronic controller is faithfully applied to the samples under test. This is generally achievedbydesigningtheexperimentalfixturetomaximisethe transmissibilityofthevibrationenergyfromtheshakertabletothe sample whilst concurrently minimising any unrepresentative cross-axis behaviour[13,28].It is widelyacknowledged as best practice to evaluate the vibration response characteristics of fixturespriortocommencingdurabilitytesting[35–37].Thisisto ensurethat noresonanceswhichcouldaffecttheaccuracy and controlofthetestoccurwithintheshakerassemblyitself.Theterm Transmissibilityisacomparisonoftheoutputsignaltotheinput signal [36] and is determined by the pre-test experimental evaluation ofthefixture.Fixtures arecharacterisedviaaswept sineresonancesearchpriortotestingtoensurethatnosignificant resonancesoccurinthethreeaxesofthevibrationfixture.Priorto evaluating thefixtures and prior to commencing any vibration study [38–40], it is also necessary to fully understand the frequencyresponseoftheMAST.ThisistoensurethattheMAST doesnotexhibita resonancewithinthefrequency rangeofthe durabilitytest.Withinthecontextofthisstudy,thiswas1–110Hz. AfurtherrequirementistoensurethattheMASTdoesnotgenerate anyvibrationspectrawhichcouldcreateunrepresentativefailure modestooccur.
Thissectionpresentstheresultsfromthepre-testevaluationof theMASTandthetransmissibilityofthecellmountingfixtures. Thisevaluationwasconductedasfollows:
AssessmentoftheMASTforresonancesintheX,YandZaxisat 1gn(gnistheforceofaccelerationequivalenttogravitywhere
1gn=9.81m/s)overafrequencyrangeof1–110Hzat1octave/
min.
Modalanalysisoffixtures
Assessmentof18650fixturesandresonancesearchplatewhen evaluatedinaccordancewithBSEN60068
2.3.1.Frequencyresponseofmulti-axisshakertable(MAST) ThevibrationfrequencyresponseoftheMASTwasmeasured usingasweptsinewaveofamplitude1gnoverafrequencyrangeof
1–110Hzat1octave/min.Fouraccelerometerswereplacedinthe X,Y andZaxisacrosstheheadexpanderoftheTEAMcubeMASTon bespokealuminiummountingblocks.Thetablewasexcitedinthe Zaxisandthecrossaxismotionwasmeasured.Uponanalysingthe
response of the MAST, no significant resonances or cross axis motion was identified that would detrimentally impact the accuracy or reliability of the durability test programme. The transmissibilityofMASTispresentedinFig.6.
2.3.2.Vibrationresponseof18650fixturesandresonancesearchplate whenassessedviaimpactexcitationmodalanalysis
Priortoplacingthefixturesontheirrespectiveshakerfacilities, theirmodalperformancewasassessedusingimpactexcitation,via hammersurveyingtechniquesusingasingleinputmultipleoutput (SIMO)method.Thiswasundertakentodeterminetheirsuitability forconductingthedurabilitytesting,throughtheidentificationof their mode shapes and the first modal frequencies. The test methodutilisedwithinthisstudyisdiscussedindetailwithinthis section,howeverthesupportingtheorybehindmodalanalysisis presentedin[41–44].
With this method several measurement accelerometers are attachedtothetestitematmultiplelocations.Theinputexcitation is appliedtothe test item at a single locationvia a calibrated aluminiumtippedimpacthammer(withbuilt-inloadcell).Each
pointisassessedindividuallybeforebeingcombinedtodetermine themodalbehaviourofthetestfixture.Theassembleddurability fixtureandtheresonancesearchplate(withasinglecellfixture installed)werebothmarkedwithasuitablegriddensitytoensure thatenoughmeasurementswererecordedtoaccuratelydetermine theirmodeshapes.Apictureofthethreedimensionalgridpattern andcorrespondingphotographsarepresentedinFigs.7and8,for the durability fixture and resonance search plate assembly respectively. The dimensions for the grid patterns (and subse-quently the basicdimensions of the fixtures)are presented in
Tables2and3.Theaxisconventionsforthemodalanalysistesting
isalsoillustratedinFigs.7aand8a.
Thetestingwasconductedusinga140g“BrüelandKjær”8206 impacthammerwithheadextender.Thiswasusedinconjunction with five multi axis “Brüel and Kjær TEDS 4535-B-001” accel-erometers.Thedataloggeranalyserusedwithinthisexperiment wasa“BrüelandKjær3053-B-120InputOutputController”which wascombinedwitha“BrüelandKjær3660typeCchassis”.The 0.14kghammerwithanaluminiumtipprovidedameasurement frequencyrangeofapproximately0–5000Hz.Thereasonwhythis
Fig.6. TransmissibilityofMASTwhenExcitedinZAxis.
frequencybandwaschosenwassothattheprimarymodescould bedeterminedwithinthetestfrequencyrangesof1to110Hz(for thedurabilityfixture)and5to3700Hz(fortheresonancesearch plateassembly).Thedatawassampledat3.2timesthedesired peakfrequencyinaccordancewithShannon’ssamplingtheorem. The weightof theaccelerometersmounted totheDUTwas no greaterthan0.001%ofthetotalweightofthelightestfixture,thus minimisingtheresultofexperimentalerrorbytheintroductionof additionalweight.Theaccelerometerswerefixedviatheuseof “petrowax”sothatnosurfaceorstructuralchangewasintroduced bytheapplicationofpermanentadhesives.
Eachfixturewasplaced onsuitablepolyurethanefoampads andtheresponseateachnodelocation(definedwithinFigs.7and 8)wasmeasuredwhenthefixturewasexcitedwithfiveimpactsat asinglepoint.Thefiveimpactswereaveragedtogenerateasingle inputandsingleoutputresponseforthegiventestlocationforthe excitedaxis.ThedurabilityfixturewasexcitedinX,YandZaxisby applyinganXaxisexcitationatnode45,aYaxisexcitationatnode 39andaZaxisexcitationatnode37.Theresonancesearchplate wasexcitedintheZaxisonlyatnode100asthisfixtureassembly was subjectedtomulti-axialvibrationduringthis study.Within the modal assessment, the durability fixture was evaluated without cells, whilst theresonance search platewas evaluated bothwithandwithout18650cellsinstalled.Thedurabilityfixture was evaluated without cells as the preliminary results of the fixturedisplayedasignificantlyhighdegreeofseparationbetween thetest frequencyrangeand thefirstnatural frequency. Itwas deemedthattheadditionalmassloadingprovidedbytheinclusion ofcellstothefixture(<1%)wouldnotimpactthesuitabilityofthe fixturefordurabilitytesting.
Withinthisinvestigationthedatawaspostprocessedusingthe “PULSEReflex Version 21” softwarewithin theBrüel and Kjær modal analyser to generate the desired information for the fixtures. A single FRF from each test position was evaluated together using itsglobal curvefittingapplication togeneratea singleFRFtracewhichrepresentedthemodalperformanceofthe wholemodule.The modalpropertiesofnaturalfrequencywere extractedbythePULSEReflexsoftware.Modeshapeswerealso estimatedfromtheFRF’susingthePULSEReflexsoftware.
Thefirstthreemodesandmodeshapesforthedurabilityfixture arepresentedinTable4andFig.9.Asdiscussedingreaterdetailin Section2.4,duringthemulti-axisdurabilitytest,thefixturewill beenexcitedtoapeakfrequencyof110Hz.Thefirstmodeofthe durability fixture occurs between 2070 and 2074Hz, which is approximately20timesgreaterthanthepeaktestfrequencyofthe durabilitytest.Thisindicatesthatthechanceofmodalexcitation withinthetestfixtureduringthedurabilitytestwillbelowand
Fig.8.a)TestSetUpandAxisConventionofResonanceSearchPlateforModalAnalysis,b)NodePositionsandNumbersofResonanceSearchPlate.
Table2
X:Y:ZNodeCoordinates/MeasurementsforDurabilityFixture.
X:Y:ZCoordinatesinmm
NodeNumber X Y Z
16 0 0 0
17 0 100 0
18 0 200 0
19 0 300 0
20 110 0 0
21 110 300 0
22 250 0 0
23 250 300 0
24 340 0 0
25 340 100 0
26 340 200 0
27 340 300 0
28 50 20 0
30 50 145 0
31 50 280 0
32 175 20 0
33 175 280 0
34 300 20 0
35 300 145 0
36 300 280 0
37 50 20 245
38 50 145 245
39 50 280 245
40 175 20 245
41 175 145 265
42 175 280 245
43 300 20 245
44 300 145 245
45 300 280 245
46 50 20 65
47 50 145 65
48 50 280 65
49 50 20 120
50 50 145 120
51 50 280 120
52 50 20 180
53 50 145 180
54 50 280 180
55 300 20 65
56 300 145 65
57 300 280 65
58 300 20 120
59 300 145 120
60 300 280 120
61 300 20 180
62 300 145 180
thatthefixtureissuitablefortesting.Theseresultsalsoindicatea goodlevelofcrossaxiscorrelationbetweenobservedmodeshapes andnaturalfrequencieswhenexcitedintheX,YandZaxis.
With respecttotheresultsfromtheresonance searchplate fixturepresentedinTable5andFig.10,thefirstmodeofthefixture
withcellsinstalledisapproximately167Hzgreaterthanthepeak frequency of the singleaxis EMS employedfor the sweptsine evaluationofthecells.Theinclusionofthecellswithinthisfixture increasedthefirstnaturalfrequencybyapproximately30Hz,and reducedthesecondnaturalfrequencybyapproximately6Hz.
In summary both fixtures, when evaluated using impact excitationinthefree–freecondition,hadfirstnaturalfrequencies andcorrespondingmodeshapesoutsidethefrequencyrangeofthe desired test excitation. However because the response of the fixtureswillchangewhenclampedtotheshakertables,additional assessments were conducted once installed totheir respective facilitiestoensuretheirsuitabilityfortesting.Theseassessments arediscussedinSection2.3.3.
2.3.3.Vibrationresponseof18650fixturesandresonancesearchplate whenassessedinaccordancewithBSEN60068
To ensure that the fixtures did not display any significant resonances once bolted to the test facility, all cell mounting fixtureswereevaluatedinaccordancewithBSEN60068priorto testing. This specification requires that the maximum vibration amplitudeinanyaxis perpendiculartothe specifiedaxis shallnot exceed50%ofthespecifiedamplitudeupto500Hz[40].The18650 fixturewasexcitedintheZ-axisviaasweptsinewaveofamplitude 1gnoverafrequencyrangefrom1to110Hzatarateof1octaveper
minuteontheTEAMCubeMAST.Themeasuredvibrationresponse inallthreeaxes,werewithinthelimitsspecifiedbyBSEN60068. The resonance search plate fixture was also evaluated in accordancewithBSEN60068andwasexcitedintheZaxisvia a1gnsweptsinefrom5to3700Hzatasweeprateof1octave/min
onthesingleaxisVP85EMS.Theresonancesearchplatewitha single18650threecellfixtureinstalledmettherequirementsofBS EN 60068from5 to3700Hz. The transmissibilityplots for the durability fixture and resonance search plate are presented in
Figs.11and12respectively.Additionalinformationrelatingtothe
resonancesearchplatefixtureevaluationispresentedin[13].Itis hypothesised that a drop in transmissibility within Fig.12 at approximately1550Hzistheresultintheinteractionbetweenthe
Table3
X:Y:ZNodeCoordinates/MeasurementsforResonanceSearchPlate.
X:Y:ZCoordinatesinmm
NodeNumber X Y Z
1 0 0 0
2 97.5 35 0
3 195 0 0
4 35 67.5 0
5 230 67.5 0
6 0 135 0
7 97.5 170 0
8 195 135 0
9 30 0 60
10 97.5 0 60
11 165 0 60
12 30 45 60
13 97.5 45 60
14 165 45 60
15 30 90 60
16 97.5 90 60
17 165 90 60
18 30 135 60
19 97.5 135 60
20 165 135 60
21 25 90 35
22 25 45 35
23 170 90 35
24 170 45 35
25 30 0 0
26 97.5 20 0
27 165 0 0
28 30 135 0
29 97.5 155 0
30 165 135 0
100 30 42 0
Table4
FirstThreeModesObservedwithinDurabilityFixturewhenExcitedinX,YandZAxis.
WhenExcitedinZAxis(Node37) WhenExcitedinYAxis(Node39) WhenExcitedinXAxis(Node45) ModeNumber Frequency(Hz) ModeShape Frequency(Hz) ModeShape Frequency(Hz) ModeShape 1 2070.56 Sidepanelspanting 2071.70 Sidepanelspanting 2074.91 Sidepanelspanting 2 2525.29 Toppanelpanting 2522.67 Toppanelpanting 2521.55 Toppanelpanting 3 2623.53 BaseTorsional 2618.37 BaseTorsional 2623.80 BaseTorsional
resonance search plate fixture and shaker table armature. All transmissibility evaluations on the fixtures were conducted withouttestsamplesinstalled.
2.4.RecreatingmeasuredvibrationcyclesfromEV’sin6DOFonMAST
With a 6DOF vibration test, measured vibration signals are appliedinthetimedomaintothesystemundertest.Thisstudy employedthevibrationmeasurementsfromtheSmartEDvehicle, which were recordedas partof theresearch presentedwithin
[2,45,46]. These measurements included the response of the
batteryassemblyasthevehiclewassubjecttospecificdurability surfacesattheMillbrookProvingGround(MPG)withintheUK.The Smart EDwas chosen for the6DOF cell durabilitystudy, asits
compact dimensions and suspension geometry results in high levelsofvibrationenergyatfrequenciesbelow5Hz.AlsotheXand Yaxisvibrationloadsaresignificantlyhigherthanothercurrent productionEVsmeasuredwithin[2,45,46].Furthermore,theSmart EDhasabatteryassemblyconstructedof18650typecells,which therefore has a greater correlation to the test samples under assessmentwithinthisinvestigation.
TheMillbrookStructuralDurability(MSD)testframework[47] wasemployedtodefinethenumberofrepeatedroadsurfacesthat wouldbesequencedtogethertoreplicateavehicledurabilitylife. This framework defines the durability surfaces and number of requiredrepeatsof eachsurface toreplicate10years oftypical European customer use. While this procedure represents an internalorganisationalstandard,ithasevolved over20 yearsof
Table5
FirstTwoModesObservedwithinResonanceSearchPlatewhenExcitedinZaxis,bothwithandwithoutCellsInstalled.
WhenExcitedinZAxiswithout18650Cellsinstalled(Node100) WhenExcitedinZAxiswith18650Cellsinstalled(Node100)
ModeNumber Frequency(Hz) ModeShape Frequency(Hz) ModeShape
1 3837.93 Bending 3867.22 Bending
2 4128.12 Bending 4122.21 Bending
Fig.10.ExamplesofModeShapesObservedinResonanceSearchPlate(withCellsInstalled)a)3867.22HzBending,b)4122.21HzBending.
experienceand is currently employed by a number of leading vehiclemanufacturerstoassesstheservicelifeoftheirproducts [45].
Unlikestandardsemployedinpreviouscellvibrationdurability studies,discussedin[13,21,28]andemployedbyotherresearchers, forexample[24],theMSDstandarddefineslifeinyearsasopposed to vehicle mileage. However, approximately 15,000 miles of provinggrounddriving represents10years oftypicalcustomer structuraldegradation[47].Asummaryofsurfacerepeatsrequired forthestandardprocedureisshowninTable6.
Testingforthisstudywaslimitedto150h.Toensurethatthe desirednumberofsurfacescouldbereplicatedwithina150htest
duration on the MAST; significant periods of “non-damaging” vibration(e.g.measuredvibrationthatwas lessthan0.1gn)was
deletedfromeachoftheaxis-signals.Thisisacommonmethod with regard toreducing test time and hence test costs [7]. To maximisetheavailabletesttime,additionalsurfacerepeatsforthe hillroute,handlingcircuitandhighspeedcircuitwereaddedtothe test schedule. The revised number of surfaces and associated numberofrepeatsaredefinedinTable6.Toensurea representa-tive even loading of surfaces was achieved the signals were replicatedinsurface-loops.Asurface-loopisdefinedasablockof multiplesurfaceswhichlastsapproximately30minandcontainsa weightednumberofrepeatsforeachsurfacedefinedinTable6.
Fig.12.TransmissibilityPlotofResonanceSearchPlateFixturefrom5to3700Hz(WithoutCellsInstalled).
Table6
SurfaceRepeatsfor10YearsEuropeanMillbrookStructuralDurability[47].
Surface Repeatsofsurfacerequiredfor10 YearsEuropeanstructuraldurability
[47]
Revisednumberof repeatsfor6DOF Study
Numberofrepeatsofsurfacefor 1loop(total300loops required)a
Editedduration ofsurface (seconds)
Totaldurationofsurfacewith surfacerepeatsforoneloop (seconds)
Hillroute(loop 1)
3365 3600 12 22 264
Citycourse 6570 6600 22 13 286
Twisthumps 1800 1800 6 33 198
Sinewaves 1204 1200 4 34 136
Randomwaves 1200 1200 4 85.5 342
Belgianpave 600 600 2 134 268
Catseyes–30 mph
600 600 2 29 58
Catseyes–50 mph
600 600 2 57 114
HSC 420 600 2 11 22
Handlingcircuit 219 600 2 110 110
Potholesb 54 60 1b 32 32
MileStraight–
Wideopen throttle
1200 0 0 0 0
MileStraight–
Partopen throttle
1200 0 0 0 0
Total – – – 1830
Eachloopofsignalswasrepeated300timestoachievethedesired 150hvibrationtestprofile.Thesurfaceswereappliedinblocksof sequencedsurfacestoensureanevenfatigueloadingduringthe testing. It is desirable to apply a balanced and mixed surface loadingsothatarepresentativeoperationallifeisreplicated,asitis unlikely that a vehicle will be exclusively subjected to one particulardrivingeventforanyprolongedperiodoftimeduring customeruse.
WhenreplicatingrecordeddatainthetimedomainonaMAST it is necessary to ensure that the six control multi-axis accelerometersareinstalledinthesameX,YandZlocationson thetestrigtothatofthemeasurementvehicle.Withinthisstudy, theupper, front rightcorner mounting holeon thecube head expanderwastakenasthe0:0:0coordinateoriginfortheX:Y:Z
measurementsforthecontrolaccelerometers.Thelocationsofthe accelerometersfromthemeasurement vehiclewerethen trans-posedtotheMASTexpander.Aphysicalframewasconstructedto ensure the accelerometers could be installed in the correct locationstoemulatethein-vehiclemeasurementlocations.The frameassemblyandtheaccelerometerlocationsarepresentedin
Fig.13andTable7.
Thisframewas suitablyridged toensurethat nodisrupting resonanceswithintheframeitselfoccurredwithinthedesiredtest frequencyrangeof1–110Hz.Duringdata-recordingthevibration signalswerefilteredwithalowpassfilterwithacut-offfrequency of 110Hz. Signal noise was clipped to 3gn in-line with the
capabilities of the shaker system. An example of the typical frequency response of the cube against the original measured
signalisshowninFig.14.PleasenotethatanASDofthesignals combinedtoreplicate100,000milesofdurabilityutilisedwithin thisstudyispresentedin[2].
2.5.Electrochemicalcharacterisationbeforetheapplicationof vibration
Thefollowingcharacterisationtestswereperformedonallcells
defined in Table 1 at SOT. Measurements were recorded in a
thermallycontrolledenvironmentatatemperatureof21C0.5 C.Theexceptiontothiswasthenaturalfrequencymeasurements whichwereconductedinalaboratoryenvironmentata tempera-tureof21C3C.
2.5.1.SOCadjustmentprocess
CellSOCwasadjustedbyfullychargingthecellswithaconstant current (CC) of amplitude 1.1A–4.2V followed by a constant voltage(CV)chargingphaseuntilthecurrentfellto0.05A.Atthe endofcharge,thecellswereallowedtorestfor4hpriortobeing dischargedat1Cfor15minor30mintoachieveaSOCof75%or 50%respectively.Thecellswerethenallowedtorestforafurther 5htoensurethattheyhadreachedequilibrium.
2.5.2.1CandC/3capacitymeasurement
ThecellswerefullychargedusingtheCC-CVprocessdefinedin Section2.5.1.Thecellswereallowedtorestfor4hpriortobeing dischargedat1Ctothemanufacturer’sdefinedcut-offvoltageof
Table7
AccelerometerX:Y:ZCo-ordinatesonHeadExpander.
X:Y:ZCoordinateMeasurementsonShakerTable(inmm)
AccelerometerLocationonVehicle X Y Z
RightHandSide“A”PostAccelerometer 0 0 530
LeftHandSide“A”PostAccelerometer 0 1260 525
BatteryAccelerometer–Front 580 550 0
BatteryAccelerometer–Rear 840 190 0
ChassisAccelerometer–Front 100 705 60
RearChassisAccelerometer–Rear 980 10 160
Fig.14.SmartED–CatsEyesat30MPH–DesiredSignalvsAchievedSignal–ZAxis.
2.75V.Theenergyextractedwasrecordedasameasureofthecell’s 1Cenergycapacity.Aftera4hresttheprocesswasrepeated,but thedischargecurrentwasreducedtoC/3andtolowervoltagelimit of2.75V.Theenergyextractedwasrecordedasameasureofthe cell’sC/3energycapacity.
2.5.3.Pulsepowerdischargeresistance
TodeterminetheDC resistanceofthecells (RDC), aseriesof
current pulses were applied to the cells after they were conditionedto50% SOC. Each current pulseis 10sin duration witha magnitudeof20, 40,60,80 and100%ofthecell’srated maximumpulsedischargecurrentof4.4A[28].Between succes-sivedischargepulses,thecellswereallowedtorestfor30min.The DCresistanceofthecellwasapproximatedusingEqs.(2)and(3) foreachexcitationpulse(n);whereVOCVdefinesthevoltageprior
totheapplicationofthecurrentpulse(Imax)andV10sisthecell
voltageattheendofthe10scurrentpulse[28].Anaveragevalueof RDCwascomputedasthemeanvalueofresistancefromeachofthe
fivepulsesasperEq.(3).
RDC¼
VOCVV10s
ð Þ
Imax ð
2Þ
RDC¼ X n¼5
n¼1 RDC
n ð3Þ
2.5.4.EISmeasurement
TheEIS measurementwasundertakeningalvanostaticmode usingModuLab1electrochemicalsystemmodel2100Afittedwith a2AboostercardanddrivenbyModulab1ECSsoftware[28].The EISspectrawerecollectedwithinthefrequencyrangeof10mHzto 10kHz with10frequencymeasurementpointsperdecade [28]. The applied route mean square (RMS) value of the excitation currentwassetto200mA.
2.5.5.Naturalfrequencymeasurement
Asinglepointnaturalfrequencymeasurementwasundertaken asatthestartandendoftestasamethodofcomparisonofthecells mechanicalperformance.Thenaturalfrequencyofeachcellwas measuredbyfasteningtherespectivecelltoaVP85 electromag-neticshaker(EMS)tableandapplyingasweptsinewavefrom5to 3700Hz,ofamplitude1gnatarateof1octave/min[13].Asshown
inFig.15,theresponseofthecellinrelationtothis1gnexcitation,
was recoded via a lightweight, single-axis accelerometer. The measurementaccelerometers(modelPCB352C65)weresecured tothecentreofthecellviaathreadedaluminiumcollarwhichwas bondedtothecellsurface.Thesewereattachedtothecellviathe use of HBM X60 adhesive. A strip of Capstantape was placed between the cell and the adhesive to enable removal of the adhesive and collar post testing whilst minimising the risk of damage to the cell wall. This test arrangement resulted in a combined additional weight of approximately 2.6g (circa: 6% additional mass for each cell). Their inclusion within the experimental set-up was not deemed to have any significant impactontestaccuracythroughtheadditionofincreasedmass. Whenthecellswereinstalled,thecollarswerelevelledusinga digitalinclinometer.Thiswas toensurethatno“offaxis” errors would occur. Two control accelerometers (model PCB 352A24) weresecuredatoppositeendsofthetestfixture,butclosetothe specimens(asshowninFig.15)usingpetro-wax.
An averaging strategy of the control accelerometers was employedduringthenaturalfrequencymeasurement.Datawas recordedat2.5timesthedesiredpeakfrequencyin accordance withNiquest rate guidelines. Eachsweepwas performed three timesandanaverageresponsewasrecorded.
2.6.Applicationofvibration
The vibration profiles, discussed within Section 2.4 were appliedtothecells(samplenumbers1–9)in6DOFfor150hin accordancewiththesurfacerepeatsdefinedinTable6.Thecells
Table8
ChangeinPulsePowerPerformance–DCResistance.
CellID Orientation DCResistance (SOT)(mV)
DCResistance (EOT)(mV)
ChangeinDC Resistance(mV)
PercentageChangeinDCResistanceDifference BetweenSOTandEOT(%)
RankingWorsttoBest: 1=Worst,9=Best
1 Z 40.56 43.51 2.95 7.28 3
2 Z 40.82 43.80 2.98 7.31 2
3 Z 40.64 43.33 2.69 6.62 5
4 Y 41.08 43.36 2.28 5.55 8
5 Y 40.91 43.45 2.54 6.22 7
6 Y 40.38 43.13 2.75 6.81 4
7 X 41.32 43.95 2.63 6.37 6
8 X 40.00 43.51 3.51 8.77 1
9 X 41.32 43.42 2.11 5.10 9
10 Control 41.90 42.98 1.08 2.58 –
11 Control 42.22 42.92 0.70 1.66 –
12 Control 41.37 42.66 1.29 3.11 –
PulsePower–MeanChange
MeanChange(mV) MeanChange(%) Ranking
MeanChangeinPulsePowerDCResistance(mV)–X 2.75 6.75 2
MeanChangeinPulsePowerDCResistance(mV)–Y 2.52 6.19 3
MeanChangeinPulsePowerDCResistance(mV)–Z 2.88 7.07 1
MeanChangeinPulsePowerDCResistance(mV)–Control 1.02 2.45 –
PulsePower–ANOVAAnalysis
Orientation ANOVAp-valueagainstControlNullHypothesis:Meanofvibratedcellsandcontrolcellsareequal.Rejectnullhypothesisifp<0.05)
X 0.018
Y 0.002
werecheckedevery4hforanysignificantchangesintemperature oranyemissionofgasesviatheremotemonitoringsystems,as discussedin2.2.2.
2.7.Electrochemicalcharacterisationaftertheapplicationofvibration
Attheend of the150hdurabilityvibrationprofile, thecells were left to rest for 4h prior to visual inspection. EOT characterisation of the cells, both in terms of the electrical performanceand mechanical attributeswas then conductedby repeatingthetestsdefinedinSection2.5.
3.Celldegradationresults
The following section introduces the data to quantify the electromechanicalattributesforallcells(samplenumbers1–12)at SOTandEOT.Trendsinthemeasurementsaremadefortestingof the NCA 18650 cells subject to 150h of vibration in 6DOF. A discussionispresentedthathighlightstheeffectofvibrationon cell performance and identifies if cell orientation can impact vibrationrobustness.
3.1.Posttestexternalconditionofcellsthroughvisualinspection
Posttesting,nosignificantmechanicaldamageordegradation wasobservedinanyofthetestsamples,thisincludednoleakingor expulsionofelectrolytewaswitnessed.
3.2.EstimationofDCresistancethroughpulsepowermeasurements
Allsamplesillustratedanincreaseinresistancepostvibration testing.Table8illustratesthechangeinestimatedDCresistance fromthepulsepowertest(PPT)forallsamples(includingcontrol samples). Based on established methods of benchmarking laboratory processes and facilities, the experimental error associatedwithpulsepowerDCresistanceestimation on18650 cells is 0.53m
V
[48]. Whilst the standard error (e.g. the confidence in the sample mean resulting fromthe experimentduetopastconfidencewiththeexperimentandcell-type)is0.62% [48].
From Table 3, the worst performing cell was sample 8
(orientatedalong theX-axis)whichshows an8.77%increasein DCresistance.ThecellwiththeleastdegradationinDCresistance postvibrationwassample9(orientatedalongtheX-axis)which hasa 5.10%increase. With all samplessubject to vibration, an increaseinpulsepowerresistancewasobservedwhichpotentially indicatesadecreaseincurrentcollectorcontactarea,possiblyasa resultofdelaminationorcrackingofinternal electrodesurfaces
[21]. A mean increase in DC resistance of 2.45% was observed
withinthecontrolsamplescomparedtoameanof7.07%,6.19%and 6.75% for the Z, Y and X oriented samples respectively. This suggeststhatvibrationhasastatisticallysignificanteffectonthe cell performance. Thisis confirmedby theanalysis ofvariance (ANOVA)presentedinTable8.BasedontheANOVAanalysisthe followinghierarchyof orientationrelatedperformance is deter-minedfromtheseresultsatthe95%confidencelevel:Y<X<Z.
3.3.EISresults
Tables9and10showtheohmicresistance(RO)andthecharge
transferresistance(RCT)ofthecellsatSOTandEOT.Acomplete
interpretationofEISresultsisbeyondthescopeofthisstudy,butis discussedwithin[49,50]forreference.Bymeansofanexample, Fig.16presentstheNyquistplotofsample4,beforeandafterthe applicationofthevibrationdurabilitytest.
The results illustrate the typical characteristics of the cell capturedbytheEISmeasurement.Table9highlightsthatallcells, includingthecontrolsamples,exhibitanincrease inROatEOT.
Sample6(orientatedalongtheY-axis)exhibitsthegreatestchange in RO of 2.12m
V
(9.39%). However, sample 4 which was alsoorientatedintheYaxisexhibitedtheleastchangeinROof0.99m
V
(4.20%).AnincreaseinRotypicallyoriginatesfromanincreasein
cell contact resistance, possibly through delamination of the material layers [49,50] or due to damage within the current collectors. However, the mean change in Ro resistance for the controlsamplesis1.59m
V
(6.92%),comparedto1.57mV
(6.89%),Table9
StartandEndofTest–RoMeasurements.
CellID Orientation SOT(mV) EOT(mV) ChangefromSOTandEOT(mV) PercentageChange(%) OverallRanking:1=Worst,9=Best
1 Z 22.99 24.45 1.47 6.39 4
2 Z 23.12 24.90 1.78 7.68 3
3 Z 23.26 24.60 1.34 5.78 6
4 Y 23.57 24.56 0.99 4.20 9
5 Y 23.25 24.61 1.35 5.82 5
6 Y 22.58 24.70 2.12 9.39 1
7 X 23.24 24.57 1.33 5.74 7
8 X 22.35 24.43 2.09 9.33 2
9 X 23.23 24.54 1.31 5.62 8
10 Control 22.66 24.12 1.45 6.41 –
11 Control 23.27 24.89 1.62 6.96 –
12 Control 22.77 24.46 1.69 7.41 –
RO–MeanChange
MeanChange(mV) MeanChange(%) Ranking
MeanChangeinRO(mV)–X 1.57 6.89 1
MeanChangeinRO(mV)–Y 1.49 6.47 3
MeanChangeinRO(mV)–Z 1.53 6.62 2
MeanChangeinRO(mV)–Control 1.59 6.93 –
RO–ANOVAAnalysis
Orientation ANOVAp-valueagainstControlNullHypothesis:Meanofvibratedcellsandcontrolcellsareequal.Rejectnullhypothesisifp<0.05)
X 0.967
Y 0.786
1.49m
V
(6.47%) and 1.53mV
(6.62%) for the X, Y and Z orientations respectively. This indicates that whilst vibration mayhavehadaneffectonthiselectromechanicalproperty,itisnot possible to isolate the level degradation observed from other ageing mechanisms, for example those associated with cell storage.ANOVAanalysisofthesignificanceofthemeanchange inohmicresistanceofthetestedcells inrelationtothecontrol samplesisshowninTable9.Basedonthisstatisticalanalysis,there isnosignificantchangeinROforanyofthethreecellorientationsatthe95%confidencelevelasaresultoftheapplicationofvibration in6DOF.
The RCT results presented in Table 10 demonstrate that all
samples(exceptsample7),includingthecontrolsamples,showa decreaseinRCTposttesting.
ThisreductioninRCTcouldbeafunctionofimprovedanodeand
cathode “wetting” over the duration of thetest [51]. However, giventhatthecontrolsamplesdisplayasimilarimprovement,this proposed mechanism is unlikely tobeas a functionof applied vibration.SamplesevaluatedintheYorientationshowasimilar averagereductioninRCTasthecontrolsamples.Theyalsohavea
significantly higher mean change of 25.89% than the X axis orienteditems(3.95%).TheZ-axisorientedsampleshavealower improvementtothatoftheYorientedsamplesandthecontrolsof 14.13%. Reviewing the ANOVA analysis results presented in
Table 10, samples mounted in the X axis orientation show a
significantdifferenceinRCTasaresultofvibrationwhencompared
tothecontrolsamples.ItisnoteworthythatRCThasreducedinall
samplespostvibrationtesting.ThereductionwaslowerinXaxis
Table10
StartandEndofTestRCTMeasurements–ItemsinItalicsIndicateaReductioninRCT.
CellID Orientation SOT(mV) EOT(mV) ChangefromSOTandEOT(mV) PercentageChange(%) OverallRanking:1=Worst,9=Best
1 Z 14.38 11.44 2.94 20.47 6
2 Z 15.64 13.97 1.67 10.68 4
3 Z 15.64 13.88 1.76 11.24 5
4 Y 16.57 12.79 3.78 22.83 7
5 Y 17.42 12.04 5.38 30.86 9
6 Y 16.46 12.52 3.95 23.99 8
7 X 12.26 12.38 0.12 0.95 1
8 X 13.73 12.50 1.24 9.01 3
9 X 17.33 16.67 0.66 3.79 2
10 Control 21.31 16.27 5.04 23.64 –
11 Control 22.25 14.62 7.63 34.31 –
12 Control 19.03 16.20 2.83 14.86 –
RCT–MeanChange
MeanChange(mV) MeanChange(%) Ranking
MeanChangeinRCT(mV)–X 0.59 3.95 1
MeanChangeinRCT(mV)–Y 4.37 25.89 3
MeanChangeinRCT(mV)–Z 2.12 14.13 2
MeanChangeinRCT(mV)–Control 5.17 24.27 –
RCT–ANOVAAnalysis
Orientation ANOVAp-valueagainstControlNullHypothesis:Meanofvibratedcellsandcontrolcellsareequal.Rejectnullhypothesisifp<0.05)
X 0.033
Y 0.618
Z 0.103
orientatedsamples,resultingin a significanteffectof vibration withinthese samples.In summary,assessing the mean change withregard to RCT, the performance of cellorientation can be
summarisedasfollows:Y=Z<X.
3.4.OCVmeasurement
NostatisticallysignificantchangesintheOCVmeasurements were observed between SOT and EOT. Typically, a change of 0.026%to0.05%wasobserved.AsdiscussedwithinSection3.2, this difference is within the error limits of the measurement method.Theseresultssupportthosepresentedin[22,23]thatalso notedthatOCVisnotadverselyaffectedbyvibrationloading.Due tothelevelofchange,theOCVresultsandANOVAanalysishave beenomittedfromthispaper.However,nosignificantchangein OCV,asaresultofvibration,wasobservedatthe95%confidence levelbetweencellsamples1–9andthecontrolcells(samples10– 12).
3.5.1Cdischargecapacityresults
All samples, including the control samples, illustrate a reductionin1Ccapacitypostvibrationtesting.Theresultsfrom the1CdischargeevaluationareshowninTable11.It isevident fromtheseresultsthatthereductionincapacityobservedinthe controlsamplesisgreaterthanthatobservedwithinsamples1–9. Thisisalsoevidentwhenviewingthemeanchangeincellcapacity ofeachcellorientationis comparedtothemeanof thecontrol samples.InterestinglytheANOVAanalysisoftheYaxissamples indicatesthatthereisastatisticallysignificantchangeincapacity performanceasaresultofvibration.However,thisisduetothe greaterreductionof1Ccapacityobservedinthecontrolsamplesas opposedtothedirecteffectofvibrationloading.
3.6.C/3dischargecapacityresults
In a similar manner to the 1C capacity discharge results discussedin3.5allsamples,includingthecontrolcells,displaya reductioninC/3capacitymeasurementpostvibrationdurability testing.ThisisillustratedinTable12.Anoteworthyobservationis thattheC/3performanceoftheYaxissamplesis,onaverage,the worstperforming orientation.Thisis likely tobeasa resultof sample6whichhasahigherlevelofcapacityreduction(4.22%). Thecontrolsamplesalsoexhibitahigherreductionincapacity atEOT.Thisindicatesthatfromacapacityreductionperspective there may be some evidence to suggest that cells excited to vibrationin6DOFmaydegradeataslowerratethancellsleftina staticcondition.However,itmustbenotedthatfurthertestingis requiredtoconfirmthishypothesis.TheANOVAanalysispresented
in Table12,highlights that there is nosignificanteffectto C/3
capacityforanyoftheorientationsasaresultofvibrationin6DOF.
3.7.Resonancesearchviasweptsineresults
NochangeinnaturalfrequencybetweentheSOTandEOTwas observed. Whilst it is likely that no significant mechanical degradation occurred which could have affected the natural frequency, it is noteworthythatthe resonanceof thecells was greater than the frequencyrange capabilities of theEMS table which was used for the natural frequency measurement. Subsequentlythefirstnaturalfrequencyofallsampleshasbeen recordedas3700HzwithinTable13.
Withregard tothe amplitudeof thefirst resonantfrequency (whichwasrecordedas3700Hz)asshowninTable14,thereisa significantchangeinthemajorityofcellsindicatingachangein damping within the cell assembly. The greatest change in amplitudewasobservedwithinsample6(Y-axis),whichhadan increase of 39.22%, highlighting a reduction in cell damping. Generally, Y axis oriented cells exhibit a greater reduction in
Table11
RankedChangeinCapacityofAllCellsItemsinItalicsIndicateaReductioninCapacity.
CellID Orientation SOT(Ah) EOT(Ah) ChangefromSOTandEOT(Ah) PercentageChange(%) OverallRanking:1=Worst,9=Best
1 Z 2.97 2.88 0.09 3.03 1
2 Z 2.98 2.90 0.08 2.68 5
3 Z 2.94 2.92 0.02 0.68 =9
4 Y 2.98 2.91 0.07 2.35 6
5 Y 2.97 2.94 0.03 1.01 7
6 Y 3.01 2.92 0.09 2.99 3
7 X 2.95 2.87 0.08 2.71 4
8 X 2.99 2.90 0.09 3.01 2
9 X 2.94 2.92 0.02 0.68 =9
10 Control 3.06 2.92 0.14 4.58 –
11 Control 3.08 2.94 0.14 4.55 –
12 Control 3.07 2.97 0.10 3.26 –
MeanChange–1CDischarge
MeanChange(mV) MeanChange(%) Ranking
MeanChangein(Ah)–X 0.06 2.13 =1
MeanChangein(Ah)–Y 0.06 2.12 2
MeanChangein(Ah)–Z 0.06 2.13 =1
MeanChangein(Ah)–Control 0.13 4.13 –
ANOVAAnalysis–1CDischarge
Orientation ANOVAp-valueagainstControl
NullHypothesis:Meanofvibratedcellsandcontrolcellsareequal.Rejectnullhypothesisifp<0.05)
X 0.0687
Y 0.0457
dampingthanothercellorientations. Thisobservationcouldbe due to a redistribution of electrolyte withinthe cell’smaterial layers due to vibration. However, further studies must be conductedtodeterminetheunderpinningreason.Assessingthe mean changewithregard totheamplitude of thefirst natural frequency,theperformanceofcellorientationcanbesummarised asfollows:X<Z<Y.ANOVAanalysiscouldnotbeperformedfor themechanicalcharacterisationresults,becauseasdiscussedin Section2.1 thecontrolsampleswerenotsubjecttoaresonance search.
Table15presentsthepost-testdimensionsofthecellswhen
comparedtothechangeinamplitudeofthepeaknaturalfrequency measured.A noteworthyobservationis thatcells witha longer lengthtypicallyhaveagreaterchangeofamplitudebetweenthe startandend oftest.Anotherobservationisthatthelengthsof samples1,3,4and8areoutsidethemanufacturer’stoleranceof maximumlengthtoleranceof65.3mm.
4.Discussion
4.1.Effectofvibrationin6DOFon18650NCALi-ioncells
The primary conclusion from this study is that both the electricalperformanceandthemechanicalpropertiesoftheNCA Li-ioncells employedarerelativelyunaffectedwhenexposedto vibrationenergythatiscommensuratewithatypicalvehiclelife. However,therearesomenuanceswithinthedatathatareworthy offurtherdiscussion.
4.1.1.Theimpactofvibrationonelectricalcellperformance It’snoticeablefromtheANOVAanalysisinSection3.2(Table8) that the value of RDC was significantly impacted by vibration
applied in 6DOF when compared to the control samples. This indicatesthatthecellsmayhavesomedegreeofdamagetothe
Table12
RankedChangeinC/3CapacityofAllCells–ItemsinItalicIndicateaReductioninCapacity.
CellID Orientation SOT(Ah) EOT(Ah) ChangefromSOTandEOT(Ah) PercentageChange(%) OverallRanking:1=Worst,9=Best
1 Z 2.98 2.90 0.08 2.68 3
2 Z 3.03 2.95 0.08 2.64 4
3 Z 3.03 2.98 0.05 1.65 7
4 Y 2.98 2.92 0.06 2.01 5
5 Y 3.02 2.97 0.05 1.66 6
6 Y 3.08 2.95 0.13 4.22 2
7 X 2.91 2.88 0.03 1.03 8
8 X 3.07 2.93 0.14 4.56 1
9 X 2.95 2.92 0.03 1.02 9
10 Control 3.02 2.90 0.12 3.97 –
11 Control 3.09 2.94 0.15 4.85 –
12 Control 3.04 3.01 0.03 0.99 –
MeanChange–C/3Discharge
MeanChange(mV) MeanChange(%) Ranking
MeanChangein(Ah)–X 0.07 2.20 3
MeanChangein(Ah)–Y 0.08 2.63 1
MeanChangein(Ah)–Z 0.07 2.32 2
MeanChangein(Ah)–Control 0.10 3.27 –
ANOVAAnalysis–C/3Discharge
Orientation ANOVAp-valueagainstControl
NullHypothesis:Meanofvibratedcellsandcontrolcellsareequal.Rejectnullhypothesisifp<0.05)
X 0.552
Y 0.673
Z 0.468
Table13
SummaryofChangeinNaturalFrequencyofObservedFirstCellResonance.
FirstResonanceFrequency(Hz)
ISRnumber Orientation SOT EOT Change Change(%) Ranking:1=Worst,9=Best
1 Z 3700 3700 0.00 0.00 =
2 Z 3700 3700 0.00 0.00 =
3 Z 3700 3700 0.00 0.00 =
4 Y 3700 3700 0.00 0.00 =
5 Y 3700 3700 0.00 0.00 =
6 Y 3700 3700 0.00 0.00 =
7 X 3700 3700 0.00 0.00 =
8 X 3700 3700 0.00 0.00 =
9 X 3700 3700 0.00 0.00 =
MeanChange(Hz) MeanChange(%) Ranking
MeanChangeFirstResonanceFrequency–X 0.00 0.00 =
MeanChangeFirstResonanceFrequency–Y 0.00 0.00 =
currentcollectororcellmateriallayersasaresultofthemulti-axis vibrationloading.However,this findingisnotconfirmedbythe degreeofchangeobservedwiththemeasurementofRO.LikeRDC,
ROis alsotakenwithinsomestudiesasanindicationofcurrent
collectorcondition and material layer integrity [52,53]. Whilst thereisdegradationwithintheobservedaveragevaluesforeach orientation,thedifferencebetweenthecontrolandtesteditemsis minimal.Thislimiteddifferencebetweenthecontrolandtested itemsimpliesthatvibrationisthenon-significantcontributorto theincreaseinRO.Tobetterunderstandwhatmaybeoccurring
within these cells furthertesting via non-destructive methods (such as computer tomography (CT) scanning) followed by
chemical and microscopic analysis of the cell layers may be employed. Exampletest methods that may beappropriate are discussedfurtherwithin[21].
AnotherobservationwithrespecttothechangeinRDC,ROand
alsocapacity(both1CandC/3)showarelativelylineardegradation post testing regardless of cell orientation. In previous results publishedwithin[13,28]therehasbeenasignificantdifferencein the performance of different cell orientations. However, unlike previousstudies(wherecellshavehadaseparatevibrationprofile appliedforeachvehicleaxisandthecellshavebeensubsequently rotatedona rig toachieve thecorrect loading), this studyhas appliedthevibrationinamorerepresentativemannerwhereall
Table14
SummaryofChangeinAmplitudeofObservedFirstCellResonance.
AmplitudeatFirstResonance(gn)
Samplenumber Orientation SOT EOT Change Change(%) Ranking:1=Worst,9=Best
1 Z 1.66 1.82 0.16 9.64 4
2 Z 1.61 1.68 0.07 4.35 8
3 Z 1.70 1.81 0.11 6.47 6
4 Y 1.80 1.97 0.17 9.44 5
5 Y 1.70 2.24 0.54 31.76 2
6 Y 2.04 2.84 0.80 39.22 1
7 X 1.70 1.78 0.08 4.71 7
8 X 1.54 1.70 0.16 10.39 3
9 X 1.89 1.90 0.01 0.53 9
MeanChange(gn) MeanChange(%) Ranking
MeanChangeinAmplitudeinFirstResonance–X 0.08 5.21 3
MeanChangeinAmplitudeinFirstResonance–Y 0.50 26.81 1
MeanChangeinAmplitudeinFirstResonance–Z 0.11 6.82 2
Table15
DimensionsofCellsPostTestinginRelationtoChangeinAmplitude.
Sample number
PositiveEndof CellDiameter (mm)
NegativeEndof CellDiameter (mm)
Lengthof Cell (mm)
PositiveEndofCellDiameter Ranking:Biggest=1, Smallest=9
NegativeEndofCell Ranking:Biggest=1, Smallest=9
LengthofCell: Biggest=1, Smallest=9
ChangeinAmplitude Ranking:Worst=1, Best=9
1 17.96 18.20 65.65 4 1 1 4
2 17.92 18.03 65.24 9 9 8 8
3 17.94 18.13 65.35 6 4 4 6
4 17.99 18.11 65.44 1 6 2 5
5 17.97 18.08 65.26 2 7 6 2
6 17.93 18.17 65.28 8 2 5 1
7 17.96 18.08 65.25 5 8 7 7
8 17.94 18.12 65.40 7 5 3 3
9 17.97 18.14 65.16 3 3 9 9
Table16
ComparisonofCellPerformanceRankingbyPostTestAssessment.
ElectricalCharacterisation MechanicalCharacterisation Cell Orientation PulsePower
ResultsRanking
EISResults RankingRo
EISResults RankingRCT
OCV Results Ranking
Capacity ResultsRanking 1C
CapacityResults RankingC/3
ResonanceResults FrequencyRanking
ResonanceResults AmplitudeRanking
1* Z 3 4 6 Nochange 5 3 Nochange 4
2 Z 2 3 4 Nochange 3 4 Nochange 8
3* Z 5 6 5 Nochange 8 7 Nochange 6
4* Y 8 9 7 Nochange 3 5 Nochange 5
5 Y 7 5 9 Nochange 5 6 Nochange 2
6 Y 4 1 8 Nochange 1 2 Nochange 1
7 X 6 7 1 Nochange 7 8 Nochange 7
8* X 1 2 3 Nochange 2 1 Nochange 3
9 X 9 8 2 Nochange 8 9 Nochange 9
degrees of motion are applied simultaneously. An interesting observationfromthisstudyisthedecreaseinRCT.Asdiscussedin
Section3.3,giventhatthecontrolsampleshavenotdisplayeda significantdecrease,itisunlikelythatthisreductionisaresultof vibration.Asobservedwithin[13,28]theOCVshowsnosignificant changeposttestingregardlessofchangestotheothermeasured attributes.
4.1.2.Theimpactofvibrationonmechanicalcellperformance Forthemechanical characterisationof thecells,noneof the cells showed a significant change in natural frequency. This supportsthegeneralfindingoftheelectricalcharacterisationdata, indicating that vibration has had a minimal impact on cell performance.However,ashighlightedinSection3.7,thenatural frequencyofthecellswasoutsidethecapabilityoftheshakertable used for thenatural frequency assessment. With regardto the changeofamplitudeofthecellsitisnoteworthythatachangein dampingwasobservedindicatingsomechangeinthestructural stiffnessmayhaveoccurred.
4.1.3.SummaryofEOTelectricalandmechanicaltestresults
Table16showstherankedperformanceofeachcellfromthe
eightsetsofcharacterisationdata.Interestinglywithinthisstudy thereissomeevidenceofconsistentlypoorperformingcellswhen rankedpertest.Forexample,sample9(testedintheXorientation) istypicallyoneofthebestperformingcells forallassessments. Converselysample 6 (subjecttovibration in theYorientation) generally is the worst performing cell. There is also some correlationbetweenthe electrical characterisationperformance and the change in resonance amplitude which has not been witnessedinanyof theprevious 18650cellvibrationdurability studies[13,28].Thereareseveralpossiblefactorsfortheincreasein theobservedcorrelationwithinthistestbetweenelectricaland mechanicalcharacterisationmethods.Firstly,thevibrationsignals withinthisstudyareareplicationofactualEVbattery measure-mentsinthetimedomain.Thevibrationspectrawithin[13,28]is appliedinarandomnaturewithingivenspectralparametersmay result in a greater variation in level degradation observed. Secondly, the test items within this study are evaluated with respecttogravity.Within[13,28]asingleaxisverticalshakerwas usedandthesampleswererotatedonthefixturetoachievethe desiredloading.Whilstthisisindustrypractice,theprocessmay resultinunrepresentativeloadingastheeffectofgravityisnot considered[29].
With respect to orientation within the vehicle, Table 17 presents a summary of each assessment’s results’, for each packaging axis. The data illustrated in Table 17 suggests that thereisnoclearoverridingorientationthatisconsistentlyworse
with respect to cell degradation. What is noticeable from this tabulateddatasetisthattheeffectisequalinalargeproportionof theperformancetests.Thisindicatesthatthecellsarepotentially robusttodifferencesofin-vehiclepackagingorientation.
4.1.4.Implicationsforvehicledesign
Whilst the cells evaluated within this study were typically unaffectedbyroadvibrationexcitationrepresentativeof10year life, there were some specific aging behaviour such as the statisticalsignificantincreaseinDCresistance(derivedfrompulse power testing)observed. Any aging behaviouras a function of vibration would have to be characterized to ensure effective battery management system (BMS) development, to maximize useful service life and to minimise potential warranty related issues.
The test methodology presented within this paper allows engineerstodeterminehowsusceptibleaproposedcell technolo-gyis tovibration. Furthermore,theunderpinning methodology withinthispapercouldalsobeappliedtootherchassismounted electrical components within the battery assembly, such as contactors, bus bars, relays, electrical control units and power electronicdevices.
5.FurtherWork
Whilstthisstudyshowsthatvibrationhasalimitedeffecton theperformance ofthecommerciallyavailable18650 NCAcells employed,it would be beneficial totest othercell chemistries, form-factorsandmodulesystemsusingthesamemethodologyto determinetheirrobustnesstomechanicalexcitation.Thiswould highlightthetransferabilityoftheseresultstoothertechnologies employed within the EV sector. One of the limitations of the methodology employedwithin this study is that electrical and mechanicalcharacterisationdatawasonlymeasuredatSOTand EOT.Asaresult,nodiscussionorconclusionscanbemadeabout the rate of degradation that may be observed throughout the vehicle’slife.Whilstitcanbeconcludedfromthisstudythatno significantdegradationwasobserved,thismaynotbethecasefor other technologies. It is therefore recommended that future studiesshouldcharacterisethecellsatintermediatepointsduring thetestprogramme,e.g.intervalsrepresentativeof10,000milesor for each year of vehicle use. This would facilitate further investigation into both the absolute value of degradation, but alsotheexpectedin-servicerateofperformancereduction.This study recommends that an alternative method of natural frequencymeasurementof18650cellsisinvestigated.Candidate methods include modal analysis via Laser Doppler Vibrometer measurementtechniques,suchasthosediscussedwithin[54,55]
Table17
AssessmentRankingofOrientationbyTest.
Assessment Test OrientationRankingByAssessment
LeastChange GreatestChange
ElectricalCharacterisation
PulsePower 6DOF Y X Z
EIS(Ro) 6DOF X=Y=Z
EIS(RCT) 6DOF Y=Z X
OCV 6DOF X=Y=Z
Capacity–1CDischarge 6DOF Y Z=X
Capacity–C/3Discharge 6DOF X=Y=Z
MechanicalCharacterisation
Resonance(ChangeinFrequency) 6DOF X=Y=Z
orthroughtheuseofasmallersingleaxisshakerwithahigher peakfrequencycapability.Thiswouldallowthevalidationofthe cellsinafree–freecondition,bothpreandposttesting.Thiswould ensurethe effects of thefixture couldbe eliminated fromthe analysisandwouldallowfortheidentificationofmodeshapesof thetestitem.
6.Conclusions
Boththeelectricalperformanceandthemechanicalproperties ofthecommerciallyavailableNCA18650Li-ioncellemployedin thisstudytypicallyshowednostatisticallysignificantdegradation asaresultofvibrationappliedin6DOF.Furthermore,noparticular cell orientation consistently displayed a significantly greater reduction in cell performance or a significant change in cell mechanical properties. Within the context of this study, cell characterisationwithintheelectrical domainwas donethrough quantificationofthecell’simpedance,theopen-circuitpotential andthecell’senergycapacity.Astatisticallysignificantchangein RDCwasobservedpostvibrationtestingonallNCA18650’sforall
threecellorientations.Theorientationwiththegreatest suscepti-bilitytoachangeinRDCasaresultofvibrationin6DOFwastheZ
axis.Theremainderofelectricalattributesemployedto character-isecellperformanceallindicateeithernostatisticallysignificant changeorareductioninperformancethatmaynotbeattributed directlytovibrationloadingofthecell.Themechanicalproperties ofthecellwereinferredthroughmeasurementofthecell’snatural frequency of vibration and the associated damping ratio and stiffness. With regard to mechanical integrity, no significant externaldamageorelectrolyteleakagewasobservedinanyofthe testedcellspostvibration.Nochangeinsamplenaturalfrequency wasobserved.However,samplesorientedintheYaxisdisplayeda significantreductioninnaturalfrequencyamplitudepostvibration indicating a possiblechangein cell stiffness.Within this study therewasadegreeofcorrelationbetweencellnaturalfrequency amplitudeand a change inthe cells’ electrical performance. In conclusion,theresultspresentedinthispaperhighlightthatthis particularcelldesign,onethatisalreadybeingusedorinvestigated by many leading vehicle manufacturers, is largely robust to vibrationexcitationthatiscommensuratewithatypical10-year vehiclelife.
Acknowledgements
Theresearchpresentedwithinthispaperissupportedbythe EngineeringandPhysicalScienceResearchCouncil(EPSRC–EP/ I01585X/1)throughtheEngineeringDoctoralCentreinHighValue, Low Environmental Impact Manufacturing. The research was undertaken in collaboration with the Warwick Manufacturing Group(WMG)CentreHighValueManufacturingCatapult(funded byInnovateUK)andJaguarLandRover.Theauthorswouldliketo expresstheirgratitudetoMillbrookProvingGroundLtd (Compo-nentTestLaboratory)fortheirsupportandadvicethroughoutthe testprogram.
References
[1]A.Dasgupta, C.Choi,E.Habtour, Simulationandtestvibration-nonlinear
dynamiceffectsinvibrationdurabilityofelectronicsystems,8thInternational
ConferenceonIntegratedPowerSystems,UniversityofMaryland,College
Park,MD20742,USA:Nuremberg,Germany,2014.
[2]J. Hooper,J.Marco,Definingarepresentativevibrationdurabilitytestfor
electricvehicle(EV)rechargeableenergystoragesystems(RESS),Electric
VehicleShow29(EVS29),Montreal,Canada,2016p.1–6.
[3]A. Halfpenny, Methods for accelerating dynamic durability tests, 9th
InternationalConferenceintheAdvancesinStructuralDynamics,nCode,
Southamption,England,2006p.1–19.
[4]S.-I.Moon,I.-J.Cho,D.Yoon,Fatiguelifeevaluationofmechanicalcomponents
usingvibrationfatigueanalysistechnique,J.Mech.Sci.Technol.25(3)(2011)
611–637.
[5]M. Firat, U. Kocabicak, Analytical durability modeling and evaluation—
complementarytechniquesforphysicaltestingofautomotivecomponents,
Eng.Fail.Anal.11(2004)655–674.
[6]A. Halfpenny, Methods for accelerating dynamic durability tests, 9th
InternationalConferenceonRecentAdvancesinStructuralDynamics,
Southamption,2006,pp.1–19.
[7]A.Vertua,A.Halfpenny,F.Kihm,Provinggroundoptimisationbasedonfatigue
damagespectra,inFrenchSocietyforMetallurgyandMaterials–SF2M,
JournéesdePrintemps2011(2011)1–13nCode.
[8]K.Muniyasamy, R.Govindarajan, N.Jayaram,R. Kharul,Vibrationfatigue
analysisofmotorcyclefrontfender(2005-32-0030),SAEInt.1(2005)1–5.
[9]P.D.Rawlinson,in:T.M.Pack(Ed.),IntegrationSystemforaVehicleBattery,
TeslaMotorsInc.,2012,pp.1–21.
[10]S.Arora,W. Shen,A. Kapoor,Review ofmechanicaldesignandstrategic
placementtechniqueofarobustbatterypackforelectricvehicles,Renew.
Sustain.EnergyRev.60(2016)1319–1331.
[11]B. Nykvist,M.Nilsson,Rapidlyfallingcosts ofbatterypacks forelectric
vehicles,Nat.Clim.Change5(2015)329–332.
[12]L.HuatSaw,Y.Ye,A.A.O.Tay,Integrationissuesoflithium-ionbatteryinto
electricvehiclesbatterypack,J.Clean.Prod.113(2016)1032–1045.
[13]J.Hooper,J.Marco,G.Chouchelamane,C.Lyness,Vibrationdurabilitytestingof
nickelmanganesecobaltoxide(NMC)lithium-ion18650batterycells,
Energies9(52)(2016)27.
[14]UnitedNations,TransportofDangerousGoods–ManualofTestsandCriteria–
FifthEdition–Amendment1,UnitedNations,NewYork,2011,pp.62.
[15]UnitedNations,ECER100—BatteryElectricVehicleswithRegardtoSpecific
RequirementsfortheConstruction,FunctionalSafetyandHydrogen,United
Nations,2002.
[16]I.Avdeev,M.Gilaki,Structuralanalysisandexperimentalcharacterizationof
cylindricallithium-ionbatterycellssubjecttolateralimpact,J.PowerSources
271(2014)382–391.
[17]L.Greve, C.Fehrenbach, Mechanical testing andmacro-mechanical finite
elementsimulationofthedeformation,fracture,andshortcircuitinitiationof
cylindricallithium-ionbatterycells,J.PowerSources214(2012)377–385.
[18]E.Sahraei,J.Meiera,T.Wierzbicki,Characterizingandmodelingmechanical
propertiesandonsetofshortcircuitforthreetypesoflithium-ionpouchcells,
J.PowerSources247(2014)503–516.
[19]X.Feng,J.Sun,M.Ouyang,F.Wang,X.He,L.Lu,H.Peng,Characterizationof
penetrationinducedthermalrunawaypropagationprocesswithinalarge
formatlithium-ionbatterymodule,J.PowerSources275(2015)261–273.
[20]H.Y.Choi,J.S.Lee,Y.M.Kim,H.Kim,AStudyonMechanicalCharacteristicsof
Lithium-PolymerPouchCellBatteryforElectricVehicle,HongikUniversity,
Seoul,SouthKorea,2013p.1–10.
[21]M.Brand,S.Schuster,T.Bach,E.Fleder,M.Stelz,S.Glaser,J.Muller,G.Sextl,A.
Jossen,Effectsofvibrationsandshocksonlithium-ioncells,J.PowerSources
288(2015)62–69.
[22]J.T.Chapin,W.Alvin,W.Carl,StudyofAgingEffectsonSafetyof18650-Type
LiCoOxCells,UnderwritersLaboratoryInc.,Northbrook,Illinois,US,2011.
[23]A.Wu,StudyonAgingEffectsonSafetyof18650TypeLICOOXCells,Product
SafetyEngineeringSociety/UL,2017.
[24]P. Svens, Methods for Testing and Analyzing Lithium-Ion Battery Cells
IntendedforHeavy-DutyHybridElectricVehicles,inDepartmentofChemical
EngineeringandTechnology,KTHRoyalInstituteofTechnology,Stockholm,
Sweden,2014p.87.
[25]D.Pohl,Lithiumironphospahate:whatfactorsinfluencethedurabilityof
storagesystems,in:A.Levran(Ed.),SolarEnergyStorage,2014.
[26]A. Suttman, Lithium Ion Battery Aging Experiments and Algorithm
DevelopmentforLifeEstimation,TheOhioStateUniversity,Ohio,USA,
2011p.122.
[27]T.Harrison,AnIntroductiontoVibrationTesting,BruelandKjaerSoundand
VibrationMeasurement,Naerum,Denmark,201411.
[28]J.Hooper,J.Marco,G.Chouchelamane,C.Lyness,J.Taylor,Vibrationdurability
testingofnickelcobaltaluminumoxide(NCA)lithium-ion18650batterycells,
Energies9(281)(2016)1–18.
[29]W.Tustin,Sumultaneousmultipleaxisvibrationtesting,TheAMMTIACQ.3(4)
(2016)13–14.
[30]C.-J.Kim,Y.J.Kang,B.-H.Lee,Generationofdrivingprofileonamulti-axial
vibrationtableforvibrationfatiguetesting,Mech.Syst.Sig.Process.26(2012)
244–253.
[31]E.Habtour,W.Connon,M.Pohland,S.Stanton,M.Paulus,A.Dasgupta,Review
ofresponseanddamageoflinearandnonlinearsystemsundermultiaxial
vibration,ShockVib.2014(2014)1–21.
[32]M.Ernst, E. Habtour, A. Dasgupta, M. Pohland,M. Robeson, M.Paulus,
Comparisonofelectroniccomponentdurabilityunderuniaxialandmultiaxial
randomvibrations,J.Electron.Packag.137(2015)011009-1–011009-8.
[33]M.Ecker,N.Nieto,S.Käbitz,J.Schmalstieg,H.Blanke,A.Warnecke,D.Uwe
Sauer,CalendarandcyclelifestudyofLi(NiMnCo)O2-based18650lithium-ion
batteries,J.PowerSources248(2014)839–851.
[34]Q.Zhang,R.E.White,CalendarlifestudyofLi-ionpouchcells,J.PowerSources
173(2007)990–997.
[35]K.Buckley,L. Chiang,Design andAnalysisof VibrationTest Fixtures for
