ML111890413

Licensee Slide Regarding Questions on LOCA Frequency Analysis and Responses to Comments

ML111890413
Person / Time
Site:	South Texas
Issue date:	07/06/2011
From:	Wright S South Texas
To:	Balwant Singal Plant Licensing Branch IV
Singal, B K, NRR/DORL, 301-415-301
Shared Package
ML111890371	List: ML111890376 ML111890380 ML111890391 ML111890408 ML111890413 ML111890420
References
TAC ME5358, GSI-191, TAC ME5359
Download: ML111890413 (8)

Text

QuestionsonLOCAFrequencyAnalysis General:

1. Howarefailuresthatdontoccuratweldsconsidered?Theseinclude,forexample,ControlRod DriveMechanism(CRDM)failures,pressurizerheatersleevefailures,SteamGenerator(SG)tube ruptures,bottommountedinstrumentation(BMI)nozzlefailures,thermalfatiguefailuresat nozzles,componenttees,andothermixinglocations?

Response:Weplantoaddressnonpiperelatedfailuresandnonweldrelatedfailuresin2012sothe focusin2011istoconsiderpipebreaks.Basedonourunderstandingoftheprevailingdamage mechanismswebelievethatLOCAfrequencieswillbedominatedbyfailuresatweldlocations includingpipetosafeendandnozzlelocations.Howeverconsideringthefactthattherearemany weldsdistributedratheruniformlyoverthepiperunstheweldlocationswillprovideanopportunity toevaluatemanydetailedpipebreaklocations.Itisnotclearthatmovingtheassumedbreak locationsbetweenweldlocationswillhaveasignificantimpactbutifitdoesthenwemayneedto considersuchbreaklocationsintheLOCAfrequencyanalysis.Thisisanissuethatcanbebetter examinedonceanumberofbreaklocationshavebeenfullyanalyzed.Weacknowledgethatwewill eventuallyneedtoincludenonweldeffectsordemonstratewhyaddingthemwillnotberisk significant.AsfarasnonweldandnonpipelocationssuchasCRDMfailuresourapproachto addressingthemhasnotyetbeendefined.SGTRaremodeledinthePRAbutitisnotclearhow tubefailureswouldgenerateanydebris.Oursubmittalwillincludeajustificationofwhichlocations wereconsideredandthosethatwerescreenedoutandwhy.

2. HowareLosscoolantAccidents(LOCAs)causedbyoverpressurizationeventssuchaswater hammerorotherPToverpressurizationevents(whichmaycausevesselfailuredueto embrittlement)considered?

Response:ThepotentialforwaterhammerintheClass1pipingwasassessedintheRIISIevaluation anddeterminednotbeacrediblefailuremechanismforLOCAsensitivepiping.ThePRAmodel explicitlyconsidersoverpressurizationduringlossofmainFW+ATWSconditionsandcalculatesa probabilityofvesselfailure.WhilesuchfailuresareinfactmodeledinthePRA,theywillnot contributetotheCHANGEinCDForLERFfromdesignoroperationalchangestoaddressGSI191-coredamagewillbeassumedforthecurrentdesignaswellasthereviseddesignifthevesselfails.

3. HowareLOCAscausedbypressurizedthermalshock(PTS)consideredintheanalysis?

Response:ThePRAmodelincludesamodelforPTSinducedvesselfailurefromthermalshockbut again,nocreditistakentojustifycoredamagepreventionfollowingvesselfailure.Our understandingoftheNRCresearchonPTSindicatesthatPTSinducedvesselfailureisnotrisk significantbutevenifitwaswhenthevesselfails,theissueofdebrisformationismootasfarasthe PRAmodelisconcerned.

NOTE:ThemodelconsidersvesselintegrityfollowingoneclassofPTSevents,namelyexcesssteam flowinresponsetoatransient.Forexample,themodelquerieswhetherthevesselintegrityis maintainedfollowingaplanttripandfailurebothoftheturbinetotripandtheMSIVstoclose.PTS orthermaltransientimpactonthereactorvesselisnotincludedinthemediumorlargeLOCAevent models.NUREG1806wouldsuggestthatafutureupdateofthemodelsshouldincludethermal shockchallengesinresponsetomediumandlargeLOCAsorequivalent.

PleasenotethatvesselintegrityisalsoqueriedafteranATWSwithfailureoftheMSIVstoisolate,in whichcasethefailuremechanismofconcernisoverpressurization.Thislogicisincludedinthe generaltransientandsmallLOCAmodelsonly.

4. HowarenonpassivesystemLOCAfrequencies(e.g.,InterfacingSystemLOCAs(ISLOCAs),seal LOCAs,activesystemLOCAs)considered?

Response:InterfacingLOCAsareexplicitlymodeledinthePRA.Whentheyoccurthepressure boundaryfailuresareoutsidethecontainmentandnotrelevanttocausingdebrisinducedcore damageinsidethecontainment.ForISLOCAsthecoredamageisduetothediversionofcoolant outsidethecontainment,bypassingthesumps,andinabilitytoestablishrecirculationflow.Seal LOCAsduetoactivefailuresareincludedinthePRAmodelandtheGSI191evaluationwillevaluate thedegreetowhichsuchfailurescangeneratedebris.Iftheyareshowntohavethecapabilityto generatedebris,theywillbeexplicitlyincludedintheanalysis.

5. Whenandhowisprobabilisticfracturemechanism(PFM)beingusedtohelpdetermineLOCA frequencyestimates(slide6)?

Response:WehavenoplanstoperformanynewPFManalysisin2011butareleavingopento performsuchcalculationsifneededbasedontheresultsthisyear.Thefocusin2011istoevaluatea widespectrumofcasesthatspantheentireClass1pressureboundary.PFManalysesaremore appropriateformorefocusedevaluationsatspecificandlimitednumberoflocations.Astheproject proceedsandwhenandifitisdeterminedthatsomespecificlocationsareespeciallyimportantand wouldbenefitfromaPFMwewillincludethatbutatthisstageitistooearlytotellwhetherthiswill benecessary.

Uncertainties:

1. Howarealeatoryandepistemicuncertaintiesconsideredandseparated(slide5)?

Response:ThealeatoryuncertaintiesarereflectedintheassumptionthatLOCAsoccurasaPoission processsothefrequencyofaLOCAisametricforthealeatoryuncertaintyaboutwhetheraLOCA willoccurornot.TheuncertaintydistributionswedeveloparoundtheLOCAfrequenciesrepresent primarilyepistemicuncertainties.Howeverwerecognizethatseparatingsourcesofuncertaintyinto thesebinsissubjecttoitsownformofuncertaintyandcanbedebated.OurLOCAfrequencymodel ofuncertaintyassumesthattheLOCAfrequencyisametricofaleatoryuncertaintyandthe

uncertaintydistributionwedevelopthatisaroundthatistheepistemictype.Werealizethisisjust amodel.

2. luncertaintiesdiscussedinslidesarecharacterizedasepistemicuncertainties.Whatisthebasis forthischaracterization?

Response:Seeaboveresponsetoitem1.

3. HowisPIPExpdatabaseusedtoresolveuncertainties(slide6)?Itsunclearfromtheslides.

Response:Wemeantosaythatinformationcontainedinthedatabaseandinsightsdevelopedover manyyearsincollectingandanalyzingthedatahelpstoreducethelevelofuncertaintythatwe expertsinputsinNUREG1829.Inadditionanearlierversionofthesamedatabaseprovideduseful inputtothelastNRCsponsoredprojectonLOCAfrequencies,NUREG/CR5750.Beforesuchdata wascollected,forexamplebackinWash1400,verylittledataonnuclearpipingsystemswaseither generatedoranalyzed.TheentireefforttoimplementRIISIprogramswassupportedbyinsights fromtheservicedata.Anearlyexampleofthekindofapplicationthatthisdatabasehasinreducing uncertaintiesistoguidethejudgmentsonhowtodefinehomogeneousweldpopulationsinto groupsforfailurerateestimation.Knowledgeofthedamagemechanismresponsibleforthe experiencedpipefailuresisessentialforguidingthesejudgments(e.g.needtoseparatebimetallic weldssubjecttoPWSCCfromotherweldsnotsubjecttothisdamagemechanism)

4. HowareNUREGs1829andCR5750beingusedtoquantifyepistemicuncertainties(slide6)?

WearepreparingaslidepresentationtouseduringourJuly7conferencecalltowalkthrough examplesofhowweplantousethesereferencestoincorporateepistemicuncertaintiesintothe STPspecificLOCAfrequencies.Tobrieflysummarizeweplantoincorporateinformationfrom NUREG1829toestablishtheuncertaintydistributionparametersofourmodelfortheconditional probabilityofLOCAvs.LOCAcategory.

5. WhydoesNihaveuncertainty?Isntthenumberofweldsknown?

Therearetworeasonsforthisuncertainty.Oneisplanttoplantvariability.Eachplanthasa differentnumberofweldsforagivencomponent,therearedifferentnumbersofcoolantloopsin thePWRpopulation(2,3,and4),differentnumberofinterfacingsystemconnectionslikeECCS,etc.

etc.Thesecondreasonisthateventhoughthesenumbersareknownwitheachplants organization,thereareonlypublicallyavailablecountsforsomespecificplants.Basedonalimited samplewhichwillbedocumentedinoursubmittal,planttoplantvariabilityisresponsiblefora factorof2aboveandbelowthebestestimateformanypipeweldcategories.

6. Whyistherelittleuncertaintyassociatedwiththenumberoffailures(nikinslide9)?Doesnt littleuncertaintyinthisparameterassumethatdatabasehascompletecoverageofallevents andthatnootherfailureshaveoccurred?

Basedonourexperience,thereisverylittleuncertaintyinpipefailurecountsfortheClass1 pressureboundarybasedonthePIPExpdataandcertainlymuchlessthantheuncertaintyweare assigningtothecomponentexposure.AlsothewholeideaofusingaBayesmethodforestimating failureratesisbasedontheideaofstartingwithapriordistributionthatmodelsaverylarge uncertainty.Inourapproachthepriorsareassumedtobelognormalwithrangefactorsof100.

CalculationofLOCAestimates:

1. Howistheintegritymanagementfactor(Iik)calculated(slide8)?IstheMarkovmodelusedto determineIik?

Response:Yes,theMarkovmodelisusedtocalculatethisfactor.Thiscalculationapproachwas workedoutfortheEPRIRIISIandisextensivelydocumentedintheattachedreferences.Thefirst reportontheMarkovmodelisReference[1]andtheinitialpipefailuredatadevelopedforuseof thismodelinReference[2].EPRIsponsoredreviewsofthisworkaredocumentedinReference[3]

whichisincludedasanappendixtoReference[1].Thismodelanddataweredevelopedinitiallyin ordertosupportestimatesinthechangeinCDFandLERFduetochangesinweldselectionsforNDE aspartoftheEPRIRIISIprogram.TheuseoftheMarkovmodeltocalculateinspectionfactorswas firstdocumentedintheEPRIRIISITopicalReportinReference[4].TheNRCsafetyevaluationofthe EPRITopicalReportincludesfindingsthatapprovetheuseofthemodelandthesupportingdatafor theEPRIRIISIevaluations.ThisreviewwassupportedbyanNRCsponsoredreviewoftheMarkov modelandtheBayesfailureratemethodverysimilartowhatweplantouseinthisprojectdoneby LANLinReference[5].Apeerreviewedjournalarticlewithmanyofthemathematicaldetailsofthis methodarefoundinReference[6].

TosummarizetheMarkovmodelisusedtoformulateordinarydifferentialequationswhichare solvedanalyticallyforthetimedependentstateprobabilities.Theinputparametersformodel whicharethecoefficientsoftheODEaredefinedintermsofpipefailuremechanismsthatproduce flaws,leaksandruptures,aswellasparametersforthefrequencyandreliabilityofprogramsto detectleaksandinspectforflaws.Thenfromthesesolutions,analyticalexpressionsareobtained forthehazardrate,whichiskindofatimedependentrateofrupture.Duetotheboundary conditionsoftheequations,thehazardrateincreaseswithplantage(asseeninAppendixDof NUREG1829).Theinspectionfactoristhehazardrateat40years(or60yearsdependingonthe application)duetosomespecificintegritymanagementprogram(combinationofleakinspection andNDE)dividedbythehazardrateatthesametimefortheaveragecomponentwithaverage integritymanagement.

2. Therelationshipbetweentheflowchart(slide12)andequations(1)-(3)(slides8and9)is unclear.Pleaseidentifywhichspecifictermsintheequationsarecalculatedbyspecificstepsin theflowchart.

WeplantowalkthroughanexampleinourJuly7presentationwhichwillclarifyeachstepin quantifyingtheLOCAfrequencies.

3. Therearemanyquestionsrelatedtotheflowchart(slide12).

a. Whydoesthenumberofleaksprovideinputtoboththefailurefrequencyand conditionalruptureprobability?

Thenumberofleakscontributestothenumeratorofthefailurerateestimate.Thenumber ofleaksalsocontributestothedenominatoroftheconditionalprobabilityofrupture estimate.

b. Whereisdegradationmechanism(DM)susceptibilityinEquations(1)-(3)?Howdoesit factorintothoseequations?

Response:Weknowfromtheserviceexperiencethatsomefailuresoccurduetosome specificdamagemechanisms.Wecancalculatetheunconditionalfailureratefromany damagemechanismsimplydividingthenumberoffailuresbythetotalcomponentyearsin theservicedata.ButafterwehaveacompletedRIISIprogramsuchasthecasewithSTP weknowonaweldbyweldbasiswhichweldsaresusceptibletoeachdamagemechanism.

Nowweneedtocalculatetheconditionalfailuregivenweknowtheapplicabledamage mechanism.Forthisweneedanestimateofthefractionofweldsinthedatabasethat producedsomayfailuresduetoeachDMhowmanyaresusceptibletoeachDM.Hencethe fractionfinthedenominatorofEquationmustbeappliedtoestimatetheconditionalfailure rate.Leavingitoutwouldyieldtheunconditionalfailurerate.

WeshallalsoaddressthisquestionintheJuly7meeting.

c. Whatexpertsarebeingusedtoprovidevariousestimates?Aresameexpertsusedto provideeachdistributionindicatedinfigure?

Response:InthecaseofinformationwebringinfromNUREG1829,itistheexpertpanel fromthatproject.Wemayalsoneedtoincorporateexpertjudgmentsfromourteamthat willbeclearlydocumentedinthesubmittal.

d. WhydotheDMsusceptibilityestimatesprovideinputtoboththegenericprior distributionandintheBayesupdatedistributions?

Response:Theestimatesofthefractionofweldsinthegenericpopulationthatare susceptibletothedamagemechanismareusedtodeterminetheparametersofthe likelihoodfunctionsfortheBayesupdateofpriordistributionswhichareintendedto representthefailureratesforcomponentssusceptibletothosemechanisms.Theyneedto beconsistent.OurJuly7presentationwillaimtoclarifythis.

e. How,specifically,istheBayesianupdateofthepriordistributionperformedusingthe threedistributionsgeneratedtoinformtheprior?

Response:Thepriorareassumedtobelognormal.WeuseaPoissonlikelihoodfunctionto update these with one set of data for each hypothesis of weld population and weld

susceptibilityfraction.Thisyieldsseveraldifferentposteriorsthatarecombinedusingwhat isreferredtoasBayesposteriorweighting.WewillexplainthismoreclearlyduringJuly7 presentation.

f.

HowistheBeliczySchultzcorrelationusedtocreatetheP(RlF)priordistribution?

Response:Asyouknow,thiscorrelationwasusedbyBengtLydellasinputtothis distributionforthebasecaseanalysesinAppendixDofNUREG1829.IntheSTPweplanto basethepriorsforthisdistributionusinginformationfromNUREG1829whichwillbe explainedduringtheJuly7meeting.

g. HowistheP(RlF)priorupdatedusingBayes?

Response:WeperformaBayesupdateforeachdiscreteLOCAcategory,whichis associatedtoabreaksize.Weuseatruncatedlognormaldistributiontorepresentthe uncertaintyintheconditionalprobabilityofLOCAateachCategoryseparately.Weupdate itwithevidenceofnormally0LOCAsandNfailureswhereNisthenumberoffailuresused tocalculatethecorrespondingfailurerates.WewillshowthisintheJuly7meeting.

1. ThecalculationprocedureandtheapplicationoftheMarkovmodelinboththeflowchartand equationsisunclear.Also,information/documentationontheusedofthismodelforotherNRC approvedapplicationsandothernuclearapplicationsshouldbeprovided.Aretheredifferences betweenthemodelbeingusedfortheseestimatesandwhathasbeenapprovedandusedin otherapplications?

Response:Pleaseseetheabovequestionresponseontheinspectionfactorandprovided references.Anydeviationsfrompreviouslyreviewedapplicationswillbefullydocumentedinthe submittalandsupportingreports.

2. HowistheMarkovmodeldifferentthanP(RlF)?IsthismodelusedtodetermineP(RlF)?

Response:P(RF)isusedtocalculaterupturefrequencies.Failureratesforflaws,leaks,and rupturesareinputtotheMarkovmodeltodeveloptheintegritymanagementfactors.

3. IntheMarkovmodelthereisnoprobabilityofrupturegivennodetectibledamage.Whyisthis termneglected?DoesntthispresumethatISIisperfect?

Response:Thismodelisonlyusedandwillonlybeappliedforrupturesduetodegradation mechanisms.ThereisanotherversionofthemodeldevelopedinReference[1]thatincludes additionaltransitionsforleaksandrupturesabsentadetectableflaw.Thereasonfornotincluding thosetransitionsisthatpipefailuresduetosevereloadingconditionsaremodeledexplicitlybynon LOCAtypeinitiatorsinthePRAmodel.

4. Thereareanumberofquestionsrelatedtoslide33

a. Whatisthebasisofthehazardrate{h(t)}equation?

Response:Thisisstandardreliabilityengineeringtheory.Thehazardrateisdefinedasthe negativeoftherateofchangeofthereliabilityfunction(probabilityofnorupture)divided bythereliabilityfunction.InthisMarkovmodel,thereliabilityisthesumofthestate probabilitiesforsuccess,flaw,andleak.SeeReference[7]forthemathematicaldetails.

b. Whatisr(t)?

Response:Thisisthereliabilityfunctionreferredtoinitema.

c. Howish(t)normalized?

Response:Seetheabovequestion/responseontheIntegrityManagementfactor

d. Whatsensitivityanalysesonh(t)areperformed??

Response:Thisreferstochangingtheassumptionsabout:whetherornotthereisaleak detectionprogramandifthereis,howoftenitisdoneandwhatistheeffectiveprobability ofdetection;andwhetherornotthereisanNDEprogram,andifthereishowoftenitis performedandwhatistheeffectiveprobabilityofdetection.Foreachcombinationofleak detectionandNDEinspectionparameters,adifferentresultisobtained.

Questionsonspecificslides:

1. Slide17:WhatisWH(waterhammer)?Yes

2. Slide19:

a. Howareunconditionalfailureratesdetermined?Byleavingfoutofthedenominatorin Equation(3)

b. Aretheconditionalestimateddeterminedthroughexpertelicitationtodetermine bumpupfactorsfortheunconditionalestimates(i.e.,bumpupofapp.5forthermal fatigue)?Ifnot,howweretheconditionalestimatesdeterminedandwhatdothey mean?NotheresultsoftheRIISIDMevaluationareusedtoresolvedeterministically whichweldsaresubjecttoeachDM

3. Slide22:WhatareexcessiveLOCAs?VesselfailuresandmultiplepipebreaksandanyLOCA thatisbeyondthecapabilitiesoftheECCSaccordingtothePRAsuccesscriteria.

4. Slide38:WhydoesinspectionstillyyieldapositiveCDFforreactorcoolantpump(RCP)?

Whileitsalowernumberthanifnoinspectionisperformedwhydoesitstilladdrisktothe plant?RCPstandsforreactorcoolantsystempiping.Thereisanincreasebecauseofthefact thatmanyweldswereremovedfromtheRIISIprogram.ThisistypicalinallRIISIprograms.

References

[1] PipingSystemReliabilityandFailureRateEstimationModelsforUseinRiskInformedInService InspectionApplications.EPRI,PaloAlto,CA:1998.TR110161.

[2] PipingSystemFailureRatesandRuptureFrequenciesforUseinRiskInformedInService InspectionApplications.EPRI,PaloAlto,CA:1999.TR111880.

[3] Mosleh,A.andF.Groen,TechnicalReviewoftheMethodologyofEPRITR110161,University ofMarylandreportforEPRI,publishedasanAppendixtoEPRITR110161(Reference[6])

[4] RevisedRiskInformedInServiceInspectionProcedure.EPRI,PaloAlto,CA:1999.TR112657, Rev.BA.

[5] U.S.NuclearRegulatoryCommission,SafetyEvaluationReportRelatedtoRevisedRiskInformed InServiceInspectionEvaluationProcedure:EPRITR112657,Rev.B,July1999,Washington, D.C.,1999.(publishedasaforwardtoTR112657(Reference[9])

[6] Martz,H.,TSA1/99164:Final(Revised)ReviewoftheEPRIProposedMarkov Modeling/BayesianUpdatingMethodologyforUseinRiskInformedInServiceInspectionof PipinginCommercialNuclearPowerPlants,LosAlamosNationalLaboratory,June1999.

[7] Fleming,K.N.,MarkovModelsforEvaluatingRiskInformedInServiceInspectionStrategiesfor NuclearPowerPlantPipingSystems,ReliabilityEngineeringandSystemSafety,Vol.83,No.1, pp.27-45,2004.