Robust PCI Monitoring During PWR Operation at ... - Studsvik

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097Robust PCI Monitoring During PWR Operation at Southern NuclearAdel Alapour, Ryan M. JoyceSouthern Nuclear Operating Company42 Inverness Center ParkwayBirmingham, AL 35242 USAaalapour@southernco.comArthur S. DiGiovine, Shaun Tarves, Nestor PatinoStudsvik Scandpower, Inc.1087 Beacon St.Ste 301Newton, MA USA 02459 USAarthur.digiovine@studsvik.comAndrew Worrall, Robbie Gregg, Glyn RossiterUK National Nuclear LaboratoryA709, SpringfieldsSalwick, PrestonLancashire, UKandrew.worrall@nnl.co.ukAbstract – Emerging industry practice is transitioning from global ramp rate monitoring tolocal power monitoring for PCI concerns. In consideration of these changes, Southern Nuclearhas worked with Studsvik and the UK National Nuclear Laboratory (NNL) to evaluate morerobust PCI monitoring capabilities during plant operation. Using Studsvik’s automated reactivitymanagement system, CMSOps and NNL's fuel performance code, ENIGMA, an actual operationalevent from one of Southern’s PWR’s was evaluated using several PCI methodologies. This studywas of particular interest since there was an indication of leaking fuel within 24 hours of the eventdescribed herein. The methodologies applied in evaluating this event range from simplisticcalculations of changes in nodal linear heat generation rates, to explicit individual pin-by-pinthermo-mechanical fuel performance assessment. The merits and deficiencies of the variousmethods are presented. The clear choice is the coupled CMS to ENGIMA capability, termedONUS. In the case evaluated here for Southern Nuclear, the analyses confirmed that the fuelfailure mechanism was not due to classical PCI, thereby, allowing the utility to eliminate thismechanism from further investigation.I. INTRODUCTIONEmerging focus on improving fuel performance andefforts to diminish fuel failures has led to better fuelmechanical designs and analysis techniques employed inanalyzing fuel failure mechanisms. Additionally, recentfocus has raised the question whether these modelingapproaches can be applied not only during fuel loadingcore design calculations, but during actual plant operationas well.This paper details recent developments in techniquesfor improving analysis of PCI during plant operation.Several modeling approaches used to evaluate PCI arepresented. The methods range from inference of margin toPCI via examining changes in LHGR to detailed thermomechanicalfuel performance analysis of actual fuel cladstress and clad crack length propagation. Finally, themethods are applied to an actual operating PWRoperational transient.

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097II. PCI ANALYSIS METHODS OVERVIEWFuel performance has long been a part of reload coredesign and analysis. Typically, only bounding analysis ofthe most limiting fuel pins is included in the core designprocess.The difference in the analysis presented here is twofold.First, an actual transient from operating reactors thatresulted in changes to LHGR is analyzed for margin toPCI. Second, since PCI is a pin-based phenomenon, allfuel pins in the core are equally analyzed (e.g., nobounding analysis of only limiting pins was used).Fig. 1. Limitations to delta kW/ft modelThe assessment demonstrates that the various methodsfor examining PCI can be applied to actual operatingevents and screening criteria for bounding analysis is notnecessary.II. A. Delta-kW/ft ModelIn a delta-kW/ft model, changes in linear heatgeneration rate (LHGR) between successive calculationpoints (statepoints) are examined to ensure the values arewithin acceptable limits. The LHGR values can be nodeaverageddata, where LHGR's for each pin in an assemblyare normalized within an assembly and divided axially intonodes.The delta-kW/ft model uses the maximum change inLHGR as a potential indicator of a PCI violation. Thedisadvantage to this approach is that it may ignore arelatively small change in LHGR that could indicate a PCIviolation if the LHGR was already near the limit.Additionally, this approach could incorrectly recognize arelatively large change in LHGR as a potential PCI issue,even though the large change occurred in a pin that wasoperating well below violation limits. This is illustrated inFig. 1.To compensate for these shortcomings, conservativevalues for the limiting delta-kW/ft are typically applied.Although this does lend itself to mitigating PCI concernsduring a power maneuver, for example, it can beeconomically inefficient, requiring a slower return topower than is actually needed to preclude PCI.The advantage of the delta-kW/ft model is that data isreadily available from the core simulation neutronicmodels, with relatively fast computational times.II. B. Threshold LHGR PCI ModelA more sophisticated method of monitoring PCI usingLHGR data is to account for the physical phenomenon ofpin conditioning, the time-dependent annealing that occurswhen a pin LHGR remains at the same power for anextended period. By modeling conditioning, greaterchanges in LHGR are achievable without violating PCIlimits, since the pellet and cladding conditioning arefactored into the calculation.Fig. 2 presents a graphical depiction of how athreshold model works. The lower dashed line representsthe conditioning threshold. The upper dashed line is thePCI violation limit, which is a delta above the PCIthreshold. If, during plant operation, a pin LHGR valueexceeds the threshold, "conditioning" begins for that pin.During conditioning the threshold increases and, since thePCI limit is a fixed delta above the threshold, the PCIviolation limit also increases. Provided the LHGRincreases at a rate slow enough to remain below the PCIlimit, no violation occurs. However, if the LHGR changesmore rapidly than the threshold changes, a PCI violationmay occur.

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097fuel performance perspective are maximum localconcentrated clad hoop stress and maximum local cladcrack length.Clad Hoop Stress - is the stress that forces a crack inthe clad wall to “open up” and propagate through thecladding. Formation of a crack is a prerequisite toPCI and is therefore an ideal indication of thesusceptibility of a fuel pin to PCI failure. However,PCI failure will not necessarily occur only byformation of the crack by itself.Fig. 2. Threshold PCI Model using LHGR dataConversely, once the LHGR is above the thresholdand then begins to decrease, the threshold and PCIviolation limits begin to decrease or de-condition.The drawback with this particular model for PWRs isthat the threshold level, the PCI violation limit delta, andthe conditioning & de-conditioning rates are not typicallysupplied from the fuel vendor. They can be determinedfrom detailed thermo-mechanical fuel performancecalculations or via experimental data. Additionally,detailed neutronic data of the actual LHGR's duringoperation has previously been difficult to obtain.II. C. Detailed Thermo-Mechanical FuelPerformance AnalysisWhen the LHGR of a fuel pin increases, the fuel pinbecomes prone to a failure mechanism known as PCI. Thespecific mechanism is referred to as stress corrosioncracking (SCC) and occurs at the inner wall of the fuel pincladding. Although the power increase causes a differentialthermal expansion in the fuel clad, which in turn generatesa tensile stress in the cladding (hoop stress), a certain finalpower level is also necessary for failure to occur.This behavior implies that a high temperature in thefuel, leading to the release of chemically aggressive fissionproducts (such as iodine), is also required for PCI. Thefailure is not immediate, suggesting that the crack has todevelop and propagate until the crack length is greater thanthe cladding thickness, requiring that the fuel pin be atpower for a certain period of time.Clad Yield Stress - the point at which the materialmoves from elastic to plastic deformation, i.e., doesnot "recover" when the stress is removed. In thisanalysis the difference in the calculated clad hoopstress to clad yield stress, or margin to clad failure, isused. The clad yield stress limit is dependent on thefuel design and conditions during plant operation.Clad Crack Length – Once a crack has formed in theclad wall, the length and growth of the crack is adirect indication of the potential for a PCI failure.However, the failure threshold occurs when thecalculated crack length exceeds the clad wallthickness.It is important to note that the clad hoop stress andclad crack length are very much dependent on theirradiation history (LHGR vs. exposure) as well as thelocal LHGR of every fuel pin. Therefore, a full powerhistory (LHGR vs. time) of every fuel pin for the entiretime it has been in the reactor core must be modeled as partof the PCI assessment. (Note this is for Zircaloy clad fueland is not necessarily applicable for other claddingmaterial.)The deficiency with this approach has been in itsapplication; typically only the most limiting pins from acore neutronic simulation are selected for this morethorough fuel performance analysis. This brings intoquestion the criteria used to select the candidate pins.During an actual operational transient it is very difficult toset inclusive criteria to ensure the most limiting pins arecorrectly chosen. As with the other methods, until now,there has not been an efficient way to perform this morerobust assessment during plant operations.Therefore, in order to properly assess vulnerability toPCI due to local power changes, the key indicators from a

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097III. OPEN VALVE TEST EVENT DESCRIPTIONTo demonstrate the application of the various PCIevaluation methods to actual operating data, a specificevent from an operating PWR in Southern Nuclear's fleetwas selected. The event is called an open valve test. Thetest itself should result in fairly mild changes in LHGRvalues. However, upon completion of the test in the actualplant, coolant off-gas activity increased indicating fuelfailure. The analysis was motivated by the need toeliminate PCI as a possible cause.In an open valve test the reactor core inlet coolanttemperature is reduced to facilitate testing on thesecondary side. The temperature reduction is mild, on theorder of a few degrees.This particular event was conducted near BOC at fullreactor power (HFP). The MTC for the core at thiscondition was slightly negative in value (meaning astemperature decreases positive reactivity is inserted). Inorder for the reactor to remain critical this positivereactivity insertion must be balanced with the insertion ofnegative reactivity by partially inserting the lead controlbank. During this event, the lead bank was partiallyinserted approximately 10% into the core.Once the testing on the secondary side is complete, thelead control bank is withdrawn to a full out position as theinlet coolant temperature is slowly raised back to nominal.During this movement of the lead control bank at HFPLHGR is redistributed axially and radially within the core.Since PCI is dependent on changes in LHGR vs. time, thisevent was chosen to apply the various PCI methodspreviously discussed.IV. ANALYSIS MODELS OF SPECIFIC OPERATINGEVENTS DESCRIPTIONIn order to assess PCI during operating events, severalcalculational components are necessary: an accurate threedimensional,cycle-specific core neutronic model, detailedcore follow data for the actual transients, an accuratethermo-mechanical fuel performance code, a systemcoupling these models together, and a data managementsystem to digest and resolve all the data points into asuccinct summary for the events. This section describesthese analysis models.IV. A. CMS Model DescriptionA cycle-specific CMS neutronics model, comprised ofStudsvik’s CASMO 1 and SIMULATE 2 codes, has beenused for several cycles at Southern Nuclear for fuel vendoroversight, operational support, and operator simulatortraining. These models employ CASMO as the crosssectionlattice code and SIMULATE as the steady-statenodal simulator.IV. B. CMSOps Model DescriptionStudsvik’s CMSOps 3 was used to integrate detailedplant operational data during the events analyzed withinthis paper 4 . CMSOps manages the high-fidelity CMSneutronics model for reactivity management, including athreshold based PCI assessment.CMSOps ensures the CMS core neutronics modelaccurately reflects the plant’s actual detailed operatinghistory by automating core follow. CMSOps reads plantmeasured signals on a minute-by-minute frequency anddetects any changes in the values of these signals. Basedon certain change criteria being met (e.g., changes inpower level, lead control bank position, inlet temperaturechange, etc.), CMSOps automatically activates executionand updating of the cycle-specific SIMULATE model.CMSOps is built on a very efficient database thatarchives all of the measured signals and all calculatedneutronic parameters, making it easy to assess PCI usingother techniques, such as the delta-kW/ft and the thresholdmodels.Studsvik, with assistance from Southern's engineers,implemented CMSOps for the plant in which the operatingevent (open valve test) was conducted. This event resultedin several SIMULATE calculations over a small amount oftime (~8 hours), each of which produced detailed threedimensionalLHGR pin-maps.IV. C. ENIGMA Thermo-Mechanical FuelPerformance Model DescriptionThe UK National Nuclear Laboratory (NNL)participated in the analysis by evaluating the CMS datafrom the operational event using their fuel performancecode, ENIGMA 5 .

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097ENIGMA is a fully validated fuel performance codethat has been used for several definitive design andlicensing assessments for both UO2 and MOX fuel,including licensing UO2 and gadolinia-doped UO2 inSizewell B (the UK’s only PWR) and the Finnish LoviisaVVER-440 reactor, and mixed oxide (MOX) fuel in theSwiss Beznau-1 PWR. ENIGMA is also used in support ofMOX production in the Sellafield MOX Plant (SMP) andto perform feasibility studies for advanced fuel concepts.Recently NNL and Studsvik have coupled the CMScore neutronics model and ENIGMA’s fuel performanceanalysis capabilities, producing ONUS 6,7 to performthermo-mechanical analysis of every fuel pin in the core.Since CMSOps automatically generates CMS casesfor the entire operational event, detailed LHGR's generatedare available for every fuel pin and axial pin node in thecore. This data was coupled directly with ENIGMA toperform PCI evaluation for every pin in the core, using itsdetailed power history. ONUS is also then used to displaythe 3-D pin-by-pin data available following the analysis.V. PCI RESULTS FROM OPEN VALVE TESTThis event provided an opportunity to explore thevarious ways of monitoring PCI limits in order to mitigatePCI violations. The three various methods explainedpreviously are applied.Fig.3. presents a graphical result from ONUS(ENIGMA/CMS), of the normalized, axially integrated(radial) pin-by-pin LHGR at the point of deepest insertionof the lead control bank. The core loading pattern is easilyrecognized as a contemporary low-leakage one.Fig.3. Relative Power Density Distribution and LeadControl Bank Locations during Event (Circles)Note also the locations of the lead control bank, whichis partially inserted during the open valve test. Since thelocations are in relatively high-powered regions, theconcern of PCI during the event is warranted.V. A. Delta-kW/ft Results for Open Valve TestUsing data from the open valve test event, the deltakW/ftanalysis was applied to assess margin to PCI limit.This analysis calculated delta-kW/ft/hr, as compared to theimmediately previous case, calculated with data for everynode in the core, as shown in Fig. 4. and Fig. 5.Fig. 4. Nodal delta-kW/ft/hr during Open Valve Test vs.Exposure..

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097Fig. 5. Nodal delta-kW/ft/hr during Open Valve Test vs.TimeThe nodal delta-kW/ft/hr values presented in thefigure coincide with the expected behavior. During theevent nodes where power is being suppressed as the leadcontrol bank is inserted experience a negative deltakW/ft/hr.Conversely, nodes where power is beingincreased as the lead control bank is withdrawn experiencea positive delta-kW/ft/hr. The magnitude of the changes (indelta-kW/ft/hr) are consistent with the expected increase inclad strain – a key component of PCI – given that the leadcontrol bank movement occurs over a relatively short time.The event was also evaluated with the delta-kW/ftapproach by using data for each fuel pin in the core (datawhich is typically not available during operationaltransients).Exploring the same plot of delta-kW/ft/hr with datafor every pin in every node in the core (Fig. 6.) highlightsone major inadequacy of this monitoring method – namelythat it ignores local (pin-level) changes, which can besignificantly larger in magnitude than more coarse nodewisedata.This pin data is compared with the nodal results andpresented as Fig.6. Note that in the following plot, the datain red is pin-by-pin delta-kw/ft/hr, while the data in blackis the nodal delta-kw/ft/hr.Fig. 6. Pin-by-Pin Margin-to-PCI during Open Valve TestThe results demonstrate that using nodal data vs.detailed pin information does not show the significantlyhigher changes in LHGR that individual pins couldexperience during a power transient. (Even though for thisparticular case there is not sufficient evidence, even withthermo-mechanical fuel performance analysis to supportPCI failure.)Finally, neither the nodal data based delta-kW/ft/hrmodel nor the pin-by-pin delta-kW/ft/hr models aresufficient for diagnosing PCI. PCI is predicated not only inchanges in kW/ft but also a time component (e.g., howquickly the changes have occurred, has the fuel claddingbeen conditioned to the higher LHGR, etc.).V. B. Threshold LHGR Results for Results for OpenValve TestUsing the same detailed pin-by-pin LHGR datagenerated from CMSOps during the open valve test, thealternative threshold model 8 was applied. As previouslymentioned, this model requires certain constants for thecomponents for monitoring (e.g., conditioning/deconditioningrates, etc.).The results presented in Fig. 7. indicates similarmargin for the most limiting pin during the transient usingthe threshold model and empirical data from StudsvikNuclear 9 .In the figure the inlet temperature vs. time during theevent is plotted against the right-side y-axis and the leadcontrol bank position on the left-side y-axis.

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097Fig. 7. Margin to PCI During Open Valve Test fromThreshold ModelAlthough every pin was evaluated for susceptibility toPCI, a simple way to summarize the event is shown in thefigure as PCIUTL (shown in blue). PCIUTL is defined asthe margin to PCI for the most limiting pin per time stepduring the event. The model accommodates the situationwhere the limiting pin location is changing from time stepto time step. Basically, the margin is calculated for everypin and only the most limiting margin value is presented inthe figure. The results show that there was sufficientmargin during the event to PCI failure.V. C. Thermo-Mechanical Fuel Performance AnalysisResultsThe last PCI analysis conducted for the open valve testwas to analyze every pin via the ONUS integrated fuelperformance analysis (ENIGMA and CMS using detailedoperational histories from CMSOps CMSOps) 10 .The most limiting parameter analyzed was the marginto maximum clad yield stress during the open valve test.The result is summarized here as Fig. 8. The ENGIMAanalysis assumed a failure yield stress of 93.5 kPSI. (Theactual limit is fuel design type- and exposure-dependent,but the variation is insignificant with the amount of marginthat exists for the fuel during this test.)Fig. 8. Margin to Yield Stress Limit of Most Limiting PinDuring Open Valve TestThe cases shown in the figure's x-axis correspond tothe time during which the open valve test was conducted,(approximately 8 hours). In the figure, the most limitingpin – in terms of smallest margin to yield stress (cladfailure) – was in one of the assemblies which the leadcontrol bank was partially inserted/withdrawn. Theminimum margin occurred axially near the core mid-plane.The figure also shows the lead control bank positionduring the test and the LHGR of the limiting pin. Theminimum margin (highest stress) occurs at the deepestinsertion of the lead control bank, most probably due to theLHGR being redistributed as the lead control banksuppresses power in the upper axial region of theassemblies into which they are inserted.Other findings from the ONUS results are that the pinswith the highest crack lengths (albeit with crack lengthsonly a fraction of 1% of the total clad thickness) are fromhigher depleted assemblies (pin-average exposures ofabout 45 GWd/T) where pellet clad gap closure will havealready occurred. However, because the higher depletedassemblies were located on the periphery of the core, thepin LHGRs before, during, and after the event wererelatively low and consequently SCC propagationnegligible.Detailed pin-by-pin results for minimum margin yieldstress during the event are presented in Fig. 9. The axiallocation where the minimum margin occurs is in thebottom half of the core. When the lead control bank ispartially inserted, the power in this lower region of thecore increases leading to a systematic increase in stress onthe clad (margin decrease). When this lead control bank isfully withdrawn, the power in this lower region decreasesagain and the stress on the clad relaxes (margin increase).

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097Pins with the lowest margin during the event are fromthe same assembly locations in which the lead control bankwas partially inserted during the event. These assemblieshave pin exposures high enough that the fuel/clad gap isexpected to have closed prior to the event.Furthermore, using these thorough and rigorousmethods, it has provided confidence that the fuel failureexperienced during the open valve test in the SouthernNuclears PWR was not caused by PCI. This analysishelped eliminated this potential cause from any furtherinvestigation and has allowed Southern Nuclear to focusresources onto other potential sources of fuel failure. (Rootcause analysis results based on poolside inspectionsconducted by Southern at the end of cycle turned out to beconsistent with results from these studies.)The techniques presented in this study were applied toanalyze a past operating event, but they can also easily beapplied to the simulation of a future power maneuvers toassess margin to fuel performance parameters (such as cladstress, strain) before the event actually occurs. The resultscould then be used to alter the power maneuver, ifnecessary, to ensure sufficient margin and help reducepotential fuel failures.Fig. 9. Minimum margin to clad yield stress (lead controlbank partially inserted)The results indicate there is still a large degree ofmargin before fuel failure would occur, even for the mostlimiting pin locations during this event. (Fuel failure mightstill be possible due to prior mechanical damage that couldhave been exacerbated during this event; missing pelletsurface, introduction of foreign material hitting the fuelpin, hydriding, etc.)VI. CONCLUSIONThree different methodologies to evaluate fuelperformance in terms of PCI were applied to an actualtransient from operating PWR. The methods range fromexamining changes in coarse nodal LHGR's (delta-kW/ftmodel) during the transient as an indicator of PCIsusceptibility, to a threshold-based pin-by-pin model, tothermo-mechanical fuel performance analysis for every pinin the entire core.The work here demonstrates that all three approachesare computationally feasible and that limitations ofperforming coarse analysis using nodal LHGR data (inwhich the margin to PCI is underestimated) can bereplaced with explicit pin-by-pin thermo-mechanical fuelperformance evaluation during plant operation.The coupled CMS-ENIGMA analysis hasdemonstrated how an explicit pin-by-pin PCI analysis canbe applied during actual operating events for fuelperformance assessment.NOMENCLATUREPCI – Pellet Clad InteractionLHGR – Linear Heat Generation RatePWR – Pressurized Water ReactorDelta-kW/ft – delta kilowatt per foot (energy)SCC - Stress Corrosion CrackingBOC – Beginning of CycleHFP – Hot Full PowerMTC – Moderator Temperature CoefficientCMS – Studsvik Core Management SystemkPSI – kilo-pounds per square inch (pressure)MPa – mega Pascals (pressure)GWd/T – gigawatt days per metric ton (exposure)REFERENCES1. J. D. RHODES, K. S. SMITH, J. D. LEE,“CASMO-5 Development and Applications,”PHYSOR 2006, Vancouver, Canada, (Sept 2006).2. K.S. SMITH et al., “SIMULATE-3 Methodology,”Studsvik Report, SOA-95/18 (1995).3. A. DIGIOVINE and A. NOËL, “GARDEL-PWR:Studsvik’s Online Monitoring and ReactivityManagement System,” Proceedings of Advances

Proceedings of 2010 LWR Fuel Performance/TopFuel/WRFPMOrlando, Florida, USA, September 26-29, 2010Paper 097In Nuclear Fuel Management III, Hilton HeadIsland, SC (October 2003).4. S. TARVES, A. DIGIOVINE, "CMSOpsReactivity Management and PCI Analysis forSelected Events During PWR Operations forSouthern Nuclear," Studsvik Report SSP-08/453C, December 2008.5. W. J. KILGOUR, J. A. TURNBULL, R. J.WHITE, A. J. BULL, P. A. JACKSON and I. D.PALMER, “Capabilities and Validation of theENIGMA Fuel Performance Code,” ANSinternational topical meeting on LWR fuelperformance, Avignon (April 1991).6. R. W. H. GREGG, “Whole Core FuelPerformance Calculations,” IAEA TCM on waterreactor fuel performance modeling, Kendal(September 2005).7. A. WORRAL, A. S. DIGIOVINE, "ONUS: On-Line Fuel Performance Surveillance – LinkingStudsvik’s CMS With UK NNL’s ENIGMA-B,"Proceedings of Global 2009, Paris, France,September 6-11, 2009.8. Studsvik Report SSP-02/438, Pellet CladInteraction (PCI) Model in SIMULATE-3, June2004.9. AEN/NEA presentation, Current PCI criteria,relation to SCIP scope, C. Vitanza, SCIP April2006. (Studsvik Clad Interaction Program –SCIP).10. R. W. H GREGG, "Whole Core Fuel PerformanceAssessment of PWR Open Valve Test Event,"National Nuclear Laboratory memo to StudsvikScandpower, August 3, 2009.