ML111890408

Licensee Handout Loca Frequencie Approach and Example Application

ML111890408
Person / Time
Site:	South Texas
Issue date:	07/06/2011
From:	Kreslyon Fleming, Lydell B KNF Consulting Services, Risk Management, Scandpower
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: ML111890408 (34)

Text

Development of LOCA Initiating Event Frequencies for South Texas Project GSI-191 (Draft Report)

Developed for South Texas Project Electric Generating Station by Karl N. Fleming KNF Consulting Services LLC and Bengt O. Y. Lydell July 2011

LOCA Frequencies for STP GSI191 Table of Contents

1. Summary of LOCA Quantification Procedure ....................................................................................... 4

2. Example Application of LOCA Frequency Methodology ....................................................................... 6 2.1. Failure Data Query ........................................................................................................................ 6 2.2. Component Population Exposure ................................................................................................ 7 2.3. Damage Mechanism Susceptibility ............................................................................................. 10 2.4. Failure Rate Bayes Updates ....................................................................................................... 10 2.5. Failure Rate Synthesis ................................................................................................................. 14 2.6. Conditional Rupture Mode Probability Model............................................................................ 14 2.6.1. Use of NUREG1829 Data .................................................................................................... 14 2.6.2. Model for Deriving Conditional Probabilities from Rupture Frequencies .......................... 17 2.6.3. Incorporation of Epistemic Uncertainties from NUREG1829 ............................................ 20 2.6.4. Bayes Update of the Conditional Probability Distributions ............................................... 28 2.7. LOCA Frequencies Associated with Surge Line Welds ................................................................ 29 2.7.1. Base Case Results ................................................................................................................ 29 2.7.2. Influence of NDE Inspections on Location Specific LOCA Frequencies ............................... 30 2.8. LOCA Frequency Summary.......................................................................................................... 32

3. References .......................................................................................................................................... 33 2 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figures Figure 21 Event Tree Model to Represent Uncertainty in Surge Line BJ Weld Exposure for Thermal Fatigue ........................................................................................................................................................ 12 Figure 22 Category 1 LOCA Frequencies for PWR Piping Systems at 25 Years of Plant Operation (Reproduced from Figure L.13 in NUREG1829) ......................................................................................... 16 Figure 23 Benchmarking Lognormal Distributions to Lydell Base Case Results - HPI Injection Line ........ 18 Figure 24 Benchmarking Lognormal Distributions to Lydell Base Case Results - RCS Surge Line............. 19 Figure 25 Benchmarking Lognormal Distributions to Lydell Base Case Results - RCS Hot Leg ................. 19 Figure 26 Definition of Target Bounds for HPI Injection Line LOCA Frequencies ...................................... 21 Figure 27 Calibration of Conditional Probability Model to Match Targets for HPI Injection Line ............. 22 Figure 28 Definition of Target Bounds for RCS Surge Line LOCA Frequencies .......................................... 23 Figure 29 Calibration of Conditional Probability Model to Match Targets for RCS Surge Line ................. 24 Figure 210 Definition of Target Bounds for RCS Hot Leg LOCA Frequencies ............................................. 25 Figure 211 Calibration of Conditional Probability Model to Match Targets for RCS Hot Leg.................... 25 Figure 212 Comparison of Conditional Probability Models for HPI Injection line ..................................... 27 Figure 213 Comparison of Conditional Probability Models for RCS Surge Line ........................................ 27 Figure 214 Comparison of Conditional Probability Models for RCS Hot Leg ............................................. 28 Figure 215 Comparison of Surge Line LOCA Frequencies with Different Damage Mechanism Inputs ..... 30 Figure 216 Comparison of Weld Failure Rates Determined by Markov Model for Different Reliability Integrity Management Approaches ............................................................................................................ 32 Tables Table 21 Results of Class 1 Failure Data Query by System and Component ............................................... 7 Table 22 Results of Class 1 Failure Data Query by Failure Mechanism ....................................................... 8 Table 23 Definition of Failure Mechanisms ................................................................................................. 9 Table 24 Service Experience by Westinghouse Reactor Type ..................................................................... 9 Table 25 Plant Data Used to Estimate Surge Line Weld Population.......................................................... 10 Table 26 Susceptibility Fractions for Surge Line Welds ............................................................................. 11 Table 27 Parameters of Bayes Updates for Weld Failure Rate Cases....................................................... 13 Table 28 Total Failure Rates for Surge Line Welds .................................................................................... 14 Table 29 NUREG1829 and STP PRA LOCA Categories............................................................................... 15 Table 210 Lognormal Distributions for Conditional LOCA Category Probabilities that Match Bengt Lydells Base Case Results ........................................................................................................................... 20 Table 211 STP Conditional Probability Models Derived From Target LOCA Frequencies ......................... 26 Table 212 Results of Bayes Update of Conditional LOCA Probabilities .................................................... 29 Table 213 Unconditional LOCA Frequencies for Surge Line Welds ........................................................... 31 3 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191

1. Summary of LOCA Quantification Procedure The technical approach to estimating LOCA initiating event frequencies is framed around the model expressed by Equations (1) and (2) for estimating the frequency of a LOCA of a given size. The parameter x is treated as a discrete variable representing different break size ranges such as those used in NUREG 1829 to describe the 6 LOCA categories. Hence x takes on values: {1,2,3,4,5,6} to correspond with the LOCA categories defined in NUREG1829 [1]. For the time being, we shall use the NUREG1829 categories with the understanding that these may be redefined later if necessary.

F ( LOCAx ) mi ix (1) i jx ix ik P( Rx Fik ) I ik (2) k where:

F ( LOCAx ) Frequency of LOCA of size x, per reactor calendaryear, subject to epistemic uncertainty calculated via Monte Carlo mi Number of pipe welds of type i; each type determined by pipe size, weld type, applicable damage mechanisms, and inspection status (leak test and NDE); no uncertainty ix Frequency of rupture of component type i with break size x, subject to epistemic uncertainty calculated via Monte Carlo ik Failure rate per weldyear for pipe component type i due to failure mechanism k, subject to epistemic uncertainty determined by RIISI Bayes method and Eq. (3)

P( Rx Fik ) Conditional probability of rupture of size x given failure of pipe component type i due to damage mechanism k, subject to epistemic uncertainty determined via expert elicitation (NUREG1829)

I ik Integrity management factor for weld type i and failure mechanism k, subject to epistemic uncertainty determined by Monte Carlo and Markov model For a point estimate of the failure rate for type i and failure mechanism k:

nik nik ik (3) ik f ik N iTi where:

nik Number of failures in pipe component (i.e., weld) type i due to failure mechanism k; very little epistemic uncertainty 4 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 ik Component exposure population for welds of type i susceptible to failure mechanism k, subject to epistemic uncertainty determined by expert opinion f ik Estimate of the fraction of the component exposure population for weld type i that is susceptible to failure mechanism k, subject to epistemic uncertainty, estimated from results of RIISI for population of plants and expert opinion Ni Estimate of the average number of pipe welds of type i per reactor in the applicable reactor years exposure for the data collection, subject to epistemic uncertainty, estimated from results of RIISI for population of plants and expert opinion Ti Total number of reactoryears exposure for the data collection for component type i; little or no uncertainty For a Bayes Estimate, a prior is updated using nik and ik with a Poisson likelihood function.

The key inputs that are needed to provide the pipe failure rate information include:

Identification of which locations will be investigated for debris formation and the groupings of locations that will be performed to support the risk evaluation.

Counts of pipe failures in applicable nuclear industry piping systems, essentially all the failure data in ASME Class 1 and 2 piping systems in PWRs in U.S. service experience and applicable international plants with similar designs and integrity management programs - from PIPExp database.[2]

Pipe exposure estimates - quantity of pipe and pipe welds and the reactor years of service experience that produced the failure counts identified above. These estimates are based on information contained in the PIPExp data base as well as the information available in risk informed inservice inspection submittals to the NRC, which include an enumeration of weld counts in different categories and the results of damage mechanism evaluations.

Estimates of the fractions of piping system components in the service data that are susceptible to different damage mechanisms. These estimates are based on NUREG1829 and supporting computer files that provide information on epistemic uncertainty about pipe rupture frequencies vs. break size for different pressure boundary components STP RIISI evaluation report and supporting calculations providing information on applicable damage mechanisms for each weld and a definition of which welds are selected for NDE.

Results of inspection reports and other evidence of any pipe failure or degradation at STP that may influence the plantspecific failure rates, as well as the information needed to estimate exposure data.

The integrity management factor Iik of Equation (2) is quantified using the Markov model for Piping Reliability that was developed to support the EPRI RIISI projects.

The methodology outlined above and the methods and databases that have been developed to implement this approach were originally developed to support the EPRI RIISI methodology that has 5 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 been implemented for many of the existing NRClicensed plants and several foreign plants. The part of this methodology that is relevant to estimating LOCA frequencies is described in detail in Reference [3]

and has been recently applied in EPRIsponsored projects to develop piping system failure rates for use in internal flooding and high energy line break PRAs, as documented in References [4] and [5]. The original EPRI study that was responsible for developing the Markov model and Bayes method for estimating pipe failure rates and rupture frequencies was documented in EPRI TR110161 [6], and an early version of the pipe failure rate data base for both conditional and unconditional pipe failure rates was published in EPRI TR111880 [7]. An independent review of these reports was carried out by the University of Maryland, which validated the methodology that was developed in these reports. These methods and data were then used as part of the EPRI RIISI technical approach as described in the EPRI RIISI Topical Report [9]. The NRC approved these methods and data for use in applied RIISI evaluations as documented in the Safety Evaluation Report [10]. The NRC SER was supported by an independent review of the Bayes failure rate method and the Markov model by Los Alamos National Laboratory [11],

which provides a second independent review of the methodology, including a validation of the Markov model solutions.

The application of the Markov model requires the development of rather complex closedform solutions to the differential equations supporting the Markov model, which were originally developed in TR 110161 and are also published in Reference [12]. Using these closedform solutions, it is straightforward to quantify the uncertainties in the resulting inspection factors using Monte Carlo simulation methods via Microsoft Excel' and Oracle Crystal Ball', which is the approach being used in this STP GSI191 evaluation.

2. Example of Applying LOCA Frequency Methodology In this section, the LOCA frequency methodology that is being used for the STP GSI191 project is described using some examples and some of the preliminary results. The examples presented here are chosen to describe the various data analyses and modeling assumptions that will be used based on some early and preliminary results that are subject to change prior to the actual NRC submittal. The purpose is to provide the NRC with a better understanding of the approach and to help identify potential review issues.

2.1.Failure Data Query The failure data query was performed on Westinghouse and Framatome PWR plant operating experience from 1970 through 2010 and included ASME Class 1 piping systems. This generally includes reactor coolant system (RCS) piping and systems that interface with the RCS inside the isolation valves that normally separate the RCS from interfacing ASME Class 2 piping. Interfacing systems include the emergency core cooling, residual heat removal, chemical volume and control system, and various other systems including RPV head vents and instrumentation lines. The preliminary results of the data query are shown in Table 21 broken down by key components along the pressure boundary and Table 22 by failure mechanism. Because roughly half of the current fleet of operating plants were designed and built 6 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 prior to the development of ASME nuclear piping codes, much of this pipe was originally designed to B31.1 design codes, and then later inspection and ISI requirements for Class 1 piping were retrofitted into these plants. So from a design and materials perspective, the LOCA sensitive piping actually reflects a mixture of B31.1 and Class 1 pipe.

Table 21 Results of Class 1 Failure Data Query by System and Component Pipe Failures by Mode (1), (2), (3)

Nominal SYSTEM Pipe Size (NPS) Crack Crack Small Large Total Leak Full Part Leak Leak 1" 7 1 6 CVC 2" ø 4" 7 1 6 1" 2 2 Safety Injection 4" ø 10" 6 3 1 1 1 PressurizerSample 2" 5 4 1 PressurizerPORV 4" ø 10" 2 2 1" 4 1 2 1 PressurizerSPRAY 4" ø 10" 3 2 1 PressurizerSRV 4" ø 10" 7 6 1 PressurizerSurge 14" 3 3 RCS 2" 76 4 10 53 4 5 RCS Cold Leg 32" 4 4 RCS Hot Leg 32" 6 5 1 RHR 1" 6 6 RHR 4" ø 10" 1 1 RC Hot Leg S/G 19 19 32" Inlet S/GSystem 2" 8 2 2 4 TOTALS 166 12 59 83 6 6 Notes (1) Query accounts for 3914 reactor years based on date of initial criticality from 1970-2010.

(2) Failure is defined as any event that required repair or replacement of damaged component.

(3) Small leaks have leak flows << 1gpm; Leaks < 1gpm; Large Leaks < 10gpm.

2.2.Component Population Exposure Pipe component exposure is evaluated in the current analysis in terms of pipe welds in the data query.

This is estimated from a combination of the reactor years of service experience and an estimate of the total number of welds per plant. In principle, the number of welds per plant is known but is seldom in the public domain. In addition, there is usually significant planttoplant variability in the number of welds for different components, one exception being the number of coolant loops or pressurizers. To address this, the component exposure, e.g., total weldyears of experience responsible for the identified failures, is treated as an uncertain parameter in failure rate development. In addition, to support the estimation of failure 7 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 22 Results of Class 1 Failure Data Query by Failure Mechanism Event Failure Count by Failure Mechanism System NPS Type Totals C-F D&C ECSCC Fret. IGSCC LC-FAT PWSCC OVLD TF TGSCC TAE V-F 1" 1 1 Crack 2" ø 4" 1 1 CVC 1" 6 1 5 Leak 2" ø 4" 6 1 5 Crack 4" ø 10" 3 1 2 Safety Injection 1" 2 2 Leak 4" ø 10" 3 3 Prz-Sample/Instr. Crack 2" 5 1 2 2 Prz-PORV Crack 4" ø 10" 2 2 Crack 1" 1 1 Crack 4" ø 10" 2 2 Prz-SPRAY Leak 1" 3 1 1 1 Leak 4" ø 10" 1 1 Crack 4" ø 10" 6 1 5 Prz-SRV Leak 4" ø 10" 1 1 Prz-Surge Crack 14" 3 3 Crack 2" 14 1 3 3 2 1 4 RCS Leak 2" 62 1 12 2 1 1 2 2 8 33 RCS Cold Leg Crack 32" 4 3 1 Crack 32" 5 5 RCS Hot Leg Leak 32" 1 1 RHR Crack 4" ø 10" 1 1 RHR Leak 1" 6 1 1 4 S/G Inlet Crack 32" 19 1 18 Leak 1" 1 1 Crack 1" 2 1 1 S/G System Crack 1" < ø 4" 2 2 Leak 1" 3 1 1 1 TOTALS 166 1 23 7 4 3 2 48 1 9 11 1 56 8 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 23 Definition of Failure Mechanisms ID Description Comment CF CorrosionFatigue D&C Design & Construction Flaws May or may not propagate throughwall ECSCC External ChlorideInduced Primarily a concern with smallbore lines SCC IGSCC Intergranular SCC Stagnant locations with high residual stresses LCFAT LowCycle Fatigue PWSCC Primary Water SCC Nibase metal locations only OVLD Overload Mechanically induced tensile loading TF Thermal Fatigue TGSCC Transgranular SCC TAE Thermal Aging Embrittlement High temp/pressure locations only VF Vibration Fatigue rates from different damage mechanisms, it is necessary to estimate the fraction of the population that is susceptible to a given damage mechanism, which is also uncertain. Results of published reports on Risk Informed InService Inspection evaluations are useful sources to sample to provide both weld count and fraction susceptible estimates.

The reactor years of service experience by reactor type responsible for the failures in Table 21 and 22 are listed in Table 24.

Table 24 Service Experience by Westinghouse Reactor Type Reactor-Calendar Years WE Type Rx Initial Grid Initial Connection Criticality 2-Loop 570.1 581.4 3-Loop 2052.6 2096.1 4-Loop 1193.9 1236.5 Total 3816.6 3914.0 One of the key steps in the analysis is to make use of insights from the service experience to break down the Class 1 weld population into homogenous classes for the purpose of failure rate development. As an example, consider the pressurizer surge line for which three component failures are listed in Table 21.

There are three distinct classes of welds in the pressurizer surge line: a single BF weld, a bimetallic weld that connects the surge line to the pressurizer nozzle, two branch connection welds that connect the surge line to one of the hot legs, and some number of BJ welds that link surge line pipetopipe connections between the branch connection and pressurizer. Table 25 depicts information used to estimate weld counts for the pressurizer surge line. As seen in this table, there is significant variability from plant to plant in the number of BJ welds.

9 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 25 Plant Data Used to Estimate Surge Line Weld Population Weld Population(1)

PWR Branch Connection Plant BF Welds Inline BJ Welds Type to Hot Leg Braidwood1 4Loop 1 8 2 Braidwood2 4Loop 1 7 2 Byron1 4Loop 1 6 2 Byron2 4Loop 1 6 2 Kewaunee 2Loop 1 6 2 Koeberg1 3Loop 1 5 2 Koeberg2 3Loop 1 5 2 STP1 4Loop 1 8 2 STP2 4Loop 1 8 2 V.C. Summer 3Loop 1 10 2 Note (1) Kewaunee surge line is NPS10; remaining plants are NPS 14 to 16.

2.3.Damage Mechanism Susceptibility Damage mechanism susceptibility is assessed based on insights from service experience and the results of RIISI evaluations on Class 1 piping systems, which have been completed for most US PWR plants. The bimetallic BF welds are inherently susceptible to PWSCC because they are based on Nibased alloys. The branch connection welds are inherently susceptible to thermal fatigue. Some of the BJ welds are also susceptible to thermal fatigue, and all of the Class 1 welds are subject to the possibility to design and construction defects. A summary of the damage mechanism susceptibility for the surge line welds is shown in Table 26. The susceptibility and weld counts of the BJ welds are uncertain. The LOCA frequency methodology used in this study and summarized in Section 2.1 uses a technique developed in the EPRI RIISI Program [9], in which the failure rates are developed for discrete combinations of estimates of weld population and damage mechanism susceptibility, which is illustrated in Figure 21.

2.4.Failure Rate Bayes Updates The next step in the LOCA frequency quantification procedure is to perform Bayes updates for each component / damage mechanism / population exposure estimate that supports the calculation. The prior distributions used in this assessment are based on those that were developed in Reference [7] for use in the EPRI RIISI evaluations that followed the methodology in the EPRI RIISI Topical Report [9],

which was reviewed by the NRC and LANL as documented in References [10] and [11]. The evidence for the updates is based on 3 failures of BF surge line welds due to PWSCC, and 0 failures for either the branch connection or BJ welds for the surge line. The parameters of the prior and updated distributions for all the cases needed to support the surge line welds are listed in Table 27.

10 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 26 Susceptibility Fractions for Surge Line Welds Confidence Weld Susceptibility Fractions System Location Level CF D&C ECSCC Fretting IGSCC PWSCC TF TGSCC TAE VF Low N/A 1 N/A N/A N/A 1 N/A N/A N/A N/A BF(1) Medium N/A 1 N/A N/A N/A 1 N/A N/A N/A N/A High N/A 1 N/A N/A N/A 1 N/A N/A N/A N/A Pressurizer Low N/A 1 N/A N/A N/A N/A 0.01 N/A N/A N/A Surge Line BJ Medium N/A 1 N/A N/A N/A N/A 0.05 N/A N/A N/A High N/A 1 N/A N/A N/A N/A 0.25 N/A N/A N/A RCHL Low N/A 1 N/A N/A N/A N/A 1 N/A N/A N/A Branch Medium N/A 1 N/A N/A N/A N/A 1 N/A N/A N/A Connection High N/A 1 N/A N/A N/A N/A 1 N/A N/A N/A Note (1) The susceptibility of BF welds to PWSCC can be effectively mitigated by application of weld overlays.

11 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Welds/Rx 6.9 Rx-yrs 3914 Base Exposure 27006.6 Fraction of B-J Welds Exposure Weld Count Exposure Susceptible to Thermal Case Exposure Uncertainty Multiplier Fatigue Probability p=.25 0.0625 0.5 13,503 weld-yrs High (.25 x Base) p=.25 p=.50 0.125 0.1 2,701 weld-yrs High (2 X Base) Medium (.05 x Base) p=.25 0.0625 0.02 540 weld-yrs Low (.01 x Base) p=.25 0.125 0.25 6,752 weld-yrs High (.25 x Base) p=.50 p=.50 0.25 0.05 1,350 weld-yrs Medium (1.0 X Base) Medium (.05 x Base) p=.25 0.125 0.01 270 weld-yrs Low (.01 x Base) p=.25 0.0625 0.125 3,376 weld-yrs High (.25 x Base) p=.25 p=.50 0.125 0.025 675 weld-yrs Low (0.5 X Base) Medium (.05 x Base) p=.25 0.0625 0.005 135 weld-yrs Low (.01 x Base)

Figure 21 Event Tree Model to Represent Uncertainty in Surge Line BJ Weld Exposure for Thermal Fatigue 12 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 27 Parameters of Bayes Updates for Weld Failure Rate Cases Weld DM Prior Distribution(1) Evidence(2) Bayes Posterior Distribution(1)

Weld Type Count Susceptibility and DM(3) Type Median RF Failures Exposure Mean 5th 50th 95th RF Case Case Surge BF SC Base Base Lognormal 8.48E07 100 3 3914 5.62E04 1.23E04 4.83E04 1.27E03 3.2 Surge BF DC Base Base Lognormal 5.46E08 100 0 3914 1.41E06 5.41E10 5.33E08 4.77E06 93.9 Surge BC TF Base Base Lognormal 2.66E07 100 0 7828 3.25E06 2.53E09 2.34E07 1.47E05 76.1 Surge BC DC Base Base Lognormal 5.46E08 100 0 7828 1.17E06 5.37E10 5.24E08 4.37E06 90.1 Low Lognormal 2.66E07 100 0 135 9.75E06 2.66E09 2.65E07 2.58E05 98.5 Low Medium Lognormal 2.66E07 100 0 675 7.17E06 2.64E09 2.61E07 2.36E05 94.6 High Lognormal 2.66E07 100 0 3376 4.48E06 2.59E09 2.48E07 1.85E05 84.5 Low Lognormal 2.66E07 100 0 270 8.70E06 2.65E09 2.64E07 2.51E05 97.4 Surge BJ TF Medium Medium Lognormal 2.66E07 100 0 1350 5.98E06 2.62E09 2.57E07 2.18E05 91.2 High Lognormal 2.66E07 100 0 6752 3.46E06 2.54E09 2.37E07 1.54E05 77.7 Low Lognormal 2.66E07 100 0 540 7.55E06 2.64E09 2.62E07 2.41E05 95.4 High Medium Lognormal 2.66E07 100 0 2701 4.83E06 2.60E09 2.51E07 1.94E05 86.4 High Lognormal 2.66E07 100 0 13503 2.58E06 2.47E09 2.22E07 1.21E05 69.8 Low Base Lognormal 5.46E08 100 0 13503 9.83E07 5.33E10 5.14E08 3.96E06 86.2 Surge BJ DC Medium Base Lognormal 5.46E08 100 0 27007 7.66E07 5.25E10 4.94E08 3.34E06 79.8 High Base Lognormal 5.46E08 100 0 54013 5.77E07 5.12E10 4.65E08 2.67E06 72.2 Notes (1) Failure rates in units of failures per weldyear.

(2) Exposure in units of weldyears.

(3) SC = stress corrosion cracking; TF = thermal fatigue; DC = design and construction defects; BF = BF weld; BC = branch connection weld; BJ = BJ weld.

13 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 2.5.Failure Rate Synthesis The total weld failure rates are calculated using a Monte Carlo posterior weighting technique to combine the distributions from the different weldcount and damage mechanism susceptibility hypotheses and then summing the contributions from different damage mechanisms. For BJ welds, the failure rate for thermal fatigue was developed by Monte Carlo sampling from a discrete distribution defined by Figure 21 to determine which of the lognormal distributions for BJ TF from Table 27 to use for that trial. Repeating this process over many trials (100,000 trials used for the current examples) yields a single distribution for the BJ weld failure rate due to thermal fatigue which incorporates a probabilistically weighted contribution from each supporting weldcount hypothesis. For the BJ weld failure rate due to design and construction defects, only 3 cases are required to model uncertainty in the weld counts because all welds are assumed to be susceptible to D&C. Then the total failure rate for BJ welds is calculated by summing the contributions from TF and D&C. For the BF welds and branch connection welds, there is no significant uncertainty for weld counts or DM susceptibility. Therefore, its necessary only to sum the random samples from the DM distributions to compute the total component failure rates for these weld types.

Table 28 Total Failure Rates for Surge Line Welds Failure Rate per WeldYear Weld Type Mean 5%tile 50%tile 95%tile BF 5.62E04 1.38E04 4.37E04 1.40E03 Branch Connection 4.54E06 9.64E09 2.88E07 1.17E05 BJ 6.29E06 9.78E09 3.26E07 1.57E05 2.6.Conditional Rupture Mode Probability Model This section illustrates how we plan to use information from NUREG1829 as input to develop the conditional rupture probability models for selected components. This material is still under review and is subject to change in the final submittal. We plan to review the supporting information from NUREG 1829 that was recently made available by the NRC, and we may modify our approach after we have completed our evaluation of that supporting information. Importantly, our approach includes a step to calculate the total LOCA initiating event frequencies from all the modeled locations and to perform a sanity check to ensure that the results obtained from this bottomup process yields reasonable results. In addition we plan to compare the aggregated results from our approach with the aggregated LOCA frequencies in NUREG1829 and other relevant sources and identify a technical basis for the accepting the results of such comparisons.

2.6.1. Use of NUREG1829 Data The expert elicitation that was performed and documented in NUREG1829 [1] provided estimates of the frequencies for loss of coolant accidents based on a set of LOCA categories selected to span the break sizes and leak rates that are normally modeled in PWR and BWR PRAs. The estimates provided in NUREG1829 included both pipe failures and nonpipe failures. However, in this example only the pipe failure part is considered. LOCAs caused by nonpipe failures will be addressed in 2012. The LOCA 14 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 categories for PWRs used in NUREG1829 are summarized in Table 29. Since the largest pipes in a PWR reactor coolant system, which correspond to the cold leg piping, are on the order of 31 nominal pipe size (NPS), the NUREG1829 LOCA categories do not include a doubleended guillotine break (DEGB). The effective break size of a cold leg pipe of 31 NPS would be about 44.

The approach to using information in NUREG1829 to develop estimates of the conditional probability of pipe ruptures is based on the following observations and information presented in that document.

Base case results are presented in the report for three welldefined piping components for PWRs, namely, hot leg piping, pressurizer surge line piping, and high pressure injection piping, which comprise part of the ASME Class 1 pressure boundary. For each component, there were four different and independent estimates provided for each applicable LOCA category, two of which were based on a statistical analysis of service data and simple models similar to those that will be used in the STP GSI191 evaluation, and two based on probabilistic fracture mechanics analyses. These base case results were provided as input to the experts, and some experts chose to use these base case results as anchors for their respective inputs. The base case results are summarized in Section 4 of NUREG1829 as well as in the supporting appendices.

Table 29 NUREG1829 and STP PRA LOCA Categories Effective LOCA Flow Rate STP PRA Category Break Size Category (gpm)

(in.)

1 Small LOCA(1) 0.5 100 2 Medium LOCA(1) 1.5 1,500 3 3 5,000 4 Large LOCA 6.75 25,000 5 14 100,000 6 31.5 500,000 Note (1) The breakpoint between Small and Medium LOCA in the STP PRA and most PWR PRAs is actually 2.

As part of the elicitation, most of the experts provided input to the estimation of LOCA frequencies for specific components in the reactor coolant system pressure boundary, including the components that were evaluated in the base case results as well as essentially all the major components on the Class 1 pressure boundary. Selected componentlevel results of this elicitation are found in Appendix L of NUREG1829. For PWRs, these results are presented for LOCA Categories 1, 3, and 5. An example of the form of this information for LOCA Category 1 is shown in Figure 22. There is also componentlevel expert elicitation information presented in this appendix for hot leg piping for LOCA Category 6. The NUREG1829 supporting information that was just recently released has additional information on component level LOCA estimates for LOCA frequencies that we have not yet been able to analyze.

In the evaluation of service data that was performed in support of NUREG1829, which includes the base case analyses performed by Bill Gallean and Bengt Lydell, none of the reviewed service 15 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 data involved the occurrence of a LOCA of any of the 6 LOCA categories. The service data we have collected in Table 21 and 22 for these systems, a total of 166 pipe failures, include flaws, cracks, and rather small leaks, but no leaks of the magnitude that would classify as a small LOCA, which corresponds to LOCA Category 1. The pipe rupture models used in the base case studies of Lydell and Gallean and the one used in this study, assumes that each pipe failure is a precursor to a LOCA. Each of these models starts with an estimate of the failure rate, which includes all pipe failures requiring repair or replacement. The bridge between the failure rates, which are estimated using service data, and the more significant pipe failures producing LOCAs, is the model for the conditional probability of a break of a given size given a pipe failure.

Another way to look at this model is that pipe failures are assumed to represent challenges to the system and upon each challenge, there is a probability of experiencing a break of a given size. By considering all the possibilities for different break sizes, all the LOCA frequency categories can be quantified.

Using the above information and insights, our approach to using information from NUREG1829 is to convert information that was presented in the form of LOCA frequencies vs. LOCA category to conditional probabilities vs. break size. This approach is applied to the three PWR components that were included in the base case results as well as in Appendix Lnamely, the RCS hot leg, the RCS surge line, and the HPI injection line. These span a representative range of nominal pipe sizes in the PWR Class 1 pressure boundary of 30, 14, and 3.75, respectively.

Figure 22 Category 1 LOCA Frequencies for PWR Piping Systems at 25 Years of Plant Operation (Reproduced from Figure L.13 in NUREG1829) 16 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 2.6.2. Model for Deriving Conditional Probabilities from Rupture Frequencies The model used to convert information on unconditional rupture frequencies to conditional failure probabilities makes use of the base case results of Lydell for each of the three selected PWR components (hot leg, surge line, HP injection line) and the following equation:

F ( LOCA j ) ml l P( R j F ) (4) l Where:

F ( LOCA j ) Unconditional frequency of LOCA Category j due to pipe failures in selected component, per reactor calendaryear ml Number of pipe welds of type l in selected component having the same failure rate l Failure rate per weldyear for pipe weld type l within the selected component in Lydells base case analysis from Appendix D in NUREG 1829 P( R j F ) Conditional probability of LOCA Category j given failure in selected component Each term in this model is subject to epistemic uncertainty, which is to be estimated. Therefore, the approach is to use this model and the base case analysis of the failure rates from Lydell to derive epistemic uncertainties on the conditional probability of pipe rupture in each LOCA category that produces the same distribution of unconditional LOCA category frequencies as presented in Appendix L.

This approach makes use of there being a technical basis for the failure rate estimates from service data and a wellreviewed and extensively applied Bayes uncertainty analysis method, and these estimates were part of the information that other experts used to anchor their inputs. Since there have been no Category 1, 2, 3, 4, or 5 LOCAs, the expert elicitation results of all the experts constitute a kind of extrapolation of the existing service data. Therefore, our approach simply assumes that the variability in the expert elicitation inputs for LOCA frequency represents the epistemic uncertainty in the LOCA frequency for each component. This epistemic uncertainty is then assumed to result from the combination of the epistemic uncertainty in the failure rate and the epistemic uncertainty in the conditional probability of each LOCA category.

This model is similar to but somewhat simplified in comparison to the Lydell base case analysis in Appendix D of NUREG1829. Lydells base case analysis uses different conditional LOCA category probabilities for different loading conditions and then combines them to produce his base case results.

So, before incorporating the expert elicitation results into this model, we shall derive an equivalent conditional probability model using Equation (4) that attempts to reproduce the Lydell base case results in order to benchmark this model against the slightly different model used in the Lydell base case results. Then we shall adjust the epistemic uncertainties in the conditional probability of a LOCA in a manner that matches target LOCA frequencies that are set to incorporate the variability among experts estimates in NUREG1829.

17 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Using the same Microsoft Excel' and Oracle Crystal Ball' files that Lydell used to develop his base case results, the simplified model of Equation (4) was applied to the same failure rate estimates that Lydell derived and documented in Appendix D of NUREG1829, assuming a lognormal distribution for the conditional LOCA category probability for each component. This resulted in lognormal parameters that essentially reproduce Lydells Appendix D results, as shown in Figures 23, 24, and 25, for the HPI injection line, RCS surge line, and RCS hot leg, respectively. The method used to obtain these was to first develop new output parameters that calculate the equivalent conditional probabilities and then to adjust the resulting distributions by trial and error until a good match was determined. The parameters of these lognormal distributions are shown in Table 22. The figures comparing the base case results from NUREG1829 with the results obtained using the equivalent lognormal distributions indicate excellent agreement. The underlying lognormal distribution parameters for the conditional LOCA probabilities in Table 210 are reasonable. It is noted that the conditional probability of a given break size is inversely proportional to pipe size.

Figure 23 Benchmarking Lognormal Distributions to Lydell Base Case Results - HPI Injection Line 18 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 24 Benchmarking Lognormal Distributions to Lydell Base Case Results - RCS Surge Line Figure 25 Benchmarking Lognormal Distributions to Lydell Base Case Results - RCS Hot Leg 19 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 210 Lognormal Distributions for Conditional LOCA Category Probabilities that Match Lydells Base Case Results Component LOCA Break Size Median Mean Range 5%tile 95%tile Category (in.) Factor RCSHL 1 .5 7.86E4 2.16E3 10.4 7.56E5 8.16E3 2 1.5 5.10E5 1.56E4 11.7 4.38E6 5.94E4 3 3 2.01E5 6.06E5 11.6 1.73E6 2.31E4 4 6.76 7.42E6 2.29E5 11.8 6.29E7 8.75E5 5 14 2.74E6 8.43E6 11.8 2.33E7 3.23E5 6 31.5 1.32E6 4.06E6 11.8 1.12E7 1.56E5 RCSSurge 1 .5 5.12E3 6.99E3 3.67 1.40E3 1.87E2 Line 2 1.5 4.02E4 6.06E4 4.44 9.06E5 1.78E3 3 3 1.59E4 2.43E4 4.53 3.51E5 7.23E4 4 6.76 5.80E5 9.03E5 4.70 1.23E5 2.73E4 5 14 2.28E5 3.44E5 4.43 5.15E5 1.01E4 HPI 1 .5 7.68E3 1.06E2 3.72 2.06E3 2.86E2 2 1.5 1.04E3 1.57E3 4.44 2.35E4 4.64E3 3 3 4.10E4 6.26E4 4.54 9.03E5 1.86E3 2.6.3. Incorporation of Epistemic Uncertainties from NUREG1829 The next step is to adjust the lognormal distributions to reflect the variability of expert opinion regarding the frequency of each LOCA category, based on the information in Appendix L of NUREG1829 and supporting information available on the NRC website that provides the results of individual experts estimates of LOCA frequencies for specific components. This process begins by superimposing the extreme maximum and minimum values from the box and whisker plots on LOCA frequencies vs. LOCA category (and hence, break size) in Appendix L of NUREG1829 for the three PWR components analyzed in the Lydell base case analysis and comparing these extreme values to the Lydell base case results. Then a target 95%tile line and a target 5%tile line are drawn on a loglog plot of the corresponding LOCA frequency vs. break size curve, in order to capture the Appendix L results of the expert elicitation. The conditional probability vs. break size model that was originally developed to benchmark the Lydell base case results is then adjusted so that the revised LOCA frequency vs. break size model approximates the targets. The application of this approach to the HPI injection lines is shown in Figure 26 and Figure 27.

The former figure shows how the results of the expert elicitation in Appendix L compare to the Lydell base case results and how the Appendix L results for LOCA Categories 1 and 3 are used to construct the target distributions for this component. The latter figure shows how the target values were used to benchmark the conditional probability models for this component.

More recently, NRC has made available more detailed information on the expert estimates of LOCA frequencies for individual components that covers more LOCA sizes than covered in the plots in NUREG 1829 Appendix L. A total of 9 experts out the total 12 panelists provided information at this level. Using this information, an analysis was performed in which each experts input was treated as a lognormal distribution and a posterior weighting procedure was applied to synthesize the 9 expert inputs into a composite uncertainty distribution giving each expert equal weight. The 95%tile and 5%tiles from this 20 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 composite distribution were then used in lieu of the values read off the box and whisker plots in Appendix L of NUREG1829, and this provided a more complete, more accurate, and appropriately probabilistically weighted characterization of the NUREG1829 input to selection of target LOCA frequencies. This is the approach that is adopted for the remaining steps of the analysis for all components.

Figure 26 Definition of Target Bounds for HPI Injection Line LOCA Frequencies This case was rather simple because the expert input for this component agrees well with the results of the Lydell benchmark, and Appendix L includes results for only two LOCA categories, which makes it easy to define a line on a loglog plot. This figure shows that conditional rupture probabilities that meet the targets in Figure 25, when combined with the Lydell HPI leak frequencies, will encompass the results of the expert elicitation. In the next update, the HPI model will be calibrated using the experts composite distribution approach described above in lieu of Appendix L information.

The application of the experts composite distribution procedure to the RCSsurge line is shown in Figures 28 and 29. Figure 28 compares the elicitation results in the form of the 95%tile and 5%tile of the experts composite distribution for this component with the Lydell base case results and also presents the proposed targets for the conditional rupture probability model.

21 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 27 Calibration of Conditional Probability Model to Match Targets for HPI Injection Line Also shown in Figure 28 are the 95%tile and 5%tile of the surge line failure rate distribution which was used in the Lydell Base Case results. In this case, there a huge variability in the elicitation results for the surge line, a full six orders of magnitude between the maximum and minimum for LOCA categories 1 and 3, and seven orders for Category 5. The targets for the 5th percentile for this component are based on the lower bound results from the elicitation, causing a straight line on the loglog plot on Figure 28.

When selecting the 95th percentile for this component, it was noted experts composite estimates are unreasonably high, as they produce, when combined with the Lydell estimates of the underlying component failure rates, conditional rupture probabilities greater than one. In other words, the experts at the extreme high end produce higher estimates of LOCA frequencies, which have never occurred, than estimates of the underlying failure rates provided in the Lydell base case results, which are well supported by service experience data. The data query presented previously includes 166 pipe failures experienced over nearly 4,000 reactoryears of service data with no leaks exceeding about 10gpm, much smaller than the high leak rates of these LOCA categories. The failure rates include any events in which repair or replacement of the weld was required, yet none of the experienced failures involved significant leakage and certainly no LOCAs in Categories 1, 2, 3, 4, or 5. As will be shown below, the target 95th percentile selected for the surge line, while set somewhat lower than the upper bound values from the experts composite distribution, still yields very high conditional rupture probabilities.

22 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 28 Definition of Target Bounds for RCS Surge Line LOCA Frequencies The application of this procedure for the RCS hot leg is shown in Figures 29 and 210. In Figure 29, the targets for the 95%tiles and 5%tiles are set to capture the range of results from the expert elicitation while assuming a straight line on a loglog plot of LOCA frequency vs. break size. The anchor point for LOCA Category 1 is taken as 10% of the 95%tile of the surge line failure rate distribution of the Lydell Base Case analysis. This characterization of the upper bound target LOCA frequencies yields, as will be shown below, very high conditional probability of LOCA values.

The analysis for the RCS hot leg component is shown in Figures 210 and 211 using the graphical procedure applied above to the hot leg. This analysis will be replaced by the experts composite approach in the next revision to this report.

The net result of this benchmarking was the derivation of the conditional probability model for these selected components, whose results are shown in Table 211. Because of the rather high upper bound LOCA frequencies embodied in the targets that were strongly influenced by the upper bound values from the expert elicitation, and the assumption of lognormal distributions for the conditional rupture probability at each LOCA size, the lognormal distributions had to be truncated at 1.0, otherwise a significant part of the upper tail of the distributions would exceed 1.0, especially for the RCS surge line and RCS hot leg models. As a result, the properties of the truncated distributions, such as medians, 23 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 means, and percentiles, are different from those of the underlying lognormal distributions prior to truncation.

A comparison of the conditional probabilities that were developed for these examples against the conditional probabilities that are linked to the Lydell base case analysis is shown in Figures 211, 212, and 213 for the HPI injection line, RCS surge line, and RCS hot leg, respectively. As expected, these plots track very closely the previous plots for the LOCA frequencies vs. break size because both sets of plots use the same failure rate model. However, this verifies that the results are internally consistent.

Figure 29 Calibration of Conditional Probability Model to Match Targets for RCS Surge Line 24 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 210 Definition of Target Bounds for RCS Hot Leg LOCA Frequencies Figure 211 Calibration of Conditional Probability Model to Match Targets for RCS Hot Leg 25 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 211 STP Conditional Probability Models Derived from Target LOCA Frequencies Distribution Input Parameters Truncated Distribution Parameters LOCA Break Component Median Range 5th 95th Category Size (in.) Type [Note (1)] Median Mean Factor Percentile Percentile 1 .5 1.39E04 7.40E+01 1.40E04 3.35E03 1.90E06 1.01E02 2 1.5 2.49E05 1.20E+02 2.48E05 1.29E03 2.05E07 2.99E03 RCSHot 3 3 8.65E06 1.63E+02 8.60E06 7.99E04 5.30E08 1.45E03 Leg 4 6.76 2.43E06 2.38E+02 2.43E06 4.32E04 1.06E08 5.73E04 5 14 8.10E07 3.23E+02 7.90E07 2.86E04 2.46E09 2.62E04 6 31.5 2.20E07 4.56E+02 2.19E07 1.46E04 4.97E10 9.92E05 Lognormal 1 .5 3.33E02 5.60E+01 2.59E02 1.05E01 5.38E04 5.36E01 truncated RCSSurge 2 1.5 4.12E03 5.48E+01 3.97E03 3.46E02 7.41E05 1.76E01 at 1.0 Line 3 3 1.36E03 6.80E+01 1.34E03 1.85E02 1.99E05 8.24E02 4 6.76 3.90E04 8.96E+01 3.88E04 9.09E03 4.34E06 3.32E02 5 14 1.30E04 1.19E+02 1.34E04 5.16E03 1.12E06 1.55E02 1 .5 5.85E03 1.52E+01 5.78E03 2.15E02 3.77E04 8.88E02 HPI Line 2 1.5 1.20E03 1.91E+01 1.18E03 5.87E03 6.34E05 2.30E02 3 3 4.56E04 2.21E+01 4.59E04 2.61E03 2.07E05 1.01E02 Note (1) These are medians to specify the input distribution to Crystal Ball' prior to truncation; the median of the truncated distribution is generally different following truncation.

26 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 212 Comparison of Conditional Probability Models for HPI Injection line Figure 213 Comparison of Conditional Probability Models for RCS Surge Line 27 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 214 Comparison of Conditional Probability Models for RCS Hot Leg 2.6.4. Bayes Update of the Conditional Probability Distributions The conditional probability model developed in the previous section is used as the basis for a prior distribution which we then update with the evidence from the service data which is no loss of coolant accidents out of 3 surge line weld failures. We are investigating whether to consider pooling the data over a larger data set for this purpose, but for the time being we take the conservative approach of limiting ourselves in the Bayes updateto the Surge line weld experience. In the next update to this report we will consider pooling the data for similar components for the purpose of characterizing the evidence for these conditional rupture probabilities. The truncated lognormal distributions described in Table 211 were used as prior distributions and then updated with 0 Category 1, 2, 3, 4, or 5 LOCAs out of 3 observed failures. The results are summarized in Table 212. It is noted that even though the evidence is rather weak, the 95%tiles of the posterior distributions are significantly reduced. This in turn influences the means of the distributions to a significant degree. Hence this procedure was important to apply.

28 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 212 Results of Bayes Update of Conditional LOCA Probabilities Distribution Parameters(2)

LOCA Break (1)

Distribution Type Range Category Size (in.) Mean 5%tile 50%tile 95%tile (3)

Factor 1 0.5 Prior Truncated Lognormal 1.05E01 5.38E04 2.59E02 5.36E01 53.6 Posterior LognormalBinomial 4.43E02 4.11E04 1.45E02 1.95E01 21.8 2 1.5 Prior Truncated Lognormal 3.46E02 7.41E05 3.97E03 1.76E01 48.8 Posterior LognormalBinomial 1.75E02 6.75E05 3.17E03 8.45E02 35.4 3 3 Prior Truncated Lognormal 1.85E02 1.99E05 1.34E03 8.24E02 64.3 Posterior LognormalBinomial 1.00E02 1.88E05 1.17E03 4.79E02 50.5 4 6.76 Prior Truncated Lognormal 9.09E03 4.34E06 3.88E04 3.32E02 87.5 Posterior LognormalBinomial 5.27E03 4.22E06 3.60E04 2.33E02 74.3 5 14 Prior Truncated Lognormal 5.16E03 1.12E06 1.34E04 1.55E02 117 Posterior LognormalBinomial 3.10E03 1.11E06 1.28E04 1.21E02 105 Notes (1) Prior lognormal distributions truncated at 1.0.

(2) Values for means and percentiles represent conditional probability of LOCA category given pipe failure.

(3) Range Factor = SQRT(95%tile/5%tile) 2.7.LOCA Frequencies Associated with Surge Line Welds 2.7.1. Base Case Results The application of this methodology to STP is a work in progress. This section illustrates the approach by developing a set of base case results as well as some sensitivity cases based on some assumptions about STP inputs.

The unconditional frequencies of LOCAs as a function of break size were obtained by Monte Carlo uncertainty propagation of the failure rate distributions and the conditional probability of break size distributions using the results of the previous sections and Equation (2). The results for each weld and for the total surge line LOCA frequency are shown in Table 213. These results are based on the input data that were assumed, which include the assumptions regarding susceptibility to damage mechanisms. The results for BF welds, for example, are heavily influenced by the inherent sensitivity to PWSCC. If this damage mechanism is mitigated, for example, by application of weld overlays, these welds would entail less frequent failure and fewer design and construction defects. The STP BF welds have in fact been repaired using weld overlays. These results also do not reflect the results of the STP RI ISI evaluation, which included a weldbyweld evaluation of damage mechanisms. The susceptibility of B J welds to thermal fatigue is assessed in the base case results based on generic data. If thermal fatigue and any other damage mechanisms to the BJ welds were ruled out, the failure rates of these welds would be determined solely by design and construction defects. As a sensitivity case, the total LOCA frequencies were reevaluated to reflect PWSCC mitigation by weld overlay and an assumption that no BJ welds are susceptible to thermal fatigue. A comparison of the mean surge line LOCA frequencies for the base case and this sensitivity case are shown in Figure 215, which shows the influence of damage 29 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 mechanism sensitivity assumptions. In the submittal, an STPspecific evaluation of damage mechanisms will be incorporated into these results.

Figure 215 Comparison of Surge Line LOCA Frequencies with Different Damage Mechanism Inputs 2.7.2. Influence of NDE Inspections on LocationSpecific LOCA Frequencies All the results presented up to this point have included the effects of piping inspections and integrity management programs only implicitly, in that the failure rate data and inputs to the expert elicitation of NUREG1829 that form the basis for our conditional probability of the LOCA model have been based on average nuclear power plant service data. These averages, benefited from reliability integrity management programs including testing and monitoring for leaks as well as nondestructive examinations that are performed in the various ISI programs on an average basis. In deriving location specific LOCA frequencies for STP, the actual status of each weld as being either included or excluded from the NDE program will be taken into account using the Markov model developed for this purpose in the EPRI RIISI program. The details of this application will be documented in the submittal. An example of the kind of changes in LOCA frequencies that can result from locationbylocation changes in the pipe inspection and leak monitoring program is shown in Figure 216 for an RCS weld subject to stress corrosion cracking [13]. As seen in this figure, the frequency of a pipe break may change by more than an order of magnitude, due to changes in the reliability integrity management program, all other factors being equal.

30 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Table 213 Unconditional LOCA Frequencies for Surge Line Welds Distribution Parameters Weld Type Parameter RF(2)

Mean 5%tile 50%tile 95%tile Failure Rate 5.61E04 1.37E04 4.38E04 1.40E03 3.2 Category 1 2.13E05 1.44E07 3.81E06 9.35E05 25.5 Category 2 1.03E05 2.44E08 1.01E06 4.26E05 41.7 BF Category 3 6.71E06 6.94E09 4.10E07 2.38E05 58.6 Category 4 3.67E06 1.55E09 1.36E07 1.16E05 86.6 Category 5 2.22E06 4.27E10 5.05E08 5.95E06 118.0 Failure Rate 4.16E06 9.75E09 2.91E07 1.17E05 34.7 Category 1 1.43E07 2.65E11 2.58E09 2.87E07 104.1 Branch Category 2 8.14E08 5.02E12 6.87E10 1.12E07 149.3 Connection Category 3 4.35E08 1.51E12 2.81E10 5.73E08 194.9 Category 4 2.87E08 3.68E13 9.32E11 2.55E08 263.1 Category 5 1.45E08 1.06E13 3.45E11 1.23E08 339.4 Failure Rate 2.03E07 2.66E11 2.86E09 3.74E07 118.6 Category 1 4.38E08 1.12E12 2.68E10 6.33E08 238.3 Category 2 2.13E08 2.34E13 7.40E11 2.38E08 319.0 BJ Category 3 1.49E08 7.21E14 2.98E11 1.15E08 398.5 Category 4 8.68E09 1.85E14 9.77E12 4.92E09 515.9 Category 5 2.07E09 1.26E13 2.23E11 3.84E09 175.0 Failure Rate 5.71E04 1.37E04 4.39E04 1.42E03 3.2 Category 1 2.19E05 1.44E07 3.81E06 9.45E05 25.6 Base Case Category 2 1.06E05 2.45E08 1.02E06 4.29E05 41.9 Total Surge Category 3 6.90E06 6.94E09 4.11E07 2.40E05 58.8 Line(1)

Category 4 3.79E06 1.55E09 1.36E07 1.17E05 86.9 Category 5 2.26E06 4.28E10 5.07E08 6.00E06 118.4 Case with BF Failure Rate 1.50E05 2.37E08 9.77E07 5.13E05 46.5 weld overlay Category 1 4.39E07 5.35E11 5.45E09 7.25E07 116.4 and no TF Category 2 3.04E07 1.01E11 1.46E09 3.02E07 172.9 Susceptibility for BJ welds Category 3 1.90E07 3.04E12 6.01E10 1.68E07 235.3 Category 4 1.46E07 7.41E13 2.02E10 8.45E08 337.8 Category 5 8.73E08 2.17E13 8.12E11 5.29E08 493.6 Notes (1) Total surge line results are based on 1 BF weld, 2 BC welds, and 6.9 BJ welds.

(2) RF = SQRT(95%tile/5%tile) 31 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 Figure 216 Comparison of Weld Failure Rates Determined by Markov Model for Different Reliability Integrity Management Approaches 2.8. LOCA Frequency Summary The technical approach to estimation of LOCA frequencies for the STP GSI191 project has been described in this section, with some preliminary results for each step. Three categories of welds in the pressurizer surge line have been selected to illustrate most of the steps of the approach. When all the components in the STP Class 1 pressure boundary are completed, the results will be combined to determine the LOCA initiating event frequencies for the PRA model, and the locationspecific results will be provided to the Casa Grande model to determine the phenomenological impacts associated with debris formation. The specific capabilities that have been demonstrated include:

The capability to estimate LOCA frequencies as a function of break size at each location.

The capability to utilize information from NUREG1829 to characterize epistemic uncertainty associated with LOCA frequencies.

A method that incorporates via Bayes uncertainty analysis the service data on pipe failures and component exposures.

A full quantification of epistemic uncertainties associated with estimating the input parameters in the model equations.

32 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191 The capability to quantify the impacts of information on degradation mechanism susceptibility at each location, based on insights from service data and results of RIISI evaluation.

The capability to adjust the results to account for locationbylocation differences in the reliability inspection program for leak monitoring and NDE.

The submittal will include a more complete application of this methodology as well as further justification for supporting assumptions and input data.

As noted earlier, the quantitative information presented is based on a work in progress and is subject to change as the project unfolds. All results are preliminary and are presented primarily to describe how the analyses will be performed.

3. References

[1] Tregoning, R., L. Abramson, and P. Scott, Estimating LossofCoolant Accident (LOCA)

Frequencies Through the Elicitation Process, NUREG1829, U.S. Nuclear Regulatory Commission, Washington, DC, April 2008.

[2] Lydell, B. O. Y., PIPExp/PIPE2010: Monthly Summary of Database Content (Status as of 31Mar 2010), RSA Technologies, Fallbrook, CA. Monthly summary reports have been issued since January 1999.

[3] Fleming, K. N. and B. O. Y. Lydell, Database Development and Uncertainty Treatment for Estimating Pipe Failure Rates and Rupture Frequencies, Reliability Engineering and System Safety, 86: 227-246, 2004.

[4] Pipe Rupture Frequencies for Internal Flooding PRAs, Revision 1. EPRI, Palo Alto, CA: 2006.

1013141.

[5] Pipe Rupture Frequencies for Internal Flooding PRAs, Revision 2. EPRI, Palo Alto, CA: 2010.

1021086.

[6] Piping System Reliability and Failure Rate Estimation Models for Use in RiskInformed InService Inspection Applications. EPRI, Palo Alto, CA: 1998. TR110161.

[7] Piping System Failure Rates and Rupture Frequencies for Use in RiskInformed InService Inspection Applications. EPRI, Palo Alto, CA: 1999. TR111880.

[8] Mosleh, A. and F. Groen, Technical Review of the Methodology of EPRI TR110161, University of Maryland report for EPRI, published as an Appendix to EPRI TR110161 (Reference [6]).

[9] Revised RiskInformed InService Inspection Procedure. EPRI, Palo Alto, CA: 1999. TR112657, Rev. BA.

[10]U.S. Nuclear Regulatory Commission, Safety Evaluation Report Related to Revised RiskInformed InService Inspection Evaluation Procedure: EPRI TR112657, Rev. B, July 1999, Washington, DC, 1999. Published as a forward to TR112657 (Reference [9]).

[11]Martz, H., Final (Revised) Review of the EPRIProposed Markov Modeling/Bayesian Updating Methodology for Use in RiskInformed InService Inspection of Piping in Commercial Nuclear Power Plants, Los Alamos National Laboratory, June 1999. TSA1/99164.

33 KNF Consulting Services LLC

LOCA Frequencies for STP GSI191

[12]Fleming, K. N., Markov Models for Evaluating Risk Informed InService Inspection Strategies for Nuclear Power Plant Piping Systems, Reliability Engineering and System Safety, 83(1): 27-45, 2004.

[13] Fleming, K.N. et al., Treatment of Passive Component Reliability in RiskInformed Safety Margin Characterization - Fiscal Year 2010 Status Report, report prepared by Pacific Northwest National Laboratory for the U.S. Department of Energy, September 2010.

[14] U.S. Nuclear Regulatory Commission, Supporting information for NUREG1829 (Reference [1])

on Individual Experts Estimates of LOCA frequencies for specific components and LOCA Categories, available on ADAMS Accession Numbers ML080560008, ML080560010, ML080560011, ML080560013 34 KNF Consulting Services LLC