ML14234A282

NRR E-mail Capture - STP risk-informed GSI-191 Meeting on 8/20/14

ML14234A282
Person / Time
Site:	South Texas
Issue date:	08/18/2014
From:	Harrison A South Texas
To:	Balwant Singal Division of Operating Reactor Licensing
References
MF2400, MF2401
Download: ML14234A282 (82)

Text

NRR-PMDAPEm Resource From: Harrison Albon [awharrison@STPEGS.COM]

Sent: Monday, August 18, 2014 7:05 PM To: Singal, Balwant Cc: Oesterle, Eric; Mitchell, Eliza

Subject:

RE: Out-of-Office (August 18 to August 29, 2014)

Attachments: 8-20-14 NRC Meeting backup slides.pptx Here are the backup slides in case we need them.

From: Harrison Albon Sent: Monday, August 18, 2014 5:55 PM To: 'Singal, Balwant' Cc: Oesterle, Eric; Mitchell, Eliza

Subject:

RE: Out-of-Office (August 18 to August 29, 2014)

Here is the main set of slides for the STP risk-informed GSI-191 meeting on 8/20/14.

Call me if you have questions.

Wayne Harrison STP Licensing (979) 292-6413 From: Singal, Balwant [1]

Sent: Thursday, August 14, 2014 2:45 PM To: Harrison Albon Cc: Oesterle, Eric; Mitchell, Eliza

Subject:

RE: Out-of-Office (August 18 to August 29, 2014)

Wayne, Wayne, Please copy the following NRC staff members on your e-mail forwarding the presentation slides:

Oesterle, Eric Eric.Oesterle@nrc.gov Mitchell, Eliza Eliza.Mitchell@nrc.gov Thanks.

Balwant K. Singal Senior Project Manager (Comanche Peak, STP, Diablo Canyon, and Palo Verde)

Nuclear Regulatory Commission Division of Operating Reactor Licensing Balwant.Singal@nrc.gov Tel: (301) 415-3016 Fax: (301) 415-1222 1

From: Harrison Albon [2]

Sent: Thursday, August 14, 2014 12:58 PM To: Singal, Balwant Cc: Lyon, Fred; Blossom, Steven; Kee, Ernie

Subject:

RE: Out-of-Office (August 18 to August 29, 2014)

Balwant, You asked yesterday when we would have slides to you for the 8/20 meeting. Well have them to you (or Fred) by COB on Monday, probably before.

Regards, Wayne Harrison STP Licensing (979 292-6413 From: Singal, Balwant [3]

Sent: Thursday, August 14, 2014 10:30 AM To: 'Hope, Timothy' (Timothy.Hope@luminant.com); Sterling, Lance; Harrison Albon; Carl.Stephenson@aps.com; pns3@pge.com Cc: Lyon, Fred; Watford, Margaret; Oesterle, Eric; Markley, Michael

Subject:

Out-of-Office (August 18 to August 29, 2014)

I will be out-of-office from August 18 to August 29, 2014. Please contact the following NRC staff members for Project Manager assistance:

Fred Lyon at 301-415-2296 for Comanche Peak, South Texas Project, and Diablo Canyon.

Eric Oesterle at 301-415-1014 for Palo Verde.

Thanks.

Balwant K. Singal Senior Project Manager (Comanche Peak, STP, and Palo Verde)

Nuclear Regulatory Commission Division of Operating Reactor Licensing Balwant.Singal@nrc.gov Tel: (301) 415-3016 Fax: (301) 415-1222 2

Hearing Identifier: NRR_PMDA Email Number: 1521 Mail Envelope Properties (8C918BCF8596FB49BD20A610FA5920CF0208358E)

Subject:

RE: Out-of-Office (August 18 to August 29, 2014)

Sent Date: 8/18/2014 7:05:10 PM Received Date: 8/18/2014 7:07:08 PM From: Harrison Albon Created By: awharrison@STPEGS.COM Recipients:

"Oesterle, Eric" <Eric.Oesterle@nrc.gov>

Tracking Status: None "Mitchell, Eliza" <Eliza.Mitchell@nrc.gov>

Tracking Status: None "Singal, Balwant" <Balwant.Singal@nrc.gov>

Tracking Status: None Post Office: CEXMBX03.CORP.STPEGS.NET Files Size Date & Time MESSAGE 2666 8/18/2014 7:07:08 PM 8-20-14 NRC Meeting backup slides.pptx 1122509 Options Priority: Standard Return Notification: No Reply Requested: No Sensitivity: Normal Expiration Date:

Recipients Received:

Backup Slides CASA Development Plan

• Aggressive SQA plan in place to achieve Appendix B certification

- Alion SQA plan meets intent of IEEE standards and q

requirements from manyy pplants

- Specifically checked for STP approved software list

• Source code under configuration management

- TortoiseSVN (windows shell to Subversion©)

- CM plan approved

• Version 1.7 to release following V&V (Nov 1st???)

- Changes made to answer RAIs

- Address issue tracking reports

- Generalize sump configuration definition

- Auto documentation feature 2

CASA V&V Elements

• Software Requirements Specification

- Functional description of capabilities

• Theory Manual

- Reference e e e ce independent depe de t from o SRS S S for o convenient co e e t updates

• Software Design Description

- Subroutine-level architecture

• Users Guide

- Installation process and verification

- Example problem, definition of user input

• Verification V ifi ti TTestt Pl Plan anddRReportt

- Comprehensive check of suitability/adequacy

• Issue Tracking Report (ITR) system

- Bugzilla report and disposition (integrated with SVN)

- Issue notification chain (Alion to users, users to NRC) 3

CASA V&V Activity Status

• Software Reqs Specification - final review

• Theory Manual - in preparation

• Software Design Description - final review

• Users Guide - v1.7 almost complete

• Verification Test Plan (with archive documentation)

- Now building checklists from SRS

- Vol 3 equation verification by independent team - >50% complete

• Confirm printed equation

• Confirm implementation

• Confirm I/O and results using test routines (partial)

- Automated Tests ((dryy runs complete) p )

• Data Arrays, Function calls, input data checks

- Vignette case studies - beginning

• Issue Tracking Reports

- Populating Bugzilla archive - in progress

- Issue disposition - in progress 4

Purpose of VISTA Correlation

• Independently confirms conservative application of NUREG/CR-6224

- No change to LAR is proposed

• Addresses NRC concerns with 6224

- Factorization F i i off porosityi ((exponents off ))

- Uniform bed compression (now differential)

- Limited range of test conditions (Re scaling)

- Stratified bed configurations (case studies) 5

Re Scaling in Viscous/Inertial Transition Lapple and Shepard (1940)

Viscous Regime Inertial Regime (Momentum Transfer)

Classic Cl i experiments i suggest that h totall h hydraulic d li d drag can b be ddescribed ib d b by a low-order function of Reynolds number in the viscous/inertial transition.

6

VISTA Attributes

• Good agreement ag ee e t with t HTVL test data

• Robust Reynolds number correlation confirms pp applicability y of existingg test data to STP

• Exponential drag law - (Reynolds 1883)

- Preserves both theoretical limits (Raleigh 1892)

• Stokes (viscous), Newton (inertial)

• Adapts to transition because coefficients are also l fit as ffunctions ti off R Re

• Maximizes use of independent debris properties

• Still S ill sensitive i i to b bedd compression/strata i /

7

VISTA Findings

• Bed ed co configuration gu at o iss tthee most ost se sensitive s t e remaining e a g assumption

- Uncertainty in spatial profile of porosity and surface area lead to largest discrepancy between prediction and measurement

- Justifies STP assumption of maximum compression

• Independent confirmation of head loss for STP Reynolds y flow conditions illustrates 6224 conservatism as applied

- Maximum bed compression

- Factor of 5 uncertainty measure 8

L* Approach an supporting CHL calculation l l

• STP CHL Correlation Objective

- Address concerns that the conservatism in chemically-induced chemically induced head loss (CHL) quantification is based on engineering judgment

- Provide technical validation based on best available data that the chemical bump up is conservative 9

LL* End Product

• Technical evaluation of bump bump up up based on engineering judgment

• Provides a supplemental methodology for chemical effects quantification that improves resolution by using strainer test data and WCAP-16530-NP calculator results 10

LL* Development

• Evaluation of both strainer and vertical head loss test

- Concept of L* developed

• grams of precipitate available to filter across a strainer surface area

• Allows comparison of strainer results to vertical head loss results

• Allows use of deterministic tools in risk informed space

• CHL investigated as a function of precipitate type

- AlOOH CHL per gram was largest response

• Enhanced conservatism duringg data miningg

- Removal of declining or non-increasing head loss when chemicals are added 11

LL* Correlation 10 9

8 7

6 CHL (ft) 5 4

3 2

STP strainer test data Correlation 1

0 0 500 1000 1500 2000 2500 3000 L* (g/m2)

• Correlation provides conservative CHL results when compared to actual test data 12

L* Evaluation Compared to Strainer 16 Thin Bed Tests 14 Full Bed 1 Full Bed 2 12 STP CHL correlation 10 CHL (ft) 8 6

4 2

0 0 200 400 600 800 1000 1200 1400 L* ((g/m2)

/ 2) 1.8 STP CHL correlation 1.6 St. Lucie 2 DBA strainer results 1.4 1.2 CHL (ft) 1.0 0.8 0.6 04 0.4 0.2 0.0 0 200 400 600 800 1000 13 1200 L* (g/m2)

L* Compared to Multiplicative Chemical h l HL Inflation fl 12 Chemical bump-up occurs Bump Up Train 1 Bump Up Train 2 Bump up Train 3 10 Correlation Train 1 Correlation Train 2 Correlation Train 3 8

NCHL Train 1 THL (N NCHL+ CHL) (ft)

NCHL Train 2 6 NCHLTrain 3 4

2 0

0 500 1000 1500 2000 2500 Ti Time (min)

( i )

NCHL - non-chemical head loss or conventional CHL- chemically induced head loss 14 THL - total head loss

Conclusions of L* and bump up comparison i

• Evaluation of the bump bump up up approach by the supplemental CHL approach, which is derived from technical data

- Provides supports that the bump up approach conservatively assesses risk

- Identifies Id tifi iimprovementt opportunities t iti ffor th the bbump-up approach

• Improvements only increase resolution of total head loss quantities and does NOT change risk 15

SSIB, Head Loss: RAI 15

• The NRC staff has generally not accepted correlations for the qualification of PWR strainers for several reasons. Please, explain why the following general concerns with the use of correlations are not an issue for the STP application:

a) Correlations have not been validated for the full range of debris loads and morphologies b) Correlations do not address nonhomogeneous debris beds c)) Correlations C l i h have not bbeen validated lid d ffor the h ffullll range of potential flow conditions and strainer geometries d)) There is significant g uncertaintyy in the model parameters used to describe physical attributes of the constituents 16

SSIB, Head Loss: RAI 16

• The staff is concerned that the validation testing is not representative of the plant. Please, provide additional information:

a) If the vertical loop tests are important to the conclusions, please, provide details why the STP HTVL tests are valid, considering that similar tests in different facilities had different results b) How was it determined that debris transported to a horizontal strainer is similar to transport to a plant strainer (with similar head loss) c) Demonstrate that the correlation used by STP is valid for plant specific geometry and plant conditions d)) Discuss how NUREG/CR-6224

/ could be used to p predict HL expected under conditions of STP flume tests 17

SSIB, Head Loss: RAI 17

• there is little or no testing that has been conducted under conditions similar to those at STP.

a) Debris constituents in validation testing are not plant specific b) Debris sizes in validation testing are not plant-specific c) Little validation testing conducted at STP velocities and none validated the correlation d) Validation testing did not include prototypical strainer geometries e) HTVL testing did not simulate potentially important aspects of debris bed formation f) Records from early testing not available, so conclusions from early testing must be limited.

limited 18

SSIB, Head Loss: RAI 18

• Implementation of the correlation in the STP model makes specific assumptions. Justify that the assumptions and use of correlation is realistic or conservative a) Beds are homogeneous and representative of the plant b) Bed is assumed to accumulate with the manufactured density. Please, explain why this is valid or requantify using new density c) Please, explain how NUREG/CR-6224 correlation compression function is applied d) Please, explain why linear mass weighting for surface-to-surface to volume ratio is acceptable e) Provide technical basis for coating material packing fractions 19

SSIB, Invessel: RAI 37

• Please provide the technical basis for assuming that 7.5 grams is an acceptable limit for a cold-leg break at STP when considering the potential for boric acid precipitation.

20

SSIB, Invessel RAI 37 Response

• 7.5

.5 g/

g/FA iss a threshold t es o d o of co concern ce

- sharp, single-value to maintain clarity on performance metric

• Full debris deposition on the fuel for conservatism

- no credit di for f barrel-to-baffle b l b ffl bypass b fib fiber d deposits i or lower plenum mixing

• Failure at 7 7.55 g/FA enters the core

• Lower than 15g/FA chosen after WCAP chemical oad added to fuel load ue 21

APLAB RAIs

• CASA Grande - Plant Configuration: RAI 1b 1b, 2b, 3

• HRA: RAI 3, 3 5

• Uncertainties: RAI 1, 2, 4, 5, 6

• Stable end state: PRA Success RAI 3c

• Use of different distributions 22

APLA, CASA CASA-Plant Plant Config: RAI 1b

• Provide technical justification for using only nominal values of time-temperature curves 23

APLA, CASA-Plant Config: RAI 1b

Response

• Not possible to choose conservative profiles across entire duration of the event

• Performed additional TH analyses to support use of nominal temperature profiles Workingg Workingg Workingg Breakk Size B Si Working W ki Case HHSI LHSI Cont. Fan Case Description (Diameter) CS Pumps Pumps Pumps Coolers 15"-9 15 inches 3 1 3 6 Dual LHSI Pump (Loops 3 & 4) Failure 15"-22a 15 inches 2 2 2 6 Single Train (Loop 4) Failure 15"-22b 15 inches 2 2 2 6 Single Train (Loop 3) Failure 15"-22a-4/6Fans 15 -22a-4/6Fans 15 inches 2 2 2 4 Single Train (Loop 4) Failure (4 Cont Cont. Fans Operating) 15"-26a 15 inches 1 2 2 6 Single Train (Loop 4) + HHSI Pump (Loop 3) Failure 15"-26b 15 inches 1 2 2 6 Single Train (Loop 3) + HHSI Pump (Loop 4) Failure 15"-43 15 inches 1 1 1 6 Dual Train (Loops 3 & 4) Failure 8"-43 8 inches 1 1 1 6 Dual Train (Loops 3 & 4) Failure 24

APLA, CASA CASA-Plant Plant Config: RAI 2b

• Provide technical justification for assuming only nominal operating conditions of flow rates and thermal hydraulic conditions 25

APLA, CASA-Plant Config: RAI 2b

Response

• Use of nominal flow rates provides a more realistic evaluation of risk when competing factors make it difficult to define conservative conditions. APLA, CASA-Plant Config: RAI 2b

• Use of nominal values provides results that are reasonable and probable for use in a holistic risk holistic, risk-informed informed evaluation evaluation.

26

APLA, CASA CASA-Plant Plant Config: RAI 3

• A qua qualitative tat e aargument gu e t iss p provided o ded whyy a combination of pumps failing in the same train is worse than the same set of pumps failing in d ff different trains.

a) Justify this assumption and clarify whether an engineering i i analysis l i was performed f d b) State if this assumption always increases conditional probability of strainer failure c) State if this assumption always increases conditional probability of in-vessel in vessel failure 27

APLA, CASA-Plant Config: RAI 3

Response

• A cursory engineering analysis based on proportion of total flow to each strainer provided the basis for Assumption 2b

• Effects of debris penetration were found to contradict traditional engineering judgment

• Extra parameter evaluations were performed

- Small increase in CDF ((1.5%)) caused byy in-vessel failures for pump failures on separate trains

- Supplement provided with detailed statistics 28

APLAB, STP PRA Model-Human Reliability l b l Analysis:l RAI 3a

• Please state if CASA Grande models plant conditions that would occur if three containment spray trains were running, running ee.g.

g

- Sump flow rates

- Washdown rates

- RWST drain-down times 29

APLAB, STP PRA Model-Human Reliability l b l Analysis: l RAI 3a Response

• Containment spray p y flows onlyy affect sump p flow rate and sump flow rate dependent physical phenomena

- Spray securement modelled as normal distribution with 20 +/- 5 minute mean and STD

• RWST switchover modelled as point values with 2 containment sprays operational (most probable)

• Failed debris washdown taken from deterministic analysis with 2 sprays operational

• Statistical sampling allows for failure of manual action to secure spray p yppump,p, but this failure is not enforced 30

APLAB, STP PRA Model-Human Reliability l b l Analysis:l RAI 3bb

• Please state if CASA Grande models plant conditions that would occur if operators fail to secure all containment spray (CS) long term e.g.

- Sump flow rates

- Washdown rates

- RWST drain-down drain down times 31

APLAB, STP PRA Model-Human Reliability l b l Analysis: l RAI 3a Response

• CASA Grand did not model plant conditions that would occur if operators failed to secure CS long term term.

• CASA Grande samples user-defined Post-LOCA time at which all CS are secure

• User-defined distribution was truncated a high CS securement time i off 7 7.5 5hhours.

• Plant conditions not modelled for failure of long term CS securement 32

APLAB, STP PRA Model-Human Reliability l b l Analysis:l RAI 3c

• If either question to CS securement RAIs RAI s is no provide technical basis, and explain how PRA meets ASME HLRHLR-HR-G HR G requirement to perform an assessment of post initiator human failure events using a well defied and self-consistent process that addresses scenario- specific influence on human scenario performance 33

APLAB, STP PRA Model-Human Reliability l b l Analysis: l RAI 3a Response

• The PRA does include to represent failure to trip one CS pump as well as failure to trip all CS pumps later in sequence

• For the STP CASA Grande evaluation these CS securement operator action failures were not modeled

- Failure F il tto ttrip i one CS pump would ld evenly l di divide id debris to all strainers resulting in lower head-loss

- This may be slightly unconservative for in in-vessel vessel effects 34

ESGB RAIs

• Coatings: RAI 1 1, 2 2, 6

• Chemical Effects: RAI 1 35

ESGB, Coatings: RAI 1

• Provide basis for the unqualified coatings epoxy distribution size.

36

ESGB, Coatings: RAI 1 Response

• Autoclave testing performed by K&L on 1

Comanche Peak supplied coatings chips

• Alion performed characterization of chip type 2

and size

- Masses M off chips hi weighed i h d iin iindividual di id l size i

categories

- The individual weight from each size category was divided by the total of all categories to yield mass fractions 37

ESGB, Coatings: RAI 1 Supporting Information

• Coatings Coat gs samples sa p es were e e tested in two t o baskets, bas ets, oneo e submerged and one unsubmerged

- Unsubmerged basket had higher mass percentage of fines and fine chips after testing

- Chip debris that escaped both baskets was taken from autoclave floor and added to unsubmerged basket for conservatism

• This resulted in a higher mass fraction of fines and fine chips

- Mass fractions from unsubmerged basket were used to define failed size fractions for unqualified epoxy in STP CASA Grande evaluation 38

ESGB, Coatings: RAI 2

• Provide ZOI used for both epoxy and IOZ qualified coatings.

39

ESGB, Coatings: RAI 2 Response

• A 4D ZOI O wasas used for o bot both qua qualified ed zincc aand d IOZ O

coatings.

Supporting Information 1

• WCAP-16568 testingg shows an average g loss of about 0.1 mil at 3.68 L/D where jet directly impacts coupon (generally within 2-inch radius)

• Extrapolation of data shows zero erosion and 4.28 L/D 40

ESGB, Coatings: RAI 6a

• Describe what has been done to ensure that STP unqualified coatings are the same as 1

coating used in EPRI testing 41

ESGB, Coatings: RAI 6a Response

• Manufacture coating type is often unavailable for unqualified coatings.

• Generic coatings types at STP (i.e. epoxy, alkyd, etc.) matched with generic coatings 1

types from the EPRI study

• Generic unqualified coatings sample types 1

labeled in EPRI study were taken from multiple plants with different visual and physical characteristics 42

ESGB, Coatings: RAI 6b

• EPRI testers stated that they made no attempt 1

to quantify debris on the filter. Provide addition justification for using this test data to assign failure time to unqualified coatings.

43

ESGB, Coatings: RAI 6b Response

• Estimated st ated failure a u e timing t g of o unqualified u qua ed coatings coat gs was based off visual inspection of filter discoloration.

• Alkyds had the highest average detachment from testing (33.9%).

- Alkyds tested were visually pigmented with a variety 1 of colors (gray, silver, red, yellow, blue, blue-green)

- Alkyd failure may conservatively bias failure timing to the maximum unqualified coatings failure rate of alkyds 44

ESGB, Coatings: RAI 6c

• STP seeks to reduce the failure of unqualified coatings from the deterministic methodology percentage of 100% to 6%.6% Provide additional justification for the current failure timing analysis.

analysis 45

ESGB, Coatings: RAI 6c Response

• Alkyds have greatest influence on interpretation of filter photographs because of distinctive coloration and highest g average g detachment

• EPRI analysis states With regard to timing of the 1

coating failures, the filters do not demonstrate a definitive time of failure however in subjective terms it appears pp that much of the failure occurred in the 24- to 48- hour timeframe 46

ESGB, Coatings: RAI 6c Response

• If much of the failures is 21% and the total average detachment is 27% for all coatings g types yp 6% failure is found at 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br />.

• Increasing the subjective Failure Percentage Vs. Time amount of much of the 100%

failures to a higher 90%

percentage would yield a 80%

Pe ercent Coatings Failure lower percentage of failed 70%

coatings at 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> than the 60%

50%

STP assumed 6%.

40%

30%

• If subjective bj ti amountt changed h d 20%

to 55% of available failure, the 10%

(24 hr., 6%)

failed percentage at 24-hours 0%

doubles to 12%. Parametric 0 20 40 60 80 100 120 140 160 180 Time (hr) evaluations have shown that risk is insensitive to this range of added particulate. 47

ESGB, Coatings: RAI 6c Response

• Artificially setting 100% failure over the 7 days and assuming much of the failures is 89% and 55% ((red and blue curves respectively) with 5% residual failure (48 to 168 hours0.00194 days <br />0.0467 hours <br />2.777778e-4 weeks <br />6.3924e-5 months <br />).

• Interpreting much of the Failure Percentage Vs. Time 100%

failures as 89% (red curve) 90%

yields the STP assumed 6% 80%

Perccent Coatings Failure f il failure att 24 h hours 70%

60%

• If subjective amount changed 50% (24 hr., 40%)

40%

to 55% (blue curve) of 30%

available failure, the failed 20%

percentage at 24-hours is 40 10%

(24 hr., 6%)

0%

%. This would contradict the 0 20 40 60 80 100 120 140 160 180 observed average detachment Ti Time (h (hr))

of 27%

48

ESGB, Chemical Effects: RAI 1a

• Please jjustifyy not correlatingg the chemical bump p up p factor to the conventional head loss since the same debris bed affects both values 49

ESGB, Chemical Effects: RAI 1a

Response

• Chemical bump-up factor

- Applies maximum head loss from chemical interaction with the highest filtration debris bed

• Based on test data (thin bed not included)

- Conventional head loss is correlated to break size with maximum chemical h i l iinteraction i ffor iincrease off totall h head d lloss

- Distributions provide opportunity for smaller breaks to have high HL

- Simplifies the chemical and bed interaction complexity with conservatism ti

• Supporting calculations

- Technical approach to support and evaluate applied engineering j d judgment

- Allows estimation or detailed quantification of the strength and weakness of bump up approach

- Assesses A weakness k off the th bump b up d do nott underestimate d ti t risk ik because of conservatisms used in generation of bump up 50

ESGB, Chemical Effects: RAI 1b

• In order to help the staff judge the magnitude of the chemical head loss bump-up factor, please provide, by performing realizations for the existing CASA Grande model, a relative frequency plot of chemical effects for STP in terms of absolute units (e.g., feet of H2O) for the SBLOCA [small

[ ll break b k LOCA], ] MBLOCA [medium

[ di break b k LOCA], ] andd LBLOCA [large break LOCA].

51

ESGB, Chemical Effects: RAI 1b 1.0E+00 LBLOCA

Response

1.0E-01 1.0E-02 1.0E-03 Frequency (%)

1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 Head Loss (Feet of Water) 1.0E+00 1.0E-01 MBLOCA 1.0E-02 1.0E-03 1.0E-04 Frequency ((%)

1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 Head Loss (Feet of Water) 52

ESGB, Chemical Effects: RAI 1c

• Please provide additional details on how the results from the Chemical Head Loss Experiment (CHLE) testing, WCAP-16530-NP, Evaluation of Post-Accident Chemical Effects in Containment Sump Fluids to Support GSI-191, calculations, and reasonable engineering judgment were used in the h ddevelopment l off the h exponentiall PDF. In addition, dd please l supply l the h

basis for choosing the exponential form of the PDF over others, e.g.,

Weibull 53

ESGB, Chemical Effects: RAI 1c

Response

• The exponential PDF was chosen:

- Shape

- Convenience of fitting statistics of the mean and truncated tail probability

• Only mean is needed to fully specify distribution

• CHL testing and WCAP WCAP-16530-NP 16530 NP calculations were used to determine the bump-up multiplier that exist at the highest probability of the PDF

• Engineering judgment based on review of STP strainer test was used to determine a mean bump up multiplier

• The maximum values were values capable of producing a quantifiable number of chemically induced failures

- 6.8X (SBLOCA), 8.1X (MBLOCA), and 10.7X (LBLOCA) higher than the DBA multiplicative response.

54

ESGB, Chemical Effects: RAI 1d

• Please provide a detailed technical basis for the mean bump up factors shown for the SBLOCA, MBLOCA, and LBLOCA. The NRC staff has observed head loss testing where the greatest chemical bump-up factors are associated with thinner beds. Please discuss why the mean bump up f

factor would ld b be h higher h ffor a LBLOCA. Please l explain l iff it is more probable b bl that a debris bed for smaller and medium breaks (assuming the bed coverage criterion is met) would consist primarily of fiber fines that are the most readily transportable to the strainer strainer. In general general, finer fiber beds tend to lead to greater head loss 55

ESGB, Chemical Effects: RAI 1d

Response

• Mean bump-up

- Based on STP strainer testing of DBA bed and 30-day worst-case chemical precipitate mode

- Conservatism applied

• SBLOCA applies multiplier based on strainer tests

• MBLOCA and LBLOCA increases the multiplier

• Bump up mean multipliers applied to thinner beds were confirmed higher than a DBA determined multiplier. However, sensitivity modeling d l required d the h thin h b bedd mean to b be muchh llarger than h that h

observed in other strainer testing

• It is not expected that small and medium break debris consists primarily of fines because

- Every break generates the same volumetric proportions of fiber glass

- One set of debris-size dependent transport fractions is applied for everyy break 56

APLAB HRA RAI 3

• Top Event OSI represents securing one train of Containment Spray p y earlyy given g three initiate;; Top p Event OFFS represents securing all Containment Spray later as directed by technical support center (a) Does CASA Grande reflect conditions if all three Containment Spray continue to run?

Statistical sampling of trip time does not preclude long term operation of three trains; however, this condition was not explicitly considered 57

APLAB HRA RAI 3, continued

( ) Does CASA Grande reflect conditions if (b)

Containment Spray is not secured in the long term?

No. While the time of tripping the pumps is sampled, truncation considerations preclude this time to go beyond 77.5 5 hours5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br /> hours. Plant conditions are not modeled for the condition where operator fail to secure long-term containment spray.

58

APLAB HRA RAI 3, continued (c) Provide justification if the answer to (a) or (b) is no.

The PRA contains the necessary logic structure to represent the conditions of interest. However, no specific CASA Grande results for these conditions were available at the time of model quantification quantification. While failure to trip one pump might lead to conservative results with respect to potential NSPH conditions due to sump screen loading (debris would be transported to two sump screens rather than three), the approach may be unconservative with respect to the filtration of debris by the sump strainers.

59

APLAB HRA RAI 5

• How o were e e Casa G Grande a de results esu ts de developed e oped to reflect combinations of success and failure of the operator actions: (1) securing one train of Containment Spray early l given three h trains initiate; (2) securing all trains of Containment Spray later in the event response; (3) switchover to sump recirculation; and (4) switchover to hot legg injection.

j How was consitencyy between the PRA scenario and information developed in CASA Grande assured?

60

APLAB HRA RAI 5, continued

• The first two actions are included in the PRA model as switches; they were included to support sensitivity analyses, if later desired

• The success or failure of the first two tasks does not impact the success or failure of the remaining two actions.

• Action 3 is required to avoid fuel damage; action 4 is only queried if action 3 was successful

• Failure of action 4 is assumed to result in fuel damage for cold leg breaks; conversely success of action 4 precludes the p possibilityy of boron p precipitation p ((after the time of switchover to hot leg injection), minimizing need for status of actions to be communicated between the PRA and CASA Grande 61

APLAB RAI 2: Arithmetic vs.

Geometric Means

• RAI Statement:

• Volume 3, Assumption 3.a (page 76 of 248) states that the geometric-mean aggregation of LOCA frequencies in NUREG-1829 is the most appropriate set of results to use for this evaluation evaluation. The basis provided is that geometric geometric-mean mean aggregation produces frequency estimates that are approximately the same as the median estimates of the panelists. There is no justification about why the median estimate is preferred and emphasis on the median conflicts with the RG 1.174 guidance that the mean values be used for decision making. Furthermore, i f information ti iin NUREG NUREG-1829, 1829 SSection ti 77.6.4 6 4 shows h th thatt th the use off th the arithmetic ith ti mean instead of the geometric mean would increase the LOCA frequency by an order of magnitude or more for some LOCA categories and may therefore substantially increase the risk estimates. Consequently, selection of the geometric mean is a key y assumption p and selection of the arithmetic mean represents p an alternative reasonable assumption as defined by RG 1.200. This is supported by RG 1.174, Section 2.5, which states that the licensee should [identify] key assumptions in the PRA that impact the application. Sensitivity studies provide important information about how some of the key assumptions affect the final results as discussed in RG 1 1.174 174 Section 2 2.5.3.

5 3 Please provide CDF CDF, LERF LERF, CDF, and LERF using the arithmetic mean aggregation of LOCA frequencies in NUREG-1829.

62

APLAB RAI 2: Arithmetic vs.

Geometric Means

• RAI Response:

• Despite the above caveat Weight Required on Expert A in NUREG-1829 contains in its th W the Weighted i ht d GM tot Achieve A hi Executive Summary, the body of the AM:

NUREG-1829, along with the literature on combining expert 5th 50th 95th opinion, makes a strong case for using GM rather than AM at 73.6% 72.4% 70.1%

least when:

• (i) the elicited probabilities concern rare-events,

• (ii) the opinions of the individual experts are disparate,

• and (iii) we seek a combination rule that represents a reasonable notion of the center of the group's opinion.

63

APLAB RAI 3A, 3B, 3C: Plant Configuration

• RAI Statement:

• Volume 3, Assumption 2b provides a qualitative argument for why a combination of pumps failing in the same train is worse" worse than the same set of pumps failing in different trains.

A) Please justify this assumption and clarify if an engineering y

analysis was p performed in support pp of this assumption.p B) Please state if this assumption always increases the conditional probability of strainer failure (i.e., is this a conservative assumption?).

C) Please state if this assumption always increases the conditional probability of in-vessel effects.

64

APLAB RAI 3A, 3B, 3C: Plant Configuration

• RAI Response:

p

• Analysis of Case 22 suggests the condition in which all pumps fail on the same train leads to the largest sump failure frequency.

• For vessel failure frequency, three cases (Cases 22-2, 22-3, and 22-5) in which the HHSI and LHSI pumps fail on different trains result in larger frequencies than when these pumps fail on the same train.

• The change in CDF reported in Volume 2 is 2.88E-08 per year.

• The CDF obtained by replacing Case 22-1 with Case 22-3 is:

2.88E-08 + [3.24E-02 x 1.17E-08] = 2.92E-08 per year, a 1.5% increase.

65

NUREG-1829 Related RAI RAIss

APLAB RAI 2: Modeling LOCA Frequency

& Break Size Under DEGB-Only Breaks

• RAI Statement:

• NUREG-1829 states that, in general, a complete rupture of a pipe is more likely than a partial rupture rupture. It appears appears, however however, that STP's methodology leads to the opposite result (i.e., a rupture of a given size is more likely to be caused by a partial rupture of a large pipe p

than a complete rupture p of a smaller ppipe).

p )

A) Please illustrate the results of your method by comparing the frequency of partial versus complete breaks for a set of representative pipe sizes.

B) Please describe whether the methodology described in the STP pilot is consistent with the assumption of NUREG-1829 or provide justification for an alternate approach.

67

APLAB RAI 2: Modeling LOCA Frequency

& Break Size Under DEGB-Only Breaks

• RAI Response:

• Under the implemented continuum" model with a hybrid approach, the probability a pipe experiences a DEGB given that it has a break is 0.165.

• Under the top-down only approach for the continuum" model, the probability a pipe experiences a DEGB given that it has a break is 0.0746.

• Using the NUREG-1829 frequencies without considering RI-ISI frequency information (no hybrid approach) increases the probability of a DEGB by more than a factor of 2.21.

68

APLAB RAI 3: Justification of the Use of 25-Year Frequency Estimates

• RAI Statement:

• Section 2.5.5 states that it is incumbent on the licensee to demonstrate that the choice of reasonable alternative hypotheses, hypotheses adjustment factors, or modeling approximations or methods to those adopted in the PRA model would not significantly change the assessment. Also,, it is assumed that the STP p plants will continue to operate for more than 25 years; RG 1.174 Section 3 states that the licensee should define an implementation and monitoring program to ensure that no unexpected adverse safety degradation occurs do to the change change.

A) Please justify the use of the 25-year frequency estimates rather than the 40-year estimates provided by NUREG-1829. Provide CDF, LERF,, CDF,, and LERF using g the 40-year y estimates.

69

APLAB RAI 3: Justification of the Use of 25-Year Frequency Estimates

• RAI Response:

• Using 40-year frequency estimates, CDF increases to 6.91E-08, a factor of about 24 2.4.

70

APLAB RAI 4A: Discrepancy Between LOCA Frequency Means for Johnson Distributions &

NUREG-1829

• RAI Statement:

• Typically, statistical sampling simulations will develop random variables that preserve the mean of the distribution from which the variables are sampled. STP has chosen to fit a Johnson bounded distribution that matches the expert-provided 5th, 50th, and 95th percentiles in NUREG-1829,, but does not match the mean values.

p The properties of the distribution are such that, as fit, the mean of the fitted distribution is always less than the experts' means from the distributions in NUREG-1829.

A) Pl Please explain l i why h th the STP evaluation l ti departs d t from f the th regulatory position in RG 1.174 regarding the use of mean values.

71

APLAB RAI 4A: Discrepancy Between LOCA Frequency Means for Johnson Distributions &

NUREG-1829

• RAI Response:

• Analyze means of Johnson distributions as a function of distributions,

• x-axis: Ratio of to elicited 95th percentile

• Means grow with , but level off for large values of

• Even at a ratio of 100, the Johnson means fall short of the NUREG-1829 means, most prominently in categories 5 and 6 72

APLAB RAI 4B: Explanation of the Johnson Di t ib ti Governing Distribution G i LOCA FrequencyF

• RAI Statement:

• Typically, statistical sampling simulations will develop random variables that preserve the mean of the distribution from which the variables are sampled.

l d STP h has chosen h tto fit a Johnson J h b bounded d d di distribution t ib ti ththatt matches the expert-provided 5th, 50th, and 95th percentiles in NUREG-1829, but does not match the mean values. The properties of the distribution are such that, as fit, the mean of the fitted distribution is always less than the experts experts' means from the distributions in NUREG-NUREG 1829.

B) The Johnson fit to 5th, 50th, and 95th percentiles is not unique.

Alternative accurate fits can be constructed with arbitrary values of the scale parameter . The scale parameter defines a bound on the maximal frequencies sampled in the Monte Carlo model. By increasing the value of , the relative proportion of large to medium to small breaks can be altered, especially in the extrapolation range beyond the 95th percentile.

percentile Please provide a technical justification for the selection of the scale parameter (other selections appear possible that could change the outputs by CASA Grande).

73

APLAB RAI 4B: Explanation of the Johnson Di t ib ti Governing Distribution G i LOCA Frequency F

• RAI Response:

p

• Using the means of the Johnson distributions as metric for relative contribution of LOCA events by category:

• Somewhat underestimate the contribution categories 5 and 6 as compared to using the implied means from NUREG-1829 NUREG 1829,

• Overestimate the contribution of these categories versus the implied NUREG-1829 means if we use the 99th or 99.9th percentiles and a large value for .

• Fixing to twice the 95th percentile: Underestimate the contribution to categories 5 and 6, even using 99th or 99.0th percentiles relative to NUREG percentiles, NUREG-1829 means.

74

APLAB RAI 4C: Impact of Johnson Distribution S l ti on CDF, Selection CDF LERF LERF, CDF, CDF LERF

• RAI Statement:

• Typically, statistical sampling simulations will develop random variables that preserve the mean of the distribution from which the variables are sampled.

l d STP h has chosen h tto fit a Johnson J h b bounded d d di distribution t ib ti ththatt matches the expert-provided 5th, 50th, and 95th percentiles in NUREG-1829, but does not match the mean values. The properties of the distribution are such that, as fit, the mean of the fitted distribution is always less than the experts experts' means from the distributions in NUREG-NUREG 1829.

C) Please provide the maximum expected difference between the CDF, LERF, CDF, and LERF developed from bounded Johnson distributions that consider alternative values of the scale parameter ,

and other distributions that would preserve mean values reported in NUREG-1829. Note, in particular, that alternative bounded Johnson distributions with large values of the scale parameter can be built to accurately fit the NUREG-1829 5th, 5th 50th 50th, and 95th percentiles, percentiles and produce mean estimates closer to the NUREG-1829 values than current fits used by STP.

75

APLAB RAI 4C: Impact of Johnson Distribution S l ti on CDF, Selection CDF LERFLERF, CDF, CDF LERF

• RAI Response:

p

• Our estimates of changes in CDF, CDF, LERF, and LERF are modest as we range the Johnson scale parameter from a factor of 1.25 up to a f t off 100 times factor ti the th 95th percentile til off th the frequencies f i elicited li it d from f

experts in NUREG-1829. More specifically, point estimates of CDF and LERF increase by no more than 2% for the specific values of we consider here.

76

APLAB RAI 4C: Impact of Johnson Distribution S l ti on CDF, Selection CDF LERF LERF, CDF, CDF LERF

• RAI Response:

77

EPNB RAI 6: Consistency of Weld Frequencies with ith RI RI-ISI ISI Program P

• RAI Statement:

• By letter dated September 10, 2012, the NRC approved the risk-p informed in-service inspection ((RI-ISI)) p program g for the third 10-year y

in-service inspection interval at STP, Units 1 and 2 (ADAMS Accession No. ML12243A343). Please discuss the following:

A) Please state if the LOCA frequency estimates used for welds in the GSI-191 GSI 191 submittal are consistent with the LOCA frequency estimates used in the RI-ISI program. If the comparison is appropriate, please provide numerical examples of the comparison.

If the comparison is not appropriate, please provide explanation.

B) If the LOCA frequencies for welds are not consistent between the two analyses, (1) please identify the differences and explain why there are differences, and (2) please discuss why the LOCA frequencies proposed in the GSI-191 submittal are acceptable if they are not consistent with that of the RIRI-ISI ISI program program.

78

EPNB RAI 6: Consistency of Weld Frequencies withith RIRI-ISI ISI Program P

• RAI Response:

p

• The mean frequencies reported in NUREG-1829 are preserved in the mean frequencies used to compute CDF. Using RI-ISI instead, these th three ffrequencies i would ld d decrease b by ffactors t off 4 4.43, 43 15 15.18, 18 and d22.27 27 ffor small, medium, and large breaks.

• If we use only RI-ISI frequencies to construct the distribution over break size and weld case, the probability mass increases by factors of 5.03, 20 38 and 20.38, d 15 15.68 68 iin th the respective ti categories t i 4 4, 55, and d66.

79