ML20212P934

Technical Evaluation of EPZ Sensitivity Study for Seabrook, Technical Rept

ML20212P934
Person / Time
Site:	Seabrook
Issue date:	03/03/1987
From:	Hofmayer C, Pratt W BROOKHAVEN NATIONAL LABORATORY
To:	NRC
Shared Package
ML20212P921	List: ML20212P918 ML20212P934
References
CON-FIN-A-3852 NUDOCS 8703160363
Download: ML20212P934 (165)

Text

__ ____ -

x 3

TECHNICAL REPORT A-3852 3-3-87 TECHNICAL EVALUATION OF THE EPZ SENSITIVITY STUDY FOR SEABROOK SAFETY AND RISK EVAltlATION DIVISION, STRUCTURAL ANALYSIS DIVISION, AND ENGINEERING TECHNOLOGY DIVISION DEPARTMENT OF NUCLEAR ENERGY, BROOKHAVEN NATIONAL LABORATORY UPTON, NEW YO X 11973

[}w 7

Prepared for the U.S. Nuclear Regulatory Commission

!W s{-) b(7,.

[g j

f,

Office of Nuclear Reactor Regulation d

Contract No. DE-ACO2-76CH00016 o

ADO K O O

D

8=.

3; g.7, s

., +

,>+

' L, ;....b '-

.:k_

~-

s lV

.t '

f r;.

I b

L m.

NOTICE This report was prepa red as an account of work sponsored by an a gency of the United States Government. Neither the United States Government nor any agency thereof.or any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for any third party's use, or the results of such use. cf any information, apparatus, product or' process disclosed in this report, or represents that its use by such third party would not infringe privately owned rights.

The views expressed in this report are not necessarily those of the U.S. Nuclear f.

Regulatory Commission.,

• L

1.J TECHNICAL EVALUATION OF THE EPZ SEllSITIVITY STUDY FOR SEABROOK W. T. Pratt and C. Hofmayer Principal Investigators 1

Department of Nuclear Energy Brookhaven National Laboratory Upton, New York 11973 March 1987 l

l l

s Prepared for U.S. Nuclear Regulatory Commission r

Washington, DC 20555 Under Contract No. DE-AC02-76CH00016 FIN A-3852 l

l l

,3 LIST OF CONTRIBUTORS Contributor Affiliation K. Bandyopadhyay Structures & Components Evaluation Group / SAD P. Bezier Civil & Structural Mechanics Group / SAD G. Bozoki Risk Evaluation Group /SRED T-L. Chu Risk Evaluation Group /SRED M. Chun Accident Analysis Group /SRED C. Hofmayer Structures & Components Evaluation Group / SAD M. Khatib-Rahbar Accident Analysis Group /SRED B. Luckas Engineering Analysis & Human Factors Group /ETD J. Pires Structures & Components Evaluation Group / SAD W. T. Pratt SRED A. Tingle Accident Analysis Group /SRED R. Youngblood Facilities Risk Analysis Group P. C. Wang Civil & Structural Mechanics Group / SAD SRED = Safety and Risk Evaluation Division ETD = Engineering Technology Division t

SAD = Structural Analysis Division iii

J' ABSTRACT A technical evaluation of the Seabrook Station Emergency Planning Sensi-tivity Study (PLG-0465) and supporting documentation has been performed. This was an evaluation which focused on those areas found to be the most influen-tial in calculating the Seabrook risk estimates. The approach taken by Brook-haven National Laboratory (BNL) was tc perform sensitivity studies to assess the impact on the results in PLG-0465 of the BNL evaluation of these areas.

s I

I l

y v

i l

1

- TABLE OF CONTENTS Page Li s t of Cont ri bu t o rs.................................................... ' 111 Abstract................................................................ v Li s t o f Fi gu re s................'......................................... i x List of Tables.......................................................... xi Preface.................................................................

xiii Acknowledgements........................................................ xv S uma ry................................................................. x v i i L

1.

INTRODUCTION.......................................................

1-1 1.1 Background......................;..............................

1-1 1.2 Scope and Focus of Revi ew......................................- 1-4

1. 3 Ap p ro ach to Re vi ew............................................

1 1.4. Organization of the Report....................................

1-7 1.5 Re f e r e n c e s....................................................

1 - 8 2.

SYSTEM EVALUATION..................................................

2-1 2.1. Int e r f a ci n g Sy s t em L0C'A...................................'..... 2-2 2.1.1 Ge n e r a l................................................ 2 - 2 2.1.2 Ot h e r I S L Pa t h s........................................ 2 - 3 2.1.3 ISL Ini ti.at or Frequencies...............~........'....... 2-4 2.1.3.1 Check valve failure frequencies...............

2-5 2.1.3.2 _ Cold. leg safety. injection path frequency......

2-8 2.1.3.3 RHR suction side f requency.............'....'... 2-9 2.1.4 Ope ra t o r Ac t i on s......................... '.....'......... 2 - 11 2.1.5 Break Loca. tion.........................................

2-15 2.1.6 Event Tree Quanti fi cati on.............................. 2-16 2.2. Accidents During Shutdown and Refueling Conditions.....'.......

2-17 l

2.2.1 Loss of Decay Heat Removal During Shutdown or Refueling. 2-20 2.2.2 Low Temperature 0verpressuri zation..................... 2-23

(.

2.2.3 Loss of. Coolant Accidents During Shutdown or Refueling. 2-24 2.2.4 PSNH Comments on BNL Dra f t Report...................... 2-25 2.2.5 Summary of the Shutdown Ri sk Revi ew.................... 2-26 2.3 Induced Steam Generator Tube Rupture (SGTR)................... 2-27 i

2.4 Ccntainment Isol a tion Fai l ure................................. 2-28 l

2.5 Su mma ry....................................................... 2 - 2 9 l

2.6 Re fe re nc e s.................................................... 2 - 3 3 3.

EVALUATION OF CONTAINMENT BEHAVI0R.................................

3-1 3.1 Ca paci ty at General Yi el d..................................... 3-1 3.2 Behavi or at Large Deformati on................................. 3-5 3.3 Capability of Penetrations....................................

3-9 l

3.4 Sunma ry of St ruct u ra l Fi nd i ngs................................

3-15 3.5 References....................................................

3-16 1

vii

L TABLE OF CONTENTS (Continued)

Page 4

CONTAINMENT EVENT TREE REVIEW......................................

4-1 4.1 Potential Containment Loads...................................

4-2 4.2 Application.to Seabrook.......................................

4-3 4.3 Su mma ry....................................................... 4 - 5 4.4 References....................................................

4-6 5.

R E V I EW OF SOURC E T ERMS.............................................

5 - 1 5.1 Fidelity to WASH-1400 Methodology.............................

5-1 5.2 Credi t for Scrubbing of Submerged Rel eases....................

5-1 5.3 Un c e r t a i n t i es................................................. 5 -2 5.4 Su mma ry.......................................................

5 - 3 5.5 Re f e r enc e s.................................................... 5 - 3 6.

S ITE CONS EQUE NC E M0 DEL............................................. 6 -1 6.1 NUR E G -0 396 8a s i s.............................................. 6 - 1 6.2 Cons equence Model i n g..........................................

6-2 l

6.2.1 Whole Body Dose Vs Di stance............................ 6-3 6.2.2 Thyroid Dose Vs Di stance...............................

6-4 6.2.3 Ri sk of Ea rly Fatali ti es............................... 6-4 6.3 Compa ri sons of Resu l t s........................................ 6-5 6.3.1 Resul ts of Seabrook Study.............................. 6-5 6.4 Sens i t i vi ty St ud i es........................................... 6-6 6.4.1 Sensitivity of Results to Multi Summa ry........................pu f f Rel ea se............ 6-7 6.4.2 6-7 6.4.2.1 Interfacing systems L0CA......................

6-7 6.4.2.2 Accidents during shutdown.....................

6-8 6.4.2.3 Induced steam generator tube rupture..........

6-10 6.4.2.4 Containment isolation failure and pre-l existing leakage..............................

6-11 6.4.2.5 Containment structural capar.ity...............

6-12 6.4.2.6 Containment loads.............................

6-13 i

6.4.2.7 Sou rc e t e rms..................................

6-15 6.4.2.8 Conse Re f e ren c es...........qu en ce Mod el..............................' 6-16 6.5 6-17 s

i I

i I

s f

i t

{

viii

J LIST OF FIGURES Figure Page 5.1 Comparison of Seabrook Station sensitivity results using WASH-1400 source tem methodology with background, safety goal i ndi vi dual and RMEPS ri sk l evel s................................ xxx S.2.

_ Comparison of median risk of early fatalities at Seabrook Station for different emergency planning' options................

xxxi S.3 Comparison of Seabrook Station results in this study and RMEPS with NUREG-0396 - 200-rem and 50-rem whole body dose plots for no immedi ate protecti ve act10ns................................. xxxi i S.4 200 rem dose versus distance curves for various failure modes assuming no immediate protective action......................... xxxiii 5.5 Comparison of 200 rem-dose versus distance curves for conservative assumption of no credit for operator recovery of open equ i pmen t ' h a t c h.........................'................... x x x i v 5.6 Comparison of BNL sensitivity studies with' PLG-0465 and NUREG-0396...................................................... xxxy S.7 Comparison of 200 rem-dose versus distance curves for conservative interpretation by PSNH of NUREG/CR-4220 data.......

xxxvi 1.1 Comparison of Seabrook Station sensitivity results using WASH-1400 source term methodology with background, safety goal i ndi vi dual and RMEPS ri sk 1 evel s...............................

1-9 1.2 Comparison of median risk of early fatalities at Seabrook Station for different emergency planning options...............

1-10 1.3 Comparison of Seabrook Station results in this study and RMEPS with NUREG-0396 - 200-rem and 50-rem whole body dose plots for no immediate protective actions................................

1-11 1.4 200-rem dose versus distance curves for various failure modes assuming no immediate protective acti0n........................

1-12 2.1 Frequency of accumulator check valve leakage events............

2-36 2.2 Comparison of 200 rem-dose versus distance curves for conservative assumption of no credit for operator recovery of open equipment hatch........................................

2-37 2.3 Comparison of BNL sensitivity studies with PLG-0465 and NUREG-0396.....................................................

2-38 2.4 Comparison of 200 rem-dose distance curves for conservative interpretation by PSNH of NUREG/CR-4220 data...................

2-39 3.1 Containment building cross-section.............................

3-17 l

3.2 Cylinder reinforcement.........................................

3-18 3.3 Containment finite element model (NFAP)........................

3-19 3.4 Pressure-radial displacement relation for containment..........

3-22 6.1 Components of NUREG-0396 curve as computed by BNL using CRAC2.. 6-19 i

6.2 Risk of death or exceeding dose levels for $1W as calculated by BNL.........................................................

6-20 i

6.3 Risk of death or exceeding dose levels for S2W as calculated i

by BNL.........................................................

6-21 6.4 Risk of death or exceeding dose levels for S6W as calculate

{

by BNL.........................................................

6-22 i

i l

[

]

lx

( --.

t l

LIST OF FIGURES (Continued)

Figure Page 6.5 Dose versus distance curve for release category S1W from Seabrook for no immediate protective action with BNL results u si ng MACCS s uperimposed....................................... 6-23 6.6 Dose versus distance curve for release category"S2W from Seabrook for no immediate protective action with BNL results us i ng MACCS su pe ri mpos ed....................................... 6-24 6.7 Dose versus distance curve for release category S6W from Seabrook for no immediate protective action with BNL results u si n g MAC CS s u pe ri mp os ed....................................... 6-2 5 6.8 Compari son of MACCS to CRAC2 codes............................. 6-26 6.9 Comparison of 200 rem-dose versus distance curves for conservative assumption of no credit for operator recovery of open equipment hatch........................................

6-27 6.10 Comparison of BNL sensitivity studies with PLG-0465 and NUREG-0396...........................................................

6-28 6.11 Comparison of 200 rem-dose versus distance curves for conservative interpretation by PNSH of NUREG/CR-4220 data......

6-29 I

i l.

i X

y LIST OF TABLES Table Page S.1 Sumary of Release Category Frequency Uncertainty Distributions. xxxvif S.2 Early Fatalities Conditional on a Release Occurring in the Population Around the Seabrook Station Site Boundary...........

xxxviii 1.1 Sumary of Release Category Frequency Uncertainty Distributions.1-13 1.2 Risk of Early Fatalities in the Poaolation Around the Seabrook St at i on Si te Bou nda ry..........................................

1 -14 2.1 Sumary of Operating Events, Emerge. Ov Core Cooling System, Isolation check Valves, Leakage FaiiJre Mode...................

2-40 2.2 Sumary of Operating Events, Emergeacy Core Cooling System, Isolation Check Valves, " Failure.o Close U Failure Mode...............................pon Demand"

.................... 2-43 2.3 Accumulator Check Valve Exposure Data..........................

2-44 2.4 Statistical Data on Leakage Events of Check Valves to AC C umu l a t o r s................................................... 2 -4 5 2.5 ISL Results Initially Assigned Plant Dama Pl ant Operati ng Modes.................... ge St ates............. 2-46 2.6

...................... 2-47 2.7 Categories of 130 Reported Total DHR System Failures When Required to Operate (Loss of Function) at U.S. PWRs 1976-1983.. 2-48 3.1_

Statistics of Rebar Yield Strength for Various Sizes...........

3-21 3.2 Reinforcement Details of the' Containment Cylinder..............

3-22 3.3 Reinforcement Details of the Containment Dome..................

3-23 3.4 Statistics of Concrete Compressive Stren Conc rete Properti es.....................gth....................

3-24 3.5 3-25 3.6 Characterization of Containment Penetrations................... 3-26 4.1 Corparison of Core Melt Frequencies and Distribution of Release Types..........................................................

4-7 5.1 Release Categories for Seabrook Station Based on WASH-1400 Sou rce Te rm Met hod ol o gy........................................ 5 -5 5.2 Revised C-Matrix for New Source Term Categories................ 5-6 6.1 Sumary of Release Categories Representin Accidents (from the RSS).................g Hypothetical 6-30 xi

3 PREFACE This report describes a technical evaluation of the Seabrook Station Emergency Planning Sensitivity Study (PLG-0465) and the Seabrook Station Risk Management and Emergency Planning Study (RMEPS) (PLG-0432).

The main objec-j tive of this technical evaluation is to assist the NRC in its evaluation of the validity of the conclusions presented in PLG-0465.

This is therefore a focused review by Brookhaven National Laboratory (BNL) of those areas identi-i fled in PLG-0465 as being the most influential in calculating the Seabrook

~

risk estimates.

However, regardless of the conclusions of this focused review, BNL cannot attest to the validity of the overall risk profiles pre-sented in PLG-0465. This follows from the observation that the risk estimates in PLG-0465 rely heavily on RMEPS, which in turn relies on the Seabrook Sta-tion Probabilistic Safety Assessment (SSPSA).

Unfortunately, the risk pro-files in the SSPSA have not been independently reassessed, requantified, and validated, by the NRC staff or their contractors. Similarly, within the scope of the review, BNL has also not validated the accident sequence probability estimates in the SSPSA.

Therefore, because these estimates form the founda-i tion for the updated risk estimates in the RMEPS and ultimately in PLG-0465, BNL has not verified the total risk estimates in PLG-0465.

This includes the predicted dose versus distance curves.

The current review should therefore be

^

regarded as an evaluation of selected issues related to the potential for a i

large early release of radioactivity at the Seabrook Station.

It is not a reassessment or validation of the total risk profile.

l 1

d 9

xiii

(

L -

=

3 ACKNOWLEDGEMENTS The authors wish to thank Dr. R. Bari and Dr. M. Reich in the Department of Nuclear Energy at Brookhaven National Laboratory for many discussions, com-ments, and suggestions related to this program.

This work was performed for the Division of PWR Licensing - A, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission.

The authors wish to acknowledge G. Bagchi, S. Long, W. Lyon, and S. Newberry for their support and guidance throughout the course of this program.

Lastly, the authors acknowledge the efforts of C. Conrad, A. Costini, E. Gilbert, and D. Votruba in preparing this document for publication.

i i

l A

XV

- - -. - - - - - ~ -. - - -

j; j

t

SUMMARY

t 1.

Purpose i

This < report ' describ'es. a technical evaluation by Brookhaven ' National' Laboratory '(BNL) of the Seabrook Station Emergency Planning Sensitivity Study i

(PLG-0465), which was prepared for the Public Service Company of New Hampshire (PSNH) by.Pickard, Lowe-and Garrick (PLG) Inc. PLG-0465 reviewed the bases'in f

iNUREG-0396 for the current.10 mile emergency planning zone (EPZ) and argued, by - taking.into account Seabrook-specific plant features and improvements in methodology..that a smaller EPZ was justified at Seabrook.

The ' results in PLG-0465 rely heavily on two earlier studies, namely the Seabrook Station Risk Management and. Emergency Planning Study (RMEPS) (PLG-0432) and the 'Seabrook Station;Probabilistic Safety Assessment (SSPSA). PLG-0432 and RMEPS were also prepared by PLG..for.PSNH.

The SSPSA was an extensive evaluation of the risk associated with opera-tion of the Seabrook Nuclear Power Station. The SSPSA investigated the conse-quences of accidents that might occur from initiating - events that could be internal to the plant and also external (e.g., seismic events).

The Nuclear Regulatory Commission (NRC) and its supporting contractors initiated an' ~

l in-depth review of the SSPSA but this effort was terminated prior to requan-l tifying the risk estimates.

RMEPS was a sensitivity. study which evaluated emergency planning options for the Seabrook Station. RMEPS focused on those areas that were found in the SSPSA to be the leading contributors to risk.

RMEPS rebaselined the SSPSA analysis of these. risk-important areas specifically to establish an updated assessment. of, the risk at Seabrook so that alternative emergency ' planning options could be developed.

Therefore, RMEPS, focused on: new-data 'and engineering insights about the initiation and progression of sequences involv-ing interfacing systems loss-of-coolant accidents (LOCAs) and on. the results of experimenta1' and ' analytical research that provide an enhanced basis for assessment of radioactive material release (i.e., source terms) for a wider spectrum of accident sequences.

Thus, RMEPS represents the applicant's best l

estimate of risk at Seabrook. RMEPS is being reviewed in conjunction with the j

review of PLG-0465.

t i

The sensitivity studies in PLG-0465 were performed to determine the' i

extent to which the conclusions of RMEPS were dependent on any new source term i

technology.

Thus, PLG-0465 only changed the source terms in RMEPS to be consistent with WASH-1400 source term methodology.

Therefore, in order for BNL to evaluate the results and conclusions in PLG-0465 we also had to evalu-l ate the results in the RMEPS and the SSPSA.

It will be shown later that the l

current BNL evaluation focused on the results in PLG-0465 and the RMEPS and did not attempt to review in detail the results of the SSPSA.

This in turn

[

led to the need to qualify the conclusions of the BNL review of PLG-0465.

I i

The principal conclusion of PLG-0465 was that an EPZ at the Seabrook Sta-tion of 1 mile radius or less is more justified in terms of its risk manage-ment effectiveness than the current 10-mile EPZ was justified by the results of NUREG-0396.

This conclusion was based on the results of the PLG-0465 Sensitivity Study, which are reproduced in Figures S.1-S.3.

These results f

xvii I

f

i o

t were constructed without accounting for any new insights about source terms since WASH-1400.

The conclusion was based on the following observations:

The individual risk of early fatalities in the population within 1 mile of the site boundary with no immediate protective actions is less than the NRC safety goal (refer to Figure S.1).

This individual risk is substantially less when a 1-mile evacuation is assumed.

The risk of early fatalities with a-1-mile evacuation is comparable to i '

the WASH-1400 results, which assumed a 25-m11e evacuation (refer to Figure S.2).

The Seabrook Station results for a 2-mile evacuation' are substantially less than those for WASH-1400.

The risk of radiological exposures for 1, 5, 50, and 200-rem whole body doses with no immediate protective actions is less at 1 mile than the corresponding NUREG-0396 results at 10 miles (refer to Figure l

S.3).

The above observations led to the statement in PLG-0465 that "there is no l

significant frequency of exceeding 200 rem beyond 1.5 miles in the Seabrook sensitivity results."

the most influential in calculating the Seabrook risk estimates:PLG-0465 ide i

i The effectiveness of the Seabrook Station primary containment to i

either remain intact or to maintain its fission product retention capability for periods much longer than required for even delayed, ad hoc protective actions.

1 A more realistic assessment of the strength and failure modes of the Seabrook containment than was possible within the state-of-the-art of j

PRA when the RSS was completed.

A more realistic treatment of the initiation and progression of inter-facing systems LOCA sequences.

Note that of the three areas identified above as being the most influen-l tial to the risk estimates in PLG-0465 the first relates to design features that are specific to Seabrook, namely the rtructural capacity of the Seabrook containment.

The other two areas refer to improvements that have been made in our ability to perform risk assessments, whir other nuclear power plant risk assessments.

therefore have application to 2.

Rational for Review At the request of the NRC, the BNL technical evaluation initially focused on the following areas in PLG-0465 and RMEPS:

- Interfacing systems LOCAs

- Containment function:

- Isolation failure

- Pre-existing leakage

- Structural capacity

- Containment loads

- Seabrook-specific WASH-1400 source t'erms

- Site consequence model.

xviii

I However, during the review process additional areas that were outside of the scope of the original BNL review were identified as potentially important to risk at Seabrook.

Two of these areas were considered sufficiently impor-4 tant to request the applicant to provide additional -information on the risk associated with such events. The two areas identified were:

i

- accidents during shutdown

- potential for induced steam generator tube rupture.

4 These areas 'were not initially included in the original BNL review l

because in. the past they were not found to be dominant risk contributors.

However, as the risk estimates in PLG-0465 and the RMEPS are relatively low,

.I events that were previously considered to be unimportant now have the poten-tial to influence the Seabrook risk estimates.

Thus, simple sensitivity studies were performed by the applicant and BNL to assess the potential influ-ence of these events on the risk estimates presented in PLG-0465.

By including consideration of these two additional areas (in addition to the other areas included in the original scope of the BNL review), it should

^

not be assumed that BNL has performed a detailed and systematic search for all

. events that might be important to risk at Seabrook.

Such a search is beyond the scope of the current BNL review.

The current review should therefore be 4

regarded as an evaluation of selected issues considered important to assessing the validity of the results and conclusions in PLG-0465 and the RMEPS.

The

- review therefore focused on assessing ways in which the Seabrook containment may fail or be bypassed early during a severe core melt accident.

3.

Findings in Each Review Area The approach taken by BNL was to perform sensitivity studies in selected areas to assess the impact on the results in PLG-0465 of the BNL review.

The BNL sensitivity studies used the conditional risk indices provided in PLG-0465 (and supporting documentation) to assess how changes in the probability of accident sequences and containment failure modes would change the Seabrook risk estimates.

The sensitivity studies calculated revi sed 200 rem-dose j

versus distance curves for comparison with those given in Figure S.3 and i

revised estimates of individual risk of early fatalities within 1 mile of the j

site boundary for comparison with the information given in Figure S.I.

l-The dose versus distance curves in Figure S.3 were constructed from dose j

versus distance curves (given in Figure S.4) for each of the source terms developed in PLG-0465.

These curves were then multiplied by their respective probabilities (given in Table S.1) and summed.

The combined dose versus distance curve was then normalized to the total core melt frequency.

To be l

consistent with the NUREG-0396 approach, which used median probabilities taken l

from WASH-1400, Figure S.3 was based on the median probabilities given in Table S.I.

The information on individual risk of early fatalities within 1 mile of i

the Seabrook site boundary given in Figure S.1 is based on the conditional risk indices given in Table S.2 for the various PLG-0465 source terms.

The f

earlier fatality risks were nultiplied by the mean frequencies in Table S.1, summed, and then divided by the population at risk.

Mean frequencies were used for this risk measure to be consistent with the NRC safety goal.

l xix

( -

The BNL review used the information in Table 5.2 and Figure S.4 to assess how changes in the probability of accident sequences and failure modes (and hence the probabilities. of the source terms in Table S.1) would change the risk estimates given in Figures S.1 and S.3.

Note that -Table S.2 also gives the total early fatality risk for each release category and that these were the only risk measures available to BNL when the first draft of this report was issued for review. Thus, the preliminary sensitivity studies in the draft report used total early fatality risk and tried to infer how changes in total.

risk might reflect changes in the early fatality risk within 1 mile of the Seabrook site boundary.

However, by comparing the early fatality risk within 1 mile of the site boundary with the total risk it is clear that virtually all of the early fatality risk for release category S2W and more than half of the risk for release category S6W occurs within 1 mile.

It is also clear that

' most of the risk of early fatalities for release category S1W occurs beyond 1 mile of the site boundary.

Therefore, a 2-mile evacuation eliminates all early fatality risk for release category S2W and virtually all for release category S6W.

However, a 2-mile evacuation - has no impact on the early fatality risk for release category S1W. Thus, it can be misleading to use the total risk of early fatalities as an indicator of the early fatality risk within 1 mile of the site boundary and this led to some confusion in the earlier draft, which has been corrected in this final version of the report.

When mean or median probabilities are used, a range of probabilities is obviously implied and the safety goal specifically states that an attempt has to be made to quantify the uncertainty associated with risk estimates.

The applicant considers the WASH-1400 source terms used in PLG-0465 to be very conservative and has a high confidence that the source terms would not be exceeded in a real accident. Therefore, in the opinion of the applicant, only uncertainty in the probabilities of the accident sequences and containment failure modes would impact the risk estimates in Figures S.1-S.3.

The appli-cant's upper bound or 95th percentile frequencies, which include consideration of the above uncertainties, are given for each of release category in Table l

S.1.

The impact of the 95th percentile frequencies in Table S.1 on the risk estimates in Figures S.1-S.3 is not great.

The leading contributor to the risk of early fatalities without evacuation in Figure S.1 is release category S2W.

The mean frequency of release category S2W increases by a factor of 5 if the 95th percentile value is used. Therefore, the early fatality risk without evacuation would increase by about a factor of 5 if the 95th percentile fre-quencies were used. However, if 1 mile evacuation is assumed, use of the 95th percentile frequencies would result in an early fatality risk below the safety goal.

Also, release category S2W is the only contributor to the 200-rem dose versus distance curves in Figure S.3 and, as this release category has no significant probability of exceeding 200 rem beyond 2 miles (refer to Figure S.4), changing its probability would not significantly change the results in Figure S.3.

In the following sections, the BNL findings related to each review area are briefly summarized.

BNL has attempted to follow the ground rules for comparison purposes (namely mean frequencies for comparison with the safety goal and median frequencies for comparison,with NUREG-0396 results), however, we have also attempted to include a discussion on the uncertainties associated with the risk estimates.

XX

s Int'erfacing Systems LOCA A major concern resulting from the BNL review of the interfacing systems LOCA analysis in PLG-0465 and the RMEPS related to the determination of initiator frequencies.

The effect of changing the initiator frequencies was determined by propagating the changes through the appropriate event trees in the RMEPS.

The revised initiator frequencies resulted in the following changes to the frequencies of release categories S1W and S7W.

Mean Frequency Per Reactor Year Release Category PLG-0465 BNL Review S1W 4.0x10 8 1.4x10 7 S7W 6.3x10-8 1.1x10 6 The above changes in release category frequencies have no impact on indi-vidual risk of early fatalities if no evacuation or 1 mile evacuation is assumed.

This is because release category S2W dominates this risk measure, and it has a frequency of 2x10 5 Only when a 2 mile evacuation is assumed (and the early fatality risk for category S2W becomes zero) do the above changes in release category frequencies change the original PLG-0465 estimates.

However, with a 2 mile evacuation the early fatality risk is very low and well below the safety goal. The 200-rem dose versus distance curve in Figure S.3 is also not influenced by the above changes in release category frequency.

This is because only release category S1W has a significant probability of exceeding a 200-rem dose, and the revised probability of this category is not sufficiently high for it to appear in Figure S.3.

There is of course uncertainty associated with predicting the frequency of interfacing systens LOCAs.

However, the frecuency of interfacing systems LOCAs resulting in release category S1W would have to increase by two orders of magnitude before the Seabrook dose versus distance curves would approach the curves given in NUREG-0396.

One can therefore conclude that interfacing systems LOCA are unlikely to increase the risk profiles presented in PLG-0465 to the level presented in NUREG-0396.

This is not too surprising because when no evacuation is assumed, the higher frequency events dominate risk and interfacing systems LOCAs did not contribute to the dose versus distance curves taken from NUREG-0396 (refer to Figure S.3.).

Accidents During Shutdown This topic was not originally addressed in PLG-0465 and a detailed assessment of such events is beyond the scope of the current BNL work on this proj ect.

However, the applicant was requested to provide information on the risk associated with accidents during shutdown.

The results of the appli-cant's assessment of such accidents were presented in the form of sensitivity studies in a draft version of this report.

The applicant provided additional frequencies to the existing release category frequencies given in Table S.1 to assess the impact on risk from accidents during shutdown.

A hase case and a bounding case were presented by the applicant.

The additional frequencies associated with these accidents are given b'elow:

xxi

I i

Mean Frequency Per Reactor Year Release Category Power Base Case Bounding

'~

Operation Events Shutdown Events Shutdown Events S.5 1.1x10 "

1.7x10 5 S.2 2.1x10-5 4.9x10 7 S.6 6.5x10 7 7.1x10 s 5x10-8 i

l BNL was not in a position to assess the above frequencies for these events because there remained fundamental questions regarding the modeling of i

f these scenarios.

However, in spite of this, the applicant's results were included in the draft report for comparison with the BNL sensitivity study results on other topics.

It should be noted that the applicant considered the

~i upper bound estimates to be very conservative.

In particular, in order to assess the impact of these events, they were included in source term categories derived for accidents from full power, which could lead to predicts of shorter times and larger quantities of fission product release than would be expected from accidents during shutdown.

In a subsequent submittal by the applicant, the consequences of accidents from shutdown were revised.

The applicant felt that 94 percent iof accidents at shutdown would occur at times later than 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> after scram.

Thus, the consequence estimates were reanalyzed assuming release-times of 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />.

The later release times resulted in dose versus distance curves which fall off 4

j at much shorter distances from the site boundary than the original dose versus distance curves.

BNL has checked this result and confirmed that if the release does occur at times greater than 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />, then the new dose versus 1

distance curves are reasonable.

i j

The results of the latest applicant's assessment of accidents at shutdown j

are reproduced in Figure S.S.

As noted above, a detailed assessment of such events is beyond the scope of the current RNL review.

However, based on our t

limited review of the applicant's assessment of these events, we still have i

reservations about the results.

These reservations are discussed in greater l

detail in the body of this report but untti they are resolved, we are unable i

to assess the validity of the risk estimates presented by the applicant in 3

Figure 5.5.

Induced Steam Generator Tube Rupture i

For accidents in which the primary system is at high pressure during core a

uncovery and melting, it is possible that large natural circulation flow patterns could develop within the primary system.

These flow patterns could in turn heat-up and degrade regions of the primary system remote from the i

reactor core. Of particular concern is the possibility of degrading the steam i

generator tubes such that the primary system could become open to the secon-j dary system.

If the secondary system were in turn open to atmosphere, then a direct path could exist between the primary system and the atmosphere, which 4

bypasses containment.

This topic was not included as part of the work scope for the current BNL review.

However, the topic has been reviewed in detail by the NRC staff and is the subject of continuing NRC and industry research j

activities.

Therefore, BNL performed sinple sensitivity studies to assess the xxii

a e

potential impact of induced steam generator tube rupture on risk at Seabrook.

The results of the sensitivity study are given in Figure S.6.

Because of uncertainty in predicting events of this nature and the fact that BNL did not evaluate this issue in detail, we were not able to develop a best-estimate frequency for induced SGTR.

The sensitivity study. therefore presents a range of possible frequencies for induced SGTR.

The frequency of high pressure squences and the conditional probabilities of failure of the operators to depressurize and induced SGTR that were used in the sensitivity study are given below:

4.0x10 s x 0.5 x 0.3 = 6.0x10 6 per reactor year 4.0x10-5 x 0.2 x 0.01 = 8.0x10-8 per reactor year.

In order to estimate the impact of the above probabilities on risk, an appropriate source tenn category had to be selected.

It was decided to allo-cate SGTR events to release category S1W, which represents a large early bypass of the containment.

It was felt that this was a conservative assump-tion because significant retention of the fission products in the secondary side could occur and this was not considered when calculating the S1W release fractions. The impact of adding the above frequencies to source term category

$1W is illustrated in Figure S.6.

The lower estimate of the frequency of induced SGTR has no impact on the risk estimates presented in Figures S.1-S.3.

The higher estimate of the fre-quency of induced SGTR has no influence on the individual risk of early fatal-ities within 1 mile of the site boundary if no evacuation is assumed but does influence the 200-rem dose versus distance curves as shown in Figure S.6 Allocating the probabilities of SGTR events to release category $1W has the largest impact on the dose versus distance curves (refer to Figure S.4).

How-ever, the impact on the risk of early fatalities within 1 mile is negligible because S1W has very little risk of fatalities within this distance (refer to Table S.2).

If the probabilities of SGTR events were added to release cate-gory 56W, the impact on the dose versus distance curves would be less but the risk of fatalities within 1 mile would increase slightly if no evacuation is assumed.

It should be noted that the range of frequencies used for the induced SGTR sensitivity study were developed to cover our lack of understanding in this area and that the NRC staff believes that the actual probability of a SGTR is closer to the lower estimate.

However, one reviewer of the BNL draft report felt SGTR to be a potentially more "significant" issue than was implied in our evaluation.

It was not BNL's intention in the draft report to minimize the potential importance of this issue, and the range we presented did not represent an upper bound.

It was an attempt to reflect the best judgments of several experts on a very difficult subject.

There is a great deal of uncer-tainty associated with predicting such events and it is therefore prudent to indicate the impact on risk of a range of assumptions.

Containment isolation Failure and Pre-existing Leakage The applicant's assessment of pre-existing leakage and containment isola-tion failure was reviewed by the NRC staff.

Based on the NRC staff review of the information available, it was concluded that the purge and vent valves in xxiii

i a fully closed configuration should provide reliable isolation of the Seabrook containment under severe accident conditions up to the pressure corresponding to 1 percent hoop strain in the containment.

The NRC staff also concluded that the applicant has presented a reasonable approach for the consideration of pre-existing leaks, both small and large.

The approach adopted by the applicant was to use information on containment unavailability developed in a study by the Pacific Northwest Laboratory (PNL) to assess the impact on risk of pre-existing leakage.

The applicant used this information to bound the effects of the data in the PNL study (NUREG/CR-4220) even though they considered that it did not apply to Seabrook.

The results of the applicant's assessment are given in Figure S.7.

From an inspection of Figure S.7, it is apparent that the impact of the NUREG/CR-4220 data on the dose versus distance curves is not great.

l Containment Structural Capacity Based on its nonlinear finite element analysis of the Seabrook contain-ment, BNL concluded that a shear failure at the base of the cylindrical wall is a potential failure mode but would not occur before reaching a pressure of 165 psig.

BNL agrees that the containment structure would reach a general yield state in the hoop reinforcing steel at a pressure of 157 psig and that it is appropriate to consider this pressure as a lower bound pressure for the hoop mode of failure.

However, BNL believes that the median hoop failure pressure should correspond to the one percent strain level in the hoop reinforcing steel, which is a pressure of 175 psig.

The above pressures are for the wet containment conditions.

For the dry containment conditions the corresponding median failure pressure is 158 psig and the lower bound pressure (general yield) is estimated to be 145 psig.

This latter value is based on the reduc-tion factor recommendation in Section 11.3.4.1 of PLG-0300 With regard to containment penetrations, BNL believes that the failure pressures should be based on containment deformations assuming no bond strength between the reinforcing steel and concrete. Based on this assumption RNL estimates median failure pressures for the wet containment condition of 159 psig and 167 psig for two critical penetrations.

For the penetration with the lower failure pressure, BNL agrees that a Type A (less than 6 square inches) leak path is appropriate for the median estimate; however a Type B (6 square inches to about 0.5 square foot) leak path should be considered as an upper bound estimate.

For the penetration with the higher failure pressure, BNL agrees that a Type B leak path is appropriate for the medium estimate; however, a Type C (greater than 0.5 square foot) should be considered as an upper bound estimate.

For the dry containment conditions, BNL estimated the median failure pressures for the above two critical penetrations to be 147 psig and 152 psig, respectively.

These values are also based on the reductior, factor recommended in Section 11.3.4.1 of PLG-0300.

Although BNL has performed some independent calculations to support its conclusions regarding the containment stren'gth, it also relied on the results of calculations performed by pSNH and its contractors.

Therefore, RNL xxiv

l

[

recommends that a complete and independent check of all relevant containment strength calculations be performed by PSNH.

PSNH committed to such a check in their letter to the NRC dated October 31, 1986 and has indicated that such a check has been completed.

Containment Loads BNL's assessment of the capacity of the Seabrook containment (described above) has to be combined with severe accident loads (pressure / temperature j

histories) to determine the potential for early containment failure.

BNL does not have Seabrook-specific containment loads and was not able to generate such loads given the limited scope of the current review.

However, BNL has been l

involved in updating (NUREG/CR-4551, Volume 5) the risk profile for the Zion l

plant for input to the NRC's " Reactor Risk Reference Document," NUREG-1150 The updating of risk for Zion was based on a methodology developed as part of the Severe Accident Risk Reduction Program (NUREG/CR-4551. Volumes 1-4) at Sandia National Laboratory (SNL).

This methodology used expert judgment in an i

attempt to estimate the uncertainty associated with determining containment loads.

The methodology was developed at SNL specifically for the Surry plant but was extrapolated to Zion at BNL.

The Zion plant is very similar to l

Seabrook in terms of the containment volume to reactor power ratio.

Thus.

I extrapolation of the Zion loads to Seabrook would give some indication of the j

impact of applying this new methodology to Seabrook.

It must be emphasized that this exercise should in no way be interpreted as a Seabrook-specific cal-culation.

It simply gives some indication of the sensitivity of the Seabrook.

results to the types of uncertainty in estimating containment loads discussed in NUREG-1150.

It should also be noted that this work is preliminary and has not yet undergone full peer review outside of NRC and its contractors.

It is, l

therefore, subject to revision.

The range of containment loads reported in Volume 5 of NUREG/CR-4551 for Zion is very wide and far exceeds the loads that would be considered credible by the applicant for Seabrook.

Of particular interest is the loads at the time of reactor pressure vessel failure.

These loads can range from about 60 psia to 200 psia depending on whether core melt is occurring with the primary i

l system at high or low pressure and on whether or not containment heat removal l

systems, CHRS (sprays and fan coolers) are operating.

The higher containment loads are postulated to occur for accidents in which the primary system pres-l sure remains high immediately before reactor pressure vessel failure.

For these accidents, direct heating of the containment atmosphere by core debris or hydrogen combustion with a steam spike at the time of reactor vessel fail-ure are possible mechanisms for failing the containment.

The applicant has presented information which indicates that these mechanisms are not credible ways of falling the Seabrook containment.

However, as noted above, BNL does not have Seabrook-specific containment loads so we cannot, at this time, elim-inate these mechanisms as potential ways of failing the Seabrook containment.

For accidents with the primary system at high pressure and without the CHRS operating an approximate median load of 135 psia (120 psig) was predicted for Zion.

if this median load is compared against the capacity of the Sea-brook containment given by the BNL review, one would conclude that the poten-tial for early containment failure at Seabrook is very low and would not influence the risk estimate in Figures S.1-S.3.

However, the range of loads estimated for Zion implies considerable uncertainty.

The 95th percentile xxv I

L

estimate of the probability of early containment failure at Zion is quoted as 0.17 in Volume 5 of NUREG/CR-4551.

If this early containment failure proba-bility were also true for Seabrook, the risk estimates in Figures S.1-S.3 would increase significantly.

However, the capacity of the Seabrook contain-nent is greater than Zion (the general yield for Seabrook is 157 psig compared with 134 psig for Zion) so the 95th percentile estimate of early containment failure should be lower at Seabrook than Zion.

However, BNL cannot at this time quantify how much lower because we have not quantified Seabrook-specific containment event trees with Seabrook-specific containment loads combined with our estimate of the structural capability of the Seabrook containment.

Source Terms The fission product source terns.used in PLG-0465 were reviewed in terms of their consistency with the approaches used in WASH-1400 and found to be appropriate. A misprint in PLG-0465 related to the release of noble gases for release category S2W was discovered.

However, correcting the noble gases release was found to have no impact on the risk profiles in PLG-0465.

In addition, the argument presented by the applicant that water in the residual heat renoval (RHR) vault is sufficiently subcooled to warrant consideration of significant decontamination was found to be reasonable.

This is an important consideration for the subset of interfacing systems LOCAs where the break location in the RHR line is low in the RHR vault. Under these circumstances, with the break location submerged considerable scrubbing of the aerosol fission products would occur.

This would result in much lower aerosol fission product release than for accidents in which the break location was uncovered.

At the start of this section, we noted that the applicant considers the WASH-1400 source terms used in PLG-0465 tci he very conservative and the appli-cant has high confidence that the source terms would not be exceeded in a real accident.

BNL found the source terms used in PLG-0465 to be consistent with WASH-1400 methodology but we are not as confident as the applicant that they could not be exceeded.

The new source term methods (refer to NUREG/CR-4551, Volumes 1-5) indicate that if the containment fails late or if there is gradual leakage from containment then the aerosol fission product release is likely to be lower than would be predicted by WASH-1400 methods.

This is because WASH-1400 methods underpredicted aerosol agglomeration and settling.

Therefore, if the new methods were applied to release categories S2W and S6W, the predicted aerosol release would be lower than WASH-1400 values.

However, the new methods also indicate that if containment fails early and the opening is large, then there is considerable uncertainty associated with predicting fission product release.

The uncertainty ranges associated with fission product release in NUREG/CR-4551 can, for certain accident sequences and early containment failure modes, exceed the WASH-1400 predictions. This uncertainty would principally affect the S1W release category at Seabrook.

Consequence Model The applicant used the CRACIT code for their consequence assessments in PLG-0465.

BNL compared CRACIT predictions of dose versus distance with pre-dictions from the MACCS code, which was developed at Sandia National Labora-tory (SNL) under NRC sponsorship.

The comparison of the dose versus distance curves for the CRACIT and MACCS codes was reasonably good.

Therefore, BNL feels that the dose versus distance modeling in PLG-0465 is fairly presented xxvi

\\

t and that the relatively 'snall ifferences between CRACIT pr? dictions and those s

computed by BNL using MACCS are explained by differences in modeling tech-

^

niques used in the two codes.

+

RNL could not check the risk of early fatalities reported in PLG-0460 because we did not have the population distribution around the Seabrook site.

Therefore, as BNL had only CRACIT results for early fatalities, it was decided to use CRACIT results for both early fatality risk and dose versus distances I

in the BNL sensitivity study.

This was done simply so that we had consistency between the two risk measures and not (as implied by the applicant's review of the draft BNL report) to present omre " conservative" CRACIT results. We found that CRACIT in general predicted dose versus distance curves that extended further than the MACCS code and in this sense CRACIT is more " conservative" than MACCS, However, we note that MACCS predlets more early health risk than CRACIT and therefore the use of the CRACIT results is probably not "conserva-a tive" for this risk measure.

In a,ddition, we are concerned about the CRACIT predictions of early fatality risk close to the site boundary for release category S1W (refer to Table S.2).

We consider these CRACIT predictions ta be lower than would be calculated using MACCS. Therefore, use of the CRACIT pre-dictions for early fatality risk in the population within one mile of the site boundary could give a more favorable comparison with the safety goal than would have been achieved using MACCS predictions.

s 4

Conclusions A major conclusion of PLG-0465 is that tharo is no significant frequency of exceeding 200 rem beyond 1.5 miles at Seabrook and therefore a significant reduction of the current 10 mile EPZ is warranted.

In order to draw this con-clusion, the applicant must have high confidence; that source terms, which result in a 200 rem dose beyond 1.5 miles (rclease categories S1W and S6W im Figure S.4), have a very low frequency (refer to Table S.1).

This in turn implies that the applicant must have confidence in their plant model and in their ability to predict low frequency events with high confidence.

j The objective of this technical evaluation was to assess the results and conclusions in PLG-0465.

The evaluation made no attempt to reassess or validate the total risk profile at Seabrook. The current review was an evalu-ation of selected issues related to the potential for a large early release of radioactivity at the Seabrook Station.

The conclusions of the BNL review for each of the selected issues are briefly given below:

1) A major concern resulting from the BNL review of the analysis of In-terfacing Systems LOCA in PLG-0465 and RMEPS related to the determinatiot of-initiator frequencies. This concern resulted in significantly shigher frequen-cies for interfacing systems LOCA than in PLG-0465.. However, the higher fre-quencies suggested by the BNL review did not influente the risk estimates in PLG-0465.

2)

Accidents during shutdown were not originally addressed in PLG-0465 4

or RMEPS but the applicant did provide an assessment of these accidents as part of an information request. A hounding analysis provided by the applicant indicated that accidents during shutdown have the potential to significantly impact the risk estimates in PLG-0465.

However, the applicant's best judgment was that if these accidents occur, they would most likely occur days after xxvii

_sm-

7,

~

o r

1shutdcwn and their 1.npact on the current Seabrook risk estimates would be min-i:nal.

However, BNL has -rnervations about the applicant's analysis of acci-s dents during shutdoW'and until these reservations are resolved BNL cannot

. assess the validity of the applicant's risk estimates for this class of acci-dents.

9

'O Sensitivity studies performed by BNL indicate that induced steam ger.eratbr tube rupture ista' potentially~ risk important issue for accidents in which the primary syttem is at high pressure.

This issue was not reviewed in detail by ENL and questions remain on whether or not induced steam generator tube rupture would occur in the event of a severe accident.

BNL considers that it has not yet been demonstrated that this issue is not risk significant for Seabrook.

4) The potential for Containment Isolation Failure and Pre-existing j

Leakage at Seabrook was ~not reviewed in detail by BNL or the NRC Staff.

The J f NRC Staff concluded that the purge and vent valves in a fully closed position should provide reliable isolation under severe accident conditions. Estimates

.m1dc by the applicant using generic data for containment isolation failure (hUREG/CA-4220) shcwed that this issue has a small impact on risk.

BNL has cot assessed the validity of the applicant's risk estimates for isolation failures.

5) Based on its nonlinear finite element analysis of the Seabrook con-tainment, BNL conclude _d that a shear failure at the base of the cylindrical wall is a potential failure. mode but that it would not occur before reaching a presr. ore of 165 psig. BNL agrees that the containment structure would reach a general yield state in the hoop reinforcing steel at a pressure of 157 psig j

and that it -is appropriate to consider this pressure as a lower bound pressure for the hoop mode of failure.

However, BNL believes that the median hoop failure pressbre should correspond to the one percent strain level in the hoop

. reinforcing steel, which is a pressure of 175 psig.

The above pressures are for the wot containment conditions.

For the dry containment conditions the

~

corresponding median failure presrure is 158 psig and the lower bound pressure (general yicid) is estimated to be 145 psig.

6) With regard to containn.ent penetrations, BNL believes that the fail-ure pressures should be based on containment deformations assuming no bond i

strength between the reinforcing steel and concrete. Based on this assumption BNL estimates median failure pressures for the wet containment condition of 159 psig and 167 psig for two critical penetrations. For the penetration with 2

the lower failure tressure,.a median leakage area of 6 in is appropriate.

l However, for the peret[ation with the higher failure pressure a larger redian leakage area of 72 in is appropriate.

For the dry containment conditions, BNL estimated the median failure pressures for the above two critical penetra-i tions to be 147 psig and 152 psig, respectively.

7) B%. did not develop Seabrook-specific containment loads given the scope of the current review.

However, BNL did develop Zion-specific contain-ment loads as part of updating (NUREG/CR-4551, Volume 5) the Zion risk profile for input to NUREG-1150.

As the Zion plant is similar to Seabrook, it was l

decided to use the Zion-specific loads to give some indication of the sensi-tivity of the Seabrook certainment to the ' types of uncertainty in estimating f-j containment loads identified in NUREG-1150.

The range of loads reported in l

xxviii i

/

i

N'JREG/CR-4551 is very wide (60-2')0, psia) and far exceeds the loads that the 1

applicant considers credible for Seabrook.

However, if the median Zion load is compared with the capacity of the Seabrook containment given by the BML,

review, the potential for early containment failure at Seabrook is predicted ' +

to be very low.

However, the range of Zion loads implies considerable uncer-tainty in estimating the probability of early containment. failure.

Most of this uncertainty is given by accidents in which the primary system pressure remains high immediately before vessel breach.

For these accidents direct V

heating of the containment atmosphere by the core debris or hydrogen.combus-tion with a steam spike at the time of reactor vessel failure have been postu-

. lated as mechanisms which could fail the containment.

The applicant has presented information which indicates that these mechanisms are not credible ways of failing the Seabrook containment.

However, as BNL has not developed Seabrook-specific containment loads, we cannot confirm that the uncertainty associated with predicting the probability of early containment failure at Seabrook is as low as that claimed by the applicant.

8) The fission product source terms used in PLG-0465 were reviewed in terms of their consistency with the approaches used in WASH-1400 and found to be appropriate.

9) The applicant used the CRACIT code for their consequence assessments in PLG-0465.

BNL compared CRACIT predictions of dose versus distance with predictions from the MACCS code, which was developed at Sandia National Laboratory (SNL) under NRC sponsorship.

The comparison of the dose varius distance curves for the CRACIT and MACCS codes was reasonably good.

There-fore, BNL feels that the dose versus distance modeling in PLG-0465 is fairly presented and that the relatively small differences between CRACIT predictions and those computed by BNL using MACCS are_ explained by differences in modeling techniques used in the two codes.

BNL could not check the risk of early fatalities reported in PLG-0465 because we dio not have the population dis-tribution around the Seabrook site.

However, BNL has questioned the validity of the very low CRACIT predictions of early fatalities close to the Seabrook site for the high energy S1W release category.

Based on the results of our focused review of PLG-0465 and RMEPS, BNL has low confidence that release categories S1W and S6W (or their equivalent) have frequencies as low as these given in Table S.I.

Therefore, BNL is, at this time, rather less confident than the applicant that there is no significant o

frequency of exceeding 200 rem beyond 1.5 miles at Seabrook.

Further work is needed in the following areas before BNL could alter its confidence level in the Seabrook risk estimates:

1) A detailed evaluation of accidents during shutdown;

2) Further evaluation of induced steam generator tube rupture;

3) An independent quantification of Seabrook-specific containment event t'rees with Seabrook-specific containment loads including the current BNL assessment of the structural capability of the Seabrook contain-ment; l

l

4) An independent systematic search f'or all accidents that might lead to early loss of containment integrity.

l XXIX l

10'2 N

S BACKGROUND ACCIDENTAL FATALITY RISK g10'3 (5 FATALITIES PER 10,000 POPULATION PER YEAR)

E S

&104 N

4

$10-s 3

SAFETY GOAL (.001 TIMES O

BACKGROUND RISK)

U y in-s PLG-0465 8

?

V"

//

=

=30-7 E

/ W!TH 1 MILE g

EVACUATION Ng 10-8 RMEPS RESULTS WITH NO IMMEDI ATE z

PROTECTIVE ACTIONS 4

N

,,-s

- 4 lilli lilli w.. a.....

.. figure S.1 Comparison of Seabrook Station sensitivity results using WASH-1400 source term cethodology with background, safety goal individual and RMEPS risk levels.

(Reproduced from PLG-0465, April 1986)

XXX

O e

10-3 LEGENO

--- PLG-0465 E*

d 10 5 WASH-1400 0

(MEDIAN RESULTS)

N a

m E

NO IMMEDIATE PROTECTIVE g

N ACTIONS

$ 10-6 __ _

3 N

o

'N 1 - MILE

\\

EVACUATION

\\

W N

\\

\\

\\

10-7

\\

{

8

\\

\\

\\

\\

I

\\

10-8 2 - MILE

\\

~ - -

EVACUATION

\\

{

\\

RMEPS RESULTS N

OFF SCALE N

\\

l

\\

10'9 I

I

\\

I \\

l I

0 1

2 3

4 5

10 10 10 10 10 10 EARLY FATALITIES Figure S.2 Comparison of median risk of early fatalities at Seabrook station for different emergency planning options.

(Reproduced from PLG-0465, April 1986)

XXXi

1 i

e ie iiial i

e i

iiaiog i

i e i a a i4 4

NUREG-0396

[

Ez PLG-0465 lEg

,...............* RM EPS R ESU LTS FOR c<

$b 1

SEABROOK STATION (200 REM CURVE OFF u', w SCALE)

.a 2 ow

-[g 0.1 e<

-7 u

25z u, m y$

T-'%

wa

\\

u. u s

oZ

\\

\\

50 REM 200 E E!

. \\ REM l

=c

\\

\\

i 8

\\

i k

gC 0.01 g

T I

05 b

\\

\\

<05 8<

l g

EE l

\\

l

}

cG C

l

_~y l

~

D

\\l 50 REM 200 R'EM

[-

ieit i

i e i i i v il i

e e i eiei ie 0.001 1

10 100 1.000 DISTANCE (MILES) i l

Figure S.3 Comparison of Seabrook Station results in this study and RMEPS with NUREG-0396 - 200-rem and 50-rem whole body dose plots for no immediate protective actions.

(Reproduced from PLG-0465, April 1986)-

XXXii

~.

I 1.0 8

U T

A L

S

/ aroe early coiitainment failure 3

0.1 or bypass (high o^

energy) b 5

Ingmdients for these curves lg" Release Category f

SIW 1.

Potentially realizable o.

Containment /

source. terms 2

8E leakage 1.5 in 2.

Site characteristics enf 0.01 :

3.

No protective action Release Category for 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> co E'

52W 4.

Release. categories S3W

• [

\\

and S5W did not exceed g,

x xe 3

~

Containment a 200 rem dose-2 leakage 50 in

[ 0.001 Release Category S6W

.t*

2 E

.0001 0.1 1

10 100 Distance (miles)

Figure S.4 200 rem dose versus distance curves for various failure modes assuming no.

immediate protective action.

(Reproduced from PLG-0465.)

o 1

a a

e ie iil e

..g.

ME NE n

.>c Q<

^

$5 m

U T o

NUREG-0396 7

li

/

e5

=>

y5 wa

u. o oz g

h{

33

-g+-- Original PLG-0465 results

%D%

bwo s\\ \\

E g 0.01 gg

\\

g g,'

Shutdown events assuming early

$ <5 l\\

\\

release,PSNHletter(SBN-1225)

~

EE l\\

\\

C llE l\\

\\

Shutdown events assuming 48 hour5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> l

t release, PSNH letter (NYN-87-002)

\\

~

l\\

g

\\

200 REM l

I

'I 0.001 1

10 100 1,000 DISTANCE (MILES)

Figure S.5 Comparison of 200 rem-dose versus distance curves for conservative assumption of no credit for l

operator recovery of open equipment hatch (calcu-lations performed by PSNH).

l l

xxxiv l

1 i iieiog i

e i i i ig i

i.....

Sh.

Ee

[

>8 8*

TU

"!I ow I8 o'

Eo NUREG-0396 55

/

w>

WG wt u

u. u i

\\

E Eo

~\\ y PLG-0465 g

88

\\

g g 0.01 g

$8 5gs

<0 N

g <5 1

s Es I

's og;;;

s l

s u8

\\

Impact of pessimistic assumptions I

ye 'r.egarding SGTR (optimistic assump-l t

tion off graph)

I I

I I

I 0.001 1

10 100 1,000 DISTANCE (MILES)

Figure S.6 Comparison of BNL sensitivity studies with PLG-0465 and NUREG-0396.

(200-rem plots with no immediate protective _ actions.)

XXXV

_=

t' 1

• • • l a

i e i i ii.:

,s -

j.

a<

~

"!I ow gg 0.1 eu NUREG-0396

!z b

$5' W5 mtu

~

ciE

~

results wQ g

88 g

g ? 0.01 g

u. u ag

\\

"a

,,,. Effects of conservative

^

interpretation by PSNH of h

NUREG/CR-4220 data,.PSNH D9 e

l\\

letter.

g2 u8 gg i

I i

l I

e i

e' iiei e

i e i i,iiI o,co 1

10 100 1,000 DISTANCE (MILES) l Figure S.7 Comparison of 200 rem-dose versus distance curves i

for conservative interpretation by PSNH of NUREG/CR-i 1

4220 data.

(Calculations performed by PSNH.)

0 l

j xxxvi

Table S.1 Sunmary of Release Category Frequency Uncertainty Distributions 1

Annual Frequency i

Release Category i

Lower Bound Median Mean Sth Percentile 50th Percentile Estimate-95th Percentile-i r

S1 - Early Containment Failure 1.5(-9) 1.5(-9) 4.0(-9)*

5.2(-9)*

1.5(-9) t S2 - Early Containment Leakage 3.5(-7) 7.5(-6) 2.1(-5) 2.0(-5) 1.0(-4) 1 S3 - Late Containment g

Overpressurization 5.1(-5) 8.3(-5) 1.4(-4) 1.4(-4) 2.3(-4) 3h SS - Containment Intact 5.5(-5) 7.7(-5) 1.1(-4) 1.1(-4 )

1.8(-4)

S6 - Containment Purge Isolation 1

Failure

<10(-10) 1.5(-8) 6.5(-7) 3.2(-7) 4.4(-6)

{

S7 - RHR Pump Seal Bypass 9.8(-10) 4.5(-9) 6.3(-8).

3.9(-8) 1.4(-7 )

• Mean influenced by right tail of distribution beyord the 95th percentile.

NOTE:

Exponential notation is indicated in. abbreviated form; i.e., 1.5(-9) = 1.5 x 10-9 i

1

Table S.2 Early Fatalities Conditional on a Release Occurring in the Population Around the Seabrook Station Site Boundary Mean Early Fatality Risk No Evacuation 1 Mile Evacuation 2 Mile Evacuation Release Category Within Within Within 1 Mile Total-1 Mile Total

.1 Mile Total S1W 9

746 9

746 9

746 S2W 121 122 8

9 0

0 S6W 385 734 193 542 2

63 l

i 1

I xxxviii 1

t

)

- 3_3' l

1.

INTRODUCTION t.

t

-1.1

Background

The - Seabrook Station Probabilistic Safety Assessment (SSPSA)1 was com-(

pleted by Pickard, Lowe,.and Garrick (PLG) Inc., for - the Public Service t

Company of New Hampshire (PSNH) and. Yankee Atomic Electric Company in' December

[

1983 and submitted 'to the Nuclear Regulatory Commission (NRC).

The NRC staff

'I and its supporting contractor initiated an in-depth review of sections of the SSPSA related.to determining those accident sequences that could lead to core j

damage.

However, this effort was terminated prior to completion of the review. 2 A. separate contract was placed with Brookhaven National Laboratory (BNL) to perform a very limited 3 review of those portions of the SSPSA related i

to core meltdown phenomenology, containment response, and radiological source terms. The BNL review 8 did not include an assessment of the physical strength f

of the Seabrook Containment.

i The key results of the SSPSA l are given below:

The mean, and median values of the uncertainty distribution for core melt frequency were found to be 2.3x10 4 and 1.9x10 4 events per l

. reactor-year, respectively.

i Both the societal and individual risk provisions of the NRC safety goals were met by wide margins; hence, the risk to public health and safety was estimated to be extremely small.

Different risk factors were found to have different key contributors.

Interfacing systems LOCA events and, to a lesser extent, seismic-induced transient events were the principal contributors to early health risk.

The contributors to core melt frequency and latent health risk were made up of a large group of initiators, including loss of offsite power, transient events, fires, and seismic events.

The dominant contributors to core melt frequency were support system faults, external events, and internal hazards that affected both the I

r t

core cooling and containment heat removal systems.

As. a result, a major fraction of the core melt frequency, 73%, was associated with l-sequences in which long-term containment overpressurization was indi-cated, while only 1% was associated with early containment failure.

t In contrast with previous containment analysis, the timing of contain-c ment _overpressurization in the above sequences was found to be measured in units of days rather than hours.

{

A major result of the SSPSA was that interfacing system LOCA events were l

the principal contributors to early health risk.

The results of' the SSPSA h

were updated in the Seabrook Station Risk Management and Emergency Planning Study -(RMEPS),. PLG-0432" to - account for new insights regarding radioactive l

release source terms and the progression of sequences involving loss of coolant events that bypass the containment.

1 l

I The purpose of the RMEPS was to present the results of a technical evalu-ation of emergency planning options and other risk management ' actions that i

were under consideration for the Seabrook Station. The principal focus of the study was the evaluation of the impact' of various protective actions such as i

evacuation and sheltering to various radial distances from the plant site.

l RMEPS rebaselined the SSPSA analysis of these risk-important areas specifical~-

l ly to establish' an updated assessment of the risk at Seabrook so that alterna-l tive emergency planning options could be developed.

Therefore, RMEPS focused t

on new data and engineering insights 'about the initiation 4nd progression of sequences involving interfacing systems loss-of-coolant accidents (LOCAs) and on the results of experimental and analytical research that provide an enhanced basis for assessment of radioactive material release (i.e., source terms) for a wider spectrum of accident sequences. Thus, RMEPS represents the applicant's best estimate of risk at Seabrook.

+

A second report related to emergency planning was published 5 (Seabrook

[

Station Emergency Planning Sensitivity Study, PLG-0465) which determined the radius of the Emergency Planning Zone (EPZ) that could be justified without consideration of any advances regarding the source term methodology since the j

completion of the Reactor Safety Study (WASH-1400)6 in 1975.

It is this i

l

1-3 second study that is the focus of the current BNL review although in order to review PLG-0465, BNL had to evaluate the results of the RMEPS and the SSPSA.

The principal conclusion of PLG-0465 was that an EPZ at the Seabrook Station of 1 mile radius or less is more justified in terms of its risk management effectiveness than the current 10-mile EPZ was justified by the results of NUREG-0396.7 This conclusion was based on the results of the PLG-0465 Sensitivity Study, which are reproduced in Figures S.1-S.3.

These results were constructed.without accounting for any new insights about source terms since WASH-1400.

The conclusion was based on the following observa-tions:

The individual risk of early fatalities in the population within 1 mile of the site boundary with no immediate protective actions is less than the NRC safety goal (refer to Figure 1.1).

This individual risk is substantially less when a 1-mile evacuation is assumed.

The risk of early fatalities with a 1-mile evacuation is comparable to the WASH-1400 results, which assumed a 25-mile evacuation (refer to Figure 2.2).

The Seabrook Station results for a 2-mile evacuation are substantially less than those for WASH-1400.

The risk of radiological exposures for 1, 5, 50, and 200-rem whole body doses with no immediate protective actions is less at 1 mile than the corresponding NUREG-0396 results at 10 miles (refer to Figure 3.3).

The above observations led to the statement in PLG-0465 that "there is no significant frequency of exceeding 200 rem beyond 1.5 miles in the Seabrook sensitivity results."

PLG-0465 identified the following three areas as being the most influential in calculating the Seabrook risk estimates:

The effectiveness of the Seabrook Station primary containment to either remain intact or to maintain its fission product retention capability for periods much longer than required for even delayed, ad hoc protective actions.

1-4 A more realistic assessment of the strength and failure modes of the Seabrook containment than was possible within the state-of-the-art of PRA when the RSS was completed.

A more realistic treatment of the initiation and progression of inter-facing systems LOCA sequences.

Note that of the three areas identified above as being the most influen-tial' to the risk estimates in PLG-0465 only the first is Seabrook specific, namely the effectiveness of the Seabrook containment.

The other two areas refer to improvements in methodology and would therefore apply equally as well to other nuclear power plant risk assessments.

1.2 Scope and Focus of Review At the request of the NRC, the BNL technical evaluation initially focused on the following areas in PLG-0465 and RMEPS:

~

- Interfacing systens LOCAs

- Containment function:

- Isclation failure

- Pr?-existing leakage

- Structural capacity

- Containment loads

- Seabrook-specific WASH-1400 source terms

-. Site consequence model.

However, during the review process additional areas that were outside of the scope of the original BNL review were identified as potentially important

-to risk at Seabrook.

Two of these areas were considered sufficiently impor-tant to request the applicant to provide additional information on the risk associated with such events. The two areas identified were:

l

- accidents during shutdown

- potential for induced steam generator tube rupture.

i 0

_m,,-%-_

-_.~,---..-m

1-5 These areas were. not initially included in the original BNL review because in the past -they were not found to be dominant risk. contributors.

However, as the risk estimates in PLG-0465 and the RMEPS are relatively low, events that were previously considered to be unimportant now have the poten-tial to influence the Seabrook risk estimates.

Thus, simple sensitivity studies were performed by the applicant and BNL to assess the potential influ-ence.of these events on the risk estimates presented in PLG-0465.

By including consideration of these two additional areas (in addition to the other areas included in the original scope of the RNL review), it should not be assumed that BNL has performed a detailed and systematic search for all events that might be important to risk at Seabrook.

Such a search is beyond t.he scope of the current BNL review.

The current review should therefore be regarded as an evaluation of selected issues considered important to assessing the validity of the results and conclusions in PLG-0465 and the RMEPS.

The review therefore focused on assessing ways in which the Seabrook containment may fail or be bypassed early during a severe core melt accident.

i 1.3.

Approach to Review The approach taken by BNL was to perform sensitivity studies in selected areas to assess the impact on the results in PLG-0465 of the BNL review.

The BNL sensitivity studies used the conditional risk indices provided in PLG-0465 (and supporting documentation) to assess how changes in the probability of accident sequences and containment failure modes would change the Seabrook risk estimates.

The sensitivity studies calculated revised 200 rem-dose versus distance curves for comparison with those given in Figure 1.3 and revised estimates of individual risk of early fatalities for comparison with the information given in Figure 1.1.

The dose versus distance curves in Figure 1.3 were constructed from dose versus distance curves (given in Figure 1.4) for each of the source terms

}

developed in PLG-0465.

These curves were then multiplied by their respective probabilities (given in Table 1.1) and summed.

The combined dose versus distance curve was then normalized to the total core melt frequency.

l To be consistent with the NUREG-0396 approach, which used median probabilities taken

~

1-6 from WASH-1400, Figure 1.3 was based on the median probabilities given in Table 1.1.

The information on individual risk of early fatalities within 1 mile of the Seabrook site boundary given in Figure 1.1 is based on the condit!ional risk indices given in Table 1.2 for the various PLG-0465 source terms.

The earlier fatality risks were multiplied by the mean frequencies in Table S.1, summed, and then divided by the population at risk.

Mean frequencies were

.used for this risk measure to be consistent with the IRC safety goal.

I,.

The BNL review used the information in Table S.2 and Figure S.4 to assess how changes in the probability of accident sequences and failure modes (and hence the probabilities of the source terms in Table S.1) would change the i

risk estimates given in Figures 1.1 and 1.3.

Note that Table S.2 also gives.

the total early fatality risk for. each release category and that these were the only risk neasures available to BNL when the.first draft of this report was issued for review. Thus,.the preliminary sensitivity studies in the draft report used total early fatality risk and tried to infer how changes in total risk might reflect changes in the early fatality risk within 1 mile of 'the Seabrook site boundary.

However, by comparing the early fatality risk within 1 mile of the site boundary with the total risk it is clear that virtually all of the early fataity risk for release category S2W and more than half of the risk for release category S6W occurs within 1 mile.

It is also clear that most of the risk of early fatalities for release category S1W occurs beyond 1 mile ~ of the site boundary.

Therefore, a 2-mile evacuation eliminates all early fatality risk for release category S2W and virtually all for release i

category S6W.

However, a 2-mile evacuation has no impact on the early fatality risk for release category S1W..Thus, it can be misleading to use the total risk of early fatalities as an indicator of the early fatality risk within 1 mile of the site boundary and this led to some confusion in the earlier draft, which has been corrected in this final version of the report.

When mean or median probabilities are used, a range of probabilities is obviously implied and the safety goal specifically states that an attempt has to be made to quantify the uncertainty associated with risk estimates.

The applicant considers the WASH-1400 source terms used in PLG-0465 to be very

~. _

o g_7 conservative and has a high confidence that the source terms would not be exceeded in a real accident. Therefore, in the opinion of the applicant, only n

uncertainty in the probabilities of the accident ' sequences and containment failure modes would impact the risk estimates in Figures S.1-S.3.

The appli-cant's upper bound or 95th percentile frequencies, which include consideration of. the above uncertainties, are given for each of release category in Table S.I.

The impact of the 95th percentile frequencies in Table S.1 on the risk estimates in Figures S.1-S.3 is not great.

The leading contributor to the risk of early fatalities without evacuation in Figure S.1 is release category S2W. The mean frequency of release category S2W increases by a factor of 5 if the 95th percentile value is used. Therefore, the early fatality risk without evacuation would increase by about a factor of 5 if the 95th percentile fre-quencies were used. However, if 1 mile evacuation is assumed, use of the 95th percentile frequencies would result in an early fatality risk below the safety l

goal. Also, release category S2W is the only contributor to the 200-rem dose versus distance curves in Figure S.3 and, as this release category has ~ no significant probability of exceeding 200 rem beyond 2 miles (refer to Figure S.4), changing its probability would not significantly change the results 'in Figure'S.3.

1.4 Organization of the Report E

The previous section identified the focus of the BNL review of PLG-0465 and indicated the limitations of the effort.

The report-is organized to address each of the areas discussed in Section 1.2.

Initially, in Section 2, those portions of PLG-0465 and the RMEPS (PLG-0432) related to system failure i

are reviewed to determine the appropriateness of the frequencies of accident sequences that could lead to early loss of containment integrity.

i Section 3 reviews the ability of the Seabrook Station primary containment i

to withstand the very severe pressure / temperature loads associated with core 4

j meltdown accidents.

This is a very important review because the applicant considers that the Seahrook containment has a significantly greater capability

(

for containing core meltdown accidents than a number of other large dry 2

I h

l

1-8 T

containments that have been reviewed by the. NRC staff over the last several

~

years.-

The sensitivity of the conclusions in PLG-0432 to uncertainties in con-tainment loads (pressure / temperatures histories) and containment performance 1

(based on the review in Section 3) is explored in Section 4.

The soure.e terms used in PLG-0432, which were based on RSS methodology, are reviewed in Section 5.

Finally, in Section 6, the site consequence model and the risk calcula-tions presented in PLG-0432 are reviewed.

1.5 References l

1.

"Seabrook Station Probabilistic Safety Assessment," Pickard, Lowe and Garrick, Inc., PLG-0300, December 1983.

2.

Garcia, A.

A., "A Review of the Seabrook Station Probabilistic Safety I

Assessment," Draft Report, Lawrence Livermore National Laboratory, dated December 12,.1984.

3.

Khatib-Rahbar, M., et al., "A Review of the Seabrook Station Probabilistic Safety Assessment:-

Containment Failure Modes and Radiological Source Terms," NUREG/CR-4540, February 1986.

l 4.

"Seabrook Station Risk Management and Emergency Planning Study," PLG-0432, December 1985.

5.

"Seabrook Station Emergency Planning Sensitivity," ~PLG-0465, April 1986.

6.

U.S.. Nuclear Regulatory Commission, " Reactor Safety Study:

An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants," WASH-1400, NUREG-75/014, October 1975.

7.

Collins, H. E., et al., " Planning Basis for the Development of State and Local Government Radiological Emergency Response Plans in Support of Light Water Nuclear Power Plants," prepared for the U.S. Nuclear Regulatory Com-l_

mission, NUREG-0396, December 1978.

r

0-1-9 10~2 o

8ACKGROUND ACCIDENTAL FATALITY RISK g10'3 (5 FATALITIES PER 10,000 POPULATION PER YEAR) iE S

4 10

5

,N 10-5

$o SAFETY GOAL (.001 TIMES o

BACKGROUND Risk) y 3o-6 PLG-0465 o

C V"

//

z k 10-7

$g WITH I MILE J.

EVACUATION j 10 8 z

RMEPS RESULTs 2

WITH NO IMMEDIATE PROTECTIVE ACTIONS

\\

10*9 l

Figure 1.1 Comparison of Seabrook Station sensitivity results using WASH-1400 source term methodology with background, safety goal individual and RMEPS risk levels.

(Reproduced from PLG-0465, April 1986.)

1-10 10'3 LEGEND 4

PLG-0465 E

s cI: 10-5 WASH-1400 0

(MEDIAN RESULTS)

U de

~~----%""===

a.

NO IMMEDI ATE PROTECTIVE ACTIONS g

$ 10-6 3

g o

N 1 - MILE N

EVACUATION

\\

W N

\\

\\

\\

N 10'7

{

E

\\

\\

\\

\\

~

.\\

\\

\\

10-8 2 - MILE

\\

, [ EVACUATION

\\

{

\\

RMEPS RESULTS N

OFF SCALE N

\\

\\

10'9 I

l

\\

l \\

{

l 0

I 2

3 4

5 10 10 10 10 10 10 EARLY FATALITIES Figure 1.2 Comparison of median risk of early fatalities at Seabrook Station for different emergency planning options.

(Reproduced from PLG-0465, April 1986.)

1-11 1

...iiiog i

e i e i isig i

e i iie i a_

NUR EG-0396 0z oE PLG-0465 O Q o<

RMEPS RESULTS FOR

@b SEA 8 ROOK STATION

~

J, w (200 REM CURVE OFF a2 SCALE) ow Im oi 3: o eu E<

a z w

w$

T~'%

U X

\\

w w

u. O 05

\\

\\

50 REM

>s 200

$9

\\ REM wQ

\\

D Q om

\\

\\

y ~ 0.01 g

BS i

i

\\

~~

\\

gh o<

\\

EE l

I 85 c.

I i

ca i

l

/k

\\l 50 REM 2co'R EM V

l

'I

I 0.001 1

10 100 1,C00 DISTANCE (MILES)

Figure 1.3 Comparison of Seabrook Station results in this study and RMEPS with NUREG-0396 - 200-rem and 50-rem whole body dose plots for no'imediate protective actions.

(Repro-duced from PLG-0465, April 1986.)

1.0 o

D X

N L

S

/ aroe early containment failure 2

0.1 or bypass (high-energy) o^

i{

i Ingredients for these curves -

Release Category f, O 51W

, 1.

Potentially realizable Containment [2 source terms gg EE 0.01 leakage 1.5 in 2.

Site characteristics 4

j "E

1 Release Category 3.

No protective action for 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> E'

S2W 4.

Release categories S3W EE

~

\\

and S5W did not exceed j

U

[5

~

Containment a 200 rem dose 2

leakage 50 in 1

0 0.001 Release -Category S6W l

2 g

E

.0001

?

0.I 1

10-100..

Distance (miles)-

i i

i Figure 1.4 200-rem dose versus distance curves for various failure modes i

assuming no immediate protective action.

(Reproduced from PLG-0465.)

i i

^

Table 1.1 Summary of Release Category Frequency Uncertainty Distributions Annual Frequency Release Category Lower Bound Median M**n Sth Percentile 50th Percentile Estimate 95th Percentile S1 - Early Containment Failure 1.5(-9) 1.5(-9) 4.0(-9)*

5.2(-9)*

.1.5(-9)

S2 - Early Containment Leakage 3.5(-7) 7.5(-6) 2.1(-5) 2.0(-5) 1.0(-4)

S3 - Late Containnent Overpressurization 5.1(-5) 8.3(-5) 1.4(-4) 1.4(-4) 2.3(-4)

SS - Containment Intact 5.5(-5) 7.7(-5) 1.1(-4) 1.1(-4 )

1.8(-4 ) ~

E3 '

S6.- Containment Purge Isolation Failure

<10(-10) 1.5(-8) 6.5(-7) 3.2(-7)-

.4.4(-6)

S7 - RHR Pump Seal Bypass 9.~8(-10) 4.5(-9).

6.3(-8) 3.9(-8) 1.4 (-7 )

• Mean influenced by right tail of distribution beyond the 95th percentile.

NOTE: Exponential notation is indicated in abbreviated form; i.e.,1.5(-9) = 1.5 x 10-9 I

1-14 Table 1.2 Risk of Early Fatalities in the Population Around the Seabrook Station Site Boundary Mean Early Fatality Risk Release Category No Evacuation 1 Mile Evacuation 2 Mile Evacuation Within Within'

.Within

-1 Mile Total 1 Mile Total 1 Mile Total S1W 9

746 9

746 9

746 S2W 121 122 8

9 0

0 S6W 385 734 193 542 2

63 1

2-1 4

2. -

SYSTEM EVALUATION In this section, those positions of PLG-04651 and the RMEPS2 (PLG--0432) related to determining the frequencies of accident sequences leading to core

. melt is reviewed. However, the review made no attempt to reassess or validate the total core melt frequency at Seabrook.

Such reassessment or validation would require an extensive review and requantification 'of the results reported 3

in the SSPSA and this was not within the scope of the current BNL evalua-tion.

The current review was therefore an evaluation of selected issues j

related to the potential for a large early release ~ of radioactivity at the Seabrook Station..The original focus of the RNL review was therefore directed to assessing:

Interfacing systems LOCAs j

Containment function:

Isolation failure Pre-existing leakage.

4 However, during the review process the following types of questions guided-the focus of the BNL - review to other areas that were not in the origi-nal work scope:

1) whether the basic assumptions of the submittal are' adequate, and whether the analysis confronts all the important issues

2) whether, given the basic assumptions, the plant-specific modeling is t

self-consistent and complete within the range of issues addressed by the modeling l

3) whether the scenarios actually modelled are properly quantified

[

(including common cause considerations).

An example of a Type,1 completeness question is discussed in Section 2.3, namely, whether a high-pressure melt scenario can lead to steam generator tube degradation and concomitant containment bypass.

This is a fairly generic question, and one that was not in the original scope of the BNL review.

However..it was considered to be potentially important to risk and therefore I

we. thought that it should be considered.

This question is presently being I

studied by the nuclear industry and the NRC staff and their contractors. BNL i

l

- - ~ - - -,. - - -.. _. -.. - -.. - - - - -

2-2

'did not have the time to review the issue in detail so we simply performed a sensitivity study to give an indication of its potential importance to risk.

An example of a Type 2 completeness question is whether the check valve between residual - heat removal (RHR) suction and the refueling water storage I

tank (RWST) is. likely to fail in scenarios wherein the RHR system is overpres-surized.

The Seabrook study raises and discusses this question, but without clearly establishing why the check valve failure probability is low enough to warrant not exploring such scenarios.

This pa'rticular question is emblematic of a family of such questions which cannot be exhaustively tallied within the 4

scope of this limited review, but which bears on the submittal's conclusions.

Finally, a particular instance of a type 3 question is discussed below in Sec' ion 2.1, wherein the frequency of multiple check valve failures as derived t

for initiating events is challenged.

Since the present concern is with the frequency of a substantial release f

of radioactivity, this section focuses on several areas which bear directly on i

this question. The strength of containment is addressed in Section3.

In this section, selected modes of containment bypass loss of containment isolation are discussed.

These include classical interfacing systems LOCA, accidents during shutdown, failure of containment-isolation, and induced steam generator tube rupture during a severe core melt accident.

2.1 Interfacing System LOCA 2.1.1 General i

l According to the Seabrook RMEPS,2 one of the principal contributors domi-nating early health risk--and one which has been subjected to extensive re-3 analysis since the SSPSA --is an Interfacing Systems LOCA that bypasses con-tainment.

From all the potential pathways through which an Interfacing Systems LOCA (ISL) may occur, the study identified six lines as possible initiators for ISLs:

g' s

2-3 '.

~

b, Four lines in. the. cold. leg. safety injection (Low Pressure Injection,, '

[LPI]/ Residual Heat: Removal [RHR] Loop Return lines)

Two lines in -the suction side of' the RHR system. '

s The corresponding. initiator frequencies as well as 'the core damage f.e,

quency due to these initiators were obtained by PLG as a result of an enhanced ' ' '

and innovative Interfacing Systems LOCA analysis, which involved new treat-ments of.various aspects of the accident.

/

The new treatments are listed below:

More complete.modeling of valve failure modes ~

New data on check valve failures versus leak' size a

More realistic treatment of dynamic pressure pulse Explicit modeling of RHR relief ' valves 0uantification of RHR piping fragilities to overpressure -

Modeling of RHR pump seal leakage

,. Operator actions to prevent. melt considerod' Thermal hydraulic and source term factors modeled using MAAP code" Uncertainties quantified.

The initiator frequencies obtained by 'the new approach are:

Cold Leg Safety Injection Path (VI) = 4.5x10 6 event / reactor year RHR Suction Side Path (VS)

= 3.7/10 6 event / reactor year which yield a core damage-frequency of-CDISL = 4.4x10 8 event / reactor year.

i BNL has performed a limited review of the new analysis, compiled a number of questions and observations, and performed a sensitivity study on the initi-ator frequencies.

The objective of this subsection is to provide the results of this ifmited review.

l 2.1.2 Other ISL Paths s

BNL also performed a cursory survey 'for other potential containment bypassing pathways for ISL.

The BNL survey identified some pathways- (e.g.,

RHR lines to the RCS hot legs, letdown line, excess letdown line, etc.) which l-w

,..v.

,,.-7

-y,,

--.-,.,,.-m-7--.

--,.,--.---gg,

-7,,

.r--.y,

2-4

'j;'. g~

[

were ignored in the Seabrook study.

Although there may be justification for ignoring these pathways,BNL believes that all the potential pathways should i,F have been. identified and the basis for the rejection of each such path 56cumented.

L A

~2.1.3 ISL Initiator Fre'quencies o

The determination of the ISL initiator frequencies is one of the most

. important parts of the Seabrook RMEPS.2 This determination depends on the

?

correct estimate of the frequencies of relevant failure modes of the valves in the various interfacing lines. These valve failure modes are:

[

Disc rupture or gross leakage of series valves (check valves) in the J

LPI lines

. Disc' rupture or gross leakage of MOVs, failure of stem mounted limit switches, and disc failing open when indicating closed, in the RHR suction lines.-

1 i

The approach applied for modeling of initiator frequencies in the Seabrook

~

study is based on two "inndvative" steps:

a) Separation of the check valve gross (reverse) leakage failure mode into

" gross reverse leakage" and " failure to raseat on demand" failure modes, which kere treated together in earlier data bases.

l b) An analysis of data on check valve leakage frequency versus leak rate for check -valves of the RCS/ECCS system boundary.

This step resulted in applying a reduced check valve leakage failure frequency in the quantifi-cation of the initiator models.

(A result which is challenged below.)

In the process of surveying data of the Nuclear Power Experience (NPE) 5 data base, no disc rupture events were identified by PLG for check valves or HOVs.

The maximum observed leak rate was 200 gpm.

Leak rates were estimated

, based on other available evidence:

the rate of boron concentration change in

'the accumulators, rate of pressure increase in the accumulators, and similari-f ty to other occurrences for which the leak rates were known.

To estimate the 6

total ~ number of check valve hours, the information provided in NUREG/CR-1363 on the number.of valves in the ECCS in various PWRs was used.

PLG's total I

number of check valve-hours was 1.08x10s,,To estimate the frequency of check

[

valve failure to reseat on demand, two types of data were used:

estimates t

l r

u..

i y

_2-5

)

j r

s from generic sources of = failure data, and. experiential data from eight U.S.

UU' nuclear plants for which PLG performed plant specific PRAs.

y 2.1.3.1 Check valve failure frequencies Since the check valve failure frequencies play a crucial role in the ISL analysis, BNL'. performed a somew' hat more detailed review of that part of the

^

Seabrook study.. As a consequenceiof the review process the following observa-tions are made:

a) After a successive screening process of check valve failure events (start-ing from a total.of 692 events at both PWRs and BWRs), PLG limited its data base "to those events involving check valve leakage in the ECCS and RCS/ECCS system boundary of PWRs.

These were judged to be the closest category to the initially seated and tested check valves modeled in the analyses." The final number of failure events were:

17 accumulator check valve failures and 4 ECCS/RCS interface check valve failures, b) To estimate the total number of check valve hours, PLG used the total

. population of check valves in the ECCS _instead'of the corresponding subset of check valves at the ECCS interfaces.

This resulted in substantial and Jnappropriate overestimation' of check valve hours and thus a substantial

onderestimation of the check valve failure rate (i.e., check valve fail-Jres divided by check valve hours).

c)

The correct exposure time for check valve failures is not merely the time

.when the plant is operating. - For. example, check valves in the RHR are almost contir.uously exposed to potentially degrading conditions (during cold shutdowns, as well).

A correction facte I-r pressure exposure of interfacing lines should be considered $t-i cat - y, in calculating the initiator frequencies.

I d)

In many cases, PLG estimated single check valve leak rates from accumula-tor inleakages.

It must be recognized that the deduced leak rates from y

accumulator inleaka;es relate to two check valves in series, rather than j

leakage through a single check valve.

A ', leakages through two check valves in series, the less-leaking valve dominates (the other valve may be even wide open).

e) The leak failure frequencies versus lesk rate curve presented in the PLG study (reproduced in Figure 2.1) is only a first approximation for a more o

=

2-6 precise leak failure frequency versus - relative leak rate curve.

In par-ticular, this curve pooled data involving a variety of check valve sizes.

A more sophisticated treatment would require knowledge of the size popula-tion of check valves at the interfacing pathways.

f) The largest leak rate in ' Figure 2.1 is of the order of 200 gpm, whereas the arena of interest ranges to 65,000 gpm. The " linear" extrapolation to higher rates is not necessarily justified.

If the shape of the distribu-tion is Pareto, the linear extrapolation is in order.

However, if it follows a Rayleigh distribution, the extrapolation is not correct (but conservative).

Seabrook-specific considerations (valve sizes, designs) are not addressed in the analysis.

g) The initiator models implicitly assume that the leak tests of the valves

" discover" all failures and valves behave as new after each test.

The study does not describe the relevant test processes and the expected "real" efficiency of these tests.

h) The report does not consider common cause failures.

Such failures can indeed happen due to boron deposition, improper maintenance such as installation of improper components (gaskets, seats, or valve disks) which may fail almost immediately or at a later time.

In order to quantitatively estimate the consequences of some 'of the above mentioned deficiencies, BNL performed an independent detailed reevaluation of relevant chbck valve failure data. The process was facilitated by the availa-bility of relevant failure events selected for an independent study of ISL at PWRs, which is presently ongoing at BNL for the NRC.

Tables 2.1 and 2.2 present failure events for High Pressure / Low Pressure isolation check valves selected by BNL.

Table 2.1 contains events for the check valve ' leakage" failure mode. Table 2.2 presents the events for the valve " failure to reclose upon demand" failure mode. Table 2.1 also includes data on the estimated leak flow rates.

These latter data are obtained essentially with the same method as those of Table 3.8 of PLG-0432.

Comparing the number of failure events of Table 2.1 with that of Table 3.8 of PLG-0432, one finds that Table 2.1 contains more events (41) than Table 3.8 of PLG-0432 (21), which may be the result of a somewhat more efficient selection procedure at BNL.

.=

2-7 I

In order to :see what is the sensitivity of the initiator frequency for t'

the. check valve failure rate, BNL selected the subset. of accumulator check valve failure events (the majority of check valve failures 35 events, summa-rized in Table 2.4) for which the total time' of exposure can.be correctly determined.

The total time of exposure of accumulator check valves for all the PWRs in the U.S. is calculated in' Table 2.3.

The time - from start of commercial operation of individual plants was used as " time of exposure" for these. check valves, since water with boric acid constantly degrades these valves.

The total number of check valve-hours obtained is 2.34x10.

7 Based on BNL-gathered data, the frequencies of accumulator check valve leakage events for various leak rate ranges are given in Table 2.4.

The corresponding frequency exceedance/hr values are plotted against the check i

valve leak rates in Figure 2.1.

For comparison, Figure 2.1 shows also the PLG data.-

The shape of the curve is almost identical with that of PLG, but shifted higher, by almost one order of magnitude, due to the higher number of failure events identified and the more precise value for check valve-hours.

It is appropriate to mention here several precautions concerning the leakage. failure characteristics derived from accumulator check valve failure events.

In the NPE data base 5 the majority of interfacing check valve leakage events involve accumulator valves.

Although this seeming bias could arise from the extra monitoring of the accumulator, it could also reflect a particularly severe environment acting on the valves.

If l

the latter is true, then leakage exceedance frequency data (ordinates f

in. Figure 2.1) may lead to overestimates of the frequency for other interfacing check valves.

The leak flow rate data (" leak sizes"; abscissas in Figure 2.1) repre-sent lower limits for these quantities, because leakage flow rates estimated from accumulator inleakages involve, in most of the cases, leakage through two check valves in series.

As a result of these factors, a more realistic leakage failure i

i

O 8

2-8 exceedance - frequency /hr versus leak ' rate curve for non accumulator interfacing paths may be somewhat lower in. frequency at - low. leak rates, but might fall off more slowly,with increasing leak rate than do the curves in Figure 2.1.

Creation of this more realistic curve was beyond the scope of the BNL effort.

Since there are no more accurate data available, BNL recalculated the initia-tor frequencies by using the data obtained for accumulator check. valves.

i Since the purpose of this calculation is to contrast the result with that of' the _PLG analysis (to see the sensitivity of the initiator frequencies for check valve failure rate), the same extrapolation and calculational techniques are used as those of PLG.

2.1.3.2 Cold 1.eg safety injection path frequency t

This section presents a revised estimate of the initiator-frequency of 1 -.

interfacing LOCA through the injection lines. The calculation presented below 1

is intended to follow the PLG analysis - step by step, except that the check valve failure statistics have been modified as indicated above. Subsequently, these modified initiator frequencies are propagated through the PLG model to illustrate new plant damage state frequencies.

The following is, then, a recalculation of the PLG result using PLG methods but modifying the single check valve failure rate as previously discussed.

From Figure 2.1, the median frequency of a single check valve failure resulting in leakage that exceeds the capacity of one charging pump (i.e.,150 gpm) is about 1.1x10 7 per hour.

Assuming a lognormal distribution for this frequency and a range factor of 10 (which may be too conservative for this increased statistic) yields:

l Frequency of Check Valve Failure Parameter (Leakage 150 gpm) 95th percentile 9.6x10 3/RY Mean 2.6x10-3/RY Median 9.6x10 4/RY 5th percentile 9.6x10-5/RY

2-9 Similarly, the median frequency of exceeding 1800 gpm is 1.4x10 8 per hour.

~ Assuming a lognormal distribution with a range factor of_14 yields:

-Frequency of Check Valve Failure Parameter (Leakage 1800 gpm) 95th percentile 1.7x10 3/RY Mean 4.4x10 4/RY Median 1.2x10 4/RY 5th percentile 8.8x10 5/RY The frequency of " failure to reclose on demand" for check valves, Ad, is taken to be the same mean value as that used by PLG:

Ad = 2.7x10 4/ demand.

By using Formula 3.14 of PLG-0432, the estimated mean frequency of failure of two series injection check valves, that produces leakage to the RHR system in excess of 150 gpm is:

4.90x10 5 events /RY.

Since there are four injection paths, the mean value for the Cold Leg Safety Injection Paths becomes VI = 1.96x10 4 events /RY.

Top event, LR in the injection path event tree (see Figure 3-4 in Reference 2) represents the fraction of the initiating event frequency, VI, in which the leakage not only exceeds 150 gpm, but also exceeds 1800 gpm.

The product of LR and VI thus represents the dominant contributor to the frequency of overpressurization challenges to the RHR system due to failure of both check valves in the four injection paths.

Based on the above values, LR has a mean value of.058 and the overpressurization frequency for the cold leg injection path becomes:

(0.058)(1.96x10 4) = 1.14x10 5/RY.

2.1.3.3 RHR suction side frequency The same introductory remarks made at the beginning of Section 2.1.3.2 also apply here.

That is, the following represents BNL's attempt to show the i

sensitivity of the PLG analysis to a modified check valve failure rate.

2-10 For an ISL to occur in the RHR hot leg suction path, failure of two series MOVs must occur.

In the PLG-model for this path, the failure involves:

a) independent failures of both M0V valves, causing excessive leakage; or b) independent failure of one of the valves and a demand failure of the second valve, or c) " valve fails open while indicating closed" failure for the first valve and excessive leakage failure of the second valve.

In the PLG treatment, the frequency of M0V valve disc leakage and failure upon demand (due to a sudden pressure loading) were assumed to be identical to that for the check valves. For the frequency of failure of a NOV to close on demand but indicate closed, a mean value of Ad = 1.1x10 4 failure / demand was used in the PLG treatment.

Applying the same approach as PLG (Formula 3.15 of PLG-0432) with the newly determined check valve leakage frequency, BNL recalculated the total (2 lines) suction side ISL frequency, VS.

The new mean frequency for the RHR suction side path is:

VS = 1.44x10 4 events /RY.

The split fraction, LR, for the fraction of VS in which the leakage past the series MOVs is greater than the capacity of the relief valves (see Figure 3.5 in Reference 2), is practically the same as in the case of the cold leg injection lines (.058) and the overpressurization frequency for the RHR suction lines (again neglecting other insignificant contributions) becomes:

(0.058)(1.44x10 ") = 8.35x10 6/RY.

l l

It is noted that for the BNL-calculated check valve leak rates, the PLG procedure of using the check valve leakage failure rates as " conservative" estimates for the leakage failure rates of the MOVs in the RHR suction lines is probably too conservative, and appropriate M0V leakage failure frequencies j

should be used if PLG redoes their analysis, i

l Furthermore, in the case of a PLG reanalysis of the RHR suction lines initiator frequency, VS, the following BNL observations should also be taken into account:

l'

a)

Inadvertent opening of the two NOVs due to-common cause failures such as improper maintenance,. malfunction of the interlock system, design error, improper tests, or testing operations.

b) Failure of the stem or other internal connections in valves equipped with limit switches or failure of a limit switch (including improper mainte-nance such as reversing indication).

4 c)

It is difficult to see why only two MOVs,have limit switches, instead of four.-

d)

It would be very useful to describe the valve -inspections that are pro-mised.each time the plant goes to cold shutdown,- or is refueled.

For example, at a plant recently investigated by NRC, Region 1, everything was tested thoroughly, but the relays for the MOVs were not inspected.

e) Considerations should be given to operating procedures and the likelihood-that the procedures will not be followed, f)

Interlock behavior.

2.1.4-Operator Actions The ability-of the Seabrook operators to diagnose, respond to, and miti-gate a Reactor Coolant System (RCS) to Residual Heat Removal (RHR) Interfacing Systems LOCA will be reviewed (without hardware reliability) in this subsection. Appropriate operator actions can mitigate the. consequences of the l

ISL sequence that result in leakage outside containment when the capacity of the RHR pump suction relief valves is exceeded and subsequent failure of the RHR pump seals results.

As discussed by PLG-0432,2 the success of these mitigative actions is dependent on the ability of the Seabrook operating staff based on their training and emergency procedures, to correctly diagnose a LOCA l

- outside containment.

The correct diagnosis may be hampered by operator i'

confusion between symptoms associated with those LOCAs inside cor.tainment j

which fill and pressurize the Pressurizer Relief Tank (PRT) by pressurizer relief or safety valve discharge flow and those associated with a RCS-RHR Interfacing Systems LOCA outside containment which also fills and pressurizes 1

the PRT via the RHR suction relief valve discharge flow.

The following is the result of a brief. BNL evaluation of operator diagno-sis and actions to mitigate the consequences of RHR pump seal failure

?

,..m.s..-,_---m

-,-,.------,_--,,,-__--_-,.--.-.._._..---_-,-__.--__m...-

_, _,.. ~ _

2-12 Interfacing Systems LOCA sequences at Seabrook and its assessment in PLG-0432,2 Section 3.1.4.3 entitled " Operator Actions and Emergency Procedures."

This evaluation was preceded by an independent and fai rly extensive familiarization preparation with the Seabrook procedures as they relate to the ISL to be studied.

This preparation was followed by observation of a series of Seabrook Simulator demonstrated accident sequences which illustrated the i

distinguishing characteristics of the LOCA outside containment and the responses expected of the Seabrook operators.

The BNL evaluation was per-formed by a former Senior Licensed Operator and Westinghouse Reactor Plant simulator Certified Engineer. A more complete evaluation of operator response would require a comprehensive Human Reliability Analysis (HRA) such as Team Enhanced Evaluation Method (TEEM)7 by a knowledgeable team of specialists providing expertise in PWR operations, PWR systems engineering and human-reliability. This team would develop a detailed task sequence analysis of the Seabrook operating staff performing the detailed tasks required to mitigate these sequences and analyze the associated human reliability of the staff response using the analysis.

2 There are three sets of operator tasks identified by PLG-0432 which are to be important to the mitigation of the sequences by the Seabrook operating staff (each with a unique Operator-Action Sequence identification number in parenthesis),namely:

Diagnose the RHR system LOCA (01)

. Isolate the RHR system-LOCA (02)

Provide makeup to the RWST (03).

i To successfully accomplish these tasks, the operating staff must follow the appropriate parts of-the following Seabrook procedures which are appli-cable to the RHR system LOCA event.

Procedure E-0 (Reactor Trip or Safety Injection), Rev. 00, dated 05/16/86.

Procedure E-1 (Loss of Reactor or Secondary Coolant), Rev. 00, dated 05/16/86.

This procedure provides guidance for long-term cooling and stabilization.

l

7...-

2-13'

. Procedure ECA-1.1 (Loss of Emergency Coolant Reci rcul ati on--ECR ),

Rev._00, dated 05/16/86. This procedure provides guidance for supply-ing adequate ECCS flow and plant stabilization.

Procedure ECA-1.2 -(LOCA Outside Containment),

Rev.

00, dated 05/16/86,_

This procedure provides _ guidance on isolating the rupture.

Please -- note that ECA-1.2, Rev 00 needs to be revised. to : ensure that valves RH-V21 and -V22, the RHR pump discharge hot leg injection cross connec-

[

tion valves, are closed prior to trying to identify and isolate a break in one of the low pressure. systems.

This need was identified by a detailed BNL review of the above four procedures.

The quantification of the three operator tasks identified by the Operator Action Sequence identification numbers 01, 02, and 03 above have been provided in PLG-0432,2 Table 3-10.

According to the accompanying discussion in Section 3.1.4.3, "These operator actions include the hardware contribution, where l

applicable, and are based on enhanced procedures and instrumentation in order.

to aid the operators in their diagnosis of the event." For each of the three operator tasks, a " base" human error probability (HEP) with a "mean" value of 0.005 has been identified as "0P".

This singular human reliability analysis HEP number is identified in PLG-04322,5 u,,.. recommended in Table 20-6 of NUREG/CR-12788,,, for fojjn,_

ing a procedure under abnormal conditions.

This human error rate is inter-preted to have a mean value of 0.005 and to be represented by a lognormal distribution range factor of 10."

Therefore, the only part of the three sets of operator tasks 01, 02, and 03 which changes their quantifications values is the hardware contribution since the human reliability quantification contribu-tion to each of these three operator tasks use the same HEP value of 0.005 with an error factor of 10.

Each HEP is based on NUREG/CR-1278,a Table 20-6 (entitled " Estimated Human Error Probability (HEP) related to failure of administrative control"), Item (4) HEP (entitled "Use written operations

{

procedures under abnormal operating conditions").

Therefore, no numerical j

differentiation is made to distinguish quantitatively among operator actions related to " diagnose," to " isolate," and to " provide." Even without using the a

7-14 Human Reliability Analysis section (4.3) of NUREG/CR-2815, more distinguish-8 able HEPs should have been selected.

The October 15, 1986 demonstration at the Seabrook Simulator with several relevant accident sequences and the abovementioned Seabrook abnormal / emergency related procedures provided (in the absence of a detailed TEEM 7 equivalent human reliability analysis performed on an actual Seabrook licensed operator shift) some reasonable assurance that a licensed Seabrook crew would adequate-ly perform the necessary, actions within the time required if the exact acci-dent sequence were programmed on the simulator.

This assurance is heightened especially since the Seabrook Training Center has recently instituted, in October 1986, a training module entitled "RHR Interface LOCA/ Student Handout,"

as part of its Requalification Training Program.

The inclusion of such a module will reinforce the importance of the RCS-RHR Interfacing Systems LOCA.

Please note that. this Figure 4 (entitled "RHR Interface LOCA Isolation Sequence") of this module confirms the need to revise Seabrook Procedure No.

ECA-1.2, Revision No. 00, dated 5/16/86 entitled "LOCA Outside Containment" to close (or verify closed) valves RH-V21 and -V22, the RHR pump discharge hot leg injection cross connection valves.

This will allow the operators to identify and isolate a break in one of the low pressure systems.

Nevertheless, there were a number of concerns raised during a plant walk-through on the same date as the simulator demonstration which the Seabrook Simulator cannot adequately address. These concerns include the following:

a) Ability of RHR pump leakage to be detected in the control room - con-cern lies with vault compartmentation design, with the Equipment Vault sump not receiving leakage promptly thereby delaying level detection input in the control room, b) Ability of RHR pump relief discharge into the PRT to be distinguish-able in the control room from the pressurizer relief and safety valve discharge - concern with the latter relief and safety valve dis-charge tailpipe temperatures.

In summary, the operator action analysis performed in PLG-0432,2 Section 3.1.4.3 appears to be superficial at besta The use of one single HEP value 8

from one table of NUREG-1278 is an example of a lack of detailed and

2-15 insufficient task analysis in assessing human performance appropriately.

A 7

detailed TEEM equivalent human reliability analysis is a far more appropriate and rigorous approach to assessing Seabrook operator actions during a RCS-RHR Interfacing systems LOCA.

Neverth'eless, the simulator demonstration empha-sized a. practical assessment of human reliability and task sequence timing.

In addition, the new training module and revised associated procedure ECA-1.2 reinforced the commitment to appropriately train the operator.

Therefore, reasonable assurance that a licensed and trained Seabrook crew would adequate-ly perform the ne' essary, actions within the timeframe required will be pro-c vided by the procedure change and training.

2.1.5 Break Location The " weakest link" of the RHR pressure boundary when subjected to acci-dental pressurization was identified by the applicant to be the RHR pump seals. A tabular listing of failure probabilities at 2250 psia showing pump seal failure probabilities ranging to 0.5 while metallic failure probabilities (piping, valves, and tubing) were 0.006 seems to support this observation.

The estimates of metallic component failure probabilities were based on:

a) accidental pressurization peak pressure limited to the initial RCS pres-sure of 2250 psia.

b) a probability of failure at the yield strength of the material to be 0.01 and the probability of failure at the ultimate strength of the material to be 0.99.

c) the characterization of the overpressurization event as a quasi static process.

d) the statement that at 2250 psia, the stresses in the limiting RHR piping are only approaching yield stresses and the heat exchanger tube and other mechanical components are at a small fraction of their respective yield stresses.

The characterization of the overpressurization event as a quasi static process with a limiting peak pressure equal to the initial RCS pressure of 2250 is based on IDCOR evaluations which have not been reviewed by BNL.

The assignment of 1% and 99% failure probabilities to the yield and ultimate

i 2-16 strengths of the material respectively is acceptable since a failure at yield is considered unlikely while a failure at ultimate is considered very likely.

The statement concerning the safety margins inherent in the RHR piping and metallic components and the basis for their calculation, including the influence of aging or time dependent effects on these safety margins was questioned during BNL's review.

Of particular concern in this regard was the capacity of the potentially corrosion-degraded or embrittled heat exchanger tubes to withstand any dynamic loads associated with the overpressurization event.

The applicant responded to our question stating that the maximum pre-dicted stresses in the RHR heat exchanger tubes due to dynamic loads were approximately 50% of the yield stress value providing a margin of greater than 50% for tube thinning due to corrosion.

They further pointed out that water chemistry is periodically sampled as part of the plant chemistry surveillance program thus minimizing the possibility of corrosive attack. Although we have not reviewed the applicant's calculations, it appears that documentation exists to support their failure probability estimates for the pressure boundary even when the effects of corrosion degradation are considered.

2.1.6 Event Tree Quantification This section summarizes the effect of observations made by BNL in previ-ous sections to the event tree quantification.

One of main problems in the quantification of various ISL scenarios related to the determination of the initiator frequencies. The other observations and questions mainly expose the overall uncertainty of the frequencies of these accident scenarios.

The effect of the change in the initiator frequencies to the plant damage states can be demonstrated if the modified initiator frequencies, VI and VS, given in Sections 2.1.3.2 and 2.1.3.3 are propagated through the corresponding event trees. Table 2.5 presents the results of the BNL requantification. The table and its notation is essentially the same as Table 3-14 of the Seabrook EPZ Study.

For convenience, in Table 2.5, the meaning of some plant damage states has been repeated.

From Table 2.6 the new value of the total core

,n

-n--.

, 17 4

damage contribution due to ISL can be determined (the sum of PDS states 8C through 1FV). This is:

CD st = 1.37X10 6 event / reactor year.

I The value obtained is much higher than the updated value (see Section 2.1.1) of the Seabrook EPZ Study.

It is much closer to the result of an earlier assessment given in the SSPSA,3 which is, CDISL = 1.8x10 6 event / reactor year.

To summarize the limited BNL review of the Seabrook ISL analysis, it is our finding that the analysis, as reviewed, is not acceptable.

BNL has shown that the check valve failure rates were underestimated resulting in over-optimistically low ISL initiator frequencies.

The BNL sensitivity study was simply an attempt to demonstrate the effect-of a more consistently calculated initiator frequency on the Seabrook analysis.

This should not be taken to mean that BNL has accepted the remainder of the Seabrook analysis because only the initiator frequencies were changed in the BNL quantification.

Having found problems with the initiators, the remainder of the model was not reviewed to an equivalent depth; although, a number of comments concerning other parts of-the analysis have also been offered.

2.2 Accidents During Shutdown and Refueling Conditions The Seabrook RMEPS2 concentrated on accidents that would occur during power operation, and did not assess the risk during non-power operation, j

Table 2.6 lists the 6 modes of plant operations as defined in the Seabrook Technical Specifications. They are listed as follows:

i Mode l

1.

Power Operation 2.

Startup 3.

Hot Standby 4.

Hot Shutdown j

5.

Cold Shutdown 6.

Refueling.

l l

i

, -,, -, - ~,., -, - -, -,.,,. - - - - - -.. -

,,-,,,,..,n,_,,,,_,v-,..__,__-.,n,,

n 2-18 i

As far as early_ releases are concerned, there are some potentially significant contributors from operation in modes.4, 5, and 6.

Seabrook Technical Specifi-cations do not address the status of containment isolation in mode 5, and.re-quire isolation in mode 6 only. during periods of fuel handling. Consequently, it is possible to have a core melt accident with the containment wide open.

Based upon Seabrook's omission of risk during shutdown from their analy-sis, BNL turned to NSAC-84 -to add perspective to the Seabrook review.

10 NSAC-84 is the only major study available that was perfor ed specifically to m

- assess the core damage frequency due to accidents during non-power operation l

at PWRs.

It is an innovative and detailed study for the Zion plant, using the plant-specific procedures and experience.

Three types of initiating events were considered:

loss-of cooling, low temperature overpressurization, and loss of coolant.

NSAC-84 results show that the dominant. core damage sequences are due to loss of the RHR system and human errors. The contribution of LOCA to core damage frequency during shutdown and refueling is approximately i

2x10 5/ calendar year.

The contribution of low temperature overpressurization is assessed to be less than 10 10/ calendar year.

The total core damage frequency during shutdown or refueling was assessed to be 1.8x10 s/ calendar year which is comparable to the frequency of core damage at Zion during power

- operations, i.e., 5.7x10 5/ reactor year (internal events only).

It was stated l

in the executive summary of NSAC-84 that "with the uncertainties involved, the j

risk of fuel damage during some period of a shutdown may be as great as the j

' risk at power."

BNL is involved in an on-going project to review NSAC-84 and investigate l

methods to improve the RHR capability of PWRs.

BNL has found that extensive l

changes to NSAC-84 are required to correct its deficiencies, and the changes tend to increase the calculated core damage frequency.

Four examples of the changes required are discussed here, j

Example 1:

NSAC-84 calculated the frequencies of station blackout during 3 types of outages:

refueling, drained maintenance, and nondrained mainte-nance.

They are 8.23x10 5, 1.96x10 6, and 2.71x10-5 per year respectively i

(Table C-27 of NSAC-84).

Given a station blackout during an outage, core r

1

,--n--.

-,.r,,

r--,_v,,..,,-,,,-,-,,_m__

,n.me,-,---_-n-,

-, - - -,. -... -, -, ~. - - - -. - -, - - - -.

6' B

~2,

damage is almost surely to result if offsite power is not recovered. NSAC-84 did not include such sequences in the list of core damage sequences (Table 6-1 of NSAC-84).

Inclusion of these sequences would have a significant impact on the calculated CDF.

Example 2:

NSAC-84 assumes that both RHR trains must fail in order to result in the initiating event of " loss of shutdown cooling."

Most of the time during shutdown or refueling,7 only one train is used and the standby train will not start automatically.

Therefore, loss of the operating train will 7

lead to loss of cooling.

NSAC-84 considers operator response to loss of j

cooling in the loss of cooling event tree, while loss of cooling is defined to be loss of both trains. This implies perfect operator response to the loss of the operating train.

Example 3:

Based on NSAC-84, the dominant cause of loss of cooling is due to j

a spurious signal that isolates the suction line from the hot leg.

It is also a dominant contributor to core damage. The evidence for such spurious signals j

is that three valve closures occurred due to unidentified causes in 27,888 hours0.0103 days <br />0.247 hours <br />0.00147 weeks <br />3.37884e-4 months <br />.

Using this evidence, strictly, the frequency can be estimated to be l

3/27,888 = 1.08x10 "/hr.

Instead of using this frequency, NSAC-84 uses a i

Bayesian approach with an inappropriate prior distribution, i.e., a distribu-tion. that applies to mechanical failure of values and does not include the L

effects from unique sources of spurious control signals.

The Bayesian updating artificially reduced the frequency by approximately an order of magnitude to 1.38x10-5/ hour.

j Example 4:

NSAC-84 analysis of low temperature overpressurization may be too optimistic.

Events such as those of Turkey Point-4 11 indicate that the i

frequency _with which a rapid pressurization occurs with the RHR system f'

isolated and the PORVs unavailable is higher than 10 3 per year.

The opera-tors have only a few minutes to respond to the event before the pressure reaches the setpoint for the safety valves. Therefore, the human error proba-bility is not going to be very small.

If the operators fail to terminate the l

overpressurization, the primary system pressure will reach the setpoint for U

the safety valves. At this pressure, the vessel rupture probability may be of

)

the order of 10 3 12 13 l

The following subsections discuss the possible initiating events from NSAC-84 which are also applicable for Seabrook.

Operational experiences and

2-20 causes of failures are provided for each type of initiating event.

Some scenarios that may lead to core damage are provided based on reports in the related area. Related safety issues are also discussed.

2.2.1 Loss of Decay Heat Removal During Shutdown or Refueling As noted previously, the RHR system is designed to remove decay heat from the primary coolant system during modes 4, 5, and 6.

Mast of the time when the system is operating, only one train is actually running; the other train is either on standby or unavailable due to test or maintenance.

If the operating train fails to continue running, the standby train will not start automatically.

Therefore, loss of the operating train leads to loss of the system and operator actions will be required to restore it.

The Office of Analysis and Evaluation of Operational Data (AE00) identified and analyzed 130 loss-of-DHR events at pWRs 14 during approximately 500 reactor years of opera-tions between 1976 and 1983.

This analysis indicates that the situation involving loss-of-DHR systems is not improving.

According to this experience base, the frequency of loss-of-DHR is estimated to be 0.25 per reactor year.

Table 2.7 lists the categories of the 130 events.

It can be seen that automa-tic closure of suction valves and inadequate RCS inventory are the two domi-nant causes of loss of DHR.

Automatic closures of suction valves were caused by spurious high pressure signals, loss of instrument bus, and human errors in calibration of pressure transmitters.

Inadequate RCS inventory was caused by human errors and inadequate vessel indications during drained-down opera-It was estimated " that approximately two-thirds of the events were tions.

human error related.

Upon loss of DHR, operators may be able to restore the failed train by i

reopening the spuriously closed suction valve, or starting the standby train.

Alternative methods for decay heat removal include use of steam generators and use of charging pumps.

Typically several hours are available before core uncovery occurs.

Therefore, it is very important that the operators must he able to recognize the loss of DHR.

In the 130 loss-of-DHR events identified in the AEOD study, the operators responded in a timely fashion, such that no serious damage resulted.

However, the duration of, loss of DHR in some cases exceeded one hour.

It was estimated " that if a loss of DHR occurs at ANO-2 i

t

F 2-21 1

(a CE plant) 36 hours4.166667e-4 days <br />0.01 hours <br />5.952381e-5 weeks <br />1.3698e-5 months <br /> after a reactor trip with the RCS in a partially drained condition, the onset of core uncovery may be 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br /> after the loss-ofDHR.

D.

C. Cook (a W plant) reported IS the results of a corresponding analysis that the onset of core uncovery would take place about one hour after the loss of I

the DHR system.

j The dominant core damage sequences in NSAC-84 represent scenarios in i

which decay heat removal is lost and the operators fail to determine that action to restore cooling is required.

For example, the accident sequence j

with the highest frequency is a sequence in which the RHR suction valve is-inadvertently closed, the operator fails to trip the RHR pumps, and also fails

}

to determine that action.to restore cooling is required.

Its. frequency is j

estimated to be 4.3x10 6 per reactor year.

Such scenarios can be postulated for Seabrook. However, quantitative assessment of the accident scenarios must 1

take into consideration the plant-specific information.

For example, operator performance is strongly affected by the indications or alarms available in the control room.

Zion does not have any alarm in the control room for inadver-tent closure of the RHR suction valves, while the Seabrook control room has an audible alarm on the video alarm system if the RHR pump is running with the j

suction valve closed.

Another difference between Zion and Seabrook is that Zion has a single drop line and Seabrook has two drop lines.

This is not expected to be a significant difference because the auto closure logic at Sea-

)

brook. will isolate both suction lines when a spurious signal is generated, l

-1.e., a single pressure transmitter provides input to the interlock logic of l

the two inner isolation valves, and a separate pressure transmitter provides input to the interlock of both outer isolation valves.

BNL performed a LER search for loss-of-DHR events due to spurious closure of suction valves at i-plants with two drop lines.

Seven events were found in approximately 40 j.

reactor years.

This indicates that the frequency of spurious closure of suction valves at plants with two drop lines is not any lower than the average frequency for all plants.

l j

In response to the NRC request for additional information, the Public Service of New Hampshire (PSNH) provided a shutdown risk analysis for Sea-p brook ' utilizing the results of NSAC-84.. Two differences between Seabrook j

and Zion were accounted for, viz., the number of hot leg suction lines, and i

i' 1

i

- - - ~.,. _. _. -

-.___,,_,...-,m

~

2-22 the support system interfaces with the RHR system.

In the analysis, credit

. was taken for the additional suction line at Seabrook, based on the statement "For spurious valve closure to cause a loss of RHR. cooling at Seabrook sta-tion, it is necessary to postulate either a. common cause event involving one valve in each suction path, or a coincidence' of a single valve closure and f

maintenance being performed on the other RHR train."

This reduced the frequency of loss of DHR by a multiplicative factor of 0.145 and the core damage frequency by approximately a factor of 2.

From the information avail-able to BNL, it is not clear that this. reduction is justified.

It is true that a single train of the RHR system is adequate for decay heat renoval.

However, the standby train is not normally ' operating and will not start automatically when the operating train becomes unavailable.

The analysis' in reference 15 assumes perfect automatic start signals for the standby train or perfect operator response to the loss of the operating train; however, it has previously been shown that. operator error is important in these sequences, and neglect of it here is inappropriate.

In fact, operator actions to restore DHR l

1s specifically modeled in the loss-of-cooling event tree.

Al so, taking credit for operator action in estimating the initiating event frequency i

represents double counting of the operator.

t The analysis of the support systems in reference 16 has not been reviewed by BNL. As was stated in the analysis, the differences in the support system f

interfaces with the RHR system are unfavorable for Seabrook.

Therefore, it was judged to be unnecessary to review this in detail given the time constraints on the BNL review.

l Two issues are related to the availability of RHR system, i.e., unre-solved safety issues A-45 and generic issue 99.

A-45 addresses the adequacy of decay heat removal systems in existing light water reactor nuclear power plants.

Generic issue 99 addresses the RHR suction line interlocks on PWRs.

BNL is currently involved in a project to investigate methods to improve the reliability of RHR systems during shutdown or refueling.

The results of the project will be used towards resolution of generic issue 99.

It is believed that these issues are applicable to Seabrook.

In particular, the PSNH analy-ts sis of shutdown risk for Seabrook is. inadequate, if not incorrect, and requires many changes.

l 2-23 t

2.2.2 Low Temperature Overpressurization L

Low temperature overpressurization may occur during shutdown as a result-of unanticipated addition of mass to the reactor coolant system, for example, inadvertent actuation of safety injection pumps, or imbalance of letdown and charging flows.

Imbalance of letdown and charging flow may be caused by f

spurious isolation of the RHR system (thus negating letdown flow) or loss of instrument air that causes the letdown flow control valve to close and the charging line flow control valve to open.

To protect the Seabrook plant against such scenarios, a low temperature overpressurization protection system is activated when the primary system is cooled down after a reactor trip. The system monitors the primary system pressure and temperature and actuates a f

main control board alarm when the pressure reaches a pre-determined fraction of the allowable pressure, and on a further increase in measured pressure, i

transmits an actuation signal to the PORVs and the PORV isolation valves.

{

Also, the safety injection pumps and one or more of the charging pumps are made inoperable during initial cooldown.

In addition to the PORVs, the relief valves in the RHR system may be available to relieve the pressure.

Each RHR l

suction line has a relief valve with 900 gpm capacity at 450 psig, and each t

RHR discharge line has a. relief valve with 20 gpm capacity at 600 psig.

How-i ever, these relief valves may be made ineffective if the RHR suction valves close automatically when the setpoint of 600 psig is reached, as was the case i

in the Turkey Point-4 events. Actually, the Seabrook Technical Specifications t

j only require either both PORVs or both RHR suction relief valves to be avail-l able.

Seabrook Technical Specifications do not address the status of the i

t j

pressure interlocks on the RHR suction valves when the PORVs are not avail-l able.

The PSNH comments" on the BNL draft version of this report claim that j

power to the RHR suction valves is removed when the RHR system is aligned for DHR.

BNL has not been given any documentation that would allow independent verification of this claim and in fact BNL does have documentation of the NRC disallowing a request by another utility to remove power to their RHR valves

- under the same circunstances.

The BNL response to some of the PSNH comments is provided in Section 2.2.4.

t l

Two generic issues are related to the subject of low temperature over-pressurization, generic issues 94 and 70 Generic issue 94 considers t

i

2-24 I

additional low-temperature-overpressurization protection ~ for light water reac ;

-tors.

It has a "high" priority ranking.13 Enclosure 1 to reference 13 is the prioritization evaluation -for the issue. -It was stated in the evaluation that before 1979 30 events in PWRs were reported where the pressure /tenperature.of the reactor. coolant system violated Technical Specifications.

After 1979, i

l following changes to operating procedures and the implementation of overpres-surization mitigation systems, there have been two reported events of over.

pressure excursion events,.i.e., the Turkey ' Point-4 events.

Based on the operational experience and the use of the Vessel Integrity Simulation Analysis (VISA). code,1s the prioritization evaluation estimated that the core damage frequency due to vessel rupture in a low-temperature-overpressurization event j

at Oconee 3 to be 4.5x10 6 per reactor year.

Generic issue 70 considers the l

reliability of PORVs.and their block valves.-

BNL is currently investigating the issue, and a draft report of the work is upcoming.

2.2.3 Loss of Coolant Accidents During Shutdown or Refueling NSAC-52" reviewed operating experience within' 5 calendar years up to the end of 1981, and ' identified 10 loss of coolant events at PWRs. They were i

caused by the following causes:

j 1.

Inadvertent manual _ initiation of RHRS supplied containment spray.

j

2.. Inadvertent loss of inventory to the containment building sump and/or f

automatic initiation of recirculation mode of low pressure safety I

injection.

I

(

3.

Inadvertent loss of inventory via the RHRS relief valves.

4.

Inadvertent loss of inventory via mispositioned crossconnect or drain i

valves.

(

5.

RHRS valve packing gland removal during plant pressurization,

{

dislodging the valve packing and gland, i

6.

Gross valve packing leakage.

As for the loss-of-cooling initiating event, LOCA during shutdown or

)

refueling requires operator response to terminate the inventory loss and to j

provide inventory make up.

The NSAC-84 analysis for Zion assessed the core l

damage frequency due to a LOCA at shutdown or fueling to be approximately 2x10 8 per reactor year.

The dominant scenario is that the operator fails to I

2-25 close the RHR return valve to the RWST after draining the cavity, on reestab-lishing RHR flow, a LOCA via the RWST vent outside the plant occurs, and the operator fails to respond to it.

BNL does not have the Seabrook procedures used during shutdown, and therefore cannot judge if the same sequence is applicable to Seabrook. However, the LOCA experience identified in NSAC-52 is applicable to Seabrook.

2.2.4 PSNH Comments on BNL Draft Report 16 PSNH provided an analysis that makes use of NSAC-84 in response to BNL's finding that risk at shutdown had not been addressed.

PSNH also provided some comments l7 on sections 2.2.1 to 2.2.3 of the draft version of this report.

This section simply provides a response to some of the PSNH input with the intent to clarify some misquoting or misinterpretation of the

[

BNL draft report by PSNH.

1.

The PSNH analysis borrowed sequences from NSAC-84 and inherited the prob-lems and mistakes of NSAC-84 such as those outlined in this report (begin-ning of Section 2.2).

2.

PSNH states "As noted by BNL, operator failure to restore the standby train was not explicitly included in the Seabrook analysis."

BNL did not make such a statement.

On the contrary, BNL believes that the Seabrook 18 analysis erroneously assumes perfect operator response to loss of the operating train of the RHR system.

3.

The PSNH analysis of spurious valve closure indicates a misunderstanding of NSAC-84.

Operator response to loss of cooling due to spurious valve closure is explicitly modelled in the loss of cooling event tree of NSAC-84.

For example, top event RT represents operator tripping the operating pump, top event RH represents reopening RHR suction valves, and restarting at least one RHR pump.

The PSNH analysis l7 first considers operator response to spurious valve closure to reduce the frequency of loss of DHR.

It then uses the loss of cooling event tree in NSAC-84 that considers operator restoration of RHR again to calculate the core damage frequency.

This way the core damage frequency is artificially reduced by taking credit for operator action twice.

2-26 4.

The PSNH response to the draft version of this report provided -a statement concerning procedural removal of power to the RHR suction' valves.

Such procedures have not-been made available to BNL.

It'is known to BNL that Diablo Canyon requested NRC permission to similarly remove power, and the NRC staff reviewed the RHR isolation valve operating procedures and found that the licensee should retain power available to the MOVs when the RHR system ' is in operation.20 The concern is that O power is removed from the RHR MOVs to remove the possibility 'of an inadvertent closure, then no ready means would be available to isolate the RHR system should it rupture or develop a leak outside containment.

Assuming such procedures are used at Seabrook, the whole shutdown risk. analysis will have to be redone, because spurious isolation of the suction valves is the dominant contribu-tor to core damage, and the - PSNH analysis does not reflect' the fact that power is removed.

17 5.

PSNH provided some discussion on the consequences of accidents during shutdown.

BNL believes that it is not very meaningful to consider the

. consequences before the analysis on core ; damage scenarios ~ is in a good shape. The calculation 17 on the mean time of scenario initiation is cor-rect, assuming that the distribution of the time of failure is uniform.

However, the distribution of the time of failure is not uniform.

A more accurate way to calculate the mean time of scenario intiation is to esti-mate the times of scenario initiation for all core damage scenarios and use the frequencies of the core damage scenarios as the weights to calcu-late the weighted average of the times of scenario initiation.

2.2.5 Summary of the Shutdown Risk Review

-The Seabrook analysis did not originally address shutdown risk.

In response to BNL questions forwarded by the NRC, Seabrook provided a shutdown risk analysis based upon selected modifications to NSAC-84.

BNL was well along on its independent review of NSAC-84 as part of a separate project when the Seabrook response was received. Based upon the scope and direction of the BNL review as outlined at the beginning of Section 2.0, BNL concludes that the shutdown risk assessment by Seabrook is unacceptable as presently documented.

The bases for this conclusion are detailed in the preceding subsections and in

2-27 sunmary are that NSAC-84 has a number of deficiencies and that the modifica-tion of NSAC-84 to represent Seabrook also has a number'of, deficiencies.

2.3 Induced Steam Generator Tube Rupture (SGTR)

For accidents in which the primary system is at high pressure during core uncovery and melting, it is possible that large natural circulation flow pat-terns could develop within the primary system.

These flow patterns could in turn heat up regions of the primary system remote from the reactor core.

As the primary system heats-up, it is possible that parts of the pressure boun-dary could degrade. Of particular. concern is the possibility of degrading the steam generator tubes such that the primary system could become open to the secondary system.

If the secondary system were in turn open to atmosphere, then a direct path could exist between the primary system and the atmosphere, which bypasses containment.

This is a very important topic for review because it could potentially lead to a relatively large early release of radioactivity.

The topic was not included as part of the work scope for the current BNL review.

However, the topic was reviewed 21 by the NRC staff and is the subject of continuing NRC and industry research activities.

Scoping studies were performed to assess the impact of induced steam generator tube rupture on risk at Seabrook. First, the frequency of accidents in which the primary system would be at high pressure had to be determined.

The applicant estimated 22 the frequency of high pressure sequences in which a SGTR might have an effect to be 4x10 5 per reactor year.

The NRC review 21 4

considered this estimate to be significant and therefore concluded that i t needed further consideration.

It was used as the basis for the BNL scoping study.

Given that core meltdown occurs with the primary system at high pressure, i

the probability that the steam generator tubes will fail had then to be deter-mined.

In addition, it is also possible (provided methods are available) for the operators to depressurize the primary systen prior to induced failure of i

i

,_.,..---.-.m,.--,-._,,,c.__.-.-.--w,.o-,,n.,,_,,.,

-.,,,,,.y___,.--,--,.,,,,,._,,.----..,e--,.

._,-..,7..w-

__e-2-28 the SGT.. The probability of successful depressurization had also to be deter-

~

. mined.

T Estimating the probabilities of the above events is subject to signifi-cant uncertainty.

However, the Severe Accident Risk Reduction Program at SNL attempted to quantify these probabilities by. use of expert judgment.

The probabilities were developed specifically for the Surry plant and reported in.

Appendix B of NUREG/CR 4551 Volume 2.23 The experts concluded that there was a conditional probability of 0.8 for the operators to successfully depres-p surize the primary system.

The BNL review team considers the 0.8 probability l

rather optimistic given that procedures do.not exist for this operation and therefore consider the conditional probability' to be indeterminant (CP = 0.5) without further analysis.

In addition, the experts felt that the probability of an induced steam generator tube rupture might be between 0.01 and 0.1 (for 4

both small and large tube ruptures) conditional on no depressurization. These i

estimates are reasonably consistent with an earlier NRC memorandum " on this 2

i subject, which suggested a conditional probability of about 0.01 to 0.3 for

{

SGTR given a high pressure core meltdown.

It was therefore decided by BNL to t

use a range of 0.2 to 0.5 for the conditional probability of failure by the operators to depressurize and a - range of 0.0'1 to ' O.3 for induced SGTR to l

assess the impact of this phenomenon on risk at Seabrook.

The results are i

summarized in Section 2.5.

i 2.4 Containment Isolation Failure i

l The applicant's assessment of pre-existing leakage and containment isola-l tion failure was reviewed s by the NRC staff.

Based on its review of the 2

j information available, the staff concluded that the purge and vent valves in a

)

fully closed configuration should provide reliable isolation of the Seahrook l

containment under severe accident conditions up to the pressure corresponding l

to 1 percent hoop strain in the containment.

)

The staff also concluded that the applicant has presented a reasonable approach for the consideration of pre-existing leaks, both small and large, j

The approach adopted by the applicant was to use information on containment l8 26 unavailsbility developed in a study by the Pacific Northwest Laboratory

- ~

2-29 (PNL) to assess the impact on risk of pre-existing leakage.

The applicant used this information to bound the effects of the data in the PNL study2s (NtlREG/CR-4220) even though they considered that it did not apply to Seabrook.

2.5 Summary In this section, the BNL findings related to each review area are briefly summarized.

Sensitivity studies have been performed using the applicant's conditional risk indices to show how the dose vs distance and risk profiles might change as a result of the concerns raised in this section.

The summa-ries included a discussion on the uncertainties associated with the risk estimates.

Interfacing System LOCA A major concern resulting from the BNL review of the interfacing systems LOCA analysis in PLG-0465 and the RMEPS related to the determination of initiator frequencies.

The effect of changing the initiator frequencies was determined by propagating the changes through the appropriate event trees in the RMEPS.

The revised initiator frequencies resulted in the following changes to the frequencies of release categories S1W and S7W.

Mean Frequency Per Reactor Year Release Category PLG-0465 BNL Review S1W 4.0x10-8 1.4x10-7 S7W 6.3x10 8 1.1x10 6 The above changes in release category frequencies have no impact on indi-vidual risk of early fatalities within 1 mile of the site boundary if no evacuation or 1 mile evacuation is assumed.

This is because release category S2W dominates this risk measure, and it has a frequency of 2x10-5 Only when a 2 mile evacuation is assumed (and the early fatality risk for category S2W becomes zero) do the above changes in release category frequencies change the original PLG-0465 estimates.

However, wi,th a 2 mile evacuation the early fatality risk is very low and well below the safety goal.

The 200-rem dose

.2-30 versus distance curve given in PLG-0465 is also not influenced by the above changes in release category frequency.

This is because only release category l

S1W has a significant probability of exceeding a 200-rem dose, and the revised I

probability of this category is not sufficiently high for it to influence the 200-rem dose versus distance curve in PLG-0465.

4 There is of course uncertainty associated with predicting the frequency of interfacing systems LOCAs.

However, the frequency of interfacing systems LOCAs resulting in release category S1W would have to' increase by two orders of magnitude before the Seabrook dose versus distance curves would approach the curves given in NUREG-0396.

One can therefore conclude that interfacing systems LOCA are unlikely to increase the risk profiles presented in PLG-0465 to the level presented in NUREG-0396. This is not too surprising because when no evacuation is assumed, the higher frequency events dominate risk and interfacing systems LOCAs did not contribute to the dose versus distance curves constructed in NUREG-0396.27 Accidents During Shutdown i

This topic was not originally addressed in PLG-0465 and a detailed assessment of such events is beyond the scope of the current BNL work on this project.

However, the applicant was requested to provide information on the risk associated with accidents during shutdown.

The results of the appli-cant's assessment of such accidents were presented in the form of sensitivity studies in a draft version of this report.

The applicant provided additional frequencies to the existing release category frequencies to assess the impact on risk from accidents during shutdown.

A base case and a bounding case were presented by the applicant.

The additional frequencies associated with these accidents are given below:

l f

Mean Frequency Per Reactor Year Release Category f

Power Base Case Bounding j

Operation Events Shutdown Events Shutdown Events S.5 1.1x10

1.7x10 5 S.2 2.1x10-5 4.9x10-7 S.6 6.5x10 7 7.1x10-8 5x10 6 L

2-31 c

v

~

BNL was not in a position tu assess the above frequencies for these events because there remained fundamental questions regarding the mo'daling of these scenarios.

However, in spite' of this, the applicant's resuits were included in the BNL draft report for comparison with the sensitivity study results on other topics.

It should be noted that the applicant considered the upper bound estimates to be very conservative.

In p' articular, in order to assess the impact of these events, they were included in source term cate s gories derived for accidents from full power, which could lead to predicts of shorter times and larger quantities of fission product release than,would be expected from accidents during shutdown.

U In a subsequent submittal by the applicant, the consequences of acci-i dents from shutdown were revised., The. applicant felt that 94 percent of acci--

dentsat'shutdownwouldoccuratifmles'laterthan48hoursafter. scram. Thus, the cor,equence estimates were reanalyzed assuming release times of 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />, The later release times resulted in dose versus distance curves which fall off at much shorter distances from the site boundary than the original dose ve sus distance curves.

BNL has checked this result and confirmed that if,.the m

release does occur at times greater than 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />, then the new dose versus distance curves are reasonable.

The,results of the latest applicant's assessment of accidents at shutdown are reproduced in Figure 2.2.

As noted above, a detailed assessment of such events is beyond the scope of the current BNL review.

However, based on our limited review of the applicant's asscrssment of these events, we still have restryations about the results.

These reservations are discussed in greater detail in Section 2.2, but until they are resolved, we are unable to assess the. validity of the risk estimates presented by the applicant in Figure 2.2.

Induced Steam Generator Tube Rupture

~

l In Section 2.3 a sensitivity study was suggested to assess the impact of induced SGTR on risk at Seabrook.

The frequency of high pressur' e sequences taken together with the conditional probabilities of failure to depressurize l

r s

__..__-__r.-

m

..-,- ~

=P 2-32

+

g. gad induced SGTR given in Section 2.3 give the following range of probabili-7 g I},ItiesofinducedSGTR:

y3 x

~

(

, p.0/10,5 x 0.5 x 0.3 = 6.0x10 s per reactor year

't.0x10 5 x 0.2 x 0.01 = 8.0x10 s per reactor year.

g In order to estkate the impact of the above probabilities on risk, an appropriate source term category had to be selected.

It was decided to' allo-cate SGTR events -to release category S1W which represents a large early I

i bypass of the c9ntainment.

It was felt that this was a conservative assump-s tion because, significant retention of the fission products in the secondary side could occur and this was not considered when calculating the S1W release fraction'. The impact of adding the above frequencies to source term category o

(

)

S1W is illustrated.in Figure 2.3.

S l

(

The lower estiniate of the frequency of ir uced SGTR has no impact on the l

risk estimates presented in PLG-0465. The higher estimate of the frequency of l

induced SGTR has no influence on the individual risk of early fatalities with-in 1 mile of the site boundary if no evacuation is assumed but does influence the 200-rem dose versus distance curves as shown in Figure 2.3.

Allocating the probabilities of SGTR events to release category S1W has the largest

. impact on the ilose versus distance curves (refer to Figure 1.4).

However, the impact on the risk of early fatalities within 1 mile is negligible because S1W j

has yery little risk of fatalities within this distance (refer to Table 1.2).

If the probabilities of SGTR events were added to release category S6W, the r

l impcct on the dose versus distance curves would be less but the risk of fatal-ftty, within 1 mile would increase slightly if no evacuation is assumed.

It should be noted that the range of frequencies used for the induced l

SGTR sensitivity study were developed to cover our lack of understanding in i

this area and that the NRC staff believes that the actual probability of a i

fp<

SGTR is cigser to the lower estimate.

However, one reviewer 2s of the BNL

{,,

draf t report felt SGTR to be a potentially more "significant" issue that was implied in our evalvation.

It was not BNL's intention in the draft report to minimize the potential importance of this. issue, and the range we presented s

.did not represent an upper bound.

It was an attempt to reflect the best

,.-...-.r._...

, ~ _ _,.. -.

,__,________.-___.-_.,,,m.,.m n

.y_ _ _.

m-

_.m--,. -. - -

mm-.m,.

~

2-33 judgments of several experts on a very difficult subject.

There is a great deal of uncertainty associated with predicting such events, it is prudent to indicate the impact on risk of a range of assumptions.

1 Containment Isolation Failure and Pre /xisting Leakage l

This issue addresses the possibility that containment may not be isolated 7

during or immediately following aq accident.

This area has not been reviewed in detail by BNL or the NRC Staff. The NRC staff concluded that the purge and vent valves in a fully closed position should provide reliable isolation under severe accident conditions.

Estimates made by the Applicant using generic data for containment isolation failure (NUREG/CR-4220) are shown in Figure 2.4 and indicate that this issue has a small impact on risk. BNL has not assessed the validity of the applicant's risk estimates for isolation failures.

2.6 References 1.

"Seabrook Station Emergency Planning Sensitivity Study," PLG-0465, April 1986.

2.

"Seabrook Station Risk Management and Emergency Planning Study," PLG-0432, December 1985.

3.

"Seabrook Station Probabilistic Safety Assessment," PLG-0300, December 1983.

4 "MAAP-Modular Accident Analysis Program Users Manual," Technical Report on IDCOR Tasks 16.2 and 16.3, May 1983.

5.

" Nuclear Power Experience (NPE)," S.

M.

Stoller Corporation. updated monthly.

6.

Hubble, W. H. and Miller, C., "Date Summaries of Licensee Event Reports of Yalves at U.S. Commercial Nuclear Power Plants," NUREG/CR-1363, June 1980 7.

" Team-Enhanced Evaluation Method (TEEM) Procedures--An Enhanced Human Reliability Analysis Process," Informal Report BNL-38585, Rev. 1 Decem-ber 1986.

8.

" Handbook of Human Reliability Analysis With Emphasis on Nuclear Power i

Plant Applications," Final' Report NUREG/CR-1278, August 1983

2-34 9.

"Probabilistic Safety Analysis Procedures Guide," NUREG/CR-2815, Rev.1, Vol. 1, August 1985.

10.

" Zion Nuclear Plant Residual Heat Removal PRA," NSAC-84, July 1985.

11.

"Overpressurization of Reactor Coolant System," IE Information Notice No. 82-17, U.S. NRC, Juna 11, 1982.

12. Burus, T.

J.,

et al.,

"Pressu rized Thermal Shock Evaluation of the Oconee-1 Nuclear Power Plant," NUREG/CR-3770, Draft, April 1984 13.

NRC Memorandum from H.

R.

Denton, Di rector, Office of NRR, to R.

M.

Bernero, Director, Division of System Integration, on the Subject of Schedule for Resolving and Completing Generic Issue No. 94, July 23,

1985,

14. Ornstein, H.,

" Decay Heat Removal Problems at U.S.

Pressurized Water Reactors," Office for Analysis and Evaluation of Operational Data U.S.

NRC, December 1985.

15.

Indiana and Michigan Electric Company, Licensee Event Report (LER) 50-316/84-014, D. C. Cook Unit 2, dated June 22, 1984.

16.

PSNH Letter (SBN-1225), dated October 31,1986, " Response to Request for Additional Information (RAls)," J. DeVincentis to S. M. Long 17.

PSNH Letter (NYN-87-002), dated January 20, 1987, " Comments on Draf t Report, G. S. Thomas to V. Nerses.

18.

" VISA--A Computer Code for Predicting the Probability of Reactor Vessel Failure," NUREG/CR-3384, September 1983.

19. Residual Heat Removal Experience Review and Safety Analysis, NSAC-52, January 1983.

20. NRC Memorandum from B. W. Sheron to Reactor Systems Branch members, " Auto Closure Interlocks for PWR Residual Heat Removal Systems," January 28, 1985.

21.

NRC Nemorandum from W. C. Lyon to C. E. Rossi, " Steam Generator Tube Rupture During Severe Accidents at Seabrook Station," March 3,1987.

22.

PSNH Letter (SBN-1237), dated November 21, 1986, " Emergency Planning Sensitivity Study," J. DeVincentis to S. M. Long.

23.

" Evaluation of Severe Accident Risks and Potential for Risk Reduction,"

NUREG/CR-4551, Volume 2, February 1986.

24 NRC Memorandum from B. W. Sheron to D. B. Liaw, " Steam Generator Tube Response duri g Severe Accidents," February 14, 1985.

l i

2-35

25. NRC Memorandum from C. E. Rossi to V. A. Noonan, "Seabrook Emergency Planning Study--Treatment of Pre-existing Leaks in Containment," Feb-ruary 9, '1987.

26. Pelto, P.

J.,

et al.,

" Reliability Analysis of Containment Isolation Systems," NUREG/CR-4220, PNL-5432, June 1985.

27. Collins, H. E., et al., " Planning Basis for the Development of State and Local Government Radiological Emergency Response Plans in Support of Light Water Nuclear Power Pl ant s," prepared for the U.S.

Nuclear Regulatory Commission, NUREG-0396, December, 1978.

. 28. Theofanous, T.. G.,

" Review Comments," University of California, Santa Barbara, dated January 12, 1987.

4 I

f

2-36 t=10

4 a 6

~

64466) iaslag i6 6

e 6 e

4*e.

lx10 i i i sit j 4

e N

s g

5 g

4 g

N Ss N

a N 'N s

N

\\

M 2

\\

g N

\\'

N s

1 = to,

\\

I lx10~

g.

w-g g

.6

~ 'N, \\' x.

N N

g o

N

\\

N 3

s s

N 2

N

'go

\\

s

\\

4 s

N

@ 1 ' 80 s

- lx10,

9

\\

N m

~

g N

y a-s U

\\

N

\\.

s-N N

\\

\\

4 y

\\

3

\\

\\.

=

, BN.

1 x 10'$

\\

~

lx10 L

ND:

\\

N N.

~

ppg 9g74 6

p 5

-- -- - sT4TisTICAL SOUNDS

\\

4 AT t,.4 CONFIDENCE

\\

\\.

3 BEST FIT

\\

- - ASSUMED 957H AND STH PERCENTILES

\\

2

\\

Experience, events selected by PLG

\\

e 1 s to-to

\\

\\

q

-9 lx10 N ".

BNL DATA N

s 5

Graphical Fit 4

+

Experience, events selected by BNL 3

~

2

,,,,.it

.,i..1

....ieiI

....,.,t

-10 lx10 1

10 100 1.000 10,0C0 CHECK VALVE LEAK RATE IGPM)

Figure 2.1 Frequency of accumulator che'ck valve leakage events.

2-37 1

i i

i i

i i i i i i i, iiiig

~

EE

~

>O

~

8" TU

-l 2E w

• 'I T o

NUREG-0396 7

u f

o<

g

/

e.

oz ri wy h5 we

~

ciE

~

>5

\\.-- Original 4

• y m

PLG-0465 results

~

gw Kh o

s\\ \\

E E 001 4

\\

BS g

g{,

Shutdown events assuming early release, PSNH letter (SBN-1225)

EE li

\\

~

Og

[\\

\\

- Shutdown events assuming 48 hour5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> lg i

release, PSNH letter (NYN-87-002)

\\

4.

lg 200 REM g

g l

'\\

I

'I 0.001 i

1 10 100 1,000 DISTANCE (MILES) i Figure 2.2 Comparison of 200 rem-dose versus distance curves for conservative assumption of no credit for operator recovery of open equipment hatch l

(calculations performed by PSNH).

I

2-38' 1

iieiit i

i a

...ig E,$

[

Ee

>d

~

8*

ab

$E ow E8 o'

u e<

NUREG-0396 z

55

/

DE

~

o i

x

=d

u. u of N

EE

~\\

PLG-0465

=0

\\/

88

\\

g g 0.01 g

to

-ps l

's

<h N

Ea 1

\\

og s

gr l

s N

og I

Impact of pessimistic assumptions gp'regarding SGTR (optimistic assump-l g

tion off graph)

I

I

I 0.001

~

1-to 1oo 1,000 DISTANCE (MILES)

Figure 2.3 Comparison of BNL sensitivity studies with PLG-0465 and NUREG-0396.

(200-rem plots with no immediate protective actions.)

2-39 1

'+ieii i

E$

$9

~

>8 8*

gn 55 ow

[g o.1 ou NUREG-0396 3:

/

s EG we

~

oE

>i b 4 0riginal

,Ey$

PLG-0465 results 8S g

g g o.01 g

in. o

\\

am h

'h interpretation by PSNH of l

Effects of conservative

$w NUREG/CR-4220 data,.PSNH 9

l\\

l letter.

5

~

l l

t l

I

~il j

o.oot i

i e i e i v il 1

10 100 1.000 OlSTANCE (MILES)

Figure 2.4 Comparison of 200 rem-dose versus ~ distance curves for conservative interpretation by PSNH of flVREG/CR-4220 data.

(Calculations performed by PSNH.)

Table 2.1 Somery of Operetlag Ev ats, toergeasy Core omettas $fstem. Isolatloa Check valves. Leekage Fall re aede Beweer of Check (stimated lief erenc e ECCS wasvos teek state (et il Plant Date Systen Event Description Reported (go=3 Iloperts Vll.A.15 Pallsados S/72 ACC Leekage late SI tank. The latermels of a Check velve on the outlet of an $1 tank 1

p3 loote 1 ees lacorrectly essembled.

Vlf.A.2S lease 12/72 ACC Lookage late Si teek. A emell place of sold slag had lodged vader the seet of the 1

y$

esote 1 Yeahoe evtlet chack velve alleelag back leekage. Olletten 1700 ppm (flelt Is 3720 peo).

Tif.A.32 Turkey S/73 WI One of tne three check valves le the St Ilmes developed a geenage of 1/3 goe.

3 y,. 33 Polet Tuo other check valves showed saly slight leakage. Failure of sof t seats.

Vll.A 63 Ginee 9/74 ACC Leekage of a check velve caused boron dilutlee la ACC. *A* (from 2250 ppm to I

y20 Isoto 1 1617 peal.

Vill.A.g5 Svery 1 4/73 ACC Check volve did met seet. ACC ("lC") level Increesed. Leekage rates "6 gem.

1 y10 Isote I ru a

vit.A.126 Zlon 2 10/75 ACC -

urong s tre Sasket testelled in the check velve_ for ACC. "A". Leak rates e,25 gpm.

1 5 2S leste 1 Jh Q

VII.A.10S flottosen 2 1/ 76 ACE Accasnelator ("s') Onleekage through leeking outlet check volve.

5 y20 Isote 1 v.A.122 Zion 1

$/76 ACC latestege to ACC. *lD* from flCS.

I y20 somte 1 Vll.A.114 Sorry 1 7/76 ACC Two check valves in series (1-51-129,1303 feeked covelag torce dIletion le 2 in series y10 leste 1 ACC. *S".

Vlf.A.120 Sorry 2 S/74 ACC goron dilution (free 1950 ppe to 19933 le SI ACC. "C" ceased by leeking check 2 le series y10 lestes 1 velves (2-58-145,147).

and 2 vil.A.225 elllistone 2 4/77 ACC laleakege of RC through outlet check valves to Si teak *4*. Lov boren 2 la series y20 Inste 2 concentratten. Five accurrena le 1977 Til.A.17S San S/78 LPI Tiltleg dish check velve (first valve leside coatelament) failed to close with enofra 1 gravity lastelled la e vertical rather then a horlaoatal pipellee.

1 yS Vill.A.142 Calvert 9/79 ACC Outlet check valves for $1 tanks 218 and 225 leaked. Baron concentration redoc-2 y10 lente 1 Cllifs 2 ties free 1724 and 1731 ppm to 1652 and 1S94 ppm le one month period, pIO e

e Table 2.1 Continued on.ner et check tstleated notere.c o scrs Valves team Rate INPE H Plant Date

$vstes Event Descriettee Reported igan)

Rearts Vll.A.262 Crystal 7/80 ACC Check volve CFV-79 to core fleed tea. failed. The Isolation valve to the N 1

100cy Notes 1 g

River 3 systen was open for Ng mislag. 600 gallon ligold entered the N2 sys en and 200 and 2

• 00 gallons oss rolessed. The corresponding Geflvlty rele8 sed estimated as 1.07 aCl.

Vlf.A.273 Deves 10/40 ACC Imat system Isoletlan check volve CF-30 leaked back sucessively. Velve 1

50sy100 slote 1 lE lato.

Besse 1 disk and era had esperated from the volve body. Bolts end locklag mechselen Notic e were alssing. Cc,re fleed tank overpressurized.

80-41 Vll.A.291 Sorry 2 1/81 ACC Accumulator l'C') baron diluted. Check velve 11-51-144) feeked. Fleshlag systen 1

y10 Note 1 Improperly set up, resulting la charging system pressure to exist en the downstream side of the check velve.

Vll.A.301 Palls des 3/01 ACC Leekage of RC late the $1 tank if-4231.

1 y5 Notes 1 N

M2

[

w VII.A.306 ItGuire 1 4/81 ACC Accussletor *A* outlet check valves Iso-159 and IN-160 were leeking. RCS pressures 2 in y10 Note 1 1800 pstg. Acc. pressure: 425 psig. Water level above eieria setpoint.

' series iril.A.307 stCulre 1 4/01 ACC Steller events alth Aces. "C* end *0".

2=2 la yc10 Note *

• serlos Ne sad 2 Vil.A.343 Point 10/81 LPI RCS/LPI leolation check velve 11-853C) leeks In escess of acceptance criterle 1

yt10 Beach 1 1 *6 gen).

VII.A.384 Calvert 7/82 A(X:

4:c. outlet che:k velve et Unit 1 f eeked due to deterioration of the disk seelleg 1

E200 Note 1 Cllfis o-ring. The e-rlag meterial hos been changed en all check valves of Unit I and 2 1&2 1/2 58-215, 229, 235, and 245 V l l. A.40 5 Sorry 2 9/82 ACC Acc. outlet check velve 12-51-1441 leaked ACS water into taak "C* during a pipe 1

y20 Notes 1 flesh resettlag in few baron concentration.

and 2 Vll.A.396 Pellsedes 9-12/

ACC Ninor leakage f ato si tent icompounded by level ladication f allures via check 1

yS Notes 1 82 valve leenages, and 2

,Toble 2.1 Coatinued leaner et Check.

Estl=eted Re f erenc e tcc5 valves teen Rate INPE il Plent Date System Event Deecrestsee Reported Igpol Romeras vil.A.407 N:Guire 1 5/83 ACC RCS weier Inleekage through outlet check velves IN-170 and l>171 resultlag 2 In series 20c y50 Note 1 la low boron eencontratlee la CLA 9".

V l l. A.4 37 Farley 2 g/83 WB

$1 Check volve to leap 3 cold leg mes excessively festleg, incomplete coatect 1

50cp100

..t.o.n,he v.lvo disk.no s.ee.

LER 84-001 Cconee 1 3/84 ACC Accepuletor ("A*3 Inleekage through leeklag volves. Adelaistrative deficiency, 2 la series y5 Note I no management control over e kneen pretten Islace 8/833 V.F.0043 Pellsades 7/94 ACC k.cumulater laleshage th:M4 looking check velves CK-3146 and CK-3116 2

y5 Notes j LER 84412 ene 2 v i l. A.4 52 St. Luc ie 12/84 ACC lateekage to Si tank. Seel plate cocked, volve seet consensattag jolat ball 1

20cp50 Note 1 2

galled.

gg 8

Vll.A.456 Calvert 1/05 ACC '

faleenage to safety injection teaks through Check velve, o=fing meterial 2

p5 Note 1 gy Cliffs degradeflon (mitt t = 1.6 goe, Unit 2 = 27.2 gaml.

20c y50 1&2 vil.4.457 N:Guire 1 4/85 ACC Lou accumule*ar borea concentratloa.

t y5 Note 1 trR 85-007 Pellsades 6/05 ACC lateekage from the RCS via e check velve. Lou level boroa concentration.

1 y5 Note 1 vit.A.474 Pellsedes 11/85 ACC A:cemulater (SIT-620) Infeekage from IICS throu8h a check velve. CK-3116.

1 y5 Note 3 Soroa dilution.

Note 1: Estimated leek rete is the resultant one through two Check valves in series.

Note 23 Not listed la Table 3.8 of PLG-0432.

Note 3: The Pellsedes salt has a chronic accueulever laleakage pr:49en.

1 e

e fable 2.2 Summery of Operating Events. Eecrgency Core Coeling Sy7en. Isoletion Check volves. "Fallure to liectose tipon Desondo Fellere ebee me6er of (beck i

lieference ECC$

ygg,gg i

(MPE di Plant Date Systen Event Description beorted Romerts j

vil.A.270 Sequoyen 1 9/80 sel

$1 check velve 63-63$ was found to be stuck ocee, it ses caused by l

8este I laterference between the disk met lockulre tack weld and the velve body.

i Val.A.20S Salem I 12/00 eel SI check vs.ve felled to close during a test. It Is en Interf ace between RCS S

Note I hot leg and $1 pings. Velve v3s found to be locked open due to beren solldtf f ce=

tion during the last ref ueling.

1 Vll.A.294 Oconee 1 2/88 Lpl Anector vessel LPI leap P lseletlen velve (IEF-12) leaked excessively during I.

Note l I

LOCA leek test. The velve elsk had h trogen et the plvet le a cocked posittm. Bulldup of deposit la the gap between the hinge and disc knob caused the free:Ing.

~

Vll.A.302 Oconee 3 3/09 LPI

$leller to event et unit I (welve involved Is 3 CF-13).

I peote I vil.A.390 8ecGuire 8 S/OS ACC Leek test deseged ecc. check velvee = seat type changed.

2 Note $

f 46 vil.A.391 tecGuire 1 3/01 ACC Acc. check velves felled.

2 Note j vil.A.313 Point 7/01 LPI flC$/LPI lealetion check valves I-853 C and D sere teund to be stuch in the full 2

Note 9 j

Beach I open position. High leakage rate.

51 Isoletten check velves 2 St.13C and 2' SG-12 stock in the open positlen during 2.

Note 1 vil.A.392 AO2 10/02 tel test reguested by lE seatice 81-30. Olsk stod protruded stove nut, disk elsaligned.

8 ente la foot listed la Table 3.8 et PLIN>d32 1

4 i

4 l

2-44 Table 2.3 Accumulator Check Valve Exposure Data Start of Number of Total Number of Commercial Number of Accumulator Check Valve i

Plant Name Operation Years Check Valves (10 Hours) 5 Arkanssas Nuclear One 1 December 1974 11.08 4

3.882 Crystal Itiver 3 March 1977 8.83 4

3.094 -

Davis-tesse 1 November 1977 8.16 4

2.859-Oconee 1 July 1973 12.50 4

4.380-Oconee 2 March 1974 11.83 4

4.145 Otonee 3 December 1974 11.08 4

3.882 Rancho Seco April 1975 10.75 4

3.767 Three Mile Island 1 September 1974 11.33 4

3.970 Three Mile Island 2 December 1978 7.08 4

2.481 Arkanese Nuclear one 2 March 1980 5.83 8

4.086 Calvert Cliffa 1 May 1975 10.67 8

7.478 Calvert Cliffs 2 April 1977.

8.75 8

6.132 Fort Calhoun September 1973 12.33 8

8.641 Milletone 2 December 1975 10.08 8

7.064 Maine Yankee December 1972 13.08 6

6.875 Palisades December 1971 14.08 8

9.867 St. Incie 1 December 1976 7.08 8

6.363 Seaver Valley 1 April 1977 8.75 6

4.599 D. C. Cook 1 August 1975 10.42 8

7.302 D. C. Cook 2 July 1978 7.50 8

5.256

- Indian Point 2 July 1974

!!.50 8

8.059 Indian Point 3 August 1976 Joseph M. Farley 1 Decesher 1977, 9.42 8

6.602 8.08 6

4.247 Newaunee June 1974 11.58 4

4.058 North Anna 1 June 1978 7.58 6

3.984 Prairie Island 1

' December 1973 12.08 4

4.233 Fratete Island 2 Deceeber 1976 11.08 4

3.882 Point Beach i December 1970 15.08 4

5.284 Point Beach 2 October 1972 13.25 4'

4.643 R. E. Cinna 1 March 1970 15.83 4

5.547.

W. 8. Robinson 2 March 1971 14.83 6

7.795 Salee 1 June 1977 8.50 8

5.957 Surry 1 December 1972.

13.08 6

6.875 Surry 2

. May 1973 12.67 6

6.659 Trojan May 1976 9.67 8

6.777 Turkey Point 3 December 1972 13.08 6

6.875 r

Turkey Point A September 1973 12.33 6

6.481 l

Yankee Rowe June 1971 14.50 2

2.540 Eton I December 1973 12.08 8

8.466 Eion 2 September 1974 11.33 8

7.940 McCutre i December 1981 4.08 8

2.859 Sequoyah I July 1981 4.50 10 3.942 Sequoyah 2 Juna 1982 3.58 10 3.136 TOTAL 2.369(2)

2-45

. Table 2.4 - Statistical Data on Leakage Events of Check.

Valves to Accumulators Leak Rate (gpm)

Number of Events Frequency of Frequency of Occurrence (per hour)

Exceedance 5

11 10 4.64(-7) 9 1.48(-6) 20 3.80(-7) 9 1.01(-6) 50 3.80(-7) 3 6.33(-7) 100 1.27(-7) 1 2.53(-7) 200 4.22(-8) 2 1.27(-7) 8.44(-8) 8.44(-8) 4 i

l I

l l

l l

2-46 Table 2.5 ISL Results Initially Assigned Plant Damage States Frequency Contribution From Plant

~

Total Damage State VI VS Frequency

- - - - LOCA - - - - - --.- 1.96-4_

1. 4 4.-_4_.,

3.4,4 DLOC 1.2-5 0

1.2-5 DILOC 9.8-8 7.7_6 7.8-6 BC 2.1-8 0

2.1-8 ~

7D 1.5-7 0

~ ~ " '" TIS-T 7FPV 7.4-8 1.7-7 2.4-7 IFPV 1.8-8 8.0-7 8.2-7 IFV 8.4-8 5.9-8 1.4-7 Totals 2.1-4 1.5-4 3.6-4 Note:

LOCA: denotes a PDS, which contains those sequences in which the RC leakage in both ISL pathways analyzed exceeds 150 gpm, but does not exceed the RHR system relief valve capacity.

The sequences are essentially medium LOCAs.

DLOC: denotes a PDS, which contains sequences in which the ISL is terminated.

DILOCA: denotes a PDS, in which coolant makeup is being supplied to the core, but the ISL has not been terminated, i

The other plant damage states are involving containment bypassing ISLs and core damage.

t l

2-47 Table 2.6 Plant Operational Modes

• Average Reactivity

% of Rated Coolant Operational Mode.

Condition, Keff Thermal Power **

Temperature

1. POWER OPERATION

> 0.99

> 5%

> (TDHR) F

2. STARTUP

> 0.99

< 5%

> (TDHR)F

3. HOT STANDBY

< 0,99 0

> (TDHR) F

4. HOT SHUTDOWN

< 0.99 0

(TDHR)F>T,yg>200*F

5. COLD SHUTDOWN

< 0.99 0

< 200*F

6. REFUELING ***

< 0.95 0

< 140'F TDHR = temperature at which the DHR system is initiated (generally 280*F -

350'F

• As defined in B&W, CE, and W standard technical specifications.

Note many plants do not use standard technical specifications.

• • Excluding decay heat.

• • • Fuel in the reactor vessel with the vessel head closure bolts less than fully tensioned or with the head removed, l

i

. 48 Table 2.7 Categories of 130 Reported Total DHR System Failures When Required to Operate (Loss of Function) at U.S.

PWRs 1976-1983 No. of Events

(% of Events)

Automation closure of suction /

37 (28.5) isolation valves Loss of inventory Inadequate RCS inventory resulting 26 (20.0) in loss of DHR pump suction Loss of RCS inventory through DHR 10 (7.7) system necessitating shutdown of DHR system Component Failures Shutdown or failure of DHR pump 21 (16.2)

Inability to open suction /

8 (6.1) isolation valve Others 28 (21.5)

Total 130 (100.0)

D

3-1 3.

EVALUATION OF CONTAINMENT BEHAVIOR 3.1.

Capacity at General Yield Seabrook Containment Building The Seabrook ' Station containment building (See Fig. 3.1) is a reinforced concrete structure consisting of a basemat, a cylindrical wall and a hemi-spherical dome.1 Th'e 'basemat is essentially a 10' thick' circular slab which supports the cylinder and other internal structures.

The cylinder has an internal diameter of 140', a height of 149' and a minimum wall thickness of 4'-6".

The dome internal radius is 69'-11 7/8" and the minimum wall thickness is 3'-6 7/8".

In addition, the containment has a mild steel liner on the inside.

The liner thickness is 1/4" at the base, 3/8" in the cylindrical wall, and 1/2" in the dome.

The containment 1.s reinforced with' ASTM A615 grade 60 reinforcing bars of various sizes, mainly #18, #14 and #11. The specified yield strength for the reinforcing bars is 60 ksi. Median yield stress and lognormal standard devia-

- tion obtained from the results of tensile tests are shown in Table 3.1.2 por the ultimate strength of #18 reinforcing bars a' mean value of 109 ksi and a CoV of 0.025 has been reported.2 The liner steel conforms witn ASME SA516 f

grade 60 for which the specified yield strength is 32 ksi.

The mean yield stress was found' to be 45.4 ksi with a CoV of 0.042.2 At 271*F a mean yield stress of 40.5 ksi and a CoV of 0.065 are reported. 2 The mean ultimate strength at 271'F was estimated to be 59 ksi and with a CoV of 0.09.2

\\

The primary membrane reinforcement in the cylindrical wall is divided into two equal groups placed near the inside and outside faces of the contain-ment wall.

Each group ~ consists of two layers of hoop bars and one layer of meridional bars as shown in Figure 3.2.

Since the cylinder basemat intersec-tion is subjected to high bending moments and shear forces, secondary meri-dional reinforcement is placed 'in this region (See Figure 3.2).

In addition, i

two layers of seismic rebars inclined 45' to the vertical axis are placed near the outside surface of the cylinder wall.

Shear ties are also placed in the l

3-2 cylinder near the cylinder-basemat intersection (See Figure 3.2).

Major rein-forcement details for the containment wall are summarized in Table 3.2.

The dome reinforcement follows the cylinder reinforcement until 9.4*

above the springline.

Between 9.4' and 79.2* the hoop reinforcement is reduced to one #18 bar near each face.

Above 79.2* the hoop reinforcement is terminated and the reinforcement pattern is orthogonal. The meridional cylin-der reinforcement is continued to 60' above the springline with an increase in its density as the elevation increases.

Above 60' every alternate meridional rebar is terminated, and they are bent such that the reinforce;nent pattern near the dome apex is orthogonal. Details of the dome reinforcement are shown in Ta,ble 3.3.

Concrete with two design strengths was used in the Seabrook containment building.

A 4,000 psi design strength concrete was used for the basemat, for the cylinder near the intersection with the basemat, and for both the cylinder and dome near the doma-cylinder intersection.

In the cylinder and dome portions where primary membrane behavior is expected a 3,000 psi design strength concrete was used.

The median and lognormal standard deviation for the 4,000 psi and 3,000 psi design strength have been reported for _28-day old cylinders and for aged congrete.2 These quantities are given in Table 3.4 which was obtained from Reference 2.

The 28-day median concrete strength values were obtained from results of cylinder compression _ tests for the concrete used in the Seabrook structures.

The median values for the aged concrete were obtained from the 28-day values using the correlation given in Figure 2.1 of Reference 6.

Seabrook Containment Model A finite element model of the Seabrook containment was developed to be used with the computer code NFAP.3 The model is shown in Figure 3.3 and is based on an axisymmetric idealization of the geometry, which is considered a good approximation for a structural failure evaluation under axisynmetric pressure loads.

The containment finite element model consists of 408 eight-noded isoparametric elements and 1354 nodes.

A set of nonlinear spring

3-3 elements with a bilinear stress-strain law are used to model reinforcing details such as shear ties.

The basemat was considered to be fixed at the bottom nodes.

Throughout the cylinder wall and dome the model' has 8 layers of eight-noded elements across -the wall thickness, as shown in Figure 3.3.

Six layers of elements were used through the thickness of most of the basemat.

The element layers and its properties were chosen to represent separately the i

liner, the plain concrete, and the reinforced concrete with hoop, meridional and diagonal rebars.

Spacing and sizes of the layers have been chosen in

[

order to model the actual rebar placements as close as possible.-

This is particularly pertinent at the cylinder-basemat intersection where high bending moments and shear forces will develop.

In addition to these criteria, the modeling requirements commonly used with finite element analysis. were also taken into consideration.

2 The inelastic behavior of the plain concrete is described by the Chen and Chen elastic-plastic-fracture model.3 Material properties for this model were estimated from the aged (as built) concrete properties and are shown in Table 3.5.

Post-cracking behavior of the concrete was modeled using a normal stiff-ness reduction factor - a of 10 , and a shear stiffness reduction factor 8=0.5/( c1/ cto), where ci, is the principal normal strain normal to the crack and cto the tensile strain at crack initiation.3 The shear stiffness reduction factor is limited to a value not less than 0.10 to account for the cummulative effect of interface shear transfer and dowel ' action.

The normal stiffness reduction factor a reduces the normal stress diagonal element in the stress-strain matrix, and 8 reduces the shear stress diagonal element in the stress-strain matrix.

The tension stiffening effect was modeled with a factor -0.1, which multiplies the concrete Young's modulus.

The elastoplastic behavior of the reinforcing bars and liner steel was modeled.by a bilinear stress-strain curve and a Von Mises plasticity model j.

with isotropic hardening.

Since the #18 reinforcement bars provide most of the reinforcement, the mean material properties for these bars were used for i

all reinforcing bars.

For the itner, the mean properties at 271*F were used.

The plain concrete properties used are shown in Table 3.5.

3 '

Loads included in the analysis are the dead weight of the containment and internal pressure. The dead weight is applied to the containment in the first load step at the beginning of the analysis, while the pressure load is applied to the containment in small increments (5.0 ' psig) in order to detect the 1

initiation of nonlinear concrete behavior and concrete cracking.

Once the concrete cracks its stiffness in the direction normal to the crack plane is reduced by the factor a defined above, and the released stresses are redistributed to the reinforcing steel. As the pressure load is increased the next nonlinearity is the yielding of the. liner steel. -

At this internal pressure the containment is cracked in both the cylinder and dome, and some flexural cracking has been initiated in the cylinder-basemat intersection region.

r Based on the results of the analysis described above, yielding of the Inside layer of vertical reinforcement at the cylinder-basemat intersection is initiated at 130 psig.

At a pressure of 154 psig yielding of the internal layer of hoop reinforcing is initiated. Above 154 psig the load increment was reduced to 1 psig and-yielding in the outside layer of the cylinder hoop reinforcement was observed at a pressure of 157 psig. From 159 psig up to 165 psig internal pressure load increments of 0.5 psig were used.

At 165 psig hoop yielding in the cylinder wall extended oher a very large portion of the l

wall in both the inside and outside layers of hoop reinforcement.

At a pressure of 165 psig the strain in the inside layer of vertical reinforcement at the basemat-cylinder intersection has reached 1 percent and the maximum concrete compressive stress at the basemat-cylinder intersection l

1s 5100 psi.

The radial shear dowel at the base has reached a strain of approximately 2 percent and the concrete cracking extends beyond the vertical compression reinforcement near the outside face of the cylinder wall.

The computer analysis did not result in a numerical instability indicating that higher pressures could still be obtained; however, the pressure increments in the analysis would have to be further decreased. Consequently, it was decided l

not to continue the analysis any further.

The extensive cracking at the base 1

l and large strains in the reinforcing indicate that a shear failure at the base is a potential failure mode.

t

3-5 4

3.?.. Behavior at Large Deformation As discussed in Section 3.1, the containment structure is predicted to reach a general yield state at a pressure 'of 157 psig, which confirms the estimate provided by SMA.

As the pressure is increased above this level the containment structure will begin to undergo large deformations.

SMA evaluated the behavior of the containment structure at such pressure levels and the results of this evaluation are summarized 'in Appendix H'.1 of the PSA,

2 The hand calculations performed by SMA and used for the probabilistic assessment primarily identify several possible weak places in the structure and determine the corresponding maximum pressure capacity in search of the controlling failure mode.

The uncertainties in the results are estimated and identified as coefficients of variation (CoV) to account for both uncertainty and randomness of material behavior and lack of knowledge regarding the exact structural. behavior.

The break of the liner plate is defined as the failure mode.

The capacity of the containment structure is computed in terms of the internal accident pressure it can withstand.

Any leakage associated with the 4

pressure level is estimated with a CoV.

Accident scenarios are postulated for both wet and dry containment condi-tions.. The corresponding containment liner temperatures are 271*F for the wet case and 700*F for the dry case.

The structural calculations are first per-I i

formed for the wet case and then modified to reflect the reduced material strengths for the dry case.

The various failure modes considered in this analysis are discussed and evaluated below.

During the course of its review, BNL observed that the SMA calculations did not show any checker's signature.

As a result, PSNH has committed to perform a complete and independent check for all containment strength calculations.

Membrane Failure 1

The cylindrical wall and the hemispherical dome are assumed axisymmetric and they take the pressure load by membrane action.

Both the hoop and the meridional pressure capacities are determined based upon the ultimate strength of reinforcing steel bars, failure of which will lead to a gross containment

_r-m--

3-6 failure.

The median pressure capacities calculated. by SMA at ;271*F are as follows:

Mode Pressure CoV cylinder,- hoop tension 216 psig (governs)

.12 dome, hoop or meridional 223 psig.

.12 cylinder, meridional tension 281 psig

.12 The govering hoop failure at 700'F corrdsponds to a med_ian pressure of 198 psig. The above capacities are based on the assumption that the membrane forces are resisted by the reinforcing bars and the liner plate, and not by concrete.

.Since the above pressure values correspond' to th'e ultimate strength of reinforcing steel (109 ksi 'at 7.57, strain), the containment will undergo a great amount of expansion before failure.- This is illustrated in Figure 3.4 which plots. containment pressure vs. radial displacement of the containment

~

I wall as calculated. by SMA.

At 216 psig the radial displacement of the con-tainment wall away from the base is in excess 'of 3.0 feet. ~ SMA believes thati

~

[

at this pressure there is a 50 percent chance that the containment liner will remain intact and there will be no gross containment rupture.

BNL believes L

that at these ~1arge containment deformations it is difficult to accurately predict the behavior of the-containment and that containment liner failure is i

much more likely.

It is also noted from Figure 3.4 that at a pressure of 216 psig the pressure-displacement-curve is almost horizontal.

Thus, any further pressure capability of the containment would have to be attributed to even greater material strength of the reinforcing steel. Although some reinforcing

^

steel may have. a greater strength, BNL believes 'that for the high strain l

levels being considered that further consideration must'be given to instances of progressive failure of the reinforcing steel.

In the light of the above discussion, BNL considers the 216 psig pressure capacity predicted by SMA to be an upper bound failure pressure. BNL believes that a more suitable median failure pressure should correspond to the pressure level at which the primary membrane reinforcing steel reaches 1 percent strain

3-7 i

(175 psig for the Seabrook containment).

Such a level recognizes the ability of the containment to withstand pressures beyond the general yield, but limits the amount of containment deformation to levels more comensurate with the current state of knowledge concerning containment performance.

The median failure pressure corresponding to the 1 percent strain level for the dry condition is 158 psig as indicated by PSNH in the response to NRC question 20 (PSNH lettar dated October 31,1986).

Shear Failure of Wall at Base The shear failure of the cylindrical wall is estimated by SMA at a median pressure value of 319 psig with a CoV of 0.29.

This pressure value is deter-mined based upon the yield strength of the reinforcing steel and on the assumption that the critical section will occur at a distance of 0.7 x effec-tive wall thickness above the base.

The shear failure corresponding to the ultimate strength of the reinforcing steel is estimated by SMA at a median pressure of 408 psig with a CoV of 0.3.

SMA assigned a large variability to this failure mode due to their uncer-tainty about the applicability of their elastic analysis when some yielding occurs.

However, BNL feels that this failure mode is more critical than assumed by SMA.

As discussed in Section 3.1, BNL investigated this mode of failure by means of a non-linear finite element analysis.

It was confirmed that such a failure is not expected to occur for pressures up to 165 psig.

However, BNL believes that a shear failure at the base is a potential failure mode at pressures above this level.

Flexural and Shear Failure of Base Slab The flexural capacity of the base slab is determined by SMA based upon the yield line theory.

The median basic capacity is estimated to be 168 psig. However, when the friction and mechanical locking between the base slab and the ring girder of the enclosure building are considered, the median overall capacity is estimated by SMA as 400 psig with a CoV of 0.25.

Conse-quently, it is concluded that flexural failure of the base slab is not a

3-8 controlling failure mode.

The shear strength -of the base slab is also calcu-lated considering restraint from the ring girder of the enclosure building.

l The median pressurc capacity is estimated by SMA as 323 psig with a.CoV of 0.23.

BNL reviewed the SMA calculations concerning the shear and flexural fail-ure modes of the base slab and agreed that these failure modes would not be controlling.

+

Containment Deformations The deformation of the containment is calculated by SMA based upon the assumption that concrete will share tensile load with steel even at the ulti-mate strength of steel.

This assumption presupposes a bond between concrete and the reinforcing bars up to the failure pressure except at the location of the cracks which are postulated to occur with' a spacing of approximately 21 5

inches. The biaxial tension test results presented by Julien and Schultz indicate that concrete will crack at an early steel stress level and the deformation at a high steel stress level is due to steel strain only.

The hoop land the meridional reinforceing bars used in these tests were of same diameter as those for Seabrook, namely, No. 18 and No. 14.

The concrete f

cracked,at a steel stress of 9.a ksi and the effect of concrete stiffness dis-i appeared beyond a steel stress of approximately 25 ksi.

Consequently, BNL is i

concerned that SMA may have underestimated the containment deformations corre-sponding to the containment pressure levels.. Since the containment deforma-l tions can result in containment penetration failures, _an underestimation of the deformations would result in higher predicted failure pressures for the penetrations.

This is so because the failure of penetrations is predicted on 4'

the basis of the amount of deformation they can withstand.

The deformation capability is then translated into a pressure capability based on the overall containment pressure-deformation response curve.

Thus, if this curve under-l estimates the containment deformation, the penetration failure pressure asso-l ciated with a given containment deformation will be too high.

l In response to this concern PSNH provided a comparison of the containment pressure displacement curve with and without bond stress (RAI 32, PSNH letter 4

= --

3-9 y

dated November. 7, 1986).

This comparison is shown in Fig.

3.4.

PSNH con-cluded that an assumption of no b'ond stress would have no effect on the con-clusions. of their studies.

The effect that this assumption has on the 4

reported capability pressures for critical containment penetrations is dis-cussed in Section 3.3.

3.3 Capability of Penetrations l

Many penetrations through the containment shell are provided.

These include. a few major penetrations such as the equipment hatch, personnel airlock and fuel transfer tube and numerous smaller penetrations accommodating system high energy piping, moderate energy piping, electrical, instrumentation and ventilation lines.

All penetrations are anchored to sleeves which are embedded in the concrete containment wall.

For the major penetrations, the containment wall is thickened into a hub around the penetration sleeve with the wall hoop and meridional reinforcing members directed past the opening in a continuous fashion and additional reinforcement provided as sleeve anchor-age. For each high energy line, the penetration is a forged member, termed a t

flued head, which forms an integral part of the piping and the containment sleeve which is welded to the containment liner.

For all other penetrations i

the closure is a flat plate welded to the containment sleeve and either welded or connected with a compression fitting to the penetrating elenmnt.

These i

flat plate closures accommodate either single or multiple penetrations.

i l

To assess the capacity of large penetrations, SMA performed an evaluation of the equipment hatch.

This hatch is the largest of the large penetrations and was-considered to represent the bounding or most critical penetration in this category.

In the evaluation, the capacity of the hatch anchorage was determined to be in excess of 300 psi.

Possible failure of the liner at the hatch juncture due to sleeve-concrete separation was also evaluated and found to be improbable due to the low magnitude of the predicted liner strain, j

These evaluations are considered acceptable.

1 j

Although the capacity of the fuel transfer tube anchorage was established

[

in the equipment hatch evaluation, the containment wall in the vicinity of j

this penetration is subject to punching shear failure since it makes hard I,

r I.

i 1

,--n--

-._4-->

,,~m_

,,_,y,.

..m m

c.,.,,m~

.,,,,____-,,.m m.-

3-10 contact with the fuel transfer building when the containment expands. Using a simple approximation to model the loading and relying primarily on doweling action of the containment reinforcement to resist the load, SMA determined a mean capacity of 320 psig in this failure mode.. Acknowledging the approximate nature of this calculation, SMA assigned a large factqr of uncertainty to the results.

Probabilistic aspects notwithstanding the crude nature of this calculation warranted further verification of the results.

Therefore, BNL performed additional calculations for this failure mode to form an independent assessment of the important force-displacement parameters.

In the BNL evaluations an approximate model of the system was again used but this model differed from that used by SMA.

The results, although differ-ent from those developed by SMA, indicated that no gross deficiencies existed in the SMA calculations.

Further, the estimate of the pressure at which contact is made by the containment shell against the fuel transfer building, a controlling parameter in the evaluations, is not subject to the large uncer-tainty associated with the force-displacement parameters mentioned above.

Consequentially the SMA calculations although approximate in nature are con-sidered sufficient to characterize the impact this failure mode has on con-tainment integrity.

To. assess the capacity of small pipe penetrations, SMA perforned evaloa-tions for three specific penetrations, X-26, X-8 and X-23.

X-26 was stated to l

be a bounding or most critical example of a single pipe moderate energy j

penetration, X-8 a bounding case for a high energy penetration and X-23 a bounding case for a multiple pipe moderate energy penetration. For each case, I

simplistic inelastic analysis methods were used to estimate the forces developed at the pipe / penetration interface as a function of containment I

internal pressure.

This data coupled with estimates of the penetration failure characteristics allowed the calculation of the probability of penetra-tion failure as a function of containment pressure in each case.

The median 2

l failure pressure and the associated median leak areas were 166 psig/0. Sin,

{

180 psig/50in and >216 psig/61n for X-26, X-8 and X-23 respectively.

2 2

+

The discussions included in the SMA evaluations provided the basis for the SMA contentions that the penetrations evaluated were the bounding cases l

s 3-11 for the penetration. types considered.

Those disc'ussions, however, did not '

adequately characterize all other penetrations.

For this reason SMA was requested to compile a list of all penetrat' ions, categorize them in accordance with design features and demonstrate that the 7erformance of each is-adequate-ly represented and bounded by the sample of three evaluated.

As a response SMA provided Table 3.6 characterizing all penetrations and the calculations considered to be pertinent for their qualification.

4 A review was made of the evaluations provided for the bounding cases.

In each instance the structural aspects lof the" calculation seemed appropriate, with the exception noted below, but the assignment of leakage area was consid-ered arbitrary.

In a'ddi tion, for each case, failure was induced Of the displacement of the containment shell. "Since the correspondence between this displacement and the containment pressure is dependent on the bonding assu:np-tion made for the containtrent ' reinforcement and since BNL has requested SMA to z

perform evaluations correspondilng to a no bonding assumption '(see discus'sion; in Section 3.2) BNL elected to further assess the failure pressure and leakagea area for the two penetrations X-8 and X-26.

Penetration X-23 w.s. not con-sidered since it exhibited a high failure pressure.

Fl.r the high energy penetration, X-8, SMA estimated the median failure pressIre to be 180 psig for the wet case with an associated median leakage area of 50 in2 and a lognormal standard deviation of 0.5.

Tha estimate of the median leakage area was based on an annular gap of 1/2 'in, for the full circumference, at the containment sleeve. The estimite of the standard devia-tion was arbitrary.

For[the no bond case BNL estimates the median failure pressure to be 167 psig for the wet case and.152.psig for the dry case.

BNL i

accepts the SMA estimate for the median leakage area but disagrees with the assumption regarding the standard deviation.

In the absence of more explicit data concerning the behavior ofi penetration sleeves at failure, BNL believes that an upper bound for the leakage area approaching the total annulus between the pipe and containment sleeve should be considered.

Bas'ed on these considerations failure of this penetration corresponds to a type B failure for the median leak (6 square inches to about 0.5 square fcot) and a type C failure (greater than 0.5 square foot) for the upper bound leak.

These

3-12 e

f categorizations agree with the assumptions in the applicant's consequence

\\

ena'iysi s.

For the Aoderate energy penetration, X-26, SMA estimated the median fail-tre pressure to be 166 psig for the wet case with an associated median leakage 2

area of 0.5 in and a lognormal standard deviation of 0.69.

The estimate of the median leakage area was based on an annular crack of 0.06 in, the machined clearance between the pipe and the thru hole in the closure plate, extending over 60% of the circumference.

The standard deviation was derived by.consid-ering 0.02 inches a miniumum crack width and full circumference cracks.

For the no bond case it is estimated that the mdf an failure pressure is 159 psig for the wet case and 147 psig.for the dry case.

Regarding leakage a'ea, the estimate for the nmdian leakage area is accepted but an assumption for the upper bound leakage area equivalent to that recommended for X-8 should be used.

Specjfically, consideration should be,given to an upper bound for the leakage area approaching the total annulus between the pipe and the contain-ment sleeve. Based.cn these considerations failure of this penetration corre-sponds to a type A failure for the median leak (less than 6 square inches) and 1 type 6 failure for the upper bound leak (6 square inches to about 0.5 square foot).

The applicant's consequence analysis assumed tyoe A for both conditions.,

As noted above, one deficiency was noted in the structural evaluations 4

for the penetrations.

In those evaluations only a simple concrete shear cone calculation for a generic case was provided to show that the penetration anchorage capacities were adequate.

Owing to the highly cracked state of the containment wall'at high containment overpressures the relitnce on normal concrete act. ion was questionec.

SMA was requested to provide additional calculatiens to demonstrate that small diameter penetration sleeves do not I

punch through the containment wall under the worst pressure cor.ditions assumed in the analysis.

The applicant's response to this request was reviewed and independent tealculations were performed by BNL to support tne conclusion that l

the anchorage provided for penetrations is adequate.

t Another potential failure mode for the piping penetrations, is the fail-ure of the pipe both inside and outside the containment.

This failure mode j..

3 13 was e" valuated by SMA for the piping in the sample considered most prone to this failure, the piping passing through penetration X-8.-

The calculations indicated that the piping failure pressure exceeded the penetration failure a

Further given the high ductility of the piping, any failures of the paessure.

piping would have gross / distortion, crushing and -section collapse associated with them limiting the size of the potential leakage areas. These evaluations seem appropriate.for the piping. considered.-

Other piping penetrations involve the containment ventilation and air i

purge systems and the containment sump system.

The containment ventilation lines have isolation valves both inside and outside of the containment.

f

~

For these penetrations, the most '*1ikely mode of failure is considered to be deterioration of the valve sealant materials at elevated temperatures.

In the event of the seal failure of the inner containment valve, the volume between the valves must fill and achieve an elevated temperature before failure at the outer isolation valve can occur.

The elapsed time for this failure mode is j

anticipated to be. long as compared to other containment failure modes and is therefore considered of little consequence.

J Y

The sump system penetration is at elevation -31'6".

A review of the

'[ drawings originally provided to BNL indicated that the pen'etration sleeve is i

. c welded to the liner at the inside of the containment and to the train A&R sump suction valve containment tank on the outside of the containment. As such the sump suction valve containment tank was considered to be a direct extension of the containment vessel and would have to have sufficient capacity to withstand the temperatures and pressures associated with containment overpressurization events.

It appeared that SMA did not consider the tank in their evaluations

~

f and therefore, they were requested to perform an assessment of the capacity of i

the sump containment vessel.to the accident conditions.

In the response PSNH provided drawings that showed that the tank is isolated,from the containment atmosphere by a welded plate closure between the penetration sleeve and the suction line piping. Because of its isolation the sump tank is not subject to accident conditions of pressure and temperature and no further evaluation of its capacity is required ph i_.__.,^ -, -

--s--

- ~ ~ ' '

^~~

~~

~

~

3-14 Another type of penetration is the ele.ctrical penetration assemblies

'(EPAs). - The applicant has indicated that they briefly reviewed these penetra-tions and that they were not a controlling mode of failure.

These types of penetrations have -been -included in the ongoing SANDIA test program sponsored by the NRC.

The NRC staff provided the results of recent tests at Sandia which are

. summarized as follows:

Test Condition Manufacturer Pressure Temperature Duration Conax _

120 psig 700*F 8 days, 15 hours1.736111e-4 days <br />0.00417 hours <br />2.480159e-5 weeks <br />5.7075e-6 months <br /> D.G. OBrien 140 psig 360*F 9-1/2 days Westinghouse 60 psig 400*

9-1/2 days i

For the Conax EPA there was no detectable leakage during or after the test.

The inner seal failed about I hour into the test while the outer seal remained leaktight throughout the test.

For the D.G. O'Brien EPA there was no signifi-cant increase in steam leakage during the test period.

There was minor and insignificant leakage found during air leak tests following cooldown after the completion of the test.

For the Westinghouse EPA there were no detectable leaks during the steam pressurized portion of the test or during the cool-down.

The results of these tests do not challenge the applicant's conclusion that EPAs would not be a controlling mode of failure. It appears that the principal mode of failure for EPAs would be degradation of the seals when exposed to very high temperatures for a long period of time.

If such failures occurred, they would be late in the accident sequence and have little conse-quence on emergency planning issues.

l I

3-15 3.4 Summary of Structural Findings -

Based on its nonlinear finite element analysis of ~the Seabrook contain-j ment 8NL concludes that a shear failure at the base of the cylindrical wall is a potential failure mode but would not occur before' reaching a pressure of 165 psig.

BNL agrees.that the containment structure would reach a general yield 4

state in the hoop reinforcing steel at a pressure of 157 psig and that it is appropriate to consider this pressure as a lower bound pressure for the hoop mode of failure.

However, BNL believes that the median hoop failure pressure should correspond to the one percent strain. level in the hoop reinforcing steel, which is a pressure of 175' psig.

The above pressures are for the wet containment conditions.

For the dry containment conditions the corresponding median failure pressure is 158 psig and. the lower bound. pressure (general yield) is estimated to be 145 psig.-

This latter value is based on the reduction factor recommendation in Section 11.3.4.1 of PLG-0300.

With regard to containment penetrations, BNL believes that the failure pressures should be based on containment deformations assuming no hond strength between the reinforcing steel and concrete. Based on this assumption BNL estimates median failure pressures for the wet containment condition of 159 psig and 167 psig for penetrations X-26 and X-8.

For penetration X-26, BNL agrees.that a Type A (less than 6 square inches) leak path is appropriate for the median estimate; however a Type B (6 square inches to about 0.5 square foot) leak path should be considered as an upper bound estimate.

For l

penetration X-8, BNL agrees that a Type R leak path is appropriate for the median estimate; however, a Type C (greater than 0.5 square foot) should be considered as an upper bound estimate.

For the dry containment conditions, BNL estimated the median failure I~

pressures for penetrations X-26 and X-8 to be 147 psig and 152 psig, respec-tively.

These values are also based on the reduction factor recommended in Section 11.3.4.1 of PLG-0300 l

i 3 16 1

Although BNL has performed some independent calculations to support its conclusions regarding the containment strength, it also relied on the results of calculations performed by PSNH and its contractors.

Therefore, BNL recom-mends that a complete and independent check of all relevant containment strength calculations be performed by PSNH. PSNH comitted to such a check in their letter to the NRC dated October 31, 1986 and has indicated that such a check has been completed.

3.5 References 1.

Containment Design Report for Public Service Company of New Hampshire.

Seabrook Station Unit Nos.1 & 2, by United Engineering Constructors Inc.,

January 1985.

2.

Seabrook Station Probabilistic Risk Assessment, Pickard, Lowe and Garrick, Inc., PLG-0300, Appendix H.1, December 1983.

3.

Sharma, S., Wang, Y. K. and Reich, M.,

" Ultimate Pressure Capacity of Reinforced and Prestressed Concrete' Containments", NUREG/CR-4149, May 1985.

4.

Hand Calculations by Structural Mechanics Associates (SMA), Originated by RP, dated December 1982.

5.

Julien, J. T. and Schultz, D.

M., Tension Test of Concrete Containment Wall. Elements, Transaction of the 7th International Conference on Structural Mechanics in Reactor Technology, Vol. J, pp. 237-244 6.

Troxell, G. E., Davis, H. E. and Kelly, J.

W., Cegosition and Properties of Concrete, McGraw-Hill, 1968.

i i

I

e.

e 3-17 (LCCEPttSCesmett e

I 8

( CCsetaaNutNT..

I N

l s

i T-s 11e*

tNetcsutt tuto N4 3

QCat l

30 '

g

( MAICM(Cusput NT LOCAtlCN PtAN"

• eer.117/s*

ON840tl f.Pt.

=

- l/T'T M rtl Nee E#II#O' I

iT[

5 5 ***C Usst EL.119".C*

f

• a O*.P'* S d' 4* WALL CTt!N0t9

.,r.e n i

I

=s/9* feeCE Liset PLAf t PCR CTt1Mott 0434C86 CP iMftFNAL SftuCTutts ST Ofpte COOT 3 aatfAL StCTICNS mffAt SECflCNS SACEt0 47 CONCltff

• SACEROSYCONCattt.

-45mt SEC. E Qiv. 2 I

ASmt 88C. M. OIV.2 8

900aa y tOyPeetest warp y e ttNe t

(.

.nettsoNN46 tCC:

~

l

• - 2P-8' L:-

P a t" etra*

2

. - ~,... m

~

~

l asifAt PCettCNS C# A(L J l====4J**C*

PtNtitAf tCasa LACIED tv

,NyrgNgLgggggAg,,

a CONCatit Am Aa.CMCaAct A$mt atCilCN IC. Cl4 2< =

-lTSf tas*aamt 3EC.ZCat 2 t17ttssaA SCU*doat?.

d, i

IM* TwsCE LINee P(Aft Anat SECTIO *e A Cav 2 pct anAf ANO Ptf g g,( mAf tNot estf C0 gt g.416' C*

t1.1.e ??.e*

CONTA&Natf N f1=

C'0"*0* **I tasCLCiust sue. DING MAf i

!k sof aes CC:es:-

- '\\

.y 8432 mar \\

. f_.30*.e*

I (Civ II-l

-etACTCt mit

_,, tt f-Le f*. ?*

..r.a

... eft, t

Figure 3.1 Containment building cross-section, b

3-18 I

l

,.,:: n.s6 N.N..

AuWe 3acAAC) 3 l

.T. -

re.~

8

.u.iv.e.

v.

• '. N

$Q-.,.

.x i-y 9

i...

g:i' s.,3 G.lu.

G, c

,e-s, 2 2 q

'o y

~i ru a v. -

7.

1.....l.......

eA

.e 1

k

,1\\

m cater

.... I r \\,'T a

s 4

)

t., y - %

- 3 r.

4.

.i,.

.3

.1.

y.

{N N

-eya

/KI,i a

( * ?..""'6' **' **

I,l

,a j$ l M

W su !w.o*1, t_i.

NMt

'h*g"*

• 6 1

i l-h l

l sv.

.e.

i Figure 3.2 Cylinder reinforcement.

3-19

%x N

\\[

ihd a

{

1 2345678

!U

!U i_l.i l:._.!

i l

_l l

.I_

j 1

LINER l

2&5 PLAIN CONCRETE

~#

3&7 HOOP REINFORCEMENT 4&6 fiERIDIONAL REINFORCEl1ENT 8

DIAGONAL SEIStilC REINFORCEMENT I

L.,

=

l l

l l FAfl

~

If I

J l

Figure 3.3 Containment finite element model (NFAP).

250 l

I includin0 effect of bond sIress (ori inal curve) 0 1

~

~

g' 200 i

fp-s

~

A 150 assaning no Mng of i

N f

hoop bars i

e i

o.

I p

100 4

Y 8

E!

50

^

0 0

10 20 30 40 Radial Displacement (inches)

Figure 3.4 Pressure-radial displacement relation for containment.

i

a 3-21 Table 3.1 Statistics of Rebar Yield Strength for Various Sizes t

Bar Size Median Yield Lognormal Standard Stress (ksi)

Deviation'

1. 4 70.5

.031

1. 5 67.9

.031

1. 6 67.9

.039

1. 7 70.0

.035

1. 8 69.6

.034

1. 9 70.3

.040

1. 10 70.9

.037

1. 11 70.6

.040

1. 14 73.6

.031

1. 18 72.3

.028

1. 4 - #11*

69.8

.040

• Sizes #4 to #11 taken as a group J

l l

l l

... ~ - -, -. - - - -. ~ - -..

q Table 3.2 Reinforcement Details of the Containment Cylinder l

i MERIDIONAL i

l Hoop Primary Secondary betsmic j

Elevation (Both Faces)

(Both Faces)

(Inside Face)

Djiagonal

-30.0' to -15.0' 2-#18 9 12"

1. 18l 0 12"

'2-#18 9 12"

1. ,18 9 11" I

-15.0' to -5.0' 2-#18 9 12."

1. 18:9 12"

1. 18 9 12" 48 9 11" w

3 j

'I

}

- 5.0' to 80.0' 2-#18 9 12"

1. 18'8 12" g

A18 9 11" 80.0' to 119.0' 2-#18 9 12"

1. 18 9 12"

1. 18 9 22"

$14922" I

(alternate) t f

a h

i l

l l

l l

9

3-23 l

Table 3.3 Reinforcement Details of the Containment Dome Hoop Meridional Seismic Elevation (Both Faces)

(Both Faces)

Diagonal 0*(S.L.) 9.4*

2-#18 9 12"

1. 18 9 12"

1. 18 9 22" 9.4*

30*

1. 18 9 12"

1. 18 9 12"

1. 18 0 22"

1. 14 9 22" (alternate) 30' 45'

1. 18 9 12"

1. 18 9 10.4"

1. 14 0 19" 45' 60'

1. 18 0 12"

1. 18 0 12" 60' 79.2*

1. 18 9 12"

1. 18 9 12" 79.2* 90'

1. 18 0 6.4"

1. 18 9 6.4" O

e i

l I

i l

i i

. Table 3.4 Statistics of Concrete Compressive Strength.

i s

j 28-Day Old Cylinders

Aged Concrete

• l t

I

~

Median Logarithmic

- Median Logarithmic Strength Standard Strength Standard j

Concrete Type

-(psi)

Deviation (psi)

Deviation j

i 3000 psi Design Strength Concre.te 4750 0.14 5700 -

0.17 4000 psi Design Strength Concrete for Containment 5450 0.10 *

• 6540 0.14 Y

5 4000 psi Design Strength Concrete for Tunnels 5780 0.096 6940 0.14 1

4000 psi Design Strength Concrete j

for Other Structures 5593,-

0.10 6710 0.14 Hedian strength and logarithmic standard deviation are obtained by multiplying the 28 day strength by.a random factor, which is assumed to be independent of the 28 day strength and has~ a median of 1.2 and a lognormal standard deviation of 0.10.

This number was estimated.-

4 4

I t

j i

u-3-25 Table 3.5 Concrete Properties MATERIAL PARAMETER f'c=3000 psi f'c=4000 psi Young's Modulus 4340ksi 4650ksi Poission's Ratio 0.19 0.19 Yield Strength.in Uniaxial Tension 0.233ksi 0.262ksi Yield Strength in Uniaxial Compression 2.46ksi 2.81ksi Yield Strength in Biaxial Compression 2.85ksi 3.27ksi Fracture Strength in Unf axial Tension 0.54ksi 0.61ksi Fracture Strength in Uniaxial Compression 5.7ksi 6.54ksi Fracture Strain in Tension 0.00045 0.0005 Fracture Strajn in Compression 0.005 0.005

+

1

]

l Table 3.6 Characterization of Containment Penetrations l

4 Penetration Closure Penetration Number Penetration Specifically Qualification Type Analyzed Nethod I.

Flued llead X-1 to X-8, X-9 to X-15, X-8 Report pages 11.1-44 to X-63 to X-66 (18 inch, sch 100 Carbon Stee1) 11. 1 - 5 0 II.

Flat PlateClosure X-25, X-26, X-27 X-26 Pages M.1-39 to H.1-44 Th i ck Wa l l - La r ge (4 inch, sch 160, stainlesa)

Bore Piping

?

III. Flat PlatsClosure X-16 thru X-24 X-23 Pages H.1-39 to H.1-43 w

Th i n Wa l l - La rge X-28 thru X-34 (12 inch, sch 40, Carbon Steel) 53 Bore and Small Bore X-39, 41, 42, 50, 60, l

Piping 61, 67 I

Page H.1-37. H.1-39 IV.

Flat Plate Closure X-35 thru X-38 X-71 Thin Wall Piping X-4 0, X-4 3, X-47, X-48, -

Hultiple Penetration X-49, X-50, X-52, X-57,'

X-71 thru X-76

{

V.

Fuel Transfer Tube X-62 X-62 Page 11.1-50 to 11.1-55 i

o 1

l i

4-1 1

4.

CONTAINMENT EVENT TREE REVIEW 4

In this section the ability of the Seabrook containment to contain severe i

accident loads is examined.

Note that ways in which the containment might be bypassed or not isolated are discussed separately in'Section 2.- This section, therefore, specifically deals with ways in which. severe accident loads might 4

i result in loss of integrity of the Seabrook. containment.

The section.is divided into two major parts.

Firstly, potential containment loads are iden-tified and discussed.

Then, the applicability of these loads for Seabrook is 2

briefly summarized.

r The bottom line of the updated assessment of containment performance for i'

Seabrook is given in Table 4.1, which was reproduced from PLG-0465.1 The con-ditional probability of a gross early containment failure given a core melt

)

accident was predicted to be 0.001 at Seabrook compared with 0.34 in the RSS.2

[

The probability conditional on core melt of early failure in PLG-0465 is an order of magnitude lower than in the SSPSA.3 This.is largely due to the reduced frequency (relative to the value in the SSPSA) of interfacing system l

LOCAs, which are discussed separately in Section 2.

Note that the containment event tree quantification in the SSPSA 3 was reviewed at BNL in NUREG/CR-4540."

The BNL review was limited in scope and did.not include at that time a detailed assessment of the Seabrook containment

}

behavior. This has subsequently been performed as part of the present review,

{

and it is documented in Section 3.

However, the review of the SSPSA was sufficiently detailed to allow BNL to conclude:

i

~

"There is negligible probability of prompt containment failure.

Failure I

l during the first few hours after core melt is also unlikely and the timing of overpressure failure is very long compared to the RSS.

Most core melt acci-i dents would be effectively mitigated by containment spray operation."

!'L The above conclusions were not based on Seabrook specific calculations

~

performed at BNL but reflected our best judgment based on extensive reviews of other similar containment designs (in particular, our review 3 of the Zion Probabilistic Safety Study ).

In this section, we critically review the above s

i i

4-2 conclusions based on our current understanding of containment loads and performance during severe accidents.

4.1 Potential Containment loads i

I During a core melt' accident, there are several. possible types of contain-ment loads that could occur. Each are briefly discussed below:

H_, combustion: During a core melt accident, significant quantities of H2 f

and other combustible gases could be generated.

If these combustible-gases accumulated to large concentrations before igniting, the resulting deflagra-tion could impose a high pressure / temperature. load on the containment.

The applicant presented information to indicate that such loads would not serious-ly challenge the-Seabrook containment.

This potential threat-to containment integrity is discussed in more detail later in this section.

Steam /noncondensible gas partial pressures': Without the containment heat removal systems operating, steam and noncondensible gases generated during the core melt accident would cause the pressure in containment to increase.

At f

the time of reactor vessel failure, there is the potential for the hot core debris to contact water.

This contact could result in rapid boiling of the water and a sharp pressure pulse in containment.

Limiting calculations were 3

performed by the NRC sponsored Containment Loa ~ s Working r,roup (CLWG), which d

7 demonstrated that the pressure pulses resulting from quenching the core debris by boiling water would not pose a threat to the Zion containment. BNL l

considers these calculations to be applicable to Seabrook and, as the calculated peak load is much lower than the general yield of the Seabrook containment (i.e.,

157 psig, refer to Section 3), we conclude that this containment load is also not a threat to the Seabrook containment.

I j

Steam explosions:

When molten core materials fall into water, experi-ments indicate that the boiling can become explosive in nature.

It has been postulated that these explosions could generate missiles which could directly fail the containment boundary.

The potential for an invessel steam explosion to occur and generate a missile capable of failing a containment building was investigated by a group of experts, and the results were published in

,t t

~

..... _ _ _. _ _ _. _ _ _ _ ~,. _ _ _.. _

_._,_..____....-__r__,-._.~.-

_,av

,i e

-A 9

h A-an uam

'a

'4-3 NUREG-1116.s The conclusion of this expert group was that such events have a relatively low probability.

The results of this expert group are consistent with the' applicant's submittals on Seabrook.

The allocation of a very low probability.(10 " conditional on core melt) to this event was considered to be reasonable by the BNL review team, i

Direct containment heating: This is an area of significant phenomenolog-ical uncertainty related specifically to core meltdown with the primary system at high pressure.

It has been suggested that if core materials are ejected 8

4 from the reactor vessel under pressure that they fonn fine aerosols, which could be dispersed into the containment atmosphere and directly heat it.

An additional concern is the oxidation of the metallic content of the core debris.

These reactions are very exothermic and would add an additional heat load to the containment. The pressure rise in containment due to direct heat-ing is proportional to the quantity of core debris ejected from the reactor vessel which is finely dispersed into the containment atmosphere.

The appli-cant considered that this phenomena is not a concern at Seabrook because of j

the design configuration of the containment, which they felt would prevent i

dispersal of the core materials into.the bulk of the containment atmosphere.

j The applicability of this phenomena to the Seabrook containment is discussed in the following section.

i 4.2 Application to Seabrook I

The combined probability (conditional on core melt) of the above phenome-l na resulting in early containment failure was determined by the applicant to l

be less than 10 4 for Seabrook.

In order to check the validity of the appli-cant's estimate of early containment failure, it would have been necessary for l

BNL to develop Seabrook-specific containment loads and combine them with our assessment of the structural capacity of the Seabrook containment. :However, given the scope of the present review, RNL has not developed Seabrook-specific containment loads. Thus we were not in a position to validate the applicant's j

estimate of early containment failure.

However, BNL has been involved in updating (NUREG/CR-4551, Volume 5) the l

risk profile for the Zion plant for input to the NRC's " Reactor Risk Reference l.

t

_ _ ~. _ _, _... _,.. _ _ _. _ _, _. - - _ _ _.., _

o 4-4 Document," NUREG-1150.30 The updating of risk for Zion was based on a method-

. ology developed as part of ' the Severe Accident Risk Reduction Program (NUREG/CR-4551, Volumes 1-4)- at Sandia National Laboratory (SNL).

This methodology used expert - judgment in an attempt to estimate the uncertainty associated with determining containment loads.

The methodology was developed at ' SNL specifically for the Surry plant but was extrapolated to Zion at BNL The Zion plant is very.s.imilar to Seabrook in terms of the containment volume to reactor power ratio.

Thus, extrapolation of the Zion loads to ' Seabrook would give some indication of the impact of applying this new methodology to Seabrook.

It must be emphasized that this exercise should in no way be interpreted as a Seabrook-specific calculation.

It simply gives some indica-tion of the sensitivity of the Seabrook results to the types of uncertainty in estimating containment loads discussed in'NUREG-1150.

It should also be noted that this work is preliminary and has not yet undergone full peer review j

outside of NRC and its contractors.

It is, therefore, subject to revision.

The range of containment loads reported in Volume 5 of NUREG/CR-4551 for i

Zion is very wide and far exceeds the loads that would be considered credible by the applicant for Seabrook.

Of particular interest is the loads at the time of reactor pressure vessel failure.

These loads can range from about"60 psia to 200 psia depending on whether core melt is occurring with the primary system at high or low pressure and on whether or not containment heat removal systems, CHRS (sprays and fan coolers) are operating.

The higher containment l

loads are postulated to occur for accidents in which the primary system pres-l sure remains high immediately before reactor pressure vessel failure.

For these accidents, direct heating of the containment atmosphere by core debris or, hydrogen combustion with a steam spike at the time of reactor vessel fail-ure are possible mechanisms for failing the containment.

The applicant has t

(

presented information which indicates that these mechanisms are not credible ways of failing the Seabrook containment.

However, as noted above, BNL does

{

not have Seabrook-specific containment loads so we cannot, at this time, elim-inate these mechanisms as potential ways of failing the Seabrook containment.

For accidents with the primary system at high pressure and without the CHRS operating an approximate median load of 135 psia (120 psig) was predicted j

for Zion.

If this median load is compared against the capacity of the

?

[

4-5 t

Seabrook containment given by the BNL review, one would conclude that the potential'for early containment failure at Seabrook is very -low and would not influence the risk estimate in PLG-0465.

However, the range of' loads esti-j mated for Zion implies considerable uncertainty.

The 95th percentile estimate 1

of-the probability of early containment failure at Zion is quoted as 0.17 in Volume 5 of:NUREG/CR-4551.

If this early containment failure probability were also true for Seabrook,- the risk estimates in PLG-046' m Id increase signiff-j cantly.- However, the ~ capacity of the Seabrook-~ cont r9ent is greater than Zion (the general yield for Seabrook is 157 psig co:..,ared with -134 psig for Zion) so the 95th percentile estimate of early containment failure should be lower at Seabrook than Zion.

However, BNL cannot at this time quantify how much lower because we have not quantified Seabrook-specific containment event j_

trees with Seabrook-specific containment loads combined with our estimate of the structural capability of the Seabrook containment.

J t

i i

4.3 Sumary 1

.j.

BNL did not develop Seabrook-specific containment loads given the scope

[

of the current review.

However, BNL did develop Zion-specific containment

{

loads as part of updating (NUREG/CR-4551. Volume 5) the Zion risk profile for i

input to NUREG-1150 As the Zion plant is similar to Seabrook, it was decided to use the Zion-specific loads to give some indication of the sensitivity of the Seabrook containment to the types of uncertainty in estimating containment loads identified in NUREG-1150.

The range of loads reported in NUREG/CR-4551 is very wide (60-200 psia). However, if the median Zion load is compared with the capacity of the Seabrook containment given by the BNL review, the poten-l tial for early containment failure at Seabrook is predicted to be very low.

However, the range of Zion loads implies considerable uncertainty in esti-l n.ating the probability of early containment failure. Most of this uncertainty f-is given by accidents in which the primary system pressure remains high ime-j diately before vessel breach.

For these accidents direct heating of the con-tainment atmosphere by the core debris or hydrogen combustion with a steam spike at the time of reactor vessel failure have been postulated as mechanisms l

which could fail the containment.

The applicant has presented information which indicates that these mechanisms are not credible ways of failing the l

Seabrook containment.

However, as BNL has not developed Seabrook-specific h

l

I 4-6 containment loads, we cannot confirm that the uncertainty associated with pre-dicting the probability of early containment failure at Seabrook is as low as that claimed by the applicant.

4.4 References 1.

"Seabrook Station Emergency Planning Sensitivity," PLG-0465, April 1986.

2.

U.S. Nuclear Regulatory Commission, " Reactor Safety Study:

An Assessment of ' Accident Risks in U.S. Commercial Nuclear.. Power Plants," WASH-1400, NUREG-75/014, October 1975.

3.

"Seabrook Station Probabilistic Safety Assessment," Pickard, Lowe and Garrick, Inc., PLG-0300, December 1983.

4.

Khatib-Rahbar, M., et al., "A Review of the Seabrook Station Probabilis-tic Safety Assessment: Containment Failure Modes and Radiological Source Terms," NUREG/CR-4540, BNL/NUREG-51961, February 1986.

5.

" Review and Evaluation of the Zion Probabilistic Safety Study," NUREG/CR-3300, Volume 2, July 1983.

6.

" Zion Probabilistic Safety Study," Commonwealth Edison Company, September 1981.

7.

" Estimates of Early Containment Loads from Core Melt Accidents," NUREG-1079, Draft Report for Comment, December 1985.

8.

"A Review of the Current Understanding of the Potential for Containment Failure from In-Vessel Steam Explosions," NUREG-1116 June 1985.

9.

" Evaluation of Severe Accident Risks and Potential for Risk Reduction,"

NUREG/CR-4551, Volume 5, February 1986.

10.

" Reactor Risk Reference Document,"

NUREG-1150, Draft for Comment, February 1987.

11. " Evaluation of Severe Accident Risks and Potential for Risk Reduction,"

NUREG/CR-4551, Volumes 1-4, February 1986.

12.

PSNH Letter (SBN-1237), dated November 21, 1986, " Emergency Planning Sensitivity Study," J. DeVincentis to S. M. Long.

l

.3 s

4-7 Table 4.1 Comparison of Core Melt Frequencies and Distribution of Release Types (reproduced from Table 2.2. of PLG-0465)

Risk Parameter -

WASH-1400 Updated PWR SSPSA Results*

Mean Core Melt Frequency-(events 9.9-5**

2.3-4 2.7-4 e

per reactor-year)

Percent Contribution of Release e

Types Gross, Early Containment 34 1

0.1 Failure Gradual Containment 66 73 60 0verpressur'ization or Melt-Through Containment Intact 0

26 40

• 8ased on RMEPS (PLG-0432).

• • 8ased on WASH-1400 uncertainty ranges.

NOTE: Exponential notation is indicated in abbreviated form; i.e., 9.9-5 = 9.9 x 10-5, i

i J

. _.. =.

s 5-1 l

5.

REVIEW OF SOURCE TERMS

]

1

=In.this section the fission product source terms developed for PLG-0465 1

are reviewed.

The source terms used in PLG-0465 are reproduced in Table 5.1.

The probabilities of each of the source terms are given in Table 5.2.

5.1 Fidelity to WASH-1400 Methodology The fission product release fractions in Table 5.1 were determined by the 2

applicant using RSS methodology.

These source terms are consistent with the point-estimate source terms used in the SSPSA.3 The SSPSA source terms were 4

reviewed by BNL in NUREG/CR-4540 and they were found to be reasonable given 5

the limitations of the RSS methodology (principally the CORRAL code).

When reviewing the.PLG-0465 source te rms ', questions were raised and p

transmitted to the applicant.

One question related to release category S2W (refer to Table 5.1).

The fractional release of the noble gases for S2W was quoted as 0.123 whereas the release fractions for Cs and Te were quoted as 0.2 and 0.19 respectively.

It appeared inconsistent to release more aerosols (Cs and Te) than noble gas, and the applicant was requested to either explain the predictions or provide revised source terms.

In reference 6, the applicant provided a response to this question.

Basically, the noble gas release in Table 4.3 (reproduced in Table 5.1 of this report) of PLG-0465 was a mis-print. The noble gas release fraction for the S2W-3 release phase should have l

been 0.23 rather than 0.023.

However, as the S2W-3 release phase occurs very i

late (approximately 20 hours2.314815e-4 days <br />0.00556 hours <br />3.306878e-5 weeks <br />7.61e-6 months <br />) after reactor scram, the impact on risk of this l

larger noble gas release would be very small.

Calculations performed at BNL have verified that the impact of increasing the noble gas release in the S2W-3 phase on both the dose versus distance curves and the risk profiles is negli-gible.

5.2 Credit for Scrubbing of Submerged Releases l

Another question related to the credit assumed for the interfacing system i

LOCA source term in which the break location was assumed to be submerged under water (S7W in Table 5.1).

In PLG-0465, the source term mitigation resulting 1

5-2 from a subcooled 30 foot deep pool. was modeled as a decontamination factor (DF) of 1000 for all release fractions except the noble gases.

In the RSS, a decontamination factor of 100 wastused for fission product scrubbing in a sub-S cooled pool.

Thus, in order to be consistent with the RSS methodology, it appeared that a lower DF should be used.

However, if the pool were indeed subcooled, calculations at BNL indicated that using a DF of 100 rather than 1000 would have no impact on the risk calculations presented in' PLG-0465.

A more important question was whether or not the pool would.be subcooled or saturated.

In the RSS, no credit was given for fission product scrubbing in a saturated pool, and therefore, the applicant was requested to provide justification for the subcooled assumption.

In reference 7, the applicant provided a response to this question.

Arguments were provided to indicate that the pool would be at least 10*C subcooled and that this degree of sub-cooling together with the large pool depth was sufficient to justify a DF of 1000 rather than a DF of 100 was used in the RSS.

However, the objective of the question was primarily to determine if the pool would be subcooled and based on the response, this appears to be the case. BNL had already concluded that changing the DF. from 1000 to 100 would not change the dose versus distance nor the risk profiles.

Finally, the conclusion (given in reference 7). that even if pool decontamination were completely ignored, that the dose versus d.istance and the risk profiles in PLG-0465 would not be significantly effected was examined at. BNL.

Calculations at BNL indicated that if the fre-quency of interfacing system LOCAs reported in reference 1 was used, then the conclusion was correct.

However, if the revised frequencies for interfacing systems LOCA suggested in Section 2 by the BNL review were used, then com-plately ignoring pool decontamination would impact the risk estimates.

5.3 Uncertainties The appilcant considers the WASH-1400 source terms used in PLG-0465 to be very conservative and the applicant has high confidence that the source terms would not be exceeded in a real accident.

BNL found the source terms used in PLG-0465 to be consistent with WASH-1400 methodology but we are not as confi-dent as the applicant that they could not be exceeded.

The new source term methods (refer to NUREG/CR-4551,s Volumes 1-5) indicate that if the

5-3 containment fails late or if there is gradual leakage from containment then the aerosol fission product release is likely to be lower than would be pre-dicted by WASH-1400 methods. This is because WASH-1400 methods underpredicte'd aerosol agglomeration and settling.

Therefore, if the new methods were applied to release categories S2W and S6W, the predicted aerosol release would be lower than WASH-1400 values.

However, the new methods also indicate that if containment fails early and the opening is large, then there is consider-able uncertainty associated with predicting fission product release.

The uncertainty ranges associated with fission product release in NUREG/CR-4551 can, for certain accident sequences and early containment failure modes, exceed the WASH-1400 predictions.

This uncertainty would principally affect the S1W release category at Seabrook.

5.4 Summary In summary, the fission product source terms used in PLG-0465 appear in general to be consistent with the approaches used in the RSS. The misprint in Table 4.3 of PLG-0465 related to the fraction of noble gas release was found by the applicant (and confirmed by BNL) to have negligible impact on the risk profiles or the dose versus distance profiles reported in PLG-0465.

Thus, the corrected noble gas-release fraction in the S2W release category would, by itself, not change the conclusions in PLG-0465.

In addition, the argument presented by the applicant that the water in the RHR vault is sufficiently subcooled to warrant consideration of significant decontamination appears reasonable.

Finally, the statement in the applicant's response that even if pool decontamination had been ignored, the risk profiles or dose versus distance profiles would not change significantly was confirmed at BNL.

Note, however, that this conclusion is based on the frequency estimates for inter-facing system LOCAs in PLG-0465.

5.5 References l

t 1.

"Seabrook Station Emergency Planning Sensitivity Study, PLG-0465, April 1986.

5-4 2.

U.S. Nuclear Regulatory Commission, " Reactor Safety Study:

An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants," WASH-1400, NUREG-75/014, October 1975.

3.

"Seabrook Station Probabilistic Safety Assessment," PLG-0300, December 1983.

4.

Khatib-Rahbar M.,

et al., "A Revie.w of the Seabrook Station Probabilistic Safety Assessment:

Containment Failure Modes and Radiological Source Terms," NUREG/CR-4540, February 1986.

5.

Burian, R.

J.,

and Cybulskis, P.,

" CORRAL 2 User's Manual," Battelle Columbus Laboratories, January 1977.

6.

PSNH letter (SBN-1234), dated November 17,1986, " Response to Request for Additional Information (RAls)," J. DeVincentis to S. M. Long.

7.

PSNH Letter (SBN-1227), dated November 7,1986, " Response to Request for Additional Information (RAls)," J. DeVincentis to S. M. Long.

r

Table 5.1 Release Categories for Seabrook Station Based on WASH-1400 Source Term Methodology

" " I

• • '9Y

• ** " I'**

n Release Time Duration Time Release

• **W (hours)

(hours)

(hours)

(MCA/5)

XE 0.1.

1-2 CS

'TE 8A RU LA SIW 2.5 0.5 1.0 11.9 0.9

'7-3

.7-

.5

.3

.06

.02 4-3 52W-1 4.8 2.0 0.5 0

.03 2.1-4 4.3-3

.023 4.2-3 2.8-3 8.4-4 8.4-5 52W-2 6.8 4.0 2.5 0

.07 5.0-4 1,3-3

.048

.039 5.5-3 3.4-3 5.2-4 52W-3 19.8 18.0 15.5 0

.023 1.6-3 2.3-3

.126

.147-

.014

.011-1.9-3 TOTAL 4.8 24.0 0.5 0

.123 2.3-3 7.9-3

.20

.19

.022

.015 2.5-3 S3W 6.0 24.0 2.0 0

4.7-4 3.3-6 3.2-5 1.7-4 1.5-4 1.9-5 1.2-5 2.0-6 56W-1 1.75 1.0 1.5 0

.15 1.1-3

.10

.11

.02

.014 4.1-3 4.1 -4 56W-2 2.75 4.0 2.5 0

42 2.9-3

.07

.19

.063

.022

.009

.001 56W-3 15.75 18.5 15.5 0

.32 2.2-3

.01

.13

.32

.011

.020 3.8-3 o,

TOTAL 1.75 23.5 1.5 0

.0 6.2-3

.18

.43

.40

. 047

.033 5.2-3 57W 8.5 7.0 2.0 0

.9 7-6 7-4 5-4 3-4' 6-5 2-5 4-6 NOTE: Exponential notation is indicated in abbreviated form; f.e.

7-3 = 7 x 10-3 m

5-6 Table 5.2 Revised C-Matrix for New Source Term Categories

/

Plant G

Damage Source Term Category s

State S1 S2 S3 SS 56 37 (frequency) 1 x

1F 1.0 (2.0-8)

(2.0-8) 1FV 1.0 (4.6-9)

(4.6-9)

IFP 1.0 (1.4-6)

.(1,4-6)

IFPV 1.0

( 2.7-8 )

(2.7-8) 2A 3.4-5 1.4-4 1.0-2 U.99 e

(1.9-6)

(6.5-11)

(2.7 10)

(1.9-8)

(1.9-6) 30/70 2.0-6 8.0-5 0.95 0.0S J

(3.8-5)

(7.6-11)

(3.0-9)

( 3.6-S),

(1.9-6) l 3F/7F 1.0 j

( 3.0-7 )

( 3.0-7 )

3FP/7FP 1.0

\\

(1.9-5)

(1.9.5) 4A/8A 3.1-6 1.3-4 5.2-3 0.997 (1.1-4)

(3.3-10)

(1.4-8)

(5.5-7)

( 1.1.-4 )

I 7FPV 1.0 (1.2-8)

(1.2-8) 80 1.1 '6 3.1-5 0.9999 (1.0-4)

( l'.1-10 )

(3.2-9)

(1.0-4)

Total 5.2-9 2.0-5 1.4-4 1.1-4 3.2-7 3.9-8 Frequency NOTES:

1. Exponential notation is indicated in abbreviated form; 1.e., 2.0-8 = 2.0 x 10-8

2. ilumners inside parenthests are unconditional frequencies (events per reactor year) based on mean values. Numbers not inside parentheses are conditional' frequencies of source term categories, given the indicated plant damage stato, also based on mean values.

Medlan values of source term categories are present.ed in Section 3.

~

i

--. =

t t l' n

f 6.

SITE CONSEQUENCE MODEL e.

6.1 NOREG-0396 Basis V

NUREG20396 1introduced the concept of generic Emergency Planning; Zones (EPZ) as a basis-for the planning of response actions which would result -in dose savings in the immediate vicinity of.. nuclear facilities-in'the event of-a serious power reactor accident.

The actions would be triggered if projected radiation deses to an individual would exceed Protective Action Guides (PAGs),

-as discussed and referenced in NUREG-0396, although ad hoc actions could be-taken'at any time.

The PAGs are 1 to 5 rem whole body dose and 5 to 25 rem thyroid dos, but are n9t intended to represent acceptable dose levels.

Fur-thermore, protective actions may not assure that PAG. levels can be prevented.

p';

It was concluded in NUREG-0396 that the objective of emergency. response plans sh'ould be to provide dose savings for a spectrum of accidents that could produce offsite doses in excess of the PAGs since no specific accident could be identified as the one for which to plan.

Tne most important guidance for emergency -planners is the size of the EPZ.

Based os factors that included risk, probability.. and accident conse-quencer, it y3s. judged that a generic distance of about 10 miles was appropri-ate

• for core melt accidents.

The less severe accidents would not have conse-quences in excess of the PAGs beyond this distance, whereas the. more severe fA b accidents would not in general cause early injuries or deaths beyond this dis-

[

tance.

Nence, protectivo actions, were judged to be most useful within this

}#

distance.

l-NUREG-0396 used the accident release categories of the Reactor Safety I

Study (RSS) to compute the risk of exceeding various dose levels 2

in the i

absence of protective actions for a spectrum of accidents.

The RSS accidents and their median probabilities are given in Table 6.1.

Using the original RSS consequence model (CRAC)3 and the accidents PWR1 through PWR7, a 200 rem whole body risk curve was constructed, shown as the heavy line marked 0396 in Figure 6.1 This is the level at which serious injuries and sone deaths can occur.

l

\\

._m.-__-._..._..___._____,_._

6-2 W

o However, CRAC did not compute the 200 rem risk directly.

The authors of NUREG-0396 had to interpolate to obtain the 200 rem curves.

Interpolation was performed for each component accident to obtain the conditional risk given the accident.

Then the conditional risk at various distances was multiplied by the probability shown in Table 6.1 and divided by the core melt probability of 5x10 5 per reactor year. The results for each were gummed to give the overall risk of the accident spectrum. Each risk was computed using about 100 weather samples from a typical series of New York City hourly annual weather data with the assumption that people would follow normal activities for one day follow-ing arrival of the first plume to reach their location. That is, people would stay at their original location and receive groundshine doses for 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br />.

4 BNL recomputed the 200 rem dose vs distance curve using an updated ver-sion of CRAC, called CRAC2," and a more detailed grid. Theconditiogicompo-nent risks were computed directly (no interpolation required) and the results for each of the WASH-1400 release categories.are shown in Figure 6.1.

PWR6 and PWR7 did not exceed 200 rem.

,7he overall 200 rem risk curve was then computed using a core melt proba-

"bility of 6x10 5 per reactor year, which is the sum of the probabilities for PWR1 - PWR7 given in Table 6.1.

The curve gives higher risk for 1-3 miles and lower ri.sk for 4-10 miles than NUREG-0396, but the curve still drops sharply beyond about 10 miles.

It should be noted that most of the core me'lt proba-bility comes from PWR7 which does not contribute to the 200 rem curve.

It can be concluded that the approximations used in NUREG-0396 are not substantially different from the more detailed calculations done by BNL using CRAC2.

6.2 Consequence Modeling 5

The applicant used the CRACIT code for their consequence assessments in PLG-0465.10 In this section CRACIT predictions are compared against conse-quences mdels currently being $ sed by the NRC and their contractors, namely CRAC, CRAC2, and 'NACCS. 6 The factors involved in consequence modeling are discussed in Appendix 6 of the RSS.

All codes compute early and delayed health effects from cloudshine, inhalation, and groundshine. The early health l

effects are based on data from the Marshall Islands, bomb tests, clinical data i

l

6-3 from radiation therapy, and lab animals (particularly for lung data).

The three CRAC models (CRACIT, CRAC2, and CRAC) use a stepwise linear function with a threshold dose for early effects as discussed in the RSS.

The MACCS code (recently developed at Sandia National Lab) uses a hazard function i

approach without a threshold as discussed in NUREG/CR-4214.

The latent effects in the CRAC models are calculated from the BEIR-1 reports which uses Japanese data plus a modification to the. linear dose response curve to account for reduced effectiveness at low doses. MACCS uses the BEIR-3 model 8 4

which is i

a linear quadratic dose response model with absolute risk of cancer for some organs such as bone marrow and relative risk for other organs, depending on population makeup.

In addition, CRAC and CRAC2 allow only a one " puff" release of radioactivity whereas CRACIT and MACCS allow "multipuff" releases.

There are also other differences in the codes, such as the shape of the plume, dry and wet deposition of particulates, weather sampling, resuspension, etc.

which can account for differences of a factor of two in the results.

BNL used the MACCS code and the source terms defined in Table 5.1 to review the calculations presented by Seabrook in PLG-0465 10 using CRACIT.

The comparisons are based 'on the 200 rem dose probability vs distance curves using the source terms and weather data supplied by Seabrook.

In addition, BNL cal-culated the individual risk of exceeding 5, 25, and 300 rem to the thyroid and the individual risk of death as a function of distance.

In all cases, it was assumed that the population was exposed to one day of groundshine following arrival of the first plume segment.

They would also be exposed to other plumes that arrived at their location within the 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br />.

6.2.1 Whole Body Dose Vs Distance The MACCS code does not have an organ defined as "whole body" so red marrow was used as a substitute.

In CRAC2 calculations it was found that the red marrow dose was about 30% higher than the whole body dose.

Also, early health effects are sensitive to the red marrow dose.

Thus, the red marrow dose is a good substitute.

MACCS does not directly calculate the risk of a dose vs distance, and it was necessary to define an appropriate risk function to obtain this curve.

In I

6-4 the MACCS calculations, the risk of exceeding 200 rem to the red marrow was set to zero if a weather sample yields a mean dose less than 200 rem and one if a weather sample yields a mean dose greater than 200 rem.

Thus the mean risk is dependent upon the annual weather data. The risks were calculated for the Seabrook release categories S1W (one puff), 52W (3 puffs), and S6W (3 puffs). The results are given in Figures 6.2, 6.3, and 6.4.

The conditional probability of.001 risks extend to 25 miles for S1W, to 2 niles for S2W, and to 4 miles for S6W.

These distances are somewhat less than those calculated by Seabrook.

6.2.2 Thyroid Dose Vs Distance The thyroid doses were not presented by Seabrook in the reviewed report but were discussed at some length in NUREG-0396.

Hence, BNL calculated the risk of exceeding the dose levels of 5, 25, and 300 rem to the thyr'oid, as was done in NUREG-0396 for Seabrook release categories S1W, S2W, and S6W.

The results are given in Figures 6.2, 6.3, and 6.4.

The results were truncated at 30 miles.

The risk of exceeding 5 rem remained above 90% for all three release categories.

The 300 rem curve shows a sharp drop at less than 10 miles for S2W and at about 15 miles for S6W.

The curves were obtained by the same hazard function definition technique as discussed in Section 6.2.1.

6.2.3 Risk of Early Fatalities MACCS uses the hazard function approach to calculate early fatalities as discussed in NUREG/CR-4214. First, the cumulative hazard is calculated as:

H = In(2) (D/D50)#

(6.1) where D is the dose and 05a is the dose required for producing an effect in 50% of the exposed individuals, and v determines the steepness of the dose effect curve.

The fatality risk is then given as:

Risk = 1 - e-IN I+H2+H3 + Hg)

(6.2)

+

s

6-5 where H is fpr red marrow, H i

2 is for lungs, and H3 and Hg are for the lower large intestine and small intestine.

The risk is assigned a threshold of

.005.

i In CRAC2 and CRACIT, the dose response is piece-wise linear due to irrad-iation of the bone marrow, lung and GI tract. The total risk is then:

R=R i + (1-R )R 2 + (1-R )(1-R )R i

2 3 (6.3) i where R, R, and R3 are the risks to the three organs, respectively.

MACCS i

2 gives somewhat higher risk, principally because the lung dose is now consid-ered more effective in producing fatalities, and also because the hazard func-tion gives some risk at lower doses.

The eff;ct of the code differences is that MACCS predicts a higher p.oba-bility of a small number of deaths while CRAC2 predicts a higher probability for large numbers of deaths.

This is shown in Figure 6.8 from a comparison calculation performed by Sandia National Laboratory for a severe ground level release.

This is. for a uniform population distribution without evacuation.

However, MACCS can predict substantially more early deaths when-evacuation is modeled since the lung dose usually becomes dominant, but evacuation scenarios are not. considered in this review.

BNL calculated.the individual risk of fatalities versus distances for S1W, S2W, and S6W as shown in Figures 6.2, 6.3, and 6.4.

It can be seen that the risk is similar to the 200 rem curve, but is not directly correlated because of the nonlinear interactions in the above equations.

6.3 Comparisons of Results 6.3.1 Results of Seabrook Study l

BNL used as a basis for comparison the 200 rem risk of whole body dose as l

calculated by Seabrook using CRACIT and 200 rem red marrow risk using MACCS at BNL.

This is the risk to a hypothetical individucl located at a particular distance and actual population distribution is not considered.

The BNL L

i

6-6 calculations were performed with meteorological data supplied by Seabrook.

Projected numbers of fatalities were not computed since BNL didn't have the actual Seabrook population data.

The Seabrook calculations for accidents S1W, S2W, and S6W are shown in Figures 6.5, 6.6, and 6.7 with the corresponding BNL results superimposed.

For S1W, MACCS predicts slightly higher risk g at less than 8 miles and much lower risk beyond 12 miles.

The differences may be partly explained by' differences in weather sampling methods and plume rise formulations, since this is a high energy release.

However, the differences in the tail of the risk curve is not considered by BNL to be significant considering the overall uncertainties in the calculation.

For S2W, the MACCS code again predicts higher risk in close and a some-what sharper dropoff in the tail.

However, the difference is that MACCS predicts a risk of.001 at 2 miles whereas the Seabrook results show this risk at 2.5 miles.

For S6W, thE. BNL results are close to those of Seabrook as in the case of S2W. MACCS predicts.001 risk at 4 miles whereas Seabrook predicts this risk of 200 rem at 6 miles.

In summary, BNL feels that the dose versus distance modeling is fairly presented by Seabrook and that the relatively small differences computed by RNL are probably explained by modeling techniques.

6.4 Sensitivity Studies Two categories of sensitivity studies have been performed as part of the BNL review.

. Fi rst, sensitivity calculations were performed to assess the affect of the duration of fission product release on the dose vs distance curves presented in PLG-0465.

These sensitivity calculations are discussed in Section 6.4.1.

Second, the impact on the dose vs distance and risk estimates in PLG-0465 of the various concerns raised by the BNL review was assessed in each section of this technical evaluation report.

These revised risk esti-mates are summarized in Section 6.4.2.

f 6-7 6.4.1 - Sensitivity of Results to Multipuff Release BNL performed ' sensitivity-calculations.with regard to the multipuff releases and the release duration for S2W and S6W.

The results are given in Figures 6.6 and 6.7, respectively using the one puff release categories defined by Seabrook.

In both cases it was found that a one puff release increased the risk and also that a shorter duration of the release (0.5 hours5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br />)'

further increased the risk.

For the one puff, 0.5 hour5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br /> release S2W, the 200 rem 001 risk distance increased from 2 miles to 7 miles and the S6W distance increased from.4 miles to 15 miles.

This demonstrates that a long release duration leads to a greater plume dilution and less risk at larger distances.

Hence, in order to have confidence in the Seabrook calculations, one must have confidence that ~ the releases will occur with rates and durations similar to those used by Seabrook.

6.4.2 Summary In the following sections, the BNL findings related to each review area are briefly summarized.

BNL has attempted to follow the ground rules for comparison. purposes discussed in Section 1 (namely mean frequencies for comparison with the safety goal and median frequencies for comparison with NUREG-0396 results), however, we have also attempted to include 1 discussion on the uncertainties associated with the risk estimates.

6.4.2.1 Interfacing systems LOCA A major concern resulting from the BNL review of the interfacing systems i

LOCA analysis in PLG-0465 and the RMEPS related to the determination of initiator frequencies.

The effect of changing the initiator frequencies was I

determined by propagating the changes through the appropriate event trees in the RMEPS.

The revised initiator frequencies resulted in the 'following changes to the frequencies of release categories S1W and S7W.

Mean Frequency Per Reactor Year Release Category PLG-0465 BNL Review S1W 4.0x10 S 1.4x10 7 S7W 6.3x10-8 1.1x10-6 4

m

. =-

6-8 The above changes in release category frequencies have no impact on indi-vidual risk of early fatalities if no evacuation or 1 mile evacuation is assumed.

This is because release category S2W dominates this risk neasure, and it' has a frequency of 2x10 5 Only when a 2 mile evacuation is assumed

.(and the early fatality risk for category S2W becomes zero) do the above changes in release category frequencies change the original PLG-0465 esti-mates.

However, with a 2 mile evacuation the early fatality risk is very low and well below the safety goal.

The 200-ren dose versus distance curve in 10 is also not influenced by the above changes in release category PLG-0465 frequency. This is because only release category S1W has a significant proba-bility of-exceeding a 200-rem dose, and the revised probability of this cate-gory is not sufficiently high for it to appear in the PLG-0465 dose versus distance curves.

^

There is of course uncertainty associated with predicting the frequency of interfacing systems LOCAs.

However, the frequency of interfacing systems LOCAs resulting in release category S1W would have to increase by two orders of magnitude before the Seabrook dose versus distance curves would approach the curves given in NUREG-0396.

One can therefore conclude that interfacing i

systems LOCA are unlikely to increase the risk profiles presented in PLG-0465 to the level presented in NUREG-0396. This is not too surprising because when no evacuation is assumed, the higher frequency events dominate risk and inter-facing systems LOCAs did not contribute to the dose versus distance curves taken from NUREG-0396.

6.4.2.2 Accidents during shutdown This topic was not originally addressed in PLG-0465 and a detailed assessment of such events is beyond the scope of the current BNL work on this project.

However, the applicant was requested to provide information on the risk associated with accidents during shutdown.

The results of the appli-cant's assessment of such accidents were presented in the form of sensitivity i

studies in a draft version of this report.

The applicant provided additional

)

frequencies to the existing release category frequencies given in PLG-0465 to assess the impact on risk from accidents during shutdown.

A base case and a bounding. case were presented by the applicant.

The additional frequencies 1

~

6-9 associated with tnese accidents are given below:

Mean Frequency Per Reactor Year Release Category Power Base Case Bounding Operation Events Shutdown Events Shutdown Events S.5 1.1x10 4 1.7x10 5 S.2 2.1x10-5 4.9x10 7 S.6 6.5x10 7 7.1x10 8 5x10 6 BNL was not in a position to assess the above frequencies for these events because there remained fundamental questions regarding the modeling of these scenarios.

However, in spite of this, the applicant's results were included in the draft report for comparison with the BNL sensitivity study results on other topics.

It should be noted that the applicant considered the upper bound estimates to be very conservative.

In particular, in order to assess the impact of these events, they were included in source term cate-gories derived for accidents from full power, which could lead to predicts of shorter times and larger quantities of fission product elease than would be expected from accidents during shutdown.

In a subsequent submittal by the applicant, the consequences of accidents from shutdown were revised.

The applicant felt that 94 percent of accidents at shutdown would occur at times later than 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> after scram.

Thus, the consequence estimates were reanalyzed assuming release times of 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />.

The later release times resulted in dose versus distance curves which fall off at much shorter distances from the site boundary than the original dose versus distance curves.

BNL has checked this result and confi rmed that if the release does occur at times greater than 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />, then the new dose versus distance curves are reasonable.

The results of the latest applicant's assessment of accidents at shutdown are reproduced in Figure 6.9.

As noted above, a detailed assessment of such events is beyond the scope of the current BNL review.

However, based on our limited review of the applicant's assessment of these events, we still have reservations about the results.

These reservations are discussed in greater

6-10 detail in the body of this report but until they are resolved, we are unable to assess the validity of. the risk estimates presented by the applicant in Figure 6.9 4

6.4.2.3 Induced steam generator tube rupture For accidents in which the primary system is at high pressure during core uncovery and melting, it is possible that large natural circulation flow patterns could develop within the primary system.

These flow patterns could in turn heat-up and degrade regions of the primary system remote from the reactor core. Of particular concern is the possibility of degrading the steam generator tubes such that the primary system could become open to the secon-dary system.

If the secondary system were in turn open to atmosphere, then a direct path could exist between the primary system and the atmosphere, which bypasses containment.

This topic was not included as part of the work scope for the current BNL review. However, the topic has been reviewed in detail by the NRC staff and is the subject of continuing NRC and industry research activities. Therefore, BNL performed simple sensitivity studies to assess the potential impact of induced steam generator tube rupture on risk at Seabrook.

The results of the sensitivity study are given,in Figure 6.10.

Because of uncertainty in predicting events of this nature and the fact that BNL did not evaluate this issue in detail, we were not able to develop a best-estimate frequency for induced SGTR.

The sensitivity study therefore presents a range of possible frequencies for induced SGTR.

The frequency of high pressure squences and the conditional probabilitiee of failure of the operators to depressurize and induced SGTR that were used in the sensitivity i

study are given below:

4,0x10 5 x 0.5 x 0.3 = 6.0x10 6 per reactor year 4.0x10 5 x 0.2 x 0.01 = 8.0x10 8 per reactor year.

In order to estimate the impact of the above probabilities on risk, an appropriate source term category had to be selected.

It was decided to allo-cate SGTR events to release category S1W, which represents a large early bypass of the containment.

It was felt that this was a conservative

6-11 assumption because-significant retention of the fission products in the secondary side could occur and this was not considered when calculating the S1W release fractions.

The impact of adding the above frequencies to source term category S1W is illustrated in Figure 6.10 The lower estimate of the frequency of induced SGTR has no impact on the risk estimates presented in PLG-0465. The higher estimate of the frequency of induced SGTR has no influence on the individual risk of early fatalities within 1 mile of the site boundary if no evacuation is assumed but does influence the 200-rem dose versus distance curves as shown in Figure 6.10.

Allocating the probabilities of SGTR events to release category S1W has the largest impact on the dose versus distance curves.

However, the impact on the risk of early fatalities within 1 mile is negligible because S1W has very little risk of fatalities within this distance (refer to Section 1).

If the probabilities of SGTR events were added to release category S6W, the impact on the dose versus distance curves would be less but the risk of fatalities within 1 mile would increase slightly if no evacuation is assumed.

It should be noted that the range of frequencies used for the induced SGTR sensitivity study were developed to cover our lack of understanding in this area and that the NRC staff believes that the actual probability of a SGTR is closer to the lower estimate.

However, one reviewer ll of the BNL draft report felt SGTR to be a potentially more "significant" issue than was implied in our evaluation.

It was not BNL's intention in the draft report to l

minimize the potential importance of this issue, and the range we presented I

did not represent an upper bound.

It was an attempt to reflect the best t

judgments of several experts on a very difficult subject.

There is a great deal of uncertainty associated with predicting such events and it is therefore prudent to indicate the impact on risk of a range of assumptions.

6.4.2.4 Containment isolation failure and pre-existing leakage The applicant's assessment of pre-existing 14akage and containment isola-tion failure was reviewed by the NRC staff.

Based on the NRC staff review of the information available, it was concluded that the purge and vent valves in a fully closed configuration should provide reliable isolation of the Seabrook

6-12 containment under severe accident conditions up to the pressure corresponding to 1 percent hoop strain in the containment.

The NRC staff also concluded that the applicant has presented a reason-able approach for the consideration of pre-existing leaks, both small and large.

The approach adopted by the applicant was to use information on 12 containment unavailability developed in a study by the Pacific Northwest Laboratory (PNL) to assess the impact on risk of pre-existing leakage.

The applicant used this information to bound the effects of the data in the PNL study (NUREG/CR-4220) even though they considered that it did not apply to Seabrook.

The results of the applicant's assessment are given in Figure 6.11.

From an inspection of Figure 6.11, it is apparent that the impact of the NUREG/CR-4220 data on the dose versus distance curves in PLG-0465 is not great.

6.4.2.5 Containment structural capacity Based on its nonlinear finite element analysis of the Seabrook contain-

~

ment, BNL concluded that a shear failure at the base of the cylindrical wall is a potential failure mode but would not occur before reaching a pressure of 165 psig.

BNL agrees that the containment structure would reach a general yield state in the hoop reinforcing steel at a pressure of 157 psig and that it is appropriate to consider this pressure as a lower bound pressure for the hoop mode of failure.

However, BNL believes that the median hoop failure pressure should correspond to the one parcent strain level in the hoop reinforcing steel, which is a pressure of 175 psig.

The above pressures are for the wet containment conditions.

For the dry containment conditions the corresponding median failure pressure is 158 psig and the lower bound pressure (general yield) is estimated to be 145 psig.

This latter value is based on the reduc-tion factor recommendation in Section 11.3.4.1 of PLG-0300 With regard to containment p netrations, BNL believes that the failure pressures should be based on containment deformations assuming no bond strength between the reinforcing steel and concrete.

Based on this assumption

6-13 BNL estimates median failure pressures for the wet containment condition of 159 psig and 167 psig for two critical penetrations. -For the penetration with

.the. lower failure pressure, BNL agrees that a Type A (less than 6 square inches) leak path is appropriate for the median estimate; however a Type B (6 square inches to about 0.5 square foot) leak path should be considered as an upper bound estimate.

For the penetration with the higher failure pressure, BNL agrees that a Type B leak path is appropriate for the medium estimate; however, a Type C (greater than 0.5 square foot) should be considered as an upper bound estimate.

For the dry containment conditions, BNL estimated the median failure pressures for the above two critical penetrations to be 147 psig and 152 psig, respectively. These values are also based on the reduction factor recommended in Section 11.3.4.1 of PLG-0300 Although BNL has performed some independent calculations to support its conclusions regarding the containment strength, it also relied on the results of calculations performed by PSNH and its contractors.

Therefore, BNL recom-mends that a complete and independent check of all relevant containment strength calculations be performed by PSNH.

PSNH committed to such a~ check in their letter to the NRC dated October 31, 1986'and has indicated that such a check has been completed.

6.4.2.6 Containment loads BNL's assessment of the capacity of the Seabrook containment (described above) has to be combined with severe accident loads (pressure / temperature

)

histories) to determine the potential for early containment failure. BNL does not have Seabrook-specific containment loads and was not able to generate such loads given the limited scope of the current review.

However, BNL has been involved in updating (NUREG/CR-4551,13 Volume 5) the risk profile for the Zion plant for input to the NRC's " Reactor' Risk Reference Document," NUREG-1150 The updating of risk for Zion was based on a methodology developed as part of the Severe Accident Risk. Reduction Program (NUREG/CR-4551, Volumes 1-4) at Sandia National Laboratory (SNL). This methodology used expert judgment in an attempt to estimate the uncertainty associated with determining containment i

e m.-

6-14 loads. The methodology was developed at SNL specifically for the Surry plant but was extrapolated to Zion at BNL The Zion plant is very similar to ~ Sea-brook in terms of the containment volume to reactor power ratio.

Thus, extrapolation of the Zion loads to Seabrook would give some indication of the impact of applying this new methodology to Seabrook.

It nust be emphasized that this exercise should in no way be interpreted as a Seabrook-specific cal-culation.

It simply gives some indication of:the sensitivity of the Seabrook results to the types of uncertainty in estimating containment loads discussed in NUREG-1150.

It should also be noted that this work is preliminary and has not yet undergone full peer review outside of NRC and its contractors.

It is, therefore, subject to revision.

The range of containment loads reported in Volume 5 of NUREG/CR-4551 for Zion is very wide and far exceeds the loads that would be considered credible by the applicant for Seabrook.

Of particular interest is the loads at the time of reactor pressure vessel failure.

These loads can range from about 60 psia to 200 psia depending on whether core melt is occurring with the primary system at high or low pressure and on whether or not containment heat removal systems, CHRS (sprays and fan coolers) are operating.

The higher containment loads are postulated to occur for accidents in which the primary system pres-sure remains high immediately before reactor pressure vessel failure.

For these accidents, direct heating of the containment atmosphere by core debris or hydrogen combustion with a steam spike at the time of reactor vessel fail-ure are possible mechanisms for failing the containment.

The applicant has presented information which indicates that these mechanisms are not credible ways of failing the Seabrook containment.

However, as noted above, BNL does not have Seabrook-specific containment loads so we.cannot, at this time, elim-inate these mechanisms as potential ways of failing the Seabrook containment.

l For accidents with the primary system at high pressure and without the CHRS operating an approximate median load of 135 psia (120 psig) was predicted for Zion.

If this median load is compared against the capacity of the Sea-brook containment given by the BNL review, one would conclude that the poten-tial for early containment failure at Seabrook is very low and would not l

influence the risk estimate in PLG-0465.

However, the range of loads esti-mated for Zion implies considerable uncertainty. The 95th percentile estimate

.a g

m 4a l

6-15 of the probability of early containment failure at Zion is quoted as 0.17 in

\\;.ume 5 of NUREG/CR-4551.

If this early containment failure probability were also true for Seabrook, the risk estimates in Figures PLG-0465 would increase i

significantly.

However, the capacity of the Seabrook containment is greater than Zion (the general yield for Seabrook is 157 psig compared with 134 psig for Zion) so the 95th percentile estimate of early containment failure should be lower at Seabrook than Zion. However, BNL cannot at this time quantify how much lower because we have not quantified Seabrook-specific containment event trees with Seabrook-specific containment loads combined with our estimate of the structural capability of the Seabrook containment.

6.4.2.7 Source terms The fission product source terms used in PLG-0465 10 were reviewed in terms of their consistency with the approaches used in WASH-1400 and found to 2

be appropriate.- A misprint in PLG-0465 related to the release of noble gases for release category S2W was discovered.

However, correcting the noble gases release was found to have no impact on the risk profiles in PLG-0465 In addition, the argument presented by the applicant that water in the residual heat removal (RHR) vault is sufficiently subcooled to warrant consideration' of significant decontamination was found to be reasonable.

This is an important consideration for the subset of interfacing systems LOCAs where the break location in the RHR line is low in the RHR vault.

Under these circumstances, with the break location submerged considerable scrubbing of the aerosol fission products would occur. This would result in much lower aerosol fission product release than for accidents in which the break location was uncovered.

The applicant considers the WASH-1400 source terms used in PLG-0465 to be J

very conservative and the applicant has high confidence that the source terms would not be exceeded in a real accident.

BNL found the source terms used in PLG-0465 to be consistent with WASH-1400 methodology but we are not as con-fident as the applicant that they could not be exceeded.

The new source tern methods (refer to NUREG/CR-4551, Volumes 1-5) indicate that if the containment fails late or if there is gradual leakage from containment then the aerosol fission product release is likely to be lower than would be predicted by WASH-1400 methods.

This is because WASH-1400 methods underpredicted aerosol i

~

F 6-16

[

agglomeration and settling.

Therefore, if the new methods were applied to release. categories S2W and S6W, the. predicted aerosol release would be lower than WASH-1400 values.

However, the new methods also indicate that if con-1-

tainment fails early and the - opening is large, then there is considerable-uncertainty associated with predicting fission product release.

The uncer-f tainty ranges associated with fission product release in NUREG/CR-4551 can,'

for certain accident sequences and early containment failure modes, exceed the WASH-1400 ' predictions.

This uncertainty would principally affect the S1W release category at Seabrook.

6.4.2.8 Consequence Model i

The applicant used the CRACIT code for their consequence assessments in PLG-0465.

BNL compared CRACIT predictions of dose versus distance with predictions.from the MACCS code, which was developed at Sandia National Laboratory (SNL) under NRC sponsorship.

The comparison -of the dose versus distance curves for the CRACIT and MACCS codes was reasonably good.

There-i fore, BNL feels that the dose versus distance modeling in PLG-0465 is fairly i

presented and that the relatively small differences between CRACIT predictions and those computed by BNL using MACCS are explained by differences in modeling techniques used in the two codes.

BNL could not check the risk of early fatalities reported in PLG-0465 because we did not have the population distribution around the Seabrook site.

Therefore, as BNL had only CRACIT results for early fatalities, it was decided to use CRACIT results. for both early fatality risk and dose versus distances in the BNL sensitivity study. This was done simply so that we had consistency between the two risk measures and not (as implied by the applicant's review of the draft BNL report) to present more " conservative" CRACIT results. We found that CRACIT in general predicted dose versus distance curves that extended l

further than the MACCS code and in this sense CRACIT is more " conservative" than MACCS. However, we note that MACCS predicts more early health risk than CRACIT and therefore the use of the CRACIT results is probably not "conserva-tive" for this risk measure.

In addition, CRACIT predicted that only 9 early fatalities would occur within 1 mile of Seabrook if release category S1W was to occur compared with an estimated total early fatality risk of 746 (refer to

6-17 Table 1.?).

The reason for the very small risk of early fatalities close to the Seabrook site boundary for the S1W release category is probably due to the plune lift-off model in CRACIT.

The S1W release category has a high energy plume and CRACIT would calculate significant elevation of the S1W plume rela-tive to the other release categories.

This reduces the probability of early fatalities close to the site but increases the probability of early fatalities f

at larger distances.

Also, the weather sampling in' CRACIT can result in increased risk of early fatalities at greater distance from the site depending on when rain is predicted to occur.

Sensitivity studies using the MACCS code at BNL indicate much less sensitivity to high energy releases than CRACIT.

It is therefore likely that if BNL had calculated early fatalities for the S1W release category using MACCS with the actual Seabrook population distribution we would have calculated more early fatalities closer to the Seabrook site boundary than predicted by CRACIT.

This statement is supported by the com-parison of the CRACIT and MACCS 200 rem dose versus distance curves in Figure 6.5, which indicate that MACCS has a higher probability of exceeding this dose level cose to the site boundary than CRACIT.

This difference in MACCS and CRACIT predictions has important implications for. comparison. with the safety goal, which deals with the risk of early fatalities in the population within one mile of the site boundary.

A major conclusion of the BNL sensitivity studies is that the safety goal compariosn is relatively insensitive to uncer-

~

tainties in estimating the frequency of release category S1W.

However, this conclusion is based on the applicant's CRACIT calculations, which predict a very low risk of early fatalities close to the site boundary.

It is not clear that if BNL had performed MACCS calculations and used these predictions in the BNL sensitivity studies that the safety goal comparison would have been so f avorable.

This is an area that requires further investigation in any follow-on effort.

6.5 References 1.

Collins, H.

E., et al., " Planning Basis for the Development of State and Local Government Radiological Emergency Response Plans i.n Support of Light Water Nuclear Power Plants," prepared for the U.S.

Nuclear Regulatory Commission, NUREG-0396, December 1978.

6-18 2.

U.S. Nuclear Regulatory Commission, " Reactor Safety Study: An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants," WASH-1400, NUREG-75/014, October 1975.

3.

Reactor Safety-Study Consequence Model, " Computer Code Users Manual,"

U.S. Nuclear Regulatory Commission, undated.

'4.

Ritchie, L.

T., et al., " Calculations of Reactor Accident Consequences Version 2.CRAC2:

Computer. Code Users Guide," NUREG/CR-2326, February 1983.

5.

Pickard, Lowe, and Garrick, Inc., "Seabrook Station Probabilistic' Safety Assessment," prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, PLG-0300, December 1983.

6.

Chanin, D.

I.,

Ritchie, L.

I., and Alpert, D.

J., "MELCOR Accident Conse-quence Code System MACCS User's Guide," Sandia National Lahoratories, Albuquerque, NM, to be published.

7.

Evans, J.

S.,

et al.,

" Health Effects Model for Nuclear Power Plant Consequence Analysis," NUREG/CR-4214, July 1985.

8..BEIR-1 Report (1972):

The Effects on Populations of Exposure to Low Levels of Ionizing. Radiation.

Report of the Advisory Committee on the Biological-Effects of Ionizing Radiation.

Division of Medical Sciences, National Academy of Sciences, National Research Council, Washington, DC.

9.

BEIR Report (1980): The Effects on Populations of Exposure'to Low Levels of Ionizing Radiation.

Report of the Advisory Committee on the Biologi-I.

cal Effects of Ionizing Radiation. Division of Medical Sciences, Nation-al Academy of Sciences, National Research Council, Washington, DC.

10..Pickard, Lowe, and Garrick, Inc., "Seabrook Station Emergency Planning Sensitivity Study," prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, PLG-0465, April 1986.

11. Theofanous, T.

G.,

" Review Comments," University of California, Santa Barbara, dated January 12, 1987.

12.

Pelto, P.

J., et al.,

" Reliability Analysis of Containment Isolation Systems," NUREG/CR-4220, PNL-5432, June 1985.

13.

Evaluation of Severe Accident Risks and Potential for Risk Reduction,"

NUREG/CR-4551, Volume 5, February 1986.

i I

6-19 1.

I 1A = PWRIA 5 Y I^

IB = PWRIB 2 = PWR2

}

3 = PWR3 4

4 = PWR4 5 = PWR5 S = Summary (6,7=0. Risk) m S 0.1 I

's

(

f3 52 g

L

.jo 0

'g k

I 8

i.

1 N

I w

\\

$.01 I

m 22

(

-5 31 1IL'

\\,

2y B

1A3 54 1

S

.001

,,,,I

,,,,,,i 1

10 100 Miles Figure 6.1 Components of NUREG-0396 curve as com-puted by BNL using CRAC2. curve is normalized to 6x10 ge sum T

core melt probability.

The result differs from NUREG-0396.

'6-20

- w f

K

}B A

C

's s

N k

's 0.1

\\

\\

\\

E

\\;

\\ D

.01r-

{

f' A: 5 Rem Thyroid g-- B: 25 Rem Thyroid g

- C: 300 Rem Thyroid

- D: 200 Rem Red Marrow }'

E: Death 1-i

\\

\\

,, i i

,1

- 001 e

l 1

10 100 Miles Figure 6.2 Risk of death or exceeding dose levels for S1W as calculated by BNL.

t r

6-21

.c

^

A

\\

t 0.1 :--

}

I

~

?

B L

C

. 01

~

A: 5 Rem Thyroid k

~

B: 25 Rem Thyroid IE C: 300 Rem Thyroid i

D: 200 Rem Red j

Marrow D

E: Death

}

.001

,t

,,,,1 1

10 100 Miles Figure 6.3 Risk of death or exceeding dose levels for S2W as calculated by BNL.

e 6-22 1.C A

I \\

B'

\\

N A: 5 Rem Thyroid

\\

B: 25 Rem Thyroid C: 300 Rem Thyroid I

D: 200 Rem Red Marrow

,\\

0.1 ---

E: Death

\\

\\

i l

.u

\\

\\

x

\\

i

,01-

\\

8 k

I

\\

p jE l

\\

I D

C

\\.

x

\\

001

.en'*I

.s i

3 to 100 Miles Figure.6.4 Risk of deaths or exceeding dose levels for 56W as calculated by BNL.

1 4

..--y,

_,,,.-.,y,_.,

6-23 16 ff

=

- N

~.-'s E

\\

3

-t

-t 5 16..

..18 N.

E f

=

5

?\\

E

-2 MACCS te --

200 rem

- '"2 g

E w

5

~]

k W

I

$ e-3.7

-3

>. t g

_.ts b

I a

2 x266.66 REM

{

cz 456.66 REN 45.66 REN or

.t.86 REr1 l

o.

4 I

4 18 18 18*1

'168

' 't S t 162 DISTANCE (MILES)

Figure 6.5 Dose versus distance curve for release category S1W from Seabrook for no immediate protective action with BNL results using MACCS superimposed.

6 '..

If w

E!

5

\\

E le l

\\

1-Puff

_te W

.g mg 0.5 hr i

duration

\\

W i

f 5

-2 I8

-2 g

g g

u Wd.

k%

3 n

w MACCS R'3 a

MACCS 200 rem r,,3 I

l 3-Puff W

200 rem

~

eee.eeREM

~

n a

+58.ee REM i

f::: ils a.

4 le

/

4 to-1

' ' 'l e s

lgi

,,)

1 DISTAffCE 01ILES) l Figure 6.6 Dose versus distance curve for release category S2W from Seabrook for no immediate protective action with BNL results using MACCS superimposed.

P o

6-25 I

l 10, y

.. i i g(

e e

i e i 3

o 7

O

~

-m > ~ s s

N "x

E

2 E 18..

\\

. t s"I w

~

E

\\

5 E

{

o

-2 le..

_tg 2 g

S

{

d 5

l Ei

-3

-Puff 18 3

g fi 1

18 0

MACCS

( 1-Puff 2

x 200. 80 REf1 200 rem a

se.e6 REn 0.5 hr 8

.+5.00 REN 3 Puff k duration oc 1.06 REf1 5

-4 10-1

't es

.L

{-

le 4

16

'181 te2 DISTAffCE (t1!LES)

Figure 6.7 Dose versus distance curve for release category S6W from Seabrook for no immediate protective action with BNL results using liACCS superimposed.

1

6-26 SST1, Unifor.m, population 100/km 2

o 10 5 ;.. ;;..

= *-.w

.r 1

Al t

10" r

3

.a, g

u n.

MEAN 10

eMACCS 750

~

aCRAC2 570 10

1 10 100 1000 10000 X Early Deaths (no evac)

Figure 6.8 Comparison of MACCS to CRAC2 codes (from preliminary benchmark calculations performed by Sandia Labs).

l I

e a

. 27 1

i i

i iiiiit i

i iiiisig Uz

[

sa og ou o r m.,a

$5 om IE 03

"~

NUREG-0396 7

i:z

/

a$5 gs-

"d oz g

b5

\\ -- Original 4

2 9 PLG-0465 results ow$

Kh g\\ \\

E g 0,01 gg

\\

$8 g

g,'

Shutdown events assuming early

<b 5<

lt

\\

release, PSNH letter (SBN-1225)

Ea li

\\

CG

~

l1

\\

- Shutdown events assuming 48 hour5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> 5

t release, PSNH letter (NYN-87-002)

\\

lg 200 REM g

g 0.001

'I 1

10 100 1,000 DISTANCE (MILES)

Figure 6.9 Comparison of 200 rem-dose versus distance curves for conservative assumption of no credit for operator recovery of open equi culations performed by PSNH.) pment hatch. (Cal-l N

i i

1 e

\\ 28-U

.t 1

• i6 il i

i iiiisig i

1 i....

,, g wa bo f

o-38 J,.

s, g4

.c

~L sw as

\\

o us

[g 0.1 o u 83

- /, NUREG-03'96 z<

'l

=>

a 3

gG

= ut

u. u Oz i

/

E52

\\

PLG-0465

\\/

wo 3o ow

\\

E Q 0.01

<05

. _- - _[, N r

u. o aw N

g4 l

\\ i.

N

-tw l

o g

og s

l g

- z-88 Impact of pessimistic assumptions I

b'regarding SGTR (optimistic assumps l-l.

tion off graph) l o.001

I I

'I i

';.1 ' ' '

1 10 100 1,000 OISTANCE (MILES) s Figure 6.10 Comparison of BNL sensitivity studies with PLG-

~(

0465 and NUREG-0396, (200-renplotswithno.

immediate protective actions.') -

6 r

4

{

wm e

/

s J

r.

A

~.

t

/-

a y; Y'.

6-29 lS ;*

- i' cj.

,i '

I i

A 3

' ' ' i*8I i

i e i e isig

._ i i,, e,_

Q-

,,, s mz-mm oc og 4

ku

~

c4os G2 dw,

._e as O,a i: 3, 0.1 --3 NUREG-0396 ou

/

o r.

ww u2,'

s, M o 1

wg u.. w-O 'f a

_ \\ <

Original 4

~

.n z '2

'PLG-0465 results x

wa 3

h u*o 0.01

\\

wu

\\

'r aw.

$a C$

-,Nl Effects of conservative

_~

f interpretation by PSNH of

['

$9 h

NUREG/CR-4220 data,cPSNH h

zi2

\\

letter.

om

\\

uo g

1

,l i

l t

s i

I L

0.001'

,'I' t

t I I t i t il i

e i e iisi

)..

I 10 100 1,000 i.

'i D!3TANCE (MILES)

)

Figure 6.11 Comparison of 200 rem-dose versus distance curves for conservative interpretation by PNSH of NUREG/CR-4220 data.- (Calculations performed by PSNH.)

4 s

4

\\

a

y a

Table 6.1 Summary of Release Categories Repmsenting Hypothetical Accidents (from the RSS) f

..........n,u,,,o...,,,.

,.a..

..~...,i.

...~i

~... ~

.~

=.

u,,,,

=i

..ici"'

a u

,,.m

.. ic i

.. ica

..w.

' i, -

.. u- >

.. w- >

..i.-*

i.,

.. u-s

.. wi

.. u-o

..n

.. w- >

.. wi.. ic.

~..

.. wi

.. wi

..n

..we

.. wi i. we

.. w.

3. ic*

~..

..u4 n..

u..

,. wi

..i.-.

.. u-.

.. wi

.. wi

...-s i. i.4

..ws u..

..we

.. w.

.. wi

..ws

....-s

i. 4

.. l.4.. u->

.. wi

.m

..we

... w.

.. w.

s. u-*

i. w.

i. w.

s. u-.

'.. w.

i.we

..we i.we i. wa i=\\

s. u-*

.. w.

u

.. wi

..=

..a

.. w.

u

.. ics

-.. a

..n

.. u-'

.. wi u

a

.. wi

.. =

..a

..b

..u*

i. w.

s..

u 3. u-.

.. u-.

..w-

.. ic'

.. w.

.. w*

i.u*

s

..i.*

3..

su

.. u-a 3. i.-*

.. i.-u

.. ic'

.. wa.. wu e'

.a.,.

-.-i- -.

.u

.a-

., s..

i....

w.-i

i...

... a a

i..

ua.u u.iu.

...i

......a m.

o u :-i. v. u. u.

u n.,.u

.e l

w 8

.-u c.....

a c o

p.s..==

, n.

en

.e,

,.,r f

s S...s.dt e a s.. a.i 8.s f S...* 3 p.s e e y..a.i42 i&.6 4. pg.h.b.

7.I.. i.* I p.t t s F-y e

3.. i.6 edhe ff.

4.h.

psee.hii y

$.C. i.4 t/he.

s.

u... a....

s.

i. e-..,. u 4 a..u,an.e.,.= m

=a a

a.,

t

.. pes

h. s.

. b..il.h.

.s

.l e

i e

t, l

(

l l

=

-