Text

Forwards,For Review & Comments,Draft Rept, Assessment of Pressurized Water Reactor Primary System Leaks, NUREG/CR-6582,to Meet Planned Issuance Schedule.Comments Requested to Be Submitted by 980130
	ML20203K560
Person / Time
Issue date:	12/18/1997
From:	Rossi C; NRC OFFICE FOR ANALYSIS & EVALUATION OF OPERATIONAL DATA (AEOD)
To:	Hodges M, Roe J, Shao L; NRC (Affiliation Not Assigned), NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES)
References
	RTR-NUREG-CR-6582 NUDOCS 9712230004
	Download: ML20203K560 (273)
	v • d • e

%

e.

s@ **%f

,y \ UNITED STATES E

j NUCLEAR REGULATORY COMMISSION WASHINGTON, D.C. tamaa "1

\.....[2 December Ul, 1997 MEMORANDUM TO: Jack W. Roe, Acting Director, DRPM:NRR Lawrence C. Shao, Director, DET:RES M. Wayne Hodges, Director, DST:RES Joseph A. Murphy, Acting Director, DRA:RES James T. Wiggins, Director, DRS:RGN-1 Johns P, Jaudon, Director, DRS:RGN-Il John A. Grobe, Acting Director, DRS:RGN-Ill Arthur T. Howell, Director, DRS:RGN IV Patrick W. Baranowsky, Chief, RRAB:SPD:AEOD .

FROM: Charles E. Rossi, director {l Safety Programs Division Office for,,aalysis and Evaluation of Operational Data

SUBJECT:

ASSESSMENT OF PRESSURIZED-WATER REACTOR PRIMARY SYSTEM LEAKS Attached for your review and comnnnt is a copy of the draft report," Assessment of Pressurized Water Reactor Primary System Leaks," NUREG/CR-6582. The study objective was to provide a comprehensive asscssment of U.S. experience related to pressurized-water recctor (PWR) primary system leaks and their rates; how aging affects leakage rates and trends; the safety )

significance of such leakages; and assessment of current leak detection methods. We believe this report contains substantial and significant operational based Information which will be useful to the NRC in developing and evaluating the risk-informed inservice inspection program, ir; allocating resources to address plant specific and generic problems, and will benefit the NRC staff during planned inspections.

The assessment was mainly based on a review of licensee event reports related to leak events from 1985 to September 1996, visits to PWR plants, and reviews of related licensee failure analyses and technical literature. The assessment covers four primary areas: (1) trends and dominant causes of leaks, (2) comprehensive analysis of different types of significant leak s

events such as those caused by thermal fatigue, (3) safety significance of leak events, and (4)

' effectiveness of leak detection systems.

V Of the 240 leak events identified,41 events are considered risk significant, and five of these events were analyzed by the AEOD Accident Sequence Precursor Procram. The conditional g core damage probability for these five events ranged from 1.3x104 to 3.3x104 Seven of the 41 L events contributed directly to the small-break loss-of-coolant accident laitiating event frequency. g Overall, the initiator frequency and location based on the leak databasa develop are accurately reflected in the typical full-scale probabilistic risk assessment. ,yji .

,_ s o

.L.,.ll..l..ll.i c mo\

CONTACT: Chuck Hsu (CCH1), AEOD:SPD:RAB (301) 415-6356 4'

9712230004 971218 7E/

65 2 PDR NI E fCHE M COPY g4

^~

.l.W. Roe, et al. 2 The assessment identified four previously (prior to 1985) unidentified degradation mechanisms:

(1) turbulent penetration and thermal cycling causing thermal fatigue damage in base metal and welds of PWR branch lines, (2) cavitation-induced vibratory fatigues of small-diameter piping, (3) primary water stress corrosion cracking of primary pressure boundary penetrations made of Alloy 600, and (4) transgranular stress corrosion cracking of a spare control element drive mechanism housing. Most of the locations associated with these four degradation mechanisms were not previously identit,ad as leak locations.

Thermal fatigue cracking of piping is significant: (1) therma! fatigue cracks have been found in base metal locations that are generally not inspected, and corresponding failure mechanisms are not well understood, (2) detection and sizing of thermal fatigue cracks in small diameter branch lines is difficult, inspection is not required, and the leak-oefore-break concept is not applicable, (3) timely htection of the cracks may become difficult if crack growth rates are high.

The study identified a statistiolly significant decreasing trend in the overall frequency of leak events since 1985. However, the frequency data for both reportable events of pressure boundary leaks end leaks through bolted connections do not show a statistically significant trend, while that of the reportable leak events related to valve packing degradation reveal decreasing trends.

To meet our planned issuance schedule, we would appreciate receiving your comments by January 30,1998. This will allow tirne to resolve comments, further edit the revised report, and publish the final NUREG/CR in March 1998.

Attt.chment: As stated cc w/o att.:

F. Gillespie, NRR R. Zimmerman, NRR B. Sheron, NRR B. Boger, NRR G. Holahan, NRR G. Lainas, NRR L. Spessard, NRR E, Adensam, NRR Distnbution w/att.: See attached list H:\RAB\HSU\ COMMENT 5. INT yo r.cy. . copy or us. oocum.ni. inee.i. m in. nox: c copy minout .tiact=neg copy min en.cnmenvee OFFICE RAB c RAB (f C:RAEV/ M % D:SPD 6 NAME CHsu:mmk , .

GLanik// V JRose'nthat CRossi C#J '

DATE 12/,1/97 12h/97 12///97 12/997 OFFICIAL RECORD iCOPY

(-,

Memorandum issued L12/18/97

SUBJECT:

ASSESSMENT 0F PRESSURIZED-WATER REACTOR PRIMARY SYSTEM LEAKS-Distribution w/att, -

Tublic::*--

File Center RAB R/F -

SPD R/F TTMartin -

FCongel KRaglin DHickman SRichards, NRR

.: RWessman, NRR MMayfield, RES REmrit, RES AThadani, DEDE JMitchell, OEDO TOMartin, OEDO

%~

NUREG/CR-6582 INEEUEXT-97-01068 1

ASSESSMENT OF PRESSURIZED WATER REACTOR PRIMARY SYSTEM LEAKS

'3 RA E V.N. SHAH G.M. GRANT M. B. SATTISON R. S. HARTLEY A. G. WARE C. L. ATV!OOD 1

4 .

Idaho Nationa! Engineering and Environmental Laboratory December 11,1997

ABSTRACT nis report presents our assessment of United States (U.S.) experience related to pressurized water reactor

- (PWR) primary system leaks and their rates, how aging affects leakage rates and trends, and the safety significance of such leakages. The report also includes our assessment of current leak detection methods.

Five spaific actions were taken to perform the assessment: (a) review of licensee event reports (LERs) i related to leak events, (b) development of a database to identify trends, distributions, and causes of leak events, (c) visits to PWR plants, (d) review of related technical literature including US Nuclear Regulatory

Commission (NRC) communications and reports prepared and/or submitted by licensees, and (e) detailed analysis of selected leak events, ne scope of the study was to review the LERs submitted during the period 1985 through the third quarter of 1996, which represents about 638 operatin,, years for U.S. PWRs.

The review included only those leak events that occurred during hot shutdown, hot standby, startup, and

( power operation. The review did not include primary system leaks through steam generator tubes.

! Our assessment of field experience related to PWR primary coolant leaks is that the USNRC licensees have

, apparently taken effective actions to reduce the number of reportable leak events. Out of 199 reportable leak events occurred during the study period, about 60% of the events (121 events) occurred during the first 4 years and the remaining 40% occurred during the last 8 years. One main reason for this decreasing trend is elimination of reportable leakage from valve packing. No packing leakage has been reported since

! 1991. Ilowever, the frequency of pressure boundary leaks did not rhow any statistic 4y significant trend during the study period.

Our assessment has identified four previously (prior to 1985) unidentified degradation mechanisms that have caused throughwall cracking of PWR pressure boundary: (1) turbulent penetration and resulting thermal cycling causing thermal fatigue damage of base metal and welds, (2) cavitation-induced vibratory fatigue failures of small-diameter pip.ng, (3) primary water stress corrosion cracking of primary pressure boundary penetrations made of Alloy 600, and (4) transgranular stress corrosion cracking of a spare control element drive mechanism housing, Most of the locations associated with these four degradation mechanisms were not previously idmtified as leak locations.

Our assessment of the risk significance of the leak events identified several events that can be regarded as precursors to a loss-of-coolant accident (LOCA). The risk impacts observed in the primary system leak database developed in this study are accurately reflected in the typical full-scope probabilistic ris<k assessment. Our review of the leak events does not indicate any dramatic difference between the standard

, perception about the risk of a small-break LOCA and the outcome of this study.

Current leakage detection systems, according to our assessment, are effective in detecting 1 gpm leakage inside the containment within an hour. A use of several leak detection systems is required for reliable characterization of leakage. However, these systems are not effective in detecting a very small leakage or in determining a letk source location inside the containment. Visual inspection is generally relied upon for detecting a very small leakage and determining the leak locations, but it requires one or more containment entries by plant personnel, which makes it a slow process. Advanced leak detection systems (Nitrogen-13 monitor, acoustic emission monitor, and local humidity monitor) can detect a very small leak and quickly determine leak location while the plant is operating and are being used IIi teveral non-U.S.

PWRs.

Job Code E8238 - Specialized Technical Assistance iii DRAFT - NUREG/CR-6582

)

DW - NUREG/Ch-6582 iv

/

k-

f CONTENTS ABSTRACT .................. ......................................iii LIST OF FIGURES ....................................................I LIST OF TABLES ....................................................xv EXECUTIVE

SUMMARY

.............................................. xvii ACRONYMS ......................................................xxiii DEFINITIONS AND TERMINOLOGY ......................................xxv 4

= ACKNOWLEDGMENTS ...... . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . x xvii I

.1. ' INTR ODUC flON . . . . . . . . . . . . . . . . . . . ... ........................1-1

2. TRENDS AND DOMINANT CAUSES OF PWR PRIMARY SYSTEM LEAKS . . . . . . . . 2-1 2.1 ' Distribution of PWR Primary System Leak E s.r.4 . . . . . . . . . . . . . . . . . . . . . . . 2-1 4

2.2 Distribution of PWR Primary System Leak Rates .......................2-2 2.3 Distribution of PWR Primary System Leak Locations . . . . . . . . . . . . . . . . . . . . . 2-3 2.4 Dominant Causes of PWR Primary System leaks ..... .................2-4 2.5 Consequences of the Leak Events . . , . . . . . . , , . . . . . . . . . . . . . . . . . . . . . . . 2-4

2.6 Actions for Leak Mitigation ....... .,...................... .. . 24

3. ANALYSIS OF LEAK EVENTS ASSOCIATED WITH THERMAL FATIGUE CRACKING OF PWR BRANCH LINES . . . . .............................3-1 3.1 Reactor Coolant Piping Design Requirements . , . . . . . . . . . . . . . . . . . , . . . . . . 3-1 3.1.1 Federal Regulations ....................................3-1 i . 3.1.2 ASME Code Section 111 Requirements .........................3-2 3.1.3 Other Sections of ASME Code and Regulatory Guides . . . . . . . . . . . . . . . 3-2 3.1.4 Design Basis (Expected) Transients ................. . . . . . . . . 3-3 3.1.5 Fatigue Analysis of Piping by ASME Code Method . . . . . ..... . . . . . . 3

.33.6 Fatigue Analyses of Branch Lines . . ......... ............... 3-6 3.1.7 Preservice Examinations - . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7 v DRAFT - NUREG/CR-6582 1

I 3.2 Phenomena Causing Thermal Fati Coolant Branch lines . . . . . . .....

. .gue Cracking in the PWR Reactor

.............. ........ ... . 3-8 3.2.1 Turbulent Penetration and nermal C 3.2.2 . ycling . . . . . . . . . . . ........

Turbulent Mixing . . . . . . . . . . ......,........ .. 3-8 3.2.3 Thermal Shock ......... ........ . . . 3-9

...................... 3- 10 3.2.4 - Thermal Stratification .. ........... .......

3.2.5 Thermal Striping ......... .. 3-10

................ .....................311 3.2.6 Flow-Induced Vibrations (Contributory Phenomenon)

........... .. 3-12 3.3 Trends in Thermal Fatigue Cracking . . . . . . . . . . . . . . .

3.3.1

..... ........ 3-12 World Wide Data for Thermal Fatigue Cracking 3.3.2 .... .... ... .. . 3-12 Thermal Fatigue Cracking Data for PWR Reactor Coolant Piping ...... 3-13 ,

3.4 Analysis of Unisolable Pressure Boundary Leak from Oconee 2 Makeup /High Pressure injection Line Weld, April 22,1997 . ...

3.4.1

............ .. ... 3-14 .

B&W Makeup /High Pressure injection Line Description ...... 3-14 3.4.2 Makeup /High Pressure Injection Nozzle Design . . . .. ......

3.4,3 .... 3-14 Earlier Cracking and Leak Events at B&W MU/HPI Lines . . . ....... 3-15 3.4.4 1997 Oconee 2 Event Description ...............

3.4.5

.. ....... 3-17 Root Cause Analysis of 199"i Oconee 2 Event 3.4.6

.. .. . . ... 3-18 Corrective Actions for 1997 Oconee 2 Event ...... . .. . . . . 3-21 3.4.7 Safety Consequences . . . . . . . . . .. ...,.... ........

... 3-22 3.5 Inservice Inspection of Branch Lines .

...... .......... ... ... 3-23 3.5.1 ASME Section XI Inspection Requirements . . . . . . . .

......,..... 3-23 3.5.2 Inspection of Austenitic Stainless Steet .

...... ...... . . 3-24 3.5.3 Past Experience / Augmented Examinations . . . . .

.... .. . ... . 3-24 3.6 USNRC Bulletin 88-08 for Unisolable Sections of Piping Connected to the Reactor Coolant System .. ... .. .. . . .... . ..., . 3-27 3.6.1 Cracking in Safety injection Line at Farley 2 and Tihange 1

. .... .. 3-27 3.6.2 Cracking in Residual Heat Removal Line at a Foreign Plant

. .... . 3-27 3.6.3 USNRC Bulletin 88-08 .. . ...

...... . .. . . .. 3-28 !

3.6.4 Responses to USNRC Bulletin 88-08 ..

. . ... .... . 3-29 3.7 Fatigue Monitoring of Branch Lines . . . i

...... ... .. . .... . 3-30 l 3.7.1 Westinghouse-designed PWRs .. .

. .. ..... ... . . . 3-30 3.7.2 Babcock & Wilcox-designed PWRs .. . . ........ ....... . 3-31 3.7.3 Combustion Engineering-designed PWRs ,.

.. .... .. .... 3-31 l 4

ANALYSIS OF LEAK EVENTS ASSOCIATED WITH VIBRATORY FATIGUE ,. . 4-1 4.1 Socket Weld Design . . .... .. .. .. .... .. .... . . .. 4-1 4.2 ASME Code Requirements for Vibration Tests of Piping Systems . .. , . 4-2 4.3 Excitation Mechanisms . . . . . . . . . . . . , . 4-2 DRAFT - NUREG/CR-6582 vi

4.4 Vibration Fatigue Damage Mechanism . . ..... ......................4-3 4.5 Trends in the Vibratory Fatigue Failures ...... ......................4-4 4.5.1 World wide Data for Vibratory Fatigue Cracking ......... . . . . . . . . 4 -4 4.5.2 U.S. Data for Vibratory Fatigue Cracking .......... .. . . . . . 4-5 4.6 Analysis of Positive Displacement Pump-Induced Vibrations Fatigue Cracking of Charging System Piping at Diablo Canyon ! . . . . . . . . . . . . . . . . . 4-6 4.7 Analysis of Cavitation Induced Vibrations of Letdown System Piping at McGuire 2 .... .... ..... ... ... .............. .. 4-8 4.8 ASME Code Section XI Inservice Inspection Requirements .. ....... . .. 4-11 4.9 USNRC Requirements / Communications . . .... ..... . ..... ..... 4-11 4.10 Mitigation and Monitoring of Vibratory Fangue Cracking . . . . . ... . . . 4-11 4.11 Industry Efforts to Manage Vibratory Fatigue . . . . ... ....... 4-13

5. ANALYSIS OF LEAK EVENTS ASSOCIATED WITH REACTOR COOLANT PUMP SEAL FAILURES . .... . .... .... . .. .......... .... 5-1 5.1 Reactor Coolant Pump Seal Design .... .... . ............ . . .. 5-2 5.2 Reactor Coolant Pump Seal Failure Modes . . .. ..... .. .. 5-3 5.3 Reactor Coolant Pump Seal Design improvements . ... . . ... , ,.. 5-4 5.4 Trends in Leak Events Caused by the Pu4np Seal Degradation . . . ..... . . 5-$

5.5 Analysis of the 1995 St. Lucie 1 Reacter Coolant Pump Seal Leak ., . . 5-5 5.5.1 Event Description . . .. . . . . . . 5-6 5.5.2 Root Cause Analysis .,. . . . . . 5-7 5.5.3 Corrective ?.ctions . . . ... 5-7 5.5.4 Safety Consequences . .. . . . . . . . 5-8 5.6 USNRC Generic Issue 23 . . . ... ... . .., ,.. .. . 5-8

6. ANALYSIS OF LEAK EVENTS ASSOCIATED WITH STRESS CORROSION CRACKING .... . . . . ... ...... ... , . ... 6-1 6.1 Transgranular Stress Corrosion Cracking Mechanism .. .. .. . 6-1 6.2 Trend in Leak Events Associated with Stress Corrosion Cracking F911ures . .... 6-2 vii DRAFT - NUREG/CR-6582

e 6.3 Analysis of 1990 Ft. Calhoun Leak Event Caused by Transgranular Stress Corrosion Cracking .............

6.3.1 Event Description .... ...............................6-2 4 . . ...... .... . ................. . 6 6.3.2 Root Cause Analysis . , . . . .....

..........................6-3 6.3.3 Corrective Action .......... .................... .......

6.3.4 Safety Significance . 6-5

.......................6-5

7. ANALYSIS OF LEAK EVENTS CAUSED BY VALVE STEM PACKING DEGRADATION . . . . . . . . .. . . . . ... . ... . ........ .. . . . . . . . . 7- 1

7.1 Valve stem Packing Design . . . . . . . . ..... . ........... . . . . . . . . . . 7-1 7.2 Valve Stem Packing Improvemen's . . . . . ... '

........ ..,... ........ 7-2 7.3 Tcends in Leak Events Catsed by Valve Packing Degradation . . . . ... . . . . . . 7-4 7.4 Analysis of 1991 North Anna 2 Residual Heat Removal Isolation Valve Packing Leak . . . . . . . . . ........,

.. ..... . 7-5 7.5 Industry Programs to Manage Packing Leaks ... ........ . . . . . . . . . . . . 7-6

8. ANALYSIS OF LEAK EVENTS CAUSED BY COMPRESSION FITTING FAILURES

......... . ......... ........ ... . ... .... .... . . 8-1 8.1 Compression Fitting Design . . . . . . . ......... .......

I

............8-1 8.2 Trend in Leak Events Caused by Compression Fitting Failures . .......... ,. . 8-2 8.3 Analysis of 1991 Oconee 3 Leak Event Caused by Compression Fitting Failure ..... . 8-3

9. SAFETY SIGNIFICANCE OF RCS LEAK EVENTS . , .

.. ..... 9-1 9.1 RCS Leaks in the Accident Sequence Precursor Program ....... ., ,... 9-1 9.2 Risk Review of RCS Leak LER Database ..,.

.. . . .. . .. . . 9-2 9.2.1 Risk Impact Categories . . . . . .. . . . . ... ...... ... . . 9-2

9.3 Conclusions ... . .. . . . . . ...... ......... . . . . . . . . . . 9-5

10. EFFECTIVENESS OF INDUSTRY LEAK DETECTION METHODS . . .. .. . 10-1 10.1 Requirements for Leakage Detection Systems .. ... . .. . . . ... 10-2 l 10.2 Distributions of Methods Used for Leak Detection and Location Determination . . .

. 10-4 DRAFT NUREG/CR-6582 viii f

_ - 9

10.3 . Effectiveness of Current Leakage Detection Systems . . . . . . . . . . . . . . . . . . . . . 10 -10.4 - Capabilities of Advanced Leakage Detection Systems L. . . . . , . . . . . . . . . . . . . . z 10 11. FINDINGS AND CONCLUSIONS ...................................-.11 1 12. REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1 APPENDIX A - Search Criteria APPENDIX B - Reactor Coo! ant Leak Data Collectior. Form APPENDIX C - Statistical Analysis of Trends of PWR Primary System Leaks

'ix DRAFT - NUREG/CR-6582

LIST OF FIGURES 2 1. Distribution of both reponable and nonreportable reactor coolant leak events during the 1985 1996 period.

2 2. Distribution of reponable leak events and their frequencies during the 1985-1996 period. The frequency for 1996 is estimated based on the operating years for 1995, accounting for data only through the third quarter of 1996.

2-3. IIistogram of the number of operating years for each calendar year of the study period.

2-4. Statistical analysis of a trend in the frequency of reponable leak events. Point estira.tes,90%

confidence bands, and 90% confidence intervals are shown.

2-5. Distribution of reponable reactor coolant leak events based on the ability to physically is.am the leak.

2-6. Reponable reactor coolant leak events inside containment panitioned by the initial leak rate.

2-7. Reportable leak events inside containment partitioned by the maximum leak rate.

2-8. Reportable leak events outside containment partitioned by the maximum leak rate.

2-9. This figure has been removed.

2-10. Distribution of pressure boundary leak events and their frequency during the INS-19% period.

The frequency for 1996 is estimated based on the operating years for 1995, accounting for data only through the third quarter of 19%.

2-11. Reponable reactor coolant leas events panitioned by the system from which the leak ocurred.

2-12. Reponable reactor coolant leak events partitioned by the component from which the leak occurred.

2-13. Reponable reactor coolant leak events panitioned by the location in the component. .

2-14. Distribution of reportable leaks through the welds in pipe and instrument lines versus their size.

2-15. Distribution of reportable leaks through the bolted connections and their frequency during the 1985-1996 period. The frequency for 1996 is estimated based on the operating years for 1995, accounting for data only through the third quaner of 1996.

2-16. Reportable reactor coolant leaks panitioned by degrauation mechanism.

DRAFT - NUREG/CR-6582 x

2 17. Distribution of the three most-s'3nificant: degradation mechanisms in terms of numbers of

^

reportable leak events during 1985-1996i Distribution includes data only through the third quarter of 1996.

2-18. Reportable reactor coolant leak events partitioned by the contributing factor for the leak.

2-19. Reportable reactor coolant leak events partitioned by the leak consequence.

2-20. Distribution of reportable leak events by the short-term corrective actions. The other category includes design changes, evaluation, inspection and testing, and proper installation.

_2 21. Distribution of reportable leak events by the long-term corrective actions.

3-1. Turbulence interaction regions (EPRI 1993b).

3-2.' A change in the penetration depth of turbulence from power variation., can cause thermal stratification and cycling in a branch line (EPRI 1993c).

3-2.1 Natural convection driven stratification in a safety injection line - Safety injection Line (EPRI 1993f).

3-3. Temperature distribution in stratified fluid.

3-4. Typical old and new designs of the makeup /high-pressure injection nozzles and thermal sleeves in the Babcock & Wilcox plants (Babcock & Wilcox 1983).

3-5. Thermal Fatigue damage in LWRs - Based on SKI's LOCA affected piping database (SLAP)

Versicn 3.0 (Nyman et al.1996),

3-6. Pipe leaks and ruptures in LWR piping caused by thermal fatigue (Nyman et al.1996).

. 3-7. Oconee 2 makeup /high pressure injection system.

4 3-8. Typical Layout of Oconee 1 MU/HPI line.

3-9. Original makeup nozzle and thermal sleeve design in a demonstration PWR plant (Flynn et al.

1975).

i~

3-10. Replacement makeup nozzle and thermal sleeve design in a demonstration PWR plant (Flynn et

- al.1975). (modified) 3-11. Observed crack on th: outside surface of the MU/HPI line weld before removing the damaged -

piping components (Redmond 1997).

3-12. - Portion of the MU/HPIline and warming line adjacent to safe end removed for the failure analysis (Redmond 1997).

xi DRAFT - NUREG/CR-6582

4 3 13. Angular distribution of depth of the circumferential crar.k ta the safe end-to MU/HPI line weld.

3-14.

Downstream end of the worn MU/IIPI nozzle thermal sleeve. The view is from the downstream side of the reactor coolant ficw direction (Redmond 1997).

3-15. Typical multidirectional cracking observed on the inside surface of the MU/HPI piping near the warming line penet ation (Redmond 1997).

3-16. Penetrant testing results revealing multidirectional cracking on the inside surface of the MU/HPI safe end (Redmond 1997).

3-17. Location of the crack in the weld of the Farley 2 safety injection piping. .

3-18. Location of the cracks in the base metal and welds in the Tihange I safety injection piping (Su 1990).

3-19. Schematz dugram of safety injection line in a three-loop Westinghouse-type plant (Su 1990).

3 20. Schenute of the residual heat temoval suction line at Genkai I (USNRC 1989).

3-21. Instrumented portion of safety injection lines on 4-loop Westinghouse plant.

3-22. Instmmented ponion (locations I through 3) of HPI line near the first check valve on B&W plant.

3 23. Instrumented portion of shutdown cooling line on older vintage Combustion Engineering plant.

3-24. Instrumented portion of safety injection line on Combustion Engineering plant.

4-1. Socket and fillet weld details and dimensioris.

(a) Minimum welding dimensions for socket welding fittings (b) Equalleg fillet weld 4-2. Examples of so;ket welding fittings.

4-3. Socket weld failme initiating at toe of fillet weld.

4-1. Socket weld failure initiating at root of fillet weld.

4-5. World-wide data for vibratory fatigue failures of small-diameter piping (s 25-mm diameter) in LWRs over the period 1972 to 1995 - based on SLAP (SKI's LOCA Affected Piping Database)

Version 3.0 (Rev.1).

4 6. Hazard function plot representing world-wide data for pipe leaks and ruptures in reactor coolant system small diameter piping (s 25-mm diameter) caused by vibratory fatigue cracking.

4-7.

Distribution of reponable vibratory fatigue-related leak events and their frequencies by calendar year.

DRAFF - NUREG/CR-6582 xii

4 8. Distribution of reportable packing degradation-related leak events and their frequencies by plant age (years in service).

4-9. Cumulative distribution of reportable leak events caused by vibratory fatigue by plant age (years in service).

4-10. Statistical analysis of a trend in the reportable vibratory fatigue related leak event fre luencies estimated by calendar year. Point estimates,90% confidence bands, and 90% confidence intervals are shown. ,

4-11. Statistical analysis of a trend in the reportable vibratory fatigue related leak event frequencies estimated by plant age (years in service). Point estimates,90% confidence bands, and 90%

confidence intervals are shown.

4-12. Typical cantilevered small-bore piping connected to a large-bore piping.

4-13. Typical Westinghouse letdown system for a 4-loop plant.

4-14. Inservice inspection region for socket welded joint.

5-1. Byron Jackson reactor coolant pump seal with an injectionless seal cooling system (USNRC Technical Training Center) 5-2. Combustion Engineering supplied Byrondackson seal and seal bleed-off system schematic (Ruger and 1.ukas 1989).

5-3. Schematic of a Byrondackson reactor coolant pump mechanical seal stage (Bell and O'Reilly 1992) 6-1 Canopy-scal weld design on Westinghouse PWRs head penetrations.

6-2. Schematic of CEDM housing showing location of cracking on the two spare housings.

7.1. Typical asbestos-based valve stem packing design originally used in nuclear power plants. Typical axial and radial stress distributions across the packing rings are shown (Ruggieri and Kelly 1988).

7-2. Typical asbestos-based valve stem packing design with a leak-off system (Ruggieri and Kelly 1988, Hart 1996).

7-3. Standard packing arrangement recommended for square packing. Gland follower is not showm (Ruggieri and Kelly 1988).

7-4. Standard packing arrangement recommended for wedge-shaped packing (Ruggieri and Kelly 1988).

7-5. Live-loading of valve stem packing using disc spring:(Ruggieri and Kelly 1988).

7-6. Distribution of reportable packing degradation-related leak events and their frequencies by calendar year. !

xiii DRAFT - NUREG/CR-6582 I

7 7. Distribution of reportable packing degradation-related leak events and their frequencies by plant age (years in service).

7-8. Statistical analysis of a trend in the reportable packing degradation related leak event frequencies estimated by :alendar year. Point estimates,90% confidence bands, and 90% co- 'ence intervals are shown.

7-9. Statistical analysis of a trend in the reponable packing degradation related leak event frequencies estimated by plant age (years in service).

8-1 Schematic of a Swagelok fitting. Crimping action of ferrules on heavy- and thin-walled tubing is shown. ,

8-2 Schematic of a Parker-Hannifin compression fitting showing nut that is torqued onto adaptor to compress tube into adaptor shoulder. A gap less than the nominal value may indicate that fite,; ,

is properly seated.

8-3 Schematic showing how six compression fittings in series were used to reduce the instrumentation line on the Oconee 3 hot leg from 19-mm (3/4-in). to 9.5-mm (3/8-in.) tubing.

8-4 Trend in U.S. nuclear power plant compression fitting leaks shows relatively stable rate of one failure per year (19851996).

10-1. Distribution of the methods used to detect the reponable reactor coolant leaks. The total exceeds the number of reportable leaks (199) because scme of the events were detected by more than one method.

10-2. Distribution of the methods used to detect the reportable leaks inside the containment. The raw data counts are given with the percentage for each category.

10-3. Distribution of the methods used to detect the reportable leak events outside the containment. The raw data counts are given with the percentage for each category.

10-4. Distribution of the methods used to locate the source of reactor coolant leakage that occurred inside or outside of containment. The taw data counts are given with the percentage for each category.

10 5. Distribution of reportable reactor coolant leak events by initial leak rate and detection method.

~

10-6. Time dependent particulate activity increases resulting from reactor coolant leakage (Aoki 1991).

10-7. Time dependent gaseous activity increases resulting from reactor coolant leakage (Aoki 1991).

10-8. Time dependent condensate flow rate increase resulting from : 3.8-lJmin (1-gpm) reactor coolant leakage (Aoki 1991).

DRAFT - NUREG/CR-6582 xiv

LIST OF TABLES 2-1. Reportable leak events caused by three different SCC mechanisms.

3-1. PWR reactor coolant leak events caused by thermal fatigue.

4 1. Summary of PWR cracking in pipes smaller than 4 inches.

4-2. Examples of fatigue failures in small ( < 2 in.) PWR lines connecting to the reactor coolant system (Stoller Corporation).

5-1. Number of reactor coolant pump seal configurations used at CEOG plants in 1983 and 1991 (Combustion Engineering 1991).

5-2. Reactor coolant pump seal leak events reported during 1985 to 1996 (3rd quarter).

91 leak eve'nts analyzed in the Accident Sequen e Precursor Program.

9-2 Risk impact categorization of the RCS leak LER database.

9-3. Risk impact parameter frequencies.

10-1. Reactor coolant leaks outside the containment that were detected by control room indication.

10 2. Capabilities of leakage monitoring methods inside containment.

10 3. Capabilities of leak detection systems in a typical Japanese PWR C 1. Summary of analyses for trend in calendar time.

C-2. Summary of analyses for trend with plant age.

xv DRAFT - NUREG/CR-6582 l

l

i DRAFT - NUREG/CR-6582 xvi

i EXECUTIVE

SUMMARY

The main objectives of this study are to review U.S. experience related to pressurized water reactor (PWR) primary system icaks and their rates, how aging affects leakage rates and trends, and the safety significance of such leakage. Another objective of the study is to assess current leak detection methods. A review of experience related to steam generator tube leaks is not included. Five specific actions were taken to accomplish the objective: (a) review of licensee event reports (LERs) related to leak events, (b) development of a database to identify trends, distributions, and causes of leak events, (c) visits to PWR plants, (d) review of related technical literature including U.S. Nuclear Regulatory Conunission (NRC) communications and reports prepared and/or submi':ed by licensees, and (e) detailed analysis of selected leak events.

The scope of the study was to review the LERs related to PWR primary system leaks submitted during the period 1985 through the third quarter of 1996, representing about 638 operating years for U.S. PWRs.

The reviev' included only those leak events that occurred during hot shutdown, hot standby, startup, and power operation. -

Our assessment of field experience related to PWR primary coolant leaks is that the USNRC licensees have app :rently taken effective actions to reduce the number of reportable leak events. One main reason for this decreasing trend is elimination of leakage from valve packing. However, the frequency of pressure boundary leaks did not show any statistically significant trend during the study period.

Our assessmem of risk significance of the leak events identified several events that can be regarded as precursors to a losa-of-coolant accident (LOCA). The risk inpacts of the observed operating experience are accurately reflected in the typical full scope probabilistic risk assessment (PRA). Our review of the leak events does not indicate any dramatic difference between the standard perception about the risk of a small-break loss-of-coolant accident (SBLOCA) and the outcome of this study.

4 Current leakage detection systems are effective in detecting i gpm leakage inside the containment within

.m hour, llowever, these systems are not effective in detecting a very small leakage or in determining a leak source location inside the containment. Visual inspection is generally relied upon for dete; ting a very small !cakage and determining the leak locations.

Our specific findings are divided into seven ereas: (1) trends of annual rates of priraary coolant leaks, (2) previously unidentified degradation mechanisms and failure modes that have caused primary coolant leaks, (3) locations and types of leaks not previously identified, (4) leaks that have a potential for relatively rapid growth, (5) safety significance of piping fatigue, (6) leak events that may be regarded as core damage precursors, and (7) effectiveness of current leakage detection system.

3 Trends of Annual Rates of Primary Coolant Leaks The study identified the following trends of primary coolant leaks that occurred during 1985 through the third quarter of 1996:

.= The study identified a statistically significant decreasing trend in the frequency of leak events since 1985, if the first four years of the analysis are excluded, the reocrtable leak events do not show any statistically significant trend.

xvii DRAFT - NUREC/CR-6582

l The frequency data for the reportable leak events related to vibratory fatigue reveal a statistically significant decreasing trend with age (years of operation), but not in calendar time.' This trend implies that the vibratory fatigue failures are caused by inadequacy of initial design and fabrication, and not aging damage.

The frequency data for the reportable leak events related to valve packing degradation reveal decreasing trends both in calerdar time and with age, but the trend in calendar time is statistically more significant than with age. Apparently, the decreasing trend is a result of valve stem packing improvements made in the early 1980s and incorporating some of them into operating nuclear power plants.

The frenuency data associated with two categories of reportable leak events, pressure boundary

~

leaks and leaks through bolted connections, did not show a statistically significant trend.

Previously (Prior to 1985) Unidentified Degradation Mechanisms and Failure Modes That Have Caused Primary Coolant Leaks The study ident,ified the following four previously unidentified degradation mechanisms or failure modes:

Turbulent penetntion and resulting thermal cycling have caused throughwall fatigue cracking in the pressure boundary of PWR branch lines; however, these phenomena are not completely understood. The affected branch lines are safety injection lines at Farley 2 an2 Tihange 1, and a residual heat removal line at a foreign PWR Apparently, these phenomena have a'so played a major role in the April 21,1997, leak event at Oconec 2. As a mWignive action, some licensees are planning or performing augmented inservice inspection of welds and base netal locations, susceptible to cracking caused by these phenomena.

Pump-induced pressure pulsations and cavitation have caused vibratory fatigue failures of socket vields on small diameter piping. The role of positive displacement pump-induced vibrations was previously identified, but the role of a cavitating component such as an orifice in causing vibratory fatigue failure of PWR letdown systems was not. In one instance, cavitation-induced vibratory fatigue and water hammer played a synergistic role in causing rupture of a letdown system drain line. Quatuitative understanding or the vibratory fatig.ie mechanism is being developed.

Primary water stress corrosion cracking has caused throughwall cracking of primary pressure boundary penetrations made of Alloy 600 (a nickel-base material). The penetrations include (a) instrument nozzles and heater sleeves for pressurizers, (b) instrument nozzles for main coolant piping, and (c) control rod drive mechanism nozzles at non-U.S. PWRs.2 The affected U.S. PWRs 8

The frequency data for trend in calendar time represent the number of reportable leak events per reactc operating year, whereas the data for trend in age represent the number of events per operating PWR.

Shah, V. N., et al.,19%. Assessment of hessurized Water Reactor Control Rod Drive Mechanism No:de Cracking, NUREG/CR 6245 EGG-271$, October.

DRAFT - NUREG/CR 6582 xviii

were designed by either Combustion Engineering or Babcock & Wilcox who employ Alloy 600 penetrations for pressurizers and train coolant piping.

Transgranular stress corrosion crackmg has caused throughwall cracking of a spare control element drive mechanism housing at Fort Calhoun, a Combustion Engineering-designed PWR. Trans-granular stress corrosion cracking of canopy-seal welds has been reported at Westinghous>

designed PWRs, but differs from the one at Fon Calhoun: a canopy weld provides sealing to the primary system and is not a structural joint, and the housing material is Type 304 stainless steel whereas the canopy weld material is Type 308L. There had been leaks from canopy seal welds before 1985.

_ Locations and Types Not Previously identified

Thermal fatigue has caused throughwall cracking in the base metal of PWR branch lines. Cracking i . was not expected at these sites, and therefore not included in the plant inservice inspection program. Two examples of such locations include a throughwall crack in a safety injection line elbow at Tihange 1, and a throughwall crack in the base metal of a straight portion of a safety injection 1ine at Dampierre 1.

Vibratory fatigue cracking of socket welds in the letdown system piping was caused by cavitation.

This cause was previously unidentified.

Primary water stress corrosion cracking has caused throughwall cracking of Alloy 500 penetrations. Susceptibility of these penetrations was previously unidentified.

Transgranular stress corrosion cracking has caused throughwall cracking of austenitic stainless neel components. Susceptibility of these components in a PWP. environment was previously uniden-tified.

Leaks That Have a Potential for Relatively Rapid Growth The following two sets of leaks have a potential for relatively rapid growth after detection, such that a SBLOCA could occurs:

Leaks through thermal fatigue cracks of branch lines.

If crack growth rate is high and uniform along its length, then a circumferential crack in a branch line may rapidly propagate through the intact ligament of the pipe cross section after a leak is detected. Our study revealed that such a rapid and uniform crack growth is possible.

The recent experience at Dampierre 1 indicates that a small, nondetectable fatigue crack in a safety injection line could become a throughwall crack in one fuel cycle, indicating a high crack growth rate. The recent cracking at Oconee 2 suggests the pssibility of a uniform crack growth. Tue growth of the circumferential crack was uniform (30% throughwall) over about 78% of the circumference. The faster growth in the throughwall portion of the crack was due to an applied cyclic bending moment, possibly from local stratification. In the absence of such a bending moment, the crack growth would have been uniform along the entire circumference.

xix DRAFT - NUREG/CR-6582

+ Leaks through reactor coolant pump seals.

An initially small leak caused by a failure of a lower seal stage may grow rapidly if the ]

remaining seal stages fall. Such an event could contribute directly to the initiation of a 1 SBLOCA. For example, a vibratory fatigue failure of a reactor coolant pump seal sensirg line at Arkansas Nuclear One Unit 2 caused a complete failure of the pump seal that resuhed in a j maxi num leak rate of 152 L/ min (40 gpmy.- !

Safety Significance of Piping Fatigen Thermal fatigue of PWR branch lines is becoming a safety issue for the following reasons:

+ Susceptible sites (welds and base metal) are generally not inspected. Additionally, it is difficult to detect cracks at these sites during inservice inspection when the plant is shutdown. Often a crack has been detected because of leakage and not inservice inspection. ,

+ If fatigue crack growth rate is about uniform over the entire circumference of a brench line cross section, a crack may rapidly propagate through the intact ligament of the piping after a leak is detected, leading to pipe rupture. Higher fatigue crack growth rate will lead to pipe rupture sooner after the lear is detected.

l + Tae leak-before-break concept is not applicable to small-diameter piping. The concept is also not applicable to piping subject to fatigue damage.

Leak Events That May Be Regarded as Core Damage Precursors Risk review of the database identified seven leak events that contributed to the SBLOCA initiating event frequency. During these events. the leakage was large enough or had a potential to develop to a size to require safety system response as modeled by a PRA SBLOCA event tree. These types of events also include small leaks that, if unattended, have a potential to become large enough to requ;re safety system or operator response. Two of the seven events were analyzed by the USNRC Accident Sequence Precursor (ASP) Program.

Risk review of the database identified six leak events that contributed to transient induced LOCA probability. One of these events was identified and analyzed as a precursor to a LOCA by the USNRC ,

ASP Program. The most typical transient-induced LOCAs involve a stuck-open relief valve or a reactor coolant pump seal failure. These failures c cur in the presence of an off-normal condition such as a pressure transient or loss of offsite power. .

Conclusions about the overall perception of the contribution of reactor coolant system (RCS) leaks to core ;

damage fol;ow: !

+

The industry leak experience has always been the basis for SBLOCA frequency calculations. The frequency of SBLOCA from this study (0.011) is nearly identical to the generic SBLOCA initiating i event frequency (0.01) used in many PRAs. !

DRAFT - NUREG/CR-6582 xx

i l

The frequency of leak-induced transients is so small (6.3E-3/ reactor year) in comparison to tha overall transient initiating event frequency (1 - 2/reactot year) that there is no discernable influence on the core damage frequency.

. The frequency of transient-induced relief valve LOCAs from the operational data (6.27E-3/ reactor year) is within the range of c.Jeulated values observed in typical PRAs.

a fihe frequency of reactor coolant pump seal leaks becoming small LOCAs is so small that none exist in the dstabase.

. Overall, the risk impacts observed in the primary system leak database developed in this ,tudy are

- accurately reflected in the typical full-scope PRA.

Effectiveness of Current Leakage Detection Systems Our huin finding is that use of several different leak detection systems is required for reliable char-acterization of leakage. Other findings follow:

The current leakage detection systems are effective in detecting I gpm leakage inside the containment within an hour. However, several small leaks [<0.38 IJmin (<0.1 gpm)] were detected by visual inspection.

. The current leakage detection systems are not effective in locating the leak source inside the containment. Generally, containment entry and visual inspection are required. During some leak events, it takes considerable time to determine the location of the leak sourco while the plant is operating; therefore, a leak source, possibly located on the pressure boundary, may remain unidentified for that period of time.

Advanced leakage detection systems based on Nitrogen-13 detection, acoustic emission monitoring and local humidity monitors can detect leaks of a fraction of a gpm and quickly determine leak location while the plant is operating. These systems have been installed at some foreign plants.

xxi DRAFT - NUREG/CR-6582

g ih DRAFr . NUREG/CR-6582 xxii

ACRONYMS

AASHTO American Association of State Highway and Transportation Officials AECL Atomic Energy of Canada Limited AEOD Office for Analysis and Evaluation of Operatiott:1 Data ASA Ameri:an Standards AssoPlion ASME Amencan Society of Mechanical Engineers ASP accides sequence precursor B&W Babcock A Wdcos BWR bodmg maser reactor

. - CFR Code of Federal Regulations CRDM control rod dme mechanism CEDM control elemem dnvc mechanism CEOG Combuston Engmeering Owners Gmup EDF $lectncne de France EPRI Electnc Power Research Institute FSAR Final Safety Analysis Report GDC general design criteria IUTC heated junction thermocouple H1AP Houston Lighting & Power HPI high pressure injection IGSCC intergranular stress corrosion cracking INEEL Idaho National Engineering and Environmental Laboratory INPO Institute of Nuclear Plant Operations ISI inservice inspection KSB Klein, Schanzlin & Becker LER Licensee Event Repor' LOCA loss-of-coolant accident LWR Light Weter Reactor MSIS main steam isolation signal MOV motor operated valve MU make up NPS nomirial pipe size NRC Nuclear Regulatory Commission xxiii DRAFT - NUREG/CR-6582 P

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

)

.4 i

. . . . i NSSS. : smclear steam supply system ..

PORY- power opcated relief valve

PRA --: probabilistic risk assesscent

- PSI preservice inspection PWR ~ pressurized water reactor

"PWSCC primary trater stress corrosion cracking .

[

RCP- reactor e colant pump RCS reacto coolant system -

RHR . residust heat removal -

. s RTD ; resistance temperature detector

RVLIS reactor vessel level indicating system 4

SCC stress corrosion crackMg SCF stress concentration factor SCSS Sequence Coding Search System SEM scanning electron microscopy i

!- -SER Significant Event Report

, SI . safety irdection '

SKI Swedish Nuclear Power Inspectorate SRP Standard Review Plan TAP Task Action Plan TASCS thermal stratification, cycling, and striping TGSCC transgranular stress corrosion cracking USAS United States of Atnerica Standard USNRC U.S. Nuclear Regulatory Commission '

. UTS ultimate tensile strength VCT volume control tank -

DRAFT - NUREG/CR-6582 - xxiv

o

DEFINITIONS AND TERMINOLOGY Identifled Leakage Identified leakage shall be:

a. Leakage into co.llection systems, such as pump seal or valve packing leaks, that is captured and conducted to a collection tank, or

b. Leakage into the containment atmosphere from sources that are both specifically locwed and known not to interfere with the operation of the leakage detection systems or not to be pressure boundary leakage.

Isolable leaks All leaks other than non-isolable leaks isolable w/ difficuhies All isolable leaks that could not be stopped until the plant operating mode was changed. Examples include i (1) escaping steam from a failed isolation valve prevented operators from placing the valve on its backseat to stop the leak; and (2) leak-induced airborne activity exceeded the muimum permissible dose and required a plant shutdown and cooldown before personnel could enter the area and stop the leak.

Non Isolable leak Any leak that can not be stopped by either placing the valve on its backseat for packing gland or body-to-bonnet leaks or closing a valve. Non-isolable leaks include all pressure boundary leaks, and leaks associated with reactor coolant pump seals and pressurizer safety valves.

Pressure Boundary Leakage Pressure boundary leakage is a leakage through a non-isolable fault (through wall crack) in a RCS component bcdy, pipe wall, or vessel wall. Pressure boundary leaks do not include leaks through gaskets, packing glands, valve rats, flanges, manway covers, or other indianical joints or fittings.

Reactor Operating Years Reactor operating years, ideally, is the time when the plant operating mode is other than the cold shutdown or refueling modes. This time is not known exactly. The NRC's database OUTINFO lists the starting and ending dates of all periods when the main generator is off-line, for each plant. The dates are found in the monthly operating report submitted to the NRC by each plant. During short generator off-line periods, it is unlikely that a, plant will enter the cold shutdown or refueling operating modes; therefore, the starting and ending days of such outages were treated as operational periods. The outages likewise were treated as operational if they spanned two calendar days or less. The operating time for a plant was estimated by xxv DRAFT - NUREG/CR-6582

u

-calendar tirne minus all periods when the main generator was off.line for more than two calend

- This method is the same rue' hod used to estimate operating time for the NRC's Performance Indicato Operating Conunercial Nuclear Power Reactors Repon.

Reportable Leak 1.

Any leak that exceeds the technical specifications for allowed leakage or:

2, any leak that results in exceeding the final safety analysis report (FSAR) limitr, for control room habitability, or 10CFR100 limits or; 3.

results m an unplanned engineered safety feature actuation.

Technical Spedfication Leakage Limits and Reportability Requirements

a. No Pressure Boundary 1.cakage .

b. 1 GPM Unidersified Leakage

c. 10 GPM Identified leakate Unidentified leakage Unidentified leakage is all leakage which is not identiSed Leakage t

DRAFT- NUREG/CR-6582 xxvi

ACKNOWLEDGMENTS xwii DRAFT - NUREG/CR-6582

i Assessment of Pressurized Water Reactor Primary System Leaks

1. INTRODUCTION A primary system leak in a pressurized water reactor (PWR) challenges the integrity of the reactor coolant system (RCS) pressure boundary, which is enential for supporting the defense in-depth concept. 'Ihe safety significance of a leak depends upon its location, rate, and duration. A location of a leak may be such that tbc leak, or leak repair action, disables or degrades a safety system and cortributes to an increased itkelihood of core damage as a result of reduced accident mitigation capability. A leak initiates a small break loss-of-coolant accident (LOCA) if its rate is higher than the charging pump capacity. Each inadvertent and uncontrolled reactor pressure boundary leak of the reactor coolant may threaten the ability to remove heat from the core. Some leaks, if not detected early, may become potenaal precursors to a small-break LOCA.

Several leaks have occurred throtagh the prhnary pressure boundaries of the Uni.ed States (U.S.) PWRs.

Some leaks have occurred at locations or been caused by failure mechanisms not previously identified.

For example, turbulent penetration and thermal cycling have caused fatigue cracking in a branch line resulting in a primary pressure boundary leakage. leakage of primary coolant has caused boric acid c.orrosion of ferritic steel components ccnstituting the primary pressure boundary.

The nuclear industry has taken several actions to mitigate and reduce PWR primary system leaks. Major emphasis has been on improving the design of the affected components. Two examples include an improved design for valve stem packing to reduce packing leakage and replacement of socket welds with butt welds to avoid vibratory fatigue cracking.

The main objective of this project is to review the U.S. experience related to PWR primary system leaks and their rates, how aging effects leakage rates, and the safety significance of such leakages. A review of steam generator tube leaks is not included. The project objective also includes an assessment of current leak detection inethods.

Five specific actions were taken to accomplish the objective: (a) review of Licensee Event Reports (LERs) related to leak events (b) development of a comprehensive database to identify trends, distributions, and causes of leak events, (c) visits to PWR plants, (d) review of related technical literature including U.S.

Nuclear Regulatory Commission (NRC) conununications and reports prepared and/or submitted by licensees and (e) detailed analysis of selected leak events.

The scope of the project was to review the LERs related to PWR primary system leaks submitted during the period 1985 through the third quarter of 1996. The review included only those leak events that took place during hot shutdown, hot standby, startup, and power operation.' Leak events that took place during cold shutdown and refueling were not included, leak events associated with intersystem leaks also were not included.

L The plant technical specifications require that the leak detection systems be operable during these four operational modes and not the other two modes.

1-1 DRAFT NUREG/CR-6582 0

INTRODUCTION The project efforts were carried out with the following eight questions in mind. The responses to these questions are presented at the end of the report. l 1

i

1. Are there data to support the existence of meanirgful trends or annual rates of primary coolant leaks?

2. Are there any previously unidentified failure mechanisms that have caused primary coolant leaks?

3. Are there any leaks in locations, or of types, not previously identified?

4. Which sets of leaks have a potential for relatively rapid growth after detection such that a small-break LOCA occurs? 4

5. Is fatigue of aging pipe becoming safety .ignificant?

6. Could any of the reportable leaks be regarded as a core damage precursor?

7. Does the review of leak events indicate that the risk of a small-break LOCA is greater than previously hypothesized?

8. 110w effecf.ive are the current leak detection systems?

The outline of the report is as follows. Section 2 presents trends, distributions, and dominant causes of primary system leaks. It also summarizes results for consequences of the leak events and the associated mitigation and vepair activities. Sections 3 through 8 present comprehensive analyses of six different types of significant leak events. Each section summarizes for a given type of leak event, relevant design characterisdr.s and requirements, degradation or excitation mechanisms, trends in the related leak events, detailed anaiysis of selected leak events, American Society of Mechanical Engineers (ASME) and USNRC requiremerits and communications, and industry efforts to mitigate the leak events. The analysis of a selected event includes event description, root cause analysis, corrective actions, and safety consequences.

Section 3 analyses the events associated with thermal fatigue cracking of PWR branch lines; the cracking has occurred at some unex},ected locations and the mechanisms causing : racking are not yet fully understood.

Secdon 4 analyses the events associated with vibrator) fatigue cracking. These events are analyzed ber.ause there was a large number of leak events related to vibratory fatigue failures prior to 1985, and several more are reported since then, in addition, the understanding of the excitation mechanisms causing vibratory fatigue failures is evolving.

Section 5 analyses the leak events associated with reactor coolant pump (RCP) seal failures. These events are analyzed because of their safety significance. One of these events has been analyzed by the USNRC Accident Squence Precursor (ASP) program.

DRAFT - NUREG/CR-6$82 12

INTRODUCTION Section 6 analyses the leak events associated with stress corrosion cracking. These events have been selected because these cracking mechanisms have caused leaks at previously unidentifieJ locations.

Section 7 analyses the leak events associated with valve-stem packing degradation. These events are selected because there were a large number of events relating to the packing degradation prior to 1985 and during the 1985-86 period, llowever, the number ofleak events related to packing degradation have been sigt.Jicatstly reduced since then because the industry has developed improved packing designs.

Section 8 analyses the leak events associated with compression fitting failures. These events are analyzed because in one event the leak was isolable but could not be isolated because of a hazardous emironment created by the leak and, as a result, a large amount of coolant leaked.

Section 9 cvaluates the safety significance of leak events. This section categorizes the leak events according to their risk significance. Then it calculates event frequency for each category and compares it with the corresponding frequency used in the current probabilistic risk analysis (PRA) studies.

Section 10 evaluates the effectiveness of industry leak detection systems. This section also presents the distribution of rnethods used for initial leak detection and location determination.

Section 11 sununarizes the project results by presenting answers to the questions presented in this section.

9 l3 DRAFT - NUREG/CR-6582

2. TRENDS AND DOMINANT CAUSES OF PWR PRIMARY SYSTEM LEAKS j The Idaho National Engineering and Environmental Laboratory (INEEL) staffhas reviewed the LERs related to PWR primary system leaks that occurred during the period January 19F to September 1996. The scope ;

of the project included the leak events that occurred during hot operat.sns, hot shutdown, hot standby, ;

startup, and power operations. Leaks found during refueling and cold shutdown periods were excluded, unless the leak initiated during hot operations.

Two hundred and fifteen LERs were included in the study. These LERs were identified by a search of the USNRC Sequence Coding and Search System (SCSS) database as potentially associated with the reactor coolant leak events, ne search criteria employed to locate these LERs are presented in Appendix A.

Some LERs were associated with more than one leak event, that is, a leak occurred at more than one location in the RCS. We found 240 leak events associated with the 215 LERs included in the study. Of the 240 events included in the study,199 were events (associated with 174 LERs) reported because ofleaks. Mostly, the leak rate in these events exceeded a plant technical specification limit, and were seportable leaks in accordancs with the LER Rule 10, Code of Federal Regulations (CfR) S0.73. In some of the 199 events ,the leak rate was smaller than the technical specification limit, but the event was reported because there was a potential for the leak to exceed the limit. Several leak events associated with RCP seal degradation fall into the later category.

He remaining 41 events were instances where a reactor coolant leak was mentioned in the LER narrative, but the LER was not prepared because ofleak, nese 41 events represent nonreportable leaks. For example, an LER was submined as a result oflate performance of a surveillance test associated with a containment isolation valve, and the test was required following adjustment of a packing gland leak.

We reviewed the 215 LERs and summarized the information in a database using the form presented in Appendix D. De information was divided into six categories: methods used for leak detection, leak rates, leak locations, leak causes, leak consequences, and corrective actions taken. Distributions of the methods used for leak detection and location are p esented in Sectica 10, whereas distribution and trends in the other categories are presented here.

2.1 Distribution of PWR Primary System Leak Events O

ne yearly distribution of all '140 leak events, including reportable and nonreportable events, is shown in Figure 2-1. As number ofleak events was significantly higher during the first 4 years of the study period

, than during the last 8 years. Other distributions and trends reported in this section include only reportable events, ne distribution of the 199 reportable leak es ents is shown in Figure 2 2. More than half of the leak events (121 events) occurred during the first 4 years of the study period. The frequency ofleak events is also shown in Figure 2-2, and is calculated by dividing the number of reportable leak events for a given calendar year by the number of PWR operating years for the same year (see Figure 2-3). Frequency of reportable leak events has significantly decreased since 1986.

21 DRAFT- NUREG/CR-6582

TRENDS .6ND DOMINANT CAUSES A histogram of the number of PWR operating years for each calendar year in the study period is presented in Figure 2 3. Estimates of operating years for a given calendar yeu arm determined by summing the plant i

specific operating time for eaa PWR in that year. Operating time is determined in the same manner used to estimate operating time for the NRC Performance Indicator Report published by the Office for Analysis and Evaluation of Operational Data (AEOD). Operating time excludes all outages greater than 72 hours8.333333e-4 days 0.02 hours 1.190476e-4 weeks 2.7396e-5 months of time. The increase in the operating years over the rtudy period is mainly because of the low power license grant:d to 16 PWRs in the first 4 years of the study and to 3 more PWRs after that period. Reduced operating years in 1993 is mainly due to decommissioning of the three PWRs in 1992. During the study period, the total operating period for the U.S. PWRs is about 638.00 years.

Investigation of the data presented in Figure 2 2 shows a statistically significant decreasing trend in the frequencies of the reportable leak events as presented in Figure 2-4. The solid line shows the estimated trend and the dotted lines show 90% confidence bands. For this investigation, the trend, if present, is assumed to have an exponential form. The data are consistent with this modeling assumption because all point estimates of 90% confidence intervals overlap the confidence bands. Excluding the first 4 years from the analysis, the

~

reportable leak events that occurred since 1988 do not show any statistically valid trend. Appendix C presents details of the statistical investigation.

The distribution of reactor enolant leak events based on the operators' ability to isolate the leak is presented in Figure 2 5. An isolable leak is defined as the ability to close a valve or place a valve on its backseat to stop the leak. A leak that was isolable with difficulties could have been isolated because of plant design; however, because of a personnel hazard associated with such a leak, it could not be stopped until the plant changed operating modes. Non isolabl: leaks include all piessure boundary leaks and some non pressure boundary leaks. Examples of non isolable, non pressure boundary leaks are leaks associated with RCP seals and safety relief valves. There were 67 unisolable leaks; 45 of which were pressure boundary leaks. Seven leaks were isolable leaks but the environment resulting from the leaks made it difficult to isolate them.

2.2 Distribution of PWR Primary System Leak Rates Distributions of the reactor coolant leak rates for 153 events, which occurred inside the containment, partitioned by the magnitude of the initial and maximum rates are presented ia Figures 2-6 and 2 7, respectively. The inual leak rate was less than 3.8 Umin (I gpm) for about 31% of the events, whereas it was greater than 76 Umin (20 gpm) for about 6.5% of the events. The initial leak rates were not reported for about 24% of the events. The maximum leak rtte was less than 3.8 Umin (I gpm) for about 29% of the ,

events, whereas it was greater than 76 Umin (20 gpm) for about 7.2% of the events. The maximum leak rates w ere not reported for about 25% of the e,'ents. The highest leak rate was 760 Umin (200 gpm), which resulted from a transient induced LOCA event that occurred at Fort Calhoun on March 7,1992. A pressure ,

transient caused a pressurizer safety valve to lift. Safety injection was demanded and subsequently the plant had to depressurire and cool down. The risk significance cf this event and the Oconee 3 evat. discussed next, is evaluated in Section 9.

A failed compression fitting on the RCS hot leg instrument line resulted in a leak event with a maximum leak rate of 494 Umin (130 gpm).11e event took place at Oconee 3 on November 23,1991. This was an isolable leak, but could not be isol3ted until the plant was placed in cold shutdown and airborne activity levels decreased sufficiently to allow containment entry. An estimated 330,600 litters (87,000 gallons) of DRAFT - NUREG/CR-6582 22

TRENDS AND DOMINANT CAUSES ,

reactor coolant was discharged to the containment before the leak was isolated. His event is discussed in Section 8.3.

The distribution of the reactor coolant leak rates for 45 events, which occurred outside the containment, partitioned by the maximum leak rates, is presented in Figure 2 8. Tu maximum leak rate was less than :

I, 3.8 Umin (I gpm) for 33% of the events, but was greater than 76 L- m (20 gpm) for about 11% of the i events. He highest maximum leak rate was 304 Umin (80 gpm), v :

was caused by a failure of a compression fitting on an instrument line in the chemical volume control system at McGuire 2 plant on March 22,1993.

2,3 Distribution of PWR Primary System Leak Locations Re distribution of the pressure boundary leak events (4 $ events) by calendar year is shown in Figure 2 10.

Investigation of the conesponding frequency data, also shown in Figure 2 10, indicates no statistically valid trend (see Appendta C) Maximum leak rate was less than 3.8 Umin(1 gpm) for 69% of the events, between 3.8 and 19 Umin (I and 5 spm) for 13% of the events, and not known for the remaining 18% of the events.

The distribution of the reportable leaks according to the system in which they occurred is presented in Figure 211. Two spiems had significantly more leaks than other systems: pressurizer (48 events) and charging / letdown (42 events). The leaks in the pressurizer system were generally unisolable, whereas those in the charging / letdown sptem were easily isolated. Three other systems had a reasonable number ofleaks:

rnain coolant piping. RCP, and charging / safety injection (SI) systems, nese five systems account for about 72% of the reportable leak events.

The distribution of the reportable leaks by the component from which they occuned is presented in Figure 2-12. He two components with the most leaks were the valves and pipes. The highest leak rate from a valve was 760 Umin (200 gpm) at Fort Calhoun as discussed. The next two highest leak rates,179 and 171 Umin (47 and 45 gpm), were caused by packing degradation and occurred in 1987 and 91, respectively.

The highest leak rate in the piping,331 Umin (87 gpm), was caused by vibratory fatigue and occurred in 1986.

The distribution of the reportable reactor coolant leaks by their location is presented in Figure 2 11 The leaks predominantly occurred either at the welds (23%) or through the valve packing (21%). The next two prominent locations were valve seats and seals, and included seven events at the RCP seals (See Section 5.4).

Ten leaks were through the base metal, considered an unusual location, with seven of these associated with

'* the pressurizer system.

The distribution of leaks through the welds in pipes and instrument lines versus the pipe size is presented in Figure 2 14. All but one leak took place in piping with a diameter smaller than 4 in., while about half occurred in piping with a diameter smaller than 25.4 mm (1 in.). Twenty eight leaks were caused by vibratory fatigue.

The distribution and frequency of the leaks that occurred through the bolted connection is shown in Figure 215. Investigation of the frequency data indicates no statistically valid trend (see Appendix C).

Thirteen of these leaks occurred through the bolted connection on valves, mainly through the gasket. The highest leak rate was 57 Umin (15 gpm), resulting from boric acid corrosion of studs on the valve closure.

2-3 DRAFT NUREG/CR-6582

_ _ _ _ _ _ _ _ ~ _ _ _ _ __ _ _ -____ _._

TRENDS AND DOMINANT CAUSES 2A Dominant Causes of PWR Primary System Leaks Degradation mechanisms were identified for 124 leak events as shown in Figure 2 16. Vibratory fatigue and packing degradation are the major degradation mechanisms causing primary system leaks, and are associated with 29 events each. The third major degradation mechanism is stress corrosion cracking (SCC), and is associated with 16 events. Distribution of leak events associated with the three major degradation mechanisms by the calendar year is presented in Figure 217. Leak events associated with packing degradation show a significant decreasing trend with none reported since 1991. Leak events associated with other two mechanisms da not show any trend m calendar time. However, the leak events associated with vibratory fatigue do show trend in age (time in senice). Trends associated with packing degradation and vibratory fatigue are analyzed in Sections 4.5 and 7.3, respectively.

The distribution of these events by the type of SCC mechanism is presented in Table 2 1. Ten events were caused by PWSCC,3 by transgranular stress conosion cracking (TGSCC),2 by intergranular SCC, and the SCC mechanism is not known for one event. For the 10 leak events caused by PWSCC, leak occurred ,

through the Alloy 600 base metal in eight events and through the weld metal used with Alloy 600 components in one event; the location for the 10th event was not reported. The maximum leak rate associated with these events is less than 1.9 Umin (0.5 gpm). Selected SCC events are discussed in Section 6.

Distribution of reactor coolant leak events by contributing factors is presented in Figure 2-18. A maximum of three contributing factors are included in the database for several leak events. The contr buting factors for 47 leak events (24% of the events reviewed)~ were not known. Installation enor was the main contributing factor affecting about 16% of the leak events. Fabrication defect and component failure were the other two major contributing factors affecting about 10% and 8% of the leak events, respectively, For about 13% of the events, the contributing factors were other than those listed in the Figure 2-18.

2.5 Consequences of the Leak Events Distribution of the reactor coolant leak events by the consequences of the events is presented in Figure 2 19.

Two or a maximum of three consequences are included in the database for several leak events. Reactor trips occurred after identifying leakage in eight events; three were manual trips and directly caused by the leaks,

whereas the remaining 5 were automatic trips possibly caused by the actions taken to manage the leaks.

l Reactor trip is not reported as a consequence of a leak if the leak was detected after the reactor trip occurred.

l Figure 219 also shows that a hot or cold reactor shutdown occurred following detection of 85 leaks.

Additionally, forcehtage, which implies either a hot or cold shutdown, occurred following 21 leak events.

Entryinto contahe <as made in 78 events to determine the leak location. Off site release of radioactive ,

gases occurred in thm events and safety in,*,ection actuation occurred in four events.

2.6 Actions for Leak Mitigation Distributions of the reactor coolant leak events by the short term and long-term corrective actions as identified in the LERs are shown in Figures 2 20 and 2 21, respectively. For most leak events, one corrective ection was identified in each category, but for some events, two corrective actions were identified.

Repair was the main short term corrective action (111 events), whereas inspection / testing and failure DRAFT - NUREG/CR-6582 2-4

TRENDS AND DOMINANT CAUSES }

evaluation were the major long term corrective actions. Other long term corrective actions included repair, replacement, procedure change, and design change.

The decreasing trend in the number of reponable leak events (see Figures 2 2 and 2-4) imply that several corrective actions were effective in leak mitigation. Conective actions for valve packing degradation are not identified in the related LERs. However, significant decrease in reportable leaks through valve packing implies that most PWR licensees have implemented some valve stem packing improvements discussed in Section 7.2.

e 4

9 2-5 DRAFT- NUREG/CR 6582

l i

3. ANALYSIS OF LEAK EVENTS ASSOCIATED WITH THERMAL FATIGUE CRACKING OF PWR BRANCH LINES ne PWR branch lines include small-diameter piping connected to the main coolant loops. The three main branch lines are the r,afety injection (also called high-pressure safety injection), residual heat removal (RHR)

(also called shutdown cooling), and charging (also called makeup) lines. De nozzles of these lines were considered susceptible to fatigue damage and were analyzed at the design stage for plant heatup and cooldown cycles and thermal shock loadings resulting from sudden injections of cold water. However, unexpected thermal phenomena have caused fatigue cracking in the unisolable portion of a few branch lines at several different US and non US PWRs. Unlike surge lines, for which the main phenomenon causing fatigue darnage is thermal stratification, the phenomenon causing fatigue damage in the branch lines is more complicated; this phenomenon is called thermal cycling. Other phenomena, turbulent mixing and flow induced vibrations, have also contributed to damage.

He design requirements for the reactor coolant piping are presented in Section 3.1. Then the presenice inspection (PSI) requirements are presented. His is followed with a description of the phenomena causing thermal fatigue to the branch lines. The main focus is on the thermal cycling. Other phenomena that play secondary roles are also discussed to provide a comprehensive picture, nen world wide data and U.S. PWR data related to thermal fatigue cracking of branch lines are presented. Sections 3.1 to 3.3 provide background information for an analysis of a recent pressure boundary leak event at Oconee 2. A detailed analysis of this event is presented. This is followed with a summary ofinsen ice inspection (ISI) requirements for the RCS branch lines for detecting and sizing fatigue cracks. Then USNRC Bulletin 88-08 and the utilities responses to the bulletin are discussed.

3.1 Reactor Coolant Piping Design Requirements The design, manufacture, processing and testing of materials and product forms for the RCS piping are generally subject to regulations, codes, and standards, which may be divided into three levels. On the top level are federal regulations. On the next level, codes and standards referenced by the federal regulations, and on the bottom level, codes and guidance (such as regulatory guides) intended to ensure quality in fabrication.

3.1.1 Federal Regulations The Code of Federal Regulations Title 10, Part 50 (10 CFR Part 50), " Domestic Licensing of Production and Utilization Facilities," contains rules for the design, construction, operation, and inspection of nuclear powcr plants. 10 CFR 50.2, Definitions, defines the reactor coolant pressure boundary. 10 CFR 50.55a, Codes andStandards, establishes a set of acceptable standards, the Boiler and Pressure Vessel Code of the AShiE, for design, fabrication and inspection. Only AShiE Code Section III, Rulesfor Construction of Nuclear Plant Components (AShiE 1995a) and Section Xl, Rulesfor Inservice inspection ofNuclear Power Plant Components (AShiE 1995b) are endorsed by the regulations, ne three appendices of 10 CFR 50 relevant to the RCS piping are Appendices A, B, and G.

31 DRAFT- NUREG/CR-6582

l TiiERMAL FATIGUE CRACKING Appendix A, GeneralJ:s ogn Critertafor Nuclear Plants, lists the criteria that the plant design must meet.

The relevant general design criteria (GDC) are as follows:

GDC 1 and 30 relate to quality standards for design, fabrication, erection and testing, e GDC 4 relates to compatibility of components with environmental conditions, and

GDC 14 and 31 relate to the extremely low probability of rapidly propagating fracture and gross rupture of the reactor coolant pressure boundary. ,

Appendix B, Quality Assurance Criteriafor Nuclear Power Plants and Fuel Reprocessing Plants, requires a description of the quality assurance program to be applied to the design, fabrication, construction, and

testing of nuclear power plant structures, systems, and components that prevent or mitigate the consequences of postulated accidents that could cause undue ilsk to the health and safety of the public.

Appendix G, Fracture Toughness Requirements, relates to materials testing and acceptance criteria for the fracture toughness of the reactor coolant pressure boundary components. It requires that the pressure retaining components made cf ferritic materials meet the requirements for fracture toughness during system hydrostatic tests and any condition of normal operation, including anticipated operational occurrences.

3.1.2 ASME Code Section 111 Requirements ne federal regulations require that the reactor coolant pressure boundary design conform to the ASME Code. He part of the ASME Code applicable to fabrication is Section 111, Rulesfor Construction ofNuclear Power Plant Components. Before the inclusion of piping design rules in the ASME Code,Section III, piping and components were designed to ANSI /ASME B31.1, Power Piping or B31.7 Nuclear Piping. He ASME Code identifies specifications for permitted materials in Section III, Appendix !! Afaterial Properties (ASME 1995a).

The part of the ASME Code applicable to ISI isSection XI, RulesforInservice Inspection.Section XI states the minimum ISI requirements for those components. PWR ISI requirements are typically written around four 10 year inspection intervals (Inspection Program B of Section XI) to cover the 40-year operating life (ASME 1995b).

3.1.3 Other Sections of ASME Code and Regulatory Guides Other sections of the ASME Code and documents such as USNRC regulatory guides are used for fabrication.

Rese incIude ASME Code Section11 Afaterials,Part A Ferrous Afaterials Spec (fications, Part B Nonferrous Afaterials Spec {fications, and Part C Specificationsfor Welding Rods. Electrodes, andFiller Afetals,Section V Nondestructive Examination, and Section 1X Welding and Bra:ing Quahfications. Here are numerous regulatory guides that give guidance, not requirements, to control weld quality, prevent sensitization of stainless steels, and other such matters. For a list of relevant regulatory guides see the USNRC Standard Review Plan, NUREG 0800, Section 5.2.3 Reactor Coolant Pressure Boundary Afaterials.

DRAFT - NUREGICR-6582 32

THERMAL FATICUE CRAClGNO 3.1.4 Design Basis (Expected) Transients Design of the RCS piping components is based on the expected number of transients during the plant operation. These expected transients are called design basis transients and include between 200 end 500 plant heatup and cooldown cycles as well as cycles for several other transients. Different plants supplied by the same vendor may have a different set of design basis transient:. Rese transients are based on maximun; anticipated events and in most cases consist of consenative assumptions both in terms of aaticipated number of cycles and in the severity of the transients. These transients were originally estimated in the plant design specifications and are included in the plant Fina10afety Analysis Reports (FSARs). He fatigue analyses of the components subjected to these design transients were performed before the plants were actually built and operated. Rese fatigue analyses serve as a basis for verification of adequate margin of safety against the initiation of a fatigue crack.

In addition to the conservative numbers of cycles, thermal, stress, and fatigue analyses often include conservative assurnptions related to the transients. As long as t'ie cumulative usage factor can be demonstrated to be less than the ASME Code allowable limit of 1.0, analyses can be simplified to reduce computational effort, if the cumulative usage factor is greater than 1.0 in the initial calculations, then refinements climinating conservative assumptions are progressively applied until the limit is met. His cumulative usage factor is reported in the FSAR, but still is greater than would be calculated if all conservative assumptions were climinated. As an example of a conservative assumptions in the thermal analysis, the actual heatup or cooldown rates may be considerably less than the corresponding design basis transient. Therefore, the temperature gradients used in the thermal analyses are more severe than actually occur. Ware et al. (1995) and Deardorff and Smith (1994) identify several other conservatisms present in the initial themial, stress, and fatigue analyses.

In calculating the cumulative fatigue usage factor for a particular component, the contributions from all applicable transients should be considered. However,in actual practice some transients are important frem fatigue considerations, whereas other transients may be relatively less important for the same component.

Experience has shown that certain types of loadings such as heatup and cooldown were anticipated at the design stage, whereas others were not. For example, thermal stratification and thermal cycling, which were not considered at the design stage, have occurred in the PWR safety injection, RHR, and charging lines and caused cracking in both weld and base metal materials in these lines.

Testing of systems and equipment can be a major stressor for some components. Some of the testing conditions were anticipated in the design phase, whereas others, for example testing of safety injection and charging system valves, were added after the plant started operating.

The plants cre also designed to accept unusual conditions such as severe earthquakes and dynamic loadings resulting from catastrophic pipe failure in connected systems.

3.1.5 Fatigue Analysis of Piping by ASME Code Method Prior to 1969, nuclear piping was designed using United States of Ameries Standard (USAS) B31.1, from 1969 to 1971 plants were designed with USAS B31.71969 as the standard, and the ASME Code has been used thereafter. Piping systems purchas-d for nuclear power plants prior to July 1,1971, generally used the rules set forth in B31.7. Those purchased after July 1,1971, generally used the rules of ASME Code, Section 33 DRAFT - NUREG/CR 6582

THERMAL FATIGUE CRACKING 111. The rules of B31.7 were incorporated L subarticle NB 3600 of the 1971 edition of Section 111. Over i

time, there has been some divergence betweet .he B31.1 and Section 111 Codes, but the basic concepts still remain very similar (Wais and Rodabou@ 1997).

USAS 831 ne Piping Code of the USAS Institute (formerly the American Standards Association (ASA)) was originally published in s935 as the American Tentative Standard Codefor Pressure Piping (ASA B31.1). It was focused on satisfying primary stress limits, and did not specifically address fatigue, which was assumed to be covered by design safety factors on primary stresses. The 1955 issue of ASA B31.1 is priticularly noteworthy in that it introduced several new concepts into the piping code. Standard equations for piping design were included; fatigue failures caused by expansion stresses were considered; and the concepts of stress range and maximum shear stress, as pertinent to the fatigue of piping systems, were used. De quantitative evaluation of local expansion stresses was introduced through stress '

intensification factore. Fatigue was addressed by stating that the expansion stress Se could not exceed tne allowable stress range Sx, which included a stress reductiou factor f, as follows:

S, = f (1.255, + 0.25S3) whcre:

S, = the basic material allowable stress at the minimum cold temperature S, = the basic innterial allowable stress at the maximum hot temperature f = 1.0 for s 7,000 cycles, gradually reducing to 0.5 at > 100,000 cycles his equation is still used by the chemical, petroleum, and power industries, with minor modifications.

When the first generation nuclear power plants were designed in C.e mid 1950s, the only basis for design and fabrication of piping was the AS A B31.1 1955, Codefor Pressure Piping. He critical nature of nuclear power plant piping demanded something beyond the minimum requirements of ASA B31.1 1955. Designers specified many requirements themselves, such as ordering materials to existing ASTM specifications. As time went on, starting in about 1962, many of these supplemental, but necessary, requirements were eventually incorporated into the Nuclear Code Cases. Much of this experience was later consolidated in the USAS B31.1 1967, Power Piping Code which was commonly referenced for early nuclear plants.

Piping designed to B31.1 is generally thicker than piping designed to the present ASME Code. This results in pressure and moment stresses tending to be lower in B31.1 piping, but stresses caused by local thermal -

gradients can be more severe. A number of PWRs have had their surge lines originally designed to B31.1 reinalyzed (most to the 1986 ASME Section til edition) to include thermal stratification transients.

Cumulative fatigue usage f actors were calculated to meet the ASME Code allowable value in all cases. Ware et al. (1995) have analyzed two B31.1 piping systems and two B31.1 branch nozzles to the 1992 ASME Code edition and found that, based on representative transients, all four would meet the ASME Code allowable value for the usage factor.

In 1969, USAS B31.7, Nuclear Power Piping, was issued specifically for nuclear piping. USAS B31.7 1969 provided design rules for three classes of piping. This included a set of rigorous design rules for Class I piping. whereas the design of Classes 2 and 3 piping were perfonned in accordance with USAS B31.1 (1967 or earlier edition), with slight modifications. USAS B31.7 1969 introduced three fatigue curves to the B31 DRAFT - NUREG/CR 6582 3-4

T11ERMAL FATIGUE CRACKING piping standards: curves for carbon and alloy steels with meal temperatures not exceeding 700'F (one for ultimate tensile strength (UTS) s 80 ksi and one for UTS 115 to 130 ksi); one for austeniti: stainless steels, nickel iron-chromium, nickel chrome iron, and nickel-copper alloys with metal temperatures not exceeding 800'F; and cun es for steel botting materials.

De USAS B31.71969 requirements were very similar to the existing ASME Section 111 requirements for nuclear vessels. The piping for some currently operating nuclear plants were designed using USAS B31.7 1969. Piping analysis requirements were incorporated into Section 111 of the ASME Code in 1971.

ASME Code. The ASME set up a committee in 1911 for the purpose of formulating standard rules for the constniction of steam boilers and other pressure vessels. In the er.rly 1960s, the rules and philosophy of ASME Code Section VI!! closely paralleled that of the power piping sections of USAS B31. A few early plants (for example, Yankee Rowe, San Onofre 1, and Haddam Neck) were built to Section Vill. As with piping, there grew the realization that more rigorous requirements were needed for nuclear vessels; consequently, Section 111, Nuc/ car l'essels, was issued in 1963 as a separate code. In this edition, a formal fatigue analysis for nuclear components was provided. Two fatigue chans were included in this version:

one for carbon and alloy stects for metal temperatures not exceeding 700'F, and one for 18 8 stainless steels and nickel chrome iron alloy for metal temperatures not exceeding 800'F. The range was from 10 to 10*

cycles.

In the 1968 edition there were three curves; one for carbon, low alloy, and series 4XX alloy steel for metal temperature not exceeding 700'F (one for UTS s 80 ksi and one for UTS 115 to 130 ksi); one for series 3XX high alloy steels, nickel-chromium iron alloy, nickel-iron chromium alloy, and nickel-copper alloy for metal temperature not exceeding 800"F; and curves for high strength steel bolting materials.

In the 1971 Code, the scope was significantly altered to approximately its presa form. The title was changed from Nuclear l'essels to Nuclear Power riant Components. Class I piping was included in Paragraph NB 3600, taken from USAS B31.71969. He Class 1 fatigue curves were placed in Appendix 1 to Section 111. He title for the ferritic steel curve was changed to substitute "high tensile" for " series 4XX alloy", and the nr.xt curve had "austenitic" substituted for

series 3XX high alloy". Both curves ranged from 10 to 10' cy;les.

De 1983 revision to Section ill extended the fatigue design curve for austenitie steels, nickel-chromium iron alloy, nickel iron chromium alloy, and nickel-copper alloy from 10' to 10" cycles, so that high-cycle fatigue analyses for components made of these materials can be performed. This extension was the first instance in the development of ASME Code fatigue curves to differentiate between base metal and weld zones. The weld zone is defined as the weld and adjacent base metal within 1,5 weld thicknesses on either side of the weld center line. His version of the Code gives three design fatigut curves (A, B, and C) for the base metal outside the weld zone, and two curves (B and C) for the weld zone, that account for the values of the primary plus secondary stress intensity range and the mean stress.

Present ASME Code Requirements. The basic stress (S) versus cycles (N) design fatigue cun es (sometimes referred to as S N curves) follow the relation proposed by Langer:

35 DRAFT- NUREG/CR 6582

TilERMAL FATIGUE CRACKINO 1

S = B N 3

S, where B and S, are constants detennined using linear, least squares regression analyses to the data (Jaske and O'Donnell,1977).

He ASME det.ign cun es were developed by applying a factor of 2 on stress or 20 on cycles, whichever is lower at a given poiat, to the best fit cun e for small, polished specimens. For less than 10,000 cycles, the factor 20 on the cycles give the lower curve. Rese factors are intended to account for size effects, surface naish, statistical scatter of the data, and differences between laboratory and industry environments, but not the effects of a specific coolant The facte 20 on cycles is a product of three subfactors: a subfactor of 2.5 for size,2.0 for d5 scatter, and 4.0 for surft.cc Gnish and atmosphere (llarvey 1980). Manjoine and Tome (1983) assign equal weighing (about 20% each) to account for surface finish, size effects, material '

variability, environrnent, and residual stresses. Large scale carbon steel vessel fatigue tests have been performed in air at room temperature for the express purpose of checking the ASME fatigue design curve (Kooistra et al. )961). It wks shown by these tests that cracks may initiate below the ASME fatigue design curves, but that wall penetration is not expected until the fatigue cycles exceed the ASME design curves by about a fsetor of 3 (Cooper lo92). Recent fatigue tests on ferritic and stainless steel cibows, subject e in-plane bending moments in air at room temperature, have also shown that cracks initiate when the fatigue design curve is reached, and that through wall penetration occurs when the fatigue cycles exceed the design curve by a factor between 2 and 3 (Kussmaul 1988).

3.1.6 Failgue Analyses of Branch Lines Fatigue analyses of branch lines were performed as part of the design. He components most susceptible to fatigue damage were branch line nozzles which were subject to thermal shock when cold water was suddenly injected. The resulting fatigue usage factors were high. Transients such as safety injection initiation, initiation of shutdown cooling (residual or decay heat removal), or loss-of letdown / loss-of-charging impose thermal shocks on the various branch nozzles. NB-3600 is only for piping fatigue analysis, whereas ND 3200 is applicable to fatigue analysis of piping, nozzles, safe ends, vessels, and other components. Ilowever, architect engineering finns, utilniet, and Westinghouse have typically used NB-3600 piping methods for fatigue analyses of branch nozzles and safe ends by treating them as the tenninal ends of branch lines. Combustion Engineering and Babcock & Wilcox (B&W) have typically used NB 3200 methods for branch nozzles and safe ends.

ne NB 3600 analyses results are more conservative than those for NB 3200 analyses for several reasons.

He NB-3600 stress indices for axial thennal gradients are assaned to be worst-case and independent of the weld location or geometry, so the r,ctual fatigue usage from a thermal transient may be less than that calculated. NB 3600 does not require a detailed finite element model for the branch nozzle, so it is considered to be the terminal end of the piping system finite element model. NB-3600 does not differentiate DRAFT - NUREG/CR-6582 3-6 1

TIIERhiAL FATIGUE CRACKING between shop and field welds. He stress indices were based c n the worst case of the AShiE Code approved nonle designs!

Nonconforming branch nonles would require a NB 3200 analysis. Detailed finite element models are used in NB 3200 analyses of branch nonles. His section states that a fatigue strength reduction factor should be used, but gives no guidance on values such as for shop or field welds. He strength reduction fsetor is the reduction in cyclic life caused by a stress concentration such as that produced by a notch. He reduction factor is a function of the stress amplitude as well as the geometry, ne stress concentration factor (SCF) is generally determined analytically whereas the strength reductien factor is determined from test data. He branch nonles are welded to *he main coolant piping in the shop and NB 3200 analyses typically do not include any fatigue strength reduction factors for these welds. So in reality, the expectancy is that the shop weld and base metal have the same fatigue resistance. Ilowever, stress indices are applied to the field welds joining the nonle and the piping. liechmer and Kuhn (1997) have proposed a matrix of weld strength reduction factors as a function of surface condition, weld types and quality levels, which would be applicable to AShiE Section 111, NB. or Section Vill analyses.

The largest cumulative fatigue usage factor for a Oconee 2 high pressure injection (liPI) nozzle was 0.88.

The fatigue anajysis included all design basis loads and stratification loads as estimated with 1989 thermocouple measurements. The usage factor for the Oconee 2 MU/liPI nonle was bounded by that for the IIPI nor.rle, but the corresponding fatigur analysis did not include the thermal fatigue loads associated with the 1997 Oconee 2 cracking event, which is described later in this section (Duke Power 1997g).

3.1.7 Preservice Examinations NH 5000 of Section ill of the AShiE Code requires nondestructive examinations to ensure that there are no fabrication induced flaws in a component that exceed the acceptance standards. These examinations are performed during fabrication of the componer,t and to a certain extent after the hydrostatic pressure test.

Nondestructive volumetric examination is normally performed throughout 100% of the volume ofeach pipe, fitting, and forging, in the US the volumetric NDE used during fabrication is genwally radiography, w hereas ultrasonic examination is generally used in other countries.

Preservice volumetric examination is universally performed with ultrasonic testing. Surface examination is performed on the entire outer and inner surface of piping. Austenitic surfaces are examined by liquid penetrant method; ferritic surfaces by the magnetic particle method. A typical examination of stainless steel piping welds includes radiographic / ultrasonic inspection of all homogeneous and dissimilar welds and ultrasonic inspection of buttered surfaces prepared for dissimilar welds. Liquid penetrant testing is used for of all weld surfaces.

2 Wals and Rodabough (1997) define stress index as the ratio of the nomin.2l bending stress praiucrngfatiguefailure in a given number of n cles in a ginh weld in a straight pipe of nomin.11 dimensions to rhar pmducingfatiguefailure in the same number of cycles in the component under considerations.

37 DRAFT-NUREG/CR 6582

~ -- - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - _ - _

TilERMAL FATIGUE CRACKING 3.2 Phenomena Causing Thermal Fatigue Cracking in the PWR Reactor Coolant Branch Lines Several different phenomena have caused thermal fatigue damage in PWR reactor coolant branch lines.

Rese phenomena include turbulent penetration and thermal cycling, turbulent mixing, thermal shock, thermal stratification, and thermal striping. nese phenomena are described here. Major emphasis has been placed on turbulent penetration and thermal cycling because it appears that this phenomenon has played a major role in the 1997 Oconee 2 leak event analyzed in this section. Flow induced vibration phenomenon, which causes mechanical fatigue f amage, is also described because it might have played a synergistic role in loosening of thermal sleeve during the Oconee 2 leak event.

3.2.1 Turbulent Penetration and Thermal Cycling Figure 31 illustrates how turbulence in a cold leg of the main reactor coolant piping penetrates into a connecting safety injection line, ne turbulence intensity decays exponentially from the head:r pipe into the branch line, but the temperature remains fairly constant ever the length of several diameters and tl en decays. De length c(turbuient penetration is greater for a higher flow velocity in the main pipe. The length also depends on the layout of the branch line. The length of turbulent penetration in typical PWR branch lines containing stagnant coolant is in the range of 15 to 25 branch line inside diameters. Turbulent penetntion may interact with stratified Guid layers, produce stratified fluid condition, or prevent the development of such conditions.

For a given operating condition, the length of turbulent penetration fluctuates about some average value by a small amount. If such fluctuations produce cyclic axial movement of an interface between hot and cold Guids, it is called thermal cyhng. Such cycling takes place when a column of hot turbulent fluid from the main piping penetrates into a connecting branch line and interacts with thermally stratified Dow. The cyclic changes in the length of turbulent penetration produce corresponding changes in the length of the stratified Guid isyers. As a resuit, the pipe in contact with the interface between the turbulent column and stratified Guid layers experiences cyclic stresses causing fatigue damage. Thermal cycling has been suspected in fatigue cracking and leakage from safety injection lines (Farley 2, Tihange 2) and makeup / safety injection lines (Crystal River 3, Oconee 2), which are connected to the PWR primary coolant piping. These incidents are discussed later in this section.

In some branch lines comaining hot stagnant Guld, such as safety injection lines, the stratified now could be produced by a srnall leakage through the first upstream check valve (in leakage). Ilowever, the temperature difference between the stratified layers can be sustained only for a short distance because of the ,

small leak rate. The cold coolant leaking through the valve will heat up to the temperature of the hot coolant on the downstream side after it travels a r,hort distance. So if the valve is at a sufficient distance away from the branch line connection with the main coolant piping, stratified Guids will not interact with the turbulent penetation colunm and thennal cycling will not take place.

it appears that turbulent penetration alone, under certain changes in operating conditions and with susceptible piping layouts, could produce thermal stratification in a branch line, liowever, the effects of plant operating conditions on the length of turbulent penetration are not well understood. He presence of thermally stratified fluids, in the absence of any valve leakage, has been detected in a branch line of a FWR plant, as illustrated in Figure 3 2. A plausible explanation for this presence of stratified Golds is as follows. De branch line containing stagnant Guid travels a certain distance vertically from the main coolant loop and then DRAIT NUREG/CR-6582 38 l

TIIERMAL FATIGUE CRACKING ri orizontally, ne turbulent penetration initially developed m the vertical section of the branch line, as wn in Figure 3 2(a). Ren, an operational transient such as a power change caused the turbulence to penetrate the full length of the vertical section and produce statification in the elbow at the end and in the adjacent horizontal section of the branch line, as shown in Figure 3 2(b). With further changes in power, the length of the turbulent penetration receded and the stratified layer was no longer present. Ti.us, the base metal and welds of the elbow and the horizontal portion of the branch line experienced cyclic thermal stresses and fatigue damage, nere is another possible scenario in which turbulent penetration can produce thermal stratification in the beanch line as shown in FiF ure 3 2.1. ne stratification is driven by natural convection. For example, a top-to bottorn temperature difference of 78'C (140'F) at the upstream of the first check valve has been reported at one PWR plarit. His is discussed in Section 3.7.

Turbaht penetration can prevent the development of a stratified fluid condition in some branch lines containing stagnant Guids, such as RifR lines, where out leakage from the first isolation valve may take place. ne out4cakage will produce a stratified fluid condition on the downstream side of the valve if the distance between the g alve and the hot leg is large enough so that tae temperature of the stagnant coolant in the vicinity of the vah e is much lower than the hot leg reactor coolant temperature. If the valve leak stops for some reason, the strstified fluid condition will also cease to exist after the layer of hot fluid cools down.

Thus, on'off leakage from the valve will impose cyclic thermal stratification loads on the pipe. This phenomena has caused thermal fatigue cracking of a RilR line in Genkal 1, a Japanen PWR plant, which is described later in this section. If the isolation valve were within the turbulent penetration length, the coolant on the downstream side of the valve would have been at the hot leg reactor coolant temperature and the stratified fluid condition would not have been developed.

He length of turbulent penetration in branch lines not containing stagnant fluid, such as a charging line, will depend on the flow rate of the charging fluid. He length decreases as the flow rate increases. For example, the depth of penetration into a 38.0-mm (1.5 in.) thermal sleeve is equal to 190 mm (7.5 in.),114 mm (4.5 in.) and 0.0 mm. respectively, for flow rates of 38,114, and 19011 min. (10,30, and 50 gpm)(Duke Power 1997a).

3.2.2 Turbulent Mixing Turbulent mixing of hot and cold coolant can induce local cyclic stresses on the inside surface of the adjacent piping wall. These stresses are caused by random oscillations of the fluid temperature in the mixing zone.

The magnitude of the metal temperatures oscillations is smaller than the difference in the hot and cold coolant temperatures because of the finite heat transfer coefficient and thermal inertia of the pipe v,all. The amplitude of the resulting thermal stresses is highest on the inside surface and attenuates rapidly through the thickness, nerefore, the high cycle fatigue damage caused by these stresses is limited to the inside surface adjacent to the mixing zone. As the direction of the thermal stresses changes randomly, the resulting fatigue cracks have random orientations.

Turbulent mixing of hot and cold coolants have caused high cycle fatigue cracking of cold leg piping in a B&W demonstration plant (Flynn et al.1975) and in the MU/IIPI injection line safe end in several other 3&W plants (B&W 1983). These events are discussed later in this section. Turbulent mixing has also caused cracking in PWR steam generator feedwater nozzles and boiling water reactor (BWR) reactor pressme vessel feedwater nozzles. Turbulent mixing of cold feedwater leaking through the thermal sleeve-37 DRAFT NUREG/CR-6582

TilERMAL FATIGUE CRACKING aozzle joint and the hot coolant in the steam generator has caused high-cycle fatigue crack initiation :t the feedwater nor21e bore and blend radius. Si nitarly,in a steam generator dere the auxiliary feedwater is discharged directly into the steam generator, leaking coolant has caused cracking at the auxiliary feedwater norile blend radius (Westinghouse 1989). Turbulent mixing has also caused similar crack initiations at the HWR feedwater nor21e blend radii at several BWR plants (Gordon et al.1987).

3.2.3 Thermal Shock initiation of relatively cold charging or safety injection flow may impose thermal shock on thermal sleeves, which are, at the beginning of the flow, at the higher reactor coolant temperature. Dermal shock loading is a dynamic loading of a short duration. In less than few minutes, the the mal shock stress magnitude increase from zero to maximum and then back to zero. Thermal shock stresses are maximum at the inside -

surface and attenuate rapidly through the thickness. So a thermal shock load can be a crack initiator but it is not likely to came significant growth of an existing erack.

Among the most severe of the charging line thermal transients is the less-of charging event followed by resumption of charging flow. This event may be caused by loss ofletdown flow, loss of the charging pumps, chemical and volume con'rol system isolation for maintenance, or leak-check procedures. When the charging flow stops, the charging inlet nozzle and thermal sleeve, which are normally at the charging water temperature, are heated to the RCS cold leg temperature of about 290*C ($$0'F). When the cha1ging flow is reinitiated, the charging nor21e and sleeve are subjected to a thermal shock as the colder fluid in the charging line flows past the nor21e. His type of severe thermal shock loading can cause considerable fatigue ,

damage to the thermal sleeve over time.

The safety injection transients of concem also include :he initiation and termination of safety injection flow in which the initial safety injection nozzle fluid temperature, which is equal to . hat of the cold leg, undergoes a step decrease to the safety injection tank (called borated water storage tank in Figure 3 7) temperature of about 50*C (120'F) when the flow is initiated. These transients include safety injection system actuations during both periodic system tests and operation.

3.2.4 Thermal Stratification Stratified flows take place in horizontal sections of branch lines containing hot fluid when cold fluid is injected at a sufficiently low rate or if valve leakage is present. Stratified flow consists of hot and cold fluids separated into two layers with an horizontal interface as shown in Figure 3 3. The lighter hot fluid is present in the upper layer and the heavier cold coolant in the lower layer. De elevation of the interface layer (height of the cold fluid layer) and its thickness depend primarily on the mass flow and hsity ratio. The propensity for stratification of a fluid in horizontal piping can be correlated to its Frouo number, which is the ratio of the inertial force (velocity head) to the force of gravity (buoyancy hea1) acting on the fluid. He velocity head is determined by the relative velocities between the two fluids. He buoyancy head is caused by the density difference between the hot (top) and cold (bottom) regions of the fluid in the pipe, and its magnitude is determined by the difference in temperature between the two regions. Low flow rates and high temperature differences (small Froud number) promote thermal stratification in horizontal piping.

Stratified fSw introduces through-wall axial and circumferential bending stresses whose magnitudes are determined by the top-to bottom temperature difference (TrTe), and the elevation and thickness of the interface layer. In long horizontal pipe supported at the cads, this stress distribuaon is comparable to the DRAFT - NUREG/CR-6582 3 10

TilERMAL FATmUE CRACKING stress distribution in a bimetallic strip subjected to a uniform temperature change. De poition of pipe below the mixing layer experiences tensile axial stresses, whereas the portion above the mixing layer experiences predominantly compressive axial stresses. The theoretical maximum axial $b esses are near the mlxing layer.

A presence of intermediate supports on the horizontal pipe will modify the stress distribution.

Circumferential bending stresses are produced because of the temperature difference between the pipe sections above and below the interface layer. Ilowever, these stresses are of much smaller magnitude than the axia) ones (Woodward 1983,Talja and Hansjosten 1990).

Flow rate variations cause the interface layer to be raised or lowered. In the charging and safety injection systems, these changes, for example, are caused by initiation and termination of the cold coolant flow. As the Dow rate of the cold coolant changes, the layer moves between the upper and lower portion of the pipe eross section and changes the through wall bending stress distribution introduced by the stratified flow, wn;ch causes low cycle fatigue damage, nese cyclic stresses are throughwall and contribute to both crack initiation and growth.

Dr pipkg inside surface temocratures, under steady state conditions, are about equal to the temperatures of the fluid that is in contact with the piping at that location. The resulting piping wall temperature distribu-tbr is piant 3pecine because, in addition to the hot and cold coolant temperature and flow rate, it depends en f$nMayout. Rese low frequency fbetuations in the piping wall temperatures can be estimated oy uud.oring the outside surface tempe atures.

When thermal stratification persists over a long horizontal section of a line, the large temperature difference (Tn-Te) will cause the a line to bow (macroscopic displacement). As the flow condition changes from stratification to no stratification and back to stratification, the amplitude of the resulting cyclic stresses will depend on the line layout, end constraints, and the support system, his type of stratification causes significant low cycle fatigue damage and a tenued global thermalstrat(fication (Su 1990).

If therm *.I stratificationi is present in a short horizontal section of a pipe, changes in the elevation of the interfan between the hot and cold coolants would impose cyclic through wall bending stresses that cause fatigue damage without any macroscopic displacement of the pipe. To differentiate this type ofstratification from global thermal stratification, it is termed cylic thermal strat(/ication.

3.2.5 Thermal Striping nermt,1 stratification can induce local cyclir tresses in the portion of the pipe near the inside surface and adjacent to the interface between the hot and cold coolant layers (shown in Figure 3-3), if the relative flow velocity between the coolants is sufficiently high. The oscillations of the fluid temperature at the interface resulting from mixing of the hot and cold Guld layers impose cyclic stresses on the adjacent pipe wall. Such interfacial mixing results in a process called thermalstriping. The onset ofinterfacial mixing that leads to thermal striping can approximately oc correlated with the initiation of a Kelvin lielmholtz instability, which occurs when inertial forces overcome stratifying density differences between the fluid layers (llafner and Spurk 1990).

Wolf et al. (1987) have conducted thermal stratification experiments in horizontal feedwater lines at the IIDR (lleissdampfreaktor) test facility in Germany. The exneriments were performed at several different now rates with thermocouples mounted on the pipe inside surface to measure the metal temperrtures.

Thermal striping was observed only at relatively high flow iates (Deardorfret al.1990). The test restats and 3 11 DRAFT- NUREG/CR 6582

THERMAL FATIGUE CRACKING l l

theory indicate that thennal striping is present when the gradient Richard:on number through the interface is less than 0.25 (Hafner and Spurk 1990, Tumer 1973). ne gradient Richardson number is the ratio of the density gradient and horizont::l velocity gradient. The typical frequency content of the metal temperature oscillations caused by thermal striping wra between about 0.01 and 21h, and the amplitude of the metal temperatures oscillations was less than 50% of the difference in the hot and cold coolant layer temperatures, ne amplitude of the metal temperature oscillations was smaller because ofthe finite heat transfer coefficient and thermal inertia of the pipe wall. The magnitude of the thermal striping stresses is highest on the inside surface and attenuates rapidly through the thickness. Therefore, the high-cycle fatigue damage caused by these stresses is limited to the pipe inside surface adjacent to the interface.

3.2.6 Flow-induced %brations (Contributory Phenomenon)

Flow induced vibrations does not cause thennal fatigue damage but it can contribute in creating conditions favorable to producing thermal fatigue loads in componentuuch as thermal sleeves in the branch nonles.

The thermal sleeves are installed to protect the noules frm thermal shock but are thin walled tubes that '

cannot withstand high mechanical loads. This stressor could lead to fatigue failure of the thermal sleeve and could cause the thermal sleeve to break loose and move through the piping system. Loss of a thermal sleeve could ultimately lend to fatigue failure of the noute.

Flow induced vibrations in several B&W plants have caused high-cycle mechanical fatigue damage to the thermal sleeves that resuhed in their breakage. nese thermal sleeves were instslied in the nonles with a soft roll (see lower part of Figure 3-4), which resulted in the loose fit. An improved thermal sleeve, shown in the upper part of Figure 3 4, was designed with a hard roll that resulted in a tight fit in the noule and eliminated the problem (B&W 1983).

3.3 Trends in Thermal Fatigue Cracking nermal fatigue cracking leading to leakage is not widespread in the nuclear power plants. Most of the reported leakage events have resulted from feedwater pipe cracking. Most of the failures have occurred because of the degradation mechanisms that were not taken into account in the original design as discussed earlier in the section. First the world wide data for thermal fatigue cracking of Light Water Reactor (LWR) piping are presented. This include both primary and secondary piping. Then the failure data for the PWR RCS piping are presented.

3.3.1 World Wide Data for Thermal Fatigue Cracking ,

ne Swedish Nuclear Power inspectorate (SKI) has compiled a world wide database on pipe failures over the period from 1972 to 1995 (Nyman et al.1996). Thermal fatigue has been reported as a failure mechanism for 69 events as shown in Figure 3 5; 36 are in PWRs and 33 in BWRs. Some of these failures are related to nignificant cracking and not leakage. The number of failures per reactor year in PWR plants is about half of that in BWR plants. In PWR plants, many of these failures are associated with feedwater piping; only a few failures have occurred in the reactor coolant piping.

Figure 3 6 shows the world wide data for leak events in nuclear power plant piping caused by thermal fatigue. Early life failures, that is, failures within the first year of commercial operation, are not included in this figure, ne failures in the PWR plants are first reported a little after 10,000 h of operation, whereas DRAFF - NUREG/CR-6582 3 12

i TIIERMAL FAT 1GUE CRACKIW1 they are first reponed in the BWR plants at about 25,000 h of operation. The failure rate in both BWR and PWR plants are about the same.

Figure 3-6 may be explained as follows.8 he figure shows the cumulative hazard function ll(t)(with the hazard function expressed as a percent) plotted against time. One interpretation of this plot uses the cumulative distribution function F(t) defined as the robability that a weld fails by time t. It can be shown that f(t) is related to the cumulative hazard function by F(t)= 1 exp[ il(tyl00).

For example, at the time when II(t) = 100, F(t) = 1 - exp( 1) = 0.632, so 63.2% of the welds in the population fail by that time.

Figure 3 6 was constructed by considering only pipes (welds and base metal) that failed, not the ones that did not fail. %us, the population considered for Figure 3 6 was all failed pipes over the period 1972 to 1995.

This leads to two warnings for users of the figure:

1. Le plants that contributed the data were at most about 33 years old. Therefore, we have no information about pipes that may fail after 33 years of age. These pipes were not considered as part of the population.

2. Figure 3 6 does not provide guidance in deciding whether a particular pipe will fail. The figure only implies "If the pipe is in the susceptible population, so that it will eventually fall (in the first 23 years),

here is the probability that it wi? fail before time t, for any t ofinterest."

Figure 3 6 may be used, with failure data from the early years, to estimate the size of the susceptible population of pipes in the entire power plant as follows. This estimate is based on the assumption that all the thermal fatigue failures will take place in the first 23 years as shown in Figure 3-6. For PWRs, the figure implies that half of the failures take place in the first 75,000 hours0 days 0 hours 0 weeks 0 months . Therefore, after 75,000 hours0 days 0 hours 0 weeks 0 months we can estimate that the same number of failures will take place in the remaining life.

3.3.2 Therrnal Fatigue Cracking Data for PWR Anactor Coolant Piping Only a few leak events caused by thennal fatigue cracking of PWR reactor coolant piping have been reponed. Rese events took place in the U.S., France, Belgium, and Japan, and are listed in Table 3-1. For all of these events, throughwall cracking was in a unisolable ponion of the reactor coolant piping. Most of the throughwall cracking was in the weld or its heat affected zone. In two cases the cracking was through the base metal of an elbow and a straight pipe.

The first leak event was reponed in 1982 at Crystal River 3 which had been in operation for 5 yrs. Among the plants listed in Table 31, Crystal River 3 had been in opera; ion for the shortest time at the time of failure. 3e throughwall leakage occurred so early in the plant's operating life possibly because the throughwall crack was initiated at both the outside and inside surfaces. The most recent leakage was reported in Oconee 2, which had been in operation for 23 years. The throughwall crack in both Crystal River 3 and Oconee 2 were circumferential and at the same location (the weld at the upstream end of the MU/HPl 8

V. N. Shah, Private communication with B. Lydell, RSA Technologies, June 30,1997.

3-13 DRAFT- NUREG/CR-6582

TilERMAL FATIGUE CRACKING i

nonle safe end), but the Oconee 2 crack was initiated only at the inside surface. This may be the reason why the leakage event occurred much later in the plant's operating life.

3.4 Analysis of Unisolable Pressure Boundary Leak from Oconee 2 Makeup (MU)/HPI Line Weld, April 22,1997 A comprehensive analysis of the 1997 Oconee 2 leak event is presented here. Sufficient background information is presented to complement the analysis of the event. First the B&W MU/IIPI system and the MU/IIPI noule desigu are described. Den the earlier cracking and leakage events associated with B&W MU/IIPI lines are summarized. nis follows with the 1997 Oconee 2 event description, its root cause analysis, a summary of the corrective actions taken by the utility, and the esahistion of the safety consequences of the event.

3.4.1 B&W MU/HPl Sysfam Description The MU/IIPI system is used during both normal and emergency operation and is shown in Figure 3 7.

During normal operation, the letdown storage tank serves as a nonnat suction source for the IIPI pumps.

Ilowever, during safety injection, the borated tank is automatically connected as an additional source. These sources provide injection water until the RCS pressure is low enough for low pressure injection.

The three IIP! pumps are connected in parallel with cross-connecting suction and discharge headers thus enabling any pump to supply any discharge line. During nonnal operation, only one pump is required for makeup and RCP sealinjection. A second pump serves as a backup for normal operation. All three pumps are started automatically by a signal from the engineered safeguard system in the event of a low RCS pressure or a high reactor building pressure condition.

Three principal flow paths are provided from the IIPI pump discharge header: an injection line to reactor coolant Loop A, a RCP injection line, and an injection une to reactor coolant Loop B. Normally, IIPl pump A or B is opecated to supply makeup flow to reactor coolant loop A and RCP seal injection fLw; pump C is used only for injection to loop B in the event of an accident. An additional emergency cross connect line is provided from the llPI pump discharge header to both reactor coolant loops as a backup for safety injection (Meyer 1989).

As the operating B&W plants do not have a regenerative heat exchanger in the makeup coolant line, the typical temperature of the makeup coolan' '49'C (120'F)) is lower than the corresponding temperatures in ,

the Westinghouse and Combustion Engineering-designed plants. Data collected in 1990 indicated that the metal (thermal sleeve wall) temperature was higher, close to 71'C (160'F), because heat conduction from the nonle side, which is at the reactor coolant temperature (Duke Power 1997h). Typical makeup flow for ,

steady state 100% operation varies from 57 to 76 L/ min. (15 to 20 gpm)(Duke Power 1997b).

3.4.2 MU/HPl Nozzle Design ne 63.5-mm (2.5-in.) MU/IIPI and llPI noules with thermal sleeves are located on cold legs of the reactor coolant piping. As uescribed, two nonles on loop A are used for both makeup and HPI, whereas the remaining two nonles are used for only IIPI flow. The noules on loop A of Unit 2 are designated 2Al and DRAFT - NUREG/CR-6582 3 14

-.. . - - . . _ - - -. - - - _ - - - . _ _ _ - _ _ . - _ . = - . _ _ - . _ -

l L

THERMAL FATIGUE CRACKING 2A2. The original design of the MUSIPI nonle is shown in the lower part of Figure 3-4. Typical layout l of Oconee MUSIPI line is shown in Figure 3 8.

Each nonle is shop welded to a Type 316 stainless steel safe end. As the non_le material is carbon steel,

, this weld is a dissimilar metal weld, ne other end of the safe end is welded to a 63.5 mm, Schedule 160, Type 316 stainless steel MUdiPI line [outside diameter = 73 mm (2.875 in.) and wall thickness = 9.5 mm (0.375 in.)). De use of a safe end allows making a dissimilar weld in the controlled shop condition, and eliminates a need for post weld heat treatment in the field.

A thermal sleeve is installed into a nonle assembly to provide a thermal barrier betwwn the cold MUS{PI fluid and the not le, which is at the reactor coolant temperature -288'C (-550'F). The thermal sleeve protects the noni. and the dissimilar metal weld from thermal shock. The sleeve is fabricated from Type 316 stainless steel and its length is 457 mm (16.125 in.), inside diameter 38 mm (1.5 in.) and minimum wall thickness 3.175 mm (0.125 in.). In the original installation, the thermal sleeve was mechanically expanded until contact was established betwwn the outside diameter of the thermal sleeve and the inside diameter of ,

the safe end. There is no documented acceptance criterion for this contact roll installation (Duke Power 1997g).

A 25.4 mm (1 in.) diameter, Type 316 stainless steel warmingline taps into the bottom of the MUSIPI pipe immediately upstream of the pipe / safe-end weld. This line permits a small continuous floit of 11.4 Umin.

(3 gpm) to reduce noule thermal transients causal by changes in normal makeup flow. He warming line flow bypasses the normal makeup flow line, and is present even with normal makeup flow stopped, ne warming line flow is present as long as the HPI pumps are running. The distance from the warming line to the MUSIPl line-to safe end weld is 75.6 mm (3 in.)(Duke Power 1997c).

3.4.3 Earlier Cracking and Leak Events at B&W MU/HPl Lines 1970 TherrnalFatlyue Cracking ofMakeup Nonle Therrnal Sleeve atIndian Point 1, Thermal shock caused low cycle fatigue damage to the circumferential and longitudinal wolas in a 102 mm (4 in.)

makeup-nonle thermal sleeve in Indian Point 1, a demonstration plant (Flynn et al.1975). The thermal shocks were caused by the intermittent makeup coolant flow and the temperature differences betwun the makeup and cold leg coolants, which might have been as high as 215'C (385'F). De design of the thermal sleeve and the failure locations are shown in Figure 3 9. In addition, turbulent mixing of the makeup and cold leg coolants caused high cycle fatigue damage to the cold leg piping. ne damage resulted in about 40 surface cracks, up to 6.4-mm (0.25 in.) deep, on the inside surface of the cold leg piping downstream of the makeup nonle. Several steps were taken to resolve this problem. The minimum makeup flow was increased from 190 to 285 Umin. (50 to 75 gpm) to reduce the severity of the temperature transients and to promote better mixing of the makeup and cold leg coolants. A redesigned thermal sleeve, shown in Figure 310, was installed to reduce the fatigue damage. This sleeve was made from a solid forging with fewer welds and reduced stress concentrations; it was held in place by hard rolling, which gives it a tight fit inside the nonte in addition, the redesigned thermal sleeve was extended an additional 12 mm (0.5 in.)into the cold leg to improve the mixing of the makeup and cold leg coolants. The makeup line connection to the cold leg was also redesigned, shown in Figure 3 10, to ethninate sharp corners.

1982 Throughwell Therinal Fatigue Cracking of Makeup Nonle Safe-End Weld at Crystal River 3. On January 21,1982, an unidentified 3.8 Umin. (I gpm) leak was found when the plant was in 3 15 DRAFT NUREG/CR 6582

TIIERMAL FATIGUE CRACKING operation. Visual inspection revealed that the leak was associated with the MU/IIPI line. As the leak was unisolable, the plant promptly proceeded to cold ' shutdown. Inspection of the safe end revealed that a throughwall, circumferential crack was present in the safe end to check valve weld ne circumferential extent of the crack at the outside surface was 140 degrees. ne crack consisted of two separate cracks; one initiated at the outside diameter and another one initiated at the inside diameter. He one on the outside surface was initiated and propagated by mechanical fatigue caused by pipe vibrations. The one on the inside surface was initiated and propagated by thermal fatigue caused mainly by turbulent mixing of hot reactor coolant and cold makeup water. Thermal shocks during periodic makeup water additions could have played some role in causing fatigue damage. liowever, as compared to IIPI nonle components, MU/IIPI nonle components are less susceptible to thermal shock damage because they are at higher temperature during normal plant operation (B&W 1983, USNRC 1982).

In addition to the throughwall crack, axial cracks with some circumferential components, resulting in checkerboard or # shaped cracking patterns, were present on the inside surface of the safe end, thermal sleeve, IIPl/MU line and warming line. De thermal sleeve cracking was confined to the roll expansion ~

region only. M?tallographic examination showed that the cracks were transgranular.

Later results from the instrumented MU/IIPl nonle indicated that thermal stratification did not play any significant role in the cracking. The results showed that no significant thermal stratification occurs even at the lowest flow rate tested 6.1 IJmin. (1.6 gpm). The safe end remained cool, and the nonle surface temperatures varied by ll'C (20*F). The original design specifications called for a minimum continuous makeup flow of 3.8 to 11.41/ min. (1 to 3 gpm).

ne thermal sleeve was found loose inside the safe end, that is, there was a Eap between the outside surface of the sleeve and the inside surface of the safe end in the region of roll expansion of the sleeve into the safe end. Extensive wear was found on these surfaces. Wear was also found at the collar region of the thermal sleeve and the weld pads. Wear was caused by flow induced vibrations. Based on these evidence of wear marks, it appeared that the sleeve had become unseated and rotated because of flow forces. Thermal cycling could have contributed to loosening of the thermal sleeve because the thermal expansion coefficients for the thermal sleeve and nonle materials are different.

Following the cracking incident at Crystal River 3, MU/IIPI and ilPI nonles were inspected at the seven other B&W plants. Five other plants had similar safe-end cracks that penetrated up to 25% of the wall thickness. Several thermal sleeves were also found to be loose. He original thermal sleeves became loose because they were installed in the nonles with a soft roll (contact roll) as shown in Figure 3-4 (lower part of the figure).

A B&W Owner's Group Safe-End Task Force was formed for g:neric investigation of the MU/liPI cracking problem. The most important finding of the task force was that all cracked safe ends were associated with '

loose thermal sleeves. Ilowever, not allloose thermal sleeves were associated with cracked safe ends. Also, all cracked safe ends were associated with MU/IIPI nonles.

The task force recommended several improvements in the safe-end and thermal sleeve design. De loose thermal sleeves were rerolled to a 5% wall reduction if no cracking was identified. ifcracking was detected, die safe end and thermal sleeve were replaced with an improved design. The replacement thermal sleeve was designed with a hard roll that resulted in a tight fit in the nonle as shown in Figure 3-4 (upper part of DRAFT NUREG/CR-6582 3 16

- - - . - - _ . - - - - - - - . - - - . - . _ - ._ =

THERMAL FATIGUE CRACKING the figure). His sleeve also had a contact roll at the collar to reduce the effects of flow induced vibrations, ne other design improvements were made to prevent the movement of the sleeve during operation.

One of the Oconee 2 thermal sleeves for the MU/HPI noule (called 2Al), which leaked later on April 21, 1997, was not found to be loose and, therefore, not terolled.

He task force also recommended an augmented, periodic, ISI program. Prior to the Crystal River 3 cracking incident, inspections of MU/HPI nozzle assemblics were not required. The augmented program included radiographic testing to ensure that the thennal sleeve is in the proper position and that no ge p exists between the thermal sleeve and the safe end. Ultrasonic examination of safe end and adjacent piping was recommended to ensure that no cracking was present. In 1985, the USNRC issued a generic letter stating '

that the task force recommendations are sufficient to preclude cracking of the MU/HPI and HPI nozzles in the future (USNRC 1985a). Duke Power, the licensee of the Oconee plants, also agreed with the task force

~

recommendations for the augmented inspections as applied to Units 2 and 3. De licensee developed a similar augmented inspection program for Unit 1 because it uses a double thermal sleeve design to which

~

the task force recommendations were not applicable.

3.4.4 1997 Oconee 2 Event Description On April 22,1997, Oconee 2 was taken off the line due to unidentified RCS leakage exceeding 3.8 Umin.

(I gpm). He leakage was initially detected on April 21 when the reactor operator nmN an increase in the reactor building normal sump rate and a change in the rate of decrease in the letdown storage tank, followed by reactor ouilding radiation monitor system alarms, ne leakage increased from 7.6 Umin. (2 gpm) to a maximum of 45.6 Umin. (12 gpm) while the reactor pressure was being reduced. Subsequent containment entry identified the area of MUSIPl line associated with reactor coolant loop A (shown in Figure >7) as the leak source. Subsequent inspection revealed that the leak was through a circumferential crack in the safe-end to MU/HPI pipe weld upstream of the 2Al loop MU/HPI nozzle, as shown in Figure 3 11. He leak was an unisolable pressure boundary leak (Oconee 21997, USNRC 1997a).

The cracked weld and the associated piping shown in Figure 3 12 were removed for a failure analysis. De safe end and thermal sleeve were r removed for the failure anelysis. Examination of the removed components revealed that the througawall crack was located at the weld centerline, contained in the weld fusion zone, and initiated at the laside surface, its circumferential extent was 360 degrees at the inside surface and 77 degrees at the outside surface, as shown in Figure 3 13. The angular orientation in this figure is set with 0 degree at the top-dead center and a clockwise convention looking in the direction of the MU/HPI flow, he throughwall portion of the crack extended from the 344-degree circumferential position to the 61 degree position. Most of the remaining portion of the crack was 30% through wall. The reactor cooant flow in the cold leg is along the 63 degree-to 247 degree position.*

ne examination found that the thermal sleeu was loose, that is, there was a gap between the thermal sleeve and safe end in the contact roll region. The radial width of the gap was about I mm (0.037 in). Note that 4

this particular sleeve was of the 1971 vintage, an originally installed component; it was not rerolled or

A somewhat different RCS flow direction (from the 45 degree to the 225-degree position)is reported by Redmond (1997).

3 17 DRAFT NUREG/CP 6582

.. -- , ...m---.-rwr.---- ..ee- , , - - - ,_m , - . , .. y. ,- .,, ,e, w- -..w_ ier -

m- e r-. v e w,r-mr++,,

l TIIERMAL FATIGUE CRACKING l

seplaced in 1982. Based on the 1982 cracking experience at Crystal River 3 and other BAW plants, it is believed thet the gap is associated with the cracking.

A review of the April,1996, radiographs of the contact roll region revealed that the gap already existed at that time but was not detected because an inappropriate acceptance criterion was used in the analysis of the radiograph. The acceptance criterion was the evidence of an intact thermal sleeve, but the status of the gap i was not addressed. The accep;ance criteria have been revised to address this deficiency and are discussed !

later (Duke Power 1997f). Based on the review of the earlier radiographs of the thermal sleeve, it is i estimated that a significant gap was developed during the 1989 to 1996 period (Duke Power 1997g).

Significant wear was noted on the outside surface of a sleeve. It was along the reactor-coolant. system downstream side (tN 245 degrees location in Figure 313) of the ricevi: as shown in Figure 3-14. The -

deepest [~ l .2 mm (~0.048 in.)] and widest (~1 $0-degree circumferential extent) poh.t of the wear was about 2$ mm (1 in.) downstream of the collar, tapering out to the outside surface of sleeve about 126 mm (5 in.)

upstream of the collar. Significant wear damage was also found both in the contact expansion region and the collar region of the sleeve, extending essentially 360 degrees eround the sleeve. He wear extended nearly through the wall at the deepest point in the collar region. It extended about 1 mm (0.04 in.) in the contset expansion region. Long axial cracks were found at the upstream end of the sleeve; the longest crack was about 229 mm (9-in.)long. Portions of the downstream end of the sleeve were broken ofTand missing; the maximum dimensions of the broken portions of the slees e were 102-mm (4-in.)icag (axial length) by 38 mm (1.5 h..) wide (circumferential extent)(Duke Power 1997c, Redmond 1997).

Multidirectional cracking, shown in Figure 3 15, was observed on the inside surface of the MUSIPI piping near the wanning line penetration. He cracking extended from the safe end to-MU/IIPI line weld to 51 mm (2 in.) beyond the warming line penetration. Penetrant testing of the inside surface of the safe end revealed similar multidirectional cracking shosm in Figure 316. The warming line was sectioned and its inside surface was examined with penetrant testing; the examination did not reveal any cracks.

3.4.5 Root Cause Analysis of the 1997 Oconee 2 Event ne fractographic examination cf :he fracture surfaces of the throughwall crack were performed both visually and with scanning electron microscopy (SEM). The examinations revealed that the crack initiated at several points on the inside smface, propagated in a transgranular mode, and exhibited essentially no branching. The crack arrest marking on the fracture surface indicated that the crack propagation was periodic. Fatigue striations were present on much of the fracture surface; the smallest striation spacing ranged from about 0.7 to 1.5 microns, whereas the largest striation ranged from 2.5 to 4 microns. He spacing ine ease 3 considerably near the very end of the crack, that is,just before the crack breaks open near the outside surface. The striations were typically bowed towards the outside surface indicating crack propagation from inside to outside surface.

He smallest striation spacing on the fracture surface of the longitudinal cracks on the hside surface of the MU/IIPI line ranged from about 0.6 to 1.5 microns. These striations were also bowed towards the outside surface indicating the crack propagation direction from inside to outside surface. Both axial and circumferential cracks were present on the inside surface of the MU/IIPI line between the safe end weld and warming line penetration. The cracking showed uniform axial and circumferential distribution. The deeper cracks were about 1.3 to 2.3 mm (0.05 to 0.09 in.) deep and spaced about 2.5 to 3.8 mm (0.1 to 0.15 in.)

apart. A few axial cracks at the bottom of the MUSIPI line were significantly deeper; the deepest one was DRAFT - NUREG/CR-6582 3-18

TIIERMAL FATIGUE CRACKING about 4.6 mm (0.18 in.) deep. Many shallower ernhs were present between the deeper cracks. All of the cracks were transgranular and exhibited little branching.

He circumferential crr.cks on the inside surface of the safe end were about 0.5 to 2.5 mm (0.02 to 0.1 in.)

deep, whereas the axial cracks were about 0.6-mm (0.025 in.) deep. These cracks were transgranular and exhibited little branching.

Micro hardness measurements were performed to characterize the changes in hardness due to locrJized cold work in the safe end and MU/HPl line material adjacent to the weld. He results indicated that the hardness at these locations was increased by a small amount.

Identification of Cracking Mechanistn. ne presence of striations observed on various fracture surfaces indicate that the damage mechanism was fatigue. De transgranular mode ofcrack propagation and the presence of discrete cracks with little branching 31so provide further support to fatigue as the main damage mechanism. In addition, the overall flatness of the examined fracturs surfaces, indicating the absence of an appreciable amount of plastic deformation, is another distinct feature associated with fatigue.

The presence of oxide films on the fracture surfaces indicates that the cracks were growiug at slow rates, ne biaxial crack pattern observed on the hiside surface indicates that the source of cracking was a thermal phenomenon such as turbulent mixing, ne striation spacings indicate that the source of thermal fatigue loading was not caused by heatup and cooldown cycles but by a higher frequency phenomena, in most of tl= fracture surfaces examined, the striation spacings were on the order of one micron or slightly less. His implies that the failed MU/IIPI line weld has experienced about 10,000 cycles during which crack growth took place. This estimate dxs not include the number of cycles needed for crack initiation, which is many times larger than 10,000. As the number of plant heatup and cooldown cycles are less than 200, the source of thermal fatigue is not directly connected to these cycles. The source hu to be connected to a higher frequency phenomenon.

However, the throughwall crtek was skewed towards the top of the MU/HPI line, whereas the remaining portion of the crack had essentially the same depth. His may imply an applied cyclic bending moment fron.

mechanical vibration or local stratification. But the analysis of the vibrational loads indicates that these loads are quite small. So thermal fatigue is considered to be the main mechanism for cracking.

Ae residual weld stresses likely had little effect on the crack growth. These stresses add to mean stress, but not to cyclic stress amplitude which is the driving force for crack growth. So the residual stresses might have played a role in crack initiation but not in crack growth. The residual stresses will be automatically relieved by crack growth since the crack face becomes a free surface and the remaining stresses are redistributed.

'Ih effect of residual stresses diminishes with increased crack depth and becomes completely insignificant when it reaches about 20% ofwall thicktiess. The stress distribution indicates that residual stress at the crack tip become zero at this crack depth; then the residual stress become compressive (Duke Power 1997c).

Phenornena Responsible for Loose Therinal Sleeve. Both heatup and cooldown cycles and flo w-induced vibrations may be responsible for loosening the thermal sleeve. The turbulent primary coolant penetrates into the annulus between the thennal sleeve and MU/HPl nozzle and tends to heat the thermal sleeve to the cold leg temperature ~288'C (-550'F). The presence of turbule it fluid in the annulus region was experimentally demonstrated earlier in 1971 while evaluating the thermal sleeve cracking at Indian Point 1 (Flynn et al.1975). De makeup water temperature is about 49'C (120'F), so there is a steep 3 19 DRAFT- NUREG/CR-6582

TIIERMAL FATIGUE CRACKING l 1

l temperature gradient across the thermal sleeve wall. During startup and shuidown the reactor coolant temperature is lower and the temperature gradient across the thermal sleeve wall becomes less steep. These thermal cycles may be responsible for loosening the sleeve. !

l Re Indien Point I therinal sleeve became loose after only 8 years of operation, whereas the Oconee 2 thermal sleeve being analyzed here became loose after a considerably longer time. (The Oconee sleeve was not found to be loose during the 1982 examination.) A possible explanation is that Indian Point I had a regenerative heat exchanEer as part ofits makeup system. As a result, the makeup water temperature was i I

fluctuating between 49'C (120'F) and 204'C (400'F) depending upon whether the letdown flow rate was lower or higher than the makeup flow rate. These fluctuations were taking place at every few minutes. So the Indian Point 1 thermal sleeve experienced many more thermal cycles during startup than the Oconee 2 thermal sleeve, which does not experience much fluctuation in 'he makeup water temperature because Oconee 2 does not have a regenerative heat exchanger in its makeup system.

Flow-induced vibrations could have contributed to further loosening of the sleeve after it was initially loosened by the therma' :ycles. The possible mechanical loading resulting from flow-induced vibrations is vortex shedding, which produces low-amplitude pressure pulsations that occur at distinct frequency bands.

He pulsation frequency is proportional to flow velocity (RCS flow). No loosening of sleeve occurs unless the vortex si.edding frequency coincides with the thermal sleeve natural frequency. The calculated vortex shedding frequency was 80 Hz, whereas the natural frequency of a tight thermal sleeve was above 200 Hz.

So, even though vortex shedding does not loosen a tight thermal sleeve, once the thermal sleeve becomes loose, its natural frequency reduces, and then vortex shedding can further loosen the sleeve (Duke Power 1997f, Olsen 1995). He wear marks observed on the outside surface of the sleeve and the possible rotation of the sleeve could have been caused by flow-induced vibrations.

High Frequency TherinalPhenomena Responsible for ThennalFati.que Crackirig. The highet frequency phenomena causing the observed fatigue damage have not been yet identified. It is suspected that turbulent mixing of hot reactor coolant and cold makeup water and thermal cycling are the likely candidates.

Turbulent mixing could have initiated fatigue cracks but its contribution to significant crack growth is not likely because the associated stresses attenuate rapidly through the pipe wall. These phenomena can take place if the turbulent reactor coolant from the cold leg penetrates into the MU/HPl nozzle-to-safe end weld.

A loose thermal sleeve is a necessary condition for such penetration to take place, otherwise the typical 57 to 76 L/ min (15 to 20 gpm) makeup flow at 100% power operation will significantly limit the turbulent penetration length (Duke Power 1997a,d,f). The hot reactor coolant in the annulus region will pass through the gap between the sleeve and the safe end and enter the MU/HPI line. He fluid in the annulus region is likely to be stratified because cold makeup water may entei the lower portion of the annulus through the gap.

So it is possible that stratihed fluid layers may be present in the MU/HPI line when the makeup flow rate is low; however, ti is can not be verified from laboratory results. As the RCS fluid is in a turbulent state, the length of turbulent penetration into the annulus region and the MU/HPI line will be continuously changing with time. This will produce thermal cycling phenomenon.

Duke Power (1997f) has estimated the number of cycles needed for fatigue crack initiation by considering

, a 222 *C (400'F) change in the fluid temperature over a 12-second period. A one-dimensional transient heat-transfer analysis was performed to estimate the temperature distribution through die wall. The amplitude of the resulting cyclic stresses on the inside surface was 172 MPa (25 ksi). With the use of the best-fit ASME fatigue curve for an austenitic material (without incorporating any ASME Code factors applied to DRAFT - NUREG/CR-6582 3-20

THERMAL FATIGUE CRACIONG develop the fatigue design curve), it is estimated that crack initiation would take more than 10' cycles. Bus the high-frequency phenomenon is assc,ciated with low amplitude stress cycles. Thus, the resulting degradation mechanism must be high cycle thermal fatigue.

Inadequacy of the Augmenteo inspection Prograrn. Inadequacy of the inspection program may be considered as a root cause of the MU/HPI lir e weld cracking. He scope of the augmented inspection program implemented by Oconee 2 after the 1982 Crystal River 3 cracking event was not as extensive t.s recommended by the task group (B&W 1983). Examples of this inadequacy are as follows. Periodic ultrasonic examinat on of the safn end included only the nozzle-to safe end weld and not the base metal. He augmented inspection pmgram failed to include the recommended ultrasonic examination of the MU/HPI piping and the pipe-to safe end weld. The inspection program added periodic radiographic testing of the safe end and thermal stone including the contact expansion region, but did tot provide adequate procedural guidance or acceptance enteria. Additions to the augmented inspection program as required by USNRC Bulletin 88 08 were implemented properly but they addressed the HPI lines only and not the MU/HPI lines (Oconee 21997).

3A,6 Corrective Actions for 1997 Oconee 2 Event ne loose thermal sleeve has been removed and replaced with a thermal sleeve having an improved design, shown in Figure 3-4. In midition, the safe end and the cracked portion of the MU/HPI piping have been replaced.

Oconee 2 will be increasing makeup flow to 228 IJmin. (60 gpm),114 Umin. (30 gpm) to each MU/HPI nozzle, to minimize the length of turbulent penetration and alleviate the concerns for thermal cycling as far as the safe-end weld and MU/HPI line region are concemed. The elimination of thermal cycling will be independent of the integrity of the safe-end/ thermal-sleeve joint. If the sleeve becomes loose, the cold makeup coolant (because of its higher flow rate) will flow into the annulus region instead of hot reactor coolant that flows out of the annulus region into the MU/HPI line. However, this may raise a concem of thermal shock to the nozzle knuckle region and the dissimilar-metal weld. In addition, administrative controls have been imposed on the makeup water flow variations. Specifically, these controls include not isolating makeup flow except in emergency conditions, and limiting variations in makeup flow as much as possible (Duke Power 1997a).

Duke Power (1997a) has been continuously assessing the impact of the potential causes of the Oconee 2 failure on the operation of the Oconee 1 and 3 plants. Review of past radiographic examination results concluded that a gap existed between the Oconee 3 therma sleeve and safe end. As a result, Unit 3 was shutdown. He review did not reveal any presence of a gap in Oconee 1 plant because the double thermal sleeve design employed in this plant provides a tighter fit between sleeve and the nozzle.

Duke Power has revised the acceptance criteria for the radiographic examination of the thermal sleeve. The revised criteria are evidence of an intact sleeve and no change in the size of the gap, if existing, between the thermal sleeve and the safe end in the hard roll expanded region (Duke Power 1997f). It may be advisable to repair the sleeve by rerolling if a gap is already present. In a radiograph, a gap is indicated by a dark line at the interface between thermal sleeve and the safe end in the contact expansion region. This line is due to the presence of air or water in the gap, which are less dense than the surrounding metal. When there is no gap, the dark line is not present.

3-21 DRAFT- NUREG/CR-6582

THERMAL FATIGUE CRACKING Duke Power also plans to assess the capability of the field ultrasonic testing to characterize the flaws in the nozzle components by comparing the ceination results with the metallurgical laboratory results (Dake Power 1997e).

Duke Power has revised the augmented inspection plan for the MU/HPI nozzle components which was developed in response to USNRC Generic Letter 85 20. The revision ensures that changes to the ISI Program will not aff ect the periodicity of inspections conducted under the Augmented Inspection Program.

In addition to radiographic examination of the thennal sleeve expansion region and ultrasonic examination of the welds, the revision include ultrasonic examination of safe end base me.al and adjacent piping base metal out to the block valve weld (see Figure 3-8)(Duke Power 1997i). Review of field experience related to metal fatigue had also recommended inspection of base metal locations susceptible to significant fatigue damage (ASME 1992, Shah and Ware 1994). .

Plant Operations has implemented several actions to assure that the potential for RCS leak are carefully monitored at Oconee 1. Any confirmed RCS leakage greater than 3.8 Umin. (1 gpm) will be treated as unisolable leak and the p! ant will be promptly shut down. Oconee units employ four different leak detection methods: (1) reactor building air particulate monitor, which is sensitive to a low leak rate [0.38 Umin. (0.1 gpm)] provided there is no fuel cladding leakage; (2) reactor building normal sump level indicator, which can detect a 3.8 Umin. (1 gpm) leak in 10 minutes; (3) letdown storage tank inventory monitoring; and (4) leakage calculations based on pressurizer water level and letdown storage tank level. All these leak detection methods were effective during the Oconee 2 event. These leak detection methods have a capability for promptly detecting a 3.8 Umin. (1 gpm) leak, as was the case in the Ococee 2 event (Duke Power 1997a).

3.4.7 Safety Consequences The leak started small,9.5 Umin. (2.5 gpm), and grew to 38 Umin. (10 gpm) in 16 hours1.851852e-4 days 0.00444 hours 2.645503e-5 weeks 6.088e-6 months . As the increase in the leak rate was slow, an orderly plant shutdown was carried out without exceeding the capability of the normal makeup system.8 However, the leak occurred in a unisolable portion of the reactor coolant pressure boundary and its rate exceeded the Technical Specifications limit of 38 Umin. (10 gpm) identified leakage.

As the leak rate increased above 38 Umin., a Notice of Unusual Event was declared. The leak was entirely within the containment and no radioactive releases occurred.

Although the leak-before-break concept can be readily applied to the PWR main coolant piphig, its application to smaller branch lines may present difficulties. Wichman and Lee (1990) reported that approximately two -thirds U.S. PWRs had approval for the application of the leak-before-break concept to the reactor coolant piping main loop, but only four plants had such approval for their auxiliary lines with inside diameters greater than 102 mm (4 in.).

A leak before-break concept (USNRC 1984) may not be applicable to the MU!rfPI line with an inside diameter of 54 mm (2.1 in.) for two reasons:

1. For small diameter piping, the flaw size associated with a 38-Umin. (10-gpm) leak is comparatively large, leaving a small undetected ligament. The large flaw size is not likely to provide a margin of safety of 2 8 he increase in the leak rate was also slow during the 1982 Crystal River 3 event (Duke Power 1997a).

DRAFT - NUREG/CR-6582 3-22

THERMAL FATIGUE CRACKING I

as compared to a critical flaw size. This limitation may be reduced if the detectable leak rate is less than 1 gpm (Maxham and Yoon 1987).

2. He MU/IIPI line has been subject to thermal fatigue and flow induced mechanical vibrations. ne leak-before-heak concept is not applicable because the geometry of a flaw introduced by fatigue may not be bounded by the postulated flaws assumed in the leak-before-break analysis.

Even though the leak-before-break methods are inappropriate for the MU/HPI line, an evaluation was performed to determine the margin that exists between a crack exhibiting a 38-Umin. (10-gpm) leak rate and the critical crack length. A crack associated with a 38-Umin leak was 69-mm (2.73-in.) long, whereas the critical crack length for a tungsten inert gas weld (the safe end-to MU/HPI pipe weld) is 113 mm (4.44 in.).

Therefore, the margin was (4.44/2.73 =) 1.63 (Duke Power 1997f).

2 2 This leak potentially could have grown to a 2.5-in. pipe break [0.0023 m (-0.025 ft )] if the leak had not been discovered or actions taken to control it were not effective. This break would have constituted a small-break LOCA. Breaks at this location are bounded by the analysis in the updated FSAR. The analysis concludes that break can be handled without core damage (Oconee 21997).

3.5 ISI of Branch Lines Nuclear power plant Class 1,2, and 3 components in the United States are subject to the requirements of Section XI of the ASME Code ( ASME 1995b) as required by the Federal Regulations [10 CFR 50.55 a(g)(4)).

Section XI contains the minimum ISI requirements for these components. This requirement includes a PSI and fourISI at 10-year interva' . iring the 40-year operating life of a nuclear plant. He specific edition of Section XI required by the Regulations is based on the start of each 10-year inspection interval. In accordance with the Regulations, the examination of components must comply with the latest edition and addenda incorporated by reference in 10 CFR 50.55a(b) on the date 12 months prior to the start of the 10-year inspection intenal.

3.5.1 ASME Section XIinspection Requirements PSI requirements. Preservice examination is performed to establish baseline examinations of welds that may be examined during ISI. The preservice and ISI requirements for Class 1 piping systems are specified in Subsection IWB of Section XI of the ASME Code. Rese requirements include examination of the Class 1 RCS piping welds as specified in Table IWB-2500-1, Examination Categories B-F and Bd.

The examination requirements and methods are identical for both the preservice and ISIS, and only differ by the extent of the examination. For the PSI, the extent of examination extends to 100% of the non-exempt pressure-retaining welds.' In accordance with Section XI, preservice examinations are to be conducted under conditions and with techniques and equipment equivalent to those that are expected to be employed for subsequent inservice examinations. The same preservice examination requirements are also applicable for components repaired or replaced during the service lifetime prior to the components return to sewice. The

Generally, Class I components that are exempt from examination are piping systems nominal pipe size 1 (NPS 1) and smaller.

3-23 DRAFT- NUREG/CR-6582

( TilERMAL FATIGUE CRACKING cxamination requirements are the same for both PSI and ISI, and the details will be discussed in the next section.

ISt requirements. The examination requirements for Class 1 piping welds are generally categorized as Examination Category B-J, " Pressure Retaining Welds in Piping". Both volumetric and/or surface examinations are specified, ne corresponding examination sample includes (a) all terminal ends in pipe runs connected to vessels, (b) all weld locations where se calculated design-basis stress intensity exceeds 2.4 S. (S. is the maximum allowable general primary membrane stress intensity as defined in Section ill of the ASME Code) er where the calculated design-basis cumulative fatigue usage factor exceeds 0 4,(c) all dissimilar metal welds, and (d) an addhional number of welds to bring the examination sample to 25%

of the weld populstion. The weld population does not include the pressure-retaining welds in pipe sizes smaller than NPS 1.

Volumetric examination is performed on the inner one-third of the weld volume and the adjacent base metal, and the surface examination includes inspection of the external surface of the weld and the adjacent base rnetaf.Section XI (1974 Edition / Summer 1976 Addenda) first recognized that the inner weld surface is the most likely site for inrnation of a service-induced flaw; thus, examination of the full weld volume is not required. For pressure retaining welds in piping smaller than NPS 4 and socket welds, only a surface examination of the exterior surface of the weld is required. No examination is required for pipirg smaller than NPS 1.

3.5.2 Inspection of Austenitic Stainless Steel Austenitic stainless steels can solidify in much larger grains (e.g., >lmm) depending on processing histon/,

making inspection far more difficult than for ferritic steels (Harker et al.1990). Inspectability can vary from difficult in the weld metal to good in the r;arotmding heat affected zor and base material, depending on the type of parent material (wrought, forged), and e e welding process ' ed, all of which influence the size and shape of the grain structure. Dendritic grains are typical withia the weld, with coarse equiaxed grains directly adjacent to the weld that become finer with increasing distance away from the weld. These regions can possess different acoustic properties that can create acoustic interfaces, cause beam redirection, false calls, and a general loss of confidence and reliability. With grain sizes larger or equal to the commonly used wavelengths of 1 to 3 mm, each grain can influence the sound beam, causing scattering and attenuation of ultrasonic energy, and creating high levels of background noise.

3.5.3 Past Experience / Augmented Examinations .

The ASME Section XI Code was written to address generic inservice degradation and intended to provide the minimum ISI requirements. As such,Section XI examinations are only performed on a portion of cntical components for degradation mechanisms postulated in the design basis (i.e., primarily thermal fatigue).

Thus, the Code does not address every location or type of denadation that cauld occur. in these cases, the USNRC may impose augmented examination requirements to supplement Code requirements. In accordance with the Code of Federal Regulations [10 CFR 50.55a(g)(6)(ii)), the USNRC may imposed augmented inspection programs for systems and components where added assurance of the structural integrity is deemed

' For volumetric and surface examination purposes, the adjxent base metal extends % of the nominal wall thickness

~

from the toe of the weld.

DRAFT - NUREG/CR-6582 3-24

THERMAL FATIGUE CRACKING necessary. As will be discussed below, this can apply to inspection methodologies, r.s well as systems and locations susceptible to certain types of degradation.

Erarn/ nation for Thennalfatigue Cracking. Fatigue occurring at several operating plants prompted the USNRC to issue several documents to inform licensees and in some cases, to request certain estions to address this aging mechanism. Documents addressing the thermal fatigue issue are USNRC Information Notices 82-09,88-01, and 97-46,Bulletin 88-08 and its three supplements, and Generic Letter 85-20. In some cases, enhanced inspection procedures were recommended.

As a result of the 1982 fttigue cracking at Crystal River 3, the B&W Owner s Group Safe End Task Force performed generic investigation and recommended an augmented, periodic ISI program for MU/HPI nozzles.

As discussed earlier, the augmented program included radiographic and ultrasonic examinations. USNRC Generic Letter 85-20 endorsed the Task Force recommendations. All B&W-designed plants implemented the augmented inspection program.

As a result of the fatigue failure at Farley 2, the use of special techniques, higher instrument gain, and examination personnel with demonstrated ability to detect and evaluate cracks in stainless steel welds were recommended by.the USNRC in Bulletin 88-08, Supplement 2 (USNRC 1988d). In this case, conventional amplitude-based ultrasonic examination procedures specified in AShE Section XI could not reliably detect or accurately size tight thermal fatigue cracks. In December 1987, Farley Unit 2 had to shut down when a leak was discovered in a weld in the stainless steel safety injection line. The throughwall crack was determined to be caused by thermal fatigue. No reportable indications were found when the weld was examined as part of the normal !31 program during a 1986 outage. Ultrasonic testing with a 45-degree transducer and a gain of 6 dB above the reference level, as required by AShE Section XI, was used in this examination. The weld was reexamined using the same procedure after the leakage was discovered, but still no reportable indications were found. Supplemental techniques using a 60-degree shear wave transducer, and an additional 8 dB of gain with the 45-degree transducer were needed to identify the crack. In a similar situation, an additional 24 dB (16 times) above the ASME Code sensitivity was needed to detect a through-wall crack at Tihange 1 in Belgium. In the latter case, the crack was discovered in the base metal of an elbow, which is not included in the AShE Section XI required ISIS, and was only detected because of the leakage.

In response to USNRC Bulletin 88-08 and its three supplements, PWR utilities in the United States have perfonned nondestructive examinations of the critical sites in the charging and safety injection lines and nozzles (Strauch et al.1990). The critical sites include the welds and the base metal sites with high stresses.

No crack like defects have been found. On-line temperature and pressure monitoring have also been implemented to detect and estimate thermal stratification loads. These measurements indicate that some thermal stratification and cycling does take place; however, the magnitude is small and no significant actions are required. Some experimental flow tests have also been carried out to assess the effect of valve leakage, and some utilities have added or removed valves in the various piping systems to reduce the potential for valve leakage problems.

Inspection of Oconee 2 MU/HPINozzle and Piping. Because of the 1982 Crystal River 3 MU/HPI nozzle weld cracking event, ultrasonic and radiographic examinations were performed at that time to detect flaws in the safe end, nozzle, and adjacent piping. No flaws or loose sleeve were detected in the 2Al nozzle amponents. A 35- to 45-degree shear wave was found sufficient to detect axial cracks in the Crystal River 3-25 DRAFT-NUREG/CR-6582

THERMAL FATIGUE CRACKING 3 safe end and adjacent piping and was used for the Oconee 2 inspections. At the time ofinspection, ASME Code Section XI had minimal qualification requirements for an inspection technique.

After the 1982 inspection, an augmented inspection program was implemented. The main elements of the augmented inspection program are radiographic examinations to ensure that the thermal sleeve is in the proper p eition and no gap exists between the thermal sleeve and safe end, and ultrasonic examinations to ensure ra cracking exists in the safe end and adjacent piping. Evaluation of the recent cracking of Oconec 2 MU/HPl nozzle safe end weld reveal the inadequacies of the augmented inspection program. The radiographic examinations did not evaluate the tightness of the thermal sleeve and the ultrasonic examination did not include the safe end-to-MU/HPI line region. As discussed earlier, these inadequacies may be considered as a root cause of the 1997 Oconee 2 cracking event.

The 2A1 MU/HPI pipir.g and 3 A 1 safe end, both containing fatigue cracks, were removed and then inspected with altra:onic examination. The ultrasonic techniques used to perform the 3Al safe end were qualified according to ASME Section XI Appendix VIII (1992 Edition with 1993 Addenda): Performance Demonstrationsfor Ultrasonic Examination Systems. The specifics of the qualified ultrasonic inspection technique are not available. The ultrasonic inspection results were compared with the metallographic examination res.ults. The comparison showed that the safe end flaws with a depth in the range of 2.8 mm (0.110 in.) to 3.8 mm (0.15 in.) can be detected; however, some flaws in this range were missed. Increased reliability in the detection capabilities are needed. Cracks shallower than 2.8 mm were not detected. The indications were in the safe-end base metal and were parallel to the safe-end axis. The accuracy of the location of the detected indications was 2.5 mm ( 0.1 in.) along the safe-end axis and *10 mm (t0.4 in.)

in the circumferential axis (Duke Power 1997f). Similar comparisons showed that the cracks in the piping with depths equal to greater than 1.27 mm (0.05 in.) can be detected.

Ultrasonic examination of small diameter piping such as the MU/HPI line to detect axial cracks is difficult because a loss of contact between the conventional flat search unit and the piping surface may take place if the search unit is not held in correct alignment. Instead, a contoured search unit fits snug on the outside surface of the piping, provides a better alignment, and can be used for detecting purely axial cracks, but it can not be oscWated to detect off-axis cracks (Duke Power 1997f).

Duke Power (1997f) does not provide information about the sizing accuracy of the inspection methods used.

Shah et al. (1997) have identified several different ultrasonic examination methods for reliable characteriza-tion of thermal fatigue cracks in piping and vessels. Tip diffraction techniques are used for reliable sizing of thermal fatigue and stress corrosion cracks. Creeping wave techniques have been used to detect very shallow defects at the inside or outside surfaces. -

Use of reliable ISI techniques is cffective in detecting fatigue cracks in a timely manner provided the crack growth rate is low, that is, a small undetectable crack takes longer than an operating cycle to become a -

throughwall crack. This was the case for the 1982 Crystal River 3 and the 1997 Oconee 2 cracking of MU/hPI safe-en.d weld. However, effectiveness of the use of ISI techniques is reduced if a small nondetectable defect become a throughwall crack within an operating cycle. It, appears that this might be the case for the recent leak exent at Dampierre 1, which is reported in Table 3- 1. Electricite de France (EDF) is developing a strategy for dealing with this type of cracking (INES 1997).

DRAFT - NUREG/CR-6582 3-26

THERMAL FATIGlJE CRACKING 3.6 USNRC Bulletin 88-08 for Unisolable Sections of Piping Connected to the RCS De cracking and leakage events that led to the issuing of USNRC Bulletin 88-08 are described first. Then the actions requested by the bulletin and its three supplements are described. Then a few examples of the utilities' responses to the bulletin are presented.

3.6.1 Cracking in Safety injection Line at Farley 2 and Tihange 1 In 1987, a leak occurred inside the containment of Farley 2 during normal power operation. ne leak was found in the unisolable location ofsafety injection line as shown in Figure 3-17. The crack was on the inside

. surface of the weld and extended approximately 120 degrees circumferentially around the underside of the pipe. About 25 mm (1 in) of this crack was throughwall. The crack was caused by thermal fatigue and had developed slowly. The leak rate was 2.7 Umin (0.7 gpm). The circumferential temperature difference in

, the region of stratification downstream of the check valve was as high as 120*C (215'F). The suspected phenomenon causing thermal fatigue is thermal cycling resulting from the interaction of hot water from tvbulent penetration with relatively cold water leaking through the globe valve. A similar occurrence caused cracking in both base metal and welds of Tihange 1 as shown in Figure 3-18. Both Farley 2 and Tihange 1 are similarly designed Westinghouse-type 3 loop plants.

The affected PWR plants have dual purpose pumps used for both adding coolant during normal reactor operation and i@cting emergency core coolant at high pressure following an accident. A schematic diagram of the high ;wssure safety injection and RHR systems in a three-loop Westinghouse plant with dual-purpose pumps is shown in Figure 3-19 (Su 1990). He charging pumps supply the coolant to the safety injection system. The safety injection and RHR systems are connected to the RCS through the same nozzles, located one on each of the three cold legs. The charging system (not shown in Figure 3-19) is connected to the cold leg through a different nozzle. Accumulator tanks, also not shown in Figure 3-19, serve a function similar to that of the safety injection tanks but are connected to the primary system cold legs through separate nozzles. All valves in the safety injection system are closed during normal operation. Globe valve A, which has leaked in Farley 2, is in a bypass line around the boron injection tank. Leakage of cold water through a closed globe valve continued down the line through the check valve and produced flow stratification (USNRC 1988a,b).

Corrective actions included replacement of the elbow and the affected pipe spool and installation of an additional valve downstream from the check valve to prevent stratified flow.

3.6.2 Cracking in RHR Line at a Foreign Plant A 203 mm (about 8 in. NPS) Schedule 140, Type 316 stainless steel RHR line was found leaking in an overseas plant (Genkai l in Japan). Figure 3-20 shows an isometric view of the RHR piping and includes the location of the throughwall crack (USNRC 1989). He unisolable leak was found in the RHR line wild joint between an elbow and a horizontal pipe section located between the hot leg and the first isolation valve.

The crack extended 97 mm (3.8 in) circumferentially around the pipe on the inside surface of the weld.

About 1.5 mm (0.06 in) of this crack extended completely through the wall. He crack initiated on the inside surface at the weld metal base metal interface and progressed through the weld toward the outside of the pipe wall. The leak rate was 0.8 Umin (0.2 gpm).

3-27 DRAFT-NUREG/CR-6582

TIIERMAL FATIGUE CRACKING The crack was caused by high-cycle fatigue that resulted from cyclic thermal stratification in the horizontal i

pipe section. He stratified flow condition was caused by a small packing leak from the first isolation valve.

The cyclic stresses may have been from intermittent valve leakage and turbulent penetration, which eventually caused the throughwall crack. A resistance temperature detector (RTD) previously installed on the valve gland leak-offline did not detect any temperature change because it was cooled by the ventilation system air flow. The effect of thermal cycling from turbulent penetration or valve leakage on piping was not considered in the design analysis.

The conective actions included:

1. The valve packing was replaced to stop the gland leak and the valve seat gap was increased to eliminate possible cyclic opening / closing of the gate. _

2. He elbow and horizontal pipe section were replaced.

3. He RTD on the leak-off line was insulated to correct the ventilation cooling problem and temperature instrumentation was installed on the outside of the RHR line to monitor the top-to-bottom temperature differentials in the pipe.

3.6.3 USNRC Bulletin 88-08 Based on the safety injection line cracking in a domestic plant, the USNRC (1988a) issued Bulletin 88-08,

" Thermal Stresses in Piping Connected to heactor Coolant Systems," in December 1988. The Bulletin requested the following three actions from licensees of operating PWRs:

1. Review and identify systems where unisolable sections of piping connected to the RCS may be subjected to thermal stratification or temperature oscillations that could be induced by leaking valves and that were not evaluated in the design analysis of the piping;

2. For susceptible locations, perform nondestructive examination of weld, heat affected zones, and high-stress locations to assure that there are no existing flaws; and

3. Develop and implement a program to provide continuing assurance that unisolable sections of piping connected to the RCS will not be subjected to combined static and cyclic stresses that could cause fatigue failure.

The assurance may be provided by:

a. redesigning and modifying these sections;

b. instrumenting the piping to monitor for adverse temperature distributions and establishing appropriate limits on these distributions; and

c. providing means to ensure that pressure upstream of block valves which might leak is monitored and does not exceed RC3 pressure.

DRAFT - NUREG/CR-6582 3 28

THERMAL FATIGUE CRACKING Actions 2 and 3 apply only to systems that were identified in Action 1 as being susceptible to thermal stratification or temperature oscillations. Various time limits were included for the actions in the Bu!!ctin, depending on whether the plant was in an extended outage.

Bulletin 88-08 was followed by three supplements: June 24,1988 (USNRC 1988c), August 4,1988 (USNRC 1988d), and April 11,1989 (USNRC 1989). Supplement 1 provided preliminary information on similar cracking in a European Westinghouse type 3-loop plant (Tihange 1) and noted that examinations of high stressed metal would include the base metah but the reporting requirements of the Bulletin were not changed. Supplement 2 emphasized the need for enhanced ultrasonic testing to detect fatigue cracks in stainless steel piping; but the reporting requirements remained unchanged. Supplement 3 discussed the cracking event in the foreign plant RHR line and noted that periodic valve seat leakage through packing glands could result in unacceptable thermal stresses; but required no additional actions.

USNRC (1991) provided following guidelines for identification of potentially susceptible piping. For sections ofinjection piping systems normally containing stagnant coolant, the susceptible piping, regardless ofits pipe size, has the following characteristics: (a) the operating pressure is higher than the RCS pressure; (b) the piping sections contain long horizontal runs; (c) the piping systems are isola'ed by one or more check valves and a closed isolation valve in series; (d) water injection into the RCS is top or side entry; and (e) the first upstream check valve is located less than 25 pipe diameters from the RCS nozzle. Examples of such system in PWR ate the safety injection and charging (makeup) lines between the reactor coolant loop and the first upstream check valve. Similar guidelines were presented for othe piping systems, such as RHR system, which normally contain stagnant coolant.

USNRC (1991) also identified acceptable responses to Action 3 identified in Bulletin 88-08. The acceptable responses include revision of system operating conditions, relocation of the first upstream check valve, and installation of temperature and pressure monitoring instrumentation for detection of valve leakage imposing thermal cycling loads on piping. The guidelines for temperature monitoring included type and location of temperature sensors, determination of baseline temperature histories, monitoring time intervals, and acceptable top-to-bottom temperature difference. Mitigative actions such as reducing valve leakage were required if the acceptable temperatures were exceeded.

3.6.4 Responses to USNRC Bulletin 88-08 Licensees took differing approaches to the Bulletin requests. Some simply reviewed the branch lines according to Action 1, concluded that no stratification from valve leakage was possible, and left the matter there. At several other plants, potentially susceptible lines have been inspected (Strauch et al.,1990), but no crack-like defects have been found. For example, on a Westinghouse two-unit facility, ten pipe lines including six 152-mm (6-in.) safety injection lines, one 102-mm (4-in.) auxiliary spray line, and three 51 mm (2-in.) reactor coolant fill lines were identified as having the potential for stratified flow, were instrumented, but no temperature oscillations that would cause a fatigue concern were identified (Bain et al.1993).

Evidence of turbulent penetration thermal cycling was found in the shutdown cooling system of an older vintage Combustion Engineering plant. A fatigue analysis was conducted to demonstrate that, even with the additional fatigue usage contributed by the turbulent penetration thermal cycling transients, the cumulative usage factor would remain less than the ASME Code limit of 1.0 for the 40-year life. Actions taken by another licensee included installing redundant manual isolation valves on small lines [76-mm (3-in.) and smaller diameter lines], verifying valve integrity through an enhanced maintenance and leak testing program, and a temporary temperature meast4 ring progrum on the RHR line.

3-29 DRAFT- NUREG/CR-6582

THERMAL FATIGUE CRACKING With regard to Action 3 of the bulletin, several utilities have based their response on an analytical methodology developed under a program sponsored by the Electric Power Research Institute (EPRI) to investigate thermal stratification, cycling, and striping (TASCS). His program is called the EPRI TASCS Program. For example, the methodology was used to evaluate the temperature monitoring results for South Texas 1 and 2 normal charging, attemate charging, and auxiliary spray piping. The conclusion of the evaluation was that the piping integrity would not be jeopardized, should inleakage of cold water through a valve into the RCS occur over the life of tb units. Subsequently Houston Lighting & Power (HL&P) informed the USNRC that temperature monitoring of these lines will be discontinued (HL&P 1996). The USNRC staff reviewed the HL&P response and concluded that the decision to remove the temperature monitoring instrumentation may have been premature.

The main objection of the staffis that the mechanism of turbulent penetration has not been fully investigated under the TASCS program and its significance in the failures at Farley 2 and Tihange I has not been clearly established. The TASCS methcoology does not predict the locations of these failures. He TASCS methodology predicts higher cyclic stresses at the end of the turbulent penetrr. tion column where thermal cycling may take place if valve inleakage is present, and lower stresses at locations within the turbulent penetration column where temperature differences approach zero. However, when the TASCS methodology is applied to the Farley 2 safety injection line cracking failure, the throughwall crack location is within the calculated length of the turbulent penetration column and not at the end ot'the column where the cyclic stresses are higher and a fatigue failure is more likely to occur. He USNRC recommended that HL&P reestablish the previous monitoring programs at both units (USNRC 1996,1997b).

Additions to the augmented inspection program at Oconee 2 were made as required by Bulletin 88-08. Even though these additions were implemented properly, they addressed the HPI lines only and not the MU/HPI lines, which did not contain stagnant coolant during normal operation.

3.7 Fatigue Monitoring of Branch Lines in response to the directives in Bulletin 88-08, several PWR RCS branch lines have been monitored to determine the thennal and fluid conditions existing in these lines during operation. He monitoring was used to detect unusual conditions such as thermal stratification in the branch lines and identify the plant operation under which the unusual conditions were present. The monitoring results also helped to identify whether valve leakage is present. Only by experimental measurements of these abnormal conditions can a reasonable estimation of the severity of the thermal loads and the numbers of cycles be developed to assure a representative fatigue analysis can be conducted. Examples of fatigue monitoring on plants designed by each of the three NSSS vendors are described below. The examples are intended to show several of the piping layouts, locations, and modes of plant operation in which thermal stratification conditions can occur.

3.7.1 Westinghouse-designed PWRs Thennocouples were installed on 38 mm (1-1/2 in.) safety injection lines downstream of the dual-purpose centrifugal charging pumps on all four loops of a PWR plant. The instrumentation for one of these loops is shown in Figure 3 21. Two check valves separate the Class I system, which is connected to the RCS piping, from the Class 2 system, which is connected to the upstream check valve. He piping between the valves is Class 1. The Class 2 piping is normally at the ambient temperature in the containment [49 C (120'F)], whereas during normal plant operation the Class 1 piping is at 292*C (558'F).

DRAFT - NUREG/CR-6582 3-30 1

1

! THERMAL FATIGUE CRACKING During nonnal operation, the reactor coolant pressure in the Class 1 piping is 2300 psi, which is higher than the pressure in the Class 2 piping; the pressure difference keeps the check valves shut. If the check valves leak, the higher pressure in the Class I portion can force hot Guid through the leaking valves to the colder Class 2 portion and produce a stratified fluid condition. This is called out-leakage.

He thermocouples upstream and downstream f the check valves were monitored to detect variations from normal readings that would indicate check valve leakage (Structural Integrity Associates 1996). For example, if the thermocouple pair SI-8T upstream of the check valve indicated a rise from the normal 49'C (120*F) temperature, out leakage from the Class I to the Class 2 portions of the piping would be indicated.

Out-leakage from the Class 1 piping to the Class 2 piping may take place during heatup where the coolant in the RCS expands and raises the pressure. Should valve leakage occur, thermal stratification, described earlier and shown in Figure 3-3, can take place. The difference in readings of thermocouples SI-STA and SI 8TB gives the thermal stratification temperature difference.

Several subtransients resulting from pump cycling (off and-on operation, that is, stopping and restarting the pumps) were identified during major transients such as heatups and cooldowns. From the measurements, it was conservatively assumed for analysis purposes that flow stratification could occur in the horizontal sections of the portion of the Class 2 piping near the check valves, but that minimal or no stratification would occur in the other part of the Class 2 piping. The maximum top-to bottom temperature difference in the Class 2 was estimated to be 95'C (171*F). The temperature and pressure measurements were used for estimating the cyclic stresses and to demonstrate that fatigue requirements were met.

3.7.2 B&W-designed PWRs Monitoring was performed on 63.5 mm (2-1/2 in.) HPI lines (BWNT 1992). Data were also collected on the pressurizer spray line and the core flood lines. Thermocouples were placed at five locations on the HPI line near a check valve as shown on Figure 3-22. The check valve is located at UD = 6 fmm the reactor coolant loop branch nozzle. Stratified flow was detected with the largect top-t >-bottom temperature difference of 182*C (327*F) at location 2.

With makeup flow off, large temperature swings were measured on the upstream side of the check valve (locations 3 and 4 in Figure 3 22), which is consistent with backflow from a leaking check valve. During the time makeup flow was turned off, the 1 A 1 pump was initially shut off and thea it was operated for a brief time. When the l A1 pump was shut off, the temperatures at locations 3 and 4 were about 260*C (500*F),

close to the reactor coolant temperature. When the pump was tumed on, the temperatures decreased to 54*C (130* F), close to the bypass temperature. As soon as the pump was shut off again, the temperatures at all locations increased to about 260*C (500*F), indicating leakage had resumed. He possible reason for the leakage is higher reactor coolant static pressure, while the pump 1 A1 was tumed off.

3.7.3 Combustion Engineering-designed PWRs In a 1993 plant cooldown, thermal stratification was detected in a 356-mm (14-in.) Schedule 140 shutdown cooling suction line in the first horizontal piping run away from the RCS hot leg (Figure 3-23) in an older vintage Combustion Engineering-designed PWR (EPRI 1993c). Thermal stratification has been described in an earlier section of this report. He stratification was detected during 4-pump operation, but not 3-pump operation. Stratification was measured in the upper horizontal run of piping (location 1) but not in the lower 3-31 DRAFT -NUREG!CR-6582

l TIIERMAL FATIGUE CRACKING portion (location 2); thus this short length of pipe was subjected to local, rather than global, stratification effects (Lubin et al.1994).

The top-to-bottom temperature difference (see Figure 3 3) was estimated to be 190'C (340'F). The root cause is believed to result from turbulent penetration of the hot RCS fluid in conjunction with a piping configuratiot that allows this phenomenon. Figure 3 2 shows a schematic of the type of therinal stratification postulated for this case. It has been hypothesized that the increase in the hot leg density (that accompanies a reduction in reactor power and a decrease in the hot leg temperature) alters the flow chas acteristics, which in turn increases the mass transport of header fluid into the branch line. 'Ihis increases the turbulent penetration distance and causes stratified flow to develop in the horizontal section. Turbulent penetration has been described in an earlier section. Valve leakage was not found to be a contributor.

~

Top-to-bottom temperatures greater than 28'C (50'F) were also measured in safety injection and power operated relicflines in a Combustion Engineering plant (Lubin and Brown 1995). A series of thermocouples was placed at axial and circumferential locations on a safety injection line shown in Figure 3-24. The check '

valve is located at an UD of about 15 from the main coolant piping branch nozzle, and there is a run of pipe with an elbow between the check valve and the motor-operated valve (MOV). Wall temperatures were recorded during heatup, normal operation, and cooldown. In general, wall temperatures between the RCS and the check valve were close to the RCS temperature and there was little angular variation of temperature.

However, between the check valve and the MOV, the top-to-bottom temperature variations were as much as 78'C (140'F) at locations 5 and 6. The temperature differences were credited to a natural convection between a source of high '.emperature at the check valve, and cold temperatures close to the MOV [see Figure 3-2.1). Luhii. w a one-dimensional model to analytically predict fluid and wall temperatures as functions of end conditions and line geometry that showed reasonable agreement with the measurements.

DRAFT - NUREG/CR-6582 3-32

4. ANALYSIS OF LEAK EVENTS ASSOCIATED WITH VIBRATORY FATIGUE Vibratory fatigue failures of small-diameter LWR piping have occurred worldwide. The failures have occurred mainly at socket welds in small-diameter pipe lines. The failures have resulted from high-cycle mechanical fatigue accompanied by low-ca-plitude cyclic stress. Review of the field experience indicate that p)ing vibrations have been a significant source of problems in nuclear power plants. The majority of pipe failures have been experienced in small piping.

The socket weld design and ASME Code requirements are described first. Then the excitation mechanisms for piping vibrations and the vibration fatigue damage mechankm are described. Rese descriptions are expected to facilitate the discussion on the trerxis in th' vibratory fatigue failures and the analyses of the

'wo selected leak events that follow, ASME Code Section XI ISI requirements, USNRC requirements, and mitigation and monitoring of vibratory fatigue are presented next. Finally, industry efforts to manage vibratory fatigue are summarized.

4.1 Socket Weld Design The socket weld joint, shown in Figure 4-1, is used for small piping (less than 2-in. diameter) because it makes the connection of the two ends of the flexible piping simple and cuy for construction. The end of a small piping is fitted into the socket and a circumferential fillet weld is used to complete the joint as shown in Figure 4-1(a). Rese fillet welds vary from convex or concave as shown in Figure 4-1(b). The concave design has a lower SCF. The theoretical throat presents the area of minimum resistance to shear forces and is used to calculate the shear stresses in the design analysis.

ASME Code Section /// Requirements. ASME Code Section III (ASME 1995a), NB-3661.2 limits the size of socket welds to pipe sizes of 2 in, or less. Socket-welded piping joints are required to conform to the requirements specified in ANSI B16.11, which specifies dimensions for standard designs which have been qualified by pressure tests. Figure NB-4427-1 of the Code (Figure 4-1), shows a socket weld design.

Figure 4-1(a) shows the minimum welding dimensions for socket welding fittings as required by the Code.

The Code also requires a gap of app;oximately 1.60 mm (0.0625 in.) be provided between the end of the pipe and the bottom of the socket before welding, as shown in Figure 4-1(a). The gap allows for the shrinkage of 'he fillet weld pulling the pipe into the coupling 8High residual stresses may be present in the weld if the welding procedure does not allow for this gap. Such high stresses might cause cracking of the weld and also make it susceptible to vibratory fatigue damage. The gap need not be present or verified after welding. However, utilities prefer to have a gap after welding because it implies a presence of a gap before welding and perform radiography of the socket weld to verify the presence of the gap.' The Code further states that socket welds shall not be used where the existence of crevices could accelerate corrosion.

Section III of the Code provides stress indices for Class 1 socket welds, and stress intensification factors for Class 2 and 3 welds. The peak stress indices for internal pressure (Ki ) and thermal loading (K2 ) are 3.0, and for moment loading (K 3) 2.0 [see Table NB-3681(a)1 of Section III). The stress analyses include B

V. N. Shah, Private conversation with Dieter Timm, G. A. L Gage Co., Stevensville, MI, July 10,1997.

V. N. Shah, Private conversation with Greg Frederick, EPRI NDE Center, July 10, 1997.

4-1 DRAFT - NUREG/CR-6582

VIBRATORY FATIGUE pressure, moment, and seismic loads, but not vibratory loads. NB-3622.3 of Section III states that piping shall be arranged and supported so that vibration will be minimbed, but gives no direct guidelines for accomplishing this.Section III relies on ASME Operation & Maintenance Code OM 3 to verify the adequacy of socket welds for vibratory loads by stating that the derigner shall be responsible, by design and by observation under startup or initial service conditions, for ensuring that sibrations of piping systems are within acceptable levels.

Section III, NB 5261 requires that socket weMs be examined by magnetic particle or liquid penetrant methods. Volumaric examinations such as ultrasonic and radiographic testing are not suitable for socket welds; the small size and complicated geometry make them difficult to apply and interpretations of the results are unreliable. Thus only the exteriors of socket welds are examined.

4.2 ASME Code Requirements for Vibration Tests of Piping Systems ASME (1994) Code OM-3 states the requirements for preoperational and initial startup vibration tests of .

piping systems, which is intended to identify and correct problems in piping vibrations before the plant begins commercial operation. There are re;quirements for both steady-state and transient vibration monitoring. Some systems are accessible for visual inspection, while local monitoring systems that transmit dau. to remote collecting and analyzing stations must be used for inaccessible (for example, because of radiation or temperature) locations. Acceptance criteria are ASME Code stress limits or limiting deflections.

4,3 Excitation Mechanisms The main excitation mechanisms for severe piping vibrations in nuclear power plants are pump induced pressure pulsations, cavitation and flashing. Pump-induced pressure pulsations occur at distinct frequencies, which are multiples of the ptunp speed. Pressure pulsations originate at the pump and are transmitted throughout the fluid the same way that sound is transmitted through air. In the case of positive displacement pumps, pulsations may be induced in both the suction and the discharge piping. When pressure pulsations coincide with a structural frequency of the piping system, severe vibratory fatigue damage may take place.

Cavitation occurs when the fluid pressure approaches its vapor pressure. Vapor pockets form and collapse in the low pressure region with extreme rapidity, generating intense shock waves. For example, it is estimated that two million cavities can collapse within one second over a small area (Sednks 1979). The .

resulting broadband (a band with a wide range of frequencies) pressure pulsation can cause severe vibration of piping downstream of a cavitating component such as an orifice in the PWR letdown system. For

example, vibration measurements in the vicinity of valves and orifices revealed high frequency, broadband .

vibration of hundreds of g's acceleration, with velocity at some points as high as hundreds of inches per second (Kelly and Davis 1995). The frequency content of the excitation is one of the main differences between the cavitation- and pump-induced pressure pulsation. In addition to severe vibrations, the collapse of the cavities on a solid surface removes material by mechanical erosion, damaging piping and other components.

Flathing occurs when the temperature of water is higher than its saturation temperature at a given pressure and the water flashes into steam. This also results in broadband pressure pulsations causing vibration of i

DRAIT - NUREG/CR-6582 4-2 l

VIBRATORY FATIGUE piping downstre.am of the flashing component. Collapse of steam bubbles may also cause water hammer which may synergistically interact with vibratory fatigue in causing failure.

These excitation mechanisms generate high frequency vibratory loads which were not considered in the design analysis because they cannot be predicted accurately. They can only be quantified during plant operation.

4,4 Vibratory Fatigue Damage Mechanism Three characteristics of welded joints contribute to the vibratory fatigue failures: weld geometry, weld discontinuities, and residual stresses. Geometric discontinuities caused by a localized change in a section intensify the stresses in a very local area. The magnitude of the stresses decay rapidly to nominal stress values away from the discorninuity. The magnitude of the stress concentration depends on the geometry of the discontinuity; a is tughest at the toe 9 the weld, if no crack like discontinuities or weld imperfections

, are present.

Fatigue cracks in a wendment initiate where the localized stress range is maximum. This location may not correspond to a locatxm wuh a geometric discontinuity where the SCF is maximum. This is so because, in addition to geometrx discontinuity, various imperfections and crack-like discontinuities present in weld metal or the heat affectal zone also act as stress raisers and drastically reduce weld-joint fatigue strength (SCFs as high as 15 have been estimated at the root of poor welds). Therefore, fatigue cracks may initiate either at surface disconunuities such as at weld toe, at embedded discontinuities such as inclusions, or at weld root.

j Fatigue cracking in a properly fabricated socket weld is usually associated with the weld toe in the case of a joint stressed in the transverse direction. In general, the crack initiates at the toe of the weld, propagates first through the weld metal, then through the heat affected rone, and finally through parent material as shown in Figure 4-3. If the weld is incorrectly proportioned, either through bad design or through faulty fabrication, the stress across the weld throat may be sufficient to initiate a crack at the weld root, Usually, such a crack propagates through the weld metal and breaks the surface near the center of the weld face as shown in Figure 4-4. Vibratory fatigue failures at socket welds have occurred predominately at weld roats. The presence of a discontinuity such as a lack of penetration at the weld root, degrades the fatigue strength greatly. However, it should be noted that the mere existence of a discontinuity does not make a weldment defective or unsuitable for a given application. Discontinuities are designated as defects only when their size, orientation, and distribution exceed specification limits and their presence affects the integrity of a component.

Residual stresses are introduced in a weldment because of the inability of the deposited molten weld metal to shrmk freely as it cools and solidifies. The magnitude of the residual stresses depends on several factors such as the size of the deposited weld beads, weld sequence, total volume of deposited weld metal, weld geometry, strength of deposited weld metal and of the adjoining base metal, and cooling rate.

Mean stresses, which include residual stresses, play a significant role in crack initiation caused by high-cycle fatig te because the amplitude of the alternating stress is below the yield stress. As a result, the presence of a mean stress lowers the fatigue life. In contrast, mean stresses do not play a significant role in crack initiation by low-cycle fatigue where the amplitude of the alternating stresses is greater than the 4-3 DRAFT - NUREG!CR-6582

j VIBRATORY FATIGUE yield strength, so that the material shakes down to clastic action and the mean stress becomes zero. For high-cycle fatigue, the greater the mean stress, the greater the effect on lowering the fatigue life. ASME Section IH has accounted for this by conservatively decreasing the fatigue design curves in the high-cycle region (at alternating stresses below the yield stress) to include the maximum effect of II ean stress. The Code conservatively assumes that the alternating stress for a given life decreases linearly with mean stress.

As the crack propagates, the residual stress will redistribute. The magnitude and orientation of the residual stresses play a part in determining the direction of crack propagation and its rate The mean stress will change if the residual stresses relax d tring cyclic loading.

4.5 Trends in the Vibratory Fatigue Failures Vibratory fatigue fallares are widespread in the nuclear power plants. First the world wide data for .

vibratory fatigue cracking are presented. Then the failure data for the U.S. power plants are presented.

Then two leak events at U.S. PWRs caused by vibratory fatigue failure are analyzed. These analyses include root cause analysis, corrective actions, ?.nd safety significance of the events. .

4.5.1 World-wide Data for Vibratory Fatigue Cracking The SKI has compiled a world wide database on pipe failures over the period from 1972 to 1995 (Nyman 1996). Vibration-induced fatigue has been reported as a failure mechanism for 425 events in small-diameter piping [ diameter smaller than 25 mm (1 in.)], as shown in Figure 4-5. The numbers of failures per reactor year are about the same for PWR and BWR plants; the PWR reactor years are about 2 % times the BWR years, and the PWR failures are about 2 % times the BWR failures, it appears that since about 1990 the failure rate has been reduced.

Figure 4-6 shows the world-wide data for leak and rupture events in small-diamuer piping (diameter smaller than 25 mm) caused by vibratory fatigue cracking. Early-life failures, that is, failures within the first year of commercial operation, are not included in this figure. The failurer. in the PWR plants are first reported little at about 10,000 h of operation, whereas they were first reported in the BWR plants at 30,000 h of operation. However, the failure rate in BWR plants appears to be higher than that in the PWR plants.

Figure 4 6 may be explained as follows." The figure shows the cumulative hazard function H(t) (with the hazard function expressed as a percent) plotted against time. One interpretation of this plot uses the cumulative distribution function F(t) defined as the probability that a weld fails by time t. It can be shown that F(t) is related to the cumulative hazard functioa by ,

F(t) = 1 - exp(-H(t)/100].

For example, at the tinx when H(t) = 100, F(t) = 1 - exp(-1) = 0.632, so 63.2% of the welds in the popu'ation fail by that time.

10 V. N. Shah, l'zivate communication with B. Lydell, RSA Technologies, June 30,1997.

DRAFT - NUREG/CR-6582 4-4

VIBRATORY PATIGUE Figure 4-6 was constnicted by considering only failed welds, not the ones that did not fall. Dus, the population considered for Figure 4-6 was all welds that failed over the period 1972 to 1995. This leads to two warnings for users of the figuie.

1. The plants that contributed the data were at most about 33 years old. Therefore, we have no infortnation about welds that may fall after 33 years of age. These welds were not considered as part of the population.

2. Figure 4-6 does not provide guidance in deciding whether a particular weld will fall. The figure only implies *1f the weld is in the susceptible population. so that it will eventually fail (in the first 33 years),

here is the probability that it will fall before time t, for any t of interest."

Figure 4-6 may be used, with failure data from the early years, to estimate the size of the susceptible population as follows. Assume that all the vibratory fatigue failures will take place in the first 33 years as shown in Figures 4-5 and 4-6. For PWRs, the figure implies that half of the failures take place in the first 40,000 hours0 days 0 hours 0 weeks 0 months . Therefore, after 40,000 hours0 days 0 hours 0 weeks 0 months we can estimate that the same number of failures will take place in the remaining life. Of course, the actual number of failures is influenced by plant operation and also by sheer randomness.

4.5.2 U.S. Data for Vibratory Fatigue Cracking Kustu and School (1980) reviewed 354 reportable occurrences involving piping and fittings to identify significant problems experienced in nuclear power plant piping along with their primary causes. These occurrences were reported in IIRs submitted during the period June 1,1976 to May 16,1979. Cracking of pipes and fittings was identified as the most recurring problem, constituting 52% of the reportable occuitences. Piping vibrations was determined to be the most significant cause. Of the 184 occurrences, 77.2% (142 occurrences) were throughwall cracks and were discovered by observed leaks. Mechanical vibration was the cause of 22.3% (79 occurrences) of all reportable occurrences involving piping and fittings.

A review of LERs from 1%9 to 1980 identified 84 events of cracking and leaking of small piping [ smaller than 100-mm (4-in.) diameter) subject to vibratory fatigue. These events are summanzed in Table 4-1.

He cracking occurred predommantly in socket welds in 19-mm to 51 mm (3/4-in. to 2-in.) diameter tap lines (for example, vents, drains, pressure taps). The cracking was predominantly near pumps, and failure was attributed to vibrational fatigue.

The September 1983 Institute of Nuclear Power Plant Operations (INPO) Significant Events Report (SER 64-83) stated that from April 1970 to September 1983, there were 234 reported small-diameter safety-related pipe failures caused by vibration-induced fatigue (Olsen 1995).

A review of the Nuclear Power Experience database has identified several different typs of vibratory fatigue problems observed in LWRs during 1974 to 1994 (Stoller Corporation). These problems are listed in Table 4 2. Note that most of the failures were at socket welds. But a few failures were in a pipe, threaded joint, and at other types of welds, 4-5 DRAFT - NUREG/CR-6532

VIBRATORY FATIGUE U.S. PWR Data for 1 85 to 1996 Period. A review of LERs from 1985 to 1996, performed as part of this study, reported 28 leak events caused by vibratory fatigue failures of socket welds; 15 in pipes with a diameter less than 25 mm (1 in.) and 13 in pipes with a diameter in the range of 25 to 102 mm (1 to 4 in.).

He leak rates van from negligible to about 151 IJmin (40 gpm). A distribution of the reportable leak events and their frequencies by calendar year is shown in Figure 4-7. A frequency for a given year is a ratio of the number of leak events and operating years for the U.S. PWR plants during that year.

De distribution of leak events by plant age rather than calendar time is shown in Figure 4-8. Here, age is defined as the time from initial criticality. A plant is said to have age n if it i: in its n' year of operatien.

A cumulative distribution of these events is shown in Figure 4-9. The data indicate vibratory fatigue failures generally occur in the first 18 years of component life; only two failures have been reported in plants older than 18 years. ,

A distribution of the frequencies of the leak events by age is also shown in Figure 4-8. A frequency for a given age is a utio of the number ofleak events that occurred in the PWR plants passing through that age ,

during the 1985 to 1996 period and the number of plants associated with that age. The number of plants that pass through a given age is equal to or smaller than the total number of PWR plants operating during the 1985 to 1996 period. Note that the frequency data for the older plants (age 24 to 30) are based on less than 10 plants (small sample size). Leak events per operating year, instead of operating plant, can be determined if data related to operating year for a given age are estimated.

Trends of the vibratory fatigue related leak events in calendar time and plant age were statistically investigated as shown in Figures 4-10 and 4 11, respectively. Details of the investigation are presented in Appendix C. The effect of calendar time reflects the evolving body of regulations, design improvements, and industy wide learning. '" effect of plant age reflects the learning of plant personnel and the aging of the hardware. Solid curves in 1 igures 4-10 and 4-11 show the estimated trend in calendar time and with age, respectively, ne dotted lines show 90% confidence bands. A trend in this investigation, if present, is assumed to have an exponential form. The data are consistent with this modeling assumption, because all point estimates of 90% confidence intervals overlap tiie confidence bands. The results in Figure 4-10 reveal no trend in calendar time, whereas the results in Figure 4-11 reveal a statistically significant decreasing tand with age. Apparently, the decreasing trend implies that the vibratory fatigue failures are because of inadequacy of the initial design and fabrication, and not aging damage.

The vibratory fatigue cracking events occurred either because of pump-induced pulsations or cavitation and flashing. Two of the leak events caused by vibratory fatigue are analyzed next. Rese leak events are

selected fer analyses because they illustrate the role played by the all three excitation mechanisms described earlier and also the synergistic role of water hammer in causing some of the failures.

4.6 Analysis of Positive Displacement Pump-induced Vibration Fatigue Cracking of Charging System Piping at Diablo Canyon Unit 1 Event Description. On July 26, 1990, an NRC inspector found a leak during a planned visual inspection of the charging pu.np pipir.g. He discovered both boric acid crystals and moisture on a 4-in.-

diameter elbow, upstream of the suction stabilizer, in the suction piping of the positive displacement pump in the charging system. The crack was located in an unisolable section of the common suction header of the positive displacement and centrifugal charging pumps. A small amount of reactor coolant was leakmg DRAFT - NUREG/CR-6582 4-6

1 VIBRATORY FATIGUE from the crack, which evaporated immediately upou exposure to air. The leak rate was about 1.51 Umin ,

(0.4 gpm)(Diablo Canyon 1990).

The crack initiated at the toe of the weld which attaches a 19-mm (0.75-in.) half coupling to the side of a b 102-mm (4.in.) ell'ow. (Some examples of half coupling connectins are shown in Figure 412.) The 19-mm cormection is a test pressure instrument tap, which was carnileured perpendiculas to the piping elbow.

The crack extended into the base metal of the c! bow !n both directions tangential to the circumference of the weld. The ultrasonic exammation an-1 visutiinspection corAned that the crack was 64 mm (2.5-in.)

long.

Roof Cause Analysis. Vibratory P.tigue was the main reason foi the cracking of the welo. The crack was not caused by intergranular stress corrosion cracking as indicated by the path followed by the crack.

An in;ergranulst stress corrosion crack follows a path along the heat affected ime of the weld, which would correspond to a circumferential crack around the toe of the weld. However, the path of the crack in this failure was along the tangential direction away from the heat-affected zone similar to the one shown in Figure 43.

Two factors contributing to vibrational fatigue were present in the Cnarging system: the positive displacement reciprocating pump an:1 the configuration of the test pressure instrtunent tap, which was cantilevered perpendicular to the piping elbow without sufficient restraining support. Some examples of cantilevered configurations are shown in Figure 4-12. Pump-induced pressure pulsations from positive displacement pumps are generally of large enough magnitudes to cause vibrational problems, unless the system is designed to mitigate these pulsations. There have been more than 80 events of cracking or leakage in charging pump and associated piping systems during the period of 1978 to the middle of 1980.

Two operating characteristics of reciprocating pumps cause vibratory fatigue problems. An instantaneous demand of fluid is created at the beginning of a suction stroke by the high acceleration of piston. This demand will accelerate the fluid and lower it pressure, which can result in cavitation and large pressure pulsations in both suction and discharge piping. In addition, the hydrogen cover gas, which is placed in the volume control tank to control oxygen in the RCS system, comes out of the solution during the suction stroke and contributes to cavitation. The cavitation causes cracking of the fluid cylinder, in addition to vibratory fatigue cracking of piping. The escape of hydrogen in the fluid has caused hydrogen embrittlement of the internal valves in a pump at Callaway; 15 to 20% of hydrogen was found in the valves removed from the pump. It is also believed that hydrogen embrittlement contributed to the cracking of the fluid cylinder at Callaway (Kochert and Rauch 1997). The solution to this reoblem is to provide an ample supply of fluid near the pump suction to satisfy the need of the instantaneous acceleration head. This problem would not likely be as severe in the event being analyzed provided the stabilizer connected to the pump suction line was effective in satisfying the instantaneous demand of fluid.

The second sot.rce of vibratory fatigue problems in piping connected to positive displacement pumps is the pressure pulsations caused by the reciprocating pistons. The periodic variations in the flow into and out of the pump may be expressed in harmonics of the pumpi ig cycle for the purpose of analysis. When one of the flow hannonics coincides with the fundamental or a higher harmonic of the suction or discharge piping system, severe pressure and flow pulsations may develcp. As the number of cylinders / pistons in the pump are Exceased, the amplitude of the periodic flow variation reduces, and as a result, the severity of the pressure pulsations decreases.

4-7 DRAFT - NUREG/CR-6582

VIBRATORY FATIGUE De pressure pulsations may be mitigated by the use of discharge dampeners. Energy absorbing dampeners such as nitrogen-filled bellows are used to mitigate these pulsations. The bellows provide cushion, reduce pressure peaks, and dramaticrily reduce pressure fluctuations in the discharge piping. If the bellows fail, the dampeners will lose their nitrogen charge, and the suction piping will vibrate severely because of pump pulsations The failures of:dtrogen-filled bellows have caused vibratory fatigue cracking of the charging pump suction drr.in line at the Catawba plants (Kelly and Davis 1995). However, the role played by the i

dampenert in the vibratory fatigue event at the Diablo Canyon 1 is not known.

The test pressure instrument tap was cantilevered perpendicular to the piping elbow without srfficient restraining support. Therefore its matral frequency was low. As a result the tap was excited by the pump-induced pulsations and expc.-ieaced vibratory fatigue cracking.

Corrective Actions. Tin crack m the elbow of the suction pioing was repaired following an ASME Section XI weld repair procedure. Additional supports were added to the Units 1 and 2 positive displacement pump piping. It was planned to install vibration monitoring instrumentation on bodi the '

section and discharge side of the positive displacement pump piping at both Units 1 and 2. The analysis of the data was te be used for determining whether any further modificatione are necessary. The main corrective actiort at the Catawba and Callawry plants was to use the centrifugal charging pump instead of positive displacement pumps. A centnfugal pump does tnt create the dramatic pressure spikes that can induce vibratiors and make hydrogen come out of solution to cause cavitation and hydrogen embrinlement.

Safety Signl#cance. Even though the crack was in an unisolable section of the common suction header of the positive displacement pump and two centrifugal charging pumps, it had little effect on the operability of there pumps because the leak rate was low (0.4 gpm). The technical specification leakage limit for the centrifugal pmps is 38 IJmin (10 gpm). whereas there is no technical specifications for the positive displacement pump.

However, to meet 10 CFR 50 Appendix R requirements for fire protection at nuclear power plants, the updated FSAR for Diablo Canyon 1 takes crulit for the positive displacemmt pump for two purposes: (1) providing an alternative and dedicated shutdown capability, and (2) maintaining the reactor coolant level within the level indication in the pressurizer. As there is no fire barrier between the two centrifugal pumps, the positive displacement pump provides shutoown capabilities should a fire occur in the centnfugal charging pump room. As a result of the crack and the associated leak event, Diablo Canyon 1 established two administrative controls to limit operation of the positive displacement pump: W not operafmg the positive displacement pump during normal plant operation to mmmuze potential darum to pipe, and (2) continue maintaining operability of the positive displacement pump so that it can be used to meet the 10 CFR 50 Appendix R requirements.

4.7 Analysis of Cavitation-Induced Vibrations of Letdown -

System Piping at McGuire 1 On August 16, 1987, a leak was discovered in the letdown system dmin line. It appears that both cavitation and water hammer played a role in this failure (McGuire 1988). To facilitate the analysis of the event, a brief description of the letdown system is presented. Then a description of the event, root cause analysis, corrective actions and safety significance of the event are desented.

DRAFT - NUREG/CR-6582 4-8

l VIBRATORY FATIGUE i

)

Letdown Systern Description. Figure 4-13 shows a portion of the letdown system which is a part of the chemical and volume control system for a four loop Westinghouse plant (USNRC Technical Training Center). The letdown fluid tapped off the Loop C cold leg travels through two series letdown flow control valves to the shell side of the regenerative heat exchanger, which provide:; initial cooling of the letdown fluid to 143'C (290*F) by preheating the returnmg charging stream. Then the letdown fluid passes through one of the three parallel flow paths, which controls the amount of coolant that is letdown (removed) from the RCS and provides the initial pressure reduction from 15.5 MPa (2250 psig) to 2.4 MPa (350 psig).

Two of the flow paths have fixed letdown orifices, designed for 170 and 284 Umin (43 and 75 gpm), and the third flow path has a control valve (also referred as a variable orifice). The letdown fluid, after passing through the letdown orifices, exits the containment and travels to the tube side of the letdown heat exchanger where the fluid is cooled by the component cooling water to a temperature [46'C (115'F)] that is compatible with an ion exchanger resin. From the leziown heat exchanger, the fluid passes through the letdown back-pressure regulating valve, which maintains 350 psig back pressure to prevent the letdown liquid flashing to steam. Then the letdown fluid passes through the ion exchangers, letdown filter, and finally to the volume control tank.

Event Description. On August 16,1987, following a reactor trip, the letdown isolatior valves were opened to establish normal letdown and charging flow. At that point, letdown flow and pressure spiked and lifted the letdown relief valve. Approximately 45 minutes later, operations pers.mnel noticed a frequent activation of automatic makeup for the volume control tank and increasing cont.inment floor and equipment sump levels, which were signs of a reactor coolant leak taking place. The estimated leak rate was 151 Umin (40 gpm). Operations personnel entered the containment building to identify a suspected leak and discovered a crack in the socket weld on the drain line off the letdown line. Normal letdown was isolated by closing the containment isolation valves to stop the leak and excess letdown initiated. The isolation valves remained closed for about 30 minutes.

Root Cause Analysis. Metallurgical analysis results indicated that the failure was initiated by vibratory fatigue followed by an undefined overload. The crack initiated at two places about 180 degrees apart. The initiation sites were separated by two regions, each about 12% of the circumference, where failure occurred because of overload. One of the overload failures was a ret'dt of mechanics removing l the sample from the system for the metallurgical analysis. The other overload failure was probably a result of water hammer from relief vahe operation. Thus, both vibratory fatigue and water hammer had contributed to the failure.

l The letdown fluid passing througn the orifices experiences a reduction in pressure from 15.5 to about 2.4 MPa. If the reduced pressure approaches the vapor pressure, cavitation takes place at the downstream end of the orifice. Vapor pockets form and collapse in the low-pressure region with extreme rapidity, generating intense shock waves and resulting in broadband pressure pulsations. This can result in severe vibration of the drain line downstream of the orifice. However, the review of vibratory testing performed one year earlier revealed that significant vibratory loads did not exist under any normal operating condnions. So it was concluded that vibratory fatigue cracking was not caused during normal operation but might have occurred during pre-operational testing. Thus, it appears that vibratory fatigue might have contributed to crack initiation but not propagation.

The occurrence of water hammers is possible in the letdown system because large changes in pressures and temperatures of the letdown fluid take place. A possible mechanism for water hammer is as follows. A reduction or interruption of charging flow through the regenerative heat exchanger will result in a higher 4-9 DRAFT - NUREG/CR-6582 l

{

VIBRATORY FATIGUE exit temperature for the letdown fluid. If this temperature is too high, flashing may take place at the downstr9am side of the letdown orifice where the pressure is reduced, and steam bubbles are formed.

During subsequent initiation of letdown and pressurization of the letdown line, steam bub 6les come ha contact with the cold letdown fluid, collapse, and water hammer may take place (Uffer et al.1982).

Investigation of the p; ant operating history indicated that water hammer might have played the main role in the failure. Inspection of the control valve revealed that its seat and the tapered plug were badly eroded and the valve must have leaked substantially in the fully closed position. So the control valve was presumably leaking during the 30-minute period when the isolation valves were closed. Under these conditions, the letdown relief valve opens as the pressure on the downstream side of the orifice builds up to 4.1 MPa (600 psig). Then, as the pressure reduces the valve closes, and the pressure builds up again.

Vapor bubbles are likely to form in the letdown line dowrcacen of the orifice when the r: lief is open and water hammer may take place when the relief valve closes. Thus, each cycle of opening and closing of the relief valve is likely to cause water hammer.

~

Other factors might have also caused water hammer. The regenerative heat exchanger was ineffective without charging flow in removing heat from the letdown fluid, and steam bubbles might have been formed. These voids could have caused water hammer if the letdown was established too rapidly. Jhen the letdon flow rate is high, interaction forces between the steam and water can create enough turbulence to form a water slug. As steam condenses, the pressure of the steam bubble reduces, the cold slug accelerates into the void, arid a water hammer would take place.

So the root cause analysis suggests that the crack was initiated by vibrational fatigue cracking at the socket weld of the drain line and eventual failure was most likely caused by water hkmmer. A somewhat similar failure occurred at Maine Yankee where a water hammer caused a fracture at an existing fatigue crack in a feedwater line (Garrity 1983).

Corrective Actions. Six additional weld / pipe locations were inspected, and one additional crack was found. The crack was not throughwall and was repaired. Sixteen pipe supports were inspected and no damage was found. A need for an additional isolation valve between the letdown orifice and the relief valve was also being considered. Procedures invoiving initiation of letdown were changed to reflect the Duke Power recommendations.

Several vibratory fatigue cracking of the letdown piping at McGuire and Catawba plants have taken place since this event. Root cause analysis was performed, vibration measurements were made, and several modifications were made. The fixed orifices were redesigned, tested, and installed. In addition, the centrol valves have been reworked and some have been replaced. The letdown pipe was changed; socket welds were replaced with the butt welds which have better fatigue resistance. It appears that these corrective actions are effective; no vibratory fatigue cracking of the letdown piping has been reported at -

these plants since implementing the corrective actions a few years ago.

Safety Consequences. The leak from the control valve was not a problem during normal operation because letdown flow was desirable. Ilowever, under contamment isolation conditions, the control valve acts as an isolation valve and the leak constituted a reactor coolant leak. The leaks from the valve and the leaking drain line were within the scope of the Final Safety Analysis Review small-break LOCA analysis.

These leaks were terminated by closing the letdown isolation valves upstream of the regenerative heat exchangers.

DRAFT - NUREG/CR-6582 4-10

L VIBRATORY FATIGUE 4.8 ASME Code Section XI Inservice inspection RequirementsSection XI of the Code (ASME,1995b) requires a surface inspection of 100% of the weld length for socket welds. Figure 4-14 (Figure IWB-2500-8(a) of Section XI) shows the extent of inspections. Table IWB-2500-1 of Section XI requires that the total number of circumferential butt welds (or branch connections or socket welds) inspected equals 25 % of the number of those welds h. the RCS. Ilowever, according to IWB-1220 of Section XI, socket welds on piping smaller than 1 NPS size, which are more likely to experience vibratory fatigue failures, are exempted from the surface examination roptirements.

Appendix D of Section XI provides advice on flush grinding piping welds to improve both fatigue life and UT inspectability. Grinding the toes of fillet welds (socket welds) is a well-known method to improve fatigue life.

Volumetric exammation of socket welds is not required and generally not performed. However, at least one utility has used radiographic examination of socket welds to detect any volumetric defect such as area oflack of fusion (Millstone 1994).

4.9 USNRC Requirements / Communications NRC vibratory test requirements for startup and initial testing were listed in Regulatory Guide 1,68 (USNRC 1978a). The requirements for FSARs were listed in Regulatory Guide 1.70 (USNRC 1978b).

These requirements are now listed in the Standard Review Plan (SRP) (USNRC 1981). The testing required in essence conforms to ASME OM-3, but the SRP contains more details in some instances cach as a description of what transients should be included. Requirements for hydrodynamic loads were incorporated into Section 3.9.3 of the SRP in 1981.

Plant technical specifications have no direct requirements for vibration monitoring or testing of small piping lines. There are requirements for leak detection and loose parts monitoring systems, which would indirectly monitor for piping failures. There are requirements to maintain piping system integrity to ASME Code Specifications, and to inspect piping in accordance wuh ASME Code Section XI requirements and frequencies.

in NUREG-1061, Volume 4 (USNRC,1984), the conclusion was made that unanticipated vibratory loads were accounted for by the dynamic testing requirements of SRP Section 3.9.2. More recently, the NRC Generic Safety issues (NUREG-0933), Task Action Plan (TAP) includes item C-12 "Prunary System Vibration Assessment". TAP C-12 recognizes that there have been a number of vibration problems in operating plants, but the NRC/NRR Mechanical Engineering Branch position is that the current guidelines in Section 3.9.2 of the SRP combined with staff positions on loose parts monitoring (Regulatory Guide 1.133) are sufficient for resolunon. The Mechanical Engineering Branch of USNRC considers that solving vibration problems on a case-by case basis is the only practical course of action.

4.10 Mitigation and Monitoring of Vibratory Fatigue Cracking Mitigative actions for vibratory fatigue cracking may be divided into three categories: (1) use of pulsation dampeners, (2) increased natural frequency of the piping system, and (3) modified weld design. Some of these actions have been mentioned in Sections 4.6 and 4.7.

4-11 DRAFT - NUREG/CR-6582

VIBRATORY FATIGUE l

Use of Pulsation Dampeners. The severity of pump-induced pulsations may be reduced by providing pulsation dampeners. VeccMo (1996) points out that at one nuclear plant, the charging system had been retrofit with pulsation dampeners on the pump suction side to p. ovide an in-line reservoir to atternate pressure pulsations. The pumps were designed to reduce the negative effects of an oscillating net positive

, suction head. One pump had also been retrofitted with a pulsation dampener en the discharge side to reduce pump cavitation. However, from subsequent failures of socket welds in the system, it was apparent that the effectiveness of the installed dampers was limited.

Effectiveness of other types of dampeners is also limited. For example, nitrogen-filled suction dampeners were used at one plant. When the bellows of the dampeners failed and the dampeners lost their nitrogen charge, the suction line experienced severe vibrations resulting in a failure of a suction drain line. The most effective solution to reduce the pump-induced pulsation is to replace the normal use of positive-displacement reciprc,cating pump with a centrifugal pump. Several plants have made such rephcements and discontinued the normal use of the reciprocating pump during plant operation.

Increased NaturalFrequency of the Piping System. The natural frequency of the piping system may be increased so that it is higher than the frequencies excited by the pump-induced pulsations or cavitations, and a resonance condition that results in a magnification of the amplitude of vibration does not occur. The natural frequency of the piping system is proportional to the square root of the stiffness of the system divided by its mass. So the frequency of the piping system can be increased by replacing the existing piping with a larger diameter and a thorter length piping, replacing the existing valves on the piping system with valves having smaller weight, and providing additional supports. For example, one plant had experienced repeated vibratory fatigue cracking of the welds on the drain line and relief valve line downstream of each charging pump. Modifications to the relief valve line included replacement of the existing 19-mm (0.75-in.) line with a 51-mm (2-in.) line, replacement of the existing discharge drain valve with a much lighter valve, and addition of supports. These changes increased the natural frequency of the charging pump discharge line beyond that of the pump-induced pulsations.

OM-3 (ASME,1994) states that experience has shown that the most effective restraint is obtained by supporting piping near bends, at all heavy masses, and at piping discontinuities. Vibrations of vents, drains, and bypass and instrument piping can be corrected by bracing the masses (for example, valves or flanges) to the main piping to eliminate vibrations. The designer must recognize that supports are also susceptible to vibrations and must be designed and inspected to account for vibrations.

Modi #ed Weld Design. The size of the fillet weld may be increased to provide higher load carryira capacity. For example, in one plant, fit up of a misaligned piping introduced high stresses in tb socket weld. As a result, a vibratory fatigue crack was initiated at the toe of the weld where a high stress concentration was present. As a corrective action,100% of the damaged weld was removed by grmdmg and replaced with an unequal-leg fillet weld having a higher load-carrymg capacity and a lower stress concentration.

DRAFT - NUREG/CR-6582 4-12

VIBRATORY FATIGUE Some plants have also replaced the socket welded pipes with butt welded pipes to increase the, fatigue resistance of the weld joint. %c combined ASME Code Section Ill stress index" for moment loats (C 2K2) for socket welds is 4.2 and for as-welded butt weld it is 1.8. So the SCF is reduced by factor 2.33. The resulting increase in fatigue life depends on the amplitude of the cyclic stress. For example, at 40 ksi amplitude, the use of butt weld instead of socket weld will increase the life of stainless steel piping from 1 05 to 10' cycles; and for 30 ksi amplitude, from 5x105 to greater than 10" cycles (ASME 1995a).

Vibration Monitoring. Vibration monitoring of susceptible locations to identify the source and magrdtude of the vibration, the modes of plant operation during which vibration takes place, and the magnitude and frequency of the response can assist in solving vibration problems. Vibration monitoring has been used at several U.S. PWRs as part of the root cause analyses of vibratory fatigue problems.

Instruments such as accelerometers and strain gages can be used for monitoring, but the size of an accelerometer model may be a problem for small-bore piping because the mass of the accelerometer affects the response of the piping.

4.11 Industry Efforts to Manage Vibratory Fatigue In the past, plant owners addressed piping vibration only as a result of leakage or ruptures. Presently, more emphasis is being placed on managing vibration-related failures. However, since there are literally thousands of small-bore lines in nuclear power plants, attention to each line is not cost-beneficial.

Some of the problems facing plant owners in dealing with vibratory fatigue are (EPRI 1995):

The failures often appear to be random,

There is a lack of historical data,

. There is seldom a thorough root-cause analysis,

Failure locations may be inaccessible for vibration monitoring,

It may not be cost beneficial to investigate the cause and prepare an adequate redesign, *

nere is no coherent industry approach, and

1.ack of communication.

The EPRI has issued a Farigue Management Handbook which includes information on managing vibratory fatigue (EPRI,1994; EPRI; 1995; Smith,1996). A fatigue database is being developed. As of 1995, a total of 446 fatigue failures are recorded, of which 80% are attributed to vibratory fatigue. An estimate of the cost of vibratory fa'igue failures is $16 million per year for U.S. nuclear plants (about $160,000 per plant per year). In addition to the nature of the failures found in other studies (that the failures occur at

, socket welds in small piping near sources of vibration), the study found that often the failures occur in Y

short, cantilevered attachments to large bore pipes (for example, vents and drains, instrument taps with socket welds, see Figure 4-8).

The ASME Code stress index for socket welds was based on tests by Markl and George in 1950, and assumed that the failures would occur at the weld toe. However, field experience has shown that the weld t1 For Class 1 piping (NB-3600), the ASME Code uses stress indices to account for stress concent ations. It is applied in the same way as a SCF.

4-13 DRAFT - NUREG/CR-6582 1

l VIBRATORY FATIGUP

(

root is the more common location of failure. Japanese test data (Higuchi et al.,1996a,b) r' ow that the weld root is the most likely failu e location. For a good weld, the EPRI han:ibook estimates that the SCF may be 2.6, for a fair weld 4.2, and for a poor weld 8.0. Thus the ASME Code indices is noncenservative for the poor weld. The Fatigue Management Handbook has a screening criteria to deterruine where vibratory fatigue failures may occur, one for compley, systems, and one for short cantilevered systems.

The screening criteria compare field measurements of maximum displacement, velocity, or acceleration to allowable values specified in the handbook. The results of studies of failed piping show that the OM 3 criteria are inadequate to prevent fatigue failures in socket welds in cantilevered attachments excited by high-frequency loads, and lower velocity limits are r--Jed to preclude failures. EPRI is continuing the program to better understand and charnterize socket-weld quality with respect to high-cycle fatigue.

Other than the EPRI Program, nuclear plant owners are addressing vibratory fatigue problems on a case-by-case, plant-by plant basis. The approach is far more often reactive, that is, repairing a damaged piping system, than proactive. Typically no root-cause analysis is performed. The failed piping may simply be replaced, or it may be strengthened with more suppom, thicker piping, or socket welds may be replaced '

with butt welds.

Other research is being cotvlucted to determine factors associated with weld failures (Higuichi et al.,1996a and 1996b; Vecchio,1996). In the Higuichi et al. papers, the authors found that relieving the socket weld residual stress by post-weld heat treatment improves the fatigue strength greatly, while moving the failure site from the root to the toe of the weld. Socket welds fail at the root because e the residual stress being tensile in this region (as well as the stress concentration effect). A thicker wall results in a higher fatigue strength. The presence of a defect in the toot can degrade the fatigue strength greatly; a fatigue strength reduction factor of 15 was estimated for a defect at the weld root, rather than the factor of 4.2 in the ASME Code. The fatigue strength can be improved markedly with a final refmement pass fimshed in a concave profile. The researchers found that, although not allowed by the standards, the fatigue strength could be increased 40 to 50% by eliminating the slip-on gap. The EPRI program is evaluating the effect of different parameters as identified by the Japanese researchers on the fatigue life of the socket welds. The fatigue testing for this evaluation is being carried out at Diablo Canyon with the technical assistance from StructuralIntegrity Asr.,,ciates."

Vecchio analyzed socket weld failures in a nuclear power plant and found that the stress levels calculated for the failure locations showed excellent agreement with data contained in American Association of State Highway and Transportation Officials (AASHTO,1989) S-N curves. These curves are based on an empirical approach that used statistically significant fatigue data from actual or full-size weld specimens, in contrast to the ASME Code approach that used data from smooth, base-metal specimens.

U V. N. Shah, Private conversat%u with Greg Frederick, EPRI NDE Center, July 10,1997.

DRAFT - NUREG/CR-6M2 4-14

5. ANALYSIS OF LEAK EVENTS ASSOCIATED WITH REACTOR COOLANT PUMP SEAL FAILURES RCP seals limit leakage of the high-pressure, high-temperature reactor coolant to the containment. This fs accomplished by directing most of the coolant leaking along the pump shaft back to the chemical and volume control system and the remainder to the reactor coolant drain tanks. Thcrefore, the pump seals are part of the prunary pressure boundary but were excluded from tl.e ASME Code requirements. Failures of the pump seals could potentially result in a small-break LOCA if leakage through the seals exceed the capacity of the normal charging systems. Complete loss of component cooling water can result in a common mode failure of RCP seals. Such failures could lead to a small-break LOCA at certain plants where high pressure coolant injection pumps are also cooled by a component cooling water system and, therefore, becomes unavailable; that is, emergency makeup capability is not available (Jackson 1988'.

In the early 1980s, the USNRC performed a prelimmary review of LERs that revealed a relatively large

, number of RCP seal and seal auxiliary system failures at operating plants. Further study indicci that the overall probabil:ry of core melt caused by small breaks could be dominated by RCP seal failures.

Concerned by this, the USNRC staff assigned a high priority to the investigation of RCP seal failures and identified it as Genenc issue 23, Reactor Coolant Pump Seal Failures.

The main objective of ttus section is to analyze the 1995 St. Lucie ! RCP seal failure. Since Bryon-Jackson pumps are employed at St. Lucie 1, this section focuses en the design and operating experience with Bryon-Jackson RCP seal, called an SU seal, used in Combustion-Engineering-designed plants.

Reviews of the field experience performed in early 1990 imply that the Byron-Jackson pump seals experienced a larger number of field failures. For example, Bell and O'Reilly (1992) performed a preliminary review of 'he RCP seal failure experience database for the period January 1,1984 through January 1,1990, and found that a large number of seal failures, particularly that of th. original SU type design described in Section 5.1, were in Byron-Jackson pumps, which are generaby employed in B&W-and Combustion-Engineering-designed PWRs and in boiling water icactors. Bell and O'Reilly (1992) also refers to the. results of an industry analy sis of RCP performance during the 19801990 period that confirm their finding. The resuhs showed PWR plants with Byron Jackson pumps had the highest total percentage of capacity factor losses due to RCP prcblems than plants with RCPs manufactured b; other vendors.

Problems with pump seals were responsible for more than 45 % of these losses. However, several design improvements in the Byron-Jackson pump seals have been made since 1983 and, as a result, the seal operating experience at all Combustion-Engineering-designed plants has improved. Our review of leak events from 1985 to 1996 identified only three events associated with Byron-Jackson pumps out of eight reportable leak events associated with RCP seals, indicating improved seal performance.

The organization of this section is as follows. He design of the SU seal is described first. Several imprvvements in the seal design are summanzed. Then U.S. PWR data related to leak events caused by RCP seal failures during 1985-1996 period are presented. Then the 1995 St. Lucie 1 RCP seal failure is analyzed, and the proposed USNRC resolutions for resolving Generic Issue 23 and industry comments are sununanzed.

5-1 DRAFT - NUREG/CR-6582

IGALTOR COOLANT PUMP SEAL FAILURES 5.1 RCP Seal Design ne Byrondackson RCP, shown in Figure 51, is a vertical, centrifugal pump with bottom suction and horizontal discharge. The pump seal assembly comists of three- or four stage seals, depending on the nuclear steam supply system (NSSS) vendor. In a typical Combustion Engineering designed PWR, the pump shaft is sealed by a four stage, controlled leakage, face-type mechanical seal; three full-pressure mechanical seal stages installed in series; and a founh low-pressure vapor seal stage installed on the top.

Each st2Et within a seal assembly has the same design and is fabricated from the same material. Each stage is designed to withstand full RCS pressure should other stages malfunction. However, the seal design uses all three mechanical seal staEcs, instead of only one seal stage, to reduce system pressure to volume con'rol tank pressure, as shown in Figure 5 2. Using pressure breakdown ori.lces, which are about 2.54 mm (0.1 in.) in diameter, the seal normally operates with 14.8-15.5 MPa (2150-2250 psi) pressure on the ,

lower stage,1010.3 MPa (1450-1500 psi) pressure on middle stage, and 4.8-5.2 MPa (700-750 psi) on the upper stage. The vyor seal stage prevents liquid or gaseous leakage from escaping to containment because it opere' at pressure lov'er than the containment pressure. The vapor seal stage normally -

operates at volumn control tank back pressure [0.14 MPa (20 psi)) and is capable of sealing against full system pressure when the pump is shut down and during coast-down (Mitra et al.1986, Bell and O'Reily 1992).

A schematic of a Byrondackson RCP mechanical seal stage is shown in Figure 5 3. The stage consists of primary and secondary seals. The primary seals limit the leakage of reactor coolant between a shaft-mounted titanium carbide rotating face and a stationary graphite face. The secondary seals (elastomer O-rings and U-cups) prevent leakage between stationary mechanical elements of the pump seals or those elements that have a very small relative velocity. Both primary and secondtry seals and pump bearings require continuous cooling while the pump is in operation or while it is stationary during hot snutdown conditions.

Methods used for cooling the seals in the B> rondackson pumps depend upon the NSSS vendor, in B&W plants and four Combustion Engineering plants (Maine Yankee and Palo Verde 1,2, and 3), an injection system is used to cool the seals, which are three-stage design. This design does not include a vapor seal state. Similarly, in Westinghouse RCPs, an injection system is used to cool the seal, which does not include a vapor seal stage. An independent source provides cooled high-pressure water to the injection system for cooling the seals. In other Combustion Engineers plants, an injection-less system is used to cool the seal:, which are frur-s* age desigt (Lucb s et al.1986).

Figure 5-2 shows the path of the reactor coolant leaking riong the pump shaft in an injection-less system.

Prior to entering the seal, the reactor coolant is cooled by component cooling water in the thermal barrier and then in the integral, concentric tube heat exchanger, shown m Figure 5-1. We auxiliary impeller ,

recirculates primary coolant from the discharge of the pump thrwgh the inner tubes of the heat exchanger and back to the suction of the auxiliary impeller, he flow rate of the recirculating water is 152 IJmin (40 gpm). Component cooling water with a flow rate of 106 IJmin (28 gpra) passes through the outer tubes of the heat exchanger and cools the primary coolant. The temperature of the primary coolant recirculates

.hrough the heat exchanger is reducH from 288'C (550'F) to 52-66'C (125-150'F) (USNRC Technical Training Center, Ruger and Luckas 199).

DRAFT - NUREO/CR-6582 5-2

REACTOR COOLANT PUhfP SEAL FAILURES in some Byrondackson pumps, the recirculation flow through the heat exchanger reases if the pump impeller spins to a stop, but the controlled bleed off flow through the seal is maintained without being cooled because of pressure drop across each stage, in other Dyroa-Jackson pumps, the recirculation flow cnntinues when the pump is stopped (Combustion Er.gineering 1991).

A flow of 3.8 Umin (1 gpm) of cooled liquid from the recirculation path is intentionally passed through the seal assemblies in order to provide cooling and lubrication for moving parts, if this liquid is not pre-cooled properly or contains abrasive contaminants, seal operation may be compromised. Approximately 99% of the 3.8 Umin flow passes through the orifice breakdown devices, and the remaining 1 % passes between the seal faces for lubrication. The recirculating liquid in the heat exchanger is replenished from the discharge of the RCP for the 3.8 Urnin leakage.

The water passing through the seals is used to cool the seal and to equalize the pressure drop across each seal stage. hiost of this water leaves the upper seal stage through the controlled bleed-off line to the volume control tank. About 1.14 Umin (0.3 gallon'h) leakage past the vapor seal stage is collected in the coolant radwaste system (hiitra et al.1986). Both of these leakages are identified leakages.

Pump Seallnstrumentation, The instmmentation provides for measurement of seal pressures and temperature and controlled bleed-off flow and temperature. Seal pressures are measured between the lower and middle seal stages, middle and upper seal stages, and upper seal and vapor seal stages- The inlet temperature to the lower seal stage is also measured. The pressure and temperature measurements can be used to determine the seal failures. When any stage is failed, controlled bleed-off flow increases as discussed above. An alarm is set for bleed-off flow of 4.75 Umin (1.25 gpm) to alert the operator of a possible seal failure (USNRC Training Center).

5,2 RCP Seal Failure Modes The RCP seal failure may be caused by loss of seal cooling or by mechanical danage. Loss of component cooling water to the heat exchangers will result in loss of seal cooling for the injection less RCP seal and lead to an increase in the controlled bleed-off temperature. At higher temperatures, the clastomers (materials for secondary seals) become hardened and their characteristics are degraded for continued use,

, and the entire seal assembly needs to be replaced flowever, the test results and field experience show that l the damaged seal can su, vive the loss of seal cooling for several hours (Combustion Engineering 1991).

For example, in one related incident at h{illstone Unit 2, the duration of the loss of seal cooling was about

, 5 h (Northeast Utilities 1991). The controlled bleed off temperature, which is normally at 54'C (130'F),

increased tu approximately 166'C (330'F), and the lower seal cavity temperature, normally at 54*C, increased beyond 182'C (360'F). The lower seal stage was failed and the middle seal stage was damaged.

The seal operated satisfactorily in the degraded stage for approximately 2 months until it reached the next refueling outage: the seal was changed at that time, hiechanical damage may be caused by excessive pump vibration, introduction of contaminants, high frictional torque, and pressure, temperature, or flow transients in component cooling water systenis.

hiechanical damage may also be caused by defective parts, faulty installation, or improper maintenance.

Blockage of the pressure breakdown orifice by loose wires inside the seal has also caused seal degradation.

When a seal stage fails, the pressure drop across that stage reduces and controlled bleed-off flow increases.

For example, when the lower stage fails, the pressure drop across that stage decreases, and leakage 5-3 DRAFT - NUREG/CR-6582 l

REACTOR COOLANT PUMP SEAL FAILURES (controlled bleed-off) through that stage increases. As bleed-off flow increases, its temperature also increases because the flow of component cooling water through the concentric tube heat eachanger remains unchanged (Northwest Utilities 1991, IAuisiana Power and Light 1985). Because of the reduced pressure drop across the lower stage, the pressure drop across the middle and upper stage increases; however, the operating pressure of the vapor seal remains at 0.145 MPa (20 ps!). As long as the vapor seal pressure is lower than the contamment pressure, a failed seal stage will result in an increase in the identified leakage, that is, increased controlled bleed-off to the volume control tank and vapor seal leakage to the coolant radwaste system, if a complete failure of all four seal stages, including the vapor seal stage occurs, unidentified leakage to l the containment may occur Field data related to pump seal failures show that the integrhy of the vapor seal has been generally maintained. But in one incident, complete failure of all four seals, including vapor

seal, occurred mi resulted m leakage past the vrpor seal into the containment building at Arkansas Nuclear One Unit 2 August 1988 (Bell and O'Reilly 1992). Vibratory fatigue caused complete severance of a seal sensing line, iruttatmp RCP seal degradation. There was not sufficient time to enter the containment and ,

isolate the leak before sigruficant seal damage. Within 10 minutes of the sensing line failure, the controlled bleed-off flow was reduced and seal degradation continued until the last two seal stages, upper and vapor stages, were indicatmg RCS prer,sure. At this point the operators manually tripped the reactor and stopped the affected RCP. The muunum leakage rate was 152 Umin (40 gpm). Most of this leakage was coming from the sensing Itne, not the seal. The sensing line leakage can be as high as 141 Umin (37 gpm). Later investigation revealed that the carbon faces in the 2d and 4* stages were broken and those in the 1" and 3" stage were cracked. In addition, there was partial U-cup extmsion (ANO 21988, Combustion Engineering 1991).

5.3 RCP Seal Gasign improvements Operating experience with Byrondackson SU seals poor to 1983 was not reliable. A variety of approaches have been taken to improve the reliability of thesocals. The approaches for enhancing seallife included improved quality control, maintenance, and istrumentation and design changes. Seven plants changed to a different seal configuration, as listed in Table 51 There were 47 Byrondackson SU pump seal assemblies in f ervice in 1983. This number was reduced to 23 by 1991, including several modified Byron-Jackson SU seals. The remaining 24 SU seals were replaced with three advanced seal designs: Byron-Jackson N-9000 seals, Bingham Willamette seals, and AECL CAN4 seals designed by Atomic Energy of Canada Limited (AECL). The advanced seal designs used in the Combustion Engineering-designed plants are multistage seals with a vapor stage as the fourth stage. The main objective of the advanced designs was ,

a longer and more reliable seal life. The Palo Verde plants, which ca.ne into operation after 1983, have RCPs including seals designed by Klein, Schanzlin, and Becher (KSB) (Combustion Engineering 1991, Bell and O'Reilly 1992).

As a result of these modifications, the RCP seal operating experience at all Combustion Engineering Owners Group plants improved since 1983. Most of the seal failures since 1983 occurred in the 1983-86 time frame, indicating an improvement trend (Combustion Engineering 1991). These seal failures were defined as occurrences when two or more seal stages were no longer able to maintain a pressure differential, the controlled bleed-off temperature or flow was not within normal operating range, or the exterral seal w as above the acceptable value in accordance with the plant specific operating procedures.

DRAFT . NUREG/CR-6582 5-4

. - -.- .. - - - = . -- .- _ - .=

REACTOR COOLANT PUMP SEAL FAILURES The modified Byrondackson SU seal included several improvements that reduced secondary seal friction and increased hydrodynamic lift at low pressure. In addition, the bleed-off flow rate was increased from 3.8 to 5.7 Umin (1 to 1.5 gpm) to improve the stability of the seal cavity pressures. As of 1992, one plant, Palisades, was operating with these seals. Operating experience indicates that these seals last approxhnately two fuel cycles, about 20 months of operation. In comparison, operating experience indicates that the SU seals last about I fuel cycle.

The Byrondackson N 9000 seal is designed for reliable operation for three fuel cycles (50,000 h). The design of this seal assumes that it will be operated in a clean system, without debris or crud in the RCS.

The N 9000 seals initially installed at Maine Yankee and Waterford 3 failed partly because of debris /ctud in the RCS. Apparently, these problems have been fixed. As of 1991, these seals have been installed on 3 pumps at Waterford 3,3 pumps at Maine Yankee, and one pump at Millstone 2 (Combustion Engineering 1991). Bell and O'Reilly (1992) reports that the operating experience with N 9000 seals, installed in late '

1989, at Maine Yankee has been good.

The Eingham Willamette seals are installed in San Onofre 2 and 3 and Calvert Cliff I and 2. The controlled bleed-off flow is 5.7 Umin (1.5 gpm). There are excess flow check valves on the controlled bleed off line, which will close as the controlled bleed-off rate reaches 38 to 57 Umin (10 to 15 gpm).

As of 1991, these seals had operated r.atisfactorily for four years at San Onofre 2 and 3 (Combustion Engineering 1991).

The AECL CAN4 seal is basically an SU seal modified by Atomic Energy of Canada in Chalk River. The modifications include several changes in the mechanical design to reduce seal friction and improve mechanical performance. The CAN4 seal is insensitive to temperature fluctuations and, therefore, should be able to handle the transients involving component cooling water. As of 1991, this seal has been installed in one RCP at Waterford 3.

5.4 Trends in Leak Events Caused by RCP Seal Degradation Eight leak events associated with RCP degrada' ion were reported during the 1985-1996 period, as listed in Table 5 2; three with Westinghouse RCPs (includes Oconee 1 pumps), three with Byrondackson pumps, and two with KSB yunps. The maximum leak rate was 152 Umin (40 gpm) during the ANO 2 event.

In this event, the initial leakage resulted from vibratory fatigue failure of the pump seal sensing line. This failure caused complete failure of the SU seal, as discussed in Section 5.2, and resulted in a small leakage into the contahunent (identified leakage). The leakage in the 7 other events was identified leakage and was less the the Technical Specification limit of 38 Umin (10 gpm). The leak rates in these 8 events were smaller than the charging pump capacities at the corresponding plants. Six of the eight events took place during the 1985-1990 period, whereas two events took place during the 1991 1996 period, it appears that this improvement in seal performance is partly due to the modifications discussed in Section 5.3.

5.5 Analysis of the 1995 St. Lucie 1 RCP Seal Leak This event has been selected for analysis even though the leakage through the failed seal was less than the allowable limit of 38 Umin (10 gpm) for identified leakage because the damaged seal could have degraded further and failed completely, resulting in a small-break loss-of-coolant accident. In addition, several problems, which were reportable according to the plant Technical Specifications, were ideuified during 5-5 DRAFT - NUREG/CR-6582

REACTOR COOLANT PUMP SEAL FAILURES this event and made the event more complicated. The event has ben analyzed in the USNRC Accident Precursor Program and its risk significance is discussed in Section 9.1. The analysis includes an event description, root cause analysis, list of corrective actions, and discussion on safety consequences of the event (Florida Power & Light 1995)."

5.6.1 Event Description On August 1,1995, both St. Lucie units were shut down in a controlled manner and cooled down to about 177'C (350'F) in anticipation of the predicted hurricane force winds from the passage of Hurricane Erin.

The maximum hurricane induced wind speed at the site was less than 45 mph, and the passage of the hurricane did not cause any flooding or wind damage to the plant. On August 2,1995, a decision was made to return both units to service. .

While Unit I was in hot shutdown (Mode 3) with a RCS pressure equal to 10.7 MPa (1550 psi), the pressure in the middle seal stage cavity of a RCP (RCP 1 A2) was approximately equal to the RCS pressure, '

indicating that the lower seal stage had failed. Frequency of monitoring the RCP parameters was increased in response to this failure.

Although plant operation with a failed lower seal stage is acceptable, it was decided to restage the failed seal stage, that is, to establish the differential pressure across the lower seal stage. While renaging the lower seal stage, it was teamed from control room indications related to controlled bleed off flow and seal cavity pressures that both the lower and middle seal stages had failed and the upper and vapor seal stages had degraded. This resulted in an increased identified leakage (controlled bleed-off leakage) rate of 7.6 Umin (2 gpm). The leakage rate was 0.95 Umin (0.25 gpm) before the event. The lower seal stage cavity presstne was quickly increased. Therefore, by the end of August 2, the operators began to cool down and depressurize the RCS to keep the lower seal temperature below 149'C (300'F).

Although the leakage rate was smaller than the Technical Specification limit of 38 IJmin (10 gpm) for identified leakage rate, there was potential for further degradation of the seal leading to increased leakage, So, an Unusual Event was declared and the state and the USNRC were notified, ne leakage rate decreased as the rector coolant system continued to depressurize. The Unusual Event was terminated about 9 hours1.041667e-4 days 0.0025 hours 1.488095e-5 weeks 3.4245e-6 months after it was declared.

Other/ncidents at St. Lucle. Four other incidents occurred during or immediately after this event:

(1) partial suspension of plant physical security safeguards, (2) unblocked main steam isolation signal, (3) failed power operated relief valves (PORVs), and (4) failure of a relief valve and subsequent unavailability of the shutdown cooling system. On August 1, plant management had partially suspended plant physical security safeguards because of potential personnel safety concems resulting from the impending hurricane, ne safeguards were fully restored after passage of the hurricane. The second incident involved a valid actuation signal for main steam isolation signal CASIS), which was not blocked as required by the procedure, llowever, all the valves that receive a MSIS signal were in their actuated positions prior to MSIS.

" Some of tne information presented in this section is from writing prepared for the Accident Precursor Program and provided by USNRC Technical Monitor Chuck lisu. The writing is based on IIR Nos. 335/95-004, -005, 006, for St. Lucie 1. This footnote will be replaced with a reference after the writing is pablished.

DRAFT - NUREGICR-6582 5-6

REACTOR COOLANT PUMP SEAL FAILURES The PORVs were stroke tested while the RCS was being depressurized and cooled down on August 3, but no increase in acoustical flow indication was observed. Both PORVs were tested again aher the pump seal was replaced and found inoperable, so the cooldown and depressurization of the plant began as required by Techn' cal Specific tion Limiting Condition for Operation 3.4.13.

When the unit was at 137'C (278'F) and 1.8 MPa (261 psi) during cooldown, the low pressure safety injection pump was started to place the shutdown cooling system in service to continue the cooldown.

Shortly thereafter, pressurizer level and letdown flow were found to be decreasing. Ilowever, there were no control room indications associated with RCS leakage nor was increase in reactor cavity sump flow detected. in addition, no leakage was observed in the low pressure safety injection pump room or in other auxiliary building areas. So, the plant operators assumed that the mismatch between charging and letdown flows was due to RCS cooldown. However, after several hours, water was found to be accumalating in the aualliary building pipe tunnel, so both shutdown cooling systems were secured. The pressurizer level and charging / letdown flow then became stable.

5.5.2 Root Cause Analysis ne root cause for the lower seal failure was under investigation but not reported in the LER. The preliminary root cause for the middle seal stage failure and degradation of the other two stages was an inadequate restaging procedure. His procedure had not been performed under the plant conditions existing at St. Lucie 1, which included RCS temperatures above 93'C (200'F) and a rotating RCP.

The root cause for the inoperability of both PORV valves was the improper installation of the main disc guides following an overhaul performed in 1994, in addition, there was inadequate post-maintenance testing before the valves were placed in service in December,1994. Since then, both PORVs were unavailable until their failed condit'on was identified ou August 3,1995.

The root cause for receiving conflicting indications for RCS leakage after placing the shutdown cooling system in service was closed floor drain isolation valves to the pump room sump. When these valves were subsequently opened, control room indications of high sump level were received. The leakage was from a relief valve located in low pressure safety injection discharge piping. The valve was common to both shutdown cooling trains. During the event, after starting the low-pressure safety injection pmnp, the operating pressure in the shutdown cooling system was above the lift-off pressure of the relief valve.

Because of the closed floor drain valves, discovery of the open relief valve was delayed for two hours.

As a result, about 15,200 litters (4,000 gallons) of reactor coolant was discharged into the auxiliary building.

5.5.3 Corrective Actions The pump seal was replaced. A seal on one other RCP at St. Lucie I was also replaced because of its degraded performance in addition, the licensee had scheduled installation of the Byrondackson N 9000 seal on one RCP during the next outage to evaluate its performance with the intention of going to this seal if upgraded performance could be achieved (USNRC 19%).

After the relief valve leak was identified, both trains of the shutdown cooling system were removed from service for 20 h to replace the valve. Reactor coolant temperature was increased to 152'C (305'F), and 5-7 DRAIT NUREG/CR-6582

REACTOR COOLANT PUMP SEAL FAILURES ,

decay heat was removed using the steam generator, which was the only source of heat removal at this point. After the relief valve replacement, both shutdown cooling systems were placed in operation, and the RCS was cooled down and depressurized to repair the PORVs.

5.5.4 Safety Consequences !

Safety cotsequences are associated with three different parts of the event and incidents described in tids section: (1) two failed RCP seal stages, (2) unavailability of both PORVs, and (3) unavailability of the safe shutdown cooling system. These problems are analyzed in the USNRC Accident Sequerre Precursor Pro 3 ram and are briefly discussed in Sc. tion 9.1. The RCP seal could have degraded further and if the degradation had catned a failure of the entire seal, including the vapor stage, it could have resulted in a small-break LOCA. .

The PORVs were unavailable for about 5,880 h for both pressure relief and for feed and bleed. So, the pressurizer code safety valve would have been demanded in the event of high RCS pressure, inasmuch as the safety valve cannot be isolated, failure of an open safety valve to close would lead to a small-break LOCA.

When the shutdown cooling system was removed from service for repair of the failed relief valve, the only source of decay heat removal was the steam generators. Feed and bleed was not available because of the failed PORVs. If both motor. driven auxiliary feedwater pumps had failed RCS heatup would allow use '

of the turbine-driven auxiliary feedwater pump. So, the associated risk with this incident was small, and, therefore, this event was not further analyzed in the Accident Sequence Precursor Program.

5.6 USNRC Generic issue 23 The Reactor Safety Study, WASH 1400, published in 1975, indicates that breaks in the reactor coolant pressure boundary with an equivalent diameter of 12.7 to 51 rnm (0.5 to 2 m.) could occur with a frequency of 10-8 per reactor year and were the largest potential contributor to PWR core melt. This estimate did not take into account the availability of makeup flow. A 1980 USNRC review of RCP seal failures experienced at operating U.S. PWRs revealed that the pump seal failures, with leak rates equivalent to those of small-break LOCAs, occurred at a frequency of 108 per reactor year, which is an order of magnitude greater than the pipe-break frequency used in WASH 1400. This huplies that the overall probability of core melt resulting from small-break LOCAs could be dominated by the pump seal failures (Jackson 1988). Because of these concerns, in 1983 the USNRC initiated a review of RCP seal failures -

as Generic Issue 23, Reactor Coolant Pump Seal failures. Another related project, Generic Issue 65, Probability of Core Melt Due to Component Cooling Water System Failures, wns incorporated into the task action plan for Generic issue 23 because of its close relationship. .

The purpose of Generic Issue 23 is to evaluate the current licensing requirements related to RCP seal integrity and to determine if further NRC action is necessary to ensure that the pump seal failures do not pose an unacceptable risk. RCP seal failure experience prior to 1985 was analyzed in support of Generic Issue 23. All three types of pumps were included: Byron-Jackson, Westinghouse, and Bingham-Willamette pumps. After several years of study, the USNRC proposed the following three resolutions for resolving Generic Issue 23 and issued them for comments in 1991 (Federal Register 1991):

DRAFT - NUREG/CR-6582 5-8

REACTOR COOLANT PUMP SEAL FAILURES

1. Treat the RCP seal assembly as an hem, performing a safety related function similar to other components of the reactor coolant pressure boundary, applying quality assurance requirements consistent with Appendix B of 10 CFR 50 and applicable General Design Criteria of Appendix A of 10 CFR 50.

2, Provide the instmmentation and instmetions for monitoring RCP seal performance and for detecting incipient RCP seal failures as recommended by RCP seal manufacturers.

3. Provide RCP seal cooling during off normal plant conditions involving loss of all seal cooling such as station blackout.

ne Combustion Engineering Owners Group (CEOG) did not support the above three resolutions. The Owners Group surveyed its members to obtain current information on reactor pump seal design and performance history. Their conunents on the three resolutions are summarized as follows:

l. The seals are being handled with the highest care to ensure that the risk of failure has been minimized, nerefore, additional quality assurance requirements will not result in improved performance. The seal materials used in the Byron Jackson pumps have been selected to perform the design function.

These materials are not necessarily listed in ASME Section III as acceptable pressure boundary materials.Section III deals with the pressure boundary materials in vessels, pumps, and valves but specifically excludes seals (Combustion Engineering 1991).

2. De CEOG believes that the existing instrumentation is sufficient for monitoring pump seal performance and the additional instrumentation proposed in Resolution 2 would not add additional confidence. The instmmentation recommended by the RCP manufacturers for monitoring the controlled bleed-off flow and temperature are in use in the CEOG plants. So, the intent of resolution has already been met and the resolution is unnecessary (Combustion Engineering 1991).

3. The CEOG believes that the addition of backup seal cooling, as recommended by Resolution 3, is not necessary because it has been demonstrated that the seals can survive station blackout conditions. The controlled bleed off is isolated during a station blackout in order to maintain the RCS inventory. This eliminates flow of hot RCS water up through the seal assembly and prevents seal heatup except by shaft conduction. Tests have shown that the temperature of the vapor seal may be as much as 111*C (200'P) lower than the temperature of the lower seal stage even after 8 h into the station blackout event (Combustion Engineering 1991).

He CEOG also pointed out that the proposed resolutions would not have prevented the complete failure of the SU seal, including failure of the vapor seal stage, that took place at ANO 2, August 1988, because that failure was caused by a fracture of a pressure sensing pipe (Combustion Engineering 1991).

5-9 DRAFT NUREG/CR-6582

REACTOR COOLANT PUMP SEAL FAILURES A draft rule that would resolve Generic issue 23 was proposed on August 24,1994, for public comments.

On March 31,1995, the USNRC Cn-mlaaion concluded that the proposed rule did not provide sufficient gain in safety to justify its h===> and decided against its publication. The USNRC staff is currently evaluating different options for the final disposition of Generic Issue 23 (UShTC l')95).

e d

e t

i e

?

DRAFT NUREG/CR-6582 5-10

6, Analysis of Leak Events Associated with SCC" i

SCC cracking has caused leakage through base metal and welds of several primary pressure boundary components worldwide. Le crteking has been reported in small-diameter piping and penetrations. In

, some instances, the throughwall cracks had circumferential orientation, suggesting a possibility of complete rupture of the component. The leakage rates through the cracks were small, as listed in Table 21, and detected by visual inspection. Review of field experience in the U.S. from 1985 through 1996 indicates that the SCC problem is not widespread in the PWR primary pressure boundary components; 16 out of 199 ,

reportable leak events were caused by SCC, as shown in Figure 216. l l

Three different SCC mechanisms are active in the PWR primary pressure boundary components: primary l water SCC in Alloy 600 components and hOrgranular and transgranular SCC in austenitic stainless steel piping and penetrations. The affected systems include instrument penetraticns in pressurizers and main coolant piping, control rod drive mechanism (CRDM) [ called control element drive mechanism (CEDM) in Combustion-Engineering designed PWRs] nozzles and housings, reactor coolant pump seal injection lines, steam generator channel head drain lines, and residual heat removal lines.

SCC has caused jeaks at locations previously not considered to be susceptible. One such event vias leakage through the base metal of a spare control element drive mechanism housing at Fort Calhoun in 1990. The cracking was caused by transgranular SCC. This event is analyzed in this section.

The section is in three paru. First, a brief description of the transgranular SCC mechanism is given. l Second, the trends and distribution in the leakage events associated with the SCC mechanism are discussed. ,

l Third, the Fort Calhoun event is analyzed, l

6,1 Transgranular SCC Mechanism Occurrences of transgranular SCC (TOSCC) have been experienced in regions of stagnant coolan: at high 4

elevation, where oxygen levels can become high. A detailed description of the leakage associated with a

~

cracked PWR CEDM housing at the top of the reactor vessel head in a Combustion Engineering-designed plant is presented in Section 6.3. TGSCC also has been found in a similar location in Westinghouse, designed PWRs. In these designs, canopy scal welds form part of the pressure boundary between the reactor vessel head and head penetration attachments such as CRDMs. Instrument port latch housings, ,

female flanges, upper probe housings, and head adapter plugs (Pezze and Wilson 1989). Leakage through the lower canopy seal region has been reported at several plants. De cracks initiated at weld fusion areas, weld inside-diameter areas, and at base-metal inside<hameter areas. Although weld defects were present, the transgranular cracking appeared to occur independently of the defects. The base metal was 3M austenitic stainless steel and the weld metal was AWS Type 308L ne combination of the susceptible material, the corrosive environment (high oxygen with chlorides present), and high tensile residual stresses from the welding process caused the TGSCC. Oxygen in the ' dead-end cavity" formed by the canopy seal can be trapped in the area during startup as the reactor vessel is being filled, or it can be introduced by a lack of chemistry control in the coolant during cold conditions. Pezze and Wilson (1989) state that

" Stress corrosion cracking of steam generator tubes is not within the scope of the project and therefore not addrened in this section.

6-1 DRAIO NUREG/CR-6582

__ __. -_ ._- ._ _ _ . ___ ._ . _ . _ _ . _ ~ _ . _ .

STRESS CORROSION CRACKING materials such as Alloy 625 or 690 could be used to replace the Type 304 stainless steel to make the canopy seal less susceptible to TGSCC. Better chemistry control and a hole drilled through the head adapter pipe to eliminate the dead-end cavity were also mentioned as other possible mitigation methods.

l 6.2 Trend in Leak Events Associated with SCC Failures ne yearly distribution of SCC failures from years 1985 through 1996 is shown in Figure 217. Of the 16 SCC leak events, less than 1/3 (5 events) occurred during the first 7 years of the 12 year period, whereas over 2/3 (11 events) occurred in the final 5 years of the period. During 1985 through 1991 the leak event rate was less than 1 eventlyr, but from 1992 to 1996 the leak event rate was over 2 events /yr. Although there is not many failure data because the rates are comparatively low, the trend in leak rates could be interpreted as increasing. This would be consistent with an aging mechmiems such as SCC, which is time- .

dependent, with more failures occurring with age.

From Table 21, the averaFe plant age at the time of TGSCC leakage (3 events) was about 20 years, .

whereas the average plant age at the time of PWSCC leakage (10 events) was 10 years, or about % the time. The average plant age for the three other events [2 intergranular stress corrosion cracking (1GSCC) and 1 SCC) was about 19 years. In all cases, the leak rates were low, most were less than 0.38 IJmin (0.1 gpm), and the highest was 1.5 IJmin (0.4 gpm).

In some instances (for example, Surry 1 pressurizer), the crack was found during planned inspections during an outage, whes cas in other instances (for example, Fort Calhoun CEDM), the plant operations staff detected a small leakage rate during plant operation. The presence of boron crystals in the area of leakage assisted the plant staff in locating the leak; the small crack sizes are too difficult to detect visually.

PWSCC occurs in inconel alloys, whereas 1GSCC, SCC, and TGSCC occur in austenitic stainless steelt.

The instances of IGSCC, SCC, and TGSCC generally occurred in welds or heat affected zones. This can be attributed to sensitization in the case of IGSCC and high weld residual stress in the case of TGSCC.

Some cracking in Westinghouse PWR canopy scal welds occurred at base-metal sites. However, only one of the 10 PWSCC leak events occurred in welds.

The cracking damage was local. Replacements were made in most cases, sometimes with an altered design. For example, in the case of the Fort Calhoun CEDM, the component was renlaced with a simpler design (CEDM housing replaced with blind flange). In other cases, pressurizer instrument nozzles, repairs were made. .

6.3 Analysis of the 1990 Ft. Calhoun Leak Event Caused by Transgranular SCC The Fort Crlhoun nuclear plant is an older vintage Combustion Engineering PWR plant that attained initial criticality in 1973. The CEDM housings, used for reactivity control, protrude above the reactor vessel head where they are contained by housings that fonn part of the primary pressure boundary. Each housing is mounted on a nozzle that is welded to the reactor vessel head. Here are a total of 41 identical CEDM housings, of which 37 are active in that they contain CEDMs that attach to the neutron-absorbing control element assemblies. The remaining four housings are considered spares, and were designed for possible future use. Two of the four (numbers 7 and 11) are now being used to house the Heated Junction DRAFT - NUREG/CR4582 6-2 l

N -w W

STRESS CORROSION CRACKING !

Thermocouple (IUTC) probes, whereas the remaining two (numbers 9 and 13) contain only irsernal natural circulation spoiler assemblies, which are considered to be essemiallypanh't. All housings are subjected j to high temperature [204'C (400'F)] and pressure [14.5 MPa (2100 psi)) at power.

ne CEDM housings are fabricated from SA 182 and SA.312 Grade 347 or 348 stainless stect. Each CEDM housing is omega-scal welded to the nozzle flange and then bolted to the nozzle flange wkh eight threaded studs (Figure 6-2). The flange is welded to a pipe, which is welded to the reactor vessel head.

He 37 active CEDM totsings are self vented through rotating mechanical seals in the CEDM seal housing. These seals allow air to vent during RCS fill and during CEDM operation. Also, when CEDMs are operated, there is an irserchange of coolant between the housing and the temainder of the RCS. he

, two housings with }UTC probes have been manually vented to removed trapped air since the probes were installed in 1984, though the venting procedure may have been inadequate. The remaining two passive housings were nevet versed; plant personnelt iid not know why venting had not been incorporated into plant procedures.

J After 17 years of plass operation, a leak developed in one of the spare housleg. Detatis of the leak event, the root cause analysts, corrective actions, and the safety significance are described below.

6.3.1 Event Description On Oct:>ber 21,1990, during full-power operation, leakage from an unknown source was identified and quantified at 0.4 to 0.8 Umin (0.1 to 0.2 gpm) (OPPD 1991). Between initial discovery and December 14,1990, the leak rate increased to and stebilized at about 1.5 Umin (0.4 gpm). On December 14,1990, the reactor was placed in hot standby mode to look for the leak, which was thought to be in the vicinity '

of the reactor vessel head. The inspection revealed a leak coming from one of the two passive spare CEDM housings (housing number 9). De reactor was then placed in cold shutdown mode for funher investigations and repair.

The leaking housing wu removed, and axially oriented cracks were identified on the inside diameter of the pressure housing, one of which had penetrated though the housing wall. The cracking was localized

, in a weld overlay area, which exists on all housings to provide positive positioning of some housing internals. Visual examination of the other passive housing also revealed axially oriented cracks in the same area. Three-quarter-meter (2.5-f:.) sections of both passive housings were cut out for detailed metallurgical analyses.

6.3.2 Root Cause Analysis Only two crack-like indications were found in each of the two housiags, as follows:

Housing number 9 (1) throughwall crack 71-mm (7.8-in.) ID and 19-mm (0.75 in.) OD (2) second crack 85 % throughwall 58 mm (2.3 in.) long 63 DRAFT - NUREG/CR-6582

STRESS CORROSION CRACKING ,

Ilousing number 13 (1) crack 70% throughwall 56 mm (1.9 in.) long (2) crack 95 % through wall 69-mm (2.7 in.) long l All four cracks had aspect ratios (ID length to depth) in the range of 3.6 to 4.0. All cracks initiated at the ID, just outside the upper edge of the weld overlay region in the heat affected zone. The cracks then propagated outward in to the wall, extending nearly symmetrically down through the a eld overlay region and up into the base metal. Although the cracks were nomhally axial, some were skewed off axial approximately 15 degrees. The fracture surfaces contained distinct rings, which could be correlated to plant shutdowns (when oxygen levels and temperatures changed), indicting that crack initiation occuned between 1981 and 1984 (1983 is the best estimate). When the plant is shut down for maintenance, the primary coolant is drained, and thus the mechanism for TGSCC is removed. When the plant is refilled, ,

a more oxygenated environment is available.

SEM of the crack surfaces and metallographic examinations showed that the cracking was TGSCC. De ,

types of austenitic stainless steels from which the housings were fabricated (SA 312 Type 348 base metal and SA 298 E348-15 weld overlay) are known to be susceptible to TGSCC (but not IGSCC) when exposed to adverse enyhonmental condition in the presence of high tensile stresses. These conditions are described in the following two paragraphs. There was no sensitization introduced during the weld overlay process, nor is sensitization required for TGSCC.

To produce TGSCC, an environment with high oxygen levels and some halogens (for example, chlorides) must be present. Very low concentrations of chlorides can produce TGSCC when the oxygen level is sufficiently high. When the chloride level is less than 0.15 ppm, the oxygen level required to produca TGSCC is 4 to 8 ppm. Since the spare CEDM housings (numbers 9 and 13) were not vented, the oxygen levels were estimated to be between 300 and 1300 ppm. In contrast, tLe vented CEDM housings are estimated to have oxygen and chlcride concentrations, both,5 ppb.

When short sections of the housings containing the indications were removed and cut longitudinally, the outside diameters decreased by 0.51 mm (0.020 in.) on housing number 9 and 0.58 mm (0.023 in.) on housing number 13. This decrease in diameter was attributed to residual stresses in the housing resulting from the veld overlay process. The corresponding residual stress was calculated to be on the order of 69 MPa (10 ksi). The tensile hoop stress contributed an additional 71.7 MPa (10.4 ksi), which resulted in a total tensile stress in th: weld overlay area greater than 140 MPa (20 ksi). When a similar longitudinal cut was made on a hous!ng section away from the weld overlay area, the measured diametrical ch1nge was only 0.04 nun (0.0015 in.). Thus, the residual stresses in the base metal away from the weld overlay were low. The yield stress of the housing material was 205 MPa, with an ultimate tensile strength of 515 MPa (Lisowyj 1993), so the total tensile stress was equal to about 70% of the yield stress at temperature. He crack branching in the metallographic analysis was moderate, indicating a moderate stress intensity at the crack tip. No weld!ng defects were observed.

Therefore, the root cause of the TGSCC and leak was the stagnant, oxygenated environment in contact with a s isceptible material under high local residual tensile stress. Studies estimate that the cracking began after about 10 years of reactor operation.

DRAIT - NUREG'CR-6582 6-4

STRESS CORROSION CRACKING 6.3.3 Corrective Actions The two spare housings were removed from the vessel head, visually inspected, sectioned, and sent to Alla-Combustion Engineering for destructive examinadon. These two locations were capped by CEDM blind hanges, which were leak tested during the following plant startup. The blind flanges had an metal Oeing seal design rather than the origina' omega-scal weld. The spoilers that inhibit oxygen removal were removed from 6e two spare housings.

Visual inspections were made of the two adjacent housing to determine if steam knpingement damage was present. No damage was found. Visual and UT inspections of the two spare housings containing the IUTC probes showed no SCC damage. Visual inspections also were made of six other (active) housings, but no SCC damage was detected. The licensee concluded that, due to their self venting features through mechanical seals, the 37 active CEDM housings are not susceptible to TGSCC. Examination and evaluatioru of instrument port housings on the reactor vessel head, which may be similarly affected by TGSCC were made; however, no e ridence of damage was detected.

Procedures were revised to vent the two IUTC housings after the reactor coolant pumps are started and the reactor coolant pump seals are vented. This will ensure that any air bubbles that may become trapped in the IUTC housinEs when the reactor coolant pumps are started will be removed. Continuing operation with the unvented blind flanges will be evaluated on a cycle-by-cycle basis. Periodic inspections of all the housing will be undertaken with UT techniques capable of detecting SCC. In addition, an enhanced RCS leakage toonitoring program was implemented.

C.3.4 Safety Significance The spare housings constituted the primary pressure boundary. Since the leakage rate from the throughwall crack was small and the cracks were oriented axially, catastrophic rupture of th: pressure boundary would not be expected.

Steam sprayed from the leaking crack could impinge on the two adjacent CEDM housings. This could potentially intubit their reactivity-control function, llowever, visual inspection showed no steam-impingement damage.

When primary coolant containing boric acid is sprayed on the reactor vessel head, boric acid-induced corrosion can result. Although a large amount of boric acid residue was found on the vessel head, inspections of the adjacent CEDMs and a few locations on the reactor vessel head revealed no damage.

6-5 DRAFT - NUREG/CR-6582

-%-e e..-- A ---~e-AJ-AAh--.4cA- W e a-- * - - 4 .42 4+.e . A J ---,2L Ah A2+AaK,._M4J4+-.4-Wea'*es- - weC+ - - +--4--.-M-444 -A~2--4-- - uW--'1-f 1

1 1

l 1

e ,

e i

h I

l i

i 5

e 6-6 F

,s...,-- ,.m~ ...-,.-m, ., --

7. ANALYSIS OF LEAK EVENTS CAUSED BY VALVE STEM PACKING DEGRADATION 4 Our review of the repertable leak events during the period 1985 through 1996 indicates that valve stem packing degradation was one of the dominant mechanisms causing PWR primary coolant system leakage during 1985 through 1991, but was not a problem thereafter; no reportable packing degradation related leak event has taken place since 1991, These earlier events had impacted overall plant performance and maintenance. Possible reasons for the improved packing performance are discussed in this section.

The original (as installed) asbestos-based valve stem packing design is described first, and the characteristics responsible for the packing leakage are identified. Den the industry efforts in making valve stem packing

,, improvements are surnmarized. The main improvements are related to packing material and application of live-loading to provide continual inservice adjustment of the valve packing. Dese design descriptions are expected to facilitate the discussions on the trends in thc ,1. king degradation and the analyses of a selected

. leak event that follow.

7.1 Valve Stem Packing Design Figure 71 shows a typical, asbestos based valve stem packing design originally used in nuclear power plants. He relevant design characteristics are the deep stuffing box filled with 8 or more packing rings of braided asbestos. If a valve is designed with an active leak-off line, then it uses both upper and lower packing sets and a lantern ring, as shown in Figure 7 2. System pressure is acting at one end of the packing ring stack, and an axial compressive force applied via a gland follower by tightening the gland bolts is acting at the other end. He axial force generates radial pressure between the packing and the stem and between the packing and the numng box. This radial pressure has to be equal to or higher than the system pressure to provide effective sealing of valve stem leakage, i

Asbestos based packing for valves for elevated temperature service have been the industry standards for a long time. Valve stem packing materials and stumng box design for nuclear power plants were largely adapted from fossil plants. }{owever, the levels of stem let kage conddered acceptable in fossil plants are not acceptable for many systems in nuclear power plants. Often, the repair of the valve packing required load reduction or forced outages. For example, our review of L'Rs shows that 13 e of 32 packing related, reportable leak events caused plant shutdown during the 1986 '996 period. T main reasons for the unacceptable leakage rates through the valve packing are stuffing be design and peking materials.

Earlier, there was a misconception that deep stumng boxes filled with pacLag ringc provided better control of valve stem leakage it was assumed that the scaling mechanism between the valve stem and the valve packing was produced by a ser;es of pressure breakdowns, similar to the labyrinth sul design, it was thought that with each packing ring the system pressure was reduced, and so the larger number of rings could totally seal off the leak path (llart 1996).

Ilowever, research has shown that deep stuffing boxes are detrimental to effective packinC performance.

De degraded packing performance may be explained by considering the stress distribution in the stumr.g bx shown in Figure 7 1. De axial compressive stress decays rapidly from the packing to-gland follower interface to the bottom of the stufling box, attributable to the higt nemeient of friction between the packing 71 DRAT ~r - NUREG/CR-6582

VALVE STEM PACKING DEGRADATION i

and the stem and between the packing and the stumng box wall. He resulting high friction forces act on the packing along the axial direction opposing the applied compressive force. As a result, most of the compressive load is transmitted to only the upper packing rings, which provide most of the sealing. He lower ring are ineffective in providing sealing. Significantly higher axial compressive force has to be applied by way of the gland follower to make the lower rings effective, llowever, the resulting higher friction forces require higher stem thrust to open or close the valves. This is more relevant for a small diameter valve than a large diameter valve because packing drag (friction force)in a small diameter valve constitutes a larger percent of the total stem thru>t. He higher friction forces can render a MOV inoperable (Cepkauskas and Garcia 1994). The high friction forces will also cause wear of the packing materialin service. The high gland force is also required to reduce the permeability of asbestos. Bralded asbestos packing rings contain fillers to reduce its permeability.1lowever, they still require a high gland force to reduce permeability sufficiently for effective sealing (Ruggieri and Kelly 1988). ,

ne asbestos packing material consolidates during service because of the loss of water of crystallization and evaporation of volatile chemicals present in the filler material. His will result in a loss of packing

compressive stress, eventually leading to valve leakage when the radial stress at the stem and stuffing box wall falls below the system pressure. To reduce the leakage, the gland bolts are further tightened to provide higher radial stresses, and eventually the valve is repacked. Hus, the deep stuffing boxes are not only ineffective but detrimental to good packing performance because they allow for larger consolidation of the packing rings.

The asbestos packing material looses its fiexibility and elasticity as it dries out and becomes hard during service. Ilowever, the valve packing must remain flexible so that it can accommodate changes in the stem diameter as it is stroked through the stuffing box and provide effective stem sealing. Valve packing research in the 1980s identified the stem site changes as a major challenge to valve stem sealing. Valve stem diameter changes because it is exposed to hot reactor coolant when the valve is closed and to the cooler extemal environment when the valve is open. So, the stem diameter is smaller when it is driven through the stuffing box to close the valve than when it is pulled out to open the valve. Once the packing material looses its flexibility and elasticity, it cannot provide the dynamic stem sealing required during stem stroke (Hart 1996).

7,2 Valve Stem Packing Improvements Several factors affect packing performance. Many of these factors, such as RCS pressure, temperature, and chemistry, are determin;d by the system design and are therefore fixed. Some other factors, such as stuffing box depth and dir, meter, necessitate valve removal and replacement. However, there are two factors that can be modified without expensive valve replacement. nese factors are packing composition and gland load.

Feur specific valve stem packing improvements have been made for reducing the valve leakage: (1) changing the packing material for effective sealing,(2) reducing the stuffing box depth to decrease inservice consolidation of the packing rings,(3) employing improved stem packing configuration for more efficient transfer of axial gland loads into radial sealing pressures, and (4) providing means of continual gland adjustment to maintain sufficiem axial gland load for long-term sealing effectiveness. These improvements are practical because they can be retrofitted. Replacement of valves is not required.

DRAFT - NUREG/CR-6582 7-2

VAIME STEM PACKING DEGRADATION Laboratory experiments and field use have shown that graphite is far superior to asbestet for effective sealing. Some of the favorable characteristics of graphite are its high density, impermeability to gases and fluids, low soefficient of friction, no loss of flexibility during service, and less consolidation in service. High density and impermeability provide sealing at a much lower gland loading than that required by asbestos.

Die formed graphite rings accommodate changes in the stem diameter as it is stroking thrsugh the stuffing box during service because the packing remains flexible. In addition, graphite packing sequires fewer in-service adjustments because it consolidates less.

In the packing design with graphite, a carbon spacer is placed at the bottom of the stuffing box so that the effective depth of the stuffing box is redeced, nis requires fewer packing rir.gs, so the ineffective lower packing rings are eliminated. The use of fewer rings reduces consolidation in service. In addition, braided end rings are placed at the ends of the packing ring stack, as shown in Figure 7 3 He braided rings prevent extrusion of the die fonned giaphite rings and act as a wiper to prevent the buildup of graphite particles adhering to the valve stem as the stem cycles. If a valve is designed with an active leak offline, then a lantern ring with a braided ring at each end is added to the graphite ring stack. The graphite packing is being used in U.S. nuclear power plants (VanTassell 1994). He cross section of the packing ring was redesigned from a square section to a wedge section, shown in Figure 7-4, so that the axial fcrees en the packing can be rnore efficieritly converted into radial scaling forces. Thus the use of graphite as a packing material having a low-coefficient of friction and packing rings with wedge section reduces packing drag (friction force) on the valve stem.

A small amount of consolidation choses significant reduction in gland load, so inservice adjustment of the gland load is required to control valve leakage. Often, adjustment of gland loads may require had reduction or forced outage. Contintal adjustment of gland load with live loading application can solve most of the leakage problems. Figure 7 5 illustrates the concept of live loading in which disc springs are installed between the gland stud puts and the gland follower. When the gland nuts are tightened to load the packing,

, the springs are compressed. As the packing consolidates in service, the springs expand to maintain a relatively :.onstant !oed on the packing. Thus, the live-load appliution provide continual in service adjustment of the gland load.

De Hve loading application was developed by AECL and has been successfully used with braided asbestos packing in Canavic.n welear power plants for the last 20 years or so. De live loading application was initially implemented because of the high cost of heavy water. However, its use has resulted in significantly more savings because of higher availability, reduced maintenance, and less personnel exposure. For example, in the Canadian plants, use of live loaaing with asbestos packing has reduced the frequency of repacking by at least a factor of 20 and frequency of gland lord adjustment by a factor of 45. He leakage rate was reduced by a factor of 10 to 40. When live loading is used with either asbestos or graphite packing, the frtquency of repacking and gland adjustment are small and equal to 0 or 1 per 100 valves per year, but the leak rate with the graphite packing as compared to the astestos packing is smaller by a factor of 50 (Ruggieri and Kelly 1988).

U.S. utilities have been hesitant to incomorate valve packing improvernents in tteir plants because of the misconceptions about valve packing performance based on the histo:ict.1 experience prior to the late 1980s (Hart 1996). However, most of the utilities have replaced asbevos packing with graphite packing and a spacer for health snd safety reasons. He replacement packMg system includes a set of five graphite packing 73 DP AFT - NUREG/CR-6582

VALVE STEM PACl3NG DEGRADATION rings with a square cross section and a carbon or stainless steel spacer. " Some utilities have also incorporated the live. loading application. The implementation of the pack.ing improvements has resulted in an absence of packing related reportable leak events at U.S. PWRs since 1991. For example, a graphite packir,g program was implemented at Susquehanna Units 1 and 2 in 1985. To date, about 22,000 valves have been repacked with graphite packing systems, and as a result packing leaks are virtually nonexistent. He success of the repacking program can be attributed to the amount of training and support given to repack crews (VanTassell 1994).

7.3 Trends in Leak Events Caused by Valve Packing Degradation Distribution of the reportable leak events by calendar year is shown itJigure 7 6. Packing degradation was a dominant cause for leakage during the 1985 86 period. About 35% of the reportable leak events were ,

associated with packing degradation during this period. Ilowever, there were three or less reportable leak events per calendar year during the next Ove years (19871991 period), and no events associated with packing degradation since 1991. A distribution of frequencies of these leak events by calendar year is also ,

show in Figure 7 6. A frequcacy for a given year is a ratio of the number ofleak events and operating years for the U.S, pWR plants auring that year.

Distribution ofleak events by plant age rather than calendar time is shown in Figure 7 7. Here, age is defined as the time from initial criticality; a plant is of age n ifit is in its n

year of operation. A distribution of the frequencies of the leak events by age is also shown in Figure 7 7. A frequency for a given age is a ratio of the number ofleak events that occurred in the PWR plants parsing through that age during the 198$

to 1996 period and the number of plants associated with that age. He number of plants that pass through a given age is equal to or smaller than the total number of pWR plants operating during the 1985 to 1996 period. Note that the frequency data for the older plants (age 24 to 30) are based on less than 10 plants (small sample sire). Leak es ents per operating year, instead of operating plant, can be determined if data related to operating year for a given age are estimated.

Trends of the packing degradation-related leak events in calendar time and plant age were statistically investigated as shown in Figures 7 8 and 7 9, respectively. Details of the investigation are presented in Appendix C. The effect of calendar time reflects the evolving body of regulations, design improvements, and indestry wide learning. The effect of plant age reflects the teaming of plant personnel and the aging of the hardware. Solid curves in Figures 7 8 and 7 9 show the estimated trend in calendar time and with age, respectively. The dotted lines show 90% confidence bands. A trend in this investigation, if present, is assumed to have an exponential form. He data are consistent with this modeling assumption, because all point estimates of 90% confidence intervals overlap the confidence bands. The results in Figures 7 8 and 7 9 reveal that the trends both in calendar time and age are decrea-ing, but the trend in calendar time is statistically more significant than in age. Apparently, the decreasing trend is a result of vah e stem packing improvements made in the early 1980s and incorporating some of these improvements in operating nuclear power plants.

15 V. N. Shah, private communication with K. A. Hart, Pennsylvania Power & Light Co.,

Pennsylvania, Septemtier 30,1997.

l DRAFT NUREG/CR4582 7-4

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

VALVE STEM PACKING DEGRADATION ,

7.4 Analysis of 1991 North Anna 2 RHR lsolation Valve Packing Leak ne main reason for analyzing this specific event is that it illustrates the perfonnance probtems related to the original asbestos-based valve packing and the utility's efforts in incorporating the valve stem packing improvements discussed in Section 7.2. ne analysis includes an event description, root cause analysis, safety consequences, and corrective actions taken by the utility.

Event Descript/on. On November 3,1991, an isolable leak of 94.61/ min (24,9 gpm) was identified at North Anna 2 while it was operating at 100% power. The leak was identified by an increased containment sump pumping froquency. De increased makeup flow to the RCS was also nuticed. A containment entry team identified that a motor operated isolation valve [a 3 % mm (14 in.) gate valve] in t' e RHR system had

, elevated leak offline temperatures [166'C (330'F)). He RCS pressure was then reduced to 2.8 Mpa (400 psig), and the valve was back-seated to stop the leak. Subsequent leak rate calculation confirmed that the leak was from that valve (North Anna 1991).

Root Cause Analysis. The root cause ofleakage was probably insufficient compressive loads on the packing rings placed at the bottom of the deep stuffmg box. There were 8 packing rings below the lantern ring and 4 rings above. When the failed packing was removed from the isolation valve, only small fragments of tha packing rings located below the lantern ring were found, nis implies that the packing rings were made from asbestos, became hard and inflexible, and consolidated during operation. Eventually, these lower rings lost elasticity and broke to pieces, ne large number of pacl:ing rings and their consolidation during operation made the lower rings inefrective in sealing the RCS leaks, as discussed earlier in the section (North Anna 1991).

Corrective Aetions. The short term corrective actioa included replacing the damaged valve packing after the valve was backscated with packing having a similar design. Then, the valve was successfully tested according to the inservice testing program, ne electrical department measured stroke traces and determined that the valve was fully operable (North Anna 1991).

Plant maintenance considered some of the improvements discussed in Section 7.2 as Impterm corrective actions to prevent recurrence of the packing leak because the subject vdve had been repacked at least every other refueling. The plant maintenance staff was well aware of the valve stem packing improvements discussed above. So they consulted a packing and seals vendor for specific recommendations. Rese recommendations included installation of a spacer bushing to replace the three lowest packing rings below the lantern ring and live loading of the packing gland. The licensee decided to install bushings in the RHR inlet isolation valves at both units during the next outages and determine the adaptability for live loading prior to completion of the next Unit 2 refueling outage (North Anna 1991). Based on Figure 7 6, no

, reportable leak has been observed at North Anna 2 since 1991.

Safety Consequences. The reactor coolant leakage was through the valve leakoff line and, therefore, it was a controlled leakage. There was no release of radioactive materials to the environment. In addition, the increase in charging flow to maintain RCS inventvy was well within the capacity of the charging pump that was in operation. So, there were no significam safety consequences from th'.s event (North Anna 1991).

75 DRAFT - NUREG/CR-6582

VALVE STV.M PACKING DEGRADATION 7.6 Industry Programs to Manage Packing Leaks Severs! U.S. utilities have developed maintenance programs for preventing reactor coolant leaks caused by packing degradstion. For example, a three phase program is being followed at McGuire plants." He program focuses on the RCS valves, nese valves are classified as criticalvalves because their failure may jeopardize continued plant operation. He nrst phase of the program identifies these valves. Inasmuch as valve packing cristitutes the primary pressure boundary, three of the criteria to identify the critical vah es refer to packing leakage:

(1) the leakage would be :lassified as unidentified leakage per Technical Specifications, OR .

(2) the leakaga would be classified as identified but would challenge cooling of collection tank, AND (3) epair of the valve based on its function and physical location and required post maintenance testing would force a un:t power reduction or shutdown.

He s:cond phase of the program involves a review of critical valve packing histories, ne third phase of the program involves monitoring critical valve packing leak-offs during walk-downs. His monitoring is l done with a temperature device for thermally Pot valvn and with valve stem leak off collection valves for those where coolant is cold. P eview of the monitoring results can reveal emergent packing leaks before they progress and cause operational problems, I

l .

l

" V. N. Shah, private commt'nication with Mike Rains, McGuire Nuclear Station, Duke Poiver, June 25,1997, DRAFT - NUREG/CR-6582 7-6

h

8. ANALYSIS OF LEAK EVENTS CAUSED BY COMPRESSION '

FITTING FAILURES Compression fittings represent a simple method of attaching different diameter tubing. neir simplicity makes them attractive to constructors, but their propensity for loosening and thereby becoming dir, connected has led to several events myolving larger than allowable plant leakage, which has resulted in plant shutdowns, ne leak es ents occurred in small instrument line tubing, which was subjected to a high temperature, high pressure, vibratory environment. Inadequate tightening of the fittings was the predominant root cause of the leakage, nis section presents the basic design principler of compression fittings, the trend in leak events caused by compression fitting failures during the period 1985 1996, and an analysis of a leak event associated wifh a compression fitting fallere at the Oconee 3 nuclear station in 1991.

8.1 Compression Fitting Design Instrument tubing is routinely used to allow a fluid connection between a primary coolant system and a device to measure a process variable such as pressure or coolant level, The tubing is typically routed on the

, most direct path possible, but changes in tubing size, elevation, or direction necessitate the use of compression fittings tojoin two pieces of tubing (Duke Power 1993). There are a variety ofdesig. A for each type fitting, and whereas only a single configuration is described here, they all use the same basic mechanism for gripping tube. The Swagelok design is shown in Figure 81, and is manufactured in sizes ranging from 1.5 to 25.4 mm (1/16 to 1 in.). It consists of four parts: the back ferrule, the front ferrule, the nut, and the adaptor, also called the fitting body. Ferrules are hollow, cone shaped subcomponents, with the thinner end positioned toward the adaptor. The body has threads on its outside diameter, and is tapered at its end. De fitting is shipped assembled, with the nut finger tight. He fitting is installed in the field by inserting the tubing into the fitting until it abuts against the shoulder of the adaptor as shown in Figure 81. He nut is tightened 1 1/4 tums from finger tight, during this tightening, several movements take place within the fitting in a pre planned sequence (Crawford Fitting Company 1985).

e ne nut moves forward, driving the back ferrule forward.

The back fenule drives the front ferrule forward.

The leading cJge of the front ferrule is forced radially inward by the conical shape of the adaptor until it contacts the tube.

The trailing edge of the front ferrule is lifted by the back ferrule.

ne leading edge of the front ferrule crimps (swages) the tube causing greater resistance to forward movement, resulting in the leading edge of the back ferrule being driven radially inward to form a second crimp on the tube.

After 1 1/4 turns, the nut has moved 1.5 mm (1/16 in.) forward. The amount of back ferrule grip is determined by the tube's resistance to front fetTule action, so the grip is much tighter when heavy wall or lutrd tubing are used, such as for high pressure, vibration, or shock applications. Thus, by design, the fitting imparts a much more secure grip on heavy wall tubing than it does on thin-wall tubing.

81 DRAFT - NUREG/CR-6582

COMPRESSION FITTING FAILURES I

A Parker Hannefin compression fitting connection, shown in Figure 8 2, is also shipped assembled with !

the nut fmger tight The fitting is installed in the field as follows. First, the tubing is inserted into the adaptor so that the left end of the tube contacts the adaptor shoaldet. A hex nut threads into the right esi of the adaptor, and the lip on the right side of the nut overlaps a ferrule which is connected to the tube.

As the nut is torqued, it moves to the left thus reducing the gap between the adaptor and the nut. A leftward force is transmitted by the lip on the nut through the ferrule to the tube, compressing the left end of the tube into the adaptor shoulder. De force in the ferrule creates a small amount of plastic deformation of the tubing, causing a slight diameter reduction of the portion of the tubing near the left end of the ferrule. The ferrule is also compressed into the right edge of the adaptor. Installation instructions by the Par 6 Hannifm Company state that the fitting should first be installed *fWer-tight", then tightened 1 1/4 rums [3/4 turns on tubing less than or equal to 4.8-mm (0.1875 in.) diameter). If the fitting is not properly aligned, biruling may occur resuhing in the fitting not being adequately s: cured. (This is t.3 true for the

Swagelok fitting.) This may result in an insufficient axial . 'rength and leak tightness between the tubing section and the adaptor. Ilowever, the nut should not be over torqued or the threads may be damaged.

No inspection criteria are normally supplied by Parker Hannifm to ensure thejoint is leak tight, ne width of the gap between the adaptor and the i,ut may be measured 'o determine the stractural integrity of the joint.

Section NB 3671.4 of the ASMF CodesSection III allows compression fittings for tubing sizes not exceeding 25.4 mm (1 in.) (ASME 1995a). The design and installation of instrument sensing lines up to 25.4-mm (1 in.) outside diameter (tubing is specified by its outside diameter) or 19-mm (3/4-in.) nominal pipe is covered by ANSI S67.02 (1980). This standard was first approved in 1977, so the tubing in earlier nuclear plants, such as Oconec 3, was not designed to the standard. FSARs do not reference this standard, in the Comanche Peak and Mllistone 3 FSARs, tubing downstream of a 9-mm (3/8 in.) tiow restrictor is designed as Class 2. Sample-line piping and tubing is presently covered by ANSI /ISA S67.10 (1986).

His standard classifies the instrument lines as Class I bp to the f;rst flow restrictor. After the flow restrictor, the line is Class 3, and downstream of the first accessible isolation valve, the line is designed according to ANSI B31.1. The flow restrictor is designed such that the loss of reactor coolant through an instrument sensing line, if ruptured in non-Class I portion of the line, does not prevent orderly reactor l shutdown assuming makeup is only by the normal makeup system.

i 8,2 Trend in Leak Events Caused by Compression Fitting Failures The distribution of reported compression fitting leaks is shown in Figure 8-3. From the years 1985 through 1996, the events per year ranged from zero to three, distributed throughout the period. Of the fourteen events, twelve were reportable and two nonreportable. The event rate was 1.0 reportable leak events per year, with 1.2 total leak events per year. Based on the relatively few numbers, and the fact that no major changes in fitting design or installation have occurred, it is expected that the failure rate in U.S. PWRs is '

about one event per calendar year. The failure rate is likely to decrease, since utilities are replacing some of the compression fittings that are most susceptible to leakage with welded joints, ne leaks all occurred on instrument or sample lines. About half occurred on lines connects! to the main ;

coolant piping (including the RCPs and steam generators), with the remainder divided equally between lines connected to the pressurizer and the charging / letdown systems. Dere also was one incident on a low pressure safety injection instrument line, ne maximum leak rates ranged from very small to 494 IJmin.

(130 gpm) in the case of the 1991 Oconee 3 leak, in most, if not all, cases the root cause was installation DRAFT - NUREG/CR-6582 8-2

COMPRESSION FTITING FAILURES error, with high temperature, vibrsion, and line configuration diagnosed as contributing facton. The tubing and fittings are relatively delicate, and in one case, a contri'.mting factor to the failure was probably personnel using a tubing hanger for their support during maintenance (Duke Power 1993).

Although Figure 8 3 begins in 1985, compression fitting failures and resulting leaks had been occurrmg long before this time. For exa4nple, in 1972 one of three hot leg RTDs on the Surry 1 phnt became disengaged from its location, causing a 10-mm (0.375 in.) hole (NPE 1973). Evidently, the ferrules were not properly swaged on the tube, and the RCS pressure completely expelled the RTD from its hole in its manifold, in 1976 a Swagelok fitting in a sensing line on the Indian Point 2 RCS leaked due to improper installativn (Nuclear Power Experience 1979).

8.3 Analysis of 1991 Oconee 3 Leak. Event Caused by Compression Fitting Failure

- A reactor vessel level mdscatmg system (RVLIS) w s installed in the Oconee 3 plant (B&W designed PWR) during an outage that ended in March 1987. Level instruments and associated impulse lines were cornected to existmg tape en the reactor vessel head and steam generators A and B, and two taps on the decay heat drop Ime (Nuclear Power Experience 1992). On the hot legs near the steam generators, a 25-mm (1 in.) line cornwets m the hot leg. At about 350 mm (12 in.) above the hot leg, there is a feejunction.

One !be exiting the tee goes to a vent, and the other to a device for measuring differential pressure. A series of tubing reducers were used for transition from the 19-mm (3/4 in.) root valve to 9.5 mm (3/8 in.)

tubing as shown m Figure 84 Six compression joints per instrument line were used to achieve this reduction. A 9-mm (3/8 in.) flow restrictor was installed on this instrument line, directly upstream of the root valve (USNRC 1992b). The normally open root valve can be used to isolate the compression fittings from the RCS, but since there is no remote actuation feature, isolation can only be accomplished when the reactor is shut down and the plant is cooled down. T;r.ts the root valve does not perform the function of an accessible isolation valve, as defined by ISA S67.02 (1980). 'ntis type of configuration was also installed on the other two Oconee units.

Over four and one-half years aft:r their installation, one of the six compression fittings (nearest the root valve) on a RVLIS line connected to one of the hot legs failed, resulting in a leak that spilled a large amount of reactor coolant into the containment. A description of the event, the root cause analysis, the corrective actions taken, and the safety significance of the event are described below.

Event Description. On Nos ember 23,1991, Oconee 3 was operating at full power. The first alert of leakage to the reactor operators were alarms that indicated failure of an instrument train that included the RVLIS (Nuclear Power Experience 1992). Shortly thereafter, the operators noticed that the letdown storage tank and pressurizer levels were decreasing, and that high pressure safety injection makeup flow had inen:ased significantly. A radiation monitor alarmed, and the leak rate was estimated to be 2661/ min (70 gpm). Operators noted through the reactor building video camera that a significant amount of steam was rising, and condensing on walls and equipment. A rapid, controlled shutdown was initiated. The operators made additions to the letdown storage tank to compensate for the leak. During the shutdown, feedwater pump oscillations resulted in an automatic reactor trip. By the time the plant was in stable hot shutdown, the leak rate was estimated to be 494 L/ min (130 gpm). When the RCS pressure had been lowered to 0.21 MPa (30 psig),the leak rate was reduced to about 19 to 381/ min (5 to 10 gpm), and the leak stopped when the RCS was fully depressurized.

83 DRAFT - NUREO/CR-6582

COMPRESSION FITTING FAILURES Upon containment entry, the source of the leak was found to be a 19-mm (3/4 in.) diameter RVLIS instrument line tbt had pulled out of a compression fitting (one of the six fittings in the line) downstream of the root valve (see Figure 8 4). The line was located at the top of the RCS hot leg pipe where it entered steam generator A. Although the leak was nianually isolable, the root valve could not be shut until the plant was placed in cold shutdown and airbome activity levels decreased sufficientiy to allow containmen' ;try (Duke Power 1991). An estimated 330,600 liters (E'1,000 gallons) of reactor coolant was discharged to the containment (USNRC 1992c). An off site release occurred when ventilating the containment. The release was within design limits, and the volume of gas was not much more than is released when going into a normal refueling outage (Nucleonics Week 1991).

Root Cause Analysis. The root valve, fitting, and affected tubing from steam generator A were removed for inspection and analysis. The measured width of the gap (see Figure 8 2) was 4.6 mm (0.182 ,

in.) versus a nominal of 3.9 mm (0.153 in.), which is required by Parker Haimifin for proper fitting.

The probable cause of the tubing separation was attributed to improper installation in that the nut had not ,

been sufficiently tightened onto the compression fitting during initial installation in 1987. Apparently, the nut had been tighte ned approximately % tum less than was recommended by Parker Hannifin. The licensee had previously caperienced problems with inadequate assembly of small tubing fittings when the plant staff had failed to apply sufficiently high torques required to properly seat the fittings.

Further inspection of the failed fitting determined that the diameter reduction of the 19-mm (3/4 in.) tubing was J.05 mm (0.002 in.), much smaller than the corresponding diameter reduction of 0.18 mm (0.007 in.)

at the other fittings on that tubing that did not fail. This indicates that the failed fitting had not crimped the t"bing sufficiently.

Other generic root causes of fitting failures (not necessarily associated with the Oconee 3 failure) noted by the USNRC (1992c,1997c) are:

Interchanging hardware from d'frerent manufacturers (incompatible components);

Installing the ferrules backwarus in the fittings, or omitting the fermles;

Failing to bottom the tubing on the shoulder of the adaptor,

Using tubing that is not cut square or that is buried scratched, deformed, or contaminated with dirt, oil, or other contaminant;

Failing to adequstely tighten the fitting to the finger tight position before making additional turns form the finger tight position;

Failure to ensure that the tubing has not moved back out of the fitting when the nut is tightened;

Over torqueing the nut; and

Searing of the seat surface.

DRAFT - NUREG/CR-6582 8-4

COhfPRESSION FITTING FAILURES A review of procedures, quality control manuals, and personnel interviews indicated that procedures provided inadequate guidance and documentation for installation or inspection of tubing fittings. De inspector stated that he typically only checked to see that fittings could not be loosened by hand, and that the tubing could not be pulled out of the fitting. The fitting had kept the instrument line f om leaking for over four and one half years, while the line was subjected to as much as 15.2 MPa (2200 psig) during pressure tests. Herefore, v ' out some specific inspection criteria such as measuring the gap width, the lack of sufficient compression wuld not have been detected.

The USNRC (1992c) also mentived that inadequate procedures for initial installation or for disconnecting and retightening during maintenance, and inadequate trcining of personnel may contribute to the failure of the fittings.

Corrective Acrlons, ne root valve, the 19 mm (3/4 in.)tubir.6, and associated fittings on the hot legs of both the steam generator A and B impulse lines were replaced (Nuclear Pcwer Experience 1992). ne redesigned configuration used welded fittings to replace fs r of the compression fittings per line, so that only two compression fittings remained in each new line.

All 264 Parker Hanmfin compression fittings on tubing connected to the RCSs and related high pressure safety injection systems were inspected (and tightened if necessary and possible). A small percentage of the Parker Ilannifin fittings (5.7%) could not be tightened to meet the neminal gap values witnout excessive force. He nominal expected gap fumished by Parker liannifin was used as an acceptance criterion, ne licensee also inspected 191 SwaSelock fittings in the RCS using gap inspection gages fumished by Swagelock. Approximately 28 percent of both types of fittings appeared to have gaps outside the nominal ranges One fitting had boron on it, indicating that it had leaked, and another had a loose nut.

An opert.bility assessment was conducted to determine if it were prudent to return the plant to service with the gap on several fittings remaining outside the manufacturer's recommended nominal range. He fitting, which had been the most out-of nominal range after retightening, was replaced and inspected. The maintenance engineers concluded that the fermle was installed properly and adequately crimped on the tube despite the gap being out of the nominal range. Based on judgement of instrument technicians that the fittings could not be further tightened without damaging the tubing and connection, and on an engineering evaluation made after disassembling and inspecting one of the questionable fittings, the 23 fittings [ sixteen 12.7-mm (K in.) and seven 6.4 mm (1/4-in.) diameter) with larger than numinal gaps w cre accepted.

The instrument line on the reactor vessel head was replaced before plant stanup. The similar portions of the RVLIS lines on Units 1 and 2 were also replaced with a configuration using fewer compression fittings.

Policy directive and procedure enhancements were to be implemented to ensure proper installation and inspection of compression fittings. All personnel who inspected, installed, made up, or remade tubing w cre to receive additional training to assure that the manufacturer's instructions were understood and complied with. During the subsequent startup, a pressure test and walk down were perfonned to detect any visible leakage.

The corrective action for the Catawba 1 pressurizer line in 1985, as well as for a reactoi coolant loop RTD line leakage in 1987, was to replace the compression fittings with welded fittings. Similarly, the pressurizer pressure and level instmmentation impulse line compression fittings on both hicGuire units were replaced with welded fittings as a result of the 1986 pressurizer instrumentation leakage event. Bush (1992) 85 DRAFT - NUREG/CR-6582

COMPRESSION FITTING FAILURES recommends that compression fitting should be forbiddu on safety reinted components and replaced elsewhere whenever reasonable to provide greater reliability for preventing primary coolant leakage.

Safety Significance. The two most significant leak events occurred at McGuire 2 and Oconee 3. In the McGuire 2 event, the maximum leakage occurred during repairs to a pipe hanger supporting the fitting.

He maximum leak rate was 304 Umin. (80 gpm) and was within the capacity of the charging pumps (a second pump had to be started). De pressurizer evel was maintained, and the total volume ofleakage was limited.

P>r the Oconee 3 r' ent, the maximurn leak rate was 494 U'nin. (130 gpm) and was within the car. ,, of the high pressure safety injection makep system. nCS inventory was maintained; however a significant amount of raakeup was required. No engineered safeguards or emergency feedwater actuations were ,

required during the entire event.

The USNRC (1992b) issued a Notice of Violation to Duke Power as a result of the leakage, with a Severity ~

Level IV classification. Activity level in the reactor building, due to iodine and noble gas, was signif cantly elevated followns the es ent, and high dose rates, as well as contamination, v ere produced in parts of the building. Ilowever, the leak rate was controlled by the flow restrictor in the line.

Duke Power performed an analvsis of the failure consequences of the problem fittings. In one case, the consequence was a reactor trip, but in most of the other cases, the consequences were simply alarms to the operaurs or loss of an instrument reading that could be compensated for by relying on other instruments (Nuclear Power Expe:ience 1992).

De 14 eveats, mentioned in Section 8.2, were not considered to have a risk impact because the reactor plant makeup flow was sufficient to maintain the RCS inventory during these events. He leak rate was highest in the Oconee 3 event; however, the conditional core damage probability was estimated to be less an 1.0E 06, and therefore, not analyzed by the Accident Precursor Program. This is further discussed in Section 9.2.

Leaks also can cause lors of the instrumentation function. In 1997, lack of redundancy and a compression fitting leak in the Oconee 3 letdown storage tank level indicating system reoitted in a nonconservative level indication. His lead to the plant being operated outside its design basis, such that the ability of the high pressure injection system to perform its safety function was not assured (USNRC 1997d).

USNRC Information Notice 92-15 (USNRC 1992c) pointed out the compression fitting problem associated with the Oconec 3 failure to licensees. However, the circumstances were not considered to be sufficiently serious to impose requirements on licensees, and no specific actions or written response were required by the Notice.

l l

DRAIT - NUREG/CR 6582 8-6 1

9. SAFETY SIGNIFICANCE OF RCS LEAK EVENTS Two questions associated with the risk aspect of RCS leaks are the focal point of this safety analysis. First, "Are there any experienced leaks that could be regarded as precursors to core damage?' To answer this question the ASP Program was reviewed for leak events that were analyzed within that program. These events have gone through substantial risk analysis and industry review. The results of this review showed that, indeed, there were several leaks that were judged to be precursors to core damage,

%e second question is, 'Does the industry leak experience change the standard perception of the likelihood of a small break LOCA?" Using the insights Se - ' from the ASP review, the entire RCS leak LER databar,e related to PWR primary system leak eveii.: snich occurred during a period from 1985 to 1996, underwent a risk based analysis nis analysis categorized the LERs according to their s.ctual or potential risk impacts from the perspective of contributing to frequencies and probabilities typically modeled in a full scope Level 1 PRA. The results of the categorization served as the raw data for estunating the frequencies and probabilities for the atsociated risk parameters. Comparisons were then made between these estimates and the typical PRA values. De ';npact of the differences on risk was then explored.

9.1 RCS Lesks in the ASP Program The ASP Program Annual Reports prepared by Minarick et (19xx) were reviewed ft analyzed events involving RCS leaks or t aptures (other than steam generator tube leaks or ruptures). Nine events fitting this description were found, ne conditional core damage probabilities for these events ranged from 1.3E-6 to 3.3E 3. Table 9.1 summarizes these nine events. The comments column indichtes the role the RCS leak played in the conditional core dantage probability calculation.

Investigation into why r5ese events were " risk significant" enough for ASP Program evaluation, while the remaining 232 leak events in the operationai data experience were not, reveals some interesting insights.

The first four of the nine analyzed events were of interest to the ASP Program because cf circumstances unrelated to the RCS leak. These events would have been analyzed even if the leak had not occurred, and furthermore, the presence of the leak did not contribute to the conditional core damage probability. These events do not give us any risk insights for the RCS leak issue.

>

The fifth event (LER 24791001) was analyzed because the RCS leak caused a plant transient (reactor

. trip) with subsequent :quipment failures unrelated to the RCS leak. These types of events contribute to the plant risk as a portion of the general transient ini'iating event, not a small break LOCA.

e ne sixth event was analyzed because during a plant transient a RCS leak developed and the subsequent repair action disabled a safety system (HPI). Events where the RCS leak or leak repair action disables or degrades a safety system or train represent one of many mechanisms that contribute to an increase in the likelihood of core damage as a result of reduced accident mitigation capability.

91 DRAFT - NUREG/CR-6582

SAFETY SIGNIFICANCE

The wcenth event was analyzed because the RCS leak was induced by a plant transient. Events of this type are transient induced LOCAs. A transient-induced LOCA starts out as a general plant transient or loss of offsite power and develop into a LOCA when plant equipment fails due to the off-normal condition. The most typical transient-induced LOCAs involve a stuck-open relief valve or a RCP seal failure. These events are modeled explicitly in the general transicat and loss of offsite power PRA event trees. This type of event in the operational database provides insights concerning the conditional probability of developing a LOCA given a plant transient has occurred.

'Ihe last two events were analyzed because the RCS leakage was great enough or had the potential to develop to a size that would require safety system response as modeled by a PRA small LOCA event _

tree. Events of this type cont-ibute directly to the small LOC /. 'uitiating event frequency.

9.2 Risk Review of RCS Leak LER Database 9.2.1 Risk G. pact Categories -

l Based on the findings of the ASP Program analyses review, the risk impact of RCS leak events can be divided into six categories. Five of these categories were observed in the precursor analyses review.

Category 4 below, while not observed in the precursor analysis review, is a reasonable extrapolation.

Should a number of RCS leak events fall into this category, then the possibility of a correlation bWeen

, RCS leaks atui the general transient initiating event would warrant investigation.

! The six risk impact categories are:

l

1. Does not contribute to any risk carameter. The leak does not require any short-term safety response, j does not impact any safety systems or functions, nor does it have a potential to do so.

2. Coatributes to reactor trio (ceneral transient) initiatine event freauency. The leak was such that a reac*or shutdown was required and a reactor trip occurred as a direct result of the leak. Reactor trips caused by other equipment failures (instrumentation, etc.) that would have occurred regardless of what initiated the rear *.or shutdown are not placed m this category. These reactor trips would be considered as initiated by other equipment / human failures, not RCS leaks.

i

3. Contributes to Ph-modeled" system decradation/inoperability. The RCS leak is such that a system or a support system typically modeled in a full scope PRA is unable to perform its mode led function to its full capacity or with full redundancy.

4. L;tt,apgi to reactor trio (ceneral transient) initiatinc event freauency and PRA-modeled system ,

dty.iMatvurocrability. This category is a combination of categories 2 and 3, tripping the reactor and impactin;; a PRA-modeled system.

U PRA-modeled system is defined as a safety system, support system or secondary plant system typically a1 deled in a full-scope PRA.

DRAFT - NUREG/CR-6582 9-2 i

I l

SAFETY SIGNIFICANCE

5. CSDtributes to transient induced LOCA rirobability. He leak occurred due ;o an improper equipment response to a transient condition, ne most common of these modeled in a typical PRA are a stuck-open relief valve and pump seal failures due to loss of suppon systems such as cooling and/or injection.

Rese failures occur in the presence of an off-normal condition such as a pressure transient or loss of offsite power. Spontaneous failures of valves or seals would not fallinto this category, but most likely would meet the requirements of Risk Impact Category 6.

6. Contributes directly to the small LOCA initiatinr,,xygtg frequeaq.y. The Icak, without safety system response or appropriate operator action, wouli enranaLy lead to core damage. This also includes small leaks that, if not attended to, have the potatial to b ? cow large enough to require safety system or operator response. For example, a RCP seal led :t%1 ting from :ignificant damage to multiple seal stages could grow to become a small LOCA. Such a a event would fall into this category. On the other

~

hand, a leak on a small valve that could not get la'ge enough to become a small LOCA, even in the worst esse scenario, would not be in this categor5 The entire RCS leak LER database was reviewed and each LER was assigned to one of the above Risk Impact Categories. Of the 240 events in the database,198 did not contribute to any risk factor (Risk Impact Category 1), and one event did r.ot involve an actual RCS leak, but rather a loss of leak detection capability. The remaining 41 events were assigned to Risk Impact Categories as indicated in Table 9.2.

The events are arranged by risk impact category. The risk impact category for each event is listed and the justification for assignment to that category is provided. Of the 41 events,29 involved reportable leaks, 12 did not. Of the 12, ten were in Risk Impact Category 3, degradation /inoperability of a PRA-modeled system. This would be expected since this category is more affected by the location of the leak and not the size of the leak. As long as the leak size is small compared to the charging system capacity, the leak location has a greater influence on the risk impact than the leak size. For example, a 20 gpm RCS leak is reportable, whereas a 0.2 gpm leak is not, even though they may both be in the same locatiou. But from a risk impact perspective, both leaks require the same plant configuration for isolation and repair and will both take about the same time for repair, and therefore, have the same risk impact. On the other hand, a 20 gpm leak in a sample line may have no risk impact, whereas a 0.2 gpm leak in the common discharge header for the HPI system could comphtely disable the HPI system, a noteworthy risk impact.

Risk Parameter Frequency Calculation. Based on the number of events in each categosy and a total of 638 reactor years of operation represented by the RCS leak LER database, the frequency of each risk parameter was calculated as shown in Table 9.3. This table also shows the number of RCS leaks events assigned to each risk impact category.

Interesting Events in Risk Impact Category 1. During the categorizing process, a number of events were found that technically did not have a risk impact as defined by the risk impact categories but are worth noting because of the circumstances or magnitude of the leak. A brief description of each of these events follows.

On November 23,1991 a 3/4-in. compression fitting on an instrument line failed at Oconee 3 (LER 28791003). The initial leak rate was 70 gpm and a reactor shutdown was begun. By the time the reactor was in hot standby, the leak rate was 130 gpm. The leak was not completely stopped until the reactor was completely depressurized. There were no engineered safeguards or emergency feedwater actuations during the entire event. A total of 87,000 gallons of reactor coolant had leaked out.

9-3 DRAFT - NUREG/CR-6582

SAFETY SIGNIFICANCE Because reactor plant makeup flow was sufficient to maintain RCS inventory the entire time, this event was not considered to have a risk impact, ne ASP Program classified this as an interesting event and made a screening evaluation that the conditional core damage probability was less than 1.0E-6, and therefore, not an accident precursor.

On Maren 22,1993 at McGuire Unit 2 a compression fitting in an instrument line was leaking as a result of a failed pipe hanger (LER 37093003). The ficia leak was stopped, however, during repairs to the pipe hanger the fitting failcd causing at 90 gym leak. A second charging pump was started to main.ain pressunzer level until the leak was is ta:ed. Even though the leak rate was substantial, it was within the capacity of the charging system and the total volume leaked was limited. This event was deemed to have no risk impact.

Risk Impact Category 2 (General Transient). This risk impact category can be directly compued to risk parameters found in a typical full-scope PRA. A good generic frequency for general transients of all hinds is on the order of I or 2 per reactor year (Gentillion et al.1994). The general transient initiating .

event frequency associated with RCS leaks (6.27E-3/ reactor year) is an insignificant fraction of the total general transient frequency. Therefore the core damage frequency attributable to leak-induced transients is also comparatively small.

Risk Impact Category 3 (Safety System impact). This category had 24 events assigned to it.

He systems most frequentl." impacted were charging, high head safety injection, low head safety injection, and power-operated reilef valves. Most impacts involved only a single train, however, five events involved multiple trains. Pump and valve leaks tended to impact a single component or train while pipe weld leaks were more likely to affect multiple trains. This analysis did not uncover any risk impacts not currently considered in typical full-scope PRAs. The presence of five events with multiple safety system train impacts emphasizes the importance of modeling common cause failures and the proper impact of single point failures such as leaks in common piping headers.

Risk Impact Category 4 (Reactor Trip and Safety System impact). This category had no observed events. While such events are possible, most safety systems are in standby and the leaks are discovered before the leak rates become large enough to trip the reactor. Rather t' tan cause a reactor trip, leak events tend to be discovered early and the component is taken out of servic: and the plant is placed in a imited condition of operation if appropriate.

Risk Impact Category 5 (Transient-Induced LOCAs). This category is usually reflected in full- .

scope PRAs in two major areas. First, in the probability of a stuck-open relief valve given a pressure transient, and second, in the probability of a RCP seal LOCA given a transient (including loss of offsite power) where one or more needed support systems are lost. Of the six events in Risk Impact Category 5, .

four were associated with relief valve failures and two were RCP seal failures. The transient-induced LOCA frequency due to relief valve failures is 6.27E-3/ reactor year and for RCP seal failures is 3.13E-3/ reactor year.

For a typical transient initiating event frequency of 2 per reactor year, a generic probability of lifting a relief valve given a transient of 4.0E-2 (Minarick et al.19xx), and a generic power-operated relief valve failure-to-close probability of 3.0E-2 per valve (Druin et al.1987), the frequency of a single relief valve transient-induced LOCA is (2/ reactor year)(4.0E-2)(3.0E-2/ valve) = 2.4E-3/ valve-reactor year. Given DRAFT - NUREG/CR-6582 9-4

SAFETY SIGNIFICANCE that there are between one and three power-operated relief valves per PWR reactor, the overall relief valve

- transient induced LOCA frequency estimate for a given plant is between 2.4E-3 and 7.2E-3/ reactor year.

This is comparable to the 6.27E-3/ reactor year value coming from the RCS leak LER database.

The two most common transient challenges to RCP seal integrity modeled in PRAs are losses of component cooling water and station blackouts. The typical PRA does not consider other causes mainly because with seal injection and seal cooling readily available, recovery before the leak size becomes threatening is very likely. Recoverv during a loss of component cooling water or station blackout is much more difficult and any delays can result in leaks bccoming small LOCAs. Thus, even though the types of seal leaks observed in the RCS leak LER database are more frequent, they are not corddered significant contributors to core damage.

~

With the frequencies of loss of component cooling water and station blackout on the order of 1.0E-5 to 1.0E-4 per reactor year, observing no seal LOCA events due to these causes in 638 reactor years of data is not surprising.

Risk Impact Category 6 (Small LOCA). The generic frequency for small LOCAs is generally accepted as being around 0.01 per reactor year (Minarick :t al.19u, Gentillion et al.1994). This compares extremely well with the RCS leak LER database result of 0.011. Based on this close comparison, the contributions to core damage from small LOCA as calcul.ated in the accepted PRAs within the industry are reasonably the best estimates available.

9.3 Conclusions The risk-based analyses discussed above were conducted to answer two questions: 1) "Are there any experienced leaks that could be regarded as precursors to core damage?" and 2) Does the industry leak experience change the standard perception of the likelihood of a r. mall break LOCA?"

In answer to the first question, there have been nine events involving RCS leaks analyzed by the ASP 4 Program. Of those nine events, four were not influenced by the RCS leak in any manner, one initiated a plant transient that was improperly responded to by the plant systems or operators, one degraded or disabled a safety system, one was a transient-induced LOCA, and two were actual RCS leaks that had the potential to become LOCAs. Thus, three operational events in the 1985 through 1995 timeframe have been identified and analyzed as LOCA accident precursors because of RCS leaks.

The answer to the second question is no, in fact, the industry leak experience has always been the basis for small LOCA frequency calculations (Gentillion et al.1994, Druin et al.1987) and any dramatic difference between the standard perception and the outcome of this study would be unexpected. The frequency of small LOCA from this study (0.011) is nearly identical to the generic small LOCA initiating event frequency (0.01) used in many PRAs.

In addition to these two questions, conclusions about the overall perception of the contribution of RCS leaks to core damage can be made. The frequency of leak-induced transients is so small (6.3E-3/ reactor year) in comparison to the overall transient initiating event frequency (1 - 2/ reactor year) that there is no discernable influence on the core damage frequency. The frequency nf tnnbic.t-induced relief valve LOCAs from the operational data (6.27E-3/ reactor year) is within the range of c:Aulated values observed l

9-5 DRAFT - NUREGICR-6582 l

l l

SAFETY SIGNIFICANCE . ]

l in typical PRAs. The frequency of RCP seal leaks W=iae small LOCAs is so snall that none exist m the database, Overall, the risk impacts observed in the RCS leak LER database are accurately reflected in the typical full-scope PRA.

i e

1 e

,[**

t DRAFT - NUREG/CR-6582 96

_ ~ -

.--r _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

10. Effectiveness of Industry Reactor Coolant Leakage Detection Systems De safety significance of PWR primary system leaks can vary widely, depending on the location of the leak, leakage rate, and its duration. He locstion of a leak may be such that the leak or leak repair action disables or degrades a safety system and contributes to an increased likelihood of core damage as a result of reduced accident mitigation capability. A leak may initiate a small-break LOCA if its rate is higher than the charging pump capacity. In the absence of a safety system response or appropriate operator actions, this leak would eventually lead to core damage. Sometime the leak location is such that a manually isolable leak could not be isolated tmtil the plant is placed in cold shutdown and airborne activity levels decreased sufficiently to allow containment entry. Such a leak took place at Oconee 3 and lasted for about 11 hours1.273148e-4 days 0.00306 hours 1.818783e-5 weeks 4.1855e-6 months (USNRC 1992c). A high-rate, long-duration leak increases fne likelihood that some other malfunction will occur and thus compound the recovery actions. The safety synificance of the leak events occurring during 1985 through Septemi.er of 1996 is discussed in Section 9.

Leak detection systems are employed in the case of leak-before-break behavior of primary pressure retammg components to ensure the detectability of leakage from a crack before it grows to critical crack size. Early leak detection and appropriate response will help to adhere to the design conditions for the power plant components and may play a potentially important role in accident prevention, Some welds are inaccessible for inservice inspection, and early leak detection and localization can timely identify the presence of a throughwall crack. Some base metal sites are susceptible to cracking but not inspected during inservice inspection; early leak detection and localization of a throughwall crack at these sites would allow for orderly shutdown of the reactor or other appropriate actions. Use ofleak detection systems with the capability to detect a smaller than 1-gpm leak would allow application of the leak-before-break concept to smaller size piping. Quick localization of leakage could reduce the radiation exposure to plant personnel inside containment. Thus, the use of a leak detection system would lead to detection of system degradation and is consistent with the defense-in-depth concept.

Sometimer, the leak detection systems have identified the presence of throughwall cracks that were not detected by the plant inservice inspection methods. Some throughwall fatigue cracks in PWR primary system branch lines, such as one in the Farley safety injection line, were found because of leakage, not inservice inspections. Similarly, many IGSCC cracks in boiling water reactor piping have been aussed during ultrasonic inservice inspection and were detected only because of leakage (Kupperman et al.1989).

He reason is that the fatigue and IGSCC cracks may close when the reactor is shut down, and their

, detection by inservice inspecti*n becomes difficult. A sensitive leak detection system can detect a presence of a throughwall crack in the reactor pressure boundary when it is still small.

he outline of the secticn is as follows. First the requirements for leak detection systems employed at the U.S. PWRs are summanzed. Then the current leak detection systems and their characteristics are discussed. Finally the advanced leak detection s; stems and their sensitivity to detect small leakage a ul capabilities to quickly localize it are discussed, t ;

l 10 1 DRAFT - NUREG/CR-6582

LEAKAGE DETECTION SYSTEMS 10.1 Recommendations and Requirements for Leakage Detection Systems General Design Criterion 30, " Quality of Reactor Coolant Pressure Boundary," of Appendix A to 10 CFR Part 50 requires in part that, "means shall be provided for detecting and to the extent practical, identifying the location of the source of reactor coolant leakage." Regulatory Guide 1.45, Reactor Coolant Pressure Boundary Leakage Detection Systems, provides recommendations for satisfying this requirement, which are implemented by means of plant technical specifications.

Leakage from the primary system are of two categories, identified leakage and unidentified leakage. Since equipment in the prunary system e mnot be perfectly leak-tight, identified leakage from valve packing, shaft seals, and other equipment is generally present during normal operation. Identified leakage includes two types: (a) leakage that is captured and conducted into a collection tank, and (b) leakage (other than pressure boundary leakage) into the containment atmosphere from sources that are both specifically located and known not to interfere with the operation of the leakage detection systems. All other primary system leakage, including pressure boundary leakage, is considered unidentified leakage. Pressure boundary leakage is dermed as leakage through a unisolable throughwall crack in a RCS component body, pipe wall, or vessel wall. Identified leakage is detectable and collectable, and, to the extent practical, isolated from the containment atmosphere so as not to mask any unidentified leak. Unidentified leakage is not isolated from the containment atmosphere out released into it. Leak detection systems are necessary to separate the unidentified leakage from the identified one to provide prompt and quantitative information to allow the plant operators to take immediate corrective action should a leak be detrimental to plant safety.

Regulatory Guide 1.45 Recommendations. The regulatory guide provides recommendations regarding separation of identified and unidentified leakages, acceptable leak detection methods, sensitisity of leak detectors, detector response time, and seismic qualification of instrumentation employed for leak detection. These recommendations are for detecting leakage inside the containment. The main recommendations are summarized as follows.

1. Flow rates of identified leakage should be monitored separately from unidentified leakage.

2. Flow rates of unidentified leakage should be monitored with an accuracy of 3.8 IJmin (1 gpm) or better.

3. At least three separate leak detection methods should be employed to ensure effective monitoring ,

during periods when some detection systems may be ineffective or inoperable Two of these methods -

should be sump level and flow rate monitoring, and airborne particulate radioactivity monitoring. A third method may be either condensate flow rate monitoring or airborne gaseous radioactivity .

monitoring. [The Japanese PWRs are designed to have all four of the leak detection methods mentioned (Aoki 1991)]. Use of humidity, temperature, or pressure monitoring of the containment atmosphere is also recommended as part of a leak detection system.

4. Each of the three selected leakage detection systems should be able to detect an unidentified leakage with a leakage rate of 3.8 L/ min (1 gpm) or smaller in less than one hour.

DRAFT - NUREG/CR-6582 10-2

I LEAKAGE DETECTION SYSTEMS i

5. The leakage detection systems should be capable of performing their function following seismic events that do not require plant shutdown.

6. Indicators and alarms for each leakage detection system should be located in the main control room.

The recommendations are limited to reactor coolant leakage into the prunary containment, not to the outside of the containment. The regulatory guidJ also recommends monitoring systems, such as monitoring of coolant radioactivity, to detect intersystem leakage, for example, steam generator tube leakage.

Intersystem leakage poses an important safety issue but it is not within the scope of this project; therefore, it is not discussed hereafter.

Technical Spec /fications. Technical specifications define the limiting conditiors for operation, actions that should be taken if these conditions are exceeded, and surveillance requirements for leakage detection systems and for operational leakage inside the containment. At present, the USNRC does not have technical specificatun requirements for leakage outside the containment. The technical specification requirements generally follow the regulatory guide recommendations discussed. For example, the requirements for the D. C. Cook plant are as follows (D. C. Cook 1995).

Technical Specifications for the D. C. Cook plant requires that the following three leak detection systems be operable during Modes 1,2,3, and 4; these modes correspond to power operation, startup, hot standby, and hot shutdown, respectively: (1) sump level and flow rate monitoring, (2) airborne particulate radioactivity monitoring, and (3) either the containment humidity monitoring or airborne gaseous radioactivity monitoring. Plant operation may continue for up to 30 days with only two of these required leakage detection systems operable. A condition for the continued operation is that grab samples of the containment atmosphere be obtained and analyzed at least once per 24 hours2.777778e-4 days 0.00667 hours 3.968254e-5 weeks 9.132e-6 months when the required particulate and/or gaseous monitoring systems are inoperable. If these conditi ons for limited continuous operation are not satisfied, the plant must be in at least hot standby within tb next 6 hours6.944444e-5 days 0.00167 hours 9.920635e-6 weeks 2.283e-6 months and in cold shutdown within the following 30 hours3.472222e-4 days 0.00833 hours 4.960317e-5 weeks 1.1415e-5 months .

The sump level and flow rate monitoring system and the contamment humidity monitoring system must be demonstrated operable at least once per 16 months. The particulate and/or gaseous monitoring systems must be demonstrated cperable at the frequencies specified in the technical specifications.

RCS leakage must be limited to 1 gpm of unidentified leakage,10 gpm of identified leakage, and no pressure boundary leakage. These limiting conditions of operation are applied to Modes 1,2,3, and 4.

With any pressure boundary leakage, the plant must be in at least hot standby within 6 hours6.944444e-5 days 0.00167 hours 9.920635e-6 weeks 2.283e-6 months and in cold shutdown within the following 30 hours3.472222e-4 days 0.00833 hours 4.960317e-5 weeks 1.1415e-5 months . With any reactor coolant leakage, except pressure boundary leakage, greater than 1 gpm unidentified or 10 gpm identified leakage, the licensee is required to reduce the leakage rate to within limits within 4 h or be in at least hot standby within the next 6 hours6.944444e-5 days 0.00167 hours 9.920635e-6 weeks 2.283e-6 months and in cold shutdown within the following 30 hours3.472222e-4 days 0.00833 hours 4.960317e-5 weeks 1.1415e-5 months .

RCS leakage must be demonstrated by the required leakage detection systems to be within the required limits at the following frequencies: (a) at least once per 12 hours1.388889e-4 days 0.00333 hours 1.984127e-5 weeks 4.566e-6 months for the containment airborne particulate radioactivity monitor, (b) at least once per 12 hours1.388889e-4 days 0.00333 hours 1.984127e-5 weeks 4.566e-6 months for the containment sump level and flow rate monitor, (c) at least once per 72 hours8.333333e-4 days 0.02 hours 1.190476e-4 weeks 2.7396e-5 months during steady state operation for the RCS water inventory balance estimates, and (d) at least once per 24 hours2.777778e-4 days 0.00667 hours 3.968254e-5 weeks 9.132e-6 months for the reactor head flange leak-off system monitor.

143 DRAFT - NUREG/CR-6582

- =. -

LEAKAGE DETECTION SYSTEMS USNRC Recommendations forLeak Detection Systems as AppIled to Leak Before Break CorNept. The accuracy and sensitivity of the leak detection system employed determines the size of a postulated throughwall flaw for leak-before-break analysis. The size of the flaw should be such that the calculated leakage rate of fluids discharged through the flaw under normal operating loads should be detectable with a margin by the leak detection system employed. The margin of the magnitude ofleakage applicable to high energy piping both within and outside of containment should be at least 10 times the capability of the leakage detecWn system used. So, if the leak detection system is capable of detecting I gpm leakage, then the hypothetical throughwall flaw size corresponds to a flaw size through which the leakage rate is 10 gpm. In addition to the capability of the leak detection system, the system should have sensitivity and reliability as recommended by Regulatory Guide 1.45 (USNRC 1985b).

Pipes with an inside diameter of 4 in, and smaller may not pass the acceptance criteria for the leak-before-break concept, mainly because of the relatively large size of the hypothetical flaw to accommodate the 10-gpm leakage. The resulting flaw angle becomes large for a small diameter piping, leaving a small uncracked ligament. However, if a plant-specific leak detection system is capable of detecting smaller than ,

I gp.n leakage, then the size of the corresponding hypothetical flaw becomes smaller, and the smaller di. meter piping may pass the acceptance criteria (Maxham and Yoon 1987).

The leak-before-break concept is not applied to piping systems outside the containment of U.S. PWRs, where recommended leak detection systems are generally not employed, it is generally difficult to install such a system because of the large number of potential leakage paths associated with the large number of pipes and fittings on the secondary side (Maxham and Yoon 1987).

The leak-before-break concept is not applicable to high energy fluid system piping, if operating experience has indicated that the piping is susceptible to failure from stress corrosion cracking, water hammer, thermal and mechanical fatigue, or other degradation mechanisms (USNRC 1985b).

10.2 Distributions of Methods Used for Leak Detection and Location Determination Figures 10-1 through 10 3 show the distribution of methods used to initially detect leaks. Figure 10-1 illustrates the reactor coolant leak event distribution by detection method for both reportable and nonreportable leaks. The term detection method used in this subsection implies how the control room operators were alerted to an increase in reactor coolant leakage. Three primary methods for identifying reactor coolant leaks are control room indication, periodic surveillance test, and personnel visual ,

observation. These three methods account for the detection of over 70% of the leak events. They are described next.

Control room Indications. The various process parameter indications, monitors, and alarms available to control room operators provided indications of leaks. Examples of the process parameters include containment sump pump out rates, volume control tank makeup rates, letdown flow rates, charging pump flow rates, and pressurizer level and pressure indication. However, these indicators are not sensitive to changes resulting from very low leak rates (such as less than 1 gallon per hour). Therefore, in addition to control room indication, containment radiation monitors and associated alarms provide operators with additional indication of the existence of increased reactor coolant leakage. These monitors have alerted operators to the appearance of leaks found to be only a few drops per hour or less. Because these monitors DRAFT - NUREG/CR-6582 104

LEAKAGE DETECTION SYSTEMS i

l and alarms generally provide indication of a very small leak, leaks identified by them are separated from the control room indication category.

Control room indicators may not provide timely detection of leakage if the supporting systems and components are not aligned correctly or functioning properly. 7or example, an event occurred at St. Lucie, Unit 1 (LER 33595004) in which the floor drain isolation valves to the sumps (outside the containment) were left closed and, thertfore, there was no indication of an increase in the sump level even though a valve was leaking for 2 hours2.314815e-5 days 5.555556e-4 hours 3.306878e-6 weeks 7.61e-7 months before it was found. Approximately 4,000 gal of reactor coolant was discharged during the 2-hour period.

Periodic surveillance tests. nese tests are required by the plant technical specifications to identify trends in the reactor coolant leakage rate. They are typically performed once a shift and consist of a RCS mass balance calculation that determines both identified and unidentified leak rates. Changes in ranctor coolant leak rates greater than 0.1 gpm are easily found by the periodic surveillance tests. As observed in the LER data, an upward trend in the leak rates of as little as 0.1 gpm has alerted operators. Typically, this change in leak rate requires operators to take actions to identify the reason for the increased rate prior to exceeding a technical specification limit and to determine whether there is a pressure boundary leak.

Personnel visual observation. Operators find leaks visually during the course of routine plant operations. Sometimes video monitoring equipment is used. Typical leaks identified in this manner are valve packing leaks and other small leaks. The sensitivity of this meth( d depends on the accessibility of the location and the frequency of inspection. However, since access to containment is limited during plant operation, the usefulness of this method for early detection ofin-containment leaks is limited.

Of the 199 reportable events studied,153 (77%) occurred inside primary containment,45 (22.5%) occurred outside of primary containment, and the location of one leak (relative to primary containment) was not identified in the LERs. Figure 10-2 shows that about 60% of the leaks that o: curred inside containment were detected by two methods: control room indications and periodic surveillance tests. Another 12% of the leaks were detected by personnel observation (visual inspection).

Figure 10-3 shows that personnel visual observation was the main leak detection method for the 62 reactor coolant leaks outside the containment. About 44% of the leaks were detected by personnel observation, whereas about 13% of the leaks (6 events) were detected by control room indications. Table 10-1 gives the names of the plants where these six events took place. It appears that the USNRC has no requirement for having a leak detection method for outside the containment. Some plants have sumps to collect outside-containment leaks from certain systems, and alarm systems for these sumps are often connected to alarms in the control room.

Figure 10-4 addresses reportable leaks both inside and outside of primary containment. He figure shows that the majority (about 80%) of the leak sources were located by personnel observation. This is an effective method, which can accurately identify the exact location of the leak. This helps to determine whether it is an identified or unidentified leak, and whether it is a pressure boundary leakage.

Figure 10-5 shows the distribution of 145 reportable leak events by initial leak rates and detection method.

For 51 other events, leak rates were not known. The remaining three leaks were detected by techniques other than the ones listed in the figure; two were detected by secondary radiation monitor and one by su area fire 10-5 DRAFf - NUREG/CR-6582

LEAKAGE DETECT 10N SYSTEMS i

alann. As shown in the figure, periodic surveillance testing was very effective at identifying small leaks up to 5 gpm. The method was attributed to finding 40% of the leaks less than I gpm and 52% of the leaks that were sized from 1 to less than 5 gpm. Control room indication is seen to be effective for all leak sizes, particularly for leaks 5 gpm or greater; 75% of the leaks greater than 5 gpm were identified with the use of control room indication.

10.3 Effectiveness of Current Leakage Detection Systems When a leak of primary system coolant occurs, a portion of the leakage evaporates and the resulting vapor is transported to the containment vessel atmosphere and mixed by the containment vessel air circulation system, and the remaining portion of the leakage, which is in the liquid phase, is routed to the containment sump. Then, the leak detection systems typically installed in both the containment atmosphere and the sump could detect the leakage. Some of the leak detection systems, such as containment sump level and flow monitor and air cooler condensate flow monitor, monitor both primary and secondary leakage, so the samples have to be analyzed to evaluate the respective part of each type ofleakage. Therefore, more than one leak ,,

detection system are used to accurately and reliably estimate the total leakage rate. Effectiveness of these leakage detection sptems is evaluated in this section.

The USNRC PWR Pipe Crack Study Group (1980) evaluated the capabilities of different available leak detection system s and concluded that none of the systems would be acceptable for complete characterization of a leak, detecting and locating a leak, and accurately estimating its rate, as shown in Table 10-2. Out of eleven systems. presented in the table, three are capable of a quantitative leak determination: sump monitoring, air cooler condensate flow monitor, and primary coolant inventory. However, the third system is only accurate if used to determine the leak rate for a 6- to 24-hour time period. 'Ihis system would be unacceptable for measuring i gpm in I hour. The best systems for leak detection and location are tape moisture sensor and acoustic monitoring, but these systems have poor capability for measuring the laak rate.

Our review of the reportable leak events during the period 1985 to 1996 indicates that the leak detection systems are used to detect a leak and determine its rate but not for determining its location, which is accomplished by visual examination. A brief description of the capabilities of the systems listed in Table 10-2 follows. The description is based on the study group report and is supplemented with recently available information.

Air Radioparticulate Activity Monitor. The containment air radioparticulate activity monitoring system monitors the radioactivity of the particulate by continuously obtaining contamment air samples. ,

Sensitivity and response time depend on several factors, including containment vessel free volume; containment background activity, which varies with RCS leakage; RCS activity, which varies with fuel rod failures; and isotope plateout, which varies with leak location and path to sampling point. The major '

contributors to the response are Rb-88 and Cs-138, which result from tramp uranium on the fuel cladding.

Figure 10-6 shows a minimum detectable leak rate and response time for this system assuming no fuel failure and 1.5x10'2 Bq/cm2 corrosion products in coolant (Aoki 1991). The minimum detectable concentration of 2

the radiopar+iculate monitor is around 4x104 Bq/cm . The figure shows that a minimum of 0.1 gpm leakage can be detected in less than 10 minutes. Duke Power (1997a) reports that the air radioparticulate monitor at Oconee 1 is sensitive to a wide range ofleakage rates, from as low as 0.1 gpm to greater than 30 gpm, provided that corrosion product activity is present and no fuel failures have occurred. Similar leak detection DRAFT - NUREG/CR-6582 10-6

LEAKAGE DETECTION SYSTEMS was also operational at Oconee 2 during the leak event discussed in Section 3.4. Table 10-3 from Aoki (1991) summarizes the leak rate measuring capabilities of this and the next three leak detection systems.

This system will not provide any response to reactor coolant leakage if the background radiation level is high, indicating fuel failures. In addition, this system cannot determine the leakage rate because the particulate activity concentration is determined by unsteady conditions such as background level, paniculate activity concentration in the coolant, and the partition factor of particulates.

Radioactive Gas Activity Monitoring System. The containment radioactive gas activity monitoring system determines the gaseous radioactivity of the air sample after passing through the particulate filter in the air radioparticulate activity monitoring system. Sensitivity and response time depend on the same factors identified for the radioparticulate monitor with the exception ofisotope plateout. The

~

major contributor to the response is Xe-133. The minimum detectable concentration for the gaseous monitor is around 2x10'2 Bq/cm2 . The sensitivity of the system depends on both the gaseous and background radioactivity. The gaseous radioactivity is determined from the containment free volume and the gaseous

~

activity concentration of the reactor coolant, which depends on the fuel failures. The background activity is from Ar-41, which always exists during power operation. The gaseous activity in the contamment atmosphere increases with the duration ofleak, as shown in Figure 10-7 ( Aoki 1991). 'Ibe data in this figure are shown for activity of Xe-133 in reactor coolant equal to 4 x 10' Bq/cm' and containment vessel free volume equal to 73,'/00 m'. The figure shows that a 2-gpm leak may be detected in about 40 minutes and a one-gpm leak in about 80 minutes. In the absence of fuel failures, the corresponding leak detection time would be longer.

Air Cooler Condensato Flow Monitor. This system collects the liquid runoff from the contamment air recirculation cooling units and the CRDM cooling units, and measures the flow rate of condensate from these units. If an is:nthalpic change of the reactor coolant leakage is assumed, then approximately 40% of the leakage mass is discharged as steam into the containment atmosphere and the remaining 60% mass is dropped in the form of water. The leaked steam raises the humidity of the containment air recirculation cooling units and the resulting increase in the condensate flow rate. Under normal operating conditions in a typical PWR, condensate can be generated in a short time after leak occurrence, and then condensate can reach a steady state in approximately 30 minutes, as shown in Figure 10-8 (Aoki 1991). So, a 1-gpm leakage from a high energy line can be detected within an hour (Aoki 1991, USNRC PWR Pipe Crack Study Group 1980). It appears that the French positimith this leak detection system is opposite of the Japanese and U.S. position presented here. Cassette et al. (1986) reports that the French PWRs were no longer equipped with this leak detection system because the system was not considered to be effective either for quantification or detection of any primary leakage.

The leak detection system provides inaccurate results if the plant operating conditions, such as changes in service water temperature and flow rate, heat load, containment purge rate, and ventilation flow rates, enange. The changes affect the condensation rates and, therefore, the corresponding estimates for leakage rate would be inaccurate.

Sump Level and Flow Monitor. Since any leakage in the containment vessel will flow into the containment sump, leakage would be indicated by a level increase in the sump. The cross section of the sump determines the sensitivity of this system. At Oconee 1, the sump capacity is 15 gallons per iach of height and each graduation on the level indicates 0.5 in. of sump height. So, this monitor is capable of 10-7 DRAFT - NUREGICR-6582

LEAKAGE DETECTION SYSTEMS detecting changes on the order of 7.5 gallons ofleakage into the sump. A 1-gpm leakage into the sump, not > '

from the primary system, can be detected within less than 10 minutes. A similar system was also operational l

, during the Oconee 2 leak event discussed in Section 3.4i In a typical Japanese PWR, the level change rate -

in the sump is about 6% cf the level indicator span per 15 minutes in case of a 3.8 Umin (1 spm) leakage into the sunp(Aoki 1991).

i If the source pfleak isfar from the sump, the transport lag time can cause a delay in chaning the sump level, The leakage may be absorbed by the thermal insulation, if the insulator is made of water-absorbable

4 '

material, and the absorption may result in a significant delay in leak detection. However, if metallic insulation is installea, and ifit is divided into small pieces to facilitate inspection of the p* ping, the leakage

. could leak out through the gaps bermen the small pieces. Calculations show that the leakage can leak out L - in about 10 rainutes(Aoki 1991). ,

Leaked steam is absorbed by the containment environment and then condensed by the air cooler, as - ;

~

previously described. The portion of the leakage that remains as water, may drop on the containment floor ,

and flow tc, the drain line in a short time since the floor has a slope to the drain lines. The drained water can

. flow to the containment sump within a few minutes because the lines are slopped to the sump. The slope ,

of the containment floor and the drain lines in a typical Japanese PWR is 1/100, so the drained water can be detected in a short time (Aoki 1991),

ne containment sump monitoring system can determine the leakage rate more precisely after the condensate flow reaches steady state, since the condensate flow from the air cooler condensate flow monitoring system flows to the containment sump.

Primary Reactor Coolant inventory. The closed-loop design ofPWR plants permits maintenance of a coolant inventory that is constant except for controlled additions, controlled discharges (identified leakage), and uncontrolled leakage (unidentified leak.ge), ne primary coolant inventory loss may be determined by estimating the weight balance of the entire primary coolant or by comparing the volume

control tank (VCT) makeup and charging rates. The weight balar.ce of the primary coolant may be estimated by monitoring pressurizer and VCT level changes, makeup and boric acid integrator changes, and system temperature changes during the same period of time. However, the large volume of the primary RCS" and instrumentation inaccuracies makes it impossible for this method to have the sensitivity to detect a 1-gpm i leak in 1 hr. The sensitivity and accuracy of the method increases when the reactor is in steady state

operation and any makeup or release are avoided. Under these conditions, Cassette et al. (1986) reports that a leak of about 150 Uh (0.66 gpm) can be detected in 2 hours2.314815e-5 days 5.555556e-4 hours 3.306878e-6 weeks 7.61e-7 months and 75 Uh (0.33 gpm) detected in 4 hours4.62963e-5 days 0.00111 hours 6.613757e-6 weeks 1.522e-6 months .

The method based on VCT makeup and charging rates requires timing the makeup pump operation. Since the volume between the level setpoints of the makeup tank is larger than the quantity of water lost from a 3.8-Umin (1 gpm) leak in I h, the response time of this system to deteet a leak of I gpm is about 6 hours6.944444e-5 days 0.00167 hours 9.920635e-6 weeks 2.283e-6 months (USNRC PWR Pipe Crack Study Group 1980).

Tape Motsture Sensor, his leak detection system uses a sensing element, which is normally placed

- next to the insulation of the piping. De element provides an electric signal activated by moisture and may 4

- " The total volume of pnmary coolant for a standard Combustion Engineering plant is about 75,000 gallons. The coolant volume for a 4-loop Westinghc9se plant is about 100,000 gallons.

DRAFT - NUREGICR-6582 10 8 7 - - ---- + -~ - g w m 3me TT- sv r wtz-r-2,- - >-- wr- h - ea -.= - ---a - -- ------m.-

LEAKAGE DETECTION SYSTEMS be used with an indicating device. The sensitivity for detecting the leakage and the capability for locating it are reported to be good; however, if the leakage runs down between the pipe and insulation prior to surfacing, the accuracy in locating die leak by this system may be questionable (USNRC PWR Pipe Crack Study Group 1980). His deficiency may be corrected if the sensing element is placed directly on the pipe wall under the insulation, which can be easily done if the insulation is removable he moisture sensitive tapes have been installed at several U.S. nuclear power plants.

Temperature Sensor. The sensitivity and response time of this leak detection system depends on the volume to be monitored, thermal transport distance, heat losses, heat loads, and sensor time constant. This method is best suited for determining the location of leakage from confined spaces and for special applications such as detecting leaks from relief valves and seal leak-off lines. Temperature sensors are installed on the relieflines in French PWRs to detect leaks from upstream and downstream of the pressurizer relief and safety valves. The sensors activate the alarms in the control room.

Visuallnspection. This leak detection method has been used to detect small leaks (<0.2 gpm) inside the containment. nese leaks were detected and located because of boric acid deposits at the leak sites. This method is mainly used for detecting the leaks outside the containment as shown in Figure 10-3. De majority of reportable leak sources are located with this method as shown in Figure 10-4. His method is mainly used to determine whether a given leak is a pressure boundary leak.

Other Leak Detection Monitors and Sensors. Two other monitors and a sensor mentioned in Table 10-2 are humidity-dew point monitor, containment pressure sensor, and acoustic emission monitor.

The sensitivity of the humidity-dew point monitors depends on containment size and proximity of the leak to the monitor. In large containments, a leakage of I gpm does not change the bulk dew point temperature sufliciently to be detected by these monitors. Similarly, in large containments, the pressure sensors cannot detect a leakage of I gpm because such a small leakage would only produce pressure charges that are within the range of normal containment pressure fluctuations. EtTectiveness of the acoustic emission monitors is discussed in the next section.

10.4 Capabilities of Advanced Leakage Detection Systems The advanced leakage detection systems are focusing on two aspects ofleakage detection: early detection of leakage having a smaller rate [<3.8 Umin (<1 gpm)] and locating the leak while the plant is operating.

Early detection ofleakage from the reactor coolant pressure boundary is vital for reliable plant operation and

, accident prevention. Reliable detection of a leak smallcr than 3.8 Umin in less than a hour may allow the application of the leak before-break concept to a smaller diameter [<102-mm (<4-in.)] piping. Use of a detection system to locate a leak while the plant is operating would help avoid a containment entry by plant personnel to determine the leak location, which is a slow process, and facilitate an earlier identification of whether the leak is a pressure boundary leak. Capabilities of three advanced leakage detection systems are summarized here: Nitrogen- 13 (N 13) monitor, acoustic emission monitor, and local humidity monitor. The sensors for these devices may be installed near welds and base metal sites that have unrepaired weld indications, and locations on the pressure boundary that are susceptible to cracking as indicated by field experience.

Nitrugen-13 Monitor. The small leak from the cracked CRDM nazzle at Bugey 3 plant identified a need for a sensitive leak detection system so that the presence ot'a throughwall crack in the nozzle pressure 10-9 DRAFT - NUREG/CR-6582

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

LEAKAGE DETECT 10N S' YSTEMS -

I

' boundary can be detected when it is still small. At the time of the leak, no such systems were installed at any PWR. As a result of the Bugey 3 nonle cracking, the French safety organization now requires a )

continuous and accurate monitoring of CRDM noule leakage so that a leak of 1 IJh (0.0044 gpm) or less . i

- can be quantified within I hour (Champion and Chauvel 1993).

EDF has developed a' system based on continuous monitoring of nitrogen 13 (N-13) to satisfy the requirement. N 13 is generated in the reactor core, and its generation rate is directly proportional to the 4

ruclear power level.' The leak rate can be estimated by measuring N 13 activity in air. De halflife ofN 13 is about 10 minutes, which is long enough to allow detection but not long enough to accumulate in the containment. The system is called Vessel Integrity Control, which uses a N 13 Sensor (VICNIS). Twenty-five of these systems have been installed at French PWRs to date, his system can detect a leak of 1 IJh in

I hour. The system has high sensitivity, quick response time, and reliability. De system can detect a leak

2 rate as low as 0.2 IJh (-0.001 gpm), and in the case of a rapid increase in the leak rate, an alarm can be ,

triggered in a few tenths of a second. The monitoring results are not affected by temperature variation or - i

- sensor location (Champion and Chauvel 1993). ,

. The installation of the N 13 monitoring system required a modified insulation system for the vessel head, which consists o.f austenitic stainless steel insulation with over 30 panels per head, ne panel joints are sealed to minimize gas leakage. A piping system for sampling the gas is incorporated into the panels (NEI 1993a).

AcousWe Emission Afontfor. When pressurized water escapes through a metal boundary, continuous structure-born noise is generated. De source of energy for this noise is the fluctuation in the flow momentum of the fluid. The noise is propagated through the affected components such as pipe and pressure vessels and is measured by the acounic emission monitors attached to these components. The magnitude of the measured signal coi+mds to the leak size. The leak noise attenuates as the distance from the leak noise increases. The leak location can be determined by comparing the leak noise measured at various monitors and the time of the measurement.

Kupperman et al. (1989) estimates that approxhnately 150 sensors are required to completely monitor the pressure boundary piping of a PWR. This estimate is based on an assumption that the piping is 150-m long divided into three sets of piping lengths having different levels of acoustic background noise: 60-m length at low ,60-m length at moderate , and the remaining 30-m length at high-background level, ne number of sensors can be significantly reduced only if the susceptible locations on the primary pressure boimdary, such as locations susceptible to fatigue and SCC damage, are monitored. ,

. Kupperman et al. (198'/) reports that the acoustic emission monitors have been installed at several U.S.

reactors. Numerous low-frequency acoustic emission monitoring systems (with high temperature accelerometers) have been used to monitor valve stem leakage, discussed in Section 7, since 1974 at U.S.

- nuclear power plants. De related field experience indicates that a leak as small as 0.5 gpm can be detected.

Acoustic emission monitors are also installed on a 28-in. recirculation line elbow at a U.S. BWR plant.

Acoustic emission monitors may be potentially used for characterizing the type of throughwall crack, whether it is an IGSCC or a fatigue crack that is leaking. The laboratory tests show that the relationship 6twa the acoustic signals from fatigue cracks and the leak rate are different than that for IGSCC cracks.

DRAFT - NUREG/CR-6582 10-10

i l

LEAKAGE DETECTION SYSTEMS In addition, acoustic signals from fatigue cracks (as well as for two valves and a flange th-t were tested) are more dependent on frequency than those from IGSCC crccks (Kupperman et al.1987).

Loca/ Humidity Monitor. Leakage from a pressurized component increases the humidity in the space surrounding the component. The increase in the humidity is the highest near the leak locr. tion, right under the insulation. This increase in local humidity can be measured with a kmidity sensor installed under the insulation along the pipe outer surface. The sensor may be used for detecting and locating a leak.

Siemens/KWU has developed an automatic monitoring system, called FLOS, employing these sensors.

FLOS includes a temperature- and radiation-resistant sensor tiibe (a flexible inctal hose) several meters long.

Porous sintered metal elements are placed at intervals of about 0.5 m along the sensor tube. Initially, dry air is injected into the tube. Moisture from a leak diffuses through the metal elements. The contents of the tube are pumped at fixed time intervals (10 to 15 minutes) through a central moisture sensor, which measures the absolute humidity level (dew point) as a function of time. By recording the time difference between start of pumping and recording of a humidity peak, the position of the leak along sensor tube can be determined by using the known air velocity inside the tube. The leak rate can be determined from the humidity profile.

The larger amount ofleakage gives a higher peak. The correlation between the leakage rate and the height of a humidity peak depends on the type ofinsulation and temperature (Streicher et al.1986, Jax 1994).

FLOS has been tested over a year at Bohunice, a VVER-440, in the Slovak Republic. The experience shows that the system can detect leakage smaller than 0.1 litters (0.026 gallons). The leaks can be located within

1% of the sensor tube length. So, for a 100-m-long sensor tube, a leak can be located within *1 m of the actual position (Jax 1994).

FLOS has been qualified for a German PWR to detect a potential leak from the reactor pressure vessel closure head. One sensor tube was placed inside the vesse.1 head insulation and another one outside. The two tubes were combined to eliminate the effect of normal fluctuations in the ambient humidity. A mockup was used to evaluate the use of FLOS to detect the leaks under the insulation of a reheater vessel. The results show that this system can reliably detect a leak rate of 0.05 L/h 0.013 g/h). The response time was a little as 15 minutes. The results also confirm the correlation between lea .kage rate and dew point (Jax 1994).

e 10-11 DRAFT - NUREG/CR-6582

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

i e

e o

e 1412

11. FINDINGS AND CONCLUSIONS ne main objectives of this study are to review U.S. experience related to PWR primary system leaks and their rates, how aging affects leakage rates and trends, and the safety significance of such leakages.

Another objective of the study is to assess current leak detection methods. A review of experience related to steam generator tube leaks is not included. Five specific actions were taken to accomplish the objective:

(a) review oflicensee event reports (LERs) related to leak events, (b) development of a database to identify trends, distributions, and causes of leak eveats, (c) visits to PWR plants, (d) resiew of related technical literature including USNRC communications and reports prepared and/or submitted by licensees, and (e) detailed analysis of selected leak events.

ne scope of the study was to review the LERs related to PWR primary system leaks submitted during the period 1985 through the third quarter of 1996, representing about 638 opchating years for U.S. PWRs.

The review included only those leak events that occurred during hot shutdown, hot standby, sta tup, and

, ' power operation. The review did not include steam generator tube leaks.

Our assessment of field experience related to PWR prunary coolant leaks is that the USNRC licensees have apparently taken effective actions to reduce the number of reportable leak events, Out of 199 reportable leak events occurred during the study period, about 60% of the events (121 events) occurred during the first 4 yr of the study period and the remaining 40% occurred during the last -8 yr. One main reason for this decreasing trend is elimination of reportable leakage from valve packing. No packing leakage has been reported since 1991. Ilowever, the frequency of pressure boundary leaks did not show any statistically significant trend during the study period.

Our assessment of risk significance of the leak events identified several events that can be regarded as precursors to LOCA. The risk impacts observed in the prunary system leak database developed in this study are accurately reflected in the typical full-scope PRA. Our review of the leak events does not indicate any dramatic difference between the standard perception about the risk of a SBLOCA and the outcome of this study.

Current leakage detection systems, according to our assessment, are effective in detecting 1 gpm leaktge inside the containment within an hour. However, these systems are not effective in detecting a very small leakage or in determining a leak source location inside the containment. Visual inspection is generally relied upon for detecting a very small leakage and determinmg the leak locations.

Our specific findings are presented as answers to questions posed in Section 1.

Question 1 Are there data to support the existence of rneaningful(statistically significant) trends or annual rates of primary coolant leaks?

Yes. The study identified the following trends of primary coolant leaks that occurred during 1985 through the third quarter of 1996:

11-1 DRAFT - NUREG/CR-6582

FINDINGS AND CONCLUSIONS :

The study identified a statistically significant decreasing trend in frequencies of the reportable leak events. The decreasing trend is due to a decreasing trend of reportable leak events that occurred inside ,

the mala ===. If the first four years of the analysis are excluded, the reponable leak events since l 1983 do not show any statistically valid trend.

. 1hr frequency data for the reponable leak ennts related to vibratory fatigue reveal a aradalcany signific.mt decreasing tmnd with age (years of operauon), but not in calendar time." Apparently, the l

decreasing trend implies that the vibratoty fatigue failures :re caused by inadequacy of initial design and fabrication, and not aging damage.

The' frequency data for the reportable leak even:s related to valve packing degradation reveal 1 decreasing trends both in calendar time and with age, but the trend in calendar time is statistically more

i significant than with age. Apparently, the decreasing trend is a result of valve uem packing I improvements made in the early 1980s and incorptrating some of them into operating nuclear power plants.

The frequency data associated with two ca'egories of reportable leak events, pressure boundary leaks and leaks through bolted connections, did not show a statistically valid trend.

Question 2. Are there any previously (prior to 1985) unidentified degradation mechanisms or failure modes that have caused primary coolant leaks?

Yes. The study identified the following four previously unidentified degradation mechanisms or failure modes:

Turbulent penetration and resulting thermal cycling have caused throughwall fatigue cracking in the l pressure boundary of PWR branch lines; however, these phenomena are not completely urmerstood.

The affected branch lines are ufety injection lines at Farley 2 and Tihange 1, and a residual haar removat line at a foreign PWR. Apparently, these phenornena have also played a major role in the April 21,1997, leak event at Oconee 2. As a mitigative action, some licensees are planriing or performmg augmanM mservice inspection of welds and base metal locations, susceptible to cracking caused by these phenomena.

Pump-induced pressure pulsations and cavitation have caused vibratory fatigue failures of socket welds on small diameter piping. The role of positive displacement pump induced vibrations was previously ,

identified, but the role of cavi*ating component such as an orifice in causing vibratory fatigue failure of PWR letdown system was not, in one mstance, cavitation induced vibratory fatigue and water hammer played a synergistic role in causing rupture of a letdown system drain line. Quantitative !

< understanding of vibratory fatigue mechanism is being developed.

Pdmary water stress corrosion cracking has caused throu;) wall cracking of prunary pressure boundary penetrations made of Alloy 600 (a nickel-base material). The penetrations include (a) instrument nozzles and heater sleeves for pressurizers, (b) instrument nozzles for main coolant piping, and (c)

" The frequency data for trend in calendar time represent the number of reportable leak events per reactor operating year, whereas the data for trend in age represent the number of events per operating PWR.

DRAFT- NUREG/CR-6582 11-2

FINDINGS AND CONCLUSIONS control rod drive mechanism nozzles at non-U.S. PWRs. De affected U.S. PWRs were designed by either Combustion Engineering or Babcock & Wilcox who employ Alloy 600 penetrations for pressurizers and main coolant piping.

Transgranular stress corrosion cracking has caused throughwall cracking of a spare control element drive mechanism housing at Fort Calhoun, a Combustion Engineering-designed PWR. He crack initiated in a weld overlay region present on the inside surface of the housing, which constitutes a part of the prunary pressure boundary, and propsgated through the base metal. The spare housing was not properly vented, resulting h a high concentration of oxygen. He housing material was Type 347 or 348 stainless steel. The leak was difficult to locate, taking the licensee about 2 months to fmd.

Transgranular stress corrosion cracking of canopy-seal welds has been reported at Westinghouse-designed PWRs, but differs in two aspects from the one at Fort Calhoun: a canopy weld provides sealing to the prbnary system and is not a structural joint, and the housing material is Type 304 stainless steel and the canopy weld material is Type 308L. There had been leaks from canopy seal welds before 1985.

Question 3. Are there any leaks in locations, or of types, not previously identified?

Yes. Several leak locations were previously unidentified. (The answers to Question 3 are related to those to Question 2.)

. Thermal fatigue has caused throughwall cracking in the base metal of PWR branch lines. Cracking was not expected at these sites, and therefore not included in the plant inservice inspection program.

Two examples of such locations include a throughwall crack in a safety irjection line elbow at Tihange 1, and a throughwall crack in the base metal of a straight portion of a safety injection line at Dampierre 1.

- Vibratory fatigue cracking of socket welds in the letdown system piping have been found. Generally, vibratory fatigue is caused by pump-induced vibration, but in this case it was caused by cavitation.

- Primary water stress co rosion cracking has caused thronshwall cracking of Alloy 600 penetrations, listed in response to Question 2. 'bmer, susceptibility of these penetrations was previously unidentified.

- Tnmsgranular stress corrosion e acking has caused throughwall cracking of austenitic stainless steel components. However, susceptibility of these components in a PWR environment was previously unidentified.

Question 4. Which ncs of leaks have a potential for relatively rapid growth after detection, such t%t a SBLOCA occurs?

The follow %g two sets of leaks have a potential for relatively rapid growth after detection, such that a SDLOCA occurs:

11-3 DRAFT - NUREG/CR-6582

FINulNOS AND CONCLUSIONS

. Leaks through thermal fatigue cracks of branch lines.

. . If crack growth rata is bigh and uniform along its length, then a circumferential crack may rapidly propagate through the intact ligament of the pipe cross section after a leak is detected.

De recent experience at Dampierre 1 indicates that a small, nondetectable fatigue crack in a safety injection line could become a throughwall crack in one fuel cycle, indicating a high crack growth rate. The recent cracking at Oconee 2 suggests the possibility of a uniform cract growth. The growth of the circumferential crack was uniform (30% throughwall) over about 78 % of the circumference.' he faster growth in the throughwall portion of the crack was due to an applied cyclic bending moment, possibly from local stratification. In the absence of such a bending moment, the crack growth would have been uniform along the entire circumference.

Leaks through reactor coolant pump seals.

- An initiaUy small leak caused by a failure of a lower seal stas;c may grow rapidly if tb- ,

remaining seal stages fail. Such an event could con:ribute directly to the SBLOCA initia+.ing event frequency. For example, a vibratory fatigue failure cif a reactor coolant pump seal sensing line at Arkansas Nuclear One Unit 2 caused a complete failure of the pump seal that resulted in a maxinnun leak rate of 152 Umin (40 gpm).

Question 5. Is fatigue of aging pipe becoming safety significant?

Yes. Thermal fatigue of PWR branch lines is becoming a safety issue for the following reasons:

(Supporting information may be found in the re.sponse to Questions 2, 3, and 4.)

Susceptible sites (welds and base metal) are generally not inspected. Additionally, it is difficult to detect cracks at these sites during inservice inspection when the plant is shutdown. Often a throughwall crack has been detected because of Icakage and not inservice inspection.

If fatigue crack growth rate is about uniform over the entire circumference of a branch line cross section, a crack may rapidly propagate through the intact ligament of the piping after a leak is detected, leading to pipe rupture. Higher fatigue crack growth rate will lead to pipe rupture sooner after the leak is detected.

The leak-before-break concept is not applicable to a small-diameter piping. The concept is also ,

not applicable to piping subject to fatigue damage.

Question 6. Could any of the reportable leak events be regarded as a core damage

- precursor?

Yes. Bree reportable leak events were identified and analyzed as precursors to LOCA by the USNRC Accident Sequence Precursor (ASP) Program; one event contributed to transient-induced LOCA probability and two contributed directly to the SBLOCA initiating event frequency. The most typicas transient-induced LOCAs involve a stuck-open relief valve or a reactor coolant pump seal failure. These failures occur in the presence of an cff-normal condition such as a pressure transient or loss of offsite power. Risk review DRAFT - NUREG/CR-6582 11-4

FINDINGS AND CONCLUSIONS of the database identified five additional leak events that contributed to transient-induced LOCA probability.

During the other two leak events, the leakage wa large enough or had a potential to develop to a size to require safety system resporise as modeled by a PRA SBLOCA event tree. Rese type of events also includes small leaks that, if unattended, have a potential to become large enough to require safety system or operator response. For example, a small reactor coolant pump seal leak could grow to become a SBLOCA if multiple seal stages fail.- Risk review of the database identified five additional events that contributed to the SBLOCA initiating event frequency.

Conclasions about the overall perception of the contribution of RCS leaks to core damage follow:

The frequency of leak mduced transients is so small (6.3E-3/ reactor year) in comparison to the overall transent uutiating event frequency (1 - 2/ reactor year) that there is no discernable influence on the core damage frequency,

Re frequency of transient-induced relief valve LOCAs from the operational data (6.27E 3/ reactor yezr) is wnhm the range of calculated values observed in typical PRAs,

The frequetry of reactor coolant pump seal leaks becoming small LOCAs is so small that none exist in the da'ahaw, i

Overall, the risk impacts observed in the primary system leak database developed in this study are accurately reflected in the typical full-scope PRA.

Question 7. Does the review of leak events indicate that the risk of a SBLOCA is greater than previously hypothesized?

No. De industry leak experience has always been the basis for SBLOCA frequency calculations and any dramatic difference between the standard perception and the outcome of this study would be unexpected.

The frequency of SBLOCA from this study (0.011) is nearly identical to the generic SBLOCA initiating event frequency (0.01) used in many PRAs.

Question 8. How effective are the current leakage detection systems?

Our main finding is that a use of several different leak detection systems is required for reliable characterization ofleakage, Other findings follow:

The current leakage detection systems are effective in detecting 1 gpm leakage inside the containment within an hour. Ilowever, several small leaks [<0.38 L/ min (<0.1 gpm)] were detected by visual inspection.

Leakage detection systems are currently not used in detecting primary system leakage outside the containment. Visual inspection is generally relied upon for detecting this leakage, t

l 11-5 DRAFT - NUREG/CR-6582 l

1 l -

.. - . . = . - - - - . . . - - - . . . . - - . - . . . ..-.- . _ _ - - - .

i FINDINGS AND CONCLUSIONS ' l I

- ' The current leakage detection systems are not effective in locating the leak source hende the manianare. Generally, waninment entry and visual inspection are required. During some leak events, it takes considerable time to determine the loution of the leak source while the plant is opetating; therefore, a leak source, poss&ty located on the pressure boundary, may remain 2 unidentified for that period of time.

- Three advanced leakage detecnon systems can detect a very small leakage and locate a leak so. tree bec.ause they are connected directly to the pressure boundary: - Nitrogen-13 monkor, acoustic emission monitor, and local humidity monitor. Nitroge>13 monitor can detect a leak rate as low as 0.2 L'h (-0.001 spm), and can detect a leak of 1 IJh (0.005 spm) in 1 bour. These monitors have been installed at several French PWPa.

- An acousuc emission monhoring system can 'de: ermine the leak location by comparing the leak -

nois; tueasured at varica monsters installed on the primary pressure bounda;y, and the time of --

the measurements 'hese monitors have been installed at U.S. nuclee. power plants to monitor . , ,

valve stem leakage.1hese monitors can detect a leak rate as low as 0.5 gpm. Acoustic emission monitors are also installed on a 711-mm (28-in.) recirculation line elbow at a U.S. BWR:

F

local humidity monitors are installed directly along the outside surface of the piping under the =

insulation. It can detect a laakage by measuring a local increase in the humidity and determine the leak location. This detection system was tested at a VVER440 plant in the Slovak Republic. The

- experience shows that the system can detect leakage smaller than 0.1 litters (0.026 gallons), and determine the leak location within 11% of the sensor tube length.

- The local humidity monitoring system is qualified for a German PWR to detect a small potential leak from the reactor pressure vessel closure hud. The test results showed that this system can reliably detect a leak rate of 0.05 IJh (0.0002 gpm) with a response time of as little as 15 rnianas.

A f

DRAFT- NUREG/CR-6582 11-6

i

-12. REFERENCES AASHTO 1989. Guide Specipcationfor Fatigue Design ofSteel Bridges.

ANO 2 (Arkansa: Nuclear One Unit 2) 1988. "Unisolatable Reactor Coolant System Leak Caused by Vibrrion Induced Fatigue FaDure of A Reactor Coolant Pump SealInstrumentation Line," Licensee Event Report 368-88-011, September 8.

Aoki, K.1991. " Reactor Coolant Pressure Mary Leak Detection Syuems in Japanese PWR Plants,"

Nuclear Engineering andDesign, 128 (1991), pp. 3542.

ASME 1992. Metal Fatigue in Operating Nuclear Power Plants, ASME Section XI Task Group on Fatigt.3 in Operating Plants, S. Leshnoff Chairman, EPRI TR 100252, Electric Power Research Institute, Palo Alto.

_ 'ASME 1994. Requirementsfor Preoperational and initial Stan Up Vibration Testing of Nuclear Power Plant Piping Systems, ASME OM SIG-1994 Standard.

' SME 1995a. ASME Boller and Pressure Vessel Code.Section III, " Rules for Construction of Nuclear 1-ower Plant Components-Division 1-Subsection NB, Class 1 Components." American Society of Mechanical Engineers, New York.

- ASME 1995b. ASME Boiler and Pressure vessel Code.Section XI, " Rules for Inservice inspection,"

American Society of Mechanical Engineers, New York.

Bartal, Y., J. Lin and R. E. Uhrig 1994. " Leakage prediction in nuclear power plants using similarity formulac and genetic algorithms," Transactions of the Nuclear Societies ofIsraelJoint Meeting, Vol.18, CONF-9411188.

t-Bartholom6, G., W. Kastner, and E. Keim 1993. " Design and calibration of leak detection systems by thermal hydraulics and fracture mechanics analysis," NuclearEngineering andDesign, Vol.142, pp.1-13.

Bell L. G., and P. D. O'Reilly 1992. Operating Experience Feedback Report - Experience with Pump L Seals installed in Reactor Coolant Pumps Manufactured by Byron Jackson, NUREG-1275 Vol. 7 l September.

Bieselt, R., B. Kuckartz, and M. Wolf 1995. " Determination of limits for smallest detectable and largest subcritical leakage cracks in piping systems," Nuclear Engineering and Design, Vol.159, pp. 27-42.

Bush, S. H.1992. " Failure Mechausms in Nuclear Power Plant Piping Systems," ASME Journal of Pressure Yessel Technology, Volume 114, pp. 389-395, i

12-1 DRAFT - NUREG/CR 6582 L

l . . _ - - . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

1 REFERENCES :j 4

i j

BWNT (Babcock & Wilcox Nuclear Technologies) 1992. 'Oconee Strati &**ian Data - 6/90 HU,"

Proprietary. =

Cassene, P., C. Giroux, H. Roche, and J. J. Seveon 1986. ' Evaluation of Primary Coolant I. mats and Aasaanmaar of D-ian Systems in Continuous Surveillance of Reactor Coolant Circuit Integrity,*. in Continuous Surwillance ofReactor Coolant Circuit integrity, Proceedmgs of A: CSNI Specialist Mtg.,

London ' England (August 1985), Nuclear Energy Agency, Paris, France,1986, pp.165-178.

Cepkauskas, M. M. and Garcia, C. M.1994. " Valve Packing 'itudy," Proceedings of de 7hird NRC/ASME Synposiw on Valve and Punp Testing, NUREG/ CPS 137, Vol.1, pp. 527-545. -

Champion G. and L. Chauvel,1993. 'EdF Installs N-13 Leak Monitoring,' Nuclear Engineering. ,

j international, October. \

Champion G., L. Chauvel, and J. Leon 1993. 'VICNIS, A New Radiation Monitoring Cimanel for Checking Vessel head Cover Penetration Integrity," Transactions ofde American Nuclear Society,1993 Winter Meetag, Vol. 69, p. 406 Combustion Engmemng I991. Genericissue - 23, Evaluation ofthe Reactor Coolant Pun =p SealIntegrity ,

Issue, CEN.408, prepared for Combustion Engineering Owners Group, September.

Connecticut Yankee Atomic Power Company Surveillance Mcedure 1995. ' Primary System I.eakage Surveillance,' SUR 5.12, Rev.16, August.

D. C. Cook 1995. D. C. Cook Nuclear Plant Unit 2 Technical Speciffcanons, November 3. ,

Crawford Fitting Company 1985. Swagelok 7kbe Fitting andinstallation Manual.

Deardortf,

A. F. and J. K. Smith,1994. Evaluations of Conservatism andEnvironmentalEfects in ASME Code Section ill, Class 1 Fatigue Analysis, SAND-94-0187, UC-523, Sandia National 12boratories.

Diablo Canyon 1990, ' Control Room Post LOCA Habitability Design Basis Potentially Exceeded due to Leakage Through a Vibration Induced Crack in Potentially CVCS Piping," LER 275-90-10.

Dickey, V. J., J. Dimmick, and P. M. Moore 1978. " Acoustic Measurement of Valve I enkage Rates,"

Mater. Eval.,3f(1),67.

Dimmick, J. G., and J. M. Cobb 1986 ' Ultrasonic Leak Detection Cuts Valve Maintenance Costs,"

Power Engineering, Vol. 90(8), August, pp. 35-38.

Druin, M. T., et nl.,1987. Analysis of Ccre Damage Frequencyfrom Internal Events: Methodology .

Gudelines, NUREG/CR 4550, Volume 1, September.

! Duke Power 1991. " Excessive Reactor Coolant Leak, Reactor Trip, and Inadvertent Protective System Actuation Result from Management Deficiencies and Equiptnent Failure," Licensee Event Report 287 '

008, December 23.

DRAFT - NUREG/CR 6582 12-2 l

E _ . _ _

REPERENCES Duke Power 1993. 'A Manual Actuation of an ECCS Pump Due to a Leak on the Charging I!eaderl

, . Caused by a Construction /Iparallation Deficiency and a Management Deficiency," LW Event Report '

~

I 370 93 003, April 22, Duke Power 1997a. Letter from J. W Hampton to USNRC, Justification for ContimiM Operation (JCO) of Oconee Unit 1 Based on Oconee Unit 2 HPI Line Leak, Docket Nos. 50 269, -270, 287, May 2.-

Duke Power 1997b. Letter from J. W. Hapton to USNRC, Justificadon for ContimiM Operation (JCO) of Oconee Unit 1 Based on Oconee Unit 2 HPI Line Leak, Oconee Nuclear Station, Docket Nos. 50-269, +

270, 287, May 7.

Duke Power 1997c. Letter from J. W. Hampton to USNRC, Response to Request for Additional Information on the High Pressure Irpection (HPI) System, Oconee Nuclear Station, Docket Nos. 50-269, 270, -287, May 8.

Duke Power 1997d. Letter from J. W. Hampton to USNRC. Response to Request for Additional Informadon en the High Pressure h$ection (HPI) System, Oconee Nuclear Station, Docket Nos. 50-269,

-270, -287, May 12.

Duke Power 1997e. Letter from J. W. Hampton to USNRC, Justification for Continued Operation (JCO) of Oconee Unit 1 Based on Ocopec Unit 2 HPI Line Leak, Supplemental Information, Oconee Nuclear

- Station, Docket Nos. 50-269, 270, -287, May 16.

Duke Power 1997f. Letter from J. W. Hampton to USNRC, Justification for Continued Operation (JCO) -

of Oconee Unit 1 Br. sed on Oconee Unit 2 HPI Line Leak, Supplemental Information, Oconee Nuclear '

Station, Docket Nos. 50-269, -270, -287, May 22.

Duke Power 19973. Letter from J. W. Hampton to USNRC, Justification for Continued Operation (JCO) of Oconee Unit 1 Based on Oconee Unit 2 HPI Line Leak, Supplemental Irformation, Oconee Nuclear

Station, Docket Nos. 50-269, -270, -287, June 5.

Duke Power 1997h. Letter from J. W. Hampton to USNRC, Response to Request for Additional information on the High Pressure Irpection (HPI) System, Oconee Nuclear Station, Docket Nos. 50-269,

-270,-287, May 9.

L 4 Duke Power 19971. Letter from W. R. McCollum to USNRC, Inservice Inspection Pr'> gram, Third Ten-Year Inservice Inspection Interval, Generic Letter 85-20 Supplemental Information, Docket Nos. 50 269,

-270,-287, August 6.

4 EPRI (Electric Power Research Institute) 1993. TASCS Workshop, Orlando FL, June 1718.

(a) B&W Owners Group Perspective by H. Cordie (b) Overview of TASCS Test Program by D. Roarty (c) ANO-2 Shutdown Cooling Line Stratification by J. Martin (d)- z TASCS Screenmg Criteria by D. Roarty :

! (e) TASCS Evaluation Methodology by D, Roarty and M. Roidt (f) CEOG Program to Address Thermal Stratification in Piping Connected to the RCS by

. B. T. Lubin 4

12-3 DRAFT- NUREG/CR-6582-

REFERENCES EPRI 1994. Fatigue Management Handbook, EPRI TR 104534.

EPRI 1995. Seminnt on Operating Fower Plant Fatigue Damage Assessment, Burlington, VT, August 9-11.

Fischer, K., and G. Preusser 1991. " Methods for Leak Detection for KWU Pressurized and Boiling Water

- Reactors,* Nuclear Engineering and Design, Vol.128, pp. 4'4-49.

Florida Power & Light 1995. ' Hurricane Erin at St. Lucie," License Event Report 335 95-004, September 27.

Flynn, A., et al.1975 ' Thermal sleeve Failure and Repairs -Indian Point # 1 Nuclear Unit (285 MW),"

Nuclear Technology, vol. 25, Januaty, pp.13 31.

Friesel, M. A., and 3. F. Dawson 1992. Acoustic Emission Monitoring ofHFIR VesselDuring Hydrostatic ,

Testing, PNie8277, August.

Fukuda, T. et al.,1995. "Early detection of leakage in reactor coolant system of miclear plant using Kalman fitter," Transactions of the Japan Society ofMechanical Engineers, Part C, Vol. 61 No. 588, pp.

3349-3354, August.

Garrity J. H., (Maine Yankee Atomic Power Company) 1983. Letter to the U.S. Nuclear Regulatory Commission, " Main Feedwater System," February 17.

Gentillion, C. D., et al.1994. Ratesfor Initiating Events in U.S. Commercial Nuclear Power Plants,

- 1989-1993, INEL-94/0270, Idaho National Engineering Laboratory, Idaho Falls, ID, December.

Germain, J. L., F. Inisy, and S. Apolzan 1994. Hygrometric Measurementfor On-line Monitoring of Fn'R Vessel Head Penetrations, EDF 95-NB-0031, June.

Gordon, B. M., et al.1987. " Service Experience of BWR Pressure Vessels," Performance and Evaluation ofLight Water ReactorPressure Vessels, PVP-Vol.119, American Society of Mechanical Engineers, New York, pp. 9-17.

Grebner, H., A. Hoefler, and H. Hunger 1994. " Essential results of analysis accorapanying the leak rate experiments E22 at HDR." MPA Seminar: Safety and Reliability of Plant Technology with Special Emphasis on Nuclear Technology-Approaches to Lifetime Extension, CONF-9410334, pp. 47.1-47.23.

Grebner, H., H. Hunger, *md A. Hofler 1994. " Experiments and calculations on crack opening and leak rate of a pipe branch within the HDR Program," Proceedings of the Annual Meeting on Nuclear Technology '91, Bonn Germany, May 14-16,1991, also Nuclear Engineering andDesign, V01.147, No.

1, pp. 79-84, Januany.

Hart, K. A.,1996. " Development of an Effective Valve Packing Program," Proceedings of the Fourth NRC/ASME Symposium on Valve and Pump Testing, NUREG/CP-0152, pp. 2B-35, 2B49.

DRAFT - NUREG/CR-6582 12 4 l

l

REFERENCES Hechmer, J. L., and E. J. Kuhn, HI,1997. " Fatigue Strength Reduction Factor for Welds based on NDE,"

PVP Vol. 353, American Society of Mechanical Engineers, July, pp. 235-242.

. Hessel, G., W. Schmitt, and F. P. Weiss 1994. " Acoustic leak detection at complicate <l topologies using neural networks," in Forschungszentrum Rossendo$, institutefor Safety Research. AnnualReport, F. P.

Weiss ed., FZR-68, June.

Hessel, G., W. Schmitt, and F. P. Weiss 1994. "A new method for acoustic leak detection r3 complicated geometrical structures," Proceedings ofIFAC Symposium on Fault Detection, Supeniston and Safetyfor Industrial Processes, Espoo, Finland, June 1316,1994.

Hessel, G., W. Schmitt, and F. P. Weiss 1995. Leak Detection at Complicated Three-Dimensional Topologies, FinalReport, F2R-67, January.

Hessel, G., W. Schmitt, and F. P. Weiss 1995. " Acoustic leak monitoring using neural networks,"

Proceedings of the 1995 Annual Meeting on Nuclear Technology, CONF-95 05329, May.

Ifiguchi, M., et.al.,1996a, " Effects of Weld Defects at Root on Rotating Bending Fatigue Strength of Small Diameter Socket Welded Pipe Joints," ASME Pressure Vessels and Piping Conference, Montreal, Canada, July 1996, PVP Volume 338, pp. 3-10.

H!guchi, M., et al.,1996b, "A Study of the Fatigue Strength Reduction Factor for Small Diameter Socket Weldui Pipe Joints," ASME Pressure Vessels and Piping Corlference, Montreal, Canada, July 1996, PVP Volume 338, pp.11-19.

HL&P (Houston Lighting & Power) 1996. Letter from S. E. Thoraas to USNRC, South Texas Project Units 1 and 2, Docket Nos. STN 50L.498, STN 50-499, Resolution of NRC Balletin 88-08, " Thermal Stresses in Piping Connected to Reactor Coolant Systems," July 15.

Holler, P. et al.,199L "New approach for location of continuously emitting acoustic emission cources by phase-controlled probe arrays," Nuclear Engineering and De. Ign, Vol.128, pp. 79-82.

Hunger, H., G. Katzenmeier, and H. Grebner 1991. " Experiments on Crack Opening and Leak Rate Behavior of Small Piping Components at the HDR Facility," in Transactions of the lith International Conference on StructuralMechanics in Reactor Technology (SMIRTII), V01. F, pp. 225-230.

Hunstein, N. H., and K. R. Reeser 1994. " Utility perspective on managing primary water stress corrosion cracking of Alloy-600 nozzles: San Onofre Nuclear Generating Station Units 2 and 3," Proceedings ofthe

~

1994 Pressure Vessels and Piping Conference, Minneapolis, MN, June 19-21, PVP Vol. 285, Part 16, pp.

149-153.

Hutton, P. H. et al.,1991. Acoustic Emission / Flaw Relatio.wh>s for inservice Monitoring of LWRs, NUREG/CR-5645, October.

Hutton, P. H.1993. " Listening to Reactor Pressurc Boundaries for the Sounds of Cracks and Leaks,"

Nuclear Engineering International, December, pp. 38-40.

12-5 DRAFT - NUREG/CR-6582

=

REFERENCES INES (The International Nuclear Event Scale) 1997. " Leakage on a Piping Connected to the Primary Circuit. Dampierre NPP. Event Date December 14,1996," May 6.

ISA (Instrument Society of America) 1980. " Nuclear Safety Related Instrument Sensing Line Piping and Tubing Standards for Use in Nuclear Po ver Plants,' ANS1/ISA S67.021980.

ISA 1986.

Sample-Line Piping and Tubing Standard for use in Nuclear Power Plants," ANSI /ISA-567.10-1986.

Jackson, J. E.,1988. " Current Status of Generic issue 23: Reactor Coolant Fump Seal Failures,"

Proceedings: Afain Coolant Pump Diagnostics, EPRI NP-6116, December, pp.1-16 to 126.

Jax, P., and W. Leaker 1992. *New leak detection and location techniques for pipework and vessels,'

VGB Krqftwerkstechnik, Vol 12 No. 6, pp. 514 517.

Jat, P., U. Kunze, and M. Miksch 1993. "New generation of diagnostic systems for nuclear power plants," Transactions of the American Nuclear Society, Vol. 67: Suppl.1, pp. 234 235.

Jax, P.,19N. '* Detecting e.nd Locating the Smallest Leak Early,' Nucleat Engineering International, December, pp. 22-24.

Jax, P., and V. Streicher 1995. " FLUES a new monitoring system for leak detection and location,"

frouedings of the 1995 2nd Regional hiceting on Nuclear Energy in Central Europ ., Nuclear Society of Slovenia, pp. 402-407.

Jenco, J. M.1996. " Advancements in Plant leak Reduction Training,' Nuclear Plant Journal, May-June, 19 26-3?.

Kalyanatuuhram, P., et al.,1992. " Leak detection in pressure tubes of a pressurized heavy water reactor by acoustic emission techniques," Britisn Journal ofNon-Destructive Testing, Vol. 34, No.11, pp. 539-543.

Kelly, M. W. and J. M. Davis 1995. " Vibration issues and Examples," presented at the Electric Power Research tnstitute (EPRI) seminar on Operating Power Plant Fatigue Damage Assessment, Burlington, VT, Augu." 9-11,199L ,

Kim, S. II., W.11. Jang, and S. J. Han 1993. Acoustic Emission hierhod to Detect Cracks and Leaks of the Nuclear Srcam Supply System. KAERI AR 381/93, October.

Kochet., S., and G. Rauch 1927. " Replacement of Positive !)irplacement Charging Pump,' Nuclear Plant Journal, May June, pp.17 21.

Kunze, U., and B. Bechtold 1095. 'New generation of monitoring systems with on-line diagnostics,"

Progress in Nuclear Energy, Vol. 29, No. 34, pp. 215-227.

Kupperman, D. S., et al.1987. " leak Detection Technology for Reactor Primary Systems,' Nuclear Safety, Vol. 28, April-June p.191.

DRAIT - NUREG/CR-6582 12-6

REFERENCES Kupperman, D. S., et al.1989.

Final Evaluation of Advanced and Current Leak Detecting Systems," in Proceedings of the USNRC Sixteenth Water Reactor Sqfety infonnation bireting, NUREGICP-0097 Vol.

2, pp.181 192, March.

Kustu, O., and R.E. Schon' 1980. "Research Needs for Resolving the Significant Problems of Light Water Reactot Piping Systems,* Proceedings ofANS/Eh!S Topicalhireting - DaermalReactor Safety, Knoxville, TN, April 1980.

Lisowyj, B.1993. " Evaluation of Cracking in Type 348 Stainless Steel Control Element Drive Mechanism 1fousings,' Sinh Intemational Sympostwn on EndronmentalDegradation ofMaterials in Nuclear Power Systems - Water Reactors, ne Minerais, Metals, & Materials Society, pp. 343-350.

lefaro, R. J.1991.

Monitoring of LWR Componem Cooling Water Systems

Proceedings of the Nuclear Power Plant and Facility braintenance Toilcal bleeting, Salt Lake City, UT, April 711,1991, NUREG/CR 45757.

1.ofaro, R. J.1992. ' Detecting and Mitigating Aging in Component Cooling Water Systems," Proceedings of the USNRCNineteenth Water Reactor Safety Informarion bieeting, NUREGICP-0119, Vol.1, pp. 275-287.

Loisy. F., J. L. Germain, and L. Channel 1995. 'Pnmary Circuit leak Detection. An application on PWR vessel head penetrations,' in Proceedings of the 7th International Symposium on Nuclear Reactor Sarveillance and Diagnostics (SMORN Vil), oECD-95-00686, Vol.1, p. 5.1118.

leuisiana Power & Light 1985. Letter from K. W. Cook to USNRC, Reactor Coolant Pump Seal leakage, Waterford 3 SES, Docket No. 50-382, April 29.

Lubin, B. T., et al.,1994, " Structural Evaluation of a Shutdo vn CooRng Line Subjected to nermal Stratification due to Turbulent Penetration," ASME Pressure Vessels and Piping Conference, Minneapolis, MN, PVP Volume 286, pp. 45-48.

Lubin, B. T., and Brown, J.,1995, " Thermal Stratification in Lines Connected to a Reactor Coolant System," ASME Pressure Vessels and Piping Conference, Honolulu, llawaii, PVP Volume 313-1, pp.19-31.

Luckas, W. J., et al.1986. Tractor Coolant Pump SealRelatedinstrumentation and Operator kesporue, NUREG/CR-4544, BNL-NUREG-51964, December.

Maeda, K.1993. *1eakage source disenmination method for nucl. e reactor,' Patent No.: JP 5-249278 A, Toshiba Corp., September 28.

Maahnm, W. D., and K. K. Yoon 1987. Application of Leak Before-Break Approach to PWR Piping Dcigned by Rabcock and Milcox, EPRI NP-4972, January, pp. 51.

McElroy, J. W.1986. "Leoust c Leaki Detection in Nuclear Power Plants," in Proceedingst Second EPRI incipient Failure Detection Cortference, EPRl-CS-4748 (October),

12 7 DRAFT - NUREG/CR-6582

REFERENCES McGuire 1988. 'A Safety injection /Reacer Trip Occurred Due to a Design Deficiency of the Main Turbine Controls - Followed by Several Equipment Malfunctions,' LER 369-8741741.

McVicker, J. P., et al.1995. ' Simplified Failure Sequence Evaluation of Reactor Pressure Vessel Head Conoding In-core insttumentation Assembly,' Structuralintegrity of Pressure Vessels, Piping, and Conponents, PVP-Vol. 318, pp. 25 31.

Meyer, l.. C.,1989. Nuclear Plant Aging Research on High Pressure injection Systems, NUREGICR.

4%7, EGG 2514, August, pp. A 1 to A 7.

Millstone 1994.

Reactor Coolant System Pressure leak Due to Instrument Line Socket Weld Failure,'

LER 423-94-012 00. .

Minnrick, J. W., et al.,19u. Precursors to Potential Severe Core Damage Accidents: A Status Report, NUREG/CR 4674 (series), Oak Ridge National Laboratory, Oak Ridge. TN. ,

- Mitra, S., et al.1986. Evaluation of Core Damage Sequences Initiated by I.oss of Reactor Coolant Pump Seal Coothig, NUREG/CR 4643. BNL-NUREG 52003, August.

Morimoto, M.1991. " leak detector for reactor pressure vessel,* Patent No.: JP 3 71095 A, Toshiba Corp., March 26.

Morishita, Y., et al.1995. ' Development of leak Detection System Using High Temperature-Resistant Microphones,' Joumal ofNuclear Science and Technology, Vol. 32, No. 3, pp. 237-244, March.

Narabayashi, T., et al.1991. ' Experimental study on leak flow model through fatigue crack in pipe,'

Proceedings of the Japanese-German Joint Seminar on Structural Strength and NDE Problent in Nuclear Engineering, CONF-8809532, also Nuclear Engineering and Design, Vol.128, No.1, pp.17 27, July.

N\ckels, R. W.199). A State-of-the-Art Review of Continuous Monitoring and Sun'elliance Techniques in Relation to Reactor Pressure CYrcuitIntegrity, Commission of the European Communities, EUR 13409 en.

North Anna 1991. ' Plant Shutdown Required by Technical Specifications due to RHR Valve Packing Failure,' Licensee Event Report No.91-011, November 3.

Northeast Utilities 1991. Letter from E. J. Mroczka to USNRC, Comments on Proposed Resolution to Generic Issue 23: Reactor Coolant Pump Seal Failure, Millstone Nuclear Power Station, Unit Nos.1,2, and 3, October 1.

s Nuclear Power Experience 1973. 'RTD Blow Out Small LOCA,' Vol. PWR 2 V. B.2.

Nuclear Power Experierice 1979. " Flow Transmitter Sensing Line leaked Due to Improper Installation,'

Vol. PWR 2, V.B.11.

DR AFT - NUREG/CR 6582 12-8

REFERENCES Nuclear Power Experience 1992. *lS lir. Alen Declared - Excessive RCS leakage, Reactor Trip, RPS Actuation - Failed Fitting on SG Instmment Line - Management Deficiency Equipment Failure," Vol.

PWR 2 V.D.575.

Nucleonics Week 1991. " Steam Leak at Oconee 3 Dumps 50,000 Gallons of Primary Coolant," November 28, pp. 2 3.

Nymnn, R., et al.,1996. Reliability of Piping System Components, Volume 4: The Pipe Failure Event Database, SKI Report 95:61, Swedish Nuclear Power Inspectorate, Stockholm, Sweden, July.

Oconee 21997. 'Unisolable Reactor Coolant Leak Due to Inadequate Surveillance Program,' Licensee .

Event Report No.97-001, May 21. s Olsen, C. T., and M. S. McCarthy 1994. " Linear Regression Leak Rate Calculatbn IIelps Avoid Premature Shutdown," Power Engineering, December.

Olsen, D. E.,1995.

Piping Vibration Experience in Pow tants," Chapter 7.4 in Pressure Vessel and Piping Technology 1993 - A Decade ofProgress, America - Society of Mechanical Engineers.

Omaha Public Power District (OPPD) 1991. " Leakage Through Control Element Drive Mechanism llousing,* Licenser Event Report 9042840.

Oshima, K.1990. " leak detection device for reactor r,oolant," Patent No.:IP 2 96695 A, Toshibs Corp.,

April 9.

Peter, A., and M. Zsenel 1992. Acoustic Leak Detection at Nuclear Plants, KFKI 1992 38/G.

Peter, A.1992. " Acoustic leakage monitoring - systems and experiences," Meres es Automatika, Voi,40, No. 3, May4une.

Peter, A., et al.,1995. " Continuous leak monitoring and long term experiments at Paks nuclear power plant, llungary,' Proceedings of the 7th internativnal Symposium on Nuclear Reactor Surveillance and Diagnostics (SMORN Vil), OECD 95-00686, Vol.1, pp. 5.2116.

Pezze, C.M. and I.L.W. Wiison 1989. "Transgranular Stress Corrosion Cracking of 304 Stainless Steel Canopy Welds in PWR Systems," Proceedings of the Founk International Symposium on Environmental Degradation ofMaterials in Nuclear Power Systems - Water Reactors, JeckylIsland, GA, August 640,

'989. The National Association of Corrosion Engineers, pp. 4-164 through 4-179.

Pot, G., nnd L. A. Sokolov 1992. 'Kazmer,' A complex Noise Diagnostic Systemfor 1000 AfWe PWR WHER Type Nuclear Power Units, KYKI-1992-1710.

Racz, A.1991. ' Detection of Small leakage in the Feedwater system of the PWR," Proceedings of the kh Brazilian Meeting on Reactor Physics and lhermal Hydraulics (Vill ENFER), CONF-910983, pp.155-158.

12 9 DRAFT - NUREG/CR-6582

REFERENCF.S Ricz, A.1993. ' Detection of Small 1.cakages by a Combination of Dedicated Kalman Filters and an Extended Version of the Binary Sequential Probability Ratio Test," Nuclear Technology, Vol.104, October, also KFKI 199136/0, December 1991.

Reason, J.1987. " Acoustic leak Detection Provides Early Warning of Piping Failure," Power, September, p. 63.

Redmond, K.,1997. ONS 2 & M llPl/biU No:zle Components, bietallurgical Analysis Repon, Report

2131, Duke Power Company, Ruget, C. J., and W. J. Luckas, Jr.,1989. Technical Findings Related to Generic issue 23: Reactor Coolant Pump Seal Failure, NUREG/CR-4948, BNIeNUREG 52144, March. ,

Roger, C. J., and J. C. liiggins 1993. ' Reactor coolant pump seal issues and their applicability to new reactor designs,' Proceedings of the Sixteenth Reactor Operations international Topical bietting, lang .

Island, NY, August 1518,1993, CONF-930807.

Ruggieri, S. K., and J. S. Kelly 1988. Valve Stem Packing Improvements, EPRI NP 5697, Electric Power Research Institute, Palo Alto.

Sedriks, A. J.,1979. Corrosion ofStainless Steels, John Wiley, New York, p. 208.

Shah, V. N., and A. G. Ware 1994. ' Inservice inspection of Base-Metal Sites Susceptible to Thermal Fatigue Dnmage.' Changsng Priorities of Codes and Standards: Failure, Fatigue, and Creep, ASME PVP-Vol. 286, pp. 9,18.

Shah, V. N., et al.1997. Review ofindustry Efons To Afanage Pressurized Water Reactor Feedwater No::le, Piping, and Feedring Cracking and Wall Thinning, NUREG/CR-6tS6, INElr96 0089, AEOD/E97-01, pp.106,121.

k.ith, J. K.,1996.

Vibration Fatigue Failures in Short Cantilevered Piping with Socket Welding Fittings,* AShiE Pressure Vessels and Piping Cortference, Afontreal, Canada, July 1996, PVP Volume 338, pp. 2124.

Stoller Corporation, Nuclear Power Experience, BWR 2 V.B and PWR 2.V.B.,1974-present Strauch, P. L., et al.1990. ' Stratification Issues and Experience in Operating Power Plants,' High Pressure Technology, Fracture hiechanics, and Service Experience in Operanng Power Plants, AShfE Pressure Vessels and Piping Cortference, Nashville, Tennessee, June 1990, PVP Vol.192, pp. 85-92.

Streicher, V., et al.1986. " localization and Sizing of Leaks by Special Techniques," Continuous Surveillance of Reactor Coolant Circuit integrity, Procee6ngs of the CSNI Specialist birg., London, England (August 1985), Nuclear Energy Agency, Paris, France,1986, pp.143-156.

Streicher, V., P. Jax, and W. Leuker 1986. " Localization and Sizing of leaks by Special Techniques,'

Continuous Surveillance of Reactor Coolant Circuit integrity, Proceedings of the CSNI Specialist hitg.,

l.ondon. England (August 1985), Nuclear Energy Agency, Paris, France,1986, pp.143156.

DRAFT - NUREG/CR-6582 12-10 i

REFERENCES Streicher, P. J. V.1995. ' FLUS - a new monitoring system for leak detection and location," Proceedings of the Annual bitering of the Nuclear Society of Slovenia - 2nd Regional bitering: Nuclear Energy in Ceatral Europe Pp. 402-8.

Structural Integrity Associates, Inc.,1996,

Evaluation of McGuire Unit 2 NI Piping Thermocouple Data,' File EPRJ 95 301.

Su, N. T. 1990. Special Study Report-Review of Thermal StratVication Operating Experience, AEOD/S902, U.S. Nuclear Regulatory Commission Office of Analysis and Evaluation of Operating Data, March.

Takahashi, K.1995. ' leaked water detection device for control rod drive and BWR type reactor," Patent No.: JP 7 294678A, linachi Ltd., November 10.

Tanoue E., asuf K. Takei 1991.

Detection device of liquid leak for nuclear power plant," Patent No.:

JP 3 44597A, Todubs Corp., February 26.

Tirelli, D.1995. ' Development and testing of on line detection systems for cavitation, subcooled boiling nnd leaks in NPP components,* Proceedings of the 7th lit?ctnational Symposium on Nuclear Reactor Surveillance and Dsagnosnes (ShiORN 171), OECD 9540686, Vol.1, pp. 5.511~8.

Uffet, R. A., et al.1% 0.unarion of Water Hammer Events in Light Water Reactor Plants, NUREG/CR.2781, QUAiN-cM8, EGG 2203, pp. 5-16,517.

USNRC Technical Training Cente;. Systemc ofanual, Pressurized Water Reactors, Combustion Engineering Design, U.S. Nuclear Regulatory Commission, pp. 3.31 to 3.3-10.

USNRC TechnicalTraining Center. SystemsbIanua! Pressurized WaterReactors WestinghouseDesign, Chapter 3.1, Volume 1.

USNRC 1973. Regulatory Guide 1.45, Reactor Coolant Pressure Boundary Leakage Detection Systems, U.S. Nuclear Regulatory Commission, May.

USNkC 1978a.

Initial Test Programs for Water-Cooled Nuclear Power Plants," Regulatory Guide 1.68, Revision 2.

USNRC 1978b. " Standard Format arxl Content of Safety Analysis Reports for Nuclear Power Plants,"

Section 3.9.2, Regulatory Guide 1.70, Revision 3.

USNRC PWR Pipc Crack Study Group 1980. Investigation and Evaluation of Cracking incidents in Piping in Pressuri:ed Water Reactors, NUREG4691, September, pp. 4-7 through 4 20.

ANSI /ISA S67.03 1982. " Standard for Light Water Reactor Coolant Pressure Boundary Leak Detection,'

American National Standard, instrument Soclety of America.

USNRC 1981. Standard Review Plen, NUREG4800, Sewtion 3.9.2, Revision 3.

12-11 DRAFT NUREG/CR-6582

REFERENCES USNRC 1982. " Cracking in Piping of Makeup Coolant Lines at B&W Plants,' USNRC Information Notice 82 09.

USNRC 1984. Report of the U.S. Nuclear Regulatory Commission Piping Resiew Committee, NUREG.

1061, Volume 4, ' Evaluation of Other Loads and lead Combinations.'

USNRC 1985a.

Resolution of Generic hsue 69: liigh Pressure injection /Make Up Nozzle Cracking in Babcock & Wilcox Plants," Generic letter 85 20.

USNRC 1985b. Report ofthe U.S. NuclearRegulatory Commission Piping Resiew Committee Evaluation of Potential Pipe Breaks, NUREG 1061, Vol. 3, March.

USNRC 1988. " Water Hammer and Possible Piping Damage Caused by Misapplication of Kerotest Packless Metal Diaphragm Globe Valves,' information Notice 8813, April 18.

USNRC 1988a. ' Safety injection Pipe Failure," USNRC In'ormation Notice 88-01.

USNRC 1988b. *nennal Sttrsses in Piping Connected to Reactor Coolant Systems," USNRC Bulletin 8848.

USNRC 1988c. 'nermal Stresses in Piping Connected to Reactor Coolant Systems," USNRC Bulletin 88-08, Supplement 1.

USNRC 1988d. *1hermal Stresses in Piping Connected to Reactor Coolant Systems," USNRC Bulletin 8848, Surplement 2.

USNRC 1989. "nermal Stresses in Piping Connected to Reactor Coolant Systems," USNRC Bulletin 88-08, Supplement 3.

USNRC 1991. I.etter from R. E. Martin to M. S. Tuckman, 'NRC Bulletin 88-08 Thermal Stresses in Piping for McGuire and Catawba," October 31.

USNRC 1992a. Standard Technical Specylcations: Westinghouse Plants, NUREG-1431 Vol.1 September, p. 3.4-33.

USNRC 1992b. Notice of Violation (NRC Inspection Report Nos. 50-269/9134,50-270/9134, and 50-287/91 +34). letter from A. R. Ileidt (USNRC) to J. W. llampton (Duke Power), January 14.

USNRC 1992c. " Failure of Primary System Compression Fitting," Infonnation Notice 9215, February 24.

USNRC 1994. "Nonconservative Reactor Coolant System Leakage Calculation," NRC Information Notice 94-46, June 20.

DRAFT - NUREG/CR-6582 12-12

REFERENCES USNRC ",95. Commission Decision on the Resolution of Generic Issue 23 " Reactor Coolant Pump Seal Ps" u,' USNRC Infonnation Notice 95 42, September 22.

USNRC 1996. USNRC Maintenance Baseline Reportfor St. L.ucle 1 and 2, Report # 50 335/96-13 nsul 50-389/96-13, October, pp. 20, 21.

USNRC 1996. Letter from T. W. Alexion to W. T. Cottle, Resolution of NRC Bulletin 88-08, " Thermal Stresses in Pipfug Connected to Reactor Coolant Systems,' South Texas Project, Units 1 and 2 (STP),

February 23.

USNRC 1997a.

Unisolable Crack in liigh Pressure Injection Piping,' USNRC Information Notice 97 46.

USNRC 1997b. Memorandum from R. H. Wessman to W. D. Beckner, Resolution of NRC Bulletin 88-08,

Thermal Stresses in Piping Connected to Reactor Coolant Systems," for South Texas Project Units 1 and 2, Evaluation of Westinghouse Responses Dated July 15,1996, March 14.

USNRC 1997c. Operating Reactors Events Briefing May 21,1997 - Briefing 97-05 Memorandum from A. E. Chaffee to M. M. Slosson, May 30.

USNRC 1997d. Region !! Augmented Inspection Team (AIT) Inspection Report 50 269-97 06,50-270/97-06, and 50-287/97 06.

VanTassell, D. M.1994. 'Argo Packing Friction Research Update," Proceedings ofthe 7hirdNRC/ASME Symposium on Valve and Pump Testing, NUREGICP-0137, Vol.1, pp. 513-526.

Vecchio R.S.,1996. " Fatigue Evaluation of Socket Welded Piping in a Nuclear Power Plant," ASME Pressure Vessels and Piping Cortference, Montreal, Canada, July 1996, PVP Volume 338, pp. 25-42.

Wals, E. A., and E. C. Rodabough 1997. " Background of Stress Intensification Factors in Piping Design,"

PVP Vol. 353, American Society of Mechanical Engineers, July, pp. 312.

Ware, A.G.,1990. ' Mechanical Damage Experience in Major LWR Systems,' ASME Pressure Vessels and Piping Conference, Nashville, Tennessee, June 1990, PVP Volume 198, pp. 7-14.

Ware, A. G., et al.1995. Applicadon ofNUREG/CR 5999 interim Fatigue Curves to Selected Nuclear Power Plant Components, NUREG/CR-6260, INEL-95/0045 Westinghouse (Westinghouse Electric Corporation) 1989. Indian Point Unit 2, Steam Generator Ginh Weld, feedmater No::les Report, Spring,1989 Outage, Consolidated Edison of New York, WCAP 12294, Revision 1.

White, D.1995. 'ASME Article 13, Continuous acoustic emission monitoring," Proceedings of the 1995 Arnetican Society for Nondestructive Testing Fall Cor;ference and Quality Testing Show, Da!!as, TX, October 16 20,1995, CONF-9510265.

12 13 DRAFT - NUREG/CR-6582

REFERENCES Wichman, K ard S.1.ce 1990. ' Development of USNRC Standard Review Plan 3.6.3 for 1.cak Belote-Bicak Applications to Nucleat Powet Plants,' internationalJournal of Pressure Vessels and Piping, 43, pp. 57-65.

Wylde, G.1990. ' Cracking Solutions - Finding and Fixing Fatigue Failures," Welding Insnture Bulletin, May/ June 1990, pp. 53 55.

Yamauchi, K., K. Kawal, and Y. Ishihara 1995. ' Coolant leakage detecting device," Patent No.: JP 7-43494 A, MitsuMshi Atomic Power Industsics, February 14.

Yu, Y J., J. II. Khn, and S. II. Park 1995. ' Conservatism on current I_eak Before-Break Application,"

International Pressure Yessels and Piping Codes and Standards, PVP Vol. 313, Vol 1, pp.145-151.

DRAFT - NUREO/CR-6582 12-14

APPENDIX A: SEARCH CRITERIA a

. _ _ _ _ - - - - . .- - _. .- -=. . ._- -- - - -

APPENDlX A: SEARCH CRITERIA A.1 OBJECTIVE ne objective of this project is to review PWR experience related to pnmary system leaks, leakage rates, how aging and component failure effect leakage rates and trends, and the safety significance of such leaks.

The project scope includes leaks of the primary system components including large and small bore piping, vessels, pump and valve bodies, flange connections, pump seals, and mispositioned or leaking valves. ,

Steam ge.serator tube leaks, as well as leaks due to errors in configuration controls during non-power operation: such as inadvertent manual opening of valves, need to be excluded.

, . The search criteria to locate all the LERs related to primary system leaks at the U.S. PWRs from 1985 to September 1996 are included in this appendix. De search includes the boundaries of the following reactor coolant systems: the reactor core system, reactor vessel system, reactor ecolant piping system including

, steam generator lower head, control rod drive system, pressurizer system, emergency boration system, residual heat removal system, chemical and volume control system, intermdiate pressure injection system, core flooding accumulator system, upper head injection system and borated / refueling water storage system.

A.2 SEARCH CRITERIA The first search and accompanying dermitions will locate all LERs that identify leakage events while the reactor is at power, excluding steam generator tube leaks. The second search and accompanying definitions will locate all LERs that identify leakage events while the reactor is shutdowm, excluding steam generator tube leaks and errors in configuration controls. Leaks that occurred at power and continued after the plant shutdown will be identified l'1 both searches.

Search 1. Leakage Events at Power (steam generator tube leaks are excluded)

1. FIND

2. < EFF> . BEG. B <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD) <T> (E F C)

3. <EFF> . BEG.11 <lSYS> (AA AB AD AE AF BD BF BK BL BS BT CD) <T> (E F C) 4, <EFF> (Al AK AM AR) <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD) <T> (EFC)

5. <EFF> (Al AK AM AR> <!SYS> (AA AB AD AE AF BD BF BK BL BS DT CD) <T> (EFC)

6. < EFF> . BEG.B < PSYS> AH

+ 7.END

8. LOCATE TYPE PWT

9. LOCATE YEAR 85 86 87 88 89 90 9192 93 94 95 %

10. END

11. STEPSCAN

12. <EFF> . BEG. B <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

A1 NUREG/CR-6582

SEARCH CRITERIA ,

13. <EFF> . BEG. B <!SYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

14. <EFF> (Al AK AM AR) <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD) .

15. <EFF> (Al AK AM AR> <lSYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

16. < EFF> .BEO.B <PSYS> AH < < COMP > .NE.TBO

17. END ne search attributes mainly deal with effect codes for leakage and PWR primary systems plus the borated / refueling water storage tank (code CD). This is a pretty broad scarth and any attempt to narrow it further would possibly eliminate the desired LERs e

DeAnktons of the search codes Lines 1 through 7. De search strategy is finding all LERs in the SCSS database that identifies a -

leak / leakage event associated with any of the f01]owing plant systems while the plant is critical /at power (E F C). AA, reactor core system; AB, control rod drive system: AD, reactor vessel system; AE, primary coolant system;"AF, Pressurizer system; BD, emergency boration system; BF, residual heat removal system; BK, chemical and volume control system; BL, intermediate pressure injection system; BS, core flooding accumulator system; BT, upper head injection system; CD, borated / refueling water storage system. Also included m the search is leakage events associated with configuration control problems associated with the same systems; AI, open; AK, Transfer open; AM, wrong position and; AR failure to close.

Line 8. Limits the search to pressurized water reactor plants only.

Line 9. Lhnits the search to events from 1985 through and including 1996.

Lines 11 through 17. Is a relational search of the SCSS step matrices, treating the data within a step as a separate and independent collection. The purpose of these lines is basically to find events that did not begin as leaks.

Lines 6 and 16. Rese two lines are used to identify leakage events associated with the steam generatcrs (AH) excluding tube leaks (.NE. TBG)

Searth 2. Shutdown leakage events excluding steam generator tube leaks and configuration control errors.

Lines in BOLD have changes from Search 1 strategy.

1. FIND NUREG/CIN5582 A2

SEARCH CRITERIA

2. <EFF> . BEG. B <PSYS> (AA AB AD AE AP BD BP BK BL BS BT CD) <T>

(5 D H I K L)

3. <EFF> . BEG. B <lSYS> (AA AB AD AE AF BD BP BK BL BS BT CD) <T>

(B D H I K L)

4. <EFF> (Al AK AM AR) <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD) <T>

(B D H I K L)

5. <EFF> (Al AK AM AR> <ISYS> (AA AB AD AE AP SD BF BK BL BS BT CD) <T>

(B D H I K. L)

6. < EFF > . BEG B < PSYS > AH

7. END

8. LOCATE TYPE PWR

9. LOCATE YEAR 85 86 87 88 89 90 9192 93 94 95 96

10. END

11. STEPSCAN

12. <EFF> . BEG. B <PSYS> (AA AB AD AE AF BD B1 3K BL BS BT CD)

13. <EFF> . BEG. B <lSYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

14. <EFF> (Al AK AM AR) <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD) ,

15. <EFF> (Al AK AM AR> <lSYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

16. < EFF> . BEG. D < PSYS > A!! < COMP > .NE.TBG

17. END

18. LINK

19. PSYS (PO and maybe PMI PT)

20. <EFF> (Al AK AM AR) <PSYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

21. <EFF> (Al AK AM AR> <lSYS> (AA AB AD AE AF BD BF BK BL BS BT CD)

20. END Lines 1 through 15. These lines are the same as the critical /at power search for leak / leakage events except the plant condition for the events have been changed to searc5 'cr shutdown events. The codes B D H I K L define the various shutdown moden identified in the SC' ' latabase.

lica 10 through 22. These lines of the search strategy we provided to remove events that are related to configuration control problems associated with the shutdown. Specific events ttat can be removed are those associated with O, operation activity; and/or PM, maintenance / repair activity; and/or Fr.

test / calibration activity.

A3 NUREG/CR-6582

SEARCH CRITERIA Note that the LERs in this group may include actual leaks as well as personnel related configuration .

problems. There is no way to distinguish LERs containing both kinds of problems in SCSS. The addition l of PM and PT to the configuration search (:.lthough plant operators frequently position valves during innia-a-e and testmg) is to find if maintenance or testing programs were deficient and ultimately caused an aging related component failure.

l l

e i

h f

I 4

i

, NUREG/CR-6582 A-4

t

~

APPENDIX B: REACTOR COOLANT LEAKS DATA COLLECTION FORM k

\

e

i APPENDIX B: REACTOR COOLANT LEAKS DATA COLLECTION FORM Plant information

~ Plant Nainer Artbitect Engineers Docket VenJor W. CE, B&W LER Nuseber Initial C iticality Date:

Reportsble Laak YevNo*

Event Dates Mode of Operat6es: p,,,, op,,,,3,,,

neartop HetStandby Het abetdows Leak Characteristics Leak Types (Dennesif theled Isolable Initial Leak Rate:(Defmes on lattialleak rue observed a se man of (Gallons'mina) could be isolmed from the RCS by Non isolable the event if tw leak rua is not stased in closure of an isolation valve. liw Internal the LI:R, the default is UNKNOEW) category ' Internal

apptws to leaks Unknown from the RCS to an interfacing system)

Maximum Leak Rate (pef *'8 Crack Slee, As published in the (Gallons / minute) the maximurn leak rate observed at any LER if provided for pomt durtng the event. If the leak rate is OncMs avail le hfMew the W ofopening area) Nksh-not stated in the LI:R the default is ITNKNOWN)

Small(<1 )

Potential Growth Rate: Pipe Stre:

Intertnediate I(14)

. (Idenones the growth potenbal of the (Gallons' minute) includes available information about leak if possible to determine from the the size of the pipe in inches. Intermediate 11(412) location and failurt nwchanism of the Large (>l2) component If tndeterminate,dw default

value is OHKNO AH)

Activity Level Identifies an airborne activity level of the areas affected by the leak. Also provides additional data relathe to the size of the leak / leak race.

See Terminology and Definitions at the end of the form.

B1 NUREG/CR-6582

- . e-- -r-, v y- ,-- .- --

DATA COLLECTION FORM Leak Characteristics Cont.

Optkos Database Code for Options !

Method of Detoetion:

Identifies the method in uhlch plant perwnnel first discovered that a Area Fire Alarm AyA reactor coolant leak exismL ControlRoom Alarm CRA Control Room Indication CRI Containment Radiation Monitor cgm High Nolu Level }{NL Periodic Surveillance Test PST PersonnelVisual Observation pyo Secondary containment Radiation Monitor SRM Other OTH Options Databaw Code for Options Locatles Deternalmation:

identifies de method in which plant personr,ellocated tlw source of the

reactor coolant leak Area Fire Alarm AFA ControlRoom Alarm CRA

. Control Room Indication CRI Containment Radiation Monitor CRM High Noiw Level HNL Periodic Surveillance Test PST Personnel Visual Observation PVO Secondary containment Radiation Monitor SRM Surface Temperature Measurement STM Other OTil Com mentoI (Additional information conceming the leak characteristics)

See Terminology and Defmitia 3 ai the end of the form.

NUREG/CR-6582 B-2

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

i OATA CO!1ECT10N FORM Leak Location information Systeen Options Database Cod ( Location relath'e tote" Synem in which the for0,Wiens pdanary costalmeneett leak was identified. t Only one system can g ,; ;g tv identified for Control Rod Ikive CRD each record' HigWintermediate Pressure Safety injection HPSI ,

14w Pressure Safety lajection LPSI p , g,g, ,

Main Coolant Piping MCP Pressuriser PZR NN8*

Reactor Coolant Charging / Letdown LTD Reactor Coolant loop Drains RCLD Y N0

. Reactor Coolant Pump RCP Reactor Coolant Sampling System RCSS Reactor Pressure Yessel RPV

~

Residual Heat Removal RHR Safetyinjection Accumulator SIA Steam Generator (primary side) SO Unknown UKN Component /14 cation: Component Options location Options Matedal Identifies the component or Identifies the material where components that failed which Heat exchanger Base metal the failure occurred.

resuhed in the leak Head vent Coupling

, Housing Fitting Material Options Instrament line Flange Alloy 600 Leakoffline Gasket Austeninc stainless steel j Manway Packing Carbon steel i Penetration O-ring Cast stabiless steel Pipe Seal Gasket material Pump Seat Seal material Tubing Stud Titanlunenickel alloy Valve (s) Weld Vessel Unknown Unknown Other Other

. Component Age: Manufacturer:

Identifies age in years of the Identifies the name of the component (if known) when it manufacturer of the feited. The default value is component that failed plant age since initial criucality and leaked date.

Comments: -

See Terminology and Definitions at the end of the form.

B3 NUREG/CR-6582

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

t DATA COLLECTION FORM Cause Nehe opaeas c. ,, % ,,,y , ,,,,,

t Mechanisant mal c, 6,., seer %,

Boric acid Gasket degradation Sul degradation :

corrosion Unkrawn Unknown Calibraion High cycle fatigue Not Applicable Not Applicable Cavitation Hydrosen Component failure embntilement Contaminants Diaphragm Pecking Dwign IOSCC Lem cycle fangue Diaphragm degradation Packing degradsuon Environment '

PWSCC Unknown Unknown Fabricmion defect i

Not Applicable Not Applicable failed to close

$CC >

Foreign material TOSCC Thermal fatigue liigh stress l

i fligh temperature

%brmory isegue Installation Unknown Inoperable leak detection sys.

Unserine Line configuration ,

lheing es aluased

~

Non-qualifad put het AMAcable Overhaul procedure Personnel Enor Proc 14aral Error Repair \ loads .

Susceptible material Valve configuration Valve vibration ,

Waterhammer Unknown Other Connements: (Addiuonal information conceming the cause and'or contributing factors of the reactor coolant leak) i NUREO/CR 6582 B-4

- . - - . - - - - - . .- . . - . . - , = _ - . - - . . . - - - . . _ _ . - . - . -.. . - . . - ._

DATA COLLECTION FOh i Leak Consequene.es l

Categor6est Forced outage . RX tr ip'cooldown i i

Comaintnant entry Safetyinjection demand Onshe Release No Offsite telsese Othee system affected flot shutdown .

Unusual event declaration Cold shutdomii Alen declaration Exceed FSAR Limits Other Eaceed TS Limits Comanents: (Addenal information cecerning the leak cornequences) 6 Corrective Ac6ons/ Measures l 2

Short-Term Actions: Laag-Term Actions:

Repalt Repair Replacement Replacement Procedure t.hange 4

Proper installeion Design chante Design change Evaluation inspection /Testmg Other 10 CFR 21 Evaluation Other >

Comments: (Additional information concen.mg the shon term and long term corrective actions) 4 y

L BS NUREG/CR e.42

DATA COLLECTION TORM Definitions and Terminology Reportable leak

1. Any leak that exceeds the technical specifications for allowed leakage or,

2. any leak that results in exceeding the FSAR limits for control room habitability, or 10CFR100 limits or,

3. results in an unplanned engineered safety feature actuation.

Pressure Boundary Leakage Pressure boundary leakage shall be leakage through a non isolable fault in a reactor coohnt system .

component body, pipe wall or vessel wall. Pressure boundaries are not gaskets, packing glands, valve seats, flanges, manway covers, or other mechanicaljoints or fittings.

Non-isolable leak Any leak that can not be stopped by closing a valve or placing a valve on its associated backseat for packing gland or body to bonnet leaks. Non isolable leaks include all pressure boundary leaks, and leaks associated with reactor coolant pump seals and pressuruer safety valves.

Isolable leaks All leaks other than non isolable leaks isolable w/ difficulties All isolable leaks that could not be stopped until that plant operating mode was changed. Examples include;(1) packing gland failure of an isolation valve that causes steam to escape from the valve and the steam prevents operators from placing the valve on its backseat to stop the leak; or (2) the leak causes airbome activity to exceed the maximum permissible dose therefore requiring a plant shutdown and cooldown before personnel can enter the area and stop the leak.

Identified Leakage identified leakage shall be:

a. Leakage into collection systems, such as pump seal or valve packing leaks, that is captured and conducted to a collection tank, or

b. Leakage into the containment atmosphere from sources that are both specifically located and known not to interfere with the operation of the leakage detection systems or not to be pressure boundary leakage.

Unidentified Leakage Unidentified leakage shall be all leakage which is not identified Leakage NUREO/CR-6582 B-6

DATA COLLECT 10N FORM

- Twhaical 5;..e18esties leakage Dants and Reportability Requiresseets ,

a) No Pressure Boundary 1.makage b) 1GPMUnidentirwul14akage c) 10 GPM Identified leakage [

I References ,

1. D.C. Cook Nuclear Plard Unit 2 Technical Speciff. cations, November 31995. :

2. Arkansas Nuclear One, Unit 2, Technical Specifications, April 1981 1

B7 NUREG/CR-6582 y . - - - , -..,u,- ,- 4 ,- w-- - - - - ----T-, --.v. ~-v--- ..-- ,y +

APPENDIX C: STATISTICAL ANALYSIS METHOD l

l l

l t

I DRAFT

--- .ad.* - -

-- A. A __~. w 4 a._- - ..-A 4

-e a

e DRAFT

APPENDIX C: STATISTICAL ANALYSIS METHOD C.1 INVESTIGATION OF TRENDS IN TIME nere are two types of Licensee Event Reports related to PWR primary system leak. Some repc

m submitted because a leak was detected whereas others were submitted for some other reason and .' L was found during the investigation of the everit. ne former type are referred to as reportable evo m whereas the later ones as nonreportable events. Only reportable leaks are considered here to ensure consistency of the database, ne occurrence rate,1, of any kind of event, such as the rate of reportable leaks caused by vibratory

. fatigue, is the average number of such events per unit time. Suppose now that we wish to investigate trends in calendar time. Then the unit time is taken as plant operating years. For example, consider only the year 1990. De occurrence rate for that year, denoted A, is the average number of events that would

. have occuned in 1990 per plant operating year, nis is a theoretical quantity, the rate that would have been observed if an infinitely large number of plants could have been observed, all in the condition of the actual plants in 1990. It can be thought of as a large-population average. He actual number of events per plant operating year in 1990 differs from this average somewhat, because of the random nature of leaks. De actual number of events reported for 1990, divided by the actual number of plant operating ,

years in 1990, is an estimate of A. This estimate is sometimes denoted by the symbol A'.

A confidence interval ca*: be constructed for A, using standard methods given in many statistics books, and by Engelhardt (1994). Confidence intervals are constructed to contain the true value of A with high probability. When many data sets are collected and 90% confidence intervals are constructed,90% of them will contain the true corresponding value of A. A wide confidence interval means that A is estirnated only with great uncertainty. A narrow confidence interval means that A is estimated with little uncertainty.

One way to investigate whether an event rate, A, is changing over time is to perform separate calculadons for each year in the study, obtaining a point estimate and a confidence interval for each year. Plot them side by side, and see if they seem to show a trend.

A more formal way to investigate for the presence of a trend is to construct a model for A that includes a possibility of a trend. In this study, the form of the occurrence rate was modeled as A = exp(a + bt) (C-1) or equivalently, log (A) = a + bt, where t denotes some measure of time. To study a possible trend in calendar time, t is set to the calendar year. His model is a loglinear model, and methods for analyzing data with such a model are explained by Atwood (1995) and by certain advanced texts. If b is negative, the trend is decreasing, and a plot of A against is an exponentially decreasing cun'e.

The SAS procedure GENMOD was used to analyze data using this model (SAS 1993). Denote exp(a +

bt) by A(t). GENMOD uses a standard statistical procedure, the maximum likelihood method, explained in many statistics books. It uses this method to estimate a and b, and thereby estimates logA(t).

C1 NUREG/CR-6582

1 STATISTICAL ANALYSIS METilOD 1

GENMOD also determines whether the estimate of b differs from 0 by more than is plausible from chance alone, A quantity called a p-value is defined as the probability of observing such an extreme b as a result of chance alone. For example, consider data in the first row in Table C 1. De estimate of b in the first row is -0.17. De p value is p.value = Prob (l estimated bl a 0.17, if there is really no trend) .

His probability is seen in the table to be 0.000), a very small number, ne probability was calculated based on a data set with 199 events. If the data set had been small, then a slope of 0.17 would not correspond to such a small probability. With very few events, it might be fairly likely that a slope of at least 0.17 could be seen from chance alone.

A very small p value indicates strong evidence that a trend is present. It is customary to use 0.05 as a cut-off; when the p value is smaller than 0.05, the trend is called statistically sign $ cant. P. values and statistical significance refer to the strength of the evidence for a trend, and therefore depend on the size ,

of data set. Engineering sign @cance, on the other hand, refers to the magnitude of the trend. De number b is the slope oflogA, and can be used to measure whether a tronJ is significant from an engineering poirtt of view. In the tables below, both the p value and the estimated slope are presented.

To display the results graphically, the following kind of plot was corstructed. First, tne confidence intervals for each year were plotted side by sMe, as discussed above Superimposed on this plot was the curve showing the estimated A(t). His curve smooths the estimatec for the individual years. Finally, using a method from page 34 of the report by Atwood (1995), a 90% confidence band for the true curve was constructed. De method for constructing confidence bands guarantees that at least 90% of all data sets would yield bands that cover the entire true curve. The calculated confidence band was also plotted, surrounding the curve for the estimated A(t).

C,2 INVESTIGATION OF TRENDS WITH PLANT AGE The same apsm;ch was used to investigate whether the occurrence rate was affected by plant age, rather than by calenk $ne. Ilere, age was defined re the time from the initial criticality, and a plant was said to have age n (fit wns in its n* year. In this case, A was regarded as a function of plant age. For example, consider plants 10 years of age, ne data for the study covered the years from 1985 through part of 1996. During this period, some of the plants passed through their 10th year. The estimated occurrence rate for plants in their 10th year was the number of events that occurred during the 10th year .

of a plant's history, divided by the number of plants that had their 10th year during the study period.

Except for these changes, the method was the same as the method for studying trends in calendar time.

C.3 RESULTS Table C 1 summarizes the data analyses for the various kinds of event, when investigating trends in time.

Figures 2-4 and 7 8, in the body of the report, show the first two cases for which a statistically significant trend has been found.

Table C-2 summarizes the data analyses for two kinds of event, when investigating trends with plant age.

Figures 4 11 and 7 9, in the body of th's report, show these two cases graphically.

NUREG/CR-6582 C-2

I C,4 DISCUSSION The effects of calendar time and plant age are intenwined, because as time advances each plant becomes older. 'Ihe effect of calendar time reflects the evolving body of regulations and industry wide learning.

The effect of plant age reflects both the teaming curve of plant personnel and the aging of the hardware.

If leaks have become less frequent in recent years, is this because calendar time has passed or because t the plants are older?

Ideally, to answer this question one should construct a model that includes both effects at once. To carry this out, one would need to construct a file showing the number of plants that 'ussed through each possible plant age during that calendar year. It was felt that the quality of the Mta did notjustify construction of such a sophisticated model.

. Therefore, a simpler approach wu carried out, comparing p values. For vibratory fatigue, the p-value for plant age is 0.055, whereas the p value for calendar time is 0.16, suggesting eat, to the extent that a trend is present at ali, the dominant factor is plant age. Apparently, the decreasing trend with age implies

. that the vibratory fatigue failures are because ofinadequacy of the initial design and fabrication, and not aging.

For packing degradation the p.value for age is 0.003, whereu the p.value for calendar year is 0.0001, suggesting that the trend in calendar time is more significant. As discussed in Section 7, the decreasing trend in calendar time is apparently a result of the valve stem packing improvements made in early 1980s and incorporation of some of these improvements in operating nuclear power plants.

C,5 REFERENCES Engelhardt, M. E.,1994. Events in Timer Basic Anclysis ofPoisson Data, EGG RAAM 11088, September.

Atwood, C. L, l995. Modeling Patterns in Count Data Using Loglinear andRelatedModels, INEL.

95/0121, December.

SAS 1993. SAS/STATSoftware: The GEN \f0D Procedure, Release 6.09, Cary, NC: SAS Institute inc.

1 C-3 NUREG/CR-6582

-ae 2 -

.u.2-- ,_au4 - ,u, -. .. .-m ,...s . - .4.. . . mu--. .m __m. - u. _ - _ --

J ASSESSMENT OF PRESSURIZED WATER REACTOR PRIMARY SYSTEM LEAKS FIGURES a

k.

b

48 -

45 -

42 - @

39 - _

36 - p[;g

. 33 - _ $2 j 30 - % ?3 % ~

5 27 - ,

5 24 -

j 21 -

$ 18 - l M1 15 -

k D F 3

9 g

1sss 1ses 1937 1ssa 1ses 1sso 1991 1ss2 1m 1m 1m 1m Calendar year ==*aa Figure 2-1. Distribution of both reportable and nonreportable reactor coolant leakage events during the 1985-1996 penod.

The data provided for 1996 is through and including the 3"' quarter.

V

! 39 - 1.2 l l

36 _

IEINumber of events

+ Frequency (events / years) _

e l 30 - , R-27 - e

, - 0.8 $ ,

! '5 2

2 4 --

V k

~

o 21 - F i m

- 0.6' il !

.8 18 -

E 8 e '

\

} 15 -

- 0.4

[2 12 - o 9- I '

1 i 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1998 l Calendar year "'

i l

Figure 2-2. Distribution of reportable leak events and their frequencies during the 1985-1996 period. The frequency for i 1996 is estimated based on the operating years for 1995, accounting for data only thr6isgh the third quarter of 1996.

I i

{

j .

j 70 l 60 50 -

E

$4 - i

@ 30 -

20 -

10 -

1985 1986 1987 19C8 1989 1990 1991 1992 1993 1994 1995 1996 Calendar year ommooi Figure 2-3. Histogram of the number of operating years for each calendar year of the study period. The data provided for 1996 is through and including the 3* quarter.

t I

t 1.2SE+00 ',

j Point est and 90% conndence interval

- Fined rete 90% contdence bend on Sie Siled rate 1,00E+00 ,

s

\ '

g 0.75E@- ,,

g" s

w 0.s0E @. ' 's c\l, s s s s

.'s. *,

's ',

~ ,T -

s -

s ,*s %

% g Il o.25E@ " 1 ,*

s, ,,

T 0.00E+00 85 06 87 88 89 90 91 92 93 94 95 96 Year "" .

Figure 2-4. Statistical analysis of a trend in the frequency of reportable leak events. Point estimates, 90% confidence bands, and 90% confidence intervals are shown. t s ,.u- , n -

. . _ . . . __..._m_ _ . . . _ _ _ _ . _ _ _ _ _ _ . . _ _ . _ _ . - _ . . _ _ . _ ---_ . _ __.__ _ _ _ _ . - _ . . _ _ _ _ _ _ . _ _ . . . _ . _ _ _ _ _ _ _

140 130 122 120 - ----- - -------- ------------ - ----------- ------------------

110 - - - - - - - - - - - - - - - - - - -- - ------ -- --------- ------------------

$C 100 - - - - - - ------------- ---- -- - -- -- -----------------------

g 90 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

@ 80 - ----- ---------- ------------ ------ ----- -------------------

O 70 - - - - - - ---------------------------67______________________

u y

60 - ----- --------- -- ---------- ---- -------------- -------

E s0 - - - - - - ---------- --- ------------- ----------------------

i h 40 - - - - - - ------------ ---------- ---- ----------------------

3a _ _____ _ _ _ _ _ _ _ _ _ . _ _ _ _ _ _ _ _ _ _ _ _ _ _ . . . __________ ____ ______

r 20 - - - - - - ---------- ------------ ---- ---------------- -----

i t

10 - - - - - - ---- ------- 7--------- ---- -------------

l 3-------

0 "

t

+/

/ / / I Category aw.n3 Figure 2-5. Distribution of reportable reactor coolant leak events based on the ability to physically isolate the leak.

I i

60 50 -- ----- - ----- ---- -----

41 y 40 _ ____

O 30 - ----

.0 E

s g 20 - ___.

11 10 -- ---- ----

------ ------------------------ -- t 6 6 1 1 1 1 0 " " " "

/ 'j/ $

/

/ /

$ 9

/ 0

/ + / q*/ /s i

s# t !'

i Leakage rate ..=_w_mo, i

Figure 2-6. Reportable reactor coolant leak events inside containment partiticned by the initial leak rate.

50 -

45 -

11 40 --

38--

E

$ 30 - ----

O !

.o E

3 20 t

Z

, 10 - ---

9--------------------------------------- -

S 2

9 9 9 9 0 . _ _ _ _

\

/e eee //// ////// ,

Leakage rate as* by_=oi i

Figure 2-7. Reportable leak events inside containment partitioned by the maximum leak rate.

I i

l l i l

l

18 16 - ,----

g - ---- ----- - --- ---~------- - -------------- -------

14 - ---- ---------

93 j 12 -- - - - - - - - - - - -

-8 so .

o b

.a a - -- --------- ----------

E y s - ---

4 -

a 2 - -- ---- - ;

, ---- ,----- 3 3-- ,----- 4----- ,----- --

0 E

/e ///////// s s Leakage rate e Figure 2-8. Reportable leak events outside containment partitioned by the maximum leak rate.

.. ,, __e .

L

[

t 8

8 o.14 (2'IPressure boundary leaks

-+-Frequency (events / years) 7 . o.12

. Y

'es s- -

??

'( - 81 x

e d

?!

-l$ $

'~ $

r i l ji.

g , ; [ p -* 1(

o ,

u 4- i g

i g 1 l

f ; ;

E

~ ,

r h ( {$

5 s j s a v

j {, - o.os -

z :

I 1 t e

a .

e l ,

3 I - o.o4 80 2_ b !

O t

^

q j-g -

i i 1 R

I l l - o.o2

'i h l le o

,, b f a,

, i 4 1l 6 I

, , , , , , ; o

/ / / / / / / / / / / /

i Caiendar year ._w, '

l Figure 2-10. Distribution of pressure boundary leak events and their frequency dur~mg the 1985-1996 period. The frequency l

for 1996 is estimated based on the operating years for 1995, accounting for data only through the third quarter of 1996.

l j -t l . . .. . .

60 i ,

50 48 42 40 5

30 -

j; 24 20 - 18 i aE sh hi 'i' xmny" 0

System do Figure 2-11. Reportable reactor coolant leak events g '9ioned by the system from which the leak occurred.

_ e e

125 1M - -

97

$ .x E \I g 75 - - -

o e ;

.c so - -

A:

E 42

:)-

z 3e E

2s - -

16- 18

-a 7 _,h c '? 8 l ,

4

- -s ._=

2 3

g _- g s 2 b, pyg

' 0 A*

/ / $ 9 4 A!

Component l x15 l

Figure 2-12. Reportable reactor coolant leak events partitioned by the component from which the leak occurred.

l l

60 -

1 50 - ~~-------- -- -------------------------- -------------------------.

e -

4s '

i c 42 0

j 40 - - ----- ------------------------- ------------------------ --

E

e

., o l $

30 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - --

o 24

! 6 22 e

.D 20 - ----------------------------18---- ---- ----- ---- ---- --

E 1s ., ,

3 amt Z

10 12 p 10 - - ----

g- - - - -

6 6 0

,/ / / / / / // /

Location xiaocore Figure 2-13. Reportable reactor coolant leak events partitioned by the location in the component.

Less than 1 inch 21 events (51.2%)

{

_ 4 to 12 inches 1 events (2.4%)

i

' 1 1 to 4 inches 19 events (46.3%)

xt tppesizeot i l

Figure 2-14. Distribution of reportable leaks through the welds in pipe and instrument lines versus their size.

l

_ . _ . . _ _ . _ _ _ _ - - - _ _ - f

8 0.12 7 -

- 0.1 F e

m 6- #

I 3 -

- o.o8 g 5~ ~

T E t E t

3

?

o 4- -

p'. j -

- o.os E t

o S

~

1 3, E 3 v

+ '

s _ .

m Z 'le -

- 0.04 2 2- -A I a l'

5 e x -

M I

/- - o.02

)g il~

~ ~

a y & i i ;

m g A ; % s 6

:t

~

ll 2 !

t >

e.-

0 - -

O 1985 1988 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Calendar year xu Figure 2-15. Distribution of reportable leaks through the bolted connections and their frequency dunng the 1985-1996 period. The frequency for 1996 is estimated based on the operating years for 1995, accounting for data only through the third quarter of 1996.

l ,

I ! i' : l i' , i l . ; ! . !

9 2

_ - - 3

- - 9

_ - - - _ _ i

_ - - _ _ v 4 e

_ i

_ - - - _ _ m i

s 9_ m n a .

_ i s h c

6 n e 1 a m

- - - _ _ h n c o s

e it a

2 m d a

n r g

o e

_ 4_ EE i t d a y

- d b a

. _ i a r d g e n

e it o

D 4 = itr

_ a p

s s

_ - - - k a

l e

- - t

_2E n

_ _ a

_ - - - _ _ lo .

- - _ _ o

- - - _ _ c

_2E r

o

- - - t c

a .

_ - - _ e r

le b

5 a 5 0 5 g s o t a

3 3 1 2 1 j r o

p e

_ R 8ptE og 6 1

2 e

r u

g i

F i

15 usSCC to. Vibratory fatigue 12 - T'--~~~'~~~~~~~~--~---~~~-~~~~~~~~~~- y g g g g g*g

.0 M g_ __ .____._______._____. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . _ _ _ _ _ . . _ . . . _ _ . ~

b

.o E 6- --- ----

3 -

3- ---- ---- ---- ----- ---- --- ----- ---------- --- -------

0 -

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 SCC 0 1 1 0 1 2 0 3 2 1 4 1 Vltwatory f atigue 2 2 5 3 3 3 2 0 2 5 2 0 PacHng degradation 12 9 2 2 0 2 2 0 0 0 0 0 Calendar year ,%

Figure 2-17. Distribution of the three most significant degradation mechanisms in terms of numbers of reportable leak events during 1985-1996. The data provided for 1996 is through and including the 3'd quarter.-

47 1

g 40 - - - - - - - - - - - - - - - - - - - - - ---------- --------- ------------- -------

e

E o 30 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - --------- --------- -------

$ 25

} 20 - ------------------------19------------ --- - ------------ -- ----

j '8 Z 10 - - - - - - - - ----------------

6 5

4 4 O -

l4 "

~$'$

/~bh$kk -

Contributing factor '

Figure 2-18. Reportable reactor coolant leak events partitioned by the contributing factor for the leak.

l l

I

, e a a 8 8 4 4 1

, t e # I I # i 4 5 6 4 I 6 t i 1 l g i f . 1 t i i t i

4 L 3 i i t i e i I i f i G I i i s 1 i l I 4 i i 1Y l

l e e i l i i t i f i 1 i 8 i 4 e i I I I i f t 1 I i 1 i l i I 8 f

. , , i , , i , , ,

a i 6 i e t i G 3

e a i t i a f 4 i d E 3 I i i 1 1 1 I ( i i f 1 i 1 4

$ I I l l 3 8 5 j , , , i i i . i .

c 4 # t i i i f Te P

Q

? p 4 I I I i 6 Q

r , i i i i s n-

, 4 1 i

8 . I i 3 g

i E i i J i I t @

j , , . . + . . , i i c

i i , i i e i i i n g n 1

i i

i

> i i i O D 4 i i , 4 i . i i i C X l l l I ! O

, , , i i . . . 3 1 , , , , , mi M.

y 0

', i i . , 4 i g .,C l l , l l l M N

, , , i i i i i C O

! i ' ' ' V i : .

O e

i i

i N O o c

i

. . i 1 i i i o

, , a i i e . i -

1 1 I l 1 I i 9

, i . , , , , m 1 . i i i . i mi Q.

j , . . i i e i . W m i i , i i . . -

l , i i . . i i C

, i i i i i i i i O

.! . m ,- -

W O i e- '

y

. . i i . . . . .

, cn i

i i i i

i i

i e i i

-g i t t e s I i -

C E

~

l l l ,

e '

O i , , i . . . . ,

g i t 6

9 i i f

8 e i i

i 2

i i i n 4<o -

o

) i . . . , ,

i . . O O i i i

m

! l I i !

m i a 4 i i e i O

o o o o o o o o o o o o

v e e t- e e e n N v- --Q CU C

O C.

O stueAs Jo .seqwnN 9 m

v= e N

O w

D

.CD s

.--.,.y

l l

s a

, v.$ <d. g[aps

f. -

Other 10.2%

- n

.. g& g . ;

W ;: @ f R: h.. a .f

~

, ,. -l. ~ f, p -13. 7 c

,'T,i b,$ '

Repair 59.4%-[ ':{. 1,p) 7_ -'. .'. .,.

,g :

t

.if

wf.<'. '%n? - [ .

\

t

. .-o; . . . .

f, L 3 s Replacement 30.5%

4 . t .

w1 Figure 2-20. Distribution of reportable leak events by the short-term corrective actions. The other category includes design changes, evaluation, inspection and testing, and proper installation.

i l

i i

l Evaluation i 22.6 %

l

/

e 19nchange ,

1 9.8%

i

-l a ;

y j Inspecilon/ testing'

~ Llh.- . .

l 28.6 % ~

l , ,

ye ': i, .

l llL

i I ,Qi ~

Procedure change ,

13.5% i i

Repair Replacernent 10.5% 15.0 %

w2 Figure 2-21. Distribution of reportable leak events by the long-term corrective actions.

l l

l .

I l

l e

Turbulent penetration Stratified coolants Leaking valve

, 4 apenm.

MI Branch line Cold leg (Main Coolant Loop) /

Cycling zone (potendal site for fatigue crack intbation)

T 5j Turbulent decay cori alation Distance from header pipe (UD)

Temperature distritx,' tion

~

. =

Distance from header pipe (UD) careras Figure 3-1 Turbulence interaction regions (EPRI 1993b).

e 4

lyg }3ff{! Main Qip? -i ygi ffNb s @l},;jxQlQ(-

', ., W:jr;4;J; cootant

! . % %-.d loop JN<-c;ge .

g g;y:e><-

.. .LMg;v4'31,g; ai N Branch line b

Q initial penetration--w 0 0.

Deeper penetrationg Stratified layer (a)Initid penetration of turbulence (b) Deeper penetration causes stratification is conhned to vertcal section- in the honzontal section.

Mgure 3-2 A change in the penetration depth of turbulence from power variations can cause thermal stratincation and cycling in a branch line (EPRI 1993c).

J

l l

i j

i

-I Warmer Warmer ONC -

DNC --

Cooler MOV- Cooler Si "

check g, Cold leg ODDC -

Convective loop so;

N stratification

Wrbulent penetration

Figure 3-2.1. Natural convection driven stratification in a safety injection line - Safety Injection Line -

(EPRI 1993f),

e A

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

-3 4

7 C

F T

s

- , - Hot water l Tm ~ -y ,

f-- Mixing zone i- 1 / (interface)- .-

7

-- ,/ /

) '

Cold water ;

Te one.

[ p 33 T+.are distnbudon in stradfied fluki 6

J l

a, e

5 W

e ,

b 8

4 h

m

,,.*:w -

m-t-s s ,em~ y n ,,,-.-w. -

+ ,a -e .,--,n .ws. s- .-a-,.-,-+v wn-e- - $n ,e- -~~m-, ~

I i

l

. I l

l l

Wold Cornact

~

Safe ~ N d***i"' l Oman > \

end Held rt3 Wold % im s -

* // w.J

- Peew

, .g yp'".re, r.

Nemal 8

.. ........... . . . ..................................ll T "

% I'f"'VI- A\.=4 -+d' u-j E k+ Z \ ,,

\

Wotd Themmi 23e. gg some S.S pee ,

I 2 16 % tt

[

te 6000e S.S. SW haft catwiG j

! 0.01M to 0.015h %

i le S.S. warmeg onamercas dearance

" (Wing) bne between theme sweve cosar, and nozra LD. VS7 tots Figure 34 Typical old and new designs of he t makeup /high-pressure injection nozzles and thermal sleeves in the Babcock & Wilcox plants (Babcock & Wilcox 1983) e i

l l

l I

1

i .

22 . ,

~ ~

l 1" =

E,Tl0 74 76 78 80 82 84 86 88 90 92 94 database (S )V n30 yman et 996) i i

. _ . _ ,, ._ _ __ . _=

i 1000

.a l l v i:,> - 98%

Ao"/

A g 100 . zy

. , .sv

. ."l

- 50% ,

g , ,

a .r u i a _;m "

g

] . 9rF dt g o 7A e z ,/ L &

m 10 -

c 3 3 ,. ,

e ,

oBWR g APWR U SI_

1 10000- 100000 1000000

= = '

Time to Failure (Hours] ,

Mgure 3-6 Pipe leaks and rupnires in LWR piping caused by thermal fatigue (Nyman et al. .

1996)

.k } te ac Seei newn 3d -

~Co*1mg S994 H8r Noitale 2M

^

-h Weld crook .

RC Letdwn & '5 2: I Seet Return + RCSL 4

,,g Res Leap 8 H pm y,,,,g g

. ) [

.# i N -

r, I ** 3 i

k# ASb M Latdovn JE4 1 RCS Loop A l

RCSLoop Stw ey 3,'

VB f are (1j X seer- +y' ~

40 pm 4 -

fp ,

,jbl RCP rde Q '

l - SEAL FLitts - N

[weta =-

- ][ D.,,' .

Y'I" Sterop N ][j !. . . ,

,, -n M

h rls N ;

74 inwrym y MMtwa

W %'

mw i AB!R9

, AB . Auxiliary Building RB . Reactor Duikling RCS ReactorCoolantSystem N Cheawwe N . Flowentting Ortfios

@ . Air Operated Wtve

. Motor Operated Wlve X Manual Valve es' em Figure 3-7 Oconee 2 makeup /high pressure injection system.

4 Safe End Vertical Check Support N Valve HPl or MU/HPI

& Safe End to Norzle "1" MUIHPl Line Weld t -

RCS

,\ ; l (Coid s e i 4

'S) )

' isolation (Block) Safe End to i ::== i Valve Nonle Weld Bimetallic) l MulHPl Line %(Warming l i

Line Snubber i L l T .- - ~

c 1 i

1

@ n 3-8 Typical Layout of Oconec 2 MU/HP! IlDe '

I (wy%,F6<ues'ify'n*[. [;

>J

,A, fezzle was to be cut off at A A

c,.;w y

- y w,7p;A . .

sweepoiet nerre '

/ ~

'% g 10 * h jh't' a% ,-y ),

/

6 '/r h 4

yp . f .. ,
4f . w . a2
. m7gggar'- w man Replacement makeup nozzle and thermal sleeve design in a demonstration PWR Figure 310 plant (Flynn et al.1975). (modified).
W e
l l
- l l
i l
I I
i l
(
I
(
! i i .
I l .
i I b ,.
. , %$U ,
~ ep . ,. ". !
%m" ., 1 s.
N 7'3.,:: ., y., - $:n:g
f ,M.p. ' Sf' ,
'j y cyef -
j
< % %'m _,
.- l
, g- .. ,y
e j.-- . - l ,3 ,
i i H'
='"
. a.~Qs
}.~ s i
I i
r. ;a 1
i I
1 l
l i
i . i I
i 1
WN l l
I I 1
i Hgure 311 Observed crack on the outside surface of the MU/HPI line weld before removing i l
j the dunaged piping components (Redmond 1997) i J
l 1
1 l _ _ _ . _ _ _ . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - -
4 Tg;; .hn .
m
.-c.. -
( * ,. ', v'
~~'{Vanin5 t e
- , , Lbe l . . . . . . ,
. - gy j , . P. s M/U Line - . ..y.;;.jj
.~
!h - ,
w y;n
[<~ ': , , . . ..,
.. .. . F'~
2
' ,,+.
I i
I J
n .
< is 4...pe
..- . m : e-j @[,,,QM.'$fi@ME!&T3N'%df&eW6p 1 ':{ l a.~
' ' ~}
i N$ 'k ! .b.* .. -. . . .--. - . . .
i I
11gure 312 Ponion of the MU/HP!line and warmmg line adjacem to safe end removed for the l
j failure analysis (Redmond 1997)
I k
j t
a j
J r
?
l I
1 J
i j
f
-, - ...-- - - . ~ - - - - - . . - . - - - . - . . - - - -
l I
i oo crack Tip @ 344'(100% TW) 0*
l W
l ,
q i sor. .
.e j .
OD rac Tip i kl3$$$$$$ l !
PJ ! e:
(100% TW) 270*- $ My Ep g lml0 ...
. 0(P
+=

:=c=: _J F
496 'N :-
Ee a >
j Metallograp y t I 4:
T

:M--
4-2 i i q ; _::zr x =8 -. ,
$ 13 5 ' '
(30% TW) {
l X
Metallography ll 1 ,.' ; , l,p yH *- . (26.5% TW) ,
24r ,, - j y ? ,
w gh., ,'
~
1 .-
I,ill ' '
l . '.
210* ,
! : ! M '150'
. Metallography (24% TW) y,g, a View is looking in direction of M/U flow Warming Line Flow cc ,
1 l'igure 3-13. Angular distribution of depth of the circumferential crack at the safe end-to-MU/HPI line weld.
I
_ _ _ . _ _ - - . . . _ _ _ - - - - - - - - - - _ . _ - - - . - . - - - . - - - - - . ~ . , . . . - - - , , , - , . . , . , . . . - - , ,
,.,-,-v. , - , , - , - - - , - . - - , , - , -
I l
1 1
l i
1 I
l i
i I
i \
j i i
i i l I
I .
l t-s-
, . ; , ,-; 9 y.:. w ? -: . . J yv. s .
I t
l
. f. ' . a,e ..
4.g s
i I
~ ., a w .
=- -,.
' '
2.t:'; b6..at+h n
, s3 g./ *'
on 's L'f._ 'y.g*,-
n.
y,Ba,,,MLMjr- >
, , ,i.: . ; - .. A r A s
' 2?U* c 5C97 1067 car em f
Hgure M4 Downstream end of the worn MUHPI noule thermal sleeve The view is from the downstream side of the tractor coolant flow direction (Redmond 1997) ,
a 4
5
.. _ _ . . _ _ _ _ , . . . _ _ , , . . . , , . - , .w- ,.---,-+-.e-,---
_ _ _ _ _ _ _ _ _ _ . . . . _ _.-__.~..__ ._..._ _ ___ _ _ . _ _ _ .
. _ _ _ _ _ _l
]
1 a
i l
1 i
1, h
i i,
3
^l 1
t 1
i j .
1 .
i ps - -e
,,'It<~ .,.. _,
r e
-.'s s - c.- a,.~ -
, fug =. .
e: ,~. . ;
,. . > , - - -- . , - ?. .-
A ., . . ..
a
~~
. . . .n .. .
. s;y,.,-.
A ... . w e .,.-
. . .9. .
r g , . a ;, , .- s *. - ,_.. ,.:-
. . [ ,_ 4 - '* * " -
/" Y..O j' .% -
s_
era. -- .e 4,-,,
;.? , -l q .., , _ .-
_.,u: g, j ..
...c, '.%.
4
.; ;. -, 7.-
g_
j .
3
. . * * . , ..7, o '
, . ;; 2 , :. . . . ;~ . w - - = .a '
. g . - ty-- ,, : . ..s? : -
e . , y' ,w
~. , ,, e; _ L , '-
,; :r . .,.,; . c . _ ; , . . j*1~ r..$ g .
s 'a '
~ . . . , -
~
.,..,,~,- ,7 ., . .n . . r. , s - T .. 4t, .s N* .* l[ $s . k'l.
^ 'y ys,& lc~;g , n'; a . -
.f & ,v f -
8 .
T >
J.
_' i. .. : 7 ,a.%, . , ,, A y% , ;,.. <L',1 J : : %. .~. . f; ..r ,s,. (e t .
4
- ,is .. - s .
.,...m: . w- ..
y. .... . . . .- x
-_ a l,f. 6.. ...y.e ..
M ..; w;t.%+; u .+4:'
W M8 90 Typical Came Cracking -3L3X 5/:LT7 cor em l
l Figure 3-15 Typical multidirectional cracking observed on the inside surface of the MU/HPI piping near the warming line pene: ration (Redmond 1997) i J
a l
l l
l l
l r
I
L i
I l
.i 1
a l
i l 6 1
}
f: .*. *~ dp "Q. ;'
.a.
i ~'
. ID TOP l ' '
- VIEW (0') .
1*;
i i .
' ec:.
, )*.y : '
I n.m -
I f.; riz jr i l w ?' 3[, ' .. -
l J
l . ,y, - . .:
KQW " ct~~n;+1
, m ..
&r
, l..I. .#. .g
[
u-I. '$$f w,
-g; . Mg, ; y
.,.v3 Ly -e, "
s i
,"'jrf c >.
..g. e T, . # .
. ,e.
' .<y-
.. r .4.,
c ;
1 . .
e%.c 4
-- .:: .n
,p , m. ...a qqq%:' v s
..,:.:p9
.:: :it:
n, . . -gy.y - v~
r ; 1.,
i .u. cc'g,;.,;^.-
-;*cp
-y y . ' 3'I Q 7 k[2;
ID :, .-
~- ,
5 lD RCS DOWNSTREAM + : RCS DOWNSTREAM VIEW (270') T - %. - #e n2%w M..%ff84-tf( 244 ..
VIEW (90') ,
O: - .: .- : . ID g@;,t if BOTTOM
~C' -
I .2 D_. VIEW (180')
.%~i ;,;x Sc:- ..**r_. -m q.,j. . - T '* ?:: :-- '
'(, ~., s :., *s . s * (
enom Mgure 3-16 Penetrant tesung results revealing multidirectional crermg on the inside surface of the MU/HPI safe end (Redmond 1997) b
( jj Check valve u
n ) bn
\ stratifiedflow Crack location Turbulence a
n y,,P7 1 l Safety injection line in a three-loop $fdcDfant Westinghouse plant r 7 Umin Figure 3-17 I ocadon of the crack in the weld of the Farley 2 safety injection piping W
4
l r
I I f ,
i .
, check . l
,'., + vaw
~~_
/ w l ,
I
[.
80 **** . k. . . . . .
Primary coolant systern .
hotleg
. 1 . . ,
[ l 1
Secton A A A
n i Check W I . .$s ,,, ~ ,
y
-e.-
Crack 3 e
l I
crack 1
j. \ Crack 1 i'
Q SJ
..... SectonB-8 8 Crack 2 B Secton C C
'i' i
/ss
' /(s Crack 3 Crack 2 4 on f 8ENE so mrn .
net om Hgure3-13 Location of the cracks in the base metal and welds in the Tihange I safety idection piping (Su 1990).
i
+
4
, m..--v.- ,4, ,rr--+- * -v+---+tw-t- r-+--,t- --e-re-w--w-*-w------*w .---a.--e =+ = c--,- -
-- -**~=-=-+*~~-=e--*-~-- -- ==+= 4 w +r
. _ . . .- _ _ . _ . . _____.._.______.m ___ _. _ . . . . . _ . _ _ . _ = _ . _ . .- ___ _ __ _. _ . - _ . _ . - _ _ . . . .
t
^
4
-+1 ::
c !
41 't
% 3' 2*
Cold leg , , C' _
, _ RHR 3*
A: P#"PM .-
loop A 7
', ' y=
' ;, 6* \;2 A-h
. 2:
Cold leg b 0* _
,! _RHR **' 2: O = M "9
loop C
- ~
P#"P '
/
g
- (See delnH *A')
-*4-- 3' b Boron i 4 , - 4 g.
y a
--e. ::
-+4--
V -++--
Leak' path vehe A Schematic diagram of safety injeo6en I 1 n m Q
l f/
\-
l Crock j
RCS I
~ cold leg I loop B Detail 'A" N92 0233 4 .
4 Maure 319 Schematic diagram of safety injection line in a three-loop Westinghouse-type plant (Su 1990)
, . . _ ~ , - - , - , . .~,,s_--y , _ . . - - - ,,,..,.......,mm ,- ,_ , , , - . - .. e,
l
/
s From reactor vessel u
l o
s
, ..... . . .c. . a,
, "..... i t as RCS Hot Leg A o
,/ .
To steam generator #
8 T 'dtimeter(Schedule 140) 8'10*
RHRisolation valve
' i s
i b
el I
,o, Pipe leak -
y =
[ Leakc4 pipe location J
.;' .,,, s
.... g -
y To reactor coolant .
3' 3'
.. drain tank Crack location -
\
t i
W i
t To DHR pump Hgure $20 Schematic of the residual heat removal suction line at Cenkai 1 (USNRC 1989) = ra
/
/ \
SI9TA*
0 SI 8TA" St.9TS
SI 8TB" U3 Downstream SI8TA Class 2 check valve piping Sl.7TAU " N SI8TS UI Upstream SI 7T8"'"
4. .__ . +y Cold leg Class 1 y
\
piping
, Notes: ") MAP 89 location of thermocouple
" MAP 90 location of thermocouple
, 'A' location is at top dead center of pipe;
'B' location is at bottom dead center Figure 3-21. Instrumented portion of safety injection lines on 4-loop Westinghouse plant.
i I
i i
1 N4
, l
' l Y y4talss' g.sAltt
{, x' ' b. -
h .
-l W
t s*
r '/
. r k - 4r-
- ~
4 y s s .j s O5 M, #r 3
N ,
1' t nozzle ?
fwaming line g#
^. '
4
'ges ,
. h mented portion (locations 1 through 5) of HPI line near the first check valve on B&W plant.
i e
b i
. . _ . _ _ _ _ _ - _ . . . .~- 4~ . _ _ _ . .. _ _ _ _ . . . _ . . . _ . . _ . _ _ _ _ . _ _ _ _ . . _ _ _ _ _ _ _ _ . _ _ . _ . .
1
-t i
4, 4 4-
.g i
n j / RCS hot leg j t
/"
/ e
././ !
'Q/ < w.r w -
tr.$o w j 4
l/
i ,
. ,\
,a > ,
w t s
t wei :
~ '
Branch ^
W line .
,,g #
N !
w e
F h
r 1' l ;
i Uninsulated
t>
. t Mgure 3-23. -
Instrumented portion of shutdown cooling line on older vintage Combustion l.
Engineering plant. >
r i T h' f
'P b
\,
d
?
epww,. g r. 4 .enwswwa .<-s.er-I v~w w r-x-r vwe n' m-"<,e-ne-rw-~wtv ro w -- m e s m- *-wr*~e'- - w- w r w-~r=~- ~ c~~~+~-.<-* ->-*--n' ~~'<v '
+ " '"
i e
Motor operated valve
.L-6' _f ,
Cnock P.
4 valve 3 .
6' 3' 2- 4 Q
Drain line ,
1 Thermocouple 2'.1 Placement for ;
- location 3 10'
/
/
v 4. 5 f
3 2'
1 Cold leg Thermocouple
. / placement for other locations .
F1gure 3 24. Instrn=>a'ad portion of safety Liection line on Combustion Engineering plant.
- - . . , , - - - - , ,. -- ,,.-,u. . - . . . - . , - . . . . . . ~ , . - . . . , , , , , - . , , - - - - - , - ,
- - . ~ . _ _ - - - _ - - - _ _ . - -
+C 1' 'I - 1. nominal pipe wall thickness l
\
, g$t\..m ,
\ g %
\
c,
\
s
% w. ,4
,D \D 1
Y QQ3 . , .
s \
\ 4 \
s X g
\ 3
\
\.
\
\ -
o, # ,s9S'
\
\
\
t
]
! (a) Minimtun welding dimensions for socket welding fittings.
1 1
Th jTheoretical throat l 7 eorebcalthroat I Surface of i Sudaco of vertcal membet i
j velocal member l / g. ^) / n ,.
b ;.N!
^
g .
/ Convex Niet weld Ni Concave fillet weld
! I ti'S,{i v weld
[S$?
,A ; Surface of honzontal rnember
[kh -f
, . 1s w. e d
l SLl*
1 General Note' The see of an equal leg flRet wtnd is the leg length of the largest inscribed d -- ->
'd 1 nght isosceies triangte. Theoretad throat = 0.7 x size of weid. l 1
j (b) Equalleg fillet weld.
i i <
i F1gure 4-1 Socket and fillet weld details and dimensions 4
i
-em- www w,w w e-ee--w -w-v s .e,a w w - w ww mww w- *~- * - -----=,wv~,-,--+v,we-w.-r-.*~m-v,%-e---
1 l
i
, jRun pipe ,
/ ,
. );/ 1 h/ Full penetration
c' and fillet weld x
Fillet weld i s - - .,
ll \ (a) m . ., ,- . ,i A \f -
l Small bore l
\
Half coupling branch pipe
/
94487r0 Coupling
, j Fillet weld , ;
i
\_ N\\W ,, <r ~, o
,, , / g ,, ,,
/ ,sr /: H ,,1
\ (b) y , g </ s:,1 sit <
%\\'\f S " 87 4 Small bore pipe I
90' elbow ;
l 5 .- ,, - 40@N y irs
'N J
/ ,,u, 8 v , , ,
8
\P \ \ (C)
Fillet weld ,f
$,I 4 4
f jQ Small bore pipe estoss Figure 4 2 Examples of socket welding fittings,
l l
l .
i-1 l
r I
I ,
s 4
\'
\
\ *
\ S**\ % fe\6 e
\
\
\
y U*
\
% , ?y i
\ J }$[. ?v?ai _*
% a:f p hf GjIkb{;b ' $i M
+x:< fR +f.#w> ,
i
~ $s: '.;T,' gy Ig - , .
\
\
\
\
S
\
x
\ q
}.y i gt*\*
G4 i at toe j '
d cnom a
)
l ,
j ngure 4-3 Socket weld failure initiating at toe of fillet weld.
i 1

,--,-u_---,--w--_n--,- a_.~~.-,------.,,+,__n---n-m.-_an-,,n,awv_m-~-<-wwww -~w,- , - . ,n - , ,---n. e---,a

i l
i
i l
', I i
i 1 1
i !
j !
!i l
i a
/
\\ ;
l l / Pipe way l
A ^
/ ,
l
'**W Villet wejg '
/
l Crack N
/
i V/
/
i kassn
\ oot R of wegg OUplin (see p;g ,
l Cmck hidated
' W lack of fu,,0n at the root .
]
cn or>o I
J j
i Figure 4 4 Socket weld failure initiating at root of fillet weld.
l l
l
)
350 -
- As of 12/05: ,
BWR- 1366 reactor years .
PWR 3226 reactor years 300 -
,a
250 sBWR '
f 2x OPWR 3 150 - -
z 100 - - - -
i - ,
50 80 82 84 86 88 90 92 94 0 72 74 76 78 c8*
Year Figure 4 5 World-wide data for vibratory fatigue failures of small-diameter piping (s 25-mm diameter) in LWRs over the period 1972 to 1995,
l E
B 1000 .
I 4 =-
96%
_w -
~#/ i r
100 =
5
[o -
50 %
a
. ase " ,
7-9 Jr '
[ oBWR = .
10 __
APWR 55 1
10000 100000 1000000 .
=
Time to Failure [ Hours]
Figure M Hazard function plot represendng world-wide data for pipe leaks and naptures in reactor coolant system small diameter piping (s 25-mm diameter) caused by vibratory fatigue cracking.
5 i
E12 GVibratory fatigue events s- + Frequency (events / years) 4
- R1 7 e
N-l
. s- - aos E 1E 3.L
$ 5 E e , v_
O - tog 3 g 4, 1 - r -
o
$ h g a
~
3 3 4 7 94 e
G-i e 2 ;! 4 2 M 2-7 l g .
f ; g l .
6 ;. J j ; - aar 5 4 l 2 e
3 y
la i 4
. E M i 4 _
/ / / / / / / / / / //
Calendar year ,yp I !
Figure 4-7. Distnbution of reportable vibratory fatigue-related leak events and their frequencies by calendis year. The data provided for 1996 is through and including the 3"' quarter.
I
. . _ . _ ~ , .
.E e
i Leak events / operating plants in age group g 2
- ! 1 p .
+ (l,3 .!
"+ .
og ves
- -r . . ,.
l
' e,$ ,
l' Ag
. Il o4 g E 'w a e, 2
o hc n x. . .- .
5e "s' n ----
w ... ,, 8 sg 5 e
6g n <
~
e u .
p o
- e .5
. ; a a .
f%
stueAs Jo JegwnN 9 9
A
.h fi
/
c
, , - . , , , , -, ., + - , - -
30 --
2s -
2 E i 8 20 -
O I U 1 o j i
E is . l S
l e !
3 '
.50 to
! S 8
i '
i o 1 5-O ll N S S S ') S f
6 4
9 @ @ e 0 + @ @ 4 s @ f f f f f f f f f f.P
Years in service r3da_*_mn
Figure 4-9. Cumulative distribution of reportable leak events caused by vibratory fatigue by plant age (years in sennee).
1 i
l I
t
- . . - . - - - . . . - - - . - - _ ~ . . . . _ _ - . ~ , - , . - . ~ . - - . - . - _ -
l I
t I
I
. i s
i i
4 i
2.80E41 l [ Point est. wW 90% contdence interval Plued rate
- - - 90% conedonoe band on tw need rate 2.00E 01 -
e a 1.60E 01 N '
s ,,
s .'
W i.00E.01 - "
I i % it w k 9he
~ - '~~ ~
% %. o 0,80E 01 o ,
- - . . .- --- . '~~----.
. o_ ,.] N %
[. ,, , ,
'~~~~l
. . ,, ., 1 0.00E+00 l 85 86 87 88 89 - 90 91 92 93 94 95 96 w w ammi .
Figure 410. Statistical analysis of a trend in the reportable vibratory fatigue related leak event -
frequencies estimated by cal-ad year. Point estimates,90% confidence bands, and 90% m% -
intervals are shown.
d 8
9 t 4
?
t
,---m- -,e ~-.-,.,-,,,,,,-~-n-,, , - - - , - - < - ~ . - - , - ,,--.y ..v--
. . , . . . _ . . - . . . . . - . . . -. ~. -- . . . - . . - . . - . . - - . - . -
r f
m m .
,_\'.~
i i .
- 1.0e+oO -
to
, { Point est. and 90% con 8dence interval 1.51.5 n n .,
Pued rem
- - - o0% aanadence band on te need raw
. 4 7.5E 01 i
g i- 5,0E-01 4
b a
4 4
2.SE 01
, .. n c
7 ,
,A,_i_I_ % ".
'I 'r e L:, ,, . ._ _ _ _ _ _.....
, , , --tttt~!r r : ,. ; -; .- ,1 _ _ _. ,l __ _ - --- --- ._ .
1 2 345 6 7 4 9 101112131415161718192021222324252627292930 y,,, ar .
i Figure 4-11. . Statistical analysis of a trend in the reportable vibratory
- f atigue related leak' even*, frequencies estimated by plant age (years in service).- Point -_ estimates, 90% confidence bands, and 90% confidence intervals =are shown.
1 S
4 G
a.
, . . _ 7,_ .- ,_ ~ , .- -, . ._ -
o t
a s
i Smalkbore pipe ,[ .
. V' M
f Half emM .g p-La5 bore pipe
(*) (
Socket welding blind-flange combination -
-Tubing (b) ,
90* elbow f'd (C) wm Figure 412. Typical cantilevered small-bore pipbg connected to a large-bore piping.
i
. _ . . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - _ _ _ _ -_ . - - - --r- m m-< r
i u
i
~
To 73
, press:rizer v ooo pais g
(vedeble)f 35o T psig Component i 75 g - molin9 e---
'k
-1: d : :
I l-'
T TC? i 2.o e 45 - -2, :: i rCv i i
-i, O FCV " _1 I I ,_.. _ { &
{ y l From residualheat Letdown heat gg 200 psig lu I removal system exchanger
Regenerative a g : td heat exchanger (
I h l I
"5T lu I s4s r ( g Charging flow g
I tw 22so peia l l From : 490 r Loop C FCV FCV $ Flow-controlvalve(FCV) @ Temperatura control g
cold leg "
g >< Normally closed valve h Pressure conkol Toloops and cold legs t* Flow measuring orifice l @ Flowindicator I
loside containment i * * " '
Figure 4-13. Typical Westinghouse letdown system for a 4 ^oop plant.
i
= -
e .w i,e
a t
L Exam. surface j l' AB ',! Profile of valve body, i
vessel nozzle, or 1 in. pump connection
y:
t i
. v2 h Weld end buttering i
!,, . ^ *
(whe"e applied)
,,.']m9
. i M
i l
Figure 4-14. Inservice inspection region for socket welded joint. i 1
I e
l l
l l i
l t
l l
l 1
1
~- _~ ^
Purrgdroer
- & i inn em4*g
\ . 8 /
" (
a U
' f A '
5 -Four Sealpackage mechancal seatsin cueurs stud l series N -
[ . seal 1hermalbanter\;
Dran e rescer
= =* i lll @
.E Toleakage senerg Te oogs resstance .
temperature a l
e f M i,] -
jiW g a 4 t-M. = "d i
c. .m _ . .c
(
g'
} y[ [99j m, s e ManImpeAx N92 0027 Figure 5-1. Byron-Jackson reactor coolant pump seal with an injectionless seal cooling system (USNRC Technical Training Center)
i l
~
1 Vapor sealleakage >t <
s
' )
to redwaste (0.3 gph) ' '
F1 g x eter h; Controlled bleedoff 20 psig mn. to VCT (1.0 gpm)
Upper seal back pressure p -
m e,:
%ddle seal a res e breakdown g
1500 psia g_ , y Lower seal n Auxiliary impeller '
Ik irculating v 2250 psia nprimary cociant Thermal barrier [J .ooooo.
^
(40 gpm)
Integral heat 3,
. exchanger Corsponent Hydrostatic bearing' 9 t.
t m 8 gpm) h, .'
" I*E C' '
Mixed flow 1.0 gpm from pump casing i
I u,u i cv een Figure 5-2. Combustion Engineering supplied Byron-Jackson seal and seal bleedoff system schematic (Ruger and Lukas 1989).
O rings
/
cG$$Y '
,f/
jf
" ,/
?f f ; /, /
/ . . //f/
/ ///jk r Stationary face
\elp['Gg hi J4 1&$- '
N g/
[g'fR:$i 1,
p
\
sph'"2 bE!
N 3r.4.- s Rotating face ring fhl'hEf, 4R;.Wq ,A -
Ll$k[hffh \ -
NEW
/M
/ w ' 'a*S '==
h'/~ Lockingrin
"*"**"$g k gStr %///z u .
mwwg-s Y7{f ..
p
[.. Rotating
. ace body h'#9 + W;'d.
h hl
/W Radial grooves Dn%?%
~'
N for Y-notch -
J/Ox #$ISJp?! '
\ U<,up f follower
/
g M[ /
U-cup
}lg[hrlhh flW /
/ b kN gg p{%t t
g mic.y
/
jwa , ,
W Coil spring
^/
==
Figure 5-3. Schematic of a ByronJackson reactor coolam pump mechanical seal stage (Bell and O'Reilly 1992)
i Gland folkmer Stem Gland nut %M M/
Gland flange ye x Gland bolt - - -
I Mal '
[ Stuffing box Comprecdvs ,
0 Stress p p -
/{ Radial Sealing Stress N
~
,; ' !/ X -
( h ' g, l -
/ -
Z, s
/ \ E5
.HHHH \ $$
8$ iHHHH \_
Q3 59 4H+ L \_ mm
$ $ ~~
b
$) W E
y 4
- Asbestos packing rings Valve body twt System pressure F - Gland pressure L- Depth of the stuffing box Figure 7-i. Typictl asbestos-based valve stem packing design origmally used in nuclear- ,
power plants. Typical axial and radial stress distributions across the packing rings are shown (Ruggied and Kelly 1988).
ll Gland follower Stem
/
C
,e Gland nut %p
- Gland flange xye s Gland bolt-j j u u
^ ~
, Stuffing box Lantem ring Leak-off port N Asbestos packing rings Vatve body c System pressure F - Gland pressure en Figure 7-i. Typical asbestos-based valve stem packing design with a leak-off sytem (Ruggieri and Kelly 1988, Hart 1996).
Stem Y ..
/p Gland flange ^
l / ////b o '1N .-
'N -
Braided packing end rings
\ Die-forned packing tir gs Stuffing box Carbon spacer si w N v ..
Figure 7-3. Standard packing arrangement recommended for square packing Gland follower is*not shown (Ruggieri and Kelly 1988). ,
4 Gland follower r.
$m .f
-era o
' 3 extemal \
dio-formed y 4 Intemal die-formed wedges .
f wedges Stumng box y Carbon spacer k /
N@B/
xt e Stem
. x v ccn ne f
- Figure 7 5. Standard packing arrangement recommended for wedge-shaped packing (Ruggieri and Kelly 1988).
4 4
~
\
Gland stad Gland nut washer : .
o Gland nut Intemal spring guide + ; j r
- <+
, r+;
i e Spring stack I I
y Gland flange
?,,
/
e t p
v
/ ['/ Gland fcilower Stumng box - /k Packing rings Braided packing Mon spacer _
end nngs l
vatve stem Figure 7-5. Live-loading ofvalve stem pachng using disc springs (Ruggieri and Kelly 1988).
14 - - 0.35 EDPacking events 2
c 12 + Frequency (events / years) _ o.3 3 '
s E y.
10 - - c.2s e ,
7 2
g s I $ :
g e-i 1
- 0.2
g. ,
. g o i 5 E.
s -- - o.15 f l f -
L %
4- { a - o.s . S
[; d
' %m
h ; 2 2 2 2 y 2- f g
- o.05
$~ g a i o = = = = o l l $$$$$$$$$$ '
Calendar year ,m, t
Figure 7-6. Distribution of reportable packing degradation-related leak events and their frequencies by calendar year. The data provided for 1996 is through and including the 3"' quarter.
i i
0 t
K1 *
7-EPacking events
- 0.2 6-
+ Frequency (events / plants) E g
3 . -
K e
G 5- o '
4 h
3 x
~
, - 0.16 E i O h
> I' 4 -
4- ,
l IU l* *
. + e.
a
0) 3 3 ;
k ) U
- 0.1
.O 3- e
' V g
E
! E ? '
1 3 l l 3 '
Z ' :
I! i 5 2 f 2 -
2 [;
3 2- - l
]
m I ,
g l i t
f , to i
' - 0.05 9 e
1-l -
11 t
i i t ,
g-1 i 9
1 1 1
, 3 l 1
g. c -
, I, ,
D 0- 'E - '
= = = = = = = = = = = =-0 ASS h S 6 6 * * $ 9 @ + s4 0 ss st S 4 s @ f f f f f f f f f f f Years in Service y., a_, _
! Figure 7-7. Distribution of reportable packing degradation-related leak events and their frequencies by plant age (years in service).
i As b
n ._ . . . , . .
b i
b C 0E 01 j Point est. and 90% conddence ir'tervet Fmed rate i
- - - 90% consdence band on the 9tted rate g
~
4.0E 01 g
\
\ ,,
\
g 3.0E-01 o
\
1 f \ i W 2.0E 01 o\
\ \
\. 1 s is , ,,
% i N 1.0E41
'd 's, ,,
~~~~
o,0E+00 --~ ' . -
85 86 87 88 89 90 91 92 93 94 95 96 ceressi Figure 7-8. Statistical analysis of a trend in the reportable packing degradation related leak event frequencies estimated by calendar year.
Point estimates, 90% confidence bands, and 90% confidence intervals are shown.
4
b i
r 1.0E+00
3.0
{
Pomt est. and 90% conRdenas interval 1.51.5 &
Rited rate a 4
- - - 90% con #dence band on the fitted rate 9
7.5E41 5.0E 01 .
W ,,
2.5E41 ,, , i
,, ~
o - -
o ,, . . , -
T y li' A
, I. i .I d -
t i j 17
r - r -f L m -T
- --~~~~~~
0.0E+0c -
r =-
1 2 3 4 5 6 7 8 9 101112131415161718192021222324252627282930 -
Year co'mer 8 Figure 7-9. Statistical analysis of a trend in the reportable packing '
degradation related leak event frequencies estimated by plant age (years in service).
l I
l l
l .
l l
h J
Adaptor Nut
> --> 4-- Gap
, .-..-.._..-...-.-.--.._... _..._..._..._.7
\
$ 'Af N N W J4 % Qyl$$j wmxxg,g , ,up .
es7oc2 T ,I : $5.,.. - ' . . . . . . .A I':
w- .-.- ,u. .
.; . c i Front ferrule Back ferrule Figure 8-1 Schematic of a Swagelok fitting. Crimping action of ferrules on heavy- and thin-wslied tubing is shown.
9 t
4 =m.
l l I l
l i
I i .
i a
i i Nut i Adapter -
4
, Femile ,
t I
! Adaptor shoulder i
l
-+ Gap +---
1 1
! Figure 8-2 Schematic of a Parker-Hannifin compression fitting showing nut that is torqued onto .
4 . adaptor to compress tube into adaptor shoulder. A gap less than the nominal value may indicate that fitting is properly seated, 4
4 d
1
}
- _ _ - - . .- .. - .. -- - - - . - ~ - . - - - - . _ - - - . - - - _ . _ _
s-4- -- - ------ --- ---- - -------- ------ - - ---- ----------- ----
i
{
$ 3- -- ------------ - --- ------------- ------------------ -------
e
< o 2_ __ ___ _____,_ ._______________ ____________________ _______
l :s 2
1- - --- ----
o 1985 1986 1987
--I----------------I--I-----------
1988 1989 1990 1991 1992 1993 1994 1995 1996 Year ElemdO1 Figure 8-3 Trend in U.S. nuclear power plant compression fittir:g leaks shows relativeN< stable rate of one failure per year (1985-1996). The data provided for 1996 is through and including the 3* quarter.
l l
Root in,strument line W"- valve Hotleg
"9 }'
r ' tubsig Compression fittings Main Main steam steam h . .
h
L a L i Steam generator .
-+ <.- .
+ + + +
l l l
l I
j 1 r 1 r 1 r 1 r g Reactor coolant pump
- - 0 Feedwater at at
+ +
V Reactor l Emergency _
feedwater Figure 8-4 Schem=+ showing how six compression fimngs in series were used to reduce the instrumentation line on the Oconee 3 hot leg from 19-mm (3/4-in), to 9.5-mm (3/8-in.) tubing.
1 . . ..
i 75
.j 70 -
65 -
CS - -
SS y 53 - -
so . .
45 - -
o ao - -
g 35 - -
y so -
25 -
g -
z 20 - -
15 - -
10 - - 8 9 :
5- -
1 0
/////
// / (/ /
Method of detection * ' = - '
i Figure 10-1. Distribution of the methods used to detect the reportable reactor coolant leaks. The total exceeds the number of reportable leaks (199) because some of the events were detected by more than one method.
l
Containment monitor Control room alarm 29 events (18.2%) :
7 events (4.4%)
Unknown /other 8 events (5.0%)
.. n.., ",. ;
i
V ; :,
^
~
g . E'Q . s Control roomind. %. * ~P' T- d 49 events (30.8%)~
ug, TN '
g; ;
^ s ,
i Periodic surveillance i 47 events (29.6%)
/
Personnel observation 19 events (11.9%)
v2 Figure 10-2. Distribution of the rnethods used to detect the reportable leaks inside the containment. The raw data counts are given with the percentage for each category.
,.- - - r -,-
t i
(
{
Control room ind:
6 events (13.3%)
r.
i Persorinel observation Unknown /other 20 events (44.4%) ' ~10 events (22.2%)
l l
\
Periodic surveillance 9 events (20.0%)
j v3 1
Figure 10-3. Distribution of the methods used to detect the reportable leak events outside the containment. The raw data counts are given with the percentage for each category. ,
l l
Control room ind. Surface pyrometer Unknown measurements 14 events (7.0%)
23 events (11.6%) 3 events (1.5%)
f yt :n:i: ,
,x, '
. " N
~i, . ' :
O!
c- ;
i
\ l Personnel observation v4
159 events (79.9%)
i i
Figure 10-4. Distribution of the methods used to locate the source of reactor coolant leakage that occurred inside or outside of containment. The raw data counts are given with the percentage for each category. !
l l
i l l
4 . . , ,
i
., , _ ., _ . _ . . . _ ~ . _ . .
-$ o "
o o .
%, E
:s e
o E EE um 8 ee 88 o o
v 'd nM f v7 %4 A LN"maem @C O *
- a e o O 80 c"J DE o o y e
O E
x E o / I m Eg y>
4 - E @
go a -- --- %, E 3
, 7 e-o o
%.e $o f,
~n -r o o e
C o
59 a
em 6,. ,.mu a i sa x
-8 e %'
s aV a u$
N ::::
4 ,O _5 o2 "o
v -
"}
~
o
%4 u
?
O 90 <
8 m z.
n ::::. .. y ei o O 6,t = ~
C n n n n o = = Q*< 8 00 0 =
s>1 eel Jo .isqwnN $
4 o
v=
m CD E
.. ~ .. . - . - _ _ _ - . - - - . - - . - . - - - . . ..
'N : .,3 : .;
'. U P
~'
pr
\ ;
}
4 r ,
.f 4
7 5
_f'
~ 1.E-03 ,
.e p-. ;___ _ *
1. Conseinment vessel tes volume - 73,700 m' - ,
2. Portculate acNty concentenon .e in reactor cooient 1.5 x 10' betyr?
3. Persson fecer 0.1 i
T 1.E 04 e /
5- i som
b. ? /
8 1.E 06 1
0.1com s
[ Detectable limit 1.E 06 - ,
r
/
1.E47 -
0 .10 20 '30 40 .. 50 - 80 - 70 *
, Time afterlook occuned (min.)
4 Figure 10-6.- Time dependent particulate actmty increases resulting from re.ctor coolant leakage (Aoki 1991).
9 o
I l -. s
^
l... _ _. .. _ , .. ;,. .. , - . _ , - - -
i s
10 Assumpeans.
1. Xe-133 ace 4 in reactor emolant 4 x 10' p'
2. Corneinment vessel free volume 73.700 m 10 gpm leakage
~
1 .
6. $
4 gpmleakage 2 opm lookege
.... ........ ............. ....... @ . p ...... ..
g 0.1 .
$ gpm i k ,,
ie K
.. ..............................................mh.m.e.... ........ .
i 0.01 0 10 20 30 40 50 60 70 80 Time after leek occunsa (min.) - cw==>
Figure l')-7. Time dependent gaseous activity _ increases resulting from reactor coolant leakage (Aoki 1991).~ ,
P u- W-- -
-e
'-- r 1
2.0 . .
~
n 1.5 ---- -----------------------
j1.0 .
0.S ,
e 0.0' r
< 0 10 20 - 30 40 50 60 Time after condensation occurred (min.)
Figure 10-8. Time dependent condensate flow rate increase resulting from a 3.8-Umin (1-gpm) reactor coolant leakage (Aoki 1991).
3 ..
- - , . = * + = - -
, e -
i d
ASSESSMENT OF PRESSURIZED WATER REACTOR PRIMARY SYSTEM LEAKS o
J
. TABLES i
e e s - - -m, - , - - , --._-, , ,,,,--_-,_,-_ _ _ _ ___ _ , _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
4-14 4.=@simJa.r-- W 3 A 4 4,--h m J . sam,Jah 412M.4.. al.h1Ma d,A ,.h ha.J 3M A hw4.# p,44 a y__M A J a. w m*w-- -
-Ah4-5..--3.4w een ,,, $4 M A,J Qb-S S J .) -- --' - 4f yM A g A4.4,4. M.Wemhw4 J
I.
I.
k i s D
E
.I t
?
L L
t I
t.
i
?
k I
i i
f, h
f4 O
e h
4 9-f-
I e
s 1
n i
l I
i
)
i ke i
1 j O
a e
P l
i i
i I
t l
I 1
l l
i l
i i
i
, . , . . - - , r,- ,e.mo r. ,, . , , - . - , , , . , ~ ,-,- ,- m.~y..,--,,_.~....,. w. -.. ., - . . ~ , , . . . . . . -...- ~ ~-. ..-., ... . - . - - - - , _, , _ , , _ .--., __s
Table 2-1.' Reportable leak events caused by three different stress corrosion cracking mechanisms.
Maximum Degradation Event Plant Leak Rate Mechanism ' Dat: Plant Name Age (yrs) (gym)' Systan Location IGSCC 12/01/94 Summer 12 3 0.2 Reactor coolant pump W eld IGSCC 02/22/96 Surry Unit 2 23.1 < 0.1 Residual heat removal Weld PWSCC 01/02/92 Palo Verde Unit i 7 < 0.1 Pressurizer Base metal PWSCC 04/24/87 ANO Unit 2 8.7 < 0.1 Pres atzer Base tretal PWSCC 05/05/89 Calvert Cliffs Unit 2 12.4 Unknown Pressurizer Base metal PW,CC 12/22/90 ANO Unit i 16.6 < 0.1 Pressurizer Base me'al i
PWSCC 02/18/92 San Onofre Unit 3 93 < 0.1 Pressurizer Base metal PWSCC 02/27/86 San Onofre Urdt 3 33 0.15 Presurrizer Base metal ,
PWSCC 03/14/92 San Onofre Unit 2 10.1 < 0.1 Pressurizer Base metal PWSCC 07/22/95 San Onofre Unit 3 1 12.7 < 0.1 Pressurizer Weld PWSCC 07/22/95 San Onofre Unit 3 12.7 < 0.1 Main coolant piping Unknown PWSCC 10/10/95 St Lucie Unit 2 12.5 < 3. I Main coolant piping Base metal SCC 01/15/93 Turkey Point Unit 3 20.5 < 0.1 Pressurizer Weld TGSCC I?lle/90 Fort Calhoun 17.4 0.4 C.nwrol rod drive Base metal !
TGSCC 09/12/95 Surry Unit 1 23 3 < 0.1 F-ider Weld 1 TGSCC 08/21/93 Surry Unit i 213 0.2 Steam generator Weld i
' I gpm = 3.8 Umin.
Taide 3-1. PWR reactor coolant leak events caused by thermal fatigue.
Plant Event Initial NSSS Flying Tweeghwell Crack Leek Date Criticality Vender Sydenn Rate Date IAt8838" S 38 '
(Uinin.)
Crystal 1/82 In7 B&W Makeup /High Check valve tody near the leadegree circumferential 3.8 ,
River 3 Pressure valve 40-safe end weld crack; two crack initiation i Injection siees: one en the inside surface and twee on the aussede surface Farley 2 12/87 5/81 W Safety Heat affected zone of clbce-to- Crack essended 120 degrees 2.7 Injection pipe weld ci.s_.i.ma ::, at the inside surface,25-mm long at the curside surface Tihange 1 6/88 2n5 ACLF Safety ERow base metal 89-ann long at the inside 23 Injection surface. 41-nue long at the Outsade surface Genkai 1 6/88 105 MHI Residual Heat Heat-affected zone of elbow-to- Crack extended 97 mm 0.8 !
Removal pipe weld cartumiume;.::j at the inside surface,1.5-inm at the outside surface i Three Mile 995 6n4 B&W Cold leg Weld between a 90-degree Crack extended 51 mm 0.06 Island 1 Drain Line elbow and a SI-mm diameter circumferent'. lfat the inside I honzontalline surface,14 mm at the otsside '
surface C. p. u 1" 12/96 3/80 Fra Safety Base metal of a straight portion < 3.8 Iniect

a of the pipe Oconee 2 4/97 lin3 B&W Makeup /High Safe-end to pipe weld Crack extended 360 degree 11.0 max Pressure u. Mum-.#y at the innde i Injectum surface, about 77 degrec 4
circumferential1y on the outside surface 8
Several odier EDF plants have eyu;u cd thermal fatigue-induced leaks in branch lines in 1990s: C_.c.. 2 in 1992; and C .p... .
3, Cmas 3, Saint-Laurent Bl. Tricastm I and 4 Bugey 3 and Gravelines 4 since 1993. l
Table 4-1. Summary of PWR cracking in pipes smaller than 4 inches IJcesssee Event Rep.st System Plants Citations Crack Iscation Probable tm' Chemical and volume Arkansas 2 1 Most cracks in vekis Fatigue caused control Calvert Chffs 1 4 located near pumps by vibration Cahcrt Cliffs 2 5 Haddam Neck 2 Fort Calhoun 1 2 indian Ptimt 2 3 Indian Poird 3 1 Kewaunee 1 2 North Anna 1 1 Palisades 1 1 Point Beach 2 1 R. E. Ginna 1 3 Salem i 1 Surry 1 1 Turkey Point 3 I Turkey Point 4 3 Yankee Rowe 4 Zion 1 1 7)on 2 1 Coolar< recirculation Arkansas 2 1 In welds at small tees Vibrction Cahert Cliffs 2 3 and nipples, etc.
Fort Calhoun 1 1 Palisades 1 2 Point Beach 2 1 Salem 1 1 Three Mile Island I 1 Indian Point 1 1 San Onofre 1 i
i i
TaWe 4-1. Summary of PWR cracking in pipes smaller than 4 inches (corninued) i i
i f Licensee
Event R* Pert sysee Pi es Cases Crack L.e a-- m6.ue Cmme l Residual heat removal '
Arkansas 1 3 Most cracks in welds Vibration ;
, D.C. Cook 1 2 near pumps or valves Indian Point 2 3 l Prairie Islan:11 1
'Ihree Mile Island 2 1 Three Mile Island 1 1 t
i ;
i Reactor coolant cleanup Calvert Cliffs 1 3 Most cracks in welds in Fatigue caused i
Calvert Cliffs 2 3 lines located near by vibration ;
! Kewaunee 1 1 pumps j Trojan 1 1 j Yankee Rowe I i ;
i i Emergency core cooling Arkansas 2 1 Most cracks in wekts of Vibration !
Beaver Valley 1 1 vent or drain lines i l Calvert Cliffs 2 1 l Farley 1 I i l Millstone 2 1 1
a
&am2 1 l Main steam supply North Anna 1 1 Cracks in weki in Not determined !
instrument imes l
Condensate feedwater Three Mile Island 1 1 Crack in socket weld Vibration
Other engineered safety Turkey Point 3 1 Crack in drain line weld Vibration features i i
l
I i'
I I e . .
Tame 4-1. Summary of PWR cracking in pipes smaller than 4 inches (continued) .
Ianm Event R* Pert Sysseus Plants Citations CrackIme=d== N Cause Reactor cos.: isolation Oconee 3 1 Crack in weld in sample Vibration cooling line 1
l Spent fuel pool Arkarsas 1 2 In welds Not determined i
1hree Mile Island 1 HAZ IGSCC 1
, Containment heat removal Indian Point 2 1 Vent to pump wc!d Vibration ;
e 4 ,
I e
9 I
T t
l i
Tame 4-2. Examples of fatigue failures in small (< 2 in.) PWR lines connecting i the reactor coolant system (Stoller Corporation) !
Year Plant Line Dimmweer Locatism Comuments 1994 Diablo Canyon 2 (W) accumulator vent 3/4 socket weld fatigue, originated at weld defect ,
i 1992 Vogtle 1 (W) drain I socket' weld fatigue
, 1991 Nine Mile Pt. 2 (GE) flex hose 3/4 socket weld fatigJe 1991 North Anna I (W) valve pressurization 3/4 weld fatigue, from external source ,
1989 Vogtle 1 (W) safety valve drain 3/4 socket weld fatigue, drain manifold vibration 1989 Yankee Rowe (W) loop bypass vent 3/4 weld fatigue j 1989 Robinson 2 (W) thermowell 0.777 threads in fatque joint j 1987 Hatch 2 instrument I socket weld fatigue -
1987 Monticello (GE) decontamination 2 fillet weki fatigue toe 1985 Rancho Seco (B&W) OTSG vent I pipe fatigue, lack of support 1984 Palisades (CE) dp sensing 3/4 socket weld fatigue, cantilever design, lack of wekt g_ ._ :.gjon I 1983 Calvert Cliffs (CE) pump seal b'eedoff I weld fatigue 1981 Dresden 2 (GE) leak-off 3/4 socket wekt fatigue 1981 Beaver Valley 1 (W) instrument weld corrosion fatigue 1979 Brunswick 2 (GE) test connection 3/4 ripe-to- fatigue elbow weld 1917 Brunswick 2 (GE) tes connection 3/4 socket weld fatigue, -_W line 1977 Millstone 2 (CE) dp pressure tap 1 socket weld fatigue, occurred at reW joint 1976 Millstone 2 (CE) do pressure tap i socket weld fatiane Y
Table 4-2. Examples of fatigue failures in small (< 2 in.) PWR lines connecting to the reactor coolant system (Stoller Corporadon)
(conunued)
Year Plant IJne Dianweer 14casfem Comuments 1975- Millstone 2 (CE) 3 dp pressure taps I socket weld fatigue, vibration frorn RCPs 1975 Millstone 1 (GE) instrument 3/4 longitEhnal fatigue, corrugated tubing weld 1974 Browns Ferry 1 (GE) flow sensing i fillet weld fatigue, improper support toe 1974 Browns Ferry 2 (GE) valve bypass 2 fillet weld fatigue, at connection to valve toe 1974 Arkansas 1 (B&W) cold leg drain I% socket weld fatigue, failed again 8 days after repair
i l
Table 51. Number of reactor coolant pump r.eal configurations tued at CEOG plants in 1983 and 1991 (Combustion Engineering 1991).
Seal %udwM!.# 7:
N:< 9dumbiiw N N M N Surebs'[^@- '$dk m - , r, .... ~ ~s 3,.,
, ' Mis (1985f ' ~; .
?* ' W!ME:l. ..g M, ,3;M; KSB 0 12 Byron Jackson SU 47 23 Byron Jackson N-9000 0 7 Bingham Willamette 0 16 ,
AECL CAN 4 0 1 k
0 t
Table 5 2. Reactor coolant pump seal leak events reported during 1985 to 1996 period.' 3 e $ ;-
Waterford 3 38285006 02/20/85 3.0 North Anna Unit 2 33986010 08/l1/86 4.3 Catawba Unit 1 41386028 06/04/86 1.8 Palo Verde Unit 2 52986041 07/01/86 < 10 Arkansas Nuclear One 36888011 08/01/88 40 2
Palo Verde Unit 3 53089001 03/03/89 6.0 0conee Unit i 26992009 07/22/92 6.1 St. lxcie Unit 1 33595004 07/31/95 1.8
1. The period does not include the last quarter of 1996.
2. I gpm = 3.8 L/ min I
1 l
l l
%'*'- T ^W-'" T'Y~ +
W' m .- y. _ _ _, _ , _ , _ _ . . _ _ _ _ _ , , _ _ , .
Table 9-1 Leak everas analyzed in the Accident Sequence Precursor Program.
PLANT LER DESCRIPTION CCDP C95 TALENTS Rancho Seco 31285025 Failure of the Integrated Contre 8 System 1.6E-5 Event analyzed because of the widespread unpact of resulted in RCS overcooling. A vahc losing the integrated control system. RCS lineup problem caused destruction of n -wwooling, and failure of a high pressure injection high pressure injection pump and seal. pump. RCS leakage was not a factor.
Catawba 1 41385041 During a power reduction due to 6.7E-6 Event a.nlyzed because of multiple secondary s>Sem unidentified RCS leakage, main feedwater failures. Dese failures were not associated with the was lost and 3 steam generator relief unidentified leakage.
valves failed to open.
Pala Verde 3 53089001 Reactor trip with failure of several non- 4.9E-5 hvent analyzed because of bus and atmospheric dump safety busses. Atmospheric dump valves i valve failures. De seal leak wasjudged to be failed to actuate. reactor coolant pump insignificant.
l (RCP) seat injectien was secured resulting in 6 gpm seal leak.
Beaver 334930s.' Ten offsite power feeder breakers opened 5.5E-5 This event was analyzed because of the dual unit Valley I resulting in LOOP for both units. Unit 1 LOOP. The RCS leak was not a factor. ,
was at power, Unit 2 was in a refueling shutdown. On a post-trip leak inspection a small leak was found on loop 1 A cold leg vent valve pressurization connection weld.
Indian 24791001 Restoration of pressure transmitter after 2.0E-6 This event was analyzed because of the safety system ,
Point 2 repair of leaking compression fitting failure during a transient. The auxiliary teedwater caused a reactor trip. An auxiliary pump failure was not linked to the RCS leak. This ,
feedwater pump failed. event contributes the frequency of a plant transient.
Salem 2 31190005 Leak on welded pipe cap on discharge side 1.3E-6 This event was analyzed not because of the leak of the boron injection tank required directly, but because leak isolation rerxlered the high isolating both high pressure injection pressure injection system inoperable.
pumps.
Fort Calhoun 28592023 During a reactor trip, a pressurizer safety 2.5E-4 This event is a transient-induced LOCA typically relief valve opened and failed to rescat. modeled in a PRA.
21,500 gallons of coolant was released from the RCS.
Table 9-1 12ak events analyzed in the Acckient Sequence Precursor Program. (continued)
PLANT LER DESCRIPTION CCDP COMMENIS 41386031 Loss of a motor control center caused loss 3.3E-3 This event was analyzed as a small LOCA with a leak Catawba 1 of control power to letdown variable rate of 130 gpm.
orifice. The orifice failed open and the resulting transient caused a rtillotine rupture at the orifice flange.
St. Imcie 1 33595004 Reactor coolant pump 1 A2 lower seal stage 5.6E-6' The seal on reactor coolant pump I A2 could have failed. During repairs two additional for degraded further and failed completely, resulting in a stages failed. Additionally there were RCS small LOCA. This was how the event was analyzed.
problems with power-operated relief valves seal and shutdown cooling. Each was analyzed leak separately. ;
' The estimated CCDP for only RCP seal leak is listed here. The estimated CCDP for the PORV unavailability present during this event is not included.
1
Table 9-2 Risk impact categorization of the RCS leak LER database.
PLANT LER DESCRIFTION CAT JUSTIFICATION Palisades 25595010 A fitting failed in a pre:sure sensing line. 2 De reactor trip was a direct result of the instruirent Irak rate was 12 gpm. A total of 800 line failure.
gallons was lost. He reactor was tripped.
Millstone 2 33688009 Leak from reactor vessel head caused 2 A dropped rod is a reactivity control event typically failure of a control element assembly considered in a full-scope PRA.
resulting in a dropped control rod.
1 Indian 24791001 Compression fitting leak on pressurizer 2 He reactor trip was a result of the leak repair.
Point 2 pressure sensing line. Operator error Without the leak the reactor trip would not have during restoration caused reactor trip. occurred.
Summer A scactor trip was caused by a pressurizer 2 The reactor tiip was a direct result of the safety valve 39589011 safety reli:f valve opening at normal failure and subsequent RCS leak.
operating pressure.
Palisades 25585009 A partially open drain isolation valve 3 The drained safety injection tank was out of wvice res"Ited in draining a safety injection until the valve could be closed and the tank refilled.
tank. l Turkey 25088024 RIIR pump seal failed. Pump was 3 Failure of an RilR/LPI pump directly impacts the Point 3 isolated to make repairs. operability of a safety system used in mitigating actions for a number of different accident types, including injection and long-term cooling.
Ginna 24494009 Pinhole leak on socket weld for 1.5-inch 3 All three SI pumps were declared inoperable.
pipe connected to SI pump recirculation line.
IIaddam Neck 21395020 A leak developed in a weld on the 3 He pump was isolated to make repairs.
discharge pressure sensing line for charging pump A.
Ilarris 40088033 A leak developed on a safety injection 3 The safety injection pump was removed from senice pump equalizing line. to make repairs.
Surry 2 28188021 Low head safety injection pump flow 3 He low head safety injection pump was taken out of element fitting leak. service.
l Table 9-2 Risk impact categorization of the RCS leak LER database. (continued)
PLANT LER DESCRIP'IlON CAT JUSTIFICATION I Harris 40093005 A leak developed on a safety injection 3 The safety injection pump was wed from service pump seal. to make repairs. ;
i McGuire 1 36986001 A leak developed from a charpng pump 3 The charging pump was isolated to make repairs.
4 shaft seal. - -
Kewaunee 30593019 A leak developed on an RHR pump 3 "Ihe RHR pump was isolated to make repairs. !
casing.
j Salem 2 31190005 Irak on welded pipe cap on discharge 3 Both HPI pumps were isolated to make repairs.
i side of the BIT required isolating both ;
) HP! pumps. !
Rancho Seco 31285025 A make-up pump seal failed. 3 The makeg pump was isolated to make repairs.
DC Cook 1 31589012 A leak developed on a safety injection 3 The safety injection pump was isola 2d to make !
pump casing. repairs.
, DC Cook 2 31686026 A leak developed in a safety injection 3 Safety injection was isolated to make the repairs.
cross-connect valve.
Callaway 1 48386038 A leak developed from a PORV block 3 'Ihe PORV block valve was closed to isolate the leak. i valve stem.
Cow rds 44596005 A weld cracked on the common header 3 Both trains of safety injection were isolated for six Peak I between the cold legs and both safety hours.
injection pumps.
Diablo 32389006 A presst:rizer PORV was weeping past its 3 The PORV was isolated to stop the leak.
CanyonI seat.
. Beaver 33485015 Failed 0-ring and gasket assembly seal on 3 Both low pressure safety injection pumps were ;
l Valley I several seismic wedge control rods for rendered inoperable. !
both low pressure SI pumps.
1 t I
j Millstone 3 42388015 A leak developed from a PORV block 3 The PORY block valve was closed to isolate the leak.
j valve stem.
i Beaver 41294008 A PORV was leding past its seat. 3 The PORV block valve was closed to stop the leak. !
Valley 2
]
t
Table 9-2 Risk impact categorization of the RCS leak LER database. (continued)
PLANT LER DESCRIITION CAT JUSTIFICATION Trojan 34488006 A leak developed from a charging pump 3 The charging pump was isolated to make repairs.
shaft seal.
Trojan '34488040 A Icak developed from a charging pump 3 The charging pump was isolated to make repairs.
shaft seal.
Trojan 34490039 A leak developed from a charging pump 3 De charb ng i pump was isolated to make repairs.
shaft seal.
Trojan 34492024 A leak developed from a charging pump 3 De charging pump was isolated to make repairs.
shaft seal.
San Onofre 3 36289008 A leak developed from a low head safety 3 The pump was isolated to make repairs.
injection pump seal.
Palisades 25586017 Letdown relief valve failed to fully close 5 Relief valve failures in response to a transiert are tv,r after a pressure transient. definition transient-induced leaks with the potential to become LOCAs.
Indian 28695014 Pressurizer PORV failed to rescat after a 5 The PORV was isolated to make repairs.
Point 3 pressure transient.
Waterford 3 38285006 RCP seal failure occurred after a loss of 5 His type of event is typically modeled explicitly in coupnent cooling water. tl.e PRA event trees involvir.g a loss of cunWa cooling water.
Palisades 25586Gi7 A reactor head vent valve failed to close 5 Relief valve failures in ioycirs to a transient are by i after a pressure transient. definition transient-i.x!uced leaks with the potersial to become LOCAs.
l Ft. Calhoun 28592023 During a reactor trip, a pressurizer safety 5 This event is a transient-induced LOCA typically l relief valve opened and failed to rescat. modeled : a PRA. This event was analyzed by the !
21,500 gallons of coolant was released ASP Program. ,
from the RCS. '
Palo Verde 3 53089001 A RCP seal leak developed due to a loss 5 His event is a typical transient-induced RCP real of seal injection flow. leak with the potential to become a seal LOCA.
I e ~
Table 9-2 Risk impact categorization of the RCS leak LER database. (continued)
PLANT LER 8%'MIPTION CAT JUSTIF1 CATION IIarris 40087058 198 gallons of coolant were lost when 6 This event had the potential to 1 ecome a small LOCA reactor vessel head vent valves spuriously had the valves failed open. The leak was terminated opened during a surveillance test. by placing the valves in pull 4o-lock.
ANO-2 36888011 An instrument line associated with a RCP 6 This event required an HP1 ptunp and a normal seal failed resulting in seal failure. charging pump to maintain pressurizer level The leak was terminated only wien the RCS pressure was reduced to atmospheric and the RCS was drained down below the seal area.
Catawba 1 41386031 less of an motor control center caused 6 This event was analyzed in the ASP Program as a loss of control power to letdown variable small LOCA with a leak rate of 130 gpm.
orifice. The orifice failed open and the CCDP = 3.3E-3.
resulting transient caused a pi!!otine rupture at the orifice flange.
St. Lucie 1 33595004 RCP 1 A2 lower sea stage failed. During 6 The seal on RCP 1 A2 could have degraded further repairs two additional stages failed. and failed completely, resulting in a small LOCA.
Additionally there were problems with This was how the event was analyzed in the ASP PORVs and shutdown cooling. Program.
I Calvert 31794007 A pressurizer safety relief valve lifted 6 The valve opened prematurely due to a manufacturing Cliffs I prematurely and did not fully rescat. defect and failed to rescat due to damage caused ;
during the lifL The leak had the potential to become a srr:II LOCA.
Calvert 31787006 A pressurizer safety re!ief valve disc and 6 The leak had the poterdial to become a small LOCA.
Cliffs I seat were stuan cut.
Oconee 1 26992009 The wrong seal was installed in a reactor 6 Irs;;allation of the wrong RCP seal could have coolant pump and subsecuently failed. resulted in a complete failure of the seal, which would easily exceed the capacity of the charging system. Such an everr woukt be a small-break LOCA.
i i
-,w -
Table 9-3. Risk impact parameter frequencies. ,
CATEGORY COUNT FREQUENCY (per reactor year) 1 199 ,
2 4 6.27E-3 [
i 3 24 3.76E-2 p/'., 4 0 0 5 6 9.4E-3 6 7 1.1E-2 ;
' Category I represents no risk impact, therefore the frequency is undermed.
n
Table 10-1. Reactor coolant leaks outside the containment that were detected by control room indication.
.1 u sv.: g pa,- ' - - -
s m ,,
, *m
. W, c3y;Mm;sg -st, ,gg3pg.317c '#
s t -~. R_;:,: o g ;c iANO Unit 2 a 36892003 i Reactor ooonant chargingSetdown 130
, iBraidwood Unit 1 456880081 Residual heet removal 12.4
Seat tN/A J
{
iMcGuire Unit 1 369870291Reactorcoolantcharging/ Letdown mi e Packing I hobinson Unit 2 26195004!Regetor coolant charging / Letdown 111.2 : Seat I
~
, iSakm Unit 1 27290022T6herging/ Safety injecuon i3.0 . Seat t iSurry Unit 1 280930131Regetor coolant charging / Letdown !25 : Seat i (1) One gallon per minute (GPM) equals 3.8 liters / minute.
(2) All six of these leaks were from valves.
Tchle 10-2. Capabilities ofleakage monitoring methods indde containment (USNRC PWR Pipe Crack i Study Group 1980). j i
Ieak Leak Leak detection measurement location Method sensitivity accuracy detection Air radioparticulate activity monitor' F(P) F (P) F(P)
Radioactive gas activity monitor F(P) F (P) F(P)
Air cooler condensate flow monitor' O(G) F (F) P (P)
Sump level and flow monitoring L8 G(G) G (G) P(P)
Primary reactor coolant inventory 8 G (N/A) G* (N/A) P (N/A) ,
(based on makeup ' low integrator)
Tape moisture sensor G (G) P(P) G (G) -
(
Temperature sensor F(F/P) P (P) F(F)
Visual inspection' F(F) P (P) G (G) llumidity-dew point monitor F(F) P (P) P (P)
Pressure sensor F (F/P) P (P) P (P)
Acoustic emission monitor G(G) P(P) G (G)
' Regulatory guide 1.45 requirements.
8 Quantitative determination possible.
8 For PWR during steady state conditions.
Leakage mear.urement accuracy is good if determined over a long time period (6-24 hr).
8 Access to cor.ainment during normal power operation restricted; therefore, usefulness of this method is limited.
NO TE: G = good, F = fair, P = poor for assessing leakage from RCPB; N/A = not applicable; designations in parentheses ( ) are for leakage from any system inside containment.
4
Tame 10-3. Capabilities ofleak detection systems in a typical Japanese PWR (Aoki 1991).
Minineone detectable lesk rate Response tisme la enee of Seeece ef tenhage to -
Iank detection s;seese withis I beer 3.5 kg/meia. Ieekage be detected
' Aircoolerwan less than I spm within I h high-teenperature liquid or secem measuring system Containment sump level less than I gpm approximately I h all monitoring system Containment air particulate approximately 0.1 gpm assuming within I h high-temperature liquid monitoring system no fuel defect and 1.5 x 10' contain~mg radioactive particulate Bq/cm' corrosion product (resesor coolant) concentration in coolant Containment radioactive gas approximately 2 gpm assuming 4 within 2 h Iquid or steam containing monitoring system x 10' Bq/cm' Xe"' concentration radioactive gas (reactorcoolant in coolant with fuel defect)
Table C t. Summary of analysis for trend in calendar time.
r%s Stat. Signincant? Eng. SigalAcant?
All reportable reactor coolant leaks, extremely significant b = -0.17 199 events p.value = 0.0001 Reportable leaks from packing degradation, extremely significant b = -0.59 29 cvents p value = 0.0001 Reportable leaks from bolted connections, almost significant b=-0.11 22 events p value = 0.096 Reportable leaks from vibratory fatigue, not significant b = 0.08 29 cvents p value = 0.016 .
Reportable leaks of pressure boundaries, not significant b = 0.002 45 events p value = 0.97 ,(
9
- . - - . - .. . . . . . - . - . - . . - - - -- - ..-..- - ~. . . - - - - - . . . .
6 Table C 2. Summary of analyses for trend wth plant age, f t
Class Stet. Signi8canee? Eng. SigalScance? e Reportable leaks from vibratory fatigue, almost significant b = -0.06 ;
29 events p-value = 0.055 Reportable leaks from packing degradation, - very significant b = 4.10 -
29 events p value = 0.002 4
?
6 h
h I
i i
. I j '
l h-i t
v 5
, . . - _ . . _ - . . . _ . - . _ . _ . . . . _ _ _ . - . . _ . . . . , , . ,_ _ . . . . _ . . , . . _ . . , . _ . . . . - - - . . _ _ . _

NUREG/CR-6582, Forwards,For Review & Comments,Draft Rept, Assessment of Pressurized Water Reactor Primary System Leaks, NUREG/CR-6582,to Meet Planned Issuance Schedule.Comments Requested to Be Submitted by 980130

Contents

Text

SUBJECT:

SUBJECT:

SUMMARY

SUMMARY

Deardortf,

=

Navigation menu