ML20236V555
| ML20236V555 | |
| Person / Time | |
|---|---|
| Issue date: | 07/28/1998 |
| From: | NRC (Affiliation Not Assigned) |
| To: | |
| Shared Package | |
| ML20236V551 | List: |
| References | |
| NUDOCS 9808040041 | |
| Download: ML20236V555 (70) | |
Text
--.
- a f
l EVALUATION BY THE OFFICE OF NUCLEAR REACTOR REGULATION RELATED TO THE REVIEW OF THE TOPICAL REPORT BY THE BOILING WATER REACTOR VESSEL AND INTERNALS PROJECT:
"BWR REACTOR PRESSURE VESSEL SHELL WELD INSPECTION RECOMMENDATIONS' (BWRVIP-05) l EXECUTIVE
SUMMARY
This safety evaluation (SE) presents an assessment of a proposal submitted by the Boiling l
l Water Reactor Vessel and Intemals Project (BWRVIP) to reduce the scope of inservice inspection (ISI) of BWR reactor pressure vessel (RPV) circumferential welds, included in this evaluation is a broader assessment of BWR reactor vessel integrity, a discussion on a class of l
beyond design basis events which dominates BWR reactor vessel failure, an assessment of the i
vulnerability to axial weld failure that requires additional study, and development of improved l
flaw density and size distribution for reactor pressula vessel welds. The identification of the vulnerability from a class of beyond design basis events confirms the value of including a broad, l
l risk-informed approach to assessing licensing changes. Based on this assessment, the staff has concluded that the change in risk associated with elimination of BWR RPV circumferential weld examinations is negligible and that the BWRVIP proposal is therefore acceptable.
The BWRVIP, a technical committee of the BWR Owners Group, proposed to reduce the scope of ISI of BWR RPV shell welds in the Electric Power Research Institute (EPRI) proprietary.Smt TR 105697, "BWR Vessel and Intemals Project [BWRVIP), BWR Reactor Pressure vomi Eht 11 Weld Inspection Recommendations (BWRVIP-05)." The report initially proposed to redra the l
scope of inspection of the BWR RPV welds from essentially 100 percent of all RPV sMil welds l
to 50 percent of the axial welds and zero percent of the circumferential welds; however, as modified, it proposes to perform ISI on 100 percent of the RPV axial shell welds, and eliminate l
the inspection of all but a few percent of circumferent!al shell welds. As originally submitted, the i
j principal basis in the BWRVIP-05 report for eliminating the inspection of circumferential welds l
was an assessment of design basis transients throughout the life of a BWR reactor pressure
?
vessel using probabilistic fracture mechanics, This assessment concluded that the 10 year l
ASME Code hydrostatic test was the most challenging transient for the RPV, but that the frequency of RPV failure was extremely low.
l On August 14,1997, the staff issued an independent safety assessment (ISA) report which l
presented the staff's assessment of the BWR VIP proposal. The staff's ISA took a broader, risk-informed approach than the original BWRVIP-05 report, and considered plant transients outside the design basis that could potentially challenge BWR reactor vessel integrity. This assessment led to the recognition that a set of beyond design basis transients, referred to as low temperature over-pressure (LTOP) events, dominates the estimated failure frequency of BWR RPVs. In these type of transients, the RPV, while in a cold condition (e.g., on the order of 100*F or 38'C) is filled solid with reactor coolant, resulting in an increase in pressure up to 1000 psi (6.9 MPa) or greater. At these relatively low temperatures, the reactor vessel material's resistance to brittle fracture is greatly reduced and the conditional failure probability for the reactor vessel is significantly higher than that associated with design basis events.
ENCLOSURE 9008040041 900728 "
PDR TOPRP EXIEPRI C
The staff's ISA r: port id:ntified one occurrznce of an LTOP svInt which occurred in a U.S.
i design BWR in a foreign country and included a precursor study which resulted in an estimated d
frequency of LTOP events of 3 x 10. The BWRVIP performed an assessment of this event and it resulted in an estimated frequency of an LTOP event of about 9 x 10d Based on the information provided by the BWRVIP, the staff estimated the frequency of LTOP events to be 4
1 x 10. The BWRVIP's assessment indicates that the major contribution to this event frequency results from unmitigated injections from the condensate or control rod drive systems and a failure to properly realign the reactor water cleanup system following a reactor trip at low temperatures. The probability that the operator falls to take action to mitigate coolant injection is a key variable in assessing the frequency of these evants. The one actual event that has.
occurred in the approximately 1700 years of worldwide BWR operation yields an observed d
frequency of 6 x 10. The LTOP pressure and temperature analyzed by the BWRVIP was 1200 psi (8.3 MPa) and 100*F (38'C). The event that occurred in the foreign BWR resulted in a pressure of 1150 psi (7.9 MPa) at 88'F (31'C).
The staff's ISA report presented conditional failure probabilities for both circumferential and axial welds under LTOP events. The staff's conditional failure probability estimates were based on generic RPV material properties associated with RPVs fabricated by different vendors. The BWRVIP subsequently provided estimawd circumferential and axial weld conditional failure probabilities on a plant specific basis for all domestic BWR plants.
As presented in this report, the BWRVIP and staff estimated limiting (i.e., for the plant with the I
worst end-of-license material properties) conditional failure probabilities for circumferential welds 4
4 subject to an LTOP event are 1 x 10 and 8.2 x 10, respectively. Corresponding conditional failure probabilities for axial welds are on the order of 1.6 x 10 to 4.4 x 10, respectively. The 4
4 difference in the estimated conditional failure probabliities is mainly due to (1) differences in the i
assumed pressure and temperature of the event in that the staff used the temperature and pressure from the actual observed event rather than the BWRVIP's estimated pressure and temperature for the LTOP event, (2) the staff's use of an updated flaw density and size distribution, and (3) differences in assumed material chemistries. The updated flaw density and size distribution utilized by the staff was based on data developed in the NRC Research Pressure Vessel Research Users Facility (PVRUF), and represents a major improvement in the RPV probabilistic assessments.
Combining the estimated frequency of LTOP events and the conditional failure probabilities for such events yields BWRVIP and staff estimated failure frequencies for BWR circumferential welds of 9 x 10* and 8.2 x 10 per year, respectively, and for BWR axial welds of 1.4 x 10 to 1
4 4.4 x 10" per year, respectively.
The staff's assessment of the effectiveness 'of inservice inspection indicates that it has a relatively small effect on the estimated failure probabilities. For axial welds, the best estimate reduction in conditional failure probability due to inservice inspection is about a factor of 2 to 3.
The effectiveness of inservice inspection may be greater for circumferential welds because circumferential welds failure probabilities are dominated by larger crack sizes (which are easier to detect) than axial. However, it is not expected based on current technology to result in more than a factor of 10 difference. In the context of the guidelines in Regulatory Guide (RG) 1.174, the marginal change in failure probability of circumferential welds, vf.:.h is already very low, is negligible.
11
f RG 1.174 provides guidelines as to how defense-in-depth and safety margins are maintained, and states that a risk assessment should be used to address the principle that proposed i
increases in risk, and their cumulative effect, are small and do not cause the NRC Safety Goals to be exceeded. The estimated failure frequency of the BWR RPV circumferential welds is well below the acceptable core damage frequency (CDF) and large early release frequency (LERF) l.
criteria discussed in RG 1.174. Although the frequency of RPV weld failure can not be directly l
compared to the frequencies of core damage or large early release, the staff believes that the estimated frequency of RPV circumferential weld failure bounds the corresponding CDF and LERF that may result from a vessel weld failure. On the above bases, the staff has concluded that the BWRVlP-05 proposal, as modified, to eliminate BWR vessel circumferential weld' examinations,is acceptable.
It should be noted that the failure frequencies for axial welds cited above are relatively high, but that there are known conservatism in these estimates. For example, these analyses were based on the assumption that the flaws in the axial weld with the limiting material properties and chemistry are alllocated at the inside surface of the BWR RPV and at the location of peak end-of license (EOL) azimuthal fluence. Since flaws are distributed throughout the weld and the EOL neutron fluence will not occur for many years, the staff has concluded that the present RPV failure frequencyis substantially below that reported by the BWRVIP,and independently l
calculated by the staff, and is not a near term safety concem. The staff is pursuing this subject further with the BWRVIP in order to assure that the estimated failure frequency of BWR vessels is significantly lower than indicated by these estimates. In a letter to Carl Terry, BWRVIP Chairman, dated June 8,1998, the staff has requested that the BWRVIP provide a plan for l
followup analyses to determine, on a more realistic basis, the potential for axial weld failures due L
to cold over-pressure events and appropriate technical approaches for addressing this concem, i
as necessary.
The ASME inservice inspection requirements for axial welds are not being modified. As discussed above, the effectiveness of these inspections may be limited in attempting to control the distribution of defects in reactor vessels; however, they will provide important information l
regarding potential unanticipated mechanistic degradation mechanisms that could potentially invalidate assumptions made in the current assessment of reactor vessel reliability. Should results of the axial weld examinations reveal the development of such a degradation f
mechanism, the need for inspection of circumferential welds will be reassessed.
The staff is developing a generic letter informing BWR licensees that relief from circumferential reactor pressure vessel weld Examinations, per the proposal in the BWRVIP-05 report, will be considered on a plant specific basis. This generic letter will be issued for public comment prior to the staff granting any final reliefs. It should also be noted that this safety evaluation is limited to the period of the current operating license. The requirements for inspection of circumferential reactor wssel welds during a license renewal period will be reassessed, on a plant specific basis, as part of any BWR license renewal application.
ill
TABLE OF CONTENTS i
EXEC UTIVE SUMMA RY........................................................ I TABLE OF CONTENTS........................................................ iv
1.0 INTRODUCTION
1 1.1 Backgrou nd.........................................................
3 1.2 Description of Independent Safety Assessment (ISA)........................
4 2.0 STAFF EVALUATION....................................................
5 2.1 Comparison of Vessel Welding Processes................................
5 2.2 Statistical Analysis of a Sampling Approach to RPV Weld Examination..........
6 2.3 RPV Degradation Mechanisms..........................................
8 2.4 Discussion of Limiting BWR RPV Transient................................
9 2.4.1 Inadvertent injection During Shutdown.............................
10 2.4.2 Hydrostatic Tests or Vessel Leak Tests.............................
11 2.4.3 injection During Maintenance Activities.............................
11 2.5 Evaluation of Probabilistic Fracture Mechanics Models......................
11 2.6 Integrated Probabilistic Assessment....................................
11 2.6.1 Frequency Estimation of Cold Over Pressurizing the RPV..............
12 2.6.1.1 Frequency Estimation of Cold Over Pressurization Events from inadvertent injections....................................
12 2.6.1.2 Frequency Estimation of Cold Over Pressurization Events from Condensate injections...................................
13 2.6.1.3 Frequency Estimation of Cold Over Pressurization Events from CRD injection...............................................
14 2.6.1.4 Frequency Estimation of Cold Over Pressurization Events from Loss of RW C U.................................................
1 5 2.6.1.5 An Actual Cold Over Pressurization Event....................
15 2.6.1.6 Estimation of a Total Cold Over Pressurization Frequency.......
16 2.6.1.7 Conclusion............................................
16 2.6.2 Conditional Probability of Vescal Failure............................
17 2.6.2.1 General Discussion of Conventional Vessel Analysis Codes.....
17 2.6.2.2 Discussion of Limiting Transients...........................
17 l
2.6.2.3 Technical input to the NRC Staff's PFM Evaluations............
17 2.6.2.4 Evaluation of Flaw Size Distribution.........................
18 2.6.2.5 Sensitivlty to the Flaw Size Distribution......................
19 2.6.2.6 Sensitivity to inservice inspection...........................
20 2.6.2.7 Comparison of Sensitivity Studies and Plant. Specific Analyses...
20 2.6.2.8 Parametric Study for Generating the Relationship Between P(FIE) and (T RT,m ).............................................
22 2.7 BWRVIP Risk-informed Assessment...................................
23 2.8 BWRVIP Recommended New Inspection Criteria and Scope.................
24 2.8.1 Successive Examinations of Flaws................................
24 2.8.2 Additional Examinations of Flaws.................................
24 2.8.3 Staff Evaluation of Successive and Additional Examination Requirements for Circumferential Wolds..........................................
25 2.8.3 Staff Evaluation of Successive and Additional Examination Requiremer ts for Axial Welds...................................................
25 iv
R_--______--_---__-__-_-__--._--------
1 i
' (
3.0 CON C LU SION S........................................................
30 4.0 IMPLEM ENTATION......................................................
33 5.0 R E FE R ENC E S.........................................................
33 APPENDICES A
PVRUF-Exponential Distribution............................................ A-1 B
Parametric Study for Generating the P(FIE) Versus (T-RT Limiting Transients.............................. 7) Curves............... B-1 C
..................... :.. C-1 D
Comparison of Flaw Density and Flaw Distributions Used in the BWRVIP Anal and Staff Analysis Method (s).......................................ysis Meth
........ D-1 l
E Additionalinformation to Be Provided to the Advisory Committee on Reactor Safeguards j
( AC R S)................................................................ E-1 TABLES TABLE 2.6-1 Comparison of P(FIE) Using PVRUF-Marshal and PVRUF-Exponential Flaw Distributions for Reference Cases.................................
26 l
TABLE 2.6-2 Results of ISI Sensitivity Analyses for Axial Flaws.....................
27 l
TABLE 2.6-3 Non Conservative Chemistry Factors from BWRVIP-05................
27 I
TABLE 2.6-4 Summary of Results of NRC Staff and BWRVIP Limiting Plant Specific Analyses (32 EFPY)....................................................
28 TABLE 2.6 5 Summary of Results of NRC Staff and BWRVIP Limiting Plant Specific Analyses I
(64 E FPY)....................................................
29 TABLE 2.6-6 Distribution of Vessel Failures as a Function of Flaw Size for Limitin Specific Cases..........................................g Plant-30 TABLE A-1 Flaws in PVRUF Welds (excluding cladding and base metal regions)...... A-2 TABLE A 2 NRC Staff Flaw Size Distributions and Density........................ A 2 i
TABLE B-1 Summary of Results for NRC Staff Parametric Study................... B-3 i
TABLE C-1 Sampling Reviews of LERs and ens................................ C-7 TABLE D 1 Flaw Size Binning Used for Evaluation of P(FIE)....................... D-3 4
TABLE D-2 Number of Flaws Within Broad Flaw Depth Ranges.................... D-4 TABLE D-3 P(FIE) for Limiting Plant Specific Case of Circumferential Flaws (Mean RTwy =
99.8'F) Using Different Flaw Parameter Sets
........................D-4 TABLE E-1 Impact of Plant Life Extension from 32 EFPY to 64 EFPY on Conditional Probability of Vessel Failure P(FIE)................................. E-6 l
FIGURES Figure B-1 BWRVIP-05 PFM Results from Staff Parametric Study................. B-4 Figure B-2 Lower Bound K, and K. Test Data for SA 533 Grade B Class 1, SA-508 Clast 2, and SA 508 Class 3 Steels................................ B-5 Figure D 1 Flaw Depth Distribution for Each Vessel Simulation.................... D-5 Figure D-2 Number of Flaws Within Each Flaw Depth Range..................... D-6 Figure D-3 Ratio of Number of Flaws to the Largest Number for the Flaw Depth Range. D-7 v
EVALUATION BY THE OFFICE OF NUCLEAR REACTOR REGULATION i
RELATED TO THE REVIEW OF THE TOPICAL REPORT BY THE BOILING WATER REACTOR VESSEL AND INTERNALS PROJECT:
"BWR REACTOR PRESSURE VESSEL SHELL WELD INSPECTION RECOMMENDATIONS' (BWRVIP-05)
1.0 INTRODUCTION
By letter dated September 28,1995, as supplemented by letters dated June 24 and October 29, 1996, May 16, June 4, June 13, and December 18,1997, and January 13,1998 (References 1, 2,3,4,5,6,7, and 8, respectively), the Boiling Water Reactor Vessel and Intemals Project (BWRVIP), a technical committee of the BWR Owners Group (BWROG), submitted the proprietary report, "BWR Vessel and Intemals Project, BWR Reactor Pressure Vessel Shell Wold Inspection Recommendations (BWRVIP-05)." The BWRVIP-05 report evaluates the current inspection requirements for the reactor pressure vessel shell welds in BWRs, formulates recommendations for attemative inspection requirements, and provides a technical basis for these recommended requirements. it initially proposed to reduce the scope of inspection of the BWR reactor pressure vessel (RPV) welds from essentially 100 percent of all RPV shell welds to 50 percent of the axial welds and zero percent of the circumferential welds; however, as modified, it proposes to perform inservice int.pections on 100 percent of the RPV axial shell welds, and essentially zero percent of the circumferential RPV shell welds, except for the intersections of the axial and circumferential welds. Approximately 2 - 3 percent of the circumferential welds will be inspected under this proposal. Revised criteria for the performance of successive and additionalinspections are also recommended. A cost benefit study was performed to compare the existing vessel shell weld inspection requirements with the recommended inspection criteria.
Reference 2 provided supplemental information in response to two Nuclear Regulatory Commission (NRC) staff requests for additional information (RAls) dated April 2 and May 20, 1996 (References 9 and 10, respectively). In Reference 3, the BWRVIP modified the BWRVIP-05 original proposal to increase the examination of the axial welds to 100 percent from 50 percent, while still proposing to inspect essentially zero percent of the circumferential RPV e
shell welds, except that the intersections of the axial and circumferential welds would have been included; approximately 2 - 3 percent of the circumferential welds would be inspected under this proposal. Reference 4 provided additional supplementalinformation in response to a third RAI dated May 20,1997 (Reference 11). References 5 and 6 provided the BWRVIP's VIPER computer code and detailed programming information for the VIPER code; the VIPER code was used by the BWRVIP in obtaining the results detailed in the BWRVIP-05 report. Reference 7 provided additional supplementalinformation in response to an NRC staff RAI dated August 14, 1997 (Reference 12). Reference 8 provided i. supplemental information in response to questions from the NRC's Advisory Committee on Reactor Safeguards (ACRS) which were forwarded to the BWRVIP in a letter dated October 10,1997 (Reference 13).
The staff met with members of the BWRVIP and their consultants on several occasions to discuss issues related to the review of the BWRVIP-05 report. These meetings were summarized in the following meeting summaries: summary of July 18,1995, meeting dated July 25,1995; summary of March 19,1996, meeting dated March 26,1996; summary of October 15,1996, meeting dated December 10,1996; and, summary of January 16,1997, dated 1
l l
f February 13,1997, (References 14,15,16, and 17, respectively). Additionally, by letter dated April 18,1997, (Reference 18), the BWRVIP requested to meet with the Commission on this issue. On May 12,1997, the Commission was briefed by representatives of the BWRVIP and the staff on the issues related to the requirements for a full inspection of reactor pressure vessel shell welds. The above meeting summaries, as well as the transcript of the May 12,1997 Commission meeting (Reference 19) and the Commission's May 30,1997, Staff Requirements Memorandum (SRM) M970512B, (Reference 20) are available in the Commission's Public Document Room,2120 L Street, N.W., Washington, D.C. 20555.
On August 14,1997, the staff forwarded to BWRVIP, an independent safety assessment (ISA) of the BWRVIP-05 document, along with an RAI (Reference 12). The staff's ISA was a multi-disciplined, risk informed review of the safety implications of reducing the inspection of the RPV shell welds as proposed in the BWRVIP-05 report. The ISA provided a description of the techniques used by the several vendors to fabricate the reactor vessels and discussions on the i
history of non-destructive examinations, of the two degradation mechanisms (fatigue and stress corrosion) which have the potential to initiate RPV cracking or to cause existing flaws to grow, and of the limiting transients of concem. Also, the probabilistic fracture mechanics model was l
discussed and an integrated probabilistic assessment was presented. Also transmitted with the ISA along with the RAI was additional guidance on what information tne staff needed to assess plant-specific requests for relief from the inspection schedule in 10 CFR 50.55a(g)(6)(ii)(A). A i
more complete description of the staff's ISA is contained in Section 1.2, below.
On August 26,1997, the staff's ISA was reviewed by the ACRS in subcommittee and by the full i
ACRS on September 4,1997. By letter dated September 10,1997, the ACRS made several recommendations regarding this review. In a letter dated October 10,1997 (Reference 13), the l
staff forwarded to the BWRVIP an additional RAI based on ACRS recommendations. Further l
work was performed by both the staff and the industry to more fully assess the risk associated I
with beyond-design-basis events for both the axial and circumferential welds at fluence levels l
projected to be reached later in life at some plants. This additional work included (1) studies of potential precursor events in order to better quantify the potential for cold over-pressure events in BWRs, (2) additional probabilistic fracture mechanics analysis to both understand the sensitivities to various parameters and to support an uncertainty analysis, and (3) assessment of the proposed changes in inspection requirements relative to the probability of vessel failure.
g On August 7,1997, the staff issued Information Notice (lN) 97-63, " Status of NRC Staff's Review of BWRVIP-05," (Reference 21) regarding licensee requests for relief. As stated in IN 97-63, the l
staff would "... consider technically Justified requests for reliefs from the augmented examination in accordance with 10 CFR 50.55a(a)(3)(l),10 CFR 50.55a(a)(3)(ii), and 50.55a(g)(6)(ll)(A)(5) from BWR licensees who are scheduled to perform inspections of the BWR RPV circumferential shell welds during the fall 1997 or spring 1998 outage seasons. Acceptably-Justified requests l
would be considered for inspection delays of up to two operating cycles for BWR RPV circumferential shell welds only. Licensees will still need to perform their required inspections of
' essentially 100 percent" of all axial welds." The acceptability of such requests was based on plant specific information submitted by the licensee that demonstrated that the expected frequency of beyond-design-basis events, which appeared to dominate the estimated frequency of BWR RPV failure, and the level of embrittlement of the RPV were low enough to assure low probability of vessel failure during the period of relief. The staff issued schedular reliefs for j
inspections of the BWR RPV circumferential shell welds due during the fall 1997 outage season 2
1
I I.
I for four units who submittid 1:chnically JustifiId rzquists, and his issutd schedular r: lists for i
two units during the spring 1998 outage season.
On May 7,1998, the staff issued IN 97-63, Supplement 1, which informed BWR licensees that the staff was extending the period in which it would "... consider technically justified requests for relief from the augmented examination in accordance with 10 CFR 50.55a(a)(3)(l),
50.55a(a)(3)(li), and 50.55a(g)(6)(ii)(A)(5) from BWR licensees who are scheduled to perform inspections of the BWR RPV circumferential shell welds during the Fall 1998 or Spring 1999 outage seasons. AcceptablyJustified relief would be considered for inspection delays of up to two operating cycles for BWR RPV circumferential shell welds only. Licensees will still need to perform their required inspections of ' essentially 100 percent" of all axial welds."
1.1 Background
in January 1991, the NRC published in the Federal Register (56 FR 3796) a proposed rule to amend Section 50.55a to Title 10 of the Code of FederalRegulations [10 CFR 50.55a), ' Codes and Standards." One purpose of this amendment was to incorporate by reference a later edition and addenda to Section XI of the American Society of Mechanical Engineers (ASME) Code.
This included the 1989 Edition of the ASME Section XI, Division 1, and addenda through 1988.
In addition, the rule proposed to create Section 50.55a(g)(6)(li)(A) to 10 CFR 50.55a
[10 CFR 50.55a(g)(6)(ii)(A)], " Augmented examination of reactor vessel," which required that all licensees perform an inspection of RPV welds in accordance with Section XI of the ASME Code on an expedited schedule, and revoked all previously granted reliefs for RPV weld examinations.
As noted in the statement of considerations for the rule, the primary reason for the augmented examination requirement was that few examinations had been performed up to that time on either BWR or on PWR RPV shell welds. Therefore, there was a concem on the part of the staff regarding the existence of original fabrication flaws in the RPV shell welds, and the initiation and propagation of flaws during sewice. The staff had no assurance that ASME Section XI flaw acceptance criteria were being satisfied in the RPV shell welds, which had never been completely examined as part of a Section XIinservice inspection program. The augmented examination was needed for BWRs because of evidence demonstrating a viable mechanism for initiating environmentally-assisted cracks in the RPV cladding, evidence that the cladding cracks
+
can propagate into the ferritic steel of the RPV base material and, evidence that BWR reactor vessels may be embrittled more by neutron irradiation than would be predicted by Regulatory Guide 1.99, Revision 2, " Radiation Embrittlement of Reactor Vessel Materials."
The intent of 10 CFR 50.55a(g)(6)(ll)(A) was to require that licensees perform an expanded RPV shell weld examination, as specified in the 1989 Edition of ASME Section XI, on an " expedited" basis. " Expedited," in this context, effectively meant during the inspection interval when the rule was approved or the first periad of the next inspection interval. The final rule was published in the FederalRegisteron August 6,1992 (57 FR 34666).
By incorporating into the regulations the 1989 Edition of the ASME Code, the staff was requiring that licensees perform volumetric examinations of " essentially 100 percent" of the RPV pressure retaining shell welds during allinspection intervals. This represented an expansion by the ASME of the requirements from previous editions of Section XI of the ASME Code which, as far back as the Winter 1975 Addenda, had required examination of 100 percent of the RPV pressure-retaining welds during the first inspection interval, then limited examinations in the 3
intervals thereafter. Requiring every licensee to perform an extensive volumetric examination of l
every RPV shell weld at least once during the service life of the RPV was therefore consistent with the philosophy that had been expressed in the ASME Code for more than 20 years. In j
those 20 years, however, BWR licensees had requested and been granted extensive relief from performing RPV shell weld examinations. As a result, only a small percentage of the beltline welds in BWR RPVs had been examined, and no BWR licensee had completed an examination which would satisfy the philosophy and requirements of the NRC and the ASME Code.
l Recognizing the small percentage of RPV welds that were being examined, the conflict between this small percentage of examinations and the ASME Code requirements, and the fact that inspection technology had evolved such that commercial systems were available to support the ASME-specified scope of RPV inspections, the previously granted reliefs were revoked in 1992 with the issuance of the rule.
1.2 Description of Independent Safety Assessment (ISA) f i
The Commission, in SRM M970512B (Reference 20), requested that the staff consider a tiered approach in gathering additional baseline information and/or implementing the rule
[10 CFR 50.55a(g)(6)(ll)(A)). The SRM recommended that the staff's assessment (a) should address the BWRVIP proposal to examine 100 percent of the axial welds which would include examinations of some circumferential weld lengths near the intersections of the weld types to determine if this proposal could provide an appropriate level of sampling of the RPV welds, (b) should provide a comprehensive evaluation of the probabilistic analysis contained in the BWRVIP proposed attemative in determining the acceptability of a proposed technical altemative and/or in pursuing changes to the rule, and, (c) should receive appropriate review, including review by the Advisory Committee on Reactor Safeguards.
The staff performed a broad risk-informed review of the BWRVIP-05 proposal. One result of this effort was the identification of a transient at a foreign BWR of U.S. design in which the RPV was subjected to high pressure (7.9 MPa or 1150 psig) at a low temperature (26 - 31'C or 79 -
88'F). This cold over pressure transient was not included as a design basis event for BWRs and was not considered in the BWRVIP-05 report which was focused only on design basis events. However, the recognition of this transient led the staff to determine that cold over pressure transients are of sufficient safety significance to be considered.
Additionally, the staff performed a review of 229 licensee event reports and 81 event notifications which involved potential BWR overcooling or over pressure events since 1980 (see Appendix C). Of the 310 events identified,35 were identified as potential precursors to cold over pressure events of the type that occurred overseas. These types of events are of particular interest because the fracture toughness of the RPV decreases at low temperatures resulting in greater potential for RPV failure. An evaluation of the foreign event indicated conditional failure probabilities for axial and circumferential welds significantly higher than those associated with the transients assumed in the BWRVIP-05 report. These staff evaluations indicated that the conditional failure probabilities for the circumferential welds, instead of being approximately 30 orders of magnitude less than the axial welds' conditional failure probabilities (as stated in the initial BWRVIP-05 report), are about four orders of magnitude lower.
On August 14,1997, the staff forwarded to BWRVIP, its ISA of the BWRVIP-05 document (Reference 12), along with an RAI. The staff's ISA provided, in Section 2.0, a description of the 4
techniques used by the several vendors to fabricate the reactor vessels. Section 3.0 contained a discussion of the history of non-destructive examinations, and Section 4.0 had a discussion of the two degradation mechanisms which have the potential to initiate RPV cracking or to cause exiQng flaws to grow, in Section 5.0 and Appendix C there was a discussion of limiting transients of concem. The probabilistic fracture mechanics model was discussed in Section 6.0, l
and an integrated probabilistic assessment was presented in Section 7.0. The safety evaluation that follows discusses and evaluates data and analyses that were submitted by the BWRVIP and generated by the staff subsequent to the issuance of the ISA.
2.0 STAFF EVALUATION 2.1 Comparison of Vessel Welding Processes The RPV inside surface is clad with stainless steel to prevent contamination of the core and reactor coolant system with corrosion products from the carbon steel RPV. In general, the circumferential rings are clad using an automated machine welding process, and then the rings are manually welded together to form a vessel. The circumferential welds are then clad using a manual welding process. Field inspections have revealed some instances of intergranular stress corrosion cracking (IGSCC) of BWR cladding. In all cases, the observed cracking was related to the manual clad region. Thus, the clad on the circumferential welds would be more susceptible to IGSCC than the clad on the axial welds. The BWRVIP probabilistic fracture analysis evaluated this difference by assuming that manual clad (circumferential weld) had a greater rate of IGSCC initiation than the automated clad (axial weid). The staff, in its probabilistic fracture analysis, conservatively assumed that both the manual and the automated clad had IGSCC cracks that extended through the clad and combined with pre-existing fabrication flaws. The staff described and reviewed the weld processes used in the fabrication of BWR RPV shells in section 2 of the ISA (Reference 12), and a summary of that review follows.
The BWR RPVs were fabricated by Chicago Bridge and Iron (CB&l), Combustion Engineering (CE), Babcock & Wilcox (B&W), New York Shipbuilding (NYS), and Hitachi. BWR RPV shell segments were manufactured from plates which were rolled into shells with axially oriented welds. Shell segments were joined together with circumferentially oriented welds.
+
1 The axial welds that were fabricated by CB&l were double V groove welds made with a spacer bar. The submerged arc welding (SAW) process was used from the shell inner diameter (ID).
The spacer bar was removed by back gouging from the outer diameter (OD), then the shielded metal arc welding (SMAW) process was used to complete fabrication of the axial weld. The CB&l fabricated circumferential welds were typically single V-groove welds with a closed root gap. In most cases, the SAW process was used from the shell OD. Back gouging was used to remove the root pass in order to reach the sound metal. The SMAW process was used from the i
shell ID to complete fabrication of the circumferential welds.
The axial welds that were fabricated by CE were double U-groove welds with closed root gap geometry. The SAW process was used from both the shsil ID and OD. After one side was completed, back gouging was used to reach the sound weld metal. The SMAW process was used as needed to shape the back gouged cavity, and the SAW process was used to complete fabrication of the axial weld. The CE fabricated circumferential welds were square groove welds with a carbon steel backing strip using the SAW process from the shell OD. The backing strip 5
1
?
e was removed and back gouging was used to reach the sound weld metal. The SMAW process was used from the shell ID to complete fabrication of the circumferential welds.
The electoslag welding (ESW) process was used for axial welds that were fabricated by B&W.
The exception was the Big Rock Point axial welds which were fabricated using the SAW process from the shell ID. The spacer bar was removed by back gouging from the shell OD. The SAW process was used from the shell OD to complete fabrication of the axial welds. The B&W fabricated circumferential weMs were single V groove welds with a backing strip, using the SAW process from the shell OD. The backing strip was removed and back gouging was used to reach the sound weld metal. The SMAW process was used from the shell ID to complete fabrication of the circumferentialwelds.
Dresden unit 2 was the only BWR fabricated by NYS The ESW process was used to fabricate the axial welds, and the SAW process was used to fabricate the circumferential welds. The NYS fabricated vessells comparable to the B&W vessels.
Hope Creek was the only BWR fabricated by Hitachl. The SMAW process *vas used to fabricate the axial welds, the circumferential welds, and a low pressure coolant injection (LPCI) nozzle weld that is in the beltline. The Hitachi fabricated vesselis comparable to the CB&l vessels.
The SAW and ESW processes are automated process and would be less susceptible to fabrication flaws than the manual SMAW process. The ESW process, which fabricates the weld using a one-pass process, would be expected to cause larger fabrication flaws than the multipass SAW and SMAW weld processes. Therefore, since axial and circumferential welds were fabricated using different combinations of processes, the fabrication flaw st sceptibilities would be different.
2.2 Statistical Analysis of a Sampling Approach to RPV Weld Examination Of the totallength of axial and circumferential welds in a BWR RPV, approximately 60 percent are circumferential welds and the remaining 40 percent are axial welds. The staff concluded in the ISA (Reference 12) that no meaningful statement about the circumferential welds can be made based only on a statistical analysis of the axial welds. Any such statement must be batad on the assumption that botn the axial and circumferential welds are random samples from some parent population of welds. If this assumption is made, an inference can be made about the expected number of defective circumferential welds based on an inspection of the axial welds.
For the best possible situation in which af! axial welds are inspected and nq defective welds are found, an inference can be made about the expected number of defective circumferential welds based on the inspection of the axial welds:
Denote the expected number of defectiveN circumferentialwelds by d, the proportion of defective we!ds in the parent population by p, the number of circumferential welds by m and the number of axial welds by n. Assuming that circumferential welds comprise 50 percent and axial welds 40 percent of the total weld length, then m = 1.5n and o m mp. Further, with the standard confidence level of 95 percent, it can be concluded that p < C/n. Therefore, it follows that, m For the purpose of this evaluation, ' defective
- is defirmd as containing a crack-like indication that could jeopardize RPV structuralintegrity.
l 6
1
y with 95 percent confidInce, d< 4.5. Based on this bound for d, a bound on the 5
probability that at least one of the circumferential welds is defective can be derived. The distribution of the number of defective circumferential welds is given by a Poisson distribution with mean d. Therefore, a bound on the probability, that at least one circumferential weld is defective, is equal to 1 - exp(-d). It follows that this probability is less than 1 - exp( 4.5) = 0.99, with 95 percent confidence.
Based on the above evaluation, the staff has concluded that no useful inference regarding the condition of the circumferential welds can be made based on a 100 percent inspection of the axial welds only. The staff also considered the effect of including a percentage of the circumferential welds in the inspection sample by assuming a sample size of 50 percent of total RPV welds (all of the axial and one-sixth of the circumferential) verses the 40 percent in the above assessment. Including this small percentage of the circumferential welds does not alter the above conclusion.
The above statistical analysis is based on the assumption that both the axial and circumferential welds are random samples from the same parent population of welds. In reality, as discussed in Section 2.1, the circumferenitial welds are not from the same populatioa due to the different weld
. processes and other sources of variances (i.e., field vs. shop fabrication); therefore, a sampling approach is not valid.
An example of the above is the results of the Tennessee Valleg8. The augme Authority's Browns Feny Nuclear Plant Unit 3 RPV shell welds augmented examination consisted of 15 axial welds and 5 circumferential welds. The axial welds were ESW and the circumferential welds were SAW. The axial welds contained no reportable indications. The circumferential welds contained a total of 600 recordable indications *,15 of which required evaluation to be dstormined acceptable by ASME Code Section XI, lWB-3600 (1986 Edition).
Since all the recordable flaws were in the SAW type welds, this examination indicates that the SAW type welds have a greater propensity for flaws than the ESW type welds. Hence, examination results from the ESW axial welds are not representative of the SAW circumferential welds.
The above evaluation of a statistical sampling approach to RPV weld examinations indicates that relatively large sample sizes (e.g., on the order of 50 percent) do not allow a meaningful conclusion with regard to the possibility of a single or a small number of isolated unacceptable defects existing in the remainder of the population. The practicalimplication of this evaluation is that a sampling approach cannot provide reliable protection against the possibility of isolated, unacceptable conditions existing in the population of uninspected welds. An illustration of this situation would be a large defect existing in a substandard weld repair. This is an important conclusion relative to assessment of the RPV which must have an extremely high reliability.
However, this evaluation should not be interpreted as meaning that no useful information can be gathered from inspection of a sample of welds. For example, a sampling inspection can provide usefulinferences regarding the possible existence of a more wide-spread mechanistic mode of
- see March 6,1995 letter from Pedro Salas, Manager of Site Licensing, regardinD rowns Ferry Unit 3 RPV Shell B
Wolds Augmented Examination and Inservice inspection (ISI) Relief Request 31S117
- ' Reportable' indications are those that are required to be reported to the NRC; ' recordable" indications are those cf a sufficient aire to be recorded by the testing equipment.
l l
7 l
y e
degradation. An instance of this might be a service-induced cracking mechanism such as stress corrosion cracking (SCC). SCC is a degradation mechanism which can initiate RPV cracking and/or cause existing cracks to grow.
The statistical analysis, presented above, considered the validity of a sampling approach for BWR vessel weld inspections. Since the amount of inspection proposed in the BWRVIP-05 report can not provide a high confidence level that an isolated or small number of critical flaws do not exist in circumferer.tial wolds of BWRs, it is important that there is a high confidence level that the flaw size distribution and density used in the probalblistic fracture mechanics analyses account for the probability of such defects. This subject is discussed further in Section 2.6.2.4, I
which has a description of the "PVRUF-Marshall" and the "PVRUF-Exponential" flaw distributions. These flaw distributions resulted from an examination of a Combustion 5rigineering (CE) fabricated PWR vessel. The data from the Pressure Vessel Research Users Facility (PVRUF) allowed realistic flaw distributions to be used in determining conditional failure probabilities.
2.3 RPV Degradation Mechanismo Based on approximately 700 reactor-years of U.S. operating experience of BWR vessels (approximately 1700 reactor years worldwide), BWRVIP identified that fatigue and stress corrosion cracking (SCC) are the two degradation mechanisms which have the potential to initiate RPV cracking or to cause existing flaws to grow. There are two fatigue mechanisms significant to the t3WR RPV: system cycling fatigue and rapid cycling fatigue. The system cycling fatigue, associated with plant start-up, shutdown, SCRAM and safety relief valve (SRV) blowdown, had been evaluated for the vessel shell cs part of the vessel design and, because the resulting usage factors were around 0.1, fatigue was considered by the BWRVIP to be insignificant. The rapid cycling fatigue associated with plant operation when feedwater sparger and nozzle cracking was initiated was discounted because the vessel shell is remote from these nozzles.
Based on the above, the BWRVIP narrowed its investigation of the degradation mechanisms to only SCC, which was suggested by RPV cracking experience (e.g., cracks found in some clad and unciad steam generator shells of PWR vessels and in some recirculation nozzles of BWR vessels). Cracking observed in the manually backclad region also suggested the involvement of SCC.
The staff reviewed the qualitative arguments in the BWRVIP-05 report supporting the elimination l
of fatigue as a significant degradation mechanism and determined that additional analysis was required. As a rsSult, the BWRVIP provided (Reference 2) a quantitative comparison of the crack growths due to these two mechanisms. The low alloy steel (LAS) SCC growth rate relation of da/dt in/hr = 1.18 x 10" K* ksi/in and the ASME Section XI reference bilinear fatigue crack growth law for carbon and low alloy steels in water environment were used in this study. The l
j crack growth history for both mechanisms were presented for crack depths of 1.27,2.54, and 7.62 cm (0.5,1.0, and 3.0 inches). The results indicate that, when the cracks were increased into the range of larger sizes of the Marshall distribution (e.g.,7.62 cm or 3.0 inches), the crack growth from stress corrosion is more dominant. This difference is even more pronounced if the staff's LAS SCC growth rate, which is approximately 10 times larger than the BWRVIP's value, is I
used in this comparism. Hence, the staff has concluded that the fatigue crack growth does not need to be considered in this application.
8
The BWRVIP evaluated SCC of flaws in the clad and flaws in the ferritic weld metal. SCC flaws in the ferritic weld metal were determined to be more limiting than SCC flaws in the clad. Hence the BWRVIP probabilistic fracture mechanics analyses only evaluated SCC flaws in the ferritic weld metal.
The initial flaw size and frequency of SCC flaws in the ferritic weld metal are those assumed in the Marshal flaw distribution (discussed in Section 2.6.2.3). The ferritic weld metal flaws are analyzed as surface flaws and are assumed to grow according to the LAS SCC growth rate discussed above. SCC growth occurs for these flaws after initiation of SCC in the clad. The initiation time for SCC in the clad is dependent upon the cladding stress according to BWRVIP's cladding SCC initiation law: (initiation time in hours) = (84.2 x 10"') x (stress in ksi)*5. This cladding SCC initiation law was based on five test data. The flaws at the clad / steel interface which ere in the LAS are conservatively assumed to be surface cracks, do not include the clad thickness, and are assumed to grow according to the LAS SCC growth rate discussed above.
In the staff's analyses, the depth of 6ach flaw in the LAS was increased by the cladding thickness without any consideration of clad crack initiation time. Although this is conservative relative to the BWRVIP-05 report's treatment of clad crack initiation time as a random variable, the staff concluded that this conservatism is appropriate since no operational data exists upon which to base such an assumption, and such data won't become available for circumferential welds if they are never inspected. The staff's model did not assume any further growth of the original fabrication crack in the LAS.
2.4 Discussion of Limiting BWR RPV Transient The BWRVIP-05 report discusses the operating characteristics of a BWR with respect to design transients and their effects on the vessel. These transients generally occur when a large steam region exists (i.e., a steam bubble is present). According to the BWRVIP 05 report, the most limiting operational transients with respect to the vessel are loss of feedwater or single safety relief valve (SRV) blowdown events for normal and upset conditions. Normal operating temperature and pressure for a BWR RPV is 260'C (500'F) and 6.9 MPa (1000 psig). During the hydrostatic test, RPV temperature and pressure is approximately 65.5 to 93.3*C (150 to 200*F) and 6.9 MPa (1000 psig) and is maintained on the pressure temperature (P-T) curve for
+
that particular vessel. The most limiting design transients for emergency and faulted operating conditions are any transients which cause or result in a rapid cooldown, and rapid depressurization, of the vessel. These transients are limiting for preexisting cracks in the vessel shell welds, in comparison, the BWRVIP-05 report states that the water solid leak test condition, or hydrostatic test, is limiting for small flaws in RPV ID. In response to the staff RAI and evaluation of beyond design-basis accident events, the BWRVIP concluded that the only contribution to a potential low-temperature over pressure (LTOP) event comes from continued operation of the CRD pump when the vesselis isolated. The staff notes that the BWRVIP conclusion is similar to the foreign reactor event and is consistent with the staff evaluation.
The BWRVIP-05 report was initially limited to design basis accident (DBA) events. To provide a broader risk-informed assessment, the staff performed a sampling review of approximately 17 years of licensee event reports (LERs) and event notifications (ens) to determine if other events (i.e., shutdown events) could be potentially more limiting to the vessel (see Appendix C). The staff's sampling review, discussed in Appendix C.2, did not encompass all LERs and ens. The staff notes that some over pressure events during shutdown, as de",cribed in Appendix C.1, may 9
3 e
not be reportable under 10 CFR 50.72. In response to the staff's RAI (Reference 12), the BWRVIP also performat a data search of LERs, INPO notices and personal experience of BWR system experts. In their December 18,1997 RAI response (Reference 7), the BWRVIP stated that most of the events that they identified were covered in the staff's ISA (Reference 12), with the exception of condensate system events. The staff notes that in order for condensate system events during cold shutdown to occur, operators would have to manually initiate the condensate pump and allow it to run until the vessel was filied since there is no automatic high level trip of the condensate pump at most plants. Based on this information, the staff believes that most of the potential precursors to cold over pressure events have been identified and that further review of LERs and ens would not result in significant findings.
Table C-1 includes the results of the staff's sampling review of LERs and ens and provides the maximum pressure and temperature reached at the end of the transient, where known. Where the information was not provided, an NP is listed. Although the transients discussed in Appendix C are examples of the types of events that could result in cold over pressure events, these events occurred at temperatures that are high relative to the reference temperature of the vessel weld material, as required by plant Technical Specifications (TS). Therefore, these events did not, in and of themselves, represent significant challenges to the RPV. This is more thoroughly discussed in Appendix C.
The staff's review also identified an actual low temperature over pressure event that occurred at a foreign plant. A more deta; led discussion of the staff's independent assessment of the potential for beyond design basis challenges to the RPV is presented in Section 2.6.1. The staff notes that the BWRVIP beyond design basis evaluation was consistent with the staff's independent evaluation.
2.4.1 Inadvertent injection During Shutdown As discussed in Appendix C.1 and shown in Table C-1, severalinadvertent injections, including both high pressure and low pressure injections, have occurred during cold shutdown in U.S.
BWRs. An uncontrolled injection of feedwater or High Pressure Core Spray (HPCS) during shutdown could potentially result in a water solid condition in the reactor and inadvertent repressurization of the RPV could occur with the potential to exceed the P-T curve limits. One
+
feedwater injection event and one HPCS injection event were identified in the staff LER search.
in addition, one HPCI injection event during an HPCI automatic initiation surveillance test at cold shutdown was reported. Once an inadvertent injection occurs during a cold shutdown, the last barrier to prevent such an event from leading to cold over pressurization and tnus challenging the vessel integrity is the operator / system recovery action; however, none of these events resulted in a violation of TS P-T lim lts.
The LER search identified a total of 14 occurrences of inadvertent low pressure injections during shutdown via Low Pressure Coolant Injection (LPCI) or Core Spray. According to Appendix C.1, l
these injection modes can also repressurize the RPV while maintaining or lowering reactor i
vessel temperature. The staff notes that low pressure systems have relatively low shutoff heads which will stop injecting once the shutoff head is reached (approximately 2.3 MPa or 340 psig).
Additionally, the shutoff head for low pressura systems is considerably lower than the RPV pressure assumed for cold over pressurization events (greater than 6.9 MPa or 1000 psig).
Therefore, low pressure systems do not represent a significant challenge to the RPV in and of thamselves. However, low head systems are characterized by high volumetric flow rates which 10 l
3 can quickly increase the water level of the RPV and will effect the time available for operator / system recovery actions. This could be significant if high head systems were also initiated, where the combination of both systems could challenge vessel integrity.
2.4.2 Hydrostatic Tests or Vessel Leak Tests A vessel hydrostatic test is performed about once every ten years and a vessel leak test is performed once every refueling cycle. During a water solid leak test or a hydrostatic test, the RPV temperature and pressure are maintained at approximately 65.5'C to 93.3*C (150'F to 200'F) and 6.9 MPa (1000 psig), respectively, and are within the P-T curve limits. These' temperature and pressure conditions are required by the plant TS with the specific limits being dependent upon the level of neutron embrittlement of the RPV.
2.4.3 Injection During Maintenance Activities A foreign BWR experienced a cold over pressurization event during shutdown when a series of operator errors of commission resulted in the control rod drive (CRD) pump continuing to run until the vessel went water solid with no outflow from the reactor. Unlike the above precursor events, this event challenged the reactor vessel, which resulted in a conditional vessel challenge probability of 1.0.
This cold (26'C to 31'O or 7g'F to 88'F) over pressure (7.9 MPa or 1150 psig) transient was not included as a design basis event for BWRs and was not originally considered in the BWRVIP-05 report, which was focused only on design basis events. However, the recognition of this transient led the staff to determine that cold over pressure transients are of sufficient safety significance to need to be considered.
2.5 Evaluation of Probabilistic Fracture Mechanics Models The evaluation of BWRVIP's probabilistic fracture mechanics models was discussed in Section 6.0 of the staff's ISA (Reference 12), which discussed the structure and underlying methodology used in the BWRVIP probabilistic fracture mechanics code, the BWRVIP's use of importance l
sampling, and the BWRVIP's input parameters used for determining the probability of vessel
+
l failure. Section 2.6.2 compares the input and results of the BWRVIP probabilistic fracture mechanics analyses to the input and results of the staff probabilistic fracture mechanics analyses.
2.6 Integrated Probabilistic Assessment l
An integrated probabilistic assessment of RPV integrity requires an evaluation to determine the frequency of events that challenge the RPV and the determination of the conditional failure probability, P(FIE), of a crack penetrating through the RPV pressure boundary as a result of the limiting event. The evaluation of the frequency of events that challenge the integrity of BWR RPVs is contained in Section 2.6.1 and the evaluation of the P(FIE) of a crack penetrating i
l through the RPV pressure boundary as a result of the limiting event is contained in Section l
2.6.2. The failure probability of the RPV is the product of the frequency of the limiting event and l
the P(FIE) of a crack penetrating through the RPV pressure boundary as a result of the limiting event.
l 11 l
2.6.1 Frequency Estimation of Cold over Pressurizing the RPV As noted in Section 2.4 and in Appendix C, the staff performed an independent review of licensee event reports and event notifications (ens) to determine if events other than those described in the BWRVIP-05 report could be potentially more limiting to the vessel. Injection of cold water into the RPV at pressure or inappropriate pressurization of the cold vessel can lead to a cold over pressurization event (a beyond-DBA event). The fracture concem arises if the temperature is sufficiently low (resulting in lower fracture toughness) and the associated I
pressure is high (creating a crack driving force). The staff review identified 35 potential precursors to cold over pressure events at U.S. BWRs and one actual cold over pressure event I
which occurred at a foreign BWR. Accounting for these precursor and actual events, the staff estimated a frequency of cold over pressurization events that could challenge the RPV integrity at cold shutdown. In addition to this staff independent analysis, an RAI was issued for EWRVIP to conduct its own probabilistic analysis to estimate the BWR cold over pressurization frequency.
In response, BWRVIP submitted its probabilistic analysis on December 18,1997. The results of this analysis are discussed in the subsequent sections.
2.6.1.1 Frequency Estimation of Cold Over Pressurization Events from inadvertent injections in its December 18,1997, RAI response (Reference 7), BWRVIP reported that although water j
can be supplied to the pressure vessel via several paths (e.g., feedwater, condensate, control rod drive [CRD), standby liquid control [SLC), reactor core isolation cooling (RCIC), High Pressure Coolant injection [HPCl], High Pressure Core Spray [HPCS], Low Pressure Coolant Injection [LPCI), and/or Low Pressure Core Spray [LPCS) pumps), not all of these injection systems can be considered as a possible initiator of a cold over pressure condition. For example, some low pressure injection systems can only operate under low pressure, and others have automatic trips. Specifically, the BWRVIP stated that most of the above systems do not j
contribute to the potential for cold over pressurization events because:
j (1) the system shutoff head is low for several systems (e.g., the LPCS/LPCI pumps) so that the vessel remains within the acceptable limits of the pressure temperature (P-T) curves even at shutdown temperatures (i.e., these systems can be activated only under low pressure),
4 (2) overfilling and pressurization to the shutoff head is very unlikely because of automatic trip of the system on high water level (e.g., HPCS),
(3) the system is steam driven and is not in use during cold shutdown conditions (e.g.,
RCIC and HPCI), and (4) operation of the system, such as standbyliquid control, requires a series of deliberate operator actions like manual pump activation and is unlikely to happen without l
adequate monitoring.
Thus, RCIC, HPCI, feedwater, HPCS, LPCl, LPCS, and the SLC systems were considered to have a negligible impact on the risk of a cold over pressure event.
As discussed in Appendix C.1 and shown in Table C-1, several inadvertent injections during cold shutdown have occurred in U.S. BWRs. Of the 35 potential precursors identified in the staff j
12
I review,17 were actual injection events (of which 3 were high pressure injections and 14 were low pressure injections). One feedwater injection event and one HPCS injection event were identified in the staff LER search. In addition, one HPCl injection event during a HPCI automatic initiation surveillance test at cold shutdown was reported. The remaining 14 injections consisted of inadvertent low pressure injections during shutdown via Low Pressure Coolant injection (LPCI) or Core Spray. Although these 17 inadvertent injections were not reported to have violated the P-T curve, the staff considers that these injections could have repressurized the RPV while maintaining or lowering reactor vessel temperature.
f However, as discussed in Section 2.4.1, the staff considers that low pressure systems do,not represent a significant challenge to the RPV in and of themselves since their shutoff head is considerably lower than the RPV pressure assumed for cold overpressurization events.
Therefore, the staff only accounted for the high pressure injections in estimating the cold overpressurization frequency from inadvertent injections. Thus, based on 3 inadvertent high pressure injections during cold shutdown conditions in approximately 750 years of BWR operating experience in the U.S., the staff estimated an inadvertent injection frequency at cold shutdown to be 0.004/yr (4 x 10'8/yr).
Once an inadvertent injection occurs during a cold shutdown, the barrier to prevent such an event from leading to cold overpressurization and thus challenging the vessel integrity is the vessel condition and the operato9 system recovery action. To estimate the probability of non-recovery, the above 3 precursor events were reviewed to note reactor conditions (e.g.,
l moderator temperature), operatar actions (e.g., operator stopping injection), and/or system recovery actions which prevented these events from propagating to a cold overpressurization condition. Due to the complexity of incorporating these and other factors such as Mtrumentation failure into estimating a non recovery probability, a conservative value of 0.015 was assumed, as suggested in NUREG/CR-1278 to depict a human error probability reflecting moderately high stress condition for dynamic interplay between the operator and system indications. Accounting foi the estimated inadvertent injection frequency at cold shutdown and the non-recovery probability, the frequency of cold overpressurization of the vessel by inadvertent injections was estimated to be about 6 x 10 Wear for a U.S. BWR.
2.6.1.2 Frequency Estimation of Cold Over Pressurization Events from Condensate injection The condensate and condensate booster pumps are not tripped on high reactor water level.
Their nomd operation during startup and shutdown would be in conjunction with the feedwater pump with the excess water being retumed to the condenser via the feedwater minimum flow line. Makeup water is supplied by the condensate system until reactor pressure necessitates the start of the feedwater pumps. Starting a condensate pump while the plant is in cold shutdown requires manual action as opposed to the automatic initiation possible with the ECCS pumps.
The pump start would then have to be subsequently ignored for the vessel to be filled. An uncontrolled condensate pump injection would be evident on the water levelindicators. The BWRVIP estimated that the condensate injection frequencyW was about 0.01/yr. The non-t M Based on INPO event database and NRC Information Notices, the BWRVIP identified 5 condensate injections (3 at Nine Mile Point 2 and 2 at WNP 2)in an estimated 516 reactor-years of BWR operation.
13
recovery probability selected was 0.015 ). Thus, the frequency of cold over pressurization due 5
to condensate injection was estimated by BWRVIP to be about 1.5 x 1G"/yr.
2.6.1.3 Frequency Estimation of Cold Over Pressurization Events from CRD Injection A vessel hydrostatic test is performed about once every ten years and a vessel leak test is l
performed once every refueling cycle. The CRD pumps are the normal means to pressurize the reactor system during these tests. As discussed in Appendix C, during a water solid leak test or a hydrostatic test, the RPV temperature and pressure are maintained at approximately 65.5'C to 93.3*C (150'F to 200'F) and 6.9 MPa (1000 Psig), respectively, and are within the pressore-temperature (P-T) curve limits. These temperature and pressure conditions are required by the plant TS; the specific limits are dependent upon the level of neutron embrittlement of the RPV.
BWRVIP response to staff RAl reported that when the combination of strict temperature limitations and oversight maintained during pressure testing is considered, the risk of cold over pressure during this controlled event is small. The BWRVIP also reported, however, that the risk of cold over pressurization due to CRD injection may be higherif a loss of station power occurs during the pressure test. If a loss of station power occurs during the pressure test, the recirculation, the RWCU and CRD pumps willlose their power. If the operator restarts the CRD pumps but does not restore the RWCU, cold CRD flow will accumulate in the lower head region and, without further operator action, the pressure will increase to the safety / relief valve setpoint of about 1200 psig. Without forced circulation, the CRD flow would fill the lower head in abo'Jt two hours but not reach the beltline region until several hours after the trip. The beltline regbn would then stay near the original 200-degree level thus maintaining the beltline P-T limits. The probability of a loss of AC during a hydrostatic leak test with the operator falling to restore RWCU suction from the lower drain was estimated to be 1 x 10'8. Assumirig that a leak test is performed every 1.5 years"), the overall cold over pressurization frequency from these pressure tests was reported to be about 7 x 10d/yr.
This estimate is similar to the result obtained in the staff's independent analysis for estimating cold over pressurization frequency contribution from CRD. Although the basis of 0.001 for non-recovery probability was not provided by BWRVIP, the staff believes that the assumption of this probability is not unreasonable since it incorporates both the probability of loss of AC during the leak test and operator failing to restore RWCU. The staff estimated the uncertainty range of this result as follows. Assuming a log normal distribution of the frequency with an error factor of 10, the median sequence frequency was estimated at 2 x 10d/yr, the 5th percentile value at 2 x 10'8/yr, and the 95th percentile value at 2 x 10'*/yr.
- Handbook of Human Reliability Analysis With Emphasis on Nuclear Power Plant Applications,' described
- moderately high level of stress in a reliability analysis to transients that involve shutdown of i
the reactor and turbine to certain tasks during startup and shutdown which must be pedormed within time constraints..? The human error probability of o.015 was suggested as a point estimate for dynamic tasks invoMng Interplay between the operator and system indications.
l l
- A refueling outage is assumed to occur once per 18 months. A hydrostatic test of RPVis performed once every l
10 years. However, for the purpose of this analysis, it is assumed to replace the RPV leak test for the year that hydrostatic test is performed.
14 t
2.6.1.4 FrIqu:ncy Estimation of Cold Over Pr!ssurization Ev:nts from Loss of RWCU BWRVIP's response to the staff RAI reported that a loss of decay heat removal with an unvented reactor vessel could also lead to pressurization. If the initiating event is a scram during startup, shutdown, or from rated conditions, the trends would be the same but the initial margins from the P-T limits would be larger. Maintenance of the P-T limits during the l
subsequent cooldown of the vessel would be dependent on operator action regarding pressure control and system restoration.
The BWRVIP reported that BWR-4 plants tend to experience momentary water level drops after turbine trips from high power that can cause trips of both the recirculation pumps and the RWCU system. These must then be restored per standard procedures while maintaining proper temperature relationships between the lower head, the core exit coolant, and the recirculation loops. Assuming that a major trip causing loss of RWCU and of the recirculation pumps could occur once per year, that the probability of failure to restore RWCU is 0.01, and that the probability that the subsequent recovery is not done correctly (i.e., the lower f'ead and later the beltline exceed P T limits) is 0.01, BWRVIP estimated an overall cold over pressure frequency d
due to RWCU to be about 1 x 10 /yr.
For BWR plants other than BWR-4s, the BWRVIP reported that the probability of losing both the RWCU and the recirculation pumps would be lower (1 in 3 years) since they have milder short term level changes. Assuming the same subsequent non-recovery probabilities, the cold over pressurization frequency for BWRs other than BWR-4s due to RWCU was estimated to be 4
about 3 x 10 /yr.
The staff notes that the bases of these assumed probabilities were not provided in the BWRVIP submittal. However, the staff believes that the assumption of these values is reasonable since m
similar values were assumed in a previous NRC study,
2.6.1.5 An Actual Cold Over Pressurization Event To date only one actual cold over pressurization event has been reported. A foreign BWR experienced a cold over pressurization event during shutdown. As described in Appendix C.2, a series of operator errors resulted in the Control Rod Drive pump continuing to run until the vessel went water solid with no outflow from the reactor. Unlike the above precursor events, this event challenged the reactor vessel, or resulted in a conditional vessel challenge probability of 1.0.
If a total of about 1700 yeam of worldwide BWR operation is assumed to be from the same population (i.e., no differences between operator training, plant management, maintenance, operations, and hardware), it can be assumed that the frequency of cold over pressurizing the reactor from this class of event is about 6 x 10d/yr.
f4 in NUREG/CR-6143," Grand Gulf Low Power and Shutdown Study," a probability of 0.011 was used to depict an operator failing to control the vessel level using CRD and RWCU.
15
i l
2.6.1.6 Estimation of A Total Cold Over Pressurization Frequency The BWRVIP estimated the total cold over pressurization frequency to be about 9.5 x 10"/yr for BWR-4 plants (about 9 x 10 /yr for other BWR plants). This total frequency estimate is d
comprised of 1.5 x 10" /yr from condensate injection, 7 x 10 d/yr from CRD injection (vessel pressure testing), and 1 x 10 /yr from loss of RWCU (3 x 10/yr from RWCU of BWRs other d
than BWR-4). The staff considers the frequency contribution from these systems to be reasonable.
The staff noted, however, that BWRVIP considered the potential contribution from other injection sources (i.e., RCIC, HPCl, feedwater, HPCI, HPCS, LPCI, LPCS, and SLC systems) to have a negligible impact on the risk of a cold overpressurization event. The staff, in general, agrees with the BWRVIP's bases for considering these injection systems to be less likely to cause overpressurization conditions than other systems accounted for in this evaluation. However, because historical data shows that actual inadvertent injections of these systems have occurred, the staff does not consider their contribution to the cold overpressurization frequency to be negligible, Therefore, the staff believes that estimated frequency of cold overpressurization due to inadvertent injection of these high head systems,6 x 10 /yr, shculd be added to BWRVIP's 5
estimate of about 9 x 10'/yr. This results in the total frequency to be approximately 1 x 108/yr.
2.6.1.7 Conclusion The BWRVIP estimated the total frequency of a cold overpressurization to be about 9 x 10 /yr.
d I
Accounting for actual injections which were not included the BWRVIP analysis, the staff estimated the total frequency to be approximately 1 x 10^*/yr (a point estimate). This represents l
a theoretical estimation of cold overpressurization frequency which is based on numerous assumptions about injection frequencies, vessel pressure test frequencies, and non-recovery (or non-mitigation) probabilities, in addition to this theoretical estimate, an actual cold overpressurization frequency based on one reported cold overpressurization event which occurred at a foreign BWR was estimated to be about 6 x 10 /yr.
d The staff notes that a significant uncertainty range may encompass these frequency estimates i
and thus these estimates should be treated as an approximation not a precise estimate.
j Considering the uncertainty associated with the above estimates, the use of 1 x 10'8/yr as an approximation for the total cold over pressurization frequency for all practical purposes may not be inappropriate, it is also noted that the two results (theoretical and " actual") represent different types of scenarios leading to a cold over pressurization condition. Because of the unique characteristics of the single foreign event (as a potential outlier), it would be premature to conclude that the theoretical estimation of cold over pressurization frequency realistically represents frequency estimate from worldwide operating data. However,if the comparison is made within the general context or the general class of cold over pressurization events, the theoretical estimate appears to be consistent with an estimate derived from the worldwide operating data.
16 l
i 2.6.2 Conditionti Probability cf V:ssil Fcilura 2.6.2.1 General Discussion of Conventional Vessel Analysis Codes The conditional probability of vessel failure, P(FlE), or the probability of vessel failure assuming that the event occurred, can be calculated using conventional vessel analysis codes, such as VISA-Il and FAVOR. These codes are based on probabilistic fracture mechanics (PFM) methodology that perform millions of deterministic vessel simulations using randomly selected l
values for the variables to determine the P(FIE) for a vessel subjected to a specific transient.
)
The vessel P(FIE) is the ratio of the number of failed vessels to the number of simulations. For l
l l
each simulation, the random variables (e.g., crack size, copper, nickel, and fluence) are '
assigned according to prescribed distributions with the form and parameters of the distributions specified by the user. Deterministic fracture mechanics analyses are then performed, and the vessel P(FIE)is determined.
2.6.2.2 Discussion of Limiting Transients The initial BWRVIP 05 report was limited to design basis accident (DBA) events. In an effort to provide a broader risk-informed assessment, the staff identified, among other things, an actual low-temperature over pressure event that occurred at a foreign plant, and had performed considerable P(FIE) evaluations using this transient. The results of this effort were reported in the ISA (Reference 12). At that time, the staff also requested the BWRVIP to conduct its own identification of beyond-DBA events and to assess the P(FIE) due to the limiting beyond DBA event. In responding to the staff's request, the BWRVIP identified the loss of AC power during a post-outage primary system pressure test as the limiting LTOP event and reported the plant-specific P(FIE) for all participating plants under this transient in Reference 7. This LTOP transient has a constant pressure of 1200 psi and a constant temperature of 100'F, which is somewhat less severe than the foreign transient, identified by the NRC staff, which had a constant pressure of 1150 psi and a constant temperature of 88'F.
2.6.2.3 Technicalinput to the PFM Evaluations The BWRVIP used the VIPER code for evaluations of P(FIE). The transient used in these evaluations had a constant pressure of 1200 psi and a constant temperature of 100'F.
All P(FIE) results reported by the staff in the ISA and this SER were from PFM analyses of RPVs using the FAVOR code, which was developed by Oak Ridge National Lat, oratory (ORNL).
Except for the parametric study reported in Appendix B, the transient used is the foreign transient, which had a constant pressure of 1150 psig and a temperature of 88'F. The parametric study reported in Appendix B was performed using the constant pressure and temperature foreign transient and other constant pressures and constant temperatures.
In their PFM evalua6ns, the BWRVIP used the "BWRVIP-Marsha!!" distribution with a flaw density of 30 flaws /m' to simulate original fabrication flaws. The number of flaws per vessel was sampled as a Poisson distribution with a menn value of 3.52. Growth of these assumed original fabrication defects was assurrM to occur by stress corrosion cracking (SCC) in the low alloy steel weldment after initiation of SCC in the clad.
17
The staff used the "PVRUF-Marshall" flaw distribution in the ISA and the "PVRUF-Exponential" flaw distribution in this SER. Both distributions are defined and discussed in Section 2.6.2.4..
The staff employed a flaw density of 995 flaws /m'(108 flaws / vessel) for the best estimate and 1143 flaws /m' (124 flaws / vessel) for the upper bound flaw distributions. The evaluation of input parameters for the BWRVIP plant-specific assessments is discussed in Section 2.6.2.7 and the i
evaluation of the BWRVIP flaw density and distribution is discussed in Section 2.6.2.4 and i
Appendix D.
l l
For each set of vessel conditions (i.e., chemistry, fluence, etc.), the staff values of P(FIE) were i
determined through simulation of a max: mum of 10 vessels, with a convergence criteria of 5 7
i percent. This convergence criteria is met faster (i.e., with fewer vessel simulations) as the l
P(FIE) increases. In general, P(FIE) must be greater than about 1.5 x 10 for the convergence d
criteria to be met within the 10 simulations.
7 2.6.2.4 Evaluation of Flaw Size Distribution j
The flaw size distribution that was used in Section 7.2.4 of the staff's ISA was based on the PVRUF data and on the assumption that, for flaws greater than 0.0787 inches (2 mm), the distribution follows a Marshall distribution with credit for preservice inspection. This flaw distribution will be referred to as "PVRUF-Marshall" throughout this SER.
To better characterize the PVRUF data the staff developed another flaw size distribution based on an exponential fit without using the Marshall distribution. This revised distribution will be referred to as "PVRUF-Exponential." This was in response to issues raised by the ACRS which indicated that additional effort was needed to address uncertainties associated with the BWRVIP-05 analyses such as showing that flaw size distribution input models are justified and l
consistent with available data including those obtained in past inspections of welds Keeping all other input variables the same, the PVRUF-Exponential distribution yields higher P(FIE) values than the PVRUF-Marshal since this revised distribution has higher probability of larger flaws.
The derivation of the PVRUF Exponential distribution is presented in Appendix A.
The staff considers the PVRUF data to be the best source available for determining flaw size distributions and density in RPV welds because the inspection technique provides for better resolution of flaws than techniques used during inservice inspection of RPV welds in operating plants. The PVRUF Exponential distribution is better than the PVRUF-Marshall distribution because the PVRUF Exponential distribution provides a more accurate upper bound distribution l
as explained in Appendix A. Hence, the staff considers the PVRUF Exponential flaw distribution and the flaw density derived from analysis of the PVRUF inspection data to be the appropriate distribution and density to be used in the evaluation of the effect of cold over pressurization events on the probability of failure of RPV welds.
The BWRVIP reported that the ratio of P(FIE) using the BWRVIP-Marshall distribution to P(FIE) using the PVRUF Marshal distribution was 2.6 and 5.2 for axial welds for two plant-specific cases. The BWRVIP comparison of the P(FIE) from the BWRVIP-Marshal and PVRUF-Marshal distributions were performed with a flaw density of 30 flaws /m'. However, the BWRVIP results are misleading because the staff and BWRVIP analyses were performed using different flaw densities. The BWRVIP analyses were performed using a flaw density of 30 flaws /m' while the staff analyses were performed using a flaw density of 995 flaws /m'. The staff flaw density and distribution will result in a greater number of small flaws and the BWRVIP flaw density and 18 1
i i
i distribution will r sult in a grsitar number of largs fliws. This is discuss:d in gr:Itsr d2tailin
+
Appendix D.
To simulate the BWRVIP analysis method, the staff repeated the analysis of the B&W reference 8
case using a flaw density of 30 flaws /m and a BWRViP-Marshal flaw distribution. As a result of this analysis, the ratio of P(FIE) using the BWRVIP-Marshall flaw distribution and 30 flaws /m' to P(FIE) using the staff's PVRUF-Marshall flaw distribution and 995 flaws /m* was 0.32. The staff performed similar calculations using circumferential flaws. This evaluation indicates that the corresponding ratio of P(FIE)s was 2.6. These evaluations indicate that the staff's PVRUF-Marshal flaw distribution and 995 flaws /m' flaw density produce more conservative P(FIE) for 3
axial welds and the BWRVIP's BWRVIP-Marshal flaw distribution and 30 flaws /m flaw density produces more conservative P(FIE) for circumferential flaws. This conclusion is caused by the fact that (a) the staff approach estimates smaller flaws more conservatively; and the BWRVIP approach estimates larger flaws more conservatively and (b) small flaws are more critical for axial welds; and larger flaws are more critical for circumferential welds. The staff compares the BWRVIP's flaws distribution and density to the PVRUF Exponential distribution in Section 2.6.2.7 and Appendix D.
2.6.2.5 Sensitivity to the Flaw Size Distribution As stated in Section 2.2, the probability of an isolated or a small number of fabrication defects existing in a circumferential weld can not be reliably inferred from the sampling of weld examinations as proposed in the BWRVIP-05 report. Therefore, it is extremely important to use a flaw density and size distribution that provides an adequate assurance that the probability of such defects is accounted for in the probabilistic fracture mechanics analysis. Therefore, the staff expended significant effort to establish the current state of the art flaw distribution and, as discussed in the following Section, performed sensitivity studies utilizing upper bounds on the flaw size distributions.
The P(FIE) values using the PVRUF-Exponential distributions are summarized in Table 2.6-1 for the three reference cases: B&W, CE, and CB&l, and for two crack orientations: axial and circumferential.' The corresponding restits using the PVRUF-Marshall distribution from the ISA i
(Reference 12) are also included in Table 2.6-1 for comparison. These results have been verified to be the same as those in the ISA with exception of the B&W vessel group. The P(FIE) for the B&W vessel group with the best-estimate PVRUF Marshall flaw distribution was reevaluated. Table 2.6-1 contains the appropriate P(FIE) value of 8.19 x 108 (8.19 E-3).
As indicated in Table 2.6 2, the P(FIE) resulting from use of the PVRUF-Exponential distribution is always higher than that for the PVRUF-Marshall distribution, for all vessel manufacturers, and for both the best-estimate and the upper bound distributions. For axial flaws, the increase in P(FIE) ranges from a factor of 1.3 to about 2.1. The lowest increases in P(FIE) (1.3 to 1.35) are for the cases of best-estimate flaw distribution without inclusion of inservice inspection (ISI). The other cases (upper bound with and without ISl, and best-estimate with ISI) have increases in P(FIE) that average 1.9.
Since the BWRVIP-05 report proposes to perform ISI's on 100 percent of the RPV axial shell welds, and eliminate the inspection of all but a few percent of circumferential shell welds, the staff evaluated the sensitivity to the flaw size in axial welds using the PVRUF Exponential distribution "with ISI" and the sensitivity to the flaw size in circumferential welds using the 19
f O
PVRUF-Exponential distribution "without ISI". The sensitivity is the ratio of P(FIE) for the upper-bound distribution to P(FIE) for the best-estimate distribution. Table 2.6-1 Indicates that the sensitivity to flaw size distribution for axial flaws is 1.31 for the B&W and 1.48 for the CE reference cases; the sensitivity to flaw size distribution for circumferential flaws is 5.65 for the B&W reference case. The values for the CB&l reference case for axial welds and for the CE and CB&l reference cases for circumferential welds cannot be calculated because either one or both of the P(FIE) using best-estimate and upper-bound distributions create no failures in 10-million (10') simulations, it should be noted that the sensitivity of 1.6 for the B&W and 2.1 for the CE vessel groups-reported in the ISA for axial flaws were for distributions without ISI. The corresponding values using the PVRUF Exponential distributions without ISI are 2.14 and 2.70. Hence, the PVRUF-Exponential distribution has greater variability than the PVRUF-Marshall distribution.
2.6.2.6 Sensitivity to inservice inspection The staff performed additional sensitivity studies to determine the effect that ISl would have on the P(FIE) using the PVRUF-Exponential flaw distribution. The flaw distributions with ISI was determined using the POD of flaws resulting from the PISC study. Additional discussion of the impact of the PlSC studies on flaw distribution and P(FIE) are discussed in Section 7.6 of the r
staff's ISA.
Table A 2 in Appendix A provides the best-estimate and the upper 95 percent confidence bound flaw size distributions with an assumption of ISI. The ISI adjusted PVRUF Exponential distributions were developed by Pacific Northwest National Laboratory (PNNL) using the l
PVRUF-Exponential distributions and the POD from the PlSC 11 study. The P(FIE) values using the ISI adjusted PVRUF-Exponential distributions are summarized in Table 2.6-1 for the three reference cases and the two crack orientations.
The effect of ISI on P(FIE) for axial flaws is indicated in Table 2.6-2. The columns entitled
" Ratio" are the ratio of the P(FIE) for the case "with ISl" to the P(FIE) for the case without ISI.
For both best estimate and upper bound PVRUF Exponential distributions, ISI reduces the P(FIE), by at least 40 percent in all cases. No failures in 10' simulations were observed for circumferential flaws with inclusion of ISI.
The reductions in P(FIE) with ISI is much greater for CB&l RPVs than for CE or B&W RPVs.
This is principally due to the lower RTa levels for CB&l RPVs, which results in lower P(FIE).
For lower P(FIE), the vessel failures are dominated by deeper flaw sizes, which are easier to detect during ISI. Therefore, the prevalence of deep flaws, required to cause vessel failure for low RT levels, is significantly reduced and the P(FIE) is likewise significantly reduced.
Conversely for CE and B&W welds, the critical flaw sizes are smaller, and ISI is not as likely to detect these flaws; therefore, the reductions in P(FIE) with ISI are not as great as for CB&l RPVs.
2.6.2.7 Comparison of Sensitivity Studies and Plant-Specific Analyses The staff's sensitivity studies discussed in Sections 2.6.2.3 and 2.6.2.4 are based on the mean characteristics of all of the vessels for each vessel manufacturer, in lieu of performing sensitivity analyses, the BWRVIP performed plant-specific analyses to determine the P(FIE) for each BWR 20
I Et th3 expirrtion of their licensis,32 effective full power years (EFPY) The plant-speelfic analyses were performed by the BWRVIP using a constant temperature and pressure of 100' F and 1200 psi, respectively. The results of the BWRVIP plant-specific analyses are reported in Tables B2-1 and B2-2 of Reference 7. The highest plant-specific P(FlE) for an axial and a circumferential weld calculated by the BWRVIP are 1.55 x 10-' and 1.00 x 10, respec.ively. The 4
BWRVIP utilized plant-specific values of unirradiated reference temperature, copper, nickel, and neutron fluence for the limiting circumferential and axial welds in each RPV. The staff verified that the unirradiated reference temperatures were the values reported by licensees and I
contained in the Reactor Vessel Integrity Database (RVID). The staff compared the amount of copper and nickel in each weld to the values contained in the RVID and the values reported in the CEOG report (Ref. 22) and the Framatome inspection report (Ref. 23). The staff dete'rmined that all copper and nickel values were conservative relative to the values in these databases except for two axial welds and three circumferential welds. The non-conservative chemistry data are identified in Table 2.6-3.
The staff performed plant-specific analyses for the plants with the highest adjusted reference temperatures for the three RPV fabricators, CE, CB&l and B&W. A total of 8 cases have been analyzed for the axial and the circumferential welds. These included axial and circumferential cases for each of the three vendors using the BWRVIP-05 chemistries. Additional axial and circumferential weld cases were evaluated for the CE vessel group since, in addition to the chemistry data reported by the BWRVIP, a different set of chemistry data was also reported by the CEOG report. These analyses were performed using the temperature and pressures associated with the foreign event (88'F and 1150 psi). These analyses used the BWRVIP chemistry data except for the two axial and three circumferential welds that contained non-conservative values. The copper and nickel values used by the staff for these welds are reported in Table 2.6-3 For these limiting plant specific evaluations, the same distributions were used for each of the input parameters as for the sensitivity studies. The results of the staff's plant specific analyses are reported in Tables 2.6-4 and 2.6-5. Table 2.6-4 contains the limiting plant-specific P(FIE) for the three RPV fabricators at neutron fluence values corresponding to 32 effective full power years (EFPY). Table 2.6-5 contains the limiting plant-specific P(FIE) for the three fabricators at neutron fluence values corresponding to 64 EFPY. To evaluate the. impact of operation beyond the current license term, the staff used the neutron fluence corresponding to 64 EFPY. The neutron fluence values for 64 EFPY are the plant specific values at 32 EFPY that were reported in Reference 7 multiplied by two. These analyses were suggested by the ACRS to determine whether there would be a precipitous increase in P(FlE) after the current license term.
The limiting plant specific P(FIE) calculated by the staff for each weld type and vessel fabricator is much greater than the values from the staff sensitivity studies discussed in Sections 2.6.2.3 and 2.6.2.4. The P(FIE) values for axial welds are greater than about 6 x 10, with the highest 2
value of 0.44 at 32 EFPY. In contrast to the results for axial welds, the P(FIE) for circumferential welds is much less with all values less than 8.17 x 10 at 32 EFPY.
4 Tables 2.6-4 and 2.6-5 contain two sets of entries for CE fabricated RPVs. The entries marked BWRVIP identifies the P(FIE) corresponding to the limiting plant-specific RTa for the axial and circumferential weld materials, as reported by the BWRVIP in Ref. 7. The entries marked CEOG identifies the P(FIE) corresponding to the limiting plant-specific RT, for the axial and circumferential weld materials, with the chemical compositions revised as reported by the CEOG in Ref. 22. For the circumferential welds, the limiting weld material is unchanged with the 21
revisions to the weld chemical compositions, but the chemical composition for the limiting weld changes, resulting in a greater RT for the limiting plant-specific circumferential. For the axial welds, the revisions to weld chemical compositions result ir, a change in the limiting weld, as reflected by the parallel changes in RTag and fluence. The revisions to the weld chemical composition result in greater RTa for the limiting plant specific cases, and small increases in the P(FIE).
As indicated above, the limiting plant-specific P(FIE) for each weld type is much greater than the values from the staff sensitivity studies. As illustrated in Appendix B, these large differences in P(FIE) can be directly related to the higher end of license (EOL) RT, for the limiting plant-l specific cases as opposed to those for the sensitivity study cases. The higher RTa values are J
due to the fact that the sensitivity studies utilized average mean values of the input parameters l
l whereas the limiting plant-specific cases tended to have several of the parameters at the upper l
bound for the respective parameter. For instance, the limiting plant-specific axial weld for CB&l i
vessels had the following input parameters (relative to those used in the sensitivity studies):
fluence (p + 140), copper content (p + 30), nickel content (p + 20), RTag (p + 1.50). The difference in these input parameters results in an increase in mean RT of 117'F. For the limiting plant specific axial weld for CE vessels, the following input parameters were used (relative to those used in the sensitivity studies): fluence (p + 30), copper content (p), nickel l
l content (p + 70), RTen (p + 30). The difference in these input parameters results in an increase in mean RTa f 97'F.
o For the limiting plant-specific axial weld and circumferential weld in Table 2.6-4, analysis of the vessel failures as a function of flaw size is presented in Table 2.6-6. As indicated, the greatest proportion of vessel failures for axial welds (79 percent) are contributed by the smallest flaws, I
with depth less than 13 mm (0.5 in.), and few failures (3.2 percent) are attributable to flaws with depth greater than 25 mm (1 in.). In contrast for circumferential welds, the failures are evenly t
l distributed above and below 25 mm (1 in.), with 5.1 percent occurring for flaws greater than 51 l
mm (2 in.).
Based on the results in Table 2.6-6, one issue that arises is the impact of inservice inspection i
(ISI) on P(FIE) and the distribution of vessel failures as a function of flaw depth for the limiting plant-specific axial weld case. The inclusion of ISI results in a decrease in P(FIE) of only 11 percent, from 0.44 to 0.39. As indicated in Table 2.6-6, this case with ISI has more than 95 percent of the failures occurring for flaws less than 13 mm (0.5 in.), and only 0.2 percent for flaws greater than 25 mm (1 in.). Narrowing the range even more, more than 90 percent of the failures for the ISl case and more than 70 percent of the failure for the no ISI case occur at a total flaw depth of 9 mm (0.3575 in.), or a depth in the base metal of only 4 mm (0.1575 in.),
under a 5 mm (0.2 in.) thickness of cladding. These results indicate that for ISI to be effective on axial welds the inspection method must be capable of finding small flaws that are at the
. clad / weld interface or penetrate through the clad into the weld material.
2.6.2.8 Parametric Study for Generating the Relationship Between P(FIE) and (T-RTa)
To assess the P(FIE) values for a wide range of BWR vessels, a fluence range up to 64 EFPY, and a range of cold over pressurization transients, the staff has conducted a parametric study for axial welds. The P(FIE) is presented as a function of the difference between the temperature (T) of the transient and the reference temperature (RTc) of the weld material, T-RTc. The results of this study demonstrate that the P(FIE)is well correlated with T RT for a variety of vessel 22
I cmbrittlimint lev 31s and constant timperatura conditions, provid:d thit th) sime constant pressure is used for each of the assumed transient conditions. The staff's parametric study is described in Appendix B.
2.7 BWRVIP Risk-informed Assessment The staff requested, in reference 13, that the BWRVIP provide a risk-informed assessment of the proposed change in RPV inspections based on the results of their analysis and the guidance in RG 1.174. This request included identifying how defense-in-depth and safety margins will be maintained with the incorporation of the proposed change.
RG 1.174 provides a guideline as to how defense-in-dooth and safety margins are maintained.
This guidelines states that defense-in-depth is maintained when: (1) a reasonable balance among prevention of core damage, prevention of containment failure, and consequence mitigation is preserved, (2) over-reliance on programmatic activities to compensate for weaknesses in plant design is avoided, (3) system redundancy, independence, and diversity are preserved commensurate with the expected frequency and consequences of challenges to the system (e.g., no risk outliers), (4) defenses against potential common cause failures are preserved and the potential for introduction of new common cause failure mechanisms is assessed, (5) independence of barrier is not degraded, and (6) defenses against human errors are preserved. The guidelines also states that sufficient safety margins are maintained when codes and standards or attematives approved for use by the NRC are met, and safety analysis acceptance criteria in the current licensing basis (e.g., FSAR, supporting analysis) are met, or proposed revisions provide sufficient margin to account for analysis and data uncertainty.
RG 1.174 also states that a risk assessment should be used to address the principle that proposed increases in risk, and their cumulative effect, are small and do not cause the NRC Safety Goals to be exceeded. The guidelines further states that expected change in core damage frequency (CDF) and large early release frequency (LERF) should be assessed.
in their response dated December 18,1997, the BWRVIP repcrted that defense-in-depth and existing safety margins continue to be maintained with the incorporation of eliminating circumferential shell weld inspections. This conclusion was based on their estimate of a low frequency occurrence of beyond design basis accident precursor events (e.g., events leading to cold over pressurization condition in the vessel) and low conditional probability of RPV failure.
Additionally, the BWRVIP stated that the consequence of an RPV failure is within the licensing basis. Accounting for the estimated frequency of cold over pressure events of 9 x 10 /yr (BWR-d
- 3) and the estimated conditional failure probability of circumferential shell weld in the limiting 4
plant (1 x 10 for BWR 3), the BWRVIP calculated the total failure frequency for circumferential shell weld limiting plant (for BWR-3) to be about 9 x 10*/yr. While this information is pertinent in understanding a significant portion of the items delineated above in addressing the question related to defense-in-depth and safety margins, the staff notes that the BWRVIP response did not specifically address several of the items outlined in the guideline.
The staff also notes that the risk assessmer.; provided by the BWRVIP does not estimate CDF and LERF and that the frequency of weld failure, as shown above, was the only risk metric
. estimated for this analysis. The BWRVIP stated that even if circumferential weld failure were to occur, the event would not result'in core damage nor loss of the containment leading to LERF.
BWRVIP45, however, did not provide a basis for this conclusion. However, as discussed in 23
l a l
Section 3.0, the failure frequency for RPV circumferential welds is sufficiently low to justify elimination of inservice inspection.
2.8 BWRVIP Recommended New inspection Criteria and Scope Section 9 of Reference 1 provides specific recommendations for revisions to the required inservice inspection program for BWR vessel shell welds, as modified in Reference 3. The recommended revisions also cover reinspection and scope expansion requirements.
r Regarding the scope of the inservice inspection program, the specific recommendations (as described previously in Section 1.0) include examination of 100 percent of the axial welds, and inspection of the circumferential welds only at the intersections of these welds with the axial l
welds, or approximately 2-3 percent of these welds. In addition, the BWRVIP recommends that the inspection procedures for these examinations should be qualified such that flaws relevant to vessel integrity can be reliably detected and sized, and the personnel implementing these procedures should be qualified in the use of the procedures.
2.8.1 Successive Examinations of Flaws The BWRVIP proposed attemative criteria to the ASME Code for IWB 2420, Successive Examinations, would eliminate these examinations for "non-threatening" flaws (e.g., such as embedded flaws from material manufacturing or vessel fabrication which experience negligible or no growth during the design life of the vessel), provided that the following conditions are met:
The flaw is characterized as subsurface (in accordance with Figure 9-1 in Ref.1),
l e
l e
The NDE technique and evaluation that detected and characterized the fisw as l
originating from material manufacture or vessel fabrication is documented in a flaw l
evaluation report, l
e The vessel containing the flaw is acceptable for continued service in accordance with IWB-3600 and the flaw is demonstrated acceptable for the intended service life of the vessel.
2.8.2 Additional Examinations of Flaws in lieu of the ASME Code requirements, as specified in IWB 2430, Additional Examinations the BWRVIP proposed the following altamative criteria:
if the detected flaw is characterized as subsurface, then no additional examinations are e
- required, if the flaw is not characterized as subsurface, then an engineering evaluation must be e
performed, addressing the following (at a minimum):
- A determination of the root cause of the flaw,
- An evaluation of any potential failure mechanisms,
- An evaluation of service conditions which could cause subsequent failure,
- An evaluation per IWB 3600 demonstrating that the vesselis acceptable for continued service.
24
If the flaw meets the criteria of IWB 3600 for the intended service life of the vessel, then additional examinations would be limited under the BWRVIP recommendations to those welds subject to the same root cause conditinns and failure mechanisms, up to the number of examinations required by IWB-2430(a). If the engineering evaluation concludes that there are no additional welds subject to the same root cause conditions, or if no failure mechanism exists, then no additional examinations would be required by the BWRVIP recommendations.
2.8.3 Staff Evaluation of Successive and Additional Examination Requirements for Circumferential Welds The staff agrees with the BWRVIP that the conditions specified in Sections 2.8.1 and 2.8.2 are adequate for determining if successive and/or additional examinations are necessary for circumferential welds.
2.8.4 Staff Evaluation of Successive and Additional Examination Requirements for Axial Welds in contrast for axial welds, the results are entirely different. From the results presented in Section 2.6, the total vessel failure frequency for axial welds in BWR vessels is relatively high, 4.4 x 10 /yr [(1 x 10'8/yr event frequency) x (4.4 x 10 conditional probability of failure)) for the d
4 limiting plant-specific case at 32 EFPY, assuming that all postulated flaws are located at the surface of the vessel. In addition, the critical flaw depths for axial welds are very small, with virtually all of the failures from flaws less than 0,5 inch (1.27 cm) in depth.
With the total failure frequency for surface fisws in axial welds very high, unmonitored growth of subsurface flaws could result in a compromise to vesselintegrity. Therefore, all flaws in axial welds shall be reinspected at successive intervals and to the requirements for additional examinations consistent with the ASME Code and regulatory requirements to assure vessel integrity.
25
I y
\\
c-TABLE 2.6-1 Comparison of P(FIE) Using PVRUF Marshall and PVRUF-Exponential l
Flaw Distributions for Reference Cases P(FlE)
AVG j
Fl.AW VESSEL RTM FLAW DISTRIBUTION PVRUF-PVRUF-ORIENT.
GROUP
{.F) "
MARSHALL EXPONENT.
t FLAW DIST.
FLAW DIST.
NoISI 8.19 E-3 1.11 E 2 Best Est.
W/ISI 3.12 E-3 6.54 E 3 B&W 52.3 i
NoISI 1.28 E-2 2.37 E-2 Upper Bound W/ISI 4.65 E 3 8.56 E 3 NoISI 1.52 E 3 1.97 E 3 Best Est.
W/ISI 4.44 E-4 8.51 E-4 Axial CE 34.7 NoISI 2.62 E-3 5.31 E-3 Upper Bound W/ ISI 7.48 E-4 1.26 E 3 NoISI 4.3 E 6 6.7 E-6 Best Est.
W/1S1 5.0 E-7 6 E-7 CB&l
-25.9 NoISI 3.8 E-6 1.65 E-5 EP
W/ISI NF (10')
- NF (10')
- No ISI 1.0 E 6 2.0 E-6 Best Est.
W/ISI NF (10')
W/ISI NF (10')
- Noist NF (10')
- NF (10')
- I Best Est.
W/ISI Cire.
CE 34.7 NoISI 1.4 E 6 Upper Bound W/ISI NF (10')
- No ist NF (10')
- NF (10 )
- 7 Best Est.
W/ISI CB&l
-25.9 NoISI Upper Bound No failures in the indicated number of vessel simulations.
26
esF a
5 1
8 9 6 Cei 5
1 7 7 0 n
1 1
5 2 l
o 9
1 2
7 3
8 6 3 3 8 it la a 5 3 4
2 0 R 0 0 0 0 0 6 5 4 2 6 F
5 9
6 3 1
C 1
0 1
7 5
itn 2 2 2 1
1 e
I 3
3 4
3 n S o
I E E E E 7 7 0
p h 4
6 1
6 E 1
7 x
t E i 5
5 5 2 6 F
Ni 9 8 3 6 1
(
W 1
5 0 0 7
-F 6
8 8
1 N
5 0 0 1
1 0
U 0-R P
I V
s SI 2
2 3 3 5
V 9
5 8 3 3
6 P
R u
W C
1 3
1 1
1 w
I E
E E E E E
0 0 0 0 0
o 1
7 7
5 1
N 1
3 9 3 7 6
B a
l 6
F 1
2 1
5 1
M l
O 1
6 2 1
2 a
x R
F i
1 0
4 1
2 A
io 1
3 2 5 7 F
C 3 2 2 3 7
8 6
9 8
1 2 2 2 2 1
t r
a 3 3
2 2
1 S
o R 0 0 0 0 0 R
f O
la s
s T
6 8
6 4
e l
n i
9 6 0 9 7
9 8
3 9
0 C
N s h io 2 y s
3 3 4 4
3A I
t 0 0 1
0 0
l S
)
E E E E 7 a
- 6. F 6 a ra I
0 2 n h
2 5
4 8 E lu 2Y A M it 1(
m E R 9
2 9
3 E
1 6
4 4
5 F LT u
1 4
9 Ly F W 1
8 t
3 4 4
7 N is Bi U
v BIS C 2 3 1
2 1
Ai R
le AM 0 0 0 0 0 Ti V
t s
T E s P s
e I
3 2 3 3 6'
6 e H
n S
IS E E E E E E v
C 6
6 f
0 0
t o 9 8
2 2 S
N 1
2 5
6 3
8 o E
a 0
4 9
0 I
V e
2 4
0 2 5 4
3 r H
1 L
0 1
3 I
8 1
1 2
I e
T
/
/
9 f
s m
V ld 4
9 B 4
o b
A 1
e 0
9 4
0 2 e
d e d e d u R
W 7
7 2 2 3 2 t
lu t
t a
n a
n a
n n E
2 2
t s
u u
u d S
e m
o m o
m o e N
R it B
it B
it B
O t
s s
s r
a E
r E
r E
e e
e ic C
ldt la e
c ix i
l e
n a
t t
t p
p p d N
ee ix i C A C s
s s
r i
r r
e p
e p
e p
O Wd A C B
U B U B U in O
i e
N h
t r
in le.
e sb E W E E
r s
sa u
e eF C & C C
le V
B t
W r
s c E
lu s a C
i ed B
a Vn f
a o
M N
3 1
n e
r e
n e
d o
m p
t o
e ls i
s r
o r
l lg i
i C C M P
fl
P I
2 2
3
)
V E
E E
'0 6
6 R
7 E
E 1
5 2
W 3
5 1
F 1
1
(
B 1
1 8
N
)YP 1
1 1
2 5
5 5
F F
E F
E E
E E
E E
7 E
2 A
-E 7
4 7
2 8
1 3
T 9
3 4
9 8
M.
2 1
(
S S
2 4
1 5
2 6
8 ESY 8
k8 6
0 0
4 1
5 8
LA 1
1 6
6 8
4 9
N T {.
1 3
9 7
8 9
4 9
A R
1 1
C IF I
8 C
8 )
T*(*
E F
0 0
0 5
0 2
0 3
0 0
6 2
d P
1 S
R l
T-e N
w A
g LP
.)
8 6
0 0
4 1
8 n
5 it TF 8
1 1
i 1
1 1
1 im 9
9 6
6 G
R (*
3 3
2 6
8 M
0 7
N A
l IT e
h I
4M t
I r
- 6. L o
)
f 2P
'm e
I E V
/c 5
0 9
2 0
0 1
9
. n 5
5 c
L R n
B 1
2 6
1 2
2 5
0 s e A W 0
0 0
0 0
0 0
0 n lu TB
'0 of 1
it n D
(
N la o
utr A
m u F
.t is ne r
5 o F
0 1
0 5
7 2
5 7
0 pl A
F ek 6
1 5
2 1
2 9
6 Pe sa T
C 7
3 3
4 5
7 0
9 r
S 2
2 1
1 1
1 1
1 se IVGep C
R Ov e RN WEoh f
t F
i 0
6 M
9 9
BCr 8
5 1
9 eg O
N 2
9 0
3 7
7 9
5 iibin nn 1
1 0
0 0
0 0
S 0
ddms u T
eeu L
t t nd r r U
e oodn S
6 9
3 ppei 0
5 3
u 1
8 0
1 E
C 2
2 1
2 1
1 1
3 eet am r r R
0 0
0 0
0 0
0 0
r c
nnie F
oodt O
it it n e Y
aaid mmes R
P l
r r
)
)
ha A
U M
O Ep W
Ep l
W oot w
)
)
f f n
B B
nni M
R g
E C
B E
C B
s a
ii U
G C
C yy r T, e
(
(
r r S
t tssuR ii l
mmian T
eef a hhoe WN L
CCNM A
AE IX LI F R A
O
IV R
W B
)E IF
)
(
Y P P
1 1
1 1
4 4
5 4
F F
F E
E E
E E
E E
E E
A 9
8 2
7 9
8 8
3 4
T 6
S 4
2 8
8 9
3 7
8 7
8 3
1 1
4 1
4
(
S ES N, k.
Y 0
4 1
9 2
5 6
4 A
5 2
7 8
3 8
0 9
L A
E T 6
7 1
9 1
2 7
2 N MR{
1 1
1 1
1 1
A C
IF I
M C
')
E
- F 0
0 0
5 0
2 0
3 0
0 6
2 P
T' (*
1 d
S R
le T-w N
g A
.F n
0 4
1 9
2 5
6 4
it L
)
P T
5 2
7 8
3 8
5 9
i G
R (*
8 7
4 8
1 2
3 0
im 1
1 1
1 1
1 1
l N
A e
IT h
I t
5M r
I o
- 6. L E ')
f 2P Cm e
c 0
0 8
5 0
0 2
9 c
9 EI N/
V n
R En 3
4 3
2 4
4 0
1 e
2 LB U'
0 0
1 0
0 0
1 0
AW L '0 lu f
TB F1 (
n D
o r
N t
A u
.t e
F 5
r F
0 1
0 5
7 2
5 7
on F
0 pk A
6 1
5 2
1 2
9 6
T C
7 3
3 4
5 7
0 9
Pea S
2 2
1 1
1 1
1 1
I r
e VGp R
l R Oe C
WEh N
i 0
8 5
1 9
9 BCg 6
4 t
9 0
F N
2 9
0 3
7 7
9 5
1 0
0 0
0 iinnn O
1 0
is 0
S ddu T
ee t t d L
r r 6
3 oo e U
9 0
5 3
0 1
ppin S
u 1
8 E
C 2
2 1
2 1
1 1
3 eem 0
0 0
0 r r R
0 0
0 0
r nne F
oot e
O i
it taad Y
P I
l W
oow mms
)
)
R U
,)
)
r r a A
O p
W p
B B
f f M
R E
C B
E C
B ii.
nn M
G C
C yy
(
(
T U
r tr t
S iimmn ssR T
ee a WN L
hhe A
AE I
CCM LI X
F R A
O l
)ll
TABLE 2.64 DISTRIBUTION OF VESSEL & ALLURES AS A FUNCTION OF FLAW SIZE FOR LIMITING PLANT-SPECIFIC CASES Flaw Orientation Flaw Size Range h'd (in.)
circ.*
NoISI W/ISI l
< 0.5 in.
1.5%
79.0%
95.1 %
l 0.5 - 1.0 in.
40.4 %
17.8%
4.7%
1.0 2.0 in.
53.0 %
3.2%
0.2%
> 2.0 in.
5.1%
0.0%
0.0%
l
' For B&W from Table 2.6-4.
l
3.0 CONCLUSION
S Due to differences in processes used to fabricate axial and circumferential welds and differences in processes used to fabricate the clad on axial and circumferential welds, the likelihood of fabrication and service-induced flaws in axial and circumferential welds are not the same. Further, statistical analysis indicates that the existence of an isolated unacceptable flaw in a circumferential weld cannot be confidently ruled out based on axial weld inspection results.
Therefore, based on the above conclusions, a statistical sampling approach is not an effective approach for inferring the possibility of a criticalisolated defect (e.g., a fabrication or repair flaw that was not identified). However, sampling inspections can provide useful information regarding the occurrence of wide spread service induced degradation mechanisms.
The flaw distribution used in the staff's independent probabilistic fracture analyses assumes the probability of a 0.5 to 2.0 inch (1.3 to 5.1 cm) deep flaw, which is in the range of flaw depths that contribute most significantly to circumferential weld failures, to be about 0.025. As stated above and discussed in Section 2.6.2.4, inspection data to infer the probability of such a defect are limited, based on the PVRUF examination and experience with defects in components that are designed, fabricated and operated to high quality standards, this is considered a reasonable estimate. Furthermore, when coupled with sensitivity studies and the assumption that the subject defect is assumed to be located on the inner surface and to penetrate the cladding on the RPV, this is considered a sufficiently conservative assumption for reaching a conclusion regarding the need for performing ISI, particularly when the decision regarding ISI is based on the relative failure probabilities with and without ISI, and for assessing the frequency of BWR RPV failures for the purpose of regulatory decisions.
The staff considers the PVRUF data to be the best source available for determining flaw size I
distributions and density in RPV welds because the inspection technique provides for better resolution of flaws than techniques used during inservice inspection of RPV welds in operating plants. Hence, the staff considers the flaw distribution and flaw density derived from analysis of I
30
the PVRUF inspection data to be the appropriate distribution and density to be used in ths evaluation of the effect of cold over pressurization events on the probability of failure of RPV welds.
{
The staff concluded that beyond design-basis events occurring during plant shutdown could lead to cold over pressure events that could challenge vesselintegrity. The BWRVIP-05 report initially considered design basis events, but did not consider cold over-pressure events. The subsequent industry response concluded that condensate and CRD pumps could cause conditions that could lead to cold over pressure events that could challenge vessel integrity. The BWRVIP's estimate of the frequency of over pressurization events that could challenge the RPV d
d is 9.5 x 10 /yr for BWR-4s and 9 x 10 /yr for BWRs other than BWR-4. Accounting for actual injections which were not included in the BWRVIP analysis, the staff conservatively estimates 4
that the total frequency could be as high as 1 x 10 /yr (a point estimate).
The initial industry review determined that the failure frequency of circumferential welds was 2.2 x 10*/yr. This frequency was determined using importance sampling, generic weld variables and design basis events. Subsequent analyses using " Monte Carlo" calculation methods, plant-specific weld variables and pressures and temperatures associated with cold over-pressure events, determined that the liniiting plant specific P(FIE) for circumferential welds 4
at 32 effective full power years (EFPY) were 1 x 10 from the BWRVIP's re-analysis and 8.2 x 4
10 from the staff's analysis. Combining the frequency of cold over pressure event with the P(FIE), the BWRVIP failure frequency for the limiting circumferential welds was 9.0 x 10*/yr [(9 4
4 x 10 /yr event frequency for a BWR 3) x (1.0 x 10 conditional probability of failure)]. The limiting plant specific failure frequency for circumferential welds at 32 EFPY was determined by 4
4 4
the staff to be 8.2 x 10 /yr [(1 x 10 /yr event frequency) x (8.2 x 10 conditional probability of failure)). As depicted in NUREG 1560, Vol. I, core damage frequencies (CDF) for BWR plants were reported to be approximately 10 /yr to 10 d/yr. In addition, Regulatory Guide (RG) 1.154 Indicates that PWR plants are acceptable for operation if the plant-specific analyses predict the mean frequency of through-wall crack penetration for pressurized thermal shock events is less 4
than 5 x 10 /yr. Since the failure frequencies for BWR plants are significantly below the criteria specified in RG 1.154 and the CDF of any BWR plant (i.e., elimination of circumferential weld examinations contributes less than the amount of change of LERF and CDF, as discussed in DG 1061), the failure frequency for RPV circumfereritial welds is sufficiently low to justify elimination of inservice inspection.
The P(FIE) for circumferential welds were calculated by the staff and the BWRVIP using the PVRUF-Exponential best-estimate flaw distribution and the BWRVIP Marshal flaw distribution, respectively. The staff's flaw sensitivity analyses in Appendix D indicates that using the PVRUF-Exponential upper-bound flaw distribution and increasing the flaw size in the BWRVIP Marshal flaw distribution by 50 percent has a small impact on the P(FIE). The P(FIE) from PVRUF-Exponential upper bound flaw distribution is approximately 4.5 larger than the P(FIE) from the PVRUF-Exponential best-estimate flaw distribution. The P(FIE) from the BWRVIP-Marshal distribution with flaw sizes increased by 50 percent is approximately a factor of 6.6 larger than the P(FIE) from the BWRVIP-Marshal flaw distribution. Although uncertainties exist with tegard to the flaw distribution, this sensitivity study, which significantly increases the number of larger flaws, suggests that the impact of this uncertainty is not so great as to significantly increase the P(FIE) and invalidate the above conclusions.
31
0 w*
The limiting plant specific conditional failure probabilities for axial welds s' 32 EFPY without inservice inspection were 1.6 x 10 from the BWRVIP analysis and 4.4 x 10 from the staff 4
4 analysis. Both the staff and BWRVIP analyses indicate that the P(FIE) for axial welds are relatively high. Because the susceptibility to failure for axial welds is higher than those for circumferential welds, it is important that inservice inspections of axial welds be performed and that they be made as effective as practical. However, the vulnerability of axial welds can come l
from relatively shallow flaws, and the staff has determined that current state-of-the-art insenrice inspections have a small impact on axial welds since inservice inspection reduces the P(FIE) for the limiting plant to 3.9 x 104 The BWRVIP and staff estimates for the limiting plant-specific RPV failure frequency from' axial l
weld failures at EOL fluence are 1.4 x 10 /yr and 4.4 x 10 /yr, respectively. These failure 4
d i
frequencies significantly exceed the goal of less than 10 for RPV failures. Insenrice inspection 4
is estimated by the staff to reduce the frequency of axial weld failure to about 3.9 x 10 /yr.
d Inspection methods have not been demonstrated to be capable of reducing the frequency of axial weld failure by by three orders of magnitude, and the possibility of qualifying such an NDE method by performance demonstration is considered very unlikely. Therefore, although ISI of axial welds should be performed as part of a defense-in-depth strategy, it does not appear to be a credible solution for resolving the issue of relatively high EOL RPV failure frequencies due to axial weld failure under postulated cold over pressure conditions, it should be noted from the staff sensitivity study in Appendix B that in order to achieve a P(FIE) 4 4
less than 10, resulting in a RPV failure frequency of 10 /yr for the transient assumed by the staff, the difference between the temperature of the transient (T) and the reference temperature of this weld material (RTc ) would have to be limited to T-RT
= 60*F. Based on the chemistries and fluences submitted by the BWRVIP, about 1/2 of the BWR plants could reach an adjusted RTc value of 28'F, corresponding to T of 88'F prior to the end of their current license. However, the number of licensees potentially affected depends upon a more detailed analysis that will have to be performed by the BWRVIP. As noted above, the staff has concluded that ISI of axial welds is not an effective solution for the issue. However, the staff recognizes that conservatism, other than those that may exist in the VIPER and FAVOR fracture mechanics codes, do exist in applying the results of the parametric study in Appendix B of this report on a plant specific basis. These uncertainties include assuming that the critical axial weld is located in the peak fluence location and could significantly alter the above conclusion. Since flaws are distributed throughout the weld and the EOL neutron fluence will not occur for many years, the staff has concluded that the present RPV failure frequency is substantially below that reported by the BWRVIP and is not a near term safety concem.
However, in a letter to Carl Terry, BWRVIP Chairman, dated June 8,1998, the staff has requested that the BWRVIP provide a plan for followup analyses to determine, on a more realistic basis, the potential for axial weld failures due to cold over-pressure events and appropriate technical approaches for addressing this concem, as necessary.
The P(FIE) for the staff and the BWRVIP evaluations are generally substantially different for both circumferential and axial welds. Reasons for these differences relate to differences in the transients (100'F or 38'C and 1200 psi or 8.3 MPa for the BWRVIP evaluations vs. 88'F or 31'C and 1150 psi or 7.9 MPa for the staff evaluations) and the flaw size distributions and densities.
32
4 '
J For circumfIrantill wilds, requiramants for sucesssive end Eddition11 axaminitions would be dependent on the conservative characterization of the location of the detected flaw after appropriate consideration of the uncertainties of the inservice inspection method. The BWRVIP recommendations for successive and additional examinations are acceptable.
For flaws in axial welds, the BWRVIP recommendations for successive and additional examinations are not acceptable under any circumstances, given the vulnerability to failure of axial welds due to small critical flaw depths for these welds. Present ASME Code and regulatory l
requirements regarding inservice inspection of axial welds are sufficient for assuring integrity of I
these welds.
The conclusions conceming the BWRVIP recommendations for successive examinations are not intended to imply or express endorsement or acceptance of ASME Code Case N-526.
Appendix E contains the BWRVIP and staff assessment of the impact on vessel failure frequency of operation of BWRs beyond the current license term. The staff analysis indicates t
that the failure frequency for the limiting plant could be as high as 5 x 10 /yr. Since the total failure frequency for the limiting circumferential weld could increase significantly after the current license renewal term and the type of flaws that could lead to vessel failure are age related, each plant requesting license renewal will be requested to perform a plant-specific assessment considering the chemistry of its limiting weld and the neutron fluence at the end of the license renewal term.
4.0.
IMPLEMENTATION Since the P(FIE) for axial welds could be relatively high at the expiration of the license of a BWR, the staff will pursue plant specific actions that will assess the need for additional regulatory action to assure that the frequency of BWR RPV failure remains acceptably low.
BWR licensees may request relief from the inservice inspection requirements of 10 CFR 50.55a(g) for volumetric examination of circumferential RPV welds (ASME Code Section XI, Table IWB-2500-1, Examination Category B A, item 1.11, Circumferential Shell Welds) by demonstrating: (1) at the expiration of their license, the circumferential welds satisfy the limiting conditional failure probability for circumferential welds in this evaluation and (2) they have implemented operator training and established procedures that limit the frequency of cold over pressure events to the amount specified in this report.
Examination of the circumferential welds shall be performed if axial weld examinations reveal an active, mechanistic mode of degradation exists. The timing and scope of these examinations are to be proposed by the licensee and approved by the NRC.
The staff will pursue clarification of this issue in future rulemaking.
5.0 REFERENCES
1.
"BWR Vessel and Intemals Project, BWR Reactor Pressure Vessel Shell Weld Inspection Recommendations (BWRVIP-05)," dated September 28,1995 33 l
I 6 1
2.
BWRVIP Response to NRC Requests for Additional Information on BWRVIP-05," dated June 24,1996 3.
" Request for Authorization of a Technical Altemative per 10 CFR 50.55a(a)(3)(1) for Boiling Water Reactors," dated October 29,1996 4.
' Transmittal of BWRVIP VIPER Noe," dated May 16,1997 5.
" Detailed Programming information for VIPER " dated June 4,1997 6.
" Responds to Requests for Additional Information Regarding BWRVIP Recommendations for BWR Reactor Pressure Vessel Shell Y!sid inspections," dated June 13,1997 7.
"BWRVIP Response to NRC Request for AdditionalInformation on BWRVIP-05,"(
a December 18,1997 8.
"BWRVIP Response to NRC Request for Additional Information on BWRVIP-05 dated October 10,1997," dated January 13,1998 9.
Request for Additional Information for Topical Report BWRVIP TP.105697, "BWR Vessel and Intemals Project, SWR Reactor Pressure Vessel Shell Weld Inspectivil Recommendations (BWRVIP-05)," dated April 2,1996
- 10. BWR Owners Group, the Second Request for Additional Information for Topical Report BWRVIP TR 105697,"BWR Vessel and intamals Project, BWR Reactor Pressure Vessel Shell Weld inspection Recommendations (BWRVic 05)," dated May 20,1996
- 11. third RAI dated May 20,1997
- 12. " Transmittal of NRC Staff's independent Assessment of the Solling Water Reactor Vessel and intemals Project BWRVIP-05 Report and Proprietary Request for Additional Information," dated August 14,1997
- 13. " Request for Additional Information Regarding BWRVIP-05," dated October 10,1997
- 14. Summary of July 18,1995, meeting dated July 25,1995
- 15. Summary of March 19,1996, meeting dated March 26,1996
- 16. Summary of October 15,1996, meeting dated December 10,1996
- 17. Summary of January 16,1997, dated February 13,1997
- 18. BWRVIP Request to Meet wita the Commission, dated April 18,1997
- 19. Transcript of the May 12,1997, Commission meeting
- 20. Commission's May 30,1997, Staff Requirements Memorandum (SRM M970512B) 34 L
- 21. Inform-ti:n Notice (IN) 97-63, " Status of NRC Staff's R;v::w of BWRVIP-05," dit:d August 7,1997
- 22. July 14,1997 letter from Robert O. Hardies, Chairman, C E Owners Group, regarding CE NPSD-1039 Revision 2, Best Estimate Copper and Nickel Values in CE Fabricated Reactor Vessel Welds and CE NPSD-1039 Appendix A, revision 2, CE Reactor Vessel Weld Properties Database Volumes 1 and 2 (changed pages)
- 23. Inspection Report No. 99901300/97-01 dated January 28,1998, regarding NRC Inspection of Framatome Technologies,Inc.
4 35
s O'*
Appendix A PVRUF-Exponential Distribution The PVRUF-Exponential distribution of flaw size, together with an upper uncertainty bound, is based only on the PVRUF inspection data. The PVRUF inspection data was from welds that had not been placed in service, but were subject to pre-service inspection. The data consists of 47 planar and volumetric flaws observed in 20.32 meters of welds, excluding flaws in the cladding and base metal regions. Except for the three largest flaws, th s flaw sizes were reported in bins of 2 mm width, and are presented in Table A-1. T% i VRUF data indicates 982 indications in the cladding, of which 4 are larger than 2 mm (0.08 inches). However, due to the uncertainties with measurements of flaw location pertaining to PVRUF data, no conclusive comments can be made regarding the lineup of flaws in the cladding with the defects in weld or HAZ. The staff's approach in considering cladding flaws can be found in Section 2.3.
The PVRUF-Exponential distribution is based on a flaw frequency function estimated from the data in Table A-1. The flaw frequency function, denoted by A(s), describes the expected l
number of flaws of size larger than s that occur in a unit length of vessel weld. This flaw frequency function is assumed to have an exponential shape given by A(8) = Soexp( - [ s - s,,,,) / S, )
where s,,,,n= 4 mm and So and Si are estimated from the data by maximum likelihood. To carry out the maximum likelihood estimation, the bins in Table A 1 are replaced by their midpoints.
This yields the best estimate C A(s). Maximum likelihood is also used to calculate a 95 percent upper confidence bound on A(s). The flaw frequency function and its upper confidence bound are converted to complementary cumulative distribution functions (CCDFs) for flaws larger than 4 mm by simply dividing by their values at s = 4 mm.
To derive the PVRUF-Exponential distribution and its upper bound, the CCDFs must be weighted by the probability that flaw size is larger than 4 mm. From the PVRUF data, there are a total of 148 near surface flaws, of which 9 are greater than 4 mm. Therefore, the probability of a j
flaw greater than 4 mm is estimated as 9/148 = 0.0608. From tables for confidence limits for a i
Poisson mean, a 95 percent upper confidence bound on this probability is given by 15.71/148 =
O.1062. Combining these values with the best estimate and upper bound CCDFs, respectively, yields the best estimate and upper bound for the revised flaw size distribution. Because the upper bound is based on two 95 percent upper confidence bounds, the confidence level associated with the upper bound is greater than 90 percent. The best-estimate and upper-bound PVRUF Exponential Distributions are given in Table A 2.
1 Since the basic source data is the same PVRUF vessel weld data, the flaw density remains unchanged. One benefit from the current approach is that a more accurate upper bound on the flaw size distribution can be obtained. In the PVRUF-Marshall analysis, the upper bound was calculated by assuming the Marshall distribution for flaw size conditional on flaw existence is correct and using a 95 percent upper confidence bound for the probability of a flaw greater than 2 mm. In the current PVRUF-Exponential analysis, the upper bound is calculated by combining a 95 percent upper confidence bound on the empirical replacement for the Marshall distribution j
with a 95 percent upper confidence bound for the probability of a flaw greater than 4 mm. The upper bound based only on the PVRUF data has a confidence level greater than 90 percent.
A-1 i
The difference between the PVRUF-Marshall distribution in the ISA and the PVRUF-Exponential
~
distribution is that a larger probability density has been assigned to deeper flaws for the PVRUF-1 Exponential distribution. The direct impact to the P(FIE) due to the use of the PVRUF-Exponential distribution is discussed in Sections 2.6.2.2 and 2.6.2.3.
Table A-1 Flaws in PVRUF Wolds (excluding cladding and bese metal regions)
Size (mm) 4-6 6-8 8 10 10-12 12-14 14 18 34
- Flaws 19 14 7
3 1
1 1
'1 TABLE A 2 NRC Staff Flaw Size Distributions and Density Best Estimate Distribution Upper Bound Distribution Flaw Depth, incl.
(995 flaws /m')
(1143 flaws /m')
Cladding (in.)
NoISI With IS!
NoISI With ISI 0.3575 0.939169 0.969519 0.893859 0.950541 0.3772 0.946579 0.974929 0.905889 0.959324 I
0.3969 0.953079 0.979441 0.916559 0.966730 0.4165 0.958799 0.983264 0.925709 0.972845 0.4559 0.968219 0.989201 0.940409 0.982113 0.4953 0.975489 0.993256 0.951679 0.988401 0.5543 0.983399 0.996869 0.964389 0.994197 0.6331 0.990129 0.998893 0.975679 0.997586 0.7118 0.994129 0.999538 0.983189 0.998794 0.7906 0.996509 0.999826 0.988369 0.999420 0.9874 0.999049 0.999953 0.995339 0.999768 1.184 0.999739 0.999987 0.998629 0.999932 1.381 0.999929 0.999996 0.999629 0.999982 1.775 0.999995 1.000000 0.999972 0.999999 2.169 1.000000 1.000000 1.000000 1.000000 A2
o Appendix B A Parametric Study for Generating the P(FIE) versus (T-RTa) Curves To assess the P(FIE) values for a wide range of BWR vessels, a fluence range up to 64 EFPY, and a range of cold over pressurization transients, the staff has conducted a parametric study for axial welds. The transients that were considered are the foreign plant transient (Transient 2), an assumed transient less severe than the foreign plant transient determined by setting the temperature 30'F higher (Transient 1), an assumed transient more severe than the foreign plant trar.sient determined by setting the constant temperature 30'F lower (Transient 3), and art assumed transient less severe than the foreign plant transient determinad by keeping the same temperature but lowering the pressure to 800 psi (Transient 4). For eaet. transient, FAVOR runs were executed for the reference cases of UAW, CE, and CB&l with a different set of fluence values up to 64 EFPY for each reference case.
The P(FIE) results are summarized in Table B-1 and plotted in Figure B 1 as a function of (T-RT
), where T is the temperature, and RT, is the mean reference temperature (without the margin term) of the material. Several entries in Table B-1 are represented by dashed lines.
They correspond to cases where simulations were not performed because more than five million simulations would be needed to arrive at convergent P(FIE) values and because adding a few more points in Figure B-1 would not change the P(FIE) versus (T-RTa) curves already established by the results in Table B-1.
Figure B 1 shows that the results pertaining to CB&l vessels have entirely different characteristics than those of CE and B&W vessels. First, the results for CE and B&W vessels for the first three transients can be represented by a single curve while the corresponding results for CB&l vessels cannot. Second, the curves for CB&l vessels for the first three transients have steeper slopes than the curves for CE and B&W vessels. Last, the curves for CB&l vessels for the first three transients do not merge with the curves for CE and B&W vessels.
To explain the first phenomenon, the staff reproduced the ASME K. curve in Figure B-2 and marked the respective range of (T-RTc) for CE and CB&l for Transients 2 and 3. Figure B-2 indicates that for CE vessels the range of (T-RT,e.) for Transient 2 overlaps the range of (T-RT ) for Transient 3 by about 65 percent while there is no overlapping for CB&l vessels. This means that for CE vessels, changing the transient from Transient 2 to Transient 3 will not significantly change how the fracture toughness, K., behaves under changing fluence. This is not true for CB&l vessels. Changing the transient from Transient 2 to Transient 3 changes the fracture toughness, K., curve from a region of a steep slope to a region of a much milder slope, creating much fewor failures forTransient 3 than what it would have if K, behaved the same as that for Transioni 2. The fewer failures associated with Transient 3 caused the offset of Transient 3 curve from the Transient 2 curves for CB&l vessels in Figure B-1. This qualitative analysis explains why the P(FIE) results for Transients 1,2, and 3 do not merge into a single curve for the CB&l vessels.
An examination of the P(FIE) results for the CB&l vessels in Figure B-1 revealed that the largest P(FlE) for Transient T2 is much greater than the lowest P(FIE) for Transient 3 even though they have comparable mean (T-RTa) values, or equivalently mean K. values. To explain this, one needs to refer to Table B-1 for more information. Table B-1 shows that the fiuence is 1.38 x 10 S 8
n/cm for the case having the largest P(FIE) for Transient T2 and 0.27 x 10" n/cm' for the case B-1
o having the lowest P(FlE) for Transient T3. The staff used the constant standard deviation (in terms of fraction of the mean fluence) that was derived in the ISA for each reference case in all current PFM simulations for the three vessel groups. For the CB&l vessels, the standard deviation is 0.1895. Applying this value, the staff calculated the one-sigma fluence to be 0.26 x 10 n/cm' for the mean fluence of 1.38 x 10 n/cm' and 0.051 x 10 n/cm" for the mean i
fluence of 0.27 x 10 n/cm'. Consequently, simulations corresponding to the largest P(FIE) for Transient T2 at a mean fluence et 1.38 x 10 n/cm tend to generate more failures, because the 8
influence due to fluence variability is much greater for this case. Therefore, in addition to the mean K. value, one should also resort to the " simulation-specific" K. value, which is statistically determined in each of the 10 million (10 ) deterministic vessel analyses in a single PFM c&se, to 7
explain the P(FIE) results presented in Figure B-1.
^
To explain the second phenomenon, refer to Figure B-2. Focus on the portion of K curve for CE vessels and the portion of K curve for CB&l vessels for Transient 2 only. Decreasing the same amount of (T-RTc) causes a much sharper decrease of K, for CB&l vessels than for CE vessels, incurring much more failures for CB&l vessels than that for the CE vessels. Hence, the CB&l vessels have steeper curves in Figure B-1.
The curves for CB&l vessels for the first three transients do not merge with the curves for CE and B&W vessels. This is best explained by the reason for the different P(FIE) for the data point related to CE vessels under Transient 1 (the first value under "Tran.1"in Table B-1) and the data point islated to CB&l vessels under Transient 3 (the sixth value under "Tran. 3" in Table B-
- 1) at approximately the same (T-RTc) value of 73.1 *F. These data points are marked in Figure B-1 as "CE/T1" and "CB&l/T3." Since the same (T RT ) value of 73.1 *F corresponds to an identical K. value as indicated in Figure B-2, we should examine the PFM simulations surrounding this mean (T RTa).
The staff calculated the " simulation specific"(T RTa) values around this mean (T-RT
) for both data points for two assumed cases: one obtained by varying in the non favorable direction the Cu, Ni, and fluence by one sigma and the second by varying the Cu, Ni, and fluence in the non favorable direction by two-sigma. Notice that the probability of having either of the cases assuming the Cu, Ni, and fluence values as described above in the PFM simulation is the same for CE/T1 and CB&l/T3. The calculated simulation-specific (T RTa) values are 50.4*F for
+
CE/T1 and 47.1'F for CB&l/T3 for the one-sigma case and 32.1 *F for CE/T1 and 22.3*F for CB&l/T3 for the two-sigma case. it is clear now that although the mean (T-RTc) corresponds to the same K value, PFM simulations away from this mean value tend to generate lower (T-RTc) values, or equivalently, lower K. values, and hence result in higher P(FIE) values for the CB&l/T3 case. This phenomenon is caused by the high sensitivity of the chemistry factor to changes in copper and nickel values for the chemical combination of low Cu (e.g.,0.04%) and high Ni (e.g.,0.93%) values associated with the CB&l vessels.
B-2
8 9 1
2 3 8
1 2 7 T
4 7 7 2 1
2 2 1
2 1
2 1
1 1
4 3 3 2 1
1 1
1 3
E E E E E E E E E E E E
.n a 0 8 3 3 4 0 0 4 2 4 5 4 5
S.
3 2 8 0 3 6 0 1
1 4
r T
5 1
3 7 4 2 4 1
1 2 3 6 3 2 2 1
5 4 3 2 2 2 1
2 n.
E E E E E E E E E E E a 2 8 4
2 4
5 9 5 9 3 5 8 2 2 9 4
4 5 5 0 7 9 r
T 5 3 9 3 8
2 1
1 5 8 2 Y
D U
4 3 2
1 5 4 3 2 2 1
T
. E E E E E E E E E S
na 3 7 5 3 0 9 4 4 9 C
4 9 3 3 3 9 3 2 2 rT I
R 2 2 1
1 9 9 5 1
8 T
E F
M
'8 A
9 7 5 6 3 1
6 8 3 1
5 3 8
)
p R
4 7 5 2 1
3 7 2 6 4 4 5
=
A R
{
4 6 8 2 2 1
5 7 8 1
P 1
e 1
r F
F u
F t
8 a
A 8
r 8
T 8
F= F. e 1
"Q*
6 6
5 p
S
'8 B
5 5
m
'8 e C
(
e 1 r E R 1 u 5t L N t
B
= a =d r
n AR ee e a TO r
r
- Q up ui 9 7 5 6 7 9 4 8 3 1
5 3 m
F t s S
0 3 1
8 4 2 8 8 1
9 9 0.
t a
ap r e r
0 2 4 7 3 4 4 5 6 7 8 2 et e0 T
(
1 1
1 1
L 0
1 pd p0 U
mn m8 S
a e=
e E
t i ts e
R 3
dpdr F
m n
nu a0as c
6 6 7 4 7 8 9 8 6 0 3 6 is1 se O
N0 5i s 1
2 3 7 2 4 6 3 0 1
1 2
Y 0 0 0 0 0 0 1
0 O 0 0 p1 pr R
0 p
1 0
=0 A
M 5
5 2
(
e1 1 r t
M 1 u1 n U
s
=
=ie s
S 1
0 8
eee F
s 5
4 9
r r r n
C 9
5 8
upua 1
1 r
s
,s sysT t
eie n r r rpepa 6
3 0
v h
Ni 7
9 6
e,e e,t 0
0 0
r sr e e
er e ve v c ev e n se s
6 7
e s
u 2
4 8
s e s s t
t r C
2 0
2 af os 0
0 0
eemle l
r ssss iiii 1 234 P
U l
t t t t O
E W
nnnn eeee R
C B
iiii C
B ssss G
nnnn aaaa r r r r TTTT i1 l!ll
= Q=x-u 3
o t
r B
B S
ll f
(
(
N c
2
!a i
r 0
i r
B x
x3 N C
t 9
i A
C e
m
]
g r-
=
m AY
=
a r
m t'
~
a P
X_
y/==
f 0
f T x3 M
1 6
a 3
)
t 4
S m
g C
a 4
n
)
r o
m %
2 a
T n
%=__
r E
=
=
F a
W
)
r 0 F T
n M-
=
=
s 3 (
t B
l l
u
/
T E
B s
D C
C e
(
(
TN R
=. g_
l l
g==._=
a a
R M
i i
x x
0 T A
A F
=_
P j
y) 3 s
=
a e
50
=
5
==_
P I
V 30
- 3
=
=
=
R W
m
=
=
=
B
=_
n
)
=
w W
1 0
B
)
6 B
y
=-
=
m_
=
l e
/
r E
B u
C C
i g
(
(
l l
F a
a g
m_
=
i i
x x
0 A
A 9
0 1
2 3
4 5
6 7
a 0
0 0
0 0
0 0
0 m
m
+
E E
E E
E E
E E
1 1
1 1
1 1
1 1
g6n.
.s l
l 220 200 T3 T2 m
m i
180 1
160 3
2 Kg (
-/ 140 l
} 120 A
4e r
T3 7
l c
Ky 100 1
T2 l
~
/
60 l
40 l
20 RTNDT I
I I
l l
l 0
- 80
- 40 0
40 80 120 160 200 t
(T-RTypp, *F Figure B-2. Lower Bound Kvand Ke Test Data for SA-533 Grade B Class 1, SA-508 Class 2, and SA-508 Class 3 Steels w _
o Appendix C:
Limiting Transients The BWRVIP-05 report initially discussed the operating characteristics of a BWR with respect to design transients and their effects on the vessel. These transients generally occur when a large steam region exists (i.e., a steam bubble is present). According to the BWRVIP-05 report, the most limiting operational design transients with respect to the vessel are loss of feedwater or single safety relief valve (SRV) blowdown events for normal and upset conditions. Normal operating temperature and pressure for a BWR RPV are 260'C (500*F) and 6.g MPa (1000 psig). During the hydrostatic test, RPV temperature and pressure are approximately 65.5'C to 93.3'C (150*F to 200'F) and 6.9 MPa (1000 psig) and are maintained in accordance with the i
pressure-temperature (P-T) curve for that particular vessel. The most limiting design transients that produce the highest stresses for emergency and faulted operating conditions are any transient which causes or results in a rapid cooldown and rapid depressurization of the vessel shell welds The BWRVIP-05 report initially stated that the water-solid leak test condition, or hydrostatic test, is limiting for small flaws in the inside diameter of the vessel.
Based on the types of events discussed above, the staff performed a sampling review of
)
approximately 17 years of licensee event reports (LERs) and event notifications (ens) to determine if other events (i.e., shutdown events) could be potentially more limiting to the vessel.
I The staff's sampling review does not encompass all LERs and ens. The staff also notes that some over pressure events during shutdown, as described in Appendix C.1, may not be l
reportable under 10 CFR 50.72. The staff sampling review is discussed in Appendix C.2.
Table C-1 includes the results of the staff's sampling review of LERs and ens. The table provides the maximum pressure and temperature reached at the end of the transient. Where the information was not provided, an NP is listed.
it should be noted that, although the following are examples of the types of events that could result in cold over pressure events, these events occurred at temperatures that are high relative to the reference temperature of the vessel weld material, as required by TS. Therefore, these events did not, in and of themselves, represent significant challenges to the RPV.
j The staff RAI requested that the BWRVIP provide an evaluation of the potential for Beyond DBA events, including low temperature over-pressure (LTOP) events, and describe the types of Beyond DBA events identified in the assessment. In response to the RAI, the BWRVIP concluded that several injection systems are capable of delivering lower temperature water to the reactor vessel while not resulting in a combined LTOP condition at the location of the circumferential vessel shell welds. These systems would not contribute to the potential for Beyond-DBA events because 1) the pump shutoff head is low enough that the vessel remains within the P T curves,2) overfilling and pressurization is unlikely due to automatic high level trips,3) some systems are steam driven and cannot be used during cold shutdown, and 4) i operation of the system requires a series of deliberate operator actions. The BWRVIP assessment included the potential LTOP events during hydrostatic testing, feedwater injection, j
HPCl/HPCS injection, RHR LPCI and core spray actuation, and CRD injection. The staff's and l
BWRVIP's conclusions are included in the description of rapid depressurization and shutdown events in Section C.1.
C-1
a
~
b C.1 Description of Rapid Depressurization and Shutdown Events l
The staff evaluated the following classes of transients and shutdown events with the potential to be more limiting than the operational transients discussed in BWRVIP-05. In addition, the BWRVIP evaluated the possible ways of injecting water into the vessel during cold shutdown in 4
response to the staff RAl. The systems evaluated by the BWRVIP include feedwater, condensate, CRD, Standby Liquid Control System (SLC), Reactor Core Isolation Cooling l
(RCIC), HPCl, Feedwater Coolant injection (FWCl), HPCS, LPCl, and LPCS.
C.1.1 Rapid Depressurization Rapid depressurization of the RPV during operation can occur as a result of a large break loss-of-coolant accident (LOCA), or lifting of a SRV. Most facilities have technical specifications (TSs) for the rate of cooldown, (i.e., cooldown limit of 37.8'C (100'F) per hour). These TSs are 1
in place to protect the vessel from pressurized thermal shock. Examples of known rapid depressurization events at domestic BWRs are as follows:
1.
Hope Creek 1 had an event which caused reactor pressure to decrease from 5.3 MPa to 689.5 kPa (765 psig to 100 psig), and reactor coolant temperature to decrease from 4
266.7'C to 164.4*C (512'F to 328'F)in 51 minutes on October 10,1987.
2.
Perry had an event which exceeded the cooldown rate of 37.8'C (100'F) per hour on October 27,1987.
3.
Quad Cities 1 had a stuck open relief valve at 10 percent power which caused a cooldown rate of 111.1 *C (200*F) per hour for the first hour and 63.9'C (115'F) for the second hour on April 17,1989.
C.1.2 Hydrostatic Testing and/or Vessel Leak Test As discussed in Section 2.4, the hydrostatic test (or vessel leak test) is performed in the water solid condition with RPV pressure and temperature maintained in accordance with the vessel P-T curve, as is generally specified in the licensee's TSs. With the vessel in the water solid condition, any addition of water to the vessel increases RPV pressure while maintaining temperature. An example of potential water addition to the vessel is control rod drive (CRD) pumps running with no reactor water cleanup (RWCU) letdown. This conditbn has the potential to exceed the P-T curve limits.
C.1.3 Feedwater injection and/or High Pressure Core Spray injection An uncontrolled injection of feedwater or High Pressure Core Spray (HPCS) during shutdown could potentially result in a water solid condition in the reactor. The staff notes that HPCS is applicable only to BWR 5 and 6. The HPCS pump is capable of delivering at least 97.8 Us (1550 gpm) at 7.9 MPa (1147 psig) reactor pressure,385.4 Us (6110 gpm) at 1.4 MPa (200 psig), and a maximum of 492 Us (7800 gpm) at runout conditions. HPCS normally is aligned to the condensate storage tank (CST) for suction. Depending on the temperature and pressure at the time of the injection, normally 48.8'C to 93.3'C (120'F to 200'F) and 0 MPa (0 psig), an inadvertent repressurization of the RPV could occur with the potential to exceed the P-T curve limits. The staff notes that HPCS injection during shutdown is usually avoided due to C2 l
l l
a precautionary tagouts during maintenance of the systems, in the BWRVIP RAI response, the BWRVIP stated that feedwater, FWCl, and HPCS have automatic high water level trips at Level 8 and operator or system errors would have to occur before the vessel experiences high pressure. Additionally, the feedwater pumps at several plants are steam driven and not available during cold shutdown conditions.
The rest of the BWR fleet has a High Pressure Core injection (HPCI) System. HPCI is turbine driven and is not designed to operate below 689.5 kPa (100 psig). Since the reactor is subcooled at atmospheric pressure during shutdown, inadvertent actuation of HPCl is generally not possible. However, the staff did identify one LER which describes an inadvertent HPCI injection during cold shutdown. This LER is discussed in Section C.2.3. The BWRVIP RAI response stated that RCIC is also steam driven, although some plants are able to operate RCIC during shutdown by manually connecting it to an auxiliary boiler.
C.1.4 Low Press ure Core injection and/or Core Spray injection A more common event during shutdown is a Low Pressure Core injection (LPCI) or Core Spray injection. Injections of this type can also repressurize the RPV while maintaining reactor vessel temperature. If the injection water is for a long duration, the potential exists that the injection may cause a reactor temperature decrease. Both LPCI and core spray take suction from the suppression pool as the primary source of water. An attemate alignment could be the CST. The suppression pool water temperature ranges from 18.3'C to 35'C (65'F tr 95'F), depending on the plant specific TSs and the time of the year. In general, each LPCl pump (RHR pump) is designed to produce a flow rate of 4S8 Us (7260 gpm) at 1.2 MPa (172 psig) and a maximum j
flow rate of 537.5 Us (8520 gpm) for runout conditions. For this pump, shutoff head is less than l
2.3 MPa (329 psig). For some facilities, core spray is capable of delivering 394.3 Us (6250 gpm) at 2.3 MPa (329 psig) and 482.9 Us (7655 gpm) at runout. Based on the normal shutoff heads of the low pressure pumps, LPCI and core spray do not represent a significant challenge to the RPV in and of themselves. However, these high flow systems are characterized by high i
volumetric flow rates which can quickly increase the water level of the RPV and will effect the time available for operator / system recovery. This could be significant if high head systems were also initiated, where the combination of both systems could challenge vessel integrity.
C.1.5 Unvented Condition During Shutdown l
An unvented reactor vessel during shutdown with reactor temperature below a certain value is a violation of some plants' TSs. In this condition, any addition of water to the vessel, from any source, without letdown, has the potential to repressurize the vessel. However, this generally is a violation of plant TSs, and therefore, it is assumed that the licensees make every effort to avoid violation of their TSs.
C.1.6 Standby Liquid Control System injection in their December RAI response, the BWRVIP discussed the potential for injection of water to the RPV via SLC. SLC requires manual operator action to initiate. The SLC system is capable of high pressure injections but has a very low flow rate and low capacity. The staff notes that if SLC was initiated, the operators have sufficient time to control reactor pressure due to the low flow rate.
C-3 1
l i
S l
C.1.7 Condensate injection The BWRVIP also discussed the potential for condensate to inject while a plant is in cold shutdown. The condensate system also requires manualinitiation and does not have an automatic high level trip. Once the condensate system is started, the operators would have to l
ignore the wide range and upset water level indicators in order to fill the vessel.
C.1.8 Cold Over Pressurization Condition As discussed above, there are several systems that are capable of injecting water to the F)PV j
during plant shutdown conditions. The staff and the BWRVIP agree that not all of these systems l
have the potential to cause a cold over pressurization condition. In the December RAI response l
the BWRVIP concluded that several actions would have to take place for a LTOP to occur in a BWR. These include operators violating the P-T curves, ignoring water level instrumentation, isolating the vessel, and continual water injection via CRDs for an extended period of time.
Based on this conclusion, the BWRVIP believes that these actions are extremely unlikely. The staff does not agree that a cold over pressurization condition is extremely unlikely as exhibited in the foreign reactor event discussed in Section C.2.1 and the potential precursors identified in the staff's LER and EN review. However, the staff and BWRVIP estimates of the frequency of the l
cold over pressurization are nearly equivalent (see Section 2.6.1).
C.2 Discussion of Staff LER and Event Notification Review The staff performed a cursory review of LERs and ens. The LER search involved reviewing potential BWR overcooling or over pressurization events since 1980. A total of 229 LERs met the initial se, arch criteria. The EN search consisted of the same review of notifications since 1985. Approximately 81 ens were identified which met the initial search criteria. Of the 310 events ider'tified, none of which were overlapping,35 events were singled out as potential precursors to cold over pressure events at domestic BWRs. These events are identified in Table C-1. In addition, the staff is aware of a cold over pressure event which occurred at a foreign BWR. The following paragraphs highlight the circumstances of some of the more significant events and what factors prevented a complete over pressurization and potential failure of the vessel.
C.2.1 Cold Over Pressure Event at a Foreign Reactor A cold over pressure event that occurred at a foreign reactor during shutdown appears to be more limiting than the leak test condition. The foreign reactor had nearly completed its sixth refueling outage with the reactor vessel head tensioned and RHR A running in the shutdown cooling mode. At the beginning of the event, reactor water level was approximately 15 centimeters (5.9 inches) below the vessel flange and the main steam line isolation valves were in the shutoff position. In addition, the safety relief valve (SRV), the vessel head vent pipe, and RWCU were administratively b!ocked. A weld overlay on a recirculation loop was in progress i
and required the B recirculation pump to be aligned to enhance heat removal capability. The l
CRD pump was started to establish sufficient seal injection flow for the recirculation pump. CRD flow was approximately 1.9 Us (30.1 gpm).
l l
As reactor water level gradually increased, operators aligned a RHR pump and related valves to i
letdown the excess water. Later on that day, maintenance personnel were informed of C-4 1
A Inaccurate level indication due to reference leg leakage. Almost concurrently, a reactor operator shift tumover occurred from shift 2 to shift 3. Records showed that the information regarding the inaccurate level indication was passed to shift 3. At shift tumover, the level indication was reading 700 centimeters (275.6 inches) based on the instrument zero line with a system pressure approximately 0 MPa (0 psig). Actual reactor water level was approximately 500 centimeters (196.8 inches).
This condition continued until reactor system pressure reached 917 kPa (133 psig) and, on high pressure, tripped the RHR pump. The operators were not aware that the RHR pump had tripped. The CRD pump continued to run until the v9ssel went water solid with no outflow from the reactor. Pressure continued to increase at a higher rate until reactor system pressure reached 7.9 MPa (1150 psig) and tripped the CRD pump due to low suction head. The B recirculation pump had tripped at 7.9 MPa (1150 psig) also. During this event, reactor coolant temperature was maintained around 26'C to 31'C (79'F to 88'F). This cold over pressure event resulted from operator error. The shift crew was not alert to meter readings for the system status key parameters such as pressure and water level.
C.2.2 Hydrostatic Tests and Vessel Leakage Tests at U.S. Plants On March 19,1989, the Clinton RPV was water solid for the leakage test and the mode switch was set to REFUEL. Concurrently, scram time testing was being performcd. At the completion of the scram time testing, the reactor mode switch was repositioned to SHUTDOWN which initiated an expected scram signal. This signal caused the scram inlet and exhaust valves to open which resulted in the control rod drive bypass flow being directed to the solid reactor. This in tum caused the reactor pressure to increase at a rate of approximately 48.2 kPa (7 psig) per second and resulted in 4 SRVs lifting per design. Operators attempted to reduce pressure, prior to the SRVs lifting by using the discharge paths established for the RPV pressure test but were unsuccessful. The maximum pressure recorded during the event was 7.8 MPa (1130 psig).
Temperature at the beginning of the event was 71.1 *C (160*F). No change in temperature was reported.
On May 11,1993, a system leakage test of the RPV was being performed at Browns Ferry Unit
- 2. Pressure was being maintained using a CRD pump and RWCU reject flow control valve. In parallel, a " Functional Line Flow Check Valve" Surveillance instruction (SI) was being performed. Due to isolations with both tests, operators received indications that reactor pressure was decreasing. The unit operator attempted to maintain pressure in the band prescribed by the ASME Section XI test by lowering RWCU reject flow. The resultant pressure l
increase resulted in an Anticipated Transient Without Scram (ATWS) scram signal that tripped l
the reactor recirculation system pumps and initiated an Attemate Rod injection (ARI) signal. The highest pressure encountered during the event was 7.7 MPa (1120 psig). Moderator temperature at the beginning of the event was 87.8'C (190*F). No change in temperature was reported.
l l
C.2.3 Feedwater injection and High Pressure Injection Events at U.S. Plants On January 25,1984, Peach Bottom Unit 3 was set up for long path recirculation (feedwater l
system flush to the condenser) after completion of maintenance on the reactor feed pump bypass valve. The operator failed to close the feedwater inlet valves to the RPV as required in the system procedure. With a condensate pump in service, the operator opened the 5th heater C-5 l
l
o i
~
c-outlet valve and inadvertently injected condensate into the reactor vessel. Reactor water level increased approximately six feet and a minimal pressure increase was noted on the wide range reactor pressure strip chart recorder. This pressure increase was estimated to be less than 69 kPa (10 psig). Minimum reactor vessel temperatures recorded were 42.2*C (108'F) in the A recirculation suction line and 46.1 *C (115'F) at the feedwater nozzle. This transient violated the technical specification limits of 48.9'O (120'F) and 0 kPa (0 psig) as referenced in Paragraph 3.6.A. Figure 3.6.2 of the Peach Bottom Unit 3 TSs.
On July 27,1990, Nine Mile Point Unit 1 was in an extended refuel outage with the core loaded and with the mode switch in the REFUEL position. The HPCI System initiated while preparing to l
perform HPCI Automatic Initiation Surveillance Test. The feedwater isolation valves were I
opened by procedure. Due to a wom cam in the feWwater flow control valve positioner, l
feedwater leaked into the reactor vessel. A high reactor water level 241.3 centimeters (95 inches), signal was received which initiated a turbine trip sign 31. HPCI initiated and injected into the reactor vessel for approximately 3 seconds, At the time of tus event, the reactor vessel i
water temperature was about 65.6'C (150'F).
On May 3,1995, during the performance of "LaSalle Unit 2 Reactor Ves3el Water Level Reference Leg Continuous Backfill Panel 2C11 P002 Operation," instrun.ent maintenance pers annel retumed a continuous backfill panel to service. The backfill par,si provides a low l
cont nuous flow of high pressure water to the reactor vessel instrument rr'erence leg piping to ensure that the reference column is filled to its proper height and that an/ non-condensable gases are flushed from the reference leg piping. During the valving sequence specified in the i
procedure, a pressure line spike caused the HPCS diesel generator anJ HPCS pump to initiate.
The HPCS pump started and injected into the vessel. Reactor water tavel rose approximately 20.3 centimeters (8 inches). If the reactor operator had not secured the HPCS pump, the HPCS l
system would have stopped injecting due to high level closure of ti 4 HPCS injection valve. The i
addition of suppression pool water did not reduce the reactor water temperature to below the 3
minimum bolt up temperature,30'C (86'F), or the minimum ter perature used in shutdown margin calculations,20'C (68'F). Reactor water temperature r amained above 48.9'C (120'F) l i
during the event.
C.2.4 Low Pressure Core injection and/or Core Spray inju: tion at U.S. Plants On May 15,1985, instrument and control personnel were in the process of performing the
" Channel Logic Response Time" procedure at Hatch Unit 2. Tiie procedure mistakenly called for a jumper to be placed between AA3 and AA4 in panel 2H11 P617; proper placement was between AAS and AA4 in panel 2H11 P627. RHR pump C was running in shutdown cooling mode. The improperly installed jumper initiated a loop A LOCA signal which caused RHR pumps B and D to start in the LPCI mode and inject water from the torus into the RPV. RPV level increased from approximately 91.4 to 254 centimeters (36 to 100 inches), referenced to instrument zero, as a result of the injection. The RPV pressure and temperature at the time of l
the event was not reported in the LER.
On June 9,1988, during performe.nce of Surveillance Procedure 6.3.4.3 on the emergency diesel generators, approximately va775 liters (15,000 gallons) of suppression pool water were injection ir,to the Cooper reactor vessel. The surveillance procedure is performed once per cycle to functionally test the emergency start of each emergency diesel generator, RHR pump and core spray pump, as well as the loading sequence of safety related equipment on the associated C-6 I
a diesti gInerator. Upon simuliting thz low raictor vissal wat2r lev:1, tha DC pow:rrd RHR loop B inboard injection valve began to open. When the critical bus was re energized, the RHR loop B outboard injection valve began to open and the RHR loop B outboard to suppression pool valve began to close. These valve operations, along with RHR pump D starting when the bus was re-energized, created an injection path into the reactor vessel via RHR loop B. Additionally, the core spray loop B outboard injection valve and the core spray loop B inboard injection valve began to open. When the core spray pump B started ten seconds later, an additional flow path via core spray loop B was created. The total injection time for HHR loop B was 1 minute 37 seconds, and 53 seconds for core spray. Approximately 41635 to 45420 liters (11,000 to 12,000 gallons) of water was injected by RHR and 11355 to 15140 liters (3,000 to 4,000 gallons) of water was injected by core spray to the RPV. Had the injection continued with no operator action, the reactor vessel would have completely filled and pressure would have stabilized at approximately 2.3 MPa (340 psig), shutoff head for core spray. Actual reactor vessel metal temperatures ranged from 76.7'C to 87.8'C (170'F to 190*F) in the flange area, with no vessel temperatures below 51.7'C (125'F).
On July 6,1994, instrument and control personnel were back filling instrument lines to support excess flow check valve testing at Washington Nuclear Plant Unit 2. Due to a lineup error, the differential pressure sensed by the inservice reactor vessel level detectors was increased creating an invalid low level indication which caused several automatic actions including a low pressure core spray system actuation and injection. The residual heat removal (RHR) system was in a shutdown cooling lineup, and therefore, no LPCI system injection to the RPV occurred via the RHR system. Reactor water level increased approximately 50.8 centimeters (20 inches) due to the core spray injection. The injection water temperature was 21.1'C (70'F) with the RPV at 54.4'C (130'F) and 730.9 kPa (106 psig).
Table C 1 Sampling Review of LERs and ens plant nata Tuna nf Fvant p nein T(A Shoreham 6/6/85 Hydrostatic test 1047 NP Nine Mile Point 1 6/21/84 Hydrostatic test 1068 NP Browns Ferry 2 5/11/93 Hydrostatic test 1120 190 Clinton 3/19/89 Hydrostatic test 1130 160 Peach Bottom 2 5/30/85 Hydrostatic test 1030 NP Peach Bottom 3 1/25/84 Feedwaterinjection 10 108 Nine Mile Point 1 7/27/90 HPCIinjection NP 150 LaSalle 2 5/3/95 HPCS injection 0
120 Shoreham 5/5/87 LPCl injection 125 160 WNP2 7/6/94 Core sprayinjection 106 130 Hatch 2 5/15/85 LPCI injection ND NP Limerick 1 3/26/85 LPCI inlection NP NP C-7
A e
Table C-1 Sampling Review of LERs and ens Plant nato Tuna nf 5'vom P ncin T 5)
Brunswick 2 10/21/86 Core sprayinjection NP NP Brunswick 2 5/30/86 Core spray and LPCI NP 150 injection FitzPatrick 4/7/87 Core spray injection NP NP Nine Mile Point 2 2/2/89 LPClinjection 0
Amb.
Dresden 2 2/5/89 Core spray and LPCI NP
> 100 injection Cooper 6/9/88 RHR and Core spray
< 340 170/190 injection Cooper 5/26/88 Core sprayinjection NP NP Hatch 2 11/14/92 RHR and Core spray NP NP injection l
Vermont Yankee 4/12/92 Core sprayinjection NP NP Hope Creek 3/8/92 Core spray and LPCl 0
NP injection Dresden 2 10,B/85 Unvented condition NP 133 Dresden 2 5/24/86 Unvented condition NP 130/140 Dresden 3 7/11/84 Unvented condition NP 130/140 Peach Bottom 3 9/28/81 Unvented condition 78
<120 Peach Bottom 3 10/26/89 Unvented condition NP 155 I
WNP2 7/4/94 Potential HPCS injection NP NP Nine Mile Point 1 8/3/86 Potential HPCIinjection NP NP Perry 7/11/86 Potential HPCS injection NP NP WNP2 6/18/89 Potential HPCS injection 122 NP WNP2 6/18/89 Potential HPCS injection NP NP WNP2 6/18/89 Potential HPCS injection NP NP Nine Mile Point 2 2/19/89 HPCS initiated - no 2
129 injection Nins Mile Point 2 10/8/88 HPCS initiated - no 35 104 i
injection NP: Not Provided C-8
A C.3 Oper; tor Training and Plint Procedures The staff requested that the BWRVIP describe the recovery actions that can be taken in response to inadvertent injections and estimate the recovery time. In response to the RAI, the BWRVIP stated that, besides the CRD actuation, inadvertent actuation of all other systems is either not possible at low pressure, or limited by water level trips. As such, the BWRVIP does i
not believe that these systems contribute to the potential of cold over pressurization. The BWRVIP also stated that the recovery from CRD actuation without RWCU available requires l
I manual CRD pump trip and restoration of the RWCU. If the RWCU is still not available, activating the automatic depressurization mode to open the SRVs will depressurize the RPV.
l The recovery time to trip the CRD pump and restore RWCU or depressurize through the SRVs is approximately five minutes, in addition, water level instrumentation and control room monitoring of the P-T curves provide the means to diagnose the inadvertent event.
The staff questioned whether current operator training and plant procedures provided guidance on preventing an inadvertent injection during shutdown conditions and preventing a cold over pressurization given an inadvertent injection occurred. In response, the BWRVIP stated that upon review of the foreign reactor event, the BWRVIP concluded that the cold over pressurization event would not have occurred if the plant procedures were followed. The i
BWRVIP believes that the existing TSs and plant procedures are adequate to avoid cold over pressure of the RPV. Additionally, the BWRVIP concluded that operator awareness of the RPV condition while the RPV is isolated for maintenance is covered in the current plant procedures and operator training. The staff notes that the BWRVIP did not provide any specifics on the type of training an operator receives or the plant procedures that are in place to prevent inadvertent injections and potential cold over pressurization of the RPV during shutdown conditions.
l 6
l i
I 1
i C-9
+
1 1
(
Appendix D Comparison of Flaw Density and Flaw Distributions Used in the BWRVIP Analysis Method and Staff Analysis Method (s) i As described in 2.6.2.3 and in Appendix A, the NRC staff evaluations of P(FIE) used either the PVRUF-Exponential best-estimate flaw distribution (with a flaw dens PVRUF-Exponential upper-bound flaw distribution (with a flaw density of 1143 flaws /m. The j
i best-estimate density represented 108 flaws per vessel and the upper-bound density represented 124 flaws per vessel. In the Monte Carlo flaw simulation technique to determine the P(FIE), the number of flaws per vessel is an important parameter since the greater the number 1
of haws in an individual vessel analysis the more flaws that will be simulated and the more likely that an individual vessel analysis will result in a failure of the vessel. In contrast, the results presented in BWRVIP-05 used the Marshall distribution, with a density of 30 flaws /m', or 3.52 flaws per vessel (this is the mean of a Poisson distribution). Due to tha nonlinear nature of l
vessel analyses, it is not possible to reach a conclusion a priori as to whether the PVRUF-3 Exponential flaw parameters (the PVRUF-Exponential flaw parameters are the PVRUF-Exponential flaw distribution and associated flaw density) or the BWRVIP-Marshall flaw parameters (the BWRVIP-Marshal flaw parameters are the BWRVIP-Marshal flaw distribution and the associated flaw density) will provide the more conservative evaluations of P(FIE). To i
provide the basis for such a conclusion, comparisons between the different flaw density and distributions are made, in conjunction with an evaluation of the relationship between flaw depth and failure probability. The comparisons of P(FIE) will be made based upon the limiting plant-specific parameters for circumferential flaws (from Table 2.6 4).
The number of flaws of a particular depth for the PVRUF-Exponential best-estimate and upper-bound distributions are compared to that for the BWRVIP-Marshall distribution in Table D-1, where the columns entitled "# of Flaws" represent the number of flaws of each depth per vessel simulation. The number of flaws of each depth eq' al the probability distribution of a particular u
flaw depth, Pr(a}, times the prescribed number of flaws per vessel. In evaluations of P(FIE),
NRC staff added 0.2 in. to each crack depth to simulate the thickness of the clad. Therefore, in l
I the staff's analyses the cracks were open to the inside surface of the vessel. The staff did not include stress corrosion cracking (SCC) of the ferritic weld metal since its analyses included cracking of the stainless steel cladding.
The BWRVIP analyses do not include a cladding thickness. Hence, cracks were assumed to Initiate at the clad / weld interface. However, the BWRVIP analyses included crack growth in the i
ferritic weld metal resulting from SCC. Analysis of SCC uses cladding residual stress and crack growth rate as random variables, with initiation in the ferritic weld metal SCC coincident with the predicted initiation of an SCC crack in the cladding metal. Due to the high variability of these evaluations of crack growth, there is no specific flaw size distribution that can be chosen to represent that used in the BWRVIP-05 analyses. Since the BWRVIP-Marshal flaw distribution in Table D 1 does not include flaw growth from SCC, the flaw sizes in Table D-1 represent a lower bound of flaw sizes that result from the BWRVIP analyses.
The number of flaws greater than a certain depth is plotted in Figure D-1. As illustrated, the BWRVIP-Marshall distribution has deeper flaws than either the best estimate or upper bound of the PVRUF Exponential. The upper bound PVRUF Exponential has a greater number of flaws at each depth than the BWRVIP-Marshall. The best-estimate PVRUF-Exponential distribution D-1
4 d
has more shallow flaws (flaws less than 0.964 in.) than that of the BWRVIP-Marshall distribution; but, the BWRVlP-Marshal distribution has more deeper flaws (flaws greater than 0.984 in.) than the best-estimate PVRUF-Exponential distribution.
To parametrically evaluate the effect of deeper flaws on the probabliistic fracture mechanics evaluations, each of the BWRVIP-Marshall flaw depths was increased by 15 percent and by 50 percent. The effect of this is to increase the number of flaws at larger flaw depths and reduce the number of flaws at smaller flaw sizes.
Grouping the data for various flaw size ranges is illustrated in Table D-2 and Fig. D-2. Fo,r flaws deeper than 2 in., the BWRVIP-Marshall distribution with flaw depths increased by 15 percent has two times the number of deeper flaws than the PVRUF-Exponential upper bound distribution and the BWRVIP-Marshal distribution with flaw depth increased by 50 percent has ten times the number of deeper flaws than the PVRUF Exponential upper bound distribution. In comparison to the best-estimate PVRUF-Exponential, the BWRVIP-Marshall distribution with flaw depths increased by 15 percent and 50 percent have factors of 10 and 50 times as many flaws with a depth greater than 2". To more easily describe the differences, the data in Fig. D-3 have been normalized to the greatest number of flaws within each range. In this figure it is easier to see that the PVRUF-Exponential upper bound distribution provides about ten times the number of flaws of the BWRVIP-Marshall distribution for flaws less than 1 inch in depth, but the BWRVIP-Marshall distribution with the flaw stres increased by 15 percent and 50 percent have substantially more flaws greater than 2 in. In depth.
From the data presented in these figures, it is expected that the PVRUF Exponential upper-bound distribution would provide the most conservative P(FIE) for all cases in which the critica!
flaw depths are smaller than 1 in. This would generally include pressurized thermal shock (PTS) and similar transients with high thermal stresses. The BWRVIP-Marshall distribution, and in particular those incorporating flaw sizes increased by 15 percent and 50 percent, would be expected to provide the most conservative P(FIE) for cases where the critical flaw sizes are greater than 2 in.
To assess the effect of each of these flaw parameter sets, the P(FIE) was evaluated using the limiting plant specific vessel parameters for circumferential flaws (mean RT,m f 99.8'F, see o
Table 2.6-4) and incorporating the flaw size distributions and densities for each of the flaw parameter sets. The results of these analyses are in Table D-3. Since the FAVOR code will not accept flaw sizes greater than 50 percent of the wall thickness (in this case 2.6 in.), all flaws greater than 2.6' in depth in the BWRVIP-Marshall flaw distribution were assumed equal to 2.6 in. This limitation provides a lower P(FIE) than that using the VIPER Code since only 2 percent of the simulated flaws at this depth fall and at greater flaw depths the likelihood of flaw failure could be much greater than 2 percent.
Using the limiting plant specific vessel parameters for circumferential welds and the temperature and pressures resulting from the low temperature event in the BWR foreign plant, the P(FIE) for the PVRUF-Exponential upper bound flaw distribution is approximately 4.5 larger than the P(FIE) for the PVRUF-Exponential best-estimate flaw distribution. The P(FIE) for the PVRUF-Exponential best-estimate flaw distribution was determined to be equivalent to the P(FIE) for the BWRVIP-Marshal flaw distribution. The staff also evaluated the impact of increased flaw size on the BWRVIP-Marshal flaw distribution. Increasing the BWRVIP-Marshal flaw size distribution by 15 percent doubles the P(FIE), and increasing the flaw size by 50 percent increases the P(FIE)
D2
4 4.
by an additional factor of 3.3. These increases are small, given the fact that the limiting flaw size distribution and density increase the number of flaws greater than 2 in, by a factor of 50 and increases the number of flaws in the size ranges of 1 to 2 in. by a factor of almost 3 over the PVRUF-Exponential best-estimate flaw distribution.
TABLE D-1 FLAW SIZE BINNING USED FOR EVALUATION OF P(FIE)
Flaw PVRUF Exponential
- BWRVIP Marphall
- i Depth Best Estimate Upper Bound
" Bin" 1
(in.)
CDF Pr(a}
ph CDF Pr(a}
CDF Pr(a}
[,"
0.0790
<W L # J *~.%
0.273504 0.273504 8.92 E 1 0.1575 0.939169 0.939169 1.c1 E 2 0.893859 0.893859 1.11 E 2 0.472417 0.198914 6.48 E 1 1
0.1772 0.946579 0.007410 8.00 E 1 0.905889 0.012030 1.49 E O 0.512971 0.040554 1.32 E 1 0.1969 0.953079 0.006500 7.02 E 1 0.916559 0.010670 1.32 E O 0.550408 0.037436 1.22 E 1 0.2165 0.958799 0.005720 6.18 E 1 0.925709 0.009150 1.14 E O 0.584798 0.034390 1.12 E 1 j
0.2559 0.968219 0.009420 1.02 E O 0.940409 0.014700 1.82 E O 0.646175 0.061377 2.00 E-1 0.2953 0.975489 0.007270 7.85 E 1 0.951679 0.011270 1.40 E O 0.698480 0.052304 1.71 E 1 0.3543 0.983399 0.007910 8.54 E 1 0.964389 0.012710 1.58 E O 0.762707 0.064227 2.09 E 1 0.4331 0.990129 0.006730 7.27 E 1 0.975679 0.011290 1.40 E O 0.827677 0.064971 2.12 E 1 0.5118 0.994129 0.004000 4.32 E 1 0.983189 0.007510 9.31 E-1 0.874808 0.047131 1.54 E 1 0.5906 0.996509 0.002380 2.57 E 1 0.988369 0.005180 6.42 E 1 0.909086 0.034277 1.12 E 1 0.7874 0.999049 0.002540 2.74 E 1 0.995339 0.006970 8.64 E-1 0.959109 0.050023 1.63 E 1 0.9840 0.999739 0.000690 7.45 E 2 0.998629 0.003290 4.08 E 1 0.981593 0.022484 7.33 E 2 1.1810 0.999929 0.000190 2.05 E 2 0.999629 0.001000 1.24 E 1 0.991728 0.010135 3.30 E 2 3750 0.999995 0.000066 7.13 E-3 0.999972 0.000343 4.25 E 2 0.996329 0.006601 2.15 E 2 1.9690 1.000000 0.000005 5.40 E-4 1.000000 0.000028 3.47 E 3 0.999663 0.001333 4.35 E-3 2.3630 s-0.999932 0.000269 8.78 E-4 2.7570 O.999986 0.000054 1.77 E-4 3.1510 e
~-
A'--
0.999997 0.000011 3.58 E 5 3.5450
-C t
t s f Ca.
. P.F
@ cic s
O.999999 0.000002 7.23 E 6 3.9390 5
~
~
1.000000 0.000000 1.46 E 6 Flaw depths in this table do not include clad thickness of 0.2 in., which was used in the l
NRC staff evaluations.
6 Flaw depths in this table do not include growth from SCC, which was included in the BWRVIP-05 evaluations.
Note: In Table D-1, the number of flaws per vessel for the BWRVIP-Marshall distribution is 3.26, reflecting 30 flaws /m* and the NRC calculated weld volume - the volume used by BWRVIP-05 is
-8 percent larger.
D-3
TABLE D 2 NUMBER OF FLAWS WITHIN BROAD FLAW DEPTH RANGES (FOR ONE MILLION VESSEL SIMULATIONS)
Number of Flaws in this Flaw Depth Range
< 0.5 in.
0.5 - 1.0 in.
1.0 2.0 in.
> 2 in.
PVRUF-Exponential Best-Estimate 105400000 2544000 102200 540 PVRUF-Exponential Upper Bound 118000000 5414000 574500 3472 BWRVIP-Marshall 2698000 501800 58910 1100 BWRVIP-Marshall + 15%
2698000 428500 127900 5446 BWRVIP Marshall + 50%
2277000 686600 269400 26967 TABLE D 3 P(FIE) FOR LIMITING PLANT-SPECIFIC CASE OF CIRCUMFERENTIAL FLAWS (MEAN RT.= 99.8'F)
USING DIFFERENT FLAW PARAMETER SETS Flaw Parameters P(FIE)
PVRUF-Exponential Best-Estimate 8.17 E-5 PVRUF Exponential Upper Bound 3.69 E-4 BWRVIP Marshall 7.91 E-5 BWRVIP Marshall + 15%
1.66 E-4
)
BWRVIP Marshall + 50%
5.54 E-4 D-4 i
l
5
%2 l
l h
2 M
n o
P i
i IV ta R
x\\
l 0
W u
im 2
B S
l V
l V
e t
s
_u' s
e
)
d
___w n
h B
c a
r
- 5. )n.
E ep r
1 i
p o
(
U f
H n
(
b N^
D E
T o
it P
P E
X u
i r
t F
W U
s N-iD A
R k
L
- 0. F V
h t
P p
1 e
D G
wa F
l
)
1 t
s
-D E
t t
e 5
s r
h"'
0 e
u B
g
(
iF P
=
X E-F U
R V
0 P
0 3
2 1
0 1
2 3
4 0
0 0
0 0
0 0
0 M
+
+
+
+
E E
E E
E E
E E
1 1
1 1
1 1
1 1
tN_oIg= Z4I y
ehh oZ u
c
e
- =-
gna l
R ll 0
la l
h 2
h t
s r
p a
e
- l. ~
M D
P w
IVR a
_= _ -
W l
)
F B
h
- 0. (n
- 1 i
Mu=
m c
2 E m
a E
- G n
- 0. N
)dn
- mN4j, 1
i A
B h
W R
t r
i ep H
p w
s T
U P
(
0 1'
,,l P
l I
- 0. E a
- l' X
+
5 D
E F
1 P
F f
I o
W
,I4 U
V 5
R R
A r
V W
.}'$[,
0 L
P B
e b
F m
u N
.j
)
2 5
t s
D 0
E
.)'
I ts e
e r
u
=I$$l i
- {!$
I B
(
g
=
5 P
1
=
i
=
F X
+
E P
-F I
3 2
1 0
1 2
3 4
5 U
V 0
0 0
0 0
0 0
0 0
R R
+
+
+
+
V W
E E
E E
E E
E E
E P
B 1
1 1
1 1
1 1
1 1
y j
$z.
d 0
n 2
B r
o r
f e
r p
e b
p U
m
(
u 5
P 1
N X
+
t s
E e
I=5-P K=
- 0. )n.
F g
I Il 2i U
V r
a L
a4 -
R R
(
E V
W
- 0. G P
B e
h 1
N t
o A
1 t
R s
w.
H ae T
lFg f
P n
E g=1 o a
- 0. D
)
rR t
W eh s
E bt
- A mp
- 5. L t
e s
u l
D
$$lrj' 0F e
iai*
l N
a w
I B
h f
(
s oa r
0 P
a 5
l oF X
M
+
i t
ae E
Rh P
P F
t I
I U
V V
R R
R 5
V WW 3
0 P
B B
D lg;E@[?Eg-g e
r p;
u l
l g
iF 0
8 6
4 2
0 1
0 0
0 0
0 u
t]<> gwE z or $5u u.o oi@
li:
lI'
r Appendix E a.
Additional Information to Be Provided to the Advisory Committee on Reactor Safeguards By letter dated September 10,1997, Robert L. Seale, Chairman of the Advisory Committee on Reactor Safeguards (ACRS), provided recommendations on several areas that the ACRS determined needed additional review before the NRC staff issues its final SE on the BWRVIP-05 report, as supplemented. The ACRSisubcommittees on Materials and Metallurgy and on Severe Accidents met with the NRC staff and representatives of the BWRVIP on August 26,1997, and the full ACRS committee reviewed tc* BWRVIP-05 report and the associated staff independent safety assessment, dated August 14,1997, during the 444th meeting on September 4,1997.
On August 14,1997, the staff issued a request for additional information and requested BWRVIP to respond to the issues raised by the ACRS. The BWRVIP response is contained in l
Its December 18,1997 and January 13,1998 letters to the NRC. The staff's evaluation of the BWRVIP response and the ACRS recommendations follows:
1)
The ACRS Indicated that additional effort was needed to address uncer'ainties i
associated with the BWRVIP-05 analyses such as showing that flaw size distribution l
input models are justified and censistent with available data including those obtained in past inspections of welds. The ACRS further indicated that the exponential decrease in flaw size distribution is chiefly responsible for the very large differences in failure i
probability between axial and circumferential welds. It is not clear that the experts who l
formulated this distribution as a bound on the frequency of large flaws intended for the l
distribution to be used to compare the relative frequencies of the approximately 2 cm flaws that lead to failure in axial welds and the approximately 4 cm flaws that lead to failure in circumferential welds. The ACRS also indicated the analyses should:
a) evaluate all available inspection data (PWRs, BWRs, and research), including those obtained in past inspections of the welds, to determine if the Marshall flaw size distribution produces realistic flaw distributions, b) address the flaw size distribution uncertainties associated with the BWRVIP-05 analyses which show that flaw size distributions inputted to the models are
+
Justifiable and consistent with available data.
Response
The BWRVIP utilized a Marshal flaw size distribution (hereafter: BWRVIP-Marshal flaw 8
distribution) and a flaw density of 30 flaws /m in its analyses. To validate the flaw distribution and density used in its analysis, the BWRVIP compared the BWRVIP-Marshal flaw distribution used in its analysis to the PVRUF-Marshal distribution that was used in the staff's interim Safety Assessment (ISA) and to the data reported from the inspection of the Browns Ferry Unit 3 RPV.
The analysis that is documented in a January 13,1998 letter indicates the BWRVIP-Marshal distribution produces greater P(FIE) for circumferential welds than the PVRUF Marshal distribution because flaws greater than 1 inch are mainly responsible for failure of circumferential welds and the BWRVIP-Marshal produces flaws at greater frequencies at depths greater than 1 inch than the PVRUF-Marshaldistribution. However,in the evaluation of the PVRUF-Marshal distribution, the BWRVIP used a flaw density of 30 flaws /m', rather than the PVRUF flaw density E-1
't e
en 8
of 995 flaws /m. In addition, the PVRUF-Marshal distribution that was used in the BWRVIP analysis did not include clad thickness. Since PVRUF-Marshal distribution used in the staff's ISA included the clad thickness in the distribution and utilized a larger flaw density, the conclusion that the BWRVIP-Marshal distribution produces greater P(FIE) than the PVRUF-Marshal distribution is not correct.
The BWRVIP reported that a total of 600 flaws were observed in the Browns Ferry Unit 3 inspection, ranging in size from 0.16 in. to 0.31 in. Tne BWRVIP observed that only small flaws were observed in the inspection and that the number of small flaws observed in the inspection was less than that predicted by the PVRUF-Marshal distribution. The number of small flaws observed in the Browns Ferry inspection exceeds the number of small flaws predicted by the BWRVIP-Marshal distribution; but since no deep flaws were observed, the BWRVIP concluded that the BWRVIP-Marshal distribution bounds the number of deep flaws (>1 in.) in the Browns Ferry inspection. Since deep flaws cause most of the failures in the probabilistic fracture mechanics analysis of circumferential welds, the BWRVIP concluded that the BWRVIP-Marshal distribution is both valid and conservative.
The staff has performed sensitivity studies to determine the impact of flaw size distribution on the conditional probability of vessel failure {P(FIE)) of RPV welds. These sensitivity studies are documented in Appendix D and Sections 2.6.2.4 and 2.6.2.5 of the staff's safety evaluation. As a result of comments by the ACRS, the staff revised the best-estimate and upper bound flaw distribution resulting from the PVRUF inspection data. The revised flaw distribution is identified as the PVRUF-Exponential flaw distribution. A discussion of the revised flaw distributions is l
l contained in Appendix A of the staff's safety evaluation. The PVRUF flaw size data was determined from inspection of actual beltline welds using a synthetic aperture technique for i
ultrasonic inspection (SAFT-UT). These data are considered by the staff to be the best source I
available for determining flaw size distributions in RPV welds because this inspection technique provides for better resolution of flaws than techniques used during inservice inspection of RPV welds in operating plants. The staff contractors performing destructive examination of the RPV welds.
The staff's inservice inspection consultant, PNNL, reports that the SAFT-UT database that was obtained on PVRUF was created to develop information on the density and distribution of reactor pressure vessel (RPV) fabrication flaws. The data was not taken based on ASME Code requirements. The data could be related to ASME Code requirements but this work has not L
been performed. However, it is still possible to provide insight as to what would be expected from ASME Code based inspections. Prior to 1986, the code requirements included using procedures based on 50 percent DAC inspection sensitivity. For examinations conducted from the inside of the RPV it was common practice to gate out the first inch of RPV material as a result of inspection transducer ring down. Therefore, none of the underciad near-surface flaws in the PVRUF data base would have been detected. The use of 50 percent DAC would not have detected the flaws in the PVRUF database fcund in the remainder of the vessel wall because the flaws had a small ultrasonic response and the results from PlSC ll show that for a flaw 50 mm deep the probability of detection would only be 10 percent. In 1986 the ASME Section XI Code was changed based on PlSC ll results to require that an effective near surface examination be performed and that a 20 percent DAC sensitivity procedure must be used.
Whenever a utility updated their ISI program to the 1986 or later editions of the ASME Code, the procedures would have met these requirements. The requirement of using an effective near surface technique would have meant that they used a transducer technique similar to that used E-2
by PNNL but without the SAFT processing of the data. Therefore, it would be expected that the larger (high signal-to-noise) flaws in the PVRUF database would have been detected in the near surface zone (first 25 mm). For the remainder of the vessel wall thickness, the use of 20 percent DAC sensitivity effectiveness must be related through the PISC !! study. In the PlSC ll study it was found that for flaws 17 mm in depth (largest validated flaw in the PVRUF database) a procedure that used a 20 percent DAC level of sensitivity would have an average probability of detection of 45 percent. Finally, inspections that would be performed today for those personnel, equipment and procedures that have successfully passed the performance demonstration test specified in ASME Section XI Code Appendix Vill, it is expected that they would detect all of the larger flaws in the PVRUF data base. This position is supported by the PISC ll data which shows for advanced techniques a probability of detection of 95 percent for a flaw 17 mm in depth.
Using the limiting plant-specific vessel parameters for circumferential welds and the temperature and pressures resulting from the limiting BWR RPV event, the P(FIE) for the PVRUF-Exponential upper bound flaw distribution is approximately 4.5 larger than the P(FIE) for the PVRUF-Exponential best-estimate flaw distribution. The P(FIE) for the PVRUF-Exponential best-estimate flaw distribution was determined to be equivalent to the P(FlE) for the BWRVIP-Marshal flaw distribution. The staff also evaluated the impact of increased flaw size on the BWRVIP-Marshal flaw distribution. Increasing the BWRVIP Marshal flaw size distribution by 15 percent doubles the P(FIE), and increasing the flaw size by 50 percent increase the P(FlE) by an additional factor of 3.3. These increases are small, given the fact that the limiting flaw size distribution and density increases the number of flaws greater than 2 in. by a factor of 50 and increases the number of flaws in the size range of 1 to 2 in. by a factor of almost 3 over the PVRUF-Exponential best-estimate flaw distribution.
2)
The ACRS indicated that additional uncertainty analysis be provided to address conservatism inherent in existing calculations, including the assumptions that weld flaws penetrate through the cladding and that all flaws are located on the inner surface of the vessel. The uncertainty analysis should also consider accident sequences that pose a threat to the welds, and operator actions that affect the probability of challenges to the integrity of BWR vessel welds.
t
Response
The BWRVIP analysis included the assumption that all flaws are located on the inner surface and that their size would increase as a result of stress corrosion crack growth. Since embedded flaws are not subject to stress corrosion crack growth and all flaws were assumed to be at the inner surface, the BWRVIP concluded that with more realistic assumptions, the P(FlE) for both axial and circumferential welds would be reduced but the relative probabilities and, therefore, the conclusions of the report, would not be changed.
Since the PVRUF inspection results indicate a high density of clad defects and the most likely-initiator of a vessel failure would be a SCC aligning with a crack in the clad or just below the l
clad, the assumption of a surface flaw is reasonable. However, the staff and licensee analyses l
assumed all flaws would be surface breaking and would be at the highest neutron fluence l
location. This is a conservative assumption since all flaws will not be surface breaking and in the highest neutron fluence location. Therefore, the staff agrees with the BWRVIP conclusion that assuming all flaws are at the t.urface results in a conservative analysis. The staff's E-3 l
L---
0 e
e consultant has completed the destructive tests of the PVRUF welds and a draft NUREG/CR on the vallation is planned for August 1998. The results of the validation can be used to evaluate the assumed flaw distribution and flaw density.
In a December 18,190, letter to the NRC the BWRVIP provided an analysis of the potential accident sequences and operator actions that pose a challenge to BWR vessel welds. This analysis is reviewed by the staff in its safety evaluation.
3)
The ACRS indicated that the staff should consider the relative value of partialinspections of the welds, including limitations in both accessibility of welds and t!.e capability to detect flaws by inspections.
Response
The CNhVIP concluded that based on the resu!ts of their analysis, inspection of the circumferential welds is of little or no value, if B'WRVIP-Marshal cracks in irradiated axial welds are assumed to be surface connected, and grow in size as a result of SCC when exposed to reactor coolant, the BWRVIP analysis indicates that 100 percent inspection produces about an order of mcgnitude decrease in the P(FIE) relative to no inspection.
The staff evaluated the capability to detect flaws in axial welds for three reference cases (B&W, CE and CB&I fabricated vessels) in Section 2.6.2.6, " Sensitivity to inservice inspection. Using the probability of detection of flaws resulting from the PISC 11 study, the staff determined that inservice inspection (ISI) reduces the P(FIE) for B&W, CE, and CB&l vessels by 40 percent,60 percent and 90 percent, respectively. The B&W and CE vessels are vessels with much greater amount of embrittlement than the CB&l vessels. The reduction in the P(FIE) for B&W and CE is not significant which indicates ISI would not have a big impact on reducing the P(FIE) for vessels with large amounts of embrittlement of axial welds. Although the reductions in the P(FIE) for CB&l vessels is significant, these represent vessels with low amount of embrittlement, which have low P(FIE) and ISI may not be needed. Partial inspections would further reduce the benefit of ISI since less weld metal would be inspected. The staff could not quantitatively evaluate the impact of partialinspection of circumferential welds because no failures were observed for circumferential flaws with inclusion of ISI. Since the P(FIE) of circumferential welds is dependent upon the frequency of large flaws and large flaws are more easily detected than small flaws, partial inspection of circumferential welds would have a greater impact on the reduction of the P(FIE) from inservice inspection than partialinspection of the axial welds.
However, since the P(FIE) of circumferential welds without inspection is already very low, the impact of partial inspection on circumferential welds is inconsequential.
Due to differences in processes used to fabricate axial and circumferential welds and differences in processes used to fabricate the clad on axial and circumferential welds, the likelihood of fabrication and service-induced flaws in axial and circumferential welds are not the same. Furtner, statistical analysis indicates that the existence of an isolated unacceptable flaw in a circumferential weld cannot be confidently ruled out based on axial weld inspection results.
Therefore, based on the above conclusions, a statistical sampling approach is not an effective approach for inferring the possibility of a critical isolated defect (e.g., a fabrication or repair flaw that was not identified). However, sampling inspections can provide useful information regarding the occurrence of wide spread service-induced degradation mechanisms.
E-
(
4)
The ACRS indicates that the staff should develop and maintain a database on vessel embrittlement specific to BWRs.
Response
The BWRVIP indicated that EPRIis maintaining two databases. One database contains a comprehensive listing of unirradiated and irradiated data. The second database contains data exclusively from surveillance programs.
I The staff is maintaining a surveillance material database for all LWRs. Since BWRs and PWRs were fabricated by the same manufacturers using many of the same materials, separate databases are not necessary nor appropriate, at least not with rogard to unirradiated mechanical properties and chemistries. However, since the neu'.ron flux for BWRs is less than that for PWRs, the relative amount of embrittlement may be different for the same material and the same neutron fluence. This will be evaluated in the staff's comprehensive evaluation of the embrittlement data.
5)
The ACRS suggested that it is important to also analyze failure probabilities beyond the current license limit. In these analyses, the contributions of base metal failure should be included.
Response
The BWRVIP compared the P(FlE) at 40 years of operation to ths P(FIE) at 60 years of operation for two axial and circumferential welds in the limiting BWR plants. For two axial welds evaluated, the P(FIE) for the 60 year cases increased by factors of 5.1 and 5.3. For the two circumferential welds evaluated, the P(FIE) was increased by factors of 12 and 9. The BWRVIP concluded that the contribution of RPV plates to the frequency of RPV failure would be very small and that the relative P(FIE) of the axial and circumferential welds would not change.
The staff evaluated the impact of operation beyond the current license term on the P(FIE) for the limiting plants that were fabricated by Combustion Engineering (CE), Babcock & Wilcox (B&W) and Chicago Bridge and Iron (CB&l). The results of these analyses are shown in Tables 2.6-4
~
and 2.6-5 of the staff's safety evaluation and are summarized in Table E 1. These analyses considered the RPV P(FIE) resulting from weld flaws and do not consider RPV plates. For the 5
limiting plant-specific case, the P(FIE) for circumferential welds increased from 8.2 x 10 to d
4.8 x 10. Combining the P(FIE) with the frequency of cold overpressurization event results in a 4
total vessel failure frequency of approximately 5 x 10. Since the total failure frequency for the li6niting circumferential weld could increase significantly after the current license renewal term and the type of flaws that could lead to vessel failure are age related each plant requesting license renewal will be requested to perform a plant-specific assessment based on the chemistry of its limiting weld and the neutron fluence at the end of the license renewal term.
Tha base metalin the beltline of all BWRs were fabricated using plate except for the Brunswick 1 and 2. Brunswick 1 and 2 have nozzle shell segments that are within the beltline region and were fabricated from forging material. The neutron fluence for these forgings are low and are not likely to significantly contribute to vessel failure.
E-5
s e
Although plates have a larger volume than welds, their contribution to vessel failure is considered to be small since most f!aws in plates w9uld be oriented in a manner that would not lead to vessel failure. It is for this reason that we direct our attention to inservice inspection of the welds and not the plates. Altnough the P(FIE) from plates would increase the P(FIE) of the RPV, the relative P(FIE) of axial welds to circumferential welds would be the same if the platas were included in the analysis.
TABLE E-1 IMPACT OF PLANT LIFE EXTENSION FROM 32 EFPY TO 64 EFPY ON THE CONDITIONAL PROBABILITY OF VESSEL FAILURE P(FIE)
P(FIE)
% increase in Flaw Orientation Group P(FIE) from 32 32 EFPY 64 EFPY EFPY to 64 EFPY CE (VIP)*
2.94 E-1 7.49 E 1 154 CE (CEOG)*
4.37 E 1 8.28 E 1 89 Axial CB&l 1.42 E 1 3.82 E 1 169 B&W 5.98 E 2 1.87 E 1 213 CE (VIP)*
2.81 E-5 1.99 E-4 608 CE (CEOG)*
6.34 E-5 4.38 E-4 592 Circumferential CB&l 2 E-7 1.78 E 5 8900 B&W 8.17 E 5 4.83 E 4 491 4
(VIP) is P(FIE) based on chemistry reported by BWRVIP and (CEOG) is P(FIE) based on chemistry reported in CEOG report for the limiting weld in a CE fabricated vessel. This is discussed in greater detail in Section 2.6.2.7.
j l
i E-6