ML16299A524

Forwards Excerpts from Draft Accident Sequence Precursor Rept for 1982-83,documenting ASP Program Analyses of Operational Events That Occurred During 1982-83

ML16299A524
Person / Time
Site:	Oconee
Issue date:	05/06/1996
From:	Labarge D NRC (Affiliation Not Assigned)
To:	Hampton J DUKE POWER CO.
References
NUDOCS 9605140378
Download: ML16299A524 (26)

Text

UNITED STATES NUCLEAR REGULATORY COMMISSION WASHINGTON, D.C. 20555-0001 May 6, 19 Mr. J. W. Hampton Vice President, Oconee Site Duke Power Company P. 0. Box 1439 Seneca, South Carolina 29679

SUBJECT:

DRAFT 1982-83 PRECURSOR REPORT

Dear Mr. Hampton:

Enclosed for your information are excerpts from the draft Accident Sequence Precursor (ASP) Report for 1982-83. This report documents the ASP Program analyses of operational events that occurred during the period 1982-83. We are providing the appropriate section of this draft report to each licensee with a plant that had an event in 1982 or 1983 that has been identified as a precursor. One of these precursors occurred at Oconee Unit 3. Also enclosed for your information are copies of Section 2.0 and Appendix A from the 1982-83 ASP Report. Section 2.0 discusses the ASP Program event selection criteria and the precursor quantification process;.Appendix A describes the models used in the analyses.

We emphasize that you are under no licensing obligation to review and comment on the enclosures.

The analyses documented in the draft ASP Report for 1982-83 were performed primarily for historical purposes to obtain the 2 years of precursor data for the NRC's ASP Program, which had previously been missing. We realize that any review of the precursor analyses of the 1982-83 events by affected licensees would necessarily be limited in scope due to (1) the extent of the licensee's corporate memory about specific details of an event that occurred 13-14 years ago, (2) the desire to avoid competition for internal licensee staff resources with other, higher priority work, and (3) extensive changes in plant design, procedures, or operating practices implemented since the time period 1982-83, which may have resulted in significant reductions in the probability of (or, in some cases, even precluded) the occurrence of events such as those documented in this report.

The draft report contains detailed documentation for all precursors with conditional core damage probabilities > 1.0 x 10-.

However, the relatively large number of precursors identified for the period 1982-83 necessitated that only summaries be provided for precursors with conditional core damage probabilities between 1.0 x 10 and 1.0 x 10-s We will begin revising the report about May 31, 1996, to put it in final form for publication. We will respond to any comments on the precursor analyses that we receive from the licensees. The responses will be placed in a separate section of the final report. Duke Power Company is on distribution for the final report.

9605140378 960506 CR V

PDR ADOCK 05000287 S

PDR

Mr. J. W. Hampton

-2 Please contact me at (301) 415-1472 if you have any questions regarding this letter. Any response to this letter on your part is entirely voluntary and does not constitute a licensing requirement.

Sincerely, David E. LaBarge, Senior Project Manager Project Directorate 11-2 Division of Reactor Projects-I/Il Office of Nuclear Reactor Regulation Docket No. 50-287

Enclosures:

1. Section B.17, "Precursor Analysis of 10/13/83 Loss of Feedwater Transient with One EFW Pump Inoperable"

2. Section 2, "Selection Criteria and Quantification"

3. Appendix A, "ASP Models" cc w/encls: See next page

Mr. J.

May 6, 1997 Please contact me at (301) 415-1472 if you have any questions regarding this letter. Any response to this letter on your part is entirely voluntary and does not constitute a licensing requirement.

Sincerely,

/s/

David E. LaBarge, Senior Project Manager Project Directorate 11-2 Division of Reactor Projects-I/II Office of Nuclear Reactor Regulation Docket No. 50-287

Enclosures:

1. Section B.17, "Precursor Analysis of 10/13/83 Loss of Feedwater Transient with One EFW Pump Inoperable"

2. Section 2, "Selection Criteria and Quantification"

3. Appendix A, "ASP Models" cc w/encls:

See next page Distribution:

Dqocket FieJ PUBLIC PD 11-2 Rdg.

SVarga JZwolinski HBerkow LBerry DLaBarge OGC ACRS EMerschoff, RH RCrlenjak, RH To receive a copy of this document, indicate in the box:

"C" = Copy without attachment/enclosure "E" = Copy with attachment/enclosure "N" = No copy OFFICE LA:PD22:pE E PM:PDII-2 6 D:PIA NAME LBERRY i

IDLABAR.kn IHBERK I

DATE

/0\\ / 6 9

/

/96

>/

/96

/

/96

/

/96

/

/96 DOCUMENT NAME:

G:\\OCONEE\\ASP8283.STD OFFICIAL RECORD COPY

Mr. J. W. Hampton Duke Power Company Oconee Nuclear Station cc:

Mr. Paul R. Newton Mr. Ed Burchfield Legal Department (PB05E)

Compliance Duke Power Company Duke Power Company 422 South Church Street Oconee Nuclear Site Charlotte, North Carolina 28242-0001 P. 0. Box 1439 Seneca, South Carolina 29679 J. Michael McGarry, III, Esquire Winston and Strawn Ms. Karen E. Long 1400 L Street, NW.

Assistant Attorney General Washington, DC 20005 North Carolina Department of Justice Mr. Robert B. Borsum P. 0. Box 629 B&W Nuclear Technologies Raleigh, North Carolina 27602 Suite 525 1700 Rockville Pike Mr. G. A. Copp Rockville, Maryland 20852-1631 Licensing - EC050 Duke Power Company Manager, LIS 526 South Church Street NUS Corporation Charlotte, North Carolina 28242-0001 2650 McCormick Drive, 3rd Floor Clearwater, Florida 34619-1035 Dayne H. Brown, Director Division of Radiation Protection Senior Resident Inspector North Carolina Department of U. S. Nuclear Regulatory Commission Environment, Health and Route 2, Box 610 Natural Resources Seneca, South Carolina 29678 P. 0. Box 27687 Raleigh, North Carolina 27611-7687 Regional Administrator, Region II U. S. Nuclear Regulatory Commission 101 Marietta Street, NW. Suite 2900 Atlanta, Georgia 30323 Max Batavia, Chief Bureau of Radiological Health South Carolina Department of Health and Environmental Control 2600 Bull Street Columbia, South Carolina 29201 County Supervisor of Oconee County Walhalla, South Carolina 29621

B.17-1 B.17 LER No. 287/83-011 Event

Description:

Loss of feedwater transient with one EFW pump inoperable Date of Event:

October 13, 1983 Plant:

Oconee 3 B.17.1 Summary On October 13, 1983, following a loss of main feedwater, the 3B motor-driven Emergency Feedwater (EFW) pump emergency started. Approximately 30 minutes later, it was discovered that there was no Low Pressure Service Water (LPSW) flow through the 3B AFW pump due to a failed solenoid valve. The pump was declared inoperable.

The estimated conditional core damage probability for this event is 2.7 x 107.

B.17.2 Event Description On October 13, 1983, following a loss of main feedwater (MFW) caused by a loss of instrument air pressure, the 3B motor driven EFW pump emergency started. Approximately 45 minutes later, it was discovered that no LPSW was flowing through the 3B EFW pump. The pump was shutdown and declared inoperable.

Investigation revealed that the solenoid valve 3SV-203 that operates the 3B EFW motor cooler outlet valve had failed. The outlet valve did not open when the pump was started. The plunger of valve 3SV-203 was found stuck. The valve was replaced and the 3B EFW pump was declared operable the next day. To help identify low LPSW flow to the EFW pumps, the computer points for flow were to be changed to provide an alarm if flow is low while the pumps are running.

On October 18, 1983, an additional problem was found related to the 3B EFW pump. While operating at 100% power, a performance test on 3B EFW pump was performed. Feedwater valve 3FDW-382 was closed per procedure. Upon completion of the test, the valve was re-opened by placing the valve position switch to its normally open position in the control room. The control panel indicated the valve was open when it was not completely open. Visual inspection of the valve during the next shift revealed that the valve was only 10%

open. Attempts to close the valve from the control room failed and the valve and pump were declared inoperable. The direct cause was unknown, but the valve was successfully cycled from the control room approximately 40 minutes later. This event was reported in LER 287/83-012.

B.17.3 Additional Event-Related Information Oconee 3 has a three train EFW system. Two trains have motor-driven EFW pumps (3A and 3B). Each motor driven pump is aligned to one steam generator. The third train has a turbine-driven pump which is aligned to both steam generators. In the event that main feedwater is lost, the EFW pumps provide secondary side cooling to the steam generators. One pump delivering flow to one steam generator provide sufficient secondary side cooling in the event that main feedwater is lost. In addition to the EFW pumps, EFW can also LER No. 287/83-011 ENCLOSURE 1

B.17-2 be supplied through manual cross-connections with Units I and 2. LPSW provides motor cooling to both motor-driven EFW pumps and turbine jacket cooling and bearing oil cooling for the turbine-driven EFW pump.

B.17.4 Modeling Assumptions The event was modeled as a transient with loss of MFW and one motor-driven EFW pump inoperable because of the loss of LPSW cooling. The valve failure reported in LER 287/83-012 was assumed not to impact the EFW system during the loss of feedwater. MFW was set to failed to reflect the loss of feedwater. Because of the loss of instrument air pressure, MFW was assumed to be non-recoverable. In revising the EFW system failure probability, it was assumed that the type of valve fault observed on pump 3B could also have affected the motor-driven pump 3A. The potential for common-cause failure exists, even when a component is failed.

Therefore, the conditional probability of a common-cause failure was included in the analysis for those components that failed as part of the event. The potential to cross-connect EFW from one of the other units was not addressed in this analysis.

B.17.5 Analysis Results The estimated conditional core damage probability for this event is 2.7 x 10Y.

The dominant sequence highlighted on the event tree in Figure B. 17.1 (to be provided in the final report) involved a plant trip with the failure of AFW, the failure of MFW and the failure of feed and bleed.

LER No. 287/83-011

TRANS AT AFW MF.W PORV PORV HPI FEED & RCV RCS CHALL RESEAT BLEED SEC SIDE COOL-RHR HPR END SEQ COOLING DOWN STATE NO OK 101 OK 102 OK 103 CD 104

._O K

10 5 CD 106 CD 107 OK 108 OK 109 OK 110 OK 111 CD 112 OK 113 CD 114 CD 115 OK 116 OK 117 OK 118 CD

• 119 CD 120 ATWS

-3.17. I

B.17-5 Train 1 Cond Prob:

2.OE-02 > Failed Train 2 Cond Prob:

1.OE-01 Train 3 Cond Prob:

5.OE-02 Serial Component Prob:

2.8E-04 afw/atus 4.3E-03 1.OE+00 afu/ep 5.OE-02 3.4E-01 MFW 2.OE-01 > 1.OE+00 3.4E-01 > 1.OE+00 Branch Model:

1.OF.1 Train 1 Cond Prob:

2.0E-01 > Failed porv.chalt 8.0E-02 1.OE+00 porv.chall/afw 1.0E+00 1.0E+00 porv.chalt/Loop 1.0E-01 1.OE+00 porv.chatt/sbo 1.0E+00 1.OE+00 porv.reseat 1.OE-02 1.1E-02 porv.reseat/ep 1.0E-02 1.OE+00 srv.reseat(atws) 1.OE-01 1.OE+00 hpi 3.0E-04 8.9E-01 feed.bteed 3.OE-04 1.OE+00 1.OE-02 enrg.boration 0.OE+00 1.OE+00 1.0E-02 recov.sec.coot 2.OE-01 1.OE+00 recov.sec.cooL/offsite.pwr 3.4E-01 1.OE+00 rcs.cooldown 3.OE-03 1.OE+00 1.OE-03 rhr 2.2E-02 7.0E-02 1.OE-03 rhr.and.hpr 1.0E-03 1.OE+00 1.OE-03 hpr 4.OE-03 1.OE+00 1.OE-03 ep 6.OE-04 8.9E-01 seal.loca 0.OE+00 1.OE+00 offsite.pwr.rec/-ep.and.-afw 2.7E-01 1.0E+00 offsite.pur.rec/-ep.and.afw 1.6E-01 1.0E+00 offsite.pwr.rec/seaL.Loca 0.OE+00 1.OE+00 offsite.pwr.rec/-seat.loca 4.5E-01 1.OE+00 sg.iso.and.rcs.cooldown 1.OE-02 1.OE-01 rcs.cool.below.rhr 3.OE-03 1.OE+00 3.0E-03 prim.press.Limited 8.8E-03 1.0E+00 branch model file

• • forced Heather Schriner 10-25-1995 10:20:51 LER No. 287/83-011

B.17-4 CONDITIONAL CORE DAMAGE PROBABILITY CALCULATIONS Event Identifier:

287/83-011 Event

Description:

Transient with one EFW pump inoperable Event Date:

October 13, 1983 Plant:

Oconee 3 INITIATING EVENT NON-RECOVERABLE INITIATING EVENT PROBABILITIES TRANS 1.OE+00 SEQUENCE CONDITIONAL PROBABILITY SUMS End State/Initiator Probability CO TRANS 2.7E-05 Total 2.7E-05 SEQUENCE CONDITIONAL PROBABILITIES (PROBABILITY ORDER)

Sequence End State Prob N Rec**

120 trans -rt AFW MFU feed.bLeed CD 2.4E-05 4.5E-01 119 trans -rt AFW MFW -feed.bteed recov.sec.cool hpr CD 2.3E-06 4.5E-01

• • non-recovery credit for edited case SEQUENCE CONDITIONAL PROBABILITIES (SEQUENCE ORDER)

Sequence End State Prob N Rec**

119 trans -rt AFW MFW -feed.bteed recov.sec.coot hpr CD 2.3E-06 4.5E-01 120 trans -rt AFW MFW feed.bLeed CD 2.4E-05 4.5E-01

• • non-recovery credit for edited case SEQUENCE MODEL:

c:\\aspcode\\models\\pwrb8283.cmp BRANCH MODEL:

c:\\aspcode\\modeis\\oconee3.82 PROBABILITY FILE:

c:\\aspcode\\models\\pwr8283.pro No Recovery Limit BRANCH FREQUENCIES/PROBABILITIES Branch System Non-Recov Opr Fail trans 2.4E-04 1.OE+00 Loop 1.6E-05 2.4E-01 Loca 2.4E-06 5.4E-01 sgtr 1.6E-06 1.OE+00 rt 2.8E-04 1.OE-01 rt(loop)

O.OE+00 1.OE+00 AFW 3.8E-04 > 5.3E-03 4.5E-01 Branch Model:

1.OF.3+ser LER No. 287/83-011

2-1 2.0 Selection Criteria and Quantification 2.1 Accident Sequence Precursor Selection Criteria The Accident Sequence Precursor (ASP) Program identifies and documents potentially important operational events that have involved portions of core damage sequences and quantifies the core damage probability associated with those sequences.

Identification of precursors requires the review of operational events for instances in which plant functions that provide protection against core damage have been challenged or compromised. Based on previous experience with reactor plant operational events, it is known that most operational events can be directly or indirectly associated with four initiators: trip [which includes loss of main feedwater (LOFW) within its sequences],

loss-of-offsite power (LOOP), small-break loss-of-coolant accident (LOCA), and steam generator tube ruptures (SGTR) (PWRs only). These four initiators are primarily associated with loss of core cooling. ASP Program staff members examine licensee event reports (LERs) and other event documentation to determine the impact that operational events have on potential core damage sequences.

2.1.1 Precursors This section describes the steps used to identify events for quantification. Figure 2.1 illustrates this process.

A computerized search of the SCSS data base at the Nuclear Operations Analysis Center (NOAC) of the Oak Ridge National Laboratory was conducted to identify LERs that met minimum selection criteria for precursors.

This computerized search identified LERs potentially involving failures in plant systems that provide protective functions for the plant and those potentially involving core damage-related initiating events. Based on a review of the 1984-1987 precursor evaluations and all 1990 LERs, this computerized search successfully identifies almost all precursors and the resulting subset is approximately one-third to one-half of the total LERs. It should be noted, however, that the computerized search scheme has not been tested on the LER database for the years prior to 1984. Since the LER reporting requirements for 1982-83 were different than for 1984 and later, the possibility exists that some 1982-83 precursor events were not included in the selected subset. Events described in NUREG -090020 and in issues of Nuclear Safety that potentially impacted core damage sequences were also selected for review.

Those events selected for review by the computerized search of the SCSS data base underwent at least two independent reviews by different staff members. The independent reviews of each LER were performed to determine if the reported event should be examined in greater detail. This initial review was a bounding review, meant to capture events that in any way appeared to deserve detailed review and to eliminate events that were clearly unimportant. This process involved eliminating events that satisfied predefined criteria for rejection and accepting all others as either potentially significant and requiring analysis, or potentially significant but impractical to analyze. All events identified as impractical to analyze at any point in the study are documented in Appendix E. Events were also eliminated from further review if they had little impact on core damage sequences or provided little new information on the risk impacts of plant operation-for example.

short-term single failures in redundant systems, uncomplicated reactor trips, and LOFW events.

Selection Criteria and Quantification ENCLOSURE 2

2-2 LERs requiring review Does the event only involve:

• component failure (no loss of redundancy)

• loss of redundancy (single system)

-seismic qualification/design error

.environmental qualification/design error Yes pre-critical event Reject

• structural degradation

.design error discovered by re-analysis

-bounded by trip or LOFW n so appreciable safety system impact

-shutdown-related event

-post-core damage impacts only No No Can event be reasonably analyzed byIdniyapoetlysgifctbu PRA-based models?

ipatclt nlz

• Yes Perform detailed review. analysis, and Define impact of event in terms of initiator ASP models system descriptions, Rejectctc

[Modfy branch probabilities to reflect event.

FCalculate conditional probability associae with event using modified event trees.

Does operational event involve:

No Dia core damage initiator o

ra total loss of a system Reject oa dos of redundancy in two or more systems a reactor trip with a degraded mitigating system Yes N

Is conditional probability _ 10-Reject based on low probability DYes Document as a precursor Figure 2.1 ASP Analysis Process Selection Criteria and Quantification

2-3 LERs were eliminated from further consideration as precursors if they involved, at most, only one of the following:

a component failure with no loss of redundancy, a short-term loss of redundancy in only one system, a seismic design or qualification error, an environmental design or qualification error, a structural degradation, an event that occurred prior to initial criticality, a design error discovered by reanalysis, an event bounded by a reactor trip or LOFW, an event with no appreciable impact on safety systems, or an event involving only post core-damage impacts.

Events identified for further consideration typically included the following:

unexpected core damage initiators (LOOP, SGTR, and small-break LOCA);

all events in which a reactor trip was demanded and a safety-related component failed; all support system failures, including failures in cooling water systems, instrument air, instrumentation and control, and electric power systems; any event in which two or more failures occurred; any event or operating condition that was not predicted or that proceeded differently from the plant design basis; and any event that, based on the reviewers' experience, could have resulted in or significantly affected a chain of events leading to potential severe core damage.

Events determined to be potentially significant as a result of this initial review were then subjected to a thorough, detailed analysis. This extensive analysis was intended to identify those events considered to be precursors to potential severe core damage accidents, either because of an initiating event, or because of failures that could have affected the course of postulated off-normal events or accidents. These detailed reviews were not limited to the LERs; they also used final safety analysis reports (FSARs) and their amendments, individual plant examinations (IPEs), and other information related to the event of interest.

The detailed review of each event considered the immediate impact of an initiating event or the potential impact of the equipment failures or operator errors on readiness of systems in the plant for mitigation of off-normal and accident conditions. In the review of each selected event, three general scenarios (involving both the actual event and postulated additional failures) were considered.

1.

If the event or failure was immediately detectable and occurred while the plant was at power, then the event was evaluated according to the likelihood that it and the ensuing plant response could lead to severe core damage.

2.

If the event or failure had no immediate effect on plant operation (i.e., if no initiating event occurred), then the review considered whether the plant would require the failed items for mitigation of potential severe core damage sequences should a postulated initiating event occur during the failure period.

Selection Criteria and Quantification

2-4

3.

If the event or failure occurred while the plant was not at power, then the event was first assessed to determine whether it impacted at-power or hot shutdown operation. If the event could only occur at cold shutdown or refueling shutdown, or the conditions clearly did not impact at-power operation, then its impact on continued decay heat removal during shutdown was assessed; otherwise it was analyzed as if the plant were at power. (Although no cold shutdown events were analyzed in the present study, some potentially significant shutdown related events are described in Appendix D).

For each actual occurrence or postulated initiating event associated with an operational event reported in an LER or multiple LERs, the sequence of operation of various mitigating systems required to prevent core damage was considered. Events were selected and documented as precursors to potential severe core damage accidents (accident sequence precursors) if the conditional probability of subsequent core damage was at least 1.0 X 106 (see section 2.2). Events of low significance are thus excluded, allowing attention to be focused on the more important events. This approach is consistent with the approach used to define 1988-1993 precursors, but differs from that of earlier ASP reports, which addressed all events meeting the precursor selection criteria regardless of conditional core damage probability.

As noted above, 115 operational events with conditional probabilities of subsequent severe core damage 1.0 X 10-6 were identified as accident sequence precursors.

2.1.2 Potentially Significant Shutdown-Related Events No cold shutdown events were analyzed in this study because the lack of information concerning plant status at the time of the event (e.g., systems unavailable, decay heat loads, RCS heat-up rates, etc.) prevented development of models for such events. However, cold shutdown events such as a prolonged loss of RHR cooling during conditions of high decay heat can be risk significant. Sixteen shutdown-related events which may have potential risk significance are described in Appendix D.

2.1.3 Potentially Significant Events Considered Impractical to Analyze In some cases, events are impractical to analyze due to lack of information or inability to reasonably model within a probabilistic risk assessment (PRA) framework, considering the level of detail typically available in PRA models and the resources available to the ASP Program.

Forty-three events (some involving more than a single LER) identified as potentially significant were considered impractical to analyze. It is thought that such events are capable of impacting core damage sequences. However, the events usually involve component degradations in which the extent of the degradation could not be determined or the impact of the degradation on plant response could not be ascertained.

For many events classified as impractical to analyze, an assumption that the affected component or function was unavailable over a 1-year period (as would be done using a bounding analysis) would result in the conclusion that a very significant condition existed. This conclusion would not be supported by the specifics of the event as reported in the LER(s) or by the limited engineering evaluation performed in the ASP Program Descriptions of events considered impractical to analyze are provided in Appendix E.

Selection Criteria and Quantification

2-5 2.1.4 Containment-Related Events In addition to accident sequence precursors, events involving loss of containment functions, such as containment cooling, containment spray, containment isolation (direct paths to the environment only), or hydrogen control, identified in the reviews of 1982-83 LERs are documented in Appendix F. It should be noted that the SCSS search algorithm does not specifically search for containment related events. These events, if identified for other reasons during the search, are then examined and documented.

2.1.5 "Interesting" Events Other events that provided insight into unusual failure modes with the potential to compromise continued core cooling but that were determined not to be precursors were also identified. These are documented as "interesting" events in Appendix G.

2.2 Precursor Quantification Quantification of accident sequence precursor significance involves determination of a conditional probability of subsequent severe core damage, given the failures observed during an operational event. This is estimated by mapping failures observed during the event onto the ASP models, which depict potential paths to severe core damage, and calculating a conditional probability of core damage through the use of event trees and system models modified to reflect the event. The effect of a precursor on event tree branches is assessed by reviewing the operational event specifics against system design information. Quantification results in a revised probability of core damage failure, given the operational event. The conditional probability estimated for each precursor is useful in ranking because it provides an estimate of the measure of protection against core damage that remains once the observed failures have occurred. Details of the event modeling process and calculational results can be found in Appendix A of this report.

The frequencies and failure probabilities used in the calculations are derived in part from data obtained across the light-water reactor (LWR) population for the 1982-86 time period, even though they are applied to sequences that are plant-specific in nature. Because of this, the conditional probabilities determined for each precursor cannot be rigorously associated with the probability of severe core damage resulting from the actual event at the specific reactor plant at which it occurred. Appendix A documents the accident sequence models used in the 1982-83 precursor analyses, and provides examples of the probability values used in the calculations.

The evaluation of precursors in this report considered equipment and recovery procedures believed to have been available at the various plants in the 1982-83 time frame. This includes features addressed in the current (1994) ASP models that were not considered in the analysis of 1984-91 events, and only partially in the analysis of 1992-93 events. These features include the potential use of the residual heat removal system for long-term decay heat removal following a small-break LOCA in PWRs, the potential use of the reactor core isolation cooling system to supply makeup following a small-break LOCA in BWRs, and core damage sequences associated with failure to trip the reactor (this condition was previously designated "ATWS," and not developed). In addition, the potential long-term recovery of the power conversion system for BWR decJ\\

heat removal has been addressed in the models.

Selection Criteria and Quantification

2-6 Because of these differences in the models, and the need to assume in the analysis of 1982-83 events that equipment reported as failed near the time of a reactor trip could have impacted post-trip response (equipment response following a reactor trip was required to be reported beginning in 1984), the evaluations for these years may not be directly comparable to the results for other years.

Another difference between earlier and the most recent (1994) precursor analyses involves the documentation of the significance of precursors involving unavailable equipment without initiating events. These events are termed unavailabilities in this report, but are also referred to as condition assessments. The 1994 analyses distinguish a precursor conditional core damage probability (CCDP), which addresses the risk impact of the failed equipment as well as all other nominally functioning equipment during the unavailability period, and an importance measure defined as the difference between the CCDP and the nominal core damage probability (CDP) over the same time period. This importance measure, which estimates the increase in core damage probability because of the failures, was referred to as the CCDP in pre-1994 reports, and was used to rank unavailabilities.

For most unavailabilities that meet the ASP selection criteria, observed failures significantly impact the core damage model. In these cases, there is little difference between the CCDP and the importance measure. For some events, however, nominal plant response dominates the risk. In these cases, the CCDP can be considerably higher than the importance measure. For 1994 unavailabilities, the CCDP, CDP, and importance are all provided to better characterize the significance of an event. This is facilitated by the computer code used to evaluate 1994 events (the GEM module in SAPHIRE), which reports these three values.

The analyses of 1982-83 events, however, were performed using the event evaluation code (EVENTEVL) used in the assessment of 1984-93 precursors. Because this code only reports the importance measure for unavailabilities, that value was used as a measure of event significance in this report. In the documentation of each unavailability, the importance measure value is referred to as the increase in core damage probability over the period of the unavailability, which is what it represents. An example of the difference between a conditional probability calculation and an importance calculation is provided in Appendix A.

2.3 Review of Precursor Documentation With completion of the initial analyses of the precursors and reviews by team members, this draft report containing the analyses is being transmitted to an NRC contractor, Oak Ridge National Laboratories (ORNL),

for an independent review. The review is intended to (1) provide an independent quality check of the analyses, (2) ensure consistency with the ASP analysis guidelines and with other ASP analyses for the same event type, and (3) verify the adequacy of the modeling approach and appropriateness of the assumptions used in the analyses. In addition, the draft report is being sent to the pertinent nuclear plant licensees for review and to the NRC staff for review. Comments received from the licensees within 30 days will be considered during resolution of comments received from ORNL and NRC staff.

2.4 Precursor Documentation Format The 1982-83 precursors are documented in Appendices B and C. The at-power events with conditional core damage probabilities (CCDPs) 1.0 x 105 are contained in Appendix B and those with CCDPs between 1.0 x 10' and 1.0 x 10- are summarized in Appendix C. For the events in Appendix B, a description of the event Selection Criteria and Quantification

2-7 is provided with additional information relevant to the assessment of the event, the ASP modeling assumptions and approach used in the analysis, and analysis results. The conditional core damage probability calculations are documented and the documentation includes probability summaries for end states, the conditional probabilities for the more important sequences and the branch probabilities used. A figure indicating the dominant core damage sequence postulated for each event will be included in the final report. Copies of the LERs are not provided with this draft report.

2.5 Potential Sources of Error As with any analytic procedure, the availability of information and modeling assumptions can bias results. In this section, several of these potential sources of error are addressed.

1.

Evaluation of only a subset of 1982-83 LERs. For 1969-1981 and 1984-1987, all LERs reported during the year were evaluated for precursors. For 1988-1994 and for the present ASP study of 1982-83 events, only a subset of the LERs were evaluated after a computerized search of the SCSS data base. While this subset is thought to include most serious operational events, it is possible that some events that would normally be selected as precursors were missed because they were not included in the subset that resulted from the screening process.

Reports to Congress on Abnotmal Occurrences2 (NUREG-0900 series) and operating experience articles in Nuclear Safety were also reviewed for events that may have been missed by the SCSS computerized screening.

2.

Inherent biases in the selection process. Although the criteria for identification of an operational event as a precursor are fairly well-defined, the selection of an LER for initial review can be somewhat judgmental. Events selected in the study were more serious than most, so the majority of the LERs selected for detailed review would probably have been selected by other reviewers with experience in LWR systems and their operation. However, some differences would be expected to exist; thus, the selected set of precursors should not be considered unique.

3.

Lack of appropriate event information. The accuracy and completeness of the LERs and other event-related documentation in reflecting pertinent operational information for the 1982-83 events are questionable in some cases. Requirements associated with LER reporting at the time, plus the approach to event reporting practiced at particular plants, could have resulted in variation in the extent of events reported and report details among plants. In addition, only details of the sequence (or partial sequences for failures discovered during testing) that actually occurred are usually provided; details concerning potential alternate sequences of interest in this study must often be inferred. Finally, the lack of a requirement at the time to link plant trip information to reportable events required that certain assumptions be made in the analysis of certain kinds of 1982-83 events. Specifically, through use of the "Grey Books" (Licensed Operating Reactors Status Report, NUREG-0200) 9 it was possible to determine that system unavailabilities reported in LERs could have overlapped with plant trips if it was assumed that the component could have been out-of-service for 2 the test/surveillance period associated with that component. However, with the link between trip, and events not being described in the LERs, it was often impossible to determine whether r not the component was actually unavailable during the trip or whether it was demandcJ Selection Criteria and Quantification

2-8 during the trip. Nevertheless, in order to avoid missing any important precursors for the time period, any reported component unavailability which overlapped a plant trip within /2 of the component's test/surveillance period, and which was believed not to have been demanded during the trip, was assumed to be unavailable concurrent with the trip. (If the component had been demanded and failed, the failure would have been reported; if it had been demanded and worked successfully, then the failure would have occurred after the trip). Since such assumptions may be conservative, these events are distinguished from the other precursors listed in Tables 3.1 - 3.6. As noted above, these events are termed "windowed" events to indicate that they were analyzed because the potential time window for their unavailability was assumed to have overlapped a plant trip.

4.

Accuracy of the ASP models and probability data. The event trees used in the analysis are plant-class specific and reflect differences between plants in the eight plant classes that have been defined. The system models are structured to reflect the plant-specific systems, at least to the train level. While major differences between plants are represented in this way, the plant models utilized in the analysis may not adequately reflect all important differences.

Modeling improvements that address these problems are being pursued in the ASP Program.

Because of the sparseness of system fa lure events, data from many plants must be combined to estimate the failure probability of a multitrain system or the frequency of low-and moderate-frequency events (such as LOOPs and small-break LOCAs). Because of this, the modeled response for each event will tend toward an average response for the plant class. If systems at the plant at which the event occurred are better or worse than average (difficult to ascertain without extensive operating experience), the actual conditional probability for an event could be higher or lower than that calculated in the analysis.

Known plant-specific equipment and procedures that can provide additional protection against core damage beyond the plant-class features included in the ASP event tree models were addressed in the 1982-83 precursor analysis for some plants. This information was not uniformly available; much of it was based on FSAR and IPE documentation available at the time this report was prepared. As a result, consideration of additional features may not be consistent in precursor analyses of events at different plants. However, analyses of multiple events that occurred at an individual plant or at similar units at the same site have been consistently analyzed.

5.

Difficulty in determining the potential for recovery of failed equipment. Assignment of recovery credit for an event can have a significant impact on the assessment of the event. The approach used to assign recovery credit is described in detail in Appendix A. The actual likelihood of failing to recover from an event at a particular plant during 1982-83 is difficult to assess and may vary substantially from the values currently used in the ASP analyses. This difficulty is demonstrated in the genuine differences in opinion among analysts, operations and maintenance personnel, and others, concerning the likelihood of recovering from specific failures (typically observed during testing) within a time period that would prevent core damage following an actual initiating event.

6.

Assumption of a 1-month test interval. The core damage probability for precursors invol% n:

Selection Criteria and Quantification

2-9 unavailabilities is calculated on the basis of the exposure time associated with the event. For failures discovered during testing, the time period is related to the test interval. A test interval of 1 month was assumed unless another interval was specified in the LER. See reference 1 for a more comprehensive discussion of test interval assumptions.

Selection Criteria and Quantification

A-1 Appendix A:

ASP MODELS ASP MODE L.S ENCLOSURE 3

A-2 A.0 ASP Models This appendix describes the methods and models used to estimate the significance of 1982-83 precursors. The modeling approach is similar to that used to evaluate 1984-91 operational events. Simplified train-based models are used, in conjunction with a simplified recovery model, to estimate system failure probabilities specific to an operational event. These probabilities are then used in event tree models that describe core damage sequences relevant to the event. The event trees have been expanded beyond those used in the analysis of 1984-91 events to address features of the ASP models used to assess 1994 operational events (Ref. 1) known to have existed in the 1982-83 time period.

A.1 Precursor Significance Estimation The ASP program performs retrospective analyses of operating experience. These analyses require that certain methodological assumptions be made in order to estimate the risk significance of an event. If one assumes, following an operational event in which core cooling was successful, that components observed failed were "failed" with probability 1.0, and components that functioned successfully were "successful" with probability 1.0, then one can conclude that the risk of core damage was zero, and that the only potential sequence was the combination of events that occurred. In order to avoid such trivial results, the status of certain components must be considered latent. In the ASP program, this latency is associated with components that operated successfuly-these components are considered to have been capable of failing during the operational event.

Quantification of precursor significance involves the determination of a conditional probability of subsequent core damage given the failures and other undesirable conditions (such as an initiating event or an unexpected relief valve challenge) observed during an operational event. The effect of a precursor on systems addressed in the core damage models is assessed by reviewing the operational event specifics against plant design and operating information, and translating the results of the review into a revised model for the plant that reflects the observed failures. The precursors's significance is estimated by calculating a conditional probability of core damage given the observed failures. The conditional probability calculated in this way is useful in ranking because it provides an estimate of the measure of protection against core damage remaining once the observed failures have occurred.

A.1.1 Types of Events Analyzed Two different types of events are addressed in precursor quantitative analysis. In the first, an initiating event such as a loss of offsite power (LOOP) or small-break loss of coolant accident (LOCA) occurs as a part of the precursor. The probability of core damage for this type of event is calculated based on the required plant response to the particular initiating event and other failures that may have occurred at the same time. This type of event includes the "windowed" events subsetted for the 1982-83 ASP program and discussed in Section 2.2 of the main report.

The second type of event involves a failure condition that existed over a period of time during which an initiating event could have, but did not occur. The probability of core damage is calculated based on the required plant response to a set of postulated initiating events, considering the failures that were observed. Unlike an initiating event assessment, where a particular initiating event is assumed to occur with probability 1.0, each initiating event is assumed to occur with a probability based on the initiating event frequency and the failure duration.

ASP MODELS

A-3 A.1.2 Modification of System Failure Probabilities to Reflect Observed Failures The ASP models used to evaluate 1982-83 operational events describe sequences to core damage in terms of combinations of mitigating systems success and failure following an initiating event. Each system model represents those combinations of train or component failures that will result in system failure. Failures observed during an operational event must be represented in terms of changes to one or more of the potential failures included in the system models.

If a failed component is included in one of the trains in the system model, the failure is reflected by setting the probability for the impacted train to 1.0. Redundant train failure probabilities are conditional, which allows potential common cause failures to be addressed. If the observed failure could have occurred in other similar components at the same time, then the system failure probability is increased to represent this. If the failure could not simultaneously occur in other components (for example, if a component was removed from service for preventive maintenance), then the system failure probability is also revised, but only to reflect the "removal" of the unavailable component from the model.

If a failed component is not specifically included as an event in a model, then the failure is addressed by setting elements impacted by the failure to failed. For example, support systems are not completely developed in the 1982-83 ASP models. A breaker failure that results in the loss of power to a group of components would be represented by setting the elements associated with each component in the group to failed.

Occasionally, a precursor occurs that cannot be modelled by modifying probabilities in existing system models.

In such a case, the model is revised as necessary to address the event, typically by adding events to the system model or by addressing an unusual initiating event through the use of an additional event tree.

A.1.3 Recovery from Observed Failures The models used to evaluated 1982-83 events address the potential for recovery of an entire system if the system fails. This is the same approach that was used in the analysis of most precursors through 1991.' In this approach, the potential for recovery is addressed by assigning a recovery action to each system failure and initiating event. Four classes were used to describe the different types of short-term recovery that could be involved:

1I ater precursor analyses utilize Time-Reliability Correlations to estimate the probability of failing to recover a failed system when recovery is dominated by operator action.

ASP MODELS

A-4 Recovery Likelihood of Non-Recovery Characteristic Class Recovery' RI 1.00 The failure did not appear to be recoverable in the required period, either from the control room or at the failed equipment R2 0.55 The failure appeared recoverable in the required period at the failed equipment, and the equipment was accessible; recovery from the control room did not appear possible.

R3 0.10 The failure appeared recoverable in the required period from the control room, but recovery was not routine or involved substantial operator burden.

R4 0.01 The failure appeared recoverable in the required period from the control room and was considered routine and procedurally based.

The assignment of an event to a recovery class is based on engineering judgment, which considers the specifics of each operational event and the likelihood of not recovering from the observed failure in a moderate to high stress situation following an initiating event.

Substantial time is usually available to recover a failed residual heat removal (RHR) or BWR power conversion system (PCS). For these systems, the nonrecovery probabilities listed above are overly conservative. Data in Refs. 2 and 3 was used to estimate the following nonrecovery probabilities for these systems:

System p(nonrecovery)

BWR RHR system 0.0 16 (0.054 if failures involve service water)

BWR PCS 0.52 (0.017 for MSIV closure)

PWR RHR system 0.057 It must be noted that the actual likelihood of failing to recover from an event at a particular plant is difficult to assess and may vary substantially from the values listed. This difficulty is demonstrated in the genuine differences in opnion among analysts, operations and maintenance personnel, etc., concerning the likelihood of recovering specific failures (typically observed during testing) within a time period that would prevent core damage following an actual initiating event.

A.1-4 Conditional Probability Associated with Each Precursor As described earlier in this appendix, the calculation process for each precursor involves a determination of imtiators that must be modeled, plus any modifications to system probabilities necessitated by failures observed 2These nonrecovery probabilities are consistent with values specified in M.B. Sattison et al., "Methods Improvements Incorporated into the SAPHIRE ASP Models," Proceedings of the U.S. Nuclear Regulatory Commission Twnry-Second Water Reactor Safety Information Meeting, NUREG/CP-0140, Vol. 1, April 1995.

ASP MODELS

A-5 in an operational event. Once the probabilities that reflect the conditions of the precursor are established, the sequences leading to core damage are calculated to estimate the conditional probability for the precursor. This calculational process is summarized in Table A. 1.

Several simplified examples that illustrate the basics of precursor calculational process follow. It is not the intent of the examples to describe a detailed precursor analysis, but instead to provide a basic understanding of the process.

The hypothetical core damage model for these examples, shown in Fig. A. 1, consists of initiator I and four systems that provide protection against core damage: system A, B, C, and D. In Fig. A. 1, the up branch represents success and the down branch failure for each of the systems. Three sequences result in core damage if completed: sequence 3 [I /A ("/" represents system success) B C], sequence 6 (IA /BC D) and sequence 7 (I A B). In a conventional PRA approach, the frequency of core damage would be calculated using the frequency of the initiating event I, 1(I), and the failure probabilities for A, B, C, and D [p(A), p(B), p(C), and p(D)J.

Assuming (I) = 0.1 yr' and p(AjI) = 0.003, p(BIIA) = 0.01, p(CII) = 0.05, and p(DjIC) = 0. 1,' the frequency of core damage is determined by calculating the frequency of each of the three core damage sequences and adding the frequencies:

0.1 yr-' x (1 - 0.003) x 0:05 x 0.1 (sequence 3) +

0.1 yr-' x 0.003 x (1 - 0.01) x 0.05 x 0.1 (sequence 6) +

0.1 yr' x 0.003 x 0.01 (sequence 7)

=4.99 x 10 yr' (sequence 3) + 1.49 x 10 yr' (sequence 6) + 3.00 x 10' yr' (sequence 7)

=5.03 x 10' yr'.

In a nominal PRA, sequence 3 would be the dominant core damage sequence.

The ASP program calculates a conditional probability of core damage, given an initiating event or component failures. This probability is different than the frequency calculated above and cannot be directly compared with it.

Example 1.- Initiating Event Assessment. Assume that a precursor involving initiating event I occurs. In response to I, systems A, B, and C start and operate correctly and system D is not demanded. In a precursor initiating event assessment, the probability of I is set to 1.0. Although systems A, B, and C were successful, nominal failure probabilities are assumed. Since system D was not demanded, a nominal failure probability is assumed for it as well. The conditional probability of core damage associated with precursor I is calculated by summing the conditional probabilities for the three sequences:

1.0 x (1 - 0.003) x 0.05 x 0.1 (sequence 3) +

1.0 x 0.003 x (1-0.010) x 0.05 x 0.1 (sequence 6)+

1.0 x 0.003 x 0.01 (sequence 7) 3 The notation p(B IA) means the probability that B fails, given I occurred and A failed.

ASP MODELS

A-6

=5.03 x 10".

1 instead, B had failed when demanded, its probability would have been set to 1.0. The conditional core damage probability for precursor IB would be calculated as 1.0 x (1 - 0.003) x 0.05 x 0.1 (sequence 3) + 1.0 x 0.003 x 1.0 (sequence 7) = 7.99 x 10'.

Since B is failed sequence 6 cannot occur.

Examle 2. Condition Assessment. Assume that during a monthly test system B is found to be failed, and that the failure could have occurred at any time during the month. The best estimate for the duration of the failure is one half of the test period, or 360 h. To estimate the probability of initiating event I during the 360 h period, the yearly frequency of I must be converted to an hourly rate. If I can only occur at power, and the plant is at power for 70% of a year, then the frequency for I is estimated to be 0.1 yr-'/(8760 h/yr x 0.7) = 1.63 x 10-s h-1.

If as in example 1, B is always demanded following I, the probability of I in the 360 h period is the probability that at least one I occurs (since the failure of B will then be discovered), or 1 -e-IM'"'""re dh1St"m= 1-e-1.6 3E-x36

= 5.85 x 10-.

Using this value for the probability of I, and setting p(B) = 1.0, the conditional probability of core damage for precursor B is calculated by again summing the conditional probabilities for the core damage sequences in Fig.

A.1:

5.85 x 10-3 x (1 - 0.003) x 0.05 x 0.1 (sequence 3) + 5.85 x 10' x 0.003 x 1.0 (sequence 7)

=4.67 x 10-'.

As before, since B is failed, sequence 6 cannot occur. The conditional probability is the probability of core damage in the 360 h period, given the failure of B. Note that the dominant core damage sequence is sequence 3, with a conditional probability of 2.92 x 10-s. This sequence is unrelated to the failure of B. The potential failure of systems C and D over the 360 h period still drive the core damage risk To understand the significance of the failure of system B, another calculation, an importance measure, is required.

The importance measure that is used is equivalent to risk achievement worth on an interval scale (see Ref 4).

In this calculation, the increase in core damage probability over the 360 h period due to the failure of B is estimated: p(cd I B) - p(cd). For this example the value is 4.67 x 10-s - 2.94 x 1(rI = 1.73 x 10s, where the second term on the left side of the equation is calculated using the previously developed probability of I in the 360 h period and nominal failure probabilities for A, B, C, and D.

For most conditions identified as precursors in the ASP program, the importance and the conditional core damage probability are numerically close, and either can be used as a significance measure for the precursor. However.

for some events--typically those in which the components that are failed are not the primary mitigating plant features--the conditional core damage probability can be significantly higher than the importance. In such cases, it is important to note that the potential failure of other components, unrelated to the precursor, are still dominating the plant risk.

ASP MODELS

A-7 The importance measure for unavailabilities (condition assessments) like this example event were previously referred to as a "conditional core damage probability" in annual precursor reports before 1994, instead of as the increase in core damage probability over the duration of the unavailability. Because the computer code used to analyze 1982-83 events is the same as was used for 1984-93 evaluations, the results for 1982-83 conditions are also presented in the computer output in terms of "conditional probability," when in actuality the result is an mportance.

A.2 Overview of 1982-83 ASP Models Models used to rank 1982-83 precursors as to significance consist of system-based plant-class event trees and simplified plant-specific system models. These models describe mitigation sequences for the following initiating events: a nonspecific reactor trip [which includes loss of feedwater (LOFW) within the model], LOOP, small break LOCA, and steam generator tube rupture [SGTR, pressurized water reactors (PWRs) only].

Plant classes were defined based on the use of similar systems in providing protective functions in response to transients, LOOPs, and small-break LOCAs. System designs and specific nomenclature may differ among plants included in a particular class; but functionally, they are similar in response. Plants where certain mitigating systems do not exist, but which are largely analogous in their initiator response, are grouped into the appropriate plant class. ASP plant categorization is described in the following section.

The event trees consider two end states: success (OK), in which core cooling exists, and core damage (CD), in which adequate core cooling is believed not to exist. In the ASP models, core damage is assumed to occur following core uncovery. It is acknowledged that clad and fuel damage will occur at later times, depending on the criteria used to defmne "damage," and that time may be available to recover core cooling once core uncovery occurs but before the onset of core damage. However, this potential recovery is not addressed in the models. Each event tree describes combinations of system failures that will prevent core cooling, and makeup if required, in both the short and long tarn. Primary systems designed to provide these functions and alternate systems capable of also performing these functions are addressed.

The models used to evaluate 1982-83 events consider both additional systems that can provide core protection and initiating events not included in the plant-class models used in the assessment of 1984-91 events, and only partially included in the assessment of 1992-93 events. Response to a failure to trip the reactor is now addressed, as is an SGTR in PWRs. In PWRs, the potential use of the residual heat removal system following a small-break LOCA (to avoid sump recirculation) is addressed, as is the potential recovery of secondary-side cooling in the long term following the initiation of feed and bleed. In boiling water reactors (BWRs), the potential use of reactor core isolation cooling (RCIC) and the control rod drive (CRD) system for makeup if a single relief valve sticks open is addressed, as is the potential long-term recovery of the power conversion system (PCS) for decay heat removal in BWRs. These models better reflect the capabilities of plant systems in preventing core damage.

ASP MODELS

A-7 The importance measure for unavailabilities (condition assessments) like this example event were previously referred to as a "conditional core damage probability" in annual precursor reports before 1994, instead of as the increase in core damage probability over the duration of the unavailability. Because the computer code used to analyze 1982-83 events is the same as was used for 1984-93 evaluations, the results for 1982-83 conditions are also presented in the computer output in terms of "conditional probability," when in actuality the result is au importance.

A.2 Overview of 1982-83 ASP Models Models used to rank 1982-83 precursors as to significance consist of system-based plant-class event trees and simplified plant-specific system models. These models describe mitigation sequences for the following initiating events: a nonspecific reactor trip [which includes loss of feedwater (LOFW) within the model], LOOP, small break LOCA, and steam generator tube rupture [SGTR, pressurized water reactors (PWRs) only].

Plant classes were defined based on the use of similar systems in providing protective functions in response to transients, LOOPs, and small-break LOCAs. Systen designs and specific nomenclature may differ among plants included in a particular class; but functionally, they are similar in response. Plants where certain mitigating systems do not exist, but which are largely analogous in their initiator response, are grouped into the appropriate plant class. ASP plant categorization is described in the following section.

The event trees consider two end states: success (OK), in which core cooling exists, and core damage (CD), in which adequate core cooling is believed not to exist. In the ASP models, core damage is assumed to occur following core uncovery. It is acknowledged that clad and fuel damage will occur at later times, depending on the criteria used to defmn "damage," and that time may be available to recover core cooling once core uncovery occurs but before the onset of cre damage. However, this potential recovery is not addressed in the models. Each event tree describes combinations of system failures that will prevent core cooling, and makeup if required, in both the short and long term. Primary systems designed to provide these functions and alternate systems capable of also performing these finctions are addressed.

The models used to evaluate 1982-83 events consider both additional systems that can provide core protection and indating events not included in the plant-class models used in the assessment of 1984-91 events, and only partially included in te asessment of 1992-93 events. Response to a failure to trip the reactor is now addressed, as is an SGTR in PWRs. In PWRs, the potential use of the residual heat removal system following a small-break LOCA (to avoid sump recirculation) is addressed, as is the potential recovery of secondary-side cooling in the long term following the initiation of feed and bleed. In boiling water reactors (BWRs), the potential use of reactor core isolation cooling (RCIC) and the control rod drive (CRD) system for makeup if a single relief valve sticks open is addressed, as is the potential long-term recovery of the power conversion system (PCS) for decay heat removal in BWRs. These models better reflect the capabilities of plant systems in preventing core damage.

ASP MODELS