ML030800607

Results of the Three Mile Island Unit 1 SDP Phase 2 Notebook Benchmarking Visit

ML030800607
Person / Time
Site:	Three Mile Island
Issue date:	03/21/2003
From:	Reinhart F NRC/NRR/DSSA/SPSB
To:	Carpenter C NRC/NRR/DIPM/IIPB
Wilson P, NRR/SPSB, 415-1114
References
Download: ML030800607 (17)

Text

March 21, 2003 NOTE TO: Cynthia Carpenter, Chief Inspection Program Branch Division of Inspection Program Management Office of Nuclear Reactor Regulation Patrick D. OReilly Operating Experience Risk Applications Branch Division of Risk Analysis and Applications Office of Nuclear Regulatory Research FROM: Mark F. Reinhart, Chief/RA/

Licensing Section Probabilistic Safety Assessment Branch Division of Systems Safety and Analysis Office of Nuclear Reactor Regulation

SUBJECT:

RESULTS OF THE THREE MILE ISLAND UNIT 1 SDP PHASE 2 NOTEBOOK BENCHMARKING VISIT During October, 2002, NRC staff and contractors visited the Exelon East corporate office in Kennet Square, Pennsylvania to compare the Three Mile Island (TMI), Unit 1 Significance Determination Process (SDP) Phase 2 notebook and licensees risk model results to ensure that the SDP notebook was generally conservative. TMIs PSA did not include external initiating events so no sensitivity studies were performed to assess the impact of these initiators on SDP color determinations. In addition, the results from analyses using the NRCs draft Revision 3i Standard Plant Analysis Risk (SPAR) model for TMI were also compared with the licensees risk model. The results of the SPAR model benchmarking effort will be documented in next revision of the SPAR (revision 3) model documentation.

The benchmarking visit identified that there was good correlation between the Phase 2 SDP Notebook and the licensees PSA. The results indicate that the TMI Phase 2 notebook was generally more conservative in comparison to the licensees PSA. The SDP Rev. 1 notebook when issued should provide similar or slightly more conservative results than the licensees PSA in 89% of the cases.

CONTACT: Peter Wilson, SPSB/DSSA/NRR 301-415-1114

C. Carpenter 2 A summary of the results of comparisons of hypothetical inspection findings between SDP notebook and the licensees PSA are as follows.

11% Notebook predicted the risk an order of magnitude less than the licensees PSA Significance 71% Notebook and licensees PSA results matched within an order of magnitude 13% Notebook predicted the risk an order of magnitude greater than the licensees PSA.

5% Notebook predicted the risk two orders of magnitude greater than the licensees PSA.

The Rev-1 SDP notebook has been greatly improved as a result of the benchmarking activity.

Number of underestimates was significantly reduced (from 15 to 5). Number of cases that Rev-1 SDP would match that of the updated licensees PSA has increased from 20 to 32.

Finally, some reduction is gained for the number of overestimates.

The licensees PSA staff was very knowledgeable of the plant model and provided very helpful comments during the benchmark visit.

Attachment A describes the process and results of the comparison of the TMI SDP Phase 2 Notebook and the licensees PSA.

Attachments: As stated

C. Carpenter 2 A summary of the results of comparisons of hypothetical inspection findings between SDP notebook and the licensees PSA are as follows.

11% Notebook predicted the risk an order of magnitude less than the licensees PSA Significance 71% Notebook and licensees PSA results matched within an order of magnitude 13% Notebook predicted the risk an order of magnitude greater than the licensees PSA.

5% Notebook predicted the risk two orders of magnitude greater than the licensees PSA.

The Rev-1 SDP notebook has been greatly improved as a result of the benchmarking activity.

Number of underestimates was significantly reduced (from 15 to 5). Number of cases that Rev-1 SDP would match that of the updated licensees PSA has increased from 20 to 32.

Finally, some reduction is gained for the number of overestimates.

The licensees PSA staff was very knowledgeable of the plant model and provided very helpful comments during the benchmark visit.

Attachment A describes the process and results of the comparison of the TMI SDP Phase 2 Notebook and the licensees PSA.

Attachments: As stated Accession # ML030800607 G:\SPSB\wilson\TMIBENCH.wpd OFFICE SPSB SC:SPSB SPSB:RGN-1 NAME PWilson:nxh2 FMReinhart/RA/ ECobey DATE 03/20/03 03/21/03 03/20/03 OFFICIAL RECORD COPY

SUMMARY

REPORT ON BENCHMARKING TRIP TO THE THREE MILE ISLAND NUCLEAR GENERATING STATION UNIT 1 Mohamad A. Azarm (BNL)

Energy Sciences and Technology Department Brookhaven National Laboratory Upton, N.Y. 11973-5000 November 2002 Attachment A

Table of Contents Page

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2. Summary Results From Benchmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

3. Proposed Revisions to Rev. 0 SDP Notebook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 3.1 Specific Changes to the Rev. 0 SDP Notebook for Three Mile Island Unit 1 . . 10 3.2 Generic Changes in IMC 0609 for Guidance to NRC Inspectors . . . . . . . . . . . 12 3.3 Generic Changes to the SDP Notebook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.3.1 Generic Insights for SDP Evaluation Process . . . . . . . . . . . . . . . . . . 12 3.3.2 Generic Insights for B&W Plants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

4. Discussion on External Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5. List of Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 List of Tables Page Table 1. Summary of Benchmarking Results for Three Mile Island Unit 1 . . . . . . . . . . . . 4 Table 2. Comparative Summary of the Benchmarking Results . . . . . . . . . . . . . . . . . . . . 9

-iv-

1 INTRODUCTION A benchmarking meeting took place at the Exelon east headquarter at Kenneth Square Pennsylvania for Three Mile Island (TMI), unit 1 on October 27, 2002. P. Wilson and E. Cobey from NRC, along with M.A. Azarm from BNL, and R. Buell from INEEL participated in this benchmarking exercise. This benchmarking report documents the overall results and insights from the benchmarking trip.

In preparation of the meeting, BNL staff reviewed the SDP notebook for Three Mile Island, evaluated the coloring results of the Rev. 0 SDP worksheets, and collected system diagrams and information. In addition, a copy of the meeting protocol was sent to the licensee by E. Cobey of the NRC prior to the meeting.

The major milestones achieved during this meeting were as follows:

1. Obtained the Risk Achievement Worth (RAW) values for basic events for the internal event model for average maintenance based on internal event model including internal flooding (6% contribution).

2. Identified a target set for the basic events for the benchmarking exercise.

3. Performed benchmarking of a subset of the target set of basic events using the Rev. 0 SDP notebook with the licensees staff participating and providing comments on the notebooks.

4. Requested some PSA model results from the licensee to determine the dominant contributors to the RAW values and to compare with the contributors captured by the notebook.

5. Obtained updated HEPs used in the TMI PSA.

6. Obtained updated dependency information for the fluid support systems and electrical systems to facilitate evaluation of the hypothetical inspection findings as a part of the benchmarking exercise.

7. Carried out the benchmarking and on-site modifications to the SDP notebook

8. Rev-0 per licensees comments.

The utility staff provided extensive comments that were resolved and will be incorporated in the SDP Rev. 1 notebook.

The Rev. 0 SDP notebook for TMI was updated and the sequences were solved prior to the site visit based on the current SDP generic guidelines. A total of 46 hypothetical inspection findings were examined during the site visit. Table 1 lists these items along with the associated risk significance based on the RAW values from the licensees PSA and the SDP notebook.

The summary results from benchmarking are shown in Table 2. The SDP Rev. 1 notebook when issued should provide similar or slightly more conservative results than the licensees PSA in 85%

of the cases. In the 11% of cases where the SDP underestimated the PSA, the main reason was attributed to round off errors. Round off errors refer to the results of a case run when the following two conditions exist:

1. The SDP notebook captures all pertinent sequences affected and identified by PSA for the specific case run (e.g. when a given component is removed from service).

2. The resulting RAW value from the PSA for the specific case run is at proximity of the assigned color threshold.

All cases where the SDP underestimated the PSA or overestimated by two orders of magnitude were examined during the site visit and they are discussed below:

• There are dedicated cooling water systems for LPI/LPR trains through the decay heat river water system (DHRW) and decay heat closed cooling water system (DHCCW). Therefore, operation of one train of LPR would require availability of the associated train of DHRW and DHCCW in addition to the availability of the LPI pump, valves, and heat exchanger.

Therefore, the PSAs credit for one train of LPR when the other train is assumed not to be available would be less than the SDP credit of one train. Since the support systems for LPR is a unique plant feature in TMI, generic credit in SDP was not changed and the case was recorded as an underestimate.

• Round off error was identified as a main reason for the SDP notebook to underestimate the seal injection crossconnect, the operator failure to pin the seal injection valves, and the recovery of offsite power in 1 and 4 hours4.62963e-5 days <br />0.00111 hours <br />6.613757e-6 weeks <br />1.522e-6 months <br />. Examination of the PSA output for the above cases showed that the SDP rev. 1 notebook will captured all relevant sequences.

• The emergency boration is shown as a match in both Tables 1 and 2 . The PSA RAW value for this case includes sequences from other initiators besides ATWS due to the way the human errors are modeled in the PSA. The specific action for emergency boration only applies to ATWS scenarios. The equivalent contribution from ATWS scenarios was re-estimated using the licensee PSA and resulted in an increase in CDF of about 2E-6 per reactor year, consistent with the results from the SDP notebook. This case was, therefore, recorded as a match.

• For HPSI pump A or B, the SDP estimated a Red and a Yellow respectively, where the PSA assigned a White color. The RAW values from the PSA were judged to be not appropriate due to effects of various levels of CCF that are modeled in the PSA. The CCF modeling technique used in the licensees PSA has impacted the RAW values in two ways:

(1) the failure probability associated with the remaining two pumps when either pumps A or B is unavailable are estimated without CCF contribution (independently), and (2) the failure probability associated with the remaining two pumps when pump C is unavailable is estimated accounting for full CCF contribution. Therefore, this treatment has resulted in artificially lower RAW values for pumps A or B but slightly higher RAW value (approximately a factor of 3) for pump C.

• There were also three cases where the SDP notebook overestimated the PSA results by two orders of magnitude. The case of HPSI pump A was just discussed and it was shown that it is not a true overestimation by two orders of magnitude. The other two cases were the failures of a SRV or a PORV to reclose. Failure of a SRV to re-close was treated in the SDP notebook as an occurrence of SLOCA with a probability of one. Failure of a PORV to re-close was treated by the SDP notebook as an occurrence of SORV with a probability of one. The PSA assumes that the SRVs/PORVs will only be demanded in a small fraction of the initiators (less than 1 in 10 reactor years). The PSA, therefore, treated this case as a SLOCA or SORV; however, with a frequency estimated by the SRV/PORV demand probability for all initiators. This frequency was estimated to be less than 1 in 10 years.

Therefore, these cases were recorded in Table 2 as overestimates.

2.

SUMMARY

RESULTS FROM BENCHMARKING This section provides the results of the benchmarking exercise. A total of 46 hypothetical inspection findings were examined during the site visit. The results of the benchmarking analyses are summarized in Table 1. Table 1 consists of seven columns. In the first column, the out-of-service components, human actions, or recovery actions are identified for the case analyses.

The second column shows the colors assigned for significance characterization from using the Rev. 0 SDP notebook. The third column shows the RAW values based on the licensees PSA for internal event initiators. The fourth column shows the assigned colors that are expected to be obtained from the Rev. 1 SDP notebook when the licensees comments are incorporated and the report is issued. The fifth column shows the results from comparison of the third and fourth columns. The results shown in fifth column are categorized in the sixth column into three groups:

O, M, and U which stand for Overestimated, Matched, and Underestimated, respectively. If an item in this column is categorized as M, it indicates that the color that will be obtained from the SDP Rev. 1 notebook will match the risk significance color obtained from the current licensees PSA model. Items categorized by O and U should also be interpreted similarly. Finally, the last column provides some comments for clarification of the SDP evaluation process, and the underlying reasons for any differences that might have occurred. The basic event names were not available and not incorporated in the last column since this plant uses the large event tree approach. In this approach RAW calculations may require changes to be done to several basic or top events, therefore identification of one basic event name may not be possible.

Table-2 shows the summary statistics of the benchmarking results for the Rev-0 SDP notebook and what would be the expected results for the Rev-1 SDP notebook. This table shows that the Rev-1 SDP notebook has been greatly improved as a result of the benchmarking activity. Number of underestimates has significantly reduced (from 15 to 5). Number of cases that Rev-1 SDP would match that of the updated licensees PSA has increased from 20 to 32. Finally, some reduction is gained for the number of overestimates.

BNL #04334 Table 1: Summary of Benchmarking Results for Three Mile Island Unit 1 Internal Events CDF is 4.02E-5 per reactor-year; therefore, RAW thresholds are: W = 1.02, Y = 1.25 , and R = 3.49 External Events CDF is 1.4 E-4 per reactor-year Component Out-of-Service or SDP SDP Human Actions Worksheet Internal Worksheet PSA/SDP Categorization Comments (1) or Results (Before) RAW (1) Results of the Results Recovery Actions (Internal only) (After)

HPSI Pump C Y (1Y+5W) 29 R R/R M HPSI Pump A Y 1.01 R W/R O The SDP assigns a red to pumps A. The PSAs RAW are Biased by the CCF Modeling. The more appropriate color for PSA should be a yellow at minimum. This case is considered as an over estimate by one color.

HPSI Pump B Y 1.07 Y W/Y O The SDP assigns yellow to pumps B. The PSAs RAW are Biased by CCF Modeling.

LPI Pump R 8.9 Y R/Y U Under: Case Run was Examined Nov. 13, 2002 TDEFW Pump Y 1.3 Y Y/Y M MDEFW Pump Y 1.7 Y Y/Y M

Component Out-of-Service or SDP SDP Human Actions Worksheet Internal Worksheet PSA/SDP Categorization Comments (1) or Results (Before) RAW (1) Results of the Results Recovery Actions (Internal only) (After)

BNL #04334 1 MFW Pump G 1.06 W W/W M PCS (Initiator) W 1.1 W W/W M (3.91E-6)

PCS (Initiator + Mitigation) W NA(2) W NA NA EDG W 1.2 (EDG-A) W W/W M 1.1 (EDG-B)

SBO DG G 1.08 W W/W M Battery Charger R 38 R R/R M Bravo side assuming that, PCS is not lost Battery Charger Alpha side R 38 R R/R M Battery W (Take SI) 38 R R/R M (B SIDE) R(Not Take SI) 1 SRV Fails to Open (FTO) W 1.028 W W/W M PZR vent line FTO G 1.01 W G/W O PZR PORV FTO G 1.2 W W/W M Block Valve Fails to Close W 1.02 W W/W M Nov. 13, 2002 (FTC)

PSV FTC R 1.07 R W/R O Demand

Component Out-of-Service or SDP SDP Human Actions Worksheet Internal Worksheet PSA/SDP Categorization Comments (1) or Results (Before) RAW (1) Results of the Results Recovery Actions (Internal only) (After)

BNL #04334 PORV FTC Y 1.015 Y G/Y O Demand LIA (Initiator) R 19 R R/R M (8.01E-4) 1 Air Compressor W P1 (1.4) Y Y/Y M Under: Round off for P2 P2 (1.2) and match for P1 LNSRW W 35 R R/R M (initiator) (1.4E-3)

Loss of 1 NSRW pump G 2.5 Y Y/Y M LRW R ~250 R R/R M (Initiator) (2.27E-2) 1 train of RW Y 2.5 Y Y/Y M 1 DHCCW pump W 19 (A) R R/R M 11 (B) 1 DHRW pump W 12 (A) R R/R M 8.6 (B)

ADV FTO G 1.08 W W/W M 1 MSIV FTC Y(1Y+2W) 1.01 G G/G M Nov. 13, 2002 1 SSCW pump (TPCS) W 1.09 W W/W M 1 SSRW pump (TPCS) W 1.0 W G/W O

Component Out-of-Service or SDP SDP Human Actions Worksheet Internal Worksheet PSA/SDP Categorization Comments (1) or Results (Before) RAW (1) Results of the Results Recovery Actions (Internal only) (After)

BNL #04334 HPR operator action R 14.2 R R/R M FB operator action R 2.2 Y Y/Y M LPR operator action R 1.08 R R/R M (LLOCA) 14.2 (MLOCA)

RCS Cooldown and Depress. R 18.4 R R/R M (SGTR/EQ)

Emergency Boration (ATWS) W 2.2 W Y/W M The emergency (see the note boration is the same column for event as manual reasons not initiation of HPI in the being an under) PSA. Contribution from ATWS scenarios which is appropriate for this entry is about 2E-6 (W).

This case therefore is considered as a match.

RCP Trip R 3.14 R Y/R O Seal injection by X-connect W 1.26 W Y/W U Under pump C 1.94 Round Off Error Nov. 13, 2002 Operator pins seal injection W 1.36 W Y/W U Under valves (LIA) Round Off Error

Component Out-of-Service or SDP SDP Human Actions Worksheet Internal Worksheet PSA/SDP Categorization Comments (1) or Results (Before) RAW (1) Results of the Results Recovery Actions (Internal only) (After)

BNL #04334 Align fire service water to R 2.25 R Y/R O decay heat river water to restore seal injection (LRW)

Isolate SG and control SI flow Y 1.76 Y Y/Y M (FWIC)

Recovery of Power in 1 hr G 1.03 G W/G U Under (REC1) Round off Recovery of Power in 4 hrs W 1.28 W Y/W U Under (REC4) Round off Local manual control of EFW W 1.94 Y Y/Y M flow (LIA) Failure to ISOSG (SGTR) NA 1.12 W W/W M (1) The RAW values obtained are based on the internal event model, including flood (about 6% contribution). The truncation limit was about 1.E-8 within the Riskman approach with residue bin calculated. The surrogate basic event names are not provided due to difficulties in extracting such information from a large event tree approach.

C NA stands for not available or not modeled.

Nov. 13, 2002

Table 2: Comparative Summary of the Benchmarking Results Total Number of SDP Notebook SDP Notebook Cases Compared Before (Rev 0)

After (Rev 1)

Number of Percentage Number of Percentage Cases Cases (46) (46) 15 32.6 5 10.9 SDP: Less Conservative 7 (O1) 15.2 6 (O1) 13.0 SDP: More Conservative 2 (O2) 4.4 2 (O2) 4.4 20 43.4 32 69.5 SDP: Matched not modeled 2 4.4 1 2.2

3. PROPOSED REVISIONS TO REV. 0 SDP NOTEBOOK A set of modifications were proposed for the Rev. 0 SDP notebook as a result of the site visit.

These proposed modifications are driven by the licensees comments on the Rev. 0 SDP notebook, better understanding of the current plant design features, allowance for additional recovery actions, revised Human Error Probabilities (HEPs), updated frequencies of the initiating events, and the results of benchmarking.

3.1 Specific Changes to the Rev. 0 SDP Notebook for Three Mile Island The earlier version of the notebook was reviewed by the utility on May 26-27, 2000. The resolution of the utilitys comments is included in the notebook. Additional comments were received during the benchmarking site visit. These comments were reviewed and incorporated into the SDP notebook to the extent possible. The following items list major comments that were incorporated.

Table 1:

3. Add the initiator for NSCCW to Row III.

4. Remove L4KV Bus from Row III.

Table 2:

1. Add a footnote on battery duration of 2 hrs without load shedding and 6 hours6.944444e-5 days <br />0.00167 hours <br />9.920635e-6 weeks <br />2.283e-6 months <br /> with load shedding.

2. Add a footnote that PCS is automatically controlled by ICS.

3. Add a footnote for PORV size (0.939 inch squared), PSV size (2.5 inches squared), and vent line (1 inch squared).

4. Add 3 condensate booster pumps to condensate/MFW row.

5. Remove all references to L4KVBUS in the initiating event column.

6. Add a row for BWST makeup from spent fuel pool through spent fuel pumps with AC and DC as support system.

7. Add a footnote that all inspection findings associated with ICCW should be colored as green. Total loss of ICCW would behave similar to a transient.

8. Note that there are 16 air bottles for EFW flow control valves that would be sufficient for 2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br /> for backup air supply.

Table 3.1:

1. Change the credit for PCS to 1 train globally. The MFW ramps back and SG level would be maintained and controlled automatically by the Integrated Control System (ICS) which should be credited as a single train.

Table 3.4:

1. Remove PSV from the title.

Table 3.5:

1. Put a footnote that the PSA HEP value for LPR is 1.7E-3.

Table 3.6:

1. Put a footnote that the PSA HEP value for LPR is 8.7E-3.

Table 3.8:

1. Change the event tree and sequences for SGTR such that it captures the licensees major sequences.

Table 3.9:

1. Change the credit and the success criteria for MDEFW for ATWS from 1/2 to 2/2 MDEFW.

2. Change the HEP value for emergency boration to 5.2 E-3.

Table 3.10:

1. The Function RCPSEAL should be credited as one train, since it is limited by a train of seal injection.

2. Change the sequence LNSRW - RCPSEAL - EIHP to LNSRW - EIHP, since the failure of EIHP implies the failure of RCP seal. Modify the event tree to explicitly show that.

Table 3.11:

1. Add explicitly to the success criteria for SEALINJ or the operator pins the ICCW containment isolation valve.

Table 3.12:

1. Change the title to say Loss of all River Water. Show in footnote 1 that the screen plugging is a major contributor to this event.

Table 3.13:

1. Add a footnote to say that MSIV isolation is not important for preventing PTS in MSLB in TMI. State that there is no equalizing line for the steam lines. It takes about 2 minutes for MSIVs to close. As long as turbine stop valves are closed, at most one SG will be emptied.

The PTS is mainly prevented by termination of feed to SG or controlling SI flow.

2. Redraw the event tree and sequences for MSLB consistent with the licensees assumption on PTS.

Table 3.14:

3. Remove the worksheet for L4KV bus and substitute it with a worksheet for loss of NSCCW (LNSCCW). The loss of the 4KV bus does not cause an initiating event.

4. Draw the event tree for LNSCCW and substitute it for the event tree for loss of 4KV bus.

3.2 Generic Changes in IMC 0609 for Guidance to NRC Inspectors No changes were identified from this site visit.

3.3 Generic Changes to the SDP Notebook None.

3.3.1 Generic Insights for SDP Evaluation Process None.

3.3.2 Generic Insights for B&W Plants No new generic insight for B&W plants was identified in Three Mile Island.

4. DISCUSSION ON EXTERNAL EVENTS In TMI, the CDF contribution from internal events is 4.0E-5 per reactor-year, seismic 3E-5 per reactor-year, fire 2E-5 per reactor-year, high winds 8E-7 per reactor-year, and aircraft crash 4E-7 per reactor-year. The CDF contribution from internal floods is 2.5E-6 per reactor-year.

5. LIST OF PARTICIPANTS P. Wilson, NRC/NRR G. Cobey, SRA/Region I M.A. Azarm, BNL R. Buel, INEEL C. Pupek, Exelon C. Adams, Exelon