ML20216J921
| ML20216J921 | |
| Person / Time | |
|---|---|
| Site: | Davis Besse  |
| Issue date: | 09/12/1997 |
| From: | Ray P NRC (Affiliation Not Assigned) |
| To: | NRC (Affiliation Not Assigned) |
| References | |
| NUDOCS 9709180152 | |
| Download: ML20216J921 (20) | |
Text
_-
September 12, 1997 LICENSEE: Toledo Edison Company
, FACILITY: Davis.Besse Nuclear Power Station
SUBJECT:
SUMMARY
OF AUGUST 21. 1997. MEETING ON INSTRUMENT DRIFT STUDY On August El 1997. NRC staff members met in Rockville. Maryland, with
- representatives of Toledo Edison Company. The purpose of the meeting was for Toledo Edison Company to discuss their response to the NRC's request for additional information (RAI) dated June 11. 1997, regarding license amendment requests (LAR) 95-27, 95-24. and 96-14. A list of the meeting participants is included as Attachment 1. and a copy of the meeting handouts is provided as Attachment 2.
The licensee opened the meeting with a detailed step by-step discussion of the Davis-Besse instrument drift dtta analysis methodology using reactor protection system reactor coolant flow measurement as an example. The methodology procedure and the exam Attachments 1 and 2. respectively ple 96-14 in LAR using dated the procedure are listed April 18, 1997. The as meeting continued with a discussion of the remainder of the RAI responses as documented in Attu.hment 2.
The meeting closed with the licensee indicating a supalement to LAR 95-24 would be submitted to clarify some minor issues that lave no impact nn the licensee's basis for the amendment-Original signed byi Phillip H. Ray. Project Manager Project Directorate III-3 Division of Reactor Projects III/IV Office of Nuclear Reactor Regulation Docket No. 50-346 Attachments: 1. List of Meeting Participants
- 2. Meeting Handouts cc w/atts: See next page j ,
DISTRIBUTION: See next page bb!hf3 OOfflf DOCUMENT NAME: G:\DAVISBES\DB98520.MTS ,
OFFICE PD3-3:Pti / l PD3-3:LA_ ,lAl PD3-3:PM _lE NAME Pray #/ E CBoyle (A6 AHansen M//
DATE 09//f/97 1 09//b /97) 09/ e /9Y
,, o n OFFICIAL Rt.COR3 COPY P M Pe h P PDR ..
l I.l lllll!l.ill.l!I.lil
Davis-Besse Nuclear Power Station, Unit No. 1 Toledo Edison Company cc:
Mary E. O'Reilly Robert E. Owen, Chief Centerior Energy Corporation Bureau of Radiological Health 300 Madison Avenue Service Toledo, OH 43652 Ohio Department of Health P.O. Box 118 Columbus, OH 43266-0118 James L. Freels Manager - Regulatory Affairs Attorney General .
Toledo Edison Company Department of Attorney General Davis-Besse Nuclear Power Station 30 East Broad Street 5501 North State - Route 2 Columbus, OH 43216 Oak Harbor, OH 43449-9760 Donna Owens, Director Gerald Charnoff, Esq. Ohio Department of Commerce Shaw, Pittman, Potts Division of Industrial Compliance
& Trowbridge Bureau or Operations and Maintenance 2300 N Street, NW. 6606 Tussing Road Washington, DC 20037 P.O. Box 4009 Reynoldsburg, OH 43068-9009 Regional Administratcr U.S. NRC, Region III Ohio Environmental Protection Agency 801 Warrenville Road DERR--Compliance Unit Lisle, IL 60523-4351 ATTN: Zack A. Clayton P.O. Box 1049 Robert B. Borsum Columbus, OH 43266-0149 Babcock & Wilcox Nuclear Power Generation Division State of Ohio 1700 Rockville Pike, Suite 525 Public Utilities Commission Rockville, MD 20852 180 East Broad Street Columbus, OH 43266-0573 Resident Inspector U.S. Nuclear Regulatory Commission James R. Williams 5503 North State Route 2 Chief of Staff Oak Harbor, OH 43449 Ohio Emergency Management Agency 2855 West Dublin Granville Road James H. Lash, Plant Manager Columbus, OH 43235-2206 Toledo Edison Company Davis-Besse Company John K. Wood 5501 North State Route 2 Vice-President-Nuclear, Davis-Besse Oak Harbor, OH 43449-9760 Centerior Service Company c/o Toledo Edison Company President, Board of County Davis-Besse Nuclear Power Station Commissioners of Ottawa County 5501 North State Route 2 Port Clinton, OH 43452 Oak Harbor, OH 43449-9760 Roy P, Lessy, Jr.
Akin, Gump, Strauss, Hauer & Feld, LLP 1333 New Hampshire Ave., NW., Ste. 400 Washington, DC 20036
.<u e
-1 DISTRIBUTION!FOR
SUMMARY
OF AUGUST 21. 1997, MEETING WITH TOLEDO EDISON ,
' COMPANY- ,
4 arf Fnvr w/atts:- !
hie; D le PD 111-3 Reading ;
PUBLIC- i GGrant,;RIIIs .
- AHansen ,
OGC :
. E-mail w/att 1:
SCollins-(SJC1) .'
FMiraglia:(FJM)
GTracy (GMT)
RZimmerman (RPZ)
EAdensam (EGA1)~
GMarcus (GHM)
- Dross (e mail SLM3)
Pray (PMR)
HGarg (HCG)-
-CLurie (DXL)
CDoutt CKD) i 4
1 1
~
t 9
t-
'A va.r-d,-e-. v s. -a,- , n 9 ,
, e , - , ,: eJ. .. a - , 4 ..-.. . . . , -. - - . - -- -
MEETING ATTENDEES NRC AND TOLEDO. EDISON COMPANY DISCUSSION OF INSTRUMENT DRIFT STUDY AUGUST El, 1997 E
Phillis Ray
- Hukam Garg Dan Lurie Cliff Doutt Toledo Edison Comoany Michael Leisure David Hooten Craig Gale Attachment 1
REQUEST FOR ADDITIONAL INFORMATION REGARDING LICENSE AMENDMENT REOUESTS (LARs) 9527.952L AND 96-14 DAVIS.BFREE NUCI F AR POWER CTATION. UNIT 1 (1) Please provide a discussion of the lastrument setpoint onethodology, including the following items: ,
l'
- a. Framatome document for instrument string error; Framatome (FTI) document 32 1172392-02," Reactor Protection System String Error Calculations", calculated the RPS string errors with revised error terms resulting from the drift study to support the transition from 18 to 24 month fuel cycles. He introduction to the document states, *FTl has reviewed the string error calculation methodology against the guidance provided in Regulatory Guide 1.105 (Reference 2) and instrun. cat Society of America Standard ISA 67.04 (Reference 3) and has determined that it is within the general guidance provided by these documents." Toledo Edison had requested that FTI perform this work in accordance with ISA 567.04, Part 1 1994 and ISA RP67.04, Part 11 1994. His was motivated by an Electrical and Controls Design Engineering document," Desktop Guide for Nuclear Safety and Non Nuclear Safety Related Instrument Setpoints", which states,"Although Toledo Edison has not committed to any standa:d, it has been decided to use the most current approved version of the standard (that is, of/SA S67.04) as the pdmary source ofinformation." Since the revised string error calculations were used in the determination of allowable values only (not trip setpoints), Method 1 and Method 2 from ISA RP67.04 are equkalent, and that was the approach taken. Toledo Edison reviewed the string error calculatkus and determined that they were performed in accordance with ISA S67.04, Part ! - 1994 and ISA RP67.04, Part 11 1994, as requested. ne uncertainties not accounted for in the FTl calculations were drift, setting tolerances, and M&TE inaccuracies associated with calibration of that portion of each instrument channel included within the scope of channel functional testing. nese uncertainties are accounted for in the determination of the field setpoints, which are not part of the Technical Specifications and are controlled by Toledo Edison. Since the allowable values were calculated for channel functional testing, which, except for the containment pressure functional unit, does not include sensors, all sensor uncertainties, except those for the containment pressure switches, were included in the string error calculations.
Framatome document 12 1257719-02," Davis-Besse Unit 1 RPS Setpoint Allowable Values Calculation", then utilized the results of the string error calculations to calculate the channel functional test allowable values. His was done in accordance with BAW 10179P-A, Rev.1
" Safety Criteria and Methodology for Acceptable Cycle Reload Analyses", with the exception of the variable low pressure trip allowable value (VLPTAV). He calculation of the VLPTAV was changed because of implementation of statistical core design (SCD), as described in Appendix D of BAW 10179P A. The non-SCD method described in BAW 10179P-A uses the "e n ," factor to account for instrument string errors, including both temperature and pressure components. When using SCD, assumed instmment string errors are incorporated into the determination of the pressure-temperature design limits. Rus, if the actual instrument string errors are less than the assumed values incorporated into the SCD analysis, thew e , factor need not be considered. His was the case for the VLPTAV calculations performed. He allowable values calculations are also consistent with ISA S67.04, Part I - 1994 and ISA RP67.04, Part 11 - 1994, as requested by Toledo Edison.
- h. Allowable value calcdation, especially the use of two allowable values for the same parameter; and Attachment 2
. De NRC staff withdrew this question, as confirmed during a July 11,1997 telephone call between A. O, llansen (NRC/NRR/DB Project Manager) and D. R. Wuokko (TF/NRAU/ Supervisor).
- c. Conformance to Regulatory Guide 1.105.
De current versiori of this Regulatory Guide, Revision 2, endorses ISA S67.04 - 1982. Toledo Edison has reviewed the calculations supporting the allowable values for LARs95-024 and 96-014, and has identified only one aspect ofISA S67.04 1982 that is not complied with. Section 4.3.1.(2).(a) states that accuracy of test equipment for measuring setpoints shall be included among the allowances between the allowable value and the safety limit. De 1994 version includes this in the allowance between the allowable value and the trip setpoint. For the RPS, these uncertainties are small enough that, had they been included between the allowable value and the safety limit, the resultant allowable values would be unchinged. For the SFAS and the SFRCS, these uncertainties ,
are small enough that, had they been included between the allowable value and the safety limit, the resultant allowable values would not be impacted significantly.
(2) Please provide details on the acerptability of the Instrument drift exceeding the design basis / reference unecrtainty during an 18-month fuel cycle, and how this will be controlled during a 24-month fuel cycle.
Tbr .lesign basis / reference uncertainty is the appropriate combination of drift, reference accurney, and MA TE uncertainty to which the 30 month projected drift is compared (see items 15b,15c,15e,15t, and 16c in " Instrument Drift Data Analysis Methodology and Assumptiont", which is Attachment I to LAR 95 024). If the 95/95% tolerance interval for as found minus as left historical data exceeds the design basis / reference uncertainty, it should not be concluded that past instrument performance has been unacceptable. De acceptability of past instrument performance is demonstrated by verifying that the as-found minus as left historical data has not exceeded the design basis / reference uncertainty except on rare occasions and, where applicable, by verifying that as-found data has not exceeded its Technical Specifications allowable value except on rare occasions (see item 15d in " Instrument Drift Data Analysis Methodology and Assumptions"), which is how Toledo Edison has complied with NRC Generic Letter 91-04, Enclosure 2, issue 1. In a number ofinstances, either all or almost all of the as-found minus as-left historical data was bounded by the design basis / reference imcertainty while the 95/95% tolerance interval for that data exceeded it. His is due to the relatively small populations of drift data available in many cases and is one reason that Toledo Edison believes the results of the drift study are generally conservative.
In the example provided (Attachment 2 to LAR 95 024), historical differential pressure transmitter drift exceeded its design basis / reference uncertainty often enough that its correctness was questioned.
It was subsequent'y determined that the drift value previously used in the string error calculation was not large enough. De revised string error calculation utilizes the value obtained from this drift study, which bounds all the historical drift values. De point illustrated here is that acceptability of historical instrument drift is evaluated by comparing indiv&uct as-found minus as-left data points (not their 95/95% tolerance interval) to the design basis / reference uncertainty.
As to how this will be controlled during a 24-month fuel cycle, the answer is essentially that provided for Generic Letter 91-04, Enclosure 2, issue 7 (see LAR 95 024, Page 15).
(3) Since some of the instruments will be calibrated at 18-month intervals, calibration for these instruments will have to be done on-line. Discuss the procedural aspects of the on-line calibration, especially addressing any effects on plant safety as well as potential plant transients and/or trips.
De Technical Specifications instruments that will be calibrated on-line at 18-month intervals are already being calibrated on-line. Existing procedures ensure that these activities are performed
7
+
without undue risk of compromising plant safety or ofinducing a plant transient and'or trip.
Performing these surveillance ter.:s on line is in no way related to the Technical Specifications changes requested by LARs95-024,95-027, and 96-014.
(4) When a sensor requires calibration, the calibration uses at least seven readings along the instrument span Since these points are not independent, the same calibration check is repeated.
Thus,it appears that the number of points used in some statistical analyses was significantly inflated. Any analysis that uses more than one point from each lastrument at each calibration check is not valid, For eammple,in LAR 95-024, testing normality (Attachment 6) uses 154 readings when there are only 22 sensor tests conducted. Please discuss.
When determining basic statistics (e g., sample mean, sample standard deviation,95/95% tolerance interval, etc.), the number of data points was never inflated in the sense described above, This is .
prohibited by items $b,6, and 7a in " Instrument DriA Data Analysis Methodology and As:umptions".
When verifying the assumption that drif data is normally distributed, the normality assessment tools (see item 12 in " Instrument Drin Data Analysis Methodology and Assumptions") were used on the sum of all data points and/or on individual data points (i.e., one column of data from the spreadsheet).
With *1.e srnall sample sites involved, the W ten often resulted in the assumption of nemality not being rejected. Since it was fairly easy to obtain this result for small sample sins, Toledo Edison reasoned that performing a D' test on the sum of all data points (which, generally, was harder to " pass" than a W test on a small sample) could provide additional support to the assumption of a normal distribution. In most cases, however, the worst case individual data point was assessed. 7%nly exceptions are the linear bridge modules and signal converter modules in the RPS RC tc m ture strings.
For the linear bridge modules, the D' test was performed on the sum of all three data points (0%,50%,
and 100% of span), and it rejected the assumption that the data is normally distributed. The histogram that plotted number of drift data points versus drin also utilized all the dats (0%,50%, and 100% of span). It had a large peak in the center, with greater than 95% of the data within two standard deviations of the mean. Given the degree to which the basic statistics for all three data points are similar, there is little doubt that a histogram for the worst case individual data point (0% of span) would have also demonstrated that de assumption of normality is bounding; however, it doesn't really matter in this case because the 95/95% tolerance interval ofic.04% span was not used in subsequent analyses. Instead, to be conservative, the 30-month projected drift was chosen to be 10.10% span since this value bounds all the data on the drift versus time since last adjustment plots (which extend out to 90 months).
For the signal converter modules, the D' test was performed on the sum of all three data points, and it rejected the assumption that the data is normally distributed. The histogram that plotted number of drift data points versus drin also utilized all the data, it was quite highly peaked in the center and clearly showed that the assumption of normality is bounding, with 96.8% of the data falling within two standard deviations of the mean A D' test has subsequently been performed and a histogram created for the worst case ladividual data point (50% of span), and they confum the earlier conclusion. Once again, the D' test (RAI Attachment 1) rejected the assumption that the data is normally distributed, and the histogram (RAI Attachment 2) was quite highly peaked in the center, with 96.6% of the d.ta falling within two standard deviations of the mean Based on the time dependency analysis, the data was extrapolated to 30 months via the square root method. For the extrapolated data set, the same approach for assessing normality was taken as for the original data Let. The D' test (performed on all the data) rejected the assumption of normality, and the histogram (utilizing all thu data) was quite highly peaked, with 93.6% of the data falling within two standard deviations of the mean, it was concluded that although the assumption of normality was slightly non-conservative, it was more than made up for by other conservative assumptions (see Attachment 3, Enclosure 3 of LAR 95-024). A D' test has subsequently been performed and a histogram created for the worst case individual data point
. . . _ . - __ ~ - -
. (50% of span), and they too confirm the earlier conclusion. Once again, the D' test (RAI Attachment
- 3) rejected the assumption of normality, and the histogram (RAI Attachment 4) was quite highly peaked in the center, with 93.3% of the data falling within two standard deviations of the mean.
(5) Charts with duplicate readings (sprh as Attachment 8) are not printed clearly. Two readings at the same calibration interval could represent different sensors or repetitions of the same sensor.
Please clarify.
He data used to generate the plots chown on Attachments 8 and 9 are available on the last four pages of Attachinent 5, titled!" Months Since Last Test"," Percent Drift Since Last Test"," Months Since Last:
Adjustment", and " Percent Drift Since Last Adjustment", respectively, ne data used to generate the plots shown on Attachments 17 and 18 are available on the last four pages of Attachment 14, with the same four titles. His data can be used to reproduce these plots, if desired. ,
Also, per the NRC stafi's request during the July 10,1997, conference call with Toledo Edison, plots similar to Attachments 8,9,17, and 18 for the worst case individual data point have been created (RAI Attachments 5 through 8). Deir inclusion does not alter the conclusions reached regarding drift time dependency for either the RC flow transmitters or the 1/E converters (see Attachment 2, pages 2 through 4 of LAR 95 024). In general, Toledo Edison believes that these types of plots are more useful for drift time dependency analysis when they include all data points not just the worst case individual data point. His is especially true for small sample sizes, which were encountered often during the course of the drift study. Dere were a number of cases where as few as half a oozen or so drift values were available for each data point in the calibration. In these situations, plouir.g drift versus time since last test and drift versus time since last adjustment for the worst case individual data point does not piovide as much insight into drift time dependency as does making these plots for the sum of all data points. His is why item 13 in " Instrument Drift Data A nalysis Methodology and Assumptions" does not require plots be made for the worst case individual data point - it merely lists it as an option.
(6) The use of time since last enlibration for the purpose of predicting drift appears to be inappropriate. Such analysis ignores sensor tests that " failed"(that is, required recalibration) at some intermediate step, and so projects an over-optimistie view of the lastrument drift. The inclusion of time since last calibration could confuse the reader. Please address this issue.
Toledo Edison agrees completely with the concem expressed here (see item 13b in " Instrument Drift Data Analysis Methodology and Assumptions"). However, the mere creation of drift versus time since last adjustment plots is not inappropriate. Proper inte pretation of these plots is the crucial point in determining the appropriateness of their use. It is important to carefully read the text related to these plots (see Attachment 2, pages 2 through 4 of LAR 95-024 for an example) when assessing the appropriateness of their use. Toledo Edison believes that the interpretations of these types of plots utilizeo in the drift study were appropriate and sufliciently conservative.
(7) Even if drift is time-independent, the uncertainty associated with the drift is time-dependent.
Attachments 8,9, and 17 show that as the time between tests increases, so does the spread of the data. Please addresses this issue.
Regarding the RC flow transmitters (Attachments 8 and 9), Toledo Edison agrees 6 hat the drift b time dependent, and this is what is stated in Attachment 2, page 2 of LAR 95-024. Dat is why the drift versus time since last test data was extrapolated to a 30-month interval (using the square root method).
Regarding the 1/E converters (Attachment 17), Toledo Edison maintains the position that the drift is not time dependent. Attachment 17 shows that the drift data spread for test intervals between 17 and 19 months is about the same as for test intervals between 15 and 16 months. Attachments 18 and 19 show that the drift data spread for adjustment intervals between 30 and 40 months is slightly bener
- . - - - . - - . . . - . - - . - - - -- - - . . - - . . _ _ ~.-- -
. s
[
.. than that for adjustment intervals between 15 and 20 months. De concem raised in the preceding question is not applicable here because only one recalibration was performed among the data for ;
adjustment intervals between 15 and 20 months (and in that case the l&C technician could have lea it .
. as it was becaux the as-found data was within the calibration tolerance). Derefore, exclusion of data -;
due to " failed" testa did not result in an over optimistic view of 1/E converter driR, it should also be _
noted that the 95/95% tolerance interval maximum of 0.15% span was not exceeded by any of the data on either Attachment 17 or 13. ;
(8) Two tests were need for normality. The first is the W test (for as;S0) or the D' test (for akS0).
'The seeded uses "blas" to stratify the data. Whereas the first test is ehjeetive,'abe second test- - --
- appears to be arbitrary (it depends on how the bias are defleed) and laseasitive to departure from normality.
The use of both tests is a concera.' If the two tests agree, then the second test is redundant.
However,if the two tests do not agree, then the bianlag method should be discarded. When the results conflict, the able" method is relied on. Please discuss.
i ne W and D' tests were the only ones used to test for normality The histograms (using " bins") were
- merely supplemental to the W and D' tests (see item 12b in " Instrument Drin Data Analysis Methodology and Assumptwns"). If the W and'or D' test rejeced the assumption of norma"ty, then the histograms were used, not to supersede the W and/or D' test results, but to demonstrate that the drin data is bounded by a normal distribution (see item 12c in " Instrument Dr A Data Analysis
- Methodology and Assumptions"). His typically was the case for samples with relatively high center peaks (i.e., high kurtosis).
l (9) On page 7 of Attachment I, item (e) states that the analysis uses the F test to check the
[ homogeneity of verisaces. The proposed test coa pares the variance associated with a long l calibration laterval to that of the shortest interval. However, other variances may also need comparison. For esemple, Attachment 8 soggests that the variance at 18 months is higher than the variance at 19 months. Therefore, it appears that a comparison against 18 months (rather than the longest calibration laterval) is warranted. Perhaps the largest and smallest variances should be compared. Also, a multiple-variance test, such as Bartlett's test for homogeseity of verlances, ought to be considered. Please address these issues.
Items 13a through 13 fin " Instrument Drift Data Aralysis Methodology and Assumptions" are +
l descriptions of tools for use in evaluating drift data time dependency; however, not all of these tools
, were used in every evaluation. De intended purpose of hypothesis testing using the F statistic for testing the equality of two variances (item 13e) was to support analysis of the groupings representing different ranges of calibration intervals (item 13d). If a group with longer cahirstion intervals had a
. slightly greater sample standard deviation than a group with shorter ones, then this test could be
- performed to determine whether or not that diff-rence was statistically significant. He objective L would be to show that a larger sample standard deviation for a group with longer calibration intervals was not due to driA time dependence but merely minor statistical variation.
(10) The results of the F test are not reported. Please discuss, i
Hypothesis testing using the F statistic for testing the equality of two variances was never used for the i purpose described above, it was used once for another purpose tojustify the combination of two sets of data from steam generator startup range level instrument strings that are identical except that the tenths digit of the indicator is covered in one set but not in the other. De discussion of this found ,
on Attachment 3, Enclosure 4, Page 10 ic LAR 95-024 is incorrect. De test has subsequently been reperformed, yielding the same result (RAI Attachment 9).
Pr ~,s E >ppee*= suns. Mw-emm.. gems ar- T- M Wm. - -*g es W--map =&+ s.-s-+ m e- >-- E
, (II) Variances for different calibration intervals do not appear to be alike. Conservatively, the higher verlance, rather than the " pooled" standard deviation,ought to be used. Even then,it appears that the entrapolation to 30 months is not appropriate. Please explain.
Although variances for different calibration intervals may not appear to be alike, Toledo Edison does not believe this is usually due to actual drift time dependency characteristics. Consider, for example, the RPS RC flow differential pressure transmitters (Attachment 8). If one accepts the notion that these differences in variances among data groupings of slightly different calibration intervals represent actual diffeiences in drin time dependency characteristics, then it would have to bc ,
concluded that these transmitters are subject to less drift during 18-month calibration intervals than i during 17-month intervals and still less drift during 19-month intervals. This is an intuitively incorrect conclusion, and even if it were correct, it would take far more drift data than Toledo Edison has available to substantiate. Instead, Toledo Edison believes the only meeningful conclusions ,
regarding drift time dependency that can saually be drawn from the available daut involve ,
comparisons among data groupings for calibrstion intervals close to a multiple of the refueling interval length (e.g., close to 18-month calibration interval data versus close to 36-month interval data), These natural groupings are easily identified when looking at drift venus time since last adjustment plots (Attachments 9 and 18).
Regarding extrapolation to 30 months, Toledo Edison believes it is both apprmriste and sufficiently conservative to utilize the linear multiplier ofitem 146 in " Instrument Drift Data Analysis Methodology and Assumptions" for any particular evaluation. This is certainly a worst case approach, as r.o instruments evaluated in the drift study exhibited either a sample mean or a sample standard deviation that continuously increased linearly with time, in addition, use of the square root multiplier ofitem 14c in" Instrument Drift Data Analysis Methodology and Assumptions" can also be appropriate and sufficiently conservative when good engineering judgement is applied in the evaluation of the drift data's time dependency. De square root extrapolation approach is similar, in concept, to an example given in ISA RP67.04, Part 11 1994, section 6.2.7 Drift. In this example, a typical manufacturer's drift value of 0.25% URL for a six-month period is converted to an 18-month value by taking the SRSS of the six month drift value three times. This follows from the assumption that drift during each six month period is random and independent of drift during any other six-month period. Toledo Edison is prepared to provide several examples illustrating use of both extrapolation multipliers during the meeting.
(12) Some of the sensors were tested immediately after being put in service (readings at or near 0 months). Please explain the rationale for these tests.
During the sixth refueling outage (in 1990), all eight RPS RC flow differential pressure transmitters were replaced due to the Rosemount fill oil loss issue. In addition, the sensin t line snubber internals were removed and the sensing lines were cleaned for seven of the eight transmitters. (ne other one had been done prior to the outage.) For IT RCI AI and IT RCID1, separate calibrations were performed after each work item. For FT RCID2, the as-left string check data was out of tolerance, which resulted in a recalibration of the transmitter alone. These additional tests did not impact the 30 month projected drift for the differential pressure transmitters because the data for test intervals less than three months was excluded from the extrapolation (see item 14d in "Instmment Drift Data Analysis Methodology and Assumptions").
(13) Please be prepared to provide a step-by-step review of a sample data calculation for at least one sensor during the proposed meeting.
Toledo Edison is prepared to provide several step by-step reviews of sample data calculations during the meeting.
v
~
RAI #:uEcc- i ! '
I I
Signal Converters (Point 2 = Worst Case)
The D* Test l
} [ Ordered Xi stdev's count: 89 T - term j -0.1700 1 -2.1I10 S^2: 0.56920 7.48 l 2 -0.1300 -1.6136 5.59 j 3 -0.1200 -1.4893 T: 117.49000 5.04 j 4 -0.1200 -1.4893 D'=T/S: 155.72924 4.92 5 -0.1000 -1.2406 4 6 -0.0800 -0.9919 average: -0.0002 3.12 7 -0.0700 -0.8676 sidev- 0.0804 2.66 8 -0.0600 -0.7432 2.22 9 -0.0500 -0.6189 1.8 10 -0.0500 -0.6189 1.75 11 -0.0400 -0.4946 D* min for 88 Points: 226.1 136 12 -0.0400 -0.4946 D' max for 88 Points: 236.2 132 13 -0.0400 -0.4946 D' Test fails low 1.28 14 -0.0300 -03702 l 0.93 15 -0.0300 -03702 D* min for 90 Points- 233.9 0.9, 16 -0.0300 -03702 D' max for 90 Points: 244 3 0.87' 17 -0.0300 -03702 D' Test fails low 0.84 18 -0.0300 03702 0.81 19 -0.0200 -0.2459 0.52 N
8
, a
, RAI Attachment 2 e
~ '
i ! l n 89 average 0.000 stdev 0.080 ,
B10 DnCIipfiqa Upper BottnA Expected OhEDd I (-e,X-2a) -0.I611 2.02 1 2 (X-20, X-1.330) -0.1075 6.09 3 3 (X-133o,X .67o) 0.0538 14.35 4 4 (X .67o, X) -0.0002 22.03 38 5 (X, X+.67o) 0.0534 22.03 38 6 (X+.67c, X+133o) ~
0.1070 1435 1
'7 (X+133o, X+2a) 0.1606 6.09 2 8 (X+20, +e) 99999 2.02 2
~
Verification of Normality - Blaning Metbod "" -
Signal Conwrters (Point 2 = Worst Case) m -
35 -
j 30 - - -.
2j="15 -------------------a
~ 1][
h 10 M D _ 1 -
$ - ik5 1 _
,s h -
bE 0- N i
! : t -
i i $i I i r--um '
~
~
(-*,X- (X-2o, (X-133o, (X-47o, (X, (X+E7o, (X+133o, (X+2o,
~~
- 20) X-IJ3o) X .67o) X) X+E7o) X+133o) X+2o) +=)
~
Bin Range I
% e e emnA s -w e .iraww<- - =,- .wm.a =.%-+.. u mr -
r RAIAMeh w 3 '
+
4 f
1 i
a i
t j Signst converers (Poire 2 - Wern Cear) exsrepossaed no 30 mueshs via separe net saaeed 1
The D' Test i sidev's count 89 T-term i -1.0274 1 -3.4180 S^2: 7.4 % 39 45.20607
! 2 -0.6853 -2.2458 29.46699 3 -0.6168 -2.0112 T: 537.31834 25.9065 l 4 -0.5154 -1.6637 D%T/S: 196.24810 21.13137 i 5 -0.4338 -13842 1735248 6 -0.4054 -1.2869 merage- -0.0298 15.51124 j 7 -03561 -1.1180 sidev- 0.2919 13.53264 8 -03022 -0.9332 11.18064
! 9 -0.2733 -0.8344 9.839942 10 -0.2467 -0.7432 !
8.635501 11 -0.2423 -0.7280 D* min for 88 Points- 226.1 8.237611 12 -0.2284 -0.6805 D* max for 88 Points- 236.2 7.538068 13 -0.2281 -0.6794 D' Test fails low 7.299009 14 -0.2199 -0.6512 { 6.815906 15 -0.2147 -0.6337 D' min for 90 Pomis- 233.9 6.44249 16 -0.2137 -0.6300 D* max for 90 Points- 2443 6.196524 17 -0.1987 -0.5787 D' Test fails low 5.563938 18 -0.1745 4.4956 4.710514 19 -0.1628 -0.4557 4.233298
1< .
1 RAI Attachment 4 l .'
l l
f a 89 average -0.030 .
stdev 0.292 dig DfEIiDt1931 LIAT1029Ad E3Dtsttd Qhyggd I (-e,X-2o; -0.6135 2.02 3 2 (X-20, X-1.330) -0.4190 6.09 2 3 (X-1.33o, X .67o) -0.2244 14.35 8 4 (X .67o, X) -0.0298 22.03 33
$ (X, X+.670) 0.1648 22.03 34 6 (X+.67o, X+ 1.33o) 0.3593 14.35 5 7 (X+133o, X+2o) 0.5539 6.09 1 8 (X+2o, +cn) 99999 2.02 3
_ Wrificatloa of Normality. Binalog Method Signal Cony's estrap to 30 soonths vis square root anethod (Point 2 = Worst Case) o e,.ct.d
- so . .
[ .] 25 - -
g 2, '
" ^
2 "
g is - 4 ca ili. io 9 _ fi. _ #
4 R g
- 5
~""
P.k
- i gJi-
~~
g .
m 0 r-- E . ..
m- i -
- &q-r
% L . - p-E w- a-u. a-au. a-a. a. a.m. a.au, a.u.
=> x-o w> x-a> x3 x. => x.i.u.> x.u> i num e- ~ - - - - c-w...,as -- wen--m =m- ..--e-
~ .
R
, . +
es a , s ,
T T e'
+ >
, o e - -
s: .
3 o +
s M
g.
h5
- y. {_
1
(5 . 3 1 p[- d j
-d
,s e -
*E 4
4
.c e n o
E 4
=
8
- 8 8 8
- s a s g
- = = a # e v v 4 9 -
(u=4 Fsom M)UNG 9 M' -**F -'mw p r#w e,yg g p$w , . p q 3r ,..,g,g4y g.g p g , ,g. , , , , , _
O
-. - dE
+ S N8d a'
qu . , . R-h
. .e R
I R A R C {
<: f fg -
R
.ga ,
3 E .g ++ ,+. 3 c . j J, g , . 3 e - a f e I a o S S % % 8 % 8 % % E % 8
= o o o o o 9 9 9 9 i i (undsy seeme upa
i
~
a e gi 9 " " e 4,
--. 1
, o 8
3 3 IE t e a >1 h b3$ k k J w
+ 5 e dh O
$ b '~ ! k $ $ k k n d 6 o 6 o o o 9 k
9 k 9 9 (medspinomg)g5aq g am,a* -ma' ** GW' a r ;w8=mm8 M *e-SM-WWW M - we ., - w e -
V I ,' . . d 8 e R
.[
3
- R R
C. 6 i 1
.1 ? .y 3 3 ti
& 5g .
<g 5 1 =
, R o e o o e o
e
}
o
- 9 e e J
o S o E $ o 4 o 4 o S o S 4 4 5
- 9 9 9 9 (w=dse naaaa)vpa
.., , , . . . ~ . . . ._ ., _ _ _. .
o o, RAI An*-hmere 9 t Drift data from LISP 9A6A and B6A
-0.4 0 J
Dria data from LISP 9A8A and 83 0.16U 0 j -0.4 0.4 0.04 0.12 Drift from LISP 9A6A and B6A
-0.4 0.16 0.4 -0.2 .. h ofpoints- 18 i
0 0.08 -0.16 0.2 2 i Variance n2 0.041830065
-O 4 02 0.2 -0.4 __ l l'
-0.4 0.2 -0.4 0 O
Qfrom LISP 9A8A arid BSA
-0.4 0.4 0.4 amnber of points- 63 0 0 0.24 0.12 Variance ci 2 0.066080082 , 0 0.4 0 0.16 '
l O 0.12 036 -0.2 Hypoe w e c #=U2* 0 0.12 -0.2 0.16
- Alternative c 3>c2
-0.4 -0.24 0.24 -036 l
-0.4 0.2 -036 0.4 Statistic Vo: 1.579726959
-0.4 -036 0.4 0 l
-0.4 0 0.16 0.12
~
Cww of freedom- (62,17) 0 0.44 0.04 0.08
-0.4 F distributson value (c) for (100,17): 2.;0 2.02.
0.04 0.04 -0.1b F diAnhition value (c) for (50,17): 2.08 0 0.44 -0.16 0.16 l l
-0.2 0.24 -0.4 Since Vo<c, we do not reject the hypothesis 024 436
-036 0 0.4 0.48 ,
f 1 - 4
,7 :
L i September- 12, 1997: LICENSEE: Toledo Edison' Company-FACILITY: Davis Besse Nuclear Power Station
SUBJECT:
SUMMARY
OF AUGUST:21, 1997, MEETING ON INSTRUMENT DRIFT STUDY-0n August 21 1997, NRC~ staff members met in Rockville, Maryland, with
-representatives of Toledo Edison Company. The purpose of- the meeting was for Toledo Edison Company to discuss their response to the flRC's request for additional information (RAI)_ dated June 11. 1997, regarding license amendment requests-(LAR) 95-27, 95-24 and 96-14. A list of the meeting participants is
. . included as Attachment 1, and a copy of the meeting _ handouts is provided as
- Attachment 2.
The 1;censee' opened the meeting with a detailed step-by-step discussion of the
. Davis-Besse-instrument drift data analysis methodology using reactor protection system reactor coolant flow measurement as an example, The nethodology procedure and the example using the procedure are listed as Attachments-1 and 2, respectively. in LAR 96-14 dated April-18, 1997. The meeting continued with a discussion of the remainder of the RAI responses as documented in Attachment 2. ;l The meeting closed with the licensee indicating a sup)lement to LAR 95 would be submitted to clarify some minor issues that lave no impact on the licensee's basis for the amendment, Original signed byl Phillip M. Ray, Project Manager Project Directorate III-3 Division of Reactor Projects III/IV Office of Nuclear Reactor Regulation Docket'No, 50-346 Attachments: 1. List of Meeting Participants 2, Meeting. Handouts
- cc w/atts:'<See next page-LDISTRIBlfTIONf'See next page DOCUMENT NAME: JGi\dAVISBES\0B98520.MTS t 0FFICE PD3-3 PK- 1 1 PD3-3:LA: slAls PD3-3:PM -lE-NAME Pray g/K'; '
C80yle 6 % AHansen M/J DATFr 09/1(/97 09//6 /94 09/ e /9V
. = OFFICIAL RECUR) COPY 1}}