ML20133C902
| ML20133C902 | |
| Person / Time | |
|---|---|
| Issue date: | 07/09/1985 |
| From: | Heltemes C NRC OFFICE FOR ANALYSIS & EVALUATION OF OPERATIONAL DATA (AEOD) |
| To: | Harold Denton, Minogue R, Taylor J NRC OFFICE OF INSPECTION & ENFORCEMENT (IE), Office of Nuclear Reactor Regulation, NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| Shared Package | |
| ML20133C904 | List: |
| References | |
| NUDOCS 8508070174 | |
| Download: ML20133C902 (10) | |
Text
, .
~
[] rz t P DP-JUL 09 E85 MEMORANDUM FOR: Harold R. Denton, Director, NRR Robert B. Minogue, Director, RES James M. Taylor, Director, IE Regional Administrators FROM: C. J. Heltemes, Jr. , Director Office for Analysis and Evaluation of Operational Data
SUBJECT:
TRENDS AND PATTERNS ANALYSIS OF FEEDWATER TRANSIENTS AT WESTINGHOUSE PWRs Enclosed is a final report of our trends and patterns analysis of feedwater transients at Westinghouse PWRs. This report is supplemented by Engineering Evaluation Report AE0D/E418.
The draft version of this report has been reviewed, and changes have been made in response to comments. The comments received and our response to them are also enclosed.
Feedwater transients comprise the most frequent cause of PWR reactor trips which in turn are the most frequent class of transients. Thus, feedwater transients as a class frequently cause situations requiring operator response and the opera-tion of backup systems to maintain the unit in a safe condition. In the worst case, loss of main feedwater without prompt recovery is part of a risk-dominant transient sequence. This study, based upon available information, was initiated to characterize the incidence of feedwater transients and, if possible, pinpoint the causes. This report covers the time period January 1981 through June 1983.
The inquiry was limited to the largest singic vendor class of nuclear units (i.e., Westinghouse PWRs) in order to keep the analysis tractable.
Based on the analysis described in this report, we conclude that 2-loop and early 4-loop Westinghouse plants share a transient rate which is a factor of 10 lower than that for 3-loop and later 4-loop units. We believe that the higher rate is due, at least in part, to the use of turbine-driven vice electric-driven main feedwater pumps in two 3-loop and all later 4-loop units. AE0D will analyze this relationship further as part of its in-depth study of the causes of reactor scrams. j N ,
ggDli 1
& o, k A d f v B508070174 850709 PDR C
TOPRP EMVWEST PDR
) {}'/ I Yj hl 1l
r
~ .
In addition, we found that, when all Westinghouse units are viewed as a group, Salem 2 is an outlier due to its high transient rate. In their response to the draft report, Region I noted that the problems had been recognized, and addressed in inspection and SALP reports. The licensee had committed in 1982 to replacing the condensate pumps with higher head models, installing a heater drain tank quench system, revising station operating procedures, and implementing additional operator training. We recommend that Region I continue to monitor the improvements underway to determine if they are effective.
We are placing a copy of this final report in the Public Document Room.
If you have any questions regarding this matter, please feel free to contact Bob Dennig of my staff. Mr. Dennig can be reached on x24491.
original s'gned by c,J. Henemes. Jr.
C. J. Heltemes, Jr., Director Office for Analysis and Evaluation of Operational Data
Enclosures:
As Stated cc w/ enclosures:
W. Dircks, EDO J. Roe, DED0 T. Rehm, A0/EDO V. Stello, DEDROGR J. Sniezek, DEDROGR E. Blackwood, DEDROGR H. Thompson, NRR R. Bernero, NRR J. Knight, NRR Distribution:
T. Speis, NRR PDR W. Russell, NRR PTB SF G. Holahan, NRR PTB CF M. Srinivasan, NRR AE00 SF D. Ross, RES AE0D CF G. Arlotto, RES MHarper E. Jordan, IE RDennig S. Schwartz, IE FHebdon E. Rossi, IE TIppolito J. Gagliardo, RTC CHeltemes J. Shea, IP H. Faulkner, IP T. Murley, RI N. Grace, RII J. Keppler, RITI R. Martin, RIV PTB AE D .. PTS....
omcr >
J. Martin, RV
- - - -- v- - - -
O g .. ... ..
"""*"4 .R .g :. .g,t , ,, NHeb.d. n ,,,,, . ,T. .. ,
'.CJ g emes,, ,MH arp,e, ,,,,,,,
' " > 4.d. . . 85. ,,,,,, 9, /,g/,8 5,,,, ,,, ,f/5785,,,,,,, ,3/,g,/,85 ... . . 6./.@as... ..
""c'""" "'""C" " OFFICIAL RECORD COPY .
Attachment 1 Trends and Patterns Analysis of Feedwater Transients at Westinghouse PWRs.
A draft report of trends and patterns analysis of feedwater transients at Westinghouse PWRs was sent out for review by the NRC staff in December 1984.
Comments on technical accuracy have resulted in changes in the final report; comments on potential areas for further study and long term changes in trends and patterns reports are addressed below.
Office of Nuclear Reactor Regulation comment:
You state in the report and your memorandum, that loss of main feedwater with-out prompt recovery is a part of a risk-dominant transient sequence. While your report does address the frequency of the feedwater transients in detail, the initial reactor power level and the degree of possible feedwater recovery, ,
which are very important to the events' risk significance, are not included.
- This information would be very helpful.
~
Response
We agree that risk significance for each event would be very useful informa-tion. This is a very difficult area to quantify, but we will begin to include this information in future trends and patterns reports. Reactor power level is expected to be included in these reports.
Comment:
We would also recommend that your threshold for recommending Region investiga-tion be reconsidered. While Salem 2 may be a statistical outlier (1.44 transi-ents per 1000 critical hours), a number of plants (e.g., Sequoyah 1, Trojan, Salem 1, Farley 2, Beaver Valley 1) are not far behind (1.00 to 1.09 transients per 1000 critical hours). Since considerations such as power level and poten-tial for feedwater recovery would appear to have as much effect on safety signi-ficance as the initial transient frequency, such that the events at these plants, and perhaps some others, could be as significant as Salem 2.
Response
Lowering the threshold for recommending Regional investigation will be consid-ered in the future as we gain confidence in the methodologies used in trends and patterns analysis.
2-Comment:
A summary table which includes all plants, age, critical time, maximum transient likelihood (with bounds) and feedwater transient cause would be useful.
Response
Summary tables of raw data will be included in future trends and patterns reports as this information is shifted onto computers.
Comment:
Units of " transients /yr" are more commonly used. Conversion to these units in the report would be helpful.
Response
The use of units for transients / year or transients / critical hour will become more standardized in future trends and patterns reports.
Comment: .
Page numbers are mixed up near the end of the report.
Response
Page numbers in the Appendix have been corrected.
Comment: (Page 6)
Last paragraph-It appears Table 2 is meant rather,than Table 1.
Response
Table number has been corrected.
Comment: (Page 7)
It is very difficult to read the Figure. The dots representing the transients
. are barely visible. In addition, the number of transients for each plant and the Figure number are miscing.
Response
The number of transients for each plant has been put back into the figure.
Comment: (Page 9)
On Table 2-the difference in Operator Errors is about the same and should not be considered different.
Response
Operator errors have been removed from the discussion of differences in Table 2.
Comment: (Page 9)
The footnote is not self explanatory.
Response
The footnote was written for statisticians, rather than non-specialists. -
Comment: (Page 10) -
Feedwater transients during startup silould be included.
Response
All feedwater transients during electricity production and all feedwater transients leading to a Licensee Event Report were included.
Comment: (Page 13) ,.
Bottom paragraph-The preceding paragraph states that 3 loop /4 loop units do not constitute a homogeneous class, yet, the 3 loop /4 loop units are treated as a class in the report.
Table 11 shows that most 3 loop plants have electric driven feedwater pumps not turbine driven pumps as implied here. Steamdriven feedwater plants are consid- ,
ered as a class in Table 11. The issue of steam driven vs. electric is confused '
by bringing in 3 loop plants (which have electric driven feedwater pumps).
Response
The implication that most 3 loop plants have turbine-driven feedwater pumps has been removed from the discussion.
Comment: (Page 14)
Same comment as previous. 3 loop and 4. loop plants were shown to be a non-homogeneous class, yet are grouped here.
Response
We agree, but 2 Loop and Early 4 Loop units have a much smaller feedwater transient rate than all other Westinghouse units.
Region I Comment:
Areas worthy of further examination in feedwater system transteat induced reac-tor scrams include:
Condenshte polishing system operation during plant start-up and early post-startu'p periods.
The reserve water inventory in a steam generator available to absorb a feed system transient. (Comparison of trips as a function of rated steam -
flow relative to mass of water in steam generator.)
The reserve capability of the feed system to sustain a transient without causing a reactor scram due to steam generator low water level.
Response
Some areas ruggested for further study of feedwater transients will be .
incorporated into the annual scram trends and patterns reports.
.' Office of Inspection and Enforcement (IE)
Comments:
We endorse the concept t.eing used in the draft report. However, we feel that the events being considered in the report transcend the feedwater system. In brief, the events cited as being feedwater transients appear to encompass the entire power conversion system, including the main steam, condensate, heater drain, reheat, electrical and control systems. It appears that the systems actually l being analyzed would provide too many degrees of freedom for the available data.
l -
If 'such is the case, one would expect that the broad characterization of feed-water transients used in the report would result in confusing trends and pat-terns.
l
Response
Systems studied in this analysis include feedwater systems, condensate systems, and feedwater control systems. They do not include the entire power conversion
5-system, the main steam system, or the electrical system, but include the systems that directly cause feedwater transients.
Comment:
One concern with the report is the fact that the " Gray Book" (NUREG-0020) is cited as being the best source of information on feedwater transients and that it was the source of almost 90% of the identified instances. The Gray Book's descriptions of transients are too abbreviated and cryptic to determine the actual causes of the transients. Further, some Gray Book transients which ap-pear to be identical (i.e., transients attributed to the same cause and that involve the same system) have occurred four times in the same day. In such cases, it is not clear whether.the draft report treats these events as one tran-sient that occurred four times, as seems reasonable, or as four discrete tran-sients.
Response
We agree that the Gray Book has many limitations in the description of events.
However, it was the only source of data for these events on a plant specific basis. Events were considered discrete, whether they happened on the same day or not, if the licensee took the trouble to report an event as a separate -
occurrence. .
Comment: -
The economic impact of feedwater transients is such that the electric utility industry may have more meaningful information on feedwater transients than that contained in the Gray Book. Potential sources of such information include the Electric Power Research Institute, the Institute for Nuclear Power Operations and the North American Electric Reliability Council (NERC) at Princeton. NERC can perform generic searches of their computerized availability data base re-
. lated to systems used in power generating stations, including feedwater systems.
Such searches are conducted on a no cost basis for federal agencies such as the NRC. ,.
Response
Generic searches of proprietary data bases may aid in increasing general understanding of transients, but are not sufficient for identifying outliers and potential problems on a plant-specific basis.
Comment:
Finding 5) of the Executive Summary reads, in part, " Salem 1 and 2 account for 19 of the 39 Main Feedwater Pump Faults" (sic). In contrast, Table A.1, enti-tied, " Main Feedwater Pump Faults" lists 53 not 39 such faults. In addition, this category (i.e., main feedwater pump faults) is misleading in that several
- of the listed faults do not appear to be pump faults per se, but faults of other components or systems such as transformers, controllers, electrical systems and condensate systems.
Response
Finding 5) has been corrected. Main feedwater pump faults are faults of the pump or its components that lead to a pump trip or other pump caused feedwater transient.
Comment:
Finding 7) of the Executive Summary reads, in part, " Robinson 2, Surry 2 and Beaver Valley account for 22 of the 53 Main Feedwater Regulating Valve Faults" (sic). In contrast, Table A.4, " Main Feedwater Regulating Valve Faults" lists 43 such faults not 53. It is presumed that the listing of Surry 2 is a typo, since Surry 1 had 6 such faults whereas Surry 2 only had 2.
Response
Both typos were corrected in the final report.
Comment:
Finding 8) of the Executive Summary reads, " Beaver Valley accounts for 6 of the .
11 Feedwater Bypass Valve problems" (sic). In contrast, Table A.5 which is entitled "Other Valve Faults" rather than "Feedwater Bypass Valve Problems" -
indicates that Beaver Valley 1 accounts for 7 not 6 of the 11 events. Further, not all of the 11 listed failures are bypass valve failures.
Response
The numbers in Finding 8) have been corrected.
. Comment:
The first sentence of the last paragraph in page 6' reads, "The data in Table 1 indicate a statistically significant difference between the distribution of causes (underlined for emphasis) for the first month transients and the distri-bution for transients occurring later. This sentence is misleading in that Ta-ble 1 merely indicates the number of transients, not their causes.
Response
The table number has been corrected to Table 2.
Office of Nuclear Regulatory Research Comment:
We would recommend the use of graphical displays in presenting results of the analyses. Insights can often be more readily gleaned from graphical presenta-tions such as the Weibull fits.
Response
We agree that reports benefit from graphical presentations. Additions of com-puter hardware and software this year have increased our ability to include more graphical displays in future trends and patterns reports.
~
Comment:
In the Executive Summary, for major finding number 5, you may want to test whether both Salem 1 and Salem 2 are outliers, since Salem I had 10 main feedwater pump faults and Salem 2 had 9 main feedwater pump faults. In findings 5 and 7, it would be useful to state totals and subtotals as follows. In finding 5, there was a total of 53 main feedwater pump faults, where 39 of these faults were in the 4-loop units and 19 of these faults were accounted by Salem 1 and 2. In finding 7 there was a total of 43 (not 53 as stated) main feedwater regulating value faults where 30 of these faults were in the 3-loop units and 22 of these faults were accounted by Robinson 2, ;urry 2, and Beaver Valley. If Beaver -
Valley's 6 feedwater bypass value faul s are highlighted in Finding 8, why not call attention to the 10 general steam cenerator level faults at Sequoyah 1 and -
the 8 operator faults at Trojan?
Response
Only Salem 1 was found to be an outlier. Findings 5 and 7 were corrected in the' final report. General steam generator level faults were not highlighted because not enough information was given to identify causes of these events.
Operator faults were not highlighted because they are such a diverse, non-homogeneous group of events.
Comment:
In the last paragraph on page 6, the differences in the indicated mix of distri-butions is characterized as statistically significant. However, the P-value of 0.0742 is borderline. That is, based upon this P-value, one would reject at a significance level of 10 percent and accept at a significance level of 5 percent.
Response
We agree with the observation, but selection of a significance level is a matter of judgment.
Comment:
For future investigations the data in Table 2 could be analyzed using waiting time theory. For example, a waiting time distribution could be fitted to the set of values for each plant.
Response
We will investigate this theory to see if it offers advantages in future investigations.
Comment:
Using the Weibull distribution to fit only 4-loop data from Table 4, the Salem 2 observation was not an outlier as stated on pages 12 and 14.
Response
We may be using the Weibull distribution in later AEOD studies to identify outliers depending on development of this technique.
Comment:
In Tables 5 and 6, pages 12 and 13, how will the information on MLE's and confi-dence bounds be used? As is noted in the footnote on page 12, the population is non-homogeneous. Therefore, MLE's and confidence bounds should be computed for each partition of the population, for example, see the partition in Enclo- i sure 2, attached. Other than containing outliers, what is the criterion for a population to be homogeneous?
Response
- The estimates and bounds were not intended for'any use beyond the report itself.
1
- Table 6 does show rates and bounds for each partition of the population.
Where additional partitioning occurred, i.e., for both loops and age, each cell was again tested for homogeneity.
- The criteria for homogeneity are given in the footnote on page 12: Reject homogeneity hypothesis if either a two-sided test based on the most signifi-cant outlier or an overall test based on Pearson ChiSquare Statistic is
.significant,at the 10% level.
.