ML20245A462
| ML20245A462 | |
| Person / Time | |
|---|---|
| Issue date: | 01/31/1989 |
| From: | Crossprather, Oreilly P, Wolf T NRC OFFICE FOR ANALYSIS & EVALUATION OF OPERATIONAL DATA (AEOD) |
| To: | |
| References | |
| TASK-AE, TASK-S804B AEOD-S804B, NUDOCS 8904250209 | |
| Download: ML20245A462 (23) | |
Text
_ _ _ - _ _ _ - _ _ _ _ _ _ _ _
-l.
Qo?l y(3)'
"l AEOD/S804B APPLICATION OF THE NPRDS FOR MAINTENANCE EFFECTIVENESS MONITORING A800 January 1989 o,,
5'
%...../
(
Prepored by: P. D. O'Reilly, T. R. Wott P. Cross-Prother Division of Safety Programs Office for Anofysis and Evoluotion of Operational Dafo U.S. Nuclear Regulofory Commission 1
i NOTE:
Appendixes A and B to this report contain plant-specific data from the Nuclear Reliability Data System (NPRDS). Therefore, they must be treated as containing PROPRIETARY commercial nuclear power plant information.
8904250209 890131 PDR ORG NEXD PDC
yy; e
,4
=
i EXECUTIVE
SUMMARY
On November 28,1988,the Commission issued a proposed rule on " Ensuring the Effectiveness of Maintenance Programs for Nuclear Power Plants," 10 CFR 50.65.
The proposed rule would require licensees to formalize their maintenance pro-grams in accordance with the definition in the rule, and to monitor the effective-ness of their programs. Specifically, the rule would require licensees to:
.. regularly assess the effectiveness of this maintenance pro-gram, and based upon this assessment, make improvements as appropriate.
Operating chorocteristics such as consistently high ovoilobility, or low equip-ment-coused forced outage rates over several operating cycles are indicators of good maintenance effectiveness. However, plant material condition con degrade significantly before these indicotors provide identification of degraded molntenance performance. A more timely indication of the effectiveness of molntenance is needed.
To support the monitoring provision of the proposed rule,the NRC's Office for Analysis and Evoluotion of Operational Data (AEOD) conducted maintenance performance indicator developmental activities and documented their results in AEODIS804A Preliminary Results of the TrialProgram on Maintenance Perform-ance Indicators, which was transmitted to the Commission by SECY 88-289 on October 7,1988. That report concluded that indicators whlch are based upon octual component reliability and failure history provide the best measure of maintenance effectiveness. It recommended that:
Licensees should be strongly encouraged to utilize on industty-wide component failure reporting system, e.g., NPRDS, as a basic element of the maintenance effectiveness monitoring activity that is to be required by the rule.
This report, AEOD/S8048, demonstrates the utility of the Nuclear Plant Reliability Data System (NPRDS) to provide useful maintenance effectiveness monitoring information. It documents the development of on indicator that is based upon the component failure reports submitted to the NPRDS, and demonstrates that the monitored indicator reflects maintenance effectiveness.
Demonstrating the validity of the condidate indicator required that the indicator be based on a reasonably complete, consistent set of NPRDS dato. In order to AEOD/S804B 2
.i 2
- 4 ensure that the data would satisfy these criterio,this study considered only major components in systems which have historically been significant contributors to forced outages. Failures of this equipment were considered most likely to be reported to the NPRDS regardless of a plant's NPRDS reporting consistency or the aggressiveness of its operations personnelin detecting failures. Using this dato, on indicator of maintenance effectiveness was then constructed that monitors increases in the failure rates within a system, and proviu s a signal when on increase exceeds a specified value. This yields a measure of the changes in the effectiveness of maintenance on a system basis. To obtain a measure of a plant's level of maintenance effectiveness, the number of indications or signols is follied across o number of systems. This folly is but one indication of the effec-tiveness of a plant's maintenance program. Other items, such as additional Indicators, systems analyses, and inspections, are needed to obtain a complete picture of the absolute level of the effectiveness of maintenance at any plant.
The validotion as to whether the condidate Indicotor reflected maintenance ef-festiveness was based upon deterministic engineering analyses and empirical methods. Engineering studies of NPRDS failure records for such compone-Js FAILURE RATE CHANGE INDICATOR
,,,,,,,, g ences in maintenance practices among the plants caused differences in failure rotes. Further, root cause
,.u=== [
onalyses of the failures com-prising the Indicator revecled molntenance effectiveness as the major cause. Figure A illustrates this port of the
,,a m
., u.
aau act
,=w. o=amav==
validation process. Finally, eau g g eg es empirically,it was shown that j
theindicator correlates reasonably wellwith other in-formation regarding mainte-
^Y nonce problems derived bE,,, 63 from Ucensee Event Reports (LERs).
Tn The usefulness of the condi-
~
dote indicator, and any otherindicator developed
{
based upon the component l
failure reports submitted to i
AEOD/S8048 3
)
1
, F O ':.
1 the NPRDS, reiles to a large degree on the quality and completeness of NPRDS reporting by licensees. Since such reporting is voluntary and subject to individual utility priorities and commitments, some limitations are Inherent in the utiltzotion of NPRDS foliure reports for maintenance effectiveness trending.
This report documents that a practical and useful maintenance performance in-dicator was developed using NPRDS dato. The obliity of the condidate indicator to reflect maintenance effectiveness was confirmed. The effect of non-uniform NPRDS reporting was shown to be acceptobly minimized through the use of a standard subset of equipment that is important to plant operoflons.
The vo!ue of the condidate indicator was confirmed through independent dato derived from maintenance-coused events reported in ERs, correlations with other studies, and correlations with the findings from maintenance effectiveness team inspections. While the focus of this report is on the use of NPRDS to monitor maintenance effectiveness, the mutually reinforcing correlation between ER-based dato and the NPRDS-based indicator points to the prospect of an addl-tional maintenance indicator. The ER-based dato used in this correlation resulted from the ongoing performance indicotor development effort aimed to demonstrate the usefulness of cause codes, one of which is maintenance.
Further development of this molntenance cause code from ERs is being pur-sued for use in monitoring molntenance effectiveness.
Although the methodology used in this study was developed using data for 28 BWRs, it should prove equally valid for other plant designs. Other valid Indicators may be developed from this data but the condidate indicator developed in this study serves as a suitable basis for describing a maintenance effectiveness tracking me+ hod which is acceptable to the staff In the forthcoming Mainte-nonce Rule reguictory guido.
In order for the NRC staff to use the condidate indicator on on Industry-wide cost-effective basis, further development is necessary to more efficiently extract the indicator from the NPRDS system and to display it in a manner which permits individual as well as generic comparisons. The staff continues to give further development efforts high priority and will shore the results of its activities with industry.
m..
AEOD/S8048 4 1
,o*.
APPL.lCATION OF THE NPRDS FOR MAINTENANCE EFFECTIVENESS MONITORING INTRODUCTION The Office for Analysis and Evoluotion of Operational Data (AEOD) recently issued AEOD/S804A, " Preliminary Results of the Trial Program on Maintenance Performance Indicators" (Ref.1). A number of condidate maintenance per-formance indicators were onclyzed, including process indicators such as correc-tive maintenance backlog, and equipment performance-based indicators such as rework and frequency of failure. A major conclusion of this study was:
Indicators that are based upon actual component reliability and failure history provide the best measure of maintenance effectiveness....
At the most fundamental level, this translates into tracking component perform-once through the construc*" n of component failure histories. Tracking equip-ment performance is also geneially accepted as a way of improving mainte-nonce. AEOD/S804A noted, however, that licensees generally were not using such data to ossess maintenance effectiveness. Independently, os shown in the following findings, recent NRC maintenance inspections otso found this to be the cose:
Improvements in problem resolution remain to be demonstrated, in view of prior and recent missed opportunities to recognize and correct the root causes of plant problems. The absence of etfec-tive eauioment performance trendina crocrams cocears to have contributed to such oversichts. (Ref. 2, emonosis added)
Work history and performance history are not integrated and repetitive failures of work on similar components cannot be I
AEOD/S804B 5
,(.o'(.
readilyidentified. Therefore, root cause analysis and prompt identification and correction of problems (are) not as effective as (they) could be. (Ref. 3) i Trending of equipment failures - the inspection team observed examples of failures to adequately assess and trend equipment failure data.... In addition to the lack of an adequate trending activity, this weakness in providing feedback to the PM (i.e.,
preventive maintenance) program also stems trom incomplete maintenance and equipment history records. (Ref. 4)
AEOD/S804A o!so noted that most plants have a molntenance work request tracking system but that such systems do not lend themselves to the ready Identi-fication and tracking of individual component failures. Consequently,it was concluded that the Nuclear Plant Reliability Data System (NPRDS) (Ref. 5) was the best available source for component failure dato. This conclusion was independently observed during one of the previously cited inspections (Ref. 3).
The inspection report noted thot:
Repetitive faGures of work on similar components cannot be readily Identified by using CHAMPS (i.e., the plant's maintenance tracking system)... NPRDSis generally used, when requested, for failure determination.
Since the NPRDS is such a valuable resource,it must continue to maintain a high quollty of component failure data. To confirm that the NPRDS remains a viable source for component failure data, the NRC periodically assesses its quality. The most recent annual appraisal is provided as on ottochment to this report.
Building on the findings of AEOD/S804A and the recent molntenance inspec-tions, work continues on the development of maintenance performance indica-tors based on NPRDS dato. This report describes the results of this work by provid-ing a detailed example of how NPRDS failure histories for selected equipment, colled outage dominating equipment (ODE), con be combined into on indico-for of maintenance effectiveness. While the indicator was developed and validated based upon NPRDS data for a single reactor type, l.e., General Electric (GE) boiling water reactors (BWRs), the principles and opproaches used are con-sidered equally appIlcoble to reactors of other designs.
The next section discusses the construction and use of the condidate indicator.
Subsequent sections provide details about the process used to validate this indicator and on examination of its use as a timely indicator of equipment forced outages.
AEOD/S8048 6
d ',.
,3 INDICATOR DEFINITION AND CONSTRUCTION The indicator constructed in this study scans the NPRDS component failure rate data within a system and signals any increase in that rate which exceeds a pre-determined threshold value. The number of these flogged failure rate increases is then tollied for all systems considered over a specified span of time to obtain a PLANT A Component Fouure Trends measure of the level nousea Geormesm System of molntenance ef-I H
I i >.,. s s s s s n..I, H,.
I festiveness at a plant. Figure 1 is on example for one j
ggg gg gg gl plant of the compo-i'* * *~
.'6~~
6-nent failure rate in-creases that were l
l flogged for five
' ' ' ' * ~ ' ' ' ~ ' ' '
**~ ~= *' ~*'
=
different systems. In
* *
this example, a total l
ll l
l l
l sss >. i, s s s s s a. e.,. s -
g g
g Increases were sig-noled during approxi-l,.,.,.,I,II,.,.,.,,.,.,.,.,.,.,d motely a three-year MONTHS time span. The following sections 4
discussin detail the Figure I definitionof the 1
Indicator,the methods onc' reasons for the selection of the equipment and follure dato used, and the cor struction of the indicator from that dato.
INDICATOR DEFINITION 1
Of the number of parameters which could be monitored as on indicator of maintenance effectiveness, the rate of reported component failures (i.e., failures 1
per month) was considered to be the most definitive measure of equipment per-formance and the one that could be directly linked to the effectiveness of the i
maintenance performed on that equipment. However, this parameter is suscep-tible to plant-to-plant inconsistencies in failure reporting. Control of such incon-sistencies, as well as dato completeness, con be exercised by measuring a plant ogainst itself. This con be done by monitoring either the deviations from on AEOD/S804B 7
p 7i*W ; p.
I overage failure rate or o change in the failure rate. This study focused on the change in the follure rate as the indicator of maintenance effectiveness.
- An increase in the rate of component failures is indicative of a change in the effectiveness of maintenance. Such a change in failure rote lends itself well to trending and, consequently, may be used as a trend indicator. This mp accomplished by tallying the number of increases in the component all rate over a given time span for o number of different systems. This tally by is but one indication of the effectiveness of a plant's maintenance program. Other items, such as additionci indicators, systems analyses, and inspections, are need3d to obtain a complete picture of the absolute level of the effectiveness of maintenance at any plant.
- It should be' pointed out that the methodology used to obtain the indicator grouped the failure dato according to porticular components in selected systems. During the course of the developmental analyses,it was found that applying the indicator at a higher level, e.g., all components together, diluted its sensitivity and resulted in relatively fewer indications than were obtained when the analysis was performed on on IndMdual system basis.- Another insight that stemmed from these analyses was that the data must be analyzed on at least a monthly basis. Viewing the failure dato on a quarterly basis resulted in a loss of the fine detail and sometimes a dampening out of pronounced increases in component failure rate that were exhibited when a monthly basis was used.
INDICATOR CONSTRUCTION in the construction of the condidate indicator, o comparative formula was developed to detect the rate of change in the failure rates of the components within a system. It was then computerized so that it would signal o component failure rate change that exceeded a predetermined value. Once the formula was computerized,it was adjusted to be sensitive to changes in the component failure rate that appeared significant based on trends observed in the historical data from 10 BWRs.
The resultant computerized indicator formula counts the number of component failures discovered during each month in a five-month span of time for each of the selected systems. Dividing the number of component failures for each of the systems in a selected time period by the number of months in the period,it then calculates the overage component failure rate for each system for (c) the first three months of the five-month time span and (b) the failure rate for the lost two months of the spon. It then compares the two average rates and,if the rate in the last two months exceeds that of the first three months by more than a threshold value, on Indicating mark is placed in the lost month of the five-month AEOD/S8048 8
i
,s**
4 span. The program then FAILURE RATE CHANGE INDICATOR odds the next more recent 1
.,.,_,,,,,mun,.....
month and drops the oldest month,i.e., the five-month span is shifted forward one month, and the failure rote calculations and comparison a-
'"""^**"'^'"a^'"
cre repeated.This moving window approach has the effect of providing multiple m
.r
..n.
Indicating marks over successive months if on i "^" """ "^" ^"" "^' 'r du' ^= """ co' 9
" """ '" """ ^"
increase in failure rate is large orifit is sustained over a number of months. Thus, the indicatorweights periods Figure 2 of t!me in proportion to the degrae of change in the component follure rate. Figure 2 shows how on increasing failure rate trend !s signaled by this method.
DATA CONSISTENCY AND COMPLETENESS The dato used in the construction and validation of the condidate NPRDS-based indicator had to satisfy two criterlo:
(1)
A consistent set of NPRDS dato had to be obtained for each plant, and (2)
The individuct plant data sets had to be reasonably complete.
Major dfferences in the level of NPRDS reporting from unit to unit have been observed. To accommodate the shortcomings of incomplete NPRDS reporting, o logic was opplied to utilize o subgroup of the equipment failures in the NPRDS.
Two factors dictated the group of equipment selected. First curing the site visits of the trial program documented in AEOD/S804A,it was noted that the operot-ing crew played a major role in molntenance work request (MWR) generation during plant operation. The NPRDS failure reports are dependent upon MWR generation and are therefore sensitive to the aggressiveness of the operating crew in the complete and timely identification of equipment problems. To minimize the effects that could be attributed to vorlations among plants, the dato analyzed was limited to failures of major components in systems that sup-port power operation. In cases where o plant was shut down, if it were to stort up without having repaired the foiled equipment, the plant would be operating with a degraded system that could eventually have on adverse impact on AEOD/S8040 9
1
y
[*
m i
, power operation. Failures of this equipment are much more likely to be identi-l:
fled for repair in a timely manner, thereby minimizing the potenticiimpact of the variations in the identificotton of failures.
The second factor that was used to assure reasonable completeness of the dato set was information obtained from the NPRDS coordinators. Generally, the NPRDS coordinators are the individuals that produce the NPRDS reports based upon the maintenance work request input, During the trial program reactor site -
1 L
. visits,it was found that, otthough the obsolute reporting rate may vary widely from unit to unit, the NPRDS coordinators generally report the important fallures.
. Important failures, in their view, were those that could influence plant.operoflon to such a degree that a plant outage could occur at their plant or another
- plant, Y
I Considering these two factors, the condidate molntenance indicator was based '
on NPRDS-reported failures from the set of equipment that historically has caused equipment forced outages. This data set represented reasonably complete component failure information for a reasonable scope of equipment.
Vorlations fiorr pW; io plant due to different NPRDS reporting philosophies were
. further lessened by using only those types of component failures that the NPRDS.
Reporting Procedures Manual (Ref. 6) requires to be reported (i.e., immediate
. and degraded failures). Incipient failures were not considered.
SCOPE The scope of the onolysis used to construct the condidate Indicator was limited to a specific subset of operating plants for o specific time period due to staff resource constraints. The subset studied was further restricted to only those plants with nuclear steam supply systems (NSSS) designed by GE that were op-erotional between January 1,1985 and March 31,1988. Further, the equipment considered was restricted to those BWR systems and components which histori-colly have been the dominant contributors to forced outages,l.e., ODE systems and components, that are within the NPRDS deportability scope. While this study was restricted to only one plant design,the methodology developed should be equally applicable to all plant designs.
The set of components in BWR ODE systems that was selected and analyzed was based on a compilation performed by the S. M. Stoller Corporation for the EPRI
. (Ref. 7). This compilation used the OPEC-2 database (Ref. 8) to determine and rank the contributing factors to plant unavailability down to the component level. A number of the dominant contributors to plant unavailability that were listed were related to either personnel or planned outoges such as refueling.
These contributors were not considered. Equipment was also eliminated that AEOD/S804B 10
}
l 1
l t
..r'e' l
was structural, such as BWR recirculation piping, or outside the current reportabil-ity scope of the NPRDS. Table 1 lists the systems and components that were selected for this study.
i TABLE 1: BWR ODE SYSTEMS AND COMPONENTS
. SYSTEM COMPONENT DESCRIPTION CorWd Rod Dtve Confrd Rod Mechcrien Contrd Rod Cor*d Rod Dtvo Flow ControlVdve Cored Rod D1ve Flow Cored Vove Operact Cormol Rod Dlve Stopfy Pump Cored Rod Dtvo Stopty Pump Motor Contrd Rod Dtve 8topry Ptsnp Motor Ocut kocker feedwater Feedwater Hgh Prese.se Heater Feedwater Pump Feedwater Pump Motor Foodwater PLmp Motcr Circuit kocker Feoowater PLmpitstano Feechster Ptsnp16stune Governor Man Stoon Man Dean Automcmc Depresulaation Safety Vdve Man Dean Autornatic Depressmanon $dety Vdvs Operator MdnStocrn Contanmentinciahon Vdve Man Stoon Contdnment isolation Vdve Operator Mdn Stean Contarvnent isolation Vove Operator Circut Demer Main Stoon Safety /Automanc Depresuuation Dschcrge Pipe Vacusn koctor Mdn Stean $ defy Vdve Neutron Marstonng Instunentation, Sstable/Mch insrunentation, indcatorg/Rorsorcars hatunentatiort fotrwnator/Primoy Detector /Bement Reactor RecucL4ation marunentation Bstctie/Mch estunentation. Indcators/Rocorders Instunentattort irawndter/Primoy Detector /Gement Reactor RecrcLJarlon Pump Reactor Recteuafhst Ptsno Motor Reacter RodrcLJation Ptsnp Motor Ocut kocher Reactor Rodrcuation Ptsnp Dschcrge Varve Recrtor Recreulcmon Pump Oschargo Vct e operator Reactor Recrcuation Ptsno Oschcrge Vatvo Operator Circut Doctor Reactcr Recrcuation PLang Strtion Vctwo Reacter Recacucmon Ptsnp sucson Vcm operarca Reactor Rodrcuarton Ptsnp Suction Vdve Cperator Ocut koc*ar Reactor Rodrcuation Ptsnp Motor Generator Set Generator Reactor RecycLJation Pling Motor Generator Set Cotseng Reactor Recrcuation Ptsnp Motor Generator Set Motor Recrtor Recacuanon Ptsno Motor Generator Set Motor Circr.it beaker The equ!oment listed in Table 1 is not on all inclusive list. Based on the results of this study, some changes are in order. For example,the BWR feedwater regulot-ing volve and its operator were not identified in the Stoller report as dominont contributors to BWR forced outoges. However,from the number of reported failures of these components found in the NPRDS during this study, these compo-nents were significant contributors to equipment forced outages of some of the
- my AEOD/S8048 11
-__________-____-___m
1.
0,
=
g l,
l l
plants considered. Consequently, they should be added to the list of key out-age-cousing equipment. Further, the NPRDS currently does not include certain balance-of-plant (BOP) systems and c omponents that have historically been sig-nificant contributors to plant outages, such as the turbine-generator and assocl.
oted support systems, the condenser, the circulating water system, non-nuclear portions of the service water and closed cooling water systems,the instrument cir system, and the service air system. At the most recent meeting of the NPRDS l-l Users Group (NUG) held in December 1988, the NUG recommended to the Institute of Nuclear Power Operottons (INPO) that the reportobility scope of the NPRDS be expanded to include the main turbhe, the moln generator, and the condenser. This action marks the first officlol step in the NPRDS scope expansion process.
As on independent check on the selected outoge dominating systems and components, published results were reviewed of a study of plant ovallobility that was done by the North American Electric Reliability Council (NERC) using their NERC-GADS database (Ref. 9). This review confirmed the basis used for selecting the equipment listed in Table 1.
As a result of queries of the NPRDS based on these systems and components,it was found that 8 of the 37 operating GE BWRs had too little data to analyze because of either limited commerclot operating history or,in some cases, due to extended shutdowns during the study period. In addition, Big Rock Point does not report to the NPRDS because of its unique design chorocteristics. Thus, the validation was based on NPRDS failure data from 28 operating GE BWRs.
The use of the indicator model and computerized algorithm developed during this study results in considerable time savings in the calculation of the condidate indicator for the number of plants analyzed. However, this process still requires the manual downloading of large amounts of component failure data from the NPRDS. Further manipulation of the downloaded dato is required to prepare the input for the algorithm. These two efforts are time-consuming and labor-intensive. The desirability of trending component failure rate has been recognized by NPRDS users. The current NPRDS user software hos the capobility to trend failure rates in on outomated way. Efforts have been initiated to see if expansion or modification of this software is possible so that it could provide the condidate Indicotor.
The following section of this report documents the validotion method that was used to confirm the relationsh o of the condidate NPRDS-based Indicator to the Commission's definition of maintenance effectiveness.
AEOD/S804B 12
i VALIDATION Validation of the condidate indicotor was accomplished through two tasks. The first task consisted of a root cause analysis of those component failure rate l
increases identified by the Indicator. This analysis was done to determine if the f
indicator is a direct or nearly direct measure of maintenance effectiveness. In the second task, the condidate indicator was compared statistically with on-i other measure of maintenance effectiveness that is currently under develop-i ment, namely,the frequency of maintenance-coused reportable events docu-mented in Licensee Event Reports (LERs) submitted in accordonce with 10 CFR 50.73 (Ref.10). While both the NPRDS-and the LER-based indicators are the subject of validation, a positive correlation between indicators based on dato from different sources would be mutually reinforcing.
i Applying the computerized algorithm technique to the NPRDS component failure data for the three year period considered resulted in between 0 and 8 indications for each of the ODE systems for a given plant,with the overage number of indications per system per plant varying between 2 and 3. About hoff of the indications were due to failures discovered during power operoflon and hoff were due to failures discovered during on outage. Forty of the component failure rate increases flagged by the algorithm were examined by AEOD contractors at the Idaho National Engineering Laboratory (INEL) to establish the relationship between the component failure rate increases and maintenance effectiveness. This involved reviewing the NPRDS descriptions of the 500 component f ailures which contributed to the 40 follure rate increases and assigning the cause of each failure to one of five distinct categories:
(1)
Ineffective Molntenance - Failures experienced while conducting, or as a consequence of, maintenance, upkeep, repair, surveillance, testing, and calibration of plant equipment. Examples include personnel errors of omission and commission by molntenance staff, procedure problems re-sutting in inadequate / improper maintenance, problems traceable to maintenance program administrative control, and equipment failures due to improper previous repair.
(2)
Random - Failures of this type usually occur in electronic equipment and are rare in operating equipment. As the term implies, there is no pottern associated with the failure and, therefore, this type of failure would not be expected to be a recurring problem.
(3)
Design / Installation / Construction Failures experienced while performing,
{
or as a consequence of, design, fabrication, construction, and Installo-tion of equipment, systems, and structures. Examples include personnel errors of omission and commission, procedure problems resulting in inode-AEOD/S804B 13
}.., *.
quote or improper design or installation, and problems traceable to de-sign or construction progrom administrative control.
(4)
Normal Aging /Woorout/End-of Ufo - Failures caused by a component or system reaching its end-of-life by normal aging or wearout.
(5)
Unknown - Insufficient information was provided in the failure norrotives to determine the root cause of the failure.
As shown in Figure 3,it was found that over three fourths of the failures involved maintenance ineffectiveness. On a plant-specific basis, the contribution as-cribed to ineffective maintenance ranged from about 25 percent to 100 per-cent.
The strong relation-i ship of these failures to maintenance inef-festiveness hos been ODE EQUlPMENT FAILURE CAUSES confirmedin other ALL PLANTS REVIEWED studies. For example, o trends and potterns analysis was com-pleted by AEOD of NPRDS failure data for Mainggance main feedwater s
]
(MFW) flow control ai y
volves, MFW flow gjjgjjjjjii'
'u f
U."d?" Q
^
/
control bypass volves, f
and MFW pumps in l
U.S. commercial pres-wearout 1
O Ra om Unknown surIZed water reoCiors (PWRs) (Refs.11,12).
The primary finding of this analysis was that (BASED ON NPRDS FAILURE NARRATIVES) differences among plants that could De traced to differences N0**3 in maintenance proc-tices had a greater influence on the failure rate of these components than any of the component design features studied. This result was independently ob-
{
toined, but echoed the results of a 1980 Electric Power Research institute (EPRI) study of MFW pump performance (Ref.13). The EPRI report concluded that the ultimate performance of a major component such as a pump is offected more AEOD/S8048 14 i
l.
f j. ~
~,.
p by how it is maintained than by the selection of a specific pump manufacturer, AEOD has also performed a trends and potterns analysis of moln steam isolation volve (MSIV) failures at both PWRs and BWRs (Ref.14). Bosed on NPRDS dato, the major finding of this analysis was that proper maintenance was a dominant means for minimizing MSIV problems at both PWRs and BWRs.
The second validotion task determined if positive correlations existed in cases where plants with high frequencies of operating events which con be ascribed to maintenance deficiencies also exhibit a high degree of ODE Indicotlon, and whether plants with moderate and low frequencies of maintenance-related events exhibit moderate and low degrees of ODE Indication, respective'ly. The events used in this comparison were those reported to the NRC in LERs. A corre-lotion was found between the condidate indicator and the LER-based mainte-nonce-coused event frequency. This correlation reinforces the conclusion that NPRDS con support a useful maintenance effectiveness indicator. A detailed explanation follows.
Using the historical LER datobose in the Sequence Coding and Search System (SCSS) (Ref.15), the Nuclear Operoflon and Analysis Center (NOAC) of the Ook Ridge National Laboratory (ORNL) developed a technique to classify the causes of the events reported in LERs. One of these causes is maintenance. The classifi-cation technique uses specific search cigorithms to produce the some results as manual cause coding of LERs by experienced engineers. Each event con be categorized by one or more causes.
The rnointenance cause category covers the entire range of programmatic deficiencies related to maintenance, survelliance, testing, and calibration.
These deficiencies are deemed attributable to poor molntenance practices or errors mode by maintenance personnel. The deficiencies include:
(1)
Maintenance oorsonnel errors Personnel errors associated with the per-formance of surveillance, testing, Calibration, or rodlotion protection activities: and (2)
Poor maintenance practices - Equipment failures that are strongly indico-tive of maintenance problems such as improper lubrication corrosion due to boric acid precipitation, short circuits, and improper prior repairs.
To eliminate the effects of the startup of NPRDS reporting on the condidate indicator count, this analysis was applied to those BWRs which begon commer-clot operation prior to January 1,1985. Hence, the number of BWRs considered was reduced from the 28 used in the first part of the condidate indicator volldo-tion to 23. The mean number of maintenance-related events occurring per month during the period of interest at each of these 23 BWRs was calculated
)
m..
AEOD/S804B 15
9
, f
... a 1
[
I:
based on the number of events in the SCSS LER database that involved molnte-nonce deficiencies (i.e., maintenance-related events). This mean molntenance-related event frequency provides some comparative measure of maintenance performance. That is, plants with the highest mean frequency of maintenance-related events seem to experience the greatest difficulty with their maintenance programs compared with other plants. Using similar techniques, the condidate indicator was also calculated.
Using a linear correlation analysis, the degree of association between the candl-date indicator and the mean molntenance event frequency for the 23 BWRs was then examined. The analysis calculated a correlation coefficient between the condidate indicator and the mean molntenance-related event frequency of 0.6. (A correlation coefficient of zero (0) Indicates there is no relationship be-tween the vorlobles. When there is perfect correlation and the vorlobles vary in the some direction,the coefficient is 1.0 (positive correlation). When there is
)
perfect correlation but the vorlobles vary in opposite directions, the coefficient is
-1.0 (negative correlation). The correlation coefficient con vary between the ex-tremes of -1.0 and 1.0 to Indicate some intermediate degree of correlation). This positive correlation was statistically significant of the 0.01 level, Indicating that the correlation was not due to rondom fluctuations in the data. Figure 4 shows how the two variables trend in the some direction. These results illustrate that the indicator correlated acceptably well with LER-based data. Thus, the second part of the validation process was satisfied.
The correlation between the NPRDS-based condidate NPRDS INDICATOR VS LER-BASED DATA indicator and the LER-based MAINTENANCE RELATED EVENTS dato,when overaged over a long period of time,is not T
entirely unexpected since h
the NPRDS failures were j,'
shown to result primarily from osR sasso om N
j j
maintenance ineffectiveness,
+
g and the some finding was
.,{
made for fodures found in
=.. +
, #.a.++
LERs in NUREG-1212,the
?
+"
staff's trend and pattern
,,, ~
A,,.*
.++
onolysis of industry mainte-
+++++ \\
nonce (Ref.16). The indl-NPRDS INDICATOR
,+
lation is quantitative confir-I PLANT rnation of the general relo-
" " ~ " " " ~ ~ " '
tionship of these two sources Figure 4
'~
AEOD/S8048 16
,e
[
.l"'.
1 l:
I for measuring maintenance effectiveness. The correlation reinforces the poten-tlal value of cause codes as on additional source of information for monitoring maintenance effectiveness.
In the construction and validation of the condidate indicator, all failures were used, including those discovered during power operoflon and those discovered during shutdown. No distinction was made regarding the mode of operoflon since all of the events were actual failures and not incipient conditions. Further, the failures most likely occurred during operation, although the discovery of some of the failures could not occur until the plants were shut down. Overall, about half of the failures were discovered during operation and hatt were -
, discovered during shutdown. Ukewise, the failure rate increases that were flogged by the condidate indicator were due to failures that were discovered approximately equally between operoflon and shutdown. In both operation and shutdown, the validation indicated that failure increases showed evidence
' ofineffective maintenance. An aggressive preventive molntenance program would seek to identify and correct problems priorto the occurrence of actual-l failures such as these.
Thus, tracking all reported failures regardless of the plant operational status when the failures were discovered showed merit for indicating molntenance effective-ness. In addition, the use of failures discovered during plant shutdowns will allow gouging of the general condition of equipment entering the outage and the I
potentialforineffective corrective maintenance during on outage. The quality of the maintenance during on outage sets the tone for operoflon in the next cycle. The next section discusses a number of situottor.s wnere increased failure rates due to failures discovered in on outage preceded on equipment forced outage experienced soon offer restart.
RELATIONSHIP WITH EQUIPMENT FORCED OUTAGES Because of the nature of the equipment whose historical data was used in the j
construction and validation of the condidate Indicator,it is a logical hypothesis
)
that there may be a relation between the indicator and the occurrence of
{
equipment forced outages (EFOs). In this study, increases in the component folture rate for a given system are viewed as indicative of the general condition of the system. This analysis postulated on increased chance of an EFO occur-ring, given on observed increase in the failure rate. The usefulness of the candl-date indicator would be enhanced if it provides a more timely indicotlon of the 4
potentialfor on EFO. This analysis examined the operational experience of the 28 plants in detail. The results for the individual plants are contained in the i
proprietary Appendix A of this report.
l l
AEOD/S8048 17
{
3
g...
i This analysis examined historical data that combined NPRDS failure information with forced outage information from NUREG-0020 Ucensed Operating Reactors Status Summary Report (Ref.17). The objective of this effort was to see whether on increase in the rote of reported system component failures preceded on EFO involving that some system. Although this seemed like a reasonable expecto-tion, there are a number of reasons why on increase in the system component failure rote Indicated by the set of dato analyzed might not result In a forced outage. Theseinclude:
(1)
The redundancy of the equipment design in each plant may be such that a specific system con tolerate a number of failures without the plant being required to shut down; (2)
An aggressive maintenance program may have discovered and fixed the problem equipment: ond (3)
A single component failure con result in on EFO with no previous warning.
Therefore,it was recognized that the tie between the condidate indicator and EFOs may not be very strong.
To perform this analysis, component failure records for each of the ODE systems were obtained from the NPRDS (see proprietary Appendix B) and a listing of all the EFOs that were related to the ODE systems was extracted from NUREG-0020 (see Appendix C). The forced outage and equ!pment failure data were com-blned and arranged chronologically for each plant, in this manner, chronolo-gies were assembled from approximately 3,000 component failures and 200 EFOs involving selected equipment in the reactor recirculation, neutron monitoring,
)
control rod drive, feedwater, and main steam systems at the 28 BWRs. The trend in the rote of component failures within each system was examined using plots of cumulative failures as a function of time (months) on which were superim-posed the historical EFOs and the operational history of the plant (i.e., all planned and unplanned outage periods).
recognizing the limitations just listed, the onclysis provided some positive results.
Ten of the 28 plants evoluoted experienced at least one EFO over the three-year period studied which was preceded by on increase in the failure rate of the components within the system that was associated with the forced outoge. The leod times observed for the folture rate increase prior to on EFO generally ranged from two to six months. While these results indicate that there may be o relationship between the condidate indicator and EFOs, this relationship is not very strong.
I e sag AEOD/S804B 1B
,e 4-In general for all of the plants considered, the best results were found for equip-ment in the reactor recirculation, feedwater, and main steam systems. Both the control rod drive and neutron monitoring systems experienced large numbers of fallures, but few EFOs. Each of these two systems is composed of highly redun-dont components and has a capacity to absorb failures up to the limits imposed by technical specifications. These systems / components did not play a major role in this plant onalysis. However, the rate of accumulation of these kinds of failures coming out of a refueling outage could be o measure of the effective-ness of the maintenance performed in the outage.
The current scope limitations of the NPRDS ruled out examining the failure experi-ence for systems such as the main turbine and the moln generator which domh noted the EFO experience at several plants. This factor impacted the number of plants for which results could be demonstrated. Another limitation was that in-ciplent failures are reported voluntarily to the NPRDS. Because of this,such follures were not used in this analysis to ensure that the results would not depend on these failures and, consequenity, be invalidated if licensees had modified their reporting practices during the study time period.
Given these limitations, from the results for some specific plants, the condidate indicator appeared to have some limited potential in providng a warning signal prior to on associated EFO. However, as anticipated,the results did not show on overall statistically strong relationship between the condidate indicator and EFOs. Nevertheless,it appears that the condidate Indicator performed as expected.
FINDINGS AND CONCLUSIONS FINDINGS The major findings of this study are:
(1)
Based on a review of the individual failures in the NPRDS, increased component follure rates within a system are generally associated with maintenance effectiveness; (2)
Detailed engineering studies that employed both statistical and deter-ministic analyses have shown a nexus between ineffective maintenance and NPRDS-reported failures of ODE equipment,l.e., feedwater regulot-ing volves, main feedwater pumps, and MSIVs:
Oh AEOD/S8048 19
o.
l.
[.
I (3)
An equipment forced outage due to a follure of a specific system was sometimes preceded by on increased rate of failure of equipment in that system; I
(4)
The frequency of maintenance problems connected with reportable events showed a positive correlation with the magnitude of the condi-dote indicator for the period onolyzed; and 1
(5)
Implementation of the condidate indicator by the NRC staff on on industry-wide basis would be labor intensive. Consequently, more effi-cient dato techniques need to be developed.
CONCLUSIONS (1)
A practical and useful molntenance performance Indicator was devel-oped using NPRDS dato. This indicator con serve os a suitable basis for describing a maintenance effectiveness tracking method which is oc-ceptoble to the staff in the forthcoming Molntenance Rule regulatory guide. Other indicators could be developed from the NPRDS data.
(2)
The ability of the condidate indicator and the NPRDS data to reflect maintenance effectiveness was confirmed.
(3)
The effect of non-uniform NPRDS reporting con be acceptobly minimized through the use of a standard subset of aquipment that is important to plant operoflons.
(4)
The value of the condidate indicator was confirmed through:
Root cause onclysis:
Independent dato derived from LER-reported, maintenance coused events:
Correlations with other studies; and Correlations with the findings from maintenance effectiveness team inspections.
While the focus of this report is on the use of NPRDS to monitor molnte-nonce effectiveness, the mutually reinforcing correlation between LER-based dato and the NPRDS-based indicator points to the prospect of an AEOD/S804B 20
(.
po odditional maintenance indicator. The LER-based dato used in this cor-relation resulted from the ongoing performance indicator development effort aimed to demonstrate the usefulness of cause codes, one of which is maintenance. Further development of this maintenance cause code from LERs is being pursued for use in monitoring molntenance effective-ness.
(5)
Although the methodology used in this study was developed using dato for 28 BWRs. it should prove equally valid for other plant designs.
(6)
For cost-effective NRC staff use of the condidate indicator on on indus-try-wide basis, further development is necessary to more efficiently ex-tract the indicator dato from the NPRDS and to display it in a manner which permits individual as well as generic comparisons. These efforts will receive high stoff priority and the results of these ocilvities will be shared with Industry.
AEOD/S804B 21
,o i
1 i
REFERENCES i
1,
" Preliminary Results of the Trial Program on Molntenance Performance i
Indicators / AEOD/S804A, Office for Analysis and Evoluotlon of Operational Dato U.S. Nuclear Regulatory Commission, Washington, D.C., October 1988.
1 2.
Letter from John B. Mortin, Regional Administrator, USNRC, Region V, to D. W. Cockfield, Vice President - Nuclear, Portland General Electric Company,
Subject:
Inspection Report No. 50-344/88-30 Enclosure 2.
dated December 2,1988.
3.
Letter from Thon.as T. Martin, USNRC, Region I, to C. A. McNeill, Executive Vice President - Nuclear Philadelphia Electric Company,
Subject:
Com-bined inspection Report Nos. 50-277/88-17 and 50-278/88 17. Enc'esure 2, dated October 11,1988.
4.
Letter from Hubert J. Miller, USNRC, Region til, to Murry Edelman, Presi-dent. Toledo Edison Company,
Subject:
Inspection Report No. 50-346/
88029(DRS). Endesure 2, dated December 16,1988.
5.
The Nuclear Plant Rollability Data System Program Description / INPO 86-010, Institute of Nuclear Power Operations, Atlanto, Georgio,1986.
6.
"NPRDS 7eporting Procedures Manuol/ INPO 84-011, institute of Nuclear Pow 6< perations. Atlanto, Ceorgia,1984 (Proprietary).
7.
Results of this compilation have been published in a series of Electire Power Research Institute (EPRO reports: EPRI-NP-5544, EPRI-NP-4368, and EPRi-NP-3480.
8.
' OPEC-2 System / The S. M. Stoller Corporation,1250 Broadway, New York, New York.
9.
" Generating AvailabHti eport 1982-86/ North American Electric 9
Reliability Council, a svo.on, New Jersey, no publication dote given.
10.
10 CFR 50.73 "Ucensee Event Report System."
11.
Trends and Pottems Program Report - Operoflonal Experience Feedback on Main Feedwater Flow C.. Strol and Main Feedwater Flow AEOD/S804B 22
,,,;ri.3 a.-
h-Bypass Valves and Valve Operators / AEOD/P701, Plumlee 111, G. L, Office for Analysis and Evoluotion of Operoflonal Data. U.S. Nuclear
- Regulatory Commission, Washington, D.C., August 1987 (Proprietary).
12.
. " Technical Evoluotion Report on Operational Experience Feedback for McIn Feedwater Pumps and Their Associated Components '
EGG-NTA-7810, Revision 1, Gentillon, C. D., Meachum, T. R., and Pace, N. E., EG&G-Idaho, Inc., Idaho Falls, Idaho, November 1987 (Proprietary).
13.
'Evoluotion of Basic Causes of Repetitive Failures of Nuclear and Fossil--
Feedwater Pumps," EPRI NP-1571. Electric Power Research Institute, Polo
. Alto, Collfornia, October 1980.
14.
" Technical Evoluotion Report on Operational Experience Feedback for Moln Steam isolation Volves and Their Associated Operators /
EGG-NTA-8126, Fish, L W., Gentillon, C. D., Mecchum T. R., EG&G-Idaho,.
Inc., Idaho Falls, Idaho, November 1987 (Proprietary).
15.
" Sequence Coding and Search System for Ucensee Event Reports /
NUREG/CR-3905, Volumes 14, Nuclear Operations Analysis Center. Ook
' Ridge National Laboratory, Ook Ridge, Tennessee, Ap111985.
16.
" Status of Maintenance in the U.S. Nuclear Power Industry 1985/ NUREG-1212 Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission, Washington, D. C., Volumes 1 and 2. June 1986.
17.
"Ucensed Operating Reactors Status Summary Report,' NUREG-0020, U.S. Nuclear Regulatory Commission, Washington, D. C., published monthly.
AEOD/S804B 23
.- _____ _____