Resolution of Generic Safety Issues: Issue 23: Reactor Coolant Pump Seal Failures (Rev. 1) ( NUREG-0933, Main Report with Supplements 1–34 )
This issue addressed the high rate of reactor coolant pump (RCP) seal failures that challenge the makeup capacity of the ECCS in PWRs. At the time this issue was identified27 in 1980, RCP seal failures in BWRs occurred at a frequency similar to that experienced in PWRs. However, operating experience indicated that the leak rate for major RCP seal failures in BWRs was smaller. The smaller leak rate, larger RCIC, HPCI, and feedwater makeup capabilities, and isolation valves on the RCP loops negated the potential problem in BWRs. The three main PWR RCP manufacturers had their own seal designs that were developed throughout the years:
BYRON-JACKSON supplies RCPs for the B&W and CE reactor systems. For a B&W system, pumps are supplied with three equally staged seals. For a CE system, the pumps are supplied with four seal stages: three stages are equally staged and the fourth stage is used as a vapor seal.
BINGHAM originally had only two stages in their RCP design. At the time of the identification of this issue, the latest Bingham seal design, developed for pumps in B&W reactor applications, used three equally staged seals.
WESTINGHOUSE used a three-stage seal design. The first seal stage takes the full system pressure, reducing the pressure from 2250 psi to 50 psi. The second stage is designed to take full system pressure in case of first-stage failure. The No. 3 seal is a vapor seal and operates at a pressure of not more than 5 psi.
The results reported in WASH-140016 indicated that breaks in the reactor coolant pressure boundary having an equivalent diameter in the range of 0.5 to 2 inches were a significant cause of core-melt. Since then, a staff study27 showed that comparable break flow rates have resulted from RCP seal failures at a frequency about an order of magnitude greater than the pipe break frequency used in WASH-1400.16 It was believed that the overall probability of core-melt due to small-size breaks could be dominated by events such as RCP seal failures.27
It was believed that development efforts could be undertaken to supply much of the missing information and thereby provide a basis for new design specifications to obtain higher reliability in future seal designs. EPRI NP-1194113 described a program to provide the information and physical insights necessary to make future RCP design considerably more reliable than existing designs. Such a program would include improved pump design, improved seal design, improved maintenance procedures, and improved seal auxiliary support systems. In addition, consideration of and coordination with Issue 9 (Reactor Coolant Pump Trip Criteria) was expected to provide a broader-based perspective of the RCP operational needs, performances, and requirements.
An apparent solution, but not necessarily the best solution,114, 195 was to replace each RCP seal annually. This solution was to be used to provide a cost estimate. The cost estimate, based on more frequent seal replacement, should bound an effective development program, or perhaps exceed the cost of needed (improved) maintenance and seal replacement procedures combined with improved instrumentation to detect incipient RCP failures.
The dominant accident sequences which follow a small-break LOCA, equivalent to a pipe break range of 0.38 to 1.2-inch diameter piping, was assumed to be representative of a RCP seal failure. The representative modeling provided by PNL64 followed the RCP seal failure analysis used in the ANO-IREP study366 in which thirty accident sequences were modeled in the RCP seal failure event tree. Assuming an RCP seal failure as the initiating event, two of the thirty accident sequences dominated the potential core-melt frequency. The two dominant accident sequences were: (1) failure of the high pressure injection system (D1); and (2) failure of the high pressure injection system (D1) and failure of the reactor building spray injection system (C). In both cases, containment failure was predicated by one of the following: vessel steam explosion (); containment overpressure due to hydrogen burn (); penetration leakage (); or base mat melt-through ().
The (D1) failure assumed that the emergency signal will not be generated prior to core uncovery. In this case, the analysis366 calculated that the pressurizer heaters could remain covered for an extended period and thus maintain RCS pressure above the emergency signal actuation set point. In the interim, the makeup (MU) tank could empty, resulting in loss of suction and failure of HPI/MU pump.
The (D1C) sequence was similar to the (D1) sequence except that the reactor building spray injection system (C) is also unavailable due to failure(s) which are common to the suction paths of the HPI pumps and spray pumps. In this sequence, all five (3 HPI plus 2 spray) pumps that take suction from the borated water storage tank could fail because of a failure of a single manual valve which is in series with two parallel MOVs.
RCP Seal Failure Frequency: The RCP seal failure frequency of 2 x 10-2/RY was used in the analysis.64, 366 This frequency represented a generic RCP seal failure frequency for major RCP seal failures that may challenge the ECCS (leaks > 50 gpm/pump). Plant-specific RCP seal failure frequencies may have been higher or lower than this generic frequency. The overall RCP seal failure frequency, including smaller leaks, was approximately (0.5/RY).114, 195 This meant that, on an average, each PWR experienced a RCP seal failure biannually and that, during a 40-year design life, each PWR could experience one major RCP seal failure that would challenge the ECCS.
Core-Melt Frequency: Two core-melt frequencies were provided in the ANO-IREP Study.366 One frequency estimate took no credit for operator recovery actions. The second estimate factored in potential recovery of failed systems. The recovery model basically considered three steps: (1) recoverability of the fault; (2) location of the fault; and (3) the critical recovery time for restoration of the component function.
The base case core-melt frequencies without recovery (W/O) and with recovery (R) for the two dominant accident sequences were:
D1(W/O) = 2.5 x 10-5/RY
D1C(W/O) = 2.0 x 10-5/RY
D1(R) = 2.8 x 10-6/RY
D1C(R) = 4.4 x 10-6/RY
Assuming a potential reduction in the major RCP seal failure frequency of 50% (1 x 10-2/RY), the above core-melt frequencies would be reduced by a factor of 2.
The major RCP seal failures contribute to 6 of the 7 PWR release categories. Based on a potential 50% reduction in RCP seal failure, the following table lists the affected PWR release categories, frequency reductions, and public risk (man-rem/RY) considered with and without recovery actions.
|Category||Core-melt Frequency (/RY)||Risk (Man-rem/RY)|
|PWR-1||2.3 x 10-9||3.6 x 1-10||12.1 x 10-3||2.0 x 10-3|
|PWR-2||1.2 x 10-5||1.8 x 10-6||5.5 x 101||8.7 x 100|
|PWR-4||0.7 x 10-7||1.5 x 10-8||11.4 x 10-1||4.2 x 10-2|
|PWR-5||0.9 x 10-7||1.0 x 10-8||0.9 x 10-1||1.0 x 10-2|
|PWR-6||0.5 x 10-5||1.1 x 10-6||7.0 x 10-1||1.5 x 10-1|
|PWR-7||0.7 x 10-5||0.7 x 10-6||3.0 x 10-2||1.5 x 10-3|
|TOTAL:||2.4 x 10-5||3.6 x 10-6||5.77 x 101||8.8 x 100|
|AVERAGE:||1.36 x 10-5||3.3 x 101|
At the time of this evaluation in 1983, the average remaining life for 47 backfit and 43 forward-fit PWRs was 28.8 years which yielded a total of 2,592 RY. Thus, the total potential public risk reduction by reducing the RCP seal failure frequency 50% ranged from 2.28 x 104 to 14.95 x 104 man-rem. From the above values, it was apparent that operator recovery actions were important to public risk reduction.
A review of the ANO-IREP study366 on RCP seal failures indicated that, in some sequences, non-conservative leak rates assumed in the IREP study could have resulted in an overestimation of the time available for an operator to take corrective action. Therefore, too much benefit may have been credited for recovery actions. In addition, if leak rates of 70 to 300 gpm were considered, as evident from some major RCP seal failures, the number of dominant accident sequences would most likely increase. Based on these observations which were to be confirmed in more detailed staff reviews, the potential public risk reduction was estimated to be 8.6 x 104 man-rem.
Industry Cost: A scheduled annual replacement of the RCP seals would involve TS changes. This one-time cost was estimated to be (2 man-weeks/plant)($2,270/man-week) or $4,540/plant. The license amendment fee for backfit plants was assumed to cost $12,300/plant. Considering 47 backfit plants and 43 forward-fit plants (no additional amendment fee), the industry cost was estimated to be $0.99M.
Operation and maintenance costs included labor and equipment (seals). The labor cost, assuming 300 man-hour/pump seal431 and an average of 3.7 pumps/plant at a rate of $2,270/man-hour, was $63,560/RY. At $57,000/pump for 3.7 pumps/RY, the equipment (seals) would cost $210,900/RY. Therefore, the total estimated cost for annual replacement of all RCP seals for 90 PWRs over an average remaining life of 28.8 years was $710M.
From the above estimates, the dominant cost was attributed to the equipment (seals) cost ($547M). No outage (replacement power) costs were assumed since the seal replacements would be part of a planned outage schedule.
NRC Cost: The NRC cost was based on a flat rate of $2,270/man-week times the estimated number of man-weeks involved in the issue. The generic resolution was assumed to require 52 man-weeks of effort. Support for implementation of the resolution was estimated to be 2 man-weeks/plant. Annual review of the operation and maintenance and related concerns was estimated to be 0.2 man-week/RY. Therefore, the total NRC cost was estimated to be $[52 + (2)(90) + (0.2)(90)(28.8)][2,270] or $1.7M.
Total Cost: The total industry and NRC cost associated with a possible solution to the issue was estimated to be $[710 + 1.7]M or approximately $712M.
Based on a potential public risk reduction of 8.6 x 104 man-rem and an estimated cost of $712M for a possible solution, the value/impact score was given by:
(1) The implementation cost impact based on annual seal replacements should bound a more effective resolution. A more effective solution would result in a greater cost benefit and lower ORE increases than annual seal replacements.
(2) Based on information in EPRI NP-2092114 and a staff report,195 the overall RCP seal failure (major and minor seal failures) frequency was ~0.5/RY. If this failure frequency and resultant unplanned outages were reduced by a factor of 2, the industry could realize a cost savings. Assuming 10 days per forced outage at a replacement power cost of $300,000/day for 90 plants over 28.8 years, the potential industry cost savings was estimated to be:
1/2 $[(5 x 10-1)(10)(3 x 105)(90)(28.8)] = $1,940M
(3) The total industry and NRC combined implementation cost of $712M was overwhelmed by the potential industry cost savings of $1,940M. Based on the above estimates, a resolution of this issue leading to a 50% reduction in RCP seal failures would result in a total combined cost benefit of approximately $1,200M.
(4) Annual replacement of all RCP seals would increase ORE. Based on information in EPRI NP-1138,431 the average ORE for one pump seal replacement was 7 man-rem. Assuming an average of 3.7 pumps/plant/year over 28.8 years for 90 reactors yielded an ORE of (7)(3.7)(90)(28.8) man-rem or 6.7 x 104 man-rem. Assuming a potential reduction of 50% in RCP seal failures for the existing failure rate of 0.5/RY provided an ORE reduction of (0.5)(0.5)(7)(90)(28.8) man-rem or 4.5 x 103 man-rem. The net ORE for annual seal replacement was an increase of 6.3 x 104 man-rem or approximately 24.3 man-rem/RY. This potential increase in ORE and operating experience which showed that more frequent RCP seal replacements were ineffective in reducing RCP seal failures195, 114 indicated the need for a more effective solution than annual RCP seal replacements.
Depending on the assumptions, this issue could have been given a low priority ranking. However, this conclusion would have been based on an optimistic assessment of operator response, the high cost and large ORE incurred by annual seal replacement, and the exclusion of the down-time and associated costs due to minor seal failures.
If operator response is poor, then the potential reduction in public risk would justify a high priority. Other solutions may have been less costly and may have incurred less ORE, two factors which would have markedly improved the value/impact ratio. Any reduction in the overall RCP seal failure frequency would reduce unscheduled shutdown and the high associated costs. Therefore, based on the potential for a large reduction in public risk, the belief that better solutions could be found, and the potential for significant saving of replacement power cost, this issue was given a high priority ranking (see Appendix C) in November 1983. Prior to this, Issue 65 was integrated into the resolution of Issue 23.
In resolving the issue, the staff elected to pursue plant-specific backfits based on the staff's plant-by-plant risk analysis of the loss of component cooling water/essential service water systems. The staff also committed to work with the industry to develop additional RCP seal models to support future risk-informed licensing decisions. Thus, the issue was RESOLVED with no new or revised requirements1763 and licensees were informed of the staff's conclusion in NRC Regulatory Issue Summary 2000-02.1767