Resolution of Generic Safety Issues: Issue 65: Probability of Core-Melt Due to Component Cooling Water System Failures (Rev. 1) ( NUREG-0933, Main Report with Supplements 1–34 )
Increasing attention has recently been focused on the integrity of RCP seals due to loss of cooling to the seals both in connection with PRA studies which are currently under review and from operating experience.27,404 The frequency of core-melt due to failures in the component cooling water system (CCWS) has been considered in PRAs associated with the Zion, Indian Point, and Sizewell plants. It is also considered in French studies in their Fessenheim plants and in their PRA on their 1300 MWe plants.405
The CCWS is relatively simple in operation. It circulates cooling water to a wide variety of equipment in the plant then rejects the heat accumulated in this water to the plant service water via heat exchangers. The CCWS is a closed system. Thus, it acts as a barrier between potentially contaminated systems and the raw water in the service water systems. The CCWS provides cooling to a great variety of equipment, both to process streams and also for auxiliary needs such as oil coolers. The plant will not operate for more than a few minutes without component cooling water. Consequently, the CCWS is designed to be highly reliable. More importantly for current purposes, the CCWS at many plants services most of the active engineered safety features. This portion of the CCWS is, therefore, required to be safety grade. Nevertheless, CCW has a finite failure probability.
The issue of concern here is a common failure which simultaneously causes a small LOCA and renders most of the ECCS inoperable. Loss of CCW will immediately cause loss of cooling water to the RCP thermal barrier heat exchangers. Loss of CCW at some plants will also render inoperable the charging pumps which usually supply the water for RCP seal injection. With neither cooling mechanism available, the RCP seals are expected to fail within a short period of time (approx. 30 minutes)405; but loss of CCW may also render the high pressure safety injection (as well as containment spray) inoperable. Eventually, if no manual mitigating actions are taken, the core will uncover and melt. Moreover, unless the plant has containment fan coolers which do not use CCW, the containment might eventually overpressurize and fail.
The solution which has been proposed for this issue is the addition of a steam turbine driven charging pump, analogous to a BWR HPCI and RCIC.405 Such a pump could maintain seal injection and thus greatly reduce the probability of seal failure, independently of the availability of CCW or AC power.
It should be emphasized that, because the study of this issue is still in its early stages, other solutions may surface later.
The core-melt frequency due to this event sequence has been estimated in several probabilistic risk analyses. The numerical value of the estimate ranges from 2 x 10-5 to 1 x 10-4 event/RY.405 These frequencies are plant-specific. Thus, this range is not a range of uncertainty in the estimate but, instead, is a measure of plant-to-plant variation. (The details of CCWS design vary significantly.) We will use the high value of 10-4/RY, recognizing that this figure will not apply to all plants. This estimate does not include the probability of station-blackout, which is another mode of failure.
Once the core melts, the consequences vary greatly depending on the mode of containment failure. It should be noted that loss of CCW may disable the containment sprays.
Loss of CCW is similar to loss of AC electrical power to the engineered safety features in that both may disable the ECCS (accumulators excepted) and (at some plants) the containment spray. Thus, for the loss of CCW sequence, we will assume the same distribution of containment failure modes as was assumed in the WASH-140016 S2B sequences:
|PWR-1||1%||1 x 10-6||5.4 x 106||5.4|
|PWR-2||,||15%||1.5 x 10-5||4.8 x 106||72|
|PWR-6||84%||8.4 x 10-5||1.5 x 105||12.6|
Therefore, the estimated total risk is 90 man-rem/RY. The proposed fix should be at least 90% effective in preventing core-melt. Thus, the net risk reduction is estimated to be about 80 man-rem/RY.
The preliminary proposed solution to this problem is the addition of a steam-driven charging pump which would be independent of both component cooling water and AC electrical power. Based on experience with steam driven auxiliary feedwater pumps, we estimate that such a modification, fully safety grade, would cost about $15M per plant. NRC costs are negligible in comparison to this figure and thus are not included.
Based on a risk reduction of 3 x 103 man-rem/reactor and a cost of $15M per reactor, the value/impact score is given by:
The frequency of core-melt is derived from the frequency of loss of CCW and the probability of core-melt after the seal fails. We will assume that the frequency of loss of CCW is uncertain to a factor of 10. Uncertainties in the two probabilities are one-sided, since the point estimate405 is a probability of one.
The probability of seal failure after loss of CCW is governed by the time it takes for at least one seal to fail versus the time it takes for the operator to diagnose the problem and restore seal cooling.
The probability of core-melt is similar in that it is a balance of time required to rig a means of injecting water versus the time available before the core melts. Although there has been considerable discussion of the time to seal failure, the time to core-melt has not been investigated in detail. (This would require a calculation of primary pressure and water inventory versus time.) The two probabilities (seal failure and core-melt) are not independent. We will assume that, given loss of CCW, the net probability of core-melt is at least 50%.
The uncertainty in consequences is a combination of the uncertainty in the calculation of man-rem and the uncertainty in the mode of containment failure. We will assume a factor of 5 for the calculational uncertainty. The distribution among the various failure modes is still a subject of discussion in the PRA field. However, it is unlikely that this distribution will introduce more than a factor of two in uncertainty.
The costs are based on actual plant experience, albeit with a different system. We will assume that a factor of 5 will bound the uncertainty in this figure.
Assuming long normal distributions, the above numbers imply the following bounds on the priority parameters:
|Man-rem/Reactor||3 x 103||2 x 102 - 5 x 104|
|Core-melt/RY||9 x 10-5||8 x 10-6 - 9 x 10-4|
|Value/Impact Score, S||200||7 - 5 x 103|
The lower bounds of the priority parameters are in the medium priority category and the "best" estimates are well into the high category. Therefore, this item was classified as high priority. However, because of the relationship between CCW system failure and RCP seal failure, Issue 65 was integrated1000 into the resolution of Issue 23 in October 1983.