Resolution of Generic Safety Issues: Issue 93: Steam Binding of Auxiliary Feedwater Pumps (Rev. 1) ( NUREG-0933, Main Report with Supplements 1–34 )
This issue was recommended635 for prioritization by DSI after a review of the AEOD engineering evaluation report (AEOD/E325)636 on vapor binding of the AFW pumps at H.B.Robinson Unit 2. Further AEOD study of the event resulted in recommendations which were documented in AEOD/C404.637
The report637 discusses thirteen occurrences reported in 1983 of steam binding of one or more AFW pumps resulting from the leakage of heated main feedwater into the AFW system. The systems are isolated by various combinations of check valves and control valves. The back-leakage occurred through several valves in series. The heated main feedwater, leaking into the AFW system, flashed to steam in the pumps and AFW discharge lines and resulted in steam binding of the AFW pumps.
Operating experience to date includes 22 events of reported back-leakage in 6 operating PWRs in the USA and at 1 foreign reactor. In other cases, backleakage has been observed but was not considered as reportable occurrences.
The potential for common mode failure is present whenever one pump is steam-bound because the pumps are connected to common piping with only a single check valve to prevent back-leakage of hot water to the second or third pump. Steam binding of more than one pump was reported to occur in 3 of the 13 events reported in 1983.
Although steam binding of the pumps was reported on only W designed plants, a back-leakage event is believed to have rendered an AFW flow sensor inoperable at Crystal River, a B&W-designed plant. The actual operating status of the pump and train during this event remains unknown. However, the AFW system in all PWRs is sufficiently similar so as to consider it a generic problem for all PWRs.
The back-leakage of steam represents a potential CCF for the AFW system that could result in the loss of its safety function.
AEOD has recommended637 that regular monitoring of the temperature of the AFW pumps be implemented to provide early detection of back-leakage of main feedwater. This will permit bleeding off the heated water and/or steam before acute steam binding of the pumps can occur. The addition of a pyrometer on the AFW discharge line at or near the pump would permit monitoring of the temperature of the fluid in the system by the plant operators during their routine visual inspections. Records of the temperature readings would show the onset of leakage at an insidious level. Trends of temperature rise times would also provide for the determination of optimal reading and recording intervals which would provide adequate assurance of system availability. The use of a pyrometer would reduce the possibility of error resulting from estimating the temperature by the operator placing his hand close to the auxiliary feedwater pumps or discharge lines.
The events experienced in 1983 are considered typical even though the number of events reported annually (prior to 1983) are less. The reporting of back-leakage is only required in those cases in which the pump has been rendered inoperable. Back-leakage which may have been detected on the steam bled from the system before a pump was rendered inoperable might not be considered a reportable occurrence. In fact, it is believed likely that the number of back-leakage events exceeds the number of events reported in 1983 and prior years. However, for this analysis 13 events will be used as the annual occurrence frequency of back-leakage events.
In the calculations, all plants will be assumed to have three auxiliary feedwater pumps although some may have two. The effect of this assumption will be that the total unavailability of the auxiliary feedwater system for those plants having only two pumps will be about 50% lower than the actual unavailability. However, due to the small number of plants having only two pumps, this error is not expected to significantly impact the results.
There were 13 events of pump unavailability reported in 1983. Based upon an expected 15 system demands/RY, 3 pumps/plant, and 47 plants, the unavailability/pump-demand (Q) is calculated as follows:
Q = 13/(47 x 3 x 15) = 6.1 x 10-3
A second pump failure occurring simultaneously was reported to have occurred in 3 of the 13 events. The failure of a second pump is then expected to occur 3 times in 13 events, or at an occurrence rate of 3/13 or 0.23. Assuming that given two pumps having become steam bound, the conditional probability that the third pump will also become steam bound is 0.1 results in a demand unavailability of all 3 AFW pumps of 1.4 x 10-4.
The original prioritization was based upon the Sequoyah RSSMAP54 study. This analysis had a TML sequence which led to core melt and the dominant containment failure mode was due to hydrogen burning. It is the belief of many in the PRA risk analysis field that the TML sequence will not lead to core-melt, and that the probability of containment failure due to hydrogen burning may be reduced by orders of magnitude. Further, to assume that the Sequoyah containment (an ice condenser) can be utilized in generic calculations may not be valid. Hence, the consequences were reexamined using the results of the Reactor Safety Study16 (RSS) and the Surry containment.
In the RSS for Surry, the unavailability of the AFW system was calculated to be 1.5 x 10-4/demand which did not include steam binding of the AFW pumps. The major sequence affected is the TMLB- sequence which is increased from 3 x 10-6/RY to 5.8 x 10-6/RY by the addition of steam binding to the AFW unavailability. In addition, a very small contribution is made by a TML sequence.
The PWR release categories are as defined in the RSS. The whole body man-rem dose is obtained by using the CRAC code64 assuming an average population density of 340 persons per square mile (which is the mean for U.S. domestic sites) from an exclusion area of a one-half mile radius about the reactor out to a 50-mile radius about the reactor. A typical midwest plain meteorology is also assumed. Based upon these assumptions, the public dose resulting from each category is as follows:
|Release Category||Dose (man-rem)|
|1||5.4 x 106|
|2||4.8 x 106|
|3||5.4 x 106|
|5||1.0 x 106|
|6||1.5 x 105|
|7||2.3 x 103|
The steam binding of the AFW pumps will increase the frequency of the following listed sequences in the categories shown resulting in the listed dose.
|Category||Sequence||Frequency Increase (RY-1)||Dose (man-rem/RY)|
|1||TMLB -||2.8 x 10-8||1.5 x 10-1|
|2||TMLB -||1.9 x 10-6||9.12|
|TMLB -||6.5 x 10-7||3.12|
|3||TML -||5.6 x 10-8||3.0 x 10-1|
|5||TML -||2.8 x 10-10||3.0 x 10-4|
|6||TMLB -||5.6 x 10-7||8.4 x 10-2|
|7||TML -||5.6 x 10-6||1.3 x 10-2|
Considering only the TMLB1 sequences the resulting dose is 12.5 man-rem/RY. The TML sequences are excluded due to the present uncertainty regarding core-melt of this sequence. For the 90 PWRs which are expected to be operating having an average life of 28.8 years, the total public dose will be 3.2 x 104 man-rem.
The assumed probability of 0.1 for the third pump failing from steam binding, given that two have failed, may not be conservative, but rather may be overly optimistic. If it is assumed that of the three events, where 2 pumps were reported to have been steam bound, that one event also involved 3 pumps, then the public dose risk would increase by a factor of 3 to the value of 9.6 x 104 man-rem.
Industry Cost: The cost estimate was based upon a number of engineering assumptions which are believed to be conservatively biased toward the high side of the costs involved. Equipment costs for the pyrometers are estimated to be $7,500/plant ($2,500 each); the selection, installation design, ordering, installation and test were estimated to be 10 person-weeks/reactor, or $22,700. No increase in operating cost is calculated. It is believed that the reading and recording of the temperature of the AFW pumps can be included as part of the plant surveillance activities which are normally accomplished each operating shift. Test and maintenance costs were estimated to be 1 man-week/RY. For the 47 backfit reactors with an average remaining life of 27 years, the maintenance costs total $2.9M. For the 43 forward-fit reactors having a life of 30 years, the maintenance costs total $2.9M. It is further estimated that each pyrometer will be replaced twice during the plant life at a cost of $32,000/plant. The total industry cost to install pyrometers at or near each pump, based upon the above, is $11.4M.
NRC Cost: The NRC cost is estimated to not exceed 1 man-week/reactor or $0.2M for all affected plants.
(1) For the scenario in which the probability of the third pump failing steam bound is 0.1, the value/impact score is given by:
(2) For the scenario in which the probability of the third pump failing steam bound is 0.33, the value/impact score is given by:
Both the total dose in man-rem and the value/impact score vary from bordering between medium to high priority for the third pump failure of 0.1 to high priority for the third pump failure, if the third pump failure were 0.33. In light of the uncertainty associated with this issue, a HIGH priority was assigned.
In October 1985, IE Bulletin 85-011112 was issued to licensees with requirements to develop procedures to detect or correct steam binding. In resolving this issue, the staff surveyed the back-leakage experience in operating plants following the implementation of monitoring procedures. Although the number of back-leakage events varied from an average of less than 1/RY at a large majority of plants to more than 100/RY at others, none of the back-leakage events that occurred during the review period resulted in the steam binding of an AFW pump. This indicated that the various monitoring methods employed can be highly effective in preventing steam binding if back-leakage occurs. For the plants with a high back-leakage event rate, the installation of continuous monitoring systems with contol room alarms was instrumental in providing for early warning to the operator and timely corrective action.
The results of the staff's regulatory analysis indicated that the following recommendations in IE Bulletin 85-011112 would ensure that the contribution of AFW pump steam binding to core-melt frequency and public risk was sufficiently low and that there was no need for new recommendations beyond those in IE Bulletin 85-01. Thus, the staff concluded that the recommended monitoring actions of IE Bulletin 85-01 should be continued. In February 1988, Generic Letter 88-031113 was issued to reinforce this conclusion. Thus, this issue was RESOLVED and requirements were established. The staff also agreed to revise NRC Inspection Procedure 71707-03C to include the matter of monitoring the AFWS pumps for steam binding as an example of a recurring operational event that should be periodically checked by the NRC inspectors.1114