Resolution of Generic Safety Issues: Issue 85: Reliability of Vacuum Breakers Connected to Steam Discharge Lines Inside BWR Containments (Rev. 2) ( NUREG-0933, Main Report with Supplements 1–34 )
In BWRs, SRVs are mounted on the main steam line inside the drywell. Each SRV discharge is piped through its own discharge line (tailpipe) to a point below the minimum water level in the primary containment suppression pool. A vacuum breaker (VB) valve is installed on each discharge line to admit drywell air into the discharge line after SRV actuation and closure. This prevents a vacuum from forming due to the condensation of leftover steam in the discharge line. Water in the suppression pool then will not be drawn up into the line. The VB valve is similar to a swing check valve with a disk that swings on a hinge pin to open, and a spring to return the disc to a closed position.
VB valves are also mounted on other steam lines (the HPCI and RCIC turbine exhaust lines) that also discharge into the primary containment pressure suppression pool. However, the VB valves on these lines admit air from above the suppression pool (i.e., wetwell airspace) rather than drywell air.
Review of recent LERs has shown several instances in which VB valves from various vendors and in different plants have failed to properly operate, indicating a potential generic problem. Based on the information provided for this safety issue,552 it appears that the cyclic impact of the disc on the VB valve seat due to steam discharge and condensing during SRV actuation or leakage presents a loading condition that has not been addressed in approved VB valve design and qualification requirements.
Failure of a SRV discharge line VB valve in the open position when combined with the failure of its SRV in the open position would result in an extended steam release to the drywell and would present the control room operators with a confusing set of indications, i.e. simultaneous indications of a stuck open or leaking SRV and a LOCA in the drywell. Several events have been reported in which the above scenario was evidenced. The most notable was the event at Hatch Unit 2 on August 25, 1982 which was evaluated in AEOD/C403.795 Similar events that occurred at Browns Ferry Unit 1 and Peach Bottom Unit 2 provided the basis for the issuance of IE Information Notice No. 83-26.789 The events at these 3 plants were also evaluated in AEOD/E322.791 Misinterpretation and confusion on the control room operators part would be expected to increase the probability of operator error in the course of response to the event and might result in an increased likelihood of the event escalating into a severe core damage accident. Steam release to the drywell may also affect safety-related valves and instruments as well as increase the drywell temperature and pressure.
Failure of a SRV discharge line VB valve in the closed or near closed position when combined with a second actuation of its SRV could result in hydrodynamic loads on the discharge line pipe and to the suppression chamber in excess of the design loads or could cause water hammer damage to the SRV. If the hydrodynamic loads are severe enough failure of the SRV discharge line or the suppression pool could occur with some probability that these failures would lead to a severe core damage event. Damage of the SRV could also lead to a greater frequency of challenges to the reactor protection system and the containment. Evaluation of the available LERs has not revealed a closed or near closed VB failure. Since a possible outcome of closed or near closed failure of a SRV VB valve would be to bypass the containment pressure suppression pool and/or loss of containment integrity, this failure mode is evaluated in Issue 61, "SRV Discharge Line Break in the BWR Wetwell Airspace of Mark I and II Containments."
Failure of HPCI or RCIC turbine exhaust line VBs in the closed or near-closed position when combined with multiple actuations of the turbine-driven pump could also result in: (1) excessive hydrodynamic loads on the turbine exhaust line or the containment wet well structure; or (2) water hammer damage to the HPIC or RCIC turbine system. Failure of HPCI or RCIC turbine exhaust line VBs in the open or leaking position could result in a suppression pool bypass leak to the wetwell air space and present the possibility of excessive containment pressurization. Since both the open and closed failure modes of the HPCI/RCIC turbine exhaust line VBs could result in bypass of the containment pressure suppression pool and/or loss of containment integrity, failure of HPCI and RCIC turbine exhaust line VBs is also evaluated in Issue 61. As a result, Issue 85 was limited to only the effects of SRV VB leakage failures.
The reliability of VBs connected to SRV discharge lines inside BWR containments was assumed to be improved by development of NRC-approved design criteria for VB valves, modification of existing valve design(s), and a prototype qualification testing program(s). Licensees were assumed to replace the existing VB valves with valves of the new qualified design. In addition, a TS that would require periodic operability testing of the VB valves was assumed. All BWRs using VBs connected to SRV discharge lines inside containment would be affected by this issue. The resolution of this issue was applicable to all BWRs.
The evaluation of this issue was based in part on an analysis performed by PNL64 and in part on an analysis by the staff. Since the issue in question applied only to BWRs, the Grand Gulf 1 PRA was utilized as the basis for estimating the risk reduction which might be achieved by the assumed resolution of the issue. This issue affected 44 BWRs with an average remaining lifetime of 27.4 years.
For the case in which the SRV discharge line VB valve was assumed to fail open (i.e., the valve will not reseat under SRV discharge flow or the disc is lost), significant pressurization and steam accumulation in the drywell will only occur for a prolonged SRV discharge. Thus, this effect will only be seen for those events for which the SRV does not properly reseat. Of the Grand Gulf dominant risk sequences, only the T•P•Q•I and T•P•Q•E sequences are affected by the failure of an SRV to reseat (i.e., event P). T is the frequency of a reactor transient (7.2/RY), P is the probability of SRV failure to reset (0.1/demand), Q is the probability of failure of the power conversion system (1/demand), I is the probability of failure of the residual heat removal system to remove decay heat from the suppression pool within 28 hours (8 x 10-5/demand) and E is the probability of failure of emergency core cooling (1.2 x 10-5/demand). Prolonged SRV release to the drywell through a failed open VB is assumed to result in a severe core damage event, through control room operator errors. Exposure of safety related instruments and equipment in the drywell was not considered because the environmental design loads for safety related equipment in the drywell exceed the expected environment following this event. Since containment failure due to H2 Burn () or steam explosion () is independent of the location of the SRV steam discharge to the containment, the T•P•Q•I• and T•P•Q•E• sequences were eliminated as affected cut sets. This leaves only the T•P•Q•I• and T•P•Q•E• scenarios where represents the probability of long-term containment failure due to steam or non-condensable vapor overpressure. Examining these scenarios reveals that in the individual cutsets the only events requiring control room operator actions are the RECOVERY probability for the T•P•Q•I. sequence and the OP probability for the T•P•Q•E• sequence. In the Grand Gulf PRA, RECOVERY is defined as failure to restore maintenance/test faults or to take other corrective actions within 28 hours and is a sub-element of event I. Since confusion over whether a LOCA had occurred or whether a SRV and VB were stuck open or leaking is assumed to hamper only the initial diagnosis of the event and very early response actions, we assumed the probability of RECOVERY would be unchanged by an open failure of the VB; therefore, the T•P•Q•I• sequence was also eliminated as an affected dominant risk sequence.
In the T•P•Q•E• sequence, OP is defined as the failure of the operator to manually initiate the automatic depressurization system (ADS)and is a supplement of event E. Since the OP event is a short-term operator response, we assumed control room confusion resulting from failure of the VB in the open position combined with a stuck open SRV would increase the likelihood of the OP event by an order of magnitude from 0.0015 to 0.015/demand.
Analysis of the VB failure data, provided in the request for classification of this concern as a generic issue by PNL, resulted in a calculated VB failure rate (X) of 0.0093/demand. Since this issue involves a failed open VB in conjunction with a malfunctioning SRV and the other failures, the affected dominant sequence becomes T•P•X•Q•E•.
The Grand Gulf PRA did not assume VB failure in the development of the event trees resulting in severe core damage. Therefore, we calculated a modified base case frequency for the T•P•X•Q•E• scenario of 2.43 x 10-8/RY. Assuming that resolution of the issue (i.e., improving VB reliability) will result in an order of magnitude reduction in the VB failure rate (i.e., X* = 0.00093/demand), the post implementation frequency of the T•P•X*•Q•E• scenario is 2.43 x 10-9/RY. Therefore, resolution of this issue would result in a reduction in the frequency of offsite release from a damage event of 2.18 x 10-8/RY and a reduction in core-melt frequency of 4.36 x 10-8/RY.
The T•P•Q•E• event results in a BWR Category 4 offsite consequence (6.1 x 105 man-rem/event). Multiplying the reduction in dominant risk event frequency (2.18 x 10-8/RY) by the Category 4 consequence (6.1 x 105 man-rem/event), the number of applicable plants (44), and their average remaining lifetime (27.4 years) results in a public risk reduction of 16 man-rem. This is the reduction in public risk and frequency of core-melt which might be provided by the improvement of SRV VB reliability for the fail open mode of VB failure. Assessment was utilized as the basis for estimating the risk reduction which might be achieved by the assumed resolution of the issue.
The solution to this issue was assumed to result in a generic program for the development of staff-approved design criteria for the SRV VB valves, modifications of the valve design, and a qualification program. The existing valves would be replaced and a periodic operability test would be required.
Industry Cost: The estimated costs for industry implementation of the solution to this issue (i.e., replace the valves and test them periodically) was estimated by PNL, based on conversations with Browns Ferry staff involved in their recent testing and maintenance effort on the VB valves at their plant.
PNL estimated the purchase of new VB valves and initial operability tests to total about $54,000/plant, for a cost of about $2.4M to the industry.64 An additional $500,000 was added to this cost to cover industry funding of a VB valve generic qualification program. Thus, the total estimated industry cost for implementation of this issue was $2.9M.
The operability test requirement will require an estimated 0.5 man-day/plant-year of industry effort over the remaining lifetime of the BWR plants, or a total testing cost of about $275,000. Thus, total industry cost was estimated to be $3.175M for resolution of this issue.
NRC Cost: It was assumed that the development of design criteria and establishment of an operability test for VBs would be sponsored by the NRC. It was assumed that a major portion of dynamic model development and engineering data was already available for use in establishing an adequate VB design criteria. The NRC effort was estimated to require 1 man-year for the development of the resolution of this issue for a cost of $100,000.
NRC review of industry implementation of the resolution of the issue, i.e., selection and installation of a new VB valve and TS for the operability testing of the valve, was assumed to require 1 man-week/plant of staff effort at a cost of $2,270/man-week or $100,000 total cost.
About 0.1 man-wk/RY was estimated for NRC review of periodic operability tests for VB valves. At $2,270/man-wk, this resulted in an estimated cost of about $275,000 over the remaining lifetime of the 44 BWRs. The total NRC cost for the resolution and implementation of the issue was estimated to be about $475,000.
Total Cost: The total industry and NRC cost for resolution and implementation of this issue was estimated to be $3.65M.
Based on a total public risk reduction of 16 man-rem, the value/impact score is given by:
(1) Replacement and operability testing of the SRV VB valves will require that the plant operators perform almost all of the work in the drywell in close proximity to the SRV valves. Using the same man-hours estimated for valve replacement and testing for only the operating BWRs and an assumed 35 millirem/hr field in the drywell, a total ORE of about 200 man-rem was estimated.
(2) It was assumed that, in every instance in which an SRV leaks or fails to close and its associated VB valve leaks to the drywell, a plant shutdown will be required and the source of drywell pressurization and temperature increase must be found and corrected. Using the previous estimates of the frequency of VB failure and SRV leakage, an estimated two-day shutdown for each VB failure and a replacement power cost of $300,000/day, it was estimated that, over the lifetime of the 44 BWRs, these events will cost the industry about $40M in lost power production. When discounted to present worth, this was equivalent to about $20M to $25M.
(3) When a SRV VB failure is detected, the valve is repaired or replaced. Resolution of this issue, i.e., placement of all existing SRV VBs with valves with an improved reliability, would therefore reduce the number of VB failures and accordingly reduce the number of VB replacements over the life of the BWRs. Using the estimated current failure rate, the frequency of challenges (transients) and an average number of SRV VBs per plant (9), it was estimated that about 725 SRV VB failures will occur over the remaining lifetime of the BWRs. Improving the reliability of the VB by an order of magnitude would be expected to eliminate about 650 VB failures and, therefore, reduce the number of VB replacements by 650. Assuming 2 man-days of labor to replace a failed VB and a 35 millirem/hr work environment, it was estimated that resolution of this issue would save about 350 man-rem of ORE.
(4) The averted ORE due to reduction in core-melt frequency was estimated to be less than 1 man-rem. Therefore, when all components of ORE are considered, resolution of this issue would result in a net savings of about 150 man-rem of ORE.
Based on the small potential reduction in public risk and core-melt frequency, this issue was DROPPED from further consideration. However, the staff believed that there could be a large economic incentive to the industry to improve the reliability of SRV vacuum breaker valves.