Issue 165: Safety and Safety/Relief Valve Reliability

DESCRIPTION

Historical Background

This issue was identified[1] by NRR when it was found that, on a number of occasions, licensees reported that spring-actuated safety and relief valves failed to meet setpoint criteria within the desired tolerance. Other

reported incidents included more seriously degraded performance of safety and relief valves. These events were

documented in AEOD/S92-02[2] in which the staff concluded that most pressurizer safety valves (PSVs), main steam safety valves (MSSVs), and BWR safety/relief valves (SRVs) did not meet the 1% setpoint drift tolerance and many were above 3%. These results suggested that other systems with safety and relief valves could be

adversely affected by setpoint drift. The staff discussed some of these systems in Information Notices 90-05[3]

and 92-64[4] and in NUREG/CR-6001.[5] More importantly, at Shearon Harris, the failure of a high head safety injection relief valve to operate at a very low setpoint resulted in the undetected loss of the entire system and would have resulted in inadequate emergency core coolant injection if a small- or intermediate-break LOCA had

occurred. This event was discussed in detail in LER 91-008-01 and Information Notice 92-61.[6]

Spring-actuated safety and relief valves provide overpressure protection for a number of systems in both PWRs and BWRs. However, failure of these valves in safety-related support systems could cause a significant diversion of flow from these systems and thus prevent the systems from performing their designed function. It was estimated that perhaps 3 to 5 (out of a total of 55 to 60) spring-actuated safety and relief valves installed in such safety-related systems of a typical PWR or BWR plant could be significant contributors to core-melt frequency. Also, due to the size of these valves (<4 inches), it was believed that most of them could be tested

at the plant site (many of them in situ), thus reducing the time and cost for testing. For these reasons, this issue addressed the unreliability of spring-actuated safety and relief valves in safety-related support systems.

Although Issue B-55 addressed the reliability of Target Rock two-stage pilot-operated SRVs and Issue 70 addressed the reliability of PORVs and block valves, there was no generic issue for spring-actuated SVs and RVs. Because significant NRC and industry resources had been spent in the past on both evaluating the risk and improving the reliability of PSVs, PORVs, MSSVs, and BWR SRVs, the focus of this issue was limited to spring- actuated relief valves in safety-related support systems and the effects of their unreliability on plant operation.

Safety Significance

Failure of a spring-actuated relief valve can lead to a core-melt from loss of core cooling and inventory makeup. Possible sources of loss include: (1) failure of a valve to close after opening; (2) failure of a valve to open when challenged, resulting in overpressure conditions that precipitate a LOCA; and (3) premature opening of a valve below setpoint resulting in a LOCA.

Possible Solution

A possible solution was to improve the periodic inspection and testing of spring-actuated relief valves in risk- significant systems.

PRIORITY DETERMINATION

Assumptions

It was assumed that 71 operating plants with a combined remaining life of 1,907 RY were affected by the issue: 47 PWRs and 24 BWRs with average remaining lives of 27.7 and 25.2 years, respectively. (This corresponded

to the number of plants existing or planned at the time of the initial publication of NUREG/CR-2800.[7]) Implementation of the solution could be achieved at future plants with minimal incremental costs and, thus, a forward-fit evaluation was not performed.

Failure of a relief valve to operate within the allowable opening and closing setpoints was considered a failure of the valve. However, not all valve failures necessarily fail the train of the system in which they operate. Therefore, it was conservatively assumed that 10% of the valve failures would fail their trains. NPRDS was used to obtain values of relief valve unreliability for various systems throughout a plant with spring-actuated relief valves. From

these data, a best estimate probability of the relief valve to fail its train was calculated to be 5 x 10^-3 /demand (based on 524 valve failures out of 10,063 events multiplied by a 10% train failure probability). The upper bound probability was 5 x 10^-2 /demand, assuming the relief valve failure always resulted in train failure. A lower bound

probability was estimated by using the AEOD report[8] which considered 9 valve failures out of 1100 events, equaling a probability of 10^-3/demand including the 10% train failure probability.

Frequency Estimate

The Surry PRA[9] was used to model PWR relief valves in SARA 4.0,[10] the Grand Gulf PRA[11] was primarily used to model BWR relief valves, and the Peach Bottom PRA[12] was used to support the Grand Gulf results.

Because the Surry PRA did not include relief valves in every system, modifications to the PRA were required to model their effects on a particular system. For those systems where relief valves were included with a component in a single train whose unavailability could fail the entire system, the failure probability of the relief

valve was added to the component's failure probability. On the other hand, for those systems where relief valves were included with components in two trains where common mode failure could occur, the failure probability

of the relief valve had to be added by taking into account the use of beta factors in the component's failure probability. A beta factor was defined as the conditional probability of a component failure given that a similar component has failed. P (the component failure probability including the relief valve reliability) and (the beta

factor for the relief valve and component) were given by =( + ) and = [( + )/( + )], where and were the beta factors and and were the failure probabilities for the component and relief valve, respectively. In

this analysis, a value of 7 x 10 was used for which was obtained from the beta factor for an SRV in the PRA. The values of and were obtained from the applicable component in the PRA. Using the above equations, the values of and were calculated and then inserted into SARA for those systems that had dual trains.

The effect of the solution would be to improve the reliability that the valves operate as designed. To reflect this, it was assumed that the solution would reduce the probability for a failure of a safety or relief valve to a negligible amount and thus bring the core-melt frequency to the values predicted by the plant-specific PRAs. As a result, in SARA the base case core-melt frequency value represented the value after implementation of the possible solution and the adjusted case core-melt frequency represented the increased risk from including the effects

of safety and relief valve unreliability. Therefore, the change in core-melt frequency computed in SARA gave the result of improving safety and relief reliability. The changes in core-melt frequency for various systems in the Surry PRA were summarized in Table 3.165-1. Diesel and emergency power includes relief valves in

the emergency diesel generator air start system (see Information Notice No. 90-18[13]). The changes for the Component Cooling Water, Containment Spray, Main Feedwater, and Essential Service Water systems were negligible.

The significant changes in core-melt frequency for various systems in the Grand Gulf PRA were summarized in Table 3.165-2. The changes for other systems studied (which included the RHR/LPI, Feedwater, Condensate, Standby Liquid Control, Control Rod Drive, Nuclear Steam Supply Shutoff, and Low Pressure Core Spray systems) were negligible. The Peach Bottom PRA was used in SARA to further validate the change from the Essential Service Water system computed in the Grand Gulf PRA. These results supported that finding.

Consequence Estimate

The containment failure probabilities and base consequences were taken from NUREG/CR-2800[14] for similar accident sequences. The results from the per-plant calculations for the changes in public risk and core-melt frequency are shown in Table 3.165-3 for the three different estimates of valve failure probability. The total public risk reduction was 10⁵ man-rem with a lower bound estimate of 2 x 10⁴ man-rem and an upper bound estimate of 10⁶ man-rem. These values would increase by about 50% if 75% of the plants had their licenses renewed for a 20-year period.

Cost Estimate

Industry Cost: Assuming that improved periodic inspection and testing of systems with relief valves were required every year and could be performed in about 2 man-days, the total annual test and inspection requirements for each system was estimated to be about 2 man-days/RY. Assuming 5 affected systems per plant, the total labor would be 2 man-weeks/RY. At a cost of $2,270/man-week, the cost for inspection and testing would be (2 man-weeks/RY)($2,270/man-week) or $4,540/RY. For the 71 affected plants, the total cost was ($4,540/RY)(1,907 RY) or $8.7M. Because testing was already required every 10 years, this value was conservatively high.

NRC Cost: Three man-days/RY (0.6 man-week/RY) were estimated for the review of test and inspection requirements associated with the solution. At a cost of $2,270/man-week, the total cost for this review was (0.6 man-week/RY)($2,270/man-week)(1,907 RY) or $2.6M. Other costs, such as work with ASME Code Committees to increase valve testing frequencies, were estimated to be negligible.

Total Cost: The total industry and NRC cost associated with the possible solution was estimated to be $(8.7 + 2.6)M or $11.3M.

Table 3.165-1Change in Core-Melt Frequency for Various PWR Systems

Table 3.165-2 Change in Core-Melt Frequency for Various BWR Systems

Table 3.165-3PWR and BWR Results for Changes in Core-Melt Frequency and Public Risk

Impact/Value Assessment

Based on a potential public risk reduction of 10⁵ man-rem and an estimated cost of $11M for a possible solution, the impact/value ratio was given by:

Other Considerations

The total ORE for implementation of the possible solution was estimated to be 380 man-rem for all affected plants.

CONCLUSION

Based on the impact/value ratio and the potential public risk reduction, this issue was given a high priority

ranking.[15] In accordance with an RES evaluation,[16] the impact of a license renewal period of 20 years was to be considered in the resolution of the issue.

In resolving the issue, the staff performed an analysis of an SRV failing its train and found the resultant CDF increase to be negligible. The staff also determined that additional testing of SRVs was included in the 1986 Edition of ASME Section XI and was later endorsed by the NRC in the 1992 revision of 10 CFR 50.55a. Thus,

the issue was RESOLVED with no additional requirements[17] and licensees were informed of the staff's conclusion in NRC Regulatory Issue Summary 2000-05.[18]