By Bill Galyean, INEEL

Bill employed a “top down” approach in his analysis of operating experience. His database represents approximately 2,600 LWR-years of operating experience. The resultant average age of a plant is 23 years. In that 2,600 years of LWR operating experience there have been no passive system LOCAs with a resultant leak rate greater than 100 gpm (380 lpm) (Category 1 LOCAs).

As part of this analysis, Bill assumed that cracks and leak events are indicators of LOCA frequencies. They indicate system susceptibility. In order to get a LOCA, Bill’s analysis assumed that a piping system must first have a leak or a crack. In the first 2,647 years of US LWR experience represented in Bill’s database there have been approximately 1,100 crack and leak events, but no LOCAs. Note, at the first elicitation meeting the demarcation between leaks and breaks was set at a 100 gpm (380 lpm).

A comment was made that small pipes are more susceptible to LOCAs than large pipes, i. e., large pipes are less likely to fail catastrophically. A question was raised as to why limit the analysis to US operating experience only. Bill did not categorically know whether there have been any 100 gpm (380 lpm)

LOCAs worldwide. Another reason to limit his analysis to US operating experience is that there are some fundamental design differences between US and other overseas plants. Pete Riccardella (Structural Integrity Associates) thought that if Bill had included foreign experience that the number of years of operating experience would have about doubled so Bill’s LWR LOCA frequency number of 1.9E-04/year be would reduced by a factor of two to approximately 1.0E-04/year. It is important to understand the basis for this 1.9E-04/year number since everything else is referenced to this number. This number is the total number of LOCAs of all sizes.

It was noted that this is a different approach than followed in NUREG/CR-5750. This analysis was not an attempt to update NUREG/CR-5750. Pete Riccardella asked if this database included all of the small diameter socket weld cracks that occur due to vibration fatigue. Bill indicated that this was the case, even though these small diameter lines could not result in a 100 gpm (380 lpm) leak. It was pointed out that there was no distinction between cracks and leaks in Bill’s analysis. Any crack deeper than 10 percent of the wall thickness was included in the analysis.

Gery Wilkowski (Emc2) asked if the analysis of the feedwater system (BWR-2) included FAC as a failure mechanism. Gery noted that on the secondary side there have been large breaks in some piping systems due to FAC. Bill indicated that he limited his database search to those systems that affected reactor coolant pressure boundary integrity.

Karen Gott was surprised at the low number of incidences for the feedwater system. Most of the problems seen to date with the feedwater systems have been outside the primary portion of system. Furthermore, cracks in nozzles are associated with the RPV and not piping. For the PWR systems there has been a lot of feedwater cracking, but those cracks have been on the secondary side. It was also pointed out to the panel that everyone has access to the SLAP database that Bill used for his analysis. It is now on the ftp site. Rob encouraged everyone to use it as part of their elicitation exercises. The SLAP database is current up to the end of 1998.

A question was raised about the validity of the single IGSCC failure reported in the carbon steel feedwater system. Bill indicated that this was the reported database value. Karen Gott said they’ve seen such cracking in Sweden as well.

The BWR recirculation system provided a unique problem for the base case analysis. From a materials standpoint, the base case was the old system. Bill segregated the data by old pipe (Type 304 stainless) versus new pipe (Type 316 nuclear grade [NG]). For the old pipe (Type 304 stainless) there were 127 events in 550 years of operating experience versus 3 events in 410 years of operating experience for the new pipe (Type 316NG).

The resultant leak/crack frequency for the old pipe (127 events/550 years = 0.231 events per year) is about a factor of 2 greater than the overall leak/crack frequency for the overall recirculation system history (old plus new), i. e., 130 events/960 years = 0.135 events per year. It was pointed out though that this improvement may be more due to other factors than pipe replacement only. The improvement could also be due to changes in water chemistry, or the installation of weld overlay repairs. Hence, it may be more appropriate to refer to the pipe systems as mitigated (new) or unmitigated (old) pipe systems.

It was stated that the base case is unrealistic in that it is for the old pipe case (Type 304 stainless) and no one uses that material anymore. Also, the base case does not account for the incorporation of water chemistry improvements which all plants have already implemented.

As part of his analysis, Bill made an assumption that the LOCA categories/sizes (e. g., 100, 1500, 5000 gpm, etc. [380, 5,700, 19,000, lpm, etc.]) are related on a logarithmic sense (1, 0.3, 0.1, 0.03, 0.01, 0.003). Half likelihood on logarithmic sense realizing that larger LOCAs are a subset of Category 1 (100 gpm [380 lpm]) LOCAs.

The 40 welds for the PWR-1 case (hot leg) include the cold leg and cross over leg welds. This is inconsistent with the assumption stated above that the PWR-1 case only considers the hot leg (not the cold leg or cross over leg). (Note, there are typically 3 loops in a PWR plant and there can be 5 to 7 welds per loop, but the loading is not the same for all these welds.)

Rob Tregoning indicated that the correlation between pipe size and LOCA size (gpm) that were originally supplied are subject to change.

Bill’s aging correction factor is for thermal fatigue and should not be used for the other failure mechanisms.

The non-pipe LOCAs that Bill included are for passive systems only (bolted flanges, etc.). He did not include active system contributions, e. g., stuck open valves, to the non-pipe break frequencies.

Bill’s results for the “Current Estimate” are significantly smaller than the LOCA frequencies reported by others (WASH-1400, NUREG-1150, NUREG/CR-5750) due to the larger database (more years of operating experience without a LOCA). The best agreement is with the NUREG/CR-5750 results. Bill employed a Bayesian approach as part of his LOCA frequency analysis by assuming a half of failure for these very low occurrence events. This is a very common data analysis practice. Since Bill’s analysis is based on the analysis of past passive-system failure data, the data implicitly include ISI and other mitigation experience implemented by industry.

Each plant has a PRA which includes LOCA frequency estimates for the plant. These LOCA frequency estimates are often based on WASH-1400 or NUREG-1150 and the ranges shown on pages 36 and 37 of Bill’s presentation are the ranges for the IPEs. Thus, the IPE range, WASH-1400 and NUREG-1150 frequencies shown by Bill on pages 36 and 37 of his handout are closely related/interlinked. As described in Slides 36 and 37, Bill expressed the uncertainty in his analysis by assuming an error factor of 10. This is somewhat crude and somewhat arbitrary, but the scope of the base case analysis did not ask for uncertainty, just a best estimate.

Bill indicated that he has no strong technical basis for extending his “Current Estimate” analysis out to 40 or 60 years. With Bill’s approach, the LOCA frequencies will go down as more years of operating experience are accumulated, unless a LOCA event occurs in the future. Bill presented information predicting 8 years to double the frequencies for thermal-fatigue events. This estimate conservatively assumes no industry-wide mitigation programs. Rob Tregoning indicated that he did not include results for Bill in the summary table for 40 and 60 years of operation because they are most appropriate for current estimates.

Bruce Bishop was very uncomfortable predicting the future out to 40 or 60 years. He felt that the uncertainties are going to increase dramatically. Lee Abramson (USNRC) responded that this is a natural experience and likely shared by others on the panel. However, it’s incumbent that each member to attempt these predictions. Rob and Lee stressed again that the panel will not be forced to answer any questions that they are very uncomfortable with. Rob further stressed that the March 2003 SRM specified that the NRC staff needs to revisit the LOCA frequency estimates every 10 years and the effort will be most concerned with the next 10 years. However, it is still important to gain longer-term insights.

A question was raised as to whether or not to have Bill extend his analysis out to 40 and 60 years? Gery Wilkowski said, yes, he wanted to see Bill’s assumptions. Fred Simonen (PNNL) would like to see more partitioning by pipe size as part of Bill’s analysis. Dave Harris (Engineering Mechanics Technology) concurred. Sam Ranganath (formerly of General Electric) specifically indicated that it would be helpful for the BWR recirculation system because they replaced the smaller diameter pipes, but not the 28-inch diameter pipes. Bengt Lydell is to check on the statistics in his database to see if any of the 28-inch diameter BWR recirculation lines had leaks and provide this information in his final base case report.

Pete Riccardella and Sam Ranganath were unaware of any General Electric large diameter recirculation pipes that leaked.