Bengt, like Bill Galyean, used passive-system failure data in his analysis, but in a much different manner. Whereas Bill Galyean followed a “Top Down” approach, Bengt followed a “Bottoms Up” approach. Bengt’s presentation assumed that all BWR welds were category D & E welds. For BWRs, Bengt indicated that Category D & E welds specify inspection criteria based on Generic Letter 88-01. Category D welds have been subjected to weld overlay repairs and Category E welds are subject to IGSCC.

Bengt’s used a different database than Bill Galyean did is his analysis. Bengt’s database is proprietary (PIPEex) and the panel will not have access to this during the elicitation. The SLAP database that Bill used is available to the panel members on the website. The cut-off date for PIPEex events is the end of 2002 while the cut-off date for the SLAP database is the end of 1998. PIPEex has about twice the number of data entries as does SLAP and includes international experience.

Bengt only looked at welds in his analysis. He did not consider base metals. His database did not show any occurrences of base metal indications. However, he invoked a wide definition for what was encompassed by the term “weld”. He included the heat-affected-zone (HAZ) and counter bore region into his definition of what a weld was. Bengt also did not consider degradation of non-piping passive components in his analysis.

For PWR systems, he assumed that the V. C. Summer and Ringhals cracks were circumferentially oriented cracks, not axially oriented. Whereas Bill did a “top down” analysis, Bengt did a “Bottoms up” analysis. The failure rates were derived for individual welds, and then an integration system level model was formed by combining the contributions from each individual weld failure to an overall pipe system failure frequency.

The term “prior” has very specific meaning in this analysis. It means “before mitigation/remedial action in response to a significant pipe failure”. Hence, failure rates that are input to LOCA frequency calculations explicitly account for reliability improvements (mitigation methods) made in response to past pipe degradation histories. Failure in Bengt’s analysis is defined as a “through-wall flaw resulting in leakage”. Bengt did not include surface cracks found during ISI in his data reduction process.

Karen Gott has a report in Swedish that may be valuable in this effort. This report gives the number of leaks and the number of ISI detected surface cracks. Gery Wilkowski thought that was important information since leaking through-wall cracks are readily detected, but the surface cracks that would grow to be long in length are more of a LOCA threat. The number of records shown on slide 10 from Bengt’s presentation includes both leaks and cracks.

There was much discussion among the group in an effort to understand slides 11 through 14. In slide 13, Bengt did not eliminate welds if mitigation was performed prior to 15 years of operation in developing his prior distribution. There were no leaks in BWRs after 15 years. From years 10 to 15, it is possible that there may have been some plants that used mitigation, but those mitigated plant weld numbers were still included in the weld failure rate analysis, i. e., that may account for why the weld failure rate was not accelerating as the number of years increase. Bengt used a Monte Carlo simulation to create this plot, where he needed to estimate the number of welds for the number of plants that were at a certain age. Results in slide 14 include results from US, Spanish, Swedish, and Japanese plants. The results were adjusted by the number of welds and type of welds.

These preliminary results are used to determine the “prior” LOCA frequencies (weld failure rate). The next step is to determine the “posterior” frequencies based on “prior” distributions. Bengt accounts for uncertainty in the knowledge base, which is fundamental difference between his analysis and Bill’s.

Dave Harris thought the “posterior” frequencies should be equal to the Prior LOCA frequencies times the Likelihood Function. Lee Abramson explained that the likelihood function was built in.

It is noted in slide 19 that the Bayesian update strategies are different for each base case. It was also noted again that the weld failures represent leaks only and not cracks. Weld failure was defined as a through-wall flaw with leakage less than or equal to the tech spec limit for undefined leakage.

Gery Wilkowski noted that in January 2003, PWSCC was found in a surge line bimetallic weld at the pressurizer in a Belgium plant. It was noted that some transient event is needed to cause a tech spec limit flaw to propagate to a higher category (well beyond 100 gpm [380 lpm]) LOCA. Slide 23 shows that the conditional probability of failure (PL/F) is a function of the nominal pipe diameter (DN), i. e., PL/F = a x DNb, much like what is in NUREG/CR-5750 (Beliczey and Schulz, i. e., PR/TWC = 2.5/DN). While the exact formulations are different, the conditional failure probability in all cases is an inverse function of pipe size.

The form of the conditional leak probability given a failure is inconsistent with the original development of the Beliczey and Schulz correlation which was developed to relate leaks to breaks as a function of pipe size. This use (see slide 24) may not be physically realistic because it assumes that larger diameter pipes are less likely to result in a category 0 leak. The implication is that larger diameter pipes are less likely to reach a Category 0 leak, and then progress to higher leak rates. Slide 26 presents information on the aspect ratios of IGSCC cracks. It was asked how the deep, full circumference cracks (a/t = 0.5, 2/B = 1.0) formed. The expectation is that these are likely crevice cracks. It was suggested that it would be nice to break down this data by pipe size in order to assess the relevance.

Bengt assumes that a through-wall crack can only propagate into a large leak if there is a large transient event. Slide 27 documents the loading categories assumed to drive the crack among various LOCA categories. Category 0 to Category 1 LOCA progression can occur assuming moderate loading, while to go from a Category 0 or Category 1 LOCA to a Category 6 LOCA, would require an extreme loading transient. The general consensus of the panel members was that this was a very subjective approach. It was noted that there were about 400 water hammer events reported, but Bruce Bishop and Guy DeBoo said that if there were this many water hammers, then the plant piping system was probably redesigned.

The extrapolation of results from the “Current Estimate” to 40 and 60 years is based on posterior analysis of prior results assuming no additional failures. If one assumes no additional failures, the failure rates will go down with time. Bengt will examine possibly extrapolating his base case results out to 60 years using another assumption.