The first part of the presentation was presented by Chris Bell (Rolls Royce) with the final few slides presented by Vic Chapman. Chris presented the general assumptions and methodology of the PRODIGAL code. Note that PRODIGAL actually has several modules. One of them is to determine weld defect size from welding information. Another is to determine failure probabilities for navy nuclear power plants. All results presented are for a per weld basis.

In PRODIGAL, surface imperfections with a depth of 0.004 inch (0.1 mm) are assumed, but there are no SCC initiation/growth models in PRODIGAL. Past study has shown that there is little sensitivity to the assumed depth and sizes much less than 0.004 inch (0.1 mm) have little effect. For surge line case, there is sensitivity to the existence of 0.04 inch (1 mm) defects, see slide 19.

Based on some work of Ritchie, Pete Riccardella felt that the 0.004 inch (0.1 mm) initial defect size was near the LB of the region where fatigue crack growth (da/dN) models were valid. It was noted that Omesh Chopra from Argonne thinks that this lower limit is closer to 0.02 inch (0.5 mm). It was suggested that this limit is dependent on the grain size of the material. Note, cast stainless steels can have very large grain sizes.

Vic and Chris felt that they could not do a generic analysis for seismic considerations since the effect of seismic has been found to be highly dependent on plant layout. Therefore they didn’t consider seismic stresses in their analysis. In addition, they only considered the three PWR base cases since they had little experience with BWRs.

They used the same stress data as Dave Harris did for consistency purposes. As Dave did, they used a second order distribution of stresses thru the thickness per NUREG/CR-5505 criteria that was done by PNNL in 1998.

The failure criterion for their instability analysis is based on the FAD approach in R6 using KIC, i. e., crack initiation would equal failure. The crack initiation used for stainless steel was closer to wrought base metal rather than weld metal which would be an order of magnitude lower, or aged cast stainless steel which could be another factor of 3 lower than the weld metal toughness.

Bruce Bishop asked what the mean temperature in the analysis is used for. It is used for material property considerations, such as flow stress, but not for subcritical crack growth. Slide 16 from Vic and Chris’ presentation shows the cumulative probability of a TWC (probability of a leak occurring). The implication from this slide is that if you are going to have a leak, it will occur in the first 25 years of operations. It was noted that slide 16 is conditional on having a crack (probability of having a crack is 1).

Rob Tregoning indicated that he wants each of the panel members in the next week to make list of what they want to see (e. g. data they used in doing their analysis) from either individual participants or from the group as a whole.

It was noted that the dominant hot leg cycles are those due to heat up and cool down. There are only on the order of 5 of these cycles per year.

Pete Riccardella thought that there were a lot of cycles on the PWR HPI/Make up nozzle each year. Dave Harris countered that he thought that there were only about 40 of these cycles in 40 years of operations. Bengt Lydell agreed with Pete and thought there were a lot of thermal cycles. Pete thought that something was missing here. Bengt said there was a nice ASME paper on the cyclic stress history of these nozzles that we could use in the analysis.

Chris indicated that although they typically keep the aspect ratio constant in their PRODIGAL runs, they have the ability to grow cracks in length, and often get very irregular crack shapes.

In slide 24, the “separation” referred to is the crack-opening displacement. In this slide, at the surge line elbow, Vic speculates that the crack starts to act as a hinge so that crack opening becomes very large for the longer crack lengths. This assertion is the basis for slide 25 illustration of the crack frequency versus

angle distribution that was assumed in the analysis. This relationship was created by Vic with guidance from metallurgist to rectify predicted and experimental crack lengths.

It was commented that for the same input, we are seeing similar results from PRAISE (Dave Harris’ results) and PRODIGAL (Vic Chapman and Chris Bell’s results). It was noted that as far as this exercise is concerned, PRODIGAL’s strength is the defect distribution analysis. PRODIGAL only looks at thermal fatigue while PRAISE can look at other failure mechanisms. PRAISE can also look at crack initiation while PRODIGAL cannot. However, neither accounts for FAC or PWSCC at this time.

PRAISE can also look at crack initiation whereas PRODIGAL assumes crack growth from weld defects or surface imperfections.

Vic pointed out that embedded cracks can straddle the compressive zone of the residual stress field through the thickness so that once they break thru to inside surface they are already through the compressive zone. This is in contrast to case where a crack is growing through the wall thickness from the inside pipe surface and the crack gets trapped in the compressive zone near mid thickness. There was also the question of whether embedded cracks are affected by the environment.

The next topic for discussion was to plan the next step for the base case calculations.

Bruce Bishop would like to bench mark the PFM results against the 25 operating-experience estimates developed by Bill Galyean and Bengt Lydell, and then use the PFM models (PRAISE and PRODIGAL) to predict the 40 and 60 year results. Rob Tregoning thought that this was an excellent idea. Rob also indicated that we could do this for some cases, but not all cases, e. g., FAC in the feedwater system or PWSCC.

Sam Ranganath would like to know when we compare results where do we get good agreement and where not. Rob indicated that we haven’t made comparisons on a consistent basis as of this date but that this would be rectified.

The next issue focused on additional stresses to consider in the PFM results. The surge line and HPI/MU cases were mentioned. Dave Harris has already done some additional analysis for the surge line, based on refined stresses developed by Art Deardorff. Gery Wilkowski noted that he had surge line stresses from a Westinghouse Owner’s Group report used for LBB analyses. Pete agreed to verify that the previous stresses provided by Art are appropriate. Pete also agreed to provide more accurate HPI/MU stresses.

Sam Ranganath would like to lower the stresses to 10 ksi (70 MPa) for the recirculation line (BWR-1) and to lower the feedwater line stresses (BWR-2) by 20 percent.

The next area where the panel thought we may want to focus is some sensitivity analysis using different material properties. Gery Wilkowski volunteered to supply some distributions of material properties (mainly toughness values for welds and aged cast stainless steels) developed as part of NUREG/CR-6004.

Gery Wilkowski also wanted to see the ratio of surface crack to through-wall cracks removed from service, and the distribution of the lengths and depths of those service removed surface cracks. He then wanted to see a comparison of the PFM probability of leaks for IGSCCs from PRAISE and the distribution of surface cracks that might exist up to the time that piping might have been replaced/repaired (15 service years?). Karen suggested looking at the more recent results on a yearly basis because ISI wasn’t sensitive enough to find surface defects in early years. Cracks were only discovered once they became a leaking crack. Also, in Sweden, even if whole pipe sections were removed, all the welds were inspected, whereas in the US if the pipe system was replaced they did not spend the effort to inspect welds that were being removed from service. Hence, the database of cracks removed from service should separate Swedish and US plants. The important aspect of this comparison is to see the population of the ISI remove surface crack lengths compared to the surface flaw sizes calculated by PFM. If the service removed crack lengths from ISI are much longer than calculated by the PFM analyses, then the PFM analysis should underestimate the future failure probabilities. Gery noted several times that failure probabilities should be controlled by development of long surface cracks, not the growth of leaking through-wall cracks. Bengt Lydell has some papers relating ISI-detected surface-crack geometries to through-wall crack leaks that he can provide on the ftp site and can provide to Rob Tregoning.

It was suggested to include PWSCC in the hot leg base case analysis. Dave Harris has initially done this using IGSCC relationships for preexisting flaws. Initiation is not accounted for. Gery Wilkowski noted that the crack growth rate through the weld metal (along the dendritic grains) is much faster than IGSCC. Gery will work with Karen Gott and Bill Cullen to see how Dave can adjust the PRAISE model to get initiation and growth for PWSCC. Gery will provide the IGSCC initiation and growth equations that PRAISE uses to Bill Cullen so that appropriate constants can be provided for PWSCC in PRAISE for base case calculations. Gery and Karen will provide information to Dave Harris on PWSCC crack initiation and growth models that he can use to evaluate the impact of PWSCC on the appropriate base case calculations.

The panel thought it was important to address Dave Harris’ strange results for weld overlay repairs. What is leading to high growth rates after the overlay is applied. Can a comparison be shown with what would happen if no weld overlay had been applied? Dave Harris is to address this concern of the unexpectedly high growth rates after the weld overlay repair is applied.

When the summary comparison tables are completed, the PFM subgroup needs to clearly identify where the PFM conditions do not agree with operating experience. Rob Tregoning indicated that the base case subgroup needs to finish up the base case calculations by the end of the month. We cannot delay individual elicitations any longer.