Dave prefaced his comments with the thought that he thinks that the base case results are surprisingly close considering the differences in the approaches. Dave indicated that an important input to any PFM analysis is the stress history. This requirement is contrary to the operating-experience approaches where the stress history is indirectly reflected in the incidence of cracking events. Dave’s analysis is performed on individual pipe locations which are then integrated to determine the overall system frequency.

Some key points from the crack initiation and crack growth portion of Dave’s presentation are that piping failures occur due to the initiation and growth of cracks. Cracks initiate due to stress corrosion or fatigue. Growth is controlled by fracture mechanics (other than early SCC). The question was asked as to what assumption Dave used as to the size of the crack once it initiates. Dave’s PRAISE code assumes that fatigue cracks are 0.3 inch (7.6 mm) deep (per a criteria proposed by Argonne National Laboratories [ANL]). This is based on an assumed 25 percent load drop definition of crack initiation from an S-N specimen test. In addition, in PRAISE the SCC rules differentiate between early SCC growth and fracture mechanics growth since the early growth is faster than calculated by fracture mechanics analysis. The SCC rules in PRAISE assume a 0.001 inch (0.025 mm) deep surface crack with some distribution function on length. The latest version of PRAISE (2002) includes updates to the S-N curves that incorporate environmental effects. Pete Riccardella noted that Art Deardorff of SAI was doing an update of some of the environmental S-N results from EPRI. For crack growth, the focus of PRAISE is semi­elliptical part-through surface cracks. PRAISE considers crack growth in both the depth and length directions (K at both the maximum depth and at the ends of crack.)

Dave pointed out that fatigue failure of welds is dominated by growth from pre-existing crack-like fabrication defects. Thus, the flaw distribution of initial fabrication defects is an important parameter to define. PRAISE stipulates that the final failure is controlled by tearing instability. PRAISE treats the stresses at maximum load as load-controlled stresses from stability analysis perspective.

The leak rate in PRAISE is computed based on the length of the leaking through-wall crack on the inside pipe surface using the SQUIRT leak-rate code. The SQUIRT code has recently been updated with numerous technical enhancements as part of the USNRC Large Break LOCA program. PRAISE includes some of the mechanistic dependent crack morphology parameters from some of the earlier versions of SQUIRT, but not the new COD-dependent roughness, number of turns, and flow path length to thickness ratio parameters. In addition SQUIRT has been made more user friendly by incorporating a graphical user interface. Note, WinPRAISE is PC-PRAISE with a Windows pre-processor for entering input parameters.

Dave used a stratified sampling technique that allows for the evaluation of extremely small probabilities events such as the case of fatigue crack growth from pre-existing defects (10E-17 frequencies). The approach also assumes that all cracks with leak rates greater than 5 gpm (19 lpm) are discovered and subsequently removed from service which implies that getting a higher Category LOCA (e. g., a 1,500 gpm [5,700 lpm] LOCA) would most likely result from the growth of a long surface crack that pops through the wall thickness and immediately becomes a TWC with a length equal to length of the surface crack on inside pipe surface. (The only other means of achieving a higher Category LOCA would be through some sort of transient event.)

Dave postulates that inaccuracies in leak-rate calculations will not significantly impact final LOCA frequencies. Even assuming the leakage detection capability equals zero should not have a large effect. Sam Ranganath felt that a surface crack grows 3 or 4 times faster in the length direction than it does in depth. He wasn’t sure if that was due to multiple initiation sites, or the surface growth rate being higher. The analysis in PRAISE considers crack growth in both the depth and length directions (K at both depth and ends of the crack) with an RMS value of K in each direction.

The detection probabilities shown in slide 14 are based on depth only, not length. It is likely that current technology has better performance. It was also noted the fatigue crack growth parameter used could be improved based on newer results. Dave Harris indicated that sensitivity studies show that using improved crack growth parameters in PRAISE (e. g., including environmental effects) will result in changes in LOCA frequencies on the order of a factor of 2. Slide 16 presents an example S-N initiation curve for a low alloy steel. Vic Chapman raised the concern that there may not be a plateau or fatigue limit with the higher number of cycles. Gery Wilkowski commented that the default flow stress values (slides 17 and 18) are very tight from a standard deviation perspective, especially in light of what was seen in NUREG/CR-6004 from an analysis of PIFRAC data.

The NUREG-6674 stresses have been downgraded from the design basis stresses by Jack Ware of INEEL to make them more realistic with fewer transients. The stresses may have been elastically calculated values, which can go well above yield and still be allowed for secondary stresses by the Code. PRAISE uses the most realistic values available. The final results (frequencies) from Dave’s analyses are very dependent on stress input values. Dave commented that one only needs stresses at the high stress locations. These high stress locations dominate the final frequency answers.

Slide 21 shows the surge line stresses. These stresses are probably for the flank of the elbow, not the weld. These values can’t be used for the girth weld location (i. e., they are too high). The stresses shown are stress amplitudes (the stress ranges will be twice these values). In addition to stresses and number of occurrences, one also needs some input as to the spatial distribution for these stresses. One of the short comings of PRAISE is that PRAISE doesn’t have a model for FAC for the feedwater lines. The fatigue initiation models are only in latest versions of PC-PRAISE, i. e., from NUREG/CR-6674 published in June 2000.

An ad hoc procedure was used with pc-PRAISE in order to obtain results for larger leak rates (stratified sampling for fatigue crack growth is not available for fatigue crack initiation). One of the handouts provided shows this ad hoc procedure.

A question arose with slide 26 concerns the fact that Dave’s analysis shows that the cumulative failure probabilities continue to increase after 20 years after a weld overlay repair is applied at 20 years. Past experience at Battelle as part of the Degraded Piping Program showed that weld overlays are very effective. They have much higher strength than the base metal. Another aspect of their application is that they apply a very high compressive stress at the crack plane. These high compressive stresses should restrict any further crack growth of the surface crack. In addition, it was noted that these high compressive stresses may preclude the environment from getting to the crack tip. Dave noted that he put in a linear approximation of the stresses through the thickness, the increased thickness of the overlay, and the crack growth equations for Type 316 NG into his analysis. WIN-PRAISE uses an adjusted weld residual stress pattern (linear gradient) for the case of post-weld overlay residual stresses (see slide 38). Dave will also present the failure probabilities without the weld overlay so that an assessment of its effect can made.

Bruce Bishop asked how much was Dave’s results affected by inspection. Dave didn’t think that the final frequencies would be affected that much. Dave’s results showed that there was a minimal change in LOCA frequencies for the hot leg as a result of the application of a 5SSE earthquake. This was not surprising to Dave since he has found similar behavior in a previous study. Lee Abramson pointed out that one means of seeing the effect of the earthquake is to compare conditions for equal probabilities. For example, for the 40 year time period analysis, no earthquake results in 1.3E-18 LOCA frequency for the no leak case while for the same 40 year time period analysis, a 5SSE earthquake results in 1.3E-18 frequency for a DEGB.

Dave, generally found that the LOCA frequencies were not highly dependent on JIc and dJ/da. Gery Wilkowski thought that if the toughness was low enough that one was operating in the EPFM regime then the LOCA frequencies may be more dependent on toughness. Gery indicated that he thought that the toughness values used in Dave’s base cases were too high for weld crack locations, or aged cast stainless steel pipe and fittings.

PRAISE can’t account for time dependent material properties. Thus, to account for aging, one would need to input aged properties in at time equal to zero.

The very low LOCA frequencies for the hot leg in slide 31 may be an artifact of the failure mechanism (fatigue) chosen for analysis. Higher frequencies may be seen for some other mechanism, such as PWSCC. This is a case that we may want to analyze in future analyses.