I.1 Introduction

The LOCA expert panel elicitation team charter includes estimating the contribution to LOCA frequency from reactor vessels and other non-piping components. Extensive analyses were performed by members of the elicitation panel to develop LOCA frequencies for five piping “base cases” that were formulated by the panel in early meetings (documented as Appendices D, E, F and G to this NUREG). The piping base cases include failures on the piping side of vessel nozzles, including safe-ends. However, they do not include small diameter, partial penetration welded nozzles such as CRDM penetrations and other small nozzles, such as instrument nozzles, that aren’t connected to piping systems. In addition, the piping base cases do not include consideration of a leak from or rupture of other regions of the reactor vessel, such as the irradiated reactor vessel beltline or the low alloy steel portions of large vessel nozzles. LOCA frequency estimates for these cases are presented in this appendix, based on prior PFM analyses performed for PWR top head nozzles [I.1, I.2], the BWR Reactor Vessel Beltline Region [I.3, I.4], and BWR reactor vessel feedwater nozzles [I.5]. These estimates are used to construct a complete set of LOCA frequency tables for BWR and PWR reactor vessels, for all LOCA categories defined in the elicitation, a comparison of them to the aforementioned piping base cases is also presented.

I.2 PWR Reactor Vessel Top Head Nozzles

Extensive PFM analyses have been conducted over the past several years to estimate the probability of leakage and rupture associated with the PWR CRDM penetration PWSCC problem [I.1, I.2]. The analysis model incorporates the following major elements:

• computation of applied stress intensity factors for circumferential cracks in various nozzle geometries as a function of crack length,

• determination of critical circumferential flaw sizes for nozzle failure,

• an empirical (Weibull) analysis of the probability of nozzle cracking or leakage as a function of operating time and temperature of the RPV head,

• statistical analysis of PWSCC crack growth rates in the PWR primary water environment as a function of applied stress intensity factor and service temperature, and

• modeling of the effects of inspections, including inspection type, frequency and effectiveness. The model has been benchmarked with respect to field experience, considering the occurrence of cracking and leakage and of circumferential cracks of various sizes. Figures I.1 and I.2 illustrate the benchmarking. Figure I.1 presents a Weibull analysis of inspection results at thirty plants, of which 14 detected leakage or cracking (data points in the figure). The remaining plants that were inspected and found clean were treated as “suspended tests” according to standard Weibull analysis theory [I.2]. The data are plotted in terms of effective degradation years (EDYs) which are equivalent operating years at 600°F (315°C), using an activation energy (Arrhenius) model [I.1] to adjust for different head operating temperatures. For plants in which multiple cracked nozzles were detected in the inspections, the data were extrapolated back to the expected time of first cracking or leakage, using an assumed Weibull slope of 3. The straight line through the data represents a medium rank Weibull regression (also with a slope of 3) upon which the probability of leakage predictions in the model are based. Figure I.2 illustrates the benchmarking process used for the crack growth analysis algorithm in the model with respect to CRDM nozzles that exhibited circumferential cracks of various sizes. (Eleven (11) nozzles out of a total of 881 inspected nondestructively through the spring of 2003 exhibited circumferential cracking. No additional

circumferential cracking has been detected in more recent inspections.) The figure shows that, when using original analysis parameters, the crack growth model under-predicted the probability of circumferential cracking somewhat, but after adjustment of selected analytical parameters, the PFM model was “benchmarked” so as to very accurately predict the field results, especially for the most important, larger crack sizes.

The benchmarked model was then used to evaluate the probabilities of nozzle failure and leakage in actual plants. A sample of the results is presented in Figures I.3 and I.4. Figure I.3 illustrates the probability of nozzle failure (ejection of a nozzle) for a head operating temperature of 580°F (304°C), the approximate average of U. S. PWRs. No inspections were assumed to be performed during the first 25 years of plant operation, resulting in the probability of nozzle failure constantly increasing with time during that period. The analysis then assumed that inspections begin after 25 years, at intervals and detection levels representative of current requirements [1.6]. It is seen from the figure that the current inspection regimen reduces the nozzle failure probability significantly.

Ejection of a 4 inch CRDM nozzle [2.75 inch (~70 mm) ID] due to a circumferential crack would yield a one-sided LOCA corresponding approximately to Category 2 LOCA [>1,500 gpm (5,700 lpm) but <

5,0 gpm (19,000 lpm)]. If periodic inspections are continued, with any nozzles in which leakage or cracking are detected repaired or the heads replaced (as is common practice), the nozzle ejection probability will be even lower in the future. Table I.1 below provides a summary of the average failure probabilities from Figure I.3, between 0 and 25 years, and from 25 to 40 years. The probability of failure for 40 to 60 years was not calculated, but was assumed to be the same as 25 to 40 years, on the basis that the current inspection regimen will be maintained, or the heads replaced. A Category 3 break was assumed to require multiple nozzle failures, the probability of which was computed via a binomial distribution for the typical number of nozzles in a head. As seen in Table I.1, the probabilities of simultaneous multiple nozzle failures is quite low.

Figure I.4 illustrates similar PFM results (based on the above Weibull model) for the probability of small amounts of leakage from a top head CRDM nozzle. The same inspection regimen was assumed as in the nozzle ejection analysis (no inspections from 0 to 25 years, inspections in accordance with current requirements thereafter). A small leak from a CRDM nozzle was assigned as a Category 0 break [less than 1 gpm (3.8 lpm)] in Table I.1, and the intermediate, Category 1 break size was obtained by logarithmic interpolation between Categories 0 and 2.