RPV nozzles constitute another potential non-piping LOCA concern. The example used to address this concern here is BWR feedwater nozzles, which have in the past been subject to thermal fatigue cracking [1.5]. The thermal fatigue problem was caused by mixing of hot reactor water and relatively cold feedwater (see Figure I.5) during reactor startups, shutdowns and other periods of low power operation, when feedwater heating is generally unavailable. Cracking of various depths, up to 1.5 inches (38 mm), was detected in a number of BWRs in the 1970s (see Figure I.6). At that time, the standard feedwater nozzle design incorporated a loose-fitting thermal sleeve/sparger configuration, as shown in Figure I.5. Since then, all U. S. BWRs have installed some type of fix, employing either welded-in spargers or multiple-sleeve designs with shrink fits and piston rings to protect the nozzle from the effects of the cold feedwater. No subsequent cracking has been discovered since the improved thermal sleeves were installed.

In order to perform a base case analysis of this problem, a modification to the software (VIPER-NOZ) was developed to estimate leakage and failure probabilities for BWR Reactor Vessel feedwater nozzles. The substantive changes to the VIPER software in VIPER-NOZ were the addition of thermal fatigue crack initiation and growth algorithms specific to the feedwater nozzle thermal cycling phenomenon, and zeroing out the effects of irradiation embrittlement, since feedwater nozzles are far enough from the reactor core region that neutron fluence effects are small. The VIPER-NOZ software was run for conditions representative of the original nozzle/sparger designs, to confirm that cracking probabilities consistent with early field experience (Figure I.3) are predicted. The boundary conditions were then modified to represent improved nozzle/sparger designs, which reduce the effects of thermal fatigue on the nozzle. The analyses were conducted for a 60 year operating lifetime, and included the effects of periodic ISI, which are performed for these nozzles on ten-year intervals. The results are given in the following table:

The predicted leakage cases were treated as Category 0 breaks in this case, and since the nozzle is attached to a 12 inch diameter pipe, the maximum credible break size was assumed to correspond to single ended rupture of a 12 inch pipe, which corresponds to a Category 4 break. A total of 1 million simulations were run, and except for LBB type failures at 40 and 60 years, no other failures were predicted. Thus a failure frequency of less than 1E-6 is given for most entries in the above table.