Comment Number: 7-9

Submitted by Zouhair Elawar — Palo Verde Nuclear Generating Station

Comment: Provide a section on probabilistic validation of the small LOCA frequency. Using a Poisson distribution with failure rate of 2.9E-03 (NUREG-1829 category 1 LOCA frequency excluding steam generator tube rupture events). And, approximately 1500 reactor-years, the probability of ZERO small LOCA events (actual industry performance) is ONLY 0.013 (i. e.1.3% chance). This result shows an excessive conservatism in the category 1 LOCA frequency. Are we to expect an average of ONE small LOCA event (with 74% chance) per 345 reactor-years (equivalent to ~4 calendar years)? I think not.

Response: This comment is similar to both GC7 and 7-8. See the responses to these comments for comparisons between the NUREG-1829 Category 1 LOCA estimates and operating experience. Section 7.10 has also been added to provide an in-depth comparison with operating experience.

Comment Number: 7-10

Submitted by Westinghouse Owners Group

Comment: It would be useful to the PRA community, and help facilitate plant to plant consistency if, for the smaller break sizes, more information or guidance were provided to help separate out the frequencies of SGTR from small break LOCA, and CRDM nozzle breaks from medium LOCAs, as well as any other contributors other than primary system piping. Although there is no current intention to use the results of the expert elicitation to update the various LOCA frequencies assumed in individual plant PRAs, as plants go forward with peer reviews of PRAs, it is likely that LOCA frequencies for small, medium and large LOCAs will be compared with NUREG-1829 results. The NUREG-1829 frequencies for the smaller LOCA sizes are several times higher than the values presented in NUREG-5750. The NUREG-1829 values listed for the small break frequency (> 100 gpm) are so high that one would expect to have seen an event in the US every three or four years, whereas to date we have seen none. It would be helpful if the conservatism in the estimates for the smaller beak sizes was discussed and some caveat provided so that plant PRAs don’t end up with excessive conservatism in their small LOCA risk estimates.

Response: This comment is similar to Comment 7-2. See this response to this comment for a discussion about separating failure of individual components like steam generator tubes and CRDMs from the total

LOCA frequency estimates. Section 7.8 in the revised NUEREG provides separate SGTR estimates.

Also, the responses to Comments GC3, GC4, GC5, GC7, 7-1, 7-3, 7-8, and 7-11 summarize comparisons between NUREG-1829 and either NUREG/CR-5750 or operating experience estimates. See also Sections 7.9 and 7.10 of the revised NUREG for more detailed comparisons.

Comment Number: 7-11 Submitted by Nuclear Energy Institute

Comment: The draft NUREG combined a variety of LOCA sources into each LOCA category. Piping LOCAs and several non-piping LOCAs were pooled together to form each of the LOCA categories. It would be useful for each of the 6 LOCA categories to add a table of LOCA sources and frequency contributions. This breakdown is particularly important for the small and medium LOCA categories.

Some contributors to the small and medium LOCAs are modeled separately in most PRA models (SGTR, RCP seals, inter-system LOCAs and others). If the end user does not subtract the separately-modeled LOCA contributors, then the contribution to CDF (core damage frequency) from those contributors would be conservatively and redundantly modeled.

Response: This comment is similar to Comment 7-2. See the response to this comment for a discussion about separating failure of individual components like steam generator tubes and CRDMs from the total LOCA frequency estimates. Section 7.8 of the revised NUREG provides separate SGTR estimates.

Comment Number: 7-12 Submitted by Nuclear Energy Institute

Comment: The various LOCA frequencies are reported in the several tables as cumulative values. In order to isolate the frequency of each LOCA category, one has to subtract the frequency of the next higher ranking category. This reporting format may lead to human errors. Some users may not become aware of the cumulative table format since that description is briefly stated at the later sections of a very large report. Please add a footnote under each LOCA frequency table explain how to obtain the frequency of each LOCA category.

Response: This comment is identical to Comment 7-4. See the response to this comment for guidance on obtaining interval values from the cumulative results reported in NUREG-1829. This point has been clarified in the revised NUREG in the Executive Summary, Section 3.4.1, and Section 7.9.

An HPI/makeup nozzle safe end from a B&W plant type was selected as one of the base case systems.

F.3.4.1 Dimensions and Welds — This type of component was considered in Reference F. 5, which identifies the component as 2 Уг inch schedule 160 pipe fabricated from Type 304 austenitic stainless steel. The location considered in Reference F.5 is in the safe end at the nozzle, which has a thickness of 11.1 mm (0.4375 inches) and a mean radius of 32.5 mm (1.28 inches) at the location of high stresses.

F.3.4.2 Stresses and Cycles -_As shown in Reference F.5, the cyclic stress history is dominated by two types of transients, with the amplitudes and frequencies shown in Table F.14.

Table F.14 Stress History for HPI/Make Up Nozzle
from NUREG/CR-6674 [F.5]

The deadweight and restraint of thermal expansion stresses for this location under normal operation that were used in Reference F.5 are

Odw-0

Cte= 63.1 MPa (9.16 ksi)

As discussed above, these stresses were composed of 310 MPa (45 ksi) uniform and the remainder the generic gradient of Equation F.6. These stresses are believed to be very conservative and are for the thermal sleeve being intact.

F.3.4.3 Results — The version of PRAISE that considers fatigue crack initiation was run for the HPI/make up nozzle. The stresses of Table F.14 were taken to be axisymmetric. Due to the small line size, only 4 initiation sites around the circumference were considered. Table F.15 summarizes the results.

Table F.15 Cumulative Probability PRAISE Results for HPI/Make Up Nozzle

(Intact Thermal Sleeve)

Table F.15 shows a cumulative leak probability of 10-5 in 25 years, which is quite low. However, leaks in this component have been observed in service, in which case the thermal sleeve in the component was failed. The results of Table F.15 use the stresses for an intact sleeve, and the stresses will be altered if the sleeve fails. A failed thermal sleeve is now considered.

F.3.4.4 Failed Thermal Sleeve — There is a thermal sleeve at the HPI nozzle, and the results in Table F.15 are for the case of the thermal sleeve not failing. The thermal sleeve has been observed to fail in service, which changes the stresses in the component.

In order to model the failure of the thermal sleeve, the following steps were taken:

1. Once the thermal sleeve fails, assume that a crack of the “initiation size” immediately appears. This size is a depth of 3.0 mm (0.12 inches). The WinPRAISE default distribution of the aspect ratio is used, as in other components.

2. A WinPRAISE run with this initial crack is performed, with the stresses that were present before the crack initiated (Table F.14), plus a uniform cyclic stress cycling each hour of sufficient amplitude to result in a high leak probability at not long times. This defines the uniform stress.

3. Use WinPRAISE to compute the leak frequencies for larger leak rates.

This procedure provides the results shown in Table F.16.

Table F.16 Cumulative PRAISE Results for HPI/Make Up Nozzle with
Failed Thermal Sleeve and Additional Uniform Cyclic Stress, ou

Table F.16 shows that a uniform stress of some 170 MPa (25 ksi) is needed to result in an appreciable leak probability within 25 years. However, the frequency of larger leak rates is actually reduced by imposing the uniform stress that is necessary to produce the high leak probabilities seen in service. This uniform stress grows cracks to leaks, so that the larger leak rate frequencies are reduced. The least favorable condition for larger leaks is a failed thermal sleeve with the original stresses (cu = 0).

F. 3.5 Recirculation Line — 12 inch

The recirculation line is one of the base case systems for a BWR. This system has developed leaks in the past due to intergranular stress corrosion cracking (IGSCC). The 12 inch line has some of the highest stresses, so is considered here. The recirculation system also has 28 inch lines, which can contribute to larger flow rate failures than possible from a 12 inch line. Hence, the 28 inch line is also considered in subsequent sections.

F.3.5.1 Dimensions and Welds — The layout of the recirculation system is given in isometrics made available to panel members. There are two recirculation loops, which are very similar to one another. There are 121 welds in this system, including field welds, shop welds and safe ends. The piping is fabricated from A-358 Class 1 Type 304, and the piping is of diameters 12, 22 and 28 inches — all schedule 80.

F.3.5.2 Stresses and Cycles — IGSCC will be the dominant degradation mechanism. Hence, time at stress is of major concern, and the number of stress cycles is of secondary importance. Estimated stresses at the highest stressed locations for the two pipe sizes of interest are given in Table F.17.

The normal operating stress (cNO) is the sum of the pressure stress, deadweight stress and restraint of thermal expansion stress. A value of 14 MPa (2 ksi) for the deadweight stress is assumed. The seismic stress is the normal operating stress plus the seismic-induced stress. Note that the seismic stresses are small in this case. The magnitude of the seismic event is unknown.

The time at stress is important for this case, with the cycling frequency being of less importance. Consistent with what is used for the PWR, the cycling is considered to be composed of heat up and cool down at 3 per year. The parameters related to stress corrosion cracking are summarized in Table F.18.

Table F.18 Stress Corrosion Cracking Parameters

Oxygen at startup (PPM) = 8.0 Oxygen at steady state (PPM) = 0.20 Heat up (100-550F) time (hrs) = 5.00 Coolant conductivity (ps/cm) = 0.20 Degree of sensitization (C/cm2) = 7.04

Residual stresses will be important, and the default residual stress distributions in pcPRAISE, which are documented in Reference F.2, are used when no remedial treatments are performed. In order to include remedial treatments that have been performed in service, a weld overlay at 20 years will be considered. This alters the thickness, crack growth kinetics (post-treatment analyses use Type 316NG crack growth defaults in PRAISE) and residual stresses. The axisymmetric through-wall residual stress distribution of Figure F.4 is employed. This figure is from Reference F.15. PRAISE can not treat the actual gradient, so the linear approximation in this figure is used. The linear gradient employed underestimates the beneficial effect of the weld overlay.

Figure F.4 Through-Wall Residual Axial Stress Distribution from Weld Overlay [F.15]

F.3.5.3 Results — Table F.19 summarizes the results obtained for the 12 inch weld in the recirculation system.

Table F.19 Cumulative Probability PRAISE Results for the 12 inch Recirculation Line Weld, with
and without Weld Overlay at 20 Years (ano = 141 MPa [20.41 ksi])

OD=12.75 inches, h=0.687 inches, wrought austenitic,
stress corrosion crack initiation and growth

5000 trials 304 full residual stress
cdw=2.0 ksi cte=13.32 ksi aDL=11.67ksi 3 HU-CD/yr p=1125 psi

The beneficial effect of the weld overlay at 20 years is not readily apparent from the results in Table F.19; such benefits are shown more clearly in Figure F.5, which provides a plot of the cumulative probability of a leak exceeding 380 lpm (100 gpm) as a function of time.

Figure F.5 Cumulative Probability of a Leak Exceeding 100 gpm as Functions of Time for the 12
inch Recirculation Line Weld with and without Weld Overlay at 20 Years

The slopes of the lines in Figure F.5 are the leak frequencies, and the slope at 40 years with no overlay is about 7 times that with overlay.

F.3.5.4 Summary of Observations from Service — Leak frequencies due to IGSCC in recirculation lines were estimated from service experience and reported in Reference F.16. Figure F.6 is Figure F.12 from that reference. With some exceptions, the results in Figure F.6 are between 10-4 and 10-3 per weld — year. The results are for times up to 15 years and do not include remedial actions. No strong dependencies on time or line size are apparent, but the smaller diameter lines appear to have a somewhat higher failure frequency.

Table F.20, which is from Charts 2 and 3 of Reference F.17, summarizes the depth distribution of observed cracks per weld-year for various pipe sizes in recirculation lines in BWRs. The remedial action of Reference F.17 is considered to consist of a weld overlay at 20 years. Observed crack sizes without remedial action, as reported in Reference F.16, are shown in Figure F.7.

Figure F.6 Leak Frequencies as a Function of Time and Pipe Size (from Reference F.16)

Table F.20 Observed Crack Depth Frequencies in Various Line Sizes in Recirculation Lines as
Percentages of the Wall Thickness (from Reference F.17)

No Remedial Action

F.3.5.5 Comparisons with PRAISE — The normal operating stress in Table F.17 of 20.41 ksi is for the highest stressed joint in the 12 inch recirculation line, whereas the observations are for all joints, including lower stressed locations. In order to generate PRAISE results that would be more representative of the population, runs were made with various stresses. Table F.21 summarizes the results.

The results for 138 MPa (20 ksi) correspond to those in Table F.19. The normal operating stress was taken to be deterministic, except for the case of 83 MPa (12 ksi), in which case the normal operating stress is normally distributed with a mean of 83 MPa (12 ksi) and a standard deviation of 0.3x(36.7+13.8) MPa (0.3x(5.32+2.00) ksi) = 15.2 MPa (2.20 ksi).

The cumulative results from Table F.21 can be compared with the observed frequencies in Table F.20 by converting the cumulative results to a frequency by dividing by the time increment involved. In the current case, the increase in the cumulative following the remedial action is relatively small, as seen from Figure F.5. Hence, the cumulative results at 25 years from Table F.21 should be divided by 20 to provide frequencies for comparison purposes. This provides the results in Table F.22.

Table F.22 Estimated Leak Frequencies Prior to Remedial Action, from Table F.21

This table shows that the mean normal operating stress of 83 MPa (12 ksi) with some variance provides the best agreement with the results of Figure F.6. This is the case that will be used for benchmarking against observed cracks.

The following steps were followed in order to provide PRAISE results for comparison with observations of part-through cracks:

1. The WinPRAISE software was modified to print out the sizes of cracks present at each time step in the analysis. The depth and length of the deepest crack and the longest crack at that time step are printed into a file, along with the number of cracks present at that time. This file contains at most a number of lines equal to the number of Monte Carlo trials times the number of time steps (which can be a lot of lines).

2. The WinPRAISE file from step 1 is then processed to provide another file that includes only the sizes of part-through cracks present at the time of interest (25 years in this case). (Cracks of zero depth, leaks and other times are eliminated.)

3. The crack size file from step 2 is then loaded into a histogram, which provides the number of cracks present at 25 years that fall within a certain depth range.

4. Since Reference F.17 reports detected cracks, the detection probability (Equation F.1) must be accounted for. This is accomplished by multiplying the number of cracks in each bin by the detection probability for a crack of depth equal to the midpoint of the bin. This provides the number of detected cracks in this bin. The contents of each bin are then divided by the number of trials times the time (25 years) to provide the crack sizes per weld year.

5. The histogram is then converted to a complementary cumulative form, which is then directly comparable to results from Reference F.17.

Figure F.8 presents the crack size results for the benchmark case. Once again, not many deep cracks are observed. A pattern is observed in Figure F.8 which shows a preponderance of cracks below about 2.5 mm (0.1 inches). This pattern is due to cracks growing to a depth of 2.5 mm (0.1 inch) and then slowing

down or arresting, which is most likely due to the transitioning from growth of “initiating cracks” to “fracture mechanics cracks” that occurs in the PRAISE modeling of initiation and growth. The transitioning criteria are discussed on page 42 of Reference F.2, and one of the criteria is “If the depth of the crack is greater than 2.5 mm (0.1 inch), its growth will always be by fracture mechanics velocity”.

Figure F.9 is a plot of the predicted complementary cumulative number of observed cracks for the benchmark case, along with a comparison with reported observations. The outstanding inspection parameters of Table F.1 were employed. In this figure, Reference F.17 results from Chart 1 (prior) and Chart 2 (posterior) are both shown, since the analysis mixed with and without remedial action (weld overlay at 20 years). The analysis results fall midway between the two results, except for shallow cracks.

о

&

nj

13

M

Figure F.8 Crack Sizes After 25 Years for the Benchmark Case (Mean oNo = 12 ksi) with Weld

Overlay at 20 Years

Figure F.9 Comparison of Results for the Benchmark Case Predicted for Outstanding Inspection
Quality with Reported Prior and Post Observations [F.17]

The agreement shown in Figure F.9 is felt to be quite good, and indicates that the PRAISE model best fits the observed crack depths when the mean stress of 83 MPa (12 ksi) is used. Figure F.10 shows that the stress has an important effect, because the agreement is not so good when a stress of 103 MPa (15 ksi) is employed.

Figure F.10 Comparison of Results for the High-Stress Case Predicted for Two Inspection
Qualities with Reported Observations (the Square Data Points are the Observations [F.17], no

Weld Overlay)

The results of Figure F.7 are for observed (detected) cracks, whereas Figure F.8 has not had the nondetection probability applied. A direct comparison is therefore not possible, and the two figures are plotted on much different scales. However, the PRAISE predictions contain a much greater proportion of short and shallow cracks than the observations. This could be somewhat affected by the nondetection probabilities, but the differences would not be removed by applying the nondetection probabilities to the cracks predicted by PRAISE.

The question immediately arises regarding the results to be used in the estimation of the recirculation system reliability; the results of Table F.20 for the highly stressed joint or the results of Table F.21 for the benchmarked stress of 83 MPa (12 ksi). Interestingly, the system and average weld leak frequencies are nearly the same whether the 83 MPa (12 ksi) or 140 MPa (20 ksi) weld is used, because the number of joints involved also depends on the stress. Table F.23 summarizes this comparison. This table is in terms of the leak frequency per year, which is obtained from the cumulative results given above by use of Equation F.2.

The system frequency and average frequency per weld are nearly the same for both cases.

Table F.23 System and Average Weld Leak Frequencies for Two Cases
of the 12 inch Recirculation Line

Of the two cases in Table F.23, the case of a mean stress of 83 MPa (12 ksi) and coefficient of variation of 0.3 (on cte + Cdw) is more representative of the population of joints as a whole, so is preferred for comparisons with observations of part-through cracks.

One of the challenges in the LOCA frequency estimation is trying to predict the probability of an event that has never happened before, but which has enormous consequences if it did. It is important to maintain a sense of balance in this effort and aim for a realistic approach that is based strictly on technical considerations. As in any probabilistic analysis, the success depends on how realistic the inputs are and how the approach reflects actual field experience. Having worked in the BWR industry for almost 30 years, Dr. Ranganath felt that his most important contribution to the elicitation process was to make sure that that frequency estimates reflect BWR field experience. For example, use of probabilistic defect distribution data is acceptable as long as the prediction is consistent with actual field behavior. Dr. Ranganath’s philosophy was to start from actual field data and to predict future behavior based on his understanding of failure mechanisms, mitigation measures and BWR systems design. Since his knowledge is mainly on BWR systems, he focused his attention on BWRs rather than PWRs. He did not want to speculate in areas where he did not necessarily have the expertise. There are other people who are more knowledgeable about PWRs and they can do a better job on the estimates. He felt that the diversity of the elicitation panel and their expertise and the open mindedness of the NRC team helped in coming up with the best estimates.

An assessment of the effect of ISI was carried out for the surge line elbow weld, the defect distribution and density being those generated by RR-PRODIGAL, see section G.4.2. A Probability of Detection (POD) curve was defined by the following equation:

This POD is shown in Figure G.7, and it can be seen that this sets the probability of detection at about 90 percent for defects 70 percent of the way through the wall thickness. This was felt to be

representative of inspections carried out to date, but for future inspections that conform to modern standards, this POD could be much better.

The results are shown in the table below and in Figure G.8 for various ISI intervals.

Table G.5 Reduction Factors Due to ISI

These results suggest that even with this quite low inspection capability, and for a weld with a high failure probability, reductions of a decade can be achieved with two or three inspections during the life of the plant. It also indicates that going beyond three inspections gives little extra return.

An interesting conclusion from this figure would be that if a fourth inspection is carried out at the end of a forty year period, then, provided this inspection was clear, there would be little gain from an inspection at fifty years for a total life of sixty years! However, at this stage such a conclusion can only be taken as tentative and would require more investigation.

G.6 References

G. 1 NUREG/CR-5505 PNNL-11898 ‘RR-PRODIGAL — A Model for Estimating the Probability of Defects in Reactor Pressure Vessel Welds.

G. 2 NURBIM (Nuclear Risk-Based Inspection Methodology) WP4. Published by the European Commission under the EURATOM programme.

Figure G.1 Leak Rate as a Function of Leakage Area (Data Supplied by USNRC)

Figure G.2 Estimated Defect Separation (COD) Based on Expert Judgment
as a Function of Defect Length Assuming Plastic Deformation

0 5 10 15 20 25 30 35

Full Defect Length (Inches)

Figure G.4 Probability of the Existence of a Defect of a Certain Length for Surge Line Base Case

ATTACHMENT 1 TO APPENDIX G (ATTACHEMENT G.1) FROM
NOTE ‘SUMMARY OF BASE CASES STRESSES’ APRIL 2003

Summary of Stress Cycles for Surge Line Elbow
(No seismic stresses)

Many of the high stress contributors in Tables 2 and 3 are from rapid excursions of the coolant temperature. The largest stress amplitude (half the peak-to-peak) is 1,310 MPa (190 ksi), so the stresses are large (but localised). These are the stresses at the peak stress location, which is not at weld. The spatial stress gradients (both along the surface and into the pipe wall) are required for a thorough analysis. The radial gradient (into the pipe wall) can be estimated by the following procedure:

1 Cyclic stresses associated with seismic loads were treated as 100 percent uniform stress.

1 Cyclic stresses greater than 310 MPa (45 ksi) were treated as having a uniform component of 310 MPa (45 ksi), and the remainder were assigned to the gradient category.

1 For those transients with more than 1,000 cycles over a 40- year life, it was assumed that 50% of the stress was uniform stress and 50% a through-wall gradient stress. In addition, for these transients, the uniform stress component was not permitted to exceed 70 MPa (10 ksi).

The gradient stress mentioned above is assumed to vary through the thickness as

o(%) = ^0 ^1 — 3£ + 2 j (G.2)

In this equation, c0 is the stress at the inner wall of the pipe, % = x / h, x is the distance into the pipe wall from the inner surface, and h is the wall thickness. The stresses and cycles are high enough that fatigue crack initiation is important, which has been considered in Reference 6. Reference 6 shows a probability of 0.981 of a leak in 40 years for this component. The LOCA probabilities will be less. The use of the gradient along the surface will reduce this.

As mentioned above, the more thorough results that include a better estimate of the radial gradient and also consider the spatial gradient along the surface are available, and could be used for the base case calculations.

Submitted by Nuclear Energy Institute (NEI)

Comment: As the NUREG-1829 report may be considered to be the “most recent applicable data” upon finalization, it is important that the final report provide an alternative to continue using operational experience data for the determination of small break LOCA frequencies. Most PRAs currently reference NUREG-5750, which used such a basis (at the time there were 1,250 reactor years of operating experience) to estimate small break LOCA frequencies. Since issuance of NUREG-5750, over one thousand additional reactor years of operational experience have confirmed the conclusions of NUREG — 5750 relative to small break LOCAs. Draft NUREG-1829 notes that, when steam generator tube rupture data are excluded, there is general correlation on small break frequencies with NUREG-5750. However, our review of the report indicated that draft NUREG-1829 estimates these frequencies over one order of magnitude higher than the estimate of NUREG-5750. Using the NUREG-1829 small break LOCA frequency estimation, the US reactor fleet should be experiencing one small break LOCA on average every 4 years. However, no such LOCAs have occurred in the operating history of the US plants. Obviously, the incorporation of this frequency estimate into existing PRAs would lead to unwarranted impacts that are out of context with reality.

Response: Comparing results from NUREG-1829 and NUREG/CR-5750 must be done with some care due to differences in how the LOCA frequencies were calculated. The SB LOCA frequencies typically reported in NUREG-1829 for PWRs include contributions from SGTRs while the SB LOCA frequencies reported in NUREG/CR-5750 exclude SGTRs. Furthermore, the LOCA frequencies in NUREG-1829 are based on threshold leak rates (Category 1 LOCAs include all leaks greater than 100 gpm [380 lpm], Category 2 LOCAs include all leaks greater than 1,500 gpm [5,700 lpm], etc.). Conversely, the SB LOCA results in NUREG/CR-5750 include only those events with leak rates between 100 and 1,500 gpm (380 and 5,700 lpm).

These distinctions were considered in making the comparisons provided in Section 7.9 that are summarized in Table 7.20. The reported SB mean LOCA frequency estimates for NUREG/CR-5750 are 4.0E-04 per calendar year and 7.4E-03 per calendar year for BWR and PWR plants, respectively. Note that these PWR SB LOCA estimates include the steam generator rupture frequencies calculated in NUREG/CR-5750. The revised NUREG-1829 SB LOCA estimates are 5.2E-04 and 6.6E-03 for BWR and PWR plants, respectively (geometric mean with overconfidence adjustment). The ratio of the NUREG/CR-5750 to the NUREG-1829 results is 0.76 for BWRs and 1.12 for PWRs. These SB LOCA estimates are therefore similar.

It is also interesting to compare the PWR SB LOCA frequency estimates after excluding SGTR frequencies from the Category 1 LOCA estimates (Table 7.19). As reported in Section 7.8 of NUREG 1829, the SGTR rupture frequencies predicted by the elicitation, NUREG/CR-5750, and operational experience are consistent. The cumulative SB LOCA Category 1 and 2 estimates without SGTR contributions are 1.9E-03 and 4.2E-04 per calendar year (Table 7.19). Therefore, the interval value that corresponds to the historical SB LOCA definition (i. e., breaks between 100 and 1,500 gpm [380 and 5,700 lpm]) is 1.48E-03 per calendar year. This value is approximately 3.7 times higher than the mean SB LOCA frequency from NUREG/CR-5750 of 4E-04 per calendar year.

The increase in the elicitation result reflects the panelists’ opinion that current PWR SB LOCA frequencies for components other than steam generator tubes are higher than historical averages. This increase primarily stems from current PWSCC concerns (Section 6.3.2). The practical implications from these differences however are not striking. The NUREG/CR-5750 estimate translates to an expected PWR SB LOCA every 36 years for the current fleet of 69 operating PWRs. The elicitation frequency corresponds to an expected PWR SB LOCA every 10 years. Additional comparisons between the NUREG-1829 and NUREG/CR-5750 estimates are contained in Section 7.9.

Because of this and similar comments, Section 7.10 has been added to NUREG-1829 to compare the elicitation LOCA frequency estimates with estimates derived from operating experience. This comparison shows that the differences between the elicitation and operating experience-based estimates of the non-SGTR, PWR Category 1 LOCA frequencies are not statistically significant. However, even though these differences are not statistically significant, this does not imply that the two approaches are estimating the same frequency. Because an operating experience-based estimate is an historical average based on many years of operation, a difference will exist if the panelists believe that the current failure frequency differs from the historical average. In fact, the increased elicitation estimate is supported by the panelists’ qualitative and quantitative responses. As noted above, the panelists indicated that medium and, to a lesser extent, small LOCAs in PWRs are most dramatically impacted by PWSCC in relatively small diameter passive system component (e. g., CRDMs, instrument nozzles, etc.) (Section 6.3.2).

Additional details on this comparison are found in Section 7.10. Related information is also found in the responses to GC4, GC5, GC6, GC7, and 7-8.

SIGNIFICANT FAILURES OF SAFETY RELATED PIPING

Table D. D.1 is a list of selected significant pipe failures during the period 1970 — 2003. The list includes failures of Code Class 1 and 2 piping systems inside the containment/drywell and auxiliary/reactor building structures of commercial nuclear power plants. The technical information has been extracted from the OPDE database (Attachment A).

Table D. D.1 Selected Historical Pipe Failure Information

PIPING BASE CASE RESULTS OF
WILLIAM GALYEAN

Figures I.7 and I.8 present plots of these RPV base cases, compared to the piping base cases from Appendices D, E, F, and G. For purposes of this comparison, a single set of piping base case LOCA frequencies were derived that are a composite of the results from the four appendices. Plots are presented for the 0-25 year (Figure I.7) and the 25-40 year (Figure I.8) periods. Since the RPV LOCA frequencies for the 40-60 year period are not significantly different than the 25-40 year results, a separate plot for that case is not included. It is seen from these figures that the RPV base cases are at the low end of the piping

LOCA probabilities for the large break Categories 5 and 6, but are at the high end for small, Category 1 and 2 breaks, due largely to the small penetration (CRDM) contributions discussed above. Note also that the small LOCA probability estimates are substantially lower in the outlying years (25-40 and 40-60) because of inspection programs implemented as a result of these issues. In general, small break LOCA frequency contributors (Categories 1 and 2) from PWR RPVs are seen to be greater than those for BWRs, due to the PWSCC concern in CRDM and other small penetrations. Large break LOCA contributors (Categories 5 and 6) are also estimated to be greater for PWR RPVs due to higher irradiation embrittlement and the potential for PTS transients.

I-7

I.6 References

1.1 Peter Riccardella, Nathaniel Cofie, Angah Miessi, Stan Tang, Bryan Templeton, “Probabilistic Fracture Mechanics Analysis to Support Inspection Intervals for RPV Top Head Nozzles” U. S. Nuclear Regulatory Commission / Argonne National Laboratory Conference on Vessel Head Penetration Inspection, Cracking, and Repairs, September 29 — October 2, 2003, Gaithersburg, Maryland.

1.2 Materials Reliability Program, MRP-105, “Probabilistic Fracture Mechanics Analysis of PWR Reactor Pressure Vessel Top Head Nozzle Cracking,” EPRI Report 1007834 (EPRI Licensed Material), May, 2004.

1.3 EPRI Report, “BWR Reactor Pressure Vessel Shell Weld Inspection Recommendations (BWRVIP — 05),” TR-105697, September 1995.

1.4 NRC Report, “Final Safety Evaluation of the BWR Vessel and Internals Project BWRVIP-05 Report (TAC No. M93925),” Division of Engineering, Office of Nuclear Reactor Regulation, May 1998.

1.5 NUREG-0619, “BWR Feedwater Nozzle and CRD Return Line Nozzle Cracking, Resolution of Generic Tech Activity A-10,” November 1980.

1.6 U. S. NRC Order EA-03-009, “Interim Inspection Requirements for Reactor Pressure Vessel Heads at Pressurized Water Reactors”, issued on February 11, 2003.

1.7 VIPER Version 1.2, Structural Integrity Associates, Report # SIR-95-098 Rev. 1, Feb. 1999.

0.90
0.63 0.50
0.20
0.10
0.05
0.02
0.01
1	10	100
EDYs

Figure I.1 Weibull Plot of Plant Inspection Data Showing Extrapolation Back to Time of First Leakage or Cracking. Plants that Performed NDE and were Found Clean are

Treated as Suspensions

1.00E-01

Figure I.2 Benchmarking of PFM Crack Growth Analyses with Respect to Field-Observed

Circumferential Cracking of Various Lengths

2.0E-03
0.0E+00
10 15 20 25 Operating Years
30	35	40
5

0

Figure I.3 RPV Top Head PFM Analysis Results for Plant with 580°F (304°C) Head Temperature — Probability of CRDM Nozzle Failure (i. e. Ejection of Nozzle from Vessel Head)

Figure I.4 RPV Top Head PFM Analysis Results for Plant with 580°F (304°C) Head Temperature —

Probability of Leakage from CRDM Nozzle

Figure I.6 Historical BWR Feedwater Nozzle Cracking Experience (circa 1980)

1.00E-04
1.00E-05
1.00E-08
1.00E-09
1500	5000	25000	500000
1.00E-10

Break Size (G PM)

Figure I.7 Comparison of RPV and Piping Base Case LOCA Frequencies Versus Break Size

(0 to 25 Years)

Figure I.8 Comparison of RPV and Piping Base Case LOCA Frequencies Versus Break Size

(25 to 40 Years)

Comment Number: 8-1

Submitted by Westinghouse Owners Group

Comment: In Chapter 8, Ongoing Work, it is noted that the LOCA elicitation results were for normal operating conditions only. The effects of Service Level D transients, of which seismic was found by NRC to be the most prominent, were not considered in the elicitation efforts. The reason seismic loading was not explicitly considered was that most of the expert panel did not believe that it would significantly change the LOCA frequencies for normal operation. Experience from probabilistic fracture mechanics calculations indicates that severe seismic loading, such as that from a design-basis safe shutdown earthquake, could increase the conditional probability of failure in flawed piping by one to two orders of magnitude. However, the probability of having the severe seismic loading during the worst time in life, such as the 40th or 60th year of operation, would be a maximum of 0.001 and would likely be much less. Thus, the maximum effect of this severe seismic loading would be to increase the LOCA frequency during normal operating conditions by 10 percent. This increase was deemed to be insignificant relative to the other uncertainties that were considered by the expert panel in the elicitation process for LOCA frequencies.

Response: The results from a separately-sponsored NRC-led seismic LOCA study (Reference: Chokshi, N. C., Shaukat, S. K., Hiser A. L., DeGrassi, G., Wilkowski, G., Olson, R., and Johnson, J. J., "Seismic Considerations For the Transition Break Size," NUREG-1903, U. S. Nuclear Regulatory Commission, February 2008) tend to support this comment. In this study, both unflawed and flawed piping analyses were conducted in order to ascertain the magnitude of any potential adjustments to the baseline TBS for the proposed rule change to 50.46a due to failures associated with seismic loading.

The principal findings from this study are that the critical flaws associated with the stresses induced by seismic events of 10-5 and 10-6 annual probability of exceedance are large. When considering the effects of mitigation strategies to preclude large flaws in service, the probabilities of pipe breaks larger than the TBS are likely to be less than 10-5 per year. Similarly, for the cases studied, the probabilities of indirect failures of large RCS piping systems are less than 10-5 per year.

These findings tend to support the contention of the commenter that seismic loading would not significantly change the LOCA frequencies under normal operation. As a result of this and related comments, the NRC report on Seismic Considerations for the Transition Break Size is now referenced in Section 2 of NUREG-1829. In addition, a summary of the seismic LOCA analysis and results is provided in the Executive Summary and in Section 7.2 of the report. Additionally, Section 2 clearly identifies that the elicitation LOCA frequency estimates do not consider rare event loading from seismic, severe water hammer, and other similar sources.

Comments Related to Section 9 of NUREG-1829

None

Comments Related to Appendix A of NUREG-1829

None

Comments Related to Appendix B of NUREG-1829

None

Comments Related to Appendix C of NUREG-1829

None

Comments Related to Appendix D of NUREG-1829

Comment Number: D-1

Submitted by Joseph Conen of the BWR Owners Group

Comment: Figure D.7 in Appendix D shows two through-wall IGSCC cases for 22 inch and 28 inch stainless steel pipe field history data. This reviewer is not aware of any through-wall IGSCC cracks in large diameter (>20-inch) BWR stainless steel pipes. A primary reason for this is the presence of mid-wall compressive weld residual stresses in such pipe that tend to retard deep cracks.

Response: Figure D.7 in the draft NUREG only shows selected IGSCC data points (only weld flaws for which detailed sizing data are available). In actuality there are have been other leaks in 22-inch and 28- inch diameter recirculation lines in BWRs. According to the expanded OPDE database used as the basis of this query (currently 1,215 records on IGSCC), there have been 10 instances of circumferential through-wall cracking in large diameter (D=22 inch to 28 inch recirculation system piping, 8 of which were leaks. Three of the leaks were in 22-inch diameter piping: a cap-to-manifold leak at Hatch-1 (LER 82-089, November 1982) and two welds at Monticello (LER 82-013, October 1982). The other five leaks were associated with the 28-inch diameter recirculation line at Brunswick-1 (LER 85-026, July 1985): Weld 1B32-RR-28-A-4, Weld 1B32-RR-28-A-14, Weld 1B32-RR-28-B-4, Weld 1B32-RR-28-B-8, and Weld 1B32-RR-28-A-15.

Comments Related to Appendix E of NUREG-1829

None

Comments Related to Appendix F of NUREG-1829

None

Comments Related to Appendix G of NUREG-1829

None

Comments Related to Appendix H of NUREG-1829

None

Comments Related to Appendix I of NUREG-1829

None

Comments Related to Appendix J of NUREG-1829

None

Comments Related to Appendix K of NUREG-1829

None

Comments Related to Appendix L of NUREG-1829

None

BIBLIOGRAPHIC DATA SHEET

NUREG-1829, Vol. 2

(See instructions on the reverse)

3. DATE REPORT PUBLISHED MONTH YEAR April 2008 4. FIN OR GRANT NUMBER ___________ N6360 6. TYPE OF REPORT

Technical

7. PERIOD COVERED (Inclusive Dates)

N/A

8. PERFORMING ORGANIZATION — NAME AND ADDRESS (If NRC, provide Division, Office or Region, U. S. Nuclear Regulatory Commission, and mailing address; if contractor, provide name and mailing address.)

Division of Engineering Battelle-Columbus

Office of Regulatory Research 505 King Avenue

U. S. Nuclear Regulatory Commission Columbus, OH 43201

Washington, DC 20555-0001

9. SPONSORING ORGANIZATION — NAME AND ADDRESS (If NRC, type "Same as above"; if contractor, provide NRC Division, Office or Region, U. S. Nuclear Regulatory Commission, and mailing address.)

Same as above

10. SUPPLEMENTARY NOTES

A. Csontos, NRC Project Manager_____________________________________________________________________

11. ABSTRACT (200 words or less)

The NRC is developing a risk-informed revision of the design-basis pipe break size requirements in 10 CFR 50.46, Appendix K to Part 50, and GDC 35 which requires estimates of loss-of-coolant-accident (LOCA) frequencies as a function of break size. Separate BWR and PWR piping and non-piping passive system LOCA frequency estimates were developed as a function of effective break size and operating time through the end of license extension. The estimates were based on an expert elicitation process which consolidated service history data and insights from probabilistic fracture mechanics studies with knowledge of plant design, operation, and material performance.

The elicitation required each member of an expert panel to qualitatively and quantitatively assess important LOCA contributing factors and quantify their uncertainty. The quantitative responses were combined to develop BWR and PWR total LOCA frequency estimates for each contributing panelist. The individual estimates were then aggregated to obtain group estimates, along with measures of panel diversity. Sensitivity studies were conducted to examine the effects of distribution shape, correlation structure, panelist overconfidence, measures of panel diversity, and aggregation method. The group estimates are most sensitive to the method used to aggregate the individual estimates.

[1] Each panelist’s quantitative elicitation responses can be found through the “Electronic Reading Room” link on the NRC’s public website (http://www. nrc. gov/) using the Agencywide Documents Access and Management System (ADAMS). The document is found in ADAMS using the following accession number: ML080560005.

[2] The nomenclature for the table and figure numbers is such that the letter B refers to Appendix B, the first number (1 or 2) refers to a figure associated with either the first or second panel meeting, and the second number refers to the numerical sequence of that particular table or figure in the text for the applicable meeting, i. e., either first or second.

[3] See for example the report EPRI NP-2472 (The Growth and Stability of Stress Corrosion Cracks in Large-Diameter BWR Piping, July 1982).

[4] The figure is reproduced courtesy of K. N. Fleming (Technology Insights, Inc., San Diego, California).

[5] Details on Bayesian reliability analysis is found in text books on statistical analysis of reliability data; e. g., Martz and Waller (1991): Bayesian Reliability Analysis, Krieger Publishing Company, Malabar (FL), ISBN 0-89464-395-9. For conjugate functions like the gamma and beta distributions a Bayesian point estimator for the failure rate is the mean of respective posterior probability density function, or:

X = (8 + r)/(p + T) — gamma X = (8 + r)/(8 + p + n) — beta

Where, (8, p) are the parameters of respective distribution and (r, T, n) correspond to new evidence (i. e., ‘r’ failures in ‘T’ hours, or ‘r’ failures in ‘n’ tests).

[6] See SKI 98:30 [D.15] for details.

[7] This table includes active leaks (= leaks detected during routine power operation) and ‘non-active’ leaks (= leaks discovered during

change of plant mode of operation), but it excludes ‘ISI-leaks.’ Appendix A, Table A-5 includes details on the through-wall cracks in NPS12, NPS22 and NPS28 Reactor Recirculation piping as included in Table D.4 above._________________________________________________________

[8] The mean of weld count in NPS20-, 22- and 24-piping.

[9] NPS30 is used to characterize the CL — and HL-piping.

[10] Table 2-3, page 2-10; Aj = 1.34E-05 (RF = 100). TR-111880: Piping System Failure Rates and Rupture Frequencies for Use in Risk Informed In-Service Inspection (September 1999).

[11] The term ‘non-active leak’ is taken to mean a through-wall flaw without visible leakage or with a small, detectable leakage that stays relatively constant over time.

[12] The database includes a single event involving the fracture of a small-diameter steam line due to seismic event (Fukushima-Daiichi Unit 6 on 07-21­2000).

[13] See for example T. V. Vo et al (1991). “Estimates of Rupture Probabilities for Nuclear Power Plant Components: Expert Judgment Elicitation,” Nuclear Technology, 96:259-270.

[14] At the time of the pipe break the primary system was at 278 C (533 F) and 15.3 MPa (2,225 psi) primary system pressure with a secondary system pressure of 6.2 MPa (900 psi)

[15] The effects of applying a load-controlled overload stress at a specified time were studied. This is called a design-limiting stress, and represents an overload event, such as water hammer or a seismic event even larger than the 5 SSE already considered for this component.

[16] This information needed to be supplied because the transient experience for the Naval Nuclear program is a) confidential and b) not applicable to commercial plants.

[17] Failures are classified using four categories: partial through-wall cracks, through-wall cracks without a significant leak rate (typically indicated by a boric acid deposit), leaks, and joint failures (i. e., non-welded connection).

[18] A half failure (0.5) was added to all degradation mechanism (DM) totals to force the representation of all DMs.

[19] Lydell, B., “Independent Review of SKI 96-20 Database,” Technical Note 1996-01, SKI Ref. No. 14.2-940477, February 1996.

[20]

Letter from Frederick P. Schiffley, II, Chairman to USNRC, “Westinghouse Owners Group Comments on Draft NUREG-1829, ‘Estimating Loss-of-Coolant Accident (LOCA) Frequencies Through the Expert Elicitation Process’ (MUHP-3062)”, WOG-05- 517, dated November 28, 2005, ADAMS Accession # ML0503340274.

Stresses and dimensions are given in the corresponding sections for the 12 inch line. IGSCC crack initiation and growth are the dominant degradation mechanisms. Table F.24 summarizes the results for this weld.

Table F.24 Cumulative PRAISE Results for the Weld
in the 28 inch Recirculation Line

odw=2.0 ksi

ote=1.75 ksi P = 1,125 psi Type 304 full residual stress 3 HU-CD/yr

F.3.7 Feedwater Elbow

The feedwater elbow is one of the base case systems. This system is subject to flow accelerated corrosion (FAC), which can be a serious degradation mechanism if left unchecked. PRAISE can not model FAC, but some analyses are provided for fatigue crack initiation and growth.

F.3.7.1 Dimensions and Welds — The layout of the feedwater system is given in the piping isometrics made available to the panel members. There are some 123 welds in the two loops of the feedwater systems, all but 6 of them in 12 and 20 inch piping. The 12 inch lines are schedule 100 (17.4 mm [0.687 inches] thick) and the 20 inch lines are schedule 80 (32.5 mm [1.281 inches] thick). The material is A — 333 Grade 6 (which is a carbon steel).

F.3.7.2 Stresses and Cycles — The feedwater line elbow is considered in Reference F.5, so this is evidently the high stress point in the system. Note that there are at least 6 such elbows in a feedwater system. (There are many more elbows, but they are likely to not be so highly stressed). The degradation mechanism is fatigue and flow accelerated corrosion (FAC). Stresses do not contribute to FAC, so are not needed for this mechanism. For fatigue, there are a considerable number of cycles of high stress amplitude. They are available from Reference F.5. Table F.25, which (except for the column of temperatures) is page A.25 of Reference F.5, summarizes the stresses. These stresses are “decomposed” according to the procedure discussed above for the surge line. The analysis reported in Reference F.5
used a temperature of 590°F (310°C), as indicated in the text at the top of Table F.25. However, Table 5­123 of Reference F.18 provides the temperatures for these transients, and it is suggested that these temperatures be used, because their use is more realistic and less conservative. They are included as the right-hand column of Table F.25. The temperature influences the strain-life curve, and has a noticeable effect on the computed failure probabilities because of its influence on the initiation probabilities.

The values of the deadweight and restraint of thermal expansion under normal operation that Reference F.5 uses for this location are

Cdw = 0

Cte = 115 MPa (16.68 ksi).

The stress history in Table F.25 most likely contains seismic events. It is not possible to eliminate them from the list using information currently available, but their influence on the calculated failure probabilities is expected to be minimal.

Table F.25 Summary of Stress Cycles for Feedwater Line Elbow
(from Page A.25 of NUREG/CR-6674 [F.5])

F.3.7.3 Results — PRAISE runs for this component were made using the version that can treat fatigue crack initiation with details of the circumferential variation of the stresses. The feedwater system is

relatively more likely to experience water hammer, so the influence of an overload event with a stress of 0.42cflo = 128 MPa (18.5 ksi) above that normally present was considered. This stress is denoted as cDL, and results were generated for one cycle of this stress at 24, 39, or 59 years. The results are summarized in Table F.26, which includes the effects of cDL (columns D & F).

Table F.26 Cumulative PRAISE Results for Feedwater Line Elbow

Case A is directly from Reference F.5, and Case B is directly from Table 4-8 of Reference F.6. Case C is Case B rerun with 108 trials. Cases D-G are variations of C with different failure criteria, overloads and reduction of stresses. The results for various failure criteria (critical net section stress only or critical net section stress and tearing instability) show that consideration of tearing instability noticeably increases the computed failure probability (compare, for instance, cases C&E). Consideration of an overload event also has a noticeable effect (E&F). The use of lower stresses markedly reduces the computed failure probabilities (G & C). In the case of an overload event, the probability of a 100 gpm failure is the same as a complete pipe break.

F.3.7.4 Alternate Procedure — The results of Table F.26 show that the probability of a large leak was obtainable from the Monte Carlo procedure only when a large overload occurred. When this did not occur, there were no leaks of even 380 lpm (100 gpm) in 107 or 108 trials. In order to obtain estimates for the larger leak probabilities, the alternate procedure discussed for the surge line was also applied to Case C of Table F.26 for the feedwater elbow.

As before, the crack length for a given leak rate, b( q), was obtained from a pcPRAISE run, along with the half-crack length of any cracks that become through-wall. Figure F.11 provides a plot of the leak rate as a function of b for the feedwater elbow.

Figure F.11 Leak Rate as a Function of Half Crack Length for Feedwater Elbow Base Case C

The results in Table F.27 are obtained from this figure and the corresponding pcPRAISE results. This table also includes the portion of the circumference that is cracked and the proportion of the crack opening area to the flow area of the pipe. It is seen that the opening area of the crack is nearly equal to the flow area of the pipe when the leak rate is 19,000 lpm (5,000 gpm). The value of b for a complete pipe break, as obtained from Equation E.7 is also included. Table F.29 defines b( q).

Table F.27 Half Crack Lengths and Areas for a Given Leak Rate
(Feedwater Elbow Base Case C)

As before, the next step is to estimate the probability of having a through-wall crack exceeding a given length as a function of time. The modified version of pcPRAISE was used to generate a table of values of b and the time at which the leak first occurred. A run was made with 107 trials, with 2,607 cracks becoming through-wall within 60 years. This corresponds to a leak probability of 2.607×10-4 at 60 years, which agrees closely with the leak probability obtained earlier. Of these 2,607 cracks, none appeared before 25 years, and 64 occurred between 25 and 40 years. The statistical distribution of these 64 cracks at 40 years provides the probability of having a through-wall crack greater than a given length within 40 years. Extrapolation is required to obtain results for the crack lengths included in Table F.27. Figure F.12 shows the complementary cumulative distribution of b at 40 years, along with the curve fit of Equation F.10.

P(> b) = e“534(b-1) (40 years) [F.10]

Note that the plot starts at a half-crack length of 25 mm (1 inch), and that the data are closely approximated by a straight line on log-linear scales.

Figure F.12 Complementary Cumulative Distribution of Half-Crack Length of Through-Wall
Cracks in Feedwater Elbow within 40 Years, Along with Fit

Figure F.13 provides a similar plot for the 2,607 through-wall cracks that occurred within 60 years. Equation F.11 is the fit of the distribution at 60 years within the range of interest.

P(> b) = 0.0274e“2’25(b_1) (60 years) [F.11]

Note that in this case the data appear bilinear and are not well approximated by a straight line on log — linear scales. To represent the data at the longer crack lengths of interest, a straight line was assumed beyond a crack length of 50 mm (2 inches). This corresponds to a probability below about 0.003.

rO

-Й

ь

яз

Ы>

Й

‘>

яЗ

Л

<4-1

0

1

-D

О

1-І

Й4

Figure F.13 Complementary Cumulative Distribution of Half-Crack Length of Through-Wall
Cracks in Feedwater Elbow within 60 Years, Along with Fit

The probability of a leak exceeding a given size within 40 and 60 years is then obtained by taking using the value of b for a given leak rate from Table F.27 in conjunction with Equations F.10 and F.11, respectively. Table F.28 summarizes the results.

Table F.28 Cumulative Results for Feedwater Elbow Case C

The leak (>0) results in Table F.28 came directly from the Monte Carlo simulation. With 108 trials, no leaks exceeding 380 lpm (100 gpm) were obtained. Hence, the Monte Carlo simulation predicts <10-8 probability of a leak exceeding 380 lpm (100 gpm) within 60 years. The alternative procedure gave a corresponding value of 6.44×10-7. This suggests that the alternative procedure overestimates the probability of a given leak, as was also the case for the surge line elbow.

The first step in the expert panel elicitation was to develop an amalgamated set of base case LOCA frequencies upon which the elicitation responses are anchored. The generic base case LOCA frequencies developed for the panel represented the work of four teams: two teams used an empirical approach based on operating plant experience with leakage and other precursor events, while two other teams used theoretical, PFM analyses. Each of these approaches has different strengths and weaknesses, such that a better estimate of base case LOCA probabilities can be achieved by selectively combining the results in a manner that optimizes the strengths of both. The method and rationale for combining the base case results of the individual teams were documented, ultimately producing a revised set of LOCA frequencies for the five piping base cases.

LOCA frequencies for each of the LOCA sensitive piping systems identified for PWRs and BWRs were then estimated. This was done by picking the base case which is most representative of the specific LOCA sensitive system, considering plant type, material of construction, operating conditions and relevant degradation mechanisms, and then scaling the base case frequencies for each LOCA category based on judgment of any substantive differences between the base case and the system under evaluation. One of the main factors accounted for in this scaling process was differences in the size of the systems, in terms of number of welds of various pipe sizes (based on a system-by-system weld census that was provided to the panel). Scaling was also used to account for system specific factors, such as whether repairs or mitigation have been applied to address degradation mechanisms considered in the base cases, and the timing of such actions. An estimate of the probability of breaks in small diameter socket welded piping (instrument, vent and drain lines) due to vibration fatigue was made, which was not included in any of the base cases. This estimate was based on prior experience with this relatively common failure mechanism.

It was felt that there is a relatively large uncertainty band in all of the above probability estimates; plus or minus an order of magnitude. Included in this uncertainty band is the potential development of new, as yet unseen degradation mechanisms in the future, which obviously weren’t considered in the base cases.

A set of base cases for non-piping LOCAs was developed, the methodology for which is documented in Appendix I to this NUREG report. These base cases included potential breaks due to small vessel penetrations such as CRD nozzles, medium size breaks due to larger diameter nozzles (excluding safe-end ruptures which are included in the piping base cases), and very large breaks due to pressure vessel ruptures (specifically addressing irradiation embrittlement of the RPV). The resulting base cases were then used to estimate contributions to LOCA frequency from non-piping LOCAs.

The detailed rationale used in developing the elicitation response for each system was documented in a report, to permit the reconstruction of the logic in the future if it becomes necessary.