ELICITATION TRAINING EXERCISE RESULTS

As part of the panel member kick-off meeting in February 2003, elicitation training was provided for the elicitation panel. The training involved the panel members answering a series of almanac-type questions for which numerical answers were available. The panel members provided both their best estimate of the answer as well as relative ratios with respect to other quantitative responses. In this way the panel members got an appreciation of the benefits of the anchoring process used throughout the elicitation process.

C. 1 Training Questions

The following questions were used in the training exercise.

Q1. According to the 2000 census, how many men 65 or over were there in the U. S.?

Q2. In 1995, how many American men age 65 or older suffered from the chronic conditions listed?

Q3. What is the ratio of the rate for men 45- 64 years old to the rate for men 65 and older for each of the conditions listed?

Q4. What is the ratio of the rate for men under 45 years old to the rate for men 45 — 64 years old for each of the conditions listed?

The answer to Q1 is 14.4 million. The chronic conditions referred to in Q2, Q3, and Q4 and the corresponding answers are listed in Table C.1.

Table C.1 Correct Value (CV) Results to Elicitation Training Questions

Dr. Gott studied metallurgy and materials science at Imperial College, London.

During the more than 20 years she has worked in Studsvik Dr. Gott studied many aspects of the environmental effects on structural materials in nuclear power plants, both through contract research projects and failure analysis. She has held a number of different types of position whilst at Studsvik including project manager, marketing manager and manger of the reactor chemistry group. She was also on periodic loan to a US subsidiary in Richland, WA, to help them establish laboratory support for their decontamination services.

The main areas of her research activities were

• Creep crack formation in stainless steels (mechanical testing, electron and light optical metallography)

• Fracture mechanics (corrosion fatigue, residual stress measurement, non-destructive testing)

• Reactor chemistry (PWR and BWR chemistry, activity build-up including field measurements, decontamination)

• Reactor materials (surveillance testing, failure analysis, metallography of Inconel 182)

In her current position at the Swedish Nuclear Power Inspectorate she has continued to work in the field of environmental degradation of nuclear power plant structural materials. The work covers both the regulatory and the research aspects. On the regulatory side she is involved in the development of regulations, inspection and safety evaluations that form the basis for decisions based on Swedish law and regulations. One of her responsibilities includes the management of the materials and chemistry research area for the Inspectorate. In addition she has built a database covering operationally induced failures and damage to mechanical components in the Swedish nuclear fleet and is responsible for its maintenance and the associated analysis of failure cases. In 2003 she was on a six month job rotation to the Materials Engineering Branch of the NRC’s Office of Nuclear Reactor Regulation working amongst other things on PWSCC problems.

She is a member of the international conference committee which arranges the regular water chemistry conferences in the nuclear field, and has also acted on the international committee for the Fontevraud conference in France. She served as chairperson of the steering committees of two large international projects concerning irradiation assisted SCC and the establishment of a pipe failure database.

Each base and reference case includes at least one transient since transients are needed to initiate a LOCA event. At this time the transients are poorly defined.

Fred Simonen has seen reports that show seismic stresses, but he has no idea of the magnitude of the non­seismic transients (e. g., water hammer, safety relief valve transients). He and Gery Wilkowski would like help in establishing a rough order of magnitude for these types of transients. This information could be best expressed as a percentage of the Service Level stresses. Pete Riccardella indicated that SRV transients could be on the order of a small earthquake, just with a higher frequency. Bruce Bishop agreed to provide some water-hammer transient stresses for the pressurizer. Gery Wilkowski volunteered to provide some summary information from past probabilistic LBB analyses

Bengt Lydell indicated that the water hammer frequency is about 5E-3. There are more water hammer events on the secondary side, but there are design basis events that can cause water hammer on the primary side. It as noted again that without a transient a large LOCA is highly unlikely. The cracks will just leak until they are detected, and then will be repaired. Long surface cracks that don’t leak are drivers for large LOCAs.

Guy Deboo (Consolidated Edison) volunteered to provide stresses and frequencies for transients (e. g., feedwater line water hammer, SRV, and seismic from the LaSalle plant). Gery Wilkowski to provide some tables from NUREG/CR-6004 showing the N+SSE stresses for about 30 piping systems. This data was originally developed for the ASME Section XI Working Group on Pipe Flaw Evaluation. Additionally, Gery Wilkowski and Guy Deboo will provide stresses (not frequency) for some large faulted loads that are not really expected to occur over the life of plant. Gery is to examine the N+SSE stresses from the USNRC LBB submittal database. Guy noted that a 1SSE amplitude earthquake, based on seismic hazard curves, is expected to occur once over 40 years (design basis). The frequency (not amplitudes) of the seismic hazard curves are generally considered to be conservative. Gery noted that the design basis apparently is conservative since he is sure that an SSE event has ever occurred at any US (or other) plant. Hence, the seismic event frequency could perhaps be down graded to 0.5/2,600 events/year rather than 1/40 events per year.

For the smaller transients, Dave Harris will extract stresses from a NUREG report by Fred Simonen. Guy Deboo, Pete Riccardella, and Sam Ranganath will provide some data on normal operating transients. Pete Riccardella will get some data showing comparison of design versus actual transients based on some thermal fatigue analysis from a Sandia report.

Bruce Bishop has some plant specific ISI data that he could provide which provides transient information, but he won’t be able to provide it expeditiously due to other commitments. In fact, it is unlikely he will be able to provide this during the timeframe of this effort. Pete Riccardella concluded that the actual transients were not as severe as the design basis transients, but there are typically more of them.

It was also agreed that the group should have isometric drawings for the LOCA-sensitive piping. Bengt Lydell will inventory his electronic drawing database. The purpose is to look at generic systems to get an idea of how many welds are involved, pipe sizes, etc. Guy DeBoo will also evaluate the ISO drawings in his archive. Bengt will coordinate with Guy Deboo on this effort. For the time being it was decided that we would not seek additional isometric drawings until we determine if we are missing any of the major piping systems. Guy Deboo will then help obtain drawings (e. g. ISI drawings) for missing systems that at least indicate the number of welds. In general, multiple isometrics of similar systems are useful since each plant design is somewhat unique.

Dave Harris thought we only needed census data, i. e., number of welds as a function of pipe sizes, etc. Dave will provide a census that he has developed of number of welds for the base case piping systems. Dave and Bengt Lydell will coordinate on this action. Gery Wilkowski indicated that there is a MRP report with locations and numbers of bimetal welds that may be useful to review. Gery will provide a table of Inconel weld locations in different piping systems from the MRP-44 report.

Rob Tregoning asked if the panel should consider redefining the base and reference cases. The biggest concern is that the loadings and the mitigation/maintenance may need to more accurately reflect the operating experience. The panel could redefine these variables in a way that is more consistent with the quantitative analyses that was presented for the base cases earlier. The approach is to define these base and reference cases load and mitigation variables to reflect historical plant operating experience for the first 25 years of plant life (e. g. the current LOCA estimate). The consensus opinion agreed with this proposal. Guy Deboo added that he would like to see the base cases run out to 40 and 60 years, with a seismic event included. It was again noted that the objective of the “current estimate” analyses (i. e., out to 25 years of plant life) was to provide a benchmark against historical data.

Rob also asked if the panel wanted to consider more than one degradation mechanism for each reference case since the operating experience contains contributions from all applicable degradation mechanisms. Rob Tregoning wants to make sure each panel member is making the same relative comparisons with the reference cases and that these relative comparisons are natural. However, the group consensus was that it is easier to use the reference cases for anchoring when only considering one degradation mechanism and that no other changes in the reference cases should be adopted.

Vic Chapman argued that if we run the probabilistic fracture models for the first 25 years for benchmarking purposes, and we see a failure, then we should exclude that result since in reality we have not seen any failures to date. Lee Abramson agreed with this thought. The operating experience-based estimates developed by Bill Galyean and Bengt Lydell are inherently benchmarked in this manner. The probabilistic fracture models (David Harris and Vic Chapman) still need to benchmark their data.

Bruce Bishop would like Karen Gott, Bill Galyean, and Bengt Lydell to extract the failure mechanisms as a function of piping system from their databases. Bill will have one of his colleagues do this. Bengt and Karen will query their databases to get a list of degradation mechanisms as a function of piping system. Bruce Bishop would like someone to publish a list of plants by design type. Rob Tregoning indicated that he would provide this information.

Several people wanted Dave and Vic to run their models using refined stress histories. Pete Riccardella agreed to redefine the loads for the HPI/MU nozzle. Gery Wilkowski volunteered to get some more realistic surge line stresses for the surge line elbow case. It should actually be for a crack in the girth weld at the elbow, not a crack in the body of the elbow.

Bill Galyean and Bengt Lydell will determine the frequency of IGSCC leaks and surface cracks as a function of time and pipe size using their respective databases. The surface crack data should be characterized by the length and depth of the flaws if possible. The Swedish data should be separately characterized from the US plant data, since the US plants may not have characterized the flaw shapes by ISI if the piping was being removed from service. In Sweden, flaws from removed pipe systems have been characterized. A further requirement of this query to examine the recirculation system piping studied in the base case would also be helpful. Bengt will provide all this information in an Excel spreadsheet.

Two types of sensitivity analysis are included in this report. The first type addresses the impact on results by an assumed incompleteness of the failure data collection. The second type relates to the sensitivity of the time-dependent LOCA frequencies to different assumptions about leak detection and in-service inspection. The sensitivity analysis results are included in Section D.6.

D.3 Service Experience Data Application to the Base Case Study

The PIPExp database documents service experience with Code Class 1, 2 and 3 and non-safety related (or Class 4) piping in commercial nuclear power plants worldwide. For the time period 1970-2002, this database was queried for service experience data specific to the Base Case piping systems. The results of the database queries are summarized here, and they form the input to the data processing and failure parameter estimation in Sections D.4 and D.5.

D.3.1 PIPExp Database, Revision 2003.1

The pipe failure database utilized in the Base Case Study is called PIPExp. It is an ACCESS database and an extension of the OPDE database [D. 12-D. 13]. Since the conclusion of the original work in 1998 [D.11,

D.17], the pipe failure database has been significantly expanded both in terms of the absolute number of event records and the depth of the database structure (Appendix A provides additional details). Lessons learned through database applications have been used to enhance the structure. In this study of HPI//NMU-, FW-, RC — and RR-piping reliability the statistical analysis is based on service data as recorded in PIPExp and with cutoff date of December 31, 2002. The analysis is inclusive of applicable worldwide BWR — and PWR- specific service experience with Code Class 1 piping. As of 12-31-2002 the database accounted for approximately 1,992 and 3,621 critical reactor-years of operating experience with commercial BWR and PWR plants, respectively.

The database is actively maintained and periodically updated. The effort involved in populating the database while at the same time assuring data quality is not trivial. As an example, changing regulatory reporting thresholds imply that an ever increasing volume of raw data reside in restricted and proprietary database systems rather than in the public domain. For an event to be considered for inclusion in the database it undergoes screening for eligibility. For example:

• The equipment failure must be positively identified as a piping component failure external to the reactor pressure vessel (RPV). A failure involves a pressure boundary degradation, which can be non-through-wall (crack with a/t-ratio > 10%, where a = crack depth and t = wall thickness) or a through-wall leak.

• There must exist documented evidence in the form of a hard copy (e. g., USNRC Inspection Report, Licensee Event Report, ISI Summary Report, Problem Identification Form, Condition Report,

ASME Code Repair Relief Request, etc.) from which a sufficiently detailed case history is developed. The documented evidence of pipe degradation/failure must contain information on its location within a piping system (e. g., with reference to an isometric drawing and/or P&ID), metallurgy, operating conditions, impact on operation, method of discovery, failure history, etc. so that a data classification may be independently verified.

• Where the documented evidence is deemed incomplete, additional information is solicited through direct contact with plant personnel or by accessing supplemental data.

• There must be sufficient technical information available to fully address the complex relationships between piping reliability attributes (or design parameters) and influence factors (e. g., fabrication/welding techniques, environmental conditions such as water chemistry, flow conditions) on the one hand and degradation/failure mechanisms on the other.

• Differentiation between UT indications versus confirmed crack indications. Only the latter are included in the database given an a/t-ratio > 10%.

Following on the initial data screening, each event selected for inclusion in the database is subjected to a classification so that the unique reliability attributes and influence factors are identified. Including memo fields, text fields, numerical fields and data filters, up to 114 database fields describe each record of the database.

Lee Abramson of the US NRC spoke on the expert elicitation process that will be followed in this exercise. Some of the specific key points from his presentation and subsequent discussion are outlined below:

• Key word is “formal” use of expert judgment. Engineers practice informal expert judgment every day.

• It was emphasized that elicitation is a structured process and that the process requires experienced practitioners to conduct the exercise. This is not a “do it yourself” activity.

Discussion: A question was raised if the results of this elicitation or past elicitations could be used as a baseline for future efforts, in much the same way that Bayesian analysis is performed. Lee Abramson indicated that there is no natural means of updating results from prior elicitations based on recent experience or new data. However, it may be appropriate to use the results of a prior elicitation as starting point for future elicitation.

• The need for comprehensive documentation was also stressed to ensure that the process approach, issues, analysis techniques, results and uncertainties are clear. Additionally, follow-on work to refine the results requires comprehensive documentation in order to understand the basis of the initial study.

• The need for an expert panel with a broad range of expertise and experiences was expressed.

Also all of the stakeholders (both utilities and regulators) must be represented.

• There are two methods of elicitation: group and individual. The problem with group sessions (versus individual sessions) is that often group dynamics lead to domination of one or two individual opinions. The results then no longer represent everyone’s input.

• Elicitation team for this exercise consists of

• Normative expert — Lee Abramson

• Substantive experts — Alan Kuritzky, Ken Jaquay, Rob Tregoning, others?

• Recorder — Paul Scott

• Documenter — Paul Scott (could be same as recorder)

• Panel members need to provide rationale for answers so others can see why certain panelists came up with certain answers. In that way other panelists have the option of changing their answers based on feedback from the group. The panel will largely be provided this feedback at the wrap-up meeting. Panel members can revise answers to any question at any time.

Discussion: It was asked if the response will be weighted in any way to account for expertise in a given area. Lee Abramson replied that the analysis will use unweighted responses so that everyone’s response is judged equally. With this size of panel, weighting should not substantially affect the final results. The elicitation will also query the panel member’s uncertainty for each answer. If inordinate uncertainty exists, then the response may be downgraded. Also, the rationale provided by each panelist will help determine if responses need to be weighted.

• Types of biases present in elicitation processes:

• Motivational biases (i. e., social pressure or group pressure to make a certain decision). These need to be recognized and avoided at all costs.

• Cognitive biases — biases can occur when people have developed an initial answer and more data becomes available which require the initial answer to be modified. Typically people underestimate the impact of the new data. This bias is referred to as anchoring. The elicitation structure will be developed in an attempt to minimize these biases. For instance, initial estimates of the total LOCA frequencies will not be asked.

• Background biases (i. e., what an individual might see as reasonable, or would expect, based on his background.) For example, an experimentalist might see a high probability of failure of a piping based on the number of experiments he has run in which he saw a failure, but typically the test conditions were such that similar conditions in the field are highly unlikely to ever occur. This bias is natural, but it is important to get each individual to consider all variables which affect the result and break them down into meaningful pieces.

• People are more than likely to underestimate the true uncertainty, by a factor of 1/2.

• People are more likely to anchor on median value, not on the extremes.

• Goal is to make the questions as unambiguous as possible (very precise) and to focus questions on the major issues affecting the LOCA analysis.

• The uncertainty range will be queried during the elicitation by asking for the “number” such that there is 5% chance that the true response is less than this number. A separate number will be provided for the UB such that there is also a 5% change that the true response is higher than this number. This corresponds to the 90% coverage interval of the variable.

• Purpose of elicitation panel members is to come up with individual answers, not a consensus.

Discussion: There was quite a bit of discussion and confusion about the definition of the coverage interval. Lee Abramson said that the uncertainty range (difference between higher and lower response to a given question) should cover the true number for that variable 90% of the time. The true value should fall below the lower response 5% of the time and the true value should land above the higher response 5% of the time. However, Lee cautioned against making the coverage interval inordinately large just to capture uncertainty. If this occurs, the coverage interval contains little useable information.

B-6

Rob started the second day at 0810 by reviewing agenda for the next two days.

Presentation #10 — Non-Piping Database Development By Bill Galyean

LER database at ORNL will no longer be available after 2/29/04; the database will be moving to INEEL but in a different format.

Failures defined as leaks or cracks.

The final portion of the meeting concentrated on developing issues associated with non-piping contributions (passive failures only) to the LOCA frequencies. Active components will be analyzed separately during this program from operating experience. Because active components have maintenance

plans, the group in general expects that the failure rate of these components will not increase in the future. Operating experience should therefore adequately represent active component failure rates. This rationale is the basis for considering only passive component failures within this elicitation.

The non-piping contributions will be combined with the piping component to determine the total LOCA frequency (Figure B.1.1). The group decided to break these issues down by component functionality.

This is an analogous approach used to tackle the piping contribution which initially segregated piping systems by functionality. Five main components were determined as candidates for passive failures: the pressurizer, the RPV, valves, pumps, and steam generators/steam systems. The valve component category encompasses both pressure isolation valves at Class 1 to Class 2 piping boundaries and also loop-stop valves. The pumps category only considers pumps in the reactor coolant or recirculating water system.

For each component category, the panel developed sub-categories which represent specific possible failure modes (e. g. what portions of the component could fail passively). Each failure mode is governed by the same variables that are important for piping systems (material, geometry, loading, degradation mechanisms, and mitigation/maintenance). Unfortunately the group did not have sufficient time or resources to fully develop comprehensive variable lists in the same manner as for piping systems.

Table B.1.13 illustrates the failure modes developed for pressurizer failures. Please note that all table abbreviations for this and all subsequent tables in this section are as previously defined unless indicated. Bold items in the failure mode sub-category of Table B.1.13 (and all following tables) indicates that operational data exists which captures that component failure. For instance, in Table B.1.13, the group thinks that data is available on heater sleeve failures.

Table B.1.13 Pressurizer Failure Scenarios

The panel identified failures in the pressurizer shell, manway, heater sleeves and nozzles as passive LOCA candidates. Also, the pressurizer bolted relief valves could fail. The group generally did not have information on component geometries and loading and mitigation/maintenance were not discussed by the panel due to lack of time. However, some specific issues were discussed for each of these failure modes. The shell failure envisioned would most likely occur by boric acid wastage from the outer diameter of the

shell. Manway failures would result by multiple bolt failures. Heater sleeves fail due to PWSCC, but as a result of their size, multiple failures are required in order to result in a LOCA. Bolted relief valves could fail due to steam cutting or localized bolt corrosion resulting from boric acid leaks.

The RPV failure modes (Table B.1.14) focused on vessel head bolt failure, failure of CRDM connections, nozzle failure, RPV wastage, and RPV corrosion fatigue. Upper head vessel head bolt failure is most likely due to human error during removal at each refueling cycle. Human error could occur as a result of improper installation procedures. Problems, however, could be identified during prestart-up inspection. The lower head bolts are not removed during refueling and they could be susceptible to common cause failure resulting from local bolt corrosion leading to several simultaneous bolt failures. A certain percentage of these bolts are inspected at each outage and the assumption is that inspection would not be effective in identifying the degradation prior to failure. These requirements may uncover the likelihood of common cause errors leading to some latent failure that is not immediately evident and shed light on other possible failure mechanisms. An example of a common cause failure is a torque wrench/tensioner which is out of calibration so that all bolts are improperly installed and then can possibly fail during operation.

CRDM connections far outside of the reactor could be welded, bolted, or threaded and seam welded. The degradation mechanism would be a function of the specific connection. For instance, welded connections would be susceptible to the mechanisms and loading discussed previously for CRDM components.

Bolted CRDM connections would be subject to steam cutting, boric acid corrosion, aging and other degradation mechanisms that are unique to bolts. It must be stressed that the CRDM connections in this table refers to the CRDM which connects to the drive mechanism. Inboard connections are considered to be part of the “CRDM piping system” discussed earlier. For bolted connections, this demarcation line is the flange joint. The group identified failure data for CRDM leakage from bolted flanged connections.

There are two RPV degradation mechanisms which were specifically discussed. The first was degradation of the shell, upper head, or lower head due to boric acid corrosion. The second mechanism was corrosion fatigue developed at through-thickness cladding cracks in the shell, upper head, or lower head. The nozzle category is subject to similar degradation mechanisms as in the attached piping. It should be stressed that the nozzle category only considers the flared portion of the nozzle up to the reactor shelf. The nozzle safe end was earlier defined as part of the piping system. The group identified that some data on nozzle issues exists.

Valve failure modes are summarized in Table B.1.15. The cast stainless steel valve bodies are susceptible to an array of potential degradation mechanisms. These include cavitation (CAV), thermal fatigue (TF), and material aging (MA). Casting defects (CD) are another particular concern. Failure due to the other

mechanisms listed could initiate at either the defects, or repairs of those defects. The main steam isolation valve (MSIV) body is associated with similar failure modes. Specific failure modes for valve bonnets, and valve bonnet bolts were not discussed. Presumably, bonnet bolt failures would be susceptible to the same failure mechanism of other bolts: aging, boric acid corrosion, steam cutting, etc. The hot leg/cold leg loop isolation valve failure modes were also not discussed. However, failures in these valves could be described in terms of the bonnet, body, or bonnet bolt failure sub-categories listed earlier. It should also be noted that valve sizes are generally consistent with the piping system where they are located.

Steam generator tube rupture (Table B.1.16) can occur from a variety of different mechanisms including thermal fatigue, mechanical fatigue, SCC, and general corrosion. The tubes can also be degraded by mechanical deformation (MECDEF), or denting, during installation, inspection, or cleaning. Steam generator tubes are too small to lead to a LOCA due to a single tube failure. Therefore, multiple tube rupture needs to also be considered in order to achieve a certain size LOCA. There is data which exists for SGTR.

Steam generator failure can also occur at the manway (specifically bolt failure), the steam generator shell, or the nozzles. These various failure modes were also not sufficiently discussed so little information has been defined in Table B.1.16. However, the nozzle failure issues will likely be similar to the associated piping system, while manway bolt failure would be caused by the same types of mechanisms as for other bolt failures.

The pump failure modes (Table B.1.17) are similar to many of the failure modes already discussed for other components. The cast pump bodies are potentially subject to the same degradation mechanisms (CAV, TF, CD, MA) as other cast components. The recirculation (RECIRC) bonnet bolts and RCP nozzle are also susceptible to mechanisms discussed earlier. The only unique mode considers an incipient failure of a pump flywheel which could initiate collateral damage in other components or in other piping systems. There was no appropriate passive pump failure data that was identified by the group.

It is obvious that the non-piping passive LOCA sources have not been nearly as well-defined as the piping system sources, and they must be better defined prior the elicitations. However, due to the number and complexity of the components, the panel realized that it may not be possible to fully define all the variables listed in the tables above. At a minimum, the group decided that it would need isometric drawings for as many of these components as possible.

The manner for arriving at the LOCA contributions of these other components will be similar to the approach followed for the piping contribution. Reference cases will be developed and absolute LOCA estimates will be assigned to those numbers. However, these reference cases will be based strictly on data. The bolded items in Tables B.1.13 — B.1.16 are component failures that are supported by passive — system failure data. This data will first need to be accumulated and analyzed. Karen Gott and Bill

Galyean are possible sources for some of this data. There is an EPRI database called PM-BASIS which consists of mainly active components, but there may be some data on bonnets and packings. Also, Spence Bush may have some data for these components that might be useful to the group. Finally, the group discussed that there may be data available for feed water nozzles.

Once the data is developed, it will be made available to the group. This data will make up the base case information for the non-piping components. The group will also be asked how representative the base case data is for future (end-of-plant-license-renewal) LOCA estimates. The LOCA propensity (for each leak threshold rate) for the components without data will also be queried relative to these base cases. This approach is identical to the development of the piping LOCA contributions discussed earlier.

As described in Section 3.3.2, the panelists were asked to supply three numbers for each question: a MV, a LB, and a UB. The MV has a nominal 50/50 chance of falling above or below the correct value. The interval (LB, UB) has a nominal 90% chance of covering the correct value.

The following tables summarize the responses made in the training exercise. There were between 15 and 17 sets of responses to each question. (Although there were only 12 panelists on the panel, members of the facilitation team were also invited to participate.) The number of respondents is indicated following each question. The table columns summarize the responses relative to the CV. The first column indicates the number of respondents where CV < LB, i. e., where the coverage interval fell above the CV; the third column indicates the number of respondents where CV > UB, i. e., where the coverage interval fell below the CV. Thus, the total of the first and third columns is the number of respondents whose coverage intervals did not cover the CV. The second column lists three numbers that summarize the set of MVs provided for each row of the table. These are the lower quartile (LQ), median and upper quartile (UQ), respectively. About one quarter of the MVs are less than the LQ and about one quarter of the MVs are greater than the UQ. Hence the interquartile interval (LQ, UQ), denoted by IQI, contains about one half of the MVs. (These three summary statistics are used to construct box and whisker plots, as described in Appendix L.) For ease of reference, the rounded correct values are listed following the conditions for the Q2 — Q4 tables.

Q1. According to the 2000 census, how many men 65 or over were there in the U. S.? (N = 17)

(CV = 14.4 million)

Respondents tended to overestimate the CV. Since LQ = 16, about three quarters of the MVs were larger than the CV. However, percent coverage at 82% was near the nominal 90%, with 3 (18%) lying above the CV and none lying below.

Q2. How many American men age 65 or older suffered from the following chronic conditions in 1995? (N = 15)

Table C.3 Summary of Respondent Results for Question Q2

Four of the six IQIs covered the CV, and the two which did not almost did. Three of the medians were above the CV and three were below. Thus, the MVs for the six conditions as a whole exhibited no systematic bias in estimating the CVs. However, the coverage intervals tended to underestimate the CVs. Of the 90 coverage intervals, 27 (30%) lay below the CV and 8 (9%) lay above. The average percent coverage of all 90 intervals was 61%. Over the six conditions, the percent coverage ranged from a low of 33% to a high of 80%.

Q3. What is the ratio of the rate for men 45- 64 years old to the rate for men 65 and older for each of the conditions listed? (N = 16)

Table C.4 Summary of Respondent Results for Question Q3

Five of the six IQIs covered the CV, but respondents tended to underestimate the CV. Five of the six medians were below the CV. Of the 96 coverage intervals, 22 (23%) lay below the CV and 5 (5%) lay above. The average percent coverage of all 96 intervals was 72%. Over the six conditions, the percent coverage ranged between 62% and 88%.

Q4. What is the ratio of the rate for men under 45 years old to the rate for men 45 — 64 years old for each of the conditions listed? (N = 16)

Table C.5 Summary of Respondent Results for Question Q4

Five of the six IQIs covered the CV, but respondents tended to overestimate the CV. Four of the six medians were above the CV, and two were equal or almost equal to the CV. Of the 96 coverage intervals, 20 (21%) lay above the CV and 8 (8%) lay below. The average percent coverage of all 96 intervals was 71%. Over the six conditions, the percent coverage ranged between 50% and 88%.

C. 3 Discussion

The results of the training exercise were consistent with several of the basic premises underlying the elicitation structure and methodology. First, apart from Q1, the responses to the other three questions as a whole did not exhibit any systematic over — or under — estimation bias. Q2 had no systematic bias, Q3 tended to underestimate, and Q4 tended to overestimate the CVs. This result is consistent with the basic premise of the elicitation process, which is that the panel responses as a whole have no systematic bias (see Section 3.3).

Second, the percent coverage of the (LB, UB) intervals were less than the nominal 90% for all four questions. Q1 had the highest percent coverage at 82%, perhaps because the question dealt with demographic data with which the respondents were relatively more familiar. Q3 and Q4 had the next highest percent coverage at about 71% each and Q2 had the lowest percent coverage at 61%. This result is consistent with the rationale for the overconfidence adjustments made to the panelists’ uncertainty intervals (see Section 5.6.2).

Third, the two questions (Q3 and Q4) that asked about ratios of rates had higher percent coverage than the question (Q2) that asked about absolute rates. This result is consistent with the rationale for the basic structure of the elicitation questions, which ask about relative rather than absolute LOCA frequencies (see Section 3.8).

Dr. Harris is a principal engineer at Engineering Mechanics Technology (EMT), Inc. and has some 30 years of experience in fracture mechanics and solid mechanics analysis and applications. His background is in mechanical engineering, and he has extensive experience in probabilistic structural mechanics, especially as related to fracture mechanics.

Dr. Harris began his career as a mechanical engineer at Lawrence Radiation Laboratory (LRL) in Livermore, California. After several years at LRL, Dr. Harris joined one of the earliest vendors of acoustic emission instrumentation, Dunegan Corporation, as Director of Research. After four years at Dunegan Corporation, Dr. Harris joined Science Application, Inc. (SAI, now known as SAIC) in their Palo Alto office. During his seven years at SAI, Dr. Harris’ efforts included performing some of the earliest applications of PFM to nuclear reactor piping. He was the principal developer of the PRAISE code, which was developed for the USNRC. The PRAISE code is based on PFM and is one of the most widely applied tools for evaluation of the reliability of weldments in nuclear reactor piping.

Dr. Harris worked at Failure Analysis Associates for over ten years. During this time he developed and applied fracture mechanics to a wide variety of problems, ranging from railroad wheels to rocket ship engines. These efforts included both deterministic and probabilistic aspects, and involved both computer software development and applications to industrial problems. He was the manager of the Fracture Mechanics section, which included some five engineers involved in fracture mechanics and related finite element stress analysis. He was the principal developer of the NASCRAC code, which is a general purpose code for deterministic analysis of crack growth that was developed for NASA.

Dr. Harris is currently a vice-president and principal engineer at EMT a company that he was involved in founding some seven years ago. EMT is an engineering consulting firm that specializes in fracture mechanics, life prediction and related software — both deterministic and probabilistic. Efforts at EMT include development of the PRAISE code in Windows (WinPRAISE), including enhancements to make the software easier to use in routine applications, and expansion of PRAISE to include crack initiation due to cyclic loading in air and water environments. He was also involved in the development of commercial fracture mechanics software — including linear and nonlinear SmartCrack. BLESS is a code for analysis of reliability of headers and piping in fossil-fired power plants that was developed with support of the Electric Power Research Institute (EPRI). BLESS is a physics-based model that considers both crack initiation and growth due to creep and cyclic loading.

Dr. Harris has been involved in ASME activities related to reliability considerations in design and inspection of nuclear reactor piping. He was an original member of the ASME Research Task Force on Risk-Based Inspection Guidelines, and was the editor of Volume 3 of a series of reports published by this committee. Volume 3 was on applications to fossil fired power plants. He is currently vice chairman of the Risk Technology Committee of the ASME. Dr. Harris is a member of ASTM as well as ASME. He has nearly 100 publications in the open literature, primarily in the areas of acoustic emission and fracture mechanics. He received a B. S. and M. S. in mechanical engineering from the University of Washington and a Ph. D. in applied mechanics from Stanford University.

Rob started by reviewing the history behind this analysis. Rob was not able to find a reference for the basis of PWR correlations developed in NUREG-1150. The BWR correlations were extracted from GE NEDO studies performed in the early 1980’s. New correlations have been based on several closed-form solutions appropriate for BWR steam and liquid lines and PWRs.

Rob still needs to add a column correlating the pipe diameter to the leak rate/area in slides 8 of 9 of this presentation. It was agreed that the pipe diameter should assume a single ended guillotine break. Rob also agreed to change the word “axial” to “transverse” on slide 9 of the original presentation which relates the pipe fracture area to transverse piping displacement. Gery Wilkowski noted that once a full DEGB occurs that the two ends of the pipe are jets that move away from each other so that he would be hesitant to use the analysis on slide 9 which assumes a transverse displacement. The base case team members will update their results using the new pipe break size to leak rate correlations developed for this presentation.