ML18306A779: Difference between revisions
StriderTol (talk | contribs) (Created page by program invented by StriderTol) |
StriderTol (talk | contribs) (Created page by program invented by StriderTol) |
||
(5 intermediate revisions by the same user not shown) | |||
Line 3: | Line 3: | ||
| issue date = 10/03/2018 | | issue date = 10/03/2018 | ||
| title = Transcript of Advisory Committee on Reactor Safeguards Structural Analysis Subcommittee Meeting - October 3, 2018, Pages 1-134 | | title = Transcript of Advisory Committee on Reactor Safeguards Structural Analysis Subcommittee Meeting - October 3, 2018, Pages 1-134 | ||
| author name = Shukla G | | author name = Shukla G | ||
| author affiliation = NRC/ACRS | | author affiliation = NRC/ACRS | ||
| addressee name = | | addressee name = | ||
Line 9: | Line 9: | ||
| docket = | | docket = | ||
| license number = | | license number = | ||
| contact person = Shukla G | | contact person = Shukla G | ||
| case reference number = NRC-3927 | | case reference number = NRC-3927 | ||
| document type = Meeting Transcript | | document type = Meeting Transcript | ||
Line 16: | Line 16: | ||
=Text= | =Text= | ||
{{#Wiki_filter:Official Transcript of Proceedings NUCLEAR REGULATORY COMMISSION | {{#Wiki_filter:Official Transcript of Proceedings NUCLEAR REGULATORY COMMISSION | ||
Date:Wednesday, October 3, | ==Title:== | ||
Advisory Committee on Reactor Safeguards Structural Analysis Subcommittee Docket Number: (n/a) | |||
Location: Rockville, Maryland Date: Wednesday, October 3, 3018 Work Order No.: NRC-3927 Pages 1-134 NEAL R. GROSS AND CO., INC. | |||
Court Reporters and Transcribers 1323 Rhode Island Avenue, N.W. | Court Reporters and Transcribers 1323 Rhode Island Avenue, N.W. | ||
Washington, D.C. 20005 (202) 234-4433 | |||
1 1 | |||
2 3 | |||
4 DISCLAIMER 5 | |||
6 7 UNITED STATES NUCLEAR REGULATORY COMMISSIONS 8 ADVISORY COMMITTEE ON REACTOR SAFEGUARDS 9 | |||
10 11 The contents of this transcript of the 12 proceeding of the United States Nuclear Regulatory 13 Commission Advisory Committee on Reactor Safeguards, 14 as reported herein, is a record of the discussions 15 recorded at the meeting. | |||
16 17 This transcript has not been reviewed, 18 corrected, and edited, and it may contain 19 inaccuracies. | |||
20 21 22 23 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W. | |||
(202) 234-4433 WASHINGTON, D.C. 20005-3701 www.nealrgross.com | |||
1 1 UNITED STATES OF AMERICA 2 NUCLEAR REGULATORY COMMISSION 3 + + + + + | |||
4 ADVISORY COMMITTEE ON REACTOR SAFEGUARDS 5 (ACRS) 6 + + + + + | |||
7 STRUCTURAL ANALYSIS SUBCOMMITTEE 8 + + + + + | |||
9 WEDNESDAY 10 OCTOBER 3, 2018 11 + + + + + | |||
12 ROCKVILLE, MARYLAND 13 + + + + + | |||
14 The Subcommittee met at the Nuclear 15 Regulatory Commission, Two White Flint North, Room 16 T2B1, 11545 Rockville Pike, at 1:00 p.m., Peter 17 Riccardella, Chairman, presiding. | |||
18 19 20 21 22 23 24 25 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W. | |||
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433 | |||
2 1 COMMITTEE MEMBERS: | |||
2 PETER RICCARDELLA, Chairman 3 MICHAEL L. CORRADINI, Member 4 RONALD G. BALLINGER, Member 5 MARGARET SZE-TAI Y. CHU, Member 6 VESNA B. DIMITRIJEVIC, Member 7 JOSE MARCH-LEUBA, Member 8 HAROLD B. RAY , Member 9 JOY L. REMPE, Member 10 GORDON R. SKILLMAN, Member 11 MATTHEW SUNSERI, Member 12 13 DESIGNATED FEDERAL OFFICIAL: | |||
14 GIRIJA SHUKLA 15 16 ALSO PRESENT: | |||
17 JOSE PIRES, RES 18 MEHDI REISI-FAHRD, NRR 19 20 21 22 23 24 25 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W. | |||
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433 | |||
3 1 C-O-N-T-E-N-T-S 2 PAGE 3 Opening Remarks and Objectives 4 Peter (Pete) Riccardella, ACRS . . . . . . 4 5 Presentation and Discussion . . . . . . . . . . . 10 6 Correlation or Dependency of Seismic 7 Performance of Similar SSCs by Jose 8 Pires (RES) and Mehdi Reisi-Fard (NRR. | |||
9 Closing Comments 10 Pete Riccardella, ACRS . . . . . . . . . 104 11 Adjourn . . . . . . . . . . . . . . . . . . . . 104 12 13 14 15 16 17 18 19 20 21 22 23 24 25 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W. | |||
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433 | |||
4 1 P-R-O-C-E-E-D-I-N-G-S 2 (12:58 p.m.) | |||
3 CHAIRMAN RICCARDELLA: The meeting will 4 now come to order. This is a meeting of the ACRS 5 Structural Analysis Subcommittee. I am Pete 6 Riccardella, Chairman of the Subcommittee. ACRS 7 members in attendance are, Vesna B. Dimitrijevic, 8 Margaret Sze, Harold Ray, Gordon Skillman, Mike 9 Corradini, Matt Sunseri, Joy Rempe, Jose March-Leuba, 10 and Ron Ballinger. We may or may not have two other 11 members come in later. | |||
12 Girija Shukla of the ACRS is the 13 Designated Government Official for this meeting. | |||
14 The ACRS reviews and advises the 15 Commission in regard to licensing and operation of 16 production and utilization facilities and related 17 safety issues, the adequacy of proposed reactor safety 18 standards, technical, and policy issues related to the 19 licensing of evolutionary and passive plant designs, 20 and other matters referred to it by the Commission. | |||
21 The purpose of this Subcommittee meeting 22 is for the staff to brief Subcommittee on the topic of 23 Correlation or Dependency of Seismic Performance of 24 Similar SSCs. This is the topic of a research 25 publication, NUREG/CR-7237, Correlation of Seismic NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W. | |||
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433 | |||
5 1 Performance of Similar SSCs. | |||
2 Members of the -- this is also the topic 3 of a research review, and I want to note that this 4 briefing that we're having here is independent of the 5 research quality review. That quality review is well 6 under way and the panel | |||
===Background=== | ===Background=== | ||
*Seismic Probabilistic Risk Assessments (SPRAs) have been conducted worldwide for over 35 years | * Seismic Probabilistic Risk Assessments (SPRAs) have been conducted worldwide for over 35 years | ||
*Two key SPRA aspects I.All possible earthquake levels, their frequencies of occurrence and potential consequential damage should be considered II.Earthquakes can simultaneously damage multiple redundant components | * Two key SPRA aspects I. All possible earthquake levels, their frequencies of occurrence and potential consequential damage should be considered II. Earthquakes can simultaneously damage multiple redundant components | ||
*Risk-quantification should account for this | * Risk-quantification should account for this common cause effect 3 | ||
=== | |||
Background=== | |||
4 1994 Northridge earthquake Haiti | |||
=== | |||
Background=== | |||
Single Motor Control Center Failure at Piti Power Plant - | |||
Guam (1993 Earthquake) 5 | |||
=== | |||
Background=== | |||
Fukushima Daini Make Up Tank Failures - | |||
Dependent Failures 6 | |||
===Background=== | ===Background=== | ||
* Treatment of dependent failures involves a set of assumptions (rules-based approach) such as: | |||
- Similar redundant components next to each other and subjected to similar seismic forces fail simultaneously | |||
- Diverse or similar components at different (far apart) locations fail independently | |||
- Sensitivity analysis to postulated bounding assumptions | |||
* This approach can mask insights and lead to varying degrees of conservatism in various places of the analysis 7 | |||
===Background=== | ===Background=== | ||
* The treatment of dependencies of seismic failures in SPRAs depends on the intended use of their results. | |||
- | * Further demonstrations of the identified methodology and its use would depend on arising uses or needs or, for example, applications to new designs, namely advanced non-light water reactors or modular reactor designs. | ||
8 | |||
Objective | |||
* As part of the 2010-2014 seismic and structural research plan, the Office of Nuclear Regulatory Research (RES) initiated research to: | |||
- evaluate and explore the seismic correlation-dependency topic and, if feasible, | |||
- to propose one or more advanced, more refined and more insightful analysis approaches for its treatment in seismic Probabilistic Risk Assessments (PRAs). | |||
9 | |||
Goals | |||
* That the evaluation of existing methodologies, the review of the related SSC-failure database, and the attributes of the proposed analysis methodology will inform SPRA analyses, peer-reviews, and staff reviews in relation to the treatment of seismic failures dependencies in seismic PRAs for operating and new reactors. | |||
* Identification of a methodology that with further demonstrations (by its potential users) and use will provide an improved and more insightful approach to seismic PRA 10 | |||
Scope | |||
* Assessment of impact of correlation assumptions made in modern SPRAs have on risk estimates (concentrating on CDF) | |||
* Review and identification of analysis methods and data sources that could be developed for improved and more insightful risk assessments | |||
* Identification of a pathway (methodology) so that improved analysis of dependent seismic failures could become a standard part of SPRAs 11 | |||
Analysis Aspects | |||
* Dependencies arise in the system analysis part of a PRA | |||
- Dependencies of SSCs arise in the calculation of the failure frequency of a sub-system of SSSs with performance defined by a Boolean (logic) statement | |||
* SSCs in parallel (redundant SSCs for example) involve AND gates | |||
- Assumption of dependency results in an upper bound estimate of the sub-system failure frequency (system fragility) | |||
* SSCs in series involve OR gates | |||
- Assuming independency results in an upper bound of the sub-system failure frequency (system fragility) 12 | |||
Analysis Aspects | |||
* Dependencies arise in the system analysis part of a PRA | |||
- Output of the analysis of the sub-system Boolean expression is essentially a fragility function for the system | |||
- Simplified (anticipated to be bounding) assumptions are typically made | |||
- An improved methodology would obtain narrower bounds on the system failure and address sensitivity of the results to defensible assumptions on dependencies and correlations | |||
* Approach: involve fragility and system analysts to obtain fragility functions for the sub-systems functions (Boolean expression) 13 | |||
=== | Analysis Aspects | ||
* Seismic Fragility: A = AmEREU | |||
- Am = median, ER= inherent random uncertainty about the median and EU = epistemic uncertainty about the median | |||
* Usually A = F x RE | |||
- F = a factor of safety and RE = a reference size of earthquake | |||
* F = FCFRS=(FSFu)(FSAFdFMFMCFECFSDFSS) | |||
Capacity Seismic Partial factor response factors factor 14 | |||
Analysis Aspects | |||
* Seismic Fragility: A = AmEREU | |||
- When there are several components in a Boolean statement, (B, C, D, ) the method involves: | |||
* Determining the correlation for the EU uncertainty between components (co-variances) | |||
* Assigning the remaining uncertainty (independent part of the uncertainty) to each component (preserving the total uncertainty) | |||
- Similarly for the uncertainties for ER | |||
- Then the methodology uses a set of rules to determine the frequency of the sub-system failure 15 | |||
Analysis Aspects | |||
* Uncertainties | |||
: 1. Determination of the correlation coefficients for the uncertainties in EU and ER | |||
: 2. Subsequent set of rules to calculate the sub-system failure probability | |||
* Item 1 - The correlation coefficients can be determined from the partial factors of safety and the combined to obtain those for Eu and ER (may reduce variability among analysts) | |||
* Item 2 - Verification of the simplified method against more rigorous existing mathematical approaches (too onerous for general use) 16 | |||
Study Approach | |||
* Review of several SPRAs in the literature (Chapter 4) | |||
* Review of the existing literature on seismic correlation and dependency analysis (Chapters 2, 6 and 7) | |||
* Review of existing data from testing and earthquake experience (Chapter 3) | |||
* Identification of a possible improved and more insightful methodology for treating dependencies in a SPRA 17 | |||
Review of SPRAs | |||
* Assessed impact of assumptions on the treatment of dependencies on the seismic failure of SSCs in risk estimates (concentrated on Core Damage Frequency (CDF) risk estimates) | |||
* Almost all SPRAs used the rule-based approach with few exceptions | |||
- The Zion SPRA for the Seismic Safety Margins Research Program (SSMRP) and the Diablo Canyon SPRA | |||
* Reviewed 10 SPRAs with CDF estimates ranging from 1E-7/year to 2xE-5/year 18 | |||
Review of SPRAs | |||
* Observations from the review (examples), which used expert judgement: | |||
- For some SPRAs, the differences in the seismic CDF could range from close (but less than) a factor of 2, but with a typical difference in the range of 30% to 60%. | |||
- For some accident sequences, the differences could be larger (as much as factor of 2 to 4) | |||
- Treatment of dependent failures has small impact if the CDF is controlled by an accident sequence that is itself controlled by a single SSC (singleton). | |||
- For several SPRAs the CDF estimates are not very sensitive to the dependencies issue but risk insights can be sensitive to the assumptions made in the analysis of dependencies | |||
- A few types of SSCs seemed to appear repeatedly among those for which the assumptions used affected the results (the CDF estimates) (although not all affect the results in every SPRA reviewed) 19 | |||
Review of SPRAs | |||
* SSCs that seem to appear repeatedly among those for which the approach to analyze dependencies in their seismic failure can affect the CDF estimates (not all appeared in this manner in all SPRAs reviewed): | |||
- Masonry walls | |||
- Electrical Motor Control Centers (MCCs) | |||
- Large tanks: condensate storage tanks of other similar tanks | |||
- Small tanks: diesel generator fuel oil day tanks | |||
- Heat exchangers: such as component cooling water heat exchangers | |||
- Mechanical: long-shafted service-water pumps, horizontal auxiliary feedwater pumps (motor and turbine driven dependency) | |||
- Batteries and racks 20 | |||
Review of Existing Literature on Dependency Analysis | |||
* A variety of approaches with different level of mathematical rigor have been proposed | |||
* SSMRP approach | |||
- Used join-lognormal distributions for the response and capacity of SSCs in a cut-set | |||
- Defines cut-set failure as all responses exceeding their capacities | |||
- Integrates the joint distributions over the failure region | |||
- Used in the Zion and SPRA and to derive a set of simplified rules (rule-based approach) called the Bohn-Lambright rules | |||
- Practice tends to use further simplified rules, which assign 100% to the situation of similar SSCs exposed to the same earthquake load (typically, SSCs located near each other), and zero everywhere else. | |||
21 | |||
Review of Existing Literature on Dependency Analysis | |||
* | * Bohn rule-based approach Rule # Description 1 Components on the same floor slab and sensitive to the same spectral frequency range (i.e., ZPA, 5-10 Hz. or 10-15 Hz) will be assigned response correlation = 1.0. | ||
-based approach) | 2 Components on the same floor slab sensitive to different ranges of spectral acceleration will be assigned response correlation = 0.5. | ||
3 Components on different floor slabs (but in the same building) and sensitive to the same spectral frequency range (ZPA, 5-10 Hz or 10-15 Hz) will be assigned response correlation = 0.75. | |||
- | 4 Components on the ground surface (outside tanks, etc.) shall be treated as if they were on the grade floor of an adjacent building 5 "Ganged" valve configurations (either parallel or series) will have response correlation = 1.0. | ||
6 All other configurations will have response correlation equal to zero. | |||
22 | |||
Review of Existing Literature on Dependency Analysis | Review of Existing Literature on Dependency Analysis | ||
*The review identified an approach that would lead to a more refined and more insightful methodology | * The review identified an approach that would lead to a more refined and more insightful methodology | ||
-Initially proposed by Reed and McCann (1985) but not further developed | - Initially proposed by Reed and McCann (1985) but not further developed | ||
-Leverages approaches widely used in the fragility analysis of SSCs that lead themselves to engineering identification of correlations in the uncertainties of variables that determine and SSC fragility estimation 23 Review of Existing Data | - Leverages approaches widely used in the fragility analysis of SSCs that lead themselves to engineering identification of correlations in the uncertainties of variables that determine and SSC fragility estimation 23 | ||
*Main observations in relation to data | |||
-driven support for determining seismic failure dependencies | Review of Existing Data | ||
*Shake-table testing data | * Main observations in relation to data-driven support for determining seismic failure dependencies | ||
-Shake-table test data is largely inadequate for purpose of understanding dependencies among failures | * Shake-table testing data | ||
-Mostly qualification testing data and when taken to failure mostly testing of single items | - Shake-table test data is largely inadequate for purpose of understanding dependencies among failures - Mostly qualification testing data and when taken to failure mostly testing of single items | ||
*Earthquake experience data | * Earthquake experience data | ||
-Largely inadequate for the purposes of understanding the dependencies among seismic failures | - Largely inadequate for the purposes of understanding the dependencies among seismic failures - Sparse data, difficult to understand and large interpretation uncertainties 24 | ||
-Sparse data, difficult to understand and large interpretation uncertainties 24 Analysis Process in a SPRA | |||
Analysis Process in a SPRA | |||
-by-one.Revised fragility functions 4.Use of the revised fragility functions in the SPRA in the usual | : 1. Performance of a SPRA 3. Use the improved methodology using standard methods to study the identified cut sets or Needs to include sufficient accident sequences one-by-sensitivity analysis studies) one. | ||
*Difficult to use methodology for the | Revised fragility functions | ||
*Variability among fragility analysts | : 2. Identification of those (presumably few) cut- 4. Use of the revised fragility sets or accident functions in the SPRA in the sequences where usual way correlations analysis May involve iterations to Step 3 or makes a difference Step 2 Difference in terms of risk quantification, importance, and safety insights 25 | ||
-Variability from analyst to analyst | |||
-Expert judgement to assign uncertainties as dependent or independent (determining coefficients of correlation) | Analysis Issues/Uncertainties | ||
-Would need further sensitivity assessments | * Difficult to use methodology for the revised fragilities except for most experienced SPRA fragility analysts | ||
*Issue of data availability and focus on matters of more or less conservative analyses (instead of, for example, looking for safety insights) | * Variability among fragility analysts | ||
*Peer-review needs 26 Summary / Remarks | - Variability from analyst to analyst | ||
*As part of the 2011 | - Expert judgement to assign uncertainties as dependent or independent (determining coefficients of correlation) | ||
-2014 seismic and structural research plan RES initiated research to explore the topic of dependent seismic failures of SSCs is SPRAs and, if feasible, to identify more refined and insightful analysis approaches for its treatment in SPRAs. | - Would need further sensitivity assessments | ||
*The research documented reviewed past SPRAs, the SSC | * Issue of data availability and focus on matters of more or less conservative analyses (instead of, for example, looking for safety insights) | ||
-failure database, and identified a more refined and insightful methodology that has promise for practical treatment of the dependency issue in SPRAs*The methodology described needs further demonstrations (by potential users) to provide an new and more insightful approach for SPRAs (in relation to the treatment of dependency of seismic failures)27 Summary / Remarks | * Peer-review needs 26 | ||
*The treatment of dependencies of seismic failures in SPRAs depends on the intended use of their results. | |||
*Further demonstrations of the identified methodology and its use would depend on arising uses or needs or, for example, applications to new designs, namely advanced non | Summary / Remarks | ||
-light water reactors or modular reactor designs. | * As part of the 2011-2014 seismic and structural research plan RES initiated research to explore the topic of dependent seismic failures of SSCs is SPRAs and, if feasible, to identify more refined and insightful analysis approaches for its treatment in SPRAs. | ||
* The research documented reviewed past SPRAs, the SSC-failure database, and identified a more refined and insightful methodology that has promise for practical treatment of the dependency issue in SPRAs | |||
* The methodology described needs further demonstrations (by potential users) to provide an new and more insightful approach for SPRAs (in relation to the treatment of dependency of seismic failures) 27 | |||
Summary / Remarks | |||
* The treatment of dependencies of seismic failures in SPRAs depends on the intended use of their results. | |||
* Further demonstrations of the identified methodology and its use would depend on arising uses or needs or, for example, applications to new designs, namely advanced non-light water reactors or modular reactor designs. | |||
28}} | 28}} |
Latest revision as of 08:25, 30 November 2019
ML18306A779 | |
Person / Time | |
---|---|
Issue date: | 10/03/2018 |
From: | Shukla G Advisory Committee on Reactor Safeguards |
To: | |
Shukla G | |
References | |
NRC-3927 | |
Download: ML18306A779 (134) | |
Text
Official Transcript of Proceedings NUCLEAR REGULATORY COMMISSION
Title:
Advisory Committee on Reactor Safeguards Structural Analysis Subcommittee Docket Number: (n/a)
Location: Rockville, Maryland Date: Wednesday, October 3, 3018 Work Order No.: NRC-3927 Pages 1-134 NEAL R. GROSS AND CO., INC.
Court Reporters and Transcribers 1323 Rhode Island Avenue, N.W.
Washington, D.C. 20005 (202) 234-4433
1 1
2 3
4 DISCLAIMER 5
6 7 UNITED STATES NUCLEAR REGULATORY COMMISSIONS 8 ADVISORY COMMITTEE ON REACTOR SAFEGUARDS 9
10 11 The contents of this transcript of the 12 proceeding of the United States Nuclear Regulatory 13 Commission Advisory Committee on Reactor Safeguards, 14 as reported herein, is a record of the discussions 15 recorded at the meeting.
16 17 This transcript has not been reviewed, 18 corrected, and edited, and it may contain 19 inaccuracies.
20 21 22 23 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 www.nealrgross.com
1 1 UNITED STATES OF AMERICA 2 NUCLEAR REGULATORY COMMISSION 3 + + + + +
4 ADVISORY COMMITTEE ON REACTOR SAFEGUARDS 5 (ACRS) 6 + + + + +
7 STRUCTURAL ANALYSIS SUBCOMMITTEE 8 + + + + +
9 WEDNESDAY 10 OCTOBER 3, 2018 11 + + + + +
12 ROCKVILLE, MARYLAND 13 + + + + +
14 The Subcommittee met at the Nuclear 15 Regulatory Commission, Two White Flint North, Room 16 T2B1, 11545 Rockville Pike, at 1:00 p.m., Peter 17 Riccardella, Chairman, presiding.
18 19 20 21 22 23 24 25 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
2 1 COMMITTEE MEMBERS:
2 PETER RICCARDELLA, Chairman 3 MICHAEL L. CORRADINI, Member 4 RONALD G. BALLINGER, Member 5 MARGARET SZE-TAI Y. CHU, Member 6 VESNA B. DIMITRIJEVIC, Member 7 JOSE MARCH-LEUBA, Member 8 HAROLD B. RAY , Member 9 JOY L. REMPE, Member 10 GORDON R. SKILLMAN, Member 11 MATTHEW SUNSERI, Member 12 13 DESIGNATED FEDERAL OFFICIAL:
14 GIRIJA SHUKLA 15 16 ALSO PRESENT:
17 JOSE PIRES, RES 18 MEHDI REISI-FAHRD, NRR 19 20 21 22 23 24 25 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
3 1 C-O-N-T-E-N-T-S 2 PAGE 3 Opening Remarks and Objectives 4 Peter (Pete) Riccardella, ACRS . . . . . . 4 5 Presentation and Discussion . . . . . . . . . . . 10 6 Correlation or Dependency of Seismic 7 Performance of Similar SSCs by Jose 8 Pires (RES) and Mehdi Reisi-Fard (NRR.
9 Closing Comments 10 Pete Riccardella, ACRS . . . . . . . . . 104 11 Adjourn . . . . . . . . . . . . . . . . . . . . 104 12 13 14 15 16 17 18 19 20 21 22 23 24 25 NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
4 1 P-R-O-C-E-E-D-I-N-G-S 2 (12:58 p.m.)
3 CHAIRMAN RICCARDELLA: The meeting will 4 now come to order. This is a meeting of the ACRS 5 Structural Analysis Subcommittee. I am Pete 6 Riccardella, Chairman of the Subcommittee. ACRS 7 members in attendance are, Vesna B. Dimitrijevic, 8 Margaret Sze, Harold Ray, Gordon Skillman, Mike 9 Corradini, Matt Sunseri, Joy Rempe, Jose March-Leuba, 10 and Ron Ballinger. We may or may not have two other 11 members come in later.
12 Girija Shukla of the ACRS is the 13 Designated Government Official for this meeting.
14 The ACRS reviews and advises the 15 Commission in regard to licensing and operation of 16 production and utilization facilities and related 17 safety issues, the adequacy of proposed reactor safety 18 standards, technical, and policy issues related to the 19 licensing of evolutionary and passive plant designs, 20 and other matters referred to it by the Commission.
21 The purpose of this Subcommittee meeting 22 is for the staff to brief Subcommittee on the topic of 23 Correlation or Dependency of Seismic Performance of 24 Similar SSCs. This is the topic of a research 25 publication, NUREG/CR-7237, Correlation of Seismic NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
5 1 Performance of Similar SSCs.
2 Members of the -- this is also the topic 3 of a research review, and I want to note that this 4 briefing that we're having here is independent of the 5 research quality review. That quality review is well 6 under way and the panel, the research panel is here 7 but they've pretty much already completed their 8 review.
9 The Subcommittee will gather information, 10 analyze relevant issues and facts. And formulate 11 proposed positions and actions as appropriate. The 12 meeting will be open to the public. We have not 13 received any written comments, or any requests for 14 time to make oral statements from members of the 15 public regarding today's meeting.
16 A transcript of the meeting is being kept 17 and will be made available as stated in the Federal 18 Register Notice, therefore we request that 19 participants in this meeting use the microphones 20 located throughout the meeting room when addressing 21 the Subcommittee.
22 Participants should first identify 23 themselves, and speak with sufficient clarity and 24 volume so that they can be readily heard. To avoid 25 interruption of the meeting, please mute your NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
6 1 individual lines during the presentations and 2 committee discussions. Also please silence your 3 phones. And I just, I've been advised that --
4 MEMBER CORRADINI: Yes, just for 5 everybody. So this test of the Emergency Broadcast 6 System via mobile is going to happen in a little bit 7 more than an hour. My recommendation is that you 8 power off your phones --
9 CHAIRMAN RICCARDELLA: Before.
10 MEMBER CORRADINI: -- before 2 o'clock so 11 it, because it will broadcast whether you have your 12 phone muted or not.
13 MEMBER MARCH-LEUBA: Yes, but 14 alternatively, you can put it in airplane mode.
15 MEMBER CORRADINI: It'll still --
16 MEMBER MARCH-LEUBA: Airplane mode --
17 MEMBER CORRADINI: Won't do it. It'll 18 still broadcast through it.
19 MEMBER MARCH-LEUBA: Airplane mode?
20 MEMBER CORRADINI: Yes if you power --
21 you're going to have to poser it off.
22 MEMBER MARCH-LEUBA: Let's try a test.
23 MEMBER CORRADINI: Okay, go ahead.
24 (Laughter.)
25 MEMBER CORRADINI: Feel free.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
7 1 CHAIRMAN RICCARDELLA: Let me, just for 2 those members who weren't on the Research Review 3 Committee, just a little bit of an introduction in my 4 simple way of looking at it, at the topic.
5 Suppose you have three pieces of 6 equipment, pumps that are required -- that are in 7 parallel, and one of them is required to work in an 8 accident sequence. And let's assume that the 9 probability of failure of each of those pieces of 10 equipment at a given level earthquake is 0.1. Okay.
11 If the components are independent then the 12 probability of none of them working is .1 of q, or 10 13 to the minus 3. On the other hand if they are 14 dependent, if the three are dependent, the probability 15 that if one fails, they're all going to fail, so the 16 probability is .1. And that's the significance that 17 we're talking about.
18 And current seismic PRA's there's a rule 19 of thumb which I'm sure Jose and others will discuss 20 that in some cases you assume they are perfectly 21 correlated, which is the 0.1. And in other situations 22 you assume that they're totally uncorrelated, which is 23 the 10 to the minus 3.
24 And I think what came to my mind when I 25 reviewed this was the potential significance with NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
8 1 respect to NuScale, where you might have 10 or 12 2 reactors in a common site. And you have a big 3 earthquake, you know, is the probability -- you know, 4 you could get into a 10 to the minus ten, 10 to the 5 minus 12th situation versus a .1 probability that 6 they'll all fail.
7 So, that's why I thought it might be a 8 good idea to have this meeting. And I think this 9 meeting will also serve as sort of some education for 10 the members of the committee who weren't on the 11 quality review.
12 MEMBER SKILLMAN: Pete, let me ask this, 13 for the record, I'm Dick Skillman. In this argument, 14 or in this example that you just explained, the way I 15 interpret that is if these three components are 16 identical, are operable, that means they're not pulled 17 along. They are in trigger, ready to go. And at the 18 same location, thus experiencing the same physical 19 behavior --
20 CHAIRMAN RICCARDELLA: Yes.
21 MEMBER SKILLMAN: Then all that you've 22 just mentioned is accurate. But of the three pumps, 23 if one is a 14 stage, one's a 13 stage, one's a 12 24 stage, one is in total lock, and one is on a different 25 set of supports. Then there is basis for the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
9 1 licensing to claim, not identical, different response, 2 and here's the response data for, you know, the shaker 3 table test I did for these. Therefore, the seismic 4 dependency while interesting, doesn't take out all of 5 my components. Is that an accurate framework for 6 this?
7 CHAIRMAN RICCARDELLA: Yes, and there's 8 current rules of thumb for how you handle that, okay.
9 MEMBER SKILLMAN: Okay, thank you.
10 CHAIRMAN RICCARDELLA: But as I understand 11 it, in fact even for those three identical pumps there 12 are aspects of their seismic fragility which are 13 correlated, and aspects that aren't correlated. Okay, 14 and so the reality is probably somewhere between zero 15 and one.
16 And for example, you know, if I think a 17 seismic qualification test is probably a lot like a 18 fatigue test. I run ten identical fatigue tests on 19 the same pieces of equipment, and the equipment can, 20 you know the samples will fail sometimes anywhere 21 between 1000 and 10,000 cycles, or 100,000 cycles.
22 And so there's a -- that's a total randomness that 23 even for identical equipment exists.
24 So, but this report is aimed at something 25 in between those extremes of zero and 1. And with NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
10 1 that I'll let Jose and the Research staff proceed.
2 MR. PIRES: Good afternoon, I am Jose 3 Pires. I am a Senior Technical Advisor in the 4 Division of Engineering, in the Office of Research.
5 I'm really the main preparer of this briefing because 6 this project was conducted by the Division of 7 Engineering. We had the lead for this project.
8 In the preparation of these slides, and in 9 preparation for this meeting, I discussed with some 10 other colleagues. I discussed with Selim, various 11 topics related to the issues of correlations. He's a 12 Seismic PRA Analyst. So he has experience in that 13 area. So if I could benefit from it. He's sitting 14 here if his help is needed to answer some questions 15 that I may not be able to answer.
16 And Mehdi, he works at NRR and he's been 17 involved in the reviews of Seismic PRAs that have been 18 submitted as part of the 2.1 Evaluation. So he can 19 provide some perspectives also. I also discussed with 20 him issues of seismic dependency, seismic failure 21 dependencies. And he can provide some perspectives of 22 what is being done nowadays in this area to address 23 this issue.
24 CHAIRMAN RICCARDELLA: Yes, perhaps some 25 perspective on the significance of it. In other words NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
11 1 if you did have a more accurate way of treating it, 2 would it have a significant effect of some PRAs or 3 not?
4 MR. PIRES: And I just in case that if we 5 talk about the applicability of some of these 6 methodologies here to these modular reactors, I also 7 talked to someone at NRO that is more familiar with 8 what is going on in relation to new reactors. So he's 9 sitting here too, and if necessary can also help 10 answer some things, if necessary, he can.
11 So, this is pretty much the outline. And 12 so I will try to focus mainly on the research before 13 us. And on the research that was then. So I'll talk 14 some about the objectives, some background leading to 15 research objectives and goals, how we thought these 16 could be used. The scope of the work.
17 And I quickly go, we're talking about the 18 approach and everything. I think it's probably a good 19 thing to talk about some aspects of the analysis. So 20 I'll, the treatment of correlations he's used in the 21 seismic PRA analysis. And what his methodology, what 22 are the enhancement that these methodology provide?
23 So this analyses aspect can go back to the 24 approach, it was a methodical approach involved in the 25 review of the existing PRAs to see what is the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
12 1 significance of the correlations on the analyses, or 2 the dependencies.
3 Review of current literature, and various 4 methodologies that I used. The project also involved 5 workshops with experts. And those workshops were 6 useful. Those are mostly seismic fragility knowledge 7 experts. They provided insights.
8 We questioned them for instance about what 9 is their data? Are they experimental data or 10 experienced data that somehow could be useful in a 11 more data driven approach? And we'll talk about that 12 later.
13 Then we'll talk about what will be the 14 process of implementing this methodology in the 15 seismic PRAs? It's really an obvious process but then 16 what could be issues associated with that, the use of 17 this methodology? Then I will summarize.
18 Right now, I say so that we have to, I 19 guess your introduction already covered that, that the 20 Seismic PRA risk, Seismic Probability Risk Assessments 21 have been done worldwide. And not just in the United 22 States for quite a long time too, more than three 23 decades. And some key aspects there, I guess one of 24 the aspects is that one has to consider all possible 25 earthquake levels in the PRA. And so the frequency of NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
13 1 occurrence of what we would say, some earthquake size 2 is what we relate to in the seismic hazard.
3 And then do a proper convolution of the 4 fragilities expressed in terms of those earthquake 5 size with the hazard, seismic hazard. That has been 6 addressed.
7 Topic today is that the earthquakes can 8 simultaneously damage multiple, redundant components 9 for example. And then also for similar components, 10 all at the same time. So, and that there is 11 quantification should account for that.
12 I have some pictures here just to try to 13 illustrate --
14 MEMBER RAY: Excuse me, I wanted to ask a 15 question on the slide before. The phrase, "all 16 possible earthquake levels". What does that mean?
17 MR. PIRES: Well, that means that the 18 seismic hazard should encompass the true scope of the 19 hazard. You know, not artificially have a limitation.
20 Is what the results of the probabilistic seismic 21 hazard analysis produce that should all be accounted 22 for in the seismic probability statistics.
23 MEMBER RAY: So there's no limit on the 24 range of probability, or is there a cutoff?
25 MR. PIRES: No, in principle there should NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
14 1 not be any, I mean. But what happens in practice is 2 that after some levels the probabilities are so small 3 that they can be neglected.
4 MEMBER RAY: Sure.
5 MR. PIRES: But they theoretically are 6 possible.
7 CHAIRMAN RICCARDELLA: Okay. You know, 8 you have a probability of occurrence of an earthquake 9 of various sizes and as the earthquake gets larger 10 obviously that probability goes down. But, you know, 11 they'll consider ground motions from .1g up to greater 12 than 1g. You know, 1g, 2g and they'll consider them 13 incrementally. And then each one of those earthquakes 14 has an uncertainty distribution around it as well.
15 Each one of those --
16 MEMBER RAY: Yes, I understand Pete, I was 17 just wondering if they at some point as he said, the 18 probability is low enough that it's cut off?
19 MR. PIRES: Yes, I --
20 MEMBER RAY: And I didn't know what all 21 possible --
22 MR. PIRES: That's what is meant. And in 23 many ways it's probably -- I don't like it. In the 24 case of the earthquake hazard, you somehow have to 25 choose now that it can come, go over the entire range, NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
15 1 and both of these very low probabilities. So, you can 2 do that.
3 CHAIRMAN RICCARDELLA: But, I think, also 4 when you say all possible earthquake levels, they're 5 really referring to the peak ground acceleration, 6 which is, you know, the ZPA, the g level of that.
7 They're not considering all multiple possible spectra 8 for a given GPA, are they?
9 MR. PIRES: Yes, the hazard will give us 10 that. For a site, it can give us the ground motion 11 response factor. And so, it dramatically could vary, 12 but normally in the bins of interest, it does not vary 13 much for a particular site. But if necessary, that 14 could be accounted for and make the process more 15 complex.
16 (Off microphone comment.)
17 CHAIRMAN RICCARDELLA: Okay.
18 MR. PIRES: So, I went and got some 19 pictures of earthquakes of first design -- looking at 20 power plant components. This is just to show that 21 sometimes really, we have structures or components 22 that are side-by-side for the same earthquake, and 23 some fail, and others do not fail, for a variety of 24 reasons.
25 And like the Northridge earthquake in NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
16 1 1994, north of Los Angeles, near the San Fernando 2 Valley. You can see there, some buildings, I think, 3 yes, the ones on the right, the first floor collapsed.
4 It's very typical on some of these old constructions.
5 First floor is soft, collapses.
6 The buildings now have one less story and 7 there are people living on that story or occupying 8 those facilities. Normally they are garages, so there 9 will be loss of life.
10 The performance faction for those 11 buildings typically is no collapse and no loss of 12 life. And the ones on the left, they do have 13 structural damage, but they did not collapse. They 14 are like very close to each other. I'm not saying 15 these are just purely statistical considerations, 16 there may be other factors. But this is an example.
17 Below is an apartment building, it was a 18 similar situation. You can see the ones on the left, 19 in the front have two story. The other on the left is 20 three story. And the first story also collapsed.
21 They are near, they're right next it, they did not.
22 I find interesting the components, the 23 electric power components, designed in a substation.
24 This was in Haiti, and you can see those components.
25 One would expect they all fail, because they typically NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
17 1 are not very robust for seismic loads. In this 2 particular case, only one failed. The others were on.
3 That does not mean that the substation did not fail 4 because there is a power flow going there, it's a 5 different issue.
6 This is in a power plant, but is not a 7 nuclear power plant in Guam, in an earthquake, and its 8 motor control centers. We don't know if they all 9 failed. But if they all failed they had different 10 failure modes because obviously one had a tilt-out 11 failure mode. The others might have had other 12 functional failure mode. I don't know, but at least 13 the failure modes were different in these phase.
14 Now, actually with the experience data, I 15 caution, sometimes it is very difficult to interpret, 16 because we don't have all information. But in this 17 case one would expect that the ground motion would be 18 similarly, the basic citation would be similar upon 19 all of them.
20 I just put one more here, these are tanks.
21 They are many makeup water tanks at the facility they 22 had built in place. And they normally tend to all 23 fail in an earthquake, but however the failure modes 24 are different for many of them. So there is a 25 possibility not all fail.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
18 1 In this case, they all had damage. There 2 is buckling off the shelf, there are some anchor 3 supports. So there may have been, they have different 4 failure modes. The one on the bottom right, sliding 5 which I don't understand. So there are issues in each 6 case. Sometimes not so obvious.
7 CHAIRMAN RICCARDELLA: The bottom left is 8 anchorage for the tank?
9 MR. PIRES: It's, I think it's an anchor 10 system that there was a lift off there.
11 CHAIRMAN RICCARDELLA: Yes.
12 MR. PIRES: Some amount of liftoff. I 13 cannot know, may not know the details of those, but 14 these was presented at one of the workshops, and was 15 a few years ago. But it illustrates that in some 16 cases you have multiple failures. Almost all the 17 components or structures fail. In other cases we 18 really don't have, in our observations of those.
19 MEMBER REMPE: Has any effort been made to 20 recently, to try and understand those differences?
21 Differences in anchorages, or there's just not data to 22 try and --
23 MR. PIRES: Oh, no. There --
24 MEMBER REMPE: -- understand this?
25 MR. PIRES: There are a lot of studies.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
19 1 With tank facilities there are quite a few studies.
2 MEMBER REMPE: Say it again. Could you be 3 closer to the mic? I'm having trouble hearing you.
4 MR. PIRES: With tank facilities there are 5 several studies to understand those behaviors. They 6 may not be looking at the same thing as we do in terms 7 of correlations. But, you know, there are a lot of 8 different, natural gas storage tanks, gas, and other 9 things like that. So these are critical facilities.
10 There is a good understanding of their behavior.
11 MEMBER REMPE: But I mean at Daini, do 12 they know why there were different failure --
13 MR. PIRES: Oh, at Daini?
14 MEMBER REMPE: -- modes as you've 15 indicated here?
16 MR. PIRES: I think this is to be expected 17 that there will be multiple failures of the water, of 18 these water tanks. Now, I have not done an analysis 19 of those. And I have not seen a report on those.
20 MEMBER REMPE: It just seems like, that an 21 important piece of information is to try and 22 understand what data are there, and can you extract 23 some insights from that data? And so examples, not 24 only over in Japan, but in the U.S. If we could try 25 and understand the differences it would be an NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
20 1 important aspect to investigate.
2 MR. PIRES: That was parts of the 3 workshops on these projects.
4 MEMBER REMPE: Right.
5 MR. PIRES: The workshops for these 6 projects had experts that typically go to every 7 earthquake that has major damage. They normally go 8 there as part of their professional work. They were 9 gathered to the workshop and they provided input. So, 10 and there is some conclusions that came out of that 11 work.
12 MEMBER REMPE: And when I reviewed those 13 conclusions, the data --
14 MR. PIRES: Well that, the data is very 15 difficult to interpret and back calculate.
16 MEMBER REMPE: And that's why I'm asking 17 this question. It seems like where there are data, 18 one should try to understand it before trying to model 19 it. And it's --
20 MR. PIRES: Yes, it's fraught with 21 uncertainties. And to the extent, Fukushima people 22 who developed models, tried to correlate them with 23 what exists in practice. But we cannot trust the data 24 by itself directly. You have to have the modeling, 25 and the knowledge, and the testing.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
21 1 (Simultaneous speaking.)
2 CHAIRMAN RICCARDELLA: You're probably 3 talking --
4 MR. PIRES: From the aspect HPS.
5 CHAIRMAN RICCARDELLA: You're probably 6 talking a big research project just to model one 7 situation like this. Model all the tanks and, you 8 know, you're probably, it's probably a research 9 project in itself that's of equal cost of this NUREG.
10 MEMBER REMPE: In other fields once you 11 get the model that's based on the data, you turn 12 around and find another place, and see if you can 13 apply that model, and if it can be validated. And so 14 that's why I'm more interested in the data before I 15 get interested in the model.
16 MR. PIRES: It is done. The experts 17 working on the field of tanks for example, they try to 18 always see how their results correlate with data. But 19 there are uncertainties on that.
20 MEMBER REMPE: Yes.
21 MR. PIRES: What happens in the field 22 normally does not have all the information that you 23 would like for a full benchmark.
24 MEMBER REMPE: Yes, I understand.
25 CHAIRMAN RICCARDELLA: Jose.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
22 1 MEMBER MARCH-LEUBA: I find interesting a 2 couple slides before, it's very visual. One more, 3 there. The picture on the right is really, that one.
4 You had to tell me what happens on the floors, the 5 missing floors. But when I look at that one I clearly 6 see one failed, three that didn't.
7 MR. PIRES: Yes.
8 MEMBER MARCH-LEUBA: Will these be 9 relevant only for only, quote-unquote, marginal 10 earthquakes? So it was, and by marginal I mean an 11 earthquake that is close to the acceleration where you 12 fail?
13 CHAIRMAN RICCARDELLA: Are you talking 14 about the Haiti?
15 MEMBER MARCH-LEUBA: The Haiti one.
16 CHAIRMAN RICCARDELLA: Okay.
17 MEMBER MARCH-LEUBA: I mean this one, it 18 tells me one failed, three didn't.
19 CHAIRMAN RICCARDELLA: Yes.
20 MEMBER MARCH-LEUBA: On the houses you 21 have to tell me what happened. Although it's clearly 22 worse.
23 MR. PIRES: Yes.
24 MEMBER MARCH-LEUBA: But it is only for a 25 small range of Gs that this will happen? If the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
23 1 acceleration had been ten percent higher, all of them 2 would have failed? The acceleration would have been 3 ten percent lower, none of them would have failed?
4 MR. PIRES: It's quite possible, it's 5 quite possible. There in this, for this particular 6 components they have prove --
7 MEMBER MARCH-LEUBA: No, but the basis of 8 their method. Is that we're going to separate this 9 plus-minus? Is this plus-minus ten percent? Is this 10 plus-minus 50 percent with random variability versus 11 --
12 CHAIRMAN RICCARDELLA: I would say, I'd be 13 surprised if ten percent made --
14 (Laughter.)
15 CHAIRMAN RICCARDELLA: -- made that much 16 of a difference.
17 MR. PIRES: I have no idea.
18 MEMBER MARCH-LEUBA: Yes.
19 MR. PIRES: In this particular, the 20 ceramic components on the substation tend to be 21 vulnerable. And so if the earthquake was stronger at 22 site, you almost would expect all of them to fail.
23 However, in this case, those cables going upwards do 24 provide some support.
25 MEMBER MARCH-LEUBA: Oh, they're holding NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
24 1 it.
2 MR. PIRES: And that might have worked 3 differently for the different components and that made 4 the difference in this case. So it also however, 5 those components, those cables going up, also may pull 6 up the bus that goes overhead. So, you have to 7 consider both cases.
8 MEMBER MARCH-LEUBA: Yes, this goes back 9 to our conclusions. And I know Vesna is going to talk 10 about that. This is going to be almost impossible to 11 recommend. Because you need to know how tight are 12 those cables on top?
13 MR. PIRES: Right. I meant, caution is 14 that as far as nuclear power plants are concerned, 15 this is not a component that appears in the PRA. But 16 it's just an example. It's normally other components 17 appear in nuclear.
18 This crisis in this case is an issue of 19 what are the anchors on these cabinets?
20 CHAIRMAN RICCARDELLA: Go.
21 MEMBER SKILLMAN: Jose, on this image, on 22 the lower left.
23 MEMBER MARCH-LEUBA: Here?
24 MEMBER SKILLMAN: No, the bolt. Very 25 commonly, the construction for that bolt is a J-bolt NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
25 1 that's been grouted or that's a Hilti that has pulled 2 loose. But in either case, I interpret that to be a 3 four or five inch displacement of the base of that 4 tank against what had been the original forked bolt.
5 That bolt has been pulled loose.
6 MR. PIRES: Yes, probably the bolt might 7 have --
8 MEMBER SKILLMAN: The force that pulled 9 the bolt loose was the ear on the tag that moved 10 upward. It pulled the bolt right out of the ground.
11 MR. PIRES: Yes.
12 MEMBER SKILLMAN: Is that what you see in 13 that image?
14 MR. PIRES: I would say that that is 15 probably the correct interpretation. I would say, 16 because I think there might have been an overturning 17 moment, and the entire force on that bolt might have 18 failed the bolt. And then it was sliding after that.
19 But you could also fail the tank involved, depending 20 on what is the weak link there? So it is, so either 21 can be a weak link. Normally, it would probably, the 22 bolt failure would be the --
23 MEMBER SKILLMAN: Likely, the bolt 24 anchorage.
25 CHAIRMAN RICCARDELLA: The anchorage of NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
26 1 the bolt, anchor fail.
2 MR. PIRES: Not, the tank link itself 3 would be the buckling. And that does not necessarily 4 mean that there is a leakage of the water. But it is 5 normally assumed that there is. But that does not 6 mean that there is. It might still hold water.
7 There, so I'll just say that normally 8 these, of course all these SPRAs normally involve a 9 set of, a rule-based approach so to speak, on how to 10 deal with the dependent failures. And typically data, 11 two parameters that are important is the similarity of 12 the components at the very high level and the 13 proximity of the components.
14 So based on those two parameters now you 15 have four cases, four combinations, similar components 16 close by, similar components far apart. But a 17 different storage, and dissimilar components close by, 18 and far apart.
19 So the rule-bases account for those cases.
20 And they try to make, I just put there two cases, and 21 not in case I need to go here with all the details of 22 the rule-bases. But obviously they are intent to be 23 close together. They are close together, they are 24 similar normally. And that's the baseline knowledge 25 is assumed correlation on those failures.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
27 1 MEMBER MARCH-LEUBA: So this is the state 2 of the err. This is what everybody has been doing for 3 the last 35 years.
4 MR. PIRES: That's what people do, yes.
5 But is not the state of the -- I would say it is the 6 state of the practice.
7 MEMBER MARCH-LEUBA: Right. People know 8 better, but they're, right.
9 MR. PIRES: Right.
10 MEMBER MARCH-LEUBA: Now would you explain 11 the logic? And I don't know anything about seismic 12 responsive buildings, but I would expect that if a 13 tank fails on the basement of a building, on floor 14 five, they will fail too. And according to this, you 15 failed them, you have correlation only if you're on 16 the same floor, but not on higher floors?
17 MR. PIRES: Yes, that is a complicated 18 situation. Because if you do a, calculate the 19 structural response of a building, the acceleration is 20 on the floors of the building. So that will be like 21 the new ground for the components mounted there.
22 You can actually calculate the correlation 23 factor between those, the accelerations on the various 24 floors. And you can see that they, the correlation 25 factor actually goes down. Can become small, around NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
28 1 .2 or .3. You know, if they are on the same floor it 2 will be higher. But if there are different floors, 3 you can start decreasing the correlation factor. It 4 is not by necessarily, it could be 1 or even .5.
5 MEMBER MARCH-LEUBA: My point, just not 6 knowing how this thing is done is, I would have done 7 if one fails, all the ones that are in weaker 8 positions, fail too.
9 MR. PIRES: If the ground motion that is 10 higher than the one with the weakest ground motion 11 fails, then that probably can happen. But there is, 12 the correlation coefficient between the motion of the 13 various floors does decrease with the --
14 MEMBER MARCH-LEUBA: So they --
15 MR. PIRES: -- floor level and the 16 distance.
17 MEMBER MARCH-LEUBA: -- kind of the G 18 forces are smaller on the 5th floor than on --
19 MR. PIRES: Well they normally are higher 20 on the fifth floor. Because there is amplification to 21 the top. But what I'm saying is the correlation 22 coefficient between them.
23 MEMBER MARCH-LEUBA: What I would do if 24 something fails on the 5th floor, doesn't mean the 25 four, three, two, and one are bad. But if something NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
29 1 fails on one, then two, three, four, and five are bad.
2 MR. PIRES: It depends on -- I see 3 particularly there is an amplification as you go up.
4 There is also a bandwidth on the ground motion. As 5 you go up and amplify, you also have more narrower 6 band ground motion. So, depending on what are the 7 frequencies of the components, and IRS, and then we 8 test those narrow bands and motions. So that also is 9 another effect.
10 CHAIRMAN RICCARDELLA: Tends to go toward 11 the natural frequency of the building.
12 MR. PIRES: Yes.
13 CHAIRMAN RICCARDELLA: And or the mode --
14 MR. PIRES: But in the lower floors the 15 bandwidth is a lot broader.
16 CHAIRMAN RICCARDELLA: Yes.
17 MR. PIRES: Tends to be broader.
18 CHAIRMAN RICCARDELLA: Yes.
19 MR. PIRES: So, and, but that is captured.
20 The response part, yes now is can capture it 21 reasonably well. Is the capacity part in my opinion 22 that is more difficult, for which it is more difficult 23 to establish the correlations.
24 CHAIRMAN RICCARDELLA: So in a seismic 25 puree, does the fragility, the person developing the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
30 1 fragility curves, he considers the difference in 2 response say to a .5 g earthquake, between the 1st 3 floor and the fifth floor? That's built into his 4 fragility --
5 MR. PIRES: Yes.
6 CHAIRMAN RICCARDELLA: -- curves. So 7 that's taken into account? So it's, the correlation 8 is different than the magnitude of the response.
9 MEMBER MARCH-LEUBA: I have been told by 10 some experts in this room that that was not the case.
11 But obviously I was wrong.
12 MR. PIRES: How it is, it is.
13 MEMBER MARCH-LEUBA: I was misinformed.
14 MR. PIRES: You try to express all the 15 fragilities in terms of the ground motion at one 16 reference location. But the calculation of that 17 accounts for the transfer function between these 18 locations.
19 So the capacity of the component may be 20 the same but the fragility can change because of the 21 location, and the ground motion varies. So there is 22 a difference there.
23 MEMBER MARCH-LEUBA: Yes, and I can see 24 how the building can because of the natural frequency, 25 can filter some. I mean you could have a high NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
31 1 frequency, you have less shock upstairs than 2 downstairs.
3 MR. PIRES: Yes, there is some of that.
4 And, you know, the question is how much of that, of 5 the process is involved some type of broaden of --
6 MEMBER MARCH-LEUBA: Yes.
7 MR. PIRES: -- of the motion. And so we 8 are careful to do that. But those effects do exist.
9 MEMBER MARCH-LEUBA: It's a complex 10 phenomenon.
11 (Off microphone comment.)
12 MR. PIRES: Those effects exist. The fact 13 that the ground motion frequency content can change 14 somewhat with the location, and is important.
15 So I guess one part of the background was 16 that of course the methodology makes this bounding --
17 simplifying assumptions. Later on when we talk about 18 the literature review, she'll review our mission that 19 initially there was a set of simply finding 20 assumptions that was more complex than these. But 21 even that set of simply applying assumptions has 22 fallen off somehow.
23 One of, of course sensitivity analysis are 24 always done. I see the analysts pick up a baseline 25 case and they do sensitivity analysis to see how NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
32 1 important those aspects are.
2 Sometimes people expressed their idea that 3 although these bands are large, you may mask some 4 insights that otherwise would not be masked. If you 5 had data more detailed than all this.
6 And now I would like to mention that the 7 treatment, well I'm not saying that the treatment of 8 dependencies the way it done now is deficient, or 9 somehow inappropriate. The treatment of dependencies 10 depend on the intended use of the results. And what 11 is the need for those results?
12 So that is one aspect, and so that I would 13 like to caution that the other thing is that the 14 methodology that was developed that here, that might, 15 will need further demonstrations and development.
16 But its use in those further 17 demonstrations will definitely, are going to depend on 18 our rising uses or needs.
19 CHAIRMAN RICCARDELLA: Yes.
20 MR. PIRES: To my understanding the 21 current uses and needs, what has been used, seems to 22 be sufficient. But that may change.
23 CHAIRMAN RICCARDELLA: Okay.
24 MR. PIRES: So, I think particular 25 applications, that some more elaborate studies have NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
33 1 been that.
2 CHAIRMAN RICCARDELLA: You know, I think 3 going back to the previous slide, I think it's, to me 4 an important point is that insights that you might 5 get. You know, for example if you have three pieces 6 of equipment, maybe it would be smart to just change 7 the mounting a little bit.
8 Make it, you know, come out and rotate 9 them or put different size gaskets in there, or 10 something. Just anything to make them not identical.
11 You might reduce the -- you know, if it turned out to 12 be significant it could have a significant --
13 MEMBER MARCH-LEUBA: But only for those 14 pieces or components of the SPRA, defined as critical.
15 (Simultaneous speaking.)
16 MEMBER MARCH-LEUBA: And which would be 17 one or two, it won't be the whole plant.
18 CHAIRMAN RICCARDELLA: Yes, I would think 19 the sensitivity analysis, you'd say well, I'll do it 20 one way with everything 100 percent correlated. I'll 21 do it a second time with everything percent zero 22 correlated. And if the core damage frequency comes 23 out the same, obviously doesn't matter, right.
24 MR. PIRES: In this, one of the slides 25 later on says, of the implementation, all the use of NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
34 1 these, what will be the process for use these? And of 2 course the process is iterative. It starts with 3 simply finding assumptions based on that, you identify 4 those few, you know, cut-sets or accident sequences 5 that need to be further refined and investigated.
6 CHAIRMAN RICCARDELLA: Yes.
7 MR. PIRES: And you'll proceed that way.
8 MR. REISI-FARD: Can I provide a comment 9 on that question?
10 MR. PIRES: Sure.
11 MR. REISI-FARD: So this is Mehdi 12 Reisi-Fard. I'm an acting team lead in the Office of 13 Nuclear Reactor Regulation. On the point that you had 14 with respect to sensitivity analysis, in fact that's 15 what the licensees do in some risk-informed 16 applications like 50.69.
17 As a part of the NEI guidance to do those 18 50.69 evaluations, when they use seismic PRA, it is 19 recommended to do, to assume full correlation and no 20 correlation on the same floor, to just get their --
21 sites.
22 So that's part of the guidance that the 23 industry is using currently for some of those 24 applications.
25 CHAIRMAN RICCARDELLA: Yes, thank you.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
35 1 MR. PIRES: Yes, I'd like to put this 2 first bullet, you may think of it as a disclaimer, but 3 I don't want to claim that what the current practice 4 is necessarily not a good practice, you know. This, 5 so I will go back what was the objective of the study.
6 You know, the 2010, 2014 seismic constructional 7 research plan in the Office of Research, initiated 8 this study.
9 And those are the objectives of the study.
10 So a lot of these topics, and if feasible to propose 11 some refinements to the methodology that could provide 12 those additional refinements, or could be more precise 13 for needs that at that time were thought that might 14 come up.
15 And at that time there were some studies 16 going on, severe accident studies going on, that 17 motivated part of this study, even though they were 18 not PRA studies. The Level 3 PRA project was going to 19 start. There was talk of modular reactors coming for, 20 even though that far back there was already talk of 21 that. And then that subsided somewhat, and then along 22 it came it back again.
23 So, the Office was being prepared to 24 address these issues. And this, the goals of course 25 are that somehow the review that the report did, the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
36 1 opinions of the experts on the components that 2 typically appear as critical, and that appear on the, 3 as important in terms of dependencies, and other 4 information there, could support staff reviews or peer 5 reviews, you know. It's tool that could be consulted 6 and used in those reviews.
7 And so, the methodology depending on the 8 needs of potential applicants, and our staff in terms 9 of their reviews could provide more insightful 10 approaches.
11 Now the scope of the work as defined in 12 the statement of work, it said, first part was, assess 13 the impact of correlation and some different 14 assumptions on the results of seismic PRAs that have 15 been done to date, and that they read about. This 16 concentrated on core damage frequency. They reviewed 17 --
18 And so then, the other part was what we 19 talked before. The review of what, an identification 20 of an analysis method that could be an improvement, or 21 thought it is now, but will not necessarily be a very 22 cumbersome method. That will not involve a level of 23 effort and work that would be another of magnitude 24 higher. But maybe 50 percent more level of effort.
25 So, and that was the intent there.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
37 1 The third way, the third is of course 2 identify a pathway so that this methodology could 3 become standard. That of course is a more, does not 4 depend on the NRC. Depends on the need of the 5 industry.
6 MEMBER SKILLMAN: Jose, before you move 7 that slide, for the first two bullets, how many 8 examples, or how many calculations, or how many --
9 what is the population of the review, as you point to 10 these two bullets?
11 MR. PIRES: Okay.
12 MEMBER SKILLMAN: Is it 100 samples, 80 13 samples, 500 samples?
14 MR. PIRES: Right, the project really 15 reviewed what existed into analysis by themselves.
16 But they reviewed 20 PRAs, 20 cycling PRAs. Some were 17 done in the last ten years, other were done probably 18 ten to 20 years ago. So there were already were the 19 ways of doing the PRA during that timeframe.
20 The analysis methods that I used, yes 21 several were reviewed. Some very simple method like 22 what they call split fraction method, to more 23 complicated methods like the Simex Safety Modular 24 Research Program, then at Livermore. I don't know if 25 you heard of that. Simex Safety Modular Research NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
38 1 Program done at Livermore, supposedly by the NRC.
2 Really involved defining for cut-sets, a 3 joint probability distribution fraction. For the 4 response in the capacity and then integrating the two 5 together to calculate what was the probability of 6 failure. Was a very complex and detailed approach, 7 and that was used in designing PRA. So this addresses 8 the second bullet. So that is more complex approach, 9 and in many ways, the more rigorous approach.
10 And that yielded a set of rules. More 11 complicated than the ones used now, but still simple.
12 So we reviewed the entire scope of methodology that 13 had been used and proposed.
14 MEMBER SKILLMAN: Thank you.
15 MEMBER REMPE: So later you cite the 16 impact of those assumptions. And you do, as you noted 17 in the first bullet, focus on CDF. What would have 18 happened, you know, how important would it have been 19 if you'd focused on the risk impact versus the 20 frequency impact. Would you still have 30 to 60 21 percent, and up to a factor of 2 to 4? I mean are 22 these important from a risk perspective as well as a 23 frequency perspective is what I'm trying to --
24 MR. PIRES: In terms of the --
25 MEMBER REMPE: Of the studies you NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
39 1 reviewed?
2 MR. PIRES: -- risk of consequences, you 3 --
4 MEMBER REMPE: Right. I mean are we 5 looking at things that are not really risk important 6 sequences? And okay, you say well it could be up to 7 a factor of 2 to 4 for some of these sequences. And 8 does it impact risk as much as it does frequency?
9 MR. PIRES: I think you're if you start --
10 right. This was outside the scope of the work, so I 11 cannot really tell you. And also there are very few 12 PRAs that have gone that far.
13 MEMBER REMPE: But you looked at those two 14 or three, or whatever you looked at. And I'm just 15 wondering what was the impact on risk of those PRAs?
16 MR. PIRES: I would be speculating.
17 (Off microphone comment.)
18 MEMBER MARCH-LEUBA: Are you implying that 19 different core damages have different risk, different 20 consequences?
21 MEMBER REMPE: Yes, I mean okay, the 22 frequency could, they're quite, this differences of 23 frequency, or what they're giving values of, and so, 24 yes, you can have core damage where you have a big 25 release. And you can have core damage when you don't NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
40 1 have squat coming out.
2 And that's what I'm asking. And are we 3 focusing on something important or not is where I'm 4 getting to?
5 CHAIRMAN RICCARDELLA: But do seismic PRAs 6 address that, or that they just compute a --
7 MEMBER DIMITRIJEVIC: Well core damage 8 frequencies, least measure. It just will at least 9 measure the large release frequency for example. Not 10 too many seismic things went to the Level 2, or Level 11 3 that you can compare it. That's at least my 12 understanding.
13 MR. PIRES: And the regular --
14 MEMBER DIMITRIJEVIC: And they're only 15 special things. Because those frequencies, also these 16 measures, so it's a large release which is next 17 running through the consequences, and then you have a 18 Level 3, which is, you know, public release which is 19 different thing.
20 But the seismic so far, I mean I only saw 21 core damage frequencies, but I'm sure that now new 22 ones are also producing land releases, right?
23 MR. PIRES: I think so.
24 MEMBER DIMITRIJEVIC: Yes, yes so.
25 MR. PIRES: Now they are.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
41 1 CHAIRMAN RICCARDELLA: What was that last 2 point? I missed that.
3 MEMBER DIMITRIJEVIC: I think that those 4 sites will now, when they're handling this seismic 5 PRAs is a part of this 2.1 initiative. I think they 6 are going to the large release too.
7 CHAIRMAN RICCARDELLA: Okay.
8 MEMBER DIMITRIJEVIC: But I don't know the 9 ones which you reviewed, these 20 PRAs, did they have 10 releases?
11 MR. PIRES: The one I, no they did not.
12 Some of the ones that -- one of the PRAs that were 13 more complex were the NUREG-1150. You know, some of 14 those went to large, to releases, but not for seismic.
15 The seismic part ended before that, was not continued 16 all the way there.
17 MEMBER MARCH-LEUBA: Because for that you 18 would have to consider correlation between this C 19 pumps over here, and failure of the containment.
20 MR. PIRES: Yes.
21 MEMBER MARCH-LEUBA: And you will have to 22 consider that correlation on top of the other one 23 which is already a difficult to have.
24 MR. PIRES: I don't want to steer away 25 from the topic, but some of the multipatient from the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
42 1 study back in 2010 time to '12 timeframe was the 2 SOARCA study. Because you had some sequences and 3 eventually the containment, there would be a release 4 and others. And so but in the end, were never used 5 there obviously.
6 The report work was turned in before and 7 that was not the risk study. It was a severe 8 accident, yes. So you can see that, yes, it was along 9 those lines. Also you have to bring evacuation 10 issues, relocation issues, and there those are also 11 another type of problem. And at this moment NUREG-12 1150 been stopped before, before that part is external 13 hazards.
14 And this, I talk a little bit about the 15 analysis aspect because we keep talking about the 16 methodology that somehow was proposed. And might be 17 better. And I felt, don't want to delay that too much 18 before going back to some of the review of the 19 literature.
20 And so it might, I think what the authors, 21 if I could put myself in their frame of mind, the way 22 I understood it, is that so these dependencies really 23 arise on the system analysis part of the -- when you 24 do the fragility analysis for a component, you don't 25 have this issue. You're talking to a single component NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
43 1 there.
2 So in the system analysis sometimes you 3 would have a set, or a subsystem that involves various 4 components, and the performance of that subsystem is 5 a Boolean expression. Simplest one would be the 6 series, the components in parallel, where if you 7 assume dependency, it will be an upper bound estimate.
8 Those are the AND gates. And OR are the ones in 9 series, which is the converse way, that is simple.
10 So in the end what you really get out of 11 that is a fragility fraction. Even though much has 12 been the fragility of one of the components. Or a 13 multiplication of the fragility, but you really get a 14 fragility fraction that represents that subsystem.
15 So the multipatient of these researchers 16 was a --
17 MEMBER RAY: Wait.
18 MR. PIRES: Yes.
19 MEMBER RAY: I waited to ask a question 20 before you changed that slide.
21 (Off microphone comment.)
22 MEMBER RAY: Well, I thought I'd pressed 23 it but I didn't, hadn't. This is a -- I started to 24 say an unfair question maybe, but nevertheless I've 25 got to ask it because it's hanging on my -- would you, NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
44 1 in talking about the AND gates, structures and some 2 components in parallel. It makes me think as somebody 3 who operated plants for a long time, what do we 4 assume, or is it relevant to even think about what the 5 tech specs require in terms of, being in service?
6 In other words, I've got three pumps but 7 I only have to have two of them in service. And I can 8 take one of those out for a week. Does that ever 9 enter into these analyses? Maybe Mehdi could shed 10 light on that --
11 MR. REISI-FARD: The short answer is yes.
12 So the reliability or availability numbers that are 13 already included in PRAs, they do take into account, 14 you know, taking certain SSCs out of service for a 15 period of time. So the numbers in the PRAs, the 16 availability numbers in the PRAs take into account 17 that factor.
18 MEMBER RAY: Okay, and that's good to 19 hear. I assumed it wasn't germane to exactly what we 20 were talking about here, but nevertheless it was a 21 question that I had because systems required to be in 22 service is sometimes a real burdensome requirement 23 that people back away from in actual practice. And 24 that's what I was interested in.
25 MR. REISI-FARD: And I used the word NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
45 1 availability, I should have said, unavailability.
2 MEMBER RAY: Yes, well that's fine. Thank 3 you.
4 CHAIRMAN RICCARDELLA: And just help me on 5 that first subbullet that you have the SSSs. Is that 6 a typo, is that supposed to be SSCs?
7 MR. REISI-FARD: Yes.
8 CHAIRMAN RICCARDELLA: I just wanted to 9 make sure we weren't introducing a new acronym here.
10 I have enough trouble --
11 MEMBER REMPE: Wise.
12 (Off microphone comments.)
13 MR. PIRES: Sorry. So what was, what's 14 done here was to try to say okay, this is, these 15 dependencies issues are right, getting the system 16 analysis process. But in the reality what you are 17 trying to do is that for that subsystem where the 18 dependencies may be significant and important, you're 19 really trying to calculate a new fragility fraction 20 for those. It is essentially what you are doing that 21 those sets of rules are doing that, using what 22 fragilities that you already have.
23 So the approach was to see okay, what, how 24 are these fragility fractions calculated for each 25 component? And then break down the process into NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
46 1 various factors that have some physical meaning or 2 engineering meaning, so that you more easily would be 3 able to break down what are the parts that would 4 likely be correlated there. So that was the approach 5 that was done here. And so then the methodology could 6 then provide this spectra between fully correlated and 7 independent.
8 I tried to put here something just to try 9 to show what I just said. And I'm not trying to come 10 much more intent in terms of these equations, but in 11 the seismic fragility, normally is expressed in terms 12 of a ground motion level. A, so it is expressed in 13 terms of the median of that ground motion level, that 14 in this case represents the capacity of the component.
15 Can it resist that ground motion level?
16 And then two random variables, one they 17 will refer to it as the random uncertainty. And the 18 other the epistemic uncertainty. And normally this 19 process is simplified by saying that the fragility can 20 be expressed in terms of effect of safety, times a 21 reference level earthquake. The spectra of safety is 22 really just scales of these, but it's just a 23 convenience. Now --
24 CHAIRMAN RICCARDELLA: The reference level 25 is like SSE, or OBE?
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
47 1 MR. PIRES: Could be the SSE, or OBE, or 2 at the given as a level, being say, the ten to the 3 minus four if you are talking about, if we think the 4 dominant being, would be around that level, ten to the 5 minus four, as a -- this, so you have that factor of 6 safety.
7 So it is component, capacity, components 8 and it is seismic response components, and those are 9 even further broken down into either the capacity as 10 some, these are strengths. The other is the 11 ductility.
12 The partial factors for the seismic 13 response have many parameters like the uncertainties 14 on the spectra of the ground motion. How you combine 15 the mode shapes, to the substructure direction. So 16 the transfer function that brings you the ground to 17 the various floors that you asked about.
18 So what this method does is that if you 19 have various components in that cut-set, in that 20 Boolean expression A, B, C & D. You are going to try 21 to find out what is the correlation coefficient on 22 those uncertainties. When you got to the epistemic 23 uncertainties that are, what part of those 24 uncertainties between A&B, and B&C and D is really a 25 correlation coefficient, the co-variance?
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
48 1 CHAIRMAN RICCARDELLA: Could we go back to 2 the previous one, and just help me with what all those 3 F factors are on the bottom.
4 MR. PIRES: Right, I know this is a bit 5 hard.
6 CHAIRMAN RICCARDELLA: I understand, you 7 know, capacity and response, okay. But then the next 8 one, are you saying that the capacity breaks down into 9 that F sub S, and F sub U?
10 MR. PIRES: Yes, that is the capacity they 11 -- it is for fragility analysis not for design. Is 12 that if you have for instance a sheer wall on a 13 building, you know what the distress of that sheer 14 wall. But you know that it may have some ductility.
15 And so because of that ductility, it can actually take 16 higher seismic loads. Not just the ones that goes to 17 the region that's strained. So the analysis is 18 rigorous and accounts for this.
19 I don't want to make your discussion on 20 all the fragility analysis done, but that it has all 21 these factors appear there. So when the analyst --
22 CHAIRMAN RICCARDELLA: Just tell me what 23 some of those subscripts stand for.
24 MR. PIRES: Okay, so the first one on the 25 response factor, the first one, SA, stands for the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
49 1 uncertainties on the response factor of the ground 2 motion. If you have a frequency, say 5Hz, you know 3 what is the median value of the spectral acceleration.
4 But there is an uncertainty about that.
5 CHAIRMAN RICCARDELLA: Okay.
6 MR. PIRES: So that they are accounted 7 there. D is the damping. We may not know as well the 8 damping of the structure as we think we know. M, now 9 I'm a little worried about what this M, I forgot, but 10 it will probably come back.
11 But MC is the model combination rule that 12 you use. You have to combine all these mode change.
13 EC is how you combine ground motion in different 14 directions, because the earthquake goes actually in 15 all directions. You know, you just cannot do an 16 analysis for one.
17 CHAIRMAN RICCARDELLA: Yes.
18 MR. PIRES: And so that's EC. And SD is 19 the depth. If the structures are embedded, you know, 20 you normally tend to reduce the forces on the 21 components but there are uncertainties also. That 22 also brought uncertainties.
23 CHAIRMAN RICCARDELLA: Yes.
24 MR. PIRES: And the SS is the 25 soil-structure interaction. You know, when you NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
50 1 interact the structure is not in -- it couples with 2 the ground and some energy is radiated away --
3 CHAIRMAN RICCARDELLA: Right.
4 MR. PIRES: -- and changes the resonant 5 frequency. So these are factors that appear on the 6 calculation of the response. Every one of those has 7 a random uncertainty and epistemic uncertainty.
8 CHAIRMAN RICCARDELLA: Yes.
9 MR. PIRES: And so what is the first is 10 trying to do here? What they try to do is if you add 11 several components, a component A and a component B, 12 or a B and C, a component B and a component C, you 13 will try to see on that ER and EU, what part of that 14 uncertainty between those two components is really 15 correlated?
16 And so you will break down the uncertainty 17 into correlated and then what it is for the component 18 itself is independent.
19 MEMBER MARCH-LEUBA: Since Pete brought 20 the math, why do you combine the certainties in a 21 multiplicity of sense, instead of being like the sum 22 or the sum of the squares, or -- does that make sense?
23 MR. PIRES: The uncertainties are going to 24 the sum of the squares.
25 MEMBER MARCH-LEUBA: I thought they NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
51 1 weren't.
2 (Simultaneous speaking.)
3 MEMBER MARCH-LEUBA: Well, those Fs are 4 not uncertainty?
5 MR. PIRES: Those Fs are not the 6 uncertainties.
7 CHAIRMAN RICCARDELLA: Those are factors.
8 MR. PIRES: Each one of those Fs is equal 9 to a median times two uncertainty factors. One for 10 random, one for epistemic. So those Fs are the 11 factors themselves.
12 CHAIRMAN RICCARDELLA: Yes.
13 MR. PIRES: And that factor is now, F, 14 those factors, each one of those factors has an 15 expression like the first one above.
16 CHAIRMAN RICCARDELLA: As a mean plus a 17 distribution.
18 MR. PIRES: Plus two factors on all you 19 vary around the mean.
20 MEMBER MARCH-LEUBA: Okay, so there was a 21 reason to the madness, and I believe you.
22 MR. PIRES: Right, right. No, it is the 23 way to account for all those factors there. So until 24 the uncertainties on all those are combined into the 25 sum of squares, and the square root. And essentially NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
52 1 what the method is trying to do is really almost like 2 finding a set of accumulating dependent random 3 variables in which they try to separate what is 4 correlated from what is not correlated between pairs 5 of components. But without necessarily throwing away 6 uncertainties. Still trying to preserve the 7 uncertainty, but allocating it.
8 Now if you go and allocate the uncertainty 9 at the high level there on the first line, that may be 10 too abrupt, may be very judgmental. The idea of 11 allocating the uncertainty through this method is that 12 each one of those components you may have more 13 engineering insights, to say what is to be correlated 14 or not. For instance in the response we can actually 15 have an analytical way to somehow calculate 16 correlation. And the capacity is more difficult.
17 MR. PIRES: All right, so I move forward.
18 And now, what are uncertainties on this process? So 19 this process demonstrated in the report for various 20 cases. But the report really does not go to the, 21 starts more or less focus on the first line there.
22 They don't go into detail to the bottom line, but 23 that's necessary. So I just want to mention that, why 24 it is necessary?
25 Because of the uncertainties, because and NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
53 1 different analysts may come up with different results.
2 So the more you try to break down and things limit 3 you, you have really engineering understanding and 4 meaning.
5 So, all you allocate what is correlated on 6 the uncertainty, to what is independent is subjective 7 in some ways. So there is engineering judgment. But 8 if you break down to those basic variables, and then 9 mathematically bring them back up to the higher level, 10 you have a process that is more transparent, so to 11 speak. That's the intent here.
12 Now, talk a little bit about the parts of 13 the study. So the review of the SPRAs was processed.
14 The impact of the assumption, the predicament of 15 dependencies. And the conclusion there was that 16 almost all SPRAs have used rule-based approach. And 17 simple rule-based approach is perfectly correlated, 18 totally dependent, and can date sensitivity analysis 19 of those.
20 There have been a few exceptions, like I 21 mentioned before, the Zion SPRA is an exception on the 22 Seismic Safety Margins Research Program. So a very, 23 these were done in the late, early, 70s. So there was 24 a lot money to do these research projects, and we were 25 learning. I mean what we do now, we learned from NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
54 1 those projects.
2 And the Diablo Canyon SPRA also uses more 3 complicated set of rules. It essentially started with 4 the SSMRP approach, and then simplified to those rule-5 based approaches, but not just a direct use of them, 6 more complex.
7 And so reviewed ten SPRAs. I think I told 8 you 20, but it was a mistake. It was only ten, and 9 that is the CDF range, so there was a broad range of 10 CDF and the conclusions were those that sometimes 11 there was a difference by a factor of two. But the 12 more typical difference in the bottom line number is 13 30 to 60 percent. But some specific accident 14 sequences could be more. And so the, for several 15 SPRAs there really was not much of a sensitivity of 16 the result to the dependencies.
17 MEMBER MARCH-LEUBA: Remind me again on 18 that first bullet the factor of two, are 30 to 60 19 percent, is that by setting all correlations to one 20 and all correlations to zero on the difference?
21 MR. PIRES: Yes.
22 MEMBER MARCH-LEUBA: Good. Thank you.
23 MR. PIRES: Great. Yes, and the --
24 CHAIRMAN RICCARDELLA: 30 to 60 percent on 25 CDF?
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
55 1 MR. PIRES: On CDF. But on particular 2 sequence is, and this where we talk about the 3 insights, is the factor was high and it could be two 4 to four.
5 MEMBER MARCH-LEUBA: Well, there are 6 sequences that don't really contribute that much to 7 the final numbers, is that correct?
8 MR. PIRES: Or there may be more than one 9 sequence that contributes. There may be three or four 10 sequences that contribute and the others are not so 11 sensitive.
12 The cases normally where the dependencies 13 were not significant, or the cases where there was a 14 sequence that controlled the CDF and one component on 15 that sequence. But they call it a single --
16 apparently that's the term they use. And those 17 happen. It's --
18 MEMBER MARCH-LEUBA: Can you give us an 19 example? Where will be the single component that is 20 the dominant?
21 MR. PIRES: I can't tell you an example 22 from my sequence --
23 MEMBER MARCH-LEUBA: From the many plans?
24 MR. PIRES: Yes, from a pilot study that 25 EPRI did. I think in that plan that there were some NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
56 1 sump pumps, there were actually several, but they all 2 are damaged by the collapse of the building. And so, 3 and those apparently were the main factor from --
4 MEMBER MARCH-LEUBA: So the collapse of 5 the building is the single --
6 MR. PIRES: Yes, and damages all of them.
7 That's the assumption. We already know by the way, 8 that buildings not all collapse entirely but that 9 there are parts of a building may collapse, others 10 not, but that was a refinement that was not used and 11 is not really used yet and I think that was one of the 12 cases. But I may be wrong but that's my recollection.
13 MEMBER REMPE: So --
14 MEMBER MARCH-LEUBA: I think it's a work 15 example.
16 MR. PIRES: Yes.
17 MEMBER REMPE: Years ago, I used to work 18 for a company that had a paper reactor design. And so 19 when they did their seismic study for the paper 20 reactor design, they could get the release because 21 they would just assume, look for the release of the 22 sequence of with all those components failed.
23 Like in the Zion PRA, it seems like one 24 could go back to the nons, the regular PRA and get 25 those releases for those sequences that were deemed to NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
57 1 be important here. If it's only a few sequences that 2 are important, I think you could get some insights on 3 is this important to risk from the information that is 4 available.
5 And so I guess I'm still pushing. I think 6 it should be explored. Are these things really 7 important to risk if it's just 10 PRAs and a few 8 sequences in those PRAs?
9 Is, I don't know how, I've seen some of 10 the Zion one and some of these PRAs that seems like 11 one could find some of that information.
12 MR. PIRES: Yes. I think you could find 13 a nice, I'm pretty sure the authors of this study will 14 know some of that information because they deal with 15 these issues predominately and so they will know some 16 of that.
17 And the, a lot of these, the benefits of 18 some of these may be more for new designs because what 19 already exist there is we don't have much in terms of 20 what to change.
21 The, so one of the interesting parts of 22 those reviews was that it looked like there was some 23 set of components that tend to keep appearing on as 24 being critical and some are relevant for these 25 dependencies.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
58 1 And they identified in the study, 2 identified if you are listing them there. I don't 3 want to go in more detail other than to say that they, 4 the workshops also questioned the participants to 5 force them to say what are the aspects of those 6 components that really may make then really be more or 7 less significant in terms of dependencies. For 8 instance, the tanks, they tend to be fabricated 9 locally so there are some variations actually.
10 So in that regard, in terms of the 11 capacity, there may be some independence. But when 12 you have some things fabricated in a factory and put 13 together there may be more correlation between them.
14 However, the anchors are done locally so that's more 15 uncertainty.
16 So but they're at least collecting some 17 input from experts there. Now the review of the 18 literature like I said before, all the methods were 19 reviewed. I'm not going to go here over all the 20 methods. I don't think there is a point in doing 21 that. I just like to provide a broad scope.
22 Of course the more rigorous one was this 23 SMRP approach in which equity joint distributions were 24 done for the capacity and the demands for components 25 in a cut-set or in a Boolean expression and those were NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
59 1 actually numerically integrated over the failure 2 region and they were used in design PRA.
3 One of the benefits of that was that one 4 of those authors of those studies, Mr. Bohn and Mr.
5 Lambright, both of them, but Bohn seems to have been 6 the name which will be associated with it, came up 7 with rules that people use and that were simpler.
8 And but those rules were even deemed too 9 difficult and too complex so they're further 10 simplified.
11 I just show you the rules here. You can 12 see that in some points these rules, these 13 correlations coefficients of .05 and .75, those were 14 abandoned in the current practice. People went back 15 to minus zero.
16 MEMBER SUNSERI: This is a more complex 17 set of --
18 MR. PIRES: A more complex set of rules on 19 rules. But even this complex set of rules is not used 20 in practice.
21 MEMBER CORRADINI: Because it's too 22 complex?
23 MR. PIRES: And to that point I think was 24 because probably the effort needed did not justify the 25 intended use of the results, of the PRA, I think.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
60 1 But now that of course as the odds have 2 increased and so forth, so in some cases it might be 3 more important to do, to at least bring some of those 4 aspects into place.
5 So when they study and review of all the 6 approaches, they come up that these older approach by 7 Mr. Reed and McCann -- by both of them, mostly Reed, 8 you won't find McCann on these there -- that this 9 approach based on what I described before, it goes 10 back to all the fragility and models done by the 11 component.
12 All those factors are isolated and get out 13 from there, because some of those factors, mostly you 14 have some engineering meaning and there's an 15 understanding of those, you could separate what could 16 be correlated from that, but there is still 17 subjectivity. There is still room for some judgment 18 there, so.
19 But and it's not very difficult because 20 you essentially are going to really, you already have 21 that information from the fragility allowed for the 22 components, and you can program the process that is 23 there and implement. So there is an advantage.
24 In my opinion, of course, because there is 25 some room from subjectivity there, it is the more you NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
61 1 break down in the basic components and then work up 2 from those, the better. The other is that the method 3 is simplified so may warrant the comparison with more 4 rigorous methods which exist, but they are very 5 complicated but are used in practice. And that is not 6 really the object of major research. That's just a 7 graduate student master's key system.
8 The other part was review of existing data 9 and I guess initially the researchers taught that 10 there could be, it would be possible to get data so 11 that the approach would be a more data-driven approach 12 but it is not possible. The data is just not there 13 and it would be very extensive.
14 MEMBER CORRADINI: Because it doesn't 15 exist or because it's too complicated to analyze?
16 MR. PIRES: Both. No, the aspect 17 experience data is too complicated to analyze. There 18 are a lot of uncertainties on those. And so far, it 19 has been proven very difficult. You know, the access 20 to the data plus the uncertainties.
21 MEMBER CORRADINI: Is there other 22 industries that care enough about seismic response 23 that this approach has been taken or a simpler 24 approach has been taken?
25 MR. PIRES: Yes. There is a, actually, in NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
62 1 our work, there is an approach that relies a lot on --
2 experience data.
3 MEMBER CORRADINI: Say again, please?
4 MR. PIRES: In the nuclear power plants, 5 I don't know if you ever heard about seismic 6 liquefaction, soil liquefaction --
7 MEMBER CORRADINI: Yes.
8 MR. PIRES: -- during an earthquake.
9 Ultimately, the approaches for soil liquefaction are 10 experience-based. There is always even, because it is 11 your technical problem involves soils and so forth.
12 The laboratory samples are typically very small.
13 There is a large effort to connect to an 14 experience data. And the uncertainty still is large 15 and disparity of many disagreements on authoring that 16 data, but it has become necessary to do that. So 17 that's a success story on use of earthquake experience 18 data.
19 MEMBER CORRADINI: What I guess I'm trying 20 to get is, is there, I'm trying to connect back to 21 your multifactor formula --
22 MR. PIRES: Okay.
23 MEMBER CORRADINI: -- which is some sort 24 of benchmarking of the approach that might be used in 25 difference to the heuristic rules that are currently NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
63 1 used.
2 MR. PIRES: Yes. I think the benchmark, 3 it probably would have to be done with shake table 4 testing data.
5 CHAIRMAN RICCARDELLA: I think the problem 6 with shake table testing though is that in equipment 7 qualification, you don't shake it to failure. You 8 just shake it up until a certain level. If it doesn't 9 fail, it passes and you're done. So you really don't 10 get any data about correlation or even about the 11 failure level of a piece of equipment, right?
12 MR. PIRES: I think that's one of the 13 issues of the test data on shake table. The more the 14 fragility analysis is amenable to analytical approach 15 is the easier it is to use this methodology. But 16 equipment, sometimes that not so easy to do.
17 CHAIRMAN RICCARDELLA: Yes, oh yes.
18 MR. PIRES: You know, for some equipment 19 changes you can do analysis, but for other equipment 20 you cannot do that. One possibility would be to, 21 there are very large shaking tables of detail that 22 maybe while there you could put multiple components on 23 the shake table, orient them in different directions 24 but you see that they have collect data. But that's 25 --
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
64 1 CHAIRMAN RICCARDELLA: You still have to 2 shake them to failure, you're not --
3 MR. PIRES: Right, you still have to shake 4 that to --
5 CHAIRMAN RICCARDELLA: You still have to 6 shake them up to a certain level and nothing fails, 7 you don't learn anything, right?
8 MR. PIRES: Yes, you will need to shake 9 them to failure until probably all of them are failed 10 in the table. And that would have correlated ground 11 motions but then you should isolate the capacity part 12 though because you --
13 CHAIRMAN RICCARDELLA: Yes, yes.
14 MR. PIRES: But to do that, you have to 15 have adjudication in that is it necessary to do that 16 for the use intended? But that will be, for some of 17 these aspect experience, in some cases you can find 18 the information but is not common.
19 MEMBER CORRADINI: So let me ask the 20 question from a justification standpoint. So I 21 thought your slide number, maybe I misunderstood the 22 conclusion you had in the slide, which was Slide 19, 23 which is it made a 30 to 60 percent difference on CDF.
24 MR. PIRES: Well, it --
25 MEMBER CORRADINI: Typical difference in NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
65 1 the range of 30 to 60 percent.
2 MR. PIRES: Yes.
3 MEMBER CORRADINI: So is that enough of a 4 difference to warrant this, excuse my English, complex 5 analysis approach?
6 MR. PIRES: It depends on the use of that 7 result. I think it, I, you know, if you start going 8 in terms of the total risk or are you just separating 9 the risk for each seismic cause, simple seismic 10 causes?
11 MEMBER CORRADINI: That last part, I'm 12 sorry?
13 MR. PIRES: Suppose you are interested on 14 the total risk for the facility, not just from the 15 seismic concept which I understand could be between 16 five to 50 percent of the total risk in terms of CDF.
17 But now you start bringing the fire, you start 18 bringing the internal events, and higher range, and 19 combining all the hazards together. Then you get that 20 risk that you may need to see where can you obtain a 21 more accurate results and so that you can open risk 22 inside safety insights --
23 MEMBER CORRADINI: Okay.
24 MR. PIRES: -- and to address that. So in 25 those situations I can possibly see that that could NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
66 1 happen for an operating plant but I've not seen any 2 case like that.
3 MEMBER MARCH-LEUBA: This is not unusual 4 for this application. We see this in the nuclear 5 industry, especially on -- an example I always mention 6 is over pressure events where you analyze how close 7 you are from the level C. All plants are like two PSI 8 from it because this started first with a cheap 9 analysis. You ran it, good. A cheap conservative 10 analysis. You run with it you stop.
11 If you fail, you go to a more 12 sophisticated analysis. You fail that one, you go to 13 even more sophisticated analysis until you pass. And 14 all of them are perfectly good analysis so this is a 15 similar thing. So this will be only applicable for 16 plants that are close to a limit and where 30 percent 17 makes a difference.
18 MR. PIRES: Yes. It, and typically when, 19 Mehdi is here, he can correct me. I went to a meeting 20 recently in which there was an issue and obviously the 21 hazard is low and the plant was designed, the job is 22 design it for a minimum capacity. In that case you 23 probably don't need to do these things. The other is 24 very high and like in the, that case of Diablo Canyon, 25 all of the studies were done in some level of rigor.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
67 1 If some of those plants closer to that 2 higher hazard situation in which the original design 3 might have missed the hazard, that would have more 4 critical. Otherwise it's new designs.
5 MEMBER MARCH-LEUBA: And my personal 6 philosophy, preference if you want to, and so for 7 spending the money doing an analysis, going forth with 8 some of those components and make them better. Spend 9 the same money and Well, just spend the money on 10 paperwork or actually reinforced beams, put some 11 reinforced beams.
12 MR. PIRES: Well, what the analyst can't 13 tell you is what each wants to do that.
14 MEMBER MARCH-LEUBA: That is where the 15 value is.
16 MR. PIRES: Yes, that would be the value 17 because it's not expected that there will be many if 18 that case would happen. I've not seen any case but 19 the analyst could tell you which one of those.
20 MEMBER MARCH-LEUBA: You know there is 21 going to be cheaper to bring forth a couple of 22 components, rotate them 90 degrees or whatever, than 23 doing all the paperwork.
24 MR. PIRES: In the end, you will end up 25 with capital investments that can do that on some NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
68 1 components and I think these analyst can identify 2 those.
3 MEMBER MARCH-LEUBA: And while it would on 4 this, NRC will not reward you for doing the right 5 thing, just reinforcing those components. There are 6 so many questions about it, whereas if you come up 7 with the paperwork analysis, they'll let you go 8 through.
9 MR. PIRES: But I have seen that in some 10 cases, people somehow identified what risk contributed 11 but maybe possible single tons, I don't, can't be 12 specific but that somehow might have led to some 13 upgrade.
14 MEMBER MARCH-LEUBA: Yes. As an example, 15 I always say that if I am operating a plant, I will 16 mount two diesel generators. One in the basement and 17 one on the fifth floor. The fifth floor will fail in 18 an earthquake, the basement will fail on a flood but 19 at least I'll have one left.
20 MR. PIRES: Right.
21 MEMBER REMPE: So if you were going to use 22 this for insights on which components to reinforce, I 23 would then say you need better data. So if you go 24 back to Slide 24, you talk about that the data from 25 actual earthquake experience in the plants is sparse NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
69 1 because people are only reporting failures, or a few 2 failures.
3 I was wondering if one would take a 4 different attitude. Maybe you ought to also be 5 documenting which components survive as well as the 6 ones who fail.
7 I think, again, I'm more into, yes, I'd 8 like to upgrade the plant but also before I go and do 9 more on data or on that modeling, I would really focus 10 on what could you learn from the real world before I 11 get a model. And so it seems like you could do 12 something that way.
13 MR. PIRES: That is done. The aspects 14 that go and collect aspect experience, they collect 15 both types. Do you remember that meeting in North 16 Carolina about a year ago and in which there was a 17 person proposing more collection of earthquake 18 experience data and more analysis of that? And he 19 emphasized and he has done that in practice for 20 industrial facilities like fossil power plants and the 21 others, and they collect both data. They collect data 22 that survived and data that did not survive.
23 MEMBER REMPE: Yes.
24 MR. PIRES: Most types of data are 25 collected. And but I think in this case, the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
70 1 reference to the, to where you might want to do some 2 capital investments is for those cases where you could 3 see that that would have a big payoff to these 4 sensitive analysis.
5 MEMBER REMPE: The sensitivity with the 6 old current models. I agree on that one, but.
7 MR. PIRES: Or even with even with 8 improved models, it's, but probably that's of course 9 my opinion.
10 MEMBER REMPE: We all have opinions, I 11 guess.
12 (Laughter.)
13 MR. PIRES: Right. But probably more 14 value of these and I move the, if I may, move to the 15 other slide. This is how the author recommended that 16 all these would be used so they would still start to 17 be the, perform SPR using the standard methods, enough 18 sensitivity studies so that you can see what, where 19 you may need to make refinements, then the 20 identification of those cut-sets or sequences where 21 you would want to make this refinement of the 22 methodologies.
23 So these are systems analysts there in 24 red. The one in blue, that would be now back to the 25 fragility analyst that would then go use these methods NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
71 1 to revise the fragility calculation for those cut-sets 2 to obtain numbers that was the methodology that is 3 referred to here.
4 And then we'll go back to the SPRA analyst 5 and do the work. So that would be a process that 6 could be done.
7 MEMBER MARCH-LEUBA: So if you only 8 propose to do this analysis for the two or three 9 cut-sets that make a difference, it's not that 10 expensive.
11 MR. PIRES: No. But you need to have a 12 good analyst.
13 MEMBER MARCH-LEUBA: Yes, well, that's 14 where the big bucks are, right?
15 MR. PIRES: Right, yes, maybe. And so you 16 have here that, so there are difficulties here with 17 the analyst and for this revised fragilities that 18 appear on the blue box and it is, there can be 19 variability among fragility analysts and the report 20 mentions that.
21 (Off microphone comment.)
22 MR. PIRES: And there is still some level 23 of expert judgment. I think by breaking down on the 24 basic variables like the fragility on all these tests, 25 you reduce some of that judgment quite significantly NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
72 1 I think. Because it's more transparent to the process 2 where you get to the breakdown of the uncertainties 3 and so.
4 (Off microphone comment.)
5 MEMBER CHU: The airplane mode works.
6 CHAIRMAN RICCARDELLA: Airplane mode 7 works, yes. I never said, that's not me.
8 MR. PIRES: And so this is, let me see, so 9 I don't know if somebody remarked that essentially 10 it's just a summary of what we have been discussing 11 about it. I don't know if you want to go through them 12 again.
13 But one item that I would like to say 14 there, is that this, we think that with this study we 15 can maintain a dialogue with the industry. If the 16 industry's interested and go into other, you know, 17 more refined methodologies. We already, it's 18 essentially our part and now the industry probably, 19 they are interested. They can either pursue this part 20 or others because we already did some staffing say 21 that there is some things that may have promise.
22 Now with the, you ask that the, so the 23 treatment of dependencies always depends on the 24 intended results. So we are not implying here that 25 what is being done now is not appropriate or correct.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
73 1 And part of the demonstrations of 2 methodology would be necessary like you said. It's 3 just the start of a dialogue and there can be arising 4 needs. And like you mentioned, the Chairman mentioned 5 at the beginning of the meeting, maybe modular 6 reactors or advanced reactors may want to consider 7 these at the early design stage.
8 I understand it's not being done now 9 because the design certification process is, there is 10 not a lot of definition yet but that is a choice of 11 the applicant.
12 And so we want some of you prepared if the 13 applicants go that way. They, our knowledge can 14 respond to that and now understand it.
15 MEMBER CORRADINI: Or they could just make 16 a more robust structure so that the residual risk is 17 small enough you don't care.
18 MR. PIRES: Right. But that is there to 19 be their choice.
20 MEMBER CORRADINI: Yes, okay.
21 MR. PIRES: And but the point is, one I 22 can hear is that you may keep adding more components 23 and not really have much benefit in terms of the 24 redundancy that it will bring up.
25 So you can from the very get-go design NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
74 1 that to get that redundancy so that they will not have 2 a dependent failure.
3 MEMBER SKILLMAN: Two or three times 4 around the table, members have mentioned that maybe 5 the ticket here is to make a robust structure. And 6 I've been involved in Europe and in the United States 7 where the exact opposite was the right thing to do.
8 We actually weakened the structure. We 9 made sure that the earthquake went under the structure 10 and the structure didn't respond and so it can go 11 either way.
12 And it boils down, at least in my mind, 13 this whole thing boils down to a probabilistic risk 14 assessment that identifies those structures and 15 systems and components that are credited for success 16 for the specific accident.
17 In one case, it might be the batteries.
18 In another case it might be high pressure injection 19 pump. In another case it might be the high head 20 safety injection pump in a boiling water reactor. So 21 it's scenario specific, it's plant design specific, 22 and it is condition specific.
23 Like Harold said, like I said early on, 24 you got to know the tech specs. You got to know 25 what's credited. You got to know which systems are NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
75 1 aligned for what purpose and when, when the event 2 occurs.
3 That is part of this rubrics to get a path 4 forward that allows examination that is meaningful in 5 terms of what Dr. Rempe said, risk, not just a failure 6 of the device.
7 You know, you can fail the water cooler, 8 you can fail the front door of the administration 9 building, but you don't want to fail the nozzle that 10 puts the water onto the fuel.
11 The risk is associated with the right 12 component, in the right place, at the right time for 13 the particular accident and the particular ground 14 motion. All of those are pieces.
15 CHAIRMAN RICCARDELLA: Core damage 16 frequency can't capture those differences though.
17 MEMBER SKILLMAN: Pardon?
18 CHAIRMAN RICCARDELLA: Core damage 19 frequency can't, I mean, the water cooler and the door 20 to the admin building isn't going to impact core 21 damage frequency.
22 MEMBER SKILLMAN: My only point is, my 23 only point is, it's very component-specific.
24 CHAIRMAN RICCARDELLA: Yes.
25 MEMBER SKILLMAN: Your risk is very, very NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
76 1 component specific.
2 MR. PIRES: I think the least that they 3 showed here on the workshop as a result of the 4 workshop and the review of past PRAs, mostly the 5 components are of course are not buildings. I guess 6 we knew that, they, this is the type of component that 7 typically appeared there as critical.
8 Of course the building's response affects 9 those because the way it amplifies the ground motion.
10 And when people talk about seismic isolation, they 11 could review some of these.
12 MEMBER CORRADINI: So let's talk about 13 seismic. You brought it up. I was going to but you 14 first. Tell me about, I was under the impression 15 there is a potential standard that, a potential 16 approach that's been in front of the staff to look at 17 for seismic isolation?
18 MR. PIRES: No, no. No, no, no. There is 19 a standard --
20 MEMBER CORRADINI: You did it. I didn't 21 do it.
22 MR. PIRES: Right. No, I know.
23 CHAIRMAN RICCARDELLA: There was a DOE 24 research part.
25 MR. PIRES: There is a standard. There is NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
77 1 an American Society of Civil Engineering Standards 2 that has a chapter there on Seismic isolation.
3 And we are not necessarily reviewing that 4 standard for seismic isolation. We are reviewing that 5 standard for many other, in terms of for more risk, 6 and from approach, performance based to the seismic 7 design, but not necessarily that particular part. We 8 have other research projects on seismic isolation 9 already at the Agency here and so we have to look at 10 those.
11 MEMBER CORRADINI: Well, what I guess I'm 12 --I'm sorry, I didn't mean to interrupt you.
13 MR. PIRES: Sure.
14 MEMBER CORRADINI: Go ahead and finish, 15 I'm sorry.
16 MR. PIRES: No, we have other research 17 projects but not necessarily aimed at reviewing a 18 particular standard but aimed at developing technical 19 consideration that should be accounted for when 20 designing seismic isolation systems.
21 MEMBER CORRADINI: Well I guess I'm a bit 22 unsure what you just said. So is there a standard in 23 front of the staff to evaluate or is it still a matter 24 of research?
25 I thought there was a standard in front of NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
78 1 the staff that the staff is evaluating, the suggested 2 standard from the civil engineering society.
3 MR. PIRES: There is a standard but we are 4 not ever tasked to evaluate that standard.
5 MEMBER CORRADINI: We --
6 MR. PIRES: We do not have a task to 7 evaluate that standard, but --
8 MEMBER CORRADINI: So there's not been a 9 request yet?
10 MR. PIRES: Not been a request. Yes, 11 there's not been a request. We are reviewing that 12 standard but that standard has so many other aspects, 13 those are the ones we are doing.
14 But we have independently done research 15 that develop technical considerations for, that should 16 be accounted for if seismic isolation were to be used 17 in a nuclear power plant design. So we have 18 independently done that ourselves.
19 MEMBER CORRADINI: But if tomorrow, if 20 reactor X comes in, a small modular reactor, and says 21 we're going to isolate the safety structures with 22 seismic isolation, is there a standard they can follow 23 that the NRC has approved or at least has reviewed?
24 MR. PIRES: No. There is not a standard 25 that we have taken a staff position on.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
79 1 MEMBER CORRADINI: Is there something that 2 --
3 MR. PIRES: We have done research in 4 relation to that.
5 MEMBER CORRADINI: So there's nothing that 6 they can point to that is yet to be reviewed but could 7 be, or is it a matter of research or is it a matter of 8 working through the standard that is out there but yet 9 to be reviewed and approved? I'm trying to understand 10 where this is.
11 MR. PIRES: Right. All right. Our 12 research project culminated in a NUREG/CR contractor 13 report that does technical considerations for the use 14 of base isolation on nuclear power plants. Those 15 technical considerations are probably better than 16 those in the standard.
17 MEMBER CORRADINI: Right.
18 MR. PIRES: This is of course our opinion.
19 But they are very closely related in many ways. So 20 actually our contractor was the one who brought most 21 up to standard.
22 MEMBER CORRADINI: Okay.
23 MR. PIRES: So maybe I should not say 24 this, but that does not mean that the standard per se 25 is sufficient either. The standard may not be NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
80 1 necessarily sufficient.
2 MEMBER CORRADINI: Okay.
3 MR. PIRES: Because there are several many 4 other details and considerations that are not 5 necessarily covered in the standard by that, neither 6 by that NUREGs.
7 MEMBER CORRADINI: Because that's some --
8 MR. PIRES: It the variety for the 9 details.
10 MEMBER CORRADINI: Well, I mean the reason 11 I'm bringing it up is, one, you did but I was going to 12 anyway, so I can't blame you totally.
13 MR. PIRES: Right.
14 MEMBER CORRADINI: But in other technical 15 meetings there has been discussion that these advanced 16 reactors, whatever they are, are going to try to use 17 this to minimize to allow the structure to be more 18 simply built --
19 MR. PIRES: Right.
20 MEMBER CORRADINI: -- because it has a 21 smaller footprint. And so the next question has been, 22 in some of these technical discussions, well, is there 23 a standard? And the answer back is yes, but the NRC 24 has not reviewed it.
25 That is what I've heard in technical NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
81 1 discussions.
2 MR. PIRES: Right.
3 MEMBER CORRADINI: And I'm trying to 4 understand where the reality is versus what I hear in 5 a --
6 MR. PIRES: The NRC has not reviewed and 7 taken a staff position on that standard. The NRC has 8 done research very similar to the research that 9 supports the standard. So we are not at zero. We 10 have a knowledge basis in Agency on that topic. So we 11 can come --
12 MEMBER CORRADINI: But nobody has come to 13 you to say we want to use this, can we use this?
14 That's what I'm hopeful for.
15 MR. PIRES: Right.
16 MEMBER CORRADINI: Eventually.
17 MR. PIRES: We are not at that situation, 18 yet. But we have done the research --
19 MEMBER CORRADINI: Okay.
20 CHAIRMAN RICCARDELLA: Yes, good.
21 MR. PIRES: -- that can take to that 22 situation if that materializes.
23 CHAIRMAN RICCARDELLA: I recall a meeting 24 I attended maybe three years ago at the Department of 25 DOE. Did DOE sponsor a big program like that?
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
82 1 MEMBER CORRADINI: Yes, there was a study 2 at DOE.
3 CHAIRMAN RICCARDELLA: You were there, 4 weren't you, Jose?
5 MR. PIRES: Yes, DOE does have recent 6 projects on that.
7 (Simultaneous speaking.)
8 CHAIRMAN RICCARDELLA: On isolation.
9 MEMBER CORRADINI: University of Buffalo.
10 There is a large group at University of Buffalo.
11 That's where I had the presentation and I --
12 MR. PIRES: He was our contact there for 13 that project.
14 CHAIRMAN RICCARDELLA: Your contact.
15 MR. PIRES: For that same group.
16 MEMBER CORRADINI: Same group?
17 MR. PIRES: Yes. So we are up to date on 18 what is happening in that world and obviously where if 19 someone is going to submit an application, there are 20 many, many more details than just the standard.
21 CHAIRMAN RICCARDELLA: Yes.
22 MR. PIRES: But we are up to date but we 23 have not developed necessarily a staff position.
24 That's, but we are up to date.
25 MEMBER CORRADINI: Okay.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
83 1 MR. PIRES: We are not, which I think we 2 are in a good position to be responsive.
3 MEMBER CORRADINI: Thanks for bringing it 4 up.
5 MR. PIRES: You are welcome, yes.
6 CHAIRMAN RICCARDELLA: I have a couple of 7 questions. I know there's these ongoing Task 2.1 8 seismic reviews of the CS plants and as I understood 9 it, about a third of the plants out there had to take 10 a new look at their seismic because they exceeded 11 their current ground motion response factor, right?
12 MR. REISI-FARD: That's about the right 13 number.
14 CHAIRMAN RICCARDELLA: Are those plants 15 having problems? I mean they're doing SPRAs or 16 seismic margins analyses. Are they having, are there 17 any plants that have issues, are having issues 18 satisfying the criteria?
19 MR. REISI-FARD: Not to my knowledge and 20 this goes back to some of the discussions earlier 21 about the level of detail or analysis that is needed 22 to do the SPRAs.
23 For the purpose of this informed decision 24 making, there is the SPRA standard out there which is 25 being peer reviewed by, you know, there's a process NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
84 1 for doing the peer review.
2 And they have been using, the industry, 3 the licensees have been using these methods for a long 4 time and, you know, it's been, their PRAs have been 5 peer reviewed against the standard and, you know, it's 6 just a matter of uncertainty, key assumption 7 uncertainty in the model. There are no, typically 8 there are not many findings associated with how the 9 correlation is considered. They either assume full 10 correlation or no correlation.
11 CHAIRMAN RICCARDELLA: Yes. Yes, I mean 12 I would think kind of like what Jose said earlier, you 13 only sharpen your pencil as sharp as it needs to be.
14 And so if none of these plants are having 15 any problem meeting their, whatever the criteria are 16 for their SPRAs then they wouldn't necessarily go back 17 and relook at this. But if they do have an analysis 18 or two that's giving them problems, maybe they might 19 try to do it.
20 MR. REISI-FARD: And I know that's out of 21 a few or several SPRAs that we've seen, there is only 22 one exception that they've used a little bit more 23 refined analysis. It's not using .05 or .75 or 24 anything like that.
25 But because the licensee decided to have NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
85 1 a look, it's more detailed final element analysis they 2 have a better understanding of the seismic demand in 3 different locations. So if there was a different 4 seismic demand, they use different correlations. But 5 other than that, they basically assume zero or 6 complete correlation.
7 CHAIRMAN RICCARDELLA: Yes. So and then 8 another potential application as I said earlier, is 9 the NuScale or some new reactors where if you haven't 10 completed the design yet, it seems like you could take 11 some of this into consideration in advance and have it 12 affect your design.
13 But, Vesna, the NuScale PRA that we'll be 14 reviewing, does that include a seismic PRA?
15 (Off microphone comment.)
16 MEMBER CORRADINI: I thought it was a 17 seismic --
18 CHAIRMAN RICCARDELLA: It concentrates on 19 a --
20 MEMBER CORRADINI: -- margin analysis. I 21 didn't think it was the seismic --
22 MEMBER DIMITRIJEVIC: Yes. I mean, 23 seismic analysis -- seismic risk analysis is a new 24 type. I think it's a seismic matching but it's 25 irrelevant because the NuScale is submitted analysis NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
86 1 based on the single unit.
2 CHAIRMAN RICCARDELLA: So --
3 MEMBER DIMITRIJEVIC: So it's irrelevant, 4 do they have a ten, 12, or 20 units. They just 5 analyze in single units. That's what their agreement 6 with NRC, so we don't, we may see in consequence, but 7 the consequence in this case means that, you know, 8 Level 3 analyzes that some combination. But on the 9 PRA, no, it's just a single use.
10 MR. PIRES: That's what I heard too. That 11 currently it's a single unit and the seismic margin is 12 what they call a PRA based seismic margin.
13 MEMBER DIMITRIJEVIC: Right.
14 MR. PIRES: But because that seismic 15 margin approach and that the guidance really involves 16 simplified methods, min-max approaches and that there.
17 So it's not really --
18 (Off microphone comment.)
19 MR. PIRES: But can be used that, but is 20 a simplified PRA based seismic margin and --
21 CHAIRMAN RICCARDELLA: But if they were to 22 come in and try to license multiple plants at a site, 23 wouldn't it be necessary for their PRA to address a 24 multi-unit site?
25 MEMBER DIMITRIJEVIC: That's my opinion NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
87 1 but I jump on this train late and disagreements had 2 already reached, I mean, with the staff. So, I mean, 3 I will bring this question to the review because it's 4 not just seismic, all other external ones, you know?
5 CHAIRMAN RICCARDELLA: Yes, yes, of 6 course.
7 MEMBER DIMITRIJEVIC: You know, like 8 external floods and which is one of the main important 9 contributions. But that's what is put into that so.
10 They have a multi-unit on the license but that 11 actually involves the analysis from ability that two 12 units fails. That didn't include the seismic common 13 cause, it just includes the single, the single 14 non-seismic common cause between each. So we will 15 have a discussion about that but that's --
16 CHAIRMAN RICCARDELLA: Okay.
17 MEMBER DIMITRIJEVIC: It doesn't say if 18 you have a challenge, what's the probability of one of 19 12 things with that.
20 CHAIRMAN RICCARDELLA: Yes. That's coming 21 up this month, right?
22 MEMBER DIMITRIJEVIC: Yes, and this 23 analysis is happening in the next --
24 MEMBER CORRADINI: The first meeting will 25 be this month.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
88 1 MEMBER DIMITRIJEVIC: -- yes, yes.
2 MEMBER CORRADINI: It won't be the last 3 meeting.
4 MEMBER DIMITRIJEVIC: No. Right, so.
5 CHAIRMAN RICCARDELLA: Okay.
6 MEMBER DIMITRIJEVIC: Well, and I have a 7 microphone on. I have one question. In your report 8 there was some example identified with three service 9 water pumps show as important.
10 You know, when we have a seismic, we 11 always assume that we have a loss of onsite power as 12 it normally is out and then these generators as you 13 mentioned are important, but, and but they cool the 14 service water so service water is important.
15 And usually single structures or tanks 16 like lube oil tank can dominate what is used at the 17 generator if you don't know correlation, or intake 18 structure when it's the pumps itself.
19 But there was an example reported in that 20 report of the sump cuts, actually the three service 21 water pumps.
22 CHAIRMAN RICCARDELLA: Yes.
23 MEMBER DIMITRIJEVIC: The three service 24 water pumps show it's very important in that site meet 25 PRA but then they didn't proceed with how did they NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
89 1 solve that issue?
2 MR. PIRES: The one I referred to was a 3 pilot study so it was not really tied into any 4 particular licensing issue or not.
5 (Simultaneous speaking.)
6 MEMBER DIMITRIJEVIC: I see.
7 MR. PIRES: It was a pilot study.
8 MEMBER DIMITRIJEVIC: It could be, yes, is 9 the early ones when done and then they tried it, yes.
10 I understand this.
11 MR. PIRES: So it was a little different, 12 the use. Well, in principle this methodology can be 13 useful modular reactors and but again it also depends 14 on the problem statement. What would it be used for?
15 What would be the intent to go all the way to 16 releases, to one model fails, all fails, do they fail 17 simultaneously? And do they need safety systems from 18 another? Are they relying on safety systems from 19 another core module to stop an accident progression on 20 one of their modules and on their other model?
21 So I think depending on the problem 22 statement is done, is what one they have to look, but 23 in principle can be done. And if you had the analysis 24 for the single model you can extrapolate it by the 25 correlations using this approach. But it has to pay NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
90 1 off, has to have a benefit for the applicant.
2 MEMBER REMPE: So when I look at the last 3 bullet here I also go back to Page 26 and see all of 4 the caveats you have about before you can be using 5 this methodology and it seems like that there's a lot 6 of work still to be done if one wanted to apply and 7 use this as part of the standard tool kit available to 8 seismic people. And I just kind of wanted to see you 9 shake your head up and down, yes.
10 MR. PIRES: I think so. There is, there 11 will be substantial work yet, but the question is at 12 this point, like I think I mentioned before, I think 13 we have done probably our part on the dialogue.
14 MEMBER REMPE: Yes.
15 MR. PIRES: And so maybe now someone else 16 has to get that additional work and maybe we, put 17 these points to some other organization that does 18 research, that can be cooperative research but maybe 19 will have a role. But if I think we only need to have 20 now the initiative coming from outside.
21 MEMBER MARCH-LEUBA: This gets me thinking 22 to the topic I wanted to change the conversation just 23 as we have plenty of time and I know you've done all 24 this because you've told me on the side.
25 MR. PIRES: Yes.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
91 1 MEMBER MARCH-LEUBA: It's still bit of 2 certainties.
3 MR. PIRES: Okay.
4 MEMBER MARCH-LEUBA: You told us that all 5 of those F parameters on the long equation have their 6 uncertainty but now you're splitting that uncertainty 7 into two.
8 MR. PIRES: Right.
9 MEMBER MARCH-LEUBA: Yes. And the estate 10 of the practice now is to split the CO1 by setting the 11 correlation to CO1 and that term means what the 12 uncertainty is. I know you've given some thought on 13 how you would do the uncertainty with the proposed 14 method. Can you talk a little bit about that?
15 MR. PIRES: How we would split this?
16 MEMBER MARCH-LEUBA: Yes. How would you, 17 I mean, what is your, what was your consideration for 18 treatment of uncertainties, the additional uncertainty 19 that this introduce by the new method?
20 MR. PIRES: Oh, alright, yes. I think one 21 of the bigger uncertainties to this, by the new method 22 is this step that you just mentioned. How you are 23 going to assign what is correlated and what is 24 uncorrelated, particularly given that you don't have 25 hard data.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
92 1 And so but the study really did not look 2 at that into detail because the study at this moment, 3 they just, all the examples they have, they assign it 4 the information to those things on the top line 5 directly. They didn't even use the bottom line.
6 They just have statements. They are 7 saying the -- interaction part of the building is not 8 the same so if all components are in the same 9 building, the -- interaction uncertainty probably is 10 correlated.
11 But they have not really provided the 12 metallic aspect how to go from the bottom line to the 13 top line in assigning those uncertainties. So but I 14 think that needs to be that. And if that, because of 15 so many factors, they had all of these factors, they 16 have many quality analysts, can eliminate much of the 17 judgment, particularly those that can be done by 18 analysis.
19 MEMBER MARCH-LEUBA: You know the field 20 which was a very large study, 1980s probably, was the 21 CSAU, Code Scaling, Applicability and Uncertainty 22 Analysis.
23 MR. PIRES: Okay.
24 MEMBER MARCH-LEUBA: Anyway, called CSAU.
25 The first step on CSAU is to provide a PIRT, Phenomena NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
93 1 Identification and Ranking Table, okay.
2 MR. PIRES: Right.
3 MEMBER MARCH-LEUBA: And all the thermal-4 hydraulics which are more of a deterministic instead 5 of a probabilistic -- understood that because there 6 will be, that you will get a different number, that he 7 will get a different number.
8 So the PIRT always requests to have at 9 least three experts in the field. And then you look 10 at the variability between different opinions.
11 When you do the separation between ER and 12 EU, probably will require at least a second opinion 13 instead of just one guy.
14 MR. PIRES: Did, one of the things and the 15 report mentions that they really recommend a peer 16 review even if it is informal. Even if that is not a 17 formal peer review that is required by a regulator or, 18 they really recommend that you should have that.
19 But I think even before you get the peer 20 review, you probably need the guidance on how to do 21 that, based on this first step, it's to have that 22 guidance and second is to have a peer review.
23 MEMBER MARCH-LEUBA: Yes, but in my mind, 24 and maybe I'm biased, but I think the largest source 25 of uncertainty will be a subjective evaluation of NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
94 1 numbers and that Jose will give a different result 2 than Josie.
3 MR. PIRES: Right.
4 MEMBER MARCH-LEUBA: And it will be, and 5 I thought that you addressed that in the report by 6 recommending a peer review or list, you get an idea 7 what the uncertainty is.
8 MR. PIRES: Right. And the report 9 recognizes that, recognizes that there will be 10 variability among the analysts.
11 MEMBER DIMITRIJEVIC: And do you know what 12 also is problem, there is not two. There don't exist 13 two fragility experts in that. You know, an agility 14 expert is very sparse. Like ten years ago all world 15 was using just one. I mean, I don't know what's the 16 situation is now but then --
17 MR. PIRES: No, there are a few.
18 MEMBER DIMITRIJEVIC: We're not going to 19 find many too complex opinions.
20 MR. PIRES: There are a few and I think 21 the breaking down of the problem allows that. In my 22 opinion, the more you can do to some analysis that are 23 having to rely directly on the use of data so you have 24 a matching off the two, you are better.
25 MEMBER MARCH-LEUBA: Okay. Well, my point NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
95 1 is in the report you've given some thought to it and 2 you have addressed the issue, that this is a problem.
3 MR. PIRES: Right. It is an issue.
4 CHAIRMAN RICCARDELLA: You know, on this 5 slide you've got two different things where you split 6 the uncertainty. In the top line it's random versus 7 uncertainty. But to me, what's equally significant on 8 the very bottom line, the first part of the equation 9 there, it's F sub C and F sub RS and do each one of 10 those have an R and a U?
11 MR. PIRES: Yes. And then each one of the 12 other F's has an R and a U. So when you, in the end 13 you can have a square root of some squares that are 14 about 15 times there. So it can be complex.
15 But the analysts are used to do that.
16 They go to their training. Typically what happens is 17 if a utility wants to do an SPRA, what I understand 18 some of them did is that they contract a firm that 19 knows how to do these and they do an example for a 20 building and the components in there.
21 They go through all this process. They 22 define templates, and then they can have other 23 engineers that go through that process systematically 24 for other components and other buildings and other 25 aspects.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
96 1 So they have a process to do that to 2 minimize errors and subjectivity. But it still can 3 happen because the next step is to assign what part 4 of, if you have two components now of those all their 5 Rs and Us, what is the part that is correlated, the 6 part of a covariance is still just a variance for each 7 one?
8 And that part is somewhat difficult but 9 there are reasons and arguments when you break down to 10 be able to move in that direction, and it partly 11 depends on the importance of the problem and what you 12 want to do with the results.
13 CHAIRMAN RICCARDELLA: Yes. It seems that 14 there's an opportunity that if you're designing a new 15 plant for a new concept, like these modular reactors, 16 that there might be some very, very minor things that 17 you could do at relatively low cost that would make 18 things less correlated.
19 MEMBER CORRADINI: But you could also 20 increase the risk of something else. You can't just 21 change this, you would have to redo everything. If I 22 took all of these things and orient them ten degrees 23 different, then my fuel handling might change --
24 CHAIRMAN RICCARDELLA: Yes, right, right.
25 MEMBER CORRADINI: -- so that I can NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
97 1 increase the risk of a fuel handling accident.
2 MR. PIRES: Yes. In some of these 3 aspects, you probably, you know, one exercise a day, 4 so I'm just thinking aloud without really --
5 But ideally it would be to have a set of 6 rules coming out from PRA studies and those set of 7 rules, then give indications if you want to obtain 8 redundancy, you need to do these things in your design 9 so that the design not necessarily have to, or define 10 the rules that, and then, but and those would be 11 exercise PRAs to come up with those rules.
12 MEMBER CORRADINI: Yes, but I think but in 13 --
14 MEMBER MARCH-LEUBA: For in --
15 MEMBER CORRADINI: If I might just say, I 16 think your Slide 25 though is the one that I got --
17 you don't have to go to it -- but the one with your 18 blue and red boxes. You don't do this for everything.
19 You find out where it's critical to do it 20 and then you delve in. I mean, it's like Jose's 21 example of where you want to sharpen your pencil and 22 where it's good enough to do a simple analysis.
23 That's what I took away from that.
24 MEMBER MARCH-LEUBA: What I think that 25 Pete is saying is that when we do fire analysis and do NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
98 1 the INC. We put one division in a different cabinet.
2 We have four different fire enclosures for each 3 division because we only, we put them only in one, we 4 have a fire, we're cooked. Okay.
5 If doing this safety PRN and you find out 6 that all your three sump pumps are on the same weak 7 building, that tells you a problem that you maybe need 8 two sump pumps in different places. And that points 9 you to small changes that maybe are not expensive that 10 make the plant better. Remove the single pump.
11 MR. PIRES: Yes, like you said, that box 12 already there in the parenthesis, but presumably 13 should be few of those where you have to do that and 14 have a benefit.
15 CHAIRMAN RICCARDELLA: But I guess the 16 first step would be to make sure that they consider 17 the multiple units in the PRA. If they're not doing 18 that --
19 MEMBER MARCH-LEUBA: Somewhere I've read 20 that they just multiply times 12 the results of one.
21 And that they assume they're fully correlated. That's 22 what I read somewhere but you're --
23 MEMBER CORRADINI: Wait until the October 24 18th meeting.
25 MEMBER MARCH-LEUBA: Yes.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
99 1 MEMBER REMPE: But if you do, I mean, it's 2 what your point was about the ship already, or the 3 train already left. There's a couple of things, again 4 just thinking off the top of your head, not only do 5 you have the issue that you might have three pumps in 6 12 units that all are not considering all 12 units in 7 the common cause, but what about just that you've got 8 12 units in the same pool and a bowling ball approach 9 where they knock out each other?
10 MEMBER SKILLMAN: Joy, we raised that 11 question.
12 MEMBER REMPE: Yes, I know.
13 MEMBER SKILLMAN: We absolutely raised 14 that exact issue.
15 MEMBER REMPE: Yes.
16 MEMBER SKILLMAN: I think we're in a 17 different --
18 (Off microphone comment.)
19 MEMBER REMPE: No.
20 MEMBER CORRADINI: I would suggest we're 21 in a different discussion, that we should wait until 22 we actually have the --
23 CHAIRMAN RICCARDELLA: Yes.
24 MEMBER SKILLMAN: I think in a way we're 25 in a different discussion but we're going after the NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
100 1 comment that Pete made that I think is right on the 2 money.
3 Early on, slight adjustments can provide 4 great risks reduction, or great reductions in risk.
5 I think that that's accurate and I think we ought to 6 be thinking about that. Just a couple of very minor 7 changes that can allow a very significant outcome for 8 an event and I think that's accurate.
9 MR. PIRES: In this case, I think for the 10 multi-module syncs, it was probably important to come 11 up with what is the problem statement that, for the 12 applicant that the applicant wants to address? Is, 13 are they worried about releases or if using systems 14 from another module to save one model? Or, you know, 15 I think that now that the problem statement has 16 expanded because you have all these modules there 17 relatively close to each other, and depending on that 18 then they would have to choose what they want to do.
19 MEMBER MARCH-LEUBA: But I would like to 20 think that the applicant is already doing that. I'm 21 sure whoever designed the plant, they're looking for 22 the low hanging fruit. I mean, if we weren't under 23 the review when it comes to ACRS, nothing can be 24 changed. So what we need is to encourage the 25 applicants to think things this way.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
101 1 MEMBER DIMITRIJEVIC: But this is just to 2 tell you it's also bad because actually they're not in 3 parallel, they're in series. So fail of any unit is, 4 you know, basically the failer. It's, you know, 5 they're not in parallel. So if you look in the 6 component combination, whatever fail that unit, if you 7 want to fail the second unit, or fail only the one 8 unit present the problem. So it's from a, you know 9 what I want try to say, the units are in the series 10 actually when it comes to the core damage frequency, 11 so.
12 MR. PIRES: This may be a case where a 13 release might make a difference because if you ever 14 sourced them for 12 modules versus from one module, 15 but I don't know. It's, I have no idea. It's a 16 problem statement.
17 CHAIRMAN RICCARDELLA: No, we have some 18 food for thought for our meeting in two weeks and we 19 can ask the applicant then.
20 Well, Jose, I thank you very much for 21 coming by and briefing us all and Mehdi, we appreciate 22 it.
23 I'm going to go around the room. But 24 first, is there anybody in the room that would like to 25 make a comment?
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
102 1 Hearing none, we'll go to the phone line.
2 Can we open the line please?
3 MEMBER CORRADINI: We'll wait for a 4 crackling sound.
5 CHAIRMAN RICCARDELLA: Yes. So when the 6 phone line is open, if there's anybody listening to 7 this, in on this call, I'd appreciate it if you would 8 let us know that you're there and you hear us.
9 MEMBER MARCH-LEUBA: I just want to take 10 -- on the line. Let's go around the table.
11 CHAIRMAN RICCARDELLA: Okay. So let's ask 12 the subcommittee members. Vesna, do you, starting 13 with Vesna and we'll go around the table.
14 MEMBER DIMITRIJEVIC: Yes, I'm all right.
15 I don't have additional questions.
16 MEMBER CHU: Nothing.
17 CHAIRMAN RICCARDELLA: Harold?
18 MEMBER RAY: To me, Pete, this is more 19 input rather than a time for us to give feedback. I, 20 you know, I think the issue always in my mind is how 21 to reduce core damage. What's the most efficient way 22 and the most appropriate way to reduce core damage 23 frequency to an acceptable level? And we're making a 24 transition now in how we think about that and I just 25 need to hear more before I can make a comment.
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
103 1 CHAIRMAN RICCARDELLA: Okay. I'm just 2 going to interrupt for a second. I think I heard the 3 crackle. Is there anybody on the line that would like 4 to make a comment?
5 MEMBER MARCH-LEUBA: Nobody phoned in on 6 the line.
7 CHAIRMAN RICCARDELLA: Nobody?
8 MEMBER MARCH-LEUBA: Not that we hear.
9 CHAIRMAN RICCARDELLA: Okay. Okay. Jose, 10 Mehdi, so thank you. That's it.
11 MEMBER CORRADINI: No comments. Thanks to 12 Jose and Mehdi.
13 MEMBER SUNSERI: I find it to be an 14 interesting advancement of the state of practice I 15 guess in this particular area.
16 I suppose my overall feed, input similar 17 to Harold here, is saying that, you know, I think we 18 need to be cautious about how we go about utilizing 19 and or applying those, because my experience is if you 20 start tweaking with hardware to try to make analysts 21 come out in different areas, it's tricky business.
22 And you can cause unintended consequences to solve a 23 problem.
24 (Off microphone comment.)
25 MEMBER SUNSERI: So that's the only NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
104 1 caution I offer at this time. I do appreciate the 2 presentation. Thank you.
3 CHAIRMAN RICCARDELLA: Joy?
4 MEMBER REMPE: I don't have any additional 5 comments but I do want to thank you for taking the 6 time and coming and talking to us.
7 MEMBER MARCH-LEUBA: I want to also 8 emphasize a thank you because that was a very 9 interesting presentation. You see it was very lively 10 and everybody wanted to hear about this.
11 My main concern when these things happen 12 is whether it may be misused. I mean, whenever you 13 give an applicant additional margin, he might come 14 down the real safety of the plan by using additional 15 margins. So that's a warning that let's make sure 16 that this results in safer plants.
17 (Off microphone comment.)
18 CHAIRMAN RICCARDELLA: Okay. Thank you.
19 MEMBER BALLINGER: No more additional 20 comments.
21 CHAIRMAN RICCARDELLA: Okay. Well, with 22 that I'll thank you again and I guess we'll adjourn 23 the meeting.
24 (Whereupon, the above-entitled matter went 25 off the record at 2:54 p.m.)
NEAL R. GROSS COURT REPORTERS AND TRANSCRIBERS 1323 RHODE ISLAND AVE., N.W.
(202) 234-4433 WASHINGTON, D.C. 20005-3701 (202) 234-4433
Correlations of Seismic Performance in Similar SSCs (Structures, Systems and Components) - NUREG/CR-7237 Information Briefing to:
Advisory Committee on Reactor Safeguards Structural Analysis Subcommittee October 3, 2018 Rockville, MD Staff presenters/speakers:
Jose Pires, RES/DE Mehdi Reis-Fard, NRR/DRA/APLB/RILIT Selim Sancaktar, RES/DRA/PRAB 1
Overview
- Background
- Objectives and Goals
- Scope
- Analysis Aspects
- Study Approach
- Review of Existing PRAs
- Review of Current Literature
- Workshops with Experts
- Analysis Process in a SPRA and Issues/Uncertainties
- Summary and Remarks
- Further Questions 2
Background
- Seismic Probabilistic Risk Assessments (SPRAs) have been conducted worldwide for over 35 years
- Two key SPRA aspects I. All possible earthquake levels, their frequencies of occurrence and potential consequential damage should be considered II. Earthquakes can simultaneously damage multiple redundant components
- Risk-quantification should account for this common cause effect 3
=
Background===
4 1994 Northridge earthquake Haiti
=
Background===
Single Motor Control Center Failure at Piti Power Plant -
Guam (1993 Earthquake) 5
=
Background===
Fukushima Daini Make Up Tank Failures -
Dependent Failures 6
Background
- Treatment of dependent failures involves a set of assumptions (rules-based approach) such as:
- Similar redundant components next to each other and subjected to similar seismic forces fail simultaneously
- Diverse or similar components at different (far apart) locations fail independently
- Sensitivity analysis to postulated bounding assumptions
- This approach can mask insights and lead to varying degrees of conservatism in various places of the analysis 7
Background
- The treatment of dependencies of seismic failures in SPRAs depends on the intended use of their results.
- Further demonstrations of the identified methodology and its use would depend on arising uses or needs or, for example, applications to new designs, namely advanced non-light water reactors or modular reactor designs.
8
Objective
- As part of the 2010-2014 seismic and structural research plan, the Office of Nuclear Regulatory Research (RES) initiated research to:
- evaluate and explore the seismic correlation-dependency topic and, if feasible,
- to propose one or more advanced, more refined and more insightful analysis approaches for its treatment in seismic Probabilistic Risk Assessments (PRAs).
9
Goals
- That the evaluation of existing methodologies, the review of the related SSC-failure database, and the attributes of the proposed analysis methodology will inform SPRA analyses, peer-reviews, and staff reviews in relation to the treatment of seismic failures dependencies in seismic PRAs for operating and new reactors.
- Identification of a methodology that with further demonstrations (by its potential users) and use will provide an improved and more insightful approach to seismic PRA 10
Scope
- Assessment of impact of correlation assumptions made in modern SPRAs have on risk estimates (concentrating on CDF)
- Review and identification of analysis methods and data sources that could be developed for improved and more insightful risk assessments
- Identification of a pathway (methodology) so that improved analysis of dependent seismic failures could become a standard part of SPRAs 11
Analysis Aspects
- Dependencies arise in the system analysis part of a PRA
- Dependencies of SSCs arise in the calculation of the failure frequency of a sub-system of SSSs with performance defined by a Boolean (logic) statement
- Assumption of dependency results in an upper bound estimate of the sub-system failure frequency (system fragility)
- Assuming independency results in an upper bound of the sub-system failure frequency (system fragility) 12
Analysis Aspects
- Dependencies arise in the system analysis part of a PRA
- Output of the analysis of the sub-system Boolean expression is essentially a fragility function for the system
- Simplified (anticipated to be bounding) assumptions are typically made
- An improved methodology would obtain narrower bounds on the system failure and address sensitivity of the results to defensible assumptions on dependencies and correlations
- Approach: involve fragility and system analysts to obtain fragility functions for the sub-systems functions (Boolean expression) 13
Analysis Aspects
- Seismic Fragility: A = AmEREU
- Am = median, ER= inherent random uncertainty about the median and EU = epistemic uncertainty about the median
- Usually A = F x RE
- F = a factor of safety and RE = a reference size of earthquake
- F = FCFRS=(FSFu)(FSAFdFMFMCFECFSDFSS)
Capacity Seismic Partial factor response factors factor 14
Analysis Aspects
- Seismic Fragility: A = AmEREU
- When there are several components in a Boolean statement, (B, C, D, ) the method involves:
- Determining the correlation for the EU uncertainty between components (co-variances)
- Assigning the remaining uncertainty (independent part of the uncertainty) to each component (preserving the total uncertainty)
- Similarly for the uncertainties for ER
- Then the methodology uses a set of rules to determine the frequency of the sub-system failure 15
Analysis Aspects
- Uncertainties
- 1. Determination of the correlation coefficients for the uncertainties in EU and ER
- 2. Subsequent set of rules to calculate the sub-system failure probability
- Item 1 - The correlation coefficients can be determined from the partial factors of safety and the combined to obtain those for Eu and ER (may reduce variability among analysts)
- Item 2 - Verification of the simplified method against more rigorous existing mathematical approaches (too onerous for general use) 16
Study Approach
- Review of several SPRAs in the literature (Chapter 4)
- Review of the existing literature on seismic correlation and dependency analysis (Chapters 2, 6 and 7)
- Review of existing data from testing and earthquake experience (Chapter 3)
- Identification of a possible improved and more insightful methodology for treating dependencies in a SPRA 17
Review of SPRAs
- Assessed impact of assumptions on the treatment of dependencies on the seismic failure of SSCs in risk estimates (concentrated on Core Damage Frequency (CDF) risk estimates)
- Almost all SPRAs used the rule-based approach with few exceptions
- The Zion SPRA for the Seismic Safety Margins Research Program (SSMRP) and the Diablo Canyon SPRA
Review of SPRAs
- Observations from the review (examples), which used expert judgement:
- For some SPRAs, the differences in the seismic CDF could range from close (but less than) a factor of 2, but with a typical difference in the range of 30% to 60%.
- For some accident sequences, the differences could be larger (as much as factor of 2 to 4)
- Treatment of dependent failures has small impact if the CDF is controlled by an accident sequence that is itself controlled by a single SSC (singleton).
- For several SPRAs the CDF estimates are not very sensitive to the dependencies issue but risk insights can be sensitive to the assumptions made in the analysis of dependencies
- A few types of SSCs seemed to appear repeatedly among those for which the assumptions used affected the results (the CDF estimates) (although not all affect the results in every SPRA reviewed) 19
Review of SPRAs
- SSCs that seem to appear repeatedly among those for which the approach to analyze dependencies in their seismic failure can affect the CDF estimates (not all appeared in this manner in all SPRAs reviewed):
- Masonry walls
- Electrical Motor Control Centers (MCCs)
- Large tanks: condensate storage tanks of other similar tanks
- Small tanks: diesel generator fuel oil day tanks
- Heat exchangers: such as component cooling water heat exchangers
- Mechanical: long-shafted service-water pumps, horizontal auxiliary feedwater pumps (motor and turbine driven dependency)
- Batteries and racks 20
Review of Existing Literature on Dependency Analysis
- A variety of approaches with different level of mathematical rigor have been proposed
- SSMRP approach
- Used join-lognormal distributions for the response and capacity of SSCs in a cut-set
- Defines cut-set failure as all responses exceeding their capacities
- Integrates the joint distributions over the failure region
- Used in the Zion and SPRA and to derive a set of simplified rules (rule-based approach) called the Bohn-Lambright rules
- Practice tends to use further simplified rules, which assign 100% to the situation of similar SSCs exposed to the same earthquake load (typically, SSCs located near each other), and zero everywhere else.
21
Review of Existing Literature on Dependency Analysis
- Bohn rule-based approach Rule # Description 1 Components on the same floor slab and sensitive to the same spectral frequency range (i.e., ZPA, 5-10 Hz. or 10-15 Hz) will be assigned response correlation = 1.0.
2 Components on the same floor slab sensitive to different ranges of spectral acceleration will be assigned response correlation = 0.5.
3 Components on different floor slabs (but in the same building) and sensitive to the same spectral frequency range (ZPA, 5-10 Hz or 10-15 Hz) will be assigned response correlation = 0.75.
4 Components on the ground surface (outside tanks, etc.) shall be treated as if they were on the grade floor of an adjacent building 5 "Ganged" valve configurations (either parallel or series) will have response correlation = 1.0.
6 All other configurations will have response correlation equal to zero.
22
Review of Existing Literature on Dependency Analysis
- The review identified an approach that would lead to a more refined and more insightful methodology
- Initially proposed by Reed and McCann (1985) but not further developed
- Leverages approaches widely used in the fragility analysis of SSCs that lead themselves to engineering identification of correlations in the uncertainties of variables that determine and SSC fragility estimation 23
Review of Existing Data
- Main observations in relation to data-driven support for determining seismic failure dependencies
- Shake-table testing data
- Shake-table test data is largely inadequate for purpose of understanding dependencies among failures - Mostly qualification testing data and when taken to failure mostly testing of single items
- Earthquake experience data
- Largely inadequate for the purposes of understanding the dependencies among seismic failures - Sparse data, difficult to understand and large interpretation uncertainties 24
Analysis Process in a SPRA
- 1. Performance of a SPRA 3. Use the improved methodology using standard methods to study the identified cut sets or Needs to include sufficient accident sequences one-by-sensitivity analysis studies) one.
Revised fragility functions
- 2. Identification of those (presumably few) cut- 4. Use of the revised fragility sets or accident functions in the SPRA in the sequences where usual way correlations analysis May involve iterations to Step 3 or makes a difference Step 2 Difference in terms of risk quantification, importance, and safety insights 25
Analysis Issues/Uncertainties
- Difficult to use methodology for the revised fragilities except for most experienced SPRA fragility analysts
- Variability among fragility analysts
- Variability from analyst to analyst
- Expert judgement to assign uncertainties as dependent or independent (determining coefficients of correlation)
- Would need further sensitivity assessments
- Issue of data availability and focus on matters of more or less conservative analyses (instead of, for example, looking for safety insights)
- Peer-review needs 26
Summary / Remarks
- As part of the 2011-2014 seismic and structural research plan RES initiated research to explore the topic of dependent seismic failures of SSCs is SPRAs and, if feasible, to identify more refined and insightful analysis approaches for its treatment in SPRAs.
- The research documented reviewed past SPRAs, the SSC-failure database, and identified a more refined and insightful methodology that has promise for practical treatment of the dependency issue in SPRAs
- The methodology described needs further demonstrations (by potential users) to provide an new and more insightful approach for SPRAs (in relation to the treatment of dependency of seismic failures) 27
Summary / Remarks
- The treatment of dependencies of seismic failures in SPRAs depends on the intended use of their results.
- Further demonstrations of the identified methodology and its use would depend on arising uses or needs or, for example, applications to new designs, namely advanced non-light water reactors or modular reactor designs.
28