ML16230A232

Responses to Apla Round 4 Requests for Additional Information Regarding STP Risk-Informed GSl-191 Licensing Application

ML16230A232
Person / Time
Site:	South Texas
Issue date:	07/21/2016
From:	Gerry Powell South Texas
To:	Document Control Desk, Office of Nuclear Reactor Regulation
References
GSI-191, TAC MF2400, TAC MF2401
Download: ML16230A232 (55)

Text

South lexas Project F.lectrlc Generatl11g St.Jffo11 f!O Sar 28fl Wadsworth. lixls 77483 U.S. Nuclear Regulatory Commission Attention: Document Control Desk Washington, DC 20555-0001 South Texas Project Units 1 & 2 Docket Nos. STN 50-498, STN 50-499 July 21, 2016 NOC-AE-16003390 10 CFR 50.12 10 CFR 50.90 Responses to APLA Round 4 Requests for Additional Information Regarding STP Risk-Informed GSl-191 Licensing Application (TAC NOs MF2400 and MF2401)

References:

1. Letter, G. T. Powell, STPNOC, to NRC Document Control Desk, "Supplement 2 to STP Pilot Submittal and Requests for Exemptions and License Amendment for a Risk-Informed Approach to Address Generic Safety Issue (GSl)-191 and Respond to Generic Letter (GL) 2004-02", August 20, 2015, NOC-AE-15003241, ML15246A126

2. Letter, Lisa Regner, NRC, to Dennis Koehl, STPNOC, "South Texas Project, Units 1 and 2-Request for Additional Information Related to the Risk Review of the Request for Exemptions and License Amendments to Resolve the Issue of Potential Debris Blockage on Emergency Recirculation During Design-Basis Accidents at Pressurized-Water Reactors", May 26, 2016, ML16125A290 Reference 2 transmitted Probabilistic Risk Assessment Licensing Branch (APLA) Branch RAls regarding the South Texas Project Nuclear Operating Company (STPNOC) application in Reference 1. This submittal responds to those RAls.

Commitments are listed in Attachment 2.

NOC-AE-16003390 Page 2 of 3 If there are any questions, please contact Mr. Wayne Harrison at 361-972-8774.

I declare under penalty of perjury that the foregoing is true and correct.

Executed on:

J;;.l._,

1 zl 1 zollt awh Attachments:

1. Response to APLA RAls

2. Commitments

3. Definitions and Acronyms 4/.~

G. T. Powell Executive Vice President and Chief Nuclear Officer

,cc:

(paper copy}

Regional Administrator, Region IV U. S. Nuclear Regulatory Commission 1600 East Lamar Boulevard Arlington, TX 76011-4511 Lisa M. Regner Senior Project Manager U.S. Nuclear Regulatory Commission Or:ie White Flint North (08H04) 11555 Rockville Pike Rockville, MD 20852 NRC Resident Inspector U. S. Nuclear Regulatory Commission P. 0. Box 289, Mail Code: MN116 Wadsworth, TX 77483 (electronic copy)

NOC-AE-16003390 Page 3 of 3 Morgan. Lewis & Beckius LLP Steven P. Frantz, Esquire U.S. Nuclear Regulatory Commission Lisa M. Regner NRG South Texas LP Chris O'Hara Jim von Suski!

Skip Zahn CPS Energy Kevin Pollo*

Cris Eug5ter L. D. Blaylock Crain Caton & James. P.C.

Peter Nemeth City of Austin Elaina Ball John Wester Texas Dept of State Health Services Helen Watkins Robert Free Response to APLA RAls NOC-AE-16003390

REQUEST FOR ADDITIONAL INFORMATION NOC-AE-16003390 Page 1 of 10 Note: The questions below use the numbering system established in the RAJ issued to STPNOC on April 11, 2016 (ADAMS Accession No. ML16082A507). The system consists of the Office of Nuclear Reactor Regulation (NRR) Branch acronym (i.e.,

Probabili$tic Risk Assessment Licensing Branch or APLA), the NRG RAJ Round No.,

then a sequential number starting with 1.

NRC RAI

References:

1. STP Nuclear Operating Company, "Enclosure 4.,2, 'South Texas Project Risk-Informed Closure of GSl-191., Volume 2, Probabilistic Risk Analysis,"' Revision 2, dated October 22, 2013 (ADAMS Accession No, ML13323A189).

2. Powell, G. T., STP Nuclear Operating Company, letter to U.S. Nuclear Regulatory Commission, "South Texas Project Units 1 and 2 - Supplement 2 to STP Pilot Submittal and Requests for Exemptions and License Amendment for a Risk-Informed Approach to Address Generic Safety Issue (GSl)-191 and Respond to Generic Letter (GL) 2004.:.02,"

dated,August 20, 2015 (ADAMS Accession No. ML15246A126).

3. Powell, G. T., STP Nuclear Operating Company, letter to U.S. Nuclear Regulatory Commission, "Description of Revised Risk-Informed Methodology and Responses to Round 2 Requests for Additional Information Regarding STP Risk-Informed GSl-191 Licensing Application," dated March 25, 2015 (ADAMS Accession No. ML15091A440).

4. Powell, G. T., STP Nuclear Operating Company, letter to U.S. Nuclear Regulatory Commission, "South Texas Project Units 1 and 2-First Set of Responses to April';

2014, Requests for Additicmal Information Regarding STP Risk-Informed GSl-191 Licensing Application Revised," dated May 22, 2014 (ADAMS Accession No. ML14149A434).

APLA-4-1:

NOC-AE-16003390 Page 2of10 NRC Regulatory Guide (RG) 1.174, Revision 2, "An Approach for Using Probabilistic Risk Assessment in Risk-Informed Decisions on Plant-Specific Changes to the Licensing Basis," May 2011 (ADAMS Accession No. ML100900006), states, in part, that, "the [probabilistic risk assessment (PRA)] should realistically reflect the actual design, construction, operational practices, and operational experience of the plant." Therefore, whether a particular accident sequence (e.g., secondary side break followed by sump recirculation) is part of a plant's licensing basis is immaterial when performing a risk analysis. For example, accident sequences involving common cause failures are not part of a plant's licensing basis since the Updated Final Safety Analysis Report (UFSAR) Chapter 15 accident analyses require considerat.ion of only a single active failure. Yet operational experience shows that these types of accidents can occur and they therefore must be modeled by the plant's PRA.

Consistent with this guidance, the risk assessment of debris should consider all hazards, initiating events, and plant operating modes.

It should not be limited to design-basis accidents, licensing basis events, specific plant operating modes, or specific initiating events such as a loss-of-coolant accident (LOCA).

In the licensee's submittal dated October 13, 2013, Enclosure 4-2, "South Texas Project Risk-Informed Closure of GSl-191, Volume 2, Probabilistic Risk Analysis" (Reference 1 ), provides the screening rationale and concluded in Section 12.3 that "Medium and large LOCAs from internal events only are retained for further consideration with respect to core damage resulting from GSI-'

191 phenomena." The Volume 2 document also stated that the full-power analysis bounds

• consequences of other plant states.

A supplemental analysis (Risk-Over-Deterministic or RoverD) was submitted in STPNOC's letter dated August 20, 2015 (Attachment 1-3; Reference 2), which did not supersede previous submittals but purported to be "standalone." This raises the question regarding the applicabliity of the earlier information.

Please confirm that the conclusion in Section 12.3 in the earlier submittal (Reference 1) *applies to the Rovero supplemental analysis (Reference 3); that is, confirm that all hazards, initiating events, and plant operating modes were screened out of consideration except medium and large:

LOCAs and that full power operation is the only operating mode that merits consideration in the qetailed RoverD ana\\yses.

STP Response:

All hazards, initiating events, and plant operating modes were screened out of consideration except medium and large LOCAs such that full power operation is the only operating mode that merits consideration in the detailed RoverD analyses.

APLA-4-2:

NOC-AE-16003390 Page 3of10 NUREG-1829, "Estimating Loss-of-Coolant Accident (LOCA) Frequencies Through the Elicitation Process,'! Apr.ii 2008 (ADAMS Accession No. ML082250436), intludes only breaks caused by long-term material degradation.

Other potential contributors to LOCA frequency such as seismically induced LOCA (both direct and indirect) should be evaluated separately. A "direct" seismically induced LOCA involves rupture of a piping or non-piping component caused by the seismic event itself. An "indirect" seismically induced LOCA is caused by, for example, failure of piping or component supports that leads to the consequential failure of the piping or non-piping component.

In its May 22, 2014, response to an NRC RAI dated April 14, 2014 (ADAMS Accession No. ML14087A075), STP provided an estimate of the frequency of seismically induced LOCA (Attachment 1, p. 24/86; Reference 4 ). However, the response did not appear to consider indirect seismically induced LOCAs. One acceptable approach for evaluating indirect seismically induced LOCAs is for the analyst to use the method described by NUREG-1903, "Seismic Considerations for the Transition Break Size," Section 4.6, "Indirectly Induced Piping Failures," February 2008 (ADAMS Accession No. ML080880140).

"Representative" values in the NUREG could be replaced with site-specific fragility and hazard information that, as appropriate, accounts for any effects of material degradation or aging.

Alternatively, it may be demonstrated that the representative values are bounding for the site with consideration of effects due to material degradation or aging.

Please clarify whether the analysis documented in the RAI response considered indirect seismic LOCAs. If not, provide an analysis accounting for indirect damage mechanics eventually leading to rupture of piping and non-piping systems and LOCA events. For both direct and indirect seismically induced LOCAs, estimate, bound or screen any increase in seismic risk due to debris.

STP Response:

STPNOC Will provide this response separately.

APLA-4-3:

NOC-AE-16003390 Page4of10 NUREG-1829 LOCA frequencies include only breaks caused by long-term material degradation.

Other potential *contributors to LOCA frequency, such as water hammer, should be evaluated separately.

One acceptable approach for evaluating water hammer is to verify that the potential for water hammer is not likely to cause pipe rupture in the break locations that can produce and transport problematic debris. Water hammer includes various unanticipated high-frequency hydrodynamic events, such as steam hammer and water slugging. To demonstrate that component failure risk due to water hammer is acceptably low, the analyst could take the following actions:

D Assess historical frequencies ofwater hammer events affecting break locations (piping and non-piping) that could generate anci transport debris.

D Evaluate operating procedures and conditions and demonstrate that they are effective in precluding water hammer~

D Alternatively, the a_nalyst can demonstrate the fd!lowing:

o Plant changes, such as the use of J-tubes, vacuum breakers, and jockey pumps, coupled with improved operating procedures, have been used to successfully mitigate water hammer events.

o Measures ysed to abate water hammer frequency and magnitude have been effective over the ljcensing period of the plant.

Please evaluate the relevance of water hammer events in the context of GSl-191 and estimate, bound, or screen any increase in risk due to water hammer events.

STP Response:

The portions of tne STP RCS that are subject to a LOCA are designed to Class 1 requirements of Section Ill of the ASME.Boiler and Pressure Vessel Code, which includes consideration of appropriate tran.sie11ts. STP specifically includes consideration and mitigation of likely water hammer in the RCS. Pressurizer piping is a primary area of consideration due its function during RCS pressure transients. The pressurizer safety and relief valve manifold assembly; including valve supports, is designed for loads d.ue to water relief, including the pc:issage of c:i water slug and the effects of water.ha.mmer. Stresses are withi11 code allowable limits. The pressurizer piping is also instrumented to monitor for indications of safety or relief valve leakage thc:it would contribute to creating a water hammer condition and the STP Technical Specifications (TS) impose.li111.its 011 excessive relief valve seat leakage.

STPNOC al.so performed a search of Corrective Action Program data for water hammer and found no issues in systems related to GSl-191 locations of concern. Based on the fact that the piping is designed to ASME Ill standards and the lack of historical data for' events, the relevance of water hammer events in the context of GSl-191 is deemed insignificant.

APLA-4-4:

NOC-AE-16003390 Page 5of10 Please provide values of total risk estimates (also including water hammer and seismically induced LOCA) for the plant expressed as core damage frequency (CDF) and large early release frequency (LERF). Those values are not available in the recent Rovero analysis (Reference 2).

This information is needed to compare pairs {CDF, ~CDF} and {LERF, ~LERF} to risk acceptance guidelines in RG t.174.

STP Response:

Water hammer events are screened from consideration due to system and component design as described previously in the response to APLA-4-3. Therefore, water hammer is assessed fo result negligible increase in risk. Seismically-induced LOCA risk estimates are provided in the response to RAI APLA-4-2, above. CDF and LERF from seismically-induced LOCA are found to b~ insignificant to the assessment.

APLA-4-5:

NOC-AE-16003390 Page 6of10 Title 10 to the Code of Federa/Regu/ations ( 10 CFR), Subsection 50.46(b )(5), Long-term cooling, and Appendix A to 1 O CFR Part 50, General Design Criterion 35, Emergency core cooling, state, in part, that the emergency core *cooling system (ECCS) must provide core cooling for extended periods following postulated LOCAs. Licensing basis analyses used to demonstrate compliance with these regulations have historically analyzed the effects of debris in a deterministic manner.

STPNOC has submitted a pilot LAR and a series of exemptions that, if approved, would change its licensing basis by using a risk-informed treatment of debris (Reference 3). RC3 1.17 4 contains five key principles for performing risk..:informed changes to a plant's licensing basis. Principle 5 states that an implementation and monitoring program should be utilized to ensure that the conclusions reached by the staff (e.g., that the increase in risk is small) remain valid.after the change is implemented.

The NRC staff has determined that it does not yet have adequate assurance that principle 5 of RG 1.17 4 *is met and. that there are sufficient regulatory controls of the key elements of the STP risk-informec:I assessment of debris.

Specifically, the NRC requires regulatory assurance of the continued applicability of the results of the risk-informed approach for Gonsideration of debris in order to issue the requestec:I license amendments and grand the associated exemptions. In order to obtain this regulatory assurance, certain aspects of the risk-informed approach must (1) be subject to an ongoing monitoring program consistent with principle 5 of RG 1.174; (2) be periodically updated; (3) continue to use methods acceptable to the NRC; and (4) be subject to reporting and corrective action when the risk-informed acceptance criteria are not met. The NRC also requires regulatory assurance that the risk-informed approach will not be employed for plant design changes that would increase the problematic debris source term without prior NRC review and approval.

Please provide assurance of appropriate regulatory considerations:

a. Prior to changing the key methods, approaches, and data of the risk-informed analysis set forth in (reference).

b. Prior to using the risk-informed approach to justify future plant.design changes that would increase the problematic debris source term compared to the level that existed as of (date).

c. STPNOC will implement and maintain a program to monitor key assumptions and data used in the risk assessment and the evaluation of defense-in-depth and safety margins.

The monitoring program must assess the effects of design or plant modifications, procedure changes, as-found conditions, identified changes or errors in the analysis, industry operating experience, and any other information that could result in increased risk, or decreased defense-in-depth or safety margins, under the alternative risk-informed approach. The results of the monitoring program should be retained onsite for inspection.

NOC.-AE-16003390 Page 7of10

d. STPNOC will update the risk-informed evaluation no later than 48 months after initial NRC approval or the latest update anq compare the risk results, CDF, LERF, ~CDF, and ALE RF, to the acceptance criteria in the safety evaluation that accompanies the requested LAR (reference).

e. In the event that the acceptance criteria for the risk-informed analysis are not met

1. STPNOG wil.1 notify the NRC in accordance with 10 CFR § 50.72 or 50,73 to notify the NRC that the acceptance criteria has been exceeded; and,

2. STPNOC will take timely action to ensure that the acceptance criteria are m-et.

These requirements are in addition to (and separate from) the reporting requirements in 10 CFR 50,46(a).

STP Response:

APLA-4-5.a, b and e.1:

STPNOC is implementing this debris assessment as a PRA Analysis Assessment. It is a bounding evaluation that shows that the effects of debris have a small effect on risk such that debris effects need not be considered for ECCS and CSS performance. This evaluation is performed to close GL2004-2 fo~ STP Units 1 and 2.

STPNOC proposes to include a change control and reporting section in the new Appendix 6A to the STP UFSAR. This control process is similar to the change control process incorporated into Ch. 13.7 of the STP UFSAR that the NRC accepted for the "Special Treatment" exemption request (ML011430090), as discussed in the staff's SE and Notice of Exemption (ML 12040370, ML 12040470).

STPNOC describes the new Appendix 6A in Attachment 3-4 to Reference 1 of the cover letter.

That attachment was provided for the staff's information. The proposed Section 2 of Appendix 6A described below will be provided for the staff's approval as part of the LAR.

APLA-4-5.e.1: Nonconforming conditions that make the strainer(s) inoperable for longer than required TS completion time will meet the 1 OCFR50.73 reporting criteria for a condition prohibited by TS. Conditions that cause the emergency sump strainers to be inoperable and result in the debris-related ~CDF or ~LERF to be greater than the RG 1.174 Region II acceptance guidance are to be reported in accordance with 10CFR50.72 or 1 QCFR 50.73. This guidance is included in the proposed Section 2 of UFSAR Appendix 6A.

APLA-4-5.e.2, timely action will be assured by the proposed TS required action time and/or the STPNOC corrective action program.

NOC-AE-16003390 Page 8of10 APLA-4-5.c:

Monitoring is addressed by the STPNOC Corrective Action Program (OPGP03-ZX-0002, "Condition Reporting Process"), the design change control process, the operating experience review program and the in-service inspection program, which are required programs that work together to address both planned and emergent conditions (STP-specific or industry experience) that could affect the results of the risk-informed resolution of debris effects.

APLA-4-5.d:

The STP licensing application evaluation is a PRA Analysis Assessment, which is a bounding evaluation of debris effects that calculated only delta-COF and delta-LERF. As an assessment, it is not incorporated into the PRA. STPNOC will meet the 48-month requirement by reviewing relevant elements of the assessment to determine if there were any significant changes that would affect the conclusions.

New Section 2.0 for UFSAR Appendix GA:

2.0 Change Control and Reporting This section describes the parts of the methodology change for Appendix 6A which have additional regulatory requirements for prior NRC approval or notification. The requirements in this section may not be changed without prior NRC approval.

2.1 Change Control 2.1.1 Changes to key methods and approaches of the risk-informed methodology set forth in [final NRG-approved Rovero description] are to be evaluated as a potential "departure from a method of evaluation described in the FSAR (as updated) used in establishing the design bases or in the safety analyses" analogous to 1 OCFR50.59(c)(2)(viii).

2.1.1.1 The key aspects of the Rovero methodology subject to Section 2.1.1 are the following:

1. The methodology for quantifying the pipe break frequencies used to calculate the change in COF and LERF.

a. The pipe break frequency source (NU REG 1829)

b. The methodology for calculation of Oi(small)

c. The methodology to identify break locations. This requirement applies to the criteria, but not to the tool or program used to identify the locations; i.e.,

programs other than CASA Grande may be used.

d. The methodology to interpolate the pipe break frequencies to quantify the LlCOF a?sociated with pipe breaks

2.1.1.2 NOC.-AE-16003390 Page 9of10

e. The methodology for assigning the RPV HL breaks to failure

f.

The methodology used to calculate.liLERF for effects of debris

2. The assumption that fine fiber is to be applied 1;1s the governing debris source

a. The methodology used to quantify the amount of fiber generated at each break location, including assumed ZOI. This requirement applies to the criteria, but not to the tool or program used; i.e., programs other than CASA Grandi? may be used.

b. The use of the July 2008 STP-specific test to establish the deterministic baseline for the quantity of fine fiber.

3. The assumptions and methods in the thermal-hydraulic analyses

a. The use of 800°F as the acceptance criterion for long-term core cooling

b. Assumptions regarding core blockage

4. The availability of key sources of defense in depth.

a. Capability for containment heat removal unaffected by debris on ECCS strainers (RCFCs)

b. Capability to refill the RWST Plant Design Changes In addition to the controls of Section 2.1.1.1, changes to plant design shall be subject to.the following requirements:

1. Procedural controls establish limits on the introduction of new debris sources Into containment, particularly fine fiber or particulate sources, to assure the deterministiq licensing basis testing and calculations performed for the Rovero methodology remain bounding.

2.2 Reporting Nonconforming conditions that make the strainer(s) inoperable for longer than required TS completion time will meet the 10CFR50.7.3 reporl:ing criteria for a condition prohibited by TS.

Conditions that cause the emergency sump strainers to be inoperable and result in the debrls-related ~CDF or ~LERF to be greater than the RG 1.17 4 Region II acceptance guidance are to be reported ir:i accordance with 10CFR50.72 or 10CFR 50.73. This guidance is included in the proposed Section 2 of UFSAR Appendix 6A.

J

APLA-4-6:

NOC-AE-16003390 Page 10of10 Please confirm the accuracy of the computation of the.LlCDF for the double-ended guillotine break (DEGB)-only model that was provided (References 2 and 3). It appears that the current approach assumes, for example, that a 12.8-in DEGB could occur in larger diameter pipes (e.g., 27.5-in, 29-in, and 31-in diameter pipes). Include arithmetic and geometric mea~ approaches.

STP Response:

The results for the DE GB-only previously reported used an extension of the continuum model that was interpreted in a way that included breaks in pipes larger than the DEGB size in smaller pipes.

The STP break frequency methodology has been revised (attached) to interpret DEGB frequencies in a way that only considers DEGB pipe breaks at the specific pipe sizes. Adopting the new interpretation results in an insignificant change to the frequency estimates in Rovero and does not change the conclusions reported previously.

Attachment:

Risk Unifying Frequency Functional (RUFF) Software Quality Assurance Documentation

Risk Unifying Frequency Functional (RUFF)

Software Quality Assurance Alexander J. Zolan, John J. Hasenbein, Jeremy J. Tejada Gr.aduate Program in Operations Research and Industrial Engineering Department of Mechanical Engineering The University of Texas at Austin Austin, Texas, USA Abstract In this software quality assurance document, we provide documentation of the llisk Unifying Fre-quency Functional (RUFF) software design, a software test plan, results of software test cases, and in-structions for using RUFF. This document is intended to satisfy the Software Quality Assurance (SQA) documentation requirements for level 2 software at South Texas Project (STP) as defined in OPGP07-ZA-0014.

2

Contents 1

• Introduction 1.1 Purpose...................

1.2 Definitions, Acronyms, and Abbreviations 1.3 Document Overview............

2 Software Requirements 3 Software Design 3.1 Design Details.........

3.1.1 LOCA Frequencies..

3.1.2 Welds and Break Sizes 3.1.3 Continuum J\\/Iodel.

3.1.4 DEGB-Only 1\\tfodel..

3.2 Source Code Listing.....

3.3 Python Implementation Details 3.3.1 J\\/Iethod ReadConfig..

3.3.2 Class NUREG_l829_Freqs 3.3.3 Class LOCAEventCalculator.

4 Software Test Plan 4.1 Testing J\\/Iethodology 4.2 Interface Programs.

4.3 RTE Test Problems.

5 Software Test Case 5.1 Test Report...

5.2 Testing Environment 5.2.1

Machine 1 5.2.2

Machine 2 5.3 Test Cases...

6 User Instructions

(

6.1 Training Needs Assessment.............

6.2 Major Assumptions and Limitations........

6.3 User Instructions for Personal Computer Software 3

6 6

6 6

7 9

g g

g 10 11 12 12 12 12 13 15 15 15 15 16 16 16 16 17 17 23 23 23 23

6.4 Input Description.

24 6.5 Output Description.

26 6.6 Known Issues and vVorkarounds 26 6.7 User Support 26 4

Acronyms CASA CDF

,6.CDF GSI-191 LAN LAR LOCA NRC PNG RCB RUFF SQA STP Containment Accident Stochastic Analysis (CASA) Grande Core Damage Frequency Change in Core Damage Frequency Generic Safety Issue 191 - the NRC Generic Safety Issue Number 191 Local Area Network License Amendment Request Loss of Coolant Accident Nuclear Regulatory Commission Portable Network Graphics Reactor Containment Building Risk Unifying Frequency Functional Software Quality Assurance South Texas Project STPNOC South Texas Project Nuclear Operating Company 5

1 Introdµction The South Texas Project Nuclear Operating Company (South Texas Project Nuclear Operating Company (STPNOC)) requires a software tool that es-timates the risk associated with the frequency of loss of coolant accident (Loss of Coolant Accident (LOCA)) events. In particular, the.software must determine the frequency of LOCA events that generate a quantity of fiber fines exceeding thresholds established though physical testing (Change in Core Damage Frequency (ACDF)). The Risk Unifying Frequency Functional (RUFF) program detailed in this document describes a method for estimat-ing this frequency given a subset of welds in the plant that can generate a sufficient amount of fiber fines (critical weld locations) and the minimum break size that generates the sufficient amount of fiber fines for each critical weld (critical break sizes). The RUFF software program was developed at the University of Texas at Austin.

1.1 Purpose The purpose of this document is to provide evidence of the SQA procedures that were utilized in the development of RUFF. The sections in this report are intended to provide the required SQA documentation in accordance with OPGP07-ZA-0014 "Software Quality Assurance Program". RUFF is classified as Level 2 software at STP based on its connection to accident analysis, and this document provides the SQA documentation required for Level 2 software.

1.2 Definitions, Acronyms, and Abbreviations Some important definitions are provided below:

• Input Variable: A variable whose value(s) are specified as an input to RUFF

• State Variable: A variable whose value(s) are calculated at each time step e Output Variable: A variable whose value(s) are calculated by the sim-ulation and displayed and/ or saved 1.3 Document Overview This document is o,rganized into a series of sections, each representing a key piece of software SQA. Section 3 details the proposed software design 6

features and the specific programming objects and that make up RUFF.

Section 2 describes the software requirements. Section 4 outlines the plan for testing RUFF to ensure the program functioning correctly using a test case and Section 5 compares the results of a spreadsheet validation to the RUFF calculations for that test case. Section 6 provides instructions for installing, setting up, and using RUFF. Lastly, a copy of the RUFF source code is available in the appendix.

2 Software Requirements

• Required Functions The software functions to estimate particular frequencies at locations in pipes of given sizes for different LOCA failure scenarios at a nuclear power plant; for breaks that only occur at the inside diameter of the nuclear power plant pipes; and for breaks that can occur at any diameter up to the inside diameter of the nuclear power plant pipes.

• Calculation Requirements The calculation must interpolate as necessary, frequencies for given sizes within the data range, divide by the number of opportunities within the inside diameter intervals to arrive at a frequency for each interval. Once all inside diamter interval frequencies are calculated, they are summed.

Several quantiles and a mean can be provided and for each, the software performs the same calculation.

• Required Output

, The summed frequencies at the mean and quantiles, as defined in the inputs, are provided in a text file.

• Required Inputs Inputs are required as enumerated below.

1. A table having two or more data sets are provided for frequency of failure at a given sizes and given quantiles and means.

2. A set of "failure size" diameters required to fall inside the range of pipe sizes (interpolates).

3. The number of possible locations and inside diameters for pipes under consideration.

7

• Assumptions No assumptions are required. The underlying method simply describes an possible way to allocate frequencies in a "top down" scheme which implies that the frequencies are the same at each interpolate.

• Independent Review The validation and verification activities are independently reviewed.

8

3 Software Design This section provides an overview of RUFF, including the details of the cal-culation and implementation of that calculation in the RUFF programming interface. RUFF is a Python-based program designed to automate the calcu-lations associated with estimating the frequency of LOCA events that gener-ate a quantity of fiber fines exceeding thresholds established tho11gh physical testing (.6.CDF). In this section, we detail a Python implementation of this methodology.

3.1 Design Details 3.1.1 LOCA Frequencies Tregoning, Abramson, and Scott [1] determined a plant-wide frequency and impact (in terms of flow rate) of different size breaks in PWRs and BWRs through elicitation of experts. The flow rates in the elicitation were converted to break diameters and presented in tables as exceedance frequencies; these data are provided as inputs to RUFF in fiat file form. Tregoning et al. are widely referenced for LOCA frequencies used in plant PRAs (see [2] for PRA LOCA initiating event frequency prior data), and are the source of inputs for RUFF calculations implemented by STPNOC; however, RUFF can accept

. any set of break sizes and their respective frequencies as input. Table 1 presents an example of the LOCA frequencies by break size used in RUFF.

Table 1: NUREG-1829 (Tregoning et al., 2008, Table 7.19) estimates for the mean, median, 5th percentile, and 95th percentile exceedance frequency values for current-day estimates Break Size (in.)

Category 5th %ile 50th %ile Mean 95th %ile 0.500 1

6.80E-05 6.30E-04 1.90E-03 7.lOE-03 1.625 2

5.00E-06 8.90E-05 4.20E-04 l.60E-03 3.000 3

2.lOE-07 3.40E-06 1.60E-05 6.lOE-05 7.000 4

l.40E-08 3.lOE-07 l.60E-06 6.lOE-06 14.000 5

4.lOE-10 1.20E-08 2.00E-07 5.80E-07

.31.000 6

3.50E-11 l.20E-09 2.90E-08 8.lOE-08 3.1.2 Welds and Break Sizes CASA Grande is a software tool that was developed to analyze accident sequences in a realistic time-dependent manner to determine the probabili-ties of various failures potentially leading to core damage from a spectrum of location-specific pipe breaks (i.e., LOCA events) [3]. As a preprocessing step to RUFF, CASA Grande runs are performed to identify all weld loca-tions, with corresponding break sizes, which produce more than the allowable 9

amount of fiber fines.

3.1.3 Continuum Model We start with the list of welds and corresponding break sizes taken from CASA Grande as described in Section 3.1.2. With this step completed, we have data that can be thought of as ordered pairs consisting of a weld index and a break size. Assume that I weld locations are in the risk-informed category and these locations are indexed by i = 1,..., I. Each weld location i then has a corresponding break size D!mall which caused it to be placed in the risk-informed category. It is possible that for a single weld, multiple break scenarios caused it to be put in this category. If so, define D!mall to be the smallest such break size.

The goal is to determine the overall frequency of events that generate too many fiber fines. First, for each weld i in the risk-informed category, the goal is to determine the frequency of breaks that exceed D!mall

• This is called F(D.~mau) and is the frequency of unacceptable events caused by that particular weld. Then, the overall frequency of unacceptable events caused by breaks in the risk-informed category is simply the sum of these frequencies:

I

<I?= LF(D!mall).

i=l The three factors that determine the frequency of breaks at any particular weld F ( D!mall) for the purposes of our calculation are:

• the percentile of the probability distribution that govern LOCA frequency by break size;

• the smallest break size for the weld that generates a sufficient amount of fiber fines D!mall; and,

• the conditional probability that the break of size D!mall came from a specific weld.

Using a given percentile as input, we calculate the frequency of breaks of a given size across the STPNOC plant, call it f (D!mazz), by looking the value up under the appropriate percentile in Table 1. In the event that a break size falls between two categories, linear interpolation is used to estimate the frequency of breaks of size D!mall across the plant; this is a conservative method of estimation of frequency, as shown in [4]. In the event that a 10

percentile other than the 5th, 50th or 95th is assessed, we can use a Johnson distribution fit to the three percentiles by category to obtain estimates for a given percentile. The methodology for fitting Johnson percentiles to the NUREG-1829 percentiles is shown in detail in [5].

Because a partial break of a weld is possible, any weld of size D!mall or greater is able to generate a break of size D!mall' We assume that any such weld in the plant has an equal likelihood of breaking. Let TWi be the total number of welds of size D!mall or greater. Then, the conditional probability that weld i is responsible for a break of size D!mall is 1

pi ( D~mall) = TWi.

Therefore, the frequency of LOCA events that generate a sufficient amount of fiber fines to be a risk-informed event at weld i is F(Di

)

f (Di

)P (Di

)

f (D!mall) small =

small i

small =

TvV.*

1, and the total risk-informed event frequency is therefore I

I

= ~

F(Di

) = ~ f ( D!mall)

L..J small L..J TvV..

i=l i=l

'/,

3.1.4 DEGB-Only Model If RUFF is used to calculate the frequency of failures related to LOCA events when it is assumed that all such events go to DEGB, we must account for the following two observations:

• the LOCA event frequency must take on a discrete nature rather than operating on a continuum; and,

• the set of candidate pipes that could be responsible for a break of a given size is restricted to the pipes of exactly that size.

To account for these, we perform the same analysis with the following changes.

Let DbEGB be the size of a DEGB at weld i, and let DYJEGB be the size of a DEGB break for the smallest pipe size that exceeds that of DbEGB' if such a size exists. Finally, let TWi+ be the number of welds with a size strictly greater than DbEGB* Then, we redefine the frequency of risk-informed events 11

at weld i caused by a DEGB to be F(Di

) _ f (DbEaB) - f (D2tFJaB)

DEGB -

TYV;* ~ TW*+

1, i

where f ( *) and TWi are defined as in Section 3.1.3. Therefore, the total risk-informed event frequency is I

I

• +

1> = L F(DbEGB) = L f(DbEGB) - f(DDEGB).

i=l i=l TWi - TWi+

Note that if DrJEGB does not exist, that is, DbEGB is the largest possible break size in the plant, then f(DrJEaB) = 0 and TWi+ = 0.

RUFF uses the DEGB-only method of calculation when the binary value DEGB_mode is set to 1 in the configuration file.

3.2 Source Code Listing RUFF is a Python script developed at the University of Texas at Austin under STPNOC grant B04425 and is implemented on OS X for production.

Apple distributes OS X (Release 10.10) with Python; however, Python was updated to a later version (Version 3.4.2) to develop and run RUFF.

The Python libraries scipy, matplotlib, and csv, and time, as well as the open-source PANDAS library (http://pandas.pydata.org/), are used in the RUFF implementation.

3.3 Python Implementation Details The RUFF module contains two classes, NUREG_1829_Freqs and LOCAEventCalculator, as well as one standalone method, ReadConf ig.

3.3.1 Method ReadConfig The following tasks are completed in ReadConfig on initialization, with the location of the configuration file as input:

• Read locations of input and output files

• Read calculation mode of RUFF (DEGB or not) 3.3.2 Class NUREG-1829..Ereqs On initialization, NUREG_1829Yreqs uses the location of the NUREG-1829 table file read by ReadConfig as input to read in the P5, P50, P95 and 12

J\\!Iean values for each category as given in the relevant table in the docu-ment by Tregoning et al. [1]. Within this class, there is a lone accessor getExceedanceFrequency, which determines the exceedance frequency of a given break size, given its size and the summary statistic (i.e., percentile or mean) as input. This is done by obtaining the frequency of the break size in the NUREG-1829 table if it matches a category size, or by perform-ing linear interpolation on the two category sizes between which the given break size resides. The two supporting functions f indFirstExceedingindex and getStat obtain the indices of the relevant categories and their related summary statistics, respectively.

3.3.3 Glass LOCAEventCalculator The following tasks are completed in LOCAEventCalculator on initialization, with the file locations provided by ReadConfig as input_:

• Read a list of welds that can generate a quantity of fiber fines exceeding thresholds established though physical testing, and the minimum such break size that does so

• Read a table summarizing the number of each weld size in the plant

• Create an instance of the class NUREG_1829_Freqs based on the input file at the location determined by ReadConf ig Within this class, there are three main accessors as described below:

1. Break size frequency retrieval:

The frequency of a given break size and summary statistic, defined as is provided by the function getBreakFrequency. In the continuous calcula-tion method, this calls getExceedanceFrequency from the NUREG_1829_Freqs class contained within the LOCAEventCalculator class, and returns f (D!mau) as described in Section 3.1.3. In the DEGB-only method of calculation, the next size higher weld is found via the function getNextWeldSize, and the difference in break frequency between the input weld and the next highest, or f (DbEcB) - f (DHEcB) as described in Section 3.1.4, is returned.

2. Number of candidate pipes retrieval:

13

The number of candidate pipes that could cause a break of a given size is provided by the function getCandidatePipes. In continuous mode, this functions returns TWi by adding the number of welds the input size or greater in the plant. In DEGB mode, this function returns T~ -T'VVi+

by only counting pipes of exactly the same size as the input.

3. Total frequency retrieval:

The total frequency of breaks generating a sufficient amount of fiber fines exceeding the thresholds given by physical testing is provided by the func-tion getSumOfAllBreaks. This function takes as input a given summary statistic and calculates the individual break frequency by dividing the frequency calculated tn getBreakFrequenGy by the number of welds cal-culated-in getCandidatePi pes, for each pipe in the list of welds provided and returns the sum.

14

4 Software Test Plan The software test plan for validating RUFF is simple and straightforward.

We utilize a simple input configuration for testing and compare the results of a hand calculation (done in Excel) to the RUFF results for that same input configuration.

4.1 Testing Methodology RUFF consists of a single Python module and relies only on a limited number of open source Python libraries, which eliminates the need for interface or module testing. Our testing methodology consists of a single system test where we take the input values from the STP submittal to the NRC. We then gather results from RUFF and compare them to an independent hand calculation in J\\!Iicrosoft Excel to validate that RUFF is producing the correct LiCDF estimates.

4.2 Interface Programs RUFF does not interface with any external programs.

4.3 RTE Test Problems In order to run RUFF correctly and accurately, the following minimum soft-ware requirements must be met.

Acceptable operating systems:

• J\\!Iac OS X Release 10.10 or later

• Windows version 8.1 or 10 Acceptable Python distributions:

• Python 2.7.9 or later

• Python 3.4.2 or later Required Python packages:

• scipy

• csv

• .Pandas 15

5 Software Test Case Two test cases were used to validate RUFF. The original test was created and performed by The University of Texas as Austin with verification and validation efforts performed by the University of Texas at Austin and STP.

In the sections below, we provide a description of the test case and results.

5.1 Test Report Two test cases were used to validate RUFF. Results of the test cases are available in Table 2. Details of the baseline test cases are available in Sec-tion 5.3.

Table 2: RUFF SQA Test Case *Report SQA Analyst:

Test Date:

RUFF Version:

Alex Zolan 0.S/15/2015.

v. 1.1 Software Requirements:

See Section 2 of this document.

Test Plan Actual Test Results Pass/Fail Baseline cases as described in Section 5.3 Checked Pass matches solution calculated by spreadsheet TWi matches welds as large or larger than Checked Pass input for continuous method TWi matches welds exactly as large as Checked Pass input for DEGB-only method Update weld size to obtain Checked Pass correct updated solution - continuous Update weld size to obtain Checked Pass correct updated solution - DEGB-only Update number of welds to obtain Checked Pass correct updated solution - continuous Update number of welds to obtain Checked Pass correct updated solution - DEGB-only 5.2 Testing Environment The test cases were run on the machines with the following hardware and software specifications:

5.2.1 Machine 1 Hardware:

l\\!Iodel Name: Mac Pro l\\!Iodel Identifier: J\\t1acPro5,1 Processor: 2.4 GHz 6-Core Intel Xeon 16

Number of Processors: 2 Total Number of Cores: 12 Memory: 64 GB Software:

System Version: OS X 10.11.2 (15C50)

Kernel Version: Darwin 15.2.0 5.2.2 Machine 2 Hardware:

Model Identifier: l\\IIacBookProl0,2 Processor Name: Intel Core i7 Processor Speed: 3 GHz Number of Processors: 1 Total Number of Cores: 2 12 Cache (per Core): 256 KB 13 Cache: 4 l\\IIB Memory: 8 GB Software:

System Version: OS X 10.11.2 (15C50) Kernel Version: Darwin 15.2.0 5.3 Test Cases The two test cases were performed on RUFF, on the revision dated April 19, 2016. The test was conducted on April 19, 2016, by Alexander J. Zolan, Jeremy J. Tejada and John J. Hasenbein at The University of Texas at Austin with verification of test results performed by The University of Texas at Austin and STP.

In order to validate RUFF, we take the list of critical welds and break sizes submitted by STP to the NRC as inputs and compute ;6.CDF estimates using both RUFF and an independently developed Excel calculator.

We now compare the output of RUFF with the hand calculation described in Section 4, using the STP baseline results submitted to the NRC. The inputs to the system are given in Table 3 and the critical welds and break sizes are given in Table 4. Results from the Excel calculator are given in Table 5, results from the Python calculator are given in Table 6, and Table 7 shows the absolute differences between the Excel calculator and the Python calculator. vVe see that the RUFF calculations from both calculation engines match several significant digits, confirming that the Python implementation is correct.

17

Table 3: Summary Data Used in the RUFF Methodolqgy Rover D Information Case 1 Case 2 Threshold (LBM) 191.78 95.89 Case Freqeuncy 4.160E-06 l.550E-09 Case Weight 9.996E-01 3.725E-04

1. of Breaks Failing 53 101 Table 4: STP Critical Welds and Break Sizes Used as Input for Validation Exercise Break Location Case 1 Case 2 1

16-RC-1412-NSS-8 12.814 9.34 2

29-RC-1401-NSS-3 13.98 9.81 3

29-RC-1101-NSS-4 14.17 9.9 4

29-RC-1201-NSS-4 14.4 9.94 5

29-RC-1301-NSS-4 14.41 9.84 6

31-RC-1302-NSS-RSG-1 C-ON-SE 14.78 9.7 7

29-RC-1101-NSS-RSG-lA-IN-SE 15.07 9.76 8

29-RC-1101-NSS-5. l 15.08 9.78 9

29-RC-1401-NSS-RSG-lD-IN-SE 15.1 10.03 10 29-RC-1401-NSS-4.1 15.11 10.04 11 31-RC-1102-NSS-l. l 15.24 9.56 12 31-RC-1102-NSS-RSG-lA-ON-SE 15.24 9.56 13 29-RC-1201-RSG-lB-IN-SE 15.33 9.99 14 31-RC-1202-NSS-1.1 15.35 9.65 15 31-cRC-1202-NSS-RSG-lB-ON-SE 15.35 9.65 16 29-RC-1301-RSG-1 C-IN-SE 15.35 9.92 17 29-RC-1201-NSS-5.l 15.35 9.99 18 29-RC-1301-NSS-5.l 15.36 9.92 19 31-RC-1202-NSS-4 15.54 11 20 31-RC-1202-NSS-2 15.58 9.99 21 31-RC-1102-NSS-2 15.63 10.07 22 RC-1302-NSS-2 15.63 10.14 23 31-RC-1302-NSS-1. l 15.7 9.9 24 31-RC-1102-NSS-4 15.71 11.1 25 31-RC-1302-NSS-4 15.92 11.17 26 29-RC-1101-NSS-1 15.95 15.95 18

27 29-RC-1101-NSS-RPVl-NlASE 15.95 15.95 28 29-RC-1201-NSS-1 15.95 15.95 29 29-RC-1201-RPVl-NlBSE 15.95 15.95 30 29-RC-1301-NSS-1 15.95 15.95 31 29-RC-1301-RPVl-Nl CSE 15.95 15.95 32 29-RC-1401-NSS-l 15.95 15.95 33 29-RC-1401-NSS-RPVl-NlDSE 15.95 15.95 34 31~RC-1402-NSS-2 15.99 10.54 35 31-RC-1402-NSS-RSG-lD-ON-SE 16.49 10.22 36 31-RC-1402-NSS-1.1 16.5 10.22 37 27.5-RC-1303-NSS-1 16.62 10.91 38 27.5-RC-1203-NSS-1 16.62 10.93 39 27.5-RC-1103-NSS-1 16.65 10.99 40 31-RC-1202-NSS-8 16.84 10.88 41 31-RC-1102-NSS-8 17.13 11.09 42 31-RC-1402-NSS-4 17.48 12.35 43 31-RC-1302-NSS-8 17.75 11.28 44 31-RC-1102-NSS-3 17.76 11.37 45 31-RC-1202-NSS-3 17.82 11.31 46 31~RC-1102-NSS-9 18.1 11.49 47 31-RC-1202-NSS-9 18.11 11.48 48 31-RC-1302-NSS-3 18.14 11.38 49 27.5-RC-1403-NSS-1 18.15 12.07 50 31-RC-1302-NSS-9 18.57 11.67 51 31-RC-1402-NSS-3 19.02 12.21 52 31-RC-1402-NSS-8 19.69 13.01 53 31-RC-1402-NSS-9 20.5 13.7 54 12-RC-1221-BBl-9 N/A 9.72 55 12-RC-1221-BBl-11 N/A 9.72 56 12-RC-1125-BBl-9 N/A 9.75 57 12-SI-1315-BBl-8 N/A 9.84 58 12-RC-1322-BBl-1 N/A 9.9 59 12-RC-1322-BBl-lA N/A 9.93 60 12-RC-1125-BBl-11 N/A 9.94 19

61 12-RC-1125-BBl-10 N/A 10.006 62 12-RC-1221-BBl-12 N/A 10.016 63 12-RC-1322-BBl-2 N/A 10.066 64 12-RC-1221-BBl-10 N/A 10.106 65 12-RC-1112-BBl-1 N/A 10.126 66 12-RC-1125-BBl-8 N/A 10.126 67 12-RC-1125-BBl-12 N/A 10.126 68 12-RC-1125-BBl-13 N/A 10.126 69 12-RC-1212-BBl-1 N/A 10.126 70 12-RC-1221-BBl-8 N/A 10.126 71 12-RC-1221-BBl-13 N/A 10.126 72 12-RC-1221-BBl-14 N/A 10.126 73 12-RC-1312-BBl-1 N/A 10.126 74 12-RC-1312-BBl-8 N/A 10.126 75 12-RC-1322-BBl-3 N/A 10.126 76 12-RC-1322-BBl-4 N/A 10.126 77 12-SI-1315-BBl-7 N/A 10.126 78 12-SI-1315-BBl-9 N/A 10.126 79 12-SI-1315-BBl-10 N/A 10.126 80 29-RC-1101-NSS-3 N/A 10.126 81 29-RC-1201-NSS-3 N/A 10.126 82 29-RC-1301-NSS-3 N/A 10.126 83 16-RC-1412-NSS-7 N/A 10.41 84 16-RC-1412-NSS-9 N/A 10.69 85 29-RC-1401-NSS-2 N/A 10.7 86 16-RC-1412-NSS-6 N/A 10.74 87 16-RC-1412-NSS-5 N/A 11.96 88 16-RC-1412-NSS-1 N/A 12.814 89 16-RC-1412-NSS-PRZ-l-Nl-SE N/A 12.814 90 27.5-RC-1103-NSS-RPV1-N2ASE N/A 22.57 91 27.5-RC-1203-NSS-5 N/A 22.61 92 27.5-RC-1203-NSS-RPV1-N2BSE N/A 22.68 93 27.5-RC-1203-NSS-4 N/A 24.65 94 27.5-RC-1103-NSS-6 N/A 24.94 20

95 27.5-RC-1103-NSS-7 N/A 25.41 96 27.5-RC-1303-NSS-RPV1-N2CSE N/A 26.78 97 27.5-RC-1303-NSS-6 N/A 26.82 98 27.5-RC-1303-NSS-5 N/A 27.5 99 27.5-RC-1403-NSS-5 N/A 27.5 100 27.5-RC-1403-NSS-6 N/A 27.5 101 27.5-RC-1403-NSS-RPV1-N2DSE N/A 27.5 Table 5: Results from Excel Calculator Continuum Break Model Geometric Means Arithmetic Means Quantile Case 1 Case 2 Delta CDF Case 1 Case 2 Delta CDF 5th 3.30E-10 4A1E-09 3.31E-10 7.SSE-09 2.68E-08 7.89E-09 50th 9.44E-09 9.92E-08 9.47E-09 2.05E-07 5.50E-07 2.05E-07 Mean 1.50E-07 5.90E-07 l.50E-07 1.90E-06 4.31E-06 l.90E-06 95th 4.38E-07 2.15E-.06 4.39E-07 5.78E-06 l.36E-05 5.79E-06 DEGB-Only Break Model Geometric Means Arithmetic Means Quantile Case 1 Case 2 Delta CDF Case 1 Case 2 Delta CDF 5th 3.47E-10 3.35E-09 3.49E-10 7.78E-09 2.23E-08 7.78E-09 50th 8.62E-09 7.60E-08 8.64E-09 l.92E-07 4.72E-07 l.92E-07 Me;i,n 9.02E-08 4.82E-07 9.03E-08 1.82E-06 3.SOE-06 l.82E-06 95th 2.84E-07 l.72E-06 2.84E-07 6.02E-06 l.19E-05 6.03E-06 21

Table 6: Results from RUFF Python Calculator Continuum Break Model Geometric Means Arithmetic Means Quantile Case 1 Case 2 Delta CDF Case 1 Case 2 Delta CDF 5th 3.30E-10 4.41E-09 3.31E-10 7.88E-09 2.68E-08 7.89E-09 50th 9.44E-09 9.92E-08 9.47E-09 2.05E-07 5.50E-07 2.05E-07 Mean 1.50E-07 5.90E-07 l.50E-07 l.90E-06 4.31E-06 l.90E-06 95th 4.38E-07 2.15E-06 4.39E-07 5.78E-06 l.36E-05 5.79E-06 DEGB-Only Break Model Geometric Means Arithmetic Means Quantile Case 1 Case 2 Delta CDF Case 1 Case 2 Delta CDF 5th 3.47E-10 3.35E-09 3.49E-10 7.78E-09 2.23E-08 7.78E-09 50th 8.62E-09 7.60E-08 8.64E-09 l.92E-07 4.72E-07 l.92E-07 Mean 9.02E-08 4.82E-07 9.03E-08 l.82E-06 3.80E-06 l.82E-06 95th 2.84E-07 l.72E-06 2.84E-07 6.02E-06 l.19E-05 6.03E-06 Table 7: Absolute Differences Between Excel Calculator and Python Calculator Continuum Break Model Geometric Means Arithmetic Means Quantile Case 1 Case 2 Delta CDF Case 1 Case 2 Delta CDF 5th 3.22E-22 8.25E-22 3.23E-22 4.23E-21 7.53E-21 4.24E-21 50th 3.58E-21 3.09E-20 3.59E-21 l.59E-19 3.69E-19 1.59E-19 Mean 9.58E-21 4.55E-21 9.58E-21 3.58E-18 l.OlE-18 3.58E-18 95th l.61E-20 6.77E-19 l.64E-20 2.82E-18 l.41E-17 2.81E-18 DEGB-Only Break Model Geometric Means Arithmetic Means Quantile Case 1 Case 2 Delta CDF Case 1 Case 2 Delta CDF 5th 4.88E-22 3.89E-23 4.88E-22 l.85E-21 3.40E-20 l.83E-21 50th 3.30E-21 l.54E-20 3.30E-21 4.22E-20 4.25E-19 4.20E-20 Mean 4.llE-20 4.90E-19 4.08E-20 2.94E-18 3.92E-18 2.94E-18 95th 7.61E-20 3.32E-18 7.73E-20 4.79E-18 3.51E-17 4.SOE-18 22

6 User Instructions This section describes the prerequisite knowledge and skills for running RUFF, the limitations of the software implementation, a user guide to running RUFF, and a description of the inputs and outputs of the software.

6.1 Training Needs Assessment In order to run RUFF, the user should be able to install a Python interpreter and some common Python packages. Further, the user should be able to execute software via the operating system's command line.

6.2 Major Assumptions and Limitations The test case represents a typical calculatio~ for RUFF, and it is expected that in most practical cases, there should be no issues with the accuracy of the calculations.

6.3 User Instructions for Personal Computer Software Use of RUFF requires the installation of a Python interpreter and the follow-ing packages: scipy, csv, and pandas.

To run RUFF, create the four input files as described in Section 6.4 and save them in the same directory as RUFF. Open a command prompt and navigate to the directory where RUFF is located. Then, execute the following command to run RUFF:

python RUFF_v1_1.py After calling the module through a Python interpreter, the following pro-cedure takes place:

• The function ReadConf ig is run, reading the locations of all input files.

Any missing or broken file references throw an exception and terminate the program.

• The class LOCAEventCalculator is initialized. This includes an initializa-tion of the NUREG_1829_freqs class, which is stored within the instance of LOCAEventCalculator.

• Th,e 5th, 50th, and 95th percentile, and mean estimates of , as defined 23

in Section 3.1.3 for the continuum model and Section 3.1.4 for the DEGB-only model, are calculated.

• The four estimates from the previous step and an echo of the inputs are printed to the output file specified in the configuration file.

6.4 Input Description The inputs are in the form of four fiat files, one of which is a configuration.

file read via the function ReadConf ig, using Python's 'csv' library. The other three fiat files are read using Python's 'pandas' library (http://pandas.pydata.org/).

The first fiat file is a configuration file read by the 'csv' library, has five lines, and takes on the following form as an example:

breaks_file,breaks1DEGB.csv welds_file,welds.csv nureg_file,NUREG_GM.csv output_file,ruff-DEGB1.csv DEGB_mode, 1 The field after the header breaks_file points to the location of the file containing the list of welds for which a break can generate enough fiber fines to exceed the threshold from physical testing. The field after the header breaks_file points to the location of the file containing the number of all welds by size in the plant. The field after the header nureg_file points to the location of the table of NUREG-1829 frequencies to use in the calculation of . The field after the header output_file points to the the location of the output file after RUFF is run. The field after the header DEGB_mode is set to one if the calculation method is for only DEGB events, and set to zero otherwise.

The second fiat file, referred to as breaks_file in the configuration file, is indexed by time and takes on the following form as an example:

1. ,Location,Break_size 1,16-RC-1412-NSS-8,12.814 2,29-RC-1401-NSS-3,13.98 24

The header # represents the weld number in the list, while the header Location indicates the location of the weld in the plant for record keeping purposes. The header Break_size indicates the size of the smallest break (in inches) that generates a sufficient amount of fiber fines to exceed the threshold, as determined by CASA Grande runs [3].

The third fiat file, referred to as welds_file in the configuration file, con-sists of inputs that are indexed on pipe size, and takes on the following form as an example:

pipe_type,number_of_welds 0.612,32 0.815,3 The header pipe_type represents the maximum break size (in inches) that a weld type in the plant can generate. The header number_of_welds rep-resents the number of such welds in the plant, including those that do not generate a sufficient quantity of fiber fines.

The fourth fiat file, referred to as nureg_file in the configuration file, consists of inputs that are indexed on summary statistic and category, and takes on the following form as an example:

Category,Break_Size,P5,P50,Mean,P95 Cat1,0.5,8.10E-04,4.80E-03,1.00E-02,3.60E-02 Cat2,1.625,4.20E-05,7.00E-04,3.00E-03,1.20E-02 The header Category represents the NUREG-1829 category number, while Break_Size represents the break size for the category in the NUREG-1829 estimates by Tregoning et al. [1]. P5, P50, and P95 represent the 5th, 50th, and 95th percentile estimates of LOCA frequency for that category break size according to the chosen NUREG-1829 table, respectively. Mean represents the estimated mean LOCA frequency estimate for the category according to the chosen NUREG-1829 table.

25

6.5 Output Description The output consists of a single flat file. The output flat file takes on the following form as an example:

Total estimated frequency at percentiles and mean:

P5,2.65353107434e-10 P50,7.53395953753e-09 P95,3.44566288184e-07 Mean,1.17670681638e-07 The headers P5, P50, and P95 denote the 5th, 50th, and 95th percentile estimates of the sum frequency of events that generate a sufficient quantity of fiber fines to exceed the given. threshold for all welds in the plant, while Mean denotes the estimated mean frequency for the same.

6.6 Known Issues and Workarounds RUFF has no known issues, provided the inputs are valid as described in Section 3.1.

6.7 User Support For technical support in running RUFF, contact John Hasenbein at jhas@mail.utex-as.edu or Ernie Kee at erniekee@ykriskllc.com.

. 26

References

[1] Tregoning, R., L. Abramson, and P. Scott (2008, April). Estimating Loss-of-Coolant Accident (LOCA) Frequencies Through the Elicitation Pro-cess. Technical Report NUREG/CR 1829, Nuclear Regulatory Commis-sion, Washington, DC.

[2] Eide, S., T. Wierman, C. Gentillon, D. Rasmuson, and C. Atwood (2007, February). Industry-Average Performance for Components and Initiat-ing Events at U.S. Commercial Nuclear Power Plants. Technical Report NUREG/CR 6928, Nuclear Regulatory Commission, Washington, DC.

[3] Allon Science and Technology (2015, April). CASA Grande Theory Man-ual. ALION-SPP ALION-I009-10, Allon Science & Technology, Albu-querque, NJ\\!L

[4] Hasenbein, John J. (2015, l\\1ay). Order Relations for Concave Interpo-lation :Methods. Technical Report, 2015-001, The University of Texas at Austin.

[5] Popova, E., D. :Morton, and Y. Pan (2013, January). Modeling and Sam-pling LOCA Frequency and Break Size for STP GSI-191 Resolution. Tech-nical report, STP-RIGSI191-V03.02, Revision 4, The University of Texas at Austin.

27

Appendix A: Source Code This implementation of RUFF is a Python script developed at the University of Texas at Austin under STPNOC grant B04425. The open-source PANDAS library (http://pandas.pydata.org/) is used in the RUFF implementation.

111111 Risk Unifying Frequency Functional (RUFF) version 1.1 Alex Zolan Updated April 19, 2016 The purpose of the program is to estimate the frequency of critical breaks that can occur We assume that any pipe that has a diameter as large or larger than any critical break size could experience such a break, and that each possible pipe has the same chance of having such a break.

version 1.1 updates:

1111 II

- added DEGB mode to conf ig file (DEGB_mode will run DEGB calcs if se-and the continuous model if set to zero).

DEGB mode defines the brea frequency as the difference between the input _size exceedance frequen exceedance frequency of the smallest pipe size larger than the input.

the number of candidate pipes is the number of pipes of exactly the i:

size in DEGB mode (unlike continuous mode which has the pipes of size greater than or equal to the inputs size).

import pandas import scipy import csv def ReadConfig(config_filename):

"""Serves as the input reader for this model.

Assumes there is one file that reads as a table of time-based inputs and another file with initial and model values.

the output is a dictionary that is used to initialize the MassCalculator class.

1111 II 28

params = {}

1. read in initials and constants file configFile = csv.reader(open(config_filename, 'rU'))

for line in configFile:

if len(line) > 1:

try: params[line[O)) = int(line[1])

except ValueError: params[line[O)) = line[1]

return params class NUREG_1829_Freqs(object):

"""This class manages the NUREG-1829 frequencies as given by an input file, which has the the break size, mean, and 5th, 50th and 95th exceedance break frequencies for a set number of categories. 111111 def __ init __ (self,nureg_file):

111111 We start with a dataframe and take the break sizes and each sumimary statistic as their own independent list.

111111 df = pandas.read_csv(nureg_file)

1. self.categories= df.Category.values self.sizes= df.Break_Size.values self.means= df.Mean.values self.PS = df.P5.values self.P50 = df.P50.values self.P95 = df.P95.values def findFirstExceedingindex(self,size):

1111 "Finds the index of the first size that is larger than the given size. --

size -- break size, in inches retval - index from sizes object"""

assert size>= self.sizes[O], "Size outside of NUREG found.

Abor for idx, sin enumerate(self.sizes):

if s >= size: return idx 29

return -1 def getExceedanceFrequency(self,size,stat):

"""Returns the exceedance frequency of a given break size uses NUREG 1829 values and linear interpolation to find the best estimate of breaks that size of larger.

size -- break size, in inches stat -- desired summary statistic retval - summary statistic frequency for break size'""'

idx =self.findFirstExceedingindex(size) assert idx >= 0, "Size outside of NUREG Found.

Aborting."

if idx== 0: return self.getStat(O,stat) lower= self.getStat(idx-1,stat) upper= self.getStat(idx,stat) frac = (size-self.sizes[idx-1])/(self.sizes[idx]-self.sizes[idx-f

1. print "Frac Calc",size,lower,upper,self.sizes[idx-1],self.sizes[

return lower + (upper-lower)*frac def getStat(self,idx,stat):

"""Returns a summary statistic based on the object desired.

idx -- index of the desire list to return stat -- desired sUinmary statistic retval - summary statistic frequency from NUREG-1829 111111 if stat --

11P6 11 : return self. P5 [idx]

if stat -- "P50 11 : return self. P50 [idx]

if stat -- "P95": return self.P95 [idx]

if stat -- "Mean": return self. means [idx]

class LOCAEventCalculator(object):

"""This class acts as the calculator for LOCA Events. It calls frequencies from the NUREG_1829_Freqs object, and determines the probability of a particular pipe breaking by finding the number of pipes that could handle such a break.

30

breaksFile -- location of the file that contains all pipes and the weld break sizes that would cause a significant event weldsFile -- location of the file that contains a summary of the number of welds of each size/type 111111 def __ init __ (self, configFile):

self.config = ReadConfig(configFile) self.breaks_df = pandas.read_csv(self.config['breaks_file'])

self.welds_df = pandas.read_csv(self.config['welds_file'])

self.nureg = NUREG_1829_Freqs (self. config[ 'nureg_file ']) _

def getCandidatePipes(self,size):

"""returns the number of pipes in from the welds dataframe that have a diameter that could be responsible for a given break In DEGB only mode, this is the number of pipes of exactly that si: _

Otherwise, it is the number of pipes that meet or exceed that siz*

if self.config["DEGB_mode"]== 1:

return scipy.sutn(self.welds_df[self.welds_df.pipe_type

== size].number_of_welds.values) else:

return scipy.sum(self.welds_df[self.welds_df.pipe_type

>= size].number_of_welds.values) def getBreakFrequency(self,size,stat):

""."Returns the frequency of a given break size uses NUREG 1829 values and linear interpolation to find the best estimate of a break of that size.

In DEGB Only mode, th the frequency of a break of an exact size (measured by the exceed.

frequency minus the exceedance frequency of the next pipe size up Otherwise, in the continuum world, this is just the exceedance frequency.

size -- break size, in inches stat -- desired summary statistic retval - summary statistic frequency for break size 111111 freq= self.nureg.getExceedanceFrequency(size,stat) 31

if self.config[

11DEGB_mode 11 ]== 0: return freq else:

next_size =self.getNextWeldSize(size) if next_size== None: return freq next_freq = self.nureg.getExceedanceFrequency(next_size,stat return freq - next_f req def getNextWeldSize(self,size):

111111finds the smallest pipe size that exceeds the pipe size given inputs. Used in DEGB mode for calculating a discrete frequency (a opposed to a continuous exceedance frequency.

size -- pipe size, in inches (given by pipe type) retval - next pipe size up, in inches 111111 if self.welds_df.pipe_type.values.max()==size: return None else: return self.welds_df[self.welds~df.pipe~type

> size].pipe_type.values.min()

def getSumDfAllBreaks(self,stat):

111111calculates the expected frequency of LDCA events based on calculating the exceedance frequency of the break size and then dividing by the number of pipes that could have a break of that size in the plant (as given by the welds file).

This term is calculated for each pipe in the pipebreaks file (when a nonzero break size is included) and then summed to get the result.

stat -- desired percentile of breaks.

retval -- expected frequency of LOCA events/CY. 111111 sum_freqs = 0.0 for i,rowdata in self.breaks_df.iterrows():

if self.breaks_df.Break_size[i]== 0: continue breakFreq =self.getBreakFrequency(self.breaks_df.Break_size[

numPipes =self.getCandidatePipes(self.breaks_df.Break_si~e[f sum_f reqs += breakFreq I numPipes 32

1. print self.breaks_df.Break_size[i],breakFreq,num.Pipes return sum_freqs def outputSummaryStats(self):

'""'outputs summary of estimated LOCA frequencies at different percentiles."""

P5Freq =self.getSum.OfA11Breaks("P5")

P50Freq =self.getSumOfA11Breaks("P50")

meanFreq =self.getSum.OfAllBreaks("Mean")

P95Freq =self.getSumOfA11Breaks("P95")

outfile = open(self.config['output_file'], 'w')

outfile:write('Total estimated frequency at percentiles and mean:*

outfile.write("P5,"+str(P5Freq)+"\\n")

outfile.write("P50,"+str(P50Freq)+"\\n")

outfile.write("Mean,"+str(meanFreq)+"\\n")

outfile.write("P95,"+str(P95Freq)+"\\n")

outf ile. close()

def outputEcholn(self):

I

"""echoes inputs to the output file"""

outfile = open(self.config['output_file'], 'a')

outfile.write("\\n\\ninputs:\\n\\nNUREG Frequencies:\\n")

outfile.write("Size,P5,P50,P95,Mean\\n")

for i in range(len(self.nureg.sizes)):

outfile.write(str(self.nureg.sizes[i])+","+

str(self.nureg.P5 [i] )+", "+

str(self.nureg.P50[i] )+", "+

str(self.nureg.P95[i] )+", "+

str(self.nureg.means[i])+"\\n")

outfile.write("\\n\\nWelds Inputs:\\n pipe_type,number_of_welds\\n")

for i in range(len(self.welds_df)):

outfile.write(str(self.welds_df.pipe_type[i])+","+

str(self.welds_df.number_of_welds[i])+"\\n")

outfile. write("\\n\\nBreaks Inputs: \\n weld_number,Location,Break_si:

for i in range(len(self.breaks_df)):

outfile.write(str(i+1)+","+

33

str(self.breaks_df.Location[i])+", 11+

str(self.breaks_df.Break_size[i])+"\\n")

def outputSolution(self):

"""outputs all calculations and echoes inputs to file. 111111 self.outputSummaryStats()

self.outputEchoin()

if __ name __==

11 __ main __ 11 :

configFile = raw_input("Please enter the name of the configuration fi locas = LDCAEventCalculator(configFile) locas.outputSolution()

print "calculations completed. output files created.

11 34 Commitments NOC-AE-16003390

Commitments Commitment The STP licensing application evaluation is a PRA Analysis Assessment, which is a bounding evaluation of debris effects that calculated only delta-GDF and delta-LERF. As an assessment, it is not incorporated into the PRA. STPNOC will meet the 48-month requirement by reviewing relevant elements of the assessment to determine if there were any significant changes that would affect the conclusions.

(Response to APLA RAI 4-5.d)

Action 11-4249-305 NOC-AE-16003390 Page 1 of 1 Due 11/30/2020 Commitments NOC-AE-16003390

Commitments Commitment The STP licensing application evaluation is a PRA Analysis Assessment, which is a

bounding evaluation of debris effects that calculated only delta-GDF and delta-LERF. As an ;;issessment, it is not incorporated into the PRA. STPNOC will meet the 48-month requirement by reviewing relevant elements of the assessment to determine if there were any significant changes that would affect the concl_usions.

(Response to APLA RAI 4-5.d)

Action 11-4249-305 NOC-AE-16003390 Page 1of1 Due 11/30/2020 Definitions and Acronyms NOC-AE-16003390

NOC-AE-16003390 Page 1of2 Definitions and Acronyms ANS American Nuclear Society EOF Emergency Operations ARL Alden Research Laboratory Facility ASME American Society of EOP Emergency Operating Mechanical Engineers Procedure(s)

BA Boric Acid EPRI Electric Power Research BAP Boric Acid Precipitation Institute BC Branch Connection EQ Equipment Qualification BEP Best Efficiency Point ESF Engineered Safety Feature B-F Bimetallic Welds FA Fuel Assembly( s)

B*J Single Metal Welds FHB Fuel Handling Building BWR Boiling Water Reactor GDC General Design Criterioil(ia)

CAD Computer Aided Design GL Generic Letter CASA Containment Accident GSI Generic Safety lssu~

Stochastic Analysis, also a HHSI High Head Safety Injection short name for the CASA (ECCS Subsystem)

Grande computer program HLB Hot Leg Break that uses the analysis HTVL High Temperature Vertical methodology Loop CCDF Complementary Cumulative HLSO Hot Leg :switchover Distribution Function or HVAC Heating, Ventilation & Air Conditional Core Damage Conditioning Frequency ID Inside Diameter ccw Component Cooling Water IGSCC lntergranular Stress CDF Core Damage Frequency Corrosion Cracking CET Core Exit Thermocouple(s)

ISi In-Service Inspection CHLE Corrosion/Head Loss IOZ Inorganic Ziilc Experiments LAR License Amen.dment CHRS Containment Heat Removal Request System LBB Leak Before Break CLB Cold Leg Break or Current LBLOCA Large Break Loss of Coolant Licensing Basis Accident (also LLOCA)

CRMP Configuration Risk LCO Limiting Condition for Management Program Operation cs Containment Spray LDFG Low Density Fiberglass CSHL Clean Strainer Head Loss LERF Large Early Release css Containment Spray System Frequency _

(same as CS)

LHS Latin Hypercube Sampling eves Chemical Volume Control LHSI Low Head Safety Injection System (ECCS Subsystem)

DBA Design Basis Actident LOCA Loss of Coolant Accident DBD Design Basis Document LOOP/LOSP Loss of Off Site Power D&C Design and Construction MAAP Modular Accident Analysis Defects Program DEGB Double Ended Guillotine MAB/MEAS Mechanical Auxiliary Building Break or Mechanicai Electrical DID Defense in Depth Auxiliary Building DM Degradation Mechanism MBLOCA Medium Break Loss of ECCS Emergency Core Coolin.g Coolant Accident (also System (also ECC)

MLOCA)

ECWS Essential Cooling Water NIST National Institute of System (also ECW)

• Standards and Technology

NOC-AE-16003390 Page 2 of 2 Definitions and Acronyms NLHS Non-uniform Latin Hypercube SI/SIS Safety Injection, Safety Sampling Injection Syste:m (same as NPSH Net Positive Suction Head, ECCS)

(NPSHA - available, NPSHR SIR Safety Injection and

-required)

Recirculation NRC Nuclear Regulatory SR Surveillance Requirement Commission SRM Staff Requirements NSSS Nuclear Steam Supply Memorandum System SSE Safe Shutdown Earthquake OBE Operating Basis Earthquake STP South Texas Project OD Outer Diameter STPEGS South Texas Project Electric OQAP Operations Quality Generating Station Assurance Plan STPNOC STP Nuclear Operating PCI Performance Contracting, Company Inc.

TAMU Texas A&M University PCT Peak Clad Temperature TF Thermal Fatigue PDF Probability Density Function TGSCC Transgranular Stress PRA Probabilistic Risk Corrosion Cracking Assessment TS Technical Specification(s)

PWR Pressurized Water Reactor TSB Technical Specification PW ROG Pressurized Water Reactor Bases Owner's Group TSC Technical Support Center or PWSCC Primary Water Stress Technical Specification Corrosion Cracking Change QA Quality Assurance TSP Trisodium Phosphate QDPS Qualified Display Processing UFSAR Updated Final Safety System Analysis Report RAI Request for Additional

• UNM University of New Mexico Information USI Unresolved Safety Issue RCB Reactor Containment UT University of Texas (Austin)

Building V&V Verification and Validation RCFC Reactor Containment Fan VF Vibration Fatigue Cooler WCAP Westinghouse Commercial RCS Reactor Coolant System Atomic Power RG Regulatory Guide ZOI Zone of Influence RHR Residual Heat Removal RI-ISi Risk-Informed In-Service Inspection RMI Reflective Metal Insulation RMS Records Management System RMTS Risk Managed Technical Specifications RVWL(S)

Reactor Vessel Water Level (System)

RWST Refueling Water Storage Tank SBLOCA Small Break Loss of Coolant Accident (also SLOCA)

SC Stress Corrosion

• .\\i>

DRP-003