ML13270A176

Redacted - Office of Nuclear Reactor Regulation (NRR) Reactor Systems Branch (Srxb) Support of Region II Inspection of H. B. Robinson Treatment of Voids in Systems That Are Important to Safety

ML13270A176
Person / Time
Site:	Robinson
Issue date:	09/24/2013
From:	NRC/NRR/DSS/SRXB, NRC/RGN-II
To:	Plant Licensing Branch II
Lingam S
Shared Package
ML13270A169	List: ML13270A176
References
Download: ML13270A176 (106)

Text

September 24, 2013 Office of Nuclear Reactor Regulation (NRR) Reactor Systems Branch (SRXB)

Support of Region II Inspection of H. B. Robinson Treatment of Voids in Systems that are Important to Safety PREFACE This report (1) contains proprietary information that is not available to some licensees of nuclear power plants, (2) contains information provided by H. B. Robinson as a result of a Region II inspection, (3) contains GOTHIC information provided to NRC by the Electric Power Research Institute (EPRI) that the NRC understands is available to authorized users of the GOTHIC computer program, and (4) provides an in-depth assessment of application of the GOTHIC computer code to predict gas movement in nuclear power plant piping systems. The report has been updated from the May 28, 2013 version originally provided to Region II to be consistent with Region IIs completion of the inspection and the restriction to NRC only has been removed since the pre-decisional aspects no longer apply.

This report was provided to Region II to support an inspection and normally would not be retained once the inspection was complete. Much of the information covered in this report was identified as proprietary in the references that were consulted. Other information was provided for the inspection by the licensee. Such information normally would not be retained by the NRC and consequently, it would not be publically available. However, this report is unique since the gas movement methodologies it addresses have not been previously reviewed in detail by the NRC staff. Making this information available will benefit both the NRC staff and licensees in other reviews and inspections. Consequently, the report has been retained. The licensee has provided an affidavit that covers withholding such material (Gideon, August 8, 2013). A vertical bar along the left side of the text, and occasionally brackets [ ], are used to identify information that is not publically available. The report will be provided to the Robinson licensee who, at its option, can make it available to any other organization authorized access to the proprietary content such as Nuclear Applications, Inc. It will also be available to the NRC staff.

Region II opened unresolved URI 05000261/2012005-03, Questions Regarding Whether GOTHIC is Sufficiently Qualified for Use in Operability Determinations, and stated that this issue will remain unresolved pending additional inspection and consultation with a GOTHIC subject matter expert at NRC headquarters to evaluate the licensees use of GOTHIC to support operability determinations. Region II requested NRR/SRXBs Warren Lyons assistance to address this unresolved item (URI) (Nesse, December 4, 2012).

Region IIs description of the URI was as follows (Musser, February 6, 2013):

Information Notice 2011-17, issued July 26, 2011, informed addressees of recent instances of gas accumulation in safety-related systems in which the resulting operability determination of the as-found condition relied on computer models (i.e., GOTHIC) that were not demonstrated to be technically appropriate for the intended application. Specifically, the computer models had not been sufficiently qualified by benchmarking against test or plant data.

Page 1

The inspectors reviewed information related to the licensees response to GL 2008-01 and determined that the licensee had found voids in the SI system, RHR system, and CS piping. In most instances, the licensee had used GOTHIC to evaluate the past operability of the subject systems with voids, and then vented the gas prior to returning the subject systems back to service. The licensee had also evaluated the continued operability of the subject systems with a void left in place until corrective actions were implemented. Specifically, in 2008, the licensee evaluated eight gas voids found following filling and venting of the subject systems that could not be successfully removed during RO-25. The inspectors observed that the licensee used the GOTHIC as part of these evaluations to perform analysis of gas movement to predict how a void volume in piping is translated into a transient void fraction at the entrance of the pumps.

The evaluations were the basis for the continued operability until corrective actions could be taken to remove the voids during the following RO-26, approximately 19 months later.

While acknowledging the NRCs concerns that the GOTHIC models may not be sufficiently qualified by benchmarking against test or plant data for the particular gas transport scenario and piping configuration being analyzed, the licensee prepared engineering change document EC 86423 to document their justifications for continued use of the GOTHIC models to support operability determinations.

The inspectors determined that this issue will remain unresolved pending additional inspection and consultation with a GOTHIC subject matter expert at NRC headquarters to evaluate the licensees use of GOTHIC to support operability determinations.

This report provides Warren Lyons assessment of selected information to support Region IIs request. The report has not been reviewed by other members of the NRC staff.

Page 2

SUMMARY

This report addresses Region IIs URI 05000261/2012005-03, Questions Regarding Whether GOTHIC is Sufficiently Qualified for Use in Operability Determinations. It evaluates the Carolina Power and Light Company use of GOTHIC at the H.B. Robinson Steam Electric plant to support operability determinations.

Robinsons current design basis with respect to voids in the subject systems is a water-solid, no gas condition. This means the subject systems must be water-solid when transitioning from an outage into power operation. Once the transition is complete, in recognition of the possibility that voids will form during operation, such voids are acceptable provided operability is reasonably maintained. Robinson has used GOTHIC to support operability assessments.

GOTHIC is a multi-dimensional, multi-component computer code with the capability to model two phase flow in nuclear power plant systems. Robinson contracted with Nuclear Applications, Inc. (NAI) to apply GOTHIC to predict behavior associated with gas in the ECCS and RHR suction and discharge pipes at the H. B. Robinson plant. The complexities in the use of GOTHIC means that the modeler must be highly qualified to ensure system modeling is consistent with test modeling and to acceptably capture nuances that are unique to the system model. NAI meets this requirement.

GOTHIC was compared to a broad range of test conditions and to tests that directly simulated aspects of behavior that may occur in plant piping. As work progressed, different GOTHIC versions were used for the calculations to address GOTHIC shortcomings and, in some cases, there were significant quantitative differences in the results. The need for continuing changes at the time of this NAI work (2008 - 2010) reflects the immaturity of GOTHIC for study of the problems of concern here. Differences between versions and with respect to test data must be considered when using GOTHIC predictions.

With respect to operability, the material reviewed during this inspection supports a finding that:

(1) GOTHIC has been qualified to predict gas transport behavior for assessment of operability provided that (a) the pump inlet void fractions and volumes predicted by GOTHIC are shown to be acceptably conservative 1, (b) appropriate, generally accepted modeling methodologies are used in the GOTHIC calculations that are consistent with the methodologies reviewed within this report, and (c) GOTHIC predicted results are consistent with simplified methodologies, such as the Froude number, when those methodologies apply.

(2) An in-depth audit evaluated GOTHICs calculation of the effect of trapped gas in an elevated 95 foot long 10 inch diameter pipe during initiation of flow from the containment emergency sump. The GOTHIC calculation was found to have acceptably determined gas volumes that would not jeopardize operability. The findings regarding modeling detail support a conclusion that the licensees use of GOTHIC to evaluate the potential movement of trapped gas with respect to operability is acceptable.

(3) Predicted water hammer results are considered to be semi-quantitative.

1 Typically, a factor of two should be used.

Page 3

TABLE OF CONTENTS PREFACE .................................................................................................................................. 1

SUMMARY

................................................................................................................................ 3 TABLE OF CONTENTS ............................................................................................................. 4 TABLE OF FIGURES ................................................................................................................. 6 ACRONYMS AND ABBREVIATIONS ........................................................................................ 8

1. INTRODUCTION ...............................................................................................................10

2. RESOLUTION OF VOID ISSUES ......................................................................................11 2.1 Regulatory Requirements............................................................................................................ 11 2.2 Void Concerns ............................................................................................................................ 13 2.2.1 Pump Response to Voids ................................................................................................ 13 2.2.2 Water Hammer ................................................................................................................ 16 2.2.3 Delay ............................................................................................................................... 16 2.2.4 Vortex Concerns ............................................................................................................. 16 2.2.5 Net Positive Suction Head (NPSH)................................................................................. 17 2.3 Prediction of Void Movement..................................................................................................... 17 2.3.1 Froude Number ............................................................................................................... 17 2.3.2 Simplified Methodologies ............................................................................................... 18 2.3.3 In-depth Thermal-Hydraulic Code Applications ............................................................ 18 2.3.4 Conservatisms Associated with Pump Void Fraction Predictions .................................. 19 2.4 Test Issues ................................................................................................................................... 19

3. USE OF GOTHIC TO PREDICT GAS / WATER BEHAVIOR IN PIPING SYSTEMS ..........20 3.1 GOTHIC Modeling ..................................................................................................................... 20 3.2 GOTHIC Configuration Control ................................................................................................. 26 3.3 Comparison of GOTHIC to Test Data ........................................................................................ 31 3.3.1 GOTHIC Version Qualification: Comparisons to Test Data .......................................... 31 3.3.1.1 Edwards Blowdown Experiment................................................................................. 32 3.3.1.2 Two Phase Flow .......................................................................................................... 34 3.3.1.3 Water Hammer ............................................................................................................ 37 3.3.1.3.1 Wave Velocity as Function of Pipe Material ........................................................ 37 3.3.1.3.2 Effect of Air on Wave Speed ................................................................................ 37 3.3.1.3.3 EPRI Waterhammer Benchmark Problems........................................................... 38 3.3.1.3.4 Delft Hydraulics Laboratory Test ......................................................................... 39 3.3.1.3.5 Fauske and Associates (FAI) Pressure Surge Tests .............................................. 40 3.3.2 Purdue Tests .................................................................................................................... 42 Page 4

3.3.2.1 Comparison of GOTHIC to Purdue 6 Inch Test Data ................................................. 44 3.3.2.2 Comparison of GOTHIC to Purdue 8 Inch Test Data ................................................. 55 3.3.2.3 Comparison of GOTHIC to Purdue 4 Inch Test Data ................................................. 56 3.3.2.4 Comparison of GOTHIC to Purdue 12 Inch Test Data ............................................... 59 3.3.2.5 Assessment of GOTHIC Prediction of Lower Vertical and Lower Horizontal Pipe Purdue Tests ................................................................................................................................ 63 3.3.3 Comparison of GOTHIC to Millstone Test Data ............................................................ 68 3.3.4 Assessment of GOTHIC Comparisons to Test Data ....................................................... 72 3.4 General Treatment of Conservatism ........................................................................................... 74 3.5 Conclusions Regarding Use of GOTHIC to Predict Piping System Behavior ........................... 75

4. ASSESSMENT OF SELECTED ASPECTS OF VOIDS IN ROBINSON SYSTEMS ...........77 4.1 The Current Design Basis (CDB) ............................................................................................... 77 4.2 Consideration of Water Hammer ................................................................................................ 77 4.3 Use of GOTHIC to Predict Suction Pipe Void Behavior at Robinson........................................ 78 4.4 Discharge Piping and Water Hammer ........................................................................................ 90

5. CONCLUSIONS.................................................................................................................97 5.1 Void Transport ............................................................................................................................ 99 5.2 Water Hammer .......................................................................................................................... 100 5.3 Summary of Inspection Findings .............................................................................................. 100

6. REFERENCES ................................................................................................................102 Page 5

TABLE OF FIGURES Figure 1. GOTHIC Volume Noding Diagram ...................................................................................... 20 Figure 2. Upper Horizontal Pipe Noding .............................................................................................. 21 Figure 3. Vertical Pipe Noding .............................................................................................................. 22 Figure 4. Lower Horizontal Pipe Noding .............................................................................................. 22 Figure 5. Lumped Volume Connection ................................................................................................ 23 Figure 6. Subdivided to Subdivided Connection ................................................................................ 23 Figure 7. Two elbows ............................................................................................................................. 24 Figure 8. Nodalization for Two Elbows ................................................................................................ 24 Figure 9. Two Elbow Configuration ...................................................................................................... 24 Figure 10. Two Elbow Nodalization ...................................................................................................... 25 Figure 11. Tee Configuration................................................................................................................. 25 Figure 12. Tee Nodalization .................................................................................................................. 26 Figure 13. Comparison of 7.2aWC(QA) and 7.2aWC2(QA) Near Top of Downcomer ................ 29 Figure 14. Comparison Near Bottom of Downcomer......................................................................... 29 Figure 15. Comparison in Lower Horizontal Pipe .............................................................................. 30 Figure 16. Edwards Experiment ........................................................................................................... 32 Figure 17. Short Term Pressure at Closed End of Pipe.................................................................... 33 Figure 18. Long Term Pressure at Closed End of Pipe .................................................................... 33 Figure 19. Void Fraction Near Middle of Pipe ..................................................................................... 34 Figure 20. Horizontal Flow Test Configuration ................................................................................... 35 Figure 21. Horizontal Flow Test Comparison ..................................................................................... 36 Figure 22. Clearing Air from a Vertical Pipe........................................................................................ 36 Figure 23. Wave Speed as a Function of Air Content ....................................................................... 37 Figure 24. Pressure Change Near Upstream End of Short Pipe ..................................................... 38 Figure 25. Pressure Change Near Upstream End of Long Pipe ..................................................... 38 Figure 26. Water Velocity Near Inlet Valve ......................................................................................... 39 Figure 27. Pressure at Exit of Top Condenser Tube ......................................................................... 40 Figure 28. Configuration for Water Hammer Test .............................................................................. 40 Figure 29. Pressure in Upper Horizontal Pipe .................................................................................... 41 Figure 30. Purdue 6 Inch Test Loop..................................................................................................... 45 Figure 31. Purdue 6 Inch Test Loop Isometric ................................................................................... 46 Figure 32. Flow Rate Through M1 ........................................................................................................ 47 Figure 33. Flow Rate Through M2 ........................................................................................................ 48 Figure 34. Differential Pressure DP1 ................................................................................................... 48 Figure 35. Differential Pressure DP2 ................................................................................................... 49 Figure 36. P2 and P3 in 6 Inch Pipe for 10%. 0.80............................................................................ 50 Figure 37. Representation of Data used for Comparison to GOTHIC Prediction ......................... 51 Figure 38. Void Data at Upper Downcomer ........................................................................................ 51 Figure 39. - Top of Vertical Pipe........................................................................................................ 52 Figure 40. - Bottom of Vertical Pipe .................................................................................................. 52 Figure 41. in Lower Horizontal Pipe .................................................................................................. 53 Figure 42. Data Near Bottom of Downcomer...................................................................................... 53 Figure 43. PW3 Data ............................................................................................................................. 54 Figure 44. Bottom Horizontal Pipe Void (PW3) .................................................................................. 55 Figure 45. Purdue 4 Inch Test Facility ................................................................................................. 56 Figure 46. Purdue 4 Inch Facility Instrumentation ............................................................................. 57 Figure 47. Layout for Purdue 12 Inch Facility ..................................................................................... 60 Figure 48. Purdue 12 Inch Facility Instrumentation ........................................................................... 61 Page 6

Figure 49. Near Bottom of Vertical 8 Inch Pipe............................................................................... 63 Figure 50. in Lower Horizontal 8 Inch Pipe ...................................................................................... 64 Figure 51. Near Bottom of 6 Inch Vertical Pipe............................................................................... 65 Figure 52. in Lower Horizontal 6 Inch Pipe ...................................................................................... 65 Figure 53. Near Bottom of Vertical 6 Inch Pipe............................................................................... 66 Figure 54. Void in Lower Horizontal 6 Inch Pipe ................................................................................ 66 Figure 55. Millstone Test Facility .......................................................................................................... 68 Figure 56. Millstone GOTHIC Model .................................................................................................... 69 Figure 57. Reducer Noding ................................................................................................................... 70 Figure 58. Reducer Noding ................................................................................................................... 70 Figure 59. Case 1 Void Fractions at Separator Inlets ....................................................................... 72 Figure 60. GOTHIC Calculation of Upper Elbow - 2 D Model .......................................................... 74 Figure 61. Typical ECCS Suction Piping Noding Diagram ............................................................... 79 Figure 62. GOTHIC Model of Robinson SI and CS Systems ........................................................... 80 Figure 63. ECCS Suction Nodalization ................................................................................................ 82 Figure 64. Void fractions for 13.1 ft3 High Point Volume ................................................................... 83 Figure 65. Void fractions for 2.9 ft3 High Point Volume ..................................................................... 84 Figure 66. Void in 12 Inch Vertical Pipe .............................................................................................. 85 Figure 67. RHR Model ............................................................................................................................ 87 Figure 68. GOTHIC RHR Nodalization ................................................................................................ 88 Figure 69. GOTHIC Model of FAI Test with No Check Valve .......................................................... 91 Figure 70. GOTHIC Load Prediction of Axial Load ............................................................................ 92 Figure 71. Fauske Test Data Corresponding to Figure 70 (Solid Line) .......................................... 92 Figure 72. Typical GOTHIC Noding for ECCS Discharge Piping .................................................... 93 Figure 73. RHR Discharge Piping ........................................................................................................ 95 Figure 74. RHR Pump Start Discharge Pressure Response ........................................................... 96 Figure 75. RHR Pump Start Axial Load ............................................................................................... 96 Page 7

ACRONYMS AND ABBREVIATIONS 2D Two dimensional 3D Three dimensional AAV_ Valve designation in Purdue test facility AIMP Arch impedance meter ANSI American National Standards Institute ANS American Nuclear Society bar Unit of pressure BEP Best efficiency point BWR Boiling water reactor C Centigrade or Celsius CA Multi-stage stiff shaft pump designation CLB Current licensing basis CFR Code of Federal Regulations CS Containment Spray D Pipe diameter DHR Decay Heat Removal DP1 Differential pressure instrument designation in Purdue test facility DP2 Differential pressure instrument designation in Purdue test facility EC Engineering change ECCS Emergency Core Cooling System EPRI Electric Power Research Institute F Fahrenheit f Friction factor ft Feet g Subscript that indicates gas gc Gravitational constant GDC General Design Criterion GL Generic Letter HHSI High head safety injection IN Information Notice JHF Multi-stage flexible shaft pump designation K Flow coefficient kg Kilogram kPa KiloPascal L Subscript indicates liquid lbsm Pounds mass LOCA Loss-of-coolant accident m meter M_ Valve designation in Purdue test facility NFr Froude number NAI Nuclear Applications, Inc.

NEI Nuclear Energy Institute NPSH Net positive suction head NRC Nuclear Regulatory Commission NRR NRC Office of Nuclear Reactor Regulation p pressure P_ Pressure instrument designation in Purdue test facility PIRT Phenomena Identification and Ranking Table Page 8

psi pounds per square inch psia pounds per square inch absolute psig pounds per square inch relative to atmospheric pressure PW_ Parallel wire induction probe PWR Pressurized water reactor PWROG Pressurized Water Reactor Owners Group Q Water volumetric flow rate RCP Reactor coolant pump RCS Reactor Coolant System RIMP Ring impedance meter RIS Regulatory Information Summary RLIJ Multi-stage flexible shaft pump designation RG Regulatory Guide RHR Residual Heat Removal RO Refueling Outage RWST Refueling water storage tank s Spacing gap sec Second SI Safety Injection SIH High head safety injection pump SR Surveillance requirement SRXB NRR Reactor Systems Branch TR Topical Report TS Technical Specification TSTF Technical Specification Task Force URI Unresolved Item USNRC United States NRC V Liquid velocity based on total pipe flow area WCAP Westinghouse report designation WDF Single stage pump designation x Horizontal coordinate y Depth coordinate y Water Depth z Vertical coordinate Void fraction Density Void Fraction Page 9

1. INTRODUCTION NRCs Generic Letter (GL) 2008-01 (USNRC 1, January 11, 2008) addressed the long-standing issue of gas accumulation in systems associated with commercial nuclear power plant operations. Part of addressing gas issues is the prediction of gas movement under transient conditions such as may be instigated by starting a pump when the suction pipe contains gas.

The complexity of such predictions often requires application of two phase, two component ,

geometrically complex computer codes. The Nuclear Regulatory Commission (NRC) has not accepted a generic computer code methodology that addresses this issue. Consequently, licensees must submit individual, plant-specific code analyses to determine pump suction piping void fractions that will not cause a loss of operability.

The Nuclear Energy Institute (NEI) has provided topical report (TR) guidance that any computer code used to develop a system specific model should be verified to be applicable to solve problems involving gas transport in piping systems via comparisons with laboratory test data or other appropriate methods. Further, a suitable safety factor should be added to predicted results to reasonably ensure the predictions encompass actual behavior. (NEI, December 2010) (NEI, October, 2012) NRC subsequently determined that the NEI TR was acceptable and endorsed the TR guidance (NRC, March, 2013).

The Electric Power Research Institute (EPRI) (Rahn, January, 2012) described GOTHIC as follows:

GOTHIC solves the conservation equations for mass, momentum and energy for multi-component, multi-phase flow in lumped parameter and/or multi-dimensional geometries. The Phase balance equations are coupled by mechanistic models for interface mass, energy and momentum transfer that cover the entire flow regime from bubby flow to film/drop flow, as well as single phase flows. The interface models allow for the possibility of thermal non equilibrium between phases and unequal phase velocities, including countercurrent flow. GOTHIIC_S includes full treatment of the momentum transport terms in multi-dimensional models, with optional models for turbulent shear and turbulent mass and energy diffusion. Other phenomena include models for commonly available safety equipment, heat transfer to structures, hydrogen burn and isotope transport.

GOTHIC was developed and is maintained under the Numerical Applications, Inc. (NAI) quality assurance program that (McGoun1, No Date) stated conforms to the requirements of 10CFR50 Appendix B and 10CFR Part 21.

Robinson contracted with NAI to use GOTHIC to predict such behavior as pump suction pipe gas volume acceptance criteria that could be used to determine the void volumes that would not jeopardize pump operability.

Modeling is strongly influenced by the modeler who must have a demonstrated capability to fully understand the code and the necessary nodalization. The NRC staff has interfaced with NAI several times, including a meeting on October 18 - 19, 2010 (Beaulieu, October 6, 2010), and has concluded that NAI is well qualified to apply GOTHIC to two phase, two component fluid flow analysis in piping systems. Further, NAI has an extensive background of comparisons to test data that support its application as discussed in (Rahn, January, 2012) and reviewed Page 10

herein. Consequently, the analyses addressed herein have been performed by a modeler with the necessary demonstrated capability.

In response to analysis methodology issues, the NRC issued Information Notice (IN) 2011-17 (NRC, June 26, 2011) that addressed that computer models may not be sufficiently qualified by benchmarking against data. In response to the IN, Robinson (McGoun1, No Date) provided a justification for continued use of GOTHIC at the H. B. Robinson Nuclear Power Plant.

2. RESOLUTION OF VOID ISSUES 2.1 Regulatory Requirements 2 The regulations in Appendix A to 10 CFR Part 50 or similar plant-specific principal design criteria 3 and in 10 CFR 50.46 4 provide design requirements. Appendix A requirements applicable to gas management include the following:

• General Design Criterion (GDC) 1 requires that systems be designed, fabricated, erected, and tested to quality standards.

• GDC 34 requires a residual heat removal (RHR) system 5 designed to maintain specified acceptable fuel design limits and to meet design conditions that are not exceeded if a single failure occurs simultaneous with failure of specified electrical power systems.

• GDCs 35, 36, and 37 require an emergency core cooling system (ECCS) design that meets performance, inspection, and testing requirements.

• GDCs 38, 39, and 40 require a containment heat removal system design that meets performance, inspection, and testing requirements.

The regulations in 10 CFR 50.46 provide specified ECCS performance criteria.

Quality assurance criteria provided in Appendix B that apply to gas management include the following:

• Criterion III requires measures to ensure that applicable regulatory requirements and the design basis, as defined in 10 CFR 50.2, Definitions, and as specified in the license application, are correctly translated into controlled specifications, drawings, procedures, and instructions.

• Criterion V requires important activities to be prescribed by documented instructions, procedures, or drawings, which must include appropriate quantitative or qualitative 2

Much of this section is copied from (NRC, March, 2013).

3 These apply to facilities with a construction permit issued before May 21, 1972 that are not licensed under Appendix A.

4 10 CFR 50.46(d) requires licensees to meet Criterion 35 of Appendix A.

5 Various licensees use decay heat removal (DHR), RHR, and shutdown cooling when referring to systems that are used to cool the reactor coolant system (RCS) during shutdown operation. These terms have the same meaning.

Page 11

acceptance criteria for determining that important activities have been satisfactorily accomplished.

• Criterion XI requires a test program to ensure that the subject systems will perform satisfactorily in service. Test results shall be documented and evaluated to ensure that test requirements have been satisfied.

• Criterion XVI requires measures to ensure that conditions adverse to quality, such as failures, malfunctions, deficiencies, deviations, defective material and equipment, and non-conformances, are promptly identified and corrected, and that significant conditions adverse to quality are documented and reported to management.

• Criterion XVII requires maintenance of records of activities affecting quality.

Furthermore, as part of the licensing basis, licensees have committed to quality assurance provisions that are identified in both their technical specifications (TSs) and quality assurance programs. Licensees have committed to use the guidance of Regulatory Guide (RG) 1.33 (NRC, February 1978) which endorses American National Standards Institute (ANSI)

N18.7-1976/American Nuclear Society 3.2 (ANS, February 19, 1976) or equivalent licensee-specific guidance. Section 5.3.4.4, Process Monitoring Procedures, of ANSI N18.7 states that procedures for monitoring performance of plant systems shall be required to ensure that engineered safety features and emergency equipment are in a state of readiness to maintain the plant in a safe condition if needed. The limits (maximum and minimum) for significant process parameters shall be identified. Operating procedures shall address the nature and frequency of this monitoring, as appropriate.

In 10 CFR 50.36(c)(3), the NRC defines technical specification (TS) surveillance requirements (SRs) as relating to test, calibration, or inspection to assure the necessary quality of systems and components is maintained, that facility operation will be within safety limits, and that the limiting conditions for operation will be met. This requires that licensees establish, implement, and maintain written procedures covering the applicable procedures recommended in Appendix A to Regulatory Guide (RG) 1.33. Appendix A to RG 1.33 identifies instructions for filling and venting the ECCS and DHR system, as well as for draining and refilling heat exchangers. Standard TSs and most licensee TSs provide SRs to verify that at least some of the piping in systems that are important to safety is filled with water 6. In response to the continuing issues with gas management, the industry owners groups have initiated changes to the TSs through the Technical Specification Task Force (TSTF). The TSTF has proposed and is developing TSTF-523, "Generic Letter 2008-01, Managing Gas Accumulation" that is applicable to all plant types (Stringfellow, March 29, 2012). The NRC found this proposal acceptable for review and requested additional information regarding TSTF-523 (Honcharik, 6

If the licensee can conclude through an operability determination that there is a reasonable expectation that the system in question can perform its specified safety function, the system piping can be considered filled with water such that the SR is met (TIA2008-03, October 21, 2008.). A condition where there is no void is described by such words as gas-free, free-of-gas, or water-solid. The current licensing basis (CLB) for most nuclear power plants requires a water-solid condition. This means, for example, that a water-solid condition should be reasonably ensured upon returning to power from an outage. Experience has shown that gas may accumulate in piping during operation and the filled with water criterion applies.

Page 12

June 13, 2012). The TSTF provided a response on August 30, 2012 (Browning, August 30, 2012).

(NRC, September 26, 2005), (NRC, April 16, 2008), and (NRC, December 7, 2009 (Approved))

state that the objective of determining system operability is to reasonably ensure that subject system operability is achieved and a reasonable expectation test applies. This means that a high degree of confidence applies but absolute assurance is not necessary. The determination can be based on analyses, test or partial test, experience, and/or engineering judgment. This is particularly applicable to void transport, pump response to voids, and vortexing, where sufficient information is not available for providing in-depth generic guidance that addresses many issues.

Consequently, a strong reliance on engineering judgment will be necessary to support an interim finding regarding operability for these issues until improved generic guidance can be developed. If an operability concern arises during operation, more reliance can be based on judgment for the short term until a more in-depth investigation is applied to assess long-term operation.

2.2 Void Concerns Voids in the piping of systems that are important to safety have several implications:

• The principal concern with a void upstream of a pump is the potential impact on pump operability. Other upstream concerns include water hammer.

• The principal concern with a void downstream of a pump is potential water hammer.

Other downstream concerns are delay of injection of water and the effect of a void injected into the RCS.

Two types of voids are generally of potential concern; gas and vapor. Gas, typically air or nitrogen, may be introduced during outages or may be generated by outgassing of fluid used to fill a system or equipment malfunctions such as leaking valves that result in a pressure reduction accompanied by outgassing. Vapor can occur when pressure is reduced below the fluid saturation temperature. In some cases, such as during flow through orifices, a non-equilibrium condition resulted in which gas was not immediately re-absorbed when pressure was restored. A combined effect is also encountered with gas that contains vapor. All of these concerns must be addressed when assessing operability.

2.2.1 Pump Response to Voids The typical starting point for determining a void volume in pump suction piping that will not jeopardize operability is the allowable void at pump suctions. Robinson used values from Tables 1 and 2 of NEI 09-10 Rev 1 (NEI, December 2010), that are consistent with the more recent Rev 1a (NEI, October, 2012) and an NRC safety evaluation (NRC, March, 19, 2013):

Page 13

Table 1 - Allowable Average Non-Condensable Gas Void Fractions, 7 (to preclude pump mechanical damage) 8 9 BWR PWR Typical Pumps Typical Single Multi-Stage Multi-Stage

% Q/Q(BEP) Pumps Stage Stiff Shaft Flexible Shaft (WDF) (CA) (RLIJ, JHF)

Steady State Operation

> 20 seconds 40%-120% 2% 2% 2% 2%

Steady State Operation

> 20 seconds <40% or >120% 1% 1% 1% 1%

Transient Operation 70%-120% 10% 5% 20% 10%

For 5 For 20 For 20 For 5 sec sec sec sec Transient Operation <70% or >120% 5% 5% 5% 5%

For 5 For 20 For 20 For 5 sec sec sec sec where:

Q = water volumetric flow rate BEP = best efficiency point Transient is averaged over the specified time span Table 2 - Allowable Average (to preclude significant reduction in discharge head)

PWR Typical Pumps BWR Single Multi-Stage Multi-Stage

% Q/Q(BEP) Typical Stage Stiff Shaft Flexible Shaft Pumps (WDF) (CA) (RLIJ, JHF)

Steady State Operation 40%-120% 2% 2% 2% 2%

Steady State Operation <40% or >120% 1% 1% 1% 1%

The transient operation criteria are based on the premise that bubbly flow exists at the pump entrance, that the initial void fraction in the pump does not exceed 0.05, that full head will be recovered after the gas has passed through the pump as substantiated by pump operation experience, and the judgment that the short times associated with the transients will not result in pump damage. Further, the NRC stated that it is the responsibility of licensees to demonstrate that a dispersed bubbly flow exists at the pump entrance throughout transients, a requirement also identified by industry for application of the criteria (NEI, October, 2012). 10 Stratified flow 7

is used in place of later in this report. They have the same meaning.

8 BWR boiling water reactor 9

PWR pressurized water reactor 10 NRC found NEI 09-10 Rev 1a to be acceptable subject to conditions identified in the NRC Safety Evaluation (SE) and endorsed NEI 09-10a as an acceptable voluntary approach to effectively prevent and manage gas intrusion and accumulation in plant systems (NRC, March, 19, 2013). NRC plans to issue a Regulatory Information Summary (RIS) that endorses NEI 09-10.

Page 14

with a small void fraction over the time of void passage through a pump can also, with justification, be used to meet the Table 1 criteria.

The difficulty of determining if bubbly flow exists will vary depending upon flow rate and the geometry. For example:

• If the Froude number (NFr) > 2.5 11, there is a potential that a void will be transported as a slug.

• If a vertical downcomer 12 is connected directly to the top of a pump and the downcomer volume is at least four times as large as the original gas volume that existed above the downcomer, then bubbly flow will exist at the pump entrance provided NFr is small enough to preclude slug flow as identified in the above bullet.

• If a horizontal pipe connects between the bottom of a downcomer and a pump entrance, a methodology should be applied that has a multi-dimensional two phase capability that has been verified by comparison to experimental data. Since phenomena in this region are not well understood, judgment may be a significant factor and a suitable safety factor must be included to reasonably ensure the prediction encompasses actual behavior.

• Horizontal pipes may introduce other concerns. For example, flow stratification in horizontal pipes can lead to an accumulation of gas, such as in an off-take or tee geometry. Once gas is accumulated, a subsequent instability can lead to a large surge in gas downstream.

The pump roadmap project (Huffman, August, 2012) stated The expert panel concluded that the use of the exiting criteria (Tables 1 and 2) is acceptable and is appropriately conservative to ensure pump operation with voids. More relaxed criteria for several pump types were suggested. Although this report was not reviewed, it has been read and it addressed behavior of voids in pumps, provided experimental data, examined pump failure modes, compared expected void capability with the Tables 1 and 2 criteria, and provided Revised Existing Criteria to preclude pump mechanical damage.

Flowserve (Kasztejna, August 5, 2008) stated that the ECCS multi-stage pumps could be run at 5% gas without distress although some performance degradation would occur and they could run at 5% to 10% for no more than an hour. Above 10% operation was stated to possibly cause loss of prime and could air bound the pump. A slug of air was stated to cause the first stage to lose prime and cause major problems (seizure) in a very short time period. No data were provided. Flowserves opinion is generally consistent with conservatism in the above criteria with one exception. It is not consistent with industry criteria for flow rates outside of stated bands relative to the best efficiency point. As illustrated in Table 1, the 2% continuous void only applies for Q/Q(BEP) from 40% to 120% and the 10% for 5 seconds is limited to Q/Q(BEP) between 70% and 120%. Note that degradation could result in stalling and complete loss of flow if the pump was exposed to a low flow rate such as may occur with some small breaks where an elevated downstream pressure exists. However, in this situation, the low flow rate may not transport gas into the pump.

11 NFr is discussed in Section 2.3.1, below.

12 Downcomer and vertical pipe are used interchangeably in this report.

Page 15

(Harvill, No date) provided qualifications for comparing GOTHIC results to the above criteria:

Transient acceptance criteria must be carefully applied since the methodology is (typically) biased for conservatism with respect to peak pump suction void fractions, but not transient duration. The primary variable in the size and shape of the downstream void fraction is flow rate; and the calculation methodology produces conservative peak void fractions by maximizing the flow rate. Lower flow rates will produce lower peak values, but can also extend the duration of void profile. For this reason, use of the transient criteria is inconsistent with use of the maximum flow rate when demonstrating that a void fraction may exceed the steady-state limit for a finite length of time. While a reduction in the flow rate may reduce the peak below the transient void limit, it may also cause the duration of the peak to extend beyond the duration limit. Application of the transient criteria must be evaluated on a case-by-case basis. This evaluation should consider the sensitivity of changes in pump flow to variations in system pressure, as well as the integrated void fraction above the steady-state limit.

2.2.2 Water Hammer Water hammers have occurred that caused relief valves to open and remain open and have caused support structures to be damaged but, to our knowledge have not caused pipe ruptures.

However, as identified in (2008-01, January 11, 2008.), conditions have occurred in which pipe rupture was possible. Water hammer is addressed in more detail in Section 3.3.1.3, below.

2.2.3 Delay The Pressurized Water Reactor Owners Group (PWROG) report (LTR-LIS-08-543, April 2, 2009) addressed the effect of gas causing a delay in the injection of water into the reactor vessel. This report was reviewed in (NRC, March, 19, 2013) and the NRC staff concluded that it establishes that an initial gas void of 5 ft3 in high pressure system piping at 400 psia and 68 °F or low pressure system piping at 100 psia and 68 °F is not of concern with respect to most aspects of injection into a PWR RCS. It does not address any other aspects of gas voids such as gas in pump discharge or suction side piping. Further, it is assumed in the report that there is no delay or reduction in ECCS flow rate beyond the point assumed in the safety analyses of record. Licensees referencing the information provided in this report must consequently establish that these assumptions are correct. The NRC also concluded that the report incorrectly stated that pumped ECCS flow during recirculation is not affected by potential voids in the ECCS piping, (and) the calculations that confirm the ability to flush the core and remove decay heat are not impacted. Consequently, Where applicable, these aspects of long term cooling must be addressed on a plant-specific basis. Further, the NRC stated that the potential for gas causing problems with RCP seals is not addressed if gas in a non-active charging path is transported to RCP seals when the path becomes active such as due to swapping charging pumps.

2.2.4 Vortex Concerns Vortexing is not well understood. No theoretical treatments have been found applicable to nuclear power plant concerns, in part because of the sporadic nature of vortexing in which even observation time influences experimental data. As a result, all vortex behavior must be based on acceptable experimental data unless an acceptable alternative is provided or conditions Page 16

clearly exist where vortexing is not of concern. For example, for sumps and tanks, vortexing is normally not of concern if an exit pipe is covered by 9 feet of water or NFr is less than 0.25 and water depth is at least one exit pipe diameter. For PWR RHR to hot leg connections, the NRC does not consider vortexing to be of concern if the correlation provided by Jennifer Gall at the 2013 Regulatory Information Conference 13 is satisfied. Consequently, vortexing must be addressed on a plant-specific basis that includes appropriate supporting detail until the NRC issues or endorses an acceptable generic method.

2.2.5 Net Positive Suction Head (NPSH)

(Robinson2, No date) states that It is acceptable to conclude that short duration transients with a 3% void fraction are not of concern with NPSH.

2.3 Prediction of Void Movement Processes that may be followed to determine void volumes in suction piping that meet the table criteria and that do not jeopardize operability include the following:

1 Froude Number (NFr) 2 Simplified methodologies 3 In-depth thermal-hydraulic code applications 2.3.1 Froude Number The Froude number is the ratio of the inertia or drag force to buoyancy force. For horizontal flow, fluid inertia versus the buoyancy determines the flow regime. When buoyancy dominates (small NFr) the flow tends toward stratification. At higher NFr, the flow tends toward bubbly or slug flow. For vertical flow, when the Froude number is small, the flow tends toward the film regime dropping into a pool. At larger NFr, the drag over comes the buoyancy and the flow tends to the slug or bubbly regime. (George, January 10, 2012)

Froude number is defined by:

V N FR =

Dg c ( L g )

L 13 Status of NEI 09-10 Rev 1 and Various NRC and Industry Gas Management Activities, presented in Technical Session T8 - Status and Path Forward on the Management of Gas Accumulation in Nuclear Power Systems. (NRC, March 12 - 14, 2013.)

Page 17

where:

D = pipe diameter V = liquid velocity based on total pipe flow area gc = gravitational constant

= density subscript L indicates liquid subscript g indicates gas NFr is useful to identify low flow conditions that will not cause gas to move, high flow conditions where gas will be carried by the flowing liquid, and to assess aspects between these two boundaries. Use of NFr that is consistent with NRC acceptance criteria is summarized in the following table (NRC, March, 2013):

NFR Effect 0.31 No gas movement in horizontal pipe if 0.20 14.

0.31 < NFR 0.65 Some gas may be transported depending on pipe geometry

> 0.54 Gas will move toward the downstream end of a horizontal pipe that has no local high points. Some bubbles may move downward in a vertical pipe.

< 0.8 Dynamic venting not effective.

0.8 < NFR < 2.0 Time to clear gas is a function of flow rate and piping geometry.

Timing is not well characterized.

1 Gas will be removed from an inverted "U" tube heat exchanger for steady state flow lasting several minutes. Criterion not applicable at bottom of vertical pipe that connects to a horizontal pipe.

> 1.2 Horizontal pipe that is open at the downstream end will run full.

2.0 All gas will be removed from pipe but localized gas pockets may remain where full flow conditions may not exist such as in the vicinity of valves or orifices.

2.3.2 Simplified Methodologies NRC has only found two generic methodologies to be acceptable for predicting pump suction void fractions that would result from voids in pump suction piping (NRC, March, 19, 2013). One that compares the volumetric flow rate for 0.5 seconds to the table acceptable void fractions is generally too conservative to be useful and the other, the Simplified Equation, has restrictions and uses proprietary information that also restricts its use. The conservative method is described in Section 3.15.2.5 of (NRC, March, 2013) and the simplified equation is discussed in Section 3.15.2.3 of (NRC, March, 2013).

2.3.3 In-depth Thermal-Hydraulic Code Applications At present, there is no acceptable generic methodology for assessing pipe void size and void transport behavior other than use of NFr and the simplified methodologies identified above.

Assessment of conditions not covered in these must usually be addressed via a computer code that is found to be acceptable on a plant-specific basis. As quoted in (NRC, March, 2013), any 14 The 0.20 criterion reasonably ensures there is sufficient flow area for liquid transport.

Page 18

computer code used to develop a system specific model should be verified to be applicable to solve problems involving gas transport in piping systems via comparisons with laboratory test data or other appropriate methods. Further, a suitable safety factor should be added to predicted results to reasonably ensure the predictions encompass actual behavior.

2.3.4 Conservatisms Associated with Pump Void Fraction Predictions The first conservatism is the use of the acceptable void fraction tables that are generally judged to be conservative. The amount of conservatism is a function of the pump and has not been well qualified in topical documentation submitted to the NRC for review.

It is desirable to use a best-estimate approach when predicting gas/liquid behavior since a conservative approach may distort results by an unknown amount. A best-estimate prediction that is without bias allows a known conservatism to applied to the predictions. Further, introducing a conservatism with respect to one parameter may be non-conservative with respect to another parameter or at another location. Unfortunately, limitations in modeling methodologies may necessitate the use of conservative assumptions. For example, aspects of flow behavior may not have been covered in tests used to benchmark the methodology yet must be addressed when the methodology is used to predict system behavior, the methodology may have known weaknesses, or the test data may be inconsistent.

2.4 Test Issues The principle concerns involve transient movement of gas in pipes containing water and the behavior of vortexing such as may occur when fluid is removed from tanks or sumps.

Theoretical capability is not available that allows calculation of the behavior. Consequently, it has been necessary to obtain test data to provide a basis to predict behavior. Tests have shown that transient behavior may be inconsistent from test to test and different test results are often obtained in otherwise apparently identical conditions. Further, although test configurations have been assembled that simulate portions of plant systems, it has been impractical to simulate entire systems. It has also often been necessary to use scaled test models.

Therefore, it is generally necessary to apply methodologies that have been established to reasonably predict or bound test behavior and, in some cases which have extrapolation or interpolation capabilities, to predict system response.

Aside from the sporadic behavior, tests have limitations that affect the data and which must be considered when applying test data to develop or assess gas/liquid behavior methodologies.

These limitations include the following:

• The test configuration may include non-prototypical components, such as a gas/water separator to remove gas from return flow before it enters a non-prototypical pump.

These may influence test results by failing to remove all gas or artificially dampening pressure oscillations. Failing to remove all gas can bias the results if gas accumulation is measured as a means to determine void fraction or gas transport rate.

• Determination of void fractions is difficult. For example, parallel wired induction meters may be most accurate when measuring stratified flow whereas a ring impedance meter may be considered for either stratified or bubbly flow. Yet there may be substantial differences between indicated void fractions between the meters in installations where the void fraction should be the same. Another problem occurs when a void does not Page 19

move at the same velocity as the liquid yet the instrumentation does not allow determination of individual component velocities.

• Determination of mass flow rate may be difficult under some conditions.

3. USE OF GOTHIC TO PREDICT GAS / WATER BEHAVIOR IN PIPING SYSTEMS GOTHIC is a computer program that solves the conservation equations for mass, momentum, and energy for multi-component, multi-phase flow. The principal modeling element is a control volume that is occupied by a fluid which may include non-condensing gases, steam, drops, or liquid water. A volume may be a lumped (single node) volume or is may be composed of one, two, or three dimensions to allow calculation of fluid property distributions and flow patterns within the volume. Volumes are defined using orthogonal coordinates (x, y, z) and volume sub-divisions are based on orthogonal coordinates. Phase balance equations are coupled by mechanistic models for interface mass, energy, and momentum transfer to cover the flow regime from bubbly flow to film/drop flow as well as single phase flows. Interface models allow for non-equilibrium between phases and unequal phase velocities. The minimum liquid temperature is 32 °F.

3.1 GOTHIC Modeling There are many aspects that must be correctly addressed when using GOTHIC to predict hydraulic aspects of fluid behavior in piping systems. For purposes of this audit, the noding will be reviewed. A typical noding diagram, in this case for the Purdue 6 inch tests that cover aspects of plant ECCS suction piping, is provided in Figure 1 (Wiles, June 26, 2009). The basic Figure 1. GOTHIC Volume Noding Diagram Page 20

Corresponding volume nodalizations for this type of analysis are shown in Figures 2, 3, and 4. 15 Figure 2. Upper Horizontal Pipe Noding 15 Figure 5 for the upper horizontal pipe is from (Wiles, June 26, 2009). In response to RAI Number 2 (Lyon, April 9, 2013), Figures 3 and 4 were stated to be in error and the correct figures are provided (Balakhnin, April 9, 2013).

Page 21

Figure 3. Vertical Pipe Noding Figure 4. Lower Horizontal Pipe Noding Page 22

There are two methodologies for modeling elbows and tees that occur parallel or perpendicular to gravitational forces. Each has strengths and weaknesses.

Figure 5. Lumped Volume Connection Figure 6. Subdivided to Subdivided Connection Page 23

Another method is necessary for elbows and tees that are not parallel or perpendicular to gravity, such as illustrated in Figure 7. A series of these would be used, for example, in a run of Figure 7. Two elbows The nodalization for this configuration is shown in Figure 8.

Figure 8. Nodalization for Two Elbows Another two elbow configuration is shown in Figure 9 with the nodalization shown in Figure 10.

Figure 9. Two Elbow Configuration Page 24

Figure 10. Two Elbow Nodalization As another possibility, the angled tee illustrated in Figure 11 would be nodalized as shown in Figure 12.

Figure 11. Tee Configuration Page 25

Figure 12. Tee Nodalization This representation will tend to move all air into the upward-oriented tee.

system. Further, special cases may require specialized treatment. This is seen, for example, in the nodalization of tapered reducers that is discussed in Section 3.3.3, below.

Additional modeling information is provided in Section 3.1.3.7 of (Harvill, No date). Other considerations are covered in Sections 3.2, 3.3, and 5.0 of that reference.

3.2 GOTHIC Configuration Control (Wiles, June 26, 2009) stated that multiple GOTHIC versions and multiple methodologies were used in the course of arriving at a successful benchmark of test data and in performing analyses Page 26

of industry suction piping systems. Some of the GOTHIC versions used for prediction of behavior were as follows:

Version Discussion 7.2aWC Used for Purdue 6 inch simulations. Good agreement with data claimed. The Purdue 6 inch tests are examined in Section 3.3.2.1, below.

7.2aWC2 Previous gas migration analyses of ECCS piping 7.2a(QA) Used in initial Purdue analyses. Shortcomings, such as under-prediction of time dependent void fraction at the pump inlet due to inadequate control of the maximum bubble size for larger pipe diameters such as used in the Purdue tests, led to patch release version 7.2aWC2(QA). This is an example of the need for applicable test data when performing plant system analyses since the issue did not arise with smaller diameter tests.

7.2aWC2(QA) NAI used for considerable analysis, including initial 6 inch and 8 inch Purdue benchmark analyses reported in Section 3.3.2, below. This is a patch release based on 7.2a(QA) to allow user control over the maximum diameter for large bubbles, to correct deficiencies affecting pressure loss in connections between subdivided volumes, and to allow an elbow to be modeled with a lumped parameter control volume.

7.2b(QA) Used for more recent analyses of 6 and 8 inch Purdue tests, 4 and 12 inch Purdue tests, some benchmarking. Used to demonstrate that methodology of lumped corners or subdivided corners for elbows provides similar results. NAI stated that future analyses could begin with 7.2b(QA). Has many new features and corrections, includes all modifications made to version 7.2a(QA) to create version 7.2aWC2(QA).

Has capability to represent elbows by lumped corners or subdivided corners.

8.0 Used for Millstone tests. NAI description of differences is that The important models for interphase drag in version 8.0 are essentially the same as those in version 7.2b.

Results from versions 7.2b and 8.0 for gas transport problems are very similar, so this validation also applies to version 7.2b. Version 8.0 includes some changes that make the code more robust for multiphase flow problems. No quantitative information was provided.

Page 27

Some of the Purdue prediction comparisons were as follows:

Description 7.2aWC2(QA) 7.2b(QA) 6-inch, 10% void, 0.80 NFr

- Void fraction, , near Good initial behavior prediction Similar with oscillations after downcomer top but a little early. Second large second prediction peak.

peak predicted not shown in data.

- near downcomer bottom Late prediction of transient and Excellent initial prediction. Peak peak . Peak factor of two later behavior identical to higher than data. 7.2aWC2(QA).

- in lower horizontal pipe Good prediction with slight over Similar with second extended prediction of peak. peak not shown in 7.2aWC2(QA) or data.

- Pressures in horizontal Good to excellent with post- Similar with larger oscillations.

pipes reduction oscillations not seen in data.

6-inch, 20% void, 1.65 NFr

- near downcomer top Excellent. Complete void at Good. Complete void at peak.

peak. Has second peak not shown in data.

- near downcomer bottom Excellent timing. Slightly over- Good timing. Over-predicts data predicts data by about 50%.

- in lower horizontal pipe Excellent but slightly under- Excellent.

predicts data.

- Pressures in horizontal Good to excellent. Good to excellent.

pipes 8-inch, 10% void, 0.80 NFr

- near downcomer top Excellent initial timing. data Same except predicts third peak show single peak with complete at 20 seconds while data void at peak, calculation shows trending to zero.

double peak, under-predicts data.

- near downcomer bottom Poor. Peak under-predicted by Same.

factor of five. This is addressed in Section 3.3.2.2, below.

- in lower horizontal pipe Good except predicts short Good. Prediction factor of two duration peak not shown in data. high.

- Pressures in horizontal Good to excellent. Small Good to excellent. Larger pipes oscillations predicted. oscillations predicted.

8-inch, 20% void, 1.65 NFr

- near downcomer top Excellent. Slightly under-predicts Excellent but under-prediction complete void. more substantial.

- near downcomer bottom Good. Void under-predicted. Excellent. Void under-prediction not as large.

- in lower horizontal pipe Excellent initial response. Void Same except peak under-duration badly under-predicted. prediction a factor of two.

Peak under-predicted by factor of three.

- Pressures in horizontal Excellent. Excellent.

pipes Page 28

The tables were used to conserve space as opposed to the alternate of providing the figures.

Figures 13 - 15 (Harville, January 26, 2009) provide typical comparisons, in this case for a Purdue 8 inch test with the initial = 20% in an upper horizontal pipe and NFr = 1.65. The lines are GOTHIC predictions and the square symbols represent experimental data. Transient flow is from the upper horizontal pipe into the downcomer into the lower horizontal pipe. Figure 13 provides predicted void fraction, , near the top of a downcomer or vertical pipe that is connected to the upper horizontal pipe. Figure 14 shows near the bottom of the downcomer and Figure 15 shows in a horizontal pipe connected to the bottom of the downcomer.

Figure 13. Comparison of 7.2aWC(QA) and 7.2aWC2(QA) Near Top of Downcomer Figure 14. Comparison Near Bottom of Downcomer Figures 13 and 14 would have been described as good to excellent agreement between GOTHIC versions. Examining in more detail, the Figure 13 peak is about 0.85 versus 0.90 and there are obvious differences in the detail of versus time. Figure 14 shows a peak of about 0.20 versus 0.23 with excellent initial timing and some difference in the time it takes for the void to pass. Figure 15 has good initial timing but poor agreement after that. With respect to agreement with the data, GOTHIC under-predicts the peaks in all locations, with the Figure 15 under-prediction about a factor of two or more. The data in Figure 13 show a broader peak than predicted but note the peak data scatter ranging from 0.35 to 1.0. Some of the disagreement Page 29

Figure 15. Comparison in Lower Horizontal Pipe can be attributed to GOTHIC predicting an average over a length of pipe while the data are at one location and illustrate variation due to individual bubbles or void accumulation Quantitative differences are observed between 7.2aWC2(QA) and 7.2b(QA). In general, overall behavior is predicted by both versions. The conclusion is that the above GOTHIC version changes result in semi-quantitative agreement between versions that is of the same variability as observed between identical tests at Purdue.

Results for Millstone calculations were obtained with Version 8.0. In response to RAI 4 ( (Lyon, April 9, 2013), the licensee stated that Version 8.0 predicted 16.8 in3 of gas went to the charging pump versus 24.8 in3 with Version 7.2b for Test Case 1, both less than the 8.6 in3 measured.

(Balakhnin, April 9, 2013). Predictions for the RHR pump were 620 in3 and 670 in3, respectively. Other predictions for Case 1 are summarized in the following table:

Item Version 7.2b(QA) Version 8.0(beta 1)

Pump flows Initiate at 10.4 sec, linear Not provided. Assumed the increase to 11 sec, decrease same based on built-in in increase rate to 11.7 sec, homologous curves and reach max at 13 sec, begin adjusted flow resistance to linear decrease at 23.2 sec, match steady state test flow.

reach 0 at 24.6 sec. Blips at 25 sec and 29 sec. Max Max RHR flow from plot = 322 RHR flow = 315 gpm gpm.

Air volume to charging pump Initiates at 13 sec, increases Initiates at 13 sec, increases linearly to ~ 18 sec, curves to linearly to ~ 17 sec, curves to 24.8 in3 by 23 sec. 16.8 in3 by 18 sec.

Air volume to safety injection 0 0 pump to charging pump Initiates at 12.9 sec, increases Similar to Version 7.2b except linearly to ~ 14 sec, smooth decreases to 0 at 19 sec and curve to peak of 0.06 at 15 area under curve judged sec, decreases to 0 at 24 sec smaller than Version 7.2b.

Page 30

Item Version 7.2b(QA) Version 8.0(beta 1)

Gas volume to RHR pump Initiates at 13.0 sec, rises Not provided.

steeply to ~ 14 sec, slope decreases slightly until ~18 sec, slope decreases slightly again until reaches 670 in3 at 23 sec.

to RHR pump Initiates at 12.7 sec, rises Initiates at 12.7 sec, rises steeply to 0.36 at ~13.5 sec, steeply to 0.47 at ~ 13.5 sec, drops to 0.07 at 14 sec, drops to 0.07 at 14 sec, decreases with bumps to decreases with bumps to 0 0.02 at 19.4 sec, second peak at 23.5 sec, second peak initiates at 21.4 sec, reaches initiates at 26.5 sec that 0.05 at 22.8 sec, drops to ~ 0 reaches 0.2 at 27.5 sec and at 23.5 sec. decreases to 0.03 at 30 sec.

Each new GOTHIC version is assessed by running a broad range of test predictions to determine if any changes cause degraded prediction results. Many of these predictions are directly applicable to using GOTHIC to predict nuclear power plant system behaviors that are of potential concern here. This subject is addressed in Section 3.3.1, below.

In general, overall behavior is predicted by the different versions but there are significant quantitative differences. The conclusion is that the need for continuing changes at the time of this NAI work (2008 - 2010) reflects the immaturity of GOTHIC for study of the problems of concern here. Differences between versions and with respect to test data must be considered when using GOTHIC predictions.

3.3 Comparison of GOTHIC to Test Data Each GOTHIC version is assessed by comparing GOTHIC predictions to more than 50 tests.

Some of these comparisons are illustrated in Section 3.3.1. Further, for the Robinson application, comparisons were made with Purdue and Millstone tests that were designed to assess specific aspects of pipe suction issues under transient conditions. These are examined in Sections 3.3.2 and 3.3.3.

3.3.1 GOTHIC Version Qualification: Comparisons to Test Data (Rahn, January, 2012) stated that the approach to qualification was based on the premise that GOTHIC is intended to be used as a best-estimate containment analysis package. It has been expanded to analyze other configurations and the library of available experimental data and analytical models used for qualification has expanded accordingly. More than 50 comparisons with various test are provided. Rahn continued by stating that code modifications are evaluated by running all relevant experimental simulations and establishing that, overall, the agreement between predictions and data had improved. In general, GOTHIC predictions range from excellent agreement where the predictions essentially overlay the data or analytic calculation results to about a factor of two difference depending upon what is being compared. Many of the comparisons may not be directly applicable to the gas/water flow issues being addressed here, such as hydrogen burn and ice modeling comparisons, while others involve modeling aspects that are more fully applicable to the models that are important to predicting behaviors addressed Page 31

here, such as water hammer or, under some conditions, degassing in pipes that include angled pipes and prediction of bubble behavior. Several qualifications involve two phase mixtures of water, steam, and air that involve conditions such as bubbly flow, film/drop, and stratified flows, including transitions between flow regimes. Test comparisons include frictional pressure loss, filling of horizontal and vertical pipes, water holdup and pressure drop in vertical pipes with upward and downward flow, water hammer, drop behavior including entrainment, many containment configuration and test conditions, and the Edwards pipe blowdown experiment, Three of these, the Edwards pipe blowdown, two phase behavior, and water hammer, will be discussed in detail as part of the audit of code capability; the first because of the short-term challenges, the second to cover some behavior that lends insight into Robinson use of GOTHIC to address pipe void movement, and the last because GOTHIC was used at Robinson to evaluate water hammer. 16 3.3.1.1 Edwards Blowdown Experiment The Edwards blowdown experiment consisted of a 13.44 ft long pipe with an internal diameter of 2.88 in. It is shown in Figure 16. An experiment consisted of filling the pipe with water, heating Figure 16. Edwards Experiment it while maintaining pressure about 500 psi above saturation pressure, and then rupturing the glass disc to allow a rapid depressurization. One of the comparisons of GOTHIC predictions to experimental data is shown in Figures 17 - 19. In general, GOTHIC provides an excellent fit to the rapid transient. The greatest deviation appears to be illustrated in Figure 19 where the initial small jump in void fraction data was missed, then generally over-predicted for the first 0.25 seconds, and followed by an accurate fit to the data as void fraction approached one.

16 All information is taken from (Rahn, January, 2012).

Page 32

Figure 17. Short Term Pressure at Closed End of Pipe Figure 18. Long Term Pressure at Closed End of Pipe Page 33

Figure 19. Void Fraction Near Middle of Pipe 3.3.1.2 Two Phase Flow Comparisons with test data were provided. A summary of selected comparisons is as follows:

Test Results Pipe configurations were 0.955 and 1.268 inch dp/dx versus quality and versus pressure were horizontal pipe and 0.5 x 1.75 rectangular predicted within +/- 40%. Much of the pipe down, horizontal, and up. Pressures uncertainty was attributed to observed test were 600, 1000, and 1400 psia. Flow rates results in combination with bias in the were from 0.25 X 106 to 2.0 X 106 lbsm/hr-ft2. predictions.

87 test parameter sets were evaluated as a function of quality ranging from 0 to 1.

Data from a geothermal well with an upper Predicted pressure was about 10% low where well casing length of 1292 m and a lower flow transitioned from bubbly to film regimes section length of 1308 m with inside diameters but fit well otherwise. Predicted outlet of 8.7 inches and 6.28 inches, respectively. temperature was at saturation but low by 10 Measured pressure and temperature at the top °C when compared to data. NAI expected of the well was 42 bar and 272 °C and at the saturation since there is water in the steam bottom was 16390 kPa and 315 °C. Flow rate and GOTHIC correctly predicted the pressure.

was 132 kg/sec. Vapor volume fraction This difference persisted from a depth of 0 to ranged from 0 at the bottom of the well to 0.9 2000 m. Temperature at deeper depths was at the top. accurately predicted.

Page 34

Test Results Wallis conducted experiments with the Figure 21 provides two comparisons where configuration and observed flow patterns GOTHIC provides an excellent fit of water illustrated in Figure 20. NFr was based on depth at the end of the 3/4 inch pipe (y) to pipe water velocity assuming the pipe was running diameter (D) for a spacing gap, s, of 19 mm full, consistent with the definition used in our and a good fit for a 2 inch pipe for s = 76 mm assessment report. at lower NFr while under-predicting y/D by about 15% at larger NFr.

Note that (WCAP-17271-P, October, 2010.)

indicated that a pipe discharging freely would run about 50% full at the outlet for a NFr = 0.6 and would run full at 1.16. For Figure 21s s =

76 mm, GOTHIC predicted y/D of about 0.5 at 0.6 but predicted only 0.8 at NFr = 1.16.

Several tests evaluated clearance of air with Both the tests and GOTHIC clearly bracketed downward flow in a vertical pipe. Critical flow critical flow rate with a slightly less flow rate rate was defined as the rate that would not removing the air and a slightly greater flow gradually remove the air over a 60 second rate removing the air. Results are shown in time. Pipe diameters were 0.75, 1, 1.5, 2, and Figure 22.

10 inches.

A test was conducted in 1 inch pipe with The GOTHIC predictions of liquid holdup were vertical sections 17 feet long with 180 degree within about 20% of the data.

bends above and below the vertical sections.

There were two risers and one downcomer. A mixture of air and water was introduced into the first riser and liquid holdup was determined.

Figure 20. Horizontal Flow Test Configuration Page 35

Figure 21. Horizontal Flow Test Comparison Figure 22. Clearing Air from a Vertical Pipe Page 36

3.3.1.3 Water Hammer GOTHIC was compared to (1) theoretical wave velocities for three pipe materials, (2) wave velocities in water with air fractions, and (3) measurements in three facilities. In Item 3, two pipe configurations were used with a rapidly closed valve and one created a pressure wave by opening a rupture disk. These are described in the following subsections.

3.3.1.3.1 Wave Velocity as Function of Pipe Material A transient was postulated as initiated in a 118 ft long 3/4 inch diameter closed pipe by postulating a 1 psi instantaneous pressure increase at one end of the pipe. Three pipe materials were used and GOTHICs predicted wave speed was compared to a theoretical determination. The difference between GOTHIC and theory was between -1.2% and +3.3%

over the range of material properties. Results are summarized in the following table:

Pipe Material GOTHIC Wave Speed, ft/sec Theoretical Wave Speed, ft/sec Rigid 4840 4897 Copper 4256 4224 Plastic 1408 1362 3.3.1.3.2 Effect of Air on Wave Speed A comparison was reported with a 100 ft long pipe with a 1 ft2 flow area. Results are shown in Figure 23.

Figure 23. Wave Speed as a Function of Air Content Page 37

3.3.1.3.3 EPRI Waterhammer Benchmark Problems The tests involved two slightly sloped 8 inch diameter pipes. One pipe was 240 ft long connected to a reservoir to maintain a constant pressure. The downstream end was either connected to a 60 ft pipe or to an air tank. Flow was initially zero and was initiated by breaking a rupture disk. One difficulty with modeling this configuration was the sloping pipes. GOTHIC was modeled with several horizontal pipes with the elevation changes represented by the connecting junctions to match the gravitational head changes. Selected comparisons to test data are shown in Figures 24 and 25.

Figure 24. Pressure Change Near Upstream End of Short Pipe Figure 25. Pressure Change Near Upstream End of Long Pipe Page 38

NAI concluded that GOTHIC is capable of determining: 1) if there is a potential water hammer problem in a piping system, 2) of estimating the magnitude of potential pressure spikes, including the effects of non-condensing gases in the system, and 3) of evaluating the effectiveness of mitigating modifications made to the piping system.

3.3.1.3.4 Delft Hydraulics Laboratory Test The test consisted of 0.09 m ID horizontal plexiglas pipe between two reservoirs that were 40 m apart. A high level water reservoir was connected to one end and a low level reservoir was connected to the other end. A simulated U tube condenser was located near the center of the horizontal pipe. A valve with a one second closure time was located at the upstream end of the pipe. The test consisted of obtaining a steady state water velocity of 1.11 m/sec followed by closing the valve. Selected comparisons are provided in Figures 26 and 27. (Solid lines are GOTHIC predictions; dash lines are data.) Excellent agreement is seen for the first 2 or 3 seconds and then the peaks differ although GOTHIC over-predicts the pressure data. NAI attributed this as probably due to piping system flexibility and simplified treatment of the 12 condenser pipes with 2 pipes in the GOTHIC model.

Figure 26. Water Velocity Near Inlet Valve Page 39

Figure 27. Pressure at Exit of Top Condenser Tube 3.3.1.3.5 Fauske and Associates (FAI) Pressure Surge Tests This test series was designed to investigate water hammer for configurations similar to the one illustrated in Figure 28. Almost 250 tests were accomplished. The test apparatus consisted of Figure 28. Configuration for Water Hammer Test Page 40

approximately 120 ft of 2 inch diameter steel piping that was closed at one end and connected to a constant pressure reservoir at the other end via a pump, a fast ball valve, and a check valve. Elevated sections of the steel piping contained one or more air bubbles. A test was initiated by starting the pump and opening the ball valve. Air in the pipe downstream of the pump allowed water to be accelerated by the pump. As the air compressed, the pressure rose and reversed flow in the line. This could close the check valve and initiate a valve closure water hammer in addition to the initial water hammer associated with the air compression / expansion.

The objective of the GOTHIC study was to show that GOTHIC could reasonably reproduce the pressure transients for cases with and without the check valve installed, followed by an investigation of the influence of the check valve location on the pressure transient. The GOTHIC comparison was performed with the single elevated pipe section configuration and compared GOTHIC predictions with two tests, one with and one without the check valve installed. Pressure in the elevated pipe section without the check valve is shown in Figure 29.

Figure 29. Pressure in Upper Horizontal Pipe Page 41

Correspondence between the initial GOTHIC predictions and test data is excellent. Timing of later oscillations drifts apart but the important aspect of this information is the accuracy of the initial prediction. The initial pressure spike was essentially identical with the check valve and the oscillations diminished more rapidly in both the test and the GOTHIC predictions. The pump shutoff head at this location was about 18 psig. Of note is the peak pressure of about 400 psig, substantially greater than the pump shutoff pressure and a clear indication that these types of water hammers could open relief valves.

Distributed gas can cause multiple peaks that did not occur when all of the gas was in one location. These peaks could be greater than discussed above. The ability of GOTHIC to assess this condition was not provided. 17 3.3.2 Purdue Tests 18 Two-phase, two-component transient fluid flow data in pipes larger than two inches in diameter were essentially non-existent before the Purdue test program that is described in WCAP-17271 (Westinghouse, October, 2010). The two inch diameter is also important because, as stated in WCAP-17271, the transition to a diameter not having an effect on the drift flux distribution coefficient for slug/froth flow is about two inches. Yet much of the concern with determination of fluid transport in nuclear power plants is in pipe diameters larger than two inches.

The WCAP provided data for 4, 6, 8, and 12 inch diameter piping in testing at Purdue University that was funded by the Pressurized Water Reactors Owners Group (PWROG). The general configuration was an upper horizontal header that was connected to a vertical downward-oriented test section that was connected to a lower horizontal pump suction header. Lengths were typically about [ ]feet, respectively. The configuration applies to many PWR suction pipe configurations although lengths differ. The tests covered 84 transient air/water test conditions with two to four repeat runs for each test condition. System flow rates approximated startup and running of an ECCS pump.

Most tests were run at about 21 degrees Celsius (°C) (70 °F). Some tests, termed "heated test section" tests, were conducted at 80 °C (128 °F) with 4 inch diameter pipes. In comparison to the low temperature tests, some heated test cases resulted in a large increase in gas volume at the top of the vertical test section and it sometimes resulted in doubling the time it took for complete gas entrainment to occur. Rapid condensation occurred in the vertical test section as pressure increased with decreasing elevation due to the head of water. The NRC staff has determined that assuming thermodynamic equilibrium for these cases is reasonable.

Test pressure was lower than generally occurs in plant ECCSs and an initial upper horizontal header pressure decrease was not typical of plant systems. The lower pressure caused voids to be a stronger function of elevation than in ECCS piping. The initial pressure decrease introduced a transient that is not present in an ECCS. The NRC staff does not judge this to detract from the data usefulness because an analysis methodology that predicts the challenging test conditions should be applicable to similar configurations in plant applications.

17 Additional information on GOTHIC modeling of these tests is provided in Section 4.4, below.

18 Taken from (NRC, March, 19, 2013) and provided here for completeness.

Page 42

The tests indicate data weaknesses exist downstream of the vertical downcomer for a range of gas volumes. For example, a hydraulic jump in the lower horizontal pipe potentially causes a significant increase in downstream gas flux in comparison to the test results.

A scaling analysis provided general correlations for the dominant phenomena observed in the testing, which included flow initialization via a vertical kinematic shock and vertical down-comer to horizontal elbow distribution. The resulting empirical correlations from the scaling analysis are acceptable for pipe diameters ranging from 4 inches to 30 inches.

WCAP Section 9.5 addressed differences between the 6 inch and other Purdue test results.

Purdue and Westinghouse believe the 6 inch piping may have been tilted and the 5 percent initial void fraction cases were not used in the scaling analysis. In light of this observation, care must be taken in using the 6 inch test results since the effect was likely to have resulted in an initial void fraction that was less than believed.

Empirical correlation predictions are limited to estimating uncertainties. Further, scaling correlation uncertainties should be increased when applied due to the effect of the assumptions, the limited data, and the stochastic nature of the data.

WCAP Section 10.3.3 provides a proprietary correlation to address gas distribution behavior in an elbow connecting a vertical downcomer to a horizontal pipe. Although this is claimed to provide some benefit when evaluating pump suction void fraction, the WCAP states that other issues must still be addressed. This includes possible gas re-accumulation associated with a kinematic shock or flow stratification in a tee or off-take and subsequent instability that may lead to a large gas flux downstream. No test data covered horizontal elbows nor did the WCAP address changes in behavior as a function of elbow radius. The WCAP identifies that (Andreychek T. S., July 1988) documented a detailed review of off-take geometry associated with the RHR connection to PWR hot legs, but the studies were performed with a low velocity in the header pipe, and are not applicable to a wide range of flow regimes. 19 The NRC staff concludes that the modeling of two phase, two component transient flow must be conducted with allowance for these data weaknesses.

A Phenomena Identification and Ranking Table (PIRT) process was discussed that concluded it is likely that additional testing efforts will be needed in areas such as horizontal flow stratification that can lead to a build-up and surge in downstream gas flux. 20 Furthermore, for conditions where large gas volumes exist, phenomena that were not investigated as part of the test program could occur. For instance, no information is available to determine what occurs when a kinematic shock reaches a downstream flow obstruction. Improved understanding was identified as needed regarding kinematic shock at vertical plane elbows, vortexing at off-takes, phase separation at tees, flow stratification in horizontal pipes, and pump entrance phenomena/piping entrance configuration.

19 Tests in this area have also been reported in (Chang, 1995), (Kim, September 1992), and (Wallis G. ,

1969).

20 The work is addressed in Swartz, M., Phenomena Identification and Ranking Table (PIRT) to Evaluate Void Fraction / Flow Regime at ECCS, RHR and CS Pump Suctions, Westinghouse Electric company LLC, WCAP-17167-NP, Rev. 0, December, 2009. The report was not provided to the NRC although members of the NRC staff have read the report and judge it to provide excellent coverage of the state of knowledge. The WCAP Section 4 summary is sufficient for the review being reported here.

Page 43

The WCAP also discussed code modeling and use of NFr. The code modeling discussion is generally consistent with the NRC staff position that codes or correlations must be validated through comparison of predictions and applicable test data or by other appropriate methods.

There are minor differences between the NFr discussion and NRC usage.

The NRC staff has found that the Purdue report provided a valuable addition to available information applicable to two-phase two-component transient pipe flow. Use of the data in acceptably verifying void transport methodologies was recommended subject to the conditions identified in the NRC staffs safety evaluation of NEI 09-10 that included an assessment of the Purdue tests (NRC, March, 19, 2013).

With respect to the Purdue tests, (George, January 10, 2012) stated that:

In reviewing the GOTHIC and data comparisons for the local void fractions , it must be recognized that the GOTHIC results represent the average void fraction over the volume of a computational cell used in the GOTHIC model (typically on the order of 1 to 2 ft long) while the data is for a comparatively thin plane at the measurement location. The GOTHIC results therefore tend to be much smoother than indicated by the fluctuation in the data. The prediction of these short duration fluctuations is neither necessary or desirable since the primary interest is the sustained gas flow into the pump. The fact that the GOTHIC results do not bound the measurements should not be construed as a non conservatism in the GOTHIC results. In fact the opposite is more likely the case. For example, a high tempory void measurement in the vertical pipe indicates a momentary hold up of some gas at the measurement location. The lower void predicted by GOTHIC indicates that void has progressed down the pipe.

Although the tests have a specified initial void in the upper pipe, the actual initial void varied somewhat from the target value. The initial void in the GOTHIC model was adjusted to approximate the void measured by the parallel wire probes in the upper pipe run.

3.3.2.1 Comparison of GOTHIC to Purdue 6 Inch Test Data As stated above, Purdue and W believe the 6 inch piping may have been tilted. Therefore, care must be taken in using the 6 inch test results since the effect was likely to have resulted in an initial void fraction that was less than believed. The examination of the 6 inch Purdue data and its comparison to GOTHIC predictions 21 will be accomplished while recognizing the potential limitation since doing so provides an excellent example of the difficulty of assessing GOTHIC for the short-time transient conditions that are of concern and much of the discussion is applicable to the Purdue tests with other pipe diameters.

Figure 30 (WCAP-17271-P, October, 2010.) illustrates the Purdue test configuration constructed with 6 inch diameter piping. The PW (parallel wire induction) probes measured void fraction and are accurate only in a stratified flow regime. The RIMP (ring impedance meter), AIMP (single arch impedance meter), and DAIMP (double arch impedance meter) also measure void fraction.

Layout dimensions are provided in Figure 31 (WCAP-17271-P, October, 2010.).

21 The latest GOTHIC version at the time of the NAI report, 7.2b(QA), was used for these analyses.

Page 44

Figure 30. Purdue 6 Inch Test Loop The important pipe lengths are the upper horizontal pipe = 30 ft, vertical pipe = 26 ft, and lower horizontal pipe = 20 ft. Note that the lengths, particularly the vertical and lower horizontal pipes, are not typical of many plant installations and this must be considered when applying GOTHIC.

Pressure was measured at P2 and P3 in the upper and lower horizontal pipes, respectively.

Test setup was accomplished by operating the pump at a steady flow rate through the secondary loop with valve AAV2 open and AAV1 and AAV3 closed with a known level of gas void introduced into the upper horizontal pipe at atmospheric pressure. The test report (WCAP-17271-P, October, 2010.) states The flow initialization valves, AAV1, AAV2 and AAV3, were air actuated butterfly valves with variable stroke time capability from 2 s to 60 s. Valve AAV2 (bypass line) was closed 4 seconds before AAV1 and AAV3 were opened to minimize its effect on the initial transient. AAV1 and AAV3 were then triggered to open together. All valves were set to stroke open or closed such that the flow initialized over a 7 s period, as measured by the M2 measurement location. No scaling of the flow initialization was performed as these times were considered prototypic.

Page 45

Figure 31. Purdue 6 Inch Test Loop Isometric NAI (Wiles, June 26, 2009) contradicted aspects of the test report as follows: The transient opening of the AAV1 and AAV3 valves is very important in the early behavior of the air bubble.

Although the test report indicates that both valves start to open automatically 4 seconds after AAV2 starts to close, the measured pressure and flow data are not consistent with this specification. For each of the tests evaluated here, the pressure data suggests that valves AAV1 and AAV3 begin to open at approximately 6 seconds. It is reasonable to expect that the flow begins shortly after the valves begin to open. However, the measured flow at M2 indicates that flow into the top horizontal pipe begins between 9 and 10 seconds. The inconsistency was resolved by adjusting the flow versus time function as necessary such that flow begins at approximately the same time as the pressure begins to decline. The inlet flow is assigned to boundary condition 1F.

Page 46

The NAI description appears to misinterpret the behavior. Figure 32 shows flow rate through AAV2 begins to decrease at 4.0 seconds 22, indicating that the valve is closing. Figure 33 shows flow rate through AAV1 initiating at 7.6 seconds, indicating that AAV1 starts opening 3.6 seconds after AAV2 starts to close, consistent with the Purdue reports 4 seconds. Figure 33s 7.6 seconds is inconsistent with NAIs 9 and 10 seconds. Figure 34 and Figure 35 are consistent and indicate a sudden pressure spike over most of the downcomer length that initiates at 7.0 seconds and peaks at 7.6 seconds, consistent within the data error of AAV1 opening and allowing flow to initiate into the end of the upper horizontal pipe. This pressure spike corresponds to the initial pressure decrease from atmospheric pressure at about 7 seconds that is shown in Figure 36 that is discussed below. The conclusion is that NAIs statement that the measured flow at M2 indicates that flow into the top horizontal pipe begins between 9 and 10 seconds is incorrect. However, NAIs pump flow table on Page A30 of A39 shows flow initiating at 6.09 seconds, increasing substantially by 11.64 seconds, and becoming almost constant after 17.07 seconds, essentially consistent with Figure 33. NAIs matching of flow initiation to the initial pressure response is correct.

Figure 32. Flow Rate Through M1 22 The times are scaled from the figures to two significant figures.

Page 47

Figure 33. Flow Rate Through M2 Figure 34. Differential Pressure DP1 Page 48

Figure 35. Differential Pressure DP2 The upper horizontal pipe pressure always immediately decreased to about -8 psig and the gas void moved to the downstream end of the upper horizontal pipe where it was then entrained over time into the vertical pipe. The experimenters attributed this initial movement to the initial pressure equilibration when AAV1 and AAV3 were opened (George, January 10, 2012).

As discussed above, GOTHIC noding is shown in Figure 1 where Volumes 2, 4, and 6 model the upper horizontal, vertical, and lower horizontal pipes, respectively. Volume 1 is an inlet volume, 7 is an outlet volume, and Volumes 3 and 5 represent long radius elbows. Connections between volumes are by 3D connectors. The 3D connectors for the corners were assigned a loss coefficient of [ ]. For each corner, the total loss coefficient is [ ], which is stated to be consistent with values given in (Crane, December, 2001) for a long radius elbow. The measured test section water flow rate was specified at flow boundary condition 1F. Water was discharged to pressure boundary condition 2P. The path that connects volume 7 to boundary condition 2P was assigned a loss coefficient of [ ]. This represented outflow into the separator tank. The total dynamic loss through the modeled portion of the test section was [ ] with the balance of the loss due to wall friction.

Crane gives resistance coefficients for standard elbows as K = 30 f where f is the friction factor, given as 0.015 for turbulent flow in clean commercial steel pipe and significantly less for smooth pipe such as drawn tubing and likely for the transparent pipe used in the tests. Thus, K

[ ] for standard steel elbows. The multiplier used in GOTHIC was [

], which is reasonable for multiple mitre bends or long radius elbows and may be high for smooth plastic pipe.

Page 49

Gothic model boundary conditions are flow through Valve 1V and pressure at 2P. Pressure at 2P was not measured and pressure at 2P was selected so that the calculated pressure at test location P3 approximately matched the data.

Noding detail was provided in Figures 2 - 4, above. (Figure 3 is incorrect in NAI -1459- 001.

Consistent with the response to the Request for Additional Information (Balakhnin, April 9, 2013), the vertical noding illustrated in NAI-1459-003 (George, January 10, 2012) is used.

Consequently, the dimensions are for the 4 inch Purdue test facility and are not consistent with the 6 inch test facility dimensions although the configuration is consistent.) The pipes are modeled using two-dimensional representations consistent with the generally expected behavior.

Two 6 inch tests were calculated, one with an initial upper horizontal initial void fraction, , of 0.10 and an end of test NFr = 0.80; the other with 0.20 and 1.65, respectively. NAI stated that Repetition 3 data were used for the 10% void, 0.80 NFr test because the first two repetitions at those conditions did not appear to achieve the desired initial void fraction in the upper horizontal pipe section. The void fraction measurements from Repetition 3 of the 10% void experiment includes some apparent bias, since the applicable instruments showed constant void fractions ranging from 1% to 5% even before the experiment began. These initial bias measurements were subtracted from the test results for the purpose of comparison with GOTHIC predictions. This is consistent with the sloped pipe observation identified in the beginning of this Section.

GOTHIC prediction of pressure behavior for = 0.10 and NF = 0.80 is shown in Figure 36 that is reproduced from (Wiles, June 26, 2009). The lines are the GOTHIC values and the symbols are representative of the data. The upper curve is the GOTHIC-calculated pressure at P3 in the lower horizontal pipe. The lower curve is the boundary condition input to GOTHIC for P2, the pressure in the upper horizontal pipe that was selected to provide an approximation of P3. P3 is accurately predicted.

Figure 36. P2 and P3 in 6 Inch Pipe for 10%. 0.80 Page 50

The symbols are somewhat misleading because the data are available as essentially continuous values versus time and the symbols are somewhat arbitrary representations of the continuous values which are often noisy. There is also sometimes a wide variation between tests. These characteristics are illustrated in Figures 37 and 38. Figure 37 illustrates the data provided by NAI when comparing to GOTHIC predictions. Figure 38, taken from Volume 2 of the Purdue report, shows the data plot for the four runs with = 10% and NF = 0.80. NAI elected to compare to R3. The author of this report believes it would have been better to compare to the average of the four runs.

Figure 37. Representation of Data used for Comparison to GOTHIC Prediction Figure 38. Void Data at Upper Downcomer Page 51

Figure 39 shows a comparison of the predicted void fraction near the top of the vertical pipe to the data. GOTHIC over-predicted the data by about a factor of two and didnt predict the decrease as rapidly as demonstrated by the data.

Figure 39. - Top of Vertical Pipe Figures 40 and Figure 41 show similar behavior near the bottom of the vertical pipe although the initial transient is well predicted whereas the behavior following the initial transient is missed and is accompanied by an over-prediction of the void fraction. Figure 42 shows the Purdue data where there is significantly better agreement between the four runs.

Figure 40. - Bottom of Vertical Pipe Page 52

Figure 41. in Lower Horizontal Pipe Figure 42. Data Near Bottom of Downcomer Page 53

It is of interest to compare the above 10% behavior near the bottom of the downcomer to observed behavior with a 5% initial volume and the same NFr of 0.8. A clear initiation of homogeneous bubbles was observed followed by an increasing for about 12 seconds where larger bubbles were observed. About 6 seconds later, a larger void was observed that initially moved upward and then gradually moved downward. Voids were essentially gone about 28 seconds after they were initially observed. 23 Timing was similar to Figure 42. Similar initial behavior was observed with a 20% initial volume and NFr = 0.6 but significantly larger voids moved up the downcomer and they broke up as they moved and the pieces were then observed to move downward. Lower elbow and lower horizontal pipe void behavior was also illustrated for a 5% initial volume and NFr = 0.6 in a 12 inch pipe. Voids were observed to collect along the inside of the elbow toward the horizontal pipe and then move back into the downcomer.

Downstream in the lower horizontal pipe, all voids were observed to move in a thin layer along the top of the pipe.

NAI mentioned that the GOTHIC predictions in the lower horizontal pipe are generally higher than the measured values for the low velocity case when the flow would tend to be more stratified. For this case, the parallel wire measurement may be more appropriate. The measurements at parallel wire indicator PW3 are in better agreement with the GOTHIC results.

The variation in the measured results by the different instruments is an indication of the difficulty and level of uncertainty in measuring pipe average voiding. The PW3 void measurement, located in the bottom horizontal pipe immediately following the vertical pipe, is provided in Figure 43.

Figure 43. PW3 Data 23 Information obtained from a 29 MB video that is too large to be included in this report.

Page 54

The NAI results corresponding to the Figure 43 location are provided in Figure 44. The predicted initiation time and peak void fraction is in excellent agreement with the Figure 43 data although the predicted peak void fraction is late.

Figure 44. Bottom Horizontal Pipe Void (PW3)

The results of comparisons for the upper horizontal pipe = 0.20 and NF = 1.65 are as follows:

Location NRC Observation P2 and P3 Excellent agreement.

- Top of Vertical Pipe Excellent agreement.

- Bottom of Vertical Pipe over-predicted by about 50%. Timing good.

in Lower Horizontal Pipe Excellent agreement.

3.3.2.2 Comparison of GOTHIC to Purdue 8 Inch Test Data Discussion of the 6 inch nodalization and assumptions applies to the 8 inch tests. GOTHIC predictions compared to data for = 0.10 and NF = 0.80 are as follows:

Location NRC Observation P2 and P3 Excellent.

- Top of Vertical Pipe GOTHIC under-predicted by close to a factor of two.

- Bottom of Vertical Pipe GOTHIC under-predicted peak by factor of five. Appears to fit much of data if scatter above prediction is neglected. Discussed further in Section 3.3.2.5, below.

in Lower Horizontal Pipe Excellent timing. Over-predicts data by about a factor of two.

Page 55

And, for = 0.20 and NF = 1.65:

Location NRC Observation P2 and P3 Excellent.

- Top of Vertical Pipe Excellent timing. GOTHIC slightly under-predicts data.

- Bottom of Vertical Pipe Excellent timing. Slightly under-predicts data.

in Lower Horizontal Pipe Under-predicts data by about a factor of two. Good initial timing but data peak and decay significantly delayed beyond prediction.

The 6 and 8 inch pipe test results show that it is necessary to further examine behavior near the bottom of the vertical pipe and in the lower horizontal pipe. This is done in Section 3.3.2.5 after examination of the 4 inch and 12 inch pipe test data.

3.3.2.3 Comparison of GOTHIC to Purdue 4 Inch Test Data Figure 45 illustrates the Purdue 4 Inch test facility. The upper horizontal pipe length = 153, the downcomer length = 129, and the lower horizontal pipe length = 76.

Figure 45. Purdue 4 Inch Test Facility Instrumentation is shown in Figure 46.

Page 56

Figure 46. Purdue 4 Inch Facility Instrumentation Noding is identical to the 6 inch and 8 inch noding illustrated in Figures 1 - 4. Subdivided volumes were used for the [

]. Boundary conditions were consistent with the 6 inch modeling with one exception. The 6 inch and 8 inch GOTHIC models employed a valve at the exit to the pressure boundary condition. The GOTHIC model for the 4 tests was simplified by removing the valve and by assuming that the pressure drop for the piping past instrument P3 would be minimal.

GOTHIC predictions compared to data for the Purdue 4 inch tests for = 0.10 and NF = 0.80 are as follows (George, January 10, 2012):

Page 57

Location NRC Observation P2 and P3 Excellent agreement between GOTHIC prediction and test data. NAI stated that initial pressure difference between P2 and P3 was due to the static water head. GOTHIC calculated this correctly in contrast to the data which indicated the elevation change was about a foot longer than shown in the piping layout diagram and that this discrepancy was observed in all tests that NAI investigated. It also attributed the larger predicted pressure oscillations as due to damping in the gas separator tanks that were not included in the GOTHIC model. These are reasonable conclusions.

- Upper Excellent agreement although initial GOTHIC value is 0.15 instead of 0.10. Data show Horizontal long-term gas holdup & some spikes near downcomer that are not predicted by GOTHIC, Pipe probably because of simplistic elbow modeling that NAI used as a conservatism to minimize holdup and maximize gas flow to pump - an approach that is somewhat inconsistent with its previously stated approach of realistic modeling of tests versus conservative modeling of GOTHIC application to plant behavior.

- Top of Good agreement with initial transient, fair agreement with decrease during one test, poor Vertical agreement with second test. Under predicts volume, especially in second test. NAI Pipe believed may have been stationary volume in second test that GOTHIC did not predict.

This would be consistent with generation of waterfall in top of downcomer.

- Center Excellent agreement with initial transient and initial decay. Completely misses second of Vertical peak. Result is significant under-prediction of volume. See next item discussing bottom Pipe of downcomer.

- Bottom Excellent prediction of initial transient. Completely under-predicted duration and volume.

of Vertical NAI stated that integration of predicted volume over time assuming gas and water move Pipe and together corrected for pressure change equals initial gas volume. Again, elbow treatment bottom that promotes gas transfer rate through elbow promotes gas transfer toward pump. NAI elbow conclusion that some gas holdup occurs that causes misleading results is reasonable.

However, promoting gas transfer rate that doesnt allow gas to accumulate may miss later gas movement as potential slug. NAI also believed parallel wire probes were misleading because their accuracy is limited to stratified flow, not the conditions that were believed to exist.

- Lower Calculation is at pump end of pipe and void measurement near the end. Accurate Horizontal prediction of void arrival time and significant over-prediction of initial volume. Accurate Pipe prediction of second peak but completely misses peak volume and post-peak behavior.

Data behavior consistent with downcomer behavior. See above discussion. Sudden data peak near zero time of essentially zero duration followed by ~ 0.01 void fraction that persists as minimum over entire transient - may be due to initial depressurization and outgassing. Perhaps bubble persists due to effect of elbow and non-uniform water flow that traps bubble - selected elbow modeling will not correctly characterize this behavior but 7 ft distance from elbow should have allowed flow pattern to become better established. Tests with other diameters have shown void followed by hydraulic jump followed by stratified flow with a small void at the top of the pipe at some distance downstream of elbow. NAI discussion that may assign behavior to data uncertainty is not convincing and this reviewer believes the data indications that are indicative of gas not moving with the water are correct. This is a region where behavior is poorly understood and more data are needed. For now, an acceptable assumption is that if the pump is upstream of where the hydraulic jump would occur, then the void will not accumulate and the GOTHIC calculation will apply. If the pump is downstream of the hydraulic jump position, then stratified flow will exist and the void fraction entering the pump will be less than predicted but will extend for a longer time assuming there is no perturbation that causes the upstream void to suddenly move downstream.

Page 58

And, for = 0.20 and NF = 1.65:

Location NRC Observation P2 and P3 Near perfect correspondence to data..

- Upper Excellent prediction of initial behavior. Prediction is pipe cleared of gas almost Horizontal immediately at upstream end but data indicate a small void persists along the Pipe pipe for longer than predicted. Downstream end shows a short-time spike in data that is accurately predicted followed by rapid gas clearing while data take longer to clear. NAI stated that the data indicated some additional time was needed to clear all of the gas but the held up volume was small as evidenced by results in the lower horizontal pipe. This is clearly correct for the upstream end but downstream, = 0.2 to 0.1 before gas was removed at about 19 seconds.

- Top of Data peak at 1 at 11 - 12 seconds consistent with behavior in downstream Vertical end of upper horizontal pipe. Prediction was slightly later with peak at 0.55 Pipe followed by decrease to 0 at 18 seconds. Data showed rapid decrease from peak to 0.5 that lasted until 18 seconds followed by a rapid decrease to 0 at 20 seconds.

- Center Data peak at 0.8 at 12 seconds followed by rapid decrease to 0.15 and of Vertical linear decrease to 0 at 22 seconds. Predicted increased from 0 at 12 seconds, Pipe peaked at 0.37 at 15 seconds and decreased to 0 at 19 seconds.

- Bottom Data differ between runs. Both voids initiated at 13 seconds. One increased to of Vertical 0.3 at 14 seconds and decreased to 0 at 24 seconds. The other increased to 0.7 Pipe at 19 seconds, decreased to 0.6 at 28 seconds, and decreased to 0 at 29 seconds. Predicted curve is similar to center prediction, with initiation at 14 seconds, peak = 0.26 at 16 seconds, and decrease to 0 at 20 seconds. NAI stated that integration of the predicted behavior in the vertical pipe agreed with the total initial volume, indicating that the data showing a void holdup involved only a small gas quantity.

in Lower Time behavior was well predicted although the data showed a slight delay in Horizontal removal of residual void. Predicted 32 inches from the elbow was low by about Pipe a factor of two whereas good agreement is observed at 63 inches. NAI stated that data from the parallel wire probes was not used since the flow was not expected to be stratified. This is a change from the NAI conclusion for the 6 inch tests at the lower NF = 0.8 3.3.2.4 Comparison of GOTHIC to Purdue 12 Inch Test Data The layout for the 12 inch Purdue facility is shown in Figure 47 and instrumentation is shown in Figure 48.

Page 59

Figure 47. Layout for Purdue 12 Inch Facility Page 60

Figure 48. Purdue 12 Inch Facility Instrumentation Noding is identical to the 4 inch, 6 inch and 8 inch noding with dimensional changes consistent with the piping diameter. Elbow modeling selections are identical to the 4 inch modeling except that 3D connectors for elbows were assigned a loss coefficient of [

]. Boundary conditions were consistent with the 6 inch modeling. The GOTHIC model for the 12 test was simplified by removing the exit valve and by assuming that the pressure drop for the piping past Instrument P3 would be minimal. To make M2 flow consistent with the P2 pressure data, [

].

GOTHIC predictions compared to data for the 12 inch tests for = 0.10 and NF = 0.80 are as follows:

Page 61

Location NRC Observation P2 and P3 Near perfect calculation with the exception of predicted oscillations and a slight under-prediction after 30 seconds in the lower horizontal pipe. The oscillations are not of concern as discussed above.

- Upper Initial is 0.12 versus it should be 0.10. Near perfect prediction of elimination of Horizontal void in upstream end. Good prediction of downstream end although data spike is Pipe higher than predicted, prediction is about 2 seconds late, and post-spike is initially over-predicted and then goes to zero while data show 0.03.

- Top of Excellent initial timing. Predicted spike is 0.57 versus data of 0.88 and decay is Vertical faster than predicted which may be due to the upper elbow modeling.

Pipe

- Upper Excellent prediction with the exception that decay below = 0.1 is faster than middle of shown in data, consistent with the top of the pipe behavior.

Vertical Pipe

- Lower Excellent prediction although peak is a little late. Data for one run show spikes middle of after prediction which may be large bubbles starting to move up the downcomer.

Vertical First spike occurs at 36 seconds. Spikes were not seen in the other runs. The Pipe reviewer postulates large bubble behavior that illustrates sporadic nature of voids under these transient conditions at relatively low NFr due to gas accumulation immediately downstream of and in the lower elbow.

- Bottom Excellent prediction although peak is a little late and predicted to be 0.16 versus of Vertical data at 0.12. Data shown spikes for one run after prediction which may be large Pipe bubbles starting to move up the downcomer. First spike occurs at 33 seconds and extended higher occurs some of the time in comparison to lower-middle.

- Lower Excellent prediction of initial behavior with significant under-prediction of data that Horizontal shown an extended region of void after the peak that is not predicted. NAI stated Pipe that prediction is consistent with all original gas assuming homogeneous flow into the pump and There is not enough gas in the original bubble to result in the high void fractions indicated by the data over the long period of gas removal.

And, for = 0.20 and NF = 1.0:

Location NRC Observation P2 and P3 Good prediction with oscillations as discussed above and slight under-prediction after 24 seconds.

- Upper Excellent prediction. Peak void at mid-pipe prediction of 0.32 versus data of 0.4.

Horizontal Data show slow increase in mid-pipe after 40 seconds that is not explained.

Pipe PW2 near downstream end of pipe shows initial = 0.26 that is not consistent with other initial s, remains at 0.02 to 0.03 for some time after peak, and increases after 50 seconds. The increase after 40 to 50 seconds was shown by two measurements and appears to be real. The implication is that something is occurring to cause gas to re-enter the downstream end of the horizontal pipe from the downcomer.

- Top of Excellent initial prediction, peak prediction of 0.7 versus data of almost 1, with Vertical significant under-prediction of post-peak decrease that ends with a prediction of Pipe zero at 24 seconds with data reaching this value at 55 seconds. Correspondence of the predicted shape versus data is outstanding.

Page 62

Location NRC Observation

- Upper Excellent prediction with slight under-prediction of toward end of transient.

middle of Vertical Pipe

- Lower Excellent timing throughout transient. 0.4 peak under-predicts data for some runs middle of of approximately one.

Vertical Pipe

- Bottom Excellent timing with peak predicted > 0.35 versus data peak of 0.25 with broader of Vertical time near peak Pipe in Lower Excellent initial prediction with peak > 0.25 with PW3 data at 0.2 or below.

Horizontal Predicted peak decays to zero at 33 seconds while data are noisy and show a Pipe void beyond 60 seconds. Further downstream, RIMP2 shows similar behavior except greater before reaching zero at 60 seconds.

3.3.2.5 Assessment of GOTHIC Prediction of Lower Vertical and Lower Horizontal Pipe Purdue Tests As identified in Section 3.3.2, above, flow stratification in horizontal pipes and pump entrance phenomena/piping entrance configuration aspects are areas where the phenomena are not well understood and the Figure 49 factor of five under-prediction of some of the data warrants further consideration.

Figure 49. Near Bottom of Vertical 8 Inch Pipe First observe that GOTHIC over-predicts the smooth data if the scattered points above the prediction in Figure 49 are neglected. Next, postulate that the high points represent bubbles that are not moving downward as fast as the water or that are actually moving upward but breaking up as they go so that they then move downward, and that GOTHIC did not capture this Page 63

behavior in part because the data are essentially for a horizontal plane and GOTHIC represents average void within the pipe over a finite vertical length.

Next consider the behavior immediately downstream of the Figure 49 location in the lower horizontal pipe as shown in Figure 50 for this case ( = 0.10, NF = 0.80).

Figure 50. in Lower Horizontal 8 Inch Pipe Here, GOTHIC significantly over-predicts the data with the largest over-prediction roughly corresponding to the Figure 49 high data points. This is consistent with the postulate of void hold-up in the lower vertical pipe that was not predicted by GOTHIC.

Figures 51 and 52 provide the corresponding information for the 6 inch tests with a significant difference in the bottom of the vertical pipe. The data show a relatively small pulse that is over-predicted by GOTHIC. Behavior in the lower horizontal pipe is similar in the 6 and 8 inch tests. However, remember that the 6 inch Purdue tests may have been conducted with a slight pipe slope in the upper horizontal pipe that would result in the initial gas volume being smaller than documented.

Page 64

Figure 51. Near Bottom of 6 Inch Vertical Pipe Figure 52. in Lower Horizontal 6 Inch Pipe Page 65

Figures 53 and 54 show behavior at higher initial = 0.20 and NF = 1.65 in the 6 inch pipe.

GOTHIC over-predicts near the bottom of the vertical pipe and provides an excellent prediction of behavior in the lower horizontal pipe.

Figure 53. Near Bottom of Vertical 6 Inch Pipe Figure 54. Void in Lower Horizontal 6 Inch Pipe Page 66

Figures 55 and 56 provide the information for the 8 inch pipe. near the bottom of the vertical pipe is slightly under-predicted and in the lower horizontal pipe is under-predicted by a factor of two.

Figure 55. Void Near Bottom Of Vertical 8 Inch Pipe Figure 56. Void In Lower Horizontal 8 Inch Pipe Page 67

The conclusion is that GOTHIC results in these regions are scattered and a factor of two margin is necessary to reasonably encompass the data.

3.3.3 Comparison of GOTHIC to Millstone Test Data 24 (George T. L., August 25, 2010) reported that GOTHIC 8.0 is used for this analysis. The important models for interphase drag in version 8.0 are essentially the same as those in version 7.2b. Results from versions 7.2b and 8.0 for gas transport problems are very similar, so this validation also applies to version 7.2b. Version 8.0 includes some changes that make the code more robust for multiphase flow problems.

A plan view (piping viewed from above) of the 1/4 scale Millstone test facility is shown in Figure 55.

Figure 55. Millstone Test Facility Water is drawn from the water storage tank through several lengths of 6 inch pipe before reaching a 90º tee with a 2 inch pipe that connects to the bottom of the 6 inch pipe and simulates the flow path leading to the high pressure safety injection (HPSI) pipes. An elbow and 6 X 4 inch reducer are located immediately downstream of the Tee. These are closely followed by a 45º downward tee that simulates the flow path to the low pressure safety injection (LPSI) pipes. There is a 4 X 2 inch reducer downstream followed by 2 inch pipe that simulates the flow path to the charging pumps which provide a high pressure safety injection function. An initial 24 Test layout, data, and GOTHIC information obtained from (George T. L., August 25, 2010).

Page 68

test condition was obtained by introducing a known level of gas in the horizontal 6 inch pipe with the plastic air dam intact to prevent air from escaping to the water storage tank.

The GOTHIC noding diagram is shown in Figure 56.

Figure 56. Millstone GOTHIC Model The complexity of this model is greater than the models discussed above and is more typical of plant system modeling. Subdivided volumes are indicated by dash line boxes and lumped or homogeneous volumes by solid boxes. Volume 1P is a constant pressure boundary condition.

Volume 1 is the water storage tank. Volume 2 models the short horizontal and vertical piping between the tank and the horizontal pipes. Volumes 3s, 4s, and 5s are the horizontal 6 inch, 4 inch, and 2 inch pipes, respectively. Volume 3s encompasses horizontal elbows, a simplification discussed in the next paragraph. Pump 1P represents the HPSI pump, 2P the LPSI pump, and 3P the charging pump. Connecting piping is labeled SIH, RHR, and CP, respectively. Pumps were represented by built-in homologous curves in GOTHIC and values were assumed consistent with the range of measured flows. Flow resistance factors for the pump flow paths were adjusted so that the flow matched the steady state test flows, stated to be equivalent to adjusting the throttling valves.

Piping that remains at a single elevation can often be reasonably approximated by a single control volume to reduce modeling complexity even if it not straight provided such items as Page 69

elbows do not introduce air / water distribution effects that must be modeled. For example, if a horizontal tee were located immediately downstream of an elbow, modeling the air / water distribution would be important in contrast to location of a tee sufficiently distant from the elbow so that the straight pipe flow profile was fully re-established. This contrasts to modeling of vertical piping where buoyancy effects are important. In addition, changes in pipe size via reducers or reducing tees require a separate control volume. Although not applicable here, minor pipe slopes are usually ignored in GOTHIC modeling due to limitations of the rectangular coordinate system. In cases where sloped piping may trap gas, other modeling features are used to address air in the system.

The reducers have a tapered flow path and reducer modeling is important. The usual approach is to use 3D connectors between two adjacent subdivided piping segments of different diameters, an approach that will minimize gas holdup at the reducers. However, when the flow rate is low, flow in the 6 inch pipe will be stratified and test results will be a strong function of two phase flow behavior at the reducers. This was addressed by the Figures 57 and 58 nodalization.

Figure 57. Reducer Noding Figure 58. Reducer Noding Forty-five degree pipe runs were modeled using connected vertical and horizontal segments.

The height of the vertical segment is equal to the vertical drop of the pipe run and the length of the horizontal segment is set to L-H, where L is the actual pipe length and H is the height of the vertical modeling segment. This was stated to preserve the total length of the actual pipe.

(Balakhnin, April 9, 2013) provided the following discussion of modeling of tees in the Millstone analyses:

Page 70

The Tee to the SIH pump has its branch oriented vertically downward. This was model with a single flow path that runs from the appropriate bottom cell of the horizontal main run to the top cell in the vertical run with the end elevations and heights set so that the connection spans the attached cell height.

The Tee to the RHR pump has its branch oriented downward at 45 degrees. The 45 degree pipe run was split into two pieces representing horizontal and vertical runs. The horizontal run was connected to the main run with a 3D connector that spanned the vertical stack of 6 cells in the main at the location of the Tee. The other end of the horizontal run connects to the vertical run with a 3D connector.

The length of the vertical run was set to match the drop in the 45 degree run. The length of the horizontal run was set so that the combined lengths of the two runs equaled the length of the 45 degree run.

This demonstrates the need for a modeler who understands the unique aspects of GOTHIC when configuring the nodalization. Other noding is understood to be consistent with noding used for the Purdue tests.

NAI provided the following comparison of test and GOTHIC results:

Case Initial Void Pump Flows (gpm) Gas Transport (in3)

(%) measured/calculated RHR SIH CHRG RHR SIH CHRG 1 8 315 23 22 na/620 0/0 8.6/16.8 9 8 170 24 21 na/460 0/0 30.1/40.3 10 8 97 24 21 na/420 0/0 98.9/83.1 11 5 27 27 22 0/0 0/0 0/5.3 12 8 0 27 22 0/0 0/0.2 30.1/35.7 21 8 0 25 55 0/0 0/0 393/320-470 With the exception of the charging path in Case 10, GOTHIC over-predicted gas transport.

NAI stated that no significant gas was transported to the SIH and that cases with full RHR flow resulted in degraded RHR pump flow because the water / gas separator was overwhelmed.

Predicted void fractions immediately upstream of the water / gas separators for Case 1 are illustrated in Figure 59. The initial RHR pump peak of 0.48 is consistent with loss of the RHR pump and the resulting flow decrease may have contributed to void reaching the charging pump with a peak of 0.06. The second RHR pump peak at = 0.20 may be due to RHR pump recovery - another postulate since no data are provided upon which to base these postulates.

In any event, the GOTHIC predictions are reasonable.

NAIs conclusion for the Millstone investigation is that This validation gives additional confirmation that GOTHIC delivers good results for gas transport provided that the model includes sufficient noding detail to allow the basic interphase drag models in GOTHIC the freedom to predict the multiphase flow patterns in the piping system. This conclusion is substantiated.

Page 71

Figure 59. Case 1 Void Fractions at Separator Inlets 3.3.4 Assessment of GOTHIC Comparisons to Test Data Test comparisons that are accomplished for each GOTHIC version include frictional pressure loss, filling of horizontal and vertical pipes, water holdup and pressure drop in vertical pipes with upward and downward flow, water hammer, drop behavior including entrainment, many containment configuration and test conditions, and the Edwards pipe blowdown experiment that is somewhat unique because of the extremely rapid test transient. Consequently, application of GOTHIC to the Edwards pipe blowdown, two phase behavior, and water hammer tests are reviewed as part of the audit of code capability; the first because of the short-term challenges, the second to cover some behavior that lends insight into Robinson use of GOTHIC to address pipe void movement, and the last because GOTHIC was used at Robinson to evaluate water hammer. Further, tests that are representative of nuclear power plant system suction piping are studied. These tests were the Purdue University tests and another that covered a Millstone operability issue.

Test data applicable to verification of GOTHIC cover many phenomena but are incomplete.

Topics that are inadequately understood for the transient conditions covered herein include elbows, tees, sloped pipes, vortices, transitions between downcomers and downstream horizontal pipes, behavior of kinematic shock in the vicinity of flow obstructions, pump entrance phenomena / piping entrance configuration behavior, and some conditions such as horizontal flow stratification that could lead to a build-up and surge in downstream gas flux. Further, it is difficult to correctly model some areas in GOTHIC such as elbows, tees, sloped pipes, and Page 72

vortexes. These weaknesses have been considered in the audit in recognition of the need for a method that will provide assessment of operability consistent with the guidance provided in Section 2.1, above.

In general, GOTHIC predictions ranged from excellent agreement where the predictions essentially overlaid the data or analytic calculation to about a factor of two difference depending upon what was being compared. Void fractions were usually within a factor of two and event timing was often within a few percent or better. Some configurations required careful consideration such as elbows, tees, reducers, and sloped pipes. With respect to elbows, representing an elbow as a lumped volume treated it as a homogeneous configuration that would accurately predict pressure drop of smooth curves with tuning but would fail to capture the gas distribution within and immediately downstream of the elbow. Modeling the gas distribution required a multi dimensional model that would over-predict pressure drop.

GOTHIC calculates average properties in a computational cell in contrast to data that represent point or thin plane locations. Consequently, GOTHIC results tend to be smoother than some data and failure to describe individual data point outliers, particularly for many of the transients of concern here, is not a concern when analyzing nuclear power plant systems.

In water hammer tests, NAI demonstrated that GOTHIC could calculate wave velocities in pipe material within 3 % and was within about 20% in air / water mixtures. It provided an excellent fit to pulse timing, magnitude, and width and to predicted water velocity. With respect to water hammer, NAI concluded that GOTHIC is capable of determining: 1) if there is a potential water hammer problem in a piping system, 2) of estimating the magnitude of potential pressure spikes, including the effects of non-condensing gases in the system, and 3) of evaluating the effectiveness of mitigating modifications made to the piping system. However, the GOTHIC analyses of the FAI pressure surge tests discussed in Section 3.3.1.3.5 only covered tests in which the initial gas was in one location. Distributed gas can cause multiple peaks that could be of greater magnitude. The ability of GOTHIC to assess this condition was not provided The Purdue data are for individual locations where individual bubbles will introduce noise and short term spikes may be higher in contrast to the GOTHIC predictions that are often an average for a two foot pipe length. A good cross-check of the GOTHIC predictions where GOTHIC appears to under predict is integration of over time assuming homogenous flow to show that the predicted gas volume is consistent with the initial upper horizontal pipe volume.

This was stated for some cases. Failure of GOTHIC to predict holdup would generally correspond to GOTHIC predicting gas to reach a pump at a higher than if gas were predicted to be held up. The exception would be if held-up gas were to move downstream at a high after being held up and accumulating.

The use of the lumped parameter approach for modeling elbows used for the Purdue predictions is not consistent with behavior near the top of the downcomer where a two dimensional flow treatment is necessary and would show gas holdup that a lumped parameter approach may miss. This could be important where plant vertical piping lengths are shorter than in the Purdue tests. NAI provided Figure 60 to illustrate GOTHICs capability to calculate a two-dimensional elbow model prediction that illustrates the waterfall effect. No information Page 73

Figure 60. GOTHIC Calculation of Upper Elbow - 2 D Model was provided beyond the picture. NAI further stated that the largest discrepancy between the code results and the data were at the location just downstream of the upper elbow.

Nevertheless, the GOTHIC results are in good agreement with the data further down stream.

The amount of gas that is held up is small relative to the total gas volume and the local holdup has minimal impact on the behavior downstream. This may not be true for all cases.

NAIs Millstone investigation provided acceptable results with most void predictions being greater than the test information.

The conclusion is that the need for continuing changes at the time of this NAI work (2008 -

2010) reflects the immaturity of GOTHIC for study of the problems of concern here. Differences between versions and with respect to test data must be considered when using GOTHIC predictions.

The demonstrated complexities in the correct use of GOTHIC means that the modeler much be highly qualified to ensure system modeling is consistent with test modeling and to acceptably capture nuances that are unique to the system model.

Overall, the demonstrated GOTHIC predictions are reasonable. Consequently, GOTHIC is acceptable for determination of system operability provided (1) modeling is consistent with the test modeling, (2) an effective factor of two is used for void predictions, and (3) predicted water hammer results are considered to be semi-quantitative.

3.4 General Treatment of Conservatism (Wiles, June 26, 2009) stated that the principle difference between the benchmark and industry applications was including conservatisms in the industry analyses that were stated to be inappropriate for benchmark analyses. The claim is that the industry application results are appropriately conservative. This is explored further in this report section.

NAI considered the Purdue tests as representing simplistic, non-prototypic piping geometry with a limited set of flow conditions when compared to plant configurations. Further, gas transport in suction piping can be influenced by parameters not explored in detail in the tests. Therefore, NAI concluded that application of GOTHIC requires conservatisms to reasonably ensure that calculation of gas transport is defensible. Areas where conservatisms were applied to GOTHIC predictions were stated to include the following:

Page 74

• Flow rates were set to the maximum possible value since (1) this may predict gas movement through a system with less attenuation in a piping run and (2) this results in a larger pressure drop. Both effects were stated to result in a larger predicted void size at downstream locations such as at the entrance to a pump. NAI did not identify that this assumption may also lead to a non-conservatism since it may preclude gas accumulating at a location where a subsequent flow rate perturbation may result in an unpredicted downstream gas slug or it may cause an incorrect gas split to be predicted at a tee.

• Gas was assumed to be located as an initial slug or compact configuration unless clearly identified to be in a different arrangement. This was intended to maximize downstream void fractions.

• Gas located upstream of an isolation valve was introduced at a different rate and with a different downstream profile than if located in open piping downstream of an isolation valve. Generally, maximum system flow rate would be established before gas was introduced. Gas was often introduced in about 0.25 seconds to bound valve stroke and pump start times. Again, this was intended to maximize downstream void fractions.

• Initial temperature and pressure was assumed to maximize void volume change during transport to a pump suction. This is important for system transients in which temperature or pressure changes due to re-alignment to a different suction source. In a static system, setting initial pressure to the low end of the operating range was claimed to maximize the impact of system pressure drop on downstream void volume.

• Friction factors and losses through configurations such as downward turning elbows, reducers, and tees may be biased to maximize pressure reduction and to predict a larger void size at downstream locations such as at the entrance to a pump. This assumption also tended to lengthen the void that would result in a smaller predicted void fraction.

The void volume increase was generally the larger effect which resulted in an increased predicted void fraction.

• Modeling elbows must be a function of the elbow configuration. Assuming a lumped parameter elbow allows assumption of realistic loss coefficients whereas assuming intrinsic (three-dimensional) elbows can cause a significant pressure drop. Lumped parameter elbow modeling can over-predict void transport with a corresponding under-prediction of void fraction near the top of vertical elbows and tees and may not accurately predict the tendency for an elbow to concentrate void on the inside of the elbow. Similarly, modeling must consider the presence of downstream components such as a tee where the modeling can influence the void fraction seen by the side entrance. In these cases, three-dimensional modeling may be best. In general, NAI uses a three-dimensional model for the first downward turning elbow following a horizontal pipe that contains air followed by lumped parameter modeling for additional elbows.

3.5 Conclusions Regarding Use of GOTHIC to Predict Piping System Behavior Available guidance that applies to use of computer codes includes the following:

Page 75

(1) (NRC, March, 19, 2013) - With respect to application of computer codes, the TR (NEI, October, 2012) states that any computer code used to develop a system specific model should be verified to be applicable to solve problems involving gas transport in piping systems via comparisons with laboratory test data or other appropriate methods.

Further, a suitable safety factor should be added to predicted results to reasonably ensure the predictions encompass actual behavior.

(2) (NRC, September 26, 2005), (NRC, April 16, 2008), and (NRC, December 7, 2009 (Approved)) state that the objective of determining system operability is to reasonably ensure that subject system operability is achieved and that the determination can be based on analyses, test or partial test, experience, and/or engineering judgment.

Thus, conclusions from this audit of GOTHICs analysis capability to address operability are to be based on use of a suitable safety factor, tests, experience, and engineering judgment.

This audit has addressed GOTHICs capability to predict void behavior in piping systems by examining its configuration modeling and comparing GOTHIC predictions to a wide range of test data. The audit has not examined such details as interactions between sub-volumes or underlying theory. Consequently, the findings are applicable to GOTHIC applications where the test configuration model applies to the application. The findings are not applicable to situations where no test configuration data were available. Stated differently, the findings are applicable where GOTHIC is used as a methodology for test interpolation or reasonable test extrapolation.

The findings are not applicable where GOTHIC has not been verified via comparison to test data.

GOTHIC has been compared to a wide range of test data. Its use here may be considered for two applications, void transport and water hammer:

(1) Void Transport. In general, GOTHIC provides good predictions of behavior timing for two phase, two component (gas and water) conditions in piping systems. Prediction of fluid flow differential pressure and void transport characteristics for these piping system conditions are generally within a factor of two when compared to data. Therefore, GOTHIC is acceptable for determination of system operability for the stated piping system conditions if the GOTHIC model is consistent with test modeling and an effective safety factor of two is applied when using GOTHIC to predict acceptance criteria.

It is not always necessary to quantitatively apply the factor of two to a GOTHIC prediction of void transport behavior if an alternate approach is valid. For example, Tables 1 and 2 provide acceptable pump void criteria that are generally conservative. In some cases it may be possible to credit the conservatism by applying margin discussed in references such as the Pump Roadmap Project (Huffman, August, 2012).

(2) Water Hammer. The test data comparisons support a conclusion that GOTHIC may be used to assess if there is a potential water hammer problem in a piping system and to estimate the magnitude of potential pressure spikes.

Additional information is provided in Section 3.3.4, above.

Page 76

4. ASSESSMENT OF SELECTED ASPECTS OF VOIDS IN ROBINSON SYSTEMS 4.1 The Current Design Basis (CDB)

For most licensees, whether stated or not, the current design basis (CDB) for the subject systems is a water-solid condition. 25 This was reiterated in (NRC, March, 19, 2013), which stated If there is no specified design limit then the design limit is no gas present."

Robinson was designed before the GDC criteria were finalized, but the GDCs that apply to Robinson, as listed in (Robinson2, No date), provide the same coverage. With respect to gas management, (Robinson2, No date) reported that review of various documents found no specific discussion of the voids or water filled condition. Therefore, It is inherently implied that the Safety Injection system needs to be sufficiently full of water in order to perform its intended function as described in the Design Basis and the supporting calculations. It went on to state that Since, currently there are no design basis associated with allowed voids within the respective systems, no periodic monitoring program associated with pipes being sufficiently full of water, there is no direct relationship between the ... GDCs and the gas management implications and No specific design requirements are imposed with respect to gas voiding in the ECCS and Spray Systems. With respect to operability, (Robinson2, No date) also stated that Consequently, all components in the (safety injection) system should be operable and ready to respond to an automatic actuation signal when the RCS is hot and pressurized and before the reactor is made critical.

To provide the above-stated rationale to conclude gas management does not need to be considered with respect to the CDB is incorrect since, for example, the Robinson statement that there is no documented discussion of voids or a water filled condition and the RNC statement that the design limit for this condition is no gas present. The stated systems must be operated consistent with the CDB requirements. Further, Appendix B provides additional applicable requirements.

Section 2.1.2 of (Robinson2, No date) repeats aspects of the CDB. It states Update the Design Basis to reflect that the Safety Injection system needs to be sufficiently full of water in order to perform its safety intended function and Incorporate into the modification process additional guidance associated with making alterations to the Safety Injection system and creating vulnerable locations with respect to voiding or gas intrusion. Section 2.1.3 states Within the Design Basis, ... It is inherently implied that the Safety Injection system needs to be sufficiently full of water in order to perform its intended function as described in the Design Basis and the supporting calculations.

On the basis of reports reviewed for this inspection, Robinsons CDB with respect to voids in the subject systems is a water-solid, no gas condition. This means the subject systems must be water-solid when transitioning from an outage into power operation. Once the transition is complete, in recognition of the possibility that voids will form during operation, such voids are acceptable provided operability is reasonably maintained.

4.2 Consideration of Water Hammer 25 The desired objective during operation is to maintain a water-solid condition but, where this is not practical, an acceptable objective of gas control measures is to limit the gas accumulation volume to a quantity that does not jeopardize system operability.

Page 77

(Robinson2, No date) stated that The discharge acceptance criteria will be based on the limitations associated with water hammers and their induced loads upon the piping and hangers, as well as the introduction of such a void into the RCS. This acceptance criteria will be xxxx volume and or xxxx void fraction. (No criteria were provided.) Section 2.4.3.1.1.2 stated that the maximum void volume would occur at the maximum flow rate for the design basis LOCA. The most likely flow condition will be with all ECCS pumps operating, not the design basis condition with an assumed single failure. Further, it is possible for the maximum void to occur with an intermediate flow rate, a possibility not considered by Robinson.

Robinson considered (FAI/08-78, August, 2008) applicability to hot leg switchover and the reference conclusion that there would be no significant challenges due to water hammer.

Robinson correctly concluded that its configuration differed from that considered in FAI/08-78 and that the FAI/08-78 discussion could not be used directly to eliminate a water hammer concern. Robinsons consideration of water hammer is addressed further in Section 4.4, below 4.3 Use of GOTHIC to Predict Suction Pipe Void Behavior at Robinson Robinson contracted with NAI to perform the GOTHIC calculations. The complexity of GOTHIC modeling requires an expert understanding of the code and its application and selection of NAI satisfied this requirement. NAI-1417-001 (Harville, January 26, 2009) described NAIs suction pipe analyses using GOTHIC version 7.2aWC2(QA). The plant systems required more comprehensive modeling than used for the above-described test verifications. NAI-1417--001 described the initial approach to address the large number of pump combinations as using a simplified model from the RWST with branch connections to each of the pumps as illustrated in Figure 61 where:

• The dimensions are a rough indication of layout but should not be taken literally. For example, RHR Pumps A and B are physically at the same vertical location.

• A dash box containing an s is a subdivided volume.

• A 3-dimensional box in a homogeneous volume.

• A solid box containing a P is a pump control volume - a boundary condition.

• Control volumes are connected by flow paths and 3-dimensional connectors.

Results from the matrix of pump combinations were stated to establish which pump combinations produced the highest voids in each branch line.

The next step was to examine branch lines in more detail to determine maximum gas voids at selected locations that could exist without exceeding the following pump acceptance criteria:

SI Pump CS and RHR Pumps 0.40 Q/QBEP 1.20: 2% continuous 0.40 Q/QBEP 1.20: 2% continuous Q/QBEP < 0.40 or > 1.20: 1% continuous Q/QBEP < 0.40 or > 1.20: 1% continuous 0.70 Q/QBEP 1.20: max of 10% for 5 Most restrictive of the following: 5% for 20 seconds seconds, max 10% for 5 seconds for 0.70 Q/QBEP 1.20, max of 5% for 5 seconds for Q/QBEP < 0.40 or > 1.20:

Q/QBEP < 0.40 or > 1.20: max of 5% for 5 seconds Page 78

Figure 61. Typical ECCS Suction Piping Noding Diagram A block diagram for a more detailed examination is illustrated in Figure 62. Here, each block corresponds to a section of piping that was described by a scaled isometric diagram provided by Z1BR2 Attachment S - NAI-1417-001 (Harville, January 26, 2009) that covered the Robinson SI and CS system suction piping from the RHR heat exchanger to the pumps. 26 26 As noted in (Balakhnin, April 9, 2013), the SH-H-5 entry should be SH-H-5/6 consistent with segments SH-H-5 and SH-H-6 being represented by a combined control volume in the models.

Page 79

Figure 62. GOTHIC Model of Robinson SI and CS Systems Lumped parameter and subdivided control volumes were used. Analysis of bubble transport in pipe sections required a subdivided control volume for each horizontal and vertical pipe segment. Volumes were used to model elbows, tees, and reducers / expanders as well as pump suctions, tanks, and other miscellaneous volumes. 3D connectors were used to connect a subdivided control volume to another subdivided control volume, or to connect a subdivided control volume to a lumped parameter control volume, across the interface between control volumes. Harvill noted that each horizontal to vertical elbow and tee was modeled with a lumped volume to conserve momentum, thus eliminating the intrinsic loss associated with the bend to allow a more realistic handbook loss to be applied to the elbow or tee. A series of flow paths were used to model vertical tees or elbows where a pipe was not purely vertical or horizontal, such as a 45-deg elbow that is not in the horizontal plane, to conserve the elevation change of the piping.

Page 80

Selected characteristics of each block in Figure 62 are provided in the following table:

The corresponding GOTHIC nodalization is shown in Figure 63.

Page 81

Figure 63. ECCS Suction Nodalization The largest void size at SH-H-1 that meets the Tables 1 and 2 pump suction void criteria for operation within the stated Q/Q(BEP) of 2% continuous for all pumps, 10% for 5 seconds for the SI pump, and 5% for 20 seconds for the CS pump was calculated to be 13.1 ft3. The calculated void fractions at the volumes upstream of the pumps are shown in Figure 64 (the references Figure 22) where AV10 corresponds to SI Pump C and AV11 corresponds to CS Pump A. The momentary void fraction spike at 40 seconds is stated to correspond to the time the void is Page 82

Figure 64. Void fractions for 13.1 ft3 High Point Volume introduced and is stated to be an artifact of the modeling. The SI pump peak void fraction at 0.072 is within the 0.10 and 5 second criteria. The CS pump peak void fraction at 0.042, while exceeding 0.02 for 17 seconds, is within the 0.050 and 20 seconds criteria, is the limiting pump for this void, and was considered by NAI to correspond to the limiting criteria. Assuming this is the limiting condition is conservative because the 0.042 peak is less than the 0.050 criterion and the 0.072 peak is within the 0.10 criterion by a significant margin.

The largest void size at SH-H-1 that meets the Tables 1 and 2 pump suction void criteria for operation outside the stated Q/Q(BEP) range is 1% continuous for all pumps, 5% for 5 seconds for the SI pump, and 5% for 20 seconds for the CS pump and was calculated to be 2.9 ft3. The calculated void fractions at the volumes upstream of the pumps are shown in Figure 65 (the references Figure 23). The momentary spike at 40 seconds is again observed and the CS pump is again the deciding factor with a flat peak void fraction of 0.010 that equals the steady state criterion but if averaged over 20 seconds would be significantly less than the 0.050 transient criterion since the peak is at the limit of 0.010. The 2.9 ft3 is too small and is also conservative.

Page 83

Figure 65. Void fractions for 2.9 ft3 High Point Volume The SH-H-1 high point pipe flow area is 0.5476 ft2 and the flow rate is 550 + 1200 = 1750 gpm 3.899 ft3/sec. Velocity is 3.899 / 0.5476 = 7.120 ft/sec. The corresponding Froude number is 1.37 indicating that a hydraulic jump will occur in the high point pipe that pushes gas rapidly toward the downstream end of the pipe and gas /water will be transported downward in SH-H-2.

The high point will be essentially cleared of gas in 94.73 / 7.120 = 13 seconds. A test for homogeneous flow out of SH-H-2 is that the vertical pipe volume must be at least four times the initial gas volume. In this case, the test is 6.50 / 13.1 = 0.5 < 4. The behavior based on these simplified criteria is that the upper part of SH-H-2 may retain some gas but it is unlikely there will be much permanent accumulation at the high Froude number. Flow exiting SH-H-2 is unlikely to be homogeneous where it enters a 14 inch pipe. Froude number in the 14 inch pipe will be 0.652 although using Froude number to characterize flow in the vicinity of the 10 inch to 14 inch tee will be misleading since the high flow rate from the 10 inch pipe will cause significant turbulence in the 14 inch pipe. The model did not include the effect of the stagnant water in the upstream part of the 14 inch pipe; a conservatism since some gas could move into that piping.

Downstream of the 10 inch to 14 inch tee, the water / gas mixture will expand into a 16 inch pipe (Froude number = 0.464) before exiting into the 12 inch pipe that functions as a header to provide connections to the three SI pumps and the two CS pumps. Again, the 16 inch piping downstream of the 16 to 12 inch tee was neglected and may collect some gas that would not move toward the pumps, an additional conservatism. The result is that significant mixing is Page 84

expected prior to fluid entering the 12 inch header. Neglecting the volume of gas and delays due to mixing at tees, gas is expected to enter the header in about 11.88 / 7.12 + 12.75 / 3.923

+ 2.75 / 2.976 = 6 seconds after initiation of the calculation. It will reach the CS pump in about an additional 4.917 / 5.017 + 23.58 / 3.440 + 12.42 / 7.067 = 9 seconds or 15 seconds after initiation of the calculation. The GOTHIC calculation was about 20 seconds for gas to reach the CS pump, consistent with the 15 seconds estimate. Further, the GOTHIC estimate of gas reaching the SI pump 8 seconds before reaching the CS pump is also consistent with the estimated 9 seconds. GOTHIC is consistent with the independent estimates.

Figure 66 provides predicted void fraction in a 12 inch vertical pipe with an initial horizontal pipe Figure 66. Void in 12 Inch Vertical Pipe void fraction of 0.20. This provides some insight into the predicted 11.88 ft long 10 inch pipe with an initial void fraction of 0.24 and a similar Froude number. It shows that the top of the pipe is essentially voided and that the 10 ft location, roughly corresponding to the bottom of Robinson 10 inch downcomer pipe, reaches a 0.4 void fraction. Calculation of the 8 inch pipe shows identical results. The mixture exiting the vertical 10 inch pipe is calculated to have a high void fraction consistent with the above hand calculation estimates. GOTHIC predicts a significant void fraction reduction by the time gas reaches the pumps, consistent with Page 85

expectations. This would not have been the case if the pumps were located immediately downstream of the 10 inch downcomer.

In conclusion, GOTHIC prediction of gas transport from the horizontal pipe to the SI and CS pumps is acceptable.

As discussed above, NAI calculated an acceptable void size in the 94.73 ft long high point pipe that met the NRC acceptance criteria of 13.1 ft3 for 40% Q/Q(BEP) 120% based on GOTHICs prediction without considering the safety factor of two that was also discussed above.

NAI also calculated 2.9 ft3 for Q/Q(BEP) < 40% or > 120% without the safety factor. The next comparison is to examine GOTHICs prediction with respect to a safety factor of two. The status as developed above is summarized in the following table:

CS Pump Void Fraction Acceptance Criteria NRC (See Tables 1 and 2) Pump Roadmap (Huffman, August, 2012)

Q/Q(BEP), % No Pump No Significant No Pump Damage Expected Void Damage Degradation Capability 40 - 120, Steady 2% 2% 4% 4%

state (> 20 sec)

< 40 or > 120, 1% 1%

Steady state Transient ( 20 sec) 5% - 8% 8%

The principal focus is on transients although the steady state criteria apply if the transient lasts longer than 20 seconds.

The following apply to consideration of conservatism:

(1) Head degradation should be insignificant for the small void fractions addressed here and, since the CS pumps will never see a large back pressure, a momentary head reduction is unlikely to be of concern. Although the expectation is that any effects will be within regulatory requirements, this should be addressed in plant safety system analyses. With this background, degradation is assumed negligible here and the No Pump Damage and Expected Void Capability criteria are applicable to consideration of the effect on safety factor. Crediting the less conservative Pump Roadmap criteria and assuming use of the less conservative criteria would result in the same change in acceptable void volume provides a factor of two to four for steady state operation and 8 /

5 = 1.6 for transients.

(2) The calculated allowable void volumes result in void fractions and time durations that are well within the acceptable void fractions. This provides an additional un-quantified contribution to the safety factor.

(3) Modeling conservatism, such as neglecting gas storage near the 14 inch and 16 inch tees, contributes an additional un-quantified contribution.

(4) GOTHIC predictions, based on test data, are reasonable estimates of expected behavior. The factor of two safety factor is intended to bracket most test data.

Page 86

Therefore, consistent with the approach to operability determination discussed in Section 2.1, above, the conclusion is that the requirement for a factor of two safety factor is met.

The GOTHIC model for the RHR suction and miniflow piping is illustrated in Figures 67 and 68.

Figure 67. RHR Model Page 87

Figure 68. GOTHIC RHR Nodalization Corresponding GOTHIC characteristics are provided in the following table:

Page 88

This modeling is consistent with the above SI and CS pump suction pipe modeling that was reviewed in more detail.

The following table was stated to summarize pump suction gas voids located during testing:

Location Void Size, ft3 Acceptance Criteria Void Fraction at Pump Suction RHR-3A. -3B 0.27 2% continuous 0 5% for 20 seconds RHR-10 0.53 2% continuous 0.14 max 5% for 20 seconds SI-6A, -6B, -6C, -6D 1.84 2% continuous Note 1 10% for 20 seconds SI-9A. -9B, --9C, -9D 54.8 2% continuous Note 2 10% for 20 seconds SI-15 0.17 2% continuous Note 3 10% for 20 seconds Notes:

1. The void is bounded by the 13.1 ft3 void that meets the 2% continuous criterion.

2. Although the void exceeds the 13.1 ft3 and 2.9 ft3 limits, the pipe containing the void is isolated and does not affect the SI and CS pumps.

Page 89

3. The void is bounded by the limits identified in Note 2 and is further isolated so that it does not affect the SI and CS pumps.

In conclusion, GOTHIC modeling of pump suction piping is acceptable and predicted results are also acceptable with the qualification that the recommended factor of two has not been addressed by Robinson but, based on the above table and the other identified considerations, the factor of two is judged to have been satisfied.

4.4 Discharge Piping and Water Hammer Typically, water hammer events associated with voids are due to the sudden increase in pressure in pump discharge piping when a system in put in to service. The void collapse associated with flow initiation creates a pressure pulse which propagates through the system.

The pulse exerts a force (water hammer) on the piping at changes in direction and size causing axial and bending stresses. Gas/water water hammers are less severe than steam/water water hammers but nonetheless can cause physical damage and are potential causes of loss of operability.

The Fauske report discussed in Section 3.3.1.3.5, above, provided generalizations that apply to water hammer. These, and several other generalizations, are as follows:

• A pressure transient initiates when water is accelerated into a gas volume. The prediction of pressure for different systems with similar configurations is insensitive to minor perturbations in system design. However, axial force is strongly affected by the system and its supports.

• Peak waterhammer pressure is determined by the pump shutoff head, the flow run-up transient, and the initial gas pressure, volume, and location(s). The principal concern with pressure is the potential to lift relief valves.

• Peak force is determined by the peak pressure and the rate of rise of the waterhammer pressurization, peak forces are a function of both the piping configuration and the initial gas volume, and a swinging check valve can cause subsequent forces, in both axial directions (upstream and downstream), that are larger than the waterhammer induced force.

• System structural properties affect waterhammer force calculations and require structural evaluation.

• For equal initial gas volumes in the Fauske tests, peak pressure increased as initial pressure decreased. Peak pressure appeared to be highest when all of the gas was located at the high point and lower if some of the volume was elsewhere in the test system. This was not the case for waterhammer forces where the force can be significantly greater if gas existed in more than one location. This was apparently due to out-of-phase oscillations of the gas volumes. In one test, the difference was a factor of seven.

• Within the range of initial Fauske test pressures from -20 to -5 inches Hg, the data exhibited an increase in waterhammer pressure followed by a decrease as initial gas volume was increased.

Page 90

NAI-1417-002 Rev. 2 (Harvill R. , April 2, 2012) used GOTHIC Version 7.2a(QA) for investigation of the Fauske tests discussed in Section 3.3.1.3.5, above. The GOTHIC model is illustrated in Figure 69:

Figure 69. GOTHIC Model of FAI Test with No Check Valve The Fauske test included a long pipe run that consisted of several essentially horizontal pipe runs that included elbows. Figure 69 illustrates water hammer modeling without including all of the elbows. GOTHIC and test data pressure comparisons were discussed in Section 3.3.1.3.5, above. An axial load calculation result is shown in Figure 70 and the comparable test data are shown by the solid line in Figure 71. GOTHIC over-predicted the initial axial load spike by about a factor of two and roughly predicted the timing. Harvill questioned the accuracy of the measured load in the test and believed GOTHIC was closer to the actual load. In the case of a check valve, GOTHIC matched the frequency but significantly over-predicted the load when compared to the data. Again, Harvill questioned the accuracy of the measured load and suspected the full axial force was not measured due to the short pressure pulse and the resulting pipe movement being limited by the inertia of the water filled pipe.

Page 91

Figure 70. GOTHIC Load Prediction of Axial Load Figure 71. Fauske Test Data Corresponding to Figure 70 (Solid Line)

Page 92

(Harvill R. , April 2, 2012) also used GOTHIC Version 7.2a(QA) to calculate generic discharge behavior where an initial void was generally modeled as a long thin ribbon since that resulted in a higher bubble pressure than a gas slug model. Figure 72 shows a typical schematic of a GOTHIC model for the ECCS discharge piping system.

Figure 72. Typical GOTHIC Noding for ECCS Discharge Piping Harvill reported that a series of models has been created to envelope Containment Spray, Residual Heat Removal, and High Head Safety Injection (including charging) systems at many plants. Subdivided control volumes were used for analysis of bubble transport and to model water hammer pressure. Lumped volumes were usually used for pump discharge volumes, reducers / expanders, tees, elbows, tanks, and other miscellaneous volumes to avoid over-prediction of pressure drop that would produce a non-conservative result. However, a limited number of 3D connectors were used when a downstream volume was in a different orientation than the upstream volume when reduced conservatism is acceptable and needed. Harvill also discussed other modeling features that are required that will not be detailed here. This illustrates the need for both modeling expertise and applicable experimental data.

Elevation changes and specific isometric piping configurations are usually not modeled for water hammer analyses. In general, gas transport is less critical when assessing water hammer in discharge piping than when assessing pump suction piping. Air compression and its effects are the primary concern.

In calculating axial load, Harvill used a 1.2 factor for conservatism in addition to the claimed conservatism in the model. With respect to validation, Harvill stated Since the models created are generic models, they cannot reasonably be validated to flow tests from each plant.

However, the basic models with no gas voids are run to ensure that the expected pump flows are produced for the curves included in the model.

Harvill provided generic tables for RHR, intermediate head SI, high head SI, containment spray, and check valve effects with the following guidance for using the tables:

Page 93

- Physical parameters such as pump head, flow rates, and pipe sizes should be matched rather than system names.

- Higher pump head is conservative. The table should be selected on the basis of the pump head. Pump flow is not as critical as head, but the modeled flow should be close to or greater than the plant pump flow.

- Lower minimum flow is conservative. Generally, the minimum flow used in the model should be less than the actual plant minimum flow. However, if the stable discharge head generated by the model is equal to or greater than the plant discharge head, the case set may be considered enveloping.

- Smaller pipe internal diameter is conservative. The plant piping should not be smaller than the modeled size(s).

- Longer pipe length is conservative. The further a void is from the pump discharge, the greater the pressure transient.

- Gas voids distributed over a long section of pipe generally produce a more severe pressure transient than voids that cover the entire cross-section of the pipe

- Higher pump suction pressure is conservative.

- Higher initial static pressure is conservative. However, if the volume of the void has been converted to standard conditions, then a lower initial pressure is conservative.

- System forward flow or feed into tanks greatly reduces water hammer. Lower forward flows are conservative.

The approach was to apply generic cases to establish test acceptance criteria for void size and then to use specific cases for evaluation where voids had been located. Acceptance criteria were limited to pressure. Axial loads must be evaluated on a case-by-case basis because the effect on structure such as hangers must be assessed outside the scope of GOTHIC calculations.

Harvill provided a number of analyses of Robinson configurations. The inspection audit examined the case of an RHR pump start with 13 ft3 of air in the RHR combined header. The analysis was keyed to not opening the RHR relief valve which had a setting between 587.9 psig and 624.1 psig. The assumed acceptance criterion was a peak pressure of 587.9 + 14.7 =

602.6 psia. The discharge piping modeling is diagrammed in Figure 73 and the calculated pressure and axial load are shown in Figures 74 and 75, respectively.

Page 94

Figure 73. RHR Discharge Piping Page 95

Figure 74. RHR Pump Start Discharge Pressure Response Figure 75. RHR Pump Start Axial Load Inertia effects, if any, that may delay relief valve opening associated with the short time near the elevated pressure were not addressed. In this sense, the calculation is conservative. Harvill addressed check valve behavior as follows:

Page 96

Due to the sensitivity of the results to the check valve characteristics, it is not possible to generalize all of these conclusions for all piping system. For the case with the check valve installed upstream of the pump, the calculated system performance is very sensitive to the valve parameters, and most sensitive to the effective inertia of the valve flapper which includes the added mass of water that must be moved as the flapper moves. For systems with check valves on the suction side of the pump only, the conservative approach would be to assume a check valve slam with the load consequences as outlined below.

If a check valve slam should occur, the GOTHIC predicted loads on the piping segments between the valve and the bubble may be as large as those given the by Joukowski formula, i.e.,

F = ADP = Apuc (4) with a duration of tload = L/c (5) where A is the pipe area, u is the velocity of the water when the valve finally closes, is the water density, c is the speed of sound, and L is the length of the pipe segment. The velocity of the water when the valve closes can be conservatively taken as the maximum water velocity observed during the bubble compression. Alternatively, a lower closing velocity may be used if it can be supported by information from the valve vendor or from valve tests.

The NRC inspector agrees with Harvills approach.

5. CONCLUSIONS On the basis of reports reviewed for this inspection, Robinsons current design basis with respect to voids in the subject systems is a water-solid, no gas condition. This means the subject systems must be water-solid when transitioning from an outage into power operation.

Once the transition is complete, in recognition of the possibility that voids will form during operation, such voids are acceptable provided operability is reasonably maintained. Robinson used GOTHIC to support operability assessments.

GOTHIC is a multi-dimensional, multi-component computer code with the capability to model two phase flow in nuclear power plant systems. Robinson contracted with NAI to apply GOTHIC to predict behavior associated with air trapped in the ECCS and RHR suction and discharge pipes at the H. B. Robinson plant. GOTHIC was compared to both a broad range of test conditions and to tests that directly simulated aspects of behavior that may occur in plant piping. With predictive modeling results demonstrated, it was then used to predict behavior if gas were trapped in Robinson system piping. The demonstrated complexities in the correct use of GOTHIC means that the modeler must be highly qualified to ensure system modeling is consistent with test modeling and to acceptably capture nuances that are unique to the system model. NAI meets this requirement.

Page 97

As work progressed, different GOTHIC versions were used for the calculations to address GOTHIC shortcomings. In some cases there were significant differences in the results. The need for continuing changes at the time of this NAI work (2008 - 2010) reflects the immaturity of GOTHIC for study of the problems of concern here. Differences between versions and with respect to test data must be considered when using GOTHIC predictions.

This audit addressed GOTHICs capability to predict void behavior in piping systems by examining its configuration modeling and comparing GOTHIC predictions to a wide range of test data. The audit has not examined such details as interactions between sub-volumes or underlying theory. Consequently, the findings are applicable to GOTHIC applications where the test configuration model applies to the application. The findings are not applicable to situations where no test configuration data were available. Stated differently, the findings are applicable where GOTHIC is used as a methodology for test interpolation or reasonable test extrapolation.

The findings are not applicable where GOTHIC has not been verified via comparison to test data.

Test comparisons that are accomplished for each GOTHIC version include frictional pressure loss, filling of horizontal and vertical pipes, water holdup and pressure drop in vertical pipes with upward and downward flow, water hammer, drop behavior including entrainment, many containment configuration and test conditions, and the Edwards pipe blowdown experiment that is somewhat unique because of the extremely rapid test transient. Application of GOTHIC to the Edwards pipe blowdown, two phase behavior, and water hammer tests has been reviewed in detail as part of the audit of code capability; the first because of the short-term challenges, the second to cover some behavior that lends insight into Robinson use of GOTHIC to address pipe void movement, and the last because GOTHIC was used at Robinson to evaluate water hammer. The audit has further covered tests that are representative of nuclear power plant system suction piping. These tests were the Purdue University tests and another that covered a Millstone operability issue.

Test data applicable to verification of GOTHIC cover many phenomena but are incomplete.

Topics where better understanding is needed for the transient conditions covered herein include elbows, tees, sloped pipes, vortices, transitions between downcomers and downstream horizontal pipes, behavior of kinematic shock in the vicinity of flow obstructions, pump entrance phenomena / piping entrance configuration behavior, and some conditions such as horizontal flow stratification that could lead to a build-up and surge in downstream gas flux. Further, it is difficult to correctly model some areas in GOTHIC such as elbows, tees, sloped pipes, and vortexes. These weaknesses have been considered in the audit in recognition of the need for a method that will provide assessment of operability consistent with the guidance provided in Section 2.1, above.

GOTHIC calculates average properties in a computational cell in contrast to data that represent point or thin plane locations. Consequently, GOTHIC results tend to be smoother than some data and failure to describe individual data point outliers, particularly for many of the transients of concern here, is not a concern when analyzing nuclear power plant systems.

Given the status of GOTHIC verification and capability, a key aspect of its acceptability in determining system operability is established NRC guidance to reasonably ensure that subject system operability is achieved and a reasonable expectation test applies. This means that a high degree of confidence applies but absolute assurance is not necessary. Thus, conclusions from this audit of GOTHICs analysis capability to address operability are to be based on use of Page 98

a suitable safety factor, tests, experience, and engineering judgment. Key aspects of acceptability of GOTHIC predictions are the depth of experimental test comparisons and the conservatism incorporated into the assessment process.

GOTHICs use may be addressed generally for two applications, void transport and water hammer.

5.1 Void Transport GOTHIC provides good predictions of behavior timing for two phase, two component (gas and water) conditions in piping systems. GOTHIC predictions compared to test data ranged from excellent agreement where the predictions overlaid the data or analytic calculations to a factor of two difference. Void fractions were usually within a factor of two and event timing was often within a few percent or better. Some configurations required careful consideration such as elbows, tees, reducers, and sloped pipes. With respect to elbows, representing an elbow as a lumped volume treated it as a homogeneous configuration that would accurately predict pressure drop of smooth curves with tuning but would fail to capture the gas distribution within and immediately downstream of the elbow. Modeling the gas distribution required a multi-dimensional model that would over-predict pressure drop.

The Purdue void fraction data are for individual locations where individual bubbles will introduce noise and short term spike data may be higher in contrast to the GOTHIC predictions that are often an average for a two foot pipe length. A good cross-check of the GOTHIC predictions where GOTHIC appears to under predict void fraction is integration of void fraction over time assuming homogenous flow to show that the predicted gas volume is consistent with the initial upper horizontal pipe volume. This was stated for some cases. Failure of GOTHIC to predict holdup would generally correspond to GOTHIC predicting gas to reach a pump at a higher void fraction than if gas were predicted to be held up. The exception would be if held-up gas were to move downstream at a high void fraction after being held up and accumulating.

The use of the lumped parameter approach for modeling elbows used for the Purdue predictions is not consistent with behavior near the top of the downcomer where a two dimensional flow treatment is necessary and would show gas holdup that a lumped parameter approach may miss. This could be important where plant vertical piping lengths are shorter than in the Purdue tests.

NAIs Millstone investigation provided acceptable results with most void predictions being greater than the test information.

It is not always necessary to quantitatively apply the factor of two to a GOTHIC prediction of void transport behavior if an alternate approach is valid. For example, Tables 1 and 2 provide acceptable pump void criteria that are generally conservative. In some cases it may be possible to credit the conservatism by applying margin discussed in references such as the Pump Roadmap Project (Huffman, August, 2012). This was the case, for example, in an in-depth audit of GOTHICs coverage of trapped gas in an elevated 95 foot long 10 inch diameter pipe.

The GOTHIC calculation was found to have acceptably determined gas volumes that would not jeopardize operability. The findings regarding modeling detail support a conclusion that the licensees use of GOTHIC to evaluate the potential movement of trapped gas with respect to operability is acceptable.

Page 99

Therefore, GOTHIC is acceptable for determination of system operability for the stated piping system conditions if the GOTHIC model is consistent with test modeling and an effective safety factor of two is applied when using GOTHIC to predict acceptance criteria.

5.2 Water Hammer In water hammer tests, NAI demonstrated that GOTHIC could calculate wave velocities in pipe material within 3 % and was within about 20% in air / water mixtures. It provided an excellent fit to pulse timing, magnitude, and width and to predicted water velocity. With respect to water hammer, NAI concluded that GOTHIC is capable of determining: 1) if there is a potential water hammer problem in a piping system, 2) of estimating the magnitude of potential pressure spikes, including the effects of non-condensing gases in the system, and 3) of evaluating the effectiveness of mitigating modifications made to the piping system. However, the GOTHIC analyses of the FAI pressure surge tests discussed in Section 3.3.1.3.5 only covered tests in which the initial gas was in one location. Distributed gas can cause multiple peaks that could be of greater magnitude. The ability of GOTHIC to assess this condition was not provided.

Robinson considered (FAI/08-78, August, 2008) applicability to hot leg switchover and the reference conclusion that there would be no significant challenges due to water hammer.

Robinson correctly concluded that its configuration differed from that considered in FAI/08-78 and that the FAI/08-78 discussion could not be used directly to eliminate a water hammer concern.

(Robinson2, No date) stated that The discharge acceptance criteria will be based on the limitations associated with water hammers and their induced loads upon the piping and hangers, as well as the introduction of such a void into the RCS. This acceptance criteria will be xxxx volume and or xxxx void fraction. (No criteria were provided.) Section 2.4.3.1.1.2 stated that the maximum void volume would occur at the maximum flow rate for the design basis LOCA. The most likely flow condition will be with all ECCS pumps operating, not the design basis condition with an assumed single failure. Further, it is possible for the maximum void to occur with an intermediate flow rate, a possibility not considered by Robinson.

The test data comparisons support a conclusion that GOTHIC may be used to assess if there is a potential water hammer problem in a piping system and to estimate the magnitude of potential pressure spikes. The magnitude of initial forces is adequately predicted but GOTHIC does not provide pipe / hanger response to the initial force nor does it address any interaction of pipe movement with forces following the initial spike.

5.3 Summary of Inspection Findings The (NRC, March, 19, 2013) stated With respect to application of computer codes, the TR (NEI, October, 2012) states that any computer code used to develop a system specific model should be verified to be applicable to solve problems involving gas transport in piping systems via comparisons with laboratory test data or other appropriate methods. Further, a suitable safety factor should be added to predicted results to reasonably ensure the predictions encompass actual behavior. Robinson did not include a safety factor in some of its Page 100

evaluations. 27 This inspection evaluated this requirement and determined that sufficient conservatism was included in the GOTHIC predictions to satisfy a factor of two safety factor.

Overall, the demonstrated GOTHIC predictions are reasonable. Consequently, GOTHIC is acceptable for determination of system operability provided (1) modeling is consistent with the test modeling, (2) an effective factor of two is used for void predictions, and (3) predicted water hammer results are considered to be semi-quantitative.

27 (NEI, October, 2012) does not provide regulatory requirements. The NRC staff requires an appropriate safety factor be applied because of the observed scatter in comparisons to experimental data and differences between tests and plant systems. Also note that (NEI, October, 2012) was published after the Robinson analyses were completed although the NRC staff has insisted on a safety factor in other inspections and the need for a safety factor was identified in earlier documents.

Page 101

6. REFERENCES 09-10, N. (December 2010.). Topical Report, NEI 09-10, Revision 1, Guidelines for Effective Prevention and Management of System Gas Accumulation, ML110240126.

2008-01, G. (January 11, 2008.). Managing Gas Accumulation in Emergency Core Cooling, Decay Heat Removal, and Containment Spray Systems, NRC Generic Letter 2008-01, ML072910759.

Andreychek, T. S. (July 1988). "Loss of RHRS Cooling While the RCS is Partially Filled," Westinghouse Electric Corporation, WCAP-11916 Revision 0, ADAMS Accession No. ML12006A164. .

Andreychek, T. S. (July, 1988.). "Loss of RHRS Cooling While the RCS is Partially Filled," Westinghouse Electric Corporation, WCAP-11916 Revision 0, ML12006A164.

ANS. ( February 19, 1976.). Administrative Controls and Quality Assurance for the Operational Phase of Nuclear Power Plants, American National Standards Institute (ANSI) N18.7-1976/American Nuclear Society (ANS) 3.2.

ANS. (February 19, 1976). Administrative Controls and Quality Assurance for the Operational Phase of Nuclear Power Plants, ANSI N18.7-1976/ANS 3.2.

Balakhnin. (April 9, 2013). No Title, response to RAI #3.

Balakhnin. (April 9, 2013). No title, Response to RAI 6.

Balakhnin. (April 9, 2013). None, Response to RAI Number 2.

Balakhnin. (April 9, 2013). None, Response to RAI Number 4.

Beaulieu, D. P. (October 6, 2010). "Forthcoming Category 2 Public Meeting with Numerical Applications Inc (NAI) ...," ML102780194.

Browning, R. A. (August 30, 2012.). "Response to NRC Request for Additional Information Regarding TSTF-523, Revision 1, 1Generic Letter 2008-01, Managing Gas Accumulation,'" ML122343A284.

Browning, R. A. (August 30, 2012). "Response to NRC Request for Additional Information Regarding TSTF-523, Revision 1, Generic Letter 2008-01, Managing Gas Accumulation," ADAMS Accession No. ML122343A284.

BWROG-TP-08-017. (April 30, 2009.). BWROG-TP-08-017, 0000-0086-7825-R0, Potential Effects of Gas Accumulation on ECCS Analysis as Part of GL 2008-01 Resolution, Proprietary ML091250362, non-proprietary ML091250361.

BWROG-TP-08-020. (April 30, 2009.). BWROG-TP-08-020, 0000-0088-8669-R0, Effects of Voiding on ECCS Drywell Injection Piping, ML091250178. .

Carlson, K. e. (June, 1990). RELAP5/Mod3 Code Manual, NUREG/CR-5535.

Chang, K. S. (1995). "An Experimental Investigation of the Air Entrainment in the Shutdown Cooling System During Mid-Loop Operation," Ann. Nucl. Energy Vol. 22, No. 9, pp 611-619. .

Chang, K. S. (1995). "An Experimental Investigation of the Air Entrainment in the Shutdown Cooling System During Mid-Loop Operation," Ann. Nucl. Energy Vol. 22, No. 9, pp 611-619.

Collier, J. a. (1994). "Convective Boiling and Condensation," Third Edition, Oxford University Press.

Crane. (December, 2001 (Reprinted) ). "Flow of Fluids Through Valves, Fittings, and Pipe," Technical Paper No. 410.

Page 102

FAI/08-70, R. (September 2008.). Gas-Voids Pressure Pulsations Program, Fauske & Associates, LLC, for the PWR Owners Group, ML090990426.

FAI/08-78. (August, 2008). Methodology for Evaluating Waterhammer in the Containment Spray Header and Hot Leg Switchover Piping, Fauske & Associates, LLC, for the PWR Owners Group, ADAMS Accession No. ML090980331. .

FAI/08-78, R. (August 2008.). Methodology for Evaluating Waterhammer in the Containment Spray Header and Hot Leg Switchover Piping, Fauske & Associates, LLC, for the PWR Owners Group, ML090980331.

FAI/09-130-P. (December, 2010.). "Technical Basis for Gas Transport to the Pump Suction," Fauske &

Associates, LLC, "Technical Basis for Gas Transport to the Pump Suction," Fauske &Associates, ML110480456.

Gall, J. (June 24, 2010.). Meeting With The Nuclear Energy Institute (NEI) And Industry Representatives To Discuss NRC Generic Letter 2008-01, Managing Gas Accumulation In Emergency Core Cooling, Decay Heat Removal, And Containment Spray Systems, NRC Memorandum, ML101650201.

George, T. L. (January 10, 2012). "Comparison of GOTHIC Gas Transport with 4" and 12" Pipe Test Data," Numerical Applications Ind., NAI-1459-003, Revision 1.

George, T. L. (August 25, 2010). "GOTHIC Simulation of the Millstone-3 Gas-Water Transport Tests,"

NAI-1459-002.

Harvill, R. (April 2, 2012). Evaluation of Gas Accumulation in RNP (Unit 2) ECCS Discharge Piping, NAI-1417-002, Rev. 2.

Harvill, R. e. (No date). "Evaluation of Gas Accumulation in Harris Nuclear Plant ECCS Suction Piping," NAI-1614-001 Revision 0 (Unsigned).

Harville, R. (January 26, 2009). "Robinson Nuclear Plant ECCS Pump Suction Void Evaluation," NAI-1417-001 Revision 0.

Honcharik, M. C. (June 13, 2012). "Acceptance for Review and Request for Additional Information....,"

NRC Letter to TSTF, ADAMS Accession No. ML12158A547.

Honcharik, M. C. (June 13, 2012.). "Acceptance for Review and Request for Additional Information....,"

NRC Letter to TSTF, ML12158A547.

Huffman, K. (August, 2012). Report of the Expert Panel on the Effect of Gas Accumulation on Pumps,"

EPRI Report Number 1026498.

INPO. (January 9, 2008.). Gas Intrusion in Safety Systems, INPO Significant Event Report 2-05 Rev 1, (Not publically available.).

Institute, N. E. (October, 2012). "Guidelines for Effective Prevention and Management of System Gas Accumulation," ML12310A010.

Joukowsky, N. (1898.). Uber den hydraulischen Stoss in Wasserleitungsrohen, Memoires de IAcademie Imperiale des Sciences de St.-Petersbourg (1900), Series 8, 9(5), 1-71.

Kamath, P. S. (September, 1982.). "An Assessment of Residual Heat Removal and Containment Spray Pump Performance Under Air and Debris Ingesting Conditions," Creare, Inc., NUREG/CR-2792, ML100110155.

Kasztejna, P. (August 5, 2008). "HB Robinson Power Plant, ECCS Pumps," Letter from Flowserve Engineering.

Page 103

Kim, S. (September 1992). "A Study on the Free Surface Vortex in the Pipe System," Journal of the Korean Nuclear Society, Volume 24, number 3. (pdf may be obtained by Google search using the following: "sang-nyung kim" "wan-ho jang. .

Kim, S. (September 1992). "A Study on the Free Surface Vortex in the Pipe System," Journal of the Korean Nuclear Society, Volume 24, number 3. (pdf may be obtained by Google search using the following: "sang-nyung kim" "wan-ho jang.

LTR-LIS-08-543. (April 2, 2009). PWROG Position Paper on Non-condensable Gas Voids in ECCS Piping; Qualitative Engineering Judgment of Potential Effects on Reactor Coolant System Transients Including Chapter 15 Events, Task 3 of PA-SEE-450, Westinghouse, ML090980303.

LTR-LIS-08-543. (April 2, 2009.). PWROG Position Paper on Non-condensable Gas Voids in ECCS Piping; Qualitative Engineering Judgment of Potential Effects on Reactor Coolant System Transients Including Chapter 15 Events, Task 3 of PA-SEE-450, Westinghouse, ML090980303.

Lyon, W. (April 9, 2013). "Office of Nuclear Reactor (NRR) Reactor Systems Branch (SRXB) Questions to Obtain Additional Information Related to Use of GOTHIC at H. B. Robinson".

McGoun, W. (No date). "Evaluate use of Gothic for Gas Transport Analyses (IN 2011-17)," PCH-EVAL Engineering Change 0000089423R0.

McGoun1, W. (No Date). "Evaluate use of Gothic for Gas Transport Analyses (IN 2011-17)," PCHG-EVAL Engineering Change 0000086423R0.

Musser, R. A. (February 6, 2013). "H.B. Robinson Steam Electric Plant - NRC Integrated Inspection Report 05000261/2012005, Letter to Carolina Power and Light Company.

NEI. (December 2010). Guidelines for Effective Prevention and Management of System Gas Accumulation, NEI 09-10 Rev 1, ML110240126.

NEI. (October, 2012). Guidelines for Effective Prevention and Management of System Gas Accumulation, NEI 09-10 Rev 1a, ML12310A009, 010.

Nesse, R. (December 4, 2012). "assistance with GOTHIC at Robinson," NRC email, ML13099A001.

NRC. (June 26, 2011). "Calculation Methodologies for Operability Determinatinos of Gas Voids in Nuclear Power Plant Piping," Information Notice 2011-17, ML11161A111.

NRC. (March, 19, 2013). "Final Safety Evaluation for Nuclear Energy Institute Topical Report NEI 09-10, Revision 1a, 'Guidelines for Effective Prevention and Management of System Gas Accumulation, Project No. 689," ML12342A368.

NRC. (March 12 - 14, 2013.). "NRC's 25th Annual Information Conference, North Bethesda, Md.

NRC. (September 26, 2005). NRC Regulatory Issue Summary 2005-20 ... 'Information to Licensees Regarding Two NRC Inspection Manual Sections on Resolution of Degraded and Nonconforming Conditions and on Operability," ML052020424.

NRC. (April 16, 2008). NRC Regulatory Issue Summary 2005-20, Rev. 1 ... Operability Determinations

& Functionality Assessments for Resolution of Degraded or Nonconforming conditions Adverse to Quality or Safety, ML073440103.

NRC. (December 7, 2009 (Approved)). Operability Determinations & Functionality Assessments for Resolution of Degraded or Nonconforming conditions Adverse to Quality or Safety, NRC Inspection Manual, Part 9900: Technical Guidance.

NRC. (February 1978). RG-1.33, Quality Assurance Requirements (Operation), Regulatory Guide (RG) 1.33, Rev 2.

Page 104

NRC/RES. (October 4, 2010.). "Generic Issue Management Control System Report for Fiscal Year 2010 4th Quarter," Office of Nuclear Regulatory Research, ML102710133.

NRC-Guidance. ( May 23, 2011.). Guidance To NRC/NRR/DSS/SRXB Reviewers For Writing TI Suggestions For The Region Inspections, ML102080675, June 4, 2010. The latest version of this guidance is Revision 11, ML111660749.

NRC-Guidance-1. (June 7, 2010.). Guidance To NRC/NRR/DSS/SRXB Reviewers For Writing TI Suggestions For The Region Inspections, ML101590268.

NUREG-0737. (November, 1980.). Clarification of TMI Action Plan Requirements .

Part-9900. (April 16, 2008.). Operability Determinations & Functionality Assessments for Resolution of Degraded or Nonconforming conditions Adverse to Quality or Safety, NRC Inspection Manual, Part 9900: Technical Guidance, Attachment to RIS 2005-20 Rev 1. .

Rahn, F. (January, 2012). "GOTHIC Thermal Hydraulic Package, Version 8.0(QA). EPRI, Palo Alto, CA:2012".

RG-1.33. (, February, 1978). Quality Assurance Requirements (Operation), Regulatory Guide (RG) 1.33 , Rev 2.

Riley, J. (September 26, 2012.). "NEI Response to Request for Additional Information on Topical Report NEI 09-10, Revision 1: Guidelines for Effective Prevention and Management of Sysstem Gas Accumulation," ML122760199.

Riley, J. H. (January 26, 2012.). " NEI Response to NRC Comments on Topical Report NEI 09-10, Revision 1: 'Guidelines for Effective Prevention and Management of System Gas Accumulation,'" Nuclear Energy Institute Letter to NRC, ML120270480.

Riley, J. H. (December 21, 2010). "Submittal of NEI 09-10, 'Guidelines for Effective Prevention and Management of System Gas Accumulation," .

Riley, J. H. (November 1, 2012). "Submittal of NEI 09-10, Rev 1a for NRC Endorsement,"

ML12310A009.

RIS-2005-20. (September 26, 2005.). ... Revision to Guidance Formerly Contained in NRC Generic Letter 91-18, Information to Licensees Regarding Two NRC Inspection Manual Sections on Resolution of Degraded and Nonconforming Conditions and on Operability, ML052020424.

RIS-2005-20-Rev-1. (April 16, 2008.). ... Revision to Guidance Formerly Contained in NRC Generic Letter 91-18, Information to Licensees Regarding Two NRC Inspection Manual Sections on Resolution of Degraded and Nonconforming Conditions and on Operability, ML073440103.

Robinson1. (No Date). "Engineering Change," PCHG-DESG, 0000070450R7.

Robinson2. (No date). Engineering Change," PCHG-DESG, 0000070450R7, Attachment A.

Stringfellow, N. J. (March 29, 2012.). "Transmittal of TSTF-523, Revision 1, 'Generic Letter 2008-01, Managing Gs Accumulation,'".

Stringfellow, N. J. (March 29, 2012). Transmittal of TSTF-523, Revision 1, Generic Letter 2008-01, Managing Gas Accumulation, Letter to NRC, ADAMS Accession No. ML12089A356. .

Stringfellow, N. J. (March 29, 2012.). Transmittal of TSTF-523, Revision 1, Generic Letter 2008-01, Managing Gas Accumulation, Letter to NRC, ML12089A356.

Page 105

TIA2008-03. (October 21, 2008.). Task Interface Agreement - Emergency Core Cooling Systems (ECCS) Voiding Relative to Compliance with Surveillance Requirements (SR) 3.0.1.1, 3.5.2.3, and 3.5.3.1 (TIA 2008-03), ML082560209.

USNRC. (January 11, 2008). 1. Managing Gas Accumulation in Emergency Core Cooling, Decay Heat Removal, and Containment Spray Systems, NRC Generic Letter 2008-01, ML072910759.

Waldrep, G. W. (1990). "Analysis of Vortex Phenomena in PWRs at Reduced Inventory Conditions,"

Transactions of the ANS, Vol 62, pp 350 - 352. .

Wallis, G. (June, 1977). "Conditions for a Pipe to Run Full When Discharging Liquid Into a Space Filled With Gas," Journal of Fluids Engineering, June, pp 405 - 413.

Wallis, G. (1969). "One Dimensional Two-Phase Flow," McGraw Hill Book Co. .

Wallis, G. (1969). "One Dimensional Two-Phase Flow," McGraw Hill Book Co.

WCAP-17271-P. (October, 2010.). Air Water Transport in Large Diameter Piping Systems: Analysis and Evaluation of Large Diameter Testing Performed at Purdue University, Rev. 1, Volumes 1, 2, and 3, Westinghouse Electric Company LLC, WCAP-17271-P, ML110490356.

WCAP-17276-P, R. 1. (January 2011.). Investigation of Simplified Equation for Gas Transport, Westinghouse Electric Company LLC, for the PWR Owners Group, WCAP-17276-P, Rev. 1, ML110480381.

Westinghouse. (October, 2010). Air Water Transport in Large Diameter Piping Systems: Analysis and Evaluation of Large Diameter Testing Performed at Purdue University, Rev. 1, Volumes 1, 2, and 3, Westinghouse Electric Company LLC, WCAP 17271-P, ADAMS Accession No. ML110490356. .

Wiles, L. E. (June 26, 2009). "Comparison of GOTHIC Gas T ransport Calculations with Test Data,"

Numerical Applications, Inc., NAI-1459-001, Rev. 0, PROPRIETARY.

Page 106