NUREG/CR-5720, "Motor-Operated Valve Research Update"

NUREG/CR-5720EGG-2643

Motor-Operated ValveResearch Update

Prepared byR. Steele, Jr., J. C. Watkins, K. G. DeWall, M. J. Russell

Idaho National Engineering LaboratoryEG&G Idaho, Inc.

Prepared forU.S. Nuclear Regulatory Commission

t I

AVAILABILITY NOTICE

Availability of Reference Materials Cited in NRC Publications

Most documents cited In NRC publications will be available from one of the following sources:

1. The NRC Public Document Room, 2120 L Street. NW., Lower Levei, Washington. DC 20555

2. The Superintendent of Documents, U.S. Government Printing Office, P.O. Box 37082, Washington,DC 20013-7082

3. The National Technical Information Service, Springfield, VA 22161

Although the Isting that follows represents the majority of documents cited In NRC publications, it is notIntended to be exhaustive.

Referenced documents available for Inspection and copying for a fee from the NRC Public Document RoomInclude NRC correspondence and Internal NRC memoranda; NRC bulletins. circulars, Information notices,inspection and Investigation notices; licensee event reports; vendor reports and correspondence; Commis-sion papers; and applicant and licensee documents and correspondence.

The following documents in the NUREG series are available for purchase from the GPO Sales Program:formal NRC staff and contractor reports, NRC-sponsored conference proceedings, international agreementreports, grant publications, and NRC booklets and brochures. Also available are regulatory guides, NRCregulations in the Code of Federal Regulations, and Nuclear Regulatory Commission Issuances.

Documents available from the National Technical Information Service Include NUREG-series reports andtechnical reports prepared by other Federal agencies and reports prepared by the Atomic Energy Commis-sion, forerunner agency to the Nuclear Regulatory Commission.

Documents available from public and special technical libraries include all open literature items, such asbooks. journal articles, and transactions. Federal Register notices, Federal and State legislation, and con-gressional reports can usually be obtained from these libraries.

Documents such as theses, dissertations, foreign reports and translations, and non-NRC conference pro-ceedings are available for purchase from the organization sponsoring the publication cited.

Single copies of NRC draft reports are available free, to the extent of supply, upon written request to theOffice of Administration, Distribution and Mail Services Section, U.S. Nuclear Regulatory Commission,Washington, DC 20555.

Copies of industry codes and standards used In a substantive manner in the NRC regulatory process aremaintained at the NRC Library, 7920 Norfolk Avenue, Bethesda, Maryland, for use by the public. Codes andstandards are usually copyrighted and may be purchased from the originating organization or, if they areAmerican National Standards, from the American National Standards Institute. 1430 Broadway. New York,NY 10018.

DISCLAIMER NOTICE

This report was prepared as an account of work sponsored by an agency of the United States Government.Neitherthe United States Government nor any agency thereof, or any of their employees, makes any warranty,expressed or implied, or assumes any legal liability of responsibility for any third party's use, or the results ofsuch use, of any Information, apparatus, product or process disclosed in this report, or represents that its useby such third party would not infringe privately owned rights.

NUREG/CR-5720EGG-2643R1, RM, 1S

Motor-Operated ValveResearch Update

Manuscript Completed: October 1991Date Published: June 1992

Prepared byR. Steele, Jr., J. C. Watkins, K. G. DeWall, M. J. Russell

G. H. Weidenhamer, 0. 0. Rothberg, NRC Project Managers

Idaho National Engineering LaboratoryManaged by the U.S. Department of Energy

EG&G Idaho, Inc.Idaho Falls, ID 83415

Prepared forDivision of EngineeringOffice of Nuclear Regulatory ResearchU.S. Nuclear Regulatory CommissionWashington, DC 20555NRC FINs A6857, B5529Under DOE Contract No. DE-AC07-761D01570

ABSTRACT

This report provides an update on the valve research sponsored by the U.S.Nuclear Regulatory Commission (NRC) that is being conducted at the IdahoNational Engineering Laboratory. The update focuses on the information applicableto the following requests from the NRC staff:

* Examine the use of in situ test results to estimate the response of a valve atdesign-basis conditions

* Examine the methods used by industry to predict required valve stem forces ortorques

* Identify guidelines for satisfactory performance of motor-operated valvediagnostics systems

* Participate in writing a performance standard or guidance document foracceptable design-basis tests.

The authors have reviewed past, current, and ongoing research programs toprovide the information available to address these items.

FIN A6857, B5529-Investigation of information used to estimatevalve response, methods used to predict valve stem forces/torques,

guidelines for MOV diagnostics systems.

iii NUREG/CR-5720

CONTENTS

ABSTRACT..................................................................... iii

LIST OF FIGURES .................. vii

LIST OF TABLES .................. xi

EXECUTIVE SUMMARY .................. xiii

ACKNOWLEDGMENTS .................. xv

1. INTRODUCTION ............................................................ 1

2. GENERAL OBSERVATIONS .................................................. 3

3. SPECIFIC OBSERVATIONS ................................................... 4

3.1 Use of In Situ Test Results to Bound the Response of a Valve atDesign-Basis Conditions ............. ........................ ............ 4

3.2 Assessment of Butterfly Valves Closing against a Compressible Fluid(Containment Purge and Vent) ........... .................................. 4

3.2.1 Background .................................................... 43.2.2 Flow Phenomena through a Butterfly Valve ........................... 53.2.3 Existing Butterfly Valve Data and Extrapolation Techniques .... ......... 73.2.4 Butterfly Valve Dynamic Flow Testing and Results ...... ............... 83.2.5 Butterfly Valve Test Results and Torque Bounding Methods .... .......... 83.2.6 Effect of an Upstream Elbow on the Torque Requirements of a

Butterfly Valve .............................................. 22

3.3 Assessment of Wedge-Gate Valves Closing against Medium to HighFlow Conditions . .............................................. 24

3.3.1 Assessment of the Disc Factor Term in the Industry Equation ..... ........ 303.3.2 Development of a Correlation to Bound the Stem Force on a Gate

Valve during Closure .373.3.3 A Correlation to Bound the Stem Force on a Gate Valve during

Closure .423.3.4 Use of the Correlation to Bound the Stem Force on a Gate Valve

during Closure .473.3.5 Identifying the Pressure Dependency Contributing to the Stem Force

on a Gate Valve during Closure .513.3.6 Low Differential Pressure Test Verification and Bounding of the

Stem Force on a Gate Valve Closing against Design-Basis Flows .513.3.7 Nuclear Maintenance Application Center (NMAC) Gate Valve Stem

Force Equation ......... 54

v NUREG/CR-5720

4. UNDERSTANDING DIAGNOSTICS AND DIAGNOSTIC TESTING OFMOTOR-OPERATED VALVES ................................................. 55

4.1 Overview . ..................................................... 55

4.2 Understanding MOV Diagnostic Testing ........... .......................... 56

4.2.1 Motor Operator Switch Position .......... .......................... 564.2.2 Motor Current ................................................... 564.2.3 Motor Voltage .................................................. 594.2.4 Motor Operator Torque ................ ........................... 624.2.5 Valve Stem Force . .............................................. 62

4.3 Load-Sensitive Motor-Operated Valve Behavior ........ ....................... 67

4.4 Direct Current Powered Motor Operators ......... ........................... 79

5. WRITING A PERFORMANCE STANDARD .. 83

5.1 Overview .83

5.2 Detailed Observations .84

6. REFERENCES .. 85

NUREG/CR-5720 vvi

'LIST OF FIGURES

1. Effect of low pressure zones on butterfly valve torque-disc oriented with the flat faceof the disc facing upstream ........................................................ 6

2. Effect of low pressure zones on butterfly valve torque-disc oriented with the curved faceof the disc facing upstream . ............. ........................... 6

3. Cross section of Valve 1, the first 8-in. butterfly valve. 9

4. Cross section of Valve 2, the second 8-in. butterfly valve, and Valve 3, the24-in. butterfly valve .10

5. Containment butterfly valve disc overlay .11

6. Typical installation-butterfly valve uniform inlet flow test section .11

7. Uniform inlet flow butterfly valve positions ............................... 11.......... I

8. Typical installation-butterfly valve nonuniform inlet flow test section .12

9. Nonuniform inlet flow butterfly valve positions .12

10. Butterfly valve differential pressure to upstream pressure ratio versus valve position .13

11. Static pressure 15 diameters downstream of a butterfly valve versus valve position .... ..... 13

12. Torque versus upstream pressure and angle for Valve 1, the first 8-in. butterfly valve,FFF orientation ............... :: .- 14

13. Torque versus upstream pressure and angle for Valve 2, the second 8-in. butterfly valve,FFF orientation .......... 15

14. Torque versus upstream pressure and angle for Valve 3, the 24-in. butterfly valve,FFF orientation .15

15. Torque versus upstream pressure and angle for Valve 1, the first 8-in. butterfly valve,CFF orientation . ........... ... -i.................. 16

16. Torque versus upstream pressure and angle for Valve 2, the second 8-in. butterflyvalve, CFF orientation. .............. o : 16

17. Torque versus upstream pressure and angle for Valve 3, the 24-in. butterfly valve,CFF orientation .......... ,.17

18. Peak torque versus static upstream pressure for Valve 1, the first 8-in. butterfly valve 17

19. Peak torque versus static upstream pressure for Valve 2, the second 8-in. butterfly valve 18

20. Peak torque versus static upstream pressure for Valve 3, the 24-in. butterfly valve .18

21. Extrapolation exponent, butterfly valve oriented with the curved face of the discfacing upstream ............................. 20

Vil NUREG/CR-5720

22. Extrapolation exponent, butterfly valve oriented with the flat face of the discfacing upstream . .............................................................. 20

23. Actual to predicted butterfly valve torque (percent) as a function of extrapolationexponent versus valve diameter ratio ............................................. 21

24. Predictions for a 48-in. butterfly valve based on extrapolating the torques of an 8-in.and a 24-in. butterfly valve at upstream pressures of 15 and 60 psig .................... 23

25. Torque versus upstream pressure and angle for Valve 3, the 24-in. butterfly valvein the CFF orientation .............. ........................................ ... 25

26. Peak torque versus static upstream pressure for Valve 3, the 24-in. butterfly valve,comparing the response of the peak torque with elbow and peak torque withoutelbow orientations . ............................................................ 25

27. Comparison of the standard industry gate valve stem force equation with selectedtest results .................................................................. 29

28. Comparison of the NMAC gate valve stem force equation with selected test results ........ 30

29. Disc factor for Gate Valve 2 closing on line break flow, effect of subcooling at1000 psig . .................................................................. 32

30. Disc factor for Gate Valve 2 closing on line break flow, effect of pressure at100l F subcooling . ............................................................ 33

31. Disc factor for Gate Valve 2 opening on line break flow, effect of pressure at100WF subcooling . ............................................................ 34

32. Gate valve disc cross section showing pressure forces and measurement locations .... ..... 36

33. Gate valve internal pressure distribution ............................................ 36

34. Gate valve disc cross section showing unbalanced forces just before wedging ...... ....... 37

35. Resolution of horizontal and vertical gate valve disc forces into surface normaland sliding forces . ............................................................ 39

36. Gate-valve stem-force history, indicating severe damage to the guides ................... 40

37. Gate-valve stem-force history, indicating damage to the guides but not to the seats .... ..... 41

38. Normalized surface sliding force versus normalized surface normal force forgate valves closing against less than 70'F subcooled fluid . ............................. 45

39. Normalized surface sliding force versus normalized surface normal force for gate.valves closing against 70'F, or greater, subcooled fluid ............................... 46

40. Stem-force history of a small gate valve during an opening test ........................ 48

41. Stem-force history of a large gate valve during an opening test .......... .. ............. 49

NUREG/CR-5720 viii

42. Gate valve load-dependent friction factor using the INEL correlation .................... 53

43. Limit switch history showing stroke time for a position-controlled valve. Limitswitch actuation points are important when analyzing other data .............. ......... 57

44. Torque switch history showing torque switch trip and the termination of current flowto the motor controller holding coil .57

45. Current history from a design-basis test showing current increasing as the valve closes;the rapid increase in current indicates torque switch trip. The test began with the valve75% open .58

46. Motor torque-speed curve showing the speed-torque-current relationship for the acmotor test results shown in Figure 45. Beyond the knee of the curve, the torque increaseis small in proportion to the speed loss and the current increase. For some motorconfigurations there is no increase; for some there is actually a decrease in torque beyondthe knee of the speed curve.58 ........ '. I

47. Current history from a design-basis test showing the motor going into a stall, saturatingthe current transducer. The test began with the valve 30% open ......................... 59

48. Motor torque-speed curve showing the speed-torque-current relationship for a dc motor.The torque (load) continues to increase as the speed drops and the current increases. .. 60

49. Disc position histories from three dc-powered closing tests showing that the stroke timeis longer with higher loads .60

50. Motor heating under load caused the current demand of this dc motor to decrease at arate of 1 amp per second during the 20-second test, resulting in a decrease in the outputtorque of approximately 1 ft-lb every 2 seconds . .61

51. Disc position histories from four dc-powered MOV closing tests, showing that the stroketime is longer with higher loads. Three of the tests did not fully seat the valve, and the testat the highest load stalled the motor ................................................... 61

52. Ratio of torque spring force to stem torque from MOVLS Test 1, with a smallstem force ........ ;..... 63

53. Ratio of torque spring force to stem torque from MOVLS Test 4 at a higher stem force ..... 63

54. Ratio of torque spring force to stem torque from MOVLS Test 5, at a high stem force, inwhich final seating was not achieved, showing that the average torque spring force tostem torque ratio was not influenced by the load on the operator .64

55. Stem force history from MOVLS Test 4, showing a 3000 lbf margin between the peakstem force and the stem force at final seating .64

56. Stem force'history'from MOVLS Test 5, showing that a 1000 lbf increase in the initialstem force eliminated the 3000 lbf final seating stem force margin observed inMOVLS Test 4 .65

ix NUREG/CR-5720

57. Typical torque spring calibration plot shows the initial force offset at zero deflection.This figure also confirms that for this spring, the spring force to deflection relationship islinear. The actual values are omitted from this figure because the data are proprietaryuntil the MOV diagnostic equipment validation effort is complete ...................... 65

58. MOVLS layout drawing .......................... ............................. 68

59. Traces of thrust versus time for Tests 8 through 10 of the MOVLS test sequence. Theseare identical low-thrust tests .................................................... 69

60. Traces of thrust versus time for Tests 10 through 14 of the MOVLS test sequence ......... 69

61. Traces of thrust versus time for the final three tests of the MOVLS test sequence.Tests 12-14, which exhibit load-sensitive MOV behavior, were followed by a series oflow-thrust tests (Tests 15-17), with resistance to closing much like the first three ofthe series .... 70

62. The design of the operator allows two potential output motion paths ......... ........... 71

63. The application of spring pack force to the operator mechanism at an offset distancefrom the stem centerline generates the torque input to the mechanism ................... 73

64. Potential causes of torque loss in the operator are identified .......... .. ............... 73

65. Traces of spring pack force versus time for Tests 8 through 10 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 59 ......................... 75

66. Traces of spring pack force versus time for Tests 10 through 14 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 60. Spring pack force doesnot change significantly from test to test ..... ...................................... 75

67. Traces of spring pack force versus time for Tests 14 through 17 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 61. Spring pack force doesnot change significantly from test to test .......... I ................................. 76

68. The ratio of spring pack force to stem torque for Tests 8 through 10 of the MOVLStest sequence. These traces correspond to the thrusts of Figure 59 ...................... 76

69. The ratio of spring pack force to stem torque for Tests 10 through 14 of the MOVLStest sequence. These traces correspond to the thrusts of Figure 60 ....................... 77

70. The ratio of spring pack force to stem torque for Tests 14 through 17 of the MOVLStest sequence. These traces correspond to the thrusts of Figure 61. Spring pack forcedoes not change significantly from test to test ................................ 77

71. Traces of coefficient of friction in the stem nut versus time for the first three tests ofthe MOVLS test sequence. Very little change in coefficient of friction is observed atthe low thrust levels of these tests ................................ ' : 78

72. Traces of coefficient of friction in the stem nut versus time for Tests 10 through 14 ofthe MOVLS test sequence. A steady increase in coefficient of friction is evident in thesetests. Coefficient of friction is highest during Test 14, the test exhibiting load sensitiveMOV behavior ............................................................ 78

NUREG/CR-5720 x

73. Traces of coefficient of friction in the stem nut versus time for the last four tests ofthe MOVLS test sequence. A full opening/closing low-thrust cycle (Test 15) is neededfollowing the test exhibiting load-sensitive MOV behavior (Test 14) before thecoefficient of friction returns to the levels recorded for the initial low-thrust level tests ...... 80

74. Typical stem thrust histories from the dc-powered MOVLS tests ....... ................ 80

75. Typical stem nut coefficient of friction histories from the dc-powered MOVLS test.Note the increase in friction between low-load Test 1 and high-load Test 12. The loadwas doubled between Tests 12 and 15 without a significant increase in stem nut coefficientof friction .................................................... 82

LIST OF TABLES

1. Comparison of torque prediction methods butterfly versus valve orientation .22

2. Ratio of peak torque to uniform flow peak torque for a butterfly valve in theCFF orientation .23

3. Phase I gate valve flow interruption test temperatures and pressures ........ ............. 26

4. Phase II gate valve flow interruption test temperatures and pressures ....... ............. 26

5. Phase II gate valve test data supporting extrapolation .................. .............. 43

6. Phase I gate valve test data supporting extrapolation . ................................ 44

7. Comparison of gate valve stem forces, estimates versus actual .......... ............... 52

xi NUREG/CR-5720

EXECUTIVE SUMMARY

The U.S. Nuclear Regulatory Commission(NRC) is supporting motor-operated valve(MOV) research at the Idaho National Engineer-ing Laboratory (INEL). The MOV test programsperformed as part of the research provide thebasis for assessing the effects various factors haveon the valves and for evaluating current industrystandards. This report discusses several researchitems, including

* Use of in situ test results to estimate theresponse of a valve at design basis conditions

* Methods used by industry to predictrequired valve stem forces or torques

* Guidelines for satisfactory performance ofMOV diagnostics systems

* Composition of a performance standard orguidance document for acceptable design-basis tests.

We have reviewed all of our past, current, andongoing valve research that included full-scale

valve test programs to address these items. Thisreview revealed that (a) use of in situ test resultsto estimate the response of gate and butterflyvalves at design-basis conditions is possible, but along list of caveats exists; (b) the methods usedby industry to predict the required stem force fora gate valve and the required stem torque for abutterfly valve are incomplete; (c) satisfactoryperformance of MOV diagnostic systems ispossible, but very few of the currently availablesystems measure enough parameters to be com-pletely effective; and (d) our participation inwriting a performance standard or guidance docu-ment for acceptable design-basis tests requires usto continue to exchange information with theAmerican Society of Mechanical Engineersstandards writing committees.

The research documented in this report formsthe basis for the technical presentations given atthe NRC inspector training courses, and has alsoenabled the INEL to create a PC-based computerprogram to assist NRC personnel during theirMOV inspections.

.x.i. NUREG/CR-5720

ACKNOWLEDGMENTS

Researchers at the Idaho National Engineering Laboratory (INEL) developed thebasis for the INEL gate valve stem force correlation presented in Section 3.3 andhave presented the results to a wide range of industry and utility experts for review.We wish to thank the following reviewers, who responded with comments andinsights to this effort:

Siemens/Kraftwerk Union Dr. Nabil Schauki, Dr. Norbert Rauffmann,Dr. U. Simon, Helmut Knoedler

Duke Power Neal Estep, W. Scheffler

Consumers Power

Portland General Electric

Nuclear Management andResources Council

Electric Power ResearchInstitute

Idaho National EngineeringLaboratory

George Smith

M. Lee Kelly

Review Team

John Hosler

Dr. Romney Duffy, Tim Boucher, Vic Berta.

We would also like to thank the following individuals from Limitorque who havereviewed our research on load-sensitive, motor-operated valve behavior: BobBailey, Pat McQuillan, Mike Bailey, Greg Pence, Charles Cox, and Walt Sykes.Their comments and insights are appreciated.

xv xv ~~~~NUREG/CR-5720

Motor-Operated Valve Research Update

1. INTRODUCTION

The Idaho National Engineering Laboratory(INEL) is performing motor-operated valve(MOV) research in support of the U.S. NuclearRegulatory Commission's (NRC's) effortsregarding Generic Issue 87, "Failure of HPCI[High-Pressure Coolant Injection] Steam LineWithout Isolation," and Generic Letter 89-10,"Safety-Related Motor-Operated Valve Testingand Surveillance."

This report updates the research reported inNUREG/CR-5558, Generic Issaie 87 FlexibleWedge Gate Valve Test Program Phase IIResults and Analysis (Steele et al., 1990). Thisupdate also provides research results on programobjectives not covered completely in that report.These objectives include

* Examine the use of in situ test results toestimate the required forces of a valve atdesign-basis conditions

* Examine the methods used by industry topredict required valve stem forces ortorques

* Identify guidelines for satisfactory perfor-mance of MOV diagnostics systems

* Participate in writing a performance stan-dard or guidance document for acceptabledesign-basis tests.

In preparing this report, we reanalyzed selectedresults from all four of the full-scale valve testprograms performed for the NRC EquipmentOperability Program, as well as results from theongoing separate effects testing currently beingconducted at the INEL. The full-scale valve testprograms include

* The Test Program for Containment Purge-and-Vent Valves is reported in NUREG/CR-4648, A Study of Typical NuclearContainment Purge Valves in an Accident

Environment (Watkins et al., 1986). In thistest program, three butterfly valves, two8-in. and one 24-in., were tested at linebreak flows at closing differential pressuresof 5 to 60 psig.

The testing reported in NUREG/CR-4977,SHAG Test Series: Seismic Research on anAged United States Gate Valve and on aPiping System in the DecommissionedHeissdampfreakior (HDR) (Steele andArendts, 1989). A 25-year old, 8-in.,'dcpowered, motor-operated Crane gate valvewas refurbished and installed in an exper-imental reactor flow loop. Test loadingsincluded flow, pressure, temperature, andseismic excitations.

* The Phase I Generic Issue 87 Test Programreported in NUREG/CR-5406, BWR [Boil-ing Water Reactor] Reactor Water CleanupSystem Flexible Wedge Gate Isolation ValveQualification and High Energy Flow Inter-ruption Test Program (DeWall and Steele,1989). This test program subjected two6-in., motor-operated, flexwedge, contain-ment isolation gate valves to (a) the appli-cable hydraulic qualification test outlined inAmerican National Standards Institute(ANSI)/American Society of MechanicalEngineers (ASME) B 16.41, (b) a test atnormal reactor water cleanup (RWCU)system flow and pressure, and (c) tests atfull line break flows at design-basis andother parametric fluid conditions.

* The Phase II Generic Issue 87 Test Pro-gram, reported in NUREG/CR-5558 (Steeleet al., 1990). This test program subjectedthree 6-in. and three 10-in. flexwedge,motor-operated gate valves to (a) applicablehydraulic qualification tests outlined inANSI/ASME B16.41, (b) tests at normalsystem pressure and flow, and (c) full linebreak flow tests at design-basis and other

I 1 ~~~~NUREG/CR-5720

Introduction

parametric fluid conditions for the BWRcontainment isolation valves. The valveswere representative of those installed in theBWR HPCI turbine steam supply line, theRWCU system, and the reactor coreisolation cooling turbine steam supply line.

The ongoing separate effects testing in ourlaboratory uses the INEL motor-operated valveload simulator (MOVLS). This device uses actualmotor operators and valve stems, which areloaded using a hydraulic cylinder to producerising stem valve loadings typical of those wehave observed in field testing. The MOVLS iscurrently instrumented to directly measure thefollowing:

* Force and torque on the stem

* Position of the stem

* Force on the torque spring

* Displacement of the torque spring

* Torque and limit switch actuation

* Root mean squared (rms) current andvoltage

* Peak to peak current

* Power, power factor, and speed of theelectric motor.

Through the Data Acquisition System, theMOVLS can also display real time calculationssuch as stem factor.

The design and calibration of the MOVLS hasbeen upgraded from strictly a research device to astandard that can be used by the industry to evalu-ate their diagnostic equipment. Current research.includes load-sensitive motor-operator behaviorand comparison testing of ac- and dc-poweredmotor operators.

NUREG/CR-5720 2

2. GENERAL OBSERVATIONS

Manufacturing tolerances, piping geometryeffects, and specific installation effects allinfluence whether a valve can be grouped withother like valves and categorically qualified bysimilarity. The following examples provide someinsights into why we believe such influences needto be considered:

* When the same 6-in. gate valve was testedwith flow in one direction, then refurbishedand tested with flow in the other direction,the failure mechanism changed fromguiding surface failure to seating surfacefailures. We believe this to be the result ofthe internal valve tolerances stacking updifferently.

* A 6-in. and a 10-in. gate valve from thesame manufacturer were design-basis flowtested. The 6-in. valve passed the tests withgood results. However, with the 1 0-in. valvethe flow forces plastically deformed thevalve body guide rails, which increased therequired stem force to close. Upon disas-sembly and inspection of the valve, it wasobserved that the guide rails were notwelded far enough down into the body. Thecurrent nuclear valve qualification standardANSI/ASME B16.41 would have allowedsimilar 3- to 12-in. valves to be qualified,based on the 6-in. qualification test.

Two like 4-in. gate valves were recently dis-assembled, inspected, and tested by a utility.The internal tolerances of both valves werewithin the manufacturer's tolerances, butthe utility lapped and polished the secondvalve until it was as similar to the first valveas possible. The first valve was subjected toboth no-flow static tests and design-basisflow tests. The second valve was alsosubjected to no-flow static tests and

required a higher torque switch setting toproduce the same stem force as the firstvalve. Both motor operators had beendynamometer tested by Limitorque,a show-ing equal output for the same torque switchsettings. The need for the higher torqueswitch setting may have been caused by adifference in the stem factor.

* Testing of containment purge-and-ventvalves revealed that butterfly valvesinstalled in certain orientations downstreamof an elbow could require up to 133% of thetorque required to close the same valve in astraight piece of pipe.

* Two like valves at the same utility, but attwo different units, performed quite differ-ently. One valve burned up two dc motors.The fault was traced to cable sizing. Thecables supplying power to the valve wereundersized, resulting in excessive voltagedrop and motor stall. The cable size at theother unit was larger,'thus avoiding'thevoltage drop, motor stall, and subsequent.motor burnout.

The list could go on, but the point is that we donot know that like valves will behave alike untilsufficient testing can show that they will performin the -generic group. Since all valves cannot bedesign-basis tested in situ, we believe it will benecessary to conduct both testing and analysis toprove that a given valve can be related to aprototype test or a generic qualification group.

a. Mention of specific products and/or manufactur-ers in this document implies neither endorsement orpreference nor disapproval by the U.S. Government,any of its agencies, or EG&G Idaho, Inc., of the use-of a specific product for any purpose.

3 3 ~~~~NUREG/CR-5720

3. SPECIFIC OBSERVATIONS

The first two research update objectives, (a) toexamine the use of in situ test results to estimatethe response of a valve at design-basis conditions,and (b) to examine the methods used by industryto predict required stem force or torques, will becovered in this section. At this time, the INEL canprovide information only on butterfly valves usedin purge-and-vent-valve applications and onwedge-gate valves used in a number of medium-and high-flow applications. Testing currentlybeing performed in Europe and at selectedutilities in the United States may provideinformation on other valve designs and flowapplications later this year.

We have found it useful to distinguish betweenpredictable valves (those whose performance isrepeatable) and nonpredictable valves (thosewhose performance is not repeatable, usuallybecause they experience internal damage whensubjected to high loads during operation). In theprevious section, we discussed the pitfallsassociated with predicting the performance ofvarious kinds of valves, both predictable and non-predictable. The remainder of this report will belimited to discussing predictable valves only.Valves that do not exhibit predictable behaviorunder load are discussed extensively in NUREG/CR-5558 (Steele et al., 1990).

3.1 Use of In Situ Test Resultsto Bound the Response of aValve at Design-BasisConditions

The results of INEL testing indicate that theresponse of a valve can be bounded for specificvalve types and fluid conditions from in situ testresults obtained from either small-scale testvalves or low differential pressure tests. Theresults of recent European testing support thisconclusion. The fact that valves of a given typerespond linearly with pressure for a specific fluidcondition leads one to believe that, with a suffi-cient amount of testing, bounding the response ofa valve is possible. On the other hand, the

equations the industry has used in the past to pre-dict the perfonnance of gate and butterfly valvesare incomplete. The INEL has confidence inbounding the stem force of predictable wedge-gate valves closing against medium to high flowsand in bounding the torque of high aspect ratiooffset disc butterfly valves used in purge-and-vent applications closing against compressibleflows.

3.2 Assessment of ButterflyValves Closing against aCompressible Fluid(Containment Purge andVent)

3.2.1 Background. The expression "butterflyvalve" is a generic term for a rotating-disc, in-linevalve. Of interest to this discussion is theapplication of butterfly valves in nuclear contain-ment purge-and-vent systems. These systemspenetrate the containment boundary and allow airto circulate through the containment; however, inthe event of an accident, these systems must closeto isolate the environment inside the containment.Consequently, the butterfly valve installed inthese systems must be functional both during andfollowing an accident. Industry operabilityassumptions have been based to a large degree onempirical information obtained from work withincompressible fluids or from small valves testedwith compressible fluids. Previous experimentalwork with incompressible fluids has, for the mostpart, been done at very low pressures, with verylow pressure drops with large valves, or withsmall valves. The operability issue concerningcontainment purge-and-vent valves was raisedafter the Three Mile Island Unit 2 incident. Thefirst question dealt with valve actuator sizing:Would an actuator stall and fail to close a valvebecause of the dynamic loads that might beproduced by a high differential pressure acrossthe containment boundary resulting from adesign-basis loss-of-coolant accident? Thesecond question dealt with stress margins to with-stand the loads imposed during the closure. The

NUREG/CR-5720 4

Specific Observations

stress margins are totally dependent on the analy-sis of the predicted loads and are not part of thisdiscussion.

3.2.2 Flow Phenomena through a ButterflyValve. As a part of the program for testing but-terfly valves, reported in NUREG/CR-4648(Watkins et al., 1986), we conducted a study intothe physics of a butterfly valve closing against acompressible flow to better understand the torqueresponse of a butterfly disc. We observed that abutterfly disc responds as an airfoil from the full-open position to the position where maximum liftoccurs (peak torque). With the exception of a discthat is perfectly symmetric and oriented so thatflow is evenly split over both faces, the flow thatpasses around the disc, and the resulting flowperturbation and pressure distribution, willdepend not only on the degree of closure of thevalve, but also on which surface of the disc isclosing into the flow. The torque characteristics ofa valve are the result of the pressure that acts overthe surface of the disc, which must be counter-balanced to move or control the motion of thedisc. With the valve partially closed and the discoriented so that the flat face of the disc is facingupstream when the valve is fully closed(Figure 1), the flow will separate around the disc,with vortices and a low-pressure region develop-ing at location "L." The result of this low-pressureregion is a torque acting on the disc in the openingdirection. However, if the valve is partly closed,with the disc oriented so that the curved face ofthe disc is facing upstream when the valve is fullyclosed (Figure 2), the flow separation andresultant low-pressure region (location "L") willresult in a torque acting on the disc in the closingdirection. However, the fact that this torque acts inthe closing direction is not necessarily helpful oreven benign, because most butterfly valve designsdo not have positive stops. Without actuatorcontrol, the disc would go beyond the fully closedposition to some partially open position in theother direction.

Immediately after peak torque occurs,. theairfoil will stall. The disc then becomes anincreasingly larger flow obstruction, with a moreuniform pressure distribution and decreasing

torque requirements as the valve completes itsclosing cycle. Based on this hypothesis, one.might believe that the pressure acting on a disccan be directly related to the torque for any degreeof valve closure, any type of fluid, or any flowvelocity through the valve. Unfortunately, testresults indicate that, unlike many other valvedesigns, there are no proven equations forpredicting the torque required for a butterflyvalve to close against a compressible fluid flow.Torque predictions must come from valid testingand extrapolation. The validity of test results isinfluenced by the following:

* A compressible flow medium must be usedduring the tests because of the phenomenaof flow separation around a disc and thecreation of a low-pressure region. Anincompressible fluid will usually cavitate,resulting in a less pronounced low-pressureregion. If the fluid is cavitating, the low-;pressure region acting on the disc will belimited to the vapor pressure of the fluid.

* The inlet pressure, rather than the dif-ferential pressure, should be used forextrapolation, because flow rates throughthe valve that cause the flow to becomechoked can result in supersonic flow down-stream of the valve. The sonic plane in thevalve will depend heavily on the degree ofvalve closure, the shape of the disc surface,the relative location of the valve seals to thedisc, and the properties of the fluid. Theseeffects should be minimal, resulting in atorque extrapolating ability for a givenvalve, provided the smaller valve isgeometrically representative of the largervalve.

* The flow rates and pressure drops achievedduring testing must be typical of conditionsexpected in the actual plant installation. Ifthe flow rate through a valve is too low, itwill produce a minimal pressure drop,which in turn eliminates the effect ofcompressibility. Consequently,.the flowfield through the valve and around each sideof the disc will be much more uniform, aswill the resultant pressure profile acting on

5s NUREG/CR-5720


0 -4 Flow

QResultanttorque

6 3210

Figure 1. Effect of low pressure zones on butterfly valve torque-disc oriented with the flat face of the

disc facing upstream.

QD

Resultant torque -4 Flow

6 3202

Figure 2. Effect of low pressure zones on butterfly valve torque-disc oriented with the curved face of

the disc facing upstream.I

NUREG/CR-5720 6


the disc. Thus, the resultant torque will beless than expected at high flow.

* For test purposes, the valve should alwaysbe oriented so that the curve face of the discwould be facing upstream when the valve isfully closed: This orientation results inabounding extrapolation of the closingtorque using the methods proposed in thisreport, whereas the other valve orientationdoes not.

* For very low flow applications, the last20 degrees of disc closure should beincluded in the test, because the torque reac-tion from the seals and bearings in the valvemay be much larger than the correspondingdynamic torque and will always workagainst valve closure.

3.2.3 Existing Butterfly Valve Data andExtrapolation Techniques. An analyticalassessment of the loads on a butterfly valveresulting from the increasing pressure environ-ment of a design-basis accident is difficultbecause of the complex geometries and flowsthrough such a valve and the lack of empiricalinformation on the dynamic response of abutterfly valve in a compressible fluid flow. Also,nonuniform inlet flow configurations will impactthe dynamic response of a butterfly valve. Con-sequently, we performed a number of tests todetermine the response characteristics of nuclearcontainment isolation (butterfly) valves duringaccident conditions, in an attempt to ensure thatthe flow dynamics of such a valve were under-stood. Prior to this work, the only public domaininformation was that produced for the AllisChalmers Company by the National Aeronauticsand Space Administration Langley ResearchCenter (Allis-Chalmers Corporation, 1979). TheLangley Research Center conducted this testingusing a 6-in. valve body and three interchange-able butterfly valve discs. The program did notinclude testing a larger-sized valve to verify theextrapolation theory. Also, Langley performedthe testing for a specific vendor, and did notcompare the results to other vendor designs.

Because of the very high volumetric flow ratesof concern, testing a large butterfly valve under asimulated accident environment is not alwayseconomical or even feasible. Consequently, thevalve manufacturers have developed a method fortesting scale-model test valves and thenextrapolating their performance to predict thetorque requirements of a larger valve. All of theextrapolation techniques used to predict thetorque requirements of a larger butterfly valvehave a common underlying assumption as to thenature of the flow. Specifically, the flow isassumed to be quasi-one-dimensional, and theresponse is assumed to be linear. Again,unfortunately, the flow through a butterfly valvehas very real and complicated three-dimensionalflow perturbations; therefore, an inherentcompromise must be accepted when extrapo-lating the performance of a scale-model test valveto a larger valve. To further complicate the flowfield, the effects of compressibility must beacknowledged. Compressibility effects can causethe flow through a valve to become choked andallow the downstream pressure to vary indepen-dent of the upstream pressure.

Each of the extrapolation techniques used bythe manufacturers contains a common extrapo-lation term that relates the size of a large valve tothe size of the scale-model test valve. This term isthe cube of the nominal diameter of the valvebeing predicted, divided by the cube of thenominal diameter of the scale-model test valve.Later sections of this report will show that thisterm will yield a bounding prediction, if data areused with the scale-model test valve oriented sothat the curved face of the disc is facing upstreamwhen the valve is fully closed. A nonboundingprediction could result if the flat face of the disc isfacing upstream when the valve is fully closed.

Most of the extrapolation techniques have adifferential pressure term. The manufacturersassume that the differential pressure across ascale-model test valve will be the same as in alarger valve. Consequently, if the differentialpressure term is eliminated from the equation, thetorque extrapolation technique reduces to thetorque being a direct function of the diameterratio cubed.

7 NUREG/CR-5720


For our testing, we selected three nuclear-designed butterfly valves typical of those used incommercial nuclear power plant containmentpurge-and-vent applications. For a comparison ofresponse, we tested two 8-in. nominal-pipe-sizebutterfly valves with differing internal designs.For extrapolation insights, we also tested a 24-in.nominal-pipe-size butterfly valve (made by thesame manufacturer who made one of the 8-in.valves). Figures 3 and 4 show cross-sectionalviews of the valves, which were ASME CodeClass III, ANSI 150 pound class, in-line, off-setdisc, elastomer-sealed, high aspect ratio butterflyvalves (thickness of the disc is relative to thediameter). These valves are typical of designs upto, and including, 24-in. nominal diameter. InFigure 3, the elastomer seal is part of the body; inFigure 4, the elastomer seal is part of the discassembly. Figure 5 is a composite cross-sectionalview of all three discs, with the 24-in. discreduced by a factor of 3.

3.2.4 Butterfly Valve Dynamic Flow Testingand Results. The results of the testing of butter-fly valve dynamic flow were analyzed to assess thebutterfly valve closing torque extrapolationmethodology used by the industry and to quantifythe influence of piping geometry on the torqueresponse of a butterfly valve.

-We performed the experiments with variouspiping configurations and disc orientations tosimulate various installation options that could beencountered in the field. As a standard forcomparing the effects of the various installationoptions, we initially performed our testing in astandard ANSI test section. The nominal oruniform inlet flow ANSI test section is shown inFigure 6, and the valve test positions for this con-figuration are shown in Figure 7. During the flatface forward (FFF) tests, the disc was oriented sothat the flat face of the disc would be facingupstream when the valve was fully closed. Like-wise, during the curved face forward (CFF) tests,the disc was oriented so that the curved face of thedisc would be facing'upstream when the valvewas fully closed.

The test section for nonuniform inlet flow isshown in Figure 8; the valve test positions for

this configuration are shown in Figure 9. As theflow bends around the elbow immediatelyupstream of the test valve, the resultant flowprofile results in higher velocities near the outsideradius of the elbow. Unlike the uniform inlet flowtest section, this nonuniform flow profile willinteract with the disc differently, depending onthe direction in which the disc is rotating towardthe closed position. As such, the clockwise (CW)and counterclockwise (CCW) notationsassociated with the nonuniform inlet flow testsidentify orientations with the disc rotating clock-wise or counterclockwise relative to the figure.

Each test was performed while the valveupstream pressure was controlled at a relativelyconstant pressure throughout the valve closurecycle. Each test cycle consisted of stabilizing thevalve upstream pressure with the valve' in thefully open (90-degree) position. We then closedthe valve at 18 degrees/second to the fully closed(0-degree) position and reopened it after a 250-msdelay. Testing cycles were performed with theupstream pressure varied up to the design-basispressure of 60 psig while monitoring the positionand torque of the valve shaft, the mass flow rate,and the temperature and pressure at various loca-tions throughout the system (shown in Figures 6and 8).

3.2.5 Butterfly Valve Test Results andTorque Bounding Methods. The test resultswere first reduced and entered into a computerdata base. This data base was then used as acommon base for presentation, comparison,analysis, and plotting. The ratio of valve differen-tial pressure to upstream pressure versus valveposition was plotted at pressures of 15, 30, 45,and 60 psig for one of the 8-in. CFF valve tests.The results (Figure 10) indicate that the responsefor the 15-psig test is very different from theresponse for the higher-pressure tests. However,the response for the higher-pressure tests is verysimilar beyond the 40-degree position. Thisdifference is indicative of choked flow at thehigher test pressures and results in a supersonicflow region downstream of the valve that fluctu-ates as the valve closes. This, in turn, results in adownstream pressure profile (Figure 11) that is

NUREG/CR-5720 8


Shaft

Shaft seals

- EPT seal ring

. Metal seat

Purge valve INEL 4 0488

Figure 3. Cross section of Valve 1, the first 8-in. butterfly valve.

9 NUREG/CR-5720


Shaft seals

Metal seat

-EPT seal ring

Figure 4. Cross section of Valve 2, the second 8-in. butterfly valve, and Valve 3, the 24-in. butterflyvalve.

NUREG/CR-5720 10


Valve 1 (full scale)Valve 2 (full scale)

-- - Valve 3 (1/3 scale)

------ ---

- …-

f ~~~~~~~~~~--,-4 IN,

Sealing surface5 2979

Figure 5. Containment butterfly valve disc overlay.

TE-1 02 . PT-1 02

ET6 0 V-102D

PT-106 |TE10

PTPDTTETV

= pressure= differential pressure transducer= temperature= test valve

PT-COQ-ORupture disc

TE-M

TE- K, PT-K 0 0

PT-D 00

PT-J, TE-J-

PT-H

.. TE-H

PT-GT0

TE-G

PT-APT-B 0 ~~PT-M

TE-C TE-D

IPDT I

PT-E PT-F _a 0a 0

O~

TE-E TE-F-

5 2999

Figure 6. Typical installation-butterfly valve uniform inlet flow test section.

PFlow-i.-ft

Flat face forward(FFF)

. ~~~~~~~~~~~~~.Flow - .-.- .

Curved face forward(CFF)

5 2980

Figure 7. Uniform inlet flow butterfly valve positions.

11 NUREG/CR-5720


TE-I 021 ao If PT-1 02

PT = pressurePDT = differential pressure transducerTE = temperatureTV = test valveV- 02D

Rupture disc

PT-J, TE-JPT-B TE-D TE-E TE-G TE-H TE F

LPDT J 5 4000

Figure 8. Typical installation-butterfly valve nonuniform inlet flow test section.

FFF-CCW _FF-CCW

FFF-CW IFFFCW 5 2981

Figure 9. Nonuniform inlet flow butterfly valve positions.

NUREG/CR-5720 12


0

jor':L an

9U 80 70 60 50 40 30 20 10 0Valve position 5 2982

Figure 10. Butterfly valve differential pressure to upstream pressure ratio versus valve position.

40

30

0.

-a

' 20a)

0L

90 80 70 60 . 50 40 30 20 10 0Valve position 5 2983

Figure 11. Static pressure 15 diameters downstream of a butterfly valve versus valve position.

13 NUREG/CR-5720


very different from that experienced duringincompressible flow and very different for eachof the valves tested. Specifically, the downstreampressure, as measured 15 diameters downstreamof each test valve, did not always recover fromthe flow perturbation during certain portions ofthe valve closure cycle. This indicates that themeasurement location was in a supersonic region.

These results suggest that torque extrapolationpractices using the differential pressure do notaccount for supersonic flow downstream of thevalve and its resulting effect on valve torqueduring a design-basis accident. Therefore, weintroduced a new parameter (upstream pressure)and developed plots of valve response, relatingvalve upstream pressure, dynamic torque, and theposition of the disc. Figures 12 through 17 arethe response plots for the three valves tested in theuniform inlet flow configuration. The figures forthe CFF orientation indicate a butterfly valveresponding with a positive or self-closing torque.In this orientation, the operator must supplytorque to keep the valve from shutting too rapidly.The figures for the FFF orientation indicate abutterfly valve responding with a negative or self-

opening torque. In this orientation, the operatormust supply torque to close the valve. Therefore,butterfly valves in the FFF orientation will beharder to close, and butterfly valves in the CFForientation will be harder to open.

Analysis of the response plots shows that themagnitude of the dynamic torque when the valvewas in the CFF orientation (a positive response)was greater than the magnitude of the dynamictorque when the valve was in the FFF orientation(a negative response). Also, the positive dynamictorque curves of the three valves in the CFForientation, as shown in Figures 12 through 14,are very similar in appearance. Conversely, thenegative dynamic torque curves of the three valvesin the FFF orientation, as shown in Figures 15through 17, are very different in appearance. Thisprovides some assurance that limited extrapola-tion is possible using the upstream pressure (ratherthan the pressure drop) across the butterfly valveif the valve is in the CFF orientation.

The peak torque for each of the three butterflyvalves tested with a uniform inlet flow con-figuration was plotted against upstream pressurein Figures 18 through 20. The results indicate

r- Angle

-50

.0

0

100

- 150L_

Figure 12. Torque versus upstream pressure and angle for Valve 1, the first 8-in. butterfly valve, FFForientation.

NUREG/CR-5720 14


Angle

-50

C,

0~

01 -

5 2986

-1o00 o

-150 -

Figure 13. Torque versus upstream pressure and angle for Valve 2, the second 8-in. butterfly valve, FFForientation. -

Angle

Or

-500 1

~0

0

I-

-1000 1-

-1500 F-

-2000 L

Figure 14. Torque versus upstream pressure and angle for Valve 3, the 24-in. butterfly valve, FFForientation.

15 NUREG/CR-5720


60

Ic~r

-.0

.0

0I3

50pI

0

Angle

10

5 2988

Figure 15.orientation.

Torque versus upstream pressure and angle for Valve 1, the first 8-in. butterfly valve, CFF

100o r

-o

c

50 -

30

0 O L

90 80 70 60 50Angle

40 3010

5 2989

Figure 16. . Torque versus upstream pressure and angle for Valve 2, the second 8-in. butterfly valve, CFF

orientation.

NUREG/CR-5720 16


2000 -

1500

-

-0I

ET

45

1000 _

30

500 15

a I

Angle

I'Torque versus upstream pressure and angle for Valve 3, the 24-in. butterfly valve, CFFFigure 17.orientation.

200

160

120

80

'a 40

40

0 -40

-80

-120

- 160

* -2000 10 20 30. 40 50 60 70 80 90

Static upstream pressure (psia) 5 2991

Figure 18. Peak torque versus static upstream pressure for Valve 1, the first 8-in. butterfly valve.

17 NUREG/CR-5720


200 III I I I I I

160 ;-

120

80 1

-,

-0

0*

01-

40 1C~~~FF

0 0

FFF

0

-40 1

-80

- 120

- 160

-200 I Il

0 10 20 30 40 50 60Static upstream pressure (psia)

70 80 905 2992

Figure 19. Peak torque versus static upstream pressure for Valve 2, the second 8-in. butterfly valve.

3200

-0.

0Hr

30 40 50 60Static upstream pressure (psia)

905 2993

Figure 20. Peak torque versus static upstream pressure for Valve 3, the 24-in. butterfly valve.

NUREG/CR-5720 18


that a linear relationship exists between the twoparameters, although the angle where peak torqueoccurs varies with pressure. The fact that theresponse is linear reinforces the confidence thatthe response of a butterfly valve can beextrapolated using the upstream pressure.

The validity of a diameter ratio cubed term, asused in typical extrapolation relationships, wasalso evaluated. The dynamic torque of the 8- and24-in. valves from the same manufacturer, withboth valves in the CFF orientation, was comparedfrom the fully open position (90 degrees) to nearthe fully closed position (20 degrees) at upstreampressures of 15, 30, 45, and 60 psig. The resultsof this evaluation are shown in Figure 21. Theextrapolation exponent was always below 3except for one position at an inlet pressure of60 psig. These results indicate that use of anextrapolation exponent of 3 will result in the pre-dicted torque of a larger valve being slightlygreater than the actual torque, provided the inletpressure does not exceed 60 psig. The trend ofthe data indicates that an extrapolation exponentof 3 could underpredict the actual torque of thelarger valve at higher inlet pressures.

This evaluation was then repeated for tests witheach valve in the FFF orientation. The results areshown in Figure 22. This evaluation indicates thatthe extrapolation exponent, when the flat face ofthe disc is facing upstream, with the valve fullyclosed, will frequently be above 3, ranging as highas 3.47 in these tests. Thus, using an extrapolationexponent of 3 could result in the predicted torqueof a larger valve being less than the actual torque.To examine the effect the extrapolation exponenthas on the estimated torque, we calculated torquesfor large valves, using an extrapolation of small-valve data and extrapolation exponents rangingfrom 2.8 to 3.5. The results are shown in Fig-ure 23. This figure indicates that the torquerequirements of a larger valve could be seriouslyunderestimated if an extrapolation exponent of 3is used when a larger exponent should be used. Forexample, if the torque requirements of a 42- to48-in. valve were being predicted from the resultsof a 3- to 4-in. scale-model test valve, an extrapola-tion range on the order of 10 to 12, the torque of thelarger valve could be underestimated by 50% or

more, if an extrapolation exponent of 3 is usedinstead of, for example, 3.2.

Consequently, a diameter-ratio-cubed formula-tion appears justified if (a) torques are obtainedwith the scale-model test valve oriented with thecurved face of the disc facing upstream when thebutterfly valve is fully closed, and (b) theupstream pressure does not exceed 60 psig. Wedeveloped the following equations to more con-sistently envelop the response of a larger valvebased on the response of a smaller scale-modeltest valve. Note that, in the CFF orientation, thebearing torque (the torque that must be suppliedto move the valve disc against the resistance ofthe bearings and seal only) and the dynamictorque (the torque that must be supplied to movethe valve disc against the resistance of the flow-ing fluid only) are acting in opposite directions,the dynamic torque assisting valve closure andthe bearing torque resisting it. Either equation canbe used, depending on the information available(i.e., whether the total and bearing torques or justthe dynamic torque of the smaller scale-modeltest valve is known). In either case, an estimate ofthe bearing torque of the larger valve must beknown.

D3

D dS

- Tbh C (1)

Tb,) - Tb, (2)

where

- valve total torque (dynamic +bearing torque)

e = a parameter of the large valve

D = valve diameter

s = a parameter of the small, scale-model test valve

Td = valve dynamic torque only

Tb = valve bearing torque only

19 NUREG/CR-5720


4 l

za,

.0

a,0

OC 3 -x

a,

0

.-

a,

(D

E

0,,

0- I

Valve position (deg) 6 3200

Figure 21. Extrapolation exponent, butterfly valve oriented with the curved face of the disc facingupstream. -

4

C

00.3x

a)C0

00-

E 2

0I2x

a)a,

o0 1Ec)

n _90 80 70 60 50 40 30 20Valve position (deg) 5 3211

Figure 22. Extrapolation exponent, butterfly valve oriented with the flat face of the disc facing upstream.

NUREG/CR-5720 20


0

0~

0)

0

1 10 100

Valve diameter ratio (D%/ Ds)6 3201

Figure 23. Actual to predicted butterfly valve torque (percent) as a function of extrap6lation exponentversus valve diameter ratio.

21 NUREG/CR-5720


A comparison between the results of themethod presented here and the results of a methodtypically used by industry is presented in Table 1.This table also provides a direct comparisonbetween the test results from this program and themethods used by industry to predict the torquerequirements of a 24-in. butterfly valve from thetest results of an 8-in. butterfly valve. The resultsindicate that, as expected, torque extrapolationsperformed with test data obtained from a valveoriented with the curved face of the disc facinginto the flow will bound the torque demands ofeither orientation. However, extrapolations basedon the results from a test valve with the flat faceof the disc facing into the flow typically do notbound the data.

We then used the proposed technique, asreflected in Equation (1), to predict the responseof a 48-in. butterfly valve using both the 8- and24-in. butterfly valves as the scale-model testvalves. The results, shown in Figure 24, indicatesome differences between the two 48-in. valvepredictions. However, the extrapolations for 8-in.to 48-in. valves generally result in higher (moreconservative) values than the extrapolations from

24-in. to 48-in. valves. This is particularly true forthe peak torque.

3.2.6 Effect of an Upstream Elbow on theTorque Requirements of a ButterflyValve. We also investigated the effects of systemupstream geometry on the closing torque require-ments of a butterfly valve. We compared the testresults for valves located immediately down-stream of an elbow to the results with uniforminlet flow. (The valves were installed as close aspossible to the elbows in order to expose them tothe maximum nonuniform flow anticipated in anactual installation.) The peak torque at 60 psigwas tabulated for all three valves in each of the sixorientations tested. These torques were thennormalized to the peak torque at 60 psig for eachvalve in the uniform flow CFF orientation andtabulated for easy comparison (Table 2).

Using this table, we can assess the effect non-uniform inlet flow relative to uniform inlet flowhas on valve torque. The worst-case elbow effectwas noted for one of the 24-in. valve orientations,1.33 times the uniform inlet torque. This wasfollowed closely by one of the 8-in. valve

Table 1. Comparison of torque prediction methods butterfly versus valve orientation.

Torque(ft-lbf)

Torque prediction method CFF orientation FFF orientation

Predicted 24-in. valve torque at 900 and 60 psig

Actual 8-in. valve torque 70.2 -52.1

Predicted 24-in. valve torque (INEL method) 1895 -1895

Predicted 24-in. valve torque (Industry method) 1895 -1407

Actual 24-in. valve torque 1754 -1828

Predicted 24-in. valve torque at 800 and 60 psig

Actual 8-in. valve torque 99.2 -58.0

Predicted 24-in. valve torque (INEL method) 2373 -2373

Predicted 24-in. valve torque (Industry method) 2984 -1875

Actual 24-in. valve torque 2257 -1986

NUREG/CR-5720 22


0 10,000 _

0

-10,000 _90 80 70 60 50 40 30 20 10 0

Valve position 5 2995

Figure 24. Predictions for a 48-in. butterfly valve based on extrapolating the torques of an 8-in. and a24-in. butterfly valve at upstream pressures of 15 and 60 psig.

Table 2. Ratio of peak torque to uniform flow peak torque for a butterfly valve in the CFF orientation.a

Valve position Valve 1 Valve 2 Valve 3

FFF 1.06

CFF 1.00

FFF-CCW' 0.90

CFF-CCW 1.29

FFF-CW 1.02

CFF-CW 1.14

a. Valve position identification is given in Figures 7 and 9.

0.81

1.00

0.83

1.00: I

0.84

0.95

0.94

1.00

1.33

1.04

0.87

0.92

23 NUREG/CR-5720

Specific Observations-

orientations, which was 1.29 times the uniforminlet torque. Based on these results, the maximumtorque expected from a nonuniform inlet flowconfiguration can be bounded by using 1.5 timesthe torque from the uniform inlet flow configura-tion, if the curved face of the disc is facingupstream when the valve is fully closed.

Next, a valve response plot was developed(Figure 25) for the 24-in. butterfly valve in theCFF orientation, with the shaft of the valveperpendicular to the plane of an upstream elbow.The similarity of the shape of this response plot tothe response plot of the same valve without anupstream elbow (Figure 17) is clear. Thiscomparison suggests that, although the torquesresulting from a nonuniform inlet flow configura-tion may be higher, the response can still bebounded with a factor of 1.5 times the torquefrom the uniform inlet flow configuration, if thecurved face of the disc is facing upstream whenthe valve is fully closed.

Finally, Figure 26 compares the linearity ofthe peak torque versus inlet pressure for the 24-in.butterfly valve with and without an upstreamelbow. Generally, the torque in the presence of anupstream elbow is higher, but the responseremains linear. This comparison provides addedconfidence that the results of the nonuniform inletflow configuration can be bounded.

This study has shown a weakness in the indus-try's extrapolation procedures for butterfly valvesclosing against a compressive fluid. The industry(a) has not identified a dominant orientation forthe small valve to be tested in, (b) has useddifferential pressure (which is influenced non-conservatively by downstream pressure) insteadof upstream pressure, and (c) has not identifiedthe effect that the upstream piping geometry hason the torque requirements of these valves, all ofwhich were shown to be important. Most butter-fly valve vendors are aware of the NRC test pro-gram, but it is not known to what extent they haveincorporated the test results into their proprietarycalculations. It is known that most large purge-and-vent valves are either closed or blocked atsmall angles of opening and that all purge-and-vent valves required plant submittals and NRC

Safety Evaluation Reports (SERs) in response toThree-Mile Island Action Plans NUREG-0660and NUREG-0737 (NRC, 1980a; NRC, 1980b).A large percentage of the purge-and-vent valveswere reviewed before these test results wereavailable, and the status of purge-and-vent valvesreplaced in the last five years is not known.Generic Letter 89-10 will not cause manyre-reviews, because many purge-and-vent valvesare not motor operated.

3.3 Assessment of Wedge-GateValves Closing againstMedium to High FlowConditions

Flexwedge gate valves were tested in two NRCtest programs referenced earlier in this report;those tests were reported in NUREG/CR-5406(DeWall and Steele, 1989) and NUREG/CR-5558(Steele et al., 1990). The latter was published insupport of Generic Issue 87. After that reportwas published, we developed a technique to(a) bound the stem force of a 5-degree flexwedgegate valve closing against medium to high flowconditions, and (b) validate a low differentialpressure closure test and then bound the stemforce of a flexwedge gate valve closing againstdesign-basis conditions.

In the two test programs mentioned above, theauthors tested six valves with a total of sevendifferent internal designs. One valve was testedwith two different discs, one with hardfaced discguides (Valve B Phase I), and one withouthardfaced guides (Valve 2 Phase II). Under thehigh loadings encountered during the testing ofthis particular valve, there was no difference inthe performance of the two discs.

The valves were tested under a broad range offluid conditions and flow rates, from normalsystem flows to design-basis line break flows.Initial conditions for each valve tested are shownin Tables 3 and 4. These conditions were estab-lished before the normal system flow isolationand the design-basis flow isolation portions ofeach test. Two of the valves, including the valve

NUREG/CR-5720 24


2000

1500

n

0-a, 1 coo

G.

5001-

psIg 5 2996

Figure 25. Torque versus upstream pressure and angle for Valve 3, the 24-in. butterfly valve in the CFForientation.

3200

2400 -

.0~~~~~~~~~~~~~~

*1600 o~~~ ~ ~ 80

r--

. 1600 . ~~~CFF-CC /

-800 90

-800 /1III0 10 *20 30 40 50 60 70 80 90

Static upstream pressure (psia) s 2997

Figure 26. Peak torque versus static upstream pressure for Valve 3, the 24-in. butterfly valve, comparingthe response of the peak torque with elbow and peak torque without elbow orientations.

25 NUREG/CR-5720


Table 3. Phase I gate valve flow interruption test temperatures and pressures.

Pressure(psia)

Temperature(OF)Valve Test Test media

AAAAA

AAAAA

23456

789

1011

10001000100014001400

14006006006001000

530480400580530

450480430350530

Hot waterHot waterHot waterHot waterHot water

Hot waterHot waterHot waterI-lot waterHot water

BBBB

2345

10001400600

1000

530580480530

Hot waterHot waterHot waterHot water

Table 4. Phase II gate valve flow interruption test temperatures and pressures.

Pressure Temperature(psig) (OF)

Valve Test Target Actual Target Actual Test media

6-in. Valve Tests

1 1 1000 900 530 520 Hot water2 1 1000 950 530 520 Hot water2 2 1000 1040 545 550 Steam2 3 1000 750 <100 <100 Cold water2 6a 600 600 300 450 Hot water2 6b 1000 1000 430 470 Hot water2 6c 1400 1300 480 520 Hot water3 1 1000 920 530 520 Hot water3 5 1200 1100 550 550 Hot water3 7 1400 1300 580 570 Hot water

10-in. Valve Tests

4 1 1000 750 545 510 Steam5 la 1000 800 545 520 Steam5 lb 1400 1040 590 550 Steam6 la 1000 990 545 580 Steam6 lb 1400 1400 590 590 Steam6 Ic 1200 1100 570 550 Steam

NUREG/CR-5720 26


that was tested with two discs, performed in amanner we have called predictable. A predictablevalve is one that does not exhibit evidence ofinternal damage during testing. In such valves,the highest stem forces occur when the disc isriding on the valve body seats just before wedg-ing. Conversely, an unpredictable valve exhibitsevidence of internal valve damage during testing,characterized by an erratic, sawtooth-shaped stemforce response. In such valves, these high stem-force requirements typically occurwhile the discis riding on the guides rather than just beforewedging. Generally speaking, the results fromtesting unpredictable valves are not useful for thekind of evaluation described here. However,through selective analyses, we were able toinclude some of the results from two unpredict-able valves, not while the disc was riding on theguides, but after the disc had transitioned to ridingon the valve body seats.

We initially evaluated the test results with thestandard industry gate-valve stem-force equation.Although some of the manufacturers modify thevariables in these equations slightly, theapplication of the equations is basically the same.

AP = differential pressure across thevalve

= area of the stem

P = pressure upstream of the valve

= packing drag force

= disc and stem weight

= dynamic coefficient of frictionbetween disc and seat

Aorifice

0

= disc area on which pressureacts

= seat angle (degrees from stemaxis)

F, = sealing force (wedging forceonly)

Rt -= friction factor at torque reac-tion surface

FS = stem factor

F1,Industry = I'dAdAP ± AP + Fp (3)= distance to torque reaction

surface.

Later in the program after we developed theINEL correlation, the Nuclear MaintenanceAssistance Center (NMAC) equation wasreleased. We also evaluated that equation, whichfollows:

F., Fp-A, P

usA 0,i,,A P,±~~~_±FcosO ± u~sinO

F., NMIAC = (4)

For wedge-type gate valves, the industry hashistorically used a disc factor (Wd) of 0.3; morerecently, they sometimes specified a more conser-vative disc factor of 0.5. The disc factor acts inconjunction with the disc area and the differentialpressure; the three multiplied together representthe largest component in the stem force equation.However, the disc area used in the area term is notuniformly applied throughout the industry. Thearea is based on the orifice diameter by somemanufacturers, on the mean seat diameter byothers, or even on the orifice diameter times oneor more factors to artificially enlarge the area onwhich the differential pressure acts. We used adisc area based on the orifice diameter when weevaluated the standard industry stem force equa-tion, because it represents the least conservativeuse of the term by the industry. This results in alower estimate of required stem force than theother disc area terms.

where

= stem force

Ed

Ad

= disc factor

= orifice area

27 NUREG/CR-5720


It is important to emphasize that the use of onedisc factor over another, or the comparison of onedisc factor to another, depends heavily on the discarea assumed. In other words, a disc factor devel-oped empirically with the mean seat area will belower and, thus, cannot be used to estimate thestem force using the orifice area. Likewise, thepoint in the closing cycle at which a disc factor isdetermined is also of great importance. A discfactor determined from the closing thrust historyprior to total flow isolation will underestimate theisolation stem force if any extrapolation is neces-sary. The disc factor should be defined at flowisolation, just before wedging.

For the standard industry equation, the esti-mated stem force is always a positive value, and itis up to the analyst to differentiate between theforce to open a valve and the force to close avalve. As a result, the stem rejection force mustbe represented as a positive value if the valve isclosing and as a negative value if the valve isopening. This is because the stem rejection forcealways acts in an outward direction relative to thevalve body, resisting valve closure and assistingvalve opening. The packing force is always repre-sented as a positive value because it alwaysopposes motion. It is typically assumed to beconstant for a given valve, but it varies from valveto valve, depending on the packing design and thepacking gland nut torque.

For the NMAC equation, a dynamic coefficientof friction of 0.35 to 0.50 is recommended. Morespecific guidance is not given. At the time weinitiated our study, the disc area term in theNMAC equation was labeled Aoi-,ce and definedas the area on which pressure acts. The exactmeaning of this term was not clear, but weinitially used the orifice area in our assessment ofthe NMAC equation. Unlike the industryequation, however, the estimated stem force isrepresented as a negative value for closing and asa positive value for opening. The packing force,the force due to the differential pressure acrossthe disc, and the sealing force can be either posi-tive or negative, depending on whether the valveis opening or closing. The weight of the disc andstem acts in an inward direction relative to thevalve body, so it is always positive. The stem

rejection force acts in an outward direction, so itis always negative.

Comparisons of the standard industry equation,Equation (3), with selected test results are shownin Figure 27. This figure presents the results ofthe same valve isolating a break at a commonupstream pressure of approximately 1000 psig,but with the fluid at various degrees of sub-cooling. The subcooling ranges from none(steam) to approximately 400'F (cold water) withintermediate values of LOTF and 1000F. Therecorded stem force is shown as a solid line; twocalculations of the stem force history, using theindustry equation and real time test data withstandard industry disc factors of 0.3 and 0.5, areshown as dashed lines. This figure shows that, atflow isolation, each test required more force toclose the valve than would be estimated using thestandard industry disc factor of 0.3. In fact, for thetests shown on this figure, the more conservativeindustry disc factor of 0.5 ranges from acceptable(the steam test) to marginally acceptable (the100F subcooled fluid test) to unacceptable (the1000F and the 400'F subcooled fluid tests). Notethat, although the results of the industry equationare presented over the entire closure cycle, theequation represents a bounding estimate of themaximum stem force. Therefore, only theestimated stem force at the final horizontal line,just before wedging, is applicable. The results arepresented for the entire closure to. aid in identify-ing trends in the recorded stemr force, not to assessthe equation throughout the closure cycle. Notealso that, although the same valve and operatorwere used for each test, the closure durations aredifferent. Because of facility inventory limita-tions, some of the tests were initiated with thevalve partially closed.

It is also interesting to note that the shape of therecorded stem force from flow initiation to flowisolation varies, depending on the degree ofsubcooling of the fluid. In tests with greater sub-cooling, the stem force during the initial portionof the closure is lower. In fact, during the test withcold water, the stem force trace was initiallypositive (i.e., the valve was self closing duringthis portion of the closure). However, just before

NUREG/CR-5720 28

20000

*__10*000

a)

IL

f--10000

U)

-20000

-30000

I I I I jII I I ,.-

- Valv'e 2, 1000 psi,430°F '(100°F ubcooled) --__ '___ Actual_------- Calculated (/L = 0.3)

*-. Calculated (/L- 0.5)

Flow initiation .........

~~ Flow isolation -

0 5 10 15 20 25 30 3

Time (s)

20000

10000

a)c-, 0

tL

E 10000

-2000 0

-30000

I.

5 0 5 10 15 - 20

Time (s)25 30 35

k)

20000

10000_

c. 00

U-

IE.10 0 0 00)

U)

-20000

"AnnnnII I 11111111111 I 1111 I ~~~~~-V- I -I1

Valve 2. 1000 psi, 530°f (10°F gubcooled)Actual

-------- …Calculated (,u - 0.3)_ *-. Calculatod (tz = 0.51

_ I

10000.0

a)Oc 00

IL

Q)

U)

-20000

-30000

' 111|1 I.I ' ' I ' ' ' ' 1) ' ' I l '' 'l- ' ' ' -1-

Valve 2. 100 psi, 545°F (Steam)-~ Actual-------- Calculated (tz = 0.3) --.---------- Calculated (/L - 0.51

.

--

I I I I I I - II-

z

Cr

PO

0

-300000 5 10 15 20 25 30 35

Time (s)0 *5 10 15 20 25 30 - 35

Time (s) MS291 RS-0491-04

Cn

D

00_

O

CD

PW0.Figure 27. Comparison of the standard industry gate valve stem force equation with selected test results.


wedging, the required stem force is generallyhigher in tests with greater subcooling of the fluidso that closure against cold water requires moreforce than closure against steam.

A comparison using the NMAC equation,Equation (4), is shown in Figure 28. In general,the estimated stem force using this equation isvery similar to the estimated stem force using theindustry equation. The trends in the stem forcetrace during a valve closure are also similar. Aswith the standard industry equation, the NMAC

equation represents a bounding estimate of themaximum stem force. The results of the NMACequation were presented for the entire closure toaid in identifying trends in the recorded stemforce, not to assess the equation throughout theclosure cycle.

3.3.1 Assessment of the Disc Factor Termin the Industry Equation. As we studied thetest results and analyzed the industry equation, itbecame increasingly evident that the disc factoror friction factor used in the equation was not well

0

CD

0W

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.i ....... _ j..........

.~~ . ... .......... * ................. ....................... ..................... , .............. .......i

Actual; HCalc late "g- . 5

.......... al u at d .. 0 0

Eci

co)-20000

-300000

....... ..... ... - - -5 10 15 20 25 30 35

Time (s)Figure 28. Comparison of the NMAC gate valve stem force equation with selected test results.

NUREG/CR-5720 30


understood. It appeared that the disc factordepended on parameters not currently beingaccounted for, such as the subcooling of the fluid.Thus, we examined the equation in more detail,specifically the disc factor term as currentlydefined. To perform this evaluation, we used boththe design-basis and applicable parametric testingto determine what influence pressure and fluidproperties, such as subcooling, had on therequired stem force of a valve.

If all of the parametric studies could haveresulted in just one parameter being varied, thenthe tests could have been compared to each otherto determine the effect of that one parameter (e.g.,fluid properties). That was not the case, however.It was impossible to provide such precise temper-ature and pressure control at the valve. This,along with other facility limitations, such as thetotal system supply volume, resulted in tests thatcannot be compared to one another without sometype of normalization.

We normalized the test results usingEquation (3) by solving for the disc factor. Thiswas possible because we knew the stem force, thesystem pressure, and the valve differential pres-sure throughout the closure cycle. The resultingequation used in this evaluation was

-5% travel position. The slight slope on theplateau between -5% and the -9% stem travel isthe result of valve inlet pressure increasingslightly with flow isolation. The plateau regionrepresents the disc factor after flow isolation.Two observations can be made from Figure 29:(a) the absolute magnitude of the disc factor forany fluid condition exceeds a 0.3, and (b) thevalve response is affected by fluid properties,steam having the lowest disc factor and coldwater having the highest disc factor. The fluid-properties effect is evident throughout the closurecycle but is most pronounced on the plateauregion, when the forces resisting closure haveessentially stabilized. This effect is contrary towhat was expected; one would expect water to bea better lubricant than steam. The industryequation and the NMAC equations do not containterms for fluid properties effects.

Our next effort was to determine if the disc fac-tor was dependent on pressure. Figure 30 showsa comparison for Valve 2 using six parametrictests where the fluid properties remained constantbut the pressure was varied. Although the discfactor did not exhibit a significant pressuredependency at the zero stem position, it did fromthe minus 5% to the minus 10% stem position,when the disc was riding on the seats just beforewedging. The figure also indicates that the discfactor was lowest during the 1400 psig test andhighest during the 600 psig test. This, too, is con-trary to what was expected; one would expect alightly loaded disc to have a thicker lubricationfilm and therefore a lower coefficient of frictionthan a heavily loaded disc.

Figure 31 depicts the effect of pressure in theopening direction, further highlighting inconsis-tencies with Equation (3). This plot is read fromleft to right as the valve opens. Although the discfactor (a positive value because the valve is open-ing) did not exhibit a pressure dependency at thezero stem position, it did from the minus 5% tothe minus 10% stem position. This trend issimilar to what was observed in the closingdirection. The figure also reaffirms our previousobservations that the disc factor is lower during ahigh-pressure test and higher during a low-pressure test. The opening disc factor is also

Ft - A, P - Fp1-d Ad AP (5)

The results of a typical comparison are shown inFigure 29, where there are two closures at eachfluid condition. The plot is read from right to leftas the valve closes. As time increases, the discfactor increases in the negative convention (indi-cating valve closure) and is plotted against stemposition. The zero stem position represents thatpoint in the valve closure where the horizontalvisual area is blocked. Recognize, however, that,although the visual area is blocked, fluid can stillflow under the disc and through the valve. At thispoint, the disc area term in Equation (5) becomesconstant. From the zero stem position to theminus 10% stem position, the stem travelinvolves full seat contact, flow isolation, andfinally wedging. The forces maximize about the

31 NUREG/CR-5720

QI

n

T4

0

C-)

Il)

ii

0.0

-0.1

-0.2

-0.3

-0.4

-0.5

-0.6

-0.7

_.

CD

0

0

0

cr_.CD0-t

-15 S -10 -5 0 5 10 15

Stem position (% of travel)

20

MS291 RS-0491-06

Figure 29. Disc factor for Gate Valve 2 closing on line break flow, effect of subcooling at 1000 psig.

0.0

-0.1

-0.2

00Co

4-)CO

(. _

-0.3

-0.4

-0.5

-0.6

-0.7-15 -10 -5 0 5 10 15

cz0

tjn

I

20

MS291 RS-0491-09Stem position (% of travel)

enlz2

0

0

0

5c_.

Figure 30. Disc factor for Gate Valve 2 closing on line break flow, effect of pressure at lOOT subcooling.

z

a

PC

;0

1.0

0.9 Effect of

pressure-I-1 1 |-1 as U I -10 Tb-cllng

fpressure at 100°F slbcooling

c:~

0Cn

la

00

_.00

Opening0.8

0.7

0+.

4-)

u)

.

,, .....

0.6G

0.5

0.4

0.3

0.2

0.1

0.0

600 psig10001400

psigpsig

-

-15 -10 -5 0 5 10 15 20

Stem posiLion (% of travel) MS291 RS-0491-10

Figure 31. Disc factor for Gate Valve 2 opening on line break flow, effect of pressure at I 00F subcooling.


observed to be higher than the closing disc factorat its peak, non-wedging value.

The previously unaccounted influence of fluidsubcooling and pressure on the disc factor is veryevident. This influence is also contrary to whatone might expect in terms of the effectiveness ofa lubricant. However, what was expected is basedon a lubrication that separates the load-bearingsurfaces with a relatively thick film of lubricant tominimize metal-to-metal contact. This type oflubrication is known as thick-film lubrication.The condition resulting from too little lubricationis known as thin-film lubrication. The defi-ciencies of this thin-film lubrication can beaggravated by valve sliding surface areas that aretoo small to carry the maximum load.

When metal-to-metal contact exists, any condi-tion that increases the ability of the lubricant topenetrate the bearing region will decrease thefriction between the two surfaces. For instance,the higher the differential pressure across abearing region, the more likely a given lubricantwill be forced into this region, thus lowering thefriction between the surfaces. Likewise, alubricant in a vapor state is more likely than thesame lubricant in a liquid state to penetrate thebearing region, thus lowering the friction betweenthe surfaces. Other researchers have noticed thesesame phenomena; however, they attribute thissensitivity to changes in the temperature of thefluid and metal. We are still trying to isolate thereason, but the effect is real and must beaccounted for.

From the results discussed above and fromsimilar results for the other valves evaluated, it isapparent that Equation (3) is incomplete and failsto identify and predict the increasing stem forcedue to fluid properties. Equation (4), evaluatedafter development of the INEL correlation, ismore complete and can bound the results,depending on the friction factor used; however,there is little guidance to selecting a friction fac-tor other than a range of 0.35 to 0.50. In addition-,the fluid subcooling and pressure dependencies ofthe disc factor or friction factor are inconsistentwith past assumptions inherent in the industry'sapplication of their equation. Thus, we concluded

that the industry equationsfailed to considerparameters that have an important effect on theobserved responses of the valves.

In response to the above conclusion, wedirected our efforts toward investigating the flowphenomena through a flexwedge gate valve andthe effect that pressures throughout the valve hadon the resultant stem force. That investigationeventually yielded a correlation that bounds therequired stem force during closure with morereliability than the standard industry equation witheither a Q.3 or a 0.5 disc factor, or than the NMACequation with a friction factor of 0.35 to 0.50.

Figure 32 shows a cross section of a typicalflexwedge gate valve and identifies those areas onthe disc and stem where the various pressureforces can act. This figure also shows where wedrilled three pressure measurement ports intoeach of the valve bodies before the Phase II test-ing, to assist in the internal pressure distribution*study. Figure 33 shows a typical pressuredistribution observed during our testing. Thepressures in both the bonnet region of the valveand under the disc are lower than the upstreampressure during most of the valve closure cycle.This reduction in pressure is due to the Bemoullieffect, the result of fluid accelerating through avalve in response to a reduction in the flow area.This phenomenon depends on the pressure andsubcooling of the fluid and on the magnitude ofthe reduction in flow area through the valve.Thus, the Bernoulli effect will be system and fluiddependent. The bonnet area also shows a lowerpressure because of the split in the disc andbecause of the gap between the disc and the valvebody seats; these structural features provide apath so that the pressure in the bonnet region canmore closely follow the pressure in the regionunder the disc.

However, from the minus 3% to the minus 10%stem position during this test, the pressures con-verge. During this portion of the valve stroke,flow has been isolated and the disc is riding on thevalve body seats; however, wedging of the dischas not yet begun. It is also during this portion ofthe valve stroke that predictable valves exhibit the

35 NUREG/CR-5720


S

Staticpressure

Flow -

Staticpressure

Staticpressure

M291 rs-0491-07

Figure 32. Gate valve disc cross section showing pressure forces and measurement locations.

1,000

900

I ' ' I 1

I . . T

U)

a)

800

700

Upstream pressure ; \

…----- Bonnet pressureUnder disc pressure

600

500

400 I .. I . . I . . I . . .I

-10 0 10 20 30 40 50Stem position (% of travel)

60 70 80

M291 rs-0491-11

Figure 33. Gate valve internal pressure distribution.

NUREG/CR-5720 36


largest stem force. Thus, we concentrated ourefforts on this segment of the valve closure cycle.Wedging forces were not considered becausethese forces are not the result of fluid dynamicand frictional effects, but instead depend on theforce capabilities of 'a given operator and on thestructural stiffness characteristics of a specificdisc and valve body.

3.3.2 Development of a Correlation toBound the Stem Force on a Gate Valveduring Closure. Our first effort was to developa relatively detailed free body diagram of the discwhile it is moving in the closing direction (afterthe flow has been isolated but before wedging),with the hope of better understanding the pressureand area terms that affect the stem force.Figure 34 presents the results of this effort and

identifies all the nonsymmetrical forces acting on.the disc. Note that, according to the free bodydiagram, the forces acting on the disc ultimately-react through the valve body seats, which are at aslight angle (for a flexwedge gate valve) relativeto the horizontal and vertical valve coordinatesystem. To account for this slight seat angle so theforces are expressed in values consistent with thedefinition of a traditional friction factor, we foundit necessary to transform the horizontal and verti-cal forces into a coordinate system that is normaland tangent to the valve body seating surfaces.

Following this logic, we theorized the exis-tence of two horizontal forces acting on the disc.The first horizontal force (FUP) is a result of theupstream pressure (Pup) acting on that area of thedisc defined by the mean diameter of the down-stream seating surface. This surface presents a

Fstem

V

M291 rs-04 9 1 -12;

Figure 34. Gate Valve disc cross section showing unbalanced forces just before wedging.

37 NUREG/CR-5720


circular profile in the horizontal plane. The meanseat diameter was selected because it best approx-imates that area of the disc in contact with thecrown of the downstream seating surface overwhich the various pressureforces act. Fup isdefined as

I70D2 \F stem rej 4 PUP~\ stem) (1 1)

_ p {:rD~mean)Fup. = Pp 4 (6)

The fourth force is a result of the pressure in thebonnet region of the valve.(Pup once the flow hasbeen isolated) acting on the area of the discdefined by the mean diameter of the seat cast inthe vertical direction. This area term is the resultof the slight angle of the seat in a flexwedge typegate valve (nominally a 5-degree angle) andresults in an elliptical area over which the pres-sure acts. The major diameter is equal to the meandiameter of the seat; the minor diameter is equalto the mean diameter of the seat times the tangentof the angle the seat makes with the vertical axisof the valve.

Resisting this force (Fup) is a horizontal force(Fdown) from the downstream pressure (Pdown),

also acting on that area of the disc defined by themean diameter of the downstream seating surface

F.o~~~r~ = -PdOH.flQ'nmean)down ~ dovvnkr4 | (7)

,7rD2\em,F~0 = P ~m In tan a (12)

These two forces represent the only horizontalforces acting on the disc and provide a far morerealistic estimate of the horizontal forcecomponent than the orifice area term often usedin the industry equation. The net horizontal forcecomponent (H) can be expressed as

Resisting this is a fifth force resulting from thedownstream pressure (Pdown) acting on the areaof the disc defined by the mean diameter of theseat cast in the vertical direction. This force isalso the result of the slight angle of the seat.

H = F,,p - Fdown (8)

The free body diagram also indicates that thereare actually five vertical forces acting on the disc.The first vertical force is due to the operator and,represents the net stem force delivered to thevalve,

Fbotro Pdo. n(7D4 ) tan aFbottomf = ~down 4 a (13)

The net vertical force component (V) duringvalve closure can thus be expressed as

Fstem = stem load (9) V = Fstem - Fpacting

+ Ftop - Fbottom

- F stem rej

(14)The second force is a result of the resistance of thepacking while the valve is closing. The effect ofthe disc and stem weights is also included in thisterm,

Fpac,, = packing drag - disc and stemweights (10)

The net horizontal and vertical forces can nowbe recast into the plane defined by the valve bodyseat and normalized to remove the effect of valvesize. Figure 35 shows these two forces and theresolution of these into forces normal and tangentto the seats. Note that the two normal forcesresulting from this transformation act in the samedirection, whereas the two tangent or slidingforces oppose each other. These can be expressedas follows for the normalized normal force (Fn)

The third force represents the stem rejectionforce, the result of the pressure under the disc (Puponce the flow has been isolated) trying to expelthe stem,

NUREG/CR-5720 38


Surface sliding

Net valve vertical load

Normal component

iding component

ig component

lormal component

M291 rs-0491-08

Surface normal load

Net valvehorizontal load

Valve disc/seat,interface surface

Si

din

Figure 35. Resolution of horizontal and vertical gate valve disc forces into surface normal and slidingforces.

and the normalized sliding forcerespectively:

H cos a + V sin aF= (3rD2 d

4/

H sina - V cosaFs = {a~(~Dnnean)

4

(15)

(16)

The analysis described above allows us tobetter characterize the normal and sliding forcesacting on a flexwedge gate valve disc just beforewedging. Our next effort was to determine if thePhase I and Phase II flexwedge gate valve testdata supported a relationship between theseforces. From our data base, we extracted the testresults of all predictable valves during the closurecycle when the disc was riding on the seats. Wedid not include the results of any testing if severeinternal damage was evident, such as shown in

Figure 36. However, if the valve exhibitedevidence of internal damage while the disc wasriding on the guides but showed no evidence ofsuch behavior while the disc was riding on thevalve body seats, as shown in Figure 37, weincluded the results.

The Phase II data extraction included testingwith three valves representing two valve sizes(6-in. and 10-in.). The full-flow isolation testsand normal operating flow isolation tests wereused; upstream pressures ranged from 600 psigto 1400 psig, and fluid conditions ranged fromsteam to 400'F subcooled water. However, theresults of tests when the differential pressure wasless than 20% of the upstream pressure, or whenboth the differential pressure and the stem forcewere increasing very rapidly while the disc wasriding on the valve seat, were not included. This isbecause the magnitude and trends of the resultingforces are obscured by relatively low loadings on

39 NUREG/CR-5720

c)

C

-0L.-

L_-

E01).4 -j

Co

20000

10000

0

-10000

-20000

-30000

C0D

0

0'

0

a,

CD

c.MvW

0

0 6 10 15 20. 25

Time (s)

30 35

S211 KGD-0490-04

Figure 36. Gate-valve stem-force history, indicating severe damage to the guides.

40000

30000

20000

o 10000

a) 00L-

LL' -10000

E '.D -20000

U)

-30000

-40000

-50000

-600000O 5 10 16 20 25

Time (s)

30 35z

MM/

0

S211 KGD-0490-10

Figure 37. Gate-valve stem-force history, indicating damage to the guides but not to the seats.

cn

0l

0

tj


the valve disc (which affect the transition fromthick-film lubrication to thin-film lubrication),and by rapid changes in both the stem force andthe differential pressure. This extraction yieldeddata from 30 tests, which are listed in Table 5.

The Phase I data extraction included testingwith two valves representing a single valvesize (6-in.). By way of comparison, these valveswere nearly the same as Valves 1 and 2 in thePhase II testing. Only design-basis flow isolationtests were used from the Phase I testing.Upstream pressures ranged from 400 psig to1400 psig, and fluid conditions ranged from100F subcooled to 140'F subcooled water. Thisextraction yielded data from 12 tests, which arelisted in Table 6.

3.3.3 A Correlation to Bound the StemForce on a Gate Valve during Closure. Wecan now rearrange the previously developed forcebalance and solve for the stem force, based on alinear relationship between the normalized nor-mal and sliding forces. Using Figures 38 and 39,the slope of the solid line (the friction factorbetween the disc and the seat) can be used torelate the normalized normal force (F,) to thenormalized sliding force (Fs) as

F, = -f F, ± C (17)

where

f = friction factor, the slope of linethat relates the normal force tothe sliding force and is equal to

The results of both data extractions were usedin the force balance developed as described aboveand presented in Figures 38 and 39. The resultsreveal two linear relationships between thenormal force on a seat (Fn) and the tangent orsliding force (Fs) necessary to induce motion.One is representative of a fluid subcooling of lessthan 70'F, while the other is representative of afluid subcooling of 70'F or greater. The twodashed lines on either side of the solid linerepresent the limits of the observed data scatter.The tight grouping of the data scatter lends confi-dence that not only can we bound the forcerequirements of a flexwedge gate valve, but wecan also devise a method where the results of lowdifferential pressure flexwedge gate valve testingcan be verified and then the design-basis condi-tions used to bound the maximum stem force.Note also that the dashed lines do not extendbelow a normnalized normal force of 400 lbf/in2.Because of the limited low pressure and lowdifferential pressure data available, and becauseof the postulated friction mechanism, the applica-bility of the INEL correlation is currently limitedto normalized normal forces of 400 lbf/in2 andabove. We are working with selected utilities thathave proven diagnostic equipment to extend theapplicability of the INEL correlation.

0.400, if the fluid is less than70'F subcooled

0.500, if the fluid is 70'F, orgreater, subcooled

C = offset bounding term reflectingthe scatter in the observed dataand is equal to

0 for no offset

50 lbf/in2 to bound the data pro-vided a normalized normal forceof 400 lbf/in2, or greater, existsaccording to Equation (15)(shown as the dashed lines inFigures 38 and 39).

Substituting the horizontal and vertical compo-nents of the normal force [Equation (15)] and thehorizontal and vertical components of the slidingforce [Equation (16)] into the above relationshipyields

V = H (f cos a + sin a) iC C ( )

(cos a - f sin a)

(18)

NUREG/CR-5720 42

Specific Observations I

Table 5. Phase II gate valve test data supporting extrapolation;

Force Pressure(lbf) (psig) Subcooled

Test Stem - --- fluidValve number Step position Stem Normal Sliding Up Delta (OF)

1 25

1 26

6b 18

6b 25

6b 26

6b 1 26

6a 18

6a 25

6a 26

6al 25

6al 26

6c 13

6c 18

6c 25

6c 26

2 18

2 25

2 26

3 25

3 26

1 25

1 26

la .. 25l ,, .

la 26

5 25

5 26

1 25

1 26

la 25

la 26

-7.34

-7.38

-7.10

-7.26

-7.08

-7.22

-7.54

-7.37

-7.14i 0.

-7.07

-7.35

-7.13

-7.08

-7.29

-7.39

*-7.46

-7.06

-7.12

-7.10

-7.27

-7.13

-7.30'1:

-7.17

12929

13578

10477

12173

12453

13654

7880

7859

7703

8688

8866

15626

15046

15798

16452

16083

13096

13079

12751

14512

4355

8628

10804

987

1021

571

888

953

1031

413

550

548

609

633

952

970

1315

1359

1234

1091

1063

793 -

936

247

- 424, 959.6 948.5

-447 1002.4 980.3

-332 887.5 540.6

-404 856.4 851.3

-409 922.4 915.8

-451. 990.6 989.6

-254 601.7 389.7

-260 525.5 526.5

-254 524.7 524.5

-288 585.7 582.8

-294 603.1 606.7

-490 1414.6 907.9

-464 1431.2 927.8

-510 1276.3 1268.3

-533 1317.7 1310.8

-528 1215.7 1186.4

-422 1056.1 1052.5

-424 1027.6 1024.3

*-438 746.4 - 753.2

-496 886.7 890.8

-113 1012.0 237.0

11.5

16.3

95.9

36.2

36.0

48.4

91.9

18.8

10.9

42.5

39.0

124.4

115.4

47.0

39.7

9.5

9.0

8.4

371.7

388.6

17.2

5.9

7.5

7.1

7.3

11.0

9.6

6.9

5.6

7.7

-7.33 ; 11032

-7.18 12235I .. .

-7.23 13281

-7.47 '24756

-7.28 22547,

-7.21 30474- .3 29

-7.33 29045

887 -298 856.1 861.0

942 -386 915.1 909.1

928 -396 894.4 893.8

1144 -435 1117.4 1106.8

1177; -477 1140.4 1136.2

904 -336 879.9 876.8

835 -304 813.0 810.9

1055 -421 1026.3 1021.3

1013 -400 985.7 981.2

43 NUREG/CR-5720


Table 6. Phase I gate valve test data supporting extrapolation.

Force Pressure(lbf) (psig) Subcooled

Test fluidValve number Step Stem Normal Sliding Up Delta (OF)

A 2 5 11550 996 -413 959.3 _-a 10

A 3 5 5347 472 -170 456.3 _-a 60

A 4 5 12884 1040 -470 998.4 _-a 140

A 5 5 17460 1573 -637 1516.3 -a 10

A 6 5 6183 416 -213 397.3 _-a 60

A 7 5 15491 1378 -563 1328.0 -- a 140

A 8 5 7406 557 -258 534.6 _-a 10

A 9 5 7735 593 -270 569.5 -a 60

A 10 5 8183 616 -289 591.1 -_a 140

A 11 5 11743 1047 -417 1009.9 _-a 10

B 2 5 13833 978 -389 938.6 _-a 10

B 4 5 8500 602 -206 580.5 _-a 10

a. Full differential pressure tests were performed; a differential pressure equal to the upstreampressure was used.

Now, substituting the horizontal and verticalforce components [Equations (8) and (14)] intoEquation (18), limiting the final result to boundthe maximum stem force, and rearranging yields

f = friction factor and is equal to

0.400 if the fluid is less than70'F subcooled

OlF + 50 (..D2 a)

F -te = Fv + 62stetn -02

(19)

0.500 if the fluid is 70'F orgreater subcooled.

where

Fv = Fpacing + F stem tej - Flp + Fbottom

(20)

Equation (19) provides a method that can beused to bound the maximum stem force require-ments of a flexwedge gate valve closing againstmedium to high flows and whose operationalcharacteristics have been demonstrated to bepredictable at design-basis pressures and temper-atures. This method will also provide the basis bywhich the results of in situ tests conducted onpredictable valves at less than design-basisconditions can be verified and then the design-basis conditions used to bound the maximumstem force. Note that the correlation applies only

01 = f cos a + sin a

Fh = Fup - Fdown

02 = cos a - f sin. a

(21)

(22)

(23)

NUREG/CR-5720 44

0

c'J.I--C\

4-

_0

-J

0)

c:--C

I 's

N

Eo

Z

-100

-200

-300

-400

-500

-600

-700

tAl

-8000 200 400 600 800 1,000 1,200

Normalized Normal Load (Ibf/in2)1,400 1,600

cz_,

CoPO~0

M455 kgd-1 091-22

Figure 38. Normalized surface sliding force versus normalized surface normal force for gate valves closing against less than 70TF subcooled fluid.

0CD0

0

0

CrV)

CD

:.

Can

z

nMU'

O

0CD

-o0

0

0

Crv]CD

cj

0=3an

.4-1

-Q

0

C:

N

E0z

-100

-200

-300

-400

-500

-600

-700

-8000 200 400 600

Normalized800 1,000 1,200

Normal Load (Ibf/in2)1,400 1,600

M455 kgw-1091-23b

Figure 39. Nornalized surface sliding force versus normalized surface normal force for gate valves closing against 70TF, or greater, subcooledfluid.


to flexwedge gate valve closure against mediumto high flows, flows which will result in a normal-ized normal force of 400 lbf/in2 or greater,according to Equation (15). This correlation hasnot yet been validated for predicting the closingforce requirements when the normalized normalloading is less, but we are working with selectedutilities to extend the applicability of the INELcorrelation.

The INEL correlation cannot be used in theopening direction. The following comparisonbest explains why. Figure 40 represents the stemforce of a design-basis opening cycle of a smallervalve. The results are representative of a valvethat experiences the highest stem force demandsimmediately after unseating and while the disc isstill riding on the seats. A correlation similar tothe one presented here may be applicable for suchvalves, subject to an assessment of the requiredfriction factor. However, Figure 41 representsthe stem force of a design-basis opening cycle ofa larger valve and clearly shows a very differentstem force response. Here, the highest stem forceoccurs when''the valve is approximately 25 to30% open. This is well outside the assumptionsimplicit in the INEL correlation, the NMACequation, or the standard industry equation anddramatizes the influence fluid dynamic effectscan have on valve response. The issue of bound-ing the maximum stem force to open a flexwedgegate valve will be investigated in the future.

3.3.4 Use of the Correlation to Bound theStem Force on a Gate Valve during Clo-sure. To use the INEL correlation to bound theclosing stem force requirements of a flexwedgegate valve, the analyst must first determine if thevalve is of the type whose operational characteris-tics are considered to be predictable. If the valveis predictable, the following information can beused to bound the stem force:

* The mean diameter of the valve body seat(the average of the inside diameter and theoutside diameter of the valve body seatsmeasured in the plane perpendicular to'thestem)

* The angle the valve body seat makes withthe vertical or stem axis

* The outside diameter of the stem

- An estimate of the-maximum packing dragexpected, less the effects of the weight ofthe disc and stem

* The maximum upstream pressure thatwould exist after flow isolation but beforewedging, typically the design-basis pressure

* The maximum differential pressure thatwould exist after flow isolation but beforewedging, typically the design-basisdifferential pressure

* The subcooling of the fluid at design-basisconditions, either less than 70'F subcooledor 70'F or greater subcooled.

With this information, the nef horizontal force(Fh) can be estimated with Equation (22) andEquations (6) and (7). The net vertical force (F,)can then be estimated with Equation (20) andEquations (10) through (13). The angle betweenthe seat and the vertical or stemn'axis can then beused with Equations (21) and (23) to estimateterms associated with transforming the horizontalforces into forces on the disc and seat and finallyinto ai vertical force. Thus, all the terms in Equa-tion (19) can be determined and the maximumstem force necessary to isolate flow at thespecified pressure and differential pressure can becalculated.

By way of example, we will use the upstreampressure and differential pressure recorded afterflowlisolati6n, but before wedging during one ofthe Valve 2 tests, and will bound the stem forcerequirements of the valve using the INEL correla-tion. Similar estimates will also be made using theindustry equation and either an orifice area or amean seat area in conjunction with a standardindustry disc factor of 0.3, as well as a more con-servative industry disc factor of 0.5. The resultsof each estimate will then be compared with theactual stem force recorded during the test.

47 NUREG/CR-5720

z

C)

MJ

la

0

0

0

Cr

en

20000

10000

-0~.a)

0U-

E.2a -10000U)

-20000

-300000 5 6 10 -15 20 25

Time (s)

30 35

S211 KGD-0490-31

Figure 40. Stem-force history of a small gate valve during an opening test.

40000 A VPI V `, 1fid 1,L- IjJ £-:iJ LV- lV H tLJ I 4 alll

.asoooo20000

n 10000

0~~oI

LL -10000 .E2-20000

Cl)-30000

-40000 Actual-------------- Calculated (,u = 0.3)50000 *-- Calculated (mu = 0.5)

-600000 5 10 15 20 25 30 35

Time (s) S211 KGD-0490-94

Figure 41. Stem-force history of a large gate valve during an opening test. 0

0D

k)~_

0 >


The results from Valve 2, Test 3, Step 25 will be used for this comparison. Since this was a cold watertest (400'F subcooled fluid), we will use the 70'F or greater subcooled friction factor in the INEL correla-tion. Pertinent valve information and the pressure and differential pressure recorded just before wedging are

Stem diameter 1.750 in.

Orifice diameter 5.187 in.

Seat inside diameter 5.192 in.

Seat outside diameter 5.945 in.

Packing drag 200 lbf

Upstream pressure just before wedging 746 psig

Differential pressure just before wedging 753 psid

We can use this information in the INEL correlation as follows:

Dmen (seat ID + seat OD) = (5.192 + 5.945) -5569

Dmean 5262 2

Dmean =7 (5.569) 2 24.358Amean 4 4

Astern = *7e - (1-750) - 2.4054 4

FIp Pu~pAmean = (746)(24.358) = 18, 171

Fdown = Pj 0dAnAmean = (746 - 753)(24.358) = - 171

FIop = PupAmeantanla = (746)(24.358) tan5 = 1590

Fbottom = PdowvrAmeantana = (746 753)(24.358) tan5 = -15

F stem rej = PtpAvstem = (746)(2.405) = 1794

Fh = FUP - Fdow:n = 18,171 - (-171) = 18,342

F, = Ppacklng + FstemrejFtop + Fbottom = 200 + 1794 - 1590 + (-15) = 389

01 = f cos a + sin a = 0.500 cos5 + sinS = 0.585

0- = cos a - f sin a = cosS- 0.500 sinS = 0.953

F = F 01 F + 5A0 mean] = 389 + [(0.585)(18,342) + (50)(24.358)] 12,926.02

NUREG/CR-5720 50

Specific Observations 1

We can also use this information with the industry equation as follows:

Industry equation with [Ld 0.3

A orifice _ .r(5.187) 2 21.131orifice - 4 4

_ s____ L. r(I.750)2_

Astem = __m = ( *4 ) = 2.405

F, Y= IAorifice AP + Astem Pup + Fpacking = 0.3(21.131)(753) + 2.405(746) + 200

= 4773 + 1794 + 200= 6767

Industry equation with Ed = 0.5

Ft = idAoriice: AP + Ase mPup + Fpacking = 0.5(21.131)(753)

+ 2.405(746) + 200 = 7956 + 1794 + 200 = 9950

During this test, the peak stem force recordedjust before wedging was 12,787 lbf. Thus, theINEL correlation bounds the actual recorded stemforce. The industry equation using either disc areaterm, with either the standard industry disc factorof 0.3 or the more conservative industry discfactor of 0.5, underpredicts the actual recordedstem force just before wedging. Additional com-parisons are presented in Table '7 for eachPhase II test evaluated. Examining these resultsindicates thatfthe INEL correlation consistentlybounds the maximum recorded stem force beforewedging, whereas the industry equation, usingeither disc area term with either disc factor,typically underestimates the stem force, exceptwith a disc factor of 0.5 and a mean seat area.

3.3.5 Identifying the Pressure Depen-dency Contributing to the Stem Force ona Gate Valve during Closure. Based on theresults of our testing and data evaluation, weidentified both a subcooled and differential pres-sure dependency on the disc or friction factor. Thesubcooled dependency has been previouslyshown; not so evident, however, is the differentialpressure dependency of the correlation. Referringto either Figure 38 or 39, the INEL correlation:follows the lower bounding curve, a higher resis-

tance for a given normal force, or a nominal fric-tion factor plus an offset. This nominal frictionfactor plus an offset can also be reduced to asingle load dependent friction factor. Figure 42displays the friction factor and offset as shown inFigure 38. Also shown on this figure is the load-dependent friction factor at two normalized nor-mal forces. The normalized normal force of400 lbf/in2 results in a 'load-dependent frictionfactor of 0.525. The normalized normal force of1400 lbf/in2 results in a load-dependent frictionfactor of 0.436. This effect is the result of the dataoffset or bounding term and its relative magnitudecompared to the normal force component. Thus,embedded in the INEL correlation is the differen-tial pressure dependency observed in our testdata.

3.3.6 Low Differential Pressure TestVerification and Bounding of the StemForce on a Gate Valve Closing againstDesign-Basis Flows. Utilities have numerousflexwedge gate valves in systems throughout anuclear power plant; many of these valves mustfunction in various design-basis events. The capa-bility of these valves to operate at design-basisconditions usually cannot be verified with in situ

-testing, especially for valves where design-basisconditions include high pressures and medium to

51 NUREG/CR-5720


Table 7. Comparison of gate valve stem forces, estimates versus actual.

Stem force(lbf)

Industry equationPressure

(psig) Orifice area Mean seat areaTest

number Step Up Delta Actual INELa [1=0.3 1[=0.5 [t=0.3 rl=0.5

Valve 21

6b6b6b6bl6a6a6a6al6al6c6c6c6c2223 .3Valve 3

1lala5

Valve 51Ilala

2526182526261825262526131825261825262526

959.6 948.5 129291002.4 980.3 13578887.5 540.6 10477856.4 851.3 12173922.4 915.8 12453990.6 989.6 13654601.7 389.7 7880525.5 526.5 7859524.7 524.5 7703585.7 582.8 8688603.1 606.7 8866

1414.6 907.9 156261431.2 927.8 150461276.3 1268.3 157981317.7 1310.8 164521215.7 1186.4 160831056.1 1052.5 130961027.6 1024.3 13079746.4 753.2 12751886.7 890.8 14512

1012.0 237.0 4355856.1 861.0 8628915.1 909.1 10804894.4 893.8 11032

1117.4 1106.8 122351140.4 1136.2 13281

879.9 876.8 24756813.0 810.9 22547

1026.3 1021.3 30474985.7 981.2 29045

134291384510556b

1218213002139207930b

8092807088069091

16558b

1 6827b

1743417975164551471814135512934b

1 5 0 4 4 b

524911283118351164914105814376

29638277123383932674

8525 125358822 129645765 80527653 112508224 120958859 130434120 57684806 70334791 70105305 77695498 80639372 132189524 13446

11307 1666611681 1722110643 156559415 138659164 134946768 99507982 11747

3180 42057038 107627421 113547292 111558954 137439171 14085

18260 2775316951 2572921116 3216720325 30943

94449771628984779111981844985316529958706086

1023810423125351295011791104341015574978844

334076218037790497049940

20133186832329622419

14067145458924

12623135731464163987883785787109043

146621494418712193361756915564151441116513185

44721173412380121751499315367

30874286153580134434

252625262526

2526 -2526

a. A friction factor (f) of 0.400 was used in the INEL correlation, except as noted.

b. 70TF or greater subcooled fluid test, a friction factor (f) of 0.500 was used.

NUREG/CR-5720 52


0

-100

ro

E

cD

'CDJ

r-

TD

- -

-300

-400

-500

-700

-8000 200 400 600 800 1000 1200 1400 1600

Normalized Normal LoadFigure 42. Gate valve load-dependent friction factor using the INEL correlation.

high flows. Usually, only low flow and low differ-ential pressure conditions can be developed nearvalve closure under in situ conditions. When suchvalves are determined to be predictable by othermeans, the INEL correlation offers a method thatcan be used to bound the stem force requirementsof the candidate flexwedge gate valve at design-basis conditions. This can be accomplished asfollows:

Step 1: Perform a differential pressure test. Usethe results of the testing to estimate thenormalized normal and sliding loads forthe valve, according to Equations (15)and (16). The valve is considered to berepresentative of the valves tested bythe INEL if (a) the upstream pressure

and differential pressure while the discis riding on the seats result in a normal-ized normal force of not less than400 lbf/in2, and (b) the resulting forcesfall within the upper and lower boundsexpected for valves of this design(see Figures 38 and 39). If the result-ing loads do not fall within the expectedband, the results of our testing are notrepresentative of the valve being testedand the INEL correlation may not beapplicable.

Step 2: If the results of Step 1 fall within theexpected band, use the actual design-basis conditions in the INEL correlationto bound the stem force requirement of

53 NUREG/CR-5720


the valve being tested. Then use thismaximum stem force estimate, alongwith the other necessary motor operatorsizing calculations, to verify. the size ofthe operator and the setting of the torqueswitch, and thus ensure that sufficientstem force is available to operate thevalve at design-basis conditions.

3.3.7 Nuclear Maintenance ApplicationCenter (NMAC) Gate Valve Stem ForceEquation. The current NMAC gate-valve stem-force equation is mathematically equivalent to thebest-estimate portion of the INEL math. Like theINEL correlation, it is based on a first-principlesanalysis, but it does not provide the additionalempirical derived guidance that is contained inthe INEL correlation. The NMAC guide providesa range of dynamic coefficients of friction thatmight be encountered in gate valves. Suggestedvalues range from 0.35 to 0.50, but more specificguidance is left to others. Our immediate concernis with the following footnote, presented in theNMAC guide, addressing the applicability anduse of a coefficient of friction:

"Extensive testing . . . using Stellite 21found no statistically significant variation

between static and dynamic friction. Thistesting also found that temperature, fluidtype, and contact pressure had no statisti-cally significant effect on friction."

Our data assessment to date has not addressed theissue of static versus dynamic friction; however,we have clearly shown that fluid type (subcool-ing) and differential pressure have a significanteffect on the friction factor. We suggest that thisstatement be reviewed in light of the INEL testresults.

The NMAC guide also presents their equationas being applicable for estimating valve openingstem forces. One of the basic assumptions statedin the guide regarding the use of their equations isthat

"The limiting thrust occurs at or near theseat, during seating or unseating, and noadditional dynamic flow effects contributeto the required stem thrust."

As mentioned earlier in this report, INEL testresults indicate that some valves experience theirhighest stem forces when the valve is 25 to30% open. Thus, this statement, too, should bereviewed in light of INEL test results.

NUREG/CR-5720 54

4. UNDERSTANDING DIAGNOSTICS AND DIAGNOSTICTESTING OF MOTOR-OPERATED VALVES

4.1 Overview

Recent recommendations made by the NRC inGeneric Letter 89-10 suggest that licenseesdevelop programs for MOV inspection, mainte-nance, and testing under pressure and flowconditions to ensure that the valves will functionwhen subjected to design-basis conditions.Development of the licensees' MOV programs isunderway, but in our opinion, it is obvious thatMOV diagnostic testing will play a large part inthese programs. Diagnostic test equipment andmethodologies are continuing to be developed byboth the diagnostic industry and by selectedutilities. Over the years, we have developed avery sophisticated diagnostic capability to aid inour MOV research. Admittedly, some of thiscapability is needed from a researcher's stand-point and is not necessary for diagnostic in situtesting. However, we have identified from this'larger capability what we believe to be theminimum diagnostic capability necessary forthorough in situ testing of rising-stem MOVs.This section of the report presents our thoughts onmeasuring and assessing the performance ofMOVs during diagnostic testing and then relatesthat performance to 'design requirements.Because MOV diagnostic test equipment is stillbeing developed, some of the judgments made inthis section are based on the capability of equip-ment that was commercially available when thisreport was written.

MOV diagnostic systems should serve the fol-lowing purposes: (a) to identify deterioration. orfaulty adjustment in the valve and operator, (b) tomaintain assurance of design-basis capability thathas been established previously for a particularMOV, and. (c) to provide information to establishdesign-basis capability based on low-pressureandlow-flow tests and/or from prototype tests.MOV diagnostic equipment is used in conjunc-tion with one of two types of tests: unloaded statictesting or testing under flow and pressure loads(dynamic testing). Both position-controlled andtorque-controlled valves will be discussed.

Unloaded static testing of position-controlledand torque-controlled valves provides mea-surements of the packing drag and stroke time..For torque-controlled valves, the static test alsoprovides measurements of torque and stem forceat torque switch trip. However, the directdetermination of the capability of an operator atdesign-basis conditions from an unloaded test ofeither a position-controlled or a torque-controlledvalve is highly suspect. These data can be used fortrending the MOV.

Flow and pressure testing provide much moreinformation on both torque-controlled and posi-tion-controlled valves than does no-load statictesting. It is well known that some valves cannotbe design-basis tested in the plant, but the higherthe in situ test loading, '-the more in-depthinformation the analyst will have to determine theperformance of the valve and the margins avail-able in each of the MOV components. Again,more information will come from the torque-controlled valves than from the position-controlled valves. For critical position-controlledvalves that cannot be significantly loaded in situ,it may be necessary to consider dynamometertesting of the motor operator.

The thoughts presented here are a product ofour experience in the four full-scale test programs,discussed in Section 1. The recommendations arealso based on the results of special effects testingbeing conducted at the INEL on the Motor-Operated Valve Load Simulator (MOVLS),described in Section 4.3. Our research to dateindicates that, for rising-stem MOVs, ideally adiagnostic system should be able to accuratelymeasure the following parameters as a minimum:position of the motor operator switch (for limitswitches and torque switches), motor current andvoltage, motor operator torque, and valve stemforce. The following discussion explores each ofthe recommended measurements and explains itsvalue to the analyst.

55 NUREG/CR-5720

Understanding Diagnostics

4.2 Understanding MOVDiagnostic Testing

4.2.1 Motor Operator Switch Position.

4.2.1.1 Limit Switches. For ac- anddc-powered position-controlled valves (includingthose valves that close on torque but open onlimit), the analyst can determine valve stroke timeand stem position in either a static test or adynamic test. The point in time when the switch isactivated provides the basis for analyzing manyof the other diagnostic measurements. Figure 43shows typical dc limit switch histories.

4.2.1.2 Torque Switches. Torque switch trip(for torque-controlled valves) establishes a timingpoint to assess the output of the operator. The stemforce at torque switch trip will typically be lowerthan the peak force produced after torque switchtrip, when the motor controller dropout time andthe motor and operator momentum result in furthercompression of the stem. The motor controllerdropout time for both ac- and dc-powered MOVsis small, but its effects are real; the length of thedropout time is further influenced by the state ofthe ac cycle when the torque switch trips. Momen-tum effects are conspicuous in static tests, but theycan be absorbed by the higher loadings in dynamictests. In addition, if the torque switch trips beforethe disc is fully seated in a dynamic test, then themomentum does not produce the additional stemforce that would have been produced in a statictest. Figure 44 is a torque switch history showinga typical ac-powered motor controller torqueswitch trip.

For a torque-controlled valve operating in theclosing direction, the exact trip time (forac-controlled circuits this must be interpretedfrom the motor controller holding coil ac sinewave) is important for analysis of all the otherquantitative measurements. It is important that thedata acquisition system have a sample rate that isfast enough to allow the analyst to correctly deter-mine the time of torque switch trip and all otherdata with respect to this time. The faster the samplerate, the less conservatism the analyst will have toadd to the measurements to account for timeerrors. We have found that a sample rate of

1000 samples/second is usually sufficient; itprovides a resolution that is faster than most of thesensor responses.

4.2.2 Motor Current.

4.2.2.1 Torque-Controlled AlternatingCurrent Motor Operators. The margin of anelectric motor can be estimated from a motor cur-rent history. The concern is that the output from anac motor can change quickly when the motor isoperating too close to stall current, where a largeincrease in required current and a large decrease inmotor speed correspond with only a small increasein torque (in some cases, no increase at all). Todetermine if the electric motor may have prob-lems, compare the peak motor current just beforetorque switch trip with the motor torque/speedcurve for the specific electric motor to help esti-mate electric motor margins. By way of example,Figure 45 is a current history for a valve closingagainst a design-basis load; Figure 46 is the acmotor torque speed curve for the operator motor.Comparing the current drawn in Figure 45 withthe motor torque speed curve, even though thecurrent transducer is nearly saturated on thisstroke, we see that the motor is operating well outon the knee of the motor torque speed curve. Fig-ure 47 shows the motor current during the nextvalve closure against a design-basis flow load.During this second test, which was performedwithin five minutes of the test shown in Figure 45,the motor went into a stall. [Careful examinationof the motor current trace from the previous cycle(Figure 45) would have alerted the analyst to theimpending problem.] Motor heat and voltagedrop, in conjunction with a marginally sized elec-tric motor, were the primary causes of the stall. Ifthe analyst cannot obtain a motor torque speedcurve for a specific motor, it may be necessary toconduct dynamometer testing to produce such acurve to determine motor margins.

4.2.2.2 Position-Controlled AlternatingCurrent Motor Operators. The analyst mustsubject the position-controlled valve to its design-basis loading, by in situ or prototype testing, tomake an analysis that is equivalent to thatpreviously described for torque-controlledvalves. Again, the analyst may want to considerdynamometer testing of the motor operator.

NUREG/CR-5720 56


2.5

2

. . . . . . . . . . . . . . . . . . . . . . . . . .

U)

t,,

1.5

1

Open limil- --- Close limi

Total stroke time i-- for a position-n'

controlled valve

i.I

0.5

0

-0.5 I I , I i' *. I . . I . I l I I . . . I l l

0 2 4 6 8 10 12 14

Time (s) S314 KGD-0291-09

Figure 43. Limit switch history showing stroke time for a position-controlled valve. Limit switchactuation points are important when analyzing other data.

0.5

0.4

=: 0.3

.= 02C,,

CD 0.1

0-

4 -a) -0.10

C3 -0.2

-0.3

-0.4

-0.511.70 11.72 11.74 11.76 11.78 11.80 11.82 11.84 11.86 11.88 11.90

Time (s) S314 KGD-0291-10

Figure 44. Torque switch history showing torque switch trip and the termination of current flow to themotor controller holding coil.

57 NUREG/CR-5720


Valve 5, Test 1A, Step 25, 1200 psi, 5700F (Steam)25

D_

E-C

C

-1Co

0

0

20

15

10

- (note: current transducer maximum range - 20 amps)

5

00 5 10 15 20 25 30

Time (s)

35

S314 KGD-0291-01

Figure 45. Current history from a design-basis test showing current increasing as the valve closes; therapid increase in current indicates torque switch trip. The test began with the valve 75% open.

2000

1600

a1)

m) 1000CL)coL_.

o

0

500

0

60

40

ECo

30Co

C.)20

0

+1

10

10

0 10 20 30 40 50 60 70

Motor torque (ft lb) S314 KGD-0291-11

Figure 46. Motor torque-speed curve showing the speed-torque-current relationship for the ac motor testresults shown in Figure 45. Beyond the knee of the curve, the torque increase is small in proportion to thespeed loss and the current increase. For some motor configurations there is no increase;, for some there isactually a decrease in torque beyond the knee of the speed curve. (Arrows indicate the applicable axis.)

NUREG/CR-5720 58


Valve 5, Test 1A, Step 26,1200 psi, 570'F (Steam)25

E-w

0)~: 4-

CCD

:t0

0

201

15

10

(note: current transducer maximum range - 20 amps)

5

0 ,,,, . . . .n . . I . . . . . . . . . . . . . . . . . . . . . . . .

0 5 10 15 20 25 30 35

Time (s) S314 KGO-0291-02

Figure 47. Current history from a design-basis test showing the motor going into a stall, saturating thecurrent transducer. The test began with the valve 30% open.

NOTE: All other readings of motor current forboth torque and position controlled valves aremade for trending. Additionally complicating thesubject, Limitorque has publicly informed theindustry that ac motors of a given model operat-ing under the same conditions can vary up to 20%in their output.

4.2.2.3 Direct Current Motor Operators. Asin the previous analysis, the dc MOV must beloaded before torque or limit switch trip. The ana-lyst can then compare the motor current to themotor torque speed curve to determine how farout on the curve the motor is operating. Figure 48shows that, unlike ac motors, dc motors continueto produce torque linearly even as they approachstall conditions; however, at the lower motorspeeds that accompany the higher operating cur-rents, the effects of motor momentum are less. Asa result, the lower motor speeds produced underhigh loads can affect the stroke time, as shown inFigure 49. In addition, dc motors operating athigh loads can heat up very quickly, which in turnreduces their output torque. Figure 50 shows thedrop in output torque over time for a 40 ft-lb dc

motor as it heated. Degraded voltage (because ofline losses) can also reduce the output of themotor. As the motor turns at lower speeds, theoverall efficiency of the operator is reduced. Fig-ure 51 shows the results of these effects. The testat the lowest loadings closed the valve, and themotor tripped out on the torque switches. Thetests at the next two higher loadings did not closethe valve, but the motor was able to generateenough torque to trip the torque switch. However,the test at the highest load not only did not closethe valve, but because of the efficiency losses inthe motor circuit, it was also unable to generateenough torque to trip the torque switch andtherefore stalled the unit. These factors, workingseparately or together, can thus produce motorstall and motor burnout.

4.2.3 Motor Voltage. Voltage drop under loadcan alert the analyst to undersized cables in thepower circuit. It has been found on a number ofoccasions that some plant circuit designers didnot realize that both ac and dc motors can operateat four to five times the running current stamped

59 NUREG/CR-5720


3,500

3,000

0.E

_0

a)a)

CL

V)

2,500

2,000

1,500

1,000

150

12 ~ 0125 a)

a)

E100 .-

4-.'C:

.. a)

75 '0

50 *2

2

25

60

MllI kod-0592-01

500

00 10 20 30 40 50

Load (lb - ft)Motor torque

Figure 48. Motor torque-speed curve showing the speed-torque-current relationship for a dc motor. Thetorque (load) continues to increase as the speed drops and the current increases. (Arrows indicate theapplicable axis.)

8

6

C:

C:0

CL)00.

6-

4

2

00 5 10 15 20 25

Time (s)

M248 kd-O291-02

Figure 49. Disc position histories from three dc-powered closing tests showing that the stroke time islonger with higher loads.

NUREG/CR-5720 60

Understanding Diagnostics'

150

125

-6

ECu

ci,

0;

100

75

50

2 5 F 1.l

0, 10 20 30 40 50 60Load (ft - lb)

248 kygd-3O1-05

Figure 50. Motor heating under load caused the current demand of this dc motor to decrease at a rate of1 amp per second during the 20-second test, resulting in a decrease in the output torque of approximatelyI ft-lb every 2 seconds.

8

6

0

0.C.,03Cn

.4

2

00 5 10 15 i 20 -25

Time (s)

.. 248 kgd-02S 1I-0

Figure 51. Disc position histories from four dc-powered MOV closing tests, showing that the stroke timeis longer with higher loads. Three of the tests did not fully seat the valve, and the test at the highest loadstalled the motor.

61 NUREG/CR-5720


on the motor nameplate, and therefore specifiedless than adequately sized power cables. Not longago, a utility traced several burned out dc motorsto this undersized-cable problem. Limitorque haspublicly stated on a number of occasions that, forthe small number of dc motor-operated valves inthe industry (compared to the large number of acmotor-operated valves), the percentage ofreplaced dc motors far exceeds that of ac motors.This increased incidence is probably a result ofthe lower dc voltages and the associated highercurrent to perform the same work as the highervoltage, lower current ac motor.

It should also be noted that in the typicalLimitorque dc motor wiring circuit, it is impossi-ble to accurately determine the voltage drop in thecircuit from one measurement. There are typi-cally four wires that conduct the motor currentand interconnect the motor controller and themotor. Two of the wires interconnect the armatureand the series field; this interconnection providesthe ability to reverse the direction of the motor.Since the resistance of each wire contributes tothe voltage drop, any single measurement ofmotor voltage will be contaminated by a voltage.drop caused by a minimum of two wires.

Previously unexplained motor burnouts may bethe result of excessive voltage drop in the powercircuit; a voltage check at the highest valveloading possible may provide the explanation.The voltage check will also aid the analyst whoneeds to know the actual line losses at normalvoltage conditions so that the additional losses atdesign-basis degraded voltage conditions can becalculated.

4.2.4 Motor Operator Torque. Load-sensitivemotor-operated valve behavior (rate of loadingexplained in Section 4.3) and the need to establishindividual valve performance margins providethe incentive to measure motor operator outputtorque. Some of the new-generation transducersmeasure motor operator output/stem torquedirectly. Where these new transducers are notavailable, special-effects testing at the INEL isproviding an increasing data base showing thatthe relationship between spring pack force or

deflection and output torque is very constant andnot affected by the loading on the operator.Figures 52 through 54 show that the ratio ofspring force divided by stem torque (the torqueapplied to the stem by the stem nut) does not varyover a wide range of loadings. Figure 52 is froma lightly loaded test with a low torque switch set-ting. Figures 53 and 54 are from tests withhigher loads and with a high torque switch set-ting. (The stem force histories from those twotests are shown in Figures 55 and 56.)

Motor operator torque measurements are usedin determining the stem factor (operator torquedivided by stem force) part of the margins assess-ment for a specific valve and provides a secondreference for the indirect- or direct-forcemeasurements made with some of today's stemforce measurement transducers. To reduce thepotential errors in indirect torque measurements,the analyst should obtain detailed calculations forspring pack force or deflection versus operatortorque. The spring pack should be calibrated todetermine any offset from the publishedLimitorque spring constant for each spring packassembly. Figure 57 shows a spring packcalibration. Using this kind of information andthe specific operator moment arm, the analyst canobtain reliable values for the operator outputtorque from many of the commercially availablespring force or deflection transducers. Torquedetermined from static testing will not provideuseful information for stem factor determination.Both torque- and position-controlled valves needto be tested with a flow and pressure load toproperly load the stem-stem nut interface.

4.2.5 Valve Stem Force. Stem force in bothposition- and torque-controlled valves can best bedetermined from measurements. While it ispossible to estimate the stem force if the outputtorque of the operator is known, our experienceshows that the stem factor varies significantlywith load. However, if the analyst knows the oper-ator torque and the stem force, the stem factor canbe determined by the operator torque divided bythe stem force. Note that the operator torque andstem force measurements must be obtained justbefore the valve hard seats; after the valve startswedging, there is almost no motion between thestem nut and the stem, so no useful value can be

NUREG/CR-5720 62


10 f -I l I I I I I I i Il i I I jI I I l l l I .

MOVLS Test 1.. , .... I., ......I I

a):30-

0

Ea)

a)00

0)

IaCo

a,:30-

0FM

9

8

7

6

5

4

' 3

2

1

,Torque swit(trip

,h

. II . . . , I . .*'. . I . .. IQ L

C....................

1 2 3 4 .5

Time (s)

I . . . . I . . . . I ... . I I , I 1 -i

6 7 8 9 10S314 KGD-0291-03

Figure 52. Ratio of torque spring force to stem torque from MOVLS Test 1, with a small stem force.

10 .11 ........ I-- ........................ I.......... . I . . . . I

a,:3

074-

Ea)

Co

04-

0)

C',

0or

MOVLe Tst4 49

8

7

6

6

4

3

2

1

I.... I .... I....

Torque -switch -t ,I-

,

I 0 1.1-1.1-1 ....

0 1 2 3 4 5 6 7 8 9 10

Time (s) S314 KGD-0291-04

Figure 53. Ratio 'of torque spring force to stem torque from MOVLS Test 4 at a higher stem force.

63 63 ~~~~NUREG/CR-5720


100)

cr0

a)4-

a)E20

.-_0)0)

0-4-

C',

0)FL

9

8

7

6

5

4

3

I I I . I I I I I I I I I r . I I ' I I I T r I I I I I . . . I I I . . I I I I I I . I . -.MOVLS Test 5

Torqueswitch/ trip

2

I I . 1 , .I I I i I i i r I ,~~~~~~~~~~~~~~~~~~~

0

4 6 6 7 8 9

Time (s)10 11 12 13 14

S314 KGO-0291-05

Figure 54. Ratio of torque spring force to stem torque from MOVLS Test 5, at a high stem force, inwhich final seating was not achieved, showing that the average torque spring force to stem torque ratio wasnot influenced by the load on the operator.

0

-5000

_ -100000)

0*° -15000E0)

u) -20000

-25000

-30000

I . I I I . r I I I . I I I I I I I I I I I I I I I I I I I I r I I I I I I I I I . I I I I I

: ~~~~MOVLS Test 4

> Seating

Margini

0 1 2 3 4 5 6Time (s)

7 8 9 10

S314 KGO-0291-07

Figure 55. Stem force history from MOVLS Test 4, showing a 3000 lbf margin between the peak stemforce and the stem force at final seating.

NUREG/CR-5720 64


0

-5000

- 100000)

0*- -16000

U) -20000

-25000

-300004 56 6 7 8 9 10 11 12 13 14

: Time (s) S314 KGD-0291-08

Figure 56. Stem force history from MOVLS Test 5, showing that a 1000 lbf increase in the initial stemforce eliminated the 3000 lbf final seating stem force margin observed in MOVLS Test 4.

0U-

-i--, . Deflection (in.)I .1I I

S314 KGD-02911-08.

Figure 57. Typical torque spring calibration plot shows the initial force offset at zero deflection.,Thisfigure also confirms that forthis spring, the spring force to deflection relationship is linear. The actual valuesare omitted from this figure because the data are proprietary until the MOV diagnostic equipment validationis complete.

65; NUREG/CR-5720


obtained for the stem factor. After the analyst hasdetermined the stem factor and how well theconversion of operator torque to stem force is tak-ing place, extremely good or bad stem factors canalert the analyst to either faulty torque or forcemeasurements or to problems at the stem-to-stem-nut interface. Limitorque selection guides (SEL)provide stem factors at several coefficients offriction for most popular stem diameter, threadpitch, and lead combinations. Accurate stem forcemeasurements can also alert the analyst to exces-sively high packing loads. Packing loads that arehigher than allowed by the design calculations canadversely affect the valve's ability to functionwhen pressure and flow loads are added. Currentand torque measurements can also alert theanalyst to high packing loads, but cannot quantifythe actual packing drag.

For torque- and position-controlled valves, thestem force measurement also can be used to deter-mine the available stem force margin if the valveis tested at design-basis flow and pressure loads.Stem force values obtained from less than fullyloaded tests cannot be used to establish fullyloaded stem force margins. Figures 55 and 56show why this is so. Figure 55 is a stem force his-tory from an MOVLS test, showing a stem forcemargin of approximately 3000 lbf (see Section 4.3for a discussion of the MOVLS). Figure 56 showsthe stem force history from an MOVLS test withan additional 1000 lbf initial loading that resultedin no seating margin. A stem force margin of3000 lbf was wiped out by a 1000 lbf increase inthe initial stem force loading because the stem fac-tor increased with the increase in load. The stemfactor increased 6% during the second test. Thisphenomenon is known in industry as the rate-of-loading effect, and is discussed in the next sectionas a load-sensitive behavior.

Although not useful to establish stem force mar-gins, stem force measurement data from unloadedtests on torque-controlled valves can be used fortrending. Changes in the measured stem force canalert the analyst to problems with the valve or withthe measurement system. Stem force measure-ments from unloaded position-controlled valveswill provide the analyst with little useful informna-tion, as the only'load in the stem is the packingload.

Inaccuracies in the stem force measurementsdo occur; this affects the reliability of data fromtesting either torque- or position-controlledvalves. Valve stem force transducers come inmany types and applications. The accuracy ofstrain gages depends on accurate characterizationof the materials involved, the expertise of theinstaller, the quality of the installation technique,and in some cases the accuracy of the in-placecalibration. Removing the stem from the valveand performing a calibration of the strain gages ina tensile machine can determine their accuracy.The accuracy of stem collar transducers alsodepends on extensive characterization of thematerials involved and the quality of theinstallation. All of these instruments measurevery small reaction loads in materials that aretypically specified only by type. Without sometype of calibration, the analyst must choose eitherthe middle of the material's properties range or aconservative high or low bound, depending on thecalculation. All of these problems are magnifiedwhen the stem force measurement must be madein the threaded portion of the stem. Using straingages installed on the yoke can result in acombination of these problems.

Placing load cells between the valve yoke andthe motor operator requires that the attaching boltmetallurgy and the structural stiffness of the yokeflange be known. The structural stiffness of theyoke flange also affects the calibration, if strainbolts are used to attach the yoke to the motoroperator. Strain bolts without other intrusivedevices will measure the stem force only in theclosing direction.

All of the above technologies measure stemforce with varying degrees of accuracy; the accu-racy depends on the degree of characterization ofthe geometrical and metallurgical considerationsand the quality of the installation. The sensiti-vities and overall accuracies of all the currentcommercial diagnostic torque and thrustmeasuring systems will be better known after theNRC/Motor-Operated Valve Users Groupcompletes its validation testing and evaluation.

NUREG/CR-5720 66


4.3 Load-SensitiveMotor-Operated ValveBehavior

The MOVLS in its simplest configuration wasbuilt to help develop our field test instrumentationand data acquisition equipment. Following ourPhase II gate valve testing, the MOVLS wasimproved to perform special effects testing. Itwas again improved for the diagnostic equipmentvalidation program. Its current configuration,shown in Figure 58, consists of a Limitorqueoperator bolted to a valve yoke, which in turn isbolted to a hydraulic cylinder. The cylinderdischarges to an accumulator, and the operatorresistive loads are varied by controlling the initialfluid level and initial gas charge in the accu-mulator. The output thrust from the operator istransferred to the hydraulic cylinder through avalve stem, a thrust bearing, and a load cell. Atorque arm is bolted to the valve stem and servestwo purposes: it acts as an anti-rotation deviceand is instrumented to measure the torque in thestem.

Instrumentation used on the MOVLS includes3-phase current (both rms and peak to peak),rms voltage, 3-phase power and power factor,motor speed, open and close limit switch posi-tions, torque switch position and trip signals,spring pack position and force, and the stemforce, torque, and position. The data acquisitionsystem typically runs at 1000 Hz, resulting in a1-ms time resolution. Loadings in the stem can bevaried from a no-load condition to design-basis-type loads. Stem thrust histories obtained fromour MOVLS compare very favorably with stemthrust histories obtained from full-scale valvetests we obtained in the field. The MOVLS pro-vides an economical means of producing realisticvalve loadings in the laboratory and allows us tostudy operator-related phenomena that we haveobserved in our full-scale valve tests.

Load-sensitive MOV behavior describes thephenomenon in which the maximum output thrustof a rising-stem MOV at torque switch tripdecreases as the load on the stem increases. This

phenomenon has been observed throughout theindustry and occurred on two occasions duringour full-scale valve test programs. Although welacked sufficient instrumentation and analyses topinpoint all of the possible causes of this phenom-enon at the time it occurred, we were able todetermine that it occurred when the torque switchwas not set high enough to fully seat the valve.Since then, we have been able to use the INELMOVLS to perform the separate effects testingnecessary to isolate the first-order cause of thephenomenon.

Using the MOVLS, we performed a series of17 simulated valve closings. The first seven weresetup tests and will not be part of this discussion.The 10 tests that will be discussed are shown inFigures 59 through 61, which show the thrusthistories for each of these tests. Each of these testshad the same torque switch setting. The only vari-able from test to test was the stem load applied bythe simulator. The first three tests shown inFigure 59 are low load closures, typical of whatcould be obtained during inplant testing while thevalve is under pressure with no flow through it.These low load tests were followed by the fourtests shown in Figure 60. During these tests, theclosing load was increased before Test 11 andagain before Test 12. Thereafter, the closing loadwas not increased. Test 11 is representative of avalve closing against a pressure and flow load andshows that increasing the stem load results in asmall decrease in the thrust margin. The thrustmargin is the difference between the thrust at thetime the valve seats and the thrust at torque switchtrip. During Test 12, the stem load is againincreased and results in a larger decrease in thethrust margin. For comparison purposes, Test 12is equivalent to slightly less than the design-basisloading of a typical predictable 6-in. BWRRWCU flexwedge gate valve.

Tests 13 and 14 represent two additional testsat the same stem loading .as Test 12 and showthat, with a constant load, the thrust margin con-tinues to decrease. Test 13 has effectively nothrust margin and the simulator was just able toseat. Test 14 shows a negative thrust margin andthe simulator does not seat. At the full design-basis loading, we have been able to create the

67 NUREG/CR-5720


/z64 Ac 'OO -/O 4L4i -. - '-.': CtOSc6D PoS* r/27a

,~~~~~~~~~~~~lj .K~ 5( I6A/rOe~5

Ot leQ6e. 4011161

/l O, ,;z ' r 4

.a~~~~~A e _, S°Att f

ZD4XZ / 4'sUT 16d1/6r4eAYdt ,,, _ .U *. . 1o.~ CdD f// A

L44-Figure 58. MOVLS layout drawing.

NUREG/CR-5720 68


-o

I-E0)

Cl)

0- -2,000

-4,000

-6,000

-8,000

-1 0,000

-12,000

-14,000

-1 6,000

-18,000

-20,000

-22,000

-24,000

-26,000

M450 rs.1 091-09

Figure 59. Traces of thrust versus time for Tests 8 through 10 of the MOVLS test sequence. These areidentical low-thrust tests.

Z-o

(n

Ea)

Cl)

0

-2,000

-4,000-6,000

-8,000

,-1 0,000

-12,000

-14,000

-16,000

-18,000

-20,000

-22,000

-24,000

-26,000

Test 10

r~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Test 11 Test 12 Test 13 Test 14

lVargin

. . Margin okar

r : Ad ' ~~~~Margin

... I I I �...........

.9Margin not- | T : seat

A Torque switch trip

I

M450 rs-1091-10

Figure 60. Traces of thrust versus time for Tests 10 through 14 of the MOVLS test sequence.

69 NUREG/CR-5720


-o

Cl)

PE)CO

0

-2,000

-4,000

-6,000

-8,000

-1 0,000

-12,000

-14,000

-16,000

-18,000

-20,000

-22,000

-24,000

-26,000M450 rs- 1091-11

Figure 61. Traces of thrust versus time for the final three tests of the MOVLS test sequence. Tests 12-14,which exhibit load-sensitive MOV behavior, were followed by a series of low-thrust tests (Tests 15-17),with resistance to closing much like the first three of the series.

Test 14 response in a single step followingTest 11. However, we went through the transitionfrom normal stem behavior to abnormal stembehavior so quickly that we were not positive ofthe root cause. Performing the test at a lowerloading, and repeating it, provided further insightabout the cause of load-sensitive MOV behavior,as we will discuss later in this section. We areaware that three near-design-basis closures backto back are not a normal design requirement.

We expected this load-sensitive MOV behav-ior, based on the setup tests. Following the abovetests, we reduced the stem load to the level ofTests 8 through 10 and repeated the closuresthree times. Tests 15, 16, and 17 are shown inFigure 61. We observed that the thrust marginduring Test 15 was less than the thrust marginduring Tests 8, 9, 10, 16, and 17. This was unex-pected, but it did indicate that the MOV load-sensitive behavior was the result of a change insome parameter that we should be able to isolate.

To better understand this phenomenon, and tobe sure it was not a combination of causes,kinematic and kinetic evaluations of the motor-operated valve were performed. Based on theseevaluations, we identified potential first-ordercauses of this phenomenon. Using measured datataken from the MOVLS testing discussed above,we then evaluated the likelihood of eachcontributing to the load sensitive MOV behavior.

The kinematic evaluation of the motor operatorrevealed that it functions like a planetary gear.There is one input path for motion, but there aretwo possible paths for the output motion, asshown in Figure 62. The motion input to the unitis the result of the electric motor; the two possibleoutput motion paths are at the stem and the springpack. Like any planetary gear, the geometry of theoperator (the diameter of the gears, the number ofgear teeth, etc.) does not completely determinethe output motion of each path. Therefore, thekinetics of the operator were considered for moreinsight into the behavior of the unit. We learned

NUREG/CR-5720 70


Spring packoutput m11otiOnl

Motor inputmotion

1I

Stem Outputmotion

Figure 62. The design of the operator allows two potential output motion paths.

that the relative resistance to motion of each pathat any given instant determines how the inputmotion is distributed between each of the possibleoutput motion paths. 'At very low stem loads, theentire input motion is transferred to the stem.Motion of the spring pack'does not occur untilthere is enough resistance to stem motion(observed as an increase in the stem thrust) toovercome the preload of the spring pack. Oncethe preload of the spring pack has been overcome,increasing the resistance to motion at the stemcauses motion in both the spring pack and thestem. If the resistance to motion in the stembecomes too high (high valve closure loads orvalve seating), then all the input motion will betransferred to the spring pack.

Perhaps this conversion of input motion to out-put motion can best be visualized with examplesof a lightly loaded MOV (Test 8), a moderatelyloaded MOV (Test 11), and a heavily loaded

MOV (Tests 12-14). In a lightly loaded MOV,the two output motions occur, but they occur inseries, one after the other. Prior to seating, theinput motion to the worm drives the worm gear,resulting in motion of the stem but 'no'motion inthe spring pack. After the valve seats, motion canno longer occur in the stem, so all the inputmotion is transferred to the spring pack' until thetorque switch trips. In this situation, all the kineticenergy of the operator is transferred to the stem asthe valve seats and results in operator torques andstem forces that exceed the values expected forthe setting of the torque switch. Increases in stemthrusts after torque switch trip are the result ofmotor controller, dropout time and post-torqueswitch trip kinetic behavior. The torque switchdoes not interrupt the electrical energy to themotor, it interrupts the power to the motor con-troller contact holding coil. Motor controllerdropout time includes the collapse time of the coil

71 NUREG/CR-5720


field and the mechanical time for the springs toopen the contacts.

The response of a moderately loaded MOV candiffer significantly. If the resistance to stemmotion is large enough, the increasing torquelevels in the operator can overcome the springpack preload before the valve seats. Output motionwill then occur simultaneously in the stem and inthe spring pack. Because there are now two simul-taneous output motion paths, the increased fric-tional losses cause a slight reduction in the overallefficiency of the unit. As the moderately loadedMOV seats, the motor controller dropout time andthe kinetic energy of the operator are transferred tothe stem, much like the lightly loaded MOV case.This results in stem forces that exceed thoseexpected for the torque switch setting.

The behavior of an MOV that is so heavilyloaded that the torque switch trips before thevalve seats is similar to that of a moderatelyloaded operator, except that the stem is not rigidlyrestrained because of valve seating. Conse-quently, the motor controller dropout time and thekinetic energy of the operator are dissipatedbecause motion of the stem following torqueswitch trip and the high stem thrusts of a rigidbody are not developed.

The above discussion of MOV behavior andcareful analyses of the motor operator internalsgenerated a number of potential candidates for thefirst-order cause of load-sensitive MOV behavior.These candidates represent one of two types ofeffects: frictional losses of the operator mecha-nism or inertial effects of the operator mechanismresulting from the stoppage of motion. We used theMOVLS-generated data to investigate each candi-date and determine the single most likely cause ofthe observed load-sensitive behavior. To do so, wegrouped each of the inertia and friction effects intoone of four categories, based in part on their pos-sible influence on motor operator behavior and inpart on the type of data available from existingMOVLS instrumentation. The four categories are(a) effects that increase the load on the motor butdo not influence the interaction between the torqueswitch and the stem, (b) effects that influence the

spring pack force at torque switch trip, (c) effectsthat influence the spring pack force needed toachieve the same level of stem nut torque, and(d) effects that influence the amount of stem nuttorque necessary to achieve the same level of stemthrust.

Based on MOVLS testing and our full-scalevalve test programs, load-sensitive MOV behav-ior was not observed to be associated with motorstall. Therefore, effects associated with the firstcategory are not related to the load-sensitivebehavior. Effects from the electric motor slowingdown will be presented in the dc-powered MOVsection.

Small inertial and large frictional possibilitiesare associated with the second category that couldcause a change in the spring pack force at torqueswitch trip. These include inertia of the springpack components and friction among them. Varia-tions of spring pack force at torque switch tripwould cascade through the mechanism and varythe operator output torque and stem thrustachieved at torque switch trip. We investigatedthis possibility.

Inertial and frictional effects associated withthe third category could cause a change in theratio of spring pack force to output torque. Input.torque to the worm gear/drive sleeve/stem nutassembly is generated by application of the springpack force at a moment arm distance from thestem centerline (see Figure 63). Elastic effects(worm shaft bending, for example) could changethis distance a bit under varying spring packforce, resulting in a change in the ratio of spring,pack force to output torque. Other effects in thiscategory include friction between the worm shaftand worm, a component of the friction forcebetween the worm and worm gear, and friction inthe drive sleeve bearings. Inertial effects that fall,into this category include those associated withacceleration of the worm or deceleration of the.worm gear/drive sleeve/stem nut assembly. Theseadditional effects are depicted in Figure 64. Thesummed result of all the third category effectswould be an instantaneous difference between thetorque input to the mechanism by the spring pack

NUREG/CR-5720 72


Mean radius,

Spring pack force

Stem torque

Figure 63. The application of spring pack force to the operator mechanism at an offset distance from thestem centerline generates the torque input to the mechanism.

Drive sleevebearing friction, Deceleration of ilie

stem nut, sleeve,and worm gear

Acceleration ofIthe worn

Figure 64. Potential causes of torque loss in the operator are identified.

73 NUREG/CR-5720


force and that output by the mechanism to thestem nut. Such a difference could cause load-sensitive MOV behavior. We investigated thispossibility.

There is only one inertial or frictional effect inthe fourth category, an effect that increases theoutput torque required to achieve the same stemthrust. This effect would be seen in the stemfactor, defined as the ratio of output stem torqueto stem thrust. The industry uses the traditionalpower thread equation to estimate the stem factor:an equation based on the stem diameter, thethread pitch and lead, and the coefficient offriction between the stem and stem nut threads.This power thread equation does not include anyinertial terms, and none are required. All of theparameters in the expression, except for thecoefficient of friction, are obviously invariantbetween a lightly loaded and a heavily loadedMOV. This leaves only the friction between thestem and the stem nut threads as this category'spossible cause of the load-sensitive behavior. Wealso investigated this possibility.

Up to this point, we have identified and catego-rized potential candidates for the first order causeof load-sensitive MOV behavior. We then useddata from each of the MOVLS tests describedearlier to evaluate the likelihood of eachcandidate causing the observed load-sensitivebehavior. In addition to acquiring stem thrustdata, spring pack force data, and stem torque data,we were also able to calculate an instantaneousratio of spring pack force to stem torque, the ratioof stem torque to stem thrust (stem factor), andthe coefficient of friction between the stem andthe stem nut. All calculations were made on areal-time basis, using measured data and valveand operator design information.

The possibility of inertial or frictional effects inthe spring pack causing load-sensitive MOVbehavior is not likely, as shown in Figures 65through 67. These figures show the samemargins in the spring pack force at torque switchtrip throughout the series of 10 tests as the stemforce plots (Figures 58-60). Torque spring forcemargin is defined as the difference in the torque

spring force history between the point where theoperator goes into single-path response andwhere the torque switch trips. There are observeddifferences in the spring pack force at torqueswitch trip, but these are within the 5% advertisedby Limitorque. The most interesting torquespring history is from Test 15, which shows that ahigher spring pack force was required during therunning portion of the stroke than was required inTests 16 and 17, even though the stem load wasthe same in all three tests. The stem nut coeffi-cient of friction discussed later in this sectioncaused the additional spring pack force during therunning portion of Test 15.

The possibility of changes in the ratio of springpack force to stem torque causing load-sensitiveMOV behavior is not likely, as shown inFigures 68 through 70. The variation of the ratiobetween individual tests was less than thatassociated with the worm interacting with differ-ent parts of the worm gear during the same test, asshown by the oscillations in the figures. Changesin the ratio indicate that the torque transferefficiency of the motor operator mechanismactually improves somewhat with increasingload. Such an improvement could result from asmall reduction in the coefficient of friction in theoperator bearing gear surfaces associated with anincreasing normal surface force, a known frictionphenomenon. Regardless, an increase in effi-ciency is counter to the losses associated withload-sensitive MOV behavior.

In contrast to the other data, the stem-to-stem-nut coefficient of friction changed significantlyand in a fashion that supports such frictionallosses as the most likely cause of load-sensitiveMOV behavior. Figure 71 shows that the stem-to-stem-nut coefficient of friction during Tests 8,9, and 10 remains relatively constant during thelow-load tests. Test 11, as shown in Figure 72,provides the first truly loaded test and the coeffi-cient of friction is observed to decrease. This isconsistent with an increase in the normal surfaceforce. However, Tests 12, 13, and 14 are not theresult of normal friction behavior; they are mostlikely the result of the lubricant being squeezedout of the stem-to-stem-nut interface. In other

NUREG/CR-5720 74


200

0Test 8 T Test 10

on-o

a)

0LLU-1C-)

CZ

a.cm. _

QLUn

-200

-400

-600

-800

-1,000

t I'I

Margin

* Torque switchtrip

Maargin Mrgin

k

-1,200

-1,400

M450 rs-1 091-06

Figure 65. Traces of spring pack force versus time forTests 8 through 10 of the MOVLS test sequence.These traces correspond to the thrusts of Figure 59.

200

0

.0

I-,0ILC-,

a _CD

U)

-200

-400

-600

-800

-1,000

-1,200

-1,400M450 rs-1091-07

Figure 66. Traces of spring pack force versus time for Tests 10 through 14 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 60. Spring pack force does not changesignificantly from test to test.

75 NUREG/CR-5720


200

0

3-

a)0)

0

0)

C:

0LC/)

-200

-400

-600

-800

-1,000

-1,200 seat I-1 ,400

M450 rs-1091-08

Figure 67. Traces of spring pack force versus time for Tests 14 through 17 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 61. Spring pack force does not changesignificantly from test to test.

81 '--....... . ,

a)

t-

ioEa)U)a)a0I-.

C-,

ELcm

. _L-

Un

7

6

5

4

3

2

1

YV�r��WW�Test 10Test 8 Test 9

- A Torque switchtrip

0 . . . . . . . .. .I ..I

M359 kgd-O7O1-19

Figure 68. The ratio of spring pack force to stem torque for Tests 8 through 10 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 59.

NUREG/CR-5720 76


8 .~~~~- - . . . . . . .- - - . .- - - ..-...llI

ai)

0

Ea)nci)2CoIL0CL

tM

CL

C')

7

6

5

4

3

2

1

1 AH

, . . . . . . . . . . . . . . .

�� vlll-��i

Test 10 s Test 11 Test 12 Test 13 Test 1 4

A Torque switch trip

0

b1359 kod-O7St*20l

Figure 69. The ratio of spring pack force to stem torque for Tests 10 through 14 of the MOVLS testsequence. These traces correspond to the thrusts of Figure 60.

8

cir0~F.

Ea)(1)C)0oLL

.wrL

cmC:

,()

6

5

4

3

2Trest 15

Test 15 Test 16 Test17i

Test 16 Test 17Test 14

4

0

T . I.A_ Torquo switch rip

b359 0d-0701-21

Figure 70. The ratio of spring pack force to stem torque for Tests 14 through 17 of the MOVLS testsequence. These traces'correspond to the thrusts of Figure 61. Spring pack force does not changesignificantly from test to test. - I

77 NUREG/CR-5720


C0

LL4-0

C

(1)00

0.20

0.18

0.16

0.14

0.12

0.10

0.08

0.06

0.04

0.02

0.00

Figure 71.MOVLS testthese tests.

0.20

0.18

0.16

M3S $gd-0791-27

Traces of coefficient of friction in the stem nut versus time for the first three tests of thesequence. Very little change in coefficient of friction is observed at the low thrust levels of

C0L)

U-

0

0)a)

00:0

0.14

0.12

0.10

0.08

0.06

0.04

0.02

0.00

FM350 1gd0791-28

Figure 72. Traces of coefficient of friction in the stem nut versus time for Tests 10 through 14 of theMOVLS test sequence. A steady increase in coefficient of friction is evident in these tests. Coefficient offriction is highest during Test 14, the test exhibiting load-sensitive MOV behavior.

NUREG/CR-5720 78


words, the stem-to-stem-nut interface goes from athick-film lubricated surface to near metal-to-metal contact. This hypothesis is further substan-tiated by Test 15, as shown in Figure 73, wherethe coefficient of friction remains high during thefirst unloaded test following the highly loadedtests. During this test, the stem-to-stem-nutinterface is relubricating itself. During subse-quent Tests 16 and 17, the lubrication returns tothick-film lubrication.

This behavior has been observed with two stemand stem nut combinations in the same operator.Other stem and stem nut combinations in otheroperators must be evaluated before any definitivestatements can be made, but if additional researchconfirms the hypothesis, this provides an addi-tional concern in bounding load-sensitive MOVbehavior. The coefficient of friction of a lightlyloaded stem varies with stem-to-stem-nut com-binations. We have observed values from 0.08 to0. 18; however, they all increase when the load onthe stem increases. In fact, we have observed thatthe coefficient of friction can 'increase from 20to 50%. A'typical in situ test is a static no-loadtest, or at best a low-load test similar to thoseshown in Figure 59. During these no-load staticor low-load valve tests, the stem and disc slideinto the valve seat basically unloaded and withlittle or no pressure on the lubricated surfaces ofthe stem-to-stem-nut interface. Consequently, theunloaded test fails to generate the load necessaryto evaluate the stem-to-stem-nut coefficient offriction that will exist when the MOV is highlyloaded. Therefore, additional margin must existto ensure closure against a high stem load. Meansto conservatively establish such an acceptablemargin still need to be developed.

Once an MOV has been tested, the no-loadrelationships can be baselined and monitored fordegradation. However, it is important to recog-nize that valves that experience load-sensitiveMOV behavior at torque switch trip are not asfully seated as those that do not.

These valves are operating in the marginbetween successful and unsuccessful closure.When possible, increasing the torque switch set-

ting enough to fully seat the valve at design-basisloadings will reduce or eliminate the effects ofload-sensitive MOV behavior.

4.4 Direct Current PoweredMotor Operators

Some of the differences in ac- and dc-poweredmotor operators have been discussed in othersections of this report. This section deals withdc-powered motor operators with respect toGeneric Issue (GI)-87 concerns. In most BWRs,one valve out of each pair of GI-87 isolationvalves is dc powered. As part of our GI-87research, we looked at dc-powered motor opera-tor performance. The results of the work showthat if the dc motor operator is correctly sized forthe valve application and the power cables arelarge enough, the dc-powered motor operator willprobably equal or exceed the performance of anequally sized ac-powered motor operator.

These results were determined from MOVLStesting and earlier work performed in the field.The MOVLS is described in Section 4.3 of thisreport. The changes to the MOVLS for this testingincluded removing the ac motor and installing a125 Vdc, 40 ft-lb dc motor, and running the dcmotor controller field coil current through thetorque and open limit switches. Appropriate dcinstrumentation was also substituted for ac instru-mentation such as voltage and current. The dctesting used the same Limitorque SMB-0 motoroperator that had previously been used in themotor-operated, load-sensitive behavior studycited earlier in this report. This ensured that theoperator would not influence the results. After aseries of check-out tests, the motor operator wassubjected to a 15-cycle test series. Figure 74represents typical thrust histories from these tests.Simulated valve loadings included no-load tests(Test I) to tests that bypassed the torque switchand nearly stalled the motor (Test 15). Asexpected from previous field experience, thestroke time increased as the motor operator loadincreased. The MOVLS monitored motor speedthrough a monopole pickup that counts the teethon the worm shaft gear. In the highest loaded test(Test 15), the motor speed was reduced from an

79 NUREG/CR-5720


0.30 - ------0.28 .0.26-0.24

c 0.22

0.20.LL 0.18a

0.16C: 0.14

0.1 2~'0.1 0

O 0.08 Test 14 Test 15 Test 16 Test170.0 60.04

A Torque switch trip0.020.00

M359 b ad0791-29

Figure 73. Traces of coefficient of friction in the stem nut versus time for the last four tests of theMOVLS test sequence. A full opening/closing low-thrust cycle (Test 15) is needed following the testexhibiting load-sensitive MOV behavior (Test 14) before the coefficient of friction returns to the levelsrecorded for the initial low-thrust level tests.

0

-5,000

-10,000

' -15,000

O -20,0000

E -25,000

' -30,000

-35,000DC Test 1 DC Test 12 DC Test 15

-40,000

-45,000

M450 rs-1091-O1

Figure 74. Typical stem thrust histories from the dc-powered MOVLS tests.

NUREG/CR-5720 80


unloaded speed of 1900 rpm to below 400 rpm,at which time the current was manually inter-rupted. The stroke time, of course, increaseddramatically, going from an unloaded stroke timeof 6.0 seconds (Test 1) to a fully loaded stroketime of 12.5 seconds (Test 15 of Figure 74). Oneof our concerns before testing a dc-poweredmotor operator was the effect of the lower stemnut speed.

As the de motor slows down and increases thetime the stem nut is highly loaded, it may increasethe load-sensitive behavior'of the'motor operator(rate-of-loading effect). This was not the case.Figure 75 compares an unloaded coefficient offriction history, Test 1 with Test 12, which showsthe increase in friction from a heavily loadedoperator, and with Test 15, which represents thecoefficient of friction resulting from an operatorthat was subjected to twice the load used inTest 12. The increase in the stem nut coefficientof friction appears to have stabilized betweenTests 12 and 15 in spite of the increasing load.This offers the possibility that the increase in thestem nut coefficient of friction under load can bebounded. However, additional testing with morestem-to-stem-nut combinations will be needed toverify such a hypothesis.

The load-sensitive behavior of a dc-poweredmotor operator is very similar to that observedduring testing with an ac-powered motor opera-tor, discussed earlier in this report. The'results ofthe dc-powered motor operator may appear to beworse because of the increase in the'stroke time;however, the final thrusts and associated stem nutcoefficients of friction are nearly equal to

similarly loaded ac tests. The output torque of adc motor, as shown in the motor torque versusspeed curve presented in Figure 48, is linear andnot significantly influenced as the unit slowsdown. On the other hand, the speed-torque rela-tionship of an ac motor is relatively constant untilit reaches the knee in the motor torque versusspeed curve and can thus stall from a relativelyhigh speed. Consequently, marginally sizedmotor operators equipped with an ac motor maystall more easily than a unit equipped-with a dcmotor. In addition, we observed that this credible-valve operation did not cause the dc motor tooverheat and adversely influence motor operatorperformance any more than with an ac motor.

Direct-current motors do have potential instal-lation problems. The higher current demands of adc motor before stalling, up to five times theirnameplate current, were not always recognized,and the larger cabling necessary to handle theselarge currents was not always installed. Unitsoperating with undersized cables at these highercurrents will experience cable heating, and subse-quent voltage drops from cable heatup willquickly influence the performance of a dc motor.The difficulty with measuring the voltage drop ina dc motor operator circuit were discussed earlierin this report.

Figure 45, shown earlier in this report, showshow quickly dc motor heating can reduce the out-put of the motor. Undersized cables and motorheating from- a marginally sized motor can signif-

" icantly reduce the output of a dc motor-operatedvalve. These installation problems have beenobserved in operating plants and support the find-ing identified here.

81 NUREG/CR-5720


0.20

0.18C:0

L_

C+)I+-

0

C

C.,

0

C_

E

(D

CO

0.16

0.14

0.12

0.10

0.08

0.06

0.04

0.02

0.00

M450 rs-1 091-02

Figure 75. Typical stem nut coefficient of friction histories from the dc-powered MOVLS test. Note theincrease in friction between low-load Test 1 and high-load Test 12. The load was doubled between Tests 12and 15 without a significant increase in stem nut coefficient of friction.

NUREG/CR-5720 82

5. WRITING A PERFORMANCE STANDARD

5.1 Overview standard, was finalized at the QME meeting inNew York in 1991.

Development of a performance standard is atwo-fold process: original qualification andmaintenance of qualification (in situ testing). Thecurrent original qualification standard for motor-operated valves is ANSI/ASME B16.41, whichwas developed in 1983 and reaffirmed in 1989.This standard is being outdated by the results ofNRC and other valve testing. The valve subcom-mittee of ASME Qualification of MechanicalEquipment (QME) has rewritten the documentunder the title "QV, Functional QualificationRequirements for Power-Operated Active ValveAssemblies for Nuclear Power Plants" (same titleas B 16.41). QV will be a section in the QMEfamily of qualification standards. The rewrittenstandard is an improvement over B16.41, butincorporates the NRC test results findings only upto the start of the analysis of Phase I gate Yalvetest results (DeWall and Steele, 1989). Subse-quent analysis of Phase I testing, the results fromthe Phase II testing, and the results of separateeffects testing on the INEL MOVLS have notbeen incorporated.

The flow interruption section of QV, Annex G,has been improved to incorporate the NRC butter-fly and early gate valve test results, but it stillallows family groupings and extrapolation thatour test results have shown to be faulted. Thequalification philosophy is still go/no-go withoutguidance for margins determination, which is adefinite concern for aging. The standard shouldalso address Generic Letter 89-10 concerns, asthere are currently no guidelines that would aid autility in procuring a replacement valve thatwould meet the recommendations of the genericletter. The QV document has been approved bysubcommittees and the main committee, over theobjections of INEL and NRC representatives. Onthe second vote, the document was approved overNRC objections. Subcommittee and main com-mittee chairmen did agree to start the rewrite ofthe standard to consider incorporating the recentfindings with regard to NRC and other testing.This new task group, formed to rewrite the

Maintenance of qualification (in situ testing ofvalves) is the responsibility of the ASME Opera-tion and Maintenance (OM) Committee, which issplit into many subcommittees and workinggroups with varying responsibilities. Those ofmajor interest here are OM-8 and OM-I0; ofminor interest are OM-18 and OM-19. OM-8 isthe working group on motor operators, OM-10 isthe working group on valves, and OM-18 andOM-19 are the working groups on air andhydraulic operators.

These OM groups have two major problems:(a) they must, consider the in situ valves qualifiedand write in situ tests accordingly, when in actual-ity every possible level of valve qualificationexists in the plants, and (b) the valve cannot beconsidered separate from its operator, as it is byOM. The three operator types, referenced above,the valves, and their functions are differentenough that they should be considered as separateintegrated units instead of being separated byoperator type and valves. The OM-10 standard,currently known as ISTC, has been issued and isreferenced in Section XI of the 1989 ASMECode. Utilities are required to update theirin-service inspection plans every ten years, so, by1999, all utilities will be using OM 10 for valvetesting requirements. Unfortunately, the require-ments of OM-10 are only marginally better thanthe previous Section XI requirements. OM-10still treats stroke time as the measure of MOVfunctionality. At the working group level, OM-10is trying to add a requirement for motor currentsignature analysis to the stroke time require-ments. However, this is not expected to beadopted. OM-8, driven by Generic Letter 89-10,has redrafted their document for motor operatorsinto an integrated MOV standard. We believe thisis an improvement over anything else that hasbeen done for in situ testing of MOVs. Represen-tatives of the INEL and NRC sit on the OM-8committee and have informed the committee of

83 NUREG/CR-5720

Writing a Performance Standard

NRC-sponsored and other testing., The standardreflecting this input is now in the balloting stage.

5.2 Detailed Observations

-The valve problem starts with qualification andcontinues with the maintenance of qualification.Most of the plants were built or in constructionbefore we had a valve qualification standard.Each valve manufacturer had its own proprietarymethods. (mostly analytic) for determiningqualification and for sizing the operator. Globevalves are more likely to be challenged in dailyoperation than gate or butterfly valves. Earlyglobe valve problems included plug guidanceproblems and bent stems. Most of these have beencorrected. As discussed earlier in the report, weare not certain whether a gate valve is predictableor nonpredictable without a design-basis test. Wealso do not know if the extrapolation of butterflyvalve response was always performed in theconservative configuration we identified in ourtests. The OM standards cannot help with these

problems. The: original qualification require-ments for a valve and operator must be upgradedfor replacement valves. The upgraded qualifica-tion requirements can then serve as guidelines forthe OM standards to provide for the maintenanceof qualification.

NRC, utility, European, and Electric PowerResearch Institute test programs may provideinsights on how to quantify the capability ofcurrent in-place valves and quantify the risk thatmight be involved with those designs. The recom-mendations contained in Generic Letter 89-10and the improved diagnostic testing describedearlier in this report can help to identify some ofthe more obvious problems. It is clear that design-basis tests cannot be performed for every valve.We expect that, during the next year, moreinformation on these remaining problems willbecome available. The INEL considers thedevelopment of adequate standards a necessarypart of this work that must be performed toimprove the reliability of MOVs in the operatingplants.

NUREG/CR-5720 84

6. REFERENCES

Allis-Chalmers Corporation, Test Report on an Allis-Chalmers 6" STREAMSEAL Butterfly Valve in AirConcerning Nuclear Containment Isolation Valves, VER-0209, December 17, 1979.

DeWall, K. G., and R. Steele, Jr., BWVR Reactor Water Cleanup System Flexible Wedge Gate Isolation lValveQualification and High Energy Flow Interruption Test, NUREG\CR-5406, EGG-2569, 1989.

MPR Associates, Inc., Application Guide for Motor-Operated Valves in Nuclear Powver Plants, NP-6660,March, 1990.

NRC, NRC Action Plan as a Result of the TMI-2 Accident, Volumes I and 2, Revision, NUREG-0660,August, 1980a.

NRC, Clarification of TMI Action Plan Requirements, NUREG-0737, November, 1980b.

Steele, R., Jr., and J. G. Arendts, SHAG Test Series: Seismic Research on an Aged United States Gate Valveand on a Piping System in the Decommissioned Heissdampfreaktor (HDR), NUREG/CR-4977,EGG-2505, 1989.

Steele, R., Jr., K. D. DeWall, and J. C. Watkins, Generic Issue 87 Flexible Wedge Gate Valve Test Program,Phase II Results and Analysis, NUREG/CR-5558, EGG-2600, 1990.

Watkins, J. C., R. Steele, Jr., R. C. Hill, and K. G. DeWall, A Study of Typical Nuclear Containment PurgeValves in an Accident Environment, NUREG/CR-4648, EGG-2459, 1986.

85 NUREG/CR-5720

NRC FORM 335 U.S. NUCLEARREGULATORYCOMMISSION 1. REPORT NUMBER(2-W (A,9Iigned by NRC. Add Vol.. Suap., Rev.,NRCM 1102, end Addendum Numbers. if eny.I3201,3202 BIBLIOGRAPHIC DATA SHEET

(See instru::,ons on the reversel NUREG/CR-57202. TITLE AND SUBTITLE EGG-2643

Motor-Operated Valve Research Update 3. DATE REPORT PUBLISHED

MONTH Y YEAR

June 19924. FIN OR GRANT NUMBER

A6857, 135529S. AUTHORJSI 6. TYPE OF REPORT

Robert Steele, Jr.John C. Watkins TechnicalKevin G. DeWall 7. PERIOD COVERED s-ceusy Dales/

Mark J. Russell

8. PERFORMING OAGAN I Z ATIO -NA MEUDAD R SS (I NR C. mefuon~ademailinnssiton. ttc AginU..NcerReuaory Co- iowidenane and mailing addrr. e

Idaho National Engineering LaboratoryEG&G Idaho, Inc.P.O. Box 1625Idaho Falls, Idaho 83415

9. SPONSORING ORGANIZATION - NAME AND ADORESS (I AIR C. tyPe 'Su- as Joue-: i consjactor., proide NRC DicIision. Oftice or Region. UV. N.~ucteRegstfrogtComnisiand ma;Iing aorddesJ

Division of EngineeringOffice of Nuclear Regulatory ResearchU.S. Nuclear Regulatory CommissionWashington, D.C. 20555

10. SUPPLEMENTARY NOTES

1 1. ABSTRACT 1200 words or less

This report provides an update on the valve research sponsored by the U.S. Nuclear Regulatory Commission (NRC)that is being conducted at the Idaho National Engineering Laboratory. The update focuses on the informationapplicable to the following requests from the NRC staff:

* Examine the use of in situ test results to estimate the response of a valve at design-basis conditions

* Examine the methods used by industry to predict required valve stem forces or torques

* Identify guidelines for satisfactory performance of motor-operated valve diagnostics systems

* Participate in writing a performance standard or guidance document for acceptable design-basis tests.

The authors have reviewed past, current, and ongoing research programs to provide the information available toaddress these items.

12. KEY WORDSIDESCR:PTORS (List woro's or phrases thai will assist researchers in ocaring rhe rworro. 13. AVAILABILITY STATEMENT

Motor-operated valve (MOV) UnlimitedGate valve-stem force 14. SECURITY CLASSIfICATION

Butterfly valve-stem torque (rAhs Pagel

MOV diagnostic systems UnclassifiedIrhAs Repoen

Unclassified15. NUMBER Of PAGES

16. PRICE

aRC FORM 335 124191

THIS DOCUMENT WAS PRINTED USING RECYCLED PAPER

UNITED STATESNUCLEAR REGULATORY COMMISSION

WASHINGTON, D.C. 20555-0001

SPECIAL FOURTH-CLASS RATEPOSTAGE AND FEES PAID

USNRCPERMIT NO. G-67

OFFICIAL BUSINESSPENALTY FOR PRIVATE USE, $300