ML15142A761

NRC Staff Letter Dated March 15, 1996, to Nuclear Energy Institute Forwarding NRC Safety Evaluation on Electric Power Research Institute (EPRI) Topical Report TR-1 0327, EPRI MOV Performance Prediction Program (Revision 1)

ML15142A761
Person / Time
Issue date:	03/15/1996
From:	Thadani A Office of Nuclear Reactor Regulation
To:	Tipton T Nuclear Energy Institute
References
NUDOCS 9608070288
Download: ML15142A761 (42)

Text

Ai UNITED STATES NUCLEAR REGULATORY COMMISSION WASHINGTON, D.C. 20566-0001 March 15, 1996 Mr. Thomas E. Tipton Vice President, Operations and Engineering Nuclear Energy Institute 1776 Eye Street, NW Suite 400 Washington, DC 2006-3708

SUBJECT:

ELECTRIC POWER RESEARCH INSTITUTE,(EPRI) TOPICAL REPORT TR-103237, "EPRI MOV PERFORMANCE PREDICTION PROGRAM" (REVISION 1)

Dear Mr. Tipton:

The NRC staff has completed its review of the subject topical report submitted by the Nuclear Energy Institute (NEI) on February 22, 1994 (Revision 0), and November 3, 1995 (Revision 1). On February 5, 1996, we forwarded a safety evaluation on the subject topical report with a request that we be notified if you considered any material in the safety evaluation to be proprietary. In response to your letter of March 6, 1996, we have revised the safety evaluation to remove material that you consider to be proprietary.

With the conditions and limitations described in the enclosed safety evaluation, the staff considers the EPRI Motor-Operated Valve (MOV)

Performance Prediction Program to provide an acceptable methodology to predict the thrust or torque required to operate gate, globe, and butterfly valves within the scope of the program, and to bound the effects of load sensitive behavior on motor-actuator thrust output.

The accepted version of the topical report should incorporate this letter, the enclosed safety evaluation, NRC staff written comments provided to NEI on the EPRI program, and the submitted EPRI responses. The staff will place this safety evaluation, the submitted non-proprietary version of the EPRI topical report, and a to-be-submitted non-proprietary version of the EPRI responses to NRC staff written comments in the NRC Public Document Room.

Sincerely, Ashok C. Thadani Associate Director for Technical Review Office of Nuclear Reactor Regulation

Enclosure:

Safety Evaluation cc w/encl : NRC Public Document Room

SAFETY EVALUATION BY THE OFFICE OF NUCLEAR REACTOR REGULATION OF ELECTRIC POWER RESEARCH INSTITUTE TOPICAL REPORT TR-103237 "EPRI MOTOR-OPERATED VALVE PERFORMANCE PREDICTION PROGRAM" ENCLOSURE 9608070288 960315 PDR TOPRP EXIEPRI C PDR

TABLE OF CONTENTS SECTION PAGE I. OVERVIEW OF EPRI MOV PERFORMANCE PREDICTION PROGRAM 1 II. NRC STAFF REVIEW OF EPRI TOPICAL REPORT 4 A. General Comments 4 B. Specific Comments on EPRI Computer Model 6

1. System Model 6

a. Model Description

b. Model Evaluation

c. Conditions/Limitations

2. Gate Valve Model 8

a. Model Description

b. Model Evaluation

c. Conditions/Limitations

3. Globe Valve Model 18

a. Model Description

b. Model Evaluation

c. Conditions/Limitations

4. Butterfly Valve Model 21

a. Model Description

b. Model Evaluation

c. Conditions/Limitations

5. Motor Actuator (Load Sensitive Behavior) 24

a. Method Description

b. Method Evaluation

c. Conditions/Limitations

TABLE OF CONTENTS (continued)

SECTION PAGE C. Specific Comments on EPRI Hand-Calculation Models 28

1. Anchor/Darling Double-Disk Gate Valve Model 28

a. Model Description

b. Model Evaluation

c. Conditions/Limitations

2. Westinghouse Flexible-Wedge Gate Valve Model 32

a. Model Description

b. Model Evaluation

c. Conditions/Limitations

3. WKM Parallel-Expanding Gate Valve Model 35

4. Aloyco Split-Wedge Gate Valve Model 35 III. CONCLUSIONS 35 IV. LESSONS LEARNED 36

I. OVERVIEW OF EPRI MOV PERFORMANCE PREDICTION PROGRAM In the 1980s, continuing problems with the performance of motor-operated valves (MOVs) at nuclear power plants raised concerns regarding MOV design, testing, and maintenance. In response to these problems, both the nuclear industry and the Nuclear Regulatory Commission (NRC) initiated efforts to improve the performance of MOVs at nuclear plants.

In Generic Letter (GL) 89-10, "Safety-Related Motor-Operated Valve Testing and Surveillance," the NRC staff requested nuclear p6wer plant licensees and construction permit holders to review MOV design bases, revise procedures for sizing and setting MOVs, test each safety-related MOV under design-basis conditions where practicable, periodically verify MOV design-basis capability, and trend MOV problems. In implementing MOV programs in response to GL 89-10, licensees and permit holders found that many MOVs could not practicably be tested under design-basis conditions.

As part of the industry response to the MOV issue, the Electric Power Research Institute (EPRI) initiated efforts in 1991 to document existing MOV maintenance and engineering evaluation technology, and studied long-term industry needs. A Technical Advisory Group (TAG) was formed of utility-industry MOV experts to provide guidance to EPRI on the formulation and execution of an MOV research program. The short-term products of the EPRI MOV program included engineering application guides for MOVs, an in-situ MOV test guide, and an MOV margin improvement guide. The long-term activity, the EPRI MOV Performance Prediction Program, focused on the development and validation of improved methods for bounding MOV performance.

The objective of the EPRI MOV Performance Prediction Program was to develop a methodology to be used in demonstrating the design-basis capability of MOVs when valve-specific design-basis test data are not available. The program included development of improved methods for prediction or evaluation of system flow parameters; gate, globe and butterfly valve performance; and motor-actuator rate-of-loading effects (load sensitive behavior). Further, EPRI performed separate effects testing to provide information for refining the gate valve model and rate-of-loading methods; and conducted numerous MOV tests to provide data for model and method development and validation, including flow loop testing, parametric flow loop testing of butterfly valve disk designs, and in-situ MOV testing. Finally, EPRI integrated the individual models and methods into an overall methodology including a computer program and implementation guide.

The key elements of the EPRI MOV Performance Prediction Program are (1) a System Flow Model to predict the differential pressure and fluid pressure for pumped flow and blowdown system configurations, (2) a Gate Valve Model to predict the thrust required to operate gate valves and potential damage at sliding interfaces, (3) a Globe Valve Model to predict the thrust required to operate globe valves, and (4) a Butterfly Valve Model to predict the torque required to operate butterfly valves.

These elements of the EPRI MOV Performance Prediction Program are implemented through computer software. Because of their unique design features, EPRI developed hand-calculation methods for the Anchor/Darling

double-disk gate valve, the Westinghouse flexible-wedge gate valve, the WKM parallel-expanding gate valve, and the Aloyco split-wedge gate valve. With respect to motor-actuator output, EPRI discusses six methods to address the potential reduction in thrust output under dynamic conditions compared to static conditions (commonly referred to as rate-of-loading effects or load sensitive behavior).

EPRI performed separate effects testing to determine friction coefficients and damage threshold levels for a range of material pairs, contact geometries and stresses, fluid media, and temperatures typical of nuclear plant fluid systems. EPRI conducted tests using an apparatus in which gate valve internal parts could be loaded with hydraulic actuators to simulate dynamic loading conditions. EPRI performed tests of motor actuators to understand more fully load sensitive behavior (rate-of-loading effects) where the thrust output of the motor actuator can be lower under dynamic conditions than under static conditions for the same actuator torque output. EPRI conducted tests to assess the friction and wear characteristics of various lubricants used for valve stem and stem nut interfaces.

EPRI conducted testing at four flow loop facilities with a total of 28 gate valves, 4 globe valves, and 2 butterfly valves tested. The gate valves ranged in size from 2-1/2 to 18 inches while the globe valve sizes ranged from 2-1/2 to 10 inches. The two butterfly valves were 6 inches in size. EPRI also conducted a series of parametric tests of six different butterfly disk designs to assess the effects of disk design and piping elbows on torque requirements. Finally, EPRI obtained results from tests of 19 gate valves, 2 globe valves, and 8 butterfly valves conducted in-situ at nuclear plants to supplement the flow loop test data.

The EPRI MOV Performance Prediction Methodology (PPM) includes computer models, software, and hand calculation models to implement individual valve and system models and methods. EPRI compared the predictions of the integrated methodology to data from the flow loop and in-situ tests.

EPRI determined that the thrust and torque predictions from the methodology bounded the measured data for 173 of the 176 total analyzed strokes. The EPRI methodology significantly overpredicted the thrust and torque requirements in some instances. The methodology suggested that 33 of 97 gate valve strokes would result in unpredictable thrust requirements because of guide or seat damage, but only four valve strokes experienced severe damage.

On February 22, 1994, the Nuclear Management and Resources Council (now the Nuclear Energy Institute (NEI)) requested NRC staff review of the topical report being prepared by EPRI on its MOV Performance Prediction Program. Subsequently, NEI submitted the proprietary EPRI Topical Report TR-103237 (November 1994), "EPRI MOV Performance Prediction Program - Topical Report," for NRC staff review. EPRI prepared 25 additional reports to support the topical report. On November 3, 1995, NEI submitted proprietary and nonproprietary versions of Revision 1 of the EPRI topical report.

2

The NRC staff and its contractor, Idaho National Engineering Laboratory (INEL), evaluated the EPRI MOV Performance Prediction Program by reviewing the EPRI topical and supporting reports, operating the computer software, and discussing the program with NEI, EPRI and its contractors, and the EPRI TAG utility members. Frequent public meetings were held between the NRC staff, NEI, EPRI, and utility representatives; minutes of these meetings have been placed in the NRC Public Document Room. In this safety evaluation, the NRC staff discusses its review of the computer models for fluid systems, and gate, globe, and butterfly valves. The staff also discusses its review of the hand-calculation methods for rate-of-loading effects and models for the Anchor/Darling double-disk gate valve and Westinghouse flexible-wedge gate valve. The staff will discuss its review of the hand-calculation models for the WKM parallel-expanding gate valve and the Aloyco split-wedge gate valve in a supplement to this safety evaluation.

The research conducted by EPRI significantly improves the understanding of MOV behavior and the ability to predict MOV performance. In a letter dated September 27, 1995, NEI forwarded a summary of important contributions and findings resulting from the EPRI MOV Performance Prediction Program. Because of the numerous reports provided to the NRC and the availability of published papers, NEI did not believe that licensees are required under 10 CFR Part 21 to further notify the NRC of the results of the EPRI MOV program. As described in an enclosure to the NEI letter, important findings (or confirmatory information) from the EPRI MOV program include the following: (1) the traditional methods for predicting gate valve performance might be nonconservative for many applications because of incomplete equations, design features, manufacturing controls, and wide-ranging friction coefficients; (2) the edge radii on disk seats and guide slots are critical to gate valve performance and predictability; (3) Stellite friction coefficients increase with differential-pressure valve strokes in cold water to a plateau level, stabilize quickly in hot water, and decrease as differential pressure increases; (4) gate valves with carbon steel guides and disk guide slots with tight clearances might fail to close under blowdown conditions; (5) many existing gate valve manufacturing and design processes and controls, and plant maintenance practices, might result in poor valve performance; (6) traditional methods for predicting globe valve performance for incompressible flow conditions are nonconservative for globe valves in which differential pressure acts across the plug guide; (7) globe valve thrust requirements for some designs can be excessive under compressible flow and blowdown conditions due to potential for plug side loading; (8) rate-of-loading effects (load sensitive behavior) can reduce static thrust output by up to 30 percent under dynamic conditions; (9) hydrodynamic torque coefficients used by some butterfly valve manufacturers might be nonconservative for certain applications, with valves located near piping elbows especially vulnerable;*and (10) butterfly valve seats should be periodically replaced to avoid hardening or degradation. In addition to these reported important findings, EPRI confirmed that thrust requirements to unwedge a gate valve can be higher under dynamic conditions than under static conditions.

3

In the following sections of this safety evaluation, the staff discusses its review of the specific elements of the EPRI MOV Performance Prediction Program. The EPRI program is the most extensive effort to date to evaluate and predict MOV performance. The NRC staff considers the EPRI program to provide significant information for the improvement of the design, setup, and testing of MOVs used in nuclear power plants.

With the limitations and conditions described herein, the staff considers the EPRI MOV Performance Prediction Program to provide an acceptable methodology to predict the thrust or torque requi'ed to operate gate, globe, and butterfly valves within the scope of the EPRI program, and to bound the effects of load sensitive behavior on motor-actuator thrust output.

As will be discussed below, EPRI was unable to provide sufficient test data or information to validate certain applications of specific models or methods for use as design standards in sizing and setting MOVs. EPRI or individual licensees may obtain additional information to support the use of the models or methods as design standards for those applications.

The staff will consider requests to update this safety evaluation based on such new information. In the interim, a licensee might use a specific EPRI model or method in an application not validated as a design standard based on the significant improvement in the understanding of MOV performance provided by the EPRI program over past MOV design practices.

II. NRC STAFF REVIEW OF EPRI TOPICAL REPORT A. General Comments The NRC staff and its contractor evaluated the EPRI MOV Performance Prediction Program through review of EPRI documentation of the program; operation of the computer software; and discussions with NEI, EPRI, and EPRI TAG utility members. On July 31, 1995, the NRC staff provided written comments to NEI on the overall operation of the EPRI computer program. On September 15, 1995, NEI submitted the EPRI response to the staff comments. NEI submitted a revised EPRI response on November 14, 1995.

Implementation of the EPRI MOV Performance Prediction Methodology and operation of the EPRI computer program require the user to understand the topical report, implementation guide, MOV system performance, training materials, motor-actuator sizing and torque-switch setting guidelines, and load sensitive behavior in. MOV performance. EPRI supplements the implementation guide and software users manual with 1-week classroom training. EPRI considers the classroom training and the classroom workbooks to provide detailed guidance for users to properly implement the EPRI methodology. The staff considers that the complexity of the EPRI methodology and the depth of knowledge required to reliably implement the methodology mandate that users of the methodology 4

complete the training provided by EPRI and periodically receive refresher training. (See NRC staff comment 1 on computer program and EPRI response.)

EPRI has established a process by which users of the EPRI computer program receive software error notices. For example, EPRI issued a software error notice to alert users to instances where, after significant changes are made to certain variables in the butterfly valve model, the computer program might not recalculate the effects of the changes. The error notice provides instruction for users to ensure that the problem will not cause incorrect results to be calculated. The staff considers it important for users of the EPRI methodology to review and take action in response to software error notices provided by EPRI. (See NRC staff comments 2 and 3 on computer program and EPRI responses.)

The EPRI computer program has weaknesses in modelling certain system conditions. For example, the pressures obtained in the System Flow Model would not converge when modeling a parallel flow path under high-temperature conditions. EPRI states that, for some system configurations and operating conditions, the system flow model solution algorithm does not converge and the computer program does not provide a calculated result. EPRI believes that, in many cases, the computer program user can take action to force the solution to converge. EPRI has established a PPM Users Group, including a newsletter, a bulletin board, and meetings where potential actions to eliminate such problems are discussed. In addition to successful completion of initial and periodic training, the staff considers participation in the PPM Users Group to be important in ensuring that operating problems with the computer software are correctly recognized and addressed in implementing the computer program. (See NRC staff comments 4, 5, 7, 8, 9, and 11 on computer program, and EPRI responses.)

In developing its methodology, EPRI omitted certain considerations to simplify the modelling effort. EPRI considers the total impact of the simplifying assumptions to be difficult to quantify because some of the assumptions have an effect only under certain conditions and might be conservative or nonconservative depending on the conditions. EPRI believes the conservatism of the overall methodology has been demonstrated in the validation of the individual models and assessment of the integrated methodology.

The EPRI models are very conservative in predicting thrust/torque requirements in some instances and marginally bounding in others.

Overall, the staff agrees with EPRI that, with some exceptions identified herein, the methodology is sufficiently conservative to account for the uncertainties and the simplifying assumptions for the large majority of valves and operating conditions. (See NRC staff comment 6 on computer program and EPRI responses.)

In prescribing implementation of the models, the EPRI MOV Performance Prediction Program focused on the operation of torque-5

switch-controlled MOVs. Therefore, the determination of the thrust/torque requirements and load sensitive behavior for limit-switch-controlled valves requires special attention. For example, diagnostic tests of limit-switch-controlled MOVs under static conditions will yield only packing load, and the stem/stem nut coefficients of friction obtained during such low loading conditions would not be reliable based on current test results.

The users of the EPRI methodology should be aware of the differences in evaluating the capability of torque-switch- and limit-switch-controlled valves when implementing the EPRI models and methods.

EPRI confirmed that valve disk-to-seat friction coefficients will typically be greater at low fluid temperatures than at high temperatures. Therefore, EPRI indicated that the user of the EPRI methodology is expected to apply the lower fluid temperature where a range of temperatures exists for various design-basis scenarios or within a specific scenario.

The staff found that certain references used in the development of the EPRI models and methods were not supported by an audited quality assurance program. On the basis of EPRI sensitivity studies of nonvalidated assumptions, consistency of the thrust/torque predictions with actual test data, and limited validation of the EPRI models and methods by the staff and its contractor, the staff determined that the EPRI models and methods are acceptable with the limitations and conditions discussed herein. (See, for example, NRC staff comment 2 on globe valve model and EPRI response.) -

B. Specific Comments on EPRI Computer Model

1. System Model

a. Model Description The System Flow Model (SFM) is a one-dimensional flow analysis model that calculates the hydraulic behavior of typical nuclear plant systems. The System Flow Model calculates the upstream pressure, pressure drop, and flow through a valve over the entire valve stroke. The flow equations used in the model are based on principles of fluid mechanics.

EPRI validated the System Flow Model using data from five tests in its flow loop test program. EPRI found that the model predicted differential pressure within 10 percent during a valve stroke.

Predictions of system upstream pressure ranged between 6 and 12 percent of the actual test value. EPRI considered the test results sufficient to validate its model.

EPRI determined the System Flow Model to be applicable to pump flow and blowdown conditions. The system must have boundaries 6

that can be modeled as constant pressure or as a predetermined pressure for a source tank. For pump-flow systems, EPRI determined that the fluid medium must be subcooled water (up to 3000 pounds per square inch absolute (psia) and 700°F). The pump-flow system must be capable of being modeled by one of four model configurations. For blowdown systems, EPRI determined that the fluid medium can be subcooled water, two-phase water and steam, or steam up to 3000 psia and 700 0 F. The blowdown system must be capable of being modeled as a single flow path from a source tank.

For subcooled water applications, EPRI developed a simplified implementation of the System Flow Model. The simplified implementation requires" only design-basis differential pressure and flow rate instead of detailed system inputs.

The System Flow Model can be implemented using the Full SFM Method, the Equivalent Resistance Method, and the Butterfly Valve Model (BFM) Steam Method. The Full SFM Method is required for gate valves in systems with steam or flashing water and in all applications with an active parallel valve. The BFM Steam Method must be used with butterfly valves installed in compressible flow service. EPRI recommends the Equivalent Resistance Method for all other gate and butterfly valve applications in systems with incompressible water flow and no active parallel valves. The System Flow Model is not required in evaluating globe valves that can use a simplified Single Point Method.

The output of the System Flow Model (or the Single Point Method when applicable) is used by the gate, globe, and butterfly valve models in predicting the thrust or torque required to operate the valve.

b. Model Evaluation The NRC staff and its contractor evaluated the System Flow Model by reviewing EPRI documentation of the model, operating the computer software, and discussing the model with NEI, EPRI, and the EPRI TAG utility members. The staff included comments on the System Flow Model together with comments on the specific gate, globe, and butterfly valve models, and operation of the computer software.

The fluid parameters generated by the System Flow Model are used in the computer model to predict thrust or torque requirements for the valve. The computer model user is responsible for ensuring that the system flow model properly reflects the design-basis conditions for the valve. For example, the model user must not reduce fluid loading on the valve being evaluated based on the effects of other valves if the design-basis condition assumes that other valves are not operating. In addition, the prediction of torque and thrust requirements can be highly sensitive to the output of the System Flow Model. Therefore, EPRI states in the implementation guide that computer model users are expected to 7

confirm that the predicted differential pressure, upstream pressure, and flow rate are consistent with the design-basis condition for the valve. (See NRC staff comment 4 on gate valve model and EPRI response, and NRC staff comment 4 on butterfly valve model and EPRI response.)

c. Conditions/Limitations The NRC staff considers the System Flow Model (and, when applicable, the Single Point Method) to provide a reasonable prediction of system parameters for use by the gate, globe, and butterfly valve models in determining the thrust or torque required to operate the valves. The staff recognizes that the System Flow Model (and Single Point Method) will have uncertainties associated with the predicted pressure and flow parameters. However, the staff does not consider these uncertainties to prevent the gate, globe, and butterfly valve models from predicting with acceptable accuracy the thrust or torque required to operate the valves.

In implementing the System Flow Model, computer model users will be expected to ensure that the model properly reflects the design-basis conditions for the operation of the valve being evaluated.

For example, each of two containment isolation valves in series might be required to close assuming that the other valve in the pipe line does not close. Further, modelling of butterfly valve systems should consider potential loss of integrity of air ducting.

2. Gate Valve Model

a. Model Description The EPRI gate valve computer model predicts the thrust required to operate gate valves throughout their stroke up to initial wedging under specified fluid conditions and differential pressure. EPRI developed the gate valve model using theoretical equations describing thrust components acting on the gate valve stem. The thrust required to operate a gate valve is calculated by summing the various force components acting along the axis of the valve stem. These force components include disk friction thrust due to differential-pressure load, stem rejection thrust, stem thrust due to valve packing friction, torque reaction thrust, and stem thrust due to disk and stem weight. Beyond initial wedging, the computer model user must determine the appropriate hard-wedging thrust requirement.

EPRI validated the gate valve model by comparing data from its flow-loop tests, tests performed at power plants, and tests performed for the NRC by the Idaho National Engineering Laboratory. A total of 29 valves and 199 test strokes were used in the validation of the EPRI gate valve computer model. Key test 8

parameters measured for model validation included stem thrust, upstream pressure, valve differential pressure and temperature, and stem position.

Based on its test program results, EPRI states that, under ambient temperature conditions, disk-to-seat friction increases with successive valve strokes until a plateau is reached. EPRI found the friction coefficient to start as low as 0.1 and reach up to 0.6 before stabilizing. EPRI stated that, prior to testing, efforts were made to precondition flow-loop gate valves by repeatedly short-stroking the valves at full differential pressure until the measured apparent coefficient of friction stabilized.

EPRI did not precondition the valves tested in-situ. Some valves tested in-situ showed lower coefficients of friction, which may have been an indication of reduced service conditions at the time of the test.

EPRI compared model predictions of valve internal damage to actual occurrences of valve damage. For ambient water tests with normal flow (13 to 22 feet per second (fps)), EPRI testing revealed only one valve to have been damaged during testing; the damage was caused by improper valve fabrication or maintenance. For ambient water tests with higher flow (33 to 72 fps), EPRI testing found no damage due to galling or gouging of the materials. However, the guide rails in two valves manufactured by Velan experienced plastic bending under the high-flow conditions. The bending caused the thrust requirements to increase by about 15 percent.

The EPRI computer model cannot predict the increased thrust requirements resulting from bent guide rails. If a valve has been dynamically tested sufficient to bend the guide rails, the increase in thrust can be observed in the static diagnostic test.

The user then adds this thrust requirement to the EPRI computer model prediction. If a valve has not been exposed to the design-basis conditions that might cause the guides to bend, the EPRI model might not sufficiently predict the thrust required to fully close and seat the valve. EPRI believes that the model prediction would be sufficient to reach a zero-stroke position where the disk is covering the valve orifice, but flow may not be isolated.

EPRI conducted hot-water blowdown tests on five valves and obtained data from INEL for three other valves. Under these test conditions, two valves sustained severe damage and the thrust requirements were deemed unpredictable. EPRI found that valves with carbon steel guide rails and guide slots showed various degrees of galling. EPRI determined that valves with carbon steel guide slots and rails must have at least a minimum specified guide clearance for galling products not to fill the available clearance. EPRI notes that the guide clearance criterion is not built into the computer model but must be determined by the user by inspection or by obtaining the worst-case tolerances from the valve vendor. EPRI also tested two valves under hot-water normal flow conditions and did not identify any significant valve damage.

9

EPRI conducted steam blowdown tests on one valve and obtained test data from INEL for three other valves. One valve revealed significant damage, and the thrust required to operate the valve was determined to be unpredictable. EPRI tested one valve under normal steam flow conditions and no damage was revealed.

EPRI considers its gate valve model to have been validated for the following gate valves: solid- and flexible-wedge gate valves with stem-to-disk connection consisting of a T-head on the stem and a parallel or a cross T-head slot on the disk, manufactured by Anchor/Darling, Borg-Warner, Crane (with insertable U-shaped guide), Pacific (now CCI), Powell, Velan, and Walworth; valve sizes 2.5 to 18 inches; ANSI pressure classes 150, 300, 600, 900, and 1500 pounds; all guide rail designs except those with slots in valve body and rails on the disk; guide rail-to-guide slot material combinations of carbon steel-carbon steel, Stellite-carbon steel, Stellite-stainless steel, Stellite-Stellite, 17-4 PH-carbon steel, and stainless steel-stainless steel; and disk-to-seat material of Stellite 6 to Stellite 6. EPRI considers the gate valve model to have been validated for the following fluid conditions: ambient water at flow rates ranging from 13 to 72 fps; 450°F (non-flashing) water at velocities ranging from 18 to 21 fps; hot-water blowdown conditions ranging from 870 psi and 525 0 F to 2660 psi and 659°F; steam blowdown conditions ranging from 720 psi and 512°F to 1260 psi and 570'F; and 580°F steam flow at 200 fps.

To implement the gate valve computer model, the user must verify applicability of the valve and fluid conditions, establish a proper piping configuration for the System Flow Model, and obtain detailed internal gate valve information by inspection or from the valve vendor. The user must perform a static diagnostic test to verify the packing load and to determine whether any additional thrust requirement is present from bent guides or other "parasitic" effects. The user may operate the computer program with default parameters or input data directly. For example, the user may determine a disk-to-seat friction coefficient from testing the valve. EPRI states that the user is responsible for the accuracy of any inputs and the subsequent effect on the accuracy of the computer model prediction. Further, EPRI states that, if the user provides a disk-to-seat friction coefficient, the computer model results and the design-basis margin are applicable only to the MOV at the time of the test. EPRI states that the user is responsible for accounting for any potential changes to the MOV which may occur and which may affect the disk-to-seat friction. EPRI believes that the default friction coefficients within the computer model will bound the disk-to-seat friction at any time, provided.the MOV is properly maintained.

The computer model will generate a "Type 2"' warning where the valve might experience minor damage but the thrust requirement remains reliable. A Type 2 warning might also be generated to 10

identify potential guide rail bending where the model might not predict sufficient thrust to fully seat the valve. The computer model will generate a "Type I" warning when substantial internal valve damage is predicted and a reliable thrust requirement cannot be determined. EPRI provides options for the user to obtain more specific valve information or to perform internal valve modifications in an effort to eliminate a Type I valve damage warning.

The computer model predicts opening and closing thrust requirements for the valve throughout the valve stroke. The computer model produces a summary output table of thrust requirements at certain stroke positions. For the valve closing stroke, EPRI states that the larger of the stem thrust at initial wedging and maximum stem thrust prior to initial wedging must be used as the design-basis required stem thrust. For valve opening stroke, EPRI specifies that the larger of the maximum-predicted stem thrust during the stroke, and the unwedging load be used as the design-basis stem thrust. As discussed below, the computer model does not predict the unwedging load, which must be estimated by a hand-calculation method. A user of the EPRI methodology must ensure that the sizing and setup of the MOV is sufficient to accommodate both the thrust predicted by the computer model for dynamic performance and the thrust predicted by the hand-calculation method to unwedge the valve.

During initial opening of a gate valve, high thrust may be needed to unwedge the disk from the valve seat. The EPRI gate valve computer model does not predict this unwedging thrust. Therefore, EPRI developed a hand-calculation method to estimate the stem thrust required to initially unwedge the valve based on previous closure thrust, packing thrust, and differential-pressure load.

EPRI compared its predicted unwedging thrust to the actual unwedging thrust for 19 valves and found the hand-calculation method to be bounding for all but one valve that was underpredicted by 1 percent. EPRI states that the effects of pressure locking and thermal binding can increase the unwedging thrust requirement beyond the EPRI calculation method, but that these effects are beyond the scope of the EPRI program.

b. Model Evaluation The NRC staff and its contractor evaluated the EPRI gate valve model by reviewing EPRI documentation of the model, operating the computer software, and discussing the model with NEI, EPRI, and the EPRI TAG utility members. On May 8, 1995, the NRC staff provided written comments to NEI on the EPRI gate valve model. On July 27, 1995, NEI submitted an initial response to the staff comments. NEI submitted additional EPRI responses to the staff comments on August 23 and October 13, 1995.

11

Based on NRC staff and contractor review of the EPRI reports, the EPRI responses to staff comments, and discussions between the staff and EPRI, the staff concludes that, with the conditions and limitations discussed below, EPRI has presented sufficient support to justify the EPRI gate valve computer model for water and steam conditions. (See NRC staff comment I on gate valve model and EPRI responses.)

EPRI obtained only limited test data for disk-to-seat friction coefficients for fluid temperatures beyond 550'F. The staff agrees with EPRI that the general trend of the available data from test specimens support the assumed friction coefficient above 550'F. However, as part of EPRI's continuing program for maintaining its methodology, EPRI is expected to evaluate new information that becomes available related to the friction coefficient at fluid temperatures greater than 550°F and make any appropriate adjustments to the model. Further, the staff considers the EPRI coefficients of friction above 550'F for water, and at 550°F and above for steam, to be applicable only to the EPRI methodology for predicting thrust and torque requirements.

The staff does not consider that sufficient evidence exists to justify applying those friction coefficients directly in industry equations to predict valve thrust requirements as valve factors or friction coefficients. (See NRC staff comment 1 on gate valve model and EPRI responses.)

Computer model users must determine the thrust requirement for unwedging forces by a hand-calculation method developed by EPRI.

Under certain conditions, unwedging thrust under differential-pressure loading conditions can exceed static unwedging forces.

In one instance, EPRI reported a 70-percent increase from static unwedging load to unwedging load under dynamic conditions. This finding might be important for licensees that have assumed that unwedging load remains constant or decreases with differential pressure. EPRI determined that the hand-calculation method satisfactorily predicts unwedging stem thrust, including those valves which may require more thrust to unwedge under differential-pressure conditions. However, EPRI had only limited data for this validation effort and licensees are expected to compare their own unwedging data in applying the EPRI hand-calculation method. (See NRC staff comment 3 on gate valve model and EPRI response, and NRC staff comment 7 on Westinghouse flexible-wedge gate valve hand-calculation model and EPRI response.)

EPRI states that the gate valve model may be applied to valves from other manufacturers than the valves tested in the EPRI program. The staff does not consider sufficient justification currently exists to apply the EPRI gate valve computer model as a design standard to valve designs without supporting test data.

Licensees will be responsible for justifying the use of the EPRI 12

model for valve designs not tested as part of the EPRI program.

(See NRC staff comment 5 on gate valve model and EPRI responses.)

EPRI states that the gate valve model may be applied to valves with other guide materials than the valves tested in the EPRI program. The staff considers EPRI to have conservatively applied the EPRI gate valve computer model to guide-disk material pairs other than those tested under the EPRI program, and therefore this is acceptable. (See NRC staff comment 5 on gate valve model and EPRI responses.)

Based on its valve testing, EPRI considers the model to apply to all orientations of gate valves and piping. Some licensees have reported operating problems with gate valves in nonstandard orientations (for example, horizontal disks in vertical pipes).

Such problems might be the result of aging or valve degradation, which may or may not be accelerated by nonstandard orientations.

EPRI has stated that the use of the computer model assumes that the valve is in good condition. Model users will need to ensure that an adequate internal valve preventive maintenance program is established for the thrust or torque requirements predicted by the model to remain valid. The user is also cautioned that aging conditions can influence valve performance. (See NRC staff comment 7 on gate valve model and EPRI response.)

User Input to the Gate Valve Model The user of the gate valve model may determine a valve disk friction coefficient from specific valve data for input into the model. In calculating the friction coefficient, EPRI allows the user to exclude the thrust requirement above the typical packing load to seat the valve during static testing. EPRI refers to this additional thrust requirement as a "parasitic" load. The user adds the parasitic load to the thrust predicted by the computer model using the user-input friction coefficient. The exclusion of the parasitic load results in a smaller test-based valve friction coefficient being input into the model. If the parasitic load changes with differential pressure, the test-based valve friction coefficient could be unreliable and, possibly, nonconservative.

In response to staff concerns, EPRI prepared a caution for model users to ensure that the "parasitic" loads observed during static tests are constant and are not affected by differential pressure.

Based on its test results, EPRI excluded from this caution gate valves manufactured by Velan because EPRI believes that these valves will have parasitic loads as a result of (1) bent valve guides or (2) performance of their cross-T head design. EPRI also excluded Borg-Warner gate valves from the caution based on their cross-T head design. The staff does not agree with EPRI that sufficient evidence exists to demonstrate that parasitic loads observed in Velan and Borg-Warner gate valves are always constant with differential pressure. In response to this concern, EPRI demonstrated that its methodology reasonably bounded the actual 13

thrust requirements for its tested valves. Therefore, the staff does not object to EPRI's omission of the Velan and Borg-Warner valves from its caution regarding parasitic loads. Nevertheless, the staff believes that it would be prudent for computer model users to carefully review their static and dynamic test data for each gate valve regardless of manufacturer when determining whether parasitic loads are constant with differential.pressure.

(See NRC staff comment 9 on gate valve model and EPRI responses.)

The computer model user must carefully evaluate the stem force trace throughout the entire stroke when a test-based valve friction coefficient is input into the gate valve model. For example, a user is expected to evaluate the significance of any observed anomaly (such as a hook in a data trace) and, where possible, confirm the results by alternate methods. EPRI agreed to caution model users to examine data traces carefully and to recommend checking the model for conservatism where the model is to be used to predict design-basis thrust requirements with available test data. (See NRC staff comment 10 on gate valve model and EPRI response.)

A valve friction coefficient determined directly from an in-situ test of the valve and input into the gate valve model might be low as a result of recent installation or maintenance, or of minimal service conditions up to that date. EPRI has stated that use of a valve-specific measured disk-to-seat friction coefficient with the EPRI methodology is consistent with the approach of demonstrating MOV capability with design-basis in-situ testing. However, EPRI notes that, in either case, thrust requirements might increase with time and valve stroking. (See NRC staff comment 11 on gate valve model and EPRI response.)

EPRI allows model users to obtain valve friction coefficients from tests under less than design-basis differential-pressure conditions in determining the design-basis thrust required to operate the valve. The staff agrees with EPRI that sufficient data exist to indicate that seat friction coefficients will not significantly increase with differential pressure provided the valve is not damaged by the flow conditions and sufficient loading is present to achieve a reliable friction coefficient. Under low differential-pressure conditions, many valves demonstrate significant scatter in their friction coefficients. EPRI has established threshold requirements in terms of an absolute value of differential pressure and a percentage of design-basis" differential pressure that must be satisfied in applying friction coefficients obtained from in-situ test data to the EPRI computer model. (See NRC comment 12 on gate valve model and EPRI responses.)

One EPRI method allows a user of the gate valve model to determine a valve friction coefficient for input into the model from hydrostatic pump tests. If pressure decreases rapidly during a 14

test, the thrust required to operate the valve might be much lower than under design-basis differential-pressure conditions. As a result, the user might obtain an unrealistically low valve friction coefficient based on this low thrust requirement assumed to occur under design-basis differential-pressure conditions.

EPRI is providing model users with guidance that requires the transient upstream pressure to be measured with sufficient sensitivity to monitor the upstream pressure based on bonnet cavity pressurization and valve disk position. If the upstream pressure decreases before the flow initiation point, the test results may result in a low friction coefficient prediction. EPRI will also recommend that the model users ensure that the fluid system is filled prior to the test. (See NRC staff comment 13 on gate valve model and EPRI response.)

Another EPRI method allows a user of the gate valve model to determine a valve friction coefficient for input into the model from the maximum opening and closing loads during static tests.

The staff encourages further development of the EPRI wedging/unwedging method for estimating the disk-seat friction coefficient. In response to staff questions regarding limited data, EPRI has established a minimum allowable friction coefficient of 0.4 when this method is used. (See NRC staff comment 14 on gate valve model and EPRI responses.)

The sampling rate and accuracy of the diagnostic test equipment need to be considered when applying a user's own test data to the gate valve model. This is particularly important in light of the need for a user to identify and quantify various stem force loads as a function of time. (See NRC comments 8 and 15 on gate valve model and EPRI responses.)

Borg-Warner Gate Valves EPRI found Borg-Warner gate valves to require more thrust to operate than observed for other manufacturers' gate valves tested in the program. The EPRI gate valve model does not bound the valve friction coefficients and thrust requirements for all Borg-Warner valves. For example, valve 8 demonstrated a 0.63 valve friction coefficient in a low-flow ambient water test under 33 percent of the design-basis differential-pressure condition; valve 9 had valve friction coefficients of 0.66, 0.65 and 0.61 in ambient water tests; and valve 10 had a valve friction coefficient of 0.60 in 33-percent and 67-percent tests. The staff questioned whether the EPRI gate valve model adequately predicts the thrust required to operate Borg-Warner gate valves. In response to staff concerns, EPRI established a 1.05 multiplier for Borg-Warner gate valve thrust predictions from the computer model after unseating or before initial wedging. (See NRC staff comment 16 on gate valve model and EPRI responses.)

15

Unpredictable Thrust Requirements Internal valve damage and higher-than-predicted thrust requirements resulted during a high percentage of the blowdown tests conducted by EPRI and previously by INEL for the NRC. In cases where the computer model identifies potential valve damage and unpredictable thrust behavior, EPRI recommends that the user round sharp valve disk edges and ensure (through inspection or manufacturer worst-case tolerances) appropriate internal valve clearances where guide surfaces are carbon steel. The staff' considers such preventive maintenance might help improve MOV performance in other instances as well. EPRI also cautions model users to have valve manufacturers evaluate any planned removal of valve guide material with respect to structural limitations. (See NRC staff comments 18, 19, and 20 on the gate valve model and EPRI responses.)

Flow Isolation EPRI has indicated that flow might continue for some valves even though the valve disk closes beyond the zero-stroke position if the disk remains sliding on the guides or the disk's inner diameter is sufficiently large to allow flow at the top of the disk. Also, because of the uncertainty in the selected location of the zero-stroke position for the tested valve, the zero-stroke position might isolate flow for one particular valve but might not isolate flow for a similar valve. EPRI requires computer model users to use the maximum calculated thrust up to the point of initial wedging as the required thrust for closure strokes. EPRI also requires that the maximum result from the two analyses with positive and negative guide offset be used. EPRI states that the model output for flow isolation is a "theoretical" flow isolation position that is for information only and is not to be used to establish thrust requirements in accordance with the EPRI methodology. The computer model user is responsible for determining an appropriate thrust requirement for hard wedging of the valve disk. (See NRC staff comment 21 on gate valve model and comment 10 on computer program, and EPRI responses.)

EPRI believes that gate valves will isolate flow even in cases where the valve guides are bent. The staff believes that, if the guides bend under high-flow conditions, the valve might allow significant leakage in that flow direction. EPRI notes, and the staff agrees, that valves with bent guides might not isolate flow in the reverse direction. (See NRC staff comment 22 on gate valve model and EPRI responses.)

c. Conditions/Limitations Limited data exist regarding friction coefficients for fluid conditions above 550°F. As part of its continuing program to maintain the computer model, EPRI is expected to evaluate new 16

information that becomes available regarding friction coefficients for fluid conditions above 550'F and make any appropriate adjustments to the model.

Model users need to ensure that unwedging loads under static and dynamic conditions are adequately considered in addition to dynamic flow loads. Users must compare their own unwedging data to the EPRI hand-calculation method.

Model users will be responsible for justifying the applicability of the EPRI model where used as a design standard for valXe designs not tested as part of the EPRI program.

Model users will need to ensure that an adequate internal valve preventive maintenance program is established for the thrust or torque requirements predicted by the model to remain valid. The user is cautioned that aging conditions or valve degradation can influence valve performance, which may or may not be accelerated by nonstandard orientations.

Model users will need to have confidence that "parasitic" loads are constant over the entire stroke of the valve and are not affected by dynamic flow conditions.

Where a model user applies in-situ test data for the valve friction coefficient, the user will need to recognize that the thrust requirements based on that in-situ data might increase with time and service, and cause the valve not to operate. A model user should particularly consider that friction coefficients can increase when the in-situ thrust data for the valve being evaluated are low compared to other plant data.

As a result of minimal data, EPRI has established a minimum allowable 0.4 friction coefficient for the wedging/unwedging method of predicting a valve friction coefficient. The staff agrees with this limitation.

EPRI has established additional margin for Borg-Warner gate valves when applying the computer model. The staff considers the 5-percent margin applied to the thrust predicted for Borg-Warner valves to be reasonable until further justification is developed to change the margin.

Model users should be aware that, where a valve must operate under blowdown or very high flow conditions, extrapolation of thrust requirements from normal pumped-flow conditions might not be sufficient to ensure that the valve can operate under its design-basis conditions.

Model users should recognize that, if the guides bend under high-flow conditions, the valve might allow significant leakage.

17

3. Globe Valve Model

a. Model Description The EPRI globe valve model calculates stem thrust required to operate globe valves throughout their stroke under specified fluid conditions and differential pressures. EPRI developed the globe valve model using theoretical equations to describe the various thrust components acting on the globe valve stem. The stem thrust for a globe valve is calculated by summing the various force components acting along the axis of the valve stem. These force components include disk differenti'al-pressure force, friction force between the disk guides and valve body, stem rejection force, gravity force on the disk and stem, valve packing friction force, and torque reaction friction force. In its model, EPRI does not include sealing loads after initial valve seating or, for globe valves with rising/rotating stems, valve packing torque.

EPRI validated its globe valve model by comparing stem thrusts calculated using the model to measured stem thrusts obtained during flow testing of globe valves under ambient fluid temperature conditions. EPRI used test data from five globe valves manufactured by different vendors under various flow conditions. Key test parameters measured for model validation included stem thrust, upstream pressure, valve differential pressure, and stem position. EPRI considered the comparison between predicted and measured stem thrust to verify that the model bounds the stem thrust developed during closing and opening strokes of globe valves. EPRI found the agreement between the model prediction and test data to be more accurate where the exact.

valve disk flow-control area is known.

EPRI considers the globe valve model to have been validated for the following globe valves and fluid conditions: T-pattern and Y-pattern valve bodies where appropriate seat-based or guide-based area terms are used; all stem orientations; incompressible flow of any pumped-flow fluid velocity (blowdown flow excluded as discussed below); body-guided or cage-guided disk with lower disk guide height to disk diameter ratio of less than or equal to 0.25; unbalanced disk, rising stem globe valves with flow over or under valve seat; unbalanced disk, rising/rotating stem globe valves with underseat flow; and balanced disk, rising stem globe valve with underseat flow.

In implementing the globe valve model, the user verifies.the applicability of the globe valve computer model. The user may apply the System Flow Model to obtain differential-pressure information throughout the entire valve stroke. However, rather than applying the System Flow Model, the user can input the single values of design-basis differential pressure and upstream pressure. This will simplify the calculation of predicted thrust requirements. The user obtains internal valve design information 18

from the valve vendor or from user measurement for input into the computer model.

For each globe valve, the computer model provides values of stroke position, differential pressure, and predicted stem thrust. The predicted thrust at each stroke position is not reliable if the System Flow Model has not been used to predict differential pressure. For closing strokes, the model predicts required thrust at disk seating and at the point of maximum predicted stem thrust prior to disk seating. For opening strokes, the model predicts required thrust at disk unseating and at the point of maximum predicted stem thrust after disk unseating. For self-actuating globe valves, no stem thrust is required to operate the valves, but the user must determine the required torque to rotate the stem nut.....

The computer model does not predict the sealing load for closure strokes and the user must determine and add this load to the model predicted closing thrust.

b. Model Evaluation The NRC staff and its contractor evaluated the EPRI globe valve model by reviewing EPRI documentation of the model, operating the computer software, and discussing the model with NEI, EPRI, and EPRI TAG utility members. On January 11, 1995, the NRC staff provided written comments to NEI on the EPRI globe valve model.

On May 24, 1995, NEI submitted responses to those comments.

Based on NRC staff and contractor review of the EPRI reports, the EPRI responses to staff comments, and discussions between the staff and EPRI, the staff concludes that EPRI has presented sufficient support to justify application of the globe valve model to the globe valves within the scope of the EPRI validation effort for cold-water pumped-flow conditions (less than approximately 150 0 F). The results from dynamic testing of globe valves by licensees in response to GL 89-10 also support the reliability of the EPRI globe valve model for cold-water pumped-flow conditions.

(See NRC staff comment 5 on globe valve model and EPRI response.)

EPRI indicates that the proper flow area term must be determined for use in applying its globe valve model. The globe valve testing by licensees reinforces the need to apply the proper area term for predicting globe valve thrust requirements whether the EPRI model or typical Limitorque actuator sizing and setting guidelines (with a 1.1 valve factor) are applied. Also, Borg-Warner has notified its customers that the proper area term must be used for predicting the thrust requirements to operate its globe valves when typical sizing and torque-switch setting methods are used. Because the globe valve model does not include margin above the thrust requirements in some instances, EPRI has stipulated that model users must input the outer diameter of the 19

disk hard-faced area to determine the area for differential-pressure application in seat-based calculations. (See NRC staff comments 7 and 9 on globe valve model and EPRI responses.)

EPRI testing of one globe valve under high-temperature and blowdown conditions resulted in much higher thrust requirements than predicted by the EPRI globe valve model or typical Limitorque actuator sizing guidelines for globe valves. EPRI could not distinguish between the thrust requirements for blowdown-flow and high-temperature conditions from this single test. Based on this limited test information, EPRI was unable to apply its methodology to globe valve operation under blowdown conditions. Further, EPRI was not able to verify the reliability of applying thrust requirements from low-temperature flow tests to obtain thrust requirements for globe valves under blowdown or high-temperature conditions. (See NRC staff comment 4 on globe valve model and EPRI response.)

EPRI conducted preconditioning tests of gate valves in its flow loops in an effort to achieve a plateau of the valves' seat friction coefficients. EPRI believes that the thrust requirements for globe valves are not strongly influenced by friction forces between the seat and disk. Therefore, EPRI did not conduct preconditioning tests of its globe valves. The staff does not object to this practice by EPRI, but notes that computer model users will need to establish appropriate preventive maintenance programs to address other aspects of valve aging. (See NRC comment 1 on globe valve model and EPRI response.)

The globe valve model focuses on thrust requirements up to the zero-stroke position. Model users must determine the appropriate valve sealing load. (See NRC staff comment 10 on globe valve model and EPRI response.)

c. Conditions/Limitations The staff has concluded that, at this time, sufficient test data have been presented to justify the applicability of the model as a design standard for globe valves within the scope of the EPRI program to cold-water pumped-flow conditions (less than.

approximately 150°F). Additional test data may be used to justify the applicability of the computer model as a design standard for globe valves beyond these conditions. Until such data are available, the model might yield the best available information when valve-specific test data cannot practicably be obtained.

Because of current uncertainties in applying thrust requirements from globe valves obtained under less severe conditions, model users must take appropriate action where a valve must operate under blowdown or high-temperature flow conditions.

20

EPRI states, and the staff agrees, that model users must identify whether the thrust requirements for each globe valve are controlled by the area of the valve seat or guide, and must apply this area term in the model.

EPRI states, and the staff agrees, that model users must determine the appropriate valve sealing load.

EPRI used the data obtained from its globe valve tests to validate its model. The EPRI test database is not sufficient to justify a modification to the Limitorque guidelines for sizing and setting globe valves to lower the typical valve factor of 1.1 assumed in the guidelines.

4. Butterfly Valve Model

a. Model Description The EPRI butterfly valve model calculates the stem torque required to open and close butterfly valves based on torque components acting on the stem and disk. The model provides predictions of the total required dynamic torque that must be applied to rotate the disk, the total required seating/unseating torque that must be applied to the valve stem to seat or unseat the disk, and the torque signature to demonstrate the capability of the model to approximately reproduce measured torque loading throughout the valve stroke.

The stem torque predictions are calculated from the torque components acting on the valve stem, including bearing torque, packing torque, seat torque, hydrostatic torque, hub seal torque, and hydrodynamic torque. The bearing torque is caused by the friction force between the stem and the bearings. The packing torque is caused by the friction between the stem and the packing.

The seating torque is caused by the contact pressure between the disk and sealing material and occurs only at disk positions near fully closed. The hydrostatic torque is caused by a difference in the hydrostatic pressure across the valve when the valve is fully closed when water levels are different on the upstream and downstream sides of a valve in a horizontal pipe. The hub seal torque is caused by the friction force that remains between the disk hub and the seat when symmetric disk designs with elastomeric seats are opening. The hydrodynamic torque is caused by the hydrodynamic forces acting on the disk when there is flow through the valve.

The torque required to operate the valve is the larger of the total required dynamic torque and the total required seating/unseating torque. The model also calculates a maximum transmitted torque for use in structural evaluation of MOV components: the larger of the torque required to operate the valve and the maximum hydrodynamic torque.

21

EPRI validated the butterfly valve model using flow data from eight valves from its Flow Loop Test Program, four valves from its In-Situ Test Program, and three valves from the INEL Purge Valve Test Program. EPRI compared the torque predictions calculated by the model for total dynamic torque and individual torque components against the test data. Based on its validation testing, EPRI considers the butterfly valve model applicable to butterfly valves with no-seat or interference-type elastomeric seats, symmetric disks, single-offset with prismatic or conical back face and flat front face, aspect ratios up to 0.35 for symmetric and single-offset disks in compressible and incompressible flow, and aspect ratios up to 0.47 for single-offset disks in incompressible flow with the shaft upstream.

To implement the butterfly valve model, the computer model user verifies the applicability of the model. The model user obtains valve-specific information from the manufacturer or by disassembly of the valve. EPRI requires that a static diagnostic test be conducted to confirm the bounding values for packing torque, hub seal friction torque and, for normally closed valves, unseating torque. The model user can select default values or may provide user-input information. EPRI states that computer model users are responsible for the accuracy of any input data and the subsequent effect on the accuracy of the model prediction.

The butterfly valve model calculates total predicted dynamic torque and its components at every I degree of disk rotation from fully open (90 degrees) to fully closed (seating). A summary table provides required total dynamic torque, required total seating/unseating torque, required actuation torque, and maximum transmitted torque. The computer model user can request the model to generate detailed output tables for each incremental stroke position. EPRI indicates that the total torque values in the detailed tables are to be considered best estimate and not necessarily representative of design-basis torque requirements.

b. Model Evaluation The NRC staff and its contractor evaluated the EPRI butterfly valve model by reviewing EPRI documentation of the model, operating the computer software, and discussing the model with NEI, EPRI, and the EPRI TAG utility members. On August 16, 1995, the NRC staff provided written comments to NEI on the EPRI butterfly valve model. On October 13, 1995, NEI submitted EPRI's response to the staff comments.

Based on NRC staff and contractor review of the EPRI reports, EPRI responses to the staff comments, and discussions between the staff and EPRI, the staff concludes that EPRI has presented sufficient support to justify application of the butterfly valve model to butterfly valves within the stated scope of the EPRI program.

22

(See NRC staff comment 2 on butterfly valve model and EPRI response.)

While reviewing the data, the staff determined that the nondimensional torque extrapolation factor adequately assessed the response of butterfly valves for aspect ratios of 0.25 and below.

For aspect ratios above 0.25, the torque coefficient varied with flow throughout the valve stroke. The butterfly valve model includes an adjustment factor to account for the effect of higher aspect ratios on the torque coefficient, and this factor appears adequate for normal system flowrates. Data are not available to verify a similar valve response for blowdown conditions. Although the limited data indicate that the torque coefficient decreases with increasing flow (tending to overpredict torque requirements at higher flow conditions), the higher flow might create fluid conditions that cause the torque response to become nonlinear.

For example, this phenomenon was observed at low flowrates for some valves. As part of its ongoing program to maintain the methodology, EPRI is expected to address any new information on butterfly valve blowdown performance in confirming its butterfly valve model.

The butterfly valve model is sensitive to errors in the output of the System Flow Model. EPRI requires in the implementation guide that the computer model user compare the calculated maximum differential pressure and flowrate to the design-basis differential pressure and flowrate for the MOV being analyzed and to confirm that the differences are small. (See NRC staff comment 4 on butterfly valve model and EPRI response.)

The butterfly valve model assumes that the seat material of the butterfly valve is in good condition. EPRI recognizes that seating material of butterfly valves may degrade (lose flexibility) and cause higher seating or unseating torque than the torque calculated by the model. For normally open valves, degradation of the seat material may prevent a leak-tight seal when the valve is closed. For normally closed valves, degradation of the seat material can increase the unseating torque to the extent that the actuator cannot open the valve. EPRI requires the computer model user to perform a static unseating test for normally closed valves to verify the assumed value of unseating torque. The staff considers that the information from the EPRI program emphasizes the need for a butterfly-valve preventive maintenance program that addresses potential seat material degradation. (See NRC staff comment 5 on butterfly valve model and EPRI response.)

As with thrust requirements for gate and globe valves, EPRI notes that the user of the butterfly valve computer model must account for any measurement uncertainty when evaluating torque requirements. (See NRC staff comment 6 on butterfly valve model and EPRI response.)

23

The EPRI program has revealed several areas where the EPRI Butterfly Valve Application Guide needs improvement or correction.

EPRI is currently revising the Application Guide and plans to include new information on flow and torque coefficients; system analysis techniques; treatment of bearing, packing, and hub-seal torque; upstream elbow modeling; and rated and survivable torque calculations. (See NRC staff comment 7 on butterfly valve model and EPRI response.)

c. Conditions/Limitations As part of its ongoing program to maintain the methodology, EPRI is expected to address any new information on butterfly valve blowdown performance in confirming its butterfly valve model.

EPRI requires in the implementation guide that the user of the butterfly valve computer model compare the calculated maximum differential pressure and flowrate to the design-basis differential pressure and flowrate for the MOV being analyzed and to confirm that the differences are small. The staff agrees with this limitation.

EPRI requires the computer model user to perform a static unseating test for normally closed valves to verify the assumed value of unseating torque. The staff considers the information from the EPRI program to emphasize the need for a butterfly-valve preventive maintenance program that addresses potential seat material degradation.

As with thrust requirements for gate and globe valves, EPRI notes, and the staff agrees, that the user of the butterfly valve computer model must account for any measurement uncertainty when evaluating torque requirements.

5. Motor Actuator (Load Sensitive-Behavior)

a. Method Description The EPRI MOV Performance Prediction Program focused on the thrust or torque required to operate gate, globe, and butterfly valves.

After predicting the thrust or torque operating requirements, the user must determine whether the output of the actuator is sufficient to overcome flow and adequately seat the valve. The user is responsible for demonstrating that the motor actuator has sufficient capability to operate the gate, globe, or butterfly valve under design-basis degraded-voltage conditions. For valves with a motor actuator controlled by a torque switch, the user must also demonstrate that the motor actuator can deliver sufficient thrust at torque switch trip toovercome the thrust required to operate the valve under design-basis conditions. The EPRI program does not provide guidance on predicting the thrust and torque output capability of motor actuators under design-basis degraded-24

voltage conditions, but does provide methods for predicting thrust output at torque switch trip under dynamic conditions.

For gate and globe valves with motor actuators controlled by the torque switch, EPRI specifies that the user of the program obtain a thrust output at torque-switch trip from a diagnostic test of the valve under static conditions. The thrust delivered by motor actuators to operate gate or globe valves under dynamic conditions can be less than the thrust delivered under static (no pressure or flow) conditions at a particular torque output. This phenomenon has been typically referred to as a "rate-of-loading" effect or "load sensitive behavior."

Depending on the specific stem, the coefficient of friction at torque switch trip in a static test may not be representative of the dynamic coefficient of friction at flow isolation. The coefficient of friction under dynamic conditions may be higher and can result in less thrust being delivered by the motor actuator for a particular torque output than delivered under static conditions. Therefore, it is important to determine the thrust that will be delivered under dynamic conditions when MOV design-basis capability is verified.

EPRI discusses six methods for users of the EPRI program to predict the reduction in thrust output under design-basis conditions from the thrust measured by diagnostic testing under less severe conditions. Method 1 involves the application of a correction factor to the thrust measured at torque switch trip during a static test. Method 2 involves the application of a correction addend to the stem friction coefficient measured at torque switch trip during a static test. In Method 3, the user applies a linear correction factor (based on the stem thread pressure achieved at initial wedging during a test) to the thrust measured at torque switch trip. The correction factor in Method 3 is eliminated if a minimum specified stem thread pressure is achieved. In Method 4, the user applies a linear correction addend to the stem friction coefficient measured at torque switch trip; the addend is based on the stem thread pressure achieved at initial wedging during a test. The correction addend in Method 4 is eliminated if a minimum specified stem thread pressure is achieved. Method 5 involves use of a reduction gear to apply the load over a minimum specified time interval prior to tripping the torque switch. Method 6 involves use of a handwheel to reduce the loading rate at torque switch trip. In Method 6, the thrust at torque switch trip using a handwheel is reduced by a speci'fied percentage to account for rate-of-loading effects.

25

b. Method Evaluation The NRC staff and its contractor evaluated the EPRI methods for addressing load sensitive behavior by reviewing EPRI documentation and discussing the methods with NEI, EPRI, and the EPRI TAG utility members. On July 17, 1995, the NRC staff provided written comments to NEI on the EPRI rate-of-loading methods. On September 15, 1995, NEI provided EPRI's response to thB staff comments. NEI submitted EPRI's revised response on November 14, 1995.

The staff considers EPRI's justification for the bounding assumptions for load sensitive behavior provided in EPRI Methods I and 2 to be limited. However, based on licensee test information, the staff believes that the 25-percent rate-of-loading margin and the approximate 30-percent stem-friction-coefficient increase of Methods I and 2, respectively, will bound most MOVs. Licensees are developing justification for their assumptions regarding load sensitive behavior (rate-of-loading) as part of MOV programs implemented in response to GL 89-10. Licensees may justify more appropriate assumptions for load sensitive behavior based on plant-specific data. (See NRC staff comment I on rate-of-loading methods and EPRI response)

The staff does not object to Methods 3 and 4, which involve adequate stem thread pressure to reach a stable stem friction coefficient. (See NRC staff comment 2 on rate-of-loading methods and EPRI response.)

The staff does not consider EPRI to have presented sufficient data to justify Methods 5 (reduction gear) and 6 (handwheel).

Licensees will be responsible for justifying their use of these methods. (See NRC staff comment 3 on rate-of-loading methods and EPRI response.)

EPRI states that Methods 1, 2, and 3 are not to be applied to globe valves without confirmatory data. (See NRC staff comment 4 on rate-of-loading methods and EPRI responses.)

EPRI states that moderate lubricant degradation can affect the determination of load sensitive behavior. EPRI states that Methods 1 through 4 require a freshly lubricated stem/stem nut to ensure proper accounting for rate-of-loading effects. If tests to assess rate-of-loading effects are performed with a degraded lubricant, the application of correction factors developed in Methods 1 through 4 might not bound potential rate-of-loading effects. The basic prerequisite for application of any of the EPRI rate-of-loading methods is.that the user maintain the stem/stem nut lubricant in accordance with an appropriate preventive maintenance program. Users need to establish adequate lubrication programs in addressing load sensitive behavior. (See NRC staff comment 5 on rate-of-loading methods and EPRI response.)

26

EPRI provides guidance on uncertainties associated with rate-of-loading measurement and other uncertainties such as diagnostic error and torque switch repeatability. EPRI also notes that uncertainty of the diagnostic equipment of the user must be considered. (See NRC staff comments 7 and 9 on rate-of-loading methods and EPRI responses.)

The staff agrees with EPRI that the primary cause of load sensitive behavior is the change in lubrication boundary conditions in the stem/stem nut interface during various forms of MOV loading. This change can result in a higher thrust being measured in a static test than will be available under a design-basis flow condition. (See NRC staff comment 10 on rate-of-loading methods and EPRI response.)

As with torque-switch-controlled MOVs, motor actuator thrust output can be lower under dynamic conditions than under static conditions for limit-switch-controlled MOVs as a result of load sensitive behavior. The staff agrees with EPRI that limit-switch-controlled valves are also a concern with respect to their design-basis.stem/stem nut coefficient of friction. Low load operation, typical of a limit-switch-controlled static test, may not be a reliable method to determine design-basis stem/stem nut coefficients of friction. Extensive laboratory testing has shown that a minimum value of stem thread pressure must be achieved to stabilize the running coefficient of friction. Valve stem packing alone is typically insufficient to achieve a stable coefficient of friction. EPRI also states that the stem friction coefficient values determined from limit-switch-controlled static testing would not be appropriate in general. EPRI states that a stem friction coefficient expected under design-basis conditions must be specified in calculating the amount of thrust which can be generated for a given amount of torque. EPRI Method 4 might be helpful in establishing a stem friction coefficient for use in determining motor actuator output capability under design-basis conditions. (See NRC staff comment 11 on rate-of-loading methods and EPRI response.)

c. Conditions/Limitations The user is responsible for demonstrating that the motor actuator has sufficient capability to operate the gate, globe, or butterfly valve under design-basis degraded-voltage conditions.

The staff does not consider EPRI to have presented sufficient data to justify Methods 5 (reduction gear) and 6 (handwheel). Users will be responsible for justifying their use of these methods.

EPRI states, and the staff agrees, that Methods 1, 2, and 3 must not be applied to globe valves without confirmatory data.

27

EPRI states, and the staff agrees, that the basic prerequisite for application of any of the EPRI rate-of-loading methods is that the user maintain the stem/stem nut lubricant in accordance with an appropriate preventive maintenance program. Users need to establish adequate lubrication programs in addressing load sensitive behavior.

EPRI states, and the staff agrees, that uncertainties associated with rate-of-loading measurement together with other uncertainties such as diagnostic error and torque switch repeatability must be considered.

Stem friction coefficient and its changes with various MOV load scenarios must be considered for both torque-switch- and limit-switch-controlled MOVs.

C. Specific Comments on EPRI Hand-Calculation Models In developing its computer model for gate valves, EPRI determined that the design of some gate valves used extensively in the nuclear industry was sufficiently different to prohibit their inclusion in the gate valve computer model. EPRI developed hand-calculation models to provide guidance for the calculation of predicted stem thrust requirements to operate four uniquely designed gate valves. These gate valves are (1) the Anchor/Darling double-disk gate valve, (2) the Westinghouse flexible-wedge gate valve, (3) the WKM parallel-expanding gate valve, and (4) the Aloyco split-wedge gate valve.

The NRC staff has reviewed the EPRI hand-calculation models for the Anchor/Darling double-disk gate valve and the Westinghouse flexible-wedge gate valve. The results of the staff review are provided below. The staff is reviewing the EPRI hand-calculation models for the WKM and Aloyco gate valves, and the results of the staff's review will be provided in a supplement to this safety evaluation.

1. Anchor/Darling Double-Disk Gate Valve Model

a. Model Description The Anchor/Darling double-disk gate valve has a disk assembly that includes an upper and lower wedge, an upper and lower wedge disk, a wedge spring, and two retainer clips. During valve closure, the lower wedge contacts a bridge (stop) in the bottom of the'valve body. Further closing of the valve results in the upper wedge sliding outward from the lower wedge against the valve seat to isolate flow. Upon valve opening, the disk assembly may unwedge or remained wedged depending on the specific characteristics of the valve.

28

EPRI developed a hand-calculation model for predicting the required stem thrust to open and close the Anchor/Darling double-disk gate valve. The approach is typical of the other EPRI gate valve models, where the individual force terms are combined to obtain the total predicted stem thrust. The disk assembly and stem weight, packing friction, piston effect load, and torque reaction friction are determined in the same manner as in the computer model. The determination of the differential-pressure stem thrust term is different for the Anchor/Darling double-disk gate valve because of its unique disk-seat configuration. The equations for determining the differential-pressure stem thrust depend on the upstream or downstream orientation of the disk assembly.

EPRI based its validation of the hand-calculation model on tests of two Anchor/Darling double-disk gate valves. One valve was tested in a flow-loop facility under ambient water conditions with simulated pumped flow, hot-water blowdown conditions, and steam blowdown conditions. The other valve was tested in-situ in an auxiliary feedwater steam supply system at a pressurized-water nuclear plant. EPRI did not report any damage to the valves during testing.

EPRI considers the hand-calculation model applicable for Anchor/Darling double-disk gate valves with Stellite 6 hardfacing at the disk-to-seat interface, water or steam flow, flow in either direction, and upper wedge to lower wedge interface materials of either Stellite 6 hardfacing or base metal (carbon or stainless steel).

Anchor/Darling noted several disagreements and concerns regarding EPRI's hand-calculation model. Anchor/Darling stated that it recommends that the lower wedge be located on the downstream side of the valve, but that most of EPRI's validation testing had the valve oriented with the lower wedge upstream. Anchor/Darling believes that EPRI has misinterpreted the thrust traces for the lower wedge upstream cases and that the hard seating thrust is lower than determined by EPRI. Anchor/Darling considers EPRI's disk-seat friction coefficients to be overly conservative based on a comparison to Anchor/Darling test data. Anchor/Darling believes that the conservatism in the EPRI hand-calculation model will result in excessive thrust loadings in many cases. A more detailed description of the Anchor/Darling disagreements and concerns is provided in the EPRI reports.

b. Model Evaluation The NRC staff and its contractor evaluated the EPRI hand-calculation model for Anchor/Darling double-disk gate valves by reviewing EPRI documentation and discussing the model with NEI, EPRI, and the EPRI TAG utility members. On September 21, 1995, the NRC staff provided written comments to NEI on the EPRI model 29

for Anchor/Darling double-disk gate valves. On November 6, 1995, NEI provided EPRI's response to the staff comments.

EPRI relied on its separate effects testing of friction coefficients and limited testing of Anchor/Darling double-disk gate valves in justifying the reliability of its hand-calculation model for these valves. In response to a staff concern regarding the extent of valve test data, EPRI discussed informathon from industry testing that provided additional support for its hand-calculation model. The staff notes that limited data exist to support the Anchor/Darling double-disk gate valve model and that the model is highly conservative in some cases and marginal in other cases. Based on NRC staff and contractor review of the EPRI reports, the EPRI responses to staff comments, discussions between the staff and EPRI, and the supporting industry information, the staff does not object to the use of the EPRI hand-calculation model in predicting the thrust required to operate Anchor/Darling double-disk gate valves. EPRI is expected to address new information as appropriate that becomes available with respect to the application and implementation of the model. For example, EPRI should consider the need for valve damage criteria in light of recent industry testing that revealed damage during a blowdown test of an Anchor/Darling double-disk gate valve. Because of the limited data available for validating the hand-calculation model for Anchor/Darling double-disk gate valves, model users are expected to compare their available data, if any, for these valves when applying the hand-calculation model. The staff also emphasizes that model users are responsible for establishing adequate margin in sizing and setting their valves. (See NRC staff comment I on Anchor/Darling double-disk gate valve model and EPRI response.)

Anchor/Darling contends that the EPRI hand-calculation model may be overly conservative in some instances. The staff agrees that the EPRI model may overpredict the thrust requirement to operate some Anchor/Darling double-disk gate valves. The staff considers EPRI's response to the Anchor/Darling concerns to be reasonable.

(See NRC staff comment I on Anchor/Darling double-disk gate valve model and EPRI response.)

The design of Anchor/Darling double-disk gate valves provides flow isolation at the differential-pressure condition present during valve operation when the disk fully overlaps the seat. If the differential pressure is reduced following valve closure, the valve may leak if the disk is not wedged into the seat. When applying the EPRI-hand-calculation model for Anchor/Darling double-disk gate valves, the model user will need to ensure that sufficient wedging is achieved to provide leak-tightness according to design requirements. EPRI indicates a theoretical flow isolation point on several data traces. The staff does not consider EPRI to have provided sufficient justification to establish those points as actual flow isolation. (See NRC staff 30

comment 2 on Anchor/Darling double-disk gate valve model and EPRI response.)

EPRI uses the friction coefficient information determined from the separate effects testing performed in developing the gate valve computer model. However, EPRI has grouped the coefficients over the temperature range to simplify the hand calculations for the Anchor/Darling double-disk gate valves. This results 4n slightly higher friction coefficients in some instances, such as for high-temperature conditions. The staff considers that this grouping of friction coefficients is more reflective of the scatter that has been observed in test data and may be more reliable than the coefficients used in the gate valve computer model. (See NRC comment 3 on the Anchor/Darling double-disk gate valve model and EPRI response.)

In addition to disk-to-seat friction coefficients as addressed in other gate valves, EPRI had to establish friction coefficients for the interaction of the internal disk wedge mechanism of the Anchor/Darling double-disk gate valve. Because these forces are less significant than the disk-to-seat friction forces, EPRI applied engineering judgment to estimate the wedge friction coefficients. The overall verification of the hand-calculation model supports the assumptions made by EPRI for the wedge friction coefficients. (See NRC staff comment 4 on Anchor/Darling double-disk gate valve model and EPRI response.)

EPRI recognized the need to precondition the Anchor/Darling double-disk gate valves to achieve the plateau of the disk-to-seat friction coefficient. Of the two Anchor/Darling double-disk gate valves it tested, EPRI believes that full preconditioning was achieved for the flow loop valve, but not the in-situ test valve.

EPRI took this into consideration in evaluating the reliability of its hand-calculation model. EPRI subsequently withdrew its use of the in-situ test valve for model validation because of uncertainty surrounding the test data. (See NRC comments 1 and 5 on the Anchor/Darling double-disk gate valve model and EPRI responses.)

EPRI found that the wedging mechanism in the Anchor/Darling double-disk gate valves can remain in a locked condition when the valve is opened. During one test by EPRI, the wedging mechanism of the Anchor/Darling double-disk gate valve remained locked.

EPRI states that the hand-calculation model will predict whether the wedging mechanism will remain in the locked condition during valve opening. (See NRC staff comment 6 on Anchor/Darling double-disk gate valve model and EPRI response.)

EPRI indicates that the orientation of the lower wedge either upstream or downstream does not affect the thrust required to reach flow isolation during the test of that valve at that differential-pressure condition. However, the orientation of the lower wedge can affect the thrust required to reach hard-seating 31

of the valve. Specifically, more thrust is required for hard-seating with the lower wedge upstream. Further, a valve not hard-seated might leak if differential pressure is subsequently lowered. Anchor/Darling recommends that the valve be installed with the lower wedge downstream. EPRI states that the model user is responsible for determining the orientation of the wedge mechanism. The wedge mechanism can be rotated to reverse the orientation without rotating the valve. Therefore, ma-intenance needs to be properly controlled when the valve is disassembled and reassembled. If the orientation of the lower wedge is unknown, EPRI recommends that the thrust be predicted assuming the lower wedge is upstream (higher thrust orientation). (See NRC staff comment 8 on Anchor/Darling double-disk gate valve model and EPRI response.)

c. Conditions/Limitations EPRI is expected to address new information as appropriate that becomes available with respect to the application and implementation of the model. EPRI is also expected to address the need for damage criteria in the model.

Because of the limited data available for validating the hand-calculation model for Anchor/Darling double-disk gate valves, model users must compare their available data, if any, for these valves when applying the hand-calculation model.

The staff emphasizes that model users are responsible for establishing adequate margin in sizing and setting their valves.

When applying the EPRI hand-calculation model for Anchor/Darling double-disk gate valves, the model user will need to ensure that sufficient wedging is achieved to provide leak-tightness according to design requirements. The staff does not consider EPRI to have provided sufficient justification to establish the points identified in its data traces as flow isolation.

EPRI states that the hand-calculation model will predict whether the wedging mechanism will remain in the locked condition during valve opening. The model user will need to address potential locking of the wedging mechanism if predicted by the model.

The model user is responsible for determining the orientation of the lower wedge mechanism.

2. Westinghouse Flexible-Wedge Gate Valve Model

a. Model Description The Westinghouse flexible-wedge gate valve has a stem-to-disk connection and guide rails that differ from those in other flexible-wedge gates. The stem disk assembly includes a double-32

pinned linkage that allows the disk to translate relative to the stem in a direction parallel to fluid flow. Two guide rails are installed in parallel slots in the valve body to guide the disk during opening and closing strokes. The link connection used in Westinghouse gate valves exhibits a tendency for the disk to remain in its current tipped or untipped configuration, compared to designs that use a T-head and a T-slot connection. For example, the disk will tend to remain in a tipped configuration when achieved.

EPRI developed a hand-calculation model for predicting the required stem thrust to open and close the Westinghouse flexible-wedge gate valve. The model uses equations from the EPRI gate valve computer model and includes specific additional equations developed for the unique Westinghouse valve stem and disk assembly in the valve closing direction. The calculations are performed only at key stroke positions rather than at 1-percent stroke intervals in the gate valve computer model.

The model evaluates each term that contributes to stem thrust and then sums the contributions to calculate the total required stem thrust. The terms considered are disk assembly and stem weight; packing friction; stem rejection load; forces due to differential pressure applied across the disk; and a torque reaction factor.

For opening strokes, the unwedging thrust is also calculated. The required thrust to open the valve is the larger of the thrust calculated by the model and the unwedging thrust.

EPRI used test data from six Westinghouse gate valves to support the validation of the model. One valve was tested in a flow loop and five valves were tested in-situ at power plants. The testing found the model to have bounded the data with the thrust for one valve determined to be unpredictable due to stem buckling caused by excessive lateral load. EPRI considers the hand-calculation model to be applicable to Westinghouse flexible-wedge gate valves with Stellite hardfacing at the disk-to-seat interface, Stellite hardfacing on the guide slots, and 17-4 PH guide rails in water or steam flow.

Westinghouse reviewed the EPRI methodology for its flexible-wedge gate valves and the EPRI test reports. Westinghouse determined that the EPRI methodology is consistent with the Westinghouse information on valves of similar size and design. Westinghouse stated that the EPRI analytical modeling of Westinghouse valve performance appeared to be conservative. Westinghouse indicated that its personnel did not attempt to verify the accuracy of all of the equations in the EPRI model. Westinghouse indicated that its review should not be considered an endorsement of the EPRI model as the only method for evaluation of operating loads for Westinghouse valves.

33

b. Model Evaluation The NRC staff and its contractor evaluated the EPRI hand-calculation model for the Westinghouse flexible-wedge gate valve by reviewing the EPRI documentation of the model and discussing the model with NEI, EPRI, and the EPRI TAG utility members. On September 21, 1995, the NRC staff provided written comments to NEI on the EPRI hand-calculation model for Westinghouse flexible-wedge gate valves. On November 6, 1995, NEI submitted EPRI's response to the staff comments.

EPRI justified the reliability of the hand-calculation model by conducting limited testing of Westinghouse flexible-wedge gate valves and evaluating the differences between the gate valve computer model assumptions and the design of the Westinghouse valves. The staff notes that limited data exist in supporting the Westinghouse flexible-wedge gate valve model and that the model is highly conservative in some cases and marginal in other cases.

Based on NRC staff and contractor review of the EPRI reports, the EPRI responses to staff comments, discussions between the staff and EPRI, and supporting statements by Westinghouse, the staff does not object to use of the hand-calculation model in predicting the thrust required to operate Westinghouse flexible-wedge gate valves. EPRI is expected to address new information as appropriate that becomes available with respect to the application and implementation of the model. Because of the limited data available for validating the hand-calculation model for Westinghouse flexible-wedge gate valves, model users are expected to compare their available data, if any, for these valves when applying the hand-calculation model. The staff also emphasizes that model users are responsible for establishing adequate margin in sizing and setting their valves. (See NRC staff comments 1 and 2 on the Westinghouse flexible-wedge gate valve model and EPRI responses.)

When the hand-calculation model predicts that the Westinghouse valve may be damaged, the model assumes a friction coefficient of 1.75 to allow the calculations to continue. However, the thrust requirement for operating the valve is considered "unpredictable" by EPRI. As with the gate valve computer model, the thrust requirement predicted under these conditions is not to be used for design-basis requirements. (See NRC comment 5 on Westinghouse flexible-wedge gate valve model and EPRI response.)

EPRI found that unwedging loads under dynamic conditions can be higher than static unwedging loads. EPRI has developed a hand-calculation method to predict unwedging loads, as discussed earlier for computer model users. The user of the hand-calculation model is expected to compare available plant data to the results of EPRI's method for calculating unwedging load. (See NRC staff comment 7 on Westinghouse flexible-wedge gate valve model and EPRI response.)

34

c. Conditions/Limitations EPRI is expected to address new information as appropriate that becomes available with respect to the application and implementation of the Westinghouse flexible-wedge gate valve model.

Because of the limited data available for validating the hand-calculation model for Westinghouse flexible-wedge gate valves, model users must compare their available data, if any, for Westinghouse flexible-wedge gate valves when applying the hand-calculation model.

The staff emphasizes that model users are responsible for establishing adequate margin in sizing and setting their valves.

When the model indicates that the thrust requirement is unpredictable, the thrust requirement predicted under these conditions is not to be used for design-basis requirements.

As discussed earlier for computer model users, the user of the hand-calculation model should compare available plant data to the results of EPRI's method for calculating unwedging load.

3. WKI4 Parallel-Expanding Gate Valve Model TO BE DISCUSSED IN SUPPLEMENT TO SAFETY EVALUATION

4. Aloyco Split-Wedge Gate Valve Model TO BE DISCUSSED IN SUPPLEMENT TO SAFETY EVALUATION III. CONCLUSIONS The EPRI program represents the most extensive effort to date to evaluate and predict MOV performance. The NRC staff considers the EPRI program to provide significant improvements for the design and setting of MOVs for nuclear power plants. The staff agrees with EPRI in the application and implementation of most aspects of the EPRI MOV Performance Prediction Program. In this safety evaluation, the staff discussed specific elements of the EPRI program and noted any limitations or conditions associated with the application or implementation of the EPRI models and methods.

With the limitations and conditions described herein, the staff considers the EPRI NOV Performance Prediction Program to provide an acceptable methodology to predict the thrust or torque required to operate gate, globe, and butterfly valves within the scope of the EPRI program, and to bound the effects of load sensitive behavior on motor-35

actuator thrust output. As with any industry model or method, EPRI will be responsible for maintaining and updating the models and methods described in its topical report as necessary in light of new information or test results. EPRI will be expected to notify the staff in the event that EPRI identifies information or test results that might affect the staff's conclusions, limitations, or conditions regarding the EPRI Performance Prediction Program discussed in this safety evaluation.

As discussed in this safety evaluation, EPRI was unable to provide sufficient test data or information to validate certain applications of specific models or methods for use as design standards. EPRI or individual licensees may obtain additional information to support the use of the models or methods as design standards in those applications.

The staff will consider requests to update this safety evaluation in light of new information. In the interim, a licensee might use a specific model or method in an application not validated as a design standard based on the significant improvements in the understanding of MOV performance provided by the EPRI program over past MOV design practices.

IV. LESSONS LEARNED The research conducted by EPRI significantly improves the understanding of MOV behavior and the ability to predict MOV performance. The EPRI program provided important information on the design, testing and maintenance of MOVs in nuclear power plants. In the overview section of this safety evaluation, the staff listed many of the important findings (and confirmatory information) of the EPRI program, as indicated by NEI.

Some of the EPRI information is applicable to gate, globe, and butterfly valves regardless of the type of actuator operating the valve. For example, the EPRI program revealed that thrust required to operate gate and globe valves may be greater than predicted by typical industry methods.

Information from the EPRI program may influence the establishment and implementation of MOV programs at some nuclear power plants. Further, the information may require nuclear power plant licensees or construction permit holders to reevaluate the performance capability of some MOVs. Examples of such information are given below:

Gate Valves Almost all flow testing conducted by licensees in response to GL 89-10 was conducted under pumped-flow conditions. The GL 89-10 test results were then extrapolated to design-basis conditions assuming predictable valve behavior and no internal valve damage. Therefore, extrapolation of test data from pumped-flow conditions to blowdown conditions (as typically performed in MOV programs) might not be sufficient to ensure that a gate valve can-operate under its design-basis conditions.

36

j Valve aging conditions can influence gate valve performance. The thrust requirements to operate gate valves under normal flow conditions can increase with time and valve stroking and might eventually cause the MOV not to operate properly.

Thrust requirements to unwedge gate valves under dynamic conditions may be greater than under static conditions.

Proper consideration of "parasitic" loads requires assurance that these loads are constant over the entire stroke of the valve and are not affected by differential pressure.

Globe Valves Based on limited testing by EPRI, higher thrust than typically predicted using the EPRI model or standard industry methods might be required to operate globe valves under blowdown or high-temperature flow conditions.

Thrust requirements for globe valves are influenced by the area of the valve seat or guide, depending on the valve design.

The EPRI test database is not sufficient to justify modifying the Limitorque guidelines for sizing and setting globe valves to lower the typical valve factor of 1.1 assumed in the guidelines.

Butterfly Valves Several areas of the EPRI Butterfly Valve Application Guide need improvement or correction. EPRI is currently revising the Application Guide and plans to include new information on flow and torque coefficients; system analysis techniques; treatment of bearing, packing, and hub-seal torque; upstream elbow modeling; and rated and survivable torque calculations.

An appropriate preventive maintenance program for butterfly valves includes consideration of potential seat material degradation.

Load Sensitive Behavior The licensee remains responsible for demonstrating that the motor actuator has sufficient capability to operate the gate, globe, or butterfly valve under design-basis degraded-voltage conditions.

An adequate lubrication program is important in addressing load sensitive behavior.

Uncertainties associated with rate-of-loading measurement and other uncertainties such as diagnostic error and torque switch repeatability need to be appropriately considered.

37

Stem friction coefficient and its variation with various MOV load scenarios need to be considered for both torque-switch- and limit-switch-controlled MOVs.

Principal Contributors:

Thomas G. Scarbrough, NRR Dr. Gerald H. Weidenhamer, RES Robert Steele, INEL 38