ML20063A340

Forwards Addl Info Re Feedwater Check Valves Functioning Following Line Break Outside Containment,In Response to Mechanical Engineering Branch Draft SER Question 3.9-42

ML20063A340
Person / Time
Site:	Perry
Issue date:	08/16/1982
From:	Davidson D CLEVELAND ELECTRIC ILLUMINATING CO.
To:	Schwencer A Office of Nuclear Reactor Regulation
References
NUDOCS 8208240296
Download: ML20063A340 (11)

Text

.,

l THE CLEVELAND ELECTRIC ILLUMIN ATING COMPANY l P o Box 5000 m CLEVELAND. ohio 44101 e TELEPHONE (216l 622-9800 m ILLUMINATING BLOG e 55 PUBLIC SOUARE Serving The Best Location in the Nation Dalwyn R. Davidson vict rRESiot NT August 16, 1982 SYSTE M ENGINEERING AND CONSTRUCTION Mr. A. Schwencer, Chief Licensing Branch No. 2 Division of Licensing U. S. Nuclear Regulatory Commission Washington, D. C. 20555 Perry Nuclear Power Plant Docket Nos. 50-440; 50-441 Revised DSER response Mechanical Engineering Branch

Dear Mr. Schwencer:

This letter is to provide additional information relative to MEB DSER Question No. 3.9-42 regarding the feedwater check valves functioning following a line break outside of containment.

It is our intention to update the feedwater check valve design information in a future amendment to our Final Safety Analysis Report.

Very truly yours, ti-Dalwyn R. Davidson Vice President System Engineering and Construction DRD:mb cc: Jay Silberg, Esq.

John Stefano, NRC Max Gildner, NRC Resident Inspector John Hilbish, gal Dave Shen, GE David Terao, NRC (kol 820 8240 N E

l DSER 3.9-42 The applicant to provide assurance that the FW check valves can function following a line break outside containment.

Response

The feedwater check valves are designed to function following a line break outside containment. The closure rate of the valves is designed to control the deceleration rate of the water in the reverse flow direction when the initial upstream pressure is 1100 psia and downstream pressure is 14.7 psia. The closure rate of both check valves shall be dampened so that the deceleratjon rate of the water in the reverse flow direction shall not exceed 20,000 lbm/sec . When the initial upstream pressure is 1100 psia, the downstream pressure is 14.7 psia and the feedwater flow rate to the reactor is 17,840 gpm. The peak pressure associated with valve closure under the conditions identified above shall not exceed 1500 psig.

Verification of operability of check valves shall be accomplished by injecting water into the dampening chamber to force the valve open and measuring closure time as detected by the audible impact of the piston against the seat.

The dashpot check valve operation is described in attached ASME paper 80-C2/PVP-27.

FSAR Figure 5.1-3 and Note 22 to Table 3.2-1 will be revised to reflect this valve arrangement in the next amendment.

i

.. w a : ; ,, .

M2./@U -

y s.. -

'THE AMERiCAN SOCIETY OF MLCHAMCAL ENGINE 8ERS J 345 E 47 St. New York. N.Y.1C317

_s aa__.m,_s. _- o. ,_s_ ..._. _. ._._

h. prMacea.3fet. o *+et. ONy fr rh. pep.r is pubhshed en  !

-_r_dchf. dog #1

- .-. . . - .-.~. .-....- p

_ _ _,_. _att A SAf. f l

m.a.xirik A Model for Check Valve /Feedwater s_,, --

System Waterhammer Analys.is R. J. Gradle Quick shutof ofa rupturedfeedwater line by a convrnitoral check valve can generate Researen frn,neer damaging surge pressure. Ilowever, slow cicsure by a conare fled closure check valve can Assoc.Mont ASME often reduce surge pressure to a tolerable level. Sir.cc a c%eckylye is self acting, the magnitude of the surge pressure depends on cine u/ ft fred &r systerr -teraction.

Apphed manacs. A n analyticalmodelofa check valve and a simphf.%f adwater;> stem vaer linebreak

• "*

• conditions is presented. This model is used to dciermine the valve dashpot capacity RockwellIntematnonal Corp.,

necessary to reduce the surge pressure to an acceptable value. Datafrom the model are p,nsy,,g pg. compared to measurementsfrom an instrumemed line rupturefacility. Results of this comparison are discussed INTRODUCTION the design of such a dashpot is not a simple task.

Although the mechanical configuration of a dashpot For many years, conventional check valves hpve may be easy to define, the sizing of the thrattling been used in feedwater lines to provide protection orifices or flow resistance devices to obtain proper against effects of reverse flow during normal plant valve closure time involves the solution of complex operations. These valves include tilting-disk, equations which are dependent on the characteristics swing and piston type check valves, and are designed

• of the valve installation. Interaction between the to close rapidly before a significant level of check valve and the rest of the feedwater system reverse flow can be established, during valve closure requires the simultaneous In recent years, attention has focused on the solution of the equations for waterhammer, valve flow postulated double-ended line rupture problem in rate, and check element motion to determine if the feedwater lines in nuclear power plants. Aaalysis dashpot is adequately sized, indicates that the rapid closure typical of This paper shows that an analytical model of a conventional check valves in a line rupture check valve with dashpot can be developed in con-situation can result in large, damaging surge junction with the waterhamer equations for the pressures (waterhansner). Line rupture tests by feedvater system to predict surge pressure data with ASEA-ATOM [1] in 1973 on a simulated feedwater reasonable accuracy, and thus indicate that the system using a tilting-disk check valve, pressurized dashpct is correctly sized. Because of the complexity cold water, and rupture disks to initiate blowdown, of the model equations, a computer program employing confirm that surge pressures can reach damaging a numerical solution procedure was written to provide levels. In these tests, peak pressures of about surge pressure and valve behavior data.

4100 psia (280 bar) were measured for an initial ,Since the intent of the computer model is to line pressure of 1100 psia (75 bar). size dashpots for check valves to be installed in One way to insure that surge pressures stay feedwater systems, the model must have a b.* sis for below damaging levels is to substantially increase qualification, preferably experimental data obtained the closure time of the check valve [ 2 ] . In [3] for that purpose, to provide the maximum match-up it is shown that just a slight increase in valve betwee, model and test system. To this end, a test closure time can actually increase surge pressure; system was built to develop the experimental data

^

so increase in closure time more like an order of base required. Computer calculated surge pressure magnitude is required (e.g. closure time of about data based on the initial conditions defined by the one second; actual time depends on the installation). individual test runs could then be compared to The addition of a dashpot to a conventional actual test data to show that the analytical model check valve is a feasible means of increasing predicted test results with acceptable accuracy.

closure time by reducing closure speed. However. The valve used for the model and the test was a Y-pattern, piston type check valve as shown in Coatnbased by the Preneure Vessels & Piping Dmsson of the Amencan Saoety Fig. 1. Thia valve was chosen because it has a of M*W Engineers for presentauce at the Century 2 Pressure vessels & natural shape for the addition of an internal dashpot Prping Coaterence. San Franc sco. Cahf., August 12-l3.1980 Manuncnpt ruerved at ASME Headquaners Man 617.1980 in the valve neck. The dashpot is formed sieply by the insertion of a stationary annular baffle plate, Copies edl be svadable uaul May l.1981. and is protected by the valve body.

4 vs swee I/ s

\\ -

susus FEEDWATER LINE WTRMR CM'RE CIECK VALVE CONSTANT REVEREE FLOW

/ ,,

-hfN ,,N 9y l rREssuRt RESERVOIR i

p, m c

Nau / ]s ,n,.f \ \ '/

M/ \ ea PIPE ID NORMAL F14W f :::,

"UNDER$ EAT" gl%; M / LINE RUPTURI M*

ca we /

Fig. 2 Model of Feedwater System disk. The dashpot acts to reduce the closing speed

'{ .n.

8 of the check element by negating most of the

• jf unbalanced closing force through pressure build-up

[ (lt in the dashpot. Closure speed thus depends on the 1eakage rate of the liquid from the dashpot through

/ /

sized clearances at the pist'on and the dashpot plate.

/ g

' 7' Description of instantaneous check element notion, surge pressure and choked flow rate comes

\.*Um #)

vie .

' from the simultaneous solution of equations governing

• i I these phenomena. Also, valve cover pressure and dashpot pressure are calculated from equations

" h representing an accounting of the mass in these cavities, piston motion, and mass flow through the equalizer tube and dashpot clearances.

amine v.a vie mean Nn s.a Check Element Motion Equations Fig. 1 Model of Y-Pattern Piston Check Valve with Internal Dashpot Summing forces on the check element, shown in Fig. 1, and substituting in Newton's Second Law A'IALYTICAL MODEL 818 Description of the analytical model begins 2 IA '

with the feedwater system, then continues with the u d - P gA'd + PdA'p -PAcp +F+C= W equations governing check element motion. This is followed by the equations for the flow rate through where Pu = effective pressure under the disk, the valve and, finally, by the waterhammer equations.

The system basis for the analysis is a check Pg = pressure in the valve at the connected valve, constant-p{essurereservoir,andinter- line side, shsolute connecting piping as shown in Fig. 2. The feed-water line rupture is assumed to occur near the Pd = dashpot cavity pressure, absolute underseat end of the valve, which is opposite to ths reservoir end. The valve is considered as a Pc = valve cover cavity pressure, absolute csvitation-choked nozzle of ever-contracting area as it closes. Also considered are the inertia Ad = underside disk area af f ects (waterhammer) in the water column, which = w(disk OD)2/4 l

can rarduce the pressure entering the valve below vrpor pressure as the flow reverses, and can cause Ag = connection tube area i surges above the reservoir pressure as the valve 5

completes closure. Frictional effects in the pipe = s(tube OD) /4 are assumed to exert only minor influences during A'd = underside disk area valve closure and are neglected.

• The check valve with the dashpot, depicted in Fig. 1, functions, for an underseat line

=Ad-At rupture, by generating a closing force on the AP = upperside piston area check element due to a difference between the = w(piston OD)2f4 valve cover pressure and the pressure under the ,

3,p = underside piston area

• It is recognized that this is a simple and ideal- =Ap-Ag istic feedwater system; the intent of this paper is to show that a valve model can be developed which F = net friction force '

gives satisfactory results, not to develop models C = gravity force of complex f eedwater systems. Once an adequate valve model is obtained, it can be merged with a U = check element weight model of any feedwater system to estimate surge V = check element velocity prsssure data.

, g = acceleration of gravity t = time 2

m

1 l

l Chsch elsmsnt lif t cbova thz body seat (x) is given It is irporttnt to point oat that tha loss by: coefficients, Ke and Fe, for tic. clearances 21 the i dashpot cannot be citimated acr ntely ushg sudden I dr c ncracticn or calargetent data for ccicentric pipes. l V I2) D e effect of any edge chamfers and eccentricity of 7t~ =

the clearances is dif ficult to assess. For this e Jel, the loss coefficients vera measured using a Pressures and Forces tot fixture simulating the da Apot clearances at ng egrees eccen ci b,and f un t be The pressure-under-ihe-disk term, P , essem a mewhat larger than those for suelden area changes tially represents the effective value of the actual pressure distribution over thi. uaderside 4;ea of ' " #' '

• the disk for choked liqu U ficv thrcugh tne valve. dashpct clearances cannot be baaed on pipe data for This pressure is deterinined f rom the erpirical two reasons. First, the clearan s ficw surfaces do

• • I*EI "I not base random roughness as in pipes, The surfaces are machined and base ribbed surface roughness (ribs Eu = RP 3g (3) perpendicular to the flow) typical of turning or boring operations. Second, the clearances are where Rg is an experimentally deteredned function of eccentric to varying degrees anI this has an effect disk lift, disk geometry, body seat diameter, and on the flow resistance L 5] even though the flow area valve internal flow geometry. This function is remains constaat. Since the hydraulic diameter dees based on critical gas flow data and is assumed to nut M age with eccentricity, this cha ve in flow cdequately represent conditions under the disk, resistaace must be represented in the f ricticn Relations of this kind car, be developed usine, a factor. As with the loss coefficients, the friction slightly modified form of equatica (1) and data factor was measured experimentally using a test f rom steady, choked, reverse flow tests with the fixare with clearance surf aces r.arhimed to match check element held stationary, at various disk manufactuslag finishes. Measured f riction factors lif ts, by a load cell connected to the top of the were much lower than that for commercial steel pipe, piston. Except for the "Pud A " terr, data for all an2 decreased as the degree of eccentricity increased.

terms in equation (1) are measured incluling an Any attempt at mMeling machined clearances of this extra term for the load cell force (the acceleration type should rely on tests to denlop flow resistance j tem is zero). The ef'ectise disk pressure Pu is factors. ,

then calculated from equation (1). S petimental The oblique tip end of the equalicer tube (see results are given in [3,Q , the experimental inset of Fig. 1) provides an aspiratio7 effect for procedure is described in [4! . normal flow through the valve at the tube opening, Pre ssure in the dashpot chamba, Fe , and which causes the pressure there to be less than the pressure in the valve cover cavity, Pc , are stream static pressure. When flow reverses due to a datermined from equations relating preu ure drop line break, the reverse effert occurs causing, to flow through the dashpot clearances and the effectively, a stagnation pressure at the tube squalizer tube, and continuity of mass equations opening. Th1 pressure at the tube opening is given for these chambers. Censidering the two clearances approximately by:

(nd the equalizer tube, the pressure drop and flow o 2 valocities are related by (see tig. 1)- Pi * = Pg + g Vf (7)

Pd-Pi=p Yl lVll 1 (4) pg ige + x, + f .)3g where Vf is the fluid velocity in the valve inlet region.

V I Pc - Pd " D 2lV2 - --tKc+Ke + fL) D (5) Piston velocity and flow velocities in the

'o 2 elea nnees and the equalizer tube are related by the conservetion of mass law applied to the valve Pg* - Pc=p 2g

% + K, + fh (6) rover cavity and the dashpot chamber. This gives 3 two additional equations:

where p = liquid density VA p = V3 A) - yA; (8)

Ke = entrance flow loss coefficient ,

p=VAg3 -Vg (9)

Ke = exit flow loss coefficient f = flow path friction factor there V = pist% selocity, subscripts 1,2,3 refer L = path or clearance length to flow areas ano flow velocities asso-ciated with the dashpot baf fle plate D = path or clearance hydraulic diameter clearance, Fistoa clearance, and equalizer Pi* = pressure at oblique tip end of tube, respectively.

equalizer tube, absolute Friction force F is the productiof the net U 3 = flow path velocity norm 1 force on the check element in reaction to j = 1,2,3 denotes dashpot baffle plate bearing on the valve bore and the coefficient of clearance, piston clearance, and friction. This forde comes from two independent equalizer tube, respectively components, one due to flow forces and the other due to gravity or:

3

1 T = p (10) whera o = chok:d cc:s flow rets through th2 v:1v2 RP2 t + W sin 0 An - equivalent nov.le area of valve for where u a coefficient of friction for the disk / choked liquid flow, which is a function bore material combination of disk lift (see L7] for typical values).

R2

• experimentally determined function of Pg = instantaneous line pressure at valve disk lift and valve internal flow geometry; it is based on data for Pv = vapor pressure based on initial fluid critical gas flow and is assumed to be C""dIII#""

adequate for choked liquid flow. See r a metastability factor

[4] for values of this function and description of the experiment.

=

0.7 [73 A1 = inside area of connected pipe 0 = counter-clockwise angle from positive x-direction to gravity vector, see pgg, g, The denominator inside the radical represents the feedwater line velocity-of-approach effect.

The gravity force component in the x-direction Mass flow rate through the valve is related 15 the G term in e p ation (1) and is expressed by: to the line flow velocity (Vfg) by the continuity

• equation, or:

G = W cos O (11) y . An (P1 - TPv)/g ,

valve Flow Rate and Waterhammer 1-( ) D Two of the primary concerns in the development Fundamental waterhammer relations, based on elastic of the analytical model were determining the choked, water column theory [ 21 are:

revarse flow rate of the subcooled liquid, generally hssing a high vapor pressure, through the check E,E BVf v:lve during closure, and determining the state of 3y g W tN line fluid irmediately af ter line rupture.

Experimental data [6] indleates that af ter the E , Le c 2 BVf line rupture occurs, a pressure reduction wave travels at g Ty~

through the valve and up the connected line at the accurtic speed of the liquid and, generally, causes where y = distance from valve along the connected pressure to f all below vapor pressure. Initially, pgp, tha fluid takes on a metastable (superheated) state baccuse of insuf ficient time for significant vapor- P = instantaneous pipe pressure bubble growth. Thereafter (several milliseconds '

t = time sftst decompression wave passage), complications set in due to vapor-bubble growth of magnitudes which Vf = instan. neous fluid velocity v:ry along the pipe. This situation, fortunately' c = acoustic wave speed of fluid Icsts only for a small fraction of the valve clost e tias as valve closure compression waves and re- These equations have the general solution:

compression waves (from the reservoir) cause the prs 2sure to rapidly rise above vapor pressure = P1 (t collapsing any bubbles formed. It is assumed in P-P 1) c + p1(t + 1) e (13) this snalysis that the fluid remains a superheated 11 quia during the small time span from rupture until vcpor pressure is reached and that, because of thia yf _ y{ , _ L [P (t 1) oc L1 c - pI(t + 1)l cj (14) sas11 time span, the effect of this assumption on thn accuracy of the solution is likewise small.

where P - Po = surge pressure related to initial Calculating the choked, reverse flow rate of line pressure Po (before rupture) tubcooled water, generally having a high vapor prassure, through the closing valve 16 an approxi- Pg = pressure disturbance (from initial mate precedure chiefly due to metastability effects. line pressure) emanating from the Fluid pressure at the valve throat can drop well valve (a function of t - y/c) below vapor pressure before significant vapor-bubble = pressure disturbance (from initial bI growth occurs. This allows for a corresponding line pressure) returning to the valve increase in mass flow rate over that for vapor (a function of t + y/c) pre:sure occurring in the valve throat. As the difference between vapor pressure and the (lower) V,

'f = initial flow velocity (before rupture).

throat pressure increases, there is an increased potsntial for vapor formation. Eventually, as this It can be showa [ 2 ] that when the check valve prsssure dif fe rential grows, vapor will form and is connected to a constant-pressure reservoir, the limit the flos rate. A discussion resulting from a P1 pressure disturbances are just the negative ,

study of this matter is given in [7] which also reflections (from the reservoir) of the P1 pressure raco= mends a flow rate equation that contains a disturbances created earlier at the valve. If the ripresentative metastability factor to account for pipe length is denoted by L, then the round trip thz increase flow rate due to the metastability travel time of a pressure disturbance moving at cffset. The equation is: wave speed c is 2L/c, thus:

21.

m - An (P1 - rPv) c/g A

pg = 4 (t 7 )

g (15) 4 1-(N")'

L

Although the waterhammer equations (13,14) apply to the entire pipeline, interest here is disks installed on the underseat end of the valve.

confined to the pipe connection at the valve The testing was performed with a size 6 valve for two where y = 0, for which the equations becomes reasons, these beings (1) The reaction (thrust) loads developed when even pi,p o , p1 g ,p1 g a size 6 valve blowdown is large. This sire valve was chosen for the test as being large V

fi

-V MLP1 (t) - p1(t)Jl fo = oc (17) en ugh to give representative results for all sizes while at the same time minimizing the problem of managing the reaction load.

where Pg = instantaneous pipe pressure at valve 6nM M hN W Vfg= instantaneous fluid velocity at valva was the same valve used to develop the inlet functional relationships between valve inlet pressure and (a) the average pressure under Sufficient equations have now been developed to the disk and, (b) the friction side bearing set up the solution procedure, but first the initial load used in the model. Since these functions conditionc must be defined. are exactly correct for this valve, use of At t = 0: x=x o (initial lift), V = 0 (disk this valve provides the best possible test of is stationary), pg = 0 (no reflected pressure waves the applicability of the choked-gas-flow-arrive at the valve until t = 2L/c). derived funt.tions to choked subcooled liquid In the computer program, a solution is obtained flow, i.e. Rg of equation (3) and R2 of equation (10).

by solving the above equations numerically using the f ourth-order Runge-Kut ta method. Program results include the lif t, velocity and acceleration of the Due to the complexity and cost of providicg high check element, the valve cover cavity, dashpot temperature water, which, admittedly, would chamber, and feedwater line pressures, the mass flow simulate actual feedwater conditions more clnnely, rate through the valve, and the total mass of fluid these tests were performed with cold water.

discharged from the valve. Since the test system corresponds closely to Preliminary verification of the computer program the underlying basis of the analytical model, was obtained in two steps. First, in the program, successful comparison of the results obtained from equation (1) for check element acceleration was the test to the predictions of the model is considered forced to have a known acceleration. The program to represent qualificatios of the analytical model.

computed the corresponding check element velocity The test provides measurement of the reservoir and displacement, which agreed with the exact pressure, to check for a constant value, and also solution. Second, input data to the program was provides reasurteent of all other pressures calculated selected such that the valve closure was, essentially, by the model, wir'. the exception of the average instantaneous. The resulting surge pressure was in pressure under the disk, for comparison to the good agreement with the exact result. calculations. Check element position during the transiert is also monitored. Due to the unavailability QUALIFICATION TESTS of a suitable flow measuring device, the flow velocity was not measured. However, total mass of fluid An analytical model is considered qualified discharged was measured for comparison to model when it can be shown, generally through comparison ""* ***

to test data, that the model provides reasonable Although not totally encompassing, the para-estimates of the performance of the system it was meters measured during the testing of the internally developed to represent. Subsequent application of dashpotted piston check valve provide sufficient data the model to represent actual physical systems to assure that underlying assumptions made in the requires a judgement decision made on a case-by- development of the model are reasonably correct.

cJse basis. The qualification of the model, and Additienally, the successful comparison of the test not its application, is of concern here. data to the model predictions demonstrates that the As discussed earlier, the model describiag the analytical model provides reasonable estimates of the dynamic performance of the check valve is bared on performance of the system it was developed to represent.

a constant-pressure reservoir, the valve, ard the Therefore, the model would be considered qualified for interconnecting piping. The line rupture is assumed determining the dynamic performance of internally to be instantaneous and to occur on the ur.derseat dashpotted check valves in simple piping systems with end of the valve. The model was written to accom- constant-pressure reservoirs.

mc3 ate subcooled liquid flow, over the range of pressures normally encountered in feeduater service. Test System In order to provide a test data base for use in qualifying the model, a test system was constructed Line rupture tests were perfo ed usin the test which corresponds closely to the basis of the facility shown in Fig. 3. The 40 f (1.1 m ) water h M ed u W " constant-pressure et stem consisted of a constant-pressure reservoi ,ad a e z a t t o he reservoir, a size 6 Y-pattern, internally dashpotted ,

check valve, and the required interconnecting piping. C'P'C Pressure was supplied to the system f rom a ecmpressed ing was 6 in. (150 mm) schedule 80, with a air storage field with sufficient capacity that the to M % d b W ' reservoir tank to the valve of reservoir pressure remained essentially constant 78 ft (24 m). A removable section of pipe near the uring the transient. Variatien of the field pressure reservoir tank also allowed test runs with a pipe vided variation of the initial test pressure. length of only 5 ft (1.5 m).

ultaneous line rupture was provided by rupture A size 6 ANSI Class 1500 Y-pattern piston type

=

5 f

i l

As th2 test systen wts being prasauriz2d for a A US -4 3 test, the cavity between the two rupture disks was l><

3 independently brought to an intermediate pressure by 4 ) an air bottle. This pressure develops the required 4 conreassi g si g,st pressure differentials across the inner disk and the ,

outer disk to prevent premature rupturing. Typically,

( the disk rupture pressure differential is about 75 percent of the test pressure. Disk pressure

.atta sessante differentials were maintained at about 50 percent of y%

n / et 8 p.in h the test pressure.

Calibration of the pressure transducers and

( associated instrumentation was initially established by the manufacturer, and checked tefore every test by following the manufacturer's recommended procedure "m sortta cf applying a known DC voltage to each transducer amplifier and checking the light-beam recorder

' matta tunP deflection.

When the test was ready, the strip chart g ,ggeogo watvt recorder motor was turned on and a switch thrown to 1 remotely open the solenoid valve connected to the l g f - .

rupture disk cavity. This allowed the rupture disk q 1

y ' 'l cavity pressure to bleed-cdf causing the inner disk j

"""""'*"""'"" to burst, followed by bursting of the outer disk.

During each test, pressures at the five points Fig. 3 Schematic Diagram of Test System indicated in Fig. 3 were measured, i.e., pressure reservoir, valve cover cavity, dashpot chamber, pipe chsek valve was used as the test valve. The valve near the valve joint, and valve discharge. Addi-vss modified for a dashpot as shown in Fig. 1, and tion' ally, the valve check element position was for a stem (not shown) connected to the piston. This recorded.

stem protruded through an 0-ring seal in the valve Following the test run, the test system was cover and served as a check element motion indicator isolated from the air storage field and de-pressurized, whzn connected to a linear potentiometer mounted to At this time, the water volume discharged from the the top of the valve. test valve during the test was measured by noting the Line rupture was simulated using a commercially difference in the water meter reading af ter the water availabic rupture disk assembly which consisted of level in the reservoir tank was raised back to the two thin metal rupture disks held separately by level switch.

flanges. Blowout area of the rupture diska was larger than the pipe internal area. RESULTS Transient pressure measurements at the points' indicated in Fig. 3 were by piezoelectric pressure Comparisona of measured to calcul.sted surge transducers having high frequency capability to assure pressure, valve closing times, and quantities of securate transient response. water discharged for all of' the tests are shown All pressure trcusducer signals and the potent. graphically in Figs. 4, 5 and 6, and are tabu-ioreter signal were recorded simultaneously by a high lated in Table 1. Data for both long and short frequency response, light-beam, strip chart recorder. Pi pe, lengths are included.

Water in the system was treated with a corrosien inhibitor to maintain cleanliness af ter initial flushing. ,,,, , o Tsst Procedure ,

E Prior to each test, the following sequence -

.a. . Cb cccurred:

The rupture disk assently was connected to the undarseat end of the test valve, and the valve opened. E Blesd points in the valve and piping were used to 3 ** -

remove any air in the system. Make-up water was I*

• pu=psd through a residential type water meter and O I

into the bottom of the reservoir tank until the water '

1sysl reached a level switch mounted just above the d'" O rsasrvoir tank. The valve isolating the water meter g f v3s closed and the water meter reading noted.

A full-ported plug valve (see Fig. 3) shown in m .

Ci Loao PIPE LE"57" tha line at the top of the reservoir tank isolates a s e t P:Ps LimoTa the air storage field from the test system until

( time for the test. The air storage field had been g previously pressurized to the desired test pressure. e, y d ,E A bypass (not shown) around the plug valve allowed tha test system to be brought up to test pressure naasueso sunst retssuet, es6 slowly. Once the test system was at test pressure. Fig. 4 Comparison of Measured and Calculated tha plug valve was opened. This allowed air flow Surge Pressure (bar = psi /14.5)

I free the storage field to maintain the water reservoir pressure nearly constant during the test run.

6

I - l . ;. ,: -.-. - g y m a w -say h p2t m w e, -l l ,,

, . I ... .. ,

i .

o a . j . . s. s .s, j .

= 0 .  ; .., . , ,,, ,,  ; .

.. .. . . u. .

Iu

• OO

. .. . ... .. i . . , .  ;

f ..

i o .

, , m.

i d" -

.. . ... i. i ,, 7

. i

.. 1 .u i . 1 . j . . , u. .

1 a "

. * :". TP

" o.. -

Table 1 Dashpot Check Valve Test Results s

u O Lo=s Pire Lansin reasons for the overprediction of surge pressure and a snoer part trasin fluid loss when long pipe lengths are involved, a study of the comparison of measured to predicted valve disk position with time reveals another reason, A typical comparison is shown in Fig. 7 (Run 7 in o

e o.: o.'a o.'6 c.s 1.0 Table 1). Total time to close the valve is cal-ntasunto watvt cLosins time. ste culated with very good accuracy but the shape of the curve is worth examining.

Fig. 5 Comparison of Measured and Calculated Apparent good agreement is shown between the Valve Closing Time measured and computed disk positions in Fig. 7 However, the vertical separation of these curves is overall agreenant between model and test is significant, especially during the latter portion considered good. Dashpot chamber pressure and valve of closure. During the entire time of closure, the cover cavity pressure comparisons also indicate good calculated disk lift is greater than the measured agreement but are not shown. value. This means that the calculated effective valve nozzle area (A n ), which varies with disk lif t, Long Pipe teneth is also larger and permits larger calculated reverse flow velocities. These larger calculated flow Surge pressure data in Fig. 4 and fluid loss velocities result in prediction of higher surge data in Fig. 6 show that the codel tends to over- pressure and increased fluid discharge from the valve.

estuate and therefore is conse*vative. Valve Therefore, it is suspected that improved agreement closure times shown in Fig. 5 appear to be cal- between the disk position curves will also result in culated with very good accuracy and with no obvious better predictability of the model.

bits. A sample comparison of measured to calculated Explanation for the conservatism of the model line pressure variation at the valve inlet (Pi in shown in Figs. 4 and 6 is offered. .Although neg- Fig. 3) is presented in Fig. 8. This graph, like lecting the effects of pipe friction is among the Fig. 7, is based on data from Test Run 7 which was initiated at a pressure of 1090 psia (75.2 bar).

The analytical model predicts that the line pressure increases in steps at the beginning of the transient,

, , as the pressure waves are reflected back and forth a between the valve and the reservoir. The steps in

• J the curve are prominent in the beginning because d O o there is little change in flow resistance of the 3

.aa .

valve initially in the transient, and the system behaves similar to a constant-pressure reservoir 3 discharging through a pipe with a fixed flow g 2D CX) resistance at the end of the pipe. As the transient 2 m -

O progresses, the curve becomes smoother because the a

continuously increasing flow resistance of the valve i

} is becoming dominant and causes continuous changes in l =, pressure at the valve.

Measured line pressure shown in Fig. 8 is h traced from a strai chart recording of raw test g data, similar to tnat shown in Fig. 9, and shows a . OLoas Part Lansin just the essential characteristics of the raw test osnoer part tensin curve. The line pressure curve in Fig. 9 snows that i

a clean, initially-steplike curve is not measured.

Reasons for the ragged appearance can be pipe

• * * * *

• friction, non-instantaneous line rupture simulation, and mechanical vibrations. These reasons are naasun to FLuin mass otstaanste, op discussed further below.

Fig. 6 Comparison of Measured and Calculated Pipe friction effects tend to round-off any Fluid Mass Discharged from Valve during sharp pressure changes and, of course, reduce Transient (kg = lb /2.2)

~

g TIME g

g. N RESERVOIR PRESSURE -4wte.

g -

NN DISK POSrTION VALW DISCHARGE -

,3 st: \s PRESSURE f Y g tac m :0 f \

\. DASHPOT PRESSURE m - E f

I

'#""'" \ VALVE COVER PRESSURE-p. fY M _g/\/ W UNE PRESSURE AT g e . , ,' N I, \ VALVE I, ref g] g .

s x

gi \

'e s: s.u ( .. (u in ...

( ( .e (u Fig. 9 Sample Strip Chart Recording of '

flME. SEC. Instantaneous Test Data Fig. 7 Comparison of Measured and Calculated between valve nozzle area and disk lift [6] . For  ;

Valve Disk Position for Test Run 7 the test valve, nozzle area decreases almost linearly (see Table 1)(sms = 25.4 x in.) alth disk lift until the disk is just above the valve body seat. At thir 79.nt nozzle area is practically pressure magnitude. zer ; here the noztia area / disk lif t relation changes The analytical model assumes a single, abruptly, because of valve geometry, and then instantaneous full line rupture. While it is true e ntinues with small slope to zero as the disk that the rupture disks used in these tests burst c ntacts the seat. Reverse flow velocity is nearly altost instantaneously (for the purposes of these terminated also at this point of disk lif t, and line tests), the time delay between the bursting of the pressure begins to drop due to the predominance of inner and outer disks was relatively long, typically waves causing decompression, which were reflected 30 ms. This delay almost coincides with the pressure frcm the water reservoir tank. These waves dominate wave travel time from the valve to reservoir to valve because strong compression waves are no longer being cnd will undoubtedly be the source of some discre- generated at the valve through reduction of reverse psccy. flow velocity.

It is suspected that me.ny of the smaller, high Line pressure curves shown in Figs. 8 and 9 frequency spikes appearing in the measured line ' exhibit the commonly observed pressure oscillations pressure curve are due to pressure fluctuations which occur after the valve is closed. Decay in the indured in the line fluid by the mechanical measured amplitude of the oscillations due to vibrations of the pressure boundaries in the test ricti n effects in the pipe is obvious. The system observed during the tests. The piezo- analytical model predicts no decay, since friction electric pressure transducers are compensated for ef ects are n t included.

aschanical vibrations imposed on the transducers Short Pipe Length and should not produce extraneous pressure signals.

Both the calculated and measured line pressure dzta indicate that the peak pressure is reached prior Calculations for the tests run with the 5 ft to complete closure of the valve as shown in Fig. 8. (1.5 m) pipe length u.derpredict the measured surge This is explained by the nature of the relation pressure (Fig. 4), and overpredict the measured valve closing time (Fig. 5) and the measured fluid mass discharge from the valve (Fig. 6). The overall 8 agreement between model and test is not as good as g

4 with the long pipe length, and is considered to be 1 due to effects not considered in the analytical  !

, model. It is not clear what these effects are, except i for pressure disturbances associated with the delay l between rupture disk bursts (for the short pipe El *""'"*' I "E " ' "* ** "" * "'#* "

f -g five pressure wave reflections between disk bursts)'.

?q I l .

gf Because the surge pressures associated with ,

j US cacau ,/ t

/\l t A short pipe lengths are ganerally small, the non-l

• s' t i ii conservatism of the surge pressure predictions is g- p / not of great concern; however, test data and model 3 / jlJ $y [1 g \. improvements are still being studied to identify the h }j ,(,,,, e g

source of the discrepancy.

l 4 -

F g L CONCLUSIONS I

I

't s:

en ..

en

,4 j

... ." Results of the tests verify the ability of the analytical model presented here to predict con-servatively, and in reasonably good agreement, the surge pressure and other associated quantities Fig. 8 Comparison of Measured and Calculated Line measured during the line rupture transients with the Pressure at Valve Inlet for Test Run 7 longer pipe length more typical of feedwater lines.

(see Table 1) dar = psia /14.5) 8

l The modal underpredicts eurge pressure for ACKNOWLEDGEMF"TS very short pipe lengths; however, these surge pressures tend to be small and may not be of much The authors wish to thank Mr. L. A. Gregory and concern. The cause for this behavior is being Mr. P. A. Nye for their assistance in the construction investigated. of the test facility, Mr. J. W. Rugh for assistance in No account was taken in the analytical model running the tests, and Mr. E. A. Bake for pertinent for the adjustment of valve parameters such as suggestions concerning the design of the test facility, nozzle area to a test-established value, or the use of ceasured fluid properties ss.ch as pressure wave REFERENCES speed from the line rupture tests to improve data c o:rpa risons. Indications are that such use would " Experimental Investigation of Pressure Transients improve the agreement. However, this information Created by a Closing Check Valve", Report No.

is not generally available when considering a valve KVB 73-473, 1973, ASEA-ATOM. Sweden.

for an installation. Use of catalog data and reference book data should provide sufficient model

2. Parmakian, J., Waterhammer Analysis, Dover Publications, 1963 e ts described in this paper verify the ability of slow-closing check valves to reduce surge pressures in feedwater systems undergoing 3. Pool, E. B. " Nuclear Containment of Postulated line rupture blowdown. The tests cited earlier by Fd e Whn",Nm A Ma %

ASEA-ATOM [l] were done on a test rig of similar Flow Line, 3rd Quarter, 1979, Rockwell Inter-national, Pittsburgh, Pa.

construction and produced a surge pressure of 3000 psi (205 bar) for an initial pressure of 1100 psia (75 bar) and a size 6 tilting-disk check valve.

4 Pool, E. B. " Nuclear Main Steam Isolation For the same initial pressure, tests on the check Valves,,, Report No. V-Rep 76-3, Flow Line, 3rd Quarter, 1976. Rockwell International, valve with dashpot produced a surge pressure of about 700 psi (48 bar). Further reduction in urge * "#8 * **

pressure can be obtained by increasing the dashpot

" '" "**"#** 5. Tao, L. N. and Donovan, W. F.,."Through-Flow in Because of the generally good agreement between Concentric and Eccentric Annuli of Fine model and test shown here, the analytical model is Clearances With and Without Relative Motion considered qualified for predicting the perf ormance f the Boundaries" Trans. ASME, November, 1955, of the type of check valve with dashpot used in the pp. 1291-1301.

tests and for a simple piping system with a constant-pressure reservoir. 6. Edwards, A. R. and Mather, D. J., "Some UK Studies Use of the analytical model for the type of Related to the Loss-of-Coolant Accident," Topical valve use in these tests in conjunction with water- Meeting n Water-Reactor Safety, Conf. 730304, hammer equations for more complex feedwater systems * * *

• b'IE

• EY' UE*"*

should provide reasonable accuracy provided the assucption of no vapor formation anywhere in the 7. Flow Calculation Handbook, Rockwell International, domain of the total model is valid. As with the Flow Control Division, Pittsburgh, Pa.1977 use of any analytical model, preliminary study and judgement are required to evaluate the validity of the ciodel application.

9 I!