AECL-6885

ATOMIC ENERGY g T O L'ENERGIE ATOMIQUEOF CANADA UMITED V ^ & J T DU CANADA LIMITEE

ENGINEERING RESEARCH IN NUCLEAR COMPONENTSRecherches d'ingenieurie sur les composants nucleaires

PART I: VIBRATION AND FRETTING WEAR OF HEAT EXCHANGERS

Partie I: Vibration et usura par frottement des echangeurs de chaleurs

P.L KO, M.J. PETTIGREW, G.A. WOLGEMUTH and A.O. CAM PAG N A

PART II. STEAM GENERATOR THERMAL-HYDRAULICS:

ANALYTICAL AND EXPERIMENTAL

Partie II: Hydraulique-thermique analytique et experimental

des generateurs de vapeur

W.W. INCH. D.A. SCOTT and M.B. CARVER

Chalk River Nuclear Laboratories Laboratoires nucleaires de Chalk River

Chalk River, Ontario

March 1980 mars

ATOMIC ENERGY OF CANADA LIMITED

ENGINEERING RESEARCH IN NUCLEAR COMPONENTS

PART I: VIBRATION AND FRETTING WEAR OF HEAT EXCHANGERS

by

V.I. K.0, M.J. VnttlQfizw, G.A. Wolgemuth, A.O. Campagna

PART II: STEAM GENERATOR THERMAL-HYDRAULICS: ANALYTICAL AND EXPERIMENTAL

by

W.W. Inch, V.A. Scott, M.8.

Prepared for the Fifth Symposium on Engineering Applications of Meahanias,Ottawa, Ontario, 1980 June 16-17. Fart I also prepared for the CanadianNuclear Association, Heat Exchanger Reliability Seminar, Toronto, Ontario,1980 May 1 and Part II for the Seventh Annual AECL-Utilities SimulationSymposium, Thermalhydraulics Section, Ecole Poly technique, Montreal,1980 May 12-13.

Printed with permission from the Conference organizers.

Chalk River Nuclear LaboratoriesChalk River, Ontario KOJ 1J0

1980 March AECL-6885

L'ENERGIE ATOMIQUE DU CANADA, LIMITEE

RECHERCHES D ' INGENIEURIE SUR LES COMPOSANTS NUCLEAIRES:

PARTIE I: VIBRATION FT USURE PAR FROTTEMENT DES ECHANGF.URS DE CHALEURS

peu

P.L. Ko, M.J. PeXtigKcw, G.A. Wotgcmuth, A.O. Campagna

RESUME

Dans cette conaunication on présente une techniciue pour l'analysedes vibrations engendrées par les écoulements et une procédurepour prédire l'usure due au frottement à partir de donnéesexpérimentales. Des examples illustrent ces techniques.

PARTIE II: HYDRAULIQUE-THERMIQUE ANALYTIQUE ET EXPERIMENTALE

DES GENERATEURS DE VAPEUR

pan.

W.W. Inch, P.A. Scott, M.5. CCLUVHK

RESUME

Dans c e t t e communication on d i s c u t e d 'un code numérique d ' o r d i n a t e u rqui modèle en détails l'hydraulique-thernique des générateurs devapeur, et on décrit des programmes expérimentaux conçus oourétudier en profondeur les écoulements tri-dimensionels biphasés.

Préparé pour le Cinquième Symposium sur les Applications Techniques de laMécanique, Ottawa, Ontario 1980 juin 16-1?, Partie I aussi préparé pour leSéminaire sur la Fiabilité des Ecltangeurs de Chaleur de l'Association NucléaireCanadienne, Toronto, Ontario, 1980 mai 1, et Partie II pour le SeptièmeSymposium Annuel de I 'EACL et des Services Electriques sur la simulation,section hydraulique-thermique, Ecole Polytechnique, Montréal, 1980 mai 12-13.

Laboratoires Nucléaires de Chalk RiverChalk River, Ontario KOJ 1J0

1980 mars AECL-6885

ATOMIC ENERGY OF CANADA LIMITED

ENGINEERING RESEARCH IN NUCLEAR COMPONENTS

PART I: VIBRATION AND FRETTING WEAR OF HEAT EXCHANGERS

by

V.L. Ko, M.J. Vzttigfiw, G.A. Wolgcmuth, A.O. Campagna

ABSTRACT

This paper presents a flow-induced vibration analysis techniqueand a procedure to predict fretting wear damage from experimentaldata. Examples are used to illustrate these techniques.

PART I I : STEAM GENERATOR THERMAL-HYDRAULICS:

ANALYTICAL AND EXPERIMENTAL

by

, W.W. Inch, P . A . Scott, M . 8 . Qo.KVQ.ti.

ABSTRACT

This paper d i s cus ses a code for d e t a i l e d numerical modell ing ofsteam generator thermal-hydraulics, and describes relatedexperimental programs designed to promote in-depth understandingof three-dimensional, two-phase flow.

Prepared for the Fifth Symposium on Engineering Applioations of Mechanics,Ottawa, Ontario, 1980 June 16-17. Part I also prepared for the CanadianNuclear Association, Heat Exchanger Reliability Seminar, Toronto, Ontario,1980 May l,and Part II for the Seventh Annual AECL-Utilities SimulationSymposium, Thermalhydraulics Section, Ecole Poly technique, Montreal,1980 May 12-13.

Chalk River Nuclear LaboratoriesChalk River, Ontario KOJ 1J0

1980 March AECL-6885

ENGINEERING RESEARCH IN NUCLEAR COMPONENTSPART 1: VIBRATION ANP FRETTING WEAR OF HEAT EXCHANGERS

P.L. Ko, M.J. Pettigrew, G.A. Wolgemuth, A.O. Canroagna

INTRODUCTION

The most important heat exchangers In CANDU*nuclear power s t a t i o n s are the steam genera tors . Atyp ica l steam generator contains over 4000 tubes .These tubes are e s s e n t i a l l y the boundary betweenthe heavy water on the primary s ide and the l i g h twater on the secondary s i d e . Thus i t i s very impor-t an t to maintain the i n t e g r i t y of the tubes over thee n t i r e l i f e of steam genera to r s . Other heat ex-changers are s i m i l a r l y important .

Heat exchanger tubes may be vulnerable toexcessive flow-^Hduced v i b r a t i o n . This in turncauses tube wear through Impacting and rubbing ontube supports and/or w^th adjacent tubes . Althoughfa i led tubes may be plugged with s p e c i a l l y develop-ed techniques , i t i s much preferable to avoid v ib r a -t ion and f r e t t i n g problems a l t o g e t h e r . This can beachieved by proper flow-induced v ib ra t i on ana lys i sof heat exchange components at the design stage andby knowing the e f f ec t s of various v i b r a t i o n a l , en-vironmental and u a t e r i a l parameters on f r e t t i n gwear. FIf>ure 1 ou t l i ne s t h i s approach.

FLOW-INDUCED VIBRATION MECHANISMS

There are three bas ic flow-induced v ib ra -t i on e x c i t a t i o n mechanisms, namely: (1) f l u i d e l a s t i ci n s t a b i l i t y , (2) random e x c i t a t i o n due to flow t u r -bulence, and (3) per iodic wake shedding. In c r o s s -flow , we have observed fluidelastic instabil i ty atidrandom turbulence excitation in both liquid and two-phase flow [ l ] . Periodic wake shedding resonance ispossible in liquid flow but has not been observed intwo-phase flow. In axial flow, response to randomflow turbulence is the dominant excitation rechanismin both liquid and two-phase flow, although axial-flow-induced random excitation is usually not aproblem in heat exchangers. Fluidelastic ins tabi l i -t ies are also possible in axial flow [2] . However,the c r i t ica l velocities for axial-flow instabi l i t iesare much higher than those normally encountered inheat exchangers.

j D E S 1 & " nOti. GEOMETR

| V I B P f l T TON E ' C I T A T I D H M t C H f l N

* " / I W S T f t B I L I T

PE S P O d S E T T : H A K E S H E D -D I N G , R AN D OM T U R B U L E N C E :

'

f

q

JLTISPAN

M P A C

RE T TING-

TEST

T T 0

[AS DATA

—4-

R C

j__S

E

ANALYTICAL•

• • V I B 1 C "

r (..„_L£ _SPA^_T[^TS;p___

* CANDU - CASadian Deuterium lJranium

Fig. 1 Vibration and Fretting of Heat Exchangerand Steam Generators

Fluidelagtic InstabilityIn a tube bundle subiected to a cross-flow,

the hydrodynamic forces on one tube are affectedby the motion of neighboring tubes. This createsan interaction between fluid forces and tubemotion. Fluidelastic ins tabi l i t ies are possible

- 2 -

when the relative motion between individual tubesis such that i t results in fluid force componentsthat are both proportional to tube displacementsand In phase with tube velocities. Instabili tyoccurs when during one vibration cycle the energyabsorbed from the fluid forces exceeds the energydissipated by damping.

I t is shown In reference 3 that, for a tubebundle subjected to the non-uniform flow velocityVr(x) « Vcl|i(x), the non-dimensional cr i t ica l velo-city V rc/fid at which instabil i ty occurs In the ithmode may be expressed by:

K.L/[2f±d2mpJ~ dx]} (1)

In the above expression fj and (J) (x) arerespectively the tube frequency and mode shape inthe i'h natural mode of vibration, d is tube dia-meter, I is tube length, m is tube mass per unitlength including the hydrodynamic mass, c is theviscous coefficient of damping, p is the fluiddensity and l|i(x) is the flow velocity distributionfunction. For the special case where the flow Isuniform over the whole bundle length equation (1)reduces to [4]:

- K(6m/pd (2)

-The instability factor K is determined fromexperimental data. Fluidelastic Instability occursat a critical flow velocity from which the value ofK may be deduced using equation (2). Many testshave been done on different tube bundle configura-tions [l]. The results are presented in terms of adimensionless reference gap velocity Vr/fd and adimensionless damping m6/pd^ in Figure 2. It showsthat for tube bundles in liquid flow, the instabi-lity factor K is generally higher than 5 regardlessof tube diameter, pitch or configuration. The samecriterion appears to apply In two-phase flow. Toallow for a realistic safety margin, a fluidelasticinstability factor of 3.3 is recommended in vibra-tion analyses of heat exchangers.

Vibration Response to Flow Turbulence ExcitationFlow turbulence In a tube bundle induces

random excitation forces. These random forces areformulated in terms of statistical parameters [3].

In liquid cross-flow the vibration responseand consequently the spectral density of the ran-dom forces is roughly related to flow velocitysquared.

Knowing the random excitation, the vibra-tion response may be estimated using expressionssuch as those formulated in reference 3.

where the logarithmic decrement S * c/(2mf ) .

200

100

50 —

20

10

d(mm)

1Lfd

T3"

cp ,p .

GiG>G>G.G iG I

G-,G i

K

V

CONNORSPETTIGREWPETTIGREWGORMANGORMANGORMANGORMANGORMANGORMANGORMANGORMAN

1 od7

aP/(p-d)

41.50.50.3357

.36

.54

.30

.23

.47

.50

Free Strenn Velocity

11.8-4030174040303030303030

0.008-01C0000000000

.156

.168

.063

.063

.071

.171

.071

.071

.071

.071

?4.012.712.719.019.013.013.013.013.013.013.0

c o

• L10UI0 FLOW• Tl

D O «1«

A t 1 1 INSTABILITY NOTfr D O REACHED

•G TEST WITH PARTIALFLOW BLOCKAGE

O SINGLE ROWA NORMAL TRIANGULAR> PARALLEL TRIANGULARa NORMAL SQUARE0 ROTATED SQUARE

mo

0 I 0 . 2 0. 5 10 20 SO 100 200

Fig. 2 Non-Dinenslonal Presentation ofExperimental Data on FluidelasticInstabilities in Cross-Flow

- 3 -

It is difficult to specify a maximum tol-erable tube respcmae as a design guideline. Wehave recommended before [3] 250 urn rtns as an out-side value with the suggestion that 100 um rmswould be preferable. Clearly, the maximum toler-able tube response depends on the resulting fret-ting wear damage. Hence, the latter must beevaluated taking into account materials, environ-ment, etc. as discussed later in this paper.

Periodic Wake She_ddingPeriodic wake shedding would generate

periodic forces in tube bundles. If the wake-shedding frequency coincides with the naturalfrequency of the tube, resonance mav occur. Thismay be a problem if the vibration response islarge enough to control the mechanism of periodicwake shedding. Then the periodic forces becomespatially correlated to the mode shape.

Periodic wake shedding excitation may beformulated in terms cf a Strouhal number S and adynamic lift coefficient. S is generally between0.34 and 0.67 [3]. In practice, resonance mav beavoided If the dimensionless numbers ftd/Vr foractual tube bundles are kept above unity. Thisallows adequate separation between tube and wake-shedding frequencies to take care of possible"locking-In" of the wake.

To avoid periodic wake shedding resonancealtogether may lead to overly conservativedesigns in many cases where localized high flowvelocities exist. Hence it is often necessarv toconsider resonance when fid/Vr is less than unity.The amplitude at resonance ts calculated from thelift coefficient.

As a design guideline we recommend keepingthe resonant amplitude below 2% of the tube dia-meter. Below this amplitude level it is unlikelythat the tube motion would be sufficient to con-trol and correlate wake shedding along the tube.

VIBRATION ANALYSIS AND MEASUREMENT

From a mechanical dynamics point of view,heat exchanger tubes are simply multi-span beams,clamped at the tubesheet and held at the baffle-supports with varying degree of constraint. Thelatter is dependent on the support geometry andparticularly on the tube to support clearance.In analysing and predicting tube vibration, weassume the Intermediate supports to be hinged.This keeps the analysis linear. The non-lineardynamics of tube to tube-support Interactions aredescribed later in the frettlng-wear section.

Analytical Technique

The dynamics of multi-span tubes and thevibration excitation mechanisms which were outlinedearlier form the basis for a computer code called'PIPEAU1 to predict the vibration response of heatexchanger tubes. The tube geometry, mechanicalproperties, and support locations and type (clamped,pinned, etc.) are input to the program. With thisdata, PIPEAU performs aiweigenvalue solution tofind the natural frequencies and corresponding modeshapes of the tube.

In order to calculate the vibration responseand the critical velocities for flui&elastic insta-bilities, the program also requires the cross-flowvelocity distribution and the overall damping ofthe tube. The damping Is a very complex parametercomposed of tube to fluid damping, structuraldamping, friction damping and squeeze film damping

at the baffle-supports. When actual measurementsare unavailable, we use a library of past cases andexperience to estimate a value.

Recently ye have reviewed the operating per-formance of ten heat exchangers from a fluidelasticinstability point of view. Both failures and satis-factory performances were considered. The resultsare presented In terms of dimensionless parametersin Figure 3. All the tube failures occurred forvalues of fluidelastic instability constant K.larger than h.U. Thus, our recommended guidelineof K = 3.3 appears very realistic In practice as itRives a reasonable safety margin.

] ] > * 1U3E l f t l l U » I S,-, M D f l HO TUBE F M U J f l E

H fi I I I INUJL*» T(J8[ fltJ

CBHl 3L . A _ . 1 ... i 1 J_ 1 _ i . i . I

Fig. 3 Fluidelastic Vibration Analyses of Field

Experiences with Heat Exchangers [5]

Field Measurement TechniquesIn situ tube vibration measurements are

still often necessary. The basic steps consist of:(1) inserting an acceleroraeter into the tube;(2) recording the generated signal on magnetic tape,and (3) analysing the taped signals In the labora-tory.

The accelerometer used (Figure 4) is a bi-axial piezoelectric type. The sensitive axes areorthogonal to each other and to the tube axis.

r Removable O-Ring Carr iers-

SignalCable

Accelerometer

Pig. Biaxial Accelerometer Probe for TubeVibration Measurements

The measurements are taken with the primaryhead removed from the heat exchanger. This allowsaccess to the tubesheet so the probe can be pushedinto a tube. Secondary-side integrity and floware maintained. The tubes vibrate as they wouldunder actual operation except for the smalldifference, which is corrected for iater, causedby the lack of fluid inside the tubes.

- 4 -

FLUIDELASTIC INSTABILITY ANALYSIS

No.

1

2

No.

1

8

8

(Hz)

17.7

42.7

69.2

VrCRIT

0.70

1.37

0.35

V, - I 82 m/s

v r - 2 8 1 m / s [ I N L E T )V, = 0 66 m/s

1 2 9 m 1 4 5 m 1 4 5 m ( 4 5 m 1 4 5 m 1 4 5 m 1 4 5 m 1 4 5 m

Fig. 5 Vibration Analysis of Process Heat Exchanger Tube

The orientation of the probe is not knownand since the motion ifl generally random, the twosignals are added vectorially. The resultant isthen frequency analysed and plots of amplitudeversus frequency in different zones of the ex-changer and different flow3 can be produced. De-pending on the reasons for the analysis, we canthen make recommandations concerning allowableflows or oroblem zones within the tube bundle.

The data are also compared with the analyti-cal predictions and any anomalies are Investigatedand resolved.

Case StudyOur methods are best i l lustrated by a

review of their application to a specific case.This one involves a triple-segmented, tube andshell , liquid-liquid process heat exchanger opera-ting in a heavy water plant [5] .

Shortly after commissioning, field measure-ments revealed very high vibration amplitudes atboth the inlet and outlet regions. At these levelstube failures would have been expected in a fewweeks of operation at 100% shell side flow. Somelong tube spans and relatively high inlet veloci-t i es , which impinged directly on the tube bundlethrough a perforated distribution plate,explainedthe problem.

The vibration analysis (Figure 5, Case 1)confirmed the above diagnosis. Note that theproblem was not due to the lower order modes ofvibration as i s often assumed but by the eighthmode (Vr/VrcRix " 1.37) where the high amplitudecoincides with the high velocity region. Stain-less steel straps were inserted between the tuberows in the two problem zones to stiffen the tubeassembly and bring the vibration below the insta-b i l i ty level. Field measurements after the modi-fication confirmed that the amplitudes were indeedlow. The unit operated successfully for aooroxl-mately 18 months at which time a leak of processfluid into the cooling water occurred.

Severe fretting damage was found on bothtubes and straps in the inlet zone (see photo inFigure 6) . Tube/strap fretting had perforated sometubes and in one instance a segment of strap hadbeen cut free. This allowed the tubes to vibratesufficiently to cause tube/tube fretting, acondition found only in cases of severe ins tabi l i ty .The lacings were not tightly fi t ted, thus providinga convenient wear surface but no adequate support.There were also unexpectedly high velocities due topartial inlet blockage.

Fig. 6 Frettlng-Wear of Heat Exchanger Tubes

Additional baffle supports were fitted inthe two problem zones. The vibration analysis(Case 2) shows there would be no further problems(Vr/VrCRIT * 1)• This heat exchanger has now runfor three years and has not encountered any problem.

FRETTING WEAR

Although large v ib r a t i on amplitudes can beavoided, flow-induced v ib r a t i ons cannot be e l i m i -nated a l t o g e t h e r . The quest ion I s : "How muchvibration i s tolerable without excessive f re t t ing-wear damage?". The prediction of fretting-wear isthe ultimate goal of our program.

Experimental TestsTo understand the tube-frett ing mechanisms,

which involve both vibration and wear, requiresstudies of a large number of parameters such asmaterials, environments, e t c . These are doneexperimentally in simple fretting-wear tes t r i p s .A tes t rig consists of a single-span cantileveredtube which i s vibrated against a tube-supportsnecimen. Excitation Is p.-ovided by a vibration

- 5 -

generator consisting of two stepping motors rotat-ing two eccentric masses in opposite directions.Some test rigs are enclosed in autoclaves foroperation at high temperatures under controlledenvironments. Others are open to atmosphere foroperation at room temperature. We have conducteda large number of frettinfi-wear tests of differenttube and tube-support materials in the room temper-ature r igs. The results have provided some insightinto fretting mechanisms as discussed in [6], Ourlong-duration tests have shown that depending onthe material combinations, while some wear ratemay diminish with time, others would increase asshown in Figure 7. In practice, this means thatsome fretting damage may occur slowly over a longperiod of time even at low vibration levels suchas those induced by random turbulence excitation.Our experience with operating components appearsto support the above. We have had a case wherethe f irs t tube failure occurred after 7 years ofoperation and the second failure after 16 years[5] . Hence fretting-wear damage predictions arerequired to assure the long-term re l iabi l i ty ofheat exchangers.

I t is often necessary to know the fretting-wear behaviour under rea l i s t i c operating conditionsto make valid fretting-damage predictions. We nowhave seven autoclaves in operation for such tes ts ,Figure 8. These autoclaves are connected to apressurized loop. The effects of temperature,oxygen content, pH, chemistry control additivesand -crud deposition on fretting-wear rates can bestudied. The tests are done for different tubeand tube-support material combinations, tube tosupport clearances, tube-support geometries, impactforce levels, e tc . Both short-and long-term testsare underway. During the long-term tes£s, fretting-wear damage is monitored periodically using eddycurrent inspection techniques to avoid disturbingthe specimens. The results of these* tests areused to develop families of curves similar tothose of Figure 7.

Fig. 8 High Temperature Autoclaves

We have found in our fretting-wear studies[6] that fretting damage correlates well with theimpact forces due to the dynamic interactionbetween vibrating tubes and tube-supports. Theseimpact forces appear to be the parameter linkingvibration response and fretting-wear.

Prediction of Impact ForcesAn analytical model in the form of a

computer code called VIBIC has been developed forthis purpose [7,8], This model considers non-l ineari t ies due to clearances between tubes andtube-supports. The model is essentially a time-domain numerical simulation of the dynamics of heatexchanger tube to support interactions. I t usesfinite-element techniques and is based on a modalsuperposition approach.

I N C O L O r a O O C A B B O N S T E E L

100 500 10 0 D 5000 1 0 . 0 0 0

TEST DURATION I , HOUR

Fig. 7 Change in Wear Race with Time, Incoloy 800/Carbon Steel

Fig. 9 Multi-span TubeTrough

- 6 -

20, 45

PREDICTED O D

MEASURED • •

PEAK E X C I T A T I O N FORCES

® 1 3 . I N AT 2 7 - 5 H ;

© 5 I N AT 1 7 . 5 H z

IN SIR

. 1 »2 '.-3 " f

TUBE SUPPORTS ( 0 . 3 8 m m D I A . CLEARANCE)

Pis. 10 Comparison of Measured and PredictedSupport Iranact Forces

The above analytical model was validated byimpact force measurements on mylti-span heatexchanger tubes subjected to simulated vibrationexcitation. The experimental apparatus consistedof a trough in which a single tube was held withreal is t ic tube-supports as shown in Figure 9.Impact forces were monitored by four miniaturepiezoelectric force transducers which supportedeach tube-support specimen at four locations 90°apart. The agreement between predictions andmeasurements is good, as shown, for example, inFigure 10. This figure presents two sets ofresults taken on a vertical steam generator tubeclamped at one end and held with four intermediatesupports. The tube was excited with sinusoidalforce at a point near the third tube-support.

Frettinn-Wear PredictionsThe prediction of fretting-wear damage is

best i l lustrated by an example. Consider thevibration response to random turbulence excitationof a hypothetical four-span tube located in theentrance region of a steam generator as shown inFigure 11. The magnitude of the random turbulenceexcitation was derived from data presented inreference 1.

V = 1.25 El's"1

73 73 85

Fig. 11 A Hypothetical Steam Generator Tube

The first four modes are considered in the analysis.The maximum random vibration response is 33 um rms.The problem now is to determine if such a level ofvibration would cause excessive fretting-weardamage over the design life of this component.Using our analytical model VIBIC, we estimate theimpact force level corresponding to the abovevibration response to be 1.2 K rms at the worsttube-support location. Assuming that the tubingand the support materials are Irtcoloy 600 andcarbon steel,respectively, the fretting-wearrelationship of Figure 7 may be used for thefretting damage prediction. Figure 7 shows afamily of three curves of depth of wear W<j vs timet for three different impact force levels in roomtemperature water. All three curves appear tofollow the relationship Wj = a-t;b. Although we donot have wear information at low impact forcelevels around 1.2 N, extrapolation of the resultsin Figure 7 indicates that below 5 N the wear rateshould be very small. We estimate that at 1.2 N rmsthe fretting-wear relationship would be 2.25 x 10~9t 2 - ! 7 um or less where t is in hours. Thus, thetube fretting-wear after 30 years under the flowconditions of Figure 11 is estimated to be 0.8 mm,assuming 7000 operating hours per year. Thisfretting damage is marginally acceptable. However,if the flow rate is 40% higher, the mid-span dis-placement would be doubled and the impact force atthe same support would Increase to about 3 N rms.The extrapolated fretting-wear relationship for 3 Nimpact force is 1.43 x 10~8 t 2 - 1 0 um resulting in a30-year tube fretting-wear ol about 2.15 mm, or athrough-wall (1.2 mm) life of about 22 years whichis obviously not acceptable.

CONCLUDING REMARKS

Most flow-induced vibration problems In heatexchangers can be avoided by thorough analyses.Fretting damage due to tube vibration response torandom turbulence excitation can be estimatedalthough additional fretting-wear information isrequired to cover a l l cases.

- 7 -

ACKNOWLEDGEMENTS

The authors wish to thank J. Tromp,J.A. Aikin and M.K. Weckwerth of the Chalk RiverNuclear Laboratories, D.J. Gorman of the Universi-ty of Ottawa and R.J. Rogers of the Universit/ ofNew Brunswick for their contribution to the workdiscussed in this paper.

REFERENCES

1. Pe t t ig rew, M.J. and Gorman, D . J . , "Vibrationof Heat Exchange Components in Liquid and Two-Phase Cross-Flow", Paper 2 . 3 , I n t . Conf.Vibrat ion in Nuclear P l a n t , Keswick, U.K.,1978, a l so Atomic Energy of Canada Limited,Report AECL-6184.

2. Pa idouss i s , M.P. and Pet t igrew, M.J.,"Dynamicsof F lex ib le Cylinders in AxisymmetricallyConfined Axial Flow", ASME, Journal of AppliedMechanics, Vol. 46, No. 1, March 1979,pp. 37-44.

3. Pe t t ig rew, M.J. , Sy lves t r e , Y., andCatnpagna, A.O., "Vibrat ion Analysis of HeatExchanger and Steam Generator Designs"Nuclear Engineering and Design, Vol. 48, (1578)pp. 97-115.

4. Connors, H . J . , " F l u i d e l a s t i c Vibrat ion ofTube Arrays Excited by Cross Flow", Proceed-ings of the Symposium on Flow-Induced Vibra-t ion in Heat Exchangers, ASME Winter AnnualMeeting, New York, Dec. 1, 1970, pp. 42-56.

Pet t igrew, M.J. and Campagna, A.O., "HeatExchanger Tube Vibra t ion : . Comparison betweenOperating Experiences and Vibration Analvses",Symp. on P r a c t i c a l Experiences with Flow-Induced Vibra t ions , Karlsruhe, Germany,Sept. 3-6, 1979, a l so Atomic Energy of CanadaLimited,Report AECL-6785.

Ko, P .L . , "Experimental Studies of TubeF r e t t i n g in Steam Generators and HeatExchangers", Trans. ASME, Journal of PressureVessels Technology, Vol. 101, pp. 125-133,May 1979.

Ko, P.L. and Rogers, R . J . , "Analyt ical andExperimental Studies of Tube/Support I n t e r -ac t ion in Multi-Sp<m Heat Exchanger Tubes",paper F9/4, Trans, of the 5th I n t e r n a t i o n a lConference on " S t r u c t u r a l Mechanics inReactor Technology", West Ber l in , Germany,Aug. 13-17, 1979.

Rogers, R .J . and Tick, R . J . , "FactorsAssociated with Support P la te Forces due toHeat Exchanger Tube Vibratory Contac t" ,Nuclear Engineering and Design, Vol. 44,No. 2, pp. 247-253, Nov. 1977.

- 8 -

ENGINEERING RESEARCH IN NUCLEAR COMPONENTS PAHT I I : STEAM GENERATORTHERMAL-HYDRAULICS, ANALYTICAL AND EXPERIMENTAL

W.W.R. Inch, D.A. S c o t t , M.B. Carver

INTRODUCTION

As rec i r cu la t ing—type steam g e n e r a t o r s a rekey components of CANDU power p l a n t s , they mustbe designed for high rel iabil i ty. Further, steamgenerator thermal-hydraulics should be optimizedto maximize heat transfer and hence plantperformance. We are actively involved inanalytical and experimental programs aimed atmeeting these objectives.

The analytical work is centered around thedevelopment of THIRST , a three-dimensional,homogeneous, incompressible two-phase flow code.THIRST predicts the thermal-hydraulics of steamgenerators, in terms of local velocity, quality,temperature, heat flux, density and pressure. Thecode has been used to assess the performance ofseveral steam generators.

An experimental program is also underwayto verify some aspects of the code. As steamgenerators may have a two-phase primary inlet,detailed void fraction measurements have also beenmade of air-water mixtures flowing in elbowssimulating the inlet feeders to the steamgenerator primary head. As a result of both thesestudies, more fundamental investigations areunderway to study the mechanism of two-pha3e flowusing both air-water and steam-water systems.

ANALYSIS OF STEAM GENERATOR THERMAL-HYDRAULICS

Steam Genera to r Design

The salient features of a natural recircu-lation steam generator with integral preheater areshown in Figure 1. Hot primary fluid flows insidethe bundle, transferring heat to the secondaryfluid en route. Subcooled feedwater from thecondenser enters the secondary side of the steam

1 CANDU - CANada IJeuterium Uranium2 THIRST - Thermal-Hydraulic Analysis J[n J

culating j>Team Generators

generator through the integral preheater oreconomizer. In the preheater, baffles forcethe flow across the tubes in a zig-zag patternto enhance heat transfer. At the preheaterexit, this flow, now raissd to saturation,mixes with flow returning from the hot side.The resulting mixture, undergoes partialevaporation as it rises through the remainingbundle section into the riser ami up into theseparator bank where the phases are separated.The steam continues up through the secondaryseparator bank, out of the steam generator, andto the turbines. The remaining saturatedliquid flows through the downcomer ennulus tothe bottom of the steam generator where itre-enters the heat transfer zone throughwindows cut into the shroud around thecircumference.

The THIRST Code

The THIRST computer code has beendeveloped to predict local flow and heat trans-fer in steam generators. THIRST provides thedesigner with a tool to analyse steam generatordesigns, pinpointing, for example, areas offlow stagnation in the secondary side, wherecorrosion would be enhanced and heat transferinadequate. Appropriate design changes canthen be made to correct the flow patterns.Designers can also examine local phenomenawhich determine overall efficiency, and thenmaximize the performance of the unit by alter-ing design features.

Typically, THIRST models a region ex-tending from the face of the tubesheet to theseparator deck and radially out to and includ-ing the downcomer annulua. Only one—half ofthe steam generator is modelled, since thedesign is symmetrical about a line through thecenter of the hot side and cold side. The codepredicts the flow, pressure and enthalpydistributions for both the shell md the tubeside. It calculates the overall pressure loss,

- 9 -

IISECOKKIY STEW SEMUTCMS I

I 1

I nima STEW SEMMTOKS I

I

I

STEPn OUOLITT CONTOURS

COLD SIDE HOT SIDE

Figure 1 Steam Generator Features Figure 2 Computed Quality Contours

overall heat transfer and local heat flux, and canalso calculate the circulation ratio for naturalrecirculation steam generator designs [l] .

THIRST is a steady state, three-dimensional homogeneous model. Presently the codecan handle roughly 5000 grid nodes with a corestorage requirement of 300,000 octal words.Frictional and heat transfer resistances are cal-culated using empirical correlations. The conser-vation equations are expressed in finitf differ-ence form and solved by using a recursivetechnique [2] . Convergence is achieved afterapproximately 50 iteratio.. steps, with each steptaking approximately 6 CPU seconds on a CDCCYBER175 computer.

Input data required for the analysis of asteam generator involve the geometric layout;primary fluid inlet enthalpy, flow and pressure;secondary feedwater enthalpy and flow, outletsteam pressure, and normal operating water level.Eight different steam generator designs have beenanalysed. These have included a number of diversefeatures such as an integral preheater, square andround U-bends, and a range of feedwater entrancegeometries.

Typical THIRST Results

Velocity and quality distributions result-ing from a typical THIRSV analysis of the designdiscussed above are shown in Figures 2 and 3«Quality values are marked directly on the con-tours in Figure 2. ' The preheater section is belowsaturation, except for a small zone at the exit.The conical shape of the contours on the hot sideillustrates the penetration of the downcomer flowinto the center of the tube bundle. The lowerhorizontal cut shows the quality pattern on the

tubesheet face. The second horizontal plane islocated just above the preheater exit. Herethe mixing of the higher quality fluid from thehot side with saturated liquid from the pre-heater results in steep gradients of quality onthe cold side. As the mixture moves up throughthe remaining bundle the quality continues torise. The hot side qualities are higherthroughout.

The influence of the U-bend geometrycan be seen by the shape of the quality pat-terns at the top of the vertical plane. Thisis more obvious from the velocity vectors inFigure 3- In the U-bend, the fluid tends tomigrate out towards the shroud where the resis-tance to flow is lower. Just below the U-bend,the velocities are primarily axial and parallelto the tube bundle. At the preheater exit thehigher flow on the hot side redistributes overthe cold side. In the preheater the zig-zagpattern of the flow around the bafflos issomewhat difficult to see because these veloci-ties are relatively small. At the tubesheetface the downcomer flow is shown as coming inon the hot and cold side and converging at thecenter of the bundle corresponding to the pointof highest quality.

Applications of THIRST

Design Assessment. To demonstrate theuse of THIRST as a design tool, the shroud win-dow size was altered to determine its influenceon performance. There is a significant effect,as all the downcoraer flow must pass throughthese windows to enter the heat transfer zone.Figure 4 shows the influence of window size onthe recireulation ratio, hPat transfer and themaximum quality on the tubesheet. Smaller

- 10 -

UELQCITY PLOTS

:QLD SIDE HOT s io

PT TUQESHEE1"

Figure 3 Computed Velocity Profiles

windows result in greater pressure loss and thuslower recirculation ratios. If the window heightis reduced, the downcomer flow is concentratedcloser to the tubesheet and although the totaldowncomer flow is less, the maximum quality on thetubesheet is reduced. As the main heat transfermode is in boiling, the overall heat transfervaries only slightly with the reduction in second-ary—side flows due to the lower recirculationratios. There is a slight decrease (0.5%) in heattransfer when the window height is increased, asthe downcomer flow is not forced to the bottom ofthe steam generator and thus the lower part of thebundle is partially by-passed. Decreasing thewindow height results in an escalating decrease inrecirculation ratio with only small decrease intubesheet quality. Increasing the window heightrapidly increases the tubesheet quality, however,the recirculation ratio becomes fixed by the pres-sure loss in other regions of the steam generator.

Effect of Operating Conditions. THIRSTcan be used to assess the influence of adverseoperating conditions such as a buildup of crud onthe tube surface, known as fouling. Fouling mar-gins are generally included in the sizing of shelland tube components. In THIRST, fouling istreated as an increase in resistance between theprimary and secondary sides, and is assumed to actuniformly throughout the tube bundle. Figure 5presents the influence of fouling levels on over-all parameters. The first set of data wasobtained for zero fouling on the tubes. Thiswould be the situation when a component is first

XRECIRCUUTION RATIO

£ OVERALL HEAT TRANSFER

O HIXIVLW QUALITY OF

TUBESHEET

J

100

V CHANCE IN ( INDOI HEIGHT

FIGURE 4 SHKUO •tHOOI HEIGHT

Figure 4 The Effect of Window Size

COX

(AT

z

ss5

15105

0- 5

- 1 0- 1 5- 2 0- 2 5- 3 0-35

4045

X RECIRCULATION RATIO

A OVERALL HE«T TRANSFER

O MAXIVUH QUALITY ON TUBESHEET

0 25 50 75 100**

* CHANGE IN FOULING ON PRIMARY TUBES

• i f a F l a u n t • • n i t s* * I 0 0 > - f l | l l l t IESISTIICE EMUS I I E I I U

HEAT T I . I S F f l HESISTIItE

Figure 5 Fouling Effects

put into service. For the last set of data, thefouling conductance was equal to the average totalconductance determined by the tube wall conduc-tivity and the primary—and secondary—side convec-tive coefficients, thus representing a resistanceincrease of 100? above zero fouling levels.

Fouling reduces the local heat flux for a givendriving temperature and degrades the overallheat transfer. The resulting lower averagesecondary qualities generate lower pressurelosses and higher recirculation ratios. Theincrease in recirculation ratio improves thepenetration and coupled with reduced heatfluxes gives the pronounced reduction inmaximum tubesheet quality. As primary fluid

- 11 -

moves along the length of the tube, its temper-ature and thus the driving temperature is sus-tained, since less heat is transferred. Atsome point, the driving temperature compensatesfor the increased resistance and the local heatflux becomes higher than the zero fouling case.This shift in local heat flux values explainsthe large reduction in tubesheet quality andrelatively small reduction in overall heattransfer.

Performance with Blocked PrimaryTubes. Over the life of the steam generator,individual tubes may develop leaks. Standardrepair procedure is to block these tubes at thetubesheet, effectively removing them fromservice.

X RECIRCULATION RATIOA OVERALL HEAT TRANSFER

O HAXtMUW DUALITY ON TUBESHEETUJ 4

§ 2

Z Q

- -4

i -6s•* -8

-

-

A\

-

-

\

0

I OF

«

\

110

THE

_ — — *

\

\20

TUBES BLOCKED

Figure 6 The Effect of Blockage

Blocking off primary tubes while maintain-ing the same overall primary flow requires theserviceable tubes to carry more flow. At thetubesheet the heat flux from the serviceable tubeschanges little with the percentage of blockedtubes because the driving temperature is the sameand the total heat transfer coefficient changeslittle with primary velocity. THIRST results,summarized in Figure 6, show that the maximumtubesheet quality varies only slightly with thepercentage of tubes blocked. Individually, theseserviceable tubes carry more energy and thus theprimary temperature and hence the heat flux issustained along their length. Secondary qualitiesaround the serviceable tubes develop at a lowerlevel in the steam generator. The effective re-duction in heat transfer area results in an over-all reduction in heat transfer. The lower averagequality yields a reduction in pressure loss andthe observed increase in recirculation ratio.

Status of the THIRST Code

The development of the THIRST code hasbeen completed, and a user—oriented version with asuitable manual will be ready for transfer topotential users in mid 19SO. Further relateddevelopment will include extending the applicationof the technique to heat exchangers. When thisdevelopment is complete, a package of detailedanalysis tools will be available to assist thedesigner in improving reliability and performanceof shell and tube components.

EXPERIMENTAL INVESTIGATIONS IN STEAM GENERATORTHERMAL-HYDRAULICS

In conjunction with the analytical work,experimental programs have been initiated at CRNLto investigate aspects of both primary andsecondary flow. We concentrate here on two partsof this program, THIRST verification experiments,and those investigating the separation oftwo-phase flow in elbows.

THIRST Verification Experiments

A first step towards experimentally veri-fying the numerical predictions in the THIRST codehas been completed. The test section, thoughsimDle is designed to incorporate many of the flowfeatures present in steam generators.

These include a gate to represent the windowsthrough which feedwater and recirculated waterenters the tube bundle, and also regions in thebundle where cross flow, axial flow or combina-tions of both exist.

Hot film anemometry was found to be themost effective technique for measuring localvelocities, and most of our efftfrt was spent ondeveloping the X-type probe, particularly inreducing the frequency of probe breakage.Details of the measuring technique are given inrefererce 3; hence we merely illustrateresults of a typical case.

The geometry of the test section, andvelocities for a particular flow and gate open-ing are shown in Figure 7. These involvedapproximately 5000 velocity measurements. Theresults demonstrate that accurate velocitymeasurements in tube bundles can be made withhot film anemometry as the overall mass flowrate can be computed fairly closely from thevelocity readings. For an inlet mass flow rateof 181.7 kg/s, flows calculated from measure-ments at eight different cross sections laybetween 167 and 220 kg/s.

Measured velocities were compared withresults calculated by a two-dimensional modelin a simplified version of the THIRST code.The average error in velocity magnitude wasapproximately 13%, a figure comparable to themeasured mass flow discrepancies. This isreasonably good agreement in view of the factthat no attempt was made to include three-dimensional effects in these calculations.Similar tests using air and water have now beeninitiated.

Two-Pha3e Flow in Elbows

A further experimental program com-pleted at CRNL studies two-phase flow on thepramarj aide of steam generators* This programcentered around a one-third scale model of theinlet head for 600 MW type reactors. Steam-waterflows were simulated by air/water mixtures atequivalent void fractions. Possible phaseseparation in the inlet elbow and at the tubesheetwas investigated.

- 12 -

T !

I 1 .. L .. J L...L _1. .J !.._.! .._._L i_.._i... 1 I L

Figure 7 Test Section and Velocities

Steam generators for the 600 Mtf GentillyII Nuclear Station are used as reference for ourmodel, shown in Figures 8(a) and (b). I" thismodel the two inlet lines are replicated by 150 mmdiameter pipes fed by two horaogenizer-mixers con-nected to the air and water supplies. The im-pedance of the tubes, through which the flow mustgo after entering the head is simulated by a per-forated plate containing 200 holes installed atthe location of the tubesheet. All measurement inthe work used the isokinetic sampling technique,two-phase flow must be sampled in such a way thatthe phases and their relative velocities are notdisturbed. Further details of the experimentaltechnique are given in reference 4.

The prevailing conditions in steam genera-tors for 600 M¥ reactors correspond to s flow of89 kg/s with a void fraction of 0.36. Ve, there-fore, studied the effect of void fractions from0.27 to 0.57 at a constant water flow of 89 kg/s,water flows from 77 to 100 kg/a at an approxi-mately constant void fraction of 0.36.

For each flow condition tested, velocityand void fraction readings were taken at sevenlocations across the diameter of the pipe, up-stream and downstream of the elbow. The upstreamreadings were to confirm that the flow enteringthe elbow was homogeneous. Results of readingstaken downstream are shown in Figures 9(a) and(b). Position 0 corresponds to the outer radiusof the elbow. As expected, phase separation ismore complete at high void fractions. Near theinside of the elbow the flow was 100? air forhomogeneous entry void fractions greater than

1 •1 1

s •4 1

1 1t 1

I 11 •| 1

, ,

1 I1 1

ICIIIII I* Illfl1' UHIEll •[•EIIIII Mllllci 1* IIIIIL HUE

ca 111 11 UK

tuic mint trimKllllilll

I E I

1 Fill I1II1U1IIII

• CIHML m «( a III till

ca flMI t.MIFICE miE

DISCHARGE TO RIVER

TRAVERSING PLANE Ct

(b)

(a)

Figure 8 Primary-Side Inlet Experimental Model

0.45, and close to 100£ for lower voidfractions. On the outside of the elbow thelocal void fraction was correspondingly lowerthan the homogeneous void fraction. The localflow velocity was fairly uniform except forlarge void fractions, where the air velocitynear the inside of the elbow was higher thanthe average velocity. Generally, the boundaryof separation between air and water was clearlydefined.

- 13 -

V Q I O FRACTION 0 . 3S

Figure 9 Void Fraction Profilesa) Void Effects b) Flow Effects

Partly because the separation found inthese tests was so extreme and partly becauseof the difficulty in relating these results tosteam-water flows, further investigations areunderway. These involve additional air-watertests, steam-water tests, and analytical model-ling of the phenomenon.

REFERENCES

1. W.W.R. Inch and R.H. Shill, THERMAL-HYDRAULICS OF NUCLEAR STEAM GENERATORS:ANALYSIS AND PARAMETER STUDY. Paper sub-mitted to ASME Nuclear Engineering DivisionConference, San Francisco, August 1980.

2. S.V. Patankar and D.B. Spalding, ACALCULATION PROCEDURE FOR HEAT, MASS ANDMOMENTUM TRANSFER IN THREE-DIMENSIONALPARABOLIC FLOWS. Int. J. Heat Transfer15, 1972, P- 1787.

3. D.A. Scott and J.D. Shaw, MEASUREMENT OFTHE FLOW FIELD IN A TUBE BUNDLE BY HOT WIREANEMOMETRY. Proceedings of the 7th AnnualCanadian Congress on Applied Mechanics,Sherbrooke, 1979, P- 725-6.

4. D.A. Scott, TWO-PHASE FLOW SEPARATION INELBOWS, Transactions Canadian NuclearAssociation, Student Conference, Ottawa,1979, p. 58-61.

ACKNOWLEDGEMENTS

This work was partially funded underthe Ontario Hydro, Atomic Energy of CanadaCommon Development Program (CAITOEV).

ISSN 0067 - 0367

To identity individual documents in the series

we have assigned an AECL- numbar to each.

Please refer to the AECL- number when re-

questing additional copies of this document

from

Scientific Document Distribution Office

Atomic Energy of Canada Limited

Chalk River, Ontario, Canada

KOJ 1J0

ISSN 0067 - 0367

Pour identifier les rapports individuals faisant

partie de cette se>ie nous avons assign^

un nume>o AECL- a chacun.

Veuillez faire mention du numeio AECL- si

vous demandez d'autres exemplaires de ce

rapport

au

Service de Distribution des Documents Officiels

L'Energie Atomique du Canada Limite'e

Chalk River, Ontario, Canada

KOJ 1J0

Price $2.00 per copy Prix $2.00 par exemplaire

698-80