Steam Bubble Collapse, Water Hammer and Piping Network Response Volume I.

Steam Bubble Collapse and Water Hammer in Piping Systems: Experiments and Analysis
by

R. Gruel, W. Hurwitz, P. Huber and P. Griffith Energy Laboratory Report No. MIT-EL 80-017 June 1980

Steam Bubble Collapse, Water Hammer and Piping Network Response Volume I. Steam Bubble Collapse and Water Hammer in Piping Systems: Experiments and Analysis by R. Gruel, W. Hurwitz, P. Huber and P. Griffith

Department of Mechanical Engineering and Energy Laboratory Massachusetts Institute of Technology Cambridge, Massachusetts 02139

sponsored by Boston Edison Company Consumers Power Company Northeast Utilities Service Company MIT Energy Laboratory Report No. MIT-EL 80-017

June 1980

-3-

Foreword

Work on steam bubble collapse, water hammer and piping network response was carried out in two closely related but distinct sections. Volume I of

wlhs report details the experiments and analyses carried out in conjunction
with tie steam bubble collapse and water hammer project. Volume II details

the work which was performed in the analysis of piping network response to steam generated water hammer.

-4-

Table of Contents

Volume

Steam Bubble Collapse and Water Hammer in Piping Systems: Experiments and Analysis

Foreword Table of Contents I. II. III. IV. V. VI. VII. Introduction . . . . . . .

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

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

3. 3
4 5
7 12

Experimental Apparatu Instrumentation . .

. . . . . . .

Operating Procedure Results Modeling Conclusions . . . . . . . . . . . . . .

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

14
16

28 42 43 44

REFERENCES Appendices

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

-5-

I.

Introduction

Water hammer incidents in conventional and nuclear steam systems are an important problem of broad general interest in piping network design and transient operation. Water hammer in PWR steam generator

sparger feed lines has, for example, been a recurrent problem when the sparger becomes uncovered during certain operational transients (Creare 1977). The central goal of this research has been to develop

experimental data and supporting analyses that will contribute to the evolving understanding of water hammer created by steam bubble entrapment in a pipe containing subcooled liquid. The first objective of this study has been to obtain a body of experimental data on water hammer initiated by steam bubble collapse. These experiments include measurement of pressure transients and high speed films of the process of bubble collapse and impact, and, in conjunction with Hurwitz (1980), records of the resultant pressure wave propagation through a variety of simple piping configurations and measurements of the induced structural response. The data that have been

obtained should be useful in benchmarking existing analytic models and numerical codes. The second objective of this study has been to formulate and test simple models for the steam bubble collapse process. The starting

point in the analysis of water hammer is obtaining a source "forcing function" which ultimately produces loads on remote as well as on nearby elements of the piping system. Back pressure (the system absolute

pressure outside the condensing steam bubble) and flow resistance in the line in which the collapse occurs have been varied experimen-

-6-

tally and analyzed via a hydrodynamic model for steam bubble collapse. Simple scaling laws correlating the effects of varying back pressure and flow resistance to impact pressures and bubble collapse times have been derived from this model and compared with the experimental data. Finally, water slug and steam bubble dynamics are combined to predict the bubble collapse dynamics as a function of heat transfer rates from the steam to the subcooled liquid. compared with the experimental results. These predictions are

Two physical limits to the

heat transfer rates can be postulated, predicted and compared with the heat transfer rates required to match theory and experiment. The

predicted limits consistantly bound the valves inferred from experiment.

-7-

II.

Experimental Apparatus

Figure 1 shows a schematic of the water hammer generator. It is constructed of standard 1-1/2" steel pipe and fittings, with a transparent lexan section (1-1/2" I.D. - 1-3/4" O.D.) in which the final phase of bubble collapse and column impact occurs. In the

initial state, a short liquid column fills the lower part of the system with its free surface adjusted to lie within the lexan section. Above this is a long steam bubble, initially in thermal equilibrium with the lower liquid-free surface. A fast-acting, pneumatically

operated piston valve separates the top of the steam cavity from the cold water reservoir. The actuation of this valve brings the steam Driven by the steam

into abrupt contact with the subcooled liquid.

condensation, water flows rapidly from the reservoir into the central structure and collides with the lower liquid column. The central structure consists of a series of nipples, crosses and unions. Two of these unions are specifically designed to provide

a rigid, watertight interface between the lexan and steel piping. Between the lexan section and the cold water reservoir lie two crosses. Three of the resulting four ports are utilized; one for

steam inlet and one for steam outlet connections; the third is a thermocouple port. A manual ball valve and an electrical solenoid

valve are located in series on both the steam inlet and outlet lines +o insure isolation during the transient. Another cross lies below the lexan section, at the end of the central structure. Two of its ports are for pressure transducers;

-8-

Pressure Ports for

Reservoir

Pi

Port

Pn

.23m
lure

T2

her

Steam Inlet
S

.65m
TI Ther
F

:am Outlet

011

.27m
-i

Ste

II

.. _, .02 n

.20m
. 6,

Press

- Outlet to
Test Section

Pi

Figure 1. Schematic of Water Hammer Generator

-9-

the other is attached to the piping network whose response to the water hammer is being tested (Figure 2). The central structure is rigidly supported by a framework constructed from 1-1/4" pipe and special interconnecting fittings. The

framework is about 1' in depth and width and stands about 6' tall. Its four legs are securely bolted to the floor. In addition, four

braces (one on each leg) run from the mid-section of the framework to the floor where they are also rigidly bolted down. Pipehangars and

threaded rods are utilized to keep the central structure from vibrating in any horizontal direction. The cold water reservoir and the

auxilliary pressure stabilization tank are also supported by this structure. The cold water reservoir, consisting of a standard 6", 125 psi drainage tee, is mounted above the central structure. It is aligned

vertically so that the side port can be used to view the water level inside the tee. Initially, water fills about half the tee; pressurAn auxilliary tank is used to insure

ized nitrogen fills the rest.

that the initial and final pressures in the system vary by less than
5%.

A pneumatic valve is mounted within the reservoir. double-acting valve:

It is a

pressure can be applied to push down the piston

sealing it against the bottom flange (closed), or to quickly lift it off the bottom flange bringing the steam into contact with the subcooled liquid (open). Nitrogen is used to operate the valve. Two

pressures are utilized; 60 psig to close the valve, 400 psig to open it. Originally, only 120 psig was used to open the valve, but when

250 psig was used there was a marked difference in the resulting

-10-

I
b, 60 IL ' B

cX

= 0

to

=a (4-

L r 0 ) u 3P3

0-0)

a.

CJ

Lu er,,

00A

o, a

11--

pressure histories.

No further change was found when 400 psig was

used, but this higher pressure was chosen to minimize any possible effect a "slow" valve might have on the pressure histories. Attached to the pneumatic valve is an aluminum rod and disk. To minimize drag and vortex formation when the disk is lifted, a cylinder of closed-cell foam, having the same diameter of the disk, encases the rod. The top of this foam cylinder lies above the water

level in the reservoir, thereby reducing the drag when the valve is opened. The bottom of the aluminum disk is covered with soft rubber which provides a good seal when compressed against the bottom flange of the reservoir. This flange provides the interface between the 6" An orifice plate

tee (reservoir) and 1-1/2" pipe (central structure).

can be mounted on the upper surface of the bottom flange - the piston would then seal against this plate. To provide a good seal between

the orifice plate and the bottom flange, the flange was machined to hold an O-ring between it and the plate. Thus, when the piston is

closed, both the soft rubber and O-ring are compressed, insuring that the cold water in the reservoir is isolated from the saturated steam in the central structure.

-12-

III.

Instrumentation

Instrumentation on the generator can be separated into two categories - initial state measurement and transient response recording. Initial state measurements were made with pressure gauges and thermocouples. The transient response was recorded using pressure trans-

ducers and high-speed films. Two static pressure gauges having a range of 0-60 psig (100-500 kPa) were used on the generator. One was used to measure The other mea-

the initial pressure in the steam bubble (Figure 1).

sured the back pressure in the cold water reservoir (P0 ; this gauge
was located on the auxilliary tank and is not shown in Figure 1).

Initial temperatures were also measured in the cold water reservoir (T2 , Figure 1) near the bottom flange and in the steam bubble (T l , Figure 1) opposite the steam inlet. Iron-constantan thermoThe

couples were used in conjunction with an artificial ice point.

signal was fed into a digital readout, displaying the voltage output of the thermocouple. recorded. Pressure transients were recorded using piezoelectric pressure transducers with response times (3ps) considerably shorter than any of the transients of interest. Two ranges of transducer sensitivity No transient temperature histories were

were used to capture details of both the steam bubble collapse transient (involving pressure changes of order 102 kPa (15 psi) over a

period of order 102 ms) and the column impact (involving pressure changes of order 104 kPa (1500 psi) over a period of order 1 ms). Four pressure transducers were used; one to measure the collapse

-13-

transient (P2 , Figure 1) and the other three to measure the column , Figure 1) and the wave propagation through the piping impact (P1 network (P1A' P3 ; Figure 2). These transducers were coupled (one or

two at a time) to a two-channel storage oscilloscope. High-speed films (1000-2000 frames per second) of the final phase of bubble collapse, column impact, and subsequent cavitation and column separation were obtained and correlated with measured pressure transients.

-1 4-

IV.

Operating Procedure

Initial conditions in the test system were accurately set and recorded before a water hammer was generated. procedure was used. The desired reservoir conditions were set first. Both the piston The following operating

valve and the drain valve were opened and water was injected into the reservoir. This flushed the air out of the piping network and brought When the water in the reservoir was

cold water into the reservoir. sufficiently cold (20-25C), closed.

the drain valve and piston valve were

The water supply was shut off when the water level in the reNitrogen was then injected into

servoir reached the desired height.

the pressure equilization tank until the desired pressure was reached (300-500 kPa). A steam pocket was formed by opening both the steam inlet and steam outlet valves. The incondensible gas and steam in the supply

line along with the water above the steam outlet port in the central structure were carried through the outlet line to the steam dump tank. This was continued for several minutes until only steam flowed through the line. The steam supply was from the M.I.T. steam lines at a preThe air fraction in this

sure of 1500 kPa and temperature of 200C. steam is typically about 10l
4

The steam outlet valve was closed and

the drain was opened slightly to lower the water level so that the water/steam interface was within the lexan section, The steam presThe

sure was set by adjusting a regulator on the steam supply line.

water/steam interface was allowed to reach a local thermal equilibrium.

-15-

The temperature and pressure in the steam were then checked to ascertain that it had reached saturation conditions, and the reservoir temperature and pressure were recorded. T2 ranged from 20-250 C and Po The

was set at various 50 kPa increments between 300 kPa and 500 kPa.

manual ball valve on the steam inlet line was closed, and immediately afterward the solenoid valve on the same line was closed simultaneous to the opening of the piston valve. This allowed the steam bubble

collapse and subsequent column impact to occur in a controlled, repeatable manner.

-i6-

V.

Results

The oscillogram

in Figure 3 shows a typical

pressure transient Pressure trans-

developed during bubble collapse and column impact. ducer P 2 shows the details

of the bubble collapse pressure drop

(upper trace) in the time interval -220 ms > t > 0.

P1 records the impact pressures (lower trace), the first of which begins at t - 0. Note the different scales for the two traces.

Figure 4 shows a sequence of line drawings traced from a highspeed film record of bubble collapse and column impact under conditions similar to those depicted by the oscillogram. In view of the

generally excellent test-to-test repeatability that has been achieved, the film records should correlate reasonably well with the details of the pressure histories, although they were obtained in separate runs. All times are referenced to the impact time t O; 0 valve opening can

also be inferred in the high-speed films from the almost immediate onset of boiling at the surface of the lower liquid column caused by the rapid depressurization of the steam cavity. Valve opening is also

characterized by the initial rapid pressure drop in the steam cavity (P2' t < -220 ms). The high degree of subcooling in the reservoir

and the large interface area and turbulent mixing produced when the valve is opened accounts for this rapid initial depressurization. The subsequent pressure drop rate decreases rapidly (t < -220 ms) due to two separate effects. When the subcooled liquid/steam inter-

face enters the central structure, a thermal boundary layer develops at the interface, slowing the rate of condensation. Boiling at the

lower interface also reduces the effective overall rate of mass

-17-

,,
4i

Eg

0 N
O

.) c
Qm . S.. S.

0 0 0
I

_I~

=@Q

IS-

(U
4')

Ia'
cn W II
L

O'
!

-ij -a.

-18-

---

STEAM

ORIGINALWATERLEVEL------

4-

.1 .. : i.

-LIOUIDO

INITIAL WATER/STEAM INTERFACE I O i 5

THE FIRST BUBBLE APPEARS

t -210 ms
cm

,4p-~
ATER SLUG - REGION OF EXTENSIVE STEAM WATER MIXING - I N TERFACE

VIGOR(XJS BOILING BEGINS

-9ml

t. -137 ms

Figure 4.

Sequence of Events through the Formation of the First Cavitation Bubbles traced from a High Speed film taken at 1000 frames per second.

-19-

''
-RESIDUAL SMALL
STEAM BUBBLES

''
..

IMPACT

+4

ms

t Oms

o 0

0 I
4e

, 'I 0o To ToTo 0

I
o e

CAVATATION-

BUBBLES . a

e o
To

'; .'

,
CAVATAT.~ CN EEGINS
t tS ms

+12 ms

Figure 4.

continued

removal from the steam cavity.

The high-speed firm

('Figure4) show

that a significant amount of water is boiled off the lower surface at this time - the interface drops 2-3 cm during bubble collapse. The total collapse time inferred from either the pressure history P 2 or the film records was approximately 220 ms for this cases
The impact time (t 0 in Figures 3 and 4) is self-evident in the

pressure histories and can be clearly identified in the films. Impact pressures are of typical acoustic levels Figure 3).

ighspeed

approxi-

mately 4000 kPa in this case (PP 1

Cavitation begins shortly after impact (t Column separation (: 30 ms long in this

8 ms, Figure 4). close

case) occurs very

to the point of initial impact as can be seen from the line drawings (Figure 4) and ends in column collapse producing a second impact pressure typically of lower amplitude than the initial impact pressure. This cavitation/collapse cycle is recurrent; the time between

pressure spikes generally remains constant while the impact pressures decay in amplitude, This process can be seen by examining Figure 3;

pressure spikes can be seen at 130 and 260 ms in the bottom trace the related cavitations can be seen in the upper trace. A series of experiments was conducted to investigate the effects of changes in the reservoir pressure or increases in the flow resistance between the reservoir and the central structure on the transient responses. Figures 5 and 6 illustrate some of the results. For the

tests shown in Figure 5, the reservoir pressure was varied from 300 kPa to 500 kPa while the flow resistance was held constant. For the tests

shown in Figure 6, the reservoir pressure was constant at 500 kPa while flow resistance was varied by inserting orifice plates in the reservoir,

-21-

-120

-80

t(ms) -40

C
Il I

300

a . Poo. 0
P

omxal

06

5UVV

6000-

2 200 (kPo)

(kPo)

PI 4000-

2000100

300_ _ _

300

6000

2000 100
,&eV

300 ow 0 8 000

.0.8

6000
200 P2
(kPo)

P 4000
(kPo)

2000
100
I

,, rWv

300

P
POMnox

8000
6000

200

P2 (kPa)

Pl 4000 (kPo)

2000

I00
I
IAA

300 00
I

-I

.. I

r V P
P2 (kPo)

-a

Iv

800(

6000
200 Pi 4000 (kPo)

2000
100 300 300

-20 0 20 40 60 60 100
t(ms)

Figure 5.

Water Hammer Pressure Signatures (P1, P2 ) for T = 1330 C, T 2 20C, and various back pressures (Po) where Pmax - 500 kPa.

-22-

A, -300 . t(mn) -200 -100 . . . , .

0.2 IU

..A

4000-

300 200
00 P2 0 (kPo)

pi 2000(kPo) 300

300-

TP Op
6000
... I

Ap

OR

4000

300 200 p 2 I00 (kPa) 0

Pi 2000 (kPa) 300

At -0.4441
Ap

Pi (kPo

300

I- 200 0
o00 P2 (kPo)

VVVt

. 0.694
Wp

4000

_L

300
200 P 2 100 (kPo) 0

P 2000 IkPo) 300

6000
At,

Ap

000 4000 Pt 2000 (kPO)

300 200 P2 100 (kPo) 0

300
0

200 100 (ma)

300

Figure

6.

Water Hammer Pressure Signatures (P1 , PO) for T1 - 133 0C, T 2 20C, P = 300 kPa, and 2 various flow resistances (At where Ap 11.4 cm .

~p

-23-

directly above the bottom flange.

Higher back pressures and lower flow

resistances clearly result in shorter collapse times and larger impact pressures. The trends exhibited by these results can be related to a simple hydrodynamic model for the steam bubble collapse and column impact. The mass flowrate through an orifice can be expressed as

KAtVpaP

(1)

where K is the discharge coefficient, At is the area of the orifice, pt is the density of the liquid, and AP is the pressure drop (P - P2 ) K is

across the orifice plate (Baumeister and Marks Seventh Edition).

assumed constant for the various orifice plates as a first approximation and the exact value of AP will depend on the heat transfer dynamics in the central structure but will scale with P. The collapse time t is inversely proportional to the volume flow-

rate; Equation (1) therefore suggests

cc

At

(2)

The impact pressure PI is proportional to the column velocity at impact and should therefore scale roughly as the inverse of the collapse
time:

PI

At

(3) (

-24-

The maximum value of At/Po in these experiments corresponded to the test in which no orifice was used (At = Ap) and the reservoir pressure was at its maximum value (Pomax = 500 kPa). Normalized variables can be for the various cases.

derived by taking the ratio of At Pr to Ap/Pmax

These variables correspond to the ratios of peak impact pressures (PI/PImax) and inversely to the ratios of the collapse times (tcmin/tc), where PImax and tcmin were the largest peak impact pressure and the shortest collapse time attainable, respectively. The experimental data

normalized in this way are tabulated in Table 1 and plotted in Figures 7 and 8. The scaling laws motivated in this very simple way appear to

be quite satisfactory, as judged from the rough linearity of Figures 7 and 8. The assumption that K is constant is, of course, true for the cases in which no orifice plate was used (K = Kp), but it becomes increasingly inaccurate as sucessively smaller orifice plates are used (Kp 0.82 if

no orifice is used and decreases to Kt = 0.68 for the smallest orifice). This correction to the data is shown in Table , and has the effect of

shifting the appropriate data points to the left on Figures 7 and 8. This shift would have little effect on these figures, although the linearity of Figure 7 is actually improved.

-25-

x
10

r_ M D _ X cr t cu

o 0 WM O D U8

0.
U4-

80000 088

U,

0
-

4 I
U,

I-

o~0COC r0 0 M 0W 0
0LLrU0OIOm * . . .

M% M
LOO

0 SCL U 4J 0.

4- C
e

. C.

O'IC

4 .'

o 0O 17 LO0 0 tO N 4 CA CD ID %0 C% r-o oot r -l 00 U)

el

0. 0
c

4i cc "fe

LfO 0 co

00 OtOt
h

L) W

le"
4-

0.
co

to c0 U
le 4i V)
**

r= koo L

0 Lfr-- M qC
c '4 l I o co oo r

F-

0 to

Q-

0o

CJ

Mu0 LDuM C rO- O F- O dt el cr0 O _C Oo

oO L) ooo - - a- a

U I--,

CZZ

-26-

O,

CO

'1 a

_.snNoUt
f
I-a

L0L. 4J

CY

c N

_ 40

a E

S. cx Y_

U-

to 4

4
d

_~~~ r
vyCr "

*I~~~ ICL

t' c
w
o to

4- U

Ui. r-CA 4-

O
L,.

O
mialr

caD

d
(-1)

Cu

podw I!l!uI JnssOJd

-27-

aw
z 0 CL

(D

d
.

_0~
4S-_

.0'
on
e%

>U q

Ow

i') d
C\

) C to UA 4-
a,

00

LL

o
m

OD (o (

( u

v o d

'M os d

0 0

ulw*4

-28-

VI.

Modeling

The steam bubble collapse process, quite complex in reality, can be modeled by applying idealized governing equations for the momentum in the water slug and the continuity, state and energy of the steam bubble (Figure 9). solved numerically. The dynamics of the water slug are determined by the unsteady Bernoulli equation The resulting four differential equations can be

P /p

+ gh = Ps/p

+ 1/2 (l+k)V

(4)

where Po is the reservoir pressure, p

is the density of the liquid,

g is the acceleration due to gravity, h is the height of water in the reservoir, Ps is the pressure in the gas bubble, k is a localized loss coefficient, V is the fluid velocity in the central structure, and x is the distance the fluid has traveled in the central structure (Figure 9). In this system the only viscous loss accounted for is at the entrance
to the central structure (x = 0).

Continuity for the steam bubble requires that the mass flux out of the steam bubble equal the rate of change of mass in the bubble:

-mc

d 5

d= t(PsA(2o-X))

(5)

where mc is

the net steam mass condensation rate,

ps is

the density of

the steam, A is the cross-sectional

area of the central

structure,

and

of the steam bubble (Figure

-29-

Po

j::
h

'.:ii
I
-

: i ::
I[ [

T i I . f: . . 2 ' Pe . :: :: ` ,:.:,:::: : .::I

Reservoir

1
Moving

..,,

:.:.'';'::.
.

~~~~r-'

: '. 1' P,

...

l~~~~r~.

~~

...... ;. '.;.
I1

X-O

Water Colu mnr

,i I I I

.I

I
-~S

Control Volume

I Steom'

I I

|T

P. I

I I,

I
--

Stationary
Steam - Water

.:.,,. ...
Stationary
Water Column
..- '.: ..... '.:

.-........

Interface

.,.::: : ::
:i : * .-'. ',.. - . ',,,.

Figure 9.

Control Volume for Collapse Dynamics

-30-

The state equation for the steam can be greatiy simplified if the
steam is modeled as an ideal gas:

Ps

PsRTI

(6)

Here R is the gas constant and T1 is the temperature of the steam, assumed uniform throughout the bubble (Figure 9). The energy equation for the steam bubble is written as

-psA( Uf~ S

o-XCvT1, o

) XCvT

Ps

pT d mccpT 1 - PS 5-t(A(L 0 -x))

(7)

where cp and c

are the specific heats of the gas under constant pres-

sure and volume, respectively. Equations (4) lowing variables: (7) can be nondimensionalized in terms of the fol-

= P/Po

= PS/Ps(O)

T* = Ps(O)RT1 /Po

Va = (Po/P)1/2

x* = X/o
V* = V/V

t*= Vat/,
a

dx*

a V

dt*

c* =

c/Ps(O)AVa

The dimensionless governing equations (see Appendix I for derivation) are thus:

1 = P* + 1/2 (l+k)(t)

dt*

(8)

-31-

-c*

d(p*(1-x*))

(9)

P* = p*T*

(10)

dt(p*(l-x*)T*)

= -c*yT* + P*(y-l)dX*

(11)

where y

s the ratio of specific heats (its value must be calculated

from the steam tables). Equations (8) - (11) have been solved numerically using the conditions of the P/Pomax = 0.6 test in Figure 5. In dimensionless terms

the necessary initial conditions and system parameters for this test are

x*(O) = 0
V*(O) P*(O) T*(O) P*(O) = 1 = 1 = 1 = 1

k = 0.5
Y = 1.3

The dimensional reference values for this test are

P0 = 300 kPa
Va 17 m/s

'o

0.9 m

It is important to note that c*, the dimensionless rate of mass transport from the steam to the liquid, is an input to the equations. Figure lob shows predicted bubble collapse pressure histories for

-32-

TIME AFTER

OfNING ( t) VALVE
0 40
I

do
I

120
1

-300
0.80 -32O0(

Pt

(a) I0.6 0.
0.2 -

I/6I

t.0

2.0
T

DMmS0NLESS TI
1.u

0.9

0.7 0.6 0.5

0.4

0.3
0.2 0.1

M'M

o0.

0.5

1.0
To,

1.5

2.0

DIMENSIONLESS TIME

Figure 10. (a) (b)

Dimensionless Bubble Collapse Pressure Histories Experimental trace for conditions in Figure 5 with Po/Pomax = 0.6. Model predictions for same conditions, varying c*.
Curve 1, c* = .404; Curve 2, c* = .485; Curve 3, c*

.566

three different values of c*:

c* = .404, .485, and .566 .

These

values correspond to heat transfer rates (-9 =ichfg/A, where hfg is

Comparing the predictions in Figure 1Ob with the experimental pressure trace in Figure 10a, -q = 30 MW/m2 . it is clear that agreement is good when

This simulation produces fairly accurate pressure am-

plitudes over most of the collapse process and produces a collapse time almost identical to that observed experimentally. Figure 11 shows corresponding calculated position, velocity and pressure histories for -; 0.68, or V I - 12 m/s. = 30 MW/m2 . The velocity at impact is V* =

This predicted impact velocity can be compared The initial

with the experimentally measured initial impact pressure.

impact pressure (PI) measured in the steel pipe is related to the impact velocity (VI) in the lexan section by (Appendix II)

PI

) Pt VICCs/(Cl+Cs

(12)

where c

and c s are the respective wave speeds in the lexan tube (cl
=

460 m/s) and the steel pipe (cs pact pressure predicted in

1370 m/s).

The amplitude of the im-

this way (PI

4100 kPa) compares remarkably as recorded

well with the actual impact pressures for these conditions

Use of a single

condensation

rate in

the numerical calculation

yields good agreement with the experimental data for the low back pressure case discussed above. Figure 5 shows, however, that as the back

pressure is increased, the collapse process is characterized by a high initial pressure drop rate and a subsequent region where the pressure

-34-

0.9 0.0

; 0.7
Cm

0.5
0.4 F I0.3

is dla 0.0

0S

i.0

1.5
To

2.0

25

DIMENSIONLESS TIME

Figure

11.

Dimensionless
(-

Bubble Pressure and Water Column

Position and elocity

histories

for c* = .485

-35-

drop rate is much smaller, approaching zero for some of the cases (Figure 5). By analyzing these observed pressure drop rates (Figure 12),

approximate net heat transfer rates from the steam to the water during different stages of the collapse process can be inferred. Modeling the

steam as an ideal gas that is undergoing a heat loss per unit crosssectional area 1 due to condensation at its interface with the liquid gives, very roughly (Appendix III),

[(ox)) 1(

_ Ps

(13)

The heat transfer rate during the initial depressurization (;o) is determined when dP/dt is given by (dP/dt)o (Figure 12) and x The heat transfer rate during the subsequent phase ( when P is
s

dx/dt

0.

) is determined o 0 /2 and o/tc,

(Figure 12) and x and dx/dt are given by

resectively.

These different phases are not very pronounced in the low

back pressure tests, but are evident in the higher back pressure tests (Figure 5). An average condensation rate () o/2, and can be found when P, x,

and dx/dt are again given by P, dP/dt is given by

o/tc, respectively, and

()

(Figure 12).

Substituting the appropriate valinto

ues of P, dP/dt, and tc which are read from the oscillogram in lOa Equation (13) along with those of y and ko gives 7 MW/m2 , and = 8 MW/m2. 2 = 40 MW/m,

-qs

These values are tabulated in Table 2 along

with the value used in the numerical solution for the low back pressure simulation. Equation (13) suggests that in the early phases of bubble collapse the condensation rate can be inferred from the initial slope of the pressure history. Experimentally, this slope is found to increase with

-36-

C --

- -- -

--

---

--------

--I

-- ----

*1M

4(1r

a~~c
o~~~
Xa a CL

s- o 00CA
0J (AB
i

:- co.

0a

~s o
P.

cna
4' *

Fly = r

rS:0 Q
.-

X cr

r)e

Cac
I

LL

aa

Q:

-37-

Table 2:

Summary of condensation rate calculations

for a
Figure 5)

Condensation rate
'qo

(MW/m2 )
40

Experimental -'s
-q

7
8

Numerical Theoretical Estimates

-q

30 600

100

-38-

higher back pressures (Figure 5), iment (Po/Pomax rate is

For the hghese

back pressure exper-

1.0, Figure 5), the inferred initial heat transfer

100 MW/m2, considerably higher than the initial heat 'o

transfer rate for the low back pressure test. The initial and subsequent heat transfer rates could be controlled by two different rate limiting phenomena for heat transfer from steam to a liquid: transport of vapor to the subcooled liquid or turbulent

convection in the liquid column. If the transport of vapor is the rate limiting process:

-vap:

'

-4

sT h

l 1

(14)

where a is a constant with theoretical upper limit of 1.0 (Merte 1973). Substituting values corresponding to the test conditions in Figure lob into Equation (14) gives -qvap 600 MW/m 2 (Appendix IV).

If turbulent convection in the liquid column provides the limiting heat transfer mechanism, a heat transfer of the following form is expected (Sonin and Kowalchuk 1978):

'qturb

c4

(T 1 - T2)

(15)

Here c and V are the specific heat and mean velocity of the liquid, D is the pipe diameter, t is the elapsed time from the initiation of the heat transfer, and is an empirical coefficient that relates the efFor turbulent

fective turbulent thermal diffusivity to the product VD.

(Sonin and Kowalchuk velocities and col-

Using V I and t

as characteristic

column

-39-

lapse times, Equation (15) yields -qturb

100 MW/m 2 (Appendix IV).

These theoretically motivated estimates (Table 2) are an order of magnitude greater than the predictions from the numerical solution which gave the best agreement with the experimental results and from Equation (13). One factor that may account for much of this difference

is boiling at the lower liquid interface during bubble collapse (Figure 4) which decreases the effective net mass transfer rates. The the-

oretical estimates do, however, provide consistant upper bounds at least for the conditions of these tests. Independent experiments by Anderson (1980) support the condensation rates inferred from these experiments (summarized in Table 2). His apparatus is shown in Figure 13. Saturated steam is injected into

a submerged vertical tube which induces "chugging" (a cyclic phenomenon whereby steam slowly fills the tube and at some point collapses rapidly). Normalized condensation rates for a pool temperature of 250 C and

saturated steam at atmospheric pressure injected at a flux of 1 kg/m2sec are shown in Figure 14. Measured condensation rates under these condi-

tions were of order 20 - 40 MW/m2 , although some rates were found as

These higher rates correspond roughly to the initial

condensation rates predicted by Equation (13) for the case of highest back pressure (-qo: 100 MW/m2), while the majority of the rates (20 -

40 MW/m2 ) correspond to the rates predicted by Equation (13) for the low back pressure test (-qo = 40 MW/m2
)

and by the numerical solution

which gave the best agreement with the experimental results (-q = 30 MW/m2 ).

-40-

Insulated Tank

PressWe
Reducer

600kPo

Steam Trap

Water Level

ccccccclAI
--LH 4 -C-C" c --cl -

T
I-

I
I
I

-I

I- --Window 135cm
I
I I I I I I ---1

_---

--

_-

Pool
Temperature

Thermocouple

-'-- I-

- _
_ _ . .

-j
115 cm

61cm
deep

Figure 13.

Schematic of "Chugging" Apparatus (Anderson 1980).

-41-

o
4-

r-

CA

4 .0 O=

4rm

CC 0
Co

a e

sO

-(,r/M)

-42-

VII.

Conclusions

Impact pressures and collapse times scale well with back pressure and flow resistance according to the proposed scaling parameters. Higher back pressures and lower flow resistances produce higher impact pressures and shorter collapse times. The bubble collapse process is also very sensitive to the condensation rates from the steam to the subcooled liquid. Good agreement

between prediction and experiment was obtained for one set of test conditions using a simple heat transfer rate as the sole empirical input to the analytical model. Other collapse conditions appear to exhibit The con-

heat transfer rates that decrease during the collapse history.

densation rates that produce the "best fit" with the experimental data are consistently below the condensation rates predicted by two theoretically predicted limits.

-43-

REFERENCES

Anderson, W. G., private communication, April, 1980. Baumeister, T., and Marks, L. S., Standard Handbook for Mechanical Engineers, Seventh Edition, McGraw-Hill, New York. Creare, Inc., "An Evaluation of PWR Steam Generator Water Hammer," NUREG-0291, NRC-1, May 1977. Hurwitz, W., "Piping Network Response to Steam Generated Water Hammer," M.S. thesis, Department of Mechanical Engineering, M.I.T., May, 1980. Merte, Jr., H., "Condensation Heat Transfer," n Advances n Heat Transfer (T. F. Irvine and J. P. Hartneft, eds.), Vol. 9, Academic Press, New York, 1973, pp. 181-272. Sonin, A. A., and Kowalchuk, W., "A Model for Condensation Oscillations in a Vertical Pipe Discharging Steam into a Subcooled Water Pool," NUREG/CR-0221, June 1978,

-44-

Appendix1;

Derivation

of Dimensionless Governing Equations

Unsteady Bernoul 1 i:

Po/Pj + gh = Ps/p + 1/2 (l+k) V2 + x dV/dt

0 JE s k~~~~~~~~

(4)

1 + pgh/P

= Ps/Po + 1/2 (l+k)(p /Po)(dx/dt)2 + (p /PO) x d2 x/dt2

For these tests

gh/P

0.007.

Gravitational effects are small

compared to back pressure and can be neglected.

1 = P* + 1/2 (l+k)

Va

(Vadx*/dt*)2 + Va

x (Va 2 d 2 x*/dt* 2 )

1 = P* + 1/2 (l+k)(dx*/dt*)

+ x d2x*/dt*2

(8)

Steam bubble continuity:

-mc =C d/dt (PsA(Z ~~S 0 -x))

(5)

-mc/Ps(O)AVa

-1 (V 19,)

d/dt*

i(p /P

())(

-x)I

-c* = d/dt* (p*(l-x*))

(9)

Ps

PsRT 1

(6)

-45-

PS/Po = (ps/Ps(O))(Ps(O)RT

1 /Po)

P* = p*T*

(10)

Energy equation:

d/dt

(psA(Qo-x)cvT)

mccpTI -Ps

d/dt (A(o -x))

(7)

AVaP

ao

[Ps

LS

Ft 0~ s(
A(Yo-X)cvPs(O)Tl

ao CT
jI~(1

AVaPo

cp

1
1

- P dt(A(Po-X) )

]
Ps Ro dx*
Po Va dt*

;9o -(p*(-x*)cT*)
Va Z

mc

Ps (O)RT

- Ps(O)AV

Cp

oP 0

d/dt* (p*(l-x*)T*)

= -c* YT* + P* (Y-1) dx*/dt*

(11)

-46-

Appendix II:

Perturbation Analysis

The linearized perturbation equations at an interface separating regions with different wave speeds give the relationships between the velocities and pressures across the interface.
t

Wave speeds:

= 460 m/s

c s = 1370 m/s

Linearized perturbation equations:

(V1 - V 3 ) (V3 - V 2 )
(V3 V)

(a)

P3 ' P2 P4
P5
-

PC

(b)
(C)

P3 = PC1
P6
=

PC s (V5 - V6 )

(d)

Boundary conditions:

V2 = V6 P2 = P6

V4 = V 5 P 4 = P5

Initial conditions:

P1

0 , P 2 = P6

0 compared

to P3

V1 = V I, V 2 = V 6 = 0

-47-

Utilizing

the initial conditions,

Equations

(a) and (b) become:

P 3
P3

pcl (V1 - V3 )
PclV 3

(e)
(f)

Combining Equations (e) and (f) yields:

V3 = V1 /2

(9)

Utilizing

the boundary conditions,

Equation (c) becomes:

P5

P3 + pcl

(V 3 - V 5 )

(h)

Utilizing

the initial

conditons, Equation (d) becomes:

'V5 = P5/pc s
Combining Equations (h) and (i)
P5
=

(i)

yields:
(V 3 - P5 /PC s )

P 3 + PC1

P5 (l + Cl/Cs)

P3 + PC 1V 3

P5 (P 3 + PclV 3 )/(1 + cl/Cs) CombiningEquations (f) and (j) yields: P 5 = (Pc1V3 + pclV3 )/(l
P5
Combining Equations
=

(j)

+ Cl/Cs)
(k)

2p(Cl Cs/cl + Cs)V3

(g) and (k) yields:

P5 From the numerical

P(CCs/Cl

+ CS)V1

simulation,

VI = 12 m/s.

Also,

P5

PI, so

PI

P(ClCs/cl

+ Cs)VI
m/s](12 m/s)

(12)

PI = (1000 kg/m3)[(4601370)/(460+1370)
PI = 4100 kPa

-48-

Appendix III:

Condensation Rate Calculation

Energy equation for steam bubble:

d/dt [psA(Qo-x)cvTl] = -c

pT

- Ps d/dt

A(ko-x)]

(7)

Modeling the steam as an -mccpTI

deal gas gives p

I = Ps/R, and noting that

can be represented as

A, Equation (7) becomes

dt

[PsA(Z-x)

y:T

= lA

Ps

dt [A(.O'X)]

Assuming constant Y gives:

YA (o-Px)
PdAd

+ PsA dt

1
q
=

dP
dt ! dt

y1 (-X)

(s

l = y-1 [(Qo X)dt

s 0- ~dx]. Y-1
YPs
d

(13)

-49-

Appendix IV:

Evaluation of qvap and

turb

~-4vap

aphfg s4

(14)

Modeling the steam as an ideal gas, RT1

Ps/Ps,

gives:

Pshfg
4 4

(8Ps
s

vap

For saturated

steam at Ps = 300 kPa, Ps

= 1.651

kg/m 3 kJ/kg

hfg = 2163.8

Using these values gives -qvap

600 MW/m2 for a = 1.

4Iturb;

-qturb=

p c

(T1

T2 )

(15)

Substituting V I and tc for V and t gives:

-qturb

C---

(T1 - T 2 )

where p.:

1000 kg/m3

c : 4200 J/kgC

8
D T 1 - T2 :

10 - 2
0.04 m 1100 C

VI = 12 m/s tc 0.1 s

Using these values gives -qturb

100 MW/m2 .