An Analytical Model for Prediction

of Two-Phase (Noncondensable)
Flow Pump Performance
During operational transients or a hypothetical LOCA (loss of coolant accident)
condition, the recirculating coolant of PWR (pressurized water reactor) may flash
Okitsugu Furuya into steam due to a loss of line pressure. Under such two-phase flow conditions, it is
TetraTech, Inc., well known that the recirculation pump becomes unable to generate the same head
Pasadena, Calif. 91107 as that of the single-phase flow case. Similar situations also exist in oil well sub-
Mem. ASME mersible pumps where a fair amount of gas is contained in oil. Based on the one
dimensional control volume method, an analytical method has been developed to
determine the performance of pumps operating under two-phase flow conditions.
The analytical method has incorporated pump geometry, void fraction, flow
slippage and flow regime into the basic formula, but neglected the compressibility
and condensation effects. During the course of model development, it has been
found that the head degradation is mainly caused by higher acceleration on liquid
phase and deceleration on gas phase than in the case of single-phase flows. The
numerical results for head degradations and torques obtained with the model
favorably compared with the air/water two-phase flow test data of Babcock and
Wilcox (1/3 scale) and Creare (1/20scale) pumps.

1 Introduction
1.1 Background. During operational transients or a Since the mechanism of pump head degradation has not been
hypothetical LOCA (loss of coolant accident) condition, the well understood, effort has simply been made on separation
recirculating coolant of PWR (pressurized water reactor) may of gas from oil. It is again desirable to understand the basic
flash into steam due to a loss of line pressure. The pressurized mechanism of head degradation so that proper pumps can be
water will now contain vapors, the amount of which will used for pumping gaseous oil from deep wells with a higher
depend upon the extent of the transient or accident condition. efficiency.
With such a two-phase flow medium as a coolant, it is well
known that the recirculation pump becomes unable to
generate the same head as that of the single-phase flow case. 1.2 Review of Past Test Programs. Extensive two-phase
The degree of head degradation depends on physical, pump tests have been conducted to date (see Table 1 and a
geometrical and thermal conditions, but it was indicated in recent review article by Kim [1]), but the geometric and
various experiments that the head was almost totally lost physical conditions were different in these testings. These
under certain conditions. From a PWR's safety point of view, include (i) scale and pump geometry, (ii) pump type (radial,
it is therefore essential to understand such phenomena and mixed, or axial flow type) and, thus, specific speed, (iii) two-
develop an analytical capability for accurately predicting the phase flow media (condensable or noncondensable), (iv)
performance of PWR's cooling pump under the two-phase pump operational conditions, and (v) line pressure. Due to the
flow conditions. Such an analytical model can then be in- different pump features and test conditions existing in these
tegrated into PWR system thermalhydraulic analysis codes. tests, the results substantially differed from one to another.
Similar situations also exist in submersible pumps for deep (/) Scale and Pump Geometry. As has been shown in
oil well pumping where a fair amount of gas is contained in Table 1, the scales of C-E (Combustion Engineering, see
oil. The head of each pump in the multistage pump con- reference [2]) and Creare pumps [3] are 1/5 and 1/20,
figuration is substantially lower than that of single-phase respectively. There existed some difference in homologous
flow, thus resulting in many hundred stages of pumps. The head between the C-E tests and Creare tests (see [3]). It is quite
same head degradation phenomena is prevalent, but in this possible that the scaling of bubble size to pump size plays an
case the two-phase flow media consists of oil and non- important role in the head degradation mechanism. Another
condensable gas, unlike the above PWR cooling systems. important factor related to the pump geometry will be that of
detailed pump blade profile shape, as was pointed out by
Chen and Quapp [4]. Although the specific speeds of the
Contributed by the Fluids Engineering Division and presented at the Winter pumps tested in the Fast Loop Facility [4] and C-E were
Annual Meeting, New Orleans, La., December 19-13, 1984 of THE AMERICAN
SOCIETY OF MECHANICAL ENGINEERS. Manuscript received by the Fluids
almost the same (see Table 1), the number of blades, exit
Engineering Division, July 8, 1983. Paper No. 84-WA/FE-8. blade angle, breadth ratio and diameter ratio were quite

Journal of Fluids Engineering MARCH 1985, Vol. 107/139

Copyright © 1985 by ASME
Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Table 1 Review of past experiments
COMPANIES PWR
PRIMARY
EG&G EG&G COOLANT
CHAR^\. (FAST (SEMI- PUMP
ACTER- LOOP SCALE (BYRON
IST1CS B&W1 C-E 2 CREARE 3 FACILITY) FACILITY) KWU JACKSON)
Scale 1/3 1/5 1/20 Not 1/5 of 1/1
Scaled German
KWU
Pump type Mixed Mixed Mixed Mixed Radial Axial Mixed
Type Type Type Type Flow Type
Type
Specific 4,317 4,200 4,200 4,150 926 6,705 4,200
speed
Two-phase Air/Water Steam/ 1) Air/Water Nitrogen/ Steam/ Steam/ Steam/
test (20-120 Water (90 psi) Water Water Water Water
condition psi) (15-1250 2) Steam/ (800 psi)*— (200-900 (15-1200 (15-2250
psi) Water psi) psi) psi)
(400 psi)
Flow rate 11,200 3,500 181 4,500 180 3,145 87,000
(GPM)
Speed 3,580 4,500 18,000 3,600 3,560 8,480 900
(rpm)

1) Babcock & Wilcox

2) Combustion Engineering
3) These pumps are geometrically similar, modeled after the Palisades LWR pumps.

different. As a result, the head degradation of these two radial, mixed and axial flow type. The pumps tested in the
pumps exhibited quite different characteristics (see [4]). Semiscale facility [4], C-E/Creare and C-E for KWU (Kraft
Werk Union of West Germany [5]) represent each of these
(//) Pump Type and Specific Speed. The type of^ three categories, as is shown in Table 1. Figure 1 shows the
hydraulic pump is usually classified into three categories, test results of these three different pumps at design condition,

Nomenclature

ah,ai = accelerations of bubble and liquid u tangential velocity

C„ = drag coefficient v absolute velocity
dn,ds = infinitesimal increments in the n- and s- v„ tangential component of the absolute
coordinates velocity
dm h,dm i = mass in the control volume, for bubble v1/ tangential component of the absolute
and liquid, respectively velocity under the single phase condition
dK,dPs = centrifugal force and pressure force , 4
acting on the control volume Vb volume of bubble
F, = forces acting on the bubble ( - •3 * • )
gravitational constant W/,,Wi flow velocity relative to the rotating
normalized head (= H/HR) blades in the s-direction
AHh,AH, energy or head increase for bubble and x m,,/mTP
liquid, respectively a void fraction
A/icp, AM jp energy or head increase for the single- aN N/NR (normalized speed)
phase and two-phase flows, respectively geometric angle of the s-coordinate made
mh,mhmTp = mass flow rate of bubble phase, liquid with the circle whose center is the center
phase . and two-phase mixture, respec- of rotation
tively nondimensional torque ( = (T/TR)-
AHK,AH„,AHS head losses attributable to the exit flow
a=
(PR/P))
velocity increase, void fraction change v Q/Qfi (normalized volumetric flow rate)
and slip velocity, respectively PI^PI densities of bubble and liquid, respec-
coordinate normal to the streamline tively
coordinate, s pTP two-phase flow density in homogeneous
N pump rotational speed (rpm) flows (= (1 -a)pi + a pb)
P static pressure p*TP two-phase flow density in separated
Po total pressure flows (= mTp/(Q, + Qb))
Q volumetric flow rate
Q„Q„ volumetric flow rate for liquid and subscripts
bubble, respectively b,l denotes the qualities related to the bubble
radial distance measured from the center phase and liquid phase, respectively
of rotation 1,2 denotes the qualities related to the
n, bubble radius suction side and discharge side,
s streamline coordinate respectively
time R denotes the quantities at the rated
t
torque condition
T

140/Vol. 107, MARCH 1985 Transactions of the ASM E

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 showing remarkably different characteristics in head —™——- CREARE S/W CORRELATION (Mixed-Type Pump)
degradation, i.e., larger head degradation for the radial-flow — ™ — —™ CREARE A/W CORRELATION (Mixed- Tvpe Pump ) ~
""" - " - = B&W CORRELATION (Mixed-Type Pump)
pumps and smaller head degradation for mixed/axial flow — * . . — KWU S/W (Kastner, et. al.rj.]) (Axial Pump) ~~

pumps (see also the paper of Lea and Bearden [6]). This seems — ««,.»» SEMISCALE/RELAP CORRELATION (Radial Pump]

to indicate that the blade-through flow dynamics have a

significant contribution to the basic mechanism of pump head
degradation under two-phase conditions.
(///') Two-Phase Flow Media (Condensable Versus Non-
condensable). There existed again a large difference in the
pump performance test results between the noncondensable
gas/water tests and condensable steam/water tests (see Fig. 1
and also the report of Creare [3]). Heidrick, et al. [7]
presented important evidence which seemed to have provided
an explanation for the difference. The reading of gamma-
densitometers installed at the inlet and outlet of the pump
showed that the steam injected at the inlet entirely condensed
through the blades and did not appear at the outlet up to the
steam-to-water mass flow ratio, x = 0.015 (equivalent to
a, = 0.2). It means that any adverse effect due to the existence Fig. 1 Comparison of Creare A/W and S/W head correlations with
of bubbles will only appear around the blade inlet area and B&W and Semiscale/ RELAP head multiplier correlations (after Creare
then disappear towards the blade exit for low inlet void Report [3]) (A/W and S/W stand for air/water and steam/water)
fraction cases. Therefore, the head degradation in con-
This question should be clearly answered before even reducing
densable steam/water media is much less than that of the
the experimental data for comparisons with theories.
noncondensable gas/water media.
1.3.2 Semi-Empirical Correlations. Based on the tests
(iv) Pump Operational Conditions. Every pump is
conducted to date, various empirical curves were constructed
designed to operate with the best efficiency at the design
to characterize the pump performance operating under two-
condition. At a LOCA condition, the recirculation loop loses
phase flow conditions. Such curves include those of Babcock
a substantial amount of coolant and also the pump-motor
and Wilcox [9], Semiscale [4], C-E [2], and Creare [3]. All
rotation speed may change. This will result in an off-design
these curves are more or less similar in that the head and
operation of the cooling pump where the blade-through flow
torque degradation of the pump are expressed by only one
pattern entirely changes. It was shown in various test results,
parameter, i.e., inlet void fraction a,, The approach of
e.g., see the paper of Runstadler and Dolan [8], that the
Wilson, et al. [10] was based on the single-phase ideal Euler's
degree of head degradation changed depending upon
head equation in combination with various empirical head
operational conditions.
loss data for both single and two-phase flow media. It was,
(v) Line Pressure. The upstream line pressure also seems however, shown by Wilson that the correlation between the
to have a significant impact on the head degradation model and test data was rather limited.
characteristics of two-phase flow pumps (see the report of C- The work by Zakem [11] is similar to the method of Hench
E [2] and also the paper of Lea and Bearden [6]). Generally and Johnston [12], who used a one-dimensional control-
speaking, the extent of head degradation is smaller for the volume method for solving two-phase diffuser flow problems.
higher line pressure. The reason for this fact may be three- Zakem applied the method to rotating machinery and then
fold, one related to possible cavitation appearance, the second simply identified a nondimensional parameter. Data
due to the condensation effect and the third due to the density reduction with this new parameter for the test data of Bab-
change of gas or steam phase. cock and Wilcox [9] resulted in some success.
1.3 Data Analysis and Existing Semi-Empirical 2 Analytical Model Development for Two-Phase Flow
Correlations
Pumps
J.J.I Data Analysis. It was pointed out in the report of
Based on the discussions provided above, it is apparent that
Creare [3] that there existed difficulty in analyzing the data
the development of a new analytical model for characterizing
taken by C-E and Creare to define the head increase across the
the performance of two-phase flow pumps is in order. This
pump. In their report, the total head increase H101a| was
should clarify the mechanism of pump head degradation
defined as
under two-phase flow conditions by including com-
SIT _ ( Pom Pin \ 1 r 2 ,, 2l pressibility, line pressure, condensation effect, flow regime,
"lolal - I j + ~ lyout ~vm J slippage of two-phase flow, and, of course, void fraction. A
V
Sc Pom Pin ' 2g t .
new analytical model which has been developed here is a first
where p, p, v denote the static pressure, average density and step for such a complete work. Although it neglects some of
velocity, respectively, and the subscripts, " i n , " and " o u t " the above aspects (as will be stated in the following section), it
identify the quantities at the pump inlet and outlet, respec- is believed that the new method is the most rigorous and
tively. When this formula was used to plot the data of C-E versatile one ever developed to date in this field. Furthermore,
and Creare, it was found that the head became negative even the final formula obtained from this new analytical model
at near design point. As a result, Creare used a simpler form, development conveniently reveals several important features
i.e, static head increase: for understanding the basic mechanism of two-phase flow
8 f, __ Pom ~~P'm pump head degradation. These features also serve as pertinent
-"sialic —
~ - information concerning what type of. new two-phase flow
Sc Pin
pump tests are to be conducted in the future.
With this definition, the data became a little more positive
and thus understandable (see [3]). A question now arises as to 2.1 Mathematical Formulation. The method used here
whether or not the above definition is a right formula for the is somewhat similar to that of Zakem [11], but the solution
real pump head increase of the two-phase flow condition. methods are quite different. Zakem applied the results of two-

Journal of Fluids Engineering MARCH 1985, Vol. 107/141

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 where
Fs = external force acting on the control volume in the s-
direction,
w,,wh = relative flow velocities of liquid and bubble,
respectively in the ^-direction.
The external force, Fs, consists of two components, namely
the centrifugal force, dK, and pressure force, dPs,
Fs = dK-dPs, (3)
where

dK = Gfrn>r»oo2»sin(3'»cos7 (4a)
dp
dP. = — 'ds'dn, (4*)
ds
dm = dm i + dm i, (Ac)
and r is the radial distance of the control volume center
measured from the center of rotation, oo is rotational speed, )3'
and 7 are the local geometric angles of the stream surface
shown in Fig. 2 and p denotes the static pressure. From
equations (2) to (4), assuming a steady state flow so that dm,
and dm,, do not change along the streamline, we find
dw, uw
dw
" b =aw/
• , »co - • a> dp
w w
dm, Ydmh — sin/3 •cos-y — 'ds'dn
dt dt ds
(5)

through the transport theorem. Equation (5) will become

dw, dwh dr_
(1 - a)p, -—w, + upi, —- wb -• [p,(\-a) + pbcx]ru2
ds sin (T cos y = dr
ds ds ds
= sinp cos y
dp
(6)
Fig. 2 Control volume method for rotating machines and force ds
diagrams where dr/ds = sin/3'«cos7, see Fig. 2 and equations (id), (lb),
phase flow equations only to a nondimensional analysis and (4c) have been used. Based on the assumption that all the
without investigating the basic mechanism of head quantities in the above equation (6) change along the
degradation, whereas the present work defined the head streamline direction, s, we will be able to write (6) as
degradation, and then analytically calculated it by solving the d (no)2
governing equations. (l-a)p, + api, {pi,a + p,(\ -a)) +P
A control volume method is employed to investigate the
head degradation for the flow going through the blades of a r >V -(roi)1 P
wh2-(ro3)2l da
(7)
pump under the two-phase flow conditions. In order to apply
such a method we need to make the assumption that the
+|P,— ~2 " 2 J^=0-
trajectory of liquid be identical to that of bubbles so that we Integration of equation (7) from the suction side to
can use a control volume bounded by two streamlines. It is discharge side, denoted by subscripts 1 and 2, respectively,
also assumed that the slip between the two phase flows be yields
allowed. Furthermore, the effects of condensation and w\i- •"2 , ^21,-"2 , ]
compressibility are neglected for simplicity so that p, and ph (\-a2)p,
— + a2p,, +p2j
are assumed to be constant.
The control volume used in the analysis is shown in Fig. 2
where the streamline, s, and the line normal to the streamline,
n, are used as the curvilinear coordinates. The control volume
(l-«l)p<
wjr •u\
+ otiPb
W\l- • « 1
+Pi Hi* wj •

is composed of dn, ds and a unit thickness in the direction da

normal to the stream surface. -Ph • ds = 0. (8)
~d7
The masses in the liquid phase and bubble phase, dm{ and
dmi, contained in this control volume are given The relationship obtained here is considered to be the
Bernoulli equation for the rotating machinery operating
dnii = (1 - a)'dn»ds'Pi (Iff) under two-phase flow conditions. There exists a slip velocity
dtnh = a'dirds'Pi, between the liquid and bubbles and therefore the void fraction
is allowed to change. It must be pointed out, however, that in
where a is void fraction; ph ph are the densities of liquid and equation (8) any nonmechanical head loss existing in actual
bubble, respectively, and the subscripts / and b are used to pump operations, such as head loss due to the frictional drag,
represent the quantities for liquid and bubbles. Then, the has been neglected. However, such an assumption will not
momentum balance applied to this control volume in the affect the final results in determining the two-phase flow head
direction of flow, i.e, s-direction, is expressed as follows, degradation, as will be explained later.
— (dm,-Wi + dm,,'Wh) = Fs (2) We now determine the head degradation of the pump
dt operating under the two-phase flow condition. The energy

142/Vol. 107, MARCH 1985 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ( 1 - a , ) ? , w\i-u] aiPi, w]„-u}
Pfpg Pfpg

— (l-x)+ x
2
S 2g

1 f 2 // !,,,?_
wj-uz 1,2 u 2 \ da
w J -wf,-u'\
-ds.
pfpg J ' ^ 2 2 / ds
(a) ENTRY VELOCITY TRIANGLE (bl DISCHARGE VELOCITY DIAGRAM (17)
Fig. 3 Velocity triangles Use of equations (10), (14), a n d (15) a n d of velocity
relationship w2 - v2 = (u - v„)2 - vu2 (see Fig. 3), provides
increase in the two-phase flow, AHTP, can be expressed as a
sum of the energy increase in the liquid phase, AH,, and that AHTP = AHSp-AHtl,-AHs-AHl„ (18)
in the bubble phase, AHh, where
AHTp-mTP = AHi-mi + AHb-mb (9) vltUi-vlUi VuhUl-VuWl
AHSP = {\-xy (19)
where rhTP, mt a n d mh denote the mass flow rate of the + *•
overall two phase flow, liquid and bubbles, respectively, and
mTP = mi + mh (10a) Avu2lu2 Ay„26«2
AH^ (l-x) + X' (20)
m, = drt'{\-a)v/i'Pi (105)
mh = dira'Wh-ph. (10c) w2j-
Assuming that the flow is inviscid and incompressible with
AHS = (1 -*).U.(^-l)
no thermodynamic effects, we can write AH, a n d AHh in
terms of the difference in the total pressures between the i,. I 1 1 — +x- ( l - a 2 ) H
suction and discharge sides, \ wu / 2g J C V w2h
Pon-Pan Pi~P\ v\r vl
AH,
Pig Pig 2g
(Ha)
0 2g \w]h / 2g )

P02b ~Po\b Pl-Pl v\i, v\b (21)

AH„ (115) 2
PbS Pi,g 2g 1 r / wj-u2 w2, — u2\ da
AH„ -pi, -ds, (22)
where g denotes the gravitational constant. fp-g\>\p<^-
pfpg ds
From equations (9) and (11), and
2
^Tp=PlZPl+v-- •vl v\h ~ v u, v
u\ = v
\!\ ~ ^vu\ > tangential component of v2
(l-x) + (12)
Pfpg 2g 2g v
u2 — V!L> ~ Av,l2, tangential component of vt (23)
where In the above equations, all velocity components are those of
the two-phase flow conditions except for t h e ones with t h e
\u) superscript " 1 0 . " They designate quantities belonging to t h e
"'" Q, + Q„' single phase conditions. It means that the velocity component,
Qi = dn-{\-~a)-w,, (14a) e.g. y','24, is different from v]„f by Av,l2 due to the two-phase
Qi, = dn-a-w,,, (14ft) flow condition as is shown in Fig. 3 . It must also be men-
tioned that the definition of AHW, AHS and AHa in equation
x = m,,/mTP, (15a) (18) has been arbitrarily m a d e . Although we can n a m e (AHW
1 -x = m,lmTp. (155) + AHS + AHa) as A// t o t a ,, it was decided t o use these three
separate terms; AHW is due to t h e increase of t h e relative
PPP is an average density in the separated two-phase flow
speed w of liquid portion at p u m p exit caused by two-phase
condition, which will be reduced to the two-phase density,
flow condition, AHS is attributable to the slip velocity bet-
pTP, of the homogeneous flow if w, = wh, where
ween t h e liquid a n d bubble phases and AHa is d u e to the
PTP = (\ -Ot)'Pi + U'ph. (16) variation of void fraction along the flow passage between t h e
It should be noted that pfP is constant since mTP, Q, a n d blades. Equation (18) will then be written
Q,, are constant owing to the conservation law of the mass AHT A//„, + A / / , + A//„
and flow rate under the incompressible flow assumption. (24)
Equation (12) should represent the head increase of the two-
AH, AHZ,
phase flow p u m p and should be compared with the one used This relation in equation (24) provides the ratio between the
by Creare [3] see Section 1. It can be seen that the formula two-phase flow p u m p head and single-phase flow p u m p head.
used by Creare for test data reduction more or less agrees with It must be pointed out here that since actually measured AHSP
equation (12) except for the dynamic head terms since at far will be used in obtaining the ratio of AHTP to AHSP in
upstream p*rP = pIN. It means that as long as the dynamic equation (24), the previously omitted nonmechanical, head
head terms a r e negligibly small compared to the static loss which was not included in Euler's formula, has now been
pressure term, the formula lof Creare [3], intuitively derived, recovered. It should also be noted that if there exists n o slip
is acceptable. velocity ( i . e . , w / = w/, a n d c o n s e q u e n t l y da/ds = Q),
The term (p2~P\) in the Bernoulli equation (8) is sub- AHu. = AHs = AHl) = Q. For such a homogeneous two-phase
stituted into the same term in equation (12), flow case, AHTP/AHSp becomes unity, indicating no head
degradation within the framework of current assumptions.
(1 - ct2)Pi w2j—u2
AH7
<*iPh
Pri'g
-j,} This new formula for the head degradation rate should be
PTPI compared to the one previously given by Creare [13], i.e.,

Journal of Fluids Engineering MARCH 1985, Vol. 107/143

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 m) 1 1 da. 1 d(dn)
Pr(dn)2 (I-a) [L(l-o!) ds (dn) ds

m \ —a ff 11 dot
da 1 d(dn)
+ pb(dn)2 a2 L a
La ds ds {dn) ds
2
1 |Y mh \ / 1 da 1 d(dn)
+
2P> \ph»drfa' \ a ds (dn) ds
mf \2 / 1 da 1 d(dn)
Fig. 4 Bubble geometry and force ( Pi'dn'(\ -a)/ VI-a ds (dn) ds
2
= (p, — pb)(\-a)ra) sin/3'»C0S7
Q 3 r mj_
AHT PTP
(25) rb 8 • (1 - a)
\-Pi'dn'(\ pb'dr\'a pt-dri'(\ - a )
An An
— (\-x)+ —X
Pi Pb (31)
pb'an'a
where Ap is the static head increase between the pump suction
and discharge sides and pTP is the two-phase flow density where the mass conservation equations (10a) and (10i>) have
under the homogeneous flow condition. The above formula been used.
of Creare was simply based on the intuitive definition, not This differential equation contains only one variable, i.e, a,
reflecting any pump flow phenomena. and all other quantities here are either constant or known as a
In order to calculate the quantities of A//,,,, AHS and AHa function of s. The quantities which belong to the former
in equation (22), we should know the detailed data for the category include mh mb, ph pb, and w, whereas those which
relative velocities for both the liquid and bubbles, as well as belong to the latter category include (dn), r sin/3' and cosy.
the void fraction along the path of blade-through flow. Hench Therefore, if the boundary conditions at the suction side are
and Johnston [12] applied the momentum equation to the known, the ordinary differential equation of (31) can be
bubble for this purpose, when they solved the problem of two- solved for a along5. Once a(s) is obtained,/?(s) can be solved
phase diffuser flows. Applying the same philosophy to the by using either equation (6) or (30). Furthermore, w, and wb
present rotating machinery problem, will be determined from the mass conservation equations in
(10i>) and (10c). Now all quantities necessary for calculating
AH„, AHS, and AH„ are made available.
(26)
Torque. The calculations of torques corresponding to the
where degraded heads of the pumps operating under two-phase flow
conditions will be made possible with a simple assumption as
\jFs = (Pressure Force) + (Centrifugal Force) + (Drag
follows. For the case of single-phase flow, the relationship
Force) + (Virtual Mass Force) (27)
between the torque, head and flow rate is written
and Vh denotes the volume of bubble. Each force acting on
TH.o> = (H^ + AHH)'Q.p.g (32)
the bubble can be written in the following form (see Fig. 4);
where HH is the actually measured single-phase flow head
dp_
(Pressure Force) = - •Vh (28a) and AHX4> is the head loss, corresponding to a portion of the
ds input torque not converted into head. Assuming that the same
(Centrifugal Force) = pb Vb ra>2 sin/3' • C0S7 (28ft) amount of head loss, A// ] 0 , can be applied to the two-phase
flow cases, we can define the torque for two-phase flow
(Drag Force) = Cd» -pr(w, — wb)' \w,-wb\wb2 (28c) conditions
Tu.o> = (Hu + AH^).Q.p*Tp.g. (33)
(Virtual Mass Force) = pl'Vh'(ab — al) (28d) By eliminating AHi4, from equations (32) and (33),
where T2,t,'u> = (H2j>-HH)'Q>p*Tp'g+T^(j>.
r,, = radius of bubble, assuming that the bubble is Normalization of each term in the above equation will yield
spherical, ft>0 _ ( hu hH \ v | /3, 0
VR + (34)
dwb dwb dwb <*N aN / aN aN-
(29a)
dt dt ds where

dw/ dwt dw, VH = hydraulic efficiency of the pump at the rated con-
at (29b) dition for the single phase flow
~dt 3/ ' ds
HRQRPZ
Assuming the steady-state flow condition and substituting
TRUR
equations (28) and (29) into equations (26) and (27), we obtain
dw,, 1 / dWj, dw, \ dp
P W + pbr•oJ2sm|3, h = normalized head ( = H/HR)
" "^+2f"V ds "'-ds-) ds aN = normalized rotational speed (= N/NR)
Cj 3
0 = normalized torque(= (T/TR) '(pH/p))
•COS7 + Pl'(W,-Wb)\Wi-Wb\. (30) v = Q/QR
rb 8
H = head
Equation (30) is now combined with the momentum N = rotational speed
equation for the two-phase mixture, equation (6), to eliminate T = torque
dp/dsterm, Q = volumetric flow rate,

144/Vol. 107, MARCH 1985 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 © CREARE AIRAVATER DATA
• Saw AIR/WATER DATA
™ " PRESENT THEORY

~^5"

INLET VOID FRACTION, a

INLET VOID FRACTION, «..
Fig. 7 Comparison of present theory for homologous heads with
Fig. 5 Comparison of present theory for homologous heads with Creare and B&W data forW« N between 1.2 and 1.3
Creare and B&W data for Wajy between 0.2 and 0.8

9 CREARE AlRftVATER DATA

O CREARE AIR/WATER DATA D B&W AIR/WATER DATA
D B&W AIR/WATER DATA — PRESENTTHEORY

— PRESENT THEORY

INLET VOID FRACTION, O INLET VOID FRACTION, a

Fig. 6 Comparison of present theory for homologous heads with Fig. 8 Comparison of present theory for homologous heads with
Creare and B&W data for WaW between 0.8 and 1.2 Creare and B&W data for vlaN between 1.3 and 2.0

and subscripts "R" denote the quantities at the rated con- the range of interest was shown to be small. For the churn-
dition. All the quantities necessary for the above equations turbulent flow regime, the drag coefficient is drastically
are available either from the single-phase pump test results or reduced. A combined relationship between Cd and /-,, was
from the results obtained with the present analytical model empirically obtained by Zuber and Hench [16]
calculations. Cd
— =2.79(1 a) 3 (in-'), o r . 1 1 ( 1 - a ) 1 (mm ').
3 Comparisons Between the Theory and Test Data
The analytical model developed here was used to calculate It must also be mentioned that the added mass term was
the head degradation and torque for the pump tested by dropped from equation (31) since the term applies only to
Babcock and Wilcox (1/3 scale) and Creare (1/20 scale), spherical bubbles and therefore is not appropriate for the two-
operating under air/water two-phase conditions. The input phase flow of churn-turbulent regime. Hench and Johnston
data necessary for such calculations included in the single- also pointed out that this term is negligibly small compared to
phase flow performance of the pump, pump geometry, rated other terms in the equation. The values for Cd/rh between
conditions, physical parameters of air and water, drag a = 0.3 and 0.4 were those of linear connection between these
coefficient, Cd and bubble size, rh. two points.
The last two values, Cd and /-,,, employed in the present Figures 5 to 8 show the homologous heads calculated from
calculations were the same as those used by Hench and the present analytical model in comparison with Babcock and
Johnston [12]. First, they classified the flow into two regimes Wilcox (1/3 scale) and Creare (1/20 scale) air/water test
depending upon the local void fraction, a, i.e., bubbly flow results. It is seen that the overall agreement is excellent,
regime for a = 0.3 and churn-turbulent flow regime for particularly for the homologous flow parameter, v/as~\,
a > 0 . 4 . In the bubbly flow regime, it was found in their i.e., near the rated condition. The trend of measured head
experiments that the bubble radius was about 0.0625" ( = degradations is well represented by the present analytical
1.59 mm) and thus the corresponding drag coefficient was model, i.e., a relatively mild head degradation up to a, (inlet
determined to be 0.54 based on the experimental data of void fraction) = 0.2, then a sudden drop between a, = 0.2
Haberman and Morton [14]. Regarding the choice of Cd and and 0.3, followed by a fairly flat low head for 0.3 = a, =
/•,, for the bubbly flow regime, it was reported in the paper of 0.9 and a return to the single-phase homologous heads as a,
Hench and Johnston [15] that the relative sensitivity of the reaches unity.
results of this analysis to bubble size and drag coefficient in It must be mentioned that the test data shown here were

Journal of Fluids Engineering MARCH 1985, Vol. 107/145

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1 1 I 1**——i i - T r—- T _ — p — ^ — ^ — _ l — „ _ , _ - T - — n — • — i 1 —
- 0.9
0.7 —
_ _ — — — • * *

- 0.9^^ #^
- ^^^0^7 ^ , ^ - 0.4^-^ *
*<^^^ ^^^^"^A
^
- -•

0.25 ————— __
-^^O^
-
^ 0 1 ^ ^ 1 1 1 ] L_ 1 L_ 1 —. 1

NORMALIZED STREAMLINE COORDINATE

"""————^^^;. 0.1 ~ <s'Stot»l>
_ Fig. 11 Void fraction changes along the flow passage for vlaN = 1.0
«,-0 - forthe inlet void fractions, a-\ = . 1 , .25, .4, .7, and .9
-
1 1 1 L 1 1 1 1 1

CREARE AIR /WATER DATA

NORMALIZED STREAMLINE COORDINATE
B&W AIR/WATER DATA
<S/Stotal> PRESENT THEORY
Fig. 9 Relative liquid velocity changes along the flow passage for
vltXtf = 1 . 0

_ _ , — i — i — i — i — i — 1
i

I^---. V°
-
,,
'^ ^---»WBiiii<^ rig

^""^^s. oT
^ ^ s T

- ^ — * ^ ^ ^
^ • ^ ^
o.i
0.4 --.
~
\ o . 2 5 ~~

: INLET VOID FRACTION, a

Fig. 12 Comparison of present theory for homologous torques with

Creare and B&W data for i / a N between 0.8 and 1.2

Figure 11 shows a continuous increase in local void fraction

1
..i i i i i _ i 1 1 along the flow passage, indicating larger population of air
bubbles towards the pump outlet. Patel and Runstadler
NORMALIZED STREAMLINE COORDINATE reported in their paper [17] that the large void in the impeller
<s'stol,l> was observed in air/water pump tests. Murakami and
Fig. 10 Relative gas velocity changes along the flow passage for Minemura [18] pointed out the association of large void
vlaf) = 1.0
formation with an extensive degradation of pump head.
Photographs taken by Manzano-Ruiz and Wilson [19] also
taken from the report of Creare [3] in which the data were showed some evidence of bubble agglomeration. The effect of
classified as a function of wide range of v/aN values, except centrifugal force on the head degradation also seems to ex-
for 1.2 ^ v/aN Sr 1.3. The test data for 0 ^ v/aN ^ 0.8, plain the more severe head degradation for the radial flow
0.8 ^ i>/aN 5? 1.2 and 1.2 S v/aN =? 2.0 were all lumped pumps in which the centrifugal force is the major head
together, as seen from Figs. 5 to 8. Although the report of generation force. The fact indicates that use of axial-flow
Creare [3] stated the insensitivity of head degradation to the pumps may provide better performance than radial- or mixed-
homologous flow parameter v/aN, theoretically there exists a flow pumps in two phase flow conditions (see Fig. 1).
clear distinction in head degradation for different values of For extreme off-design condition cases, such as v/aN =
v/aN. It seems that the difficulty in classifying the test data as 0.4 (Fig. 5) or v/aN ^ 2.0 (Fig. 8), some discrepancy between
a function of v/aN is attributable simply to wide scattering of the analytical results and test data exists. It is not surprising to
the test data. find the discrepancy at such extreme conditions. Despite the
The basic mechanism of these head degradations shown in physical fact that the streamline profile shapes may be sub-
Fig. 5 to 8 has already been explained in the previous section. stantially different near the blade inlet area between the
However, it will more clearly be understood by investigating design condition and off-design conditions, the present model
the detailed quantities of blade-through flow obtained in the only used those of the design condition. It is indicative based
present calculations. Figures 9 to 11 show change of liquid on this fact that the future analytical model may need to
and gas flow relative velocities and local void fraction, employ flow passages at off-design conditions different from
respectively, along the flow passage. Homologous flow that of design condition.
parameter is v/aN = 1.0 and the inlet void fractions were ax Figure 12 shows the calculated homologous torques
= 0.1, 0.25, 0.4, 0.7 and 0.9. Due to the difference of flow corresponding to the cases of Fig. 6. Again the overall
media densities between water and air, the former is subject to agreement with test data, particularly with those of Creare, is
larger external force, in this case, centrifugal force, than the good. It is not known why the test data of Babcock & Wilcox
latter. Consequently, the water is accelerated faster and the are so scattered. The.limiting values of homologous torque
air is decelerated slower than in the case of single-phase flow. test data as o, — 0 do not match with the single-phase
This acceleration of the water contributes to the major part of measured data which are shown by the analytical results at at
head degradation, i.e., in terms of AHw. = 0. This fact indicates substantial error in measurements.

146/Vol. 107, MARCH 1985 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Although such errors were observed in the homologous head for carrying out all numerical computations for the present
measurements (see Figs. 5 to 8), the degree of the errors was work.
much smaller. As far as torque measurements are concerned,
those by Creare are less scattered and compare favorably with References
the analytical results. 1 Kim, J. H., "Perspectives on Two-Phase Flow Pump Modeling for
Nuclear Reactor Safety Analysis," presented at Cavitation and Multiphase
4 Conclusions Flow Forum of ASME Conference, Houston, June 1983, edited by J. Hoyt, pp.
29-33.
Based on the one dimensional control volume method, an 2 Combustion Engineering, Inc., "Pump Two-Phase Performance
analytical model has been developed to determine the per- Program" Vol. 1 through VIII, EPRI NP-1556, Sept. 1980.
3 Creare, Inc., "Two-Phase Performance of Scale Models of a Primary
formance of pumps operating under two-phase flow con- Coolant Pump," EPRI NP-2578, Sept. 1982.
ditions. The analytical method developed here has in- 4 Chen, T. H., and Quapp, W. J., "Centrifugal Pump Performance Under
corporated pump geometry, void fraction, flow slippage and Simulated Two-Phase Flow Conditions," The Symposium on Polyphase Flow
flow regime (implicitly through the bubble diameter and drag and Transport Technology, Aug. 1980.
5 Kastner, W., and Seeberger, G. J., "Pump Behavior and its Impact on a
coefficient) into the basic formula, but neglected the com- Loss-of-Coolant Accident in a Pressurized Water Reactor," Nuclear
pressibility and condensation effects. Technology, Vol. 60, Feb. 1983.
During the course of model development the basic 6 Lea, J. F., and Bearden, J. L., "Effect of Gaseous Fluids on Submersible
mechanism of head degradation was identified. Three Pump Performance," Journal of Petroleum Technology, Vol. 34, Dec. 1982.
7 Heidrick, T. R., et al., "Centrifugal Pump Behavior in Steady and
separate terms contributing to the head degradation are called Transient Steam-Water Flows," The Symposium on Polyphase Flow and
here A//„., AHS, and A//„. AH„, is due to the relative flow Transport Technology, The Winter Annual Meeting of the ASME, Dec. 1978.
speed increase of liquid particles at pump exit, AHS is at- 8 Runstadler, P. W., Jr., and Dolan, F. X., "Two-phase Flow Pump Data
tributable to the slip velocity between two phases at inlet and for a Scale Model NSSS Pump," ASME Symposium on Polyphase Flow in
exit and AHn stems from the void fraction change along the Turbomachinery, Dec. 10-15, 1978, San Francisco.
9 Babcock & Wilcox, "One-Third-Scale Air-Water Pump Program, Test
flow passage. It has been found in the numerical calculations Program and Pump Performance," EPRI NP-135, July 1977.
that the A//„,'s contribution to the head degradation is much 10 Wilson, D. G., et al., "Analytical Models and Experimental Studies of
larger than the sum of (A//s. + A//„). Centrifugal-Pump Performance in Two-Phase Flow," EPRI NP-677, May
1979.
The numerical results for head degradations and torques 11 Zakem, S., "Analysis of Gas Accumulation and Slip Velocity in a
were obtained and favorably compared with the test data of Rotating Impeller," ASME Cavitation and Polyphase Forum, 1980.
air/water two-phase flow conditions obtained by Babcock & 12 Hench, J. E., and Johnston, J. P., "Two-Dimensional Diffuser Per-
Wilcox (1/3 scale) and Creare (1/20 scale). The overall formance with Subsonic, Two-Phase, Air-Water Flow," ASME Journal of
agreement is excellent despite the fact that the test data were Basic Engineering, Mar. 1972.
13 Creare Inc., "Review and Analysis of State-of-the-art of Multiphase
so scattered and inconsistent. Pump Technology, " E P R I NP-159, Feb. 1976.
For the extreme off-design conditions, some discrepancy 14 Haberman, W. L., and Morton, R. K., "An Experimental Investigation
existed in the above comparison. It seems that the pre- of the Drag and Shape of Air Bubbles Rising in Various Liquids," DWTN-
SRDC, Report 802, Sept. 1953.
determined stream-surface used in the analysis will need 15 Hench, J. E., and Johnston, J. P., "Two-Dimensional Diffuser Per-
modification for such extreme off-design conditions. formance with Subsonic, Two-Phase, Air-Water Flow, General Electric Report
Furthermore, in order for the analytical model to be able to APED 5477, Jan. 1968.
determine the performance of steam/water two-phase flow 16 Zuber, N., and Hench, J. E., "Steady State and Transient Void Fraction
of Bubbling Systems and Their Operating Limits. Part I, Steady State
pumps, the condensation and compressibility effects will have Operation," General Electric Report No. 62GL100, 1962.
to be properly incorporated. 17 Patal, B. R., and Runstadler, P. W., Jr., "Investigation into the Two-
Phase Flow Behavior of Centrifugal Pumps," ASME Symposium on Polyphase
Acknowledgment Flow in Turbomachinery, Dec. 10-15, 1978, San Francisco.
The author had many useful discussions with Prof. A. J. 18 Murakami, M., and Minemura, K., "Effects of Entrained Air on the
Performance of a Centrifugal Pump," Bulletin of the Japanese Society of
Acosta of Caltech and this is gratefully acknowledged. He Mechanical Engineers, Vol. 17, No. 110, Aug. 1974.
also thanks Drs. J. H. Kim and Bill K. H. Sun of EPRI, for 19 Manzano-Ruiz, J. J., and Wilson, D. G., "Experimental Study of Two-
their various support as well as encouragement during the Phase, Two-Component Flow in Centrifugal Pumps," 2nd International
course of the project. The author thanks Mr. Shin Maekawa Topical Meeting on Nuclear Reactor Thermal-Hydraulics, Jan. 1983.

Journal of Fluids Engineering MARCH 1985, Vol. 107/147

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use