Conguration dependence and optimization of the entrainment performance

for gasegas and gaseliquid ejectors

Cui Li, Yanzhong Li
*
, Lei Wang
School of Energy and Power Engineering, Xian Jiaotong University, Xian 710049, China
a r t i c l e i n f o
Article history:
Received 9 March 2011
Accepted 19 November 2011
Available online 26 November 2011
Keywords:
Ejector
Conguration dependence
Optimization
CFD simulation
Pseudo-shock length
a b s t r a c t
This paper describes a numerical study of entrainment behavior and its conguration dependence for
gasegas and gaseliquid ejectors. A computational uid dynamics (CFD) model is developed and
experimental validation is undertaken over a wide range of operation conditions for ejector with
different congurations. The predicted values by CFD simulation prove to be in good agreement with the
experimental data. The investigation results indicate that pseudo-shock length has a dominant effect on
entrainment performance and geometry optimization. Signicant difference is noted in pseudo-shock
length for gasegas and gaseliquid ejectors, and this is mainly because the viscosity similarity mark-
edly differs within the range of 0.01e1.0, depending on the primary and secondary uids of usage.
Therefore the optimum mixing tube length to diameter ratio is about 1e2 for general gaseliquid ejectors
while 5e7 for gasegas ejectors. As an exception to the general gaseliquid ejectors, the optimum L/D
ratio in HeeLH
2
ejector is about 4, lying between that of the gasegas ejector and gaseliquid ejector but
still consistent with the pseudo-shock length. If the maximum allowable length of ejector mixing tube is
less than the optimum value, placing the primary nozzle exit upstream of the mixing tube can greatly
improve the entrainment performance.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
An ejector is a device in which a uid at high pressure is
accelerated by a primary nozzle and then the high speed primary
jet entrains and compresses a secondary uid at low pressure. The
objectives for ejector design could be (a) To get large entrainment of
the secondary uid, (b) To produce intense mixing between the
primary and secondary uids, or (c) To pump uids froma region of
low pressure to a region of high pressure, depending on its area of
application. Ejectors can be operated with compressible or
incompressible uids, and have found many applications in engi-
neering, such as refrigeration, aerospace, chemical and biochemical
process industries [1]. Their application in air-conditioning or
refrigeration, either totally replace mechanical compressors or
simply be used for cycle optimization (for example, the combined
ejector-absorption refrigeration system), has a long-established
history and as an environment-friendly technology, it has become
a recent focus of research in response to the problem of global
warming or ozone depletion [2e12]. In aerospace engineering,
ejectors are used for thrust augmentation, exhaust noise
suppression, altitude testing [13e15] or to mix exhaust gases with
fresh air in order to reduce the thermal signature [16]. The incentive
for their application in natural precooling system of cryogenic
rocket engine is to intensify the precooling circulation and improve
the cooling effect of rocket engine [17]. In chemical and biochemical
process industries, ejectors are widely used for entraining and
pumping corrosive uids, slurries, fumes and dust laden gases, etc.,
which are otherwise difcult to handle [18], say, recovering natural
gas from gas/oil mixture in oil storage tank. Ejectors may also be
used for mass transfer, namely, liquideliquid extraction, gas
absorption, stripping, fermentation, hydrogenation, chlorination,
etc [19]. One of the most promising applications is the biotechno-
logical treatment of industrial gaseous wastes which concerns
many industries: automobile industry, petrochemical, printing, ne
and heavy chemicals industry, etc [20]. Advanced reactor systems
utilizing steam-driven ejector as an emergency core cooling system
have also been proposed [21]. When compared with others, an
ejector systemhas the advantages of high reliability, lowcost and at
the same time, being easy to maintain and operate, therefore has
attracted many research activities.
Theoretical study of ejector can be classied into two categories:
constant-area mixing methods and constant-pressure mixing
methods, both of which were rst proposed by Keenan et al.
[22,23]. The latter are believed to give superior performance and
* Corresponding author. Tel.: 86 29 82668725; fax: 86 29 82668789.
E-mail address: yzli-epe@mail.xjtu.edu.cn (Y. Li).
Contents lists available at SciVerse ScienceDirect
Applied Thermal Engineering
j ournal homepage: www. el sevi er. com/ l ocat e/ apt hermeng
1359-4311/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2011.11.041
Applied Thermal Engineering 48 (2012) 237e248
are therefore more widely used. In fact, as a result of this nding,
progress on research into constant-area mixing was nearly halted
at one time. The constant-pressure mixing model reported by
Keenan et al. [23] is considered as the rst comprehensive theo-
retical analysis of ejector design. It was based on ideal gas dynamics
with the assumption of isentropic ow in the primary nozzle and
diffuser. Effects of molecular weight, specic heat ratio and friction
were ignored, so were the nozzle and diffuser efciencies. This
model has gained widespread acceptance ever since and further
improvements have been made by various researchers [6e8,24,25].
Munday and Bagster [8] put forward a model to account for the
choking phenomena which was neglected by Keenan et al. [23]. A
constant hypothetical throat area, or effective area, was dened
for the entrained ow, and mixing was assumed to occur beyond
the hypothetical throat with a uniform pressure. Huang et al. [7]
computed the effective area and found that it does not remain
constant but varies with the operation conditions. By applying the
same theory with ammonia, Rogdakis et al. [25] included the
possibility of mixing and supersonic compression taking place in
the mixing chamber. Although these methods are under the
assumption that the velocities of both the primary ow and
secondary ow are uniform at every cross section of the ejector,
they are helpful when analyzing the entrainment efciency of
a xed geometry ejector. However, this kind of studies are still
unable to correctly reproduce the ow physics locally along the
ejector, nor could they explain the effects of most geometric
parameters on the ejector performance, for example, the mixing
tube length and primary nozzle position, owing to their one-
dimensional simplications. Over the years considerable effort
has been devoted to the two-dimensional aspects of ejector
performance. Mikhail [26] assumed different velocity proles at
each stage of the mixing process in a constant-area tube to solve the
linearized integral momentum equation. And a 2D exponential
expression is adopted for velocity distribution by Zhu [27] to
approximate the viscosity ow near the ejector inner wall. Hick-
man et al. [28] included in their study the effects of temperature
difference between the primary and secondary ows. This kind of
studies involved solving a number of integral equations, and
exhibited a large reliance on experimentally determined constants.
In 1974, Hedges and Hill [29] developed a nite-difference scheme
to model the ow process, which was the rst venture into what is
recognized today as computational uid dynamics (CFD). In
contrast to integral methods, the nite-difference model places no
restrictions on the initial conditions and does not need to specify
the proles of velocity and temperature. Gilbert and Hill [30]
furthered this method and concluded that a nite-difference
model was both accurate and more exible than the previous
integral methods. Nilavalagan et al. [31] presented a similar nite-
difference model and claimed that their method allowed the
optimum mixing tube length to be predicted. In recent years the
commercial CFD software programs have received increasing
attention in that they are capable of dealing with the complicated
supersonic owing and mixing problem and at the same time,
revealing the internal local phenomena clearly. As compared to
experimental measurements and observation, the CFD technique
turns out to be a more efcient diagnosis tool for ejector analysis
which gives a better insight into how ejectors function yet at
a reasonable cost [4].
Recently, a great deal of research has been made on optimal
ejector geometry in various application elds, mainly by means of
experimental [3,12,32e40] and numerical (CFD) [9,10,34,41e47]
methods. As is seen from Table 1, the mixing tube length has
been paid close attention due to its signicant effect on the mixing
of the primary and secondary ows. In the experimental study of
Dirix and Wiele [37], the mixing tube length seems to have no
obvious effect: the length to diameter ratio L/D extended from 2 to
10 results in a slight decrease in the individual efciency of ejector
and an equal performance of the reactor system. But research of
Cramers and Beenackers [38] clearly illustrates that ejector with
longer mixing tube creates higher volumetric mass transfer coef-
cients compared to that with a shorter one. Bando et al. [39] have
reported increasing entrainment rate with an increase in mixing
tube length, when the L/D ratio is in the range of 20e30, the
entrainment rate reaches the maximum value, and it then
decreases with any further increase in L/D. However, the experi-
ment conducted by Havelka et al. [40] shows that as L/D ratio
increases the suction rate increases rst and then reaches a plateau
when L/D is greater than 6. Likewise, Keenan and Neumann [22]
reported that an optimum mixing tube length is about 7 times
the mixing tube diameter. On the other hand, the CFD studies of
Kandakure et al. [41] and Balamurugan et al. [42] indicate that the
entrainment rate is the highest when the ejector L/D ratio is equal
to zero, i.e., no constant-area mixing tube at all. Fromthese studies,
it is clear that the optimumlength varies greatly with the operation
conditions and it is difcult to nd a universal value that meets all
the conditions.
Similar problems are encountered in research referring to
optimization of other geometry parameters, say, the primary nozzle
Table 1
Optimum ejector geometry from references.
Reference Ejector typePrimaryeSecondary Geometry parameter Results
Dirix[41] LiquideGas L/D 2e10 Extension of L/D results in a slight decrease in ejector efciency.
Cramers[42] LiquideGas L/D 2, 10 Mass transfer rate increases when longer mixing tubes are used.
Bando[43] LiquideGas L/D 550 L/D of 20e30 gives higher suction ow rate.
Havelka [44] LiquideGas L/D 010 The optimum ratio is 0 for ejector with swirls and 6 for ejector
without swirls.
Keenan [26] GaseGas L/D The optimum mixing tube length is about 7 times the mixing tube diameter
Kandakure[45] LiquideGas L/D 0, 4, 8 Highest entrainment rate is obtained for L/D 0.
Balamurugan[46] GaseLiquid L/D 0, 3.9, 7.8 Ejector with L/D 0 has the maximum entrainment rate.
Li[17] GaseGas/Liquid L/D 0e10 The optimum L/D ratio is 5e7 for gasegas ejectors and 1e2
for gaseliquid ejectors.
Watanabe [52] GaseGas NXP The optimum position is within the mixing section.
Aphornratana [3] GaseGas NXP The optimum (NXP)
r
is in the range of 0e0.83.
ESDU [53] GaseGas NXP The optimum (NXP)
r
varies between 0.5 and 1.0.
Zhu[51] Gase Gas NXP The optimum (NXP)
r
ranges from 1.7 to 3.4.
Acharjee[57] LiquideGas NXP The optimum (NXP)
r
is 2
Biswas [58] LiquideGas NXP The optimum (NXP)
r
is 2
Henzler [59] GaseGas L/D The optimum L/D ratio is 7.5 for 4 < 2.5.
NXP The optimum (NXP)
r
is 0.4e0.9.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 238
exit position NXP (see Table 1). Watanabe [48] has made observa-
tions regarding the effect of primary nozzle position on ejector
performance. His experiments showthat the primary nozzle has an
optimum position within the mixing section, and this maximizes
the ejector efciency. Aphornratana and Eames [3] have conducted
experimental studies on a small-scale steam ejector refrigerator
using ejector with adjustable primary nozzle position. The
optimum NXP is found in the range of 0e15 mmwithin the mixing
chamber inlet section. These conclusions are in contrast to the
recommendation from ESDU [49] which suggested placing the
nozzle exit 0.5e1.0 length of throat diameter upstream of the start
of the mixing chamber. Likewise, Croft and Lilley [50] have
discovered that when the nozzle moves closer to the mixing tube,
the entrainment ratio decreases, and the experimental studies of
Keenan et al. [11,22,23], Sun [11], Chunnanond and Aphornratana
[33], Chaiwongsa and Wongwises [51], as well as the CFD study
reported in [4,9,10,44,45,52], have also indicated that in ejector
refrigeration system, moving the nozzle into the mixing chamber
causes the cooling capacity and COP to decrease. Moreover, the
optimum NXP of steam ejector described in Ref. [47] is 1.7e3.4
length of throat diameter upstream of the start of the mixing
chamber. And studies by Acharjee et al. [53] and Biswas et al. [54]
show that the optimum NXP normalized by the throat diameter
is about 2.0 for liquid jet ejectors. However, according to the
experimental results for refrigerant R123 given in Ref. [12], the
optimum relative primary nozzle position is only about 0.55, rela-
tively small when compared with the above studies. In addition,
Henzler [55] has ever summarized the experimental data of
different investigators, and the optimum relative position sug-
gested by him is in the range of 0.4e0.9. Hence it should be said
that the selection of primary nozzle position is not fully understood
so far and it still awaits further work.
In the present work, conguration dependence of the entrain-
ment performance for gasegas and gaseliquid ejectors are studied.
A compressible CFD model has been formulated and calibrated by
actual experimental data, and the effects of ejector geometry as
well as uid property on entrainment performance are investigated
in detail. Distinct characteristics of entrainment performance
between gasegas and gaseliquid ejectors are demonstrated and
analyzed, which is of signicance to ejector design and operation.
2. CFD modeling
The commercial package, Gambit 2.3 and Fluent 6.3, are used as
the grid generator and the CFD solver, respectively. The initial mesh
is composed of about 30,000e50,000 quadrilateral elements
depending on the length of ejector, and the cells close to primary
nozzle and walls are small enough to capture the complex ow
phenomenon. In order to better predict the internal ow status,
solution-adaptive mesh renement is employed and more cells are
added at locations with signicant ow changes, for example, the
interaction region between the supersonic stream and the low-
speed secondary stream, and the region with adverse pressure
gradient caused by shocks, etc. The resulting mesh thus enables the
features of the ow eld to be better resolved. Since the axisym-
metric solver is adopted, the three-dimensional effect has been
taken into account by this two-dimensional computational domain.
2.1. Algorithm
The CFD simulation is performed on the basis of steady-state
compressible turbulent ows and ve pairs of working uids, as
shown in Table 2, are used for analysis of ejector entrainment
performance. Solution of primary gas in the liquid phase is
neglected in the present study on account of its poor solubility. For
multiphase ows in gaseliquid ejectors, mixing process is accom-
plished via viscous diffusion and small-scale turbulence transport.
The mixture model is used to model the interphase interactionwith
the assumption that local equilibrium is established over short
spatial length scales. Manninen assumption is applied to compute
the slip velocity and the determination of drag function is achieved
by using symmetric model. In the cases of Pair HeeLO
2
and
HeeLH
2
, the entrained liquid oxygen/hydrogen is at subcooled
condition and no evaporation occurs before encountering the
primary streamwhen adiabatic wall condition is applied. If the inlet
subcooling DTsub decreases, for example to 5.8 K for HeeLO
2
, gas
oxygen is observed in the interaction region of helium and liquid
oxygen, and the entrained liquid oxygen decreases slightly because
gas oxygen will occupy larger region. It is worth mentioning here
that the inlet subcooling in the present study is relatively large
when compared with the onset one of phase change, and thus it is
rarely possible for evaporation to take place.
The governing equations, linearized in an implicit manner, are
solved in a pressure based solver. SIMPLEC (SIMPLE-Consistent)
algorithm is employed for the pressure-velocity coupling. And the
second order upwind discretization scheme is used for the
convection terms of each governing equation except that the
volume fraction is solved using QUICK scheme. The steady-state
continuity equation for the mixture can be written as:
v
_
r
m
u
m;i
_
vx
i
0 (1)
And the momentum equation:
v
vx
i
_
r
m
u
m;i
u
m;j
_

vP
vx
j

v
vx
i
_
m
m
_
vu
m;i
vx
k

vu
m;k
vx
i
__
F
j

n
k 1
v
vx
i
_
a
k
r
k
u
dr;k;i
u
dr;k;j
_
(2)
where F is a body force, m
m
is the viscosity of the mixture, and u
dr
,
the drift velocity for secondary phase.
The energy equation for the mixture takes the following form:

n
k 1
r
k
E
k
p
v
vx
i
_
a
k
u
k;i
_

v
vx
i
_
k
eff
vT
vx
i
_
(3)
Here, the effective thermal conductivity, k
eff
, is computed from:
k
eff
k
l

c
p
m
t
Pr
t
Where, k
l
is the thermal conductivity, c
p
is the specic heat, and Pr
t
,
the turbulent Prandtl number. In this case, Pr
t
0.85. The turbulent
viscosity, m
t
, is computed by combining turbulence kinetic energy k
and its rate of dissipation 3 , as follows:
m
t
r
m
C
m
k
2
3
And the volume fraction equation for the secondary phase
(phase p):
Table 2
Working uid pairs used for experimental and CFD studies.
Working pair Primary uid Secondary uid Characteristic
N
2
eN
2
Nitrogen Nitrogen Single-phase
HeeHe Helium Helium Single-phase
N
2
eH
2
O Nitrogen Water Two-phase
HeeLO
2
Helium Subcooled liquid oxygen Two-phase
HeeLH
2
Helium Subcooled liquid hydrogen Two-phase
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 239
v
vx
i
_
a
p
r
p
u
m;i
_

v
vx
i
_
a
p
r
p
u
dr;p;i
_

n
q 1
_
_ m
qp
_ m
pq
_
(4)
where _ m
qp
denotes the mass transfer from phase q to phase p, and
_ m
pq
, from p to q. If the evaporation or condensation process occurs,
the mass transfer is modeled with the mechanistic model (reported
in Ref. [56]) by using user dened functions.
The robust, economical, and reasonably accurate Standard ke 3
model is employed to model the turbulent behavior of ow in
ejector. The main reason for this is that in our previous work
(Ref. [17]) the Standard ke 3 model predicts the entrainment
performance satisfactorily for gaseliquid ejectors while the
entrained mass ow rates calculated by the SST kes model and the
Standard kes model show obvious deviation from the experi-
mental results. Besides, the CFD modeling of Hemidi et al. [57]also
concludes that for a supersonic air ejector the ke 3 model provided
best results while the results for the SST kes model are in a less
agreement. The turbulence kinetic energy k, and its rate of dissi-
pation 3 , are obtained from the following transport equations:
v
vx
i
_
r
m
ku
m;i
_

v
vx
i
__
r
m
C
m
k
2
3
_
vk
vx
i
_
r
m
C
m
k
2
3
_
vu
m;i
vx
l

vu
m;l
vx
i
_

vu
m;i
vx
l
r 3 2r 3 M
2
t
(5)
v
vx
i
_
r
m
3 u
m;i
_

v
vx
i
__
r
m
C
m
k
2
3
_
v 3
vx
i
_
C
1 3
C
m
r
m
k
_
vu
m;i
vx
l

vu
m;l
vx
i
_

vu
m;i
vx
l
C
2 3
r
3
2
k
(6)
The model constants C
1 3
, C
2 3
and C
m
in Equations (5) and (6) have
the following default values:
C
1 3
1:44; C
2 3
1:92; C
m
0:09
And M
t
is the turbulent Mach number that accounts for
compressibility effect in the ke 3 model.
In addition, the near-wall region is modeled by the standard
wall function approach, which gives reasonably accurate results for
the high-Reynolds-number, wall-bounded ows.
2.2. Boundary condition
With respect to the compressibility, the primary owis assumed
to obey the ideal gas law. And the properties of liquid phase such as
viscosity, specic heat and thermal conductivity are derived from
the uid thermo-physical properties provided by NIST (2010).
The pressure inlet boundary condition is applied to the primary
owinlet as well as that of the secondary ow, and the outlet of the
ejector is assigned to be a pressure outlet boundary condition.
Adiabatic thermal condition is applied for the ejector wall by
setting a zero heat ux at the wall surface.
The solution is iterated until convergence is achieved, that is,
residual for each equation falls below 10
3
(10
8
for single-phase
operation) and changes in mass ow rates for both primary ow
and secondary ow become negligible. In general, it is observed
that the residual for continuity equation is below 10
6
while for
other equations the residuals are well below 10
8
.
2.3. Experimental validation
The validation experiment is carried out with nitrogen as the
motive uid and water as the entrained uid, and a view of the
ejector installed in the experimental rig has been presented in
Fig. 1. As shown, the ejector under investigation consists of four
parts: the convergingediverging primary nozzle, the suction
chamber, the constant-area mixing chamber (namely, the mixing
tube) with conical inlet section, and the diffuser. The primary
nozzle used in the present study has an inlet diameter (d
1
) of 6 mm,
throat diameter (d
2
) of 2.85 mmand exit diameter (d
3
) of 3.2 mm. It
is placed upstream of mixing tube which has a length (L) of
193 mm, and the distance between the nozzle exit plane and the
start of constant-area mixing chamber is 8 mm. A schematic
diagram of the experimental setup is shown in Fig. 2. The main
elements of this test facility are a gaseliquid separator, an ejector,
a centrifugal pump, and a group of nitrogen cylinders. Compressed
nitrogen from cylinders is accelerated when owing through the
convergingediverging primary nozzle and undergoes successive
changes from subsonic to supersonic. The supersonic jet ow then
gets mixed with the low-speed water stream, which eventually
resulting in the suction of water via entrainment. Consequently,
a mixture of nitrogen and water enters the gaseliquid separator
and the separated nitrogen is directly discharged into the atmo-
sphere while the water is used repeatedly.
The secondary mass ow rate and the entrainment ratio
(dened as u m
S
/m
P
) by experiment and by CFD calculation are
shown in Fig. 3. Nitrogen and water is utilized as the primary and
secondary uids respectively for ejectors with different area ratios,
i.e. 60, 90 and 126. The primary owpressure varies approximately
Fig. 1. Wall contours and principle dimensions of supersonic ejector.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 240
between 1 MPa and 3 MPa, and the compression ratio x(the ratio of
exhaust pressure to secondary pressure, expressed as P
C
/P
S
ranges
from 1 to 1.25. A detailed description of the operation condition is
available in Ref. [17]. As seen from Fig. 3, the discrepancy between
the measured and calculated data falls in a reasonable range, and
the relative errors are less than 25% in almost all cases. It thus can
be concluded that, the CFD model proposed in the present study
predicts the entrainment performance satisfactorily for gaseliquid
ejectors without phase change taking place.
3. Investigation of ejector performance: conguration
dependence and optimization
3.1. Performance analysis
The ejector performance is mainly inuenced by three factors,
i.e., the operation condition, the ejector conguration and the
working uids used in the ejector system. The entrainment
behavior under the inuence of operation condition was previously
investigated [17]. In the present study, ve pairs of working uids,
as shown in Table 2, are utilized to analyze the conguration
dependence of ejector entrainment performance. The distinct
characteristics of entrainment performance between gasegas and
gaseliquid ejectors are demonstrated and analyzed, aiming at
providing useful information for ejector optimization design. For
a viewof the owelds in gasegas or gaseliquid ejectors, see Fig. 4.
The entrainment ratio for single-phase N
2
eN
2
ejector with
different area ratios is plotted against the mixing tube L/D ratio in
Fig. 5(a), and HeeHe ejector in Fig. 5(b). It is found that for xed
operation condition (P
P
1 MPa, P
S
0.25 MPa), the optimum
mixing tube L/D ratio is about 6 for both N
2
eN
2
and HeeHe ejec-
tors. The entrainment performance can be greatly improved when
the mixing tube length increases from 0 to the optimum value,
namely 6 times the mixing tube length; little difference in
Fig. 2. Schematic diagram of experimental setup.
Fig. 3. Comparison of experimental data and CFD results for different operation conditions.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 241
entrainment ratio is observed when mixing tube length is further
increased. Similar to this point of view, Havelka et al. [40] have
reported that the gas suction rate increased with the increasing
length of the mixing tube and the increase was almost negligible for
L/D greater than 6. And the experimental investigation of Elkady
et al. [58] also indicates that a turning point occurs when L/D
increases to 7.5, which means after this point the rapid growth of
entrainment ceases to exist and the entrainment ratio reaches
a constant value. Since great similarity is found between N
2
eN
2
and
HeeHe ejectors, the results mentioned above can be extended to
ejectors that use other gases as working uids and be served as
a general guideline for gasegas ejector design.
Conguration dependence of entrainment performance for
gaseliquid ejectors is studied under constant operation condition
to make sure the maximum efciency is achieved by the so-called
optimum mixing tube geometry. The variation of entrainment ratio
with mixing tube L/D ratio for gaseliquid N
2
eH
2
O ejector is shown
in Fig. 5(d) and the HeeLO
2
ejector in Fig. 5(e). From these two
gures, it is clear that for N
2
eH
2
O and HeeLO
2
ejectors the mixing
tube length is a very important design parameter and can exert
a remarkable inuence on ejector performance. The optimum
mixing tube length for N
2
eH
2
O ejector is about 1e2 times the
mixing tube diameter. Evidently, the ejector performance is sharply
increased when the mixing tube length increases from 0 to the
optimum value, that is, 1e2 times the mixing tube diameter and
further increase in mixing tube length causes a gradual but obvious
decrease in entrainment performance, which is quite different from
that of the gasegas single-phase ejector. This behavior agrees well
with the experimental results published by Bando et al. [39]. And
the experiments performed by Dirix et al. [37] on liquid jet two-
phase ejector in loop reactors also comes to the conclusion that
extension of L/D ratio from 2 to 10 results in a slight decrease in
ejector efciency. Still, as is evident from Fig. 5(d) w (e), the
performance curves of HeeLO
2
ejector are quite similar to those of
the N
2
eH
2
O ejector and an analysis of the similarity characteristic
makes it possible to conclude that: the optimumlength to diameter
ratio (L/D) of gaseliquid ejector is in between 1 and 2, and deviation
from the optimum value (actually a small range) can dramatically
degrade its entrainment performance. This may be explained by the
fact that in excessively long mixing tube the pressure decrease due
to frictional ow is greater than the increase due to mixing process
and as a result excessive loss is caused under such circumstance. If,
on the other hand, the mixing tube is too short, momentum
transfer does not intensively occur or the primary and secondary
ow get inadequately mixed, for such cases the pressure recovery
region becomes shorter and thus the ejector efciency decreases
dramatically.
An exception to the general gaseliquid ejectors is ejector using
Pair HeeLH
2
as working uids. Although the primary uid and
secondary uid in this kind of ejectors are of different physical
phase, the u curve in terms of (L/D) possesses a shape similar to that
of gasegas ejector, and the optimum L/D ratio, about 4 as indicated
Fig. 4. Characteristics of oweld for gasegas and gaseliquid ejectors: (a) velocity distribution of gasegas (N
2
eN
2
) ejector; (b) velocity distribution of gaseliquid (N
2
eH
2
O) ejector;
(c) volume fraction of primary uid in gaseliquid (N
2
eH
2
O) ejector.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 242
in Fig. 5(c), is much closer to that of the single-phase gasegas
ejector rather than general gaseliquid ejector.
In summary, optimum mixing tube length gives rise to the
occurrence of maximum entrainment ratio and is found to depend
on the working uids to a great degree. Mixing tube that is too long
or too short can either cause additional viscous losses or be
disadvantageous to pressure recovery, which nally degrades the
entrainment efciency of ejector with gaseliquid-type working
uids (except for HeeLH
2
). But in the case of gasegas entrainment,
no obvious decay of entrainment ratio is observed when the ejector
mixing tube is longer than its optimum value.
3.2. Optimization
In order to get a more comprehensive understanding of the
conguration dependence and to optimize the ejector performance,
the effects of primary nozzle position on the entrainment ratio are
investigated for both gasegas and gaseliquid ejectors. The nozzle
position (NXP) is dened as the distance between the primary
nozzle exit plane and the inlet plane of constant-area mixing
chamber, i.e. the mixing tube (see Fig. 1). It has a positive value when
the primary nozzle is placed forward into the mixing tube, and is
negative when located backward from this constant-area mixing
Fig. 5. Effect of mixing tube length on ejector performance in different entrainment modes (b 4; x 1; 4 30, 60 and 90, respectively): (a) N
2
eN
2
; (b) HeeHe; (c) HeeLH
2
; (d)
N
2
eH
2
O; (b) HeeLO
2
.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 243
section. The nozzle position corresponding to the maximum
entrainment ratio is assumed to be the optimum nozzle position.
Importance of the primary nozzle position in maintaining high
entrainment performance is numerically investigated (see Fig. 6)
with emphasis on the off-design ejectors, which in the present
study can be classied into two categories. The rst category
associates with mixing tube that is too long compared to the
optimum value, say, L/D 3 for N
2
eH
2
O and HeeLO
2
ejector. For
such cases, the entrainment ratio increases as the primary nozzle
approaches the inlet of mixing tube and reaches the highest level
when the two planes, i.e. the primary nozzle exit and the mixing
tube inlet, coincide with each other. Further movement of the
primary nozzle into the mixing tube results in a slight rise of the
entrainment ratio. This suggests that increasing the mixing tube
length from its optimum value leads to the moving of optimum
nozzle position into the mixing tube. It should be pointed out
however, that under these circumstances the effect of NXP is not so
prominent because the mixing tube is long enough to obtain the
maximumattainable pressure recovery by itself. On the other hand,
for ejectors with relatively small L/D ratios, the primary nozzle
should be placed at a certain distance upstream of the mixing tube
inlet. For example, in HeeLO
2
ejectors the optimum relative posi-
tion of primary nozzle is about 0.36 (NXP 8 mm) for the case of
L/D 1. Similar results are found for gasegas ejectors. When L/D
ratio decreases to 1, the primary nozzle in ejector using N
2
or He as
working uids even needs to be moved to a place with (NXP)
r
1 in
order to maximize the entrainment performance. Still, however,
when L/D becomes equal or slightly greater than the optimum
value, say 5e7, effects of NXP is not so prominent. For these cases,
the optimum relative primary nozzle position is in the range of
Fig. 6. Effect of primary nozzle position (NXP) on ejector performance(b 4; x 1; 4 60; L/D 1, 3, 5, 7, respectively): (a) N
2
eN
2
; (b) HeeHe; (c) HeeLH
2
; (d) N
2
eH
2
O; (b)
HeeLO
2
.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 244
0e0.36 (equivalent to an NXP of 0 to 8 mm). From the results
mentioned above as well as the experimental and computational
investigations [3,47e50,53e55] that have been reported so far, it
can be concluded that varying the position of primary nozzle could
give marked improvements in performance at off-design condi-
tions, but it is impossible to dene a single optimum nozzle posi-
tion to meet all conditions. Physically this means the primary
nozzle position should be selected properly according to the actual
situation, mainly including mixing tube geometry and the working
uids being adopted.
4. Analysis and discussion
The entrainment performance of ejector has proven to be
strongly affected by the mixing tube L/D ratio, but the optimum L/D
ratio widely differs within the range of 1e7, depending on the
working uids utilized. It is therefore necessary to investigate the
complicated ow status in the supersonic gasegas or gaseliquid
ejector and to make clear the principle of optimum L/D ratio
determination.
4.1. Flow structure in ejector
Pressure along the ejector axis is schematically illustrated in
Fig. 7. As shown, the pressure at the centerline falls and rises
repeatedly due to the presence of successive shocks. Such a series of
shocks, including incident and reected ones, is referred to as shock
train. It is accompanied by abrupt change not only inpressure but in
density and velocity. After these successive changes the primary jet
ow is still supersonic or mixed supersonicesubsonic while the
secondary ow outside the jet boundary layer remains low-speed
(as shown in Fig. 4). The resulting radial nonuniformity in the
velocity distribution leads to intensive momentum exchange.
Actually this process is driven by the combined action of turbulent
eddies and interphase drag. That is, in the primary jet, dispersed
droplets get caught up in the turbulent eddies of continuous gas
phase, and are transported by the effect of interphase drag. Like-
wise, outside the jet boundary layer bubbles that are dispersed in
the liquid phase move from high to low concentration regions. In
this way interphase exchange intensively occurs and pressure rise
caused by this mixing process is initially greater than the decrease
due to friction effect. Consequently, after the shocks cease to exist
in the core ow the centerline pressure continues increasing until
a local maximum pressure is achieved at some point of the mixing
tube. This point, followed by a gradual decrease in pressure, is the
end of the pseudo-shock (the term used to indicate the whole
interaction region from the primary nozzle exit with the shock
fronts of the shock train included in it) [59]. Unlike the uctuation
in the jet core, the pressure at the wall varies mainly with the cross
section area of the ow passage. Behind the pseudo-shock,
however, the friction effect plays a dominant role in axial pres-
sure, and the pressure at the centerline is nearly the same as that at
the wall, that is, the pressure has a uniform radial distribution till
changes of section occur, for example transition from the mixing
tube to diffuser.
The behavior of shock train and pseudo-shock, such as structure
and location, is affected by the uid status, for example, primary
ow alone or with the existence of the secondary ow as well as
their individual phases. Therefore a plot of axial pressure for the
above-mentioned entrainment modes (HeeHe, HeeLH
2
and
HeeLO
2
, respectively) is shown in Fig. 8 to illustrate the similarities
and differences of pseudo-shock between gaseliquid and single-
phase entrainment. The curves demonstrate that, although the
shock train and pseudo-shock markedly differ in their uctuation
amplitude and size of occupied region, they exhibit a striking
similarity in the oscillation manner. Behind the pseudo-shock, the
friction effect plays a dominant role in HeeLO
2
entrainment mode,
and thus a gradual decrease in pressure can be observed for this
case till changes of section area occur. When helium is employed as
Fig. 7. Schematic illustrations of static pressure distribution along ejector centerline
for (a) Gasegas ejector and (b) gaseliquid ejector.
Fig. 8. Plot of axial pressure distributions for different entrainment modes, i.e., HeeHe,
HeeLH
2
and HeeLO
2
, respectively (b 4; x 1; L/D 8).
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 245
the secondary uid, i.e., HeeHe mode, the friction effect can be
negligible and the pseudo-shock, although in a similar shape with
HeeLO
2
mode, occupies a relatively longer region. Still, the pseudo-
shock behavior for HeeLH
2
entrainment mode (represented in
Fig. 8 by the red dash-dot curve) lies between those for HeeLO
2
and
HeeHe ejector, in both uctuation manner and occupied region
size. Conclusions from pressure distribution of N
2
eH
2
O and N
2
eN
2
ejector are consistent with those from HeeLO
2
and HeeHe ejector,
respectively, and have further elucidated the discrepancy between
general gaseliquid entrainment mode and gasegas mode.
4.2. Investigation of pseudo-shock: characteristic parameter and its
major determinant
The interphase exchange occurring in ejector is caused by the
combined action of turbulent eddies and interphase drag, and
depends greatly on the primary and secondary uid properties
especially the viscosity. For the purpose of clear and comparative
demonstration, the viscosity of uids involved in the present study,
derived fromthe uid thermo-physical properties provided by NIST
[27], are listed in Table 3. And the Tanimoto coefcient, also known
as the extended Jaccard coefcient, is introduced to determine if
the primary and secondary uids are similar to each other or not. It
is represented as
Tm
P
; m
S
m
P
$m
S

__
m
2
P
m
2
S
m
P
$m
S
_
and returns a score indicating the level of similarity between the
two uids under comparison. As shown, the viscosity of primary
uid differs greatly from that of the secondary uid in gaseliquid
ejectors (with an exception of HeeLH
2
ejector). For instance, the
Tanimoto coefcient is 0.0155 for N
2
eH
2
O ejectors, while in single-
phase ejector the viscosity ratio is approximate to 1. Therefore,
different characteristics and behaviors of the pseudo-shock, which
is strongly affected by the interaction and mixing process between
the primary and secondary ows, are observed in single-phase and
two-phase ejectors.
In gaseliquid ejectors, the signicant difference in uid physical
properties leads to rapid shock attenuation and in turn, a shorter
pseudo-shock length which is dened as the length necessary to
fully achieve the pressure recovery. This is clearly demonstrated in
Fig. 9. As shown, the pseudo-shock length (L
PS
) normalized by the
mixing tube diameter is in the range of 1e2 for either N
2
eH
2
O or
HeeLO
2
ejector, while for single-phase ejector, the non-
dimensional pseudo-shock length is 6. An exception to the
general gaseliquid ejectors is that using Pair HeeLH
2
as working
uids. In this kind of ejectors, minor difference in viscosity between
the primary and secondary uids is observed, which is much closer
to that of gasegas ejector than to gaseliquid ejector. The viscosity
similarity described by Tanimoto coefcient for HeeLH
2
is about
0.898, while less than 0.1 for general gaseliquid ejector and
approximate to 1 for gasegas ejector. The pseudo-shock length
normalized by the mixing tube diameter for HeeLH
2
ejector is
about 4, which lies between that of the gasegas ejector and
gaseliquid ejector. It can be concluded that the pseudo-shock
length is closely related to the viscosity similarity between
primary ow and secondary ow, and greater viscosity dissimi-
larity leads to smaller non-dimensional pseudo-shock length.
Moreover, the pseudo-shock length in gaseliquid ejector decreases
slightly when the pressure ratio increases and even can be assumed
to be constant for N
2
eH
2
O or HeeLO
2
ejector, but in gasegas
ejector it increases obviously with the increasing pressure ratio
b and varies from 5 to 7 times the mixing tube diameter with
b ranging from 2 to 12 (see Fig. 9(b)). The pseudo-shock length
appears to be a signicant parameter, and has a dominant effect on
the entrainment performance and efciency of ejector as well as
the geometry optimization.
4.3. Pseudo-shock length and determination of optimum L/D ratio
The optimum mixing tube length shows a close relation to the
pseudo-shock length L
PS
. Dependence of entrainment ratio on
mixing tube length and the optimum value of (L/D) ratio have
reected the dominating effect of the pseudo-shock on the global
Table 3
Similarity of uid viscosity for different entrainment modes.
N
2
eN
2
HeeHe N
2
eH
2
O HeeLO
2
HeeLH
2
m
P
(uPa s) 19.417 17.490 19.417 19.417
17.490
m
S
(uPa s) 19.387 1142.0 197.13 13.885
17.364
m
P
/m
S
1.001 0.0153 0.098 1.398
1.007
T (m
P
, m
S
) 0.999997 0.015549 0.108097 0.898061
0.999948
Fig. 9. Non-dimensional pseudo-shock length versus working uids for: (a) Ejector
with different area ratios, and (b) different working conditions.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 246
ejector performance, and the relation between the mixing tube
length and the optimum primary nozzle position has further
proved this conclusion. In fact, the discrepancy of optimum mixing
tube length in gasegas and gaseliquid ejectors is a consequence of
the discrepant pseudo-shock length, and the optimum value of
mixing tube length is equivalent to L
PS
. For example, the optimum
mixing tube length is 1e2 times the mixing tube diameter for
N
2
eH
2
O or HeeLO
2
ejector, which is well consistent with the
pseudo-shock length observed in these two kinds of ejectors. Good
accordance between optimum mixing tube length and pseudo-
shock length has also been found for gasegas and HeeLH
2
ejec-
tors. Therefore ejector, whether single-phase or not, should be
designed with a mixing tube length equal to or slightly greater than
L
PS
to obtain the maximum attainable pressure recovery and to
ensure high entrainment performance. In other words, the
optimum value of mixing tube length is actually a small range and
L
PS
is its lower limit. If the maximum allowable length of ejector is
less than the pseudo-shock length, retracting the primary nozzle
from the mixing tube inlet will help to ensure pressure recovery
and thus greatly improve the entrainment performance. Moreover,
under the inuence of operation condition (say, the pressure ratio
b) the variation of optimum mixing tube length (reported in [17])
appears similar to that of pseudo-shock length (depicted in
Fig. 9(b)), and this further proves the equivalent relation between
the optimum mixing tube length and the pseudo-shock length.
5. Conclusion
In this paper, computational uid dynamics (CFD) modeling
studies have been carried out to investigate the entrainment
behaviors of gasegas and gaseliquid ejectors. Experimental vali-
dation has been performed on different ejector congurations by
using N
2
eH
2
O as working uids. It is proved that the CFD model
used in the present study is efcient and can predict the entrain-
ment performance satisfactorily for gaseliquid ejectors. Effects of
mixing tube geometry and primary nozzle position are then
studied with the usage of different working uids, and over 300
cases are calculated to systematically analyze the conguration
dependence and optimization of entrainment performance. The
calculation results have conrmed the limitations of xed geom-
etry ejector and show that an appropriately designed ejector can
provide high entrainment performance.
Essentially it is the interior pseudo-shock that strongly affects
the ejector entrainment performance. Since the great difference in
Tanimoto coefcient (a measure of viscosity similarity between the
primary and secondary uids), the pseudo-shock length in gasegas
ejector differs signicantly from that in gaseliquid ejector, i.e. 5e7
and 1e2 times the mixing tube diameter respectively. As an
exception to the general gaseliquid ejectors, the pseudo-shock
length in HeeLH
2
ejector is about 4 when normalized by the
mixing tube diameter, lying between that of the gasegas ejector
and gaseliquid ejector.
The optimum mixing tube length should be selected to be equal
to or slightly greater than L
PS.
The deviation from the optimum
value, actually a small range, may dramatically degrade the
entrainment performance. If the lower limit requirement of the
optimum mixing tube length range cannot be satised due to
external restrictions, the primary nozzle position then plays an
important role in pressure recovery. Under this condition, negative
NXP is required to ensure high entrainment efciency. It should be
pointed out however, that effect of NXP is not so prominent when
the mixing tube is too long when compared with the optimum
length. Therefore the following recommendations is proposed for
an optimum ejector design: the mixing tube length is equal to or
slightly greater than the pseudo-shock length with moderate value
of NXP, say 0e8 mm upstream of the mixing tube inlet; if the
maximum allowable length is less than the optimum value,
retracting the primary nozzle backward from the mixing tube inlet
can greatly improve the entrainment performance; but if the
mixing tube is too long when compared with the optimum length,
moving the primary nozzle into the mixing tube is more appro-
priate. And evidently, the concept of exible construction of NXP
should be a proper choice for design, which can easily be realized
by ejector with a movable primary nozzle.
Acknowledgements
This work was supported by the Specialized Research Fund for
the Doctoral Program of Higher Education of China under Grant No
20100201110012. The assistance of China Academy of Launch
Vehicle Technology (CALT) is greatly appreciated.
Nomenclature
D mixing tube diameter, mm
d
1
primary nozzle inlet diameter, mm
d
2
primary nozzle throat diameter, mm
d
3
primary nozzle exit diameter, mm
E total energy
L mixing tube length, mm
NXP primary nozzle position, mm
(NXP)
r
relative primary nozzle position, (NXP)/D
m
P
mass ow rate of primary ow, kg/s
m
S
mass ow rate of secondary ow, kg/s
P Pressure, Pa
P
P
primary ow pressure, MPa
P
S
secondary ow pressure, MPa
P
C
exhaust pressure of ejector, MPa
T Temperature, K
u velocity vector
Greek symbols
r density, kg/m
3
m
P
viscosity of primary uid, mPa s
m
S
viscosity of secondary uid, mPa s
a volume fraction
g adiabatic exponent of gas
4 area ratio, (D/d
2
)
2
b pressure ratio, P
P
/P
S
x compression ratio, P
C
/P
S
u entrainment ratio, m
S
/m
P
Index
i summation index
j free index
l summation index
m index for mixture
k index for phase k
p index for phase p
q index for phase q
References
[1] D.W. Sun, I.W. Eames, Recent developments in the design theories and
applications of ejectors e a review, Journal of the Institute of Energy 68 (1995)
65e79.
[2] Z. Aidoun, M. Ouzzane, The effect of operating conditions on the performance
of a supersonic ejector for refrigeration, International Journal of Refrigeration
27 (2004) 974e984.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 247
[3] S. Aphornratana, I.W. Eames, A small capacity steam-ejector refrigerator:
experimental investigation of a system using ejector with movable primary
nozzle, International Journal of Refrigeration-Revue Internationale Du Froid
20 (1997) 352e358.
[4] Y. Bartosiewicz, Z. Aidoun, Y. Mercadier, Numerical assessment of ejector
operation for refrigeration applications based on CFD, Applied Thermal
Engineering 26 (2006) 604e612.
[5] K. Chunnanond, Ejectors: applications in refrigeration technology, Renewable
and Sustainable Energy Reviews 8 (2004) 129e155.
[6] I.W. Eames, S. Aphornratana, H. Haider, A theoretical and experimental-study
of a small-scale steam jet refrigerator, International Journal of Refrigeration-
Revue Internationale Du Froid 18 (1995) 378e386.
[7] B.J. Huang, C.B. Jiang, F.L. Hu, Ejector performance-characteristics and design
analysis of jet refrigeration system, Journal of Engineering for Gas Turbines
and Power-Transactions of the Asme 107 (1985) 792e802.
[8] J.T. Munday, D.F. Bagster, New ejector theory applied to steam jet refrigera-
tion, Industrial & Engineering Chemistry Process Design and Development 16
(1977) 442e449.
[9] S.B. Riffat, S.A. Omer, CFD modelling and experimental investigation of an
ejector refrigeration system using methanol as the working uid, Interna-
tional Journal of Energy Research 25 (2001) 115e128.
[10] E. Rusly, L. Aye, W. Charters, A. Ooi, CFD analysis of ejector in a combined
ejector cooling system, International Journal of Refrigeration 28 (2005)
1092e1101.
[11] D.W. Sun, Variable geometry ejectors and their applications in ejector
refrigeration systems, Energy 21 (1996) 919e929.
[12] R. Yapici, H. Ersoy, A. Aktoprakoglu, H. Halkaci, O. Yigit, Experimental deter-
mination of the optimum performance of ejector refrigeration system
depending on ejector area ratio, International Journal of Refrigeration 31
(2008) 1183e1189.
[13] E.J. Roshke, F.P. Massier, H.L. Gier, Experimental Investigation of Exhaust
Diffuser for Rocket Engines, Jet Propulsion Laboratory, Pasadena, 1962.
[14] M. Alperin, J.-J. Wu, Thrust augmenting ejectors, part I, Aiaa Journal 21 (1983)
1428e1436.
[15] M. Alperin, J.-J. Wu, Thrust augmenting ejectors, part II, Aiaa Journal 21 (1983)
1698e1706.
[16] B. Zhou, B.A. Fleck, F. Bouak, J.E.D. Gauthier, Comparison of swirling effects on
ejector performance using four turbulence models, Canadian Aeronautics and
Space Journal 46 (2000) 178e182.
[17] C. Li, Y.Z. Li, Investigation of entrainment behavior and characteristics of
gaseliquid ejectors based on CFD simulation, Chemical Engineering Science
66 (2011) 405e416.
[18] R. Yadav, A. Patwardhan, Design aspects of ejectors: effects of suction
chamber geometry, Chemical Engineering Science 63 (2008) 3886e3897.
[19] S. Balamurugan, M. Lad, V. Gaikar, A. Patwardhan, Hydrodynamics and mass
transfer characteristics of gaseliquid ejectors, Chemical Engineering Journal
131 (2007) 83e103.
[20] G.M. de Billerbeck, J.S. Condoret, C. Fonade, Study of mass transfer in a novel
gaseliquid contactor: the aero-ejector, Chemical Engineering Journal 72
(1999) 185e193.
[21] N. Beithou, H.S. Aybar, A mathematical model for steam-driven jet pump,
International Journal of Multiphase Flow 26 (2000) 1609e1619.
[22] J.H. Keenan, E.P. Neumann, A simple air ejector, Journal of Applied Mechanics-
Transactions of the Asme 64 (1942) 75e81.
[23] J.H. Keenan, E.P. Neumann, F. Lustwerk, An investigation of ejector design by
analysis and experiment, Journal of Applied Mechanics-Transactions of the
Asme 72 (1950) 299e309.
[24] B.J. Huang, J.M. Chang, C.P. Wang, V.A. Petrenko, A 1-D analysis of ejector
performance, International Journal of Refrigeration-Revue Internationale Du
Froid 22 (1999) 354e364.
[25] E.D. Rogdakis, G.K. Alexis, Design and parametric investigation of an ejector in
an air-conditioning system, Applied Thermal Engineering 20 (2000) 213e226.
[26] S. Mikhail, Mixing of coaxial streams inside a closed conduit, Journal of
Mechanical Engineering Science 2 (1960) 59e68.
[27] Y. Zhu, W. Cai, C. Wen, Y. Li, Shock circle model for ejector performance
evaluation, Energy Conversion and Management 48 (2007) 2533e2541.
[28] K.E. Hickman, P.G. Hill, G.B. Gilbert, Analysis and testing of compressible ow
ejectors with variable area mixing tubes, Journal of Basic Engineering 94
(1972) 407e416.
[29] K.R. Hedges, P.G. Hill, Compressible ow ejectors. 1. Development of a nite-
difference ow model, Journal of Fluids Engineering-Transactions of the Asme
96 (1974) 272e281.
[30] G.B. Gilbert, P.G. Hill, Analysis and Testing of Two-dimensional Slot Nozzle
Ejectors with Variable Area Mixing Sections, 2nd Symposium on Jet Pumps
and Ejectors and Gas Lift Techniques, Cambridge, England, United Kingdom,
1975, 45e64.
[31] S. Nilavalagan, M. Ravindran, H.C. Radhakrishna, Analysis of mixing charac-
teristics of ow in a jet pump using a nite-difference method, Chemical
Engineering Journal 39 (1988) 97e109.
[32] S. Aphornratana, S. Chungpaibulpatana, P. Srikhirin, Experimental investiga-
tion of an ejector refrigerator: effect of mixing chamber geometry on system
performance, International Journal of Energy Research 25 (2001) 397e411.
[33] K. Chunnanond, An experimental investigation of a steam ejector refrigerator:
the analysis of the pressure prole along the ejector, Applied Thermal Engi-
neering 24 (2004) 311e322.
[34] T. Narabayashi, W. Mizumachi, M. Mori, Study on two-phase ow dynamics in
steam injectors, Nuclear Engineering and Design 175 (1997) 147e156.
[35] T. Narabayashi, M. Mori, M. Nakamaru, S. Ohmori, Study on two-phase ow
dynamics in steam injectors - II. High-pressure tests using scale-models,
Nuclear Engineering and Design 200 (2000) 261e271.
[36] J. Yan, S. Shao, J. Liu, Z. Zhang, Experiment and analysis on performance of
steam-driven jet injector for district-heating system, Applied Thermal Engi-
neering 25 (2005) 1153e1167.
[37] C.A.M.C. Dirix, K. Vanderwiele, Mass-transfer in jet loop reactors, Chemical
Engineering Science 45 (1990) 2333e2340.
[38] P. Cramers, A. Beenackers, Inuence of the ejector conguration, scale and the
gas density on the mass transfer characteristics of gaseliquid ejectors,
Chemical Engineering Journal 82 (2001) 131e141.
[39] Y. Bando, M. Kuraishi, M. Nishimura, M. Hattori, I. Takeshita, The character-
istics of ow of a bubble column with a gas-suction-type, simultaneous gas-
liquid injection-nozzle, International Chemical Engineering 30 (1990) 9.
[40] P. Havelka, V. Linek, J. Sinkule, J. Zahradnik, M. Fialova, Effect of the ejector
conguration on the gas suction rate and gas hold up in ejector loop reactors,
Chemical Engineering Science 52 (1997) 1701e1713.
[41] M. Kandakure, V. Gaikar, A. Patwardhan, Hydrodynamic aspects of ejectors,
Chemical Engineering Science 60 (2005) 6391e6402.
[42] S. Balamurugan, V. Gaikar, A. Patwardhan, Effect of ejector conguration on
hydrodynamic characteristics of gaseliquid ejectors, Chemical Engineering
Science 63 (2008) 721e731.
[43] R. Matysko, D. Butrymowicz, W. Angielczyk, M. Trela, M. Bergander, Model of
Supercritical Two-Phase Steam-Water Injector, 7th World Conference on
Experimental Heat Transfer, Fluid Mechanics and Thermodynamics, Krakow,
Poland, 2009, 491e498.
[44] S.B. Riffat, G. Gan, S. Smith, Computational uid dynamics applied to ejector
heat pumps, Applied Thermal Engineering 16 (1996) 291e297.
[45] T. Sriveerakul, S. Aphornratana, K. Chunnanond, Performance prediction of
steam ejector using computational uid dynamics: part 1. Validation of the
CFD results, International Journal of Thermal Sciences 46 (2007) 812e822.
[46] T. Sriveerakul, S. Aphornratana, K. Chunnanond, Performance prediction of
steam ejector using computational uid dynamics: part 2. Flow structure of
a steam ejector inuenced by operating pressures and geometries, Interna-
tional Journal of Thermal Sciences 46 (2007) 823e833.
[47] Y. Zhu, W. Cai, C. Wen, Y. Li, Numerical investigation of geometry parameters
for design of high performance ejectors, Applied Thermal Engineering 29
(2009) 898e905.
[48] I. Watanabe, Experimental Investigations Concerning Pneumatic Ejectors,
with Special Reference to the Effect of Dimensional Parameters on Perfor-
mance Characteristics, Symposium on Jet Pumps and Ejectors, London,
England, United Kingdom, 1972, 97e120.
[49] ESDU, in: Ejector and Jet Pump, Data Item 86030, ESDU International Ltd.,
London, 1985.
[50] D.R. Croft, D.G. Lilley, Jet pump design and performance analysis, in: AIAA
14th Aerospace Sciences Meeting, American Institute of Aeronautics and
Astronautics, Washington, D.C ; United States, 1976.
[51] P. Chaiwongsa, S. Wongwises, Experimental study on R-134a refrigeration
system using a two-phase ejector as an expansion device, Applied Thermal
Engineering 28 (2008) 467e477.
[52] Y. Bartosiewicz, Z. Aidoun, P. Desevaux, Y. Mercadier, Numerical and experi-
mental investigations on supersonic ejectors, International Journal of Heat
and Fluid Flow 26 (2005) 56e70.
[53] D.K. Acharjee, P.A. Bhat, A.K. Mitra, A.N. Roy, Studies on momentum-transfer in
vertical liquid-jet ejectors, Indian Journal of Technology 13 (1975) 205e210.
[54] M.N. Biswas, A.K. Mitra, A.N. Roy, Studies on Gas Dispersion in a Horizontal
Liquid Jet Ejector, 2nd Symposium on Jet Pumps and Ejectors and Gas Lift
Techniques, Cambridge, England, 1975, p. E3-27-42.
[55] H.J. Henzler, Design of ejectors for single-uid material systems, German
Chemical Engineering 6 (1983) 292e300.
[56] W.H. Lee, Pressure iteration scheme for two-phase ow modeling, in:
T.N. Verizoglu (Ed.), Multiphase Transport: Fundamentals, Reactor Safety, Appli-
cations, Vol. 1, Hemisphere Publishing, Washington DC, USA, 1980, pp. 407e432.
[57] A. Hemidi, F. Henry, S. Leclaire, J.-M. Seynhaeve, Y. Bartosiewicz, CFD analysis
of a supersonic air ejector. part I: experimental validation of single-phase and
two-phase operation, Applied Thermal Engineering 29 (2009) 1523e1531.
[58] M. Elkady, A. Karameldin, E.S. Negeed, R. El-Bayoumy, Experimental investi-
gation of the effect of ejector geometry on its performance, International
Journal of Nuclear Desalination 3 (2008) 215e229.
[59] K. Matsuo, Y. Miyazato, H.D. Kim, Shock train and pseudo-shock phenomena
in internal gas ows, Progress in Aerospace Sciences 35 (1999) 33e100.
C. Li et al. / Applied Thermal Engineering 48 (2012) 237e248 248