New Journal and we have not received input yet 25 (2021) 101016

Contents lists available at ScienceDirect

Thermal Science and Engineering Progress

journal homepage: www.sciencedirect.com/journal/thermal-science-and-engineering-progress

One dimensional steam ejector model based on real fluid property

Borirak Kitrattana a, Satha Aphornratana a, *, Tongchana Thongtip b
a
Department of Mechanical Engineering, Sirindhorn International Institute of Technology, Thammasat University, P.O. Box 22 Thammasat Rangsit PO, Patumthani
12121, Thailand
b
Advanced Refrigeration and Air Conditioning Laboratory (ARAC), Department of Teacher Training in Mechanical Engineering, King Mongkut’s University of
Technology North Bangkok, 1518 Phacharat 1 Rd., Bang Sue, Bangkok 10800, Thailand

A R T I C L E I N F O A B S T R A C T

Keywords: The one-dimensional ejector model based on the real fluid property is proposed to design the steam ejector
Steam ejector geometries and to predict its performance. This model is developed from the one-dimensional compressible flow
Ejector refrigeration of steam-water which was firstly proposed by Stoecker (1958). The ejector geometry and performance deter­
Jet pump
mined by the proposed model are validated with the experimental results to ensure that the proposed model can
be used efficiently for designing the steam ejector. Also, the results obtained from the one-dimensional model
based on the ideal gas are compared with those obtained from the proposed model. It is found that both models
can efficiently be used to design the ejector geometry and to predict the performance (in terms of entrainment
ratio and critical condenser pressure). However, the one-dimensional model based on the ideal gas provides
relatively low temperatures at each state compared to the proposed model. The calculated temperature is well
below the freezing point while that calculated by the proposed model provides fluid properties at equilibrium
condition. In addition, the condensation process of the supersonic flow is involved in many processes inside the
ejector. This indicates that the ideal gas assumption for the wet fluid (steam is classified as the wet fluid) might
not suitable for analysing the process of the ejector.

an ideal gas while the conservative equations of mass, momentum and

1. Introduction energy were used simultaneously. The Keenan theory was further
developed by Munday and Bagster [4], B.J. Huang et al. [5], and
An ejector refrigeration system is a heat powered refrigeration sys­ Ruangtrakoon and Aphornratana [6] for providing a reasonably accu­
tem which can convert thermal energy into useful refrigeration. The rate result as compared with the experimental results.
power source of the ejector refrigeration system is available from in­ With the rapid development in personal computers and commercial
dustrial waste heat, solar energy, or geothermal energy, etc., which is software, the performance of the designed ejector is easily predicted by
cheap or even free. However, performance of the system, in terms of implementing the computational fluid dynamics (CFD) simulation. The
COP, is relatively low compared with an absorption refrigeration system previous research by Ruangtrakoon et al. [7] and Sriveerakul et al. [8]
(the most widely used heat powered refrigeration system). However, the showed that the predicted steam ejector performance via CFD simula­
ejector refrigeration system is much simpler, both in system configura­ tion agreed well with the experimental results at the same working
tion and operation. In order to make the ejector refrigeration system condition. Also, performance prediction of the ejector working with
more competitive with the absorption refrigeration system, the system R141b and R134a as working fluid using the CFD technique [9,10]
COP should be improved. showed good agreement with the experimental result. In addition, the
One aspect to improve the system performance is through the precise contour of relevant parameters, such as the Mach number contour, can
design of the ejector geometries because it strongly affects the entire efficiently represent the flow process inside the ejector. This can be used
system performance. The change in the ejector geometries has a sig­ to discuss the parameter which affects the ejector performance as pro­
nificant influence on the ejector efficiency as reported in many previous posed by previous work [7]. Even though the CFD simulation technique
studies [1,2]. To design the ejector geometry, one of the commonly used has been extensively used to discuss and to predict the ejector perfor­
methods is a one-dimensional flow theory which was first proposed by mance, it cannot be used for directly designing the ejector geometries.
Keenan et al. [3]. Using this theory, the working fluid was assumed to be This is because a particular ejector geometry must initially be designed

* Corresponding author.
E-mail address: satha@siit.tu.ac.th (S. Aphornratana).

https://doi.org/10.1016/j.tsep.2021.101016
Received 24 September 2020; Received in revised form 7 July 2021; Accepted 7 July 2021
Available online 15 July 2021
2451-9049/© 2021 Elsevier Ltd. All rights reserved.
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Nomenclature W˙pump Work input at the pump (kW)

x Dryness quality
A Cross-sectional area (m2)
COP Coefficient of performance Greek letter
D Diameter (mm) ρ Density (kg⋅m− 3)
h Specific enthalpy (kJ⋅kg− 1) υ Specific volume (m3⋅kg− 1)
M Mach number Subscript
m˙ Mass flow rate (kg⋅sec-1) boiler Boiler
P Absolute pressure (kPa, bar, mbar) cri Critical
Q˙boiler Heat input to the boiler (kW) evap Evaporator
Q˙evap Cooling load at the evaporator (kW) f Saturated fluid
Rm Entrainment ratio fg Difference between saturated vapour and saturated fluid
S Slip ratio g Saturated vapour
s Specific entropy (J⋅K− 1) p Primary fluid
T Temperature (◦ C, K) s Secondary fluid
V velocity (m⋅sec-1)

for creating the CFD modelling. Later, performance of the ejector is properties for achieving the solution. Therefore, at that time, it was quite
predicted by means of numerical methods when all boundary conditions difficult to calculate the flow state of the working fluid with only the use
are specified. Hence, the one-dimensional flow ejector theory must still of a calculator and a conventional fluid properties table (especially at
be used to design the ejector geometries. This is the reason why at very low absolute pressure and close to the freezing temperature). Thus,
present, one-dimensional theory is still widely used for designing the the designed model for the steam ejector proposed by Stoecker [11] has
ejector geometry. been long forgotten and was even removed from his text book in the
The one-dimensional ejector theory was first developed for working later editions (this also may be due to the unpopularity of steam ejector
with air which is assumed to be an ideal gas. Thus, most of the mathe­ refrigeration systems at that time).
matical models which have been developed by many researchers [4–6] Nowadays, various models used to determine the real fluid proper­
are based on the ideal gas assumption in which the working fluid ties are available and reliable. Moreover, the saturation properties can
properties can easily be determined by the ideal gas equation of state. be achieved by means of a computer software package. This makes it
However, some of the working fluids used with the ejector, such as easily available as well as providing high accuracy. In addition, it re­
steam-water, cannot be assumed to be ideal gas. This is because the quires little calculation time for achieving the real working fluid prop­
condensation process is involved in the flow processes within the erties under various working conditions. This makes Stoecker’s ejector
ejector. Moreover, the temperature calculated by the ideal gas model feasible today. However, Stoecker’s ejector model has not been
assumption is well below the fluid freezing point. Therefore, the ideal developed comprehensively for the steam ejector based-refrigeration.
gas assumption might not be appropriate for the ejector working with There is still no evidence to prove that the model is efficient for
steam-water. designing the whole steam ejector.
In 1958, Stoecker [11] proposed a mathematical model for designing In this paper, an ejector model based on Stoecker’s theory is used to
the ejector geometry which employs the real fluid property instead of design and to analyse the steam ejector which is used for a refrigeration
those obtained by the ideal gas assumption. This model requires itera­ application. The mathematical models of two-phase flow based on a
tion to calculate the flow states occurring inside the ejector and across homogeneous equilibrium assumption [12] are developed for designing
the normal shock wave. The calculation process requires several fluid the ejector geometries. The Stoecker model was to include the loss

Fig. 1. Schematic diagram of an ejector refrigeration system.

2
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Fig. 2. Processes within the steam ejector on a Temperature–Entropy chart.

coefficient of the primary nozzle, the mixing chamber, and the diffuser model based on ideal gas (Keenan’s model).
to be more realistic. The calculated results are validated with experi­ The results show that both the Stoecker’s model and the Keenan’s
mental data for indicating the efficient use of the developed mathe­ model provide accurate results (in term of critical condenser pressure,
matical models. The calculated results based on this model are also entrainment ratio, and geometries) when compared with the experi­
compared with those based on the classical one-dimensional ejector mental data. Even though both models can be used to design a steam

3
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Fig. 3. Schematic view of a steam ejector used in the analysis.

Fig. 4. Schematic view at the primary nozzle exit and the mixing chamber inlet.

ejector, the calculated temperatures of the steam-water within the and accelerated to supersonic velocity at the nozzle exit. The low-
ejector from Keenan’s model are extremely low; for example, at the pressure region is created by this primary fluid supersonic stream (jet
primary nozzle, the temperature of the supersonic primary steam was stream) at the nozzle’s exit plane (3). The low-pressure region draws the
calculated to be around − 149 ◦ C at the primary nozzle exit plane. This is low pressure and low temperature fluid, known as the secondary fluid
in contrast to the result obtained from Stoecker’s model which is 2.4 ◦ C. (4), into the mixing chamber.
Moreover, the process within the ejector (especially the fluid tempera­ Within the mixing chamber, the converging duct which is formed
ture) is significantly different. The Stoecker’s model demonstrates that between the primary fluid jet core and mixing chamber wall is created as
the state of the steam-water within the mixing chamber is in the two- shown in Fig. 4. The secondary fluid is accelerated along this converging
phase region which is significantly different from the ideal gas duct until its velocity reaches sonic velocity (Choked flow). Munday and
assumption. Bagster [4] proposed that the primary fluid and the secondary fluid
started to mix together when the secondary fluid was accelerated to
2. Background sonic velocity (choked flow). At the mixing chamber throat section (6), it
is assumed that the two fluid streams are completely mixed and their
Fig. 1 shows a schematic diagram of an ejector refrigeration system. velocity is still at supersonic level. Later, the mixed fluid stream (low
It consists of 5 main components: a boiler, an ejector, a condenser, an pressure supersonic stream) experiences a high-pressure region which is
evaporator, and a circulating pump. caused by the condenser pressure. This causes the creation of the normal
From Fig. 1, it can be seen that the system consists of an evaporator, a shock wave (which is classified as the compression shock wave). Across
condenser, and an expansion valve which is similar to an electrically the normal shock (7), flow velocity of the mixed fluid is reduced sud­
powered vapour compression refrigeration system. However, the denly to subsonic level. The kinetic energy of the flow is converted back
elevation of the refrigerant pressure is achieved by means of the ejector to enthalpy. This causes the temperature, pressure, and density of the
and boiler instead of a mechanical compressor as is used in a vapour mixed fluid to increase sharply. The velocity of the mixed stream is
compression refrigeration system. further reduced to almost a stagnation state at the subsonic diffuser exit
Fig. 2 shows the flow state of the working fluid on a temperature- (8) before discharging to the condenser.
entropy diagram. Figs. 3 and 4 typically show the location of each From the temperature-entropy diagram shown in Fig. 2, it can be
state within the ejector. From the figures, a high pressure and high seen that, upstream of the normal shock all the processes involved are on
temperature primary fluid (1) flows and expands through the primary the two-phase region, and, therefore, the behaviour of the working fluid
nozzle. The primary fluid accelerates to sonic velocity (Mach number of is far from that based on the ideal gas assumption. This means the flow
1.0) at the primary nozzle throat (2). The fluid is then further expanded state based on the ideal gas assumption may not correctly represent the

4
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

process occurring inside the steam ejector. Herein, the Stoecker’s model
is employed to explain the actual process (occurring on two-phase re­
gion) for which the ideal gas model proposed by Keenan is limited.
As explained earlier, upstream of the normal shock all the processes
involved occur under a two-phase region (saturated liquid–vapour
mixture region). Thus, this steam ejector model must deal with the two-
phase flow model to be more realistic. The fundamental two-phase flow
[14] model can be categorized into two main types: the homogenous
flow model and the separate flow model.
For the homogenous flow model, it is assumed that both phases are
well mixed and flow with the same velocity. This model is simpler than
the separate flow model. However, this flow model is not valid for all
cases because the velocity and property of each phase is different. For
the separate flow model, it is assumed that each phase is separated from
each other, and, thus, the velocity of each phase is different. In this
situation, the proportion of the vapour phase to the liquid phase plays an
Fig. 5. Performance of a steam ejector based on experimental data provided by
important role in the flow behaviour. The factors which have significant Kitrattana et.al. [13]
influence on two-phase flow are defined as the “void fraction” and the
“slip ratio”. The void fraction is the fractional area occupied by gas
phase. The slip ratio is defined as the ratio between the velocities of the
vapour phase to that of the liquid phase. The slip ratio S = 1 for the
homogenous flow model where both phases travel at the same velocity.
Usually, the slip ratio is indirectly determined by experiment since
the velocity of each phase cannot be measured directly. Many re­
searchers have developed the method to calculate the slip ratio [15–17].
The slip ratio depends on many factors, such as the mixture quality, mass
flow rate, temperature, pressure, flow direction, heat loss or heat gain,
and shape of the cross-section flow area.
Moreover, the effect of supersaturation takes place within the su­
personic nozzle [18]. The supersaturated flow or metastable flow occurs
when the vapour expands through the nozzle without condensation at
the saturation line. This is because the fluid flows with very high ve­
locity. The time of the vapour traveling through the nozzle is very short.
Hence, the time may not be sufficient for forming the liquid droplet
(condensation). The condensation is then delayed and the properties of Fig. 6. Effect of the boiler saturation temperature and the evaporator satura­
the supersaturated vapour do not rely on the equilibrium condition. This tion temperature on the system performance.
makes it more complicated for calculating the fluid properties due to the
lack of availability of the fluid properties in supersaturated state. Based on the operating temperatures of the boiler and the evapo­
In this study, for simplicity of calculation, the two-phase flow in the rator, it can be assumed that the enthalpy change at the evaporator is
steam ejector is assumed to be homogenous equilibrium [14]. This approximately equal to that of the boiler:
means the vapour and liquid phases travel together with the same ve­ Δh@evap ≈ Δh@boiler (4)
locity, at the same time, the pressure and temperature of both phases are
the same. This is because the flow process in the ejector occurs at very and therefore, it may be assumed that:
high velocity and at relatively high dryness quality. Moreover, the effect
COP ≈ Rm (5)
of supersaturated flow is not applied to this present work because of its
complexity. However, the loss coefficient of the nozzle is included Basically, the system COP or the entrainment ratio is plotted against
during the development of the model in order to compensate for the the condenser pressure for determining the critical point of operations.
effect of supersaturation. This provides a much simpler calculation and a This critical point indicates the highest possible condensation pressure
reasonably accurate result when compared with experimental data. while the entrainment ratio is still kept constant at the highest possible
Performance of the ejector is usually described by the entrainment value. The typical performance curve of a steam ejector obtained from
ratio which is the ratio of secondary mass flow rate to primary mass flow the previous experiments [13] is shown in Fig. 5. The experimental
rate: steam ejector has a mixing chamber throat diameter of 19 mm and the
primary nozzle has a throat diameter of 2.0 mm. Such a performance
ṁevap
Rm = (1) curve was obtained when the boiler (primary fluid) and the evaporator
ṁboiler
(secondary fluid) saturation temperatures were fixed at 130 ◦ C and
Coefficient of performance (COP) of the ejector refrigeration system 7.5 ◦ C, respectively. During the test, the condenser pressure was
is: increased from the lowest possible value (depending on the temperature
and flow rate of the cooling water supplied) until the ejector failed to
Q̇evap operate.
COP = (2)
Q̇boiler + Ẇ pump From Fig. 5, there are three operating regions: choked flow in the
The mechanical work required for the circulating pump is negligible mixing chamber; unchoked flow in the mixing chamber; and reversed
since it is relatively small compared to the heat input required at the flow in the mixing chamber [1]. Since the primary nozzle is always
boiler. The COP is: choked due to the boiler saturation temperature being fixed, the mass
flow rate of the primary fluid generated by the boiler remained constant
COP = Rm⋅
Δh@evap
(3) throughout the test. In the choked flow region, the choked flow of sec­
Δh@boiler ondary fluid at the effective area is found. In this region, the secondary

5
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Fig. 7. The experimental entrainment ratio at critical condenser pressure operation.

mass flow and entrainment ratio are constant and independent from the used to predict the ejector’s performance when it is operated at the
condenser pressure. The sudden drop of entrainment ratio with critical condenser pressure. In the next section a 1-D theory based on the
increasing the condenser pressure is found when operating on the steam-water real properties as proposed by Stoecker will be described.
unchoked flow region. This is because the impact of the shock wave
disturbs the mixing process, causing the ejector to draw less secondary 3. A steam ejector model based on thermodynamic properties of
fluid. In the reverse flow region, the ejector malfunctions and the sec­ steam-water
ondary fluid flows back to the suction port.
Fig. 6 shows the effects of the boiler saturation temperature and the As mentioned earlier, the process inside the ejector is liquid–vapor
evaporator saturation temperature on the system performance. When mixture, thus the Keenan’s model (assumption of working fluid as a
the boiler saturation temperature is decreased while the evaporator perfect gas) might not be appropriate. The model using real fluid
saturation temperature is fixed, a slight increase in the secondary fluid is properties is more suitable. The aim of this study is to develop the simple
achieved (the cooling capacity is also increased). Moreover, the critical mathematical model using the real fluid properties. The model is
mass flow rate of the primary fluid (also the boiler heat input) is developed from the Stoecker’s model [11] with loss coefficient to be
dramatically reduced. Therefore, the entrainment ratio and the COP are more realistic. One dimensional flow theory and the conservative
significantly increased. However, the ejector can operate at a lower equations of mass, momentum, and energy conservation are employed
critical condenser pressure, and therefore, a lower temperature of the for developing the mathematical models.
cooling water is required. However, for simplicity, some assumptions have been made for this
If the evaporator saturation temperature is increased while the boiler model. The assumptions of this model are as follows:
saturation is fixed, the ejector entrains more secondary fluid (the cooling
capacity is significantly increased). This results in an improvement of 3.1. General assumption
the entrainment ratio and the COP. Moreover, the ejector can operate
with a higher critical condenser pressure. Therefore, the system can be • The process is at steady state.
operated with relatively higher cooling water temperature. • Primary and secondary fluid are saturated vapor.
It may be said that, for a specified boiler saturation temperature • Homogenous equilibrium two-phase flow model is considered.
(depending on the thermal energy supplied), and evaporator saturation • Fluid at the ejector outlet is at stagnation state (velocity = 0).
temperature (depending on the refrigerating temperature required), the • No heat loss between the fluid to surroundings.
ejector should be designed to be operated at the critical condenser • Constant pressure mixing process is assumed (constant pressure from
pressure (depending on the temperature of the cooling water supplied). state 3 through state 6).
Fig. 7 shows a performance map or carpet plot of the tested ejector. Any • The solidification of working fluid is ignored.
point resulting from the intersection between the boiler isothermal line • Super saturated, mixing loss and other losses are accounted by the
and the evaporator isothermal line is the entrainment ratio or COP at the isentropic efficiency and mixing efficiency
relevant critical condenser pressure.
Performance of a steam ejector is usually analysed by using 1-D
3.2. Primary nozzle
ejector theory. It must be noted that, in this theory, the normal shock
is induced at the centre of the throat of the mixing chamber after the two
A supersonic nozzle or De Laval nozzle was first invented by the
fluid streams are completely mixed. Due to this, the theory only can be
Swedish inventor Gustaf de Laval in 1888. In the steam ejector, the De

6
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Fig. 8. De Laval nozzle and the variation of Mach number.

Laval nozzle is known as the primary nozzle which is used to produce a Once the pressure (P2) and enthalpy (h2) are obtained, the actual
supersonic jet stream, resulting in producing a low-pressure environ­ state at the nozzle throat can be determined. Normally, when a two-
ment within the mixing chamber. Fig. 8 shows a drawing of a supersonic phase fluid flows through the nozzle, the velocity of each phase is
nozzle. High temperature and high-pressure steam, which is the dry different as explained by [12]. The liquid phase flows at a slower ve­
saturated steam produced by the boiler, enters the nozzle on the left- locity than the gas phase. However, in this present work, the different
hand side (1). It accelerates and expands through the elliptical velocity of each phase is not considered and the amount of fluid in the
converging duct profile. The steam velocity reaches sonic value and liquid phase is relatively low compared to fluid in the vapour phase
chokes at the throat of the nozzle (2). The steam further expands and (dryness quality more than 95%) at the nozzle throat. Thus, for
accelerates through the conical profile, diverging duct and discharges as simplicity, the flow can be assumed to be homogenous equilibrium. This
a supersonic jet stream on the right-hand side (3). means the two-phase fluid is well mixed while the pressure and tem­
For specified pressure and temperature at the inlet, when the nozzle perature for each phase are identical. The velocity of steam at the nozzle
is being choked, the mass flow rate of the steam reaches its maximum throat is obtained from
value. Thereafter, it cannot be increased even when the nozzle down­
V12 V2
stream pressure is reduced. This maximum mass flow rate is called a h1 + = h2 + 2 (11)
2 2
critical mass flow rate(ṁcri ).
The critical mass flow rate is usually calculated based on the ideal gas The flow velocity at the nozzle’s inlet is normally negligible (V1 ≈ 0
assumption as proposed by many previous studies [1,2]. However, this m⋅sec− 1). The actual dryness quality is
may not be perfectly correct for steam as mentioned previously. Theo­ h2 − hf @P2
retically, when the high temperature and high-pressure saturated steam x2 = (12)
hfg@P2
(1) expands isentropically through the nozzle, part of the vapour is
condensed. Therefore, during the expansion process, steam is two-phase The cross-sectional area of the nozzle throat is
fluid whose behaviour is far from that based on the ideal gas assumption.
ṁcri ⋅υ2
In this paper, the steam is treated as a homogenous equilibrium fluid for A= (13)
V2
simplicity as mentioned previously. Meanwhile, the loss coefficient of
the nozzle is also used for compensation of the non-isentropic flow and
supersaturation. 3.3. Secondary fluid at the primary nozzle exit plane
To design the primary nozzle based on the two-phase model pro­
posed in this present work, the properties of the inlet steam (1), the Based on the experimental investigation provided by Ruangtrakoon
nozzle efficiency, and the critical mass flow rate are all required. Since and Aphornratana, [2], at the primary nozzle exit it is assumed that the
the supersonic nozzle is always operated with the choked flow condi­ pressure is uniform over the cross section and is approximately equal to
tion, the critical mass flow rate and the flow state of the steam at the 70% of the saturation pressure inside the evaporator (P3 = P5 = 0.7⋅P4).
throat are independent of the downstream condition of the throat. The It is thought that the secondary fluid is expanded and accelerated
pressure at the primary nozzle throat (P2) is initially assumed and the isentropically from stagnation state (4) to some velocity at the nozzle
properties of the saturated vapour and saturated liquid are later defined. exit plane (5); therefore,
If the flow process is isentropic, the state of steam at the nozzle throat is
s5 = s4 (14)
s2′ = s1 (8)
Since the steam is wet vapour, the dryness quality is calculated from
Since the steam is the wet vapour, the dryness quality is calculated s5 − sf @P5
from x5 = (15)
sfg@P5
s2′ − sf @P2
x 2′ = (9) State of the secondary vapour at the primary nozzle exit plane (5) is
sfg@P2
then defined; therefore, the pressure (P5) and enthalpy (x5) are obtained.
If the nozzle’s isentropic efficiency is considered, the enthalpy at the The velocity of the secondary vapour is
nozzle throat (h2) can be calculated by
V42 V2
h1 − h2 h4 + = h5 + 5 (16)
ηnozzle = (10) 2 2
h1 − h2′
The flow velocity V4 is normally negligible. The cross-sectional area

7
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

for the secondary vapour at the primary nozzle exit plane is Table 1
Actual operating condition of the experimental ejector.
ṁs ⋅υ5
A5 = (17) Boiler saturation temperature 130 ◦ C
V5
Boiler saturation pressure 270.1 kPa
Primary steam mass flow rate 4.58 kg⋅hour-1
3.4. Mixing process of the two fluid streams inside the mixing chamber Evaporator saturation temperature 7.5 ◦ C
Evaporator saturation pressure 1.036 kPa
Secondary vapour mass flow rate 1.19 kg⋅hour-1
Within the mixing chamber, primary stream with supersonic velocity
Ejector discharged pressure (condenser saturation pressure) 4.8 kPa
and the secondary vapour with subsonic velocity are mixed together Primary nozzle throat diameter 2.0 mm
under the constant pressure condition from the inlet (3 and 5) to the Mixing chamber throat diameter 19.0 mm
throat section (6). At the throat section, it is assumed that the two fluid
streams are completely mixed. The velocity of the mixed stream is ob­
tained from the conservation of momentum equation which is
(ṁp + ṁs )⋅V6 = ηmix ⋅(ṁp ⋅V3 + ṁs ⋅V5 ) (18)

where ηmix is mixing efficiency (which was introduced by [20]). The

mixing efficiency value of ηmix ≈ 90%is recommended for this present
work. The enthalpy of the mixed stream can be determined as

V62
(ṁp + ṁs )⋅(h6 + ) = ṁp ⋅h1 + ṁs ⋅h4 (19)
2
Since P6 and h6 are obtained, the state of mixed stream at the throat
section (6) is defined. The mixing chamber throat cross-sectional area is
(ṁp + ṁs )⋅υ6
A6 = (20)
V6

3.5. Normal shock at the mixing chamber throat

Within the throat, it is assumed that the primary stream and the
secondary stream are completely mixed and their velocity is still at su­
personic level. Usually, when the supersonic stream experiences high
downstream pressure (the condenser saturation pressure), the normal
shock wave is induced. Across the normal shock, the velocity of the
mixed stream drops suddenly to subsonic level. Pressure, temperature,
and density of the mixed stream are increased suddenly (normal shock Fig. 9. Calculation flow chart of the primary nozzle.
has zero thickness). To determine the downstream state of the normal
shock, mass conservation equation, energy conservation equation, and given as
momentum conservation equation are used simultaneously.
Mass conservation equation ηdiffuser =
h8′′ − h7
(24)
h8′ − h7
V6 V7
= (21)
υ6 υ7 4. Calculation procedure and calculation example
Energy conservation equation
For the input data and base-reference case, the experimental data
V62 V2 obtained from the authors’ previous study [13] is used. In such a case, an
h6 + = h7 + 7 (22)
2 2 experimental steam ejector with mixing chamber throat diameter of 19
Momentum conservation equation mm and primary nozzle throat diameter of 2.0 mm was tested with the
boiler saturation temperature of 130 ◦ C and the evaporator saturation
V6
(P7 − P6 ) = ⋅(V6 − V7 ) (23) pressure of 7.5 ◦ C. The details of the test are provided in Table 1. The
υ6
temperature, pressure, and mass flow rate of the primary fluid and those
of the secondary fluid, are used as the input data. The critical condenser
3.6. Subsonic diffuser
pressure and the ejector geometries are the calculated output results.
Thermodynamic properties of steam-water are available from the soft­
At the end of the mixing chamber, which is the downstream of the
ware package, Computer-Aided Thermodynamic Tables (CATT version
normal shock wave (7), velocity of the mixed fluid is at subsonic level.
3) [19].
To make further pressure recovery of the mixed fluid, its velocity must
By varying the static pressure between nozzle exit and primary flow
be slowed down while travelling through a subsonic diffuser to almost
pressure, the primary nozzle throat is where the cross-sectional area
stagnation state (8). Therefore, during this process, the kinetic energy is
reaches the minimum value. If the static pressure is further reduced, the
converted to enthalpy, resulting in further compression effect. The
steam’s velocity is accelerated to supersonic level in the diverging duct
compression process through the subsonic diffuser is shown in Fig. 2
section. The pressure at the nozzle exit is the outlet boundary condition.
(process 7–8).
Its value must be specified during the calculation. Based on the experi­
When the mixed fluid is brought to the stagnation state, the highest
mental results provided by Ruangtrakoon et al. [2], the nozzle exit
possible discharge pressure is obtained. In practice, the compression
pressure has been proven to be approximately 70% of the secondary
process within the subsonic diffuser may not be reversible, and, thus, the
fluid pressure. Thus, the nozzle exit pressure is about 0.725 kPa (P3 =
loss of stagnation pressure is the result. This can be indicated by isen­
0.7 × 1.036 kPa). It must be pointed out that, if 70% of the secondary
tropic efficiency of the subsonic diffuser (diffuser’s efficiency) which is

8
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

supersonic level (due to the rapid increase in the specific volume). Thus,
the diverging duct is required for achieving the supersonic flow. At the
nozzle exit, the steam fans out with supersonic velocity (V3). Hence, the
converging–diverging profile duct can be achieved by calculating the
flow area as shown in Table 2.
After obtaining primary nozzle throat area and properties, the pro­
cess of calculation for the rest of the ejector is shown as a flowchart in
Fig. 12.
Table 3 shows the calculation results for the secondary vapour at the
primary nozzle exit plane using Eqs. (14)–(17). It must be noted that the
total size of the mixing chamber at the primary nozzle exit plane is
composed of the flow area for the primary fluid (A3) and the flow area
for the secondary fluid (A5) as shown in Fig. 13.
The mixing process is assumed to occur at the constant area section.
Table 4 shows the calculation results of the mixing process between the
Fig. 10. Variation of the primary nozzle flow area from the inlet to the exit. primary fluid and the secondary fluid within the mixing chamber using
Eqs. (18)–(20).
To determine the downstream state of the normal shock wave (7),
these three governing equations must be employed simultaneously (Eqs.
(21) to (23)). However, there are two solutions when using these gov­
erning equations to determine the downstream state of the shock wave.
For the first solution, the mass conservation equation is combined with
the energy conservation equation. This combined equation is used when
the downstream velocity of the normal shock is decreased from the su­
personic level (V6) to the stagnation state. Hence, the series of the
calculated specific volume (υ7) and specific enthalpy (h7) are obtained.
When these calculated values are shown on an Enthalpy–Entropy chart,
the process line called the “Fanno line” is plotted.
For the second solution, the mass conservation equation is combined
with the momentum conservation equation for determining the down­
Fig. 11. Variation of the primary fluid velocity, dryness quality, and specific stream state of the shock wave. The series of the calculated specific
volume through the primary nozzle from the inlet to the exit. volume (υ7) and calculated pressure (P7) are obtained. These calculated
values are shown on the Enthalpy–Entropy chart, and the process line
fluid pressure is equal to or less than 0.613 kPa (saturation pressure at called the “Rayleigh line” is determined.
0.01 ◦ C), a nozzle exit pressure of 0.613 kPa will be used, since it is the The calculated result is tabulated in Table 5. From Fig. 14, it can be
lowest pressure at which water remains in a liquid phase according the seen that when two process lines (Fanno line and Rayleigh line) are
steam-water properties table. The calculation flow chart of the primary plotted together, there are two intersecting points. The first intersecting
nozzle is shown in Fig. 9. point can indicate the upstream state of the normal shock (6). Mean­
Fig. 10 shows the variation of the calculated cross-sectional area of while, the second intersecting point can indicate the state downstream
the primary nozzle. Fig. 11 shows the variation of primary fluid velocity, of the normal shock (7). Hence, the downstream state of the normal
specific volume, and dryness quality along the nozzle from inlet to exit. shock wave, which is the goal of using the above governing equations,
The calculated data are listed in Table 2. From Fig. 10, when the pres­ can be determined accurately
sure is decreased, the calculated flow area is the cross-sectional area It can be seen from Fig. 14 and Table 5 that across the normal shock
along the nozzle. From the inlet, the area is reduced to the minimum wave, the flow velocity drops sharply from 873 m sec− 1 to 202 m sec− 1.
value at the throat (P2 = 160 kPa). The steam is accelerated from The temperature and pressure increase suddenly from 2.4 ◦ C and 0.7252
stagnation (V1 ≈ 0 m⋅sec− 1) at the nozzle inlet to sonic value (V2) at the kPa to 79.0 ◦ C and 4.3906 kPa, respectively. This indicates that the
nozzle throat. If the pressure continues to decrease, the flow area is shock wave causes a major compression effect in the ejector. This has a
increased while the steam’s velocity continues to accelerate to significant impact on the condenser working pressure. However, further

Table 2
Calculation results for the primary nozzle.
Boiler’s saturation temperature 130 ◦ C
Boiler’s saturation pressure 270.1 kPa
Isentropic efficiency 90 %
Critical mass flow rate 4.58 kg⋅hour-1

State P (kPa) T (◦ C) V (m⋅sec-1) h (kJ⋅kg− 1) x υ (m3⋅kg− 1) A (mm2)

1 270.1 130.0 0 2720.5 1.00 0.87 –

200.0 120.2 315.4 2670.3 0.98 0.96 3.52
180.0 116.9 366.4 2652.9 0.98 1.06 3.32
2 160.0 113.3 417.2 2633.0 0.97 1.07 3.23
150.0 111.4 438.9 2623.7 0.97 1.19 3.26
100.0 99.6 566.3 2559.7 0.95 2.98 3.62
50.0 81.3 722.9 2458.7 0.92 12.62 5.24
3 0.725 2.4 1220.0 1975.8 0.79 138.06 144.06

9
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Table 3
Calculation results for the secondary fluid at primary exit plane.
Evaporator saturation temperature 7.5 ◦ C
Evaporator saturation pressure 1.036 kPa
Primary nozzle exit pressure 0.725 kPa
Secondary vapour mass flow rate 1.19 kg⋅hour-1

State P T V h x υ A
(kPa) (◦ C) (m⋅sec- (kJ⋅kg− 1) (m3⋅kg− 1) (mm2)
1
)

4 1.036 7.5 0 2515.0 1.00 124.90 –

5 0.725 2.4 9.5 2469.9 0.99 172.76 602.1

Fig. 13. Cross sectional view of the mixing chamber at the primary nozzle
exit plane.

Table 4
Calculation results for the mixing process between the two fluid streams inside
the mixing chamber.
State P T V h x υ A
(kPa) (◦ C) (m⋅sec- (kJ⋅kg− 1) (m3⋅kg− 1) (mm2)
1
)

3 0.725 2.4 1220 1975.8 0.79 138.06 –

5 9.5 2469.9 0.99 172.76 –
6 873 2287.2 0.91 159.9 295.7

30–50 m⋅sec− 1, otherwise, the diffuser may be too large. However, the
diffuser is designed to have an exit cross-sectional area of approximately
5 times greater than that at the inlet cross-sectional area (A8 ≈ 5⋅A7).
This is to provide a suitable size for practical use. Table 6 shows the
calculation results for the process through the subsonic diffuser.

5. Validation with experimental results

In a previous section (Section 3), the proposed equation based on

Stoecker’s model is employed for designing the ejector’s geometries. The
procedure to determine the flow area (ejector geometry) in each
component is also provided. The ejector geometries are designed asso­
ciated with the input data which is available from the authors’ previous
work. However, to ensure that the proposed ejector model can effi­
ciently be used to design the ejector or to predict the performance under
various working conditions, validation of the results is essential. This
Fig. 12. Calculation flowchart of the process in the mixing chamber to sub­
sonic diffuser. will be presented in this section.
In this validation, the designed ejector’s geometries based on
Stoecker’s model and those based on a Keenan’s model are both vali­
pressure recovery process is achieved when travelling through the
dated with the experimental data at identical working conditions. The
subsonic diffuser. This will be presented later.
Keenan’s model for designing the steam ejector was explained clearly by
The diffuser’s efficiency is usually around 90 to 95 % which is rec­
Ruangtrakoon and Aphornratana [6].
ommended by Borgnakke et al. [21]. In this paper, the diffuser’s effi­
The mass entrainment ratio and critical condenser pressure, which
ciency of 90% is used throughout. It is also assumed that the flow is
are obtained experimentally, the primary nozzle, and mixing chamber
brought to stagnation state at the diffuser exit. In practice, the diffuser
throat diameter are compared with those obtained from both models.
must be designed to discharge the mixed fluid with velocity of around

10
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Table 5
Calculation results for the normal shock.
State Fanno line Rayleigh line
-1 3 − 1 − 1 − 1
V (m⋅sec ) υ (m ⋅kg ) h (kJ⋅kg-1) T ( C)
◦
S (kJ⋅kg ⋅K ) υ (m3⋅kg− 1) P (kPa) T (◦ C) s (kJ⋅kg− 1⋅K− 1)

6 873 159.90 2287.2 2.4 8.278 159.90 0.7252 2.4 8.278

600 109.87 2488.4 9.3 8.810 109.87 2.2172 254.7 9.813
400 73.25 2588.4 46.7 8.899 73.25 3.3094 252.1 9.619
210 38.45 2646.3 77.7 8.732 38.45 4.3469 89.3 8.779
205 37.54 2647.4 78.3 8.723 37.54 4.3742 82.9 8.742
7 202 36.99 2648.0 79.0 8.720 36.99 4.3906 79.0 8.720
200 36.62 2648.4 78.8 8.714 36.62 4.4015 76.5 8.705
100 18.31 2663.4 87.1 8.427 18.31 4.9476 32.7 5.570

diameter of 19 mm is equipped with a primary nozzle with throat

diameter of 2.0 mm and 2.3 mm, respectively. The mixing chamber with
throat diameter of 13.4 mm is equipped with the primary nozzle with
throat diameter of 1.4 mm and 1.7 mm. The boiler temperature is varied
ranging from 115 to 135 ◦ C with increments of 5 ◦ C.
Table 9, and Figs. 15 and 16 show comparisons of the calculated
values of the primary nozzle throat diameter and the mixing chamber
throat diameter with those obtained experimentally. For the case of the
primary nozzle throat diameter, both models show good agreement with
the actual nozzle sizes used in the experiment. The average percent error
of the Stoecker’s model and the Keenan’s model are 2.13% and 1.73%,
respectively. In this case, the results based on the Keenan’s model are
slightly more accurate than the Stoecker’s model. However, the method
used to measure nozzle throat diameter is limited because a drill bit set
with incremental diameter of 0.1 mm was used to measure the primary
nozzle throat diameter. Therefore, the actual size of the primary nozzle
Fig. 14. Plot of calculated Fanno line and calculated Rayleigh line on an Ent­ comes with an uncertainty of ± 0.1 mm or approximately ± 5% to ± 7%
halpy–Entropy chart.
of the nozzle throat diameter which is greater than the percentage error
obtained from both models. Hence, both models could be used as an
Table 6
efficient tool for designing the primary nozzle.
Calculation results for the subsonic diffuser. For the mixing chamber throat diameter, both models provide almost
the same error. The average percent error for the Stoecker’s model and
state P T V h x (m3⋅kg− 1) A
(kPa) (◦ C) (m⋅sec- (kJ⋅kg− 1) (mm2)
the Keenan’s model are 4.2% and 4.04%, respectively. The Keenan’s
1
) model provides a slightly smaller mixing chamber throat diameter size
for all cases. However, the difference between the two models is not
7 4.391 79.0 202 2648.0 1.00 36.99 293.7
8 4.890 89.9 0 2668.4 1.00 34.21 – significant. Thus, for designing the ejector, both models are useful tools
for determining the primary nozzle throat diameter and mixing chamber
throat diameter.
Table 10 and Fig. 17 show comparisons of the calculated critical
Table 7
Dimension of the mixing chamber referring to Fig. 3.
condenser pressures with those obtained experimentally. It is seen that
the Stoecker’s model provides a slightly more accurate result than the
Mixing D5 D6 D8 L1 L2 L3
Keenan’s model with average percent error of 6.98% and 8.54%,
chamber (mm) (mm) (mm) (mm) (mm) (mm)
respectively. In addition, at a relatively high critical condenser pressure,
19 24 19 40 130 114 180
both models provide lower critical condenser than the experimental
13.4 17 13.4 38 110 70 220
values. It is noted that the Keenan’s model has more deviation than the
Stoecker’s model. It is thought that the mixing efficiency may be varied
by using different upstream conditions for achieving more accurate re­
Table 8 sults. However, the mixing efficiency of 90% is used for calculation
Dimension primary nozzles.
throughout this present work. Overall, the difference of calculated and
Nozzle Throat diameter, Nozzle’s area ratio (Aexit: Calculated exit Mach experimental result is acceptable.
mm Athroat) number Validation of the results shows that both models can be used to
D1.4 1.4 20:1 4 design the ejector geometries and efficiently predict the ejector perfor­
D1.7 1.7 mance. However, the calculated result of the flow state of the working
D2.0 2.0
fluid is significantly different. In the next section, the detailed compar­
D2.3 2.4
ison of the flow state (fluid temperature and pressure) will be discussed
in an attempt to show the advantage of the Stoecker’s model.
The experimental data used for validation are obtained from the au­
thors’ previous work [13]. Tables 7 and 8 show dimensions of the mixing 5. Comparisons between the Stoecker’s model and the Keenan’s
chamber and primary nozzles used for experiments. All of the primary model
nozzles have identical calculated exit Mach numbers of 4. The calculated
exit Mach number is obtained by assuming that the fluid is an ideal gas In this section, the ejector designs based on the Stoecker’s model and
(Keenan’s model). Two mixing chambers with throat diameter of 19 mm Keenan’s model are both discussed while concentrating on the flow
and 13.4 mm are used for validation. The mixing chamber with throat process within the ejector. The main purpose is to demonstrate an

11
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Table 9
Calculated and experimental primary nozzle and mixing chamber throat diameters.
Evaporator Temp. (◦ C) Boiler Temp. (◦ C) Nozzle Throat diameter (mm) Mixing chamber throat diameter Error (%)
(mm)

Stoecker Keenan Experiment Stoecker Keenan Experiment Nozzle Throat Mixing chamber throat
diameter diameter

Stoecker Keenan Stoecker Keenan

7.5 115 2.037 2.001 2 17.8 17.05 19 1.85 0.05 6.32 10.26
120 2.019 1.979 18.3 17.65 0.95 1.05 3.68 7.11
125 2.036 2.003 18.8 18.2 1.80 0.15 1.05 4.21
130 2.029 1.997 19.3 18.77 1.45 0.15 1.58 1.21
135 1.985 1.952 19.4 18.9 0.75 2.40 2.11 0.53
115 2.388 2.345 2.3 19.1 18.35 19 3.83 1.96 0.53 3.42
120 2.406 2.358 19.5 18.85 4.61 2.52 2.63 0.79
125 2.378 2.339 19.7 19.05 3.39 1.70 3.68 0.26
130 2.374 2.336 20.3 19.62 3.22 1.57 6.84 3.26
135 2.372 2.333 20.7 20 3.13 1.43 8.95 5.26
115 1.555 1.558 1.5 13 12.5 13.4 3.67 3.87 2.99 6.72
120 1.54 1.539 13.6 13.1 2.67 2.60 1.49 2.24
125 1.547 1.552 14.2 13.7 3.13 3.47 5.97 2.24
130 1.528 1.532 14.3 13.9 1.87 2.13 6.72 3.73
135 1.532 1.536 14.5 14.1 2.13 2.40 8.21 5.22
115 1.714 1.717 1.7 13.5 13 13.4 0.82 1.00 0.75 2.99
120 1.722 1.721 14 13.5 1.29 1.24 4.48 0.75
125 1.703 1.708 14.4 14 0.18 0.47 7.46 4.48
130 1.728 1.733 14.9 14.4 1.65 1.94 11.19 7.46
135 1.719 1.724 15 14.6 1.12 1.41 11.94 8.96

10 115 2.037 2.001 2 18.2 17.27 19 1.85 0.05 4.21 9.11

120 2.019 1.979 18.2 17.45 0.95 1.05 4.21 8.16
125 2.036 2.003 18.6 17.87 1.80 0.15 2.11 5.95
130 2.029 1.997 19.2 18.43 1.45 0.15 1.05 3.00
135 1.985 1.952 19.3 18.77 0.75 2.40 1.58 1.21
115 1.555 1.558 1.5 13.4 12.8 13.4 3.67 3.87 0.00 4.48
120 1.54 1.539 13.5 12.9 2.67 2.60 0.75 3.73
125 1.547 1.552 13.8 13.3 3.13 3.47 2.99 0.75
130 1.528 1.532 14 13.5 1.87 2.13 4.48 0.75
135 1.532 1.536 14.2 13.8 2.13 2.40 5.97 2.99
Average % error 2.13 1.73 4.20 4.04

Error (%) = 100 × (Calculated – Experimental)/Experimental.

Fig. 15. Calculated primary nozzle throat diameter. Fig. 16. Calculated mixing chamber throat diameter.

advantage of using the Stoecker’s model over the Keenan’s model. The primary nozzle throat (quite accurately as mentioned in previous sec­
temperature, pressure and velocity of the working fluid at the points of tion). However, it is obvious that the temperature calculated from
interest are calculated and compared with the experimental data. In Keenan’s model is much different from those of Stoecker’s model. At the
comparison, the results obtained from using mixing chamber 19 mm primary nozzle throat (state 2), the calculated temperature and pressure
with nozzle D2.0 tested with boiler temperature of 130 ◦ C and evapo­ from the Keenan’s model are 73.4 ◦ C and 141.0 kPa, respectively, while
rator of 7.5 ◦ C are used as the reference case. The results are presented in those calculated by the Stoecker’s model are 113.1 ◦ C, 159.0 kPa and
Table 11. The flow process within the ejector is shown on temperature- with dryness quality of 0.97, respectively. At the pressure and temper­
entropy charts which are shown in Figs. 18 and 19, respectively. ature obtained from Keenan’s model, the state of the water is in a sub­
It can be seen from Table 11, Figs. 18 and 19 that at the primary cooled liquid phase. However, for the pressure and temperature values
nozzle where the temperature, pressure and critical mass flow rate of the obtained from the Stoecker’s model, the state of the water is a two-phase
primary steam are all specified, both the Stoecker’s and Keenan’s models mixture with a relatively high dryness quality.
can efficiently be used to calculate the cross-sectional area (2) of the

12
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Table 10
Calculated and experimental critical condenser pressure.
Nozzle Evaporator Temp. (◦ C) Boiler Temp. (◦ C) Critical condenser pressure (mbar) Error (%)

Stoecker Keenan Experiment Stoecker Keenan

2.0 7.5 115 3.65 3.59 3.2 14.06 12.19

120 3.94 3.81 3.7 6.49 2.97
125 4.5 4.2 4.2 7.14 0.00
130 4.89 4.52 4.8 1.88 5.83
135 5.41 4.88 5.6 3.39 12.86
2.3 115 4.3 4.09 4.4 2.27 7.05
120 4.89 4.51 5.2 5.96 13.27
125 5.5 5 6 8.33 16.67
130 6.02 5.42 6.8 11.47 20.29
135 6.61 5.91 7.8 15.26 24.23
1.4 115 3.53 3.67 3.4 3.82 7.94
120 4.02 4.02 3.9 3.08 3.08
125 4.62 4.47 4.4 5.00 1.59
130 4.93 4.79 5 1.40 4.20
135 5.43 5.07 5.7 4.74 11.05
1.7 115 3.96 3.83 3.2 23.75 19.69
120 4.14 3.97 3.8 8.95 4.47
125 4.54 4.26 4.2 8.10 1.43
130 4.94 4.62 4.8 2.92 3.75
135 5.74 5.18 5.4 6.30 4.07

2.0 10 115 4.4 4.17 4.1 7.32 1.71

120 4.89 4.48 4.8 1.88 6.67
125 5.22 4.81 5.5 5.09 12.55
130 5.9 5.32 6.4 7.81 16.88
135 6.63 5.89 7.1 6.62 17.04
1.4 115 3.76 3.84 3.4 10.59 12.94
120 4.22 4.2 4 5.50 5.00
125 4.8 4.6 4.4 9.09 4.55
130 5.2 4.99 5 4.00 0.20
135 6 5.48 5.6 7.14 2.14
Average % error 6.98 8.54

Error (%) = 100 × (Calculated – Experimental)/Experimental.

This is the result of the formation of the expansion wave which causes
the jet stream to further expand with the series of the combined shock-
expansion fan. Therefore, the velocity of the primary stream is accel­
erated further after leaving the primary nozzle. This causes the pressure
within the mixing chamber to drop lower than the primary nozzle exit
pressure. In other words, the real value of the mixing chamber pressure
is the result of the expansion wave formation which is lower than the
nozzle exit pressure. As a result of this effect, the calculated primary
nozzle exit area (using both Keenan’s and Stoecker’s model) is larger
than that used for the experiment.
At the primary nozzle exit (3), the exit temperature obtained from
the Keenan’s model is − 149.2 ◦ C which is well below the freezing
temperature of water. On the other hand, the temperature calculated by
the Stoecker’s model is 2.4 ◦ C, which is the saturation temperature at
0.725 kPa. These two values are much different. The value obtained
Fig. 17. Plot of experimental and calculated critical condenser pressure. from the Keenan’s model seems to be impossible since the temperature is
well below the freezing point; however, in reality, it is possible that the
From the Stoecker’s model, at state 3, the velocity of the primary fluid is still in vapour phase at temperature below freezing point. This is
stream must be accelerated to around 1220 m⋅sec-1 so that the exit because the velocity of the fluid at this point is very high, therefore there
pressure of 0.725 kPa is produced. Hence, the nozzle exit diameter must is not enough time for the fluid to be condensed [18] or crystalized [22].
be as large as 13.5 mm so that it is consistent with the exit pressure. On the other hand, the results obtained from the Stoecker’s model is
However, the nozzle exit diameter used for the experiment is 8.9 mm (as under the fluid thermodynamic properties at equilibrium state. These
shown in Table 11) which is quite a large difference. Theoretically, when thermodynamic properties are realistic for most of the time. However,
the exit diameter of 8.9 mm is used for this calculation, the exit velocity there are some extreme working conditions in which the fluid is not
of 1220 m⋅sec-1 and pressure of 0.725 kPa cannot be achieved. However, based on these thermodynamic properties, e.g., supersaturation condi­
for the experiment, the pressure of 0.725 kPa can still be produced tion [18], and super cooling condition [22]. Therefore, it cannot be
within the mixing chamber even when the nozzle exit diameter of 8.9 concluded which model is more correct because there is no experiment
mm is used. This is the result of further expansion of the primary jet proof.
stream after leaving the primary nozzle. This phenomenon can be clar­ The mixing process of both the Keenan’s and Stoecker’s model is
ified by the CFD simulation provided by Ruangtrakoon, et al. and Sri­ assumed to occur under the constant pressure conditions. At the mixing
veerakul, et al. [7,8] which is shown in Fig. 20. It is obvious that there is chamber throat or the upstream of the normal shock (state 6), the
a further expansion of the jet stream after leaving the primary nozzle. pressure is maintained constant at 0.725 kPa. Based on the Stoecker’s

13
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Table 11 (upstream of the normal shock, 6) to superheated vapour (downstream

Comparison between the calculated results obtained from the Stoecker’s model of the normal shock, 7).
and from the Keenan’s model. The subsonic mixed fluid (7) is then further slowed down to almost a
Boiler saturation temperature 130 ◦ C stagnation state as it travels through the subsonic diffuser (8). This
Boiler saturation pressure 270.1 kPa causes further pressure recovery. The pressure raised via the subsonic
Critical mass flow rate 4.58 kg⋅hour-1 diffuser is around 10% of the total pressure recovery process.
Evaporator saturation temperature 7.5 ◦ C
At state 7 and 8, the results obtained from both models are similar.
Evaporator saturation pressure 1.036 kPa
Secondary vapour mass flow rate 1.19 kg⋅hour-1 This is because the fluid at this state is the superheated steam based on
Primary nozzle exit pressure 0.725 kPa the Stoecker’s model of which behaviour is close to ideal gas. This may
Primary nozzle efficiency 90 % imply that for the dry fluid in which the expansion process occurs under
Mixing efficiency 90 %
the superheated region, the ideal gas assumption is acceptable [23].
Subsonic diffuser efficiency 90 %
In practice, if the mixed fluid is brought to stagnation state, a very
Stoecker Keenan Experiment large exit cross sectional area and very long length of diffuser are
Primary Nozzle required. Therefore, it is not suitable for practical use. Hence, the
Throat (2) T (◦ C) 113.1 73.4 calculation of the pressure recovery process of the mixed fluid is slowed
P (kpa) 159.0 141.0
down to the velocity around 30 to 50 m⋅sec-1 so that the proper geometry
V (m⋅s− 1) 420 460
υ (m3⋅kg− 1) 1.07 1.13 of the subsonic diffuser is achieved. From the calculations, the ejector
x 0.97 – discharge pressures calculated from both models are very close to those
A (mm2) 3.23 3.13 3.14 obtained experimentally as can be seen in the validation section.
D (mm) 2.03 2.00 2.00 Figs. 19 and 20 obviously show the difference of the flow state ob­
Primary Nozzle tained from both models. From Fig. 19, which shows the flow process of
Exit (3) T (◦ C) 2.4 − 149.2 the Stoecker’s model on the temperature-entropy chart, it can clearly be
P (kpa) 0.725 0.725
seen that all the processes upstream of the normal shock occur in a two-
V (m⋅s− 1) 1220 1022
υ (m3⋅kg− 1) 138.06 78.77 phase region. This is different from that of the flow process based on the
x 0.79 – Keenan’s model as shown in Fig. 20.
A (mm2) 144 98.1 62 Based on the Stoecker’s model, it can also be seen that temperatures
D (mm) 13.5 11.1 8.9 of the steam-water at any state in the ejector are higher than the freezing
Normal Shock temperature. This is in contrast to that obtained based on the Keenan’s
Upstream (6) T (◦ C) 2.4 − 60.1 model in which the minimum temperature of − 149.2 ◦ C is found (at the
P (kpa) 0.725 0.725
primary nozzle exit (3)). Moreover, the mixing process between the
V (m⋅s− 1) 873 785.5
υ (m3⋅kg− 1) 159.9 135.49
primary and secondary fluid occurs at a temperature far below the
x 0.91 – freezing point of the water.
A (mm2) 293 277 283 Another interesting point is that, for the Stoecker’s model, the mix­
D (mm) 19.3 18.8 19.0 ing process upstream of the normal shock occurs at a constant pressure
Normal Shock and also constant temperature. This is due to the fact that both primary
Downstream (7) T (◦ C) 78.8 87.55 fluid and secondary fluid are in a two-phase region (saturates liquid­
P (kpa) 4.39 3.81 –vapour mixture condition). On the other hand, for the Keenan’s model,
V (m⋅s− 1) 202 253
υ (m3⋅kg− 1) 36.99 43.65
only the pressure is maintained constant. Even though the velocity and
x Superheated – pressure obtained from both models are not much different, the tem­
A (mm2) 293 277 284 perature obtained from the Keenan’s model is much lower than that
D6 (mm) 19.3 18.8 19 obtained from the Stoecker’s model. However, it cannot be ensured
Discharge (8) T (◦ C) 89.61 104.72
which model yields a more accurate result or reflects a more realistic
P (kpa) 4.89 4.51 4.8
V (m⋅s− 1) 0 0 process. The ideal gas assumption based on the Keenan’s model may not
υ (m3⋅kg− 1) 34.21 38.57 be suitable for the wet fluid because the expansion process is in liquid­
x Superheated – –vapour mixture region because the fluid’s behaviour is far from ideal
gas. On the other hand, the limitation of Stoecker’s model is that all
processes are under the equilibrium state and the fluid properties are
model, the mixed fluid is the wet steam with dryness quality of 0.91 and
associated with the common thermodynamic properties. In reality, there
the temperature is 2.4 ◦ C (saturation temperature at 0.725 kPa).
is some extreme condition at which the process is under the non-
Meanwhile, for the Keenan’s model, temperature of the mixed fluid is
equilibrium state (the thermodynamic properties are not available).
− 60.1 ◦ C, which is below the freezing temperature of water. Even
Hence, the flow process inside the ejector obtained from these two
though the state of water obtained from the two models is significantly
models is significantly different as shown in Figs. 19 and 20. To clarify
different, the calculated velocity of the mixed fluid obtained from both
this, the calculated temperatures from both models need to be validated
models is not much different. Due to the higher specific volume for the
with the experimental result.
Stoecker’s model, the cross-sectional area of the ejector throat is slightly
Fig. 21 shows the performance comparison based on the perfor­
larger than that obtained from the Keenan’s model. However, the
mance map of the experimental result with those obtained from both
calculated cross-sectional area obtained from both models is very similar
models. In Fig. 21a, it can obviously be seen that the at higher boiler
to that of the experimental ejector used.
saturation temperature, the calculated result obtained from the Keen­
Across the normal shock, the velocity of the mixed fluid stream drops
an’s model is largely different from the experimental result. On the other
suddenly from supersonic to subsonic value. Thus, the temperature,
hand, the calculated result obtained from the Stoecker’s model is closer
pressure and density of the mixed stream increase sharply. This means
to the experimental result. This shows that the Stoecker’s model can be
the normal shock wave creates a compression effect or pressure recovery
used to predict the performance of the ejector for a wider range of the
process. It is believed that the pressure recovery via the normal shock
working condition. However, the Keenan’s model might not be suitable
wave is around 90% of the overall pressure recovery process (the
for high boiler saturation temperature. This shows the benefit of using
remaining 10% is obtained via subsonic diffuser). According to the flow
the Stoecker’s model over the Keenan’s model which is useful for
state, the mixed fluid stream is changed from a two-phase mixture
developing the modern design theory of the steam ejector.

14
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Fig. 18. Processes on Temperature-Entropy chart based on Stoecker’s model.

Fig. 19. Processes on Temperature-Entropy chart based on Keenan’s model.

15
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

Fig. 20. Expansion wave form CFD simulation by Ruangtrakoon, et al. [7].

Fig. 21. Performance for the tested ejector calculated from the data obtained based on experiments, Keenan model, and Stoecker model.

16
B. Kitrattana et al. Thermal Science and Engineering Progress 25 (2021) 101016

• The temperature and flow pressure based on the Stoecker’s models is

associated with the liquid–vapour mixture and superheated steam at
equilibrium state.

However, even though both models show a significant difference in

the temperature and flow process inside the ejector, it cannot be ensured
which models is more accurate. The experimental proof of the temper­
ature measurement inside the ejector is needed for validation.

Declaration of Competing Interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence
the work reported in this paper.
Fig. 22. Plots of calculated critical mass flow rate through two different pri­
mary nozzles and compared with the data obtained experimentally.
Acknowledgements
Normally, the Keenan’s model is widely used to calculate the critical
mass flow rate of saturated steam through the primary nozzle. This is The first author would like to thank to Sirindhorn International
because the calculation method is relatively easy and provides reason­ Institute of Technology for funding support for this study (Scholarship
ably accurate results. Fig. 22 shows the experimental and calculated Number: ETS-G-S1Y17/064).
critical mass flow rate of saturated steam through two different size
nozzles. It can be seen that the critical mass flow rate obtained from both References
the Keenan’s and Stoecker’s model is in good agreement with the
[1] S. Aphornratana, I.W. Eames, A small capacity steam-ejector refrigerator:
experimental results. This comparison indicates that both models can be
experimental investigation of a system using ejector with movable primary nozzle,
used as the efficient tool for predicting the critical mass flow rate of the Int. J. Refrig 20 (5) (1997) 352–358.
steam through the primary nozzle. [2] N. Ruangtrakoon, S. Aphornratana, T. Sriveerakul, Experimental studies of a steam
The calculated result shows that the ideal gas assumption for steam jet refrigeration cycle: effect of the primary nozzle geometries to system
performance, Exp. Therm Fluid Sci. 35 (4) (2011) 676–683.
provides relatively low temperature compared to the real fluid proper­ [3] Keenan, J. H. (1950). An investigation of ejector design by analysis and
ties. Moreover, the flow process from these two models is significantly experiment. Journal of Applied Mechanics, 17, 299.
different as shown in Figs. 19 and 20. However, the calculated geome­ [4] J.T. Munday, D.F. Bagster, A new ejector theory applied to steam jet refrigeration,
Ind. Eng. Chem. Process Des. Dev. 16 (4) (1977) 442–449.
tries from these two models is not much different. This is because the [5] B.J. Huang, J.M. Chang, C.P. Wang, V.A. Petrenko, A 1-D analysis of ejector
pressure at the nozzle exit is assumed to be 70% of the evaporator performance, Int. J. Refrig. 22 (5) (1999) 354–364.
pressure and is being kept constant throughout the mixing process for [6] N. Ruangtrakoon, S. Aphornratana, Design of steam ejector in a refrigeration
application based on thermodynamic performance analysis, Sustain. Energy
both models. This causes the specific volume determined from both Technol. Assess. 31 (2019) 369–382.
models to be not much different. Therefore, the calculated flow area [7] N. Ruangtrakoon, T. Thongtip, S. Aphornratana, T. Sriveerakul, CFD simulation on
from both models is not much different. Across the shock wave, the state the effect of primary nozzle geometries for a steam ejector in refrigeration cycle,
Int. J. Therm. Sci. 63 (2013) 133–145.
of the fluid is the superheated vapour based on the Stoecker’s model. [8] T. Sriveerakul, S. Aphornratana, K. Chunnanond, Performance prediction of steam
Hence, the fluid’s behaviour is close to the ideal gas. As a result, the ejector using computational fluid dynamics: Part 1. Validation of the CFD results,
process at the downstream of the shock wave obtained from two such Int. J. Therm. Sci. 46 (8) (2007) 812–822.
[9] T. Thongtip, S. Aphornratana, Development and performance of a heat driven
models are not much different and, therefore, the critical condenser
R141b ejector air conditioner: Application in hot climate country, Energy 160
pressure is not much different. (2018) 556–572.
[10] J. Yan, S. Li, Z. Liu, Numerical investigation on optimization of ejector primary
6. Conclusions nozzle geometries with fixed/varied nozzle exit position, Appl. Therm. Eng. 175
(2020) 115426, https://doi.org/10.1016/j.applthermaleng.2020.115426.
[11] W.F. Stoecker, Steam-jet refrigeration, Refrig. Air Condition. (1958) 194–205.
From this study, the Keenan’s model and Stoecker’s model are used [12] Fauske, H. K. (1962). Contribution to the theory of two-phase, one-component
for analyse the steam ejector. The Keenan’s model uses ideal gas critical flow (No. ANL-6633). Argonne National Lab., Ill..
[13] B. Kitrattana, S. Aphornratana, T. Thongtip, The performance of steam ejector
assumption, while the Stoecker’s model uses real fluid properties. These refrigeration system based on alternative analysis, Energy Procedia 138 (2017)
two models have been modified to be more precise for designing ejector 482–487.
geometry and analysing the flow process. The calculated results from [14] P. Whalley, Boiling, Condensation, and Gas-Liquid Flow, Clarendon, Oxford, 1990.
[15] S.M. Zivi, Estimation of steady-state steam void-fraction by means of the principle
these two models show both differences and similarities. It can be of minimum entropy production, J. Heat Trans. 86 (2) (1964) 247–251.
concluded as follows: [16] Chisholm, D. (1972). Equation for velocity ratio in two-phase flow. Nat. Eng. Lab.
Rep.(U. K.), (535), 13.
[17] Premoli, A. (1970). An empirical correlation for evaluating two-phase mixture
• The validation result shows that both models can be used to design density under adiabatic conditions. In European Two-Phase Flow Group Meeting,
ejector geometries and predict the ejector performance (in term of 1970.
entrainment ratio and critical condenser pressure) [18] Sarkar, B. K. (1998). Thermal engineering. New Delhi: TATA McGraw-Hill.
[19] VanWylen, G. J., Sonntag, R. E., & Borgnakke, C. (1994). CATT: Computer-aided
• The calculated pressures at each state from both models are not much
thermodynamic tables. Fundamentals of classical thermodynamics.
different. [20] I.W. Eames, S. Aphornratana, H. Haider, A theoretical andexperimental study of a
• The Stoecker’s model shows that condensation occurs in the ejector, small-scale steam jet refrigerator, Int. J. Refrig. 18 (6) (1995) 378–386.
so an assumption of ideal gas might not be valid. [21] Borgnakke, Claus, et al. Fundamentals of Classical Thermodynamics. Wiley, 1994.
[22] F.C. Frank, Supercooling of liquids, Proc. R. Soc. Lond. A 215 (1120) (1952) 43–46.
• The calculated temperature from the Keenan’s model is much lower [23] J.A. White, S. Velasco, Characterizing wet and dry fluids in temperature-entropy
than the freezing point of water. diagrams, Energy 154 (2018) 269–276.

17