You are on page 1of 11

Applied Thermal Engineering 115 (2017) 444–454

Contents lists available at ScienceDirect

Applied Thermal Engineering


journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

Ejector performance prediction at critical and subcritical operational


modes
Fenglei Li a,⇑, Qi Tian a, Changzhi Wu b, Xiangyu Wang b, Jae-Myung Lee c
a
College of Environmental Science and Engineering, Taiyuan University of Technology, Taiyuan, Shangxi 030024, China
b
Australasian Joint Research Centre for Building Information Modelling, School of Built Environment, Curtin University, Perth, WA 6845, Australia
c
Department of Naval Architecture and Ocean Engineering, Pusan National University, Busan, Republic of Korea

h i g h l i g h t s

 Novel models for ejector performance analysis at critical and breakdown points are developed.
 The models are integrated to predict ejector performance over entire operational range.
 An analysis method is developed for identifying the ejector component efficiencies.
 The performance prediction errors over entire operational range are within ±10%.

a r t i c l e i n f o a b s t r a c t

Article history: Traditional ejector models are focusing on the ejector performance predictions at critical mode under
Received 27 September 2016 design conditions. In reality, ejector systems cannot be operated under these conditions perfectly.
Revised 18 December 2016 Thus, the study of ejector performance at subcritical mode under off-design conditions is important. In
Accepted 23 December 2016
this paper, novel models for ejector performance predictions at critical point and breakdown point are
Available online 29 December 2016
developed based on constant-pressure mixing and constant-pressure disturbing assumptions. Then,
the two models are integrated as the model to predict ejector performance at critical and subcritical oper-
Keywords:
ational modes. In order to determine the ejector component efficiencies in the models, a novel concept,
Ejector performance prediction
Critical mode
the effect of the change (EOC) of efficiency, is introduced to identify the efficiencies which affect ejector
Subcritical mode performance significantly. Then, the identified efficiencies are determined by sparsity-enhanced opti-
Efficiency mization method. The predicted results obtained by our model are much more accurate than those
obtained by existing methods.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction mix at a uniform pressure without heat and friction losses in their
model. Munday and Bagster [4] defined a hypothetical throat of the
Ejector refrigeration systems (ERSs) driven by low-grade energy secondary flow in their model. The theoretical results of the model
is an attractive solution to the reduction of electricity consumption can agree with the plant-scale tests data. Eames et al. [5] modified
in air-conditioning system [1]. The ejector is the key component of Keenan’s model [2] by using isentropic efficiencies to consider irre-
an ERS, and plays a significant role for the performance of ERS. versibility of the process inside the ejector. Their theoretical results
Ejectors can be classified into two types: constant-area mixing are within 85% of the experimental data derived from a small-scale
ejectors, and constant-pressure mixing ejectors [2]. Generally, the steam jet refrigerator. Huang et al. [6] introduced four component
performance of a constant-pressure mixing ejector is superior to efficiencies in their 1-D ejector model to account for the losses
that of a constant-area mixing ejector as claimed in [2,3]. For this inside ejector. For working fluid R141b, the maximum error of ana-
reason, extensive studies focus on developing theoretical models lytical results is 22.99% in comparison with their experiment data.
of the constant-pressure mixing ejectors [3]. Zhu and Li [7] adopted a simple linear function to approximate the
Keenan [2] proposed a model for constant-pressure mixing real velocity distribution inside the ejector. In the meanwhile, they
ejector. It was assumed that the primary and secondary flows took the expansion efficiency, /e , as the equation of operating
parameters. Compared with the experimental data in [8,6], the
⇑ Corresponding author. Root-Mean-Square of relative error and the maximum error of
E-mail address: lifenglei@tyut.edu.cn (F. Li). entrainment ratio are 8.41% and 13.8%, respectively. Cardemil

http://dx.doi.org/10.1016/j.applthermaleng.2016.12.116
1359-4311/Ó 2016 Elsevier Ltd. All rights reserved.
F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454 445

Nomenclature

A area (m2) e expansion, evaporation


Cp specific heat at constant pressure (J/kg K) exp experiment
Cv specific heat at constant volume (J/kg K) m mixing flow
d diameter (m) p primary flow, condensing pressure
m mass flowrate (kg/s) p0 primary flow at inlet of ejector
M Mach number p1 primary flow at nozzle exit
P pressure (Pa) pb breakdown condensing pressure
R gas constant (J/kg K) r ratio
T temperature (K) s secondary flow
u entrainment ratio s0 secondary flow at inlet of ejector
V velocity (m/s) t throttle
g efficiency relating to isentropic efficiency u critical entrainment ratio
/ efficiency account for losses y position of the hypothetical throat
c ¼ C p =C v
Abbreviations
Subscripts EOC the effect of the change (of efficiency)
b breakdown ERS ejector refrigeration system
c to condenser, critical
d diffuser

and Colle [9] proposed an ejector model for both dry and wet vapor the systemic methods of analyzing the effect of them to select and
working fluids. The calculation method of the sound speed in two- optimize them.
phase mixtures, the real gas equation and the correlation equations Optimization is a useful tool to find the relationship underneath
of the ejector component efficiencies, /e and /m , were introduced [21]. For example, firefly algorithm is used to schedule scaffolding
in their model. In comparison with the experiment for R141b, projects [22] and sparsity enhanced optimization is used to
steam and carbon dioxide in [6,10,11], the prediction errors of improve the ejector prediction accuracy [23]. In this paper, we will
entrainment ratios and critical condensing pressures are all within use the sparsity enhanced optimization in the optimization of ejec-
9% and 6%, respectively. Chen et al. [12] developed an ejector tor component efficiencies. The contribution of this paper is two-
model to predict the optimum COP, entrainment ratio and the cor- fold: developing an ejector model that can predict performance
responding ejector area ratio simultaneously. Three ejector compo- over entire operational range and proposing a method to analyze
nent efficiencies were introduced in the model and the calculated the effect of ejector component efficiencies on the ejector
results are in very good agreement with the experimental data at performance. Firstly, novel ejector models at critical point and
optimum conditions. breakdown point are developed based on constant-pressure
All the aforementioned models are only suitable to the critical mixing and constant-pressure disturbing assumptions, respec-
mode of ejector under design conditions. However, ERSs can be tively. Then, the two models integrated as the model for predicting
operated at subcritical mode under off-design conditions. Better ejector performance over entire operational range. Furthermore,
control and regulation of an ERS require the ejector model for the effect of the change (EOC) of efficiency is presented and
entire operational regime, which consists of the models at critical employed to identify the ejector component efficiencies which
and subcritical modes. Therefore, developing ejector models for have more influence on the validity of the model. The identified
subcritical mode is important. Ouzzane and Aidoun [13] proposed efficiencies are utilized to develop the model through our recently
a 1-D ejector model for entire operational regime based on the developed sparsity-enhanced optimization method in [23]. The
assumption that the mixing takes place at uniform pressure corre- prediction results show that the ejector performance predicted
sponding to the critical pressure of the secondary flow. However, by our method is much more accurate than those obtained by
since the secondary flow is not choking at subcritical mode [14], existing methods.
the mixing pressure does not equal to the critical pressure of the
secondary flow. Boumaraf and Lallemand [15] also developed a
1-D model for the prediction of ejector performance for entire 2. Models for the analysis of ejector performance
operational regime, but the assumptions for subcritical mode were
not clearly defined and the validation of the model were not avail- The critical point and the breakdown point are two key opera-
able. Based on the constant-pressure mixing assumption, Chen tional points in the ejector performance. This can be seen in the
et al. [16] carried out a 1-D analysis on the ejector performance typical performance curve of vapor ejectors at fixed primary pres-
over the entire operational range. They have considered the differ- sure Pp0 and secondary pressure Ps0 in Fig. 1. Three operational
ence of mixing pressure between the critical and subcritical mode, modes are divided by the two points:
but the ejector component efficiencies were taken as fixed values.
The prediction errors for subcritical mode were about 20% in com- (1) critical mode where the entrainment ratio is constant and
parison with the experimental data in [16,17]. These existing equals critical entrainment ratio at critical point, while
results show that ejector models at subcritical mode have some P c 6 P cc ;
limitations in physical descriptions, prediction accuracy, etc. Fur- (2) subcritical mode where the entrainment ratio is between the
thermore, since the ejector component efficiencies have dramatic critical entrainment ratio and zero, while P cc < Pc < Pcb ;
influence on the validity of ejector models [18,19], some research- (3) back-flow mode where the entrainment ratio is less than or
ers have done some sensitivity analyses to determine the influence equal to zero, while P c 6 P cb .
of them on theoretical results [20], but few studies are dealing with
446 F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
Critical mode Subcritical mode Back-flow mode Pp0 At c 2 ðcþ1Þ=ðc1Þ pffiffiffiffiffi
mp ¼ pffiffiffiffiffiffiffi gp ð1Þ
(double chocking) (single chocking) T p0 R cþ1
u =constant u <0
where gp is the isentropic efficiency of the primary flow from inlet
Entrainment ratio u

to nozzle throttle. Using isentropic relation as an approximation,


Critical point the following relation of the primary flow can be obtained:
 c=ðc1Þ
Pp0 =P py  1 þ M2py ðc  1Þ=2Þ ð2Þ

Breakdown point Introducing /e to account for the losses of primary flow from
Pcc Pcb inlet to section y-y, then the diameter of the primary flow at sec-
Condensing pressure Pc tion y-y can be expressed as follows:
" #ðcþ1Þ=½4ðc1Þ  1=2
Fig. 1. Operational modes of ejector. /e dpy 2 þ ðc  1ÞM 2py 1
¼ ð3Þ
dt 2 þ ðc  1Þ Mpy

It can be observed that the ejector loses its function completely


at back-flow mode. Thus, the entire operational range actually 2.2.2. Secondary flow from inlet to section y-y
includes critical and subcritical modes. The schematic diagram of The pressure at section y-y can be obtained by using the follow-
ejector used in developing the ejector model is shown in Fig. 2. ing isentropic relation as an approximation:
 c=ð1cÞ
2.1. Assumptions
Ppy ¼ Psy  Ps0 1 þ M2sy ðc  1Þ=2 ð4Þ

The area Asy of the secondary flow at section y-y can be calcu-
To simplify the modeling process, without loss of generality, the
lated as follows:
following conditions are assumed to be satisfied.
Asy ¼ A3  Apy ð5Þ
(1) The vapor working fluid is an ideal gas with constant C p and
The secondary flow is chocked at section y-y. Using isentropic
c; efficiency gs to account for the loss from inlet to section y-y, then
(2) The flow inside the ejector is one dimensional and steady;
the flow rate of the secondary flow is:
(3) The kinetic energy of the primary and secondary flow at the sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
inlets and the mixing flow at the outlet are negligible; Ps0 Asy c 2 ðcþ1Þ=ðc1Þ pffiffiffiffiffi
(4) Constant-pressure mixing. At critical point, after exhausting ms ¼ pffiffiffiffiffiffiffi gs ð6Þ
T s0 R cþ1
from the nozzle, the primary flow induces but does not mix
with the secondary flow before the cross section y-y inside Then, the entrainment ratio can be derived as:
the constant-area section, where the secondary flow is chok-
uc ¼ ms =mp ð7Þ
ing. Then, the two streams start to mix with a uniform pres-
sure (Pm ¼ P py ¼ P sy ) before the shock;
(5) Constant-pressure disturbing. At breakdown point, the pri- 2.2.3. Temperature and velocity at section y-y
mary flow fans out of the nozzle without mixing with the The temperature of the primary flow and the secondary flow at
secondary flow before the cross Section 2. Then, the primary section y-y can be determined as:
flow is disturbed by the secondary flow which mixes in and
T p0 =T py ¼ 1 þ M2py ðc  1Þ=2 ð8Þ
out of the primary flow with a uniform pressure
(Pm ¼ P p2 ¼ P s2 ) before the shock;
(6) The inner wall of the ejector is adiabatic. T s0 =T sy ¼ 1 þ M2sy ðc  1Þ=2 ð9Þ

At section y-y the velocity relations are:


2.2. Ejector performance at critical point qffiffiffiffiffiffiffiffiffiffiffiffi
V py ¼ M py cRT py ð10Þ
2.2.1. Primary flow from inlet to section y-y
qffiffiffiffiffiffiffiffiffiffiffiffi
For a given inlet stagnant pressure Pp0 and temperature T p0 , the
V sy ¼ Msy cRT sy ð11Þ
mass flow rate of the primary flow through the nozzle is expressed
as the gas dynamic equation:

Suction chamber Constant-area section Diffuser

throat

Primary to
p0
flow condensor
t 1
nozzle 2 y m 3
s0
aerodynamic throat c
Secondary
flow

Fig. 2. Schematic diagram of ejector [23].


F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454 447

   
2.2.4. Mixing chamber /mb mp V p2 þ ms V s2 ¼ mp þ ms V m ð21Þ
The energy conservation equation during the mixing process is:
! ! The energy conservation equation during the mixing process
V 2py V 2sy is:
mp C p T py þ
þ ms C p T sy þ ! !
2 2 V 2p2
! V2
mp C p T p2 þ
þ ms C p T s2 þ s2
  V 2 2 2
¼ mp þ ms C p T m þ m ð12Þ !
2   2
V
¼ mp þ ms C p T m þ m ð22Þ
The momentum conservation equation can be expressed as: 2
   
/m mp V py þ ms V sy ¼ mp þ ms V m ð13Þ The mach number at section m-m can be calculated using Eq.
(14). At some section after section m-m, a supersonic shock takes
where /m is the momentum transfer efficiency accounting for the
place with a sharp pressure rising. Assuming that the flow under-
loss during the mixing of the two streams. The velocity of the mixed
goes an isentropic process between section m-m and section 3 and
flow can be calculated as:
has a uniform pressure P3 after the shock inside the constant area,
pffiffiffiffiffiffiffiffiffiffiffiffi
V m ¼ Mm cRT m ð14Þ P3 and M 3 can be derived by Eqs. (15) and (16).

The mixed flow undergoes a supersonic shock process at some 2.3.3. Diffuser
section after section m-m with a sharp pressure rise. The variation Assuming the flow undergoes an isentropic process, the pres-
of the mixed flow across the supersonic shock can be expressed as: sure at the exit of diffuser can be obtained by the following
 
relation:
P3 =Pm ¼ 1 þ ð2c=ðc þ 1ÞÞ M 2m  1 ð15Þ
 c=ðc1Þ
The Mach number M 3 of the mixed flow follows: Pcb =P3 ¼ 1 þ M 23 ðc  1Þ=2 ð23Þ

1 þ ððc  1Þ=2ÞM 2m
M 23 ¼ ð16Þ
cM2m  ðc  1Þ=2 2.4. Prediction of the performance at both critical and subcritical mode

When the condensing pressure P c is not higher than the critical


2.2.5. Diffuser pressure P cc , the ejector is working at critical mode, then
Assuming the mixed flow undergoes an isentropic process, the
u ¼ uc ð24Þ
pressure at the out let of the diffuser can be calculated using the
following relation: If P c is between the critical pressure Pcc and the breakdown
 c=ðc1Þ pressure Pcb , the entrainment ratio u can be obtained by the follow-
Pcc =P3 ¼ 1 þ M 23 ðc  1Þ=2 ð17Þ ing relationship:
u ¼ uc ðPcb  Pc Þ=ðPcb  Pcc Þ ð25Þ
2.3. Ejector performance at breakdown point The ejector performance prediction at critical mode is actually
determined by the critical point model and that at subcritical mode
2.3.1. Primary flow from inlet to section 2 is determined by both the critical point and breakdown point mod-
It is an approximate isentropic process for the primary flow els. In other words, the ejector performance over entire operational
from the inlet to section 2. Then the Mach number M p2 can be range can be predicted by Eqs. (24) and (25) when the uc ; Pcc and
obtained by the following relation: Pcb have been obtained from the ejector models at critical and
 c=ðc1Þ breakdown points. The calculation process of the performance pre-
Pp0 =Pp2  1 þ M 2p2 ðc  1Þ=2Þ ð18Þ diction is illustrated in Fig. 3.
The entrainment ratio equals 0 at breakdown point. It means that
the flow rate of the secondary flow equals 0 and the pressure of 3. Analysis of ejector component efficiencies
the secondary flow at section 2 is equal to Ps0 , i.e., P s2 ¼ Ps0 . Hence,
it can be derived that P p2 ¼ P s0 according to the assumption (5). The Different forms of the ejector component efficiencies are
mass flow rate of the primary flow through the nozzle is the same as adopted in different ejector models. In [24], they are taken as fixed
that at critical point, which can be calculated using Eq. (1). The tem- values. In [9,25], they are expressed as correlation equations of
perature and the velocity of the primary flow at section 2 follow: geometric parameter Ar and operating parameter P r . In fact, not
all the efficiencies have the same influence on the accuracy of an
T p0 =T p2 ¼ 1 þ M 2p2 ðc  1Þ=2 ð19Þ ejector model. The concept of EOC will be proposed and employed
to identify which efficiency can be taken as a fixed number and
qffiffiffiffiffiffiffiffiffiffiffiffi
which one should be expressed as a correlation equation.
V p2 ¼ Mp2 cRT p2 ð20Þ
3.1. The concept of EOC
2.3.2. Mixing chamber
Although the flow rate of the secondary flow equals 0, the sec- For the critical point model, there are four efficiencies, i.e.,
ondary flow has influence on the primary flow as assumption (5). gp ; gs ; /e , and /m , and two prediction parameters, namely, the crit-
In order to consider the effect of the secondary flow, it is assumed ical entrainment ratio uc and the critical condensing pressure Pcc . If
that the secondary flow rate is small enough to satisfy the geometric and operating parameters are given, uc and Pcc can
ms ¼ 1e  6 kg/s. be expressed as the functions of gp ; gs ; /e , and /m , i.e.,
Using /mb to account for the losses from section 2 to section m-
uc ¼ uðgp ; gs ; /e ; /m Þ; Pcc ¼ Pðgp ; gs ; /e ; /m Þ: ð26Þ
m, the momentum balance conservation relation can be expressed
as:
448 F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454

Similarity, denote
Start
EOCpb ¼ ðPð1Þ  Pð0:5ÞÞ=Pcb;exp ð33Þ

Input operating and geometric parameters of


3.2. EOC of the efficiencies on ejector performance
an ejector, Pp0, Ps0, Tp0, Ts0, At, A3
For the given Ar and P r , EOC is the relative variation of a predic-
tion parameter, such as uc ; P cc or P cb , to experimental result where
Calculate the critical entrainment ratio uc and the efficiency varies from 0.5 to 1 while the other efficiencies equal
1. The change of an efficiency can induce the variation of the ejec-
critical condensation pressure Pcc
tor performance. For example, when gp increases, the pressure of
mixed flow will increase since mp increases and ms remains
unchanged based on the ejector model. Then, the benefit of gp
Calculate the breakdown condensing pressure increasing is the rise in critical condensing pressure P cc at constant
ms although the critical entrainment uc decreases. Thus, EOC can
Pcb
indicate whether a prediction parameter is sensitive to an effi-
ciency. It is worth to mention that the value of EOC may be chan-
ged with the change of Ar or Pr . For simplicity, EOCu is taken as an
Yes No example to illustrate the attributes of EOC as follows:
Pc Pcc
(1) The sign of EOCu . When the efficiency changes from 0.5 to 1,
uð1Þ may be larger or smaller than uð0:5Þ, or equal to uð0:5Þ.
Yes Since uc;exp is positive, the value of EOCu can be positive, neg-
Pc Pcb No ative or zero. In particulary, if uð1Þ > uð0:5Þ, EOCu is positive.
For this case, the critical entrainment ratio uc increases with
Pcb Pc the increasing of the efficiency. If uð1Þ < uð0:5Þ, EOCu is neg-
u uc u uc u 0 ative which means that uc decreases with the increasing of
Pcb Pcc
the efficiency.
(2) The absolute value of EOCu . EOCu is the relative variation
value of uc to uc;exp . Hence, if the absolute value of EOCu is lar-
ger, uc is more sensitive to the efficiency, i.e., the efficiency
has larger influence on prediction accuracy of uc .
(3) The variation of EOCu . If EOCu varies with one parameter,
such as P r , then the sensitivity of uc to the efficiency is
End affected by Pr . It can be induced that if the efficiency is chan-
ged with the change of P r , the prediction error of uc may be
Fig. 3. Flow chart of ejector performance prediction. unchanged at different P r . Clearly, if a proper relationship
between the efficiencies and Pr can be found, the prediction
accuracy can be improved significantly.
Since gp ; gs , /e and /m are the ejector component efficiencies used
to account for losses inside ejector, they should satisfy the following 3.2.1. EOC of the efficiencies for the critical point model
constraints: Figs. 4 and 5 illustrate the calculation results of EOC for critical
point using air as a working fluid based on the experimental data
0 6 gp 6 1; 0 6 gs 6 1; 0 6 /e 6 1; 0 6 /m 6 1: ð27Þ [26]. The variation of EOCu in terms of the pressure ratio Pr at
In order to find the effect of each efficiency on the validity of the Ar ¼ 4:51 are shown in Fig. 4. According to the sign of EOCu for each
model, take one efficiency, such as gp , as variable and the others efficiency it can be deduced that the prediction value of uc
increases with the increasing of /e and gs and decreases with the
are constant C, for example, C ¼ 1. Then
increasing of gp .
uc ¼ uðgp Þ; P cc ¼ Pðgp Þ: ð28Þ For /e , the value of EOCu is large and it varies with the variation
In practice, the efficiencies are more possible to be between 0.5 of P r . Therefore, it is sensitive to /e and closely correlates to Pr . For
to 1. Thus, we denote gp and gs , since the values of EOCu is much lower than that for /e ,
they have much less influence on the prediction value of uc than /e .
EOCu ¼ ðuð1Þ  uð0:5ÞÞ=uc;exp ð29Þ Furthermore, they have no significant correlation with Pr since the
EOCp ¼ ðPð1Þ  Pð0:5ÞÞ=P cc;exp ð30Þ values of EOCu is near constant with the variation of Pr . For /m , it
has no influence on the prediction value of uc since EOCu equals 0.
For breakdown point model, only one ejector component effi-
Fig. 5 shows the variation of EOCp in terms of the pressure ratio
ciency /mb is used to account for the losses inside the ejector and
Pr at Ar ¼ 4:51. It can be seen that the values of EOCp are positive
only the breakdown condensing pressure Pcb is the prediction
for /m and gp , and negative for /e and gs . Thus, the prediction value
parameter. Thus, if the geometric and operating parameters are
of Pcc increases with the increasing of /m and gp , and decreases
given, P cb can be expressed as the function of /mb , i.e.,
with the increasing of /e and gs . The absolute value of EOCp for
Pcb ¼ Pð/mb Þ: ð31Þ gp and gs is much less than those for /m and /e . This indicates that
As those efficiencies in critical point model, /mb satisfy Pcc is much more sensitive to /m and /e than gp and gs . Moreover,
/m and /e have significant correlation to P r due to the sensitivity of
0 6 /mb 6 1 ð32Þ
EOCp with P r .
F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454 449

2.5

2 p e A r =4.51 Ar =7.73
2.0
p e
s m
s m
1.5

1
EOCu

1.0

EOCu
0.5

0
0.0

-0.5
-1
3.0 3.5 4.0 4.5 5.0 8 9 10 11 12 13 14 15 16
Pr
Pr
Fig. 6. Variation of EOCu with pressure ratio using R141b as a working fluid.
Fig. 4. Variation of EOCu with pressure ratio using air as a working fluid.

1.0
1

0.5

0
0.0
EOCp
EOCp

-1 -0.5

Ar=4.51
Ar=7.73
-1.0
-2 p e p e

s m s m

-1.5
8 9 10 11 12 13 14 15 16
3.0 3.5 4.0 4.5 5.0
Pr
Pr
Fig. 7. Variation of EOCp with pressure ratio using R141b as a working fluid.
Fig. 5. Variation of EOCp with pressure ratio using air as a working fluid.

Figs. 6 and 7 illustrate the variations of EOCu and EOCp in terms 3.0
of P r , respectively. They are based on the experimental data, which Pr =12.85
using R141b as working fluid, in [6]. It can be found that for each 2.5 p e
efficiency, the variation trends of EOC in Figs. 6 and 7 are similar
s m
with those in Figs. 4 and 5, respectively. This is because air and 2.0
R141b are all the dry working fluid and their characteristics are
similar with the ideal gas. Since the proposed model is applicable 1.5
EOCu

to dry working fluid, it can be deduced that the results of EOC anal-
ysis derived by air and R141b can be used for other dry working 1.0

fluid, such as R245fa.


0.5
Figs. 8 and 9 describe the variations of EOCu and EOCp with Ar ,
respectively. They are also based on the experimental data of 0.0
Huang et al. [6]. In Fig. 8, the absolute value of EOCu for /e is larger
than the other efficiencies. This means that the critical entrainment -0.5
ratio uc is much more sensitive to /e than the other efficiencies,
and /e is a key parameter for performance prediction of the ejector 6 7 8 9 10 11
Ar
model. Since the value of EOCu for /e drops sharply with the
increasing of Ar , the efficiency /e has significant correlation to Ar Fig. 8. Variation of EOCu with area ratio using R141b as a working fluid.
and should not be taken as a fixed value in the ejector model.
In Fig. 9, it is observed that the absolute values of EOCp for /e
and /m are larger than those for gp and gs . This indicates that /e increasing of Ar ; P cc decreases with the increasing of /e and /e is
and /m have lager influence on the critical pressure Pcc than gp sensitive to Ar . Therefore, /e should not be taken as a fixed value
and gs . For /e , since EOCp is negative and increases with the in the ejector model. For /m , although the variations of EOCp with
450 F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454

1.0 2.5

0.5

2.0
0.0

EOCpb
EOCp

-0.5 1.5

-1.0 Ar =4.51
P r=12.85 1.0
-1.5 p e

s m

-2.0 0.5
6 7 8 9 10 11
3.0 3.5 4.0 4.5 5.0
Ar
Pr
Fig. 9. Variation of EOCp with area ratio using R141b as a working fluid.
Fig. 10. Variation of EOCpb with pressure ratio using air as a working fluid.

Ar are smaller than those for /e , it should not be taken as a constant


etc. How to select basic parameters in the correlations of the effi-
due to large positive values of EOCp .
ciencies is still lacking. Recently, a sparsity-enhanced optimization
From the above analysis, gp and gs have much less influence on
method is proposed to pick out the most important factors in all
the variations of uc and P cc than the other efficiencies and the influ- parameters [23]. Based on the method, the correlation equations
ence keeps almost unchanged with the variation of Ar or Pr . Thus, of /e and /m in the proposed model at critical point can be the sup-
for dry working fluid, such as air, R141b and R245fa, gp and gs posed as follows:
can be taken as fixed value as those in [6], namely
/e ¼ xe1 þ xe2 A1 1
r þ xe3 P r þ xe4 ðAr =P r Þ þ xe5 ðAr =P r Þ
1

gp ¼ 0:95; gs ¼ 0:85 ð34Þ


þ xe6 ðAr P r Þ1 ð35Þ
However, the variations of /m and /e will lead to significant
variations of critical entrainment ratio uc and critical pressure
/m ¼ xm1 þ xm2 A1 1
r þ xm3 P r þ xm4 ðAr =P r Þ þ xm5 ðAr =P r Þ
1
Pcc . Furthermore, /e closely relates to Ar and Pr . Therefore, /m
and /e should be taken as correlation relationship s in which Ar þ xm6 ðAr Pr Þ1 ð36Þ
and P r are basic parameters.
Then, the critical entrainment ratio and critical condensing
pressure can be expressed as
3.2.2. EOC of the efficiencies for the breakdown point model
Fig. 10 describes the variation of EOCpb with the pressure ratio uc ¼ uðxe ; xm Þ; Pcc ¼ Pðxe ; xm Þ ð37Þ
Pr at Ar ¼ 4:51 for the breakdown point. It is obtained based on where xe1 ; xe2 ; . . . ; xe6 , xm1 ; xm2 ; . . . ; xm6 are coefficients and xe and xm
the experiment of Federico Mazzelli et al. [26]. In the experiment, are the corresponding coefficient vectors. Denote
air is used as a working fluid. It can be observed that EOCpb varies  2
 
from 1.3 to 2.0 when P r increases from 3.0 to 5.0. Based on the def- ^ j Þ=u
Ej ðxe ; xm Þ ¼ ðuj ðxe ; xm Þ  u b j Þ= P
^ j 2 þ ðPj ðxe ; xm Þ  P bj ; ð38Þ
inition of EOCpb , the following statements can be deduced from this
variation. Cðxe ; xm Þ ¼ ðkxe k0 þ kxm k0 Þ: ð39Þ

where uj and u ^ j are the jth theoretical and experimental value of the
(1) The value of EOCpb is positive. It means that the prediction
entrainment ratio, respectively, Pj and P b j are the jth theoretical and
value of P cb increases with the increasing of the efficiency
experimental data of the critical condensing pressure, respectively,
/mb .
k  k0 is the number of the non-zero elements of . Then, to deter-
(2) The absolute value of EOCpb is large. This indicates that the
mine the critical model, the optimization problem can be formu-
prediction value of P cb is much sensitive to the efficiency
lated as:
/mb .
(3) The value of EOCpb is not constant but increasing sharply X
n

with the increasing of Pr . This indicates that /mb closely minEðxe ; xm Þ ¼ ð1  bÞEj ðxe ; xm Þ þ bCðxe ; xm Þ; ð40Þ
xe ;xm
j¼1
relates to P r .
s:t:ð1Þ  ð17Þ; ð27Þ; ð34Þ  ð39Þ: ð41Þ
In summary, the efficiency /mb has significant influence on the
where b is used to balance the ejector performance prediction and
validity of the breakdown point model and it closely relates to Pr .
sparsity, while n is the number of the experimental results.
Thus, /mb should be optimized and taken as a correlation equation.
For breakdown point model, the only efficiency /mb has much
influence on the prediction performance of ejector. Similar with
3.3. Correlations of the efficiencies the efficiencies /e and /m , we suppose that

For the critical point mode, it has been identified that /m and /e /mb ¼ xmb1 þ xmb2 A1 1
r þ xmb3 P r þ xmb4 ðAr =P r Þ þ xmb5 ðAr =P r Þ
1

should be expressed as correlation equations including Ar and Pr . In þ xmb6 ðAr P r Þ1 ð42Þ
literature, the equations of efficiencies are determined empirically
and are taken as correlations of Ar [6] or Ar and P r [27] or Pr =Ar [9], Denote
F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454 451

 2
b bj Þ= P
b bj ; from 90 to 100 °C and from 8 to 16 °C, respectively. As R245fa is
Ebj ðxmb Þ ¼ Pbj ðxmb Þ  P ð43Þ
also treated as ideal gas, gp and gs are taken as 0.95 and 0.85,
Cðxmb Þ ¼ kxmb k0 : ð44Þ respectively, as those using R141b as working fluid. Based on the
spasity-enhanced optimum method [23], the correlation equations
where xmb1 ; xmb2 ; . . . ; xmb6 are coefficients and xmb are the corre-
of /e and /m obtained as follows:
sponding coefficient vector. P bj and P b bj are the jth theoretical and
experimental data of the breakdown condensing pressure, respec- /e ¼ 0:6896 þ 0:0414Ar =Pr þ 0:1064Pr =Ar ð52Þ
tively. Then, to determine the subcritical model, the optimization
problem can be formally stated as: /m ¼ 0:9212  0:0596Ar =Pr ð53Þ
X
n They should be used within the following boundaries
minEðxbm Þ ¼ ð1  bÞEj ðxbm Þ þ bCðxbm Þ; ð45Þ
xbm
j¼1 7:25 6 Ar 6 12:89; 10:64 6 Pr 6 14:79
s:t:ð1Þ; ð14Þ  ð16Þ; ð18Þ  ð23Þ; ð32Þ; ð42Þ  ð44Þ: ð46Þ For the breakdown point model, optimum form of the efficiency
/mb and the boundaries of it are:
The coefficient vectors, xe ; xm and xmb can be obtained through the
sparsity-enhanced optimization method proposed in [23]. /mb ¼ 0:9212  0:0596Ar =Pr ð54Þ

3.3.1. The efficiencies for ejector using air as a working fluid Ar ¼ 9:55; 12:45 6 Pr 6 14:79
The efficiencies of the model using air as a work fluid can be
derived by the spasity-enhanced optimum based on the experi-
mental data in [26]. The ejector in the experiment is with a rectan- 4. Results and discussion
gular cross-section nozzle, constant area section and diffuser. The
throat height and depth of the ejector are 6.0 and 48.8 mm, respec- The variable(s) in all correlation equations of /e ; /m and /mb are
tively, and the height and depth of the ejector constant area section Ar =Pr or/and P r =Ar . This indicates that the geometric parameter Ar
are 27.06 and 48.8 mm, respectively. The secondary pressure is and operating parameter Pr affect the ejector component efficien-
1.0 bar and the primary pressures are from 3.0 to 5.0 bar with a cies and ejector performance simultaneously. These correlation
step 0.5 bar. Since gp and gs have less influence on the variations equations were introduced in the proposed models. Then, predic-
of critical entrainment ratio and critical pressure, they are taken tion results will be discussed in the following section.
as 0.95 and 0.85 as those in [6]. The optimum results of efficiencies
/e and /m are: 4.1. Air ejector
/e ¼ 0:9927  0:1236Ar =Pr  0:0036Pr =Ar ð47Þ
In order to verify the proposed model using air as working fluid,
the prediction results were compared to the experimental data in
/m ¼ 0:8574  0:0120Ar =Pr þ 0:0446Pr =Ar ð48Þ
[26]. The operation conditions of calculation and experiment are
The optimum efficiency at breakdown point is: the same as those reported in Section 3.3.1. Fig. 11 illustrates the
comparison of 6 experimental characteristic curves with the pre-
/mb ¼ 0:5528 þ 0:1333Pr =Ar ð49Þ
diction results over entire operational range. It can be seen that
These correlations of efficiencies should be used within the fol- the differences between the prediction curves and experiment
lowing boundaries: curves are much lower at critical mode than those at subcritical
mode. In addition, the prediction errors are less than 10% at sub-
Ar ¼ 4:51; 2:5 6 Pr 6 5:0
critical mode. Compared with the prediction errors of entrainment
ratios at subcritical mode, the errors at the critical point and the
3.3.2. The efficiencies for ejector using R141b as a working fluid breakdown point are much lower. More specifically, for critical
For the critical points, gp and gs are take as 0.95 and 0.85 as points, the prediction errors of critical entrainment ratios uc and
those in [6]. The equations of /e and /m are obtained by the
spasity-enhanced optimum method based on experimental data
1.2
in [6]. They are:
experiment
/e ¼ 0:8091 þ 0:05354Pr =Ar ð50Þ
1.0 Pr =3.0
Pr =3.5
/m ¼ 0:8811  0:04322Ar =Pr þ 0:05053Pr =Ar ð51Þ
0.8
Pr =4.0
They should be used within the following boundaries which are Pr =4.5
the same as those of experimental data: Pr =5.0
0.6
0:400 MPa 6 Pp0 6 0:604 MPa; 0:040 MPa 6 Ps0 6 0:047 MPa Ar =4.51
u

model
6:44 6 Ar 6 10:64; for 2:64 mm 6 At 6 2:82 mm
0.4 Pr =3.0
Pr =3.5
The correlation equation for breakdown point cannot be 0.2 Pr =4.0
obtained because of laking experimental data. Pr =4.5
Pr =5.0
0.0
3.3.3. The efficiencies for ejector using R245fa as a working fluid
The ejector component efficiencies can be obtained based on 1.0 1.2 1.4 1.6 1.8 2.0
the experimental data in [28]. The experiments were carried out
Pc /bar
for 9 different ejectors at 11 critical points and 3 breakdown points.
The area ratio Ar varies from 7.25 to 12.89, correspondingly, the Fig. 11. Comparison of predictive performance with experimental performance
generating temperature tg and the evaporating temperature te vary using air as a working fluid.
452 F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454

critical condensing pressures P cc are within ±5% and ±3%, respec- results coincide fairly well with the experimental data and the
tively; for breakdown point s, the errors of breakdown condensing errors are all within 5%. Since the efficiencies are selected by
pressures P cb are within ±3%. EOC analysis and models are determined through advanced opti-
mization method, it is certain that the prediction accuracy of our
4.2. R141b ejector model is better than traditional 1-D model [6].

The prediction results of entrainment ratio at critical point 4.3. R245fa ejector
using R141b as working fluid are reported in Table 1. The geomet-
ric parameters and operating parameters were taken as those in [6] The experimental data using R245fa as refrigerant in [28] are
during calculation. From Table 1, it can be seen that the maximum used to verify the proposed model. Taking the same operation con-
error of prediction entrainment ratio is 7.95% which is much better ditions as those in the experiment, the ejector performance predic-
than 22.99% in [6]. The average absolute error is 2.77% in compar- tion is carried out. The theoretical results are in good agreement
ison with 8.70% in [6]. The comparison between prediction critical with the corresponding experimental results. For critical points,
condensing pressure and experimental critical condensing pres- the prediction errors of critical entrainment ratios uc and critical
sure has been shown in Fig. 12. It can be seen that the prediction condensing pressures P cc are within ±6% and ±3%, respectively.

Table 1
Comparison of performance results between present model and the experiment of Huang et al. (1999).

A3 =At T p0 (°C) T s0 (°C) Experiment u Huang et al. u Present u Huang et al. errors (%) Present errors (%)

AA ejector
6.44 95 8 0.1859 0.1554 0.1826 16.43 1.78
6.44 90 8 0.2246 0.2156 0.2204 3.99 1.89
6.44 84 8 0.2880 0.2880 0.2760 0.23 4.17
6.44 78 8 0.3257 0.3525 0.3440 8.24 5.63
6.44 95 12 0.2350 0.2573 0.2373 9.49 0.98
6.44 90 12 0.2946 0.3257 0.2831 10.54 3.92
6.44 84 12 0.3398 0.4147 0.3500 22.04 3.00
AB ejector
6.99 90 8 0.2718 0.2093 0.2577 22.99 5.20
6.99 84 8 0.3117 0.3042 0.3197 2.39 2.57
6.99 78 8 0.3922 0.4422 0.3952 12.74 0.77
AG ejector
7.73 95 8 0.2552 0.2144 0.2604 15.98 2.04
7.73 90 8 0.3040 0.2395 0.3087 21.22 1.56
7.73 84 8 0.3883 0.3704 0.3793 4.61 2.32
7.73 78 8 0.4393 0.4609 0.4648 4.93 5.81
7.73 95 12 0.3503 0.3434 0.3304 1.97 5.69
7.73 90 12 0.4034 0.4142 0.3883 2.67 3.75
7.73 84 12 0.4790 0.4769 0.4724 12.09 1.38
7.73 78 12 0.6132 0.6659 0.5737 8.6 6.44
AC ejector
8.28 95 8 0.2814 0.2983 0.2944 6.01 4.63
8.28 90 8 0.3488 0.3552 0.3472 1.84 0.46
8.28 84 8 0.4241 0.4605 0.4241 8.58 0.01
8.28 78 8 0.4889 0.5966 0.5170 22.03 5.74
AD ejector
9.41 95 8 0.3457 0.3476 0.3654 0.56 5.70
9.41 90 8 0.4446 0.4178 0.4273 6.02 3.90
9.41 84 8 0.5387 0.5215 0.5170 3.19 4.03
9.41 78 8 0.6227 0.6944 0.6250 11.51 0.36
9.41 95 12 0.4541 0.4708 0.4549 3.67 0.18
9.41 90 12 0.5422 0.5573 0.5284 2.78 2.54
9.41 84 12 0.6350 0.6906 0.6345 8.75 0.07
9.41 78 12 0.7412 0.8626 0.7617 16.37 2.77
EG ejector
6.77 95 8 0.2043 0.1919 0.2022 6.06 1.03
EC ejector
7.26 95 8 0.2273 0.2078 0.2317 8.57 1.94
7.26 95 12 0.3040 0.3235 0.2961 6.41 2.59
ED ejector
8.25 95 8 0.2902 0.2658 0.2926 8.39 0.81
EE ejector
9.17 95 8 0.3505 0.3253 0.3502 7.2 0.08
9.17 95 12 0.4048 0.4894 0.4370 10.55 7.95
EF ejector
9.83 95 8 0.3937 0.3774 0.3921 4.13 0.41
9.83 95 12 0.4989 0.5482 0.4865 9.89 2.49
EH ejector
10.64 95 8 0.4377 0.4627 0.4439 5.7 1.42
F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454 453

1.6 range was obtained by integrating ejector model at critical point


with that at subcritical point. Since ejector component efficiency
is one key factor affecting the validity of the model, a novel EOC
1.4 analysis method was proposed and employed to identify the ejec-
+5% tor component efficiencies which have more influence on the ejec-
Pcc/Mpa (model)

tor performance and the model’s accuracy. It is found that the


1.2 ejector performance at critical point is much more sensitive to
the expansion efficiency /e and mixing efficiency /m than the isen-
-5% tropic efficiencies gp and gs ; the breakdown condensing pressure is
1.0
sensitive to /mb at the breakdown point. Then, the correlation
equations of identified efficiencies /e ; /m and /mb were deter-
mined by sparsity-enhanced optimization method. For the entire
0.8
operational regime, compared with exiting methods, our obtained
model can achieve more accurate prediction performance than
those obtained by existing models. More specifically, the maxi-
0.6
0.6 0.8 1.0 1.2 1.4 1.6 mum prediction error obtained by our model is less than 10%,
Pcc/Mpa(experiment) which is much lower than those of existing models.

Fig. 12. Predictive critical condensing pressure v.s experimental critical condensing
pressure. Acknowledgement

This research is supported by International Science and Tech-


0.6 nology Cooperation Project of China (2013DFA61580), the National
experiment Research Foundation of Korea (NRF) grant funded by the Korea
0.5 te government (MSIP) through GCRC-SOP (No. 2011-0030013), Key
te Science and Technology Program of Shanxi Province, China
0.4 te (20140313006-6), Research Project Supported by Shanxi Scholar-
ship Council of China (2016-032).
Ar=9.55
0.3
tg
References
u

0.2 model
[1] J. Kasperski, B. Gil, Performance estimation of ejector cycles using heavier
te
hydrocarbon refrigerants, Appl. Therm. Eng. 71 (1) (2014) 197–203.
0.1 te [2] Joseph Henry Keenan, An investigation of ejector design by analysis and
te experiment, J. Appl. Mech. 17 (1950) 299.
[3] D.W. Sun, I.W. Eames, Recent developments in the design theories and
0.0 applications of ejectors, J. Inst. Energy 68 (475) (1995) 65–79.
[4] J.T. Munday, D.F. Bagster, A new ejector theory applied to steam jet
0.18 0.19 0.20 0.21 0.22 0.23 0.24 0.25 refrigeration, Ind. Eng. Chem. Process Design Develop. 16 (4) (1977) 442–449.
[5] I.W. Eames, S. Aphornratana, D.W. Sun, The jet-pump cycleA low cost
Pc\MPa refrigerator option powered by waste heat, Heat Recov. Syst. CHP 15 (8)
(1995) 711–721.
Fig. 13. Comparison of predictive performance with experimental performance [6] B.J. Huang, J.M. Chang, C.P. Wang, V.A. Petrenko, A 1-D analysis of ejector
using R245fa as a working fluid. performance, Int. J. Refrig 22 (1999) 354–364.
[7] Y. Zhu, Y. Li, Novel ejector model for performance evaluation on both dry and
wet vapors ejectors, Int. J. Refrig 32 (2009) 21–31.
[8] Y.M. Chen, C.Y. Sun, Experimental study of the performance characteristics of a
For breakdown point, the maximum error of breakdown condens- steam-ejector refrigeration system, Exp. Thermal Fluid Sci. 15 (4) (1997) 384–
ing pressure P cb is less than 1% compared to experimental data. 394.
[9] J.M. Cardemil, S. Colle, A general model for evaluation of vapor ejectors
Fig. 13 describes the ejector performance over entire opera-
performance for application in refrigeration, Energy Convers. Manage. 64
tional range. By analyzing the ejector performance, it can be (2012) 79–86.
observed that prediction errors at critical mode are determined [10] I.W. Eames, S. Aphornratana, H. Haider, A theoretical and experimental study
by those of entrainment ratio and condensing pressure at critical of a small-scale steam jet refrigerator, Int. J. Refrig 18 (6) (1995) 378–386.
[11] X. Xu, G. Chen, L. Tang, Z. Zhu, S. Liu, Experimental evaluation of the effect of an
point. The prediction errors at subcritical mode are determined internal heat exchanger on a transcritical CO2 ejector system, J. Zhejiang
by those at critical point and the prediction error of condensing University Sci. A 12 (2) (2011) 146–153.
pressure at breakdown point. Further observation discovers that [12] J. Chen, H. Havtun, B. Palm, Investigation of ejectors in refrigeration system:
optimum performance evaluation and ejector area ratios perspectives, Appl.
the prediction errors at critical mode are less than those at critical Therm. Eng. 64 (s 1C2) (2014) 182–191.
point, whereas, the prediction errors at subcritical mode are larger [13] M. Ouzzane, Z. Aidoun, Model development and numerical procedure for
than those at critical and breakdown points. Thus, the maximum detailed ejector analysis and design, Appl. Therm. Eng. 23 (18) (2003) 2337–
2351.
prediction error is at subcritical mode. It is observed that predic- [14] G. Besagni, R. Mereu, F. Inzoli, Ejector refrigeration: a comprehensive review,
tion errors are less than 10% over the entire operational range Renew. Sustain. Energy Rev. 53 (2016) 373–407.
and much lower than those in literature. [15] L. Boumaraf, A. Lallemand, Modeling of an ejector refrigerating system
operating in dimensioning and off-dimensioning conditions with the
working fluids R142b and R600a, Appl. Therm. Eng. 29 (2) (2009) 265–274.
[16] W.X. Chen, M. Liu, D.T. Chong, J. Yan, A.B. Little, Y. Bartosiewicz, A 1D model to
5. Conclusions
predict ejector performance at critical and sub-critical operational regimes,
Int. J. Refrig. 36 (6) (2013) 1750–1761.
In this paper, we developed ejector models for predicting the [17] A. Hemidi, F. Henry, S. Leclaire, J.M. Seynhaeve, Y. Bartosiewicz, CFD analysis of
ejector performance at critical point and breakdown point base a supersonic air ejector. Part I: experimental validation of single-phase and
two-phase operation, Appl. Therm. Eng. 29 (8) (2009) 1523–1531.
on constant-pressure mixing and constant-pressure disturbing [18] S. Varga, A.C. Oliveira, B. Diaconu, Numerical assessment of steam ejector
assumptions. Then, a novel ejector model for entire operational efficiencies using CFD, Int. J. Refrig 32 (6) (2009) 1203–1211.
454 F. Li et al. / Applied Thermal Engineering 115 (2017) 444–454

[19] G. Besagni, R. Mereu, P. Chiesa, F. Inzoli, An Integrated Lumped Parameter-CFD [24] B. Saleh, Performance analysis and working fluid selection for ejector
approach for off-design ejector performance evaluation, Energy Convers. refrigeration cycle, Appl. Therm. Eng. 107 (2016) 114–124.
Manage. 105 (2015) 697–715. [25] F. Liu, E.A. Groll, D. Li, Modeling study of an ejector expansion residential CO2
[20] H. Zhao, K. Zhang, L. Wang, J. Han, Thermodynamic investigation of a booster- air conditioning system, Energy Build. 53 (2012) 127–136.
assisted ejector refrigeration system, Appl. Therm. Eng. 104 (2016) 274–281. [26] F. Mazzelli, A.B. Little, S. Garimella, Y. Bartosiewicz, Computational and
[21] J. Liu, K.L. Teo, X. Wang, C. Wu, An exact penalty function-based differential experimental analysis of supersonic air ejector: turbulence modeling and
search algorithm for constrained global optimization, Soft. Comput. 20 (4) assessment of 3D effects, Int. J. Heat Fluid Flow 56 (2015) 305–316.
(2016) 1305–1313. [27] F. Liu, E.A. Groll, D. Li, Investigation on performance of variable geometry
[22] L. Hou, C. Zhao, C. Wu, S. Moon, X. Wang, Discrete firefly algorithm for ejectors for CO2 refrigeration cycles, Energy 45 (2012) 829–839.
scaffolding construction scheduling, ASCE J. Comput. Civil Eng. (2016), Article [28] K.O. Shestopalov, B.J. Huang, V.O. Petrenko, O.S. Volovyk, Investigation of an
ID 04016064. experimental ejector refrigeration machine operating with refrigerant R245fa
[23] F. Li, C. Wu, X. Wang, Q. Tian, K.L. Teo, Sparsity-enhanced optimization for at design and off-design working conditions. Part 2. Theoretical and
ejector performance prediction, Energy 113 (2016) 25–34. experimental results, Int. J. Refrig 55 (2015) 212–223.

You might also like