A novel special distributed method for dynamic refrigeration

system simulation
F.Q. Wang
a,
*
, G.G. Maidment
a
, J.F. Missenden
a
, R.M. Tozer
b
a
London South Bank University, Faculty of Engineering, Science & the Built Environment, 103 Borough Road, SE1 0AA London, UK
b
Waterman Gore-Mechanical and Electrical Consulting Engineers, Versaillers Court, 3 Paris Garden, SE1 8ND London, UK
Received 20 August 2004; received in revised form 13 September 2006; accepted 5 October 2006
Abstract
Anovel dynamic mathematical model based on spatially distributed approach has been developed and validated in this paper.
This model gives good agreement in predicting the system COP and other parameters. The validated model has been used to
enhance the prediction of the micro variations of superheat and sub-cooling. The novel spatial distributed model for the con-
denser and evaporator in refrigeration system, calculates the two-phase region in gas and liquid eld separately since the gas
and liquid in the two-phase region have different velocities. Previous researchers have used a pre-dened function of the
void fraction in their spatially distributed model, based on experimental results. This approach results in the separate solution
of the mass and energy equations, and less calculation is required. However, it is recognized that the mass and energy equations
should be coupled during solving for more accurate solution. Based on the energy and mass balance, the spatial distribution
model constructed here solves the velocity, pressure, refrigerant temperature, and wall temperature functions in heat exchangers
simultaneously. A novel iteration method is developed and reduces the intensive calculations required. Furthermore, the con-
denser and evaporator models have shown a parametric distribution along the heat exchanger surface, therefore, the spatial dis-
tribution parameters in the two heat exchangers can be visualised numerically with a two-phase moving interface clearly shown.
2006 Elsevier Ltd and IIR. All rights reserved.
Keywords: Refrigeration; Condenser; Evaporator; Two-phase ow; Superheating; Subcooling; Modelling; Heat transfer; COP
Simulation dynamique dun syste`me frigorique a` laide dune
nouvelle methode distribuee
Mots cles : Refrigeration ; Condenseur ; E

vaporateur ; E

coulement diphasique ; Surchauffe ; Sous-refroidissement ; Modelisation ; Transfert

de chaleur ; COP
* Corresponding author. Present address: Fulcrum Consulting, 62e68 Rosebery Avenue, London EC1R 4RR, UK. Tel.: 44 0207 520 1305;
fax: 44 0207 520 1355.
E-mail address: fuqiao.wang@fulcrumrst.com (F.Q. Wang).
0140-7007/$35.00 2006 Elsevier Ltd and IIR. All rights reserved.
doi:10.1016/j.ijrefrig.2006.10.010
ARTICLE IN PRESS
International Journal of Refrigeration xx (2006) 1e17
www.elsevier.com/locate/ijrefrig
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
1. Introduction
Generally, in refrigeration system simulation, heat ex-
changer models dealing with unsteady compressible two-
phase ow fall into one of two categories: the lumped
parameter approach or the spatially distributed approach
[3]. The lumped parameter model (LPM) divides the heat
exchangers into mainly three regions, the superheated gas
region, the two-phase region, and the sub-cooling region.
In each region, average values are used to represent the para-
meters of the whole region. The spatial distributed parameter
method considers the parameters in each region to vary. The
complexity of the spatially distributed approach is much
greater than the lumped parameter model. The main merit
of the spatial model is that the parameter distribution over
distance is detailed and the moving two-phase interface is
clearly dened. Therefore, the heat exchange processes in
the refrigeration system can be displayed numerically but
also more precise results can be produced.
During transient operation, the refrigerant mass ow rate
is continuously changing, which causes spatial variations in
the refrigerant distribution in the system components as well
as variable refrigerant states at the inlet and outlet of each
component. It is seen that the mass distribution and other pa-
rameters within the heat exchangers are functions of time
and space. Hence the spatially distributed model has been
used by various researchers. The transient simulation of
the refrigeration system was reviewed by a number of re-
searchers and their key conclusions in relation to spatial
distribution modelling is described below:
Chi and Didion [1] simulated an air-cooled system with
moving refrigerant phase boundaries, with each phase region
treated as a counter-ow heat exchanger. Refrigerant ow in
both heat exchangers is one-dimensional and homogenous.
Internal resistances of metallic components were neglected.
A direct expansion evaporator for water-cooling was simu-
lated by Yasuda et al.s [9] model. The refrigerant in each
phase in the two-phase region was treated using lumped pa-
rameters, while the superheated region was simulated using
a nite difference approach. The method considers the heat
transfer distribution within the evaporator tube material, sec-
ondary uid and refrigerant phases. This approach was used
Nomenclature
A area (m
2
)
C specic heat (kJ kg
1
K
1
)
D diameter (m)
F force (N)
G refrigerant mass ow ux (kg m
2
s
1
)
g gravity constant (m
2
s
1
)
h enthalpy (J kg
1
)
i control variable in a loop
L length (m)
_ m mass ow rate (kg s
1
)
Q heat ux (W m
2
)
P pressure (Pa)
R gas constant (J kg
1
K
1
)
Re Reynolds number
r radius (m)
T temperature (K)
T
r
reduced temperature
t temperature (

C)
T
f
heat transfer uid temperature (K)
U overall heat transfer coefcient (Wm
2
K
1
)
VV volume (m
3
)
v specic volume (m
3
kg
1
)
W velocity (ms
1
)
x coordinate (m)
X dryness fraction
Xdif the cell coordinate (m)
Xcv the cell interface coordinate (m)
Y the displacement of the spring in TEV
Greek symbols
a heat transfer coefcient (Wm
2
K
1
)
r density (kg m
3
)
l thermal conductivity (Wm
1
K
1
)
h
f
n effectiveness
t time (s)
J void fraction
6 wetted perimeter (m)
Sufx
Cr refrigerant side
w wall
ca air side
tp two-phase region
cond condenser
tev thermal expansion valve
comp compressor
sub sub-cooling
evap evaporator
sat saturation or saturation point
f heat transfer uid
s constant entropy process
g gas
r reduced parameters, refrigerant
i inside
r3 the compressor outlet
L liquid
r4 the condenser outlet or the TEV inlet
n the nth iteration
r6 the evaporator outlet
o outside
v vapour
p constant pressure
ARTICLE IN PRESS
2 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
in order to be able to predict the evaporator exit condition pre-
cisely, especially when observing the hunting phenomenon.
Inthe evaporative regiona meanvoidfractionwas determined
from the Hughmark correlation, assuming uniform heat ux.
Macarthur [4] modelled refrigerant heat exchangers on
a nite-volume basis by discretizing the length of the heat
exchanger into phase-independent control volumes and ap-
plying the mass and energy balances on the refrigerant,
tube material and secondary uid in each control volume.
In contrast to the condenser, the evaporator is modelled as
a two-phase region, the liquid and vapour being handled sep-
arately while being in equilibrium. Heat from the secondary
uid passes exclusively to the liquid refrigerant to cause
evaporation. Mass and energy balances on the vapour and
liquid phases are linked through the evaporation rate. The
single-phase region in the evaporator was modelled in the
same way as in the condenser. In Sami et al.s [6] model,
the total system is divided into a number of lumped param-
eter control volumes. The condenser is modelled as a com-
bination of the three different phase regimes: the fully
superheated, the two-phase and the sub-cooled. The vapour
and liquid phases in each regime were modelled by a set of
coupled mass and energy balances.
Rossi and Braun [5] modelled the heat exchanger on
a xed-boundary, nite difference basis. Elemental mass
and energy balances are written for each element of the
heat exchanger, including the heat transfer equations be-
tween the refrigerant, the tube-wall and the air. The momen-
tum equation does not appear as the pressure drops are
considered negligible. The heat exchanger model uses the
inlet and outlet refrigerant ow-rates and the inlet enthalpy
as boundary conditions. The air side heat transfer is mod-
elled differently for heat transfer in the condenser and evap-
orator. For the condenser the dry analysis was based on the
temperature difference between the wall and the air, while
for the evaporator the wet analysis is based on the enthalpy po-
tential between the refrigerant and the air. An effectiveness-
NTU formulation was employed in the analysis.
Svensson [7] modelled a condenser consisting of two
zones: the condensing and the sub-cooling, both of which
are assumed tohave xed volumes. Energy andmass balances
are obtained for each of these zones as a set of coupled ordi-
nary differential equations. These are then combined to give
an equation for the pressure difference in terms of: tube-wall
temperature, incoming mass ow rate and saturated refriger-
ant properties. The tube-wall and thewater volume are discre-
tized into n volumes in the condensing zone and 1 volume in
the sub-cooling zone and nite difference forms of the energy
balance equations are used. For each of these n volumes, the
refrigerant side conditions are considered identical.
Most researchers have previously used a pre-dened func-
tion of the void fraction in their spatially distributed model,
based on experimental results. This approach results in the
separate solution of the mass balance and energy equations,
and less calculations are required. However, it is recognized
that the mass and energy equations should be coupled during
solution. The spatial distribution model constructed here is
based on the principle of energy and mass balances, to solve
the velocity, pressure, temperature, and wall temperature
functions in heat exchangers simultaneously.
2. Condenser model
2.1. The assumptions of the spatial distributed model
The following assumptions are made in this study:
(1) Axial conduction in the refrigerant is negligible.
(2) Liquid and vapour refrigerant phases in the heat
exchangers are in thermal equilibrium.
(3) The effects of pressure wave dynamics are negligible.
(4) The pressure drops in the condenser, evaporator and
short connecting pipe work are neglected. (This has
a small, variable effect on COP.)
(5) Thermal resistances of metallic elements in the system
are negligible in comparison with other thermal resis-
tances, however, their capacitance is important.
(6) In the two-phase region, axial conduction in the tube is
neglected. However, this is considered in the single-
phase ow region.
(7) Annular ow in the condenser and evaporator.
2.2. Mass and energy balance equations
The condenser has three different heat-dissipating zones:
a cool down zone for the superheated gas, a condensation
zone and a sub-cooling zone for liquid refrigerant. Each re-
gion was divided into two control volumes. The condenser
model is shown in Fig. 1.
According to above assumption the refrigerant owin the
heat exchanger can be simplied to one-dimensional ow.
The mass and energy equations can be expressed as follows:
vr
vt

vrw
vx
0 1
vrh
vt

vrwh
vx

vP
vt

4
D
a
i
T
w
T 2
The pipe wall energy equation can be written as:
C
w
r
w
A
w
vT
w
vt
a
i
pD
i
T T
w
a
0
A
o
h
f
T
a
T
l
w
A
w
v
2
T
w
vX
2
3
The above three equations include four parameters: pres-
sure P, pipe wall temperature T
w
, refrigerant temperature T
and velocity w. The refrigerant property state equation
must be included.
f P; v; T 0 4
Thus with four independent variables, the four equations
are closed. In the two-phase region, the pressure and the
ARTICLE IN PRESS
3 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
temperature are dependent parameters and can be easily de-
termined, however, the void fraction or the dryness fraction
need to be calculated. This group of equations is difcult to
solve, and numerical simulation is required. In order to
obtain convergent solutions speedily, numerical methods
are discussed below.
2.3. Equations in the sub-cooling liquid
region and superheat region
In both these regions, the uid is in single phase, and Eqs.
(1e4) can be used directly. In this model, the liquid density
is a function of pressure and temperature, which is calcu-
lated by the thermodynamic properties model as described
elsewhere [8].
2.4. Equations in the two-phase region
In the two-phase region, the model assumes that the liq-
uid and gas are in equilibrium, the temperature and pres-
sure in the gas region are the same as that in the liquid
region. However, the liquid density is much greater than
that of gas and the condensed liquid will accumulate at
the bottom of the tube due to gravity. The two-phase
ow has its own particular features: the gas velocity and
the liquid velocity are not the same, with the liquid velocity
being lower than the gas velocity due to its higher viscosity.
Although this phenomenon is neglected by most re-
searchers who assume the same velocity in the two-phase
region. In this model the individual liquid and gas veloci-
ties were needed to accurately predict performance. To
achieve this the void fraction was calculated in Eq. (5) to
split the liquid and gas phases.
j
V
g
V
5
where V
g
is the gas volume, and V is the total volume at the
control volume unit.
Macarthur [4] assumed heat fromthe heat transfer uid is
transferred exclusively to the liquid refrigerant in the evap-
orator, (the wall being completely wet) a similar assumption
was made in this condenser model, hence the liquid and gas
equations can be written as:
Vapour region:
v

jr
g

vt

jr
g
w
g

vx

_ m
V
0 6
v

jr
g
h
g

vt

jr
g
w
g
h
g

vx

vP
vt

_ m
V
h
fg
0 7
Liquid region:
vr
l
1 j
vt

vr
l
w
l
1 j
vx

_ m
V
0 8
vr
l
h1 j
vt

vr
l
w
l
h
l
1 j
vx

vP
vt

_ m
V
h
fg

4
D
a
i
T
w
T 9
where _ m is condensed liquid mass ow rate.
3. The energy and mass balance
equations in the evaporator
The model developed in the evaporator is based on the
following assumptions:
(1) Axial conduction in the refrigerant is negligible.
(2) The liquid and vapour refrigerant phases in the heat
exchangers are in thermal equilibrium.
(3) The effects of pressure wave dynamics are negligible.
(4) The pressure drop in the evaporator is neglected.
(5) The thermal resistances of metallic element in the sys-
tem are negligible in comparison with other thermal
resistances, however, their capacitance is important.
V
V
_
c
o
n
d
1

L_sub L_sat L_sup
m_r4
h_r4
m_r3
h_r3
P,w
h, t,
x
VV_condL
VV_condG
P,w
h, t,
x
X
Q_cond1 Q_cond2 Q_cond3
Metal tube
VV_cond3
Fig. 1. The conceptual model of the condenser.
ARTICLE IN PRESS
4 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
(6) In the two-phase region, the axial conduction of the
tube is neglected. However, this is considered in the
single-phase ow region.
(7) Refrigerant enters the evaporator directly from the TEV.
(8) Theuidstate at the entrance tothe evaporator is two-phase
already, some refrigerant has ashed into gas in the TEV.
As with the condenser model, the conceptual model of
the evaporator is similar. However, in the evaporator nor-
mally only two regions exist: the two-phase region and the
superheat region. The mass, energy, and pipe wall energy
equations are the same with the condenser; the only differ-
ences are their initial and boundary conditions.
As in the condenser, for the liquid continuously changing
phase, the heat from outside is assumed exclusively to be
transferred to the liquid. Then, the energy and mass balance
equations in the two-phase region are exactly the same as in
the condenser. Here the equations are omitted as they are
already dened in Section 2.
As the refrigerant pressure drops due to the suction of
compressor, more liquid ashes into gas in the direction of
the evaporator, this phenomenon causes the liquid tempera-
ture to drop and the initial thermal balance to break down.
The heat from outside the tube then balances the energy
needed by the ashing process. This energy stored in the
gas is then removed away by the compressor.
4. The other parameters calculation
in two-phase region
4.1. Void fraction and dryness fraction
As dened in Eq. (5), the void fraction is the ratio of
vapour volume to the total volume in the two-phase region.
With the void fraction, the density of the uid can be
expressed as:
r r
l
1 j r
g
j 10
Dryness fraction is dened as:
X
m
g
m
g
m
l
11
The dryness fraction is used to determine the properties
of the refrigerant mixture. For liquid and vapour mixture,
the thermodynamic parameters can be expressed as:
h 1 Xh
l
Xh
g
12
v 1 Xv
l
Xv
g
13
4.2. Relation between the void fraction
and dryness fraction
Dryness fraction and void fraction are not independent
parameters. From the void fraction, the dryness fraction
can be calculated as:
X
r
g
j
r
g
j1 jr
l
14
Similarly, the void fraction can be calculated as:
j
v
g
X
v
g
X 1 Xv
l
15
Therefore calculation of variable dryness fraction can be
calculated from the variable void fraction.
5. Boundary conditions
5.1. Boundary conditions at inlet (x 0) and outlet (x L)
of the condenser
In the model, the inlet of the condenser is set as the outlet
of the compressor. This assumes there is no heat loss between
the compressor and condenser. Some extra heat exchanger
surface between the compressor and the condenser has
been added to the actual condenser surface to account for
the real thermal conditions (a similar addition was made to
evaporator area). At the outlet of the condenser, the mass
ow rate is assumed as that passing through the TEV. The
boundary conditions can be described mathematically as:
x 0; h
r3
h
comp2
16
_ m
r3
_ m
comp2
17
x L; _ m
r4
_ m
tev
18
5.2. The boundary conditions at inlet (x L) and
outlet (x 0) of the evaporator
At the outlet of the evaporator (x 0) the parameters are
assumed the same as those at the inlet to the compressor. The
refrigerant mass stored in the compressor is neglected, and
the inlet mass ow rate is equal to the outlet mass ow rate
(storage only therefore occurs in the heat exchangers). At
the inlet to the compressor the mass ow rate is equal to
the mass owrate of the TEV, and the parameters at this point
are equal to the outlet parameters of the TEV, therefore the
boundary conditions can be expressed as follows:
x 0; h
r6
h
comp1
19
_ m
r6
_ m
comp1
20
x L; _ m
r4
_ m
tev
21
5.3. Boundary conditions on the heat exchanger surface
At the outside of the condenser or evaporator coil, the air
is drawn through the heat exchanger by a fan and heat is
taken away by convection. The boundary condition can be
expressed as:
q a
o
t
w
t
air
22
ARTICLE IN PRESS
5 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
where a
o
is the heat transfer coefcient which can be found
in Section 10, and is dened differently for both the evapo-
rator and condenser.
6. Initial conditions
6.1. Initial conditions in the condenser
In a dynamic model, all the parameters are considered as
time dependent, and the initial conditions affect system-
starting performance. The initial condition plays an important
roleinthe results. The initial mass distributionandtemperature
also affect the solution. Unfortunately, it is very difcult to
track the initial mass distribution in the system. The initial
condition is only based on the assumption that the condenser
is in two-phase equilibrium or in a superheated state.
The mass distribution will initially have an effect on
system performance. If a quantity of mass is stored in the
condenser at the initial state, the rate at which the pressure
in the condenser will rise will be affected by the mass pres-
ent. The initial transient period will be shorter with less mass
contained, and this has been demonstrated by a lumped
model investigation detailed elsewhere [8].
6.2. The initial conditions of evaporator
The initial condition in the evaporator is somewhat differ-
ent from the condenser. When a system starts, the evaporator
is normally always charged with refrigerant. This is because
the evaporator temperature and pressure are always less than
the condenser temperature and pressure when the compressor
is stopped. The refrigerant therefore migrates to the evapora-
tor due to this pressure difference. However, the exact refrig-
erant charge in the evaporator is difcult to predict. The
initial conditions always assume the uid exists in the two-
phase state at a saturation temperature equivalent to the initial
temperature of the calorimetric chamber. For initial values,
the refrigerant was distributed 80% by mass in condenser,
20% in evaporator, residue being vapour of low density.
7. Discretization equations
7.1. The method of meshing
The heat exchanger is meshed along the refrigeration
ow direction using a method by which the rst node is at
the entrance of the heat exchanger and the last node is at
the exit of the heat exchanger. The node coordinates value
can be calculated and stored as:
dx
L
N 1
23
Xdifi i 1dx; i 1; 2; .; N 24
The coordinates value of the interface of the control
volume can then be calculated:
Xcvi
Xdifi 1 Xdifi
2
; i 1; 2; .; N 1 25
7.2. The method of discretization
The model uses a control volume based technique to con-
vert the governing equations to algebraic equations that can
be solved numerically. This control volume technique con-
sists of integratingthegoverningequations about eachcontrol
volume, yielding discrete equations that conserve each quan-
tity on a control volume basis. For the time-based equations,
all the discrete equations will be expressed implicitly with
time and this method will enable the iterations to converge.
For the non-steady, one-dimensional problem, all
parameters in the equations vary with space and time. In a
discrete expression these parameters will be stored in a two-
dimensional array, the subscript represents space, and the
superscript represents time. For example, t
1
2
presents the
second node temperature at present time, and t
0
2
expresses
the second node temperature at the previous time. In
order to save memory, the maximum-element in each two-
dimensional array is N2. All the parameters will be sent
to an external data le at each time step or selected time step.
7.3. The discretization of the equations
The nite difference method was employed in all equa-
tion discretizations. Mass and energy balance method was
used on all boundary conditions over the cells.
7.3.1. Condenser
Fig. 2 shows the two-phase region mesh, the dashed lines
are the interfaces of the control volumes, the solid lines are
the nodal position for each control volume. There are three
regions in the condenser including superheated gas, two-
phase and liquid region. For each region, the energy and
mass balance equations in the discrete control volume terms
can be expressed as:
(1) Superheated gas:
r
1
i1
w
1
i1
r
1
i
w
1
i
Dx

r
1
i
r
0
i
Dt
26
r
1
i1
w
1
i1
Ah
1
i1
a
ref
pDDx

t
1
wi
t
1
i

r
1
i
w
1
i
Ah
1
i
ADx
r
1
i
h
1
i
r
0
i
h
0
i
Dt
ADx
P
1
i
P
0
i
Dt
27
Dx Xcvi Xcvi 1 28
(2) Two-phase region:
In this region, the liquid and gas ow at different
velocities. The energy and mass equations will be
conducted separately.
For the gas:
j
1
i1
r
1
g i1
w
1
g i1
j
1
i
r
1
g i
w
1
g i
Dx

_ m
ADx

j
1
i

r
1
g i
r
0
g i

Dt
29
ARTICLE IN PRESS
6 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
j
1
i1
r
1
g i1
w
1
g i1
Ah
1
g i1
j
1
i
r
1
g i
w
1
g i
Ah
1
g i
_ mh
fg
j
1
i
ADx
r
1
g i
h
1
g i
r
0
g i
h
0
g i
Dt
j
1
i
ADx
P
1
i
P
0
i
Dt
30
For the liquid:

1 j
1
i1

r
1
L i1
w
1
L i1

1 j
1
i

r
1
L i
w
1
L i
Dx

_ m
ADx

1 j
1
i

r
1
L i
r
0
L i

Dt
31

1j
1
i1

r
1
Li1
w
1
Li1
Ah
1
Li1

1j
1
i

r
1
Li
w
1
Li
Ah
1
Li
a
ref
pDDx

t
1
wi
t
1
i

_ mh
fg

1j
1
i

ADx
r
1
Li
h
1
Li
r
0
Li
h
0
Li
Dt

1j
1
i

ADx
P
1
i
P
0
i
Dt
32
Dx Xcvi Xcvi 1 33
(3) Sub-cooled region:
r
1
i1
w
1
i1
r
1
i
w
1
i
Dx

r
1
i
r
0
i
Dt
34
r
1
i1
w
1
i1
Ah
1
i1
a
ref
pDDx

t
1
w i
t
1
i

r
1
i
w
1
i
Ah
1
i
ADx
r
1
i
h
1
i
r
0
i
h
0
i
Dt
ADx
P
1
i
P
0
i
Dt
35
Dx Xcvi Xcvi 1 36
(4) Wall energy equation:
a
ref
pD

t
1
i
t
1
wi

a
o
F
o
h
f

t
air
t
1
wi

l
w
A
w
t
1
wi1
t
1
wi
xdifi 1xdifi
l
w
A
w
t
1
wi1
t
1
wi
xdifixdifi 1
c
w
r
w
A
w
xcvixcvi 1
t
1
wi
t
0
wi
Dt
37
7.3.2. Evaporator
There are two regions in the evaporator, that is, the super-
heated gas and the two-phase region. For each region, the en-
ergy and mass balance equations in discrete control volume
terms can be dened as:
(1) Superheated gas:
r
1
i1
w
1
i1
r
1
i
w
1
i
Dx

r
1
i
r
0
i
Dt
38
r
1
i1
w
1
i1
Ah
1
i1
a
ref
pDDx

t
1
w i
t
1
i

r
1
i
w
1
i
Ah
1
i
ADx
r
1
i
h
1
i
r
0
i
h
0
i
Dt
ADx
P
1
i
P
0
i
Dt
39
Dx Xcvi Xcvi 1 40
(2) Two-phase region:
In this region, the liquid and gas ow at different
velocities. The energy and mass equations will be
conducted separately.
For the gas:
j
1
i1
r
1
g i1
w
1
g i1
j
1
i
r
1
g i
w
1
g i
Dx

_ m
ADx

j
1
i

r
1
g i
r
0
g i

Dt
41
j
1
i1
r
1
g i1
w
1
g i1
Ah
1
g i1
j
1
i
r
1
g i
w
1
g i
Ah
1
g i
_ mh
fg
j
1
i
ADx
r
1
g i
h
1
g i
r
0
g i
h
0
g i
Dt
j
1
i
ADx
P
1
i
P
0
i
Dt
42
For liquid:

1 j
1
i1

r
1
L i1
w
1
L i1

1 j
1
i

r
1
L i
w
1
L i
Dx

_ m
ADx

1 j
1
i

r
1
L i
r
0
L i

Dt
43

1j
1
i1

r
1
Li1
w
1
Li1
Ah
1
Li1

1j
1
i

r
1
Li
w
1
Li
Ah
1
Li
a
ref
pDDx

t
1
wi
t
1
i

_ mh
fg

1j
1
i

ADx
r
1
Li
h
1
Li
r
0
Li
h
0
Li
Dt

1j
1
i

ADx
P
1
i
P
0
i
Dt
44
(3) Wall energy equation:
a
ref
pD
i

t
1
i
t
1
w i

a
o
F
o
h
f

t
air
t
1
w i

l
w
A
w
t
1
w i1
t
1
w i
xdifi 1 xdifi
l
w
A
w
t
1
w i1
t
1
w i
xdifi xdifi 1
c
w
r
w
A
w
xcvi xcvi 1
t
1
w i
t
0
w i
Dt
45
Control
Control
Volume
I + 1
Xdif (i+1) Xcv (i) Xcv (i-1) Xdif (i-1) Xdif (i)
Volume I
Volume
I - 1
Control
Fig. 2. The two-phase region mesh in condenser.
ARTICLE IN PRESS
7 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
7.3.3. The inlet and outlet boundary conditions
discretization
7.3.3.1. Condenser entrance (node i 1). The refrigerant
at the entrance of the condenser will be the superheated gas
from the compressor; then the energy and mass balance
equations can be expressed as follows:
Mass balance equation:
m
r3
r
1
1
w
1
1
A
A dx

r
1
1
r
0
1

Dt
46
Energy balance equation in the refrigerant:
m
r3
h
r3
a
i
pDdx

t
1
w 1
t
1
1

r
1
1
w
1
1
Ah
1
1
A dx
r
1
1
h
1
1
r
0
1
h
0
1
Dt
A dx
P
1
1
P
0
1
Dt
47
Energy balance equations in the wall:
a
ref
pDDx

t
1
1
t
1
w 1

a
o
Fh
f
Dx

t
air
t
1
w 1

l
w
A
w
t
1
w 2
t
1
w 1
Xdif2 Xdif1
c
w
A
w
r
w
dx
t
1
w 1
t
0
w 1
Dt
48
In above equations, Dx Xcv1 Xdif1.
7.3.3.2. Condenser exit (node i Nc)
Mass balance equation:
r
1
n1
w
1
n1
Am
r4

ADx

r
1
n
r
0
n

Dt
49
Energy balance equation in the refrigerant:
m
r4
h
1
n
a
ref
pDdx

t
1
w n
t
1
n

r
1
n1
w
1
n1
Ah
1
n1
ADx
r
1
n
h
1
n
r
0
n
h
0
n
Dt
ADx
P
1
n
P
0
n
Dt
50
Energy balance equations in the wall:
a
ref
pDDx

t
1
n
t
1
w n

a
o
Fh
f
dx

t
air
t
1
w n

l
w
A
w
t
1
w n1
t
1
w n
Xdifn Xdifn 1
c
w
A
w
r
w
Dx
t
1
w n
t
0
w n
Dt
51
In the above equations, Dx Xdifn Xcvn 1.
7.3.3.3. Evaporator entrance (node i 1). The gas enter-
ing the evaporator from the TEV will be wet, with dryness
fraction X, and then the energy and mass balance equations
can be expressed separately in the liquid and gas phases as
follows.
For the vapour, mass balance equation:
m
r4
X j
1
n
r
1
g n
w
1
g n
A _ m
1
n

j
1
n
A Dx

r
1
g n
r
0
g n

Dt
52
The energy balance equation in the refrigerant:
Xm
r4
h
1
g
_ m
1
n
h
1
g
r
1
g n
w
1
g n
Ah
1
g
j
1
n
Adx
r
1
g n
h
1
g n
r
0
g n
h
0
g n
Dt
Aj
1
n
dx
P
1
n
P
0
n
Dt
53
For liquid, the mass balance equation:
m
r4
1 X

1 j
1
n

r
1
L n
w
1
L n
A _ m
1
n

1 j
1
n

A Dx

r
1
L n
r
1
L n

Dt
54
The energy balance equation in the refrigerant:
m
r4
1 Xh
L 4
a
i
pD dx

t
1
w n
t
1
n

1 j
1
n

r
1
n
w
1
n
Ah
1
n
_ m
1
n
h
1
L n

1 j
1
n

A dx
r
1
n
h
1
L n
r
0
n
h
0
L n
Dt

1 j
1
n

A dx
P
1
n
P
0
n
Dt
55
The energy balance equations in the wall:
a
ref
pDDx

t
1
n
t
1
w n

a
o
Fh
f
Dx

t
air
t
1
w n

l
w
A
w
t
1
w n1
t
1
w n
Xdifn Xdifn 1
c
w
A
w
r
w
dx
t
1
w n
t
0
w n
Dt
56
In the above equations, Dx xdifn xcvn.
7.3.3.4. Evaporator exit (node i N)
The mass balance equation:
r
1
2
w
1
2
Am
r6

ADx

r
1
1
r
0
1

Dt
57
The energy balance equation in the refrigerant:
m
r6
h
1
1
a
ref
pDdx

t
1
w 1
t
1
1

r
1
2
w
1
2
Ah
1
2
ADx
r
1
1
h
1
1
r
0
1
h
0
1
Dt
ADx
P
1
1
P
0
1
Dt
58
The energy balance equation in the wall:
a
ref
pDDx

t
1
1
t
1
w1

a
o
Fh
f
dx

t
air
t
1
w1

l
w
A
w
t
1
w 1
t
1
w 2
Xdif2 Xdif1
c
w
A
w
r
w
Dx
t
1
w 1
t
0
w 1
Dt
59
In the above equations, Dx Xcv1 Xdif1.
ARTICLE IN PRESS
8 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
7.3.4. The phase transition point discrete equations
7.3.4.1. The transition point in condenser. Because of the
different heat transfer mechanisms, the rate of single-phase
heat transfer in a sub-cooled uid is always less than the
heat transfer coefcient taking place in the condensing pro-
cess. This makes establishing the point of saturation impor-
tant in accurate model performance.
During normal system operation this point always varies
with the condenser ambient air condition. When the system
starts, the delivery temperature and condenser pressure are
low, and the low-pressure ratio enables the compressor to
deliver a large mass ow rate. Under start up conditions the
condensing point will be somewhere just after the entrance
of the condenser. However, with further system running,
the delivery temperature and condenser pressure both rise.
With greater superheat, the condensing point would move
towards the exit of the condenser. The sub-cooling transition
point also moves with the running conditions. Initially, this
point moves because the surface area available for condens-
ing decreases.
To establish the points, an enthalpy-comparison method
was developed and is used, where the gas enthalpy from
the energy equation is compared with the saturation vapour
enthalpy at the same condenser pressure. If it is less than the
saturated vapour enthalpy, the point is then marked as a start-
ing point for the condensing process. The recorded point will
be stored as a variable, and the moving boundary of the in-
terface is gradually revealed. The point of sub-cooling is
found in a similar way.
7.3.4.1.1. Gas condensing point. At the point of conden-
sation, the gas is condensing to liquid and this will be
attached to the wall in a very thin lm. The liquid velocity
is assumed to be 0 at the initial condensing point. When
the void fraction at this point reaches 0.99, the liquid begins
to move. The equations of the initial condensing point are
a little different from those in the two-phase region and at
this point they are recorded and marked as I ktr1 in the
program code. The mass and energy balance equations at
this point are given below.
For the gas:
j
1
i1
r
1
g i1
w
1
g i1
j
1
i
r
1
g i
w
1
g i
Dx

_ m
A Dx

j
1
i

r
1
g i
r
0
g i

Dt
60
j
1
i1
r
1
g i1
w
1
g i1
Ah
1
g i1
j
1
i
r
1
g i
w
1
g i
Ah
1
g i
_ mh
fg
j
1
i
A Dx
r
1
g i
h
1
g i
r
0
g i
h
0
g i
Dt
j
1
i
ADx
P
1
i
P
0
i
Dt
61
Dx Xcvi Xcvi 1
For liquid:
_ m

1 j
1
i

r
1
L i
Dt
Dx 62
a
ref
pDDx

t
1
w i
t
1
i

_ mh
fg

1 j
1
i

A Dx
r
1
L i
h
1
L i
Dt

1 j
1
i

A Dx
P
1
i
P
0
i
Dt
63
Dx Xcvi Xcvi 1
7.3.4.1.2. The sub-cooling liquid point. The last cell in
the two-phase region, as marked I ktr2 in the model
code, is recognized as a transition point from the two-phase
region to liquid region. The energy and mass balances for
this cell in gas region are:
j
1
i1
r
1
g i1
w
1
g i1
j
1
i
r
1
g i
w
1
g i
Dx

_ m
A Dx

j
1
i

r
1
g i
r
0
g i

Dt
64
j
1
i1
r
1
g i1
w
1
g i1
Ah
1
g i1
j
1
i
r
1
g i
w
1
g i
Ah
1
g i
_ mh
fg
j
1
i
A Dx
r
1
g i
h
1
g i
r
0
g i
h
0
g i
Dt
j
1
i
A Dx
P
1
i
P
0
i
Dt
65
Dx Xcvi Xcvi 1
r
1
g i
r
1
g i1
r
1
g
66
For liquid:

1 j
1
i1

r
1
L i1
w
1
L i1

1 j
1
i

r
1
L i
w
1
L i
Dx

_ m
A Dx

1 j
1
i

r
1
L i
r
0
L i

Dt
67

1 j
1
i1

r
1
L i1
w
1
L i1
Ah
1
L i1

1 j
1
i

r
1
L i
w
1
L i
Ah
1
L i
a
ref
pDDx

t
1
w i
t
1
i

_ mh
fg

1 j
1
i

A Dx
r
1
L i
h
1
L i
r
0
L i
h
0
L i
Dt

1 j
1
i

A Dx
P
1
i
P
0
i
Dt
68
Dx Xcvi Xcvi 1 69
Some attention needs to be paid to the rst cell in the
sub-cooled region. This cell on subsequent iterations can
become sub-cooled, just saturated or two-phase. The analy-
sis of this must be careful to ensure accurate prediction of its
subsequent state. The liquid and gas enter the cell simulta-
neously from the upper and lower positions of the cell.
The energy and mass balance equations in this cell can be
expressed as:
r
1
L
w
1
L i1

1 j
1
i1

w
1
g i1
r
1
g
j
1
i1
r
1
L
w
1
i
Dx

r
1
i
r
0
i
Dt
70
ARTICLE IN PRESS
9 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
r
1
L
w
1
L i1
Ah
1
L

1 j
1
i1

Ar
1
g
w
1
g i1
j
1
i1
h
1
g
a
ref
pDDx

t
1
w i1
t
1
i1

r
1
i1
w
1
i1
Ah
1
L
A Dx
r
1
i1
h
1
i1
r
0
i1
h
0
i1
Dt
A Dx
P
1
i1
P
0
i1
Dt
71
Dx xcvi 1 xcvi 72
For this point, if the calculated enthalpy h
1
i1
is larger
than the saturated liquid enthalpy then the point will be in
the two-phase region, and this point will be marked as
I ktr2 and recognized as the last point of the two-phase
region. If the enthalpy of the refrigerant is less than the satu-
rated liquid enthalpy, this point is still in the liquid region.
By recording this point, the moving interface can be traced
properly.
7.3.4.2. The transition point in evaporator. In a similar way
to the condenser model, this point was also traced in the
model. However, in the evaporator, there is only one phase
transition point, from the two-phase region to the super-
heated gas region The transition point is a balance point af-
fected by the refrigerant mass ow passing through the TEV
and the thermal load exerted on the evaporator. For a smaller
thermal load, less mass of refrigerant evaporates in the heat
exchanger, and this point will be adjacent to the outlet of the
evaporator, the heat transfer surface for the superheated gas
will be less and lower superheat will be obtained resulting in
a smaller TEV mass ow.
For a higher thermal load on the evaporator, more mass of
refrigerant will be evaporated and the balance point will be
further away from the outlet of the evaporator, resulting in
more heat exchanger surface for the superheatedgas. Ahigher
outlet temperature of the superheated gas is possible which
will increase the mass ow rate in the TEVin the system.
In the code, the control cell i represents this point. To
establish this point, the liquid mass is used as a criterion,
such that if the liquid mass is less than or equal to zero,
the cell will be in the dry region. However, it may be
more complex while the liquid is moving. This is because
the liquid can move forward because of gas friction or
backward because of high thermal load on the evaporator
or gravity (in this model the gravity force is neglected).
The energy and mass balance for this point can be
expressed as follows.
For the gas:
j
1
i1
r
1
g i1
w
1
g i1
j
1
i
r
1
g i
w
1
g i
Dx

_ m
A Dx

j
1
i

r
1
g i
r
0
g i

Dt
73
j
1
i1
r
1
g i1
w
1
g i1
Ah
1
g i1
j
1
i
r
1
g i
w
1
g i
Ah
1
g i
_ mh
fg
j
1
i
A Dx
r
1
g i
h
1
g i
r
0
g i
h
0
g i
Dt
j
1
i
A Dx
P
1
i
P
0
i
Dt
74
Dx Xcvi Xcvi 1 75
For liquid:
A

1 j
1
i1

r
1
L i1
w
1
L i1
_ m

1 j
1
i

m
1
L i
m
0
L i

Dt
76

1 j
1
i1

r
1
L i1
w
1
L i1
Ah
1
L i1
a
ref
pDDx

t
1
w i
t
1
i

_ mh
fg

1 j
1
i

A Dx
r
1
L i
h
1
L i
r
0
L i
h
0
L i
Dt

1 j
1
i

A Dx
P
1
i
P
0
i
Dt
77
Dx Xcvi Xcvi 1 78
uxli 0 if a
1
i
> 0:998 x > 92% 79
With the above equations, a
1
i
can be calculated. If a
1
i
is
greater than 0.998, the dry fraction is 0.9 or more, and
the ow is recognized as fog ow in the evaporator. The liq-
uid forms very small particles and will move with the gas at
the same velocity. If a
1
i
is less than 0.998, the liquid and gas
will separate with a dened interface and the liquid will
move at a speed different to that of the gas.
7.3.5. The conditions for iteration
In a set of differential equations with a close chain of
implicit parameters, the numerical solution is based on
initial assumptions. Under these initial values, the next
values can be calculated from the equations and then com-
pared with the assumed values. If the chosen tolerance is
not met, the initial assumptions can be modied, then cal-
culation should be carried out again, until the chosen tol-
erance is met. For example, in this model, at each time
step, the initial pressures of the condenser and the evapo-
rator are assumed, then the mass ow rate of the compres-
sor and TEV can be calculated, the condensing and
evaporator temperatures can also be calculated, then heat
transfer process can be analysed in the heat exchangers,
with known mass ow-rates. The condenser and the
evaporator model can be carried out with the results of pa-
rameters at each tube section, these parameters then affect
the TEV and compressor mass ows and parameters.
The assumed pressures in the condenser and evaporator
are correct when the required tolerance is met; otherwise
the simulation needs to be repeated until the convergence
tolerance is met.
In the condenser and evaporator models, an iterative pro-
cedure is also required. The initial parameters such as refrig-
erant temperature, wall temperature, and gas void fraction
distributions are assumed, then the differential equations
are solved and these parameters may be then recalculated.
Comparing the calculated parameters with the assumptions,
the assumptions are then modied again till the chosen tol-
erances are met. A relative error is used to compare with the
chosen tolerance (usually 1.0 10
6
). Once the iteration
conditions are met, the solution can then be stored and the
solution for the next time step considered.
ARTICLE IN PRESS
10 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
8. The novel triangle and tri-point
method developed for iteration
In order to compare the new method with general
methods for iteration, the general methods including Bisec-
tion, Secant, False Position, and NewtoneRaphson methods
are listed below as reviewed by EFUNDA [2].
8.1. Bisection method
The idea of the bisection method is based on the fact
that a function will change sign when it passes through
zero. The method can be expressed by the following
program:
To nd a root of f(x) 0 in the interval of (a
0
, b
0
)
With which f(a
0
) f(b
0
) <0
Pick tolerance 3
X
k1
(b
k
a
k
)/2, k 0,1,2,3,.
If (jf(x
k1
)j <3) root found, stop iteration.
Else
If (f(x
k1
) f(b
k
) < 0) a
k1
x
k1
; b
k1
b
k
Else a
k1
a
k
; b
k1
x
k1
End if
End if
When an interval contains a root, the bisection method is
the one that will not fail. However, it is amongst the slow-
est. When an interval contains more than one root, the
bisection method can nd one of them. When an interval
contains a singularity, the bisection method converges to
that singularity.
8.2. Secant method
To improve the slow convergence of the bisection
method, the secant method assumes that the function is ap-
proximately linear in the local region of interest and uses
the zero-crossing of the line connecting the limits of the in-
terval as the new reference point. The next iteration starts
from evaluating the function at the new reference point
and then forms another line. The process is repeated until
the root is found.
To nd a root of f(x) 0 in the interval of (x0, x1)
With which f(x0) f(x1) < 0
x
k1
x
k

x
k
x
k1
f x
k
f x
k1

f x
k
; k 1; 2; 3; .
Mathematically, the secant method converges more rap-
idly near a root than the false position method (discussed be-
low). However, since the secant method does not always
bracket the root, the algorithm may not converge for func-
tions that are not sufciently smooth.
8.3. False position method
Similar to the secant method, the false position method
also uses a straight line to approximate the function in the
local region of interest. The only difference between these
two methods is that the secant method keeps the two most
recent estimates, while the false position method retains
the most recent estimate and the next recent one which has
an opposite sign in the function value.
To nd a root of f(x) 0 in the interval of (a
0
, b
0
)
With which f(a
0
) f(b
0
) < 0
Pick tolerance 3
x
k1
b
k

b
k
a
k
f b
k
f a
k

f b
k
; k 1; 2; 3; .
If (jf(x
k1
)j <3) root found, stop iteration.
Else
If (f(x
k1
) f(b
k
) <0) a
k1
x
k1
;b
k1
b
k
Else a
k1
a
k
; b
k1
x
k1
End if
End if
The false position method, which sometimes keeps an
older reference point to maintain an opposite sign bracket
around the root, has a lower and uncertain convergence
rate compared to the secant method. The emphasis on brack-
eting the root may sometimes restrict the false position
method in difcult situations while solving highly non-
linear equations.
8.4. NewtoneRaphson method
The NewtoneRaphson method nds the slope (the
tangent line) of the function at the current point and uses
the zero of the tangent line as the next reference point.
The process is repeated until the root is found.
To nd a root of f(x) 0 with the initial guess x
0
,
x
k1
x
k

f x
k

f
0
x
k

where k 0; 1; 2; 3.
The NewtoneRaphson method is much more efcient
than other simple methods such as the bisection method.
However, the NewtoneRaphson method requires the calcu-
lation of the derivative of a function at the reference point,
which is not always easy. Furthermore, the tangent line often
shoots wildly and might occasionally be trapped in a loop
and the solution does not converge. It is recommended to
monitor the step obtained by the NewtoneRaphson method.
When the step is too large or the value is oscillating, other
more conservative methods should take over the case. How-
ever, this method cannot be used in cycle iterations, because
the function is not known explicitly.
ARTICLE IN PRESS
11 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
8.5. The novel method
The general iteration methods are listed above. The
Bisection method is too slow and the NewtoneRaphson
method is not effective when the function is explicit. The
other two methods are also not effective when tried by the
author repeatedly. This is because in either the condenser
or evaporator, both the pressures are changing. When using
the Secant method, one heat exchanger reaches a converging
point, however, the other might be still far away from con-
vergence. When the other heat exchanger is near to conver-
gence, the previous converged point has drifted away. In
order to solve this problem, a new method of iteration was
developed. This method is based on the false position
method, but the minimum positive and negative errors of
each iteration are recorded, respectively. In the condenser it-
erations, this method was initially used, and in the evapora-
tor a simple constant factor is used initially, this makes the
iterations in the evaporator relatively slow. When the con-
denser iteration reaches the tolerance then the same method
is applied to the evaporator, if the iterations in the condenser
diverge, the recorded minimum positive and negative errors
will be applied to the condenser iteration. This method en-
sures the iteration continues and makes the iteration three
times faster than the other methods mentioned here.
9. Simulation of TEV and compressor
The compressor generates a mass ow from the low-
pressure side of the cycle to the high-pressure side of the
cycle. The heat generated by the electric motor is absorbed
by the refrigerant rather than transferred to the ambient.
Thus the compression is assumed to be adiabatic and the
isentropic efciency can be found from the manufacturer,
which can be expressed as a function of pressure ratio:
h
i

h
2
h
1
h
0
2
h
1
f

p
2
p
1

80
The refrigerant mass ow in the compressor is given by
the following equation:
_ m r$VV$h
v
81
where VVis the swept volume rate of the compressor, which
is calculated by:
VV
p
4
D
2
S n=60 82
where S is stroke of piston, D is the piston diameter; n is the
rotational speed of the shaft (rev/min), h
v
is the volumetric
efciency of the compressor, which was determined from
makers data as:
h
v
f

p
2
p
1

83
The compressor power, p
e
, can be calculated:
p
e

_ m

h
0
2
h
1

h
m
84
The TEVshown in Fig. 3 is a proportional controller with
the control signal from the evaporator outlet superheat. It re-
sponds to the difference between the pressure of the refrig-
erant at the evaporator outlet and the pressure developed
in the temperature-sensing remote phial attached to the out-
let of the evaporator. The phial is normally charged with the
same refrigerant as the plant. A spring setting in the valve is
provided to adjust the required superheat of the vapour leav-
ing the evaporator.
The TEVoperation can be modelled by a set of algebraic
equations, which are given below. The forces acting on the
TEV spindle, when it is in equilibrium, are shown in Fig. 3:
F
s
F
2
F
1
85
where F
1
is the force exerted as a result of the pressure of the
liquid/vapour mixture in the remote phial, which is the sat-
urated pressure at the outlet temperature of the evaporator
calculated by the state equation. F
2
is the outlet pressure
of the evaporator (or inlet pressure of the evaporator minus
the pressure drop in the evaporator). F
s
is the closing force
exerted by the spring.
The spring pre-tension can be adjusted. The force exerted
when the valve starts to open is denoted by F
0
. Therefore, the
force exerted by the spring is given by:
F
s
F
0
KY 86
where Y is the displacement in millimeter and K is the spring
stiffness in N/mm. Manipulating Eqs. (85) and (86) gives:
KY F
2
F
1
F
0
87
or KY P
2
P
1
AF
0
where A is the area of the diaphragm, P
1
is the pressure in the
remote phial and P
2
is the evaporator pressure at the position
of the pressure-sensing connection. When Y 0, the spring
pre-tension is known as the static pressure of the valve,
which is usually expressed as:
F
0
DP
0
A 88
where Y is given by:
Y C
0
P
2
P
1
DP
0
89
where C
0
A/K; also Y is the adjustment of the TEV.
The ow area A of the TEVis a function of Y. The actual
ow area of the TEV can be determined with the calculated
value of the displacement Y. The mass ow rate can then be
calculated:
M C
d
A

r DP

90
A

p
4
DY Ctga
2

p
4
D
2

sina 91
ARTICLE IN PRESS
12 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
10. Heat transfer coefcient used in the model
In this model the Shah correlation and the Dobson and
Chato method will be used to predict the heat transfer behav-
iour in the condenser.
The Shah correlation for condensation is a two-phase
multiplier approach valid for annular ow. It is based on
the liquid heat transfer coefcient, which is shown in Eq.
(92). The liquid heat transfer coefcient is calculated from
the DittuseBoelter correlation.
a
cond
a
L

1 X
0:8

3:8X
0:76
1 X
0:04
Pr
0:38

92
where
Pr
P
sat
P
c
; reduced pressure
a
L

l
D

0:023

Re
L
1 X

0:8
Pr
0:4
L

Re
L

G1 XD
m
L
> 350
Pr
L

C
pL
m
L
l
L
In the sub-cooling region, the heat transfer coefcient
on the refrigerant side can be calculated from the standard
DittuseBoelter equation:
Nu 0:023 Re
0:8
Pr
0:4
93
The Kandlikar correlation [Kandlikar 1987] is shown:
a a
L

1:136C
0:9
o
25F
r

0:3
667:2B
0:7
o
F
fl

94
where F

is uid dependent parameter, for R22,

F
fl
2:2
a
L
0:023 Re
0:8
L
Pr
0:4
L
l
L
=d
C
o

1 X
X

0:8

r
G
r
L

0:5
; X is dryness fraction
F
r

G
2
r
2
L
gD
; G is mass flux; kg=m
2
s
B
o

q
G$h
fg
Gungor and Winterton correlations were found to pro-
duce mean deviations of 19% [Kandlikar [12]], therefore,
in this model, Kandlikar [11] and Gungor and Winterton
[10] correlations are used.
11. Model validation
The test plant is based on an IMI Marstair model C60/E,
direct air cooling split system. The indoor unit comprised of
an evaporator (IMI impact CU6E) and a fan. The outdoor
unit (IMI CUE60) comprised of a condenser, compressor
and propeller fan.
The evaporator was sited inside a calorimetric chamber
and was a twin circuit air-cooled coil with 26 tubes, two
of which were blanked. The tubes had external aluminium
alloy plate ns at a pitch of 551/m. The total external heat
transfer surface area was 11.432 m
2
, while the total internal
heat transfer area was 0.533 m
2
with refrigerant tube diam-
eters of 9.52 mm, giving a total metal mass of 6.8 kg.
Remote bulb
To evaporator
Superheat adjustment
Liquid
refrigerant in
Fs
Fs
F
2
F
1
Pressure in bulb P2
Connect to
evaporator outlet
Pressure in
evaporator P1
Fig. 3. A typical TEV and Forces balance on TEV spindle.
ARTICLE IN PRESS
13 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
The air was driven over the evaporator by a fan having
three speeds. These gave 0.233, 0.283 and 0.312 kg/s of
air, respectively. The measurements were all taken with
the evaporator fan operating at the top speed (0.312 kg/s).
The condenser was an air-cooled coil having 63 tubes
(one blanked off), arranged in three parallel circuits. The
tubes had external aluminium plate ns at a pitch of 472/m.
The total external heat transfer area was 17.955 m
2
and the
total internal heat transfer area was 0.867 m
2
. The refriger-
ant tube diameter was 9.53 mm and the total metal mass
was 8.2 kg. The air was driven over the condenser by a
propeller fan having two speed settings. The air mass ow
rate at the low speed was 0.563 kg/s and 0.7 kg/s at the
high speed setting, at which the experimental work was
performed.
The compressor was a Tecumseh model, No. TAH
5527E, having a swept volume of 0.0023611 m
3
/s, at
2900 rpm. The unit was a reciprocating, suction gas cooled
hermetic type.
The thermostatic expansion valve is a Danfoss model
(with orice 03), having a capacity of 5.2 kW. The outside
diameter of the suction line is 15.88 mm, this was insu-
lated to minimise the heat gain from the surroundings.
The outside diameter of the liquid line to the evaporator
pipe is 9.525 mm. All the valves are are valves made
by Danfoss.
The spatial model was validated against the test results
for the system. The validation was carried on one group of
test data. Like the lumped parameters model, both the inlet
air temperature of the condenser and evaporator were fed
in as known parameters into the model. The calculated and
test parameters were represented on the same diagram
against time. The results are shown in Figs. 4e6.
Fig. 4 shows the calculated pressures with the test data
in the condenser and the evaporator. The calculated pres-
sures agreed with the test pressures as shown in Fig. 4.
Fig. 5 shows the temperature before the TEV for the model
and the test data. This shows there is 3K difference. This
error is caused by a number of factors. The boundary and
initial conditions inuence the results. Unlike the lumped
parameters, the spatially distributed model is more depen-
dent on this condition, and includes detailed geometric
parameters as well. Fig. 6 shows the temperatures from
simulated and tested results in the evaporator. The
calculated evaporating temperature is 2 K less than the
test data. This error mainly comes from the heat transfer
coefcient, but also initial and boundary conditions do in-
uence the results. For example if the mass ow rate
through the compressor (the boundary condition) is higher
than the actual rate, the calculated evaporating temperature
would be lower than the actual one.
12. The other parameters shown against space
and time in two heat exchangers
As mentioned in Section 1, one additional benet of the
spatial model is that the spatial distributed parameters can be
visualized against time and space. If reasonable initial and
boundary conditions are given, a valid result will be obtained
by this model. Figs. 7e12 show some of these parameters
distributed in the heat exchanger. Fig. 7 shows the dryness
fraction distribution along the condenser tube. As time in-
creases, the dryness fraction distribution will be stabilized
as long as running conditions remain unchanged. At the en-
trance of the heat exchanger (x 0), it clearly shows that the
length of the superheated gas section is changing. As the out-
let temperature of the compressor increases, the gas (super-
heat) length increases, and the position of the two-phase
region interface is moving. Fig. 8 shows the refrigerant tem-
perature distribution in the condenser, it is a function of time
and space. The horizontal line represents the phase changing
temperature. The length of each region including gas,
0.0
0.3
0.6
0.9
1.2
1.5
1.8
0 200 400 600 800 1000
Time (sec)
P
r
e
s
s
u
r
e

(
M
P
a
)
Pc_test Pe_test
Pc_cal Pe_cal
Fig. 4. Pressure validation.
15
20
25
30
35
40
45
0 200 400 600 800 1000
Time (sec)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
TEV inlet_Test
TEV inlet_Cal
Fig. 5. Validation of the temperature before TEV.
5
9
13
17
21
25
0 200 400 600 800 1000
Time (sec)
T
e
m
p
e
r
a
t
u
r
e

(
C
)
Test_Te Test_Evaporator outlet
Cal_T3 Cal_Evaporator outlet
Fig. 6. Validation of the temperature in the evaporator.
ARTICLE IN PRESS
14 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
two-phase, and sub-cooled can be clearly shown in this g-
ure. Fig. 9 shows the wall temperature distribution along the
tube against time. These lines are similar to the refrigerant
temperature distribution. The only difference is that their rel-
ative values are lower than the refrigerant temperatures. In
these the refrigerant can be seen owing from right to left,
the exit is at x 0 and the inlet is x L
evap
. From the dia-
gram, the lowest temperature in the evaporator is around
the superheat and two-phase interface in the initial stage.
This is because a quantity of refrigerant vapour was drawn
away by the compressor at the initial stage, and liquid refrig-
erant is being ashed into gas intensively around this point,
which reduces the wall temperature.
13. The moving boundary
Fig. 11 shows the length of superheat section in the
evaporator. It can be seen that this length changes rapidly
at the beginning of the cycle. This is because a large quan-
tity of vapour is drawn into the compressor, and the
pressure in the evaporator drops quickly. This then results
in much liquid ashing into vapour and liquid owing
backwards in the evaporator against vapour ow direction.
This results in an increase in the superheat section. Fig. 12
shows how the two-phase region varies with time. The
length of this region is important to the accuracy of the
pressure calculation in the condenser. The longer this sec-
tion is, the more gas that is condensed and the lower con-
densing pressure. The main parameter determining the
length is the heat transfer coefcient, which can be up to
15% different than the test data.
14. Conclusions
A spatial distribution model in two heat exchangers has
been presented in this paper. The complexity of the model
is much greater than the lumped parameter model. The
main merit of the spatial model is that the parameter distri-
bution over distance is detailed and the moving two-phase
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40
Condenser length (m)
D
r
y
n
e
s
s

f
r
a
c
t
i
o
n
0.1 Sec
0.4 Sec
0.6 Sec
2 Sec
120 Sec
360 Sec
Fig. 7. Dryness fraction distribution ong the condenser tube.
280
290
300
310
320
330
340
350
360
370
380
0 10 20 30 40
Condenser length (m)
T
e
m
p
e
r
a
t
u
r
e

(
K
)
0.1 sec
0.5 sec
1.5 sec
8.0 sec
31.9 sec
120.4 sec
Fig. 8. The refrigerant temperature distribution in the condenser.
280
284
288
292
296
300
0 5 10 15 20 25
Evaporator length (m)
T
e
m
p
e
r
a
t
u
r
e

(
K
)
0.1 Sec 0.5 Sec 6.0 Sec
15.4 Sec 30.9 Sec 120.9 Sec
Fig. 9. The wall temperature distribution in the evaporator.
276
278
280
282
284
286
288
290
292
294
296
0 4 8 12 16 20 24
Evaporator length (m)
T
e
m
p
e
r
a
t
u
r
e

(
K
)
0.5 Sec 1.0 Sec 3.0 Sec
10 Sec 60.9 120.9
Fig. 10. The refrigerant temperature distribution in the evaporator.
ARTICLE IN PRESS
15 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
interface is clearly dened. This would allow visualisation
of the modelled system, useful both in design and optimisa-
tion. As the model has been experimentally validated it
stands alongside previous lumped parameter models. For
the future, three-way comparisons between the two model-
ling methods and experiments would be interesting.
Most researchers have previously used a pre-dened
function of the void fraction in their spatially distributed
model, based on experimental results. This approach results
in the separate solution of the mass balance and energy eq-
uations, and less calculations are required. However, it is
recognized that the mass and energy equations should be
coupled during solving for more accurate solution. The spa-
tial distribution model constructed here is based on the prin-
ciple of energy and mass balance to solve the velocity,
pressure, temperature and wall temperature functions in
heat exchangers simultaneously. Although more calcula-
tions are required, this method presents a clearer understand-
ing of two-phase ow region. If the initial and boundary
conditions are accurate, better precision should be achieved
in the simulation.
The lumped parameter model gives an energy balance for
each section namely superheat region, two-phase region and
sub-cooling region. This means there are only a maximum
of three mesh points in each heat exchanger. The spatial
distribution model gives more mesh points in heat
exchangers and therefore the solution should be more
accurate. The results show that mesh size is one factor in
the models performance. By examining the results from the
two models, it is believed that the results depend on the
following factors:
(1) The accuracy of heat transfer coefcient used;
(2) The boundary conditions;
(3) The initial conditions for dynamic model; and
(4) The physical model assumptions.
Although a number of heat transfer coefcients are avail-
able, there are some discrepancies and more work is needed
in this area. The models used here are not reliable enough to
predict variable superheat. This may be due to the effect of oil
on the superheat exchanger surface or the inuence of droplets
in the compressor suction.
Iteration is the only procedure for numeric solution in
system modelling; a skilled iteration method will reduce
computation time. A novel iteration method for reducing
computation time has been proposed here.
Acknowledgements
The nancial support of the EPSRC is gratefully
acknowledged.
References
[1] J. Chi, D. Didion, Asimulationmodel of a heat pumps transient
performance, International Journal of Refrigeration 5 (3)
(1982) 176e184.
[2] EFUNDA website, 2003, <http://www.efunda.com/math/
num_rootnding>.
[3] E.W. Grald, J.W. Macarthur, A moving boundary formula-
tion fpr modeling time-dependant two-phase ows, Inter-
national Journal of Heat and Fluid Flow 13 (3) (1992)
266e272.
[4] J.W. Macarthur, Analytical representation of the transient en-
ergy interactions in vapour compression heat pumps, ASHRAE
Transactions 91 (part 1B) (1984) 982e996.
[5] T. Rossi, J.E. Braun, Development and validation of a real-
time transient model for air-conditioners with on/off control,
in: Proceedings of 20th International Congress of Refrigera-
tion (Sydney), paper no. 743, 1999.
[6] S.M. Sami, T.N. Duong, Y. Mercadier, N. Galanis, Prediction of
the transient response of heat pumps, ASHRAETransactions 93
(1987) 471e490.
[7] M.C. Svensson, None-steady-state modelling of a water-to-
water heat pump unit, in: Proceedings of 20th International
Congress of Refrigeration (Sydney), paper no. 263, 1999.
1.5
2.5
3.5
4.5
5.5
0 200 400 600 800 1000
Time (sec)
T
h
e

l
e
n
g
t
h

o
f

s
u
p
e
r
-
h
e
a
t

s
e
c
t
i
o
n

(
m
)

Fig. 11. The length of the superheat section in the evaporator.
20
21
22
23
24
25
26
27
28
0 200 400 600 800 1000
Time (sec)
T
h
e

l
e
n
g
t
h

o
f

t
w
o
-
p
h
a
s
e

r
e
g
i
o
n

(
m
)

Fig. 12. The length of the two-phase section in the condenser.
ARTICLE IN PRESS
16 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010
[8] F.Q. Wang, The passive use of phase change materials (PCMs)
in refrigeration system, PhD thesis, London South Bank
University, 2003.
[9] H. Yasuda, S. Touber, C.H.M. Machielsen, Simulationmodel of
a vapour compression refrigeration system, ASHRAETransac-
tions 89 (1983) 408e425.
[10] K.E. Gungor, R.H.S. Winterton, Simplied general correlations
for saturated ow boiling and comparisons of correlations
with data, The Canadian Jouranal of Chemical Engineering
Research & Design 65 (2) (1987) 148e156.
[11] S.S. Kandlikar, A general correlation for saturated two-phase
ow boiling heat transfer inside horizontal and vertical tubes,
ASME Winter Annual Meeting, 1987, pp. 14e18.
[12] S.S. Kandlikar, A general correlation for saturated two-phase
ow boiling heat transfer inside horizontal and vertical tubes,
Journal of Heat Transfer 112 (1990) 219e228.
ARTICLE IN PRESS
17 F.Q. Wang et al. / International Journal of Refrigeration xx (2006) 1e17
Please cite this article in press as: F.Q. Wang et al., A novel special distributed method for dynamic refrigeration system simulation, Int. J.
Refrigeration (2006), doi:10.1016/j.ijrefrig.2006.10.010