International Journal of Refrigeration 146 (2023) 15–27

Contents lists available at ScienceDirect

International Journal of Refrigeration

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

Dynamic modelling of the steady state and load processing operation of a

domestic refrigerator cooled through natural convection
Modélisation dynamique du fonctionnement en régime permanent et en charge d’un
réfrigérateur domestique refroidi par convection naturelle

Salman Mustafa HUSAIN a, b, Mutlu İPEK a, *, Akin ÇAĞLAYAN a, Tolga Nurettin AYNUR a,
Serdar KOCATÜRK a, Hasan BEDİR b
a
Arçelik Research and Development Center, Tuzla, 34950 Istanbul, Turkey
b
Boğaziçi Üniversitesi, Bebek, 34342 Beşiktaş/İstanbul, Turkey

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

Keywords: A dynamic model of a domestic refrigerator is developed in this study and is used to analyse both the steady state
Domestic refrigerator (cyclic) and warm load cooling (load processing) operation. The compressor is modelled semi-empirically using
Dynamic modelling loss parameters, while the capillary tube is modelled by incorporating both friction and momentum pressure
Modelica
drop. Heat exchangers (condenser and evaporator) are modelled as tubes divided into finite volume cells. The
Load processing
Natural convection
capillary tube-suction line heat exchanger is treated by connecting part of the capillary tube to the suction line
via a heat resistor. The cabinet model consists of submodels for walls, air and shelves connected in a network to
Mots clés:
Réfrigérateur domestique
allow for thermal interaction. All the components are then arranged in a cycle, and cyclic operation is validated
Modélisation dynamique at 32 ◦ C using a fixed speed compressor. A sensitivity analysis is then carried out by varying the compressor
Modelica speed. In the second part of the study, the model is modified to include the effect of placing a warm load of water
Traitement de la charge in the refrigerator and simulating the subsequent processing operation. This is validated using an experiment in
Convection naturelle which 20 L of water at 27 ◦ C is placed inside the Fresh Food compartment. Results show that the power con­
sumption value is captured well, with the average power consumption deviating less than 10% for all tests. The
energy consumption is within ±5% for all tests. For load processing, the deviation in water temperature was a
maximum of 1.2 ◦ C for the whole cooling period.

Tosun and Tosun (2020) and Yan et al. (2020). Experimentation is time
consuming and costly. Modelling of the refrigeration cycle can speed up
1. Introduction the optimisation process greatly, and several studies pertinent to
refrigerator modelling have been caried out. Ding (2007) and Flores
One of the most important global issues facing us today is global et al. (2015) reviewed the progress of simulation techniques for
warming and depletion of the earth’s energy resources. Therefore, these vapour-compression refrigerant systems. Davis and Scott (1976) laid out
valuable resources must be used as efficiently as possible. To this end, general guidelines for refrigerator modelling; the authors described
optimising the performance of a domestic refrigerator can have a large losses inside the compressor and showed how to calculate the mass flow
impact, since they account for 6% of the world’s electrical energy con­ rate using volumetric efficiency. Several other modelling studies used
sumption (Choi et al., 2018a). steady state approaches to model the refrigeration cycle, often focusing
The bulk of such optimisation has been carried out through experi­ on one component in detail (like the cabinet), such as those by Chen
mental means hitherto. Different parameters such as compressor speed, et al. (1996) and Gupta et al. (2007). Hermes et al. (2009) used the
refrigerator charge amount, capillary tube geometry and ambient tem­ steady state approach to predict the energy consumption of household
perature have been analysed experimentally in studies like the ones by refrigerators. The mass flow rate was calculated using volumetric
Choi et al. (2018b), Diniz et al. (M.C. 2018), Harrington et al. (L. 2018),

* Corresponding author.
E-mail address: [email protected] (M. İPEK).

https://doi.org/10.1016/j.ijrefrig.2022.08.015
Received 11 March 2022; Received in revised form 20 August 2022; Accepted 21 August 2022
Available online 28 August 2022
0140-7007/© 2022 Published by Elsevier B.V.
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Nomenclature RT Run Time (%)

T temperature (K)
A area of wall (m2) t thickness of the wall layer (m) or time (s)
c specific heat capacity (J kg− 1 K− 1) x quality
D diameter (m)
E Energy Consumption (Wh) Greek Symbols
f Darcy friction factor ε absolute surface roughness of the capillary tube (µm) or
G mass flux rate (kg m− 2 s− 1) emissivity
hc convective heat transfer coefficient (W m− 2 K− 1) ρ density, kg m− 3
hc,liq convective heat transfer coefficient in liquid state σ Stefan-Boltzmann constant (W m− 2 K− 1)
(W m− 2 K) Subscripts
hr radiative heat transfer coefficient (W m− 2 K− 1) cold cold wall
k thermal conductivity (W m− 1 K− 1) hot hot wall
L axial length of the capillary tube (m) in inside
Lc Characteristic Length (m) inner inner side of the cabinet
m mass (kg) j wall layer
Nu Nusselt Number l left
P Power (W) out outside
p pressure (Pa) outer outer side of the cabinet
Pr Prandtl Number r right
Q heat transfer rate (W) FF Fresh Food
R thermal resistance (K W− 1) FRZ Freezer
Ra Rayleigh Number
Re Reynolds Number

efficiency while the compressor power was obtained through isentropic refrigerator modelling study heretofore, Çağlayan et al. (2021), devel­
efficiency. Heat exchangers (evaporator and condenser) were modelled oped a dynamic model of a single door refrigerator in Dymola. A
using heat exchanger effectiveness relations. The cabinet model was semi-empirical compressor model was developed with all the loss pa­
treated by equating the heat gain to the total cooling capacity (from the rameters treated explicitly. Also, an algorithm block was added, allow­
evaporator) minus the work done by the fan. The model was validated ing temperature feedback to the compressor (through cut-in and cut-out
against experiment, and the results deviated by 5%. In a similar study, temperatures). The maximum deviation in temperatures and energy
Gonçalves et al. (2009) used empirical relations to obtain compressor consumption were 2 ◦ C and 2% respectively. The present study extends
efficiencies which were used to predict the cooling capacity and power the above by making modifications necessary to model a top-mounted
consumption. The heat exchangers were treated by using different double-door refrigerator having a serial refrigeration system, which
overall heat transfer coefficient values for the subcool, superheat and has not been modelled in studies hitherto. The cyclic operation is
saturated phases. For the capillary tube, experimental data was used to simulated, and a sensitivity analysis for different compressor speeds is
obtain a set of π-groups which were used to calculate the mass flow rate. carried out. In the second part of the study, load processing operation of
The results for compressor power and internal air temperature showed the refrigerator is simulated. To the authors knowledge, no study
5% and 4.5 ◦ C deviation respectively. heretofore has treated the placement of load and door opening as part of
Ding et al. (G. 2004) developed a dynamic model of a natural con­ a dynamic model.
vection bypass cycle refrigerator. In this study, the component models
were presented in detail. The compressor model comprised of two parts: 2. Model
the shell and the cylinder. For the heat exchangers, a multizone model
was used. The capillary tube was modelled into two parts; the adiabatic Each refrigerator component - compressor, condenser, evaporator,
part was treated as a simple steady state tube, while the diabatic part capillary tube, capillary tube-suction line heat exchanger (CT-SL HEX)
was modelled by introducing extra subcooling. As a continuation of the and cabinet - is modelled and then combined into a cycle. Most
above study, Lu et al. (Z. 2004), carried out the actual simulation and component models are presented briefly here; the details can be found in
validation. . Hermes and Melo (2008) performed another dynamic Çağlayan et al. (2021)
modelling study in which the deviation in energy consumption was 10%.
They carried out a subsequent study (Hermes and Melo, 2009) in which 2.1. Compressor
a sensitivity analysis was completed. Bruno et al. (2010) used a some­
what different approach by creating a quasi-steady state model. This was The compressor is treated by determining parameters required to
done to capture the transient behaviour of the air temperature with calculate the losses associated with the compression process. These pa­
reasonable computation effort. In this approach, the system components rameters are calculated through statistical means, by model fitting
(namely, compressor, condenser, capillary tube and evaporator) were simulation results to experimental data (Çağlayan et al., 2021). The
modelled as steady state, whilst the cabinet was modelled as transient, model captures the discharge temperature within 1 ◦ C, and the mass
by treating the compartments as lumped masses. Results were satisfac­ flow rate, input power and cooling capacity within 5%.
tory, with a maximum deviation of 2% in the compressor run-time and
energy consumption. A sensitivity analysis was also carried out for the 2.2. Evaporator and condenser
evaporator and condenser conductance values. In another model by
Heimel et al. (2016), a mean deviation of 7% was observed in the energy The heat exchangers (condenser and evaporator) are treated as tubes
consumption and run-time ratio. Similar studies include the ones by divided into finite volume cells. The condenser is connected to the
Bergera et al. (2015) and Lin et al. (2011). In arguably the most recent ambient, while the evaporator is connected to the cabinet. The heat

16
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

transfer coefficient between the refrigerant and wall is calculated using resistor.
the Shah-Chen (Collier and Thome, 1996) and Shah (Shah, 1979) cor­
relations (for evaporation and condensation respectively). For 2.5.1. Walls of the cabinet
single-phase flow, the well-known Dittus-Boelter correlation is used. As described in the study by Çağlayan et al. (2021), the cabinet walls
Shah-Chen: are made up of three layers: steel, polyurethane and ABS plastic. To
[ ]
0.76
hc = hc,liq (1 − x)0.8 + 3.8x Pr(1− x)0.04
(1) model the thermal capacity of each layer, the layers are divided as
0.38
shown in Fig. 1 below. The relevant heat flows and temperature values
Dittus-Boelter:
are also labelled.
4
Nu = 0.023Re5 Pr3
1
(2) Each layer is divided into two parts (shown by the dashed line) so
that the heat entering the layer is not equal to the heat leaving it. This
naturally implies that each layer has a thermal capacitance associated
2.3. Capillary tube
with it. Each layer is treated as a lumped mass, and the temperature is
therefore given by the 1st Law energy equation for lumped masses:
The capillary tube is modelled as an adiabatic tube. The tube is
divided into discretised cells, and the continuity and momentum equa­ mj cj
dTj
= Qj (5)
tions are solved for each cell. The pressure drop is the main quantity dt
governing the behaviour of the capillary tube, and the equation for
where j denotes the subscript attached to each layer and takes values
pressure drop consists of two main terms- friction pressure drop, and
from 1 to 3, m is the mass of a layer, c is the specific heat, T is the
momentum pressure drop (Çağlayan et al., 2021):
temperature, t is time, and Q is the net heat transfer across a layer.
[( ) ]
f ρ − ρout G2 The net heat transfer across each layer is by conduction, and, using
Δp = ΔL + in (3)
2D ρ ρ Fourier’s law:
The friction coefficient is calculated using Swamee-Jain correlation Tr − Tj T − T
Qj = − kj Aj / + kj Aj j / l (6)
(Swamee and Jain, 1976): tj 2 tj 2
[ ( )]
ε/D 5.74 − 2 where j again denotes the layer in question while r denotes the side to
f = − 2log + 0.9 (4)
3.7 Re the right of the layer and l to the left. For instance, for layer 1, r is 23 and
l is outer. k, A and t are a layer’s thermal conductivity, area, and
2.4. CT-SL HEX thickness respectively.
Finally, at the extreme ends of the wall, heat transfer occurs through
The suction line heat exchanger is modelled separately from the convection, and Newton’s Law of Cooling gives:
capillary tube. The tube heat exchanger model is used again and con­ For the left side (layer 3 side):
nected to another tube heat exchanger at the evaporator outlet to
Qouter = houter Aouter (Tambient − Touter ) (7)
represent the suction line heat exchanger. A resistance element is used to
connect the two tubes. This model is validated against data by Apaydın For the right side (layer 1 side):
(2016), and a deviation less than 2 ◦ C and 10% was observed for the
Qinner = hinner Ainner (Tinner − Tinnerair ) (8)
CT-SL HEX outlet temperature and mass flow rate respectively. Details
can again be found in Çağlayan et al. (2021).
2.5.2. Convection coefficients used in the cabinet model
The outside heat transfer coefficient houter is taken from ASHRAE’s
2.5. Cabinet handbook as 9.26 W/m2 K. This value includes the effect of radiation as
well. For the inside combined heat transfer coefficient hinner , an appro­
The cabinet model consists of numerous wall models, each connected priate way to treat the natural convection and radiation inside the
to an air cell. The air cell in turn is connected to the evaporator via a

Fig. 1. Heat transfer across and temperature of each layer in the wall model.

17
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

compartments is required. For natural convection, two main approaches coefficient is obtained using Eq. 13:
were found in literature (Karatas and Derbentli, 2018; Laguerre et al.,
kNu
2005; Laguerre and Flick, 2004; Ostrach, 1972,)[28–31]): treating the hc = (13)
Lc
compartment as a 2D cavity or applying the vertical and/or horizontal
wall natural convection correlations on the cold evaporator wall, by Radiation also plays a major role in heat transfer and cannot be
treating the core air as stagnant. The latter approach was found to yield ignored [31]. Hence, the radiation heat transfer coefficient is also
better results according to Laguerre and Flick (2004). Laguerre et al. calculated for both the compartments, according to Stefan-Boltzmann
(2008) again validated this claim by measuring the air velocity at Law (Eq. (14)):
various locations. They concluded that the air is indeed almost stagnant ( 4 4
)
σ ∈ Thot − Tcold
at the core. Therefore, this approach is used for both compartments. hr = (14)
Thot − Tcold
The correlations presented by Cengel (2002) were used, as shown in
Eqs. (9)-12.
where ∈= 0.9 according to [31].
For vertical walls and entire Rayleigh number range:
In this equation Thot and Tcold are two wall temperatures. For the
⎧ ⎫
⎨ 1 ⎬ Fresh Food compartment, the natural choice is the back wall (the coldest
0.387Ra6
Nu = 0.825 + [ ] 8 (9) wall behind which the evaporator is placed) and the wall directly
⎩ 9 27 ⎭
1 + (0.492/Pr)16 opposite, namely, the door, since the view factor is maximum in this way
(the evaporator does not "see" the other side walls). For the Freezer, the
For lower surfaces of a cold horizontal plate: top and bottom wall are used, since previous experimental data showed
that the temperature difference between those walls is much higher than
(10)
1
Nu = 0.54Ra4 between the two side walls and the back and front walls.
Ra range: 104 - 107. Once both hr and hc are calculated, the final step is to combine them
using the resistor network approach, as used by Laguerre and Flick
(11)
1
Nu = 0.15Ra3 (2004):
7 11
Ra range: 10 – 10 . 1 1
For the upper surface of a cold horizontal plate: hinner = = (15)
Requivalent h1c + h1r

(12)
1
Nu = 0.27Ra4
2.5.3. The model in Dymola
Ra range: 105 – 1011. The cabinet model in Dymola is as shown in Fig. 2. The important
After calculating the Nusselt number, the convective heat transfer parts of the model are labelled using dashed black boxes. As the

Fig. 2. The cabinet model in Dymola.

18
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Figure shows, the cabinet is divided into two compartments, Freezer and • Wall thicknesses
Fresh Food. The bulk of both compartments is comprised of wall models • Mass of items
(a). In addition, several goods (b) can be seen, which represent the • Air volume
different materials present inside each compartment. The total mass, • Initial temperature values
length and gasket heat transfer coefficient can be entered in the gasket • Heat exchanger and capillary tube geometry
model (c). The air cell model (d) represents the lumped air mass, and is • Compressor loss parameters obtained semi empirically
connected to the evaporator through a heat port (e). The other end of the
air cell is connected to the walls, goods and gasket. Finally, (f) shows the Outputs:
fan model, where the desired combined heat transfer coefficient values
can be entered for both compressor on and off states. • Condensation temperature
• Evaporation temperatures
2.5.4. Cabinet model validation • Cooling Capacity
A temperature rise test was carried out on the refrigerator to validate • Compressor Power input
the cabinet model. In this test, a refrigerator is left to cool to a certain • Mass flow rate
temperature, after which the appliance is unplugged, and the tempera­ • Air temperatures
ture rise curve is obtained. This experimental data was compared to the
simulation results for both compartments. The results can be seen in
Fig. 3. A deviation of less than 2 ◦ C was obtained. 2.7. Modified cabinet model for simulating load processing

2.6. Complete cycle model To perform this simulation, the cycle model was left unchanged,
since the nature of those components do not change with the addition of
The components described above were used to develop the serial goods in the cabinet (the operating conditions, however, do change
system refrigeration cycle in Dymola. Fig. 4 shows the cycle structure. since the inclusion of warm goods increase the evaporation temperature,
The compressor, heat exchangers (as tubes), capillary tube, CT-SL HEX thereby increasing the mass flow rate and also the compressor power
and cabinet can be seen labelled from a through d. The other compo­ input). The cabinet model, however, had to be modified to represent the
nents have been described in detail in Çağlayan et al. (2021). following phenomena:
Inputs, including cabinet geometry, heat exchanger geometry and
capillary tube geometry are entered in the model. In addition, loss pa­ • Opening of the Fresh Food compartment door for a specified time
rameters as outlined in Section 2.1. are entered. The model then makes • An increase in the amount of plastic goods (since the water were
an initial calculation based on initial values given for the evaporation, stored in plastic containers) and “water” good when the water was
condensation, and cabinet temperatures to calculate the mass flow rate, placed inside
cooling capacity and power consumption of the compressor. The cooling
capacity and mass flow rate, along with the capillary tube geometry
determine the evaporation and condensation temperatures and pres­ 2.7.1. Modelling door opening
sures. These new values are then used to update the mass flow rate, As described in the wall model above, all the walls of a refrigerator,
compressor power input and cooling capacity, and this iterative process including the door, are made of plastic, polyurethane and steel, each of
continues until a balance point is reached. The inputs and outputs are which provide a certain resistance to heat conduction from the outer
summarized in the following list: ambient air to the air inside the refrigerator. When the door is opened,
Inputs: the conduction resistance has to go to zero since the door essentially
ceases to exist. To model this phenomenon, the conductivity of the
layers making up the door are connected to a step function, through

Fig. 3. FF and FRZ temperature rise- simulation vs experiment.

19
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Fig. 4. Full cycle model.

which they are set to extremely high values (approaching infinity) while additional components described above are labelled and defined in the
the door is open. With zero conduction resistance, heat gain is maxi­ figure caption:
mised through the door, just would occur in reality. Once the door is
closed the values are returned to the default values. 3. Experimental details
Besides conduction, there is a certain convective heat transfer coef­
ficient inside and outside the refrigerator which sustains convection at To validate the model and to gather data for calculating empirical
both ends. When the door opens, air infiltration occurs, and the parameters such as the heat transfer coefficient, experiments were car­
increased mixing between the cold interior air and the warm exterior air ried out on a top-mounted conventional refrigerator. Details of the
causes the heat transfer coefficient to increase. To capture this, the refrigerator, the experimental setup, and the method used to obtain data
convection values are also modified when the door is opened. In this are now presented.
case, the value was fine tuned to fit the experimental data. After the door
is closed, the values are returned to those before the door was opened. 3.1. Details of the refrigerator

2.7.2. Modelling the increase of mass inside the cabinet The refrigerator consisted of a Freezer (FRZ) compartment and a
Plastic containers filled with water are placed inside the refrigerator Fresh Food (FF) compartment. The Freezer compartment was located at
after it has been running for around 6 h. This is modelled by including the top, while the Fresh Food was located at the bottom of the cabinet
the total amount of water and additional plastic as “goods” in the model. Details:
These goods, however, are connected to the air cell through a resistor
with a variable resistance value. This variable resistance value acts as a • Freezer Volume: 74 L
control switch, effectively turning “off” the effect of goods during the • Fresh Food Volume: 190 L
initial cooling of the refrigerator (i.e., when the additional load is not • FF Evaporator type: Plate-on-tube
present). This is done by setting the resistance value to a very high value, • FRZ Evaporator type: Skin evaporator wrapped around the top, side
approaching infinity. As the resistance approaches infinity, the heat and bottom walls
transfer rate approaches zero, thereby suppressing the presence of • Condenser type: Wire-on-tube
additional goods. After the load is placed and the door is closed, the • Compressor type: Reciprocating Fixed Speed Compressor (3000
resistance value is set such that it approaches zero, so that the heat RPM)
transfer rate between the goods and air returns back to the value gov­ • Refrigerant: 61 g of R600a (Isobutane)
erned by the original convective heat transfer coefficient defined for the
air. This effectively turns “on” the effect of goods.
The final modified cabinet model can be seen in Fig. 5. The

20
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Fig. 5. Modified cabinet model for load processing. a) The FF door connected to various inputs; b) Thermal conductivity inputs for each wall layer; c) Convective
heat transfer inputs for inside and outside the cabinet; d) Resistance inputs for the resistor between goods and air; e) Water and plastic goods; f) variable resistances
between water and plastic goods and air.

3.2. Experimental setup

3.2.1. Temperature measurement

The refrigerator was prepared by placing thermocouples at various
locations (Fig. 6). After preparation, the refrigerator was placed inside a
climate chamber, on a station made of wood -in accordance with the
International Electrotechnical Commission (IEC) Standard for House­
hold Refrigerators (IEC 62,552).].

3.2.2. Other measurements

Besides temperature, the voltage, current, power and energy values
were also measured during the experiments. The uncertainty values for
each sensor are shown in Table 1.
The energy consumption of the refrigerator is calculated by
E = (24)(P)(RT) (16)

where E is the energy in Wh/24 h and P is the power in W. The run time
of the refrigerator is calculated by
on time
Run Time (RT) = (17)
total time

3.3. Experimental method

After the refrigerator was prepared with thermocouples and placed

on the platform in the climate chamber, data was recorded every minute Fig. 6. Thermocouples (marked with blue circles) inside the fresh food
using a Data Acquisition System. To ensure that sufficient cycles are compartment.
present for validation, data was collected for 24 h at an ambient of 32 ◦ C.
To assess the sensitivity of the model for different compressors and

21
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Table 1 • Cyclic operation using a variable speed compressor operated at

Uncertainty values of sensors used in the experiment. different speeds at 32 ◦ C:
Sensor Uncertainty • At 3000 rpm
• At 2500 rpm
Voltage 0.25%
Current 0.25% • At 2000 rpm
Power 0.05% • Load processing test at 32 ◦ C
Temperature (T-type thermocouples) ±0.5 ◦ C
The results and discussion are now presented for each case. Note that
the experimental air temperature values in the relevant graphs below
compressor speeds, simulations were carried out by replacing the orig­
are the average of all the thermocouples placed inside the refrigerator to
inal fixed speed compressor with a variable one. This variable speed
measure the air temperature. Error bars denoting ±0.5 ◦ C (the ther­
compressor was run at three different speeds (3000 rpm, 2500 rpm and
mocouple uncertainty) are also placed on the graphs showing temper­
2000 rpm). Experiments were carried out with this compressor to assess
atures. Error bars for power consumption are not shown because power
the accuracy of the model’s response.
has a very low uncertainty of 0.05%; hence, the bars were not visible on
The second part of this study involved simulation of the load pro­
the graph.
cessing test. A specific amount of warm water is placed in the FF
compartment and the water temperature after 6 h is noted. To validate
4.1. Cyclic operation using a fixed speed compressor at 32 ◦ C
this model; the refrigerator was left to cool down at 32 ◦ C. The ther­
mostat was deactivated inside the refrigerator to prevent cyclic
Fig. 8 shows that the air temperatures are very close, with a
operation.
maximum deviation of 2 ◦ C (for both FF and FRZ). This occurs primarily
After 6 h of pull down, the door of the FF compartment was opened
because air is treated as a “lumped” mass in the model. Also note that the
for 6 min and a total of 20 L of water at 27 ◦ C was placed inside the FF
steep rise at the onset of the heating period (i.e., when the compressor
compartment. Thermocouples were placed inside each container, at the
first switches off) is because of numerical errors as the compressor
very centre of the volume of water. The refrigerator was then left to cool
switches from on to off.
down the water for another 6 h. Again, data was recorded every minute.
Fig. 9 shows the compressor power consumption. A deviation of
Fig. 7 shows the FF compartment in the refrigerator filled with water
3.7% occurs between the average power consumption values, with the
pots.
model slightly understating the average value. This happens because
compressor power is a function of the evaporation temperature. The
4. Results and discussion
evaporator temperature in the simulation reaches the steady state
evaporation temperature (set by the system parameters such as the
Using the modelling and experimental methods described in the
capillary tube length and diameter) more quickly than in reality (see
previous chapters, the following simulations were validated:
Fig. 10). The more gradual drop of the evaporator temperature in reality
is because the actual evaporator is attached directly to the walls of the
• Cyclic operation using a fixed speed compressor (3000 rpm) at 32 ◦ C
compartments and is therefore directly exposed to heat gain from the
outside ambient. This heat gain opposes the tendency of the evaporator
to cool as low pressure refrigerant flows in it. In the model, the evapo­
rator is treated as a separate tube and is not directly exposed to the heat
gain. Another cause for the delay is that in reality the refrigerant takes
some time to fill the evaporator, while in the model, this filling is much
more rapid.
Fig. 10 shows that the actual evaporation temperature (i.e., the
bottom peak for the experimental values, around -28 ◦ C) differs from the
simulated value by a maximum of 0.5 ◦ C. As explained above, in the
simulation the steady evaporation temperature value during compressor
on-time is reached more quickly than in the experiment. This occurs
because of the reasons mentioned above while discussing power,
namely, the direct attachment of the evaporator to the wall, and the time
taken by the refrigerant to fill the evaporator.
Table 2 presents the run time and power for both experiment and
simulation, two essential factors in determining the energy consumption
of a refrigerator. Note that the time-averaged power consumption values
were compared for both the on and off states. The run time shows very
little deviation, with the percentage difference between the two ratios
less than 1.5%. The slightly longer cooling times in simulation are
balanced by the almost equally longer heating times, thereby cancelling
the discrepancy, and leading to nearly equal run times. The power
consumption value shows a deviation of 3.7% during on time and is
mainly responsible for the 2.4% deviation in the energy consumption.

4.2. Cyclic operation using a variable speed compressor

A variable speed compressor was used to test the model’s sensitivity

to a different compressor and different compressor speeds. The ambient
was kept at 32 ◦ C for all simulations (and validation experiments). The
Fig. 7. Fresh food compartment filled with water pots for the load process­ air temperatures were found to be within 3 ◦ C for all speeds. Graphs of
ing test. power consumption can be seen for each speed in Figs. 11–13, while

22
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Fig. 8. FF and FRZ Air temperatures for 32 ◦ C (fixed speed compressor).

Fig. 9. Compressor power consumption for 32 ◦ C (fixed speed compressor).

Fig. 10. Evaporator temperature for 32 ◦ C (fixed speed compressor).

23
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Table 2 4.3. Load processing test results

Comparison of runtime, power and energy- 32 ◦ C, fixed speed compressor.
Experiment Simulation Difference Figs. 14 and 15 show the FF air temperature and FF water
temperature.
Time (min) ON 37.0 39.3 2.3 min
OFF 56.5 58.7 2.2 min It can be seen from the graph above that the refrigerator is left to cool
Run Time (%) – 39.5 40.1 1.4% down until 365 min (~6 h), upon which the door of the FF compartment
Power (W) ON 108 104 -3.7% is opened, and water load is placed inside. A peak is reached, which is
OFF 0 0 0.0% also captured by the simulation, as heat is gained rapidly by the FF
Energy (Wh/24 h) 1025.0 1000.9 -2.4%
compartment. The door is closed, and the warm water load then causes
–

the air temperature to rise further. The temperature peaks at 14 ◦ C,

Table 3 summarizes the results. where after it starts to cool down again. After 6 h, the air temperature in
Table 3 shows the run time, power and energy values for all three the simulation is 10.2 ◦ C, while that of the experiment is 10.9 ◦ C. Also,
rpms. The maximum difference in each is 4%, 5% and 1.6%. The reasons the overall deviation throughout the test is 2 ◦ C, occurring during the
for deviation have been discussed at length in the previous sections. For initial unloaded pull down. Considering the thermocouple uncertainly of
3000 rpm, it is noteworthy that despite the lower power consumption 0.5 ◦ C, these results are quite acceptable.
for this compressor the run time is very similar to the fixed speed Again, as with the air temperatures for the cyclic case, this deviation
compressor (comparing Table 3 with Table 2). This happens because this can be mainly attributed to the lumped approximation of the air inside
new compressor has a greater COP than the previous one. Also, as ex­ the FF cabinet.
pected there is a reduction in power from 86 W at 3000 rpm to 70.7 W at After 6 h, as mentioned above, water is placed inside the FF
2500 rpm. At 2000 rpm, the high runtime is not surprising, since the compartment. Although the door of the FRZ compartment is not opened,
lower cooling capacity leads to longer cooling times. the FRZ air temperature starts to rise. This happens because the

Fig. 11. Compressor power consumption (variable speed compressor- 3000 rpm).

Fig. 12. Compressor power consumption (variable speed compressor- 2500 rpm).

24
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Fig. 13. Compressor power consumption (variable speed compressor- 2000 rpm).

5. Conclusion
Table 3
Run time, power and energy for 3000 rpm, 2500 rpm and 2000 rpm.
A top-mounted, serial system conventional refrigerator was
Experiment Simulation Difference modelled in this study by creating a sub-model for each component
3000 Time (min) ON 37.2 36.5 − 0.7 min present in the refrigerator and then combining all the models into a
RPM OFF 53.8 53.7 − 0.1 min refrigeration cycle. The TIL library in Dymola, based on the program­
Run Time (%) 40.9 40.5 − 1.0%
–
ming language Modelica was used to carry out the modelling.
Power (W) ON 86.5 83 − 4.0%
OFF 0 0 0.0%
The cyclic state of the refrigerator was simulated at 32 ◦ C and vali­
Energy (Wh/ – 848.7 806.1 − 5.0% dated using experiments carried out in a climate chamber. To test the
24 h) sensitivity of the model to a different compressor type at different
2500 Time (min) ON 71.9 73 1.1 min compressor speeds, a variable speed compressor was used to simulate
RPM OFF 72.9 71.7 − 1.2 min
the model at three different speeds. Validation was carried out again,
Run Time (%) – 49.7 50.4 1.6%
Power (W) ON 70.7 69.6 − 1.6% and the results were indeed very close.
OFF 0 0 0.0% As a second part of the study, the cabinet model was modified to take
Energy (Wh/ – 842.5 842.7 0.02% into account the effect of loading a warm load of water into the FF
24 h) compartment and simulate the subsequent cooling. The effect of door
2000 Time (min) ON 296.7 293.8 − 2.9 min
RPM OFF 108.7 110.3 1.6 min
opening was also included in the model. Again, validation was carried
Run Time (%) – 73.2 72.7 − 0.7% out by performing experiments, and the results showed less than 2 ◦ C
Power (W) ON 53.9 55.8 3.5% difference in the air and water temperatures.
OFF 0 0 0.0% Some key takeaways from this study are as follows:
Energy (Wh/ – 946.7 973.6 2.8%
24 h)
• The slight difference in air temperatures and water temperature
(~2–3 ◦ C) occurs due to the lumped approximation method.
evaporator temperature in the FF compartment rises, causing refrigerant • The evaporator temperatures in the model reaches the actual steady
temperature in the FRZ evaporator to rise as well. In reality, the evap­ state evaporation value almost instantaneously, compared to the real
orator in the FRZ is directly attached to the walls of the FRZ, which case, in which there is a time lag. This lag occurs because the real
causes the walls and consequently the air inside to rise steeply. In the evaporator is attached directly to the wall of the compartments,
simulation, the evaporator is treated as a separate tube, causing no rise causing a heat gain through the wall and thus a more gradual tem­
in the temperature. perature drop. Another cause is the time taken by the refrigerant to
At the 365th minute, after about 6 h of unloaded pull down, water at fill the evaporator.
27 ◦ C is placed inside the cabinet. The maximum deviation of 1.2 ◦ C • This difference in evaporation temperature causes the power values
occurs around the 450th minute. The primary cause of this is once again in the model to reach the final value much quicker compared to the
the lumped mass approximation used in the simulation. Four water experiment and leads to the slight deviation in power.
containers are placed close together inside the refrigerator. Neverthe­ • The run time is very well captured by the model for all cases. The
less, some variation of temperature is to be expected, both within each energy values are also in very good agreement, with a maximum
container as well as across all the containers. This slight deviation, deviation of 5% for the variable speed compressor operated at
however, is hardly a problem, especially while evaluating the cooling 3000 rpm.
performance using this test, in which the temperature after a predefined • The load processing simulation is also in excellent agreement with
time (for example 6 h) is more important. experiment, with a deviation of 1.2 ◦ C for water temperature. The

25
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Fig. 14. FF air temperature for the load processing test.

Fig. 15. FF water temperature for the load processing test.

major cause can be attributed to the lumped mass approximation of References

water.
Apaydın, T., 2016. Modelling For Refrigerant Flow Of R600a Through Diabatic Capillary
Tube In Vapor Compression System. Yıldız Technical University.
This model can be used as a basis for future studies in which the Bergera, E., Poscha, S., Heimela, M., Almbauera, R., Eichingera, M., Stupnik, A., 2015.
evaporator may be modelled in more detail, so that the lag in evaporator Transient 1D heat exchanger model for the simulation of domestic cooling cycles
temperature may be captured. Also, a discretised cabinet model may be working with R600a. Sci. Technol. Built Environ. https://doi.org/10.1080/
23744731.2015.1057084.
created, so that the spatial variation in air temperature can be modelled Bruno, B., C. Hermes, C. Melo and G. Goncalves, “Transient simulation of household
as well. The model may also be used to optimise the algorithm of a refrigerators: a semi-empirical, quasi-steady approach”. international refrigeration
refrigerator. Finally, frost accumulation on the evaporator and cabinet and air conditioning conference. Paper 1018. http://docs.lib.purdue.edu
/iracc/1018, 2010.
inner walls (especially the FRZ compartment) may also be studied in the Cengel, Y., 2002. Heat Transfer- A Practical Approach, 2nd Edition. McGraw-Hill, New
future by extending this model to create more sophisticated ones. York.
Chen, L., Sun, F., Wu, C., Kiang, R.L., 1996. A generalized model of a real refrigerator and
its performance. Appl. Therm. Eng. 17, 401–412.
Declaration of Competing Interest Choi, S., Han, U., Cho, H., Lee, H., 2018a. Review: recent advances in household
refrigerator cycle technologies. Appl. Therm. Eng. 132, 560–5746.
Choi, Y., Min, B., Kim, H., Choi, G.M., 2018b. Experimental study on effects of lubricant
The authors declare that they have no known competing financial oil in a domestic refrigerator-freezer. In: 17th International Refrigeration and Air
interests or personal relationships that could have appeared to influence Conditioning Conference at Purdue.
the work reported in this paper.

26
S.M. HUSAIN et al. International Journal of Refrigeration 146 (2023) 15–27

Collier, J.G., Thome, J.R., 1996. Convective Boiling and Condensation, 3rd ed. Oxford Karatas, H., Derbentli, T., 2018. Natural convection and radiation in rectangular cavities
University Press, New York. with one active vertical wall. Int. J. Therm. Sci. 123, 129–139.
Çağlayan, A., Husain, S.M., İpek, M., Aynur, T.N., Çadırcı, S., 2021. Dynamic modeling Laguerre, O., Amara, S.B., Flick, D., 2005. Experimental study of heat transfer by natural
and experimental validation of a domestic refrigeration cycle. J. Therm. Sci. Eng. convection in a closed cavity: application in a domestic refrigerator. J. Food Eng. 70,
Appl. 14 (7), 071007. ASME. 523–537.
Davis, L., Scott, T.C., 1976. Component, modeling requirements for refrigeration system Laguerre, O., Ben Amara, S., Charrier-Mojtabi, M.C., Lartigue, B., Flick, D., 2008.
simulation. Int. Compress. Eng. Conf., Paper 221. https://docs.lib.purdue.edu/icec Experimental study of air flow by natural convection in a closed cavity: application
/2211976. in a domestic refrigerator. J. Food Eng. 85, 547–560.
Ding, G., 2007. Recent developments in simulation techniques for vapour-compression Laguerre, O., Flick, D., 2004. Heat transfer by natural convection in domestic
refrigeration systems. Int. J. Refrig. 30, 1119–1133. refrigerators. J. Food Eng. 62, 79–88.
Ding, G., Zhang, C., Lu, Z., 2004. Dynamic simulation of natural convection bypass two- Lin, E., Ding, G., Zhao, D., Liao, Y., Yua, N., Yamashita, J., 2011. Dynamic model for
circuit cycle refrigerator–freezer and its application. Part I: component models”. multi-compartment indirect cooling household refrigerator using Z-transfer function
Appl. Therm. Eng. 24, 1513–1524. based cabinet model. Int. J. Therm. Sci. 50, 1308–1325.
Diniz, M.C., Melo, C., Deschamps, C.J., 2018. Experimental performance assessment of a Lu, Z., Ding, G., Zhang, C., 2004. Dynamic simulation of natural convection bypass two-
hermetic reciprocating compressor operating in a household refrigerator under circuit cycle refrigerator–freezer and its application. Part II: system simulation and
on–off cycling conditions. Int. J. Refrig. 88, 587–598. application”. Appl. Therm. Eng. 24, 1525–1533.
Flores, J.M.B., Maldonado, J.M.B., Muñoz, A.P.R., Vázquez, G.C., 2015. Enhancements in Ostrach, S., 1972. Natural Convection in Enclosures. Adv. Heat Transf. 8, 161–227.
domestic refrigeration, approaching a sustainable refrigerator – a review”. Ren. Shah, M.M., 1979. A general correlation for heat transfer during film condensation inside
Sustain. Energy Rev. 51, 955–968. pipes. Int. J. Heat Mass Transf. 22, 547–556.
Gonçalves, J.M., Melo, C., Hermes, C.J.L., 2009. A semi-empirical model for steady-state Swamee, P.K., Jain, A.K., 1976. Explicit equations for pipe flow problems. J. Hydraul.
simulation of household refrigerators. Appl. Therm. Eng. 29, 1622–1630. Div. 102, 657–664.
Gupta, J.K., Gopal, M.R., Chakraborty, S., 2007. Modeling of a domestic frost-free Tosun, T., Tosun, M., 2020. Heat exchanger optimization of a domestic refrigerator with
refrigerator. Int. J. Refrig. 30, 311–322. separate cooling circuits. Appl. Therm. Eng. 168, 114810.
Harrington, L., Aye, L., Fuller, B., 2018. Impact of room temperature on energy Yan, G., Liu, G., Zhou, M., Yu, J., 2020. Experimental research on configuration
consumption of household refrigerators: lessons from analysis of field and laboratory optimization of a frost-free refrigerator-freezer with the parallel circuit cycle. Int. J.
data. Appl. Energy 211, 346–357. Refrig.
Heimel, M., Berger, E., Posch, S., Stupnik, A., Hopfgartner, J., Almbauer, R., 2016.
Transient cycle simulation of domestic appliances and experimental validation. Int.
J. Refrig. 69, 28–41.
Further reading
Hermes, C.J.L., Melo, C., 2008. A first-principles simulation model for the start-up and
cycling transients of household refrigerators. Int. J. Refrig. 32, 1341–1357. ASHRAE Handbook of Fundamentals, Ref. 1, Chap. 22, Table 1. 2022.
Hermes, C.J.L., Melo, C., 2009. Assessment of the energy performance of household IEC 62552:2015, 2015. IEC 62552:2015 Household Refrigerating Appliances
refrigerators via dynamic simulation. Appl. Therm. Eng. 29, 1153–1165. –Characteristics and Test Methods. International Electrotechnical Commission (IEC).
Hermes, C.J.L., Melo, C., Knabben, F.T., Gonçalves, M., 2009. Prediction of the energy
consumption of household refrigerators and freezers via steady-state simulation.
Appl. Energy 86, 1311–1319.

27