Applied Thermal Engineering 31 (2011) 1114e1124

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

A semi-empirical model for the unsteady-state simulation of reciprocating

compressors for household refrigeration applications
Cezar O.R. Negrão a, *, Raul H. Erthal a, Diogo E.V. Andrade a, Luciana Wasnievski da Silva b
a
Thermal Science Laboratory e LACIT, Federal University of Technology-Paraná e UTFPR, Av. Sete de Setembro, 3165 Curitiba, PR, Brazil
b
Whirlpool S/A e Unidade EMBRACO Rua Rui Barbosa, 1020 Caixa Postal 91, Joinville, SC, Brazil

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

Article history: The current work presents a semi-empirical mathematical model to simulate the unsteady behavior of
Received 5 August 2010 mass flow rate and power of reciprocating compressors. The model is based on thermodynamic equations
Accepted 1 December 2010 fitted to thirteen calorimeter data sets of two compressors with different capacities. The curve fitting
Available online 8 December 2010
suggests linear correlations between the measured values and the thermodynamic equations, indicating
that only two calorimeter data sets are necessary for curve fitting. Additionally, the proposed model was
Keywords:
also fitted to data sets of 21 other compressors obtained directly from the catalog of two different
Reciprocating compressor
manufactures. Besides, comparisons of computed and measured values of mass flow rate and power, in
Modeling
Household application
transient regime, were conducted for the two fitted compressor curves. A good agreement of results was
Transient state found for both compressors in either cycling or start-up tests. It was found that the proposed semi-
empirical model can be satisfactorily applied to dynamic simulations of the whole refrigeration system.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction second approach, on the other hand, is based on simple thermo-

dynamic correlations fitted to experimental data [4e11]. The third
According to Eletrobrás [1], the household appliances are class is used to study details of the compressor design such as valve
responsible for 33% of the whole energy consumed in Brazil and flows, cylinder heat transfer, cylinder-piston leakage, bearing los-
domestic refrigerators have a great importance on it. Although the ses, among many others, but requires large amount of data, e.g.,
unitary consumption is low (for example, a typical refrigerator valve geometry and stiffness, bearing clearances, which are only
expends about 30 kWh per month) 11% of all produced electricity is available to manufacturers [12e15]. Computational Fluid Dynamics
employed in domestic refrigeration. Despite the considerable (CFD) models are included in the last class, in which fluid and heat
improvement the performance of refrigeration systems has ach- flows within the cylinder and through the valves are considered
ieved in the last few years, there is still a great potential for energy three dimensional and turbulent [16,17].
consumption reduction as the system efficiency is smaller than 20%. Although the household refrigeration compressors work most of
The compressor most widely used in domestic refrigeration is the time in unsteady-state mode, e.g., system start-up and cycling
the reciprocating one. This component expends most of the system- transients, the above-mentioned models were developed for
required energy, if not all, and therefore, its performance enhance- steady-state situations as compressor design is based in a single
ment is one of the main issues of system designers. Prediction of the condition. The compressor transient modeling is usually useful
compressor behavior can help designers to find out alternative when the concern is the performance of the whole refrigeration
solutions to reduce the refrigerator energy consumption [5,6,16]. system. The complexity of transient process models, such as
Several works have addressed the modeling of reciprocating condensation, evaporation and compressible fluid expansion,
compressors for refrigeration appliances. Such mathematical entails time-extensive system simulations. As detailed compressor
models can be divided into three categories: polynomial fits, semi- models also demand high computing times they are generally
empirical and detailed models. In the first case, the correlations are avoided in transient system simulations. The polynomial fits, on the
fitted to calorimeter data, but they do not have any physical other hand, should not be used as they do not have any physical
meaning and cannot be applied out of the test range [2,3]. The meaning. Some authors [18e22] applied semi-empirical models to
transient simulations of the whole refrigeration system, as the
* Corresponding author. Tel.: þ55 41 3310 4658; fax: þ55 41 3310 4852. computing time is quite adequate for this kind of simulation.
E-mail address: negrao@utfpr.edu.br (C.O.R. Negrão). Nevertheless, validation is still a matter of concern.

1359-4311/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2010.12.006
 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124 1115

time of the whole refrigeration system. Therefore, the current model

Nomenclature considers the compression process as a quasi-steady one [20,21].

a linear coefficient of Eq. (6)

2.1. Mass flow rate
b angular coefficient of Eq. (6)
c clearance fraction
At the top dead position of a compressor piston, there always
k refrigerant isentropic coefficient
remains a small volume of vapor to accommodate valve ports and
m_a actual mass flow rate, kg s1
production tolerances. This clearance volume reduces the amount
m_i ideal mass flow rate, kg s1
of vapor that enters the cylinder. In order to account for the effect of
pd discharge pressure, Pa
the clearance volume, the ideal mass flow rate (V_ sw =vs ) is multi-
ps suction pressure, Pa
plied by a volumetric efficiency [24]:
V_ sw compressor swept volume rate, m3 s1
vs specific volume at the suction port, m3 kg1
V_ sw
wa actual compressor work, J kg1 _i ¼
m h (1)
vs vi
wi isentropic compressor work, J kg1
W_ compressor losses, W where m _ i is the mass flow rate of an ideal compressor with clear-
l
_
Wa actual compressor power, W ance, V_ sw is the compressor swept volume rate, vs is the specific
volume at the suction port, computed as a function of pressure and
Greek symbols temperature by employing the refrigerant superheating properties.
hg compressor overall efficiency hvi is the volumetric efficiency of the ideal compressor with clear-
hva actual volumetric efficiency ance, given by:
hvi ideal volumetric efficiency

 1=k 
hv volumetric efficiency ratio pd
hvi ¼ 1  c 1 (2)
ps
Some comparisons of model predictions with experimental data c is the clearance fraction, pd and ps are, respectively, the discharge
have been found in the literature [20,21], but they are usually and suction pressures and k is the refrigerant isentropic coefficient.
conducted at the system level. Validation at the component level is Similarly to the ideal compression, the actual mass flow rate can
also important because two poor component models may provide be defined as
reasonable global results, as their discrepancies can be counter-
balanced by one another. This is particularly true for the semi- V_ sw
_a ¼
m h (3)
empirical compressor models which are quite popular in transient vs va
system simulation but have not yet been completely validated [23].
where hva is the volumetric efficiency of the real compressor.
Considering this lack in the literature, the current work
However, due to compression irreversibility, piston-cylinder leak-
proposes a semi-empirical model to predict the performance of
ages, suction and discharge valve throttling, suction gas heating
reciprocating compressor in transient regime. The model is based
and gas-to-cylinder wall heat transfer, hva will be smaller than the
on thermodynamic equations fitted to manufacturer data by using
ideal counterpart and consequently, the actual mass flow rate, m _ a.
linear correlations. Comparisons with experimental data show that
these simple fitted correlations are quite accurate on the prediction
of compressor mass flow rate and power. 2.2. Compressor work

2. Mathematical model The specific compression work can be computed by the inte-
H
gration ðvdpÞ over an indicator (p-V) diagram of an ideal recipro-
Because of the compressor high rotational speed, the refrigerant cating compressor with clearance. By considering the compression
compression is quite instantaneous in comparison to the response as an isentropic process (pvk ¼ const), one can find [24]:

Table 1
Calorimeter test results for compressors X and Y. Ambient temperature ¼ 32  C. Refrigerant R134a.

Evaporating Measured variable Compressor X Compressor Y

temperautre ( C)
Condensing temperature ( C) Condensing temperature ( C)

45 55 60 45 55 60
35 Mass flow rate (kg h1) 2.21 1.92 1.75 2.48 2.01 1.79
Compressor power (W) 102.68 102.53 100.65 92.0 87.0 83.5
Compressor shell temperature ( C) 64.4 65.9 66.8 63.9 64.5 60.0
Discharge line temperature ( C) 66.2 67.3 67.1 63.6 64.2 60.3
Suction line temperature ( C) 39.4 40.6 39.3 40.5 44.5 41.2
25 Mass flow rate (kg h1) 4.41 3.90 3.95 4.64 4.28 4.03
Compressor power (W) 145.9 148.9 150.0 126.4 130.8 130.7
Compressor shell temperature ( C) 67.4 70.1 72.3 63.5 65.1 66.0
Discharge line temperature ( C) 75.1 81.4 82.5 72.9 74.8 76.6
Suction line temperature ( C) 35.6 39.8 40.5 38.8 40.0 41.0
15 Mass flow rate (kg h1) 7.69 7.04 6.74 7.77 7.38 7.16
Compressor power (W) 192.5 202.5 210.1 162.4 175.0 179.8
Compressor shell temperature ( C) 68.1 71.9 74.1 62.3 65.9 66.5
Discharge line temperature ( C) 80.9 87.0 91.6 71.2 78.9 80.6
Suction line temperature ( C) 30.9 32.5 34.8 29.9 34.6 34.0
1116 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124

Table 2 on the compressor performance, three other tests were performed

Calorimeter tests for three shell temperatures. Ambient temperature ¼ 32.0  C, by varying the compressor shell temperature and the results are
condensing temperature ¼ 55  C, evaporating temperature ¼ 35  C.
presented in Table 2. An additional test with a 43  C ambient
Measured variable Compressor X Compressor Y temperature was conducted to check if the shell temperature could
Compressor shell Compressor shell be really used as an independent variable. Those results are shown
temperature ( C) temperature ( C) in Table 3.
60 75 90 60 75 90 As expected and shown in Table 1, both mass flow rate and
Mass flow rate(kg h1) 1.96 1.91 1.87 2.01 2.01 1.98
compressor power are more sensitive to the evaporating than to the
Compressor power (W) 101.8 101.2 101.45 87.17 86.35 86.72 condensing temperature, because low pressures affects more
Discharge line 63.8 76.3 84.0 58.5 68.7 83.2 significantly the pressure ratio. On the other hand, the mass flow
temperature ( C) rate varies slightly and the compressor power does not alter with
Suction line 39.1 41.0 43.1 40.2 43.4 44.6
the shell temperature, as shown in Table 2.
temperature ( C)
Besides, the comparison of Tables 2 and 3 shows that the mass
flow rate is as much insensitive to the ambient temperature as to

 k1  the shell temperature. Note that the shell temperature of
k pd k compressor X in Table 3 is approximately 75  C and that the mass
wi ¼ ps vs 1 (4)
k1 ps flow rate in such condition is identical to mass flow rate for the
shell temperature of 75  C in Table 2. The insensitivity of the mass
The actual compressor work, which is always larger than its
flow rate must be related to the type of gas suction; semi-direct for
isentropic counterpart due to the thermodynamic and mechanical compressor X and direct for compressor Y. Additionally, both
losses, can be simply calculated from
compressor powers do not change with either ambient or shell
_ a temperatures. As the product of Eqs. (1) and (4) shows that the
W
wa ¼ (5) compressor power is independent of the specific volume, the
_a
m compressor shell temperature should not affect it.
_ a is the actual compressor power (measured).
where W
4. Model calibration
3. Experimental work
The difference between Eqs. (1) and (3) is the volumetric effi-
By analyzing the isentropic model presented above, it can be ciency. As already mentioned, the value of the actual volumetric
seen that the variables affecting the compressor performance are: efficiency is lower than its ideal counterpart because of thermody-
evaporating and condensing pressures (or temperatures), and gas namic losses and gas leakages that depend on operating conditions.
specific volume at the suction port (indirectly, the suction pressure
and temperature). Additionally, ambient conditions (air speed and
temperature) also affect the performance of the real compressor a 1

because of heat losses.

In order evaluate the performance of the two compressors 0.9
manufactured by EMBRACO (nominated here X and Y), thirteen
calorimeter tests were conducted by controlling condensing, evap- 0.8
orating, suction line, ambient and compressor shell temperatures, as
well as the air speed. The tests were performed according to [3]. Nine
0.7
tests were performed at the ambient temperature of 32  C,
combining three evaporating temperatures (35, 25, 15  C) and
three condensing temperatures (45, 55 and 65  C). Table 1 shows the 0.6
measured values of mass flow rate, compressor power, shell,
discharge line and suction line temperatures for those nine tests. The 0.5
uncertainties of the measured powers and mass flow rates are 5 10 15 20 25 30
þ/5%. The refrigerant employed was the R134a. The compressor pd p s
shell temperature was measured at the shell middle height. The
suction and discharge line temperatures were measured 100 mm b 1

away from the compressor shell.

Assuming that the shell temperature depends directly on the 0.9
ambient condition (air temperature and convection coefficient),
which is quite different from that observed when the whole 0.8
refrigeration system is tested, it was then considered to be an
independent variable. To verify the effect of the ambient condition
0.7

Table 3
Calorimeter tests for ambient temperature ¼ 43  C. Condensing temperature ¼ 55  C, 0.6
evaporating temperature ¼ 35  C.

Measured variable Compressor X Compressor Y

0.5
Mass flow rate (kg h1) 1.91 2.02 5 10 15 20 25 30
Compressor power (W) 102.1 87.0
Compressor shell temperature ( C) 74.8 68.6
pd p s
Discharge line temperature ( C) 77.1 68.6
Suction line temperature ( C) 49.0 47.5 Fig. 1. Actual-to-ideal mass flow rate ratio as a function of the pressure ratio. (a)
Compressor X and (b) compressor Y.
 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124 1117

Table 4
Fitted constants of Eqs. (6) and (7) for compressor X and Y.

Compressor Mass flow rate Compressor power

a () b () R2 Largest Difference [%] _ (W)

W hg () R2 Largest Difference [%]
l

X 1.0282 0.01781 0.985 2.5 31.59 0.7860 0.997 2.8

Y 1.0579 0.01733 0.971 4.6 25.09 0.9398 0.999 1.5

As heat transfer and leakages are intensified with the increase of the
discharge-to-suction pressure ratio, one may suggest that the
a 220
actual-to-ideal volumetric efficiency ratio is also pressure ratio
dependent. Fig. 1, for example, shows the actual-to-ideal mass flow
Measured Compressor Power (W)

200 rate ratio as a function of the discharge-to-suction pressure ratio, for

compressors X and Y. The clearance fractions of compressor X and Y
180
needed for the calculation of the ideal volumetric efficiency were
160 provided by the manufacturer. As can be seen, the higher the pres-
sure ratio the lower the mass flow rate ratio and its value is always
140
less than one. The coefficients of determination (R2) of the straight
120 lines of Fig. 1a and b are, respectively, 0.985 and 0.971. As the suction
port temperature is not usually measured in calorimeter tests, the
100
compressor shell temperature was used to evaluate the gas specific
80 volume in Eq. (1). This implies the mass flow rate ratio is a linear
50 60 70 80 90 100 110 120 130 140
function of the pressure ratio.
Mass flow rate x Isentropic work (W)
Thus, from Fig. 1, a straight line can be proposed as a correlation
b 200 between the mass flow rate ratio and the pressure ratio:
Measured Compressor Power (W)

 
180 _a
m p
¼ hv ¼ a þ b d (6)
_i
m ps
160
where hv is the volumetric efficiency ratio (¼hva/hvi), a and b are the
140 linear and the angular coefficient of the straight line which are
fitted to calorimeter data. The a and b coefficients for compressors
120
X and Y are shown in Table 4. According to equation (6), hv gets
100 close to the unity as the discharge pressure approaches the suction
pressure. This would be expected as the losses and leakages
80 decrease with the drop in the pressure ratio. Table 4 also shows the
50 60 70 80 90 100 110 120 130 140 150
coefficient of determination and the largest difference between the
Mass flow rate x Isentropic work (W)
measured and calculated mass flow rates. Note that the largest
Fig. 2. Compressor power as a function of the product of the mass flow rate and the difference lies within the 5% uncertainty estimated for the calo-
isentropic compressor work. (a) Compressor X and (b) compressor Y. rimeter tests.

Table 5
Calibration data for EMBRACO and TECUMSH catalog compressors.

Model Manufacturer Mass flow rate (kg/s) Power (W)

a b R 2 _
W hg R2
l

1 AE1390YS-AE390KS TECUMSEH 0.7639 0.01623 0.974 85.17 0.6863 0.983

2 AEA2410YXA-AE600AR TECUMSEH 0.5683 0.01205 0.974 61.07 0.4344 0.988
3 AZ1320DS-AZ224TS TECUMSEH 0.7284 0.02426 0.991 8.46 0.3896 0.975
4 AZ1320YS-AZ410NS TECUMSEH 0.7434 0.01630 0.951 24.36 0.3871 0.982
5 AZ1345YS-AZ440NS TECUMSEH 0.7199 0.01579 0.950 36.23 0.4757 0.983
6 AZ1355YS-AZ445RS TECUMSEH 0.7834 0.01718 0.952 45.47 0.5170 0.983
7 EG75HLR EMBRACO 0.8788 0.01354 0.963 16.62 0.6053 0.991
8 EGX70HLC EMBRACO 1.0132 0.01594 0.958 12.89 0.7716 0.997
9 EGZS80HLC EMBRACO 1.0100 0.01854 0.980 19.39 0.7880 0.998
10 EMU26CLC EMBRACO 0.9363 0.01449 0.944 23.05 0.7369 0.975
11 EMY26CLC EMBRACO 0.9112 0.01172 0.934 12.95 0.6988 0.996
12 FFI85HAK EMBRACO 0.8813 0.01072 0.927 47.11 0.7189 0.995
13 FFL10HAK EMBRACO 0.8907 0.01609 0.972 60.56 0.7231 0.995
14 THG1330MDS-TH611DS TECUMSEH 0.8289 0.01720 0.963 24.79 0.6566 0.990
15 THG1340MGS-TH631GS TECUMSEH 1.0079 0.02114 0.961 31.16 0.6959 0.986
16 THG1345MDS-TH671DS TECUMSEH 0.8055 0.01702 0.959 37.25 0.7067 0.985
17 TP1360MS-TP302DS TECUMSEH 0.9600 0.01815 0.921 47.19 0.7536 0.999
18 TP1370MS-TP303DS TECUMSEH 1.1263 0.02133 0.922 54.53 0.8264 0.999
19 TP1380MS-TP305RS TECUMSEH 0.8886 0.01682 0.926 61.99 0.7918 0.998
20 TPH1415YGS-TP156GS TECUMSEH 0.7409 0.00727 0.869 66.80 0.6911 0.995
21 NB2121GK EMBRACO 0.7741 0.02328 0.985 51.41 0.5211 0.966
1118 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124

18 Table 6
Characteristics of the refrigeration system.
16
Refrigeration system Vertical freezer
14 Capacity 300 L
Measured mass flow rate (kg/h)

Evaporator type Roll-bond (free convection)

12 Condenser type Wire-and-tube (free convection)
Capillary tube-suction line heat exchanger Concentric counter-flow
10 ± 10% Fluid refrigerant R134a

4 Table 7
Measuring instruments and their uncertainties.
2
Variable Instrument Uncertainty
0 Temperature T type thermocouple 0.2  C
0 2 4 6 8 10 12 14
Discharge pressure Absolute pressure transducer 0.03 bar
Calculated mass flow rate (kg/h) Suction pressure Absolute pressure transducer 0.03 bar
Mass flow rate Coriolis 0.012 kg h1
Fig. 3. Comparison of measured and computed mass flow rates for different compressor
Compressor power Wattmeter 0.2% of the measuring
models.
value

Despite losses and thermodynamic irreversibility, the com-

pressor power must be related to the thermodynamic compressor
work. In order to check the correlation, the measured compressor
power was plotted against the product of mass flow rate and
isentropic compression work for both compressors X and Y, as a 10 20

presented in Fig. 2. Straight lines are fitted to the points of Fig. 2a

6.5 8

Mass flow rate [kg/h]

6
and b and the coefficients of determination are, respectively, 0.997 6

Difference [%]
5.5 15
and 0.999. The following correlation was thus proposed for the 8 5
4

calculation of compressor power: 4.5

2
Mass flow rate [kg/h]

4 0
10
_ þm_ a wi

Difference [%]
_ a ¼ m
W _ a wa ¼ W (7) 0 10 20 30 40 50
l
hg 6 Time (min)

5
where W _ and 1/hg are, respectively, the linear and angular coeffi-
l
cients of the straight line. W _ is suggested to be the power
l Experiment
consumption for the unloaded compressor. The unloaded 4 Model 0
compressor still consumes energy to overcome the losses even if Difference

the refrigerant is not been compressed. On the other hand, hg is

-5
a thermodynamic efficiency of the compression process. The values
of W_ and 1/hg are also shown in Table 4, together with the coef- 2
l
ficient of determination and the largest difference between the -10
measured and calculated power for compressors X and Y. 0 50 100 150 200 250 300
The linear functions (6) and (7) suggest that only two conditions Time (min)
would be necessary to calibrate either the mass flow rate or the
power model. To avoid extrapolations, the maximum and the
b 10 20
6

15
Mass flow rate [kg/h]

5.5
600 8
5
Mass flow rate [kg/h]
Measured compressor power (W)

500 10
Difference [%]

4.5
6
400 4
0 10 20 30 40 50 5
± 10% Time (min)

300 4 Experiment
Model 0
Difference
200
2
-5
100

0 0 -10
0 100 200 300 400 500 0 50 100 150 200 250 300
Calculated compressor power (W) Time (min)

Fig. 4. Comparison of measured and computed compressor powers for different Fig. 5. Comparison of the measured and computed mass flow rate for the start-up test.
compressor models. (a) Compressor X and (b) compressor Y.
 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124 1119

minimum values of mass flow and power found within an estab- Other compressors
lished range of evaporating and condensing temperatures should
be used in the calibration. Note that the maximum and the As the required data for the curve fitting are those available in
minimum mass flow rate take place at the lowest and the highest manufacturer catalogs, the current model can be easily applied to
pressure ratio, respectively. other compressors. Curve fitting was conducted for nine EMBRACO
According to Eq. (7), the highest and the lowest measured [25] and twelve TECUMSEH [26] compressors. Their capacities
compressor powers occur at the maximum and minimum isen- range from 70 to 400 W and they employ three types of refriger-
tropic powers, respectively. The product of Eqs. (1) and (4) can be ants: eleven use R134a, eight R600a and the other two R404a.
used to identify the maximum and minimum values of isentropic Those tests were performed in different conditions, depending on
power within an established range of evaporating and condensing the compressor type. In general, the evaporating temperatures
temperatures. The analysis of Eqs. (1) and (4) shows the isentropic range from 40 to 10  C and the condensing temperatures, from
power is either zero at zero mass flow rate (the highest pressure 35 to 65  C. The shell temperatures are not available at the catalog
ratio) or zero compressor work (pressure ratio equal to zero). in order to evaluate the gas specific volume. As discussed in Section
Therefore, the minimum power can either take place at the lowest 4.1, the temperature chosen for the evaluation of the specific
or at the highest pressure ratio within the established range of volume does not affect significantly the curve fitting and therefore,
evaporating and condensing temperatures. On the other hand, the the ambient temperature is used instead. Besides, the compressor
maximum isentropic power usually occurs at the highest evapo- clearance fractions were also not available and consequently, they
rating pressure combined to the highest condensing pressures. In were estimated. A sensitivity analysis conducted for all compres-
order to assure a good fit for both mass flow and power, four sors showed that the lower the clearance fraction the higher the
conditions are suggested for the calibration: the four combinations coefficient of determination of the fitted curves. Thus, the largest
of minimum and maximum evaporating and condensing temper- coefficients of determination were obtained for the clearance
atures of the established range. fractions equal to zero. That means the ideal volumetric efficiency is

a 20 20
a 300 25

15 20
250
15 15
10
Compressor power [W]
Mass floow rate [kg/h]

200
Difference [%]

Difference [%]
Experiment 10
Model 5
10 Difference 150 5

0
0
100

-5 -5
5 Experiment
50
First approach model -10
-10 Difference

0 -15
0 50 100 150 200 250 300
0 -15
0 0.5 1 1.5 2 2.5 3 Time [min]
Time [min] 25
b 300

20 20
b 20
Experiment 250
Model 15
Difference 15
15
Compressor power [W]

10 200
Mass floow rate [kg/h]

10
Difference [%]
Difference [%]

5
150 5
10
0 0
100

-5 -5
5 Experiment
50
First approach model -10
-10 Difference

0 -15
0 -15 0 50 100 150 200 250 300
0 0.5 1 1.5 2 2.5 3 Time [min]
Time [min]
Fig. 7. Comparison of the measured with the calculated compressor power by
Fig. 6. Comparison of the measured and computed mass flow rate during the first employing the measured mass flow rate. The start-up test. (a) Compressor X and (b)
3 min of test of Fig. 5. (a) Compressor X and (b) compressor Y. compressor Y.
1120 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124

discharge lines. In order to measure temperatures, T type ther-

a 300 10
mocouples were placed at the compressor shell and at the suction
5
and discharge lines. The suction line thermocouple was located
250
0 100 mm away from the compressor shell and the discharge line
one, 150 mm away. A Coriolis flowmeter was installed at the
Compressor power [W]

-5
200 compressor discharge line and was thermally isolated from

Difference [%]
-10 the ambient. To check the influence of the flowmeter on the system
performance, the freezer was tested without and with it. A reduc-
150 -15
tion on the condenser-superheating region was observed, which
-20 did not affect significantly the system performance as a whole.
100
-25
The compressor electric power was measured by a wattmeter.
Experiment A data acquisition system was used and all variables were recorded
First approach model
-30 every 4 s. The instruments uncertainties are shown in Table 7.
50 Difference
-35 Comparisons of measured and computed values of mass flow
rate and compressor power were conducted for both compressors
0 -40 X and Y. The computed variables were based on measured
0 0.5 1 1.5 2 2.5 3
boundary conditions such as, suction and discharge pressures and
Time [min]
compressor shell temperature. Because these boundary conditions
present measuring uncertainties, the computed values are not
b 300 10
exact. Therefore, the uncertainties of the calculated mass flow rate
5
and compressor power can be evaluated according to [27]:
250
0

a 250 20
Compressor power [W]

-5
200
Difference [%]

-10
15
150 -15 200
Compressor power [W]

-20
10

Difference [%]
100
-25 150

Experiment -30 5
50 First approach model
Difference 100
-35
0
0 -40
0 0.5 1 1.5 2 2.5 3
Time [min] 50 Experiment
-5
Second approach model
Fig. 8. Comparison of the measured with the calculated compressor power by Difference [%]
employing the measured mass flow rate. The first 3 min of Fig. 7.
0 -10
0 50 100 150 200 250 300
equal to one and that Eq. (6) is a correlation for the actual volu- Time (min)
metric efficiency. Table 5 shows the calibration data for all 21
compressors. b 250 20
Figs. 3 and 4 compare, respectively, the measured mass flow rate
and compressor power to their computed counterpart for such
compressors. 363 points were used in the comparisons of Figs. 3 15
200
and 4. 91% of the points of Fig. 3 lie under the þ/ 10% error bars
Compressor power [W]

in which 77% are under þ/5% differences. In Fig. 4, on the other Experiment
Second approach model 10
hand, 83% of the points are under þ/5% error bars and only 14%
Difference [%]

150 Difference [%]

are above the þ/10% differences.
5

5. Model validation 100

0
In order to validate the model, two types of transient experi-
mental tests were carried out in a 300 L vertical freezer: thermal 50
cycling and start-up tests1. The freezer was tested with both -5
compressors X and Y at ambient temperature of 32  C. The char-
acteristics of the refrigeration system are described in Table 6. Two
0 -10
pressure transducers were installed at the compressor suction and 0 50 100 150 200 250 300
Time (min)
1
The tests were carried out at the facilities of the POLO Research Laboratories for Fig. 9. Comparison of the measured with the calculated compressor power by
Emerging Technologies in Cooling and Thermophysics at the Federal University of employing the computed mass flow rate. The start-up test. (a) Compressor X and (b)
Santa Catarina, Brazil. compressor Y.
 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124 1121

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2  2 the compressor power is 3 W (about 3% of lowest measured value).
DZ ¼  DZjTsh þ DZjpe þ DZjpc (8) In this case, the uncertainties of suction and discharge pressures
have the same importance on the compressor power deviation. The
where DZ is the uncertainty of mass flow rate (Eq. (6)) or shell temperature (or suction temperature) does not affect the
compressor power (Eq. (7)). DZjTs , DZjpe and DZjpc are changes on compressor power. Note that the uncertainties of the computed
mass flow rate or power, caused, respectively, by the uncertainties values are much larger than the measured ones (Table 7). Therefore,
of shell temperature DTsh, suction Dpc and discharge pressures Dpe to decrease the uncertainty of the correlations, the uncertainty of
given in Table 7. DZjTs , DZjpe and DZjpc are calculated according to: the pressure measurements must be reduced.
Fig. 5 shows comparisons of the measured and computed mass
DZjTsh ¼ ZðTsh þ DTsh ; pc ; pe Þ  ZðTsh ; pc ; pe Þ (9) flow rates of both compressors for the start-up test. Not only the
results are quite close e most differences lie within 15% error
bands e but also the curve shapes are very similar (see the zoom at
DZjpc ¼ ZðTsh ; pc þ Dpc ; pe Þ  ZðTsh ; pc ; pe Þ (10)
the upper part of the figure), meaning the quasi-steady state
hypothesis is adequate. Fig. 6 shows the first 3 min of the results of
DZjpe ¼ ZðTsh ; pc ; pe þ Dpe Þ  ZðTsh ; pc ; pe Þ (12) Fig. 5. Although measured and computed values show different
trends in the first 3 points, the model represents quite well the
The maximum deviation of mass flow rate, as calculated by mass flow rate. However, one would expect the model behavior to
Eq. (8), is 0.07 kg h1 (4.7% of 1.5 kg h1 e the lower measured value be closer to the physics as the mass flow rate should reach its
of mass flow in the transient tests), which is mainly caused by maximum value at the compressor start-up, as a consequence of
the suction pressure uncertainty. The deviations produced by the the highest volumetric efficiency (equal to one) and the low specific
uncertainties of discharge pressure and shell temperature volume at the suction port. As the pressure ratio growths, the
are smaller than one tenth of the ones induced by the suction specific volume increases and the volumetric efficiency decreases,
pressure variations. On the other hand, the maximum deviation of causing a monotonic reduction of the mass flow rate, as predicted

a 300 20
8 20
15 a
10 7 Experimental
250
Model 15
5
Difference
Compressor power [W]

0 6
Mass flow rate [kg/h]

200
-5 10
Difference [%]

Difference [%]
5
-10
150 -15
4 5
-20
-25 3
100
-30 0
Experiment
Second approach model -35 2
50 Difference [%]
-40 -5
1
-45
0 -50
0 0.5 1 1.5 2 2.5 3 0 -10
0 2 4 6 8 10
Time (min)
300 10
Time [min]
b b 8 20

5
250 Experimental
Model 15
0 Difference
Compressor power [W]

6
Mass flow rate [kg/h]

200
10
Difference [%]

-5
Difference [%]

150 -10
4 5

-15
100
0
Experiment -20
2
Second approach model
50
Difference [%] -5
-25

0 -30
0 0.5 1 1.5 2 2.5 3 0 -10
0 2 4 6 8
Time (min)
Time [min]
Fig. 10. Comparison of the measured with the calculated compressor power by
employing the computed mass flow rate. The first 3 min of Fig. 9. (a) Compressor X and Fig. 11. Comparison of the measured with the calculated mass flow rate for the cycling
(b) compressor Y. test. (a) Compressor X and (b) compressor Y.
1122 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124

by Eq. (6). One could argue that the measured mass flow rate peak curves of compressor power and mass flow rate are not exactly in
is due to the compressor rotor inertia. Nevertheless, the necessary phase because the power depends not only on the mass flow rate
time to reach the steady-state rotation is only a fraction of a second but also on the compressor work.
and consequently, the 4 s data acquisition period is not fast enough The second approach, which employs the computed mass flow
to pick up such start-up peak. Therefore, the initial behavior of the rate on the calculation of the compressor power, is now used to
measured mass flow rate seems incorrect. This must be caused by compare the measured and the computed power values (see Fig. 9).
the low response time of the Coriolis flowmeter. In comparison to Fig. 7, the curves are again very similar though the
Two approaches were adopted to verify the compressor power computed values have different magnitudes: the steady-state
model (Eq. (7)). Firstly, the compressor power is evaluated by differences have changed from 5% to þ10% for compressor X and
employing the measured mass flow rate and secondly, by using the from 3% to 5% for compressor Y. Fig. 10 shows the first 3 min of
computed mass flow rate. A comparison of the measured with the results of Fig. 9. One can see the measured and computed curves
the first approach computed power is presented in Fig. 7. The are in better agreement than they were in Fig. 8, indicating once
curves are very similar to each other with the discrepancies within more the transient measurement of the mass flow is not accurate.
þ/15% for both compressors. The same comparison was also performed for the cycling
The first 3 min of the results of Fig. 7 are shown in Fig. 8. The experiments. The oneoff cycles take place approximately every
largest value of the compressor power must take place at the start- 21.5 min for compressor X (10.1 min running) and every 16.4 min
up in order to overcome the rotor inertia. Besides, due to the (7.5 min running) for compressor Y. As the cycles are approximately
reduction of the mass flow rate, despite the slight increase of the of the same duration, only one of them was chosen for the analysis.
compressor work, the power should decrease after the start-up, as Fig. 11 compares the measured and computed mass flow rates,
observed in the measured curves of Fig. 8. As the calculated power during the “on” period. Note that the compressors are switched on
is based on the measured values of the mass flow rate, its peak is at time zero and off after 10.1 and 7.5 min, respectively. Although the
a result of the curve shape shown in Fig. 6. Note that the measured differences are quite high soon after the compressor start-up, these

a 200 20

Experimental
a 200 30

First approach model

Difference 15 25
Experiment
150 Second approach model
Compressor power [W]

150 Difference 20
Compressor power [W]
Difference [%]

10

Difference [%]
15
100
100 10
5

5
50
0 50 0

-5
0 -5
0 1 2 3 4 5 6 7 8 9 10 11 0 -10
Time [min] 0 1 2 3 4 5 6 7 8 9 10 11
Time [min]
b 200 20
18
b 200 30

Experimental 16 25
First approach model Experiment
150 Difference 14 Second approach model
Compressor power [W]

150 Difference 20
Compressor power [W]

12
Difference [%]

10
Difference [%]

15
8
100
6 100 10

4
5
2
50 0 50 0

-2
-5
-4
0 -6 0 -10
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Time [min] Time [min]

Fig. 12. Comparison of the measured with the calculated power based on the Fig. 13. Comparison of the measured with the calculated power based on the
measured values of mass flow rate for the cycling test. (a) Compressor X and (b) computed values of mass flow rate for the cycling test. (a) Compressor X and (b)
compressor Y. compressor Y.
 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124 1123

disappeared. Although the differences dropped soon after the start-

a 8 5
up they increased after the power peak in comparison to those in
Experiment Fig. 12.
7 Second approach model
Usually the mass flow rate is not measured during the tests of
Difference
domestic refrigeration systems. As the compressor power is always
6
measured, the mass flow rate can be alternatively computed by
Mass flow rate [kg/h]

0 employing Eq. (7). Besides, Eq. (7) has shown a better curve fitting

Difference [%]
5
than Eq. (6) for almost all cases; the coefficients of determination
for Eq. (6) were smaller than they were for Eq. (7). Fig. 14 compares
4
the measured mass flow rate to the one computed by Eq. (7). Note
that the isentropic compressor work was based on the measured
3 values of suction and discharge pressures and also on the shell
-5
temperature. Not only the curve shapes of Figs. 11 and 14 are quite
2 similar but also the differences on Fig. 14 are smaller than those
found in Fig. 11.
1
Concluding remarks
0 -10
0 1 2 3 4 5 6 7 8 9 10 11 In the current work, a semi-empirical model to predict the
Time [min] transient mass flow rate and the power of domestic refrigeration
b 8 5 compressors was presented. The model was based on thermody-
namic equations and they were fitted to thirteen calorimeter test
Experiment data sets of two compressors. The analysis showed that the actual-
7
Second apprach model
Difference to-ideal volumetric efficiency ratio is linearly dependent on the
6 discharge-to-suction pressure ratio and that the compressor power
is a linear function of the product of mass flow rate and compressor
Mass flow rate [kg/h]

0
isentropic work. These linear correlations require only two calo-
Difference [%]

5
rimeter data sets to be fitted.
4 Data fitting was also conducted for other 21 compressors whose
data were obtained directly from the catalog of two manufacturers.
363 data sets were employed on the fitting. The comparison of the
3
-5 catalog values to the model results showed that 77% and 83% of
the mass flow rate and compressor power differences were with
2
the error bands of þ/5%, respectively.
Two types of transient experiments were used to validate the
1
current model: the start-up and the cycling tests. Despite a delay
between the measured and computed mass flow rates, at the
0 -10
0 1 2 3 4 5 6 7 8
beginning of the tests, the results agreed quite well. Excluding the
Time [min] first five measuring points, the differences were within 4% to
þ18% for the compressor X and 10 to þ12% for compressor Y in
Fig. 14. Comparison of the measured and the calculated mass flow rate based on the the transient period. In steady-state, the observed differences for
measured compressor power for the start-up test. (a) Compressor X and (b) compressors X and Y were 8% and 10%, respectively. The delay,
compressor Y. however, might be related to the Coriolis flowmeter response time.
In order to compare the computed and measured compressor
differences reduce considerably and stabilize at approximately 7% powers, two different approaches were employed to calculate the
for compressor X and 5% for compressor Y. After the peak observed power: the first based on the measured and the second, on the
in the measured values, the curves agree quite well and the differ- computed mass flow rate. In the first approach, the measured and
ences are even smaller than those observed in the start-up test. Note computed curves are quite similar but there is a time shift between
that the slope change at about 1.5 min is observed in both measured them, probably related to the measurement of the mass flow rate.
and calculated curves, for both compressors. With exception to the first 3 measuring points, all differences are
As before, the power is either based on the measured or within e9 to 12% and 13 to 12% for compressors X and
computed values of mass flow rate. Fig. 12 presents the measured Y, respectively.
power and its computed counterpart based on the measured mass Not only the measured values of mass flow rate and compressor
flow rate. Despite the time shift, the measured and calculated power were liable to uncertainties but also their computed coun-
curves are very similar to each other. The calculated curves are terparts, as they were based on measured values of pressure and
delayed in comparison to the measured ones; the maximum temperature. The uncertainties of the computed values were esti-
calculated powers take place a few seconds after the measured mated to be much larger than the measured ones. Therefore, most
values. The time shift makes the differences of measured and of the differences found in the comparisons can be attributed to the
computed curves quite high at the start-up. As soon as the curves computed value uncertainties, which can only be decreased if the
get in phase, the differences reduce significantly, reaching 4% for pressure transducers uncertainties (mainly the suction pressure)
compressor X and becoming almost unnoticeable for compressor Y. are reduced.
The power was again computed by employing the calculated Considering the good agreement with experimental values, one
mass flow rate and compared to the measured figures (see Fig. 13). concludes that the proposed semi-empirical model can be applied
Regarding their shapes, the measured and computed curves are to dynamic simulations of reciprocating compressors. Therefore,
very similar to each other, as the time shift has completely the process can be considered quasi-steady because the pressure
1124 C.O.R. Negrão et al. / Applied Thermal Engineering 31 (2011) 1114e1124

changes affect quite instantly the mass flow and the compressor [9] E. Navarro, E. Granryd, J.F. Urchueguía, J.M. Corberán, A phenomenological
model for analyzing reciprocating compressors, International Journal of
power. The model can be used to dynamic evaluations of the whole
Refrigeration 30 (2007) 1254e1265.
refrigeration system performance. [10] E. Navarro, J.F. Urchueguía, J.M. Corberán, E. Granryd, Performance analysis of
In refrigeration systems’ tests, the compressor electric power is a series of hermetic reciprocating compressors working with R290 (propane)
usually measured but not the mass flow rate. Therefore, the mass and R407C, International Journal of Refrigeration 30 (2007) 1244e1253.
[11] M.E. Duprez, E. Dumont, M. Frère, Modelling of reciprocating and scroll
flow rate can be obtained from the power correlation. This esti- compressors, International Journal of Refrigeration 30 (2007) 873e886.
mation can be quite accurate, as it is better fitted to calorimeter [12] R. Dufour, J. Der Hagopian, J. Lalanne, Transient and steady state dynamic
tests than the mass flow correlation. Consequently, the dynamic behaviour of single cylinder compressors: prediction and experiments, Jour-
nal of Sound and Vibration 181 (1) (1995) 23e41.
system capacity can be continuously computed based on the [13] M.N. Srinivas, C. Padmanabhan, Computationally efficient model for refrig-
compressor measured electric power. eration compressor gas dynamics, International Journal of Refrigeration 25
(2002) 1083e1092.
[14] G.A. Longo, A. Gasparella, Unsteady state analysis of the compression cycle of
Acknowledgements a hermetic reciprocating compressor, International Journal of Refrigeration 26
(2003) 681e689.
[15] M. Elhaj, F. Gu, A.D. Ball, A. Albarbar, M. Al-Qattan, A. Naid, Numerical
The authors acknowledge the financial support of EMBRACO and simulation and experimental study of a two-stage reciprocating compressor
CNPq (PIBIC e Program) and also thank POLO Laboratories that for condition monitoring, Mechanical Systems and Signal Processing 22
(2008) 374e389.
carried out the experimental tests with the refrigerator.
[16] O. Yaşar, M. Koças, Computational modeling of hermetic reciprocating
compressors, International Journal of High Performance Computing Applica-
tions 21 (1) (2007) 30e41.
References [17] E.L.L. Pereira, C.J. Deschamps, F.A. Ribas Jr., Performance analysis of recipro-
cating compressors through computational fluid dynamics, Journal of Process
[1] Eletrobras Annual Report (2008).www.eletrobras.com.br in portuguese. Mechanical Engineering 222 (4) (2008) 183e192.
[2] A.R.I. Standard 540-99, Positive Displacement Refrigerant Compressors and [18] R.N.N. Koury, M. Machado, K.A.R. Ismail, Numerical simulation of a variable
Compressor Units. Air-Conditioning and Refrigeration Institute, 1999. speed refrigeration system, International Journal of Refrigeration 24 (2001)
[3] ASHRAE/ANSI Standard 23-1993, Methods of Testing for Rating Positive 192e200.
Displacement Refrigerant Compressors and Condensing Units. American [19] G. Ding, C. Zhang, Z. Lu, Dynamic simulation of natural convection bypass two-
Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., Atlanta, circuit cycle refrigeratorefreezer and its application Part I: component
1993. models, Applied Thermal Engineering 24 (2004) 1513e1524.
[4] P. Popovic, H.N. Shapiro, A semi-empirical method for modelling a recipro- [20] Z. Lu, G. Ding, C. Zhang, Dynamic simulation of natural convection bypass two-
cating compressor in refrigeration systems, ASHRAE Transactions 101 (2) circuit cycle refrigeratorefreezer and its application Part II: system simulation
(1995) 367e382. and application, Applied Thermal Engineering 24 (2004) 1525e1533.
[5] D.I. Jähnig, D.T. Reindl, S.A. Klein, A semi-empirical method for representing [21] C.J.L. Hermes, C. Melo, A first-principles simulation model for the start-up and
domestic refrigerator/freezer compressor calorimeter test data, ASHRAE cycling transients of household refrigerators, International Journal of Refrig-
Transactions 106 (2000) 122e130. eration 31 (2008) 1341e1357.
[6] M.H. Kim, C.W. Bullard, Thermal performance analysis of small hermetic [22] Z. Lei, M. Zaheeruddin, Dynamic simulation and analysis of a water chiller
refrigeration and air-conditioning compressors, JSME International Journal, refrigeration system, Applied Thermal Engineering 25 (2005) 2258e2271.
Series B 45 (4) (2002) 857e864. [23] S. Porkhial, B. Khastoo, M.R. Modarres Razavi, Transient characteristic of
[7] E. Winandy, C.O. Saavedra, J. Lebrun, Simplified modelling of an open-type reciprocating compressors in household refrigerators, Applied Thermal
reciprocating compressor, International Journal of Thermal Sciencies 41 Engineering 22 (2002) 1391e1402.
(2002) 183e192. [24] W.B. Gosney, Principles of Refrigeration. CambridgeUniversity Press, 2002.
[8] C.J.L. Hermes, C. Melo, How to get the most out from a reciprocating [25] Embraco, Embraco Catalog (2009).www.embraco.com.br.
compressor semi-empirical model using a minimum data set, in: Proceedings [26] TECUMSEH, Tecumseh Catalog (2009).www.tecumseh.com.br.
of the International Conference on Compressors and Coolants, Papiernicka, [27] R.J. Moffat, Describing the uncertainties in experimental results, Experimental
Slovak Republic (2006). Thermal and Fluid Science 1 (1) (1988) 3e17.