Professional Documents
Culture Documents
DOI 10.1007/S12273-009-9209-x
Research Article
Lotfi Seghouani, Nicolas Galanis( )
Abstract Keywords
A quasi-steady model for the refrigeration system of an indoor ice rink was developed based on a refrigeration system,
combination of thermodynamic relations, heat transfer correlations and data available in the heat exchangers,
manufacturers catalogue. The system includes five compressors, rejects heat to the ambient air ice rink,
and uses R-22 to keep a stream of brine at a temperature of approximately 9C . The model has thermal load,
energy efficiency
been validated by comparison with measured values and with data from the manufacturers
catalogue. It was then used to simulate its performance over a typical meteorological year. Results
for a representative day include the number of compressors in operation at any given moment, Article History
Received: 12 February 2009
their power consumption, the coefficient of performance (COP) of the cycle and of the system, as
Revised: 21 April 2009
well as the heat rejected by the condensers and the corresponding mass flow rate of the cooling
Accepted: 5 May 2009
air. They show that the evaporation pressure is essentially constant while the condensation
pressure varies from about 1600 to 2000 kPa. The COP of the system varies between 1.9 and 2.5.
Tsinghua University Press and
Results for the entire year show that the heat rejected during phase change is approximately four
Springer-Verlag 2009
times that due to desuperheating and demonstrates the interest of recovering heat from both
processes. Finally, the model is used to illustrate the advantages of a control strategy which limits
the maximum number of simultaneously operating compressors to four. This strategy results in a
10% decrease of the energy used by the compressor motors and a 20% decrease of the peak
power demand but increases the temperature of the brine at the exit from the chillers by
approximately 0.5C during short periods following the ice resurfacing operations.
List of symbols
compressor shell, the refrigerant oil mixture concentration is readily available in manufacturers catalogues.
and the corresponding thermophysical properties are Cecchini and Marchal (1991) proposed a simplified
almost analytically indescribable. For this reason, some numerical code for simulating refrigerating and air-
authors used and alternative method which falls between conditioning equipment. This model did not describe in
the two types of models previously described. Some articles detail the operation of the systems components but
illustrating these approaches are briefly reviewed in the characterizes it with a few parameters deduced from
following paragraphs. Most describe steady, or quasi-steady, experimental data. The simulation calculates heating and
models that are more appropriate for annual energy cooling capacities and electrical power input for any
calculations. operating condition.
Fisher and Rice (1983) presented a steady-state model Stefanuk et al. (1992) developed and used a model to
for an air-to-air reciprocating vapour compression heat simulate a steady-state water-to-water vapour compression
pump based on thermodynamics principles and heat heat pump with superheat control. It was derived entirely
transfer relations. They used manufacturer maps and curve from the basic conservation laws of mass, energy, momentum
fitting for the compressor model and developed their and the equation of state as well as fundamental correlations
model with the objective of only requiring input data that of heat and mass transfer. It predicts system performance
Seghouani and Galanis / Building Simulation / Vol. 2, No. 2 121
over the full operating range of a heat pump. Good agreement assumptions, mathematical description, solution technique,
with experimental results was found for the cases studied. applicability, discussion, and references. It therefore provides
Nyers and Stoyan (1994) modeled the evaporator of a a wealth of information in a simple format which gives a
refrigeration system with detailed heat transfer equations. useful overview of the evolution of the modeling approaches
The focus was on the response of the evaporator to system applied to such systems.
controls. The expansion valve and compressor were modeled The principal goal of the present paper is to describe
using algebraic equations with lumped parameters. Other and validate a numerical model of an indoor ice rink
interesting works using first principles to model the refrigeration system, which can be used to simulate its
complete refrigeration system have been published by Sami operation, to estimate its annual energy consumption and to
and Dahmani (1996), Xiandong et al. (1997), and Xuejun et analyse different energy saving strategies. The system under
al. (1999). consideration includes several compressors, evaporators,
Popovic and Shapiro (1995) described a compressor and condensers while previous publications treated simple
model which requires eight input quantities and determines systems with only one component of each type. The
mass flow rate, refrigerant outlet state and compressor modeling approach uses a combination of the Fundamental
power with a 10% relative error. method especially for heat exchangers where well
Bourdouxhe et al. (1997, 1994) produced a simulation established correlations for phase changes were used and
toolkit for ASHRAE which uses a small number of the Curve Fitting method coupled with thermodynamics
parameters that are accurate enough for annual energy laws for the compressor. The expansion valve is modeled as
calculations of building systems. They used a quasi-static a perfect isenthalpic device maintaining constant superheat.
approach and assumed an ideal thermodynamic cycle with In future work this model of the refrigeration system will
isothermal refrigerant in the condenser and evaporator. be coupled to the models of the building and the concrete
Williatzen et al. (1998) presented a mathematical model slab with brine pipes described in our previous articles
describing the transient phenomena of two-phase flow for (Daoud and Galanis 2006; Daoud et al. 2008; Seghouani et
evaporators and condensers in a refrigeration cycle with al. 2009).
application to a domestic refrigerator. The heat exchanger
was separated in three zones (liquid, two-phase, and vapour 2 Refrigeration system description and basic
regions). assumptions
Svensson (1999) modeled a basic vapour compression
system with several simplifying assumptions for the The refrigeration system under study is that of the Camilien
compressor. The condenser was separated in two zones Houde ice rink situated in Montreal, Canada. The thermal
(condensing and subcooling) while the evaporator was behaviour and refrigeration loads of this building have
treated as a single zone. The expansion device was modeled been analysed previously (Daoud and Galanis 2006; Daoud
as a perfect isenthalpic process maintaining constant et al. 2008; Seghouani et al. 2009).
superheat. Figure 1 shows a schematic representation of this
Sreedharan and Haves (2001) compared three chiller system which is essentially constituted of:
models (including the ASHRAE toolkit model and the Two parallel shell and tubes evaporators with three
DOE-2 chiller model) with experimental data and concluded compartments each, shell diameter 0.5 m, evaporator
that all three displayed similar levels of accuracy. length 1.778 m, baffle spacing 0.393m, 342 tubes in two
Jin and Spitler (2002) presented a steady-state simulation passes, total surface of tubes 49.1 m2, internal and
model for a water-to-water reciprocating vapour compression external tube diameters respectively 15.75 mm and
heat pump, based on thermodynamics principles and heat 19.05 mm, tube length 2.4 m, pitch 23.812 mm, triangular
transfer relations. It includes several parameters that are configuration of tubes
estimated from catalogue data using a multi-variable Five open-drive four-cylinder compressors (Carrier
optimization and was developed with the objective of 5H40 model) with maximum speed 1750 rpm, piston
only requiring input data that is readily available in displacement 0.04359 m3/s, clearance 0.08
manufacturers catalogues. Five air-cooled condensers (Larkin RC8, 0768 model)
The literature review of models of vapour compression with 2.743 m length, 2.038 m width, 168 copper tubes in
equipment by Bendapudi and Braun (2002) includes 40 six passes and three rows, internal and external tube
papers spanning a period of 23 years from different journals diameters respectively 9.525 mm and 11.525 mm, tube
and conference proceedings. They present a summary of length 2.667 m, total surface (tubes + fins) 91.338 m2,
each of these papers using the same general template which number of aluminum fins 390 per meter, fin pitch
includes eight headings: equipment description, purpose, 2.56 mm, fin thickness 0.25 mm, height and width of fins
122 Seghouani and Galanis / Building Simulation / Vol. 2, No. 2
Fig. 1 Schematic representation of the refrigeration system and Fig. 3 Temperature-enthalpy diagram for the basic refrigeration unit
the secondary loop (Camilien Houde ice rink)
Seghouani and Galanis / Building Simulation / Vol. 2, No. 2 123
QR* = QR N C (2)
Fig. 4 Procedure for the calculation of the systems performance Pev = f (Tev )sat (4)
124 Seghouani and Galanis / Building Simulation / Vol. 2, No. 2
1
Pcd
(
vol = 1 + C C ( ))
Pev
n
(11)
PD
m R = vol (12)
v1
n1
W pol = m R
(
n
P v ((PP ) 1))
n 1 ev 1
cd
ev
n (13)
2
The other thermodynamic properties of the refrigerant
(enthalpy, entropy, density ) at states 1 and 1' are also
W m = W pol (0.77423 + 0.03969(PP ) 0.00442(PP ))
cd
ev
cd
ev
ev
n (10) data and correlation)
Seghouani and Galanis / Building Simulation / Vol. 2, No. 2 125
5. This step uses the evaporator model to calculate The fanning friction factor is calculated from the correlation
the heat exchange surfaces for the two-phases (2p) and for smooth circular ducts given by
superheated (sup) zones. The corresponding equations are
f = ( 1.58ln Re 3.28 )2 (19)
m
m R ( i1 i4 ) = b C pb (Tb,out Tb,1 )
NC
= U ev,2p Aev,2p CFev LMTDev,2p (15a) Equation (19) predicts the results in the range 104 < Re <
5 106 and 0.5 < Pr < 200 with 5% to 6% error, and in
the range 0.5 < Pr < 2000 with 10% error.
m b
m R ( i1 i1 ) = C (T Tb,in ) For the transition region where the Reynolds numbers
N C pb b,1
are between 2300 and 104, the Gnielinski correlation
= U ev,sup Aev,sup CFev LMTDev,sup (15b)
(Gnielinski 1976) is recommended:
Dev,o Dev,o ln(Dev,o Dev,i ) 1 1 In the two phase zone of the evaporator the heat transfer
U ev,phase = (D h
ev,i ev,phase
+
2 k wall
+ Rf,o +
hev,b
) coefficient is evaluated from the Shah correlation (Shah
1982):
(16)
8. The condenser is modeled by expressing energy The calculation in this step proceeds as follows. First the air
conservation for each of three zones: desuperheating (2 to temperatures Tair,3 ,Tair,2 , and Tair,out are calculated from
2' ), condensing (2' to 3' ) and subcooling (3' to 3). The Eqs. (24). Then the logarithmic temperature differences
corresponding expression are LMTDcd,sub , LMTDcd,2p , and LMTDcd,des are evaluated from
their definition. Next Eqs. (26) (28) are used to calculate
m R ( i2 i2 ) = N fan m air C pair ( Tair,out Tair,2 ) the convection heat transfer coefficients and the overall
= U cd,des Acd,des CFcd LMTDcd,des (24a) conductance for each zone is calculated from Eq. (25).
Finally the heat transfer surface for each zone ( Acd,des , Acd,2p ,
m R ( i2 i3 ) = N fan m air C pair ( Tair,2 Tair,3 ) and Acd,sub ) is calculated from Eqs. (24).
= U cd,2p Acd,2p CFcd LMTDcd,2p (24b) 9. The third test is performed by comparing the sum of
the three calculated surfaces ( Acd,des , Acd,2p , Acd,sub ) to the
m R ( i3 i3 ) = N fan m air C pair ( Tair,3 Tair,in ) given total surface of the condenser Acd. If the sum is not
= U cd,sub Acd,sub CFcd LMTDcd,sub (24c) within 5% of Acd, the number of fans is increased and the
calculation continues with step 7b. If the sum is within
The overall heat transfer conductance for each zone is 5% of Acd and N fan - 4 the calculation proceeds to
Acd,o ln(Dcd,o Dcd,i ) Rf,o 1 step 10. On the other hand, if N fan > 4 the solution is
Acd,o 1
U cd,phase= (A cd,i hcd,phase
+
2kwall L
+ +
o o hcd,air ) unacceptable since there are only four fans per condenser
in the system under study. In such a case, the number of
(25) compressors (refrigeration units) is increased and the
procedure restarts at step 2.
Next it is necessary to calculate the convection heat transfer 10. If the above procedure leads to N C - 5 the model
coefficients for the refrigerant in each zone and for the air. has converged to physically acceptable operating conditions.
8.a Refrigerant side of the condenser (three zones are Time is then incremented and the procedure is repeated for
considered) the next timestep. On the other hand if N C > 5 the solution,
For single phase refrigerant (desuperheating and although mathematically and thermodynamically correct,
subcooling zones) the convection heat transfer coefficient is not acceptable since there are only five compressors (five
was calculated using Eq. (18) or (20) depending on the refrigeration units) in the system under study. This solution
value of the Reynolds number. implies that the system can not deliver the desired outlet
For the two phase refrigerant we use the correlation by brine temperature Tb,out . Therefore its value is increased and
Cavallini and Zecchin (1974) the whole procedure is repeated for the same timestep by
kR,des replacing its previous value by the new one.
hcd,2p = 0.05ReR,eq
0.8 0.33
PrR,des (26) Since the calculations are repeated every 6 minutes for
Dcd,i
an entire year the results are in the form of a set of 87600
where the equivalent Reynolds number is defined by: vectors (365 days 24 h/day 10 calc/h) which include the
calculated values of all the systems operating variables
( )( ) + Re
0.5
(evaporation and condensation pressures, compressor
ReR,eq = Redes R,des R,sub
R,sub (27a)
R,sub R,des polytropic power, power input to the motor, exit brine and
air temperatures, etc.) at each instant.
Gev (1 x ) Dcd,i
ReR,sub = (27b)
R,sub 4 Model validation
Gev xDcd,i The developed model for the Camilien Houde ice rink
ReR,des = (27c)
R,des refrigeration system was validated in two steps.
Firstly, the performance of the system was compared to
8.b Air side of the condenser the manufacturers data. Eight of the 87600 vectors
The air side convection heat transfer coefficient was containing the results were chosen at random and the
estimated using Webbs correlation (Webb 1990) for heat calculated values of the evaporation and condensation
exchangers with flat plate fins temperatures as well as those of W m, the power input to the
motor driving the compressor, and QR, the refrigeration
0.23
Nucd,air = 0.4 Gz ( s Dc )
for Gz - 25
0.73
N R0.23 capacity, are presented in Table 1. The corresponding
(28) values of W m and QR obtained for the same Tev and Tcd by
0.23
Nucd,air = 0.53 Gz ( s Dc )
for Gz > 25
0.62
N R0.31 interpolation in the manufacturers catalogue are also
Seghouani and Galanis / Building Simulation / Vol. 2, No. 2 127
specified in this Table. The agreement is similar to that is analysed and discussed for two time periods. A short
reported in other studies. Thus, in the case of W m the period of 240 timesteps corresponding to a typical day
difference in 7 of the 8 cases is less than 7% while in the (the 1st of March was chosen for this purpose) illustrates
case of QR it is less than 5% in 5 out of 8 cases. Overall this the variation of operating conditions (evaporation and
is deemed to be acceptable in view of the assumptions condensation pressures, number of compressors in operation,
incorporated in the model and the uncertainty of the etc.) due to the diurnal variations of the refrigeration load
empirical correlations. and of the ambient air temperature. A long period of 87600
Secondly, calculated values of the evaporation and timesteps corresponding to a typical year is used to
condensation pressures, refrigerant mass flow rate, and compare monthly energy consumption by the compressors,
number of compressors in operation were validated by the fans, etc. for two control strategies of the system. The
comparing them to those obtained experimentally by results were calculated using yearly distributions of the
Ouzzane et al. (2006). The measured outdoor air temperature, ambient temperature and refrigeration load described in an
inlet and outlet brine temperatures, and brine mass flow earlier publication (Seghouani et al. 2009). The pumping
rate were taken as inputs for the model. Table 2 specifies power was added to this load resulting in QR while the
the values of these inputs and compares the calculated and corresponding brine temperature at the inlet of the evaporator
measured values of these four significant outputs. The was obtained from Eq. (29).
agreement is considered to be very good. The small
differences between calculated and measured evaporation 5.1 Transient results for a typical day (March 1st)
and condensation pressures and mass flow rates are due to
the imprecision of the measurements and to the The time dependent input data is shown in Fig. 7. The
simplifying assumptions of the model. outside air temperature Tair,in for the day under consideration
(from ENERGYPLUS) decreases from 5.6 to 11
5 Results and discussion during the night, starts rising after sunrise, reaches a
maximum of 1 around 16:00 and then decreases. The
In this section the response of the modeled refrigeration corresponding total refrigeration load QR, calculated with
system to the time-dependent input data (QR, Tb,in and Tair,in) the code described in a previous article (Seghouani et al.
2009), is the sum of heat fluxes into the ice (by convection,
radiation, and phase changes), heat gains from below the
concrete slab and the pump power (5 kW) which is dissipated
by friction. As seen in Fig. 7 this quantity decreases slightly
from 0:00 to 8:00 due to the fact that there are no activities
in the ice rink during the night (no lights, etc.) and then rises
from 8:00 until noon when the first resurfacing operation
takes place. During the first half of the day the load is
between 100 and 75 kW. The second half of the day is
characterised by high peaks which reach more than 300
kW and correspond to the resurfacing of the ice which is
achieved by spreading water at 60 on its surface. The
temperature of the brine entering the refrigeration system
Tb,in is the third input and is calculated from the relation
Fig. 7 Diurnal variation of the inputs for March 1st
QR = m b C pb ( Tb,in Tb,out ) (29)
Fig. 10 Power required by the fans and the compressors on March Fig. 11 Coefficients of performance for the refrigeration system
1st and for the refrigeration cycle for March 1st
130 Seghouani and Galanis / Building Simulation / Vol. 2, No. 2
Fig. 12 Inlet and outlet brine temperatures for two functioning Fig. 13 Average daily energy at the evaporator for two operating
strategies strategies
returning brine temperature is only visible during the short Figure 14 compares the average daily energy consumption
resurfacing periods and is rather small: when the maximum by the motors driving the compressors for the two operating
number of simultaneously operating compressors is limited strategies under consideration. Logically, this quantity is
to four the returning brine temperature during these periods lower when the maximum number of units in simultaneous
is slightly warmer (8.35) than if all five compressors are operation is 4. Thus the annual energy consumption of the
allowed to operate simultaneously (8.77). During the motors driving the compressors can be reduced by almost
rest of the day both operating strategies meet the design 10% by this simple control strategy with a small increase
requirement of a 9 returning brine temperature. The in the temperature of the ice. The values in Fig. 14 rise
effect of each of these operating strategies on the annual progressively from January to May and can exceed
energy consumption is analyzed in the next section of the 1400 kWh/day during the summer months. After September
present paper while their influences on the ice sheet they decrease significantly due to the decrease of the
temperature will be studied in the future by coupling the refrigeration load (Fig. 13).
present model with the one described by Seghouani et al. The heat rejected by each of the three regions of the
(2009). condensers (desuperheating, phase change, and subcooling)
is presented in Figs. 15(a) and 15(b) for the two operating
5.2 Monthly energy calculations for a typical year strategies under consideration while Fig. 15(c) compares
the corresponding totals.
Figure 13 shows a comparison of the average daily energy According to the results in Figs. 15(a) and 15(b) the
transferred to the refrigerant in the evaporators for the two heat rejected by the subcooling regions is insignificant for
operating strategies described earlier (maximum number of
units in simultaneous operation equals 4 or 5). It clearly
shows that with a maximum of five compressors in operation
the refrigeration load treated by the system is higher than
with four compressors. The difference can reach almost
150 kWh/day during the summer months while in winter it
is slightly smaller than 100 kWh/day. As already mentioned,
the peak power demand when the maximum number of
compressors in simultaneous operation is lowered from 5
to 4 is reduced by approximately 20% while the maximum
temperature of the brine at the outlet of the evaporator
increases by approximately 0.5. The seasonal effect on
this energy quantity is clear and qualitatively predictable.
Its value is lower in winter than in summer due to the
colder ambient and soil temperatures which reduce heat Fig. 14 Average daily energy consumption by the compressors for
fluxes towards the brine in the ice rink. two operating strategies
Seghouani and Galanis / Building Simulation / Vol. 2, No. 2 131
6 Conclusions
since the developed code is very flexible and allows such Lavoie M, Suny R, Gigure D (2000). Potentiel dconomies dnergie
modifications. en rfrigration dans les arnas du Qubec. Report prepared by
the CANMET Energy Technology Center, Varennes, Quebec. (in
French)
Acknowledgements Nyers J, Stoyan G (1994). A dynamical model adequate for controlling
the evaporator of a heat pump. International Journal of Refrigeration,
This study was financed by the Natural Sciences and 17 (2): 101 108.
Engineering Research Council (NSERC) of Canada through Ouzzane M, Suny R, Zmeureanu R, Gigure D, Scott J, Bellache O
the Strategic Project Grant STPGP 306792 (Title: (2006). Cooling load and environmental measurements in a
Canadian indoor ice rink (RP-1289). ASHRAE Transactions, 112
Development of design tools and of operation guidelines (2): 538 545.
for the heating, ventilation, air conditioning and refrigeration Petukhov BS (1970). Heat transfer and friction in turbulent pipe flow
systems of ice rinks). with variable physical properties. In: Advances in Heat Transfer,
Vol 6. New York: Academic Press.
References Popovic P, Shapiro HN (1995). A semi-empirical method for modeling
a reciprocating compressor in refrigeration system. ASHRAE
Transactions, 101(2): 367 382.
Bendapudi S, Braun JE (2002). A review of literature on dynamic models
REFPROP (Reference Fluid Thermodynamic and Transport Properties
of vapor compression equipment. ASHRAE Report #4036-5.
Database): Version 8.0.
Bourdouxhe JP, Grodent M, Lebrun JJ (1997). HVAC1 Toolkit:
Sami SM, Dahmani A (1996). Numerical prediction of dynamic
Algorithms and Subroutines for Primary HVAC System Energy
performance of vapor compression heat pump using new HFC
Calculations. The American Society of Heating, Refrigerating
alternatives to HCFC-22. Journal of Applied Thermal Engineering,
and Air Conditioning Engineers.
16: 691 705.
Bourdouxhe JP, Grodent M, Lebrun JJ, Saavedra C, Silva KL (1994). A
Seghouani L, Daoud A, Galanis N (2009). Prediction of yearly energy
toolkit for primary HVAC system energy calculationPart 2 requirements of indoor ice rink. Energy and Buildings, 41(5):
Reciprocating chiller models. ASHRAE Transactions, 100(2): 500 511.
774 786. Shah MM (1982). Chart correlation for saturated boiling heat transfer:
Cavallini A, Zecchin R (1974). A dimensionless correlation for heat equation and further study. ASHRAE Transactions, 88(1): 185 196.
and mass transfer in forced convection condensation. In: Sreedharan P, Haves P (2001). Comparison of chiller models for use in
Proceedings of the 5th International Heat Transfer Conference, model-based fault detection. In: Proceedings of International
309 313. Conference for Enhanced Building Operation, Austin TX
Cecchini C, Marchal D (1991). A simulation model of refrigerant and (http://www.osti.gov/bridge/servlets/purl/829977-eNbJEo/native
air-conditioning equipment based on experimental data. ASHRAE /829977.pdf).
Transactions, 97(2): 388 393. Stefanuk NBM, Aplevich JD, Renksizbulut M (1992). Modeling and
Daoud A, Galanis N (2006). Calculation of the thermal loads of an ice simulation of a superheat-controlled water-to-water heat pump.
rink using zonal model and building energy simulation software. ASHRAE Transactions, 98(2): 172 184.
ASHRAE Transactions, 112(2): 526 537. Svenssen MC (1999). Non-steady-state modelling of a water-to-water
Daoud A, Galanis N, Bellache O (2008). Calculation of refrigeration heat pump unit. In: Proceedings of the 20th International Congress
loads by convection, radiation and condensation in ice rinks of Refrigeration, Sydney, Paper No. 263.
using a transient 3D zonal model. Applied Thermal Engineering, Webb RL (1990). Airside heat transfer correlations for flat and wavy plate
28: 1782 1790. fin-and-tube geometries. ASHRAE Transactions, 96(2): 445 449.
ENERGYPLUS, http://www.eere.energy.gov/buildings/energyplus/. Williatzen M, Pettit NBOL, Ploug-Sorensen L (1998). A general
Fisher SK, Rice CK (1983). The Oak Ridge heat pump design models: dynamic simulation model for evaporators and condensers in
1. a steady-state computer design model for air-to-air heat pumps. refrigeration. Part I: moving-boundary formulation of two-phase
Report No. ORNL /CON-80/R1., Oak Ridge, Tenn., Oak Ridge flows with heat exchange. International Journal of Refrigeration,
National Laboratory. 21(5): 398 403.
Gnielinski V (1976). New equations for heat and mass transfer in He XD, Liu S, Asada HH (1997). Modeling of vapor compression
turbulent pipe and channel flow. International Chemical Engineering, cycles for multivariable feedback control of HVAC systems. ASME
16: 359 368. Journal of Dynamic Systems, Measurement and Control, 119 (2):
Jin H, Spitler JD (2002). A parameter estimation based model of 181 191.
water-to-water heat pumps for use in energy calculation programs. Xuejun S, Liu W, Que X, Chen Z (1999). Researches on dynamic
ASHRAE Transactions, 108 (1): 3 17. simulation and optimization of automobile air-conditioning
system, part 1: mathematical model and dynamic simulation of
Kaka S, Liu H (2002). Heat Exchangers Selection, Rating and Thermal
automobile air-conditioning system. In: Proceedings of the 20th
Design, 2nd edn. Washington, DC: CRC Press.
International congress of refrigeration, Sydney, Paper No. 577.