Accepted Manuscript

Performance Optimization of Absorption Refrigeration Systems Using Taguchi,

ANOVA and Grey Relational Analysis Methods

A.S. Canbolat, A.H. Bademlioglu, N. Arslanoglu, O. Kaynakli

PII: S0959-6526(19)31536-7

DOI: 10.1016/j.jclepro.2019.05.020

Reference: JCLP 16785

To appear in: Journal of Cleaner Production

Received Date: 21 October 2018

Accepted Date: 03 May 2019

Please cite this article as: A.S. Canbolat, A.H. Bademlioglu, N. Arslanoglu, O. Kaynakli,
Performance Optimization of Absorption Refrigeration Systems Using Taguchi, ANOVA and Grey
Relational Analysis Methods, Journal of Cleaner Production (2019), doi: 10.1016/j.jclepro.
2019.05.020

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Performance Optimization of Absorption Refrigeration Systems Using Taguchi,

ANOVA and Grey Relational Analysis Methods

A.S. Canbolat1, A.H. Bademlioglu2, N. Arslanoglu1, O. Kaynakli1*

1 Department of Mechanical Engineering, Engineering Faculty, Bursa Uludag University,

Turkey
2 Department of Energy Systems Engineering, Faculty of Engineering and Natural Sciences,

Bursa Technical University, Turkey

*Corresponding author: kaynakli@uludag.edu.tr, +90 224 2941903 (fax)

Abstract
There are various factors having an impact on the energetic and exergetic performance (i.e.,
COP and eCOP) of an absorption refrigeration systems (ARS) such as the temperatures of the
generator, condenser, evaporator and absorber, effectiveness of solution, refrigerant and
solution-refrigerant heat exchangers and isentropic efficiency of the solution pump. Many
studies have focused on these process parameters, but the importance order and contribution
ratios of the parameters due to thermodynamic performance have not been determined by using
statistical methods. Firstly, in this study, cycles’ thermodynamic model is established and the
variation of the COP and eCOP are calculated for different working conditions with different
parameters ranges. The effects of these parameters on the COP and eCOP are examined
separately on a statistical basis. The importance order of the parameters are determined by using
Taguchi and ANOVA methods and the results are compared. Optimum operating conditions
are determined by means of statistical analysis for the COP and eCOP. Under these operating
conditions, the COP and eCOP of the system are calculated as 0.697 and 0.2829, respectively.
Furthermore, for the simultaneous maximization of these two performance indicators, Taguchi-
Grey Relational Analysis (GRA) is used. By using this analysis, importance order of the
examined parameters on multiple performance characteristics are determined. The absorber and
evaporator temperatures are the most efficient parameters on multiple performance
characteristics with a contribution ratio of 29.66% and 26.34% of the total effect while the least
efficient parameters are the pump efficiency and effectiveness of solution-refrigerant heat
exchanger with a contribution ratio of 0.48% and 2.41%, respectively. For the best condition
considering the multiple performance characteristics, COP and eCOP of the system are found
as 0.6255 and 0.2829, respectively.

Keywords: Absorption Refrigeration System; Taguchi Method; Grey Relational Analysis;

ANOVA; COP; eCOP

1. Introduction
The decrease of fossil fuels such as natural gas, coal, oil and the increase of the negative
impact of these fuels increase the need for renewable energy sources day by day. Therefore, in
the last few years, the use of absorption refrigeration systems (ARSs) instead of vapor
compression refrigeration systems is currently gaining momentum. The most important
advantages of ARSs are as follows: They do not destroy the ozone layer depending on the
working fluid pairs used in the system and can benefit from various renewable energy sources
(i.e., geothermal energy or solar energy). Tugcu et al. (2016) optimized the single stage
geothermal energy assisted ARS, working with NH3-H2O, for different solution concentrations
 ACCEPTED MANUSCRIPT

and design parameters. In this study, for the optimum design, COP of the system was
determined as 57.22% while the exergy efficiency was calculated as 62.01%. Saleh and Mosa
(2014) examined the single-effect ARS powered by a flat-plate collector for hot regions and
optimized the performance of the system. They found that the overall system performance takes
its optimal value at temperatures between 75°C and 80°C, adopting typical values encountered
in hot regions.

The studies on the energy and exergy analysis of ARS

In literature, there are many studies on the thermodynamic analysis of ARSs and the
performance characteristics of the cycle. Karamangil et al. (2010) presented a comprehensive
literature review on the ARS and they examined the influence of the effectiveness’s of solution,
refrigerant and solution-refrigerant heat exchangers (SHE, RHE and SRHE), the operating
temperatures (generator, evaporator, condenser, and absorber) and the selection of working
fluid (LiBr-H2O, NH3-H2O, NH3-LiNO3) on the system performance indicators (COP and
circulation ratio, CR). In that study, it was concluded that SHE has the most significant effect
on COP since it increases the system COP by 66% compared to RHE and SRHE. Li et al.
(2017) performed a thermodynamic analysis of a novel air-cooled non-adiabatic ejection-
absorption refrigeration cycle with R290/oil mixture driven by exhaust heat. Ouadha and El-
Gotni (2013) performed the thermodynamic analysis of an ARS driven by waste heat from a
Diesel engine. The thermodynamic study of the cycle performed for several working conditions
by changing the temperatures of generator, condenser, absorber and evaporator. They
determined that higher performance of the system is obtained at high generator and evaporator
temperatures and also at low condenser and absorber temperatures.
Kaynakli and Yamankaradeniz (2003) investigated the effect of heat exchangers, which are
used to recover heat energy in the ARS, on the coefficient of performance (COP). In this study,
it was found that the solution heat exchanger (SHE) is the most effective heat exchanger on the
system performance. Abed et al. (2015) focused on the optimization of the utilization of the
internal heat recovery of the ejector-flash tank-ARS working with NH3-H2O. They found that
the refrigeration capacity increment of the proposed cycle with added RHE is 4.85%. Sencan
(2007) performed the performance analysis of NH3-H2O ARS based on the artificial neural
network model. Novella et al. (2017) performed the thermodynamic analysis of an absorption
refrigeration cycle used to cool down the temperature of the intake air in an internal combustion
engine using the exhaust gas of the engine as a heat source. In general, in these studies, the
different parameters affecting the first law efficiency of the ARS were examined and the effects
of the system on the COP were analyzed.
Aman et al. (2014) developed a thermodynamic model based on a 10kW air-cooled NH3-
H2O absorption chiller driven by solar thermal energy. In this study, the first and second law
analyses were conducted to evaluate the performance of this residential scale cooling system
and it was found that the absorber is the component where the most exergy loss occurs (63%)
followed by the generator (13%) and the condenser (11%). Joybari and Haghighat (2016)
performed the exergetic analysis of a base case single effect LiBr-H2O absorption system. In
this study, it was determined that as the mass flow rate of water circulated in the system became
lower, the exergy losses in the main components of the system decreased. Modi et al. (2017)
investigated the energetic and exergetic performance of the absorption refrigerant system in
LiBr-H2O solution for each component. For this purpose, they developed a numerical program
for energy and exergy calculations of the ARS. Kumar et al. (2017) performed a
thermodynamic analysis of the single effect NH3-H2O ARS. They determined the optimum
 ACCEPTED MANUSCRIPT

generator temperature corresponding to maximum COP and minimum exergy loss. Fernández-
Seara and Sieres (2006) investigated the effects of the ammonia purification, the liquid
entrainment and blow-down from the evaporator in the ARS. They developed a mathematical
model based on a single stage system with complete condensation. It was concluded that small
values of liquid entrainment or blow-down fractions increase the operating range of the
absorption system significantly, and smoothes the efficiency requirements of the distillation
column. In these studies, the energetic and exergetic performance of the ARS were investigated
and the effect of different parameters on COP and eCOP was examined.
Kaynakli and Kilic (2007) performed a detailed thermodynamic analysis of the LiBr-H2O
ARS. They investigated the impacts of operating temperature and effectiveness of heat
exchanger on the system performance. In this study, it was obtained that the SHE increases the
COP value up to a maximum of 44%, while the RHE had an effect of only 2.8%. Ketfi et al.
(2015) studied on the modeling and simulation of a 70 kW Yazaki absorption refrigeration
machine working with LiBr-H2O solution. They examined the effect of system parameters on
the coefficient of performance (COP). The parameters and operating ranges examined in the
studies summarized above are listed in Table 1. This table is used as a reference for the factors
affecting ARS performance and for the data to be used in statistical evaluation in this study.

Table 1
Various parameters examined in the literature.
Working Parameters
Study
Solution Tgen (°C) Tcon (°C) Tevap (°C) Tabs (°C) ƐSHE ƐRHE ηp
Tugcu et al. (2016) NH3-H2O 70 30 -4 30 0.95 - 0.6
Karamangil et al. (2010) NH3-H2O,LiBr- 75 - 105 30 - 50 5 - 15 30 - 50 0-1 0-1 0.9
H2O,NH3-LiNO3
Ouadha and El-Gotni (2013) NH3-H2O 60 - 120 20 - 45 -10 - 10 20 - 45 0.7 - 1 - 1
Kaynakli and Yamankaradeniz (2003) NH3-H2O 75 - 120 20 - 50 -1.25 - 20 20 - 50 0 - 0.6 0 - 0.6 1
Abed et al. (2015) NH3-H2O 60 - 120 20 - 50 -15 - 15 20 - 50 0.5 0.6 1
Sencan (2007) NH3-H2O 60 - 90 20 - 40 2.5 - 7.5 20 - 40 0.8 - 1
Novella et al. (2017) NH3-H2O 165 48 -10 41 0.7 0.7 0.8
Aman et al. (2014) NH3-H2O 80 30 2 30 0.8 - 1
Modi et al. (2017) LiBr-H2O 75 - 110 40 8 - 15 40 0.7 - 1
Kumar et al. (2017) NH3-H2O 60 - 140 30 - 45 -5 - 5 30 - 45 0.8 - 0.95
Fernández-Seara and Sieres (2006) NH3-H2O 115 30 -10 30 0.7 0.7 0.5
Kaynakli and Kilic (2007) LiBr-H2O 65 - 104 30 - 48 4 - 10 30 - 48 0-1 0-1 1
Ketfi et al. (2015) LiBr-H2O 90 40 7 40 0.7 - 1
Táboas et al. (2014) NH3-H2O, NH3- 85 25 -15 25 0.7 0.7 0.5
LiNO3, NH3-
(LiNO3+H2O)
 ACCEPTED MANUSCRIPT

The studies on the statistical evaluation

In general, analysis of variance (ANOVA) and Taguchi method are used to determine the
impact weight of the parameters statistically by considering the possible combinations of the
parameters affecting the system. However, studies on the statistical analysis of parameters
affecting the performance of thermal systems are limited in the literature. In many of these
studies, the design parameters of heat exchangers were determined and performance analyses
were performed with the help of Taguchi and ANOVA methods. Turgut et al. (2012)
determined the optimum design parameters of the concentric heat exchanger with injector
turbulators using Taguchi experimental design method. They investigated the effects on the
heat transfer and pressure loss of the injector-shaped turbulators having different angles,
diameters and numbers. Verma and Murugesan (2014) analyzed the performance of a solar
assisted ground source heat pump using Taguchi technique. In the study, the design parameters
were optimized to obtain solar collector area and ground heat exchanger length for space
heating application with optimum COP. Yakut et al. (2006) examined the effects of various
kinds of design parameters on the heat transfer and pressure-drop characteristics of the heat
exchanger using the Taguchi method. Zeng et al. (2010) investigated the influence of various
design parameters on the heat transfer and flow friction characteristics of a heat exchanger with
vortex-generator fins and optimized the parameters of vortex-generator fin-and-tube heat
exchangers by using the Taguchi method.
Bademlioglu et al. (2018) examined the impact weights of parameters on ORC's first-law
efficiency by utilizing Taguchi and ANOVA methods. In this study, the most efficient
parameters on the thermal efficiency of the ORC (evaporator temperature, condenser
temperature and turbine isentropic efficiency) were determined and the total effect ratios of
these parameters were calculated to be 70%. Coskun et al. (2012) analyzed the performance of
waste heat recovery application with the aid of the Taguchi method, and determined the
significant parameters and optimum operating conditions. Arslanoglu and Yigit (2017)
examined the parameters that have the most significant effect on the optimum insulation
thickness, in accordance with the importance order by utilizing Taguchi method. Moreover,
impact ratio for each parameter was determined with the help of ANOVA.
Another important statistical analysis method, which is different from Taguchi and ANOVA,
is Grey Relational Analysis (GRA). In general, the GRA model allows the simultaneous
evaluation of different objective functions and enables the determination of the optimum
parameters for all purpose functions. Naqiuddin et al. (2018) performed a computational fluid
dynamics analysis to evaluate the performance of a micro-channel heat sink. In that study, the
effects of various design parameters on the heat transfer and flow characteristics were optimized
using Taguchi-GRA method. Celik et al. (2018) experimentally investigated the corrugated tape
type turbulators with different pitches, weights and thicknesses inserted in a concentric pipe
heat exchanger. In that study, the multi-performance analysis was carried out by using GRA
method for heat transfer and pressure loss.
Acir et al. (2017) determined the optimum factors having an impact on the first and second
law efficiencies for a unique design solar air heater using the GRA method. They calculated the
contribution ratios of each parameter on the system performance by using ANOVA. Kuo et al.
(2011) performed the design by determining the optimal operating parameters of the flat-plate
collector with the help of the Taguchi method and GRA. Chamoli et al. (2016) investigated the
effect of processing parameters on the performance of a heat exchanger tube with the help of
the Taguchi based GRA. In addition, they conducted an experimental study by using the optimal
design condition.
Studies in the literature show that there are various parameters affecting the energetic and
exergetic performance of ARS such as the generator, evaporator, condenser and absorber
temperature, the effectiveness of SHE, RHE and SRHE and pump isentropic efficiency.
 ACCEPTED MANUSCRIPT

Nonetheless, a detailed study analyzing all these parameters and determining their contribution
ratios on the system’s performance with a statistical approach has not been encountered in the
literature. For this reason, the purpose of this study is to examine the parameters that present
the most significant effect on the ARS's COP and eCOP values and determine the importance
order of these parameters by utilizing Taguchi-GRA and ANOVA methods. Moreover, the best
and worst working conditions are determined by different statistical analysis methods and the
results are compared.

2. Mathematical Model
2.1 Determination of Absorption Refrigeration System Performance Characteristic
The schematic diagram of the single stage ARS for which the thermodynamic analysis is
carried out is given in Fig. 1. In this study, RHE cools the saturated liquid from the condenser
temperature to the evaporator temperature, while SHE helps to increase the temperature of the
solution entering the generator. A process similar to that in the SHE is also carried out in the
SRHE and these heat exchangers increase the thermodynamic performance of the cycle (COP
and eCOP) by reducing the thermal energy input required in the generator.

Fig. 1. The schematic of single-stage ARS.

In this study, a thermodynamic analysis of an ARS using NH3-H2O has been performed. The
basic equations used in the analysis of the first and second laws of thermodynamics are
presented here. More detailed information can be found in the literature (Karamangil et al.,
2010; Aman et al., 2014; Kaynakli and Yamankaradeniz, 2003; Ketfi et al., 2015).
Mass balance at the generator is defined as follows:
𝑚𝑠𝑠 = 𝑚𝑤𝑠 + 𝑚𝑁𝐻3 (total mass balance) (1)
𝑚𝑠𝑠𝑋𝑠𝑠 = 𝑚𝑤𝑠𝑋𝑤𝑠 + 𝑚𝑁𝐻3 (ammonia balance) (2)
The flow rates of the strong (ṁss) and weak (ṁws) solutions are calculated with the help of
Eqs. (1) and (2) respectively:
1 ― 𝑋𝑤𝑠
𝑚𝑠𝑠 = 𝑋𝑠𝑠 ― 𝑋𝑤𝑠𝑚𝑁𝐻3 (3)
 ACCEPTED MANUSCRIPT

1 ― 𝑋𝑠𝑠
𝑚𝑤𝑠 = 𝑋𝑠𝑠 ― 𝑋𝑤𝑠𝑚𝑁𝐻3 (4)
The circulation ratio (CR) can be described as the ratio of the mass flow rate of the solution
through the pump to the mass flow rate of the working fluid. The circulation ratio (CR) is
expressed as follows:
𝑚𝑠𝑠 1 ― 𝑋𝑤𝑠
𝐶𝑅 = 𝑚𝑁𝐻 = 𝑋𝑠𝑠 ― 𝑋𝑤𝑠 (5)
3

In the ARS, the main components’ specific energies are calculated as follows:
𝑞𝑔𝑒𝑛 = ℎ1 + (𝐶𝑅 ― 1)ℎ12 ―𝐶𝑅ℎ11 (6)
𝑞𝑎𝑏𝑠 = 𝐶𝑅ℎ8 ― (𝐶𝑅 ― 1)ℎ14 ― ℎ7 (7)
𝑞𝑒𝑣𝑎𝑝 = ℎ6 ― ℎ5 (8)
𝑞𝑐𝑜𝑛 = ℎ3 ― ℎ2 (9)
Pump work per unit refrigerant mass flow rate is defined as:
𝐶𝑅 𝐶𝑅
𝑤𝑝 = (𝑃𝑐𝑜𝑛 ― 𝑃𝑒𝑣𝑎𝑝)𝑣𝑠𝑠 𝜂𝑝 = (ℎ10𝑠 ― ℎ9) 𝜂𝑝 (10)
In this equation 𝜂𝑝 is the isentropic efficiency of a pump and can be calculated as follows:
𝑊𝑝𝑠 ℎ10𝑠 ― ℎ9
𝜂𝑝 = 𝑊𝑝 = ℎ10 ― ℎ9
(11)
where ℎ10𝑠 and ℎ10 are the enthalpies at the exit for the isentropic and actual process,
respectively.
The energy balance and specific energy equations of the SHE can be written as:
𝑇13 = 𝜀𝑆𝐻𝐸𝑇10 +(1 ― 𝜀𝑆𝐻𝐸)𝑇12 (12)
𝑞𝑆𝐻𝐸 = ℎ12 ― ℎ13 (𝐶𝑅 ― 1) = ℎ11 ― ℎ10 𝐶𝑅
( ) ( ) (13)
The RHE increases the thermodynamic performance of the evaporator by reducing the
temperature of a saturated liquid separated from the condenser. The energy balance for the RHE
can be written as:
ℎ7 = ℎ6 + 𝜀𝑅𝐻𝐸(ℎ7 ∗ ― ℎ6) (14)
ℎ4 = ℎ3 ― 𝜀𝑅𝐻𝐸(ℎ7 ― ℎ6)
∗ (15)
where ℎ7 is the enthalpy of superheated vapor at the condenser temperature and the
∗

evaporator pressure. The specific energy of the RHE is calculated as follows:

𝑞𝑅𝐻𝐸 = ℎ3 ― ℎ4 = ℎ7 ― ℎ6 (16)
In the SRHE, the temperature of the solution from the absorber is increased slightly by using
the heat of the superheated steam coming out of the generator. Thus, the system’s COP
increases and the condenser capacity decreases. The energy balance for the SRHE can be
expressed as:
ℎ2 = ℎ1 ― 𝜀𝑆𝑅𝐻𝐸(ℎ1 ― ℎ8 ∗ ) (17)
𝜀𝑆𝑅𝐻𝐸(ℎ1 ― ℎ8 ∗ )
ℎ9 = ℎ8 + 𝐶𝑅
(18)
In case the absorber temperature is greater than the condenser temperature, ℎ8 ∗ is the
enthalpy of superheated vapor at the absorber temperature. Otherwise, ℎ8 ∗ is the enthalpy of
saturated vapor at the condenser temperature.
The system performance is evaluated by the COP, defined as follows:
𝑞𝑒𝑣𝑎𝑝
𝐶𝑂𝑃 = 𝑞𝑔𝑒𝑛 + 𝑤𝑝 (19)
Exergy is described as the maximum theoretical work that can be obtained from a system as
it comes to equilibrium with the environment. Specific exergy for each point in the cycle, 𝑒𝑖 is
described as follows:
𝑒𝑖 = ℎ𝑖 ― ℎ0 ― 𝑇0(𝑠𝑖 ― 𝑠0) (20)
 ACCEPTED MANUSCRIPT

In this equation, 𝑇𝑜 is dead state temperature for exergy calculations, ℎ𝑜 and 𝑠𝑜 are enthalpy
and entropy of the fluid used the in cycle under dead state conditions, respectively. For each
individual component of ARS, the exergy lose rate equations are written as follows:
∆𝐸𝑔𝑒𝑛 = 𝑚𝑠𝑠𝑒11 ― 𝑚𝑤𝑠𝑒12 ― 𝑚𝑁𝐻3𝑒1 + 𝑞𝑔𝑒𝑛[1 ― (𝑇0 𝑇𝑔𝑒𝑛)] (21)
∆𝐸 = 𝑚𝑁𝐻 𝑒 + 𝑚𝑤𝑠𝑒 ― 𝑚𝑠𝑠𝑒 ― 𝑞 [1 ― (𝑇0 𝑇 )]
𝑎𝑏𝑠 3 7 14 8 𝑎𝑏𝑠 𝑎𝑏𝑠 (22)
∆𝐸𝑐𝑜𝑛 = 𝑚𝑁𝐻3𝑒2 ― 𝑚𝑁𝐻3𝑒3 ― 𝑞𝑐𝑜𝑛[1 ― (𝑇0 𝑇𝑐𝑜𝑛)] (23)
∆𝐸 𝑒𝑣𝑎𝑝= 𝑚𝑁𝐻 𝑒 ― 𝑚𝑁𝐻 𝑒 + 𝑞
3 5 [ 1 ― (𝑇 0 𝑇
3 6 𝑒𝑣𝑎𝑝 )]
𝑒𝑣𝑎𝑝 (24)
∆𝐸𝑝 = 𝑚𝑠𝑠(𝑒9 ― 𝑒10) + 𝑤𝑝 (25)
∆𝐸𝑆𝐻𝐸 = 𝑚𝑠𝑠(𝑒10 ― 𝑒11) + 𝑚𝑤𝑠(𝑒12 ― 𝑒13) (26)
∆𝐸𝑅𝐻𝐸 = 𝑚𝑁𝐻3(𝑒3 ― 𝑒4) + 𝑚𝑁𝐻3(𝑒6 ― 𝑒7) (27)
∆𝐸𝑆𝑅𝐻𝐸 = 𝑚𝑁𝐻3(𝑒1 ― 𝑒2) + 𝑚𝑠𝑠(𝑒8 ― 𝑒9) (28)
The exergetic efficiency (eCOP), also called as second law efficiency, is the ratio between
the useful exergy obtained from a system and the useful exergy supplied to the system. eCOP
can be written as follows:
𝑇0
[ ( )]
𝑞𝑒𝑣𝑎𝑝 1 ― 𝑇𝑒𝑣𝑎𝑝

𝑒𝐶𝑂𝑃 = ― 𝑇0 (29)
[ ( )]
𝑞𝑔𝑒𝑛 1 ― 𝑇𝑔𝑒𝑛 + 𝑤𝑝

2.2 Taguchi Method

The Taguchi method, which was developed by Genichi Taguchi, is one of the most
commonly used statistical approaches to obtain the order of importance of different factors for
the target function (Taguchi, 1987). Although, in the beginning, this method was applied to
experimental studies, it can also be applied to numerical and theoretical studies (Arslanoglu and
Yigit, 2017; Naqiuddin et al., 2018; Bao et al., 2013; Hu et al., 2018).
In total, there are eight different parameters considered in this study, such as evaporator
temperature, condenser temperature, generator temperature, absorber temperature,
effectiveness of solution heat exchanger, effectiveness of refrigerant heat exchanger,
effectiveness of solution-refrigerant heat exchanger and pump efficiency. Parameter ranges
(levels) chosen for statistical analysis (see Table 2) are obtained by performing a literature
review which is listed in Table 1.

Table 2
Parameters and levels used in analysis
Parameters Levels 1 2 3
A Evaporator Temperature, Tevap -5 2.5 10
B Condenser Temperature, Tcon 28 33 38
C Generator Temperature, Tgen 90 110 130
D Absorber Temperature, Tabs 28 33 38
E Effectiveness of Solution Heat Exchanger (SHE), ɛshe 0.6 0.75 0.9
F Effectiveness of Refrigerant Heat Exchanger (RHE), ɛrhe 0.6 0.75 0.9
G Effectiveness of Solution-Refrigerant Heat Exchanger (SRHE), ɛsrhe 0.6 0.75 0.9
H Pump Efficiency, ηp 0.6 0.75 0.9

The Taguchi method allows to optimize the system performance and determine the best
options of parameters with less number of experiments or analyses rather than a full factorial
analysis. Significant advantages of Taguchi method is to reduce time, cost and effort. Taguchi
 ACCEPTED MANUSCRIPT

method contains certain orthogonal arrays tables to provide a comprehensive performance of

full factorial analysis. Selecting the orthogonal array type is an essential part of the Taguchi
method. The appropriate orthogonal array table can only be selected after calculating the total
degree of freedom (DOF) which is the sum of the singular degree of freedom of each factor.
The number of each factor levels minus 1 gives the individual DOF for the factor (Bademlioglu
et al., 2018; Tutar et al., 2014). Because there are eight factors with three levels, the total DOF
number is equal to 26. According to the Taguchi methodology, the DOF of the selected
orthogonal array must be higher than the total DOF and for our optimization problem, 27 cases
are enough to determine the order of importance, contribution ratios and optimum working
condition. Thus, the orthogonal array of L27 (38), which was recommended by the Taguchi
technique, is established in this study as shown in Table 3.
In this study, the Taguchi method is applied to ARS to maximize the COP and eCOP of the
system. At the beginning of the Taguchi analysis, the results of the target functions are
converted to the S/N (signal to noise) ratio. Three types of performance characteristics can be
used for the calculating of the S/N ratio i.e., the lower is the better, the higher is the better and
nominal is the best (Ross, 1996). Since both target functions (COP and eCOP) are desired to
be maximized in this study, the higher is the better performance characteristic is chosen. The
S/N ratios for the higher is the better situation is calculated using the Eq. (30):
𝑆 1 𝑛
𝑁 = ―10𝑙𝑜𝑔 ∑ (
𝑛 𝑖 = 11/𝑦𝑖
2
) (30)
where 𝑛 indicates the number of case and 𝑦𝑖 describes the result value for the 𝑖th performance
characteristics.

2.3 Analysis of Variance (ANOVA)

The main goal of ANOVA, which is another statistical approach, is to determine the
contribution ratios of each parameter on the performance characteristic. Due to the fact that
ANOVA demonstrates the importance order of influencing factors on the response, it can also
be used to validate the results obtained from Taguchi method. In this study, the effect of each
parameter on the COP, eCOP and multiple performance characteristics of ARS are determined
using this method and calculation results are analyzed. In the ANOVA analysis, the significance
level of the statistical analyses is 0.05 which corresponds to 95% confidence level. For the
statistical reliability of the results, F-test has been carried out. The F-test (𝐹𝑓𝑎𝑐𝑡𝑜𝑟) values, which
are the ratio between the regression mean square and the mean square error, have been
determined by using ANOVA method. If the calculated F-test value is greater than F value from
the appropriate standard confidence table, the relevant parameter is considered statistically
significant. Following equations can be used for calculating F values, sum of squares (SS),
mean of squares (MS) and DOF of each factor (Ross, 1996):
𝑉𝑓𝑎𝑐𝑡𝑜𝑟
𝐹𝑓𝑎𝑐𝑡𝑜𝑟 = 𝑉𝑒𝑟𝑟𝑜𝑟
(31)
𝑆𝑆𝑓𝑎𝑐𝑡𝑜𝑟
𝑉𝑓𝑎𝑐𝑡𝑜𝑟 = 𝐷𝑂𝐹𝑓𝑎𝑐𝑡𝑜𝑟 (32)
𝐷𝑂𝐹𝑓𝑎𝑐𝑡𝑜𝑟 = 𝑘 ― 1 (33)
∑𝛽𝑓𝑎𝑐𝑡𝑜𝑟,𝑖2 (∑𝛽𝑖) 2

𝑆𝑆𝑓𝑎𝑐𝑡𝑜𝑟 = 𝑁 ― 𝑛 (34)
where 𝐹𝑓𝑎𝑐𝑡𝑜𝑟 indicates whether the factor is associated with the response. The larger 𝐹𝑓𝑎𝑐𝑡𝑜𝑟
value states that the parameter has a greater effect on the response. 𝑉𝑓𝑎𝑐𝑡𝑜𝑟 and 𝑉𝑒𝑟𝑟𝑜𝑟 values are
the variance of the factor and error, respectively. 𝐷𝑂𝐹𝑓𝑎𝑐𝑡𝑜𝑟 is number of factor’s degree of
freedom, 𝑆𝑆𝑓𝑎𝑐𝑡𝑜𝑟 is the sums of squares due to factor, 𝛽𝑓𝑎𝑐𝑡𝑜𝑟,𝑖 is the sum of the S/N ratio at the
𝑖th level of the factor, 𝛽𝑖 is the S/N ratio at the 𝑖th level of the factor, 𝑁 is repeating number of
each level’s factor, 𝑛 is the number of tests. In these equations, “factor” represents the name of
 ACCEPTED MANUSCRIPT

the individual factors (Celik and Turgut, 2012). MS is equal to the ratio of the SS values of each
parameter to the DOF of each parameter.
2.4 Grey Relational Analysis (GRA)
Taguchi is not an applicable method for simultaneously optimizing multiple purpose
functions. Therefore, in this study, GRA method is applied to the ARS to obtain the order of
importance of each design parameters on multiple performance characteristics of the cycle by
maximizing both COP and eCOP, simultaneously. GRA method allows converting an
optimization problem, which has multiple performance characteristics, into a single objective
optimization problem (Deng, 1989).
Firstly, the analysis results of the responses need to be normalized in the range of 0-1. In
GRA, there are three types of equations used for normalization procedure i.e., the lower is the
better, the higher is the better, and nominal is the best (Acir et al., 2017). In this study, the
higher is the better equation described in Eq. (35) is used for both objective functions because
it is desired to maximize both COP and eCOP.
The higher is the better;
𝑥0𝑖 (𝑘) ― min 𝑥0𝑖 (𝑘)
𝑦𝑖(𝑘) = max 𝑥0(𝑘) ― min 𝑥0(𝑘) (35)
𝑖 𝑖

where 𝑦𝑖(𝑘) represents the normalization value of grey relational generation, max 𝑥0𝑖 (𝑘) and
min 𝑥0𝑖 (𝑘) express the maximum and minimum value of 𝑥0𝑖 (𝑘), respectively. Lastly, 𝑥0 denotes
the optimum value.
After the normalization process, the grey relational coefficient (𝜉𝑖) needs to be calculated
using the following Eqs. (36)-(39). The relationship between the ideal and actual normalized
value is expressed by the grey relational coefficient (Acir et al., 2017).
Δmin + 𝜑Δmax
𝜉𝑖(𝑘) = Δ0𝑖(𝑘) + 𝜑Δmax (36)
Δ0𝑖 = ‖𝑦0(𝑘) ― 𝑦𝑖(𝑘)‖ (37)
Δmax = max max ‖𝑦0(𝑘) ― 𝑦𝑖(𝑘)‖ (38)
∀𝑗 ∈ 𝑖 ∀𝑘
Δmin = min min ‖𝑦0(𝑘) ― 𝑦𝑖(𝑘)‖ (39)
∀𝑗 ∈ 𝑖 ∀𝑘
In these equations, φ is the identification coefficient that is limited in the range 0 < 𝜑 < 1.
In GRA, taking any identification coefficient value between 0 and 1 does not change the order
of importance of the parameters. In general, φ=0.5 is used (Turgut et al., 2012; Naqiuddin et
al., 2018; Acir et al., 2017; Kuo et al., 2011; Chamoli et al., 2016; Tosun, 2006). Δ0𝑖 is the
deviation value between 𝑦0(𝑘) and 𝑦𝑖(𝑘). 𝑦0(𝑘) is the referential sequence and 𝑦𝑖(𝑘) is the
comparative sequence. Δmax and Δmin is the maximum and minimum value of the Δ0𝑖,
respectively.
Grey relational grade (𝛾𝑖) denotes the level of correlation between the referential sequence
and comparative sequence. If they are the same, then the grey relational grade equals to 1. The
highest grey relational grade gives the best result where maximum COP and eCOP are achieved
in this case. Grey relational grade (𝛾𝑖) is defined for the different weight factor in Eq. (40)
1 𝑛
𝛾𝑖 = 𝑛∑𝑘 = 1𝑤𝑘𝜉𝑖(𝑘) (40)
where, 𝛾𝑖 is the overall grey relational grade, 𝑛 is the number of test, 𝑤𝑖 represents the
weighting factor for the ith performance characteristic.
Determining the weight factors for each response is very important in GRA. Because in real
engineering problems, the importance of various responses is different. However, most of the
researchers use equal weight to determine the grey relational grade of multiple responses. This
situation makes the reliability of the results questionable. Consequently, it is necessary to follow
an applicable procedure to calculate the weight factors according to the effect of parameters’
variations on COP and eCOP.
 ACCEPTED MANUSCRIPT

Firstly, each parameter’s average S/N ratio ranges (max-min) which is called Delta has to
be calculated for each response (in the fourth line in Table 4 and Table 6). Then, it is necessary
to calculate the sum of the Delta for each response (in the sixth line in Table 4 and Table 6).
Lastly, the ratio of sum of the Delta of each response to the total Delta of all responses (Eq. 41)
gives the weight factors as shown in the last lines in Table 4 and Table 6. The process steps
described above are formulated as follows (Yan and Li, 2013):
𝑝
∑𝑗 = 1𝐷𝑒𝑙𝑡𝑎𝑖,𝑗
𝑤𝑖 = 𝑚 𝑝 (41)
∑𝑖 = 1∑𝑗 = 1𝐷𝑒𝑙𝑡𝑎𝑖,𝑗
In this equation, 𝑚 is the number of response, 𝑝 is the number of parameters and 𝐷𝑒𝑙𝑡𝑎
represents the S/N ratio range.

2.5 Validation
Before proceeding with the statistical analysis, for the same thermodynamic model used in
this study, performance coefficients (COP and eCOP) of the cycle have been calculated for
different operating conditions. The COP values obtained according to the generator
temperature, and eCOP values obtained for different operation parameters are compared with
different studies in the literature (Sozen, 2001; Táboas et al., 2014; Aman et al., 2014; Kumar
et al., 2017) to check on the validity of the model. The operation parameters of the studies given
in Fig. 2 are presented in Table 1. In the study of Sozen (2001) given in Fig 3, the absorber and
condenser temperatures were taken 25°C and 27°C respectively, and the variation of the COP
value based on the generator temperature was investigated for the different evaporator
temperatures. As seen in Figs. 2 and 3, the results of the present study for different operating
conditions are in good agreement with the data available in the literature.

0.65

0.6

0.55
COP

0.5

0.45
Táboas et al. (2014)
Present Study
0.4 Aman et al. (2014)
Present Study
Kumar et al. (2017)
Present Study
0.35
60 80
Generator 100
Temperature (°C) 120

Fig. 2. Variation of COP depending on generator temperatures (Tevap=-15°C, Tcon=Tabs=25°C

in Táboas et al. (2014), Tevap=2°C, Tcon=Tabs=30°C in Aman et al. (2014), Tevap=-5°C,
Tcon=Tabs=30°C in Kumar et al. (2017))
 ACCEPTED MANUSCRIPT

0.6
Sozen (Tevap=-20°C) (2001)
Present study (Tevap=-20°C)
Sozen (Tevap=-10°C) (2001)
0.5 Present study (Tevap=-10°C)
Sozen (Tevap=0°C) (2001)
Present study (Tevap=0°C)
0.4 Sozen (Tevap=10°C) (2001)
Present study (Tevap=10°C)

eCOP
0.3

0.2

0.1

0
90 100 110 120 130 140
Generator Temperature (°C)

Fig. 3. Variation of eCOP depending on generator temperatures (Tabs = 27°C, Tcon =25°C)

3. Results and Discussion

In the first part of the analysis, the Taguchi method has been applied to the ARS to find out
the influence of process parameters on the COP and eCOP, separately. In this context; the COP,
eCOP values are calculated from the thermodynamic model, and S/N ratios are determined for
each case as indicated in Table 3. Totally, there are 27 cases which have different levels of
parameters as presented in Table 2. For instance, the COP of the ARS for Case 1 has been
calculated as Tevap = -5°C, Tcon = 28°C, Tgen = 90°C, Tabs = 28°C, ɛshe = ɛrhe = ɛsrhe = ηp = 60%.
COP values are obtained by applying this procedure for other cases, and all analysis results are
shown in Table 3. Next, the effect of each parameter is obtained for the system’s COP. Average
S/N ratios and the importance order of the factors (Rank) on the cycle’s COP are determined,
as shown in Table 4.
 ACCEPTED MANUSCRIPT

Table 3
COP and eCOP variation with their S/N ratios for the L27 orthogonal array
Parameters
A B C D E F G H Results S/N Ratio

Case Levels COP eCOP COP eCOP

1 1 1 1 1 1 1 1 1 0.5291 0.2407 -5.529 -12.370
2 1 1 1 1 2 2 2 2 0.5746 0.2608 -4.813 -11.674
3 1 1 1 1 3 3 3 3 0.6255 0.2829 -4.075 -10.967
4 2 2 2 1 1 1 1 2 0.5458 0.1718 -5.259 -15.300
5 2 2 2 1 2 2 2 3 0.5914 0.1856 -4.562 -14.628
6 2 2 2 1 3 3 3 1 0.6377 0.1923 -3.908 -14.320
7 3 3 3 1 1 1 1 3 0.5632 0.09965 -4.987 -20.030
8 3 3 3 1 2 2 2 1 0.6061 0.1041 -4.349 -19.651
9 3 3 3 1 3 3 3 2 0.6557 0.1127 -3.666 -18.962
10 2 1 3 2 1 2 3 1 0.5567 0.1443 -5.088 -16.815
11 2 1 3 2 2 3 1 2 0.5834 0.1538 -4.681 -16.261
12 2 1 3 2 3 1 2 3 0.6063 0.1596 -4.346 -15.939
13 3 2 1 2 1 2 3 2 0.6333 0.09079 -3.968 -20.839
14 3 2 1 2 2 3 1 3 0.6409 0.09373 -3.864 -20.562
15 3 2 1 2 3 1 2 1 0.6524 0.09311 -3.710 -20.620
16 1 3 2 2 1 2 3 3 0.4860 0.1793 -6.267 -14.928
17 1 3 2 2 2 3 1 1 0.5267 0.1928 -5.569 -14.298
18 1 3 2 2 3 1 2 2 0.5622 0.2046 -5.002 -13.782
19 3 1 2 3 1 3 2 1 0.5990 0.07537 -4.451 -22.456
20 3 1 2 3 2 1 3 2 0.6167 0.07763 -4.199 -22.199
21 3 1 2 3 3 2 1 3 0.6226 0.07929 -4.116 -22.016
22 1 2 3 3 1 3 2 2 0.5004 0.1635 -6.014 -15.730
23 1 2 3 3 2 1 3 3 0.5220 0.1701 -5.647 -15.386
24 1 2 3 3 3 2 1 1 0.5565 0.1806 -5.091 -14.866
25 2 3 1 3 1 3 2 3 0.5640 0.1246 -4.974 -18.090
26 2 3 1 3 2 1 3 1 0.5739 0.1215 -4.823 -18.308
27 2 3 1 3 3 2 1 2 0.5940 0.1310 -4.524 -17.655
 ACCEPTED MANUSCRIPT

Table 4
Factorial effect and weight factor for COP (Taguchi)
Parameters
Level A B C D E F G H
1 -5.334 -4.589 -4.476 -4.572 -5.171 -4.834 -4.847 -4.724
2 -4.685 -4.669 -4.815 -4.722 -4.723 -4.753 -4.691 -4.681
3 -4.145 -4.907 -4.874 -4.871 -4.271 -4.578 -4.627 -4.679
Delta (max-min) 1.189 0.318 0.398 0.299 0.9 0.256 0.22 0.079
Rank 1 4 3 5 2 6 7 8
∑ Delta 3.659
Weight 22.75%

Fig. 4 indicates each parameter’s effect on the COP. According to the Taguchi method, the
effect of each factor on the target function can be symbolized by the variation of each
parameter’s S/N ratio. If the variation between the highest and lowest S/N ratios is small, then
the effect of the factor on the target function is relatively low. Additionally, the optimum
conditions of the system are obtained from the level of design parameters which have the
maximum S/N ratio. Therefore, optimum values of design parameter for the maximum COP
are determined as Tevap = 10°C (A3), Tcon = 28°C (B1), Tgen = 90°C (C1), Tabs = 28°C (D1), ɛshe =
90% (E3), ɛrhe = 90% (F3), ɛsrhe = 90% (G3), ηp = 90% (H3). The highest COP within the studied
operating condition range is found as 0.697 from the thermodynamic analysis when the cycle
is operated under the optimum working condition (A3B1C1D1E3F3G3H3).

Fig. 4. Effect of each parameter on COP (Taguchi)

The contribution ratios for each parameter on the ARS performance are determined and the
Taguchi analysis results are validated by using the ANOVA method for the statistical reliability.
In the ANOVA analysis, the impacts of process parameters on the efficiencies have been found
as a result of F-test and 95% reliable test comparisons. The contribution ratios obtained based
on the ANOVA method are presented in Table 5 and Fig. 5. According to the ANOVA results,
it is shown that the evaporator temperature and the effectiveness of solution heat exchanger are
the most effective parameters on the COP of the ARS with an impact ratio of 50.37% and
28.8%, respectively. Analysis results also indicate that the importance order of the operation
parameters for COP of the cycle is determined based on the impact ratios: Tevap > ɛshe > Tgen >
Tcon > Tabs > ɛrhe > ɛsrhe > ηp. Moreover, it can be seen that Taguchi and ANOVA results give
the same ranking for COP.
 ACCEPTED MANUSCRIPT

Table 5
ANOVA results and contribution ratios on COP
Parameters DOF SS MS F Contribution
(%)
Evaporator Temperature, Tevap 2 6.3752 3.1876 92.81 50.37
Condenser Temperature, Tcon 2 0.4929 0.24644 7.18 3.89
Generator Temperature, Tgen 2 0.8319 0.41596 12.11 6.57
Absorber Temperature, Tabs 2 0.4021 0.20103 5.85 3.18
Effectiveness of SHE, ɛshe 2 3.6446 1.82231 53.06 28.80
Effectiveness of RHE, ɛrhe 2 0.3072 0.15362 4.47 2.43
Effectiveness of SRHE, ɛsrhe 2 0.2301 0.11503 3.35 1.82
Pump Efficiency, ηp 2 0.0284 0.01421 0.41 0.22
Error 10 0.3435 0.03435 2.71
Total 26 12.6558 100

60

50
Contribution Ratio (%)

40

30

20

10

0
A B C D E F G H
Fig. 5. Contribution ratio of each parameter on COP (ANOVA)

Taguchi analysis is performed for eCOP after obtaining significant parameters and their
impact ratios for COP. The same procedure is applied to eCOP results and the importance order
of the design parameters on the eCOP are determined by the Taguchi method and the impact
ratios are determined by ANOVA. The order of parameters (Rank) based on the magnitude of
the impact on eCOP is shown in Table 6.
 ACCEPTED MANUSCRIPT

Table 6
Factorial effect and weight factor for eCOP (Taguchi)
Parameters
Level A B C D E F G H
1 -13.78 -16.74 -16.79 -15.32 -17.4 -17.1 -17.04 -17.08
2 -16.37 -16.92 -17.1 -17.12 -17 -17.01 -16.95 -16.93
3 -20.82 -17.3 -17.12 -18.52 -16.57 -16.85 -16.97 -16.95
Delta (max-min) 7.04 0.56 0.34 3.2 0.83 0.25 0.09 0.14
Rank 1 4 5 2 3 6 8 7
∑ Delta 12.43
Weight 77.25%

According to response results given in Table 6, the importance order of the design
parameters for eCOP is determined as follows: Tevap > Tabs > ɛshe > Tcon > Tgen > ɛrhe > ηp >
ɛsrh. Fig. 6 shows each parameter’s effect on the eCOP. The effect of other parameters on the
eCOP except the evaporator and absorber temperatures is seen to be very low.
Optimum values of process parameter for the maximum eCOP are determined as Tevap = -
5°C (A1), Tcon = 28°C (B1), Tgen = 90°C (C1), Tabs = 28°C (D1), ɛshe = 90% (E3), ɛrhe = 90% (F3),
ɛsrhe = 90% (G3), ηp = 90% (H3). The eCOP value is calculated as 0.2829 from the
thermodynamic model under A1B1C1D1E3F3G3H3 operating conditions which give the
maximum eCOP.

Fig. 6. Effect of each parameter on eCOP (Taguchi)

As mentioned before, the ANOVA method allows obtaining the impact ratios of the
parameters for a target function. From the ANOVA method, it is found that the evaporator
temperature has a considerable impact on the eCOP of the cycle with an impact ratio of 81.38%
as shown in Table 7. The absorber temperature follows the evaporator temperature with an
impact ratio of 16.53%. The effect of the other six parameters on the eCOP is quite small as
shown in Fig. 7. According to the results of Taguchi and ANOVA analysis, it is observed that
both methods give the same ranking (importance order) for eCOP.
 ACCEPTED MANUSCRIPT

Table 7
ANOVA results and contribution ratios on eCOP
Parameters DOF SS MS F Contribution
(%)
Evaporator Temperature, Tevap 2 228.017 114.009 3290.68 81.38
Condenser Temperature, Tcon 2 1.459 0.73 21.06 0.52
Generator Temperature, Tgen 2 0.544 0.272 7.84 0.19
Absorber Temperature, Tabs 2 46.311 23.155 668.34 16.53
Effectiveness of SHE, ɛshe 2 3.069 1.535 44.3 1.10
Effectiveness of RHE, ɛrhe 2 0.297 0.149 4.29 0.11
Effectiveness of SRHE, ɛsrhe 2 0.039 0.019 0.56 0.01
Pump Efficiency, ηp 2 0.113 0.057 1.64 0.04
Error 10 0.346 0.035 0.12
Total 26 280.196 100

90

80

70
Contribution Ratio (%)

60

50

40

30

20

10

0
A B C D E F G H
Fig. 7. Contribution ratio of each parameter on eCOP (ANOVA)

The optimum working conditions obtained from the above analysis are found to be different
for COP and eCOP. Therefore, the multi objective function approach is applied by using an
alternative method. In this study, the GRA method is used to perform an optimization study by
reducing the two objective functions (maximizing the COP and eCOP of the cycle) to a single
function, which is called as multiple performance characteristics. It is desirable to maximize
the multiple performance characteristics by obtaining the highest possible COP and eCOP,
simultaneously. In the mathematical model section, the process steps of the GRA method are
explained. In this section, normalized results, grey relational coefficients (GRC), grey relational
grade (GRG) and order are calculated and presented in Table 8. As mentioned in the previous
section, determining the weight factors of each objective function (COP and eCOP) on multiple
performance characteristics is crucial to obtain realistic results in the GRA. Using the equal
weight factors of each objective function (50%-50%) may not be an appropriate approach.
Consequently, before determining the grey relational grade, the weight factors of COP and
eCOP have been calculated as 22.75% and 77.25% by using Eq. (41), respectively, and they
 ACCEPTED MANUSCRIPT

are presented in Table 4 and Table 6. Thus, the equation for the grey relational grade is
determined as follows:
𝐺𝑅𝐺 = 0.2275 𝐺𝑅𝐶𝐶𝑂𝑃 +0.7725 𝐺𝑅𝐶𝑒𝐶𝑂𝑃 (42)
In this equation, 𝐺𝑅𝐶𝐶𝑂𝑃 and 𝐺𝑅𝐶𝑒𝐶𝑂𝑃 represent the grey relational coefficient of COP and
eCOP, respectively.

Table 8
Normalized results, grey relational coefficients, grey relational grade and order

Case Normalized Grey relational

Results coefficients
Grey relational Order
COP eCOP COP eCOP grade
1 0.2540 0.7967 0.4013 0.7109 0.6405 3
2 0.5221 0.8935 0.5113 0.8244 0.7532 2
3 0.8220 1.0000 0.7375 1.0000 0.9403 1
4 0.3524 0.4647 0.4357 0.4829 0.4722 12
5 0.6211 0.5312 0.5689 0.5161 0.5281 6
6 0.8939 0.5634 0.8250 0.5339 0.6001 4
7 0.4549 0.1170 0.4784 0.3615 0.3881 27
8 0.7077 0.1384 0.6311 0.3672 0.4272 21
9 1.0000 0.1799 1.0000 0.3788 0.5200 7
10 0.4166 0.3321 0.4615 0.4281 0.4357 20
11 0.5740 0.3779 0.5399 0.4456 0.4671 14
12 0.7089 0.4059 0.6320 0.4570 0.4968 9
13 0.8680 0.0743 0.7911 0.3507 0.4509 17
14 0.9128 0.0885 0.8515 0.3542 0.4673 13
15 0.9806 0.0855 0.9626 0.3535 0.4920 11
16 0.0000 0.5008 0.3333 0.5004 0.4624 15
17 0.2398 0.5658 0.3968 0.5352 0.5038 8
18 0.4490 0.6227 0.4757 0.5699 0.5485 5
19 0.6659 0.0000 0.5994 0.3333 0.3939 26
20 0.7702 0.0109 0.6851 0.3358 0.4152 24
21 0.8049 0.0189 0.7194 0.3376 0.4244 22
22 0.0849 0.4247 0.3533 0.4650 0.4396 19
23 0.2121 0.4565 0.3882 0.4791 0.4585 16
24 0.4154 0.5071 0.4610 0.5036 0.4939 10
25 0.4596 0.2372 0.4806 0.3960 0.4152 25
 ACCEPTED MANUSCRIPT

26 0.5180 0.2223 0.5092 0.3913 0.4181 23

27 0.6364 0.2681 0.5790 0.4059 0.4452 18

The variation of the grey relational grade is illustrated in Fig. 8. The case that has the highest
grey relational grade among all cases gives the maximum multiple performance characteristic.
In this study, it is seen that Case 3 gives the maximum multiple performance characteristic that
can be obtained under these working conditions. For the best condition (A1B1C1D1E3F3G3H3)
COP and eCOP of the system are found as 0.6255 and 0.2829, respectively. Moreover, the
worst condition in terms of first and second law efficiencies has been determined as
A3B3C3D3E1F1G1H1. For the worst condition COP and eCOP of the system are found as 0.5559
and 0.06883, respectively.

Fig. 8. Grey relational grade for the maximum performance characteristic

Response table is generated for the grey relational grade to determine the importance order
of each parameter. According to data, the importance orders are determined as follows: Tgen >
Tevap > Tabs > ɛshe > Tcon > ɛrhe > ɛsrh > ηp as shown in Table 9. Each parameter’s impact on the
multiple performance characteristics of the cycle is presented in Fig. 9. Regarding the effect on
the cycle’s multiple performance characteristics, the absorber and evaporator temperature are
in 1st and 2nd place among the parameters examined as shown in Fig. 9.

Table 9
Response table for grey relational grade
Parameters
Level A B C D E F G H
1 0.5823 0.5519 0.5581 0.5855 0.4554 0.4811 0.4780 0.4895
2 0.4754 0.4892 0.4832 0.4805 0.4932 0.4912 0.4994 0.5031
3 0.4421 0.4587 0.4585 0.4338 0.5512 0.5275 0.5224 0.5090
Delta (max-min) 0.1402 0.0932 0.0995 0.1518 0.0959 0.0464 0.0443 0.0196
Rank 2 5 3 1 4 6 7 8
 ACCEPTED MANUSCRIPT

Fig. 9. Effect of each parameter on multiple performance characteristics.

The results of grey relational grade are evaluated within the ANOVA method and
contribution ratios of each parameter are presented in Table 10. ANOVA method indicates that
the absorber and evaporator temperatures are two parameters with the greatest effect on
multiple performance characteristics with an impact ratio of 29.66% and 26.34%, respectively
(Fig. 10). The generator temperature, effectiveness of solution heat exchanger and condenser
temperature have almost the same influence on the response with a contribution ratio of 13.20%,
11.45% and 11.08%, respectively. Besides, the effectiveness of refrigerant heat exchanger,
effectiveness of solution-refrigerant heat exchanger and pump efficiency affect the performance
coefficients of ARS slightly. These results are in accordance with Table 9, which is the response
table for the grey relational grade.

Table 10
ANOVA table for grey relational grade
Parameters DOF SS MS F Contribution
(%)
Evaporator Temperature, Tevap 2 0.096547 0.048273 53.62 26.34
Condenser Temperature, Tcon 2 0.040618 0.020309 22.56 11.08
Generator Temperature, Tgen 2 0.048378 0.024189 26.87 13.20
Absorber Temperature, Tabs 2 0.108729 0.054364 60.38 29.66
Effectiveness of SHE, ɛshe 2 0.041979 0.020989 23.31 11.45
Effectiveness of RHE, ɛrhe 2 0.010696 0.005348 5.94 2.92
Effectiveness of SRHE, ɛsrhe 2 0.008839 0.00442 4.91 2.41
Pump Efficiency, ηp 2 0.001747 0.000873 0.97 0.48
Error 10 0.009004 0.0009 2.46
Total 26 0.366536 100
 ACCEPTED MANUSCRIPT

35

30

Contribution Ratio (%)

25

20

15

10

0
A B C D E F G H
Fig. 10. Contribution ratio of each parameter on multiple performance characteristics
(ANOVA)

4. Conclusion
This study has presented the detailed results of the statistical analysis performed to analyze
the effect of various parameters on the ARS performance. The effects of the process parameters
on the first and second law efficiencies (COP and eCOP) of the cycle are investigated,
separately. Furthermore, the GRA method is used to obtain the order of importance and
contribution ratios of the parameters on the multi-objective function (considering the COP and
eCOP, simultaneously). The validation is performed with the studies that use the same
thermodynamic model with this study.
Analysis results show that the importance order of the operation factors for COP is
determined as follows: Tevap > ɛshe > Tgen > Tcon > Tabs > ɛrhe > ɛsrhe > ηp. According to the
ANOVA results, it is shown that the evaporator temperature and the effectiveness of solution
heat exchanger are the most effective parameters on the COP with the impact ratio of 50.37%
and 28.8%, respectively. Moreover, the optimum working condition for the maximum COP is
obtained in the case of A3B1C1D1E3F3G3H3 and under these conditions, the highest COP is
found as 0.697 from the thermodynamic analysis.
The order of parameters based on the magnitude of the effect on eCOP is found as: Tevap >
Tabs > ɛshe > Tcon > Tgen > ɛrhe > ηp > ɛsrh. The evaporator and absorber temperatures have a
dominant effect on eCOP such that these two parameters are found equal to 97% of the total
impact. Furthermore, the maximum eCOP value is calculated as 0.2829 under optimum
operating conditions for eCOP which is A1B1C1D1E3F3G3H3.
Additionally, an optimization study that maximizes both COP and eCOP is performed with
GRA in this study. The effect of absorber, evaporator and generator temperatures, effectiveness
of solution heat exchanger, condenser temperature, effectiveness of refrigerant heat exchanger,
effectiveness of solution-refrigerant heat exchanger and pump efficiency on the multiple
performance characteristics are with contribution ratios of 29.66%, 26.34%, 13.20%, 11.45%,
11.08%, 2.92%, 2.41% and 0.48%, respectively. The best and worst operating conditions of
ARS for multiple performance characteristics are determined as A1B1C1D1E3F3G3H3 and
A3B3C3D3E1F1G1H1, respectively. The best condition corresponds to the highest first and
second law efficiencies simultaneously. For the best condition, COP and eCOP of the system
are found as 0.6255 and 0.2829, respectively. For the worst condition, COP and eCOP of the
system are found as 0.5559 and 0.06883, respectively.
 ACCEPTED MANUSCRIPT

Consequently, this optimization study will be a useful resource for researchers because it
allows designing more efficient ARS considering the optimum working conditions and the
contribution ratios of each parameter.

Nomenclature

ARS absorption refrigeration system

COP coefficient of performance
CR circulation ratio
DOF degree of freedom
eCOP exergetic coefficient of performance
e specific exergy (kJ/kg)
∆𝐸 exergy lose rate (kW)
GRA grey relational analysis
GRC grey relational coefficient
GRG grey relational grade
h specific enthalpy (kJ/kg)
𝑚 mass flow rate (kg/s)
MS mean of squares
P pressure (kPa)
q specific energy (kJ/kg)
RHE refrigerant heat exchanger
SHE solution heat exchanger
SRHE solution-refrigerant heat exchanger
s specific entropy (kJ/kg K)
SS sum of squares
S/N signal to noise
T temperature (°C)
v specific volume (m3/kg)
𝑋 solution concentration
W work (kW)
w weight factor
ε effectiveness of heat exchanger
η efficiency (%)

Subscript
abs absorber
con condenser
evap evaporator
gen generator
p pump
s isentropic process
ss strong solution
ws weak solution
 ACCEPTED MANUSCRIPT

i point
0 dead state

References

Abed, A.M., Alghoul, M.A., Sirawn, R., Al-Shamani, A.N., Sopian, K., 2015. Performance
enhancement of ejector-absorption cooling cycle by re-arrangement of solution streamlines and
adding RHE. Appl. Therm. Eng. 77, 65-75.
https://doi.org/10.1016/j.applthermaleng.2014.12.003

Acir, A., Canli, M.E., Ata, I., Cakiroglu, R., 2017. Parametric optimization of energy and
exergy analyses of a novel solar air heater with grey relational analysis. Appl. Therm. Eng. 122,
330-338.
https://doi.org/10.1016/j.applthermaleng.2017.05.018

Aman, J., Ting, D.S.K., Henshaw, P., 2014. Residential solar air conditioning: Energy and
exergy analyses of an ammonia-water absorption cooling system. Appl. Therm. Eng. 62, 424-
432.
https://doi.org/10.1016/j.applthermaleng.2013.10.006

Arslanoglu, N., Yigit, A., 2017. Investigation of efficient parameters on optimum insulation
thickness based on theoretical-Taguchi combined method. Environ. Prog. Sustain. Energy 36,
1824-1831.
https://doi.org/10.1002/ep.12628

Bademlioglu, A.H., Canbolat, A.S., Yamankaradeniz, N., Kaynakli, O., 2018. Investigation of
parameters affecting Organic Rankine Cycle efficiency by using Taguchi and ANOVA
methods. Appl. Therm. Eng. 145, 221-228.
https://doi.org/10.1016/j.applthermaleng.2018.09.032

Bao, Z., Yang, F., Wu, Z., Nyamsi, S.N., Zhang, Z., 2013. Optimal design of metal hydride
reactors based on CFD–Taguchi combined method. Energy Convers. Manage. 65, 322-330.
https://doi.org/10.1016/j.enconman.2012.07.027

Celik, N., Turgut, E., 2012. Design analysis of an experimental jet impingement study by using
Taguchi method. Heat Mass Transfer 48, 1408-1413.
https://doi.org/10.1007/s00231-012-0989-7

Celik, N., Pusat, G., Turgut, E., 2018. Application of Taguchi method and grey relational
analysis on a turbulated heat exchanger. Int. J. Therm. Sci. 124, 85-97.
https://doi.org/10.1016/j.ijthermalsci.2017.10.007

Chamoli, S., Yu, P., Kumar, A., 2016. Multi-response optimization of geometric and flow
parameters in a heat exchanger tube with perforated disk inserts by Taguchi grey relational
analysis. Appl. Therm. Eng. 103, 1339-1350.
https://doi.org/10.1016/j.applthermaleng.2016.04.166

Coskun, S., Motorcu, A.R., Yamankaradeniz, N., Pulat, E., 2012. Evaluation of control
parameters’ effects on system performance with Taguchi method in waste heat recovery
application using mechanical heat pump. Int. J. Refrig. 35, 795-809.
 ACCEPTED MANUSCRIPT

https://doi.org/10.1016/j.ijrefrig.2011.12.008

Deng, J., 1989. Introduction to grey system. J. Grey Syst. 1, 1-24.

Fernández-Seara, J., Sieres, J., 2006. Ammonia-water absorption refrigeration systems with
flooded evaporators. Appl. Therm. Eng. 26, 2236-2246.
https://doi.org/10.1016/j.applthermaleng.2006.03.011

Hu, Y., Yang, J., Wang, J., Wang, Q., 2018. Investigation of hydrodynamic and heat transfer
performances ingrille-sphere composite pebble beds with DEM-CFD-Taguchi method. Energy,
155, 909-920.
https://doi.org/10.1016/j.energy.2018.05.018

Joybari, M.M., Haghighat, F., 2016. Exergy analysis of single effect absorption refrigeration
systems: The heat exchange aspect. Energy Convers. Manage. 126, 799-810.
https://doi.org/10.1016/j.enconman.2016.08.029

Karamangil, M.I., Coskun, S., Kaynakli, O., Yamankaradeniz, N., 2010. A simulation study of
performance evaluation of single- stage absorption refrigeration system using conventional
working fluids and alternatives. Renew. Sustain. Energy Rev. 14, 1969-1978.
https://doi.org/10.1016/j.rser.2010.04.008

Kaynakli, O., Yamankaradeniz, R., 2003. Effect on performance of heat exchangers used in
absorption refrigeration systems. Uludag Univ. J. Fac. Eng. Arch. 8, 111-120.

Kaynakli, O., Kilic, M., 2007. Theoretical study on the effect of operating conditions on
performance of absorption refrigeration system. Energy Convers. Manage. 48, 599-607.
https://doi.org/10.1016/j.enconman.2006.06.005

Ketfi, O., Merzouk, M., Merzouk, N.K., El-Metenan, S., 2015. Performance of a single effect
solar absorption cooling system. Energy Procedia 74, 130-138.
https://doi.org/10.1016/j.egypro.2015.07.534

Kumar, V., Pandya, B., Patel, J., Matawala, V., 2017. Cut-off temperature evaluation and
performance comparison from energetic and exergetic perspective for NH3-H2O absorption
refrigeration system. Therm. Sci. Eng. Prog. 4, 97-105.
https://doi.org/10.1016/j.tsep.2017.09.007

Kuo, C.F.J., Su, T.L., Jhang, P.R., Huang, C.Y., Chiu, C.H., 2011. Using the Taguchi method
and grey relational analysis to optimize the flat-plate collector process with multiple quality
characteristics in solar energy collector manufacturing. Energy 36, 3554-3562.
https://doi.org/10.1016/j.energy.2011.03.065

Li, K., Cai, D., Liu, Y., Jiang, J., Sun, W., He, G., 2017. Thermodynamic analysis of a novel
exhaust heat-driven non-adiabatic ejection-absorption refrigeration cycle using R290/oil
mixture. Energy Convers. Manage. 149, 244-253.
https://doi.org/10.1016/j.enconman.2017.07.021

Modi, B., Mudgal, A., Patel, B., 2017. Energy and exergy investigation of small capacity single
effect lithium bromide absorption refrigeration system. Energy Procedia 109, 203-210.
 ACCEPTED MANUSCRIPT

https://doi.org/10.1016/j.egypro.2017.03.040

Naqiuddin, N.H., Saw, L.H., Yew, M.C., Yusof, F., Poon, H.M., Cai, Z., Thiam, H.S., 2018.
Numerical investigation for optimizing segmented micro-channel heat sink by Taguchi-Grey
method. Appl. Energy 222, 437-450.
https://doi.org/10.1016/j.apenergy.2018.03.186

Novella, R., Dolz, V., Martín, J., Royo-Pascual, L., 2017. Thermodynamic analysis of an
absorption refrigeration system used to cool down the intake air in an Internal Combustion
Engine. Appl. Therm. Eng. 111, 257-270.
https://doi.org/10.1016/j.applthermaleng.2016.09.084

Ouadha, A., El-Gotni, Y., 2013. Integration of an ammonia-water absorption refrigeration

system with a marine Diesel engine: A thermodynamic study. Procedia Comput. Sci. 19, 754-
761.
https://doi.org/10.1016/j.procs.2013.06.099

Ross, P.J., 1996. Taguchi Techniques for Quality Engineering, second ed. McGraw Hill, New
York.

Saleh, A., Mosa, M., 2014. Optimization study of a single-effect water-lithium bromide
absorption refrigeration system powered by flat-plate collector in hot regions. Energy Convers.
Manage. 87, 29-36.
https://doi.org/10.1016/j.enconman.2014.06.098

Sencan, A., 2007. Performance of ammonia-water refrigeration systems using artificial neural
networks. Renew. Energy 32, 314-328.
https://doi.org/10.1016/j.renene.2006.01.003

Sozen, A., 2001. Effect of heat exchangers on performance of absorption refrigeration systems.
Energy Convers. Manage. 42, 1699-1716.
https://doi.org/10.1016/S0196-8904(00)00151-5

Táboas, F., Bourouis, M., Vallès, M., 2014. Analysis of ammonia/water and ammonia/salt
mixture absorption cycles for refrigeration purposes in fishing ships. Appl. Therm. Eng. 66,
603-611.
https://doi.org/10.1016/j.applthermaleng.2014.02.065

Taguchi, G., 1987. Taguchi Techniques for Quality Engineering, Quality Resources, New
York.

Tosun, N., 2006. Determination of optimum parameters for multi-performance characteristics

in drilling by using grey relational analysis. Int. J. Adv. Manuf. Technol. 28, 450-455.
https://doi.org/10.1007/s00170-004-2386-y

Tugcu, A., Arslan, O., Kose, R., Yamankaradeniz, N., 2016. Thermodynamics and economical
analysis of geothermal assisted absorption refrigeration system: Simav case study. J. Therm.
Sci. Tech. 36, 143-159.
 ACCEPTED MANUSCRIPT

Turgut, E., Cakmak, G., Yildiz, C., 2012. Optimization of the concentric heat exchanger with
injector turbulators by Taguchi method. Energy Convers. Manage. 53, 268-275.
https://doi.org/10.1016/j.enconman.2011.09.011

Tutar, M., Aydin, H., Yuce, C., Yavuz, N., Bayram, A., 2014. The optimisation of process
parameters for friction stir spot-welded AA3003-H12 aluminium alloy using a Taguchi
orthogonal array. Mater. Des. 63, 789-797.
https://doi.org/10.1016/j.matdes.2014.07.003

Verma, V., Murugesan, K., 2014. Optimization of solar assisted ground source heat pump
system for space heating application by Taguchi method and utility concept. Energy Build. 82,
296-309.
https://doi.org/10.1016/j.enbuild.2014.07.029

Yakut, K., Alemdaroglu, N., Sahin, B., Celik, C., 2006. Optimum design-parameters of a heat
exchanger having hexagonal fins. Appl. Energy 83, 82-98.
https://doi.org/10.1016/j.apenergy.2005.01.007

Yan, J., Li, L., 2013. Multi-objective optimization of milling parameters – the trade-offs
between energy, production rate and cutting quality. J. Cleaner Prod. 52, 462-471.
https://doi.org/10.1016/j.jclepro.2013.02.030

Zeng, M., Tang, L.H., Lin, M., Wang, Q.W., 2010. Optimization of heat exchangers with
vortex-generator fin by Taguchi method. Appl. Therm. Eng. 30, 1775-1783.
https://doi.org/10.1016/j.applthermaleng.2010.04.009
 ACCEPTED MANUSCRIPT

HIGHLIGHTS

 The process parameters of ARS and their effects are examined statistically.
 The effects of the process parameters on COP and eCOP are investigated, separately.
 Contribution ratios of the parameters on the multi-objective function are obtained.
 The best/worst conditions of different parameters are determined for COP and eCOP.
 Absorber and evaporator temperatures are found to be the most effective parameters.