International Communications in Heat and Mass Transfer 63 (2015) 35–40

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer

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

Modeling of thermal conductivity of ZnO-EG using experimental data

and ANN methods☆
Mohammad Hemmat Esfe a,⁎, Seyfolah Saedodin b, Ali Naderi b, Ali Alirezaie b, Arash Karimipour a,
Somchai Wongwises c,⁎, Marjan Goodarzi d, Mahidzal bin Dahari e
a
Department of Mechanical Engineering, Najafabad Branch, Islamic Azad University, Isfahan, Iran
b
Faculty of Mechanical Engineering, Semnan University, Semnan, Iran
c
Fluid Mechanics, Thermal Engineering and Multiphase Flow Research Lab. (FUTURE), Department of Mechanical Engineering, Faculty of Engineering,
King Mongkut's University of Technology Thonburi, Bangmod, Bangkok 10140, Thailand
d
Young Researchers and Elite Club, Mashhad Branch, Islamic Azad University, Mashhad, Iran
e
Department of Mechanical Engineering, Faculty of Engineering, University of Malaya, 50603, Kuala Lumpur, Malaysia

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

Available online 7 February 2015 In the present study, the thermal conductivity of the ZnO-EG nanofluid has been investigated experimentally. For
this purpose, zinc oxide nanoparticles with nominal diameters of 18 nm have been dispersed in ethylene glychol
Keywords: at different volume fractions (0.000625, 0.00125, 0.005, 0.01, 0.015, 0.02, 0.03, 0.04, and 0.05) and temperatures
Thermal conductivity (24–50 °C). The two-step method is used to disperse nanoparticles in the base fluid. Based on the experimental
MgO-EG data, an experimental model has been proposed as a function of solid concentration and temperature. Then, the
Artificial neural network
feedforward multilayer perceptron neural network has been employed for modeling thermal conductivity of
Nanofluid
ZnO-EG nanofluid. Out of 40 measured data obtained from experiments, 28 data were selected for network train-
ing, while the remaining 12 data were used for network testing and validating. The results indicate that both
model and ANN outputs are in good agreement with the experimental data.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction and the power consumption of the pump will increase. Hemmat Esfe et al.
[6] have studied the heat transferring of functionalized COOH DWCNTs-
It was 20 years ago when, for the first time, nanoparticles were used in water nanofluids and obtained the friction factor and pressure drop for
common fluids as additives to increase their heat transferring [1]. Since different Reynolds numbers in the experiment. Fuoc et al. [7] have also
then, a wide range of studies have been done in this area and different studied MWCNT nanofluids, which were suspended by chitosan. They
nanoparticles have been used in heat-transferring fluids. Therefore, a col- presented their result based on the cut rate and concluded that this
lection of data has been obtained to produce the fluids commercially and nanofluid shows non-Newtonian behavior. A lot of different studies
to use them in the industrial systems. However, after spending so much have been conducted about the issue [8–10].
money and doing a lot of research in this area, some scholars have claimed Diameter size of the nanoparticles is one of the other factors that has
that studies done on nanoparticles do not have a necessary cohesion, and been noticed in the studies of nanofluids. The investigations [11–14] in-
the obtained results are very different from each other in some cases. dicate that the smaller the diameter of the nanoparticles, the larger the
Most of the research done in this area has focused on thermal con- thermal conductivity. Nasiri et al. [15] reported that in carbon nano-
ductivity in the calculation of nanofluids because the scholars believe tubes, when the diameter size of nanotubes is smaller, the nanofluid
that when thermal conductivity of nanofluids increases, their heat shows greater thermal conductivity.
transferring increases too [2–5]. Additionally, the effect of temperature on the thermal conductivity
However, the studies have not stopped here, and some scholars have of nanofluids has been studied in some studies [16–20], and most of
investigated other features of these fluids. The viscosity of nanofluids is them proved that there is a direct relationship between thermal con-
the second important issue from the scientists' point of view because ductivity fraction and temperature.
adding solid particles to the fluid will cause a pressure drop in the system, The volume fraction of the solid particles is also another feature that
has been investigated by the researchers [21–23]. Thermal conductivity
of nanofluids increases in a non-linear way by an increase in solid vol-
☆ Communicated by W.J. Minkowycz.
⁎ Corresponding authors.
ume fraction.
E-mail addresses: [email protected] (M. Hemmat Esfe), Simultaneously, in experimental studies, analytical and numerical
[email protected] (S. Wongwises). studies have been done on nanofluids, as well [24–26]. Different models

http://dx.doi.org/10.1016/j.icheatmasstransfer.2015.01.001
0735-1933/© 2015 Elsevier Ltd. All rights reserved.
36 M. Hemmat Esfe et al. / International Communications in Heat and Mass Transfer 63 (2015) 35–40

have, thus far, been proposed for predicting thermal conductivity and vis- ZnO-EG nanofluid has been studied by developing a non-linear regres-
cosity of nanofluids. The other method of predicting nanofluids behaviors sion equation that includes the effect of the volume fraction and the
involves using an artificial neural network for conductivity fraction and fluid temperature. Measuring thermal conductivity of the ZnO-EG
viscosity of nanofluids. Hojjat et al. [27] conducted an experimental nanofluid with average particle diameters of 18 nm has been performed
study on thermal conductivity of three different nanofluids, alumina, in different volume fractions and temperatures.
titania, and CuO. Then, they modeled the experimental data using the The desired relationship in this study is KKnbf ¼ f ðT; φÞ, which fits the
neural network. They used three factors—temperature, volume fraction thermal conductivity ratio in terms of temperature and volume fraction.
of nanoparticles, and thermal conductivity of nanoparticles. In another Accordingly, multiple relations were derived. Finally, the best model
study, Hemmat Esfe et al. [28] measured the thermal conductivity of was selected in terms of accuracy. In order to select the best curve-
magnesium oxide-ethylene glycol nanofluids and modeled experimental fitting equation, various criteria can be employed. In this study, the
data by an artificial neural network. Their variables in this experiment mean squared error (MSE) and the mean absolute error (MAE) were
were temperature, volume fraction, and the diameter of the nanoparti- used as the main criteria for investigating the correlated model perfor-
cles. They attained a model that is very accurate. mance, which is calculated by the following relations:

2. Experimental   !2
1X N
K n f  K n f 
MSE ¼ − ð1Þ
2.1. Nanofluid preparation N i¼1 K b Exp K b pred

Measuring thermal conductivity, ZnO-EG nanofluid was provided by

   
N 
two-step methods. Furthermore, the XRD pattern and a transmission of 1X K n f  K n f  
MAE ¼  −  ð2Þ
N i¼1  K b Exp K b pred 
a microscope electron image of the particles, which take from the com-
pany, are displayed in Fig. 1.
First, the ZnO nanoparticles required for the desired volume concen-
trations (i.e., 0.05 (5.0%), 0.04 (4.0%), 0.03 (3.0%), 0.02 (2.0%), 0.015 where N is the number of experimental data used in the correlation,
 
(1.5%), 0.01 (1.0%), 0.005 (0.5%), 0.0025 (0.25%), 0.00125(0.125%), and Kn f  
K  and KKn f  are thermal conductivity of the experimental results
0.000625 (0.0625%)) were weighed and then added with the base fluid. b Exp b pred

After mixing the ZnO nanoparticles with ethylene glychol, the samples and thermal conductivity of the correlated model results, respectively,
were subjected to an ultrasonic vibrator(400 W, 20 KHz) for about 3–5 h. whose difference shows the curve-fitting error between the real and
predicted values.
In order to select the best curve-fitting equation, the MSE and MAE
2.2. Thermal conductivity of the nanofluid
performance criteria were investigated for multiple curve-fitting equa-
tions developed for predicting thermal conductivity in terms of temper-
Thermal conductivity of the ZnO-EG nanofluid has been investigated
ature and volume fraction from regression equations. Finally, Eq. (3)
experimentally. The ZnO-EG nanofluid with particle diameters of 18 nm
was selected as the best correlation equation, which is, indeed, the
has been examined by the KD2 pro instrument at different volume frac-
best equation for the results of the regression analysis of the thermal
tions up to 5% over the temperature ranging from 24 to 50 °C. Fig. 2
conductivity data at different temperatures and volume fractions com-
shows the thermal conductivity ratio with respect to a solid volume
pared to other equations.
fraction at different temperatures.
Knf 2 2
¼ 1:00475 þ 2:26216  φ þ 1:57146  T  φ þ 481:646  φ
2.2.1. Proposed new correlation Kb
In this study, experimental data have been employed as a correlation  expð−66:7522  φÞ− 0:0100301  T  φ  cosð1560:99φÞ
pattern for the experimental results. The thermal conductivity of the ð3Þ

2000

1500
Lin (Counts)

1000

500

0
20 40 60 80
2 Theta Scale

Fig. 1. TEM image and XRD patterns of ZnO nanoparticles.

 M. Hemmat Esfe et al. / International Communications in Heat and Mass Transfer 63 (2015) 35–40 37

1.4 1.4

o
Temperature = 24 C
Temperature = 30 oC
1.3 Temperature = 40 oC 1.3

Thermal Conductivity Ratio

Thermal Conductivity Ratio

Temperature = 50 oC

1.2 1.2

1.1 1.1

1.0 1.0

0.9 0.9
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.9 1.0 1.1 1.2 1.3 1.4
Volume Fraction Experimental Results

Fig. 2. Thermal conductivity ratio via volume fraction in different temperature.

Fig. 3. Correlation regression diagram.

where T and φ represent temperature and volume fraction of the temperature and volume fraction variables are considered as inputs to
nanofluid, respectively. the neural network. Fig. 5 depicts the network training process.
Performance criteria for Eq. (3) have been presented in Table 1. In this study, the feedforward multilayer percepetron neural net-
Fig. 3 compares the present experimental model prediction with ex- work has been used for modeling the thermal conductivity ratio of the
perimental data for thermal conductivity of the ZnO-EG nanofluid at dif- ZnO-EG nanofluid. This network has a good ability in estimating non-
ferent temperatures and volume fractions. It can be observed that the linear relations and is one of the most common neural network models
thermal conductivity ratio obtained from the correlated model is in used in engineering applications.
good agreement with the thermal conductivity ratio measured for all The multilayer perceptron neural network consists of multiple
studied temperatures and volume fractions. layers, each of which has a number of neurons. Each neuron in each
Fig. 4 compares the experimental results with those obtained by the layer is connected to the neurons in the next layer by weight coeffi-
correlated experimental model. As it can be seen, the correlated model cients. An activation function is determined for the neurons in each
properly predicts the thermal conductivity ratio of the nanofluid. It is layer, which is used for calculating the sum of the input weights
determined that the maximum error is about 0.0098 at a volume frac- and the biases of each neuron in order to produce the output neuron.
tion of 4% and in temperature of 40 °C. Developing an ANN involves three steps. The first step is developing
According to the experimental results, a correlation was developed the data required for network training. The second step is evaluating
to estimate the thermal conductivity ratio using a regression equation. different neural network structures in order to select the optimal
A comparison of the results from the experimental data with those of one. Finally, the third step is testing the neural network using data
the correlation model shows that there is good agreement between not previously used in network training. Biased neurons are linked
them, indicating the accuracy of the relation provided in this study. to other neurons in the next layers to develop a constant bias. The
most exact and accurate prediction of neural networks is made
using the tan-sigmoid function for the hidden-layer neurons and
3. Artificial neural network modeling the purelin function for the output-layer neurons. Therefore, the

Artificial neural network modeling is a non-linear statistical tech-

nique that can be used for solving problems that are not solved using
1.4
conventional statistical methods. Neural networks can be used for
Experimental Results
modeling complex relations between outputs and inputs or finding Correlated Model Prediction
patterns between data. Important benefits of ANN compared to other 1.3
Thermal Conductivity Ratio

specialized systems include high speed, simplicity, the capability of

modeling multivariate problems for solving complex relations between
variables, and the ability to derive non-linear relations through training 1.2
data.
The proper selection of input parameters plays an important role in
the ANN method and can improve the predicting quality. This selection 1.1
is usually based on the physical background of the problem. In order to
predict the thermal conductivity ratio for every type of nanofluid,
1.0

Table 1
0.9
Performance of the correlated model.
0 10 20 30 40
Mean square error 1.0829906*10−5 Data Number
Mean absolute error 0.0022070592
Maximum error 0.0098439036
Fig. 4. Comparison between experimental results with those obtained by correlated ex-
Correlation coefficient 0.99949742
perimental model.
38 M. Hemmat Esfe et al. / International Communications in Heat and Mass Transfer 63 (2015) 35–40

Fig. 6. Three layer neural network with two hidden layers and one output layer.

Volume fraction (φ): 0.000625, 0.00125, 0.005, 0.01, 0.015, 0.02,

0.03, 0.04, 0.05

Temperature (T): 24, 30, 40, 50

Out of the 40 data obtained from the experiments, 28 of the data are
used for network training, while the remaining 12 data are used for net-
work testing and validating.
The existence of the input parameters in different ranges may lead to
inadequate network training. This should be considered regarding the
inputs of this study—that is, volume fraction varying from 0.0025 to
0.25 and temperature varying from 24 to 50 °C. To remove this problem,
input data are normalized in [−1, 1] intervals.
The number of optimal neurons in the hidden layer is determined
trying different networks and comparing their performance functions.
The neural network structure with the lowest MSE and MAE (the per-
formance functions selected in this study) is considered as the most op-
timal network structure. Table 2 presents MSE and MAE values for
artificial neural networks with different structures. According to
Table 2, the three-layer structure with two neurons in each hidden
layer (Fig. 6) is the most optimal network structure for modeling the
thermal conductivity ratio of the ZnO-EG nanofluid in terms of temper-
ature and volume fraction. This structure has been selected based on the
smallest difference between the model outputs and the experimental
results and predicts a thermal conductivity ratio with MSE and MAE
values equal to 6.37 × 10−6 and 1.8 × 10−3, respectively.
Fig. 5. ANN process flowchart.
Fig. 7 shows a network regression diagram after training, the value of
which is over 0.999 in all cases. This regression value confirms good per-
formance of the trained network.
The experimental data and predicted results of the neural network
tan-sigmoid activation function is used in the hidden layer while lin- have been compared with the tested data in Fig. 8. As it can be seen,
ear activation functions are used in the output layer. In hidden layers, the experimental results and model outputs are in good agreement
each neuron calculates the sum of the corresponding bias and the with each other.
input weights and transfers them via the activation function. Fig. 6
shows a three-layer neural network with two hidden layers consisting
Table 2
of three neurons and one/an output layer.
Performance of ANN with different structures.
In training algorithms, training data are given to the network and the
network updates the weights and biases repeatedly until the predicted Number of hidden Number of neurons in each MSE MAE
layers hidden layer
values agree with the desired values in an acceptable and logical man-
ner. The back-propagation algorithm and the Levenberg–Marquardt 1 2 5.50E−05 0.00703
training method have been employed in this modeling. 1 3 5.65E−05 0.0072
1 4 6.70E−05 0.00833
The performance function is a function that is used for calculating er- 1 5 2.96E−05 0.0043
rors during the network training process. Two of the most common per- 1 6 7.28E−05 0.00895
formance functions used in feedforward neural networks are the mean 1 7 4.45E−05 0.0059
squared error (MSE) and the mean absolute error (MAE) (Eqs. (1) and 1 8 1.42E−04 0.0164
2 2 4.89E−05 0.00638
(2)). These two criteria are used for investigating neural network per-
2 3 6.37E−06 0.0018
formance in this study, as well. 2 4 2.16E−05 0.00344
A multilayer perceptron neural network has been used in which the 2 5 8.45E−05 0.0102
volume fraction and the temperature have been considered as network 2 6 1.02E−05 0.00221
inputs for predicting the thermal conductivity ratio (Fig. 5). The thermal 2 7 1.11E−04 0.0131
2 8 3.27E−05 0.00463
conductivity ratio has been measured using the following conditions.
 M. Hemmat Esfe et al. / International Communications in Heat and Mass Transfer 63 (2015) 35–40 39

1.4 1.4

1.3 1.3
ANN Model Prediction

Thermal Conductivity Ratio

1.2 1.2

1.1 1.1

1.0 1.0

0.9 0.9
0.9 1.0 1.1 1.2 1.3 1.4 0.00 0.01 0.02 0.03 0.04 0.05 0.06
Volume Fraction
Experimental Results
Fig. 9. Predicted values for thermal conductivity ratio ZnO-EG nanofluid obtained through
Fig. 7. ANN regression diagram. correlation method and ANN.

Given the above-mentioned points, it can be realized that this model the thermal conductivity ratio value of the ZnO-EG nanofluid can be
is capable to predict the thermal conductivity ratio with high precision. properly predicted using the ANN and correlation models.
In addition, training and test errors are properly close to each other, in-
dicating that the data have been appropriately divided into training and
5. Conclusion
test data.

In this paper, we have mentioned that the artificial neural network is

4. Comparing correlation model and artificial neural network with one of the approaches to obtain an accurate model in nanofluids. Fur-
experimental results thermore, to estimate the thermal conductivity of the ZnO-EG nanofluid,
a correlation with two parameters (temperature and volume concentra-
Fig. 9 shows the predicted values for the thermal conductivity ratio tion) with a very high accuracy has been provided. For that aim, nine dif-
of ZnO-EG nanofluid obtained through the correlation method and the ferent volume concentrations from 0.0625% to 5% and four various
ANN, as a function of the volume fraction, in different temperatures. temperatures have considered. The accuracy of the ANN model and the
In order to better visualize the prediction quality of the models used proposed correlation were up to 10−5 and 10−4, respectively.
in this study, an experimental parity plot and thermal conductivity ratio
prediction for ZnO-EG have been presented in Fig. 10. As it can be seen, Acknowledgment
the predicted values for both models are in good agreement with the
experimental thermal conductivity ratio value. The authors gratefully acknowledge High Impact Research Grant
In this study, we aimed at modeling the thermal conductivity ratio of UM.C/HIR/MOHE/ENG/23 and Faculty of Engineering, University of
the ZnO-EG nanofluid through an artificial neural network (ANN) and Malaya, Malaysia for their support in conducting this research work.
correlation methods using experimental data in the temperature rang- The sixth author would like to thank the “Research Chair Grant”
ing from 24 to50°C and volume fractions between 0.0625% and 5%. In- from the National Science and Technology Development Agency, the
vestigating the results of the developed models shows that there is Thailand Research Fund (IRG5780005) and the National Research
good agreement between them and the experimental data. Therefore, University Project for the support.

1.4 1.4
Experimental Results Correlation
ANN Outputs ANN
1.3 1.3 Fit Line
Thermal Conductivity Ratio

Themal Conductivity Ratio

1.2 1.2

1.1 1.1

1.0 1.0

0.9 0.9
0 10 20 30 40 0.9 1.0 1.1 1.2 1.3 1.4
Data Number Experimental Results

Fig. 8. Network regression diagram. Fig. 10. Experimental parity plot and thermal conductivity ratio prediction for ZnO-EG.
40 M. Hemmat Esfe et al. / International Communications in Heat and Mass Transfer 63 (2015) 35–40

References [16] W. Duangthongsuk, S. Wongwises, Measurement of temperature-dependent ther-

mal conductivity and viscosity of TiO2-water nanofluids, Exp. Therm. Fluid Sci. 33
[1] S.U.S. Choi, Enhancing thermal conductivity of fluids with nanoparticles, Dev. Appl. (4) (2009) 706–714.
Non-Newtonian Flows 31 (66) (1995) 99–105. [17] H. Jiang, H. Li, C. Zan, F. Wang, Q. Yang, L. Shi, Temperature dependence of the sta-
[2] W. Wang, A Comprehensive Model for the Enhanced Thermal Conductivity of bility and thermal conductivity of an oil-based nanofluid, Thermochim. Acta 579
Nanofluids, 2012. 5. (2014) 27–30.
[3] A. Ghadimi, R. Saidur, H.S.C. Metselaar, A review of nanofluid stability properties and [18] S. Halelfadl, P. Estellé, B. Aladag, N. Doner, T. Maré, Viscosity of carbon nanotubes
characterization in stationary conditions, Int. J. Heat Mass Transf. 54 (17–18) (2011) water-based nanofluids: influence of concentration and temperature, vol. 712013.
4051–4068. 111–117.
[4] E. Timofeeva, A. Gavrilov, J. McCloskey, Y. Tolmachev, S. Sprunt, L. Lopatina, J. [19] M. Hemmat Esfe, S. Saedodin, An experimental investigation and new correlation of
Selinger, Thermal conductivity and particle agglomeration in alumina nanofluids: viscosity of ZnO–EG nanofluid at various temperatures and different solid volume
experiment and theory, Phys. Rev. E 76 (6) (2007) 061203. fractions, Exp. Therm. Fluid Sci. 55 (2014) 1–5.
[5] T.P. Teng, Thermal conductivity and phase-change properties of aqueous alumina [20] S.S. Khaleduzzaman, I.M. Mahbubul, I.M. Shahrul, R. Saidur, Effect of particle concen-
nanofluid, Energy Convers. Manag. 67 (2013) 369–375. tration, temperature and surfactant on surface tension of nanofluids, Int. Commun.
[6] M. Hemmat Esfe, S. Saedodin, O. Mahian, S. Wongwises, Heat transfer characteristics Heat Mass Transfer 49 (2013) 110–114.
and pressure drop of COOH-functionalized DWCNTs/water nanofluid in turbulent [21] A. Indhuja, K.S. Suganthi, S. Manikandan, K.S. Rajan, Viscosity and thermal conduc-
flow at low concentrations, Int. J. Heat Mass Transf. 73 (2014) 186–194. tivity of dispersions of gum arabic capped MWCNT in water: influence of MWCNT
[7] T.X. Phuoc, M. Massoudi, R.-H. Chen, Viscosity and thermal conductivity of concentration and temperature, J. Taiwan Inst. Chem. Eng. 44 (3) (2013) 474–479.
nanofluids containing multi-walled carbon nanotubes stabilized by chitosan, Int. J. [22] M. Hemmat Esfe, S. Saedodin, O. Mahian, S. Wongwises, Thermophysical properties,
Therm. Sci. 50 (1) (2011) 12–18. heat transfer and pressure drop of COOH-functionalized multi walled carbon nano-
[8] I.M. Mahbubul, R. Saidur, M.A. Amalina, Latest developments on the viscosity of tubes/water nanofluids, Int. Commun. Heat Mass Transfer 58 (2014) 176–183.
nanofluids, Int. J. Heat Mass Transf. 55 (4) (2012) 874–885. [23] M. Hemmat Esfe, Seyfolah Saedodin, Omid Mahian, S. Wongwises, Thermal Conduc-
[9] T. Yiamsawas, O. Mahian, A.S. Dalkilic, S. Kaewnai, S. Wongwises, Experimental tivity of Al2O3O3/ Water Nanofluids: Measurement, Correlation, Sensitivity Analysis,
studies on the viscosity of TiO2 and Al2O3 nanoparticles suspended in a mixture of and Comparisons with Literature Reports, J. Therm. Anal. Calorim. 117 (2014)
ethylene glycol and water for high temperature applications, Appl. Energy 111 675–681.
(2013) 40–45. [24] J.B. Menaa, A.A.U. de Moraesb, Y.R. Benitoa, G. Ribatskib, J.A.R. Parise, Extrapolation
[10] M. Kole, T.K. Dey, Effect of aggregation on the viscosity of copper oxide–gear oil of Al2O3–water nanofluid viscosity for temperatures and volume concentrations be-
nanofluids, Int. J. Therm. Sci. 50 (9) (2011) 1741–1747. yond the range of validity of existing correlations, Appl. Therm. Eng. 51 (1–2)
[11] M. Hemmat Esfe, S. Saedodin, O. Mahian, S. Wongwises, Efficiency of ferromagnetic (2013) 1092–1097.
nanoparticles suspended in ethylene glycol for applications in energy devices: ef- [25] J. Sarkar, A critical review on convective heat transfer correlations of nanofluids,
fects of particle size, temperature, and concentration, Int. Commun. Heat Mass Renew. Sustain. Energy Rev. 15 (6) (2011) 3271–3277.
Transfer 58 (2014) 138–146. [26] A.R. Sajadi, M.H. Kazemi, Investigation of turbulent convective heat transfer and
[12] M.J. Pastoriza-Gallego, C. Casanova, J.L. Legido, M.M. Piñeiro, CuO in water nanofluid: pressure drop of TiO2/water nanofluid in circular tube, Int. Commun. Heat Mass
influence of particle size and polydispersity on volumetric behaviour and viscosity, Transfer 38 (10) (2013) 1474–1478.
Fluid Phase Equilib. 300 (1–2) (2011) 188–196. [27] M. Hojjat, S.G. Etemad, R. Bagheri, J. Thibault, Thermal conductivity of non-
[13] A.A. Abbasian Arani, J. Amani, Experimental investigation of diameter effect on heat Newtonian nanofluids: experimental data and modeling using neural network,
transfer performance and pressure drop of TiO2– water nanofluid, Exp. Therm. Fluid Int. J. Heat Mass Transf. 54 (5–6) (2011) 1017–1023.
Sci. 44 (2013) 520–533. [28] M. Hemmat Esfe, S. Saedodin, M. Bahiraei, D. Toghraie, O. Mahian, S. Wongwises,
[14] L.S. Sundar, K.V. Sharma, M.T. Naik, M.K. Singh, Empirical and theoretical correla- Thermal conductivity modeling of MgO/EG nanofluids using experimental data
tions on viscosity of nanofluids: a review, Renew. Sustain. Energy Rev. 25 (2013) and artificial neural network, J. Therm. Anal. Calorim. 118 (1) (2014) 287–294.
670–686.
[15] A. Nasiri, M. Shariaty-Niasar, a.M. Rashidi, R. Khodafarin, Effect of CNT structures on
thermal conductivity and stability of nanofluid, Int. J. Heat Mass Transf. 55 (5–6)
(2012) 1529–1535.