Content uploaded by Wassila Grine
Author content
All content in this area was uploaded by Wassila Grine on Jul 10, 2022
Content may be subject to copyright.
PREDICTION OF THERMAL CONDUCTIVITY OF ALUMINA WATER-BASED
NANOFLUIDS USING EXPERIMENTAL DATA AND ARTIFICIAL NEURAL NETWORK
(ANN)
Wassila Grine, Mohamed El Hocine BENHAMZA
Laboratory of Industrial analysis and Materials Engineering. University of Guelma,
.
INTRODUCTION DISCUSSIONRESULTS
Figure 2. Actual by predicted plot for TCR models
Figure 3. The TRC predict versus X1(φ), X2(dp) and X3 (T)
ABSTRACT
METHODS AND MATERIALS
CONCLUSIONS
REFERENCES
CONTACT
The nanofluids are
solutions composed of
particles of a nanometric
size suspended in a liquid.
Studies of these composite
fluids show interesting
aptitudes specifically: a
better thermal conductivity
and a coefficient of
convective exchange
significantly increased
compared to traditional
liquids, water in particular.
Nanofluids, containing
nanometric metallic or
oxide particles, exhibit
extraordinarily high thermal
conductivity that can be
used for enhancing heat
transfer performance of
conventional systems. This
work presents a proposed
method for calculating the
effective thermal
conductivity of nanofluids.
The thermal conductivity of
nanofluids primarily
depends on the properties
of base fluids and
nanoparticles, the volume
fraction of nanoparticles,
the interfacial layer, non-
uniform sizes of
nanoparticles, fractal
dimension of particles,
Brownian motion and
temperature.
The aims of the present
study are to develop and
validate an artificial neural
network (ANN) approach to
estimate the thermal
conductivity ratio (TCR) of
alumina water-based
nanofluids as a function of
temperature, volume
fraction and diameter of the
nanoparticle.
Key Words: Nanoparticle,
nanofluid, thermal
conductivity, Artificial Neural
Network, Alumina water-
based nanofluids.
Neural Report
Estimation
Validation of ANN model
Artificial neural network (ANN) model:
The artificial neural network is the non-linear
mathematical models which get great attention due
to its simplicity, flexibility, availability a various
training algorithms as well as its large modeling
capacity, An artificial neural network, which is
derived based on the activity procedure of human
brain, has been employed for modeling of many
scientific disciplines up to now.
In this section, an artificial neural network (ANN)
modeling is performed to predict the thermal
conductivity of Al2O3-water nanofluid as a function
of temperature, volume fraction and diameter of
nanoparticles.
In the present work a systematic procedure
based on artificial neural network is developed
to predict the thermal conductivity of Alumina-
water nanofluid as a function of temperature,
volume fraction and diameter of nanoparticles
(φ,Tand dp) . The excellent correlation
between predicted and observed TCR
response, high and significant (R² = 0,72 and
RSME= 0.038 for training and R² = 0,60,
RSME= 0.045 for validation) give a good
accordance between the model and
experimental data. Additionally, the selected
ANN approach can predict the training with
AARD% of 0.553. These small value of the
proposed model confirm its excellent
performance in modeling and simulation of TCR
of considered nanofluid. Comparison with the
recommended correlations confirms that the
proposed model has superior performance in
prediction of TCR of alumina water-based
nanofluids and can be considered as a practical
tool for the considered task.
Over the last several decades, many
researchers have been trying to create
new kinds of heat transfer fluid in order
to increase the heat thermal
performance of the common fluids, such
as water, ethylene glycol or engine oil.
One of the recent techniques to improve
heat transfer is suspending small
amount of some species of metallic or
non-metallic nanoparticles with high
intrinsic thermal conductivity in these
common fluids.
These exclusive features of
nanoparticles have been motivated
many researchers to investigate the
thermal/physical performance of various
nanoparticles in different base fluids
regarding variety of variables affecting
thermal properties of the nanofluids.
The thermal conductivity enhancement
is one of the noteworthy effects
originating from the suspension of
nanoparticles into a base fluid. The
thermal conductivity enhancement ratio
is defined as the ratio of thermal
conductivity of the nanofluid to that of
the base fluid (knf/kf).
The objective of this paper is to develop
an ANN model to predict the thermal
conductivity of Alumina–water
nanofluids, based on own experimental
data and then validated against data
coming also from other authors to
exploit the consistency of the modelling
approach.
1. Ravi Sankar.B, Nageswara Rao. D,Srinivasa
Rao.Ch, Nanofluid Thermal Conductivity-A
Review, International Journal of Advances in
Engineering & Technology, (Nov. 2012) 2231-
1963.
2. Xiang-Qi Wang, Arun S. Mujumdar, Heat
transfer characteristics of nanofluids: a review,
International Journal of Thermal Sciences 46
(2007) 1–19
3. J.A. Eastman, S.U.S. Choi, S. Li, W. Yu, L.J.
Thompson, Applied Physics Letters 78 (2001)
718–720.
<Wassila Grine>
Affiliation:Laboratory of
Industrial Analysis and Material
Engineering (L.A.I.G.M)
University: University of 8 may
1945 GUELMA; ALGERIA
Email: grinewassila@yahoo.fr
Phone: +213663748506
Journée Nationale
« Etude des Matériaux : Elaboration et Modélisation »
Dec.13, 2018, Guelma, Algeria
Training Validation
Knf/KfMeasures Knf/KfMeasures
R20.725 R20.60
RMSE 0.0387 RMSE 0.045
Mean Abs Dev 0.0327 Mean Abs Dev 0.0354
-Log likelihood 208.69 -Log likelihood 96.80
SSE 0.171 SSE 0.1205
Sum Freq 114 Sum Freq 58
Training Validation
Figure 4. Plot of Residuals by Predicted for Training and
Validation
Figure 5. Surface Profiler
Figure 6. Contour Profiler
Validation
Method
The
Training
set
The
Validation
set
Is the part that
estimates
model
parameters
Is the part that
estimates the
optimal value
of the penalty,
and validates
the predictive
ability of the
model
In this research, a network structure with
three hidden layer is used to predict thermal
conductivity of the nanofluid. The number of
neurons in the hidden layer has been
determined through minimizing RMSE and
AARD% and maximizing R2values of both
test and training data sets.
Figure 7. Experimental and predicted values of
TCRs versus volume fraction of Al2O3
nanoparticles (dp= 38.4 nm, T = 51 °C).
Figure 4. Structure of the neural network
used in this study.