Content uploaded by Wassila Grine
Author content
All content in this area was uploaded by Wassila Grine on Jun 08, 2023
Content may be subject to copyright.
1st International Seminar on Process Engineering & Environment (ISCPE2022) December 06-08-2022
Using perceptron feed-forward Artificial Neural Network (ANN) for predicting the
thermal conductivity of (Al2O3/Water) Nanofluid
GRINE Wassila1,*, SAHRAOUI Abderrahmane2, BENHAMZA M E Hocine 1
1 Laboratory of Industrial Analysis and Materials Engineering, University 8 May 1945, Guelma, 24000, Algeria,
2 Laboratoire de Génie Mécanique (LGM), Université de Mohamed KHIDER-BP 145, 07000 Biskra, Algérie
(*grinewassila@yahoo.fr / 0663748506)
Abstract
Nanofluids are nowadays the most widely used working heat transfer fluids. Therefore, a more accurate assessment
of their thermophysical properties, as well as their performance, is required. Thermal conductivity is the most
influential thermophysical property in the application of nanofluids. The increase in nanofluids' thermal conductivity
cannot be accredited only to a better and excellent thermal conductivity of nanoparticles in suspension. Still, it also
comes simultaneously from several physical factors of varying importance. Additionally, nanoscale thermal
behavior does not follow models applied to larger structures, thus further research is needed to design appropriate
models.
In this paper, an Artificial Neural Network (ANN) model for predicting the thermal conductivity of (Al2O3/Water)
nanofluid was developed. The model accounts for the effect of temperature, nanoparticle volume fraction,
nanoparticle diameter and nanoparticle shapes. Feed forward ANN has been used to predict the effective thermal
conductivity of nanofluid. The network was trained, tested, and validated using a total of 105 experimental data
points. The results show that the best architecture obtained in hidden layers for the Thermal Conductivity Ratio
(TCR) is 15 neurons. TCR model provides an excellent correlation between predicted and experimental values, with
coefficient of determination (R2) values superior to 0.99 for both learning and validation and insignificant Mean
Square Error (MSE) values (equal to 0.000018). Moreover, the selected ANN approach provides learning with an
Absolute Average Relative Deviation (AARD) of 0.013 %, confirming the validity of the adopted method. The
comparison with numerous empirical correlations also confirms that the model proposed in this study predicts the
TCR of Alumina/Water nanofluids with better performance and therefore can be considered a practical tool for the
considered tasks.
Keywords: Nanofluids, Nanoparticle, Metal Oxides, Thermal Conductivity, ANN.