Artificial neural network modeling of thermal characteristics of WO3-CuO (50:50)/water hybrid nanofluid with a back-propagation algorithm

Abstract

Thermophysical properties such as thermal conductivity (knf) make the use of fluid suitable for heat transfer. Fluids such as water have limited applications due to their low thermal conductivity. One of the new methods to improve the properties of fluids is to add nanoparticles with high thermal conductivity and create a nanofluid. Nanofluids combine the suspension of two or more nanoparticles in a base fluid or the suspension of hybrid nanoparticles in a base fluid. This study investigates the thermal behavior of WO3-CuO (50:50)/water nanofluid using an artificial neural network (ANN) and back-propagation algorithm. The results show that increasing the volume fraction of nanoparticles (φ) (due to increasing the surface-to-volume ratio) increases the knf. In this study, ANN modeling for WO3-CuO/water (50:50) hybrid nanofluid was performed to investigate the effect of nanofluid on knf. These two important parameters are φ and temperature. The results show that increasing the φ increases the knf due to increasing the surface-to-volume ratio and the collision between nanoparticles. Increasing the temperature shows a similar effect and improves the knf by increasing the interaction between the nanoparticles. The effect of temperature on the knf is more significant than the φ, equal to 16.33% and 6.72%, respectively. Function parameters such as correlation and error value for hidden layer 7 and 12 neurons are about 0.982, 0.981, and 10−6, respectively. As a result, ANN models offer acceptable performance in estimating knf, and the correlation coefficients and error values are 0.96 and 10−6, respectively. Given the absolute error value, it can be concluded that the proposed models can predict the knf of WO3-CuO (50:50)/water hybrid nanofluid.