Optimization of turbulent convective heat transfer of CuO/water nanofluid in a square duct

Abstract

Analysis of the turbulent forced convective heat transfer of CuO/water nanofluid has been conducted employing an artificial neural network. A backpropagation network with two hidden layers has been used to investigate and optimize the effects of the Reynolds number, volume fraction and nanoparticle diameter on heat transfer, pressure drop and entropy generation. In order to train the network, the Levenberg–Marquardt algorithm has been applied. The results of a total of 168 sample models, considering the extended range of input data, have been prepared as the training data for the network. The precision of the network based on the mean squared error for a set of random experimental data has been approximately 0.05. The obtained results of the artificial neural network are presented in an extended range of Re, $$\varphi$$ and particle diameter. Results indicate that heat transfer enhancement is achieved by increasing Re and volume fraction. While augmenting Re reduces the friction factor, it is increased at higher volume fractions. The optimum conditions concerning Nusselt number, friction factor and entropy generation have been investigated. For instance, at low volume fractions, the ratio $$Nu/f^{1/3} S^{2}$$ has the highest value in the range $$5 \times 10^{4} < Re < 6 \times 10^{4}$$ , which is a proper range of performance for the nanofluid.