This paper aims to three dimensionally investigate flow and heat transfer characteristics of the Al2O3–water nanofluid in channels with discrete heat sources, in which variable thermophysical properties are used in the simulations. The effects of Reynolds number, aspect ratio of the channel, volume fraction and particle size are assessed. The results show that thermal conductivity of the nanofluid varies periodically along the length of the channels due to its dependency on temperature and the amplitude of these variations intensifies by increasing the volume fraction and decreasing size of the particles. As a result, in comparison with pure water, a more uniform temperature distribution is caused along the length of the channel since in the sections with heat source, thermal conductivity of the nanofluid increases, which leads to better cooling as compared to other sections. Due to having this superior characteristic, the nanofluid under study is called a “smart fluid.” It means that the nanofluid automatically possesses better heat transfer characteristics in sections where higher heat dissipation is required. Average convective heat transfer coefficient and pressure drop increase by raising Reynolds number, volume fraction and aspect ratio of the channel. Nevertheless, their relative values, which are calculated in comparison with water, do not vary considerably by changing the aspect ratio of the channel. In addition, a neural network model is developed to model the average convective heat transfer coefficient and pressure drop, which is able to predict the output variables with a great accuracy.
