Modelling and optimisation of thermophysical properties and convective heat transfer of nanofluids by using artificial intelligence methods

Abstract

Nanofluids are modern heat transfer fluids which can significantly increase the thermal performance of a thermal system. It enhances the thermal conductivity of working fluids due to adding solid nanoparticles to the base fluid. In order to use nanofluids widely in industrial applications knowing the thermophysical properties of these new heat transfer fluids are essential. In this research, the GA-PNN and FCMANFIS methods are employed to present models for thermophysical properties of nanofluids. Furthermore, modified NSGA-II technique has been used to optimise the convective heat transfer of nanofluids in a turbulent flow regime. In recent years considerable correlations have been suggested by different researchers for thermophysical properties of nanofluids based on the experimental and theoretical works, which a large number of those correlations are failed to predict the thermophysical properties of nanofluids for a wide range of particle size, temperature and nanoparticle volume concentrations. In this thesis, experimental data available in literature have been used to propose models for thermophysical properties of nanofluids to overcome this problem by using artificial intelligencebased techniques. Two models based on FCM-ANFIS and GA-PNN techniques have been proposed for the thermal conductivity and viscosity of nanofluids. To show the accuracy of the proposed models, the predicted result has been compared with experimental data as well as well-cited correlations in literature. Furthermore, the convective heat transfer of nanofluids was studied and different models based on artificial intelligence techniques have been proposed to model the Nusselt number and pressure drop of nanofluids in a turbulent regime. Finally, a multi-objective optimisation technique was used to optimise the convective heat transfer characteristics and pressure drop of nanofluids to find the best design point base on the Pareto front of the results. The predictions of the models for all cases agreed with the experimental data much better than the available correlations.