This paper presents a theoretical study of the electrical properties of two different samples of nanofluids (MgO and Si–TiO2 nanoparticles in ethylene glycol EG as the base fluid) using an artificial neural network (ANN) model. Experimental data were extracted from previous experimental studies and used as inputs. A learning ANN method was applied based on the back propagation technique. The optimal network structure, which produces the most acceptable performance, was attained. Electrical conductivity and permittivity in terms of nanoparticle concentration, temperature and frequency were simulated and predicted using the ANN model. A new nonlinear equation describing the behavior of electrical properties of nanofluids was obtained. The simulation results of the ANN model are highly accurate in comparison with experimental data. Predictions of values that are not involved in the experimental data range were carried out and provide excellent results. The mean squared error and regression coefficient were also determined. Their values support the success of the ANN model. The main purpose of the paper is to show the ability and power of an ANN model in estimating and predicting electrical properties of nanofluids. According to this research, the ANN model can be utilized as an efficient tool to predict the electrical properties of nanofluids. It can also achieve great links between practical and theoretical branches.
Read full abstract