Experimental study for developing an accurate model to predict viscosity of CuO–ethylene glycol nanofluid using genetic algorithm based neural network

In this paper, the viscosity of CuO–ethylene glycol nanofluid is measured at different concentrations and temperatures. The nanofluid is prepared via the two-step method by employing the shear homogenization and ultrasonication. The viscosity is measured at concentrations of 0, 0.125, 0.25, 0.5, 0.75, 1, and 1.5%, as well as temperatures of 27.5, 30, 35, 40, 45, and 50 °C. The model of viscosity for this nanofluid is developed using genetic algorithm based neural network with the aid of the data obtained from the experiments. The results show that the viscosity increases with the volume concentration increment, while reduces by increasing the temperature. Moreover, the temperature has a more significant effect at the lower concentrations. By investigating the performance of different neural network configurations, a neural network with two hidden layers and 8 neurons in each layer was selected as the best ANN, which predicts the viscosity with a very good accuracy. For the test data set, this model estimates the viscosity with R2 value of about 0.999 and mean relative error of about 0.0175. In addition, the sensitivity analysis reveals that the volume concentration is the more effective factor on the viscosity in comparison with the temperature. [Display omitted] •Viscosity of CuO nanofluid is measured at various concentrations and temperatures.•Nanofluid is prepared by employing shear homogenization and ultrasonication.•Model of viscosity is developed using genetic algorithm based neural network.•Viscosity rises by concentration increment, while reduces by temperature increase.•The concentration is more effective factor in comparison with the temperature.