Experimental investigation of rheological behavior of water/graphite nanofluid and presenting a new empirical relation and predicting data using artificial neural network

Abstract

The present study examines the changes in the viscosity of graphite nanoparticles (GNPs) with distinct insights. The temperature range of the nanofluid (NF) from 20 to 50 °C and the volume fraction (VF) of GNP, as compared with the fluid, varying from 0.1 to 0.5%, are the two key variables to achieve optimal NF characteristics. The experimental results confirm that the viscosity of the NF is inversely proportional to the temperature changes, whereas GNP VF follows a direct trend upon changes in viscosity. The intermolecular attraction forces decreased with an increase in temperature, helping to dilute the fluid. At high concentrations, the inhibitory role is lost, and the nanoparticles themselves contribute to the increase in yield stress by coagulation. The optimal condition corresponds to 0.5% VF of GNP at 20 °C. Under such conditions, a 22.6% increase in viscosity was achieved compared to the water-based fluid. Utilizing varying NF temperature and GNP VF (as input), a set of experimental viscosity data (as the target function) was primarily collected. A curve-fitting technique was then employed to develop a theoretical correlation based on the input and target functions. The obtained correlation coefficient (R), i.e., 0.99753, indicates a strong correlation between the experimental data and the proposed relation. In the present study, the perceptron neural network, the Purelin, and the tangent sigmoid functions were used. The algorithm considered was the Levenberg–Marquardt (LM) algorithm, and 32 neurons were used for both optimization and prediction. The root mean squared error (RMSE), mean squared error (MSE), correlation coefficient (R), and mean absolute error (MAE) for the proposed relation, as well as artificial neural network (ANN) data, are shown. These values for the proposed relation data are reported as follows: 0.00782, 0.00000342, 0.997, and 6.25 × 10−7. Accordingly, the ANN is: 8.74 × 10−3), 7.6 × 10−5, 0.998, and 6.939 × 10−18, respectively. The margin of deviation (MOD) was calculated as -2.1831 < MOD < 3.1420.