Measurement of the dynamic viscosity of water-based nanofluids containing Al2O3, TiO2, and ZnO using the Artificial Neural Network method

Nanofluids are strong candidates as heat carriers due to their excellent thermophysical properties. Among these thermophysical properties, viscosity is critical in heat transfer and pressure loss calculations. In this study, three different water-based nanofluids, Al2O3, TiO2, and ZnO, were prepared with volumetric concentrations ranging from 0.1% to 1%. The dynamic viscosities of these nanofluids were experimentally measured within a temperature range of 20 °C to 50 °C. Artificial neural networks (ANN) were employed to predict the results based on the experimental data. Two different approaches were applied in the implementation of the ANN method. The first approach involved creating three separate ANN models, each dedicated to predicting the viscosities of the three different nanofluids. The second approach used a single generalized ANN to predict the viscosities of all nanofluids. The results were evaluated using the criteria of R-squared (R2) and root mean square error (RMSE) values. In all models, R2 values exceeded 99%, while the RMSE values were calculated for the Al2O3, TiO2, and ZnO nanofluid ANN models and the generalized ANN model to be 0.40%, 0.30%, 0.04%, and 0.28% respectively. These results demonstrate that a nanofluid's viscosity can be effectively predicted individually and multiple nanofluids using an ANN model.