An accurate RBF-NN model for estimation of viscosity of nanofluids

Abstract

Viscosity is one of the most important properties of nanofluids which play an important role in many applications of these fluids. There is no general model or equation for reliable prediction of viscosity of nanofluids in various thermodynamic conditions. In this work an intelligent model named radial basis function neural network (RBF-NN) is proposed to estimate the viscosity of nanofluids with different nanoparticle types and base fluid types. The model was constructed based on 1490 experimental data gathered from literature. The performance of model and its accuracy was investigated by utilizing various graphical and statistical approaches. The predictions of RBF-NN model were also compared with a literature model and several correlations. Results showed that the model provides good degree of accuracy. The overall R2, AARD% and RMSE of developed model are 0.99996, 0.2 and 0.0089 respectively. In addition, the developed model successfully outperforms literature model and correlations for viscosity prediction of nanofluids and present more accurate and reliable results. The sensitivity analysis of predictions of developed model also shows that the volume fraction of nanoparticle has the greatest impact on viscosity of nanofluid and temperature has the lowest impact.