Locally weighted moving regression: A non-parametric method for modeling nanofluid features of dynamic viscosity

Abstract

The objective of this study is to evaluate the variations of CuO/ ethanol nanofluid dynamic viscosity with temperature and shear rate experimentally. In this study, the nanoparticles average size, their mass fraction, temperature, and shear rate are varying from 10 to 50 nm, 0.005 to 5 wt. %, 25 to 70°C, and 2.6 to 64.6 s−1, respectively. The obtained results showed that as the nanoparticles mass fraction increases and the temperature decreases, the nanofluid dynamic viscosity enhances. Then to predict the nanofluid dynamic viscosity as a function of temperature and nanoparticles mass fraction, the locally weighted moving regression (LWMR) method is applied to the data alongside with applying a classic cubic interpolation for comparison. It was found that although both methods have acceptable performance in dealing with the trained data, the LWMR model results in a more precise prediction. In addition, the numerical viscosities obtained by both methods were evaluated at the trained dataset. Finally, it was concluded that the LWMR may be a suitable model for the investigation of nanofluid features.