Performing regression-based methods on viscosity of nano-enhanced PCM - Using ANN and RSM
Nidal H. Abu‐Hamdeh, Ali Golmohammadzadeh, Aliakbar Karimipour
Abstract
Evaluation of the use of linear and nonlinear regression-based methods in estimating the viscosity of MWCNT/liquid paraffin nanofluid was investigated in this study. At temperature range of 5–65 °C, the viscosity of samples containing MWCNT nanoparticles at 0.005–5 wt.% which is measured by a Brookfield apparatus, was first evaluated to determine the response to the shear rate. The decrease in viscosity due to the increase in shear rate indicated that the rheological behavior of the nanofluid was non-Newtonian and therefore, in addition to temperature and mass fraction, the shear rate should be considered as an effective input parameter. Linear regression was performed by response surface methodology (RSM) and it was observed that the R-square for the best polynomial was 0.988. The results of nonlinear regression also showed that the neural network consisting of 3 and 13 neurons in the input and hidden layers was able to estimate the viscosity of the nanofluid more accurately so that the R-square value was calculated to be 0.998.