A novel hybrid nanofluid including MWCNT and ZrO₂ nanoparticles: implementation of response surface methodology and artificial neural network

Abstract

In this study, the thermal behavior of a hybrid nanofluid including nanoparticles of multi-walled carbon nanotube ($\mathrm{MWCNT}$ ) and ${\mathrm{ZrO}}_{2}$ was discussed. The nanoparticles of $\mathrm{MWCNT}$ and ${\mathrm{ZrO}}_{2}$ at the mass fraction 40:60 were added to pure water with CMC surfactant to prepare samples at 0.009, 0.018, 0.036, 0.072 and 0.14 vol%. The thermal conductivity of $\mathrm{MWCNT}-{\mathrm{ZrO}}_{2}/\mathrm{water}$ $\left( {{\text{k}}_{{{\text{ZrO}}_{2} - {\text{MWCNT}}/{\text{water}}}} } \right)$ was measured at temperature range of 20–50 °C and compared with ${\mathrm{k}}_{\mathrm{water}}$ to investigate the effectiveness of adding $\mathrm{MWCNT}-{\mathrm{ZrO}}_{2}$ nanoparticles to the base fluid by evaluating the parameter of thermal conductivity ratio $\left( {{\text{TCR = }}\frac{{{\text{k}}_{{{\text{ZrO}}_{2} - {\text{MWCNT}}/{\text{water}}}} }}{{{\text{k}}_{{{\text{water}}}} }}} \right)$. The results showed that an increase in temperature results in a higher $\mathrm{TCR}$. The use of $\mathrm{MWCNT}-{\mathrm{ZrO}}_{2}$ led to an increase in thermal conductivity up to 15.4%. To boost the amount of improvement, the sonication time increased from 30 to 60 min. The best sonication time was 50 min so that $\mathrm{MWCNT}$ and ${\mathrm{ZrO}}_{2}$ nanoparticles could enhance ${\mathrm{k}}_{\mathrm{water}}$ by 59.6%. To estimate TCR, the response surface methodology was used, which based on linear regression, provides a function consisting of independent variables, and it was found for the best function with an error of less than 1.52% and by providing R² = 0.998, can estimate TCR very well. Then artificial intelligence was used, and it was determined that the neural network consisting of eight neurons with a maximum error of less than 0.4% could predict TCR.