Analysis and prediction of thermo-physical properties in water-based MWCNT-ZnO hybrid nanofluids using ANN and ANFIS models

In this study, the thermal conductivity and viscosity of multiwalled-carbon nanotubes/zinc oxide water hybrid nanofluid across volume concentrations varying from 0.2% to 0.8% and temperatures from 25°C to 65°C were experimentally studied. Three mathematical models such as multivariable regression, artificial neural network, and adaptive neuro-fuzzy modeling were employed for the prediction of the thermal conductivity of the water baes multiwalled-carbon nanotubes/zinc oxide hybrid nanofluid. Volume concentration and temperature of the nanofluid are the input parameters for the models. Despite the complexity of the input data, which encompassed extensive ranges of temperature and volume concentration, adaptive neuro-fuzzy modeling exhibited superior predictive performance than the other two models. It achieved conductivity values closely aligned with experimental results, characterized by the lowest mean square error compared to regression and artificial neural network models. Notably, the adaptive neuro-fuzzy modeling method facilitated the resolution of the neural network layer's hidden structure without the need for extensive trial and error.