Developing dissimilar artificial neural networks (ANNs) to prediction the thermal conductivity of MWCNT-TiO2/Water-ethylene glycol hybrid nanofluid

Abstract

In this paper, we developed dissimilar artificial neural networks (ANNs) by suitable architectures and training algorithms via sensitivity analysis to predict the thermal conductivity MWCNT -TiO2/ Water-Ethylene glycol nanofluid. Forecasting of thermal conductivity of MWCNT –TiO2/ Water-Ethylene glycol nanofluid based on changes in temperature and concentration using ANN and stability analysis is done. MWCNTs-TiO2 hybrid nanoparticles were also used at a 50:50 volume ratio. The dataset of ANN was divided into three main parts including 70% for the train, 15% for test and 15% for validation and the results of the optimum ANN are in a better agreement to the empirical dataset, and it can predict the thermal conductivity of MWCNT-TiO2-Wa-EG(50–50) better than the correlation. The empirical dataset, ANN outputs, and correlation results were presented. There is a small difference between correlation results and ANN outputs, and it can be concluded that ANN outputs are can predict the empirical results better than the correlation formula.