Synthesis of ternary nanofluids and optimization of their thermophysical properties using artificial neural network

Abstract

This work focuses primarily on the two-step synthesis of ternary nanofluids consisting of silver (Ag), graphene oxide (GO), and multi-walled carbon nanotubes (MWCNT) in volume fractions ranging from 0.005 to 0.03, their stability and structural (morphological) analysis, and appraisal of their thermophysical properties such as thermal conductivity, viscosity, density and specific heat capacity in the temperature range from 20 to 80°C. The thermal conductivity and viscosity were found to be 0.7845 W/m.K and 0.8718 cP at 30°C for 3 vol %. The work also involved the optimization and validation of these thermophysical parameters using Artificial Neural Network (ANN). The ANN was constituted and tested with the application of Levenberg-Marquardt (LM) algorithm. The titular network size has been inherently optimized as per the relative error to enhance the model’s ability to predict thermal conductivity and viscosity. The Levenberg-Marquardt feed-forward network possessing the optimal network design, 6 – 1 (hidden layer nodes – Output layer nodes) for thermal conductivity and 2 – 4 for viscosity have been identified as the best training approach. The ANN results exhibit the coefficient of regression (R²) to be significant at 0.99736 and 0.99725 for thermal conductivity and viscosity respectively, and the upper limit of relative error was negligible. The same data set subjected to the standard fitting model gave the R² of 0.9931 and 0.9944 with mean square error of 0.0064 and 0.0170, respectively.