Modeling and optimization of thermal conductivity ratio of Al2O3–water nanofluid using artificial neural network and Box-Behnken design based response surface methodology with canonical analysis

Superior thermal characteristics, including increased thermal conductivity, enhanced convective performance, and improved thermal stability, make nanofluids attractive substitutes for enhancing the effectiveness of heat transfer. It is therefore possible to circumvent the thermo-physical constraints of regular fluids by scattering appropriate nanoparticles. This study predicts and optimizes the thermal conductivity ratio of water-aluminum oxide nanofluids using statistical response surface methodology (RSM) and artificial neural networks (ANN). A Box-Behnken design (BBD) within the RSM framework was employed to explore the relationship between independent variables, such as nanoparticle concentration (1–4 %), temperature (293-323 K), and surfactant weight (776-3104 mg), and the response function thermal conductivity ratio. Canonical analysis was also conducted to identify significant interactions among variables. For ANN, the Levenberg-Marquardt (LM) algorithm is employed to optimize the network's performance with six neurons in the hidden layer. To create second-order polynomial equations for predictive modeling, a total of 17 experiments were conducted. The accuracy of the predictive performance of RSM and ANN was evaluated using the margin of deviation (MOD), mean squared error (MSE), root mean squared error (RMSE), and coefficient of determination (R²). The optimal ANN configuration exhibited a high R² (0.9945) and a low MSE error (0.0030) as compared to the RSM model. Moreover, the average error for the ANN was 1.8192 %, which is significantly less than the 3.9773 % error of RSM. Both methods were successful in forecasting the thermal conductivity ratio of aluminum oxide–water nanofluids, although the ANN method was more accurate. According to these results, ANN is a practical and effective tool for evaluating and improving heat transfer systems based on nanofluids in industrial applications.