Innovative simulation of Al2O3 nanofluid heat transfer using advanced machine learning methods

Abstract

In both turbulent and laminar pipe flows, we were able to accurately forecast the beginning range of the convective thermal transferring coefficients of Al2O3 magnetized nanofluids using machine learning approaches. The simulations utilized two machine learning techniques: radial basis function-backpropagation (RB) and multiple linear regression analysis. First, we used multiple linear regression analysis to fit the polynomial equation. Afterwards, grid search cross-validation was employed to determine the optimal RB model with six hidden layer neurons. To evaluate the RB model, we compared numerical patterns of the parameters used to measure accuracy. The regression coefficient and mean square error were the most commonly utilized parameters in Reynolds number mass percentage simulations, R2. In the case of a laminar flow, these numbers were found to be 0.99994 and 0.34, respectively. Additionally, the results for laminar flow conditions using Reynolds number-magnetic field strength simplification were ideal, with an mean square error of 3.85 and an R2 value of 0.99993. By comparing the predicted values with the experimental results visually using 3-D smoothed surface plots, we were able to further prove that the model was valid and accurate. These revolutionary findings could spark new developments and encourage substantial improvements in nanotechnology and machine intelligence. These findings are an important asset for driving future research and development, which in turn makes significant contributions to the ever-expanding frontiers of these innovative fields.