Optimization of convective heat transfer and thermal storage in ternary hybrid nanomaterials using machine learning-driven exogenous neural networks with radiation effects

Abstract

Ternary hybrid nanomaterials (THNFs) have emerged as effective enhancers of the physical properties of base fluids, significantly improving thermal efficiency. A THNF (CoFe2O4, Al2O3, Cu/H2O) is formulated by dispersing three distinct nanoparticles into a base liquid, leading to advanced features such as enhanced flow control, electrical conductivity, and magnetic properties. This study utilizes Artificial Neural Networks (ANNs) to analyze the 3-D magnetized stretched flow of THNF within a porous medium. The Darcy-Forchheimer model is employed to investigate the complex movement of THNF through the porous space. A convective boundary condition is applied, with water as the base fluid and a combination of copper, polyphenol-coated, and aluminum oxide nanoparticles as the dispersed phase. In order to transform PDEs into the intended ODEs, similarity variables are used. The Wolfram Mathematica program is used to implement the ND-Solve technique as a numerical solver tool. Using a stochastic artificially intelligent neural network in MATLAB, the findings are validated and cross-checked until they further converge to the Levenberg-Marquardt backpropagated model. Full samples (100 %) are split into training data (70 %), and testing and validation data (15 % each) before BPLMS is applied. Ten hidden neural units are used to continually train the data in order to get the mean squared error. The statistical tabular record contains regression expressions, fitness state functions, error histograms, state transition functions, and performance analysis. The graphical results are obtained from matrix data for all three functions, dimensionless primary velocity profile $f^{\prime }\left(\eta \right)$, dimensionless temperature profile $\theta \left(\eta \right)$ and dimensionless secondary velocity profile $g^{\prime }\left(\eta \right)$ along with Skin friction coefficients and Nusselt number by using ANNs on THNF CoFe2O4, Al2O3, Cu/H2O with variation of different parameters.