This study aims to develop a hybrid computational framework combining high-fidelity numerical modeling with a Levenberg–Marquardt–trained artificial neural network (LM-ANN) to analyze magnetohydrodynamic boundary-layer transport of a Cu–TiO2–Al2O3/ethylene glycol trihybrid nanofluid. The objective is to capture the effects of magnetic field, porous resistance, radiation, Soret diffusion, viscous dissipation and buoyancy while offering a fast surrogate tool for predictive thermal-fluid analysis.
Similarity transformations are applied to reformulate the governing equations, which are solved numerically using BVP4c. The resulting data set trains the LM-ANN, achieving near-perfect predictive accuracy (MSE = 4.97× 10−9, R =1). Comparative analyses between numerical and ANN outputs confirm consistency, while parametric studies examine how nanoparticles (NPs), magnetic field and porosity influence velocity, temperature and concentration fields.
Results indicate that increasing NPs concentration significantly improves heat transfer, whereas stronger magnetic and porous effects suppress fluid motion. Radiation and Soret diffusion enhance thermal and mass transport, while viscous dissipation slightly diminishes cooling performance. The LM-ANN provides rapid and accurate predictions, validating the reliability of the dual-model strategy.
This work introduces a novel dual-model approach that unifies numerical simulations with ANN surrogates for complex MHD trihybrid nanofluid transport. Unlike conventional single-method studies, the framework offers both physical insight and computational efficiency, enabling predictive modeling and optimization of nanofluid-based porous media flows in advanced thermal management and energy systems.
