Neural network-driven analysis of MHD Eyring–Powell nanofluid dynamics with heat and mass flux theory: a machine learning strategy

Abstract

This investigation examines the three-dimensional magnetohydrodynamic flow of an Eyring–Powell nanofluid with reactive transport, modeled using the Cattaneo–Christov heat and mass flux theory. The aim is to investigate the combined effects of the magnetic field, rotation, and reactive parameters on flow, heat, and mass transfer, and to validate the predictive capability of artificial neural networks experimentally. The governing equations are solved numerically using MATLAB’s bvp4c solver, and the resulting data are used to train artificial neural networks based on the Levenberg–Marquardt algorithm. It is observed that increasing the magnetic parameter decreases velocity by about 12–18 %, whereas increasing the rotation factor increases velocity by 9–14 % owing to Coriolis forces. Temperature drops by 10–13 % with increased heat generation, while concentration gradients decrease by 8–11 % with an increase in the Schmidt number. The ANN model achieves the lowest MSE of 6.81 × 10 ⁻⁹ in close agreement with numerical predictions. Model validation against benchmark cases confirms the model’s accuracy. The results provide new insights into advanced machine learning integrated with reactive nanofluid modeling for industrial, environmental, and biomedical thermal systems.