Coupling magnetic braking, exothermic chemistry and latent storage in a porous-finned annulus: a CFD–machine-learning study of NEPCM–water transport

Latent-heat thermal-energy-storage exchangers underpin compact solar-thermal receivers, waste-heat recovery loops, and building-integrated HVAC modules, yet the coupled interaction among magnetic braking, exothermic chemistry, porous-media damping, and NEPCM latent storage in a finned annular exchanger remains largely unaddressed. The present study closes this gap by delivering a coupled numerical and data-driven analysis of irreversibility production and thermal transport in a double-pipe exchanger charged with a NEPCM–water slurry that is simultaneously exposed to magnetic actuation, heat-releasing chemistry, and a permeable matrix. A Galerkin finite-element discretisation resolves the coupled transport system, and the solver is benchmarked against two established literature datasets. Broad parametric windows are traversed: Rayleigh (10³–10⁵), Darcy (10⁻¹–10⁻⁶), Frank-Kamenetskii (0–1.5), Lewis (0.1–10), buoyancy ratio (1–7), Hartmann (0–85), and capsule loading (0.01–0.035). Three learners—XGBoost, Random Forest, and a multilayer perceptron are trained on 1579 CFD-derived samples to reproduce Nu, Sh, S , and Be from the nine control groups. Enriching the capsule fraction from 0.01 to 0.035 lifts thermal transport by 55.8% at Ra = 10³ , whereas magnetic braking curtails the Nusselt gain from 79.0% down to 4.1% once Ha attains 75. Thermal and species transfer respond in opposite senses to the reactive term: pushing FK to 1.5 depresses Nu by 15.2% while lifting Sh by 9%. The boosted-tree learner emerges as the most faithful surrogate (test R² ≥ 0.9995 ), while Ra dominates the ranking for Nu, S, and Be, and Le governs Sh . The findings offer quantitative guidance for tuning latent-storage units through balanced capsule enrichment and magnetically regulated flow control.