Machine learning-assisted analysis of magnetohydrodynamic double-diffusive convection and exothermic reaction in nano-encapsulated phase change materials-water filled arched-top cavity for evacuated tube solar collector manifold

Integrating latent thermal energy storage with evacuated tube solar collectors remains a major challenge because the time-mismatch between solar input and demand limits practical deployment, while the influence of realistic manifold geometries on storage performance has rarely been quantified. This study aims to characterize the coupled heat- and mass-transfer behavior of a nano-encapsulated phase change material suspended in water inside an arched-top cavity that emulates the cross-section of an evacuated tube solar manifold, under the joint action of an inclined magnetic field and an Arrhenius-type exothermic reaction. The governing dimensionless equations were solved using a validated Galerkin finite-element procedure. Nine controlling parameters were varied: the Rayleigh number, Hartmann number, magnetic-field inclination angle, Frank–Kamenetskii parameter, Lewis number, buoyancy ratio, Stefan number, fusion temperature, and nanoparticle volume fraction. The findings show that the Rayleigh number boosts the average Nusselt number by 196%, while raising the nanoparticle volume fraction simultaneously enhances heat transfer by up to 202% and reduces entropy generation by 13%. A flow bifurcation was identified at moderate field intensity and 60° tilt, raising heat transfer by 19%. Three machine-learning surrogates trained on 10,993 simulations achieved accuracies above 97%, offering a fast and interpretable design aid for solar-manifold engineering.