Neural network prediction of MHD thermo-solutal natural convection in ternary hybrid nanofluids within a curved enclosure

Abstract

In an era when optimizing energy efficiency and ensuring effective thermal management are essential, this study addresses the complex dynamics of MHD thermo-solutal natural convection (TSNC) within a curved cavity containing a unique ternary hybrid nanofluid (THNF) (Cu + CuO + Al₂O₃/H₂O). We explore the effects of key parameters, including Rayleigh number (Ra: 10⁴–10⁷), Hartman number (Ha: 0–100), volume fraction of nanoparticles (ϕ: 0–0.04), Lewis number (Le: 1–10), heat source length (H: 0.2–0.6), and heat source location (R: 0.25–0.75), on convective flow patterns, heat transfer, and mass movement. An artificial neural network (ANN) model is developed to investigate the influences of these parameters on the thermal and solute transport properties of THNF. The ANN model is trained using high-fidelity numerical simulation data to predict the thermosolutal convection behavior of THNF. This methodology accelerates predictive analysis but requires significant computational effort for training. Our precise representation of streamlines, isotherm lines and concentration profiles reveals that complex relationships between these variables. The results demonstrate that higher the $\mathrm{Ra}$ values improve heat and mass transfer, whereas increasing Ha suppresses convective motion due to electromagnetic damping from the Lorentz force. Increased $\varphi$ enhances heat and mass tranfer, leading to stronger convection. $\mathrm{Le}$ analysis reveals that elevated values promote mass transfer over heat transfer. Although variation in heat source parameters affects flow structure and thermal gradient differently, the overall convective behavior depends on their interplay. Our comparison research show that the ternary hybrid nanofluid ($\mathrm{Cu}+CuO+{\mathrm{Al}}_2{O}_3$) exhibits a higher average Nusselt number compared to the base fluid, the ternary hybrid nanofluid shows an increase in heat transfer efficiency of up to 78.33 % under specific conditions. These findings emphasize the potential of THNF and the modification of operational parameter, giving important insights for improving heat and mass transfer technologies across diverse industries.