Magnetic field influence on heat transfer of NEPCM in a porous triangular cavity with a cold fin and partial heat sources: AI analysis combined with ISPH method

Abstract

This study employs the Incompressible Smoothed Particle Hydrodynamics (ISPH) method and an Artificial Neural Network (ANN) model to examine the thermal and fluid dynamics behavior of nano-enhanced phase change material (NEPCM) within a triangular cavity containing a fin. The research investigates how varying physical parameters optimize heat transfer efficiency. The analysis spans partial heat source length $\left(L_B:0.2\text{to}0.9\right)$, Darcy number $\left(\mathrm{Da}:{10}^{-2}\text{to}{10}^{-5}\right)$, Hartmann number $\left(\mathrm{Ha}:0\text{to}50\right)$, Cattaneo-Christov heat fluxes $\left({\delta }_{\mathrm{Ht}}:0\text{to}0.1\right)$, fusion temperature $\left({\theta }_f:0.25\text{to}0.95\right)$, and nanoparticle concentration $\left(\varphi :0\text{to}0.06\right)$. Key findings demonstrate that increasing $L_B$ by 350 % enhances temperature distribution and nanofluid velocities, reducing the heat capacity ratio $\left(\mathrm{Cr}\right)$ by approximately 20 %. The addition of cooling fins decreases peak temperatures by around 15 %. Higher Darcy numbers improve circulation and convection by up to 30 %, creating more uniform thermal distributions, whereas lower $\mathrm{Da}$ values restrict fluid motion, intensifying temperature gradients. Increasing the Hartmann number reduces flow and heat transfer efficiency by 40 %, causing sharper temperature gradients, while lower $\mathrm{Ha}$ values promote natural convection and more uniform temperature distributions. The fusion temperature $\left({\theta }_f\right)$ stabilizes thermal profiles through latent heat absorption, adjusting $\mathrm{Cr}$ by 25 %. A higher nanoparticle concentration boosts the average Nusselt number $\left(\overline{\mathrm{Nu}}\right)$ by 10 %, improving overall heat transfer efficiency. The ANN model’s training, reflected in a decreasing mean squared error (MSE), demonstrates prediction accuracy, and regression analysis reveals high model reliability, with predictions closely aligning with theoretical $\left(\overline{\mathrm{Nu}}\right)$ values.