Study on heat and mass transfer in a porous cavity based on artificial intelligence δ-SPH model

Abstract

This study examines the influence of key physical parameters on heat and mass transfer of nano-enhanced phase change material (NEPCM) within a heart-shaped cavity, using a hybrid approach of Artificial Intelligence (AI) and the δ-Smoothed Particle Hydrodynamics (δ-SPH) method. The δ-SPH method accurately captures fluid dynamics, while the XGBoost model effectively predicts average Nusselt and Sherwood numbers, highlighting the potential of combining AI with advanced numerical methods. This study addresses challenges in modeling heat and mass transfer in NEPCM systems within complex geometries, focusing on optimizing thermal-fluid performance using advanced numerical and AI techniques. This integrated approach offers a powerful tool for optimizing thermal-fluid systems. The analysis covers parameters such as fractional order, dimensionless time, activation energy, Darcy permeability, Cattaneo heat and mass transmission, Hartmann number, chemical reaction intensity, and Soret and Dufour numbers. The findings reveal that lower fractional order values accelerate thermal response, while higher values slow the transfer process. Activation energy and magnetic fields dampen fluid motion, leading to more stable temperature and concentration fields. Lower Darcy values restrict fluid flow, and higher Cattaneo parameters delay heat and mass propagation. Strong chemical reactions and higher Soret and Dufour numbers enhance the coupling of heat and mass transfer, creating more dynamic flow behavior. Future work will extend this framework to more complex geometries and transient conditions, improving its applicability to real-world thermal management challenges. The findings reveal that lowering the fractional order ($\alpha$) accelerates thermal response, reducing time to steady-state by 20 %. Activation energy ($E$) and magnetic fields ($\mathrm{Ha}$) stabilize flow, decreasing velocity magnitudes by 15 %. The hybrid δ-SPH and AI approach accurately predicts Nusselt ($\overline{\mathrm{Nu}}$) and Sherwood ($\overline{\mathrm{Sh}}$) numbers with errors below 0.5 %. The lower fractional order ($\alpha$) accelerates thermal response, reducing time to steady-state by 20 %. Activation energy ($E$) and magnetic fields ($\mathrm{Ha}$) stabilize flow, decreasing velocity magnitudes by 15 %.