Magnetic effects on solid particle dispersion from circular cylinders and double-diffusive convection in a NEPCM-filled porous thermal chamber

Purpose This study aims to examine the effects of magnetic fields, buoyancy ratios (N), Darcy numbers (Da) and thermal radiation parameters (Rd) on the dispersion of solid particles and double-diffusive convection in a porous cavity filled with nano-enhanced phase change materials (NEPCM). The cavity includes three high-temperature and high-concentration circular sources and a fixed cylinder, creating complex convective interactions. The goal is to quantify heat and mass transfer characteristics and improve thermal storage system efficiency by integrating numerical simulations with machine learning for enhanced predictive accuracy. Design/methodology/approach The incompressible smoothed particle hydrodynamics (ISPH) method is used for solving the governing equations of heat, mass and momentum transfer in the porous cavity. The study incorporates the Soret and Dufour effects and examines convective behavior across a range of N, Da and Rd values. To enhance predictive analysis, a machine learning framework based on ensemble regression with bagging is developed to estimate the average Nusselt (Nu¯) and Sherwood (Sh¯) numbers with high accuracy. The computational approach is validated against benchmark studies, ensuring reliability in thermal energy storage applications. Findings Results indicate that increasing N enhances thermal and solutal transport, driven by thermal buoyancy effects, whereas negative N values lead to solutal-dominant behavior. Higher Da values promote strong convective mixing, while lower Da induces a conduction-dominated regime, suppressing particle dispersion. Thermal radiation (Rd) significantly enhances heat and mass transfer, optimizing energy transport efficiency. The machine learning model achieved low prediction errors (MSE ∼ 10−3), successfully identifying critical feature interactions such as τ · Rd and τ · N, influencing transfer dynamics. Originality/value This study presents a novel integration of ISPH-based numerical simulations with machine learning to analyze solid particle dispersion and double-diffusive convection in an NEPCM-filled porous cavity. The research uniquely quantifies the interplay between thermal radiation, magnetic effects and convective transport in the presence of multiple heat sources. By leveraging machine learning, the study enhances predictive capabilities for Nusselt and Sherwood numbers, offering practical insights into thermal energy storage optimization. These findings contribute to next-generation heat transfer models, providing a framework for improving thermal management in industrial and renewable energy applications.