ISPH simulation and XGBoost prediction of double diffusion of NEPCM with sloshing fins under thermal radiation, Soret, and Dufour effects

This work develops an integrated framework combining incompressible smoothed particle hydrodynamics (ISPH) and eXtreme Gradient Boosting (XGBoost) to analyze double-diffusive convection of nano-encapsulated phase change materials (NEPCM) in a porous cavity with sloshing fins under thermal radiation, Soret, and Dufour effects. A clear research gap is addressed by treating the coupled impact of dynamic sloshing fins, cross-diffusion (Soret–Dufour), and radiation in porous NEPCM systems topics commonly examined in isolation in prior studies thereby establishing the study’s novelty. The ISPH solver captures unsteady thermo-solutal transport, and XGBoost provides fast surrogate predictions of the average Nusselt and Sherwood numbers. Model credibility has been reinforced through benchmark validation for natural convection in porous enclosures and a particle-resolution independence study (variation ≤ 2% in Nu avg and Sh avg ). A fractional-time derivative is retained as a theoretical generalization for possible memory effects in porous/NEPCM media; it is not experimentally calibrated here and results are interpreted accordingly. Key trends indicate that increasing the Darcy number from 1 0 − 5 to 1 0 − 2 enhances Nu avg and Sh avg by about 36.2% and 28.4%, respectively, whereas raising the Hartmann number from 0 to 60 reduces them by 17.5% and 14.3%. Thermal radiation promotes heat transfer, with Nu avg increasing by 42.7% as Rd rises from 0 to 5. The trained XGBoost models, fitted to the validated ISPH database, reproduce target metrics with low errors, enabling rapid parametric screening. Main limitations include the two-dimensional setting, sinusoidal fin kinematics, and the absence of direct experiments; prospective extensions include 3D configurations, non-Newtonian NEPCM rheology, and active fin-control optimization.