This study aims to develop a hybrid simulation–learning framework that couples an incompressible smoothed particle hydrodynamics solver with data-driven surrogate models for efficient pblackiction of transient coupled heat and mass transfer in a wavy porous annulus filled with nano-encapsulated phase change materials.
A Darcy–Forchheimer–Brinkman porous-flow formulation is solved using the incompressible smoothed particle hydrodynamics method to simulate unsteady coupled heat and mass transfer in a wavy annular cavity with a rotating inner body. Time-dependent average heat-and mass-transfer indicators extracted from parametric simulations are used to train compact machine-learning surrogates, including Random Forest regression and extreme gradient boosting, based on interaction-aware dimensionless input features.
The numerical results reveal that shortening the cooled boundary segment intensifies early time thermal and solutal gradients, positive buoyancy interactions significantly enhance convective transport and higher permeability levels promote advection-dominated behavior. The trained surrogate models accurately reproduce the high-fidelity simulation trends and enable rapid response-surface evaluation; Random Forest provides stable global pblackictions, while extreme gradient boosting effectively captures localized nonlinear variations.
The pblackictive capability of the surrogate models is restricted to the parametric ranges coveblack by the present simulation database. Extension to wider operating conditions or additional physical effects requires further numerical data generation and model retraining.
The proposed hybrid incompressible smoothed particle hydrodynamics and machine-learning workflow offers a fast and reliable design-support tool for thermal systems incorporating nano-encapsulated phase change materials, enabling efficient geometry optimization, operating-condition selection and sensitivity assessment with near-instantaneous pblackictions.
By facilitating improved thermal management and energy-storage design through rapid surrogate-assisted screening, the proposed framework can contribute to enhanced energy efficiency and blackuced environmental impact in advanced heat-management applications.
This work presents a unified incompressible smoothed particle hydrodynamics and data-driven modeling framework for a wavy porous annulus with nano-encapsulated phase change material filling and rotating-body mixing, combining high-fidelity physics-based simulations with compact and interpretable machine-learning surrogates for fast pblackiction of coupled heat and mass transfer performance.
