Hybrid incompressible SPH–machine learning approach for simulating natural convection in a porous intricate domain
This study explores the double-diffusive convection of nano-enhanced phase change material (NEPCM) inside a intricate-shaped cavity partially filled with porous media, integrating Incompressible Smoothed Particle Hydrodynamics (ISPH) with an Artificial Intelligence (AI)-based predictive model. The n...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Case Studies in Thermal Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2214157X25009256 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study explores the double-diffusive convection of nano-enhanced phase change material (NEPCM) inside a intricate-shaped cavity partially filled with porous media, integrating Incompressible Smoothed Particle Hydrodynamics (ISPH) with an Artificial Intelligence (AI)-based predictive model. The novelty of this work lies in developing a hybrid numerical-AI approach to analyze complex heat and mass transfer phenomena influenced by buoyancy effects, thermal radiation, and thermosolutal interactions, offering a computationally efficient alternative to traditional simulations. A comprehensive mathematical model is developed based on double-diffusive convection equations, incorporating Soret and Dufour effects, Darcy–Brinkman–Forchheimer modeling of the porous medium, and latent heat release from NEPCM. The ISPH method is employed for numerical simulations, ensuring accuracy in handling fluid-porous interactions and phase change processes. Additionally, an ensemble machine learning model (Bagging Regression) is trained on ISPH-generated data to predict average Nusselt number (Nuavg) and average Sherwood number (Shavg) numbers, enhancing the computational efficiency of parametric studies. Key findings reveal that buoyancy ratio, Darcy number, Rayleigh number, and Dufour-Soret interactions significantly affect thermal and solutal transport. Increasing buoyancy strength enhances convective mixing, while high permeability accelerates heat transfer. The model is validated by a mesh independence study and benchmark comparisons. Results indicate that increasing the Rayleigh number from 103 to 106 enhances average Nusselt number by over 50 %. AI-based predictions achieve an accuracy exceeding 97 %, demonstrating their reliability in modeling nonlinear transport dynamics. The study provides valuable insights for thermal management systems, energy storage applications, and porous media heat exchangers. The integration of machine learning with ISPH modeling offers a computationally efficient framework for optimizing NEPCM-based cooling and storage technologies in electronics, solar energy systems, and industrial heat exchangers. |
|---|---|
| ISSN: | 2214-157X |