Neural modeling of nano-encapsulated PCM convection in a wavy porous enclosure with thermal radiation

The reliance on energy storage systems is a cornerstone of energy efficiency, establishing them as a vital technology in modern times. Maintaining thermal stability and ensuring effective heat transfer through cooling are critical for their optimal operation. Nano-encapsulated phase change materials (NEPCMs) have gained remarkable attention in energy storage and cooling applications because of their considerable latent heat properties during phase changes. For this reason, NEPCMs are frequently used. They are thought to be among the most promising nanomaterials in this field. The present paper aims to investigate buoyancy-driven convection along with a second law examination within a NEPCMs-occupied U-shaped porous enclosure with a cold obstacle inside and a hot stair-like wavy heater. The flow in the porous medium is predicted via the Brinkman-Forchheimer-extended Darcy formulation, and the impacts of thermal radiation are considered. The finite element method (FEM) is implemented to obtain accurate solutions of the governing equations, and an artificial neural network (ANN)-based multi-layer perceptron (MLP) learning algorithm is applied to predict mean heat transfer rates. The results show that cold obstacle placement and domain inclination angle strongly affect natural convection. The average Nusselt number intensifies with Rayleigh number, radiation number, porosity, Darcy number, and NEPCM concentration, but decreases with Stefan number, with the maximum value occurring for obstacle location χ = 0.4 and orientation angle λ = 0°. Moreover, the ANN-based MLP model achieved a best validation performance of 1.7682e-4 at epoch 9, confirming its predictive accuracy. These results provide promising insights for optimizing thermal energy storage and cooling system designs using NEPCM suspensions.