Computational simulation of natural convection of NEPCM nanofluid in a finned cylindrical annulus using GA-assisted machine learning

Purpose Lithium-ion batteries present significant thermal management challenges due to their high-energy density, making efficient battery thermal management systems (BTMS) essential. While phase change materials (PCMs) have been used to control overheating, their low thermal conductivity and potential leakage limit performance. Nanoencapsulated PCMs (NEPCMs) enhance the heat transfer, and the addition of fin structures further improves convective cooling. However, conventional BTMS research predominantly uses straight fins, which often misalign with buoyancy-driven plumes, reducing overall efficiency. To overcome this limitation, the purpose of this study is to examine natural convection of NEPCM nanofluids in a cylindrical annulus featuring slanted and W-shaped fins, systematically assessing how fin width affects heat transfer to generate quantitative design guidelines for advanced BTMS. Design/methodology/approach This study explores natural convection of NEPCM nanofluids within a cylindrical annulus equipped with slanted and W-shaped fins, varying in width from 0.02 to 0.08. The governing equations for buoyancy-driven flow and thermal transport are solved via an artificial-compressibility finite difference method. Thermal performance is assessed across a broad spectrum of dimensionless parameters, including Rayleigh, Hartmann and Darcy numbers, fusion temperature and nanoparticle volume fraction. The resulting data are analyzed using a machine-learning framework integrated with genetic algorithm optimization, enabling precise prediction and systematic optimization of fin geometry and width. Findings By combining genetic algorithms optimization with machine learning, the model achieves R2 = 0.98 with a markedly reduced mean squared error, accurately capturing the nonlinear effects of Rayleigh number, Darcy number, Hartmann number, fin width and NEPCM volume fraction on heat transfer. At Ra = $10^6$, slanted fins enhance heat transfer by 73% compared to W-fins, while narrow fins (width = 0.02) consistently deliver the highest Nusselt numbers. In the slanted-fin arrangement, increasing the width from 0.02 to 0.04 raises performance by only 2.07%, whereas the same width change in the W-fin case produces a 21.11% variation. Raising NEPCM concentration to 4.5% further boosts performance, yielding up to a 49% increase in the slanted configuration. Thermal buoyancy peaks at 43.9% for fins with width 0.02 and decreases with larger widths, showing a similar but weaker trend for W-fins. Originality/value The present work demonstrates that narrow slanted fins paired with NEPCM yield the highest enhancement of buoyancy-driven heat transfer in cylindrical annuli. In contrast to previous studies that primarily use straight fins or conventional numerical optimization, this research leverages a GA-ML framework, offering a reliable predictions. The findings provide quantitative, design-focused insights for advanced BTMS and can be applied to related fields, including solar thermal systems and latent heat energy storage, presenting novel strategies for the development of compact, high-performance passive cooling solutions.