Integrated optimization of fin geometry and nanoparticle-enhanced PCMs in shell-and-tube thermal storage systems: A CFD–ML framework

This study presents a comprehensive numerical and machine learning-based investigation of a shell-and-tube latent heat thermal energy storage (LHTES) system enhanced through the integration of optimized fin geometry and nanoparticle-enriched phase change materials (NEPCMs). Using Docosane as the base PCM, the effects of various nanoparticles - graphene nanoplatelets (GNP), aluminum oxide (Al2O3), and multi-walled carbon nanotubes (MWCNT) - were evaluated in conjunction with geometric variations in fin number, thickness, length, and material. A detailed computational fluid dynamics (CFD) model was developed using the enthalpy-porosity approach to simulate the phase change process. Results show that increasing fin thickness and length significantly accelerates PCM melting, with 22 mm-long, 5 mm-thick copper fins achieving the best performance. Among NEPCMs, MWCNTs at 0.6 wt% concentration reduced melting time by up to 44% compared to the pure PCM baseline. However, diminishing returns were observed beyond certain fin counts and nanoparticle loadings. To enhance generalizability and design insight, machine learning regression models were trained on CFD data, achieving high prediction accuracy (R2 = 0.99). Feature importance analysis revealed that fin count and thickness had the strongest influence on melting time, followed by nanoparticle type and concentration. The combined CFD-ML framework enabled the development of optimized design guidelines that balance thermal performance, material cost, and system scalability. These findings demonstrate the potential of integrated fin-nano design strategies in advancing the efficiency of compact thermal energy storage systems for applications such as smart buildings, solar thermal systems, and industrial heat recovery. The integration of ML not only improves predictive capabilities but also reveals non-obvious interactions between fin geometry and NEPCM composition, enabling synergistic thermal optimization beyond conventional parametric studies.