ML-Assisted Thermal Screening of Green Fin Materials for Passive Cooling under Natural Convection: An Analytical Sustainability Study

Abstract

This paper presents a submission-ready journal manuscript on passive cooling using green fin materials under natural convection. A vertical plate-fin heat sink was analysed as a no-CAD, no-laboratory design problem so that the thermal logic remained reproducible and suitable for early-stage mechanical-engineering screening. Four material cases were compared: copper, aluminum, recycled aluminum, and recycled aluminum with a high-emissivity black coating. A deterministic design space of 108 configurations was generated by varying fin height, fin spacing, and base temperature while keeping the overall geometry simple enough for spreadsheet or script-based implementation. Classical heat-transfer equations were used to estimate convection, linearized radiation, fin efficiency, total heat dissipation, thermal resistance, mass, and heat dissipation per unit mass. The resulting dataset was then used to build a lightweight machine-learning layer in which linear regression served as an interpretable baseline and random forest regression acted as a nonlinear surrogate for rapid thermal-performance prediction. Graphical results show that fin height improves heat rejection with diminishing returns, fin spacing governs the trade-off between buoyancy access and fin population, and emissive surface treatment becomes increasingly valuable as the base temperature rises. Within the analysed design space, the coated recycled-aluminum configuration produced the strongest absolute thermal result, while aluminum achieved the highest weighted sustainability score under the selected multi-criteria scheme. The manuscript therefore demonstrates that a short mechanical-engineering paper can still integrate thermodynamics, material selection, sustainability reasoning, data visualization, and a modest AI layer in a form aligned to a fast-track multidisciplinary journal submission.

Introduction

The text presents a study on passive thermal management using a plate-fin heat sink under natural convection, aimed at low-power and compact electronic systems. It emphasizes that passive cooling performance depends heavily on geometry, material selection, and surface emissivity rather than just thermal conductivity.

An analytical model is developed instead of a physical prototype, generating 108 design configurations by varying fin height, spacing, material type, and base temperature. Heat transfer is evaluated using standard natural convection and radiation relations, and performance metrics include heat dissipation, thermal resistance, mass-normalized output, and a sustainability score. Materials compared include conventional aluminum, recycled aluminum, and coated recycled aluminum.

A small machine-learning component is added after the physics-based dataset is generated. Linear regression and random forest models are trained to predict heat dissipation from design parameters. The random forest performs better, showing strong agreement with analytical results and capturing nonlinear effects, with base temperature and emissivity identified as the most influential factors.

Results show that increasing fin height improves heat dissipation but with diminishing returns due to fin efficiency effects, while fin spacing creates a trade-off between surface area and airflow. The best-performing configuration is coated recycled aluminum with tall fins and tight spacing, achieving the highest heat rejection and low thermal resistance, largely due to enhanced radiative heat transfer.

From a design perspective, aluminum-based and recycled aluminum systems are highlighted as competitive, especially when mass and sustainability are considered. The study concludes that surface emissivity improvements can sometimes outperform simply using higher-conductivity materials like copper in passive systems. The machine-learning model is positioned as a fast surrogate tool for early-stage thermal design screening rather than a replacement for physics-based analysis.

Conclusion

This paper reformulates a longer final-year thermal report into a concise journal-style study and shows that a no-CAD, no-laboratory workflow can still deliver technically meaningful results. Classical heat-transfer modelling produced an interpretable design space for passive cooling, while a lightweight ML layer converted that design space into a rapid prediction framework. Within the analysed cases, the coated recycled-aluminum configuration achieved the best absolute thermal outcome, whereas aluminum secured the highest weighted sustainability score under the chosen criteria. Random forest outperformed linear regression for heat-dissipation prediction, confirming the usefulness of a small nonlinear surrogate in this type of engineering problem. Future work can extend the same framework through CAD refinement, CFD validation, prototype testing, and application-specific sustainability weighting.

References

[1] F. P. Incropera, D. P. DeWitt, T. L. Bergman, and A. S. Lavine, Fundamentals of Heat and Mass Transfer. Hoboken, NJ, USA: John Wiley & Sons. [2] J. P. Holman, Heat Transfer. New York, NY, USA: McGraw-Hill. [3] Y. A. Cengel and A. J. Ghajar, Heat and Mass Transfer: Fundamentals and Applications. New York, NY, USA: McGraw-Hill. [4] A. D. Kraus, A. Aziz, and J. Welty, Extended Surface Heat Transfer. New York, NY, USA: John Wiley & Sons. [5] M. F. Ashby, Materials and the Environment: Eco-Informed Material Choice. Oxford, U.K.: Butterworth-Heinemann. [6] L. Breiman, \"Random Forests,\" Machine Learning, vol. 45, no. 1, pp. 5-32, 2001. [7] G. James, D. Witten, T. Hastie, and R. Tibshirani, An Introduction to Statistical Learning. New York, NY, USA: Springer. [8] F. Kreith, R. Manglik, and M. Bohn, Principles of Heat Transfer. Stamford, CT, USA: Cengage Learning. [9] Y. Joshi and A. Fedorov, \"Thermal management of electronics through compact heat sinks and passive cooling strategies,\" Journal of Electronic Packaging, vol. 130, no. 4, pp. 041001-1-041001-9.

Copyright

Copyright © 2026 Shubhangam ., Shivam Kumar, Vasu Bansal. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.