A Review on Boiling Heat Transfer in Channels and Metal Foams: Experimental, Numerical, and AI Model Studies

ABSTRACT

Effective thermal management is a significant engineering challenge in high‐flux applications, including microelectronics, small modular reactors (SMRs), and electric vehicle (EV) batteries, where heat fluxes are very high. Subcooled flow boiling uses latent heat for energy transfer and has become a highly effective cooling method. This review assesses two primary approaches to improve the performance of boiling heat transfer (BHT): utilizing sustainable binary fluid mixtures and incorporating porous metallic structures. Binary mixtures are of interest because they show thermal and physical properties that vary and have a lower negative impact on the environment. However, their unique attributes, such as resistance to diffusion of mass and Marangoni convection, create difficulties for predicting their behavior. On the other hand, porous foams enhance BHT by increasing the density of nucleation sites and improving capillary wicking. This can raise the critical heat flux (CHF) by over 50%. However, these structures cause higher pressure drops and the potential for vapor entrapment. The present review examines the limitations of current approaches, such as analytical and empirical methods, and explores advanced computational models, including the volume of fluid (VOF) method, the Lattice Boltzmann method (LBM), and AI‐based models, to gain a deeper understanding of the complicated dynamics of the bubble. The review also highlights the need to identify generalized modeling that are applicable to binary mixtures and recognize AI‐based models as a means to develop robust, scalable, and environmentally friendly thermal solutions. Overall, the review finds that porous metallic foams raise the CHF by 50%–58% through higher nucleation‐site density and capillary wicking, whereas sustainable binary mixtures such as water–ethanol reduce environmental impact but introduce composition‐dependent nucleation and Marangoni effects that existing empirical correlations cannot generalize across fluids or geometries. Physics‐informed and data‐driven (AI/ML) models are identified as the most promising route to close this predictive gap. Building on these findings, the review recommends future work on generalized correlations for binary‐mixture and porous‐media boiling, additive‐manufactured graded‐porosity foams, transient/cyclic‐load testing relevant to EV batteries and SMRs, and standardized, shared experimental datasets to train and validate next‐generation AI‐based predictive models.