Machine learning prediction of thermal transport in porous media with physics-based descriptors

•The significant effect of pore distribution and shape on the effective thermal conductivities of porous media are identified.•Five structural descriptors with explicit physical meanings are proposed: shape factor, bottleneck, channel factor, perpendicular nonuniformity, and dominant paths.•These descriptors effectively quantify the anisotropy of pore morphology and strongly correlate with effective thermal conductivities.•The proposed descriptors are incorporated into machine learning models to predict the effective thermal conductivity of porous media and show significantly improved accuracy than using porosity alone. Understanding the thermal transport mechanism in porous media is important for various engineering and industrial applications. The effective thermal conductivity of porous media is known to be related to the morphology of porous structures. However, existing effective medium approaches usually miss the morphology effects, and numerical simulations are expensive and not physically intuitive. Machine learning methods have recently been successful in predicting effective thermal conductivity, but the lack of descriptors limits physical insights. In this work, we investigate structural features that have significant effects on thermal transport in porous media and identify five physics-based descriptors to characterize the structural features: shape factor, bottleneck, channel factor, perpendicular nonuniformity, and dominant paths. These descriptors can effectively quantify the anisotropy of pore morphology and strongly correlate with effective thermal conductivities. The proposed descriptors are incorporated into machine learning models to predict the effective thermal conductivity of porous media, and the results are shown to be fairly accurate. They provide new insights into the thermal transport mechanisms in complex heterogeneous media.