A generalized characterization of radiative properties of a porous media using engineered features and neural network

The assessment of radiative properties of materials and devices at high-temperature levels is an imperative and inevitable part of various engineering applications. Particularly, the study of radiative heat transfer in porous media has attracted the attention of researchers for decades. Advanced computational algorithms have provided a reliable approach to tackle the cumbersomeness of experimental measurements. In this study, Monte Carlo ray tracing (MCRT) method is implemented to generate supervised labeling data for random overlapping and non-overlapping circular packed beds by solving the classical radiative transfer equation. A highly generic artificial neural network (ANN) model based on engineered physical and geometrical features is designed to predict the radiative properties of a given arbitrary porous media at significantly lower computational costs. Results demonstrate the generalizability and applicability of the present model for the calculation of radiative characteristics of porous media.