The effective thermal conductivity of random isotropic porous media analysis and prediction

Effective thermal conductivity of porous media is a crucial parameter for heat transfer within them. Many studies have characterized various porous media by adjusting the control parameters generated through the Quartet Structure Generation Set method. The porous media effective thermal conductivity is then determined through Computational Fluid Dynamics calculations, which, however, necessitate significant computational resources and time. Thus, following an exploration of the influence of control parameters (i.e., porosity, core growth probability, and their coupling) on the effective thermal conductivity of isotropic porous media generated by the Quartet Structure Generation Set method, this study developed a multi-layer perceptron prediction model. The aim was to establish a prediction model from the porous media control parameters to effective thermal conductivity, thereby reducing the time spent on iterative calculations. The findings indicate that the effective thermal conductivity does not uniformly increase with the growth of core probability, and instead fluctuates after a certain threshold. Notably, the trained model exhibits a high prediction accuracy, with average deviations of 0.0887 and 0.0748 for the training and testing datasets, respectively.