Freeze-Casting of Alumina and Permeability Analysis Based on a 3D Microstructure Reconstructed Using Generative Adversarial Networks.

Abstract

In this study, alumina ceramics with hierarchical pores were successfully fabricated using freeze casting. Experimental studies show that both the solid loading of the slurry and the thermal insulation layer at the interface of the slurry and cooling plate can influence the pore characteristics of cast samples. In order to examine the pore characteristics and evaluate the permeability of the freeze-cast samples fabricated under different conditions, a generative adversarial network (GAN) method was employed to reconstruct the three-dimensional (3D) microstructure from two-dimensional (2D) scanning electron microscopy (SEM) images of the samples. Furthermore, GAN 3D reconstruction was validated against X-ray tomography 3D reconstruction results. Based on the GAN reconstructed microstructures, the permeability and pore distribution of the various samples were analyzed. The sample cast with 35 wt.% solid loading shows an optimal permeability.