A NOVEL COMPUTATIONAL FRAMEWORK FOR THE EFFECTIVE TRANSPORT PROPERTIES OF HETEROGENEOUS MATERIALS RECONSTRUCTED FROM DIGITAL IMAGES

In computational material science, Digital Image Processing and Big Data analysis play a crucial role in Microstructure Characterization and Reconstruction (MCR), especially in the estimation of structure-property relationships. In this work, we are interested in the calculation of the effective transport properties of composite materials from digital images. Because most MCR techniques are heavily statistical, they may suffer from significant microstructure information loss as they are incapable of reproducing exact images of microstructures. Here, we take advantage of pattern recognition algorithms to extract nearly exact morphological information pertaining to the interphase boundaries from digital micro structural images, thereby minimizing information loss. The data extracted then serves as the basis for our effective transport property module for calculating the effective properties of two-phase composite materials in a way that is automated, fast, stable, memory efficient, and accurate. Our current formulation is limited to circular or near-circular inclusions, with very large contrast properties. Preliminary numerical tests on four cases of 2D, two-phase microstructure images yielded relative errors ranging from 0.1%o to 2.0%, for image pixel density around 1000 x 1000 pixels. These relative errors are perfectly acceptable without having to resort to unrealistic image sizes.