Microstructure Characterization Using X-Ray Microtomography of Wet Unsaturated Granular Material

Unsaturated wet granular materials exhibit intricate microstructures composed of solid particles, liquid phases, and void spaces. Understanding the morphological and rheological characteristics of these materials is essential for various applications, from geotechnical engineering to environmental science. By employing X-ray tomography, we aim to understand the relationship between the microstructural features and rheological properties of these materials, shedding light on their behavior and interactions in various conditions. Our focus is on slightly polydisperse polystyrene beads mixed with a minimal amount of liquid, specifically in the pendular regime (the ratio between the liquid volume and the volume of polystyrene beads is less than 0.075), where capillary bridges are the primary liquid morphology, although other morphologies also exist in smaller proportions. A custom shear device, compatible with an X-ray microtomography imager, has been designed to observe microstructural evolution under imposed confining stress and shear rate. Through these X-ray microtomography experiments, we capture detailed 3D images at different deformation stages. Employing advanced image segmentation techniques, which integrate machine learning and deep learning, we can analyze these complex microstructures accurately and comprehensively. Our segmented images offer a deeper understanding of grains and liquid distribution, as well as the different liquid morphologies. In particular, we have developed an automatic tool for classifying the different liquid morphologies within the sample. This method enables us to analyze the 3D spatial distribution of the grains and liquid fractions, in addition to the changes in the liquid morphologies, providing insights into their responses under shear conditions.

Keywords: X-ray microtomography, Artificial Intelligence, U-Net model, Random Forest model, Unsaturated wet granular materials.