The research supported by this award is directed toward the development of a suite of stereological algorithms to allow for the robust quantification of the three-dimensional structure of granular soils from the results of computed tomography (CT) scans of specimens tested in the laboratory. These algorithms will allow for the unambiguous description of the solids and void space at multiple scales within the specimen. Such an approach is potentially transformative with respect to the manner in which the structure of granular materials are measured, described, quantified, and modeled. These algorithms, when coupled with existing approaches to the quantification and simulation of particle shapes, will allow for unprecedented amounts of information to be extracted from laboratory measurements and, if desired, used for discrete numerical simulations. There is currently a significant disconnect between the experimental and numerical approaches that are exploited for the study of granular material microstructure. An experimental approach relies upon observation of the response of real materials in the laboratory while a numerical approach is typically applied to an idealized assembly of particles but is capable of rapid replicate tests and continuous monitoring of microstructure as a function of boundary deformation. However, there is not currently a robust and consistent procedure for linking the two research strategies across multiple scales. The same algorithms used to quantify microstructure in real specimens can be used to guide the generation of granular assemblies for DEM simulations such that they are structurally and statistically similar to those measured in the laboratory.
This suite of algorithms will be valuable to researchers in geotechnical engineering and other fields where characterization of microstructure at similar resolutions is necessary: e.g., petroleum engineering, materials processing, composite materials, ceramics, and some consumer goods. While some individual algorithms currently exist in the literature, these have often been developed on proprietary (and varying) platforms and are generally not widely distributed. Significantly, these disparate approaches have not been synthesized across a single coherent data set to allow for comparison of results, inference of higher-order behavior, or realization of any computational efficiencies that may arise when performing multiple characterizations of a single microstructure. Thus, development and dissemination of the proposed algorithms may have significant broader impacts. In addition, the proposed work will provide valuable training to the students working on the project team and equip them with a set of skills that is portable to a variety of problems.
