School of Civil and Environmental Engineering

Ph.D. Thesis Defense Announcement

Computational Techniques for Enhanced Characterization of Granular Material Microstructure

By

M. Mahdi Roozbahani

Advisor:

Dr. J. David Frost (CEE)

Committee Members:

Dr. Arun M. Gokhale (MSE), Dr. Duen Horng Chau-Polo (CSE), Dr. Sheng Dai (CEE), Dr. Andrew Fuggle (Golder Associates, Inc.)

Date & Time: Monday, January 7, 2019, 10:00am

Location: Mason 1132

The behavior of granular materials is of overarching engineering importance, given its ubiquitous presence in industrial activities and nature. In soils, the various forms of particle associations give rise to a range of mechanical and conductive behaviors with great implications to the built environment. The complex structure of packed particles can be regarded as a binary assembly of solid and void. The interplay of these two elements governs the micro-phenomena from which macroscopic behavior emerges. Hence, the characterization of microstructure is fundamental to advance the understanding of geo-systems.Ún effective technique to conduct such characterization consists of numerically generating synthetic specimens following a set of control parameters. In this study, dynamic and geometrical sphere packing (GSP) algorithms were employed to produce a wide variety of synthetic granular material structures. Random close and loose packing simulations were developed to rapidly pack particles geometrically, and discrete element method (DEM) was used to generate specimens dynamically. The generated specimens were then evaluated through new computational approaches, that provide insight into the filtration systems, pore structure, and particles interaction with themselves and their environment.Mechanical trapping (or straining) of fine particles is a key mechanism in many filtration systems. Using an assembly packed via GSP, the pore network and the associated pore size distribution were analyzed using geometrical approaches. Results showed that fine particles between 15% and 25% of the coarse particle size can be physically strained within the randomly packed bed. The technique provides an efficient yet accurate alternative for understanding how fine particles transport through a particulate medium.Pore scale modeling plays a key role in fluid flow through porous media and associated macroscale constitutive relationships. The polyhedral shape and effective local pore size within granular material microstructure are computed in this study by means of the Euclidean Distance Transform (EDT), a local maxima search (non-maximum suppression), and a segmentation process. Various synthetic packed particles are simulated and employed as comparative models during the computation of pore size distribution (PSD). Reconstructed un-sheared and sheared Ottawa 20-30 sand samples are used to compute PSD for non-trivial and non-spherical models.With no more than a couple of thousand years of experience, humankind has developed some innovative techniques to leverage the subsurface for a variety of beneficial functions. In contrast, nature has had the benefit of several billion years to initially design and subsequently evolve the manner in which flora and fau