Date & Time: Monday, November 08, 2021, at 12:00 PM
Location:  Sustainable Education Building (SEB), Room 122/ Virtual via Zoom:
https://us02web.zoom.us/j/89856743663?pwd=dFR6UzVVWlRiMDBkZHhzWWdybXZBUT09

ABSTRACT
The   behavior   of   particulate   materials   is   of   overarching   importance   across   
multiple   science   and engineering  fields,  given  its  ubiquitous  presence  in  nature.  These 
materials  are  typically  composed of two phases, solids and voids, and are therefore described as 
complex multi-phase materials that exhibit non-linear responses when subjected to varying boundary 
conditions. While the attributes of the solid phase of particulate materials have been extensively 
characterized both experimentally and numerically, there  is  much  less  understanding  of  the  
attributes  and  behavior  of  the  pore  phase.  Furthermore, classical pore models incorporate 
idealized assumptions of feature geometries, limiting the accuracy of the   information   that   
can   be   obtained   from   these   features.   This   study   aims   to   advance   digital 
characterization  capabilities  for  particulate  microstructures,  focusing  on  characterizing  
the  geometry and  topology  of  the  highly  complex  pore  space  within  packed  particle  
systems.  A  new  and  robust computational  algorithm  is  proposed  that  quantifies  various  
characteristics  of  the  three-dimensional pore space of a given particulate media, which is 
unimpeded by assumptions of feature shapes or user dependency. The method is validated against 
packings of known pore geometries and implemented on real,  simulated,  and  fabricated  
microstructures  of  different  packing  densities,  particle  sizes,  shapes, gradation, and 
following different specimen preparation techniques to measure its ability in capturing multi-scale 
responses of microstructures.
The study also leverages the emergence of machine learning techniques to scale up the findings to 
real- world field-scale applications comprising particle-pore systems with 10's of millions of 
particles. In this regard,  the  use  of  deep  learning  tools  for  the  rapid  estimation  of  
pore  space  properties  from  three- dimensional  images  is  sought.  Finally,   the  developed  
techniques  and  tools  are  implemented  on  real granular  soils  to  strengthen  the  
understanding  of  macro-geomechanical  phenomena.  The  findings highlight  the  importance  of  
accounting  for  pore  space  properties  when  interpreting  the  macroscopic
response of granular assemblies subjected to external mechanical and precipitational loading.