Investigating the Simple Shear Response of Gravelly Soils Using the Discrete Element Method

Abstract

Understanding the simple shear behavior of soils is very important in many geotechnical engineering problems such as earthquakes and slope stability. The response of gravelly soils is particularly critical for understanding the performance of earth and rockfill dams, and other gravel embankments and fills. The observed gravelly soil liquefaction during recent earthquakes (2008 Wenchuan, China, 2014 Cephalonia, Greece, and 2016 Kaikoura, New Zealand) have highlighted the need for improved understanding of the seismic response of gravels to improve infrastructure. Given the advances in computational power, particle-based discrete element methods have been progressively used in analyzing soil behavior. However, these models have only been validated by comparing simulation and experimental results using idealized granular materials. When real soil is introduced, existing laboratory tests have used sandy soils that cannot be 1:1 modeled in discrete elements due to the extremely large number of particles that render the needed computational time unrealistic. This dissertation addresses these limitations by validating 3D DEM analyses using results from laboratory tests on real soils, where each particle is accounted for and is characterized by size and shape. Results from 3D Discrete Element Modeling, large scale stacked-ring simple shear laboratory tests and Translucent Segregation Table (TST) tests are integrated to investigate the monotonic and cyclic behavior of Pea gravel specimens. 3D DEM simulations of this study, by considering a realistic algorithm for modeling the movement of stacked rings, indicate that extra caution should be taken and necessary modifications should be applied for testing rounded to sub-rounded granular materials in stacked-ring simple shear device under constant volume condition to ensure proper imposition of simple shear deformation on the specimen. It is also shown that incorporating both the irregularity and non-uniformity of particles shape by assigning a distribution of rolling resistance to the equivalent spherical particles (used to reduce the computational effort) is crucial for acknowledging the diversity of particle shapes and providing a realistic representation of soil assembly in the numerical simulations. Such simulations are more reliable in predicting the behavior of actual soil and provide valuable information on the complex behavior at the micro- and meso-scale, which can eventually be used in developing more robust constitutive models based on the micromechanical response of granular assemblies. Results from the DEM simulations in this study allow for a better understanding of soil response during constant volume simple shear testing. For example, evaluating the stress state at the specimen core, it is shown that a single assumption should not be made about the stress state inside the specimen during the whole course of shearing. The DEM simulation results also suggest that the level of density can affect the level of shear strain at which each of the assumptions may be realistic and can be confidently used in interpretation of simple shear test data. The simulation results also provide deeper insight into the existing non-uniformities of stress and strain inside the specimen and into how they compare to the boundary measured ones as done in laboratory. For example, the simulations show that boundary measured shear strain overestimates the average induced shear strain inside the specimen and that the boundary measured pore pressure ratio is more positive in value than the actual one generated inside the specimen. Such information helps in making more accurate interpretation of experimental data for analysis of soil response.