Boundary effects in discrete element method modeling of undrained cyclic triaxial and simple shear element tests

Abstract

Cyclic triaxial and simple shear tests are both commonly used to study the behavior of granular assemblies subjected to dynamic loading. These two stress paths have also been simulated with the discrete element method (DEM) using varying boundary conditions. In this work, stiff boundaries and periodic boundaries were implemented in DEM simulations of constant volume cyclic triaxial and simple shear tests. The model response was analyzed at the specimen scale by comparing stress paths, decrease in mean effective stress, and number of cycles to failure across simulations. Particle scale analyses, including quantification of the contact force network, entropy of the local void ratio distribution, and spatial variability in the distribution of void ratio, contact number, and normal contact force are used to provide insight into model response. Simulation results imply that periodic boundaries are more appropriate for consolidating specimens to a homogenous state and minimizing boundary effects during cyclic loading. Significant uneven “squeezing” where particles are pinched between the horizontal and lateral plates in the corners of the specimen with stiff boundaries was observed in the cyclic simple shear test. After liquefaction initiation, increased entropy indicated a more disordered particle assembly in all cases. In addition, specimens in cyclic simple shear tests and cyclic triaxial tests behave significantly differently at the microscale. With the same level of deviatoric shear strain amplitude, the simple shear specimens liquefied in fewer cycles than the triaxial specimens. Cyclic triaxial tests are not able to produce pure shear waves, while cyclic simple shear tests do, micromechanically indicating the better suitability of cyclic simple shear tests to simulate shear wave induced liquefaction. The findings from this work will have implications for the future modeling of cyclic element tests using DEM.