3D DEM-based particle-scale analysis of drained and undrained triaxial behaviors of granular materials

Some numerical simulations of drained and undrained triaxial tests on granular materials with different initial densities are carried out with the three-dimensional discrete element method. An in-depth particle-scale analysis is performed quantitatively to illustrate the physical mechanism of the shear mechanical behaviors, with a special attention paid to the characteristics of quasi-steady state and critical state. The simulation results show that the initial density and shear drainage condition both have significant effects on the evolution of stress-strain, coordination number, fabric anisotropy factor, force chains and clusters. The chained grains ratio and the mean length of force chains in the specimens are constantly adjusted to bear and transfer the changing external loads. The transitions between small clusters and large clusters are also continually taking place in varying degrees, correlating to volumetric contraction or dilation. For the loose undrained triaxial specimen presenting quasi-steady state during shearing, the coordination number decreases obviously to nearly 4 and then increases again; the chained grains ratio decreases after a slight increase in the initial loading stage, and then begin to increase again after a period of lower value of around 0.285; the volume ratio of small, submedium and medium clusters all first decreases and then increase gradually, meanwhile volume ratio of large clusters increases sharply to as much as 0.28 and then decreases gradually. The macroscopic critical state of granular materials is a comprehensively external manifestation when the microscopic coordination number and mesoscopic force chains and clusters all evolute to a dynamic equilibrium. At the critical state, the deviator stress, void ratio, coordination number, fabric anisotropy factor, and the volume ratio of small clusters and large clusters all manifest a respectively unique linear relationship with the mean effective stress.