Tang‐Tat Ng

Abstract Using arrays of ellipsoidal particles in the Discrete Element Method (DEM) is one of the means to enhance simulations of granular materials. A major challenge in implementing ellipsoidal element to DEM is the development of an efficient and stable contact detection algorithm in order to properly detect contact formation and compute contact forces on the elements. In view of the current available methods in two‐dimensional (2D) condition, two different contact detection algorithms for ellipsoidal particles in 3D modelling are identified. Their accuracy and efficiency are compared and the results favour the algorithm based on the geometric potential concept. This algorithm has been implemented in the recently developed DEM code ELLIPSE3D.

The paper explores the use of discrete element simulations to model granular soil response to monotonic and cyclic loading. Two‐ and three‐dimensional random arrays of quartz spheres of various diameters are used that crudely represent rounded uniform quartz sand. Computer program CONBAL, developed by the writers from existing code TRUBAL, is used. All simulated granular specimens are first isotropically consolidated, and are then subjected to monotonic drained loading or constant volume (undrained) cyclic “simple shear” simulations. The monotonic results exhibit similar pressure‐dependent shear strength and dilation behavior to that found in actual sands, but with the simulated specimens being stiffer and failing at a smaller strain. The simulated cyclic loading results closely resemble the “pore water pressure” buildup to initial liquefaction, hysteresis loop formation and degradation, “banana loop” shapes, and lines of phase transformation observed in sand experiments. The effect of intergranular friction coefficient μs=tanφu and particle rotation on the results of simulated material is also studied, and they are found to be very important.

This paper presents a sensitivity analysis of the input parameters of a program based on the discrete element method (DEM). Triaxial compression simulations were conducted on an assembly of ellipsoids with two particle shapes. We examine four input parameters including shear modulus of particles, density of particles, time step, and damping. Generally, these parameters are chosen by calibrating the result with certain known behavior of granular materials. In dynamic simulations, these input parameters are bounded by their physical attributions that should not be altered. However, in static simulations, they do not have the same physical implication. Validity of results may be questionable when input parameters are used without justification. A sensitivity analysis of the input parameters should shed light on this issue. In this paper, we will study the effect of the input values within the range of (1∕10)–10 times the benchmark value. The benchmark values are commonly used by the writer. The results are presented against the benchmark simulation. The unbalanced forces in the simulations are kept below a prescribed value to enforce equilibrium. The result shows that the effect of all input parameters used in this paper is negligible as long as the small unbalanced forces in the system can be achieved. The runtimes are different. However, there are two simulations (one with low damping and the other with a large time step size) that cannot maintain the required small unbalanced force. In other words, equilibrium cannot be achieved for these simulations.

Abstract The microscopic and macroscopic behaviors of assemblages of monodisperse ellipsoids with different particle shapes were studied using the discrete element method. Four samples were created with 1170 identical prolate ellipsoids. The samples were compressed isotropically to 100 kPa. Then triaxial compression tests were carried out to very large strains until the ultimate state was reached. This paper presents typical macroscopic result including stress–strain relationship and volumetric behavior. In addition, the fabric of the samples was examined at the initial state, at the peak shear strength state, and at the ultimate state. We studied the evolution of three vector‐typed micromechanical arguments with strain including the particle orientation, branch vector, and normal contact force. The normal contact force (micromechanical argument) was found to have a direct relationship with the principal stress ratio (macroscopic parameter). The angles between these vectors were also investigated. The maximum angle between vectors is related to particle shape. The results indicate that the distributions and the maximum values of these angles do not change with loading. Copyright © 2008 John Wiley & Sons, Ltd.