Investigating the micromechanics of granular soils subjected to cyclic loading using discrete element method

he soil experiences millions of load cycle from nature, i.e., earthquake or many geo-engineering structures are subjected to cyclic loading during normal operation, e.g., renewable offshore wind turbines are likely to experience millions of load cycles, with variations in cycle magnitude and frequency, during their service lives. These forces are transmitted to the soil which may cause unacceptable soil displacement and, in extreme cases, it may lead to soil liquefaction. This is a major concern to geotechnical engineers, therefore determination of soil response to cyclic loading has great importance. This research investigates the undrained behaviour of sand subjected to monotonic and cyclic loading using DEM (Cundall and Strack, 1979). A series of constant-volume undrained simulations of sand subjected to monotonic loading at different initial stress ratio, confining pressure and void ratio were performed to gain the understanding of the monotonic behaviour of sand which is an essential precursor to the cyclic loading tests. One problem arises when shearing dense samples are the generation of unrealistically high stresses. Four alternatives are hypothesised to address the shortcomings of the constant-volume method are explored, each of which has a physical basis: particle crushing, the presence of highly compressible air within the sample, or the reduction of stiffness due to particle surface asperities or non-spherical particle shapes. In situations where a significant amount of particle crushing occurs, it is important to incorporate this in the simulations so that stresses are not over-estimated. In the absence of particle crushing, the most effective method to achieve more realistic stress–strain responses is to reduce the particle shear modulus substantially. This approach has the added computational benefit of enabling an increase in the simulation time-step. A Design of Experiments (DOE) approach was adopted to systematically investigate the behaviour of sand subjected to cyclic loading using DEM. Detailed simulations results are presented that related the influence of different parameters such as frequency, mean cyclic load, cyclic amplitude, confining pressure and void ratio on the dynamic properties of granular materials. Based on those DOE analyses, prediction of cyclic responses for randomly selected input parameters are presented. The void ratio was found to have the most significant effect on the shear modulus and coordination number of the sample. The influence of frequency on cyclic response quantities was found to be insignificant. In addition, energy terms were computed in a set of undrained cyclic triaxial discrete-element method simulations which form a parametric study of five factors: void ratio, initial mean effective stress, mean deviator stress, deviator stress amplitude and compressive/extensive initial loading. Void ratio is the only one of these factors which significantly affects the relationship between the excess pore water pressure and the unit energy. By increasing the void ratio or decreasing the initial mean effective stress, both the number of complete cycles and the energy dissipated per unit volume up to the onset of liquefaction, respectively denoted as Nl and δWd, were reduced. Initial stress anisotropy reduces Nl but increases δWd. Increasing the deviator stress amplitude also reduces Nl but has no significant effect on δWd. All of these observed trends in Nl and δWd match data from physical experiments, where available. The preferred contact orientation for frictional dissipation is between 30° and 40° for these cyclic simulations. There is a greater heterogeneity for extension than for compression, regardless of whether the initial phase of loading is compressive or extensive. Immediately following a shear reversal, the boundary work decreases and there is a period of negligible frictional dissipation which lasts for around 0.04% axial strain. If an energy-based model is being applied for liquefaction assessment of anisotropic samples, a significant improvement in the accuracy of the model may be achieved by including the mean deviator stres