Mingjing Jiang

Loess soils undergo collapse when the material is loaded and wetted, which leads to engineering problems and disasters. To understand better the collapse mechanisms that govern loess and loess-like deposits under complex stress paths, a distinct element model has been established. A contact model considering water content effect introduced for structural loess was used to cement contacted particles together to simulate unsaturated structural loess under biaxial compression conditions. The numerical structural loess was tested for different wetting and loading paths, i.e., loading-quick wetting (LW^Q), loading-gradual wetting (LW^G), and wetting-loading (WL) paths. The simulated macroscopic mechanical behavior agrees approximately quantitatively in terms of the deviatoric deformation and qualitatively regarding the volumetric deformation when compared with available experimental results. The tests under different mean stresses indicate the existence of initial, maximum, and terminating collapse stress states in a biaxial wetting test. The influence of the wetting procedure, i.e., gradual wetting or quick wetting, has a considerable effect on the mechanical performance for numerical samples. The axial strain of a sample during a loading-wetting (LW) path is larger than that of a sample in a WL path. The wetting-induced volumetric deformation corresponds with the failure of bonded contacts. The deviator fabrics (fabric anisotropy) of total, bonded, frictional, and broken bond contacts were also investigated.

The surfaces of real sands are often rough and a discrete element modeling of rough granulates is very important to understand the behavior of real sands. This paper presents a two-dimensional discrete element method (DEM) to capture the roughness of particles. The model consists of two parts: equations governing the motion of rough particles and mechanical contact models controlling rough-contact behavior. The key point of the theory is that the assumption in the original DEM proposed by Cundell and Strack in 1979 that a pair of particles are in single-point contact, is here replaced by that the particles are in rough contact over a width. By making the idealization that the contact width is homogeneously distributed with a finite-number of the normal/tangential basic elements (BEs) (each BE is composed of spring, dashpot, slider, or divider), we relate the governing equations to local equilibrium and also establish a rolling contact model together with normal/tangential contact models. The three main fe...

A discrete element modelling of naturally microstructured sands is very important to geomechanics. This paper presents a simple discrete element model for naturally microstructured sands with the aim to efficiently capture the effect of cementation between particles (bonds). First, a simple bond contact model was proposed by introducing a rigid‐plastic bond element into the conventional contact model for dry granular material. Second, efficient numerical techniques were investigated to implement this contact model into the distinct element method (DEM). Then, a two‐dimensional DEM code was developed to simulate a series of isotropic compression tests on the materials of different densities and bonding strengths. Finally, the DEM results were examined in comparison with the experimental data on artificially bonded sands obtained by Rotta et al. (Géotechnique 2003; 53(5):493–502). In addition, we discussed the yielding mechanism, the Coop and Willson criteria on weak/strong bonding (J. Geotech. Eng. (ASCE) 2003; 129(11):1010–1019) and the strong bonding phenomenon observed by Rotta et al. based on the DEM data. The study shows that the DEM model is able to capture the main features of naturally microstructured sands, such as variations of yielding and bulk modulus against bonding strength or material density. In addition, it is shown that the gross yielding (the yielding defined in terms of strains) is largely related to bond breakage; Coop and Willson criteria are generally reasonable; and the strong bonding in the experimental data obtained by Rotta et al. comes from that their bonded materials start at different points on the same compression line. Copyright © 2006 John Wiley & Sons, Ltd.