3.2 Simulation conditions
Large and small glass bead particles, described by the median sizes of the actual particles, 180 and 40 µm respectively, are used in the current study. Both particles are assumed to be spherical and possess the same material properties. Initially, a number of large spheres and small spheres are randomly generated in the simulation domain, periodic in two dimensions, with a specified mixture composition. The size of the domain is \(30d_{S}\times 30d_{S}\times 15d_{S}\), in which \(d_{S}\)is the diameter of the small particle. Based on our previous work2, it is believed the simulated shear stress results are independent of the domain size, although the domain size is smaller than the shear cell size. Cases with mass fraction of large particles ranging from 0 to 1 are conducted. After all particles deposit on the bottom wall of the DEM shear cell and achieve a negligible velocity, a normal pressure (2000 Pa) is specified on the top wall and a velocity (shear rate of 0.2 s-1) is applied to the bottom wall to start the shearing process as shown in Figure 5. The shear rate is higher than the experimental one to reduce the computational expense but it also assures a quasi-static state in which the shear stress is independent of shear rate2,20. Shearing bars (blades) on both top and bottom walls are also created to mimic the configuration of the realistic shear cell and prevent the sliding of particles. Simulations run sufficient time steps until the shear process reaches a steady state and the stresses are nearly constant with small fluctuations. The particle properties in the simulations are given in Table 1.