Estimating the Mechanical Properties of Gravel Deposits by Integrating Field Investigation and DEM Simulation

    The gravel deposit, widely distributed in Taichung City, central Taiwan, is a composite material in which gravels and cobbles are embedded within a fine-grained soil matrix. Due to the presence of large particles ranging from 100 mm to 300 mm, the characterization of shear strength commonly relies on large-scale direct shear tests or triaxial tests. In practice, however, these measurements are costly and difficult to perform in sufficient numbers to represent the high variability of mechanical properties at the site. To address this limitation, this study proposes an integrated framework that incorporates the hardness value of the matrix, obtained from the Yamanaka soil hardness tester, and the geometric features of gravels into Discrete Element Method (DEM) simulations to estimate the shear strength of the gravel deposit.

    Specifically, three-dimensional DEM simulations were conducted using EDEM, employing the Hertz-Mindlin contact model coupled with the Bonded Particle Model (Bonding V2) to replicate a full-scale direct shear test. The gravel deposit was discretized into matrix (< 4.75 mm) and gravel (> 4.75 mm) fractions based on the field investigation results. To represent the geometric heterogeneity, image analysis was employed to characterize the gravel fraction and extract morphological parameters for the simulation. Under this classification, distinct approaches were adopted for the micro-parameters. Contact parameters were specified for each fraction, while bonding parameters for all interaction types (Matrix-Matrix, Matrix-Gravel, and Gravel-Gravel) were governed by the matrix properties. To determine the appropriate bonding and contact parameters specific to the matrix fraction, the Yamanaka soil hardness tester was utilized to bridge the gap between field conditions and numerical simulations. By employing Response Surface Methodology (RSM), a quantitative relationship between micro-parameters and penetration depths was established to identify the parameters from the field data. Subsequently, the calibrated micro-parameters corresponding to the target in-situ penetration depth were assigned to the composite model for full-scale direct shear test simulations to evaluate the shear strength. Preliminary verification confirms the feasibility of the penetration test simulation in EDEM. Furthermore, complementary uniaxial compression tests demonstrate that the calibrated bonding parameters correspond to realistic physical properties, thereby ensuring the reliability of the shear strength estimation in the full-scale DEM simulations.