Scour by hydraulic plucking is a fundamental process in landscape evolution in which large, competent rock blocks are eroded from a fractured rock mass by flowing water. This process also affects engineered structures interacting with water, such as dams and bridges, and often leads to operational and safety concerns because erosion of large volumes of material can compromise structure foundations and serviceability. To assess potential scour at a site, present methods either are empirically derived, assume a specific failure mode, or significantly simplify the geometry of potentially eroding rock particles. This limits the broader applicability of these methods and their ability to offer actionable insight into scour risk. Therefore, the discrete-element method coupled with the lattice Boltzmann method was applied to assess hydraulic plucking of fractured rock. In this approach, the three-dimensional shape of rock particles was considered explicitly, including how each particle interacts dynamically with fluid. Additionally, the highly turbulent flow conditions at which plucking often occurs were modeled using large-eddy simulation. Comparison of numerical results with scaled flume experiments show that this modeling methodology is able to capture the correct kinematic failure mode in rock block removal without restricting the potential failure mechanism, and naturally captures the governing response. This capability makes this scour assessment technique broadly applicable since site-specific characteristics can be input directly into scour risk assessments to understand the influence of local features on the plucking process.

This contribution presents an improved two-dimensional numerical model of the dynamic roller-granular subsoil interaction system that facilitates the numerical prediction of both the achieved soil compaction (“improvement depth”) and the dynamic drum response in terms of acceleration (“measurement depth”) of a specific oscillation roller. In this plane-strain model, the intergranular strain enhanced hypoplastic constitutive model captures the nonlinear inelastic behavior of the soil below the drum. The numerical simulations are performed with the Finite Element software suite ABAQUS/Standard by implementing the hypoplastic soil model using an in-house Fortran code (UMAT). Thus, the linear elastic layer applied to the soil surface (“protective foil”) proposed by the author in the original model is no longer required to ensure the numerical stability of the model. In order to study both the soil compaction and the drum response, the soil layer to be compacted is modeled with linearly increasing thickness resting on a fully compacted subsoil and a loose subsoil, respectively. The effect of a roller pass at standard excitation frequency on an initially loose soil is investigated for selected roller speeds in terms of the reduction of the void ratio. The influence of the predicted soil compaction on the drum response is simultaneously analyzed in the time and frequency domain in terms of the drum center acceleration. In addition, an experimentally found Continuous Compaction Control (CCC) parameter for dynamic rollers with an oscillatory drum is evaluated. It is shown that the developed model qualitatively predicts the fundamental response characteristics of the interacting oscillation-subsoil system observed in field tests. Moreover, the numerical model is capable of predicting the depth of influence of the selected oscillation roller.

Solving optimization and inverse problems are critical for engineering design and analysis. Despite the computational power, calculating derivatives, i.e., evaluating the forward simulation multiple times with small perturbations, is computationally very expensive and is prone to numerical instability. Traditional forward simulations cannot be used in machine learning models for optimization, as they cannot compute gradients with respect to the input parameters (reverse mode). We propose a novel Differentiable Programming simulator that combines automatic reverse differentiation with a second-order gradient-based optimization algorithm, such as L-BFGS, to develop a fully-differentiable Material Point Method (MPM) simulator. Using the differentiable simulator, we can identify the input material properties by iteratively updating the input parameters by minimizing a loss function. Typically the loss function is the norm of the difference to the target observation. We then minimize this gradient (loss corresponding to the input parameters) using a second-order optimization algorithm. The differentiable MPM is a novel tool to provide gradients through a simulator, which can be coupled with existing machine learning algorithms to generate real-time decisions and optimization in robotics.

We exploit physics-embedded differentiable graph network simulators (GNS) to accelerate particulate and fluid simulations and solve challenging forward and inverse problems. GNS represents the domain as graphs with particles as nodes and learned interactions as edges. GNS allows learning localized physics compared to global dynamics, improving generalization. GNS achieves over 500x speedup for granular flow prediction compared to parallel CPU simulations. The differentiable GNS enables solving inverse problems through automatic differentiation, identifying material parameters that result in target runout distances. Granular column collapse experiments demonstrate friction angle inversion based on final runout profiles. Additionally, we also derive material parameters and loads from videos by constructing 3D point clouds using Neural Radiance Fields. We then iteratively minimize diffGNS to derive material properties and loading conditions.

Swelling clays are found all over the world. In the presence of water, these clays swell and exert swelling pressure when constrained. These clays cause enormous damage to the infrastructure of the order of 20 billion dollars annually in the United States. These clays are also used as barrier materials in environmental engineering, drilling muds in petroleum extractions, and drug delivery carriers in the pharmaceutical industry. The predominant clay minerals in swelling clays are typically the Smectite group of clay minerals, with montmorillonite clay being the most common. We have developed a multiscale approach that includes molecular dynamics (MD), steered molecular dynamics, and discrete element modeling tightly integrated with experiments at various length scales to elucidate the fundamental mechanisms that influence macroscale properties of swelling clays that include swelling, swelling pressure, permeability, compressibility, shear strength, etc. The molecular models of Na-montmorillonite clay developed in our studies mimic SWy-2 from the Clay Minerals Society and correspond to the clay used in the experiments. The MD simulations of the hierarchical elements of clay, including the dual-stacked clay sheets, the clay tactoids, and clay aggregates, are used to evaluate the mechanical response of the clay with various levels of hydration, and the relationships between binding interaction energies, interlayer swelling, and exfoliation. Long MD simulations evaluate fluid flow into the interlayer and identify the associated mechanisms. Discrete element modeling simulations describe the role of particle breakdown on swelling and swelling pressure. Molecular dynamics simulations of clay with fluids with a wide range of dielectric constants and experiments further illustrate the crucial role of molecular interactions between clay and fluids on the macroscale properties. Our results indicate that the molecular interactions alter the clay microstructure, the molecular interactions between the clay and fluids, which in turn influence the macroscale properties of swelling clays.

The sharp crack topology can be regularized into the diffusive crack topology, resulting in an outstanding challenge in computational contact mechanics. But in the realistic scenario, contact with normal and tangential forces are ubiquitous in geological discontinuous structure. Hence, it (is) meaningful to pixelized interfaces in (the) voxel-based models for modeling compressive failure process of (a) rock mass. In this study, we propose a phase-field model for frictional fracture phenomena in rock masses. Our model has two novel features: (i) a level-set function characterized by the phase-field variable for interface; (ii) a Drucker–Prager type plastic criteria for stick and slip state. It is worth noting that the interfacial indicator function proposed in this study will be updated with the phase-field evolution to reflect topology changes due to crack propagation. We assume a linear elastic (no-tension) response for the relationship of the diffused interface in the normal direction. In the case of tangential contact, the stick and slip distinction is done by a slip condition, which is analogous to the yield condition in plasticity theory. In addition, an alternative splitting strategy of energy takes an analogous expression with Helmholtz free energy in continuum breakage mechanics is proposed. It continuously interpolates the damaged bulk and residual interface systems so that enables us to model the contact behaviors in the regularized interface region. We also provide a robust numerical solution strategy to treat the spatiotemporal evolution of frictional damage and sliding. Our model is validated by five benchmark problems, and the numerical simulations are compared with some published data. We proceed to apply this model to study the complex failure mechanism of crack initiation and propagation in CCBD specimens with various boundary conditions, where the effects of confining pressure and friction are discussed.