In this work, an advanced hypoplastic model incorporating the critical state is integrated into smoothed particle hydrodynamics (SPH) for the numerical investigation of geotechnical large deformation problems. Besides, a return mapping strategy is proposed to regulate the stress state to avoid calculation failure. To this end, a closed-form solution to predicting the failure surface implicitly incorporated in the hypoplastic model is derived. The proposed SPH method is examined with element tests such as the oedometer test, simple shear and biaxial compression test, from which good performance of this SPH model is observed. Finally, the SPH model is applied to numerically reproduce the sand column collapse and rigid footing problem to investigate the effect of the initial void ratio on soils' behavior.

MiniFrac tests, Diagnostic Fracture Injection Tests (DFIT), and their revisions like DFIT flowback tests are commonly used in the field to determine in-situ stress and other rock properties such as permeability. However, there have been ongoing debates over the interpretation methods for DFIT, primarily driven by the simplifying assumption of uniform fracture closure. Despite significant research on fracture propagation, there have been limited advancements in understanding fracture closure mechanisms. Fractures may close uniformly from tip to mouth or may start receding from the tip or closing at the mouth (near the wellbore). Depending on the rock permeability, fracture conductivity, and other factors, each of these closure modes can occur, and the resulting pressure signature would be significantly different. This lack of understanding has led to confusion in the interpretation of DFIT data. To address this challenge, this study proposes first a detailed analytical solution and second a fast method to describe or predict fracture closure modes. Inspired by field data and previous literature, a scaling analysis is conducted to define dimensionless numbers that can determine different fracture closure modes under various flowback rates. Fully coupled 2D and 3D geomechanics and fluid flow models are built to validate the proposed analytical and scaling analysis. The model, similar to the real test, consists of two stages: injection and flowback. The impact of fracture roughness on fracture conductivity changes during closure is also incorporated into the model to achieve results comparable with field data. This work aims to provide a better understanding of fracture closure mechanisms and offer a practical approach to interpret DFIT data more reliably, leading to improved characterization of in-situ stresses and rock properties.

Reconstructing the microstructure of granular materials is an essential approach to understanding and elucidating granular mechanics. However, a persistent challenge has been the accurate reconstruction of arbitrary granular materials from three-dimensional (3D) X-ray micro-computed tomography (μCT) images. This study first proposes an innovative framework to bridge the reconstruction of granular materials and vision foundation models. An automatic multi-mode generator is tailored to extract two-dimensional (2D) low-quality mask maps denoting granular grains from raw μCT images, utilising vision foundation models. These 2D low-quality mask maps are subsequently transformed into high-quality ones, well-preserving interior textures and boundaries, by one three-step strategy. The stacked 2D mask maps are fed into an optimal-transport-based algorithm to achieve one accurate 3D mask map. All procedures are applied to μCT images from three orthogonal orientations of granular materials, yielding three 3D mask maps. Finally, the complete 3D mask map is acquired by combining three orthogonal mask maps. Our method is tested on four granular materials of varying sizes and shapes (including porous) and compared with the benchmark watershed segmentation. The results indicate that our method is state-of-the-art and significantly outperforms the benchmark in all metrics. Over-segmentation can be repaired automatically with our method. The performance of a foundation model in extracting 2D mask maps is not necessarily proportional to training data scaling. The proposed generic framework is robust enough to reconstruct arbitrary granular materials in different scenarios, even with high porosity or extremely irregular morphology.

Particle breakage significantly influences the mechanical properties of granular materials at both macro and micro scales, while particle shape also plays a crucial role in various simulations. In current discrete element method (DEM) approaches, clumps, which are composed of multiple overlapping spheres, have been widely adopted as an efficient technique to represent real particle shapes. However, due to the inherent nature of clumps, which do not consider internal interactions between sub-particles, they cannot be directly employed to simulate particle breakage. To address this limitation and bridge the gap between particle shape representation and breakage simulation, this study proposes a novel clump-based method for simulating particle breakage in DEM. The proposed method introduces a breakage criterion based on the statistical stress analysis of clump particles. The location of breakage initiation and propagation is determined by evaluating the stress level of the sub-spheres constituting the clump. By leveraging extensive single-particle crushing studies and summarizing the observed patterns, a breakage mode criterion that integrates three-dimensional Voronoi tessellation is proposed. The sub-spheres within the clump are utilized to capture the actual stress concentration characteristics of the particle, enabling the definition of breakage modes. The accuracy and robustness of the proposed method are validated through a comprehensive series of single-particle and multi-particle experiments, demonstrating its ability to reproduce realistic breakage patterns and force-displacement responses. Compared to existing DEM breakage methods, the proposed model better considers the local stress concentration characteristics of particles and captures the influence of multiple contact forces on continuous crushing, while the clump-based breakage algorithm efficiently balances computational efficiency. This innovative approach provides a solid foundation for future research on breakage simulations considering particle morphology, offering a powerful tool for accurately modeling particle breakage in granular materials.