The evolution of particle migration in porous media is widely observed in various engineering applications, such as internal erosion in embankment dams and mud pumping under rail tracks. The former is usually considered a consequence of increased hydraulic gradients, while the latter is due to cyclic train loads. How particles migrate under coupled hydraulic gradient and cyclic loading remains unclear. This paper presents an experimental study on gravel-sand mixtures with two different gradations under coupled hydraulic gradient and static/cyclic loading conditions. The effects of load frequency and mean load magnitude on the progression of particle migration were quantitatively explored. The results indicated that the migration of fine particles was substantially promoted when loading changed from static to cyclic. Furthermore, an increase in the mean load magnitude and loading frequency will lead to a larger number of migrating particles, with more and larger particles migrating from the deeper layer within the soil matrix. Finally, an empirical model for predicting the accumulated mass of fine particles considering hydraulic gradient, load frequency, and load magnitude was developed and validated against the data both from this paper and in the literature, which demonstrates its satisfactory performance.
Bentonite is widely used as a buffer material in deep geological repositories because its low hydraulic conductivity and swelling pressure help limit advective radionuclide transport. Elevated temperature, however, can cause drying and shrinkage, which may alter swelling behavior and reduce mechanical integrity. Fiber reinforcement has been proposed to mitigate these effects, but the behavior of fiber-reinforced bentonite at elevated temperature remains poorly understood. This study investigates the swelling pressure, compressibility, and consolidation behavior of glass microfiber-reinforced Na-bentonite at 22 °C and 50 °C with fiber contents of 0, 1, and 3 wt
A modified triaxial test (MTT) was developed to investigate soil–structure interface (SSI) behaviour under controlled triaxial conditions. The modification integrates custom-fabricated material pieces and a specialised loading cap, enabling direct vertical shearing of the soil interface within a standard triaxial apparatus. A dedicated calibration procedure was implemented to ensure accurate determination of soil interface shear response. Comparison with the modified direct shear test (MDST) showed that the MTT produced interface friction angles comparable to those obtained from the MDST, with differences of only 0.6–4.6
Deep excavation is widely applied in large cities for underground space development, but its construction may threaten the safety of adjacent engineered structures. Efficient engineering decision making should therefore ensure the safety of both the excavation and the structures. To support this goal, the performance of adjacent engineered structures should be considered in the reliability assessment of deep excavation, whereas the lack of such consideration can mislead engineers in practice. The present paper aims to propose a novel reliability assessment framework for deep excavation with the input of the performance of adjacent engineered structures (RIPAS). In the framework, we first conduct probabilistic analysis for deep excavation with explicit consideration of causal coupling relationships among the associated indicator data, from which the reliability of the deep excavation alone can be derived. To introduce the performance of adjacent engineered structures into the reliability assessment of deep excavation, a probabilistic analysis for structures is performed in a manner similar to that used for the excavation, based on a directed acyclic graph–graph neural network (DAG–GNN)-based causal decoupling method. Subsequently, an intermediary for mapping the relationships between structural performance and excavation states is established through Bayesian bidirectional probabilistic learning. Based on the learned relationships, an inverse probabilistic analysis of excavation states given acceptable structural performance is implemented via Bayesian inference. Together, these steps systematically constitute the proposed RIPAS, and its efficiency is verified through a real-world deep excavation project of No. 2 Xuzhou Metro. Neglecting the performance of adjacent engineered structures can overestimate the assessed reliability of deep excavation by more than 500
Geotextile tube dam technology has increasingly been used in tailings dam construction because of its construction efficiency, material adaptability, and potential for fine tailings containment. However, the layered arrangement of geotextile tubes introduces pronounced interface-controlled behaviour, and the tube interfaces may deviate from a horizontal configuration during construction, making the failure mechanism and stability assessment more complex. This study investigates the stability of geotextile tube tailings dams with inclined stacking through reduced-scale model tests and theoretical analysis. The model tests reveal a composite failure mechanism consisting of arc-shaped shear within the tailings and interfacial sliding along geotextile tube layers. Outward interface inclination promotes upward migration of the slip-surface exit, intensifies deformation localisation along tube interfaces, and reduces the bearing capacity, whereas inward inclination enhances lateral confinement and produces a more contained failure mode. Three analytical methods based on the composite slip surface were compared: the rigid geotextile tube surcharge method, the upper-bound limit analysis method, and the limit equilibrium method. Among them, the limit equilibrium method was selected for its balanced accuracy and practicality and was used as the basis for developing an oblique horizontal slice method, in which the slices are aligned with the inclined geotextile tube interfaces. The predicted composite slip surfaces and safety factors agreed well with the model test results, and the calculated safety factors showed a nonlinear decrease with increasing outward interface inclination.
