ABSTRACT This study explores the predictive performance of two ensemble learning models, Adaptive Boosting (AdaBoost) and Gradient Boosting (GBoost), optimised using Bayesian Optimization (BO), for estimating the bearing pressure (P) of spread foundations on clayey soil. A dataset comprising plate load tests (PLTs) with 576 data points across 16 literatures was utilised for model development. Input parameters included settlement (S), undrained cohesion (Cu), plate width (B), footing embedment depth (Df), depth to groundwater table (Dw), and soil unit weight (γt). GBoost-BOA outperformed AdaBoost-BOA and standalone models, and Bayesian Optimization played a crucial role in mitigating overfitting and enhancing model generalisability by systematically selecting optimal hyperparameters (number of estimators, maximum depth, a learning rate and maximum features). The implementation was carried out using Python in a Jupyter Notebook environment. To further address overfitting, a diverse dataset was used to capture complex patterns and variations, ensuring reliable and transferable predictions across varied soil conditions. The consistency of performance metrics across training and testing datasets highlights the model’s robustness and adaptability. The training and testing of diverse dataset performances of the optimal hybrid model is benchmarked against an Artificial Neural Network (ANN) model used in previous study.

ABSTRACT Microbially Induced Calcite Precipitation (MICP) has emerged as an ecologically friendly geotechnical technology widely applied for soil reinforcement. However, few researches on rock joint reinforcement by using the MICP method have been reported. In this study, artificial rock joint specimens with the morphology characteristics of the tenth Barton’s standard profile were prepared by using cement slurry rock-like material. Both MICP and cement slurry grouting methods were used to reinforce the artificial rock joint specimens, and direct shear tests under three levels of normal stress were performed to investigate the shear strength behaviours. The results reveal that both the peak and residual shear strength of MICP-treated rock joint increase with the specimen curing time as a result of the increasing bonding strength of the calcium carbonate layer generated during the MICP process, and the MICP method is more effective than the cement slurry grouting method to enhance the shear strength of rock joint. However, both the shear failure of both MICP-treated and cement-slurry-treated rock joint firstly occurs along the interface between rock surface and the infill layer, and then the asperities are involved to form a mixture contact of rock-infill and rock-rock during the subsequent shear process.

ABSTRACT Two-dimensional soil consolidation involves coupled vertical and horizontal drainage, posing challenges in accurately predicting consolidation coefficients due to complex boundary conditions and data limitations in real-world applications. This study addresses these issues by evaluating data-driven physics-informed neural networks (PINNs) for solving forward and inverse problems in 2D consolidation with coupled consolidation coefficients. The methodology integrates hard constraints for drainage boundaries and soft constraints for both drainage and impermeable boundaries, enabling precise modelling of excess pore pressure and coefficient interactions. Forward problems were analysed with and without observed data, optimising hyperparameters and assessing sensor arrangement effects on predictions. Inverse problems utilised observed data to estimate vertical and horizontal consolidation coefficients and incorporated noise to simulate practical scenarios. Results show that PINNs accurately predict the full consolidation process, determining coefficients from initial-period data alone and robustly handling noise levels typical in field measurements. This work contributes a novel, efficient PINN-based approach for consolidation analysis with coupled coefficients, offering practical advantages for geotechnical engineering where long-term data is scarce and measurements are error-prone.

ABSTRACT This study evaluates soil nailing design methods using a 3D finite element model (FEM) and data from an instrumented vertical excavation. Eight methods were analysed: four based on Rankine’s and Coulomb’s Lateral Earth Pressure Methods (LEPM and Modified Coulomb’s), three regional methods from the 1980 s (German, Davis, and French), and a Limit Equilibrium Method (LEM) using software. The methods were compared in terms of (i) critical slip surface; (ii) Factor of Safety (FoS); (iii) nail tensile loads; and (iv) location of maximum load in the nails. The FEM model was calibrated using field monitoring data and served as a benchmark. Results showed that the LEM software and French method best matched the FEM model regarding slip surface shape and FoS, with differences under 5%. For tensile loads and their distribution along the depth, one LEPM method aligned best with FEM results. The remaining LEPM approaches, along with Modified Coulomb’s, Davis, and German methods, showed only limited agreement with the FEM results.

ABSTRACT Intense rainfall is a leading cause of failures in deep excavations and adjacent shallowly buried pipelines, underscoring the critical need to evaluate their adverse impacts. To calculate settlements of pipelines adjacent to anchored deep excavations under rainfall, an integrated analytical solution was developed combining the two-stage method, the Multivariate Adaptive Regression Splines (MARS) algorithm and the stratified Green-Ampt (SGA) model. After validation against field data, systematic parametric analyses were conducted. Results showed that the pipeline’s maximum settlement ( u z , max ) exhibited a decreasing trend with increasing burial depth, bending stiffness and strata elastic modulus. u z , max exhibited a non-monotonic trend with the pipeline-to-excavation distance, first increasing and then decreasing; its peak value occurred when the pipelines were at a distance of 0.3–0.4 times the final excavation depth. Moreover, a 12-h, 19.4 mm/h rainfall event increased u z , max by 3.2–32%. These effects were more pronounced under conditions of a smaller maximum lateral deflection ( δ hm ). To quantify these effects, a simplified formula linking u z , max and δ hm was fitted through regression analysis. By incorporating the rainfall intensity formula, an indicator-calculation flowchart was summarised and illustrated with an example, thereby demonstrating how the method enhances the application of the current indicators in heavy rainfall scenarios.

ABSTRACT This study assesses the water durability of tuff-based mortars stabilised with clay and cement. Three formulations are tested: M1 (15% cement, 10% clay), M2 (10% cement, 15% clay), and M3 (5% cement, 20% clay). The results demonstrate a predominant influence of the clay/cement ratio. Formulation M1 exhibits the best performance, with a 28-day compressive strength of 8.15 MPa, a low capillary absorption coefficient (Cb of 9.78 g/cm2.min0.5), and minimal mass loss (7.28%) after 12 drying – wetting cycles. In contrast, mortar M3, which is richer in clay, exhibits poor compressive strength (2.97 MPa), strong capillarity (Cb = 19.24 g/cm2.min0.5), and significant degradation (20.88% mass loss). A paradox is resolved: although the literature indicates that the identified clay minerals (kaolinite/muscovite) are typically associated with low cation exchange capacity and limited swelling potential, the fine fraction of the clay induces textural swelling and destructive capillary pressures during hydraulic cycles. SEM-EDS analyses reveal a heterogeneous microstructure with impurities (Fe, K) that may intensify degradation. The durability is governed by the composite microstructure rather than mineralogy alone. Cement content greater 15% is essential to ensure the long-term stability of earth-based mortars.

ABSTRACT This study aims to introduce a new way of weighting geomechanical parameters that pose a risk to the stability of underground openings. Understanding and managing geomechanical risks play a vital role in the design and safety of underground structures. The creation of effective methods for assessing such risks is crucial, given the large degree of uncertainty present in geomechanical data. Therefore, the Criteria Importance Through Intercriteria (CRITIC) method was first used as a strategy for objectively weighting geomechanical parameters. Field data from the Ridder-Sokolny mine in Kazakhstan were used to demonstrate the method. Results showed that the internal friction angle and Rock Mass Rating (RMR) are the most influential parameters, together accounting for 40% of the total weight. This research demonstrates that the proposed method offers a replicable, data-driven approach for geomechanical risk assessment, serving as a reliable weighting strategy for underground mining, and can be integrated into design procedures without expert input, supporting its broader adoption in practice.

ABSTRACT This study investigates the life cycle environmental impact of using waste foundry sand (WFS) as backfill material for cantilever retaining walls. Life Cycle Assessment (LCA) methodology compares the environmental footprint of WFS backfill with conventional sand backfill. The openLCA software tool is utilised along with the Ecoinvent v3.8 database. A functional unit of a 1-metre-long and 3-metre-high retaining wall construction is established. The IMPACT 2002+ method assesses the environmental impact across various mid-point categories. It is found that the use of WFS and the optimal sand-WFS mixture reduces the renewable and non-renewable energy demand by 89.14% and 37.58%, respectively, and decreases the environmental impact in all 15 mid-point impact categories. Concrete production and sand mining are found to be significant contributors in all categories and cases. The cost of the construction of the wall with WFS backfill was also reduced by 63%. This study indicates that the use of WFS backfill is technically feasible, socially acceptable, and environmentally beneficial.

ABSTRACT This study presents an enhanced Nishihara model for precisely describing rock creep under three-dimensional stress, emphasising the accelerated creep phase. Unlike the classical model-which combines elastic, viscoelastic, and viscoplastic elements but poorly captures nonlinear acceleration under high stress-the revised version adds a strain-activated nonlinear Newtonian component. We derive its 3D constitutive equations and calibrate parameters using sandy mudstone creep tests. Results show the modified model accurately reproduces initial, steady-state, and accelerated creep, outperforming the original formulation. Sensitivity analysis highlights the influence of deviatoric stress and nonlinear viscosity on creep evolution. Fitting additional data from granite and sandstone confirms the model’s versatility across rock types. These improvements enhance predictive accuracy for engineering applications, including tunnel-surrounding-rock collapse risk assessment.

ABSTRACT To investigate deformation and damage mechanisms of jointed surrounding rock under transient unloading, a 2D plane-strain model (100 m × 100 m) with a central circular chamber (radius = 5 m) was established using LS-DYNA. Non-reflective boundaries and 20 MPa in-situ stress were applied, with graded mesh refinement (2–10 cm) implemented near the chamber to balance accuracy and efficiency. Validated through laboratory model experiments, the implicit-explicit-implicit method simulated excavation-induced unloading. Comparative studies revealed three critical advancements: (1) Joint length dominates deformation through stress wave trapping, where increasing length amplifies near-joint displacement by 25% per unit length and drives tensile damage connectivity from joint centres to chamber walls – a mechanism previously unreported in static analyses. (2) A critical joint spacing threshold (1.5a) regulates damage evolution, transitioning reflective damage zones from interconnected to isolated patterns as spacing increases, despite minimal displacement variation (5.3 mm). (3) Stiffness selectively controls long-joint stability, where enhancing stiffness from 0.05E to 0.25E reduces convergence displacement by 9.4% through wave energy reflection (>65%), offering a novel criterion for targeted reinforcement. These findings establish dynamic interaction principles between transient unloading and joint networks, advancing predictive models for deep excavation in fractured rock masses.