Landslides frequently occur on cutting slopes with weak rock during the rainy season. An understanding of the failure mechanism is essential for designers to adopt effective measures to mitigate potential impacts caused by landslides. Based on an enormous number of highway slope practices in Guangdong Province, China, this paper presents a whole train of numerical investigations on the hydraulic response and stability of weak rock slopes subject to a subtropical climate where the average annual rainfall is approximately 1,800–2,000 mm. This study explored a different failure mechanism of three different types of weak rock slopes in a modeling approach. A physics-based slope model coupled with a hydrological model is used to simulate the factor of safety and porewater pressure development of unsaturated slopes with various rock masses under diverse rainfall scenarios. The input parameters were determined from the rainfall intensity-duration-frequency (IDF) curves of regional hydrological conditions to comprehensively consider the impact of rainfall. The general suction range of rock masses was acquired using three hydraulic characteristic curves associated with sandstone, siltstone, and mudstone. Furthermore, the failure models of weak rock slopes under different rainfall conditions were assessed. Based on the theory of regional hydrology, a unified rainfall and intensity-duration threshold for different weak rock slopes was established. The results indicate that both siltstone and mudstone slopes are more sensitive to long-term rainfall than short-term rainstorms. However, the sandstone slope appears to be unstable under regional rainfall conditions of 40-year return periods. This also proves that the instability mode of the siltstone slope exhibits a retrogressive type in which the sliding surface of the slope increases with an increase in rainfall duration. Translational shallow landslides occurred in the transient saturation zone of the sandstone slope, and the depth of the failure plane was independent of rainfall patterns and duration. This study revealed the instability mechanism and provided a robust early warning measure for weak rock slopes by fusing engineering geology and regional hydraulics, providing insightful information for mitigating potential landslide disasters. The proposed model can also be used to predict the possible infiltration depth of rainfall and the spatial location of the sliding surface, prior to construction to assess whether slope stabilization and reinforcement measures are needed.

The subgrade performance assessment and targeted maintenance of a highway during operation is very important and challenging. This paper focuses on the performance of the whole life-cycle of a highway subgrade during the operational period. Four roads with different traffic volume and geological conditions were selected; 20 test sections of these 4 roads were examined for a three-year distress survey, and 18 specific subgrade distresses of the 5 assessment objects were tracked and collected. First, based on the analytic hierarchy process (AHP), the subgrade performance of the selected section is evaluated, and the subgrade performance index (SPI) at different time periods is obtained. Then, based on the internal and external factors which affect the subgrade, three algorithms to determine the optimal support vector machine (SVM) model were proposed to train and predict the SPI. The results show that the SPI predicted results based on the data time series and particle swarm optimization–least squares SVM (PSO–LSSVM) model are better than those based on grid search (Grid-SVM) and genetic algorithm (GA-SVM) models. Finally, this paper provides a detailed idea for the rational layout of subgrade life-cycle assessment and decision-making by establishing a subgrade performance assessment–prediction–maintenance–management architecture system.

Due to their proximity to track structures, over-track buildings in depot areas may experience adverse vibrations from train operations, which can negatively impact residential comfort and the performance of precision equipment. As the test line typically involves the highest train speeds within the depot and features distinct operating conditions such as constant-speed cruising, accelerating, and braking, this study aims to investigate the influence of different train operation modes on the vibration excitation of over-track high-rise buildings through field measurements at both the central and end sections of the test line. Three operational modes were examined, including cruising, accelerating, and braking. Vibration signals were systematically collected using sensors installed in the free field, on the ground above the building foundation, and at the bases of the shear walls. A time-varying frequency phenomenon was observed at measurement points on the ground above the foundation and at the shear wall bases during train acceleration and braking. Field data indicate that vertical vibrations in the free field at the end section of the test line were significantly greater than those at the middle section, whereas vertical vibrations at the shear wall bases were comparable between the two sections. Notable vibration transmission loss occurs as the vibration propagates into the building structure, with peak values reaching 20 dB at 80 Hz during cruising operations and 35 dB at 125 Hz during acceleration and braking. This study offers valuable insights for assessing train-induced vibration transmission into high-rise over-track buildings.

Accurate estimation of the geological strength index (GSI) is essential for rock mass characterization in tunnelling and underground engineering, yet remains challenging under datalimited conditions due to multiple sources of uncertainty, including parameter uncertainty, modelform uncertainty, and input variability. This paper proposes a Bayesian multimodel probabilistic inference (BMPI) framework to estimate GSI from sparse data. The framework integrates twelve candidate deterministic models derived from the RMR, Qsystem, RMi, and their hybrid formulations. Parameter uncertainty is quantified via Markov Chain Monte Carlo (MCMC) simulation, which draws posterior samples of model coefficients. A Dirichletbased model averaging scheme is then employed to synthesise predictions across all candidate models, explicitly accounting for modelform uncertainty. The probabilistic characteristics of GSI predictions are further evaluated under different input distributions generated from fitted models of the seven input parameters. The proposed methodology is demonstrated on a realworld case study of argillaceous clastic rock from the Three Gorges Reservoir region, China, where only 67 field measurements are available. Results show that BMPI substantially outperforms the best single deterministic model (M7), increasing the coefficient of determination (R2) from 0.79 to 0.89 on the training set and from 0.74 to 0.85 on the validation set, while improving the coverage proportion of the 95% credible interval (CP95) from 0.90 to 0.94 (training) and from 0.86 to 0.91 (validation). Moreover, the BMPI framework produces the widest and smoothest predictive distributions among eight compared models, closely matching the empirical distribution of the actual GSI data. It also exhibits interpretable uncertainty behaviour by widening credible intervals in sparsely sampled regions of the input space, thereby avoiding overconfident predictions. The proposed BMPI framework provides a rigorous, uncertaintyaware tool for GSI estimation under sparse data conditions, with promising applications in slope stability assessment, tunnel support design, and underground excavation planning.