The escape of pedestrians in the subway station is hampered by floods created by heavy rain. In order to explore the critical flood level in a subway station so that pedestrians can escape safely, the case study of the Mingxiu Road subway station in Nanning, China, was conducted using numerical simulation techniques. In total, 30 groups of sample pedestrians with different walking speeds and numbers were randomly generated by the Monte Carlo method, and 3D simulation software was used for escape simulation. The simulated escape data were put into the SVM model, and the maximum pedestrian capacity and minimum speed of pedestrians were solved successfully with different conditions of the Mingxiu Road subway station. Then, a 1:1 contour model of the pedestrian was constructed to simulate the flood resistance of the pedestrian escaping at the minimum speed. The flood resistance and the friction force between the pedestrian and the ground were compared to calculate the critical escape flood level height, and the critical escape flood level height of an adult, child, and elder was 87.4 cm,75.5 cm, and 83.0 cm, respectively.

With the rapid development of urbanization and infrastructure construction, the requirements for the foundation design of high-rise buildings and large bridges are increasing. Pile foundations, as important supporting structures, are widely used in weak foundations and high-rise buildings. However, pile groups show significant advantages in bearing capacity, settlement control, and structural stability, while also bringing complex pile–soil interactions and group pile effects. Based on an FLAC3D numerical simulation (version 3.0), this paper constructs a pile group composite foundation model under different pile length conditions and analyzes the influence of pile–soil interaction on the group pile effect. The results show that pile length has a significant impact on the settlement and bearing capacity of the pile group composite foundation. When the pile length exceeds a certain critical value (23.4 m in this study), the interaction between piles is enhanced, the bearing capacity of the soil between piles is improved, the pile–soil stress ratio is reduced, and the overall settlement is effectively controlled. Moreover, there are obvious differences in settlement and stress distribution between pile group composite foundations and single-pile composite foundations, and the group pile effect can lead to greater settlement and more complex stress distribution. Therefore, when designing pile group composite foundations, factors such as pile length, pile spacing, and geological conditions should be fully considered to optimize foundation performance. This study provides a theoretical basis and reference for the design and optimization of pile group composite foundations, highlighting the importance of considering pile length and pile–soil interaction in practical engineering applications.

Accurate joint roughness coefficient (JRC) estimation is crucial for understanding rock mass mechanical behavior, yet existing predictive models show limitations in capturing complex morphological characteristics of geological surfaces. This study developed an advanced hybrid ensemble learning methodology (HELIOS-Stack) to enhance JRC prediction accuracy by integrating multiple machine learning models and statistical analysis techniques. The research implemented a hybrid ensemble approach combining random forest regression, XGBoost, LightGBM, support vector regression, multilayer perceptron models, and meta-learner using LightGBM as the final estimator. The study analyzed 112 rock samples using eight statistical parameters. Model performance was evaluated against 12 empirical regression models using comprehensive statistical metrics. HELIOS-Stack achieved exceptional accuracy with R² values of 0.9884 (training) and 0.9769 (testing), significantly outperforming traditional empirical models and alternative machine learning models. Also, the HELIOS-Stack statistical evaluation demonstrated superior performance across multiple metrics, including mean absolute error (training: 1.0165, testing: 1.4097) and concordance index (training: 0.99, testing: 0.987). The analysis identified three distinct roughness clusters: high (JRC 16–20), moderate (JRC 8–15), and smooth (JRC 0.4–7). The HELIOS-Stack methodology significantly advances rock discontinuity characterization, establishing a new benchmark for geological surface analysis. This innovative approach offers transformative applications in geotechnical engineering, rock mass stability assessment, and geological modeling through its unprecedented precision in JRC prediction.

Previous joint test studies have mainly been conducted under the condition of uniformly distributed loads, but in engineering, overlying loads are often non-uniformly distributed. It is necessary to investigate the shear mechanical properties of joints under non-uniform loads. This paper establishes three typical mechanical models: far-field concentration (FFC), near-field concentration (NFC), and focus on center (FOC). Direct shear test simulations using the DEM software PFC reveal that the location of load concentration affects main shear parameters. The closer the load concentration is to the far end along the shear direction, the greater the deformation of rock mass during failure is, the higher the proportion of shear cracks is, and the greater the strength and energy required for failure are. Compared to a uniformly distributed load (UD) condition, FOC shows similar mechanical properties; NFC yields inferior outcomes, while FFC results in superior mechanical performance compared to UD. By utilizing machine learning, five prediction models for non-uniform load shear strength are developed, with the Genetic-XGBoost algorithm demonstrating the highest accuracy. Weight calculation results indicate that load distribution form is the most critical factor influencing shear strength.

Structural planes widely developed in slope rock masses are key geological elements governing deformation, failure modes and engineering stability. Traditional manual logging suffers from low efficiency, high safety risks and inadequate data integrity, failing to meet large-scale and refined survey needs. This paper proposes a cross-modal collaborative recognition system for slope discontinuities. The principal methodological contribution is the cross-modal ROI-guidance mechanism itself: 2D detection bounding boxes are back-projected through pixel-to-point-cloud registration to construct region-of-interest constraints in 3D space, transforming intractable global blind-search segmentation into localized oriented analysis within bounded volumes—to the best of the authors’ knowledge, the first systematic establishment of such a “visual detection → ROI-guided 3D analysis” framework for slope discontinuity characterization. Within this paradigm, established modules are adapted to the discontinuity recognition task rather than newly invented: channel attention, bidirectional multi-scale fusion and angle-aware regression are integrated into the detection backbone to address the weak texture contrast, large-scale span and extreme aspect-ratio morphology of discontinuity targets, while a PCA–DBSCAN–RANSAC cascade operating within the ROI volumes extracts dip direction, dip angle, spacing and trace length. Validated on two typical slopes in Hunan Province, the improved network achieves a mAP@0.5 of 89.4%, the average IoU of point cloud segmentation is 82.6–86.3%, the dip angle RMSE is 2.46° and the spacing average relative error is 6.8%. The full workflow takes about 86 min, a 19.5-fold efficiency gain over manual methods, and provides an automated pipeline from heterogeneous remote sensing data to engineering-usable structural parameters. The resulting outputs are organized in a tabular schema compatible with mainstream discrete-element software such as 3DEC and UDEC, where they serve as geometric inputs to downstream stability modelling once site-specific mechanical calibration is performed. The two-site validation reported here should accordingly be read as a proof of operational feasibility within the limestone and sandstone–mudstone envelope examined, with broader deployment to other lithologies identified as the natural next phase of evaluation.