X-SEL: a stacked ensemble learning architecture for forecasting soil slide displacement in flume simulations

Physical and data hybrid-driven modeling for the evolution state prediction of reservoir colluvial landslides with step-like movement patterns: a case study in the Xiluodu Reservoir area, China

A Physics-Informed Data-Driven Soiling Prediction Model for Heliostat Fields in High-Altitude CSP Plant

Soil Moisture Prediction Based on LSTM with a Sliding Window Strategy

: Soil moisture is a key indicator of land surface hydrological processes and plays an essential role in agricultural irrigation management and ecological monitoring. However, traditional monitoring approaches mainly rely on fixed sensors and lack the capability to predict future soil moisture dynamics. To address this issue, this study proposes a soil moisture time series prediction method that integrates a sliding window strategy with a Long Short-Term Memory (LSTM) network. Daily soil moisture data from a single observation station are preprocessed through data cleaning and min – max normalization, and then converted into supervised learning samples using a 14-day sliding window. The LSTM model is constructed to capture nonlinear temporal dependencies in the time series, and its performance is evaluated using RMSE and MAE. Experimental results show that the proposed model achieves an RMSE of 0.0476 and an MAE of 0.0369 on the validation set, demonstrating satisfactory prediction accuracy and stability.

Ensemble Machine-Learning-Based Horizontal Permeability Prediction and Hydraulic Flow Unit Characterization of the Hugin Sandstone

Understanding permeability is essential for evaluating reservoir quality and field development planning. Reliable permeability estimation can reduce the uncertainty in reservoir characterization, particularly in intervals where core data are limited. As the industry relies on log-based interpretations and empirical correlations, the limitations of these approaches become apparent. Data-driven approaches offer a promising alternative to conventional empirical methods. The data set in this study comprises 252 samples with seven features derived from conventional well logs. Data preprocessing includes handling missing values, smoothing logs, feature engineering to add an extra input, and transformation with the Yeo-Johnson technique. A center moving average filter was used to reduce variance and improve data consistency. Ensemble machine-learning (ML) and baseline models were developed and evaluated using a 75-25 train-test split, four-fold cross validation, and model complexity assessment. Ensemble methods outperformed baseline models, with extremely randomized trees (ET) and random forest (RF) emerging as the most stable, achieving a mean R² of 0.93 and 0.89 and a low R² standard deviation (0.3). Multilinear regression (MLR) and artificial neural networks (ANNs) show limited accuracy, while gradient boosting (GB) and extreme gradient boosting (XGBoost) methods exhibit overfitting despite perfect training scores. Predicted kh values were compared with core data. Both linear and nonlinear empirical equations were derived using MLR, polynomial regression, and a power-law model. The power-law model (empirical equation) achieved an R² value of 0.79 and can therefore be used to estimate permeability. Additionally, a Gaussian mixture model (GMM) was used for unsupervised classification of hydraulic flow units (HFU) using the flow zone indicator (FZI), computed from the continuous permeability curve obtained from the best ML model. Thus, ML-based permeability prediction is an indispensable component of HFU modeling. The model identified three distinct flow zones, enabling HFU clustering and defining their corresponding petrophysical properties and depositional environments.

Machine Learning Prediction of Soil‐Water Characteristic Curve and Its Application in Lateral Earth Pressure Calculation of Retaining Walls

The soil‐water characteristic curve (SWCC) is a fundamental parameter that governs the hydro‐mechanical behavior of unsaturated soils. Conventional laboratory measurement of SWCC is time‐consuming and labor‐intensive, while traditional lateral earth pressure design for retaining walls frequently relies on the saturated soil assumption, neglecting the effects of SWCC and resulting in significant systematic deviations in calculations. This study develops a statistically rigorous machine learning (ML) framework for efficient SWCC prediction and its application to lateral earth pressure calculations for pile‐supported box counterfort retaining walls. Four ML algorithms with distinct methodological frameworks were employed: extreme learning machine (ELM), least squares support vector machine (LSSVM), projection pursuit regression (PPR), and Bayesian ridge regression (BRR). These algorithms were utilized to construct SWCC prediction models using the cleaned UNSODA database. Model performance was assessed through multi‐metric evaluation, paired t ‐tests for statistical significance, and robustness analysis involving 30 independent runs, with validation conducted on measured silty clay data across 12 suction levels. Results indicate that the ELM model achieves the highest prediction accuracy, demonstrating statistically significant superiority over LSSVM, PPR, and BRR and excellent robustness. Independent validation reveals an average relative error of only 2.25% for ELM‐predicted SWCC. The SWCC‐based earth pressure calculation rectifies the bidirectional deviations of the traditional saturated method and identifies a neutral point at a depth of 17.5 m for a 25 m‐high retaining wall. This study offers a reliable technical approach for rapid SWCC acquisition and refined lateral earth pressure design for retaining structures.

An AI-Driven Framework for State-Dependent Constitutive Modeling of Soils Integrating Particle Swarm Optimization and Machine Learning

Accurately predicting soil stress-strain behavior remains challenging due to the nonlinear, path-dependent, and evolving nature of soil properties. This study develops an AI-driven State-Dependent modeling framework (AI-D-SD-F) that integrates Particle Swarm Optimization (PSO) and Machine Learning (ML) for real-time parameter evolution and adaptive stress-strain simulation. PSO dynamically calibrates plastic potential parameters under varying stress states, while Gaussian Process Regression (GPR) and Broad Learning System (BLS) models establish nonlinear mappings among stress paths, hardening variables, and initial conditions to enable data-driven parameter updating. An adaptive strain-step implicit algorithm further improves the stability of nonlinear stress-return computations. Validation using triaxial tests on Hangzhou clay shows that the framework improves prediction accuracy by more than 32% compared with the Tsinghua and Modified Cam-Clay models. Engineering-scale simulations of shallow foundation failure exhibit deviations within 5% of Terzaghi’s local bearing capacity, confirming strong predictive reliability. The proposed framework provides a unified, data-enhanced foundation for adaptive constitutive modeling, improving both numerical robustness and accuracy in geotechnical analyses.