Reservoir landslides pose significant risks to hydropower projects, potentially leading to catastrophic disasters that threaten downstream lives and properties. Landslide susceptibility assessments are critical for effective regional disaster prevention and mitigation. However, the complexity, model uninterpretability, and data scarcity related to reservoir landslides, particularly when adapting models across diverse geographic regions, present significant challenges. This study proposes an interpretable Deep Transfer Learning model coupled with multi-source data and Physical methods (DTLP). The model is trained on multi-source data from the Three Gorges Reservoir Area (TGRA) and Lower Jinsha River Basin (LJRB), tested in Baihetan Reservoir Area (BHT), addressing the issues of limited data and cross-regional generalization. The physical method captures the effect of dynamic water level changes on slope stability. SHAP values are used to interpret the model, providing clear insights into its internal mechanisms. Results demonstrate that DTLP outperforms TrAdaBoost in data-scarce regions, achieving higher accuracy (AUC=0.953, Accuracy=0.941) with better feature generalization and susceptibility zone identification. Incorporating dynamic water level changes into the physical model enhances identification of high-susceptibility areas and reduces misclassifications. SHAP analysis indicates that elevation, lithology, and distance to river significantly influence the model decisions. Using TGRA as the source domain further validates the superiority of DTLP framework. However, due to the initial discrepancies between TGRA and the target domain, the transferability is constrained to some extent, resulting in models trained on LJRB data outperforming those trained on TGRA data.
Modeling the lateral behavior of piles in clay remains a significant challenge because of the complexity of soil-structure interactions. In recent years, machine learning (ML) has emerged as a powerful tool for tackling nonlinear problems in pile-soil interaction by leveraging the high-dimensional characteristics inherent in experimental and field-measured geotechnical datasets. However, as the number and diversity of input parameters grow, the interpretability of these models diminishes. To address this limitation, this study develops an interpretable ML-based model for predicting p-y curves of monopile foundations. The proposed approach integrates eXtreme Gradient Boosting (XGBoost) with Gaussian Process Regression (GPR) and is trained on a comprehensive dataset compiled from existing tests. The ML model demonstrates agreement with field measurements, where the predicted lateral response of monopile foundations derived from p-y curve methodology matches well with the benchmark results documented in the PISA (Pile Soil Analysis) project. This validation confirms the robustness and practical applicability of the model. Moreover, to enhance interpretability, Local Interpretable Model-Agnostic Explanations (LIME) are employed to quantify the influence of individual input variables on the predictions, providing deeper insights into the prediction-making process of the model.
Dust accumulation on solar panels markedly diminishes their energy conversion efficiency, especially in dry and semi-arid regions such as Saudi Arabia. Conventional maintenance techniques, including manual cleaning and fixed-interval schedules, frequently exhibit inefficiency, high costs, and environmental unsustainability. This paper presents a machine learning approach for real-time dust level detection employing a Random Forest Classifier, trained on a synthetically created dataset that simulates various climatic and operational situations in Dammam, Saudi Arabia. A DJI Matrice 300 RTK drone was utilized to implement this model in practical situations, facilitating airborne, real-time observation of dust deposition on photovoltaic (PV) modules. This UAV-enabled system markedly improves scalability and diminishes the requirement for terrestrial labour. The technology utilizes the normalized voltage-to-irradiance ratio to categorize dust levels and activate automatic cleaning processes accordingly. The results indicate a classification accuracy of 91.3
Floating multi-module (FMM) systems, like floating offshore photovoltaic and floating city, provide an innovative solution to challenges posed by resource scarcity and population growth. These systems consist of multiple identical floating modules, interconnected by connectors and anchored by a mooring system. Connectors, as critical components, are vulnerable to failure due to factors such as extreme and cyclic loads. Such failures can result in accidents like module overturns, collisions, and mooring line failures, which threaten the safety of engineering structures. Therefore, a numerical model of an FMM/connector/mooring system was established, based on 3-D potential flow and multi-body dynamic theory, to investigate the effects of connector failure on system safety. A model test was conducted to validate the accuracy of the developed numerical model. The numerical results show that the maximum surge, sway, and yaw motions are notably increased by connector failure, particularly in the weather-side modules connected to the failed connector. Furthermore, the modal characteristics of the system are markedly changed after failure, with the amplitude curves for surge, sway, and yaw exhibiting multiple peaks, increasing the resonance risk of engineering structures. Meanwhile, due to the connector failure, the horizontal motion trajectories of the modules are transformed from straight lines into complex patterns, such as ellipses or 'infinity'-shapes. Finally, the maximum tensions in the remaining connectors positioned opposite the failed connector are considerably higher than those observed under intact conditions, increasing the risk of cascade connector failures and consequently endangering the safety of engineering structures.
Due to their advantageous biocompatibility and adjustable degradation rates, zinc alloys are increasingly recognized as potential materials for biodegradable implants. Nevertheless, the suboptimal mechanical properties of traditional Zn-Mg alloys pose a significant limitation to their broader clinical use. This study proposes a dual optimization strategy that integrates rare earth neodymium (Nd) alloying with room-temperature rolling, systematically exploring the effects of varying rolling reductions (50 %, 60 %, 70 %, 80 %) on the microstructureproperty relationships in Zn-1Mg-0.3Nd alloys. With increasing rolling reduction, the synergistic evolution of eutectic structures and NdZn11 phases intensifies basal texture development, leading to exceptional mechanical properties at 80 % reduction: tensile yield strength of 247.9 +/- 2.5 MPa, ultimate tensile strength of 339.1 +/- 1.9 MPa, and fracture elongation of 13.8 +/- 0.13 %. Mechanistic analysis reveals that dislocation pinning by NdZn11 contributes to strength elevation, while their microcrack strain-accommodation capability and multi-slip system coordination effectively mitigate strain softening. These findings elucidate the pivotal role of rare earth element Nd in toughening Zn alloys through thermomechanical processing, providing critical insights for developing next-generation biodegradable metallic implants.
