Search Results (1,235)

In silico prediction of ADME parameters, Lipinski’s Rule of Five, toxicity, or the reliability level of the predicted biological activity of a given structure against a specific pathogen are an integral part of modern drug discovery to accelerate the identification of promising lead compounds while reducing development costs and toxicity risks. In this context, multivariate statistical methods such as Principal Component Analysis (PCA) play an important role in the evaluation of large chemical and biological datasets, facilitating the identification of structural similarities, physicochemical trends, or predictions of pharmacological behavior. Natural products remain one of the most important sources of structurally diverse bioactive scaffolds, serving as inspiration for the development of novel therapeutics. Buchenavianines are a rare class of piperidine-flavonoid alkaloids isolated from tropical woody species of the genus Buchenavia (Combretaceae). Scientific interest in buchenavianines was initially stimulated by the discovery of potent anti-HIV activity in extracts of Buchenavia capitata, which demonstrated their ability to interfere with viral replication pathways and highlighted their potential as antiviral drug candidates. Studies have suggested that buchenavianines may possess anti-inflammatory, antioxidant, and anticancer potential; therefore they represent important lead compounds for medicinal chemistry in the development of a new generation of drugs. PCA showed that the first two principal components explained 92.7% of the total variance (PC1 = 56%, PC2 = 36.7%), indicating that the selected descriptors adequately captured the variability among the studied buchenavianines. This study underscores the importance of these secondary metabolites as lead compounds in modern medicinal chemistry and drug discovery. Full article

Background: Composite materials are not easy to model using traditional experiments and physics-based approaches due to their diverse properties and multiscale nature. This review summarizes how artificial intelligence (AI) methodologies, including machine learning, deep learning, and generative AI, have been increasingly transforming and how they have been increasingly applied to composite materials research. It combines their possible use across the prediction of their properties, damage detection, structural health monitoring, and materials design. Methods: Searches were performed in Scopus and Google Scholar for articles published using combined keywords such as “(artificial intelligence OR machine learning OR deep learning OR Generative AI) AND (composite OR fiber reinforced) AND (prediction OR optimization OR characterization)”. The search strategy was designed to capture both fundamental developments in AI methodologies relevant to materials research and their specific applications to composite materials. The identified publications underwent a two-step selection process: (i) an initial review of titles and abstracts, and (ii) a detailed full-text evaluation. Studies were included if they provided explicit descriptions of AI models, clearly specified datasets, and applied quantitative performance assessments to composite materials-related applications such as material property prediction, damage detection, characterization, or manufacturing process optimization. For each selected article, relevant information was extracted in a structured and consistent manner, including the type of composite material, AI methodology, input parameters, dataset characteristics and size, validation methods, and application domain. These data were used to support both quantitative trend analysis and the qualitative identification of research gaps and future research opportunities in composite materials research. Results: Supervised learning methods, in particular artificial neural networks, support vector machines, random forests, and decision trees, were commonly used and exhibited durable potential accuracy in predicting properties of composite materials. Unsupervised methods such as principal component analysis (PCA) and clustering were comparatively underused. Generative AI, including generative adversarial networks (GANs) and variational autoencoders (VAEs), and physics-informed models emerged as growing approaches for synthetic data generation and improved interpretability, respectively. Limitations: The reviewed studies were frequently limited by small or heterogeneous datasets, weak generalizability testing beyond training conditions, high computational requirements and limited model interpretability (“black box” behavior); a risk-of-bias assessment across included studies was not formally conducted. Conclusions: AI is shifting composite materials research from trial-and-error toward data-driven, predictive design. Realizing its full potential will require open, well-annotated datasets, physics-aware and explainable models, and closed-loop, active-learning workflows linking prediction to experimental validation. Full article

Chest X-ray (CXR) imaging remains a widely used diagnostic modality for the detection of thoracic diseases, including pneumonia, tuberculosis, pleural effusion, lung cancer and cardiomegaly. Despite the extensive use of CXR examinations in clinical practice, the development of robust multi-label artificial intelligence (AI) systems remains constrained by limitations in dataset availability, annotation consistency and reproducible labeling methodologies. Existing public datasets frequently rely on heterogeneous labeling protocols, limited pathology coverage or annotations that inadequately account for linguistic negation and uncertainty within radiology reports. In this paper, we introduce MultiCaRe-Thorax, a curated thoracic benchmark dataset derived from the publicly available MultiCaRe repository, with a reproducible negation-aware natural language processing (NLP) framework for automated generation of multi-label thoracic disease annotations. Unlike existing CXR benchmarks that are distributed in their pre-curated form, MultiCaRe-Thorax provides a fully reproducible pipeline from raw case reports to curated dataset, establishing a new resource that is both scalable and transparent. Our pipeline performs thoracic case identification, image–report matching, duplicate removal, data cleaning, disease-specific label extraction and quality-control verification to produce a curated cohort comprising 5252 chest radiographs from 3415 unique patients. Sixteen clinically relevant thoracic pathologies were automatically annotated using a dictionary- and pattern-based NLP framework incorporating explicit negation detection and n-gram mining, achieving 95% accuracy against 500 manually reviewed reports (Cohen’s $\kappa =0.89$). Sixteen clinically relevant thoracic pathologies were automatically annotated using a dictionary- and pattern-based NLP framework incorporating explicit negation detection and n-gram mining, achieving 95% agreement against 500 manually reviewed reports. Five ImageNet-pretrained convolutional neural network (CNN) architectures (ResNet50, DenseNet121, EfficientNetB0, EfficientNetB7 and InceptionV3) with two hybrid ensemble configurations were evaluated under a patient-level data partitioning protocol. Among individual architectures, DenseNet121 attains the best AUC on tuberculosis (0.757) and pleural effusion (0.767) with roughly one-third the parameters of ResNet50, while ResNet50 attains the higher mean AUC across all sixteen pathologies (0.688 vs. 0.682) and on interstitial lung disease (0.769) and lung mass/cancer (0.777); we report this as an accuracy–efficiency trade-off rather than a single best architecture. The DenseNet121-InceptionV3 ensemble achieves the highest reported AUC of 0.821 for lung mass/cancer detection. We report 95% confidence intervals for the evaluated ensemble and remaining backbones and note substantial estimation uncertainty for the rarest classes (e.g., congenital heart disease, $n=8$ test cases). Grad-CAM visualizations were also employed to provide clinically interpretable explanations of model predictions. Full article

Groundwater level responses to solid tides in confined wells provide a means for investigating aquifer parameter variations and earthquake-related effects. Using hourly water-level observations from eight confined wells of the Beijing seismic monitoring well network (January 2017–December 2025), this study applied the Baytap-G program to extract the M2 tidal constituent contribution, amplitude, tidal factor, and phase lag, and analyzed their temporal variations and responses to earthquakes. The eight wells are classified into five groups: Group 1 (BQ, CP), stable during non-interference intervals but with permanent instrument-induced jumps; Group 2 (MF, XJ), stable background references; Group 3 (LX), reflecting genuine hydrological processes; Group 4 (DHC, XXZ), signals too weak for reliable analysis; and Group 5 (WLY), dominated by persistent instrumental disturbances. Using mean ±3σ thresholds derived from seismically quiet periods, nine persistent ≥3σ anomaly segments were identified; none coincides in time with the 14 Ms ≥ 4.0 earthquakes or the 41 coseismic water-level response events, and all can be attributed to documented instrument events. The hypothesis that earthquakes alter M2 parameters before or after their occurrence is not supported, nor is the hypothesis that coseismic water-level responses produce short-term M2 parameter changes; the hypothesis that instrument operational conditions and long-term hydrological changes dominate the observed variations is supported. The absence of earthquake-triggered M2 parameter responses is attributed to basin attenuation, limited strain conversion efficiency, response mechanism decoupling, and observational resolution. These results provide a background reference for applying well water-level tidal parameters to earthquake precursor monitoring in the Beijing area, showing that independent baselines must be established for each well and that instrument operational records are necessary for anomaly identification. Full article

Cargo items at different longitudinal positions experience different local base excitations because of vehicle-body bounce and pitch, creating position-dependent demands on vibration isolation and support stroke. This study presents a response-guided equivalent-design framework that links critical-position identification to the selection of nonlinear support characteristics represented by an equivalent force model for X-type cargo supports. A coupled model comprising prescribed stochastic road inputs, linear tire stiffness and damping, a four-degree-of-freedom half-car subsystem, and five vertically supported cargo masses of 2000 kg each is established. Cargo acceleration, relative support displacement, interaction force, and energy-related indicators are evaluated. Under the nominal condition used for design-target extraction, P5 is identified as the critical position, with a dominant local-base frequency of 2.3994 Hz and an RMS-equivalent displacement amplitude of 15.339 mm. These response characteristics, together with prescribed design constraints, guide the selection of equivalent support properties. The support force law includes basic stiffness, delayed hardening, a displacement-activated limiting term, displacement-dependent damping, and regularized friction. In the nominal system-level comparison, the low-frequency-compliant X-type scheme reduced the maximum P5 stroke from 35.68 mm to 31.81 mm relative to the feasible low-frequency linear reference, accompanied by small increases in acceleration and interaction-force RMS. Under the investigated perturbed conditions, the grouped X-type configuration reduced the largest P5 stroke from 56.05 mm to 47.83 mm relative to the linear reference; however, it still exceeded the prescribed 40 mm limit. Further parameter optimization, independent benchmark comparison, and experimental validation are required before engineering feasibility can be established. Full article

Accurate identification of low-grade graphite ore is important for improving resource utilization and intelligent mineral sorting. To reduce the computational burden of existing instance segmentation models, this study proposes a lightweight model, RVE-YOLO-seg, based on YOLOv12-seg. GhostConv is introduced for lightweight downsampling, C3k2-RVE is designed to enhance fine-grained feature representation, and Segment-SEAM is employed to strengthen mask-oriented feature extraction. Experiments on a self-constructed dataset of 1978 conveyor-belt images and 19,614 annotated ore instances show that RVE-YOLO-seg achieves 92.7% mAP50 for bounding boxes and 86.3% mAP50 for masks, comparable to YOLOv12n-seg. Meanwhile, the parameter count, FLOPs, and model size are reduced by 56.7%, 24.6%, and 52.5%, respectively, with an inference speed of 80 FPS. Multi-seed experiments and retraining on an external mineral-image dataset further demonstrate stable performance and cross-dataset applicability. These results indicate that RVE-YOLO-seg achieves a favorable accuracy–efficiency trade-off for resource-constrained graphite ore sorting. Full article

Applying augmented reality (AR) glasses to overhead crane inspection can significantly improve the efficiency and accuracy of inspection operations. However, this technology also introduces new and complex risks stemming from the intricate interactions between people, machines, the environment, and management factors. Because traditional risk assessment methods implicitly assume factor independence, they struggle to capture this complexity. To overcome this deficiency, this study develops a Consequence-Prioritized Weighted Borda integrated with DEMATEL (CPWB-DEMATEL) hybrid framework specifically designed for risk assessment in AR-assisted inspection environments. First, an application-oriented risk identification framework is proposed by expanding SHELL model from a traditional four-dimension structure to include dedicated Personnel (P) and Management (M) dimensions, which leads to the systematic identification of 27 critical risk factors. Second, the weighted Borda severity ranking is combined with DEMATEL network centrality, ensuring that the comprehensive risk priority reflects both high inherent severity and high systemic influence. At the same time, this integration further enables a critical risk propagation path identification procedure that extracts multi-hop transmission chains from the DEMATEL influence matrix, revealing system-level risk propagation structures. Third, a case study conducted at a special equipment inspection agency validated the applicability of the framework. The results show that operator/rigger violations, signal personnel coordination errors, multi-task coordination failures, inadequate AR process supervision, and wire rope recognition errors are the most critical risk factors, with management deficiencies and communication-related factors also playing significant roles in risk propagation. Comparative analyses and sensitivity tests confirm the stability and robustness of the derived risk rankings across various parameter settings. Full article

The evaluation of rock properties is an indispensable part of the process of developing deep oil and gas resources. Traditional lithology identification methods have limitations such as high cost and time consumption. During the drilling operation, the interaction between the drill bit and the formation occurs, and the dynamic vibration response of the drill bit can represent the geological characteristics of the formation. In this paper, the differences in vibration response during the drilling process are explored by experimental methods, and the feasibility of using vibration signals to judge the change in rock stratum is investigated. A full-scale drill bit rock breaking test rig is constructed, and laboratory experiments on full-size Polycrystalline Diamond Compact (PDC) bit rock breaking are carried out, during which vibration response data under different drilling parameters are collected. The differences among experimental datasets are then analyzed and compared, and key data characteristics are extracted. Based on these, a new method for discriminating formation lithology is proposed. The results show that the amplitude of vibration response is closely related to drilling parameters and rock properties. The frequency composition of the vibration is only affected by the properties of the rock strata and is not affected by drilling parameters. Furthermore, the model established based on differences in vibration responses can effectively identify lithological changes, with an identification accuracy rate exceeding 95%. This study offers a theoretical basis for identifying lithologic changes in field production and serves as a valuable supplement to existing methods. Full article

Reliable transition of laboratory-developed gas-sensing nanomaterials into environmental monitoring systems remains limited by the lack of integrated testing platforms capable of reproducing realistic operating conditions while providing automated, high-quality sensor characterization. In this work, we present a comprehensive framework for environmental gas sensor development centered on a custom-designed gas sensor chamber (GSC) and its connection to standalone measuring stations (sMSs) intended for smart city air quality monitoring. The GSC was developed as a dedicated platform for controlled evaluation of sensing nanomaterials through automated regulation of temperature, relative humidity, gas concentration, and pressure, combined with impedance measurements over a broad frequency range (10 Hz–100 kHz). The system integrates dedicated chamber hardware, environmental control loops, an LCR/impedance-based measurement unit, and in-house software for instrument control, data acquisition, synchronization of environmental and impedance data, and post-processing with visualization and correlation analysis. The functionality of the platform was demonstrated using graphene oxide-based nanocomposites deposited on interdigital electrodes and tested under variable parameters of temperature, humidity, and carbon monoxide concentration. The chamber enabled stable and reproducible extraction of frequency-dependent sensor responses and identification of optimal operating conditions for selected materials. Building on these results, the same sensing concept was translated into an energy-autonomous sMS architecture based on modular interdigital sensor elements, impedance-based indirect gas quantification, remote communication, and photovoltaic power support. By integrating material evaluation, environmental simulation, automated data handling, and field-oriented station design within a single workflow, the presented approach bridges the gap between laboratory gas sensor research and practical deployment in distributed smart city monitoring networks. Full article

The detection and identification process in radar is susceptible to disturbances from numerous non-cooperative jamming signals in addition to target signals. Accurately estimating the parameters of these non-cooperative signals is crucial for extracting the target signal. Addressing the current lack of effective solutions for estimating parameters of different types of uncooperative combined signals, this paper proposes a method for parameter estimation of non-cooperative composite signals based on joint time-polarization domain separation. This method combines amplitude and phase information from both the time and polarization domains of non-cooperative signals and integrates the JADE separation algorithm to achieve the separation and estimation of specific non-cooperative signal combinations. Experiments compare the separation performance of this method with other signal separation algorithms for non-cooperative jamming signal combinations. The results indicate that, at jamming-to-noise ratio of −10 dB to 15 dB, this method achieves better separation accuracy and precision for specific non-cooperative jamming combinations, while also accurately estimating the parameters of the non-cooperative jamming signals. Full article

Accurate identification of pavement distress in UAV imagery is crucial for road safety but is hindered by scale disparities, background noise, and isotropic extraction constraints that neglect highly directional distress textures. This paper proposes MAG-YOLO, a lightweight detection network driven by multi-scale and anisotropic gating perception. The architecture integrates three core innovations: a Re-parameterized Multi-scale Attention (RMA) module, which employs multi-branch depthwise convolutions to enhance fine-grained texture capture while mitigating parameter redundancy; an Anisotropic Gating Feature Pyramid Network (AGFPN), which incorporates orthogonal feature decomposition and dynamic gating mechanisms to extract multi-scale topological features and suppress environmental noise; and a Cross-stage Multi-scale Gated Linear Attention (CMGLA) module designed for global context aggregation and local feature recalibration. Extensive experiments on the RDD2022_China_Drone and UAPD datasets demonstrate that MAG-YOLO outperforms the YOLOv11n baseline by 5.4% and 7.0% in mAP50, respectively, while reducing the parameter count by 29.1%. Deployment tests on the RK3588 edge platform further confirm that MAG-YOLO achieves a favorable balance between accuracy and real-time efficiency. This work provides a robust and deployment-ready solution for UAV-based pavement distress detection, supporting automated road inspection and intelligent pavement maintenance. Full article

With the growing penetration of renewable energy, accurately characterizing the frequency support performance of grid-following wind and photovoltaic (PV) units has become increasingly important. However, conventional methods for assessing frequency support parameters often overlook practical dynamic effects, making it difficult to determine the support parameters actually realized during disturbances. To address the time-domain coupling, differential noise amplification, and parameter distortion problems in the online identification of virtual primary frequency regulation and virtual inertia coefficients, this paper establishes a frequency response model for wind and PV units that incorporates these support mechanisms together with practical physical constraints. On this basis, a two-stage forgetting-factor recursive least squares (FF-RLS) method is proposed to identify realized frequency support parameters. Exploiting the difference in response time scales between primary frequency regulation and inertial support, a quasi-steady-state frequency regulation window and a transient inertia window are constructed to decouple the two parameters. Meanwhile, Tustin phase compensation and band-limited differentiation are introduced to mitigate measurement noise and the phase mismatch between frequency and power responses. Finally, a stable window criterion is developed to adaptively extract reliable identification intervals. Simulation studies on a modified IEEE 24 bus system, together with comparisons against conventional identification methods, demonstrate the effectiveness and accuracy of the proposed method. Full article

Objective: Falls among the elderly constitute a major global public health issue. Research has therefore focused on two complementary areas: fall risk prevention and fall detection. Existing solutions mainly rely on body-attached sensors confined to laboratory settings, whereas recent research has shifted toward wearable e-textiles with non-invasive, long-term-wear sensors. In this context, the study presents a wearable sensor platform to both predict fall risk and detect falls, based on sensorized clothing integrating inertial measurement units and surface electromyography sensors. Methods: Fall risk was estimated from gait parameters extracted during a 10 m walking test as the probability of belonging to a faller (vs. non-faller) group, using a logistic regression model trained on the G-STRIDE dataset, complemented by neuromuscular parameters extracted from sEMG and associated with fall risk. Fall detection, focused on improving pre-impact identification, was framed as a binary classification between activities of daily living and falls, using a reduced Spatio-Temporal Attention Network trained on the FallTL dataset and refined with our own platform’s data. Results: The obtained results were: fall-risk assessment, Area Under the Curve = 77.8%, Accuracy = 68.7%; fall detection, Accuracy = 96.7%, F1 Score = 77.3%, lead time = 390 ms. Conclusions: As a single-subject proof of concept, these results support the feasibility of a unified platform integrating objective fall-risk screening and fall event identification, with a predicted time before impact suitable for protective systems intervention; validation on a larger, representative cohort is required before any clinical claim can be made. Full article

Existing parameter extraction methods for photovoltaic modules are mostly developed for specific module types or structural configurations, making it difficult to achieve unified modeling across different materials, structures, and operating conditions. To address this issue, this study proposes a unified five-parameter identification method based on the single-diode model of photovoltaic cells and combines it with a genetic algorithm. The photocurrent, reverse saturation current, series resistance, shunt resistance, and diode ideality factor are selected as the identification parameters, and the parameter extraction problem is formulated as a global optimization problem. On this basis, a cell-unit-based parameter correction model considering the effects of irradiance and temperature is established, and unified modeling of photovoltaic modules with different architectures is achieved according to their internal electrical connections, enabling the prediction of their electrical characteristics. The proposed method is validated using experimental data from a half-cell mono-crystalline silicon module under different shading conditions, as well as mono-crystalline silicon, multi-crystalline silicon, and thin-film modules under varying irradiance and temperature conditions. In addition, the RTC France solar cell and Photowatt-PWP201 module benchmark datasets are employed to further assess the reliability of the parameter identification procedure. The results demonstrate that the proposed method can effectively reproduce the electrical characteristics of the investigated photovoltaic modules, with maximum relative errors of 2.48%, 2.17%, and 4.32% for open-circuit voltage, short-circuit current, and maximum power, respectively. The benchmark validation further demonstrates that the proposed method achieves fitting accuracy comparable to that of other representative optimization algorithms while exhibiting good repeatability and convergence performance. These results collectively demonstrate the feasibility of the proposed cell-unit-based five-parameter modeling framework for photovoltaic modules with different materials and structures under the investigated operating conditions, providing a simple and feasible approach for unified parameter identification, performance characterization, and engineering modeling. Full article

Large-scale photovoltaic (PV) power plants are expanding rapidly across arid desert regions, driving land-cover transformations and associated changes in surface properties. Most existing remote sensing studies focus primarily on mapping PV distribution, while quantitative assessments of PV-induced land-cover transitions and associated surface property changes remain limited. This study develops an integrated assessment framework that combines time-series Sentinel-2 imagery and Moderate Resolution Imaging Spectroradiometer (MODIS) Bidirectional Reflectance Distribution Function (BRDF) prior parameters. Three semantic segmentation models are compared for PV extraction, with independent generalization validation conducted over desert areas in Xinjiang. Constrained by coarse-resolution BRDF products, 10 m broadband white-sky albedo (WSA) is retrieved. The results show that SegFormer outperforms the other two models for PV identification. From 2021 to 2025, the PV-covered area of the Talatan region expanded from 160.35 km² to 303.26 km². For the newly converted PV area, surface albedo decreased by 0.0391 and 0.0310 during 2021–2023 and 2023–2025, respectively, corresponding to local albedo-driven shortwave energy changes of 27.12 W·m⁻² and 23.55 W·m⁻². Consistent variation patterns are observed in the Xinjiang validation site. This study offers an integrated framework for characterizing PV expansion-induced land-cover changes and associated surface property variations, while providing an observation-based assessment of local shortwave energy variations related to surface albedo changes in arid regions. Full article