Search Results (4,841)

This paper presents a multi-sensor framework for condition monitoring and intelligent fault diagnosis in rotating industrial machinery, integrating heterogeneous sensing, machine-learning-based diagnosis, MQTT communication, historical data management, and a synchronized 3D Digital Twin. The experimental platform employs feature-level fusion by combining statistical and spectral descriptors extracted from vibration and acoustic signals with temporally aligned temperature and rotational-speed measurements to classify four machine-health states: Healthy, Minor, Moderate, and Severe. To assess the robustness of the diagnostic approach under different operating conditions, experiments were conducted at seven controlled operating frequencies ranging from 15 to 70 Hz. Sensor signals were segmented and characterized using statistical and spectral features, and seven supervised machine-learning classifiers were comparatively evaluated. The diagnostic contribution of sensor fusion was assessed through an ablation analysis, with classification accuracy increasing from 94.8% using vibration alone to 98.1% when vibration, acoustic, and temperature features were combined. Among the evaluated models, Random Forest achieved the highest overall classification performance, reaching 98.9% accuracy in the comparative evaluation. Under the stricter Leave-One-Operating-Frequency-Out validation, the model achieved a mean accuracy of 98.04 ± 0.48% and a mean Macro-F1 score of 97.94 ± 0.48%, demonstrating stable classification across unseen operating conditions. The architecture separates high-frequency signal acquisition and local processing from supervisory communication, with extracted features, diagnostic states, and selected signal blocks exchanged through MQTT for historical monitoring and Digital Twin synchronization. The MQTT supervisory layer achieved a mean communication latency of 21.8 ms, with no message loss observed during the reported 1000-message test sequence. Experimental results obtained from controlled imbalance and bearing-fault conditions demonstrate the feasibility of the proposed framework for multi-condition fault diagnosis and operator-oriented maintenance support. The developed platform provides an experimental basis for future cross-machine validation, prognostics, remaining-useful-life estimation, and bidirectional Digital Twin capabilities. Full article

Credit card fraud detection is challenging due to the extreme class imbalance and the limited number of fraudulent transactions. To address this problem, this paper proposes a fraud detection framework that combines generative AI-based data augmentation with machine learning classification models. We generated synthetic fraud samples at augmentation ratios of 1%, 5%, and 10% using three generative approaches: a Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP), a Variational Autoencoder (VAE), and a Large Language Model (LLM)-based data generator. The generated samples were evaluated using the Discriminator Test, Distance-to-Closest-Record (DCR) analysis, and t-SNE visualization to compare them to real fraud data and identify differences in their distributions. The augmented datasets were then used to train Random Forest, XGBoost, and LightGBM classifiers. The baseline models trained on the original dataset achieved F1-scores of 0.857 for Random Forest, 0.856 for XGBoost, and 0.877 for LightGBM. After adding synthetic fraud samples to the training data, the results showed improvements in fraud detection performance in some cases. LightGBM achieved the highest F1-score of 0.897 when trained with VAE-generated data at the 10% augmentation ratio. Similarly, XGBoost achieved an F1-score of 0.880 using VAE-generated data at the 5% ratio. Overall, the findings demonstrate the potential of generative AI-based augmentation to improve fraud detection performance in highly imbalanced financial datasets, particularly when combined with LightGBM and XGBoost classifiers. Full article

Chronic obstructive pulmonary disease (COPD) is characterised by persistent airflow limitation arising from chronic inflammation, oxidative stress, mucus hypersecretion, protease–antiprotease imbalance, airway remodelling, and emphysematous destruction. This review evaluates plant-derived bioactive compounds with reported activity against molecular processes implicated in COPD. Resveratrol, kaempferol, quercetin, fisetin, epigallocatechin gallate, gallic acid, ursolic acid, hesperidin, apigenin, and luteolin are examined with emphasis on Nrf2, NF-κB, MAPKs, PI3K/Akt, SIRT1, TLR4, and NLRP3 signalling. In selected cellular and animal models, these compounds have been reported to reduce oxidative stress, inflammatory mediator release, mucus production, metalloproteinase activity, and tissue injury. However, these experimental findings do not establish therapeutic efficacy in patients with COPD. Some also affect autophagy, corticosteroid sensitivity, and pathways associated with cellular senescence. Overall, the evidence is derived mainly from cultured cells and animal models, with limited clinical data. Poor oral bioavailability, rapid metabolism, uncertain pulmonary exposure, formulation variability, dose standardisation, and potential drug interactions remain major translational limitations. Future studies should define pharmacokinetics, lung tissue exposure, optimal formulations, safety, and interactions with standard COPD therapies, followed by adequately powered clinical trials assessing exacerbations, lung function, symptoms, exercise capacity, quality of life, and treatment-related adverse effects in well-defined COPD populations. Full article

Labour-driven grading of apples can be inconsistent because quality judgments may vary between inspectors, particularly for internal attributes such as sweetness that cannot be determined reliably from appearance alone. Artificial intelligence offers a more consistent alternative; however, its performance can be affected when the available training data are limited and unevenly distributed across quality classes. The multispectral apple dataset used in this work comprises images acquired in eight discrete spectral bands for non-destructive apple sweetness classification. For sweetness classification, the available samples remain imbalanced across Brix ranges, making minority class recognition more difficult. Five deep learning architectures—AppleNet, MobileNetV3-Small, EfficientNet-B0, ResNet18, and Swin-Tiny—were first compared using five-fold cross-validation on the real dataset. EfficientNet-B0 and Swin-Tiny achieved mean accuracies of approximately 94%, while AppleNet showed substantially weaker and more class-biased predictions. EfficientNet-B0 was then evaluated using classical augmentation, GAN, DDPM, and DDPM2 with different synthetic data proportions. The generated samples were further examined using structural, perceptual, feature distribution, diversity, and class fidelity measures. Augmented training improved class-wise consistency and fold-level performance, although larger synthetic proportions gradually reduced correspondence with the real data distribution. The results indicate that synthetic augmentation is most effective when sample quantity is controlled together with fidelity and class preservation. Full article

Objective: Female athletes have a higher prevalence of menstrual disorders due to hormonal imbalances, often related to physical stress, low energy availability, and inadequate nutritional habits. These disorders can affect both health and athletic performance. The aim of this study is to evaluate the frequency of menstrual cycle disorders in female athletes and their relationship with nutritional variables. Design: Observational, descriptive, cross-sectional study based on an anonymous online questionnaire administered to female athletes aged 18–55 years (n = 418; average age = 28 ± 9 years). Methods: The questionnaire collected self-reported data on menstrual cycle characteristics, diet, supplementation, and training-related habits. Chi-square tests, Fisher’s exact test, and Kruskal–Wallis tests were used for statistical analysis, with significance set at p < 0.05. Results: Premenstrual symptoms were reported by 57.0% of participants, regular clot presence by 27.4%, and cycle irregularities by 22.8%. In unadjusted analyses, self-reported dietary balance was associated with cycle regularity and bleeding duration; after false discovery rate correction, only the association with altered bleeding duration (binary coding) remained statistically significant. No clear associations were observed between sports frequency and menstrual cycle alterations. Conclusions: Menstrual disorders were common among female athletes. Although an association between self-reported dietary balance and bleeding duration persisted after correction for multiple comparisons, it derived from an exploratory binary recoding and warrants cautious interpretation. These findings may help guide future research on nutrition and menstrual health in female athletes. Full article

Over recent decades, the veterinary profession has undergone substantial feminisation. However, this demographic transformation has not been accompanied by proportional representation of women in leadership and decision-making roles. This descriptive retrospective study examined women’s representation in leadership positions within Italian veterinary scientific societies, the National Federation of Veterinary Orders (FNOVI), the National Veterinary Social Security and Welfare Institution (ENPAV), and the Provincial Veterinary Orders. Data were extracted from publicly available institutional websites and official organisational records. Gender was inferred from names and verified, when necessary, using publicly available professional and institutional information. The analysis of veterinary scientific societies covered the three most recent presidential terms, whereas FNOVI, ENPAV, and Provincial Veterinary Orders were analysed over longer historical periods. Veterinary scientific societies showed an increase of 19.2 percentage points in the proportion of women presidents. In 2025, women accounted for 36.8% of society presidents and 36.8% of executive board members. Women also accounted for an average of 18.4% of Provincial Veterinary Order presidents between 2012 and 2025. Neither FNOVI nor ENPAV has ever had a woman president. Overall, the findings indicate persistent gender imbalance in leadership representation across Italian veterinary governance structures. Full article

This paper presents a single-layer coupled-line Marchand balun designed for a nominal transformation from a 50-$\Omega$ unbalanced source to a 300-$\Omega$ differential load at 2.25 GHz. The circuit combines two inductively terminated coupled-line sections with an independent input matching network. Symmetric termination inductors provide a means of tuning the coupled-line response without introducing additional cascaded sections. The design was developed using even- and odd-mode transmission-line analysis and optimized in Keysight Advanced Design System (ADS). Conventional and compensated prototypes were fabricated for comparison. At 2.25 GHz, the compensated prototype yielded a reported input reflection level of $-28.05$ dB, with output transmission levels of $-3.11$ and $-3.05$ dB. Its reported amplitude imbalance was 0.052 dB, and the output phase difference was 180.52${}^{\circ }$ (0.52${}^{\circ }$ phase error). The corresponding conventional prototype exhibited an input reflection level of $-21.51$ dB, an amplitude imbalance of 0.217 dB, and a phase error of 9.29${}^{\circ }$. The results indicate improved balance and input matching at the design frequency. Because the retained records do not establish whether the VNA data were renormalized from physical 50-$\Omega$ ports to the intended 150-$\Omega$ output references, the measurements cannot independently confirm the differential impedance transformation. Full-band performance, common-mode rejection, and circuit-area advantages are therefore not quantified. Full article

Automated waste image classification remains challenging because of substantial class imbalance, large intra-class visual variability, and high inter-class visual similarity. We propose a backbone-agnostic weak-to-strong training framework that combines class-balanced focal supervision over weak, strong, and activation-guided mixed views, asymmetric Kullback–Leibler (KL) consistency, and Activation-Guided CutMix. The detached weak-view prediction distribution serves as the reference for aligning the strong-view prediction, while Activation-Guided CutMix centers the replacement region at the highest-response location derived from an intermediate feature map of the base weak view rather than selecting its location randomly. The framework was evaluated on two waste-classification datasets using seven CNN backbones. ConvNeXt-Base achieved 98.88% mean accuracy and 98.44% macro-F1 on the corrected Waste Classification dataset, while ConvNeXt-Tiny achieved 79.20% mean accuracy and 76.73% macro-F1 on WaRP-C. Ablation studies demonstrated performance gains across different backbone architectures, with component-level contributions varying by dataset and backbone. The t-SNE, confusion-matrix, and Grad-CAM analyses provided complementary qualitative evidence of more clearly organized feature distributions, reduced off-diagonal class confusion, and more object-focused activation patterns. Because all auxiliary operations are confined to training, the proposed framework preserves the original single-branch, single-forward-pass inference architecture. Full article

Continuous scalp EEG monitoring enables automated epileptic seizure detection, but extreme class imbalance, inter-patient covariate shifts, and intra-subject data leakage heavily hinder clinical deployment. This study introduces an unsupervised, cross-patient feature selection and patient-independent one-class anomaly detection under Leave-One-Group-Out (LOGO) cross-validation enhanced by a novel One-Class Hybrid (G-Cramer’s V) Statistical Gate to resolve these gaps. The architecture models normal baseline neural activity as the solitary learned class, employing a dynamic feature gate that maximizes baseline purity via a One-Class Gini Index and drives robust cross-patient generalization through a Cramer’s V filter. To eliminate patient-level data leakage and ensure reliable cross-patient performance, a strict Leave-One-Group-Out (LOGO) cross-validation protocol is enforced. Benchmarked on the open-source CHB-MIT database against five core unsupervised estimators (OCSVM, kNN, LNND, LOF, IF) and a reference Morlet Wavelet baseline, the proposed Local Outlier Factor (LOF) estimator utilizing a Gated Fourier framework delivers high precision and efficiency. It achieves a mean cohort F1-score of 0.79, a precision of 78%, and a False Alarm Rate of 1.5 alarms per hour on completely unseen patients, while simultaneously yielding a 35.5% reduction in feature-extraction latency. Full article

With the advancement of satellite remote sensing and related techniques, the resolution of remote sensing satellites and aerial remote sensing has been continuously improved, providing more detailed ground observation information. The significance of semantic segmentation for high-resolution remote sensing images has also increased with the expansion of their applications. Hence, a new research direction in image interpretation and remote sensing data processing is to establish a connection between the pixel-level features and higher-level semantics in order to achieve precise semantic segmentation of remote sensing images. It is still difficult to extract fine details from the complicated remote sensing data by applying the traditional region-based and edge-based approaches. Although deep learning can enhance the semantic representation, further improvement in segmentation accuracy is usually accompanied by a larger model size and a heavier computational burden. To overcome this problem, this paper investigates targeted structural modifications of a standard segmentation network, and a specially designed DeepLabV3+ architecture is suggested for high-resolution remote sensing. The SE-ASPP module takes the place of the conventional method to deal with scale differences, while the CFF module integrates the hierarchical features. CARAFE upsampling, which is intended to retain the fine texture features, is utilized to solve the issue of spatial reconstruction. Moreover, an integrated loss function combining Dice and cross-entropy is employed to solve the inherent class imbalance in urban objects. In the end, the improved architecture demonstrates good potential for accurate and efficient remote sensing applications owing to its ability to handle the complex spatial relationships more effectively than the basic one. The effectiveness of this approach is evaluated on the ISPRS Potsdam and Vaihingen datasets using the official ISPRS train/test tile partition, with the proposed model and the DeepLabV3+ baseline averaged over three independent runs and reported as mean ± standard deviation. The improved network attains overall accuracies of 87.98% and 85.58%, mean intersection over union (mIoU) of 73.58% ± 0.03 and 64.99% ± 0.26, and mean F1-scores (mF1) of 83.49% ± 0.03 and 77.21% ± 0.30 on Potsdam and Vaihingen, respectively. Relative to the standard DeepLabV3+ baseline trained under the same protocol, the proposed model improves mIoU by 1.02 and 2.82 points and yields the highest car-class IoU among the four models with per-class results on Potsdam, while its boundary F1-score rises consistently, indicating that the main benefit lies in boundary delineation and small-object recovery rather than in a uniform accuracy lead over all competing models. The outcomes show that the proposed structure provides a balanced and reproducible solution for complex urban scene interpretation, with a moderate and explicitly quantified increase in parameters and computational cost. Full article

The growing worldwide penetration of solar energy into power grids has increased the need for accurate forecasting to reduce supply–demand imbalances and grid congestion. Physics-based models perform well for average conditions. However, edge cases such as very low or high irradiance are challenging for traditional physics-based models, which often underperform under these conditions. Data-driven models offer improved performance in these conditions but typically lack interpretability and insight into the underlying physical mechanisms. To address these limitations, this article presents a combined modeling framework that integrates physics-based principles with machine learning inspired by Physics-Informed Neural Networks (PINNs). The main novelty of this work is a machine learning architecture for solar energy power prediction that explicitly isolates angle of incidence losses and temperature losses in the form of the respective parameters entering a PV power equation as a strict output constraint, which improves interpretability and generalizability, while maintaining high forecasting precision. The methodology is demonstrated using only commonly available commercial system data and validated on both a small-scale research system and a commercial floating solar installation. The results demonstrate significantly improved accuracy over basic physics-based models (0.018 and 0.018 versus 0.023 and 0.029 normalized mean absolute error), particularly for non-standard conditions, and isolate key loss mechanisms. Full article

Accurate mapping of coastal wetland vegetation is essential for ecological conservation and blue carbon assessment, yet it remains challenging due to fragmented boundaries, spectral similarity among species, pronounced phenological variability, and limited generalization under small-sample conditions. We propose a spatio-temporal attention hybrid network (STAH-Net) for multi-source time-series classification. A 29-band feature cube is constructed from Sentinel-1 and Sentinel-2 data, integrating monthly Normalized Difference Vegetation Index (NDVI) and vertical–vertical (VV) polarization time series with static spectral and polarimetric indices. An improved U-Net backbone with dual-branch 3D convolutions extracts optical and Synthetic Aperture Radar (SAR) temporal features in parallel, incorporating a Temporal Enhancement Attention Module (TEAM) for phenological weighting, a Local Channel Attention (LCA) module for channel recalibration, and Atrous Spatial Pyramid Pooling (ASPP) for multi-scale context. An edge-weighted hybrid loss combining edge-weighted cross-entropy and Dice loss mitigates class imbalance and boundary blurring. Experiments in Hangzhou Bay, Liaohe Estuary, and Minjiang Estuary achieves overall accuracy (OA) values of 98.16%, 97.23%, and 95.76%, respectively, with a maximum Kappa of 97.49%, outperforming ConvLSTM, U-Net, SegFormer, SITSMamba, and GLMambaNet under identical settings. Ablation and visual analyses confirm the effectiveness of the proposed modules, with NDVI phenology providing primary discrimination and VV offering complementary structural information. Cross-year and cross-region tests show promising transferability, with OAs of 97.21% and 94.75% in Liaohe Estuary for 2024 and 2020, respectively, and 93.83% in Shanghai. These results demonstrate that STAH-Net provides an effective and transferable workflow for coastal wetland vegetation mapping. Full article

Accurate prediction of hybrid press dynamics requires the integration of data-driven modeling with the governing principles of constrained multibody mechanics. This study proposes a physics-informed deep neural network (PIDNN) for predicting five engineering responses, including slider displacement, velocity, acceleration, actuator force, and required motor torque, from six operating inputs. The proposed framework integrates supervised learning with deterministic reconstruction of generalized mechanical states, incorporating kinematic consistency, dynamic equilibrium, constraint reaction estimation, frictional effects, and energy-rate conservation into a unified optimization objective. The model was developed and evaluated using 75 operating trajectories generated in MATLAB/Simscape, comprising 375,000 samples distributed across independent training, validation, and testing subsets. Experimental evaluation demonstrated that the proposed PIDNN achieved normalized root mean square errors (nRMSEs) of 0.118, 0.129, 0.154, 0.111, and 0.138 for the five predicted responses, respectively, yielding an overall mean nRMSE of 0.130 compared with 0.221 for the conventional data-driven artificial neural network, representing a 41.2% reduction in prediction error. Furthermore, the proposed model achieved a root mean square normalized dynamic residual of 0.018 and a mean energy imbalance of 0.91%, demonstrating improved consistency with the governing mechanical equations. Controlled ablation experiments confirmed the effectiveness of the combined physics-informed constraints in improving predictive accuracy and mechanical consistency. Computational evaluation demonstrated single-sample inference latency of 0.17–0.20 ms, indicating substantial computational advantages over conventional numerical integration. These findings demonstrate the effectiveness of the proposed PIDNN as an accurate, computationally efficient, and physically consistent surrogate model for simulation-based prediction and analysis of hybrid press system dynamics. Full article

This study examines the profile and leadership style of women in leadership positions at a university in northern Chile. This cross-sectional descriptive study combined institutional administrative records with self-reported MLQ Form 6S data from 45 of the 48 women occupying leadership positions at the institution, representing a response rate of 93.75%. Women occupy 45.7% of leadership roles; however, women’s representation remains low in several leadership categories, including the Rector’s Office, Vice-Rector’s Office, and school or department director positions. This pattern points to the persistence of vertical segregation in university governance. Among the women surveyed, high self-reported scores predominated across the four transformational dimensions and both transactional dimensions, whereas laissez-faire scores were predominantly moderate. The study highlights the coexistence of leadership styles within a context marked by persistent gender asymmetries and limited access for women to positions of greater authority and decision-making. Practically, the study supports targeted institutional measures to improve women’s representation in leadership categories where they remain underrepresented. Socially, it underscores the need to address persistent gender imbalances in higher education leadership. Full article

Condition documentation for traditional stone carvings requires comparable and repeatable records, yet field images are often limited by small samples, class imbalance, multi-defect co-occurrence, and diffuse defect boundaries. This study uses 643 field images from 26 courtyards in the Guanzhong region to examine classification granularity, model performance, and conservation applicability. Three tasks were defined: M1 for defect presence, M2 for six-class multi-label recognition, and M3 for eleven-class fine-grained recognition. Under courtyard-grouped nested cross-validation, in which the held-out fold is never used for model selection or threshold tuning, ResNet-50 achieved an F1 of 0.8908 ± 0.0055 on M1, with macro-F1 falling to 0.5378 ± 0.0148 on M2 and 0.3274 ± 0.0159 on M3. Model and training variants did not consistently outperform the baseline, and fine-tuned vision–language models reached comparable accuracy on M2 (0.5719) at far higher training cost. Class-wise analysis showed pronounced differences in reliability, with artificial damage at an F1 of 0.7361 and abrasion at 0.3409; false positives were more frequent in images containing other defect labels, while abrasion and soiling showed high missed-detection rates. M1 is suitable for defect screening and M2 can support defect typing and expert review, whereas M3 is presented as a data-sufficiency assessment. Full article