Search Results (72)

Bond wire degradation represents the predominant failure mechanism in IGBT modules, accounting for approximately 70% of power converter failures and posing significant reliability challenges in modern power electronic systems. Existing monitoring techniques face inherent trade-offs between measurement accuracy, implementation complexity, and electromagnetic compatibility. This paper proposes a physics-constrained ensemble learning framework for non-intrusive bond wire health assessment via V_ce-on prediction. The methodological innovation lies in the synergistic integration of multidimensional feature engineering, adaptive ensemble fusion, and domain-informed regularization. A comprehensive 16-dimensional feature vector is constructed from multi-physical measurements, including electrical, thermal, and aging parameters, with novel interaction terms explicitly modeling electro-thermal stress coupling. A dynamic weighting mechanism then adaptively fuses three specialized gradient boosting models (CatBoost for high-current, LightGBM for thermal-stress, and XGBoost for late-life conditions) based on context-aware performance assessment. Finally, the meta-learner incorporates a physics-based regularization term that enforces fundamental semiconductor properties, ensuring thermodynamic consistency. Experimental validation demonstrates that the proposed framework achieves a mean absolute error of 0.0066 V and R² of 0.9998 in predicting V_ce-on, representing a 48.4% improvement over individual base models while maintaining 99.1% physical constraint compliance. These results establish a paradigm-shifting approach that harmonizes data-driven learning with physical principles, enabling accurate, robust, and practical health monitoring for next-generation power electronic systems. Full article

The construction industry remains a cornerstone of the global economy; however, it continues to face persistent challenges, including low productivity, frequent workplace accidents, and environmental degradation. This study employs a systematic literature review to explore how digital technologies can enhance these three areas in construction project management, focusing on their benefits and challenges. The study adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A total of 18 articles were retrieved from Scopus and Web of Science databases. The findings highlight Building Information Modeling (BIM) as the most influential digital innovation supporting productivity gains, improved safety standards, and progress towards sustainable practices. Among the three focus areas, productivity remains the most extensively investigated, while sustainability is relatively underexplored. The identified benefits include increased productivity, enhanced safety, improved risk management, data-driven decision-making, improved sustainability, real-time monitoring, and stronger collaboration. Conversely, significant barriers include high implementation and training costs, data privacy concerns, a limited number of skilled workers, and resistance to change among construction stakeholders. The review emphasizes the need for further empirical studies that investigate underrepresented technologies and regional contexts. It further suggests that industry practitioners and policymakers should prioritize digital capacity building, policy incentives, and regulatory frameworks to strengthen the sustainable digital transformation of construction project management. This review presents a unique, integrated perspective by synthesizing outcomes related to productivity, safety, and sustainability. It not only delineates critical research gaps but also provides actionable guidance for industry practitioners and policymakers by prioritizing strategic areas such as digital capacity building, policy incentives, and regulatory frameworks. Full article

The rapid and accurate identification of radionuclides and the quantitative analysis of their activities have long been key research areas in the field of nuclear spectrum data processing. Traditional nuclear spectrum analysis methods heavily rely on manual feature extraction, making them highly susceptible to interference from factors such as energy resolution, calibration drift, and spectral peak overlap when dealing with complex mixed-radionuclide spectra, ultimately leading to degraded identification performance and accuracy. Based on multi-nuclide energy spectral data acquired via Geant4 simulation, this study compares the performance of partial least squares regression (PLSR), random forest (RF), a convolutional neural network (CNN), and a hybrid CNN–Meta model for radionuclide identification and quantitative activity analysis under conditions of raw energy spectra, Z-score normalization, and min-max normalization. To maximize the potential of each model, principal component selection, Bayesian hyperparameter optimization, iteration tuning, and meta-learning optimization were employed. Model performance was comprehensively evaluated using the coefficient of determination (R²), root mean square error (RMSE), mean relative error (MRE), and computational time. The results demonstrate that deep learning models can effectively capture nonlinear relationships within complex energy spectra, enabling accurate radionuclide identification and activity quantification. Specifically, the CNN achieved a globally optimal test RMSE of 0.00566 and an R² of 0.999 with raw energy spectra. CNN–Meta exhibited superior adaptability and generalization under min-max normalization, reducing test error by 70.8% compared to RF, while requiring only 49% of the total computation time of the CNN model. RF was relatively insensitive to preprocessing but yielded higher absolute errors, whereas PLSR was limited by its linear nature and failed to capture the nonlinear characteristics of complex energy spectra. In conclusion, the CNN–Meta hybrid model demonstrates superior performance in both accuracy and efficiency, providing a reliable and effective approach for the rapid identification of radionuclides and quantitative analysis of activity in complex energy spectra. Full article

Background: Extracorporeal membrane oxygenation (ECMO) can affect the disposition of drugs, notably by sequestering them in a circuit. This review aimed to provide a comprehensive summary of existing ex vivo studies investigating the impact of contemporary ECMO circuits on drug sequestration, and to examine the associations between the physicochemical properties of drugs, the features and settings of ECMO devices, and the extent of drug sequestration. Method: A comprehensive search was conducted to identify ex vivo studies that determined drug concentrations in ECMO circuits. Studies that did not allow for the proper assessment of drug loss by degradation were excluded. Drug characteristics and experimental conditions were recorded. Drug sequestration in the circuit was calculated as the difference between the drug loss measured in the ECMO circuit and the drug loss due to spontaneous degradation measured under control conditions. To identify predictors of drug sequestration, a stepwise multiple linear meta-regression was applied by testing the physicochemical properties of drugs and ECMO device features/settings. Results: A total of 40 studies were identified, of which 21 were included in the analysis, covering 41 drugs. The Maquet membrane oxygenator was the most used brand (73%). About half of the circuits were adult and half were pediatric. Our final regression model retained lipophilicity, and to a lesser extent ionization at a physiological pH, as significant predictors of drug sequestration (R² 0.44, relative standard error 23%). Protein binding had no additional effect. Anti-infectives were the most studied class of drugs (n = 28). Antibiotics were overall not significantly sequestered, while lipophilic drugs such as posaconazole, voriconazole, paracetamol, fentanyl, sufentanil, propofol, thiopental, dexmedetomidine and amiodarone were highly sequestered (≥50%). However, this sequestration occurred mainly within the first few hours of the experiments, possibly reflecting a saturation effect. Conclusions: Lipophilic drugs are significantly sequestered in ex vivo ECMO circuits, although this effect may be limited by early saturation. Full article

Improving nitrogen efficiency in dairy cattle requires a better understanding of the dietary and ruminal factors that regulate nitrogen partitioning. This meta-analysis evaluated the effects of ruminal pH and dietary characteristics on microbial nitrogen (MN), non-microbial non-ammonia nitrogen (NANMN), and NAN flows to the small intestine in lactating cows. A dataset was assembled from 44 peer-reviewed in vivo studies (163 data points), with dietary intake and ruminal variables standardized across trials. Mixed linear models were developed for each N fraction, and the relative contribution of each predictor to the explained variance was assessed using semipartial coefficients of determination (pR²). Efficiency of microbial protein synthesis (EMPS), rumen undegraded protein intake (RUPI), and organic matter truly digested in the rumen (OMTDR) were the most relevant predictors of NANMN and NAN. Although the ruminal pH itself was not statistically significant in the models, the dietary components that influenced pH, starch concentration, physically effective fiber, and RUP supply were strongly associated with nitrogen flow profiles. Nitrogen utilization was not affected by ruminal pH, but rather by the combination of fermentable substrates and the supply of rumen-degradable and undegraded protein. Full article

Effective spam detection systems are essential in online social media networks (OSNs) and cybersecurity, and they directly influence the quality of decision-making pertaining to security. With today’s digital communications, unsolicited spam degrades user experiences and threatens platform security. Machine learning-based spam detection systems offer an automated defense. Despite their effectiveness, such methods are frequently hindered by the “black box” problem, an interpretability deficiency that constrains their deployment in security applications, which, in order to comprehend the rationale of classification processes, is crucial for efficient threat evaluation and response strategies. However, their effectiveness hinges on selecting an optimal feature subset. To address these issues, we propose a lightweight, explainable spam detection model that integrates a nature-inspired optimizer. The approach employs clean data with data preprocessing and feature selection using a swarm-based, nature-inspired meta-heuristic Rüppell’s Fox Optimization (RFO) algorithm. To the best of our knowledge, this is the first time the algorithm has been adapted to the field of cybersecurity. The resulting minimal feature set is used to train a supervised classifier that achieves high detection rates and accuracy with respect to spam accounts. For the interpretation of model predictions, Shapley values are computed and illustrated through swarm and summary charts. The proposed system was empirically assessed using two datasets, achieving accuracies of 99.10%, 98.77%, 96.57%, and 92.24% on Dataset 1 using RFO with DT, KNN, AdaBoost, and LR and 98.94%, 98.67%, 95.04%, and 94.52% on Dataset 2, respectively. The results validate the efficacy of the suggested approach, providing an accurate and understandable model for spam account identification. This study represents notable progress in the field, offering a thorough and dependable resolution for spam account detection issues. Full article

Basalt fiber-reinforced composites are increasingly utilized in sustainable construction due to their high strength, environmental benefits, and durability. However, the long-term tensile performance of these composites in alkaline environments remains a critical concern. This study investigates the degradation performance of basalt fibers exposed to different alkaline solutions (NaOH, KOH, and Ca(OH)₂) with varying concentrations (5 g/L, 15 g/L, and 30 g/L) over various exposure periods (7, 14, and 28 days). The performance assessment is carried out by mechanical properties, including tensile strength and modulus of elasticity, using experimental techniques and Response Surface Methodology (RSM) to find influential factors on tensile performance. The findings indicate that tensile strength degradation is highly dependent on alkali type and concentration, with Ca(OH)₂-treated fibers exhibiting superior mechanical retention (max tensile strength: 938.94 MPa) compared to NaOH-treated samples, which showed the highest degradation rate. Five machine learning (ML) models, including Tree Random Forest (TRF), Function Multilayer Perceptron (FMP), Lazy IBK, Meta Bagging, and Function SMOreg (FSMOreg), were also implemented to predict tensile strength based on exposure parameters. FSMOreg demonstrated the highest prediction accuracy with a correlation coefficient of 0.928 and the lowest error metrics (RMSE 181.94). The analysis boosts basalt fiber durability evaluations in cement-based composites. Full article

Wastewater treatment plants (WWTPs) function as engineered ecosystems in which microbial consortia mediate nutrient cycling, xenobiotic degradation, and heavy metal detoxification. This review discusses a forward-looking roadmap that integrates microbial ecology, multi-omics diagnostics, and artificial intelligence (AI) for next-generation treatments. Meta-analyses suggest that a globally conserved core microbiome indicates sludge functions, with high predictive value for treatment stability. Multi-omics approaches, including metagenomics, metatranscriptomics, and environmental DNA (eDNA) profiling, have integrated microbial composition with greenhouse gas (GHG) emissions, showing that WWTPs contribute 2–5% of anthropogenic nitrous oxide (N₂O) emissions. Emerging AI-enhanced eDNA models have achieved >90% predictive accuracy for effluent quality and antibiotic resistance gene (ARG) prevalence, facilitating near-real-time monitoring and adaptive control of effluent quality. Key advances include microbial strategies for degrading organic pollutants, pesticides, and heavy metals and monitoring industrial effluents. This review highlights both translational opportunities, including engineered microbial consortia, AI-driven digital twins and molecular indices, and persistent barriers, including ARG dissemination, resilience under environmental stress and regulatory integration. Future WWTPs are envisioned as adaptive, climate-conscious biorefineries that recover resources, mitigate ecological risks, and reduce their carbon footprint. Full article

Accurate brain tumor classification from MRI is often constrained by limited labeled data. We systematically compare conventional machine learning, deep learning, and few-shot learning (FSL) for four classes (glioma, meningioma, pituitary, no tumor) using a standardized pipeline. Models are trained on the Kaggle Brain Tumor MRI Dataset and evaluated across dataset regimes (100%→10%). We further test generalization on BraTS and quantify robustness to resolution changes, acquisition noise, and modality shift (T1→FLAIR). To support clinical trust, we add visual explanations (Grad-CAM/saliency) and report per-class results (confusion matrices). A fairness-aligned protocol (shared splits, optimizer, early stopping) and a complexity analysis (parameters/FLOPs) enable balanced comparison. With full data, Convolutional Neural Networks (CNNs)/Residual Networks (ResNets) perform strongly but degrade with 10% data; Model-Agnostic Meta-Learning (MAML) retains competitive performance (AUC-ROC ≥ 0.9595 at 10%). Under cross-dataset validation (BraTS), FSL—particularly MAML—shows smaller performance drops than CNN/ResNet. Variability tests reveal FSL’s relative robustness to down-resolution and noise, although modality shift remains challenging for all models. Interpretability maps confirm correct activations on tumor regions in true positives and explain systematic errors (e.g., “no tumor”→pituitary). Conclusion: FSL provides accurate, data-efficient, and comparatively robust tumor classification under distribution shift. The added per-class analysis, interpretability, and complexity metrics strengthen clinical relevance and transparency. Full article

With the rapid development of renewable energy and the continuous pursuit of efficient energy utilization, distributed photovoltaic power generation has been widely used in village-level microgrids. As a key platform connecting distributed photovoltaics with users, energy storage systems play an important role in alleviating the imbalance between supply and demand in VMG. However, current energy storage systems rely heavily on lithium batteries, and their frequent charging and discharging processes lead to rapid lifespan decay. To solve this problem, this study proposes a hybrid energy storage system combining supercapacitors and lithium batteries for VMG, and designs a hybrid energy storage scheduling strategy to coordinate the “source–load–storage” resources in the microgrid, effectively cope with power supply fluctuations and slow down the life degradation of lithium batteries. In order to give full play to the active support ability of supercapacitors in suppressing grid voltage and frequency fluctuations, the scheduling optimization goal is set to maximize the sum of the virtual inertia time constants of the supercapacitor. In addition, in order to efficiently solve the high-complexity model, the reason for choosing the snow goose algorithm is that compared with the traditional mathematical programming methods, which are difficult to deal with large-scale uncertain systems, particle swarm optimization, and other meta-heuristic algorithms have insufficient convergence stability in complex nonlinear problems, SGA can balance global exploration and local development capabilities by simulating the migration behavior of snow geese. By improving the convergence effect of SGA and constructing a multi-objective SGA, the effectiveness of the new algorithm, strategy and model is finally verified through three cases, and the loss is reduced by 58.09%, VMG carbon emissions are reduced by 45.56%, and the loss of lithium battery is reduced by 40.49% after active support optimization, and the virtual energy inertia obtained by VMG from supercapacitors during the scheduling cycle reaches a total of 0.1931 s. Full article

As a critical ecological security barrier in the Indo-China Peninsula, the Lao People’s Democratic Republic (Lao PDR) is increasingly threatened by forest degradation, frequent geological hazards, and intensified anthropogenic disturbances. To address the urgent need for a scientific evaluation of eco-geological environmental quality, this study develops a comprehensive assessment framework integrating multi-source remote sensing imagery, geological maps, and socio-economic datasets. A total of ten indicators were selected across four dimensions—geology, topography, ecology, and human activity. A stacking ensemble learning model was constructed by combining seven heterogeneous base classifiers—AdaBoost, KNN, Gradient Boosting, Random Forest, SVC, MLP, and XGBoost—with a logistic regression meta-learner. Model interpretability was enhanced using SHAP values to quantify the contribution of each input variable. The stacking model outperformed all individual models, achieving an accuracy of 91.14%, an F1 score of 93.62%, and an AUC of 95.05%. NDVI, GDP, and slope were identified as the most influential factors: vegetation coverage showed a strong positive relationship with environmental quality, while economic development intensity and steep terrain were associated with degradation. Spatial zoning results indicate that high-quality eco-geological zones are concentrated in the low-disturbance plains of the northeast and southeast, whereas vulnerable areas are primarily distributed around the Vientiane metropolitan region and tectonically active mountainous zones. This study offers a robust and interpretable methodological approach to support ecological diagnosis, zonal management, and sustainable development in tropical mountainous regions. Full article

In precision agriculture, semantic segmentation enhances the crop yield by enabling precise disease monitoring, targeted herbicide application, and accurate crop–weed differentiation. This enhances yield; reduces the overuse of herbicides, water, and fertilizers; lowers labor costs; and promotes sustainable farming. Deep-learning-based methods are particularly effective for crop and weed segmentation, and achieve potential results. Typically, segmentation is performed using homogeneous data (the same dataset is used for training and testing). However, previous studies, such as crop and weed segmentation in a heterogeneous data environment, using heterogeneous data (i.e., different datasets for training and testing) remain inaccurate. The proposed framework uses patch-based augmented limited training data within a heterogeneous environment to resolve the problems of degraded accuracy and the use of extensive data for training. We propose an attention-driven and hierarchical feature fusion network (AHFF-Net) comprising a flow-constrained convolutional block, hierarchical multi-stage fusion block, and attention-driven feature enhancement block. These blocks independently extract diverse fine-grained features and enhance the learning capabilities of the network. AHFF-Net is also combined with an open-source large language model (LLM)-based pesticide recommendation system made by large language model Meta AI (LLaMA). Additionally, a fractal dimension estimation method is incorporated into the system that provides valuable insights into the spatial distribution characteristics of crops and weeds. We conducted experiments using three publicly available datasets: BoniRob, Crop/Weed Field Image Dataset (CWFID), and Sunflower. For each experiment, we trained on one dataset and tested on another by reversing the process of the second experiment. The highest mean intersection of union (mIOU) of 65.3% and F1 score of 78.7% were achieved when training on the BoniRob dataset and testing on CWFID. This demonstrated that our method outperforms other state-of-the-art approaches. Full article

Motion blur is a complex phenomenon caused by the relative movement between an observed object and an imaging sensor during the exposure time, resulting in degradation in the image quality. Deep-learning-based methods, particularly convolutional neural networks (CNNs), have shown promise in motion deblurring. However, the small kernel sizes of CNNs limit their ability to achieve optimal performance. Moreover, supervised deep-learning-based deblurring methods often exhibit overfitting in their training datasets. Models trained on widely used synthetic blur datasets frequently fail to generalize in other blur domains in real-world scenarios and often produce undesired artifacts. To address these challenges, we propose the Spatial Feature Selection Network (SFSNet), which incorporates a Regional Feature Extractor (RFE) module to expand the receptive field and effectively select critical spatial features in order to improve the deblurring performance. In addition, we present the BlurMix dataset, which includes diverse blur types, as well as a meta-tuning strategy for effective blur domain adaptation. Our method enables the network to rapidly adapt to novel blur distributions with minimal additional training, and thereby improve generalization. The experimental results show that the meta-tuning variant of the SFSNet eliminates unwanted artifacts and significantly improves the deblurring performance across various blur domains. Full article

Mollusks are among the most ecologically and economically significant invertebrates; yet, their associated microbiomes remain understudied relative to those of other metazoans. This scoping review synthesizes the current literature on the diversity, composition, functional roles, and ecological implications of molluscan microbiomes, with an emphasis on three major groups: gastropods, bivalves, and cephalopods. Drawing on studies from terrestrial, freshwater, and marine systems, we identified the dominant bacterial phyla, including Proteobacteria, Bacteroidetes, and Firmicutes, and explored how microbiota vary across different habitats, diets, tissue types, and host taxonomies. We examined the contribution of molluscan microbiomes to host functions, including digestion, immune modulation, stress responses, and nutrient cycling. Particular attention was given to the role of microbiota in shell formation, pollutant degradation, and adaptation to environmental stressors. The review also evaluated microbial interactions at different developmental stages and under aquaculture conditions. Factors influencing microbiome assembly, such as the host’s genetics, life history traits, and environmental exposure, were mapped using conceptual and graphical tools. Applications of molluscan microbiome research in aquaculture, conservation biology, and environmental biomonitoring are highlighted. However, inconsistencies in the sampling methods, taxonomic focus, and functional annotations limit the generalizability across taxa. We identify key knowledge gaps and propose future directions, including the use of meta-omics, standardized protocols, and experimental validation to deepen insights. By synthesizing emerging findings, this review contributes to a growing framework for understanding mollusk–microbiome interactions and their relevance to host fitness and ecosystem health. It further establishes the importance of mollusks as model systems for advancing microbiome science. Full article

The integration of artificial intelligence (AI) in medical imaging has revolutionized diagnostic capabilities, yet the black-box nature of deep learning models poses significant challenges for clinical adoption. Current explainable AI (XAI) approaches, including SHAP, LIME, and Grad-CAM, predominantly focus on post hoc explanations that may inadvertently undermine clinical decision-making by providing misleading confidence in AI outputs. This paper presents a systematic review and meta-analysis of 67 studies (covering 23 radiology, 19 pathology, and 25 ophthalmology applications) evaluating XAI fidelity, stability, and performance trade-offs across medical imaging modalities. Our meta-analysis of 847 initially identified studies reveals that LIME achieves superior fidelity (0.81, 95% CI: 0.78–0.84) compared to SHAP (0.38, 95% CI: 0.35–0.41) and Grad-CAM (0.54, 95% CI: 0.51–0.57) across all modalities. Post hoc explanations demonstrated poor stability under noise perturbation, with SHAP showing 53% degradation in ophthalmology applications (ρ = 0.42 at 10% noise) compared to 11% in radiology (ρ = 0.89). We demonstrate a consistent 5–7% AUC performance penalty for interpretable models but identify modality-specific stability patterns suggesting that tailored XAI approaches are necessary. Based on these empirical findings, we propose a comprehensive three-pillar accountability framework that prioritizes transparency in model development, interpretability in architecture design, and a cautious deployment of post hoc explanations with explicit uncertainty quantification. This approach offers a pathway toward genuinely accountable AI systems that enhance rather than compromise clinical decision-making quality and patient safety. Full article