Search Results (12,378)

This paper evaluates machine learning (ML) classifiers for predicting preeclampsia (PE) and pregnancy-induced hypertension (PIH) using first-trimester screening data from 1383 pregnant women in Plovdiv, Bulgaria (2018–2020). Three classifiers—Logistic Regression (LR), Extra Trees Classifier (ETC), and Voting Classifier (VC)—are compared across multiple prediction targets and feature configurations. The impact of SMOTE oversampling strategies on model performance in the context of a severe class imbalance (2.46% PE prevalence) is assessed. Logistic Regression achieves the highest AUC of 0.853 for preterm PE prediction without oversampling, while SMOTE significantly improves tree-based models (ETC: +0.058 AUC). A non-screened control cohort of 533 patients is evaluated separately using maternal characteristics alone (AUC 0.751). The ML model showed promising discrimination for preterm PE in this local cohort and warrants direct comparison with FMF-based risk stratification in future studies. These results support further validation of ML-based tools as potential components of future clinical decision support systems to extend systematic PE screening in resource-constrained settings. Full article

Online flatness detection of aircraft rudder surfaces serves as a pivotal core procedure for ensuring the manufacturing precision, aerodynamic performance and operational safety of aeronautical components. Traditional plane fitting-based detection approaches are constrained by low detection efficiency, susceptibility to local optimal solutions, weak anti-noise robustness and limited automation capability, which fail to satisfy the micron-level high-precision online detection requirements for curved composite rudder surfaces in batch manufacturing scenarios. To address the aforementioned technical bottlenecks, this study proposes a bidirectionally coupled PSO-SA hybrid optimization algorithm for non-convex minimum zone flatness evaluation of curved rudder surfaces, which overcomes the unidirectional open-loop iteration limitation inherent in conventional serial PSO-SA composite frameworks. Two targeted algorithmic improvements are elaborated in this work: a residual-adaptive nonlinear inertia weight strategy, which dynamically balances global exploration and local exploitation capabilities based on the fluctuation characteristics of free-form surface measurement residuals; and a measurement noise-modified Metropolis acceptance criterion, which substantially enhances the algorithm’s anti-interference performance against on-machine trigger sampling noise. Integrating with the trigger-type on-machine detection hardware of computer numerical control (CNC) machine tools, an integrated online detection system is established to realize the full-process functions of point cloud data acquisition, error compensation, intelligent plane fitting and flatness error evaluation. Meanwhile, the complete technical workflow involving measurement path planning, probe calibration and algorithm iterative solution is systematically illustrated. Comparative simulation experiments implemented on the MATLAB platform demonstrate that the proposed algorithm exhibits superior performance in convergence speed, fitting accuracy and optimization stability over five mainstream algorithms, including standard particle swarm optimization (PSO), standard simulated annealing (SA), comprehensive learning PSO (CLPSO), adaptive cooling SA and conventional serial PSO-SA. On-machine physical measurement experiments are conducted on 24 aircraft rudder workpieces covering aluminum alloy skins and assembled riveted components. After multi-dimensional systematic calibration, the overall detection error of the developed system is controlled within 1 μm. The experimental results indicate that the average flatness error calculated by the proposed bidirectionally coupled PSO-SA algorithm is 29.7 μm, which is 30.1% and 38.5% lower than that of standard PSO and standard SA, respectively, fully complying with the aviation flatness tolerance specification of 0.1–0.3 mm. Moreover, the full detection cycle for a single workpiece is only 2.1 min, achieving a 34.4% reduction in detection time compared with standard PSO and effectively improving the efficiency of online in-process inspection. One-way analysis of variance (ANOVA) combined with Tukey’s posthoc test further verifies that the accuracy superiority of the proposed algorithm is statistically significant. This research provides a targeted theoretical basis and complete engineering implementation scheme for intelligent flatness detection of aerospace curved thin-walled parts, and offers a valuable technical reference for form and position error evaluation of irregular industrial components under noisy measurement conditions. Full article

In wheat threshing, reducing total grain loss and energy consumption is crucial for both economic and sustainable food security. This study investigates the effects of threshing drum type (straight and helical row), drum peripheral speed (36.73–48.98 m s⁻¹), and drum-concave clearance (35–50 mm) on total grain loss and specific fuel consumption in a stationary threshing machine using a full factorial design. The optimum machine settings (drum type, peripheral speed and drum–concave clearance) that simultaneously minimize these two outputs were then determined. We systematically compared three surrogate modelling approaches—Response Surface Model (RSM), Gaussian Process Regression (GPR), and Artificial Neural Network (ANN)—to identify the most effective method for small-dataset optimization in threshing machine design. The best model was selected through cross-validation, and optimization was performed using the NSGA-II multi-objective genetic algorithm. GPR yielded the highest prediction accuracy for both outputs (R² in prediction data: 0.99 for total grain loss and 0.91 for specific fuel consumption). Multi-objective optimization revealed a conflict between the two objectives; the best balance was achieved for the helical drum at a peripheral speed of approximately 41.5 m s⁻¹ and a drum–concave clearance of 50 mm (predicted total grain loss approximately 3.7%, specific fuel consumption approximately 2.98 mL kg⁻¹). Compared to the straight-row drum, the helical drum provided lower losses and fuel consumption, as well as approximately 3.5 times wider safe operating range. It should be noted that this optimum was predicted by the surrogate model and agreed closely with the best measured treatment; it was not confirmed by an independent validation experiment. The results demonstrated that combining a surrogate model with a genetic algorithm is an effective tool for optimizing threshing machine parameters. Full article

Background: Digital behavioral phenotyping of autism spectrum disorder (ASD) offers a promising approach for developing more scalable diagnostic frameworks across diverse global contexts. Machine learning (ML) models show promise for ASD diagnosis using behavioral videos, but critical questions remain regarding whether models trained on data from one country work in another, and how the background of the raters affects the accuracy. Our work addresses these questions by testing whether ML models can accurately diagnose ASD across different populations and rater groups. Methods: This work evaluates the performance of a supervised ML framework for binary classification of ASD versus non-ASD [speech, language and communication disorders (SLC) + neurotypical (NT)] in a cohort of 227 children in Bangladesh. We first assessed the cross-domain model transferability of a clinical-instrument-trained logistic regression model (LR-9) on behavioral ratings that were based on videos of Bangladeshi children interacting with caregivers and toys at two major child development centers in Dhaka, Bangladesh. We then trained five diverse classifiers (Logistic Regression, Random Forest, XGBoost, SVM, and RuleFit) on the full annotated Bangladeshi dataset. Using SHAP-based consensus elbow feature selection, we identified a compact set of features that maintained the performance. Finally, we developed ensemble models to improve predictive stability. Results: The LR-9 model, originally trained on U.S. clinical instrument data, was evaluated on video-based behavioral ratings from 214 Bangladeshi children. When tested on Bangladeshi clinician ratings, the LR-9 model achieved a sensitivity of 86.1% (95% CI: [0.78–0.93]) and AUC of 0.79 (95% CI: [0.73–0.86]). The distinction across rater groups was between trained raters (clinicians and students) and crowd workers, who showed lower sensitivity 28.5% (95% CI: [0.21, 0.39]). When tested on the aggregated ratings from all groups, the model achieved an AUC of 0.78 (95% CI: [0.72–0.84]). Inter-rater reliability followed the same pattern: individual agreement was fair (Krippendorff’s α = 0.26), but the multi-rater consensus was reliable (ICC(1,k) = 0.84), with Bangladeshi clinicians showing the highest agreement (α = 0.34) and crowd workers the lowest (α = 0.20). We then trained new models directly on the Bangladeshi ratings. All model types achieved similar AUC values (0.86–0.89), with overlapping confidence intervals. Using just 8–11 key behaviors kept the similar performance while cutting the features by 66–75%. Combining ensembles gave similar results (e.g., Bayesian averaging: AUC 0.88 [0.78, 0.95]) but with more stable predictions. Conclusion: This study provides evidence that mobile video-based ASD diagnosis can achieve comparable performance (AUC: 0.89 [0.76, 0.96]) to models trained on clinical instrument data. This work contributes to the development of broader adaptable autism detection tools, bypassing the dependence on traditional clinical instrument data. Full article

Mycotoxin contamination by aflatoxin B₁ (AFB₁) and deoxynivalenol (DON) in maize silage threatens feed safety, requiring rapid, non-destructive monitoring tools. This study developed a visible-light machine vision approach combined with machine learning to quantify AFB₁ and DON and classify contamination levels. A total of 210 silage samples were imaged, and 111 RGB-based color and texture features were extracted, followed by correlation analysis and model-based feature selection. Using the 10 selected features, support vector regression (SVR) achieved the best quantitative performance for AFB₁ (R² = 0.9945, RMSE = 3.18 µg·kg⁻¹), while XGBoost performed best for DON (R² = 0.9816, RMSE = 45.50 µg·kg⁻¹). For classification, random forest and XGBoost correctly identified AFB₁ contamination levels with an Accuracy of 90.48%, whereas SVM achieved 97.62% Accuracy for DON. Comparison of correlation-based and model-based feature importance confirmed the complementary value of statistically significant and nonlinearly predictive features. Biological interpretation suggested that color and texture responses may reflect fungal pigmentation/browning and multiscale surface-structure alterations, respectively, with selected texture descriptors changing earlier than color descriptors during aerobic exposure. The proposed low-cost RGB imaging strategy, coupled with multi-feature fusion, offers a promising approach for high-throughput preliminary mycotoxin screening in feed production. Full article

In recent years, interest in machine learning applications has grown rapidly, particularly in the medical domain, where large amounts of data are available for training these models. This review focuses on the potential of machine learning for early diagnosis and prognosis of chronic kidney disease (CKD) by examining the most recent literature. Articles published from 2016 to 2025 were collected from online databases such as PubMed, Web of Science, and Embase. After abstract and full-text screening, 57 articles were included in the results section. Machine learning was applied to clinical and laboratory data, medical imaging, urine samples, retinal images, and at-home measurements to diagnose CKD and predict CKD progression and related complications. Although many studies reported high discriminatory performance, the evidence base was dominated by retrospective, single-center, and methodologically heterogeneous studies, with frequent high-risk-of-bias findings and limited external validation. Although machine learning has been applied to a broad range of diagnostic and prognostic tasks in CKD, the current evidence is dominated by retrospective, single-center, and methodologically heterogeneous studies with a high risk of bias and limited external validation. Furthermore, most published models are not yet sufficiently validated for clinical deployment. Before these tools can be adopted in routine care, prospective, multicenter studies are required that report calibration and clinical utility, adhere to established reporting standards, and demonstrate added value over the current standard of care. Full article

Background: Postoperative hemoglobin (Hb) reduction reflects the physiologic extent of perioperative blood loss after total knee arthroplasty (TKA). Whereas previous studies have relied on transfusion as a binary endpoint, transfusion decisions are highly variable across institutions, obscuring the underlying hematologic trajectory. This study aimed to develop and interpret machine learning (ML) models to characterize and quantify the determinants of postoperative Hb reduction in a standardized cohort of non-transfused TKA patients. Methods: A retrospective cohort of 866 patients who underwent primary TKA under a standardized operative protocol—with identical cemented posterior-stabilized implants and uniform cementing technique—was analyzed (1 January 2014–31 March 2024). During the study period, a consistent 1 g intra-articular tranexamic acid (TXA) regimen administered through the drain was introduced and applied to a subset of patients, allowing TXA use to be modeled as a binary predictor. Four ML algorithms (Linear Regression, Random Forest, XGBoost, and Stacking Regressor) were trained using preoperative, demographic, and perioperative variables. Fivefold cross-validation assessed model performance, and SHapley Additive exPlanations (SHAP) values were used to identify influential predictors and enhance interpretability. Results: Across all ML models, preoperative Hb emerged as the strongest determinant of postoperative Hb reduction, followed by TXA use, body mass index (BMI), and platelet count. Ensemble models captured non-linear and interacting effects more effectively than linear regression. Test-set performance was modest (best R² = 0.330), consistent with the influence of unmeasured physiologic factors such as hidden blood loss, fluid dynamics, and inflammatory responses. Accordingly, the primary value of the framework lies in the exploratory and transparent assessment of determinant importance rather than in individual-level prediction. Conclusions: This study provides an interpretable, exploratory ML framework for identifying factors associated with percentage Hb reduction after TKA. Preoperative Hb was the dominant determinant, while TXA use and higher BMI were recurrently associated with smaller predicted percentage reductions. Given the modest test-set performance and the absence of external validation and clinical utility assessment, the models should not be interpreted as tools for individual-level prediction or clinical decision making. Full article

The digital transformation of manufacturing associated with Industry 4.0 and the human-centric paradigm of Industry 5.0 requires advanced predictive tools that can improve the performance, reliability, and sustainability of tribological systems. In this context, accurate prediction of friction behavior in rolling contacts is crucial for intelligent monitoring and optimization of bearing components. This study presents a machine learning-based methodology for predicting the coefficient of friction in rolling contact of 100Cr6 steel bearing discs as a function of surface roughness and rolling distance parameters. Experimental studies were conducted using discs with different surface topography under controlled rolling contact conditions. Surface roughness characteristics and rolling distance data were correlated with experimentally measured friction coefficients to create a comprehensive dataset for artificial intelligence (AI) modeling. Several dozen machine learning (ML) algorithms, including random forest, support vector regression, and artificial neural networks, were developed and comparatively evaluated to capture nonlinear relationships between operational and surface parameters. The predictive ability of the models was assessed using statistical metrics such as the coefficient of determination (R²), mean absolute error (MAE), and root mean square error (RMSE). The obtained results demonstrate that ML methods provide high prediction accuracy and effectively identify the combined effects of surface roughness and rolling distance on rolling contact friction. Feature importance analysis revealed that roughness parameters dominate friction behavior during the run-in phase, while rolling distance becomes increasingly important under stabilized operating conditions. The proposed approach supports the development of intelligent tribological systems, predictive maintenance strategies, and data-driven decision-making frameworks aligned with Industry 4.0 and Industry 5.0 concepts. The presented methodology can contribute to the implementation of intelligent manufacturing solutions, the sustainable operation of bearing systems, and AI-assisted monitoring of machine components in modern industrial environments. Full article

Drought is a recurrent hazard in the Chi River Basin, northeastern Thailand, causing agricultural losses despite its low-lying floodplain setting. This study developed a machine learning framework integrating Sentinel-2 indices (SMI, NDVI, MNDWI, VCI, NDMI) and static spatial variables (DEM, TWI, prox2river, slope) aggregated to H3 hexagonal grids (≈5.96 km²) for village-relevant analysis. Drought reports from 1482 observations (2019–2024) were aggregated to 203 grid cells. Three models—Random Forest, XGBoost, and LightGBM—were evaluated using temporal (training: 2019–2023; test: 2024) and spatial holdout validation. LightGBM achieved the best performance with AUC = 0.783 (temporal) and 0.714 (spatial), accuracy = 78.3%, and balanced accuracy = 76.4%. Five-class severity classification showed declining accuracy from 71.4% (Very Low) to 25.0% (Severe), limited by rare event sample sizes. SHAP analysis revealed static topographic variables dominated importance (76.1%) over remote sensing indices (23.9%), with weak individual correlations (|r| < 0.10). The framework is best characterized as a drought risk mapping tool for identifying persistently vulnerable areas rather than an operational early warning system. The methodology is transferable to similar floodplain environments with local re-estimation and validation. Full article

The metalworking industry faces growing regulatory pressure to quantify and verify its product carbon footprint (PCF), driven by frameworks such as the European Union’s Carbon Border Adjustment Mechanism (CBAM) and emissions trading schemes (ETS). Traditional static accounting methods, based on generic emission factors and annual averages, are insufficient given the dynamic, multi-stakeholder nature of the sector’s supply chains. This study presents a systematic review conducted under the PRISMA 2020 protocol to identify, classify and critically evaluate the digital tools and technologies used to calculate, manage and verify PCF in this sector. From 495 records screened, 53 thematically relevant studies were analyzed and 5 sector-specific cases examined in depth. The results indicate that the Internet of Things (IoT) and smart sensor networks constitute the primary data-capture layer, while Machine Learning, Big Data and Digital Twins are the predominant processing technologies. Blockchain and verifiable digital credentials emerge as governance mechanisms that ensure the transparency, auditability and immutability of emissions inventories, enabling compliance through Digital Product Passports (DPPs). We conclude that digital decarbonization requires interoperable architectures integrating real-time capture, distributed traceability and common semantic standards; viability in small and medium-sized enterprises (SMEs) and interoperability across heterogeneous platforms remain the main research gaps. Full article

High-entropy alloys (HEAs) offer attractive combinations of mechanical performance, thermal stability, and compositional flexibility, making them promising candidates for advanced structural applications. Laser-based additive manufacturing, particularly laser powder bed fusion (LPBF) and laser-directed energy deposition (LDED), enables the fabrication of geometrically complex HEA components with non-equilibrium microstructures. However, the distinct thermal histories of LPBF and LDED, with typical cooling rates of approximately 10⁵–10⁷ K s⁻¹ and 10²–10⁴ K s⁻¹, respectively, strongly govern solidification behavior, elemental segregation, residual stress development, defect formation, and mechanical properties. Although previous reviews have discussed additively manufactured HEAs, an integrated framework linking composition design, printability, LPBF/LDED processing, microstructural evolution, post-processing, and industrial qualification remains limited. Therefore, this review establishes a unified composition–process–structure–property framework for laser additively manufactured HEAs. Fundamental HEA concepts, LPBF/LDED process characteristics, solidification behavior, phase formation, defect evolution, and mechanical performance from ambient to elevated temperatures are systematically discussed across representative FCC, refractory, and dual-phase HEA systems. This review emphasizes that printability should be considered during alloy design by correlating composition-dependent solidification characteristics, cracking susceptibility, phase stability, and defect formation with mechanical performance. Post-processing treatments are shown to modify residual stress, microsegregation, precipitation behavior, porosity, and deformation mechanisms, although their benefits must be balanced against thermal softening or brittle phase formation. Finally, CALPHAD, integrated computational materials engineering (ICME), machine learning (ML), and in situ monitoring are identified as promising tools for accelerating alloy and process optimization, while reproducible process windows, defect-control criteria, databases, and qualification protocols remain essential for industrial implementation. Full article

Accurate tool wear monitoring can effectively improve machining quality and reduce tool costs. In this paper, tool wear monitoring was studied using multi-signal fusion based on an integrated machine learning model. Firstly, tool holder strain, acceleration, and AE signals are selected as tool wear monitoring signals based on different types of physical quantities and acceptable installation convenience. Tool wear experiments are conducted to synchronously acquire these signals. After the signal denoising process, 102 features from these signals are extracted, which include time domain, frequency domain, and wavelet packet time-frequency domain features. Then, 15 key features are selected using the minimum redundancy maximum relevance (mRMR) method to realize multi-signal fusion at the feature level. Subsequently, an integrated machine learning model is proposed for tool wear monitoring. Three complementary models, extra trees, random forest, and ridge regression, are selected to construct the integrated model. The results indicate that this strategy achieves a tool wear state classification accuracy of 96.77%, exhibiting higher accuracy than single models. Full article

Aluminum honeycomb structures are widely used in the aeronautical, aerospace, marine, and automotive industries due to their excellent stiffness-to-weight ratio. However, machining these structures remains highly challenging because their thin, highly flexible cell walls are susceptible to plastic deformation and geometric defects. To overcome these limitations, this study proposes an innovative machining approach that combines longitudinal-torsional ultrasonic vibration-assisted machining (LT-UVAM) with a 55-tooth CZD10 cutting tool. A three-dimensional finite element model was developed using Abaqus/Explicit 2017 to simulate the dynamic interactions between the cutting tool and the honeycomb cell walls during the milling process. Following experimental validation on a high-speed machining center, the model was employed to investigate the effects of cutting and vibration parameters on the machining performance. The results demonstrate that longitudinal-torsional ultrasonic vibration coupling significantly reduces the cutting forces, resulting in a 26% to 42% reduction in the axial force component (Fz). Furthermore, vibration assistance effectively limits cell wall deflection, reducing the stress levels by up to 60% in the thinnest walls while maintaining them below the critical Euler buckling load. Furthermore, an ultrasonic vibration frequency of 22.5 kHz almost completely eliminates plastic deformation, while a vibration amplitude of 25 µm significantly reduces tool wear by promoting intermittent tool–workpiece contact, thereby facilitating chip evacuation. Ultimately, the LT-UVAM process produces finer and more uniform chips, leading to improved machining quality, enhanced dimensional accuracy, and extended tool life. Full article

Water quality analysis plays an essential role in maintaining the health and sustainability of river ecosystems, especially in semi-arid regions like the Arroyo Colorado Watershed in South Texas. Since the river is a vital source of water supply for local communities, agriculture, and wildlife, it faces significant challenges and pollution from land use changes, climate variation, and agricultural runoff. Continuous monitoring and assessment of water quality parameters and their temporal variability are essential to ensure the drinking water supply and aquatic ecosystem health. However, comprehensive laboratory-based water quality investigations are often constrained by higher costs, logistical complexity, and limited manpower. As a result, monitoring datasets are often not available for all water quality parameters, or the datasets may be incomplete. To address such challenges, the objective of this study was to evaluate the potential of water quality index (WQI)-based assessment supported by machine learning algorithms as an alternative decision-support tool for water quality evaluation. The analysis compared four monitoring stations in the Austin and Arroyo Colorado Watersheds, with particular emphasis on one gauging station at Port Harlingen. Datasets were collected from the Texas Commission of Environmental Quality (TCEQ). A complete exploratory data analysis (EDA) was performed to understand the TCEQ water quality datasets containing sixteen parameters, and seven water quality parameters were selected based on multicollinearity checks. It was observed that seven independent water quality parameters (dissolved oxygen, ammonia, nitrate, phosphorus, temperature, fecal coliform, and residual non-filterable material concentrations) were identified as sufficient to define the WQI of the Austin monitoring stations. Moreover, U.S. Environmental Protection Agency (EPA)-based guidelines were utilized to scale individual parameters to a range of 0–100 to remove their magnitude and correlation-based bias. These parameters were further analyzed using machine learning techniques, i.e., principal component analysis, K-means, and one-class support vector machine, to compute the relative importance based on their fluctuation within the temporal dataset. Finally, the mean WQI model was developed for Port Harlingen and achieved a strong agreement with the National Sanitation Foundation (NSF) WQI (R² = 0.91). These findings demonstrate the applicability of the proposed data-driven WQI framework for regional water quality assessment and comparative analysis across watersheds. Full article

This study employs machine learning methods to estimate possible source regions from Andean lead isotopes, a novel approach for tracing the provenance of metal artifacts in this region. Two Random Forest coordinate models were trained to approximate geographic provenance. The latitude model uses ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, and ²⁰⁸Pb/²⁰⁴Pb isotope predictors, plus three derived ratios. The longitude model uses the same predictors together with the latitude predicted by the first model, allowing the regression to incorporate broad spatial coherence in the Andean ore-lead system. The Random Forest models reached, on validation, 3.50° and 1.94° MAE for latitude and longitude, respectively. These moderate values support the use of the framework as a screening tool that generates a ranked shortlist of nearest isotopic matches, rather than as a source of single definitive coordinates. Additionally, K-means clustering and Euclidean distance analysis were used to link artifact isotope compositions to known sources. The models’ limitations and scope were documented to ensure appropriate use and interpretation. Full article