Search Results (5,180)

Crack detection in composite structures remains a central challenge in structural health monitoring, particularly when sensing must rely on a small number of embedded multiplexed fiber Bragg gratings (FBGs). Here, we present a physics-guided convolutional neural network (CNN) framework for crack-associated event detection from multiplexed FBG interrogator signals acquired during the three-point bending of glass-fiber-reinforced polymer (GFRP) beams. The dataset was constructed from raw interrogator recordings and synchronized force–displacement metadata while preserving the cracked and non-cracked loading stages present in the experiments. Each candidate response was encoded by 13 synchronized optical, loading, and mechanics-guided descriptors, including Euler–Bernoulli expected strain and residual terms, where the residual denotes the difference between the measured response and the elastic response predicted by beam theory. A compact one-dimensional CNN operating on 30-response sequences was evaluated on 64 experimental runs under strict leave-one-run-out validation. At the selected operating point, the model reached window-level precision of 0.900, recall of 0.910, F1 score of 0.905, and balanced accuracy of 0.942, while the corresponding run-level decision reached a precision of 0.833, a recall of 1.000, an F1 score of 0.909, and a balanced accuracy of 0.969. Bootstrap resampling over runs yielded 95% confidence intervals of 0.787–0.978 for window-level F1 and 0.769–1.000 for run-level F1. To probe generalization beyond the initial fabrication batch, the final frozen pipeline was also tested once on seven later-batch runs from two newly manufactured specimens, where it reached a window-level precision of 0.908, a recall of 1.000, an F1 score of 0.952, a balanced accuracy of 0.969, an ROC-AUC of 0.979, a PR-AUC of 0.955, and perfect run-level classification. These results show that a compact sequence CNN, enriched with mechanics-guided strain interpretation, can extract robust crack-event signatures from multiplexed FBG measurements while preserving a simple and reproducible modeling pipeline. Full article

The electrocatalytic reduction of CO₂ (CO₂RR) into value-added chemicals represents a promising strategy for achieving carbon-neutral energy conversion. However, it is fundamentally limited by sluggish reaction kinetics, insufficient product selectivity, and the competitive hydrogen evolution reaction (HER). Herein, density functional theory (DFT) calculations were employed to systematically investigate transition-metal single-atom catalysts anchored on boron and nitrogen co-doped graphene (TM@BNG), with the aim of elucidating the role of heteroatom-induced coordination engineering in modulating catalytic performance. The results demonstrate that B, N co-doping effectively tailors the electronic structure of the metal active sites, thereby optimizing the adsorption energetics of key intermediates and dictating the reaction pathways. Among the 27 candidates examined, Pd@BNG, Ag@BNG, Sc@BNG, Cu@BNG, Co@BNG, Cd@BNG, and Y@BNG exhibit superior catalytic activity and selectivity toward CO or HCOOH production, featuring low limiting potentials down to −0.06 V while simultaneously suppressing HER. Mechanistic analysis reveals that product selectivity is governed by the relative stabilization of *COOH and *HCOO intermediates during the initial proton-coupled electron transfer step. Furthermore, a physically interpretable descriptor (φ), derived from intrinsic electronic properties using machine-learning approaches, establishes a volcano-type correlation with the limiting potential and provides an effective activity-screening criterion within the investigated TM@BNG dataset. Collectively, these findings clarify the electronic-structure modulation of TM@BNG single-atom catalysts and provide a system-specific framework for screening related B/N-coordinated CO₂RR electrocatalysts. Full article

Background: Oromucosal films are thin polymeric dosage forms designed to hydrate rapidly in the oral cavity and enable local or systemic drug delivery. Their performance depends on coupled processes including wetting, swelling, polymer relaxation, matrix softening, and structural failure. Because these phenomena depend strongly on the formulation composition and polymer-network organization, a mechanistic framework linking conventional characterization data to film performance is needed. This study aimed to develop a Madelung-type multifractal swelling–disintegration–release-readiness model for chitosan/hydroxypropyl methylcellulose (HPMC) films and to examine its relevance using a twelve-formulation experimental series. Methods: Twelve films based on chitosan (film-forming polysaccharide), HPMC K-4M (hydrophilic swelling polymer), glycerin (plasticizer), and starch (disintegrant) were prepared via solvent casting. The films were characterized for loss on drying, surface pH, mass and thickness uniformity, wetting time, swelling behavior, structural-disintegration onset, elongation response, rupture resistance, folding endurance, and surface roughness. The proposed model described water uptake, swelling-front motion, matrix integrity, local release-readiness activation, and hydration-induced loading as coupled fields across the film thickness. Results: Formulation markedly influenced hydration behavior, mechanical performance, structural stability, and surface morphology. Films F2 and F7 emerged as the most promising complementary unloaded matrix platforms for future active-compound incorporation and experimental release evaluation. F2 behaved as a high-swelling, mechanically stable starch-free matrix, whereas F7 combined faster wetting, starch-assisted structural destabilization, and favorable flexibility. Conclusions: This framework provides a quantitative link between empirical film characterization and formulation-level mechanistic interpretation. It translates conventional characterization parameters into descriptors related to the apparent water penetration, swelling capacity, matrix-failure tendency, mechanical suitability, and structural heterogeneity. The present results support candidate selection for future Active Pharmaceutical Ingredient-loaded studies but do not constitute validation of drug-release kinetics. Full article

Background/Objectives: Osteoporosis is a major contributor to vertebral compression fractures (VCFs) and other skeletal complications, yet quantitative bone mineral density (BMD) assessment using dual-energy X-ray absorptiometry (DEXA) is not routinely available in many spine surgery workflows. This study proposes an anatomically guided and optimization-refined radiomics framework for opportunistic osteoporosis assessment from routine lumbar spine magnetic resonance imaging (MRI). Methods: The proposed pipeline employs a hierarchical template-matching strategy to automatically localize the L1–L4 vertebral region, followed by an optimization-based refinement procedure that adapts vertebral regions of interest (ROIs) using intensity, texture, boundary, and geometric constraints. Anatomically consistent ROIs are subsequently used for extraction of handcrafted radiomic descriptors, including statistical, textural, gradient-based, frequency-domain, and shape-related features. The extracted features were evaluated using conventional support vector classification (SVC) and a NeurodynamicSVMRBFTanh classification framework for osteoporosis-related classification. Results: Experimental results demonstrated robust and anatomically consistent vertebral localization across heterogeneous lumbar MRI acquisitions. The NeurodynamicSVMRBFTanh framework achieved the best screening-oriented performance, yielding 85.2% classification accuracy and 100.0% sensitivity on an independent test set. In addition, exploratory BMD regression analysis demonstrated the feasibility of estimating DEXA-derived BMD directly from MRI-derived radiomic features, achieving mean absolute percentage errors of approximately 15–20% across lumbar vertebral levels. Conclusions: These findings suggest that anatomically guided vertebral radiomics extracted from routine lumbar spine MRI contain clinically meaningful information associated with osteoporosis-related bone quality changes and may provide a practical tool for automated opportunistic osteoporosis assessment in settings where DEXA measurements are unavailable. Full article

Regional landslide susceptibility mapping commonly relies on static conditioning factors, including terrain, geology, hydrology, land cover and human activity. These factors describe long-term instability settings but cannot directly represent recent or ongoing ground deformation. Interferometric Synthetic Aperture Radar (InSAR) can provide spatially distributed deformation information, yet mountainous InSAR evidence is affected by uneven observation availability, vegetation decorrelation, terrain-induced geometric distortion and confounding between observation support and static environmental conditions. These issues make it difficult to determine whether radar-derived variables represent deformation signals or mainly indicate where observations are reliable. This study develops a scale-sensitive and confounding-audited Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) evidence representation framework for regional landslide susceptibility mapping. In Pingwu County, China, a 1607-record landslide inventory was converted into 1579 unique 30 m landslide cells, with 1393 for training and 186 for inventory-concentration validation. Fourteen static factors formed the baseline model. Deformation evidence from 98 Sentinel-1A descending acquisitions was represented using point-based line-of-sight (LOS) variables, neighbourhood component descriptors, a compressed deformation-intensity and observation-availability index, and a separated deformation intensity (DI) plus observation availability (RI) representation. Results show that the static factors already provided a strong first-order baseline. Adding SBAS-InSAR evidence did not produce uniform paired improvements across models, metrics or neighbourhood scales. In 10 km spatial-block cross-validation, random forest models using static factors, 300 m component descriptors and 300 m DI + RI features achieved similar mean area under the receiver operating characteristic curve values of 0.948, 0.949 and 0.950, with mean Matthews correlation coefficient values of 0.791, 0.795 and 0.795. Broader neighbourhood representations may expand the attainable performance boundary, but only as diagnostic evidence. SBAS-InSAR-derived information should therefore not be treated as a simple additional conditioning factor; its value depends on jointly interpreting deformation intensity, observation availability and their confounding with static context. Full article

The empirical descriptors of high-entropy alloys and oxides, from the mixing entropy and atomic-size mismatch to the Miedema enthalpies and the $\mathsf{\Omega }$, $\mathrm{\Phi }$, and $\phi$ stability parameters, are quoted in nearly every design study, yet they are reimplemented ad hoc by individual groups, by closed web calculators, and now inside language-model agent frameworks, where fabrication of property values is a documented failure mode. The resulting numbers disagree and cannot be traced or reproduced. We present HEA-Bench, an open calculator in which every descriptor is a closed-form expression over a curated, literature-cited element-property table, with the six canonical phase-prediction rules reported alongside their thresholds and sources rather than as predictions. One calculation core is delivered as a dependency-free Python (version 3.10 or later) library, a zero-install browser application, an offline desktop executable, and a Model Context Protocol server that exposes it to AI agents as deterministic tools, returning every value with its unit, citation key, and version so an agent’s reasoning trace can be audited. The implementation reproduces published per-alloy and per-oxide anchor values to their printed precision and extends to high-entropy oxides in four structure families. The numerical instability of $\mathsf{\Omega }$ near zero mixing enthalpy is quantified and exposed as a callable check. Full article

Natural fiber-reinforced polymer composites are attractive for lightweight and sustainable engineering applications; however, property scatter remains a major barrier to reliable design. Mean tensile properties alone are insufficient when material selection depends on repeatability and lower-tail performance. This study presents a statistical variability and lower-tail reliability assessment of flax, jute, and carbon fiber composite laminates using 590 open-access tensile test records from a published natural-fiber composite dataset. Flax and jute were selected as representative bast-fiber systems covering a range of woven, unidirectional, and short-fiber architectures; carbon fiber was included as a synthetic-fiber reference system. Three mechanically important properties were analyzed: the recalculated tensile modulus, tensile strength, and axial failure strain. Normal, lognormal, and two-parameter Weibull distributions were screened for each material–property combination using the Akaike information criterion (AIC); empirical fifth percentiles (P5) and bootstrap 95% confidence intervals (CI) were computed as lower-tail descriptors. The results show that Carbon-0 has the highest lower-tail modulus and strength, with empirical fifth percentiles of 104.95 GPa and 989.64 MPa, respectively. Among the natural fiber systems, Flax-0 and Flax-VE-0 provided the highest lower-tail strengths, whereas Flax-Twill and Flax-CP showed the highest lower-tail failure strains. The lowest tensile strength coefficient of variation was observed for Flax-90 (2.41%), followed by Flax-Twill (3.43%), Flax-0 (4.50%), Jute-Satin (4.83%), and Jute-Plain (4.92%). A balanced reliability ranking that combined lower-tail property ranks and coefficient of variation ranks identified Flax-0, Flax-VE-0, Flax-Twill, Flax-CP, and Jute-Satin as the most favorable natural-fiber systems. The lower coefficient of variation values observed in aligned and satin-weave architectures relative to short-fiber and plain-weave systems reflect the role of fiber orientation uniformity in moderating property scatter at the laminate scale. This study provides a reproducible statistical framework based on lower-tail performance descriptors for comparative screening purposes, not on formal design allowables for distinguishing high mean performance from reliable minimum-level performance in natural fiber composite laminates. Full article

Compact spatial heterodyne spectroscopy (SHS) is useful for atmospheric remote sensing, but repeated interferogram acquisition alone does not guarantee stable recovered spectral intensity. An intensity-preserving framework based on self-calibrated quality-gated robust fusion (SC-QGRF) was evaluated for multi-frame SHS spectral recovery. Frame weights are derived from image and spectral quality descriptors while avoiding framewise spectral normalization. Test spectral measurements were carried out using potassium and xenon lamps, which served as controlled laboratory cases for line spectrum and broadband spectrum recovery, respectively. Random Single, Image Average, Spectrum Mean, Spectrum Median, and the proposed strategy were compared under a common recovery and evaluation setting. Complete frame and exposure response analyses retained the main spectral profiles and showed no evident exposure-dependent intensity bias in this laboratory setting. Repeated subsampling provided the main stability evidence. SC-QGRF achieved the lowest variability in all four core metrics: coefficient of variation (CV) of K-AreaRatio = 0.0445%, CV of the full width at half maximum of the second K peak (K-FWHM2) = 0.0395%, CV of Xe-BandIntegral = 1.1770%, and standard deviation (STD) of Xe-SpectralCentroid = 0.0786 nm. The method can be used as a spectral quality control and fusion step for repeated SHS observations toward atmospheric remote sensing, while gas cell and field measurements remain necessary for retrieval-level validation. Full article

The NIOSH Lifting Equation is widely used to evaluate manual material handling tasks by identifying lifting risk through load reduction multipliers based on task geometry. However, while it provides a means of classifying risk, it does not estimate spinal forces and therefore cannot quantify the biomechanical load experienced at the lumbar spine. In contrast, biomechanical simulations can estimate spinal compression with high fidelity, but they require motion capture systems and specialized software that are not practical for most workplace assessments. This study aimed to bridge these approaches by developing a family of mixed-effect statistical models that predict L4/L5 spinal compression forces using the geometric parameters of the NIOSH framework combined with posture-specific biomechanical descriptors at peak-loading poses extracted from digital simulations. Data were aggregated from multiple experimental lifting studies in which standardized NIOSH parameters and corresponding spinal compression forces were obtained through validated digital human modeling. Mixed-effects regression was used to establish the relationship between task geometry, joint posture, load weight, and spinal compression. The resulting predictive equation demonstrated strong agreement with simulation-derived forces and effectively captured the contributions of subject, task and body positioning to the spinal compression force across diverse lifting tasks. Importantly, the variance structure of the model’s coefficients allows the contribution of risk to be attributed either to task-related factors or to subject-specific movement behaviors, reinforcing the safety relevance of the framework. Horizontal reach, vertical hand height, and load magnitude emerged as dominant predictors, with trunk posture providing additional explanatory power. The model offers ergonomists a practical, biomechanics-informed tool that extends the descriptive capacity of the NIOSH framework by enabling direct estimation of spinal compression forces without the need for full biomechanical simulations. Full article

Image feature matching is a cornerstone of computer vision, yet robust correspondence estimation under rotation, low texture, and illumination variation remains challenging. Detector-free methods such as LoFTR reduce the dependence on repeatable keypoints, but their standard convolutional backbones are still sensitive to orientation changes and weak local structures. To address these limitations, we propose RA-LoFTR, which enhances LoFTR with Rotational Coordinate Convolution (RCC) and Adaptive Fusion Weighting (AFW). RCC partitions feature neighborhoods into concentric annular regions and aligns dominant orientations through channel-wise cyclic shifts, transforming rotation handling from discrete angle classification into channel-phase alignment. AFW dynamically fuses RCC-derived local structural cues with Transformer-based global positional information, and MAGSAC++ is further used as a geometric verification step for outlier rejection. On MegaDepth, RA-LoFTR achieves pose-estimation AUCs of 37.4, 53.6, and 66.0 at 5°, 10°, and 20°, improving over LoFTR by 1.1, 1.7, and 1.3 points, respectively. On HPatches, it improves homography AUCs over LoFTR by 2.2, 1.5, and 1.9 points at 3, 5, and 10 pixels, respectively. Robustness experiments under low-light, motion blur, Gaussian noise, and low-texture conditions further validate the proposed annular convolution design. Full article

Fly ash (FA) recycling into cementitious binders offers a promising route to simultaneously reduce the environmental burden of high-impact construction materials, divert industrial waste from disposal pathways, and generate engineering value from an existing aluminosilicate residue. FA has therefore been widely incorporated into ordinary Portland cement (OPC), calcium sulfoaluminate (CSA) cement, magnesium potassium phosphate cement (MKPC), and alkali-activated/geopolymer systems, where it can improve workability, mechanical strength, durability and promote hydration reactions, depending on the binder chemistry. However, these benefits are not always successful: FA addition may also reduce early-age strength, delay setting, limit reactivity, impair fluidity, or produce under-reacted matrices when ash properties are mismatched with the chemistry window of the target binder. Rather than revisiting these systems as isolated application categories, this review develops a property-guided framework for interpreting and selecting FA across major cementitious routes. It first highlights why FA should not be treated as a single material. Key descriptors include glass content, fineness, calcium level, carbon residue, mineralogy, and beneficiation state. The review then compares four representative binder environments to clarify the role of FA shifts from pozzolanic contributor to filler-dominated hydration modifier to functional regulator, and finally to reactive precursor, including the calcium hydroxide (CH)-rich Portland systems, CH-poor CSA systems, phosphate-bonded MKPC systems, and alkali-activated/geopolymer systems. The central conclusion is that the key question is not simply where FA can be used, but which FA is best matched with which binder chemistry and for what performance objective. Full article

Background/Objectives: Postural control in elite athletes may reflect sport- and side-related balance demands. Conventional Centre of Pressure (CoP) path length alone offers only limited information about directional and frequency-domain sway characteristics. This cross-sectional study described CoP path length, mediolateral/anteroposterior (ML/AP) directionality, and power spectral density (PSD)-derived frequency-domain descriptors in elite athletes from sports with distinct movement demands. Methods: A total of 116 asymptomatic elite athletes from volleyball, football, short track, ice hockey, and field hockey were assessed during single-leg stance. CoP path length, the ML/AP index, and PSD outcomes were analysed. PSD was calculated in LabVIEW using a fast Fourier transform (FFT) routine from the complete 60 s trial acquired at 1 kHz after removal of the DC component. Spectra were not normalised and are reported as absolute spectral-density values in mm²/Hz. PSD outcomes were summarised in low-frequency (0.02–0.6 Hz) and higher-frequency (1–5 Hz) windows, and the PSD quotient was defined as PSD 0.02–0.6 Hz/PSD 1–5 Hz. Results: Observed sport–sex groups differed in subject-averaged CoP path length (F(5,110) = 22.26, p < 0.001, eta_p² = 0.503), ML/AP index (F(5,110) = 4.07, p = 0.002, eta_p² = 0.156), PSD 0.02–0.6 Hz (F(5,110) = 38.67, p < 0.001, eta_p² = 0.637), PSD 1–5 Hz (F(5,110) = 4.83, p < 0.001, eta_p² = 0.180), and the exploratory PSD quotient (F(5,110) = 3.33, p = 0.008, eta_p² = 0.132). Paired-side comparisons showed greater right-side CoP path length, greater right-side PSD 0.02–0.6 Hz, and a higher right-side ML/AP index, whereas PSD 1–5 Hz and the PSD quotient did not differ significantly between sides. Conclusions: The combined analysis of CoP path length, ML/AP directionality, and PSD-derived descriptors characterised sport-, sex-, and side-specific postural-control profiles in this cohort. Mechanistic interpretations of segmental neuromuscular control remain tentative because the study was cross-sectional and did not include electromyography, kinematics, or prospective injury data. Full article

Computer vision technology has emerged as a promising approach for multipoint displacement monitoring of civil infrastructure, owing to its inherent noncontact operation and remote measurement capabilities. However, its measurement accuracy is greatly affected by ambient temperature variations during long-term monitoring applications. Therefore, this study investigates temperature-induced displacement measurement errors in vision-based measurement techniques and proposes an error compensation method based on a time series forecasting model. First, a deep learning-based feature descriptor is employed and enhanced for image feature extraction and displacement extraction, followed by a quantitative analysis of the impact of temperature variations on vision-based displacement measurement errors. Second, the ModernTCN deep learning model is used to establish the nonlinear relationship between temperature differences and temperature-induced measurement errors. Finally, the established model is employed to predict and compensate for the temperature-induced measurement errors. The effectiveness of the proposed method is validated via indoor laboratory experiments, outdoor temperature tests, and field tests on a long-span bridge. The results reveal a nonlinear relationship between temperature differences and vision-based displacement measurement errors, and this relationship can be effectively mitigated using the proposed time series forecasting model. Experimental results indicate that correlation coefficients and cosine similarities obtained by the proposed method outperform conventional Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and state-of-the-art architectures (i.e., Transformer, Informer, PatchTST, TimesNet, and N-BEATS). In field tests, the proposed method achieved a Mean Absolute Error (MAE) of approximately 0.002 mm, representing an error reduction of over 90% compared to the Transformer model. The proposed method enables robust multipoint displacement monitoring with integrated temperature self-compensation, providing a reliable data foundation for safety evaluation and health monitoring of engineering structures. Full article

Upstream–downstream nutrient connectivity strongly regulates water-quality risk in reservoir cascades, yet its quantification remains difficult where discharge, reservoir release, and hydraulic residence-time records are unavailable. This study developed a causal–explainable framework to diagnose total nitrogen (TN) connectivity between the upstream Shimantan Reservoir and downstream Banqiao Reservoir in the Huai River Basin, China. Long-term water-quality records, meteorological variables, land-cover indicators, seasonal descriptors, and lagged upstream predictors were integrated within a leakage-safe analytical workflow combining nonlinear causal inference, time–frequency coupling diagnostics, predictive modeling, SHAP-based attribution, and counterfactual analysis. Convergent Cross Mapping identified statistically significant asymmetric bidirectional nonlinear coupling, with a CCM skill of ρ = 0.746 for Shimantan TN → Banqiao TN and p = 0.730 for the reverse direction (p = 0.002 for both directions). Wavelet coherence showed that upstream–downstream TN coupling was concentrated mainly at short temporal scales, with a cone-of-influence-restricted mean squared coherence of 0.7114 in the 2–6-month intra-seasonal band. The validation-selected Gradient Boosting model provided interpretable test-period predictive skill, on independent source-derived observations from 2021–2023, achieving R² = 0.567, RMSE = 0.391 mg/L, and MAE = 0.312 mg/L. The GAN-generated 2024–2025 segment was excluded from empirical model evaluation and retained only for exploratory future-scenario assessment. SHAP decomposition indicated that upstream-related predictors accounted for 77.04% of the total absolute model attribution during high-TN events, while seasonal conditioned counterfactual upstream neutralization produced mean model-predicted changes of 0.162 mg/L across all test observations and 0.807 mg/L during high-TN events. Together, these results demonstrate that hidden upstream-downstream TN connectivity can be diagnosed through convergent causal, temporal, predictive, and model-attribution evidence in data-limited reservoir cascades. The findings support asymmetric coupled dynamics and downstream inheritance of upstream information but should not be interpreted as proof of exclusively one-way physical nutrient transport. The proposed framework offers a diagnostic decision-support approach for reservoir systems with sufficiently long monitoring records where direct hydraulic observations are unavailable. Full article

Liquid chromatography–tandem mass spectrometry (LC-MS/MS) is widely employed in pesticide residue analysis. Machine learning methods for automated spectral interpretation depend on large, well-curated training datasets; however, publicly available pesticide mass spectrometry data are fragmented across heterogeneous repositories, lack standardized preprocessing, and suffer from incomplete metadata. We introduce PAID (Pesticide AI-ready Dataset), comprising two curated LC-MS/MS spectral collections derived from 15 public sources (GNPS, MassIVE, MoNA, MassBank). Starting from 91,420 raw spectra, after initial pesticide-directed screening, a seven-step reproducible pipeline—spanning multi-source integration, spectral cleaning, deduplication, metadata standardization, quality scoring, stratified splitting, and feature engineering—yields PAID-Strict (7527 spectra; 3197 compounds) and PAID-Extended (21,292 spectra; 3224 compounds). Both versions cover eight pesticide categories across QTOF and Orbitrap platforms, with core metadata fields (SMILES, InChIKey, molecular formula) exceeding 98% completeness. A feature suite of 32 chemoinformatic descriptors, 2214 molecular fingerprints, and 30 spectral features is provided alongside the spectra. Benchmark classification of the eight pesticide categories using XGBoost and LightGBM achieved 81.5% accuracy. The dataset, code, and pre-computed features are publicly available under CC BY 4.0 and MIT licenses. Full article