Search Results (107)

In recent years, numerous studies have reported on intelligent analytical applications for food quality and safety. To identify the underlying patterns and development trends in this field, this paper presents a comprehensive review from the perspectives of both data-driven and mechanism-driven paradigms. Under the data-driven paradigm, detection modalities may involve either single-modal or multimodal approaches. By integrating measured detection data with appropriate intelligent learning algorithms, specific tasks for food quality or safety assessment can be achieved. Research efforts in this area encompass the development of detection techniques, optimization of measurement parameters, construction of high-dimensional spectral features, design of feature extraction methods, selection and tuning of algorithms, and formulation of multimodal data fusion strategies. This paradigm is characterized by high computational speed and superior prediction or classification efficiency. Nevertheless, it is constrained by several limitations, including poor model interpretability, limited extrapolation and generalization capabilities, and a heavy reliance on high-quality annotated data. In contrast, the data–mechanism hybrid-driven paradigm integrates physical laws and other prior knowledge as constraints that are deeply embedded into neural network training. By combining data-driven mining capabilities with theoretical prior knowledge, this approach achieves improved predictive performance and decision-making reliability. This paradigm offers notable advantages, such as enhanced interpretability, greater trustworthiness, improved data efficiency, and reduced computational costs. It is particularly well-suited for small-sample or data-sparse scenarios, and thus represents a promising and important direction for future research in this domain. Full article

Atmospheric microplastic pollution is an emerging environmental concern, yet data on ecological risks posed by snow cover pollution remain sparse. This study presents the first assessment of ecological risks associated with snow cover pollution in Western Siberia. Pollution and risk levels were quantified using the pollution load index (PLI), the polymer hazard index (PHI), and the potential ecological risk index (PERI). Microplastics were detected in all snow samples; however, the pollution level was rated as ‘low’ across all sampling sites throughout the observation period. Although median PHI values for the snow cover indicated medium hazard, the presence of samples referred to the hazard categories ‘danger’ and ‘extreme danger’ (up to 27% in 2023) pointed to localized pollution by potentially toxic polymers. In contrast, the average PERI values remained low (1.92–19.7), as this index accounts for both polymer toxicity and microplastic abundance. However, the index showed a consistent upward trend over the three-year observation period, driven by the increase in both the proportion of hazardous polymers and the total microplastic load. Thus, although the quantitative content of microplastics in the snow cover of Western Siberia was low, the high PHI values at some sampling sites and the occurrence of polymers containing toxic monomers like polyacrylonitrile (PAN) and polyurethane (PU) indicate a potential ecological risk that requires further monitoring. Full article

Accurate state-of-charge (SOC) estimation is essential for range prediction, power allocation, and operational safety in electric vehicles. Low-temperature operation introduces coupled effects of capacity degradation and polarization dynamics, which challenge conventional fixed-capacity models and single-timescale estimation methods. This paper proposes a hierarchical multi-timescale SOC estimation framework that combines data-driven capacity prediction, online parameter identification, and nonlinear state estimation. At the upper layer, a temperature-aware temporal self-attention long short-term memory network (TS-LSTM) estimates the available battery capacity. At the lower layer, forgetting-factor recursive least squares (FFRLS) and a gain-scheduled unscented Kalman filter (UKF) jointly track impedance variations and recursively estimate SOC. The framework is validated on public lithium-ion battery datasets at 4, 24, and 43 °C under pulse and sparse high-current conditions. Under the challenging 4 °C scenario with a 10% initial SOC bias, the method achieves an SOC root mean square error (RMSE) of 1.35%. Additional perturbation tests yield SOC RMSEs of 1.35–2.40% under −5% to +10% initial-SOC offsets, 1 mV voltage noise, a +2 °C temperature bias, and a +20% impedance-prior error. Furthermore, the computational overhead remains substantially lower than the data sampling interval, indicating promising potential for online estimation in battery management systems. Full article

Phased arrays are key for next-generation mobile communication, satellite communication, and radar systems. Radiation patterns are essential performance metrics for evaluating antennas, but conventional mechanically scanned measurements become inefficient when repeated for every beam of the phased array. Existing active-element-pattern-based methods reduce this burden by grouping elements with similar coupling environments, yet their grouping and representative-element selection are mainly based on physical intuition, and on–off-mode-based measurements may differ from the actual all-on operating state. This paper proposes a data-driven adaptive grouping method based on K-means unsupervised learning. Element positions and broadside phase information construct the feature matrix, and a composite evaluation function considering group size and electric-field variation subdivides high-dynamic regions. Representative element patterns are acquired in the all-on mode, with sparse angular sampling to further reduce measurement time. Simulation results of an $8\times 8$ dipole array show that the proposed method reconstructs multi-beam array patterns using 16 representative elements, fewer than the 25 representative elements required by a $5\times 5$ grouping strategy, with the same accuracy in the main lobe and improved accuracy in sidelobe nulls. The effectiveness of the proposed grouping method has also been demonstrated for the measured $4\times 4$ mmWave phased array. Full article

Micromachined ultrasonic transducers (MUTs) represent a notable advance in miniaturized sensing, enabling compact, low-power, and complementary metal-oxide-semiconductor (CMOS)-integrated platforms that extend ultrasonic capabilities into wearable, implantable, and edge-computing domains. The integration of artificial intelligence (AI) has introduced new approaches for signal interpretation, adaptive control, and data-driven optimization, enhancing performance in specific areas such as compressed sensing, neural beamforming, and learned image enhancement that complement conventional signal processing. Meanwhile, sensor fusion strategies that combine ultrasonic data with complementary modalities have improved robustness, contextual awareness, and diagnostic accuracy across applications ranging from industrial monitoring to clinical diagnostics. This review provides a comprehensive analysis of this active research area, systematically covering transducer hardware platforms, design methodologies, and intelligent signal processing frameworks. While traditional bulk piezoelectric transducers remain the benchmark for high-power applications, capacitive and piezoelectric micromachined variants offer superior acoustic impedance matching and monolithic CMOS compatibility essential for portable systems. We examine the evolution from deterministic analytical and numerical modeling toward AI-powered inverse design, which enables the discovery of non-intuitive, high-performance geometries beyond human intuition. Furthermore, the integration of machine learning (ML) for signal recovery, image enhancement, and multi-modal sensor fusion is discussed as a pathway to compensate for hardware constraints such as limited aperture, sparse sampling, and low signal-to-noise ratio (SNR), while pointing out that AI technology cannot overcome fundamental physical limits including acoustic attenuation, thermal noise floors, and transduction efficiency boundaries. By synthesizing recent advancements, this review demonstrates how the convergence of classical acoustic physics and data-driven intelligence is guiding the development of of intelligent ultrasonic systems. Full article

The quality and safety supervision of agricultural products at the grassroots level has long faced the triple superposition dilemma of small-sample sampling, multi-source evidence conflict, and risk chain evolution. Although existing data-driven models have considerable accuracy, they are difficult to leverage for intervention decisions, and the simple serial connection of traditional Bayesian networks and evidence theory cannot respond to dynamic scenarios. Aiming at this research gap, this paper constructs a dynamic risk assessment model, CIBE-DR, that deeply couples Bayesian networks with evidential reasoning. It contains three core innovations. First, the structure learning method of the causally identifiable Bayesian network embeds a graded do-calculus identifiability score covering both back-door and front-door criteria into the BDeu scoring function and combines this reward with an expert-prior divergence penalty that breaks Markov equivalence so as to realize the transition from relevance modeling to intervention decision modeling. Second, the conflict-aware adaptive evidence synthesis rule orthogonally decomposes multi-source conflict into an epistemic component and an ontological component, which are modeled respectively by Tsallis belief entropy and abductive inference over a discrete twenty-seven-point heterogeneity hypothesis space and are then fused under a reparameterized Dempster–Yager interpolation in which the two endpoints recover the two named rules under a single consistent interpretation. Third, the bidirectional closed-loop coupling mechanism between BN and ER realizes the mutual calibration between the conditional probability table and the evidence credibility prior under a Lyapunov monotone descent argument with the explicit Lipschitz bound $L_{\theta }$ ≤ 0.028 < 1, endowing the model with time-varying self-correction ability. Based on experiments on 156,847 sampling samples from counties and townships in East China, Central China, and Southwest China from 2021 to 2024, the proposed method achieved the best value in six of the seven evaluation indicators, with a minority recall of 0.864 ± 0.014, an intervention effect estimation error of 0.063 ± 0.005, and a dynamic response delay of 2.8 ± 0.3 days, significantly ahead of eleven mainstream baselines under the McNemar test on classification ($p$ < 0.001) and the Wilcoxon signed-rank test on intervention-effect estimation ($p$ < 0.001). The only indicator on which CIBE-DR does not lead is overall accuracy, which is 0.002 lower than that of Transformer; this difference does not reach statistical significance under the McNemar test ($p$ = 0.32) and does not weaken the value of grassroots supervision in the strong-imbalance scenario where the positive rate is only 1.04%. The robustness advantage of the model is particularly prominent in the scenarios of sparse data, adversarial perturbation, and prior-graph incompleteness, and the intervention-effect estimates were additionally validated against two post-2022 policy interventions with absolute deviations of 1.4 and 1.2 percentage points respectively. These results verify the product gain and grassroots deployability of the three mechanisms. Full article

The prediction of shear strength in seasonally frozen silty clay is restricted by complex physical mechanisms and sparse experimental data. A self-supervised contrastive physics-informed neural network is proposed to overcome these limitations. Robust latent features are extracted from limited datasets via contrastive pretraining. Time-dependent constitutive equations and physical boundary conditions are simultaneously embedded into the loss function. This mathematical constraint ensures strict physical consistency during the modeling process. The proposed framework was validated using 100 independent laboratory samples prepared under controlled moisture content, freezing temperature, and thawing duration. The experimental results demonstrate the superior predictive accuracy of the proposed model. A coefficient of determination ($R^2$) of 0.988 was achieved on the test set, accompanied by minimized error metrics compared to conventional data-driven approaches. Consequently, a highly accurate and reliable methodology is established by this architecture for evaluating soil stability and supporting infrastructure design in cold regions. Full article

In recent years, data-driven mineral prospectivity mapping (MPM) has become a key technical approach for delineating exploration targets. Nevertheless, two major challenges persist. First, sampling is often sparse and irregularly distributed, making it difficult to represent spatial relationships effectively. Second, many existing models implicitly assign equal importance to all geochemical elements. They also fail to jointly model heterogeneous elemental contributions and spatial neighborhood relationships. To address these issues, we propose the Spatial–Element Transformer (SETransformer), which integrates SHapley Additive exPlanations (SHAP)-derived element-importance attention bias with adaptive-bandwidth spatial-decay attention for data-driven MPM. We evaluate the proposed method using data from the Jiama porphyry–skarn Cu-polymetallic deposit in Tibet, China, within two a priori defined metallogenic feature subspaces: Mo-type and Cu-type. SETransformer is benchmarked against conventional machine learning models (Random Forest (RF), eXtreme Gradient Boosting (XGBoost), support vector machine (SVM)), and deep learning baselines (graph convolutional network (GCN) and TabNet). Model performance is assessed under two spatial validation schemes: a south–north hold-out validation and a five-fold spatial block cross-validation. Under both schemes, SETransformer achieves the best overall performance in the two subspaces. The predicted MPMs are consistent with regional geological evidence and exhibit clearer boundaries, providing a useful reference for subsequent target delineation in the study area. Full article

The Upper Blue Nile Basin contributes about 60% of the Nile River’s annual streamflow but faces severe sediment-related challenges driven by intense monsoonal erosion and reservoir siltation. Accurate estimation of suspended sediment concentration (SSC) and sediment load in large, data-scarce watersheds remains difficult due to sparse monitoring and complex supply-limited transport dynamics. This study develops a hybrid machine learning (ML) and process-based approach for the Kessie watershed (65,784 km²), a major sediment source upstream of the GERD. The approach combines Random Forest (RF) based SSC reconstruction from 251 intermittent samples spanning 1995–2011, approximately 70% collected during the wet season (June–October) and 94% concentrated in 2008–2011, covering a wide range of observed streamflow conditions at the time of sampling (120–5897 m³/s), with a two-stage calibration of the WASA-SED model. Using hydrologically informed predictors, the RF algorithm substantially outperformed the bias-corrected traditional sediment rating curve and other ML algorithms, increasing the validation coefficient of determination (R²) from 0.274 to 0.693. The reconstructed daily SSC yielded a mean annual sediment load of 180.7 Mt/yr. The model performed well, particularly at monthly scales for 1995–2011, achieving good to very good performance (NSE up to 0.83/0.71 for streamflow and 0.86/0.63 for sediment load, calibration/validation, respectively) and reproducing dominant hydrological and sediment regimes using duration curves. Mann–Kendall trend analysis (α = 0.05) indicated no statistically significant monotonic trends in annual rainfall (p = 0.90), mean annual streamflow (p = 0.24 observed; p = 0.84 simulated), or mean annual sediment load (p = 0.66 simulated); the reconstructed sediment load series showed a near-significant increasing tendency (p = 0.06) that falls below the accepted significance threshold and should be interpreted with caution given the short 17-year record. This hybrid approach effectively captures monsoon-driven sediment fluxes and provides model-based daily-to-monthly sediment load estimates with quantified uncertainty. It supports improved reservoir sedimentation assessment, erosion-risk evaluation, and transboundary water-resources planning in data-scarce tropical highlands. Full article

Semantic segmentation of 3D point clouds remains difficult when LiDAR or depth-camera data are sampled unevenly. This paper presents DANet, a 3D semantic segmentation framework built on joint density- and semantics-adaptive convolution. Its core operator, Density-Adaptive Radius Convolution (DAR-Conv), predicts point-wise neighborhood radii before feature aggregation by combining density-driven initialization with semantics-aware modulation. In this way, dense regions can use compact receptive fields, whereas sparse or semantically complex regions can draw on broader contextual support. DANet also includes a Gated Adaptive Cross-Layer Fusion (GACF) module, which aligns encoder–decoder features and performs gated fusion with residual refinement. Experiments on S3DIS and NPM3D show that DANet obtains the highest reported mean accuracy (mAcc) among the compared methods on S3DIS, and high mean Intersection over Union (mIoU) and overall accuracy (OA) on NPM3D, supporting the usefulness of density- and semantics-aware receptive-field adaptation. Full article

Background: Artificial intelligence (AI)-driven gait analysis using plantar pressure and ground reaction force (GRF) signals may provide objective digital biomarkers for rehabilitation, but clinical translation remains uncertain. This scoping review and evidence map aimed to summarize clinical applications, compare evidence maturity, and identify methodological and translational gaps. Methods: PubMed, Web of Science, Embase, and Scopus were searched from the earliest available indexed records in each database to May 2026. Original clinical studies using plantar pressure- or GRF-derived signals with AI methods for disease recognition, severity assessment, risk prediction, rehabilitation monitoring, or decision support were included. Results: Fifteen studies met the eligibility criteria. Evidence was concentrated in Parkinson’s disease (PD), particularly PD recognition and freezing of gait prediction, where relatively more mature evidence was supported by multiple studies and participant-level or cross-dataset validation. Evidence for PD severity assessment, knee osteoarthritis monitoring, chronic ankle instability rehabilitation, fall-risk stratification, sarcopenia screening, peripheral artery disease recognition, and functional gait disorder classification remained less mature. Translation was limited by small or single-center samples, unclear participant-level data splitting, limited external validation, absent calibration, sparse explainable AI reporting, and insufficient real-world workflow testing. Conclusions: Future studies should prioritize prospective, externally validated, interpretable, calibrated, and clinically embedded models before routine rehabilitation implementation. Full article

High-dimensional change point detection is a fundamental problem in modern statistical learning, particularly when distributional changes are driven by a small and unknown subset of variables. In heterogeneous settings, uniform aggregation across coordinates may suffer from signal dilution, because stable or noisy variables can mask the evidence carried by structurally unstable coordinates. Moreover, many existing procedures primarily focus on temporal localization and provide limited information about the variables responsible for a detected structural break. To address these challenges, we propose WAVE, a weighted adaptive variable selection procedure for interpretable change point detection. WAVE constructs variance-standardized global CUSUM evidence and locally standardized exponentially weighted evidence for each coordinate and then adaptively maps intervalwise coordinate evidence into a continuous weight vector. The learned weights strengthen coordinates with persistent or local evidence of change while downweighting nuisance coordinates with weak evidence. The resulting weighted scan statistic is calibrated by a residual moving block bootstrap that preserves temporal and cross-sectional dependence and re-applies the weighting rule within bootstrap samples to account for data-adaptive aggregation. Detected change points are further equipped with coordinate-level attribution through a multi-criteria fusion rule combining adaptive weights, local standardized effect sizes, and marginal testing evidence. Simulation studies show that WAVE achieves accurate localization and reliable support recovery in both single and multiple change point settings, particularly under sparse and heterogeneous alternatives. An empirical analysis of S&P 100 stock returns in 2020 further demonstrates that WAVE identifies economically meaningful market regime shifts with interpretable coordinate-level attribution. Full article

Under 10 min sparse Automatic Identification System (AIS) sampling, the reliability of point-wise motion statistics degrades substantially, and conventional classification methods rely on trajectory interpolation, which may introduce spurious motion patterns. This study proposes a feature-driven framework for fishing vessel type identification that eliminates the need for interpolation preprocessing. A 39-dimensional feature set is constructed using robust statistics, including the median and interquartile range, to characterize trajectory-level behavioral patterns. Adaptive speed interval thresholds are derived through a data-driven approach grounded in Bayesian decision boundaries, thereby removing the dependence on manually defined cut-off values. A backward ablation procedure guided by feature importance ranking identifies a lightweight 12-dimensional feature subset that retains 98.7% of the classification accuracy at a compression rate of 69%. Evaluated on 18,320 fishing vessel trajectories in the East China Sea, the full 39-dimensional feature set achieves a 5-fold cross-validation accuracy of 91.92% (Macro-F1 = 0.919, Kappa = 0.879), with inter-fold standard deviations ranging from 0.002 to 0.004. Comparative experiments demonstrate that three tree-based classifiers all exceed 90% accuracy on the same feature set, confirming that feature robustness, rather than model selection, constitutes the dominant performance factor. LightGBM achieves the optimal trade-off between accuracy and training efficiency, whereas the cross-validation standard deviation of LSTM is approximately 7.5 times greater, indicating that hand-crafted robust features provide superior stability under sparse sampling conditions. The proposed framework requires no fishery-specific prior knowledge and offers a transferable paradigm for sparse AIS trajectory analysis. Full article

Vibration sensor networks deployed on bridges continuously generate large volumes of unlabelled measurements under healthy operation, whereas labelled damage records on any specific target bridge remain extremely scarce—a chronic data asymmetry that constrains data-driven structural health monitoring (SHM). Existing remedies either require labelled source-bridge data or borrow augmentation pipelines and encoders from computer vision that are poorly matched to one-dimensional vibration signals. This study proposes a two-stage framework—siamese contrastive pretraining followed by few-shot fine-tuning on the target bridge—that learns environment-invariant representations from unlabelled source-side sensor signals and transfers them to a new bridge using only a handful of labelled samples. Three contributions are advanced: (i) a signal-domain augmentation policy that decouples sensor-level corruptions from operational-level fluctuations, including a frequency-band stochastic masking scheme designed to emulate cross-bridge perturbations; (ii) a one-dimensional self-calibrated convolutional encoder embedded in a stop-gradient siamese learner, providing the enlarged receptive field and inter-channel coupling required to capture sparse damage signatures in multi-sensor recordings; and (iii) a transferability analysis that formally links the contrastive invariance objective to a bound on the expected cross-bridge risk. On the Z24 benchmark and an in-house four-configuration laboratory bridge population, the method attains a 5-shot macro-F1 of 0.913 (Z24 → Lab) and 0.892 (Lab → Z24), outperforming eleven baselines by 3.4–37.1 percentage points. Full article

The southeastern Tibetan Plateau (SETP) lies at the intersection of extreme topography and the Indian summer monsoon, producing intense orographically driven precipitation. Resolving these extremes requires convection-permitting simulations (CPMs) due to the combined influence of complex terrain and vigorous convection. However, kilometer-scale simulations over SETP remain limited to short periods because of computational cost, preventing robust estimation of precipitation extremes. We analyze a decade of 1 km CPMs and apply the Simplified Metastatistical Extreme Value (SMEV) framework, which uses all wet-day precipitation rather than annual maxima, increasing the effective sample size by an order of magnitude. We apply the SMEV framework to estimate daily precipitation return levels up to 100 years. While SMEV increases the effective sample size, uncertainty remains non-negligible for long return periods when derived from a decadal record. Results show that CPMs’ estimates align with observations within 90% bootstrap confidence intervals (CIs). For instance, at a representative station (Obs: 51.6 mm/d), the 50-year return level is estimated at 57.1 mm/d (CI: 47.1–68.9 mm/d). In contrast, coarse-resolution products systematically overestimate these extremes by 50–100%, with their estimates often falling far outside the observed range beyond 20-year return periods. CPMs also reveal a model-derived, non-monotonic elevation dependence absent in coarse datasets. Instead of monotonic decline, three phrases emerge: a weak increase below 2700 m, a sharp decrease across mid-elevations, and a reversal above ~5300 m where orographic uplift enhances extremes, yielding a 2.3-fold increase. These results show that CPMs alter not only magnitude but also the spatial structure and elevation scaling of precipitation extremes, providing a physically constrained framework for extreme-value estimation in data-sparse mountains. Full article