Search Results (2,288)

Open-world classification must account for samples whose classes are absent from the current frame of discernment. Although generalized evidence theory can represent such possibilities, the conventional base belief function assigns reference support only to nonempty propositions and therefore cannot directly support source-level treatment of unknown information. This work proposes a generalized base belief function that incorporates the open-world proposition into its reference domain and uses it to modify generalized basic probability assignments before fusion. The modified evidence is combined through rules compatible with generalized evidence theory and followed by a known/unknown decision, forming a complete classification procedure for incomplete frames. Experiments on Iris and Seeds show that source-level modification and generalized fusion jointly support known/unknown classification. On Waveform, classification remains stable as controlled perturbation increases in a larger, higher-dimensional, noisy environment. On the Wall-Following Robot dataset, the procedure gives a selective known/unknown decision across different class compositions, preserving known samples while retaining overall discrimination close to GCR. These results support the proposed GBBF-based method for open-world classification under incomplete frames. Full article

This study seeks to overcome the limitations of traditional gas well production peak forecasting approaches, particularly in the context of handling multi-dimensional features, complex dynamic behaviors, and non-linear relationships. Accordingly, a deep learning-based multi-scale adaptive fusion perception network is proposed in this work. Specifically, a time-series processing framework based on multi-scale decomposition is first constructed, where wavelet transform is employed to perform multi-scale decomposition of the gas well production data, thereby effectively extracting both long-term trends and short-term fluctuations. Subsequently, a multi-dimensional feature collaborative encoding and adaptive fusion module is designed. This module jointly encodes the multi-dimensional features in gas well production forecasting. By utilizing a dual-path PatchTST encoder, it captures the long-term trends and short-term fluctuations within the time-series data. Finally, to enhance the accuracy of gas well production peak forecasting, a personalized multi-loss training function is developed. This mixed loss function balances the robustness to outliers with the control of percentage errors, effectively mitigating the sensitivity of conventional loss functions to outliers. Systematic experiments conducted on real gas well production samples demonstrate that the proposed model outperforms existing state-of-the-art models across various evaluation metrics. Furthermore, quantile accuracy analysis reveals that the proposed model consistently achieves the highest proportion of qualifying samples under error thresholds ranging from 5% to 20% while exhibiting stable adaptability across different production scenarios, including low, typical, and high production rates. The proposed method provides a high-precision, effective solution for gas well production peak forecasting, offering reliable decision support for production process optimization. Full article

Local short-duration heavy-rainfall warning requires early assessment of hazardous rainfall within a specified neighborhood. Geostationary meteorological satellites provide frequent, spatially continuous observations of cloud systems and storm evolution, but only indirectly reflect surface rainfall. Rain gauges measure surface accumulation directly, but only at irregular locations. Deep learning fusion methods commonly convert gauge observations to gridded representations before combining them with satellite observations. This treatment merges gauge measurements and local gauge support into a single representation, limiting how local surface conditions inform warning decisions and increasing the likelihood that local heavy-rainfall risk is underestimated or overestimated. We present PLUME (Precipitation Nowcasting with Late-Stage Residual Updates from Rain-Gauge Measurements for Short-Duration Heavy-Rainfall Events), a staged fusion model with gridded and point-origin gauge pathways. PLUME first combines a gauge-derived gridded rainfall background with FY-4A AGRI observations to form a base rainfall representation. The point-origin pathway uses local gauge features and observation support to update this representation through signed residuals, which the station decoder then maps to neighborhood-event probabilities. This staged design preserves the complementary roles of continuous spatial context, local rainfall measurements, and gauge availability. We evaluated PLUME for 0–3 h local short-duration heavy-rainfall warning over central and eastern China using an independent test set from May to September 2023. Across Barnes analysis, inverse distance weighting, and kriging, PLUME consistently improved the critical success index (CSI) over matched satellite–grid baselines. Relative CSI gains were 2.6–3.3% under complete gauge input and 3.6–5.9% under simulated gauge missingness. PLUME also reduced the associated CSI loss by 45.8–67.4%. Analysis of the variant with frozen batch normalization (BN) statistics showed that, under all three backgrounds and both input conditions, the largest positive probability updates were concentrated among baseline misses and converted some to hits, whereas the largest negative probability updates were concentrated among baseline false alarms and converted some to correct negatives. The dominant conditional CSI contribution shifted from false-alarm suppression at 0–1 h to missed-event recovery at 1–3 h. By using gridded and point-origin gauge information at different fusion stages, PLUME improves local heavy-rainfall warning through lead-dependent bidirectional refinement. Full article

The styling design of engineering equipment is jointly constrained by functional requirements, structural characteristics, and professional application contexts. Therefore, design decisions need to consider both subjective Kansei preferences and objective form characteristics. However, Kansei semantics and objective aesthetic indicators belong to different evaluation spaces, making their direct integration difficult. To address this issue, a design decision method for engineering equipment styling based on the mapping between Kansei semantics and objective aesthetic indicators is proposed. After establishing subjective and objective indicator systems, an explanatory contribution mapping matrix between Kansei semantic sub-criteria and objective aesthetic indicators is constructed using the LMG relative importance analysis method. The subjective weights are then projected into the objective aesthetic indicator space, and the mapped subjective equivalent weights are fused with objective weights through the entropy-modified Dempster–Shafer (D-S) evidence theory to obtain a comprehensive decision model for styling scheme evaluation. A high-speed railway contact-wire inspection vehicle (HRCIV) is selected as a case study. The proposed method exhibits high consistency with the preference ranking of an independent participant group, with a mean rank deviation of 0.533 rank positions on a 1–15 scale. The results demonstrate that the proposed method enables interpretable mapping and unified fusion of heterogeneous subjective and objective information, providing quantitative support for styling design decisions for function-constrained engineering equipment. Full article

Artificial intelligence (AI) for thyroid nodules should be evaluated against the decisions it changes, not diagnostic accuracy alone. This critical narrative review examines ultrasound, cytology, histology, and molecular models and distinguishes modality-specific prediction from implemented multimodal learning. The original PubMed/MEDLINE search through 10 August 2026 was supplemented on 25 September 2026 by targeted OpenAlex searches and checks of primary publications, reporting standards, and regulatory sources. Selected studies are compared by sample size, reference standard, acquisition setting, testing design, and performance with confidence intervals where available. We explain feature concatenation, intermediate representation learning, cross-attention, and decision-level fusion, and examine whether integration improves on the strongest component. Reported gains are inconsistent, and complete-case multimodal datasets may exclude the low-risk nodules most relevant to avoiding biopsy. High discrimination for histological malignancy or occult nodal disease does not demonstrate reduced overdiagnosis, fewer operations, or improved survival. A thyroid-specific translation roadmap therefore addresses active surveillance, partial verification and spectrum bias, Hashimoto thyroiditis and multinodular disease, cytology preparation shifts, category-specific thresholds, operator dependence, reporting frameworks, and regulatory change control. The priority is a calibrated, externally tested tool using information available at a defined clinical decision, followed by prospective evaluation of patient and workflow outcomes. A unified imaging–pathology–omics system and reliable AI prediction of surveillance progression remain research objectives rather than established clinical capabilities. Full article

Stroke is a leading cause of death and long-term disability, and auxiliary methods for rapid patient identification are needed. Previous studies based on electrocardiography (ECG) have attempted stroke classification using quantitative measurements and future stroke risk prediction using raw signals. However, measurements do not preserve all waveform detail, and models using raw signals alone may not sufficiently learn quantitative characteristics from limited data. We therefore propose a hybrid fusion framework that integrates the raw 12-lead signal and the quantitative measurements of the same ECG record. It combines a signal-only model, a feature-only model, and an interaction early fusion model that couples the two inputs at the raw data and representation levels, averaging their logits for the final prediction. Evaluation used 4147 ECG records from 4020 patients in a single-institution retrospective cohort, with patient-level splits. The proposed model achieved an F1-score of 77.6%, an area under the receiver operating characteristic curve (AUC) of 0.832, and a Brier score of 0.169. Among the evaluated models, the proposed framework achieved the highest AUC and lowest Brier score. The AUC improved by 0.034 over the best unimodal model, and an AUC of 0.820 was maintained under a year-based temporal holdout. These results suggest that integrating complementary representations of the same ECG at multiple levels can improve stroke classification. The framework could potentially serve as an auxiliary decision-support tool for identifying patients who require further evaluation, using ECG characteristics statistically associated with stroke. Full article

Accurate pruning target decision and pruning point localization are essential for intelligent walnut pruning but remain challenging because slender and crossing branches are difficult to segment during the dormant season. This study proposes an improved semantic segmentation method combined with a three-stage progressive rule inference strategy for pruning target decision and pruning point localization. SAA-TransUNet was developed by incorporating an enhanced atrous spatial pyramid pooling module (EABP-ASPP) and a context-aware gated feature fusion module (CAGF) to improve trunk and branch segmentation. The predicted masks were skeletonized to construct branch topology based on endpoints, bifurcation points, and branch segments. Geometric and topological features were then used to sequentially determine pruning targets for upright, inward-growing, and competing branches and localize their pruning points. SAA-TransUNet achieved an mIoU of 84.93% and an IoU_branch of 80.41%, with the latter increasing by 13.05 percentage points over the original TransUNet. On the independent test set, the pruning target decision achieved an overall precision of 86.34%, with a mean pruning point localization error of 3.52 pixels and an average processing time of 2.154 s per image. The proposed method integrates branch segmentation, topology analysis, pruning target decision, and pruning point localization, providing a basis for subsequent robotic pruning operations in dormant walnut trees. Full article

Physiological sensing can support electric micro-mobility safety research. Elevated heart rate during riding, however, may reflect physical exertion, psychological stress, motion artefacts, road context, or several of these factors. This paper presents RideSafe-AI, a non-clinical architecture-feasibility prototype. It separates data quality, exertion/load proxy estimation, public-dataset stress-label classification, and motion reliability before applying deterministic fusion and abstention. The Python reference checks input and output fields for raw physiological streams and direct identifiers, while the ESP32-S3 (Espressif Systems, Shanghai, China) checks incoming fields; Android constructs bounded inputs internally without an equivalent schema guard. PAMAP2 supports exertion/load proxy modelling, WESAD supports low/high laboratory condition-label classification, and official PhysioNet stress/exercise records support a subject-wise laboratory feature-set experiment. On the PhysioNet binary stress-versus-exercise task, full-feature XGBoost achieved 0.963 accuracy, 0.944 macro-F1, and a 0.015 exercise false-stress rate on 10,824 windows from 36 subjects. These results describe that public-dataset task; PhysioNet does not contain bike-frame motion, GPS terrain, or power telemetry and does not evaluate the complete RideSafe-AI pipeline. Deterministic scenarios verify selected decision rules, while two same-day technical rides by one participant demonstrate prototype integration and logging. These findings support architecture feasibility. Rider-stress accuracy, safety impact, and performance across riders and road conditions remain to be evaluated. Full article

Distributed sensing systems must ultimately reach a single shared decision, whether an alarm, a classification, or a control action. Achieving this is challenging because the two underlying technologies rely on conflicting assumptions. Quorum-based consensus protocols provide a consistent, totally ordered decision log, but they depend on relatively stable network connectivity and lose availability during network partitions. In contrast, disruption-tolerant communication relies on store-and-forward mechanisms that continue operating despite intermittent connectivity, yet these mechanisms are not designed to support the synchronous message exchanges required by consensus protocols. This work investigates how these two approaches interact by deploying the Raft consensus protocol over a resilient, multi-transport, store-and-forward communication layer. The system is evaluated in NS-3 under varying degrees of network partitioning, adversarial source participation, and node failures, while also validating the transport abstraction through packet-level analysis. The study yields four main findings. First, the integrated system maintains safety: committed logs remain consistent in every experiment, including scenarios where isolated leaders rejoin the network and discard uncommitted log entries. Once connectivity is restored, the system converges to a common state within a bounded recovery time. Second, intermittent connectivity can trigger excessive Raft term growth, a phenomenon effectively controlled by enabling PreVote together with CheckQuorum. This configuration reduces post-recovery convergence time by approximately an order of magnitude while also revealing a trade-off between the CheckQuorum timeout and the transport carry deadline. Third, decision accuracy follows the predictions of the Condorcet jury theorem: as partitions reduce the effective number of participating nodes, the system becomes more susceptible to adversarial influence, lowering the threshold at which incorrect decisions can dominate. Finally, we quantify the resilience of different data-fusion strategies against confidence-inflation attacks. Our evaluation focuses on crash-fault tolerance in the presence of malicious data sources, and we explicitly discuss the limitations of this model. Full article

Flow-based intrusion detection in unmanned aerial vehicle (UAV) networks can underuse protocol relationships and model complementarity. The Protocol-Semantic and Confusion-Pair-Aware Heterogeneous Ensemble (PS-CAHE) combines semantic features, training-only confusion-pair selection and two fusion weights. On 113,952 deduplicated UAVIDS-2025 flows, PS-CAHE achieved 95.58% Macro-F1 and 93.88% Matthews correlation coefficient on the fixed holdout. The full pipeline improved Macro-F1 by 0.83 percentage points over Original XGBoost; the pair-aware increment over global out-of-fold (OOF) weighting was only 0.15 points. Across 25 endpoint-pair-disjoint folds, this increment averaged 0.14 points (corrected 95% confidence interval: 0.06 to 0.22), positive in every fold. Three-base and matched two-base probability stacking scored higher in both nested protocols, with row-stratified advantages of 0.08 and 0.05 points. Strict no-port node-disjoint evaluation yielded 93.54% versus 93.78% for Original XGBoost; the corrected difference interval spanned zero, establishing no unseen-node advantage. Training-only temperature scaling reduced negative log-likelihood from 0.1097 to 0.1074 and expected calibration error from 0.88% to 0.49% without changing decisions. The small fusion increment has no demonstrated operational benefit. PS-CAHE offers directly inspectable, low-dimensional fusion on one simulator-generated benchmark; real-flight data, independent external datasets, temporal/mission shifts, open-set attacks and adversarial robustness remain unevaluated. Full article

Stock price forecasting is important for financial decision-making but remains challenging because of complex and volatile market dynamics. Although existing forecasting methods have achieved considerable progress, they still face several limitations. Decomposition-based approaches are sensitive to manually specified parameters, deep learning models often require high computational resources, and individual lightweight models may not adequately capture diverse temporal characteristics. To address these limitations, this study proposes an improved whale optimization-based variational mode decomposition and dual-branch normalized linear forecasting (IWVMD-DNFLinear) framework, which integrates adaptive sequence decomposition, complementary linear forecasting, and learnable fusion. An improved beluga whale optimization (IBWO) algorithm is employed to adaptively optimize the parameters of variational mode decomposition (VMD). The decomposed components are then processed by parallel decomposition-based linear forecasting (DLinear) and normalized linear forecasting (NLinear) branches with a learnable fusion mechanism. Extensive experiments on four stock datasets under multiple forecasting horizons demonstrate that IWVMD-DNFLinear achieves consistently competitive forecasting performance against the trained baseline models across different datasets and forecasting horizons. On the AAPL dataset under single-step forecasting, IWVMD-DNFLinear achieves the lowest root mean square error (RMSE) of 5.1220. The DNFLinear forecasting network contains only 743 trainable parameters and requires 0.0110 MB of model storage. The proposed framework achieves high forecasting accuracy while maintaining a parameter-efficient forecasting structure, providing a practical solution for stock price forecasting. Full article

As the core rotating components of the air supply system, the compressor and its blades are prone to crack initiation under long-term cyclic loading and high-temperature erosion. Failure to detect cracks in a timely manner may lead to blade fracture, subsequently causing catastrophic unit failure. Affected by environmental noise and the weak vibration signatures of blade cracks, the existing methods for detecting blade cracks lack sufficient accuracy and reliability. To address this problem, a cracked blade detection method is proposed based on a vibration deep fusion framework for compressors. Specifically, the proposed vibration deep fusion framework includes two parts, namely data-level and decision-level fusion. The proposed data-level fusion method divides the original vibration signals into low-frequency and high-frequency bands based on the sampling frequencies of the two vibration sensors and fuses these two frequency bands to suppress noise and enhance defect features. Moreover, the proposed decision-level fusion method can further fuse the preliminary results obtained from the one-dimensional convolutional neural network with two raw vibration signals and the data-level fusion signal, thereby obtaining the final blade crack detection result. To verify the proposed method, a compressor blade crack detection experimental platform is built to simulate the crack detection effect of blades under different working conditions, which covers blades with artificial cracks of different lengths in noisy and noiseless environments. The results show that the proposed method has a detection accuracy of over 96% with 10-fold cross-validation, and the ablation experiment verifies that the proposed method is superior to traditional single signal or single-stage fusion methods, which demonstrates its effectiveness for compressor blade crack detection. Full article

Osteosarcoma is an aggressive bone malignancy in children and adolescents, where early and accurate classification is essential for improving clinical outcomes. While deep learning methods have shown promising results in medical image analysis, most existing approaches rely on single-modality data or assume patient-level correspondence in multimodal learning, which is rarely available in real-world clinical settings. In addition, many fusion strategies are designed for paired datasets and depend on fixed weighting schemes, limiting their applicability to heterogeneous and independently collected data. To address these challenges, this study proposes a modality-agnostic ensemble learning framework across independent cohorts for osteosarcoma classification using histopathology and X-ray images. Each modality is first modeled independently using EfficientNet-B0, and decision-level integration is performed without requiring cross-modal alignment. Two fusion strategies are investigated: fixed-weight probability fusion and a proposed adaptive sample-level gating network that dynamically learns fusion weights based on modality-specific predictive confidence. Experiments conducted on publicly available histopathology (TCIA) and X-ray (BTXRD) datasets demonstrate that the proposed adaptive fusion approach achieves superior performance compared to fixed fusion. The adaptive gating network reaches an accuracy of 97.08%, outperforming fixed-weight fusion (94.15%) and unimodal baselines. These results confirm that adaptive decision-level fusion can effectively leverage complementary information across unpaired datasets. Overall, the proposed framework provides a practical solution for multimodal medical image classification under data fragmentation conditions, where paired datasets are not available. Full article

Polyethylene (PE) pipes are extensively used in gas- and water-distribution systems because of their corrosion resistance, toughness, and installation flexibility. However, the field reliability of PE pipeline systems often depends on the quality of butt fusion joints, whose integrity is highly sensitive to welding parameters and thermal history. Among process-induced defects, cold welding and over-welding are particularly critical because they are usually concealed, difficult to identify by visual inspection or short-term pressure testing, and strongly associated with long-term brittle failure. Existing reviews have discussed butt fusion joining or polymer non-destructive testing in a broad sense, yet a defect-oriented synthesis linking process deviation, interfacial evolution, material degradation, and microwave response remains limited. This review therefore focuses specifically on cold welding and over-welding as the two most safety-critical hidden defects in PE butt fusion joints. First, the paper summarises how insufficient heat input, poor wetting, and limited molecular-chain interdiffusion lead to weak interfacial healing and low effective entanglement density, whereas excessive heat input and unfavourable thermal history promote chain scission, thermo-oxidative degradation, abnormal recrystallisation, and microvoid formation. It then compares how these mechanisms influence tensile behaviour, slow crack growth resistance, and long-term service performance. On this basis, conventional radiography and ultrasonics, phased-array ultrasonics, nonlinear acoustics, and microwave non-destructive testing (MNDT) are critically compared. Particular emphasis is placed on the structure–property–signal pathway that links local perturbations in complex permittivity to changes in microwave scattering parameters, including |S11|, |S21|, phase drift, resonance shift, and quality-factor reduction. Finally, a hierarchical evaluation framework is proposed to connect microwave features with defect severity and engineering decisions, and the major barriers to wider engineering adoption—benchmark datasets, quantitative inversion, lift-off robustness, and standardisation—are discussed. By organising the field around a mechanism-based structure–property–signal framework, this review aims to provide a more rigorous basis for defect-oriented quality control and future intelligent inspection of PE butt fusion joints. Full article

This research presents an integrated, multidisciplinary framework for transitioning from traditional preventive maintenance to predictive (PdM) and risk-based maintenance (RBM) strategies applied to heavy excavation machinery (Liebherr R9250 and R9350 fleets). The study merge’s reliability engineering principles with advanced Industry 4.0 technologies—including Digital Twin architectures, Stacking Ensemble Machine Learning algorithms, and IoT sensor networks—to maximize Mechanical Availability (MA ≥ 85%) and Overall Equipment Effectiveness (OEE ≥ 75%). The study evaluates these performance targets as standard benchmarks synthesized from state-of-the-art literature for optimized fleets, rather than as values derived from a single isolated field experiment. Special emphasis is placed on the direct correlation between mechanical degradation (wear of Ground-Engaging Tools—G.E.T. and hydraulic systems) and ergo-physical impacts on operators (structural vibrations and elevated acoustic emissions), evaluated using the HFACS framework and biometric data fusion. From a financial perspective, mathematical modeling of cumulative cost functions (Total Cost of Ownership—TCO) demonstrates a break-even point at t = 1.428,57 operating hours and a net maintenance cost reduction of 18–22% over a 5000 h operational cycle. Furthermore, the implementation of Fault Tree Analysis (FTA) and 5 × 5 risk matrices confirms up to a 66% risk score reduction for critical failure modes. The synthesized outcomes validate the practical execution framework designed to support the ‘Triple Zero’ strategic paradigm (zero accidents, zero unplanned downtime, zero environmental compromise) as a long-term management vision and progressive operational objective, offering a sustainable decision-making model for surface mining operations. Full article