Search Results (1,050)

Aero-engine remaining useful life (RUL) prediction supports condition-based maintenance, yet most data-driven models assume fixed sensor availability. Power-supply, acquisition, or communication failures can invalidate this assumption. We propose the Remaining Useful Life Dual-Attention Robust Network (RUL-DARNet), which combines a convolutional neural network–long short-term memory (CNN–LSTM) backbone with training-stage whole-channel Sensor Dropout (SD), Mask-Aware (MA) feature attention, and temporal attention. SD exposes the model to reduced sensor sets, whereas MA excludes unavailable channels from feature-attention normalization using an explicit availability mask. Ten seeds and ten paired masks were evaluated across four Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) subsets under controlled synthetic sensor unavailability. Trajectory metrics use a 125-cycle label cap and equal engine weighting. At the prespecified FD001 40% Missing Completely at Random endpoint, RUL-DARNet attained an RMSE of 17.474 ± 1.410 cycles, compared with 19.023 ± 0.706 for SD-only. Adding MA after SD reduced RMSE by 1.549 cycles in nine of ten seeds after Holm correction. Benefits weakened or reversed under value-related missingness, multiple operating conditions, and several trajectory-level outages. Training-stage exposure accounts for most of the observed robustness, while mask-aware reweighting provides a smaller, conditional benefit within the tested C-MAPSS protocols when availability labels are reliable and the remaining channels retain degradation information. Full article

This study addresses a regime-conditioned question in electricity theft detection: when only monthly billing series and inspection-confirmed labels are available, which supervised model families recover irregular consumption without relying on the temporal resolution of advanced metering infrastructure (AMI)? The working sample comprises 4000 utility customers and 864 confirmed theft cases, each represented by 53 monthly kWh values from January 2021 to May 2025. After majority-class undersampling that retains all theft observations and construction of a balanced 1:1 learning set, eight classifiers are compared under an 80/20 stratified split: K-Nearest Neighbors, Decision Tree, Support Vector Machine, Random Forest, two dense multilayer perceptrons, Long Short-Term Memory, and a one-dimensional Convolutional Neural Network. Performance is assessed through threshold-optimized accuracy together with precision, recall, F1-score, the area under the receiver operating characteristic curve (AUC), and confusion matrices. On the hold-out test set, Random Forest and the compact dense network both reach an accuracy of 0.685; Random Forest attains the highest AUC (0.748) and F1-score (0.677). Even so, these models miss about one-third of the hold-out theft accounts (59 and 65 false negatives out of 173). Sequential deep models underperform on this short monthly regime. The results support ensembles and compact dense networks for monthly theft screening and indicate that AMI-oriented sequential gains do not transfer automatically to 53-point billing vectors under the present protocol. Full article

This study used explainable Random Forest models to identify predictors of adolescents’ expected active political participation using ICCS 2022 data from 1487 students in Schleswig-Holstein, Germany. Three predictor specifications were examined to distinguish prediction based on variables conceptually close to the outcome from prediction based on more general civic characteristics. The Full Random Forest achieved the highest predictive accuracy (R² = 0.426), with expected legal and electoral participation as the strongest predictors. Removing proximal participation measures reduced predictive accuracy but revealed a stable set of predictors. In the most restrictive model, citizenship self-efficacy, beliefs about missing conventional citizenship, and civic knowledge consistently ranked highest under both permutation importance and SHAP. Civic knowledge showed substantial predictive importance but a consistently negative association with the outcome, supported by the correlation (r = −0.139), SHAP dependence, and ALE analyses. Random Forest’s predictive advantage over benchmark models was modest and specification-dependent. Full article

Barrett’s esophagus (BE), the intestinal metaplasia arising from chronic gastroesophageal reflux disease, is the principal identifiable precursor of esophageal adenocarcinoma (EAC), whose incidence rose from 0.4 to 2.8/100,000 person-years between 1975 and 2017 and whose prognosis, once symptomatic, remains poor. Because outcomes depend on intercepting the metaplasia–dysplasia–carcinoma sequence, diagnostic accuracy is decisive. White-light endoscopy with Seattle-protocol biopsy remains the reference standard, yet it is constrained by the following three interrelated weaknesses: sampling error, as random forceps biopsies interrogate only about 3.5% of the Barrett’s mucosa; poor reproducibility of dysplasia grading, with interobserver agreement of only κ 0.24–0.27 for the pivotal distinction of low-grade dysplasia; and a substantial burden of missed disease, with roughly one-quarter of EACs diagnosed within a year of an index endoscopy reported as nondysplastic. This review synthesizes the technologies converging to address these gaps. Advanced imaging, encompassing high-definition endoscopy, narrow-band imaging, acetic acid chromoendoscopy, and the optical-biopsy platforms confocal laser and volumetric laser endomicroscopy, raises dysplasia yield by approximately 34% over standard white-light examination. Image-enhanced endoscopy improves targeted detection while remaining complementary to structured biopsy sampling. Molecular, genetic, and epigenetic biomarkers, notably DNA-content abnormalities, p53 immunohistochemistry, and multi-gene methylation panels, add an objective read on progression risk. Their pairing with non-endoscopic sampling, including Cytosponge-TFF3, capsule endoscopy, exhaled volatile organic compounds, and circulating microRNA liquid biopsy, is reshaping screening at population scale, while artificial intelligence standardizes interpretation and narrows the expert–nonexpert gap. Together these advances point toward a risk-stratified, multimodal paradigm, though prospective validation and cost-effectiveness evidence remain prerequisites for guideline adoption. Full article

Indoor exposure to fine particulate matter (PM2.5) is a major environmental-health concern in Northern China, where severe ambient pollution, coal-based district heating and diverse residential building stocks coexist. Previous studies have been constrained by small numbers of independent residences, limited geographic coverage, and predominantly pairwise (bivariate) analyses. This study presents a seasonal monitoring campaign covering 48 residences across eight Northern Chinese cities. Paired indoor and outdoor PM2.5 was measured at 5 min intervals and aggregated to 31,094 valid hourly observations (from 32,256 scheduled hourly slots after exclusion of 1162 h with missing or invalid 5 min readings) over seven consecutive days in four seasons, alongside building, behavioural and meteorological covariates. Because repeated hourly measurements are clustered within residences, the primary analysis is a linear mixed-effects model (LMM) with a residence-level random intercept and a first-order autoregressive residual structure (marginal R² = 0.71; conditional R² = 0.91; intra-class correlation = 0.34), supported by ordinary-least-squares (OLS) models with CR2 finite-cluster-robust standard errors and a daily average sensitivity analysis. Adjusted LMM coefficients indicate that outdoor PM2.5 (β = 0.872, p < 0.001), window-open fraction (β = 0.548, p < 0.001) and cooking events (β = 0.046, p < 0.001) were positively, and air-purifier operation (purifier-on fraction β = −0.902, p < 0.001) was negatively, associated with indoor PM2.5; continuous purifier operation was associated with 59.4% lower indoor PM2.5 (an observational association, not a causal effect). Building-level infiltration factors (F_inf) averaged 0.28 ± 0.15. Estimated outdoor contributions to indoor PM2.5 ranged from 54.8% to 66.3% across an assumed penetration factor of p = 0.6–1.0 (60.7% at p = 0.8), peaking in the heating season. The mass-balance analysis estimated a median deposition rate of 0.53 h⁻¹ and a median air change rate of 0.52 h⁻¹ (window-open periods > 2 h⁻¹); these are model-derived, not directly measured quantities. These findings provide multivariable, cluster-aware evidence for envelope airtightness, informed window operation and correct air-purifier use to reduce residential PM2.5 exposure in Northern China. Full article

Colonoscopy is a key screening method for colorectal cancer (CRC), but its effectiveness is limited. Computer-aided detection (CADe) and computer-aided diagnostics (CADx), as part of an artificial intelligence (AI) system, improve the detection and optical characterization of lesions. This review maps and analyzes the evidence on the application of AI in colonoscopy, with a focus on the detection, segmentation, and characterization of colon neoplasms, available platforms, architectural models and implementation. The review was conducted in accordance with JBI and PRISMA-ScR guidelines, using the PCC framework. Meta-analyses, randomized controlled trials, systematic and narrative reviews, observational studies, guidelines, and consensus documents on the use of AI systems in different phases of colonoscopy were searched. CADe significantly improves adenoma detection and reduces the number of missed lesions. CADx, segmentation, depth of invasion assessment, and detection of learned lesions remain limited and heterogeneous. CADe has strong evidence for improving ADR, whereas current evidence for CADx remains insufficient to support a “resect-and-discard” strategy. Further cost-effectiveness studies are needed. Commercial platforms vary in their features and level of clinical validation. Colonoscopy using AI is a current topic with rapid development. This scoping review comprises heterogeneous literature covering clinical applications, technical aspects, the potential benefits and limitations of AI in improving colonoscopy performance and reducing the burden of colorectal cancer. Further trials involving diverse patient populations across different countries are needed to validate and extend the current evidence. Full article

Wi-Fi stations (STAs) repeatedly scan channels to maintain access-point information, but a conventional full Wi-Fi scan incurs substantial delay and energy overhead. SplitScan mitigates this cost by allowing neighboring STAs to exchange reports through Bluetooth Low Energy (BLE) advertisements and divide channel-set scanning responsibilities. We develop a graph-theoretic and stochastic framework for this cooperation over a tri-band channel-set space. For fixed BLE graphs, the framework models cooperation through closed-neighborhood coverage, yielding topology-dependent load limits and a centralized min–max reference for evaluating distributed assignments based on neighbor reports. For spatially random deployments, a random geometric model captures density-dependent load scaling and yields an energy break-even condition. To address report loss, Robust SplitScan-r uses redundant scan assignments to improve reliability. Monte Carlo simulations show that increasing the mean BLE neighbor count from 0.45 to 9.0 reduces the mean Responsible Channel Set (RCS) load from 21.9 to 5.1 sets, a 77% decrease. At an empirical mean degree of 6.68, the mean maximum load is 13.8 sets for SplitScan and 25 sets for SplitScan without the load threshold, while the centralized greedy and exact min–max assignments achieve 8.2 and 7.8 sets, respectively. When each BLE report is delivered with probability 0.9, increasing redundancy from $r=1$ to $r=3$ lowers the bound on missing scan information for a channel set from ${10}^{-1}$ to ${10}^{-3}$, while increasing the scan energy. Overall, BLE connectivity determines the attainable cooperative-scanning gain, and the centralized comparisons quantify the maximum-load difference between distributed decisions based on neighbor reports and centralized assignments. Full article

Automatic reseeding in spoon-chain potato planters requires reliable classification of miss-seeding, normal, and multiple-seeding events. Oblique viewing and continuous spoon motion cause scale changes, boundary truncation, blur, and occlusion, which destabilize single-frame predictions. The proposed system treats each seed-spoon passage as one control event. It combines lightweight detection, track-level learning, and multi-frame classification with air-blow clearing and missed-seed reseeding. Based on the target-size distribution, the detector removes the P3 head from YOLOv8n and restores shallow details through space-to-depth rearrangement and gated addition. Frame quality combines boundary distance, sharpness, adjacent-box intersection over union, and box-area stability. High-quality frames form a track prototype, while low-quality frames receive stronger constraints. A quality-gated temporal network classifies each spoon as containing 0, 1, or at least 2 seed potatoes. Each model was trained with three random seeds under the same data split. Compared with YOLOv8n, the detector reduced the parameter count and computational cost by 35.8% and 48.3%, respectively. Track-level learning improved event accuracy by 3.67 percentage points over frame-level training. The seven-frame model achieved 93.41% event accuracy and 93.80% macro-F1. At 0.5 m/s, the closed-loop recognition, air-blow, and reseeding success rates were 93.7%, 96.4%, and 97.1%, respectively. These results provide a practical basis for real-time, event-level planting-quality control in spoon-chain potato planters. Full article

This paper investigates the distributed fusion estimation problem for multi-sensor cyber-physical systems (CPSs), where the communication channels from local estimators to the fusion center are subject to random packet dropouts. Packet dropouts induced by either network congestion or intermittent denial-of-service (DoS) attacks are modeled as Bernoulli random variables. When a local estimate is lost, a prediction compensation strategy is adopted at the fusion center, where the missing data are replaced by their one-step predictors. By constructing an augmented state consisting of the original state, local prediction errors, and virtual measurements, the multi-sensor system is transformed into a stochastic system with random parameter matrices and one-step autocorrelated noises. Based on the transformed system, a distributed state fusion (DSF) filter is proposed via the innovation analysis method, whose filter gain depends on the successful-reception probabilities. The stability of the proposed DSF filter is analyzed, and a sufficient condition for the existence of a steady-state filter is obtained. The steady-state gain can be pre-computed offline, thereby reducing the online computational burden. Simulation results validate the effectiveness of the proposed algorithm. Full article

Background: WordPress plugins account for the large majority of disclosed CMS vulnerabilities, and learning-based detectors report high accuracy on synthetic corpora and random splits. Methods: We build a benchmark from 1757 real plugin CVEs (6666 indexed identifiers, 1281 plugins), yielding 30,860 labeled PHP functions in two negative-sampling variants. An audit of the released artifacts found exact normalized-code duplicates crossing test folds (166 groups/1086 records in Variant A; 151/676 in Variant B) and identical code carrying contradictory labels (730 and 456 groups). After removing conflicting groups, merging exact duplicates and collapsing token- and AST-level clone families, the corpora contain 15,294 (Variant A) and 24,917 (Variant B) records with zero shared hashes, clone families or plugins across folds. Ten detectors are compared under plugin-grouped five-fold cross-validation on complete held-out folds: a static taint analyzer, TF-IDF Random Forest and SVM, Bi-LSTM, unidirectional LSTM, a taint-augmented neural model, a nested out-of-fold stacking ensemble, and three baselines (majority class, a source/sink presence rule, and a single-feature function-length classifier). Results: The best macro-F1 is 0.563 (Variant A, binary, SVM) and 0.601 (Variant B, binary, nested stacking). Neural models do not lead in any of the six configurations, and the single-feature length baseline matches or exceeds the TF-IDF Random Forest in four of six. The source/sink rule attains an MCC close to zero, and the taint analyzer misses about 98% of vulnerabilities. Manual review of 400 functions finds 48.8% of cleaned positive labels to be genuine vulnerabilities. Conclusions: Under duplicate-controlled evaluation, the differences between detector families are small compared with the variation attributable to label quality, which appears to be the more binding constraint. All data, folds, predictions and fingerprinted result files are released. Full article

Automated concrete-surface classification increasingly employs multimodal sensing, although a wide range of systems assume that all sensors remain continuously available, reliable, and suitable for fusion. This study presents Reliability-Aware Selective Multimodal Fusion (RASMF), an image-first framework in which supplementary acoustic evidence is requested conditionally, assessed for signal-quality anomalies, fused when appropriate, and rejected when necessary. RASMF was evaluated on 4094 class-associated image-acoustic observations using 5 duplicate-aware outer folds and three random seeds under clean, simulated-degradation, and missing-modality conditions. At the nominal 10% acquisition budget, realised acoustic acquisition was 25.67%. Relative to the predefined missing-aware image-first baseline, mean error decreased from 1.9777% to 1.4499%, an absolute reduction of 0.5279 percentage points and a relative reduction of 26.69%, with a two-way bootstrap 95% confidence interval of −1.6220 to −0.0661 percentage points. The mean Brier score decreased from 0.03033 to 0.01688. RASMF produced lower mean error in 12 of 16 simulated-degradation and modality-loss conditions, with 6 condition-specific confidence intervals excluding zero in its favour. Increasing acoustic acquisition progressively reduced mean error and calibration error but increased processing demand; estimated pipeline time was 24.77 ms at the nominal 10% setting compared with 20.60 ms at the nominal 0% setting. A simpler logit stacker achieved a slightly lower mean error of 1.4295%, and the difference from RASMF was not statistically supported. The results therefore support RASMF as an explicit reliability-aware selective decision architecture relative to the designated baseline, without establishing universal predictive superiority over simpler fusion strategies. Full article

This study examines whether possession-oriented tactical profiles are associated with reported absolute injury burden across Europe’s Big Five domestic football leagues. Team-level tactical statistics were integrated with publicly available injury data covering eight seasons (2017–18 to 2024–25), yielding 780 team-season observations. Three outcomes were examined: total reported injury events, total days lost, and total matches missed. League-specific feature selection used Elastic Net and Random Forest models combined through a performance-weighted Meta Score, followed by K-Medoids clustering with cosine distance to identify relatively higher- and lower-possession tactical profiles. Injury outcomes were compared using Welch’s t-tests and permutation tests with false discovery rate correction, supplemented by club-clustered bootstrap confidence intervals. After multiplicity correction, higher-possession profiles showed greater injury-event counts and matches missed in the Premier League and La Liga, all three outcomes in Ligue 1, and matches missed in the Bundesliga. The La Liga total-days-lost difference did not remain robust under club-clustered resampling. No adjusted difference was identified in Serie A. These findings indicate exploratory associations between possession-oriented tactical profiles and reported absolute injury burden rather than causal or exposure-adjusted effects. As player exposure, fixture volume, and major club- and competition-level confounders were not fully controlled, the results should not be interpreted as estimates of epidemiological injury risk. Full article

Fault diagnosis in distributed systems is challenged by complex service dependencies, cascading anomaly propagation, and similar symptom patterns. This paper proposes a knowledge-enhanced iterative reasoning framework that integrates large language models (LLMs) with a numerical domain knowledge graph (KG). The KG encodes fault–symptom relations, anomaly directions, and training-derived mean and standard-deviation intervals. Structured prompting first generates candidate faults; interval verification then rejects numerically inconsistent candidates. For retained candidates, counterfactual reasoning constructs hierarchical causal chains, KG traversal refines missing or inconsistent links, and a deterministic evidence score supports acceptance, exclusion, early stopping, and fallback across at most five iterations. Under the common 68-case evaluation protocol for eight known single-root-cause faults in the controlled Redis-based testbed, the complete framework achieved 100.00% Accuracy, Macro-F1, and Balanced Accuracy with GPT-4o and GPT-5.2, compared with 91.18% accuracy for KG-only reasoning and 85.29–89.71% for Random Forest, XGBoost, and Transformer baselines. GPT-3.5 reached 98.53%, whereas LLaMA-3.1-8B reached 80.88%, showing that the incremental KG–LLM gain is backbone-dependent. Five GPT-4o repetitions and three GPT-5.2 repetitions yielded 100.00% ± 0.00, and all three metrics remained at 100.00% across the evaluated Z-score thresholds, iteration limits, and interval tolerances. The framework therefore provides highly accurate, stable, and traceable diagnoses within the evaluated Redis-based distributed-service protocol, while providing explicit intermediate reasoning and solution retrieval. Full article

Rapid, reliable, and energy-efficient data collection is essential for disaster response, where terrestrial communication networks may be disrupted or unavailable. Unmanned Aerial Vehicles (UAVs) provide a flexible means of collecting critical sensing data, but their operation is constrained by limited onboard energy, stochastic wireless conditions, complex three-dimensional environments, and stringent latency requirements. This paper presents a structured multi-UAV framework that separates mission optimisation into spatial, temporal, and safety layers. In the spatial layer, a three-dimensional Travelling Salesman Problem with Neighbourhoods (3D-TSPN) formulation enables UAVs to collect data by entering valid sensing regions rather than visiting exact sensor coordinates. An Age of Information (AoI)-aware Genetic Algorithm (GA) optimises the sensor-visitation sequence, while Rapidly Exploring Random Tree Connect (RRT-Connect) generates obstacle-aware feasible paths in the three-dimensional environment. In the temporal layer, a Lyapunov-based controller selects between local processing and binary offloading to a single Mobile Edge Computing (MEC) node according to queue backlog, processing delay, energy consumption, information freshness, wireless-link feasibility, and task deadlines. In the safety layer, continuous-time conflict detection and bounded temporal or spatial adjustments are used to monitor and mitigate inter-UAV and obstacle-related risks. The framework is evaluated under stochastic wireless, mobility, computation, and obstacle conditions using 20 independent random seeds. Across the corresponding 20 proposed-policy runs, it achieves a 100% mission-validity rate, complete sensor coverage, no dropped tasks, and zero final collision or near-miss events. Compared with planning-oriented and MEC-oriented baselines, the proposed framework achieves lower information age, average delay, processing delay, energy consumption, and system cost under the evaluated conditions, while maintaining reliable multi-UAV coordination. The layered design also clarifies the contribution of each component: 3D-TSPN provides spatial flexibility, the AoI-aware GA improves route sequencing, RRT-Connect supports obstacle-aware path feasibility, Lyapunov control enables queue-aware processing decisions, and safety monitoring supports coordinated multi-UAV operation. These results indicate that integrating spatial planning, computation control, and safety coordination within a clearly separated layered architecture can provide an effective solution for multi-UAV disaster-response data collection in complex three-dimensional environments. Full article

GRACE and GRACE-FO satellite gravimetry provide unique observations of terrestrial water storage (TWS), but their coarse effective resolution and intermittent temporal gaps limit water-resource applications at subregional and basin scales. This study presents a framework to temporally reconstruct and spatially downscale GRACE TWS anomalies for the transboundary Bug River Basin (Poland–Ukraine–Belarus), a region where in situ monitoring is limited and further disrupted by the 2022 war in Ukraine. First, missing monthly GRACE TWS anomalies (2002–2024) are imputed using a Random Forest model driven only by lagged GRACE values (1–3 months) and seasonal timing, thereby avoiding potential information leakage. Second, the continuous GRACE signal is downscaled to 0.1° using an independent set of hydroclimatic predictors with lagged and rolling features, together with elevation, land type and lithology. Model performance is evaluated under strict spatiotemporal holdouts and cross-validation. The key methodological advance is a bias-aware, block-wise mass-conserving correction that reconciles downscaled fields with the original GRACE water mass at coarse resolution. After downscaling to 0.1°, systematic residual biases between aggregated high-resolution estimates and GRACE observations are quantified monthly and redistributed within spatial blocks using river-runoff-based weights. This procedure enforces exact mass closure while preserving physically meaningful sub-grid variability. Full article