Search Results (7,478)

To address the limited endurance of unmanned underwater vehicles (UUVs) during long-duration missions, this study proposes a foldable and retractable wave energy converter (WEC) conformally integrated with the UUV hull. A two-degrees-of-freedom heave-coupled dynamic model of the float–UUV system is established, and parameter-matching relationships are derived using complex dynamic stiffness and impedance-matching theory. A bidirectionally coupled STAR-CCM+-AMESim co-simulation framework resolves the nonlinear viscous flow field, relative motion, and PTO dynamic response in closed loop. Under regular wave conditions defined based on a representative Bohai Sea state, the effects of the transmission ratio and spring stiffness on the coupled motion and equivalent resistive load power output are systematically investigated. Under the specified wave condition, average electrical power varies unimodally with both parameters, reaching 70.8 W at a transmission ratio of 15 and a spring stiffness of 4642 N/m; the corresponding peak power is 161.2 W. The system is more sensitive to increases than decreases in transmission ratio, suggesting a value slightly below the theoretical optimum for engineering design. The instantaneous power shows an asymmetric double-peak pattern, indicating a shift in dominance between direct float-driven generation and spring-mediated energy release. Agreement between theory and co-simulation provides numerical cross-validation and offers a theoretical basis and numerical methodology for designing and optimizing WECs on mobile UUV platforms. Full article

With the rapid development of sustainable photovoltaic power generation, energy-storage-coordinated grid-connected photovoltaic systems have been widely adopted to stabilize power output and enhance grid adaptability. Aiming at the low fault tolerance of conventional photovoltaic grid-connected systems under grid voltage swell disturbances, this paper designs a single-stage power management system for grid-connected photovoltaic generation and studies its energy-storage-coordinated high-voltage ride-through (HVRT) technology. The single-stage topology boasts simple structure, low cost and high conversion efficiency, yet faces prominent stability risks under voltage swell faults. The system integrates photovoltaic units, energy storage modules and grid-connected interfaces to implement flexible bidirectional power dispatching. A three-phase AC/DC converter realizes photovoltaic maximum power point tracking (MPPT), and the energy storage module connects to the DC bus via a dual half-bridge (DHB) converter to restrain power fluctuations. Under HVRT faults, the energy storage coordination strategy elevates DC bus voltage to maintain stable grid-tied operation without disconnection. Different from schemes requiring extra hardware or complicated control optimization, the proposed method realizes stable bus voltage regulation and flexible energy scheduling with zero additional hardware cost. Simulations and experiments validate the rationality, feasibility and outstanding fault-ride-through performance of the designed system. Full article

Academic dropout is costly for students and institutions, and support actions are most effective early, when little information beyond enrollment records is available. However, the most informative predictors are derived from academic trajectory data, such as grades and course progression, which only become available after students have completed one or more terms. This study investigates dropout prediction using supervised machine learning (ML) applied to records of 66,820 students across 45 undergraduate programs of the State University of Maringá, Brazil (2002–2022), trained exclusively on static enrollment-time variables, since these were the only institutional data available. Although this restricts the information the models can use, it also allows students who may be at risk to be flagged very early. Decision Tree, Random Forest, and eXtreme Gradient Boosting (XGBoost) models were evaluated in ten training configurations, covering the complete dataset, the complete-generations subset, a temporal cohort split, academic centers, individual programs, and resampled variants. Evaluation was based on per-class precision, recall, and F1-score, together with threshold-free and calibration measures (PR-AUC, ROC-AUC, and Brier score). XGBoost performed best in the institution-wide configurations, reaching a dropout recall of 0.64 at a precision of 0.51 with undersampling and, without resampling, a PR-AUC of 0.586 against a dropout prevalence of 0.364, which corresponds to 1.61 times the performance of random ranking. Under the temporal split, recall at the default threshold fell from 0.37 to 0.18, while ROC-AUC changed little (0.708 to 0.679); this loss reflects miscalibration under changing dropout prevalence and can be corrected without retraining the model. The main contributions are an evaluation methodology for enrollment-time dropout prediction that controls information leakage and includes temporal validation, and a deployment protocol that uses the model outputs to prioritize student support when follow-up capacity is limited. Full article

Agricultural machinery operates under variable loads, impacts, dust, and changing soil–crop interactions, allowing faults to propagate through energy, material, information, and control pathways. This qualitative review synthesizes 195 research publications across fault formation, trustworthy diagnosis, prognostics and proactive risk control, maintenance and recovery, and case-based evidence assessment. Empirical findings are reported separately from review-derived recommendations, the authors’ conceptual requirements, and mandatory provisions of applicable standards or law. The literature most consistently supports controlled-fault identification, selected single-machine field monitoring, and localized operational compensation. Evidence is weaker for transfer across machines and seasons, calibrated prognostics, safety authorization, post-repair verification, and fleet-scale deployment. On this basis, the review recommends mission profile-specific fault boundaries, traceable diagnostic outputs, explicit uncertainty and abstention, and risk decisions linked to remaining work and resources. It also proposes a diagnostic passport, risk evolution trajectory, repair-effectiveness label, case-evidence matrix, and closed-loop repair workflow as review-level organizing constructs. No included study continuously followed the same machine or fleet through diagnosis, prognosis, authorization, maintenance, acceptance, and feedback; the framework therefore connects evidence-supported stages theoretically rather than claiming end-to-end validation. Full article

Large language models and AI agents are extending data-engineering automation beyond isolated artifact generation toward end-to-end processes in which agents interpret requirements, select data, generate transformations, invoke tools, validate results, and communicate analytical outputs. This shift introduces risks that conventional notions of data quality and execution success do not fully capture. A dataset may satisfy established quality standards, and a generated query may execute without technical errors, while the agent still selects an incorrect metric, combines incompatible analytical grains, accesses unauthorized data, or draws conclusions that are insufficiently supported by evidence. This paper develops a conceptual framework for Agentic Data Engineering centered on Quality of Agentic Data Use , defined as the extent to which an agent uses and communicates data in accordance with task, semantic, quality, security, governance, and provenance requirements. An evidence-informed analysis of Data Contracts, Semantic Layers, Data Quality, Guardrails, AI Governance, and Data Provenance shows that these foundations provide essential but fragmented capabilities. The proposed framework integrates and extends them through four core artifacts: Agentic Data Contracts, Agentic Expectations, Agentic Data Provenance, and Agentic Data Governance. It also introduces an execution lifecycle, a reference architecture, a failure taxonomy, and a multidimensional evaluation framework. A governed sales-analysis scenario illustrates how the proposed artifacts interact throughout an agent-mediated data process. In addition, a controlled Databricks prototype and a complementary benchmark comprising 10 cases and 40 executions demonstrate the framework’s technical feasibility and support the independent computation of enforcement indicators. The benchmark highlights the value of separating generation from validation while also showing that the current validation and automated-repair mechanisms require further calibration. These preliminary findings do not establish generalized improvements in safety, correctness, or reliability. Rather, they provide an operational foundation for broader empirical evaluation of trustworthy agent-mediated data-engineering processes. Full article

Aiming at the problems of large speed synchronization error and prominent speed overshoot existing in conventional PID control algorithms widely adopted for diesel-engine parallel-unit speed-governing systems, this paper proposes a speed control algorithm based on Model Predictive Control (MPC) for dual-diesel-engine parallel operation. A quasi-steady-state method is employed to establish the coupled state-space model for dual-engine parallel operation. Leveraging the prediction-optimization and multi-constraint regulation characteristics of MPC, the fuel-injection outputs of the two diesel engines are regulated respectively by two independent SISO MPC controllers, which share the identical speed reference and are coordinated at the logic level through the Stateflow engagement/disengagement state machine to achieve speed synchronization. Comparative simulations under different operating conditions are carried out on the Matlab/Simulink platform. The simulation results show that the proposed MPC algorithm can effectively suppress speed fluctuations between the two diesel engines under the tested operating conditions compared with the traditional PID control. Furthermore, a hardware-in-the-loop test platform is constructed for validation in the embedded environment. The test results reproduce all operating conditions of offline simulation, achieving zero speed overshoot and restricting the dual-engine synchronization deviation within ±3 rpm, which are consistent with the offline simulation results. The real-time computational capability and engineering feasibility of the proposed algorithm are therefore verified. The proposed method is applicable to stable speed-governing scenarios of marine dual-diesel-engine parallel-unit sets. Full article

Human body static electricity is one of the major hazards in electrostatic-sensitive industrial environments. Existing noncontact measurement methods often rely on explicit physical modeling and generally require accurate prior knowledge of the sensor layout, target motion state, and scene geometry, which limits their applicability in complex dynamic scenarios. To address this issue, this paper proposes a neural network-based data-driven method for estimating human body electrostatic potential from noncontact electrostatic measurements. The proposed method uses four-channel noncontact electrostatic sensor signals as inputs and the synchronously measured reference body potential as the target output. Sixteen neural network architectures, including recurrent neural networks, convolutional neural networks, attention-based networks, and hybrid models, are systematically evaluated. The Gray Wolf Optimizer is further used to optimize key hyperparameters of each model. A 5 m × 5 m experimental scene is established, and 36 groups of synchronized time-series signals are collected from three subjects under two motion states. Training and validation datasets are constructed using a sliding time-window method, and the effects of model architecture, window length, and subject–motion condition on inversion performance are analyzed. The results show that the proposed method can effectively estimate human body electrostatic potential from noncontact measurements. Among the evaluated models, Trans-LSTM achieves the best performance. With a window length of 1 s, it obtains a validation normalized root-mean-square error of approximately 0.139 and a coefficient of determination of approximately 0.72. Compared with a representative existing method under the same coverage area and sensor layout, the proposed approach reduces the NRMSE from 0.22 to 0.139. These results demonstrate that the proposed method provides a feasible approach for remote, real-time, and noncontact monitoring of human body electrostatic potential in complex environments. Full article

Explainable artificial intelligence (XAI) is increasingly important in scientific and engineering applications where predictive performance alone is insufficient and model outputs must also be physically credible, transparent, and reliable under unseen conditions. This study proposes a domain-informed XAI framework for predicting total suction in xanthan gum-treated clays using 139 experimental observations covering different mineralogical, moisture, polymer-dosage, and curing conditions. Eleven linear, kernel-based, ensemble, boosting, and physics-guided algorithms were evaluated using leakage-resistant five-fold grouped cross-validation, including a matched constrained–unconstrained HGB comparison with identical model settings. The methodological contribution is an evidence-linked XAI validation protocol in which model explanations are not accepted from feature attribution alone, but are audited through their agreement with leakage-resistant grouped generalization, physically constrained response directions, matched experimental contrasts, residual behavior, predictive uncertainty, and applicability-domain support. Selective monotonic constraints, physics-guided residual learning, SHAP explanations, and nonlinear response visualization are integrated within this protocol as complementary sources of evidence rather than treated as independent indicators of interpretability. The unconstrained histogram–gradient-boosting model achieved the highest out-of-fold predictive performance (R² = 0.958, RMSE = 0.098, and MAE = 0.070 in log10(MPa)). The corresponding monotonic model produced R² = 0.935, RMSE = 0.121, and MAE = 0.094 but eliminated the directional violations detected in the unconstrained response, revealing a measurable trade-off between predictive accuracy and guaranteed physical consistency. Explanations identified moisture content as the dominant negative control and revealed that xanthan-gum effects were non-monotonic and dependent on curing, moisture, and mineralogy. The residual model remained interpretable but underperformed the leading ensembles. Overall, the framework validates explanations against experimental contrasts, physical directions, grouped generalization, residual behavior, uncertainty, and domain support, offering a transferable strategy for trustworthy XAI in structured scientific datasets. Full article

Multi-channel electroencephalogram (EEG) emotion recognition in intelligent human–computer interaction has gained substantial attention. However, deep-neural-network-based approaches can involve substantial architectural and parameter-tuning complexity. Conventional cross-validation uses pseudo-random fold construction, whereas best-discrepancy systematic cross-validation (BDSCV) is used here as a deterministic and systematic data-partitioning strategy for fold construction. We present an integrated parallel gcForest-based EEG emotion recognition framework that processes complementary 1D and 2D EEG representations in parallel and fuses their class-vector outputs. On the public Database for Emotion Analysis using Physiological Signals (DEAP) dataset, the framework reported accuracies of 93.65%, 94.78%, 93.23%, and 95.01% for arousal, valence, dominance, and liking, respectively. The reported accuracy and F1-score results indicate competitive recognition performance under the adopted subject-dependent segment-level evaluation setting. Full article

Transformer lifecycle data across organizations are typically vertically partitioned among material suppliers, manufacturers, logistics service providers, testing agencies, and operation and maintenance units. This study presents a reproducible multi-layer vertical federated learning (VFL) prototype that integrates salted-hash identifier matching, additive sharing of local score vectors over finite fields, and a local public key infrastructure with a signature-based audit verification mechanism. A deterministic synthetic dataset is first constructed, comprising 5200 aligned records and 31 predictor variables, which are partitioned among five participants with varying numbers of features per participant. Second, across five validation runs, the VFL models under both the standard block-wise and score-sharing paths achieved an AUC of 0.8825 ± 0.0119, an F1 score of 0.7367 ± 0.0249, and an accuracy of 0.8102 ± 0.0183 on the test set. The classification results of both paths were fully consistent with the centralized gradient-descent logistic regression baseline. Notably, the score-sharing path exhibited a maximum log-odds deviation of only 2.22 × 10⁻⁸ on the test set, with no prediction discrepancies observed. Third, across 10 independently generated synthetic populations, the nonlinear output mechanism highlights the limitations of linear models: the AUC of vertical federated learning (VFL) drops to 0.6457 ± 0.0171, while Extra Trees and HistGradientBoosting achieve 0.7731 ± 0.0149 and 0.7743 ± 0.0139, respectively. Finally, in a separate residual-sharing diagnostic test, when 1 to 4 participants jointly shared the residuals, the label inference AUC remained around 0.499–0.500; however, when all five participants shared or plaintext residuals were used, the labels could be fully recovered. Both simple membership inference diagnostic tests yielded results close to random. The local signature log verifier rejected all 700 injected faults and accepted the 400 clean control log events. These results validate the feasibility of numerical reproducibility and local audit functionality under synthetic data and single-process conditions, yet they are insufficient to demonstrate end-to-end label privacy protection, malicious security, effectiveness on real data, or real-time ledger performance. Full article

This study examines how the vertical spacing of a multi-height 360° camera array affects novel-view synthesis quality in indoor scenes reconstructed with 3D Gaussian splatting (3DGS). Although 3DGS enables high-quality real-time scene representation, its output depends on the geometric arrangement of the input views, which single-lens workflows secure through repeated captures at several heights. A rig of three vertically arranged 360° cameras was evaluated in four indoor spaces differing in ceiling height, structure, and capture path. Three spacing conditions were applied in each space, denoted as Set-L, Set-D, and Set-H. These are ordinal positions within the vertical range each space allows rather than fixed absolute distances, since an identical spacing sits differently in rooms of different scale. The 360° footage was stitched into equirectangular video and reframed into multi-view image sequences, yielding 120 datasets from four spaces × three conditions × ten repeated captures. Novel-view synthesis quality was measured with the PSNR, SSIM, and LPIPS on validation views withheld from training. Because normality and homogeneity of variance were not satisfied, a robust two-way factorial analysis of variance based on 20% trimmed means was used, with robust post hoc comparisons. Spacing, spatial characteristics, and their interaction were significant for all three metrics, so no single spacing was preferable across every space. The wide setting performed best in the low-ceiling repetitive space, the narrow setting in the open space, and the baseline setting in the largest space with variable ceiling height. In one space, comprising stepped, near-symmetric seating, no condition was distinguishable, and dispersion across repeated captures was an order of magnitude larger than elsewhere, indicating that, where a repetitive structure is extensive, the limiting factor is capture stability rather than the choice of spacing. The findings are consistent with a trade-off between vertical viewpoint separation and inter-view overlap, and provide exploratory, space-conditional reference points for indoor 3DGS capture rather than a standardized specification, with relevance to virtual exhibitions, architectural visualization, and digital-twin construction. Full article

Efficient parsing of manufacturing process data is a key enabler for the informatization of intelligent manufacturing systems. However, network isolation in aerospace-specific job shops makes large volumes of unstructured shop-floor data, such as handwritten production reports and equipment logs, inaccessible to existing information systems. To tackle this, we propose a comprehensive parsing framework that covers data acquisition, parsing, and structured output. For handwritten report parsing, we devise a collaborative pipeline comprising text detection via the Differentiable Binarization Network (DBNet); text recognition using an enhanced Convolutional Recurrent Neural Network (CRNN) that incorporates Asymmetric Convolution (AC) and a Simple Attention Module (SimAM)-based residual module (SimRes, short for SimAM ResNet), referred to as AC-SimRes-CRNN; and table structure extraction via TableMaster. The predicted table cell coordinates, detected text-region coordinates, and recognized text contents are subsequently aggregated to reconstruct complete tables, which are then exported as Excel files. Experiments on the CASIA-HWDB2x and IAM datasets show that AC-SimRes-CRNN achieves an Accurate Rate (AR) of 91.36% and a Correct Rate (CR) of 93.17% on Chinese handwritten text recognition and a Character Error Rate (CER) of 7.85% and a Word Error Rate (WER) of 26.83% on English handwritten text recognition, demonstrating competitive performance against representative methods. Ablation studies validate the contributions of both AC and SimRes. A case study on an aerospace equipment maintenance report further illustrates the component-level feasibility of the proposed workflow. Full article

Precise prediction of pressure drop and drag-reduction performance is essential for improving the hydraulic efficiency and reducing the energy demand of crude-oil pipeline transportations. Therefore, this study aims to develop an integrated computational fluid dynamics (CFD)–machine learning (ML) framework for predicting pressure drop (∆p), drag reduction (DR), pumping power reduction (PPR), energy savings (ES), and flow-rate enhancement (Q) in turbulent crude-oil pipeline flow containing drag-reducing agents (DRAs). The investigated system considers the effect of pipeline length (L), diameter (D), surface roughness (ε), operating temperature (T), and DRA concentration (25–200 ppm). The Reynolds-average Navier–Stokes equations (RANS) were solved using the shear stress transport (SST) k-ω turbulence model approaching near-wall resolution of y⁺ ≈ 1 for DRA3 at 20 ppm. The CFD modelling was first used to validate an experimental benchmark and subsequently used to expand the available dataset over the investigated operating conditions. The combined experimental–CFD dataset was then employed to develop a multi-output Kolmogorov–Arnold network (KAN) surrogate model. The proposed framework predicted DR up to 44.2%, PPR of approximately 55 W, ES of 30%, and flow-rate enhancement up to 5–10(${\displaystyle \frac{\mathrm{L}}{\mathrm{d}\mathrm{a}\mathrm{y}}}$). The KAN model effectively captured the nonlinear relationships among DRA characteristics, pipeline geometry, and operating conditions, achieving R² = 0.9318 for PPR prediction. The novelty of the proposed work lies in integrating a validated, near-wall-resolved SST k-ω CFD model with a multi-output KAN surrogate model, combining physics-based flow analysis with rapid data-driven prediction of hydraulic and energy-performance indicators. Full article

Atmospheric effects substantially influence remote-sensing reflectance retrieval in optically complex inland waters. This study evaluated seven atmospheric correction approaches (QUAC, FLAASH, Sen2Cor, LaSRC, 6S, C2RCC, and ACOLITE) for Sentinel-2 MSI and Landsat-8/9 OLI imagery over the Danjiangkou and Luhun reservoirs. The evaluation used 67 quality-controlled, temporally matched in situ spectral observations and satellite matchups. Performance was quantified using the squared Pearson correlation coefficient (r²), root mean square error (RMSE), and average unsigned relative error (AURE). Laboratory-measured chlorophyll-a (Chl-a) concentrations were used to develop sensor-specific retrieval models and to examine how atmospheric-correction differences propagated into Chl-a estimates and spatial patterns. Because residual aerosol and sun-glint effects may remain after atmospheric correction, an exploratory SWIR-based adjustment was evaluated for the C2RCC visible-band outputs. In the pooled-band analysis, C2RCC yielded the most favorable balance of the evaluated metrics for both sensor datasets. However, performance varied among bands, and the Landsat-8/9 B5 output showed very weak covariation with the in situ measurements. Within the model-development dataset, Sen2Cor achieved the highest Sentinel-2 r² (0.762), whereas C2RCC achieved the lowest Sentinel-2 RMSE (2.30 mg/m³). C2RCC achieved both the highest Landsat-8/9 r² (0.689) and the lowest RMSE (3.18 mg/m³). Independent temporal validation used 14 Luhun observations from 2024. Sen2Cor yielded the lowest Sentinel-2 RMSE and AURE (1.658 mg/m³ and 29.28%). For Landsat-8/9 OLI, C2RCC yielded the highest r² (0.536), lowest RMSE (3.468 mg/m³), and lowest AURE (44.08%). Relative errors increased in weak-signal near-infrared bands, underscoring the need for band-specific interpretation. The SWIR-based adjustment improved both RMSE and AURE for Sentinel-2 MSI but did not provide a consistent improvement for Landsat-8/9 OLI. An exploratory comparison of quality-screened imagery from 2016 to 2025 showed broadly similar reservoir-scale Chl-a patterns in C2RCC-derived products from the two sensors. These results provide reservoir-specific evidence for atmospheric-correction selection and Chl-a retrieval under the sampled conditions. Full article

Continuous operation is increasingly expected of Internet of Things (IoT) and wireless sensor network (WSN) nodes, yet practical solar front ends must account for source variability, intermediate storage, conversion losses, sensing overhead, and battery-management constraints. This article develops and evaluates a discrete, two-stage photovoltaic front end for such nodes: a perturb-and-observe (P&O) buck stage tracks the maximum power point of a 20 W Solarland SLP020-12U module (rated 17.2 V, 1.16 A) and feeds an intermediate storage bus, while a PI-compensated SEPIC stage regulates the IoT rail to 3.2 V independently of that bus voltage. Closed-loop MATLAB/Simulink simulations are reported at 1000, 800, and 600 W/m². The reported conversion figures originate from an idealized switching model and should therefore be interpreted as simulation-only values rather than measured prototype efficiency. A low-cost Arduino-based prototype confirms correct switching behavior and a 20.0048 kHz PWM signal, but the available captures lack synchronized, calibrated input/output power logging; consequently, no hardware efficiency, MPPT tracking efficiency, regulation error, ripple, or settling-time figure is claimed. The revised manuscript makes this simulation-to-hardware boundary explicit, adds the power cost of sensing and data conversion to the loss discussion, strengthens the battery-management and deployment caveats, and defines the measurements required for full quantitative validation. Full article