Search Results (87,087)

To establish an accurate and interpretable prediction framework for soybean lodging grade and clarify the core regulatory traits and differentiated driving mechanisms of soybean lodging under high-density drip irrigation cultivation, 356 spring soybean germplasm accessions were used as experimental materials in this study. Morphological and mechanical traits including plant height (PH), stem pulling force (SPF), internode number (IN) and petiole length (PL) were measured over two consecutive years of field phenotyping. Two composite evaluation indices, plant height/stem pulling force ratio (PH/SPF) and plant height/internode number ratio (PH/IN), were further constructed. Four machine learning algorithms were adopted to develop multi-classification models for soybean lodging grade prediction. SHAP analysis combined with three global sensitivity approaches (perturbation analysis, Sobol’ method and Morris screening) was applied to decipher the regulatory patterns of key traits. The results showed that lodging grade significantly affected soybean grain yield and explained 25–28% of the phenotypic yield variation; yield reduction tended to plateau under severe lodging. Compared with single indicators such as SPF and PL, the two derived composite indices could stably distinguish soybean accessions with different lodging grades and exhibited stronger discriminatory power. Model comparison revealed that the XGBoost model achieved optimal prediction accuracy and generalization stability for lodging grade, with a weighted F1-score of 95.34% on the test set, significantly outperforming the conventional linear model. Interpretability analysis demonstrated that the PH/IN, PH, and PH/SPF acted as the primary positive traits promoting lodging, while SPF was the sole protective trait. Driving factors of lodging presented obvious gradient heterogeneity: mild lodging was dominated by the imbalance of plant architecture ratio, whereas severe lodging was governed by the cumulative effects of PH and IN. Strong interactions existed among all measured traits. The interpretable machine learning framework established in this study can provide theoretical support and technical references for lodging-resistant germplasm screening and targeted plant architecture regulation for densely planted soybean under drip irrigation systems. Full article

Background: The leading cause of cancer-related deaths globally is lung cancer, and the P2X7 receptor (P2X7R) is a promising therapeutic target due to its role in the disease progression. Methods: A deep learning-based molecular generation framework that integrates a fragment-based drug method with Relational Graph Convolutional Networks (RGCNs) and a Wasserstein Generative Adversarial Network (WGAN) was employed. Known P2X7R targeting drugs were fragmented to construct a fragment library, which was used to generate new candidate molecules. The generated molecules from the model were evaluated for chemical validity, novelty, Lipinski’s Rule of Five compliance, quantitative estimate of drug-likeness (QED), lipophilicity (LogP), similarity using the Tanimoto coefficient, and binding affinity through molecular docking. Results: The model generated 4498 chemically valid molecules, including 968 unique and 384 novel molecules. Approximately 97% satisfied standard drug-likeness criteria, with QED values predominantly above 0.6 and LogP values within acceptable pharmacokinetic ranges. The novel molecules demonstrated an improved docking score against P2X7R compared to the seed molecules. Conclusions: Despite training on 5000 SMILES due to limited computational resources, the model achieved high validity, strong molecular diversity, and drug-like physicochemical properties, demonstrating the feasibility of a scalable, target-specific AI pipeline for lung cancer drug discovery using fragment-based molecular generation, RGCN and WGAN. Nevertheless, the biological activity of the generated molecules remains experimentally unvalidated, and the findings are based solely on computational analyses. Full article

In advanced manufacturing, designing multi-pulse composite driving waveforms for piezoelectric inkjet (PIJ) printing presents a constrained, high-dimensional, physical black-box optimization challenge. The feasible jetting region within the 12-dimensional parameter space is highly sparse; furthermore, traditional unconstrained optimization algorithms are prone to triggering nozzle flooding or actuator fatigue damage. To overcome this bottleneck, this paper proposes CA-TuRBO-m, a closed-loop collaborative architecture based on classifier-assisted multi-trust region Bayesian optimization. This architecture reconstructs the deposition morphology features on the substrate into a composite visual feedback source that implicitly incorporates fluid dynamics. Furthermore, it repurposes a Random Forest classifier into a dynamically iterating physical safety topological gating mechanism to actively intercept high-risk parameter combinations. Simultaneously, a multi-trust-region parallel exploration mechanism is introduced to balance global exploration and local exploitation. Experimental results demonstrate that over 200 online physical printing iterations, the proposed architecture reduces the number of invalid prints leading to system failures to an average of 3.8, achieving a high effective sampling rate of 98.1%. Without relying on complex fluid dynamic models, this approach enables precise morphological control over droplets of varying sizes and mitigates printing defects, successfully achieving multi-target adaptive regulation within a limited budget on a single physical platform. Full article

Leveraging advances in artificial intelligence algorithms, distributed acoustic sensing (DAS) based on phase-sensitive optical time-domain reflectometry (Φ-OTDR) has achieved high event-recognition accuracy through a variety of learning models. Nevertheless, further improving the accuracy of multi-event recognition remains a persistent challenge. In this paper, we propose a Dual-Domain Fusion Network (DD-FusNet) for vibration event recognition in Φ-OTDR sensing systems. To fully capture signal dynamics, the model simultaneously processes time- and frequency-domain representations, employing a crucial cross-attention mechanism to bridge these branches and enable dynamic, learnable interactions. Experimental results based on a six-class field engineering vibration event dataset collected by Φ-OTDR, containing car events, manual tapping, road breaker, excavation, leaking and noise, demonstrate that the proposed method achieves an average accuracy of 99.12%, significantly outperforming baseline methods by approximately 3 to 10 percentage points in accuracy, thereby ensuring the accuracy of multi-event recognition. We believe the proposed DD-FusNet will advance the recognition capabilities of Φ-OTDR systems in complex industrial sensing applications. Full article

Permanent Magnet Direct Current (PMDC) motors are widely used in applications requiring precise speed control due to their efficiency and high torque-to-inertia ratio. This work proposes a feedforward speed control strategy for PMDC motors based on model inversion, complemented by a disturbance rejection feedback term. A gray-box model is developed using only four concentrated parameters, avoiding the overdetermination problem of classical seven-parameter identification. These parameters are identified experimentally from step-response data using a nonlinear optimization approach. The proposed control law is validated on two commercially available PMDC motors with distinctly different dynamics: a fast-response motor (FC130SA) and a slower motor with a gearbox (GM25-370). Experimental results show that the proposed feedforward controller with disturbance rejection achieves lower or comparable Integral Squared Error (ISE) than optimally tuned PID/PI controllers, while significantly reducing overshoot (up to 66% in the fast motor) and maintaining lower or comparable Control Input Area (CIA), a metric commonly used in the literature to provide an indirect indication of control effort. Unlike classical controllers, the proposed method requires no per-reference gain tuning. These results show compelling evidence that inversion-based control with disturbance rejection is a viable, energy-efficient alternative to PID control for PMDC motor speed regulation. Full article

Evacuated-tube solar collectors charged with hybrid nanofluids can raise thermal output, but their coupled thermal and hydraulic response depends on many interacting variables, making the design an optimisation rather than a prediction problem. This study recasts it as a constrained, mixed-variable, three-objective task that maximises thermal efficiency and the Nusselt number while minimising pumping power over the hybrid pair, base fluid, weight fraction, component-one share and flow rate, and develops a memetic metaheuristic: the Adaptive Memetic Hybrid (AMH). A histogram gradient-boosted surrogate trained on 54,432 reduced-order runs, with held-out coefficients of determination of at least 0.9999, provides a fast screen, while a continuous reduced-order model validated to within 0.02 percent serves as the objective; the surrogate is accurate off-grid for efficiency but not for pumping power or the Nusselt number. Nine optimisers, comprising four baselines, three recent metaheuristics, and two AMH variants, were validated on twelve ZDT, DTLZ, and constrained problems over thirty trials using the hypervolume, generational distances, and spacing, and analysed with Friedman, Nemenyi, and Holm-corrected Wilcoxon tests. AMH attained the best mean Friedman rank of 3.08 (chi-square 65.7, p = 3.6 × 10⁻¹¹), significantly outperforming the recent methods and NSGA-III and remaining competitive with the strongest classical algorithms. On the collector, the reduced-order front recovers the 1512-design brute-force maximum efficiency to within 0.02 percent and improves the trade-off through continuous flow rates. The study is a deterministic, model-based optimisation process: the surrogate serves as a tool for fast screening and diagnostics, while the reconstructed reduced-order model is the objective for the final continuous optimisation. The collector application has a low effective design dimension, being governed mainly by the base fluid and the loop flow rate, so the decisive separation of the algorithms is established on the benchmark suite rather than on the collector. Experimental validation remains a task for future work. Full article

To effectively address the challenges of complex spatiotemporal conflicts, dynamic obstacle avoidance, and coordinated multi-objective optimization in preemptive multi-UAV logistics delivery within complex airspace, this study proposes a Hybrid Improved Multi-Objective Cuckoo Search algorithm (HI-MOCS) for preemptive multi-UAV cooperative logistics scheduling and planning. To overcome the limitations of conventional MOCS, including a low proportion of feasible solutions under complex constraints, susceptibility to local optima, and uneven distribution of multi-objective solution sets, a multi-constraint physical model and a multidimensional evaluation framework are established for preemptive scheduling. A positive knowledge-transfer mechanism based on the co-evolution of primary and auxiliary populations is developed, in which constraint-violation information is used to guide infeasible solutions toward the feasible region. A hybrid heuristic population initialization strategy combining emergency-order priority and spatial scanning rules is introduced to increase the proportion of high-quality feasible solutions in the initial population. In addition, a nonlinear dynamic adaptive parameter-adjustment strategy is designed to balance global exploration and local exploitation, while an iterative truncation-based environmental selection mechanism using the shortest-distance criterion is employed to improve the distribution quality of the Pareto solution set. The experimental results show that, in the benchmark scenario, HI-MOCS achieves an average increase of 33.26% in the total order completion rate and an average reduction of 15.34% in emergency response time compared with 11 multi-objective optimization algorithms, while also exhibiting favorable performance in terms of flight distance per completed order. The fleet-size analysis shows that the 15-UAV configuration achieves the lowest best mean fitness. The safety-distance analysis indicates that, compared with the other safety-distance settings, the 30 m setting increases the total order completion rate by an average of 26.55%, while reducing emergency response time and flight distance per completed order by 27.36% and 33.72%, respectively. The task-scale analysis shows that the 50-order scenario achieves the lowest best mean fitness. Further ablation experiments demonstrate that, compared with the average performance of MOCS and the four single-strategy variants, the complete HI-MOCS improves the total order completion rate by 20.27%, while reducing emergency response time and flight distance per completed order by 20.71% and 36.18%, respectively. The HV, IGD, and Pareto-front results further confirm that the synergistic effects of the four improvement mechanisms effectively enhance the multi-objective optimization performance and the quality of the nondominated solution set. The current study is still validated under simulation conditions assuming reliable GNSS positioning and communication links, without explicitly considering communication delays. Full article

Robotic Aerial Work Platforms (RAWPs) are subjected to multi-source excitations including wind gusts and inertial loads, which result in strongly nonlinear and time-varying coupled triaxial attitude vibrations. Conventional recurrent architectures such as LSTM and GRU suffer from gradient vanishing when processing long sequences, while Transformer utilize self-attention mechanisms to learn simple periodic correlations; however, the vibrations in RAWPs exhibit a complex time-series pattern composed of low-frequency oscillations superimposed with high-frequency impacts and accumulates errors through autoregressive decoding. To address these limitations, this paper proposes an improved Transformer model featuring dual-channel periodic positional encoding and global–local hybrid multi-head attention for one-shot multi-step long-sequence prediction of RAWPs attitude vibrations. The proposed method designs a dual-channel independent sine–cosine positional encoding with a tunable periodic modulation factor to explicitly embed the multi-scale periodicity priors of vibration signals and introduces a global–local hybrid attention mechanism that parallelly extracts transient amplitude impact features in the time domain and periodic fluctuation features in the frequency domain. A full-scale aerial experimental platform is established to collect triaxial vibration data under two operating conditions at a sampling frequency of 20 Hz. The results determine the optimal periodic modulation factor and input window length, and ablation studies validate the synergistic gains of the two proposed modules. Comparative results demonstrate that the proposed model achieves substantially reduced prediction errors. In terms of pitch angle, the proposed model achieves a performance improvement of 53.89% over Transformer, 52.11% over LSTM, and 57.67% over GRU. The proposed model effectively provides a reliable data-driven prediction framework for attitude monitoring and active vibration suppression of aerial work platforms. Full article

This study meta-analyzed the incremental contribution of artificial intelligence (AI) to flipped learning on students’ academic achievement, relative to flipped learning without AI. Following a protocol registered on the Open Science Framework (OSF), systematic searches were conducted in Web of Science and Scopus, supplemented by an ERIC search, with coverage through 25 July 2026; eligible studies used experimental or quasi-experimental designs comparing an AI-supported flipped condition with a matched non-AI flipped comparator on an academic-achievement outcome. Applying the comparator criterion strictly, ten studies (two randomized controlled trials and eight quasi-experimental studies; N = 937 learners) met all criteria; one further study whose comparator was conventional, non-flipped instruction was excluded from the primary analysis and retained only as a sensitivity check. Using Hedges’ g in a random-effects model (REML with the Knapp–Hartung adjustment), the pooled effect was positive, moderate and statistically significant: g = 0.67, 95% CI [0.32, 1.02], t(9) = 4.36, p = 0.002. Heterogeneity was substantial (I² = 64.6%, τ² = 0.086) and the 95% prediction interval ranged from −0.10 to 1.43, indicating that the true effect in a new setting could plausibly be large or close to negligible. Egger’s test was statistically significant (t = 3.25, p = 0.01), consistent with a small-study effect; excluding studies with fewer than 15 participants per arm attenuated the estimate to g = 0.53. Because the comparator was itself flipped learning, this estimate approximates the incremental contribution of the AI component rather than the combined effect of AI and the flipped model. Overall certainty of the evidence (GRADE) was low. AI can add value to flipped learning, but the benefit is conditional on implementation quality and context rather than an inherent property of the technology, and the findings should be read with caution given a small, heterogeneous and partly small-study-influenced evidence base. Full article

Predicting target intent in complex, dynamic multi-agent environments remains a formidable challenge due to incomplete perception and the highly dynamic nature of multi-target interactions. Conventional approaches—such as D-S evidence theory, expert systems, and Recurrent Neural Networks (RNNs)—are often constrained by data incompleteness and rigid behavioral assumptions, limiting their adaptability to dynamic high-value target identification and multi-target situational awareness on the ground. To address these challenges, a novel framework termed Threat Field–Gated Recurrent Unit (TF-GRU) is proposed. The TF-GRU framework integrates threat field modeling with a dynamic repair mechanism to enhance intent prediction under partial perception. Specifically, threat field modeling associates target attributes with intentions through the construction of static and dynamic threat fields, effectively capturing the temporal and semantic relationships among multiple targets. A particle filtering and dynamic time warping fusion strategy (PF-DTW) is employed to repair data gaps via short-term filtering and long-term trajectory matching, further refined by a neighborhood-angle constraint for accurate multi-target state estimation. In addition, trajectory and threat field features are processed using a Mish activation function and a threat-adaptive gating mechanism, which dynamically regulate information flow within the recurrent unit to model behavioral evolution. Experimental evaluations demonstrate that TF-GRU significantly enhances intent prediction accuracy under incomplete data conditions, thereby improving comprehensive situational awareness and supporting high-confidence decision-making in dynamic multi-target scenarios. Full article

The behavior of petroleum pool fires has important implications for fire safety and environmental protection due to heat release, smoke generation, and pollutant emissions. In this study, controlled pool-fire experiments were conducted using representative petroleum fuels with multiple pan diameters to investigate the coupled effects of fuel properties and geometric scale on combustion behavior and emission characteristics. Key parameters, including the heat release rate, smoke production rate, mass loss rate, major gaseous emissions, and soot characteristics, were systematically measured. The results show that increasing pan diameter accelerated fire development, increased combustion intensity, and generally enhanced cumulative gaseous emissions. Compared with kerosene, gasoline exhibited more rapid combustion and higher smoke production, whereas kerosene produced a more sustained heat-release process and higher cumulative gaseous emissions. Correlation analysis, principal component analysis, and principal component regression revealed that fuel thermophysical properties and geometric scale are the dominant factors governing combustion behavior and pollutant formation. The proposed statistical framework provides a practical approach for quantitatively relating fuel properties to heat-release characteristics. These findings improve the understanding of the coupled effects of fuel composition and fire scale on petroleum pool-fire behavior and provide experimental support for fire hazard assessment and combustion modeling. Full article

Tetrapod piled jacket foundations, widely adopted for large-capacity offshore wind turbines, are frequently affected by scour, which substantially alters their lateral mechanical responses. Nevertheless, existing studies on this subject remain limited and mostly adopt simplified uniform scour assumptions that deviate significantly from actual field conditions. This study conducted lateral-loading model tests on scoured tetrapod piled jacket foundations, with the local scour geometry idealized based on the non-uniform scour-hole morphology reported in field monitoring and flume test studies. The evolution law of the lateral bearing capacity of the foundations with scour development is revealed, and the differences in lateral bearing performance under uniform and non-uniform scour are systematically compared. A three-dimensional finite element model is established and validated against test data to verify its accuracy and reliability. Additionally, comprehensive parametric analyses are performed to supplement the experimental results, exploring the influences of flow angles, corresponding scour-hole morphologies and lateral load directions on the lateral bearing performance of tetrapod piled jacket foundations. The pile bearing mechanism and internal force distribution characteristics are further clarified. The research findings can provide a theoretical basis for the safe service of offshore wind turbines supported by tetrapod piled jacket foundations. Full article

Hydrogen embrittlement (HE) remains one of the key challenges limiting the safe and reliable deployment of steels in hydrogen production, storage, transportation, and utilization systems. Artificial intelligence (AI) and machine learning (ML) have emerged as powerful tools for predicting HE behavior, accelerating materials development and selection, while supporting engineering decision-making. This paper systematically reviews and critically evaluates AI/ML applications for HE in steels through a structured analysis of all studies published between 2010 and 2026. The review examines experimental, literature-derived, and computational datasets together with data preprocessing, feature engineering, AI/ML models, validation strategies, and prediction objectives. Experimental datasets remain the dominant source for predicting HE susceptibility, hydrogen concentration, fracture behavior, and hydrogen-assisted cracking, whereas computational datasets provide complementary mechanistic insights into hydrogen diffusion, trapping, crack propagation, and atomistic interactions across multiple scales. Image- and signal-based modalities within these datasets further enable computer vision and automated defect characterization. Beyond systematically synthesizing the current literature, this review provides a critical and analytical evaluation of AI/ML datasets, model families, and prediction applications. It also establishes a practical framework for selecting appropriate AI/ML approaches according to dataset characteristics and engineering objectives. Future research should focus on standardized HE databases, rigorous external validation, explainable and uncertainty-aware AI, and closer integration of data-driven and physics-based approaches to improve predictive reliability, mechanistic understanding, and the safe deployment of hydrogen-compatible steels. Full article

Background: Vascular endothelial growth factor (VEGF) has emerged as a potential therapeutic target in osteoarthritis (OA). Although several anti-VEGF agents are widely used in clinical practice, their comparative effects on osteoarthritic cartilage remain unclear. This study compared the histopathological effects of intra-articular bevacizumab, ranibizumab, and aflibercept in an experimental rat model of OA. Methods: Experimental OA was induced in 36 rats using anterior cruciate ligament transection. One month later, animals received two intra-articular injections of bevacizumab (2 mg/kg), ranibizumab (0.5 mg/kg), aflibercept (3.2 mg/kg), or saline at 3-week intervals. Histopathological evaluation was performed using the Osteoarthritis Research Society International (OARSI) grading and staging system by two blinded pathologists. The primary outcome was the total OARSI score. Group comparisons were performed using the Kruskal–Wallis test followed by Dunn–Holm post hoc analysis. Results: Three animals were lost during anesthesia and surgical induction of OA before treatment allocation, leaving 33 rats for analysis. Significant overall differences were observed among the groups for the primary outcome of total OARSI score (p = 0.026) and the secondary outcomes of OARSI grade (p = 0.042) and OARSI stage (p = 0.019). Ranibizumab demonstrated numerically lower OARSI grade and total scores than the control group, whereas both groups had the same median OARSI stage score. Bevacizumab showed numerically lower OARSI grade and total scores than the control group but a slightly higher median stage score. Pairwise analysis showed significantly lower OARSI stage and total OARSI scores in the ranibizumab group than in the aflibercept group, whereas no significant differences were observed between any treatment group and the control group after correction for multiple comparisons. Conclusions: Anti-VEGF agents demonstrated differential histopathological responses in experimental OA. Although no treatment significantly improved histopathological outcomes compared with the control group after adjustment for multiple comparisons, ranibizumab showed lower OARSI stage and total scores than aflibercept. These findings indicate that the evaluated anti-VEGF agents may not exert equivalent biological effects and warrant further investigation in adequately powered experimental studies. Full article

Anisotropy is a first-order control on rock mass behaviour in foliated, veined, and fabric-controlled ground, yet it remains underrepresented in routine geotechnical practice. This paper presents a review of the literature on anisotropic rock mass behaviour, organised around three identified gaps: the absence of an integrated explanation linking structure, and excavation response; the underdeveloped treatment of spatially variable anisotropy in mine-scale numerical models; and the limited incorporation of precursory monitoring evidence into mainstream geomechanical design. The review is selective rather than encyclopaedic, covering experimental characterisation of directional strength and stiffness, the mechanical role of microstructure and fabric morphology, excavation damaged zone development, constitutive and spatial modelling strategies, seismic source characterisation, laboratory precursor detection, and data collection practice. The review finds that anisotropy changes the path to failure and the spatial organisation of damage, not merely the peak strength, and that constitutive sophistication without a geologically admissible representation of fabric is insufficient to reproduce the mechanism, location, and temporal evolution of damage around underground excavations. Full article