Search Results (830)

Despite the extensive application of machine learning (ML) methods to educational datasets, few studies have provided a systematic benchmarking of the available algorithms with respect to both predictive performance and interpretability of the resulting models. In this work, we address this gap by comparing a range of supervised learning methods on a freely available dataset concerning two high schools, where the goal is to predict student performance by modeling it as a binary classification task. Given the high feature-to-sample ratio, the problem falls within the small-data learning regime, which often challenges ML models by diluting informative features among many irrelevant ones. The experimental results show that several algorithms can achieve robust predictive performance, even in this scenario and in the presence of class imbalance. Moreover, we show how the output of ML algorithms can be interpreted and used to identify the most relevant predictors, without any a priori assumption about their impact. Finally, we perform additional experiments by removing the two most dominant features, revealing that ML models can still uncover alternative predictive patterns, thus demonstrating their adaptability and capacity for knowledge extraction under small-data conditions. Future work could benefit from richer datasets, including longitudinal data and psychological features, to better profile students and improve the identification of at-risk individuals. Full article

Predicting the severity of traffic accidents remains challenging due to the limited ability of existing methods to extract deep semantic information from unstructured accident narratives, as traditional approaches typically depend on structured data alone. This study proposes a severity prediction approach enhanced by semantic risk reasoning derived from large language models (LLMs). A prompt-engineering template is designed to guide LLMs in extracting proxy semantic features from accident descriptions, forming an enriched feature set that incorporates causal logic. These semantic features are fused with traditional structured features through three integration strategies—direct feature concatenation, optimized feature selection, and model-level fusion. Experiments based on 4013 accident records from expressways in Yunnan Province, China, demonstrate that models using LLM-derived semantic features significantly outperform those relying solely on structured features. Notably, the LightGBM model utilizing semantic features within a balanced learning framework achieves a severe accident recall of 77.8%. While model-level fusion proves optimal for XGBoost (improving Macro-F1 to 0.6356), we identify a “feature dilution” effect in other classifiers, where high-quality semantic reasoning is compromised by low-quality structured noise. These findings indicate that the proposed approach effectively enhances the identification of high-risk accidents and offers a novel semantic-aware solution for traffic safety management. Furthermore, the obtained results provide actionable insights for traffic management agencies to optimize emergency response resource allocation and formulate targeted accident prevention strategies. Full article

Accurate mapping of localization error distribution is essential for assessing passive sensor systems and guiding sensor placement. However, conventional analytical methods like the Geometrical Dilution of Precision (GDOP) rely on idealized error models, failing to capture the complex, heterogeneous error distributions typical of real-world environments. To overcome this challenge, we propose a novel data-driven framework that reconstructs high-fidelity localization error maps from sparse observations in TDOA-based systems. Specifically, we model the error distribution as a tensor and formulate the reconstruction as a tensor completion problem. A key innovation is our physics-informed regularization strategy, which incorporates prior knowledge from the analytical error covariance matrix into the tensor factorization process. This allows for robust recovery of the complete error map even from highly incomplete data. Experiments on a real-world dataset validate the superiority of our approach, showing an accuracy improvement of at least 27.96% over state-of-the-art methods. Full article

As an emerging immersive media format, point clouds (PC) inevitably suffer from distortions such as compression and noise, where even local degradations may severely impair perceived visual quality and user experience. It is therefore essential to accurately evaluate the perceived quality of PC. In this paper, a no-reference point cloud quality assessment (PCQA) method that uses complexity-driven patch sampling and an attention-enhanced Swin-Transformer is proposed to accurately assess the perceived quality of PC. Given that projected PC maps effectively capture distortions and that the quality-related information density varies significantly across local patches, a complexity-driven patch sampling strategy is proposed. By quantifying patch complexity, regions with higher information density are preferentially sampled to enhance subsequent quality-sensitive feature representation. Given that the indistinguishable response strengths between key and redundant channels during feature extraction may dilute effective features, an Attention-Enhanced Swin-Transformer is proposed to adaptively reweight critical channels, thereby improving feature extraction performance. Given that traditional regression heads typically use a single-layer linear mapping, which overlooks the heterogeneous importance of information across channels, a gated regression head is designed to enable adaptive fusion of global and statistical features via a statistics-guided gating mechanism. Experiments on the SJTU-PCQA dataset demonstrate that the proposed method consistently outperforms representative PCQA methods. Full article

Industrial surface defect detection is essential for ensuring product quality, but real-world production lines often provide only a limited number of defective samples, making supervised training difficult. Multimodal anomaly detection with aligned RGB and depth data is a promising solution, yet existing fusion schemes tend to overlook modality-specific characteristics and cross-modal inconsistencies, so that defects visible in only one modality may be suppressed or diluted. In this work, we propose DCRDF-Net, a dual-channel reverse-distillation fusion network for unsupervised RGB–depth industrial anomaly detection. The framework learns modality-specific normal manifolds from nominal RGB and depth data and detects defects as deviations from these learned manifolds. It consists of three collaborative components: a Perlin-guided pseudo-anomaly generator that injects appearance–geometry-consistent perturbations into both modalities to enrich training signals; a dual-channel reverse-distillation architecture with guided feature refinement that denoises teacher features and constrains RGB and depth students towards clean, defect-free representations; and a cross-modal squeeze–excitation gated fusion module that adaptively combines RGB and depth anomaly evidence based on their reliability and agreement.Extensive experiments on the MVTec 3D-AD dataset show that DCRDF-Net achieves 97.1% image-level I-AUROC and 98.8% pixel-level PRO, surpassing current state-of-the-art multimodal methods on this benchmark. Full article

The increasing volume of produced water (PW) generated by oil extraction activities has intensified the need for environmentally sustainable strategies that enable its reuse and valorization. Biotechnological approaches, particularly those involving the microbial production of value-added compounds, offer a promising route for transforming PW from an industrial waste into a useful resource. In this context, bacterial exopolysaccharides (EPS) have gained attention due to their diverse functional properties and applicability in bioremediation, bioprocessing and petroleum-related operations. This study evaluated the potential of Lelliottia amnigena to synthesize EPS using oilfield PW as a component of the culture medium in stirred-tank bioreactors. Three conditions were assessed: a control using distilled water (dW), PW diluted to 25% (PW25%) and dialyzed PW (DPW). Batch experiments were conducted for 24 h, during which biomass growth, EPS accumulation and dissolved oxygen dynamics were monitored. Post-cultivation analyses included elemental and monosaccharide composition, scanning electron microscopy and rheological characterization of purified EPS solutions. EPS production varied among treatments, with dW and DPW yielding approximately 9.6 g L⁻¹, while PW25% achieved the highest productivity (17.55 g L⁻¹). The EPS samples contained fucose, glucose and mannose, with compositional differences reflecting the influence of PW-derived minerals. Despite reduced apparent viscosity under PW25% and DPW conditions, the EPS exhibited physicochemical properties suitable for biotechnological applications, including potential use in fucose recovery, drilling fluids and lubrication systems in the petroleum sector. The EPS also demonstrated substantial adsorption capacity, incorporating salts from PW and contributing to contaminant removal. This study demonstrates that PW can serve both as a substrate and as a source of functional inorganic constituents for microbial EPS synthesis, supporting an integrated approach to PW valorization. These findings reinforce the potential of EPS-based bioprocesses as sustainable green technologies that simultaneously promote waste mitigation and the production of high-value industrial bioproducts. Full article

Iodine is essential for thyroid hormone synthesis and is particularly critical during pregnancy, where excess and mainly its deficiencies can impair fetal neurodevelopment and increase maternal complications. Bromine has also gained attention due to its potential to interfere with iodine metabolism and contribute to adverse health effects when present in excess. Monitoring iodine and bromine in biological samples, especially urine and serum, is therefore important for assessing thyroid function and population health. This work presents a simple and robust ICP-MS method for simultaneous determination of bromine and iodine in urine and serum. The procedure uses a 20-fold dilution with 10 mmol L⁻¹ ammonia containing 0.1% (w/w) EDTA-2Na, ensuring solution stability, minimizing sample-to-sample variability, and eliminating the need for matrix-matched calibration. EDTA-2Na effectively prevents precipitation of metal species at high pH, avoiding blockages in the sample introduction system. Method accuracy was confirmed through certified reference materials and spike-recovery experiments, both showing suitable agreement for the two analytes. Precision was consistently strong (RSD < 6%), and low detection limits were achieved (0.78 μg L⁻¹ for Br and 0.24 μg L⁻¹ for I). The use of a high-efficiency nebulizer enabled analysis with only 50 µL of sample, making the method suitable for limited-volume specimens. Overall, this approach provides a sensitive, accurate, and practical solution for large-scale population studies and clinical applications. Full article

The present study aimed to investigate the effect of Turnera diffusa supplementation on the quantitative and qualitative characteristics of semen in Duroc boars (n = 4). The experiment was divided into two periods, each corresponding to the duration of one spermatogenic cycle: a control period (40 days) (CP) and an experimental period (40 days) (EP). Nutrition and environmental conditions were kept constant throughout both periods. During the experimental period, each boar received a daily supplement of 7 g of Turnera diffusa extract. In each period, five ejaculates were collected from each boar included in the study. The ejaculates were evaluated for volume, sperm concentration, motility, agglutination, number of insemination doses obtained per ejaculate after dilution, and sperm viability after 24, 48, and 72 h of storage. The results of a two-way repeated measures ANOVA showed that the combined effect of boar × treatment significantly influenced ejaculate volume (p < 0.01) and viability after 48 h of storage (p < 0.05). The results of the two-way repeated measures ANOVA showed that treatment with the tested additive T. diffusa significantly affected sperm survival during storage for 24 h (p < 0.01), 48 h (p < 0.001), and 72 h (p < 0.05). Bonferroni post hoc analysis indicated that T. diffusa significantly affected only the parameters related to sperm viability, namely survival rates at 24 h (p < 0.001), 48 h (p < 0.01), and 72 h (p < 0.01). The findings of this study demonstrate that the application of the tested supplement, at the specified dose and duration, has a positive effect on semen quality in boars. Full article

A quantitative understanding of methane leakage has become essential for safety design as eco-friendly fuel systems expand in modern ship applications. To address this need, controlled methane-release experiments were conducted in a large indoor chamber (30 × 16 × 20 m) to evaluate how sensor-network resolution (1 m vs. 0.5 m spacing) influences dispersion measurement and 5% Lower Explosive Limit (LEL)-based risk assessment. Initial tests with a 1 m grid showed that most sensors detected only low concentrations except for near the release nozzle, demonstrating that coarse spatial resolution cannot capture the primary dispersion pathway or transient peaks. This limitation motivated the use of a 0.5 m high-density sensor network, which enabled clear identification of the dispersion centerline, concentration-gradient development, early detection behavior, and the evolution of diluted regions, particularly under buoyancy-driven plume rise. Experimental results were compared with CFD simulations using the RNG k–ε and k–ω GEKO turbulence models. Strong agreement was obtained in peak concentration, concentration-rise rates during the accumulation phase, and LEL-based dispersion distances. These findings confirm the suitability of the selected turbulence models for predicting methane behavior in large enclosed spaces and highlight the sensitivity of model–experiment agreement to measurement resolution. The results provide an experimentally grounded reference for sensor layout design and verification of gas-detection strategies in ship compartments, fuel-gas preparation rooms, and modular supply units. Overall, the study establishes a methodological framework that integrates high-resolution experiments with CFD modeling to support safer design and operation of methane-fueled vessels. Full article

Droplet microfluidics enables high-throughput, compartmentalized reactions using minimal reagent volumes, but most implementations rely on precision-fabricated chips and external pumping systems that limit portability and accessibility. Here, we present a handheld vibrational droplet generator that repurposes a consumer electric toothbrush and a modified disposable pipette tip to produce nearly monodisperse water-in-oil droplets without microfluidic channels or syringe pumps. The device is powered by the toothbrush’s built-in motor and controlled by a simple 3D-printed adapter and adjustable counterweight that tune the vibration amplitude transmitted to the pipette tip. By varying the aperture of the pipette tip, droplets with diameters from ~100–300 µm were generated at rates of ~100 droplets s⁻¹. Image analysis revealed narrow size distributions with coefficients of variation below 5% in typical operating conditions. We further demonstrate proof-of-concept applications in digital loop-mediated isothermal amplification (LAMP) and microbiological colony-forming unit (CFU) assays. A commercial feline parvovirus (FPV) kit manufactured by Beyotime Biotechnology Co., Ltd. (Shanghai, China), three template concentrations yielded emulsified reaction droplets that remained stable at 65 °C for 45 min and produced distinct fractions of fluorescent-positive droplets, allowing estimation of template concentration via a Poisson model. In a second set of experiments, the device was used as a droplet-based spreader to dispense diluted Escherichia coli suspensions onto LB agar plates, achieving uniform colony distributions across the plate at different dilution factors. The proposed handheld vibrational generator is inexpensive, easy to assemble from off-the-shelf components, and minimizes dead volume and cross-contamination because only the pipette tip contacts the sample. Although the current prototype still exhibits device-to-device variability and moving droplets in open containers complicate real-time imaging, these results indicate that toothbrush-based vibrational actuation can provide a practical and scalable route toward “lab-in-hand” droplet assays in resource-limited or educational settings. Full article

The tasks of community detection in complex networks have garnered increasing attention from researchers. Concurrently, with the emergence of graph neural networks (GNNs), these models have rapidly become the mainstream approach for solving this task. However, GNNs frequently encounter the Laplacian oversmoothing problem, which dilutes the crucial neighborhood signals essential for community identification. These signals, particularly those from first-order neighbors, are the core source information defining community structure and identity. To address this contradiction, this paper proposes a novel training strategy focused on strengthening these key local signals. We design a multi-branch learning structure that injects a gradient into the GNN layer during backpropagation. This gradient is then modulated by the GNN’s native message-passing path, precisely supplementing the parameters of the initial layers with first-order topological information. Based on this, we construct the network structure-informed GNN (NIGNN). A large number of experiments show that the proposed method achieves a 0.6–3.6% improvement in multiple indicators compared with the basic model in the community detection task, and performs well in the t-test. The framework has good general applicability and can be effectively applied to GCN, GAT, and GraphSAGE architectures, and shows strong robustness in networks with incomplete information. This work offers a novel solution for effectively preserving core local information in deep GNNs. Full article

Net-shaped casting processes in the automotive industry have proved to be difficult to simulate due to the complexities of the interactions amongst thermal, fluid, and solute transport regimes in the solidifying domain, along with the interface. The existing casting simulation software lacks the necessary real-time estimation of thermophysical properties (thermal diffusivity and thermal conductivity) and the interfacial heat-transfer coefficient (IHTC) to evaluate the thermal resistances in a casting process and solve the temperature in the solidifying domain. To address these shortcomings, an axial directional solidification experiment setup was developed to map the thermal data as the melt solidifies unidirectionally from the chill surface under unsteady-state conditions. A Dilute Eutectic Cast Aluminum (DECA) alloy, Al-5Zn-1Mg-1.2Fe-0.07Ti, Eutectic Cast Aluminum (ECA) alloys (A365 and A383), and pure Al (P0303) were used to demonstrate the validity of the experiments to evaluate the thermal diffusivity (α) of both the solid and liquid phases of the solidifying metal using an inverse heat-transfer analysis (IHTA). The thermal diffusivity varied from 0.2 to 1.9 cm²/s while the IHTC changed from 9500 to 200 W/m²K for different alloys in the solid and liquid phases. The heat flux was estimated from the chill side with transient temperature distributions estimated from IHTA for either side of the mold–metal interface as an input to compute the interfacial heat-transfer coefficient (IHTC). The results demonstrate the reliability of the axial solidification experiment apparatus in accurately providing input to the casting simulation software and aid in reproducing casting numerical simulation models efficiently. Full article

Over the last decade, numerical simulations and experiments have confirmed the existence of a novel class of vibrationally excited solid-particle attractors in cubic cavities containing a fluid in non-isothermal conditions. The diversity of emerging particle structures, in both morphology and multiplicity, depends strongly on the uni- or multi-directional nature of the imposed temperature gradients. The present study seeks to broaden this theoretical framework by further increasing the complexity of the thermal “information” coded along the external boundary of the fluid container. In particular, in place of the thermal inhomogeneities located in the center of otherwise uniformly cooled or heated walls, here, a cubic cavity with temperature boundary conditions satisfying the D₂h (in Schoenflies notation) or “mmm” (in Hermann–Mauguin notation) symmetry is considered. This configuration, equivalent to a bipartite vertex coloring of a cube leading to a total of 24 thermally controlled planar surfaces, possesses three mutually perpendicular twofold rotation axes and inversion symmetry through the cube’s center. To reduce the problem complexity by suppressing potential asymmetries due to fluid-dynamic instabilities of inertial nature, the numerical analysis is carried out under the assumption of dilute particle suspension and one-way solid–liquid phase coupling. The results show that a kaleidoscope of new particle structures is enabled, whose main distinguishing mark is the essentially one-dimensional (filamentary) nature. These show up as physically disjoint or intertwined particle circuits in striking contrast to the single-curvature or double-curvature spatially extended accumulation surfaces reported in earlier investigations. Full article

The decomposition of silicates in sulfuric acid to extract rare earth elements (REE) is typically characterized by the formation of an amorphous silica layer surrounding the receding crystal that may act as a passivation barrier limiting the rate of mineral dissolution. In this context, sulfuric acid treatment experiments coupled with detailed characterization of the evolution of the decomposition reaction were performed on natural allanite (CaREEAl₂Fe²⁺Si₃O₁₁O[OH]), as well as synthetic neodymium disilicate (Nd₂Si₂O₇), orthosilicate (Ca₂Nd₈(SiO₄)₆O₂), and orthophosphate (NdPO₄) phases in order to investigate if there are key factors, operating on a wide range of silicates, that negatively impact REE recovery. While, as expected, the acid strength is the driver in promoting the decomposition of the orthophosphate, for the silicates investigated, no matter their crystalline structure and chemical resistance, there is a severe passivation mechanism at play in concentrated H₂SO₄. However, in all cases, this effect can be minimized by water dilution, which strongly enhances sulfate-forming cation transfer across the produced amorphous silica layer. Taking into consideration this distinct characteristic of the mode of decomposition of silicates in sulfuric acid should help in defining optimal extraction strategies. Full article

The non-uniformity of the inlet velocity profile (referred to as inlet distortion) in a gas turbine combustor critically influences the outlet temperature distribution, which is a key factor for the operational safety and durability of the turbine blades. To investigate the influence of inlet velocity distortion on the outlet temperature distribution factor (OTDF) and the hot streak evolution in a combustor, scaled-adaptive simulations (SAS) and experiments were conducted at an inlet temperature of 400 K, an inlet total pressure of 0.20 MPa, and a fuel–air ratio (FAR) of 0.018. RP-3 aviation kerosene was used as fuel for this investigation. The results show that in the primary zone, the heat release rate is quite low in the counter-current region, while it is very high in the co-current region. In the area downstream of the primary zone, intense heat release mainly takes place near the primary and dilution jets. The substantial penetration of the jets results in a relatively low FAR at the mid-height part of the liner, while the FAR is relatively high near the wall leading to the formation of hot streaks. Critically, experimental data demonstrate that the defined inlet distortions substantially increase the OTDF by 40 percentage points (from approximately 10% to 50%), highlighting a significant challenge for combustor design. This work provides validated insight into the linkage between inflow distortions and critical thermal loads, which is essential for developing more robust combustion systems. Full article