Search Results (160)

Background/Objectives: Inter-platform variability in beam characteristics and low-dose exposure may arise from differences in linear accelerator head design, multileaf collimator geometry, and dose calculation algorithms. This study aimed to evaluate system-level dosimetric differences between two widely used linear accelerator platforms under clinically commissioned conditions. Methods: A comparative dosimetric analysis was performed between Elekta Synergy and Varian TrueBeam linear accelerators. Beam data were acquired using a SunSCAN™ 3D water phantom, and patient-specific quality assurance was conducted with the Octavius^® 4D system. Treatment plans were generated for left-sided breast, prostate, and head and neck cases using clinically commissioned treatment planning systems. Beam flatness, symmetry, penumbra width, low-dose exposure, conformity, homogeneity, and organ-at-risk dose metrics were evaluated. Results: Platform-dependent differences were observed in penumbra behavior and out-of-field dose, primarily attributable to intrinsic linac head design and collimation characteristics. These differences propagated into clinical plans, with greater variability observed for breast and head and neck cases, while prostate plans showed higher consistency between platforms. Algorithm-dependent trends were noted for conformity and homogeneity indices; however, all plans met institutional clinical acceptance criteria during quality assurance. Stricter gamma evaluation criteria revealed systematic but limited inter-platform deviations. Conclusions: Elekta Synergy and Varian TrueBeam demonstrated clinically acceptable dosimetric performance, with modest platform-dependent differences. While target coverage and overall plan quality were comparable, these variations were primarily observed in peripheral dose regions and may be relevant for platform-specific planning optimization and quality assurance. This supports the importance of comprehensive commissioning and QA procedures in both mixed- and single-platform clinical settings, particularly for highly modulated techniques. Full article

To examine how grain-size distribution affects the mechanical response and fracture behavior of Lac du Bonnet (LdB) granite under uniaxial compression, numerical simulations were conducted using the particle flow code (PFC) with a grain-based model. By displacing grain centroids in different directions along the y-axis, four LdB granite models with distinct grain sizes were generated, with grains delineated by Voronoi tessellation. The main findings are as follows: (1) The flat-jointed constitutive model reproduces the experimental response well, and introducing unbonded contacts (micrometer-scale gaps) improves the simulation of crack-closure behavior during loading. (2) Secondary cracks initiate predominantly at grain boundaries, and the yield stress is strongly associated with the evolution of intragranular tensile cracks. (3) Grain size governs the sequence of crack accumulation (tensile vs. shear), the growth rate and spatial correlation of damage, and the distribution and intensity of local failures; smaller grains hinder macroscopic damage, whereas larger grains are more readily penetrated and filled by microcracks. (4) Mechanical cutting tests show that grain-size combinations produce several dominant secondary-failure modes; the failure thickness is controlled by the penetration depth of the subsequent cutting head, and the stress concentration near the cutting head is sensitive to grain size. Full article

Battery energy storage systems (BESSs) play a vital role in contemporary smart grids, but their increasing digitalisation exposes them to sophisticated cyberattacks. Existing anomaly detection approaches typically treat sensor measurements as flat feature vectors, overlooking the intrinsic relational structure of cyber–physical systems. This work introduces an enhanced Graph Neural Network (GNN) autoencoder for unsupervised BESS anomaly detection that integrates multiscale graph construction, multi-head graph attention, manifold regularisation via latent compactness and graph smoothness, contrastive embedding shaping, and an ensemble anomaly scoring mechanism. A comprehensive evaluation across seven BESS and firmware cyberattack datasets demonstrates that the proposed method achieves near-perfect Receiver Operating Characteristic (ROC) and Precision–Recall Area Under the Curve (PR AUC) (up to 1.00 on several datasets), outperforming classical one-class models such as Isolation Forest, One-Class Support Vector Machine (One-Class SVM), and Local Outlier Factor on the most challenging scenarios. These results illustrate the strong potential of graph-informed representation learning for cybersecurity monitoring in distributed energy resource infrastructures. Full article

Accurate grading of corn kernels is critical for seed certification, directional seeding, and breeding, yet it is still predominantly performed by manual inspection. This work introduces CornViT, a three-stage Convolutional Vision Transformer (CvT) framework that emulates the hierarchical reasoning of human seed analysts for single-kernel evaluation. Three sequential CvT-13 classifiers operate on $384\times 384$ RGB images: Stage 1 distinguishes pure from impure kernels; Stage 2 categorizes pure kernels into flat and round morphologies; and Stage 3 determines the embryo orientation (up vs. down) for pure, flat kernels. Starting from a public corn seed image collection, we manually relabeled and filtered images to construct three stage-specific datasets: 7265 kernels for purity, 3859 pure kernels for morphology, and 1960 pure–flat kernels for embryo orientation, all released as benchmarks. Head-only fine-tuning of ImageNet-22k pretrained CvT-13 backbones yields test accuracies of 93.76% for purity, 94.11% for shape, and 91.12% for embryo-orientation detection. Under identical training conditions, ResNet-50 reaches only 76.56 to 81.02 percent, whereas DenseNet-121 attains 86.56 to 89.38 percent accuracy. These results highlight the advantages of convolution-augmented self-attention for kernel analysis. To facilitate adoption, we deploy CornViT in a Flask-based web application that performs stage-wise inference and exposes interpretable outputs through a browser interface. Together, the CornViT framework, curated datasets, and web application provide a deployable solution for automated corn kernel quality assessment in seed quality workflows. Source code and data are publicly available. Full article

Although wavelet neural networks (WNNs) combine the expressive capability of neural models with multiscale localization, there are currently few theoretical guarantees for their training. We investigate the weight decay (${\mathrm{L}}_2$ regularization) optimization dynamics of gradient descent (GD) for WNNs. Using explicit rates controlled by the spectrum of the regularized Gram matrix, we first demonstrate global linear convergence to the unique ridge solution for the feature regime when wavelet atoms are fixed and only the linear head is trained. Second, for fully trainable WNNs, we demonstrate linear rates in regions satisfying a Polyak–Łojasiewicz (PL) inequality and establish convergence of GD to stationary locations under standard smoothness and boundedness of wavelet parameters; weight decay enlarges these regions by suppressing flat directions. Third, we characterize the implicit bias in the over-parameterized neural tangent kernel (NTK) regime: GD converges to the minimum reproducing kernel Hilbert space (RKHS) norm interpolant associated with the WNN kernel with ${\mathrm{L}}_2$. In addition to an assessment process on synthetic regression, denoising, and ablations across λ and stepsize, we supplement the theory with useful recommendations on initialization, stepsize schedules, and regularization scales. Together, our findings give a principled prescription for dependable training that has broad applicability to signal processing applications and shed light on when and why ${\mathrm{L}}_2$-regularized GD is stable and quick for WNNs. Full article

Objective: This in vitro study aimed to evaluate and compare the trueness and precision of four extraoral 3D facial scanning systems using a standardized 3D-printed human head model. Methods: A 3D-printed head model with 16 anatomical landmarks and 17 inter-landmark linear distances was fabricated using a high-resolution 3D printer. Caliper measurements were used as reference standards. The model was scanned 15 times by four systems: a handheld scanner (MetiSmile, Shining 3D, Hangzhou, China), a desktop scanner (RAYFace v2.0, Ray Co., Seongnam, Gyeonggi-do, Republic of Korea), and two mobile applications (Heges and Polycam, iPhone 15, Apple Inc., Cupertino, CA, USA). All digital distances were measured in Blender software. To assess intra-observer reliability, all measurements were repeated twice by the same examiner with a 3-week interval between sessions, and intra-class correlation coefficients were calculated using a two-way mixed-effects, single-measurement, absolute-agreement model (ICC 3,1). Trueness, defined as the absolute deviation from the reference caliper values, was compared across scanners using the Kruskal–Wallis test due to its non-normal distribution. Precision, regional trueness and precision values across the four scanners defined as the standard deviation of repeated scans, was analyzed using One-way ANOVA with Tukey post-hoc comparisons for normally distributed datasets (α = 0.05). Distances were measured digitally in Blender software, and trueness (absolute deviation from reference) and precision (standard deviation of repeated scans) were analyzed using the Kruskal–Wallis test and One-way ANOVA with Tukey post hoc comparisons (α = 0.05). Results: The Polycam application demonstrated the highest trueness (0.49 ± 0.32 mm), followed by MetiSmile (0.51 ± 0.36 mm), RAYFace (0.58 ± 0.39 mm), and Heges (0.73 ± 0.42 mm). The MetiSmile scanner showed the highest precision (0.12 ± 0.07 mm), while RAYFace and Polycam exhibited moderate precision (0.28 ± 0.19 mm and 0.15 ± 0.06 mm, respectively). Vertical measurements tended to be more accurate than horizontal ones, and the lower facial region showed smaller deviations; however, these differences were not statistically significant (p > 0.05). Conclusions: MetiSmile achieved the highest precision and Polycam the highest trueness. Although all systems showed mean deviations < 1 mm, only three demonstrated <0.6 mm accuracy (except for Heges scanner). These results suggest that professional and mobile-based scanners can provide clinically acceptable facial data for educational and preliminary digital workflow applications, though further validation under clinical conditions is required. This study provides quantitative evidence on the accuracy and repeatability of commonly available extraoral 3D facial scanning systems under controlled laboratory conditions. The results indicate that both professional-grade and mobile-based scanners can reproduce facial morphology with clinically acceptable deviations, particularly in flat and stable regions such as the forehead and chin. Although only three systems achieved mean trueness below 0.6 mm, all demonstrated errors within 1 mm, sufficient for diagnostic visualization, digital smile design, and preliminary virtual patient modeling. These findings support the safe and cost-effective adoption of extraoral facial scanning in dental education and treatment planning, while emphasizing the need for further validation in real clinical environments where motion, lighting, and soft-tissue variability may affect accuracy. Full article

In Global Navigation Satellite System (GNSS)-denied environments, reconstructing three dimensional trajectories using only an Inertial Measurement Unit faces challenges such as heading drift, stride error accumulation, and gait recognition uncertainty. This paper proposes a path estimation method with a nine-axis inertial sensor that continuously and accurately estimates an agent’s path without external support. The method detects stationary states and halts updates to suppress error propagation. During motion, gait modes including flat walking, stair ascent, and stair descent are classified using vertical acceleration with dynamic thresholds. Vertical displacement is estimated by combining gait pattern and posture angle during stair traversal, while planar displacement is updated through adaptive stride length adjustment based on gait cycle and movement magnitude. Heading is derived from the attitude matrix aligned with magnetic north, enabling projection of displacements onto a unified frame. Experiments show planar errors below three percent for one-hundred-meter paths and vertical errors under two percent in stair environments up to ten stories, with stable heading maintained. Overall, the method achieves reliable gait recognition and continuous three-dimensional trajectory reconstruction with low computational cost, using only a single inertial sensor and no additional devices. Full article

In-vehicle voice assistants face usability challenges due to limitations in delivering feedback within the constraints of the driving environment. The presented study explores the potential of Rich Visual Feedback (RVF) on Head-Up Displays (HUDs) as a multimodal solution to enhance system usability. A user study with 32 participants evaluated three HUD User Interface (UI) designs: the AR Fusion UI, which integrates augmented reality elements for layered, dynamic information presentation; the Baseline UI, which displays only essential keywords; and the Flat Fusion UI, which uses conventional vertical scrolling. To explore HUD interface principles and inform future HUD design without relying on specific hardware, a simulated near-field overlay was used. Usability was measured using the System Usability Scale (SUS), and distraction was assessed with a penalty point method. Results show that RVF on the HUD significantly influences usability, with both content quantity and presentation style affecting outcomes. The minimal Baseline UI achieved the highest overall usability. However, among the two Fusion designs, the AR-based layered information mechanism outperformed the flat scrolling method. Distraction effects were not statistically significant, indicating the need for further research. These findings suggest RVF-enabled HUDs can enhance in-vehicle voice assistant usability, potentially contributing to safer, more efficient driving. Full article

To address the problems of low inversion accuracy and poor noise resistance in pulsed eddy current (PEC) grounding grid depth detection, this study proposes a novel inversion model (IE-CBiLSTM). This model integrates the Informer Encoder with the CNN-BiLSTM for the first time to detect the depth of the PEC grounding grid and conducts experimental verification based on an independently designed pulsed eddy current detection device and a dedicated coil sensor. The model design employs a two-dimensional convolutional neural network (CNN) to extract local spatial features, combines a bidirectional long short-term memory network (Bi-LSTM) to model temporal dependencies, and introduces a multi-head attention mechanism along with the Informer structure to enhance the expression of key features. In terms of data construction, the design integrates both forward simulation data and measured data to improve the model’s generalization capability. Experimental validation includes self-burial experiments and field tests at a substation. In the self-burial test, the IE-CBiLSTM inversion results show high consistency with actual burial depths under various conditions (1.0 m, 1.2 m, and 1.5 m), significantly outperforming other optimization algorithms, achieving a coefficient of determination (R²) of 0.861, along with root mean square error (E_RMS) and mean relative error (E_MR) values of 17.54 Ω·m and 0.061 Ω·m, respectively. In the field test, the inversion results also closely match the design depths from engineering drawings, with an R² of 0.933, E_RMS of 11.30 Ω·m, and E_MR of 0.046 Ω·m. These results are significantly better than those obtained using traditional Occam and LSTM methods. At the same time, based on the inversion results, a three-dimensional inversion map of the grounding grid and a buried depth profile were drawn, and the spatial direction and buried depth distribution of the underground flat steel were clearly displayed, proving the visualization ability of the model and its engineering practicality under complex working conditions. This method provides an efficient and reliable inversion strategy for deep PEC nondestructive testing of grounding grid laying. Full article

The extraordinary mechanical flexibility, high electrical conductivity, and nanoscale instability of freestanding graphene make it an excellent candidate for vibration energy harvesting. When freestanding graphene is stretched taut and subject to external forces, it will vibrate like a drum head. Its vibrations occur at a fundamental frequency along with higher-order harmonics. Alternatively, when freestanding graphene is compressed, it will arch slightly out of the plane or buckle under the load. Remaining flat under compression would be energetically too costly compared to simple bond rotations. Buckling up or down, also known as ripple formation, naturally creates a bistable situation. When the compressed system vibrates between its two low-energy states, it must pass through the high-energy middle. The greater the compression, the higher the energy barrier. The system can still oscillate but the frequency will drop far below the fundamental drum-head frequency. The low frequencies combined with the large-scale movement and the large number of atoms coherently moving are key factors addressed in this study. Ten ripples with increasing compressive strain were built, and each was studied at five different temperatures. Increasing the temperature has a similar effect as increasing the compressive strain. Analysis of the average time between curvature inversion events allowed us to quantify the energy barrier height. When the low-frequency bistable data were time-averaged, the authors found that the velocity distribution shifts from the expected Gaussian to a heavy-tailed Cauchy (Lorentzian) distribution, which is important for energy harvesting applications. Full article

The continuous advancement of fire protection technologies has necessitated the development of comprehensive safety standards, leading to an increasingly diversified and specialized regulatory landscape. This has made it difficult for fire protection professionals to quickly and accurately locate the required fire safety standard information. In addition, the lack of effective integration and knowledge organization concerning fire safety standard entities has led to the severe fragmentation of fire safety standard information and the absence of a comprehensive “one map”. To address this challenge, we introduce FT-FLAT, an innovative CNN–Transformer fusion architecture designed specifically for fire safety standard entity extraction. Unlike traditional methods that rely on rules or single-modality deep learning, our approach integrates TextCNN for local feature extraction and combines it with the Flat-Lattice Transformer for global dependency modeling. The key innovations include the following. (1) Relative Position Embedding (RPE) dynamically encodes the positional relationships between spans in fire safety texts, addressing the limitations of absolute positional encoding in hierarchical structures. (2) The Multi-Branch Prediction Head (MBPH) aggregates the outputs of TextCNN and the Transformer using Einstein summation, enhancing the feature learning capabilities and improving the robustness for domain-specific terminology. (3) Experiments conducted on the newly annotated Fire Safety Standard Entity Recognition Dataset (FSSERD) demonstrate state-of-the-art performance (94.24% accuracy, 83.20% precision). This work provides a scalable solution for constructing fire safety knowledge graphs and supports intelligent information retrieval in emergency situations. Full article

As dogs are often required to be leashed in public, guardians need to choose between various restraint devices. While using collars typically considered aversive (such as choke, prong, or electric collars) is generally discouraged due to welfare concerns, guidance is less clear when it comes to selecting among other devices such as collars, harnesses, and head collars. This review examined 21 full-text articles and two abstracts on the effects of commonly used restraint devices, aiming to offer practical guidance for guardians and identifying areas for future research. The impact of these devices was examined in terms of walking kinematics, pressure distribution on the body, and behavioural signs of stress. The findings suggest there is no one-size-fits-all device, and selection should consider the individual needs of guardian and dog. For dogs that pull, non-tightening front-clip harnesses appear to offer the best balance between discomfort and reduction in pulling. Tightening harnesses, martingale collars, and head collars can pose greater discomfort and should be used with caution. For brachycephalic breeds or when pulling is not a concern, back-clip harnesses are suitable, especially chest-strap or Y-shaped ones. Flat collars are also appropriate for dogs that do not pull as they produce the least body restriction. Full article

As a crucial characteristic waterfowl breed, the egg-laying performance of Lion-Headed Geese serves as a core indicator for precision breeding. Under large-scale flat rearing and selection practices, high phenotypic similarity among individuals within the same pedigree coupled with traditional manual observation and existing automation systems relying on fixed nesting boxes or RFID tags has posed challenges in achieving accurate goose–egg matching in dynamic environments, leading to inefficient individual selection. To address this, this study proposes YOLO-Goose, an improved YOLOv8s-based method, which designs five high-contrast neck rings (DoubleBar, Circle, Dot, Fence, Cylindrical) as individual identifiers. The method constructs a lightweight model with a small-object detection layer, integrates the GhostNet backbone to reduce parameter count by 67.2%, and employs the GIoU loss function to optimize neck ring localization accuracy. Experimental results show that the model achieves an F1 score of 93.8% and mAP50 of 96.4% on the self-built dataset, representing increases of 10.1% and 5% compared to the original YOLOv8s, with a 27.1% reduction in computational load. The dynamic matching algorithm, incorporating spatiotemporal trajectories and egg positional data, achieves a 95% matching rate, a 94.7% matching accuracy, and a 5.3% mismatching rate. Through lightweight deployment using TensorRT, the inference speed is enhanced by 1.4 times compared to PyTorch-1.12.1, with detection results uploaded to a cloud database in real time. This solution overcomes the technical bottleneck of individual selection in flat rearing environments, providing an innovative computer-vision-based approach for precision breeding of pedigree Lion-Headed Geese and offering significant engineering value for advancing intelligent waterfowl breeding. Full article

This study presents a comprehensive numerical investigation into the influence of blade profile geometry on the internal flow dynamics and hydraulic performance of Cross-Flow Turbines (CFTs) under varying runner speeds. Four blade configurations, flat, round, sharp, and aerodynamic, were systematically evaluated using steady-state, two-dimensional Computational Fluid Dynamics (CFD) simulations. The Shear Stress Transport (SST) k–ω turbulence model was employed to resolve the flow separation, recirculation, and turbulence across both energy conversion stages of the turbine. The simulations were performed across runner speeds ranging from 270 to 940 rpm under a constant head of 10 m. The performance metrics, including the torque, hydraulic efficiency, water volume fraction, pressure distribution, and velocity field characteristics, were analyzed in detail. The aerodynamic blade consistently outperformed the other geometries, achieving a peak efficiency of 83.5% at 800 rpm, with improved flow attachment, reduced vortex shedding, and lower exit pressure. Sharp blades also demonstrated competitive efficiency within a narrower optimal speed range. In contrast, the flat and round blades exhibited higher turbulence and recirculation, particularly at off-optimal speeds. The results underscore the pivotal role of blade edge geometry in enhancing energy recovery, suppressing flow instabilities, and optimizing the stage-wise performance in CFTs. These findings offer valuable insights for the design of high-efficiency, site-adapted turbines suitable for micro-hydropower applications. Full article

The aerodynamic loading of four as well as of six tilted flat plates-panels arranged in tandem and in close proximity to the ground is examined through force and pressure measurements. In the four-plate set up, conducted in an open-circuit wind tunnel, a movable floor is used to vary the ground clearance, and a one-component force balance is employed to measure the drag coefficient Cd of each plate for tilt angles 10° to 90° and for two head-on wind directions, 0° and 180°. An increase in the ground clearance from 20% to 60% of the plates’ chord length, results in a Cd increase of over 40% in the downstream plates, and up to 20% in the leading one. For tilt angles below 40°, the drag on the first plate is up to 25% higher under the 180° wind direction compared to the opposite direction. Pressure distributions are also presented on a series of six much larger plates, examined in a closed-circuit wind tunnel at tilt angles ±30°. While the windward surfaces exhibit relatively uniform pressure distributions, regions of low pressure develop on their suction side, near the plates’ tips leading edge, tending to become uniform streamwise. Full article