Search Results (315)

Many methods have been proposed for visualizing and interpreting the results of artificial intelligence (AI) algorithms. AI explainability (XAI) methods vary in mathematical basis, effectiveness, and scope of application. Knowing this, an important question arises: how do their results differ from a statistical point of view, and are some of them more useful than the others in certain scenarios? Our article aims to assess the robustness of the most popular AI models’ explainability visualization methods and to identify differences in the results obtained. We did this by analyzing fundamental convolutional neural network models that classified 598 cat images from the Oxford III-T Pet database and 580 filtered pictures of Boeing planes from the Aircraft Images Dataset. We performed a comparative analysis of the similarities between methods based on Class Activation Mapping (CAM), gradients, and Local Interpretable Model-agnostic Explanations (LIME). To evaluate them, we used Pearson Correlation Coefficient (CC), Matthews Correlation Coefficient (MCC), Spearman’s Rank, Structural Similarity Index Measure (SSIM), Kullback–Leibler divergence, Intersection over Union (IoU), and Soft IoU. To check the fidelity and robustness of the XAI methods, we used RandomCAM and ran an ablation test, checking for a decrease in prediction confidence as we gradually removed the least significant regions. Our results provide an up-to-date and broad comparative analysis of this field. They can serve as a reference point for machine learning scientists and engineers. Full article

Background/Objectives: Dynamic navigation and robot-assisted implant workflows depend on robust intraoral perception. Marker-based tracking introduces workflow complexity and is sensitive to occlusions, motivating markerless alternatives. This study evaluates whether a single-stage YOLO instance segmentation model (YOLO-seg) can provide a practical markerless perception layer for dental navigation, combining accurate per-tooth delineation with low, predictable inference latency. Methods: YOLO-seg was trained end to end on an intraoral RGB corpus of 400 training, 20 validation, and 100 testing images, combining a public source and a partner-hospital in-house set. A two-stage YOLO + SAM baseline was implemented for comparison. Segmentation quality was evaluated on a 50-image held-out clinical test set at three complementary levels (per-instance matching, per-class union, and global union), with paired Wilcoxon signed-rank tests, Cliff’s delta effect sizes, and 95% bootstrap confidence intervals. Runtime was assessed under matched inference-only and end-to-end conditions on N = 100 frames at a 640 × 640 resolution on an NVIDIA RTX A2000 GPU. Results: YOLO-seg significantly outperformed YOLO + SAM across all primary metrics, with very large effect sizes (Cliff’s delta: 0.76–0.94; Wilcoxon p < 10⁻⁸ on every metric except precision at IoU ≥ 0.5). YOLO-seg reached AP₅₀ = 0.716 and recall = 0.973 versus 0.383 and 0.398 for YOLO + SAM. Under matched inference-only timing, YOLO-seg ran at 27.08 ms per frame (36.9 FPS) versus 1302.78 ms (0.77 FPS), an approximately 48-fold latency gap intrinsic to the two-stage forward pass. Conclusions: YOLO-seg shows strong potential as a 2D perception module for dental navigation, balancing per-instance segmentation fidelity with real-time feasibility under the tested conditions. These results support its use as a 2D perception front-end for future integration with stereo-based 3D reconstruction and robot-assisted navigation; 3D registration accuracy, implant-placement error, and robotic execution remain outside the scope of the present study. Full article

Tomato leaf diseases substantially reduce tomato yields and quality and remain a persistent challenge for efficient crop management. Although deep learning-based detectors have achieved strong accuracy in controlled benchmarks, many existing solutions are still difficult to transfer to resource-constrained agricultural systems because they rely on high-end GPUs, consume considerable power, and often lose performance after deployment on embedded devices. To address this practical gap, this study proposes HGS-YOLO, a system-oriented deployable lightweight adaptation of YOLOv11 for leaf-level tomato disease detection, together with an end-to-end edge sensing pipeline for low-power agricultural deployment. The main contribution lies in the coordinated system-level co-design of model structure, optimization, and deployment rather than in a novel detector architecture. Specifically, YOLOv11 is adapted through three coordinated modifications: an HGNetV2 backbone for efficient feature extraction, an HS-FPN neck with channel attention for lightweight multi-scale fusion, and an MPDIoU loss function for more stable localization optimization. Beyond the model architecture, the study establishes a complete engineering pipeline that includes training, optimization, post-training quantization, and hardware deployment with BPU acceleration on a D-Robotics RDK X5 handheld platform. Comprehensive benchmark experiments indicate that HGS-YOLO achieves 93.6% ${\mathrm{mAP}}_{50}$ and 72.1% mAP@[0.5:0.95] with 86.5% recall, only 1.3 M parameters, and a 3.1 MB model size, substantially reducing the model complexity and storage cost relative to the YOLOv11 baseline. A three-seed retraining comparison shows that HGS-YOLO trades roughly 0.5 ${\mathrm{mAP}}_{50}$ points for this compactness (a statistically significant but small concession) and recovers the cost on the deployment side: on the RDK X5 chip, HGS-YOLO is the fastest, most memory-efficient, and lowest-power model among all compared detectors. Indoor deployment tests using separately collected tomato leaf samples further achieve 90.3% ${\mathrm{mAP}}_{50}$, 82.3% recall, 89.0% precision, 25.0 ± 0.4 ms end-to-end latency, 40.0 ± 0.6 FPS, and 9.8 ± 0.4 W average system power. After PTQ, the ${\mathrm{mAP}}_{50}$ drops from 93.6% to 93.0% on the same benchmark; because this figure was measured under controlled imaging conditions, it is presented as an in-distribution reference point rather than as evidence of robustness in the open field. We also took the handheld system into a working tomato greenhouse for a small outdoor field round, where it ran end-to-end and produced on-device disease detections under natural sunlight, specular highlights, partial occlusion, background clutter, and handheld motion blur. These results show that HGS-YOLO reaches a good balance of accuracy, efficiency, and deployability and that it works in the field on an independent small-scale test; validating it more widely across sites, seasons, and weather is left to future work. Full article

Vertical axis wind turbines (VAWTs) have regained attention for distributed, urban, and floating offshore applications, yet the literature remains fragmented across competing rotor concepts and modelling traditions. This review consolidates the principal archetypes—Savonius, H-Darrieus, troposkein Darrieus, helical Darrieus, and Savonius–Darrieus hybrids—through five governing parameters: drag-versus-lift-driven operating principle, tip speed ratio $\mathsf{\lambda }=\mathsf{\omega }\mathrm{R}/{\mathrm{V}}_{\infty }$ (0.6–1.2 for Savonius; 2.5–5.0 for Darrieus), solidity $\mathsf{\sigma }=\mathrm{N}\mathrm{c}/\mathrm{R}$ (0.1–0.4), chord-based Reynolds number ${\mathrm{R}}_{\mathrm{e}\_\mathrm{c}}$ (${10}^5–{10}^6$), and peak power coefficient ${\mathrm{C}}_{\mathrm{p}\_\mathrm{m}\mathrm{a}\mathrm{x}}$ (0.15–0.25 for Savonius; 0.35–0.45 for optimized H-Darrieus). Off-design performance is dominated by unsteady mechanisms that quasi-steady streamtube models cannot resolve—leading edge vortex shedding, dynamic stall hysteresis, blade–wake interaction, and flow-curvature-induced virtual camber—each examined for its contribution to the instantaneous torque ${\mathrm{C}}_{\mathrm{T}}\left(\mathsf{\theta }\right)$ and the cycle-averaged ${\mathrm{C}}_{\mathrm{p}}$. Turbulence closures are benchmarked against phase-locked PIV and torque measurements: $\mathrm{k}–\mathsf{\omega }\mathrm{S}\mathrm{S}\mathrm{T}$ URANS captures peak-region ${\mathrm{C}}_{\mathrm{p}}$ to within $\pm 5–10\%$ but over-predicts torque below ${\mathsf{\lambda }}_{\mathrm{o}\mathrm{p}\mathrm{t}}$; the $\mathsf{\gamma }–{\mathrm{R}}_{\mathrm{e}\_\mathsf{\theta }}$ transition SST model reduces this error to $\pm 3–5\%$; DES, DDES, and LES reach $\pm 2–3\%$ at one to two orders of magnitude higher cost. Best practice computational fluid dynamics (CFD) guidelines are consolidated: domain extents of 15D upstream, 10D downstream, and 20D lateral; rotating sub-domain ${\mathrm{D}}_{\mathrm{r}\mathrm{o}\mathrm{t}}$ $\approx 1.5\mathrm{D}$; ${\mathrm{y}}^{+}\le 1$; $\Delta \mathsf{\theta }\le 0.1°$; and 20–30 revolutions before sampling. Performance enhancement strategies (variable pitch, guide vanes, helical twist, and hybridization) are reviewed quantitatively, with reported ${\mathrm{C}}_{\mathrm{p}}$ gains of $5–30\%$. Four research priorities are identified: (i) transition-sensitive turbulence closures validated below ${\mathrm{R}}_{\mathrm{e}\_\mathrm{c}}$ = $5\times {10}^5$; (ii) coupled aero-hydro-servo-elastic models for floating offshore VAWTs; (iii) machine-learning-augmented turbulence modelling—including physics-informed neural networks (PINNs) and neural-network-corrected RANS closures—to improve unsteady flow prediction at sub-LES cost; and (iv) integrated aeroacoustic–aeroelastic frameworks for urban and building-integrated deployment. Full article

This study investigates whether aerodynamic interaction effects in an interlocked (negative-gap) counter-rotating dual Savonius rotor configuration can improve the efficiency of drag-based vertical-axis wind turbines in urban wind conditions. Two-dimensional Computational Fluid Dynamics (CFD) simulations were performed in ANSYS Fluent 2025 R2 using both steady and unsteady RANS approaches, including dynamic meshing to enable collision-free rotation in the interlocked overlap region. The numerical setup was first validated for a single two-bucket reference rotor against published experimental data of torque and power coefficients and subsequently applied to dual-rotor configurations with negative gap distances. The results show that the dual-rotor arrangement redistributes torque production over the azimuth angle and yields a smoother and consistently positive mean static torque coefficient, indicating improved self-starting behavior compared to the single rotor. Under transient operation, the dual-rotor configuration yields higher power coefficient values across the entire investigated tip-speed ratio range. The highest performance gain is observed at a tip-speed ratio of $\lambda \approx 1.0$, where the peak power coefficient increases from $c_p\approx 0.25$ (single-rotor) to $c_p\approx 0.32$ (dual-rotor), corresponding to an improvement of the power coefficient of about $\mathsf{\Delta }{c}_p/c_{p0}\approx 28\%$. Full article

Hydrogen enrichment of compression ignition (CI) engines has emerged as a promising strategy to simultaneously enhance thermal efficiency and reduce carbon-based emissions. This study numerically investigates how hydrogen enrichment affects engine performance and emissions in methanol–diesel dual-fuel CI engines, a combustion mode gaining increasing attention for replacing fossil diesel with sustainable fuels, particularly in hard-to-abate sectors such as maritime transport. The simulations are based on the Unsteady Reynolds-Averaged Navier–Stokes (URANS) equations, incorporating the RNG k-ε turbulence model, the Eddy Dissipation Concept (EDC) for turbulence–chemistry interaction, and the G-equation for turbulent premixed flame propagation. The numerical model is validated against experimental data for in-cylinder pressure and heat release rate at 45% methanol substitution ratio (by energy). The results indicate that increasing the hydrogen enrichment ratio (HER, defined on an energy basis) from 5% to 20% raises the Sauter mean diameter (SMD) of the diesel fuel from 20.2 µm to 28.0 µm (+38%), driven by reduced aerodynamic breakup intensity associated with modified gas-phase properties under hydrogen enrichment. Furthermore, hydrogen’s elevated adiabatic flame temperature and superior mass diffusivity intensify combustion, raising peak in-cylinder pressure from 75.2 to 79.1 bar (+5.2%), amplifying the peak heat release rate from 129 to 211 J/°CA (+63.6%), and elevating maximum in-cylinder temperature from 1542 to 1735 K (+193 K). Under the investigated CFD operating conditions, these thermodynamic gains translate into an engine-level 6% improvement in indicated thermal efficiency and a 14% reduction in indicated specific fuel consumption (accounting for hydrogen, methanol, and diesel) at HER 20%. On the emissions front, CO₂ declines by 24% in direct proportion to the carbon-containing fuel mass displaced by hydrogen substitution, while NO_x increases approximately twofold from 0.10 g/kWh at HER 0 to 0.21 g/kWh at HER 20, driven by peak temperature elevation. These findings establish hydrogen-enriched methanol–diesel dual-fuel combustion as a viable pathway toward high-efficiency, low-carbon CI engine operation for heavy-duty transport applications. Full article

Highway merging and platoon formation are critical scenarios in heavy-duty vehicle aerodynamics. This study presents a transient computational fluid dynamics (CFD) analysis of two trucks undergoing a merging maneuver and subsequent platoon formation. A three-dimensional unsteady Reynolds-Averaged Navier–Stokes (uRANS) approach with the SST k–ω turbulence model is employed under zero-crosswind and yawed inflow conditions. The present work provides a time-resolved characterization of truck–truck aerodynamic interactions during dynamic spacing evolution, enabling the capture of unsteady wake effects that are not accessible in steady-state formulations commonly used in cooperative driving studies. Unlike previous steady analyses, the approach resolves transient wake development, vortex shedding, and their direct impact on instantaneous aerodynamic loads. Results identify three interaction regimes: weak interaction, strong wake interaction during wake impingement, and wake recovery at larger spacing. Under zero-crosswind conditions, significant drag reduction is observed, confirming platooning benefits. However, crosswind conditions substantially reduce this benefit and increase lateral loads due to asymmetric pressure distribution and wake deflection. A non-linear spacing–drag relationship is observed, governed by wake evolution and shear-layer interaction. These findings provide quantitative insight into transient aerodynamic interactions and highlight the importance of accounting for unsteady and crosswind effects in platoon performance assessment. Full article

Future aircraft propulsion concepts (e.g., water-enhanced engines and fuel cells) will depend on efficient water recovery to enhance cycle efficiency and environmental performance. Operating conditions commonly involve droplet (mist) transport in turbulent air and wall-bounded films formed by droplet–wall interactions. This work develops an Eulerian–Lagrangian model within the RANS/URANS framework that accounts for air–droplet–wall phenomena—interfacial shear, impingement, and film advection. A dynamic contact-angle model, implemented and calibrated from static contact angle measurements performed in this study, represents wall wetting at the liquid–solid interface. The model is validated against experiments using two design metrics: pressure loss across the unit and recovered water mass fraction. At a low Mach number ($Ma=0.1$), saturated and dry air produce nearly identical pressure losses in the circular test section, whereas the separation lip geometry exerts a strong influence via local acceleration and separation. The simulations reproduce measured pressure drops and water mass recovery with close agreement. Full article

This paper presents a numerical investigation of vortex-induced vibration (VIV) of six elastically mounted circular cylinders in oscillatory flow, three smooth and three biofouled with triangular surface roughness elements. The study aims to characterise the influence of the longitudinal spacing ratio ($L/D=3,4,$ and $5$) on the two-degree-of-freedom (2DOF) vibration response at a constant Keulegan–Carpenter number of $KC=10$. Simulations are performed using the transient RANS equations with the SST $k–\omega$ turbulence model, and structural motion is resolved using a dynamic mesh approach. Lock-in behaviour is observed over the reduced velocity range $5\le U_r\le 10$. Biofouled cylinders generally exhibit higher in-line displacement amplitudes than smooth cylinders in the initial and lower lock-in branches, whereas smooth cylinders tend to attain higher in-line amplitudes in the upper lock-in branch. The spacing ratio $L/D$ is found to significantly influence the response, with peak vibration amplitudes varying non-uniformly across the array and no single spacing configuration being optimal for all cylinders. This behaviour is further supported by analyses of trajectories, frequency content, and vorticity fields. Among the smooth cylinders, the middle cylinder exhibits the largest in-line displacement amplitude of $3.28D$ at $L/D=5$ and the largest cross-flow displacement of $1.34D$ at $L/D=3$. For the biofouled configurations, the middle and upstream cylinders show the highest in-line displacement amplitude of $2.69D$ at $L/D=4$, while the maximum cross-flow displacement of $1.27D$ is observed for the upstream cylinder at $L/D=5$. Full article

Understanding the unsteady aerodynamic behavior of quadrotor unmanned aerial vehicle (UAV) is critical for improving flight stability, control, and performance, particularly in complex operational environments. In closely spaced multirotor configurations, coherent tip vortices shed from each blade convect downstream and form helical vortex streets that interact with subsequent blades and neighboring rotors. These interactions induce rapid fluctuations in local inflow velocity and effective angle of attack, resulting in transient lift variations, increased vibratory loads, and elevated acoustic emissions. This study presents a comprehensive computational investigation of quadrotor rotor interactions and wake dynamics using a large-eddy simulation (LES). Detailed analyses reveal that the formation and evolution of tip vortices and blade–vortex interaction phenomena significantly influence lift fluctuations and aerodynamic loading. The simulations capture transient wake structures and their effects on neighboring rotors, highlighting unsteady aerodynamic mechanisms that are not adequately predicted by conventional RANS or URANS approaches. Parametric studies examining vortex-street offset distance demonstrate the sensitivity of wake-induced instabilities to design and operational parameters. The results provide new physical insights into multirotor wake dynamics and establish the LES as a predictive framework for quantifying unsteady aerodynamic loading in quadrotor drones. The findings provide insights into the complex flow physics of multirotor systems, offering guidance for more accurate modeling, rotorcraft design optimization, and the development of control strategies that mitigate adverse unsteady aerodynamic effects. This study provides new insights into rotor–vortex-street interactions, with applications to multirotor UAVs, by isolating multi-vortex coupling effects and quantifying the influence of horizontal vortex spacing on unsteady aerodynamic loading, complementing existing high-fidelity LES research. Full article

Transonic buffet is a critical self-sustained shock/boundary-layer instability limiting the flight envelope of modern transport aircraft. This study investigates the interaction between shock buffet and forced wing motion on the Airbus XRF-1 wind tunnel model, using unsteady Reynolds-Averaged Navier–Stokes (URANS) simulations with the DLR TAU code. The investigation is carried out in deep buffet condition ($\mathrm{Ma}=0.84$, $\alpha =4.5^{\circ }$, $\mathrm{Re}=25\times {10}^6$) and validated against wind tunnel data at the same flow condition. The buffet flow is superimposed with forced wing motions derived from a symmetric wing eigenmode at $\mathrm{Sr}=0.164$. Two different amplitudes scaled with the half-span s are considered: $A_{tip}=0.0025·s$ and $0.01·s$. The baseline no-forcing URANS captures the buffet flow quite well with only small deviations in the standard deviation of the surface pressure coefficient $c_{p,rms}$. A special variant of the Discrete Fourier Transformation for the whole wing upper surface $c_p$ distribution revealed that the typical buffet frequencies are also matched. The analysis of the forced simulations revealed a strong influence of the local wing motion on the increase of $c_{p,rms}$. The spectral content showed a shift and damping or amplification of different buffet modes, which is relevant for the interaction of motion induced and buffed induced aerodynamic forces. Full article

A common solution for reducing the tonal noise annoyance caused by fans is to change the circumferential blade spacing from even to uneven. However, this technique requires predictive tools to simulate and assess their acoustic performance at a lower cost compared to experimental tests, which remain very costly. In this study, a hybrid analytic/numeric (HAN) approach for predicting the tonal noise of fans is proposed. It is based on the acoustic interference law, which is applied to the sound pressure generated by each blade, and Computational Aeroacoustics (CAA). This model allows for the analytical construction of a fan’s acoustic pressure spectrum from the numerically computed response of a single blade, significantly reducing computation time. An optimization procedure is then implemented to minimize the prominence of tonal noise peaks, where the decision variables are the blades’ angular positions and the constraints are rotor balance and the minimum angular distance between adjacent blades. The results show that the developed method may help designers reduce tonal noise annoyance by optimizing blade spacing. Full article

Visual perception in Unmanned Surface Vehicles (USVs) suffers from drastic lighting changes and missing texture features. These factors lead to depth scale drift and motion estimation bias. Moreover, existing multi-modal fusion models are computationally complex and unfit for resource-limited edge devices. To address these problems, a lightweight Radar–Vision Fusion (Geo-RVF) algorithm is proposed. To supplement spatial information, point clouds are projected to build sparse depth maps. A probability consistency-based depth correction module is designed to suppress water noise. This helps extract accurate geometric anchors to guide visual depth propagation. Subsequently, a Recurrent Autoregressive Network (RAN) fuses radar and image features in the temporal dimension. This resolves dynamic positional deviations caused by texture degradation and distant small targets. After real-time optimization, Geo-RVF achieves 23 FPS on the Jetson Orin NX. On a collected dataset, the method attains a mean average precision (mAP) 50–90 of 44.2% and a mean intersection over union (mIoU) of 99%, outperforming HybridNets and Achelous. Full article

Drifting fish aggregating devices (DFADs) are central to tropical tuna purse-seine fisheries, yet their hydrodynamic performance under realistic seas has not been adequately addressed, particularly for emerging eco-friendly designs. A three-dimensional framework based on computational fluid dynamics is developed to assess the motion response and mooring loads of full-scale DFADs comprising raft buoys, biodegradable cotton rope, and iron sinkers, using four buoy layouts (Models A to D). Unsteady Reynolds-averaged Navier–Stokes (URANS) simulations are performed with a realizable k–ε closure, volume of fluid (VOF) free-surface capturing, the Euler overlay method, dynamic overset meshes, and catenary mooring coupling. Regular waves representative of operational conditions (T = 1.40 to 2.40 s, H = 0.10 to 0.40 m) are imposed via a VOF wave-forcing technique, and mesh/time-step sensitivity analyses demonstrate the accurate reproduction of the first-order wave elevation (error < 0.8%). Surge drift per cycle and heave response amplitude operators, with the relative mooring force, are evaluated as functions of the relative wavelength (λ/L_a) and wave steepness (H/λ). The results reveal that the buoy layout exerts first-order control on DFAD dynamics, whereas short, steep waves dominate motion and line loads. The intermediate end-point sinker mass achieves a favorable balance between motion suppression and mooring load control, whereas distributing a fixed total sinker mass along the rope reduces heave response and mooring force by improving the tension redistribution and overall stability. Across all sea states, Models A and D reduced motion envelopes and mooring forces, indicating their suitability as robust, low-impact configurations. The proposed framework and design recommendations provide quantitative guidance for optimizing eco-DFAD geometry and deployment strategies, supporting safer and more sustainable DFAD-based tuna fisheries. Full article

Vertical axis wind turbines (VAWTs) suffer from dynamic stall (DS) at low tip-speed ratios (λ), where cyclic variations in angle of attack (α) dominate the blade aerodynamics, severely undermining aerodynamic performance and power extraction. The coupled influence of airfoil parameters on DS remains unexplored. To address this gap, a fully coupled parametric study using 126 incompressible URANS simulations is conducted, examining three geometric parameters of symmetric airfoils: maximum thickness (t/c), chordwise position of maximum thickness (x_t/c), and leading-edge (LE) radius index (I). The results show that coupled geometric modification fundamentally alters the stall mechanism, shifting it from abrupt, LE-driven separation toward a gradual, trailing-edge (TE)-controlled process as airfoils transition from thin, forward-x_t/c profiles to thicker configurations with aft x_t/c and reduced I. This transition enhances boundary-layer (BL) stability, delays DS onset, weakens dynamic stall vortex (DSV) formation, and mitigates unsteady aerodynamic loading. Within the investigated design space, the best-performing configuration (NACA0024–4.5/3.5) achieves a 73% increase in turbine power coefficient (C_P) relative to the baseline airfoil (NACA0018–6.0/3.0), mainly through passive control of BL separation and vortex development. These findings highlight the limitations of single-parameter optimization and establish a physics-based, coupled-design framework for mitigating DS-induced performance losses in VAWTs. Full article