This study demonstrated that when a semi-dilute polymer solution is injected into a microchannel with pillar arrays, oscillatory flow in the spanwise direction emerges throughout the channel when the Weissenberg number (Wi) exceeds a certain threshold. Previous studies have reported that viscoelastic fluids flowing through channels with obstacles exhibit unstable flow due to inherent instabilities. However, the dominant factors governing this transition and the influence of the solution’s rheological properties remain unclear. Therefore, our focus was on a semi-dilute polymer solution that generates dead zones (DZs) around pillar arrays during flow. In this study, a semi-dilute polymer solution was introduced into a microchannel with pillar arrays, and flow analysis was conducted using micro-particle image velocimetry (µPIV). At Wi ≤ 5, the flow remained steady along the pillars. In contrast, when Wi > 5, oscillatory flow in the spanwise direction was observed across the entire region of the pillar array in the channel. This oscillatory behavior is attributed to the spanwise oscillation of the DZs formed upstream of the initial pillars. Furthermore, when the orientations of multiple DZs upstream of the initial row of pillars align, strong diagonal flow emerges, which appears to trigger the oscillatory flow in the channel. As Wi increased, oscillatory flow at Wi ≥ 10 shifted toward higher frequencies. Higher Wi accelerates stretch–relaxation cycles in the downstream flow which destabilize the upstream DZs faster and cause this high-frequency behavior. These results demonstrate that channel obstacles may exacerbate flow instability. This provides fundamental insights into the flow physics of viscoelastic fluids around micropillar arrays and may contribute to the design of passive mixing and flow control strategies in microfluidic devices employing viscoelastic fluids.
This paper introduces a web-based library HiVis, designed for financial data visualizations. With the increasing complexity of financial data and the growing demand for intuitive authoring tools, there is a need for rapid development and high user usability financial data visualization solutions. HiVis offers a range of visual designs, dynamic charts, interactions, and customizable visualizations tailored to financial data, enabling users to explore and analyze complex finance datasets with ease. Moreover, the library provides a series of well-designed strategies for financial features and scenarios, including financial characteristics visualization, data compression, and data aggregation. HiVis aims to empower developers to effectively construct financial data visualizations that provide financial analysts, researchers, and professionals with powerful visualizations for data-driven decision-making.
Deep reinforcement learning (DRL) has achieved remarkable success in robotic autonomous control applications, owing to its superior performance. While existing visual analytics methods primarily focus on 2D single-agent environments, research on 3D multi-agent environments remains limited. In these environments, the increased dimensionality of agent action spaces and the complexity of dynamic interactions result in higher-dimensional data, demanding more sophisticated visualization techniques. This paper presents 3DMAViz, a visual analytics system designed to offer comprehensive insights into the training data of DRL models in 3D multi-agent environments. The system aims to enhance the understanding, diagnosis, and evaluation of these models, thereby improving their interpretability. To achieve this, 3DMAViz introduces an innovative ring target view that enables effective analysis of agent strategies through circular and multi-level visualizations, particularly in environments with increased dimensionality of agent action spaces. Additionally, it integrates a clustering and dimensionality reduction method with a deep Gaussian process model to classify high-dimensional patterns and quantify data correlations, thus improving the efficiency of identifying high-value information. Through case studies conducted in a 3D multi-agent simulation environment with domain researchers, we validate the effectiveness of 3DMAViz.
With the advent of polarization imaging sensors, flow-induced birefringence has emerged as a promising method for measuring fluid shear stress distributions. A primary challenge in establishing this measurement method is the determination of the photoelastic coefficient, which links measured retardation to the shear stress. In this study, we demonstrated quantitative imaging of a two-dimensional shear stress distribution based on flow-induced birefringence using polarized-image sensing. Cellulose nanocrystal (CNC) suspension was employed as the working fluid, and its photoelastic coefficient was calibrated in a circular Couette flow where the theoretical shear stress distribution can be analytically calculated. Based on the stress-optic law, shear stress was reconstructed from polarization intensity measurements using a camera equipped with a polarization image sensor. Using the calibrated photoelastic coefficient, we conducted quantitative measurements of the cumulative shear stress distribution of the fluid flowing vertically through a rectangular channel. The experimental results showed a distribution consistent with expectations, exhibiting low shear stress at the channel center and high stress near the walls. Furthermore, a numerical simulation was performed to verify the experimental findings. The simulation results showed good agreement with the experimental data in terms of the minimum values and the overall trend of the cumulative shear stress. In summary, the present study successfully demonstrates the fundamental feasibility and potential of polarized-image sensing for the quantitative evaluation of local shear stress fields.
With rising living standards and growing public concern, food safety has become a critical issue, prompting regulatory systems to shift from manual, reactive approaches to intelligent, data-driven risk prediction. However, the multi-source, heterogeneous nature of food safety data poses significant challenges for traditional machine learning methods, which struggle to capture complex relational semantics and provide interpretable results. To address these limitations, we propose a visual analytics framework that integrates heterogeneous graph neural networks (HGNNs) with metapath-driven semantic modeling. By constructing a heterogeneous graph of food samples and attributes, our approach captures cross-type associations and models multi-relational structures with enhanced interpretability. A tailored visualization strategy further bridges data structures, embedding representations, and predictive behaviors, enabling users to intuitively explore attribute relationships, trace metapath influences, and understand risk prediction logic through interactive, multi-level visual analysis. Experiments on real-world food safety datasets demonstrate that our method significantly improves both prediction accuracy and model transparency, offering practical decision-making support for intelligent supervision and contamination prevention.
The wake characteristics of a solid particle featuring a through-hole were investigated using particle image velocimetry. The particle had a diameter of d = 25.4 mm, with a through-hole diameter ratio of γ = d_h/d = 0.24 , where d_h represents the diameter of the through-hole. The particle was placed in a uniform flow with a velocity U_0 = 3.6 m/s, yielding a Reynolds number of approximately 6000. The orientation of the through-hole relative to the uniform flow α was varied from 0 deg to 90 deg. For the solid particle without a hole, the wake demonstrated a pronounced velocity deficit and a pair of symmetric vortices behind the particle. Introducing a through-hole resulted in the formation of a jet emerging from the hole into the wake at 10 ≤α≤ 60 deg. The jet velocity decreased as α increased, and the jet disappeared for α≥ 70 deg. Downstream of x/d ≥ 2 , the velocity field became independent of α . At α =0 deg, two pairs of vortices were generated by the interaction of the shear layers and hole jet. For α≥ 10 deg, one vortex of each pair disappeared. For 20 ≤α≤ 40 deg, the vortex size associated with the hole jet increased with increasing α . For α≥ 70 deg, the wake structure closely resembled that of a solid particle. These findings demonstrate that the through-hole angle exerts a significant influence on the near-wake structure, and they provide further support for the interference model between the hole jet and shear layer proposed in previous studies.
Current large language models, due to their inherent reliance on statistical correlations, are primarily optimized for generating coherent text rather than understanding deep semantic meanings, which hinders their effectiveness in solving complex reasoning tasks. Although popular prompting methods such as Chain-of-Thought (CoT) have been investigated to enhance the performance of LLMs for reasoning tasks, the intermediate reasoning steps generated remain uncontrolled, making it difficult to trace and treat hallucinations. Thus, we propose VisFIT, a Visual analysis method inspired by analogical transfer from Family Inheritance scenarios, utilizing Tree-structured prompts. VisFIT creatively draws an analogy to the dual rules of inheritance. On one hand, LLM assigns a “preference level” to each child node while performing automated reasoning based on the hierarchical prompts, which enables it to execute “testamentary inheritance". On the other hand, Users can act as “judges", transparently monitoring and modifying the reasoning process through interactive interventions, thereby implementing “statutory inheritance" as a form of guided adjustment. Specifically, VisFIT first interactively constructs a tree structure named FIT from the underlying logic in the relevant text, using it as a structural prompt for LLMs to tackle complex reasoning tasks. Visualizations at varying levels of granularity enable the tracing back of erroneous reasoning steps to their sources. Then, users can then iteratively adjust FIT to refine the prompts, thereby enhancing the model’s performance in specific domains. Additionally, we conducted quantitative study, case analysis, and expert review to comprehensively evaluate VisFIT. The results show that it improves the average accuracy by 30.37
Retrospective public opinion analysis provides a scientific basis for subsequent public opinion management and decision optimization. However, current analyses overlook the reflection of popular news and their reading data on the daily concerns of the public, and existing visualization systems are overly complex, hindering their use by domain experts. We designed and implemented a highly usable retrospective public opinion visualization system focusing on massive news data. Based on domain experts’ needs and design principles, the system features intuitive interfaces and interactions. For instance, we provide a temporally stable news map, which integrates contour, time brush, and lens-based interaction to enable rapid hotspot localization, comprehension, and division of evolution phases. To support exploring event evolution in sentiment and popularity, we proposed sentiment metrics to identify events with features like volatility and reversal, plus an event map that clusters events to promote understanding of evolutionary patterns. A comprehensive evaluation systematically demonstrates our system’s effectiveness; a case study on early COVID-19 public opinion in China offers a perspective for analyzing the pandemic via public opinion, facilitating the excavation of pandemic-related narratives.
Accurate numerical prediction of subcooled boiling flows remains challenging due to the complex interaction between phase change, interfacial transport, and turbulence. This study presents a conjugate fluid–solid phase-change solver for the simulation of subcooled boiling flow. The solver couples heat conduction in the solid region with phase change in the fluid domain using a volume-of-fluid (VOF)-based formulation. The α-transport equation is solved using both fully explicit and semi-implicit MULES formulations. The solver is first validated against classical Stefan problems to verify interface propagation. Subsequently, vertical subcooled boiling flow under two operating conditions is simulated and compared with experimental data and ANSYS Fluent predictions. The model accurately captures key boiling characteristics, including the onset of nucleate boiling, transition to two-phase flow, and axial wall temperature evolution. The vapor volume fraction is reasonably predicted, although deviations in magnitude are observed under certain conditions. Performance analysis shows that for Stefan problems, the semi-implicit formulation significantly reduces computational time, particularly at higher Courant numbers. For conjugate subcooled boiling simulations, a moderate improvement in computational efficiency of approximately 5–14
This study presents a novel Y-junction micromixer configuration that enhances mixing of high-viscosity liquids through 3D acoustic streaming flows. Micro-scale flow visualization experiments were conducted to investigate the impact of viscosity variations on liquid-liquid mixing behavior, specifically the challenges posed by high-viscosity laminar flows. The acoustic micromixer's mixing performance was evaluated using water-ethanol and ethylene glycol-water liquid mixtures, yielding a range of viscosity ratios from 1 to 8.21 between the two distinct miscible liquids. A parametric study was also conducted to explore the influence of various operating parameters, including fluid viscosity, volume flow rate, oscillation frequency, and amplitude. This study employed the digital in-line holographic microparticle tracking velocimetry (DIH-µPTV) technique to visualize the streaming flow and mercury lamp-induced epi-fluorescence mixing, and to measure scalar transport concentrations using an in-house MATLAB code. The findings indicate that the combined effects of acoustic streaming and molecular diffusion induced by concentration gradients within the liquid significantly enhance the mixing efficiency. The optimal conditions were identified at a frequency of 12 kHz, an amplitude of 34 V, a flow rate of 80 L/min, and a liquid viscosity ratio of 8.21, resulting in a significant improvement in the mixing index to 96.26
The pervasive use of social media has surpassed traditional temporal and spatial boundaries, fundamentally reshaping how user-generated content is produced, disseminated, and consumed. Posts from bloggers and other content creators typically contain multi-dimensional information, including semantic content and geographic tags. However, existing studies often analyze textual data or trajectory data in isolation, overlooking the correlations between these dimensions. To bridge this gap, we propose a novel visualization approach that captures the spatio-temporal-semantic coupling embedded in users’ social media data. Our method integrates a honeycomb layout for constructing semantic maps with customized visual encodings, enabling the simultaneous exploration of individual trajectory patterns and semantic attributes. We develop an interactive visualization system based on a dataset of Xiaohongshu blogger posts, supporting multi-scale analysis from macro-level trends to micro-level details. Case studies demonstrate the system’s effectiveness in revealing bloggers’ content creation patterns and behavioral characteristics.
Left ventricular assist devices (LVADs) are mechanical circulatory support used in case of heart failure; however, their implantation can lead to hemodynamic complications, including aortic regurgitation (AR) and vascular damage. In this study, we analyze the impact of LVAD outflow graft inclination angles on aortic flow patterns, wall shear stress (WSS) distribution, and turbulent kinetic energy (TKE) using computational fluid dynamics (CFD) simulations. Simulations were performed in a patient-specific aortic geometry for four idealized graft angles (45°, 60°, 75°, and 90°) across three cardiac phases: systole, diastole with a closed aortic valve (AV-closed), and diastole with aortic regurgitation (AR). A full patient-specific model incorporating a real graft inclination of 89° was also analyzed for comparison. Our simulation results demonstrated that smaller inclination angles (e.g., 45°) induce elevated WSS at the brachiocephalic artery (BCA) branch and pronounced recirculation zones in the aortic lumen. In contrast, steeper angles (e.g., 90°) cause high WSS and recirculating flow near the aortic root and valve, which may exacerbate the risk of AR. The patient-specific 89° graft closely resembles the 90° idealized model, but exhibits broader WSS distribution at epiaortic branch inlets, higher outer wall WSS, and more complex vortices resulting from sharply bent and narrowed cannula. The left common carotid artery (LCCA) consistently exhibited the highest WSS among epiaortic vessels, which might be attributed to its smaller diameter and higher blood flow velocity. Hemolysis was assessed using shear strain rate and scalar shear stress (SSS), showing strong sensitivity to graft inclination, with the 75° configuration producing the largest high-shear volume fractions, while steeper angles sustained elevated shear under regurgitant conditions. These findings highlight the importance of optimizing LVAD outflow graft positioning to mitigate adverse hemodynamic effects on both aortic valve (AV) and carotid arteries.
Proton exchange membrane (PEM) water electrolyzer is a promising technology for green hydrogen production. However, its performance is often limited by the accumulation of gas bubbles at the porous transport layer (PTL) surface, especially at high-current densities. Effective management of bubble transport is therefore essential for improving two-phase flow and electrochemical performance. To address this, 3D flow channel designs are being explored to enhance flow and advection for superior bubble stability. In this study, three different flow channels, such as a conventional straight channel (Basic) and two 3D wave-shaped channels (Wave 1 and Wave 2), are experimentally investigated to analyze bubble dynamics and their impact on system performance. The effects of operating conditions, including temperature (40–80 °C) and flow rate (15–75 mL/min) are studied. Bubble behavior between the PTL-flow channel interface is captured using high-speed imaging and analyzed through an artificial intelligence-based segmentation model (nnU-Net). The results show that the flow field geometry influenced the bubble transport. In particular, Wave 2 channel exhibited superior performance by promoting finer bubble dispersion and more regular detachment compared with the Wave 1 and Basic channel designs. This reduces bubble coverage at the PTL surface and facilitates effective bubble removal, leading to lower cell voltage and enhanced system efficiency. The present study highlights the importance of flow field design in controlling two-phase flow behavior and offers insights for optimizing flow channel geometries in PEM water electrolyzers.
Over the past two decades, light sheet fluorescence microscopy has developed to a powerful tool for studies of dynamics in biological systems. In a new development, we apply light sheet fluorescence microscopy as a novel experimental measurement technology within the field of multiphase process engineering and fluid dynamics. The technology enables a number of novel studies of single- and multiphase transport phenomena on a microscopic scale. In the current publication, we introduce the first implementation of light sheet fluorescence microscopy in the field and demonstrate its applicability on one exemplary measurement of diffusive oxygen mass transfer from an oxygen bubble to degassed water. The results prove that such measurements can be conducted with a high spatial resolution with a submicron pixel pitch and enable precise studies on microscopic transport phenomena. Besides measurements of mass transfer, on which we lay the focus here, light sheet fluorescence microscopy further enables studies of fluid dynamics on a microscopic scale.
The inducer is a key component of aerospace centrifugal pumps, and its performance directly affects cavitation characteristics and overall efficiency of the pump. Based on the theory of reverse engineering, a parametric intelligent optimization (PIO) platform for the inducer was constructed, achieving precise extraction of geometric parameters, numerical simulation, and automatic optimization design. The head of the inducer was chosen as a constraint; the maximum hydraulic efficiency and the minimum critical net positive suction head (NPSHr) were chosen as the objectives, and the Latin hypercube sampling method and the Particle Swarm Optimization algorithm were used to perform multi-objective optimization on the inducer. The results demonstrate that the proposed reverse reconstruction and optimization framework achieves high predictive accuracy, with relative errors of only 1.54
Buoyancy-driven instabilities often occur at the interface between two miscible reactive solutions due to complex density stratifications arising from differences in solutal concentrations and diffusion coefficients. In this study, to visualize and characterize these instabilities, two-dimensional concentration maps of sulfuric acid (H2SO4), sodium hydroxide (NaOH), and sodium sulfate (Na2SO4) were simultaneously obtained for a system where an aqueous H2SO4 solution was layered over aqueous NaOH in a microfluidic channel and a modified Hele–Shaw cell. This measurement method exploits the distinct spectral variations of the near-infrared water absorption band induced by these species. The resulting maps revealed that the H2SO4 and NaOH regions remained separated, with the product Na2SO4 forming at the interface. Subsequently, the interface underwent fluctuations, triggering upward convection plumes from the H2SO4–Na2SO4 boundary and downward finger-like plumes from the Na2SO4–NaOH boundary. Density profiling revealed distinct driving mechanisms for these asymmetric instabilities: a locally unstable stratification (heavy over light) at the upper interface drove Rayleigh–Taylor convection, whereas the lower interface remained statically stable, indicating that the downward fingers were driven by double-diffusive instability.
With the increasing scale and complexity of graph data, graph sampling has become a crucial dimensionality reduction technique, while the evaluation of its effectiveness has also garnered significant attention. However, traditional sampling evaluation methods primarily focus on preserving topological structures while neglecting the integrity of node attributes. To address this, we introduce an attribute-aware evaluation framework for assessing graph sampling differences. First, a novel taxonomy is proposed to categorize sampling differences into two major types: Attribute-Isolation and Attribute-Structure. The former is further divided into Attribute-Distribution and Attribute-Correlation, while the latter includes three subcategories: Attribute-Node, Attribute-Path, and Attribute-Community. Based on this, we design a set of computable metrics to quantify each type of attribute-related differences. Furthermore, we develop an interactive visualization system that integrates multiview visual modules to visualize sampling impacts on attributes and structural correlations. Case studies and quantitative evaluations demonstrate the effectiveness of our method in characterizing the impact of different sampling strategies on attribute information and aiding users in network analysis and strategy selection.
This study employs large eddy simulation to systematically investigate the primary flow patterns and flow field characteristics in a cross-pipe configuration featuring various adjacent closed-cavity length ratios (La/Lb = 1, 2, and 0.5) at a Reynolds number of 10,000. The spatial distribution of acoustic sources is qualitatively evaluated based on the extracted λ-vector and Lighthill’s acoustic analogy. Spectral proper orthogonal decomposition (SPOD) is employed to elucidate the correlation between flow structures and noise generation. The results highlight the impact of cavity geometry on flow behavior and flow-induced noise. For La/Lb = 1, the flow is characterized by three unstable shear layer structures, which dominate the distribution of vorticity and Reynolds shear stress. The noise sources generated by fluid–wall interactions span the entire downstream region, the cross-junction, and the straight cavity (cavity A). For La/Lb = 2, the extended cavity length enables full flow recirculation, which reduces vortex strength and weakens acoustic source intensity, while increasing its distribution range. However, the increased length of the side cavity (cavity B) enhances flow separation at the downstream trailing edge, which in turn generates additional acoustic sources near the outer downstream wall. SPOD analysis reveals that the dominant mode across all configurations is a large-scale wavepacket structure originating from flow separation at the downstream leading edge. An extended straight cavity enhances internal recirculation, reducing the main flow energy, thereby weakening the wavepacket structures and accelerating their downstream dissipation. For La/Lb = 0.5, the wavepackets’ behavior below the cross-junction is intensified. At high frequencies, the downstream wavepackets become smaller and more concentrated, with a slower rate of dissipation. Therefore, valve installation in hydrogen pipelines should prioritize extending the straight cavity. This study provides valuable guidance for optimizing structural vibration and noise generation in hydrogen pipeline systems. Contours of a the instantaneous and b the time averaged lambda magnitude for three configurations with La /Lb = 1, 2, and 0.5
Geospatial heatmaps have become a predominant visualization modality for urban data analysis, particularly for representing population distributions and mobility patterns. While offering rich spatial insights, these visualizations exhibit greater structural complexity compared to conventional charts, and are frequently compromised by non-expert designers contravening perceptual design principles. We propose a novel four-stage pipeline for data recovery: (1) Layer decoupling through generative adversarial networks (GANs) that disentangle geospatial and density map components; (2) Dual extraction process combining SLIC (Simple Linear Iterative Clustering) superpixel analysis for color decoding with hybrid OCR-image matching for coordinate alignment; (3) Iterative generative estimation for density map deconstruction; (4) Multimodal fusion integrating color, spatial, and density features for source data reconstruction. Our experimental evaluation leverages a multimodal corpus comprising 6670 synthetic samples and 67 real-world heatmaps curated from scientific literature and news media. The proposed method demonstrates three transformative applications: (1) Vectorization of raster heatmaps enabling multiscale interaction; (2) Dynamic recoloring for perceptual optimization; (3) Explanatory overlay generation through automated statistical annotation. Quantitative evaluations show 0.94 correlation ( R^2 ) in data recovery accuracy, with 61
Spatial data analysis is essential in domains such as construction, architecture, and geodesy, to check for differences between construction stages of a building. In this context, real-world environments are often digitally recorded as point clouds, providing a basis for digital processing and analysis. Furthermore, aspects of spatial cognition, often measured with eye tracking, are also relevant to many research questions related to how people perceive and understand space and spatial structures, e.g., in quality inspection of construction sites or buildings. We present a workflow to combine eye tracking and point cloud data to analyze user attention during inspection of spatial data in virtual environments. This type of data enrichment enables the deployment of established open-source software for cloud and mesh processing to analyze eye tracking data. Our workflow employs the investigation of point cloud data and simultaneous user gaze recording in virtual reality, followed by point-based gaze data analysis, interactive visualization, and authoring and production of high-quality dissemination materials for stakeholders. We showcase our approach with a usage scenario on how people investigate the construction of a pavilion building in virtual reality.
