Journal of Visualization最新文献

Abstract

The explosive growth in supercomputers capacity has changed simulation paradigms. Simulations have shifted from a few lengthy ones to an ensemble of multiple simulations with varying initial conditions or input parameters. Thus, an ensemble consists of large volumes of multi-dimensional data that could go beyond the exascale boundaries. However, the disparity in growth rates between storage capabilities and computing resources results in I/O bottlenecks. This makes it impractical to utilize conventional post-processing and visualization tools for analyzing such massive simulation ensembles. In situ visualization approaches alleviate I/O constraints by saving predetermined visualizations in image databases during simulation. Nevertheless, the unavailability of output raw data restricts the flexibility of post hoc exploration of in situ approaches. Much research has been conducted to mitigate this limitation, but it falls short when it comes to simultaneously exploring and analyzing parameter and ensemble spaces. In this paper, we propose an expert-in-the-loop visual exploration analytic approach. The proposed approach leverages: feature extraction, deep learning, and human expert–AI collaboration techniques to explore and analyze image-based ensembles. Our approach utilizes local features and deep learning techniques to learn the image features of ensemble members. The extracted features are then combined with simulation input parameters and fed to the visualization pipeline for in-depth exploration and analysis using human expert + AI interaction techniques. We show the effectiveness of our approach using several scientific simulation ensembles.

Graphical abstract

We present an approach that supports the analysis of flow dynamics in the neighborhood of curved line-type features, such as vortex core lines, attachment lines, and trajectories. We achieve this with continuous deformation to the flow field to straighten such features. This provides “deformed frames of reference”, within which qualitative flow dynamics are better observable with respect to the feature. Our approach operates at interactive rates on graphics hardware, and supports exploration of large and complex datasets by continuously navigating the additional degree of freedom of deformation. We demonstrate the properties and the utility of our approach using synthetic and simulated flow fields, with a focus on the application to vortex core lines.

Graphical abstract

Abstract

Data dashboards with intuitive visualizations make information more accessible and provide a more in-depth explanation. They have emerged as a crucial tool for effectively communicating pandemic information to wide-ranging audiences. However, the urgency and high-impact nature of pandemics requires rapid and trustworthy dashboard creation. Studies shown that information transparency plays a pivotal role in building trust. Therefore, in this paper, we present FlowDashboard, a domain-specific visualization framework that enables users to create pandemic dashboards quickly and transparently. Our design for FlowDashboard is guided by qualitative analysis of 207 practical pandemic dashboards. Based on the identified key requirements of speed and transparency, a novel transparent flow model called TransFlow is proposed as the core dashboard creation approach. This model formalizes intuitive flow diagram design to construct interactive dashboards, making it easy to learn and revealing the underlying data and interaction flows at property level. Additionally, the FlowDashboard framework accommodates all common components used in practical pandemic dashboards, and incorporate the pandemic gallery as an interface to facilitate users quickly learning the design space. Through use cases, user study and comparisons to state-of-the-art works, we demonstrate the usability and effectiveness of our framework.

Graphical abstract

This study investigates the three-dimensionality of synthetic jet flow control over a NACA 0025 profile wing using horizontal and vertical smoke wire visualization. The stalled flow in the baseline case is visualized, providing insights into the shear layer roll-up process, the transition to turbulence, and vortex shedding in the wake. In the controlled flow study, two actuation frequencies, (F^+=1.18) and (F^+=11.76), are investigated, with a focus on spanwise control authority and the role of coherent structures in flow reattachment. The results indicate that while the control is effective at the midspan over the entire chord length, its effect diminishes with increasing distance from the midspan. Both control cases result in significant spanwise velocities, observed by a contraction of the flow toward midspan. Lastly, the high-frequency actuation results in unique small-scale structures at the shear layer-freestream interface.

Graphical abstract

The rotating stall generated by the centrifugal pump impeller reduces efficiency and causes severe flow fluctuations and noise due to surging. In the present work, the six-bladed impeller in the centrifugal pump was simulated by RANS, LES, and Hybrid RANS/LES turbulence models using ANSYS CFX. The turbulence models considered were the Shear Stress Transport (SST), Detached Eddy Simulation (DES), Stress-Blended Eddy Simulation (SBES), Scale Adaptive Simulation (SAS), and Wall-Adapted Local Eddy-viscosity (WALE). The design load condition and the quarter-load condition were applied for the boundary conditions, and the experimental results were compared and analyzed using velocity profile and turbulent kinetic energy at the impeller mid-height. Under the design load condition, all turbulence models predicted results similar to the experimental results. In the off-design load condition, LES predicted the experimental value most accurately, followed by SST-RM of RANS with high accuracy, and the hybrid RANS/LES model showing lower prediction accuracy; SBES predicted excessive recirculation flow. However, if sufficient grid resolution is achieved, hybrid RANS/LES model can simulate the rotating stall under the off-design flow condition than RANS models. Both DES and SAS model show relatively low mesh dependent results with acceptable accuracy.

Graphical abstract

Abstract

Visual analysis of epidemic spread data is crucial in understanding the process of epidemic transmission, tracing the source of infection, evaluating the development of epidemic, and formulating reasonable policies. However, due to limited capabilities in collecting comprehensive data on the spread of the epidemic, it remains challenging to fully and visually comprehend the spatial and temporal changes in virus transmission, which, in turn, hampers efforts in exploration of macro-level pattern analysis and validation of micro-level facts. To fill this gap, in this paper, we propose a virtual simulation and visual analysis system, named VIVIAN. The user-friendly interaction design of the system enables rapid infection traceability and accurate investigation of close contacts. Furthermore, it establishes synchronous correlations between model parameters and epidemic prevention and control measures. The system supports users in simulating the formulation of epidemic prevention and control policies and evaluating their effectiveness. The system automatically generates the movement trajectories and contact situation based user-defined thresholds, which relaxes the data challenge and scalability of the system. The system is equipped with multiple linked, intuitive, and interactive visualization charts for rapid infection traceability and accurate close contact investigation. In addition, the system supports users in formulating and evaluating of epidemic prevention and control policies. Case studies and expert interviews based on simulated data have demonstrated the effectiveness and practicality of the system, which make it, as a foundation, possible to be employed in certain scenes for epidemic prevention and controls.

Graphical abstract

Runback ice can lead to significant aerodynamic deterioration and affects aircraft flight safety. The frozen rivulets can form during the aircraft icing process as a part of the runback ice. So far, the effect of frozen rivulets on the aerodynamic performance of airfoils has not been fully studied. The present study experimentally investigated the effects of the length of frozen rivulets on the flow field structures and aerodynamic performance of an airfoil. Detailed measurements were performed in a low-speed reflux wind tunnel by utilizing the particle Image Velocimetry technique and a high-sensitivity six-component balance. The results showed that the length of frozen rivulets had minor effects on lift and drag coefficients. The presence of the frozen rivulets delayed the trend of decreasing pitching moment and reduced the maximum lift–drag ratios of airfoils. Moreover, the frozen rivulets affected the separation bubble length. The separation of the airflow became more pronounced as the length of frozen rivulets increased. The spanwise vorticity distributions were more continuous and smoother with the increase of the length of frozen rivulets. In addition, the transition onset positions showed a strong dependence on the frozen rivulets. Besides, the length of frozen rivulets had limited influences on the turbulent kinetic energy and Reynolds stress distributions.

Graphical abstract

Most users are able to obtain exploratory ideas from a data table but cannot clearly declare their analysis tasks as visual queries. Visualization recommendation methods can reduce the demand for data and design knowledge by extracting or referring information from existing high-quality views. However, most solutions cannot identify analysis tasks, which limits the accuracy of their recommendations. To address this limitation, we propose a deep learning and answer set programming-based approach to guide visualization recommendations by tracking potential analysis tasks and field preferences in exploration interactions. We demonstrate this approach via Dowsing, a mixed-initiative system for visual data exploration that automatically identifies and presents users’ potential analysis tasks and recommends visualizations during exploration. Additionally, Dowsing allows users to confirm and edit their intentions in multiple ways to adapt to changing analysis requirements. The effectiveness and usability of our approach are validated through quantitative experiments and two user studies.

Graphical abstract

As one of the effective means of representing geographic information, geographic visualization can directly improve the cognitive efficiency of users who are perceiving geospatial data. The existing geographic information visualization relies heavily on the background knowledge and visualization skills the data workers own. Therefore, the geographic visualization task is usually very time-consuming and challenging. To lower the barrier of visualization of geographical data, we propose a novel recommendation system of geographic information visualization called GeoVis. This system extracts the distribution characteristics with adaptive kernel density estimation and recommends the map type (scatter, bubble, hexbin and heatmap) that can best reflect the regularity of data distribution based on latent code. The key idea of how the data-driven recommendation works is to use latent code to express and decouple data features and then learn the mapping between data features and visual styles. At the same time, this system recommends design choices (e.g., map styles and color schemes). Users only need to browse the recommendation results to realize explorations and analyses of the dataset, which will greatly improve their work efficiency. We conduct a series of evaluation experiments on the proposed system, including a case study. The experiment results show that the system is practical and effective and can perform the task of recommending informative and esthetic geographical visualization results well.

Graphical abstract

Abstract

In the field of deep learning, pre-trained BERT models have achieved remarkable success. However, the accompanying problem is that models with more complex structures and more network parameters. The huge parameter size makes the computational cost in terms of time and memory become extremely expensive. Recent work has indicated that BERT models own a significant amount of redundant attention heads. Meanwhile considerable BERT models compression algorithms have been proposed, which can effectively reduce model complexity and redundancy with pruning some attention heads. Nevertheless, existing automated model compression solutions are mainly based on predetermined pruning program, which requires multiple expensive pruning-retraining cycles or heuristic designs to select additional hyperparameters. Furthermore, the training process of BERT models is a black box, and lacks interpretability, which makes researchers cannot intuitively understand the optimization process of the model. In this paper, we propose a visual analysis system, BHPVAS, for pruning BERT models, which helps researchers to incorporate their understanding of model structure and operating mechanism into the model pruning process and generate pruning schemes. We propose three pruning criteria based on the attention data, namely, importance score, stability score, and similarity score, for evaluating the importance of self-attention heads. Additionally, we design multiple collaborative views to display the entire pruning process, guiding users to carry out pruning. Our system supports exploring the role of self-attention heads in the model inference process using text dependency relations and attention weight distribution. Finally, we conduct two case studies to demonstrate how to use the system for Sentiment Classification Sample Analysis and Pruning Scheme Exploration, verifying the effectiveness of the visual analysis system.

Graphical Abstract