Research Project D01

Visualization of multi-field processes in porous media

People

Thomas Ertl

Steffen Frey (since 01.07.2020 at the University of Groningen)

 

Alexander Straub (from 01.10.2020)

Stefan Scheller (until 30.09.2020)

Steffen Frey (University of Groningen, Bernoulli Institute, Scientific Visualization and Computer Graphics)

Publications

Publications in scientific journals

  1. Frey, S. (2020). Temporally Dense Exploration of Moving and Deforming Shapes. Computer Graphics Forum, 40(1), 7--21. https://doi.org/10.1111/cgf.14092
  2. de Winter, D. A. M., Weishaupt, K., Scheller, S., Frey, S., Raoof, A., Hassanizadeh, S. M., & Helmig, R. (2020). The Complexity of Porous Media Flow Characterized in a Microfluidic Model Based on Confocal Laser Scanning Microscopy and Micro-PIV. Transport in Porous Media. https://doi.org/10.1007/s11242-020-01515-9
  3. Gralka, P., Grottel, S., Staib, J., Schatz, K., Karch, G. K., Hirschler, M., Krone, M., Reina, G., Gumhold, S., & Ertl, T. (2018). 2016 IEEE Scientific Visualization Contest Winner: Visual and Structural Analysis of Point-based Simulation Ensembles. IEEE Computer Graphics and Applications, 38(3), 106–117. https://doi.org/10.1109/MCG.2017.3301120
  4. Zhang, H., Frey, S., Steeb, H., Uribe, D., Ertl, T., & Wang, W. (2018). Visualization of Bubble Formation in Porous Media. IEEE Transactions on Visualization and Computer Graphics, 1–1. https://doi.org/10.1109/TVCG.2018.2864506
  5. Frey, S. (2018). Spatio-Temporal Contours from Deep Volume Raycasting. Computer Graphics Forum, 37(3), 513–524. https://doi.org/10.1111/cgf.13438
  6. Frey, S., & Ertl, T. (2017). Flow-Based Temporal Selection for Interactive Volume Visualization. Computer Graphics Forum, 36(8), 153–165. https://doi.org/10.1111/cgf.13070

Research

About this Project

The long-term goal of this project is to develop new visualization techniques to support the projects of the CRC in understanding flow, transport and deformation phenomena occurring at interfaces in and around porous media. This requires the integrated visual analysis of processes captured across multiple fields to relate different quantities, analyze simulation ensembles, and compare simulations against experiments. The interactive exploration of large and heterogeneous volumes of data is achieved by efficiently using parallel architectures as well as insitu data aggregation. To bridge the gap between the techniques used by domain scientists and the approaches available from visualization research, we intend to continue developing visualization approaches for specific projects, and combine them into a flexible porous media visualization framework.

Results

Volume Rendering

For the visualization of time-dependent volume data and as a basis for visualization tools developed later in the project, we implemented a free and open source volume renderer. This is a universal tool for the visualization of 3D scalar data and has already been used to visualize profiles at fluid-fluid and fluid-solid interfaces.

Example datasets rendered with our volume renderer, showing fluid-fluid (left) and fluid-solid (right) interface density profiles.
Spatio-temporal contours for time-dependent aggregation

Introducing a contour visualization that indicates spatial outlines of continuous processes, we allow aggregating dynamic fields from spatio-temporal processes to provide an overview and support interactive exploration. This aggregation of all time steps in one visualization enables us to efficiently analyze processes in the data. The visualization can be combined with volume rendering for context. To finally obtain a temporally dense representation that still maintains shape information, we convolve a noise texture along shape outlines at every time step, thus conveying shape changes over time.

Static visualization of spatio-temporal processes for a two-phase flow simulation of colliding droplets (top) and a laser pulse experiment (bottom), showing selection (left), contours (center) and shapes (right).

S. Frey 2020, Temporally Dense Exploration of Moving and Deforming Shapes

Bubble extraction and comparative visualization

We investigated the influence of sequestrating bubbles on the seismic characteristics of porous rock formations based on micro-CT data of liquid CO2 filled sandstone. To this end, we developed an automatic technique that extracts emerged bubbles and their surrounding structure. To analyze how the structure of the porous medium affects the occurrence and shape of formed bubbles, we automatically classify and relate them in terms of morphology and geometric features. The results are visualized using 3D volume rendering and coordinated views. Bubbles are then clustered by similarity and are either visualized side-by-side or in a single view for comparison.

Overview of our approach for the analysis of bubble formation. After the detection and ex-traction of bubbles and their surroundings (a), they are classified, similar structures identified (b), and clustered (c). Bubbles and their surrounding structures are registered and shown from different angles(d), and can finally be visualized together in one integrated view (e).

H. Zhang et al. 2019, Visualization of Bubble Formation in Porous Media

Stream surfaces for flow analysis

Extracting and visualizing stream surfaces between different flow regions and illustrating transport mechanisms with animated renderings, we investigated the behavior of single-phase flow in microfluidic structures. This includes the experimental recording of flow profiles, and a comparison between measured and simulated behavior.

Streamlines show the instantaneous flow of mass-less particles, with color mapped to integration time (blue to red). Imaginary barriers (stream surfaces; grey) depict boundaries between bundles of streamlines.

D.A.M. de Winter et al. 2021, The Complexity of Porous Media Flow Characterized in a Microfluidic Model Based on Confocal Laser Scanning Microscopy and Micro-PIV

Abstract visualization of displacement in porous media

For the investigation of a series of two-phase flow displacement experiments, we developed a visual analysis approach with the goal of better understanding the involved processes and relating them to the experiment parameter space. With a novel node-link diagram based representation, we temporally aggregate the propagation of the displacing fluid through the porous medium, yielding the usage rate. The two-dimensional parameter space can be analyzed using a small multiples representation of these diagrams. Transport graphs also enable to quantify the relation between displacement and pore-space characteristics, such as pore throat width.

Applied parameter space exploration

Our visual analysis method for a coalbed-methane production model allows to analyze the results of multiple simulation runs in several coordinated views. Rather than visualizing the data concurrently in space, we represent them as high-dimensional vectors in the parameter space of the underlying model. The influence of model parameters can be explored via dimensionality reduction and clustering techniques, while investigating the results in a combined view of scatter plots, line plots, and parallel coordinate plots.

Exploration of the parameter space of a microbially enhanced coalbed-methane production model. From left to right: The high-dimensional parameters are projected into 2D space, clustered to identify groups of similar behavior (color-coded), the results are visualized in line and parallel coordinates plots to identify regimes with characteristic temporal evolution of methane production.
Visual analysis of parameter spaces

To deal with large numbers of simulations or experiments and especially high-dimensional parameter spaces, we developed a generic approach to scalable visual analyses. We cluster ensemble data from simulations in computational fluid dynamics and thermodynamics, and relate them to their position in parameter space. Employing glyphs in a novel and intuitive visual representation, we show the transition between homogeneous regions in parameter space.

Kármán Vortex Street ensemble overview. From left to right, adjacent parameter space regions can be seen turning from turbulent to laminar, until they get turbulent again. The transition colors shown in the interfaces allow a quick assessment of the reason: first lowering the Reynolds number is responsible (mainly green transitions), while the right half of transitions imply that the offset of the obstacle from the symmetry axis causes the turbulence (orange).

O. Fernandes et al. 2019, Visual Representation of Region Transitions in Multi-dimensional Parameter Spaces

High performance visualization

High performance visualization covers approaches from maintaining interactivity for medium-size data sets to being able to process large-scale data sets at all. Here, we first focused on the analysis of large spatio-temporal data and on interactivity. For the Pretty Porous exhibition, we developed an interactive demonstrator for the investigation of geothermal energy sites. Through touch input, visitors could interactively modify the properties of two sets of injection and extraction wells, and explore the underground heat transport at geothermal energy sites. The underlying Polynomial Chaos Expansion (PCE) model is evaluated dynamically on the GPU. This allows to compute the temperatures in each cell of the domain on the graphics accelerator hardware and to visualize the results with interactive frame rates.

In situ coupling for visualization

Further, we have worked on an in situ visualization approach that makes use of preCICE to ingest the simulation data. Our current software prototype is coupled to a DuMux solver for fractures in porous media benchmark scenarios, providing a uniform way to visualize the results of arbitrary solvers that implement these benchmark cases.

For further information please contact

This picture showsThomas Ertl
Prof. Dr. rer. nat. Dr. techn. h.c. Dr.-Ing. E.h.

Thomas Ertl

Principal Investigator, Research Project D01, Project Public Relation

This picture showsSteffen Frey
Dr. rer. nat.

Steffen Frey

Principal Investigator, Research Project D01

To the top of the page