Facebook Logo Twitter Logo RSS Logo
Issue: Volume: 24 Issue: 1 (January 2001)

Fluid Motion in Focus




Scientists and engineers are increasingly looking to computer visualization to gain a better understanding of how substances flow in, through, and around complex structures. Such information is critical to a broad range of applications, because the magnitude, force, and direction of flow can have a significant impact on the state and operation of any number of objects and assemblies.

Unfortunately, while techniques for visualizing complex flow phenomena have become quite sophisticated in recent years, they are still lacking on a number of fronts. Chief among these is an inability to achieve both a local and global view of flow behavior within the same visualization. For example, traditional vector graphics effectively uses icons such as arrows, particles, and stream ribbons to reveal local flow features, but these techniques can preclude a comprehensive understanding of the evolution of the overall flow over time. This is because limited spatial resolution and the perceptual constraints of 2D display media (computer monitors) restrict the number of icons that can be displayed at once. On the other end of the spectrum, available global-imaging techniques, such as texture-based methods that highlight the curves of a vector field to reveal flow structure and direct volume rendering, are ill-suited for depicting certain local characteristics, such as flow magnitude at a given point.
Two-sided lighted isosurfaces are used to represent an engine dataset. The new visualization technique computes multiple time surfaces from the flow around such structures to gain an accurate picture of the representative fluid dynamics.




In an effort to ease and improve the flow-visualization process, a trio of university-based researchers has developed a system that provides both a big picture of overall flow behavior and a magnifying glass into specific aspects of that big picture.

The technique, developed by Rudiger Westerman from the University of Technology Aachen, Christopher Johnson from the University of Utah, and Thomas Ertl from the University of Stuttgart, provides information about the speed, direction, and magnitude of a given flow at every point along its path. The system gathers this information by converting dense flow fields into scalar quantities, which consist of single real numbers used to measure size. In this case, size refers to the magnitude of particles "dropped" into the flow at a given time. By identifying significant areas of distortion in the streamlines used to represent the magnitude of particles at each time step (sharp curves, for instance), the researchers are able to extract pertinent information about changes in the flow structure and behavior through its evolution.

As with all flow visualization techniques, the new system creates visual representations of numerical simulations. The fundamental difference is that most existing methods attempt to analyze the complete vector field of flow data as a whole, taking into account both magnitude and directional information, while the new system repackages that information into the measurement of a single quantity-the magnitude of flow over time. This is determined by simulating the size of particles that have been introduced into the flow. When the path lines of several of these particles join, they form a flow surface. In this application, the researchers are interested in analyzing the temporal domain of such flow surfaces. That is, they are looking at the amount of time it takes particles in a flow surface to reach a given boundary. Collectively, the results are used to generate time surfaces, the positions and distortions of which provide insight into changes that are happening as the flow evolves.

To interactively display and analyze the continuous evolution of time surfaces, the researchers apply 3D texture maps to cutting planes that are parallel to the viewing plane. Hardware-supported texture interpolation is used to simulate lighting physics.
To obtain information on both the overall patterns and specific aspects of the flow field around the back of a car, a new technique converts the dense flow fields into a scalar format representing flow magnitude over time, then displays the computed volum




Once the time surfaces are visualized, the data is analyzed to extract homogenous streams in the flow-crucial information that generally cannot be determined by analyzing the vector data locally. "Even if the vector data is locally homogenous in terms of direction and speed, we will find regions where different streams proceed parallel to each other over a certain distance, but will then separate again," says Westerman.

To illustrate the concept, Johnson describes each particle moving along a path line as a container filled with liquid. "At the beginning, each container is empty. At each position in the field, the particle carries all the liquid that was injected along the path line up to this position," he says. "We try to extract the exact regions in which adjacent particles carry different amounts of liquid, because this implies that the particles have a different history in terms of what they collected along their paths." By studying the curvature of the time surfaces and identifying areas where differences occur, researchers can get a more accurate picture of the characteristics of the flow under consideration.

Once the streams have been extracted, iterative smoothing refines the curvature of the time surfaces to enhance the distinction among divergent surfaces. Minor, inconsequential deviations are removed to enhance clarity.

Next, a sparse set of particles-at least one particle per separated stream-is placed in the field to generate streamlines, which together with the time data for each vertex are used to animate the flow dynamics along the streams.

The new system does have certain drawbacks, however. For example, says Westerman, "Our approach is expensive in terms of numerical operations and storage for large-scale 3D flow fields." Also, the method fails if the data under consideration is highly turbulent with no detectable regular stream boundaries, because no homogenous streams exist.
Multiple "time surfaces" in a dynamic flow volume are represented using three-dimensional texture-based surface rendering. Regions of various thickness are illustrated using a two-sided lighting technique through which surface variations can be easily ide




Despite these deficiencies, the ability to generate multiscale flow-field representations holds significant potential. The researchers are investigating how the technique could lead to the development of a multiresolution framework for flow that would further ease the perception of distinct flow behaviors. "The multiresolution aspect means that we can do large-scale visualization very rapidly," says Westerman. "As the size of simulation data continues to grow, we think our method will be one way to still offer fast visualization that scales." Additionally, the team is considering how the multiscale nature of the analyzed flows, as well as their geometric and topologic structure, might lead to more efficient particle-tracing techniques.

Diana Phillips Mahoney is chief technology editor of Computer Graphics World.
Back to Top
Most Read