Looking into the Vortex

By Herbert Morgan, NCSA

When the F-18 fighter first came out, it was having problems with its twin vertical tails. They cracked. At a high angle of attack there were two vortices that came off the leading-edge extension and impinged on those twin vertical tails causing them to vibrate. The ensuing stress caused metal fatigue, which in turn resulted in the cracks. The ability to predict vortex cores such as these is an important development in engineering flows.

David Semeraro, senior research scientist at NCSA's Visualization and Virtual Environments Division, has been using feature-extraction techniques that were developed for engineering fluid dynamics to find features in tornadic thunderstorm simulations.

The idea behind Semeraro's work is to use mathematical-based vector topology techniques to locate vortex cores, or the centers of swirling flow. As in the example of the F-18, wingtip vortices or the trailing vortices that emanate from wingtips of large airplanes can be quite strong, which is one reason air-traffic-control regulations require minimum distances between landing airplanes.

"You don't want to get buffeted by these strong vortices coming off the wingtips," says Severaro. "If you can identify computationally where these vortices are, and their strength, you can use this information to design better systems."

Vortex cores also occur naturally in tornados and hurricanes. Enter Bob Wilhelmson, senior research scientist at NCSA's Application Technologies, whose interest lies in predicting tornados.

Semeraro approached Wilhelmson a few years ago and asked him if he knew of Robert Haimes' (a principal research engineer at MIT) work in computational techniques for identifying centers of swirling flows of vortex cores. While familiar with it, Wilhelmson had never really tried it.

Semeraro says, "I had always been intrigued to find out whether that [Haimes' algorithm] would be an effective means in identifying where in the thunderstorm vortex cores form." The techniques used by Semeraro are automatic: they find as many vortices as exist in the flow field and define them by a line that traces the core at the center of rotation. Semeraro has been parallelizing this algorithm and trying to incorporate some of the software into an application called ParaView, a parallel visualization package.

Semeraro implemented Haimes' algorithm in a serial program and incorporated different datasets supplied by Wilhelmson and Leigh Orf, assistant professor of Atmospheric Science in the Geography Department of Central Michigan University. The algorithm was very efficient at finding even the weakest rotations.

The algorithm is also cell-based; that is, it does its calculations on a cell-by-cell basis, making it, in Semeraro's words, "embarrassingly parallel." It can be dissected into many smaller cells and efficiently distributed on a distributive memory system.

The algorithm goes through the flow field and the output is geometry—line segments that indicate where the vortex cores are in three-dimensional space. That information is fed into a graphics program, which draws in 3-D. To date, Semeraro has analyzed an entire time-series run, consisting of 1600 time steps from which he has generated a couple of animations.

One piece in progress is the distributive-memory parallel version of the program that generates the geometry. A serial version works, and a cluster-based version is being developed, which would allow researchers to run batch-processed jobs to extract the geometry from large numbers of datasets. "We are also trying to build a module," says Semeraro, "that can be plugged into ParaView, which would allow people to do this interactively."

Atmosphere. Ocean. Perhaps even electromagnetic fields. This research has the potential to shed light on the physics of many fields of science.