The next time you sit through a meeting shivering or gasping breaths of stagnant air, think of Ping-Pong balls. Imagine your room filled with these small spheres riding along on the currents of air, swooping down along the walls, or swirling about the vortices of whirlpools. Watch them as they collide and are propelled along paths that become more chaotic with each brief encounter. In their restless motion, you will find a distraction from your discomfort as well as a glimpse of its causes.
In NCSA's virtual environment CAVE (Cave Automatic Virtual Environment), researchers are making Ping-Pong ball-like spheres circulate about a 3D model of a conference room. They do so not for entertainment but as part of a project to eliminate drafts, dead air pockets-even annoying hums. Their goal is to produce software that engineers and designers can use to diagnose circulation problems as well as to remedy and prevent them. Eventually, they will have a tool for designing conference rooms that are ideal from the start.
"Architects and consultants don't have access to this technology," says William ("Tilt") Thompkins, assistant director of research for Mechatronics at United Technologies Research Center (UTRC), which is funding the flow modeling portion of the project. "We want to be able to quickly custom-design conference and auditorium areas in which both physical comfort and acoustics are important," he explains.
NCSA industrial partner UTRC has good reasons to care about comfort. As owners of Carrier Corporation, one of the world's largest manufacturers of commercial and residential heating, ventilating, and cooling systems, it wants to be able to offer these capabilities to its customers. Discovering the tools for making a difference-quickly-led them to NCSA.
"NCSA has a major thrust in computational analysis and visualization," says Thompkins. "What I'm doing with this project is exploring how far we can push both of those technologies and how well we can integrate them. This project is a one-of-a-kind effort to find out how far these disparate technologies can be combined and integrated."
United Technologies initiated the conference room project after learning of a new computing code at NCSA called IDLEST that could predict flows more accurately without overwhelming the company's computational resources. IDLEST (Incompressible Direct and Large Eddy Simulation of Turbulence) was developed by NCSA Research Scientist Danesh Tafti, team leader of the Center's Computational Fluid Dynamics Group and an adjunct assistant professor in the UIUC Department of Mechanical and Industrial Engineering, for simulating a variety of fluid flow fields.
Though few of us notice air's movement unless it whips up into a wind or subsides into stillness, its flow follows complex paths. The complexity stems from a phenomenon called turbulence.
"Up to a certain point, flow is ordered and predictable," explains Tafti. "But then, a smooth flow will develop instabilities that rapidly transition into turbulence. Turbulent flow is random and chaotic."
Within a span of only a few millimeters, small, random fluctuations in a flow can spawn large vortices of turbulence. Think of what happens to smoke from a cigarette. It trails smoothly upwards for a little while, then suddenly bends and swirls in snakelike writhing.
To predict precisely air's tortuous path, engineers would have to resolve a flow on a computational grid with zones fine enough to capture the minute fluctuations. For the UTRC conference room, that would mean solving the partial differential equations that govern fluid flow (called the Navier-Stokes equations) at potentially billions of grid points-an undertaking that is computationally prohibitive.
Instead, for most complex flows, engineers and designers rely on computing codes that average the effects of turbulence on momentum, mass, and heat transfer, then lump these length- and time-dependent scales into one variable (a process called Reynolds stress averaging). Tafti likens the effect to taking all the frames of a motion picture and superimposing them to obtain a single frame that encapsulates the entire movie. Averaging may simplify computations, but it reduces their accuracy and precludes much of the important flow physics. This is true especially when flows are predominantly turbulent-as in conference rooms where air is blown into the room and must negotiate among numerous people, chairs, and tables with the flow patterns constantly evolving.
The approach Tafti is adapting in IDLEST falls between the two extremes. Called a large eddy simulation, IDLEST solves the Navier-Stokes equations but without the degree of detail that will deadlock a computer system. Large eddy simulations, as the name implies, resolve the large, energy-containing scales in a flow field and filter out the smaller scales. Because the activity on small scales influences the flow, Tafti and his colleague Fady Najjar, NSF Division of Advanced Scientific Computing Postdoctoral Fellow, model the effects of the small scales on the system. Preserving the effects of the smaller scales without adding computational complexity, power, and cost is the essence of large eddy simulations.
"We eliminate the small scales, but then we model their effect on the large scales by using information from the large, or resolved, scales," says Tafti.
Even with its trimmed down approach to calculating flows, IDLEST still puts out a lot of data. For the conference room simulation, the data output by IDLEST involves thousands of frames, each containing about 80 megabytes of data. Some of these data are saved for visual postprocessing.
Here's where the Ping-Pong balls come in.
Doug Britton, UIUC undergraduate in computer science, and Derek Storr, UIUC senior in industrial design, receive from Tafti the dimensions of the conference room, the placement of vents and furniture, and data on the location of the particles as well as their temperatures. Using a graphics interface program, Britton transforms this raw data into a format that can be viewed on Silicon Graphics' workstations or in the CAVE. For presentations, the images are enhanced to more realistically depict a conference room. Storr uses Wavefront to make the table appear wooden and the walls screened. In both instances, the results are stunning.
The simulation begins with the Ping-Pong ball-like particles aligned along the vents on both sides of the room. The particles begin a slow descent, with each line closely mirroring the other. Though their movement is slow, the particles (when color coded for temperature) rapidly change from purple to red, indicating a rise in temperature as they encounter the warm air from the room. The particles cascade downwards as most dive under the simulated chairs and begin swooping up towards the stream of air coming from the other side of the room. As two air streams from either side of the room collide, the particles scatter and ricochet chaotically. As time passes, some particles come under the influence of the exhaust vents and are drawn out of the room.
From this brief simulation, Tafti can draw several conclusions about the design of the room and the arrangement of the vents and furniture (all of which seem good). His goal, however, is to streamline the process so designers can experiment with different heating and cooling scenarios. He and his colleagues also are working on making the CAVE interface more interactive.
"We will use the stored database from the simulation to calculate trajectories of particles injected in the CAVE at arbitrary locations and times," says Tafti. "Later, we will move from representing the flow field with particles to actually modeling the field."
To cancel a sound, the researchers first mechanically listen to a room and identify those frequencies associated with undesirable sounds. Once these frequencies have been isolated, they then generate signals of equal frequency but opposite phase and project them into the room.
"By adding another sound . . . it's like having so many air molecules coming at you. You want to send exactly the same number of air molecules back in the same direction so they meet dead on, and you don't hear anything. That's cancellation," says Robin Bargar, a composer who directs the audio group that is part of NCSA's Virtual Environment Group.
Ideally, cancellation is done at every place in the room for every listener. Instead, microphones are put in several places in the room to get an acoustical overview of the room. A generalized signal is then fed back into the room to cancel the bothersome noise.
So far, the Audio Development Group has simulated multiple sources of sounds in the CAVE with independent localization of each source without relying on headphones-a first in a virtual environment. Next, they will move from simulated cues for distance and direction to numerically modeled cues. A long-range goal is to compute the characteristics of the room and model how sounds reverberate in the space-for example, wood versus marble.
As with the circulation model, the sound model will yield interactive software that can identify problems before rooms are built. Thompkins (UTRC) wants to put engineers, architects, and building designers in the CAVE and show them why a design will not work. "There are no obvious reasons why noise and drafts have to be tolerated," says Thompkins, who predicts UTRC will use this technology to design heating and cooling systems in three years.
In a few meeting rooms, some lucky people won't be seeing Ping- Pong balls.
Holly Korab is a science writer in the Publications Group.
