Researchers at ORNL, in collaboration with the National Highway Traffic Safety Administration and George Washington University, are developing detailed computer models of a variety of vehicles. In the last few years, they have completed models of the Ford Taurus and Explorer, both among the top sellers in the United States. Researchers are now modeling an Audi A8, an all-aluminum car that is one of the first to use a lightweight material that may be prominent in future cars.
The shell of the Audi was on display at the recent "Supercomputing '98: High Performance Networking and Computing" conference in Orlando, Fla.
"We use these models combined with lightweight materials models to analyze material performance in a wide variety of crashes," said Srdan Simunovic of ORNL's Computer Science and Mathematics Division. "We can substitute different materials in individual parts and compare the results."
Researchers develop models by disassembling a vehicle, scanning the shape and measuring the weight and inertia of each component. They program those data into a computer and perform a number of simulated crashes, all at a fraction of what it would cost to perform actual crashes.
Results gained from these and other tests will allow the U.S. government to assess safety as U.S. automakers strive to triple the efficiency of today's cars without sacrificing performance, utility, cost of ownership or safety. To accomplish these goals, experts estimate that the weight of automobiles will have to be reduced by 40 percent.
By using computer models, which are validated with a controlled crash, researchers can gain information identical to that from actual crashes that cost up to $75,000 per crash. Researchers compare simulations using high-speed films of collisions and traces from accelerometers that are placed throughout the vehicles. They also disassemble and analyze parts of the car after the crash.
ORNL's capabilities in materials modeling and parallel computing make the lab uniquely qualified to perform this work, Simunovic said. He points to the lab's development of the parametric finite element model as an example of innovative solutions.
"With parametric finite element models, we can tune the grid - in which the vehicle is divided into hundreds of small sections - according to the kind of crash we're going to simulate and kind of computer resources available," Simunovic said. "This makes it more manageable for the computer."
At the conference in Orlando, researchers demonstrated their work in a variety of areas, including in spark emissions, improved catalytic converters and developments to enhance engine efficiency.
"We have a number of disciplines that are coming together to help in the design of a better automobile," said Thomas Zacharia, director of the Computer Science and Mathematics Division. "And we're demonstrating how supercomputing has relevance to people's lives."
ORNL is a DOE multiprogram research facility and is managed by Lockheed Martin Energy Research Corporation.
.