by Jill R. Aitoro

MISSISSIPPI STATE, Miss. - All matter, whether Silly Putty or the steel frame of an automobile, is made up of atoms. But it's the state of matter determined by the strength of the bonds between atoms - i.e. solid, liquid, gas or plasma - that largely determines how something will respond to exerted forces. That explains why Silly Putty thrown against a brick wall doesn't cause the same damage as a car crash, regardless of how powerfully it's launched.

Researchers at Mississippi State University keep that in mind as they develop a methodology for simulating car crashes at the atomistic level. Advanced visualization software is used by the researchers to identify crash variables and produce animations showing their impact.

The Transition to "Virtual Manufacturing"

The computerized crash simulations are conducted as part of the university's Center for Advanced Vehicular Systems (CAVS). CAVS was established in 2001 by the State of Mississippi to enhance its relationship with the automotive community - most notably Nissan, which manufactures a large chunk of its vans and trucks in the state. CAVS research aims at discovering manufacturing and design methods for producing technologically advanced vehicles at reduced costs.

A key aspect of the research involves designing and analyzing "virtual products" and "virtual manufacturing operations," according to Mark F. Horstemeyer, CAVS chair in solid mechanics and professor of mechanical engineering at Mississippi State.

It All Starts with the Atom

Atomistic-level simulations are central to CAVS' efforts to create more accurate and efficient virtual models.

"Using atomistic-level simulations to determine mechanics variables is a recent area of research that fits into the bigger picture of multi-scale modeling," Horstemeyer says.

Researchers take the results of mechanism behavior from atomistic simulations and incorporate them into the properties for higher-scale finite element analysis (FEA) of automobile parts. Starting at the smallest common denominator, the atom, allows researchers to better simulate the manufacturing and design process in which a physical model is built and tested. The atomistic simulations help determine the microstructure-property relations for the FEA.

The process is time-consuming. CAVS researchers have to build atom upon atom to reach any substantial amount of material.

"Getting up to two microns in size requires millions of atoms," Horstemeyer says.

CAVS researchers rely on parallel computing to handle their extensive data sets. A Sun workstation accesses computing power from Sun, Hewlett-Packard and IBM server clusters located at the university.

"When running parallel computing, you have to be very careful about how you use your code," Horstemeyer says. "The larger the code the harder it is to manipulate and use. There are a lot of details to pay attention to."

In atomistic simulations, the individual atom acts like the node of the computational fluid dynamics' (CFD) world.

"But unlike CFD, we have no elements in the atomistic code," Horstemeyer says. "We essentially just take one atom at a time and add new atoms to build the material structure. Then we load in tension, compression, torsion, or the combination of these loads on the atomic structure and watch what happens."

The CAVS' approach relies a great deal on the Embedded-Atom Method (EAM), which describes the bonding of an atom in terms of the local electronic density and incorporates electrostatic and repulsive interactions.

Animating the Damage

Horstemeyer used the EAM formula as well as his own algorithms to develop a mathematical tool to calculate such crash variables as stress, strain, centrosymmetry parameter, deformation gradient, and grain morphology. Those calculations are then loaded into CEI's EnSight Gold (www.ensight.com) visualization software.

"Variables are developed to analyze the atomistic simulations, and EnSight allows those variables to be viewed," Horstemeyer says. The visualizations enable researchers to watch and analyze the mechanisms before adding the atomistic simulation information from a given region into a finite element mesh.

"Sometimes we put in a defect and watch how it will initiate failure upon deformation," Horstemeyer says. "Showing animations of the damage progression is very helpful in developing mathematical functions at the higher scales. And because EnSight functions the same regardless of hardware or the number of processors, it's very robust in a parallel computing environment."

The process of microstructure-property relations will some day allow researchers to run complete crash simulations with accuracy to the smallest molecular detail. The atomistic simulations fit into higher scale FEA models of car parts, which eventually will fit together like pieces of a jigsaw puzzle to form a complete automobile. So far, CAVS has completed an FEA model for one car part - the control arm that connects to the axle and shock absorbers.

Although some might think the CAVS approach is excessive, Horstemeyer believes that the group's work is essential for more accurate FEA.

"Running FEA simulations without the microstructure-property relations - the focus of the atomistic-level simulations - is deceiving because the physical results could be wrong," Horstemeyer says. "It's like analyzing a plane crash with the plane made out of jelly."

###