Team pounds pavement
with concrete calculations

(page 2 of 3)

Concrete and its vicissitudes

These studies are particularly challenging because concrete contains particles of many sizes. Cement particles are made from a crushed combination of limestone, shells, chalk, clay and slate. The smallest cement particles can be submicron, while the largest in the full concrete mixture, which includes rocks and gravel, can be roughly 100,000 times that size &ndash a few centimeters in diameter. With such a span in scales, tracking the behavior of both types of particles at the same time without a supercomputer would be impossible, much like following individual football fans from a blimp above a packed stadium. Modeling such a large range of particle sizes makes parallel computing a necessity.

To address particle variability, researchers need a large modeling space, and they must set up a multiscale system that characterizes the chunky-ice-cream-like mixture in several stages. Imagine concrete is a scoop of rocky road: Modelers first must understand the forces at work in the plain chocolate ice cream before they mix in larger components &ndash the concrete equivalents of marshmallows or nuts.

In concrete, the first ingredient is cement &ndash tiny calcium silicate particles mixed with water. This wet cement mixture serves as the fluid in which sand &ndash the next particle size up &ndash is simulated to yield the properties of mortar. Finally, the modelers add rocks to the mortar for a complete picture of concrete. Running on the Blue Gene/P computer at DOE's Argonne National Laboratory near Chicago, they can model the interactions of up to 100,000 "suspended groupings" &ndash the large particles &ndash in the mixture and up to 1 million fluid (small) particles.

The team's computations are examining the details of how those suspended groupings interact. To describe such a complicated mixture, the researchers use algorithms based on dissipative particle dynamics (DPD). Similar to molecular dynamics (MD), another set of algorithms used for molecular interaction simulations, DPD allows researchers to model the motion of complex mixtures, such as cement or mortar, as "particles" that interact with larger components such as rocks.

The team's main interests are viscosity and yield stress. Because concrete is a non-Newtonian fluid, the viscosity (thickness of its flow) varies with the relative motions of the particles that come into contact as it moves. Yield stress refers to the amount of force required for a material to flow.

At each stage of the calculation, the researchers must validate their findings against similar theories or, more often, by experimenting with concrete, NIST computer scientist Bill George says. They want to ensure that simulation data matches what people actually see in concrete.

Although some validation data comes from experiments their group conducts, the research also is supported by the Virtual Cement and Concrete Testing Laboratory. The laboratory is a consortium that works with NIST groups, various concrete companies and associations and government agencies such as the Federal Highway Administration. These partners can test what the NIST researchers have learned with specific concrete mixes. The fundamental computational models from NIST help the concrete industry understand the rheological parameters &ndash viscosity and shear stress &ndash their equipment experiences as concrete is mixed, poured and pumped.

Diving into concrete

Exploratory visualization is another critical part of model validation. The images NIST physicists, computer scientists and visualization experts create allow researchers to "stick their heads in a bucket of concrete," says Steve Satterfield, who helped develop the institute's immersive visualization software platform.

This virtual concrete bucket reconstructs their computational results in a three-dimensional space &ndash an "immersive visualization environment," with walls and screens that form a corner and a floor. Researchers wear virtual reality goggles that process stereo images projected on the walls. Like other virtual reality technology, the goggles contain a transmitter that tells the computer where the observer is and where he or she is looking.

« Previous       1   |   2   |   3   |   Print       Next »