Research Feature
August 11, 2004
View Animation (410KB MOV)
The terrible loss of the space shuttle Columbia posed great challenges to NASA and to the nation's engineering community. To meet the challenges, the agency assembled engineering teams composed of NASA experts as well as industry, national laboratory, and university colleagues, to search for solutions to some difficult problems. Could their collective engineering skills keep better score in the battle between chance and causality? One engineering team member was Eric P. Fahrenthold of The University of Texas at Austin (UT Austin).
Fahrenthold and his group in the Mechanical Engineering department have long studied problems of hypervelocity impact to help NASA design shielding for orbiting spacecraft. Above the earth's atmosphere, the impact of a tiny, pebble-sized object at a speed of ten or eleven kilometers per second can be fatal for spacecraft with insufficient shielding.
"Our modeling of these situations has meant that we had to find a way to cross the usual boundary between continuum mechanics, the study of material response to stresses and strains, and discrete system dynamics, which deals with what happens where the stresses and strains have become too great for material integrity," Fahrenthold says.
About five years ago, Fahrenthold designed a hybrid computational code, EXOS, that uses both finite elements and a particle representation of the masses within those elements. "EXOS is computationally intensive, because it carries a lot of information through many calculations at each time step," Fahrenthold says, "but it is well suited to a particular class of problems." A typical EXOS simulation might take several days on Longhorn at UT Austin's Texas Advanced Computing Center (TACC) or one of the comparable NASA supercomputers.
Eric L. Christiansen, Fahrenthold's program manager at the NASA Johnson Space Center in Houston, knows that EXOS can handle challenges. He calls it a valuable addition to NASA's own arsenal of codes. "EXOS is fully three-dimensional, solves our problems very nicely, works on a wide range of problems where other codes crash, and is unique in its wide applicability to our goal of making orbiting vehicles safer," Christiansen says.
After the shuttle Columbia disintegrated on re-entry in February 2003, an Accident Investigation Board convened to analyze the tragedy. The board concluded that the most likely cause for the loss of the orbiter was an impact on the leading edge of the shuttle's wing. The impactor, a block of foam insulation, was dislodged from the rocket booster on liftoff, apparently hitting the wing edge with sufficient force to create the damage that proved fatal on re-entry.
Christiansen and his colleagues called on NASA, industry, national laboratory, and university engineering groups to apply their modeling and simulation skills to reconstruct the impact in detail. Many of the NASA researchers used the well-known finite-element code DYNA-3D, but Christiansen wanted to know how EXOS could treat the same problem. "Because this was a comparatively low-velocity impact, and because we had never modeled the skin of the space shuttle, it took us about six months after the accident to get our codes ready to analyze the incident," Fahrenthold says. "The availability of computer time at TACC and the NASA centers enabled us to make timely comparisons."
The space shuttle's thermal protection system consists of a combination of ceramic tiles and reinforced carbon-carbon (RCC) panels, with the latter molded to form the aerodynamic leading edge of the wing. Fahrenthold conducted a series of simulations of the impact of a one- to two-pound hexahedral block of foam, coming in at various angles and hitting the underwing tiles or the RCC panels. The models employed one to two million particles and the several-day-long simulations covered five to six milliseconds of real time. While the particular cases that were run, even with a heavier projectile, had little impact on the ceramic tiles, the RCC panels failed readily or developed giant cracks under all postulated impacts.
Fahrenthold's initial simulations agreed well with the limited experimental data then available, and he is continuing to simulate RCC impacts with better material models. Now that he has begun using Lonestar, the 856-processor, 5.3-teraflop Cray-Dell Linux cluster at TACC, he can also use more particles, obtaining what is effectively higher resolution.
"The hybrid method in EXOS combines the general contact-impact modeling capabilities of particle methods with a true description of material strength effects provided by the finite-element techniques," Fahrenthold says. Purely particulate representations of material dynamics are subject to instabilities when dealing with materials under tension, while finite-element representations cannot cope with material failure--the elements that fail simply vanish from the calculations. "We use both particles and elements throughout the calculation," Fahrenthold says, "but each represents distinct physical effects. Where the finite elements collapse, particles not associated with any intact element are free to flow under the general physical constraints of the model."
The entire NASA multisite team met about a year ago to share their initial results, and Fahrenthold and graduate student Young-Keun Park will publish their findings in a upcoming issue of the American Institute of Aeronautics and Astronautics Journal of Spacecraft and Rockets. In a parallel and more publicized effort, NASA scientists conducted lab and field experiments, still ongoing, to test against the simulations.
Fahrenthold and Park will continue, using EXOS on Lonestar, to model space shuttle impacts and also the effects of introducing new, multilayered designs for the RCC panels. "What we learned has also given us more insight into our normal work of studying hypervelocity impacts on orbiting spacecraft," Fahrenthold says, "and we've made our code more versatile and nimble in the process. TACC has provided exemplary support for all of our efforts."
