NASA Discovers Disaster Simulation Falls ShortBy David F. Carr | Posted 2005-09-01 Email Print
Re-Thinking HR: What Every CIO Needs to Know About Tomorrow's Workforce
Insulating foam that won't stay put has grounded the space shuttle. Now NASA has to rethink how it uses computer models to predict safety hazards.When the space shuttle Discovery came back to Earth on Aug. 9, the crew owed their safe return partly to luck.
After Discovery's July 26 launch, NASA was forced to admit that two years of work had not solved the problem of how to keep insulating foam from flaking off the shuttle's external tank, the cause of the Columbia disaster in 2003. Now the shuttle is grounded.
NASA predicted the tank would not shed anything larger than 0.03 pound, compared with the 1.67-pound piece of foam believed to have punched a hole in the heat shield on Columbia's wing.
But one of the four large chunks observed during Discovery's launch was about 0.9 pound30 times the predicted maximum and big enough to be deadly.
As in 111 prior successful shuttle missions, debris either missed Discovery or hit a less vulnerable area.
Computer models played a role in deciding the Discovery was safe to fly. But models are only as good as the underlying assumptions.
These assumptions can change, such as when the predictions of a computer model of foam impacts on a wing panel are verified with ballistic testingfiring chunks of foam from high-speed air guns.
Still, human judgment ultimately played a large role in weighing the results of the simulations and tests.
So how did the assessment wind up being too optimistic?
At a press conference after Discovery's launch, former astronaut Richard Covey, who co-chaired the Return to Flight Task Group, an advisory panel that reviewed NASA's preparations for the mission, dismissed the notion that there was too much reliance on computer models.
Simulations are the logical option when full-scale testing is impractical, such as trying to fit the entire external tank into a wind tunnel, Covey says.
However, seven members of the task group added a dissenting opinion to its final report, specifically critiquing NASA's use of computer models "without the attention to the interdependencies between the models necessary for a complete understanding of the end-to-end result."
To draw the line between tests and simulations, NASA had to decide how many millions of dollars of shuttle components it could deliberately ruin. In ballistic tests, the agency managed to stretch dollars by firing several shots at each wing panel, says Darwin Moon, manager of the orbiter stress group at Boeing. Computer models then extrapolated the effect of other impacts at various speeds and angles.
Now, the detailed imagery that showed where and when foam shed from Discovery's tank can help improve the next round of models. "We're still learning how foam sheds," Moon says.
Already, one fix has been suggested by a computer model of the foam created by AlphaStar Corp., working with Boeing and Lockheed Martin. The fix: spraying the foam over a fishnet structure to stop it from shedding, "or if it sheds, it sheds in smaller pieces," says AlphaStar CEO Frank Abdi.
AlphaStar has worked on several other models related to the shuttle program, including a step-by-step thermal simulation of how the heat of reentry burned through Columbia's wing.
After that accident, an investigation board skewered NASA for having grown complacent. Previous shuttles survived foam damage to heat-resistant tiles, but Columbia suffered a puncture to one of the reinforced carbon heat-shield panels used on the shuttle's nose and wings to withstand extreme heat approaching 3,000 degrees Fahrenheit.
The board said mission planners were analyzing the danger to shuttle launches based on statistics about the damage caused by foam strikes on the tiles, not the panels.
Read more about NASA's redefinition of its space mission: Should NASA Open Near-Orbit Space to Business?
Today, most shuttle impact analysis is done with Livermore Software Technology Corp.'s LS-Dyna software and "material models" of the behavior of foam, tile, heat shields and ice.
Kelly Carney, a NASA aerospace engineer, says modeling ice, water and vapor took eight months and rework of the LS-Dyna software. The payoff: The model showed ice posed a fatal risk, prompting NASA to use heated shuttle tanks.
And although the foam debris recorded on Discovery was an unpleasant surprise, "the response to it was a lot different this time," Carney says. His hope: The data recorded during Discovery's flight will allow NASA to build better models and correct problems for future flights.