NASA considered three redesigns to replace a long piece of foam insulation on the space shuttle's external tank where a chunk broke off during the launch of Discovery, according to NASA officials and records, but the space agency decided against all of them because the old design was deemed "safe to fly."

The long ridge of foam running down the side of the tank was one of several areas that got special attention after the Columbia disaster 21/2 years ago because it was in a critical area where dislodged foam could slam against the orbiter's vital heat shielding.

NASA grounded the shuttle fleet Wednesday after analysts discovered that Discovery's tank lost pieces of foam insulation similar to those that fatally damaged Columbia. The foam ripped off in an area known as a protuberance air load (PAL) ramp, a 36.5-foot-long triangular wedge that acts as a windbreak to protect the tank's cables and pressurization lines from being ripped off by air pressure in flight.

The new foam mishap, which NASA officials do not believe caused serious damage, illustrates the difficulty of eliminating every potential problem with the tank design, which has had a long history of losing foam chunks. Records show that after spending millions of dollars making repairs, some officials at the space agency had lingering doubts that they had done enough and continued to investigate potential fixes for the foam covering.

"After 21/2 years of testing, the foam shouldn't have come off. They should have done something to get that foam to stick," said Paul Czysz, a professor emeritus of aerospace and mechanical engineering at Saint Louis University. "I would have thought after 21/2 years it would have been solved."

Chunks of foam from the same part of the external tank had fallen off during two shuttle flights in 1982 and 1983, said NASA spokeswoman June Malone at the Marshall Space Flight Center in Huntsville, Ala. But NASA officials decided then that those incidents, coupled with another detailed analysis, were not enough to warrant changing the design, Malone said.

Since the Columbia accident, however, records on NASA's Web site and interviews show that redesign of that area had been under careful consideration. A fact sheet dated August 2004 said "three redesign options were under consideration for future flights" and one would be selected. But an April 2005 fact sheet said the old design of the foam wedge, used on Columbia, was acceptable.

"Based on material analysis, testing and flight data, NASA is satisfied that the current design configuration of the PAL ramp is safe to fly," the April fact sheet said.

The earlier instances of trouble with the PAL wedges occurred during a Columbia flight, launched June 27, 1982, and on a Challenger flight, launched June 18, 1983. In those cases, NASA determined the foam came off because of the configuration and repairs that had been made to the foam on that area of the tank.

Joseph W. Cuzzupoli of Kistler Aerospace Corp., who served on a task force that helped oversee NASA's redesign, said his group agreed with the agency's decision not to change the PAL ramp area design because that section was not prone to foam breakage and because NASA planned to perform more detailed inspections for tiny air pockets that might have formed when the foam was installed. Pockets, or voids, were believed to have caused the foam breakage that doomed Columbia, Cuzzupoli said, but had never been detected in the PAL ramp area.

"I think what they have learned from this incident is that there are other failure modes they should look at," he said.

In 2003 and 2004, NASA officials conducted extensive wind-tunnel tests on three possible redesigns at NASA's Glenn Research Center in Cleveland, said G. Scott Williamson, the engineer who conducted the tests. They considered eliminating the foam wedges, reducing them or building a kind of "fence" on the back edge of the rectangular tray that holds cables to redirect the wind.

Lockheed Martin, which operates the Michoud Assembly Facility in New Orleans that applies the foam to the tank, declined to answer specific questions about the foam incident.

NASA officials indicated in news releases before the launch that they were still looking closely at redesigning the area where the problem occurred Tuesday. There are two similar foam wedges on the tank -- one near the top section that holds liquid oxygen, and the other on the bottom section, which holds hydrogen. The part that ripped off was near the hydrogen tank.

On the current flight, NASA continued its testing by placing devices to measure the wind speed or pressure, called accelerometers, on the tank.

The measurements were to "be used to determine whether there is need for the ramps in future tank modifications," according to a news release before the launch.

One focus for investigators looking into Tuesday's mishap will be a 10-foot section of the PAL wedge that had been dissected to make another repair that was also designed to reduce the possibility of losing foam, Malone said. The foam was then reapplied.

The foam-shedding problem had long plagued the space shuttle and proved to be one of the most vexing issues NASA faced after the Columbia disaster, several experts noted yesterday. NASA spent millions of dollars and thousands of hours implementing dozens of changes and safeguards that included improving the fitting that connects the external tank to the orbiter; a video camera on the liquid oxygen feed line to photograph liftoff; reversed bolts on the flange of the tank's midsection and a new process for spraying thermal foam on that area; redesign of the joints for movement along the liquid oxygen feed line to the main engines; and more rigorous procedure of spraying on the foam. The foam is needed to insulate the tank and keep the fuel at extreme subzero temperatures.

"They had a lot of different issues they were working after Columbia," said George Whitesides, executive director of the National Space Society. "Foam was one they threw quite a bit of resources at, and I think to a certain extent never got fully comfortable that they had solved it."

But several members of the Return to Flight Task Group, which helped oversee NASA's response to the Columbia disaster, said that although they were surprised and disappointed by the latest developments, the agency had taken a number of steps that seemed to have addressed the issue.

"I certainly believe from what I know that they had done whatever they could to fix the problem," said Kathryn Thornton of the University of Virginia. "You can think you have a problem fixed, but until you have a chance to test it you don't really know, and this was the test."

Nevertheless, the agency clearly had failed to solve the problem and now faces the daunting task of starting the process again, several experts said.

James N. Hallock, a member of the Columbia Accident Investigation Board and aviation safety manager of the Department of Transportation, said NASA needs to figure out how foam reacts to supersonic speeds.

"Do they really, really understand what's causing the problem?" Hallock said. "I suspect they really don't understand the mechanisms the way they really need to."

"Obviously they didn't do enough," said Jerry Grey, director of science and technology policy for the American Institute of Aeronautics and Astronautics. "What they have to do now is go back and review the whole process."

An image of the space shuttle Discovery during its ascent to orbit shows a large piece of foam separated from the external fuel tank, detaching from an area called the protuberance air load (PAL) ramp. The area of missing foam is indicated by a light spot near the upper edge of the tank.