On the Leading Edge

SwRI ballistics tests help investigators determine the cause of Columbia loss

By James D. Walker, Ph.D., and Donald J. Grosch     image of PDF button

"... I'm sure that Columbia, which had traveled millions of miles and made that fiery re-entry 27 times before, struggled mightily in those last moments to bring her crew home safely once again. She wasn't successful ..." Robert Crippen at the Columbia Memorial Service, who with John Young flew Columbia on the first space shuttle flight into orbit in April 1981.

Donald Grosch, left, is manager of the Ballistics and Explosives Range for the Engineering Dynamics Department of the Mechanical and Materials Engineering Division. He manages numerous technical programs involving ballistic testing and armor development and testing. Dr. James Walker, right, a staff scientist in the Computational Mechanics Section of the Engineering Dynamics Department, focuses on impact physics and how projectiles go through armors. He has analytically modeled the penetration of projectiles into metals, ceramics and fabrics, and the Walker-Anderson model is in fairly widespread use for these problems. Walker's interests include the mechanical response of systems and materials, and his work includes and combines large scale numerical simulations, analytical techniques and experiments.

On Jan. 16, 2003, 81.7 seconds into ascent, foam separated from the bipod ramp of the external tank (A) and impacted Columbia's wing leading edge (B). At the time the foam broke away, the shuttle was traveling 1,568 mph (Mach 2.46) and was at an altitude of 65,900 feet. Based on film evidence, the foam traveled the 58-foot distance in 0.16 seconds. Assuming constant acceleration, a simple calculation shows an impact velocity of 494 miles per hour, very close to the more refined 528 mph used during the impact tests. The thermal protection system comprises silica tiles on most of the orbiter body, including thousands of tiles on the bottom surface of the wing, and 22 reinforced carbon-carbon (RCC) panels on each wing's leading edge. Early in the investigation, the team believed the impact occurred near the left main landing gear door; later analysis indicated an impact near RCC panels 5 to 9. The final conclusion of the investigation was that the foam impacted RCC panel 8.

The space shuttle orbiter Columbia was lost during re-entry on Saturday, Feb. 1, 2003. Researchers at Southwest Research Institute (SwRI) were contacted two days later to begin their role in the accident investigation. SwRI had been involved in impact experiments on the orbiter thermal protection system since the early 1980s. These impacts were with various materials, including foam insulation, ablator and ice, impacting carbon-carbon materials and the silica thermal protective tiles that cover most of the orbiter. Most recently, in 1999, SwRI had performed an experimental program of foam insulation impacts into tiles over a wide range of velocities and impact angles. Thus, SwRI had the experience and expertise to carry out an impact investigation to determine whether the foam impact observed on ascent could have played a role in the loss of Columbia.

Tests performed at the SwRI Ballistics Range used the SwRI-designed large compressed-gas gun to launch foam material. The blue chamber was filled with compressed nitrogen, and the foam was launched down the rectangular barrel towards the wing leading edge test structure. Lights provided extra illumination needed by the high frame-rate cameras, some running up to 7,000 frames per second (as a comparison, film for motion pictures is shot at 24 frames per second and television is essentially 30 frames per second).

Week One: Capability Demonstration

During the accident investigation teleconferences of the first week, in which the authors participated, much of the discussion dealt with the size, shape, mass and velocity of a piece of foam that was seen in launch film and video. A NASA Accident Investigation Team (NAIT) pursued the investigation and analyzed the film and video footage in great detail. The NAIT worked in conjunction with the Columbia Accident Investigation Board (CAIB). Early estimates of the size of the foam would prove to be quite accurate after months of detailed analysis. However, there was a clear difference from testing that SwRI had previously performed: The volume of the foam that impacted Columbia on ascent was at least 400 times greater than any foam projectiles launched in previous SwRI tests.

Researchers knew it would be extremely difficult to launch foam insulation in a traditional manner, where research projectiles are launched in a sabot, or launch package, that fits within the gun barrel and then separates from the projectile as it leaves the barrel (see "Ballistics Research and Computational Physics," by James D. Walker, in the September 1993 issue of Technology Today). Typically, sabot separation is accomplished through air resistance. However, the foam insulation material is so light - around 0.038 grams per cubic centimeter (2.4 pounds per cubic foot) - that it is extremely difficult to get a separation of the sabot from the foam impactor.

Five impact tests were conducted against thermal tiles attached to Enterprise's left main landing gear door. Impacts showed minimal damage to the tiles.

SwRI's approach had been to construct a barrel for each impactor design, where the cross section of the barrel exactly fit the cross section of the foam impactor, removing the need for a sabot. This technique was adopted in the Columbia accident investigation. Because the size of the foam was only approximately known, to save time and cost the team selected a barrel cross section matching that of standard, off-the-shelf, rectangular structural steel tubing. The tests involved a tube with a cross section of 5.5 inches by 11.5 inches.

Five days after the loss of the Columbia, SwRI had fabricated the barrel and used its large compressed-gas gun to launch a large piece of foam the size of the foam in question, at velocities of interest to the accident investigation team.

Months One and Two: Preparations

During the next two months the team improved launch techniques. Elsewhere, the investigation team was determining the size of the foam impactor and its velocity. In addition to detailed film analysis, NASA began large-scale computational fluid dynamics studies to examine the flight of the foam from the bipod ramp area (near where the front of the orbiter attaches to the external tank) to the left wing. These computations, along with the visual evidence, were used to determine the impact conditions for the foam.

Computational modeling revealed the stress levels achieved over time during simulated strikes of thermal tile by foam insulation material.

It became clear to the team that analysis was to be an important part of the investigation, and NAIT set up computational and analysis teams to study the impact. In response, CAIB contracted with SwRI to support the investigation with impact modeling to complement the extensive modeling work by NAIT. In particular, the CAIB intended to support an independent analysis of the impact event. One reason was that, in the event the results of the investigation relied heavily on analysis, they wanted an independent verification of the computational results.

Early telemetry from Columbia during its descent showed a temperature rise in the left wheel well. Because of this, plus the fact that no images directly showed the location of the foam impact, early focus was on the thermal tiles in the vicinity of the left main landing gear door. The team removed a corresponding door from the Enterprise, currently in storage by the Smithsonian Air and Space Museum. The Enterprise had been used only for atmospheric flight tests (it was dropped from a modified 747) and never had re-entered from orbit, so it did not have a thermal protection system. Shuttle technicians placed thermal tiles on the door as they would on an operating orbiter. The thermal tiles in the vicinity of the landing gear door are very light, with a formal density of 0.15 grams per cubic centimeter (9 pounds per cubic foot).

Personnel from NASA, Boeing, Lockheed and others gathered at SwRI to assemble instrumentation for the impact targets. More data recorders and high-speed cameras than are typically used in such tests were acquired so that images of the impact could be taken from many angles. During this time, impacts were conducted against aluminum plates to test the strain gages, load transducers and cameras.

The hydrocode CTH was modified to allow accurate modeling of both the foam insulation and the thermal tile. These images from a simulation (at 0, 1, 2 and 5 milliseconds) show no damage, as was observed in the corresponding test. The normal stresses vs. time during the impact are displayed for various locations along the tile surface: The stresses are below the 50 pounds per square inch (345 kilopascals) required to begin crushing the thermal tile.

Months Three and Four: Thermal Tiles

When the left main landing gear door was ready, the SwRI team carried out foam impacts using a projectile of 5.5 inches by 11.5 inches by 19 inches, weighing 1.67 pounds. Lockheed Martin had prepared the foam insulation the same way it is prepared for the shuttle external fuel tank. It was then machined to fit the barrel. The impact velocity was 775 feet per second (528 mph) with an impact angle of 5 to 10 degrees for the underside of the wing where the landing gear door was located.

Concurrently, SwRI developed detailed analytic and numerical models of foam insulation impact on thermal tiles. The models provided a damage-no damage transition curve in the impact speed-impact angle plane. In particular, tests showed that the component of the foam impactor velocity that is normal (perpendicular) to the impact surface determined whether there was tile damage. Given an impact speed and angle for an incoming piece of foam insulation, the model determined whether tile material would be damaged. The SwRI model agreed extremely well with previous tile impact tests and would agree with the tests performed during the Columbia accident investigation of foam insulation impacting tiles. The models were used not only to study the impact of the foam on tile as occurred in the tests, but also to explore the role of rotation of the foam insulation as it tumbled toward the wing and to determine how the test might be adjusted to reflect the rotation of the foam.

The team performed five impact tests against the left main landing gear door, with impact angles from 5 to 13 degrees and speeds from 717 to 827 feet per second (489 to 564 mph). Both the tests and the computations indicated that an impact on the tiles beneath the wing would not lead to extensive tile damage, and most likely did not cause the loss of the orbiter. The models showed that, given the speed of the foam, at least a 17-degree impact angle would be required to extensively damage tiles. Following the tests SwRI performed nondestructive evaluation of the tiles, and would later similarly examine the reinforced carbon-carbon (RCC) panels, to determine the extent and type of damage caused by the impact. In addition, three-dimensional optical scans provided detailed post-test geometries of tile gouging.

Impact on RCC panel 6. The interior rib fractured.

The impact against RCC panel 6 caused an internal broken rib, but minimal damage at the face of the panel. White tabs with yellow wires are strain gages. Numbered black dots are calibration markers for cameras. The fracture can be seen to the right of the apex (area circled in red).

During this time, two developments led interest away from the thermal tiles on the underside of the wing and towards the left leading edge. First, Columbia's modular auxiliary data system recorder was found containing data, recorded from hundreds of sensors, that had not been transmitted to the ground. It became clear that the earliest indications of thermal problems were from the leading edge and not the left landing gear wheel well. Meanwhile, the groups that were analyzing the foam trajectory were concluding that the foam must have hit very near the leading edge.

Months Four, Five and Six: Leading Edge

As the focus shifted to the leading edge, a new test apparatus was built to reflect the structure of the wing and to hold the leading edge panels as the orbiter would hold them. RCC, a very different material than the light silica foam tiles that line most of the orbiter's body, makes up the wing's leading edge. The RCC is denser (1.6 grams per cubic centimeter) than the tiles and is brittle. A series of 22 RCC panels lines the leading edge of the left wing. The simulated wing structure held panels 5 through 10.

SwRI developed an analytic model that demarcated the damage-no damage region of the plane: for impact speed and angle above the curve, tile damage occurs. For impact speed and angle below the curve, no damage occurs. This model (blue curve), agreed extremely well with previous test data generated at SwRI (black data points), computational results (blue data points) and with the impact tests performed during the investigation (red data points). The model and tests showed that an impact on the lower surface of the wing would not cause extensive tile damage and was not the cause of the accident.

Two tests were performed using Enterprise's non-RCC, fiberglass panels that had been instrumented. Diagnostics were worked out and comparisons were made with analysis. Because the leading edge is hollow, high-speed cameras were placed inside to measure any deflection of the panels during impact. These tests used up to 15 high-speed cameras, with frame rates up to 7,000 frames per second. Up to 240 channels of strain gage, accelerometer and load cell data were recorded for each test.

Following the fiberglass panel tests, an RCC panel 6 that had flown 30 missions on Discovery was impacted by foam insulation at 768 ft/s (524 mph) and an angle of around 20 degrees. The curvature of the panels makes it difficult to assign a single impact angle: The impact angle changes because the foam impacts the leading edge over a large area. The impact broke the interior rib of panel 6, and the crack ran so that it could (barely) be seen on the exterior leading edge. The team thought this amount of damage, unless the crack grew during re-entry, was insufficient to lead to a loss of the shuttle.

SwRI developed a code to analyze the impact of the foam on the RCC panels. These four frames are from the analysis of the impact into RCC panel 8, with the images at 0.1, 1.2 and 2.5 milliseconds (a front and back view at this later time). Failure of the panel occurred in roughly 2.5 milliseconds.

The reconstruction of the final minutes of Columbia, both through the data recorders and the recovered pieces, now focused on RCC panel 8 as the most likely failure site. Thus, the next tests were performed against panel 8 - first using one of Enterprise's fiberglass panels, then an RCC panel 8 which had flown 26 missions on Atlantis. A 1.67-pound piece of foam insulation struck the RCC panel 8 at 777 ft/s (530 mph) and about 20 degrees. The barrel had been rotated 30 degrees to increase the load delivered to the panel. This time the impact resulted in a large hole in the panel, some 16 inches by 16 inches. Thermal analysis performed by the NAIT indicated that a hole 10 inches across would likely bring the orbiter down and would be consistent with the sensor data on Columbia's last flight.

As had occurred with the tiles, SwRI modeled the impact of foam insulation on RCC panels. A numerical model simulated the panel, and an analytic boundary condition simulated the pressure load from the impacting foam. Once again, central to the load delivered and the stresses calculated is the normal component of the foam impact velocity. Comparison with the two tests performed against RCC panels led to estimates of failure stresses within the panel material. Parametric studies performed with the model determined the impact location that led to the most extreme stresses in the rib and in the panel. Other computations investigated the effect of foam impactors with rotational velocity. It was shown that a nonzero rotation velocity for the foam impactor nearly always increased the stresses on both the panel face and the rib of the panel.

Impact on RCC panel 8 resulted in a hole.

The impact against RCC panel 8 produced a hole in the RCC panel roughly 16 inches by 16 inches. NAIT analysts estimated that a hole 10 inches across could have caused a loss of the orbiter on re-entry.


With the Columbia accident investigation concluding, SwRI has transitioned to testing and analysis in support of the return-to-flight effort, to certify that the current space shuttle is safe to fly. Looking toward the future, the Institute team is continuing to support the design and safety analysis of the next-generation space shuttle.

Upon release of Volume 1 of the CAIB report, two conclusions were stated as factors in the cause of the accident. While the second conclusion addressed policies and procedures within NASA, the first was described thus: "The physical cause of the loss of Columbia and its crew was a breach in the thermal protection system on the leading edge of the left wing. The breach was initiated by a piece of insulating foam that separated from the left bipod ramp of the external tank and struck the wing in the vicinity of the lower half of reinforced carbon-carbon panel 8 at 81.7 seconds after launch."

The report stated that during re-entry, superheated air penetrated the leading-edge insulation and progressively melted the aluminum structure of the left wing, "until increasing aerodynamic forces caused loss of control, failure of the wing, and breakup of the orbiter."

Technologies developed at SwRI over many years were instrumental in demonstrating that the impact of the foam during launch was the cause for the breach of the left wing leading edge. In addition, the team demonstrated that an impact into thermal tiles on the underside of the wing was not the cause of the accident.

Comments about this article? Contact Walker at (210) 522-2051, or james.walker@swri.org, or Grosch at (210) 522-3176, or donald.grosch@swri.org.


Numerous SwRI staff members were involved in the accident investigation, including Scott Mullin, Erick Sagebiel, Carl Weiss, Larry Bishop, Juan Magallan, Shaun Schraeder, Jerry Nixon, Joe Elizondo, Ray Burgamy, Art Nicholls and Dick Sharron, and 3-D imaging and nondestructive evaluation teams led by Dr. Ernest Franke and Dr. Glenn Light, respectively. NASA and its contractors supported the experiments and data collection. The authors thank Justin Kerr, of NASA Johnson Space Center, who led the impact testing for the NAIT; Scott Hubbard, who represented CAIB during the impact tests; and Paul Wilde, who supported the CAIB. Thanks also to NASA and CAIB for photographs used in this report.

Further Reading

The CAIB reports can be found at www.caib.us. Volume 1, Chapter 3, entitled "Accident Analysis" in Section 8 "Impact Analysis and Testing" discusses the work performed at SwRI. Detailed discussion of the test plan, tests and NAIT analysis will be found in a forthcoming NASA document. CAIB Report Volume 2, Appendix D.12, includes a chapter on the SwRI analysis.

Published in the Fall 2003 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Joe Fohn.

Technics Fall 2003 Technology Today
SwRI Publications SwRI Home