Columbia

This week I started the second course in my EdX Model Based Systems Engineering course series, "Models in Engineering". Part of the week's assignment was to examine the models used in the post-accident analysis of the reentry breakup of Columbia during STS-107. This is an excerpt of my analysis.

The Columbia Report

After the loss of the Columbia orbiter during STS-107 a comprehensive review of the accident was performed. The review board was tasked with determining what caused the orbiter to break up upon reentry into Earth's atmosphere and suggest mitigation strategies that could prevent the loss of future flights. As part of the causal analysis, the review teams utilized a variety of models to analyze the effect of foam strikes on the Space Shuttle orbiter.

Image analysis of video footage captured during the ascent phase of the mission indicated a large section of foam debris separated from the external fuel tank and collided with the leading edge of Columbia's left wing. Based on this analysis, the team identified a range for the speed of impact between 625 and 840 feet per second, and debris dimensions of between 21 to 27 inches long by 12 to 18 inches wide. These parameters were then used to guide mathematical analytical and simulation models to try to infer missing information such as the location of the strike. These models included a computational fluid dynamics (CFD) simulation performed in CART-3D which provided a position and velocity estimate of the foam section, analytical estimates of the aerodynamic properties of the foam section based on its inferred ballistic and aerodynamic properties, and computed trajectory estimates based on imagery data. In addition to these mathematical models, the board performed physical modeling of the foam strike using a compressed gas gun to launch foam projectiles at a representative wing section, using the estimated debris properties from the previous analyses to inform the model parameters.

Some of these models were available during flight - CFD analyses had been performed previously, but not with data obtained from the external tank. The review board noted that no instrumented external tanks were flown prior to STS-107, and so there was a lack of data on the bipod strut available to validate the computational models. In all, the models seem to have the following credibility and fidelity characteristics:

  • Image analysis
  • Credibility: Low
  • Fidelity: High
  • Available after launch
  • CFD
  • Credibility: High
  • Fidelity: Low
  • Available during flight
  • Ballistic coefficient verification
  • Fidelity: Low
  • Credibility: High
  • Available during flight
  • Foam gun test
  • Credibility: High
  • Fidelity: High
  • Not available during flight

The results of each of these models helped to inform the accident review board's analysis and enable the return to flight of the shuttle. In their report, the board indicated that there was sufficient data and models available during the flight to determine the effect of the foam strike on the RCC panels. However, they did also suggest in their recommendation R3.8-2 that NASA develop more physics-based models for use during flight. These models could be used to assist in on-orbit repair for future missions where foam strikes or other wing damage events were suspected.

Had the models available during flight been utilized and indicated a possibility of damage to the wing, a space walk could have been performed to confirm the model predictions and identify the extent of the damage. If major damage was detected, the crew could attempt on-orbit repairs or be rescued by Atlantis; however, both of these options carried attendant risks. The report concludes:

"If Program managers had understood the threat that the bipod foam strike posed [...], these repair and rescue plans would most likely have been developed, and a rescue would have been conceivable."

If you like what I write, you can subscribe to my mailing list to get new posts in your inbox: