What happened to Columbia?

The 'not-Starfire' image.

At right is a link to a pair of simple image processing results on the 'Starfire Group' image of Columbia, purportedly taken with a 3.5" Questar telescope aimed at a computer controlled tracking mirror, as she approached from the west on the morning of February 1, 2003. A digital copy of the original image was retrieved from the NASA web site, rotated until symmetric about a horizontal line in NIH Image, and smoothed to reduce pixilation. A copy of the upper (right) half of the orbiter was made, flipped, and pasted onto the lower left half of the orbiter image. Two different methods of image combination are displayed to highlight anomalous features. I could not easily turn the protrusion into an indentation by interactive alignment in the software package.

The azimuth of the shuttle when the image was taken is not stated, so the precise rotation which needs to be applied could not be determined, and neither, therefore, the precise geometry. I have not been satisfied with the projections of the forward fuselage geometry against the sky for any VRML orientation that i have tried. The version i worked with (obtained from an on-line space reporting service) is of unknown photometric scaling. Nevertheless one can speculate that the protuberance showing on the leading edge of the left wing might be optically bright material emanating from the interaction of the plasma with the wing.

A note of caution should be stated regarding the above speculation: one can actually get configurations of Columbia where the vertical stabilizer peeks out from behind the wing, jutting forward of the leading edge...right where the leading edge, left wing anomaly occurs! Likewise, it is possible to get configurations where the rear of the vertical stabilizer juts behind the left wing. Furthermore, if the protrusion isthe vertical stabilizer, then the anomaly trailing the left wing could be the ionization tail coming off of the nose!! Until the precise geometry is made clear, the nature of the leading-edge left-wing anomaly should be treated skeptically.

AWST Feb 7 issue states that a 75-degree left-bank had been completed by the time this image was taken. The NASA groundtrack & events summary timeline shows that occurred at approximately 13:56:55. Albuquerque overflight occurred at approximately 13:57:23. From the image, one can measure an aspect ratio of vehicle length to wingspan of 1.69:1. The approximate true aspect ratio is 1.51:1. From this, one can compute that the vehicle must have been in at least a 27-degree bank relative to the ground observer. The true bank relative to the ground observer is not constrained with this information since the vehicle is likely pitched down at some angle (nose closer to the observer), making the likely relative bank angle higher than 27-degrees. This assumes no distortions have been introduced in the creation of this public release version of the image. No distortions are apparent in the above NIH Image rotation and overlays.

A trigonometric calculation from this and the information provided in the ABQ Journal article yields that the image was taken approximately 16 seconds before this time, depending upon the location of the ground telescope relative to Albuquerque and the actual Columbia ground track.

Note that the location of the anomaly extends from the junction of the double delta outward along the leading edge of the wing, but does not obviously extend forward along the highly swept chine segment of the wing. Since we do not know anything about the photometric scaling or zero-point, in the image, one can't say anything about what is seen between Columbia and the camera. Note that the outer edge of the main landing gear wheel well box is approximately in line with the junction of the double-delta.

As part of trying to understand Columbia's demise, one needs a model. I created a simple 6-element, 3-dimensional volumetric model (triangular plates for tail and wings, cylinders for mid- and aft-fuselage, and a frustum for the forward-fuselage) in an attempt to understand numerically what would happen during a wing failure event. Although crude, the results are interesting and to a limited extent applicable to Columbia.

Aero-Thermal Environment, Shuttle, Mach 18

Control during re-entry: how the shuttle does it.

Shuttle re-entry is a delicate balancing act where the attitude of the spacecraft is adjusted to balance the center of mass close enough behind the center of pressure to minimize the aerodynamic torques on the vehicle. Technically, this is dynamically unstable, but the shuttle design provides sufficient control authority over a specific range of entry profiles.

Lift generated over the wing changes with speed due to different flow conditions at different speeds, therefore the angle of attack must be varied as the shuttle decelerates. Even though the shuttle is travelling very fast, at high altitudes where the atmospheric density is low, the elevons have little authority and so are not very useful. Reaction Control System (RCS) thrusters are relied upon in this regime for attitude control.

However, once the aerodynamic pressure builds, the elevons and body flap can be used to adjust the center of pressure on the underbody of the shuttle. This is achieved by actuating the elevons and body flap upward or downward, fractionally increasing or decreasing the drag area on the back of the vehicle, thereby adjusting the center of aerodynamic pressure at a given attitude. In some altitude range between entry interface and stratospheric flight, the Reaction Control System (RCS) shares duty with elevons and body flap for attitude control.

At low and sub-Mach numbers, the aerodynamics of the lift are different, the angle of attack of the vehicle is reduced since thermal protection is no longer required and the elevons change their role transitionally to become more conventional flight control surfaces.

The primary goal at the early stage of reentry is to bleed off orbital speed without undergoing high accelerations. Unfortunately, this has the tradeoff of high heating rates and total amounts of heat that are transferred to the spacecraft. Thus the need for the TPS (Thermal Protection System) which we have all heard so much about.

Columbia was in the flight regime where pitch RCS activity was nominally deactivated, although yaw control was the domain of the RCS. This is because at a 40-degree angle of attack travelling at Mach 18, the vertical stabilizer is shielded from the flow and is ineffectual at yaw control. Roll control was possible with both RCS and elevon actuation. What the domain control for each in this flight regime, i confess ignorance.

6-element, 3-D Inertial Shuttle model, 40-degree AOA.

This model was implemented in a spreadsheet form (Excel 4.0 on a macintosh at home) and used to compute approximate center of mass, principal moments of rotational inertia (uniform and cylindrical mass distribution assumptions), and center of pressure during entry at a 40-degree angle-of-attack--both with and without a left wing. Using standard classical mechanics, the model predicts angular acceleration rates starting from the stationary nominal entry attitude around all three axes of varying magnitudes.

What would happen with a wing failure?

A number of different type of wing failures are possible, and for the time being i will focus on wing separation or partial wing separation (or even disintegration) events. If you look at diagrams showing the interior wing structure of the shuttle, you can see that there are 4 large spars from wing-edge to mid-fuselage in the torque box which support the bulk of the wing loading.

Wing loading and design limits.

According to the AIAA Aerospace Design Engineers Guide, the shuttle wings are designed to withstand 4.5-g loads. From the NASA timeline charts, one can compute an aerodynamic deceleration of about 0.6-g near the time of breakup. However, there are also components of 1-g (downward) and centripetal acceleration (~-0.45-g) to factor in, bringing the total wing loading to ~0.8-g.

Types of wing failures:

• Simple wing separation at the root.

  If Columbia was still maintaining a nominal entry attitude at the time of the wing separation event (which has not been concluded from Columbia's telemetry), then the model can be used to predict the initial attitude evolution. From general principles, even without a sophisticated model, one can conclude that in a wing separation event, Columbia would have yawed to the right (reduced left wing area producing drag) and begun rolling to the left. The roll can be understood because when the wing breaks off, the center of pressure shifts significantly toward the other wing. Center of mass also shifts toward the other wing, but not by as much since each wing weighs only about 10 tonnes. The moment of rotational inertia about the roll axis of the shuttle is approximately some five times smaller than the moment of inertia about either the yaw or pitch axes, therefore roll oscillations are easier to excite than others. The aerodynamic force on the bottom of the orbiter then produces a torque on it of magnitude ~ hundreds of kN-m. Because of the 40-degree angle-of-attack with no nominal yaw, the components of this torque are mostly around the orbiter X- (roll) and Z-axes (yaw), although a small component of pitch is induced, depending upon the amount of center of pressure shift fore/aft.

• Outboard wing separation.

  The general instability outlined for wing separation at root still applies for separation further outboard, but rates induced in all three axes are reduced. The crude spreadsheet model predicts a roll of up to 30-degrees in 1.5 seconds, and a yaw of 30-degrees in as little as 3 seconds for a root separation (at the fuselage/wing junction). Since these rates are derived from assuming separation at the wing-root, and since the Starfire image and pre-LOS telemetry indicate the burn-through locus was further outboard along the wing, these rates probably represent upper limits. The 25-second telemetry dropout time might well be the result of the antenna beam patterns no longer painting TDRS as the orbiter rolled over upside down (in this scenario).

  Predicting the detailed time-evolution of the reentering orbiter much beyond an initial phase is perilous when using a simple model, but general principles can be applied to understand something about the stability of the configuration at various different attitudes and angular rates with respect to the local aerodynamic flow.

  Directionally Damped, Oscillating Roll?

  After rolling through 90 degrees, the aerodynamic torque would have then acted to slow the roll as the upper portion of the right wing became exposed to the aerodynamic flow. Before this even, as the vertical stabilizer (assuming it was still attached) rotates into the flow, it will act as a drag and roll-rate reducer. It is likely that the re-entering Columbia would not have rolled completely 180-degrees upside down because of the wind-brake affect the tail would have. Once the roll was stopped, the torque would keep acting to accelerate the roll in the opposite direction, reversing the roll. The vertical stabilizer would act to give more torquing power for the reversed roll than was in the initially induced roll, but in the opposite direction.

  The simple model indicates that yaw and pitch rates may not have been as high as the roll rate. However the evolving location of the center of pressure with respect to the center of mass coupled with the changing attitude of Columbia is the domain of a higher fidelity model, and high-confidence conclusions cannot be reached about Columbia's behaviour based on this simple model.

  If the model is a reasonable representation of Columbia's state at the time of such a catastrophic failure, we can further speculate that the resumption of 2-seconds of telemetry may have been the result of Columbia being pushed back upright by the aerodynamic force that acted on the 'upper' side of the right wing as the reentering space plane rolled onto its back, effectively allowing a restoring torque to push Columbia back upright. One can think of this behaviour as a manifestation of a harmonic oscillation in roll, although it would be damped on one side of the roll due to the presence of the vertical stabilizer (if still attached after the separated left wing impacted the back of Columbia. The 25-second dropout period is about two-times longer than the simple model predicts. But recall that in the simplified model, wing separation at the root was assumed. The Starfire group image suggests that separation probably occurred further outboard. This would drive the predicted period of oscillation upward.

  What we all saw on February 1, 2003: the videos.

  There is compelling video evidence for a wing separation event early in the breakup timeline. In one of the amateur videos shot in Texas from the north side of the ground track, one can clearly see a shimmering object behind the orbiter as the video starts, near the right edge of the screen, flickering at about a third or half a hertz, moving slower than the main spacecraft, but still at a substantial fraction of it's speed. After only about 5 seconds, the object has separated enough to be out of the field of view with the head of the reentry path in the center of the field of view. I have been presuming this might be the left wing or a fragment of it, although it is not evident in other angles, perhaps due to unfavorable sun glint angles. The left wing is not the only possibility: this object could have been other large flat surfaces such as the vertical stabilizer, top or bottom wing panels, or even a payload bay door.

  Many other events are notable in this and other video segments...puffs in the trail are either RCS firings or pressurized tanks or lines rupturing; the blooming and mushrooming of the trail later on may be due to cabin or Spacehab depressurization. One can note a distinct reddish-brown coloration that takes over the trail suddenly around this time--the RCS and/or OMS N₂O₄ tanks failing. Also appearing near this time, many debris fragments emerge from the locus of the vehicle.

  Powerful arguments against wing separation

  The wing torque-box (which carries the bulk of the wing loading) contains 4 main spars running from mid-fuselage to wing-edges. In order for wing separation to occur, all 4 of these spars would have had to be attacked by the heat, either burned through or heated to the softening point. While this may have happened at some stage during loss-of-vehicle, it is likely that other control wiring and/or hydraulic lines would also have been attacked under such conditions.

  Evolution of a wing separation.

  Other concrete conclusions can be arrived at from general physical principles: if the wing did not separate instantaneously, it would have folded upward still attached tenuously to the orbiter along the fracture line, rotating through 90-degrees in approximately 1-second. If separation was not clean, it is likely that it would have struck the orbiter body, left OMS pod, and/or vertical stabilizer. There were unconfirmed reports that telemetry indicated a leak in the RCS system. A wing separation scenario predicts that a failing outboard wing segment possibly folded up and struck the rear of the left aft OMS pod, which could (if confirmed) produce an RCS leak.

• Lower wing panel separation.

  The wing may not have separated before the entire orbiter began to disintegrate. Other failure modes are imaginable. For example, the high temeratures could have softened the attachments of the upper and lower aluminum wing panels to the wing frame enough for the dynamic pressure to blow them off.

  The previously mentioned simple wing separations do not seem to correspond with at least one report that Columbia was believed to be in a flat-spin when telemetry was regained near the end of the post-MCC-LOS 32-seconds. Wing panel separations might go some measure toward an explanation if a lower wing panel peeled back and down underneath the orbiter. This might torque the vehicle into an initial hard left yaw. Does someone volunteer to run the numbers on this?

• Upper wing panel blowout: also a possibility.

  Upper panel blowout might still do it. An upper panel blowout would marginally reduce drag on the left side and induce (an initial) rightward yaw, and possibly roll. After the right yaw is induced, the vertical stabilizer eventually moves into the stream and breaks the yaw, possibly eventually reversing the yaw rate. If the angular rates work out correctly, Columbia could have been spinning back to the left during the final two seconds of telemetry.

• Forward outboard torque box separation.

  Another possibility is that the plasma torched through such a large aerodynamic chord and bled laterally through the wing enough to soften the entire forward outboard section of the torque box. Dynamic pressures could then have overwhelmed the local ability to sustain load, shearing off the forward section of the outboard wing, or possibly just bending it back and up above the plane of the wing.

• Hydraulic or electrical elevon systems failure

  In this scenario, the plasma could have shorted out the elevon electrical control wiring, burned through all three hydraulic systems tubing, or produced a related mechanical failure in the hydraulic pressure system. The flight control surfaces presumably relax to the minimum resistance to the aerodynamic flow. A simple extrapolation from the 6-element model shows that this has the result of inducing pitches (pitch up), rolls (since at MCC LOS there was more than a 6-degree difference in elevon trim starboard/port), and a restoration of a yaw if such was being produced by asymmetric drag.

  A hydraulic system failure has a number of possible consequences:

  • Speed brake opening.
  • Elevons relax up against stops.
  • Left yaw, right roll, pitch up.
  • right rear RCS thrusters fire to counteract left yaw.
  • Comm hits if the vertical stabilizer moves in line-of-sight of TDRS.

  The integrated timeline states the following: "13:59:31.7 Speedbrake Channel 4 OI position Measurement Indicated Successively 19 deg, 20 deg, 24 deg Over last Three Samples Prior to LOS (Should be Closed/0 deg). No Change Seen in Actuator Motion nor Change in Secondary Delta Press that would Indicate a CMD or Position Feedback Discrepancy in Channel 4. Signature May be Possible Indication of Pending Failure of ASA 4."

  A fraction of a second later, the timeline states: 13:59:32 GMT Max Observed Aileron Trim (-2.3 deg). LOS for: M-FUS LT BL Temp at x1215; LH Aft Fus Sidewall Temp at x1410; LMG Brake Line Temps [A (172.2F), B (154.2F), C (104.8F), D (88.3F)]; LMG Strut Actuator Temp (76.3F); Hyd Sys 1 LMG Uplk Actr Unlk Ln Temp (52.2F); Sys 2 LH Brake Sw Vlv Return Temp (AFT) (62.8F), Sys 3 LMG Brake Sw Vlv Return Ln Temp (FWD) (67.3F)"

• Tire rupture

  In this scenario, the tires overheated and exploded, sending shrapnel through the wing in various directions. This scenario needs to stand the test of calculations. Temperatures in the wheel well were rising at MCC LOS. The hydraulics are routed through the wheel box and could all suffer catastrophically from this failure, but an explanation must be made for how this could be a part of a chain leading to a flat spin (if that state obtained).

Evolution of the breach.

Even though we do not know the extent of the damage before entry, basic physical principles can be applied to examine the limits of the evolution of damage within the wing after reentry began given what is known about sensor readings, known failures, and nominal behaviour.

The Timeline charts for off nominal behaviour show the evolution of sensor dropout and temperature trending prior to MCC LOS (official Mission Control Center console Loss of Signal at 13:59:32.136 UT).

The first signs of trouble were several temperature probes in the left main gear box showing slow, linear rise rates, probably indicating excessive heating of the gear box, but not indicating a gear box burn-through yet. It is likely that plasma was entering the wing by this time, but the penetration point(s) are not yet identified.

Probes in the aft portion of the wing went "off-scale" low throughout the entry period starting at 13:52:59, about 30 seconds before Columbia passed over the California coast, indicating power or data failure. This probably indicates that wires connecting these probes were burned through since the probes in the rear of the wing were not showing abnormal temperature readings and many of them had wire bundles routed around the outer and forward sides of the left main gear box. Other temperature (and wheel pressure) probes began showing temperature rises during the interval before MCC LOS, and others dropped off-line including ones in the left main gear box.

Curiously, a few others started trending down in temperature. This indicated a disturbance in the nominal heating pattern on and in the wing. This is potentially an important clue to the state of Columbia's wing, although i do not understand what it is telling us.

By 13:55, heat was penetrating to the join of the mid-fuselage and the wing, between the first and second struts. At about 1 minute before MCC LOS, a large number of sensors went off scale low in the wheel box.

Prior to MCC LOS, the left main gear brake line temperatures were elevated significantly, possibly indicating heating of the fluid in the lines or the line itself by direct exposure of the external line surface to plasma.

Did Columbia scream?

For five seconds after MCC console LOS at 13:59:31.167, telemetry continued coming down from Columbia, but suffered from high bit-error rates and so was not displayed on MCC consoles. James Oberg's Feb 25 article describes Columbia as flying on at this point, largely unchanged. Could some event associated with loss of vehicle have produced the high bit error rates before telemetry ceased?

Not necessarily. The TDRS satellite in use for telemetry downlink was near the vertical stabilizer from the view of the upper right S-band quad antenna located on the upper portion of the forward fuselage. Communication interference could have been caused by enhanced plasma density along line of sight or perhaps by partial occlusion by the vertical stabilizer. Brief periods of Columbia downlink dropout occurred throughout reentry.

Fuselage still intact at MCC LOS + 30 seconds.

For the brief two seconds of telemetry received following the extend complete LOS period, telemetry showed loss of pressure in all 3 hydraulic systems. Fuel cells (forward lower mid-fuselage), APUs (aft fuselage), and flight control computers (forward fuselage) aboard Columbia were all operational.

RCS thruster firings around MCC LOS

There are four types of rocket engines aboard shuttles: main engines (SSMEs) which produce some of the thrust to reach orbit, OMS engines which are used for orbit insertion and deorbit maneuvers, and two types of reaction control system (RCS) engines: verniers (100 N force) for fine control and R-40As (4000 N force) for higher rate maneuvers. The high rate RCS thrusters are the ones which were firing, each capable of producing about 70 kN of torque about the pitch and yaw axes of the orbiter, and about half of that around the roll axis when used in oppositely directed pairs. Comparing the roll rate of a pair of thrusters with the roll rates induced during a wing separation in the simple model, it is clear that RCS authority is insufficient to counter the roll with even a portion of the outboard left wing separating.

A one-second firing of a single high-rate RCS thruster produces about a half a degree per second yaw or pitch. Six yaw motors (4 aft, two forward) would produce about 3-degrees per second yaw. Starting about one-and-a-half seconds prior to MCC LOS, 2 right rear high rate RCS thrusters began firing. These would nominally have produced a 1-degree per second-squared angular acceleration--not enough to counter the yaw rate produced in the wing separation in the simple model, but approaching the threshold for yaw control for separation of the outboard segment of the wing. Combined with the newer information that an additional pair of RCS thrusters were active during the 5-seconds of post-MCC-LOS and before telemetry dropout, we could conclude that the entry guidance system was correcting for at least a large fraction of yaw that would have been induced from a hypothetical outboard wing separation.

Forward RCS units had been disabled by this time of flight, and RCS propellants possibly 'dumped'. The rear RCS pitch jets had also been commanded off as a nominal part of entry, since elevons nominally control the pitch of the orbiter during this stage of reentry. Jenkins states that rear roll thrusters are off by this stage as well, since aileron/elevons are used to control vehicle roll. The integrated timelines do not show when or if this did occur during Columbia's entry. To summarize, nominally only rear RCS yaw thrusters are available for vehicle control to supplement aileron/elevon and body flap actuation.

Columbia, at MCC LOS, had limited ability to manage unusual drag or center of pressure shifts during this phase of reentry. Yaw thruster ability was available, but at a nominal 40-degree angle of attack, any unusual aerodynamic torques would likely have had components in roll as well as yaw. If elevon hydraulic pressure was lost, Columbia might temporarily have been able to correct for yaw excursions and rates, but she probably would not have had an option for maintaining nominal entry attitude.

I won't pretend to know. We have seen 3 video frames ~ 80 seconds into launch showing at least two significantly sized chunks of debris between the orbiter and ET in one frame, and another chunk in another frame, the debris disappearing under the wing and perhaps another chunk of debris passing under the orbiter--possibly one from a previous frame, and then in the third frame something reappearing under the wing as a spray of widely dispersed particles. We have heard reports of an object tracked separating from the shuttle while on-orbit. Below, are some of the more popular theories of what went wrong, not all of which are mutually exclusive.

• Foam or ice separated from the ET (external tank), possibly from around the base of the strut connecting the orbiter to the ET, impacted either the underside or leading edge of the orbiter left wing, compromising the thermal protection system on the wing. This is the approximate area where the central New Mexico image of Columbia shows unusual aero-thermodynamic or mechanical appearance. Either tiles on the underside or a damaged leading edge may be the source of entry of high-temperature plasma into the wing.

• Impact of orbital debris during the first day of on-orbit operations, possibly on the left wing. An object currently presumed to be 1-2 feet across was tracked near the shuttle and at some point separating from it at 5 metres/second, and reentering just two days later. Ice breaking off from the waste dump ports would have had to be filamentary or sheet-like in order to have a sufficiently low ballistic coefficient to re-enter within just a few days.

  An alternate theory to impact for the source of this debris is that Columbia could have been shedding material all the way to orbit. The fragment tracked by ground radar near Columbia may have been pinned by launch acceleration to the wing.

• Excessive heating on Columbia's wings due to a phenomenon called Early Boundary Layer Transition, perhaps caused by excessive 'roughness' of wing surface. Scenarios include either a purely roughness-induced burn-through or damage from debris impact sufficient enough to produce turbulence during reentry with heating rates high enough to melt or slump tiles. Note AWST Feb 17 and 24 articles which explore this possibility to some depth.

• Electrostatic discharge. In this scenario, large electrical potentials which can be produced in the upper atmosphere are induced to equalize from the passage of the ionization envelope and trail of a reentering shuttle. Essentially, a lightning bolt strikes Columbia, presumably on the leading edge, damaging the TPS.

• Terrorism. How? This would have required either an insider deliberately damaging the spacecraft before launch, or being able to remotely damage the vehicle upon ascent. I have heard of no plausible scenario for damaging STS upon the early phases of reentry.

• Improper maintenance. Here, a number of possibilities can be debated, ranging from inadequate or failed attachment of RCC leading edge segments or tiles, to inadequate ice inspection pre-launch, to improper forming or bonding of ET insulation.

  ---

  Links

  There are many, many links on the web that contain relevant information, and i will not even try to provide comprehensive links, but i will provide a few to those of technical and news-source interest.

  News Links:

  • NASA Columbia site.
  • Good coverage by Spaceflightnow
  • Aviation Week & Space Technology News Site
  • Space.com web site.
  • Spaceref.com

  • Columbia Loss FAQ.

  Specific Stories:
  • Feb 25 ABCNews story...
  • Feb 25: An unusual looking tile.
  • Feb 28 Spaceflightnow left wing debris story & pic
  • April 16 story on breach evolution.

  ---

  Technical Links

  • NASA Office of Human Spaceflight Shuttle Web
  • Technical Reference Document Index
  • Sensor timeline charts, Feb 07 briefing.
  • STS-107 Groundtrack & Events Summary
  • STS-107 Investigation Reference Page
  • Columbia Accident Investigation Board

  ---