3 Seconds to Disaster: The SpaceShipTwo Accident

Pete Siebold under canopy approaches the desert floor. (Credit: Mark Greenberg/Virgin Galactic)
Pete Siebold under canopy approaches the desert floor. (Credit: Mark Greenberg/Virgin Galactic)

On Oct. 31, 2014, the crown jewel of Sir Richard Branson’s ambitious space effort, SpaceShipTwo, broke up during a flight test over the Mojave Desert. Scaled Composites co-pilot Mike Alsbury was killed while Pete Siebold managed to parachute to safety.

This week, Parabolic Arc is publishing a multi-part series on that accident. The series is based upon the accident investigation conducted by the National Transportation Safety Board and Parabolic Arc‘s own reporting about the SpaceShipTwo program.

The stories will take you inside the pre-flight briefings as the pilots and their colleagues at Scaled Composites and Virgin Galactic prepared for the program’s riskiest flight test.

You will be in Mission Control, the cockpit of SpaceShipTwo, and in the skies above the Mojave Desert as a flight that began so well went tragically wrong.

You will see Pete Siebold’s miraculous survival as the ship breaks up around him and his terrifying descent in temperatures colder than on Mount Everest. You will follow emergency responders as they raced to the stricken pilot’s location.

Finally, there is a detailed look at what happened in the aftermath of the tragedy that took the life of Mike Alsbury.

The Series

  • Pasi Jokela

    So.. can anyone explain to me why the tail booms rose to the “feathered” position when they were unlocked? They are motorized, right? I’m not really sure how this system works (or is supposed to work). Is the “feathered” position their aerodynamical “rest” position or what gives?

  • Douglas Messier

    The locks are needed during the transonic region around Mach 1. The actuators are not strong enough to hold the tail booms in place due to aerodynamic pressures up to around Mach 1.2.

    After Mach 1.2, the feather can be unlocked safely. Alsbury was supposed to do that at Mach 1.4. The additional .2 Mach was a safety feature. They unlock the feather at Mach 1.4 to make sure it unlocks. They don’t want to be up at maximum altitude and have the locks stick. A feather down reentry would be very difficult. If the feather doesn’t unlock on the way up, they need to cut the engine and abort the flight.

    The feather is a good system for re-entry, but it’s complicated in operation. The engineers fell into a trap of worrying about one issue (reentry) and not about something going wrong on the way up. NASA did the same thing with Apollo, worrying more about a fire in space than one on the ground.

  • Pasi Jokela

    Aah, so the tail booms really have a property of rising “on their own” in a certain flight regime. I’m not quite sure if that is a good desing. I would rather come down in an unstable configuration (until enough atmosphere) rather than have the craft “shuttlecock” without my full consent or due to any kind of error or malfunction. Anyways, releasing the locks too early should be preventable easily enough, even with a simple pressure sensor that overrides pilot’s actions. Or maybe the whole thing should be made automatic.. why bring an errant human in the loop at all? 😉

  • Douglas Messier

    NTSB concluded they had failed to properly take human error into account when designing the system.

  • Aerospike

    A pressure Sensor sounds like a simple solution at first glance, but if you think about it, it just introduces another failure mode into an already complicated system.

  • Pasi Jokela

    You are right, of course. But one could have several pressure sensors, One could have pressure sensor(s) override switch. They quite likely have the craft position and velocity vector from GPS as well..

    The best thing, however, would be to have the tail booms assume glide position by the aerodynamic forces alone, when the craft is in the lower atmosphere. Maybe that’s not possible with SS2, but it would be the ideal.

    Or.. you could just forget the whole shuttlecock idea and have the computer actuate the flight control surfaces during descent to keep the craft relatively stable and the passangers fairly comfortable.

  • Kirk

    The tail booms are actuated pneumatically, not hydraulically, presumably as a weight saving measure.

    Hydraulic fluid is incompressible, so a hydraulic system will remain locked in place when its control valves are closed (up until the force where the system is compromised, such as by rupturing the hydraulic lines, or where the system’s pressure relief valve lifts). This is true even if the motive force for the system is derived from a compressed gas (as with some lightweight, short use, open circuit systems — such as the Falcon 9 first stage grid fin control).

    Gases, on the other hand, are compressible, so even when its control valves are closed, a pneumatic system can be overpowered by forces below that which would actually damage the control system.

  • ThomasLMatula

    In short use a design similar the old X-15, but with the application of a modern fully automated flight management systems which is probably the smart decision.

    But I see two problems. First the shuttlecock design won the Ansari X-Prize, it must be the best design for suborbital tourism, prizes always pick the best answer to a problem. 🙂

    Second, Burt Rutan’s strength has always been in airframe design, not computerize cockpits. Look at how ancient the controls look on SpaceShipTwo. I recall someone stating it was like stepping back into the 1950’s. Such a design change would require Scaled Composites to change their entire paradigm on how it designs aircraft and spacecraft.

    Bottom line. If Virgin Galactic did make this decision they would basically be going back to square one, to when they were founded in 1995 before they allowed the Ansari X-Prize to select their design. This means basically putting out a RFP and starting out as if there never was an Anasari X-Prize. Probably a good decision for the long run, but you would be looking at 4-5 years before entering commercial service. And more importantly an admission they messed up buying into the Ansari X-Prize hype.

    No, the more likely option is for VG/scaled Composites to just add kluges to what they have, hopefully get it safe enough to fly out their reservation list (150-200 flights) to an altitude they are able to claim is close enough to space to be space without any incident and then quietly retire it to a museum like SpaceShipOne and declare victory.

  • ThomasLMatula

    Doug,

    You know that reminds of the John Denver crash in one of Scaled Composite’s designs. If I recall the cause of the crash was running out of fuel in one of the tanks and losing control when he had to turn his body around to reach a fuel selector switch located behind the pilot, where the fuel gauge was also located…

  • ThomasLMatula

    Doug,

    Great job with this series. You are shining a light into what has been darkness.

  • Doug Weathers

    “In short use a design similar the old X-15, but with the application of a modern fully automated flight management system”

    Ironically, the fatal crash that ended the X-15 program was caused by the newfangled flight control computer. Computers aren’t very good at handling unexpected situations, and that’s why we still use test pilots instead of test robots.

    Once you know you’re putting a human in the cockpit, you need to design it around the human (strengths and weaknesses) and train the human how to fly that vehicle. This is what the NTSB says that VG didn’t do well enough.

  • ThomasLMatula

    True, but flight computers have advanced greatly in the last 50 years. Aircraft like the F-117, B2, F-22 and F-35 would not be able to get off the ground without them. The key is they are not replacements for the pilot but complements that reduce the risk of pilots getting into situations that are not recoverable.

    That said avionics are expensive, often the most expensive part of a modern aircraft and so its understandable why small firms would minimize their use.

    And yes, I do agree, that lacking a good flight control computer, and it would probably be as difficult to integrate one into SpaceShipTwo as a new engine, a much better designed cockpit would be in order.

  • stoffer

    Shuttlecock reentry in a craft with variable configuration was considered already in the 50s. It was never implemented, because the engineers were wary of exactly this type of accident and capsules became standard. It is also what killed the DC-3 shuttle concept by Maxime Faget – the airforce people hated the idea of stalling through the reentry, and possibly ending up in an unrecoverable spin.

    I don’t think that the whole shuttlecock reentry is needed for suborbital flight. it is not like the heating loads are unbearable. It would be probably better to make an intrinsically stable airframe. A delta wing and good weigh distribution would be good enough probably. This is why it was chosen for the shuttle. It is also what XCOR are doing with the Lynx.

  • TimR

    Doug. The speed is definitely one factor in the force imparted on the tail booms that require the locking mechanism but it is not after Mach 1.2 that the booms can be unlocked but rather the combination of speed and altitude. You are referring to Mach 1.2 as a reference to the flight profile and the vehicle presumably being at a high altitude where overall the dynamic pressures are low despite the speed.

  • TimR

    Great Q. Thanks for posting it.

  • I always felt that many of the problems associated with SpaceShipTwo would have been discovered if SpaceShipOne had something akin to a normal flight testing cycle. Seventeen total flights, six powered and only three into space — the first two of which had serious issues. Then it gets shipped off to a museum. Not enough. Not nearly enough to really understand the motor, feather and operational requirements in order to build a much larger ship based on a different design.

    SpaceShipTwo was heading in the same direction before the accident. A couple of more flights, all of one to maximum altitude, turn it over to Virgin Galactic in December, then fly Richard Branson and his son Sam by the end of March in New Mexico on the first commercial flight. I quote Chuck Yeager from his autobiography:

    “By definition, a prototype was an unproven, imperfect machine….Some defects were obvious….But other problems…might be discovered late in a program, only after hundreds of hours of flying time. The test pilot’s job was to discover all the flaws, all the potential killers. Testing was lengthy and complicated, resulting in hundreds of major and minor changes before an aircraft was accepted in the Air Force’s inventory.”

  • TimR

    If the feathering remains a manual process, the simplest safety feature introducing little more complexity might be to introduce a time delay between the (co)pilot initiating the unlock and the unlock execution. Feathering of the vehicle is not such a time critical operation such that, e.g. 5 second delay could not be introduced. Assume such a delay was in place during this last flight. An automated audio warning could be heard by both pilots accompanied by a common blinking text message on panel. Additionally, sensor measurements could define the audio message – raising the level of alert if outside the safe envelope of operation. Given this feedback, to both pilots, for 5 seconds, I think Seabold and/or Alsbury would have realized the impending error and halted the unlock release.

  • Douglas Messier

    That’s true. The Mach 1.2 was for this flight profile.

  • Doug Weathers

    An excellent reply. I agree with you in all respects.