And the (probable, not final) winner is, a failure in a support strut
holding an He bottle *down* against buoyancy in the LOX. Bottle then
apparently shot to top of tank, releasing enough He to overpressure the
tank and cause the failure.
On 6/29/2015 7:36 AM, Henry Vanderbilt wrote:
This strongly implies there was no clear cause in the data before the
final milliseconds, which rules out a whole class of things that might
gradually overpressure the second stage LOX tank: Frozen pressure-relief
valves, a stuck-on helium regulator, a heat source boiling the LOX.
(Come to think of it, the latter two would also require stuck-closed
pressure-relief valves.)
As for overpressure sources that split a tank within milliseconds of
onset, the first thing that comes to mind is a failure under flight
loads in the high-pressure helium storage bottles-plus-plumbing
submerged in the LOX tank.
A second possibility is some sort of ignition inside the tank - EG,
something breaking loose and scraping an aluminum surface, or a brace or
baffle cracking, so a significant area of unoxidized aluminum is
suddenly exposed to LOX.
A distant third would be detonation of just that tank's segment of a
flight-termination linear shaped charge. (That assumes the charges are
in separate segments; I couldn't find a detailed description of the
system.) (Note that any such detonation would not likely have been
commanded - I can think of no reason to design in the ability to
separately command local charge segments in such a system.)
A secondary implication of "final milliseconds" is that whatever
happened was violent enough to stop data transmission from the stage, or
at least from the relevant parts of the stage.
Mind, a crack developing in the tank outer skin under flight loads then
unzipping rapidly would also explain the results - it just wouldn't
explain the statement about an "overpressure event".
On 6/29/2015 6:12 AM, Ian Woollard wrote:
*Elon Musk* @*elonmusk* <https://twitter.com/elonmusk> 5h5 hours ago
<https://twitter.com/elonmusk/status/615431934345216001>
"Cause still unknown after several thousand engineering-hours of review.
Now parsing data with a hex editor to recover final milliseconds."