Upper-Stage Starship Failure, Test 8

The video on the SpaceX website is disabled to black.  I cannot watch it.

As for the two upper stage “Starship” failures in a row,  bear in mind that most flight vehicle failures in flight test are due to multiple things acting together.  Having only a single cause is unlikely in the extreme.

That being said,  after flight 7,  they made some changes to the vehicle for flight 8, which apparently  did not work.  Which strongly suggests that the real causes (plural!) were not the leaks in engine plumbing that they assumed after flight 7.

There was a significant vehicle design change after flight 6.  The vehicles for flights 7 and 8 were longer,  with larger propellant tanks.  Which raises the specter of some sort of slosh or other mode in the propellant tanks,  causing the excessive vibration,  and perhaps causing fatal leaks in weld joints of the lower tank aft bulkhead.  The larger propellant masses involved would likely amplify any such effect.  This is only speculation,  but it is a real possibility that they need to explore.

If it were me,  I'd re-fly the older Starship upper stage design on flight 9,  with the smaller tank volumes and shorter length.  If that design makes the ascent successfully,  when the “improved” design did not,  twice in a row,  that would pretty well “nail it” to the dynamics of the larger tanks vs smaller tanks.

If it's liquid-sloshing in the tanks that really is the source,  then change the baffles.  The tank structures in and adjacent to the aft bulkheads may need reinforcement to better resist the unanticipated loads.  They need to instrument these locations for possible effects,  probably to include in-tank camera views,  and some strain gages and pressure sensors.

That's not to say there might not also be failures in the engine plumbing,  too!  But whatever is going on there, is apparently being overwhelmed by something else they have not identified yet.

This sort of thing happens often in experimental flight test.  It shows up more frequently when you make too many changes to the vehicle between tests,  too early in the program.  In this particular case,  it may also trace to believing too strongly in computer code outputs,  by engineers who cannot detect a garbage-in/garbage-out problem,  because they have not done enough (or cannot do) “old-timey” pencil-and-paper design analysis. 

Same day update:  there was some sort of pogo-mode excessive low-frequency vibration that afflicted the first 2 Saturn-5 flights.  I do not remember what the cause and cure for this were,  but it did not afflict the next test flight,  which was manned.  That was Apollo 8 around the moon December 1968. 

Update 3-13-2025:  I looked up the history of the Saturn-5 and POGO oscillations.  The unmanned test flight of Apollo-6 had severe POGO in the first stage,  doing enough damage to the second and third stages as to compromise the mission entirely.  Had it been manned,  this likely would have been an abort.  

There was severe-enough POGO in the Apollo-13 second stage as to cause a premature engine shutdown,  compensated by longer burns of the remaining 4 second stage engines,  and of the third stage engine.  These failures were overshadowed by the oxygen tank explosion later in the mission that led to the crippled craft aborting home without entering lunar orbit and doing the landing.  Man-rating the Titan-2 booster for Gemini was delayed because of POGO instabilities.  This was a rather common problem in a lot of rockets.

I will add this:  detecting this correctly requires the right instrumentation,  and people who know how to use it.  In the tactical solids I worked on,  combustion noise has frequencies of a few hundred Hertz,  of amplitude about 1-5% of the basic pressure level.  If it got to 10% or more of the pressure level,  we considered it "combustion instability" to be remedied.  Usually there was a discernible higher-amplitude instability signal at a frequency we could determine,  standing out from the lower-amplitude combustion noise at a "broad hash" of frequencies,  up to something approaching a significant fraction of a  MegaHertz.  

To see this behavior in the data,  you simply cannot use modern digital data acquisition!  That "pixelates" the signal too much to see what is going on.  Nobody wants to pay for digital systems capable of resolving those high frequencies.  You must use an analog tape recorder that uses FM data processing,  with a response of about 1 MegaHertz or higher.  You play that data back and display it in multiple plot formats until you can see what you need to see.  There are few today who actually have equipment like that anymore,  much less who know how to use it to diagnose combustion instabilities.  And yet,  POGO instability is said to result from combustion instability interacting with any or most of an abundance of longitudinal modes in assorted tanks and plumbing.

I had long experience with data from equipment like that,  diagnosing these kinds of phenomena.  While too old to be seeking regular employment,  I would be happy to consult in this area.  Please contact me if you need help.