Cause of AMOS-6 failure

On September 1 last year, a SpaceX Falcon 9 rocket exploded during fueling for a engine test that SpaceX does before every launch.

Since then, SpaceX has been analyzing the data that they gathered before and during the anomaly.  Recently they posted this report detailing what happened.

The report says that the second stage exploded due to heat or pressure coming from something called COPV.  COPV stands for composite overwrapped pressure vessel, which is a kind of tank that, rather than being made out of a strong material, uses tightly wrapped carbon fiber wrapped around a thin internal aluminum tank.  This is simply the lightest way of making high pressure tanks (in this case, around 350 bar, or ~5000 psi), and was used extensively on the space shuttle.

This is what a cutaway of one looks like:

Here's a video of one being wrapped:

This kind is filament wound, which is what the Falcon 9 has.  Braided ones are stronger and don't fail as explosively (not that it would have mattered) and are even more fun to watch being made (this is just carbon fiber being braided, not over a tank): https://www.youtube.com/watch?v=E9_21CDLoBo

SpaceX uses these on their Falcon 9 in the second stage to hold helium that is used to fill up the tank as the fuel is pumped out, as the tank is airtight, it would crumple if pressure was not maintained.

There are three tanks in the second stage, visible in this image:

Here's a size comparison of one of those which survived reentry:

So, the report says that liquid oxygen got between the liner and the overwrap.  Normally this wouldn't be a problem, but somehow a buckle formed in the liner, like the way you can pop in the side of a plastic water bottle.  When the liquid oxygen pooled in that buckle, it may have also frozen, because SpaceX uses liquid oxygen that is not just liquid, but much colder and closer to solid oxygen (the kerosene fuel is also much colder than usual), which increases the amount of propellant that can be stored in a tank of the same size, and the performance of the engines, which is improved because more propellant can flow into the engine with pipes of the same size.

SpaceX will first change the fueling procedure, which worked perfectly on Iridium NEXT, and later they will change the COPV design.

Some articles have said that SpaceX's failures have been caused by a more silicone valley iteration based mentality, however I think that they have more to do with the fact that SpaceX is developing the first new rocket technology since the '70s.  If you look at failure rates from rockets from that time, you see even higher failure rates.  The Falcon 1 (the first SpaceX rocket) had a failure rate of 60% (3 out of 5 launches over 3 years) and the Falcon 9 has a 6.897% launch failure rate (1 out of 28, over 7 years)  not including partial failures.  The AMOS-6 failure is counted here despite not being a launch failure, as the payload was destroyed.

The total failure rate for all Falcon rockets is 14.71%, 5 out of 34, over 11 years.

Now, compare that to the Atlas ICBM variant D which flew 67 times in its ICBM (non Mercury-Atlas) configuration over 4 years (1959-1962), and had a 35.82% failure rate.  In Febuary of 1962, after 68 flights with 36.76% or 25 of those resulting in failure, NASA actually began launching this rocket with humans aboard.  This isn't as crazy as it sounds, as the launch escape system in use at the time had never failed.

Most of this information came from this image:

So, in short, I think that SpaceX's higher failure rate are caused by flying the most recently developed rocket.  There's actually a term for how new technologies fail more often while new, called the bathtub curve.