[AR] Re: What blew up Crew Dragon...

  • From: John Schilling <john.schilling@xxxxxxxxxxxxxx>
  • To: arocket@xxxxxxxxxxxxx
  • Date: Fri, 19 Jul 2019 17:21:49 -0700

On 7/15/2019 10:03 PM, Henry Vanderbilt wrote:

On 7/15/2019 7:56 PM, Henry Spencer wrote:

The problem is that check valves don't reliably block slow reverse flow of *gas*, and so a volatile propellant can seep up past the check valve and condense in colder plumbing upstream. This is a known problem, and has been for decades!  In the case of N2O4, such seepage can also corrode upstream components. (This is almost certainly what really happened to Mars Observer, whose helium pressure regulators were *not* rated for N2O4 exposure -- when the pressurization system was activated, the corroded regulators failed to control the helium flow, and the propellant tanks burst.  Once this possibility was noticed, the regulator failure was successfully duplicated in the lab.)  So just taking it slow on the pressurization is not sufficient.

The one time I was involved in an incident investigation where fuel and oxidizer had met illicitly and noisily, I ended up spending some considerable time testing out that migrate-as-vapor-then-condense-in-a-bad-place possibility.  FWIW, in the real world it's quite difficult to make that happen.

Check valves are prone enough to just flat-out sticking and leaking fluids that more exotic explanations aren't often needed.

Note that the propellants don't have to condense as liquids to do bad things.  Vapor-phase reaction of hydrazine and nitrogen tetroxide produces a solid residue that precipitates on the nearest convenient surface.  Which will very often be just inside one of your check valves, where the leaking vapor of one component has its first opportunity to meet a pre-existing saturated vapor of the other.  The precipitate is a detonable high explosive, but you'll probably get a boring failure before you ever see that exciting one - because the precipitate is also either a gummy or hard crystalline solid depending on e.g. how methylated your hydrazine was, but either one in the working parts of a check valve will tend to make the valve stick.  If the check valve sticks closed, you probably can't pressurize your feed system and you can't run your engines properly.

This was implicated in the failure, er, "anomalous behavior leading to premature shutdown" of the main propulsion systems of the Juno and Akatsuki space probes, and several other missions that I'm going to be vague about.  Common pressurization systems with check-valve isolation are good for about a month with high confidence.  At least with traditional hypergols; if you come up with something new, you'll have different reactions at different rates with different products.  It's still a fairly common scheme for communications satellites that are only supposed to use their bipropellant systems for orbit transfer in the first few weeks of the mission, but probably best to just not do that.

Also, and particularly for series-redundant check valves that get to arguing about which one of them is supposed to be checking the flow, they  are prone to chatter among some flow conditions, which damages the seats and produces leakage.   Figure out what flow conditions do that, and avoid them.  Do NOT, NOT NOT NOT, use audible chatter of the check valves as a ground-test diagnostic for "yep, the check valves are unstuck and working".  And don't ask me how I know that.

And don't trust them below -40 deg C, if that comes up (and  it might in a fast-flowing blowdown helium system).  If the spec sheet says they are qualified for that, it was probably twenty years ago with a grade of teflon seat material that current production can't match.

        John Schilling
        john.schilling@xxxxxxxxxxxxxx
        (661) 718-0955



Other related posts: