Flight test is a harsh, unforgiving business.
Modern computer modeling can sometimes give you an incorrect appreciation of this. You think you have all the angles figured. You think you've accounted for everything. You think you've analyzed the stresses down to the last nut, bolt, rivet and weld. But there's a small problem ... one of the unfortunate implications of Godel's Incompleteness Theorem is that every model is necessarily incomplete. That is, the only complete model of a thing is the thing itself.
So ... the only way to find out how the thing behaves is to go try it out for real. To try it in the real world, where there are few (and sometimes only one) way to succeed, and innumerable ways to fail.
That's not always a bad thing, by the way; failure can be instructive. But those lessons sometimes come with a staggeringly high price tag.
On the 28th of October, Orbital Sciences Corporation had an Antares launch vehicle sitting on the pad at Wallops Island in Virginia, set to carry a Cygnus resupply pod to the International Space Station. The Antares rocket's first stage is powered by two AJ-26 rocket engines. The AJ-26 is a repackaging by Aerojet General of the Soviet-built NK-33. Originally, the NK-33 was intended to be the main engine for the N-1 lunar booster, with thirty in the first stage and eight in the second stage.
Someone ... oh, God, someone should have thought this through. Because the N-1 "flew" four times, without much variation in the outcome.
Anyway, recycling old rocket bits is Orbital's stock in trade. When they found a bunch of perfectly good rocket engines sitting in a warehouse, someone thought "Score!" Which is how two ex-Soviet engines ended up in an American rocket. Engines, by the way, that were notorious for trying to eat their own turbopumps. Which is more or less how this happened.
At some point you have to wonder if the actual purpose of the NK-33 was the conversion of fuel and structure into shrapnel and combustion by-products. Needless to say, Orbital is now looking for a new engine for Antares. (Addendum, 27Dec14: I'd have expected them to look for a non-Russian engine, but ... yeah, they're using another Russian engine. On the upside, the RD-181 is built in the same factory as the RD-180. The RD-180 has proven fairly reliable so far. On the downside, this puts Antares at the mercy of whatever the hell Putin decides to pull this month. They may yet regret this choice.)
The good news is that nothing was hurt, apart from OSC's pride. The same can't be said for the test mishap suffered by VSS Enterprise.
On October 31st, VSS Enterprise dropped from its White Knight Two carrier airplane for its fourth powered flight, and its thirty-sixth flight overall. Virgin and Scaled Composites had, by all accounts, been pursuing a long and fairly conservative flight test program, expanding the flight envelope bit by bit with each new powered flight. It should have been a fairly routine outing. The full NTSB report won't be available for some time, but what seems to have happened is that the copilot hit the "feather" switch too soon, while the rocket engine was still burning. The sudden nose-down pitch caused the ship to break up in flight, killing both crew on board. (Addendum, 27Jan15: Rarely have I been happier to have been wrong. Peter Siebold, while seriously injured, was not killed in the crash.)
I'm reserving judgement, here. I don't know enough about how it operates ... but it does seem to me that switch should have a guard of some kind. By way of comparison, on the C-130, the paratrooper "jump" light switch has a metal flange that will not allow you to turn that light on unless you've already opened the jump door. And maybe there's already something like that there, and the crew got confused.
Truth is, we just don't know yet, and probably won't know until the report is released.
There is some good news to go along with the bad, though. In early December, there was a test flight of the new Orion crew capsule. NASA looks like they've shook off whatever was ailing them back in their Ares-1X days. On December 5th, a test version of the Orion spacecraft was launched on a Delta IV Heavy booster on a two-orbit test flight. The launch and re-entry both came off without a hitch.
I had heard Ares-1X described as "a low-fidelity test of a bad design." It's not really fair to say that of Orion. Orion took two passes through the Van Allen radiation belts, and the Delta second stage fired a second time on the last orbit, to accelerate into re-entry. The peak deceleration was 8.5 times the force of gravity. This was a fairly rigorous shakedown cruise. It was a legitimate test that NASA could have failed.
Now, it's probably still true that Dragon V2 will beat a manned Orion into space, because Orion is paced by NASA funding and Dragon is paced by the whims of an eccentric billionaire, but once it does fly, Orion will be a top-notch ship. Of that, I'm very confident ... and a few years ago, I wasn't.
Flight test is a harsh, unforgiving business. Its lessons are paid for in blood. But we do learn, and in the end, the lessons are worth it.