SoylentNews is people
SoylentNews
Boeing's failed Starliner mission strains 'reliability' pitch:
Boeing Co’s (BA.N) stunted Friday debut of its astronaut capsule threatens to dent the U.S. aerospace incumbent’s self-declared competitive advantage of mission reliability against the price and innovation strengths of “new space” players like Elon Musk’s SpaceX.
Boeing, the world’s largest aerospace company, has anchored its attempt to repel space visionaries like Musk and Amazon.com (AMZN.O) founder Jeff Bezos partly on its mission safety record built up over decades of space travel.
While SpaceX and Bezos’ Blue Origin are racing to send their own crewed missions to space for the first time, Boeing or Boeing heritage companies have built every American spacecraft that has transported astronauts into space. And the single-use rockets it builds in partnership with Lockheed Martin Corp (LMT.N) have a virtually unblemished record of mission success.
“We are starting from a position of mission reliability and safety,” Boeing Chief Executive Dennis Muilenburg told Reuters earlier this year when asked about SpaceX and other insurgents aiming to disrupt Boeing on everything from astronaut capsules to rockets to satellites.
“There is a difference between putting cargo in space and putting humans in space, and that’s a big step. Our very deliberate, safety-based approach for things like CST-100, that will be a differentiator in the long run,” Muilenburg said.
The actual technical glitch that stunted Friday’s CST-100 Starliner mission to the International Space Station was a timer error though Boeing said it was too early to determine the exact cause of the fault.
Boeing was already working to surmount other technical and safety-related challenges on the multibillion-dollar NASA human spaceflight program. A government watchdog report in November found Boeing demanded “unnecessary” new contract funds from NASA.
Friday’s glitch adds to a year of intense scrutiny over how Boeing developed its money-spinning 737 MAX jetliner following twin crashes that killed 346 people in five months.
While there is no link between the 737 MAX crashes and the Starliner setback, one rocket industry executive told Reuters that in both cases problems arose as Boeing was racing to catch up with fast-moving rivals.
(Score: 2) by Coward, Anonymous on Monday December 23 2019, @09:01PM (3 children)
With one trial and one failure you can actually put a decent limit on the reliability using Bayes' rule and Jeffreys' prior. If I've done the math right, the chance that the reliability is better than 90 % is only about 1.4 % (assuming a Bernoulli process).
Of course they will improve the reliability.
(Score: 2) by JoeMerchant on Monday December 23 2019, @09:37PM (2 children)
Just a few of the problems with reliability statistics are: incorrect assumptions are made, the people doing the calculations don't usually provide enough documentation to check or reproduce their work, the people who order the statistics to be calculated have an agenda, the people who report the statistics have an agenda, the people who repeat the reported statistics have no idea what they actually mean but only repeat them when they back up their agenda.
In this case, the process that developed and tested that rocket is not static, which I assume is not factored into your Bernoulli process? To correctly model a dynamic process would require a tremendous amount of data, or assumptions, usually both, and the extrapolations would still be fraught with very wide uncertainty windows.
All in all, would I rather fly on this rocket which has an N of 1 demonstration of no life threatening failures, or another rocket with 0 test flights? I'd go with this one every time - particularly with the post-failure lessons learned training which should illuminate what went wrong and improve training in the areas that had the problem.
At what point do you abandon the program in favor of a competing program that is farther from flight readiness? That's all about where the pork flows and what else is on the docket this session.
🌻🌻🌻 [google.com]
(Score: 2) by Coward, Anonymous on Monday December 23 2019, @10:48PM (1 child)
I wrote already that reliability would improve and I'm not disagreeing with you on that.
The calculation is more out of idle curiosity, because I've always been interested in what can be deduced in a rigorous way from very little data. If you're interested, here is a PDF [google.com] of the rather short Mathematica calculation notebook (since you mentioned documentation).
I'm not sure I would make that choice. It was a bad error that throws their whole process into question. I guess it depends on who built the other rocket. But Muilenberg's statement that Boeing was "starting from a position of mission reliability and safety" is turned on its head now.
(Score: 3, Interesting) by JoeMerchant on Monday December 23 2019, @11:18PM
For a couple of years my job was to shoot down proposals that tried to justify "process improvements" with insufficient data to back up the safety of the changes. The typical proposal that came to my desk was an N of 5 or 10 with binary pass-fail results, to which I would typically reply "you took the same company stats class that I did, you know we need 95-5 and you propose this?" to which the typical answer was "62 samples is impractically expensive" to which I would reply "can't you at least put a micrometer on it or something to get some continuous data where you might obtain 95-5 with 10 or 15 samples?" to which they would reply "I don't have time for that, this has to be done within 5 days or I lose my quarterly bonus" to which I would have to reply "sorry, out of my jurisdiction..." Once, I actually went to bat for the guy to keep his bonus because he spun a sufficiently compelling sob story, but usually they're just trying that as an emotional appeal to let the rules slide - and it was in my job description (and theirs) to not let the rules slide.
Absolutely, and they deserve a timeout in the penalty box to reflect on the error of their ways - ways of their errors, etc. If the other rocket were built by a roughly equivalent capability company with a roughly equivalent safety record and recent performance history... it's still a tough call, assuming that the N of 0 test rockets before manned launch is such because they were rushed into launching... Apollo I is what you get for rushing.
🌻🌻🌻 [google.com]