Re: Boeing and Dreamliner

[aek@vib.usr.pu.ru (Alexander Kopilovitch)]

Robert I. Eachus wrote:

>Hyman Rosen wrote:
>
> > Ada made the Ariane 5 crash!
>
>No, stupid management decisions made Ariane 5 crash.

No, simple "stupid management" is not enough here. Something more technical
was rotten. There was not single decision (or couple of decisions) taken by
top level manager at the last moment - there was nothing like ordering a
launch despite inappropiate weather. It was a long project with many presumably
competent people involved; and it is important to investigate more deeply:
why consequences of initial wrong decision (which was made by incompetent
people) were not recognized by scientific and technical staff before the
actual failure happened.

>This is one of 
>those stories where the truth really needs to catch up to the rumor. 

It must be not so simple, because the "truth" position is vulnerable - I think
that I'd feel myself rather comfortable in devil's advocate role for that
dispute (despite of my ignorance - your excellent description/explanation of
the case is enough for me -:) .

>Some brilliant management type had the idea that reusing the flight 
>control software from the Arianne 4 on Ariane 5 would save lots of money 
>on testing and verification.  As a result, and for political reasons, 
>there was no Ariane 5 contractor who even got to see the Ariane 4 source 
>code.

Beautiful decision. But engines were new, so there was a contractor for them,
right? And that contractor probably has own scientists and engineers, right?
Were those scientists and engineers interested in successful flights? Probably
yes. Were they informed about the decision to use old flight control software
unchanged and untested for their new engines? If yes then how can they agree
to that decision? If not, why they were not alarmed by the absence of combined
testing?

>...
>The engines were 
>commanded to deflect beyond what the stack could take, and the Arianne 5 
>broke up.  How could THAT happen?  The reuse included the flight 
>dynamics profile for the Arianne 4!  Since the Arianne 4 had smaller 
>moments and a larger tolerance for guidance inputs, ANY significant 
>correction sent to the engines, such as hitting wind shear, would have 
>pushed the software into a regime where errors would build up. 
>Eventually the commanded input would exceed the stack's structural 
>limits and destroy everything.
>
>There was some indication that such errors had occured during the 39 
>seconds of flight, but all had been small enough errors to be damped out 
>by other deviations.

Obviously, all that would be caught during the tests (even if engines were
simulated, those errors would be caught)... Perhaps this is (in)famous
"culture barrier", but I can't get how can one even think about avoiding
combined system testing for such complex and costly systems.



Alexander Kopilovitch                      aek@vib.usr.pu.ru
Saint-Petersburg
Russia