Quote from: joek on 11/06/2015 09:58 pmAll of the capabilities in the table below are required. Note that the only provider who can meet all requirements is SNC (based on best information available at this time).SpaceX or Boeing could tick the "pressurized disposal" box just by saying they can pack it in their vehicle, land it, and take it to the nearest landfill.
All of the capabilities in the table below are required. Note that the only provider who can meet all requirements is SNC (based on best information available at this time).
How is disposal on a dump or incineration facility on earth less desirable than disposal in a container burning up in the high atmosphere?Or: Why would disposal by SpaceX on the ground not be equal to disposal by Orbital Sciences Cygnus burning up in the high atmosphere?
If you want a more accurate analysis, we could use (n+1)/(k+2) on both launch vehicle and spacecraft independently to derive the probability of failure of at least one system on the return to flight.
There are other advantages to having a disposal vehicle like Cygnus. Running experiments like the one where they plan start a fire inside Cygnus.
Quote from: TrevorMonty on 11/07/2015 10:12 pmThere are other advantages to having a disposal vehicle like Cygnus. Running experiments like the one where they plan start a fire inside Cygnus.Always wondered why not giant room sized garbage bags with some disposable thruster device to guide them down to a fiery death?
after a failure an attempt is made to fix the root cause, so there's a smaller chance of failure after the fix.
Quote from: ChrisWilson68 on 11/07/2015 10:00 pmafter a failure an attempt is made to fix the root cause, so there's a smaller chance of failure after the fix.Prove that?Why is an implementation of the fixed design less likely to fail than the implementation of the original?
Thanks, that example is really helpful in seeing this from your perspective.I don't see it that way, though. Are you by any chance familiar with what's called the "Swiss Cheese" model?https://en.wikipedia.org/wiki/Swiss_cheese_modelIf you replace one slice of cheese, even if it has fewer holes, isn't there a good possibility that you're opening a new passageway for hazards to get through?
Suppose, for example, you have 1,000 design decisions and the chances of a fatal flaw being made in the design of each and not being caught is 1/10,000. Then you have a bit under a 10% chance that at least once fatal design flaw will get through. But if you find that flaw and replace it, if the replacement again has a 1/10,000 chance of being a fatal flaw that isn't caught, your chances of the replacement also being bad are 1/10,000.
Quote from: ChrisWilson68 on 11/08/2015 04:51 amSuppose, for example, you have 1,000 design decisions and the chances of a fatal flaw being made in the design of each and not being caught is 1/10,000. Then you have a bit under a 10% chance that at least once fatal design flaw will get through. But if you find that flaw and replace it, if the replacement again has a 1/10,000 chance of being a fatal flaw that isn't caught, your chances of the replacement also being bad are 1/10,000.The chance of an attempted design fix not working is much greater than 1 in 10,000. There have been order 500 failed orbital launches in the history of rocketry and Orbital's Taurus had two consecutive fairing failures, so from that incident alone the chance of a fix failing is at least 1 in 500.
Falcon 9 - 90%
Quote from: ncb1397 on 11/07/2015 05:58 pmFalcon 9 - 90%18 out of 19 is 90%?
you need to make some assumption about prior probabilities
ncb1397 decided to make up an arbitrary formula for calculating probabilities
if n == k it would give a 100% chance of success if so far there hasn't been a failure. Adding 1 to the numerator and 2 to the denominator is a fudge to try to get it to give a non-zero chance of failure even if there hasn't been a failure yet, but a decreasing chance of failure for every trial which does not result in a failure.
Quote from: ChrisWilson68 on 11/09/2015 05:43 amyou need to make some assumption about prior probabilitiesThis is a really good point, and shouldn't be overlooked.Quotencb1397 decided to make up an arbitrary formula for calculating probabilitiesNo he didn't. Ed Kyle has been using this approach on his spacelaunchreport website for years.
Quoteif n == k it would give a 100% chance of success if so far there hasn't been a failure. Adding 1 to the numerator and 2 to the denominator is a fudge to try to get it to give a non-zero chance of failure even if there hasn't been a failure yet, but a decreasing chance of failure for every trial which does not result in a failure.It does more than that, it provides the assumed probability distribution. (I think a Bayesian would say, "The prior being used is 1/2.")
----Although this appears unrelated to CRS-2, the connection is that NASA ///might/// be delaying the decision until after it has data on F9-FT and/or Cygnus on Atlas V. So if I may beg your indulgence (and that of other readers)? Suppose repeated identical trials of a launch system had a perfect record of 99 successes, and then one failure. The root cause is determined and a fix is designed. Would you rather bet your life on another trial identical to the 99 successes and the one failure, or would you choose to bet your life on the system that had never flown?
Suppose repeated identical trials of a launch system had a perfect record of 99 successes, and then one failure. The root cause is determined and a fix is designed. Would you rather bet your life on another trial identical to the 99 successes and the one failure, or would you choose to bet your life on the system that had never flown?