Author Topic: Abort options for Starship and Starship/SuperHeavy  (Read 311129 times)

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1400 on: 07/03/2023 03:06 pm »
I want to point out that the risk that SpaceX would achieve by flying Starship over 100 times in a row without failure first will be lower than the 1:75 threshold NASA has for Artemis 3, and likely would be in line with the 1:150-270 standard for commercial crew missions.

Just because you keep asserting this doesn't make it any truer.

100/100 reliability:

95% CI Adjusted Wald Interval: 96.83% - 100.00%.  Best estimate: 99.02%. Empirically, that would be pLOC=1:125 pLOC=1:102

If failures are random and independent, then the probability of 100 consecutive successes is 51% if the true failure rate is 1:143. This is the "more likely than not" confidence level, and any higher confidence level is double dipping.

What confidence level does NASA use for reliability estimates?

You were the one who had that chart for what SMD required for the launchers used on the various classes of science missions, weren't you?  I apparently didn't bookmark that and can't find it any more.  That's a really useful chart.  Do you still have a pointer to it?

IIRC, that was all based on 50% CI confidence level, which made sorta-kinda sense for robotic missions on new launchers.  Whether that same reasoning would apply for human spaceflight, I don't know.
« Last Edit: 07/03/2023 03:08 pm by TheRadicalModerate »

Offline Robotbeat

  • Senior Member
  • *****
  • Posts: 39463
  • Minnesota
  • Liked: 25589
  • Likes Given: 12245
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1401 on: 07/03/2023 03:34 pm »
It's weird to have a "confidence level for a confidence (reliability) level." Trying to reach a 95% confidence interval is, to me, just navel-gazing about probability distributions, and the same resources would be better spent improving (reducing) the expectation value for the fatality rate.

In fact, because "confidence interval" is talking about spread of the data, it's actually possible for a MUCH improved number to REDUCE the 95% CI reliability.

For instance, if I have a population of 91, 89, 91, 89, 90, the mean will be 90, but the 95% CI lower bound will be 89.123, but if I instead have a population of 91, 89, 91, 89, 99, then the mean will be higher (91.8),  BUT the 95% CI lower bound will be WORSE at 88.165. So purely by adding a much better value in the population (and making NOTHING worse), you can make the 95% CI look worse simply because the spread is greater!


That's why I think it's wrong to use 95% CI and instead you should use the expected value (the mean). Don't let navel gazing about how tight your probability distribution is get in the way of improving the actual expected value. It sounds good and conservative to use 95% CI, but everyone is much better served to simply use a more conservative expected reliability threshold instead.
« Last Edit: 07/03/2023 03:44 pm by Robotbeat »
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline Coastal Ron

  • Senior Member
  • *****
  • Posts: 9266
  • I live... along the coast
  • Liked: 10734
  • Likes Given: 12343
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1402 on: 07/03/2023 06:39 pm »
The Titan/Oceangate accident may represent a failure ratio that we hope SpaceX never accepts, but it does represent how much risk "normal" people will accept in pursuit of unique experiences.
You're assuming that the normal people were well-informed, and the information they received was accurate.

Adventure-seeking billionaires are not capable of being "well-informed"?  ::)

Look, we all know that liability release forms will be signed before getting into a Starship, so I'm not talking about financial risk to SpaceX. What I'm saying is that people are allowed to assume risk that is above what "normal" people would assume, and that is not against the law. It is not even immoral as long as the person going has been allowed to access the internet and see for themselves the history of SpaceX test flights.

It would be the rare person today that was not aware of the potential risks to flying to space, on any form of transportation, much less one that is so completely different.

People have the "freedom" to make risky decisions. We have whole industries that promote and extoll risky behavior, and spaceflight is just one small category of them. Let's keep some perspective here...
If we don't continuously lower the cost to access space, how are we ever going to afford to expand humanity out into space?

Offline DigitalMan

  • Full Member
  • ****
  • Posts: 1734
  • Liked: 1222
  • Likes Given: 76
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1403 on: 07/03/2023 06:49 pm »
The Polaris 3 crew is likely to be the first to fly in Starship, you know who to ask about risk management

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1404 on: 07/03/2023 08:00 pm »
It's weird to have a "confidence level for a confidence (reliability) level." Trying to reach a 95% confidence interval is, to me, just navel-gazing about probability distributions, and the same resources would be better spent improving (reducing) the expectation value for the fatality rate.

The two have nothing to do with one another.  One's a measurement, and one's an activity.  You use the measurement as a guide to how well you're doing with the activity.

But the people who care about the measurement are the people who plan to use the system that you're trying to improve.  The measurement is really all they have to make informed judgments on whether what they want to do has a low enough risk.

Quote
In fact, because "confidence interval" is talking about spread of the data, it's actually possible for a MUCH improved number to REDUCE the 95% CI reliability.

For instance, if I have a population of 91, 89, 91, 89, 90, the mean will be 90, but the 95% CI lower bound will be 89.123, but if I instead have a population of 91, 89, 91, 89, 99, then the mean will be higher (91.8),  BUT the 95% CI lower bound will be WORSE at 88.165. So purely by adding a much better value in the population (and making NOTHING worse), you can make the 95% CI look worse simply because the spread is greater!

That's why I think it's wrong to use 95% CI and instead you should use the expected value (the mean). Don't let navel gazing about how tight your probability distribution is get in the way of improving the actual expected value. It sounds good and conservative to use 95% CI, but everyone is much better served to simply use a more conservative expected reliability threshold instead.

Let's start with what we'd like to measure, which isn't really a mean.  Instead, because we're dealing with just success or failure, we want the true success (or failure) rate of a Starship mission.  That's a single statistic, lim(N→∞, Ns/N), where Ns is the number of successes in N missions.

Since we obviously don't have infinite missions to sample, we use a finite sample size, i.e., a series of missions, over which we measure the success rate.  Then the task is to decide how close the measured success rate of our sample is likely to be to the true (unknowable) success rate.

To quantify how close, we use the confidence interval, computed to satisfy a particular confidence level.  For a particular sample (i.e., for one set of a series of sets of missions, each with the same number of missions), the true success rate should fall within the confidence interval a <confidenceLevel> proportion of the time, if you were to fly the same number of missions over and over.

So a 95% CI will be wider than a 55% CI, because there has to be a 95% probability that the true success rate falls in the former, but it only has to fall in the latter interval 55% of the time.  At the same time, increasing the number of missions in the sample will make both intervals narrower.

The CI is less and less useful as the sample sizes get smaller, and they're particularly unhelpful when you're dealing with a system where the success rate is engineered to be particularly high, because then only one of the two tails that lie outside the confidence interval are available.  But that's why you use Wald intervals for smaller sample sizes of success/failure trials.

None of this changes the fact that any reasonable sample size of launches is retrospective, not prospective.  So it has failures in it for things that you fixed, which makes the sample not truly independent.  That's why, once you have a good feel for how your system behaves, and what failures cascade in which directions, PRA starts giving you a better estimate for the reliability than pure empirical measurement.

But there's no engineering without measurement.  You use what you have until you have something better.  The real question is when "something better" (likely PRA) actually is better than empirical measurement.

As I said above, we have good PRA models for launches, and they map relatively cleanly onto Starship.  By mission #100, I'd expect that our handle on ascent reliability would be quite good.

But that's likely not the case for Starship EDL, which is a relatively unexplored failure space.  We're going to have "normal accidents" during EDL, based on the complexity of the system and our lack of experience with it.  There's really no good way to quantify how likely those accidents are, because most of them will result from failures of imagination.  In that case, the best data we have for decision-making will likely be good ol' frequentist success rates, and the confidence intervals associated with them.  What confidence level is considered good enough for that, we'll have to see.

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1405 on: 07/03/2023 08:19 pm »
Quote
You're assuming that the normal people were well-informed, and the information they received was accurate.

Adventure-seeking billionaires are not capable of being "well-informed"?  ::)

It's a two-part predicate I put on there.  You can only be well informed if the information you're using is accurate.

Quote
Look, we all know that liability release forms will be signed before getting into a Starship, so I'm not talking about financial risk to SpaceX. What I'm saying is that people are allowed to assume risk that is above what "normal" people would assume, and that is not against the law. It is not even immoral as long as the person going has been allowed to access the internet and see for themselves the history of SpaceX test flights.

It would be the rare person today that was not aware of the potential risks to flying to space, on any form of transportation, much less one that is so completely different.

People have the "freedom" to make risky decisions. We have whole industries that promote and extoll risky behavior, and spaceflight is just one small category of them. Let's keep some perspective here...

I have no doubt whatsoever that SpaceX can find people willing to take large risks to go into space more cheaply.  But there are two risks to SpaceX, in allowing them to do so:

1) The informed consent release is only good if SpaceX can prove that the information provided in the consent was accurate, and that SpaceX wasn't being reckless with the launch.  I'd bet that SpaceX would win that set of lawsuits, but I wouldn't bet overwhelmingly on it.  And the PR would be bad.

2) The even worse PR comes from the social media feeding frenzy, which can result in regulatory interference. 

So, yeah, people have freedom to make risky decisions.  And companies have the freedom to decide that their being involved in the risk isn't worth the benefit.

Offline Robotbeat

  • Senior Member
  • *****
  • Posts: 39463
  • Minnesota
  • Liked: 25589
  • Likes Given: 12245
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1406 on: 07/03/2023 08:25 pm »
SpaceX has a greater incentive to have high reliability than the government (NASA) does. That explains partially why Falcon 9 is the most reliable launch vehicle ever, by about a factor of 2.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline Coastal Ron

  • Senior Member
  • *****
  • Posts: 9266
  • I live... along the coast
  • Liked: 10734
  • Likes Given: 12343
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1407 on: 07/03/2023 08:32 pm »
Quote
You're assuming that the normal people were well-informed, and the information they received was accurate.
Adventure-seeking billionaires are not capable of being "well-informed"?  ::)
It's a two-part predicate I put on there.  You can only be well informed if the information you're using is accurate.

In the Titan accident, I'm not a billionaire (yet!) but even I know that carbon fiber construction is not good at resisting compression, and any billionaire would have access to plenty of professional engineers that could have evaluated the risks ahead of time. And if they don't, there is no amount of regulation that can fix stupid...  ;)

Quote
I have no doubt whatsoever that SpaceX can find people willing to take large risks to go into space more cheaply.  But there are two risks to SpaceX, in allowing them to do so:

1) The informed consent release is only good if SpaceX can prove that the information provided in the consent was accurate, and that SpaceX wasn't being reckless with the launch.  I'd bet that SpaceX would win that set of lawsuits, but I wouldn't bet overwhelmingly on it.  And the PR would be bad.

Elon Musk is pretty good about talking publicly about risk, and he has been warning people about the riskiness of going to Mars for many years now, so no one is going to be able to plead ignorance about going to Mars, and somehow they have to get to space first...

Quote
2) The even worse PR comes from the social media feeding frenzy, which can result in regulatory interference. 

So, yeah, people have freedom to make risky decisions.  And companies have the freedom to decide that their being involved in the risk isn't worth the benefit.

Yes, PR will likely be the biggest issue, but not by scaring off potential customers, but in causing the U.S. Government to step in and increase regulation in some way.
If we don't continuously lower the cost to access space, how are we ever going to afford to expand humanity out into space?

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1408 on: 07/03/2023 08:58 pm »
In the Titan accident, I'm not a billionaire (yet!) but even I know that carbon fiber construction is not good at resisting compression, and any billionaire would have access to plenty of professional engineers that could have evaluated the risks ahead of time.

Beg pardon, but aren't you making my point for me?  An "ordinary" billionaire got "informed" of the risks, which he believed were an accurate representation.  I imagine that the litigation that's sure to follow will cause OceanGate to implode faster than the walls of its submersible.

I doubt that similar litigation against SpaceX would kill the company.  But it sure wouldn't do it any good.

Quote
And if they don't, there is no amount of regulation that can fix stupid...  ;)

We fix stupid with regulation all the time.  What do you think seat belt and motorcycle helmet laws are for?

Quote
Elon Musk is pretty good about talking publicly about risk, and he has been warning people about the riskiness of going to Mars for many years now, so no one is going to be able to plead ignorance about going to Mars, and somehow they have to get to space first...

Even if Elon is good about publicly communicating risks about going to Mars (a debatable assertion), that's pretty far in the future.  Going to Mars has nothing to do with going to LEO on a Starship.  Jared Isaacman doesn't want to go to Mars (yet).  Neither do Maezawa and Tito.  Neither does NASA (yet).  But they'd all be very interested in using Starship for launch to LEO and EDL, if the risk profile were low enough.

Quote
Yes, PR will likely be the biggest issue, but not by scaring off potential customers, but in causing the U.S. Government to step in and increase regulation in some way.

Which is exactly what I said, wasn't it?

Offline envy887

  • Senior Member
  • *****
  • Posts: 8210
  • Liked: 6922
  • Likes Given: 2975
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1409 on: 07/04/2023 11:52 am »
I want to point out that the risk that SpaceX would achieve by flying Starship over 100 times in a row without failure first will be lower than the 1:75 threshold NASA has for Artemis 3, and likely would be in line with the 1:150-270 standard for commercial crew missions.

Just because you keep asserting this doesn't make it any truer.

100/100 reliability:

95% CI Adjusted Wald Interval: 96.83% - 100.00%.  Best estimate: 99.02%. Empirically, that would be pLOC=1:125 pLOC=1:102

If failures are random and independent, then the probability of 100 consecutive successes is 51% if the true failure rate is 1:143. This is the "more likely than not" confidence level, and any higher confidence level is double dipping.

What confidence level does NASA use for reliability estimates?

You were the one who had that chart for what SMD required for the launchers used on the various classes of science missions, weren't you?  I apparently didn't bookmark that and can't find it any more.  That's a really useful chart.  Do you still have a pointer to it?

IIRC, that was all based on 50% CI confidence level, which made sorta-kinda sense for robotic missions on new launchers.  Whether that same reasoning would apply for human spaceflight, I don't know.

I have referenced.that document, yes. It is NASA policy directive 8610.7.

https://nodis3.gsfc.nasa.gov/displayDir.cfm?t=NPD&c=8610&s=7D

50% confidence in 95% reliability (14 consecutive successes) is required for flagship payloads in the absence of any other flight data.

The failure rate requirements for crew rating are apparently quite different from those for flagship payloads, judging by NASA's requirement of a 1:270 threshold for commercial crew.
« Last Edit: 07/04/2023 02:20 pm by envy887 »

Offline envy887

  • Senior Member
  • *****
  • Posts: 8210
  • Liked: 6922
  • Likes Given: 2975
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1410 on: 07/04/2023 12:01 pm »
In the Titan accident, I'm not a billionaire (yet!) but even I know that carbon fiber construction is not good at resisting compression, and any billionaire would have access to plenty of professional engineers that could have evaluated the risks ahead of time.

Beg pardon, but aren't you making my point for me?  An "ordinary" billionaire got "informed" of the risks, which he believed were an accurate representation.  I imagine that the litigation that's sure to follow will cause OceanGate to implode faster than the walls of its submersible.

I doubt that similar litigation against SpaceX would kill the company.  But it sure wouldn't do it any good.

Similar litigation would get quickly thrown out of court, at least in Florida. State law provides broad immunity to the space flight provider for civil liability for death or injury of crew or passengers during a space flight.

They would have to show at least gross negligence, and lack of an full-envelope abort system would not qualify, since it is not an industry standard. SS2 is operating without one, and Shuttle did the same for decades.

Offline Coastal Ron

  • Senior Member
  • *****
  • Posts: 9266
  • I live... along the coast
  • Liked: 10734
  • Likes Given: 12343
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1411 on: 07/04/2023 04:00 pm »
In the Titan accident, I'm not a billionaire (yet!) but even I know that carbon fiber construction is not good at resisting compression, and any billionaire would have access to plenty of professional engineers that could have evaluated the risks ahead of time.
Beg pardon, but aren't you making my point for me?  An "ordinary" billionaire got "informed" of the risks, which he believed were an accurate representation.

My point was that the billionaire had the ABILITY to be fully informed, if they so desired. If they choose to ignore those risks, isn't that on them?

Quote
Quote
And if they don't, there is no amount of regulation that can fix stupid...  ;)
We fix stupid with regulation all the time.  What do you think seat belt and motorcycle helmet laws are for?

Seat belts and motorcycle helmets are not mandated because people are stupid, it is because they are affordable ways to increase the likelihood of surviving an accident - even one that you had no part in creating. Like crumple zones on a car are affordable ways to increase survival rates in crashes, regardless who was at fault.

In this case "stupid" is when you don't rationalize the risks. For instance, being someone that LIKES to take risks, like the billionaire in the Titan accident, would show a court that he was someone that understood the risks, and had repeatedly put himself in positions where he could have died. I think he let his sense of adventure overcome his sense of safety, and he rationalized that the prior adventurers had survived, regardless of the risks of the design of the Titan.

My point being that every day we make choices that put our lives at risk, whether it is walking to the store instead of driving, driving instead of flying to our vacation destination, eating a new food, turning left instead of right, etc., etc., etc. Future Starship passengers will be rationalizing their choice to fly on a Starship that has greater-than-zero chance of death in case of an accident. And I don't think they will have a shortage of paying customers.

Quote
Quote
Yes, PR will likely be the biggest issue, but not by scaring off potential customers, but in causing the U.S. Government to step in and increase regulation in some way.
Which is exactly what I said, wasn't it?

I was agreeing with you  :D

I think it will be government regulation that will determine the amount of risk that SpaceX will be allowed to offer, with the question being how quickly that kicks in. The fewer the people that die in Starship accidents, the longer it will take for government regulation to be implements, so let's hope that SpaceX is able to build a reliable transportation system.
If we don't continuously lower the cost to access space, how are we ever going to afford to expand humanity out into space?

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1412 on: 07/04/2023 09:39 pm »
I have referenced.that document, yes. It is NASA policy directive 8610.7.

https://nodis3.gsfc.nasa.gov/displayDir.cfm?t=NPD&c=8610&s=7D

50% confidence in 95% reliability (14 consecutive successes) is required for flagship payloads in the absence of any other flight data.

The failure rate requirements for crew rating are apparently quite different from those for flagship payloads, judging by NASA's requirement of a 1:270 threshold for commercial crew.

Thanks for re-posting.

The regs specifically state that the 50% confidence level is computed assuming a binomial distribution, which is weird, since only a confidence interval is meaningful for this kind of computation.  I assume that they're taking the midpoint of the confidence interval and waving their arms.

Based on a calculator I found online, the 50% CI for 14 consecutive successes is 90.57% - 100%, which does indeed make the midpoint pretty close to 95%.  Apparently literally close enough for government work.

It's also notable just how much NASA seems to trust their independent verification and validation, design certification review, and engineering review processes.  Note also that the Class A mission option with 3 missions / 2 consecutive mission successes requires  a "full vehicle fishbone" vehicle analysis, which I assume is NASAese for a trustworthy failure tree, against which to run PRA.

I agree that human spaceflight certification will be considerably different, but this is an interesting window into how their thinking runs.  If the 1:270 pLOC (or, for our purposes, the 1:500 launch and EDL pLOCs, i.e., 99.8% pSuccess) are midpoints on a 50% confidence level, the success confidence interval would be 99.6%-100%, which would require 350 consecutive successes.

However, I'm sure that the "full vehicle fishbone" is an absolute requirement for crew certification.  Still, I'd expect that they'd want an empirical starting point to make them comfy with the PRA.

Knowing what the 50% CI midpoint they'd accept as an empirical starting point would be really nice to know.

95%?  We already know that one--14 consecutive successes
98%?  34
99%?  69
99.5%?  136

Can we get an inkling from D2 and Starliner?

The first Starliner OFT was Atlas V mission #80.
SpaceX Demo-2 was F9/FH mission #85, which was 57 consecutive successes.

So... 50% CI midpoint=99%?  (Yeah, this is an extremely thin inference.)

I guess 70 consecutive EDL successes would start to make one feel like the associated PRA had most of the fishbones.  But again, capsules and parachutes are pretty well-understood.  Flip-and-land is going to cause some handwringing.

Offline envy887

  • Senior Member
  • *****
  • Posts: 8210
  • Liked: 6922
  • Likes Given: 2975
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1413 on: 07/04/2023 11:26 pm »
Confidence, reliability, and number of consecutive successes are related by the formula:

log(confidence)/log(reliability)=successes

Plug in 50% and 95%, and the number of successes is 13.51. Since you can't have a fractional success, they would take the next integer. That's where NASA got 14.

Online meekGee

  • Senior Member
  • *****
  • Posts: 15698
  • N. California
  • Liked: 15851
  • Likes Given: 1443
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1414 on: 07/05/2023 03:17 am »
I can't believe this claim is serious.  One of the things we learned from Shuttle is not to try to make one vehicle do too many things.  Did we really forget that lesson already?
Of course it is serious.

For Starship to be successful (which means delivering huge numbers of people to Mars) it needs to do a whole bunch of things: take off from Earth, refuel, travel to Mars, EDL at Mars, take off from Mars, travel to Earth, EDL at Earth. That's a tall order, no question, but it is the fundamental design goal for the vehicle.

But that's a stupid goal.  The goal (if you think colonizing Mars is a good idea) should be to do all the things necessary to colonize Mars.  It doesn't all have to be done by the same vehicle and, in fact, doing it that way is very inefficient.

I have used, for work, vehicles between 62 pounds and 320,000 pounds, each one for a different purpose.

To use the 737 analogy, if that's such a good idea, why do we have CRJs, PC12s, and 777s if the 737 fleet can do it all?
There may be different vehicles in the future, like a 12 or 15 m starship, or a smaller 7 m "small craft" size.

But until then, variants on the 9 m ship (cargo, manned, tanker) is what's going to happen, and it will turn out that one basic design will do everything.

STS failed in so many ways, it's hard to say that the main lesson was not to do too many things.  I think the main lesson was not to have the design dictated by too many external influencers.
ABCD - Always Be Counting Down

Offline KilroySmith

  • Full Member
  • ****
  • Posts: 410
  • Phoenix, AZ, USA
  • Liked: 637
  • Likes Given: 414
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1415 on: 07/05/2023 04:21 am »
STS failed in so many ways, it's hard to say that the main lesson was not to do too many things.  I think the main lesson was not to have the design dictated by too many external influencers.

I think the other thing STS failed at was flexibility - there was no reasonable way to change anything about the stack other than the payload inside the orbiter.  In the SH/SS architecture, SH is just the worlds largest dumb(?) booster, like Saturn V, Atlas, Vulcan.  It doesn't much care what the thing you bolt on top of it looks like.  Wanna want to turn it into a three stage rocket?  Wanna turn it into an unmanned tanker?  A bus carrying 200 passengers?  SH is fine with that;  the thrust of 33 Raptors will be there and ready for you.  You just(?!?) need a new SS design to do what you want. 

Online meekGee

  • Senior Member
  • *****
  • Posts: 15698
  • N. California
  • Liked: 15851
  • Likes Given: 1443
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1416 on: 07/05/2023 04:27 am »
STS failed in so many ways, it's hard to say that the main lesson was not to do too many things.  I think the main lesson was not to have the design dictated by too many external influencers.

I think the other thing STS failed at was flexibility - there was no reasonable way to change anything about the stack other than the payload inside the orbiter.  In the SH/SS architecture, SH is just the worlds largest dumb(?) booster, like Saturn V, Atlas, Vulcan.  It doesn't much care what the thing you bolt on top of it looks like.  Wanna want to turn it into a three stage rocket?  Wanna turn it into an unmanned tanker?  A bus carrying 200 passengers?  SH is fine with that;  the thrust of 33 Raptors will be there and ready for you.  You just(?!?) need a new SS design to do what you want.
Sure, but like the guy said, let's abort this side conversation ..
ABCD - Always Be Counting Down

Offline CorvusCorax

  • Full Member
  • ****
  • Posts: 1944
  • Germany
  • Liked: 4225
  • Likes Given: 2869
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1417 on: 07/05/2023 06:26 am »

...

As I said above, we have good PRA models for launches, and they map relatively cleanly onto Starship.  By mission #100, I'd expect that our handle on ascent reliability would be quite good.

...

Great post, thanks for the launch failure statistics 101 :-)

Yeah. Starship has a good chance of finding "abort solutions" for well understood failure cases like pad abort or in flight abort in case of booster issues - especially now with the introduction.of hot staging and the planned 9 engine ship variants.

The real "unknowns" in literally every sense of the words are the.potential failure cases of starship, especially on EDL.

We have only a very rough idea in which way things could go wrong ( mainly thanks to SN 4-11) and no idea about their probability, nor about unknown unknowns aka"what else could go wrong"

Without knowing the risks, the answer "which risks need to be mitigated by abort systems" cannot really be answered, so the question of "how" becomes kinda mute.

Obviously, for any known flaw, the best possible mitigation is simply to fix the design so the flaw is gone. Theres remainder-risks such as "generic engine failure" which can be addressed by redundancy ( more engines, better isolation ) and/or backup systems, but if they are necessary requires quantitative  knowledge of these risks.

So effectively , we wont know wether starship will need an abort system, nor for what modes, until we habe seen 50-100 landings and probably a few more exemplary failures.


« Last Edit: 07/05/2023 06:34 am by CorvusCorax »

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1418 on: 07/05/2023 05:15 pm »
Confidence, reliability, and number of consecutive successes are related by the formula:

log(confidence)/log(reliability)=successes

Plug in 50% and 95%, and the number of successes is 13.51. Since you can't have a fractional success, they would take the next integer. That's where NASA got 14.

When you exponentiate, that becomes:

confidenceLevel = reliabilitysuccesses

...which, if you replace "confidenceLevel" with "probability", is the formula for the probability that <successes> consecutive missions will succeed.

That's... kinda weird.  I'm not saying that it isn't a reasonable (albeit non-intuitive--I hate statistics so much) approximation, but it's not exactly what the regulations call for.

Offline TheRadicalModerate

  • Senior Member
  • *****
  • Posts: 5153
  • Tampa, FL
  • Liked: 3771
  • Likes Given: 705
Re: Abort options for Starship and Starship/SuperHeavy
« Reply #1419 on: 07/05/2023 10:19 pm »
Yeah. Starship has a good chance of finding "abort solutions" for well understood failure cases like pad abort or in flight abort in case of booster issues - especially now with the introduction.of hot staging and the planned 9 engine ship variants.

Here's an algorithm for figuring out whether hot-staging will be good enough:

1) Figure out the minimum safe distance from a 1σ (~84th percentile) explosion.
2) Figure out the 1σ "uh-oh-to-boom" time.
3) Figure out the Raptor startup time.
4) Compute acceleration, using aEscape = 2*minSafeDist/(uhOhToBoom - raptorStartupTime)²
5) Figure out the lightest amount of prop able to get a crew to LEO, to do an RPOD or an emergency EDL.  Use that to compute TW = totalRaptorThrust/(dry + payload + prop).
6) If aEscape < TW-1, you have a winner.  Otherwise, not.  I don't think you can crew-certify without pad escape of some kind.

Quote
Without knowing the risks, the answer "which risks need to be mitigated by abort systems" cannot really be answered, so the question of "how" becomes kinda moot.

Eh... you can do somewhat better than that.  You can identify points in the conops which are especially vulnerable to unknown failures, and then talk to your customers about how comfortable they are with uncertainty in the failure tree at those points.

The thing about escape is that, even though it has a pEscapeSuccess that's substantially lower than pNominalSuccess, it's largely independent of the nominal path, so pLOC = 1 - pEscapeSuccess*pNominalSuccess.  pEscapeSuccess may not be able to be quantified perfectly, but there's a pretty wide range of acceptable probabilities.

From there, you can ask the question, "How much would I discount pNominalSuccess, given the amount of system uncertainty I guess is likely?"  A guess is indeed as good as you get, but you now have the tools to figure out how good your discounted guess needs to be without the escape system.  If your guess is kinda big, it's probably time to start working on escape.

I think there are critical nodes at the following spots in the conops:

1) On the pad through about 500m after launch
2) At max-q
3) During flip-and-land process.

There's also one during hypersonic, but there's nothing to be done.  Only continuous Starship risk reduction does anything there.  Of those, being pessimistic about the uncertainty of flip-and-land seems rational.

Quote
So effectively , we wont know wether starship will need an abort system, nor for what modes, until we have seen 50-100 landings and probably a few more exemplary failures.

I don't think they need to make a decision right now, but if they wait to make the decision to start on escape 100 EDLs from now, it's gonna be painful.  Ultimately, it's a development decision that needs to be made with too little information.

Tags: LAS black zones 
 

Advertisement NovaTech
Advertisement
Advertisement Margaritaville Beach Resort South Padre Island
Advertisement Brady Kenniston
Advertisement NextSpaceflight
Advertisement Nathan Barker Photography
0