Abort options for Starship and Starship/SuperHeavy

[Lee Jay]

A mechanism sitting on the top of its normally used staging system is reasonably simple (plus or minus hypergolic explosions).   They still fail.  And they fail when not needed as well as when needed.

When was the last time that happened operationally?

Probably https://spaceflightnow.com/2022/02/10/research-satellites-destroyed-in-astra-rocket-failure/

In the very early days SpaceX commissioned Futron to study causes of rocket failures (to help design Falcon). In their sample 28% of all launch failures were due to separation systems.

I'm pretty sure I was talking about abort systems.

[TheRadicalModerate]

A mechanism sitting on the top of its normally used staging system is reasonably simple (plus or minus hypergolic explosions).   They still fail.  And they fail when not needed as well as when needed.  The former increases probability of LOC.

The probability of an uncommanded escape failure is going to be two or three orders of magnitude less than the probabilities of any of the events for which it enables escape.  If that's what your argument rests on, it's not a very good argument.

And, as Lee Jay said, even if the escape system is only 90% successful, it reduces the pLOC for any given event by another order of magnitude.  That requires some testing, but not very much--even if it's a completely new system.  And if you can re-kludge a D2 to be 80% of the solution, that should require even less testing.

One that has to eject out of a Starship, not so simple.

I've acknowledged the difficulty.  The problem is that you haven't made a compelling case for why it's unnecessary, other than to wait for a bunch of missions to have flown and everything'll be fine.

Do me a favor:  Go through the following flight phases and tell me how you make each one of them safe enough for humans (it's the same list as up-thread) without some kind of escape system.

1) On-pad.
2) Low altitude, before Starship can get any staging separation.
3) Mid altitude, after it can get staging separation.
4) Max-q.
5) Hypersonic ascent to orbit.
6) Orbital emergency requiring immediate de-orbit.
7) Hypersonic entry.
8) The belly-flop descent phase.
9) The flip-and-burn phase.
10) Off-target landing.
11) Post-landing.

A couple of these probably have no solution and you have to live with them.  But you also have to get to 1:1000 for pad to orbit, 1:1000 for orbit to landing, and 1:270 for a six(ish)-month mission.

[InterestedEngineer]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

[InterestedEngineer]

The problem is that you haven't made a compelling case for why it's unnecessary, other than to wait for a bunch of missions to have flown and everything'll be fine.

You clearly don't believe in iteration, which defies all modern software practices and what SpaceX, Tesla, and many other modern hardware companies have been intentionally¹ doing for over a dozen years.

The word is "iterate", not "wait".  Which means, try, fail, improve, try again.  Repeatedly.   Mathematically it's equivalent to annealing. Simulated annealing is one of the ways we converge high variable count systems like neural nets.

You also have the burden of proof backwards.  "Best part is no part" means the burden of proof is on the side of need the part.  The burden of proof is not on "remove the part" side.

¹  disclaimer:  I hate the word "intentional" but I'm using it anyways.

[InterestedEngineer]

<list of things to solve>

1) On-pad -

iterate 200 launches. 

The cryo is what makes this more dangerous than fueling an airplane (but even that is not completely safe - https://archive.ph/wip/9MfPA)

2) Low altitude, before Starship can get any staging separation.

multi-engine out capability.

Has Falcon-9 ever failed here?  200 iterations.  Airplanes don't have low altitude aborts either.  A common cause of crashes.   Don't use motors that can't be shut down (like SRBs)

https://skybrary.aero/sites/default/files/bookshelf/493.pdf

Airplanes have a V1 decision speed.  But they also have wheels, wider weather ranges, lots of moving wing surfaces, and can't launch on engine out very easily. The equivalent for a rocket is hold down clamp release decision.

3) Mid altitude, after it can get staging separation

Has a falcon 9 ever failed here?  multi-engine out capability.

Separate Starship and land on water after expending fuel.

Separation part involves iteration involves destroying a Starship on ascent.  Not going to happen very much.  but capsules have the identical problem.   Dragon-2 was tested exactly once in this mode.

4) Max-q.

has a falcon 9 or STS ever failed here?  multi-engine out capability.

Separate Starship and land on water after expending fuel.

5) Hypersonic ascent to orbit.

has a falcon 9 or STS every failed here?  multi-engine out capability.

Separate Starship and land on water after expending fuel.  May be able to do a suborbital flight to a landing tower if feasible.

6) Orbital emergency requiring immediate de-orbit.

Land on water nearest a port with rescue ships if can't be scheduled to a normal landing tower

7) Hypersonic entry.

Triple redundancy.  Tiles fail?   Saffil blanket.  Saffil blanket fail too?  Stainless steel.   Shuttle had no redundancy, yet only had 1 LOC in 135 missions.   Triple redundancy should get to much better than 1:1000.

Iteration also helps.  monitor Hot spots using internal infrared cameras.  See what failure of tile actually results in.  Fix and iterate.

The belly-flop descent phase.

The aero control surfaces of Starship is by far simpler than an airplane, and airplane failure rates are still low but do happen from failure of aero surfaces.  PRA+200 iterations should be able to easily prove pLOC < 1:1000

9) The flip-and-burn phase.

2 engine out redundancy.  An airplane stalling at low altitude also has no recovery.   Iterate 200 times and measure pressure of tanks to make sure variation is outside 5-6 sigma of min engine startup pressure (as an example of one thing they should measure and get real data on, instead of PRA guesses).

10) Off-target landing.

Always land near water.  Water landing abort, same as Falcon 9.

11) Post-landing.

when has this happened to STS of falcon 9?  I note the paranoia they have with Dragon-2 landing and its hypergolics.   Best part no part applies here, not hypergolics.

A couple of these probably have no solution and you have to live with them.  But you also have to get to 1:1000 for pad to orbit, 1:1000 for orbit to landing, and 1:270 for a six(ish)-month mission.

You can see why I'm a huge fan of making water landing aborts work.   It eliminates a massive number of parts, probably only at the expense of beefing up the cargo section and maybe a small bit of the upper tank and dome that is traveling at > 10m/sec during rotation.  It solves the landing phase of various aborts scenarios

Transform the SpaceX navy into a rescue navy (to answer that obvious concern)

I'm amused to note that you missed staging failures, which was the reason for almost all Soyuz aborts (except one pad fire):

https://en.wikipedia.org/wiki/Soyuz_abort_modes

Perhaps there's something complicated about Soyuz stage separation.  But I'll note again that 200 iterations with various metrics on how 'clean' separations are will likely get pLOC to < 1:1000.

[Lee Jay]

1) On-pad -


iterate 200 launches.

And if some of those 200 launches fail?

Airplanes don't have low altitude aborts either.   
Airliners do.

4) Max-q.


has a falcon 9 or STS ever failed here?
Yes.

11) Post-landing.
has this happened to STS of falcon 9?   


Yes.

[TheRadicalModerate]

The problem is that you haven't made a compelling case for why it's unnecessary, other than to wait for a bunch of missions to have flown and everything'll be fine.

You clearly don't believe in iteration, which defies all modern software practices and what SpaceX, Tesla, and many other modern hardware companies have been intentionally¹ doing for over a dozen years.

The word is "iterate", not "wait".  Which means, try, fail, improve, try again.  Repeatedly.   Mathematically it's equivalent to annealing. Simulated annealing is one of the ways we converge high variable count systems like neural nets.

You also have the burden of proof backwards.  "Best part is no part" means the burden of proof is on the side of need the part.  The burden of proof is not on "remove the part" side.

¹  disclaimer:  I hate the word "intentional" but I'm using it anyways.

I believe wholeheartedly in iteration, but it's not a panacea.  You have to have your architecture sorta-kinda right to begin with, or all iteration does is pile kludge on top of kludge.

F9 didn't start out as a reusable vehicle, but it made a lot of architectural choices at the outset (restartable engines, lots of moderately-sized engines that gave it a wide range of T/W, common first and second stage engines, and a 2nd stage with a lot more delta-v than was common at the time) that made it possible to iterate to get to reusability.  Making different choices from these might easily have left F9 as an expendable vehicle forever, iteration or no iteration.

It is entirely possible that the architecture of Starship is close enough to correct that engineering in crew safety after the fact won't be that hard.  But they've made some very odd choices if that was the goal.  To be fair, those choices are considerably less odd for landing on Mars and returning to Earth, but they all seem to point to a vehicle where total pLOC is going to be well over 1%--which would be great for an early Mars trip, but fairly terrifying for any LEO application.

We'll see what happens.  But iteration itself dictates that if things are looking iffy 50-100 flights in, it's time to build an escape system.

And I'm so sick of "the best part is no part" chant that I could scream.  Some requirements aren't stupid and, after carefully considering whether they're stupid and concluding that they're not, you account for them in your architecture.  Sometimes, that's not conducive to getting your minimum viable product out, and you pay for it later.  I think that's what's going to happen with crew-rating.

[Lee Jay]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

[RamsesBic]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

[chopsticks]

multi-engine out capability.

You better hope that they never have any tankage issues, like ullage collapse which takes out all engines..

See SN10

(or any other event that could take out all or most of the engines at once)

[chopsticks]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

Maybe because we have come to expect a higher level of safety than we used to?

[Lee Jay]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

The statistics of flying were never as bad as the statistics of rocketry are right now, plus we probably place a higher value on human life now than we did 100 years ago when lots of people were dying of (by today's standards) minor diseases, industrial accidents and pollution.

[eriblo]

The premise of this thread is that while all signs point to SpaceX not planing an abort system and inclusion of one might well be incompatible with their vision for Starship that is boring and we will therefore discuss the "how" of Starship abort systems without bothering with "if" or "why" (slight paraphrasing by me ).

That means that the relative safety of Starship is technically off topic as is the question of whether a Starship with an abort system can ever be made as safe as one without. Discussion of risks and possible mitigations during different stages of flight on the other hand should be okay as they pertain to which abort modes the abort system will be designed for. Obviously, care has to be taken to not mitigate too many of them such that one accidentally eliminates the abort system completely.

With this in mind my opinion is that a likely escape system incorporates possibly modified versions of commercially available systems such as those attached together with additional Raptors, redundant actuators, structural reinforcements and possibly redundant header tanks

[RamsesBic]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

The statistics of flying were never as bad as the statistics of rocketry are right now, plus we probably place a higher value on human life now than we did 100 years ago when lots of people were dying of (by today's standards) minor diseases, industrial accidents and pollution.

What are the death statistics of rocketry? I remember flying in the 50's - it was a clear and known risk every time. Have we become so risk averse that we are becoming paralysed? Is zero accidents the only acceptable outcome? Why not leave it to the ones who wish to sit in the rockets to decide? Do we apply that standard on other areas of life, like traffic?

I am not saying that we should stop trying to make it safe, but somewhere there has to be a balance between total safety and reasonable safety.

[InterestedEngineer]

With this in mind my opinion is that a likely escape system incorporates possibly modified versions of commercially available systems such as those attached together with additional Raptors, redundant actuators, structural reinforcements and possibly redundant header tanks

 

Ha, that made me laugh very much.  Hats off sir

[InterestedEngineer]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

The statistics of flying were never as bad as the statistics of rocketry are right now, plus we probably place a higher value on human life now than we did 100 years ago when lots of people were dying of (by today's standards) minor diseases, industrial accidents and pollution.

I'm curious, when you make such assertions about flying, do you do any research at all, even a 30 second Google search, that finds things like this?

https://www.kaggle.com/code/garydee/who-not-to-fly-with

(scroll way down to get deaths per year).  I note this isn't per-departure or flight hour.

[Robotbeat]

multi-engine out capability.

You better hope that they never have any tankage issues, like ullage collapse which takes out all engines..

See SN10

(or any other event that could take out all or most of the engines at once)

Yes, this is something they’ll have to devote more margin to, especially for crewed vehicles.

Of course, this belongs in the same category as “better hope your LAS doesn’t blow up in a way that would kill your crew,” which has happened before, ie with the Dragon LAS.

SRMs can and have failed in catastrophic ways. An SRM is needed even for the jettison motor on the Orion LAS, and the answer there is just to build it with enough margin and testing that its probability of failure is extremely remote.

[Lee Jay]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

The statistics of flying were never as bad as the statistics of rocketry are right now, plus we probably place a higher value on human life now than we did 100 years ago when lots of people were dying of (by today's standards) minor diseases, industrial accidents and pollution.

I'm curious, when you make such assertions about flying, do you do any research at all, even a 30 second Google search, that finds things like this?

https://www.kaggle.com/code/garydee/who-not-to-fly-with

(scroll way down to get deaths per year).  I note this isn't per-departure or flight hour.

Rocket failures are in the 1% to 5% range.  Where on that page does it give a failure *rate* (fatal crashes per flight) for airplanes?

Today, it's about 1 in 10,000,000 for commercial airliners.  I think it's safe to say it wasn't 1 in 100 in 1920.

[Robotbeat]

Yeah - SS is a bad architecture for humans.

So were airplanes circa 1920

And they had to fundamentally change to become safe for the public to fly. Are you accepting that SS is the rocket equivalent of a 1920s airplane, and thus needs to fundamentally change to become safe for people?

These statistics did not stop people from flying. Why are we so scared now?
https://en.wikipedia.org/wiki/Aviation_accidents_and_incidents#/media/File:ACRO_incidents.svg

The statistics of flying were never as bad as the statistics of rocketry are right now, plus we probably place a higher value on human life now than we did 100 years ago when lots of people were dying of (by today's standards) minor diseases, industrial accidents and pollution.

I'm curious, when you make such assertions about flying, do you do any research at all, even a 30 second Google search, that finds things like this?

https://www.kaggle.com/code/garydee/who-not-to-fly-with

(scroll way down to get deaths per year).  I note this isn't per-departure or flight hour.

Rocket failures are in the 1% to 5% range.  Where on that page does it give a failure *rate* (fatal crashes per flight) for airplanes?

Today, it's about 1 in 10,000,000 for commercial airliners.  I think it's safe to say it wasn't 1 in 100 in 1920.

Falcon 9 is less than 1% right now for full failure.

I think long range flights in 1920 may have had a 1 in 100 failure rate, FWIW.

[edzieba]

Today, it's about 1 in 10,000,000 for commercial airliners.  I think it's safe to say it wasn't 1 in 100 in 1920.

The main difficulty in comparison is lack of data on total passenger numbers and total flight numbers for early aviation (total accident and fatality numbers are easier to source). 1/100 is not an unreasonable ballpark for early commercial aviation, given early commercial aviation was putting one or two people in the open back of early experimental aircraft.