Abort options for Starship and Starship/SuperHeavy

[TheRadicalModerate]

People are getting hung up because they have a fixed schedule in mind.

The rational way to do this is to launch and recover SS enough that it becomes reliable.  How many flights that takes and how long does not matter as much as that it gets done.  You have to do this if you intend to use SS for Starlink deployment, so you get a large part of your reliability "for free".

If launch and EDL actually become reliable enough to forgo an escape system, then this might work.  However, while I suspect that launch will probably become reliable enough to get crew-rated, it remains to be seen if EDL will ever become reliable enough.  At this point, we don't even know if EDL is possible, to say nothing of reliable.  You could wait a long time.

And I do think that there are time limits on how long SpaceX can wait.  The most obvious one is that they can't end-of-life F9/D2 until Starship can take its place.

The Maezawas and Isaacmans and Titos of the world are willing to be patient--up to a point.  But they're all going to want to go to cislunar space, if not the lunar surface, and they're going to want to do it soon enough that it means something special.  But the per-seat cost of a lunar mission is going to be very high, because you have to factor in the tanker launch costs.  Those can be spread out over more seats only if Starship can handle 10-20 passengers at a time.  But doing so requires launch and EDL crew-certification.

If SLS/Orion finally dies (IMO this is likely), then NASA would be happy to get Artemis crews to the surface by ferrying four of them at a time up with an F9/D2.  But that may cost $250M/seat.¹  Since that's about 20% of per-seat cost for an SLS/Orion/LSS mission, it's a terrific deal for NASA.  But it's a pretty hefty cost even for a motivated billionaire.

For comparison, using the same kind of SWAGs for a direct launch/EDL Starship mission with 12 crew, still using the LSS to do LEO-LS-NRHO-refuel-LEO², you get $110M/seat.

There's an argument to be made that SpaceX doesn't need those private missions.  But I think they need them to put pressure on NASA to step up their lunar surface cadence, which in turn gives SpaceX the launch and BEO experience to get ready for Mars.  My guess is that they'll need something crew-certified by 2030.  That could be anywhere from 400 launches to a couple thousand, depending on how things go.  But the couple thousand launches is much more likely if they have crew-certification early.  That may make the expense of engineering a launch escape system worthwhile.

______________
¹$240M for each F9/D2 launch.
15 tanker launches @ $40M/launch.
1 LSS, and let's say it costs that same as an F9/D2: $240M
Total Price:  $1320M.  For 4 crew:  $330M/seat.

²Refueling in NRHO is actually cheaper than refueling in LLO.  Figure that the LEO ferry Starship and the LSS both cost $240M/mission.

[TheRadicalModerate]

Getting a fairing that can carry reentry loads and the canards for a nominal landing but still blow away to enable capsule propulsion is also a challenge.

Yeah, I'd like to see how separating header tank piping is going to work.

Don't put a header tank in the nose.  Put ballast in the nose, and put the header tanks midships.  (They're probably going to have to do the midships headers for LSS anyway.)

Restricting crewed launch/EDL only to LEO gives you gobs of mass margin, which in turn allows you to do very inefficient things if it improves safety.

[Negan]

The landing is still risky.   There are still too many single points of failure

Maybe it would be possible to preemptively deploy a small, short-ranged aircraft to offload the people before landing. Wouldn't be instant but could be used well before meeting up with the chopsticks.

[Anguy]

______________
¹$240M for each F9/D2 launch.
15 tanker launches @ $40M/launch.
1 LSS, and let's say it costs that same as an F9/D2: $240M
Total Price:  $1320M.  For 4 crew:  $330M/seat.

I don't think D2 mission actually costs that much. SpaceX is now likely just gouging NASA a little, because they have no competition. But if D2 launches are actually necessary for SS to make money I expect the price to come down drastically.Also the tanker launch should be 10m max, or SX Mars plans will never work...

[DanClemmensen]

______________
¹$240M for each F9/D2 launch.
15 tanker launches @ $40M/launch.
1 LSS, and let's say it costs that same as an F9/D2: $240M
Total Price:  $1320M.  For 4 crew:  $330M/seat.

I don't think D2 mission actually costs that much. SpaceX is now likely just gouging NASA a little, because they have no competition. But if D2 launches are actually necessary for SS to make money I expect the price to come down drastically.Also the tanker launch should be 10m max, or SX Mars plans will never work...

Once Starship is operational, D2 mission cost will go up a lot, because SpaceX will move as many payloads as possible to Starhip, where their profits will be higher. This means the D2 missions must pay for an increased percentage of the fixed operating and infrastructure costs of F9.

[Robotbeat]

People are getting hung up because they have a fixed schedule in mind.

The rational way to do this is to launch and recover SS enough that it becomes reliable.  How many flights that takes and how long does not matter as much as that it gets done.  You have to do this if you intend to use SS for Starlink deployment, so you get a large part of your reliability "for free".

What if it takes a couple thousand successful landings without a failure (as it does for airliners), and you *never* get a run that long without a failure during LEO launch operations?

”What would you do if the Earth gets hit by a comet the size of Texas?”
“Well, I guess I’d just die?”

Sometimes, you just fail. The vision Starship was created for requires the launch vehicle to be that reliable in order to meet its goals of costs anyway. They might never get there, but Falcon 9 is approximately within an order of magnitude of that already (in terms of consecutive successful launches and consecutive successful landings), so it is a reasonable goal.

I think the longest string of successful landing is in the high 70s. That's 25 times fewer than airliner certification, and over 10,000 times fewer than airliners in service.

Yes, it's 79. But so what? 80 is a lower limit of what can be achieved on a vehicle where the booster is still even intentionally expended sometimes (i.e. to save the mission if there's an engine out as in the last 2 landing failures, or just to push max performance), it's not an upper limit. 2000 flights is only an order of magnitude away. Simply allowing enough margin for successful landing with engine-out, along with improved robustness features (including some type of abort, perhaps) and extra QA on crewed flights, and thousands of cargo flights would go a long way to making it roughly as safe as, say, general aviation per mile. Achieving 2022 passenger aviation safety levels off the bat isn't feasible for literally anything, and certainly wasn't when air travel first was invented, either.

Asking what would they do if they fail to achieve that goal… well, then they failed to achieve the goal. Not a hard question to answer.

Which means it's never capable of launching the public, only those willing to accept a far higher chance of dying that the general public accepts.

No, it proves nothing. You were literally begging the question by asking what would happen if it can't ever do 2000 in a row LOL.

That's conceptually fine, but killing people once in a while also puts a big stain on the whole HSF industry...

Where did I even claim that? Doing even 400 flights in a row would make it safer than anything NASA has ever built. Yes, SpaceX has a massive job to do to prove Starship can launch and land reliably and they might never be able to do that. But you're just concern-trolling.

[Barley]

And I do think that there are time limits on how long SpaceX can wait.  The most obvious one is that they can't end-of-life F9/D2 until Starship can take its place.

Of course they can.  NASA got rid of Apollo before they had the Shuttle.  NASA got rid of the Shuttle before they had Dragon.   It's a bit much to hold SpaceX to a different standard.

But there really is no need to end of life F9/D2 at all.  As long as NASA keeps buying SpaceX can keep building.  If NASA is the sole remaining F9 customer, you raise the price to cover costs.

If launch and EDL actually become reliable enough to forgo an escape system, then this might work.  However, while I suspect that launch will probably become reliable enough to get crew-rated, it remains to be seen if EDL will ever become reliable enough.  At this point, we don't even know if EDL is possible, to say nothing of reliable.  You could wait a long time.

This is true, we don't know if EDL is possible.  If EDL is not possible you don't have a StarShip.  If EDL is possible but unreliable, you'll at least know what you're working with and can try to make it more reliable or design an escape system around a known failure, not a laundry list of theoretical failures.

You need to test, and test, and test.  Before you give up you should be sure it won't work.  Plan A should not include the assumption that plan A will not work.

IMHO any escape system will never be temporary.  Once you pay the Danegeld you will never get rid of the Dane.  Giving in too soon on a kludged escape system will cripple HSF for decades.  Almost any cost of not engineering an escape system is cheap at twice the price.

[Barley]

The landing is still risky.   There are still too many single points of failure

Maybe it would be possible to preemptively deploy a small, short-ranged aircraft to offload the people before landing. Wouldn't be instant but could be used well before meeting up with the chopsticks.

Just like Gagarin.  Of course the FAI probably won't count it as a space flight.  Oh the humanity.

[Lee Jay]

Yes, it's 79. But so what? 80 is a lower limit of what can be achieved on a vehicle where the booster is still even intentionally expended sometimes (i.e. to save the mission if there's an engine out as in the last 2 landing failures, or just to push max performance), it's not an upper limit. 2000 flights is only an order of magnitude away.

It's more than an order of magnitude, to minimum for certification, and it's a far more difficult booster landing approach combined with adding EDL and landing of the upper stage.  So, in addition to being more than an order of magnitude away, it's also a couple of orders of magnitude more complex and difficult.

[InterestedEngineer]

Yes, it's 79. But so what? 80 is a lower limit of what can be achieved on a vehicle where the booster is still even intentionally expended sometimes (i.e. to save the mission if there's an engine out as in the last 2 landing failures, or just to push max performance), it's not an upper limit. 2000 flights is only an order of magnitude away.

It's more than an order of magnitude, to minimum for certification, and it's a far more difficult booster landing approach combined with adding EDL and landing of the upper stage.  So, in addition to being more than an order of magnitude away, it's also a couple of orders of magnitude more complex and difficult.

How many flights does an airplane require for certification?

The initial 777, which was the first full long-haul ETOPS airplane, underwent 3,000 flights for certification.  That's an upper limit on the number, and far too pessimistic.

Fact: When the first 777 takes flight next June, it will begin the most extensive flight-test program ever for any commercial jetliner. Three 777s will each fly 1,000 flights

Mind you this is the number of flights required for passenger airplane in an already extremely safe environment with fatal accident rates per departure of 1 in 5,000,000.

For rockets, from 1961-2020, there were 5 accidents per 327 manned flights, or 1:65 flights.   That's 77,000 times worse than airline flights.

It's ridicules to try and jump from 1:65  to 1:5,000,000 in one jump.    1:650 would be amazingly better than we have today.   That would require about 1000 flights.  equaling today would be about 150 flights.



EDL and landing the upperstage is not 20x-30x more complicated than landing a modern airliner.  Twice as hard, maybe.   The Space Shuttle would never have worked if it was 20x harder.



https://web.archive.org/web/20110811212724/http://community.seattletimes.nwsource.com/archive/?date=19930816&slug=1716209
https://www.mdpi.com/2226-4310/9/11/675
https://www.airlines.org/dataset/safety-record-of-u-s-air-carriers/

[Lee Jay]

Yes, it's 79. But so what? 80 is a lower limit of what can be achieved on a vehicle where the booster is still even intentionally expended sometimes (i.e. to save the mission if there's an engine out as in the last 2 landing failures, or just to push max performance), it's not an upper limit. 2000 flights is only an order of magnitude away.

It's more than an order of magnitude, to minimum for certification, and it's a far more difficult booster landing approach combined with adding EDL and landing of the upper stage.  So, in addition to being more than an order of magnitude away, it's also a couple of orders of magnitude more complex and difficult.

How many flights does an airplane require for certification?

The initial 777, which was the first full long-haul ETOPS airplane, underwent 3,000 flights for certification.  That's an upper limit on the number, and far too pessimistic.

Fact: When the first 777 takes flight next June, it will begin the most extensive flight-test program ever for any commercial jetliner. Three 777s will each fly 1,000 flights

Mind you this is the number of flights required for passenger airplane in an already extremely safe environment with fatal accident rates per departure of 1 in 5,000,000.

For rockets, from 1961-2020, there were 5 accidents per 327 manned flights, or 1:65 flights.   That's 77,000 times worse than airline flights.

It's ridicules to try and jump from 1:65  to 1:5,000,000 in one jump.   

Right, which is why PTP and colonization won't happen with this system.

EDL and landing the upperstage is not 20x-30x more complicated than landing a modern airliner.  Twice as hard, maybe.   The Space Shuttle would never have worked if it was 20x harder.

It may be more than 20-30x harder.  Entering the atmosphere at Mach 25 or higher if from a BEO trajectory is an extreme environment, to say the least.  Heat, plasma, hypersonics, vacuum, transition from thruster or mass control to aerodynamic control, transition from supersonic to subsonic are all complications a landing airliner doesn't have to deal with.

[Robotbeat]

Again, this laughable begging the question.

[Negan]

The landing is still risky.   There are still too many single points of failure

Maybe it would be possible to preemptively deploy a small, short-ranged aircraft to offload the people before landing. Wouldn't be instant but could be used well before meeting up with the chopsticks.

Just like Gagarin.  Of course the FAI probably won't count it as a space flight.  Oh the humanity.

No, more like SpaceShipOne except instead of using a second aircraft to get you up, you use it to get you down.

Edit: A Cirrus Vision SF50 with folding wings might fit.

[chopsticks]

I wish we would stop this argument of comparing rockets to airplanes and the safety requirements for each.

Rocket travel to space and back is always going to be more dangerous than air trave no matter how safe you make it, you just can't get around that. In terms of LOS or LOM we should instead be looking at what might NASA, ESA, commercial astronauts, etc. require, and build off of that. What are the actual probabilities of a failure (of any mode of launch to EDL) and how might we avoid them, instead of making irrelevant equivalencies to airplanes.

Airplanes ≠ Rockets


We can all have our opinions on what might or might not be a suitable abort system or scenario, but these analogies get tiring IMO.

[InterestedEngineer]

In terms of LOS or LOM we should instead be looking at what might NASA, ESA, commercial astronauts, etc. require, and build off of that.

I think it is already well proven that NASA's requirements are what Elon calls "bad requirements"

NASA's requirements are based off of up-front-get-it-perfect-in-simulation with very expensive and low iteration rates.  Just look at SLS for example.  Everything perfectly simulated, program insanely long and expensive, and they still severely damaged the launch mount.

Precisely the wrong thing for long term economics of Space Flight, and everything Elon is fighting for.

Iteration is how you get things right long term, for the least amount of money.

Elon himself says he wants to emulate the logistics of commercial flight.   I can't argue with it, because it's the right thing to do.

Most successful products I've helped design did what Elon does, not "ask the customers for their requirements".

[spacenut]

If launching or landing becomes a problem for NASA or even SpaceX, they can always go back to 6 seats on Dragon and ferry crews using Dragons and dock with Starships.  SpaceX might even go back to land landings of the capsules to save recovery time. 

There are a lot cheaper options than making a giant capsule to sit on top of a Starship.  Ultimate goal is to make Starships land and be completely reusable. 

[Lee Jay]

I wish we would stop this argument of comparing rockets to airplanes and the safety requirements for each.

The reason to do that is because of the application. If the rocket's application is to carry the general public from point to point on the Earth, then it's entirely reasonable to demand from it that which we demand from the other means we use to do the same thing.

Same thing if it's carrying the general public to go colonize Mars.

If it's professional astronauts, paid to learn the systems and missions and to take the risks, that's different and the bar doesn't have to be as high.

Rocket travel to space and back is always going to be more dangerous than air trave no matter how safe you make it, you just can't get around that.

Probably true, which is one of many reasons I think doing the two things I mentioned above simply won't happen with this system.

In terms of LOS or LOM we should instead be looking at what might NASA, ESA, commercial astronauts, etc. require, and build off of that.

Again, astronauts are *paid* to take these risks.  The general public, as payING customers, are not.

[sebk]

I wish we would stop this argument of comparing rockets to airplanes and the safety requirements for each.

The reason to do that is because of the application. If the rocket's application is to carry the general public from point to point on the Earth, then it's entirely reasonable to demand from it that which we demand from the other means we use to do the same thing.

Possibly.

Same thing if it's carrying the general public to go colonize Mars.

Absolutely not. If the chances of death were similar to commuting by ones car for a year that would be absolutely acceptable.

I'd like to note that those odds are about 1:8000. That's within one order of magnitude of modern (i.e. Dragon) rocket systems.

If it's professional astronauts, paid to learn the systems and missions and to take the risks, that's different and the bar doesn't have to be as high.

That's a wrong argument. People are paying in the order of $100k to  maybe climb Mount Everst which has chances of dying worse than taking part in a spaceflight like Inspiration 4.

Or even stay with aviation. People pay quite a lot for stuff like air taxi or helicopter rides. Both are incomparably more dangerous than regular transport planes. It's actually more dangerous than driving a car.

Rocket travel to space and back is always going to be more dangerous than air trave no matter how safe you make it, you just can't get around that.

That's unlikely to be true in the long term. Airplanes have to travel through a capricious dynamic environment hard to precisely predict. Vacuum is harsh but much more regular. In engineering harsh but regular is easier to design for than less harsh but irregular.

In terms of LOS or LOM we should instead be looking at what might NASA, ESA, commercial astronauts, etc. require, and build off of that.

Here I agree. You could also add there there would be various categories of commercial astronauts.

[TheRadicalModerate]

The reason to do that is because of the application. If the rocket's application is to carry the general public from point to point on the Earth, then it's entirely reasonable to demand from it that which we demand from the other means we use to do the same thing.

Same thing if it's carrying the general public to go colonize Mars.

If it's professional astronauts, paid to learn the systems and missions and to take the risks, that's different and the bar doesn't have to be as high.

There's a middle ground here, where you're not a common carrier (what the airlines are, and likely what SpaceX would be with a P2P business, or even one that carries thousands of people to the Moon or Mars), but you're also carrying more than just professional astronauts.  Those people are permitted to fly under a regime of informed consent on experimental spacecraft.

The Space Act was very good about giving the commercial industry enough leeway to handle these experimental passengers, and it continues to do so, having been extended multiple times.  But this won't last forever.  Either the FAA will go full common-carrier on Starship and other crewed commercial systems, or it will create a new regulatory class.  I suspect that the industry has until about 2030 before somebody puts their foot down.

Meanwhile, NASA has regulations for its professional astronauts:

pLOC <1/500 during ascent
pLOC <1/500 during EDL
pLOC <1/270 during a six-month mission, after factoring in MMOD and other on-orbit hazards.

1/500 for each of launch and EDL is almost certainly what NASA would expect, and I'd be surprised if SpaceX didn't hold itself to the same standard for private missions.  There's too much downside and not enough upside to do anything else.

pLOC for an early Mars mission is clearly going to be much, much larger.  But you can't do a Mars mission unless you can get crews to and from LEO.  My recommendation would be to ignore Mars pLOCs for the time being and concentrate on the LEO problem, which is a necessary precondition for doing Mars (or even high-scale Moon missions).

Now:  what method will NASA require to demonstrate the reliability?  Seems to me that there are three options:

1) Probabilistic risk assessment, which will be what NASA wants eventually.  They can probably do this with launch after a few dozen launches, because Starship isn't a particularly weird platform on launch.  But I don't even know how you'd go about constructing a credible failure network for EDL, to say nothing of assigning failure probabilities to enough nodes to make PRA useful.

2) If PRA won't work, they'll have to do empirical measurements of reliability.  The traditional way to do this is to simply measure the successes and failures, develop a standard error for the instantaneous reliability success/(success+failure), and display a range that represents the confidence interval for the reliability.  The problem with this is that it weights past failures too heavily, especially considering that engineering improvements should largely fix whatever caused the early failures.

3) This is the point where my statistical knowledge fails me, but there is presumably some sort of quasi-bayesian way of improving our knowledge of the reliability over time, such that whatever we use as a not-very-well-informed prior (say, the reliability of the first 50 missions) gets washed away as we get new data.

Eventually, we'll have so much data about the various kinds of failures that a PRA can be constructed to make NASA comfy, and SpaceX can be confident that they understand the risk to which they're subjecting all of their crews, not just the NASA ones.  But until that time, we need to know how many missions we have to run to get a decent empirical measurement.

That number of missions will obviously be affected by how many failures there are at various stages of the mission sample.  If failures are seen to have clustered early in the program and are virtually nonexistent after the bugs have been worked out, then less than a couple hundred missions may easily be all that's needed.  But if there are persistent failures after the early debugging--even if they're fairly infrequent--then it may indeed take thousands of missions, not so much to get to a reliable system, but more to get to the point where the system can be demonstrated as being adequately reliable.

[Robotbeat]

I wish we would stop this argument of comparing rockets to airplanes and the safety requirements for each.

Rocket travel to space and back is always going to be more dangerous than air trave no matter how safe you make it, you just can't get around that. In terms of LOS or LOM we should instead be looking at what might NASA, ESA, commercial astronauts, etc. require, and build off of that. What are the actual probabilities of a failure (of any mode of launch to EDL) and how might we avoid them, instead of making irrelevant equivalencies to airplanes.

Airplanes ≠ Rockets


We can all have our opinions on what might or might not be a suitable abort system or scenario, but these analogies get tiring IMO.

It highly depends on the method of air travel. I agree that passenger air travel is a terrible comparison as it’s effectively infinitely safe. It’s not necessary to get to that safety level. General aviation includes private business jets, helicopters, air taxis, and other utilitarian uses as well as instruction and personal use.
General aviation is about 1 fatal accident per 100,000 flight hours, which lets say is about 1 fatal accident per 20,000,000 miles.

So if you circumnavigate the globe with general aviation flight reliability statistics, your odds of fatality are about 1 in 8000, the same as the odds of dying in a car crash per year.

I fail to see why Starship couldn’t reach that reliability level in principle. I’m really surprised Falcon 9 has achieved the reliability that it has, and that’s the first VTVL-based reusable rocket. Future iterations like Starship could do much better (if they can reach a higher flight rate, like 1000 per year).