[dpifke]

Preliminary indication is that we had an oxygen/fuel leak in the cavity above the ship engine firewall that was large enough to build pressure in excess of the vent capacity.

Apart from obviously double-checking for leaks, we will add fire suppression to that volume and probably increase vent area. Nothing so far suggests pushing next launch past next month.

https://x.com/elonmusk/status/1880060983734858130

[perihelions]

Reminds me of one of NASA's reckless ideas, abandoned after Challenger in 1986, to put a liquid hydrogen stage inside the cargo bay of the Shuttle orbiter [0]. That would have likely leaked inside that confined volume, and could plausibly have exploded in a similar way as Starship.

[0] https://en.wikipedia.org/wiki/Shuttle-Centaur

- "The astronauts considered the Shuttle-Centaur missions to be riskiest Space Shuttle missions yet,[85] referring to Centaur as the "Death Star".[86]"

[raverbashing]

I'm not sure there's fire suppression effective enough for this type of leak (especially given rocket constraints)

[echelon]

Replying to this comment so people can see the incredible video of the breakup taken from a diverting aircraft:

https://www.reddit.com/r/aviation/comments/1i34dki/starship_...

[psunavy03]

Aerospace fire suppression is generally Halon, which would purge the cavity with inert gas.

[m4rtink]

Actually the Super Heavy (first stage) already uses heavy CO2 based fire suppression. Hopefully not that necessary in the long term, but should make it possible to get on with the testing in the short term.

[raverbashing]

That's interesting

However if you see the stream you can see one of the tanks rapidly emptied before loss of signal

It seems this was not survivable regardless of fire or not

[Alive-in-2025]

What is a long term solution for this? Is there something more than "build tanks that don't leak"? I'm sure spaceX has top design and materials experts, now what ;-).

[m4rtink]

I think its likely not the tanks but rather the plumbing to engines and the engines themselves leaking (sense lines, etc).

Next engine revision (Raptor 3) should help, as it is much simplified and quite less likely to leak or get damaged during flight.

[metalman]

just increased venting to keep any vapor concentrations of fuel and oxidiser below that capable of igniting, even simple baffling could suffice as the leaks may be trasitory and flowing out of blowoff valves, so possibly a known risk. Space x is also forgoeing much of the full system vibriatory tests, done on traditiinal 1 shot launches, and failure in presurised systems due to unknown resonance is common. Big question is did it just blow up, or did the automated abort, take it out, likely the latter or there would be a hold on the next launch.

[vessenes]

There’s no way that was anything but the automated abort — it was a comprehensive instantaneous rapid event. Or I guess I’d say, however it started, the automated abort kicked in and worked.

[spandrew]

It might not even be about fire suppression. Oxygen and different gases can pool oddly in different types of gravity. If oxygen was leaking, it may be as simple as making sure a vacuum de-gases a chamber before going full throttle.

We know nothing, but the test having good data on what went wrong is a great starting point.

[varjag]

If you can displace the oxidizer/air remaining in the volume why not.

[littlestymaar]

The initial tweet says:

> we had an oxygen/fuel leak

If that's correct, then you can't just remove air. The only option would be to cool things down so it stops burning.

[shellfishgene]

If it was really an oxygen/fuel mix burning I don't think you can do much of anything to stop that.

[littlestymaar]

If you cooled the mixture at low enough temperature, you'd stop it from burning (like when you pour water on top of a camp fire), but it's not clear how you're supposed to do that in a spaceship where you can't carry a few tons of water for your sprinklers.

[ben_w]

> If you cooled the mixture at low enough temperature, you'd stop it from burning (like when you pour water on top of a camp fire), but it's not clear how you're supposed to do that in a spaceship where you can't carry a few tons of water for your sprinklers.

Also water would make it hotter, given this is liquid oxygen.

[varjag]

There are other methods too, e.g. fire inhibitors (like Halon or whatever is allowed now) or shockwave to disrupt fire boundary. But I doubt they are very practical on a spaceship.

[anentropic]

I wonder if it's related to the loose panel flapping about at the left of the screen here: https://youtu.be/qzWMEegqbLs?si=aUlI6zfkH3bZCmVm&t=111

[Alive-in-2025]

This sounds like one of those "and also" things. I'd say you add fire suppression AND ALSO try more to reduce leaks. It's got to be really difficult to build huge massive tanks that hold oxygen and other gases under pressure (liquid methane too will have some vapor of course). Are leaks inherently going to happen?

This is meant to be a human rated ship of course, how will you reduce this danger? I know this stuff is hard, but you can't just iterate and say starship 57 has had 3 flights without leaks, we got it now. Since I have no expertise here, I can imagine all kinds of unlikely workarounds like holding the gas under lower pressure with humans on board or something to reduce the risk.

[wat10000]

This might be one of those components where it just needs to be built without problems, and improved safety means fixing individual design and manufacturing flaws as you find them, until you’ve hopefully got them all.

This can work. Fundamental structural components of airliners just can’t fail without killing everyone, and high reliability is achieved with careful design, manufacturing, testing, and inspection. I’m not sure if a gigantic non-leaky tank is harder to pull off that way, but they might have to regardless.

We’re going to have to accept that space travel is going to be inherently dangerous for the foreseeable future. Starship is in a good position to improve this, because it should fly frequently (more opportunities to discover and fix problems) and the non-manned variant is very similar to the manned variant (you can discover many problems without killing people). But there are inherent limitations. There’s just not as much capacity for redundancy. The engines have to be clustered so fratricide or common failure modes are going to me more likely. Losing all the engines is guaranteed death on Starship, versus a good chance to survive in an airliner.

All other practical considerations aside, I think this alone sinks any possibility of using Starship for Earth-to-Earth travel as has been proposed by SpaceX.

[WalterBright]

High reliability of airliners is achieved by having redundancy of all critical parts. The idea is no single failure can cause a crash.

For example, if system A has a failure probability of 10%, if A is redundant with another A', the combined failure probability is 1%.

That of course presumes that A and A' are not connected.

[wat10000]

Yes for systems, not always for structure. A failed wing spar means everybody dies. For real-world examples, there were two 747 crashes caused by improper repairs to a rear pressure bulkhead or aircraft skin. When the repairs eventually failed, the explosive decompression caused catastrophic damage to the tail in one instance, and total structural failure resulting in a mid-air breakup in the other.

The response to this was to make sure repairs are carried out correctly so the structure doesn’t fail, not to somehow make two redundant bulkheads or two skins.

[WalterBright]

The wing spar is dual, too.

The idea is to design the airplane to survive an explosive decompression failure, not pretend that explosive decompression doesn't happen. For example, on the DC-10, the floor collapsed from explosive decompression, jamming the control cables and causing a horrendous crash.

The fix was not preventing explosive decompression. The fix (on the 757) was to locate the redundant set of control cables along the ceiling. Also, blowout panels were put in the floor so the floor wouldn't collapse.

It's not always practical to fix an older design like the 747. When it isn't practical, a stepped-up inspection protocol is added.

P.S. The 747 was designed to survive a decompression. The oversight was nobody realized that a failure of the rear bulkhead could destroy the tail section. Things like that happen in complex systems, and an airliner is incredibly complicated.

P.P.S. When I was a newbie at Boeing, I asked about the wing spar, too. That's how I know it is dual!

[mavhc]

Given that a) most human rated rockets have had 0 flights before use, and b) I'd expect each starship to have at least 10 flights, and at least 100 in total without mishap before launching, the statistics should be good

[wat10000]

I don’t think (a) is true. The Shuttle flew with people on its maiden voyage, but that’s the only one I can think of.

(b) is true and should make it substantially safer than other launch systems. But given how narrow the margins are for something going wrong (zero ability to land safely with all engines dead, for example) it’s still going to be pretty dangerous compared to more mundane forms of travel.

[laverya]

Most rockets flew test flights before sticking people inside the same model, but most rockets are also single use and so each stack is fundamentally new.

A future starship could plausibly be the first rocket to fly to space unmanned, return, and then fly humans to space!

[WalterBright]

Lindbergh's Spirit of St Louis had the main fuel tank directly in front of him. This was in spite of his primal fear of being burned alive. In some airplanes you sit on the fuel tank.

[api]

Would be unpleasant if there was crew. Of course this thing is pretty far from human eating.

[onion2k]

Would be unpleasant if there was crew.

19 people have died in the 391 crewed space missions humans have done so far. The risk of dying is very high. Starship is unlikely to change that, although the commoditization of space flight could have reduce the risk simply by making problems easier to spot because there's more flights.

[gr3ml1n]

The higher frequency of launches seems likely to have a big impact on reliability. It's no different than deploying once per day vs once per month. The more you do it, the more edge cases you hit and the more reliable you can make it.

SpaceX also has a simplification streak: the Raptor engines being the canonical example. Lower complexity generally means less unexpected failure modes.

[api]

My #1 rule for all engineering: simplicity is harder than complexity.

I truly wish more software engineers thought this way. I see a lot of mentality in software where people are even impressed by complexity, like "wow what a complex system!" like it's a good thing. It's not. It's a sign that no effort has been put into understanding the problem domain conceptually, or that no discipline has been followed around reducing the number of systems or restraint over adding new ones.

I've seen incredibly good software engineers join teams and have net negative lines of code contributed for some time.

If we ever encountered, say, an alien race millions of years ahead of us on this kind of technology curve, I think one of the things that would strike us would be the simplicity of their technology. It would be like everything is a direct response and fit to the laws of physics with nothing extraneous. Their software -- assuming they still use computers as we understand them -- would be functional bliss that directly represented the problem domain, with every state a pure function of previous state.

We might get to this kind of software eventually. This is still a young field. Simplicity, being harder than complexity, often takes time and iteration to achieve. Often there's a complexity bloat followed by a shake out, then repeat, over many cycles.

[psunavy03]

"Perfection is achieved not when there is nothing more to add, but when there is nothing left to take away."

-Antoine de Saint-Exupéry

[exe34]

> Their software -- assuming they still use computers as we understand them -- would be functional bliss that directly represented the problem domain, with every state a pure function of previous state.

I love that this is also a model of reality. Everything is made of differential equations.

[WalterBright]

I've written something good when others look at it and say: "pshaw, anyone could have written that!"

[londons_explore]

> SpaceX also has a simplification streak: the Raptor engines being the canonical example.

Not necessarily. Your engine which used to have 200 sensors perhaps now only has 8. But you now don't know when temperatures were close to melting point in a specific part of the engine. When something goes wrong, you are less likely to identify the precise cause because you have less data.

Many of those sensors are not to enable the rocket to fly at all, but merely for later data analysis to know if anything was close to failure.

In yesterdays launch, if the engines had more sensors musk probably wouldn't have said "an oxygen/fuel leak", but would have been able to say "Engine #7 had an oxygen leak at the inlet pipe, as shown by the loud whistling noise detected by engine #7's microphone array"

[api]

I could imagine the risk going down to a few times air travel after 50+ years of operating a mature launch system.

[BurningFrog]

Modern space ships are very likely to change that, as designs mature and improve.

Early aviation was extremely dangerous. Now a plane is among the safest places to be.

[mark-r]

Planes are safe because investigators were diligent about going to every crash and investigating to find the root cause. That's a lot more difficult when your evidence is dispersed to the stars or burns up on re-entry.

[BurningFrog]

Modern space ships are full of sensors measuring everything interesting.

That data is streamed to Earth, so you usually have pretty good indication of what happened, even if your ship is now a puff of stratospheric smoke .