5 ms·

I agree with this assessment and probably the biggest problem in the test. After that, they still have a fair way to go on Raptor reliability; I was a bit disa

I agree with this assessment and probably the biggest problem in the test. After that, they still have a fair way to go on Raptor reliability; I was a bit disappointed with the number of engine outs: clearly the engines have not arrived. I think the big win of the day was the overall resilience of the vehicle. It seemed to handle (what looked to be) possibly explosive engine failures and kept going and stayed intact... even through the tumbling. I think this is a case where the engineers can look at the numbers and calculate for a limited variety of well defined (and foreseen) scenarios, but that they continue to fail to have an intuitive sense of the power of the thing that they built. Given the complexity of the real world, they can "engineer" themselves into an undue sense of certainty about what will happen when they light a rocket of this power... and I think we see the result of that. As an aside, this is precisely where I think we'll some of the biggest failings for LLM based AI... it can get you some nice words to describe a situation and may be able to do all the calculations... but having that gut feeling that maybe there's more to a situation than the numbers would have you believe is still something that humans can do better (though not SpaceX this time around). In closing, if I had a binary choice of "successful" or "unsuccessful" for this test... I'd still probably call it successful. But only just; it's a very qualified success given some of the test parameters SpaceX had set for themselves. (of course we could find out about even more damage to "stage 0" which might shift my assessment... but we know that they're in the process of putting in a deluge system at least already, so I'm being a bit more charitable regarding expectations than I might otherwise be.)

ceejayoz 4y ago
The Raptor issues may well be just a symptom of the concrete issue; large chunks flying up into complicated plumbing isn't ideal.
- nerpderp82 4y ago
  This will be more of an issue on landing (the moon, mars, mercury, the bricks in red square).
  - ceejayoz 4y ago
    Not necessarily; low gravity plus lighter weight means much, much lower thrust on landing in those locations. F9 lands on the drone ships with one engine, for example. It also looked like today's launch involved a hold-down period while they lit them all; that wouldn't be the case for landing either.
    - nerpderp82 4y ago
      That makes sense, but on mars? Full payload, very little atmospheric drag, little atmosphere to keep the blastolith from scrubbing the bottom of the vehicle. I am talking generically since presumably the first stage won't be landing on mars.
      - m4rtink 4y ago
        Lower pressure could help help the exhaust spread more.
      - ceejayoz 4y ago
        Mars is far off enough it's entirely speculative on how they'll handle this. Drop some robot bulldozers first? Specialized landers for landing pad construction crews? Bring the heavy cargo via a separate non-atmospheric craft built in Earth orbit? Hover from height for a while to blow debris off a rocky flat spot? I don't have the answer, but IMO cheap mass to Earth orbit is the biggie that enables a whole bunch of options.
        yetihehe 4y ago
        There was an idea of adding aluminum powder to exhaust and hovering a little over the pad, so that you deposit a solid metal surface where you intend to land.
        gridspy 4y ago
        You missed the moon solution - use engines high off the ground (most of the way up the ship) for landing. Another idea (not starship specific) is to introduce cement (or similar) into the engine burn to build a landing pad from middair.
samwillis 4y ago
> After that, they still have a fair way to go on Raptor reliability I suspect the debris from blowing a whole in the ground was the biggest contributing factor for the engines out. It looked like at lease 4 we're gone before it left the pad.
- rmu09 4y ago
  Isn't it pretty unlikely that this debris, flying against the exhaust, can reach the engines with significant energy?
  - wkat4242 4y ago
    Considering the energy they have comes from precisely that exhaust, I don't think so. I'm pretty sure some pieces would make it through.
  - malfist 4y ago
    Rocket "engines" aren't like engines in your car. The bell and the plumbing for it is the engine. There's not too many moving parts besides a pump. Any sort of damage to the bell can cause a failure. The bells are cooled by pumping super cold fuel over them, a rock hitting those thin lines of fuel could rupture a line, or divert on causing a failure in the bell which eventually (seconds) ends in an engine rich fuel mixture.
  - beerandt 4y ago
    That's my thinking, but ignition is a staggered 6 second duration, so there's at least a window for adverse directional flow. But looking at that pad and the damage to it- without deluge dampening, I bet the RC spalling might have been a bigger issue than just ground blowback. Put a blowtorch to some concrete a and watch how quickly trapped moisture starts causing miniature explosions and flying shrapnel. But that's just my first uninformed impression.
    - cubefox 4y ago
      They have special concrete now which shouldn't explode from moisture. But apparently it didn't hold up at all. I don't think anyone expected this.
jackmott42 4y ago
> but having that gut feeling that maybe there's more to a situation than the numbers would have you believe is still something that humans can do better In my life experience so far, humans having gut feelings it 99% of the time complete nonsense. But people forget the 99 times and that one time their gut was right they tell that story forever about how smart they were.
- angelbar 4y ago
  Like survivor bias.
- trey-jones 4y ago
  Correct gut feelings are just your subconscious running the numbers without bothering you about it. A human having a (correct) gut feeling that the numbers are wrong just means that the human has not chosen the correct numbers to look at, or possibly that a human has made an error in telling the computer how to calculate the numbers. Yes, computers are really good at making calculations, but we have to tell them which calculations to make, so there's still (will always be?) a weak link in the chain.
nerpderp82 4y ago
> I was a bit disappointed with the number of engine outs Who knows if they were controlled tests, you need failures to understand the reliability domain. Engineering is about handling failure, not success. You learn very little from success, esp with so many free variables.
- tibbydudeza 4y ago
  Reminds me of the flash memory cells SSD's are made of - overprovision them because some are going to become/or is bum anyways.
chasd00 4y ago
> clearly the engines have not arrived. I think Raptor2 as an engine has definitely arrived. Igniting and running this many at the same time has some work remaining however.
double0jimb0 4y ago
Pretty sweet that they got to test 33 Raptors at a time. You can find failure modes approximately 33 times faster that way.
- nickpp 4y ago
  Plus the combinations.
- robryan 4y ago
  There was a time where people questioned if 33 of them packed in like this would work at all. Pretty impressive that it does and survives multiple engine failures.
mr_toad 4y ago
> After that, they still have a fair way to go on Raptor reliability They seem to have worked fine individually and in smaller groups. There could be issues with the plumbing, which they seem to be iterating on quite a lot with the newer boosters.

ceejayoz 4y ago

The Raptor issues may well be just a symptom of the concrete issue; large chunks flying up into complicated plumbing isn't ideal.

nerpderp82 4y ago
This will be more of an issue on landing (the moon, mars, mercury, the bricks in red square).
- ceejayoz 4y ago
  Not necessarily; low gravity plus lighter weight means much, much lower thrust on landing in those locations. F9 lands on the drone ships with one engine, for example. It also looked like today's launch involved a hold-down period while they lit them all; that wouldn't be the case for landing either.
  - nerpderp82 4y ago
    That makes sense, but on mars? Full payload, very little atmospheric drag, little atmosphere to keep the blastolith from scrubbing the bottom of the vehicle. I am talking generically since presumably the first stage won't be landing on mars.
    - m4rtink 4y ago
      Lower pressure could help help the exhaust spread more.
    - ceejayoz 4y ago
      Mars is far off enough it's entirely speculative on how they'll handle this. Drop some robot bulldozers first? Specialized landers for landing pad construction crews? Bring the heavy cargo via a separate non-atmospheric craft built in Earth orbit? Hover from height for a while to blow debris off a rocky flat spot? I don't have the answer, but IMO cheap mass to Earth orbit is the biggie that enables a whole bunch of options.
      - yetihehe 4y ago
        There was an idea of adding aluminum powder to exhaust and hovering a little over the pad, so that you deposit a solid metal surface where you intend to land.
      - gridspy 4y ago
        You missed the moon solution - use engines high off the ground (most of the way up the ship) for landing. Another idea (not starship specific) is to introduce cement (or similar) into the engine burn to build a landing pad from middair.

samwillis 4y ago

> After that, they still have a fair way to go on Raptor reliability I suspect the debris from blowing a whole in the ground was the biggest contributing factor for the engines out. It looked like at lease 4 we're gone before it left the pad.

rmu09 4y ago
Isn't it pretty unlikely that this debris, flying against the exhaust, can reach the engines with significant energy?
- wkat4242 4y ago
  Considering the energy they have comes from precisely that exhaust, I don't think so. I'm pretty sure some pieces would make it through.
- malfist 4y ago
  Rocket "engines" aren't like engines in your car. The bell and the plumbing for it is the engine. There's not too many moving parts besides a pump. Any sort of damage to the bell can cause a failure. The bells are cooled by pumping super cold fuel over them, a rock hitting those thin lines of fuel could rupture a line, or divert on causing a failure in the bell which eventually (seconds) ends in an engine rich fuel mixture.
- beerandt 4y ago
  That's my thinking, but ignition is a staggered 6 second duration, so there's at least a window for adverse directional flow. But looking at that pad and the damage to it- without deluge dampening, I bet the RC spalling might have been a bigger issue than just ground blowback. Put a blowtorch to some concrete a and watch how quickly trapped moisture starts causing miniature explosions and flying shrapnel. But that's just my first uninformed impression.
  - cubefox 4y ago
    They have special concrete now which shouldn't explode from moisture. But apparently it didn't hold up at all. I don't think anyone expected this.

jackmott42 4y ago

> but having that gut feeling that maybe there's more to a situation than the numbers would have you believe is still something that humans can do better In my life experience so far, humans having gut feelings it 99% of the time complete nonsense. But people forget the 99 times and that one time their gut was right they tell that story forever about how smart they were.

angelbar 4y ago
Like survivor bias.
trey-jones 4y ago
Correct gut feelings are just your subconscious running the numbers without bothering you about it. A human having a (correct) gut feeling that the numbers are wrong just means that the human has not chosen the correct numbers to look at, or possibly that a human has made an error in telling the computer how to calculate the numbers. Yes, computers are really good at making calculations, but we have to tell them which calculations to make, so there's still (will always be?) a weak link in the chain.

nerpderp82 4y ago

> I was a bit disappointed with the number of engine outs Who knows if they were controlled tests, you need failures to understand the reliability domain. Engineering is about handling failure, not success. You learn very little from success, esp with so many free variables.

tibbydudeza 4y ago
Reminds me of the flash memory cells SSD's are made of - overprovision them because some are going to become/or is bum anyways.

chasd00 4y ago

> clearly the engines have not arrived. I think Raptor2 as an engine has definitely arrived. Igniting and running this many at the same time has some work remaining however.

double0jimb0 4y ago

Pretty sweet that they got to test 33 Raptors at a time. You can find failure modes approximately 33 times faster that way.

nickpp 4y ago
Plus the combinations.
robryan 4y ago
There was a time where people questioned if 33 of them packed in like this would work at all. Pretty impressive that it does and survives multiple engine failures.

mr_toad 4y ago

> After that, they still have a fair way to go on Raptor reliability They seem to have worked fine individually and in smaller groups. There could be issues with the plumbing, which they seem to be iterating on quite a lot with the newer boosters.