My guess is that it’s basically to make it safe against failure. If the engines are fully lit up, this amount of water isn’t going to come close to hitting them, no matter the angle. But if there’s an ignition fault or something, then that won’t be the case and you don’t want to be hosing them down. I don’t think the angle matters much at full thrust–it’ll be pressing the water blast down significantly, and the water won’t have much opportunity to move away from the pad no matter the angle (until it vaporizes).
I’d also guess that they have limited control of the water flow once the system’s activated. Water hammer means they can’t just close some giant valves.
You also want to start the water before you hit the ignition. The water needs to be in place in the air below the vehicle before the engines go from nothing to lots in the span of a just couple seconds.
Very true. If the water spray were designed to go upward into the engine bells unless countered by active thrust, you can’t start the water until a fraction of a second before ignition. Which is pretty fine timing, the kind of thing engineers hate because they’re so hard to trust and so prone to go wrong.
A launch abort after the water starts would be a huge mess to recover from.
Did a bit of math since I was curious how much of the water we could expect to vaporize.
Raptor engines get 327 s Isp at sea level. They’re continually increasing the thrust on the Raptors, but 250 tons each is probably a reasonable guess for the next launch. Therefore it’s consuming 250,000 kg * 33 / 327 s = 25,200 kg/s of propellant. At a 3.6:1 mixture ratio, that’s 5480 kg/s of methane.
Methane has a heat of combustion of 55.5 MJ/kg. If we assume 10 seconds of thrust going directly into the pad, that’s 3 TJ of energy. Which is an insane amount–equivalent to a 0.7 kiloton bomb.
At any rate, let’s say the water starts at 20 C and turns to steam. That requires 4.2 kJ/kg * 80 K + 2257 kJ/kg = 2593 kJ/kg. Which means it can vaporize up to 1,160,000 kg (liters) of water.
Reports are that the deluge system can put out around 1.3 million liters of water. Which is only a tad higher! So we are at least in the right ballpark.
My estimate was higher than it will be in practice, since in reality the steam will be heated well past 100 C. If it reaches 700 C, it roughly halves the amount of water needed.
Still, this is an impressive amount of energy. The deluge system will be enough, but not by as much as I’d have guessed off the top of my head.
Starship has repeatedly shown they hold very little truck with traditional conservative engineering safety margins, at least during development. Better to underengineer, fail, and retry than to expensively overengineer (and still possibly fail, if we’re being perfectly honest).
I wouldn’t say that it’s “margins” exactly, but rather that there is a huge array of practices and establishment knowledge that has been embedded in the industry, and without a continual effort to reevaluate it. Some of those practices are good, hard-won knowledge. Others had utility at one time but are now obsolete. And others were never good, but it was just decided to do it like that at some point and no one ever pushed back on it.
Early on, these things are relatively innocuous. They might cause a program to cost a little more or reduce the performance by an imperceptible amount, but over time they add up to something significant.
Musk has said that if you aren’t adding things back in 10% of the time, then you aren’t deleting enough. And I expect that applies more generally; that is, if you aren’t reversing 10% of your engineering decisions, then they probably aren’t aggressive enough.
It’s something that I experience a lot of in the software world, where it’s fairly easy to reverse a decision (although even here, I find that institutional conservatism makes experimental engineering harder than it should be). It seems to be less common in the physical world.
BTW, a new static fire just happened:
Mmm, nice, clean, white steam. Raptor reliability still isn’t where it should be, but the water system looks great.
Personally, I’ve thought for a while that the whole concept of a static “safety margin” is inherently flawed, and that proper engineering should look in detail at what sorts of failures occur, how often, and with what consequences. Like, a safety margin of making everything 50% stronger than it needs to be sounds reasonable, right? Well, maybe. What if there’s some sort of intermittent flaw in the manufacturing process that sometimes results in a batch of material that’s half the strength it’s supposed to be? If that’s the case, and that failure mode is common enough, then a 50% safety margin isn’t enough. On the other hand, maybe the manufacturing process is so mature and reliable that only one time in ten billion is it any weaker than 95% of its design strength: In that case, then a 6% safety margin would be plenty.
Or maybe you’re so far out on the cutting edge that you don’t know how common failures and defects are in your process, but for which failures would nonetheless be utterly catastrophic. In that case, you shouldn’t be designing one system and making it stronger than you think it needs to be; you should be designing two independent systems, that share none of the same materials, just in case you DO get a case where a whole batch is badly defective.
SpaceX has been subject to exactly this problem. The Falcon 9 CRS-7 launch failed in flight due to the failure of an internal strut. The strut was rated at 10,000 lbs load, but failed at 2,000 due to material defect. Ultimately, the fault was on SpaceX for not screening the parts properly, and for inappropriate material selection. But still, this was a case where even a heavy margin was insufficient.
While I agree with your overall point, I think that aerospace is basically operating as you suggest already. There isn’t a fixed safety margin; instead it’s related to how well the system is understood. Poorly understood systems get a higher margin.
Making two independent systems is fine when you aren’t mass-constrained, but isn’t reasonable for the main airframe for a rocket (which might only have a 25% structural margin).
While my entire career has been spent pushing change into risk averse production, I also know that there are reasons to go slow and limit experimentation,
The first is a lesson I took from “To Engineer is Human” many years ago. The book, subtitled The Role of Failure in Successful Design, discusses design changes that led to catastrophic results. One of the themes in the book is that the failure to fully analyze the effects of a design change on the total system can result in unintended consequences from a change that looks simple and straightforward.
Some of the inherent conservatism that frustrates innovators is based on the inability or unwillingness to perform the extensive analysis necessary to identify these unintended consequences at the system level. But this conservatism isn’t irrational, just risk averse.
Second, in aerospace many of the products we make are one-offs or limited production of a handful of deployed systems. This limits the opportunity to iterate at the final product level. What works for introducing change into smart phones doesn’t work for introducing changes into the James Webb. Space-X has managed to move launch systems more towards the smart phone end and away from the one-off end, which is a good thing and allows more experimentation and change in the product, but it’s still limited production compared to most commercial production.
Agreed. What I’d add, though, is that it’s important to distinguish between cases where the risk aversity is appropriate vs. not. And further, whenever possible, to manage projects in a way where increased risk is acceptable.
The booster reusability program is a great example of this. They made no effort to get things right on the first try. They built some prototypes, and started running experiments on their flights–installing grid fins, testing hypersonic entry, soft water “landings”, etc. Almost all of this was on the customer’s dime–that is, they were running these experiments on flights that someone else paid for.
That was fine, since the experiments themselves carried no risk. Many of them were “failures” in the sense that they didn’t achieve the hoped-for outcome. But these failures also cost almost nothing, especially since SpaceX has no one to answer to (unlike government projects).
So they learned an enormous amount, and was able to iterate rapidly, while also costing very little and not imposing any of that risk on their customers. It’s not something that every project can get away with, but it’s not as common an approach as it should be.
Thanks for the book reference. I’ll give it a read.
And in recent Starship news, SpaceX filed their final mishap report to the FAA:
It should be noted that this is just the final report. The FAA will have had been given draft reports along the way as well as other progress reports. So it’s not the case that the timer starts ticking now; it started ticking months ago. SpaceX may have completed all of the likely corrective actions already. SpaceX seems to be hoping for Aug 31 for their next flight, and while I think that’s very unlikely, it’s probably not as far off as one might otherwise think.
Also, SpaceX turned a profit last quarter:
They’ve had profitable quarters in the past. But this is fairly remarkable given that they are still building out the Starlink constellation and are funding Starship development. Both are multi-billion-dollar projects. But it seems that Starlink is pretty profitable these days, especially now that they’ve expanded to commercial and military customers.
Oh, and for fun, check out this hilariously large tap:
That’s them threading holes in the water deluge plates. The tap is about the size of a watermelon and the employees are taking turns jumping off the plate while holding onto a cheater bar.
Yes. There’s a terrible bias in Aerospace that the processes and risk tolerance you practice for spacecraft with lifetimes of 10+ years are just the thing for producing experimental spacecraft meant to test for months.
Also, despite the publication dates on the book, it was written in the early 80’s. The lessons are still timely, but don’t expect examples from the last 40 years.
They iterate so rapidly, it’s hard to keep track sometimes…
One advantage of the stainless steel construction is that you can afford to build things proactively, just to have something to test with. It’s just laser-cut sheet metal and non-exotic welding. The frowny-face version probably allowed some fit checks and other things while they were still figuring out the rest of it. They’ll just recycle the old one.
There was at least one image (posted earlier) that had those vents above the grid fins:
And the large vent area (which one of the technicians is leaning out of) is in a corresponding location. And the other technician seems to be sitting on top of the inverted “steelpan” deflector bulkhead y’all were speculating about.
It’s literally the hot-staging setup many anticipated.
