Starship Raptor Reliability - The Wonder of Engine-Out Redundancy
Вставка
- Опубліковано 11 жов 2024
- The three last Starship flights have given us some good information on the reliability of Raptor.
Mia and the martians:
www.indiegogo....
Off Nominal Podcase episode
offnom.com/epi...
Rocket Factory Augsburg test anomaly update: • Rocket Factory Augsbur...
Rockets Behaving Badly - Pogo
• Pogo - Rockets Behavi...
@Eager_Space on Twitter
www.reddit/r/EagerSpace on Reddit
/ eagernetwork
/ eager-space-1038430522...
“Blew up because of something weird nobody had seen before” is exactly the kind of thing that SHOULD NOT be in the list of failure modes. But it is the kind of thing you’d expect an organization to run up against occasionally if they move fast and break things. For Crew Dragon, that’s exactly why that test happened on the ground with nobody in the capsule. I’d say they’re lucky they experienced it. Having it be a lessons-learned document instead of a plaque at the launch site is a really big PR win.
I think you should expect it to happen if you are pushing the boundaries and doing really new stuff.
Agree whole heartedly. Im going to suggest this new failure mode to our engineering manager tomorrow morning! I expect enthusiasm to this new revelation! 🤣😎🔥
@@keithrange4457Honestly, I’ve been in design reviews getting grilled with stupid questions from people in military uniforms, things like “how are you verifying that your flight software doesn’t have sign errors in it” and the only response I could come up with was “if we aren’t smart enough and careful enough to get THAT part of things right, we will fail at basically every other aspect of the project too.” I wish they would have done me the honor of asking “what if something nobody has ever seen before happens?” My response would have been “this is a single-string mission with no redundancy. If ANYTHING truly weird happens, it’s probably toast. But in the off-chance it can be recovered, we’ll work on recovery from the anomaly. Turns out we had a lot of anomalies. And we did see some new ones. But mostly they were stupidity-induced.
If I remember correctly, the failure of the COPV was literally "new science" as analyzed by NASA and SpaceX - a failure mode never seen before in an otherwise standard industry part. Nothing to do with "going fast and break things", it's just that SpaceX was unlucky to experience this newly discovered feature of COPVs before anyone else ever had.
You can assign semi arbitrary number based on overall history failures and technological maturity.
The number won't be accurate, in fact it should be overestimated, but it will still be useful when calculating risk assessments.
1:05 When calculating failure probabilities, I like to insert a box for "stuff we didn't think of". It isn't just a joke, however, because it really is possible to roughly(!) estimate the value in this box.
Pick one number based on past experience with things that have surprised you in the past on similar projects, add a second number based roughly on how much confidence you have that you have carefully thought through everything, and finally add a third number based on how novel the situation is.
All three of those numbers are wild guesses, of course, but at least it gives you a number instead of an unwritten hidden assumption, so I prefer to write it in.
If you're going to do that, you should probably be reporting 3 numbers instead of one. The regular failure probability without "stuff we didn't think of", the "stuff we didn't think of" probability, and of course the combined probability, where I presume you assume the probability of known failure modes is independent from the probability of unknown failure modes, though an example of when that would not be the case would be, say, an engine failure with engine-out capability nonetheless results in a cascading failure mode that has not been foreseen during analysis, or a sensor failure causes a software glitch that had not been found in prior simulations or code review. Like the integer overflow on Ariane flight V88.
Wait, no, V88 was an independent software failure, not caused by a separate, known failure mode. So nevermind.
@@J7Handle That's actually a reasonable idea, but we tended to just call the whole thing "Fudge Factor", with no formal method for deciding on what number to put in. The three steps I mentioned are just what I used.
We didn't always do it this way, but one advice I have about fudge factors is that it's better to have one big factor for the entire project than a bunch of little ones. It's rare for every step of a project to be late by the same amount, but common for one or two steps of a project to be way, way off.
Excellent video as always. It really highlights the fact that even though Starship is far from being operational, the Raptor might already be more flight tested & more reliable than the engines of most OPERATIONAL launch vehicles: Vulcan, Ariane 6, Vega, Firefly, all the Indian launchers, most or all of the Chinese launchers, the Japenese H-2/3, Proton, Angara... That's the power of a hardware-rich development program vs. a paperwork-rich development program.
And this is all data from the Raptor 2 engine, the new Raptor 3 is supposed to be quite a bit more reliable and have more thrust combined with increased efficiency
I’m an airline pilot, in the industry we joke that ETOPS stands for Engines Turn Or Passengers Swim.
I thought about adding that in but decided not to...
The sudden jump in reliability every nth additional engine makes sense if engine failure occurred only on the pad or directly after liftoff. However, an engine may be more likely to fail late in flight, when a failure is more tolerable. Any rocket can complete the mission after an early shutdown; it's just a question of how long before MECO that can happen (Super Heavy can lose engines before liftoff; a one-engine core stage might only have a second or two).
Therefore, if we count possible engine failures at any point throughout flight, I think the chances get much better and the vehicle reliability graph smooths out somewhat. Either way, the math gets much more complicated, especially if the chance of engine failure over time in flight is not constant.
@@kevinjohnston4923 Also because 9 is not, in fact, a magic number, as this video pointed out, just a shorthand.
Yes. And it's worse than that because it depends on the details of the mission and the payload that's being carried. I played around with trying to add that in and couldn't come up with anything that wasn't very confusing.
I might replied to the wrong comment... I won't delete this, but meh, my bad.
Excellent analysis as per usual, looking forward to IFT5
I suspect the only real example of cascading failure was the N1. Those were spectacularly unreliable engines, with no computer controls to shut them down early.
The fact that they had to shut down a second engine every time they lost an engine to balance out thrust also didn't help.
And that they couldn't actually static fire the engines to test them. They tested one in four (IIRC) and hoped they caught all the issues with that.
Just an allround shitshow really.
During the first flight of Starship as well, multiple engines taken out by other engines.
They had both an automatic shutoff and engines that were fairly reliable. A complicated bird's nest of plumbing and a panicky automatic shutoff were the issue, not having no control or unusually unreliable engines.
Exceptional content as always! Small nit: internally we’ve stopped using the IFT/OFT acronyms and refer to missions as Flight 1/2/etc. The fact that the acronym has stuck around so long after we’ve stopped using it is somewhat puzzling to me!
Who is we?
@@TheMoneypresident I assume edward works at SpaceX
Noted. And thanks.
@@TheMoneypresident There is an Edward Jacobs who is a production operations engineer at Starbase...and is tolerably active on X, so it's quite possibly him here.
2AM, Eager Space, on it. Watch at first sight baby
Among other things, it makes me happier about the upper Starship landing successfully.
NASA and SpaceX recently stated a SuperDraco propulsive landing is possible and will be used in the extremely unlikely case of multiple parachute failures. IIRC Dragon can land with 6 out of 8 SDs firing as long as the 2 dead ones aren't in the same pod. We have no empirical in-flight data for SDs. The cause of the pad explosion has been removed, the valves are replaced with burst disks. Can we calculate the odds of failure? Of course, if your parachutes have failed that badly I suppose even a 1 in 1,000 chance of a successful propulsive landing is preferable to certain full aqua-braking.
We don't have any empirical data for Super Dracos (SpaceX probably has some), so without that it's pretty hard to know. Though hypergolics tend to be very simple and very reliable.
I'd probably model just assuming that the SuperDraco is 1-engine-redundant.
I wonder about 2nd stage landings that include some degree of cargo. They may need at least 2 engines to run or might have an emergency redundancy that involves lighting up the vacuum engines.
That's an interesting question and would of course depend on the amount of cargo and other factors. Vacuum engines seem problematic as you can't gimbal them to control the thrust axis and they're a long way from the centerline of the vehicle.
There are some gaps in my understanding, but if the superdraco's are set to work with a burst disk does that not mean there is no shutting them off until fuel depletion?
Incorrect. The propulsive landing IS A POSSIBILITY IN THE CURRENT, MAN-RATED DESIGN. The Dracos (Super Dracos??) will be modified slightly and put in place of the original launch abort system. Then, and only then, Dragon will be capable of post-liftoff abort and will scoot away from the Roman Candle like it's ass is on fire ORRRRRRR the system can be used for dry-land landing. SpaceX has said it will land softer than Soyuz, which is famous for it's near about car crash levels of impact. The seats are made to do some moving upon landing to try and help absorb the same force time after time (landing) over a longer duration in order to drop the levels of force as much as possible.
But again, SpaceX has not put the modifications in to do this yet. But they are supposed to after consultations with NASA {As of October 10, 2024}
The attempted booster landings so far had a lot more engine failures than 1 in 100 I believe. (I think the cause had something to do with ice filtration in the tanks.) Also, those failures were correlated (they are positively statistically dependent), so we can't simply multiply the individual engine reliability probabilities to get the probability that all work. In other words, it's more likely for engines to fail in groups rather than as single engines. At least for the ice issue during booster landing.
It's unknown why the engine in Flight 4 blew up. Ice wouldn't have made the engine blow up. It would've destroyed the pump but that's about it
If I'm not mistaken, it is expected to correct the generation of ice in the tanks for the Raptor 3, by stopping recycling oxygen from the engine back into the liquid oxygen tanks, and replacing it with another gas.
@@APMI-OFICIAL Nope. Still tapping from the oxygen preburner. But once the filters are dialed in, it won't be an issue any longer. Kind of like the grid fins on Falcon 9 that used to be made of aluminum and would sometimes melt but now they're not even an issue
@snakevenom4954 Well that is technically a multi-point failure which you can include in the analysis, and as mentioned during the video will create a graph of failure probability that needs to be computed along the flight path.
The first assessment of reliability is the ability to achieve a successful mission (i.e. payload to orbit), once you have stage separation, or engines reburns the term 'reliability' needs to be reevaluated since the aim is not towards a payload but preventing an RUD or uncontrolled flight path.
@@snakevenom4954 I would assume they would switch it up because if you pull from the pre-burner, you know have CO2 and H2O in your tanks that you have to purge before filling up again. If they can do it like the fuel side with a heat exchanger, then you wouldn't need to purge the tanks of contaminates after every launch
Clicking like while the adverts run
Sameee
use adblock bruv
@@kipkipper-lg9vl I have it on PC, I'm currently on mobile
Worth noting that the raptor that didn't start on IFT4 was NOT due to an issue with the engine, but rather an issue with the GSE side high pressure gas spin up system.
Hadn't heard that. Thanks.
Weird that it left the pad. Sounds like a double failure in that sense.
@@ahgflyguy almost certainly programmed to keep going if only one engine failed to start.
The whole point of the video is that Super Heavy is fine if it has lost 3 engines. They did run slightly more risk that they would lose two more engines during the burn.
Where did you get that from? I don't recall Spacex saying anything about what caused that single failure
Cascading failure has happened on Starship already. SN11 hard start, B7 engine bay fires that led to loss of computer control, B9 explosive shutdowns, and possibly B10 as they lost connection with it mid air.
Yes, development is hard especially when you are pushing into new areas and doing new stuff.
Falcon 9 landings are a decent analogy. Things were breaking all over the place in different ways and then it started working better and very soon it almost never broke.
@@EagerSpaceWhat Im saying is that you said its hard to find examples of cascading failure from engine failure, but its already happened to starship several times
And I'm saying that failures in development are not indicative of failure probability in production. This is especially true for rocket engines; you can go back and look at the history of the RS-25 and it was failing all the time in development.
The difference is that SpaceX was actually putting development engines on actual vehicles and testing them in a very dynamic flight environment.
I love using the binomial probability calculator on the internet. It makes it easy to quickly try lots of different combinations.
This is, again and again, such valuable content! I need to watch it multiple times in order to really understand the magnitude of the calculation and transfer the findings to my own (comparably superficial) works. The way you explain it: just wow!
7:18 your 2 engine failure calculation is not how you correctly do the combinatorics, and as a result it's off by a factor of about 2. I think you might want to check some of the calculations.
We are assuming a binomial distribution of engine failures with n = 9 and p = 0.01. The correct probability for exactly 2 failures would be 9C2 * 0.01^2 * 0.99^7 = 0.003355... and the correct probability for at least 2 failures (which is the important number) would be 1 - 0.99^9 - 9C1 * 0.01 * 0.99^8 = 0.003436...
New bucket list item, CTP or multi-crew with Eager Space!
Great video! I learned a lot. Thank you.
I welded some fuel tanks together in my garage, and I was ready to light this candle.
But now you say I have to consult a professional rocket engineer? You just love spoiling everyone's fun, don't you?
Yes. My middle name is "Fun Spoiler"
Nice analysis
another great video, thanks man
Glad you enjoyed it
In regards to Starship landing, im pretty sure its a 2-step process.
From the test articles, they need 2 engines to flip, and 1 engine to land, with a flight profile thay initially lights 3 engines for the flip and down-selects to 2 for landing. This means they only have 1 engine of redundancy throughout the landing process, not 2, and the only way they have 2 is if they get 1 failure in each stage of the flight.
Or I guess if they have enough time to start one on the landing burn after one has failed.
They light all three engines and then cut it down to one because there's no time to try engines sequentially.
@@EagerSpace It's light all 3 and cut it down to 2 IIRC, at least after SN10.
7:28 your math here is.... pretty bad. 0.99^9 is the chance of having 0 engine failures, so 1-0.99^9 = 0.0865 is not the chance of one engine failing, but the chance of at least one engine failing. The chance of exactly one engine failing is 9×0.99^8×0.01 = 0.0830.
The chance of exactly two engines failing is (9×8/2)×0.99^7×0.01^2 = 0.00336, while the chance of at least 2 engines failing is P(>=1 engine failure) - P(1 engine failure) = 0.00344.
What you did is just not how binomials work.
Yes. I think the level of math that I did here is pretty much the limit that my audience is capable of understanding or is prepared to watch. I generally put in something about how this answer isn't the actual right answer but is only for illustration purposes but for some reason I forgot this time.
I do find it to be a bit of a strange criticism. I'm making unwarranted assumptions about engine reliability without empirical data, I'm pretending that engine failure during flight has the same effect as as engine failure at launch, and I'm *totally* ignoring the effect of differing payloads, different target trajectories, different margins for different flights, different approaches for crewed and cargo flights, and whole host of other factors.
The answers I end up with a wrong for a host of reasons.
@@EagerSpace Yeah, I get that this is very much a qualitative argument supported by oversimplified approximations.
But I see your channel as a sort of fusion of approachable educational content and areospace enthusiasm. And in that context of educational content, I think its especially important that you follow through on your sometimes questionable assumption with mathematically and logically sound argumentation.
I can totally understand if you want to be more like Scott Manley and less like 3Blue1Brown, and you don't need to explain every mathematical concept you come across in rocket science. But at the very least skip the explaination instead of putting incorrect maths up on the screen.
And in the case of this video, the whole argument is essentially just an applied excercise in binomial distributions. If I wanted to make an interesting 10th grade maths class, this is the type of problem I would choose to explain/ practice binomial distributions with. So it's baffeling to me that you would chose not to touch on the topic at all.
Shrug. The main point of the video is that engine redundancy is a really useful concept, far more useful than many people realize. I could have spent more time trying to be more rigorous in how I did the math but it's not going to improve the quality of the overall answer because all of the other simplifications I made are more wrong than the math is, so my judgement is that it adds complexity with no increase in quality.
You are basically arguing for a different audience than I'm trying to target. Which is fine - I'm always happy to have feedback on what line I've chosen for a particular video.
@EagerSpace I'm arguing you have a responsibility to your audience not to spread mathematical misinformation. I'm arguing that showing no mathematical calculations and skipping straight to a correct mathematical conclusion is better than showing incorrect mathematics.
@@EagerSpaceYour audience obviously wants the mathematics corrected, that's why we're commenting. We are capable of understanding, and the video wouldn't have been any longer or more complicated if you used the correct values.
There are comments by people who didn't notice your error, planning to apply your faulty math to their real-world problems. Your incorrect judgement is hurting people, and most of them aren't going to comment about it.
You can start by pinning one of the comments correcting the maths.
On that shuttle mission that had one engine failure, it turned out that the engine didn't fail, it was a bad sensor.
Suspecting this, when a second engine also showed imminent failure, a ground controller overrode the shutdown command.
If she hadn't, not only would the mission have failed, but it would have probably resulted in LOV, loss of vehicle (and crew). The shuttle was not in a survivable abort mode.
"Bad sensor that causes the engine to shutdown" is an engine failure.
Another consideration is an engine not just failing, i.e., needing to be shut down, but actually exploding. In that case just one engine exploding can cause loss of mission. In the last test flight, an engine exploded during the booster landing burn. The booster itself did not immediately explode but it is likely the engine exploding damaged the structure causing the booster to explode soon after ocean touchdown.
Great video, as usual.
Something you didn't mention
More engines + more flights + landings leads to data collection and design improvements!
they could have just made engines that don't melt to begin with
great work
An additional remark on the failure chance of Starship v3.
If I remember correctly, the v3 booster was advertised for a 10,000tf liftoff thrust, and a Raptor 3 sea-level performance was marked to achieve 280tf thrust. This implies that a v3 booster requires a total of 36 raptor 3 engines, which would further decrease the risk of failure plotted in 12:30.
Additionally, for people discussing failures after stage seperation or payload insertion. Failures after those could include multi-point failures not directly associated with the engine which you must include in the analysis, and as mentioned during the video will create a graph of failure probability that needs to be computed along the flight path.
The first assessment of reliability is the ability to achieve a successful mission (i.e. payload to orbit), once you have stage separation, or engines reburns the term 'reliability' needs to be reevaluated since the aim is not towards a payload but preventing an RUD or uncontrolled flight path. Of course, for starship this must also have a low probability since the aim is to achieve soft landing for both stages.
Very informative video. Now it makes sense to me of SpaceX's multi engine redundancy designs.........you got me on Tachyon Tablets, I was about to look for them when after a few more seconds the joke hit me.💊🐙
Like the video. Three nitpicks I didn't see other people bring up.
First, the cascading failure already did happen on Flight 1, with exploding engines taking out multiple engines to their sides. Of course, a lot has changed since then, but it isn't exactly unprecedented.
Second, you might not have engine redundancy during the final stage of the landing. If you're hovering twenty meters above ground and your engine malfunctions, you might not have the time to spin up another engine. B1062 may be an example of that.
Finally, and this is a personal opinion, but I have doubts on Raptor 3 achieving the same reliability as Merlin. That is because the best way to increase reliability is to not push at the 'edge of physics' as SpaceX likes to say.
Thanks.
First, I don't think IFT-1 is representative of what we can expect from raptor. If raptor was that bad the chance we would see the performance we've seen in the following flights is very low.
Second, this is a fair criticism. I'll only note that engineers that start and get into steady state and appear healthy rarely have issues.
Third, I think flight hours and post flight analysis will help a ton here, but the high power does make it harder.
The best part is that most likely no engines between starship flights 2-4 really failed. The fuel supply and filtering systems did
This is the 4th time I've clicked the bell, let's hope it _actually_ sticks this time...
Algorithm, yes I said it and that's why I'm commenting
This is why I always try to design redundancy into KSP RP-1 designs. Having 5 out of 7 engines fail on a stage before burnout can be nerve-racking though.
Mainly in that it causes more engines to go beyond their burn time and therefore become more likely to fail.
What about in-flight engine shutdowns? A lot less consequential on multi engine setups
I think this video is inspired by the IFT-4 video of the pressure fed. Anyway engine redundancy reduces payload.
Ift-1 is a good example of cascading failure. It's engines didn't run hinky, according to sensors. But in reality, due to loose flange connections, hot gases were escaping from the gas generators and slowly melting the hydraulic lines and other engines
Where did you find out what the engine instruments were reporting?
Discovering bugs in Fortran would happen in a plot of rate of discovery versus time since it was released. The rate of discovery would drop linearly on a log scale. This allowed estimating how many more bugs remain undiscovered. Would it be possible to estimate how many technical surprises could remain undiscovered in rocket ship design & operation by using a similar approach?
My first job in 1986 was in Fortran...
I think the answer is probably no. Software projects using a given language and environment have a lot of similarities, but rockets are bespoke projects that are all quite different.
Compare dragon to starliner...
Of course, the graphs you are showing are for simplicity where you have a step at every multiple of 9 is for simplicity.
For one, the step ratio probably goes down as the number of engines goes up.
And in reality, engineers would work out the number of engines they need, then add the number of engines they can conveniently add, meaning they are always at the top of the step.
The way you present it make it sound like a 27 engine starship would be more reliable than a 33.
Yes, I'm ignoring a whole bunch of factors; for topics like this there's always a big tension between an answer with enough rigor to be at least "sortof right" with an explanation that is going to go over my audience's head.
15:30 "Current Starship has three sea level engines in the middle with small nozzles - Current Starship has three sea level engines in the middle with small nozzles..." You need to cut out this redundant repeat.
PS:
In every instance where you believe "losses," the real word is "loss," gravity loss is loss related to gravity as the cause if loss and the two words being paired together is called a phrase. In this phrase, multitude is implied, you don't need to make it appear in ugly/redundant emphasis as "losses."
It seems like the old N1 failures convinced a lot of people that Superheavy will never work with so many engines.
it's not the number of engines but the design, why go for max performance in an engine that by definition has to be supremely reliable
Did you see about the documents for a 35 engine superheavy? This would be a lot more redundant and also i belive the booster is able to lose a max of 4 engines. Also yes it was confirmed by elon that ift 4 engine was an early shutdown
>inb4 video takedown for showing the dragon test explosion.
Is that a thing?
It's been on the wide internet for a while now
16:10 SN9 crashed because it only had one working engine, Starship with V1 Raptors at least needs at least 2 engines to land, maybe higher thrust engines could do it but Starship itself will get bigger as well.
Ift 1 which wasn't mentioned in the video left many people worried about Raptor reliability. It turn out rather than focusing on switching raptors all spaceX did was to improve the engine startup sequence on their ground units and...problem solved, IFT 2 went flawlessly in regards to engines
I thought the problems with IFT 1 were leaking gases behind panels which led to fires and explosion due to not properly sealed flanges/connections and a very much undersized fire surpression system
@@meinking_sensei3807 This is what led to loss of vehicle. But IFT1 already lauched with missing engines.
What if they came in aft first, which would be passively stable because of the weight of the engines, and did the methane transpiration thing through the already existing pipes that feed propellant to the nozzles- but didn't ignite it, just let it form a buffer against the plasma. Then they don't need tiles *or wings
Great analysis, but I'm going to be *that* guy. Your analysis makes the assumption that engine failures are random and independent. For instance, on that space shuttle 51-F, I think that the same sensor issue was happening on both engines, or maybe all three. On the golden bullet flight, they did have two independent failures, but if I recall, one of the failures actually mitigated the other a little bit. Also one of the Falcon launches back when they were still using the square engine pattern did have an explosive engine failure where they were lucky that it was contained to just the one engine. They ended up putting in extra armor in between the engines at least for a little bit.
Did the number of engines out on Starship flight one influence your choice of 1-in-9 is survivable?
There are many ways to have a bad mission, and given the benefit of engine redundancy those other ways are going to dominate the reliability of the stage as they will be much more likely to happen.
Flight 1 was concerning and surprising until I later realized it was a throwaway flight. They didn't have the engines they needed at that time and the rest of the vehicle wasn't in great shape either. It was really a pathfinder vehicle that they decided to launch anyway.
Wow 1680 Soyuz launched?!! I knew it was a large number but that is a LOT.
786 of the Soyuz launch attempts were with the Soyuz U variant, which flew from 1973 to 2017 - almost 44 years.
"Shutdown doesn't mean failure, just look at Apollo 13"
I mean, to be fair, it ended in a desperate rescue mission using lunar landing engines to speed up the trip back to Earth so they don't suffocate to death. If the engine failure wasn't a sign the mission was cursed, I don't know what would be.
You're connecting dots that don't actually connect. So it's not 'fair' it's meaningless.
What about partial engine failures that reduce thrust output (unlikely), or much more likely, midflight engine failures? A flight that has one immediately engine out and loses another halfway through first stage burn could be considered a 1.5 engine failure flight. A 33-engine booster will undoubtedly be more resilient to a non-integer engine failure count than a 27-engine booster, so your graph should probably look more like a regular ascending exponential curve, rather than several disconnected descending exponential curves.
If you consider engine failures during flight, basically every engine except one can fail during flight and you could still have a successful mission. It's all about when they go out - as an example, if an engine shuts down every second for the last 33 seconds of the booster burn, that's probably recoverable by the second stage.
The reality is worse than that. Different missions have different payload masses and fly to different trajectories and you may fly with different margins for different payloads, and it gets worse if you fly people because then abort sites matter.
The challenge I have in topics like this is that I can't actually give a truly correct answer or the real way of doing things, because it's not really feasible to do that in the amount of time I have and at least 95% of my audience wouldn't understand the math or stick around for it anyway.
Beside engine redundancy, the complex plumbing it took to feed 33 engines is not trivial, and the cause of the N1 rocket's downfall if I remember correctly. Would we consider the potential plumbing failure as part of engine failure statistics or is it a trivial matter once the plumbing is figured out? Just food for thought.
The N1 had a bunch of problems all over the place. It did have plumbing issues but it had engine issues and control issues as well.
I do think it's a little odd to be concerned about it when Falcon has been flying with 9 engines very successfully and we've seen three flights of Super Heavy now with *maybe* one engine failure on boost. There is a lot of complexity in the plumbing but I dont' expect it to be an ongoing problem
Have you run these numbers for the N1, which had a similar number of engines to Superheavy, but less success trying to launch? Plus there are a small number of the follow-on engines that could be used to check the maths. (Though some of the problem was more with the control system than the engines).
N1 was less robust. Because it used differential thrust for control, in the event of an engine failure it needed to shut down the engine on the opposite side. So one engine failing meant two engines shutting down. But even so, had they gotten engine reliability up sufficiently, it should have worked. They also had lots of issues that weren't engine related, though.
What would be the point in running them for the N1? He didn't run numbers for any specific system even here.
This vehicle flies empty and basically runs out of fuel. This means no payload. While full reausibility is great, it comes at a cost. In this case, the cost is really big. Hopefully, a lower weight vehicle, more thrust, more Isp, more fuel, less drag, will fix the problem. This is a long way from an operational vehicle and this is assuming they can catch it reliably, turn it around quickly and that reentry will work as designed, all things that have to be demonstrated
Of course. One would hope it's obvious that this is a system IN development still. After all they're still using low fidelity throw away test articles..
All is proceeding (mostly) as Jerry Pournelle had seen.
Eager raptor~
35 engines on Super Heavy now make a lot more sense 😅
How would reliability change with reuse? For example, I can imagine a case where a Booster is flown 5 times (arbitrary #) uncrewed before being flown crewed, as the reliability of both the Raptors and vehicle is higher -- similar to how with Falcon 9 there is a goldilocks zone of boosters that have been reused enough times to be reliable but not too often to have wear and tear that compromises that reliability.
Hard to talk about without a lot more data, but I think the closest analogy is airplane engines. The big factor with reuse is that you get failed or weird engines back and can examine them and find the issues.
12:44 What if we were to take this graph and plot it differently, where redundancy is gained for every 8th engine? having 33 engines would mean that 4 engines could fail, leading to closer to one in two billion chance of mission failure.
Isn't starship only rated for 3 engine failures?
@@an.ma.2937 It is at take off. I'm pretty sure it will till work after 4th engine goes off, but with more than 3 offline at launch, made the risk just too high.
I don't understand why 27+6 is so much worse than 27.
Surely each additional engine gives you some redundancy, like 1/9th additional potential failures. But maybe that makes the math too complicated.
If you have 33 engines and you lose 1 of the outer ones you have enough ability to gimbal that you don't need to shut down one on the other side, as we saw on IFT-5.
When you add an engine your reliability is worse because of more engines, and it only goes up when you hit a threshold that gives you more engine redundancy
And that's probably why SpaceX will change Superheavy to have 35 engines, they have already published this in their plans. Right now they need the three center engines (the ones connected to the header tank) to relight for the landing burn. They have no engine out capability during landing.
It's from environmental assessment which account for worst case scenario. For the existing 2022 EA they listed 37 Super Heavy engines even though in reality they only used 33
Don't jump on the guns just yet
The booster does not have header tanks. Only the upper stage does.
@@debott4538The booster does have landing tanks integrated with/ near the bottom of the methane downcomer.
@@debott4538 There's an internal oxygen tank right above the methane sump. There's no methane header tank because the downcomer is enough.
Someone should probably show this to Thunderf00t
16:15 is there enough time to detect an engine failure, fire up an other one and correct for landing?
No. In the starship tests we've seen, all three engines are started and then they shut down the ones they don't need.
@@EagerSpace so in case of mid landing procedure failure there is a total loss?
7:28 Actually, it's even better. The chance for k engines out of n total to fail, with each having a reliability p, is (n choose k)*(1-p)^k*p^(n-k) - you have k engines fail with probability 1-p, n-k engines not fail with probability p, and there are n choose k ways for those engines to be distributed among the n total engines.
For n=9 and p=0.99, the chance of one engine failing is 0.083, of two engines 0.0034, of three engines 0.000079, of four 0.0000012, and it keeps going down. So, with 99% engine reliability, nine engines and one-engine-out capability, the odds of loss of mission are about one in 287
The 0.086 figure you calculated as 1-0.99^9 is actually the chance for any number of engines to fail.
15:31 also a redundancy?
why go for max performance engines on something that is supposed to be supremely reliable and reusable?
Because it doesn't work otherwise.
See my why is starship late video
@@EagerSpace sounds like bad design philosophy to me
@@kipkipper-lg9vl Sounds like you might not know much about rocket or rocket engine design... Like a lot of us.
@@TheEvilmooseofdoom it's not that complicated to be fair, simple rockets blow up less, complex ones blow up more
if you want humans to fly on it better keep it relatively simple, else you get columbia again
I used to agree with you, and then I realized that the reason they are pushing Raptor so hard is that they can't get to full reusability without the high performance, at least with their current architecture. Or, to put it another way, the raptor performance is only as high as it needs to be to reach their goals.
That's not quite true, but it's the right spirit.
I talk more about this in the "why is starship so late? " video.
Blopper at 15:35 haha
Long March 2/3/4/5 family have 500+ launches. They shared the same first stage and shared the same 4 + 1 YF-21 (sea lefel + vacume) engine (and 2/4 more with launches with booster) That should accounted for a very large number of engines in flight (They are hyperbolic and super reliable)
Also Superheavy / Starship does have a single failure point on the engine system, the TVC system. If any TVC got stucked in position, the vehicle will loss it's TVC. Hopefully the reliability of the TVCs are really good.
Thanks for the info.
There are many ways to fail outside of the engines.
@@EagerSpaceNobody minds when you leave in a repeated prepositional phrase at the beginning of a sentence, especially nobody who has ever cut together a video of themself talking. The content is what matters, and it is excellent.
In aggregate, they have had a number of launch failures sprinkled among the versions 2, 3, 3b, 4b, 4c, 5 (and maybe others)... Note Intelsat 708.
Now maybe the first stage wasn't the problem, but as Wikipedia noted about the Long March rocket family as a whole: "518 were successful, 10 were failures, and 9 were partial failures. The cumulative success rate is 96.5%."
I’m not sure your failure mode is likely. If one TVC fails, they can turn that engine off. If turning it off would result in an abort, they would probably be able to compensate by using most of the remaining TVC engines or adjoins throttles on all of the remaining engines to effect the steering.
@@PetesGuide What I mean is that the TVC moves together, if one engine is not moving, other engine will vector into it
Unfortunately the redundancy charts like at 7:00 only assume that the engine has failed before the launch has even started. This saw tooth chart scenario becomes more unrealistic the more engines you add. More realistically, engines may fail at a slightly higher rate within the first few seconds and then become less likely to have anomalies (and get shut down) as time goes on (until they reach their maximum intended burn time). This would not be the bent saw tooth you show. Considering this, the reliability would not reduce as you add engines, but only increase beyond a certain number of engines. I guess your trying to keep it simple, while misleading people in the process.
Yes. I did mention that as a simplification. The problem with doing it the complex way - to try to get an answer that is "right" - is that the calculations are very complex, depend on information about the vehicle we don't have, and - probably worst - are different depending on the payload mass, the target trajectory, and what sort of margins they chose for the launch.
Jeez. Planes can't emergency land in Ireland? They have airports don't they? 😄
I believe it's too hard to avoid the leprechauns...
@@EagerSpace Oh. And thanks for letting that long clip of the explosions play out of the AMOS-6 Static Fire so we could experience how long it takes for the sound to catch up to the light. What a crazy entertaining reality we live in. I love that.
The clip is really surreal - you just watch and watch and watch and then all of the sudden the sound gets to you.
5:47 "Sim-are-o-lee" - Huh!?
Similarly is pronounced sim-ew-lar-lee, why is there an "are-o" in the middle of your pronunciation?
Don't switch to A.I voice, keep on keeping on in trying to master your own speech centres.
Thus isn't "hate."