We had the pleasure of watching the event from Titusville, right across the Indian River, where there is always a crowd gathered for launches. Yes, incredibly cool - the booster really did create that stream of orange light, and those with sharp enough eyes could see the first stage separation. The chest-shaking noise (quieter than an Atlas V but hey), the thunder to the lightning, arrived on schedule, surprisingly much later. Suddenly another stream of light closer to the ground, then it shut off while we all wondered if that had gone wrong. But it came on again, slower and even closer to the ground, while it visibly decelerated and shut off - success! Applause! Then the sonic boom that we all thought at first was a heartbreaking explosion until we remembered it was supposed to happen. A fun time was had by all.
Just one of the side benefits of living in Orlando.
It’s overstating it. The Apollo program was a few orders of magnitude more difficult than what SpaceX did.
…however…
Apollo was an evolutionary dead end. Although it continues to be an unmatched standalone achievement, when the program stopped, that was it. We learned a lot from it, of course, but not much that wouldn’t have been learned anyway. If and when we go back to the moon, it won’t look like the Apollo program.
On the other hand, it may well be that in a couple of decades, all rockets are reusable. And if so, today’s achievement will be seen as huge jump in the evolutionary heritage of those rockets, just as Robert Goddard’s tests of liquid fueled rockets are in the history of every rocket today. In that sense, it may be just as important or even more so.
This isn’t a guaranteed future–there’s still a long ways to go. There was a lot of optimism at the start of the Shuttle program, and that was another dead end. The promise is there, though. If SpaceX succeeds, they will reduce the costs to space by a large factor, possibly an order of magnitude, and this will enable all kinds of new applications for space.
In terms of pure technical achievement, Doc is quite right. But I was struck by the similarity, and differences, with the old footage from NASA’s Mission Control in the '60s. There are still rows of consoles, and people monitoring them, but they’re computers and flatscreens now. The controllers don’t have crew cuts, white shirts, and skinny ties anymore. They hug each other when things go right. The room has glass walls so everyone can watch.
The glee that they showed last night is what I hope their counterparts felt 50 years ago, but expressed in a different way for a different time.
Apollo led to Skylab. And it only dead-ended there because NASA stopped developing the capability further to focus on the Shuttle. If NASA continued to build and update the Saturn-V, they may not be trying to reinvent it now.
To be fair, the decision to end the Apollo program was taken out of the hands of the then-current NASA administration. Funding for Apollo 18 through 20 was deallocated (despite the fact that most of the hardware fabrication and subsystem testing costs had already been sunk), and NASA was put to the course of developing a spaceplane-like shuttle which could also perform various military and defense functions that the Air Force in no way asked for or wanted and which caused the already questionable decision toward a quasi-reuseable spaceplane concept to be burdened with ridiculously conflicting requirements that hobbled its basic functionality and reliability. The development of the Space Transportation System (“Shuttle”) was supposed to presage a permanent orbiting platform but Skylab decayed before the first flight of Columbia to boost it into higher orbit, and it wasn’t until the late 'Nineties that the ISS even began to be assembled. (By the way, there was not “a lot of optimism” about the Shuttle from the technical people working on it, who saw time and again technical compromises being made for political expediency and promises that it would never be able to deliver upon. One former Rocketdyne engineer who headed one of the development of one of the SSME critical subsystems told me that the engine was known to be “a self-cannibalizing piece of shit” from the start due to the machinations which forced Rocketdyne to partner with opponent Pratt & Whitney on the high pressure staged turbopump design rather than the lower pressure aerospike engine that Rocketdyne had designed, built, and qualified in-house using IR&D funding.)
If there is a failure that can be attributed to NASA, it was the failure to plan for and demonstrate a viable and necessary path for a continued crewed space program following the Apollo landings. And to be fair, it isn’t clear that we need or can support a sustained crewed space program at the current state of the art either then or now; an ultimate sustained human presence demands either an absurdly cheap bulk material launch capability or in situ utilization of space resources; shipping consumables and critical hardware into space at even hundreds of dollars a kilogram (never mind the many thousands it costs now at the cheapest) is just not a sustainable model beyond transitory exploration for its own sake. (The SLS certainly isn’t going to provide that capability; it is an absolute pig of a vehicle that shouldn’t cost a tenth of the estimated per-flight cost.)
The failure of the destination-oriented model of the Apollo program should be a cautionary tale for advocates of a crewed Mars mission in the “soonest” timeframe; even if we could achieve such a goal with conventional propulsion and habitat technology (doubtful based on the studies I’ve worked on), even the cheapest budgets for a single mission (at a cost that would exceed the entire Apollo program) are vastly too expensive to support repeated missions. We need the technology and infrastructure development to perform a practical, useful interplanetary mission. Fortunately, this technology would come with its own benefits, including detection and mitigation of space hazards such as solar weather and potentially hazardous asteroids, fabrication of structures and complex processing of materials in space, enhanced communication and Earth surveillance, et cetera, all of which could provide a benefit to the world at large.
Back to the topic of the SpaceX landing, it is certainly an outstanding technical achievement, but I remain dubious that this alone will result in substantial reduction in flight costs. It may seem that being able to ‘reuse’ a stage would be a big savings, but the analogy to “throwing away an airliner after one flight” misses the point that launch vehicles are not airliners; rockets experience extremes of vibration, shock, thermal stress, and operating pressures that are beyond any other conventional engineered system. The useful lifetimes of many components in rocket engines such as valves, manifolds, seals, et cetera are measured in minutes rather than hundreds or thousands of hours. The operating and structural margins are necessarily slim in order to avoid excess mass. The use of supercooled (densified) propellants–done solely to allow enough remaining propellant for the flyback operation–presents additional processing challenges and risks which add to the expense and effort of processing the vehicle for flight, and it is this cost which is most significant in launch costs, and are potentially most amenable to reduction through simplification and automation.
To be clear, in no way have I intended to denigrate NASA for their achievements–Apollo, Shuttle, and otherwise. They behave according to the political environment they exist in, which does enable them to achieve great things, but low costs are not one of those things. I think the current approach of commercial partnerships is a good one and SpaceX would likely not exist without it.
There aren’t enough public details to say how much SpaceX has been able to simplify their ground operations. The few public bits are promising, however. For instance, SpaceX avoids the use of pyrotechnic fasteners. Although straightforward to use, they are one shot deals and can’t be reused. They’re also unsuitable for non-destructive ground testing. Similarly, SpaceX uses pneumatic actuators for the stage separators and nitrogen cold gas thrusters for the ullage motors, as compared to the more usual solid motors.
From what I’ve seen, when given a choice, they’ve always taken the approach of spending extra mass to have a system which is resettable and which is more easily tested on the ground. A nitrogen thruster can be cycled and the tank refilled; the tank has sensors to verify that it is pressurized and there are no leaks. For a solid motor, you have to hope that your quality control several months ago was good enough and that nothing happened in the meantime.
The ideal is a rocket where you can press a button, it does a bunch of self-tests, and when it says OK you have high confidence that it will succeed. I’m sure SpaceX is still a ways off from this, but it’s been a goal of theirs from the very beginning.
I posit that one reason rocket parts have historically been so fragile is that up until now, no-one has ever had any reason to design them otherwise. Now they do.
Well, that’s certainly one reason, but the other reason given by Stranger is sufficient all on its own. Lower mass is so far up the list of design parameters for orbital rockets that the others are barely on the same page. Every gram spent on something other than fuel is a gram of payload capacity you’ve lost.
That’s only true on the final stage. The ratio is much less on the boost stage(s).
SpaceX has said that second-stage resusability is a ways off for just this reason. The heat shields and stuff you need to deorbit are also heavy.
The first stage has much more comfortable margins. They can afford to have landing legs, aerodynamic controls, and reserve fuel. A gram spent here is only 0.2 grams lost on the payload.
Is it the same engine on the second stage (but fewer of them) as on the first stage. Maybe the plan is to put new engines on the first stage, reuse them, and as they get near the end of their design life put them on the second stage.
It’s a very similar engine. The first stage uses nine Merlin 1D (M1D) engines, while the second uses a single Merlin 1D Vacuum (M1DVac) engine. The most obvious difference between the two is that the M1DVac has a very large nozzle bell to take advantage of the vacuum environment. Larger bell => more exhaust expansion => higher thrust => higher efficiency.
I would be surprised if the M1D could be converted to a M1DVac, however. I’m sure the differences are more significant than bolting on a new bell. Also, reliability is more important on the second stage–whereas the first stage can afford to lose an engine for a significant part of the flight envelope, if the second stage loses the engine it’s all over.
It can’t be converted. The shape of the nozzle and combustion chamber is different, it’s not just a matter of bolting a larger bell on, and you can’t really replace the combustion chamber without essentially rebuilding the entire engine anyway. Maybe they could re-use some of the parts, valves and electronics and other small bits, but I expect they probably won’t.
Yeah, this doesn’t surprise me. In principle, they could have designed the M1Vac to be something very close to “bolt on a new bell”, with some attendant loss of efficiency, but there’s really not much point to it. Most of the complexity is in the turbopumps and such, which are common, but at the same time you’re right that swapping out the combustion chamber would entail a full rebuild.
SpaceX has a quite detailed guide to their rocket here. I’ve actually yet to do more than skim it, but it has an impressive amount of information.
Pneumatic actuators are used for payload deployment or regulated delta-V in many vehicles, as are cold gas (typically N[SUB]2[/SUB]) thrusters; this is in no way an innovation of SpaceX. The notion that a reusable delatching or actuating device must be more reliable than a one-shot pyrotechnic or or combustion device seems intuitive at the layperson level from the standpoint of being able to verify function, but in fact the opposite is true. I can address this topic in some depth insofar as I began my aerospace career designing mechanical delatching/latchup mechanisms and later worked with a wide range of pyrotechnic initiators, actuators, and separation/cutting/penetrating mechanisms.
Purely mechanical (non-pyrotechnic or non-mechatronic) mechanisms for latching and delatching have two ostensible states; open and closed (or latched and released, or whatever nomenclature the program uses). However, a mechanism with loose enough tolerances to function without binding may often have a range of various states in which it may settle, and these can have different stored and initiation energy states which are different enough to require separate characterization. In addition, the action of a latching mechanism in releasing stored energy is highly dynamic, to the point that normal mechanism analysis (which generally assumes quasi-rigid or linear elastic response) fails. Mechanisms released or actuated under load often see high stress and exhibit highly nonlinear mechanical behavior; linkages twist or flex, surfaces may experience such extreme friction that they reach melting temperatures, et cetera. Mechanisms are often very sensitive to random vibration, high shock, and coefficient of thermal expansion (CTE) environments. Under repeated actuation, mechanisms often exhibit significant wear which may change the degree of stored energy or the required initiation energy. Any change to the design or variation in manufacture or materials can also have a significant change to the behavior of the mechanism. Do you like the subdued but meaty “clunk” your car door makes when you close it? That’s the result of several tens of thousands of person-hours to get that particular effect, and that is a mechanism that operates under just a few foot-pounds of received or stored energy. A device to restrain and release a separation ring or large payload fairing will operate with hundreds or thousands of foot-pounds of stored energy. In addition, most commanded actuation systems require a power source, which is frequently hydraulic or pneumatic in nature. This requires a pressurized system and pressure lines to each mechanism, which adds mass and complexity with the attendant reliability concerns. (Loss of pressurization or pressure anomalies probably make up at least a third of actual mechanical flight anomalies that I’ve seen; pressurized systems are some of the least reliable and most failure prone systems on a launch vehicle.) In the case of simultaneous or time sequenced release, multiple mechanisms such as fairing delatch are extremely difficult to develop.
Pyrotechnic systems, while not capable of pre-use functional verification, are relatively simple. The typical initiation requirements are just electrical leads to an initiator or interface to an ordnance transfer line. Pyrotechnics, as one shot devices, can be designed to operate with high energy margins since you don’t have to worry about over-stroking or buckling a reusable mechanism (although the desire to limit the shock response or contain debris may limit the energy input). It is also almost trivial to design a pyrotechnic system with redundancy. Although you can’t functionally test a pyrotechnic without expending it, good quality control and lot acceptance testing (LAT) has resulted in modern pyrotechnic devices that are exceedingly reliable. As an example, the NASA Standard Initiator (NSI), an initiator that NASA uses in virtually all pyrotechnic applications consisting of a threaded housing, hermetically sealed charge, hot bridgewire-initiatied charge, and four pin electrical interface, has been build and functioned in quantity exceeding hundreds of thousands of units, and it may be the single most produced aerospace ordnance device in existence. As far as I’m aware, there have only been two instances of a failure attributed to defect or inadequate function of the NSI, and both of these in testing under extreme cold (-290 °F) conditions. (There have been other failures associated with the use of the NSI but those are all attributable to incorrect application and or requirements mismatch.) A R=0.99998 realized reliability in a purely mechanical mechanism operating under such extreme conditions and stored energy would be ambitious to say the least.
There are legitimate reasons to eschew pyrotechnics; specifically, the personnel hazards (both real and presumed), the ESD concerns with pyrotechnic systems, the cost associated with lot acceptance testing; the cost to reject and destroy an entire production lot that fails testing; the high shock response conditions resulting from an ordnance device, et cetera, but the presumption of reliability isn’t among them.
As it happens, there is well-known launch vehicle contractor that aimed to do just this with a specific, now quasi-retired launch system. They built an elaborate set of self-testing and self-diagnostic systems which attempted to identify problems as all stages of integration. The system actually more-or-less worked as advertised insofar as reducing diagnostic efforts. The problem, however, was that the focus of effort then shifted to getting the diagnostic systems to work correctly and debugging the issues it came up with, and the contractor did not end up saving any money or reducing the time to launch cycle. The ultimate lack of success of the system (both in terms of not reducing cost, and actual mission failures not related to problems with the vehicle integration) rendered their efforts moot, but even if they had continued it wasn’t clear that there would be a significant savings. The fundamental flaw is that in trying to implement this automation they added to the complexity rather than removed it. Any real approach to reducing cost and effort needs to focus first and foremost on simplicity; reducing the number of operations and functions in order to reduce potential for failure. Adding more systems to increase reliability is like fucking for virginity; it just doesn’t give you the result you are aiming for even if it is fun doing it.
“Rocket parts” are not “fragile”; most components on a rocket (save for the delicate avionics that are unavoidably sensitive to dynamic and thermal environments) have higher margins than many parts on your automobile or household tools. However, they experience very high environments that are vastly beyond the normal experience of any ground based equipment. For instance, the MIL-STD-810G most aggressive transportation vibration test profile has an overall room mean acceleration level of 4.43 g[SUB]RMS[/SUB] up to 500 Hz. (The root mean acceleration is the square root of the sum of the acceleration power spectral density; basically, a measure of how much overall acceleration per unit octave fraction of spectral bandwidth in the vibration profile.) The NASA General Environmental Verification Standard (GEVS) requires a bare minimum vibration test profile of 6.8 g[SUB]RMS[/SUB] (Table 2.4-4) and recommends qualification at a level of 14.1 g[SUB]RMS[/SUB] and acceptance at 10.0 g[SUB]RMS[/SUB]. In practice, we often see levels on primary structure-mounted components exceeding 24 g[SUB]RMS[/SUB] and I’ve seen measured levels up in the 60 g[SUB]RMS[/SUB] range. In other words, the kind of vibration conditions experienced by launch vehicles are 2 to 15 times as much as that experienced by anything in a ground vehicle driving on the worst washboard road conditions. Shock response can be even worse; the highest shocks in any non-ordnance terrestrial application tend to be around 700 g at 1000 Hz. I’ve seen MPE shock environments on flight vehicles (due to the aforementioned pyrotechnic separation systems) that exceed 50,000 g at 900 Hz and climbing thereafter at the environment source (usually attenuated by distance and traverse through sprung joints).
Components on launch vehicles are not designed for “infinite life” per se, but they are designed to be as robust as feasible given the mass and volumetric envelope constraints. The problem with reuse isn’t that component design is weak; it is that rocket launch vehicles operate at the basic limits of the materials being used, and making them more robust entails adding significantly more mass, which is detrimental to the essential function of a rocket, e.g. carry payload to orbit. The focus on reusability as some kind of game changer is based on the presumption that rockets can be designed for long (if not infinite) life the way airliners are. But even in extreme operations airliners are not exposed to the kinds of conditions and loads that that are experienced by a rocket during any flight. Complaining that we should make rocket launch vehicles capable of surviving repeated flights without refurbishment is like suggesting that we should genetically modify chickens to lay eggs that can withstand being dropped from six feet onto concrete; you might be able to do it, but the result isn’t going to have much utility as an egg insofar as you are going to need a bandsaw to open the thing.
The actual cost of building the physical hardware, while not insignificant, is not the driving cost of a launch vehicle. Actual fabrication costs on systems that I’ve reviewed are somewhere about 10% of the total launch cost. The cost of propellant is typically less than 1% for RP-1/LO[SUB]x[/SUB]. The real costs is all of the effort that goes into mission analysis and verification, and the integration and testing of building up the vehicle and assuring that it functions correctly. Vehicle reuse, by itself, doesn’t save much if any of this cost. The savings to be had is in simplifying and automatic (as much as reasonably possible without adding complexity) the processing and integration process, and spreading the amortized costs of manufacturing and launch facilities across as many launches as possible.
While the Merlin Vac-D (second stage engine) is of the same essential construction and heritage as the Merlin 1-D (first stage engines), they are different engines with different functional requirements. The MVac-D is optimized for vacuum operations, so it has a number of physical differences. Nor would it make sense to put a previously used engine with potential damage into a critical upper stage application where any engine failure or even reduced function may result in a complete mission failure. Rocket components are not Tinkertoys, and unlike Kerbal Space Program, you can’t just swap out components or put in a bunch of struts to make things work and not break.
To be clear, I did not intend to imply that pneumatic actuators, cold gas thrusters, etc. were SpaceX innovations. My point was that when given the choice, SpaceX has always chosen resettable systems. If we want to call something an innovation, it is their design philosophy, as opposed to the design of any particular subsystem.
I also did not mean to imply that resettable systems were somehow fundamentally more reliable than one-shot devices. The difference is in how you verify that a particular part will work. As you note, there are high costs involved with lot acceptance testing and the other mechanisms you need to ensure reliability for one-shot devices. Resettable systems are more forgiving.
As you’ve noted before, the same difference exists between solid vs. liquid fueled rockets. Solid boosters are more reliable than liquid in actual practice. But this reliability comes at great cost as compared to liquid fueled engines, which can (usually) be tested via static firing (the Lunar Module Ascent Engine notwithstanding).
At any rate, I do appreciate the detailed response.
Interesting. Now I’m curious who you’re talking about…
I certainly agree with the simplicity aspect, with a caveat. One shouldn’t just look at a parts count to conclude that one system is simpler than another. It is just as important to look at commonality. Which is simpler: a rocket with a single staged-combustion engine, or one with nine gas generator engines? The latter certainly has more parts, but the former has more unique parts, and those parts are subject to harsher conditions.
On the F9, the pneumatic pushers are certainly more complicated than pyrotechnics. But perhaps not as much as one would think, since it uses the existing high-pressure helium system. Helium is complicated and tricky, but in the F9 is used for at least four subsystems: pushers, latch release, tank pressurization, and grid fin hydraulic pressurization. They amortize the cost of the system over several applications, and when they test it, it increases their confidence in the function of all dependent subsystems.
I do know that the DC-X project had a high focus on automated ground operations, and supposedly required only three support personnel. It’s too bad the project was cancelled, but I think that at least some of the lessons carried over to other parties.
The shuttle was a bad idea. There was no escape tower for the crew if the launch went bad. That meant there could be almost no margin for error. It carries heavy wings and landing gear all the way to orbit. It was said that the three main engines running liquid hydrogen and LOX were really screaming and on the high end of what was safe. Then they dumped the solid boosters in the ocean, which could not have been good.
Stranger, that was all fascinating. Tomorrow morning, when I hear/feel the “subdued but meaty clunk” of my car door closing, I’ll pay a momentary silent homage to the many whose man-hours created that; and I’ll enjoy presuming to make too much of the tenuous (but real) link between this and the engineering of great rocket systems.
