3D Printing An Entire Rocket

If you’re ever flying into LAX and have the left side window seat, just a few minutes before landing, look out the window. You’ll see a small airport just below you and what appears at first glance to be a smokestack. That’s not a smokestack, though: that’s a rocket, and that’s where SpaceX is building all their rockets. Already SpaceX has revolutionized the aerospace industry, but just down the street there’s another company that’s pushing the manufacturing of rocket engines a bit further. Relativity Space is building rockets. They’re 3D printing rocket engines, and they’re designing what could be the first rocket engine made on Mars.

Bryce Salmi is an avionics hardware engineer at Relatively Space, and he made it out to the 2018 Hackaday Superconference to tell us all about manufacturing rockets. It’s an entirely new approach to manufacturing rockets and rocket engines with a clean-slate design that could eventually be manufactured on Mars.

There is a lot of work that goes into manufacturing a rocket. There are jigs that could cost millions of dollars, there are tanks that have to be pressure tested, and there is a vast amount of labor involved for what is essentially a very carefully controlled explosion. This is changing with Relativity Space’s first rocket, the Terran 1. This is a 100-foot-tall, 7-foot diameter rocket that will launch 1250 kg to Low Earth Orbit. It uses Oxygen and Methane — the same fuel that SpaceX plans to use to return from the surface of Mars — and is almost entirely 3D printed. In fact, Relativity aims to build a rocket from raw materials and have it fly in two months.

If you’re printing a rocket engine, you need a 3D printer, and for this Relativity is using Direct Metal Laser Sintering, where metal powder is spread across a bed, a laser melts the powder into a pattern, and another layer of powder is deposited. While this sounds futuristic, it’s becoming a fairly standard industrial process, with Boeing and Airbus either looking into DMLS for manufacturing parts or using it in planes already. Relativity has designed their AEON 1 engine to be entirely 3D printed, reducing the thousands of parts that would go into a rocket engine down to just three. The entire engine is designed to be 3D printed, and there really is no other way to make this engine.

But Relativity is talking about 3D printing an entire rocket, not just an engine. You can’t buy a machine that will print something 100 feet long and seven feet in diameter; you’ve got to build one. That’s what Relativity Space did with the creation of the Stargate. The Stargate (named after StarCraft, by the way), is a wire deposition 3D printer designed to print tanks and structural bodies for the Terran 1 rocket.

Inside the Stargate, there are several robotic arms, each equipped with a wire deposition head. These arms can be used together where one arm can print, another arm machines away the excess, and another inspects the work, all at the same time.

What Relativity Space is doing is special. Making rockets is expensive, because no company will ever make very many — a low volume — and the labor that goes into these rockets is very high. By reducing the part count of the engine, manufacturing tanks and structures with 3D printing, and reducing the tooling cost, Relativity space is building rockets that are cheaper and possibly lighter than anything else out there. Since these rockets can be manufactured with robots, this might just be the rocket we launch from Mars.

43 thoughts on “3D Printing An Entire Rocket

    1. The same people also think that 3D printers are going to be able to reproduce themselves endlessly. Grey goo and all. Conceptually with the right hardware and raw materials and everything else in place? Sure, it’s technically possible. In practice in 2019? Not even close. Even DMLS has fairly poor surface finish and needs to be further machined after it is done printing. You are also resolution limited but the raw materials at least are great. You can also use titanium powder to actually get near net shape without needing to machine 95% of a block away. Saves a ton of raw material costs (which are a part of but not the major cost of space exploration). But it’s still in a very inert atmosphere and being made on five or six figure machines that are not going to suddenly become $500 from China or wherever.

      Also “with Boeing and Airbus either looking into DMLS for manufacturing parts or using it in planes already.” makes no sense or at least is so overly broad to make no sense and shows that fact was not researched at all. They are either looking into it or they are using it already. Hint, it’s the latter.

      All that said, what Relativity Space is trying to do here is amazing and a great use of certain 3D printing technologies. You also see similar things in other niche markets right now. Supercars and the like. Has to start somewhere.

      1. Boeing already uses the technology at great advantage.

        My friend’s wife works there as an industrial patent lawyer.

        The advantage of 3D printing a part is that it reduces the amount of QC control issues and inspection steps that are required. (I am vastly understating the number of parts)

        In the earlier manufacturing process the process went like this.
        a) Design the parts as an assembly for manufacturability requiring ten base components that will eventually be one assembly
        b) Review the design
        c) Manufacture ten separate parts that will form the engine (these are the base components)
        d) inspect all ten (inspection cost) Inspection may include radiography/NDT methods
        e) the ten (10) base components are welded together to form four (4) subsubcomponents.
        f) Inspect AND CERTIFY each subsubcomponent formed. NDT including radiography or X-rays will DEFINITELY BE PERFORMED. Four (4) inspections performed in total
        g) Weld the sub-sub compenents together into two (2) subcomponents
        h) Inspect and CERTIFY the two (2) subcomponents. NDT radiography and X-rays will DEFINITELY BE PERFORMED
        i) Weld the subcomponents into the final component
        j) Inspect and CERTIFY the final design. NDT radiography and X-rays will DEFINITELY BE PERFORMED
        j) Add any remaining bits and test
        k) Ship to client.

        In this scenario
        Steps ( d = 10), (f = 4), (h = 2), and (j = 1) involve the time and cost for 17 inspection steps alone
        (Assume that each block eats a minimum of 3 days, plus shipping and handling efforts). This is a major time eater, and steps C,e,G,i are points where something could go wrong (assuming the design was done correctly). Remember that this is inspection and certification to AEROSPACE LEVEL STANDARDS.

        Using 3D metal printing for complex manifolds
        a) Design the parts as a single block design
        b) Review the design
        c) 3D print the design as a single block
        d) inspect AND CERTIFY the single block design. Inspection includes radiography/NDT methods & X-rays to test for occlusions and fail points
        e) Add any remaining bits and test.
        f) Ship to client

        Now we only have one or two inspection and test operations, and only one NDT inspection stage.
        This radically reduces costs to aerospace industry manufacturers.

        In this case your improvements are
        a) reduced time to market
        b) reduced tooling costs (it’s not the printer its the jigs and fixtures)
        c) reduced inspection and certification costs

  1. I don’t know, anymore it’s hard to take a rocket company seriously if they aren’t onboard for re-usability. SpaceX has shown it can be done, and while we can still argue about how much you save, I don’t think anyone is really arguing it’s BETTER to throw the things in the ocean after each launch. Even if you are squirting them out of a printer.

      1. A common pedantic point, but an invalid one. The Shuttle was not a true reusable booster, no matter how much you want it to be. Other “reused” vehicles have been sub-orbital, or one off proof of concepts. Nobody has ever repeatedly flown reused orbital boosters like SpaceX is currently doing.

        Their launches are now routinely using reflown hardware (booster and capsule), with a refurb time measured in months. It even looks like fairing recovery is now viable. Second stage isn’t going to be recovered, but with a rumored cost of only 10M, it’s hardly a big loss.

        We don’t know how much they are saving on refurb versus new hardware, but the fact that they are continuing to do it at all as a commercial company proves they are at least saving enough for it to make sense. Why else would they stop producing Dragon V1, or the Block 4 boosters? If reuse wasn’t saving them money, they would have just kept building them as before.

        1. I was talking about the previous tail-landing tech demos that already proved the concept was technically feasible, and the only question remaining was about how much fuel (payload) you want to waste to return each stage. That also factors in, because the returned rocket won’t lift as much, so you pay the cost for less capacity making each launch count for less, and the Tsiolkovsky rocket equation is a hard one to bargain with.

          Ultimately, if you wanted to return the entire rocket in one piece, your payload fraction would pretty much be the paint on the skin.

          But onto the point: even according to Musk’s rather dubious calculations that often ignore half the costs, they’d need to fly each booster about 10-12 times to make it profitable, but if they’re following the industry average curve of 4-5% failure/loss rate (after many more decades of experience than SpaceX), there’s only about 50/50 chance that any one booster will break even.

          They’ll probably do worse because after all, they’re using already flown boosters which can have any number of faults that simply aren’t discovered because it costs money to strip them to bits and check each part (a task that would be made all the more difficult with 3D printed monolithic engines).

          1. Nice math. If there’s also 5% change that the booster will break (not even), it can happens in any of the 10 flights that it’d be used to. It’s more likely than it happens in the last flights (because, well, when it’s broken, it’s not going to be reused…).
            So let’s say their process is good enough to reduce that 5% failure rate to only 4% (that’s a small improvement here), then instead of 20 working flights, they would get 25 of those, so more than twice the minimum required for profitability.

            In the end, when it’s profitable, it’s profitable enough to justify the technology. With a 3% error rate (or 33 flights per rocket), that’s more than 3 times the profitability. With a 2% error rate, it’s 5 times more profitable and so on.

            Every percent lost on the failure rate is a huge gain for the system.
            Said differently, the better their process of manufacturing, the better the profitability, so yes, I think Musk is right here, in the long run.

          2. “the only question remaining was about how much fuel (payload) you want to waste to return each stage. ”

            This is the exact same issue I face every day working with firefighting aircraft, and people just don’t get it.

            For any aircraft fuel is payload. Take more fuel and you less capacity available for your cargo, which in our case is water. The further we fly, the more fuel we need, so the less water we can carry. This is critical, because dumping that water is our sole purpose. Dumping more water in less time can literally be the difference between people living and dying.

            If we loaded enough fuel onto one of our water bombers to fly to the fire *and* fly back to an airstrip, we would be reducing the maximum amount of water we could dump on the fire by as much as 30%, depending on the location.

            What we do is load enough fuel to get to the fire, dump the maximum amount of water and then bail out and ditch the aircraft. This also eliminates the expenses of maintaining an aging fleet of airplanes. Since we only fly brand new airplanes, we don’t have those drastic expenses of inspecting and refurbishing an airplane that jas been flown to it’s structural limits through what are practically the fires of Hell.

            Ditching the planes does sometimes start new fires, but since we get paid to fight them, it makes good economic sense.

          3. >”With a 3% error rate (or 33 flights per rocket)”

            That’s not how it works. If you take 100 rockets with a 5% failure rate, the probability of any one surviving the launch is (1 – 5%)^n for n launches or 0.95^n

            On average you will have 0.95^n rockets surviving for another go. The number of rocket flights you can manage out of this fleet is the integral of that from 0 to n. For a^n the integral is a^n/ln(a) + C where at n=0 we must have a zero result because we haven’t launched anything yet. Hence C becomes approximately 19, and we get the equation:

            Y = 0.95^X / ln(0.95) + 19

            What this tells us, if you have an infinite fleet of rockets, you’ll get about 19 launches per rocket before they fail. If your breakeven point is 12 launches, you are paying 12/19 = 63% per launch compared to single-use, or saving approximately 1/3 of your cost.

            That’s not a huge advantage, considering you lose 2/3 of the lifting capacity to return the booster. Thanks to having to separate early. In other words, you build a rocket that can lift 3 and it costs 3, but you can only use it to lift 1 and for that it costs 2.

            If you needed to lift just 1 you could build a much smaller rocket, and thanks to the rocket equation it would be exponentially easier to make because the smaller it is, the smaller it can be.

          4. The above might be difficult to visualize, because the entire fleet of rockets is normalized to one rocket, but it works the same.

            For example, you start with a number of rockets, then fly all of them 100 times over. If you have 100 rockets, you’d expect to make 10 000 launches, but at 5% failure rate the cumulative number of launches per rocket is already very close to 19, so you’ll actually get 100*19=1900 launches because each time your fleet will shrink and you have less rockets than you started with.

            That’s another way of analyzing the situation. If refurbishment cost you nothing, you could build 526 rockets and achieve the same 10 000 launches for just 5% of the cost. However, since refurbishment only saves you 1/12th of the price of the rocket (break-even at 12 lauches), re-using those 526 rockets actually costs you the same as 6312 single-use rockets and we arrive at the same conclusion.

        2. Wikipedia knows:

          “As of 6 August 2018, SpaceX had recovered 21 first-stage boosters from previous missions, of which six were recovered twice”
          ” In total, 14 boosters have been re-flown as of August 2018″

          At best they’ve used any one booster three times, while the target for commercial feasibility is 10+, so they haven’t actually proven it to be profitable at all. What really counts in this matter is the dog and pony show that excuses continued investments and NASA contracts that would stop if they cut the program short.

          Well, at least they still have plausible deniability at this point.

    1. The thing about re-usability is what the Shuttle taught: the re-certification and refurbishing process can take just as many man-hours and materials as simply making another rocket from scratch, or more.

      After it became clear – after numerous attempts of denying it – that the Space Shuttle basically incurred so much damage each launch that it had to be completely overhauled, the bottom fell out of the program and it was maintained only to save face and put up a middle finger to the Soviets.

      The same is probably true of the Falcon rockets, knowing how Elon Musk has a habit of playing fast and loose with statistics and ignoring obvious issues because the numbers he picks are telling (his investors) that the thing must be working as promised. Eventually though, people lose money while others get killed.

          1. Conventional rockets are pretty much unsuitable for re-use. It’s like trying to make a dragster that runs on nitromethane and does a quarter mile in three seconds, into a school bus.

            A conventional rocket is engineered for minimum mass and maximum thrust/impulse. The optimal rocket has the structural integrity to survive exactly one flight, because the relationship between the rocket mass and the payload fraction follows a non-linear curve – you have to shave tons off the rocket to lift one more kilo to orbit. so the whole thing is like a tower of twigs, made of aluminum alloys that don’t have a threshold against fatigue cracking in the first place. When it is launched the vibrations and forces basically shake the thing to the point of fatigue failure every time, and most likely every time something breaks just after it has done its job. It may look intact, but the whole structure is like a jumbo jet flown for 20 years of service – after one launch.

            You have to go through every weld and strut with an x-ray machine to make sure it’s still safe to fly. Miss one crack and you’ll hear a big one in the sky. That’s what makes it expensive, and the turnaround times long – yet people like Elon Musk with their half-completed engineering degrees come around going “I bet you I can do it in 24 hours”.

            Yes you can, but you’ll kill people and destroy property due to your negligence.

    2. It’s going to be a case-by-case basis forever. Some payloads are just totally bonkers to get into orbit with no disposable staging (and without using something completely insane and dangerous like Project Orion). We’re close enough to the theoretical limits of chemical rockets that we can be reasonably sure these issues won’t go away with further advances in technology, barring new propulsion methods like ground-based lasers or something else equally exotic. Or nukes, as I just mentioned. There’s always nukes. Space is the perfect place for nukes; but for a launch vehicle? Eeeeehhh… I don’t think anyone will ever get past that PR cataclysm. Maybe in 120 years when we are all scrambling to evacuate the charred husk of a planet we’ve ruined by that time, but not before.

      Launching a commercial LEO satellite and a few cubesats? Sure, it’s tough but as SpaceX has proven it’s feasible. Launching a space station module? Translunar vehicle? God forbid part of a Mars mission? Those bottom stages are going to be massive as skyscrapers and they will be operating on the most razor-thin margins.

      The tyranny of the rocket equation is a huge mean bitch and turns conventional ideas about re-use on their heads. But I’m super pumped that we’ve even made it work for small, routine payloads–that’s huge. And it’s really the only thing we ever do in space anymore, so maybe this is all academic anyway.

      1. Air breathing rockets that fly horizontally to get out of the first ten miles of atmosphere circumvent the rocket equation quite nicely.

        There’s just the problem where the Brits invent a suitable engine but don’t have the money to build it or facilities to test it, the French and the Germans refuse to fund it, the Russians are too busy robbing each other to take notice, China has already copied it but doesn’t understand how it works, and the Americans are ignoring it because it wasn’t invented there.

      2. Nuclear water rockets, as depicted in “Destination Moon” back in 1950. Water is very dense and of course non-volatile and non-toxic. Heat it to the point where it spontaneously breaks down into oxygen and hydrogen.

        The volume increase alone provides a huge amount of thrust. Then you get to set the exhaust on fire. :) Nuclear heated flaming steam rocket. Doesn’t matter if that adds much to the thrust, it’s more thrust which the engine nozzle would be designed to take best advantage of. Plus… it’s fire.

        The main technical obstacle is material for the heat exchanger that won’t melt at temperatures over 3,000C, along with something to embed the nuclear material in that also can withstand such temperatures for long periods of time.

          1. Why?

            Oxygen is heavy compared to hydrogen and it’s chemically aggressive. So do it the other way round: Use (liquid) hydrogen without any (per)oxide. Burning the H2 and O2 after the nuclear super-heating does not do any good, it does not even work. Don’t forget, you have a super hot gas or plasma which underwent thermal decomposition just now. So the chemical balance is not in favor of the reaction 2H2 + O2 -> 2H2O. And if the gas mix burns, the number of molecules decreases 3 to one, so your volume and pressure decreases.
            AFAIK any idea of a nuclear thermal rocket was just about using the super light hydrogen and leaving out the heavy (16 times) oxygen.

        1. It must not only resist melting at 3000C, it must be chemically stable enough to hydrogen, oxygen and water (highly supercritical) at this temperature. So graphite is already ruled out.

      3. There is conjecture today and theoretical work that one can make solid hydrogen – with enough pressure. The solid will be stable, much like diamond. Solid H is about 15% denser than liquid. Over all, you get small lighter fuel systems with no dewar or other insulation. I wonder what it could do? Oh, there is also a load of binding energy due to the compressing.

  2. >”wire deposition 3D printer”

    What they’re doing is basically welding the whole tank out of wire, and that presents obvious problems of thousands and thousands of inclusions, cracks, pockets, and the heat affected zone will require a full-on heat treatment to even out the material properties and relieve the stresses left behind from the thermal expansion and shrinking, which can then warp the whole thing when it’s done. To add insult to injury, you need a continuous supply of welding gas for the days of printing time it takes, which isn’t exactly cheap.


    I’ll believe it when they fly the first one.

    1. I am with you and the “I’ll believe it when it flys”. Look at CNN every week they are claiming some new air craft is the future. They have been doing this for almost a decade now. Yet not a single one of the air crafts they are claiming are the future and are soon to be delivered. Have even started to be built. Every single one is searching for capital, and that is pretty much it.

      Sadly I feel many NASA and DARPA aircraft are the same way at this point. I feels as if long gone are the days of actually producing something. Vaporware is the new product that seem to be put out for public consumption and usage. While yeah i am being a little harsh on NASA and DARPA. Cause often very ambitious projects reach a standstill in order to solve other problems. Its the fact that so many of these projects don’t seem to have any useful sub projects that are even beneficial anymore.

      1. I feel that way whenever I see another “LOOK! ANOTHER AMAZING AND TOTALLY NEW FLYING CAR!!!!!!” project in the media. Look, if the failure mode of your transportation device is “it falls out of the sky” and you want to combine that with unskilled operators and cost cutting efforts and low maintenance cycles, don’t be surprised at the results. Have you seen what people do (or don’t) do to their cars? People have been believing that flying cars are just around the corner for decades. Sorry, but until we figure out how to levitate things like a science fiction shuttle craft, that’s not a viable transportation mechanism for unskilled operators.

        1. Ah. We now have the computing power to fly them autonomously.
          Flying in the air might be considerably simpler than driving on street level with other non-autonomous vehicles.
          I am thinking it will also work a lot better if we use supercomputers to direct air traffic.
          All the additional air traffic would be overwhelming to human air traffic operators.
          Therefore, we should have a low altitude autonomous air traffic operators.

          1. Haven’t you heard of “To err is human, but it takes a computer to royally f— up.”

            Adding another buzzword to the bull salad isn’t going to make it happen. First you’ve got to deal with the “AI revolution” that is turning out anything but intelligent. Remote operating flying vehicles by a supercomputer has the same problem of playing a video game via a dumb terminal – too much lag. Horrible things happen in a split second.

          2. If I use it, I want it mostly self contained, not “cloud connected”. To much data is generated, which can – and WILL – be misused and it only needs somebody with a radio jammer to take me out of the air.

    2. I assume that for something like this it will be done in a sealed chamber filled with an inert gas–it wouldn’t make sense to just spray away the argon all damn day and let it float out the window. Even more exciting–someday it could happen in the vacuum of space, don’t need no stinkin’ shielding gas there! Or heat for that matter. Stuff welds together spontaneously in the absence of an oxide layer. But that’s another story.

      I’ve seen some attempts at automated x-ray testing of parts made using these techniques. And working prototypes of engines have been tested, so it’s not like it’s completely infeasible. IIRC they actually performed pretty well.

      There’s ways to do this. It’s going to be strictly controlled robotic welding, so it won’t be anything like the welds that clumsy humans make, even the most experienced of old hands. And it’ll all be designed to occur in the ideal environment, ideal circumstances, ideal angle and speed, etc–it’ll be a top-down weld in probably a closed humidity- and temperature- controlled environment filled with argon or whatnot. I expect it produces a fairly homogeneous product when the process is designed with adequate funding and a good engineer at the helm.

      1. >”It wouldn’t make sense to just spray away the argon all damn day and let it float out the window.”

        As opposed to filling up an entire sealed building large enough to house the object being printed?

        >”working prototypes of engines have been tested”

        Parts of engines, and tiny copies of larger ones.

        >” Stuff welds together spontaneously in the absence of an oxide layer. But that’s another story.”

        Only if the surfaces can match to atomic precision. I.e. just pushing a random loop of wire out of a spool won’t make it stick to itself, and even if it did, that would still be downright useless because it only catches over a tiny area, the wire being round.

    3. Of course your points are correct. Only the supply of welding gas would be no problem on Mars: Just do it in the open atmosphere :-) But also on Earth CO2 is not difficult to get, coal fired power plants produce more than enough of it. Some purification may be necessary, but in case of stick welding it’s also just the CO2 produced by the burning of components in the coating of the stick that protects the weld from oxygen.

  3. So TL;DR they’re just using MIG welding and then machining it as they go. That’s ugly.

    Also it’s the same thing that “3D printed bridge” in Holland did. It’s a really awful way of making something unless you have some particularly good reason – I don’t really know how it stacks up as being a better solution if you’re on Mars other than you can break the robots & consumables down for transport.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.