One Press or Ten Thousand

In Fremont, California, there is a machine the size of a house. It is sixty feet long, twenty-four feet wide, and it weighs close to a million pounds. Its job is to squeeze two halves of a steel die together with about six thousand tons of force while molten aluminum is shot in between them. When the die opens, out comes the entire rear underbody of a car, one piece, where the old way took seventy stampings and a hall full of welding robots.¹²

That is Elon Musk's bet. Build the biggest press anyone has ever built, and make a car the way a toy car is made. He has said so in almost those words.³ It is a good bet, and half the car industry is now copying it.⁴

Star Bound's bet, laid out in A Thousand Garages, runs the other way. Instead of one machine the size of a house, ten thousand machines the size of a microwave, in ten thousand garages, each turning out the same ten-inch triangle. Where he centralizes, we scatter.

I set out to write this article as one bet against the other. By the end of it I had changed my mind, and I am going to show you where, because the place I changed it is the most useful thing in here. But first the part that has to come before any of it: the panel has never been tested. There are three drawings and two print files sitting on this site, and not one of those panels has ever been broken on purpose to find out what it takes. Everything else in this magazine is an argument. This is the one thing that could be a measurement.

The plastic panel proves one thing only

If you download the file and print it, you get a panel that looks right and slides together right. It proves the geometry: that the tongue finds the groove, that two panels lock in the plane, that the third one laid on top blocks the first from backing out. That matters. A joint that does not fit is a joint you never get to test for strength. And it is the test anybody with a six-hundred-dollar printer can run this weekend.

But a plastic panel tells you nothing about what the structure will hold. The real thing is stainless steel, and the honest thing to say about a printed and sintered stainless panel is that we do not know what it holds either. The parent article said a sintered part comes out with the properties of ordinary three-sixteen-L. The follow-up, Cold, Vacuum, and a Hundred Thousand Joints, corrected that: a binder-jetted three-sixteen-L part comes out of the furnace around three percent porous, and the pores cluster in layers where the print was laid down,⁵ which means the part has a grain to it, like wood. One study found parts loaded along their print layers nearly fifty percent stronger than parts loaded across them.⁶

None of that changes the plan. The garage makes a green part, the yard makes the strength, and if the steel is half as strong you put more of it in the tongue. But "more of it" is a number, and nobody has that number. Which is why the test has two stages.

Stage one, plastic: does it fit? Print fifty. Slide them together. Stack a second course. Measure the gaps, note where they bind, note where they rattle.

Stage two, steel: what does the joint hold? This is the one that counts, and it is the one nobody has run.

Which way do you push?

A panel in a spinning ring is loaded three ways, and each one is a different test.

You can crush the panel, straight down on its face. That tells you about the panel, and the panel is just a printed part. If the steel is good, the panel is good. Not very interesting.

You can pull the joint apart in the plane: grip two locked panels and drag them away from each other along the direction they slid together. In the ring, that is the hoop load. The spin is trying to fling every panel outward, and every panel is hanging onto its neighbors to stop it. This is the load that matters most, and it is the one that No bolts, no welds said the joint would carry.

You can peel one course off the course below: lift a top panel straight up off the one it is locked into. In the ring, that is what the air inside and everything bolted to the inside wall is doing, pushing outward through the layers. Sixteen Inches of Steel put a number on how much.

The joint is the test. Not the panel. So the fixture is a frame that holds two or three locked panels and pulls or lifts on them in one direction until something lets go, with a gauge reading the force. A hydraulic shop press with a load cell will do it. A university materials lab will do it better and give you a proper trace of force against stretch.

Fifty panels, not one

Here is the part most people skip, and it is the whole point.

If you break one panel and it takes eight hundred pounds, you have learned almost nothing. If you break fifty and they all let go between seven-eighty and eight-twenty, you have learned something enormous: the process is repeatable. Every panel that comes out of that printer and that furnace is going to behave like every other one, and you can design the ring around eight hundred pounds with confidence.

If you break fifty and they let go anywhere between four hundred and twelve hundred, you have learned something just as important, and much worse. The average is still eight hundred. But you cannot trust any single panel to hold eight hundred, and there are a hundred thousand of them in the ring. You have to design to the weakest one you might get, which means designing to four hundred, which means you just threw away half your strength margin on not knowing.

There is a whole branch of engineering statistics built on this, and the short version is that for a structure made of a great many identical parts, the scatter in strength matters more than the average, because the structure fails at its weakest part, not its typical one.

Scatter usually has a cause, and the causes are the boring ones. Wet filament. A bed that was five degrees off. A nozzle worn oval. A furnace that ran a little hot on Tuesday. A green part handled roughly on its way to the box. The fifty-panel test finds these things, because the panels that fail low will have something in common if you go looking. That is what a qualification program is. It is not proving the part is strong. It is proving the process is boring.

Could you press them instead?

While I was working on this, the obvious question came up: if one machine can press a whole car underbody in a shot, why not press panels? Not in a garage, in a plant. How many panels would a machine like Musk's turn out per stroke?

The answer surprised me, so I want to walk through it.

There are two ways to make a part in a press. The old way is powder metallurgy: pour dry steel powder into a die, squeeze it at enormous pressure until it holds together, then sinter it. It has been making gears and bushings for a century, and it skips the plastic binder entirely. The catch is the pressure. Steel powder is compacted at up to fifty tons per square inch.⁷⁸ A ten-inch triangle has about forty-three square inches of face. That is thirteen hundred to twenty-two hundred tons for one panel.⁹ A six-thousand-ton machine the size of a house gets you three to five panels a stroke. The nine-thousand-ton one built for the Cybertruck, maybe four to seven. A hundred thousand panels is twenty thousand strokes of the biggest press on Earth.

That is the thing worth noticing. Pressing powder does not reward bigness. The pressure is per square inch of part, and it does not care how big your machine is. A giant press just presses a few more parts at once.

The other way is what Musk actually does: liquid metal. Die casting does not press the metal at all. You inject it molten, and the clamp force only holds the die shut against the injection pressure, which is hundreds of pounds per square inch, not tons. So a six-thousand-ton clamp could hold a die with dozens of panel cavities in it, and the press would fill all of them every few seconds. That is the machine that makes a hundred thousand of anything.

Except it cannot make them out of stainless. Molten stainless is up around twenty-six hundred degrees, and no die survives being hit with it a few thousand times. Steel gets sand cast or investment cast, one mold per part, which throws away the whole speed argument. Aluminum, on the other hand, die-casts beautifully. It is the reason the Giga Press exists.

So here is the fork. Aluminum gets you the giant press. Stainless gets you the garages and the furnace. And that would be the end of it, one bet against the other, if the habitat were one wall.

Four walls, not one

It is not one wall. I should have said this two articles ago, because anyone who has worked around a tank farm already knows it.

A tanker does not have a hull. It has a double hull, because the whole point is that when something punctures the outer skin, you have lost a skin, not the cargo. A habitat a mile across, with a hundred thousand joints, spinning in a belt full of dead satellites, is not going to be one wall either. It is going to be three or four structural walls with gaps between them, insulation in the gaps, and each wall doing a different job.

Once you see it that way, the press question stops being a fight and becomes an assignment.

The outer wall is the bumper. Its job is to get hit, and to be cheap enough that getting hit does not matter. That is an aluminum wall, and it is exactly the wall you cast by the thousand in one enormous machine. Musk's bet pays off here.

The structural walls inside it carry the spin and the air. Those are the stainless walls, the ones that cycle under load for fifty years, the ones that need the strength and the fatigue life that aluminum does not have. Those are printed, green, in ten thousand garages, and sintered at the yard. Our bet pays off here.

Behind the metal, a third wall never sees ultraviolet, radiation, or debris, because the outer walls have already eaten all of it. That wall can be plastic. Plastic is a poor material outside and a fine one inside, and its main weakness simply does not apply there.

And there may be a fourth. Carbon fiber is the one material genuinely stronger than steel per pound in tension, and a spinning ring is pure tension. But I would not put it in a wall that cycles. Carbon fiber is fine on the first cycle and quietly worse on every one after, and you cannot see it happening. The Titan submersible proved that with five people in it. Steel handles pressure, relax, pressure, relax. Carbon fiber, as far as anyone has shown, does not. If newer materials change that, this magazine will say so.

How thick each wall is, how wide the gaps are, and what goes in them is its own article, and it is next. But the material fork closes here: different walls, different metals, different machines. The house-sized press and the ten thousand garages are not competing for the same job.

What we are asking

The two print files are still on the site, the four-inch test piece and the full ten-inch panel. Print them. If you have a press, or know someone who does, lock two together and pull until one lets go. Write down the number and the way it failed. Send it to hello@starboundmagazine.com with a picture of the break.

When we have fifty numbers, we will publish the plot. It will be the first honest data anyone has on whether a printed habitat panel is a real idea or a good drawing. If it is a good drawing and nothing more, I would rather know now than at a hundred thousand joints.

Ideas in Star Bound are published open — take them, build them better. Our open proposal.

Comments

Star Bound is not a forum, so comments run the old way: if you have comments, please write to us at hello@starboundmagazine.com and they will be published here in the next update, with your name on them. Say if you would rather stay anonymous.

No comments posted on this one yet. Yours could be the first.

Share this article

Copy link Facebook X Email

A note on where this comes from. I am not a metallurgist, a spacecraft engineer, or a degreed engineer of any kind. But I am a practical one, with more than forty years in heavy industrial construction: welding, millwright work, hydraulics, mining, and thirty-three years of electrical construction and commissioning on the North Slope of Alaska, in a place that punishes anything built wrong. Much of that time I have worked as a field engineer, which means I am the one standing in front of the equipment when the drawing turns out to be wrong, and the fix has to be made on the spot, with what is on hand, and has to work the first time. Anyone who has done that long enough learns that a lot of engineering gets corrected in the field by people with more experience than the people who drew it. That is the eye I bring to these subjects: not what should work on paper, but what actually gets built, and what breaks. The facts I take from other people are footnoted so you can check them and read further. The ideas built on top of those facts are my own, and I would rather see them knocked down by somebody who knows better than left standing because nobody looked.

Sources

1. Clamping force of 5,600 to 6,200 tons; assemblies of as many as 70 parts replaced by two or three castings. InsideEVs, Learn About Tesla's Giga Press From IDRA, January 2022. ↩
2. IDRA OL 6100: 60 ft × 24 ft × 18 ft, 950,000 lb, 6,000 metric tons rated clamping force. HotCars, How Tesla's Gamechanging Giga Press Shows Other Automakers The Way, December 2022. ↩
3. Tesla on the machines enabling full-size cars to be made the way toy cars are made. Silicon UK, Tesla Shows Giant 'Giga Press' In Action, September 2022. ↩
4. Toyota, Geely, GM, Volvo and Hyundai adopting giga casting. Autobody News, The Giga Press: Tesla's Game-Changing Manufacturing Process Goes Mainstream, January 2024. ↩
5. Binder-jetted 316L: about 3 percent initial porosity, pores of roughly 20 µm clustered in layers perpendicular to the build direction. Roth et al., Rate-dependent strength and ductility of binder jetting 3D-printed stainless steel 316L, International Journal of Mechanical Sciences, 2021. ↩
6. Layering parallel to the load about 48 percent stronger than layering perpendicular to it. Applying Binder Jetting Technology to 316L Stainless Steel Materials and Testing Its Mechanical and Dimensional Properties Depending on the Printing Method, Materials, 2024. ↩
7. Conventional powder metallurgy: powder compacted at room temperature in a rigid die at up to 50 tons per square inch. Metal Powder Industries Federation, Industry Facts. ↩
8. Iron-based powder compositions compacted at about 20 to 50 tsi (275 to 700 MPa). Hoeganaes Corporation, U.S. Patent 5,298,055. ↩
9. Editor's arithmetic. Face area of a 10-inch equilateral triangle: 0.433 × 10² = 43.3 sq in. At 30 to 50 tsi: 1,300 to 2,165 tons per panel. Panels per stroke at 6,000 tons: 2.8 to 4.6; at 9,000 tons: 4.2 to 6.9. The eight-hundred-pound figures in the scatter section are illustrative, not measured. ↩

The press-per-stroke arithmetic (note 9) is my own hand calculation from the published compaction range and is meant to give scale, not to size anything. The two-stage test, the three loading directions, the fifty-panel scatter argument, and the four-wall assignment of materials to processes are my own proposals. The double-hull principle is borrowed from every tanker afloat. The objection to carbon fiber in a cycling wall is mine, on the evidence of the Titan; a materials engineer who can show a carbon-fiber laminate surviving fifty years of pressure cycles is invited to knock it down.