Hacker News new | past | comments | ask | show | jobs | submit login
3D-printing a stainless steel bridge full of sensors (mx3d.com)
101 points by pasta on Oct 22, 2018 | hide | past | favorite | 41 comments



This looks incredibly expensive with a lot of touch labor. The surface is incredibly gnarly due to being basically made entirely of welds. You'd want to grind that puppy smooth on the railings, etc.

One thing that I think is finally entering the mainstream is that 3D printing is actually a MUCH more labor-intensive process than typical mass production techniques.


The amount of time investment is really surprising coming into it, even as a hobby developer. The media portrayal seems to imply that you download a model, press a button, and out pops this perfectly printed part hot off the bed. The reality is more like

1. Calibrate the Z-axis again because that end-stop is glitchy 2. Apply some sort of adhesive to the print surface 3. Print with support (wait 5-20 hours) and hope nothing goes terribly wrong causing you to start over at step 1 4. Carefully remove the support material, hope the delicate bits of the part don't snap off in the process, if so start over at step 2 5. Sand, polish, and generally clean the part

That's just for my hobby printer making, let's not mince any words, rather low quality prototype parts and figurines. I can't imagine the complexity that goes into printing these huge structures with production quality requirements. It's impressive, and I'm sure they're making the most of the tech since it can accomplish certain things that would be quite difficult in other mediums (the embedded sensors seem like such an awesome idea) but the time investment seems like it would certainly be quite high.

Other than the coolness factor, I'm not sure it's more efficient than ordinary construction techniques, but I'm also just a hobbyist with a desk printer; I'm sure the folks actually working in that industry have much better insight.


>The media portrayal seems to imply that you download a model, press a button, and out pops this perfectly printed part hot off the bed

I happily do exactly this all day with my several year old Makerbot. I print in 3D just about as reliably as I print something with ink on paper.


A lot of that work could also be automated though if they wanted. There are machines out there today for smaller parts that will use wire deposition like this to build the bulk rough outline of the part then machine away the rough outer layer for a precision part.

Leaving it rough though I think is an intentional choice part of showing off the materials and processes used instead of hiding it. In reality I bet the surface isn't much worse than an unsmoothed stone or concrete.


> There are machines out there today for smaller parts that will use wire deposition like this to build the bulk rough outline of the part then machine away the rough outer layer for a precision part.

Right, so all you're really saving is the first rough cut, which is normally pretty fast anyway. (And the chips can be recycled.) But now you also have to live with the material properties of welds, instead of a high quality, grain-oriented piece of forging stock. And you also have to move the part from one machine to another, then re-register it (dial it in to make sure the build is oriented correctly in the CNC machine's coordinate system). Not as bad for the hybrid machines which can do both operations, but still not great.

Back to this bridge: Without post-processing by grinding/sanding/smoothing, you're reducing the material strength due to increased localized stress concentrations. This doesn't matter much in stiffness-dominated structures, but significantly reduces the performance for strength-dominated structures and structures which experience a lot of cyclical loading.

I just don't think this is terribly scalable. Neat art project, though.

EDIT:If you want to automate things, I think it probably makes more sense to automate using more conventional building methods... Cut standard beams to size, place and weld automagically, etc. It'd be much faster..... Like this: https://www.youtube.com/watch?v=rD9__jheNLg


You don't have the move machines actually, as of about 3 years ago there are machines that'll do both in a single machine. [0] On top of that though moving from one machine to another really isn't the end of the world especially if you're able to setup the process right, like putting all the 3D printed parts onto a standard sized base plate that you can easily index off of.

By bulk rough outline I mean they're basically just machining off the outer millimeter at most to smooth out the surface. It actually can save a lot of material as the size of the part increases because unless you're taking cast/forged pieces then milling them down to tighter tolerances the initial work piece will be the smallest bounding cube/rectangle or cylinder that'll fit the part which for some things is a LOT of waste.

And yeah they will be slightly weaker than the same piece made out of a forged piece but that can be mitigated by part design and some annealing either after the milling or between printing and machining.

[0] https://www.youtube.com/watch?v=oaIOrQi2HLM


Likewise, I'm curious about the energy consumption of fabricating structures of this size (and greater) via laser sintering, when compared to conventional techniques of casting i-beams and welding only at the joints. The power required to effectively melt each cubic millimeter inside the structure, and also to position the robot arm throughout, doesn't look like it would be streamlined much if done at scale.


Every day folks have no idea how much of welding actually consists of grinding. It's always fun when you encounter spatter/beads on a weld to ask the person you are with(assuming they are not and never have been a welder) what they think all those little beads are.... The most common answer I get is "it looks like balls of paint". There is a huge disconnect between how people think things are made, and reality.


The last time I bought welding wire it was 3-4 dollars/pound for cheap flux core stuff, so I can't imagine how much this cost to 'print'.

Most metal printers seem to work in a box full of helium/argon, but in this case they were just venting the shielding gas and probably using several times more than necessary.


> This looks incredibly expensive with a lot of touch labor.

This isn't as much as a bridge as it is a monument.


> One thing that I think is finally entering the mainstream is that 3D printing is actually a MUCH more labor-intensive process than typical mass production techniques.

I suppose you could 3D-print a mold, and from there do cheap production.


"Autodesk is supplying the cloud services that will power the bridge’s data collection and processing" So the first question is what happens when all these services eventually become obsolete over the bridges life time or if autodesk goes bankrupt?(Of course that may be unlikely if this proposed 'infrastructure as a service' business takes off) "These sensors will collect structural measurements such as strain, displacement and vibration," Now since the bridge has vibration sensors, could one not potentially do gait recognition on people walking across the bridge? And since all this data is stored in 'the cloud,' we now have a bridge which could potentially leak user data. Now to make a ridiculous slippery slope argument, the end result of this may be the house of cards construction building use in Vernor Vinge's Rainbow's End[0], where building s are literally constructed like a house of cards and computer controlled stabilization is used to prevent them from falling over.

[0]https://en.wikipedia.org/wiki/Rainbows_End


Hopefully by the time Autodesk goes under, we'll have good understanding of how it behaves. Also, no need to dramatize, we have bridges without sensors. I guess surveyors would have to include this one as well.


3D-printed structures behave very differently compared to your usual, continuous materials which are bolted, jointed or welded together, in that re-balancing loads and reactions. Another issue is how they cope with aggressive environments. This is a real-life experiment, something someone would however do, so let’s what happens. I’m confident they will be fine, we are even having 3D printed components for nuclear power plants already.


In this case, it behaves like one giant weld, because that's what it is. Fairly well understood, and this has large margins so will be fine.


Do you have experience with 3D printing? I have pretty extensive experience with 3D printing plastics and I can assure you that it's nothing like plastic weld. How is 3D printed steel like a contiguous weld?


They literally use a welding machine on a robot arm, that's how. Powder bed metal 3D printing is different, however.

(And I would argue that FDM is essentially plastic welding, too.)


And a weld executed properly with the same filler material acts more or less like a single piece of metal (minus any issues with the local heating of the pieces.


There must be a large vision for something down the road given the partnership list and that the application is effectively frivolous. They solved 12m pedestrian bridges a long time ago.


This looks like it's a massive PR stunt more than anything. The kind of cool project big companies want to have their name associated with. I was hoping to find more details about the technology used to build it, the compromises involved, etc.


I was wonder how much of our world would change if we had bidirectional material design. Things that would communicate in one way or another their internal status just a bit, to avoid finding faults too late.


I think external investigation will have the edge on self reporting sensor for quite a while, because bringing a xray/ultrasound machine place to place for a yearly inspection allow to split the machinery investment over a lot of structures.

> to avoid finding faults too late

even before fault discovery there's a lot of value in embedding sensors to tune simulators, so that the next generation product can improve in reliability, material costs or both while being cheaper to design.


Or use the data and feed it to the self learning program that will design the next bridge.


true but I appreciate objects that speak a bit instead or planning in advance and being in the blind afterwards


I would be curious to know the technological differences with https://www.relativityspace.com/stargate


MX3D describes their machine as "Standard Robot + Standard Welding Machine + MX3D’s proprietary software". The stargate machine appears superficially similar aside from using multiple 'print heads'. From their pictures and description it looks like one is a welder and one is a milling machine, no clue what the third one is.


Testing in production with good enough metrics is fine unless it's AN ACTUAL BRIDGE.


I agree, but the bridge design appears to be extremelly robust (i.e., extraordinarily oversized) and it will be subjected to foot traffic, which has a negligible service load.

It's a pretty sculpture that people can walk on. That's it.


>It's a pretty sculpture that people can walk on. That's it.

Famous last words.

https://www.miamiherald.com/news/local/community/miami-dade/...

Bridges are supposed to be boring and functional. Fluffing them up with all sorts of over engineering to make them pretty tends to introduce unnecessary variables that are less well understood or less easy to account for or may be "unknown unknowns" which have the potential to reduce the strength of the structure.

https://www.nytimes.com/2018/08/14/world/europe/italy-genoa-...

If you want to make a pretty art piece then make a fancy ornate bridge deck and slap it on top of your boring but functional steel I-beams that reliably carry rail, road and foot traffic worldwide.


Robust aka oversized does not counter brittle fracture, though, which is the main concern with 3D-printed artifacts. It depends on the material / the alloy but also on the technological process we generally call printing.


It actually does, since it reduces the local stress. Risk of fracture due to defects and repetitive loading is dramatically reduced (by order of magnitude or more) by increasing the material margins.

It's a short foot bridge, so it's pretty easy to design it with significant enough margin.

That said, it's not a very good idea nor a scalable solution to, like, anything. More like an art project.


Yes, you are basically saying that this particular artifact is just like a bridge-shaped pancake made of polystyrene and hosting very few ants at any given time. Which is actually an art project.


I think that's overstating it. This bridge is plenty strong, just very inefficient and expensive.


> Risk of fracture due to defects and repetitive loading is dramatically reduced

In this case I believe the design is due to concerns regarding the additive manufacturing process, which is prone to introduce a lot more defects than standard steel sections, and ina way that are hard to spot and to track.


Walked on it last weekend. No more than 10 ppl were allowed on it at the same time. Guess they are still gathering data


Its not uncommon to leave stress strain sensors on bridges to know the state of the bridge even as it ages. Testing doesnt end when it goes into production for actual bridges. The best metric for measuring the bridges safety is strain under actual loading.


https://vimeo.com/239238204

Video shows physical material tests and simulated structural tests. What else would you suggest?


Well the bridge is approved and will be placed in Amsterdam


"Approved and placed" is hardly the same thing as "safe".

Case in point: https://www.miamiherald.com/news/local/community/miami-dade/...



Kind of cool but also the type of wasteful project that gives "3D printing" a bad name.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: