Podcast: Sheet Metal Forming Goes High-Tech with Robotics & AI

author avatar

19 Apr, 2023

Image credit: Machina Labs

Image credit: Machina Labs

In this episode, we talk about how robotics and artificial intelligence are being used in the sheet metal forming industry to improve safety, efficiency and precision.

In this episode, we talk about how robotics and artificial intelligence are being used in the sheet metal forming industry to improve safety, efficiency and precision. The use of these technologies is allowing manufacturers to produce high-quality products at a faster rate, while also reducing costs and waste. 


(1:00) - Robotics & AI in Sheet Metal Forming

What's up everyone? In today's episode, we're talking all about how robots and AI can improve sheet metal forming. If you've ever designed a sheet metal part or worked with someone who has this might scratch some pain points for you like it did for me. I think it's really interesting. And we're talking about one of my favorite startups. So, I'm super excited about it. Let's jump in.

I'm Daniel and I'm Farbod, and this is the NextByte podcast. Every week, we explore interesting and impactful tech and engineering content from Wevolver.com and deliver it to you in bite sized episodes that are easy to understand, regardless of your background.


Daniel: What's up everyone, like we said, today, we're talking about really, really interesting startup. It's actually one of my favorite startups in the startup ecosystem was talking about and in the hard tech space. They're called Machina, and they're going to, we're going to explain how their robots and AI are helping improve sheet metal forming

Farbod: Wait. Um, really quick, sorry, before we get into what Machina is, I want to give the folks some context, the genesis of this podcast was Dan and I would have conversations about companies that we thought were exciting. Now, Dan has not shut up about Machina for the past, like three months, probably. He's been very passionate about this. We've been talking a lot about it. He's really excited. Just wanted to let everyone know that, that this is a passionate episode.

Daniel: Yeah, it's all passion here. So, what I mean, let's stay zoomed out where you are, right, we've given everyone a little bit of background around my excitement for this startup. And this this space that they're working in. Let me explain a little bit of why I'm so excited, so, especially since starting to work in the automotive industry, I've started to really understand how many inefficiencies and limitations there are in traditional sheet metal forming processes. So traditionally, you know, I'm thinking in automotive like making a car or something, you know, big stamping, you need that's called like draw forming, you need a positive dye, a negative dye, you need high pressure dies to stamp this metal. And a lot of times, you've got humans helping load the metal, the metal blanks into the stamping machines to stamp these big parts, you can already start to tell the amount of machinery that's required, the amount of human intervention that's required, this is a huge investment, if you're going to make any new sheet metal parts, especially if they're large in size, it's tons and tons of investment, it's not uncommon to find that you when you're making a huge sheet metal part, you need to spend, you know, anywhere in the order of hundreds of thousands to millions of dollars, making the tool to be able to form the sheet metal parts. And the really, really frustrating part about that is it's not like injection molding, where you can usually like 3d print a part, and maybe it doesn't have the exact material properties that you want. But it's close enough in terms of getting a representative design you can look at with sheet metal, it's really, really hard to prototype too. So, you've got to commit to spending hundreds of thousands, if not millions of dollars on a giant part, you've got very few reliable ways to prototype them in the meantime, which means you better make sure your designs 100% correct. And you don't have a great way to verify that. And more so than that if you've got to make changes to the design. They're also really, really time intensive and expensive. So, there's, to me, there's a lot of passion in that because I feel those pain points, trying to work with sheet metal. And I know a lot of engineers that are trying to design large sheet metal parts that encounter those pain points on a day-to-day basis.

Farbod: I mean, I'm coming at it with like an outsider's knowledge of what I guess not a super outsider. But like based on what we learned in engineering classes, right? Sheet metal forming is very common. auto industry is a big one. We talked a lot about it. Now, back in material science. It's like big sheets of metal, like you're talking about, you're using these dies, they're usually custom for like a specific part, they cost a lot of money to get. So, like, if you're a hardware startup, like let's say you're starting your own boutique car company, this is a big expense for you. And you don't have like you said a lot of ways to iterate and get it right on top of that. You have the machines that are actually using this thing. And it's not very safe, right like by the nature of what it's doing. You're applying a lot force to like stamp something out. Injuries can happen, or like just manipulating and moving that stuff around. It's not easy. So, if I'm not mistaken, this process has been around for like almost a century and it's just not been changed whatsoever. Right? Like a missionary there. It's coming to play but like for the most part, it's remained the same

Daniel: That I would say we've, we've had an increase in the amount of automation, we've had an increase in the amount of machines and specialized equipment that's used. But the general processes that are used to manufacture sheet metal, have generally stayed the same for a while. And there's kind of an overview of what these processes are, there's bending, you can do this with a press break. Or you can do this with tools. But it's like kind of bending things along straight lines like origami, they're stretch forming, where you stretch out the sheet of material, there's a roll forming where you continuously curve and bend the material over a series of rollers. They're stamping, which is what we talked about using a blank and a die hydroforming, which uses high pressure fluids against a one-sided tool to kind of form sheet metal, I think that's interesting. Then there's superplastic, forming where you heat a sheet of metal. And then you press it against a one-sided tool using gases and their spin forming. This is the one that's like usually really manual, and it looks really cool, we've actually got a cool photo of it in the article, it's linked in the show notes, you should check it out, you can have a giant sheet of metal spinning on what looks like a lathe that's like the size of three or four people. And people are using tools to try and like bend this piece of metal. To me, it kind of looks like someone like molding clay on a potter's wheel. But it's you know, the scale of four or five people pretty, pretty interesting. But again, none of these lend themselves particularly well to high product quality, reduced cost, increased production speed, rapid prototyping, and short development cycles. These are all things that are very capital intensive, very time intensive. And, you know, part of the barrier to entry and things like aerospace and automotive, where a lot of these big sheet metal pieces are used, is the fact that you need to spend so much investment just to get off the ground just to be able to produce your parts at that scale in the representative material. And then representative shape and dimensionality. So, there are ways of prototyping kind of, there's like crash form tools and stuff like that, which use a very, very cheap version of the tool for stamping on a very, very thin set of material to just kind of try and confirm the geometry. But that's not production intent. So if you're trying to do prototyping in a manner that allows you to do production, intent, validation, or even use something in a production environment, right, create sheet metal parts that you're going to, you know, go put into your product and sell to someone, there's no real “Agile” is the way that I like to like to think of it, there's no real agile way to get into it, which is where Machina enters with robotics and AI to help with sheet metal forming and prototyping.

Farbod: Want to get into the sauce?

Daniel: Yeah, I would say there's like three main ingredients in the sauce. One of them is AI algorithms, one is robotics, and then one of them is sensors. And Machina is using these three separate ingredients to help drive enhanced frizz precision and quality and sheet metal products as well as these addressing these pain points, we're talking about like high manufacturing costs, high lead times. And one other twist that I think is pretty interesting that machina talks a lot about is the fact that there are so many humans involved in the sheet metal forming industry today, the amount of skilled workers wanting to enter that field have been decreasing steadily over the last 10 to 20 years. So not only is there a challenge where everything costs a lot, and everything takes a long time, there's also heavily relying on humans. So, this sounds ripe for automation that sounds ripe for a new process, which is what Machina is doing,

Farbod: Dude, so, I was curious about this because I don't have again, a lot of insight into like machining, the manufacturing realm, all that jazz. So, I did some digging. And according to the National Association of Manufacturers, there was a study done in collaboration with Deloitte and the Manufacturing Institute, something like over 2 million manufacturing jobs could be left unfilled by 2030 in the US, because we just don't have enough people going into this into this career anymore. And apparently the pandemic wiped out a lot of those skilled workers like they just didn't come back to work. And there's no influx of new workers. So, we've always like harped on the fact that automation can be good. If it's collaborating with people, and it's not taking away from like, roles that people should be having. It's not negatively impacting society as a whole. And this looks like an opportunity that's ripe for automation, right? There’s real need and the benefits, it almost seems like a win-win across the board. Right? Real quick. I think I want to get into the flow of how Machina does a request for manufacturing. Is there anything else you want to cover before that?

Daniel: Now? Let's jump into it.

Farbod: Okay, so I was reading this article, and I mean, you primed me quite a bit on it before I did. but it's just as amazing as you made it out to be. So, you have this new request that comes in, let's say, Daniels, the client and I'm Machina Labs, you come in, you're like, hey, I want this part. And apparently, within hours or maybe even a couple of days, I can get you your first port. So, you have these two robots working in collaboration with each other on two different sides of the same piece of sheet metal, by the way, the material that you're using, it's, the process is material agnostic, they, I'm going to quote them their website, it said, “if it bends, we can work with it.” So that's pretty promising. You have these two robots that have a CAD file that's gone through the Machina Software System loaded to them, they start forming and doing whatever processes is required. And at the end of it, they start scanning the part. And what the scan does is that it goes back to that Machina software and analyzes the actual part with the CAD design. And if there's anything that is not perfect, if there's anything that's standing out, they’re optimize themselves before they start attacking the next piece of work until they get it right. So, you have this what I was most interested in, you know, given my role, I'm like a test engineer, basically. So, I'm always shooting for like highest quality whenever possible, they've built in quality control into the manufacturing process. And it happens so much more seamlessly, than if you were to have someone physically like pick up the part and start doing analysis bit by bit. So that's the part that like, really blew me away. And because you have robots working at this thing, you don't have to worry about loading and unloading of materials, you don't have to worry about like, you know, I'm an aerospace company, I need this part, like over the weekend to be done. You can't have humans do that. Or if you do, you have to pay an extra like, you know, service charge. Now, these robots run 24/7,365, they can run away from people as long as they have the material there, and you come back and you have the parts that you need.

Daniel: And the entire process, right, loading, forming, scanning, cutting, finishing unloading, then reloading the blank material for the next one, the entire thing's automated. Yep, we don't have humans in there, like I've been in factories where there are stamping going on. A lot of the times there's humans reloading the blanks, and then humans taking the parts out after they're stamped and loading them onto a pallet. None of that exists in this Machina solution. But honestly, before we even dive any further into it, I want to try and describe exactly what it is that's going on. When they're doing this, they call it Robo-forming is this new, new sheet metal forming method, let's say that it's proprietary to the work that they've been doing. This Robo forming, it's a little bit challenging to explain, but I'm going to do my best here. But let's take the analogy that you've got like a sheet of aluminum foil, and your hands stretched out, right, you're gonna hold the foil, you're gonna hold it tight, to make sure that doesn't move around or anything like that. And what I'm going to do is come at it with, like, let's say that the top of a marker, a marker cap on one hand, and then the top of the marker cap and, and other hand and I'm on each side of the aluminum foil, one of the backside one of the front side, I'm going to press these markers down. And I'm going to use the pressure between those to bend the metal. And I could do something like let's say, I want to put t and b, etch that into the aluminum foil that you're holding for the next bite, right, the robots apply pressure from either side, and they apply pressure, twist it and turn it to form the metal into the desired shape. That's exactly what they're doing at the huge scale here with large sheets of metal that can be used on airplanes that can be used in automotive, and they're doing the same exact thing. But with two seven axis robotic arms. So highly precise with their movement. They've got proprietary end effectors. So those that's the very, very end of that seven-axis robot that's actually coming into contact with the metal, they've got proprietary end effectors that they use to do different types of forming. Basically, what you've got here is the super cyber punk version of what we talked about with two markers and the sheet of aluminum foil. They're doing this to help create giant sheet metal parts in a repeatable manner. That honestly, you should just check out the link that's in the show notes and watch the video because it's super interesting to watch them. Turn a sheet of flat metal into these parts that they can do repeatedly load, form, scan, cut, finish, unload, again and again and again. And the incredible part about this is it doesn't require weeks and weeks of lead time. It doesn't require millions of dollars of investment to make these dies that are able to stamp it doesn't require huge hydraulic presses. All it requires is these two robotic arms, some frame to hold the metal and like you said any metal that bends they can do this forming process with which is really cool.

Farbod: Usually like our main call to action is go and read the article, right? If you're interested gone read the article, I want to change it this time and say you don't even need to read the article. But watch the video, please watch the video. The video does so much of the explanation. And honestly, it's freaking cool man. Like just seeing that thing in action making what it makes the precision. You mentioned the precision. At first, I was like, alright, well, they're talking about how this is mainly suited for big parts. And I'm assuming because like, that's just what makes sense, like financially for the volume that you're doing whatever you're not gonna do for like an iPhone or something were small. And then I was like, it's probably because they don't have the precision required to manipulate parts so small. No, no, they have submillimeter precision on the parts of their manufacturing, which is incredibly impressive. And on top of that,

Daniel: I mean, just on that note quickly, right, spoiler alert, the state-of-the-art stamping processes don't have submillimeter level precision. until weeks and weeks and weeks of tooling. And tuning have happened after they do trial and error. To understand how thick is this metal? How hot should our tools be? Are the tools the right shape? Or the processes, right? Are we pressing long enough? Are we pressing hard enough? It takes weeks of tuning after you've spent months tooling up these multimillion dollar tools to be able to get these reliable parts produced with submillimeter level precision. Machina can do it on the first try.

Farbod: Which is mind blowing dude. And again, you have this closed, closed loop feedback system, where you're manufacturing, learning and iterating. All within like, I mean, depending on how complex your part is, like hours or days, which is wild to me. And there's no human involved, it's just a machine doing the thing over and over again and comparing it with a designer's CAD model. And for what it's worth, the sheets that they can manipulate are up to 12 feet in length and four feet in depth, but not width like actual depth of how far they can stretch it out. And quarter inch thick. So up to a quarter inch thick sheet metal they can manipulate. I don't know what the industry standard is. But that seems like fairly thick for sheet metal.

Daniel: Yeah, very thick. Okay. So, I mean, it's, it's incredible. All this that they're able to achieve, I would love to, I think we've kind of hit on what it is how they're achieving it, right, what's in their secret sauce, they've got these AI algorithms that look at the material, they're able to make sure that they're handling it properly. using machine learning that that informs how much pressure they need to form the geometry that they want. They do process optimization like you're mentioning, right, they before they even start forming the metal, they've got AI algorithms, helping them understand the exact optimized path to be able to perform this and to make sure that it's got the right dimensional stability. They've got robotics involved. That's the secret part of the secret sauce that does precision handling, high precision robotic arms, and then the proprietary end effectors, they do this flexible forming process that uses, you know, maybe one type of end effector for these detailed areas, and then another type for rolling and then another type for smoothing it out. And then again, the whole thing is robotically automated. And then the really, really interesting part, there's the sensors, which tie it all together, they get real time monitoring using the smart sensors to make sure that they're applying the appropriate amount of force. And then again, they're getting feedback immediately, during the process back in for closed loop control around is this giving us the exact dimensions that we want, comparing it against the CAD and then uses that to inform future iterations to make sure that maybe even if the first one was 100%, dimensionally stable, you know, match matches every single critical dimension, maybe they need to do some process adjustments to make sure that they can do it faster next time. Yep. And the AI is tying in with the sensors to help make sure that it's doing it in the most efficient manner possible and making sure that it's doing it in a way that completely matches the design intent of the designer that sent the parts to them to be produced. And that, honestly, is one of the most interesting parts about everything. Because in traditional sheet forming processes, you don't have this closed loop feedback during manufacturing, right, you've got a giant hydraulic press coming in and stamping parts, you know, crashing two pieces of metal together with a sheet of metal in between. And then afterward after that's done after it's been trimmed and finished and everything, then you take that part over to a CMM machine or a scanner or something like that, to do some measurements to make sure that it's dimensionally correct. You have no level of feedback during the process. Machina has gotten every single step along the way to make sure that every detail is dimensionally correct, which is really interesting. For someone like me who has been a designer before and experienced the frustration out After you spend all this money in tooling with a supplier, after you wait all this time for the parts to arrive, and then someone tries to build with it, and they say, oh, sorry, this hole is in the wrong location, or hey, sorry, the bend is in the right radius. That's super frustrating, because all you've wanted this whole time is to get this part that builds into this assembly, do you know to make a cool product that you want to put out into the world?

Farbod: Yep. And by the way, I feel like, we've assumed the audience knows this by us just talking about it. But these robots can easily be reprogrammed to do a different job that has a completely different material, different geometry and all that pretty much on the fly. So, you can repurpose the same robotic duo to do job after job after job after job. You have these adaptable heads, you have this reprogramming, re optimization, all of that happens. One thing that came to my mind, I think we hinted on it earlier, too. But like, what are the we're going to be the outcomes of this, like you're obviously have the safety stuff that happens you have the efficiency, but the affordability aspect, like I think about all these hardware startups that don't have the capital to like, get something set up in the traditional sheet forming tooling. But with this, it could be a solution that enables a lot of those endeavors to happen. I remember what 14-year-old Farbod was watching Top Gear. And Jeremy is showing like this new Jaguar F type, and he's holding the entire side panel on one finger, because it's a sheet of aluminum that's been formed to like the perfect geometry. And I'm like, That's sick. And then you realize that to like, get something like that off the ground you need like millions and millions and millions of dollars before your first design happens. Imagine if like, you know, one day you and I just start opening up our own car company, we could potentially get sheet metal made on the fly pretty quickly get our design validated, and get it all end down within a week. Like that's so exciting, right?

Daniel: Yeah, I agree. And I really want to harp on something you talked about here, they've spoken about specific applications, right, this is a huge value add for people doing things that they're not ready to be super capital intensive, they're not ready to be super high volume, or they're not ready to wait 52 weeks for their tooling to be complete before they ever get a part off their production tool. What they've got here is providing this, this segment of the industry with an immediate solution that produces parts faster, produces them quicker with higher quality and at less cost than any solution in the market today. But they also talk about how this can interface how this can play with the traditional manufacturing methods and current manufacturing tech. So, they use you can use Machina kind of as like a middleman between craftsmen, which are like really skilled laborers making things by hand one by one by one. And then current manufacturing technology, which after the tuning after the millions of dollars of investment, it's really precise, and it's really scalable, you can make hundreds of thousands of parts really, really easily. Machina bridges the gap between those two, so it's not trying to replace craftsmen, it's not trying to replace current manufacturing tech, it's the fact that there's a huge chasm between those two. And people that are trying to work in that space in between don't have a great solution. And Machina can solve that problem can bridge the gap between craftsman and current manufacturing tech, using their AI using Material Science innovations using the robotic control and then using their sensors, I think you can tell I'm really excited about it. And I hope I'm not coming off as like, just extremely biased around the technology. Because we try to give an unbiased look at like this technology could be cool. This is where I'm kind of talking about the limitations, right? It's not going to replace craftsmen that create one off parts for a bespoke car in their garage, right? It's not going to replace these huge stamping factors that create hundreds of thousands of parts every month. But there's a big gap between those two. That's where I see Machina ticking at a huge hold and making it sincere difference for engineers starting as soon as today. Right? They've already got contracts with the Air Force and with other with other people using their technology today to create parts.

Farbod: So, my take away from this and chime in if you don't feel the same way is what these folks have proposed with their robotic system feels to me like what scaled or I guess more common additive manufacturing has been to injection molding, right? Yeah, you can prototype much quicker get feedback on what your design looks like iterate before you're ready. And if you want to scale up and you want to go big, you can finally do your injection molding with a design that you know you want to do it with. Or like we've seen some companies that pretty much build their entire product line which is limited, right? Like it's not the thousands but maybe in the hundreds to be just around additively manufactured parts. And yeah, that's how I'm seeing it. Does that make sense to you?

Daniel: Yeah, no, I'm with you. So, it's not only just for prototyping, I actually think this is less so for prototyping and even more so can add a lot of value for people in that low to medium volume production range. Because you're using production intent materials, you're getting your production intent geometry, getting the scenario to do it in a way that doesn't require tons of investment to get that same level of precision and quality. Yeah.

Farbod: So, I feel like we just kind of did it. But do you want to give a quick recap?

Daniel: Yeah, I'll wrap it up here. So, we talked a lot about the inefficiencies and limitations in sheet metal forming processes, there's not a great way to do it in a way that's high quality, low cost and high speed. Enter Machina, right, they use robots and AI to improve this sheet metal forming process. They've made the process faster, cheaper, and more accurate. They have robots that precisely shape and handle the metal that the whole thing is automated, while AI and sensors help them optimize the process along the way. This has led to better quality products and quicker development of new designs and is bridging the gap between the Craftsman which are creating one off parts. And the current manufacturing technology, which creates hundreds of thousands of parts. This gap in between there, Machina is helping revolutionize what sheet metal forming looks like. For engineers and for companies.

Farbod: That's money, man. Right on the money as always.

Daniel: Well, I hope everyone that's listening to this is even a fraction of excited about this technology as I am. As you can tell this is hitting some personal pain points for me. So that's probably why I'm so excited about the technology, but I don't know. I'm just happy. My final is gonna be big deal.

Farbod: I'm just happy with finally this episode. We've been talking about it for a while.

Daniel: So now I can shut up. Like Oh, yeah.

Farbod: All right. If there's nothing else, I'm going to wrap up. Yeah, I didn't check again. But I don't know if we're trending anywhere. I'm just gonna thank you guys for listening, like always. And, yeah, we'll catch you in the next one. Peace.


That's all for today The NextByte Podcast is produced by Wevolver, and to learn more about the topics with discussed today visit Wevolver.com. 

If you enjoyed this episode, please review and subscribe, via Apple podcasts Spotify or one of your favorite platforms. I'm Farbod and I'm Daniel. Thank you for listening and we'll see you in the next episode.

More by Daniel Mitchell

👋🏼 Hi, I’m Daniel Scott Mitchell. I'm an entrepreneur, engineer, and a lifelong learner. I've helped to build and launch products at Tesla, Rivian, Formlabs, and startups. I host the Next Byte Podcast with Farbod Moghaddam in collaboration with Wevolver. You can find me on the web at links.dan...