Freeform
Freeform is an insane metal 3D printing factory(!) startup, has raised $60 million to date, and is on the prowl for killer MechE talent (read all the way through to see what roles)!
Here’s why they are going to win!
🌎 The State Of Metal 3D Printing
Heralded as a turnkey piece of technology that could theoretically democratize industrial manufacturing (think of the prevalence of FDM printers but now make them metal), metal 3D printing has had a turbulent couple of decades.
It is a piece of technology that has to date been reserved for very costly and complex systems where no other manufacturing methods are capable of producing the same parts (usually due to internal geometries or other manufacturing constraints).
**I was lucky enough to see the good and the bad when designing and sourcing metal 3D printed parts during an internship in 2020, but I’ve now revised several of those opinions. Let’s get into it!**
âś… The Good:
1. Combining multiple part geometries simplifies supply chain and can reduce weight. There are numerous examples of this, but let me just show you what it helped the raptor 3 engine achieve:
2. Organic geometries for structural components. Generative design, if you’re unfamiliar, is the process by which one applies load cases and fixed surfaces (similar to FEA setup) on a basic component, and the software iterates through to place the mass exactly where it best counters the load. Theoretically, this enables the best stiffness (or whatever design parameter you’re targeting) to weight ratio possible. Practically speaking, it is mostly used for organic aesthetics, and rarely for structural components at this time unfortunately, mainly due to traditional costs and difficulties associated with metal 3D printing at volume reliably.
3. Internal geometries and channels. These are critical to thermal control, particularly in heat exchangers (and rocket propulsion more generally). Internal geometry design is a whole new world that was costly if not impossible to access previously, and required significant fixed tooling.
4. Less material gets wasted in most additive processes due to powder recycling! When machining, in most cases, the chips need significant processing to get back to a usable state. Processing with high energy and time cost. With printing, the untouched powder remains usable, and can be recycled relatively easily without ever leaving the machine shop.
5. Non-dependency on part geometry. No CAM or tooling changes are required based on component design. Splicing is fully automated more-or-less, so well designed parts should theoretically be able to be uploaded and manufactured with the click of a button!
❌ The Bad:
1. Speed
Frankly, there just weren’t enough print shops capable of producing parts in the U.S. With maybe 10 reputable companies spread all across the continental 48, I could quote a part to all 10 of them and get lead times of anywhere from **6 - 15 weeks**. That is absurd. For most startups, 6-15 week turnaround for hardware just doesn’t work (especially during an internship) as a solution unless it’s something that literally can’t be made any other way.
Heat exchangers are a great example of this - more or less we just can’t get the same performance with other manufacturing methods due to lack of ability to form internal geometries. Old heat exchangers used to be made by routing tiny traces on a surface then attaching a cover, but printing lends a stupendous amount of additional flexibility.
2. Supply Chain
For a process literally designed to be as automatic as possible, there sure are a lot of non automatic steps. **A metal 3D printed part might touch 3-4 other vendors for EDMing, 5-Axis CNCing, and heat treating before making it to the customer.** After printing, the metal part is fused to a build plate. Build plates can be bigger or smaller depending on the thermal mass of the part. Fun fact: as bigger parts cool, the thermal strain can cause build plates to warp (or Pringle as it is sometimes affectionately referred to). Luckily, heat treating can relieve some of those internal stresses leading to higher part quality. To separate the component from the build plate, EDM is the method of choice. Machining typically introduces some amount of force on a part, force which can induce dimensional issues relatively easy. If you, like me, have ever bandsawed a thin walled pipe and ended up with an oval instead of a circular cross section, you know what I’m talking about. Regardless, the real problem, is that when my printing vendor in Florida sends the part to Virginia for heat treatment (1 week) then to Alabama to get EDM’ed (1 week) before it gets shipped to me and I have to coordinate post machining (2-6 weeks), my timeline has just gone from 6-15 up to 10-23 weeks. Worse than that, each step introduces quite a lot of risk. Shipping fragile parts between vendors with no oversight can be a problem, and if the machine shop messes up the machining a little, you’re out of luck and out of time. As a result I’ve literally known companies to parallel path a traditionally machined component and a 3D printed one. That degree of unreliability makes it difficult to integrate printed components en masse into most assemblies.
Additionally, infrastructure was so terrible, that the company is sometimes responsible for sourcing the powder that is used. **I recall driving to an undisclosed location, loading my Prius with 200 lbs of metal powder and driving it to the vendor so that they could begin work immediately.** Imagine buying a high grade leather wallet or purse and being forced to procure the leather itself so that the craftsman can begin work. I don’t know anything about leather and there’s no reason for me to have to become a subject matter expert on the leather when I should be the customer.
3. Quality
This is a big one. At least at the time, there was really no guarantee of quality. If I wait 6 weeks for parts which then warp and become out of spec (OOS), what am I to do? Unless I parallel pathed another component, I have to just make do with whatever the vendor provides me. What an incentive misalignment. Same goes for some very common print defects. Print defects usually occur across entire layers. Advanced systems nowadays can check for print defects after each layer and attempt to correct for the next layer, but the bottom line is that it’s already too late at that point. A layer wide defect acts as a crack propagation point, threatening part performance in tension and fatigue (two extremely common load cases). By printing tensile bars alongside the metal parts (standard practice), at least the strength impact can be measured, but if it’s not as good as desired due to defects, it’ll still cause delays or danger depending on how much margin was built into the part initially.
I will note that porosity and lack of characterization was a big issue for a long time. My impression of the industry (not to say that this is reality), is that they are now generally capable of low porosity production and have at least decent characterization curves, though I presume some knockdown is still being used in analysis to account for imperfections.
🚀 Freeform
We had the epic opportunity to go and tour Freeform’s facility a couple weeks ago and interview one of their engineers to learn a bit more about the company. They are working on metal 3D printing as a technology to address all the concerns listed above.
A typical metal 3D printing supplier buys really, really expensive printers from a company like EOS or SLM and then sets up a web portal (or more often just an email chain) to coordinate running the machines. They are completely dependent on the machine manufacturers for servicing broken machines, and frankly, those manufacturers aren’t often incentivized well to minimize machine downtime and provide a fix in a timely manner. I’ve visited factories with “upgraded” sections of the assembly line that were completely unusable and taken back out of the line. Furthermore, machine manufacturers have to make their machines easy to use by unfamiliar technicians and therefore don’t focus on speed and automation. The incentives to get you your parts quickly aren’t super aligned.
This is where Freeform is different.
There are two fundamental strategies a company developing manufacturing equipment can take: sell the machines, or sell the parts that the machines produce. Freeform is selling the parts. One interesting by-product of that decision which I wouldn’t have guessed without visiting, is that it actually creates even better incentive alignment (if you know anything about me, it’s that I love incentive alignment). When selling machines, the company wants to make sure it works for a year (or however long the warranty is) up to the projected specification. When selling parts, they care way more about the machine being the absolute best it can be! THEY are the end customer for how well their printing technology works. They are responsible if the system goes down in the middle of the night, so they invest in designing for robustness and reliability. Any increases in speed are felt immediately and can be rolled in instantly. This incentive alignment is a force to be reckoned with and likely plays no small part in what is enabling such rapid innovation by such a small company (~40 - 50 people).
As the machines are developed and reach maturity, they enable Freeform’s larger mission to create autonomous printing factories all over the world. High quality metal parts, produced locally, delivered quickly and cheaply with high design freedom (and zero capital equipment investment by the customer). That is the long term vision and why Freeform's approach to metal 3D printing has the potential to cause so much disruption to the metal manufacturing industry. As they like to put it..."Everyone else is making printers, we are making the printing press."
More Throughput = More Revenue = Your Parts To You Faster!
Metal additive has the advantage of a high degree of flexibility. Most parts that can be manufactured via casting and CNC machining, can also be manufactured via additive. Historically, it would just cost a lot more on a per part basis if any amount of volume is desired. If printing technology becomes fast enough, since they are component shape agnostic, there is theoretically almost unbounded market potential. But how can we make them faster?
There are 2 main metrics that drive metal 3D printing throughput: Power and Uptime.
Power
Freeform has built the most energy dense metal 3D printing system in the world. It harnesses 18 1 kw lasers to maximize the amount of “lasing” possible per unit of time (and they are already working on next gen systems with many, many more lasers!). Keep in mind more power isn’t always better as it can lead to a messy melt process without real-time in-situ process control (we will cover this later).
Uptime
The OG factory innovation came from Henry Ford’s Model T line. Since then, Toyota has worked to perfect the factory line and is known as a sparkling example of what automotive lines should look like (if you’re interested, “The Toyota Way” is a book about this that I’ve had recommended to me on numerous occasions, but much to my own chagrin have not gotten around to reading yet). One of the tenets of a factory line, is maximizing uptime. Downtime is the enemy of efficiency, and must be dealt with by any means necessary. One of the biggest inefficiencies with traditional metal 3D printers, is the recoating process. While fresh powder is being deposited after each layer of printing, the lasers idles. Depending on the cross section of the part(s), in can take longer to recoat than it does to melt the metal! Freeform's factory system architecture not only uses many more lasers which can all operate in parallel but also leverages multiple print beds so that one can undergo recoating while the other is being lased. As a result, their lasers operate at near 100% uptime and huge efficiencies in throughput are realized (literally orders of magnitude!).
The Maxim Of Metal
While on-site, we encountered a brilliant maxim of Freeform’s: “The Metal Doesn’t Lie.”
Engineers are empowered to implement any additions or changes they want as long as the following two tenets hold true:
1. Does it make the print faster?
2. Does it make the quality better?
If the answer to both of those questions is no, it was not a good add, regardless of what it “should” have done.
Data & Defects
Another interesting thing about Freeform, is how they deal with print defects. Print defects often occur at the layer level. Once it begins, it propagates all the way across. It isn’t until the next layer that the print typically recovers. The real issue there, is that a print defect propagated across an entire layer critically compromises any part in tension and fatigue since it basically acts as a crack. Freeform is addressing this via an insane computer vision system, capturing images at a super fast frame rate (we aren't allowed to disclose the exact number) and feeding them through on-board FPGAs which analyze the “melt pool” for quality and correct the lasers accordingly. The kicker is the timescale at which this process occurs - they are collecting, analyzing, and adjusting the printing process in microseconds (literally one millionth of a second!), giving them the ability to control the quality of the printed parts with extreme precision. This takes a LOT of compute, so they’ve actually just announced a partnership with NVIDIA to help get that compute! Huge! Fun fact: take a guess how much data they store on a weekly basis due to these high speed cameras? I’ll give you a hint, it’s in the petabytes…
What’s especially exciting about having all that data? Training! Neural nets take a LOT of data, and luckily it seems like Freeform probably has more high quality stored data on metal 3D printing than anyone else on the planet at the moment. While it will take a ton of compute to parse even some of that data, especially given how rich of a data format video is, they should be able to leverage software to refine the printing process to make it faster and higher quality, developing a robust model of the physical interactions at a fundamental level (solids, gases, liquid and plasma are all at play here).
The vibe at Freeform was electric.
SO. MUCH. HARDWARE!!!
We interviewed one of their younger engineers, Jose Martinez, who joined Freeform back in April and were absolutely astounded by the scope of what he gets to work on (also, many of the insights from this article were drawn from conversation with him).
Within his very first month, he designed and installed the depowder station (the module to the far right of the above picture, you can see the glove box holes). He is also the RE for the build module (the system controlling and moving the build plates).
There are a LOT of challenges associated with that (powder never works how you want it to), and it is an absolutely crucial subsystem of the printer. He has the awesome opportunity to design, prototype, install, test and iterate hardware everywhere from the build plate to the laser optical stack. There is so much exposure to the overall project. More exposure = more learning.
Learning things early compounds impact over time. I cannot overstate the importance of that as a young engineer. If your goal is to become the best engineer, you need to find somewhere that will throw you in the deep end (ideally with senior engineers ready to toss you a life preserver and lend a hand as needed).
We were genuinely impressed by the facility and depth of hardware being designed at Freeform. If you’re a passionate engineer looking to reshape the future of what additive manufacturing looks like, they’re hiring!
đź’Ľ Jobs!
They are hiring for these roles:
Additive Manufacturing Development Engineer
Freeform is looking for young mechanical/aerospace engineers (fresh grads welcome!) with full life cycle hardware development experience to fill this role. They want someone with strong engineering fundamentals and a passion for additive manufacturing who is ready to dive off the deep end of metal 3D printing.
Mechanical Engineer
Freeform is looking for young mechanical/aerospace engineers (fresh grads welcome!) with full life cycle hardware development experience to fill this role. They want someone with strong engineering fundamentals and a passion for complex mechanical design who is excited to design the world’s first autonomous printing factories.
Materials Engineer (Metal 3D Printing)
Freeform is looking for young materials engineers with 2+ years of experience in metals to fill this role. They want someone with a passion for metal 3D printing who’s excited to work at the cutting edge of laser powder bed fusion technology.
Manufacturing Operations Engineer
Freeform is looking for young manufacturing engineers with 2+ years of experience working in a fast-paced R&D hardware development environment. They want someone who is excited to implement and scale the production processes required for the world’s first autonomous metal printing factories.
*** Sponsored by Freeform.