Aaron Slodov isn’t the first person who wanted a plastic part made that stumbled with bewilderment upon the arcane world of injection mold design and fabrication, but along with Lou Young Jr., he could be the one to reshape the world of molding going forward.
Slodov’s introduction to tools, injection molding and how ideas normally go from a napkin sketch to a product on a shelf came five years ago. At that time, he came to Young, then at moldmaker Linear Mold & Engineering, Livonia, Michigan, seeking a means of production for a plastic Dungeons & Dragons figurine. “He needed to get a mold built, and was like, ‘This is crazy!’” Young explains. “He couldn’t, as a normal person, easily work out this supply chain of getting CAD models built, then a tool built and eventually finding somebody to mold your widgets for you.”
Atomic Industries “test bed” tooling center aims to have a wide variety of metal cutting technologies. Source: (all) Plastics Technology
As Young explained the ins, outs and inefficiencies of injection molding, Slodov, whose background was in venture capital and startups, told Young, “I think there’s probably a startup company here.” At that time, Linear had made a name for itself as an innovator applying additive manufacturing to create conformal cooling in molds. While Young was explaining the benefits and challenges of conformal cooling, Slodov would help Young arrive at an epiphany of his own.
“I was telling [Slodov] that, in general, all plastic parts would be molded up to 50% more efficiently if we always did conformal cooling, but that’s not feasible because you don’t have enough designers or time in the day to do it,” Young recalls, “and Aaron’s like, ‘We could make software that designs water lines automatically; it would just create the CAD geometry in the insert.’ I went home that night, and I couldn’t sleep. I started talking to him every day. It was a light-bulb moment, like, ‘Of course you could. It’s heat transfer; it’s fluid flows, it’s equations.’”
Atomic uses AI to help it design conformally cooled cavity inserts which are then 3D printed with Velo3D laser powder-bed fusion technology.
And just like that, Atomic Industries was born with Young and Slodov as co-founders and serving as head of manufacturing and CEO, respectively. In August 2023, it took over the former Proper Tooling’s 32,000-square-foot facility on 11 Mile Road in Warren, Michigan to serve as a “test bed” for tool design and manufacturing. Last September, it raised $25 million in Series A funding, and this April it added a greenfield 72,000-square-foot site in Shelby Township, which houses 16 injection molding machines, ranging in size from 55 to 2,000 tons, with 14 LS Mtron’s and two from Milacron, including a Roboshot. These machines run molds designed and built in Warren. When Plastics Technology visited those sites in July, the company was actively scouting a location for a planned 1-million-square-foot mega facility.
[Tools need] to be designed and built in a way that OEMs can count on them for production
This rapid growth and planned expansion are tied to its business proposition — apply Nucleus, its in-house AI-powered software — to streamline quoting, tool design, tool fabrication and injection molding, rethinking how parts get made and addressing what Young calls “scar tissue” from the industry’s status quo.
As this LS Mtron injection molding cell runs at Atomic’s Shelby Township facility, its Nucleus program provides real-time insights into production on the adjacent flat screen monitor.
“Why don’t companies just care about whatever their widget is?” Young asks, before answering his own question. “It’s the scar tissue from previous bad mold experiences. If the tool isn’t built correctly, the line stops for any tiny little part — the tool becomes this huge issue because it’s such a complex engineering problem. It needs to be designed and built in a way that OEMs can count on it for production. It’s such a critical point in getting that little widget off the production line as fast as possible that everybody spends so much time on. We’re trying to make that go away. We just worry about your part. We’re quoting you the part, and ultimately, future state, nobody’s worried about tools anymore. They’re just giving us their part design. The tool is going to be perfectly optimized because of this tech stack that we’ve built.”
The Barbell
Part of how Atomic plans to deploy that tech stack is to attack the barbell-shaped issue that Young currently sees in the molding industry. A production CNC shop — making the same detail in the same part over and over again — can create one cutter pass, load in blocks of steel and walk away, letting a machine run unattended for a week. Whereas every injection tool build is a one-of-one unique scenario, with different cutters, different paths and different “fabrication strategies”, as Young calls them.
A worker inspects a part molded for the recreational vehicle industry.
“What Nucleus is really doing, as far as fabricating the tool, is turning it into more of a productionized process,” Young says. “It doesn’t matter that the fabrication strategy is done differently for every detail we’re making because the software is just developing the fabricated strategy. It’s relieving that pain point.”
Young says if you consider a 20-week lead time for a new tool, there’s typically six weeks of design/simulation iteration, eight weeks of actual tool fabrication, and then another six weeks of first shots and trials. Atomic’s time savings and efficiencies are primarily found on the front and back ends of those processes.
“Where we’re making the biggest impact is that six weeks of simulation and design up front, and that six weeks of trial and error at the end,” Young says, “because if you can come up with a perfectly optimized tool design in a day because it’s all just cloud and compute, you’re going to have a lot less trialing because it should be perfectly optimized for that part already — all the water lines are perfect; the gate location was optimized. I would say it’s a barbell. The biggest blocks of time for a program launching are all this design and simulation up front and then this trial and error and tuning at the end, and we’re turning those from weeks into days. The fabrication time of the mold itself, that block in the middle, is being impacted too, but not as much as those other pieces.”
Atomic builds these press-side data-collection units to help it gather process information to feed back into its Nucleus software.
Feeding Nucleus
Key to Atomic’s compression of the tool-development cycle is the gathering, interpretation and application of data it gathers from every aspect of its operation. Sensors feed data from every job — whether it’s printing an insert, cutting a cavity or molding a part — back into Nucleus. All molded parts are inspected, with the software learning from good and bad parts equally. On the molding floor in Shelby Township flat screen TVs in front of the molding machine display real-time data as the mold runs, including screw position, injection pressure, cycle time, injection speed, clamp force, oil temperature and more. Nucleus also provides insights into the financial performance of the part run, down to margin on each shot. Cameras around the work cell track all aspects of the press and production operations required to make parts. Additional vision systems are trained to watch the cell, monitoring mold open/close for more productivity insights. All this information is fed into data-collection systems, which Atomic made, that sit next to the press.
Which pins are actually doing work and which ones are just along for the ride?
Young uses ejector pins, and their impact on cooling channel design, to show how Nucleus and Atomic can apply all this data into making better molds and better parts. Pressure sensors under the ejector pins measure pressure in cavity but also track ejection force. “Which pins are actually doing work and which ones are just along for the ride,” Young says, noting that the knowledge gained regarding the ejector pins and everything else will be applied to future tool designs. “If you don’t have enough ejection you have zero parts coming off your mold. This is why, traditionally, people put ejector pins everywhere because if you don’t have enough of them, you’re in trouble. Whereas if you don’t have enough water, your cycle time is just longer than it needs to be. People become complacent, and say ‘48 seconds is a pretty good cycle time.’ That’s how optimizing all these systems ultimately makes you make widgets faster, better. When our algorithm is deciding where to put ejector pins, it’s ultimately learning how a part is shrinking and warping on a core, and where pins actually need to go to eject it. Today companies often put pins everywhere, and then try to design the water system around that. What if you just put the exact right amount of pins in? Now you have more real estate to optimize water and cooling, and now you improve your cycle time.”
For the parts themselves, Atomic applies 100% inspection for the data that is needed to continuously grow, teach and refine their proprietary AI model, Nucleus. “We learn what is making the best part,” Young says. “Within one week of starting production, we’re already learning, even for parts that passed inspection.”
Automated Quoting, Cost Estimating
Atomic sees its future as ownership of the entire process, with vertical integration which has included building out its own solvers and simulation tools, benchmarked against existing programs. “Not only is part production the end market we want to affect, but to truly build what we’re building, you need to own the whole thing,” Young says. “You cannot be the best part molder if you’re not getting the best tools, and you can’t get the best tools unless you’ve got the best tools designing them correctly and optimizing them. You need to own that whole thing, and you need to know how this part ran to feed right back in to the beginning of Nucleus to retrain those models and tweak those solvers.”
As those models and solvers are continuously tweaked and improved, the software opens up additional opportunities, where if you can simulate highly accurate cycle times, which is also part of Nucleus, it can enable precision cost estimating. “It’s automated quoting too, built on physics,” Young says. “Fully autonomous quoting is the future where customers are just uploading part files.”
As part of an experiment, Atomic created this tool with various cooling technologies — human and and AI designed, as well as regular and conformal — to test the efficiency of various methods.
