Using an AI-driven vision system with proprietary software, Teqram’s EasyGrinder robot can identify a part on the pallet, pick it, place it on the worktable, grind it, and re-palletize it automatically. Images: Accurate Metal Products
If you’re ranking the most tedious jobs in a fabrication shop, it’s hard to beat removing slag with a hand grinder. Not that it’s low or somehow demeaning—just that it can be tiring, dirty, and a little hazardous if you don’t know the ropes.
So, when Accurate Metal Products needed a way to stay on top of its parts cleaning without hiring an endless string of temp workers who would get one taste of the job and soon leave, Teqram’s EasyGrinder robotic grinding machines emerged as a possible solution.
AMP operates facilities in Rockford, Ill., Milwaukee, and Menomonee Falls, Wis. It specializes in flame cutting parts for a variety of industries, including mining, defense, and energy. The parts it cuts for those customers vary widely in size, but one it makes frequently is a hook for hydraulic loading docks flame-cut from 1.5-in.-thick carbon steel that is about 12 in. long and 8 in. wide.
The hooks might be relatively small, but moving them on and off a grinding table gets taxing, even for parts cleaners with experience. Manager Will Alverson was quick to point out that the EasyGrinder isn’t faster than a human grinder—but speed isn’t the point. The key benefit to AMP lies in taking work that’s either inherently unsafe or that people simply don’t want to do and automating it.
“It’s physical, it’s dangerous—sometimes you’re moving things around or having to handle awkward pieces,” said Alverson, whose family owns the company. “It’s a really hard thing to hire for, and it was kind of the bottleneck. So, when we saw that Teqram was producing vision-guided robotics that sort of solved this problem, you know, we were very interested.”
Common Operations
AMP’s two robots—one in Rockford, one in Milwaukee—work with several different grinding discs, some of which are slightly larger than most shops use. These include flap wheels and also ceramic discs, in addition to chisels for larger bits of slag.
A robotic arm with a magnetic gripper loads pieces autonomously onto an EasyFlipper folding table that is used to flip pieces and comes in handy for the Blanchard grinding that AMP commonly does. The machine also uses an AI-driven vision system that can identify different workpieces on a pallet, which allows users like AMP to run the system lights-out if necessary. The combination of magnetic gripper and folding table enables the system to handle a variety of different jobs with no human intervention.
“It’s very common for us to grind two sides and then grind a lead-in so you can do vertical grinding as well, which is nice,” Alverson said. “We flip the part, finish it, re-palletize it, and move on to the next one. We tend to like it either for large-quantity production runs that are just tedious for human operators or things that are a little bit larger and are going to be awkward to move around with a crane.
“So, it’s either awkward to handle, so we throw it on the robot, or there’s a lot of consistent jobs where … sometimes we’re cutting 800 at once.”
AI-driven Vision and Software
The robotic grinder is controlled by Teqram’s proprietary PC-based software stack (no PLCs to be found). The company also developed its own AI algorithms to guide the machine’s vision system.
The EasyGrinder’s robotic arm’s magnetic gripper can be switched to any grinding tool to complete the grinding and cleaning process.
“We found that the vision packages you can buy or the open-m Director Frans Tollenaar. “We have a very specific vision issue: We have to find the metal parts on pallets, and the shades of brown and gray are quite tricky. So, we developed the vision algorithms ourselves and also implement the AI ourselves to recognize features on parts that are relevant to the grinding process
“There’s a lot of machine logic that needs to be 100% predictable every single time—you don’t need LLMs or deep learning algorithms for that. What we use AI for are just very specific subtasks where AI excels—recognizing details on the parts. The 3D sensors all have their limitations, and using AI to overcome those limitations is very beneficial.”
In addition, with skilled operators ever more difficult to find and retain, the ability to run the machine’s AI-driven vision system without programming helps the shop keep product moving through the automated cells easily.
However, adding the robots isn’t about automating people out of the shop; it’s about automating for the people in the shop. In fact, the robots have proven such a hit that some members of the parts cleaning team don’t want to leave, despite the job’s reputation as being a sort of lower-tier task.
“We’ve been able to hire some really good people as parts cleaners to the point where we probably have too much parts cleaning capacity right now,” Alverson said. “And we keep offering to move them up and train them on other machines—and people don’t do it. It used to be kind of the bottom of the totem pole. And now we’ve got a ton of really good people in that role who are actively not wanting to do other things.
“We’re not using this to replace people,” Alverson said. “We’re augmenting them. If you see more throughput per employe at a workstation, then you can compensate them better. And what the robots have done is they’ve removed the most dangerous and tedious parts of that job.”
