AI-FORGE uses incremental forming, a heat-assisted metalworking process that permits users to manufacture small lots of customized manufactured parts for military aircraft. The addition of artificially intelligent software allows the robotic system to make significant forming decisions on its own without the need for a human operator, offering near-term cost- and time-saving benefits as well as an improved ability to replace hard-to-find aircraft structural parts.
OSU’s College of Engineering received $500,000 in funding from the Advanced Robotics for Manufacturing, or ARM, Institute to develop the system, while AFRL’s Materials and Manufacturing Directorate provided project leadership and oversight as well as an additional $150,000 in post-doctoral research support to OSU.
A multidisciplinary team based in OSU’s Artificially Intelligent Manufacturing Systems, or AIMS, Laboratory, primarily led development efforts, oversaw system integration and advised on the use of artificial intelligence decision making software. Additional collaborators CapSen Robotics, who supplied the system’s computer vision components and motion-planning software, and Yaskawa Motoman, who provided robotic hardware, bore an additional cost share to see the project to completion. Researchers outfitted the system with custom sensors to control and track temperature and material shape changes.
After the system was developed and integrated onsite at OSU, it was transported to WR-ALC, a military maintenance and sustainment depot, where it ran autonomously under environmentally taxing conditions for approximately six hours without human intervention.
“Parts break, and they’re critically important — if you don’t have them, the mission fails,” Donegan said. “And these are not always the kinds of things that you can just go and pick up off the shelf at a big box home hardware store. These are things that were, at one time, custom-made for that given aircraft platform, and now maybe you can’t get them anymore because the company that made them went out of business, or the forging house lost [the specific tools] needed to recreate that specific part. These kinds of things happen more frequently than we’d like.”
The AI-FORGE system offers a solution that is analogous to a human metalworker using heat and pressure to robotically forge replacement parts. In this case, AI-FORGE uses artificial intelligence to automate specific portions of the forging process, including placing, holding and deforming a part to create desired formations.
“When a human blacksmith forges a piece of metal into a specific shape, they have to make all kinds of decisions on the fly,” Donegan said. “How to orient the component, where to apply force, how much heat to use, when to put it back into the furnace — all of this has to be decided in real time. A skilled artisan blacksmith can make decisions like these without even thinking about them, and our research team is especially interested in how to replicate that kind of decision-making process [with artificial intelligence].”
For the initial demonstration, project engineers chose to highlight the system’s ability to run autonomously rather than focus on its capability to create overly complex parts, said Shane Groves, lead automation engineer at WR-ALC. To this end, for its first run, engineers tasked the robot with the creation of a geometrically basic but still critically relevant component that is currently in short supply at some local air bases.
“We picked a relatively simple part as the demonstration unit because, first, we had to see if it is even possible to do this type of work in a depot environment, and we determined that it absolutely is,” Groves said. “That’s why I think this was one of the most successful ARM demonstrations we have ever seen here at WR-ALC because we established that yes, the system was able to operate autonomously and complete the task in a very taxing environment.”
With an initial test run complete, researchers can begin to think about ways to build a more sophisticated system that allows users to better control the inner microstructure of these parts, including ones that may have been additively manufactured initially, said Dr. Andrew Gillman, research materials engineer in AFRL’s Materials and Manufacturing Directorate.
Incremental forming, a process that is performed in a series of locally controlled steps, is of particular interest to members of the materials science community, largely due to its potential to deliver repeatable results, Gillman said.
“If we can locally control the properties of these components, it could transform the front end of tech development,” Gillman said. “Perhaps we can come up with new designs, think of entirely new ways to use the material.”
The inner microstructure of additively manufactured parts can be unpredictable, Groves said.
“Unpredictable results do not build confidence, but what [incremental forging] allows us to do is to not only create a shape, but also align its microstructure in a way that is repeatable,” Groves added.
Form, fit and function are all aspects that both human and robotic blacksmiths must take into account when shaping specific components, and function boils down to whether the part can properly perform its job, said Dr. Michael Groeber, associate professor of the Integrated Systems Engineering and Mechanical and Aerospace Engineering departments at the Ohio State University.
“Function largely comes from the inner properties of the material,” Groeber said. “So, once I have my shape, I have to ask, well, can it hold the load that it needs? Can it withstand stresses and things like that? And that’s really driven by the internal structure of the material, which is highly impacted by processing. The end goal is, we don’t only want to be able to make shapes. We want to make shapes but also control how we get there, in order to produce repeatable results. Then we can start to get into some really clever designs that will ensure performance [in aircraft or other systems].”
The Ohio State University team included co-principal investigator Dr. Steve Niezgoda, an associate professor of materials science and engineering who co-led the AI decision-making portion of the work, and Walter Hansen, a mechanical and systems engineer at the Center for Design and Manufacturing Excellence, who spearheaded the physical system integration and hardware design. Dr. Toby Mahan’s AFRL-funded post-doctoral research efforts contributed significantly to the success of the project, while Dr. Glenn Daehn developed the initial concepts upon which the project was based, Groeber said.
In the future, artificially intelligent incremental forming methods could be applied to the development of personalized medical devices, Groeber added.
“We can imagine smaller systems where this AI decision-making of incremental forming applies, such as metal implants,” Groeber said. “If a surgeon needs jaw reinforcement plates or hip implants for a patient, right now, they have to eyeball what they think might be the best size, choose from limited options, and then bend the component manually to fit the patient. In the future we could see robotic systems like this doing the bending. There is potential for accuracy and quality improvements when a robot does this versus a surgeon because everyone’s anatomy is different.”
Groves said the demonstration effort was the direct result of an effective partnership between AFRL, academia, industry, the ARM Institute and the depot.
“It’s the way it was designed to work from the start,” Groves said. “AFRL develops the technology with the help of their partners, and then we help them test it, and then we transition it together, as a team, into production.”
Groves said his team’s mission is to support the warfighter, no matter the challenge.
“The people who are successful are those who are agile and can adapt to changes,” Groves said. “And the systems, then, have to do the same.”