Skip to content

A.I. Fuels Aerospace Manufacturing Automation

Ilene Wolff
By Ilene Wolff Contributing Editor, SME Media

Robots have become able co-pilots on the factory floor as painters, sanders, machine tenders, and materials transporters

Automation expert Phil Freeman, senior technical fellow, Boeing Research & Technology sees a trend for more autonomy in the robots and systems used in the aerospace production process. (Provided by Boeing Co.)

When Phil Freeman started his aerospace career toward the end of the 20th century, there were no robots drilling holes for fasteners on aircraft.

“We learned a lot about why; (it’s because of) the way the robots work in terms of their accuracy and their stiffness,” explained Freeman, senior technical fellow, Boeing Research & Technology (Charleston, S.C.), Boeing Co. “Then, through that, we figured out how to overcome those challenges. And now, I don’t know of an aerospace company that doesn’t have robots drilling on aircraft.”

In addition to drilling holes for fasteners, robots are painters, machine tenders, materials transporters, sanders, and polishers for aerospace OEMs and their suppliers.

Advanced composites maker Airborne Aerospace BV (The Hague, Netherlands) uses automated programming for a robot that seals the edges of composite parts by devising its own glue path based on CAD drawings.

Satyandra “S.K.” Gupta, GrayMatter Robotics’ chief scientist, said with physics-informed A.I.-driven software, known physical principles can stand in for a lack of data in manufacturing. (Provided by Gupta)

Montreal-based software company RoboDK enabled automation solutions provider Wilder Systems Inc., Austin, Texas, to develop a robotic aerowash—like a car wash except the robots move while the vehicle is stationary—to scrub down dirty craft. Neither Airborne nor Wilder Systems responded to requests for an interview.

Not only are robots taking on the repetitive and more dangerous work most humans don’t like, they don’t need a security clearance to work in restricted industries like aerospace and its more secretive twin, defense. Robots also pitch in to make up for the lack of qualified workers.

Aerospace, with legacy processes such as hand layup, may seem like it’s late to the automation game, but Freeman points out that’s not accurate.

“Aerospace actually has a lot of automation in it,” he said. “Most fabrication is automated, right? So on the low end, sort of classical, you have a lot of CNC work. But even in things like, you know, automated fiber placement of composites, and things like that, there’s quite a bit of automation.

“And then, as you get into assembling and joining, there’s still a considerable amount of automation. So you see automated trim and drill, automated drilling and fastening of aircraft. And then you see automated inspection with NDI (non-destructive inspection) and non-destructive evaluation. You see a lot of automation with workholding. At the (Boeing) Composite Wing Center we’ve done more in material handling. We’re using AGVs (automated guided vehicles) to move large parts around, robotic fixturing, and things like that.”

Freeman—who’s been involved on automation projects for the F-18, C-17, F-15 and 787 Dreamliner—sees aerospace automation trending toward more autonomy in the robots and systems used in the production process.

“I think autonomy is really one of the areas that we’re growing in,” he said. “The ability of systems to work in less defined and less constrained environments.”

Otto Motors’ 1500 automated mobile robot (AMR) is a workhorse designed to move payloads as heavy as 1900-kg (2.1 short tons) through demanding environments faster than any other AMR on the market, with zero compromise to safety, according to the company. (Provided by Otto Motors)

For example, Freeman noted, automated mobile robots (AMRs) can navigate the factory on their own, get to the location where work needs to be done, and then perform the work based on their own artificial intelligence.

“That involves a lot of machine learning, artificial intelligence, algorithmic robotics technologies such as automated task planning, automated path planning. These are ideas that we’re borrowing from other areas like self-driving cars,” he said. “That’s where I see a lot of emerging areas in robotics—moving from sort of the classic, rigidly scripted set of steps and sequences to more autonomous systems that are able to react to the environment, make decisions on their own, and continue to work.”

As with most researchers, Freeman and his colleagues are constantly trying to push the boundaries of what’s possible.

That’s exactly what one supplier did during a decade of trying to implement robotic sanding to polish acrylic aircraft transparencies.

One Size Doesn’t Fit All

In a quest to find automated process solutions for its production of aircraft transparencies (windows and canopies), U.K.-based GKN Aerospace Services Ltd. worked for 10 years with experienced automation integrators using off-the-shelf robots and controls. Despite its efforts, the number of failed systems configured “well outnumbers” the successful deployments, according to Martin Philo, principal research engineer.

So, the company set out to identify the cause.

Capriol's model MNT combines an Otto Motors 1500 automated mobile robot with a FANUC R1000 series robotic arm designed for handling medium-sized payloads. (Provided by Capriol)

“After integrating a few of our own projects, GKN Aerospace has come to the realization that ‘one-size-fits-all’ systems provided by many of the robot suppliers simply do not have the capability to precisely execute the more complicated processes that GKN Aerospace needs to automate, to stay competitive, and protect our workforce,” Philo explained. “Further study brought us to the conclusion that the hardware was not the issue, but it was the robot software that cannot adequately execute the automation plans due to the complexity of our processes. What is needed is a comprehensive systems solution inclusive of the software, component drivers, and A.I. developed specifically to handle complex processes.”

No newcomer to automation, GKN Aerospace already uses CNC machines, forming automation equipment, and automated inspection. An additional task up for automation at GKN Aerospace’s Garden Grove, Calif., plant is sanding and polishing acrylic military canopies and cockpit windows, and passenger aircraft windows. The job is extremely labor intensive: It takes more than a work shift to sand one part.

GKN Aerospace is confident it’s found the right partner to answer its need for automated sanding and polishing in California’s GrayMatter Robotics Inc., with its optimism bolstered by working directly with the developers.

GrayMatter’s experts agreed with the aerospace supplier’s conclusion that providing a robot as a complete solution along with software for off-the-shelf robotics were the sources of failure in previous projects. Its Scan&Sand technology uses optical scanning and custom, physics-informed A.I.-driven software to support industrial robotic arms mounted on a gantry and equipped with an abrasive tool.

Satyandra “S.K.” Gupta, GrayMatter’s chief scientist, said, “If you see a lot of A.I. that’s being used in advanced industry, movies and Facebook image detection, those are all purely data-driven technologies. They have millions and billions of pieces of data available to them and they simply mine the data looking for patterns.”

In a manufacturing center where data is limited, though, known physical principles can stand in for the lack of data. For example, it’s known that when sanding, applying more force removes more material.

That level of information, or data, is enough for Scan&Sand. “You don’t know exactly what the relationship is but you know at least that that’s a trend I’m expecting,” said Gupta. “So you don’t need to mine the data to get the trend. You already know that these physical relationships exist, and beyond that I would need data. So, whatever information that you know about the process you want to exploit.”

As promised with automation, GKN Aerospace expects to churn out work faster and of consistently higher quality with GrayMatter’s help.

Based on initial estimates, Scan&Sand will increase productivity by completing a part in less than four hours, giving GKN Aerospace’s production a boost by a factor of three or four. In addition, automation has the potential to significantly reduce its scrap, repair, and rework costs associated with sanding, which can reach $5 million yearly.

“We chose to partner with GrayMatter on this project because they have a focus on sanding, an appetite for achieving high quality and not just productivity, and the software technology to support the project objectives,” Philo said.

GKN started using Scan&Sand in January 2022 and plans to deploy it in production next year once it achieves the necessary technology and manufacturing readiness levels.

“The project team is working to develop the technology to achieve precise quality, while exploring new inspection methodologies to incorporate into the system for production sanding and polishing,” said Philo.

Saving Time

At GE Aviation in Asheville, N.C., highly skilled operators were their own gofers as they conducted CT non-destructive testing of aircraft engine shrouds. At the same time, the plant’s production of the shrouds was scheduled to balloon to meet customer demand.

Something needed to change to enable the increased productivity and it needed to be scalable throughout the plant.

“This process was manually interactive with a lot of people interaction, process iteration,” said Stephen Rice, GE Aviation lead engineer for testing, in a case study video about the 2018 start to fix the problem. “We were looking at a rate of 30-60 (shrouds) a week at that point in time. Our future rate looking into 2022 was going up to 1,500 a week, so we knew we had to find a path forward.”

With each of those manually interactive processes taking 20-30 minutes each, the target rate of 1,500 shrouds every seven days would take more hours than there are in a week.

The solution, and the subject of the video, was to create a fleet of mobile machine tenders by installing Fanuc collaborative robotic arms atop AMRs from Otto Motors, (Kitchener, Ont.), a division of Clearpath Robotics Inc.

“By bringing in automation, we saw a 20 percent decrease in the direct labor going into the part,” said Evan Bryant, GE Aviation process engineer, in the video.

The OEM saved a lot more than time.

According to the case study video, the company realized a $1.3 million return on investment in 2019, the first full year of the mobile fleet’s operations. In addition, mobile machine systems integrator Capriol Ltd., West Bloomfield, Mich., proposed a solution to save $8 million in capital investment to get to final production volumes, Tom Post, Capriol’s operations manager, said in an interview.

“In the case study video that you watched, they developed the Fanuc collaborative robot attachment that went on top of our vehicle to suit the very specific load handling needs of interfacing to a CNC machine and taking a very important, expensive part out of that machine and moving it downstream in the process,” added Otto Motors CEO Matt Rendall in a separate interview.

The setup for GrayMatter Robotics' automated Scan&Sand technology includes two robotic arms, one for the process and another for inspection (Provided by GrayMatter)

Who’s Otto?

Otto Motors has a longstanding relationship with GE, with the American multinational being an early investor and partner/customer since the Canadian company’s founding in 2015. Otto’s specialty is providing “some of the largest AMR fleets on the planet” for material handling purposes, according to the company.

While Otto Motors builds its AMRs to offer customers the best total cost of ownership, longest life, and highest uptime, its real power lies in its software.

“The software actually is the most important reason to integrate our product into an environment,” Rendall said. “You’ve got two layers to it: There’s the software that runs on the vehicle, but from a process level, where the magic actually lies is in the fleet management software.”

For the AMR’s software, operators want to have the most intelligent, most resilient and robust software onboard. With that kind of smarts, the robot can respond to real-world situations like turning a corner and working in chaos that’s impossible to simulate, all in a manner that still allows the right part to get to the right place at the right time. Fortunately, even newly deployed Otto AMRs benefit from the hive mind-like quality of the software that’s currently accumulated more than 3 million hours of production driving experience.

“Think about a teenager who is learning how to drive, and maybe it takes them two weeks worth of driving experience before they can get their driver’s license,” said Rendall. “Well, at the machine scale, you have shared intelligence, right? Picture the ability to plug a cable into the driving instructor’s head and download all of that driving experience into the student.

Now every single vehicle that ever hits the shop floor is able to plug into the central intelligence of our software, and all of a sudden have 3 million hours of driving experience. That’s incredibly important and potent and differentiated from our perspective.”

The technology behind Rendall’s description is appealing: The fleet management software is one of the biggest reasons why customers choose Otto over its competitors, he said.

“It’s the fleet management software that interfaces your AMR fleet into your manufacturing execution system, your SCADA system, your PLC network,” he said. “It is what gives you seamless, end-to-end integration and handoff of materials from a piece of processing equipment to a material transport solution like ours.”

In the GE Aviation case study, the plant’s AMRs interface with CNC machines, which means the robots and machine tools have to communicate. At the same time, the AMRs and fleet manager are signalling back and forth.
“So you have as close to a lights-out flow of materials as you possibly can,” Rendall said.

Running Past GE

While Otto’s AMRs use lidar and software to navigate, Capriol incorporated an additional layer of vision in the end effectors it manufactured for the robotic arms atop the mobile robots in the GE Aviation project.

“It’s an enabling technology for mobile robots,” said Post. “Because we’re driving from station to station, we need to get the location for the robot positioning. So we use the robot vision to change the offsets on the robot, find the part to be worked on, and set the offsets at the location.

“When it drives up, the AMR positions (itself) ±2-cm, which is huge in the robot world. To be able to position within half a millimeter of the design specifications of the robot, we have to use fiducials on the equipment or at the station, which is a barcode or a dot matrix the robot uses to set its offsets. And then we are able to precisely grab parts and do work.”

Like Otto Motors, Capriol also has a longstanding link to GE. “We (the founders) worked together at General Electric Global Research in the advanced manufacturing group,” said Post. “And our director, Roland Menassa, instructed us to make mobile robots. So we started building their first units.”

When GE’s fortunes took a downturn, it eliminated Post’s group. “We were like, we have customers, internal customers, that are interested in this product,” he said. “Can we have it? And they said yes. So we continued to work on the project under Capriol.”

Some time later, representatives from GE Capital, the company’s financial services division, were touring the GE Aviation plant and saw the Otto Motors-Capriol material handling robots.

“One of the guys asked me how did you guys do this?” said Post. “GE dropped the ball. We picked it up and ran it.”

  • View All Articles
  • Connect With Us

Always Stay Informed

Receive the latest manufacturing news and technical information by subscribing to our monthly and quarterly magazines, weekly and monthly eNewsletters, and podcast channel.