Northrop Grumman, like Boeing and Raytheon, is working to link together its digital environments, starting with the design phase, and “making sure we put quality up front in the design in order to create a much more efficient downstream set of activities related to manufacturing and then into the sustainment environment,” Bissell Smith, director of integrated enterprise solutions at Northrop Grumman, said during a March panel talk at AeroDef Manufacturing in California.
A big part of the effort concerns integrating the PLM and ERP worlds—“where we see opportunities to embed different capabilities to achieve higher quality, … more efficient use of people’s time and better automation in the factory areas.”
Chief among those capabilities is artificial intelligence, or AI: the theory and development of computer systems that can handle tasks that previously required human intelligence, including decision-making.
The “Northrop Grumman integrated digital environment” initiative is aimed at applying new information technology (IT) along with imbedded AI capabilities in it, he said. Right off the bat, automation and product quality should be improved.
By “sustainment,” Smith essentially was referring to maintaining profitability by making good decisions “in a very complicated field of activities that are going on in the maintenance area,” he said. “It’s everything from parts management and the logistics associated with that all the way down to using information to predict longer lifecycles of the product. And then in the customer integration area.”
AI is part and parcel with data science.
A&D manufacturer are “hiring a plethora of data scientists today that are descending within our organizations and finding new ways to work with information and produce insights,” Smith said.
AI is also part and parcel with machine learning—where teams like the one Smith leads are “trying to create predictions” that, among other things, “can really help you manage your ROI,” he said.
In general, he added, “we want to get more out of our analytics and making use of the data that we have” to positively impact efficiency and quality.
“I can see where AI can help a lot around capacity management, and getting a better sense situationally of what’s going on in a particular work cell or factory area in order to provide a better way to improve efficiency and make better use of our assets and our people,” Smith said.
There are so many benefits associated with predictive maintenance and location analysis that AI is “irresistible,” he suggested.
Situational awareness is paramount not just because supply chains and manufacturing facilities for individual manufacturers often dot the globe but also because while automated machinery is programmed to perform a repetitive task “it knows not what is around it,” Smith said.
AI technology can also help machinery sense when workers are nearby and when someone inadvertently puts another asset next to it—that it could knock over if not careful, he added. “It’s no different than what they’re doing with some of the automated vehicles and sensing collisions.
“No matter how good your procedures and how good your training is of your employees, they’re still leaving materials behind,” Smith said.
AI has “a great opportunity … doing real-time inspection in the automation process and then … taking actions associated with that learning,” he added.
Supply chain management “is probably one of the most complex things we do” and involves decision trees that boggle the mind, he said. “There’s no doubt there’s a lot of errors in the process. … And so … AI capabilities in the application can go a long way in helping people make the right decisions very quickly and then speed up the process in order to get a contract out the door. I’ve not seen anybody develop that yet, but I think the opportunity space is huge in that area.”
Then there is cybersecurity.
AI technology “can really help” manufacturers investigate attacks, as well as detect fraud, Smith said, noting that security and privacy concerns rise along with increased automation and interconnected architecture.
Boeing uses AI for a variety of applications, including diagnostics, Bala Chidambaram, capability technology leader at Boeing, said. AI can be thought of as “machines that act like humans” and “rational agents,” he added. “They have some objective function that we give them, and they can accomplish that function. And in the process of completing that function, they act like humans.”
As operations technology (OT) and IT systems converge, “the machines are getting more and more integrated with the IT network,” creating new challenges related to cyber security that AI can address.
Process optimization is another good practical example of AI in manufacturing, Plataine CEO Avner Ben-Bassat said at AeroDef. The more AI software vendors are able to produce a complete business workflow, the more value they generate.
Infosys helps aerospace manufacturers “achieve excellence across the value chain by using the digital thread,” which is tied to AI, Nitesh Bansal, senior VP and industry head for manufacturing at Infosys, said at AeroDef.
The world will by 2020 have about 20 billion connected devices, he said. And those devices will produce data day in and day out. “That basically translates to about 163 zettabytes of data. And that’s actually more data produced every day than was produced in the last 10 years taken together.”
The good news is aerospace and defense manufacturers see it coming. According to a small Infosys study, adoption of AI grew to 50% this year from 25% last year, Bansal said. “That means the data is being utilized, that companies are getting ready to do more and more productive stuff with AI technologies.”
The bad news is “there are barriers” because adoption is only at of 50%, he added.
Data is often a mess. “There are lots of data scientists trying to figure something out but they don’t know what.”
The idea that “data is the new oil” is interesting, Bansal said. “If that is true, then … the aerospace industry has a lot of crude oil.”
Thanks to government regulations and “the extensive research that goes into producing the final end product, the amount of data that is collected across the aerospace value chain is immense.”
To make AI real, “you need to marry it along with the business outcomes,” Bansal said. “You need to know why you want to use a certain kind of data married with a certain technique to achieve an outcome. Whether you are going toward reducing risk or improving efficiency. Are you increasing safety or increasing value in a certain case? It’s only with a proper understanding of what ROI you are trying to drive that you will be able to bring it to the right kind of data, used with the right techniques, to produce an ROI” in the product lifecycle area.
One example is Infosys’s work on engine balancing.
“Engine balancing is a very critical aspect of making an aircraft fly-ready,” he said. “Engines are very unique. Despite being fairly standard, each engine is unique. And you need to carry out multiple test flights to understand the variance and progressively bring the variance down.”
That involves conducting many test flights—and collecting “a ton of data.”
The problem cannot be “completely eradicated or solved,” Bansal said. Test flights are necessary “because engines being unique, you don’t have historical data that can provide a prediction.
“But you can use AI algorithms to learn and narrow down the variance in a significant way” so that fewer test flights are needed.
AI is “not just a shiny object with a promise of the future,” Bansal said. “This is real today, to be used by all of us to make a real bottom-line difference on our books.”
Process optimization, identifying quality issue in real time and predictive maintenance are good practical examples of AI in manufacturing, Ben-Bassat said.
Process optimization is “the most comprehensive area,” he said. Aerospace manufacturers are looking for the strategic – top line – benefits, such as increased throughput and improved quality, let alone bottom line benefits such as cost reductions.
Impediments that need to be overcome include “the pure maturity of the capabilities,” Smith said. “This is a relatively immature area. So it’s really important that you’ve got teams of people looking at the technical readiness of these capabilities and not deploying them faster than you should.”
In fact, Northrop executives take pride in being “a little bit late adopters,” he said. “We love it when Lockheed and Boeing go out there and spend billions of dollars on some of these things, and then we just come around and pick them up later for a lot less money.”
To grow AI-area embedding technologies, it is important to try to discover data quality errors in the upfront transactional process, Smith said.
Northrop uses a tool called “information steward” that does just that—as the transactions are occurring.
“You’ve got to put capabilities in place like that,” he said, to gather the kind of data needed for predictive analytics and improving process efficiencies.
Smith recommends hiring experts upfront to install AI technology, rather than counting on employees to learn on the job.
If the installers don’t understand “the inner workings and the math and the logic” associated with the technology, there will be “way too much learning after deployment,” which will create variations and higher costs and higher deployment activities, he said.
AI as a “black box” that takes complete control is a concern, Ben-Bassat acknowledged.
The question that needs to be answered is: How do manufacturers maintain visibility into what AIs are doing and what they are changing?
“It’s a question of trust, like any relationship,” he said. “So this relationship between AI and us takes time to build. People need to understand to an extent how it works.
“One of the challenges of machine learning as one of many AI techniques is the fact that it’s in many ways a black box. Sometimes even the data scientists don’t even know how it worked, but it did.”
The building of trust “also calls for domain expertise,” Ben-Bassat said, suggesting that care be taken to carefully match people working with AI to subjects they know well.
And it is important to measure the outcome, as opposed to what the AI did, to ensure the AI is bringing value, he added.
The changing workforce is helping to advance AI in manufacturing.
“Newer colleagues—Millennials and Gen Ys and Gen Zs—are used to consuming technology on a day-to-day basis, for their daily lives,” Bansal said. “And they expect to work in the same manner.”
Young workers want to be able to accomplish a workflow activity with an app on their phone rather than a PC—and for the phone to recognize what they are doing automatically, he said.
AI, at least for these people, creates a “human experience” in manufacturing, he added. “I think cultural change is certainly needed, and it’s going to be an important catalyst to adoption of AI within the organization.”
Connect With Us
