Why don’t more manufacturers in the United States use smart manufacturing technologies like AI and machine learning to reduce waste, achieve predictive maintenance and enhance their automation systems?
That question was the focus of “The Role of AI in Manufacturing,” a roundtable sponsored by CESMII, the Smart Manufacturing Institute. Panelists represented Procter & Gamble, Raytheon Space and Airborne Systems, Microsoft, UCLA and CESMII.
Panelists talked about overcoming impediments like “pilot purgatory,” platform-dependent software, data scientists with no manufacturing domain knowledge and data “anarchy.”
“There’s still a lot of hype around this space, which isn’t unusual when you have the potential to disrupt an entire industry,” said CESMII CEO John Dyck, the moderator of the roundtable. “But that makes it even more important to understand the distinction between what’s still aspirational and what’s pragmatically achievable.”
Which takes us back to the question of what’s constraining manufacturers in broader adoption of AI?
Many companies report trying to incorporate AI in their enterprise but being stuck in “pilot purgatory,” unable to scale a small, successful project.
Even a Fortune 500 company like Procter & Gamble admits to having problems. Still early in adopting AI, the company has managed to deploy some machine learning algorithms. But P&G is stymied by scaling them.
“You realize, I don’t have an infrastructure that can get the data out in the way I want it,” said Jeff Kent, leader of P&G’s smart platforms. “I don’t have a platform that contextualizes well, and I don’t have a place where the algorithm can be developed very easily by non-data science experts. So, we’re at that stage where we have [some] successful pilots but they kind of stay within the context of who developed them.
“I think we are coming out of pilot purgatory,” he added, “but we’re only going to sustain a scalable adoption and achieve the whole promise of Industry 4.0 when all those come together into a full work process, and a full set of applications that P&G, our suppliers and OEMs all can participate in.”
Kelly Dodds, advanced manufacturing tech director at Raytheon, said the military contractor has had success integrating AI into robotic applications using machine vision.
“The ability for machine vision to pick up what you want to pick up every time and improve itself is a significant endeavor,” she said.
To help meet the AI-adoption challenge, Dodds said Raytheon is promoting data science educational programs with a manufacturing context.
“We need some data scientists that have domain expertise,” he said. “So, growing that pipeline of folks that have that context is important.”
Speaking of data scientists, Jonathan Wise, CESMII’s VP of technology, is a software developer who thinks his fellow coders must start their designs with deployment flexibility in mind.
“We have a workforce that has developed intelligence into a PLC for the past couple decades, but those platform vendors have built no hardware abstraction layer (HAL) between the code and the PLC,” he said. “In IT software, by contrast, for the past few years, we’ve been building our software … out of components, and those components are loosely coupled into a particular architecture and attached by well-defined interfaces.”
The antidote to platform-dependent software, Wise said, is to build algorithms that are platform independent and built with common information interfaces.
In addition, the data itself needs to be in a standardized format, he said.
Jim Davis, vice provost of IT at UCLA, said, “The ways to do that ... is to think about how you exchange data, share it, pool it, combine it, work with it in a collective fashion. So, there’s an industry wide strategy.”
With the pandemic and climate change disrupting supply chains, reducing waste and using resources efficiently may be one critical benefit to adopting AI, said Walid Ali, an AI in manufacturing expert at Microsoft.
“With industrial processes accounting for almost half of our human species’ consumption of energy and one-fifth of global greenhouse gas emissions, it is the ethical thing to do, as well as the right business decision with certain economies driving us to collaborate on closed-resource loops and the life cycle of products, post production into consumption,” he said.
“This is a time of unprecedented opportunities that allows us to do the right thing from a technology standpoint with AI and smart manufacturing, as well as for sustainability opportunities in the environment we live in.”