Georgia-Pacific is implementing an AI-powered solution known as the Digital Mezzanine across many of its paper converting operations to analyze data and identify actionable insights on how to make meaningful performance improvements.
In today’s fast-paced industrial landscape, identifying meaningful improvements is key to staying ahead of the competition. At Georgia-Pacific, their innovative homegrown tool known as Digital Mezzanine is helping to do just that — transforming raw machine data into actionable insights through the power of AI and machine learning.
The Digital Mezzanine ingests a continuous stream of data from converting equipment across multiple facilities, meticulously analyzing when, why, and for how long a machine enters a fault state. But it’s not just about spotting problems; it’s about understanding which problems can be fixed to generate real value and helping solve those problems more quickly.
Adam Dunigan, creator of Digital Mezzanine, said they call these “compressible opportunities.” By comparing fault data over time and across machines and facilities, the system can identify and differentiate between issues that are part of normal baseline operations and those that exist outside of normal and can be improved. It's like when your smart watch notifies you that you didn't get as much sleep last night as you normally do.
“Prior to Digital Mezzanine, it was all about going after your top losses,” he said. “We’d be banging our heads against the walls trying to fix a systemic problem, like a paper tear. Such problems are unavoidable. You will always see those occur to some extent, so you’re working on a problem that may be impossible or unrealistic to eliminate if you only follow a top losses approach. Digital Mezzanine identifies and prioritizes opportunities with proven value to capture, not just top losses.”
It’s a paradigm shift for many facilities. Instead of focusing on just the problems that cause the most downtime but tend to occur regardless of interventions, Digital Mezzanine zeroes in on those issues that, once addressed, drive measurable improvements in productivity, reliability and safety.
Adam shared one example of the system highlighting a fault that happened at a palletizer – an automated machine that stacks and arranges cases of product onto a pallet for handling, storage and shipping. In this case, the fault was a slip sheet lost by gripper. The AI model looks at the fault, knows what product was running, and analyzes historical data to see what “good” looks like. Should the operators expect five of those faults a day or 50 or never any at all?
Adam said in this case, they shouldn’t really be seeing any of those at all, and in the last four hours, they’ve seen five faults.
“If they continue to see this fault at that rate, they’ll have 80 stop conditions over the next 24 hours,” he said. “So, they’ve got something that’s really bugging them.”
Now that an issue has been recognized and highlighted, a sophisticated large language model leverages a vast repository of machine manuals and documentation to generate tailored troubleshooting steps. These proposed solutions are validated by senior technicians and subject matter experts to ensure safety and feasibility before operators see them.
“The Digital Mezzanine software consolidates all that information and generates a fairly succinct output for us that says, when you see this failure mode start to tick up, here are the steps that you should take to resolve it,” explained Adam.
He said it’s been one of the most successful implementations of AI tools across Georgia-Pacific to date. Georgia-Pacific facilities that have embraced this technology have reported an average 15% improvement in machine efficiency with a limited initial investment.
The platform’s scalability is another key strength. Georgia-Pacific is deploying Digital Mezzanine across as many machines as possible and even sharing this advanced technology with other Koch companies, like Guardian.
Ultimately, the Digital Mezzanine is more than just a tool for loss reduction. It is a paradigm shift — turning mountains of complex data into clear, prioritized pathways for improvement. It helps engineers and operators alike understand not just what is going wrong, but why, and how to fix it safely and efficiently.
“This helps folks understand the problems quickly so they can be addressed when they are small rather than letting them progress into really large problems,” Adam said. “And the technology and our capabilities keep evolving, so we’ve got no shortage of things that we want to go after.”
