Jaxson O’Brien, maintenance supervisor at CHS Laurel Refinery in Montana, told attendees at the 2025 AFPM Summit about a compressor problem that initially seemed minor but ultimately required a lengthy overhaul when hidden damage was uncovered.
He explained that his team relied heavily on vibration and performance monitoring data, which suggested the machine was operating normally. Because of that, they assumed they were being proactive when, in fact, deeper inspection might have revealed issues earlier. O’Brien said the experience forced him to reflect on how easily maintenance teams can overvalue technology while undervaluing hands-on checks and traditional practices.
The situation illustrated what O’Brien calls a breakdown in maintenance "situational awareness" — a failure to balance tribal knowledge and hands-on inspection with emerging tools like condition-based monitoring and AI. "As technology increases, so does vulnerability," he said. "We can rely too much in one area of technology. The truth can be uncomfortable."
O’Brien’s fellow presenter, Mark Sannes, analytics and monitoring leader with Flint Hills Resources in Corpus Christi, Texas, expanded on the theme by describing the uncertainty companies face when they embark on digital transformation. "We knew the mountain was over there. We knew it was generally in that direction. We also knew we couldn’t see very far," Sannes said.
He emphasized that while uncertainty is part of the journey, digital tools excel at handling the flood of information that overwhelms human teams. "The computer doesn’t get tired of looking at mountains of data. Take away the mundane so that we can be freed up to be creative, to solve problems."
Flint Hills, CHS: Predictive maintenance requires culture, not just tech
Mark Sannes with Flint Hills Resources and Jaxson O’Brien with CHS Inc. share insights during the 2025 AFPM Summit held at the Gaylord Texas Resort & Convention Center in Grapevine, Texas.
Sannes said Flint Hills built its anomaly detection capability atop a foundation of maintenance culture reform. "It’s going to get busier before it gets calmer," he said. He credited prior gains in reducing backlog and improving cycle times for enabling the team to absorb the new workload from predictive insights.
Still, even with dedicated teams and strong leadership support, early wins didn’t come easy. "I’ve been a huge proponent of root cause analysis as a key part of the reliability toolbox, so we’ve incorporated that into monitoring," Sannes said. "Sometimes we’ll find that we didn’t have a model on it, but we can add one. Or we have a model on it, but the bounds were too sloppy. Let’s go tune it up."
That constant feedback loop between data, diagnostics and field action is essential, O’Brien said. "You start that journey with the people you need to really have buy-in from, from the ground up," he added. "Without the operations and maintenance teams having bought into it … it’s really hard to push the agenda."
Audience members raised practical challenges about how predictive maintenance fits into daily work. One point of discussion centered on deciding whether newly detected anomalies should go first to frontline craftspeople or to engineers for review, with different refineries experimenting with both approaches to see what works best.
Ultimately, the panelists emphasized that predictive maintenance isn’t a plug-and-play solution. It’s a cultural shift that demands investment, adaptation and humility.
"Don’t underestimate the technology part," Sannes said. "But the technology is a small part of the tip of the iceberg. Another 70% is people, work process and culture."
