As experienced engineers retire and operational demands grow, refineries are turning to AI to bridge the gap. In this session, Tim shares how anomaly detection is being used to surface early signs of equipment issues, enabling proactive maintenance, reducing unplanned downtime and supporting a shift towards predictive maintenance.

Using a dynamic risk analyzer, the team used a hands-off AI tool that automates data cleansing, modeling, and real-time monitoring. It flags subtle shifts in process behavior, such as changes in reactor temperature slopes, that may signal emerging equipment issues. By surfacing these early indicators, engineers can plan interventions more effectively and reduce the risk of costly failures.

The session also explores how to simplify model building and analysis, enabling engineers to extract insights from thousands of data points without writing code. While the pilot has shown promising results, human oversight remains critical, with a “trust but verify” approach guiding the path toward enterprise-wide adoption.