IMC Presentation (43:58)

Cris Crosby, Aveva

When it comes to everyday challenges in operations and maintenance, like improving PM (planned maintenance) completion rate, AI/ML (artificial intelligence/machine learning) is often relegated to experimental solutions to justify “big data” initiatives that eventually fail to deliver real value.

AI/ML can help but only when you resist the hype and discard the notion that predictive maintenance is the only analytics-based use of IoT data or that you need prolonged PoCs (proof-of-concept) to validate the AI/ML approach. The oft-heard “just give me all your data, let’s put it in a data lake and we’ll figure it out…” fails to recognize and incorporate existing knowledge about your plant-floor machinery and process - leading and lagging indicators for equipment failure, signal vs. noise in the data, P-F curve, machine-condition data vs. process data, and others.

What works is a layered, fit-for-purpose approach to analytics – simple, automated engineered-analytics (first principles, SME (subject-matter-expert) heuristics, basic math, STATS101 including simple predictive) combined with AI/ML analytics where it is a must for its unique capabilities and multivariate models. Even here, purpose-built proven off-the-shelf AI solutions ease operationalizing and enterprise-wide scaling for improved equipment reliability and optimized maintenance practices. Join us for this session as we walk-through the steps needed to transform AI/ML into an operationalized solution.

We will discuss success stories from different industries where customers have benefited from a layered, fit-for-purpose analytics and a sober mix of “usage-based”, “condition-based”, “simple-predictive” and “advanced-predictive” maintenance. Several customers have also been successful in implementing their own centralized monitoring and diagnostic centers to improve asset reliability without elaborate DevOps/MLOps cycles.