Predictive Maintenance

AI monitors equipment condition data to flag likely failures early so maintenance can be scheduled before breakdowns interrupt operations.

Business impact

• Unplanned maintenance — Wärtsilä says its predictive-maintenance service can cut unplanned maintenance by as much as 50%.
• Advance warning before breakdown — CNH Industrial reported advance warning of 30 minutes to 3 hours; Nissan reported 2 weeks to 6 months.
• Sensitivity — The CNH Industrial pilot reported sensitivity higher than 90%.
• Specificity — The CNH Industrial pilot reported specificity higher than 90%.
• Return on investment — Nissan reported ROI under 3 months.
• Production downtime avoided — Stanley Black & Decker reported about 4 production hours of downtime avoided.
• Savings — Stellantis reported around $2 million in savings during the first 50 days.
• Accuracy in identifying low production wells — Origin Energy reported about 80% accuracy.
• Run-life of wells — Origin Energy reported a simulated increase of 300+ days.

Data requirements

• Sensor data (sensor) — Used to monitor equipment, passenger facilities, or plant conditions.
• Telemetry (tabular) — Used especially in engine and vehicle monitoring cases.
• PLC data (tabular) — Used in the CNH Industrial pilot.
• Manufacturing execution data (tabular) — Used in the Rockwell Automation case.
• Images (image) — Used in BMW and airport-related examples.
• Diagnostic data (tabular) — Used to assess equipment condition and anomalies.
• Historical operational data (tabular) — Used to train models and predict demand or failure patterns in some deployments.

AI methods and techniques

• Anomaly detection — Used to spot unusual equipment behavior and flag potential failures.
• Predictive maintenance modeling — Used to estimate when equipment may fail or need intervention.
• Machine learning — Used across several cases to learn failure patterns from historical and live operational data.
• Computer vision — Used in BMW's case to analyze visual and other production data for failure prediction.
• Advanced diagnostics — Used with AI to interpret equipment condition and support proactive response.
• Digital twin modeling — Used in Rolls-Royce to assess engine condition alongside telemetry and analytics.

Industries

Real-world evidence

20 documented case studies on record.

Companies using this: BMW, CNH Industrial, Danone, GM, Infosys, Nestlé, Nissan, Nokia, Rivian, Rockwell Automation, Rolls-Royce, Schneider Electric, Stanley Black & Decker, Stellantis, Dubai Airports and 5 more.