Shell scales global predictive maintenance across critical equipment
Shell implemented an AI-driven predictive maintenance system across more than 10,000 pieces of equipment spanning its global operations. Leveraging Microsoft Azure, Azure Databricks, and C3 AI, Shell ingests over 20 billion rows of sensor data weekly, running more than 10,000 machine learning models in production. The solution detects anomalies, extends asset life, ensures environmental and human safety, and integrates with digital twin technology for comprehensive asset management. Shell's journey included significant technical and organizational challenges, ultimately creating a scalable, robust solution embedded in workflows. Shell has now made this solution commercially available via the Open AI Energy Initiative, in partnership with C3 AI, Baker Hughes, and Microsoft. The deployment demonstrates real cost reduction, improved equipment uptime, operational efficiency, and a positive environmental impact. Future plans involve extending predictive maintenance to renewables and broader use cases for AI within Shell's businesses.
- Organization
- Shell
- Industry
- Energy & Utilities
- Location
- Global
- Published
- May 2025
Reported outcomes
Strategic outcomes
Primary read
Use case focus
Showing 3 of 3
- 1Global Predictive Maintenance for Industrial Equipment
- 2Automated Anomaly Detection in Oil & Gas Assets
- 3AI-driven Equipment Health Monitoring
- Shell needed to monitor and maintain over 10,000 pieces of critical equipment worldwide.
- Manual or reactive maintenance resulted in unplanned downtime, high costs, and operational inefficiencies.
- There was a need to process and analyze massive volumes of sensor data to detect equipment anomalies proactively.
- Ensuring environmental and human safety with timely detection and prevention of failures was a priority.
- Building an internal scalable AI platform was deemed too resource-intensive.
- Deployed C3 AI running on Microsoft Azure as the scalable core platform.
- Used Azure Databricks for processing large sensor data pipelines.
- Developed and operationalized over 10,000 machine learning models for equipment anomaly detection.
- Integrated AI predictions into digital twin systems for advanced asset management.
- Embedded predictive maintenance capabilities into workflows company-wide, leveraging global community and governance.
- More than 10,000 pieces of equipment now benefit from predictive maintenance.
- 20 billion sensor data rows processed per week across 3 million data streams.
- Over 15 million predictions generated daily, significantly reducing unplanned downtime.
- Substantial reductions in operating costs and improvements in safety and environmental outcomes.
- AI solution now available industry-wide through a commercial initiative.
Architecture
Shell's solution leverages C3 AI running on Microsoft Azure, ingesting data from the Shell Sensor Intelligence Platform, which is based on Delta lake (Azure Databricks). Over 10,000 production ML models analyze data from global assets (including upstream and downstream sites), with anomaly detection predictions integrated into digital twin platforms for visualization and advanced asset management. The models, embedded in operational workflows, generate 15 million predictions per day and scale across Shell's entire asset ecosystem.
Implementation partners2
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?