Leading Fertilizer Company Improves Asset Uptime with AI

Challenges

A leading fertilizer company operates the world’s largest single-site export of urea and produces 5.6 million tons (MT) of urea and 3.8 MT of ammonia annually. With 6 world-class plants, the company aims to become the world’s largest urea producer by 2030. With a commitment to safe and efficient asset operations, the company is using AI to predict and prevent failures: transforming its operations and condition maintenance strategy to achieve goals of increased asset availability and reduced maintenance costs.

Prior to partnering with BakerHughesC3.ai (BHC3), the company took a predominantly reactive approach to asset maintenance. However, due to an aging fleet, these assets frequently experienced unplanned downtime, forcing the maintenance crews to perform costly emergency repairs. Unplanned downtime also introduced significant risks to production continuity. As a result, the company faced reliability and performance losses 60% higher than their operational target. To improve asset uptime and reliability, the company needed an innovative solution that could integrate data from various sources, focus attention on the relevant information, predict potential failures, and provide actionable insights.

Approach

Over a 24-week period, the BHC3 team partnered with the fertilizer company to configure BHC3 Reliability to enable predictive monitoring for 27 production assets across 4 plants.

First, the team identified high-priority rotating equipment across 4 key asset types (e.g., compressors, turbines) and the relevant data sources. The team integrated over 5 years of historical data along with live sensor data from Baker Hughes Cordant Platform to create a unified, federated data model of the asset fleet. During data ingestion, the platform’s normalization engine improved data quality by imputing missing data and reordering time series data.

Using the unified data image, the BHC3 team configured machine learning models to predict system failures and detect anomalies for the 4 key asset types. BHC3 machine experts guided the model development process by identifying significant signals and anomalies that were incorporated into each asset class’s anomaly detection models.

To ensure high user adoption and value, BHC3 conducted a series of training workshops, provided documentation, enabled executive visibility, and implemented change management strategies.

Today, the condition monitoring and asset reliability team leverages AI-based performance metrics from BHC3 Reliability alongside rule-based analytics for dynamic decision making.

About the Company

• Largest single-site exporter of urea
• 6 world-class plants
• 5.6 MT of urea and 3.8 MT of ammonia produced annually

Project Objectives

• Enable predictive monitoring to improve asset uptime of critical equipment such as compressors and turbines
• Integrate and unify data from available sources, including historical and live data
• Apply advanced AI to predict potential failures and enable early warnings for 27 assets across 4 plants
• Configure the BHC3 Reliability user interface and expose actionable AI insights
• Deploy the sensor health module within BHC3 Reliability to monitor sensors in near real-time

Project Highlights

• 27 assets across 4 plants
• 1.3B historical records integrated
• 3.5 million incremental records from 2,400 sensors ingested daily
• 233 ML models in production
• BHC3 Reliability user interface configured
• Co-deployed with Baker Hughes Cordant Platform
• Planning to scale to additional 92 assets
• 10 users onboarded

Solution Architecture