Enterprise AI for Predicting HVAC Chiller Failures

• Analytics & Modeling - Predictive Analytics

• Equipment & Machinery

• Predictive Quality Analytics

• Data Science Services

• Load and cluster sensor data for use in a predictive model
• Train a machine learning model to predict chiller failure
• Demonstrate speed of development and deployment by completing project in < 1 week

Not Disclosed

• Global equipment manufacturer and services company
• 100,000 employees
• $30 billion in revenue

• Configured C3 AI Reliability in 4 days
• Loaded 3 years of sensor data for 165 HVAC chillers (40-50 sensor feeds per chiller)
• Developed 163 analytics as inputs for failure prediction algorithm
• Trained and tuned a machine learning model to predict chiller failures with 73% precision and 71% recall

• [Data Management - Data Analysis]
  

  The manufacturer deployed C3 AI Reliability for 165 of its chillers to address the objectives. The customer selected C3 AI for its proven ability to rapidly integrate sensor data, normalize and cluster disparate readings, and run machine learning algorithms to identify deteriorating conditions before failures occur.

• $178M annual benefit identified. 73% model precision. 71% model recall

• In 4 days, C3 AI and the customer loaded, normalized, and mapped 3 years of sensor data for all 165 chillers, created custom analytics on these data, and configured a machine learning algorithm to predict chiller failure events