Creating an automated chiller fault detection and diagnostics tool using a data fault library

Area of Science:

• Mechanical Engineering
• HVAC Systems
• Artificial Intelligence

Background:

• Vapor compression refrigeration cycle (VCRC) equipment requires reliable fault detection and diagnosis (FDD) for efficient operation.
• Automated FDD tools are highly desirable for equipment owners and operators to minimize downtime and maintenance costs.

Purpose of the Study:

• To develop and validate a fault detection and diagnostic (FDD) tool for a 70-ton helical rotary, air-cooled chiller.
• To present methodologies for simulating and experimenting with common VCRC faults.
• To demonstrate the effectiveness of a neural network-based FDD approach.

Main Methods:

• Assembled a fault library of empirical data from normal and fault conditions.
• Developed a fault detection and diagnostic (FDD) tool utilizing a neural network (NN) classifier.
• Conducted sophisticated fault experiments simulating air-cooled condenser, refrigerant, and oil-related issues.
• Analyzed thermodynamic and empirical data to assess performance degradation.

Main Results:

• Successfully simulated and documented procedures for air-cooled condenser, refrigerant, and oil-related faults.
• Provided evidence of significant chiller performance degradation under fault conditions.
• The neural network (NN) FDD classifier accurately detected and classified performance degradation.

Conclusions:

• The developed neural network (NN) based fault detection and diagnostic (FDD) tool effectively identifies performance issues in VCRC chillers.
• The experimental procedures offer a valuable guide for simulating and studying VCRC faults.
• Automated FDD systems are vital for maintaining optimal chiller performance and operational efficiency.