Incremental Clustering for Predictive Maintenance in Cryogenics for Radio Astronomy

Area of Science:

• Radio astronomy instrumentation
• Mechanical engineering
• Artificial intelligence

Background:

• Maintaining optimal performance in radio astronomy receivers requires consistent operation of cooling systems.
• Cold heads and compressors are critical components susceptible to mechanical deterioration.
• Monitoring motor power currents can indicate early signs of mechanical issues.

Purpose of the Study:

• To develop an intelligent system for monitoring the health of cold heads in radio astronomy cooling systems.
• To detect and predict mechanical deterioration by analyzing power current anomalies.
• To ensure the reliable and consistent performance of sensitive radio astronomy equipment.

Main Methods:

• Utilized Hall effect sensors to measure motor power currents.
• Developed a microcontroller-based electronic board for data acquisition.
• Implemented an unsupervised artificial intelligence model based on incremental clustering for anomaly detection.
• Trained the model initially on known operational categories, allowing for adaptation to new data and anomalies over time.

Main Results:

• The system successfully detects and predicts anomalies in cold head motor power currents.
• The incremental clustering approach allows the model to adapt to evolving operational conditions and new fault scenarios.
• The system can categorize new anomalies, forming new clusters as needed.
• Demonstrated the potential for precise and reliable long-term health monitoring of cooling system components.

Conclusions:

• The developed AI system provides a robust solution for monitoring the health of radio astronomy cooling systems.
• The unsupervised incremental learning approach enhances adaptability and long-term effectiveness in detecting diverse and evolving anomalies.
• This technology contributes to improved reliability and reduced downtime for critical radio astronomy infrastructure.