Deep Learning-Based Prognostics and Health Management Model for Pilot-Operated Cryogenic Safety Valves.

Area of Science:

• Industrial Safety Engineering
• Artificial Intelligence in Engineering
• Mechanical Systems Reliability

Background:

• Modern industries, especially petroleum and LNG, rely heavily on safety valves for system integrity under extreme conditions.
• Traditional empirical methods for safety valve maintenance have limitations in predicting failures.
• There is a growing need for advanced, data-driven approaches to ensure operational safety.

Purpose of the Study:

• To develop and validate a deep learning-based prognostics and health management (PHM) model for safety valves.
• To enhance the reliability and safety of industrial systems through early detection of valve malfunctions.
• To provide a data-driven solution for real-time performance monitoring and anomaly detection.

Main Methods:

• Collection and analysis of sensor data from safety valves.
• Development of a deep learning model for performance monitoring and lifespan prediction.
• Integration of data for real-time anomaly detection algorithms.

Main Results:

• Demonstrated capability of the deep learning model in predicting safety valve lifespan.
• Successful real-time anomaly detection in safety valve performance.
• Validation of the model's effectiveness in preventing potential accidents.

Conclusions:

• The proposed deep learning PHM model significantly improves safety valve reliability and operational safety.
• This data-driven approach offers a robust solution for anomaly detection in industrial settings.
• The research contributes to safer operations of pilot-operated cryogenic safety valves.