Sensor-based performance analysis and intelligent fault diagnosis of pneumatic-hydraulic actuated valves

Pneumatic-hydraulic actuated ball valves are critical components in pipeline rupture protection and emergency shutdown systems for large-diameter natural gas transmission pipelines. Traditional maintenance methods, relying on periodic inspections and reactive maintenance, often result in delayed fault detection, which increases safety risks. To address this challenge, this study proposes an innovative intelligent fault diagnosis approach that significantly enhances early fault detection and predictive maintenance, without requiring structural modifications to the valve. The main contributions of this work are: (1) the development of a novel time-frequency analysis-based feature extraction method for current and pressure signals, which improves fault signature reliability and distinguishes fault patterns more effectively; and (2) the application of long short-term memory (LSTM) networks optimized using an improved particle swarm optimization (IPSO) algorithm, achieving 100% diagnostic accuracy for both solenoid valve and mechanical faults. Experimental results demonstrate the superiority of the proposed approach, with field tests successfully detecting torque anomalies and identifying risks related to solenoid valve power-off logic. This approach provides a robust solution for transitioning from preventive to predictive maintenance, significantly improving operational safety and pipeline reliability.