A Convolutional Autoencoder Based Fault Diagnosis Method for a Hydraulic Solenoid Valve Considering Unknown Faults

Area of Science:

• Mechanical Engineering
• Electrical Engineering
• Data Science

Background:

• Hydraulic solenoid valves are critical electromechanical components in industrial systems.
• Valve failures, caused by electrical issues, wear, contamination, or assembly errors, lead to downtime and safety hazards.

Purpose of the Study:

• To develop a nondestructive diagnostic technology for rapid and accurate identification of hydraulic solenoid valve failures.
• To mitigate system downtime and safety risks associated with valve malfunctions.

Main Methods:

• A data-driven model utilizing voltage and current signals from normal and faulty valves.
• Implementation of a convolutional autoencoder for feature extraction.
• Application of hypersphere-based clustering for fault classification and identification.

Main Results:

• The diagnostic model achieved a 98% accuracy rate in classifying seven distinct fault modes.
• The approach successfully identified both known and potentially new fault types.
• Data augmentation with white noise demonstrated the model's robustness.

Conclusions:

• The proposed diagnostic method offers an effective solution for accurate and prompt fault identification in hydraulic solenoid valves.
• This technology can significantly improve the maintenance and operational safety of hydraulic systems.