Simulation-Guided Interpretable Fault Diagnosis of Hydraulic Directional Control Valves Under Limited Fault Data...

Area of Science:

• Mechanical Engineering
• Control Systems
• Fault Diagnosis

Background:

• Delayed switching faults in hydraulic directional control valves (DCVs) degrade system performance.
• Existing data-driven methods lack interpretability and struggle with limited industrial fault data.

Purpose of the Study:

• To propose an interpretable fault diagnosis framework for DCVs with delayed switching faults.
• To address challenges of limited fault samples and insufficient physical interpretability in industrial settings.

Main Methods:

• Developed a simulation-guided feature construction method using physical simulation models.
• Analyzed fault mechanisms to derive diagnostic features from inter-sensor pressure differences.
• Evaluated features using classical machine learning classifiers (RF, SVM, KNN, LR) with limited fault samples.

Main Results:

• The proposed method effectively identifies structural imbalances caused by internal valve faults.
• Achieved high diagnostic accuracy and robustness compared to single-sensor and black-box models.
• Demonstrated successful fault diagnosis under small-sample conditions.

Conclusions:

• The framework offers a practical, physically interpretable solution for hydraulic valve fault diagnosis.
• Provides potential value for industrial quality inspection and maintenance.
• Overcomes limitations of traditional methods in small-sample, interpretable fault diagnosis.