Fault Classification for Cooling System of Hydraulic Machinery Using AI

Area of Science:

• Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Hydraulic systems are crucial in various industries like manufacturing and robotics.
• Faults in hydraulic systems can lead to significant operational downtime.
• Artificial Intelligence (AI) offers potential for fault prediction and classification.

Purpose of the Study:

• To propose a novel AI-based approach for classifying cooling system behavior in a hydraulic test rig.
• To investigate the effectiveness of deep learning models for hydraulic fault diagnosis.

Main Methods:

• Generated spectrograms from time-series data of a hydraulic test rig under three distinct fault conditions.
• Employed a Convolutional Neural Network (CNN) variant, specifically a Residual Network (ResNet-18), for fault classification.
• Extracted performance metrics such as F-score, precision, accuracy, and recall using a Confusion Matrix.

Main Results:

• The ResNet-18 model demonstrated a high classification accuracy, approaching 95%, after rigorous testing, validation, and training.
• The model successfully classified three different fault conditions in the hydraulic system's cooling behavior.
• Performance evaluation confirmed the model's efficacy in distinguishing between normal and faulty operational states.

Conclusions:

• The developed AI-based approach, utilizing ResNet-18, is highly effective for classifying hydraulic system cooling faults.
• This method can significantly contribute to preventing operational downtime and ensuring the sustainable operation of hydraulic systems.
• The study highlights the potential of deep learning for advanced fault diagnosis in industrial hydraulic equipment.