A Fault Diagnosis Methodology for Gear Pump Based on EEMD and Bayesian Network

Area of Science:

• Mechanical Engineering
• Condition Monitoring
• Reliability Engineering

Background:

• Gear pumps are critical components in many industrial systems.
• Traditional fault diagnosis methods often rely solely on sensor data, limiting their effectiveness.
• Integrating diverse information sources can enhance diagnostic accuracy.

Purpose of the Study:

• To develop an advanced fault diagnosis methodology for gear pumps.
• To improve diagnostic accuracy and capacity by fusing multi-source information.
• To leverage ensemble empirical mode decomposition (EEMD) and Bayesian networks for robust fault detection.

Main Methods:

• Ensemble Empirical Mode Decomposition (EEMD) for signal processing.
• Bayesian network construction with fault, fault feature, and multi-source information layers.
• Fusion of vibration signal features (IMF energy) with qualitative data (visual inspection, maintenance records).

Main Results:

• The proposed EEMD-Bayesian network model demonstrated superior diagnostic performance compared to artificial neural networks and support vector machines when using sensor data alone.
• Incorporating information from human observation and maintenance records significantly aided fault diagnosis.
• The methodology proved effective and efficient in diagnosing gear pump faults, even with uncertain or incomplete data.

Conclusions:

• The multi-source information fusion approach enhances gear pump fault diagnosis accuracy.
• Bayesian networks provide a robust framework for integrating diverse data types in fault diagnosis.
• This methodology offers a promising solution for reliable condition monitoring of rotating machinery.