Research on Twin Extreme Learning Fault Diagnosis Method Based on Multi-Scale Weighted Permutation Entropy

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Check valve fault diagnosis in high-pressure diaphragm pumps is challenging due to non-stationary and non-linear vibration signals.
• Traditional methods like MPE and ELM struggle with accurate fault feature extraction and classification.
• Existing methods lack reliability in identifying complex fault characteristics.

Purpose of the Study:

• To develop an accurate and reliable fault diagnosis method for check valves in high-pressure diaphragm pumps.
• To enhance fault feature extraction from complex vibration signals.
• To improve the classification accuracy of fault diagnosis models.

Main Methods:

• Utilized Multi-scale Weighted Permutation Entropy (MWPE) to extract enhanced multi-scale fault and arrangement pattern features from vibration signals.
• Developed a fault diagnosis model using the Twin Extreme Learning Machine (TELM).
• The TELM model was designed to find non-parallel classification hyperplanes for improved applicability.

Main Results:

• The proposed MWPE method effectively enhances fault features by combining amplitude and arrangement pattern information.
• The MWPE-based TELM model demonstrated superior performance in identifying check valve fault states.
• Achieved a high fault diagnosis accuracy rate of 97.222%.

Conclusions:

• The combined MWPE and TELM approach offers a robust solution for check valve fault diagnosis in high-pressure diaphragm pumps.
• This method overcomes limitations of traditional techniques in characterizing complex vibration signal dynamics.
• The study validates the effectiveness and high accuracy of the proposed diagnostic model.