A Novel Signal Separation Method Based on Improved Sparse Non-Negative Matrix Factorization.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Vibration Analysis

Background:

• Accurate diagnosis of rotating machinery faults is crucial for operational reliability.
• Separating compound fault features from single-channel vibration signals presents significant challenges.
• Traditional sparse non-negative matrix factorization (SNMF) methods struggle with underdetermined blind source separation.

Purpose of the Study:

• To propose a novel signal separation method for extracting compound fault features from single-channel vibration signals.
• To enhance the performance of SNMF for underdetermined blind source separation using a constraint reference vector.
• To improve the efficiency and accuracy of fault diagnosis in rotating machinery.

Main Methods:

• A novel signal separation method based on improved sparse non-negative matrix factorization (SNMF) is proposed.
• A constraint reference vector, generated by the pulse method and updated via feedback, is introduced into SNMF.
• Time-frequency distribution is used for feature extraction, followed by improved SNMF factorization and envelope analysis for diagnosis.

Main Results:

• The improved SNMF method effectively separates compound fault features from complex vibration signals.
• The proposed method demonstrates superior feature extraction and separation capabilities compared to traditional SNMF.
• Simulation and experimental results confirm the method's effectiveness in reducing dimensionality and improving diagnostic efficiency.

Conclusions:

• The developed improved SNMF method offers an effective solution for compound fault diagnosis in rotating machinery.
• The integration of a constraint reference vector significantly enhances SNMF's performance in underdetermined scenarios.
• This approach provides a more efficient and accurate tool for analyzing vibration signals and identifying machinery faults.