Sparse decomposition method based on time-frequency spectrum segmentation for fault signals in rotating machinery

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Impulse signals from damaged rotating machinery are often sparse, weak, coupled, and non-periodic, complicating fault diagnosis.
• Sparse Decomposition (SD) is effective for complex signal description but suffers from slow atom pursuit speed.
• Existing time-frequency analysis methods like Generalized S Transform (GST) have resolution limitations affected by noise.

Purpose of the Study:

• To develop an efficient and precise method for extracting complex fault signals from large rotating machinery.
• To overcome the limitations of traditional Sparse Decomposition (SD) in terms of speed and accuracy.
• To enhance the analysis of vibration signals for improved machinery fault diagnosis.

Main Methods:

• Introduced Sparse Decomposition based on Time-Frequency Spectrum Segmentation (SD-TFSS).
• Utilized Multiresolution Generalized S-transform (MGST) to generate multi-resolution time-frequency spectrums, mitigating noise-induced resolution issues.
• Implemented spectrum fusion, segmentation, and optimal atom library construction for signal representation.

Main Results:

• The proposed SD-TFSS method demonstrated lower pursuit complexity compared to traditional SD.
• Achieved higher decomposition efficiency and better approximation precision in simulations and experiments.
• Successfully extracted complex impulse signals characteristic of machinery faults.

Conclusions:

• The SD-TFSS method offers a significant improvement for machinery fault diagnosis by efficiently extracting complex signals.
• The approach enhances the speed and accuracy of sparse decomposition in analyzing vibration data.
• This technique provides a robust solution for identifying faults in large rotating machinery, even under intermittent operation.