Acoustic Signal Recognition of Partial Discharge Optical Fiber Sensors Using Time-Frequency Phase Composition.

Area of Science:

• Electrical Engineering
• Materials Science
• Signal Processing

Background:

• Partial discharge (PD) in insulation systems can lead to equipment failure.
• Acoustic emission (AE) monitoring using optical fiber sensors offers a promising non-destructive testing method.
• Accurate recognition of PD acoustic signals is crucial for effective insulation defect detection.

Purpose of the Study:

• To propose a novel method for recognizing acoustic signals from PD optical fiber sensors.
• To enhance the accuracy and reliability of PD detection through advanced signal processing and machine learning.

Main Methods:

• Utilized Cohen's bilinear time-frequency transformation to obtain the Wigner-Ville distribution of acoustic signals, providing high time-frequency resolution.
• Implemented a time-frequency phase composition property by combining Wigner-Ville distributions at different power cycle phases.
• Employed a Vision Transformer with an attention block for signal identification, focusing on energy concentration areas in acoustic features.

Main Results:

• The proposed method achieved an accuracy of 99.56% in identifying acoustic signals of PD optical fiber sensors across three test sets.
• The Wigner-Ville distribution effectively reflected insulation defect properties influenced by medium dispersion and acoustic propagation.
• The attention block in the Vision Transformer improved the model's focus on critical acoustic signal features.

Conclusions:

• The developed time-frequency phase composition method combined with a Vision Transformer is highly effective for recognizing PD acoustic signals.
• This approach demonstrates significant potential for detecting various insulation defects in electrical equipment using AE feature analysis.
• The high accuracy validates the method's promise for advanced non-destructive testing and condition monitoring applications.