Vibration Event Recognition Using SST-Based Φ-OTDR System.

Area of Science:

• Optical Engineering
• Signal Processing
• Machine Learning

Background:

• Phase Sensitive Optical Time-Domain Reflectometry (Φ-OTDR) systems are crucial for detecting physical events.
• Accurate vibration event analysis and identification are essential for robust Φ-OTDR performance.
• Existing methods like Continuous Wavelet Transform (CWT) and Short-Time Fourier Transform (STFT) have limitations in time-frequency resolution and phase information extraction.

Purpose of the Study:

• To propose and evaluate a novel Synchrosqueezing Transform (SST) based method for vibration event analysis in Φ-OTDR.
• To enhance the time-frequency resolution and phase information for distinguishing subtle vibration events.
• To assess the performance of deep learning classifiers (VGG, ViT, ResNet) when applied to SST-transformed Φ-OTDR data.

Main Methods:

• Application of Synchrosqueezing Transform (SST) to Φ-OTDR signals to improve time-frequency representation.
• Utilizing deep learning models including Visual Geometry Group (VGG), Vision Transformer (ViT), and Residual Network (ResNet) for classification.
• Experimental validation using diverse vibration events with varying intensities and attenuation levels.

Main Results:

• The proposed SST method demonstrates superior performance compared to CWT and STFT based approaches.
• Residual Network (ResNet) emerged as the most effective deep learning classifier for the analyzed data.
• High recognition rates were achieved across different signal strengths, event types, and attenuation conditions.

Conclusions:

• The SST-based method offers significant advantages for vibration event analysis and identification in Φ-OTDR systems.
• The integration of SST with deep learning classifiers, particularly ResNet, provides a powerful tool for robust Φ-OTDR monitoring.
• This approach holds considerable value for enhancing the reliability and accuracy of Φ-OTDR applications.