A Hybrid Deep Learning Approach: Integrating Short-Time Fourier Transform and Continuous Wavelet Transform for...

Area of Science:

• Engineering
• Computer Science
• Signal Processing

Background:

• Pipeline integrity is critical for fluid transportation safety.
• Effective leak detection is essential for preventing environmental damage and economic losses.
• Traditional methods often struggle with complex acoustic data and noise.

Purpose of the Study:

• To develop a robust hybrid deep learning model for accurate pipeline leak detection.
• To enhance signal processing using STFT and CWT for improved feature extraction.
• To validate the model's effectiveness in real-time leak identification and classification.

Main Methods:

• A hybrid deep learning approach combining STFT and CWT for acoustic signal analysis.
• Application of Sobel and wavelet denoising filters to enhance signal quality.
• Feature extraction using convolutional neural networks (CNNs).
• Dimensionality reduction via Principal Component Analysis (PCA).
• Leak classification using t-distributed Stochastic Neighbor Embedding (t-SNE) and Artificial Neural Networks (ANNs).

Main Results:

• The hybrid model demonstrated high accuracy and reliability in detecting and classifying pipeline leaks.
• The combined STFT and CWT approach effectively captured both spectral and temporal signal details.
• Feature space dimensionality reduction improved computational efficiency and discriminatory power.

Conclusions:

• The proposed hybrid deep learning model offers a promising solution for real-time pipeline leak detection.
• This approach significantly contributes to advancing pipeline monitoring and maintenance strategies.
• The method shows potential for application in diverse industrial fluid transportation systems.