Time-Frequency Distribution Map-Based Convolutional Neural Network (CNN) Model for Underwater Pipeline Leakage
Yingchun Xie1, Yucheng Xiao1, Xuyan Liu1
1College of Engineering, Ocean University of China, Qingdao 266000, China.
Sensors (Basel, Switzerland)
|September 9, 2020
Summary
This study introduces a novel acoustic detection method for underwater pipeline leaks. A Convolutional Neural Network (CNN) effectively identifies leak severity using processed acoustic signals, offering a new approach for subsea systems.
Area of Science:
- Ocean Engineering
- Acoustics
- Signal Processing
Background:
- Underwater pipeline leakage poses significant risks to subsea production systems.
- Effective detection technologies are crucial for operational safety and efficiency.
Purpose of the Study:
- To propose and validate a new acoustic-based method for detecting underwater pipeline leakage.
- To investigate the characteristics of acoustic leak signals and their relationship with pressure and leak size.
Main Methods:
- Collected acoustic leak signals using a hydrophone during pipeline leakage tests.
- Processed acoustic signals using Ensemble Empirical Mode Decomposition (EEMD) and Hilbert-Huang Transform (HHT) to generate time-frequency images.
- Utilized a two-layer Convolutional Neural Network (CNN) for automated leakage detection and severity identification.
Main Results:
- Acoustic radiation noise from leaks exhibits a continuous medium to high-frequency spectrum.
- Increased pipe pressure and leak hole diameter narrow the spectral structure and shift the center towards lower frequencies.
- Pipe pressure significantly impacts noise levels, with a 6-7 dB increase per 0.05 MPa pressure rise.
Conclusions:
- The proposed acoustic detection method, combined with CNN analysis, accurately identifies underwater pipeline leakage.
- This approach offers a promising new solution for pipeline integrity monitoring in subsea engineering applications.
