Acoustic Emission-Based Pipeline Leak Detection and Size Identification Using a Customized One-Dimensional DenseNet

Sensors, Journal Year: 2025, Volume and Issue: 25(4), P. 1112 - 1112

Published: Feb. 12, 2025

Effective leak detection and size identification are essential for maintaining the operational safety, integrity, longevity of industrial pipelines. Traditional methods often suffer from high noise sensitivity, limited adaptability to non-stationary signals, excessive computational costs, which limits their feasibility real-time monitoring applications. This study presents a novel acoustic emission (AE)-based pipeline approach, integrating Empirical Wavelet Transform (EWT) adaptive frequency decomposition with customized one-dimensional DenseNet architecture achieve precise classification. The methodology begins EWT-based signal segmentation, isolates meaningful bands enhance leak-related feature extraction. To further improve quality, thresholding denoising techniques applied, filtering out low-amplitude while preserving critical diagnostic information. denoised signals processed using DenseNet-based deep learning model, combines convolutional layers densely connected propagation extract fine-grained temporal dependencies, ensuring accurate classification presence severity. Experimental validation was conducted on real-world AE data collected under controlled non-leak conditions at varying pressure levels. proposed model achieved an exceptional accuracy 99.76%, demonstrating its ability reliably differentiate between normal operation multiple severities. method effectively reduces costs robust performance across diverse operating environments.

Language: Английский

Data-driven reliability evolution prediction of underground pipeline under corrosion

Reliability Engineering & System Safety, Journal Year: 2025, Volume and Issue: 261, P. 111148 - 111148

Published: April 16, 2025

Language: Английский

Gas Pipeline Leak Detection by Integrating Dynamic Modeling and Machine Learning Under the Transient State

Energies, Journal Year: 2024, Volume and Issue: 17(21), P. 5517 - 5517

Published: Nov. 4, 2024

This study focused on developing machine learning models to detect leak size and location in transient state conditions. The model was designed for an onshore methane–hydrogen blending gas pipeline Canada. Base case simulations revealed significant effects mass flow pressure due leaks, with the system taking approximately 6 h reach a steady from made it essential analyze characteristics during state. Trend data pipeline’s inlet outlet were examined, considering location. To better represent over time, method used create two-dimensional images, which then fed into CNN (convolutional neural network) training. model’s accuracy assessed using classification confusion matrix. By refining acquisition process implementing targeted screening procedures, increased 80%. In conclusion, this demonstrates that can enable rapid accurate detection findings are expected complement existing methods support operators making faster, more informed decisions.

Language: Английский