Real-time monitoring of CO2 transport pipelines using deep learning DOI Creative Commons
Juhyun Kim, Hyunjee Yoon, Saebom Hwang

и другие.

Process Safety and Environmental Protection, Год журнала: 2023, Номер 181, С. 480 - 492

Опубликована: Ноя. 17, 2023

Real-time pipeline monitoring is important for the safe transportation of captured CO2. A dynamic modeling method, which one methods, can provide reliable diagnostic results various anomalies. In anomalies are detected by comparing predictions and observations variables. However, licensing costs associated with use flow simulators that provides high. this study, we developed a real-time deep-learning-based method save cost simulators. The obtained using deep-learning models where simulator required only in training step. Two improvements were made to enhance both prediction anomaly detection accuracies. First, accuracy variables be improved considering delay time interval between inlet outlet points pairing input output data. Second, also conditionally choosing based on normal operation ranges observations. As part field demonstration, proposed was applied CO2 transport located Donghae-1 gas field. showed more than 25%.

Язык: Английский

On the mixed acoustic and vibration sensors for the cross-correlation analysis of pipe leakage signals DOI
Xiwang Cui, Yan Gao, Xiaojuan Han

и другие.

Applied Acoustics, Год журнала: 2023, Номер 216, С. 109798 - 109798

Опубликована: Дек. 9, 2023

Язык: Английский

Процитировано

15
Roadmap for Recommended Guidelines of Leak Detection of Subsea Pipelines DOI Creative Commons
Ahmed Reda, Ramy Magdy A. Mahmoud, Mohamed A. Shahin

и другие.

Journal of Marine Science and Engineering, Год журнала: 2024, Номер 12(4), С. 675 - 675

Опубликована: Апрель 18, 2024

The leak of hydrocarbon-carrying pipelines represents a serious incident, and if it is in gas line, the economic exposure would be significant due to high cost lost or deferred hydrocarbon production. In addition, leakage could pose risks human life, have an impact on environment, cause image loss for operating company. Pipelines are designed operate at full capacity under steady-state flow conditions. Normal operations may involve day-to-day transients such as pumps, valves, changes production/delivery rates. basic detection problem distinguish between normal operational occurrence non-typical process conditions that indicate leak. To date, industry has concentrated single-phase flow, primarily oil, gas, ethylene. application leak-monitoring system particular pipeline depends environmental issues, regulatory imperatives, prevention company, safety policy rather than pipe size configuration. This paper provides review recommended guidance subsea context integrity management. also presents capability various techniques can used offer roadmap potential users systems.

Язык: Английский

Процитировано

6
Spatio-Temporal Feature Extraction for Pipeline Leak Detection in Smart Cities Using Acoustic Emission Signals: A One-Dimensional Hybrid Convolutional Neural Network–Long Short-Term Memory Approach DOI Creative Commons
Saif Ullah,

N. Ullah,

Muhammad Siddique

и другие.

Applied Sciences, Год журнала: 2024, Номер 14(22), С. 10339 - 10339

Опубликована: Ноя. 10, 2024

Pipeline leakage represents a critical challenge in smart cities and various industries, leading to severe economic, environmental, safety consequences. Early detection of leaks is essential for overcoming these risks ensuring the safe operation pipeline systems. In this study, hybrid convolutional neural network–long short-term memory (CNN-LSTM) model leak that uses acoustic emission signals was designed. model, are initially preprocessed using Savitzky–Golay filter reduce noise. The filtered input into where spatial features extracted CNN. then passed an LSTM network, which extracts temporal from signals. Based on features, presence or absence determined. performance proposed compared with two alternative approaches: method employs combined time domain bidirectional gated recurrent unit model. approach demonstrated superior performance, as evidenced by lower validation loss, higher accuracy, enhanced confusion matrices, improved t-distributed stochastic neighbor embedding plots other models when tested industrial data. findings indicate more effective accurately detecting leaks, offering promising solution enhancing safety.

Язык: Английский

Процитировано

5
Natural Gas Induced Vegetation Stress Identification and Discrimination from Hyperspectral Imaging for Pipeline Leakage Detection DOI Creative Commons
Pengfei Ma, Ying Zhuo, Genda Chen

и другие.

Remote Sensing, Год журнала: 2024, Номер 16(6), С. 1029 - 1029

Опубликована: Март 14, 2024

Remote sensing detection of natural gas leaks remains challenging when using ground vegetation stress to detect underground pipeline leaks. Other stressors may co-present and complicate leak detection. This study explores the feasibility identifying distinguishing gas-induced from other stresses by analyzing hyperspectral reflectance vegetation. The effectiveness this discrimination is assessed across three distinct spectral ranges (VNIR, SWIR, Full spectra). Greenhouse experiments subjected plant species controlled environmental stressors, including leakage, salinity impact, heavy-metal contamination, drought exposure. Spectral curves obtained underwent preprocessing techniques such as standard normal variate, first-order derivative, second-order derivative. Principal component analysis was then employed reduce dimensionality in feature space, facilitating input for linear/quadratic discriminant (LDA/QDA) identify discriminate Results demonstrate an average accuracy 80% gas-stressed plants unstressed ones LDA. Gas leakage can be discriminated scenarios involving a single distracting stressor with ranging 76.4% 84.6%, treatment proving most successful. Notably, derivative processing VNIR spectra yields highest

Язык: Английский

Процитировано

4
Real-time monitoring of CO2 transport pipelines using deep learning DOI Creative Commons
Juhyun Kim, Hyunjee Yoon, Saebom Hwang

и другие.

Process Safety and Environmental Protection, Год журнала: 2023, Номер 181, С. 480 - 492

Опубликована: Ноя. 17, 2023

Real-time pipeline monitoring is important for the safe transportation of captured CO2. A dynamic modeling method, which one methods, can provide reliable diagnostic results various anomalies. In anomalies are detected by comparing predictions and observations variables. However, licensing costs associated with use flow simulators that provides high. this study, we developed a real-time deep-learning-based method save cost simulators. The obtained using deep-learning models where simulator required only in training step. Two improvements were made to enhance both prediction anomaly detection accuracies. First, accuracy variables be improved considering delay time interval between inlet outlet points pairing input output data. Second, also conditionally choosing based on normal operation ranges observations. As part field demonstration, proposed was applied CO2 transport located Donghae-1 gas field. showed more than 25%.

Язык: Английский

Процитировано

9