A Study on Pipeline Leakage Detection Using Bayesian Deep Learning

Zhou, Taotao; Lan, Fengyi; Li, Yuntao; Ye, Yingchun; Zhang, Laibin

Permalink

https://hdl.handle.net/2142/121802 Copy

Description

Title

A Study on Pipeline Leakage Detection Using Bayesian Deep Learning

Author(s)

• Zhou, Taotao
• Lan, Fengyi
• Li, Yuntao
• Ye, Yingchun
• Zhang, Laibin

Issue Date

2023

Keyword(s)

• Pipeline leakage
• Leakage detection
• Bayesian deep learning
• Oil and gas

Abstract

Pipelines act a key role in assuring the safe and economical transportation of oil and gas. However, pipelines are typically operated under harsh working conditions and are hence vulnerable to leakage owing to a variety of causes, such as corrosion, fatigue, and third-party interference. Moreover, the consequences of pipeline leakage may lead to serious production loss, environmental pollution, and even injuries. Therefore, an effective pipeline leakage detection method is of great importance to maintain the integrity of the oil and gas transportation systems. The objective of this paper is threefold. First, we will provide a comprehensive review of pipeline leakage detection methods involving various types of sensors and machine learning techniques in the oil and gas industry. Second, we will develop a Bayesian deep learning-based method to effectively identify the pipeline leakage and meanwhile characterize the predictive uncertainty. This can present distinct advantages against the Frequentist deep learning-based pipeline leakage detection methods, which would present over- confident predictions and lead to errors in the decision-making process. Third, we will demonstrate the proposed method using real-field pipeline monitoring data, and a comparative study between Bayesian deep learning and its Frequentist counterp

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121802

Owning Collections