Machine Learning Methods for Pipeline Surveillance Systems Based on Distributed Acoustic Sensing: A Review

Javier Tejedor,Juan Pastor-Graells,Javier Macias-Guarasa,Sonia Martin-Lopez,Hugo Martins,Pedro Corredera

doi:10.3390/app7080841

Abstract

There is an increasing interest in researchers and companies on the combination of Distributed Acoustic Sensing (DAS) and a Pattern Recognition System (PRS) to detect and classify potentially dangerous events that occur in areas above fiber optic cables deployed along active pipelines, aiming to construct pipeline surveillance systems. This paper presents a review of the literature in what respect to machine learning techniques applied to pipeline surveillance systems based on DAS+PRS (although its scope can also be extended to any other environment in which DAS+PRS strategies are to be used). To do so, we describe the fundamentals of the machine learning approaches when applied to DAS systems, and also do a detailed literature review of the main contributions on this topic. Additionally, this paper addresses the most common issues related to real field deployment and evaluation of DAS+PRS for pipeline threat monitoring, and intends to provide useful insights and recommendations in what respect to the design of such systems. The literature review concludes that a real field deployment of a PRS based on DAS technology is still a challenging area of research, far from being fully solved.

Highlights

The most sustainable and safest transmission method to transport energy sources from the producing facilities to the various end-users relies on pipeline transmission
This paper has presented a review of the main approaches for the application of machine learning techniques in pipeline surveillance systems based on distributed acoustic sensing
We have first addressed a general review of related work, concluding that, in general, there is a lack of understanding of the adequate and rigorous methodology to apply in the design and evaluation of Distributed Acoustic Sensing (DAS)+Pattern Recognition System (PRS) strategies

Summary

Introduction

The most sustainable and safest transmission method to transport energy sources from the producing facilities to the various end-users relies on pipeline transmission. Despite all safeguard measures taken by the system operators, zero risk does not exist, being the energy transmission an industrial activity, so that extra care must be taken to avoid the pipeline to be damaged This is especially important if we take into account that most incidents involving natural gas transmission infrastructures occur due to external interference (between 50% and 60% of the reported incidents according to [1,2], well above the second main cause (construction defect or material failure, which account for between 16% and 25% of the cases, respectively)), mainly due to third party works in the pipeline vicinity, some of which lead to human casualties. This set of data is typically employed to build statistical models that can be used to make predictions.

Objectives

Methods

Results

Discussion

Conclusion