Principal component analysis and artificial neural network based approach to analysing optical fibre sensors signals

Lewis, Elfed; Sheridan, Cormac; O’Farrell, Marion; King, D.; Flanagan, Constance A.; Lyons, W. Berry; Fitzpatrick, Colin

doi:10.1016/j.sna.2007.02.012

Cited by 39 publications

(22 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For example, in [15,16] the authors first computed the FFT of the acquired signals, and then the power spectral density (PSD) was calculated (the first 12 values of the PSD form each feature vector).…”

Section: Frequency Domain-based Feature Extraction Methodsmentioning

confidence: 99%

“…These methods are typically based on artificial neural networks [12][13][14][15][16]19,27] or support vector machine [20,45,46] models. In [12], a 3-layer backpropagation and momentum backpropagation ANNs is employed, while in [13] the authors use a momentum backpropagation ANN.…”

Section: Discriminative Model-based Methodsmentioning

confidence: 99%

“…In [12], a 3-layer backpropagation and momentum backpropagation ANNs is employed, while in [13] the authors use a momentum backpropagation ANN. The authors of [14][15][16] employed a 3-layer feed forward ANN for pattern classification. In [19,27], 3-layer backpropagation ANN is employed; while in [20,45,46] a SVM was used as the pattern classification algorithm, employing a radial basis function (RBF) as kernel function in the last two references.…”

Section: Discriminative Model-based Methodsmentioning

confidence: 99%

“…The classification accuracy was 95%. The proposal described in [16] aimed to classify different ethanol quantities in water. The length of the optical fiber cable was 500 m. 5 different sensors were employed for signal acquisition, located each at 87 m, 215 m, 307 m, 397 m, and 436 m. A combination of 9 ethanol quantities from the 5 sensors was intended to be detected, which derives in a 9-class classification problem.…”

Section: Multi-class Classificationmentioning

confidence: 99%

“…Although some works have been presented in this direction [3,[12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27], there is still, many times, a lack of understanding of how a reliable and efficient system can be built, and what is the adequate and rigorous methodology to apply in the system evaluation processes. In addition, to the best of our knowledge, no thorough written report that summarizes the main characteristics of these systems exists in the literature applied to DAS+PRS technology.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

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

Tejedor

Macías-Guarasa

Martins³

et al. 2017

Applied Sciences

View full text Add to dashboard Cite

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.

show abstract

Section: Frequency Domain-based Feature Extraction Methodsmentioning

confidence: 99%

Section: Discriminative Model-based Methodsmentioning

confidence: 99%

Section: Discriminative Model-based Methodsmentioning

confidence: 99%

Section: Multi-class Classificationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

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

Tejedor

Macías-Guarasa

Martins³

et al. 2017

Applied Sciences

View full text Add to dashboard Cite

show abstract

Recent Advances in Machine Learning for Fiber Optic Sensor Applications

Venketeswaran

Lalam

Wuenschell

et al. 2021

Advanced Intelligent Systems

115

View full text Add to dashboard Cite

The latest digital revolution involves the rise of smart devices composed of sensor hardware and artificial intelligence (AI) software for performing intelligent tasks. Smart sensors have become ubiquitous in our lives with varied applications ranging from voice-enabled home devices (Google Home, Alexa, etc.) to the Industrial Internet of Things (IIoT). This revolution has been fueled by 1) miniaturization of sensing hardware, 2) easy access to cloud and high-performance computing, 3) development of big data storage and analytics technologies, and 4) the latest breakthroughs in machine learning (ML) and AI technologies. The emergence of AI since 2012 and its major breakthroughs can be attributed to the research and development (R&D) in deep learning, a subfield of ML that uses biologically inspired neural networks to perform learning tasks. [1] The performance of conventional ML algorithms depends on the individual selection of specific features, while deep neural networks (DNN) automatically generate features as part of the learning process. Deep learningbased AI technologies are increasingly showing performance

show abstract

Interrogation system for optical sensor using filter bank and artificial neural network

Jucá

Haddad

Santos

2020

Micro & Optical Tech Letters

View full text Add to dashboard Cite

Considering the increasingly wide application of optical fiber sensors, this article aims to present an alternative form of interrogation without the use of an optical spectrum analyzer or any other high-cost devices. The sensor studied here is a long-period grating being used to measure temperature. The interrogator is composed of optical filters and photodetectors whose responses are processed by a suitably trained artificial neural network. Results show that this technique enables effective interrogation of a range limited only by the optical bandwidth of the light source.

show abstract

Principal component analysis and artificial neural network based approach to analysing optical fibre sensors signals

Cited by 39 publications

References 26 publications

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

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

Recent Advances in Machine Learning for Fiber Optic Sensor Applications

Interrogation system for optical sensor using filter bank and artificial neural network

Contact Info

Product

Resources

About