XClose

Institute of Communications and Connected Systems

Home
Menu

Convolutional Neural Networks to classify oil, water and gas pipe fluid using Acoustic signals

ISSPIT 2019 | Vahabi N, Selviah D | Identifying the fluid type and predicting the amount of each fluid in the fluid mixture in the pipes and boreholes are of critical importance for system contro...

1 February 2020

Convolutional Neural Networks to classify oil, water and gas pipe fluid using Acoustic signals

Abstract


Identifying the fluid type and predicting the amount of each fluid in the fluid mixture in the pipes and boreholes are of critical importance for system control and management of wells by Oil and Gas production energy industry, Hydraulic fracturing and shale gas extraction energy industry, Borehole water supply industry and for Gas pipelines. Therefore automating this process will be very valuable for the oil industry. The current study contributes to our knowledge by addressing this important issue. The presented paper investigates the classification algorithms that identify the fluid type in oil, water and gas pipes using acoustic datasets. The datasets analysed in this study are recorded from real oil and gas pipes which means there is no controlled environment, with lots of noisy signals due to unpredicted events under the sea. The data is collected during 24 hours from optical laser acoustic sensors which is attached alongside the 4000 m of oil, water and gas pipes. In this research we used the sample of data from 1000 m of pipes and during 20, 000 s. We implemented Artificial Neural Networks (ANN) and Conventional Neural Networks (CNN) algorithms to recognise the patterns of each fluid type by analysing its acoustic energy. Both algorithms were trained on three datasets (oil, gas and water) and validated on another dataset from different water pipe. The result of this study shows ANN and CNN algorithms classify the fluid type with the accuracy of 79.5% and 99.3% respectively when applied on the test dataset. This is a novel application of modern machine learning algorithm on audio signals which has a direct industry impact.

Publication Type:Conference
Authors:Vahabi N, Selviah D
Publisher:IEEE
Publication date:01/02/2020
Pagination:1-6
Published proceedings:Proceedings of the 2019 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)
Status:Published 
Name of Conference:IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)
Conference location:Ajman, UAE
Conference start date:10/12/2019
Conference end date:12/12/2019
Print ISSN:1932 -4553
Language:English
Keywords:
Convolutional Neural Networks, Artificial Neural Networks, Fluid Flow Classification, Signal Processing
DOI:http://dx.doi.org/10.1109/ISSPIT47144.2019.9001845
Full Text URL:

https://discovery.ucl.ac.uk/id/eprint/10043441/


Explore how UCL research is advancing the future technologies of a connected world: