7th IMRMPT

Abstract

Abstract information

English title:	Pattern recognition of two-phase flow in vertical pipes using artificial intelligence techniques
Spanish title:	Reconocimiento de patrones de flujo bif�sico en tuber�as verticales mediante t�cnicas de inteligencia artificial
English abstract:	The structuring of flow maps in the hydrocarbons industry has been developed from empirical correlations that integrate flow characteristics, such as superficial velocities and volumetric fractions, together with physical properties of the substances under analysis such as density and viscosity, in addition to using geometric parameters such as inclination and internal diameter of the pipes. Currently, technological advances have been developed that allow optimizing the flow pattern identification process, based on the implementation of artificial intelligence techniques, due to the fact that there are limitations to monitor the flow pattern generated inside the pipelines in offshore platforms, given the extreme working conditions that occur there. In this study, a long- and short-term memory (LSTM) recurrent neural network was developed, capable of predicting flow patterns generated inside vertical pipes. The LSTM network was trained and subsequently validated with the information contained in a database obtained from the literature, showing an error generated in the model predictions of less than 1%.
Spanish abstract:	La estructuraci�n de mapas de flujo en la industria de hidrocarburos se ha desarrollado a partir de correlaciones emp�ricas que integran caracter�sticas del flujo, como velocidades superficiales y fracciones volum�tricas, en conjunto con propiedades f�sicas de las sustancias en an�lisis como densidad y viscosidad, adem�s de utilizar par�metros geom�tricos como la inclinaci�n y el di�metro interno de las tuber�as. En la actualidad, se han desarrollado avances tecnol�gicos que permiten optimizar el proceso de identificaci�n de patrones de flujo, a partir de la implementaci�n de t�cnicas de inteligencia artificial, debido a que existen limitaciones para monitorear el patr�n de flujo generado al interior de las tuber�as de las plataformas instaladas en altamar, dadas las condiciones de trabajo extremas que all� se presentan. En este estudio se desarroll� una red neuronal recurrente de memoria a largo y corto plazo (LSTM), capaz de predecir los patrones de flujo generados al interior de tuber�as verticales. La red LSTM fue entrenada y posteriormente validada con la informaci�n contenida en una base de datos obtenida de la literatura, mostrando un error generado en las predicciones del modelo inferior al 1%.
Topic:
Presentation Type:
Speaker:

Authors

Name	Institution	Email
Octavio Andr�s Gonz�lez-Estrada	Universidad Industrial de Santander	agonzale@uis.edu.co
Carlos Ruiz D�az	Universidad de Sao Paulo	carlosruiz978@hotmail.com
Brayan Quispe	Universidad Industrial de Santander	brayan.nos@hotmail.com

Presentation

Room:	Principal Room
Date:	June 7, 2023
Hour:	08:20 - 08:40

Back

7^th IMRMPT
Seventh International Meeting for Researchers in Materials and Plasma Technology

Previous editions:

        1st IMRMPT 2011
        2nd IMRMPT 2013
        3rd IMRMPT 2015
        4th IMRMPT 2017
        5th IMRMPT 2019
        5+1 IMRMPT 2021

Contact information:
imrmpt@foristom.org

Copyright 2022 FORISTOM
All rights reserved