Predicting pressure losses in the water-assisted flow of unconventional crude with machine learning. (17th November 2021)

• Record Type: Journal Article
• Authors: Rushd, Sayeed; Rahman, Moklesur; Arifuzzaman, Md; Ali, Sherif Abdulbari; Shalabi, Faisal; Aktaruzzaman, Md
• Is Part Of: Petroleum science and technology. Volume 39:Number 21/22(2021)
• Page Start: 926
• Page End: 943
• Publication Date: 2021-11-17
• Subjects: Core-annular flow -- pipeline transportation -- heavy oil -- viscosity -- friction -- artificial neural network -- support vector machine -- artificial intelligence; Liquid fuels -- Periodicals; Petroleum -- Periodicals; 665.505
• DOI: 10.1080/10916466.2021.1980012
• Languages: English
• ISSNs: 1091-6466

Abstract

Abstract: Machine learning (ML) is recognized as an efficient prediction tool. However, very few attempts have been made to apply it to model pressure losses in the water-assisted pipeline transportation of unconventional crudes. The performances of conventional ML algorithms for predictions were analyzed in the current study based on a dataset comprised of 225 data points and seven input parameters: pipe diameter, average velocity, densities of oil and water, viscosities of oil and water, and water content. Among the algorithms tested, the artificial neural network demonstrated the most promising performance with the coefficient of determination (R 2 ) of 0.99 and mean squared error (MSE) of 0.009.