Please use this identifier to cite or link to this item:
|Title:||Computational intelligence models for PIV based particle (cuttings) direction and velocity estimation in multi-phase flows||Authors:||Tombul, H.
Özbayoğlu, Ahmet Murat
Özbayoğlu, M. E.
|Issue Date:||2019||Publisher:||Elsevier B.V.||Source:||Tombul, H., Ozbayoglu, A. M., and Ozbayoglu, M. E. (2019). Computational intelligence models for PIV based particle (cuttings) direction and velocity estimation in multi-phase flows. Journal of Petroleum Science and Engineering, 172, 547-558.||Abstract:||In multi-phase flow, the gas phase, the liquid phase and the particles (cuttings) within the liquid have different flow behaviors. Particle velocity and particle direction are two of the important aspects for determining the drilling particle behavior in multi-phase flows. There exists a lack of information about particle behavior inside a drilling annular wellbore. This paper presents an approach for particle velocity and direction estimation based on data obtained through Particle Image Velocimetry (PIV) techniques fed into computational intelligence models, in particular Artificial Neural Networks (ANNs) and Support Vector Machines (SVM). In this work, feed forward neural networks, support vector machines, support vector regression, linear regression and nonlinear regression models are used for estimating both particle velocity and particle direction. The proposed system was trained and tested using the experimental data obtained from an eccentric pipe configuration. Experiments have been conducted at the Cuttings Transport and Multi-phase Flow Laboratory of the Department of Petroleum and Natural Gas Engineering at Middle East Technical University. A high speed digital camera was used for recording the flow at the laboratory. Collected experimental data set consisted of 1080 and 1235 data points for 15° inclined wellbores, 1087 and 1552 data points for 30° inclined wellbores and 885 and 1119 data points for horizontal (0°), wellbores respectively to use in estimation and classification problems. Results obtained from computational intelligence models are compared with each other through some performance metrics. The results showed that the SVM model was the best estimator for direction estimation, meanwhile the SVR model was the best estimator for velocity estimation. The direction and speed of the particles were estimated with a reasonable accuracy; hence the proposed model can be used in eccentric pipes in the field. © 2018 Elsevier B.V.||URI:||https://www.sciencedirect.com/science/article/pii/S092041051830826X?via%
|Appears in Collections:||Bilgisayar Mühendisliği Bölümü / Department of Computer Engineering|
Scopus İndeksli Yayınlar Koleksiyonu / Scopus Indexed Publications Collection
WoS İndeksli Yayınlar Koleksiyonu / WoS Indexed Publications Collection
Show full item record
checked on Sep 23, 2022
WEB OF SCIENCETM
checked on Sep 24, 2022
checked on Nov 28, 2022
Items in GCRIS Repository are protected by copyright, with all rights reserved, unless otherwise indicated.