(75b) A CFD Based High-Dimensional Deep Neural Network Surrogate Model for Oil/Water Separation in Horizontal Oil Production Pipelines

Tsai, K., Shell International E&P

Eisnot McSquare Normal Tsai, Kuochen T SIEP-PTX/D/I 2 9 2019-04-13T01:47:00Z 2019-04-13T01:47:00Z 1 464 2651 22 6 3109 16.00

Clean Clean false false false false EN-US ZH-TW X-NONE

/* Style Definitions */ table.MsoNormalTable {mso-style-name:"Table Normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-parent:""; mso-padding-alt:0in 5.4pt 0in 5.4pt; mso-para-margin-top:0in; mso-para-margin-right:0in; mso-para-margin-bottom:8.0pt; mso-para-margin-left:0in; line-height:107%; mso-pagination:widow-orphan; font-size:11.0pt; font-family:"Calibri",sans-serif; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin;}

A CFD based high-dimensional deep neural network surrogate model for
oil/water separation in horizontal oil production pipelines

-Kuochen Tsai and Ligang Lu, Shell International Exploration
& Production, Inc.

Deep neural network (DNN) model has been proven to be highly
successful in modeling computational fluid dynamics data for oil/water
separations in oil production pipelines in 4-dimensional parametric space [Ref
1]. Further development demonstrates the feasibility for a 9-dimensional DNN
model with physical parameters that covers the entire range of operating
conditions, including mixture velocity, water cut, initial droplet size, pipe
diameter, fluid viscosity, density and surface tension.

The computational fluid dynamics (CFD) model is based on the
Eulerian Interfacial Area Exchange (IAC) model of Ishii and Kim [Ref 2]. The
CFD model development was presented in 2017 with comprehensive validation with
published experimental studies [Ref 3, 4, 5]. The validations showed the accuracy
and scale-up capability of the CFD model.

The combination of DNN and CFD is a powerful tool that can
be further integrated with a web model using public domain software (Python and
Dash Plotly). The result is a general- purpose, user
friendly real-time web model for oil/water separation predictions.

Similar ideas can be extended for many other types of
physics-based simulation tools including finite element models and process
simulators. The concept of extracting 1-D or 2-D subset data from a 3-D model
is highly flexible and can be a powerful tool for converting time-consuming
simulations into simple real-time general-purpose web models.


1] K. Tsai and L. Lu, “A Deep Learning Neural Net Model for Oil/Water Separation in Oil
Production Pipelines”, Paper 152g, 2019 AIChE Annual
Meeting, Pittsburth, PA.

2] Wu Q., Kim, S. and Ishii, M.: “One-group interfacial area transport in
vertical bubbly flow,” International
Journal of Heat and Mass Transfer
 (1998), 41(8–9), 1,103–1,112

3] K. Tsai, “Modelling the separation of oil and water in pipelines”,
Paper 452a, 2017 AIChE Annual Meeting, Minneapolis, USA

4] Chesters, A. K.: “The modelling of coalescence
process in fluid–liquid dispersions: A review of current understanding,” Chemical
Engineering Research and Design
 (1991), 69 (A4), 259–270

5] Simmons, M. J. H. and B. J. Azzopardi: “Drop size distribution in dispersed
liquid–liquid pipe flow,”International
Journal of Multiphase Flow
 (2001) 27(5), 843–859

6] Fairuzov Y. V., Arenas-Medina, P., Verdejo-Fierro,
J. and Gonzales-Islas, R.: “Flow pattern transitions in horizontal pipelines
carrying oil-water mixtures: Full-scale experiments,” Journal
of Energy Resources Technology
 (2000) 122(4), 169–176


This paper has an Extended Abstract file available; you must purchase the conference proceedings to access it.


Do you already own this?



AIChE Members $150.00
AIChE Graduate Student Members Free
AIChE Undergraduate Student Members Free
Non-Members $225.00