PAPER (OPENFOAM) CFDModelingSlugCatcherMultiphase 2016
PAPER (OPENFOAM) CFDModelingSlugCatcherMultiphase 2016
net/publication/303790530
CITATION READS
1 83
5 authors, including:
All content following this page was uploaded by Gianluca Montenegro on 07 June 2016.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Journal of Applied Fluid Mechanics, Vol. 9, Special Issue 1, pp. 1-9, 2016.
Selected papers from the 7thInternational Exergy, Energy
and Environment Symposium, IEEE7-2015
Available online at www.jafmonline.net, ISSN 1735-3645, EISSN 1735-3645.
ABSTRACT
In the oil & gas industry, the traditional procedure for slug catcher design is based on the Stokes' law. Design
equations are obtained from a 1-D analysis and validated with experimental data. Therefore, this method
basically relies on simplified models and empirical correlations. For this reason, an over margin factor from
20 to 40% is usually applied. In this paper, a simplified CFD procedure for the modelling of the gas-liquid
separation is presented. Steady state and transient models have been considered for single phase and
multiphase fluids, using OpenFOAM. The influence of flow model and mesh grid on results have been
evaluated as a trade-off between solution accuracy and computational efforts, in order to assess the
applicability of these models to industry. A comparison of the industrial validation procedure with the CFD
analysis has been realized, focusing on the pros and cons of the two different approaches. A new application
solver has been constructed and programmed in order to get the most accurate results with the minimum
computational efforts. This solver is based on a completely new and innovative approach to the Navier-Stokes
equations for multiphase flow. New model proposed has been used for the evaluation of design for the two
slug catchers studied, in order to get a better separation and fluids management.
NOMENCLATURE
2. THE SLUG CA
ATCHER
2
G Montenegro et
G. e al. /JAFM, Vol. 9, Special Isssue 1, pp. 1-9, 20016.
The drag
d model can be chosen fromm a variety of
differeent experimentaal or semi-empirrical relations.
In thiss work the Schhiller Neumann formulation is
considdered [5]:
24
Ree p 1000
0.687
Re 1 0.15 Re p
Cd p (2)
0.44
Re p 1000
Fig. 2. Floww regimes in horrizontal pipes Wheree the Reynolds nnumber depends on the relative
The first step needed to prooceed with a CFD
T C velociity between the ggas and the liquid phases:
a
analysis of the multiphase
m flow inside
i a slug cattcher
i the study of th
is he actual design procedure for finger l d droplet U r
t
type slug catch her already useed in the indusstrial Re p (3)
l
p
practice. It is based
b on simplee concepts of ffluid
3
G. Monttenegro et al. /JA
AFM, Vol. 9, Sppecial Issue 1, ppp. 1-9, 2016.
4
G Montenegro et
G. e al. /JAFM, Vol. 9, Special Isssue 1, pp. 1-9, 20016.
hhypothesis of incompressiblee flow, the m mass the seettling Vt . Thee value of Vt is, therefore,
c
conservation equuation is rearrangged in the follow
wing mined by means oof an iterative prrocedure.
determ
f
form, which is implemented in the solver ccode
(
(Malalasekera, 2007;
2 Ferziger): An adv vantage of this ssimplified approoach is that the
system
m of equations iss now decoupledd. There in no
U 0 (9) more interaction beetween the moomentum and
t
contin
nuity equations oof the two phasess, but there is a
T general forrm of the mom
The mentum conservaation
single phase system of o equation pluus a decoupled
e
equation, obtainned with the Eulerian approoach
transport equation w which exploits thet convective
f
fixing a cell vo olume in the space, can be wrritten
term. This means thatt the system cann be modelled
w the followin
with ng notation:
also ass a steady state process,
p resultingg in a dramatic
U
UU U T S f (10) reducttion of the compuutation demand.
t
w
where T is the sttress tensor of a Newtonian fluidd. 5
5. PRELIMIRA
ARY SIMULATIIONS
A
Applying the steeady state and incompressible
i f
flow
c
condition and neglecting the diiffusive transporrt of
t liquid particcles, which are transported onlyy by
the To vaalidate the simpplified approachh, simulations
c
convection, the final equation for the transporrt of were carried
c out on a pipe with sqquared section.
t droplet volu
the ume fraction caan be formulated as This allowed
a having a calculation mesh
m free from
f
follows: typical issues that generate
g errors,, in order to
quicklly optimize the t CFD appplication. The
U S p (11) geomeetry is representeed by a 16 meteers pipe, with a
t squareed cross section cross section off 0.36 m2 (Fig.
S n this way it is possible to simplify the
4). In
The source term
T m p , is mathem matically introduuced calculaations, reducingg the time necessary to solve
a a scalar quaantity inside th
as his formulation and the prooblem, and provviding the possib bility to easily
s
should mimic thhe separation of the t liquid phasee due analysse the results. The mesh in this case is
t the gravity efffect. Considerinng the physics off the
to compo osed by a simplee block.
s
sedimentation prrocess, it is posssible to undersstand
t the gravity produces
that p a flow of liquid particlees in
t direction off the gravity. Thhis flow producces a
the
f
flux that can be conveniently represented byy the
d
divergence operaator. For this reaason, it is possible to
r
rearrange the souurce term as a flux
fl term, given by a
v
vertical velocityy multiplied by the orthogonal area
o the cells:
of
U Vt . (12)
t Fig. 4:Square duct ggeometry and mesh
m detail.
The volumetric flux of this neew divergence tterm
T
c
comes from the terminal
t velocityy calculated in E
Eq.1. The simulations are m managed focusing the attention
T face center value of the terrminal velocity used
The u on thee influence on tthe final solutioon of different
i the discretizzation of the divergence
in d term
m is parammeters, such as thhe mesh size, thee discretization
o
obtained by lineear interpolationn of the cell ceenter schemmes and the soluution algorithm ms. The results
v
values of the terrminal velocity field. Equationn 12 obtainned have allowed the comparisoon of both the
c be also writtten in a more compact
can c formulaation CFD approaches: thee eulerian-euleriian two phase
c
considering the volumetric
v flux given by the sum m of approaach and the simmplified single-pphase one. The
t two convecctive contributioons: gas convecction
the simulaated cases are illlustrated in the Figs 5 and 6,
a deposition:
and where a threshold ffilter highlights a separation
U - Vt 0 ((13) efficieency of 99%. It can be noticed that
t the length
t of the complete segreegation of the liquid
l phase is
IIn this frameworrk, it is really im
mportant to definee the approxximately the saame for the tw wo approaches,
c
correct boundarry on the terminal velocity. A As a approxximately 14 m. In this particulaar case there is
M
Matter of fact, im
mposing at the boundary a Neum mann no expperimental correelation that cann be exploited,
c
condition wouldd imply to assiggn an incoming and therefo ore, the CFD reesults are compaared to the 1D
o
outgoing flux att those patches where there shhould Stokess’ approach. Thee particle diametter used for the
n be any kind of flow. To acco
not ommodate this iissue calculaations has been cchosen equal to 150 μm.
a slip type bouundary conditio on on the term minal
v
velocity field hass been imposed.
The drag coefficcient is computeed with the Schiiller-
T
N
Neumann modell previously presented. In this ccase,
s
since there is onnly one velocityy (the gas veloccity),
a
assumptions neeed to be made ini order to estimmate
t
the relative vellocities. In partticular, it has bbeen
a
assumed that thee two phases arre moving one with Fig. 5:Singlle-phase approaach.
r
respect to the oth
her only by the term responsiblee for
5
G. Monttenegro et al. /JA
AFM, Vol. 9, Sppecial Issue 1, ppp. 1-9, 2016.
experimental database.
d
An analysis ofo the velocity vvector in the dirrection
orthogonal to the main flux reevealed the preseence of
a lifting effectt at the end of tthe downcomer, where
the rotational flow caused byy the variation of the
geometry distuurbs the depositiion process. Thee liquid
Fig. 6. Two-phases approach
h. droplets, moddelled as sphericcal particles, aree lifted
by this flow causing the ooverestimation of the
The 1D D Stokes’ meethod underesttimates the separation length. This effect iis shown in Fig. 9.
segregattion length, provviding a value off 9.68 m. The
maximum m error obtaineed is about 40% %. The main
reason of
o this discrepanccy is due to the fact that the
1D theoory approximatees the flow as a 1D flow,
whose component
c of the velocity in i the axis
direction
n is uniformly diistributed over thhe flow area.
This is obviously
o not reaalistic and thereffore does not
account that in the midddle on the pipe thhe velocity is
higher, causing
c an increease of the lengtth necessary
to allow the separaation. Imposin ng uniform
distribution of the velociity and setting sllip boundary
conditionns, the case beecomes equivaleent to a 1D
case, ressulting in a sepparation length close
c to one
given byy the 1D theory.
After havving validated thhe two approachhes, a simple Fig. 8. a) Sin
ngle-phase thresshold filter; b) Two-
T
part of a real slug catcheer, composed off a finger and phase thresholdd filter.
a 45° incclined downcom mer has been considered (Fig.
7). This geometry perm mits to test the behaviour
b of
CFD sollvers, comparinng the results obbtained with
the exxperimental inndustrial standdards. The
downcom mer is 4 m lonng, with a diam meter of 30
inches. The
T finger is incclined by 1.15°, and is 40 m
long, wiith a diameter ofo 48 inches. Thhe mesh size
has beenn chosen after a sensitivity anaalysis on the
volumetrric fraction disstribution. The calculation
grid, connsidering these two parameterrs, has been
generateed using the messh cartesian mesh generator
of Open nFOAM, namelly snappy HexxMesh. This
mesh gennerator creates a high quality, heex-dominant
mesh sttarting from a background mesh m and a Fig. 9. Rootational flow in
nside the fingerr.
surface representation
r o the geometry that will be
of
studied. Furthermore, surface refinnement and The solver does not take into account that thee liquid
boundary y layers have been included in the final deposited on the bottom off the pipe cann not be
mesh. Also
A in this case several test sim
mulations are dragged in thee same way of siingle droplets. ForF this
performeed, in order to further optimiize both the reason, a coallescence model has been implem mented
CFD sollvers. The attenntion has been posed
p on the to simulate thhe real behaviouur of the liquid already
a
influencee of the mesh anda the boundarry condition collected at the
t bottom of tthe geometry, liimiting
characteristics. this lifting efffect. The idea inn this case is to find a
threshold valuue of the volum me fraction of liquid,
above whichh the droplets start to agggregate,
generating parrticles with largeer diameter. The model
implemented individuates the αlim value whenn a cell
is completelyy full of spherrical particles without
w
merging phenomena (Fig. 10)).
Fig. 7. Downcomer
D + finger
f geometryy and final
mesh
h detail.
The finaal results are shown
s in Fig. 8. As it is
possible to see, in this case
c the results obtained, in
terms off liquid volumetrric fraction, usinng the single-
phase ap pproximation is not in agreemeent with the
Euleriann-Eulerian approoach. In particullar, it seems
to overestimate the to otal length off separation,
whereas the two-phasse simulation computes c a Fig.
F 10. α limit ccondition.
separatioon in agreem ment with thee industrial
6
G Montenegro et
G. e al. /JAFM, Vol. 9, Special Isssue 1, pp. 1-9, 20016.
5. CONC
CLUSION
As it is possible
p to see (T
Tab. 2), the current layout of
the slugg catcher seem ms to gather annd distribute
homogennously the incooming flow from the well.
Having evaluated the fluuid dynamics off the device,
it is noww possible to prroceed with the multi-phase
simulatioon using the single
s phase appproximated
olver (FFig. 15) and thee multiphase app proach (Fig.
16). Thee average particcle diameter is set to 1000
μm, in agreement withh the specificattions of the
extractioon site.
8
G. Montenegro et al. /JAFM, Vol. 9, Special Issue 1, pp. 1-9, 2016.
separation mechanism, allowing both to improve Engineering Data Book, Gas Processor
design and better handle change in operating Association.
conditions. The latter can be easily translated in the
ability to support facilities management and Gossman A. D., R. I. Issa, C. Lekakou, M. K.
operation problem solving. Looney and S. Politics (1992)
Multidimensional modelling of turbulent two-
On the other side it has to be mentioned that CFD phase flows in stirred vessels, AIChE J.
simulations are time consuming operations. As a
consequence a trade-off between calculation efforts Hill D. P. (1998), The computer simulations of
and result accuracy has to be achieved. dispersed two-phase flows, Imperial College of
London, London,
In the present paper, the single-phase steady state
application has been recognized as the best Malalasekera, H. K. and W. Versteeg (2007) An
compromise among the aforementioned parameters. introduction to computational fluid dynamics,
Main focus points are on the coalescence PEARSON Prentice Hall.
mechanism of the liquid particle and on the average Naumann, L. and A. Schiller (1933), Uber die
droplet diameter, individuated as the most critical grund legend en Berechnungenbei,
parameters responsible for reliable and consistent Vereinsdeutscher Ing.
results.
OpenCFD Ltd., “OpenFOAM Programmer’s
This activity has allowed the evaluation of a new Guide”. Online available:
workflow for slug catcher design, allowing to http://www.openfoam.org
enhance a standard procedure which in the past has
always been based on empirical correlations. OpenCFD Ltd., “OpenFOAM User’s Guide”.
Online available: http://www.openfoam.org
REFERENCES Rusche H. (2002), Computational Fluid Dynamics
of Dispersed Two-Phase Flows at High Phase
B. P. “Energy Outlook 2035”, 2014. Online Fractions, Imperial College of London,
available: http://www.bp.com London.
Ergun S. (1952). Fluid flow through packed Weller H. G. (2002a), A code independent for finite
columns, Chem. Eng. Progr. volume algorithm, Technical report-Nabla Ltd.
Ferziger J. H., and M. Peric, Computational Weller H. G. (2002b), Derivation, modelling and
Methods for Fluid Dynamics, Springer, third solution of the conditionally averaged two-
edition phase flow equations, Technical report-Nabla
Ltd.
Gas Processing Supplier Association (2004)