Severe Slugging 1 PDF

G Model
JOCS-225; No. of Pages 14
ARTICLE IN PRESS
Journal of Computational Science xxx (2013) xxxxxx
Contents lists available at ScienceDirect
Journal of Computational Science

journal homepage: www.elsevier.com/locate/jocs
Modeling and simulation of severe slugging in air-water systems

including inertial effects
J.L. Balino
Departmento de Engenharia Mecnica, Escola Politcnica, Universidade de So Paulo, Av. Prof. Mello Moraes, 2231, CEP 05508-900, Cidade Universitria,
So Paulo, SP, Brazil
a r t i c l e
i n f o
Article history:
Received 31 January 2013
Received in revised form 29 July 2013
Accepted 25 August 2013
Available online xxx
Keywords:
Severe slugging
Pipeline-riser system
Air-water ow
Stability
Lumped and distributed parameter systems
Switched systems
Petroleum production technology
a b s t r a c t
A mathematical model and numerical simulations corresponding to severe slugging in air-water pipelineriser systems are presented. The mathematical model considers continuity equations for liquid and gas
phases, with a simplied momentum equation for the mixture. A drift-ux model, evaluated for the local
conditions in the riser, is used as a closure law. In many models appearing in the literature, propagation
of pressure waves is neglected both in the pipeline and in the riser. Besides, variations of void fraction in
the stratied ow in the pipeline are also neglected and the void fraction obtained from the stationary
state is used in the simulations. This paper shows an improvement in a model previously published by the
author, including inertial effects. In the riser, inertial terms are taken into account by using the rigid waterhammer approximation. In the pipeline, the local acceleration of the water and gas phases are included in
the momentum equations for stratied ow, allowing to calculate the instantaneous values of pressure
drop and void fraction. The developed model predicts the location of the liquid accumulation front in
the pipeline and the liquid level in the riser, so it is possible to determine which type of severe slugging
occurs in the system. A comparison is made with experimental results published in literature including a
choke valve and gas injection at the bottom of the riser, showing very good results for slugging cycle and
stability maps. Simulations were also made assessing the effect of different strategies to mitigate severe
slugging, such as choking, gas injection and increase in separation pressure, showing correct trends.
2013 Elsevier B.V. All rights reserved.
1. Introduction
Severe slugging is a terrain dominated phenomenon, characterized by the formation and cyclical production of long liquid slugs
and fast gas blowdown. Severe slugging may appear for low gas and
liquid ow rates when a section with downward inclination angle
(pipeline) is followed by another section with an upward inclination (riser). This conguration is common in off-shore petroleum
production systems. Main issues related to severe slugging are:
(a) High average back pressure at well head, causing tremendous
production losses, (b) High instantaneous ow rates, causing instabilities in the liquid control system of the separators and eventually
shutdown, and (c) Reservoir ow oscillations.
For steady state and low ow rates, the ow pattern in the
pipeline may be stratied, while it may be intermittent in the riser,
as shown in Fig. 1(a).
A cycle of severe slugging can be described as taking place
according to the following stages [1]. Once the system destabilizes and gas passage is blocked at the bottom of the riser, liquid
E-mail addresses: jlbalino@usp.br, balinoj@yahoo.com
continues to ow in and gas already in the riser continues to ow

out, being possible that the liquid level in the riser falls below
the top level at the separator. As a consequence, the riser column
becomes heavier and pressure at the bottom of the riser increases,
compressing the gas in the pipeline and creating a liquid accumulation region. This stage is known as slug formation (Fig. 1(b)).
As the liquid level reaches the top while the gas passage is kept
blocked at the bottom, pressure reaches a maximum and there is
only liquid owing in the riser. This is the slug production stage
(Fig. 1(c)).
Since gas keeps owing in the pipeline, the liquid accumulation
front is pushed back until it reaches the bottom of the riser, starting
the blowout stage (Fig. 1(d)).
As the gas phase penetrates into the riser the column becomes
lighter, decreasing the pressure and then rising the gas ow. When
gas reaches the top of the riser, gas passage is free through the
stratied ow pattern in the pipeline and the intermittent/annular
ow pattern in the riser, causing a violent expulsion and a rapid
decompression that brings the process to slug formation again. This
stage is known as gas blowdown (Fig. 1(e)).
Fig. 1(f) shows the different stages in the pressure history at the
bottom of the riser corresponding to an experiment under laboratory conditions [2].
1877-7503/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jocs.2013.08.006
Modeling and simulation of severe slugging in air-water systems including inertial effects,
Please cite this article in press as: J.L. Balino,
J. Comput. Sci. (2013), http://dx.doi.org/10.1016/j.jocs.2013.08.006
G Model
ARTICLE IN PRESS
J.L. Bali
no / Journal of Computational Science xxx (2013) xxxxxx
Fig. 1. Stages for severe slugging (from [1,2]).
A classication of severe slugging can be made, according to the

observed ow regime, as follows:
Severe Slugging 1 (SS1): the liquid slug length is greater to or
equal to one riser length and maximum pipeline pressure is equal
to the hydrostatic head of the riser (neglecting other pressure
drop terms).
Severe Slugging 2 (SS2): the liquid length is less than one riser
length, with intermittent gas penetration at the bottom of the
riser.
Severe Slugging 3 (SS3): there is continuous gas penetration at the
bottom of the riser; visually, the ow in the riser resembles normal slug ow, but pressure, slug lengths and frequencies reveal
cyclic variations of smaller periods and amplitudes compared to
SS1.
Oscillation (OSC): there are cyclic pressure uctuations without
the spontaneous vigorous blowdown.
Most of the models for severe slugging were developed for vertical risers and assume one-dimensional, isothermal ow and a
mixture momentum equation in which only the gravitational term

is important.
In [3] a model was presented considering constant mean values for the gas density and void fraction in the riser, allowing to
calculate time variations of pipeline pressure, position of the accumulation region, ow rate into the riser and mean holdup. It was
found that as the operation point moves closer to the stability
line the numerical procedure did not converge, giving gas mass
ows going to innite as the spatial discretization was decreased.
Experimental data were obtained from a facility for different buffer
volumes (simulating equivalent pipeline lengths) and a comparison
was made with the simulation results, showing good agreement
except for the blowout/blowdown stage. Setting apart the nonconvergence problems, lumped parameter models seem to work
ne for short risers, where the local variations of variables are small,
but are not successful in long risers, typical of offshore systems.
In [4] a model with a distributed parameter formulation for
the riser was presented. Considering continuity equations for the
liquid and gas without phase change and a gravity-dominant mixture momentum equation, the model was capable of handling
G Model
ARTICLE IN PRESS
J.L. Bali
discontinuities such as liquid accumulation in the piping and liquid

level in the riser. The resulting equations were solved by using the
method of characteristics. A comparison of simulations with different experimental data showed reasonable agreement, although the
model also suffered from non-convergence in the unstable region.
In [5] a model for severe slugging valid for risers with variable inclination was presented. The model was used to simulate
numerically the air-water multiphase ow in a catenary riser for the
experimental conditions reported in [6]. Stability and ow regime
maps in the system parameter space for the multiphase ow in a
catenary pipeline-riser system were built.
In [7] the model developed in [5] was extended to investigate the
dynamics of gas, oil and water ow in a pipeline-riser system. Mass
transfer between the oil and gas phases was calculated using the
black oil approximation. The properties of uids were calculated
by analytical correlations based on experimental results and eld
data.
The stationary solution for a given point in the system parameter space is given as initial condition for the numerical simulation;
if the numerical solution does not go away from the initial condition with time, the stationary solution is stable and it is the system
steady state. If the numerical solution goes away with time, the
stationary state is unstable, there is no steady state and an intermittent solution develops with time. By changing the point in the
system parameter space and repeating this process, the stability
map can be built. For unstable ow, the analysis of the limit cycle
leads to the determination of the ow regime map, showing the
regions corresponding to the different types of intermittency.
A recent monograph [8] reviews different issues related to
severe slugging for air-water systems.
In all the models reviewed above, propagation of pressure waves
is neglected both in the pipeline and in the riser; this constitutes
the no-pressure-wave (NPW) approximation [9]. As a result of the
NPW approximation, pressure changes are felt instantaneously at
any point in the domain (pressure waves travel at innite speed).
Also in the models reviewed above, a stratied ow pattern
is assumed at the pipeline and variations of void fraction are
neglected. The void fraction is obtained from a momentum balance in the gas and liquid phases, resulting an algebraic relation
between the mean variables [10]. The void fraction determined in
the stationary state is assumed as constant in the simulations, so
the momentum balance equation is not satised and variations of
the void fraction cannot be calculated in the transients.
As stated in [5], a riser model with a simplied momentum
equation in which only the gravity force is considered has limitations to deal with general boundary conditions. In particular, for
such a model to handle a discontinuous pressure boundary condition, it should react with a discontinuous time variation in the
void fraction distribution; to do so, distributional (Diracs delta)
supercial velocities would be necessary. A gravity dominant riser
behaves like an ideal mechanical spring, which reacts with a distributional velocity (whose time integral is a displacement) to a
discontinuous applied force. This limitation probably gave rise to
the non-convergence problems reported in [3,4] as innite gas
penetration at the bottom of the riser.
A discontinuous pressure boundary condition occurs when a
choke valve is located at the top of the riser and the valve opening
is changed discontinuously. For a constant valve opening, a discontinuity in pressure also occurs in the severe slugging slug formation
stage (see Fig. 1(b)): pressure drop across the choke valve is negligible for gas ow only, but becomes important when the mixture
level reaches the top of the riser. For these cases, discontinuous
supercial velocities appear, giving unrealistic results.
In [5] simulations were made, in which the separation pressure
was constant and there was no choke valve; in this case, continuity in pressure (and also in pressure time derivative) at the bottom
of the riser was assured by suitable pipeline switching conditions

between the two possible congurations (see Section 2.1). The
simplied friction term implemented in the model improved the
accuracy of the results and had an stabilizing effect. The implicit
numerical procedure used to solve the model equations (including
the time step and the gas velocity used to displace the nodes in the
nonlinear iterations) also contributed to the numerical stability and
convergence.
It becomes evident that a more sophisticated riser model is
necessary to deal with general boundary conditions, taking into
account additional terms in the dynamic equations. In addition to
gravitational and friction, inertial and compressibility effects can
be taken into account.
Compressibility effects are necessary for rapid transients in
which discontinuous velocity boundary conditions exist, such as a
sudden closure of the choke valve, giving rise to pressure waves and
increasing the number of the characteristic values and the complexity of the numerical solving procedure. Inertial effects, associated to
the uid acceleration, can deal with discontinuous pressure boundary conditions and can be added as a good approximation for slow
transients.
This paper shows an improvement in the model previously published by the author [5], including inertial effects.
In the riser, inertial terms are taken into account by using the
rigid water-hammer approximation [11]. In this approximation,
the acceleration terms for the liquid and gas phases are taken into
account in the momentum equation, but compressibility of the liquid phase is neglected in the mass conservation equation. In the
pipeline, convective acceleration terms are neglected but the local
acceleration terms for the water and gas phases are included in the
momentum equations for stratied ow, allowing to calculate the
instantaneous values of pressure drop and void fraction.
The model includes additional devices, such as a valve located
at the top of the riser and a gas injection line at the bottom of the
riser. In this way it is possible to evaluate valve closure and gas lift
as mitigating actions for severe slugging.
2. Model
The model considers one-dimensional ow in both pipeline and
riser subsystems. The liquid phase is assumed as incompressible,
while the gas phase is considered as an ideal gas. Both phases ow in
isothermal conditions. The ow pattern in the pipeline is assumed
as stratied. The model is capable of handling discontinuities in the
ow, such as liquid accumulation in the pipeline, liquid level in the
riser and void fraction waves.
2.1. Pipeline
The pipeline, shown in Fig. 2, can be either in a condition of liquid
accumulation (x > 0) or in a condition of continuous gas penetration
(x = 0), where x is the position of the liquid accumulation front. The
existence of a buffer vessel with volume e is considered in order
to simulate an equivalent pipeline length Le = Ae , where A is the
ow passage area (A = 14 D2 , where D is the inner diameter).
Three locations in the pipeline can be identied, with corresponding pressures and supercial velocities for the liquid and gas
phases: the bottom (Ppb , jlb and jgb ), the stratied region located at
the liquid accumulation front (Pps , jls and jgs ) and the top (Ppt , jlt and
jgt ).
Mass and momentum conservation equations are applied for the
control volumes corresponding to the liquid and gas phases. In the
mass equation, a mean pressure P g (t) = 12 (Ppt + Pps ) is assumed. In
mass and momentum equations, variations of the void fraction with
position are neglected and a mean void fraction p (t) is considered.
G Model
ARTICLE IN PRESS
J.L. Bali
Fig. 2. Denition of variables at the pipeline.
These approximations are made in order to get a lumped parameter

model for the pipeline.
The supercial velocities at the top of the pipeline can be written
as:
Q
jlt = l0
A
jgt
(1)
g0 Rg Tg
m
=
Ppt A
(2)
g0 and Ql0 are respectively the gas mass ow and the liquid
where m
volumetric ow injected in the pipeline and Rg and Tg are respectively the gas constant and temperature.
2.1.1. Condition x > 0
For this condition there is no gas owing out the pipeline, resulting:
jgb = 0
(3)
Applying mass conservation equation for the liquid and gas phases,
it is obtained:
Ql0
A
jlb
+ p
dp
=
Lx
dt
dP g
=
dt
P g
jlb
Ql0
A
dx
dt
(4)
+
Rg Tg
A
g0
m
(5)
(L x) p + Le
where L is the pipeline length and t is time. A mass balance for the
liquid in the stratied region yields:
dp
Q
(L x)
+ jls l0 = 0
A
dt
(6)
From Eqs. (4) and (6), it results:
jgs = jgb
(10)
jls = jlb
(11)
Pps = Ppb
(12)
Applying mass conservation equation for the liquid and gas

phases, it is obtained:
dp
1
=
L
dt
dP g
=
dt
dx
dt
(8)
Applying the momentum conservation equation in the control

volume of liquid region at the penetration front and including friction and inertial terms, we get:
2f
dj
= Pps + l x g sin ll jlb |jlb | lb
D
dt
(9)
where fll is the Fanning friction factor (assuming that only liquid
is lling the cross sectional area), g is the gravity acceleration constant, l is the liquid density and is the pipeline inclination angle
jlb
P g
Ql0
A
jgb + jlb
(13)
Ql0
A
Lp + Le
Rg Tg
A
g0
m
(14)
2.1.3. Switching between conditions

The supercial velocities must be determined when the pipeline
commutes between the states of liquid accumulation and continuous gas penetration.
Assuming that the commutation from the state x > 0 to the state
x = 0 happens at time t0 , it can be shown that the supercial velocities at times immediately before t0 and immediately after t0+ can
be determined as follows:
+
dx
=0
dt
(15)
jgb
=0
(16)
(7)
From the kinematic condition at the liquid penetration front, we

get:
Ppb
2.1.2. Condition x = 0
For this condition there is no liquid penetration front, resulting:
+
jgb
= p
dx
jls = jlb + p
dt
jgs = p
(positive when downwards). The Fanning friction factor is calculated by using the correlation from [12].
dx
dt
jlb
= jlb
+ p
(17)
dx
dt
(18)
Assuming that commutation from the state x = 0 to x > 0 happens

at time t0 , it can be shown that the supercial velocities at times
immediately before t0 and immediately after t0+ can be determined
as follows:
+
jgb
= jgb
=0
(19)
jlb
= jlb
(20)
dx
=0
dt
(21)
In Eqs. (15)(21) the superscripts and + denote variables evaluated respectively at t0 and t0+ . In this way, in any commutation
G Model
ARTICLE IN PRESS

J.L. Bali
Fig. 3. Stratied ow at the pipeline.
the total supercial velocity, pressure and time derivative of the

pressure at the bottom of the riser are continuous.
2.1.4. Local equilibrium condition for stratied ow
In previous models, the void fraction at the pipeline is determined from an algebraic relationship evaluated from the stationary
state, derived from the momentum balance in stratied ow [10].
This relationship can be generalized by including the local inertial terms. Assuming stratied ow (see Fig. 3), the generalized
relationship can be written as:
Sg
Sl
wl
+ i Si
p
1 p
wg
+ (l g )Ag sin A
ug =
l
1
1
+
1 p
p
dug
dul
g
dt
dt
jg =
jg
(24)
Pg
Rg Tg
(25)
1
(jgt + jgs )
2
Q
l0
+ jls
(26)

(27)
where jg and jg are respectively the mean supercial velocities

for gas and liquid, Sg , Si and Sl are respectively the gas, interfacial and liquid wetted perimeters, wg , i and wl are respectively
the wall-gas, interface and wall-liquid shear stresses, ug and ug are
respectively the mean phase velocities for gas and liquid and g is
the mean gas density.
The pressure gradient along the stratied region in the pipeline
can be obtained from the linear momentum equation for the liquid
phase as:
S
S
P
= wl l + i i + l
Al
Al
s
g sin
dul
dt
Ppt = Pps
0
Lx
Continuity equations for the phases are considered at the riser

(see Fig. 4). This results in the following set of equations:
jl
+
=0
t
s
(29)
In Eqs. (22)(29) the wetted and interfacial perimeters are determined considering a stratied geometry, while the shear stresses
are related to the supercial velocities of the phases through friction factors based on the hydraulic diameters for each phase [13].
(31)
where jg , jl and j are respectively the gas, liquid and total supercial
velocities, P is the pressure, s is the position along the riser, is the
void fraction and ug and ul are respectively the gas and liquid phase
velocities.
The supercial velocities for the phases are determined by using
a drift ux correlation, assumed to be locally valid:
jg = ug = (Cd j + Ud )
(32)
jl = j jg = ul (1 ) = (1 Cd )j Ud
(33)
It will be assumed that the drift parameters Cd and Ud depend

at most on the
local owconditions and
inclination
angle = (s),
this is, Cd = Cd , P, j, and Ud = Ud , P, j, [14,15].
The drift coefcients used in the model are [14]:
For Frj < 3.5:
Cd = 1.05 + 0.15 sin
Ud =
gD 0.35 sin + 0.54 cos
(34)
(35)
For Frj 3.5:

(36)
Ud = 0.35
gD sin
(37)
where the Froude number Frj is dened as:

Fr j =
P
P
ds = Pps
(L x)
s
s
(30)
(P) +
(Pjg ) = 0
t
s
Cd = 1.2
(28)
Finally, the pressure at the top of the pipeline can be obtained by

integrating Eq. (28) as:
2.2. Riser
(23)
1
jl =
2
(22)
=0
jl
ul =
1 p
g =
Fig. 4. Denition of variables at the riser.
(38)
gD
Considering as the state variables in the riser the void fraction,

pressure and total supercial velocity (functions of position and
time), Eqs. (30) and (31) can be nally rewritten as:
Dg
j
+ (Cd j + Ud )
=0
Dt
s
s
(39)
G Model
ARTICLE IN PRESS
J.L. Bali
Dg P
j
+P
=0
Dt
s
(40)
where:
Dg
=
+ ug
Dt
t
s
(41)
A mixture momentum equation is considered, in which the

inertial terms corresponding to the liquid and gas phases were
included:
Dg ug
fm

P
Du
(1 )l l l
= m g sin + 2 j
j
g
D
Dt
Dt
s
m = l (1 ) +
Rg Tg
(42)
(43)
Dl
=
+ ul
Dt
t
s
(44)
where fm is the Fanning friction factor (dependent on the Reynolds

number and the relative roughness /D, where is the pipe
roughness), calculated from [12] using a homogeneous mixture
two-phase model and m is the mixture density.
2.2.1. Gas lift
Gas lift is a process used to articially lift uids from wells where
there is insufcient reservoir pressure. By injecting gas, it is possible
to aerate the liquid column and to reduce the pressure gradient. Due
to this effect, it is acknowledged that gas lift can stabilize the ow
in severe slugging, although relatively large gas ow rates are necessary [16]. In the model it is considered the possibility of injecting
gas in a position sl along the riser. Considering the balance conservation equations with a gas mass source term, it can be shown
that the gas injection introduces the following discontinuities in
position in the gas supercial velocity, pressure and void fractions:
jg+ =
Rg Tg
A
gl + P jg
m
P+
P+ 1
P =
1
Rg Tg
(45)
j 2 +
g
1
+ =

2
l jl
1
Rg Tg
1
1+
j 2
1
1
(46)
jg+
(Cd j + Ud )
(47)
gl is the injected gas lift mass ow and the superscripts

where m
and + denote evaluation respectively at location upstream and
downstream of position sl .
2.2.2. Gas region
In transients where the liquid level falls below the top of the
riser (su < st ) a gas region is formed, as shown in Fig. 5. This region
is modeled considering a constant mean pressure P gr = 12 (Pt + Pu )
for the mass balance equation and friction, gravitational and inertial
terms in the momentum balance equation; Pt and Pu are respectively the pressure at the top and at the liquid level in the riser. The
resulting equations for the mass balance depends on the location
of the gas lift injection sl compared to the liquid level position su .
If su > sl , we get:
dP gr
P gr
=
(jgt ju )
st su
dt
(48)
If su sl , we get:
dP gr
P gr
=
st su
dt
jgt ju
Rg Tg
P gr A

gl
m
(49)
Fig. 5. Denition of variables at the gas region.
where jgt is the gas supercial velocity at the top of the riser, ju is
the total supercial velocity at the liquid level and st is the position
of the top of the riser.
The momentum conservation equation results, for both cases:

Pu = Pt + g g(zt zu ) +
jgr =
1
(ju + jgt )
2

djgr
2fgg
jgr
jgr

D
dt

(st su )
(50)
(51)
where fgg is the Fanning friction factor considering only gas owing,
jgr and g are respectively the mean gas supercial velocity and the
mean gas density at the gas region; zt and zu are respectively the
vertical positions at the top of the riser and at the liquid level.
2.2.3. Riser geometry
The riser geometry is characterized by the coordinates X and Z
corresponding to the top of the riser and a set of functions furnishing the ordinate z and local inclination angle as a function of the
local position s along the riser (see Fig. 4). For a constant angle riser,
for instance, it results:
= arctan
Z
X
st = (X 2 + Z 2 )
z = s sin
1/2
(52)
(53)
(54)
The denition of geometry for a catenary riser can be seen in [5].

2.2.4. Choke valve
In normal operation in petroleum production systems the choke
valve controls the ow, allowing a production compatible with the
reservoir characteristics. In severe slugging, it is acknowledged that
choking can stabilize the ow by increasing the back pressure [2,1].
For low pressures, typical of air-water laboratory systems, the valve
operates in subcritical condition; in this case, the ow depends on
the pressure difference across the valve (see Fig. 6). The model
G Model
ARTICLE IN PRESS

J.L. Bali
Fig. 8. Discretization along the characteristic directions.
Fig. 6. Denition of variables at the choke valve.
considers a valve characteristic based on the homogeneous ow

model:
1
Pt Ps = Kv mt jt
jt
(55)
mt = t gt + (1 t )l
(56)
where jt , t , mt , gt are respectively the total supercial velocity,

void fraction, mixture density and gas density, all of them evaluated
at the top of the riser, Ps is the separator pressure and Kv is the valve
constant.
Another expression for the pressure drop across the valve was
used in [16]:
Pt Ps = Cjlt
jlt
and total supercial velocity. The pipeline imposes the supercial velocities for the gas and liquid phases at the bottom of the
riser, while the riser imposes the pressure to the pipeline; these
variables are the boundary conditions for the corresponding subsystems. Additional boundary conditions are the liquid volumetric
ow rate and the gas mass ow rate at the pipeline, as well as the
gas lift mass ow rate and separation pressure at the riser.
As initial conditions, the stationary conditions were chosen, this
is, the solution of the system of equations obtained after setting
equal to zero the time derivatives.
3. Discretization and numerical implementation
(57)
where jlt is the liquid supercial velocity at the top of the riser and
C is a dimensional valve constant. According to this relationship, it
is neglected the contribution of the gas phase to pressure drop.
2.3. Coupling between pipeline and riser
Assuming the same ow passage area for the pipeline and riser,
the pressure and supercial velocities at the bottom of the riser are
continuous:
P (s = 0, t) = Pb (t)
(58)
jg (s = 0, t) = jgb (t)
(59)
jl (s = 0, t) = jlb (t)
(60)
The boundary condition for the void fraction can be obtained

from Eq. (32) evaluated at the bottom of the riser:
(s = 0, t) = b (t) =
jgb
Cdb jb + Udb
The system of equations corresponding to the stationary state,

as well as the system of dynamic equations, were discretized and
numerically implemented using the software MATLAB [17].
In the riser a moving grid method was adopted (see Fig. 8),
in which node i (1 i N 1) moves with the corresponding gas
velocity (red lines), in order to calculate the directional derivatives
of Eq. (41). Last node N moves with the liquid velocity if the liquid level falls below the top of the riser (su < st ), or remains xed
at position st otherwise. The time step tk+1 is chosen as the time
step such that the characteristic propagated from the N 1th node
intersects the position su at time tk + tk+1 if the liquid level falls
below the top level in the riser, or as the time step such that the
characteristic propagated from the N 1th node intersects position
st otherwise. Values at time tk are interpolated in order to calculate
the directional derivatives of Eq. (44), corresponding to the liquid
velocity (blue lines). An implicit scheme was used, with a predictorcorrector method for treatment of the nonlinearities. Details of the
procedure can be seen in [5].
(61)
4. Simulations
Fig. 7 shows the state variables and the coupling between the
subsystems. State variables for the pipeline are the average gas
pressure, void fraction and position of the liquid accumulation
front, while for the riser they are the local pressure, void fraction
In this section, simulations corresponding to experimental data

for a vertical riser are shown, considering the effects of the choke
valve and gas lift injection at the bottom of the riser.
Fig. 7. Coupling between subsystems.
G Model
ARTICLE IN PRESS
J.L. Bali
Fig. 9. Simulation results for jg0 = 0.063 m/s and jl0 = 0.124 m/s (case 1, Table 1).
After a nodalization study, the riser was discretized in N = 21

nodes. Input ow variables are dened in terms of the supercial velocities jg0 and jl0 at standard conditions (pressure
P0 = 1.013 bara, temperature T0 = 293 K); these supercial velocities
are related to the ows as:
jg0 =
g0
Rg T0 m
P0 A
(62)
jl0 =
Ql0
A
(63)
4.1. Data from Taitel et al. [3]

The following parameters were chosen for a comparison with
experimental data of [3]: uid parameters are
l = 1 103 kg/m/s,
g = 1.8 105 kg/m/s, l = 1 103 kg/m3 , Rg = 287 m2 /s2 /K and

Tg = 293 K; pipeline parameters are L = 9.1 m, Le = 1.69 m and = 5 ;
riser height is Z = 3 m; common parameters for pipeline and riser
are D = 2.54 102 m and = 1.5 106 m; separation pressure is
Ps = 1.013 bara. No choke valve or gas injection was considered.
Fig. 9 shows results of a simulation for jg0 = 0.063 m/s and
jl0 = 0.124 m/s for representative variables: gas (jg1 and liquid (jl1 )
supercial velocities at the bottom of the riser, position (x) of the
liquid penetration front, void fraction (1 ) and pressure (P1 ) at the
bottom of the riser, liquid level at the riser (su ) and void fraction (p )
and pressure drop (Ppe Ppt ) in the stratied region at the pipeline.
It can be seen that, in this case, the stationary state used as the
initial condition is not stable and the system goes to a limit cycle. It
can be observed a large variation of supercial velocities, compared
to the initial stationary values. Variations in the void fraction at the
pipeline are relatively small, supporting the assumption of constant
void fraction made in previous models. On the other hand, variations in pressure drop at the pipeline can be large compared to the
stationary values, particularly in the blowdown stage. It can also be
observed a liquid penetration front and a plateau in the pressure at

the bottom of the riser, characterizing the ow regime SS1.
Many parameters corresponding to the transient can be calculated from Fig. 9. Considering that the slugging cycle begins when
the gas passage at the bottom of the riser is blocked, the severe slugging period and times corresponding to different stages described
in Section 1 can be calculated. In this case, it can be seen that the
liquid level does not remain at the top of the riser. From the simulations, it is also possible to determine the pressure amplitude at
the bottom of the riser, the maximum position of the liquid penetration front in the pipeline and the minimum position of the liquid
level in the riser.
Table 1 shows a comparison of experimental and model simulated severe slugging periods for different gas and liquid supercial
velocities. The periods calculated with the model are in very good
agreement with the experimental ones. There are some cases (cases
1724, 26 and 27) in which the simulation predicts a unstable condition while the experiment reports a stable condition; as will be
seen later in this Section, these points are located close to the stability curve.
In Table 1 are also shown simulation results using the models
of [5] and [3]. As stated before, in [5] inertia was not taken into
account; besides, it was considered a constant void fraction and
negligible pressure drop across the stratied region at the pipeline.
In [3] a further simplication was made, considering the riser as a
lumped parameter system with a mean void fraction and ignoring
the friction term.
The model of [5] also predicts as unstable some stable cases close
to the stability curve (cases 17, 18, 2123) and shows overall errors
slightly higher than the present model. The model of [3] captures
better the stability curve, although predicts as stable two points
that are unstable (cases 11 and 15); the overall errors are higher.
In summary, it can be observed that the model with inertia
shows slightly better overall performance than the model of [5]
G Model
ARTICLE IN PRESS

J.L. Bali
Table 1
Comparison with experimental results [3].
Case
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Experiment
Simulation
Simulation [5]
Simulation [3]
jg0 (m/s)
jl0 (m/s)
Texp (s)
Tsim (s)
Error (%)
Tsim (s)
Error (%)
Tsim (s)
Error (%)
0.063
0.064
0.123
0.124
0.062
0.063
0.063
0.064
0.065
0.122
0.123
0.126
0.187
0.188
0.188
0.19
0.058
0.063
0.122
0.126
0.126
0.184
0.185
0.187
0.188
0.19
0.313
0.314
0.319
0.321
0.43
0.433
0.124
0.209
0.183
0.212
0.679
0.367
0.679
0.535
0.226
0.374
0.621
0.228
0.226
0.466
0.502
0.312
0.705
0.698
0.730
0.673
0.085
0.127
0.161
0.551
0.755
0.685
0.433
0.347
0.614
0.744
0.604
0.701
24
20
15
14
6
13
9
10
19
11
8
13
11
8
7
10
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
25.4
19.5
15.5
14.6
7.36
13.3
7.35
9.27
18.7
10.8
7.36
14.0
11.9
8.22
7.86
10.1
7.19
7.21
6.45
14.0
18.2
4.24
12.3
7.36
Stable
6.24
14.0
Stable
Stable
Stable
Stable
Stable
5.8
2.5
3.3
4.3
22.7
2.3
18.3
7.3
1.6
1.8
8.0
7.7
8.2
2.8
12.3
1.0
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
24.2
18.4
14.7
13.7
6.6
12.0
6.5
8.3
17.3
9.8
6.8
13.0
10.8
7.3
7.0
9.2
6.4
6.5
Stable
Stable
18.2
13.0
12.0
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
0.8
8.0
2.0
2.1
10
7.7
27.8
17.0
8.9
10.9
15.0
0.0
1.8
8.8
0.0
8.0
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
22
17
14
13
6
11
6
7
16
9
Stable
12
10
7
Stable
8
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
8.3
15
6.7
7.1
0.0
15.4
33.3
30.0
15.8
18.1
NA
7.7
9.1
12.5
NA
20.0
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
which, in turn, shows better overall performance than the model

of [3].
The errors appearing in the comparison are reasonable for a
two-phase dynamic system, considering the simplifying assumptions and that the model does not have any tting constant adjusted
from the experimental data to be predicted. Among the simplifying
assumptions, it can be mentioned the use of a drift ux correlation
valid for constant ow in ducts [14] and the determination of the
friction pressure drop with a homogeneous two-phase ow model.
As far as the author knows, there is no data for drift ux coefcients
under accelerating ow, but a more sophisticated model to calculate the friction pressure drop might improve the simulation
results.
With the simulations, it is also possible to obtain the numeric
stability curve by keeping constant a value of liquid or gas ow rate
and varying the other in xed increments until passing from one
condition (stable or unstable) to another; when this happens, the
procedure is repeated with half the increment until achieving convergence. The procedure is laborious and computationally costly.
Stability maps generated for catenary risers can be seen in [5,7].
Fig. 10 shows the numerically generated stability map for the
conditions corresponding the experimental conditions of [3]. In the
same gure the experimental data points are shown; these unstable
data points were classied as unstable fall (su < st in the transient)
or unstable no fall (su = st always in the transient), based on a
visual observation. It can be seen that the numerically generated
stability curve includes all the unstable data points. Data points
for which there is a discrepancy between experiment and stability
simulation prediction are located close to the stability curve; for
these points pressure amplitudes and liquid penetration lengths
are small and liquid level in the riser remains at the top or close
to it, making difcult to differentiate between the severe slugging
instability condition and the uctuations associated to the intermittent ow based only on a visualization. Besides, the experimental
determination of the stability curve requires a very careful control
of variables such as separator pressure and input ows; cases 5
and 7 in Table 1, for instance, show a large discrepancy in experimental period for almost the same values of supercial velocities,
indicating that these or other variables were not kept constant in
the experiment.
Fig. 10. Numerically generated stability map and data from [3].
G Model
ARTICLE IN PRESS
10
J.L. Bali
Table 2
Comparison with experimental results, choke valve, C = 1.2 105 Pa s2 /m2 [16].
Case
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Experiment
Simulation
Simulation [16]
jg0 (m/s)
jl0 (m/s)
Texp (s)
Tsim (s)
Error (%)
Tsim (s)
Error (%)
0.0753
0.1147
0.1739
0.0781
0.1181
0.0809
0.1209
0.1713
0.1209
0.1734
0.2493
0.1698
0.2474
0.2502
0.251
0.0959
0.0949
0.0959
0.0497
0.0487
0.1704
0.1693
0.0497
0.2365
0.2354
0.2386
0.1693
0.1704
0.0959
0.0497
46.6
37.6
31.8
47.5
38.5
45
39.9
31.5
Stable
Stable
Stable
Stable
Stable
Stable
Stable
54.9
43.7
32.8
56.2
43.7
50.4
42.4
31.8
43.0
36.7
30.6
35.5
28.5
25.0
22.0
17.8
16.2
3.1
18.3
13.5
12.0
6.3
1.0
NA
NA
NA
NA
NA
NA
NA
44.9
38.1
Stable
50.7
Stable
35
31
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
3.6
1.3
NA
6.7
NA
22.0
22.3
NA
NA
NA
NA
NA
NA
NA
NA
4.2. Data from Jansen et al. [16]

In [16] it was used the same test facility as in [3]. The simulations parameters were the same, except for Le = 10 m and
= 1 .
A choke valve with C = 1.2 105 Pa s2 /m2 , see Eq. (57), was
introduced at the top of the riser in order to study the inuence
of choking on severe slugging; no gas injection was considered.
velocities. As in Section 4.1, the experimental and simulated periods
are in very good agreement, with some cases (915) in which the
model predicts as unstable some stable cases close to the stability
curve. According to the model presented in Section 2.2.4, pressure drop at the choke valve is not dependent on the gas ow,
what seems to be a too simplistic assumption and a source of
uncertainties; a more sophisticated valve model could improve the
simulation results.
In Table 2 are also shown simulation results using the model
of [16], which is a modication of the model of [3] (without inertia
and friction) to take into account the choke valve. It can be observed
a better overall performance for small liquid supercial velocities
(cases 1, 2 and 4) and a better capture of the stability condition
(cases 915). On the other hand, higher errors appear for higher
liquid supercial velocities (cases 6 and 7) and some cases exist in

which a unstable ow is predicted as stable (cases 3, 5 and 8).
In another experimental campaign, a constant gas supercial
velocity at standard condition jg0gl = 0.091 m/s was injected at the
bottom of the riser in order to study the inuence of gas lift on
severe slugging; no choke valve was considered. The gas supercial
velocity and the corresponding gas mass ow are related by:
jg0gl =
gl
Rg T0 m
(64)
P0 A
velocities. Compared to the former simulations, errors are higher
for the cases with low gas supercial velocities, comparable to the
injection gas supercial velocity (cases 13, 5 and 6). Again, there
are some stable cases close to the stability line (1417) in which
the model prediction is unstable. In general, the agreement with
the experimental data is pretty good.
In Table 3 are also shown simulation results using the model of
[16], which is a modication of the model of [3] (without inertia
and friction) to take into account the gas injection. It can also be
observed higher errors for low gas supercial velocities. Besides,
there are some cases in which a unstable ow is predicted as stable
(cases 713).
Table 3
Comparison with experimental results, gas injection, jg0gl = 0.091 m/s [16].
Case
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Experiment
Simulation
Simulation [16]
jg0 (m/s)
jl0 (m/s)
Texp (s)
Tsim (s)
Error (%)
Tsim (s)
Error (%)
0.0808
0.1153
0.1702
0.2515
0.0791
0.1147
0.3125
0.1695
0.248
0.1129
0.1737
0.3215
0.2489
0.366
0.4115
0.369
0.3141
0.2528
0.2549
0.2571
0.2582
0.152
0.152
0.1542
0.1013
0.0949
0.0487
0.0476
0.0916
0.0465
0.1552
0.1552
0.0981
0.0444
21.5
18.9
14.9
13.4
24.1
20.5
10.8
16.8
13.2
27
18.8
9.5
14
Stable
Stable
Stable
Stable
29.8
23.8
18.1
13.2
36.9
27.3
11.0
19.4
13.9
28.0
19.7
11.1
13.6
9.82
8.96
9.41
10.4
38.6
25.9
21.5
1.5
53.1
33.2
1.9
15.5
5.3
3.7
4.8
16.8
2.9
NA
NA
NA
NA
29.1
21.8
16.1
11.6
33.8
24.4
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
Stable
35.3
15.3
8.1
13.4
40.2
19.0
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
G Model
ARTICLE IN PRESS

J.L. Bali
11
Fig. 11. Simulation results for jg0 = 0.063 m/s, jl0 = 0.124 m/s and kv = 100.
4.3. Severe slugging mitigation strategies

In this section some simulations are shown, assessing different
strategies used to mitigate or suppress severe slugging. Choking,
gas injection and separation pressure increase are studied through
a comparison between the case shown in Fig. 9 (considered as reference) and simulations in which suitable parameters are varied.
4.3.1. Choking
Fig. 11 shows simulation results for the conditions of Fig. 11,
except that a choke valve with kv = 100 was added. The following
observations can be made by making a comparison with Fig. 9.
It can be observed that the ow is also unstable and a limit
cycle develops from the stationary state. As choking (as well as any
stabilizing action) modies the stability map and shifts the operating point closer to the stability line, it takes longer to develop the
limit cycle. As the ows of the phases are kept constant, there is
an increase in pressure in the stationary state and also in the mean
pressure during the transient. This pressure increase gives rise to a
decrease in the severe slugging period and in the uctuations of the

variables of interest. For this condition, it can be observed that the
liquid level remains at the top of the riser. Besides, there is a much
smaller liquid penetration front but the pressure at the bottom of
the riser does not show any plateau, indicating that the riser is not
completely lled with liquid and characterizing the ow regime
SS2.
except that a choke valve coefcient was set to kv = 150. It can be
observed that the small numerical perturbations (coming from the
simulation with a dynamic computer code using an initial condition
calculated with a stationary computer code) fade away, reaching a
steady state. In this case, the ow is stable.
4.3.2. Gas injection

except that gas was injected at the bottom of the riser with a supercial velocity jg0gl = 0.1 m/s.
Fig. 12. Simulation results for jg0 = 0.063 m/s, jl0 = 0.124 m/s and kv = 150.
G Model
ARTICLE IN PRESS
12
J.L. Bali
Fig. 13. Simulation results for jg0 = 0.063 m/s, jl0 = 0.124 m/s and jg0gl = 0.1 m/s.
Fig. 14. Simulation results for jg0 = 0.063 m/s, jl0 = 0.124 m/s and jg0gl = 0.3 m/s.
Fig. 15. Simulation results for jg0 = 0.063 m/s, jl0 = 0.124 m/s and Ps = 1.3 bara.
G Model
ARTICLE IN PRESS
J.L. Bali
13
Fig. 16. Simulation results for jg0 = 0.063 m/s, jl0 = 0.124 m/s and Ps = 1.7 bara.
Although the quantity of injected gas in the riser is greater than

the quantity of gas in the pipeline, the ow is still unstable. Making
a comparison with Fig. 9, it can be observed a longer time to achieve
the limit cycle and a shorter severe slugging period. Although there
is a liquid penetration front in the pipeline, the gas supercial velocity and void fraction are always positive in the riser due to the
gas injection. Pressure drop across the riser and uctuations of the
variables of interest are reduced; the liquid level remains at the
top of the riser. Increasing the gas injection supercial velocity to
jg0gl = 0.3 m/s the ow stabilizes, as shown in Fig. 14.
4.3.3. Separation pressure
except that the separation pressure was increased to Ps = 1.3 bara.
For this condition, the ow is still unstable. Making a comparison
with Fig. 9, it can be observed a longer time to achieve the limit
cycle and a shorter severe slugging period. There is a reduced liquid
penetration front in the pipeline, while the liquid level remains at
the top of the riser. Pressure drop across the riser and uctuations
of the variables of interest are reduced. Increasing the separation
pressure to Ps = 1.7 bara the ow stabilizes, as shown in Fig. 16.
5. Conclusions
A mathematical model and numerical simulations corresponding to severe slugging in air-water pipeline-riser systems, are
presented. The model is an improvement of the one previously published by the author [5], including inertial effects. Inertial effects
are taken into account by using the rigid water-hammer approximation, which was numerically implemented without increasing
substantially the complexity of the model. The inclusion of inertial effects allows the riser for dealing with discontinuous pressure
boundary conditions without showing unrealistic discontinuities
in the supercial velocities and void fraction.
A comparison is made with experimental results published in
literature for vertical risers including the effect of a choke valve
at the top and gas injection at the bottom of the riser, showing
very good results for slugging cycles and stability maps. The present
model predicts the period of severe slugging with reasonable accuracy. In some cases, the model predicts as unstable, cases that
are stable; in this way, the unstable region in the stability map is
overestimated. Other simpler models, not taking into account inertia and friction [3,16], predict some unstable conditions as stable,
which is not safe for design purposes. Simulation performance can
be improved by using more sophisticated models for the friction

term in the riser, as well as for the choke valve. An important issue
in making a comparison between experimental data and simulation
is the measurement and control of the inlet ows, in order to have
accurate and constant parameters used as boundary conditions.
Simulations were made assessing the effect of different strategies to mitigate severe slugging, such as choking, gas injection and
increase in separation pressure, showing correct trends. Through
these studies it is possible to built stability and ow pattern maps
for different sets of system parameters.
Acknowledgements
This work was supported by Petrleo Brasileiro S. A. (Petrobras).
The nancial support of Fundaco de Amparo Pesquisa do Estado
de So Paulo (FAPESP) to attend IMAACA 2012 (where a shorter
version of this paper received the Best Paper Award) is deeply
acknowledged.
References
[1] Y. Taitel, Stability of severe slugging, International Journal of Multiphase Flow
12 (2) (1986) 203217.
[2] Z. Schmidt, Experimental study of two-phase slug ow in a pipeline-riser system, The University of Tulsa, Tulsa, USA, 1977 (Ph.D. thesis).
[3] Y. Taitel, S. Vierkand, O. Shoham, J.P. Brill, Severe slugging in a riser system:
experiments and modeling, International Journal of Multiphase Flow 16 (1)
(1990) 5768.
[4] C. Sarica, O. Shoham, A simplied transient model for pipeline-riser systems,
Chemical Engineering Science 46 (9) (1991) 21672179.
K.P. Burr, R.H. Nemoto, Modeling and simulation of severe slugging
[5] J.L. Balino,
in air-water pipeline-riser systems, International Journal of Multiphase Flow
36 (2010) 643660.
[6] C. Wordsworth, I. Das, W.L. Loh, G. McNulty, P.C. Lima, F. Barbuto, Multiphase
Flow Behavior in a Catenary Shaped Riser. CALTEC Report No.: CR 6820, Bedford,
UK, 1998.
Modeling and simulation of severe slugging with mass
[7] R.H. Nemoto, J.L. Balino,
transfer effects, International Journal of Multiphase Flow 40 (2012) 144157.
[8] S. Mokhatab, Severe Slugging in Offshore Production Systems, Nova Science
Publishers, Inc., New York, 2010.
[9] J.M. Masella, Q.H. Tran, D. Ferre, C. Pauchon, Transient simulation of two-phase
ows in pipes, International Journal of Multiphase Flow 24 (1998) 739755.
[10] Y. Taitel, A.E. Dukler, A model for predicting ow regime transitions in horizontal and near horizontal gasliquid ow, American Institute of Chemical
Engineers Journal 22 (1) (1976) 4755.
[11] M.H. Chaudhry, Applied Hydraulic Transients, Van Nostrand Reinhold, New
York, USA, 1987.
[12] N.H. Chen, An explicit equation for friction factor in pipe, Industrial and Engineering Chemical Fundamentals 18 (1979) 296297.
G Model
ARTICLE IN PRESS
14
J.L. Bali
[13] S.L. Kokal, J.F. Stanislav, An experimental study of two-phase ow in slightly

inclined pipes I. Flow patterns, Chemical Engineering Science 44 (1989)
665679.
[14] K.H. Bendiksen, An experimental investigation of the motion of long bubbles in
inclined tubes, International Journal of Multiphase Flow 10 (4) (1984) 467483.
[15] B. Chexal, G. Lellouche, J. Horrowitz, J. Healer, A void fraction correlation for
generalized applications, Progress in Nuclear Energy 27 (4) (1992) 255295.
[16] F.E. Jansen, O. Shohan, Y. Taitel, The elimination of severe slugging experiments and modeling, International Journal of Multiphase Flow 22 (6) (1996)
10551072.
[17] E.B. Magrab, S. Azarm, B. Belachandran, J.H. Duncan, K.H. Herold, G.C. Walsh,
An Engineer s Guide to MATLAB, Pearson Prentice Hall, New Jersey, USA, 2010.
was born in Buenos Aires, Argentina.

Jorge Luis Balino
He graduated (1983) and made a PhD (1991) in Nuclear
Engineering at Instituto Balseiro, Argentina. He worked
for echint S.A. (19831984), Centro Atmico Bariloche and
Instituto Balseiro (19852000) in Argentina and Instituto
de Pesquisas Energticas e Nucleares (20012003) at So
Paulo, Brazil. Since 2004 he is Professor at Universidade
de So Paulo. His research interests are Fluid Dynamics,
Heat Transfer and Multiphase Flow.

Severe Slugging 1 PDF

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Severe Slugging 1 PDF

Încărcat de

Drepturi de autor:

Formate disponibile

G Model

JOCS-225; No. of Pages 14

Contents lists available at ScienceDirect

Journal of Computational Science

Modeling and simulation of severe slugging in air-water systems

E-mail addresses: jlbalino@usp.br, balinoj@yahoo.com

continues to ow in and gas already in the riser continues to ow

Fig. 1. Stages for severe slugging (from [1,2]).

A classication of severe slugging can be made, according to the

mixture momentum equation in which only the gravitational term

discontinuities such as liquid accumulation in the piping and liquid

of the riser was assured by suitable pipeline switching conditions

Fig. 2. Denition of variables at the pipeline.

These approximations are made in order to get a lumped parameter

From Eqs. (4) and (6), it results:

Applying mass conservation equation for the liquid and gas

Applying the momentum conservation equation in the control

2.1.3. Switching between conditions

From the kinematic condition at the liquid penetration front, we

Assuming that commutation from the state x = 0 to x > 0 happens

JOCS-225; No. of Pages 14

no / Journal of Computational Science xxx (2013) xxxxxx

Fig. 3. Stratied ow at the pipeline.

the total supercial velocity, pressure and time derivative of the

+ (l g )Ag sin A

where jg and jg are respectively the mean supercial velocities

Continuity equations for the phases are considered at the riser

It will be assumed that the drift parameters Cd and Ud depend

gD 0.35 sin  + 0.54 cos 

For Frj 3.5:

where the Froude number Frj is dened as:

Finally, the pressure at the top of the pipeline can be obtained by

Fig. 4. Denition of variables at the riser.

Considering as the state variables in the riser the void fraction,

A mixture momentum equation is considered, in which the

where fm is the Fanning friction factor (dependent on the Reynolds

gl is the injected gas lift mass ow and the superscripts

Fig. 5. Denition of variables at the gas region.

The denition of geometry for a catenary riser can be seen in [5].

JOCS-225; No. of Pages 14

no / Journal of Computational Science xxx (2013) xxxxxx

Fig. 8. Discretization along the characteristic directions.

Fig. 6. Denition of variables at the choke valve.

considers a valve characteristic based on the homogeneous ow

mt = t gt + (1 t )l

where jt , t , mt , gt are respectively the total supercial velocity,

The boundary condition for the void fraction can be obtained

The system of equations corresponding to the stationary state,

In this section, simulations corresponding to experimental data

Fig. 7. Coupling between subsystems.

After a nodalization study, the riser was discretized in N = 21

4.1. Data from Taitel et al. [3]

g = 1.8 105 kg/m/s, l = 1 103 kg/m3 , Rg = 287 m2 /s2 /K and

observed a liquid penetration front and a plateau in the pressure at

JOCS-225; No. of Pages 14

no / Journal of Computational Science xxx (2013) xxxxxx

which, in turn, shows better overall performance than the model

4.2. Data from Jansen et al. [16]

liquid supercial velocities (cases 6 and 7) and some cases exist in

JOCS-225; No. of Pages 14

no / Journal of Computational Science xxx (2013) xxxxxx

4.3. Severe slugging mitigation strategies

gD 0.35 sin + 0.54 cos