Chinese J. Chem. Eng.

, 14(5) 626—633 (2006)

Development of Liquid Slug Length in Gas-Liquid Slug Flow along

Horizontal Pipeline: Experiment and Simulation*

WANG Xin(王鑫), GUO Liejin(郭烈锦)** and ZHANG Ximin(张西民)

State Key Laboratory of Multiphase Flow in Power Engineering, Xi’an Jiaotong University, Xi’an 710049, China

Abstract The liquid slug length distribution is crucial for designing the downstream processing system with mul-
tiphase pipeline. Experiments were conducted in a 133m long horizontal test loop. The measurements were per-
formed by conductivity probes to determine the liquid slug length distribution. The data covered both the slug and
plug flow regimes. From experimental results, the mean liquid slug lengths were relatively insensitive to gas and
liquid flow rates in the higher mixture velocity range. But in the lower mixture velocity range, the mean liquid slug
length decreased and then increased with mixture velocity. It shows that the development length of slug flow was
longer than x/D=1157. A slug tracking model was adapted to study the evolution of liquid slug length distribution in
a horizontal pipeline. In the present model, the wake effect of elongated bubble and the pressure drop due to accel-
eration are taken into account and random slug lengths are introduced at the entrance. The results of the model are
compared with the measured slug length distributions of slug flow regime. It shows that the predicted mean and
maximum slug lengths are in agreement with the experimental data at x/D=1157 and the form of the slug length
distributions is also predicted well by the model.
Keywords two-phase flow, slug flow, liquid slug length, slug tracking

1 INTRODUCTION tional velocity of elongated bubble, while the transla-

In petroleum industry, the slug flow is a fre- tional velocity is related to the maximum local veloc-
quently encountered flow regime in multiphase flow ity of the slug tail. At the back of short liquid slug, the
pipeline. For pipeline designers, the liquid slug length velocity profile is not developed as fully as the longer
distribution is important for the proper design of one, therefore short slugs are unstable and easily
downstream facilities, such as slug catcher and sepa- overtaken by the following elongated bubble. When a
ration system. However, for its transient and unsteady short slug collapsed, the trailing liquid slug scoops the
nature, it is a great challenge for engineers to correctly liquid and its length increases. Moissis and Griffith[2]
predict the flow parameters of slug flow, especially studied the collapsing rate of liquid slugs due to wake
the maximum liquid slug length. effect in upward vertical flow. Cook and Behnia[3]
The unit cell model for slug flow in horizontal studied this effect in nearly horizontal pipe.
pipe was first proposed by Dukler and Hubbard[1]. From Prudhoe Bay pipeline data, Brill et al.[4]
Based on this approach, some researchers developed observed that the liquid slug length could be fitted by
modified models to predict flow parameters of slug a log-normal distribution and Nydal et al.[5] obtained
flow. But all the models assumed the constancy of the same conclusion from experimental data. Zheng et
liquid slug length and could not predict the liquid slug al.[6] proposed a primary slug tracking model to study
length distribution. the slug flow behavior in hilly terrain pipeline. Barnea
From observation, short liquid slugs generated in and Taitel[7] also used a simple slug tracking model to
the entrance section of horizontal pipe may be several calculate liquid slug length evolution along a vertical
times the pipe diameter and the frequency of liquid pipe. In this model the liquid film in Taylor bubble
slugs is relatively high. When the short slugs move region and the dispersed bubble in liquid slug were
downstream, some elongated bubbles disappear and neglected and an assumed relation of the translational
short slugs merge to form longer liquid slugs. In this velocity was specified to consider the wake effect.
course the frequency of liquid slug decreases and Cook and Behnia[3] compared this model with ex-
reaches a stable value when most liquid slugs are perimental slug length distribution for nearly horizon-
longer than the minimum stable slug length. The sur- tal pipe. van Hout et al.[8] used this model to study
vival of liquid slug is controlled by the pick-up proc- liquid slug length distribution in vertical flow.
ess at the slug front, where the moving slug front Nydal and Banerjee[9] developed a Lagrangian
scoops the liquid film in elongated bubble region, and slug tracking model, which was based on the balances
by the liquid shedding process at the tail of liquid slug. of mass and momentum. The important assumption
The liquid shedding rate is determined by the transla- was that no gas was entrained into the liquid slug.

Received 2005-09-12, accepted 2006-04-14.

* Supported by the National Natural Science Foundation of China (No.50536020 and 50323001).
** To whom correspondence should be addressed. E-mail: lj-guo@mail.xjtu.edu.cn
 Development of Liquid Slug Length in Gas-Liquid Slug Flow along Horizontal Pipeline: Experiment and Simulation 627

Figure 1 Schematic diagram of the experimental facility

P1, P2, P3—pressure transducers; CP1, CP2, CP3, CP4—conductivity probes;
1—centrifugal pump; 2—gear pump; 3—screw compressor; 4—electromagnetic flowmeters;
5—mass flowmeters; 6—vortex flowmeter; 7—orifice plate flowmeters

－1
and VGSO=1－20m⋅s 1.
－
Predicted results of the model were compared with the sponded to VLS=0.16－1.5m⋅s
measured data of slug flow in hilly terrain pipeline.
Based on the unit cell model presented by Dukler and 2.1 Experimental facility
Hubbard[1], Taitel and Barnea[10,11] proposed a slug Figure 1 illustrates the experimental facility. Wa-
tracking model to study the slug flow in hilly terrain ter was pumped and measured by an electromagnetic
pipeline. Al-safran et al.[12] verified the slug tracking flowmeter with precision of 0.5%. Oil was supplied
model with two-phase flow experimental data ac- by a gear pump from an oil tank and then measured by
quired in a 420 m long hilly terrain flow loop and the a CMF050 or CMF200 micro motion mass flowmeter
agreement between the measured slug length distribu- with precision of 0.1%. Air was compressed by the
tion along pipeline and the predicted data was good, screw compressor into a buffer vessel. Then it was
although only two cases were provided. However, the filtered and measured by an orifice plate or a digital
model[10,11] only took into account the terrain slugs vortex flowmeter (accuracy 1%) before entering the
and neglected the wake effect of hydrodynamic slugs. mixer. A horizontal plate was fixed in the mixer pipe
Slugs could dissipate at top elbows or be generated at to get stratified flow regime at the initial part of the
bottom elbows due to the gravity force. In addition, test section, as shown in Fig.1.
the pressure drop due to acceleration was not included The experimental facility consisted of a 133m
in this model[10,11]. Also the lengths of the liquid slug long horizontal test section and finally a 16m vertical
and the elongated bubble that enter the pipe were pipe. Most segments of the test loop were made of
specified and a constant stable slug length 50D or stainless steel pipe, but several segments of the
150D was used in the model. straight pipe sections were made of PVC pipe and
In this work, the measurement of liquid slug were transparent for visual observation. The inner
length conducted in a 133m long horizontal and diameter of the test loop was 50mm. At the end of the
φ 50mm test loop was presented, over a wide range of vertical pipe, the gas-liquid mixture was separated in a
input gas and liquid velocities. The slug tracking separator which was operated under atmospheric con-
model[10,11] was adapted to take into account the hy- dition. The air was vented to atmosphere and the liq-
drodynamic slugs that assumed a stochastic normal uid was returned to the liquid-liquid separator tank,
distribution of liquid slug lengths at the pipe entrance. from which oil and water flowed back to their storage
Also, the wake effect and the acceleration pressure drop tanks separately.
were included. The liquid slug lengths predicted by the Air and water were used as the test fluids in this
model were then verified with the measured data. study. The flow regime was slug flow or plug flow in
the horizontal pipeline. Measurement instruments in-
2 EXPERIMENTAL cluding Keller pressure and differential pressure
An experimental facility was designed and con- transducers (accuracy 0.1%) were installed along the
structed in this study to determine liquid slug length test loop. The liquid slug and elongated bubble inter-
distributions. The parameter values investigated corre- faces were measured by conductivity probes[13]. The

Chinese J. Ch. E. 14(5) 626 (2006)

628 Chinese J. Ch. E. (Vol. 14, No.5)

hardware of the data acquisition system consisted of a

PC and a high speed multifunction board PCI-6071E
made by National Instruments Inc. USA. The data
acquisition and analyzing program was developed in
the graphical programming system of LabVIEW6.0.
All the concerned signals were sampled at 2000 Hz.

2.2 Measurement techniques and data processing

The measurement of electrical impedance in
two-phase flow is a very common method to deter-
mine the local or area average void fraction in multi-
phase flow research. In this study the technique was
used to study the elongated bubbles in slug flow. A
single sensor conductivity probe is shown schemati-
cally in Fig.2. The needle is electrically isolated by a
thin layer of lacquer but the tiny tip is exposed. It is
assembled into a thin stainless steel tube and the tube
is sealed at both ends. The circuitry is designed that a
high voltage signal is output when the tip is in gas Figure 3 Normalized output signals of two successive
phase and a low voltage signal occurs when the tip conductivity probes (VLS=0.72m⋅s－1, VGSO=10.5m⋅s－1)
contacts liquid phase. Therefore, the sharp voltage
transitions indicate the interfaces between gas and passing the adjacent probes, as reported by Fagundes
liquid. The tips of the probes were close to the upper et al.[14]. The uncertainty of Δt is equal to the sampling
wall of the pipe and the distance was 5mm. Two pairs period, i.e. 1/2000s. For all experimental conditions,
of conductivity probes mounted in the PVC pipe sec- the maximum of the relative uncertainty of the elon-
tions were used to detect the gas-liquid interfaces. The gated bubble velocity is 2.8%, which is smaller than
first pair(CP1 and CP2 in Fig.1) was located at the maximum error 5% permitted in the calculated
x/D=1157 after the entrance and the second pair(CP3 velocity, proposed by Pinto and Campos[15]. The rela-
and CP4) was located at x/D=2609. The longitudinal tive uncertainty of the liquid slug length is always
spacing of two probes is L=0.50m. The signals of the smaller than 3.3%.
probes are shown in Fig.3. By cross-correlation of the
signals, the transit time Δt is determined and the 3 MODEL
translational velocity of the elongated bubble can be A typical slug unit consists of a liquid slug region
obtained by Vt=L/Δt. Then the elongated bubble length of length LS and elongated bubble region of length Lf,
and the liquid slug length can be calculated. Several under the gas bubble there is a liquid film. The main
hundreds of slug units were recorded and processed assumption of the unit cell model[1] is that slug flow is
for each set of test conditions and the mean values and comprised of a series of slug units in which every liq-
the statistical distributions of each hydrodynamic pa- uid slug has constant length. However, in the slug
rameter were obtained. tracking model[10,11] every liquid slug has its individ-
The accuracy in determining the translational ual length and the velocities of liquid slug front and
velocity of elongated bubble depends on the accuracy tail are calculated respectively. Therefore, the lengths
in the estimation of the transit time Δt for a bubble of all liquid slugs in the pipeline can be determined.
Some liquid slugs may collapse when their lengths are
shorter than a critical length. To simplify the calcula-
tion a uniform liquid film thickness is used and it
equals the equilibrium one.
As shown in Fig.4, the position of slug front is Zfr
and the slug tail is Zta; the front and tail have velocity
Ufr and Uta respectively. These parameters are calcu-
1.2mm
1.2mm lated, as well as the fluid velocities, pressures and
holdups in every liquid slug and elongated bubble re-
gion. Because a constant stable slug at the entrance
was assumed in the model of Taitel and Barnea[10,11]
0.4mm
0.4mm and the wake effect was also not included, the model
5mm
5mm would not predict the evolution of the slug length dis-
tribution along a horizontal pipe due to no collapsing
Figure 2 Conductivity probe assembly of short liquid slugs. In this study, a stochastic normal
October, 2006
 Development of Liquid Slug Length in Gas-Liquid Slug Flow along Horizontal Pipeline: Experiment and Simulation 629

Figure 4 Sketch of slug flow units in horizontal pipe[10]

distribution of input liquid slug lengths is employed slug:

and the input slug units at the entrance are generated Fr − F0
by a steady state unit cell model. The wake effect is RS = 1 − (4)
Fr + F1
included and short liquid slugs may collapse and the
coalescence of elongated bubbles occurs. Due to the with
underprediction of pressure drop at high flow rates Fr = U S gD
when the acceleration pressure drop is ne-
F0 = 2.6 ⎡⎣1 − 2 ( D0 D ) ⎤⎦ , D0 = 2.5cm
2
glected(He[16]), it is considered in this study.
⎛ 1 ⎞
3.1 The translational velocity of elongated bubble F1 = 2400 ⎜1 − sin ϕ ⎟ B0−3 / 4
and the wake effect ⎝ 3 ⎠
The correlation proposed by Nicklin et al.[17] is B0 = ( ρ L − ρG ) gD 2 / σ
applied to calculated the translational velocity of elon- 3.3 Pressure drop calculation
gated bubble: There are three contributions to the pressure drop
Vt = C0Vm + V0 (1) in slug flow: the frictional and the gravitational pressure
where V0 is the drift velocity of elongated bubble in drop and the acceleration pressure drop. Dukler and
stagnant liquid, and Vm is the mixture velocity Hubbard[1] and Nicholson et al.[20] proposed that the
(Vm=VLS+VGS). The coefficient of C0 and the drift ve- acceleration pressure drop results from the acceleration
locity V0 is predicted by the correlation reported by of the slow moving liquid film to the slug velocity:
Bendiksen[18]: ΔPacc = ρ L RS (U S − U f )(Vt − U S ) (5)
C0 = 1.05 + 0.15sin 2 ϕ The study of He [16]
showed that the pressure
Vm2 would be underpredicted if the acceleration pressure
V0 = gD ( 0.35sin ϕ + 0.54 cos ϕ ) for Fr 2 ≡ ≤ 3.52 drop was neglected at high mixture velocities. How-
gD
ever, the pressure would be overpredicted when the
C0 = 1.2 acceleration pressure drop was included. From ex-
Vm2 perimental observation, when the input mixture ve-
V0 = gD ( 0.35sin ϕ ) for Fr 2 ≡ > 3.52 (2) locities was low, it was found that only partial of liq-
gD
uid film would be accelerated to the slug velocity and
With the wake effect of a leading bubble, a short
the slug may move on a thin layer of liquid film which
trailing liquid slug is not stable, so that the elongated
had slow velocity. Therefore, a coefficient Cacc is used
bubble following it will accelerate. Cook and Behnia[3]
to consider the proportion of the liquid film scooped
studied the relation between the translational velocity of
by the mixing vortex in the liquid slug. Bendiksen et
elongated bubble to the liquid slug length ahead of it in
al.[21] observed that the mixing zone at the slug front is
a 16m tube inclined at +5° to the horizontal. They found
short when the mixture velocity is lower, but in the
that the wake effect can be correlated by the equation:
higher mixture velocity range the mixing zone is
Vtb ⎛ L ⎞ longer and maybe equal to half of liquid slug length.
= 1.0 + 0.56exp ⎜ −0.46 S ⎟ (3)
Vt ∞ ⎝ D⎠ For simplification, we assume that the acceleration
where Vt∞ is the elongated bubble translational velocity pressure drop only exists in the front half of liquid
behind a long developed liquid slug and calculated by slug. Due to the force balance between points i and
Eq.(1). This correlation is used in this study. i+1, the pressure difference is given by
1 S
3.2 Liquid holdup in the liquid slug region PS,i+1 − PS,i = f S,i ρ S,i LS,i U S,i U S,i +
4 A
In this work the correlation proposed by An-
dreussi and Bendiksen[19], which is based on a physi-
1 1 S
ρ S,i gLS,i sin ϕ + f S,i+1 ρ S,i+1 LS,i+1
cal model for the entrainment and transport processes 2 4 A
for the dispersed bubbles in the liquid slug region, is 1
U S,i+1 U S,i+1 + ρ S,i+1 gLS,i+1 sin ϕ +
used to calculate the liquid holdup within the liquid 2
Chinese J. Ch. E. 14(5) 626 (2006)
630 Chinese J. Ch. E. (Vol. 14, No.5)

⎛ τ f Sf τ G SG ⎞ was found that the collapse of liquid slugs occurred

⎜ A + A + ρf g sin ϕ ⎟ Lf ,i + and the flow was still developing after x/D=1157. The
⎝ ⎠f ,i
minimum stable slug length was longer than 10D that
Cacc ΔPacc,i+1 (6) suggested by Cook and Behnia[3].
where ρf is defined as
ρf = Rf ρ L + (1 − Rf ) ρG (7)
A more detailed description of the other equa-
tions to calculate the parameters, such as the slug front
velocity Ufr, the slug velocity US, the liquid film ve-
locity Uf and holdup Rf , is found in Taitel and Ba-
rnea[10,11] .
The model was implemented in C++ using ob-
ject-oriented techniques. Dynamic memory allocation
and linked lists were used. In the calculation, the input
air and water flow rates and the pressure at the outlet
were specified. From calculation it was found that the
mean slug length and the standard deviation of the
predicted length distributions were insensitivity to the
chosen form and range of input slug lengths. In this
study the slug lengths introduced into the pipe were
random between 3D and 5D and followed a normal
distribution. The code was run and the Cacc was tuned
until the averaged inlet pressure (P1 in Fig.1) was
correctly predicted. It was found that Cacc varied from
0 to 0.3. Then the development of liquid slug length
along the pipeline of a long time was simulated and
the data of thousands of slugs were collected.

4 RESULTS AND DISCUSSION

Some researchers reported that slug lengths were Figure 5 Measured mean liquid slug length vs. mixture
quite constant for a large range of mixture velocities velocity at two locations (The error bar represents the standard
deviation from the mean value)
for air-water flow in short horizontal pipes. Dukler
and Hubbard[1] observed Ls=12－24D. Nicholson et
al.[20] reported Ls=12－30D. Andreussi and Bendik- Ferre[22] observed that the mean liquid slug
length increases in the range of 0.5＜VGSO＜4m⋅s 1 in
－
sen[19] obtained Ls=12－24D. The result of Nydal et
[23]
al.[5] was Ls=15－20D and He[16] was Ls=10－34D. a 4.5cm pipe that was 50m in length. Woods re-
Fig.5 shows the mean liquid slug lengths measured at ported that the mean liquid slug length increases
slightly in the range of 1.0＜VGSO＜3.0m⋅s 1 in a pipe
－
x/D=1157 and x/D=2609. It is found that the mean
slug lengths from both locations are almost independ- with a diameter of 7.63cm and a length 19.8m. The
ent on the mixture velocities only in the higher mix- trend was also observed in this study. Fig.6 shows the
ture velocity range, but in the range of lower mixture mean liquid slug lengths measured at x/D=1157 as a
velocity, the slug lengths are smaller than the values at function of VGSO for four constant liquid superficial
velocities. While VGSO is higher than 6m⋅s 1, the mean
－
higher mixture velocities and the slug length decreases
and then increases as the mixture velocity increases. liquid slug length is weakly dependent on VGSO and
VLS. However in the range of 2.2＜VGSO＜6 m⋅s 1, the
－
This characteristic is considered to be due to the tran-
sition of flow regime from slug flow to plug flow, mean liquid slug length increases with the increasing
which was observed that the small bubbles in the liq- of VGSO and decreases with the increasing of VLS. In
the range of VGSO＜2.2m⋅s 1, the mean liquid slug
－
uid slug region disappeared in the lower mixture ve-
locity range from the signal of conductivity probes. length increases with the decreasing of VGSO and the
From both figures, it is observed that the mean liquid translational velocities of elongated bubble corre-
slugs grew from 15D－27D at x/D=1157 to 23D－ sponding to the minimum of mean liquid slug length
are 1.9, 2.6, 2.5m⋅s 1 for VLS=0.7, 1.0, 1.49m⋅s 1 re-
－ －
40D at x/D=2609 in the higher flow rates range. The
[24]
slug length measured at x/D=1157 varied in the same spectively. Grotjahn and Mewes reported that the
range as the other authors reported, but the range at transition between plug and slug flow occurs for
Vt =2.2m⋅s- 1 and longer liquid slugs appear in the
－
x/D=2609 was higher. From the minimum slug lengths
and the slug frequencies measured at both locations, it plug flow region. It is shown that the transition Vt in

October, 2006
 Development of Liquid Slug Length in Gas-Liquid Slug Flow along Horizontal Pipeline: Experiment and Simulation 631

average value of the calculated maximum slug lengths

is 48.3D and its variance is 14.3. The average of the
calculated maximum slug length is about 0.7% shorter
than the measured length. The average absolute error
is 9.2% and the maximum absolute error is 20.5%.
Note that the distance of the location from the en-
trance, the agreements between the predicted and the
measured mean and maximum liquid slug lengths are
good.

Figure 6 Mean liquid slug length at x/D=1157 vs. gas

superficial velocity for various liquid superficial velocities
VLS, m·s 1: △ 0.50; ▼ 0.70; ☆ 1.00; ○ 1.49
－

this research are close to the value suggested by the

authors[24] and the reverse trend of the mean liquid slug
length against VGSO in the range of VGSO＜2.2m⋅s 1 is
－

due to the flow regime transition.

The experimental data of slug flow regime were
simulated by the slug tracking model. Fig.7 shows the Figure 8 Comparison of measured and calculated
mean liquid slug length at x/D=1157
predicted mean liquid slug lengths at x/D=1157 for a measured; calculated
wide range of mixture velocities. As the measured
data [Fig.5(a)], the predicted mean slug lengths are
insensitivity to the mixture velocities and fairly con-
stant. Fig.8 compares the mean liquid slug lengths
predicted by the model with the measured data of 21
combinations of liquid and gas flow rates at x/D=1157.
For each experimental condition, the left unfilled
column represents the measured mean liquid slug
length and the right filled column represents the pre-
dicted value. The average value of the measured mean
liquid slug lengths is 23.4D and the variance is 2.8.
While the average value of the calculated mean liquid
slug lengths is 22.5D and the variance is 2.5. The av- Figure 9 Comparison of measured and calculated
erage of the calculated mean slug length is about 3.9% maximum liquid slug length at x/D=1157
shorter than the measured length. The average abso- measured; calculated
lute error is 5.2% and the maximum absolute error is
14.7%. The predicted and the measured maximum
The typical distributions of measured and pre-
liquid slug lengths at x/D=1157 are shown in Fig.9.
dicted liquid slug length at x/D=1157 and x/D=2609
The average value of the measured maximum slug
are presented in Fig.10 and 11. Although in the calcu-
lengths is 49.1D and the variance is 31.8. While the
lation, the normal distribution of input liquid slug
length (3D—5D) was employed at the entrance, the
statistical distributions of measured and calculated
slug length at both locations maintain right skewed
and the shapes can be fitted by a lognormal curve,
which agrees with the reports of Brill et al.[4] and Ny-
dal et al.[5]. The model predicts the standard deviations
of distribution at both locations and the mean slug
length at x/D=1157 quite well. But the mean liquid
slug length at x/D=2609 is underestimated. In the
model, the minimum stable slug length was 10D be-
cause the correlation [Eq.(3)] was used to relate the
Figure 7 Calculated mean liquid slug length of slug flow wake effect of elongated bubble. However, this corre-
regime vs. mixture velocity at x/D=1157 (The error bar lation was proposed by Cook and Behnia[3] from the
represents the standard deviation from the mean value) experiments conducted in a short inclined pipe and the
Chinese J. Ch. E. 14(5) 626 (2006)
632 Chinese J. Ch. E. (Vol. 14, No.5)

slug lengths increased clearly after x/D=1157 and a

minimum stable slug length longer than 10D was ex-
isted. Therefore the short minimum stable slug length
assumed in the model is an important reason to un-
derpredict liquid slug length at x/D=2609.
Figure 12 presents the variation of the lengths of
9 consecutive liquid slugs along the pipeline from the
model output. The initial slug lengths were random
between 3D to 5D. When a slug length was shorter
than the critical length, 1D specified in this study, the
liquid slug dissipated. While all the liquid slugs were
initially shorter than the minimum stable length of
10D assumed in the model, some slugs disappeared
and the other grew to stable length and survived in the
pipeline. For instance, the collapse of the second slug
caused the following slug, the third, to grow to 8.7D
as it accumulated the liquid from the second slug,
however, its length was less than 10D and finally dis-
appeared. The fourth slug also collapsed. The liquid of
the second, the third and the fourth slug were collected
by the fifth slug and then its length increased to 18.5D
that was longer than 10D and stable. Thereafter, be-
cause there were no slugs ahead of it collapsing again,
its length only increased 2.5D when it moved along
the pipeline. It is seen that the minimum stable slug
Figure 10 Distributions of measured liquid slug length is a key factor in the slug tracking model to
length at two locations (VLS=1m⋅s 1, VGSO=5.8m⋅s 1)
－ －

simulate the hydrodynamic slug flow.

Figure 12 Development of a sequence of liquid slugs

(VLS=1m⋅s－1, VGSO=10.2m⋅s－1)

5 CONCLUSIONS
A slug tracking model, based on the work of
Taitel and Barnea[10,11], was adapted for the hydrody-
namic slug flow in horizontal pipeline, that included
the wake effect and the pressure drop due to accelera-
tion and assumed a random slug length distribution
introduced at the entrance. The model was imple-
mented in C++ using object-oriented techniques and
the evolution of the slug length distribution along the
pipeline was simulated.
Figure 11 Distributions of calculated liquid slug
length at two locations (VLS=1m⋅s－1, VGSO=5.8m⋅s－1) Experimental study of gas-liquid slug flow was
carried out in a 133m long horizontal test loop and the
maximum mixture velocity was only 2.5m⋅s 1, which
－
measurements were performed by conductivity probes.
were different from the conditions in this study. From The liquid slug length and the other parameters were
the experimental results, it was found that the mean determined from two locations positioned at the
October, 2006
 Development of Liquid Slug Length in Gas-Liquid Slug Flow along Horizontal Pipeline: Experiment and Simulation 633

middle and near the outlet of the pipeline. The mean horizontal gas-liquid intermittent flow”, Chem. Eng. Sci.,
liquid slug lengths are relatively insensitive to the gas 55(11), 2009—2018(2000).
4 Brill, J.P., Schmidt, Z., Coberly, W.A., “Analysis of
and liquid flow rates in the higher mixture velocity two-phase tests in large-diameter flow lines in Prudhoe
range. But in the lower mixture velocity range, the Bay field”, Soc. Pet. Eng. J., 271, 363—377(1981).
mean liquid slug length decreases and then increases 5 Nydal, O.J., Pintus, S., Andreussi, P., “Statistical charac-
with the mixture velocity. The transition from slug terization of slug flow in horizontal pipes”, Int. J. Multi-
phase Flow, 18(3), 439—453(1992).
flow to plug flow occurred in the lower mixture ve- 6 Zheng, G., Brill, J.P., Taitel, Y., “Slug flow behavior in a
locity. The development length of slug flow was hilly terrain pipeline”, Int. J. Multiphase Flow, 20(1),
longer than x/D=1157 and after this distance the liquid 63—79(1994).
slug collapse and growth were also observed. 7 Barnea, D.A., Taitel, Y., “A model for slug length distri-
bution in gas-liquid slug flow”, Int. J. Multiphase Flow,
The results of the model were compared with the 19(5), 829—838(1993).
experimental data of slug flow regime. The compari- 8 van Hout, R., Barnea, D., Shemer, L., “Evolution of sta-
son revealed that the predicted mean and maximum tistical parameters of gas-liquid slug flow along vertical
slug lengths were in agreement with the measured data pipes”, Int. J. Multiphase Flow, 27(9), 1579 —
1602(2001).
at x/D=1157 and the forms of the distributions were 9 Nydal, O.J., Banerjee, S., “Dynamic slug tracking simu-
also predicted well by the model. However, the mean lations for gas-liquid flow in pipelines”, Chem. Eng.
slug length at x/D=2609 was underestimated, which Comm., 141—142, 13—39(1996).
was caused by the short minimum stable slug length 10 Taitel, Y., Barnea, D., “Effect of gas compressibility on a
assumed in the model. slug tracking model”, Chem. Eng. Sci., 53(11), 2089—
2097(1998).
11 Taitel, Y., Barnea, D., “Slug tracking in hilly terrain
NOMENCLATURE pipelines”, In: SPE Ann. Tech. Conf. Exhib. Prod. Oper.
A area, m2 and Eng., Houston, USA, 321—330(1999).
C coefficient 12 Al-safran, E.M., Taitel, Y., Brill, J.P., “Prediction of slug
D internal pipe diameter, m length distribution along a hilly terrain pipeline using
slug tracking model”, In: Proceedings of ETCE/OMAE,
f friction factor
Petroleum Production Technology Symposium, Houston,
Fr Froude number
USA (2002).
g gravitational constant, kg⋅m⋅s－2
13 Wu, Q., Ishii, M., “Sensitivity study on double-sensor
L length; longitudinal spacing, m conductivity probe for the measurement of interfacial
P pressure, Pa area concentration in bubbly flow”, Int. J. Multiphase
R holdup Flow, 25(1), 155—173(1999).
S wet perimeter, m 14 Fagundes, Netto, J.R., Fabre, J., Peresson, L.M., “Shape
t time, s of long bubbles in horizontal slug flow”, Int. J. Multi-
U velocity, m⋅s－1 phase Flow, 25(6), 1129—1160(1999).
VGS gas superficial velocity, m⋅s－1 15 Pinto, A.M.F.R., Campos, B.J.L.M. M., “Coalescence of
VGSO gas superficial velocity at atmospheric condition, m⋅s－1 two gas slugs rising in a vertical column of liquid”,
VLS liquid superficial velocity, m⋅s－1 Chem. Eng. Sci., 51(1), 45—54(1996).
Vm mixture velocity, m⋅s－1 16 He, L.M., “An investigation of the characteristics of
Vo drift velocity of elongated bubble, m⋅s－1 oil-gas two-phase slug flow in horizontal pipes”, Ph. D.
Vt translational velocity of elongated bubble, m⋅s－1 Thesis, Xi’an Jiaotong Univ. (2002).
Vtb translational velocity of elongated bubble behind a 17 Nicklin, D.J., Wilkes, M.A., Davison, J.F., “Two-phase
liquid slug, m⋅s 1
－
flow in vertical tubes”, Trans. Instn. Chem. Engrs., 40,
x axial distance from the mixer, m 61—68(1962).
Z axial coordinate, m 18 Bendiksen, K.H., “An experimental investigation of the
φ inclination angle, deg motion of the long bubbles in inclined tubes”, Int. J.
τ shear stress, Pa Multiphase Flow, 10(4), 467—483(1984).
ρG gas density, kg⋅m－3 19 Andreussi, P., Bendiksen, K., “An investigation of void
ρL liquid density, kg⋅m－3 fraction in liquid slugs for horizontal and inclined
σ surface tension, N⋅m－1 gas-liquid pipe flow”, Int. J. Multiphase Flow, 15(6),
Subscripts 937—946(1989).
acc acceleration 20 Nicholson, M.K., Aziz, K., Gregory, G.A., “Intermittent
two-phase flow in horizontal pipes: predictive models”,
f liquid film
Can. J. Chem. Eng., 56, 653—663(1978).
fr front
21 Bendiksen, K.H., Malnes, D., Nydal, O.J., “On the mod-
i slug unit number eling of slug flow”, Chem. Eng. Comm., 141/142, 71—
S liquid slug 102(1996).
ta tail 22 Ferre, D., “Ecoulements diphasiques a poches en conduite
horizontale”, Rev. Inst. Fr. Pet., 34, 113—142(1979).
REFERENCES 23 Woods, B.D., “Slug formation and frequency of slugging
1 Dukler, A.E., Hubbard, M.G., “A model for gas-liquid in gas-liquid flows”, Ph. D. Thesis, University of Illinois,
slug flow in horizontal and near horizontal tubes”, Ind. USA (1998).
Eng. Chem. Fundam., 14, 337—345(1975). 24 Grotjahn, K., Mewes, D., “Measurement and calculations
2 Moissis, R., Griffith, P., “Entrance effects in a two-phase on the transient behavior of two-phase slug flow in a
slug flow”, J. Heat Transfer, 84, 366—370(1962). horizontal pipe”, In: the 4th International Conference on
3 Cook, M., Behnia, M., “Slug length prediction in near Multiphase Flow, New Orleans, Louisiana, USA (2001).

Chinese J. Ch. E. 14(5) 626 (2006)