Experimental Thermal and Fluid Science 96 (2018) 329–346

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science

journal homepage: www.elsevier.com/locate/etfs

Non-uniformity of gas/liquid flow in a riser and impact of operation and T

pipe configuration on slugging characteristics
⁎
Suifeng Zou, Tian Yao, Liejin Guo , Wensheng Li, Quanhong Wu, Hongliang Zhou, Chen Xie,
Weizhi Liu, Shicai Kuang
State Key Laboratory of Multiphase Flow in Power Engineering, Xi’an Jiaotong University, Xi’an 710049, China

A R T I C LE I N FO A B S T R A C T

Keywords: A pipeline-riser system is a typical complex pipe configuration in which the non-uniformity of gas/liquid flow is
Pipeline-riser system observed to be significant under certain flow rates. This article interprets slugging flow, the most common flow
Temporal and spatial non-uniformity regime in a pipeline-riser system, in the context of non-uniformity, which differs from the approaches used to
Differential pressure characterise slugging cycles in previous studies. By comparing data collected by several transmitters positioned
Pipe configuration
locally on the riser and comparing flow in a simple vertical pipe, both the temporal and spatial non-uniformity of
Flow boundary condition
the gas/liquid distribution in the riser is presented and interpreted; the relationship between them is also illu-
Active flow control
strated. The non-uniformity of the flow provides an effective means to quantitatively evaluate the severity of
slugging in a pipeline-riser system under different operating conditions. In addition to flow uniformity, specific
flow characteristics can be affected by operating conditions as well as pipe configuration. Several factors closely
related to the design of experimental facilities and procedures are investigated by a case study of three ex-
perimental loops with different layouts and dimensions, and several general findings are summarised. The results
may contribute to a better and more specific understanding of gas/liquid flow mechanisms in a pipeline-riser
system in addition to serving as a means to evaluate the features of laboratory pipeline-riser systems and the
applicability of laboratory results in industrial cases.

1. Introduction made possible by pipe design optimization and operation, or by real-

A pipeline-riser system is a common type of offshore petroleum and
gas transportation system that enables flow from a wellhead to a plat-
form or between platforms. Owing to the elbow at the bottom of the
riser, which connects the seabed pipeline and the floating production
storage and offloading (FPSO) unit at sea level, gas/liquid flow in the
riser can experience both temporal and spatial non-uniformity. Altered
flow direction projected into the vertical plane and gas/liquid inter-
facial instabilities result in temporal and spatial non-uniformity.
Slugging is the most common flow regime in a pipeline-riser system. If
the section upstream of the riser is inclined downward due to the
presence of downward-sloping terrain [1], or the diameter of the pipe is
large enough to support the formation of long hydrodynamic slugs [2],
strongly intermittent regimes characterised by long liquid slugs and
sharp fluctuations in pipeline pressure may occur for a certain range of
flow rates. This can have a range of negative impacts on the pipeline
and facilities on an FPSO unit, such as separator overflow, gas com-
pressor overload or gas waste, lower liquid production, pipeline fatigue,
and increased corrosion [3]. The avoidance of undesirable regimes
time monitoring and slug control, requires a better understanding of
flow mechanisms in the pipeline-riser structure. Therefore, gas/liquid
flow in pipes within a pipeline-riser configuration is worthy of in-
vestigation.
Since 1977, it has been observed that the flow regime in a pipeline-
riser system is quite different from that in a simple pipe for a certain
range of flow rates [1]. A simple pipe is one whose direction remains
unchanged from the inlet to the outlet. The phrase ‘global flow regime’
was proposed [4] and can be defined as the phase distribution along a
pipe over time for a certain pipe configuration. The definition has two
aspects: one is the phase distribution along the pipe at a specific mo-
ment in time, and the other is the phase distribution at a local position
over time. In other words, the global regime is temporal and there is a
spatial distribution of two-phase flow, or there is a temporal and spatial
distribution of basic flow patterns such as bubble, slug, churn, and
annular flow.
Many investigations have been conducted for gas/liquid flow in a
pipeline-riser system. Previous studies mainly focused on the categor-
ization and description of global flow regimes [1,5–7]. Regime maps

⁎
Corresponding author.
E-mail addresses: zousuifeng.2011@stu.xjtu.edu.cn (S. Zou), lj-guo@xjtu.edu.cn (L. Guo).

https://doi.org/10.1016/j.expthermflusci.2018.03.021
Received 8 September 2017; Received in revised form 13 March 2018; Accepted 19 March 2018
Available online 20 March 2018
0894-1777/ © 2018 Elsevier Inc. All rights reserved.
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Nomenclature sep separator

w water
Variables or symbols
Abbreviations (also subscripts)
C choke coefficient, Pa·s2·m−2
DP differential pressure transmitter DI downward inclined
Δp pressure drop, kPa DV downward vertical
P pressure transmitter H horizontal
p pressure, kPa, MPa IR Irregular regime, described in Section 2
Q mass flow rate, kg·min−1 OSC Oscillation regime, defined in Section 2
u velocity, m·s−1 R Riser
SS Severe slugging regime
Subscripts SS1 Type 1 of Severe slugging regime, defined in Section 2
SS2 Type 2 of Severe slugging regime, defined in Section 2
a air SS3 Type 3 of Severe slugging regime, defined in Section 2
in inlet ST Stable regime, defined in Section 2
s superficial (at 0 °C, 101.325 kPa) UST Unstable regime, defined in Section 2

have been obtained for different experimental pipe systems, and several
transition criteria have been presented and modified. Periodic para-
meters of intermittent flow were analysed, such as flow period or slug
frequency, slug velocity, slug length, and the liquid holdup of a slug. In
addition, amplitude-domain characteristics of pressure, liquid fraction,
and outflow [1,8–12] were analysed.
Although the non-uniformity of two-phase flow was involved in
several published studies on pipeline-riser systems, it was not in-
dependently treated. In reality, discussions on flow non-uniformity had
long been limited to a temporal aspect, or more precisely, the de-
scription of flow stages. This was the case until we recently proposed a
quick approach to recognise the global flow regime in a riser [4] and
optimised the measurement positions [13]. The basis of the method is
the temporal and spatial correlation of local flow and global flow re-
gime, with spatial non-uniformity playing a more significant role. A
greater focus on the relation between local flow characteristics and
global flow regime contributes to not only the fast recognition of global
flow regimes but also a better understanding of different flow me-
chanisms. The extent of spatial non-uniformity can be easily evaluated
by an array of transmitters along the pipe [14,15]. In addition, the
effect of slug control can be assessed from the perspective of non-uni-
formity, separate from the fluctuation amplitude of pressure. Therefore,
non-uniformity alone is worthy of investigation.
As Schmidt [1] observed that the SS1 regime would not occur
without a downward-inclined section upstream of the riser, most ex-
perimental studies were conducted on a downward-inclined pipe-riser
system or a downward-inclined pipe-riser system that combined with a
horizontal pipe. Severe slugging has been observed in both long and
short and both large and small pipe systems. Published studies provide
results from facilities with an inner diameter of 25.4–254 mm and a
total length of approximately 20–1700 m (see Table 1). There are some
general findings, such as the observation of similar regimes in both
large- and small-scale pipe configurations, the detection of large am-
plitude pressure oscillations under severe slugging, and similar changes
in flow behaviour relative to certain flow rates. However, the flow re-
gime map and several global flow characteristics such as cycle period
and equivalent slug volume have been found to be dependent on pipe
structure parameters. These parameters include diameter, riser type,
riser height, and inclination angle of the pipeline. Nevertheless, the
discussion on pipe layout and geometry is still inadequate, especially
for local and transient flow behaviours as well as for regimes other than
typical severe slugging.
In addition, operating conditions (not limited to elimination
methods for slugging) are observed to affect flow characteristics such as

Table 1
Layout and dimension of experimental/virtual facilities reported in literature.

Researchers Length (m) or height (m) of horizontal section/downward inclined Inclination angle to horizontal Inner Diameter (mm) Riser type
section/riser/downward vertical sectiona (°)b

Schmidt [1] None/30.48/15.24/Long −5 to 5 50.8 Vertical

Taitel et al. [31] None/9.1/3/None −5 25.4 Vertical
Fabre and Peresson [32] None/25/13.5/None −0.9 to 0.9 53 Vertical
Tin [5] None/60/33/Short −2 50.8 Catenary, S-shape
Montgomery [8] None/57.4/9.98/None −2 50 S-shape
Tengesdal [33] None/19.81/14.9/Short −5 to 0 76.2 Vertical
Wang [9] 114/20.4/16.3/None −5 to 0 50 Vertical
Mokhatab [34] None/53.84/10.5/None −2 101.6 Catenary
Ali [19] 36/None/12.2/None 0 254 Vertical
Gao [35] 16.7/0.61/4/None −2 26 Catenary
Malekzadeh et al. [6,36] 65/35/15.5/None [6] −2.54 ([6]) 50.8 (pipeline) Vertical
66/34/15/None [36] 0 ([36]) 45 (riser/[6])
50.8 (riser/[36])
Li et al. [7] 114/16/15.3/None −5 to −1 50 Flexible S-shape
Pedersen [37] 30/12/6/None (3rd generation) −20 to 0 (1st generation) 63 Vertical
Zou et al. [4] 380/25/21.5/Long −5 75 Vertical
Li et al. [38] 1657/30/11.2/Short −7 46 S-shape

Notes:
a
Length of downward-inclined section and riser height are for a certain inclination angle, and they are slightly different for other angles. For the last value, ‘Long’, ‘Short’, and ‘None’
are determined by observation of the figures provided.
b
Non-continuously adjustable within the range for most of the facilities, i.e. only discrete values are available.

330
S. Zou et al.
the effect of outlet pressure on the flow regime boundary [8]. With
regard to the response behaviour of flow to transient operational con-
ditions, the influence of inlet conditions on pipe flow was not empha-
sised although published studies have included the transient process of
outlet pressure [16] and inlet flow rates [17].
Therefore, a comparative study focusing on the impact of these
factors would help to distinguish which findings are general and which
are specific; then, the results from previous and current studies can be
evaluated. Moreover, the designing of new experimental facilities and
the updating of existing facilities will benefit from these results.
The pipe layout showcases the joint of sections with different or-
ientations by bends. The influence of bends on local flow behaviour can
be illustrated by the distribution of a local flow pattern along the entire
pipe system. The distribution is the outcome of both pipe configuration
and flow rates. Occasionally, the transition of a local flow pattern will
occur almost immediately after the bend, while sometimes the transi-
tion will cover some distance downstream of the bend [18]. For a pi-
peline-riser system in particular, configuration-induced slugging and
hydrodynamic slugging correspond to short and long flow pattern
transition distances, respectively. The former case is well known, while
the latter case manifests as the movement of long liquid aerated slug
from the upstream pipeline to the riser, which was observed in a large-
diameter pipe [19]. For a shorter and smaller horizontal pipe-vertical
riser [20], the persistence of slug frequency upstream and downstream
was observed under high liquid velocity.
Flow pattern development or evolution in undulated piping has also
been studied. The dissipation of slugs in a downward section and the
generation of slugs in an upward section were investigated. Al-Safran
et al. [21] classified the flow regimes into five categories while con-
sidering the contribution of the hilly-terrain effect. The effects of both
configuration and velocity on the transition distance of a local flow
pattern were observed. Huang et al. [22] also noticed that both struc-
ture and flow rates are predominant parameters for three-phase flow
regimes. Slug dissipation would be enhanced by bigger inclinations and
longer pipes [21,23–25]. For viscous oil/water two-phase flow, the
impact of undulation would be insignificant [26]. Flow development
from a horizontal pipe to consecutive wavy sections was also numeri-
cally and experimentally investigated, and a mitigation of hydro-
dynamic slug flow was discovered [27]. These studies are helpful for a
discussion on flow evolution from a horizontal section to downward-
inclined section in the present study.
In addition, more precise two-fluid and multi-field models [28–30]
can benefit from an investigation of local flow characteristics and the
distribution of flow patterns in both short and long pipes. Of course,
numerical models must be verified through experimental results.
The above review of published studies indicates that flow char-
acteristics can be affected by many aspects in such large-scale
Fig. 1. Sketch of experimental systems.
331
Experimental Thermal and Fluid Science 96 (2018) 329–346
experimental apparatuses. The influences involve steady and transient
boundary conditions, pipe layout, and pipe geometry. Some influences
and their mechanisms are apparent, while the others have not received
much discussion.
As previously mentioned, the first aim of the present study is to
explain flow phenomena in a pipeline-riser from the perspective of non-
uniformity. Thus, this aspect will be discussed first (in Section 3). Then,
examples are given as an application of non-uniformity analysis in as-
sessing the severity of slugging under several operations (Section 4.1).
Detailed flow behaviour is also observed to be affected by operation as
well as pipe configuration, and related discussion is present in Section
4.2. This section follows the second aim of this study, which is to dis-
tinguish the similarities and differences in gas/liquid flow in different
pipes and find the causes. Mechanisms are also involved in the dis-
cussion. The results may contribute to developing a more in-depth and
specific understanding of gas/liquid flow in pipeline-riser systems.
2. Experiment
Experiments were conducted on three pipeline-riser systems under
different flow rates, separator pressures, and choke coefficients. All
systems are gas/liquid two-phase loops, as sketched in Fig. 1, despite
the fact that their facilities and pipe configurations (see the table in
Fig. 1) are different. Loop B is the oldest among the three loops; later
on, Loop A and Loop C were constructed for study of flow in larger pipe
and longer pipe, respectively. All three loops are made of stainless steel,
except for a transparent section on the riser of Loop B. The horizontal
section (dashed line in Fig. 1) for each loop is sketched in Fig. 2. Table 1
shows that the most common value for pipe I.D. is approximately
50 mm (2 in.), which includes Loop B and Loop C; however, Loop A has
larger inner diameter. The lengths of the three loops are among the top
of all the facilities in Table 1, concerning the evolution of flow in longer
pipe closer to field condition. The total length of the S-shaped riser of
Loop C is 16.7 m, and the ‘S’ section was drawn to scale in [38]. For
detailed descriptions of Loops A, B, and C and their data acquisition
systems, see [4,39], and [38], respectively. The installation of several
pressure transmitters and differential pressure transmitters is presented
in Fig. 3 and Table 2. Both ‘B1’ and ‘B2’ in Fig. 3 are positioned a few
centimetres upstream of the lowest position of the loop; for loop B, the
distance is 7 cm. There are two positions on Loop B for gas injection,
namely ‘Pos-R’ and ‘Pos-DI’, and the two positions are also marked in
Fig. 3.
Generally, the purpose of the choke valve is to develop techniques
to eliminate slugging; however, it could also be used to increase pres-
sure in the loop along with the separator. Loop A and Loop C were
originally designed for high-pressure flow experiments. Because high-
pressure experiments on a large closed loop would be much more
S. Zou et al.
Fig. 2. Configuration of Section H ((a) Loop A, (b) Loop B, (c) Loop C).
Fig. 3. Installation of several pressure/differential pressure transmitters and positions for
gas injection.
dangerous, both loops are open. This means that a large pressure drop
will be incurred when high-pressure fluids are discharged into an at-
mospheric environment. Therefore, two-stage depressurization was
originally designed. This process is characterised by the use of valves at
both inlet and outlet of the separator. Several relevant impacts of this
process will be discussed in Section 4.2. After the last experiment at
5 MPa on Loop A [12], the valve at the separator inlet was removed for
experiments on Loop C under an atmospheric outlet condition [38], as
illustrated in Fig. 1.
Air and water were used as working fluids. The operating conditions
during the experiments are listed in Table 3. Data were acquired at a
frequency of 100 Hz, which was sufficient for the investigation of the
global flow regime in this work. The acquisition time for each test de-
pended on the period of the flow cycles, and varied from 12 min to 6 h.
The accuracy of the pressure transmitters and differential pressure
transmitters was either 0.075% or 0.06% of full scale, which was sa-
tisfactory for the analyses presented in Section 3.
During the experiments, pressure and differential pressure were
332
Experimental Thermal and Fluid Science 96 (2018) 329–346
measured at different pipe positions. Different types of flow regimes
occur under different gas and liquid velocities. Those regimes can be
roughly divided into two types: unstable regimes (USTs) and stable
regimes (STs). The UST regime can be further categorised into severe
slugging 1 (SS1), severe slugging 2 (SS2), severe slugging 3 (SS3), and
oscillation (OSC). The definitions of the UST (SS1, SS2, SS3, and OSC)
and ST regimes are provided as follows:
UST – Cyclic flow whose fluctuation amplitude of differential pres-
sure along the riser is above 20% of the hydrostatic pressure dif-
ference when the riser is filled with stagnant water;
SS1 – a UST regime with periodic pure liquid slug blockage at the
bottom of the riser and with a maximum liquid slug length greater
than the riser height;
SS2 – a UST regime with periodic pure liquid slug blockage at the
bottom of the riser but with a maximum liquid slug length no
greater than the riser height;
SS3 – a UST regime with a maximum liquid slug length greater than
riser height, but the liquid slug is aerated with gas slugs or large
bubbles;
OSC – a UST regime with aerated liquid slug that is shorter than the
riser height and with continuous gas entry into the riser.
ST – Cyclic or stochastic flow whose fluctuation amplitude of dif-
ferential pressure along the riser is no more than 20% of the hy-
drostatic pressure difference when the riser is filled with stagnant
water. Bubble, slug, churn, annular pattern, and non-severe slugging
are representative of a stable flow regime.
The definitions of SS1 and SS2 are consistent with the long-accepted
definitions, and the threshold of fluctuation amplitude (20%) for the
determination of UST and ST regimes is in accordance with the de-
scription made by Schmidt [1]. Among all the UST regimes, SS1 is the
most typical severe slugging case that contains clear stages of slug
growth, liquid outflow, blowout, and liquid fallback (Fig. 4). Both SS1
and SS2 are universally considered to be typical severe slugging, while
OSC is considered to be a transitional regime. SS3 is controversial; the
conditions under which it occurs are limited, and it is occasionally re-
garded as SS1 or OSC.
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Table 2 regime, which was conducted on Loop A, the results of this loop will be
Installation position of several pressure/differential pressure transmitters. presented first.
Pressure Position of taps (DP5 Length (DP1 to DP4) Full scaleb
transmitters or and DP11) or centre or altitude difference (abs, kPa) 3.1. Non-uniformity of flow in Loop A
differential position of (DP5 to DP11) ofc
pressure measurement section measurement section Fig. 5 presents the mean (time-averaged value) and range (ampli-
transmitters (the rest) (m) tude) of pressure drop signals collected on Loop A at psep = 1 MPa (g).
Loop A Both the temporal and spatial non-uniformity of flow in the riser can be
DP1 68 m from ‘A’ 113 200 obtained from Fig. 5. The coloured region of all contour maps, which
DP3 8.1 m from ‘A’ 0.36 2 was not a rectangle, refers to experimental range.
DP4a 15.2 m from ‘A’ 1.50 10 Large intermittency, or temporal non-uniformity, is significant
DP5 Lower tap: ‘A’ 0.5 10
under the SS, OSC, and IR regimes, as illustrated by the large amplitude
Upper tap on riser
DP11a ‘B1’ and ‘C’ 21.5 240 (range) of the riser pressure drop presented in Fig. 5(b) and (d). The
DP12 (not ‘C’ and bottom of 190 200 difference between the contours of Δp11 and Δp5 can help to explain the
presented in section DV dominant mechanism of each regime. The SS1 and SS2 regimes show a
Fig. 3)
larger fluctuation of Δp11, but smaller fluctuations of Δp5 compared to
PDV (not Bottom of Section DV N/A 5000 (8000
presented in for the OSC regime. A larger fluctuation of Δp11 indicates longer liquid
Fig. 3) Fig. 17(b)) slugs, which could only be the result of liquid blockage at the riser
PR ‘B1’ N/A 5000 (8000 bottom for all three of the loops. Because liquid blockage is induced by
for the elbow, it is certain that the pipe configuration is the primary cause
Fig. 17(b))
of the SS1 and SS2 regimes. A larger fluctuation of Δp5 indicates that
Pin (not presented Inlet of pipeline N/A 5000 (8000
in Fig. 3) for there is a longer transition distance for slugs traveling from upstream.
Fig. 17(b)) The transient slugs in section H can be induced by pressure fluctuation
Psep (not Inlet of separator N/A 5000 (8000 at the riser bottom or by Kelvin–Helmholtz instability [40]. Because the
presented in (downstream of for
pressure fluctuation at the riser bottom still suggests that the pipe
Fig. 3) choke valve) Fig. 17(b))
configuration has an effect on it, the mechanism of the OSC regime is
Loop B more complex, and the dominated mechanism of the regular OSC re-
DP1a 51 m from ‘A’ 90 110
DP2a 10.4 m from ‘A’ 19.9 90
gime cannot be distinguished. The irregular OSC regime will be dis-
DP3 4.06 m from ‘A’ 2 9 cussed in Section 4.2. It is widely accepted that the dominant me-
DP5 Lower tap: ‘B1’ 0.8 20 chanism of the ST regime is Kelvin–Helmholtz instability. Hence, the
Upper tap on riser mechanism of the SS3 regime can also be regarded as mixture of con-
DP6 11 m from ‘C’ 0.75 20
figuration and hydrodynamics, because it is a transitional regime from
DP7 8.6 m from ‘C’ 0.75 20
DP8 7.4 m from ‘C’ 0.75 20 SS1 to ST. As ‘hydrodynamic slug/slugging’ is a common phrase in this
DP9 3.3 m from ‘C’ 0.75 20 field, the mechanism of the Kelvin–Helmholtz instability is written as
DP10a 1.6 m from ‘C’ 0.75 30 ‘hydrodynamics-dominated mechanism’ in this article.
DP11a ‘B1’ and ‘C’ 16.3 220 Temporal non-uniformity can be visually interpreted by the inter-
PR ‘B1’ N/A 500
Pin (not presented Inlet of pipeline N/A 500
mittency of liquid outflow (Fig. 4), i.e. the non-conservative transient
in Fig. 3) liquid flow rate at the inlet and outlet of the pipeline. However, even
the time-averaged mass flow rates at the inlet and at outlet in a cycle
Loop C
DP1 945 m from ‘A’ 1299 2000 might not be equal. The latter form of non-conservative flow has not
DP11a ‘B2’ and ‘C’ 11.2 200 been emphasised until very recently [39]; now, the irregular SS regime
PR ‘B2’ N/A 2000 and some cases of irregular OSC regimes can be explained. Different
Pin (not presented Inlet of pipeline N/A 2000 mechanisms for irregular regimes will be discussed in Section 4.2.
in Fig. 3)
Transient spatial non-uniformity, or instant distribution of basic
Notes: flow patterns, can be easily confirmed visually (Fig. 4) through mea-
a
At least one pair of flanges between two taps. surement of liquid holdup (or differential pressure), or even through
b
For differential pressure transmitters, this value refers to max. set value minus min. numerical work with a two-fluid model [28]. However, in the present
set value; min. set value can be negative. work, it is discovered that spatial non-uniformity still exists after the
c
Positions ‘A’, ‘B1’, ‘B2’, and ‘C’ are illustrated in Fig. 3.
fluctuations are wiped out by time average, as demonstrated by the
different trends in Fig. 5(a) and (c). Qualitatively, the time-averaged
The irregularity of the flow regime was also observed in the ex- pressure drop (liquid holdup) of the riser (Δp11) and that at the riser
periment. The irregularity manifests as an alternate occurrence of an ST bottom (Δp5) show different trends with superficial water velocity
regime and an SS or OSC regime. An irregular regime (IR) is char- under the SS regime. The pressure drop, Δp11, is in consistent with li-
acterised by either non-conservative liquid flow at the inlet and outlet quid holdup at inlet, while Δp5 first shows a contrary tendency with it
of the pipeline during the blowout stage, or as occasional hydro- but then becomes consistent with it as liquid velocity increases. It is also
dynamic slugs that form in the pipeline and move to the riser. Both noticed that liquid velocity at the inversion point of the trend is ap-
cases will be discussed later in this article. proximately at the boundary of severe slugging and stable regime.
The trend presented in Fig. 5(c) indicates that the effect of gravity is
3. Spatial and temporal non-uniformity of flow in the riser still dominant under low flow rates. Configuration- or gravity-domi-
nated mechanisms under low and medium gas and liquid flow rates
The pressure drop of a vertical section can indicate liquid holdup were interpreted as a criterion for the transition to severe slugging [41].
when the flow rate is not too high (friction can be neglected). Thus, From the view of monotony, the difference between Fig. 5(a) and 5(c)
spatial and temporal non-uniformity of flow in the riser can be illu- suggests that effect of pipe configuration (gravity) would become
strated by the difference in pressure drop at different positions on the weaker than that of hydrodynamics beyond the downstream neigh-
riser (Δp5 to Δp11). Because the investigation into non-uniformity was bourhood of the elbow. Because Loop A suffered from a limited ex-
motivated by our previous research on fast recognition of a global flow perimental velocity range, a limited number of transmitters, and

333
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Table 3
Operating conditions during experiments.

Item Symbol Unit Value Note

Superficial air velocity usa m·s−1 0.056–22 (Loop A) at 0 °C, 101.325 kPa; dissolution neglected
0.045–8.5 (Loop B)
0.055–10 (Loop C)
Superficial water velocity usw m·s−1 0.014–0.86 (Loop A)
0.018–2.0 (Loop B)
0.020–2.1 (Loop C)
Separator pressure psep MPa 0, 0.5, 1 (Loop A) Gauge
0 (Loops B, C)
Choke coefficient C Pa·s2·m−2 7.0 × 105(*), 2.5 × 106 (Loop A) Tested by single-phase water;
(or Local resistance factor) 4.0 × 102(*)–9.3 × 105 (Loop B) (*)
valve open = 100%
N/A (Loop C)
Max pressure in the pipeline pmax MPa 3.0 (Loop A) Gauge
0.49 (Loop B)
1.8 (Loop C)
Temperature in the pipeline T °C 10–40 (Loops A, B)
20–40 (Loop C)
Pressure at pump outlet ppump MPa 2–6(*) (Loop A) (*)
2–6 MPa for psep = 0 MPa; 6 MPa for psep = 0.5 and 1 MPa
1 (Loop B)
1–2 (Loop C)
Avg. pressure of gas buffer tank pv MPa 14 MPa (Loop A) Gauge
0.8 MPa (Loop B)
9 MPa (Loop C)

choking even at the fully open valve condition, successive investigation
on non-uniformity were conducted on Loop B.
3.2. Non-uniformity of flow in Loop B
The non-uniformity of the flow in Loop B will be discussed from the
viewpoint of parameter distribution along the riser and non-mono-
tonous behaviour of parameter with liquid flow rate, as mentioned in

Fig. 4. A typical case of severe slugging ((a) sketch of a cycle; (b) pictures captured at transparent section on Loop B).

334
S. Zou et al.
Fig. 5. Time-averaged pressure drop vs. superficial air and water velocities (at 0 °C and 101.325 kPa) for riser and riser bottom (Loop A, psep = 1 MPa (g), C = 7.0 × 105 Pa·s2·m−2).
Section 3.1. Thanks to the array of recently installed differential pres-
sure transmitters, it is possible to quantitatively describe spatial non-
uniformity. In addition, the relation between spatial and temporal non-
uniformity can be illustrated qualitatively. To avoid inaccuracies in the
estimation of liquid holdup through measured pressure drop under the
homogeneous hypothesis, a control group of experiments involving
two-phase flow in a simple vertical pipe were conducted, i.e. cases
where gas and liquid were mixed at the riser bottom (‘Pos-R’ in Fig. 3).
It was assumed that time-averaged frictional loss was equal for the two
mixing conditions under the same gas and liquid flow rates. By com-
paring the pressure drops under the two mixing conditions, the mag-
nitude relation between the pressure drops (or liquid fractions) in the
two cases was qualitatively (but reliably) derived, as presented in
Fig. 6.
Fig. 6(a) shows that at the riser bottom, the liquid holdup is larger
than that in a simple vertical pipe under lower liquid velocity, corre-
sponding to long-time liquid blockage; and it is smaller under higher
liquid velocity, corresponding to movement of liquid slug. Similar to
the results presented in Fig. 5(c), a left ripple pattern (‘ < < < ’) is
also observed at the upper portion of this subplot. For Loop B, the
vertex of each ripple (‘ < ’) is approximately located on the boundary of
the SS3 and ST regimes. This indicates a transition of the dominant flow
mechanism. As the gas velocity increases, the liquid holdup becomes
smaller than that in a simple vertical pipe. However, the major cause
for this cannot be determined because pure liquid slug remained be-
tween the lower tap of the DP5 transmitter and the injection point.
Fig. 6(b–f) reveals that the area indicating less than unity (1.000)
increases as the altitude increases, and the lower boundary of unity on
the contours also increases with altitude. This tendency can be ex-
plained by a longer duration of vacant volume at a higher riser position
under low and medium liquid flow rates. Therefore, spatial non-uni-
formity shows the difference in temporal non-uniformity among dif-
ferent positions. From another perspective, because the time-averaged
335
Experimental Thermal and Fluid Science 96 (2018) 329–346
liquid holdup of the entire riser is still different from that in a simple
vertical pipe (Fig. 6(g)), the transient outflow rate of each phase defi-
nitely changes sharply so as to meet the requirement of mass con-
servation, which manifests as a large amplitude of pressure fluctuation
and severe intermittency. Therefore, the temporal non-uniformity of
gas/liquid flow can be seen from spatial non-uniformity. In addition,
Fig. 6(b–f) suggests that the liquid velocity corresponding to the highest
ratio increases with altitude position of the riser.
Five local differential pressure transmitters (DP6 to DP10) are suf-
ficient to quantitatively evaluate time-averaged spatial non-uniformity
in the riser. Because of the compressibility of gas, liquid holdup de-
creases with altitude, which means that phase distribution is still non-
uniform under the same flow rates even if gas and liquid are mixed at
the riser bottom. Therefore, when quantifying spatial non-uniformity in
a riser, local pressure drops are replaced by the ratios of their time-
averaged magnitudes (those in Fig. 6) at these measurement positions.
Spatial non-uniformity can be illustrated by max./min. and standard
deviation of the five ratios presented in Fig. 7. The two similar contours
indicate that the SS1, SS2, and OSC regimes are the most spatially non-
uniform, and the highest level of non-uniformity is found at the oc-
currence of the SS2 regime, which is in consistent with the fluctuation
amplitude (Fig. 5(b) and [42]). As for the SS3 regime, it seems that even
though the fluctuation amplitude of the riser pressure drop (i.e. tem-
poral non-uniformity) is still moderate, the spatial distribution of gas
and liquid is more uniform at each stage of an SS3 cycle; consequently,
time-averaged spatial uniformity arises.
Fig. 5(c) presents a pressure drop trend at the riser bottom that is
opposite to that of the liquid flow rate at the inlet under a low gas flow
rate for Loop A. In the most recent experiment on Loop B, the sign of the
pressure drop differential with respect to liquid superficial velocity
(∂Δp/∂usw) at each measurement position has been obtained. The re-
sults are presented in Fig. 8. At the riser bottom (Fig. 8(a)), the region of
‘−1’ (decrease in pressure drop with increase in liquid flow rate) covers
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Fig. 6. Difference in pressure drop between riser and simple vertical pipe, characterised by magnitude ratio (Loop B).

those of the SS1, SS2, and SS3 regimes. Fig. 8(a–f) shows that the region negative trend also arises at the upper part of the riser under an irre-
contracts as the altitude increases, and non-monotony in the region of gular regime. Therefore, it can be concluded that a hypothetical area
SS3 regime disappears last. It is also observed in Fig. 8(c–f) that a that envelops all the areas related to ∂Δp/∂usw < 0 in Fig. 8

336
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Fig. 7. Level of spatial non-uniformity (Loop B; (a) max./min. of ratios of Δp6 to Δp10, (b) standard deviation of ratios of Δp6 to Δp10).

approximately coincides with the velocity range of an unstable regime.
In addition, ∂Δp/∂usw < 0 is experienced at all the measurement po-
sitions on the riser, even at the position of the DP10 transmitter where
the local resistance of the flange accounts for a significant portion of the
pressure drop. This indicates that the impact of the pipe configuration
can spread to as far as 0.9 times the riser height for Loop B.
The above results show similar transition behaviour from SS1 to
SS3, and to the ST regime in different pipeline-riser systems with a
vertical riser. Moreover, the liquid flow rate for the inversion of the sign
of ∂Δp/∂usw can be regarded as the approximate transition boundary

Fig. 8. Sign function of partial derivative of pressure drop with respect to superficial water velocity (Loop B).

337
S. Zou et al.
from the SS3 to ST regime.
Even when hydrodynamics are the dominant mechanism under high
gas and/or liquid velocity, the impact of pipe configuration on flow
characteristics is still visible. A comparison was made based on the
observation of the transparent section where Δp7 was measured. The
difference between local flow patterns for a stable regime and that for a
simple vertical pipe under the same inlet flow rates is illustrated in
Fig. 9(a). The difference manifests as different velocity ranges for
bubble and slug patterns. Fig. 9(b) and (c) presents different distribu-
tions of amplitude and frequency parameters between the two condi-
tions, indicating some inapplicability of a recognition model based on
data from a simple vertical pipe to pipeline-riser [43]. Related results
can be found in [14,18,20]. The reason why the impact of pipe con-
figuration is not likely to completely vanish lies in the inhomogeneous
gas/liquid mixture at riser bottom determined by upstream flow rates.
4. Impact of operation and pipe configuration on slugging
characteristics
Although previous studies have confirmed that severe slugging can
occur in both large- and small-scale pipes, gas/liquid flow character-
istics vary among different laboratory pipe systems and operating
conditions, such as steady and transient boundary conditions, and pipe
geometry (layout and dimension). Active control for elimination of
severe slugging is also among the boundary conditions. The extent of
flow uniformity and detailed flow behaviours can both be affected.
4.1. Effect of operation on extent of flow uniformity
Operation can influence global flow regime, thus affecting the
Fig. 9. Different characteristics of local flow patterns in a riser and in a simple vertical pipe [43] (Loop B; (a) local flow pattern map at position of DP7 transmitter; (b) and (c) distribution
of amplitude and frequency parameters for riser and simple vertical pipe; parameters are calculated from samples plotted in subplot (a)).
338
Experimental Thermal and Fluid Science 96 (2018) 329–346
extent of flow uniformity. Severe slugging and oscillation regimes are
undesirable, and some active control methods are implemented in field
circumstances. Operating pressure is also an important parameter for
pipe transportation. The extent of flow uniformity is also different in
pipes with different configurations, but the effect of pipe configuration
will not be discussed because there are too many variables.
Nevertheless, the analysis method presented in Section 3.2 is still ap-
plicable.
4.1.1. Effect of active control
The effect of choking is presented in Fig. 10, whose data were col-
lected from Loop B. The control group consisted of experiments con-
ducted using a simple vertical pipe with a valve open condition equal to
the values in Fig. 10, rather than those with valve open = 100. The
numerical values marked on Fig. 10 refer to the valve open condition
for reducing the fluctuation amplitude of Δp11 to 20% of the hydrostatic
pressure difference when the riser is filled with stagnant water (i.e.
transition to stable regime); the colours refer to quantitative non-uni-
formity parameters that have been introduced in Fig. 7. Obviously, the
flow in a riser is more like that in a simple vertical pipe when choking is
applied. Fig. 10 also shows that the highest level of spatial non-uni-
formity occurs at lower usa compared with a condition without choking.
However, the level of spatial non-uniformity is smaller at higher usw,
even though the fluctuation amplitude of Δp11 is roughly the same for
all nine cases.
The effect of gas-lifting is presented in Fig. 11, whose data were also
collected from Loop B. The 4-D plots were generated from 16 tests. The
surface refers to a total superficial gas velocity of 2 m/s (sum of su-
perficial velocities at inlet and at injection position), and the colours
refer to the non-uniformity parameter introduced in Fig. 7(a). The plots
S. Zou et al.
Fig. 10. Level of non-uniformity under permanent choking (Loop B, fluctuation of Δp11 = approx. 34 kPa).
clearly show a smaller fluctuation amplitude and a higher level of
spatial uniformity when gas is injected at Pos-R than at Pos-DI. Both
plots exhibit no significant relation between fluctuation amplitude and
spatial non-uniformity. To say a few more words, although severe
slugging and oscillation regimes (whether regular or irregular) can be
eliminated by gas injection at Pos-R for Loop B, different results were
reported for different pipe systems, as summarised by Ogazi [44]. The
advantage of gas-lifting over choking is the fast response of flow regime
in the riser.
From the two cases above, the two proposed parameters describing
spatial non-uniformity prove useful for assessing the severity of un-
desirable flow regime before and after the implementation of control
methods. Not only permanent control, but also dynamic control can be
evaluated from the view of non-uniformity of gas/liquid flow, though
this article does not involve response to transient operation process.
4.1.2. Effect of operating pressure and validation of existing criteria of
regime transition
The results were obtained from Loop A where experiments under
different operating pressures were conducted. Owing to the absence of
an injection device at the riser bottom, the controlled experiment was
not applicable. Instead, a drift-flux model was adopted to calculate the
gas fraction in the hypothetical simple vertical pipe. Drift-flux models
are widely used in the prediction of multiphase pipe flow (classic
models are presented in [45–47]), and more precise formulae are still
being proposed [48,49]. A drift-flux model was also considered to be
suitable for predicting the void fraction in a serpentine pipe in a vertical
plane [50]. A model proposed by Shi et al. [51], which was also
adopted by Malekzadeh et al. [52], was selected for the calculation of
Fig. 11. Level of non-uniformity under permanent gas-lifting (Loop B, (a) injection at Pos-R, (b) injection at Pos-DI).
339
Experimental Thermal and Fluid Science 96 (2018) 329–346
liquid holdup of a simple vertical pipe.
A comparison between the experimental results of the riser and the
calculated results for the corresponding simple vertical pipe is pre-
sented in Fig. 12. The ratio above unity is observed around the transi-
tion boundary from SS1 to SS2 at an atmospheric outlet condition in
both Figs. 6(g) and 12(a); the left bottom sections of the two plots are
similar. Thus, this approach is qualitatively reliable for most of the area
in Fig. 12, except for the region at the bottom right of Fig. 12(a), where
friction cannot be neglected. Fig. 12 indicates that the difference be-
tween the time-averaged liquid holdup in a riser and that in a simple
vertical pipe would be more significant with higher operating pressure,
although the slugging actually becomes less severe. Fig. 12(a) exhibits
an unusual result above the upper usw bound of SS1 regime, which
could be ascribed to an atypical regime that will be discussed later.
Although measurement devices are limited on Loop A, a validation
of existing criteria for regime transition is still applicable. The para-
meters of the criteria proposed by Taitel [53], Pots et al. [41], Jansen
et al. [54], Montgomery [8] and Malekzadeh et al. [6] can be calculated
with the measured pressure, differential pressure, and flow rate of each
phase. A criterion based on the linear stability analysis of the governing
equations has been recently presented by Azevedo et al. [55], though,
in this work the conventional criteria are still analysed by experimental
results.
As the pressure at the riser top has been measured, it is directly used
for validation of criteria, and the pressure at the separator is aban-
doned. Therefore, the criteria of Taitel [53] and Jansen et al. [54] have
the same expression:
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Fig. 12. Difference in Δp11 between riser and simple vertical pipe, characterised by magnitude ratio (Loop A, C = 7.0 × 105 Pa·s2·m−2; (a) psep = 0 MPa (g), (b) psep = 0.5 MPa (g), (c)
psep = 1 MPa (g)).

2 The predictions of the velocity range of severe slugging through the

αL 2Cusw
ptop > ρw g ⎡ (ϕ− )−ϕH⎤ four available criteria are plotted in Fig. 13. The properties of gas and
⎢ α′ ρw gH ⎥ (1)
⎣ ⎦
liquid were determined by measured pressure and room temperature,
Criterion of Malekzadeh et al. [6] is given by and it was assumed that the solution of gas in the water is saturated
2 after mixing. The gas fraction in pipeline α was calculated by Taitel’s
αL 2C′u mix
ptop > ρw g ⎡ (ϕ− )−ϕH⎤ formulae [40], and for all the criteria involved, the gas fraction in the
⎢ α′ ρ gH ⎥ (2)
⎣ w ⎦ pipeline was weighted by the lengths of horizontal section and down-
where ward-inclined section (Eq. (5)); L is the total length of section H and
section DI. The average liquid fraction in the riser, φ, and the local gas
ρmix ρ usw + ρa usa
C′ = C =C w fraction at the riser bottom, α′, were calculated using the drift-flux
ρw ρw (usw + usa ) (3) model [51] based on the time-averaged measured pressures under the
Criterion of Montgomery [8] is given by assumption of a simple vertical pipe. In Eq. (4), the velocity of gas
entering the riser, usa,riser_bottom, was calculated according to Mon-
gαLsinθ (ρw −ρa ) ⎤ tgomery’s procedures [8]. The inclination angle at the nose of gas slug
usa,inlet ⩽ ⎡1− u
⎢ pR ⎥ sa,riser bottom (4) entering the riser, θ, will be discussed in the next paragraph.
⎣ ⎦

Fig. 13. Comparison between experimental results and predictions of criteria of global flow regime (Loop A; (a) and (d) psep = 0 MPa (g), (b) psep = 0.5 MPa (g), (c) psep = 1 MPa (g); (a),
(b) and (c) C = 7.0 × 105 Pa·s2·m−2, (d) C = 2.5 × 106 Pa·s2·m−2).

340
S. Zou et al.
αH LH + αDI LDI
α=
LH + LDI (5)
The results show that the prediction by Malekzadeh’s criterion [6]
agrees well with the boundary of severe slugging for the cases of
psep = 0.5 MPa (g) and 1 MPa (g), while the velocity range of severe
slugging at atmospheric operating pressure of separator is under-
estimated. On the other hand, Jansen’s criterion [54] overestimates the
velocity range of severe slugging for all four operating conditions. The
difference between the two criteria lies in the velocity adopted for the
choking calculation. Superficial water velocity is adopted by Jansen
et al. [54], while the mixture velocity (the sum of superficial water and
air velocities) is taken by Malekzadeh et al. [6]. The two-phase mixture
at the bottom of downward vertical section is between pure water and
homogeneous air/water mixture owing to the gravity effect; thus, the
transition boundary obtained from experiments is expected to lie be-
tween the predicted boundaries of the two criteria. The performance of
Pots’s criterion [41] is poor for this loop; however, the criterion can be
combined with Jansen’s criterion [54] to achieve better prediction
under atmospheric separator pressure. As for severe slugging enhanced
by the downward connection between the riser top and separator, this
particular regime will be discussed in Section 4.2. It is interesting to
discover that Montgomery’s criterion [8] can predict the boundary
between regular unstable regime and irregular regime. However,
Fig. 13 demonstrates that the inclination angle θ differs much among
operating pressure of the separator. This result suggests that the posi-
tion of gas slug nose at the elbow at the riser bottom might be influ-
enced by pressure, which requires validation by visualization or im-
pedance tomography in the future. The predicted boundary varies with
θ, but a value for prediction of the boundary between irregular regime
and stable regime cannot be found for Loop A.
4.2. Effect of operation and pipe configuration on detailed flow behaviours
Because detailed flow behaviours can be characterised by pressure
and pressure drop signals, a signal difference would potentially lead to
different performance for flow regime recognition and dynamic anti-
slugging control models. The flow regime depends on the inlet and
outlet conditions originally determined by operation. The pressure
fluctuation in the pipe can also be influenced by inlet and outlet con-
ditions, but the reaction would be insignificant under proper operation.
The relation involved will be discussed later. Flow cycles are not exactly
the same in pipes with different configurations, although some beha-
viours are general. For each influencing factor, effects on behaviours
will be discussed first by case study; then, general findings are sum-
marised or inferred by detailed behaviours in each loop.
Fig. 14. An SS1 case without and with choking (Loop B, usw = 0.21 m/s, usa = 0.19 m/s).
341
Experimental Thermal and Fluid Science 96 (2018) 329–346
4.2.1. Non-synchronization of pressure and pressure drop signals induced
by choking and section DV
Fig. 14 shows non-synchronization of pressure and pressure drop
signals induced by choking. The inner diameter of the valve is equal to
that of the pipe, indicating no choking at the full open position. Under
this circumstance (Fig. 14(a)), the fluctuation of pressure at the riser
bottom is almost the same and is synchronised with that of riser pres-
sure drop; the blowout stage of severe slugging is fast and clear. When
choking is applied (Fig. 14(b)), the pressure signal lags behind the
pressure drop signal. This difference is that the outcome of different
pressure drop across the valve when pure liquid, pure gas, or gas/liquid
mixture flows across it. Therefore, choking-induced non-synchroniza-
tion is a common behaviour for different pipeline-riser systems. For
dynamic anti-slugging control, the outlet flow rate has been found to be
effective for stabilizing the control system [56], and a system based on
topside pressure may also work within a limited range of valve open
scenarios [37,44,57]. Both flow rate and topside pressure are associated
with the pressure drop across the valve; this suggests that possible so-
lutions may be realised based on the pressure drop.
Fig. 15 shows non-synchronization induced by a downward vertical
section (Loop C). The pressure signal also lags behind the pressure drop
signal; however, the non-synchronization under this circumstance is not
identical to that under a choking condition. Fig. 15(b) shows that the
pressure drop decline starts ahead of the pressure decrease, but the
growth of both pressure drop and pressure may occasionally begin at
roughly the same moment. The delay of the pressure decrease can be
tied to the lag time between the blowout and the moment when the
long liquid slug flows out of section DV. Therefore, the occurrence of
non-synchronization is not restricted to the downward ‘vertical’ section,
but should be a more general behaviour when there is a downward
connection at any angle between riser top and separator, or even more
complex connection with several bends in vertical plane. The horizontal
connection between the bottom of section DV and the separator is also
likely to enhance non-synchronization because liquid blockage can
more easily form at the bottom of section DV before it flows out.
4.2.2. Prolonged gas exhaust by choking and section DV
Both causes of non-synchronization are related to the impediment of
gas exhaust. It is also noticed in Fig. 14(b) that blowout stage of severe
slugging will last longer under the occurrence of shorter slugs, which
manifests as high-frequency oscillation of pressure and pressure drop in
the middle and later period of the blowout stage. Choking can hinder
flow velocity from sharp increase during blowout stage, so the gas ac-
cumulated in the pipeline cannot exhaust quickly after the liquid slug is
pushed into the separator, and a short period of slug flow appears.
Under a choking condition, the gas in the riser is still exhausting when
liquid in the riser starts to fall back, so the pressure drop rises ahead of
S. Zou et al.
Fig. 15. An SS2 case without and with downward vertical section that connects riser and separator (usw = 0.15 m/s, usa = 1.51 m/s).
pressure. Meanwhile, gas is still accumulating when it starts to push
liquid out of the riser, which creates a time lag between the decrease in
pressure drop and pressure. With slug flow passing through the valve,
the fluctuation of its pressure drop arises and results in the fluctuation
of pressure at the riser top and pressure drop along the riser. In sum-
mary, choking will induce a non-synchronization of pressure with
pressure drop, and prolong the blowout stage by bringing about oscil-
lation.
Non-synchronization induced by the volume of connection between
riser top and separator has been illustrated in Section 4.2.1. When the
downward connection is combined with choking at its bottom, the
pressure signal will significantly differ from pressure drop signal.
Fig. 16 displays more than one blowout in a cycle, with pressure
stepping down correspondingly but with smaller amplitude. The oc-
currence and number of steps depends mainly on the volume of
downward vertical section, but also on choke coefficient. For Loop A,
multiple blowouts arise because the volume of the downward vertical
section is less than that of the riser, thus the section cannot hold the
whole slug; then, the long slug will be truncated by the air that enters
the riser and rises to the top. When the liquid column in the downward
vertical section is drained, the next blowout comes about, and so on
until the remaining air in the pipe is unable to create another blowout.
The pressure amplitude is significantly smaller than that of the pressure
drop as a result of compressible air pocket at the reversed U-bend at the
riser top.
Figs. 14 and 16 clearly show that the liquid production stage that
characterises the SS1 regime can still be determined by the pressure
drop signal, but not the pressure signal. Hence, pressure drop has to be
adopted for the determination of flow regime when the pipe is non-
Fig. 16. Examples of SS1 regime with downward vertical section and choking (Loop A).
342
Experimental Thermal and Fluid Science 96 (2018) 329–346
transparent.
4.2.3. Siphoning and enhanced fluctuation of pressure drop
Siphoning is an issue that cannot be avoided in a discussion about
the downward connection between a riser top and a separator. Negative
pressure can be found at the riser top for single-phase flow and two-
phase flow under low velocity in both Loop A and Loop C. Siphoning
becomes more significant for longer downward pipes, as evidenced by
the pressure decline witnessed at the liquid production stage in Fig. 16.
However, when the mixture velocity is very low, pressure will not de-
cline because the bend at the riser top is not blocked by liquid slug, and
there is still a proportion of gas occupation through the downward
vertical section from riser top to the separator (stage II of Fig. 4(a)). The
above characteristics can help researchers to qualitatively estimate
whether expected results or conclusions are independent from existence
or absence of siphon in future experimental work.
Notice that small pulses appear on pressure drop signal when
pressure starts descending. This phenomenon can be interpreted as
another mutual effect of choking, and the volume of the downward
connection between the riser top and separator. The cause and effect
can be illustrated by Fig. 17(a). The red dashed lines correspond to the
complete exhaust of the liquid column in section DV (see Δp12). The
blue dashed lines correspond to the moment when gas phase starts to
flow across the horizontally-installed valve. Apparently, a sudden de-
crease in pressure drop across the valve leads to a decrease in pDV. Then,
the liquid level grows quickly like a pulse owing to the transient
pressure gradient.
Similar cases are marked as ‘severe slugging enhanced by section
DV’ in the flow regime maps for Loop A (Fig. 13). ‘Enhanced’ refers to
S. Zou et al. Experimental Thermal and Fluid Science 96 (2018) 329–346

Fig. 17. Effect of outlet condition on global flow regime (Loop A, (a) mutual effect of section DV and choking, (b) effect of manual control of compressor pressure).

an enhancement of pressure drop fluctuation. Fig. 13 also shows that
this phenomenon is more significant with a higher choke coefficient
where the draining of downward vertical section is slower. Although
the temporal non-uniformity is enhanced, the pressure fluctuation is
mitigated; thus, it is an atypical case of severe slugging. Its occurrence
is beyond the prediction made by those criteria (Fig. 13). However, for
flow regime recognition, it was not regarded as an individual regime
but as either severe slugging or oscillation regime [4].
Similar to the non-synchronization behaviour induced by the pipe
configuration discussed in Section 4.2.1, ‘enhanced severe slugging’ is
not restricted to the condition of downward ‘vertical’ section either. For
a field case, this phenomenon may also be encountered because the

Fig. 18. Propagation of pressure wave and fluctuation of water flow rate.

343
S. Zou et al.
valve may not be installed at the highest position of the pipe config-
uration, although it still installed on the floating platform.
4.2.4. Interaction between transient inlet/outlet conditions and global flow
regime
The gas/liquid separator at the end of the pipe system will be jeo-
pardized under severe slugging; however, flow instability in the pipe
can be induced by the operation of the separator. Fig. 17(b) shows that
ineffective control of separator pressure will give rise to additional
transient pressure-gradient driving force and result in larger fluctua-
tions of pressure drop or phase distribution.
Inlet flow rates are among the decisive parameters of flow regime in
both simple and complex pipe configurations; however, not much at-
tention has been placed on the reaction of flow regime on inlet condi-
tions even though fluctuations in inlet flow rates were observed by
several researchers. Fig. 18(a–d) presents an inverse fluctuation of inlet
water flow rate against that of inlet pressure. The pressure signal at the
inlet lags behind the signal at the riser bottom, indicating a propagation
of the pressure wave from downstream to upstream. For Loop A and
Loop B, the waveform of the pressure signal is roughly the same from
the riser bottom to the inlet (Fig. 18(a), (d) and (e)). Meanwhile,
Fig. 19. Irregular regimes ((a) IRSS2 [39], (b) and (c) IROSC, (d) IRSS1).
344
Experimental Thermal and Fluid Science 96 (2018) 329–346
significant attenuation can be observed in Loop C (Fig. 18(b)), and even
the elimination of cyclic fluctuation can be found at higher flow rates
(Fig. 18(c)). For shorter pipes, if the fluctuation of inlet flow rates is
desired to be mitigated, increasing the reservoir pressure would be an
effective measure. For the case presented in Fig. 18(e), the fluctuation
amplitude of inlet pressure (90 kPa) is small enough compared to the
outlet pressure of the pump (5 MPa). In our experiments, increasing the
pressure of the water reservoir was achieved by turning down the inlet
valve and bypass valve upstream of the flowmeter; while the pressure of
air reservoir was maintained by a large buffer tank at the outlet of the
compressor. The methods adopted in this study would be better than
inserting a buffer tank downstream of air/water mixer [6,31,33,36].
The above results indicate that sometimes inlet and outlet condi-
tions cannot be regarded as steady; this suggests that more precise inlet
and outlet conditions are needed for integrated numerical simulations.
For a field case, reservoir pressure and operating pressure decrease with
time, and the impact of downstream flow on upstream condition would
possibly become more significant. Thus, pressure control at the inlet
should be more like field condition for a closer approximation.
S. Zou et al.
4.2.5. Flow irregularity caused by long pipeline
An irregular flow regime was found in all three loops, and this type
of regime is difficult to monitor in real-time. IRSS1 clearly appeared in
Loop C and faintly in Loop A, IRSS2 occurred in all the three loops, and
IROSC appeared clearly in Loop A and faintly in Loop B. IRSS1, IRSS2,
and IROSC are presented in Fig. 19.
A measurement of liquid volume at the separator outlet found non-
conservative liquid volume flow within an SS2 cycle when the IRSS2
regime occurred [39]. The outflow volume was much bigger than the
inflow volume within an SS2 cycle (Fig. 19(a)); thus, little water re-
mained in the pipeline after several cycles, and a period of stable re-
gime arose.
The above mechanism is believed to be the cause of IRSS1, IRSS2,
and some cases of the IROSC regime. The length ratio of the section (H
plus DI) over the riser is 18.8, 8.16, and 101 for loops A, B, and C,
respectively, indicating more irregularity in a longer pipe according to
the description in the first paragraph of this section. A large volume of
gas compared with liquid volume in the long pipeline enhances the
uncertainty of severe slugging cycles. It can be inferred that this type of
irregular flow will occur in a pipe longer than Loop B and with the
existence of Section DI.
4.2.6. Long hydrodynamic slugs caused by large pipeline
For the rest of IROSC cases, the mechanism is different. Fig. 19(b)
presents the generation of isolated liquid slugs in the horizontal section
and its movement from section H to the riser via section DI. This case
clearly shows the influence of upstream flow on downstream flow,
which is opposite to the cases presented in Figs. 14–16, where pulses of
DP1 are induced by blowouts in the riser.
As no evidence is found to prove that downstream flow conditions
influence the generation of slugs, it is certain that these slugs are hy-
drodynamic slugs. However, this phenomenon is not so clear in loop B,
whose diameter is smaller (Fig. 19(c)). The phenomenon eventually
vanishes in Loop C, which has an even smaller diameter, where long
liquid slugs are absent in the horizontal section. The tendency of long
liquid slugs agrees with Brill et al. [2], who reported that slug length
was positively correlated to pipe diameter. Ali and Yeung [58] reported
a case in a large-diameter pipe (254 mm) where the flow pattern in the
horizontal section was slug flow, while a large fluctuation of pressure
was encountered at the riser bottom.
‘Hydrodynamic slugging’ was also observed in a small diameter pipe
by Gao [35], but the riser height was only 4 m. According to the defi-
nition in this article, a liquid slug longer than 0.8 m could be called
‘unstable regime’ for Gao’s case, and his case also matches the de-
scription of ‘irregular regime’ in this article. Therefore, the more precise
summary would be that a hydrodynamics-dominated unstable or irre-
gular flow regime is more likely to occur in a larger pipe when the riser
height is the same.
It can be derived from criteria of flow pattern transition [40] that a
slight downward inclination to the horizontal can significantly extend
the velocity range of stratified pattern, and thus extend velocity range
of (typical) severe slugging. However, if there is a horizontal section
upstream of the downward-inclined section, the flow would be some-
what different because the transition of flow patterns will cover some
distance from the intersection of the two sections (point ‘A’ in Fig. 3).
Therefore, if slug flow arrives at the riser bottom before the transition
to stratified flow is completed, then the flow regime in the riser would
possibly be different from that under the condition without the hor-
izontal section. The cited publications in the Introduction [21–26] may
help to improve the understanding flow development in a pipeline-riser
system.
5. Summary and conclusions
Gas/liquid flow in a pipeline-riser system was studied from the
viewpoint of spatial and temporal non-uniformity. Then, the impact of
345
Experimental Thermal and Fluid Science 96 (2018) 329–346
operation on uniformity was investigated, and the impacts of both
operation and pipe configuration on detailed flow behaviours were
discussed. The results provide a novel understanding of flow mechan-
isms.
(1) The most intense non-uniformity occurs under the SS2 regime but
very close to the transition boundary between the SS1 and SS2 re-
gimes. It is validated for a vertical riser through this study in co-
ordination with published studies.
(2) The difference between temporal non-uniformity for the whole riser
and that for the riser bottom indicates a different flow mechanism,
i.e. configuration-dominated and hydrodynamics-dominated me-
chanisms. The trend of the pressure drop at the riser bottom against
superficial liquid velocity is related to the transition from the SS3 to
the ST regime.
(3) The spatial non-uniformity and temporal non-uniformity are asso-
ciated by mass conservation. A quantitative description of spatial
non-uniformity achieved by the max./min. and standard deviation
of pressure drop ratio at five local positions was demonstrated to be
effective for evaluating the severity of slugging in a pipeline riser
with or without flow control. This approach is inapplicable to the ‘S’
section of an S-shaped riser, but it is possible to apply it to the
vertical section of an S-shaped riser.
(4) The non-synchronization of pressure and pressure drop signals can
be induced by choking and downward connection between the riser
top and the separator. An alternate outflow of pure liquid, pure gas,
and a gas/liquid mixture leads to the oscillation of pressure drop
across the choke valve, resulting in the further oscillation of pres-
sure and riser pressure drop. When choking and the downward
connection are combined, the fluctuation of riser pressure drop will
be enhanced.
(5) An irregular regime can manifest as non-conservative liquid flow at
the inlet and outlet within a cycle. This type of irregularity occurs in
all the three loops whose length ratio of section (H plus DI) over the
riser is above 8 and whose absolute length of section (H plus DI) is
above 134 m. The critical length ratio and absolute length for the
occurrence might be found in the future.
(6) Another irregular regime manifests as occasional slugs moving from
the horizontal section to the riser via downward-inclined section.
The occurrence of this type of irregularity depends on the pipe
diameter. It can be speculated that for a riser higher than 10 m, the
inner diameter of pipeline must be no smaller than 50 mm. The
length and inclination angle of section DI also affects the occur-
rence.
Through this study, we wish to provide better and more specific
understanding of gas/liquid flow in pipeline-riser systems. As there are
many variables in pipeline-riser experiments, this research offers a
qualitative reference for the design of experimental apparatus and
procedures as well as a rough approach to evaluate the features of la-
boratory pipeline-riser systems. Some flow characteristics in industrial
cases can also be qualitatively predicted. The prediction of flow char-
acteristics, flow regime recognition, and the elimination of undesirable
regimes would benefit from this study because flow behaviours are
reflected in signals.
Acknowledgement
Support of National Natural Science Foundation of China (No.
51527808) and long-term support of Chinese National Science and
Technology Major Project (No. 2011ZX05026-004-002 and No.
2016ZX05028-004-002) are gratefully acknowledged. The authors also
wish to express gratitude to Mr. Wang, Yubo, Mr. Guo, Zhengbin, Mr.
Xie, Xiangdong, Mr. Liu, Zhigang, Mr. Qi, Chao, Mr. Zhou, Wenbo, Mr.
Guo, Longjin, and Mr. Zhao, Xiangyuan for their participation in the
experimental work.
S. Zou et al.
References
[1] Z. Schmidt, Experimental Study of Two-Phase Slug Flow in a Pipeline-Riser Pipe
System, Ph.D. Dissertation University of Tulsa, U.S.A, 1977.
[2] J.P. Brill, Z. Schmidt, W.A. Coberly, J.D. Herring, D.W. Moore, Analysis of two-
phase tests in large-diameter flow lines in Prudhoe Bay field, SPE J. 271 (1981)
363–378.
[3] S. Pedersen, P. Durdevic, Z. Yang, Challenges in slug modeling and control for
offshore oil and gas productions: A review study, Int. J. Multiph. Flow 88 (2017)
270–284.
[4] S. Zou, L. Guo, C. Xie, Fast recognition of global flow regime in pipeline-riser system
by spatial correlation of differential pressures, Int. J. Multiph. Flow 88 (2017)
222–237.
[5] V. Tin, Severe slugging in flexible riser, in: Proceedings of 5th International
Conference on Multiphase Technology, Cannes, France, 1991, pp. 507–525.
[6] R. Malekzadeh, R.A.W.M. Henkes, R.F. Mudde, Severe slugging in along pipeli-
ne–riser system: experiments and predictions, Int. J. Multiph. Flow 46 (2012) 9–21.
[7] N. Li, L. Guo, W. Li, Gas-liquid two-phase flow patterns in a pipeline-riser system
with an S-shaped riser, Int. J. Multiph. Flow 55 (2013) 1–10.
[8] J.A. Montgomery, Severe Slugging and Unstable Flows in an S-Shaped Riser, Ph.D.
Thesis Cranfield University, U.K., 2002.
[9] X. Wang, Study on the Characteristics of Oil-Gas-Water Slug Flow in Horizontal
Pipes and Severe Slugging in Pipeline-Riser System, Ph. D. Dissertation, Xi’an
Jiaotong University, China., 2006 (in Chinese).
[10] S. Blaney, H. Yeung, Investigation of the exploitation of a fast-sampling single
gamma densitometer and pattern recognition to resolve the superficial phase ve-
locities and liquid phase water cut of vertically upward multiphase flows, Flow
Meas. Instrum. 19 (2008) 57–66.
[11] J. Ye, L. Guo, Multiphase flow pattern recognition in pipeline–riser system by sta-
tistical feature clustering of pressure fluctuations, Chem. Eng. Sci. 102 (2013)
486–501.
[12] C. Xie, L. Guo, W. Li, H. Zhou, S. Zou, The influence of backpressure on severe
slugging in multiphase flow pipeline-riser systems, Chem. Eng. Sci. 163 (2017)
68–82.
[13] S. Zou, T. Yao, L. Guo, Q. Wu, Z. Liu, Effect of measurement position on signal
length for recognition of global flow regime in pipeline-riser, in: Proceedings of
10th International Symposium on Measurement Techniques for Multiphase Flow,
Hong Kong SAR, China, 2017, No. ISMTMF-R001-116.
[14] A.M. Aliyu, A.A. Almabrok, Y.D. Baba, L. Lao, H. Yeung, K.C. Kim, Upward ga-
s–liquid two-phase flow after a U-bend in a large-diameter serpentine pipe, Int. J.
Heat Mass Transf. 108 (2017) 784–800.
[15] A.A. Almabrok, A.M. Aliyu, L. Lao, H. Yeung, Gas/liquid flow behaviours in a
downward section of large diameter vertical serpentine pipes, Int. J. Multiph. Flow
78 (2016) 25–43.
[16] E. Al-Safran, L. Kappos, C. Sarica, Experimental and numerical investigation of
separator pressure fluctuation effect on terrain slugging in a hilly terrain two-phase
flow pipeline, ASME J. Energy Resour. Technol 130 (2008) 033001.
[17] S. Zou, L. Guo, W. Liu, T. Yao, S. Kuang, Experimental study on transient process of
gas and liquid flow rates in pipeline-riser system, J. Eng. Thermophysics 38 (2017)
587–594 (in Chinese).
[18] D. Zhao, R. Omar, M. Abdulkadir, L.A. Abdulkareem, A. Azzi, F. Saidj, V.H. Perez,
B.N. Hewakandamby, B.J. Azzopardi, The control and maintenance of desired flow
patterns in bends of different orientations, Flow Meas. Instrum. 53 (2017) 230–242.
[19] S.F. Ali, Two Phase Flow in Large Diameter Vertical Riser, Ph. D. Thesis Cranfield
University, U.K., 2009.
[20] F. Saidj, R. Kibboua, A. Azzi, N. Ababou, B.J. Azzopardi, Experimental investigation
of air–water two-phase flow through vertical 90° bend, Exp. Therm. Fluid Sci. 57
(2014) 226–234.
[21] E. Al-Safran, C. Sarica, H.-Q. Zhang, J. Brill, Investigation of slug flow character-
istics in the valley of a hilly-terrain pipeline, Int. J. Multiph. Flow 31 (2005)
337–357.
[22] S. Huang, B. Zhang, J. Lu, D. Wang, Study on flow pattern maps in hilly-terrain
air–water–oil three-phase flows, Exp. Therm. Fluid Sci. 47 (2013) 158–171.
[23] Y. Taitel, C. Sarica, J.P. Brill, Slug flow modeling for downward inclined pipe flow:
theoretical considerations, Int. J. Multiph. Flow 26 (2000) 833–844.
[24] E. Al-Safran, Investigation and prediction of slug frequency in gas/liquid horizontal
pipe flow, J. Petrol. Sci. Eng. 69 (2009) 143–155.
[25] G. Ersoy, C. Sarica, E. Al-Safran, H.-Q. Zhang, Three-phase gas-oil-water flow in
undulating pipeline, J. Petrol. Sci. Eng. 156 (2017) 468–483.
[26] A.B. Desamala, A.K. Dasmahapatra, T.K. Mandal, An appraisal of viscous oil–water
two-phase flow through an undulated pipeline in peak configuration, Exp. Therm.
Fluid Sci. 68 (2015) 177–186.
[27] L. Xing, H. Yeung, Y. Geng, Y. Cao, J. Shen, Study on hydrodynamic slug flow
mitigation with wavy pipe using a 3D–1D coupling approach, Comput. Fluids 99
(2014) 104–115.
[28] P. Han, L. Guo, Numerical simulation of terrain-induced severe slugging coupled by
hydrodynamic slugs in a pipeline–riser system, Int. J. Heat Fluid Flow 56 (2015)
346
Experimental Thermal and Fluid Science 96 (2018) 329–346
355–366.
[29] A.H. Akselsen, O.J. Nydal, Applying multiple grids to a multi-field model – the
resolution requirements of individual fields in the two-fluid model for 1D pipe flow,
J. Disper. Sci. Technol. 36 (2015) 1378–1387.
[30] V.R. Almeida, M.G. Conte, F.A.A. Barbuto, C. Cozin, R.E.M. Morales, Numerical
simulation of two-phase slug flow from horizontal to downward inclined pipe using
a hybrid code based on slug tracking and two-fluid methodologies, in: Proceedings
of ASME 2017 Fluids Engineering Division Summer Meeting, 2017, No.
FEDSM2017-69335.
[31] Y. Taitel, S. Vierkandt, O. Shoham, J.P. Brill, Severe slugging in a riser system:
experiments and modeling, Int. J. Multiph. Flow 16 (1990) 57–68.
[32] J. Fabre, L.L. Peresson, J. Corteville, T. Bourgeois, R. Odello, Severe slugging in
pipeline/riser systems, SPE Production Eng. 5 (1990) 113–120.
[33] J.Ø. Tengesdal, Investigation of Self-Lifting Concept for Severe Slugging
Elimination in Deep-Water Pipeline/Riser Systems, Ph. D. Thesis University of
Tulsa, U.S.A., 2002.
[34] S. Mokhatab, B.F. Towler, Severe slugging in flexible risers: review of experimental
investigations and OLGA predictions, Petrol. Sci. Technol. 25 (2007) 867–880.
[35] S. Gao, Research on Gas-liquid Severe Slugging Characteristic in Pipeline-Riser
Systems, Ph. D. Dissertation, Shanghai Jiaotong University, China, 2012 (in
Chinese).
[36] R. Malekzadeh, R.F. Mudde, R.A.W.M. Henkes, Dual-frequency severe slugging in
horizontal pipeline-riser systems, ASME J. Fluids Eng 134 (2012) 121301.
[37] S. Pedersen, Plant-Wide Anti-Slug Control for Offshore Oil and Gas Processes, Ph. D.
Dissertation, Aalborg University, Denmark, 2016.
[38] W. Li, L. Guo, X. Xie, Effects of a long pipeline on severe slugging in an S-shaped
riser, Chem. Eng. Sci. 171 (2017) 379–390.
[39] T. Yao, L. Guo, Z. Liu, Q. Wu, Experimental study on differential pressure fluc-
tuations characteristics of unsteady flow pattern in pipeline-riser system, J. Eng.
Thermophysics 38 (2017) 1250–1256 (in Chinese).
[40] Y. Taitel, A.E. Dukler, Transitions in horizontal and near horizontal gas-liquid flow,
AIChE J. 22 (1976) 47–55.
[41] B.F.M. Pots, I.G. Bromilow, M.J.W.F. Konijn, Severe slug flow in offshore flowline/
riser systems, SPE Production Eng. 2 (1987) 319–324.
[42] X. Luo, L. He, H. Ma, Flow pattern and pressure fluctuation of severe slugging in
pipeline-riser system, Chin. J. Chem. Eng. 19 (2011) 26–32.
[43] Q. Wu, L. Guo, S. Zou, T. Yao, Different characteristics of pressure drop signal of
flow patterns in a simple vertical pipe and in a riser, in: Proceedings of 10th
International Symposium on Measurement Techniques for Multiphase Flow, Hong
Kong SAR, China, 2017, No. ISMTMF-R001-115. (extended abstract).
[44] A.I. Ogazi, Multiphase Severe Slug Flow Control, Ph. D. Thesis Cranfield University,
U.K., 2011.
[45] N. Zuber, J.A. Findlay, Average volumetric concentration in two phase flow sys-
tems, ASME J. Heat Mass Trans. 87 (1965) 453–468.
[46] R.C. Fernandes, R. Semiat, A.E. Dukler, Hydrodynamic model for gas-liquid slug
flow in vertical tubes, AIChE J. 29 (1983) 981–989.
[47] K.H. Bendiksen, An experimental investigation of the motion of long bubbles in
inclined tubes, Int. J. Multiph. Flow 10 (1984) 467–483.
[48] B.J. Azzopardi, H.K. Do, A. Azzi, V.H. Perez, Characteristics of air/water slug flow
in an intermediate diameter pipe, Exp. Therm. Fluid Sci. 60 (2015) 1–8.
[49] J. Escrig, B. Hewakandamby, B. Azzopardi, Influence of diameter and inclination of
the pipes on the velocity of periodic structures in gas-liquid intermittent flows, in:
Proceedings of 9th world conference on experimental heat transfer, fluid mechanics
and thermodynamics, Iguazu Falls, Brazil, 2017, No. OC025.
[50] A.A. Almabrok, A.M. Aliyu, Y.D. Baba, L. Lao, H. Yeung, Void fraction development
in gas-liquid flow after a U-bend in a vertically upwards serpentine-configuration
large-diameter pipe, Heat Mass Transfer (2017), http://dx.doi.org/10.1007/
s00231-017-2118-0 (in press).
[51] H. Shi, J.A. Holmes, L.J. Durlofsky, K. Aziz, L.R. Diaz, B. Alkaya, G. Oddie, Drift-flux
modeling of two-phase flow in wellbores, SPE J. 10 (2005) 24–33.
[52] R. Malekzadeh, S.P.C. Belfroid, F. Mudde, Transient drift flux modelling of severe
slugging in pipeline-riser systems, Int. J. Multiph. Flow 46 (2012) 32–37.
[53] Y. Taitel, Stability of severe slugging, Int. J. Multiph. Flow 12 (1986) 203–217.
[54] F.E. Jansen, O. Shoham, Y. Taitel, The elimination of severe slugging: experiment
and modeling, Int. J. Multiph. Flow 22 (1996) 1055–1072.
[55] G.R. Azevedo, J.L. Baliño, K.P. Burr, Linear stability analysis for severe slugging in
air–water systems considering different mitigation mechanisms, Int. J. Multiph.
Flow 73 (2015) 238–250.
[56] E. Storkaas, S. Skogestad, Controllability analysis of two-phase pipeline-riser sys-
tems at riser slugging conditions, Control Eng. Pract. 15 (2007) 567–581.
[57] E. Jahanshahi, Control Solutions for Multiphase Flow – Linear and nonlinear ap-
proaches to anti-slug control, Ph. D. Thesis Norwegian University of Science and
Technology, Norway, 2013.
[58] S.F. Ali, H. Yeung, Experimental investigation and numerical simulation of two-
phase flow in a large-diameter horizontal flow line vertical riser, Petrol. Sci.
Technol. 28 (2010) 1079–1095.