sensors

Article
Characterization of Gas–Liquid Two-Phase Slug Flow Using
Distributed Acoustic Sensing in Horizontal Pipes
Sharifah Ali 1 , Ge Jin 2 and Yilin Fan 1, *

1 Petroleum Engineering Department, Colorado School of Mines, Golden, CO 80401, USA; [email protected]
2 Geophysics Department, Colorado School of Mines, Golden, CO 80401, USA; [email protected]
* Correspondence: [email protected]

Abstract: This article discusses the use of distributed acoustic sensing (DAS) for monitoring gas–liquid
two-phase slug flow in horizontal pipes, using standard telecommunication fiber optics connected
to a DAS integrator for data acquisition. The experiments were performed in a 14 m long, 5 cm
diameter transparent PVC pipe with a fiber cable helically wrapped around the pipe. Using mineral
oil and compressed air, the system captured various flow rates and gas–oil ratios. New algorithms
were developed to characterize slug flow using DAS data, including slug frequency, translational
velocity, and the lengths of slug body, slug unit, and the liquid film region that had never been
discussed previously. This study employed a high-speed camera next to the fiber cable sensing
section for validation purposes and achieved a good correlation among the measurements under all
conditions tested. Compared to traditional multiphase flow sensors, this technology is non-intrusive
and offers continuous, real-time measurement across long distances and in harsh environments, such
as subsurface or downhole conditions. It is cost-effective, particularly where multiple measurement
points are required. Characterizing slug flow in real time is crucial to many industries that suffer
slug-flow-related issues. This research demonstrated the DAS’s potential to characterize slug flow
quantitively. It will offer the industry a more optimal solution for facility design and operation and
ensure safer operational practices.

Keywords: distributed acoustic sensing; distributed fiber-optic sensing; slug flow characterization;
Citation: Ali, S.; Jin, G.; Fan, Y.
multiphase flow sensors; gas–liquid slug flow; horizontal pipe flow; flow monitoring; multiphase
Characterization of Gas–Liquid
flow meter
Two-Phase Slug Flow Using
Distributed Acoustic Sensing in
Horizontal Pipes. Sensors 2024, 24,
3402. https://doi.org/10.3390/
1. Introduction
s24113402
Multiphase flow in pipes is a common but also complex phenomenon occurring in
Academic Editors: Michael P. Buric various fields, such as petroleum engineering, geothermal applications, nuclear engineering,
and Nageswara R. Lalam
etc. For example, in the petroleum industry, oil and gas flow is inevitable due to the
Received: 24 April 2024 pressure reduction and temperature variations along the wellbore and pipeline systems.
Revised: 19 May 2024 As energy demand increases and the world explores deeper harsh zones to increase oil
Accepted: 22 May 2024 and gas recovery, multiphase-flow-related safety issues are becoming more complex and
Published: 25 May 2024 concerning. Critical parameter monitoring in real time is essential for better production
management and optimization. Detecting and forecasting the circumstances of the well
at earlier periods have a substantial effect on well control strategies; health, safety, and
environment (HSE); and risk management [1]. This will give the operators the ability to
Copyright: © 2024 by the authors. take the appropriate action at the right time [2].
Licensee MDPI, Basel, Switzerland.
There are several types of point sensors that can characterize slug flow, such as γ-ray
This article is an open access article
sensors, wire-mesh sensors, electrical impedance (capacitance and conductance) sensors,
distributed under the terms and
microwave sensors, optical sensors, etc. They are either not safe (γ-ray sensors have a
conditions of the Creative Commons
radioactive source), intrusive (such as the wire-mesh sensors), or have special requirements
Attribution (CC BY) license (https://
on the type of fluids or flow pattern (see details in [3]). One significant drawback is that
creativecommons.org/licenses/by/
they all require installation directly into the piping system, necessitating the cutting of
4.0/).

Sensors 2024, 24, 3402. https://doi.org/10.3390/s24113402 https://www.mdpi.com/journal/sensors

Sensors 2024, 24, 3402 2 of 17

pipes and their subsequent connections which adds another failure factor. Moreover, these
technologies can only provide point measurements at a specific location where the sensor
is installed. Considering the cost of each sensor, it is not economically feasible to install one
for each individual pipeline/well that requires detailed monitoring.
The fiber-optic sensing technology is developing rapidly and signifies the future of
multiphase flow supervising and monitoring. Fiber optics can prevail over the limitations
of conventional electrical sensor arrays due to their smaller size, non-intrusive features,
corrosion resistance, isolation, and ability to function effectively in intense environmental
situations [4]. The small size of these sensors enables them to be securely utilized over
extended distances with minimized future maintenance operations [5]. One of the crucial
benefits of fiber optics is that the same sensor can carry dual purposes. It can act in the
same way as the sensing element for measuring the critical physical parameters and as a
transmission medium for the detected signal [6].
Generally, distributed fiber-optic sensing (DFOS) can be classified into three types:
(a) distributed temperature sensing (DTS), (b) distributed acoustic sensing (DAS), and
(c) distributed strain sensing (DSS) [7]. DAS employs an optical time-domain reflectometer
(OTDR) system to record high-frequency strain rates over long distances. The sensing
range along the fiber length of a DAS system can be more than 10 km, with a spatial
resolution less than 1 m and a temporal sampling frequency up to 10 kHz [8]. Since a
well-protected sensing fiber can endure harsh environments, DAS is desirable in wellbore
diagnostics to examine several aspects. It can serve as a noise log to detect leakage or
estimate injection rate—e.g., [9,10], supporting the differentiation between the channeling,
borehole flow, matrix flow, and fracture flow [11]. It can also be used for hydraulic fracture
monitoring [12] and seismic acquisition [13]. DFOS is commonly used and has many
applications in engineering and geophysics disciplines.
van der Horst et al. (2013) [14] reported using DAS for flow monitoring for tight
gas well producers. The acoustic noise was monitored, and the production profiles were
generated by the gas phase flowing from the perforations, which is then converted to
flowrates using the DAS signal in a specified frequency band. This method has provided
some quantitative production measurement but does not apply to oil well producers. A
study shows the flow velocity estimation using low-frequency DAS signals for tracking the
wellbore thermal slugging. Jin et al. (2019) [15,16] demonstrate the use of a combination of
DTS and DAS for production monitoring. The method is used for oil producers with low
producing rates where the DTS was utilized to measure steady-state borehole temperatures,
and the DAS is used to measure transient borehole flow velocities by tracking the signals
generated by thermal slugging. This research was further developed for liquid and gas
two-phase flow, specifically for gas–oil two-phase slug flow [17–19]. The experiments’
findings using a vertical flow loop indicated that the acoustic and thermal signals measured
by DAS are sensitive to low air injection. Also, the thermal signatures obtained from the
DAS data mainly respond to the water phase in the flow, while the air phase provides
unique characteristics in the acoustic domain.
Another study shows that DAS measurements for two-phase gas–liquid slug flow
are used to estimate the velocity and extract velocity patterns. The research includes two
methods, k-f transform, and distributed cross-correlation. The results are validated by a
commercial conductance-based phasemeter [20]. Many DFOS data processing are based
on qualitative approaches only. In this article, experiments are conducted to develop
a quantitative analysis for multiphase flow in pipes using DAS, which can expand the
applications of DFOS data and contribute to better multiphase flow characterization.
Flow patterns in horizontal two-phase flows can range from dispersed, stratified, in-
termittent, and annular flow patterns depending on the rates of the phases, fluid properties,
and pipe diameter [21]. Slug flow is a type of intermittent flow pattern, in which the gas
and liquid phases flow alternatively. It is also one of the most common flow patterns
encountered in the oil and gas transportation system [22]. This type of flow pattern have
been extensively studied in the laboratory [23], such as [24–27], to name a few. The authors
 Sensors 2024, 24, x FOR PEER REVIEW 3 of 18
Sensors 2024, 24, 3402 3 of 17

been extensively studied in the laboratory [23], such as [24–27], to name a few. The authors
ofof[21,28]
[21,28]provide
provideaadetailed
detailedreview
reviewofofthis
thisflow
flowpattern
patternandandthe
therelevant
relevantmodeling
modelingstudies.
studies.
AAtypical
typical slug unit consists of a slug body region and a film region (also calledgas
slug unit consists of a slug body region and a film region (also called gaspocket
pocket
region),
region),asasdepicted
depictedininFigure
Figure1.1.Some
Someprevious
previousstudies
studiesclaimed
claimedthat
thatthe
theslug
slugregion
regioncould
could
bebeconsidered
consideredas as aa dispersed
dispersed bubble
bubbleflow,
flow,andandthe
thefilm
film region
region as as segregated
segregated flowflow
[29].[29].
The
The entire
entire slugslug
unitunit is moving
is moving forward
forward at theattranslational
the translational velocity
velocity (vT ), which
(vT), which is also
is also the slug
the slug
front front velocity.
velocity. Figure 1 illustrates the typical characteristic parameters for slug
Figure 1 illustrates the typical characteristic parameters for slug flow, in-
flow, including the translation velocity. From Figure 1, LU is the slug unit length, LS is the
cluding the translation velocity. From Figure 1, LU is the slug unit length, LS is the liquid
liquid slug body length, Lf is the length of the liquid film region, vs. is the average mixture
slug body length, Lf is the length of the liquid film region, vs. is the average mixture ve-
velocity inside the slug body, vF is the average liquid velocity of the liquid film region, and
locity inside the slug body, vF is the average liquid velocity of the liquid film region, and
vC is the velocity of the gas pocket in the liquid film region.
vC is the velocity of the gas pocket in the liquid film region.

Figure1.1.Physical
Figure Physicalstructure
structureofofslug
slugflow
flowwith
withitsitscharacteristic
characteristicparameters.
parameters.

ItItisisessential
essentialtotocharacterize
characterizethe
theslug
slugflow,
flow,notnotonly
onlybecause
becauseofofits
itscommon
commonoccurrence
occurrence
ininthe
the oil and gas transportation system, but also due to its alternatingbehavior
oil and gas transportation system, but also due to its alternating behaviorofofliquid
liquid
slugs
slugsand andgas gaspockets
pockets which could
which couldpose a risk
pose to the
a risk production
to the productionsystem and compromise
system and compromisethe
safety of the pipelines, especially to the joints and elbows [30–32]. By monitoring
the safety of the pipelines, especially to the joints and elbows [30–32]. By monitoring the the slug
flow
slugcharacteristics,
flow characteristics,such as frequency,
such translational
as frequency, velocity,
translational and the
velocity, lengths
and of the of
the lengths slug
the
body and the liquid film regions, we will be able to assess their risks to the
slug body and the liquid film regions, we will be able to assess their risks to the pipeline pipeline systems
and the piping
systems and thecomponents, react immediately
piping components, if the slugifflow
react immediately behavior
the slug flow is determined
behavior to
is deter-
be damageable, and assess the multiphase flow rates for individual pipelines.
mined to be damageable, and assess the multiphase flow rates for individual pipelines.
In this research, low-frequency strain-rate analysis of the DAS is used to detect gas–oil
In this research, low-frequency strain-rate analysis of the DAS is used to detect gas–
two-phase slug flow and quantify its characteristics. Novel algorithms were developed to
oil two-phase slug flow and quantify its characteristics. Novel algorithms were developed
characterize the slug translational velocity, and the lengths of slug body and the liquid film
to characterize the slug translational velocity, and the lengths of slug body and the liquid
regions that had never been discussed in previous studies. These parameters are essential
film regions that had never been discussed in previous studies. These parameters are es-
for slug-induced risk assessment and flow rate estimation. This study also employed
sential for slug-induced risk assessment and flow rate estimation. This study also em-
a high-speed camera next to the fiber cables for validation purposes, achieving a good
ployed a high-speed camera next to the fiber cables for validation purposes, achieving a
correlation among the measurements under all conditions tested.
good correlation among the measurements under all conditions tested.
2. Materials and Methods
2. Materials and Methods
The two-phase gas–oil experiments were conducted in a three-phase flow loop at the
ColoradoTheSchool
two-phase gas–oil
of Mines, experiments
consisting were conducted
of a transparent in a three-phase
polyvinyl chloride (PVC) flowhorizontal
loop at the
Colorado
test sectionSchool
measuring of Mines,
14 mconsisting
long, a 5.25 ofcm
a transparent polyvinyl
inner diameter, and achloride
6.03 cm(PVC)
outer horizontal
diameter.
test section measuring 14 m long, a 5.25 cm inner diameter, and a 6.03
Figure 2 shows the schematic of the facility. Figure 3 is a picture of the test section, cm outer diameter.
high-
Figure
speed camera, and the pipe helically wrapped with yellow fiber cable. A fiber cable high-
2 shows the schematic of the facility. Figure 3 is a picture of the test section, was
speed camera,
wrapped helicallyandonthe pipe
a 35 cmhelically wrapped
long section and with yellow fiber
was connected to cable. A fiber
a Terra15 cable
Treble DASwas
wrapped helically
interrogator. on a 35
The winding cm is
pitch long section
0.112 cm. Theanddata
waswere
connected
acquired to awith
Terra15 Treble rate
a sampling DAS
ofinterrogator.
10 kHz and The winding
a special pitch isof0.112
sampling 0.816cm. The data
m along the were
fiber acquired
cable. Thewithgaugea sampling rate
length that
of 10
was kHz and
applied a special
is 2.4 m. The sampling
flow loopof had0.816 m along the
a differential fiber cable.
pressure The gauge
transducer length the
to measure that
was applied
pressure dropisin2.4 them.horizontal
The flow test
loopsection.
had a differential
Pressure andpressure transducer
temperature to measure
transducers werethe
pressureatdrop
installed in the
the test horizontal
section test section.
to monitor Pressure
the pressure and and temperature
temperature transducers
during were
the tests. The
installed
sensors at the
were all test section to monitor
connected the pressure
a data acquisition and temperature
system and recorded during the tests. The
at a frequency of
10sensors
Hz. were all connected to a data acquisition system and recorded at a frequency of 10
Hz.
 nector,
nector, and
and then
then the
the building
building airair from
from aa compressor
compressor was
was introduced
introduced afterward.
afterward. Various
Various
rates
rates for the gas and liquid phases were examined by varying the control valve for each
for the gas and liquid phases were examined by varying the control valve for each
phase
phase or
or the
the pump
pump motor
motor through
through thethe variable
variable frequency
frequency drive
drive (VFD).
(VFD). The
The superficial
superficial oil
oil
velocity
velocity (defined
(defined asas the
the volumetric
volumetric flow
flow rate
rate divided
divided by
by the
the pipe
pipe cross-sectional
cross-sectional area)
area) var-
var-
Sensors 2024, 24, 3402 ied
ied from
from 0.2
0.2 to
to 0.8
0.8 m/s,
m/s, while
while the
the gas
gas phase
phase varied
varied from
from 0.16
0.16 to
to 0.78
0.78 m/s.
m/s. The
The test
test matrix
4 of 17
matrix
presented
presented inin this
this work
work is is shown
shown inin Table
Table 1.
1.

Figure 2.
2. The
Figure 2. The schematic of
schematic of the
of the three-phase
the three-phase multiphase
three-phase multiphase flow
multiphase flow loop.
flow loop.
loop.
Figure The schematic

(a)
(a) (b)
(b)
Figure 3.
3. (a)
Figure 3. Pictures
Pictures of
(a) Pictures of the testing
the testing section
testing section with
section with the
with the helically
the helically wrapped
helically wrapped yellow
wrapped yellow fiber
yellow fiber cables,
fiber cables, and
cables, and the
and the
the
Figure
high-speed(a)camera of thevideos.
taking (b) A closer look at the wrapped fiber cable.
high-speed camera taking videos. (b) A closer look at the wrapped fiber cable.
high-speed camera taking videos. (b) A closer look at the wrapped fiber cable.
Table
Table 1.
1. Test
Test matrix
matrix in
in the
the current
current study.
study.
Mineral oil and compressed air were used as testing fluids in this study. The mineral
Case# ofSuperficial
810 kg/mOil 3 and ◦ C Velocity
Oil Velocity [m/s] Superficial
Pa·s at 19Gas [m/s]
1
oil had a density
Case# 1 Superficial a viscosity
Velocity [m/s]of 0.012 Superficial Gas and atmospheric
Velocity [m/s]
pressure. While the air had a density of 1.2 kg/m 3 and a viscosity of 1.83 × 10 −5 Pa·s
1–5
1–5 0.2
0.2 0.16,
0.16, 0.31,
0.31, 0.47,
0.47, 0.62,
0.62, 0.78
0.78
◦
at 19 C and atmospheric pressure. The oil was introduced to the system through a wye
6–9
6–9 0.5
0.5 0.16,
0.16, 0.31,
0.31, 0.47,
0.47, 0.62
0.62
inlet connector,
10–13 and then the building
0.8 air from a compressor was
0.16, introduced afterward.
10–13 0.8 0.16, 0.31,
0.31, 0.47,
0.47, 0.62
0.62
Various
1 The caserates
1 The case number
for the gas and liquid phases were examined by varying the control valve for
number ## in in each
each row
row corresponds
corresponds to to different
different superficial
superficial gas
gas velocity.
velocity. For
For example,
example,
each
Case#1phase or the pump motorsuperficial
through the velocity
variable frequency drive (VFD). The superficial
Case#1 corresponds
corresponds to to an
an 0.2
0.2 m/s
m/s superficial oil oil velocity and
and 0.16
0.16 m/s
m/s superficial
superficial gas
gas velocity;
velocity; Case#3
Case#3
oil velocity (defined
corresponds
corresponds to to an
an 0.2
0.2 m/s
as the volumetric
m/s superficial
superficial oil
flow rate
oil velocity
velocity and
divided
and 0.47
0.47 m/s
by the pipe
m/s superficial
superficial gas
cross-sectional
gas velocity.
velocity.
area)
varied from 0.2 to 0.8 m/s, while the gas phase varied from 0.16 to 0.78 m/s. The test matrix
The
The facility
presented in this was
facility was equipped
work is shownwith
equipped aa high-speed
in Table
with 1.
high-speed camera
camera (Phantom
(Phantom VEO640,Wayne,
VEO640,Wayne, NJ, NJ,
USA)
USA) to acquire a side view of the fluid flow behavior in the acrylic test
to acquire a side view of the fluid flow behavior in the acrylic test section
section (Figure
(Figure
1. Test matrix
3a).
3a). The
Table
The video
video wasinrecorded
was the current
recorded atstudy.
at 200
200 HzHz with
with aa resolution
resolution of of 1024
1024 ×× 700.
700. A A light
light source
source waswas
fixed behind
fixed behind the pipe to provide illumination. The camera was used to record the flow
Case#the 1 pipe to provide illumination.
Superficial Oil VelocityThe [m/s]cameraSuperficial
was usedGas to record
Velocitythe flow
[m/s]
pattern
pattern ofof testing
testing fluids
fluids that
that were
were circulated
circulated through
through the the pipe.
pipe. Camera
Camera videos
videos were
were also
also
processed 1–5 analyzed to obtain the slug 0.2 flow characteristic parameters
0.16, 0.31, 0.47, 0.62, 0.78
processed and and
6–9
analyzed to obtain the slug
0.5
flow characteristic parameters and
and used
used
0.16, 0.31, 0.47, 0.62
to
to val-
val-
idate
idate the
the ones
ones obtained
obtained from from DAS.
DAS.
10–13 0.8 0.16, 0.31, 0.47, 0.62
1 The case number # in each row corresponds to different superficial gas velocity. For example, Case#1 corresponds
to an 0.2 m/s superficial oil velocity and 0.16 m/s superficial gas velocity; Case#3 corresponds to an 0.2 m/s
superficial oil velocity and 0.47 m/s superficial gas velocity.

The facility was equipped with a high-speed camera (Phantom VEO640, Wayne, NJ,
USA) to acquire a side view of the fluid flow behavior in the acrylic test section (Figure 3a).
The video was recorded at 200 Hz with a resolution of 1024 × 700. A light source was fixed
behind the pipe to provide illumination. The camera was used to record the flow pattern of
testing fluids that were circulated through the pipe. Camera videos were also processed
and analyzed to obtain the slug flow characteristic parameters and used to validate the
ones obtained from DAS.
Sensors 2024, 24, x FOR PEER REVIEW 5 of 18

Sensors 2024, 24, 3402 5 of 17

3. Data Processing Workflow and Results
In this section, the data processing is discussed in two sections based on measure-
ment
3. Datatechnology.
Processing The first part,and
Workflow which is also the main one, describes the automated work-
Results
flow Indeveloped in this
this section, studyprocessing
the data to detect is
and track the
discussed inslugging
two sectionssignals in on
based themeasurement
gas–oil two-
phase slug flow and obtain the slug flow characteristic parameters from
technology. The first part, which is also the main one, describes the automated workflowthe DAS data. It
also includes a discussion of the workflow process and its accuracy. The
developed in this study to detect and track the slugging signals in the gas–oil two-phase second part
shows the high-speed camera data processing of the tested cases and the
slug flow and obtain the slug flow characteristic parameters from the DAS data. It also slug parameters
obtained.
includes aThe comparison
discussion of theofworkflow
the results from DAS
process and and the high-speed
its accuracy. camera
The second is shows
part discussed
the
afterward,
high-speedfollowed by aprocessing
camera data subsectionofshowing thecases
the tested slugandflowthevariations with theobtained.
slug parameters flowing
conditions.
The comparison of the results from DAS and the high-speed camera is discussed afterward,
followed by a subsection showing the slug flow variations with the flowing conditions.
3.1. DAS Data Processing Workflow
3.1. DAS
This Data Processing
workflow uses Workflow
low-frequency strain-rate analysis of the DAS signal to demon-
strateThis workflow uses
its capabilities low-frequency
of quantifying strain-rate
two-phase analysis
slug flow of the DAS
characteristic signal to demon-
parameters, includ-
strate
ing theitsslug
capabilities
frequency, of quantifying two-phase
slug body length, slugslug flow characteristic
translational velocity, parameters,
slug unit, andinclud-
the
liquid film region lengths. The DAS can capture these signals due to the sensitivitythe
ing the slug frequency, slug body length, slug translational velocity, slug unit, and of
liquid film
helically region fiber
wrapped lengths.to the The DASpipe
slight can diameter
capture these
change signals
that isdue to the
caused bysensitivity
the pressure of
helically wrapped
differential exertedfiber
by a to the slight
passing slug.pipe diameter change that is caused by the pressure
differential
The raw exerted
DAS databy a are
passing
read slug.
into the workflow for the selected beginning and ending
times.TheToraw DASnoises
reduce data areandread into the
enhance the workflow
analysis, thefor median
the selected
value beginning
is computed and and
ending re-
times. To reduce noises and enhance the analysis, the median
moved from the channel direction, and then a low-pass filter is used to filter the data, value is computed and
removed from
displaying the that
signals channel direction,
are lower thanand then
20 Hz. a low-pass
Figure 4a shows filter
theislow-frequency
used to filter thedata data,
for
displaying signals that are lower than 20 Hz. Figure 4a shows the
gas–oil two-phase slug flow. The horizontal axis is time, while the vertical axis shows thelow-frequency data for
gas–oil two-phase slug flow. The horizontal axis is time,
channel number, which corresponds to the fiber length from the interrogator. The while the vertical axis shows
the channel
wrapped testnumber,
section iswhich
from corresponds
around 45 m to to the
110 fiber
m. A length
zoomedfrom the interrogator.
portion of the data over The a
wrapped
fifteen-second period in Figure 4b, presented four slugs passing which are indicated by thea
test section is from around 45 m to 110 m. A zoomed portion of the data over
fifteen-second
red period inlines.
and blue inclined Figure tSU4b, presented
denotes four interval
the time slugs passing
betweenwhich the are
twoindicated
adjacentby the
slugs
red and blue inclined lines. tSU denotes the time interval between the
at a position in the test section around 80 m in fiber length from the interrogator (near the two adjacent slugs at
a position in the test section around
middle of the fiber cable wrapped section). 80 m in fiber length from the interrogator (near the
middle of the fiber cable wrapped section).

Figure 4.
4. (a)
(a) Low-frequency
Low-frequencyDAS DASsignal
signalfor
forslug
slug flow;
flow; (b)(b) fifteen-second
fifteen-second time
time selected
selected to illustrate
to illustrate the
the passing slugs in the signal.
passing slugs in the signal.

Figure 55 is
Figure is aa visualization
visualization ofof the
the DAS
DAS signal
signal of
of aa single slug passing through and the
identification
identification of ofthe
theslug
slugfront
frontand
andtail.
tail.One
Onewayway
to to obtain
obtain thethe slug
slug translational
translational velocity,
velocity, vTB ,
visTBto
, iscalculate
to calculate the slope
the slope of theofblue
the line,
blue which
line, which indicates
indicates thefront,
the slug slug front, given
given in in Equa-
Equation (1):
tion (1):
∅( L2 − L1 )
v TB = , (1)
t2 − t1
 correction factor that converts the fiber cable distance to the actual distance in the
section; (t2 − t1) represents the time required for this slug traveling from L1 to L2.
tSS in Figure 5 is the time interval between slug front and tail, 𝑡 −𝑡
Sensors 2024, 24, 3402 representing the time required for the slug body to pass a particular point6 of on17the test
tion. It can be used to determine the slug body length, given in Equation (2):

where (L2 − L1 ) represents the = 𝑣 𝑡between

𝐿 distance = 𝑣 two 𝑡 points − 𝑡
shown ,
in the DAS data; ∅ is
a correction factor that converts the fiber cable distance to the actual distance in the test
section; (t2 − t1 ) represents the time required for this slug traveling from L1 to L2 .

Figure 5.5.DAS
Figure DASsignal of aofsingle
signal slug slug
a single in thein
gas–oil two-phase
the gas–oil slug flow.slug flow.
two-phase
 
tSS in Figure 5 is the time interval between slug front and tail, tSlugTail − tSlugFront ,
The slug
representing the unit length, for
time required LU,the
canslug
be body
determined using Equation
to pass a particular point on (3), where
the test tSU is the t
section.
interval between
It can be used two adjacent
to determine the slugslugs
body (length, 4b): in Equation (2):
Figuregiven

𝐿 =𝑣 𝑡 , 
LS = v TB tSS = v TB tSlugTail − tSlugFront , (2)
The length of the film region, LF, are then calculated using Equations (4):
The slug unit length, LU , can be determined using Equation (3), where tSU is the time
𝐿 =𝐿 −𝐿 ,
interval between two adjacent slugs (Figure 4b):

UL =v TB SUt , (3)
In this paper, we propose a more automatic workflow for the determination of
translational
The length of the film region, LF , are then the
velocity and the lengths of slug unit,
calculated usingslug body,(4):
Equations and the film region.
following paragraphs describe this workflow.
We selected the data between LF = LU − LS 45
channels , to 110 in the data processing(4)to elimi
the excessive noise
In this paper, beyond athis
we propose morerange. These
automatic channels
workflow for are associated with
the determination the wrap
of the
fiber sectionvelocity
translational near the
andcamera section.
the lengths of the A sum
slug unit,isslug
then taken
body, andin
thethe
filmchannel direction a
region. The
taking
followingtheparagraphs
absolute describe
value ofthis
theworkflow.
signal and plotted against the time to identify the tim
eachWe selected
slug, whichthe data between channels
are presented as the45 peaks
to 110 ininthe data processing
Figure 6. Each of to eliminate
the slugsthewas giv
excessive noise beyond this range. These channels are associated
unique slug index number (ID#) to be used in comparison with the camera with the wrapped fiberdata at l
section near the camera section. A sum is then taken in the channel direction after taking
stages.
the absolute value of the signal and plotted against the time to identify the time of each
slug, which are presented as the peaks in Figure 6. Each of the slugs was given a unique
slug index number (ID#) to be used in comparison with the camera data at later stages.
The peaks in Figure 6 were counted to determine the slug frequency, fS , which is the
number of slugs divided by the recording time. Note that the peaks that are very close are
counted as one peak, which should correspond to a single slug but probably with a more
complex structure. This prevents the overcounting of slugs due to adjacent peaks when
calculating the slug frequency.
 Sensors 2024, 24, x FOR PEER REVIEW 7 of 18
Sensors 2024, 24, 3402 7 of 17

Figure 6. The peak value of the slugs over a three-minute recorded time for the case of 0.2 m/s oil
velocity and 0.16 m/s gas velocity.

The peaks in Figure 6 were counted to determine the slug frequency, fS, which is the
number of slugs divided by the recording time. Note that the peaks that are very close are
counted as
Figure6.6.The
one
Thepeak
peak,
peakvalue
which
valueofofthe
should
theslugs
correspond
slugsover
to a single
overa athree-minute
three-minuterecorded
slugtime
recordedtime
but for
probably
thecase
with a m/s
caseofof0.20.2
more oil
Figure for the m/s oil
complex structure. This prevents
velocity and 0.16 m/s gas velocity. the overcounting of slugs due to adjacent peaks when
velocity and 0.16 m/s gas velocity.
calculating the slug frequency.
To
The
To
determine
peaks inthe
determine
the slug translational
Figure were countedvelocity,
slug6 translational
we
to determine
velocity,
used
we used thesemblance, aaquantitative
slug frequency,
semblance, fS, whichmeas-
quantitative is the
mea-
ure of
number the ofwaveform
slugs similarity
divided by the that estimates
recording the
time. consistency
Note that the
sure of the waveform similarity that estimates the consistency of the waveforms from of the
peaks waveforms
that are very from
close dif-
are
ferent
countedchannels.
as one The
peak, process
which is explained
should in
correspond Figure
to a7 , which
single includes
slug but
different channels. The process is explained in Figure 7, which includes four plots: (a) slug four
probably plots:
with (a)
a slug
more
waterfall
complexplot,
waterfall plot, (b) semblance
structure.
(b) semblance
This preventsvs. velocity,
vs. velocity, (c)waterfall
waterfallofplot
the overcounting
(c) plot after
slugs due
after applying
to adjacent
applying linear
linear moveout
peaks
moveoutwhen
using velocity
calculating
using velocity that
thethat
slug gives the
frequency.
gives highest semblance value from the previous
the highest semblance value from the previous plot, (d) vertical plot, (d) vertical
stacking
To in
stacking inthe
thedistance
determine distance direction
the slug to to
translational
direction determine the the
velocity,
determine we width
widthused of negative
the negative
semblance,
of the apeak peak
quantitative
for slug forbody
slug
meas-
body length calculation.
ure ofcalculation.
length the waveform similarity that estimates the consistency of the waveforms from dif-
ferent channels. The process is explained in Figure 7, which includes four plots: (a) slug
waterfall plot, (b) semblance vs. velocity, (c) waterfall plot after applying linear moveout
using velocity that gives the highest semblance value from the previous plot, (d) vertical
stacking in the distance direction to determine the width of the negative peak for slug
body length calculation.

Figure7.7.An
Figure Anexample
exampleofofquality
qualitycheck
checkplots forfor
plots workflow
workflowevaluation at slug#45.
evaluation (a) Slug
at slug#45. waterfall
(a) Slug plot;
waterfall
plot;
(b) (b) semblance
semblance as a function
as a function of testing
of testing velocity;
velocity; (c) waterfall
(c) waterfall plotapplying
plot after after applying linear moveout
linear moveout using
using velocity
velocity thatthe
that gives gives the semblance
highest highest semblance
value in value
(b); (d)invertical
(b); (d)stacking
verticalinstacking in thedirection.
the distance distance
direction.
The plot generation processes are described as follows:
(a) The waterfall,
The plot generation
a plot processes
of distanceare described
versus time, as follows: for each slug.
is generated
Figure 7. An example of quality check plots for workflow evaluation at slug#45. (a) Slug waterfall
(a)
(b) The
plot;A waterfall,
(b)linear moveout
semblance a aplot
as of distance
correction
function versus
(LMO)
of testing time,(c)is
is applied,
velocity; generated
which is
waterfall a for
plot each
velocity
after slug.
correction
applying linear to shift
moveout
(b)
usingA linear
traces
velocityinmoveout
timegives
that correction
basedtheon (LMO)
an assumed
highest is velocity.
semblance applied,
value inwhich
The isvertical
best
(b); (d) a velocity correction
reflects
stacking theto
in the shift
most
distance
traces in time
consistency
direction. of based on an assumed
the waveform among velocity. The best
traces in time aftervelocity reflectsThe
the correction. the following
most con-
sistency
equationof the waveform
represents among
the linear tracescorrection:
moveout in time after the correction. The following
The plot represents
equation generationthe processes are described
linear moveout as follows:
correction:
d(n)
(a) The waterfall, a plot of distancetversus LMO ( n,time,
v) = is generated
, for each slug. (5)
v
(b) A linear moveout correction (LMO) is applied, which is a velocity correction to shift
traces ninrepresents
where time based theontrace
an assumed
number in velocity. The best
the channel velocity
direction, d(n)reflects the most
is the fiber con-
distance
sistency
of of then,waveform
the channel v represents among tracesvelocity,
the testing in time and
aftert LMO
the correction.
represents The following
the time shift
equation
applied torepresents
the channel. the linear moveout correction:
Sensors 2024, 24, 3402 8 of 17

(c) The semblance is a quantitative measure of the waveform similarity from different
channels, which is a metric commonly used in seismic processing. It is calculated
using the following equations:
!2
t=t0 +∆t N
S= ∑ ∑ f n (t − t LMO (n, v)) , (6)
t=t0 −∆t n =1

where S is the summation of all waveforms in the channel direction, which is then
squared and summed again in the time direction. fn (t) represents the data value of
channel n at time t. N is the total number of channels, ∆t is the half window length in
time for the semblance calculation, and t0 is the center of the time window.
t=t0 +∆t N
E= ∑ ∑ f n 2 (t − t LMO (n, v)) , (7)
t=t0 −∆t n=1

where E(t0 ) is the summation of the square of all waveforms in both channel and
time directions.
S−E
R(v) = , (8)
( N − 1) E
Finally, R is the normalized semblance value of all the channels for the testing velocity
v. An array of the different testing velocities versus semblance can be obtained, and
the best velocity is chosen at the maximum semblance value which represents the
highest consistency of waveforms after linear moveout correction. This can be used as
the estimation of the slug translational velocity after applying a fiber-length-to-pipe-
length ratio of 169:1.
(d) All the signals are stacked vertically in the channel direction after the linear moveout
correction using the best velocity, and the negative peak is identified with its start and
end times, which represent the slug’s front and tail.
The slug characteristic parameters, including the slug body length, slug unit length,
and the length of the film region, are calculated using Equations (2)–(4), respectively. An
example of one of the slugs with its calculated parameters from the process in this section
is shown in Table 2.

Table 2. Characteristic parameters for slug#45 determined from the DAS signals using the pro-
posed approach.

Parameters Value Parameters Value

Slug#ID 45 tSlugFront (s) 0.406
Slug occurrence time (s) 170.26 tTail (s) 0.668
Semblance before correction −0.008588 tSS (s) 0.262
Best velocity (m/s) 265.52 Slug Frequency (s−1 ) 0.239
Semblance after correction 0.602624 Slug Body Length, LS (m) 0.410
Slug translational velocity, vT (m/s) 1.565 Liquid Film Region Length, LF (m) 6.125
tSU (s) 4.176 Negative Peak Time (s) 0.500
Slug Unit Length, LU (m) 6.536 Negative Peak Value −3.487 × 10−6

To increase the accuracy of the quantitative analysis on the slug characteristic param-
eters, a threshold for the calculated semblance after correction is selected at 0.1 to filter
out some of the bad signals that may not represent an actual or complete slug signal and
do not give a good quantitative analysis. An example is shown in Figure 8, where the
automated workflow does not predict a reliable measurement for slug#12. This is also
observed in Table 3 where the semblance after correction is very low (highlighted in red).
In this example, the DAS signal in the selected window only contains a little part of a
slug body shown at the bottom left in the waterfall plot (Figure 8a), resulting in unreliable
quantification of the slug characteristics as anticipated. Nonetheless, these “bad” slugs
 some of the bad signals that may not represent an actual or complete slug signal and do
not give a good quantitative analysis. An example is shown in Figure 8, where the auto-
mated workflow does not predict a reliable measurement for slug#12. This is also ob-
served in Table 3 where the semblance after correction is very low (highlighted in red). In
Sensors 2024, 24, 3402
this example, the DAS signal in the selected window only contains a little part of a9 of 17
slug
body shown at the bottom left in the waterfall plot (Figure 8a), resulting in unreliable quan-
tification of the slug characteristics as anticipated. Nonetheless, these “bad” slugs repre-
represent a low percentage of the total population of the detected slugs that have reliable
sent a low percentage of the total population of the detected slugs that have reliable quan-
quantitative analysis and velocity measurement. Table 4 summarizes the percentage of
titative analysis and velocity measurement. Table 4 summarizes the percentage of the
the slugs with reliable quantitative analysis. The characteristic parameters of the slug
slugs with reliable quantitative analysis. The characteristic parameters of the slug were
were analyzed for these slugs and evaluated with data from the high-speed camera. The
analyzed for these slugs and evaluated with data from the high-speed camera. The com-
comparison is discussed in the later sub-section.
parison is discussed in the later sub-section.

Figure
Figure8.8.An
Anexample
exampleofofaabad
badslug
slug(ID
(ID#12)
#12) that
that does
does not give good
good quantitative
quantitativeanalysis.
analysis.(a)
(a)Slug
Slug
waterfall
waterfallplot;
plot;(b)
(b)semblance
semblanceasasaafunction
functionof
oftesting
testing velocity;
velocity; (c)
(c) waterfall plot after
waterfall plot after applying
applyinglinear
linear
moveout
moveoutusing
usingvelocity
velocitythat
thatgives
givesthe
thehighest
highest semblance
semblance value
value in (b); (d)
in (b); (d) vertical
vertical stacking
stackingin
inthe
the
distance direction.
distance direction.

Table
Table3.3.Characteristic
Characteristicparameters
parametersfor
forslug#12
slug#12determined
determined from
from the DAS signals
the DAS signals using
usingthe
theproposed
proposed
approach that does not give good quantitative analysis.
approach that does not give good quantitative analysis.
Parameters Value Parameters Value
Parameters Value Parameters Value
Slug#ID 412 tSlugFront (s) 0.000
Slug#ID 412 tSlugFront (s) 0.000
Slug occurrence time (s)
Slug occurrence time (s)
43.074
43.074
tTail (s)
tTail (s)
0.673
0.673
Semblance
Semblancebefore
before correction
correction −0.004933
−0.004933 tSS(s)
tSS (s) 0.673
0.673
Best
Bestvelocity
velocity (m/s) 279.31
279.31 SlugFrequency
Slug Frequency (s−(s
1 )−1
) 1.429
1.429
Semblanceafter
Semblance aftercorrection
correction 0.024887
0.024887 Slug Body Length,S L(m)
Slug Body Length, L S (m)
1.108
1.108
Slug translational velocity, vT (m/s) 1.6464 Liquid Film Region Length, LF (m) 0.046
Slug translational tSUvelocity,
(s) vT (m/s) 1.6464
0.7 LiquidNegative
Film Region
Peak Time Length,
(s) LF (m) 0.046
0.045
t (s)
Slug Unit Length, LU (m)
SU 0.7
1.154 Negative
Negative Peak Value (s)
Peak Time 0.045
0.363 × 10−6
Slug Unit Length, LU (m) 1.154 Negative Peak Value 0.363 × 10−6

Table 4. Percentage of the slugs with reliable quantitative analysis from the DAS automated algorithm.
Table 4. Percentage of the slugs with reliable quantitative analysis from the DAS automated algo-
rithm. Case # v [m/s] v [m/s] P * [%]
SO SG

Case # 1 vSO [m/s] vSG [m/s] 0.16 P * [%]

78
1 2 0.16 0.31 7884
3 0.2 0.47 88
2 4 0.31 0.62 8473
3 5 0.2 0.47 0.35 8863
4 6 0.62 0.16 7381
5 7 0.35 0.31 6390
0.5
6 8 0.5 0.16 0.47 8188
9 0.62 94
10 0.16 94
11 0.31 94
0.8
12 0.47 95
13 0.62 98
* P: Percentage of slugs with reliable quantitative analysis and used for slug characteristic parameter determination.
 10 0.16 94
11 0.31 94
0.8
12 0.47 95
13 0.62 98
Sensors 2024, 24, 3402 * P: Percentage of slugs with reliable quantitative analysis and used for slug characteristic
10 parameter
of 17
determination.

The
The summary
summary of of
thethe workflow
workflow steps
steps in the
in the DAS DAS
datadata processing
processing andslug
and the the charac-
slug charac-
teristicparameters
teristic parameters that
that cancan
be be obtained
obtained is presented
is presented in Figure
in Figure 9. 9.

Figure
Figure Workflow
9. 9. of the
Workflow DAS
of the datadata
DAS processing for slug
processing flow flow
for slug characterization.
characterization.
3.2. High-Speed Camera Data Processing
3.2. High-Speed Camera Data Processing
The camera data processing involves the recorded videos from the high-speed camera
of the The
flowcamera
at each data
of theprocessing involves flow
gas–oil two-phase the recorded videos in
cases presented from
Tablethe1high-speed
previously. cam-
era of the flow at each of the gas–oil two-phase flow cases presented in
Each video is recorded after reaching the stabilization of the flow for about three minutesTable 1 previously.
atEach video isof
a frequency recorded
200 Hz.after reaching
Figure the stabilization
10 presents a capture ofof the flow
a slug for about
passing by forthree minutes
a liquid
at a frequency
superficial velocityofat200 Hz. and
0.2 m/s Figure presentsvelocity
gas10superficial a capture of am/s.
at 0.16 slugApassing
scale wasbyattached
for a liquid
tosuperficial velocity
the pipe which wasatused
0.2 m/s
as a and
pointgas superficial
of entry for thevelocity at 0.16
slug’s front andm/s.
tail.AThe
scale was attached
videos are
to the
then pipe which
analyzed was
entirely used
and asslug
each a point
that of entry forwould
is passing the slug’s
have front andslug
a unique tail. index
The videos
and are
the following
then analyzed data processing
entirely and for it:slug that is passing would have a unique slug index and
each
the The
(a) following
time when data the
processing
slug frontfor it: is determined once reaching the scale, tSF .
starts
(b)
(a) TheThetime
timewhenwhen thethe
slug tailfront
slug reaches theisscale
starts is documented,
determined tST .
once reaching the scale, tSF.
(c)(b) The
The time when the slug tail reaches the scale is documented,camera
time when the slug front reaches the beginning of the tST. exposure is
recorded,
(c) The time SF_In t .
when the slug front reaches the beginning of the camera exposure is rec-
(d) The time when the slug front reaches the end of the camera exposure is recorded, tSF_Out .
orded, tSF_In.
(e) Translational velocity, vT , is obtained by finding the length of the horizontal section
(d) The time when the slug front reaches the end of the camera exposure is recorded,
that is exposed to the camera divided by the duration of exposure of each slug, i.e.,
vTtSF_Out
= L/(t.
SF_Out − tSF_In ).
(f)(e) Slug
Translational
unit lengthvelocity, vT, isby
is determined obtained
LU = vT by finding
(tSF1 the
− tSF2 ), length
where of −
(tSF1 the horizontal
tSF2 section
) is the time
that is exposed to the camera
interval between two adjacent slugs. divided by the duration of exposure of each slug, i.e.,
(g) The vT =length
L/(tSF_Out − tSF_In
of the slug). body is determined by LS = vT (tSF − tST ), and the length of the
film region is determined using Equation 4 as presented previously.
(h) Slug frequency, fS , is determined by counting the number of slugs divided by the
corresponding recording time.
(i) For each of the slug characteristic parameters obtained, the average and the median
values were calculated over the full three-minute duration of the recorded video.
 (g) The length of the slug body is determined by LS = vT (tSF − tST), and the length of the
film region is determined using Equation 4 as presented previously.
(h) Slug frequency, fS, is determined by counting the number of slugs divided by the
corresponding recording time.
Sensors 2024, 24, 3402 11 of 17
(i) For each of the slug characteristic parameters obtained, the average and the median
values were calculated over the full three-minute duration of the recorded video.

Figure10.
Figure 10.Slug
Slugflow
flowcaptured
capturedfrom
fromaarecorded
recordedvideo
videoby
byaahigh-speed
high-speedcamera
cameraatat0.16
0.16m/s
m/s gas
gas velocity
velocity
and 0.2 m/s oil velocity.
and 0.2 m/s oil velocity.

3.3.
3.3. Data
Data Validation
Validation
The
The accuracy
accuracyof of the
the DAS
DAS workflow
workflow process
processisis validated
validated usingusing the
the high-speed
high-speed camera
camera
videos
videos which
which serve
serve as
as the
the ground
groundtruth.
truth. An
An example
example of of the
the comparison
comparison isis presented
presented inin
Figure 11, which shows the comparison of v , L , L , and L , between
Figure 11, which shows the comparison of vT, LU, LS, and LF, between the data obtained
T U S F the data obtained
from DAS (red points) and the high-speed camera videos (black points). The oil superficial
Sensors 2024, 24, x FOR PEER REVIEWfrom DAS (red points) and the high-speed camera videos (black points). The oil superficial 12 of 18
velocity
velocityisis0.2
0.2m/s,
m/s, and
and the
the gas
gas superficial
superficialvelocity
velocityisis0.16 0.16m/s.
m/s. A A good
good match
match between
between the
the
DAS
DAS data
dataand
andthe
thehigh-speed
high-speedcamera
cameradata
dataisisobserved.
observed.

Figure 11. Comparison between the slug characteristic parameters from the DAS data and the camera
Figure
for the11. Comparison
case of 0.2 m/s between the slug
oil superficial characteristic
velocity and 0.16 parameters from thevelocity.
m/s gas superficial DAS data and the cam-
era for the case of 0.2 m/s oil superficial velocity and 0.16 m/s gas superficial velocity.
One might have noticed that the slug characteristics are not the same at the same flow-
ing condition.
One mightThis havefluctuation
noticed thatis the
thenature of slug flow. Dozens
slug characteristics are notofthe
previous
same atstudies have
the same
tried to condition.
flowing develop statistical models using
This fluctuation is the laboratory dataflow.
nature of slug to predict
Dozens theofslug characteristics,
previous studies
especially
have thedevelop
tried to slug length whichmodels
statistical is a critical
usingparameter
laboratoryfordata
facility design the
to predict [33–35].
slug With the
charac-
development
teristics, of DAS,
especially we can
the slug nowwhich
length monitor
is a the slugparameter
critical characteristics directly
for facility in the[33–35].
design field in
real time.
With the development of DAS, we can now monitor the slug characteristics directly in the
field in real time.
To further evaluate the performance of DAS and the data processing workflow, we
generated cross plots for all the gas–oil two-phase experimental data points, and the com-
parison of the median values is shown in Figure 12. Considering the median value of each
slug parameter in the 13 experiments over the three-minute period of recorded data, a
 Sensors 2024, 24, 3402 12 of 17

To further evaluate the performance of DAS and the data processing workflow, we
generated cross plots for all the gas–oil two-phase experimental data points, and the
comparison of the median values is shown in Figure 12. Considering the median value of
each slug parameter in the 13 experiments over the three-minute period of recorded data,
a very good match can be observed with the camera data with an error bar of less than
20% in the majority of the cases. One astonishing observation is that the DAS can detect
the number of slugs or quantify the slug frequency, very successfully for the conditions
investigated in the current study. For the other parameters, the error seems to increase with
the increase in gas flow rate or decrease in liquid flow rate. One of the reasons could be
the reduced number of slugs investigated as the gas flow rate increases or the liquid flow
rate decreases, so a smaller number of measurements are averaged. Another reason could
be the change in the physical properties of the slug body, i.e., more gas is present in the
Sensors 2024, 24, x FOR PEER REVIEW slug body as the gas flow rate increases or liquid flow rate decreases, and the liquid 13 ofslugs
18
become more chaotic, frothy, and shorter, making the signal less clear and more difficult to
be processed [31,32,36]. This phenomenon is illustrated in the images from the high-speed
camera in Figure 13. On the other hand, we also assessed the measurement uncertainty
uncertainty from the high-speed
from the high-speed camera.
camera. The The
errors forerrors for translational
translational velocity,velocity, slug
slug unit, andunit,
body
and body lengths range from 0.87% to 1.1%. These errors are not visible
lengths range from 0.87% to 1.1%. These errors are not visible in Figure 12.in Figure 12.

Figure
Figure 12.12. Comparison
Comparison between
between thethe median
median value
value from
from DAS
DAS and
and high-speed
high-speed camera.
camera.
Sensors 2024, 24, 3402 13 of 17
Figure 12. Comparison between the median value from DAS and high-speed camera.

Figure 13. Flow

Figure13. patterncaptured
Flow pattern captured from
from thethe high-speed
high-speed camera
camera at 0.2at 0.2oilm/s
m/s oil superficial
superficial velocity velocity
and and
differentgas
different gas flow
flow rates.
rates.
Sensors 2024, 24, x FOR PEER REVIEW 14 of 18
3.4. Slugs at Different Flowing Conditions
In this subsection, we will show the slug flow behavior at different flowing conditions.
Figure3.4.
14Slugs at Different Flowing Conditions
shows the DAS signal after applying the low-pass filter and the median value
In this subsection, we will show the slug flow behavior at different flowing condi-
removal in the channel direction, for three different liquid flow rates and four gas flow
tions. Figure 14 shows the DAS signal after applying the low-pass filter and the median
rates. value
The removal
time duration is 3 s for all the figures. It can be clearly noticed that the slug
in the channel direction, for three different liquid flow rates and four gas
frequency increases
flow rates. as duration
The time the liquid is 3flow rate
s for all theincreases. Moreover,
figures. It can the lines
be clearly noticed thatbecome
the slug steeper as
the gasfrequency
or liquid velocity
increases increases,
as the liquid flowindicating an increase
rate increases. Moreover,in the
the slug
lines translational
become steeper velocity.
as the gas or liquid
This observation velocity increases,
is consistent with other indicating
previousan increase
studiesinforthegas–liquid
slug translational
flowve-in horizontal
locity. This observation is consistent with other previous studies for gas–liquid flow in
pipes [37]. The slug body length cannot be directly read from the plots, as it is a function of
horizontal pipes [37]. The slug body length cannot be directly read from the plots, as it is
time duration
a function and also
of time the slug
duration and translational velocity. velocity.
also the slug translational

Figure 14. DAS signals at different flowing conditions (vSL: liquid superficial velocity, vSG: gas su-
14. DAS
Figureperficial signals
velocity. at different
Superficial velocityflowing
is definedconditions (vSL : flow
as the volumetric liquid
rate superficial velocity,
divided by the pipe vSG : gas
superficial velocity.area).
cross-sectional Superficial velocity is defined as the volumetric flow rate divided by the pipe
cross-sectional area).
The pictures of some slugs in Figure 14 are shown in Figures 15 and 16 for 0.5 m/s
and 0.8 m/s liquid superficial velocities, respectively. As the gas velocity increases, more
gas bubbles are entrained inside the slug body due to the high turbulence in the mixing
zone at the slug front [38]. This increases the noise in the DAS signal, as depicted in Figure
14. We suspect that the intensity of the DAS signal within the slug body may correlate
with the turbulence and entrained gas bubbles within the slug body. Further investiga-
tions are necessary to better understand this relationship quantitatively.
Sensors 2024, 24, 3402 14 of 17

The pictures of some slugs in Figure 14 are shown in Figures 15 and 16 for 0.5 m/s and
0.8 m/s liquid superficial velocities, respectively. As the gas velocity increases, more gas
bubbles are entrained inside the slug body due to the high turbulence in the mixing zone
at the slug front [38]. This increases the noise in the DAS signal, as depicted in Figure 14.
We suspect that the intensity of the DAS signal within the slug body may correlate with
Sensors 2024, 24, x FOR PEER REVIEW
the turbulence and entrained gas bubbles within the slug body. Further investigations 15 ofare
18
Sensors 2024, 24, x FOR PEER REVIEW 15 of 18
necessary to better understand this relationship quantitatively.

Figure 15.
Figure 15. Pictures
Picturesfrom
fromhigh-speed
high-speedcamera
cameravideos
videosfor
forthe
theslugs
slugsshown
shownininFigure
Figure1414for
forv vSL==0.5
0.5m/s.
m/s.
Figure 15. Pictures from high-speed camera videos for the slugs shown in Figure 14 forSL vSL = 0.5 m/s.

Figure 16. Pictures from high-speed camera videos for the slugs shown in Figure 14 for vSL = 0.8 m/s.
Figure16.
Figure 16.Pictures
Picturesfrom
fromhigh-speed
high-speedcamera
cameravideos
videosfor
forthe
theslugs
slugsshown
shownininFigure
Figure1414 for
for vSL==0.8
vSL 0.8m/s.
m/s.
4. Discussion
4.
4. Discussion
Discussion
Slug flow characterization is of great importance to many field applications. In the
Slug
Slug flow characterization is
is of
of great importance to to many field applications. In
In the
oil and gasflow characterization
industry, knowing the sluggreat
flowimportance
characteristics, manysuchfieldas the applications.
slug body length, the
oil
oil and
and gas
gas industry,
industry, knowing
knowing the
the slug
slug flow
flow characteristics,
characteristics, such
such as
as the
the slug
slug body
body length,
length,
translational velocity, and frequency, is crucial for facility design and operation. For ex-
translational
translational velocity, and
velocity, and frequency, isis crucial crucialfor forfacility
facility designand and operation. ForFor
ample, the size and internalfrequency,design of the separator should design
be able to handle operation.
the slugs ex-
to
example,
ample, the
the size
size and and internal
internal design
design of of
the the separator
separator should
should be be
able ableto to handle
handle thethe slugs
slugs to
avoid flooding or ineffective separation. Slug catcher design also requires information on
to avoid
avoid flooding
flooding or or ineffective
ineffective separation.
separation. Slug Slug catcher
catcher design
design alsoalso requires
requires information
information on
the slug flow characteristic parameters to achieve optimal gas–liquid separation [39]. Top-
on
thethe
slugslug
flow flow characteristic
characteristic parameters to achieve optimal gas–liquid separation [39].
side
Topsideslugging
slugging control
control aparameters
for for riser
a system
riser
to requires
system
achieve
requires
optimal
real-time gas–liquid
real-time slugslug flow separation
flow monitoring
monitoring
[39]. Top-
thatthatis
side slugging
automated with control
the for a valve
control riser system
systems requires
[40]. real-timeslug
Moreover, slugflowflow canmonitoring
accelerate that
corro- is
is automated
automated with
with the the control valve systems [40]. Moreover, slug flow can accelerate
sion and erosion
corrosion and erosion duecontrol
todue theirvalve
highsystems
to their translational
high
[40]. Moreover,
velocities
translational
slugthe
and
velocities
flow
and
can accelerate
highly turbulent
the highly
corro-
zoom
turbulent
sion
at theand
slug erosion
front due
that to
could their high
possibly translational
lead to velocities
cavitation and
[41,42]. the
Knowing highly the turbulent
slug flow zoom
zoom
at the at thefront
slug slugthatfrontcouldthatpossibly
could possibly
lead to lead to cavitation
cavitation [41,42]. [41,42]. Knowing
Knowing the slug the char-
flow slug
char-
acteristics,
flow particularly
characteristics, in real time,
particularly in empowers
real time, us to more
empowers us toconfidently
more assess the
confidently assessrisk of
acteristics, particularly
slug-flow-induced or in real time,
-accelerated empowers
corrosion and us to more
erosion and confidently
adjust the assess theparam-
operating riskthe of
risk of slug-flow-induced or -accelerated corrosion and erosion and adjust the operating
slug-flow-induced
eters accordingly or -accelerated
to minimize corrosion
the risks. and erosion and adjust the operating param-
parameters accordingly to minimize theThe current
risks. most widely
The current most used widely method in the field
used method in
eters accordingly
for field
multiphase to minimize
surface facilities the risks.
design and The current
flowand most
assurance widely used
is basedisonbased method
predictive in the
modelsfield
the for multiphase surface facilities design flow assurance on predictive
for multiphase
that are that
developed surface based facilities
onbased design and
laboratory dataflow or assurance
very is based on predictive models
models are developed on laboratory data limited
or very fieldlimited data. Their
field data.accuracy
Their
that
becomesare developed
questionable based
as it on
is laboratory
scaled up to data
the or
field very limited
application. field
Because data. theTheir
slug accuracy
flow be-
accuracy becomes questionable as it is scaled up to the field application. Because the slug
becomes
havior is questionable
transient and as it is scaled diverse
dramatically up to the in field
both application.
the spatial Because
and time the slug flow
domains, even be-
at
flow behavior is transient and dramatically diverse in both the spatial and time domains,
havior is transient and dramatically diverse in both the
the same flowing conditions [22,43], real-time monitoring in the field will empower us tospatial and time domains, even at
the same
more flowingtrack
effectively conditions [22,43], real-time
their behaviors, allowing monitoring
us to better in optimize
the field will empower
facility designus andto
more effectively
operation. track their behaviors, allowing us to better optimize facility design and
operation.
DAS is advanced in many aspects compared with other point sensors, such as γ-ray
DASwire-mesh
sensors, is advanced in many
sensors, aspectsimpedance
electrical compared(capacitance
with other point sensors, such sensors,
and conductance) as γ-ray
Sensors 2024, 24, 3402 15 of 17

even at the same flowing conditions [22,43], real-time monitoring in the field will empower
us to more effectively track their behaviors, allowing us to better optimize facility design
and operation.
DAS is advanced in many aspects compared with other point sensors, such as γ-ray
sensors, wire-mesh sensors, electrical impedance (capacitance and conductance) sensors,
microwave sensors, optical sensors, etc., as previously mentioned in the introduction.
It is completely non-intrusive and does not disturb the existing facility. It can provide
measurements at multiple different locations and over a long distance in real time, instead
of a point measurement like the other types of sensors, making it more cost-effective.
Another big advantage is that DAS can work in harsh environments like downhole, where
the other type of sensor can hardly survive.
In this paper, we introduce a new method for characterizing the slug translational
velocity using semblance, showing great alignment with the observations from the high-
speed camera that serve as the ground truth. Moreover, we demonstrated a new algorithm
to characterize lengths of slug body, slug unit, and the liquid film region, which are the
most critical parameters required in field design and operation as discussed above. The
algorithm requires the fiber to be densely wrapped around the pipe, which increases the
sensitivity of the strain rate measurement due to the pipe diameter variations caused by
the slug-induced pressure fluctuations (see Figure 5).
The slug flow behavior should be related to different types of fluids and compositions,
and the DAS signal should respond differently depending on the fluid properties and the
slug flow behaviors. In the next phase, we are going analyze the DAS signals for other
different fluid compositions, including gas–water, oil–water, and gas–oil–water flows.
Furthermore, we will investigate the causes of the close peaks in Figure 6. There is a
possibility that these close peaks are induced by large gas pockets within long slug bodies,
but we will further investigate this phenomenon from both fluid flow dynamics and data
processing perspectives. Frequency domain dynamic averaging (FDDA) and/or activation
function dynamic averaging (AFDA) methods will be tested.

5. Conclusions
A series of experiments to investigate the capability of DAS to quantify the slug
characteristics were conducted in a horizontal pipe under different flowing conditions. A
low pass filter and median value removal were applied to better extract the slug signals. We
introduce a new method for characterizing the slug translational velocity using semblance,
showing great alignment with the observations from the high-speed camera that serve
as the ground truth. Moreover, we demonstrated a new algorithm to characterize the
lengths of the slug body, slug unit, and the liquid film region. The algorithm requires
the fiber to be densely wrapped around the pipe, which increases the sensitivity of the
strain rate measurement due to the pipe diameter variations caused by the slug-induced
pressure fluctuations.
Real-time slug flow monitoring and characterization are crucial for many industries
that involve slug slow, such as petroleum engineering, geothermal wells that have two-
phase flow production, nuclear engineering, etc. This research demonstrates the capabilities
of DAS to quantitatively measure the slug flow characteristic parameters, especially the
lengths of the slug body, slug unit, and the liquid film region, which are the most critical
parameters required in field facility design and operation. This technology will offer the
industry a more optimal solution for facility design and ensure safer operational practices.

Author Contributions: Conceptualization, S.A., G.J. and Y.F.; methodology, S.A., G.J. and Y.F.;
software, S.A., G.J. and Y.F.; validation, S.A., G.J. and Y.F.; formal analysis, S.A., G.J. and Y.F.;
investigation, S.A., G.J. and Y.F.; resources, G.J. and Y.F.; data curation, S.A.; writing—original draft
preparation, S.A.; writing—review and editing, G.J. and Y.F.; visualization, G.J. and Y.F.; supervision,
G.J. and Y.F.; project administration, Y.F.; funding acquisition, Y.F. All authors have read and agreed
to the published version of the manuscript.
Sensors 2024, 24, 3402 16 of 17

Funding: This research received no external funding.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data may be provided upon request.
Acknowledgments: The authors acknowledge Ge Jin’s and Yilin Fan’s research groups’ support
throughout this work. We also acknowledge Denghong Zhou, Kanat Karatayev, and Mouna-Keltoum
Benabid for their significant help with running the experiments. We appreciate the scholarship
provided by the Kuwait Petroleum Corporation (KPC), which supported Sharifah Ali’s Ph.D. study
at the Colorado School of Mines.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Stalford, H.; Ahmed, R. Intelligent Casing-Intelligent Formation (ICIF) Design. In Proceedings of the Offshore Technology
Conference, Houston, TX, USA, 5–8 May 2014; p. D011S012R008.
2. Zhong, Z.Y.; Zhi, X.L.; Yi, W.J. Oil Well Real-time Monitoring with Downhole Permanent FBG Sensor Network. In Proceedings of
the 2007 IEEE International Conference on Control and Automation, Guangzhou, China, 30 May–1 June 2007; pp. 2591–2594.
3. Falcone, G.; Hewitt, G.; Alimonti, C. Multiphase Flow Metering: Principles and Applications; Elsevier: Amsterdam, The Nether-
lands, 2009.
4. Mukhopadhyay, S.C. New Developments in Sensing Technology for Structural Health Monitoring; Springer: Berlin/Heidelberg,
Germany, 2011.
5. Hill, D.; Fellow, S. Managing Oil and Gas with Fibre-Optic Sensing; EIC Energy Focus: London, UK, 2014.
6. Kinet, D.; Mégret, P.; Goossen, K.W.; Qiu, L.; Heider, D.; Caucheteur, C. Fiber Bragg grating sensors toward structural health
monitoring in composite materials: Challenges and solutions. Sensors 2014, 14, 7394–7419. [CrossRef]
7. Hartog, A.H. An Introduction to Distributed Optical Fibre Sensors; CRC Press: Boca Raton, FL, USA, 2017.
8. Boering, M.; Braal, R.; Cheng, L.K. Toward the next fiber optic revolution and decision making in the oil and gas industry. In
Proceedings of the Offshore Technology Conference, Houston, TX, USA, 6–9 May 2013. [CrossRef]
9. In’t Panhuis, P.; den Boer, H.; van der Horst, J.; Paleja, R.; Randell, D.; Joinson, D.; McIvor, P.; Green, K.; Bartlett, R. Flow monitoring
and production profiling using DAS. In SPE Annual Technical Conference and Exhibition? SPE: Kuala Lumpur, Malaysia, 2014.
[CrossRef]
10. Raab, T.; Reinsch, T.; Aldaz Cifuentes, S.R.; Henninges, J. Real-time well-integrity monitoring using fiber-optic distributed
acoustic sensing. SPE J. 2019, 24, 1997–2009. [CrossRef]
11. Boone, K.; Ridge, A.; Crickmore, R.; Onen, D. Detecting leaks in abandoned gas wells with fibre-optic distributed acoustic sensing.
In Proceedings of the IPTC 2014: International Petroleum Technology Conference, Doha, Qatar, 19–22 January 2014; European
Association of Geoscientists & Engineers: Rotterdam, The Netherlands, 2014. [CrossRef]
12. Jin, G.; Roy, B. Hydraulic-fracture geometry characterization using low-frequency DAS signal. Lead. Edge 2017, 36, 975–980.
[CrossRef]
13. Mateeva, A.; Lopez, J.; Mestayer, J.; Wills, P.; Cox, B.; Kiyashchenko, D.; Yang, Z.; Berlang, W.; Detomo, R.; Grandi, S. Distributed
acoustic sensing for reservoir monitoring with, V.S.P. Lead. Edge 2013, 32, 1278–1283. [CrossRef]
14. van der Horst, J.; den Boer, H.; in’t Panhuis, P.; Kusters, R.; Roy, D.; Ridge, A.; Godfrey, A. Fibre Optic Sensing for Improved
Wellbore Surveillance. In Proceedings of the International Petroleum Technology Conference, Beijing, China, 26–28 March 2013.
[CrossRef]
15. Jin, G.; Friehauf, K.; Roy, B. The calibration of double-ended distributed temperature sensing for production logging purposes.
In Proceedings of the SPE/AAPG/SEG Unconventional Resources Technology Conference, Denver, CO, USA, 22–24 July 2019.
[CrossRef]
16. Jin, G.; Friehauf, K.; Roy, B.; Constantine, J.J.; Swan, H.W.; Krueger, K.R.; Raterman, K.T. Fiber optic sensing-based production
logging methods for low-rate oil producers. In Proceedings of the SPE/AAPG/SEG Unconventional Resources Technology
Conference, Denver, CO, USA, 22–24 July 2019. [CrossRef]
17. Titov, A.; Fan, Y.; Kutun, K.; Jin, G. Distributed Acoustic Sensing (DAS) Response of Rising Taylor Bubbles in Slug Flow. Sensors
2022, 22, 1266. [CrossRef]
18. Titov, A.; Jin, G.; Fan, Y.; Tura, A.; Kutun, K.; Miskimins, J. Distributed fiber-optic sensing based production logging investigation:
Flowloop experiments. First EAGE Workshop on Fibre Optic Sensing. Eur. Assoc. Geosci. Eng. 2020, 2020, 1–5. [CrossRef]
19. Titov, A.; Fan, Y.; Jin, G.; Tura, A.; Kutun, K.; Miskimins, J. Experimental investigation of distributed acoustic fiber-Optic Sensing
in production logging: Thermal slug tracking and multiphase flow characterization. In SPE Annual Technical Conference and
Exhibition? Day 4 Thu, October 29, 2020; SPE: Kuala Lumpur, Malaysia, 2020. [CrossRef]
Sensors 2024, 24, 3402 17 of 17

20. Weber, G.H.; dos Santos, E.N.; Gomes, D.F.; Santana, A.L.B.; da Silva, J.C.C.; Martelli, C.; Pipa, D.R.; Morales, R.E.; de Camargo
Júnior, S.T.; da Silva Junior, M.F.; et al. Measurement of Gas-Phase Velocities in Two-Phase Flow Using Distributed Acoustic
Sensing. IEEE Sens. J. 2023, 23, 3597–3608. [CrossRef]
21. Al-Safran, E.M.; Brill, J.P. Applied Multiphase Flow in Pipes and Flow Assurance: Oil and Gas Production; Society of Petroleum
Engineers: Kuala Lumpur, Malaysia, 2017.
22. Al-Kayiem, H.H.; Mohmmed, A.O.; Al-Hashimy, Z.I.; Time, R.W. Statistical assessment of experimental observation on the slug
body length and slug translational velocity in a horizontal pipe. Int. J. Heat Mass Transf. 2017, 105, 252–260. [CrossRef]
23. Raimondi, L. Gas/liquid two-phase flow in pipes: Slugs, classical flow-map, and 1D compositional simulation. SPE J. 2022, 27,
532–551. [CrossRef]
24. Ekinci, S. Pipe Inclination Effects on Slug Flow Characteristics of High Viscosity Oil-Gas Two-Phase Flow. Master’s Thesis, The
University of Tulsa, Tulsa, OK, USA, 2015.
25. Brito, R. Effect of Medium Oil Viscosity on Two-Phase Oil-Gas Flow Behavior in Horizontal Pipes. Master’s Thesis, The University
of Tulsa, Tulsa, OK, USA, 2012.
26. Liu, Y.; Upchurch, E.R.; Ozbayoglu, E.M. Experimental Study of Single Taylor Bubble Rising in Stagnant and Downward Flowing
Non-Newtonian Fluids in Inclined Pipes. Energies 2021, 14, 578. [CrossRef]
27. Gokcal, B. An Experimental and Theoretical Investigation of Slug Flow For High Oil Viscosity in Horizontal Pipes. Ph.D. Thesis,
The University of Tulsa, Tulsa, OK, USA, 2008.
28. Shoham, O. Mechanistic Modeling of Gas-Liquid Two-Phase Flow in Pipes; Society of Petroleum Engineers: Kuala Lumpur,
Malaysia, 2006.
29. Zhang, H.-Q.; Wang, Q.; Sarica, C.; Brill, J.P. Unified Model for Gas-Liquid Pipe Flow via Slug Dynamics—Part 1: Model
Development. J. Energy Res. Technol. 2003, 125, 266–273. [CrossRef]
30. Ludwig, E.E. Applied Process Design for Chemical and Petrochemical Plants: Volume 1; Elsevier Science: Amsterdam, The Nether-
lands, 1995.
31. Fan, Y.; Soedarmo, A.; Pereyra, E.; Sarica, C. A comprehensive review of pseudo-slug flow. J. Pet. Sci. Eng. 2022, 217, 110879.
[CrossRef]
32. Fan, Y.; Pereyra, E.; Sarica, C. Experimental study of pseudo-slug flow in upward inclined pipes. J. Nat. Gas Sci. Eng. 2020,
75, 103147. [CrossRef]
33. Bernicot, M.F.; Drouffe, J.-M. A Slug-Length Distribution Law for Multiphase Transportation Systems. SPE Prod. Eng. 1991, 6,
166–170. [CrossRef]
34. Barnea, D.; Taitel, Y. A model for slug length distribution in gas-liquid slug flow. Int. J. Multiph. Flow 1993, 19, 829–838. [CrossRef]
35. Al-Safran, E.M.; Sarica, C.; Zhang, H.Q.; Brill, J.P. Probabilistic/mechanistic modeling of slug-length distribution in a horizontal
pipeline. SPE Prod. Facil. 2005, 20, 160–172. [CrossRef]
36. Fan, Y.; Aljasser, M. Unified modeling of pseudo-slug and churn flows for liquid holdup and pressure gradient predictions in
pipelines and wellbores. SPE Prod. Oper. 2022, 38, 350–366. [CrossRef]
37. Abdulkadir, M.; Hernandez-Perez, V.; Lowndes, I.S.; Azzopardi, B.J.; Sam-Mbomah, E. Experimental study of the hydrodynamic
behaviour of slug flow in a horizontal pipe. Chem. Eng. Sci. 2016, 156, 147–161. [CrossRef]
38. Al-Safran, E.; Kora, C.; Sarica, C. Prediction of slug liquid holdup in high viscosity liquid and gas two-phase flow in horizontal
pipes. J. Pet. Sci. Eng. 2015, 133, 566–575. [CrossRef]
39. Suleimanov, R.I.; Ya Khabibullin, M. Optimization of the design of the scrubber separator slug catcher. IOP Conf. Ser. Mater. Sci.
Eng. 2020, 905, 012090. [CrossRef]
40. Nnabuife, S.G.; Tandoh, H.; Whidborne, J.F. Slug flow control using topside measurements: A review. Chem. Eng. J. Adv. 2022,
9, 100204. [CrossRef]
41. Sun, J.; Jepson, W.P. Slug Flow Characteristics and Their Effect on Corrosion Rates in Horizontal Oil and Gas Pipelines; SPE: Kuala
Lumpur, Malaysia, 1992. [CrossRef]
42. Sarac, B.E.; Stephens, D.S.; Eisener, J.; Rosselló, J.M.; Mettin, R. Cavitation bubble dynamics and sonochemiluminescence activity
inside sonicated submerged flow tubes. Chem. Eng. Process.—Process Intensif. 2020, 150, 107872. [CrossRef]
43. Olbrich, M.; Bär, M.; Oberleithner, K.; Schmelter, S. Statistical characterization of horizontal slug flow using snapshot proper
orthogonal decomposition. Int. J. Multiph. Flow 2021, 134, 103453. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.