Professional Documents
Culture Documents
net/publication/312204825
CITATIONS READS
11 661
4 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Aliyu M Aliyu on 29 May 2018.
Slug frequency in high viscosity oil-gas two-phase flow: Experiment and MARK
prediction
⁎
Yahaya D. Babaa,b, , Archibong E. Archibonga, Aliyu M. Aliyua, Abdulhaqq I. Ameenc
a
Oil and Gas Engineering Centre, Cranfield University, Bedfordshire MK43 0AL, United Kingdom
b
Chemical/Petroleum Engineering Department, Afe Babalola University, PMB 5454, Nigeria
c
Chemical Engineering Department, Texas A & M University, P.O. Box 23874, Qatar
A R T I C L E I N F O A BS T RAC T
Keywords: The number of slug units that traverses a particular point at a given time within a defined pipe cross-section is
High viscosity oil known as slug frequency. The behaviour of this critical parameter for two-phase flow in high viscosity oils is
Gamma densitometer significantly different from those of conventional oils (of less than 1 Pa s). In this experimental investigation,
Slug flow new data on slugging frequency in high viscosity oil-gas flow are reported. Scaled experiments were carried out
Slug frequency
using a mixture of air and mineral oil in a 17 m long horizontal pipe of 0.0762 m ID. A high-speed Gamma
Densitometer of frequency of 250 Hz was used for data acquisition over a time interval of 30 s. For the range of
flow conditions investigated, increase in oil viscosity was observed to strongly influence the slug frequency.
Comparison of the present data with prediction models available in the literature revealed discrepancies. A new
correlation incorporating the effect of viscosity on slug frequency has been proposed for horizontal flow. The
proposed correlation will improve the prediction of slug frequency in high viscosity oils.
1. Introduction Brito et al., [6] investigated the effects of high oil viscosity on oil-
gas two flows behaviour, Kora et al., [27] proposed a correlation for
Unconventional oil resources constitute heavy oil, extra heavy oil, the prediction of slug liquid holdup for high viscosity oil in
oil sand, tar sands, oil shale and bitumen which account for a greater horizontal pipeline. Foletti et al., [12] carried out experimental
portion of the world remaining oil reserves as illustrated in Fig. 1. Their investigation using a viscous oil (μ o =0.896 Pa s) and air in a 0.022-
production in recent years have gained more attention considering m ID horizontal pipe. It was observed that the fluid properties have
their huge reserves. This is amidst increasing world energy demand a strong influence on the flow pattern maps. The measured pressure
and the decline in the reserves of conventional low viscous oils. drop from their study were compared with those from several
However, the physical nature (i.e. high viscosity) of heavy oil makes empirical and theoretical models and were found to be in poor
its production and transportation difficult. agreement. Khaledi et al., [26] also investigated medium oil
Intermittent flows (which include most viscous oil flows) have been viscosity effects on two-phase flow characteristics using synthetic
widely investigated by several researchers both experimentally and commercial oils (0.032 and 0.1 Pa s). The experiments were per-
theoretically due to their practical relevance in the oil and gas industry. formed in a straight 0.518-m long horizontal pipeline with 0.069-m
Most of these studies are focused low viscous liquids (i.e. water or oils ID. New data set base on their investigation were presented for high
with viscosities less than 0.1 Pa s). More recently Nadler and Mewes viscosity oil.
[31] conducted experiments to study the effects of liquid viscosity on Findings have showed that only a few of these publications address
the phase distribution of slug flow in a 0.059-m ID horizontal pipe. Air, liquid viscosities above 1.0 Pa s [4,51]. In addition to the effects of
water and oil (0.014–0.037 Pa s were used as the working fluids. liquid liquid viscosity, Archibong [4] also studied pipe inclination and scaling
holdup measurement was done using a multi-detector gamma densit- effects on slug frequency. He noted that both cases result in the
ometer. Result obtained showed that liquid holdup increased with decrease of slug flow region hence slug flow characteristics. Table 1
increasing liquid viscosity. below presents the summary of the experimental investigations focus-
Gokcal [14], Kora et al., [27] and Brito et al., [6] conducted air– ing on medium and slightly high viscous oil.
oil experiments, using a 0.0508-m ID horizontal pipe for oil Slug flow, schematically shown in Fig. 2, is one of the most
viscosity ranging from 0.1 – 0.587 Pa s. While Gokcal [14] and common and undesired flow regimes occurring in hydrocarbon
⁎
Corresponding author at: Oil and Gas Engineering Centre, Cranfield University, Bedfordshire MK43 0AL, United Kingdom.
http://dx.doi.org/10.1016/j.flowmeasinst.2017.01.002
Received 25 June 2016; Received in revised form 18 December 2016; Accepted 3 January 2017
Available online 10 January 2017
0955-5986/ © 2017 Elsevier Ltd. All rights reserved.
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
A Area m2 μ Viscosity cP
C Constants ρ Density kg/m3
D Pipe diameter m τ Shear stress Pa
Fr Froude number ε Relative error
fs Slug Frequency s−1
g Acceleration due to gravity m. s−2 Subscripts
L length m
Nμ Viscosity number f Film zone
HL Holdup g Gas phase
N f Inverse viscosity number l Liquid phase
Re Reynolds number m Mixture phase
St Strouhal number ns No slip
V Velocity m/s s Superficial
t Translational
production and transport. The intermittent nature of slug flow frequency has significant dependency on the liquid viscosity and
causes difficulties for phase separation and produces transient increases with increase in liquid viscosity. Most of these studies are
forces which are capable of damaging pre-process equipment. Hill however all limited to viscosity range less than 1.0 Pa s and were
et al., [22] noted that high slug frequency can cause corrosion in conducted in smaller diameter pipelines.
pipeline. The chaotic flow behaviour at the slug front can cause The purpose of the present work is to provide a new experimental
breakage of protective coatings resulting in exposure of the pipe dataset for high viscosity oil-gas two-phase flow for oil viscosity ranging
wall to corrosive elements already present in the multiphase flow from 1.0 – 5.5 Pa s. In addition, a new closure relationship for the
Sun and Jepson, [39]. Corrosion rates in pipeline are directly prediction of slug frequency is proposed. This will improve the
proportional to slug frequency thus making it a significant necessity predictive capacity of previous correlations.
for accurate measurement of slug frequency for the precise predic-
tion of corrosion rates to aid corrosion inhibitor programs. 1.1. Slug formation mechanism: initiation and dissipation
Slug frequency prediction models (i.e. empirical or mechanistic)
forms an integral part of any slug flow model. The knowledge of The formation of slugs can be classified into three main groups,
slug frequency most especially in high viscosity liquids is very vital namely; hydrodynamic slugging, terrain slugging and severe slug-
considering the fact that the behaviour of two-phase flow char- ging depending on the pipe orientation and geometrical configura-
acteristics in high viscosity oils differs significantly when compared tions. However, this investigation is focussed on hydrodynamic
to those of conventional oils. More so, they serve as input para- slugging. This slugging mechanism is commonly encountered in
meters for existing mechanistic models for the prediction of two- horizontal pipelines as a result of the flow disturbances on the gas–
phase flow characteristics such as liquid holdup and pressure liquid interface. The experimental set-ups for hydrodynamic slug-
gradient in pipes. ging investigation by most researchers [19,28,45] are designed to
A considerable number of correlations has been developed from favour the formation of stratified flow; a flow pattern characterized
different data sources for the prediction of slug frequency, extend- by the parallel flow of gas and liquid in the pipeline with the less
ing from simple correlations like those of ([17,18,21] and [43]) to dense phase (i.e. gas) flowing at the top while the denser phase (i.e.
more complex ones such as [23,24,30,38]. Slug frequency, as oil) flow at the bottom owing to gravity effects. A detailed under-
estimated by these correlations presented in Table 2 is expressed standing of this mechanism has been developed from an extensive
by a limited number of flow parameters such as the flowing liquid visualization study conducted by [10,41] in a 0.0375 m ID hor-
fraction and Froude number. However, recent investigations have izontal pipe of 19.8 m long.
shown that the length and diameter of pipe, densities and flow rates In the description depicted in Fig. 3, they noted that beyond the
gas and liquid and most importantly the viscosities of the liquid entrance (mixing) region, there exist small perturbations on the
phase considerably influence slug flow characteristics. The effects interface which transforms into growing waves due to suction
of liquid viscosity on slug frequency have also been studied by effects caused by increasing superficial gas velocities. A point is
[11,16,2–4,7]. All these authors unanimously concluded that slug reached when the growing waves momentarily bridge the top of the
pipe thereby resulting in the build-up of the upstream pressure and
eventual blockage of the pipe cross-section as can be seen depicted
in Fig. 3c. With further increase in the gas phase, the liquid is then
accelerated by the gas velocity. Once this happens, the fluid
blockage appears to be accelerated uniformly across the cross
section and consequently acting as a scoop (i.e.) picking up all
the slow moving liquid in the film ahead of it. By this process, the
fast moving liquid builds up in volume resulting in the formation of
slugs (See Fig. 3d). As the formed slug moves downstream of the
pipe, shedding of liquid occurs uniformly from its back and forms a
film with a free surface. The liquid in the film decelerates rapidly
from the slug velocity to a much lower velocity as regulated by the
interfacial and shear. Further movement down the pipe results in
Fig. 1. Total world oil reserves by percentage [1]. sweeping of all excess liquid which had entered the pipe since the
110
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
formation of the last slug. From that point on, it picks up liquid film
Models/Correl-ations Proposed which has been shed from the preceding slug. The length of the slug
Slug Frequency
Slug Frequency
Slug Frequency
Slug frequency
2. Experimental setup
length
NA
NA
NA
NA
Fig. 4 below shows the schematic of the experimental setup
Pressure Gradient Slug Frequency Slug Holdup
phases is positioned at the end of the pipe. The mixture of the two
Slug Liquid Holdup
fluids (i.e. oil and gas) upon injection is achieved at the T-junction
Pressure Gradient
Slug Frequency
Slug Frequency
Flow pattern
Plexiglas
Acrylic
Acrylic
Acrylic
NA
NA
NA
NA
–
–
The oil is usually recirculated to the oil tank via a by-pass aimed
towards achieving a uniform oil viscosity just before been injected
Inclination
into the main test line (Fig. 5). The oil temperature is regulated
(Degree)
0, 90
0
0
0
0
0.0508
0.0508
0.0228
0.0508
0.0508
0.0508
0.0508
0.022
0.051
0.074
Pipe
Air
Air
Air
Air
Air
Air
Air
Air
air
Air used as the gas phase was supplied from a screw engineering
compressor with a maximum supply capacity of 400 m 3 /h free air
Density (kg/
886
890
916
–
–
–
–
–
–
–
Viscosity (cP)
is injected into the mainline via a 2-in. steel pipe about 150 pipe
1000- 7500
1000, 3500
75, 150
1–590
587
900
896
Oil
Oil
Oil
Oil
Oil
Oil
Oil
Oil
Oil
Oil
Oil
oil
Wang, 2012
Author (s)
source emitting gamma rays at 662 Kev high energy level (hard
Table 1
[27]
[37]
[51]
spectrum) and the soft spectrum, lower energy level with range of
100 ~ 300 Kev. Caesium 137 acquired via a sodium iodide
111
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
scintillator was used in the study. The energy source is attenuated in comparison to Electrical Capacitance Tomography and wire
through a steel wall in the measurement section. A proprietary DAS mesh sensor. Its applicability to oil-gas flow is an advantage
is used for voltage signal acquisition produced by the detector, ICP considering the limitation inherent in the use of the wire mesh
i-7188 programmable logic controller which is used to convert the sensor.
raw voltage to gamma counts signals (counts are the remainder of
1
the attenuation signals after absorption by the media it passes Pth= [max(Pa )+min (Pa )]
2 (2)
through).
The Beer-Lambert equation represented by Eq. (1) is used for
linear attenuation coefficients computation and hence, the liquid 2.3. Test matrix
hold up. For an empty pipe, the gamma radiation beam's intensity
remains unchanged inside the pipe because of the non-existence of The test fluid used for this investigation with generic name
an attenuating media, however; some of the incident beam is CYL680 is manufactured by Total Limited, UK with physical
attenuated at the entrance and exit of the beams due to the pipe properties at 25 °C given as 917 kg/m3 and 1.83 Pa s for density
walls. and viscosity respectively. The oil's minimum and maximum
⎡ ⎛I ⎞⎤ viscosity were given as 0.333 and 15.33 Pa s at temperatures of
⎢ ln ⎜ IM ⎟ ⎥ 50 °C and −2.5 °C respectively. A summary of the test fluid proper-
⎢ ⎝ A⎠⎥
λL =⎢ ties and the matrix used for experimental investigation is presented
⎛ ⎞⎥
⎢ ln ⎜ IL ⎟ ⎥ in Table 3. The uncertainties in the superficial gas and liquid
⎣ ⎝ IA ⎠ ⎦ (1) velocities as presented in Table 4 were obtained based on manu-
facturers’ specification of flow meters, viscometer, and gamma
Where
sensor. These agreed with values obtained upon repeatability tests
IM = average gamma count obtained from liquid-gas mixture in the conducted to ascertain accuracy of values.
pipeline
IA = average calibration gamma data obtained for empty pipe 3. Results
(i.e.100% Air)
IL = average calibration gamma data obtained for pipe containing 3.1. Pipe inlet effects on slug frequencies
pure liquid
λl = Liquid Hold Up In order to study the effects of pipe inlet on the slug flow
development of, the inlet pipe was modified by using a stratified plate
A typical plot from of the Gamma Densitometer liquid holdup (4 m high) from the ground. Three different inlet plates as depicted in
time series exhibits an intermittent behaviour for slug flow as Fig. 8 with different split fractions (see Table 5) for oil and gas were
presented in Fig. 7 below. This is characterized by crests and used to achieve different flow area for both phases. The frequencies of
troughs. The crests are indicative of the passage of slug liquid while slug flow which forms just after the mixing region were observed at 4.0-
the troughs represent the slug film region. slug frequency is m and 14-m upstream and downstream respectively away from the
obtained by counting the quantity of instantaneous slug liquid injection point. The effects of the various pipe inlet modifications on
body passing over a period of time, this is validated by counting the flow behaviours were compared with that obtained using the normal
number of slug body from the recordings using high definition inlet (inlet without the stratified plate) and the results are presented
video camera recordings. So as to differentiate travelling waves below.
from slug liquid body most researchers are known to have adopted Fig. 9a, b and c below shows the slug frequency obtained for
specific values representing a threshold for slug count. For this three different pipe inlets and normal inlet for both the upstream
study, the technique used by [4,51] was adopted. This technique and downstream location. The result shows relatively higher values
defines the threshold for liquid holdup in the slug body and slug of slug frequency for all the three stratified pipe inlets used when
film region Pth by the expression (Eq. (2)). Where Pa is the compared to the normal inlet in the liquid dominated region
instantaneous liquid volume fraction obtained from gamma den- corresponding to low gas superficial velocities thus highlighting
sitometer. The choice of gamma densitometer over other two-phase the effects of effects of pipe inlet on slug flow development.
flow monitoring instrument is dependent upon it fast sampling rate Stratified pipe inlets as depicted in Fig. 8 above enhances the
112
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Table 2
Summary of existing correlations in the literature for slug frequency.
[43] – fS =0.61 −
ρG UG
ρL (D − hL )
Where UG is the actual gas velocity; hL = is the equilibrium liquid depth for stratified flow.
[32] Air/water, VSL 0.6–3.5 m/s VSG = 0.5 to 20 m/s. (VSL +1.5) 2
fS =0.088
gD
[23] Air/water fS D
=0.275 × 10(2.68Hle) whereHle is stratified equilibrimHL
ID = 0.05 m VM
fS D H
=2.74 le
(vg − vl ) 1 − Hle
[24] – ⎛ fS D ⎞*
⎜ ⎟ =−24.729 + 0.00766e9.91209HeLe
* +24.721e 0.20524HeLe
*
⎝ vm ⎠
⎛ f D ⎞* f D ⎛ ⎞
* = HLe ⎜1 − 0.068 ⎟
Where ⎜ S ⎟ = S (1 − vsg ) D 0.3 ; HLe
⎝ vm ⎠ vm ⎝ VSL ⎠
SL SG
(continued on next page)
113
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Table 2 (continued)
[4] Air/oil, μ = 0.1–5.5 Pa. s, ID= 0.025 and 0.0762 m, ln fs=ψ1 ln where
VSG = 0.3 to 9.0 m/s and VSL =0.07 −3.0 m/s, θ =0-30 °C ⎡ ⎤
Vsg ⎢ [(ρl − ρg ) / ρg ] ⎥
Vm ⎣ ⎦ Resl ∙ Vsl 2 Vsl
Fr = ; Vsgd = ; MR= ; λ= ;
gD gD Resg ∙ Vsg2 VM
And ψ1,ψ2 , ψ3 and ψ4 were obtained from his dataset as 0.138, 0.801, 1.661 and 0.277
respectively.
[42] Air/water, , ID= 0.025 m ⎡ L ⎤−0.14
fS =1.52*10−8 ⎢ ⎥ (ReSL )1.46 (ReSG )0.28
⎣D⎦
[3] Air/oil, μ = 0.1–0.587 Pa. s, ID= 0.0508 m fS =exp [1.51 − 17.04D + 0.77Ln (VSL )−0.181Ln (S )]
Vg ⎛ μ ⎞ β ⎛ Vsg ⎞γ
where S = =1 + α⎜ L⎟ ⎜ ⎟
VL ⎝ μ0 ⎠ ⎝ VsL ⎠
Where μ0 = reference viscosity 1 Pa s, α = 1.95; β =0.01; γ = 1.5
downstream of the pipe for all the pipe inlets investigated shows a
slight difference as illustrated in Fig. 9(d, e & f) suggesting that the
inlet effects are eliminated as the flow progresses downstream.
Nevertheless, at low superficial gas velocity, slug frequency was
found to be slightly lower for the Tee-inlet when compared to those
(a) Wave just passes out of view to the right of stratified inlets. This could be as a result of reduced increased
viscous effect which dampens the initial superficial velocity. For the
Tee-junction, the volume of oil entering into the pipeline is full in
contrast to the stratified plates which are reduced.
From Fig. 10 below, it can be seen that at most superficial liquid
velocities, the slug frequency is always lower downstream. This is
(b) Level rebuilds and wave nearly bridges the pipe
attributable to slug evolution and development as described by
[10]. They noted that at the upstream of the pipe, the unstable slug
with a high frequency tend to be evolved and merged with each
other as the slug body travels further down the pipeline thus
resulting in the formation of a relatively stable slug with a low
(c) Bridging of pipe by liquid, Slug formation frequency at downstream. The discrepancy is magnified as liquid
velocity increases and this manifests more at gas superficial
velocities between 0.3 and 1.3 m/s. However, at high gas velocities,
it appears the inlet effects are nullified. This is because the flow
quickly develops within the pipe inlet plates. Fig. 10 below also
shows the error bars for the measured slug frequency. These are
(d) Slug sweeps up liquid, level drops; Slug given as 5% of the measured value corresponding to the estimated
error by the gamma densitometer in determining slug units. To
determine this error level, comparisons were made with manually
counting slug units using recorded high speed videos of the flow.
For every 60 slug units counted a difference of between 2 to 3 slug
3
(e) Slug passes out of view level drops, level drops units were observed giving a maximum of 60 ×100% or 5% error in
slug frequency.
Fig. 3. Slug Formation Process [41].
formation of slug flow and thence slug frequency when gas and oil Fig. 11 shows an examination of the effects of liquid viscosities
are injected by a T-junction. Albeit, there is no significant differ- on the measured slug frequency for the given set of flow conditions
ence in the obtained slug frequencies between the three pipe inlets investigated. Within the investigated test matrix, slug frequency
possibly attributed the relative velocity of oil and gas phase. Also, increases with an increase in oil viscosity. This is because an
beyond gas superficial velocities of 1.5 m/s, there is a slight increasing the oil viscosity results in an increase in the liquid
difference in the measured slug frequency for all the different inlets height since there is increasing resistance to flow. However, it is
and this is because the gas phase having swept most of the liquid, noted that at V sg of about 2.6 m/s, the slug frequency of viscosity
becomes highly turbulent and thus dominates the process of mixing range 4.3–4.8 Pa s overlaps with that of 3.3–3.8 Pa s, we may
of gas and oil flow. Similarly, the slug frequency obtained at the consider this behaviour to be out of trend within the entire dataset.
An explanation for this may be the viscosity of liquid at this test
114
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
115
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Fig. 8. (a) Pipe inlet plates (b) cross-sectional view when inserted in test pipe.
Table 5
Oil and gas fractions based on pipe inlet plates.
Table 3
Inlet Oil split Gas split
Experimental test matrix and fluid properties.
1 1/2 D 1/2 D
S/N Test fluids Density Density Viscosity Test API
2 3/4 D 1/4 D
(kg/m3) (kg/m3) (Pa.s) matrix gravity
3 1/4 D 3/4 D
(m/s)
•
fs Ls
I= Froude number to account for the inertia forces
C (2)
Vd ρL
Several researchers such as [29,33,47,48,9] have shown from Frm=
(gD )0.5 (ρL − ρg ) (6)
experimental investigation that the intermittency can roughly be
estimated by
C =1. 2(Vsl +Vsg )≈VT (3) • And the viscosity number to account for viscous forces
Vd μL
From Eq. (2) above, C can be substituted as translational velocity VT Nμ=
gD 2 (ρL − ρg ) (7)
and can be correlated as a function of velocity ratio thereby yielding
fs Ls Vsl Dividing Eq. (6) by Eq. (7), we obtain a dimensionless inverse
I= ∝ viscosity number
VT Vsl + Vsg (4)
Vd ρL
From the relationship above in Eq. (3), a plot of intermittency Frm (gD)0.5 (ρL − ρg )
Nf = =
against velocity ratio for different viscosities at a constant liquid Nμ Vd μL
2
velocity of 0.06 m/s as shown in Fig. 13 indicate and approximate a gD (ρL − ρg ) (8)
116
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Fig. 9. Slug frequency for different inlets at upstream & downstream position of pipe.
117
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Fig. 11. Slug frequency vs. gas superficial velocity for different oil viscosities.
Fig. 12. Comparison of slug frequency data for this study with [15,36] data.
Fig. 10. Comparison of Slug frequencies for 1/2 gas and oil inlet at upstream and
downstream positions.
118
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Fig. 14. Dimensionless Slug Frequency vs. Liquid Ratio at Different Viscosities for Vsl Fig. 15. Dimensionless slug frequency vs. liquid ratio and inverse viscosity number for
=0.06 m/s. Vsl=0.2 m/s.
strongly depends on the superficial fluid velocities. We hence define noting that based on previous investigations and the literature
a velocity ratio being the ratio of the liquid superficial velocity VSL to review above, it can be said that there is no one correlation that can
that of the mixture velocity Vm=VSG+VSL . Therefore, we can correlate be used for all conditions. Therefore, the validity of tested correla-
the dimensionless slug frequency as a function of the inverse tions is limited to the individual range of experimental conditions
viscosity number and the velocity ratio, i.e. from which they were developed as was earlier highlighted on
Table 2 above.
⎧⎛ 3 ⎞ ⎫ ⎛ 3 ⎞b
fs D ⎪ ⎜ D 2 ρL (ρL − ρg ) g ⎟ ⎛ VSL ⎞ ⎪ ⎜ D 2 ρL (ρL − ρg ) g ⎟ ⎛ VSL ⎞c The correlations whose performance were evaluated include;
=f ⎨ ⎜ ⎟⎟ , ⎜ ⎟ ⎬=A ⎜ ⎟⎟ ⎜⎝ V ⎟⎠ [16–18,2,21,32,34,35,49,51] with their performances (mostly un-
VM ⎪⎜ μL ⎝ Vm ⎠ ⎪ ⎜ μL m
⎩⎝ ⎠ ⎭ ⎝ ⎠ der predictions) exhibiting different magnitudes of prediction. The
(10) prediction of these correlations can be divided into three categories
based on their prediction performance and the range of liquid
A plot of the dimensionless frequency against a combination of
viscosity datasets used. The first set is that proposed by
dimensionless inverse viscosity number and velocity ratio are
([17,18,21,32] and [49]a). This category was developed based on
presented in Fig. 15 for the same velocity of liquid. From the plots
experimental data from low viscosity ( > 0.01 Pa s) liquids and the
it can be seen that a simple linear relationship exists between
correlations exhibit similar prediction trends in that they signifi-
dimensionless slug frequency and the right hand side of Eq. (10),
cantly under-estimate slug frequency. The prediction of those
and the plots are seen to be clearly segregated by viscosities. Linear
proposed by ([2] and [16]) with R 2 values of 0.55 and 0.66
regression was used to obtain the coefficient A, the indices b, and c.
respectively which fall under the second category. They are fairly
The coefficient of variation (R2 ) obtained from the plot of dimen-
accurate when compared to the first set as can be seen from Fig. 14.
sionless groups for all the viscosities investigated were found to be
The reason for their slight improvement in prediction over the first
very close to unity as illustrated in Fig. 15. The coefficient's value
set can be attributed to the fact that the liquid viscosities investi-
usually lies between 0.0 and 1.0. An R 2 value of 0 is an indication
gated in these studies were higher than those of the first set (i.e.
that there is no relationship between the dimensionless groups
between 0.1 and 0.6 Pa s). As for the correlations developed by
while a unit value indicates strong dependence between ordinate
[34,35,51], despite using the same facility as the data source, their
and abscissa.
poor performance can be attributed to the fact that [34,35] had a
The final correlation proposed for the estimation of slug
limited data base, and instrumentation limitation (i.e. low sam-
frequency is a modified form of that proposed by [16]. The
pling frequency) in the case of [51]. In all, the proposed correlation
modification is based on the present data for this study and those
for slug frequency in this investigation out performed all the tested
obtained by [16] for medium oil viscosity. After linear regression
prediction correlations evaluated as presented in Fig. 16 and
analysis, the values of A, b, and c were determined to be 1.81012,
Table 6. The definitions of the statistical parameters ε 1 –ε 6 are
−0.4687, and 1 respectively. Therefore, the new correlation for slug
given in the appendix.
frequency is given below
Finally, the error margin between the measured slug frequencies result
1 VSL and model prediction becomes more significant with increase in Reynolds
fs =1. 8102
Nf −0.4687 D (11) number thus highlighting the sensitivity of the proposed model to change in
liquid viscosity. Nevertheless, based on the use of higher viscosity oil in the
experiments, and the incorporation of published data from [15], the new
4.2. Evaluation and comparison of slug frequency prediction correlation has a more robust experimental database. Hence, it is concluded
correlations that the proposed correlation gives better predictions, which the statistical
analysis revealed that it gave the least deviations among the correlations
Measured slug frequency data for high viscosity oil obtained in tested. Also the proposed correlation performed best with each of the errors
this investigation have been compared with a number of well- being the least.
known empirical correlations earlier summarized above. It is worth
119
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Fig. 16. Comparison of prediction correlations for slug frequency vs measured slug frequency.
5. Conclusion this investigation shows that slug frequency has strong dependency on
liquid viscosity with an increasing trend as viscosity and gas superficial
The literature is short of adequate theoretical and experimental velocities increases. Performance evaluation of most existing prediction
studies relating to the effects of high viscosity oil on slug frequency. A models against present data revealed wide discrepancies attributed to
study on the effects of liquid viscosities on slug frequency has been unaccounted liquid viscosity effects in the models. The proposed
conducted for oil viscosities ranging from 1.0–5.5 Pa s in a 0.0762 m correlation exhibits a better prediction of the dataset and as such, will
ID horizontal pipe using an advanced instrumentation (Gamma serve as a better alternative for the prediction of slug frequency than
Densitometer) with a sampling frequency of 250 Hz. The result from many of those found in the literature.
120
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
Table 6
Performance evaluation of proposed model against existing prediction Correlations.
Gregory 1969 Henwood & Richardson 1971 [18] [49] [32] [2] [16] [34,35] [51] Proposed Correlation
ε1 −57.46 −18.83 −54.24 −56.66 −57.24 −40.39 −30.73 1122.29 −22.32 −0.91
ε2 57.46 33.88 54.24 56.66 57.24 40.39 30.73 1122.29 24.35 7.40
ε3 32.23 48.65 30.64 31.81 32.11 25.16 18.38 847.81 21.33 14.45
ε4 −0.44 −0.23 −0.42 −0.44 −0.44 −0.28 −0.24 10.63 −0.21 −0.01
ε5 0.44 0.26 0.42 0.44 0.44 0.28 0.24 10.63 0.22 0.05
ε6 0.52 0.41 0.50 0.52 0.52 0.30 0.27 16.73 0.31 0.09
Appendix
Statistical Parameters
Six statistical parameters were used to evaluate the performance of predictive correlations relative to the experimental data acquired. These
parameters were also used by several researchers such as [16,2,27,50] and are evaluated based on two types of errors; actual and relative error
defined in Eqs. (13) and (14) respectively. Results are given in Table 6 and the best performing correlations are those with the least magnitude of the
statistical parameter concerned. They are:
ypredicted − ymeasured
εi= *100
ymeasured (13)
121
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
122
Y.D. Baba et al. Flow Measurement and Instrumentation 54 (2017) 109–123
[28] E.S. Kordyban, T. Ranov, Mechanism of Slug Formation in Horizontal Two-Phase in horizontal oil and gas pipelines. in: Proceedings of the SPE Annual Technical
Flow, J. Basic Eng. 92 (1970) 857–864. Conference and Exhibition. SPE, Washington DC, pp. 215–228, 1992.
[29] G.E. Kouba, W.P. Jepson, The flow of Slugs in horizontal, two-phase pipelines, J. [40] Y. Taitel, A.E. Dukler, A model for predicting flow regime transitions in horizontal
Energy Resour. Technol. 112 (1990) 20–24. and near horizontal gas-liquid flow, AIChE J. 22 (1976) 47–55.
[30] G. Manolis, M. Mendes-Tatsis, G.F. Hewitt, The Effect of Pressure on Slug [41] Y. Taitel, A.E. Dukler, A model for slug frequency during gas-liquid flow in
Frequency in Two-Phase Horizontal Flow. in: Proceedings of the 2nd International horizontal and near horizontal pipes, Int. J. Multiph. Flow. 3 (1977) 585–596.
Conference on Multiphase Flow. Elsevier, Kyoto, Japan, pp. 347–354, 1995. [42] J. Thaker, J. Banerjee, Characterization of two-phase slug flow sub-regimes using
[31] M. Nadler, D. Mewes, Effects of The liquid viscosity on The phase distributions in flow visualization, J. Pet. Sci. Eng. 135 (2015) 561–576.
horizontal gas-liquid slug flow. Int. J, Multiph. Flow. 21 (1995) 253–266. [43] E. Tronconi, Prediction of slug frequency in horizontal two phase slug flow, AIChE
[32] O.J. Nydal, An Experimental Investigation on Slug Flow, University of Oslo, 1991. J. 36 (1990) 701–709.
[33] O.J. Nydal, S. Pintus, P. Andreussi, Statistical characterization of slug flow in [44] G.B. Wallis, One-dimensional Two-phase Flow, McGraw-Hill Book Comp.
horizontal pipes, Int. J. Multiph. Flow. 18 (1992) 439–453. American Institute of Chemical Engineers, Newyork, New York, 1969.
[34] C. Okezue, Application of the gamma radiation method in analysing the effect of [45] G.B. Wallis, J.E. Dodson, The onset of slugging in horizontal stratified air-water
liquid viscosity and flow variables on slug frequency in high viscosity oil-gas flow, Int. J. Multiph. Flow. 1 (1973) 173–193.
horizontal flow, WIT Trans. Eng. Sci. 79 (2013) 447–461. [46] X. Wang, L. Guo, X. Zhang, An experimental study of the statistical parameters of
[35] C. Okezue, Application of the gamma radiation method in analysing the effect of gas–liquid two-phase slug flow in horizontal pipeline, Int. J. Heat. Mass Transf. 50
liquid viscosity and flow variables on slug frequency in high viscosity oil-gas (2007) 2439–2443.
horizontal flow. in: C.A. Brebbia, Vorobief, P. (Eds.), in: Proceedings of the 8th [47] B.D. Woods, Z. Fan, T.J. Hanratty, Frequency and development of slugs in a
International Conference on Computational and Experimental Methods in horizontal pipe at large liquid flows, Int. J. Multiph. Flow. 32 (2006) 902–925.
Multiphase and Complex Flow. València, Spain, pp. 447–461, 2013b. [48] B.D. Woods, T.J. Hanratty, Relation of slug stability to shedding rate, Int. J.
[36] J. Pan, Gas Entrainment in Two-Phase Gas-Liquid Slug Flow (Ph.D. thesis). Multiph. Flow. 22 (1996) 809–828.
Imperial College London, 2010. [49] G.J. Zabaras, Prediction of Slug Frequency for Gas-Liquid Flows. in: Proceedings of
[37] R. Schulkes, Slug Frequencies Revisited. in: Proceedings of the 15th International the SPE Annual Technical Conference and Exhibition. Society of Petroleum
Conference on Multiphase Production Technology. BHR Group, Cannes, France, Engineers, Houston, Texas, 1999.
2011. [50] Y. Zhao, High Viscosity Liquid Two-phase Flow (Ph.D. thesis). Cranfield University,
[38] R.H. Shea, H. Eidsmoen, M. Nordsveen, J. Rasmussen, Z.G. Xu, J.O. Nossen, Slug United Kingdom, United Kingdom, 2014.
frequency prediction method comparison. in: Proceedings of the 4th North [51] Y. Zhao, H. Yeung, L. Lao, Slug frequency in high viscosity liquid and gas flow in
American Conference on Multiphase Technology. pp. 227–237, 2004. horizontal pipes. in: Proceedings of the 16th International Conference on
[39] J.-Y. Sun, W.P. Jepson, Slug flow characteristics and their effect on corrosion rates Multiphase Production Technology. BHR Group, Cannes, France, 2013.
123