Beruflich Dokumente
Kultur Dokumente
ÇİĞDEM ÖMÜRLÜ
MAY 2006
MATHEMATICAL MODELING OF HORIZONTAL TWO-PHASE FLOW
THROUGH FULLY ECCENTRIC ANNULI
A THESIS SUBMITTED TO
THE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCES
OF
MIDDLE EAST TECHNICAL UNIVERSITY
BY
ÇİĞDEM ÖMÜRLÜ
MAY 2006
Approval of the Graduate School of Natural and Applied Sciences
I certify that this thesis satisfies all the requirements as a thesis for the degree
of Master of Science.
This is to certify that we have read this thesis and that in our opinion it is fully
adequate, in scope and quality, as a thesis for the degree of Master of Science
Signature:
iii
ABSTRACT
Ömürlü, Çiğdem
M.Sc., Department of Petroleum and Natural Gas Engineering
Supervisor: Assist. Prof. Dr. M. Evren Özbayoğlu
iv
flow patterns were identified visually. Meanwhile a mechanistic model has been
developed. The flow pattern identification criteria proposed originally for two-
phase flow through pipes by Taitel and Dukler1 has been inherited and modified
for the eccentric annular geometry. The complex geometry of eccentric annuli
has been represented by a new single diameter definition, namely
representative diameter dr. The representative diameter has been used while
calculating the pressure losses. A computer code based on the algorithm of the
proposed mechanistic model has been developed in Matlab 7.0.4. Both the flow
pattern prediction and the frictional pressure loss estimation are compared with
the gathered experimental data. Moreover, friction factor correlations have
been developed for each flow pattern using experimental data and statistical
methods. The performance of the proposed model and the friction factor
correlations has been evaluated from experimental data. The mechanistic
model developed in this study accurately predicts flow pattern transitions and
frictional pressure losses. The model’s pressure loss estimations are within
± 30% for two different annular flow geometries.
v
ÖZ
Ömürlü, Çiğdem
Y.Lisans, Petrol ve Doğal Gaz Mühendisliği Bölümü
Tez Yöneticisi: Y.Doç. Dr. M. Evren Özbayoğlu
Bu çalışmadaki başlıca hedef, iki fazlı akışın yatay halkasal ortamlardaki akış
mekanizmasını, hidroliğini ve karakteristiğini anlamaktır. Bu amaca ulaşırken
yoğun bir şekilde hem teorik hem de deneysel çalışmalar yapılmıştır. ODTÜ-
PETE-KTÇFAL (Orta Doğu Teknik Üniversitesi Petrol ve Doğal Gaz Mühendisliği
Bölümü Kesinti Taşıma ve Çok Fazlı Akış Laboratuvarı) iki fazlı akış deney
düzeneği 0.1143m iç çaplı akrilik muhafaza borusu - 0.0571m dış çaplı sondaj
borusu ve 0.0932m iç çaplı akrilik muhafaza borusu - 0.0488m dış çaplı sondaj
borusu geometrik özellikteki yatay halkasal ortamdan oluşmaktadır ve uzunluğu
4.84m’dir. Deneyler sırasında çeşitli gaz ve sıvı akış debilerinde oluşan basınç
kayıpları dijital ve analog olarak kaydedilmiştir. Akış biçimleri ise görsel olarak
tespit edilmiştir. Bir yandan da mekanistik model oluşturulmuştur. Borulardaki
vi
iki fazlı akış için Taitel ve Dukler1 tarafından geliştirilen akış biçimi tayini
kriterleri tercih edilmiş ve tam eksentrik halkasal ortama uyarlanmıştır. Eksentrik
halkasal ortamın karmaşık geometrisi tek bir yeni çap terimi, temsili çap dr ile
temsil edilmiştir. Temsili çap, basınç kayıpları hesaplamaları sırasında da
kullanılmıştır. Matlab 7.0.4 programı kullanılarak oluşturulan mekanistik modele
dayalı bilgisayar kodu yazılmıştır. Hem akış biçimi tayinleri hem de sürtünme
kaynaklı basınç kayıpları hesaplamaları deneysel verilerle karşılaştırılmıştır.
Ayrıca, her bir akış biçimi için sürtünme faktörü bağıntıları istatistiksel yöntem
kullanılarak oluşturulmuştur. Oluşturulan modelin ve sürtünme faktörü
bağıntılarının performansı deneysel veriler kullanılarak değerlendirilmiştir. Bu
çalışmada oluşturulan mekanistik model akış biçimleri geçişlerini ve basınç
kayıplarını doğru bir şekilde tayin edebilmektedir. Basınç kayıpları ± 30% hata
sınırları arasındadır.
Anahtar Kelimeler: İki Fazlı Akış, Sürtünme Basınç Kayıpları, Sürtünme Faktörü,
Mekanistik Model, Eksentrik Halkasal Ortam, Çok Fazlı Akış Deneyi
vii
To My Family
viii
ACKNOWLEDGMENTS
Finally my special and sincere thanks go to my whole family for their endless
love and support, to my grandmother who stands always by me with her great
understanding and prayers, to my father and mother for their care and
tolerance, to Tolga Metin who is always beside me with joy and love along with
kindness and never-ending support, and help, to my aunt for great
encouragement and guidance, to my sister İpek Ömürlü and my cousin Berk
Gercek, and to my friends İlkay Uzun and Sevtaç Bülbül for their supports,
encouragements and considerations. I would like to thank to Tolga Metin once
more for drawing the excellent figures in this thesis.
ix
TABLE OF CONTENTS
ABSTRACT ..................................................................................................IV
ÖZ..............................................................................................................VI
ACKNOWLEDGMENTS ..................................................................................IX
NOMENCLATURE .......................................................................................... 1
CHAPTER .................................................................................................... 5
1.INTRODUCTION...................................................................................... 5
2.LITERATURE REVIEW .............................................................................. 8
x
5.1 Flow Pattern Prediction................................................................... 21
7.COMPUTER WORK................................................................................. 57
8.RESULTS and DISCUSSION.................................................................... 59
xi
8.1. Validation of Flow Pattern Identification of Proposed Model with
Experimental Data ............................................................................... 59
RECOMMENDATIONS.................................................................................. 83
REFERENCES.............................................................................................. 85
APPENDIX.................................................................................................. 90
xii
LIST OF FIGURES
xiii
Figure 8.1.4- Comparison of flow pattern maps generated using dhyd and OOM
for configuration 2...................................................................................... 63
Figure 8.1.5- Validation of flow pattern maps generated using dr and modified
Beggs and Brill3 method with experimental data for configuration 1 ............... 64
Figure 8.1.6- Validation of flow pattern maps generated using dr and modified
Beggs and Brill3 method with experimental data for configuration 2 ............... 65
Figure 8.2.1- Comparison of frictional pressure loss estimations of the proposed
model with experimental data and mostly used models for stratified flow
through configuration 1 .............................................................................. 66
Figure 8.2.2- Comparison of frictional pressure loss estimations of the proposed
model with experimental data and mostly used models for stratified flow
through configuration 2 .............................................................................. 67
Figure 8.2.3- Comparison of frictional pressure loss estimations of EPDM with
experimental data and mostly used models for intermittent flow through
configuration 1 ........................................................................................... 69
Figure 8.2.4- Comparison of frictional pressure loss estimations of EPDM with
experimental data and mostly used models for intermittent flow through
configuration 2 ........................................................................................... 69
Figure 8.2.5- Comparison of frictional pressure loss estimations of OM with
experimental data and mostly used models for intermittent flow through
configuration 1 ........................................................................................... 71
Figure 8.2.6- Comparison of frictional pressure loss estimations of OM with
experimental data and mostly used models for intermittent flow through
configuration 2 ........................................................................................... 71
Figure 8.3.1- Friction factor and mixture Reynolds number relation of
experimental stratified flow data ................................................................. 75
xiv
Figure 8.3.2- Friction factor and mixture Reynolds number relation of
experimental intermittent flow data ............................................................. 76
Figure 8.3.3- Friction factor and mixture Reynolds number relation of
experimental intermittent flow data for NRemixλ < 100000............................. 77
Figure 8.3.4- Friction factor and mixture Reynolds number relation of
experimental intermittent flow data for NRemixλ ≥ 100000 ............................ 78
Figure 8.3.5- Comparison of pressure losses determined by the empirical
correlations and experimental data for stratified flow.................................... 79
Figure 8.3.6- Comparison of pressure losses determined by the empirical
correlations and experimental data for intermittent flow ............................... 79
Figure A.1- The analysis of the forces during wave growth in the conduit ...... 98
Figure A.2- Stratified smooth flow through configuration 1.......................... 101
Figure A.3- Stratified wavy flow through configuration 1 ............................. 101
Figure A.4- Stratified wavy flow through configuration 1 ............................. 102
Figure A.5- Intermittent flow through configuration 1 ................................. 102
Figure A.6- Intermittent flow through configuration 1 ................................. 103
Figure A.7- Intermittent flow through configuration 1 at high liquid and gas flow
rates........................................................................................................ 103
Figure A.8- Intermittent flow through configuration 1 at high liquid and gas flow
rates........................................................................................................ 104
Figure A.9- Stratified smooth flow through configuration 2.......................... 104
Figure A.10- Stratified smooth flow through configuration 2 ........................ 105
Figure A.11- Stratified wavy flow through configuration 2 ........................... 105
Figure A.12- Stratified wavy flow through configuration 2 ........................... 106
Figure A.13- Intermittent flow through configuration 2 ............................... 106
Figure A.14- Intermittent flow through configuration 2 ............................... 107
xv
LIST OF TABLES
xvi
NOMENCLATURE
A area (L2)
AN Annular Flow
Bo Bond number
cg constant for gas phase
cl constant for liquid phase
DB Dispersed Bubble Flow
dC critical gas bubble diameter (L)
dCB critical bubble size below which bubbles can not migrate (L)
dCD critical bubble size above which the bubble is deformed (L)
dg hydraulic diameter for gas phase(L)
dhyd hydraulic diameter (L)
di outer diameter of inner pipe, drillpipe (L)
dl hydraulic diameter for liquid phase(L)
do inner diameter of outer pipe, casing (L)
dr representative diameter (L)
El liquid volume fraction in the slug unit
Els liquid volume fraction in the slug body
f friction factor
f friction factor
FE entrainment fraction
ff friction factor
g gravitational acceleration (L/T2)
g gravitational acceleration (L/T2)
1
GVF gas void fraction
Hl Liquid holdup
hl liquid level (L)
I Intermittent Flow
Lentrance length required from entrance for fully developed flow (L)
Lexit length required from exit for fully developed flow (L)
Lf length of gas zone (L)
Ls length of liquid slug body (L)
Lu length of a slug unit (L)
m constant for gas phase
n constant for liquid phase
NB dimensionless term
P pressure (M /(L T2))
Re Reynolds number
S contact perimeter (L)
s sheltering coefficient
SS Stratified Smooth Flow
SW Stratified Wavy Flow
v velocity (L/T)
z axial direction
dAl
dhl derivative of liquid area with respect to liquid level
N FrL
Froude number
Re∞ high Reynolds number
∆P
pressure gradient (M/T2L2)
∆L
2
Subscripts
c core
d drift
d∞ drift at high Reynolds numbers
f film
ff formation fracture
fgp frictional pocket/liquid film
fsL frictional slug
g gas pocket
g related with the gas phase
i related with the interface
l related with the liquid phase
lf liquid film
ls liquid slug
m mixture
mixλ mixture based on liquid holdup
ml mixture related with liquid phase properties
r representative
sg superficial gas
sl superficial Liquid
t translational
wg wall gas
wl wall liquid
3
Greek
δ dimensionless liquid film thickness
Τµ weighting factor
δ liquid film thickness (L)
φ angle
η weighting factor
λl no-slip holdup
µ viscosity (M /(L T))
θ angle
ρ density (M/L3)
σ interfacial surface tension (M/T2)
τ shear stress (M /(L T2))
4
CHAPTER 1
INTRODUCTION
Two-phase flow is the flow phenomenon of two different fluid phases flowing
simultaneously through a conduit. Generally, liquid and gas phases are the
components of this commonly encountered flow type. Since 1950’s, the flow
mechanism of two-phase fluids has been the subject of research in many
different engineering practices. In petroleum industry, the applications of two-
phase flow start from drilling and continue till the refining process.
5
diminishes the risks of contaminating the reservoir and eliminates the potential
pollution of drilling mud to environment2.
6
methods provide more accurate results with better understanding of two-phase
flow systems.
7
CHAPTER 2
LITERATURE REVIEW
8
determined from single-phase gas and liquid properties through the liquid
holdup.
The opposite approach was taken in the “Separated Flow Model” proposed by
Lockhart and Martineli5. In this model the gas phase and the liquid phase were
assumed to flow separately from each other. Thus, each of the phases was
analyzed utilizing single-phase flow methods such as friction factor concept.
Four flow mechanisms were established and transition criteria between these
flow mechanisms were suggested. Curves were presented for the prediction of
pressure drop and the liquid level in the pipe.
“The Drift Flux Model”, developed by Wallis3, treated the two phases as a
homogeneous mixture. However, it allowed slippage between the gas and the
liquid phases. This was a significant improvement of the homogeneous model.
However, additional information was required about the relative movement of
the two phases and, this information was not always available.
9
Correlations were the most common procedure to predict the flow properties in
annulus, in the past. The predictions either have applied correlations developed
for flow in pipes by use of the hydraulic diameter concept or have applied
correlations developed from experimental data. Baxendell7, Gaither et al.8,
Angel and Welchon9, and Winkler10 presented empirical correlations for annular
two-phase flow.
Beggs and Brill14 studied two-phase flow through pipes in an entire range of
inclination angles, i.e., vertical upward, horizontal and vertical downward. They
developed correlations for flow pattern determination by using Froude number
and no-slip holdup. The proposed model estimated the actual holdup and the
pressure loss for each flow pattern separately. Dukler et al11’s method was
referred to calculate the two-phase friction factor.
Taitel and Dukler1 analyzed the prediction of transition boundaries between the
flow patterns. The model begins with the equilibrium stratified flow assumption.
The equilibrium liquid holdup is determined by using the Lockhart and Martinelli
parameter. They modified the Kelvin-Helmholtz inviscid theory in order to
10
predict the initiation of slugs. The transition of intermittent to annular flow is
assumed to be dependent only on liquid level. Jeffrey’s theory for wave
initiation is used to determine the transition of stratified smooth to stratified
wavy flow pattern. Turbulent and buoyant forces acting on a gas pocket are
investigated for the boundary between dispersed bubble flow and intermittent
flow. The transition conditions were also expressed as dimensionless
parameters. This model, developed for Newtonian flow, was verified with the
experiments conducted in small diameter pipes under low-pressure conditions.
Dukler and Hubbard15 investigated the mechanisms and the hydraulic behavior
of slug flow. An idealized slug unit concept was introduced. Two main zones
that constituted a slug unit were defined as a liquid slug and stratified liquid
film/gas pocket. The proposed model estimated the liquid holdup, pressure loss
and velocity distributions within the slug unit.
Barnea16 studied the transition mechanisms for each individual boundary and
proposed a unified model. The flow chart of the model begins with the
dispersed bubble transition. The applicability of the developed mechanisms was
presented for the whole range of pipe inclinations. The results were compared
with the experimental data. The effects of flow rates, fluid properties, and pipe
size and inclination angle were incorporated in dimensionless maps.
11
i.e., Baker et al19 and Andritsos and Hanratty20. Also the effect of pipe
roughness was taken into consideration during friction factor calculations.
Uniform liquid level in the film zone was assumed for intermittent flow.
Empirical correlations were used in order to predict the slug length and liquid
holdup of the slug body. Annular flow was treated as stratified flow with
different geometrical configuration. Liquid entrainment fraction (Oliemans et
al21) was also considered while calculating the liquid holdup in the gas core.
12
Garcia et al25 studied a large amount of two-phase flow data and developed
composite analytical expressions for friction factor covering both laminar and
turbulent flow regimes. Two different approaches were presented. The first
method is the universal composite correlation for friction factor estimation
regardless of the flow pattern. The second method represents the friction factor
correlations for a given flow pattern. This Fanning friction factor definition was
based on the mixture velocity and density.
13
They concluded that for a small ratio of diameters of casing and tubing, the gas
void fraction in vertical annular conduit was similar to the gas void fraction in
circular pipes.
Caetano et al29 investigated the upward vertical flow of two-phase fluid systems
through an annulus. This experimental study covered the flow of water-air and
kerosene-air mixtures through concentric and eccentric annuli. The flow
patterns were identified visually. They proposed a mechanistic model for flow
pattern prediction by applying the Taitel and Dukler’s1 model to annular
concentric and eccentric geometries. Moreover, average liquid holdup and
pressure loss determination methods were developed. The effect of fluid
properties was observed as a result of the comparison of the developed flow
pattern maps for water-air and kerosene-air mixtures. Experiments were
conducted in a small scale experimental setup.
14
Sunthankar32 conducted experiments in a field scale experimental setup for
water–air mixture through horizontal annuli. He modified the unified model
developed by Xiao et al.17 to predict the flow patterns for horizontal and nearly
horizontal annular flow. Hydraulic diameter concept was used while modifying
the model. The effects of eccentricity and inner pipe rotation were also
investigated. A simulator was developed based on the work of Gomez et al22 for
pressure loss estimations. The performance of the proposed model was
evaluated with the experimental data. It was concluded that the intermittent
flow was different than that was defined for pipe flow, i.e., the Taylor bubble
was distorted and the liquid slug was highly aerated. The developed flow
pattern maps showed shifts when compared with the flow pattern transition
boundaries of pipe flow.
Zhou33 studied cuttings transport with aerated mud in horizontal annulus under
elevated pressures and temperatures. Taitel and Dukler’s1 model was modified
for annular two-phase flow. A mechanistic model was developed to predict the
volumetric cuttings concentration in the annuli and the critical pressure gradient
for preventing cuttings from deposition. Experiments were also conducted
during this study in order to verify the accuracy of developed model. The
predictions of developed mechanistic model were in agreement with measured
data. It was concluded that liquid flow rate, gas liquid ratio and temperature
essentially affected the cuttings transport efficiency. Comparisons between
predictions and measurements for aerated mud flow showed an average error
of 12.2%.
15
Rodriguez3 carried out an experimental study in order to find the minimum air
and water flow rates that effectively transport cuttings through highly inclined
and horizontal wells. The experiments were carried out in a low pressure field
scale flow loop. The model proposed in the study of Sunthankar32 was inherited
for flow pattern identification and pressure loss determination. The model’s
results were compared with experimental data. It was concluded that the flow
patterns of cuttings are dependent on the total flow rate of the liquid and gas
phase. It was also concluded that in order to avoid the formation of a stationary
cuttings bed, an approximate boundary of minimum flow rate of each phase
can be determined. The minimum requirements for gas and liquid flow rates
were found to be always in the intermittent flow regime.
16
Flow pattern data and pressure drop measurements were compared with the
mechanistic models predictions. The results showed good agreement even
though the number of data points did not permit the development of a
complete and precise flow pattern map. The model performance was also
compared with Beggs and Brill14 correlation and modified Aziz et al36 method. It
was concluded that proposed model had better performance.
17
CHAPTER 3
18
to comprehend the flow behavior of two-phase flow through horizontal annular
conduits using techniques different than hydraulic diameter.
19
CHAPTER 4
20
CHAPTER 5
THEORY
In this chapter, the flow pattern determination and the pressure loss estimation
methods are presented in details. New approaches are developed hence
yielding in a mechanistic model for both flow pattern identification and frictional
pressure loss estimation of horizontal two-phase flow through eccentric annuli.
Accurate mapping of the flow patterns is the first step for determination of the
frictional pressure losses correctly. Major concern in mechanistic modeling is
the determination of flow patterns accurately. As discussed in previous section
most of the studies carried out inherited the flow pattern transition definitions
proposed by Taitel and Dukler1 and Barnea16. Since the flow area is fully
eccentric annuli; hydraulic diameter (Equation 1) approach yields to significant
errors during flow pattern determination.
d hyd = do − di (1)
21
dr = (d 2
o − di2 )
(2)
∆P 2cl ρl v sl2
= (3)
∆L sl d r Rensl
∆P 2cg ρ g v sg2
= (4)
∆L sg d r Remsg
gas phases less than 2100. For greater values of superficial Reynolds numbers
the constants cl = cg = 0.046 and m = n = 0.2 . The superficial liquid and gas
d r vsl ρl
Re sl = (5)
µl
d r vsg ρ g
Re sg = (6)
µg
22
Then, after the determination of Hl from chart presented by Lockhart and
Martinelli5 parameter using the ratio of superficial liquid to superficial gas
pressure gradient, representative pipe liquid level hlr is calculated as
hlr = H l d r (7)
Equating the liquid pipe flow area to the liquid area in annular geometry (Alr=
Al), the liquid level hl in eccentric annular geometry (Figure 5.1.1) is calculated
using geometrical equations, which are functions of Al.
Equations 8-20 are given for determination of the liquid flow area Al and gas
flow area Ag,, as well as liquid contact perimeter Sl,, gas contact perimeter Sg
23
and interfacial length Si perimeters for a given liquid level in the annular
conduit. Figure 5.1.2 clearly represents the parameters used in these equations.
I II III
do
−1
2 − hl do
θ = cos if hl < (8)
do 2 2
di
−1
2 − hl di
φ = cos if hl < (9)
di 2 2
24
do
−1
2 − hl do
θ = cos − if hl ≥ (10)
do 2 2
di
−1
2 − hl di
φ = cos − if hl ≥ (11)
di 2 2
where the angles θ and φ are in radians. For a given liquid level in the annular
conduit, the geometrical parameters can be determined using appropriate
equations presented as follows;
do d
if hl < and hl < i
2 2
2
d o2 d d d d d
Al = θ − o o − hl sin θ − i φ + i i − hl sin φ (12)
4 2 2 4 2 2
Sl = d oθ + diφ (13)
π
Ag = (d o2 − di2 ) − Al (14)
4
25
Si = d o sin θ − d i sin φ (15)
S g = π ( d o + d i ) − Sl (16)
where the equations given for determination of Ag, Si and Sg are same for all
liquid level cases in the annular conduit.
do d
if hl < and hl ≥ i
2 2
2
d o2 d do di di di π di2
Al = θ − o − hl sin θ + φ + − hl sin φ - (17)
4 2 2 4 2 2 4
Sl = d oθ + di (π − φ ) (18)
do d
if hl ≥ and hl ≥ i
2 2
2
d2 d d d d di π (do2 − di2 )
Al = − o θ − o o − hl sinθ + i φ + i − hl sin φ + (19)
4 22 4 2 2 4
Sl = d o ( π − θ ) + d i ( π − φ ) (20)
26
The liquid flow area Alr in the pipe of representative diameter dr is determined
using geometrical equations (Equations 8-20) and replacing do by dr and
equating di to 0.
Once the liquid level and the liquid level dependent parameters for annular
geometry are determined accurately, flow pattern transitions can be checked
accordingly.
27
In this study this transition model is modified in two different methods. The first
approach is
1
2
h ( l ρ − ρ g ) g A g
vg > 1 − l (21)
d o dA
ρg l
dhl
1
2
h ( l ρ − ρ g ) g A g
vg > 1 − l (22)
dr dAl
ρg
dhl
28
A
vg = vsg (23)
Ag
hl
< 0.35 (24)
d
29
Figure 5.1.2.1- Intermittent Flow
In this study, it is suggested that the critical liquid level in annular geometry is
the liquid level representing the half of the flow area as the geometrical
difference is involved. The transition criterion for intermittent flow to annular
flow is as follows
30
Al
< 0.50 (25)
A
The transition takes place when the turbulent fluctuations are strong enough to
overcome the buoyant forces, which keep the gas at the top of the annulus. At
sufficiently high liquid velocities, i.e., superficial liquid velocity greater than 10
m/s, gas pocket is broken into small dispersed bubbles mixing with the liquid
phase (Figure 5.1.3.1).
This transition is originally proposed for pipe flow by Taitel and Dukler1.
Similarly, in this study the transition is given by;
31
1
4 gAg ( ρl − ρ g ) 2
vl ≥ (26)
S f ρ
i l l
Al
vl = vsl (27)
A
The friction factor and the Reynolds number of the liquid phase are defined by;
16
fl = if Rel ≤ 2100 (28)
Rel
6 3
1
fl = 0.001375 1 +
10
if Rel > 2100
Rel (29)
The critical Reynolds number for each phase is taken as 3000 in order to
guarantee the turbulent flow. Reynolds number for liquid phase is as follows
Rel =
( Al Sl ) vl ρl
(30)
µl
32
5.1.4 Stratified Smooth Flow to Stratified Wavy Flow Transition
The mechanism of this transition is based on Jeffrey’s theory for wave initiation,
as suggested by Taitel and Dukler1. The pressure and shear forces exerted by
the gas phase overcomes the viscous dissipation force in the liquid phase, as a
result waves occur on the interface.
1
4 µl ( ρ l − ρ g ) g 2
vg > (31)
s ρl ρ g vl
After the identification of the flow patterns, the frictional pressure losses can be
estimated. The flow mechanism of each flow pattern is studied independently.
33
As the momentum balance is investigated (Figure 5.1.1.1) following equations
are obtained for liquid and gas phase respectively.
∆P
− Al − τ wl Sl + τ i Si = 0 (32)
∆z l
∆P
− Ag −τ wg Sg −τ i Si = 0 (33)
∆z g
The wall shear stress of liquid phase τ wl , and the wall shear stress of gas
2
f ρv
τ wl = l l l (34)
2
2
f g ρ g vg
τ wg = (35)
2
Similar to the liquid friction factor and Reynolds number, gas phase friction
factor and Reynolds number are calculated as
16
fg = if N Re g ≤ 2100 (36)
Re g
34
The fanning friction factor for gas phase is given in Equation 37.
1
10
6 3
f g = 0.001375 1 + if N Re g > 2100 (37)
Re g
and
Re g =
(A g ( S g + Si ) ) v g ρ g
(38)
µg
determination of interfacial friction factor, fi, and the interfacial shear stress τi .
The interfacial friction factor fi, is assumed to be equal to gas friction factor fg
for stratified smooth and stratified wavy flow patterns. Then τi is given by;
f g ρ g (vg − vl )2
τi = (39)
2
The frictional pressure loss of the two-phase flow system through the eccentric
annuli may be calculated using either Equation 32 or 33.
35
5.2.2 Intermittent Flow
Intermittent flow consists of two main zones, i.e., liquid slug body and the gas
pocket/liquid film region (Figure 5.1.2.1). In this alternate flow of gas pockets
and liquid slugs, gas phase in large bullet-shaped pockets flows in the upper
part of the pipe. The liquid film flows below the gas pocket. Dukler and
Hubbard15 investigated the flow mechanism of intermittent flow through pipes.
Due to the complexity of the phase distributions, numerical iterations are
required in order to determine the liquid holdup distributions within the slug
body and gas pocket/liquid film zone. Several different pressure loss calculation
methods were presented for intermittent flow afterward. Xiao et al17 assumed
the film thickness to be uniform along the gas pocket/film zone. This region
was treated to be analogous with stratified flow. Even with this simplifying
assumption, the iterative solution procedures are not reliable. Petalas and Aziz24
claimed that, this method contradicted with the experimental results and
developed empirical equations for pipe flow. In this study, two methods are
proposed. The first approach is similar to the one suggested by Petalas and
Aziz24. As the momentum balance over a slug unit is investigated Equation 40 is
obtained.
∆P 1 τ lsπ d r τ lf Slf + τ g S g
= Ls + Lf (40)
∆z Lu A A
36
∆P ∆P ∆P ∆P ∆P
=Tµ + − − (41)
∆z ∆z fsl ∆z fgp ∆z fsl ∆z fgp
Tµ =f ( H l ) (42)
The translational velocity of liquid slug is calculated from equation (43) given by
Bendiksen37:
vt = C o v m + v d
(43)
Co is taken as 1.2 as suggested by Niklin et al38. The drift velocity of gas pocket
is determined from Zukoski39 correlation:
vd = f m vd ∞
(44)
37
1.76 gd r ( ρ l − ρ g )
vd ∞ = 0.54 − 0.56 (45)
Bo ρl
( ρl − ρG )
Bo =
σ
( gd )
2
r
(46)
Otherwise,
fm = 1 (48)
where
ρl vd ∞ d r
Re∞ = (49)
2 µl
Once these parameters are determined, liquid volume fraction in the slug and
the average liquid volume fraction of the slug unit can be calculated from
38
Gregory et al41 correlation and mass balance equation of liquid phase over the
slug unit.
1
Els = 1.39 (50)
v
1+ m
8.66
El s vt + vg (1 − Els ) − vsg
El = (51)
vt
The entrainment fraction, FE, is estimated using Petalas and Aziz24 correlation:
0.2
FE 0.074 v sg
= 0.735 N B (52)
1 − FE vsl
where
µl 2 vsg 2 ρ g
NB = (53)
σ 2 ρl
39
1 ( FE ) vsl + vsg
δ = 1 − (1 − El ) (54)
2 vsg
ρm = El ρl + (1 − El ) ρg (55)
µm = El µl + (1 − El ) µ g (56)
Once all the necessary parameters are defined, the frictional losses for slug
body can be determined using Equation 57.
2
∆P 2 f ml vm ρ m
=
∆z f sl dr
(57)
dr ρl vm
Reml = (58)
µl
40
∆P 4τ wl
=
∆z f gp dr
(59)
2
∆P f f ρl v f
=
∆z f gp 2
(60)
vsl
vf = (61)
El
From equation (41), total pressure loss is calculated. As the pressure loss
estimations of this method are compared with the experimental data, it is
observed that the method EPLDM gives more accurate results for 0.1143m -
0.05715m annular geometry than 0.0932m - 0.0488m annular conduit.
Therefore the second method, Omurlu Method (OM), is developed introducing
dimensionless group. The proposed equation in this study is as follows
41
vsl vsg ρm ∆P vsl vsg ρm ∆P
El + 1 − E l
∆P vm2 ρl ∆z fsl vm2 ρl ∆z fgp
= 1/ 4
∆z di (62)
d
o
Pressure gradients in the liquid slug and the liquid film/gas pocket zone are
calculated as described in the previous method. Then the pressure gradients
are multiplied with a new dimensionless group and divided by ratio of the
casing diameter to drillpipe diameter.
At very high gas flow rates, gas phase flows in a core of high velocity, which
contains entrained liquid droplets. The liquid phase flows as a thin film around
the pipe wall with a greater thickness at the bottom of the pipe than that at the
top (Figure 5.1.2.2). For practical purposes the film thickness δ, is assumed to
be uniform and equal to the average film thickness. When the momentum
balance equations of the liquid phase and the gas phase are examined, it is
observed that the mechanism of this flow pattern is analogous to stratified flow.
As the appropriate geometrical parameters are defined with empirical closure
equations, liquid film thickness can be determined from the combined
momentum equation by trial and error procedures. For liquid phase,
momentum equation is
42
∆P
− Af − τ w f S f + τ i Si = 0 (63)
∆z
For the gas phase in the gas core, momentum equation can be derived as
∆P
− Ac −τ i Si = 0 (64)
∆z
Sf 1 1
τ wf − τ i Si + =0 (65)
Af A
f Ac
It is assumed that the liquid droplets in the gas phase have the same velocity
with the gas core. Then, gas void fraction in the core, GVFc, is:
vsg
GVFc = (66)
vsg + vsl FE
43
µc = µg GVFc + µl (1 − GVFc ) (68)
respectively. With the geometrical configuration, liquid film velocity and core
velocity are defined as:
(1 − FE ) d r2
v f = vsl (69)
4δ ( d r − δ )
(vsg + vsl FE ) d r2
vc =
(d − 2δ )2 (70)
r
Petalas and Aziz24 correlation is preferred for the interfacial friction factor, fi:
0.085
fi σ 0.305
= 0.24 2 N Re (71)
fc ρc vc dc f
The pressure loss can be estimated from either Equation 63 or 64, once the film
thickness is appropriately determined.
44
5.2.4 Dispersed Bubble Flow
This flow pattern occurs at very high liquid flow rates, i.e., superficial liquid
velocity higher than 10 m/s. The gas phase is dispersed as discrete gas bubbles
within the continuous liquid phase (Figure 5.1.3.1). A simple homogeneous flow
is assumed since the gas bubbles are moving at the same velocity as liquid
phase. Pressure loss can be calculated from equation (72).
2
∆P 2 f m ρ m vm
= (72)
∆z dr
d r ρ m vm
Rem = (73)
µm
The mixture density, velocity and the mixture viscosity are calculated using the
no-slip liquid holdup λl term. The no-slip hold up is given by;
vsl
λl = (74)
vsl + vsg
Where
45
As mostly used in literature, mixture density and mixture viscosity are:
ρ m = λl ρl + (1 − λl ) ρ g (76)
µm = λl µl + (1 − λl ) µ g (77)
After the calculation of the necessary parameters, the frictional pressure loss of
the dispersed bubble flow can be calculated from Equation 72.
46
CHAPTER 6
EXPERIMENTAL WORK
47
Figure 6.1.1- Schematic view of the experimental setup
A thin plate is welded parallel to flow direction into the flange at the entrance of
the loop in order to contribute to the establishment of fully developed flow. An
air compressor and a centrifugal pump fed the two-phase air-water system. A
separator (Figure 6.1.2) is connected to the exit of the loop; hence water is
separated from two-phase mixture before being carried to the liquid tank. The
centrifugal pump is used with a magnetic flow meter and an electropneumatic
48
control valve (Figure 6.1.3) to measure and control the desired liquid flow rate.
Similarly, the air compressor is used with a volumetric flow meter and an
electropneumatic control valve to deliver required amount of gas into the loop.
The compressed air mixed with the water before entering to the annular
section. A pressure regulator is mounted before the gas flow meter as a safety
measure and to keep the air pressure controlled prior to entering to the test
section. The pressure of the gas phase is kept usually at 25 psi. The pressure of
the loop, frictional pressure losses, liquid and gas flow rates are measured
using the data acquisition system. Data logger and data acquisition software
are used to gather and store the experimental data digitally. The capacity and
brand name of each component in the experimental setup are presented in
Table 6.1.2.
49
Figure 6.1.3- Electropneumatic control valve
50
Table 6.1.2 Capacity and brand name of experimental components
Component Brand Name Capacity
Air Compressor TAMSAN 3000 l/min at 6 atm
Centrifugal Pump DOMAK 1.136 m3/min
Liquid Tank 2000 m3
Magnetic Liquid Flow TOSHIBA 1.136 m3/min
Meter
Volumetric Gas Flow COLE-PARMER INST. CO 0-1000 l/min at 25 psi
Meter
Electropneumatic Control SAMSON
Valves
Digital Differential COLE-PARMER INST. CO 0-1 psi
Pressure Transducers
Differential Pressure ASHCROFT 0-1” water
Gauges 0-2” water
0-5” water
0-10” water
The test section is 4.84 m. long and made of acrylic casing and steel drillpipe
(Figure 6.2.1). The geometrical configuration of the test section is given in
Table 6.1.1. The transparent casing allowed the observation and identification
of the flow patterns. Digital 1 psi differential pressure transducers and
differential pressure gauges of are used in order to measure frictional pressure
losses. Moreover, digital pressure transducers are mounted on the gas line at
51
the entrance of the gas flow meter and on the annular test section with the aim
of monitoring the system pressure at different locations. The average pressure
inside the loop was kept less than 20 psia. Table 6.1.2 clearly shows the
capacity and brand name of the pressure transmitters.
52
Lentrance = 50d hyd (78)
Equation 78 is porposed by Knudsen and Katz42 for pipe flow. In this study it
was modified by using hydraulic diameter.
The distance from the exit required to eliminate the chimney effect is given in
Equation 79 (reference 43). Similarly, the equation proposed for pipe flow was
modified for annular geometry inserting hydraulic diameter.
1
Lexit = 4.4 N Re 6 d hyd (79)
Thus, a fully developed region of 1.22 m for configuration1 and 0.61 m for
configuration 2 annular test sections are obtained.
The most important components of the experimental study are the frictional
pressure loss measurement and flow pattern visualization. The verification of
fully developed flow in the test section is essential. To ensure the accuracy of
the experimental data, frictional pressure losses are recorded for single phase
water flow through the annular test section. The pressure and flow rate
readings are recorded and compared with theoretical estimations. Figure 6.3.1
and Figure 6.3.2 show the accordance between the calculated pressure losses
and the experimental data for configuration 1 and configuration 2 eccentric
annular flow.
53
0.045
0.04
DP/DL (inch of water/inch)
0.035
0.03
0.025 MEASURED DP/DL
0.02 CALCULATED DP/DL
0.015
0.01
0.005
0
0 20 40 60 80 100
Q (gpm)
Figure 6.3.1- Frictional pressure loss gradient versus flow rate data of
water flowing through configuration 1
0.1
0.09
DP/DL (inch of water/inch))
0.08
0.07
0.06
MEASURED DP/DL
0.05
CALCULATED DP/DL
0.04
0.03
0.02
0.01
0
0 50 100 150 200 250
Q (gpm)
Figure 6.3.2- Frictional pressure loss gradient versus flow rate data of
water flowing through configuration 2
54
The pressure losses are calculated for concentric annuli using narrow slot
approach. However, the experimental data were collected from fully eccentric
conduit. The slight difference between the experimental data and estimated
pressure losses may be due to this fact.
The calibration of liquid and gas flow meters are checked and the readings are
found to be within the 1% accuracy as given in the calibration data sheets. The
differential analogue pressure gauge is designed to measure the pressure
losses to an accuracy of 2% of full scale of reading. Similarly, the pressure
readings of digital differential pressure transducers and the pressure gauges are
within the 0.25% and 0.13% accuracy of the full scale, respectively.
In the experimental work, air and water are used during two-phase flow tests.
Pressure losses and flow patterns are recorded at different gas and liquid flow
rates through horizontal fully eccentric annuli. The procedure of the two-phase
tests is as follows.
1) The water is pumped from the liquid collection tank to the loop using
centrifugal pump.
2) The water inside the water lines connecting the differential pressure
transducers’ low and high ends to the test section is flushed to prevent
the trapping of air bubbles in these lines.
55
3) Once the differential pressure transmitters are ready and the water flow
in the system is stabilized, control valve on the air line is opened and
adjusted to the desired flow rate.
4) After waiting for the stabilization of the flow rate of both phases, the
frictional pressure losses, system pressures and the flow rates are
recorded using data logger and data acquisition software. The record
time for each set of flow condition is one minute after steady state
condition is established. During the analysis of the experimental data the
average of this one minute recorded data is used.
5) Meanwhile, the observed flow pattern is recorded using high resolution
Sony Digital Video Camera Recorder.
A test matrix covering the minimum and maximum capacities of the system is
established to perform the experiments. While conducting the tests, air flow
rate is increased gradually keeping the water flow rate constant. Then the
water flow rate is increased to the next level and same procedure is followed.
56
CHAPTER 7
COMPUTER WORK
57
START
Ql,Qg, di,
do,µl, µg, ρl,ρg
1
1
2
NO
hl ( ρl − ρg ) g Ag
YES 4µl ( ρl − ρg ) g 2 YES
vg ≤ 1− vg > Stratified Wavy
dr ρg dAl
dhl sρl ρg vl
NO
1
NO NO
Al 4gAg ( ρl − ρg ) 2
< 0.50 vl ≥
A Si fl ρl
YES YES
Figure 7.1- Flow chart of Matlab code for the flow pattern
identification and frictional pressure loss determination
58
CHAPTER 8
After the calibration of experimental setup is finished and the validation of the
data acquisition system and the experimental setup with single phase water
flow experiments is carried out, two-phase flow measurements are conducted.
Firstly, flow pattern maps are generated for 0.0932m I.D acrylic casing –
0.0488m O.D drillpipe (configuration 1) and 0.1143m I.D acrylic casing –
0.05715m O.D drillpipe (configuration 2) annular conduits. The validation of the
proposed model’s flow pattern prediction with the experimental data is given in
detail in the following section. The comparison with generated flow pattern
maps using hydraulic diameter is also discussed. The accuracy of the proposed
model’s frictional pressure loss estimations are explained using experimental
pressure loss measurements. Empirical friction factor correlations are presented
for each observed flow pattern. The results and the discussion are given in
details in the following sections.
The flow pattern map is generated using hydraulic diameter and the transition
criterions proposed by Taitel and Dukler1. Figures 8.1.1 and 8.1.2 represent the
59
comparison of flow pattern maps for hydraulic diameter, dhyd and representative
diameter, dr approaches (GDM) for two different annular geometries, i.e.,
configurations 1 and 2.
100
GDM DB
10
dr
vsl (m/s)
SS
1
I SW
I
AN dhyd
0.1
SS
SW
0.01
0.1 1 10 100
vsg (m/s)
60
100
DB
10
GDM
dr
vsl (m/s)
SS
1 I SW
I
dhyd
0.1
AN
SS SW
0.01
0.1 1 10 100
vsg (m/s)
The flow patterns observed experimentally are stratified smooth (SS), stratified
wavy (SW) and intermittent (I) flow. Some of the flow pattern examples are
given as pictures taken during experiments in Appendix A.5. As compared with
the experimental data (Figures 8.1.1 and 8.1.2), it is observed that a significant
shift in flow pattern map generated using dhyd (blue dashed lines in Figures
8.1.1 and 8.1.2) occurred in stratified to non-stratified flow pattern boundaries
and stratified wavy to stratified smooth flow pattern transitions. Stratified
smooth and stratified wavy flow patterns are developed at considerably higher
liquid and gas flow rates than those estimated with hydraulic diameter
approach. It can be observed that the proposed first approach in this study
predicts accurately the flow pattern transitions for the conduit configuration 2.
The experimental data fits exactly within the predicted boundaries. However, as
the annular geometry changes, i.e., the flow area decreases, the accuracy of
the proposed GDM diminishes. From Figure 8.1.1, it is clearly observed that the
61
transition from intermittent flow to stratified smooth and stratified wavy flow
begins at higher liquid superficial flow rates than predicted velocities. Similarly,
annular flow pattern transition should shift towards right, to higher superficial
gas velocities. However, one should note that due to the experimental setup
limitations, i.e., capacity of experimental setup components given in Table
6.1.2, the majority of the data collected from configuration 1 annular conduit
consist of intermittent flow patterns. Therefore, more stratified flow data are
required in order to evaluate accurately the flow pattern transition models. The
liquid superficial velocity was greater than 0.1 m/s during the experiments. At
higher superficial liquid velocities, gas phase could not flow through the mixing
line since the pressure of the gas phase was kept less than 25 psi at the
entrance of flow meter. In case of hydraulic diameter used maps, the flow
pattern identifications are not correct for configurations 1 and 2 annular
conduits when compared with experimental results (blue dashed lines in Figures
8.1.1 and 8.1.2).
When the overall flow patterns in both annular geometries are observed it is
noted that the flow patterns are independent of the conduit dimensions.
However, the proposed first method (GDM) is geometry dependent because of
the presence of do in the transition equation, i.e., Equation 21. Therefore, the
proposed model is modified by replacing do with di (Equation 22). This second
method (OOM) becomes geometry independent. The comparison of the results
of OOM with the experimental data is presented for configuration 1 and
configuration 2 annular geometries in Figure 8.1.3 and Figure 8.1.4.
62
100
OOM SS
10
DB
SW
I
vsl (m/s)
I
1 dhyd
dr, 2nd aproach
0.1 SS AN
0.01
SW
0.1 1 10 100
vsg (m/s)
100
DB OOM
10
dr, 2nd approach
SS
vsl (m/s)
I
1 SW
I
dhyd
0.1 SS AN
SW
0.01
0.1 1 10 100
vsg (m/s)
As an expected result, the flow pattern predictions of OOM in this study are
highly accurate when compared with the experimental data. Similar to GDM,
63
the experimental results fit exactly within the appropriate flow pattern
boundaries configuration 2. In case of configuration 1, this geometry
independent method, OOM, gives more accurate flow pattern estimations than
GDM. Yet, more stratified flow data are needed in order to discuss and evaluate
the accuracy of OOM for configuration 1 annular geometry. The proposed
model still estimates the flow patterns more correctly than hydraulic diameter
approach.
The final part of the flow pattern comparison is the evaluation of Beggs and
Brill14 method (Appendix A.3) for horizontal fully eccentric annular two-phase
flow. Hydraulic diameter is used to represent the annular geometry. Figure
8.1.5 and Figure 8.1.6 show the comparison of the flow patterns generated
with the observed data.
100
10 DISTRIBUTED
SS
vsl (m/s)
1 SW
I I
modified
0.1
SEG-I TRANSITION Beggs &Brill
0.01
SEGREGATED
0.1 1 10 100
vsg (m/s)
64
10
DISTRIBUTED SS
SW
1
I
vsl (m/s)
0.1
SEG-I TRANSITION modified
Beggs & Brill
0.01
SEGREGATED
0.1 1 10 100
vsg (m/s)
The empirical flow pattern prediction model of Beggs and Brill14 is modified
using dr. This widely used model in industry could not predict correctly the flow
patterns observed. Especially in annular conduit configuration 1, the results are
not accurate. In case of configuration 2, intermittent flow data are within the
appropriate boundaries but stratified flow data are also in this region. Stratified
flow takes place at higher liquid superficial velocities according to the
experimental observations.
The flow patterns observed during experiments are only stratified (SS and SW)
flow and intermittent (I) flow. Although pressure loss determination equations
65
for annular (AN) and dispersed bubble (DB) flow are given in chapter 5, only
intermittent and stratified flow frictional pressure loss estimations are discussed
in this chapter.
The comparison of the experimental results and proposed model estimation for
pressure losses of stratified flow of water-air mixture flowing through fully
eccentric horizontal annulus of configurations 1 and 2 is presented in Figures
8.2.1 and 8.2.2.
350
300 +30%
calculated DP/DL(Pa/m)
250
modified Aziz et al.
200 modified Garcia et al.
-30% Ti=f(fG),this study
150
modified Beggs & Brill
100
50
0
0 50 100 150 200 250 300 350
measured DP/DL(Pa/m)
66
250
calculated DP/DL (Pa/m) +30%
200
150
modified Aziz et al.
-30% modified Garcia et al.
100
Ti=f(fG),this study
modified Beggs & Brill
50
0
0 50 100 150 200 250
measured DP/DL (Pa/m)
The dashed lines in Figures 8.2.1 and 8.2.2 are ±30 % error margin, and the
solid line represents the perfect match between the experimental stratified flow
data and calculated results for configuration 1 and configuration 2 annular
conduits respectively. Mostly used models of Petalas and Aziz24, Garcia et al25
and Beggs and Brill14 for pipe flow are modified for annular geometry by using
dr instead of pipe diameter as shown in Appendix A.3. Modified Petalas and
Aziz24 and Beggs and Brill14 models give underestimated pressure loss results
when compared with the experimental results. The correct determination of
interfacial shear stress is the most important step during the pressure loss
calculations. Instead of using empirical correlation for interfacial friction factor
(Petalas and Aziz24), gas friction factor is preferred in this study. Although
modified Garcia et al’s model25 gives accurate results the proposed model’s
performance is agreeable when validated with experimental data. It can be
67
observed that the procedure presented in this study for stratified flow estimates
the frictional pressure losses with a reasonable accuracy for both annular
geometries. However, it should be remarked that the amount of data is not
sufficient to generalize the results due to the limitations of the experimental
setup, i.e. Table 6.1.2.
In chapter 5 two methods, i.e., EPDM and OM are presented for pressure loss
determination in case of intermittent flow. In this study, in EPDM an empirical
equation (Equation 80) is developed.
Figure 8.2.3 and Figure 8.2.4 represent the comparison of the experimental
results and EPDM’s estimations for pressure losses of intermittent flow. The
performances of the modified Petalas and Aziz24, modified Garcia et al.25 and
modified Beggs and Brill14 models are also evaluated with the experimental
data.
68
calculated DP/DL (Pa/m) 2500
2000
500
0
0 500 1000 1500 2000 2500
measured DP/DL (Pa/m)
800
700
calculated DP/DL(Pa/m)
600
modified Aziz et a.
500
modified Garcia et al.
400 EPDM, this study
300 modified Beggs and Brill
200
100
0
0 200 400 600 800
measured DP/DL(Pa/m)
69
Similarly, in the figures above, the dashed lines are ±30 % error margin, and
the solid line represents the perfect match between the experimental and
calculated results. From Figures 8.2.3 and 8.2.4, it is observed that EPDM in
this study estimates accurately the frictional pressure losses for intermittent
flow. The results are more correct for configuration 2. As in case of
configuration 1, at relatively high measured pressure losses the model
overestimates the frictional pressure losses. However, the modified models, i.e.,
Garcia et al.25 and Beggs and Brill14, highly underestimate the pressure losses.
As the measured frictional pressure loss increases these modified models still
calculate very low pressure loss estimations. Modified Petalas and Aziz24 model
gives closer pressure results to the experimental data than other modified
models. However, EPDM proposed in this study is the most accurate model
among others.
70
calculated DP/DL (Pa/m) 2500
2000
500
0
0 500 1000 1500 2000 2500
measured DP/DL (Pa/m)
800
700
calculated DP/DL(Pa/m)
600
modified Aziz et al.
500
modified Garcia et al.
400 OM, this study
300 modified Beggs and Brill
200
100
0
0 200 400 600 800
measured DP/DL(Pa/m)
71
In Figures 8.2.5 and 8.2.6, the dashed lines are ±30 % error margin, and the
solid line represents the perfect match between the experimental and
calculated results. It is remarked that OM proposed in this study estimates the
frictional pressure losses with a high accuracy for intermittent flow through
configurations 1 and 2 horizontal eccentric annular conduits. As discussed in the
validation of EPDM, the modified models, i.e., Garcia et al.25 and Beggs and
Brill14, highly underestimate the pressure losses, and modified Petalas and
Aziz24 model gives closer pressure results to the experimental data than other
modified models. However, OM proposed in this study is the most accurate
model among others, and most of the calculated pressure losses are within the
±30 % error margin. The reason for inaccurate pressure loss estimations of
modified models may be the fact that these models were originally developed
for two-phase pipe flow. Also, there are empirical equations used in these
models which are similarly proposed for circular pipes.
72
Table 8.2.1 represents quantitatively the average, maximum and minimum
errors for pressure loss estimations. As remarked from this table, for
intermittent flow EPDM proposed in this study has an average error of 35.1%
and 31% for configurations 1 and 2. However, OM proposed in this study
determines the frictional pressure losses in both annular configurations more
accurately with an average error of 30.5% and 15.1%. Other compared models
have very low accuracy in pressure loss estimations. In stratified flow the
proposed procedure in this study calculates the pressure drop accurately with
an average error of 51.1% for configuration 1 and 25.3% for configuration 2.
In configuration 1 modified Garcia et al25 model estimates the pressure drop
more accurately than the proposed model with an average error of 36.7%. In
both annular geometries and for both flow patterns, proposed mechanistic
model has the least minimum errors. Moreover, as the amount of data is
considered, the proposed method in this study can be considered as accurate
and applicable. This proposed mechanistic model is highly useful for practical
purposes, since the calculation procedure is simple and accurate for both
annular geometries.
In this study, although the simplicity and the accuracy of the model proposed
for stratified flow and OM for intermittent flow, for practical purposes empirical
friction factor correlations are developed. The frictional pressure losses are
estimated theoretically for each flow pattern independently using representative
diameter, dr. The correlations are developed based on data obtained from
METU-PETE-CTMFL multiphase flow loop using statistical methods. The data
73
obtained include only stratified and intermittent flow data due to the
geometrical and two-phase system restrictions, i.e., Table 6.1.2.
For stratified flow, the friction factor correlation is given in Equation 81.
−1.1734
f f = 9880 ( Re m ) (81)
Where Rem, the mixture Reynolds number can be calculated using Equation 73.
The friction factor versus mixture Reynolds number relation of the experimental
stratified flow data is presented in Figure 8.3.1. The friction factor is calculated
from Equation 72.
74
1
0.1
ff
0.01
0.001
10000 100000 1000000
Rem
75
0.008
0.007
0.006
0.005
0.004
ff
0.003
0.002
0.001
0
100000 1000000
Rem
N Remixλ = λl N Re sl + (1 − λl ) N Re sg (82)
where liquid holdup, λl is calculated from Equation 74. As the results are
investigated, it is remarked that at NRemixλ = 100000, the slope of the curve
changes. Figures 8.3.3 and 8.3.4 show the relation between the friction factor
and the Reynolds number for each case.
76
1
0.1
ff
0.01
0.001
10000 100000
NRemixλ
λ
−0.0594
f f = 0.0085 ( N Remixλ ) (83)
−0.4676
f f = 0.9435 ( N Remixλ ) (84
77
0.006
0.005
0.004
0.003
ff
0.002
0.001
0
100000 1000000
NRemixλ
λ
Then, pressure loss can be determined using Equation 62. The performance of
the empirical correlations is compared with experimental data. Figures 8.3.5
and 8.3.6 represent the accuracy of the proposed correlations. Dashed lines
represent ±30% error margin. It can be noted that all of the estimated pressure
losses are within this error range. The empirical model can estimate the
frictional pressure losses accurately.
78
400
350
calculated DP/DL (Pa/m)
300
250
200
150
100
50
0
0 50 100 150 200 250 300 350 400
measured DP/DL (Pa/m)
1600
1400
calculated DP/DL (Pa/m)
1200
1000
800
600
400
200
0
0 200 400 600 800 1000 1200 1400 1600
measured DP/DL (Pa/m)
79
CHAPTER 9
CONCLUDING REMARKS
80
with the experimental data. Geometry dependent model (GDM) model
identifies flow patterns correctly for configuration 2, but the
experimental stratified data of configuration 1 are not within the
appropriate flow pattern boundaries predicted by GDM.
• Interfacial shear stress should be well defined for stratified flow. Gas
friction factor is used as the interfacial friction factor. The model’s
pressure loss estimations are accurate when compared with
experimental results.
81
dimensionless group respectively. The pressure losses are estimated
accurately using EPDM for 0.1143m I.D – 0.05715m O.D horizontal
eccentric annuli when compared with experimental data. However, for
flow through 0.0932m I.D – 0.0488m O.D annular conduit, EPDM
overestimates the pressure losses at high liquid and gas flow rates, i.e.,
1 m/s superficial liquid velocity and 2.5 m/s superficial gas velocity. OM
determines the frictional pressure losses with a high accuracy for two-
phase flow through both annular geometries. OM is more accurate and
reliable as the amount of data compared is considered.
82
RECOMMENDATIONS
• Water and air was used during the experiments. Experiments with more
viscous liquid phase can be conducted in order to analyze the effect of
viscosity on flow pattern transition boundaries. Moreover, different fully
eccentric annular geometrical configurations are needed to generalize
the accuracy of the mechanistic model.
83
• The effect of eccentricity on flow pattern identification and pressure loss
determination may be analyzed using concentric and eccentric annular
conduits with different eccentricities.
84
REFERENCES
1. Taitel, Y., and Dukler, A.E.: “A Model for Predicting Flow Regime
Transition in Horizontal and Near Horizontal Gas-Liquid Flow,” AIChE J.
(1976) 22 No. 1, 47.
2. Guo, B., and Ghalambor, A.: “Gas Volume Requirements for
Underbalanced Drilling, Deviated Holes”, PennWell Corporation, 2002.
3. Rodriguez, P. V.: “Experimental Determination of Minimum Air and Water
Flow Rates for Effective Cutting Transport in High Angle and Horizontal
Wells”, MSc. Thesis, 2001, The University of Tulsa.
4. Wallis, G.B.: “One-Dimensional Two-Phase Flow,” McGraw-Hill Book Co.
Inc., New York City (1969).
5. Lockhart, R.W., and Martinelli, R.C.: “Proposed Correlation of Data for
Isothermal Two-Phase, Two-Component Flow in Pipes,” Chemical
Engineering Progress 45, No. 1, 39-48 (1949).
6. Duns, H.Jr. and Ros, N.C.J.: “Vertical Flow of Gas and Liquid Mixtures in
Wells,” Proc., Sixth world Petroleum Congress, Frankfurt-am-Main,
Germany (1963) 451.
7. Baxendell, P.B. “Producing Wells on Casing Flow-An Analysis of Flowing
Pressure Gradients,” JPT (August 1958) 59;Trans., AIME,213.
8. Gaither, O.D.,Winkler, H.W.,and Kirkpatrick,C.V.: “Single-and Two-Phase
Fluid Flow in Small Vertical Conduits Including Annular Configurations,”
JPT (March 1963) 309; Trans., AIME,228.
85
9. Angel, R.R. and Welchon, J.K.: “Low-Ratio Gas-Lift Correlation for
Casing-Tubing Annuli and Large-Diameter Tubing,” Drill.& Prod. Prac.
(1964) 100.
10. Winkler, H.W.: “Single-and Two-Phase Vertical Flow Through 0.996 X
0.625-Inch Fully Eccentric Plain Annular Configuration,” PhD dissertation,
U. of Texas, Austin, Texas (1968).
11. Dukler, A.E., Wickes, M. III, and Cleveland, R.G.: “Frictional Pressure
Drop in Two-Phase Flow: An Approach Through Similarity Analysis”
AIChE J. (1964) 10, No. 1, 44.
12. Eaton, B.A. and Brown, K.E.: “The Prediction of Flow Patterns, Liquid
Holdup and Pressure Losses Occurring During Continuous Two-Phase
Flow in Horizontal Pipelines,” Technical Report, The U. of Texas (October
1965).
13. Flanigan, O.: “Effect of uphill flow on pressure drop in design of two-
phase gathering systems,” Oil and Gas J. 56-132 (1958).
14. Beggs, H.D., and Brill, J.P.: “A Study of Two-Phase Flow in Inclined
Pipes,” JPT (May 1973) 607; Trans., AIME, 255.
15. Dukler, A.E., and Hubbard, M.G.: “A Model For Gas-Liquid Slug Flow in
Horizontal and Near Horizontal Tubes,” Ind. Eng. Chem. Fundam. (1975)
14, 337.
16. Barnea, D.: “A Unified Model for Predicting Flow Pattern Transitions for
the Whole Range of Pipe Inclinations,” Int. J. Multiphase Flow (1987)
13, No. 1, 1.
17. Xiao, J.J., Shoham, O., and Brill, J.P.: “A Comprehensive Mechanistic
Model for Two-Phase Flow in Pipelines.” Paper SPE 20631 presented at
86
the 1990 SPE Annual Technical Conference and Exhibition, New Orleans,
23-26 September.
18. Barnea, D., Shoham, O., and Taitel, Y.: “Flow Patter Transition for
Vertical Downward Inclined Two-Phase Flow; Horizoontal to Vertical,”
Chem. Eng. Sc. 37, No. 5, 735-740 (1982).
19. Baker, A., Nielsen, K., and Gabb, A.: “Pressure loss, Liquid Holdup
Calculations developed,” Oil & Gas J., 55-59 (March 14, 1998).
20. Andritsos, N., and Hanratty, T. J.: “Influence of Interfacial Waves in
Stratified Gas-Liquid Flows,” AIChE J. 33, No. 3, 444-454 (1987).
21. Oliemans, R.V., A., Pots, B.F., and Trope, N.: “Modelling of Annular
Dispersed Two-Phase Flow in Vertical Pipes,” Int. J. Multiphase Flow 12,
No. 5, 711-732 (1986).
22. Gomez L.E., Shoham O., Schmidt Z., Chokshi R.N. and Northug T.:
“Unified Mechanistic Model for Steady-State Two-Phase Flow: Horizontal
to Vertical Upward Flow,” SPE Journal 5, No. 3, 339-350 (September
2000).
23. Ouyang, L.B., and Aziz, K.: “Development of New Wall Friction Factor
and Interfacial Friction Factor correlations for Gas/Liquid Stratified Flow
in Wells and Pipes,” paper SPE 35679 presented at the 1996 SPE
Western Regional Meeting, Anchorage, 22-24 May.
24. Petalas, N., and Aziz, K.: “A Mechanistic Model for Multiphase Flow in
Pipes,” JCPT 39, No. 6, 43-55 (2000).
25. Garcia, F., Garcia, R., Padrino, J.C., Mata, C., Trallero, J.L. and Joseph,
D.D.: “Power Law and Composite Power Law Friction Factor Correlations
for Laminar and Turbulent Gas-Liquid Flow in Horizontal Pipelines,” Int.
J. Multiphase Flow 29, 1605-1624 (2003).
87
26. Theofanous, T.G. and Hanratty, T.J.: “Appendix 1: Report of Study
Group on Flow Regimes in Multifluid Flow,” Int. J. Multiphase Flow 29,
1061-1068 (2003).
27. Sadatomi, M., Sato, Y., and Saruwatari, S.: “Two-Phase Flow in Vertical
Noncircular Channels,” Int. J. Multiphase Flow 8, 641-655 (1982).
28. Hasan, A.R., and Kabir, C.S.: “Two-Phase Flow in Vertical and Inclined
Annuli,” Int. J. Multiphase Flow 18, 279-293 (1992).
29. Caetano, E.F., Shoham, and O., Brill, J.P.: “Upward Vertical Two-Phase
Flow Through An Annulus, Part-I: Single Phase Friction Factor, Taylor
Bubble Velocity And Flow Pattern Prediction,” ASME J. Energy Resour.
Technol. 114, 1-13 (1992).
30. Salcudean, M., Chun, J.H., and Groeneveld, D.C.: “Effect of Flow
Obstructions on Void Distribution in Horizontal Air-Water Flow,” Int. J.
Multiphase Flow 9, 91-96 (1983).
31. Salcudean, M., Chun, J.H., and Groeneveld, D.C.: “Effect of Flow
Obstructions on the Flow Pattern Transitions in Horizontal Two-Phase
Flow,” Int. J. Multiphase Flow 9, 87-90 (1983).
32. Sunthankar, A.A.: “Study of the Flow of Aerated Drilling Fluids in Annulus
under Ambient Temperature and Pressure Conditions,” Ms. Thesis, The
University of Tulsa, 2000.
33. Zhou, L.: “Cuttings Transport with Aerated mud in Horizontal Annulus
under Elevated Pressure and Temperature Conditions,” PhD Dissertation,
The University of Tulsa, 2004.
34. Gucuyener, I.H.: “Design of Aerated Mud for Low Pressure Drilling,”
paperSPE 80491 presented at the SPE Asia Pacific Oil and Gas
Conference, Jakarta, 15-17 April, 2003.
88
35. Lage, A.C.V.M.; Rommetveit, R.; Time, R.W.: “An Experimental and
Theoretical Study of Two-Phase Flow in Horizontal or Slightly Deviated
Fully Eccentric Annuli,” SPE 62793, IADC/SPE Asia Pacific Drilling
Technology, Kuala Lumpur, Malaysia, 11–13 September 2000.
36. Aziz, K., Govier, G.W., and Fogarasi, M.: “Pressure Drop in Wells
Producing Oil and Gas,” J. Can. Petrol. Technol 11, 38 (1972).
37. Bendiksen, K.H.: “An Experimental Investigation of the Motion of Long
Bubbles in Inclined Tubes,” Int. J. Multiphase Flow 10, 467 (1984).
38. Niklin, D.J., Wilkes, J.O., and Davidson, J.F.: “Two Phase Flow in Vertical
Tubes,” Trans. Inst. Chem. Engr. 40, 61-68 (1962).
39. Zukoski, E. E.: “Influence of Viscosity, Surface Tension, and Inclination
Angle on Motion of Long Bubbles in Closed Tubes,” J. Fluid Mech. 25,
821-8337 (1966).
40. Weber, M.E.: “Drift in Intermittent Two-Phase Flow in Horizontal Pipes,”
Canadian J. Chem. Engg. 59, 398-399 (1981).
41. Gregory, G.A., Nicholson M.K., and Aziz K.: "Correlation of the Liquid
Volume Fraction in the Slug for Horizontal Gas-Liquid Slug Flow," Int. J.
Multiphase Flow 4, No. 1, 33-39 (1978).
42. Knudsen, J.G., and Katz, D.L.: “Fluid Dynamics and Heat Transfer,”
McGraw-Hill Book Co. Inc., New York City (1959)
43. ANSYS CFX version 8.0, Tutorial, Section Laminar to Turbulent Flow.
89
APPENDIX
Pressure loss for intermittent flow can be determined using Equation A.1
∆P ∆P ∆P
− =η + (1 − η )
∆z ∆z fsl ∆z f gp
(A.1)
η = λl (0.75− E )
l
(A.2)
vt = Co vm + vd (A.3)
vd = f m vd ∞ (A.4)
90
where the correlation of vd∞ proposed by Bendiksen37 has only
horizontal velocity term and is determined by using Weber40
correlation:
1.76 gd r ( ρl − ρ g )
vd ∞ = 0.54 − 0.56 (A.5)
Bo ρl
( ρl − ρ g )
Bo =
σ
( gd )
2
r (A.6)
fm = 1 (A.8)
where
ρl vd ∞ d r
Re∞ = (A.9)
2µl
91
Once these parameters are determined, liquid volume fraction in
the slug and the average liquid volume fraction of the slug unit
can be calculated from Gregory et al41 correlation and mass
balance equation of liquid phase over the slug unit.
1
Els = 1.39 (A.10)
v
1+ m
8.66
0.2
FE 0.074 vsg
= 0.735 N B (A.12)
1 − FE vsl
where
µl 2 vsg 2 ρ g
NB = (A.13)
σ 2 ρl
92
The dimensionless liquid film thickness is determined from
equation (A.14)
1 ( FE ) vsl + vsg
δ = 1 − (1 − El ) (A.14)
2 v
sg
2
∆P 2 f ml vm ρ m
= (A.15)
∆z f sl g dr
∆P 4τ wl
= (A.16)
∆z f gp dr
2
∆P f f ρLv f
=
∆z f gp 2
(A.17)
93
For the stratified flow, pressure loss determination procedure is
similar the one presented in chapter 5. The differences are the
determination of liquid friction factor and interfacial friction factor
which are presented in Equations A.18 and A.19.
0.731
fl = 0.452 f sl (A.18)
1.335
ρl d r g
fi = (0.004 + 0.5 x10−6 Re sl ) N Frl 2
(A.19)
ρ v
g g
4 Al
dl = (A.20)
Sl
4 Ag
dg = (A.21)
Sg
Then, liquid and gas phase Reynolds number are calculated using
Equation A.22 and Equation A.23.
dl vl ρl
Rel = (A.22)
µl
d g vg ρ g
Re g = (A.23)
µg
94
The pressure loss for stratified flow can be calculated from
Equation 32 or Equation 33.
Pressure loss for intermittent flow can be determined using Equation A.24 and
A.25 for stratified flow.
∆P vm2
= ff gar 2 ρ m (A.25)
∆z dr
The frictional pressure loss for this flow pattern is calculated from Equation
A.25.
95
A.3 Modified Beggs and Brill Model
The flow regime is Segregated if λl<0.01 and NFrl<L1 or λl ≥ 0.01 and NFrl<L2
Where L1 and L2 are determined from Equations A.27 and A.28.
L1 = 316λl0.302 (A.27)
L3 = 0.1λl−1.4516 (A.28)
The transition zone between segregated and intermittent flow takes place if
λl ≥ 0.01 and L2 ≤ NFrl ≤ L3. L2 can be determined from Equation A.29.
L2 = 0.0009232λl−2.4684 (A.29)
L3 − N Frl L − N Frl
H l (transition ) = H l ( segregated ) + 1 − 3 H l (int ermittent ) (A.30)
L3 − L2 L3 − L2
The intermittent flow takes place if 0.01 ≤ λl ≤ 0.4 and L3< NFrl ≤ L1 or λl ≥ 0.4
and L3< NFrl ≤ L4 where L4 is:
L2 = 0.5λl−6.738 (A.31)
96
Similarly, the transition criterion for distributed flow is in case λl<0.4 and
NFrl ≥ L1 or λl ≥ 0.4 and NFrl ≥ L4. Then dimensionless liquid holdup Hl(o) is
determined.
aλlb
H l (o) = c (A.32)
N Frl
For all flow conditions Hl(o) ≥ λl should be satisfied. The friction factor is
determined from Dukler et al’15s method. Then the total pressure gradient is
calculated from Equation A.33.
∆P
∆P ∆z f
= (A.33)
∆z ρ v v
(1 − m m sg )
gP
97
Where hg and hg’ are the gas level in the conduit as shown in Figure A.1.
vg’ hg’ P’ vg
hg
P
hl’ vl hl’
Figure A.1- The analysis of the forces during wave growth in the
conduit
1 2 2
P − P' = ρ g (v 'g − vg ) A.35
2
2 2 ρl − ρ g
(v 'g − vg ) > 2(hg − hg ')( )g A.36
ρg
2
' 2 2 hg
v = vg '
g h A.37
g
98
Equation A.37 can be expressed for round pipe in terms of Equation A.38.
2
' 2
Ag
2
v = vg '
g A A.38
g
and
dA
Ag = Ag + (hl − hl' ) g A.40
dhl
Also writing Equation A.41, the derivative of gas flow area Ag to liquid flow area
Al can be taken.
Ag = A − Al A.41
dA
Ag' = Ag − (hl − hl' ) − l A.42
dhl
99
Substituting the parameters calculated above in Equation A.36, Equation A.43
the criterion for transition to intermittent flow from stratified flow is obtained.
12
Ag ρl − ρ g
'
vg > gAg
Ag dA A.43
ρ l
g dh
l
Where
Ag' hl
≅ 1− A.44
Ag dr
If the equilibrium level approaches to top of the conduit, the ratio of areas in
the left hand side of Equation A.44 goes to 0. Also, the right hand side of this
equation goes to 0. Therefore, the approximation presented in Equation A.44
can be used.
100
A.5 Pictures Taken During Experiments
101
Figure A.4- Stratified wavy flow through configuration 1
102
Figure A.6- Intermittent flow through configuration 1
103
Figure A.8- Intermittent flow through configuration 1 at high liquid
and gas flow rates
104
Figure A.10- Stratified smooth flow through configuration 2
105
Figure A.12- Stratified wavy flow through configuration 2
106
Figure A.14- Intermittent flow through configuration 2
107