A New Comprehensive, Mechanistic
A New Comprehensive, Mechanistic
ments conducted at nearly atmospheric conditions. However, dur- in which Dep⳱DIC + DOT is the equiperiphery diameter.
ing UBD operations, gas and liquid are injected simultaneously at When Dep is greater than the right side of Eq. 5, bubble flow
a high injection pressure. This high pressure has the same effect of can exist. Therefore, the agglomeration or coalescence of small gas
reducing gas velocity as that in upward flow. The downward tur- bubbles into large Taylor bubbles, which occurs when the in-situ
bulent flow generally maintains a dispersed gas phase, and gas gas rate increases (void fraction increases), is the basic transition
injection results in gas void fractions of less than 0.78. Therefore, mechanism from bubble to slug flow. Except for Caetano,17 who
one can assume that dispersed bubble, bubble, and slug flow, as suggested that the bubble-to-slug transition occurs at a void frac-
suggested by Hasan,7 are also the dominant flow patterns in down- tion of approximately 0.20, other investigators2,18 and recently
ward flow through the drillstring. Lage et al.9 agree that such transition occurs at a void fraction of
approximately 0.25.
Flow-Pattern Prediction Models Although there is wide agreement in the gas void fraction value
Upward Flow in Annuli. Caetano,17 Hasan and Kabir,2 Kelessi- at which bubble-to-slug transition occurs, there is an inconsistency
dis et al.,18 and recently Lage et al.8,9 agree that flow patterns in the criterion used to express this transition in terms of measur-
observed in vertical, concentric annuli are similar to those seen in able variables, such as superficial phase velocities.
pipes. They also agree in using the framework developed by Taitel Zuber and Findlay24 stated that the effect of the nonuniform
et al.13 to predict the flow-pattern transitions by adapting annular flow and concentration distribution across the pipe and of the local
geometrical parameters. Based on these different works, flow pat- relative velocity between the two phases affect two-phase flow
terns can be predicted by defining the transition boundaries be- systems and define a velocity profile coefficient, C0. On the other
tween them (see Fig. 2). Although these authors consider five hand, Wallis25 determined that bubble swarm affects the motion of
different flow patterns (dispersed bubble, bubble, slug, churn, and a single bubble and introduced the concept of bubble swarm effect,
annular), for the reason explained previously we have emphasized HLn. Different from other authors,2,9,17,18 who separately consider
only dispersed bubble, bubble, and slug flow. However, to avoid these effects, both are taken into account in the present model.
convergence problems during the calculations, a transition to churn Therefore, the bubble-slug transition is defined by
and annular flow are considered. If churn flow occurs, it is treated
as slug flow. For the annular flow occurrence, a simplified annular uG − C0um = u⬁HLn, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
flow model proposed by Taitel and Barnea22 was implemented. In
UBD operations, these simplistic assumptions have a negligible in which n⳱the swarm-effect exponent.
effect on the overall calculations because when churn or annular To fit experimental with analytical data, different authors2,8,9,17
flow occur, they are relatively close to the surface. used different values for the velocity profile coefficient C0, but all
Bubble-to-Slug Transition. During bubble flow, discrete agree in using 0.5 for the swarm-effect exponent. In the current
bubbles rise with the occasional appearance of a Taylor bubble.13 model, the most widely used values, (C0⳱1.2)2,6 and (n⳱0.5),26
The discrete bubble rise velocity (after Harmathy23) is given by were used.
冋 册
Although most authors agree that the bubble-slug transition
共L − G兲g 1Ⲑ4
occurs at a void fraction of approximately 0.25, better results were
u⬁ = 1.53 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
L2 obtained in this study when considering a void fraction of 0.20, as
suggested by Caetano17 and Lage et al.8 This probably happens
The rise velocity of the Taylor bubbles, on the other hand, is given by because the studies reported previously were carried out with
冋 册冉冊冉 冊
ward flow.
1.6 0.5
L 0.6 2fFH 0.4
Similarly, as stated by Taitel et al.,13 Hasan7 considers that re-
um1.2
共L − G兲g Dh gardless of the existing turbulence forces, the gas void fraction can
冉 冊 uSG 0.5 not exceed 0.52 without causing transition to slug flow. Therefore,
= 0.725 + 4.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8) Eq. 10 also may be used to predict the transition boundary between
um dispersed bubble and slug flow for downward flow conditions.
Because the bubble rise velocity in dispersed bubble flow is very
small compared to the local velocity values, a nonslip homoge- Flow Behavior Prediction Models
neous mixture flow description represents the flow parameters Models that allow accurate prediction of pressure and phase con-
relatively well.17 Therefore, the homogeneous Fanning friction centration are required for each particular flow pattern predicted
factor, fFH, in Eq. 8 is calculated with the nonslip liquid-holdup previously. Considering that the three dominant flow regimes in
concept defined by UBD operations are dispersed bubble, bubble, and slug, six inde-
pendent models are required to handle both downward two-phase
uSL flow through the drillstring and upward two-phase flow through
HL = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
um the annulus. Additionally, a two-phase flow-bit model is required
to predict the pressure drop through the nozzles.
Dispersed Bubble-to-Slug Flow Transition. Taitel et al.13
stated that regardless of how much turbulent energy is available to
Annular-Bubble Flow Model. This model is based on the drift-
disperse the mixture, bubbly flow can not exist at in-situ gas rates
flux approach, which considers the velocity difference between the
so high that bubbles are packed close enough to be in contact.
phases or between a phase and the average volumetric velocity of
Assuming the bubbles are spherical and arranged in a cubic lattice,
the mixture.28 Similar to the bubble-slug transition model (Eq. 6),
they determined that the maximum allowable gas void fraction
the implemented model takes into account both the velocity profile
under bubbly conditions is 0.52. Higher void fraction values will
coefficient, C0, and the bubble swarm effects, HLn. Thus, expressed
cause the transition to slug flow. Thus, considering this gas void
in superficial velocities, the bubble drift-flux model used to predict
fraction limit and the dispersed-bubble homogeneous flow condi-
the liquid holdup is given by9
tions, Eq. 10 gives the transition boundary between dispersed
bubble and slug flow. This is shown as Transition C in Fig. 2. uSG
HLnu⬁ = − C0um. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14)
uSL = 0.923uSG. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10) 共1 − HL兲
Slug-to-Churn Transition. Eq. 11, proposed by Tengesdal et After the liquid holdup is calculated with Eq. 14, the total pressure
al.,27 is used to predict this transition. At a gas void fraction of gradient can be estimated. For steady-state flow, it consists of
0.78, the slug structure is completely destroyed, and the two dis- gravity, friction, and convective acceleration losses and is given by
冉 冊 冉 冊 冉 冊 冉 冊
tinct regions, liquid slug and Taylor bubble, no longer exist, causing
the transition to churn flow, represented as Transition D in Fig. 2. dp dp dp dp
= + + . . . . . . . . . . . . . . . . (15)
dZ dZ dZ dZ
uSL = 0.0684uSG − 0.292公gDep. . . . . . . . . . . . . . . . . . . . . . . . (11)
T Hy Fric Acc
search in upward, two-phase flow in pipes, there are only a few As suggested by Caetano,17 the Fanning friction factor, fF, is cal-
investigations of gas-liquid mixtures in downward flow in pipes. culated with the Gunn and Darling17 procedure for turbulent flow,
Moreover, these investigations have been carried out at nearly which is a function of the diameter ratio K ⳱ DIC/DOT and the
atmospheric conditions, which greatly differ from those occurring mixture’s Reynolds number, which is defined by
during jointed-pipe UBD operations in which liquid and gas are
injected simultaneously at a high injection pressure. With these mumDh
NRe = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (19)
limitations, the approach of Hasan,7 which was extensively vali- LHL + G 共1 − HL兲
冉 冊
dp
dZ Acc
=
mumuSG dp
p dZ
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (20)
The hydrodynamic parameters required in the pressure gradient
equations for both FDTB and DTB are deduced in Appendices A
Annular-Dispersed Bubble Flow Model. Because of the high and B, respectively. Thus, for slug flow, the gravitational compo-
turbulence forces during dispersed bubble flow, the dispersed gas nent is given by12
bubbles do not exhibit significant slippage through the liquid
phase, and the velocity profile is approximately flat.17 Therefore,
the slip velocity is negligible (u⬁ ≈0), and the velocity profile
coefficient is approximately one.28 Thus, liquid holdup can be
冉 冊
dp
dZ Hy
= 关共1 − 兲 mLS + mTB兴g. . . . . . . . . . . . . . . . . . . . . . (21)
Annular-Slug Flow Model. Fernandes et al.26 developed the first in which mLS⳱LHLLS+G(1 − HLLS) and the friction factor is
mechanistic model for slug flow in vertical pipes, and then Caet- calculated as before with a Reynolds number defined by
ano17 implemented this model for vertical annuli. These works are
adopted and modified in this study. Different from these works, the mLSumDh
present model considers variable liquid holdup in the liquid slug NRe = . . . . . . . . . . . . . . . . . . . . . . . . . . (23)
and incorporates the bubble drift-flux model represented by Eq. 14 LHLLS + G 共1 − HLLS兲
to predict the in-situ gas velocity in the liquid slug zone. Moreover,
the current slug flow model takes into account both fully devel- The pressure drop caused by acceleration across the mixing
oped Taylor bubble (FDTB) flow, depicted in Fig. 3, and devel- zone at the front of the liquid slug is given by
冉 冊
oping Taylor bubble (DTB) flow, illustrated in Fig. 4. HLLSL
dp
As shown in Figs. 3 and 4, slug flow is characterized by the
presence of large cap bubbles (Taylor bubbles), which occupy dZ
=
LSU 共uLLS + ⱍuLTBⱍ兲共uT − uLLS兲, . . . . . . . . . . . . . (24)
Acc
most of the cross section available to flow. Between the Taylor
bubble and the pipe walls, liquid flows downward as a falling film. in which, for the FDTB flow condition,  ⳱L TB /L SU , and
Liquid slugs containing small gas bubbles, which bridge the entire m T B ⳱ G ; for the DTB flow condition,  ⳱L d T B /L d S U ,
cross-sectional area, separate successive Taylor bubbles. Conse- mTB⳱LHLdTB+G (1 − HLdTB) and uLTB⳱uLdTB.
quently, the hydrodynamic parameters that describe this flow be- The selection of which parameters are used is based on which
havior are required to calculate the pressure drop in slug flow. flow condition exists, as explained in the previous paragraph.
When the calculated bubble cap length is less than the calcu- The average liquid holdup for the entire slug unit for either
lated Taylor bubble length for existing flow parameters at a point developed or developing Taylor bubbles can be calculated by
in the annulus, FDTB (see Fig. 3) is considered, and a constant
film thickness may be assumed. On the other hand, when the uSG + 共1 − HLLS兲共uT − uGLS兲
HLSU = 1 − . . . . . . . . . . . . . . . . . . (25)
calculated length of the Taylor bubble cap exceeds that calculated uT
for an FDTB for the same in-situ flow parameters, a DTB (see Fig.
4) is considered, and the film thickness varies continuously along Drillstring-Bubble Flow Model. For downward flow, buoyancy
the bubble zone. opposes the flow of the gas phase. According to Hasan,7 Eq. 14
冋 册
vertical lines at the extreme right of Fig. 6. As expected, in the
LTB LLS annulus churn flow (CH) occurred at surface, slug flow (SL) ex-
HLSU = 1 − 共1 − HLTB兲 + 共1 − HLLS兲 . . . . . . . . . . . . (29)
LSU LSU tends to 730 m (2,395 ft) from the surface, bubble flow (B) oc-
curred in almost 80% of the annulus, and dispersed bubble flow
Akagawa and Sakaguchi30 show that the average volume fraction of (DB) occurred at the bottom in the casing and drill collar annulus.
gas in the liquid slug (␣LLS/LSU) is approximately equal to 0.1 when On the other hand, only dispersed bubble flow was predicted to
uSG > 0.4 m/sec and equal to 0.25 uSG for lower superficial gas occur in the drillstring.
velocities. Hasan7 validated that this approximation may be extended
to downward slug flow in pipes. Thus, applying Eq. 27 for the Full-Scale Data Validation. The data obtained by Lopes31 from
gas void fraction in the liquid slug and knowing that LLS≈16DIT,13 an two experiments performed in a full-scale well located at Louisi-
equation for the slug unit length may be obtained. Thus:
160DITuSG
LSU = for uSG ⬎ 0.4 mⲐsec, . . . . . . . . . . . . . . . . . (30a)
C0um − u⬁
64DIT
LSU = for uSG ⱕ 0.4 mⲐsec. . . . . . . . . . . . . . . . . . . (30b)
C0um − u⬁
Considering FDTB flow and knowing that LLS/LSU+⳱1, the
gravitational and frictional components can be calculated with Eqs.
21 through 23. The acceleration component in the drillstring is
small and may be either neglected or calculated with the approach
suggested for bubble flow (Eq. 20).
Fig. 7—The first experiment’s wellbore pressure. Fig. 8—The second experiment’s well pressure.
Fig. 11—In-situ liquid velocity comparison. Fig. 12—In-situ gas velocity comparison.
冋 册
Theory and Application,” SPEPE (May 1991) 171.
12. Ansari, A.M. et al.: “A Comprehensive Mechanistic Model for Upward gDep 共L − G兲 0.5
uT = 1.2um + 0.345 . . . . . . . . . . . . . . . . . (A-1)
Two-Phase Flow in Wellbores,” SPEPF (May 1994) 143. L
共uT − uLLS兲 HLLS = 共uT + ⱍuLTBⱍ兲HLTB. . . . . . . . . . . . . . . . . . . . . (A-2) A net volumetric flow rate across the plot B-B in Fig. 4 gives
ⱍ ⱍ
uLRTB = uT + uL TB = 公2gl , . . . . . . . . . . . . . . . . . . . . . . . . . . (B-5)
behave analogous to fully developed bubble flow conditions, from
Eq. 14, the in-situ gas velocity in the liquid slug is
in which l⳱the distance from the nose of the bubble to the point
uGLS = C0um + HLn LSu⬁. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-4)
of interest. Combining Eqs. B-4 and B-5 and taking l⳱Lc , the
Instead of assuming that the liquid holdup in the liquid slug is length of the Taylor bubble cap can be calculated by
冉 冊
constant, different methods12,32−34 were implemented to calculate uG TBAG TB u A 2
it as the slug flow progresses. For the actual well conditions used 1 m p
LC = uT + − . . . . . . . . . . . . . . . . . . . . (B-6)
to validate the model, better performance was obtained with the 2g AL TB AL TB
approach followed by Ansari et al.12:
From the comparison of LC and LTB, if LCⱖLTB, the slug flow is
uSG in its developing stage; therefore, different hydrodynamic param-
HLLS = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-5)
0.425 + 2.65um eter values (LdTB, uLdTB, HLdTB, LdSU)are required for pressure-drop
estimations.
Considering that the free-falling film thickness in the Taylor The gas volume in the developing Taylor bubble is given by
bubble reaches a terminal constant value, Fernandes et al.26 proved
L dTB
that the film thickness equation proposed by Wallis25 can be used
to estimate its value. Thus, for turbulent flow, VG dTB = 兰A G dTB dl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-7)
冋 册冉 ⱍ ⱍ 冊
0
L2 1Ⲑ3 4L uLTB ␦ 2Ⲑ3
␦ = 0.0682 . . . . . . . . . . . (A-6) Because liquid holdup can be expressed as a function of areas,
g共L − G兲L L combining Eqs. A-2 and B-5, the area of gas in the DTB is
冋 册
Based on the annular slug flow geometry, Caetano17 proposed the 共uT − uL LS兲 HL LS
following expression for liquid holdup in the Taylor bubble zone. AG dTB = 1− Ap . . . . . . . . . . . . . . . . . . . . (B-8)
4␦ 共DIC − ␦兲
公2gl
HLTB = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-7) Because the gas volume in a developing slug unit is equal to the
DIC − DOT
2
gas volume in the DTB plus the gas volume in the liquid slug, the
Fernandes et al.26 and Taitel et al.13 have shown that in a wide gas volume in the DTB can be also expressed as
冉 冊
range of flow conditions, the slug length in upward two-phase flow
in pipes has a fairly constant value equal to 16 pipe diameters. L dTB+ L LS L LS
VG dTB = uSG Ap − u GLS Ap 共1 − HL LS兲 . . . . . (B-9)
Later, Caetano17 confirmed this for annuli with the hydraulic di- uT uT
ameter concept. Hence, the liquid slug length is given by
Substituting Eqs. B-9 and B-8 into B-7 and performing the inte-
LLS = 16Dh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8) gration, an implicit equation for the developing length of the Tay-
lor bubble is obtained.
If fully developed slug flow is considered, the numerical solution
of Eqs. A-1 through A-8 gives the necessary hydrodynamic pa-
rameters (LLS, LTB, uT, uLLS,uLTB, HLLS, HLTB) for pressure-drop
calculations.
冉 1−
u SG
uT
冊 L dTB −
2共uT − uL LS兲 HL LS
公2g
1Ⲑ2
LdTB
␦N = 冋 3uLTBALTBL
DICg共L − G兲
册 1Ⲑ3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-1) Because the gas volume in the DTB also can be expressed as
VGdTB⳱AGdTBLdTB, solving Eq. B-7 after substituting Eq. B-8, the
Once the film thickness decreases to the Nusselt film thickness, the average liquid holdup in the DTB is
resulting Taylor bubble area in an annulus is17 2共uT − uL LS兲 HL LS
HL dTB = . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-12)
AGTB = 0.25关共DIC − 2␦N兲 − 兴. . . . . . . . . . . . . . . . . . . . . (B-2) 公2gL dTB
2 2
DOT
兰 兰 v dp = 0 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-1) E+00 ⳱
1 psi × 6.894 757 kPa
udu +
gc *Conversion factor is exact.
0 pup
冉 冊
for more than 20 years. His primary research interests are drilling
u2n 共1 − wG兲 wG zRT pbh related, including well control and design, deep drilling, bit
+ 共pbh − pup兲 + ln = 0 . . . . . . . . . . (C-4) performance, and wellbore integrity. Smith holds a BS degree
gc L MG pup
in electrical engineering from the U. of Texas at Austin and MS
For steady-state flow conditions that use the continuity equation, and PhD degrees in petroleum engineering from Louisiana
the nozzle velocity may be express by State U. He was an SPE Distinguished Lecturer on deep drilling
from 1999−2000 and is currently serving on the SPE R&D Advi-
qG G + qL L sory Committee. Jeremy K. Edwards is currently an assistant
un = v . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-5) professor of petroleum engineering at Louisiana State U.
An
e-mail: [email protected]. He has teaching and research inter-
Eqs. C-4 and C-5 can be solved numerically to obtain the pressure ests in production engineering and multiphase flow in pipes.
upstream of the nozzle pup by knowing the corresponding bottom- Edwards holds BS, MS, and PhD degrees in mechanical engi-
hole pressure pbh. neering from the U. of Tulsa.