Sie sind auf Seite 1von 16

THIS IS A PREPRINT - SUBJECT TO CORRECTION

A MECHANISTIC MODEL
FOR MULTIPHASE FLOW IN PIPES

Nicholas Petalas
Stanford University

Khalid Aziz
Stanford University

PUBLICATION RIGHTS RESERVED


TH
PAPER NO. 98-39. THIS PAPER IS TO BE PRESENTED AT THE 49 ANNUAL TECHNICAL
MEETING OF THE PETROLEUM SOCIETY OF THE CANADIAN INSTITUTE OF MINING,
METALLURGY AND PETROLEUM HELD IN CALGARY, ALBERTA, CANADA ON JUNE 8-10,
1998.

facilities. Mechanistic models, on the other hand, are based


ABSTRACT on fundamental laws and thus can offer more accurate mod-
eling of the geometric and fluid property variations.
Mechanistic models for multiphase flow calculations can
improve our ability to predict pressure drop and holdup in All of the models presented in the literature are either in-
pipes especially in situations that cannot easily be modeled complete2,3, in that they only consider flow pattern determi-
in a laboratory and for which reliable empirical correlations nation, or are limited in their applicability to only some pipe
are not available. In this paper, a new mechanistic model, inclinations4,5. A preliminary version6 of the model proposed
applicable to all pipe geometries and fluid properties is pre- here that overcomes these limitations was presented in 1996.
sented. New empirical correlations are proposed for liq- For most of the flow patterns observed, one or more em-
uid/wall and liquid/gas interfacial friction in stratified flow, pirical closure relationships are required even when a
for the liquid fraction entrained and the interfacial friction in mechanistic approach is used. Where correlations available
annular-mist flow, and for the distribution coefficient used in in the literature are inadequate for use in such models, new
the determination of holdup in intermittent flow. correlations must be developed. In order to be able to achieve
this, access to reliable experimental data is important.
INTRODUCTION
A large amount of experimental data has been collected
Empirical models often prove inadequate in that they are through the use of a Multiphase Flow Database7 developed at
limited by the range of data on which they were based and, Stanford University. The database presently contains over
generally, cannot be used with confidence in all types of flu- 20,000 laboratory measurements and approximately 1800
ids and conditions encountered in oil and gas fields. Fur- measurements from actual wells. Based on subsets of these
thermore, many such models exhibit large discontinuities1 at data, the previously proposed model6 included a detailed
the flow pattern transitions and this can lead to convergence investigation of the annular-mist flow regime and new corre-
problems when these models are used for the simultaneous lations for the liquid fraction entrained and for interfacial
simulation of petroleum reservoirs and associated production friction. This model has since been refined based on addi-
tional investigations of the stratified and intermittent flow  dp  g
regimes, and is the subject of this paper. − AL   − τ wL S L + τi Si − ρ L AL sin θ = 0 Eq. 4
 dL  gc
FLOW PATTERN DETERMINATION  dp  g
− AG   − τ wG SG − τi Si − ρG AG sin θ = 0 Eq. 5
 dL  gc
The procedure for flow pattern determination begins with
the assumption that a particular flow pattern exists and is These can then be combined, eliminating the pressure gra-
followed by an examination of various criteria that establish dient terms, and expressed in terms of the dimensionless liq-
the stability of the flow regime. If the regime is shown to be ~
uid height, hL = hL D , using the geometric relationships
unstable, a new flow pattern is assumed and the procedure is
outlined by Taitel and Dukler. The shear stresses are given
repeated. Figure 1 shows flow pattern transitions based on
by the following relationships:
the superficial velocities of the phases where the stability
criteria (transition boundaries S1, S2, etc.) considered in this f ρ V2
τ wG = G G G Eq. 6
model are sketched. The procedure for flow pattern determi- 2gc
nation is illustrated in Figure 3, where it is seen that the ex-
f L ρ LVL2
amination of the dispersed bubble flow regime is the first to τ wL = Eq. 7
be considered. 2 gc
f i ρGVi Vi
Dispersed Bubble Flow τi = Eq. 8
2gc
The dispersed bubble flow region is bounded by two crite- These definitions differ from those proposed by Taitel and
ria. The first is based on the transition to slug flow proposed Dukler mainly in that pipe roughness is not ignored when
by Barnea2 where a transition from intermittent flow occurs determining the friction factors. In addition, the definition for
when the liquid fraction in the slug is less than the value as- the gas/liquid interfacial shear does not require the assump-
sociated with the maximum volumetric packing density of tion that the gas phase moves faster than the liquid phase
the dispersed bubbles (0.52): (thereby assuming the shear stress to be based on the gas
E Ls < 0.48 Eq. 1 phase velocity alone). A new approach is also used when
The same mechanism is adopted in this model with the determining the liquid /wall interfacial friction factor, f L , as
exception that the liquid volume fraction in the slug is not discussed below.
obtained from the correlation proposed by Barnea, but from The friction factor at the gas/wall interface (Eq. 6) is de-
the Gregory et al.8 correlation given below: termined from an approach similar to that used in single-
1 phase flow with the actual pipe roughness and the following
E Ls = 1.39
 V  Eq. 2 definition of Reynolds number:
1+  m  D ρ V
 8 .66  Re G = G G G Eq. 9
where V m is expressed in meters/sec. This transition is µG
shown as line I1 in Figure 1. where DG is the hydraulic diameter of the gas phase.
A transition from dispersed bubble flow to froth flow can For the liquid/wall interface (Eq. 7) it was found that the
also occur when the maximum volumetric packing density of use of an approach similar to single phase flow is not appro-
the dispersed gas bubbles is exceeded (line D1 in Figure 1): priate and, instead, the following empirical relationship is
V used for calculating the wall/liquid interfacial friction factor:
CG = SG > 0.52 Eq. 3
Vm f L = 0.452 f SL0.731 Eq. 10
The friction factor based on the superficial velocity, f SL , is
If the criteria given by Eq. 1 and Eq. 3 are not satisfied,
dispersed bubble flow is not possible and the possibility of obtained from standard methods using the pipe roughness
stratified flow is examined next. and the following definition of Reynolds number:
Dρ LVSL
Re SL = Eq. 11
Stratified Flow µL

Determining the stability of the stratified flow regime re- During downhill flow, it is possible for the dense phase to
quires the calculation of the liquid height, which can be ob- flow faster than the lighter phase. For this reason, the defini-
tained by writing the momentum balance equations for the tion of the gas/liquid interfacial shear (Eq. 8) is based on the
gas and the liquid phases as was done by Taitel and Dukler3: quantity Vi = VG − VL , which can become negative under

2
certain conditions. The interfacial friction factor is calculated will form on the liquid surface once the gas velocity is in-
from the empirical relationship: creased beyond (line S2 in Figure 1):
 ρ Dg  4µ L (ρ L − ρG ) g cos θ
f i = (0.004 + 0.5 × 10− 6 Re SL ) FrL1.335  L 2  Eq. 12 VG ≥ Eq. 15
 ρGVG  sρ LρGVL
VL
The Froude number is defined as FrL = . The sheltering coefficient, s, is given as 0.01. In Xiao et
gh L al. 5 and in the present model, s is taken as 0.06, based on a
study by Andritsos9. This value is said to be more suitable,
Once the liquid height is known, the stability of the strati- especially for gas flow with high viscosity liquids.
fied flow pattern can be determined. The approach used by
Taitel and Dukler, which uses an extension of the Kevin- During downflow, waves can develop on the flowing liq-
Helmholtz wave stability theory, is also used in this model. uid independent of interfacial shear from the gas flow. The
This attempts to predict the gas velocity at which waves on criterion for the appearance of waves can be expressed in
the liquid surface are large enough to bridge the pipe: terms of a critical Froude number which varies from 0.5 to
2.2 depending on roughness and whether the flow is laminar
 h 
VG = 1 − L 
(ρ L − ρG )gAG cos θ or turbulent. Barnea2 recommends a limiting value of 1.5 for
 D
ρG
dAL Eq. 13 the critical Froude number. When interfacial effects are con-
dhL sidered in the calculation of the liquid height, this limit can
This transition is represented by line S1 in Figure 1. The predict smooth flow even at high liquid rates where the flow
transition criterion expressed by Eq. 13 does not apply when is known to be wavy. Reducing the limit to 1.4 appears to
θ = −90° . Experimental data, however, suggest that strati- resolve this problem. Thus the transition from stratified
fied-like annular flow can occur at such inclinations. In order smooth to wavy flow based on this mechanism is (line S3 in
to account for this, when cos θ ≤ 0.02 , cos θ = 0.02 is substi- Figure 1):
tuted in Eq. 13. VL
Fr = > 1. 4 Eq. 16
ghL
At steep downward inclinations, Barnea proposes a
mechanism whereby stratified flow can change to annular,
even at relatively low gas rates. This occurs when the liquid Annular-Mist Flow
height is small and the liquid velocity is high. Liquid droplets
are sheared off from the wavy interface and deposited on the The treatment of the annular-mist flow regime is similar to
upper pipe wall, eventually developing into an annular film. the approach used for stratified flow and is based on the
The condition for this type of transition to annular flow, work of Taitel and Dukler3 and Oliemans et al.10. The model
shown as line S4 in Figure 1, is given as2: is based on the assumption of a constant film thickness and
~ accounts for the entrainment of the liquid in the gas core.
gD(1 − hL ) cos θ Slip between the liquid droplets in the gas core and the gas
VL > Eq. 14
fL phase is not accounted for. Momentum balance on the liquid
It should be noted that f L is calculated as per Eq. 10; not the film and gas core with liquid droplets yields:
 dp  g
definition proposed by Barnea. − A f   − τ wL S L + τi Si − ρ L Af sin θ = 0 Eq. 17
 
dL g c
At the higher upward pipe inclinations the predicted liquid
height has the tendency, given the transition criterion of Eq.  dp  g
− Ac   − τi Si − ρc Ac sin θ = 0 Eq. 18
13, to predict stratified flow where none is known to exist.  dL  gc
For this reason, and to ensure continuity between flow pat-
tern transitions, the present model limits stratified flow to The geometric parameters can be expressed in terms of the
~
horizontal and downhill angles only. This approach is also dimensionless liquid film thickness, δL = δ L D , and the
supported by the fact that stratified flow is only observed for liquid fraction entrained, FE. The shear stresses are given by:
small upward angles in large-diameter pipes.
f f ρ LV f2
τ wL = Eq. 19
Thus, when θ ≤ 0 , if the gas phase velocity is less than 2 gc
( )
the transitional value given by Eq. 13 and the liquid phase
velocity is less than that of Eq. 14, the flow pattern is strati- f iρc Vc − V f Vc − V f
τi = Eq. 20
fied. Although no distinction is made in this model between 2 gc
stratified smooth and stratified wavy flow for the purposes of
determining pressure drop and liquid volume fraction, the The friction factor for the liquid film is computed using
transition between these two regimes is considered in flow any of the standard correlations with the pipe roughness and
pattern predictions. Taitel and Dukler propose that waves the film Reynolds number as expressed by:

3
D f ρLV f Bubble Flow
Re f = Eq. 21
µL When the liquid fraction in the slug (Eq. 2) is greater than
0.48 and the stratified, annular and dispersed bubble flow
In order to solve Eq. 17 and Eq. 18, two additional quanti-
regimes have been eliminated, the flow will either be inter-
ties need to be determined: the interfacial friction factor, f i ,
mittent, froth or bubble flow.
and the liquid fraction entrained, FE. These are determined
empirically and are given by: Bubble flow is encountered in steeply inclined pipes and is
0.2 characterized by a continuous liquid phase containing a dis-
FE V 
= 0.735N B0.074  SG  Eq. 22 persed phase of mostly spherical gas bubbles. It can exist if
1 − FE  VSL  both of the following conditions are satisfied:
0.085
fi  σ  1. The Taylor bubble velocity exceeds the bubble veloc-
= 0.24  Re 0f.305 Eq. 23 ity. This is satisfied in large diameter pipes (Taitel et
 ρcVc Dc 
2
fc
al.11) when
Where the dimensionless number, N B , is defined as 1

 (ρ − ρ )σ  2

µV ρ
2 2 D > 19 L 2 G  Eq. 28
NB =  ρL g
L SG G
Eq. 24 
σρ 2
L
2. The angle of inclination is large enough to prevent
Having determined the liquid film thickness, it is now migration of bubbles to the top wall of the pipe (Ba-
possible to test for the presence of annular-mist flow. Ba- rnea et al.12):
rnea2 presents a model for the transition from annular flow
3  C γ2 
based on two conditions. The same mechanisms are used in cos θ ≤ Vb2  l  Eq. 29
the present model, although they are revised to account for 4 2  gdb 
the differences in the modeling assumptions.
The lift coefficient, C l , ranges from 0.4 to 1.2, the bubble
The first of the transitions proposed by Barnea is based on
the observation that the minimum interfacial shear stress is distortion (from spherical) coefficient, γ, ranges from 1.1 to
associated with a change in the direction of the velocity pro- 1.5 and a bubble size, d b , between 4 and 10mm is recom-
file in the film. When the velocity profile becomes negative mended. For this model, C l is taken as 0.8, γ as 1.3 and a
stable annular flow cannot be maintained and the transition bubble diameter of 7 mm is used. The bubble swarm rise
to intermittent flow occurs. This transition mechanism is velocity in a stagnant liquid, V b , is given by13:
only relevant during uphill flow. The minimum shear stress 1
∂τ
condition may be determined by setting ~ i = 0 .  g (ρ L − ρG )σ  4
Vb = 1.41  sin θ
Eq. 30
∂δ L  ρ2L 
2ff
ρ L VSL2 (1 − FE )
2
E3 1 − 3 E
= f 32 f
( ) When both of the above conditions are satisfied, bubble
Eq. 25
ρ L − ρ c gD sin θ 2 − 2 Ef flow is observed even at low liquid rates where turbulence
does not cause bubble breakup. The transition to bubble flow
The liquid fraction in the film is given by: from intermittent flow as suggested by Taitel et al.11 occurs
Ef =
Af
A
~ ~
(
= 4 δL 1 − δL ) Eq. 26 when the gas void fraction (during slug flow) drops below
the critical value of 0.25 (line I3 in Figure 1). The calculation
of the gas void fraction for slug flow is discussed below.
Eq. 25 can be solved using an iterative procedure to obtain
the liquid film height at which the minimum shear stress oc-
~ Intermittent Flow
curs, δ L (line A1 in Figure 1).
min
The intermittent flow model used here includes the slug
The second mechanism proposed by Barnea for annular and elongated bubble flow patterns. It is characterized by
flow instability occurs when the supply of liquid in the film alternating slugs of liquid trailed by long bubbles of gas. The
is sufficient to cause blockage of the gas core by bridging the liquid slug may contain dispersed bubbles and the gas bub-
pipe. This is said to take place when the in situ volume frac- bles have a liquid film below them.
tion of liquid exceeds one half of the value associated with
the maximum volumetric packing density of uniformly sized As stated above, a transition from intermittent flow occurs
gas bubbles (0.52). Hence, the transition from annular flow when the liquid fraction in the slug exceeds the value associ-
occurs when (line A2 in Figure 1): ated with the maximum volumetric packing density of the
dispersed bubbles (Eq. 1, line I1 in Figure 1). The same
E L ≥ 12 (1 − 0.52) or E L ≥ 0.24 Eq. 27
mechanism can occur at low liquid rates when sufficient liq-

4
uid is not available for slug formation. To account for this Bendiksen15 gives the elongated bubble drift velocity at
situation, an additional transition criterion is imposed (line I4 high Reynolds numbers as:
in Figure 1). Vd = Vdh cos θ + Vdv sin θ
∞ ∞ ∞
Eq. 38
E L ≤ 0.24 Eq. 31
The drift velocity of elongated bubbles in a horizontal sys-
The liquid volume fraction calculated for slug flow is dis- tem at high Reynolds numbers is given by Weber19 as:
cussed below. Although it is not treated as a separate flow
 1.76  gD (ρ L − ρG )
pattern for the purposes of phase volume fractions and pres- Vdh∞ = 0.54 − 0.56  Eq. 39
sure drop determination, the elongated bubble flow regime is  Bo  ρL
defined here as the portion of intermittent flow for which the
The Bond number, Bo = L
(ρ − ρ G ) gD 2
liquid slug contains no dispersed bubbles of gas. This condi- σ
tion is arbitrarily represented in the model by the region
where E L ≥ 0.90 (line I2 in Figure 1). The drift velocity of elongated bubbles in a vertical system
s
at high Reynolds numbers is obtained from a modified form
The liquid volume fraction may be determined by writing of the Wallis20 correlation
gD(ρ L − ρG )
an overall liquid mass balance over a slug-bubble unit. As-
suming that the flow is incompressible and a uniform depth ∞
(
Vdv = 0.345 1 − e −β ) ρL
Eq. 40
for the liquid film14:
E V + VGdb (1 − E Ls ) − VSG The coefficient, β, is given by:
E L = Ls t Eq. 32 β = Bo e (3.278 −1.424 lnBo ) Eq. 41
Vt
VGdb represents the velocity of the dispersed bubbles, V t is Finally, the volume fraction liquid (Eq. 32) can be calcu-
the translational velocity of the slug, and E Ls is the volume lated once the velocity of the dispersed bubbles in the liquid
slug is obtained from:
fraction liquid in the slug body (Eq. 2). All of these quanti-
ties need to be determined from empirical correlations. VGdb = C0Vm + Vb Eq. 42

The translational velocity of the elongated bubbles is C 0 in Eq. 42 is determined from Eq. 34 and the rise ve-
given by Bendiksen15 as: locity of the dispersed bubbles is calculated from21:
Vt = C0Vm + Vd Eq. 33 1

 g (ρL − ρG )σ  4

The parameter C 0 is a distribution coefficient related to the Vb = 1.53  sin θ Eq. 43


 ρ2L 
velocity and concentration profiles in dispersed systems and
under special conditions is related to the inverse of the The empirical nature of the correlations used for determin-
Bankoff K factor. Zuber and Findlay16 have confirmed em- ing the liquid volume fraction (Eq. 32) requires that certain
pirically its application to other flow patterns, including slug limits be imposed on the calculated values. The first such
and annular flow. Nicklin et al.17, in their study of the rise condition affects Eq. 42 where it is possible, under certain
velocity of Taylor bubbles, have found that for liquid Rey- downflow conditions for the calculated value of VGdb to be-
nolds numbers greater than 8,000, C 0 =1.2, whereas at lower
come negative. In these situations, VGdb is set to zero. In
Reynolds numbers C 0 approached 2.0. It is generally taken
other situations, it is possible for Eq. 32 to yield values for
to be 1.2, although for this analysis it is determined from the EL that are greater than 1.0. In these cases, EL is set equal to
following empirically derived correlation:
CL .
C0 = (1.64 + 0.12 sin θ) Re −mL0.031 Eq. 34
The modified Reynolds number in Eq. 34 is based on the When none of the transition criteria listed above are met,
mixture velocity and liquid properties: the flow pattern is designated as "Froth" implying a transi-
ρV D tional state between the other flow regimes.
Re mL = L m Eq. 35
µL
CALCULATION OF PRESSURE DROP AND LIQ-
UID VOLUME FRACTION
The elongated bubble drift velocity, Vd , can be calculated
from the Zuboski18 correlation: Once the flow pattern has been determined, the calculation
Vd = f mVd ∞
Eq. 36 of the pressure drop and phase volume fractions can be de-
termined as detailed below.
Where f m = 0.316 Re ∞ for f m < 1 , otherwise f m = 1 , and
ρ LV d ∞ D
Re ∞ = Eq. 37
2µ L

5
Dispersed Bubble Flow The pressure gradient is obtained from either Eq. 17 or Eq.
18.
The calculation of the liquid volume fraction in dispersed
bubble flow follows the procedure used for the dispersed Bubble Flow
bubbles in the slug in intermittent flow. Thus,
VGdb = C0Vm + Vb Eq. 44 The volumetric gas fraction during bubble flow is ob-
tained from:
C 0 is determined from the empirical correlation given in V
EG = SG Eq. 53
Eq. 34, and the rise velocity of the dispersed bubbles, Vb , is Vt
calculated from Eq. 43. The volume fraction is then obtained The translational bubble velocity is defined as:
from: Vt = C0Vm + Vb Eq. 54
V
E L = 1 − SG Eq. 45 Zuber and Findlay16 have shown that the distribution pa-
VGdb
rameter, C 0 , for dispersed systems can range from 1.0 to 1.5,
If VGdb ≤ 0 , the volume fraction is then obtained from:
the higher values being associated with high bubble concen-
VSG trations and high velocities at the center line (laminar flow).
E L = 1− Eq. 46
C0Vm When the flow is turbulent and the velocity and concentra-
tion profiles are flat C 0 approaches 1.0. For the present
In cases where the value of EL calculated by Eq. 45 or Eq.
method, C 0 is taken as 1.2. The bubble swarm rise velocity
46 is greater than 1.0, EL is set equal to CL .
in a stagnant liquid, Vb , is given by Eq. 30. The value of
Once the liquid volume fraction is known, the pressure EG thus obtained, is limited to the range:
gradient is determined from:
VSG
 dp  2 f V ρ
2
g 0 ≤ EG ≤ CG = Eq. 55
−   = m m m + ρm sin θ Eq. 47 Vm
 
dL g c D g c

The friction factor, f m , is obtained from standard methods The pressure gradient is given by:
 dp  2 f V ρ
2
using the pipe roughness and the following Reynolds num- g
−   = mL m m + ρm sin θ Eq. 56
ber:  dL  gc D gc
Dρ mVm
Re m = Eq. 48 The friction factor, f mL , is obtained from standard meth-
µm
ods using the pipe roughness and the following Reynolds
The mixture density and viscosity are calculated in the number:
usual way: Dρ LVm
Re mL = Eq. 57
ρm = ELρ L + EGρG Eq. 49 µL
µm = E Lµ L + EGµG Eq. 50
Intermittent Flow
Stratified Flow
The calculation of the volume fraction liquid for intermit-
The liquid volume fraction during stratified flow is sim- tent flow has already been described (Eq. 32). The pressure
ply, from geometric considerations: drop may be obtained by writing the momentum balance
A over a slug-bubble unit:
EL = L Eq. 51
A  dp  g
−   = ρm sin θ
 
dL g c
The pressure gradient is obtained from either Eq. 4 or Eq.
5.   τ Ls πD  
 Ls    Eq. 58
 A 
+  
1
Annular-Mist Flow
Lu   τ Lf SLf + τGdb SGdb 
 + L f  
As is the case for stratified flow, the volume fraction liq-   A 
uid during annular-mist flow is determined from geometric
considerations once the liquid film thickness is known: Unfortunately, no reliable methods exist for the calcula-
~ VSG tion of the slug length, Ls , nor for the length of the bubble
E L = 1 − (1 − 2 δL )2 Eq. 52
VSG + FE VSL region, L f . Furthermore, although it is known that the fric-

6
tional pressure gradient in the gas bubble is normally small The friction factor, f m , is obtained from standard methods
compared to that in the liquid slug, no reliable method is using the pipe roughness and Reynolds number given in Eq.
available for calculating it. Xiao et al.5 have modeled the 48. The mixture density and viscosity are obtained from Eq.
bubble region by assuming it to be analogous to stratified 49 and Eq. 50, respectively.
flow. This treatment contradicts observations made in the
laboratory. In view of these uncertainties, the following sim- Froth Flow
ple approach is selected:
 dp  g  dp  Froth flow represents a transition zone between dispersed
−   = ρm sin θ + η  bubble flow and annular-mist flow and between slug flow
 dL  gc  dL  fr
SL
Eq. 59 and annular-mist. The approach used in this model is to in-
 dp 
+ (1 − η) 
terpolate between the appropriate boundary regimes in order
 dL  fr
AM
to determine the transition values of the in situ liquid volume
fraction and pressure drop. This involves a number of itera-
Here, the quantity η is an empirically determined weighting tive procedures in order to determine the superficial gas ve-
factor related to the ratio of the slug length to the total slug
locities at the dispersed bubble, annular-mist and slug transi-
L tions to froth. Once V SG at each transition is known, the vol-
unit length, s and is calculated from:
Lu ume fraction and pressure drop values at the transitions are
η = CL(0.75 − E )
L
Eq. 60 calculated and a log-log interpolation between these values is
with the condition that η ≤ 1.0 . made for each quantity.

The frictional pressure gradient for the slug portion is ob- RESULTS
tained from:
The model’s overall performance has been evaluated using
 dp  f V 2ρ
  = 2 mL m m Eq. 61 the following approaches.
 dL  fr
SL
gD
a) The behavior of the model was examined over a wide
The friction factor, f mL , is calculated from standard methods range of flow rates and fluid properties using three-
using the pipe roughness and the Reynolds number given by dimensional surface plots. This was done over the com-
Eq. 57. plete range of upward and downward pipe inclinations
and both, pressure gradient and volumetric liquid frac-
The term that still needs to be defined in Eq. 59 is the fric-
tion were analyzed.
tional pressure gradient calculated for annular-mist flow.
This is obtained by using the liquid fraction given by Eq. 32 b) Data were extracted from the Stanford Multiphase Flow
to determine the liquid film height, assuming the flow pattern Database for which pressure gradient, holdup and flow
to be annular-mist. Thus, from Eq. 52, pattern observations were available. This resulted in a
~ 1
δL = 1 − (1 − EL )
(FE VSL + VSG )  total of 5,951 measurements consisting of variations in
 Eq. 62 fluid properties, pipe diameters, and upward as well as
2  VSG  downward inclinations. The model was then compared
with these experimental observations.
The frictional pressure gradient based on annular-mist
flow is then calculated from: c) Finally, these same experimental data were analyzed
 dp  4τ using a number of existing methods and the results com-
  = wL Eq. 63 pared to the new model.
 dL  fr
AM
D
Flow pattern maps and three dimensional surface plots are
The shear stress τ wL is obtained from Eq. 19. Note that shown in Figures 4 to 27 for an air/water system at standard
Eq. 63 is obtained by adding Eq. 17 to Eq. 18 (to eliminate conditions and for an oil/gas system at reservoir conditions
the interfacial component) and removing the hydrostatic (see Table 1). The pipe inclinations shown include horizon-
term. tal, 10° upward, vertical upflow (+90°), and 10° downward.
Each plot covers a range of superficial gas velocity of 0.01
When the calculated film height (Eq. 62) is less than ft/sec to 500 ft/sec and of superficial liquid velocity from
1 × 10 −4 , a simple homogeneous model with slip is used, 0.01 ft/sec to 100 ft/sec.
where:
The coloring of the three dimensional plots is consistent
 dp  2 f V 2ρ
  = m m m Eq. 64 with that of the flow pattern maps so as to show the location
 dL  fr
AM
gc D of the flow pattern transitions. The superficial velocity axes
appear on the X-Y plane with the appropriate parameter
(pressure gradient or liquid volume fraction) plotted on the

7
vertical axis. Over one hundred different gas and liquid rates always result in the “statistically best” correlation being
have been calculated per plot, equating to over 10,000 calcu- adopted.
lated points. This was done to insure that the model behaves
Solving the momentum balance equations in stratified
predictably over the entire practical range of flow rates and
flow (Eq. 4 and Eq. 5) and annular-mist flow (Eq. 17 and Eq.
pipe inclinations. The effect of fluid properties on flow pat-
18) poses certain problems because of the presence of multi-
tern transitions can easily be seen in these plots. Although
some discontinuities in pressure gradient and liquid volume ple roots. Primarily, it is necessary to determine which of the
fraction are present at the transitions from stratified flow, roots is the physical one. Xiao et al.5 assume that it is the
lowest root. There does not however seem to exist a clear
overall, the model exhibits generally smooth behavior and
rationale for this assumption. Figure 2 shows the variation of
consistent trends between flow patterns.
liquid height versus superficial gas velocity at a fixed super-
The distribution of experimental data points according to ficial liquid rate for an air/water system at standard condi-
angle of inclination is shown in Figure 28. The convention tions (2.047” I.D. pipe, at 2° upward inclination). It can be
used in expressing the range of inclinations in this figure is seen that in the range where multiple roots occur ( V SG =
that the lower number in the specified range is inclusive, 57.144 to 112.67 ft/sec), the selection of the lowest root ver-
whereas the higher number is not. Thus, “0° to 10°” implies sus the highest root results in significant changes to the pre-
the range where 0° ≤ θ < 10°. Although more than half of dicted liquid height. This large variation in liquid height
the data fall in the range of 0° to 10° inclination, it can be based on different roots suggests that the selection of one
seen that there is also a fair sampling (1,164 points, or ~20%) root over the other simply affects the value of V SG at which a
of downhill data.
transition to another flow pattern occurs. In order to prevent
The model predictions for liquid volume fraction are plot- discontinuities, it is important to ensure that, whether the
ted against the experimental measurements in Figure 29. lowest or the highest root is used, the same root is used in all
Figure 30 shows a similar plot for the pressure gradient cal- the calculations. In the present model, the lowest root is se-
culations. The model is able to predict the in situ liquid vol- lected for the ensuing calculations. Having to determine all
ume fraction to within an accuracy of 15% in 3,663 of the of the roots presents a further complication as can be seen by
5,951 cases (62%). This is shown in Figure 31, where these examining Figure 2 when V SG =58.0 ft/sec. It is seen that
numbers are compared with other methods. The pressure three roots exist under these conditions: 0.0586, 0.0646, and
gradient is predicted to the same accuracy for 2,567 cases 0.508. The lower two roots can only be detected by investi-
(43%), as shown in Figure 31. ~
gating values of hL within less than 0.005 of each other. For
Some of the methods with which comparisons are being the range of validity of the solutions (0 to 1.0) this could re-
made are limited to specific ranges of pipe inclination. To quire over 200 iterations. For certain combinations of fluid
make the comparisons of Figure 31 more meaningful, they rates and pipe inclination the required resolution of liquid
are shown grouped according to downward, near-horizontal height or liquid film thickness is of the order of 0.001, which
and upward pipe inclination in Figure 32, for the volume can require over 10,000 iterations.
fraction liquid, and in Figure 33, for the pressure gradient.
A final note is warranted regarding discontinuities that
COMMENTS arise from friction factor calculations when the flow changes
from laminar to turbulent. The traditional approach is to use
The empirical correlations introduced in this paper, the laminar flow friction factor when the Reynolds number is
namely: less than 2,000. In the present model, the approach followed
• Eq. 10 and Eq. 12 for stratified flow, is to use the turbulent friction factor wherever it is greater
• Eq. 22 and Eq. 23 for annular-mist flow, than the laminar flow value. This results in a smoother corre-
• Eq. 34, Eq. 40 and Eq. 60 for intermittent flow, lation and since the application of single-phase friction factor
were developed in accordance with the following two objec- correlations to multiphase flow situations is, at best, arbi-
tives. trary, the approach is believed to be reasonable.

a) Accurately reflect the expected/observed behavior of the CONCLUSIONS


quantity being estimated.
A new mechanistic model has been presented which is ap-
b) Ensure that the correlation’s behavior results in smooth
plicable to all conditions commonly encountered in the petro-
transitions between adjacent flow patterns.
leum industry. The model incorporates roughness effects as
The application of both of these criteria involved relying well as liquid entrainment, both of which are not accounted
on statistical analysis of differences between predicted and for by previous models. The model has undergone extensive
experimental values as well as the study of numerous surface testing and has proven to be more robust than existing mod-
plots (such as those shown in Figures 4 to 27). This did not els and is applicable over a more extensive range of condi-
tions.

8
The empirical correlations that are necessary within the db relating to the dispersed bubbles
model can only be improved with accurate and consistent G relating to the gas phase
data over a wide range of conditions of commercial interest. i relating to the gas/liquid interface
To this end, efforts are in progress to obtain additional data L relating to the liquid phase
in order to expand the Stanford Multiphase Flow Database. m relating to the mixture
SG based on superficial gas velocity
Further testing of the new model will be undertaken which
s relating to the liquid slug
will include as much actual field data as can be obtained. It is
SL based on superficial liquid velocity
expected that results from these tests will more adequately
wL relating to the wall-liquid interface
demonstrate the ability of the model to predict reasonably
wG relating to the wall-gas interface
accurate pressure drops and holdup under operating condi-
tions.
REFERENCES
ACKNOWLEDGEMENTS
1. Aziz, K. and Petalas, N., “New PC-Based Software
The work described in this paper has been made possible
for Multiphase Flow Calculations,” SPE 28249, SPE
through the support of the Reservoir Simulation Industrial
Petroleum Computer Conference, Dallas, 31 July-3
Affiliates Program at Stanford University (SUPRI-B) and the
August, 1994.
Stanford Project on the Productivity and Injectivity of Hori-
zontal Wells (SUPRI-HW). Portions of this research have 2. Barnea, D. “A Unified Model for Predicting Flow-
also been supported by the U.S. Department of Energy con- Pattern transitions for the Whole Range of Pipe Incli-
tract DE-FG22-93BC14862. The development of the model nations,” Int. J. Multiphase Flow, 13, No. 1, 1-12
has also greatly benefited through numerous discussions with (1987).
Mr. Liang-Biao Ouyang of Stanford University.
3. Taitel, Y., and Dukler, A. E. “A Model for predicting
Flow Regime Transitions in Horizontal and Near
NOMENCLATURE
Horizontal Gas-Liquid Flow,” AIChe Journal, 22, 47
A Cross-sectional area (1976).
C0 Velocity distribution coefficient
4. Ansari, A. M., Sylvester, A. D., Sarica, C., Shoham,
D Pipe internal diameter
O., and Brill, J. P., “A Comprehensive Mechanistic
E In situ volume fraction
Model for Upward Two-Phase Flow in Wellbores,”
FE Liquid fraction entrained
SPE Prod. & Facilities, pp. 143-152, May 1994.
g Acceleration due to gravity
hL Height of liquid (Stratified flow) 5. Xiao, J. J., Shoham, O., Brill, J. P., “A Comprehen-
L Length sive Mechanistic Model for Two-Phase Flow in Pipe-
p Pressure lines,” paper SPE 20631, 65th ATC&E of SPE, New
Re Reynolds number Orleans, September 23-26, 1990.
S Contact perimeter 6. Petalas, N., and Aziz, K., “Development and Testing
V SG Superficial gas velocity of a New Mechanistic Model for Multiphase Flow in
V SL Superficial liquid velocity Pipes,” Proceedings of the ASME Fluids Division
δL Summer Meeting, Volume 1, FED-Vol. 236, pp. 153-
Liquid film thickness (Annular-Mist)
159, July, 1996.
ε Pipe roughness
η Pressure gradient weighting factor (intermittent 7. Petalas, N., and Aziz, K., “Stanford Multiphase Flow
flow) Database - User’s Manual,” Version 0.2, Petroleum
θ Angle of inclination Engineering Dept., Stanford University, 1995.
µ Viscosity 8. Gregory, G. A., Nicholson, M.K. and Aziz, K., “Cor-
ρ Density relation of the Liquid Volume Fraction in the Slug for
σ Interfacial (surface) tension Horizontal Gas-Liquid Slug Flow,” Int. J. Multiphase
τ Shear stress Flow, 4, 1, pp. 33-39 (1978).
~
x Dimensionless quantity, x
9. Andritsos, N., “Effect of Pipe Diameter and Liquid
Viscosity on Horizontal Stratified Flow,” Ph.D. Dis-
Subscripts
sertation, U. of Illinois at Champaign-Urbana (1986).
b relating to the gas bubble
c relating to the gas core 10. Oliemans, R. V. A., Pots, B. F., and Trope, N., “Mod-
f relating to the liquid film eling of Annular Dispersed Two-Phase Flow in Verti-

9
cal Pipes,” Int. J. Multiphase Flow, 12, No. 5, 711- Table 1. System Properties for Flow Pattern Maps
732 (1986).
Air/Water Oil/Gas
11. Taitel, Y., Barnea, D. and Dukler, A. E. “Modeling System System
Flow Pattern Transitions for Steady Upward Gas- Pipe diameter 2.047 in 6.18 in.
Liquid Flow in Vertical Tubes,” AlChe Journal, 26,
Gas Density .08 lb/ft3 8.139 lb/ft3
pp. 345-354 (1980).
Liquid Density 62.4 lb/ft3 52.53 lb/ft3
12. Barnea, D., Shoham, O.,Taitel, Y. and Dukler, A.E.,
“Gas-Liquid Flow in Inclined Tubes: Flow Pattern Gas Viscosity 0.01 cP 0.018 cP
Transitions for Upward Flow,” Chem. Eng. Sci., 40, 1 Liquid Viscosity 1.0 cP 2.757 cP
pp. 131-136 (1985). Interfacial Tension 72.4 dyne/cm 20 dyne/cm
13. Zuber, N., Staub, F.W., Bijwaard, G., and Kroeger, (Absolute) Pipe
0.00015 ft 0.01 ft†
P.G., “Steady State and Transient Void Fraction in Roughness
Two-Phase Flow Systems,” 1, Report EURAEC- † This high value of roughness is used to represent an open-hole well com-
GEAP-5417, General Electric Co., San Jose, Califor- pletion.
nia, January 1967.
14. Govier, G. W. and Aziz, K. “The Flow of Complex D1
I1
Mixtures in Pipes,” Van Nostrand, Reinhold (1972),
reprinted by Robert E. Kriger Publishing Co., Hunt- A2
I2
ington, New York, 1977.
15. Bendiksen, K. H. “An Experimental Investigation of
VSL
the Motion of Long Bubbles in Inclined Pipes,” Int. J.
Multiphase Flow, 10, pp 1-12 (1984). S4
I3 S2 S1
16. Zuber, N., and Findlay, J.A., “Average Volumetric A1
Concentration in Two-Phase Flow Systems,” J. Heat.
S3
Transfer, Trans. ASME, Ser. C, 87, pp. 453-468 I4
(1965).
17. Nicklin, D. J., Wilkes, J. O., and Davidson, J. F.,
VSG
“Two Phase Flow in Vertical Tubes,” Trans. Inst.
Chem. Engrs., 40, pp. 61-68, (1962). Figure 1. Transitions used in flow pattern determination

18. Zukoski, E.E., “Influence of Viscosity, Surface Ten-


sion, and Inclination Angle on Motion of Long Bub- 1.00
VSL= 0.05 ft/sec
bles in Closed Tubes,” J. Fluid Mech., 25, pp. 821-
837 (1966).
19. Weber, M. E., “Drift in Intermittent Two-Phase Flow
in Horizontal Pipes,” Canadian J. Chem. Engg., 59,
h L/D

pp. 398-399, June 1981. 0.10

20. Wallis, G. B. “One-Dimensional Two-Phase Flow,”


McGraw-Hill, New York, 1969. Air/Water at Standard
Conditions
21. Harmathy, T.Z., “Velocity of Large Drops and Bub- 2.047" Pipe, +2° Inclination
bles in Media of Infinite or Restricted Extent,” AIChE
J., 6, pg. 281 (1960). 0.01
0.01 0.1 1 10 100 1000
VSG
Figure 2. Stratified Flow: Multiple Roots in Liquid Height
Calculation for Increasing VSG (Air/Water System)

10
Begin Mechanistic Model

DISPERSED
E Ls < 0.48 Yes CG ≤ 0.52 Yes
BUBBLE

No

VL ≤
( ~
)
gD 1 − h L cosθ
Calculate
Flow downward fL
or horizontal?
Yes ~
hL  ~ 
θ ≤ 0 1 2 dA L 
F 
2
VG ≤1
 2~ ~ 
 C A G dh L 
No
Yes
Calculate
~ No
δL

Flow is
STRATIFIED
4µL( ρL − ρG) gcosθ
VG ≤
sρLρGVL Yes
STRATIFIED
SMOOTH

VL
≤14
.
ghL
No
STRATIFIED
WAVY
EL A−M
≤ 0.24
ANNULAR
Yes
~ ~ MIST
δL < δL max

No FROTH

EL > 0 . 24
No SLUG

Yes

ELs ≥ 0.48 No ELs ≤ 09


. Yes SLUG

ELONGATED
2
No
No
2
3 Cl γ V0 BUBBLE
cosθ ≤
4 2 dbub g

Yes
D > 19
(ρ L
− ρG σ ) Yes BUBBLE
2
ρL g
FROTH EL > 0.75
slug

Figure 3. Flow pattern determination for mechanistic model

11
100
100 Dispersed Bubble
Dispersed Bubble
Froth I
Froth I
Slug
Slug 10
10

Elongated Bubble
Annular-
V sL (ft/sec)

Elongated Bubble Annular-

V SL
Mist 1 Mist
1

Stratified
Wavy Stratified
0.1 Wavy
0.1
Stratified Smooth
Stratified Smooth

0.01
0.01 0.01 0.1 1 10 100
0.01 0.1 1 10 100
VSG
VsG (ft/sec)
Figure 4. Flow pattern map for air/water system at 0° Figure 7. Flow pattern map for oil/gas system at 0° inclina-
inclination (horizontal) tion (horizontal)

100
10
1 100 0.1
10
∆Pfr (psi/ft) 0.01 ∆Pfr 0.001
1
0.1
0.0001 1E-005
0.01 ∆Pfr (psi/ft) 0.001
∆ Pfr
1E-006 1E-007
0.0001 1E-005
10 1E-006 10 1E-007

1 1
VsL (ft/sec) 10 VsL 10

0.1 1 0.1 1
VsG(ft/sec) VsG
0.1 0.1

0.01 0.01 0.01 0.01

Figure 5. Frictional pressure gradient for air/water system at Figure 8. Frictional pressure gradient for oil/gas system at 0°
0° inclination (horizontal) inclination (horizontal)

1 1
0.8 0.8
1
1
0.6 0.6
0.8 EL EL
0.4 0.8 0.4
0.6 0.2 0.6 0.2
EL EL 0
0.4 0 0.4
0.2 0.2
10 10
0 0
0.01 0.01 1
1
0.1 0.1
VsL (ft/sec) VsL
1 1
0.1 0.1
10 10
VsG (ft/sec) VsG

0.01 0.01

Figure 6. Liquid volume fraction for air/water system at 0° Figure 9. Liquid volume fraction for oil/gas system
inclination (horizontal) at 0° inclination (horizontal)

12
100 100
Dispersed Bubble Dispersed Bubble

Froth I
Slug
Slug
10 10

Annular-
Mist

V SL
V SL

1 1
Annular-
Mist
Elongated Bubble
0.1 Elongated Bubble 0.1
Froth

0.01 0.01
0.01 0.1 1 10 100 0.01 0.1 1 10 100
VSG VSG
Figure 10. Flow pattern map for air/water system at 10° Figure 13. Flow pattern map for oil/gas system at 10° upward
upward inclination inclination

100
10
1 100 0.1
10
∆Pfr 0.01 ∆Pfr 0.001
1
0.1
0.0001 1E-005
0.01 ∆Pfr ∆ Pfr
1E-006 1E-007 0.001
0.0001 1E-005
10 1E-006 10 1E-007

1 1
VsL 10 VsL 10

0.1 1 0.1 1
VsG VsG
0.1 0.1

0.01 0.01 0.01 0.01

Figure 11. Frictional pressure gradient for air/water system at Figure 14. Frictional pressure gradient for oil/gas system
10° upward inclination at 10° upward inclination

1 1
0.8 0.8
0.6 0.6
EL 1 1 EL
0.4 0.4
0.2 0.8 0.8
0.2
0 0.6 0.6 0
EL EL
0.4 0.4
10 10
0.2 0.2

1 0 0
0.01 1
VsL 0.1 VsL
0.1 10 1 0.1
1 10
VsG VsG
0.1
0.01 0.01 0.01

Figure 12. Liquid volume fraction for air/water system at 10° Figure 15. Liquid volume fraction for oil/gas system at 10°
upward inclination upward inclination

13
100 100
Dispersed Bubble Dispersed Bubble

Froth
Froth

10 10
Slug Annular- Slug
Bubble
Mist
V sL (ft/sec)

V SL
1 1
Annular-
Mist
Bubble
0.1 0.1 Elongated Bubble
Elongated Bubble

0.01 0.01
0.01 0.1 1 10 100 0.01 0.1 1 10 100
VsG (ft/sec) VSG
Figure 16. Flow pattern map for air/water system at 90° Figure 19. Flow pattern map for oil/gas system at 90° upward
upward inclination inclination

100
10
1 100 0.1
10
∆Pfr (psi/ft) 0.01 ∆Pfr 0.001
1
0.1
0.0001 1E-005
0.01 ∆Pfr (psi/ft) ∆ Pfr
1E-006 1E-007 0.001
0.0001 1E-005
10 1E-006 10 1E-007

1 1
VsL (ft/sec) 10 VsL 10

0.1 1 0.1 1
VsG(ft/sec) VsG
0.1 0.1

0.01 0.01 0.01 0.01

Figure 17. Frictional pressure gradient for air/water system at Figure 20. Frictional pressure gradient for oil/gas system at
90° upward inclination 90° upward inclination

1 1
0.8
0.8
1 0.6
0.6 EL
0.8 EL 0.4
0.4 1
0.2
0.6 0.2 0.8 0
EL
0.4 0
0.6
0.2 EL 10
10 0.4
0 0.2 1
0.01 1
0.1 0 VsL
VsL (ft/sec) 0.01
1 0.1
0.1 0.1
10 1
VsG (ft/sec) 10
VsG
0.01 0.01

Figure 18. Liquid volume fraction for air/water system Figure 21. Liquid volume fraction for oil/gas system at 90°
at 90° upward inclination upward inclination

14
100 100
Dispersed Bubble Dispersed Bubble

Froth Froth
Slug Slug
10 10

Annular-
Elongated Bubble
Mist
Elongated Bubble Annular-
V SL

V SL
1 1 Mist

Stratified
0.1 Wavy 0.1 Stratified
Wavy

0.01 0.01
0.01 0.1 1 10 100 0.01 0.1 1 10 100
VSG VSG
Figure 22. Flow pattern map for air/water system at 10° Figure 25. Flow pattern map for oil/gas system at 10° down-
downward inclination ward inclination

10 10

0.1 0.1
∆Pfr 10 ∆Pfr 10

0.001 0.001
0.1 0.1
∆ Pfr ∆ Pfr
1E-005 1E-005
0.001 0.001

10 1E-005 10 1E-005

1 1
VsL 10 VsL 10

0.1 1 0.1 1
VsG VsG
0.1 0.1

0.01 0.01 0.01 0.01

Figure 23. Frictional pressure gradient for air/water system at Figure 26. Frictional pressure gradient for oil/gas system
10° downward inclination at 10° downward inclination

1
0.8 1 1

1 0.6 0.8 0.8


EL
0.8 0.4 0.6 0.6
EL EL
0.2 0.4
0.6 0.4
EL 0 0.2
0.4 0.2
0
0
0.2 10 0.01
0 0.1 10
0.01 1
0.1 1 1
VsL
1 VsL
0.1 VsG 10
10 0.1
VsG

0.01 0.01

Figure 24. Liquid volume fraction for air/water system Figure 27. Liquid volume fraction for oil/gas system at 10°
at 10° downward inclination downward inclination

15
3000 2895

61.6%
2500 49.4%
Number of Data Points

38.0%
EL
EL 37.9%
2000
40.3%
1500 22.0%

1000 43.2%
633 41.2% Mechanistic Model
500 407 382 31.7%
293 Xiao, Shohamand Brill
212 242
146 ∆P DP Mukherjee and Brill
46 118 56 112 67 83 101 90 68 33.0%
Beggs and Brill
0 41.3% Dukler, Wicks and Cleveland
-90° to -80°

-80° to -70°

-70° to -60°

-60° to -50°

-50° to -40°

-40° to -30°

-30° to -20°

-20° to -10°

-10° to 0°

10° to 20°

20° to 30°

30° to 40°

40° to 50°

50° to 60°

60° to 70°

70° to 80°

80° to 90°
0° to 10°

27.6% Homogeneous Model

0% 10% 20% 30% 40% 50% 60% 70% 80%


Inclination Angle Range
Percentage of Experimental Data Predicted to 15%Accuracy

Figure 28. Distribution of inclination angle for experimental Figure 31. Comparison of selected methods’ ability to predict
data experimental volume fraction liquid and pressure gradient to
within 15% accuracy
+15%
1.00

-15%

0.75 SU-21, 23
SU-24 to 29
Mechanistic Model
SU-53 to 56
Calculated

SU-66
SU-96 Xiao, Shohamand Brill
0.50 SU-101
SU-108
SU-111 to 113 Mukherjee and Brill
SU-114 to 117
0.25 SU-120 to 124
SU-138
Beggs and Brill
SU-175 to 198
SU-199 to 209
5,951 Data Points
SU-210 to 215
0.00 Dukler, Wicks and Cleveland -90° to -30°
0.00 0.25 0.50 0.75 1.00
-30° to +30°
Experimental
Homogeneous Model +30° to +90°
Figure 29. Mechanistic model volume fraction liquid calcula-
tions compared with experimental data 0% 10% 20% 30% 40% 50% 60% 70%
Percentage of Experimental Data Predicted to 15% Accuracy
+15%
1.50

1.25 -15% Figure 32. Comparison of selected methods’ ability to predict


experimental volume fraction liquid to within 15% accuracy
1.00 SU-21, 23
SU-24 to 29 (grouped by angle of inclination)
0.75 SU-53 to 56
Calculated

SU-66
SU-96
0.50
SU-101
SU-108
0.25 SU-111 to 113 Mechanistic Model
SU-114 to 117
0.00 SU-120 to 124
SU-138 Xiao, Shohamand Brill
SU-175 to 198
-0.25
5,951 Data Points SU-199 to 209
Mukherjee and Brill
SU-210 to 215
-0.50
-0.50 -0.25 0.00 0.25 0.50 0.75 1.00 1.25 1.50 Beggs and Brill
Experimental

Figure 30. Mechanistic model pressure gradient calculations Dukler, Wicks and Cleveland -90° to -30°
compared with experimental data -30° to +30°
Homogeneous Model +30° to +90°

0% 10% 20% 30% 40% 50% 60% 70% 80%


Percentage of Experimental Data Predicted to 15%Accuracy

Figure 33. Comparison of selected methods' ability to predict


experimental pressure gradient to within 15% accuracy
(grouped by angle of inclination)

16

Das könnte Ihnen auch gefallen