Professional Documents
Culture Documents
A 2-in. (5.l-cm) diameter, 550-ft (168-m) long pipeline was designed and con-
structed in a hilly terrain configuration. Two-phase-flow liquid holdup and pressure-
loss correlations were evaluated using gas/water data obtained from experiments.
Accurate predictions were obtained using the Beggs and Brill correlation and a
combination of Beggs and Brill and Guzhov et al. correlations.
1198 JOURNAL OF PETROLEUM TECHNOLOGY
phase density that will determine elevation pressure correlation that requires an iterative procedure. The
loss. In plug flow, the stratified uphill holdup ex- holdup is used to determine friction and acceleration
pression is used for holdup in both uphill and pressure drops. The two-phase friction factor is a
downhill flow. This results in complete recovery of function of no-slip holdup and a smooth pipe friction
the elevation pressure loss. A friction loss expression factor.
is given. Evaluation of the acceleration term requires Eaton's? correlation also is used for pipeline
an iterative solution. design. This method is based on data taken in 2-, 4-,
Beggs and BrilP investigated two-phase flow at and 17-in. pipe under field conditions. Eaton
several inclinations between + 90 and - 90° from proposed a holdup correlation that is a function of
horizontal. Test sections of 1- and I.5-in. pipe were several dimensionless groups. This holdup is used to
used. Holdup and pressure drop were measured, and determine the acceleration component of the pressure
holdup and friction factor correlations were drop. The friction factor is also a function of several
proposed. A horizontal flow pattern map consisting dimensionless groups, but this must be used with
of segregated, intermittent, and distributed flow caution because the friction factor becomes un-
regimes is used to determine a horizontal holdup. bounded as single-phase flow is approached.
The horizontal holdup is corrected for inclination
and then used to determine the elevation pressure Experimental Facility
loss. Friction and acceleration terms also are A schematic diagram of the hilly-terrain-pipeline test
provided. facility is shown in Fig. 1. The test facility consisted
Recently, Robinson4 showed that the accuracy of of 1,200 ft of 2-in. Schedule 40 line pipe (ID = 2.067
the Beggs and Brill correlation could be improved in.) with associated gas compressor, water pump,
when applied to directional wells by using a tran- meter runs, and separator. The entire system was
sitional zone between the segregated and intermittent closed - i.e., both gas and liquid were recirculated.
flow regimes. When the flow is in this region, a The test section was 400 ft long and arranged in three
weighted average of the segregated and intermittent hills. Entrance and exit effects were avoided by
holdup values is used. extending the pipe about 75 ft (450 pipe diameters)
Sometimes, horizontal flow correlations are used on each end of the test section. Fig. 2 is a profile of
with Flanigan's elevation factor to design pipelines in the test section, and Table 1 gives data about that
hilly terrain. The American Gas Assn. Design section.
Manuals recommends using Dukler's6 horizontal
correlation with Flanigan's elevation factor. Dukler's General Procedure for Two-Phase Flow Testing
correlation is based on a large amount of ex- The system was pressurized initially to 400 or 500
perimental data. Dukler developed a holdup psi a with a three-stage Worthington compressor.
N - 1
from each segment and then weighed. The ball valves
were opened and the equalizing valves were shut. If To make any statistical inference from ex-
only pressure loss data were desired, the holdup perimental data, the data usually must be distributed
procedure was omitted. A complete discussion of the normally. For this reason, the actual values of
experimental equipment, including pressure traverse percent error for predicted pressure drop in the total
and holdup measurement procedure, is found in test section were sorted into ranges, and the
Refs. 8 and 9. The holdup data taken by PalmerS and cumulative relative frequency of the percent error
the pressure drop data taken by Payne9 are presented was plotted vs the value of percent error on normal
in Tables 2 and 3. probability paper. A normal distribution will plot as
a straight line intersecting the 50th percentile at the
Evaluation Techniques mean. It also will intersect the 16th percentile and the
PVT Properties 84th percentile one standard deviation on either side
Because the liquid used here was water, we did not of the mean. Figs. 3 and 4 show that the errors do
approximate a normal distribution.
Because the average percent error is based on a
finite sample, we cannot expect it to coincide exactly
12 -Flow
with the actual population mean. However, since the
percent errors approximate a normal distribution, it
(6)
(68)
(OA)
(4) (3)
(2)
is possible to determine an interval that will bracket
(1)
the true value of the average percent error, given
100 200 300 400 specified odds. For a normal distribution where the
LENGTH (FT)
sample variance is used as an estimator of the
population variance, this interval is given by the
Elevation
Section Length (tp Chang' Ot) Angle(degr •• ) following expression:
3
50
50
50
3,68
-3.76
8.11
4.23
-4.30
1.02
%E ± (t I-a/
.N - 1) (...£).
z' YN
4 50 -3.31 -3.80
5A 100 0.24 -0. t 4
58 46 6.69 8.24
-5.91
Using the above confidence limits and IX = 0.01
6 54 -6.56
(i.e., a 99% confidence interval), we can show that
Note that •• ctlon 5 contain a two inclinations ovar the the Beggs and Brill correlation true average percent
total length corre.pondlng to the pre,aur. drop data
error lies between -20.51 and -28.53%. Con-
Fig. 2 - Test section profile. sequently, at best this correlation underpredicts
99 # 99
#
> >
()
()
z 95 z 95
w w
:> :>
()
()
w w
a:
... 80 ...a: 80
w w
::l- 80
::I- 80
< <
..J
..J
W 40 w 40
a: a:
w
::l- 20
• w
::I- 20
< <
..J
..J
:> :>
:l: :l: 5
5 :>
:> ()
()
$ERROR $ERROR
Fig. 3 - Cumulative frequency of percent errors in Fig. 4 - Cumulative frequency of percent errors in
pressure drop (total test section) using the Beggs pressure drop (total test section) using the Eaton·
and Brill correlation. Dukler-Flanigan combination.
%E
-50 -20 -10 0 10 20 50 100
Beggs and Brill (modified by Robinson) 4 61 87 94 100
Beggs and Brill (neglect pressure recovery) 0 13 33 50 57 93 100
Beggs and Brill (rough pipe friction factor) 0 26 39 63 73 84 99 100
Dukler (HL), Dukler (FF), Flanigan (EF) 0 0 0 11 31 54 97 100
Eaton (HL), Eaton (FF), Flanigan (EF) 0 10 20 44 64 79 96 100
Guzhov et al. 70 84 84 89 89 90 96 100
HL = liquid holdup.
FF = friction factor.
EF = elevation factor.
pressure loss at least 20070 for 2-in. pipe. caused by slugging. The Guzhov et al. correlation
A similar analysis for the Eaton-Dukler-Flanigan gave the lowest average percent error of the three
combination gives a 99% confidence interval of 1.16 correlations. Using the Guzhov et al. criteria for
to 10.80%. Table 4 gives the percent error distribu- stratified flow, liquid holdup in downhill flow is
tion of the other methods used to compare calculated predicted most accurately by Guzhov et al. in
pressure drop with the measured pressure drop. stratified flow and by Beggs and Brill elsewhere.
Statistics for this combination also are shown in
Holdup Analysis Table 5.
Liquid holdup was measured in three uphill and three The Beggs and Brill correlation was used in a
downhill sections for each experiment. Calculated modified form to see if the downhill holdup could be
values of liquid holdup corresponding to the predicted more accurately. The correlation was
measured values were determined from the Beggs and forced to a segregated flow condition for those
Brill,3 Flanigan, I and Guzhov et al. 2 correlations. The experiments where stratified flow was indicated,
results of this analysis are provided in Table 5 and in according to the Guzhov et at. criteria. This modified
Figs. 5 through 10. form proved to be less accurate than the original
Examination of the individual percent errors for Beggs and Brill correlation for downhill flow. The
the uphill sections indicates that the correlations consistently large positive percent errors for liquid
predict the holdup more accurately in some sections holdup in the downhill section indicates that the
than others. This fluctuation could result from the correlation is overpredicting the downhill holdup.
presence of slugs in some parts of the test section. As Table 5 shows that the Flanigan correlation was
can be seen in Table 5, the Beggs and Brill correlation the least accurate of the three correlations. As
was the most accurate in predicting the liquid holdup Flanigan stated, the correlation was a crude ap-
in the uphill sections. proximation, but this was one of the first steps taken
In downhill flow, the fluctuation of the holdup to predict liquid holdup. Flanigan's correlation was
percent error in individual sections is similar to that developed for uphill flow, and holdup in downhill
found in the uphill sections. This also could be flow was ignored.
1. 0 r---.--,.--r---.--,.--r----r-,..--,-.....,.
0.8
A. A.
'"
Q
..J
'"
Q
..J
o o
:z: :z:
0.8 0.8
..J ..J
..J ..J
:z: :z:
A. A.
'"
Q
'"
Q
0.4 0.4
'"
a: '"a:
.,'" .,'"
-< -<
'":Ii '"
:Ii
0.2 0.2
0.0 0.2 0.4 0.8 0.8 1.0 0.0 0.2 0.4 0.8 0.8 1.0
Fig. 5 - Measured uphill holdup vs calculated uphill Fig. 6 - Measured uphill holdup vs calculated uphill
holdup using the Beggs and Brill method. holdup using the Flanigan method.
Q.
Q.
"...
Q
"...
Q o
o ...'" 0.8
......'" '=
z"
;:
'"
Q. o
Q
"
Q
Q
w
w a:
a:
.,"
".,< <
w
w ::!i
::!i
1. 0 ,..---,----,.---.--,--.,.--r----r---,----.--?I
Q.
0.8
Q.
"...
Q
"...
Q
o
o
......'"
......'" 0.8
0.8
'"z
'"z
;:
;:
oQ
oQ Q
w
0.4
Q
a:
w
a: .,"
".,< <
w
::!i
w 0.2
::!i
0.0
0.0 0.2 0.4 0.8 0.8 1.0
0.0 0.2 0.4 0.8 0.8 1.0
CALCULATED DOWNHILL HOLDUP
CALCULATED DOWNHILL HOLDUP
Fig. 10 - Measured downhill holdup vs calculated downhill
Fig. 8 - Measured downhill holdup vs calculated downhill holdup using the Beggs and Brill/Guzhov et a/.
holdup using the Beggs and Brill method. method.
HL = liquid holdup,
FF = friction factor.
EF = elevation factor.
Dukler friction factor, and Flanigan elevation factor same results for these data. The Beggs and Brill
combination. This must be a direct result of the correlation next is analyzed in detail.
Dukler holdup correlation. This agrees with a
previous evaluation of Dukler's holdup correlation Beggs and Brill Method
by Marcano 13 where the Dukler correlation un- When analyzing the holdup data, we found that the
derpredicted by 33070. Underprediction of liquid Beggs and Brill correlation over predicted holdup in
holdup would result in overprediction of friction downhill flow. This would result in too much
loss. pressure recovery and a corresponding under-
The combination using Eaton's friction factor was prediction in overall pressure loss. In view of this, the
fairly accurate since the correlation was developed Beggs and Brill correlation was evaluated with the
with data taken in 2-in. pipe. elevation term set at zero in downhill flow. This
The original Beggs and Brill correlation and the resulted in considerably better performance. The
version modified by Robinson4 gave essentially the average percent error was reduced to -1.45% with a
(!) -24
-24 Q
<&
Q
'" '"
Co ta
t'" (!)
-20
Co
-20 0
Co
a:
0 Q GG
a:
Q GG W -16
a: G
w -16 GG
G :>
a: ~ G
.,
:> (!)
eefl'
'"'"w -12
GGG
'"a:w -12
a:
Co ~G
G
'" <!>
Co G '"
€ '" Q
w
a-s 0
Q
w
a: -8
ee0
(!)
(!)~
a:
.,:> -6 "'i
"'.~
.,
:> G$ .
<
w
G
< ~ £
!~
w
~ ~ -4
-4
I~
-8 -12 -16 -20 -24
-4 -8 -12 -18 -20 -24
CALCULATED PRESSURE DROP (PSID)
CALCULATED PRESSURE DROP (PSID)
Fig. 12 - Measured vs calculated pressure drop (total test
Fig. 11 - Measured vs calculated pressure drop (total test section) using the Beggs and Brill correlation as
section) using the 'Beggs and Brill correlation. modified by Robinson.
., .,
-24 -24
0 ., 0
t'" G>
...'"
-20
...0 -20
...0
a: a:
0 0
W w -18
a: a:
:> :>
'"<w '"<w
::! ::!
Fig. 13 - Measured vs calculated pressure drop (total test Fig. 15 - Measured vs calculated pressure drop (total test
section) using the Beggs and Brill correlation section) using the Eaton·Dukler-Flanigan
with no pressure recovery. combination .
.,
., .,
-24 -24
0
., ., !:
.,
t o
t I!l
-20
...0 -20
...0
a:
0
., ., a:
0
., .,
w -18 .,~
w -18
a: a:
.,., .,.,
:> :>
., G> "p
., .,
w -12 w -12
a:
... ., ...a:
0 ., 0
w w
a: -8 a: -8
.,
:>
.,
:>
< <
w w
::! ::!
-4 -4
Fig. 14 - Measured vs calculated pressure drop (total test Fig. 16 - Measured vs calculated pressure drop (total test
section) using the Beggs and Brill correlation section) using Dukler-Dukler-Flanigan com-
with a rough-pipe friction factor multiplier. bination.
-.,...
Q
-24
.,
Q
-24
0
-...
0
-20 t
... -20
C!I
C!I
a: 0
Q 0" a:
w -18
Q 0 e
a: w -18
.,.,
:::l a: ~
C!I
.,.,
:::l
w 0
a: -12 w -12 e
... ...a: 0 C!I GI
Q 0
W Q
r9
E>El C!I
a: -8 w 0
.,
:::l a:
.,
:::l
-8 GI
<
W <
~ W C!I
-4 ~
-4
-20 -24
-12 -18 -20 -24
CALCULATED PRESSURE DROP (PSID)
CALCULATED PRESSURE DROP (PSID)
Fig. 17 - Measured vs calculated pressure drop (total test
section) using the Eaton·Eaton·Flanigan com· Fig. 18 - Measured vs calculated pressure drop (total test
bination. section) using the Guzhov et a/. correlation.
Introduction
The following comments pertain to two methods of liquid velocity number, N LV' defined by
holdup prediction discussed in "Evaluation of In-
clined-Pipe, Two-Phase Liquid Holdup and N LV = VSL (PL/(1)Y4,
Pressure-Loss Correlations Using Experimental
Data" by Payne et at. (Sept. 1979 JPT, Pages 1198- where VSL is the superfacialliquid velocity, PL is the
1208). liquid density, and (1 is the interfacial tension of the
two phases.
Beggs and Brill Correlation The slip between the pipeline phases basically is
Payne et at. 1 concluded that the Beggs and Brill caused by the difference in properties of the two
correlation accurately predicts holdup in uphill flow phases. The most important physical properties are
and over predicts holdup in downhill flow, but density, viscosity, and interfacial tension. When the
remains the most accurate method for all flow pressure of a gas-condensate mixture increases and
regimes with the exception of stratified flow. Use of approaches the critical condition, the difference in
the Beggs and Brill correlation also has been the properties of the two phases is reduced and the
recommended by others (such as Gould and Ram- holdup for such a gas-condensate mixture at high
sey2) for gas-condensate offshore pipelines. pressure is definitely lower than the air/water system
The Beggs and Brill correlation was used at the used in the Beggs and Brill experiment at the same
Abadan Inst. of Technology 3 to predict the per- flow rates of gas and liquid. Hence, it can be con-
formance of a 24.5-in. (622-mm) ID hilly-terrain gas- cluded that the method, which does not consider the
condensate pipeline 30 miles (48 km) long at 2,500- effects of physical properties, overpredicts the
psi (17 236-kPa) inlet pressure. When the predicted horizontal holdup for a high-pressure gas-condensate
holdup results were studied, many unrealistic values pipeline.
were noticed. The method predicted negative values For inclined flow where the effects of physical
and values greater than one. It also predicted properties are considered, the inclination correction
horizontal holdup may decrease in the uphill section factor for horizontal segregated flow shows that the
of the pipe. Careful examination of the calculations increase of (p L / (1) increases the uphill holdup. While
revealed most of these incorrect results originated for the gas-condensate line, the increase of (p L / (1),
from use of the inclination correction factor. which is the pressure increase, should reduce the
In the Beggs and Brill method, the two parameters holdup. For example, at the critical condition the
of input liquid content and Froude number are flow becomes homogeneous and the slip is zero,
sufficient to predict only the horizontal flow regime while the correction factor approaches infinity. For
and the horizontal holdup, and they are independent horizontal intermittent flow, the correlation shows
of the physical properties of the fluids. But the in- that the holdup decreases with the increase of (P L / (1).
clination correction factor, which adjusts the This further magnifies problems associated with the
horizontal holdup for the inclination, depends on the discontinuity that occurs from methods giving
different correlations for different flow regimes. 1,3
Gregory4 showed that the effect of the angle of
0149-2136/80/0001-8782$00.25
inclination on holdup is small for angles less than
© 1980 Society of Petroleum Engineers of AIME 10°. Therefore, for uphill flow it is recommended
JANUARY 1980
169
that the inclination correction factor be ignored for a
high-pressure gas-condensate hilly-terrai~ pipeline til>E=EHPLg Ez+
and the horizontal holdup be used. This recom-
mendation may reduce the accuracy of the holdup = [PLHL +PC (l-HL)]g Ez+ ,
prediction at some conditions but ensures that large
errors are not introduced in the evaluation of the
which gives
holdup. It is worth mentioning that setting con-
straints in the calculation procedures (such as
0:5 H L :51) and other constraints given in the r~vised
method of Beggs and Brill is not a reliable solutIOn to
where til>E is elevation pressure drop, P~ and Pc are
the problem. We discarded only the obviously liquid and gas density, respectively, ~z is the sum
erroneous results. Proper constraints should be in the
of uphill rises, HL is liquid holdup, and g is
f~rm of limits on flow properties and conditions, gravitational acceleration. The equation shows that
such as physical properties, diameter, Reynolds H L is always smaller than E H as E H < 1, and E lj
number, etc. given by Flanigan can be taken as HL when PC/PL IS
The final comment on the Beggs and Brill method, very small. The correlation of E H was developed
which can be generalized to most of the available using data of Flanigan as well as other investigators.
correlations for two-phase flow, is that these Flanigan's data were taken from a high-pressure gas-
correlations are based on some forms of dimen-
condensate line, and the density ratio of the two
sionless groups or factors introduced by simply
phases should not be ignored, especially at low
combining the parameters or exactly duplicating the
holdup. The properties of the two phases are not
forms used in single-phase flow. To have a reliable
given by Flanigan, but the density ratio may be
correlation, the dimensionless groups should be
estimated to be about 0.1, which results in the
developed by rigorous analysis of the multiphase
following values.
flow and a thorough study of the pertinent forces.
Otherwise, these correlations should be used only Liquid head factor 0.3 0.4 0.5 0.6 0.7
within the tested range of the variables. The ap- Liquid holdup 0.22 0.33 0.44 0.56 0.67
plication of these correlations becomes further Error percent 35 20 13 8 5
limited when parameters are employed that were not
varied at all (e.g., PL and (J in the Beggs and Brill Flanigan used his data to develop the correlation
correlation). for E H >0.35, and for E H <0.35 he used 10:-"-
Flanigan Correlation pressure gas-liquid data of others. Hence, the denSity
correction for E H < 0.35 is not required. Fig. 6 in
Flanigan defined the liquid head factor E H as the Ref. 1 shows that the predicted liquid holdup values
fraction of the elevation pressure drop that would by the Flanigan method are higher than the measured
exist if all uphill sections of the pipe were filled with values for E H >0.35; therefore, the density
the liquid. Flanigan did not measure the liquid correction definitely improves the results.
holdup but evaluated the elevation pressure drop by
subtracting the calculated friction pressure drop References
from the total measured pressure drop, assuming no
I. Payne, G.A., Palmer, G.M., Brill, J.P., and Beggs, H.D.:
downhill pressure recovery. Therefore, the values of "Evaluation of Inclined-Pipe, Two-Phase Liquid Holdup and
E H evaluated by Flanigan ~epend on t?e pressure Pressure-Loss Correlations Using Experimental Data," J. Pet.
recovery in the downhill sectIOn of the pipe and the Tech. (Sept. 1979) 1198-1208; Trans., AIME, 267.
calculated friction pressure drop, which are not 2. Gould, T.L. and Ramsey, E.L.: "Design of Offshore Gas
Pipelines Accounting for Two-Phase Flow," J. Pet. Tech.
pertinent to the liquid holdup in uphill section of the (March 1975) 366-374; Trans., AIME, 259.
pipe. Neglecting these effects, 3. Danesh, A. and Noghrekar, G.R.: "Hilly Country Gas-
Condensate Pipelines," Oil Service Co. of Iran Report No.
P-3062 (1976).
4. Gregory, G.A.: "Comparison of Methods for the Prediction of
Liquid Holdup for Upward Gas-Liquid Flow in Inclined
Original manuscript (SPE 8782) received in SOCiety of Petroleum Engineers Pipes," Cdn. J. Chern. Eng. (1975) 384-388.
office Oct. 22, 1979.
This paper will be included in the 1980 Transactions Volume. JPT