You are on page 1of 4

IMPROVING OPERATIONS

An improved revision to the Hagedorn and


Brown liquid holdup correlation
GHASSAN H. ABDUL MAJEED
Computer Centre, University of Baghdad
Baghdad, Iraq .
NIMAT BEHNAM ABU AL-SOOF and JAWAD R. ALASSAL *
Petroleum Research Centre, Scientific Research Council
Jadiriyah, Baghdad, Iraq

ABSTRACT Several two-phase, flowing pressure loss, evaluation studies were


Many two-phase flowing pressure drop evaluation studies have performed in which the modified Hagedorn-Brown correlation was
shown that the modified Hagedorn-Brown correlation is the best included in addition to other existing correlations. A summary of
over-all predictor. However, on average, this correlation tends to these studies is shown in Table 1. The results of these studies
under-predict pressure drop. showed that, the modified Hagedorn-Brown correlation was the
In this study, the Hagedorn-Brown liquid holdup correlation best over-all predictor. However, it was found, based on the statisti-
was revised using 51 pressure profiles containing 540 pressure loss cal results presented in these studies, that on average, the modi-
measurements. The revised correlation gave higher value of liquid . tied Hagedorn-Brown correlation tended to under-predict pressure
8> mentioned that the under-prediction was
holdup than the original for the same value of correlating junction. drops. Rossland<
The pressure drops for 157 well test data (not used in the devel- mainly due to the under-prediction of liquid holdup. Thompson(9)
opment of the revised liquid holdup correlation as they were taken added that, the modified Hagedorn-Brown correlation tended to
from another field), were calculated for different cases using the over-predict pressure loss in bubble flow (Griffith), while it tended
original and the revised liquid holdup correlation. Tn comparing to under:..predict for slug flow.
the calculated pressure drop results with the field data, the revised In the present study an attempt was made to revise the
correlation gave considerably better results. Hagedom-Brown.liquid holdup correlation using unpublished field
data. The original and the revised liquid hold-up correlations were
then tested against 157 well test data using different cases.
Introduction
Several published correlations can be. used to predict pressure drops Measured Data
in vertical oil well pipes for the simultaneous flow of oil, water
and gas. One of these correlations is that of Hagedorn and The well test data were drawn from unpublished sources. These
Brown(l>. Their correlation was developed from 475 tests in a data consisted of 51 field pressure profiles (for vertical wells). In
1500-ft experimental well using fluids having viscosities up to 110 each profile, the well was divided into equal depth interval and
cp. An average mixture density corrected for downhole conditions the pressure was measured at the end of each interval. The total
was used for calculating estimates of pressure losses caused by fric- number of pressure measurements in the collected profiles was 540
tion and acceleration. Liquid holdup was then calculated from the points. The ranges of flow variables covered by the measured data
total measured pressure loss and the calculated values of friction are given in Table 2. Distribution of the measured data according
and acceleration losses. These holdup values were correlated with to liquid flow rate, pipe diameter and gas/liquid ratio is shown
various flow variables and fluid properties. Because liquid hold- in Table 3.
up was not measured directly, values of holdup given by the corre-
lation are not always physically significant. Development of the Correlation
Further work by Brill and Hagedom(2) recommended that the
The gradient equation adopted in the present work is the same
pressure .gradient should be calculated by the Griffith<3> correla-
as that developed by Hagedorn-Brown:
tion for the bubble regime. Also, they suggested that the density
of the mixture calculated using the Hagedorn-Brown holdup corre- y2
lation should be compared with that calculated using no-slip hold- Llp PmL1[~]
up. The largest of these two values was then considered. 144 = Pm + +
Llh 2.9652 X 1011 d5 Pm Llh

................................................ (1)
*~rrent Address: Petroleum Department, College of Engineering, Uni-
versity of Baghdad, Iraq. . where
Keywords: Reservoir engineering, Pressure loss, Liquid holdup, Statistics. Pm = HIPL + PgCl- HJ .................................................... (2)

Paper reviewed and accepted for publication by the Editorial Board of the Journal of Canadian Petroleum Technology.

November-December 1989, Volume 28, No.6 25


TABLE 1.
No. of Data Used
Evaluation Study Date in the Study Reference
Messulam 1970 434 4
Lawson-Sri II 1973 726 5
Robinson 1974 268 6
Ibe 1979 891 7
Ross land 1981 130 8
Thompson 1982 1191 9
Ghassan-Zeiad 1987 1276 10

TABLE 2. Ranges of parameters covered in this Cf Nlv


B. I
P Cn I

study Ngv
,575
Pa
B.l
Nd

Parameters Field Units 51 Units


FIGURE 1. Revised Hagedom-Brown liquid holdup correlation.
Oil flow rate, m3fD (STBO/D) 0-20000 0-3180
Water flow rate, m3JD (STBW/D) 0-9000 0-1430
Gas flow rate, m3JD (Macf/D) 1-1200 28-3400
Tubing diameter, m (inch) 1-6.5 0.0254-0.165
Oil gravity, gm/cc (API) 9-55 1.007-0.758
Gas-Liquid ratio, m3fm3 (Scf/STB) 25-10000 4.5-1781
Wellhead pressure, KPa (psig) 15-1684 103-11610
Bottomhole pressure, KPa (psig) 130-6200 896-42747
Tank gas gravity (air= 1) 0.6-1.25 0.6-1.25
Well depth, m (ft) 7000-14070 2133-4288 H / 'f'
Wellhead Temp., oc (0 F) 60-150 15.5-65.5
Bottomhole Temp., oc (0 F) 85-320 29-160
Notes
1. Oil, gas and water viscosities are evaluated using empirical corre-
lations(11>.
2. Roughness of the pipe are taken = 0.00015 ft.
B. I
Cf Nlv P Cnl
.575 8.1

TABLE 3. Summary of data bank composition Ngv Pa Nd

Test Point Category No. of Points FIGURE 2. Comparison of revised and original liquid holdup cor-
relations.
Total number of points 540
High flow rate > 795 m3JD (5000 STB/D) 169
Low flow rate < = 795 m3fD (5000 STB/D) 371
Small ID tubing < = 0.062 m (2.441 inch) 298 0.0001 0.34 0.40
Large ID tubing > 0.062 m (2.441 inch) 242 0.0002 0.44 0.52
High GLR > 178 fn3Jm3 (1000 Scf/STB) 255 0.0005 0.65 0.71
Low GLR <= 178 m3Jm3 (1000 Scf/STB) 285 0.0001 0.82 0.84
0.0002 0.92 0.92
0.0003 0.96 0.96
The two-phase friction factor (f) was calculated on the basis 0.001 1.00 1.00
of a two-phase Reynolds number using the standard Moody dia-
gram. The two-phase Reynolds numberO> used was:
Calculation Procedure
QLM
NR~P = 2.2 X I0- 2 ................................. (3) A.main Fortran program was written for calculating the liquid
d /LrL JL~I- HL) holdup to satisfy the measured pressure losses for each data point,
with fluid property correlations handled as subroutines. These
Using this definition for the Reynolds number, and the con- subroutines were developed by Brill and Beggs<ll>. In this study,
ventional relationship between f and NRe for single-phase fluid, the liquid mixture properties, liquid surface tension, viscosity and
the liquid holdup is calculated from equations (1) and (2) for the density were handled ori a weighted average basis for the volume
540-test point data. These values of liquid holdup in terms of if; fraction of oil and water in the liquid phase. The following steps
were plotted vs the correlating function (Cf) suggested by are suggested for calculating the liquid holdup of each data point:
Hagedorn-Brown, as show in Figure 1. 1. Assume a value of liquid holdup.
2. Calculate the two-phase Reynolds number from equation (3).
Abscissa = Cf = __N_L_v_) (-P-)o.I 3. Using the value of NRetp and Eld, calculate f from Moody
NJ.575 14.7
gv diagram.
4. Calculate Pm from equation (1).
The comparison between the original and the revised liquid hold- 5. Calculate HL from equation (2).
up correlations is given in Figure 2. This figure reveals that, for 6. If the assumed value of HL and the value calculated from
any value of Cf, the revised correlation gives the higher value of equation (2) agree within 1o/o, consider the value of HL as
liquid holdup. This, of course, reduces the under-prediction of calculated from equation (2). If these do not agree, repeat
liquid holdup and, hence, of pressure drop. For computer appli- the procedure with the new value of HL until agreement
cation, the two correlations are numerically given below. within 1% is obtained.

Cf
Results and Discussions
0.000002 0.04 0.17
0.000005 0.09 0.12 The original and the revised Hagedorn-Brown correlations are
0.00001 0.15 0.17 tested aginst 157 well test data. These tests, were collected from
0.00002 0.18 0.22 15 flowing oil wells taken from a large undersaturated oil reser,.
0.00005 0.25 0.30 voir in Iraq. The wells are partially tubed and flowed on annulus.

26 The Journal of Canadian Petroleum Technology


The ranges of flow parameters are shown in Table 4. In order TABLE 4. Range of flow parameters of the present
to determine the effect of the two modifications introduced by data set
Brill and Hagedorn(2) on the performance of the original and the
revised correlation, the calculations are performed using the fol- Parameters Field Units Sl Units
lowing cases: Oil flow rate, m3JD (STBO/D) 3930-37400 625-5946
1. Original Hagedorn-Brown correlation(!). Tubing diameter, m (inch) 2.441 0.062
2. Original Hagedorn-Brown correlation(!) with the restriction Casing diameter (ID), in. 4. 778-7.825 0.1213-0.1987
on liquid holdup(2). Oil gravity, gm/cc (API) 32-34 0.862-0.855
Gas-Liquid ratio, m3Jm3 (Scf/STB) 630-870 112-155
3. Original Hagedorn-Brown correlation(!) with Griffith corre- Wellhead pressure, KPa (psig) 595-1815 4102-12514
lation(3) for bubble flow. Bottomhole pressure, KPa (psig) 4310-4980 29716-34335
4. Modified Hagedorn-Brown (two modifications introduced by Tank gas gravity (air= 1) 0.84-0.959 0'.84-0.959
Brill and Hagedorn(2). Well depth, m (ft) 10200-1 0500 3109-3200
5. Revised Hagedorn-Brown correlation. Wellhead Temp., oc (0 F) 126-170 52-77
6. Revised Hagedorn-Brown correlation with the restriction on Bottomhole Temp., oc (0 F) 203-210 95-99
liquid holdup(2).
7. Revised Hagedorn-Brown correlation with Griffith correla-
tion for bubble flow(3). TABLE 5. Summary of statistical results for the data
8. Revised Hagedorn-Brown correlation with the two modifi- used (157 points)
cations introduced by Brill and Hagedorn(2).
The statistical results for the above cases applied to all157 well Case APD AbAPD SD
tests are given in Table 5. Defiriitions of per cent difference, PD, 1 -19.96 30.76 31.10
arithmetic average of per cent differences, APD, absolute aver- 2 - 7.64 16.58 23.30
age of per cent differences, AbAPD, and standard deviation of 3 -11.40 23.60 25.84
4 - 7.21 17.46 24.13
per cent difference valued from the average per cent difference,
5 - 6.77 14.75 16.05
SD; are given below. 6 - 4.20 11.73 15.30
7 - 3.46 13.37 15.97
LWei- ~Pmi 8 - 2.60 12.52 15.73
PDi= X 100.
~Pmi

n
PDi TABLE 6. Statistical results for wells with slug flow
r;
i=1 at wellhead (129 points)
APD=
n Case APD AbAPD SD
5 - 6.35 13.63 15.26
n - 4.77 12.10 14.95
IPDilr; 6
1
AbAPD = _i=-- - -
n
TABLE 7. Statistical results for wells with bubble flow
at wellhead (28 points)
~ ~ (PDi - APD)2
SD = ~i=~l_ _ _ _ ___ Correlation APD ABAPD SD
n Hagedorn-Brown with revised holdup -9.2 19.8 21.0
Griffith 7.0 14.7 19.8
No-slip calculation 1.6 9.2 16.6
where:
LWei = Calculated pressure loss for the Ith point.
~Pmi = Measured pressure loss for the Ith point. well. For the measured data used in the present study (157 points),
it was felt that this assumption, while not perfectly accurate, would
An examination of the results given in Table 5 reveals that: reveal prevailing trends for the Hagedorn-Brown correlation. Table
1. The results were considerably better when using the revised 6 shows the performances of the revised correlation (cases 5 and
correlation (cases 5, 6, 7 and 8). 6) for the wells with slug flow at the tubing exit. Table 6 indicates
2. The original Hagedorn-Brown correlation (case 1) yielded the that, the use of the restriction on liquid holdup (case 6) has a lit-
poorest results and showed a tendency toward sigriificant under- tle effect on the performance of the revised correlation. This rev-
prediction of pressure drop. ealed that, for most data in slug flow, the no-slip liquid holdup
3. On the basis of the lowest absolute average per cent error and was lower than that calculated from the revised holdup correlation.
the standard deviation, the revised correlation with the restriction Table 7 shows the statistical results for the wells with bubble
on liquid holdup (case 6), showed the best performance. flow at the tubing exit (28 points) for:
4. Results obtained in using the restriction on liquid holdup (cases 1. Hagedorn-Brown correlation with the revised holdup correla-
2 and 6) were found to be better than adopting Griffith correla- tion (case 5).
tion for bubble flow (cases 3 and 7). The reason was that, the 2. Griffith(3) correlation .
.majority of our data (129 points) was in slug flow and few (28 3. No-slip calculation.
points) were in bubble flow.
5. Results obtained in using the restriction on liquid holup (cases Table 7 indicates that the revised liquid holdup correlation
2 and 6) were found to be slightly better than using the two modifi- tended to underpredict liquid holdup, while Griffith correlation
cations (cases 4 and 8). This was due to the overprediction(9) of showed a tendency to overpredict liquid holdup. The no-slip cal-
the liquid holdup calculated by the Griffith correlation for bubble culation showed the best performance.
flow.
In addition to the results of Table 5, the performance of
Hagedorn-Brown correlation was compared for wells with the same
Conclusions and Recommendations
predicted flow regime at the tubing exit. The flow regimes were 1. The Hagedorn-Brown liquid holdup correlation was revised to
defined using the criteria proposed by Orkiszewski(l2). It was satisfy 51 pressure profiles consisting of 540 field measured pres-
assumed that the flow pattern at the tubing exit was representa- sure losses, collected from 51 vertical oil wells from different geo-
tive of the predominant two-phase flow pattern throughout the graphical areas.

November-December 1989, Volume 28, No. 6 27


2. The original Hagedorn-Brown correlation underpredicted pres- Sl METRIC CONVERSION FACTORS
sure loss by nearly 200Jo. This was primarily because of an under- oAPI 141.5/131.5 + oAPI) = g/cm3
prediction of liquid holdup. bbl X 1.589873 E-01 = m3
3. The revised Hagedorn-Brown correlation (case 5) gave results cu. ft X 2.831685 E-02 = m+
much better than the original with the two modifications intro- oF eF-32)!1.8 = oc
duced by Brill-Hagedom (case 4). in. x 2.54* E+OO = em
4. The no-slip calculations accurately predicted pressure drop in psi X 6.894757 E + 00 = kPa
bubble flow.
5. Pressure drop calculations by Hagedorn-Brown correlation can *Conversion factor is exact.
be improved by using the revised holdup correlation when slug
flow is indicated and by using no-slip holdup calculations at bubble REFERENCES
flow.
1. HAGEDORN, A.R., and BROWN, K.E., Experimental Study of
6. A comparison study such as this one depends a great deal on
Pressure Gradients Occurring During Continuous Two-Phase Flow
the quality and range of basic well data. Different well data may in Small Diameter Vertical Conduits; Journal of Petroleum Tech-
result in different conclusions, therefore, additional field verifi- nology, p. 475, Apr. 1965.
cation was needed to check the validity of the above conclusions. 2. BRILL, J.P., and HAGEDORN, A.R., In The Teclrnology of Arti-
fical Lift Methods; K.E. Brown, ed., Vol. 1, 114 p., 1977. ,
NOMENCLATURE 3. GRIFFITH, P., In The Technology of Artificial Lift Methods; K.E.
A = pipe cross section area, sq. ft Brown, ed., Vol. 1, 114 p., 1977.
CNL = coefficient for liquid viscosity number 4. MESSULAM, S.A.G., Comparison of Correlations for Predicting
d = .internal diameter of tubing or casing, ft Multiphase Flowing Pressure Losses in Vertical Pipes; M.S. thesis,
f = Moody friction factor of Fanning friction factor Univ. of Tulsa, 1970.
gc = gravitational constant, lbm/sec.lbf. 5. LAWSON, J.D., and 'BRILL, J.P., A Statistical Evaluation of
HL = liquid holdup fraction Methods used to Predict Pressure Losses for Multiphase Flow inVer-
M = total mass of oil, water and gas, associated with 1 bbl of tical Oil Well Tubing; Journal of Petroleum Technology, pp. 903-913,
liquid flowing into and out of the flow string, lbrnlbbl Aug. 1974.
Nd = pipe diameter number, d(pL g/a)0.5 6. ROBINSON, J.R., Development of a Two-Phase Flow Well Data
Ngv = gas velocity number, Vsg (pdga)0.25 Bank and an Evaluation Study of Pressure Loss Methods Applied
NLv = liquid velocity number, V5i(pLJga)0.25 to Directional Oil Wells; M.S. thesis, _Univ. of Tulsa, 1979.
NL = liquid viscosity number, JLL(g/pL~)0.25 7. IBE, M.C., Determination of the Best Combination of Pressure Loss
NRt;p = Renolds number, two-phase flow and PVT Property Correlations for use in Upward Two-Phase Flow;
M.S. thesis, Univ. of Tulsa, 1979.
p = pressure, psia
Pa = atmospheric pressure, psia . 8. ROSSLAND, L., Investigation of the Performance of Pressure Loss
= liquid flow rate, STBL/Day Correlations for High Capacity Wells; M.S. thesis, Univ. of Tulsa,
QL
1981.
QLM
= average velocity of the mixture, ft/sec., _ __ 9. THOMPSON, L.G., Determination of Optimum Combination of
ApM Pressure Losses and PVT Property Correlations for Predicting Pres-
sure Gradients in Upward Two-Phase Flow; M.S. thesis, Univ. of
Greek Symbols Tulsa, 1982. ,
E = roughness of the pipe, ft 10. GHASSAN H., ABDUL-MAJEED, and ZEIAD, A.R. ASWAD,
..:ip/dh = total pressure drop over a given elevation change, psi/ft Comparison of Correlations for Predicting Multiphase Flowing Pres-
Jl.g = viscosity of gas, cp sure Losses in Vertical and Inclined Pipes; Al-Muhandis, Journal of
ILL = viscosity of liquid, cp the Iraqi Society of Engineres, Serial 91, 1987.
Pg = density of gas, lb/ cu. ft 11. BRILL, J.P., and BEGGS, H.D., Two-Phase Flow in Pipes; Two-
PL = density of liquid, lb/cu. .ft Phase Flow Text, Univ. of Tulsa, 1984.
Pm = mixture density, lb/cu. ft 12. ORKISZEWSKI, J., Predicting Two-phase Pressure Drops inVer-
I/; = secondary correction factor tical Pipes; Journal of Petroleum Technology, p. 829, June 1967.

28 The Journal of Canadian Petroleum Technology

You might also like