Professional Documents
Culture Documents
I
J. Eng. Sci., King Saud Univ., Vol. 14 (2), pp. 423-435 (1408/1988)
! i I ! i
t ! !
The correlation of Lockhart and Martinelli in its present form cannot be used to
study a large set of data because it requires the use of charts and hence cannot be simulated numerically. A correlation between Lockhart and Martinelli parameters </>and X lor a two phase pressure drop in pipelines was developed using the Statistical Analysis System (SAS). This enabled the development of a computer program for the analysis of data using the Lockhart and Martinelli correlation. Field data from Saudi flowlines were then analysed using the program and the results show that the impoved Lockhart and Martinelli correlation predicts accurately the downstream pressure in flowlines with an average percent difference of 5.1 and standard deviation of 9.6%.
Nomenclature
Ap
d e f L
NRe P Ppr PI Pz q Qg R
Cross sectional area of pipe, sq. ft Pipe diameter, in The absolute roughness, in Friction factor Pipe length, ft Reynold's number Pressure, psi The reduced pressure Upstream pressure, psi Downstream pressure, psi Flow rate, cu ft/sec Gas flow rate, scf/hr. The multiple correlation coefficient, dimensionless
423
~.Y'-" I ~
w~
L~i MI ,~..dl
i}J\ ~
424 t T
VsL Vsg X y
Z Y
t1 p J1 cp
The reciprocal pseudo reduced temperature Average temperature, OR Superficial liquid velocity, cu ftjsec. Superficial gas velocity, cu ftjsec. Lockhart and Martinelli parameter The reduced density Compressibility factor Specific gravity Difference Density, Lbmjcu ft Viscosity, cp Lockhart and Martinelli parameter
Subscripts
g L 0 TP
gas phase liquid phase oil two-phase viscous flow mechanism turbulent flow mechanism
Introduction Many of the two-phase flow correlations [1-4J are in use, but the Lockhart and Martinelli correlation [4J, because of its simplicity, has been widely used in industry to predict two-phase flow in horizontal pipes especially for low gas and liquid flow rates and small pipe sizes. The pressure drop resulting from two-phase flow, is predicted by calculating the single-phase pressure drop for each phase as ifeach of them was flowing alone through the pipes; and correcting these values for two-phase flow by Lockhart
425
ASSUME LIQUID-PHASE FLOW ONLY, CALCULATE qL' vsL' PL' (NRe)L' fL' t.PL
ASSUMe GAS-PHASE FLOW ONLY, CALCULATE qg' Vsg' Pg' (NRe)g' ~Pg
I. I I I : t , i j 1 I I ~
~P 9
CALCULATE
(APL).5,~PTP(9)'~TP(L)
Lockhart
and
Martinelli
correlation.
~f--" .:..ill I
~ ,
d\
~.
,~.dl
(}.oJ ~ I
426
Add Hemeida and Faisal Sumait and temperatures by the Lagrangian interpolation method [5J, while the other is to calculate the compressibility factor of gas using the Hall-Yarborough equation [6, 7J, which were developed by using the Starling-Carnahan equation of state: z= 0.06125 Ppr t e -1.2(1 y
-c)2
(l)
The value of reduced density (y) can be obtained from the solution of the following equation using the Newton-Raphson iterative technique:
2 3 4
t2+42.4 t3)y<2.18+2.821)=0
Fig. (2)shows the subroutine for calculating the gas compressibility factor. By assuming the gas phase flowing alone, downstream pressure is calculated by Weymouth's equation [7]:
O.5
(3)
Therefore, the pressure drop is: /1P = upstream pressure (P 1)- downstream pressure (P 2) (4)
= 43.5fqlLPL
(5)
wherefis the friction factor determined from the equation for Moody friction factor [7J in terms of Reynolds number and relative roughness (Table I). Reynold's number for each phase is obtained as follows:
-(NRe)L = 123.9~
where
VsL=
dv P
flL
(6)
qL
Ap
(7)
and
where
(N Re)g = 123.9dvsgPg
flg
q v =---L sg Ap
The type oflow mechanism is determined from Table (2).The X parameter is calculated
Journal of Eng. Sci.. Vol. 14, No.2 (1988). College of Eng., King Saud Univ.
427
CALCULATE A, S, C,
YES
YES
(
Fig. (2).
RETURN
)
factor.
dl
LI)I..\.60\ ,~..dl
pWI ~
i ~ " " ~ 11 e. i t , t i i
for calculating
! ~
NR<
critical transition
( ;-)
200~
1.16
9.34
,fl = 1.14-2
1
log [ d + NRe,flJ
turbulent
[NRe>C~dr16J
,fl = 1.14-
210g~
Flow mechanism
turbulent, gas turbulent viscous, gas viscous turbulent, gas viscous viscous, gas viscous
x=
( )
APg
APL
O'5
(8)
Figure (3) is used to obtain the Lockhart and Martinelli parameter </>. The curves for </>L are used when the pressure gradient for the liquid is used and vice versa. The two-phase pressure drop is then calculated as:
.
() ()
AL
-</>2
TP
AP
AP
AL
(9)
</> using SAS software:
LorG
The following
equation
was derived
to calculate
the parameter
</>
= exp
(10)
where a, b, and c are constants and they have been selected according to the type of fluid and flow mechanisms (Table 3). Table (4) shows a sample of the computer outputs. The validity of the program was tested. The results obtained, showed a percentage error of 1.8 in pressure drop. The square of the multiple correlation coefficients for </>'sre close to unity and hence the a equation derived can replace the charts with high accuracy. Journal of Eng. Sci., Vol. 14, No.2 (1988). College of Eng.. King Saud Univ.
429
1000.0
i
~
"
'"
5 100.0
, "
"; ,
I
<t>lvl ;
i
i '
I
II I
I'
ii'
i Iii
q,gtt
100.0
~
~
1
.. : :-'
& E E /1.
~I;; ~
--"
I 'o "'90'
,--<t>lIv
! " I
I
'
'
<t>lvv
"
I
i "
I
,
10.0
''''..
"
'" gll
I
i
.."
I
~ ..
='-
I "
I !
I
I ;
j,
! I'
i
; I
LUfi1
,--<t>lit <t> Ivl <t> v II <1>' vv Ilil
, 10.0
,
I I
I<"
i
1.0
I
11111 ;
I
1.0 100.0
1.0 0.01
I
0.1 Parameter X
10.0
=J(:~Y(~:)G =ff!i;
and Martinelli).
Fig. (3).
Correlation
Table (3). Values of a, b, and c of equation (10) for different flow mechanisms Parameter q, gvv gvt gtv gtt Lvv Lvt Ltv Ltt a 0.4625 0.5673 0.5694 0.6354 0.4048 0.5532 0.5665 0.6162 b 0.5058 0.4874 0.4982 0.4810 0.4269 -0.4754 -0.4586 -0.5063 c 0.1551 0.1312 0.1255 0.1135 0.1841 0.1481 0.1413 0.124 R2 0.996576 0.999395 0.999324 0.999928 0.984280 0.999905 0.999718 0.992167
Field Data
Tests were done on 101Saudi flowlines of 4,6 and 8 inches diameter [9]. The pipe length varies from approximately 2000 to 35000 ft and flow rate varies from approximately 400 to 18000STBjD. Most of the PVT data were taken from the experimental PVT analysis, such as formation volume factor, gas solubility, gas-liquid ratio and average specific gravity of the gas. API gravity and viscosity of oil were taken from actual field data. Gas viscosity was taken from reference [10] as a function of gas
:'y<--'!.illl4 6:- , . xJ.1~
. (.J>
,t ./\) Jl:!1 : wl
'?
~ w 0
'c ;::
~ ~.,
~
C/J
.., ;:::
t"r1 ;::: Table (4). Dependent variable: Y DF 2 8 10 OF Sum of squares 4.84029187 0.01662831 Mean square 2.42014593 0.00207854 F value 1164.35 PR>F 0.0001 ROOT
0.04559100
~.
."....
Computer
rPgn
~
""
Source
Model Error Corrected Source total
R-square 0.996576
i::>..
"" ""
;:t...
<: ~
N
........
::J::
MSE
::!
4.85692018 -Type I SS 4.26274560 0.57754627 F Value 2050.84 277.86 PR>F 0.0001 0.0001 DF 1 1
I::>
\Q Co Co '--
(]
XI X2
. ~ ~ ""
.-----------------
(;;. C/J
;::
Tror HO:
Parameter Intercept (a) X (b) X2 (c) Estimate 0.46251578 0.50578379 0.15507377 Parameter 22.49 46.86 16.67
=0
~
~
t"r1 ;:::
.--
::!
'"
~
;:'
-;:::
,....
;:;.
Two-Phase
431
gra vity. Its average value was estimated as 0.0127 cp after correction for the existance of N2, CO2, and H2S gases. Results and Discussion The computer output using calculated gas compressibility factor, gas solubility and oil formation factor at operating pressures and temperatures shows a high degree of accuracy in results with an average percentage difference of 1, 2 and 0.5%
respectively.
Table (5) shows samples of test results obtained using improved Lockhart and Martinelli correlation. The statistical results yielded an average percentage error of 5.1 %, standard deviation 9.6% and average absolute error 8.4% for downstream pressures.
Table (5). Lockhart and Martinelli correlation - test analysis Test 6 10 11 12 13 14 15 16 17 18 19 20 21 23 25 26 27 28 29 30 38 39 40 41 42 44 45 51 52 66 67 75 76 86 Diam. 4.0260 4.0260 4.0260 4.0260 4.0260 4.0260 4.0260 7.8264 7.8264 7.8264 7.8264 7.8264 7.8264 7.8264 7.8264 4.0260 4.0260 4.0260 4.0260 4.0260 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 6.0000 Flow rate 941 8647 4513 4370 4022 3892 2216 15440 15522 12212 7184 2408 16882 7793 3667 12884 7804 4176 1618 984 3533 2540 6750 4570 3590 11508 8086 3837 12562 3812 3112 11008 3812 14319 P2(Measd.) 276 231 208 204 201 203 195 480 480 462 428 421 380 288 269 326 288 273 271 264 183 186 195 193 186 435 436 264 294 307 304 302 267 422 P2(Calc.) 278.4262 195.1303 178.8734 181.5468 176.8641 178.4189 185.4497 433.1257 438.1752 431.6255 435.0644 449.3134 321.3018 269.0238 284.9198 261.9651 243.6818 254.3103 267.7656 262.6232 174.5437 186.6878 172.7394 178.1591 173.5823 378.5863 386.5228 252.3550 262.8994 296.6782 301.6867 265.5043 260.5042 366.5192 %Err press. drop 35.37472 -20.15150 -49.36710 - 36.80856 -48.27182 - 53.43724 - 53.05745 - 51.51025 -43.56753 -46.02198 14.71748 83.27458 -45.85797 -42.16941 48.24184 - 23.20105 - 32.34903 - 36.64644 - 32.34446 - 34.42096 - 35.23472 3.619967 -32.261820 -43.649860 -59.131900 -29.691430 -48.987350 - 23.290090 -10.798830 -86.014780 -17.795080 - 38.825210 -38.210560 - 77.056630 % Err P2 -0.8971837
15.5280000 14.0031700 11.0064800 12.0079200 12.1089300
6.5889680 -5.9181410
19.6426000 15.3882600
7.6896070 6.6761790
12.9686700 11.3479900
4.4110040
10.5784400
3.3621420 0.7609719
12.0846700
2.4328800
13.1471000
~ ,
,J.:J,\ ~
~~1 ~I
432
Adel Hemeida
The effects of some important variables such as pipe size, liquid flow rate and liquid Reynolds number, on percentage errors in downstream pressure calculations, were studied. The results are presented in Figs. (4 and 5). The minimum percentage error in downstream pressure prediction at different liquid flow rates for 4 in, 6 in and 8 in pipes, are presented in Fig. (4). As liquid flow rate increases the percentage error increases. The rate of increase in percentage error is higher for smaller diameter pipes because of the increase in superficial liquid velocities. It can be seen that at low liquid flow rates the absolute value of percentage error decreases to zero and then starts to rise. It should be noted that the variation in percentage error gradually fades. Since a small range in pipe diameters was considered in the Lockhart and Martinelli correlation (0.0586to 1.017 in) it was safe to assume that the accuracy in pressure drop calculations was a function of flow rate irrespective of the prevailing flow pattern. For large pipe diameters, the flow pattern changes for a constant flow rate which leads to more erroneous results when using the above Lockhart and Martinelli correlation. In other words the ratio of the fraction of the space occupied by the liquid decreases with the increase in pipe diameter. When going to larger diameters, radial static pressure difference may exist and may produce differences between static pressure drops of the liquid and gas phases which also lead to erroneous results. Since Reynolds number is used to calculate cp,according to the flow mechanism, Reynolds number for the liquid phase was calculated and plotted against the
30
.
0 X
4 in 6 in 8 in
20
cr: a cr: cr: UJ 10
IZ UJ U cr: UJ
a..
_10
LI QUID
F LOW RATE,
1000 BB L I DAY
_20
Fig. (4). Effect of liquid flow rate on percentage diameters. errors in downstream pressure prediction at various pipe
Improving
433
30
20
a: 0 a: a: UJ
..... 00
10
0 0
UJ u a: UJ a..
0 . . _10
REYNOLDS NUMBER FOR LIQUID PHASE
_20
Fig. (5). Effect of Reynolds number on percentage errors in downstream pressure prediction.
percentage errors in downstream pressure prediction. Fig. (5)shows that the increase in Reynolds number causes an increase in percentage errors. This reflects the effect of liquid superficial velocities on the pressure drop calculations. As the liquid superficial velocity increases the percentage error increases.
Conclusions
The Lockhart and Martinelli correlation was improved by allowing predictions of <p be made without the use of charts. The program developed accordingly was used to to study a large set of field data and the following conclusions were drawn: (1) The equations for parameter <p,developed by the SAS method, are in good agreement with Lockhart and Martinelli experimental graphs. Therefore, the program thus developed proved its accuracy and capability of analysing large sets of data. (2) As liquid flow rate increases, the percentage error in downstream pressure prediction increases. The range of increase is higher for smaller diameter pipes.
References
[1]
[2]
Beggs, H.D. and Brill, J.P. (1973) A study of two-phase flow in inclined pipes. J PT(May), 607.
Dukler, A.E., Wicks, M. and Cleveland, R.G. (1964) Frictional pressure drop in two-phase flow, (A)
~.Y'--".!.ill I "'-'lo:-
,"
434
[3]
Eaton, B.A.; Andrews, D.E.; Knowles, C.R.; Silberberg, I.H. and Brown, K.E. (1967) The prediction of flow patterns. liquid holdup and pressure losses occuring during continuous two-phase flow in horizontal pipelines, J.PT (June): 815-828; Trans. AIME, 24 (June 1967). Lockhart, R.W. and Martinelli, R.C. (1949) Proposed correlation of data for isothermal two-phase,
two-component flow in pipes, Chemical Engr. Progress, 45, N-l:39--48 Gerald, Curtis F. (1978) Applied Numerical Analysis, Reading,
Wesley Publishing Company. of state for Z-factor, SPE Reprint No. 13, 2"d Ed. Massachusetts: Addison-
Yarborough, L. and Hall, K.R. (1977) How to solve equation I, p. 233. Texas.
[7]
Ikoku, Chi u. (1980) Natural Gas Engineering, A System Approach. Tulsa, Oklahoma: Pennwell
Publishing Company. Book Company.
Katz, D.L. (1959) Handbook of Natural Gas Engineering, New York: McGraw-Hili Caponegro,
Microfilm
flow data on wells and flowlines, SPE Paper No. 1655. G, Univ.
Brown, K.E. (1977) The Technology of Artificial Lift Methods. 1, Tulsa: Petroleum Publishing Company.
(Manuscript Received: 5.7.1987; Accepted: 4.11.1987)